content
stringlengths 1
15.9M
|
|---|
\section{References}\vspace*{-10pt}\list
{[\arabic{enumi}]}{\settowidth\labelwidth{[#1]}\leftmargin\labelwidth
\advance\leftmargin\labelsep
\usecounter{enumi}}
\def\hskip .11em plus .33em minus .07em{\hskip .11em plus .33em minus .07em}
\sloppy\clubpenalty4000\widowpenalty4000
\sfcode`\.=1000\relax}
\let\endthebibliography=\endlist
\def\@nomath#1{\ifmmode \fi}
\def\@ifnextchar [{\@tempswatrue\@mmycitex{\@ifnextchar [{\@tempswatrue\@mmycitex}
{\@tempswafalse\@mmycitex[]}}
\def\@mmycitex[#1]#2{\if@filesw\immediate%
\write\@auxout{\string\citation{#2}}\fi
\def\@citea{}\@mmycite{\@for\@citeb:=#2\do
{\@citea\def\@citea{,}\@ifundefined
{b@\@citeb}{{\bf ?}\@warning
{Citation `\@citeb' on page \thepage \space undefined}}%
\hbox{\csname b@\@citeb\endcsname}}}{#1}}
\def\@mmycite#1#2{{{\scriptsize#1}\if@tempswa , #2\fi}}
\def\mycite#1{$^{\protect\@ifnextchar [{\@tempswatrue\@mmycitex{#1}}$}
\def\@cite#1#2{{#1\if@tempswa , #2\fi}}
\def\arabic{section} {\arabic{section}}
\def\section#1{\addtocounter{section}{1}\setcounter{subsection}{0}
\bigskip\medskip{\noindent\bf\arabic{section}.\ #1}
\medskip}
\def\arabic{section}.\arabic{subsection} {\arabic{section}.\arabic{subsection}}
\def\subsection#1{\addtocounter{subsection}{1}
\medskip{\noindent\arabic{section}.\arabic{subsection}.\ #1}
\medskip}
\makeatother
\topmargin=0.25in\oddsidemargin=0.25in\evensidemargin=0.25in
\textheight=8.50in\textwidth=6.00in\headheight=0.00in
\headsep=0.00in\thispagestyle{empty}
\newcommand{\bold}[1]{\mbox{\boldmath $#1$}}
\defbegin{eqnarray}{begin{eqnarray}}
\defend{eqnarray}{end{eqnarray}}
\begin{document}
\vspace*{0.3in}
\begin{center}
{\bf NEUTRINOS FROM PROTONEUTRON STARS: \\
A PROBE OF HOT AND DENSE MATTER } \\
\bigskip
\bigskip
SANJAY REDDY AND MADAPPA PRAKASH \\
{\em Physics Department \\
State University of New York at Stony Brook \\
Stony Brook, NY 11794-3800, USA }
\bigskip
\end{center}
\smallskip
{\footnotesize
\centerline{ABSTRACT}
\begin{quotation}
\vspace{-0.10in}
Neutrino processes in dense matter play a key role in the dynamics,
deleptonization and the early cooling of hot protoneutron stars formed in the
gravitational collapse of massive stars. Here we calculate neutrino mean free
paths from neutrino-hyperon interactions in dense matter containing hyperons.
Significant contributions to the neutrino opacity arise from scattering
involving the $\Sigma^-$ hyperon, and absorption processes involving the
neutral and $\Sigma^-$ hyperons. The estimates given here emphasize
the need for (a) opacities which incorporate many-body
effects in a multi-component mixture, and (b) new calculations of thermal and
leptonic evolution of protoneutron stars with neutrino transport and equations
of state with strangeness-rich matter.
\end{quotation}}
\vskip 5pt
\section{Introduction}
The general nature of the neutrino signature expected from a newly formed
neutron star (hereafter referred to as a protoneutron star) has been
theoretically predicted\mycite{BL} and confirmed by the
observations\mycite{SN87A} from supernova SN1987A. Although neutrinos
interact weakly with matter, the high baryon densities and neutrino energies
achieved after the gravitational collapse of a massive star ($\geq 8$ solar
masses) cause the neutrinos to become trapped on the dynamical timescales of
collapse\mycite{sato,maz}. Trapped neutrinos at the star's core have
Fermi energies $E_\nu \sim 200-300$ MeV and are primarily of the $\nu_e$
type. They escape after diffusing through the star exchanging energy with the
ambient matter, which has an entropy per baryon of order unity in units of
Boltzmann's constant. Eventually they emerge from the star with an average
energy $\sim 10-20$ MeV and in nearly equal abundances of all three flavors,
both particle and anti-particle.
Although the composition and the equation of state of the hot protoneutron star
matter are not yet known with certainty, QCD based effective Lagrangians have
opened up intriguing possibilities. Among these is the possible existence of
matter with strangeness to baryon ratio of order unity. Strangeness may be
precipitated either in the form of fermions, notably the $\Lambda$ and
$\Sigma^-$ hyperons, or, in the form of a Bose condensate, such as a
$K^-$-meson condensate (see Ref.~[\cite{prak}] for detailed discussion and
extensive references). In the absence of trapped neutrinos, strange
particles are expected to appear around $2-4$ times the nuclear matter density
of $n_0=0.16~{\rm fm}^{-3}$. Neutrino-trapping causes the strange particles
to appear at somewhat higher densities than in neutrino-free matter. The
compositions shown in Figs. 1 and 2 highlight the influence of hyperons in the
neutrino trapped regime. The results shown in these figures were
calculated\mycite{prak} using a field-theoretical model in which baryons
interact via the exchange of $\sigma , ~\omega$ and $\rho$ mesons. With the
appearance of hyperons in matter, the electron-neutrino fraction increases with
density in contrast to the case in which matter contains nucleons and leptons
only. A similar behavior is observed in kaon condensed matter\mycite{tpl} and
also in matter where a phase transition to quark matter occurs\mycite{pcl}.
This behavior is associated with the presence of non-leptonic negatively
charged particles in matter\mycite{prak}, such as the $\Sigma^-$ hyperon, or
$K^-$meson, or $d$ and $s$ quarks.
Keil and Janka\mycite{KJ} have recently investigated the influence of the
equation of state on the cooling and evolution of the protoneutron star. They
find that the neutrino luminosity depends sensitively on the composition and
on the stiffness of the equation of state at high density. In particular, the
influence of hyperons, which introduces a softening of the high density
equation of state, was examined. In many cases, the protoneutron star
collapsed
to a black hole causing an abrupt cessation of the neutrino signal. Although
a clear distinction between the different equations of state could be achieved
on the basis of the calculated neutrino signals, Keil and Janka conclude that
``none of the models could be considered as a good fit of the neutron star
formed in SN 1987A''.
We note, however, that in these studies, the sensitivity of
the neutrino signals due only to the structural changes caused by the
equation of state was assessed. In the presence of hyperons, the transport of
neutrinos is also affected due to the changes in the composition,
and additionally, to the interactions of neutrinos with strange particles.
These effects were ignored in Ref.~[\cite{KJ}]. Previous
work\mycite{max,pplp} involving neutrino interactions with hyperons was
concerned with charged-current reactions only. Studies of the opacities and
the transport processes of neutrinos out of the core containing nucleons,
leptons, and in some cases, pion condensates may be found in
Refs.~[\cite{Saw1}] through [\cite{Bur}].
It is our purpose here to study neutrino mean free paths in matter containing
hyperons. Specifically, we will estimate scattering and absorption mean free
paths of neutrinos when matter is under degenerate conditions. We find that
significant contributions to the neutrino opacity arise from scattering
involving the $\Sigma^-$ hyperon, and absorption processes involving the
neutral and $\Sigma^-$ hyperons.
Compared with ordinary matter, the presence of strangeness in
matter is expected to lead to an excess of $\nu_e$ neutrinos relative to other
types. Calculations of the complete thermal and leptonic evolution of a newly
formed neutron star incorporating neutrino transport and equations of state
with strangeness-rich matter will be taken up separately. With new generation
neutrino detectors capable of recording thousands of neutrino events, it may
be possible to distinguish between different scenarios observationally.
\begin{figure}
\vspace*{7.5in}
\caption[]{\footnotesize
The composition, electron chemical potential and entropy per particle in
nucleons only matter with a lepton fraction $Y_{Le} = Y_e + Y_{\nu_e}= 0.4$,
where $Y_i=n_i/n_b$. Results are from Ref.~[\cite{prak}]. }
\end{figure}
\begin{figure}
\vspace*{7.5in}
\caption[]{\footnotesize
The composition, lepton chemical potentials and entropy per particle
in strangeness-rich matter with a lepton fraction
$Y_{Le} = Y_e + Y_{\nu_e}= 0.4$, where $Y_i=n_i/n_b$.
Results are from Ref.~[\cite{prak}]. }
\end{figure}
\vskip 5pt
\section{Neutrino interactions with strange baryons }
Neutrino interactions with matter proceed via charged and neutral
current reactions. The neutral current processes contribute to
elastic scattering, and the charged current reactions result in neutrino
absorption. The interaction Lagrangian for these reactions
is given by the Wienberg-Salam theory:
\begin{eqnarray}
{\cal L}_{int}^{nc} &=& ({G_F}/{2\sqrt{2}}) ~~l_\mu
j_z^\mu\, \qquad {\rm ~for} \qquad \nu + B \rightarrow \nu + B \nonumber \\
{\cal L}_{int}^{cc} &=& ({G_F}/{\sqrt{2}}) ~~l_\mu^e j_w^\mu\,
\qquad {\rm for} \qquad \nu + B_1 \rightarrow \ell^- + B_2 \,,
\end{eqnarray}
where $G_F\simeq 1.436\times 10^{-49}~{\rm erg~cm}^{-3}$ is the weak coupling
constant, $\nu$ is a neutrino, $B_1$ and $B_2$ are baryons, and $\ell^-$ is a
lepton. The leptonic and hadronic currents appearing above are:
\begin{eqnarray}
l_\mu &=& {\overline \psi}_\nu \gamma_\mu
\left( 1 - \gamma_5 \right) \psi_\nu \nonumber\\
l_\mu^e &=& {\overline \psi}_e \gamma_\mu
\left( 1 - \gamma_5 \right) \psi_\nu \nonumber\\
j_w^\mu &=& {\overline \psi}_i \gamma^\mu
\left( g_{Vi} - g_{Ai} \gamma_5 \right) \psi_i \nonumber\\
j_z^\mu &=& {\overline \psi}_i \gamma^\mu
\left( C_{Vi} - C_{Ai} \gamma_5 \right) \psi_i \,,
\end{eqnarray}
where $i=n,p,\Lambda,\cdots$.
The neutral current process couples neutrinos of all types ($e,\mu$ and $\tau$)
to the weak neutral hadronic current $j_z^\mu$. The charged current processes
of interest here are electron and muon neutrinos coupled to the charged
hadronic current $j_w^\mu$. The vector and axial vector coupling constants
are listed in Table 1. Numerical values of the parameters that best fit the
experiments are: D=0.756 , F=0.477, $\sin^2\theta_W$=0.23 and $\sin\theta_c =
0.231$.
\vskip 0.2in
{\centerline {TABLE 1}
\vskip 0.1in
{\centerline {NEUTRAL CURRENT VECTOR AND AXIAL COUPLINGS } }
\vskip 0.1in
\begin{center}
\begin{tabular}{c|cccc}
\hline \hline
{Reaction } & $C_V$ & $C_A $ & ${\cal R}_{nc}^{(1)}$ & ${\cal R}_{nc}^{(2)}$
\\ \hline
$\nu_i + n \rightarrow \nu_i + n$ & $-1$ & $-D-F$ & 1 & 1 \\
$\nu_i + p \rightarrow \nu_i+ p$ & $ (1-4\sin^2\theta_W)$ & $D+F$ & 0.7504 &
0.8597 \\
$\nu_i + \Lambda\rightarrow \nu_i +\Lambda$ & 0& 0 & 0 & 0 \\
$\nu_i + \Sigma^-\rightarrow \nu_i +\Sigma^-$ &$ (-2+4\sin^2\theta_W)$ & $-2F$
& 0.7392 & 0.6788 \\
$\nu_i + \Sigma^+\rightarrow \nu_i +\Sigma^+$&$ (2-4\sin^2\theta_W)$ & 2F &
0.7392 & 0.6788 \\
$\nu_i + \Sigma^0\rightarrow \nu_i +\Sigma^0$&$ 0$ & 0 & 0 & 0 \\
$\nu_i + \Xi^-\rightarrow \nu_i +\Xi^-$ & $ (-1+4\sin^2\theta_W)$ & D
& 0.2845 & 0.3238 \\
$\nu_i + \Xi^0\rightarrow \nu_i +\Xi^0$ & 1 & $-D+F$ & 0.2861 & 0.1852
\\ \hline \hline
\end{tabular}
\end{center}
\begin{quote}
NOTE.-- The quantity ${\cal R}_{nc}^{(1)} = [C_V^2+2C_A^2]_{ \nu + B
\rightarrow
\nu + B } / [C_V^2+2C_A^2]_{ \nu + n \rightarrow \nu + n }$ and
${\cal R}_{nc}^{(2)} = [C_V^2+4C_A^2]_{ \nu + B \rightarrow
\nu + B } / [C_V^2+4C_A^2]_{ \nu + n \rightarrow \nu + n }$.
\end{quote}
\vskip 5pt
\newpage
\vskip 5pt
{\centerline {TABLE 2}
\vskip 0.1in
{\centerline {CHARGED CURRENT VECTOR AND AXIAL COUPLINGS }}
\vskip 0.1in
\begin{center}
\begin{tabular}{c|ccc}
\hline \hline
{Reaction } & $g_V$ & $g_A$ & ${\cal R}_{cc}$ \\ \hline
$\nu_i + n\rightarrow e + p$ &$ 1$ & $F+D$ & 1 \\
$\nu_i + \Lambda \rightarrow e + p$ &$-\sqrt{3/2}$ &$-\sqrt{3/2}(F+D/3)$ &
0.0394 \\
$\nu_i + \Sigma^-\rightarrow e + n$ &$-1$ &$ -(F-D)$ & 0.0125 \\
$\nu_i + \Sigma^-\rightarrow e + \Lambda $ & $ 0$ & $\sqrt{2/3}$ & 0.2055 \\
$\nu_i + \Sigma^-\rightarrow e + \Sigma^0$ &$ \sqrt{2}$ & $\sqrt{2}F$ & 0.6052
\\
$\nu_i + \Xi^-\rightarrow e + \Lambda$ &$ \sqrt{3/2}$ & $\sqrt{3/2}(F-D/3)$ &
0.0175 \\
$\nu_i + \Xi^-\rightarrow e + \Sigma^0$ &$ \sqrt{1/2}$ & $(F+D)/\sqrt{2}$ &
0.0282 \\
$\nu_i + \Xi^-\rightarrow e + \Xi^0$ &$ 1$ & $F+D$ & 0.0564 \\
$\nu_i + \Xi^0\rightarrow e + \Sigma^+$ &$ 1$ & $F-D$ & 0.2218 \\
\hline\hline
\end{tabular}
\end{center}
\begin{quote}
NOTE.-- The quantity ${\cal R}_{cc} =
[C^2(g_V^2+3g_A^2)]_{ \nu + B_1 \rightarrow
\ell + B_2 } / [C^2(g_V^2+3g_A^2)]_{ \nu + n \rightarrow \ell + p } $.
\end{quote}
\vskip 5pt
In what follows, we consider the lowest order (tree level) processes for
both elastic and absorption reactions.
The squared matrix element for neutral current reactions is given by
\begin{eqnarray}
{\overline {|{\cal M}_{12\rightarrow 34}|^2} }
= 16G_F^2~ [(C_V+C_A)^2 ~(p_1 \cdot p_2)~(p_3\cdot p_4) \nonumber \\
\hspace*{3cm} + (C_V-C_A)^2 ~(p_1\cdot p_4)~(p_2\cdot p_3) \nonumber \\
\hspace*{3.5cm} - (C_V^2-C_A^2) ~(p_2\cdot p_4)~(p_1\cdot p_3) ] \,,
\label{mes}
\end{eqnarray}
where the overline on ${\cal M}$ denotes a sum over final spins and an average
over the initial spins, and $p_i$ denotes the four momenta of particles
$i=1,4$. The squared matrix element for the charged current
reactions is given by a similar relation, but with the replacement
$C_V\rightarrow g_V$, $C_A\rightarrow g_A$, and $G_F \rightarrow G_FC$,
where $C=\cos\theta_C$ for a change of strangeness $\Delta S=0$ and
$C=\sin\theta_C$ for $\Delta S=1$, consistent with the Cabibbo theory.
\vskip 5pt
\section{Neutrino mean free paths in degenerate matter}
We turn now to consider the mean free path of neutrinos in stellar matter
comprised of degenerate baryons (neutrons, protons and hyperons) and leptons
under conditions of charge neutrality and chemical equilibrium. For the
estimates below, we employ the non-relativistic approximation
for the baryons, so that the squared matrix element takes a simple form.
We treat the neutrinos only in the degenerate and non-degenerate limits.
Results for arbitrary neutrino degeneracy and with the full matrix element in
Eq.~(\ref{mes}) will be reported elsewhere.
For elastic collisions $1 + 2 \rightarrow 3 + 4$, where 1(3) denotes the
initial (final) neutrino and 2(4) denotes the the initial (final) baryon $B$,
the scattering relaxation time may be calculated by linearizing the Boltzmann
equation. For small departure from equilibrium, the inverse relaxation times
from the various components are additive; thus
\begin{eqnarray}
\frac{1}{\tau_s} =\sum_2 g_2 \int \prod_{i=2}^4 \frac {d^3p_i}{(2\pi)^3}
~{\cal W}_{fi}~[n_2(1-n_3)(1-n_4)~-~(1-n_2)n_3n_4] \,.
\end{eqnarray}
Above, the sum is over all species of baryons, $g_2$ is their degeneracy,
$n_i$ are the equilibrium
Fermi-Dirac distributions, and ${\cal W}_{fi}$ is the scattering rate
\begin{eqnarray}
{\cal W}_{fi} &=& \left({\prod_{i=1}^4~2E_i}\right)^{-1}~ (2\pi)^4
\delta^4 (p_1+p_2-p_3-p_4)~
{\overline {|{\cal M}_{12\rightarrow 34}|^2} } \,.
\end{eqnarray}
The relaxation time $\tau_s$ characterizes the rate of change of
the distribution function $n_1$ due to interactions with species 2, and may be
used to define a scattering mean free path $\lambda_s = c\tau_s$.
For elastic scattering on heavy fermions, the momentum transfer is small.
Thus,
the scattering rate may be expressed as a function of the neutrino energy
$E_{\nu}$ and the neutrino scattering angle $\theta$. In degenerate matter,
where the participant particles lie on their respective Fermi surfaces, the
phase space integration can be separated into angle and energy
integrals\mycite{GP,Good}. Thus for {\em degenerate neutrinos} and
when $k_BT \ll E_\nu v_{Fi}/c$, where $v_{Fi}$ is the
velocity at the Fermi surface of species $i$, the inverse relaxation time is
given by (see Ref.~[\cite{IP,Saw1}] for the result in a single component
system)
\begin{eqnarray}
\frac {1}{\tau_s} &=& \sum_i \frac {G_F^2}{12\pi^3} (C_{Vi}^2 + 2C_{Ai}^2)
m_{Bi}^2 (k_BT)^2 E_\nu
\left[\pi^2+\frac{(E_\nu-\mu_\nu)^2}{(kT)^2}\right] \,,
\label{scat1}
\end{eqnarray}
where the sum is over the baryonic components present in the system.
For {\em non-degenerate neutrinos} and when $k_BT \ll E_\nu v_{Fi}/c$, the
result of Ref.~[\cite{IP}] may be generalized to give
\begin{eqnarray}
\frac {1}{\tau_s} &=& \sum_i \frac {G_F^2}{15\pi^3} (C_{Vi}^2 + 4C_{Ai}^2)
p_{Fi}^2 E_\nu^3 \,.
\label{scat2}
\end{eqnarray}
The relaxation time for absorption through charged current reactions can be
calculated in a similar fashion. When neutrinos are degenerate, absorption
on neutrons\mycite{Iwa2} and similarly on hyperons\mycite{pplp}, is
kinematically allowed. In this case, the inverse absorption
length is given by\mycite{pplp}
\begin{eqnarray}
\frac {1}{\tau_a} &=& \sum_j \frac {G_F^2C^2}{4\pi^3} (g_{Vj}^2 + 3g_{Aj}^2)
m_{B_1} m_{B_2} (k_BT)^2 \mu_e
\left[\pi^2+\frac{(E_\nu-\mu_\nu)^2}{(kT)^2}\right] \Xi \nonumber\\
{\rm with} ~~ \Xi &=&
\theta (p_{B_2} + p_e - p_{B_1} - p_\nu) \nonumber\\
&+& {\displaystyle\frac {p_{B_2}+p_e - p_{B_1} + p_\nu}{2E_\nu} }
{}~\theta(p_\nu - |p_{B_2}+p_e-p_{B_1}|)
\label{abs1}
\end{eqnarray}
where $\theta(x)=1$ for $x\geq 1$ and zero otherwise.
When neutrinos are {\em non-degenerate} and when absorption is
kinematically allowed, the relaxation time is
given by
\begin{eqnarray}
\frac {1}{\tau_a} = \sum_j \frac {G_F^2C^2}{4\pi^3} (g_{Vj}^2 + 3g_{Aj}^2)
m_{B_1} m_{B_2} (kT)^2 \mu_e
\left[\pi^2+\frac{E_\nu^2}{(kT)^2}\right]
\displaystyle{\frac {1}{1+e^{-E_\nu/kT} } } \,.
\label{abs2}
\end{eqnarray}
We note that for nucleons-only matter, neutrino
absorption on single nucleons can proceed only if the proton concentration
exceeds some critical value in the range $(11-15)\%$ (see Ref.~[\cite{lpph}]).
For matter with lower proton concentrations, neutrino absorption occurs on two
nucleons. However, as shown in Ref.~[\cite{pplp}],
neutrinos may be absorbed on single hyperons as long as the
concentration of $\Lambda$ hyperons exceeds a critical value that is less than
$3\%$ and is typically about $1\%$.
\vskip 5pt
\section{ Results and discussions }
The calculations above give the mean free path in a mixture in which
all baryons are under degenerate conditions. Several effects of strong
interactions must
be included before the results in Eq.~(\ref{scat1}) through Eq.~(\ref{abs2})
may be utilized. The renormalization of the density of states at the Fermi
surfaces results in the baryon masses $m_B$ being replaced by $m_B^* =
p_F/v_F$. As pointed out by Iwamoto and Pethick\mycite{IP}, baryon-baryon
interactions introduce further Fermi liquid corrections. Specifically, the
axial vector interaction is significantly suppressed, which increases the mean
free path of neutrinos in dense matter. So far, such an analysis has been
restricted to pure neutron matter only. In a multicomponent system, this
formalism must be extended to include correlations between the different
species present. The effect of density correlations in the long wavelength
limit can be related to the compressibility of the system\mycite{Saw1}. In
addition, the medium dependence of $g_A$ must also be considered. For
example, it has been argued\mycite{WRB} that $|g_A|$ is quenched for nucleons
in a medium. Whether $g_A$ for hyperons is similarly affected is not known and
is worth studying. Finally, depending on the momentum transfers involved,
final-state interactions may modify the weak-interaction matrix element. Of
the many corrections mentioned above, those due to the effective mass are the
easiest to incorporate, since it would be contained even in a mean field
description of the equation of state. Pending a more complete analysis, we
consider below the modifications introduced by the multicomponent nature of
the system incorporating only the effective mass corrections.
When neutrinos are degenerate, the relative abundances of the individual
components do not play a significant role in determining the total mean free
path. However, the extent to which the neutrino mean free path is altered by
the presence of hyperons depends on the number of hyperonic species present.
This may be illustrated by using the results for
${\cal R}^{(1)}_{nc} = [C_V^2+2C_A^2]_{ \nu + B \rightarrow
\nu + B } / [C_V^2+2C_A^2]_{ \nu + n \rightarrow \nu + n }$
and ${\cal R}_{cc} = [C^2(g_V^2+3g_A^2)]_{ \nu + B_1 \rightarrow
\ell + B_2 } / [C^2(g_V^2+3g_A^2)]_{ \nu + n \rightarrow \ell + p} $
listed in Tables 1 and 2. For example, in matter containing $\Lambda,\Sigma^-$
and $\Sigma^0$ hyperons, the scattering mean free path is reduced by about
$30-50\%$ from its value in nucleons only matter. This reduction is
achieved mainly by scattering on the $\Sigma^-$ hyperon, since the $\Lambda$
and $\Sigma^0$ hyperons do not contribute to scattering in lowest order.
Similarly, up to $50\%$ reduction may be expected from absorption reactions.
Again, reactions involving the $\Sigma^-$ hyperon, particularly the one leading
to the $\Sigma^0$ hyperon, give the largest contribution.
The relative importance of the scattering and absorption reactions by
degenerate neutrinos may be inferred by noting that
\begin{eqnarray}
\frac{\lambda_{a}}{\lambda_{s}}&=&\frac{1}{3}\displaystyle
\frac{E_\nu}{\mu_e}~
\left(\frac{\sum_i~(C_{Vi}^2+2C_{Ai}^2)m^{*^2}_{B_i} }
{\sum_j~C^2(g_{Vj}^2+3g_{Aj}^2)m^*_{B_1}m^*_{B_2} } \right) \,.
\end{eqnarray}
For a fixed lepton fraction and for neutrinos of energy $E_\nu \sim \mu_\nu$,
the factor $\mu_\nu/\mu_e \leq 1$ (see Figs. 1 and 2). Inasmuch as the baryon
effective masses are all similar in magnitude, the factor in the parenthesis
above may be approximated by the ratio of the coupling constants. From the
results in Tables 1 and 2, it is easy to verify that the factor containing the
coupling constants is of order unity (but generally less than unity) and is not
very sensitive to the number of hyperonic species present. We thus arrive at
the result that the absorption reactions dominate over the scattering reactions
by a factor of about three to five, even in the presence of hyperons. This
conclusion is not affected by the inclusion of the baryon effective masses in
the calculation of $\lambda_a/\lambda_s$.
For non-degenerate neutrinos, and when $k_BT \ll E_\nu v_{Fi}/c$,
\begin{eqnarray}
\frac{\lambda_{a}}{\lambda_{s}} = \displaystyle
\frac{4}{15}~
\left(\frac{\sum_i~(C_{Vi}^2+4C_{Ai}^2) p_{Fi}^2E_\nu^3 (1+e^{-E_\nu/kT}) }
{\sum_j C^2(g_{Vj}^2 + 3g_{Aj}^2)
m^*_{B_1} m^*_{B_2} (kT)^2 \mu_e
\left[\pi^2+ {\displaystyle\frac{E_\nu^2}{(kT)^2}} \right] } \right) \,.
\end{eqnarray}
In this case, the abundances of the various particles play a significant role
in determining whether or not the absorption reactions dominate over the
scattering reactions. Some insight may be gained by examining the
concentrations in Figs. 1 and 2, and also the ratio ${\cal R}^{(2)}_{nc} =
[C_V^2+4C_A^2]_{ \nu + B \rightarrow \nu + B } / [C_V^2+4C_A^2]_{ \nu + n
\rightarrow \nu + n }$ listed in Table 2.
\vskip 5pt
\section{ Outlook }
Neutrino signals in terrestrial detectors offer a means to determine the
composition and the equation of state of dense matter. Many calculations of
the composition of dense matter have indicated that strangeness-rich matter
should be present in the core of neutron stars. Possible candidates for
strangeness include hyperons, a $K^-$ condensate, and quark matter containing
$s$-quarks. Neutrino opacities for strange quark matter have been calculated
previously\mycite{Iwa2}. In this work, we have identified the relevant
neutrino-hyperon scattering and absorption reactions that are important new
sources of opacity. Much work remains to be done, however. Many-body
effects which may introduce additional correlations in a mixture are worth
studying. Full simulations, of the type carried out in Refs.~[\cite{BL,KJ}]
including possible compositions and the appropriate neutrino opacities, will be
taken up separately.
\vskip 5pt
\section{Acknowledgements}
This work was supported by the U.S. Department of Energy under contract number
DOE/DE-FG02-88ER-40388. We thank Jim Lattimer for helpful discussions.
|
\section{Introduction}
In a thorough paper
(Ref. \cite{Kuchar}, hereafter referred to as KVK)
Kucha\v{r} has examined the
canonical reduction of the most general action functional
describing the geometrodynamics of the maximally extended
Schwarzschild geometry. This reduction yields the true degrees
of freedom for vacuum general relativity subject to
ansatz of spherical symmetry. The key technical ingredient
in Kucha\v{r}'s analysis is a canonical transformation to
a certain chart on the
gravitational phase space which features the Schwarzschild
mass parameter $M_{\scriptscriptstyle S}$,
expressed in terms of what are essentially
Arnowitt-Deser-Misner (ADM) variables,\cite{ADM} as a
canonical coordinate. (Kucha\v{r}'s paper
complements earlier work\cite{Thiemann} by Kastrup and
Thiemann, based mostly on Ashtekar variables, which has
also explicitly isolated the true degrees of freedom for
vacuum SSGR.)
Potential applications of the new reduced
formalism include examinations, from the canonical viewpoint,
of spherically symmetric collapse, the Hawking effect, and
equilibrium gravitational thermodynamics.
Indeed, Louko and Whiting have already made such an application.
Applying Kucha\v{r}'s method to a spatially bounded exterior
region of the Schwarzschild black hole, they have constructed
the Schwarzschild thermodynamical (canonical) partition
function completely within the Lorentzian Hamiltonian
framework (Ref.\cite{LW}, hereafter referred to as LW).
Their canonical partition function is in agreement with
previous results derived via the Euclidean-path-integral
method.\cite{BMYW} The starting point in LW is the
``thermodynamical action,'' which is of central importance
in the path-integral formulation of gravitational
thermodynamics. A very delicate issue in the analysis of LW
concerns the treatment of the thermodynamical action's
boundary terms under the canonical reduction via the
KVK method.
In this paper we discuss the geometric interpretation of
Kucha\v{r}'s canonical transformation. By appealing
to notions of quasilocal energy and momentum in general
relativity which have been given by Brown and
York\cite{BY,BYL}, we interpret Kucha\v{r}'s canonical
transformation as a ``sphere-dependent boost to the rest
frame," where the ``rest frame'' is defined by vanishing
quasilocal momentum. It seems to us that KVK finds essentially
the same interpretation via an alternate route involving
subsequent canonical transformations and a ``parameterization
at infinity.'' We believe that our approach complements the
one given by KVK. The center-stone of ours is
the following observation. On
an arbitrary (spherically symmetric) spatial slice
$\Sigma$ the parameter $\varphi$ describing the local boost
between the slice Eulerian observers at a point and the
rest-frame observers at the same point can be constructed
from the canonical variables of $\Sigma$ (if known in a tiny
spatial region surrounding the point of interest).
Furthermore, we work in a framework which is general enough
to cover Witten's two-dimensional dilaton gravity
(2dDG).\cite{Witten} (Further still, many of our preliminary
results concerning the theory of quasilocal energy-momentum are
unaffected by the inclusion of minimally coupled matter and
hence are relevant for the two-dimensional dilaton-plus-matter
model of Callan, Giddings, Harvey, and Strominger.\cite{CGHS})
Therefore, besides reviewing some of Kucha\v{r}'s original
work for Schwarzschild black holes from the framework of
hyperbolic geometry, we present new results concerning the
canonical reduction of Witten-black-hole geometrodynamics.
We show that the canonical transformation of KVK can also be
made in the context of (vacuum) 2dDG. Therefore, the potential
applications of the KVK
formalism, listed in the first paragraph, are also relevant
for 2dDG.
Finally, with our general framework we address some of
the delicate points, first considered in LW, concerning the
canonical reduction of the ``thermodynamical action.''
Our conceptual framework supports the difficult technical
steps taken in LW. All of our results are given
for both the Schwarzschild and Witten-black-hole cases.
A few technical points demand some comment at the outset.
As mentioned, the analysis of KVK concerns the full Kruskal
spacetime, the maximally extended Schwarzschild geometry.
The canonical variables used in KVK are defined on spatial
slices which cut completely across the Kruskal diagram, and
therefore have to obey appropriate boundary conditions in the
asymptotic regions. In crossing from one spatial infinity to the
other, the slices of KVK are allowed to cross the horizon in
a completely general way. This introduces some technical
difficulties at the horizon, especially when one is considering
Kucha\v{r}'s canonical transformation. However, as demonstrated
in KVK, with care these difficulties may be surmounted.
We choose
to confine our attention entirely to the right static region of
the Kruskal diagram. At first, we work with the time history of a
static region lying between concentric spheres. Thus we avoid
many of the technical difficulties faced by Kucha\v{r} at the
outset. We could, of course, work in the full Kruskal diagram,
but the essential points of this paper do not demand that we do
so. However, since we do chose to bypass a technical
treatment of the horizon, questions concerning how to handle such
horizon difficulties remain for the Witten black holes of 2dDG.
However, notationally we adopt nearly the same conventions as KVK.
Therefore, we expect that with the present paper as a stepping stone,
the interested reader could -with minimal effort- convert any and
all of the horizon arguments given in KVK into corresponding
arguments applicable to the Witten black-hole case.
The layout of this paper is as follows. In $\S$ {\sc II}, the preliminary
section, we describe the relevant kinematics of our spacetime
geometry. Since the spacetime geometry is spherically symmetric,
it proves convenient to work with a toy $1+1$ dimensional
spacetime $\cal M$. In reality, the points of $\cal M$ are round
spheres. In $\S$ {\sc III} we derive quasilocal\footnote{We
use the adjective ``quasilocal" because from
the four-dimensional viewpoint the energy and momentum
are associated round two-spheres. Thus these ``quasilocal"
quantities are associated with points of $\cal M$.} energy
and momentum expressions for the physical fields
defined on a generic spatial
slice $\Sigma$ of $\cal M$. The method used to derive the quasilocal
expressions is a Hamilton-Jacobi analysis of an appropriate action
principle for $\cal M$. In $\S$ {\sc IV} we use the developed notions of
quasilocal energy and momentum to underscore the geometric
significance of Kucha\v{r}'s canonical transformation. This
section also considers the reduction of the canonical action
with the boundary conditions adopted in this work.
Finally, $\S$ {\sc V} considers the canonical reduction of the
thermodynamical action.
\section{Preliminaries}
\subsection{Spacetime $\cal M$ and foliations}
Consider a $1+1$ dimensional spacetime region $\cal M$ which is
bounded spatially. The region $\cal M$ consists of a collection
of one dimensional spacelike slices $\Sigma$. The letter $\Sigma$
denotes both a foliation of $\cal M$ into spacelike slices and a
generic leaf of such a foliation. However, for the initial
spacelike slice we reserve the special symbol $t'$ (also the
value of the coordinate time on the initial slice), and,
likewise, for the final spacelike slice we reserve the symbol
$t''$ (also the value of the coordinate time on the final slice).
On spacetime $\cal M$ we have coordinates $(t, r)$,
and a generic spacetime point\footnote{One may consider each point
of $\cal M$ to be round two-sphere with radius $R(t,r)$, where $R$
is the radius function.} is $B(t,r)$. Every constant-$t$ slice $\Sigma$
has two boundary points $B_{i}$ (at $r = r_{i}$) and $B_{o}$
(at $r = r_{o}$). Assume that along $\Sigma$ the coordinate $r$
increases monotonically from $B_{i}$ to $B_{o}$. We represent the
timelike history $B_{i}(t) \equiv B(t, r_{i})$ by $\bar{\cal T}_{i}$
(unbarred $\cal T$ is reserved for another meaning) and refer to it
as the {\em inner boundary}. Likewise, we represent the timelike
history $B_{o}(t) \equiv B(t, r_{o})$ by $\bar{\cal T}_{o}$ and refer
to it as the {\em outer boundary}. Later on, when we deal with
black-hole solutions, we will ``seal" the inner boundary. In other
words, the time development at the inner boundary will be arrested,
and the point $B_{i}$ will correspond to a bifurcation point in a
Kruskal-like diagram. We denote the {\em corner} points of our
spacetime as follows: $B'_{i} \equiv B(t',r_{i})$, $B'_{o} \equiv
B(t',r_{o})$, $B''_{i} \equiv B(t'',r_{i})$, and $B''_{o} \equiv
B(t'',r_{o})$.
We can also consider a {\em radial} foliation of $\cal M$
by a family of one-dimensional timelike surfaces which extend
from $\bar{\cal T}_{i}$ outward to $\bar{\cal T}_{o}$ (for
black-hole solutions $\bar{\cal T}_{i}$
may be a degenerate sheet).\cite{Hayward}
These are constant-$r$ surfaces.
Like before, we loosely use the letter $\bar{\cal T}$ both to denote
the radial foliation and a generic leaf of this foliation. We
call $\bar{\cal T}$ leaves {\em sheets}, whereas we have called
$\Sigma$ leaves {\em slices} (this is an informal convention).
Abusing the notation a bit, we also often let the symbol
$\bar{\cal T}$ denote the total timelike boundary
$\bar{\cal T}_{i} \bigcup \bar{\cal T}_{o}$.
However, when the symbol $\bar{\cal T}$ has this meaning, it
always appears in the phrase ``the $\bar{\cal T}$ boundary.''
\subsection{Metric decompositions}
The spacetime metric is $g_{ab}$. The metric on a generic
$\Sigma$ slice is $\Lambda^{2}$, and the metric on the
$\bar{\cal T}$ boundary is denoted by
$- \bar{N}^{2}$. In terms of the $\Sigma$ foliation, the
metric may be written in ADM form\cite{ADM}
\begin{equation}
g_{ab}{\rm d}x^{a} {\rm d}x^{b}
= - N^{2} {\rm d} t^{2} + \Lambda^{2}\left({\rm d}r + N^{r}
{\rm d}t\right)^{2}\, ,\label{ADM}
\end{equation}
with $N$ and $N^{r}$ denoting the familiar {\em lapse}
and {\em shift}. The vector field
\begin{equation}
u^{a} \partial/\partial x^{a}
= N^{-1}\left(\partial/\partial t
- N^{r} \partial/\partial r \right)
\end{equation}
is the unit, timelike, future-pointing normal to the
$\Sigma$ foliation.
In terms of the $\bar{\cal T}$ foliation, the $\cal M$ metric
takes the form
\begin{equation}
g_{ab}{\rm d}x^{a} {\rm d}x^{b}
= \bar{\Lambda}^{2} {\rm d} r^{2}
- \bar{N}^{2}\left({\rm d}t +
\bar{\Lambda}^{t}{\rm d}r\right)^{2}\, ,
\end{equation}
where $\bar{\Lambda}$ and
$\bar{\Lambda}^{t}$ are the {\em radial
lapse} and {\em radial shift}.
The unit, spacelike, $\bar{\cal T}$-foliation
normal is
\begin{equation}
\bar{{{\boldmath \mbox{$n$}}}}^{a} \partial/\partial x^{a} =
\bar{\Lambda}^{-1}\left(\partial/\partial r
- \bar{\Lambda}^{t}\partial/\partial t\right)\, .
\end{equation}
We also define the $\bar{\cal T}$ {\em boundary}
normal $\bar{n}^{a}$ on $\bar{\cal T}_{i}$ and $\bar{\cal T}_{o}$
by the requirement that it always be {\em outward}-pointing on
these boundary elements. On the outer boundary
$\bar{\cal T}_{o}$ the outward normal is
$\bar{n}^{a} = \bar{{{\boldmath \mbox{$n$}}}}^{a}$, while on the inner boundary
$\bar{\cal T}_{i}$ the outward normal is
$\bar{n}^{a} = - \bar{{{\boldmath \mbox{$n$}}}}^{a}$. (For a handful
$\{n^{a}, \bar{n}^{a}, v, w, \eta, \varphi, E\}$ of symbols
we use this convention throughout the paper. Regular letters
represent objects associated with the boundary and have the
appropriate sign for each boundary element built in.
Boldface versions of the same letters represent the same objects
but with the fixed sign appropriate for the outer boundary.
Perhaps it would have been a more natural choice to
adopt the reciprocal notation and let the boldface letters
possess the sign flexibility. We have made the seemingly unnatural
choice simply because it leads to the minimal use of boldface
letters. In other words, in what follows the sign-flexible
quantities happen to appear more frequently than their fixed-sign
counterparts.)
By equating the coefficients of the above forms of $g_{ab}$, we obtain
the following relations between the ``barred" and ``unbarred" variables:
\begin{eqnarray}
\bar{N} & = & N/\gamma \label{transformation}
\eqnum{\ref{transformation}a} \\
\bar{\Lambda} & = & \gamma \Lambda
\eqnum{\ref{transformation}b}\\
\Lambda N^{r}/N & = & - \bar{N}
\bar{\Lambda}^{t}/\bar{\Lambda}\, .
\eqnum{\ref{transformation}c}
\addtocounter{equation}{1}
\end{eqnarray}
Here $\gamma \equiv (1 - {{\boldmath \mbox{$v$}}}^{2})^{-1/2}$ is the
local relativistic factor
associated with the velocity ${{\boldmath \mbox{$v$}}} \equiv \Lambda N^{r}/ N
= - \bar{N} \bar{\Lambda}^{t}/\bar{\Lambda}
= - \bar{{{\boldmath \mbox{$v$}}}}$.\cite{Hayward} The timelike normal
associated with the foliations
$B_{i}(t)$ and $B_{o}(t)$ of the boundary sheets
$\bar{\cal T}_{i}$ and $\bar{\cal T}_{o}$ is
$\bar{u}^{a} \partial /\partial x^{a}
= \bar{N}^{-1} \partial/\partial t$. Note
that on the $\bar{\cal T}$ boundary the vector fields
$u^{a}$ and $\bar{u}^{a}$ need not
coincide. Also, fixation of the $t$ coordinate gives a
collection of points $B(r)$ which foliates the slice $\Sigma$.
The normal associated with this foliation of $\Sigma$ is
${{\boldmath \mbox{$n$}}}^{a} \partial/\partial x^{a}
= \Lambda^{-1}\partial /\partial r $.
Again, we define a boundary
normal $n^{a}$ such that at the inner boundary
$n^{a}$ is $- {{\boldmath \mbox{$n$}}}^{a}$, the outward-pointing
normal of $B_{i}$ as embedded in
$\Sigma$, while at the outer boundary
$n^{a}$ is ${{\boldmath \mbox{$n$}}}^{a}$, the
outward-pointing normal of $B_{o}$ as embedded in $\Sigma$.
On the inner and outer boundaries $n^{a}$ and
$\bar{n}^{a}$ need not coincide.
It is easy to verify the following point-wise boost relations:
\begin{eqnarray}
\bar{u}^{a} & = & \gamma u^{a} + v \gamma n^{a}
\label{boundaryboost} \eqnum{\ref{boundaryboost}a} \\
\bar{n}^{a} & = & \gamma n^{a} + v \gamma u^{a}\, ,
\eqnum{\ref{boundaryboost}b}
\addtocounter{equation}{1}
\end{eqnarray}
where $v = - {{\boldmath \mbox{$v$}}}$ on $\bar{\cal T}_{i}$ and
$v = {{\boldmath \mbox{$v$}}}$ on $\bar{\cal T}_{o}$.
\subsection{Extrinsic curvatures}
The extrinsic curvature of a $\Sigma$ slice as embedded in
$\cal M$ is ${{\cal K}} \equiv - \nabla_{a}u^{a}$, where
$\nabla_{a}$ denotes the torsion-free covariant derivative
operator compatible with $g_{ab}$.
One may also consider spacelike slices
$\bar{\Sigma}$ which are everywhere orthogonal by assumption
to the $\bar{\cal T}$ sheets.
Since the spacetime-filling extension
of the $\bar{\cal T}$ sheets from $\bar{\cal T}_{i}$
to $\bar{\cal T}_{o}$ is arbitrary, the $\bar{\Sigma}$ slices
are almost as general as the $\Sigma$ slices. However, the
$\bar{\Sigma}$ slices are restricted by the requirement that
their normal vector field coincides with $\bar{u}^{a}$ on
the $\bar{\cal T}$ boundary. We describe such slices as
{\em clamped}. When the velocity $v$
defined above is set to zero on
the boundary, then the $\Sigma$ slices are clamped. (In which
case, there is no longer a need to make a distinction
between barred and unbarred slices.) Also define
$\bar{{{\cal K}}} \equiv - \nabla_{a}\bar{u}^{a}$,
the extrinsic curvature associated with the $\bar{\Sigma}$
slices.
The extrinsic curvature associated with the $\bar{\cal T}$
boundary elements is defined by $\bar{\vartheta} \equiv -
\nabla_{a}\bar{n}^{a}$. We may also
consider a foliation $\cal T$ generated by the $u^{a}$ Eulerian
histories of points in the $\Sigma$ slices. At the boundary,
the $\cal T$ sheets may be ``crashing into" or ``emerging from"
the actual boundary elements $\bar{\cal T}_{i}$
and $\bar{\cal T}_{o}$. Nevertheless, one can define an
extrinsic curvature
$\vartheta \equiv - \nabla_{a} n^{a}$.
The value of $\vartheta$ at a particular boundary point of
the $\bar{\cal T}$ boundary is associated with the $\cal T$ sheet
intersecting that point. We have that $\vartheta = - n_{b}
a^{b}$, where $a^{b} \equiv u^{a}\nabla_{a} u^{b}$ is the
spacetime acceleration of $u^{a}$. Since by assumption the
metric-compatible connection associated with $\cal M$ is
torsion-free, $- \vartheta = n^{a} \nabla_{a} [\log N]$.
With our transformation equations (\ref{boundaryboost}),
one can derive the following ``splitting" formulae for
$\bar{{\cal K}}$ and $\bar{\vartheta}$:
\begin{eqnarray}
\bar{{{\cal K}}} & = & \gamma {{\cal K}} + v \gamma \vartheta -
\bar{n}^{a} \nabla_{a} \eta \label{splittings}
\eqnum{\ref{splittings}a} \\
& &\nonumber \\
\bar{\vartheta} & = & \gamma \vartheta + v \gamma {{\cal K}} -
\bar{u}^{a} \nabla_{a} \eta\, , \eqnum{\ref{splittings}b}
\addtocounter{equation}{1}
\end{eqnarray}
where $\eta \equiv \tanh^{-1} v
= \frac{1}{2}\log |(1 + v)/(1 - v)|$.
Note that $\eta = \tanh^{-1} - {{\boldmath \mbox{$v$}}}$
on $\bar{\cal T}_{i}$ and $\eta = \tanh^{-1}
{{\boldmath \mbox{$v$}}}$ on $\bar{\cal T}_{o}$. In accord with our conventions
also set ${{\boldmath \mbox{$\eta$}}} \equiv \tanh^{-1} {{\boldmath \mbox{$v$}}}$.
\section{Action and quasilocal energy-momentum}
We begin this section by precisely defining the type
of action functional associated with our bounded
spacetime region $\cal M$ which is of interest in
this work. We will discuss in detail the action's
associated variational principle, paying strict
attention to all boundary terms. This is the
background work necessary to derive expressions for
the quasilocal energy and momentum
associated with a generic bounded slice $\Sigma$.
\subsection{Variational principle}
Our analysis begins with the following action functional:
\begin{eqnarray}
{\cal S}^{\it 1} & = &
{\textstyle \frac{1}{4}}\,\alpha\!\int_{\cal M}
{\rm d}^{2}x \sqrt{-g} e^{- 2 \Phi}\left[ {\cal R} +
2y g^{ab} \nabla_{a}\Phi \nabla_{b}\Phi
+ 2\lambda^{2}y \exp\left[(2 - y)2\Phi\right]\right]
\nonumber \\
& &
+ {\textstyle \frac{1}{2}}\,\alpha\!\int^{t'}_{t''}
{\rm d}r \Lambda e^{- 2 \Phi}{{\cal K}}
- {\textstyle \frac{1}{2}}\,\alpha\!\int_{\bar{\cal T}}
{\rm d}t \bar{N} e^{- 2 \Phi} \bar{\vartheta}
- \left. {\textstyle \frac{1}{2}}\,\alpha e^{- 2 \Phi} \eta
\right|^{B''}_{B'}
\label{baseaction}\, ,
\end{eqnarray}
where ${\cal R}$ is the scalar curvature of $\cal M$ built
from the metric $g_{ab}$, and the scalar field $\Phi$ is the
celebrated {\em dilaton}. The variable $y$ is an as-yet
unspecified number, $\lambda$ is a positive
constant with dimensions of inverse length, and
$\alpha$ is another positive (and possibly dimensionful)
constant. The ${\cal M}$ integral in our action corresponds
to a subclass of models within the larger framework
of generalized dilaton theories.\cite{Kummer}
In (\ref{baseaction}) and throughout the rest of this
paper we use the following short-hand notations:
\begin{eqnarray}
\int_{t'}^{t''} & = & \int_{t''} - \int_{t'}
\eqnum{\ref{Tshort}a} \\
\int_{\bar{\cal T}} & = & \int_{\bar{\cal T}_{i}} +
\int_{\bar{\cal T}_{o}} \label{Tshort}
\eqnum{\ref{Tshort}b} \\
\Bigl.\Bigr|^{B''}_{B'} & = &
\Bigl.\Bigr|_{B''_{i}} - \Bigl.\Bigr|_{B'_{i}}
+ \Bigl.\Bigr|_{B''_{o}} - \Bigl.\Bigr|_{B'_{o}}
\,\,\,\, ,
\eqnum{\ref{Tshort}c}
\addtocounter{equation}{1}
\end{eqnarray}
where (\ref{Tshort}a) is used only when
$t'$ and $t''$ represent the initial and final slices
and not integration parameters.
The boundary terms in the above action
ensure that its associated variational
principle features fixation of
induced metric and the dilaton on the
boundary. Symbolically, we
could collect the boundary terms into one expression
\begin{equation}
\left({\cal S}^{\it 1}\right)_{\partial {\cal M}} =
{\textstyle \frac{1}{2}}\,
\alpha\!\int_{\partial {\cal M}}{\rm d}x
{\textstyle \sqrt{|{}^{1}\! g|}}
e^{-2\Phi} {{\cal K}}\, ,
\end{equation}
where here ${{\cal K}}$ is used for the extrinsic
curvature over the whole
boundary $\bar{\cal T}_{i} \bigcup \bar{\cal T}_{o}
\bigcup t' \bigcup t''$.
The corner-point contributions in (\ref{baseaction})
are included because,
though the corner points are a set of measure zero
in the integration of
$e^{-2\Phi} {{\cal K}}$ over all
of $\partial {\cal M}$, the term ${{\cal K}}$ becomes
infinite at the corners,
since the boundary normal changes discontinuously
from $u^{a}$ to $\bar{n}^{a}$ at
these points.\cite{Hayward}
Finally, we write ${\cal S}^{\it 1}$ for the action, because we
anticipate the need to append to the action a Gibbons-Hawking
subtraction term $- {\cal S}^{\it 0}[\bar{N},\Lambda,\Phi]$ (a
functional of the fixed boundary data).\cite{Gibbons,BY}
In this case the full action would be ${\cal S} \equiv
{\cal S}^{\it 1} - {\cal S}^{\it 0}$. We briefly
consider the more general action
later in an appendix. Also, we could add to the
action a matter contribution ${\cal S}^{\it m}$. Most of our work
in this section on quasilocal energy-momentum would be unaffected
by an ${\cal S}^{\it m}$ contribution to the action,
as long as the matter fields were minimally coupled. However,
an ${\cal S}^{\it m}$
contribution would affect the following sections
devoted to the canonical reduction. Therefore, for the sake
of simplicity we do not consider an ${\cal S}^{\it m}$ further.
We are interested in (vacuum)
spherically symmetric general relativity
(SSGR). As was first noted by Thomi, Isaak, and
H\'{a}j\'{\i}\v{c}ek,\cite{action} a suitable
action for SSGR is given by (\ref{baseaction})
with the choices $y = 1$, $\alpha = \lambda^{-2}$.
Note that for the SSGR case, we set $\alpha$
equal to the {\em dimensionful} constant $\lambda^{-2}$. This gives
our action the units of action in four dimensions.
An appendix shows that this action can be obtained via a
reduction by spherical symmetry of a covariant first-order
version\cite{York,Hayward}
of the Einstein-Hilbert action (where the {\em four dimensional}
action principle is associated with the time history ${}^{4}\!{\cal M}$
of a spatial region bounded by concentric spheres).
In this correspondence with SSGR, it turns out that the radius of
a round sphere is given by
\begin{equation}
R = \sqrt{\frac{A}{4\pi}} =
\frac{e^{-\Phi}}{\lambda}\, ,\label{radius}
\end{equation}
where $A$ stands for the proper area of the sphere.
In this paper we also consider (vacuum) two-dimensional
dilaton gravity (2dDG), which corresponds to $y = 2$ and
$\alpha$ taken to be a dimensionless positive number.
Often $\alpha$ is chosen to be $2/\pi$, but $\alpha$ remains
essentially arbitrary.
The arbitrariness of $\alpha$ for vacuum 2dDG
corresponds to the freedom to shift the dilaton
by a constant $\Phi \rightarrow \Phi - \log \sqrt{\alpha'}$
(a freedom not present for the SSGR action). Under such a shift,
$\alpha \rightarrow \alpha'' = \alpha\alpha'$. We would actually
prefer to set $\alpha = 2$ for reasons which become clear
later. Adding the right matter contribution $S^{m}$ to the vacuum 2dDG
action, we would obtain the CGHS model.\cite{CGHS} Though we
do not consider this matter contribution, the expressions for
quasilocal energy-momentum that we derive in this section
are nevertheless valid for the full CGHS theory.
For the SSGR case the action
${\cal S}^{\it 1}$ has dimensions of length-squared, while for
2dDG the action ${\cal S}^{\it 1}$ is dimensionless.
This difference in units will propagate throughout
all the formulae to follow.
However, the freedom of allowing $\alpha$
to be either $\lambda^{-2}$
or a plain number will automatically
keep track of the correct units for both cases. For the SSGR
case $(y = 1, \alpha = \lambda^{-2})$ all of our conventions
have been tailored to match those of KVK and LW.
The first step is to compute the variation of the action.
One can compute the variation in a number of ways, but the fastest
way is the following. Note that $\cal R$ is a pure divergence,
and then perform an integration by parts on the $\cal R$ term in the
action.
This leads to cancelation of
most of the boundary terms. This short calculation and resulting
form of the action are given later in the discussion in the text
preceding (\ref{3+1action}). Vary the resulting form of the action
(\ref{3+1action}) to find
\begin{eqnarray}
\delta {\cal S}^{\it 1} & = &
\left({\rm terms\,\,giving\,\,the\,\,equations\,\,
of\,\,motion}\right) \nonumber \\
& & + \int^{t''}_{t'}{\rm d}r
\left( P_{\Lambda}\delta\Lambda
+ P_{\Phi}\delta\Phi\right)
+ \int_{\bar{\cal T}}{\rm d}t
\left( \bar{\Pi}_{\bar{N}}\delta\bar{N}
+ \bar{\Pi}_{\Phi}\delta\Phi\right)
+ \left. \alpha
e^{- 2 \Phi} \eta \delta\Phi
\right|^{B''}_{B'}\, , \label{variation}
\end{eqnarray}
where we have defined the momenta
\begin{eqnarray}
P_{\Lambda} & \equiv & \alpha
e^{- 2\Phi} u^{a} \nabla_{a} \Phi \label{momenta}
\eqnum{\ref{momenta}a} \\
P_{\Phi} & \equiv & - \alpha
e^{- 2 \Phi} \Lambda
\left( {{\cal K}} + y u^{a} \nabla_{a} \Phi\right)
\eqnum{\ref{momenta}b} \\
\bar{\Pi}_{\bar{N}} & \equiv &
- \alpha e^{- 2 \Phi} \bar{n}^{a} \nabla_{a} \Phi
\eqnum{\ref{momenta}c} \\
\bar{\Pi}_{\Phi} & \equiv & \alpha
e^{- 2 \Phi}\bar{N}
\left(\bar{\vartheta}
+ y \bar{n}^{a} \nabla_{a} \Phi\right)\, .
\eqnum{\ref{momenta}d}
\addtocounter{equation}{1}
\end{eqnarray}
Inspection of the
variation of the action shows that $P_{\Lambda}$
is the gravitational momentum conjugate to
$\Lambda$. Likewise, $\bar{\Pi}_{\bar{N}}$ is the
gravitational momentum conjugate to $\bar{N}$,
where now conjugacy is defined with respect
to the $\bar{\cal T}$ boundary. The momentum conjugate to
the dilaton field is $P_{\Phi}$. The momentum
$\bar{\Pi}_{\Phi}$ is the $\bar{\cal T}$ boundary
analog of $P_{\Phi}$.
Note that one might try to define a momentum
$p_{\Phi} \equiv \alpha e^{- 2 \Phi} \eta$ in some
sense conjugate to $\Phi$ at the corners.
However, we chose not to do this. Therefore,
strictly speaking, the canonical action we consider
later has a mixed Hamiltonian-Lagrangian form.
If the corner terms
in the action (\ref{baseaction}) had not been
included, then the variational principle would
have featured fixation of $\eta$ on the corner
points. Fixing $\Phi$ at the corners seems to be
more in harmony with the fact that $\Phi$ is
fixed on $t'$, $t''$, and the $\bar{\cal T}$
boundary.
Our momenta $P_{\Lambda}$ and $P_{\Phi}$ agree
with the analysis of KVK.
To see this take the SSGR case and use the
fact that $P_{R} = -(1/R) P_{\Phi}$. One then finds
precisely Kucha\v{r}'s momenta,
\begin{eqnarray}
P_{\Lambda} & = &
- N^{-1} R \left( \dot{R} - N^{r} R'\right)
\label{KVKmomenta} \eqnum{\ref{KVKmomenta}a}\\
P_{R} & = &
- N^{-1}\left[R\left(\dot{\Lambda}
- (\Lambda N^{r})' \right)
+ \Lambda\left(\dot{R} - N^{r} R'\right)\right]\, .
\eqnum{\ref{KVKmomenta}b}
\addtocounter{equation}{1}
\end{eqnarray}
To get the last expression, we have used the definition
of ${{\cal K}}$ given in the preliminary section to find
${{\cal K}} = - (N\Lambda)^{-1}\left[\dot{\Lambda}
- (\Lambda N^{r})'\right]$.
(Note, however, that due to well-entrenched notation
for the dilaton, we unfortunately must break with
the KVK convention of only using Greek
letters to represent spatial densities like
$\Lambda$. Though represented by a Greek letter, the
dilaton $\Phi$ is a scalar and its momentum $P_{\Phi}$
is a density.)
\subsection{Quasilocal energy-momentum}
As advertised, our variational principle has been rigged
so that the induced metric
($\Lambda^{2}, -\bar{N}^{2}$) and the dilaton $\Phi$
are fixed as boundary data. In particular, the lapse of
proper time between the initial and final slices
is fixed in the variational principle, since this
information is encoded in the $\bar{\cal T}$ boundary metric
$- \bar{N}^{2}$. This is the key feature exploited
in the Brown-York theory\cite{BY,BYL}
of quasilocal energy-momentum in general
relativity. Following the basic line of reasoning in
this theory, we ``read off" from the variation of the action
what geometric expressions play the role of quasilocal
energy and momentum in our theory.
We begin by writing the boundary terms in the
variation (\ref{variation}) of the action in the
following suggestive way:
\begin{equation}
\left(\delta {\cal S}^{\it 1}\right)_{\partial {\cal M}}
= - \int^{t''}_{t'}{\rm d}r
\left( J\delta\Lambda + \Lambda T \delta\Phi\right)
- \int_{\bar{\cal T}}{\rm d}t
\left(\bar{E}\delta\bar{N}
+ \bar{N} \bar{{S}} \delta\Phi\right)
+ \left. \alpha
e^{- 2 \Phi} \eta \delta\Phi
\right|^{B''}_{B'}\, ,
\end{equation}
where we have defined the {\em scalars}
\begin{eqnarray}
\bar{E} & \equiv & - \bar{\Pi}_{\bar{N}}
= \alpha e^{-2 \Phi}
\bar{n}^{a} \nabla_{a} \Phi \label{qldensities}
\eqnum{\ref{qldensities}a} \\
J & \equiv & - P_{\Lambda}
= - \alpha e^{-2 \Phi} u ^{a} \nabla_{a} \Phi
\eqnum{\ref{qldensities}b} \\
\bar{S} & \equiv & - \bar{\Pi}_{\Phi}/\bar{N}
= - \alpha e^{-2 \Phi}(\bar{\vartheta}
+ y \bar{n}^{a} \nabla_{a} \Phi)
\eqnum{\ref{qldensities}c}\\
T & \equiv & - P_{\Phi}/\Lambda
= \alpha e^{-2 \Phi}({{\cal K}}
+ y u^{a} \nabla_{a} \Phi) \, .
\eqnum{\ref{qldensities}d}
\addtocounter{equation}{1}
\end{eqnarray}
We interpret $\bar{E}$ as the {\em quasilocal energy}
associated with the $\bar{u}^{a}$ observers at the boundary.
It is convenient to associate a separate energy with
each boundary point of $\bar{\Sigma}$. For instance,
$\bar{E}_{o} \equiv \bar{E}|_{B_{o}}$
is the energy associated with the
outer boundary point. However, properly speaking, the
full quasilocal energy associated with the
$\bar{\Sigma}$ gravitational and dilaton fields is the
sum
\begin{equation}
\left.\bar{E}\right|_{B_{o}} +
\left.\bar{E}\right|_{B_{i}} =
\left. \alpha e^{-2 \Phi}\, \bar{{{\boldmath \mbox{$n$}}}}^{a} \nabla_{a} \Phi
\right|_{B_{o}} - \left. \alpha e^{-2 \Phi}\,
\bar{{{\boldmath \mbox{$n$}}}}^{a} \nabla_{a} \Phi
\right|_{B_{i}}\label{totalEbar}
\end{equation}
of the inner and outer boundary point contributions.
Notice that the full energy above is associated with a
slice $\bar{\Sigma}$ which has $\bar{u}^{a}$ as its timelike
normal {\em at the $\bar{\cal T}$ boundary}.\footnote{Technically,
$\bar{E}_{i} + \bar{E}_{o}$ is the
energy associated with {\em any} slice in the
equivalence class determined by this condition.
The slice $\bar{\Sigma}$ can be ``wiggled" in the
interior as long as its ends remain clamped to the $\bar{\cal T}$
boundary. Similar comments are relevant for the $\Sigma$
energy $E_{i} + E_{o}$.}
Also notice that, when evaluated on solutions to the field equations,
the outer-boundary contribution to the energy $\bar{E}_{o}$,
for example, is minus the rate of change of the classical
action (or Hamilton-Jacobi principal function) with respect
to a unit stretch in $\bar{N} |_{\bar{\cal T}_{o}}$, where
$\bar{N} |_{\bar{\cal T}_{o}}$ controls the lapse of proper time between
neighboring points on $\bar{\cal T}_{o}$.\cite{BY}
The quasilocal energy associated with the $u^{a}$ observers at the
boundary is
\begin{equation}
E \equiv \alpha e^{-2 \Phi}\, n^{a} \nabla_{a} \Phi\, ,
\end{equation}
and sum $E_{i} + E_{o}$ is the full quasilocal energy
for a $\Sigma$ slice. The energy $E$ depends solely
on the Cauchy data of $\Sigma$ in the same way that $\bar{E}$
depends solely on the Cauchy data of $\bar{\Sigma}$.
It proves useful in the next section
to have an energy expression for {\em each} point\footnote{Recall that
the $\Sigma$ points are spheres, at least in the SSGR case. So we
are not really defining a local energy, and certainly not a
local energy {\em density}.} of $\Sigma$. However, we have a sign
ambiguity, because each $\Sigma$ point could be viewed as an inner
or an outer boundary point. For the sake of definiteness, define
\begin{equation}
{{\boldmath \mbox{$E$}}} \equiv \alpha e^{-2\Phi} {{\boldmath \mbox{$n$}}}^{a} \nabla_{a} \Phi =
\alpha e^{-2\Phi} \Phi'/\Lambda\, . \label{sansE}
\end{equation}
Of course, $E = \pm {{\boldmath \mbox{$E$}}}$, depending on whether it is evaluated
at the outer or the inner boundary. For some expressions, like $E^{2}$,
the sign ambiguity cancels, so we can use $E^{2}$ or ${{\boldmath \mbox{$E$}}}^{2}$.
We consider $J$ to be a {\em quasilocal momentum}.
Notice that on-shell $J |_{t''}$, for instance, is minus
the rate of change of the Hamilton-Jacobi principal function
with respect to a unit stretch in $\Lambda |_{t''}$, where
$\Lambda |_{t''}$ controls the lapse of proper distance between
neighboring radial leaves $B$ of $t''$. Such a variation in the
boundary data corresponds to a dilation of $t''$. Again, we
associate a $J$ with each point of a generic $\Sigma$ slice.
At a glance, one sees that $J = - P_{\Lambda}$ depends solely on
$\Sigma$ Cauchy data.
\subsection{Boost relations and invariants}
The Eulerian observers of $\bar{\Sigma}$ at the
$\bar{\cal T}$ boundary coincide with
the natural observers in the boundary. We may define a set
\begin{eqnarray}
\bar{E} & = & \alpha e^{-2 \Phi}
\bar{n}^{a} \nabla_{a} \Phi \label{barobs} \eqnum{\ref{barobs}a} \\
\bar{J} & = & - \alpha
e^{-2 \Phi} \bar{u} ^{a} \nabla_{a} \Phi
\eqnum{\ref{barobs}b} \\
\bar{S} & = & - \alpha e^{-2 \Phi}
( \bar{\vartheta}
+ y \bar{n}^{a} \nabla_{a} \Phi)
\eqnum{\ref{barobs}c} \\
\bar{T} & = & \alpha e^{-2 \Phi}(\bar{{\cal K}}
+ y \bar{u}^{a} \nabla_{a} \Phi)
\eqnum{\ref{barobs}d}
\addtocounter{equation}{1}
\end{eqnarray}
of quasilocal quantities for such observers.
Now the slice $\Sigma$ need not be clamped to the
$\bar{\cal T}$ boundary in our formalism.
Hence, in general $\Sigma$ and $\bar{\Sigma}$ are different
slices which intersect at the same boundary point of interest.
We will refer to a switch of the spatial slice spanning a
particular boundary point as a {\em generalized boost} or
simply a boost. Properly speaking, a generalized boost is a
switch of the equivalence class of spanning slices.
The behavior
\begin{eqnarray}
\bar{E} & = & \gamma E - v \gamma J\label{boost}
\eqnum{\ref{boost}a} \\
\bar{J} & = & \gamma J - v \gamma E
\eqnum{\ref{boost}b}\\
\bar{S} & = & \gamma S - v \gamma T +
\alpha e^{-2 \Phi}\bar{u}^{a} \nabla_{a} \eta
\eqnum{\ref{boost}c} \\
\bar{T} & = & \gamma T - v \gamma S -
\alpha e^{-2 \Phi}\bar{n}^{a} \nabla_{a} \eta
\eqnum{\ref{boost}d}
\addtocounter{equation}{1}
\end{eqnarray}
of the quasilocal quantities under boosts follows
immediately from the boost relations (\ref{boundaryboost})
and the splittings (\ref{splittings}). In (\ref{boost}c,d)
the unbarred version of $\bar{S}$ is
\begin{equation}
S \equiv - \alpha e^{-2 \Phi}\left(- a_{b} n^{b}
+ y n^{b} \nabla_{b} \Phi\right)\, .
\end{equation}
Notice that $E$, $J$, $S$, and $T$ have the same dependence on
$\Sigma$ Cauchy data that $\bar{E}$, $\bar{J}$, $\bar{S}$ and
$\bar{T}$ have on $\bar{\Sigma}$ Cauchy data. (However,
due to the appearance of acceleration terms, both $S$
and $\bar{S}$ do not depend solely on Cauchy data.)
Clearly the expression $- E^{2} + J^{2}$ is invariant
under boosts.
We may multiply it by any function of
the dilaton field or add to it any function of the
dilaton field and retain a boost-invariant expression.
Therefore, it is not completely unnatural to introduce the
invariant
\begin{equation}
M = (2\alpha\lambda)^{-1} e^{y\Phi}
\left[ - E^{2} + J^{2} + (\alpha\lambda)^{2}
\exp\left[-2y\Phi\right]\right]\, . \label{invariant}
\end{equation}
Note that $M$ has units of length for the SSGR case and units of
inverse length for 2dDG.
It turns out that on-shell (on solutions to the field equations)
$M$ is a completely conserved quantity (constant in time and space).
Moreover, for the case of SSGR we find that
$M = M_{\scriptscriptstyle S}$,
where
$M_{\scriptscriptstyle S}$ is Kucha\v{r}'s canonical expression
for the Schwarzschild mass parameter,
\begin{equation}
M_{\scriptscriptstyle S} = \frac{P_{\Lambda}^{2}}{2R}
- \frac{R}{2}\left(\frac{R'}{\Lambda}\right)^{2} + \frac{R}{2}\, .
\end{equation}
From the four-dimensional spacetime perspective, the expression for
$M_{\scriptscriptstyle S}$ corresponds
to several mass definitions in general relativity, when spherical
symmetry is assumed. One is the Hawking mass\cite{Hawking2,SHayward}
\begin{equation}
M_{\scriptscriptstyle H} = \frac{1}{8\pi}\sqrt{\frac{A}{16\pi}}\int_{B}
{\rm d}^{2}x\sqrt{\sigma}\left[ - {\textstyle \frac{1}{2}}( k^{2}
- \ell^{2}) +{}^{2}\! R\right]\, ,
\end{equation}
where $^{2}\! R$ is the intrinsic scalar curvature of the two-surface $B$
(in our case a round sphere), and $k$ is the trace of the intrinsic
curvature of $B$ as embedded in a spanning (three-dimensional) $\Sigma$
slice. Likewise, $\ell$ is the trace of the extrinsic curvature
of $B$ as embedded in the timelike three-surface generated by the
integral curves of the $\Sigma$ normal, denoted
by $u^{\mu}$ for four-dimensions.
The boost-invariant
combination ${\textstyle \frac{1}{2}} ( k^{2} - \ell^{2})$ is more
often written as a product of the expansions associated with the
ingoing and outgoing null normals to $B$, and the factor $A$, the
area of $B$, ensures that the whole expression has units of
energy. For Schwarzschild $M_{\scriptscriptstyle H}$ equals
the mass parameter of the solution even for finite two-spheres.
For general closed two-surfaces in general spacetimes,
the Hawking expression can be ``built"
as a combination of ``quasilocal boost invariants.''\cite{BYL,Lau4}
For the 2dDG case set $M_{\scriptscriptstyle W} = 2M/\alpha$.
By expressing $M_{\scriptscriptstyle W}$ in covariant form,
\begin{equation}
M_{\scriptscriptstyle W} = \lambda^{-1} e^{-2\Phi}
\left( \lambda^{2} - g^{ab} \nabla_{a}\Phi \nabla_{b}\Phi \right)\, ,
\end{equation}
we see that it is the ``local mass" of Tada and Uehara.\cite{Tada}
Such a quantity was also considered by Frolov in Ref.\cite{Frolov}.
With an argument originally given by KVK for the Schwarzschild case,
the appendix shows that $M_{\scriptscriptstyle W}$ is the canonical expression
for the
mass parameter of the Witten black hole.
The appendix also shows that the ADM energy \cite{ADM} at spatial infinity
(associated with the preferred static-time
slices) of the Witten black hole is the on-shell value of $M$. This is
the reason we would prefer to set $\alpha = 2$.
\section{Canonical theory}
This section is devoted to the canonical form of the theory. We first
sketch the Legendre transformation which yields the canonical form of the
action. We then vary the canonical action, paying strict attention to all
boundary terms. Finally, we consider the canonical transformation of
KVK and write down a new-canonical-variable version of the action
(\ref{baseaction}) which is particularly amenable to canonical
reduction.
\subsection{Form of the canonical action}
Passage to the canonical form of the action (\ref{baseaction})
demands that we write the
action in $1+1$ form as a preliminary step. This is easily done with
three ingredients.
The first is the splitting result (\ref{splittings}b) for
the $\bar{\cal T}$ boundary extrinsic curvature
$\bar{\vartheta}$. The second
is the identity
\begin{equation}
2y g^{ab}\nabla_{a}\Phi \nabla_{b}\Phi =
2y\left[ - \left(\frac{\dot{\Phi}}{N}\right)^{2}
+ 2 N^{r} \left(\frac{\dot{\Phi}\Phi'}{N^{2}}\right)
+ \left(\frac{\Phi'}{\Lambda\gamma}\right)^{2}\right]\, ,
\end{equation}
where $\gamma$ is the relativistic factor of the preliminary section.
The third and final ingredient is the realization that for our two
dimensional spacetime the Ricci scalar is a pure divergence,
\begin{equation}
{\cal R} = - 2 \nabla_{a}\left( {{\cal K}} u^{a}
+ a_{b} n^{b} n^{a}\right)\, .
\end{equation}
Using these three ingredients, one can quickly cast (\ref{baseaction})
into the form
\begin{equation}
{\cal S}^{\it 1} = {\textstyle \frac{1}{4}}\,
\alpha\!\int_{\cal M} {\rm d}^{2}x
N\Lambda e^{- 2\Phi}X
+ \int_{\bar{\cal T}} {\rm d}t\, \alpha e^{-2\Phi}
\eta \dot{\Phi}\,\, , \label{3+1action}
\end{equation}
where $X$ is the expression
\begin{eqnarray}
X & = & - \frac{4}{N\Lambda}\left[\Lambda
{{\cal K}}\left(\dot{\Phi} - N^{r}\Phi'\right)
+ \frac{N'\Phi'}{\Lambda}\right] \nonumber \\
& & +
2y\left[ - \left(\frac{\dot{\Phi}}{N}\right)^{2}
+ 2 N^{r} \left(\frac{\dot{\Phi}
\Phi'}{N^{2}}\right) +
\left(\frac{\Phi'}{\Lambda\gamma}\right)^{2}\right]
+ 2\lambda^{2} y \exp\left[(2 - y)2\Phi\right]\, .
\end{eqnarray}
Performing the usual calculations with (\ref{3+1action}), one
finds the following canonical action:
\begin{equation}
{\cal S}^{\it 1} =
\int_{\cal M}{\rm d}^{2}x \left( P_{\Lambda}\dot{\Lambda} +
P_{\Phi}\dot{\Phi} - N {\cal H} - N^{r} {\cal H}_{r}\right) +
\int_{\bar{\cal T}} {\rm d}t \left( \alpha e^{-2\Phi} \eta \dot{\Phi}
- \bar{N} \bar{E}\right)\, , \label{Hamilton}
\end{equation}
where here $\bar{E}$ is short-hand for $\gamma E - v\gamma J$. It is
important to realize that in the canonical picture the equation
$\bar{E} = \alpha e^{- 2 \Phi} \bar{n}^{a} \nabla_{a} \Phi$
is not necessarily valid,
for equality implicitly assumes the {\em canonical equation of motion}
$P_{\Lambda} = \alpha e^{- 2\Phi} u^{a} \nabla_{a} \Phi$.
Respectively, the Hamiltonian
constraint and the momentum constraint have the form
\begin{eqnarray}
{\cal H} & = & \alpha^{-1}
e^{2\Phi}\left[P_{\Phi} P_{\Lambda}
+ {\textstyle \frac{1}{2}}
y \Lambda (P_{\Lambda})^{2}\right] \nonumber \\
& & +\, \alpha e^{-2\Phi} \left[\left(2 - {\textstyle
\frac{1}{2}}y\right)\frac{\Phi'^{2}}{\Lambda}
- \frac{\Phi''}{\Lambda} + \frac{\Phi'\Lambda'}{\Lambda^{2}} -
{\textstyle \frac{1}{2}}\lambda^{2}\Lambda y
\exp\left[(2-y)2\Phi\right] \right] \label{constraints}
\eqnum{\ref{constraints}a}\\
{\cal H}_{r} & = & P_{\Phi} \Phi' - \Lambda P_{\Lambda}' \, .
\eqnum{\ref{constraints}b}
\addtocounter{equation}{1}
\end{eqnarray}
Notice that it is $P_{\Lambda}$ which appears differentiated in
the momentum constraint ${\cal H}_{r}$. This is to be expected,
as $\Lambda$ is a scalar density.
\subsection{Variation of the canonical action}
Straightforward manipulations establish that
the variation of the canonical
action is
\begin{eqnarray}
\delta {\cal S}^{\it 1} & = &
\left({\rm terms\,\,which\,\,enforce\,\,
the\,\,constraints\,\,and\,\,give\,\,the}\right.
\nonumber \\
& & \left. {\rm \,\,{\em Hamiltonian}\,\,
equations\,\,of\,\,motion}\right)
+ \int^{t''}_{t'}{\rm d}r \left(P_{\Lambda}\delta\Lambda +
P_{\Phi}\delta\Phi\right) \nonumber \\
& & - \int_{\bar{\cal T}}{\rm d}t
\left[\bar{E}\delta\bar{N} + \bar{N}\bar{S}\delta\Phi
- \left(\bar{N}\bar{J} + \alpha
e^{-2\Phi}\dot{\Phi}\right)\delta\eta\right] +
\left. \alpha e^{-2\Phi} \eta
\delta\Phi\right|^{B''}_{B'}\, , \label{cvariation}
\end{eqnarray}
where here $\bar{E}$, $\bar{J}$, and $\bar{S}$ are
short-hand for the expressions
\begin{eqnarray}
\bar{E} & = & \gamma E - v\gamma J \label{shorthand}
\eqnum{\ref{shorthand}a} \\
\bar{J} & = & \gamma J - v\gamma E
\eqnum{\ref{shorthand}b}\\
\bar{S} & = & \gamma S - v\gamma T + \alpha
e^{-2\Phi} \bar{u}^{a} \nabla_{a} \eta\, .
\eqnum{\ref{shorthand}c}
\end{eqnarray}
As mentioned, one should be careful, for while
$E$, $J$, $S$, and $T$ are built from the
canonical variables $(\Lambda, P_{\Lambda} ;
\Phi, P_{\Phi})$ in the same
way as before (but $S$ does not depend solely on the
canonical variables),
{\em now the momenta $P_{\Lambda}$ and $P_{\Phi}$ need not
have the forms given in} (\ref{momenta})
(which are canonical equations
of motion). The term which appears multiplied
by the variation
$\delta\eta$ vanishes when the canonical
equation of motion
for $P_{\Lambda}$ holds. Therefore,
$\eta$ is not a quantity which is held fixed in
our variational principle.
\subsection{Canonical Transformation}
In this subsection we perform Kucha\v{r}'s canonical transformation
on the phase-space pairs $(\Lambda, P_{\Lambda} ; \Phi, P_{\Phi})$.
In order to grasp the underlying hyperbolic geometry of this canonical
transformation, we first need to collect a few results and
observations.
Consider a black-hole solution which extremizes the
action (\ref{baseaction}) (either a Schwarzschild black hole or a
Witten black hole, depending on whether $y$ is $1$ or $2$). Associated
with this solution, there is a preferred family of static-time
slices, the collection of constant-Killing-time level surfaces.
For the Schwarzschild-black-hole case let $T(t,r)$ denote the Killing time,
and for the Witten-black-hole case let $\tau(t,r)$ denote the Killing
time. Now, given a particular $\cal M$ point $B$, we may interpret it as a
boundary point of the static-time slice $\tilde{\Sigma}$ which
contains it. Our construction defines the {\em rest frame}
$(\tilde{u}^{a} , \tilde{n}^{a})$ at $B$, where $\tilde{u}^{a}$
is the normal of
$\tilde{\Sigma}$ as embedded in $\cal M$ and $\tilde{n}^{a}$ is the
normal of $B$ as embedded in $\tilde{\Sigma}$. If $B$ is also
considered to be a point of the $\bar{\cal T}$ boundary, then in general
$\tilde{\Sigma}$ does not define the same frame at $B$ as
the slice $\Sigma$ or the slice $\bar{\Sigma}$
considered before. We know how to compute the energy
$\tilde{E} $ and momentum $\tilde{J}$ of the bounded
static-time slice $\tilde{\Sigma}$,
\begin{eqnarray}
\tilde{E} & = & \alpha
e^{-2\Phi}\tilde{n}^{a} \nabla_{a} \Phi
\nonumber \\
\tilde{J} & = & - \alpha
e^{-2\Phi}\tilde{u}^{a} \nabla_{a} \Phi\, .
\end{eqnarray}
Clearly $\tilde{E}$ and $\tilde{J}$ depend
on the Cauchy data of $\tilde{\Sigma}$ in the same way that $E$ and
$J$ depend on the Cauchy data of $\Sigma$.
Now, it is a fact that $\tilde{J} = 0$, which is why we refer
to $(\tilde{u}^{a}, \tilde{n}^{a})$ as the rest frame at $B$. The
existence of the rest frame at $B$ leads
to our key observation:
{\em at $B$ the parameter
$\varphi$ associated with the boost from the
frame $(u^{a} , n^{a})$ defined
by $\Sigma$ to the rest frame
$(\tilde{u}^{a} , \tilde{n}^{a})$ is
determined by the canonical
variables of $\Sigma$}.\footnote{Which need to be known
only in a tiny neighborhood of $B$.} Indeed, with
$w \equiv J/E$ the boost from the $\Sigma$
frame to the rest frame is parameterized by
\begin{equation}
\varphi \equiv {\textstyle \frac{1}{2}}
\log\left|\frac{1 + w}{1 - w}\right|
= {\textstyle \frac{1}{2}}
\log\left|\frac{E + J}{E - J}\right|\, .
\label{varphiboost}
\end{equation}
Notice that $\tilde{u}^{a} = \psi u^{a} + w \psi n^{a}$ with
$\psi = (1 - w^{2})^{-1/2}$. We know the expression
(\ref{varphiboost}) is correct, because our general theory
of boosted quasilocal energy-momentum implies that
$E = \psi \tilde{E}$ and $J = w \psi \tilde{E}$. To get these
last relations, assume that the $\bar{\cal T}$ boundary is
generated by the $\tilde{u}^{a}$ Eulerian history of $B$
points (identify ``tildes'' with ``bars'' and $w$
with $v$)
and use (\ref{boost}a,b).
At this stage we have a
sign ambiguity in our expressions, since we did not say
whether $B$ is an inner or an outer boundary point.
For the sake of definiteness in what follows,
we often want to assume that $B$ is taken as an outer
boundary point and make this distinction notationally.
Therefore, we use ${{\boldmath \mbox{$E$}}}$,
defined before in (\ref{sansE}), and
also ${{\boldmath \mbox{$w$}}} = J/{{\boldmath \mbox{$E$}}}$
which defines ${{\boldmath \mbox{$\varphi$}}} \equiv
\frac{1}{2}\log|(1 + {{\boldmath \mbox{$w$}}})/(1 - {{\boldmath \mbox{$w$}}})|$.
Notice that the sign ambiguities
cancel in the expression $E \varphi
= {{\boldmath \mbox{$E$}}} {{\boldmath \mbox{$\varphi$}}}$. One should be careful
when dealing with
${{\boldmath \mbox{$w$}}}$, $\psi$, and ${{\boldmath \mbox{$\varphi$}}}$.
Since for a classical solution the
dilaton is a ``bad" radial coordinate in the
sense that $\Phi' < 0$, our
{\em unreferenced} energy ${{\boldmath \mbox{$E$}}}$ is negative. The
(in our case positive) expression for the
quasilocal energy with the
flat-space reference
contribution is considered in the appendix and
at length in Ref. \cite{BY}.
Therefore, since square roots are by
convention positive, $\psi =
- {{\boldmath \mbox{$E$}}} /\sqrt{E^{2} - J^{2}}$.
With our canonical expression for ${{\boldmath \mbox{$\varphi$}}}$ we
straightforwardly write down a new set of constraints which
generate proper unit displacements with respect to the static-time
slices,
\begin{eqnarray}
{\sf H} & = & \psi {\cal H}
+ {{\boldmath \mbox{$w$}}} \psi ({\cal H}_{r}/\Lambda)
\label{parthenon} \eqnum{\ref{parthenon}a} \\
{\sf H}_{\vdash} & = & \psi ({\cal H}_{r}/\Lambda) +
{{\boldmath \mbox{$w$}}} \psi {\cal H}\, . \eqnum{\ref{parthenon}b}
\addtocounter{equation}{1}
\end{eqnarray}
Since these new constraints depend only on the canonical
variables of $\Sigma$, we can consider them off the
constraint surface in phase space.
However, when computing a Poisson bracket $\{G, {\sf H}\}$,
where $G$ is a functional of the canonical variables, one finds
that all of the ``dangerous'' brackets $\{G, {{\boldmath \mbox{$w$}}}\}$
which arise come multiplied by either by factor of $\cal H$ or
${\cal H}_{r}$.
Therefore, on-shell $\sf H$ generates proper unit displacement of the
$\Sigma$ slice in the $\tilde{u}^{a}$ direction.
Likewise, ${\sf H}_{\vdash}$ generates proper unit displacement of the
$\Sigma$ slice in the $\tilde{\boldmath \mbox{$n$}}^{a}$ direction.
As we show below, these generators are closely related to the
momenta $P_{T}$ and $P_{\sf R}$ considered in KVK.
Define the new canonical variables in terms of the old ones as
follows:
\begin{eqnarray}
M & = & {\textstyle \frac{1}{2}}\,\alpha\lambda
e^{- y\Phi} (1 - F)
\label{canonical} \eqnum{\ref{canonical}a} \\
P_{M} & = & (\alpha\lambda)^{-1} e^{y\Phi} F^{-1}
\Lambda P_{\Lambda} \eqnum{\ref{canonical}b} \\
& & \nonumber \\
\Psi & = & \Phi \eqnum{\ref{canonical}c}\\
P_{\Psi} & = & P_{\Phi} + {\textstyle \frac{1}{2}}
y (1 + F^{-1})\Lambda
P_{\Lambda} + \alpha e^{-2\Phi}{{\boldmath \mbox{$\varphi$}}}'\, ,
\eqnum{\ref{canonical}d}
\addtocounter{equation}{1}
\end{eqnarray}
where $M$ is the boost invariant (\ref{invariant}) and in terms
of the old variables $F$ and ${{\boldmath \mbox{$\varphi$}}}$ are short-hand for
\begin{eqnarray}
F & = & (\alpha\lambda)^{-2} \exp\left[(y - 2)2 \Phi\right]
\left[ (\alpha\Phi'/\Lambda)^{2}
- (e^{2\Phi} P_{\Lambda})^{2}\right]
\label{oldFandphi} \eqnum{\ref{oldFandphi}a} \\
& & \nonumber \\
{{\boldmath \mbox{$\varphi$}}} & = & {\textstyle \frac{1}{2}}\log
\left|\frac{\alpha e^{-2\Phi}\Phi' -
\Lambda P_{\Lambda}}{\alpha e^{-2\Phi}\Phi' +
\Lambda P_{\Lambda}}\right|\, .
\eqnum{\ref{oldFandphi}b}
\addtocounter{equation}{1}
\end{eqnarray}
Notice that ${{\boldmath \mbox{$\varphi$}}} = \frac{1}{2}\log|F_{-}/F_{+}|$, where
\begin{equation}
F_{\pm} = (\alpha\lambda)^{-1} e^{y\Phi}({{\boldmath \mbox{$E$}}} \mp J)\, .
\end{equation}
Evidently then, another expression for $F$ of key importance is
\begin{equation}
F = F_{+} F_{-} = (\alpha\lambda)^{-2} e^{y2\Phi}
\left( E^{2} - J^{2}\right)\, .
\label{Fexpression}
\end{equation}
For the SSGR case
$e^{-\Psi} = \lambda {\sf R}$ and $P_{\Psi} =
- {\sf R} P_{\sf R}$, in the notation of KVK and LW. Also for this case,
one finds that $P_{M} = - T'$,
because as shown in KVK the canonical expression for (minus) the radial
derivative of the Killing time is given by $- T' = (RF)^{-1}\Lambda
P_{\Lambda}$. The situation is the same for 2dDG, as in this
case $P_{M} = -\tau'$. We show in the appendix that the canonical
expression for (minus) the radial derivative of the Witten-black-hole
Killing time is $-\tau' = (\alpha\lambda)^{-1} e^{2\Phi}
F^{-1}\Lambda P_{\Lambda}$. One may prove that the transformation $(\Lambda,
P_{\Lambda} ;\Phi, P_{\Phi}) \rightarrow (M, P_{M} ; \Psi, P_{\Psi})$ is
canonical
for our boundary conditions by verifying the identity
\begin{equation}
P_{\Lambda}
\delta \Lambda + P_{\Phi} \delta \Phi
- P_{M} \delta M - P_{\Psi} \delta \Psi
= \delta\left[ \Lambda (- J + E\varphi)\right]
- \left(\alpha e^{-2\Phi} \delta \Phi {{\boldmath \mbox{$\varphi$}}}\right)'\, ,
\label{generating}
\end{equation}
which upon integration over $r$ shows that the difference between the
old Liouville form and the new Liouville is an exact form. Hence, the
transformation is canonical.
Recall that for simplicity we wish to restrict our attention to
the right-static region of the relevant Kruskal diagram. Now,
in fact, for both SSGR and 2dDG $F$ is the canonical
expression for $\tilde{N}^{2}$, where $\tilde{N}$ is the lapse
function associated with the static-time slices. On-shell, the
event horizon of a particular black-hole solution is the locus of
points determined
by $F = 0$. We may ensure that we are working exclusively in a static
region of the Kruskal diagram by choosing our boundary conditions
appropriately and by excluding solutions for which $F = 0$
somewhere on $\cal M$. Where $F$ is nonzero, the above transformation
may be inverted,
\begin{eqnarray}
\Lambda & = & \left[\lambda^{-2} F^{-1}
(\Psi')^{2}\exp\left[(y - 2)2 \Psi\right] - F (P_{M})^{2}\right]^{1/2}
\label{invertedcanonical} \eqnum{\ref{invertedcanonical}a} \\
& & \nonumber \\
P_{\Lambda} & = & \frac{\alpha\lambda e^{- y \Psi}
F P_{M}}{\left[\lambda^{-2} F^{-1}
(\Psi')^{2}\exp\left[(y - 2)2 \Psi\right]
- F (P_{M})^{2}\right]^{1/2}}
\eqnum{\ref{invertedcanonical}b}\\
& & \nonumber \\
\Phi & = & \Psi \eqnum{\ref{invertedcanonical}c} \\
P_{\Phi} & = & P_{\Psi} - {\textstyle \frac{1}{2}} \alpha\lambda
y e^{- y\Psi} P_{M}(1 + F) - \alpha
e^{-2\Psi}{{\boldmath \mbox{$\varphi$}}}'\, ,
\eqnum{\ref{invertedcanonical}d}
\addtocounter{equation}{1}
\end{eqnarray}
where in terms of the new variables
$F$ and ${{\boldmath \mbox{$\varphi$}}}$ are short-hand for
\begin{eqnarray}
F & = & 1 - 2 (\alpha\lambda)^{-1} e^{y\Psi}M
\label{newFandphi} \eqnum{\ref{newFandphi}a} \\
& & \nonumber \\
{{\boldmath \mbox{$\varphi$}}} & = & {\textstyle \frac{1}{2}}\log
\left|\frac{\Psi' - \lambda \exp\left[(2 - y)\Psi\right]
F P_{M}}{\Psi' + \lambda \exp\left[(2 - y)\Psi\right]
F P_{M}}\right|\, .\eqnum{\ref{newFandphi}b}
\addtocounter{equation}{1}
\end{eqnarray}
As mentioned in the introduction,
KVK considers the canonical transformation and its inverse in all
regions of the Kruskal diagram for SSGR. Moreover, this reference
provides a detailed treatment of the (singular) behavior of the
transformation at the horizon. We expect that a similar treatment
with essentially the same results can be carried out for the 2dDG
case.
As shown in KVK for the SSGR case,
the payoff obtained by using the new variables comes when considering
the constraints (\ref{constraints}).
Since on solutions to the constraints $M$ is a
constant, one knows that $M'$ must be a sum of constraints. Indeed,
direct calculation establishes that
\begin{equation}
M' = - F^{1/2}{\sf H}\, .
\end{equation}
Since $F^{1/2}$ is the lapse $\tilde{N}$, we see that $- M'$ is the
generator of Killing-time evolution. For this reason, KVK uses the
notation $P_{T} = - M'$ (and we would likewise set $P_{\tau} =
- M'$ for the 2dDG case). Furthermore,
\begin{equation}
P_{\Psi} = - \lambda^{-1} \exp\left[(y - 2)\Phi\right]
F^{-1/2}{\sf H}_{\vdash}\,
\end{equation}
is also a sum of constraints and so weakly vanishes. KVK and LW set
$P_{\sf R} = F^{-1/2}{\sf H}_{\vdash}$. As described in KVK, the
momenta $P_{T}$ and $P_{\sf R}$ respectively generate coordinate-scaled
displacements of the $\Sigma$ slice along lines of constant
${\sf R} = R$ and $T$ (or $\tau$ for 2dDG) in the Kruskal diagram.
Kucha\v{r} obtains these interpretations for $P_{T}$ and $P_{\sf R}$
via a ``parameterization at infinity'' and canonical transformations
subsequent to the main one (\ref{canonical}). We have obtained the
exact same interpretations via a different route.
The new canonical
variables are related to the new constraints (\ref{parthenon})
in a very simply way: ${\sf H} = - F^{-1/2} M'$ and
${\sf H}_{\vdash} = - \lambda \exp\left[(2 - y)\Psi\right]
F^{1/2} P_{\Psi}$. Using these relations, we may write the old constraints in
terms of the new variables as
\begin{eqnarray}
{\cal H} & = & \psi {\sf H} - {{\boldmath \mbox{$w$}}}\psi {\sf H}_{\vdash}
\nonumber \\
{\cal H}_{r} & = & \Lambda\left( \psi {\sf H}_{\vdash} - {{\boldmath \mbox{$w$}}} \psi
{\sf H} \right)\, ,
\end{eqnarray}
where here $\Lambda$ is given in (\ref{invertedcanonical}a)
and ${{\boldmath \mbox{$w$}}}$ must be expressed in terms of the new variables.
With the list (\ref{invertedcanonical}), it is not hard to show
that
\begin{equation}
{{\boldmath \mbox{$w$}}} = - \lambda
(\Psi')^{-1}\exp\left[(2 - y)\Psi\right] F P_{M}\, .
\end{equation}
It is now straightforward, if tedious, to
express the old constraints in terms of the new variables,
\begin{eqnarray}
{\cal H} & = & \frac{\lambda^{-1} F^{-1} \Psi'
\exp\left[(y - 2)\Psi\right] M'
+ \lambda F \exp\left[(2 - y)\Psi\right]
P_{\Psi} P_{M}}{\left[ \lambda^{-2}
F^{-1} (\Psi')^{2} \exp\left[(y - 2)2\Psi\right]
- F (P_{M})^{2}\right]^{1/2}}
\nonumber \\
& & \\
{\cal H}_{r} & = & P_{M} M' + P_{\Psi} \Psi'\, .
\nonumber
\end{eqnarray}
As noted in KVK, in terms of the new variables it is relatively simple
to show that the Poisson bracket of $M$ with either $\cal H$ or
${\cal H}_{r}$ vanishes weakly.
\subsection{Canonical reduction}
The goal of this subsection is to use the new canonical variables
to find a reduced action principle which -in a certain sense-
corresponds to the canonical action (\ref{Hamilton}). However,
we will be adding boundary terms to (\ref{Hamilton}) before the
reduction is made, so we should clearly state what we have in mind
to begin with and why.
Several aspects of the path-integral
formulation of gravitational thermodynamics motivate fixation of
${\bar N}$ and $\Phi$ on the $\bar{\cal T}$ boundary as
the features of central importance which need to be preserved as we
modify the original action (\ref{Hamilton}). In path-integral
expressions for gravitational partition functions, the sum over
histories includes only spacetimes for which the initial and
final slices are identified. In this scenario the gauge-invariant
information of $\bar{N}$ (the lapse of proper time between the identified
initial slice $t'$ and final slice $t''$) is essentially
the inverse temperature,
which is fixed in the canonical ensemble.\cite{BMYW}
Regarding the fixation of the dilaton $\Phi$ on the
$\bar{\cal T}$ boundary, from a four-dimensional
perspective this feature allows the area of the boundary of the
system to be fixed as a boundary condition.
In what follows we modify our original canonical action
(\ref{Hamilton}) in several steps. Each step is -at least
heuristically- justified. The result will be an action
${\cal S}^{\it 1}_{\ddagger}$ (for the SSGR case this action is closely
related to one considered by LW) which is particularly amenable to
canonical reduction via the new variables. Moreover, as we will
explicitly demonstrate, the new action ${\cal S}^{\it 1}_{\ddagger}$
retains fixation of $\bar{N}$ and $\Phi$ on the $\bar{\cal T}$ boundary,
important for the above mentioned reasons, as features of its
associated variational principle. Our analysis provides some
conceptual justification for several technical steps taken in LW.
Let us go through the steps of modifying ${\cal S}^{\it 1}$. We know from
(\ref{generating}) that addition of the boundary terms
\begin{equation}
\left. - \omega[\Lambda, P_{\Lambda}, \Phi]\right|^{t''}_{t'}
= \int^{t''}_{t'} {\rm d}r \Lambda
\left(J - E\varphi\right)
\end{equation}
to the canonical action (\ref{Hamilton}) gives the new action
\begin{equation}
{\cal S}_{\dagger}^{\it 1} \equiv \int_{\cal M}{\rm d}^{2}x
\left( P_{M}\dot{M} +
P_{\Psi}\dot{\Psi} - N {\cal H}
- N^{r} {\cal H}_{r}\right) +
\int_{\bar{\cal T}} {\rm d}t
\left( \alpha e^{-2\Psi} \eta \dot{\Psi}
- \bar{N} \bar{E}\right)\, , \label{newHamilton}
\end{equation}
where here we consider all the quantities as expressed in terms
of the old variables. The vanishing of the
original set of constraints is equivalent to the vanishing of
$M'$ and $P_{\Psi}$. To take advantage of this fact,
re-express the constraint terms in the action as
$N {\cal H} + N^{r} {\cal H}_{r} = N^{M} M' +
N^{\Psi} P_{\Psi}$, where the new Lagrange multipliers are
\begin{eqnarray}
N^{M} & = & - F^{-1/2}\left(\psi N -
{{\boldmath \mbox{$w$}}} \psi \Lambda N^{r}\right) \label{newlapseshift}
\eqnum{\ref{newlapseshift}a} \\
N^{\Psi} & = & - \lambda \exp\left[(2 - y)\Psi\right]F^{1/2}
\left(\psi
\Lambda N^{r} - {{\boldmath \mbox{$w$}}} \psi N\right)\, . \eqnum{\ref{newlapseshift}b}
\addtocounter{equation}{1}
\end{eqnarray}
At this stage the new Lagrange multipliers still depend on the
old multipliers and the old canonical variables. In particular,
note that in terms of the old variables
\begin{equation}
N^{\Psi} = - \alpha^{-1} e^{2\Phi} \bar{N}
\left( \gamma J - v \gamma {{\boldmath \mbox{$E$}}}\right)\, ,
\end{equation}
where we have used the expression (\ref{Fexpression}) and the fact that
${{\boldmath \mbox{$w$}}} = J/{{\boldmath \mbox{$E$}}}$. With the canonical equation of motion for
$P_{\Lambda}$, one can show that
$N^{\Psi} = - \alpha^{-1} e^{2\Phi} \bar{N} \bar{J} = \dot{\Phi}$.
Had we merely passed to the new canonical variables, without
redefining the Lagrange multipliers,
the variational principle associated with (\ref{newHamilton}) would have
featured fixation of $\bar{N}$ and $\Phi$ on the $\bar{\cal T}$
boundary automatically.
However, with the Lagrange multipliers
redefined (and in a way which absorbs some
of the canonical variables) all
bets are off. We must cleverly choose the
$\bar{\cal T}$ boundary term
appropriate for the new-variable version of the action.
In passing from the old constraints to $M'$ and $P_{\Psi}$, we are
effectively performing the Lorentz boost from
the frame $(u^{a}, {{\boldmath \mbox{$n$}}}^{a})$ to the rest
frame $(\tilde{u}^{a}, \tilde{{\boldmath \mbox{$n$}}}^{a})$ at each
point on $\Sigma$. A point-wise boost\footnote{Or, depending on the
viewpoint, a sphere-wise boost.}
has been performed on the old constraints and Lagrange multipliers.
However, we have not included the boundary term in
the boost. It seems that the correct way to incorporate the effect of
the boost into the boundary term is to reference the existing boost parameter
$\eta$ against the parameter $\varphi$ associated with the boost to the rest
frame. This is achieved by adding a $\bar{\cal T}$ boundary term to the action
${\cal S}^{\it 1}_{\dagger}$, with the result
\begin{equation}
{\cal S}_{\ddagger}^{\it 1}
\equiv \int_{\cal M}{\rm d}^{2}x \left( P_{M}\dot{M} +
P_{\Psi}\dot{\Psi} - N^{M} M' - N^{\Psi} P_{\Psi}\right) +
\int_{\bar{\cal T}} {\rm d}t \left( \alpha e^{-2\Psi} \rho
\dot{\Psi} - \bar{N} \bar{E}\right)\, , \label{doubledagaction}
\end{equation}
On-shell, the boost parameter in the new action
\begin{equation}
\rho \equiv \eta - \varphi =
- {\textstyle \frac{1}{2}}\log\left|\frac{\bar{E} + \bar{J}}{\bar{E} -
\bar{J}}\right| \label{rhoparameter}
\end{equation}
is associated with the local boost between the rest frame
$(\tilde{u}^{a}, \tilde{n}^{a})$ and the boundary frame
$(\bar{u}^{a} , \bar{n}^{a})$.
At this stage, the terms $\bar{E}$ and
$\bar{J}$ are still short-hand expressions for $\gamma E - v \gamma J$
and $\gamma J - v \gamma E$, respectively.
We now wish to express all terms in the action
${\cal S}_{\ddagger}^{\it 1}$ solely in terms
of the new variables and freely variable $N^{M}$ and $N^{\Psi}$.
Using the expressions (\ref{invertedcanonical}), one can easily
express the quasilocal energy and momentum in terms of the new
variables,
\begin{eqnarray}
E & = & \frac{\epsilon \alpha e^{-2\Psi} \Psi'}
{\left[\lambda^{-2} F^{-1}(\Psi')^{2}\exp\left[(y - 2)2 \Psi\right] - F
(P_{M})^{2}\right]^{1/2}}
\label{newEandJ} \eqnum{\ref{newEandJ}a} \\
& & \nonumber \\
J & = & \frac{- \alpha\lambda e^{- y\Psi} F P_{M}}{\left[
\lambda^{-2} F^{-1} (\Psi')^{2}\exp\left[(y - 2)2 \Psi\right] -
F (P_{M})^{2}\right]^{1/2}}\, , \eqnum{\ref{newEandJ}b}
\addtocounter{equation}{1}
\end{eqnarray}
where the factor $\epsilon \equiv (\bar{n}_{a} \bar{{{\boldmath \mbox{$n$}}}}^{a})$
takes care of the appropriate
sign on each of the boundary elements. Moreover, we must now regard
$N$ and $N^{r}$ (which along
with $\Lambda$ are hidden in the $v$'s and $\gamma$'s which are in turn
hidden in $\rho$) as depending
on the new variables. It is not difficult to invert the relations
(\ref{newlapseshift}) to get the needed expressions. Also, a short
calculation shows that the boundary lapse $\bar{N} = N/\gamma$ has
a fairly nice expression in terms of the new variables,
\begin{equation}
\bar{N}^{2} = F \left(N^{M}\right)^{2} - \lambda^{-2}
\exp\left[(y - 2)2 \Psi\right] F^{-1}\left(N^{\Psi}\right)^{2}\, .
\label{newNbar}
\end{equation}
However, the situation at hand remains
quite problematic. We would
like to use the fact that $M' = 0$ to
define a radially independent
${\bf m}(t) \equiv M(t)$ with conjugate momentum
${\bf p}(t) \equiv \int_{\Sigma} {\rm d}r P_{M}(t,r)$.
Since we have seen that $P_{M}$ is minus the radial
derivative of the Killing time, the momentum $\bf p$
would then be the difference between
the Killing-time values for the
boundary points $B_{i}$ and $B_{o}$.
Indeed, our canonical-reduction goal
is to insert the solutions of the constraints
into the action to find a reduced action which is
expressed in terms of the pair $({\bf m}, {\bf p})$.
However, $P_{M}$ appears in the $\bar{\cal T}$ boundary
terms explicitly. Therefore, even if we perform the $r$
integration in the action to
define $\bf p$, the boundary
terms still contain factors of
$P_{M}$.\footnote{The reader might suspect
that this problem was introduced when we performed
the heuristically
justified reference
$\eta \rightarrow \rho = (\eta - \varphi)$ to
get the action
${\cal S}_{\ddagger}^{\it 1}$.
But the boundary term in the action
${\cal S}_{\dagger}^{\it 1}$
suffers from the same problem.
Indeed, $\eta$ is built from $v$
which is in turn built from $\Lambda$,
and hence in term of the new
variables $\eta$ depends on $P_{M}$.}
A solution to the problem at hand is to make an
appeal to the
equations of motion. For the moment,
let us go back to considering the action
${\cal S}_{\ddagger}^{\it 1}$ as
depending on the old variables. Notice that
{\em using the canonical equation of motion for}
$P_{\Lambda}$, one may write
\begin{eqnarray}
\bar{J} & = & - \alpha e^{-2\Phi}
(\dot{\Phi}/\bar{N}) \label{anotherEandJ}
\eqnum{\ref{anotherEandJ}a} \\
& & \nonumber \\
\bar{E} & = &
- \epsilon \alpha e^{-2\Phi}\sqrt{\lambda^{2}
\exp\left[(2 - y)2\Phi\right] F
+ (\dot{\Phi}/\bar{N})^{2}}\, ,
\eqnum{\ref{anotherEandJ}b}
\addtocounter{equation}{1}
\end{eqnarray}
where we have appealed to the form (\ref{Fexpression}) for $F$
and again used $\epsilon$ to take care of
the appropriate sign for each boundary
element. Again, note that by convention we take the positive
square root in (\ref{anotherEandJ}b) and
$\bar{E}$ is negative.
It is trivial to write these new expressions
for $\bar{E}$ and $\bar{J}$
in terms of the new variables. Fortunately,
these expressions have no
dependence on $P_{M}$. Using these expressions
instead of those in
(\ref{newEandJ}), we find that in terms of the new variables
our action is again (\ref{doubledagaction});
but now with (i) $\bar{N}\bar{E} = - \epsilon
\alpha e^{-2\Psi}\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right] \bar{N}^{2} F
+ (\dot{\Psi})^{2}}$, (ii) $\bar{N}^{2}$ short-hand
for the expression (\ref{newNbar}), (iii)
$F$ short-hand for expression (\ref{newFandphi}a),
and (iv) the parameter specifying the boost between
the boundary and rest frames given by
\begin{equation}
\rho = - {\textstyle \frac{1}{2}}
\log \left|\frac{\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right] \bar{N}^{2} F
+ (\dot{\Psi})^{2}}
+ \epsilon\dot{\Psi}}{\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right]
\bar{N}^{2} F +
(\dot{\Psi})^{2}}
- \epsilon\dot{\Psi}}\right|\, . \label{newrho}
\end{equation}
Now let us verify that the variational principle
associated with (\ref{doubledagaction}) and the just-given list
(i)-(iv) does in fact possess the features
that we demand.
The constraints and
equations of motion are
\begin{eqnarray}
M' & = & 0 \label{newEOM} \eqnum{\ref{newEOM}a} \\
P_{\Psi} & = & 0 \eqnum{\ref{newEOM}b} \\
\dot{M} & = & 0 \eqnum{\ref{newEOM}c} \\
\dot{P}_{M} & = & N^{M} \eqnum{\ref{newEOM}d} \\
\dot{\Psi} & = & N^{\Psi} \eqnum{\ref{newEOM}e} \\
\dot{P}_{\Psi} & = & 0\, . \eqnum{\ref{newEOM}f}
\addtocounter{equation}{1}
\end{eqnarray}
In terms of the old variables, we have already seen that the
equation (\ref{newEOM}e) holds when the canonical equation of
motion for $P_{\Lambda}$ is assumed.
Upon variation of the action, we find the boundary terms
\begin{eqnarray}
\lefteqn{\left(\delta
{\cal S}_{\ddagger}^{\it 1}\right)_{\partial {\cal M}}
=} & & \nonumber \\
& & \int^{t''}_{t'}{\rm d}r
\left(P_{M} \delta M + P_{\Psi}\delta\Psi\right)
+ \int_{\bar{\cal T}}{\rm d}t\left( \bar{\Pi}_{\bar{N}} \delta \bar{N}
+ \bar{\Pi}_{\Psi} \delta \Psi + \bar{\Pi}_{M} \delta M\right) +
\left. \alpha e^{-2 \Psi}\rho
\delta\Psi\right|^{B''}_{B'}\, , \label{deltaddagger}
\end{eqnarray}
where now the $\bar{\cal T}$ boundary momenta are the following:
\begin{eqnarray}
- \bar{E} = \bar{\Pi}_{\bar{N}} & = & \epsilon \alpha
e^{-2\Psi}\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right] F + (\dot{\Psi}/\bar{N})^{2}}
\label{dagmomenta} \eqnum{\ref{dagmomenta}a} \\
\bar{\Pi}_{\Psi} & = & {\textstyle \frac{1}{2}}
\bar{N}\left[ y\bar{E} (1 + F^{-1}) -
2 \alpha e^{-2\Psi} \bar{u}^{a} \nabla_{a}\rho\right]
\eqnum{\ref{dagmomenta}b} \\
\bar{\Pi}_{M} & = & - \epsilon N^{M}
+ (\alpha\lambda)^{-1} e^{y\Psi}
F^{-1} \bar{N}\bar{E}\, . \eqnum{\ref{dagmomenta}c}
\addtocounter{equation}{1}
\end{eqnarray}
Note that $\bar{\Pi}_{\Psi}$ is not the same as the $\bar{\Pi}_{\Phi}$
in the list (\ref{momenta}). Although the $\bar{\Pi}_{\bar{N}}$
found here is not the expression (\ref{momenta}c), it agrees with
this expression on-shell.
Recalling that $\bar{N}^{2}$ stands for (\ref{newNbar}),
one finds that $\bar{\Pi}_{M}$ vanishes when the equation of motion
(\ref{newEOM}e) holds. Therefore, $M$ need not be held fixed on
the $\bar{\cal T}$ boundary in the variational principle associated with
${\cal S}_{\ddagger}^{\it 1}$, as the equations of motion ensure that
the boundary terms with $\bar{\Pi}_{M}$ vanishes for arbitrary
variations $\delta M$ about a classical solution.
The reduced action $I^{\it 1}_{\ddagger}$, expressed in terms
of ${\bf m}$ and ${\bf p}$ defined earlier, is obtained
by solving the constraints and inserting these solutions
back into the action ${\cal S}_{\ddagger}$. From the result
(\ref{deltaddagger}) for the
variation of the action ${\cal S}_{\ddagger}^{\it 1}$, we know that
the reduced action,
\begin{eqnarray}
I_{\ddagger}^{\it 1} = \int^{t''}_{t'}{\rm d}t\, {\bf p} \dot{\bf m}
& + & \int_{\bar{\cal T}}
{\rm d}t\, \alpha e^{-2\Psi}
\left[\epsilon\, \sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right] \bar{N}^{2} F +
(\dot{\Psi})^{2}} \right. \nonumber \\
& & \left.
- {\textstyle \frac{1}{2}}\dot{\Psi}
\log \left|\frac{\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right] \bar{N}^{2} F
+ (\dot{\Psi})^{2}}
+ \epsilon \dot{\Psi}}{\sqrt{\lambda^{2}
\exp\left[(2 - y)2 \Psi\right]
\bar{N}^{2} F +
(\dot{\Psi})^{2}}
- \epsilon\dot{\Psi}}\right|\,\, \right]\, ,
\end{eqnarray}
possesses the variational principle we desire.
In the reduced action $\bar{N}$ is positive and independent, and
$F = 1 - 2 (\alpha\lambda)^{-1} e^{y\Psi} {\bf m}$. Also, in the expression
for $I^{\it 1}_{\ddagger}$ the $t'$ and $t''$ represent integration parameters
rather than spacelike slices as they did before.
\section{The thermodynamical Action}
For both the Schwarzschild and the Witten black hole
we are interested in applying the canonical
action principle to a static exterior region with spatial boundary
({\em including} the bifurcation point)
of the relevant Kruskal diagram. Such an application of the action
principle is the appropriate one
for studying the equilibrium thermodynamics
of black holes.\cite{BMYW} In such a scenario, with the
covariant form of the action there is no inner boundary,
since the bifurcation point is a set of measure zero in the
integration over $\cal M$. Nevertheless, in the canonical picture
the bifurcation point is a boundary point of every spacelike slice,
which implies that the canonical coordinates must obey certain fall-off
conditions as the bifurcation point is approached.
Moreover, in the thermodynamical paradigm, when the initial and final
spacelike slice are identified, one must worry
about {\em regularity conditions} at the bifurcation point which
ensure that the geometry is smooth.\cite{Gibbons,BMYW,micro}
To handle these issues, we will use a technique due to
Brown and York.\cite{BYpathint} The basic idea is to work
with an inner boundary, but with boundary conditions which
effectively seal it.
The main ingredient in this technique is a new action functional,
which differs from (\ref{baseaction}) by boundary terms. The
purpose of this section is to introduce this new action principle
and to study its canonical reduction via the new canonical
variables.
\subsection{Alternative canonical action principle}
Starting with the canonical action (\ref{Hamilton}), we define
the new action
\begin{equation}
{\cal S}^{\it 1}_{*} \equiv {\cal S}^{\it 1}
- {\textstyle \frac{1}{2}}
\int_{\bar{\cal T}_{i}} {\rm d}t\bar{N}\bar{S}
+ {\textstyle \frac{1}{2}} \left. \alpha
e^{-2\Phi}\eta\right|^{B''_{i}}_{B'_{i}}\, .
\label{astaction}
\end{equation}
Note that only inner boundary and inner corner terms
have been added to the original action.
It is easy to show that the canonical form of
${\cal S}^{\it 1}_{*}$ has the boundary terms
\begin{equation}
\left({\cal S}^{\it 1}_{*}\right)_{\partial {\cal M}} =
\int_{\bar{\cal T}_{o}} {\rm d}t
\left[ \alpha e^{-2\Phi} \eta \dot{\Phi}
- \bar{N} \bar{E} \right] + \int_{\bar{\cal T}_{i}}
{\rm d}t \left[
{\textstyle \frac{1}{2}}\, \alpha e^{-2\Phi}\dot{\eta}
- \bar{N}\left( \bar{E} + {\textstyle \frac{1}{2}}
\bar{S}\right)\right]\, .
\end{equation}
Using the expressions (\ref{shorthand}a) and
(\ref{shorthand}c), one finds
that the inner boundary term is
\begin{equation}
\left({\cal S}^{\it 1}_{*}\right)_{\bar{\cal T}_{i}}
= - \int_{\bar{\cal T}_{i}} {\rm d}t
N\left[ E - v J
+ {\textstyle \frac{1}{2}}( S - v T)
\right]\, . \label{astinnerbt}
\end{equation}
It is also relatively straightforward to compute
the variation of ${\cal S}^{\it 1}_{*}$.
With the result (\ref{cvariation}) it follows that
the $\bar{\cal T}$ boundary and corner contributions to
the variation of
${\cal S}^{\it 1}_{*}$ are
\begin{eqnarray}
\lefteqn{(\delta {\cal S}^{\it 1}_{*})_{\bar{\cal T}, B'', B'} =}
& & \nonumber \\
& & - \int_{\bar{\cal T}_{o}}{\rm d}t
\left[\bar{E}\delta\bar{N} +
\bar{N}\bar{S}\delta\Phi
- \left(\bar{N}\bar{J} + \alpha
e^{-2\Phi}\dot{\Phi}\right)\delta\eta\right]
+ \left. \alpha e^{-2\Phi} \eta
\delta\Phi\right|^{B''_{o}}_{B'_{o}}
\nonumber \\
& &
- \int_{\bar{\cal T}_{i}}{\rm d}t
\left[\bar{E}\delta \bar{N}
+ {\textstyle \frac{1}{2}}
e^{-2\Phi}\delta
\left(e^{2\Phi}\bar{N}\bar{S}\right)
- \left(\bar{N}\bar{J} + \alpha
e^{-2\Phi}\dot{\Phi}\right)\delta\eta\right]
+ {\textstyle \frac{1}{2}}\left. \alpha e^{-2\Phi}
\delta\eta\right|^{B''_{i}}_{B'_{i}}\,\, .
\end{eqnarray}
We describe the boundary conditions at the inner boundary
as {\em completely open} because $\bar{N}$ and
$e^{2\Phi}\bar{S}$ are held fixed, as opposed to closed
or {\em microcanonical} boundary
conditions characterized by fixation of the energy $\bar{E}$
and fixation of $\Phi$
(effectively the $B$ surface area).\cite{BYpathint}
With some work and the formulae in (\ref{shorthand}),
one can show that
the inner-boundary and inner-corner-point
contributions in the above
variation may be combined into the expression
\begin{equation}
(\delta {\cal S}^{\it 1}_{*})_{\bar{\cal T}_{i}, B''_{i}, B'_{i}} =
- \int_{\bar{\cal T}_{i}} {\rm d}t
\left[\left(E - v J\right)\delta N
+ {\textstyle \frac{1}{2}}
e^{-2\Phi} \delta\left(Ne^{2\Phi}S
- v Ne^{2\Phi} T\right)
- (N/\gamma^{2}) J\delta\eta\right]\, .
\label{openbc}
\end{equation}
Now we introduce fall-off conditions on the fields which seal the
inner boundary. What we have in mind is a general foliation of
our spatially
bounded static region $\cal M$. All of the spatial slices meet
at the bifurcation point, but otherwise are essentially arbitrary.
Our phase space is the set of fields $(\Lambda, P_{\Lambda};
\Phi, P_{\Phi})$ with the appropriate fall-off conditions
near the bifurcation point. The needed fall-off conditions have
already been given in LW for the specific case
of the Schwarzschild geometry.
For convenience and without loss of generality,
take the inner boundary, the bifurcation point, to be
located at $r = r_{i} = 0$
and the outer boundary to be located at $r = r_{o} = 1$.
The boundary conditions given in LW are the following:
\begin{eqnarray}
\Lambda(t,r) & = & \Lambda_{0}(t) + O(r^{2}) \label{top}
\eqnum{\ref{top}a}\\
\Phi(t,r) & = & \Phi_{0}(t) + \Phi_{2}(t) r^{2} + O(r^{4})
\eqnum{\ref{top}b} \\
P_{\Lambda}(t,r) & = & O(r^{3})
\eqnum{\ref{top}c}\\
P_{\Phi }(t,r) & = & O(r)
\eqnum{\ref{top}d}\\
N(t,r) & = & N_{1}(t)r + O(r^{3})
\eqnum{\ref{top}e}\\
N^{r}(t,r) & = & N_{1}^{r}(t)r + O(r^{3})\, ,
\eqnum{\ref{top}f}
\addtocounter{equation}{1}
\end{eqnarray}
where $O(r^{n})$ stands for a term whose magnitude as
$r \rightarrow 0$ is bounded by $r^{n}$ times a constant.
Also, as $r \rightarrow 0$, the $k$'th derivative of such a
term is similarly bounded by $r^{n-k}$ times a constant for
$1 \leq k \leq n$. Note that the time development at the inner
boundary has been arrested as the lapse vanishes there.
One can show that
these boundary conditions are consistent with the equations of
motion. In other words, the Hamiltonian evolution preserves the above
boundary conditions, provided that the initial data obeys
both the above set of fall-off conditions
and the constraints (\ref{constraints}) on the initial spacelike
slice $\Sigma$ and provided that the lapse and shift also obey the
above fall-off conditions.\footnote{Note that
one must appeal to (\ref{Lambdanaught})
when showing that the boundary
conditions (\ref{top}) are consistent with
the $\dot{P}_{\Lambda}$ equation of motion.}
Moreover, the dynamical equation for $\Phi$ and
the above fall-off conditions imply that $\dot{\Phi}_{0} = 0$.
Imposition of the Hamiltonian constraint (\ref{constraints}a)
as $r \rightarrow 0$ yields the relation
\begin{equation}
(\Lambda_{0})^{2} = - 4 y^{-1} \lambda^{-2} \Phi_{2}
\exp\left[(y - 2)2\Phi_{0}
\right]\, , \label{Lambdanaught}
\end{equation}
which shows that $\Phi_{2}$ is negative for classical solutions.
Let us quickly compare this result with the Schwarzschild result found
in LW. Setting $R(t,r) = R_{0}(t)
+ R_{2}(t) r^{2} + O(r^{4})$ near the bifurcation point,
one finds that $\lambda R_{0} = e^{- \Phi_{0}}$
and $\lambda R_{2} = - \Phi_{2}\, e^{- \Phi_{0}}$. Therefore, for the
SSGR case the above expression is $(\Lambda_{0})^{2} = 4 R_{0} R_{2}$,
which is the LW result.
Application of the fall-off conditions (\ref{top}) to the inner boundary
term (\ref{astinnerbt}) shows that now the new action (\ref{astaction}) has
the following form:
\begin{eqnarray}
{\cal S}^{\it 1}_{*} & = & \int_{\cal M} {\rm d}^{2}x\left(P_{\Lambda} \dot{\Lambda}
+ P_{\Phi} \dot{\Phi} - N {\cal H} - N^{r} {\cal H}_{r}\right) \nonumber \\
& & + \int_{\bar{\cal T}_{o}} {\rm d}t
\left( \alpha e^{-2\Phi}\eta \dot{\Phi}
- \bar{N} \bar{E} \right) +
{\textstyle \frac{1}{2}}\int^{t''}_{t'}
{\rm d}t \left. \alpha e^{-2\Phi} (N'/\Lambda)\right|_{r = 0}\, ,
\label{thermoaction}
\end{eqnarray}
where in the last integral $t'$ and $t''$ now represent
integration parameters and not manifolds as they
have before.
We refer to (\ref{thermoaction}) as the {\em thermodynamical action}
because of its importance in the path-integral formulation of
gravitational thermodynamics. This is the appropriate
action with which to study the canonical ensemble for spherically
symmetric black holes.\cite{BMYW} Using (\ref{openbc}) in tandem
with the fall-off conditions (\ref{top}), one finds the following
boundary terms in the variation of the thermodynamical action:
\begin{eqnarray}
(\delta {\cal S}^{\it 1}_{*})_{\bar{\cal T}, B'', B'} & = &
{\textstyle \frac{1}{2}}\int^{t''}_{t'} {\rm d}t \left.
\alpha e^{-2\Phi}\delta\left(N'/\Lambda\right)\right|_{r = 0}
\nonumber \\
& & - \int_{\bar{\cal T}_{o}}{\rm d}t
\left[\bar{E}\delta\bar{N} +
\bar{N}\bar{S}\delta\Phi
- \left(\bar{N}\bar{J} + \alpha
e^{-2\Phi}\dot{\Phi}\right)\delta\eta\right]
+ \left. \alpha e^{-2\Phi} \eta
\delta\Phi\right|^{B''_{o}}_{B'_{o}}\,\, .
\nonumber \\
& &
\end{eqnarray}
For the $y = 1$ case this is precisely the action and variational
principle considered in LW. Notice that we have the same boundary
conditions at the outer boundary as before: $\bar{N}$ and $\Phi$ are
held fixed on this surface in the variational principle associated
with (\ref{thermoaction}). As spelled out in LW, the quantity
$N'/\Lambda$ which is fixed at the bifurcation point in the
variational principle has a direct physical interpretation. In fact,
$N'/\Lambda$ is the time rate of change of a certain boost parameter.
Each $\Sigma$ slice defines its own timelike normal $u^{a}$ at the bifurcation
point. As the $\Sigma$ slicing develops in time this vector is
continuously boosted at the rate $\left. (N'/\Lambda)\right|_{r = 0}
= N_{1}/\Lambda_{0}$.
\subsection{Canonical reduction of the thermodynamical action}
We want a new-variable version of the thermodynamical action which
is amenable to canonical reduction. From the work in $\S 4$ we already
know how to modify/handle the outer boundary term when passing to
the new canonical variables. Except for a minor difference, we will handle
the outer boundary term in the thermodynamical action just like in
the previous section. Therefore, the last section has already
addressed several of the delicate issues concerning the canonical
reduction of the thermodynamical action. However, there is a
new feature of the thermodynamical action which we need to worry about.
This new feature concerns the quantity $N'/\Lambda$ which is fixed at the
bifurcation point. We have already discussed why
fixation of $\bar{N}$ and $\Phi$ (now only at the outer boundary)
are important features of the variational principle.
The boundary integral at the bifurcation point is also of importance.
Indeed, in the thermodynamical paradigm black-hole entropy arises from
this term.\cite{BMYW,micro}
For the Schwarzschild case LW has shown that for applications
to gravitational thermodynamics is is crucial to retain fixation of
$N'/\Lambda$ at the bifurcation point when passing to a new-variable
version of the action. We also regard this feature of the action
principle as the feature of central importance which needs to be
preserved. Though essentially reviewing
the work of LW, this subsection shows how the Louko-Whiting formalism
extends to the 2dDG case. Therefore, our discussion has
relevance for the thermodynamics of pure dilaton gravity.
Let us present the quantities with which we will construct a
new-variable version of the thermodynamical action. The new canonical
variables are the same as before, since the transformation
(\ref{canonical}) remains canonical with the boundary conditions adopted
for the thermodynamical action. As before, the term
$- \left(\alpha e^{-2\Phi} \delta \Phi {{\boldmath \mbox{$\varphi$}}}\right)'$ in the identity
(\ref{generating}) gives two boundary terms. The one at
the outer boundary vanishes as $\Phi$ is held fixed on this
boundary element. Moreover,
${{\boldmath \mbox{$\varphi$}}}$ vanishes as $r \rightarrow 0$ so the inner boundary
term also vanishes.
Hence, upon integration (\ref{generating}) still shows that the difference
between
the old and new Liouville forms is an exact form. The new shift
$N^{\Psi}$ is again defined by (\ref{newlapseshift}b). However, following Louko
and Whiting, we define a {\em different} new lapse
\begin{equation}
{\sf N} = - N^{M} \left(\frac{yM}{\alpha}\right)
\left(\frac{\alpha\lambda}{2M}\right)^{2/y}.
\end{equation}
Recall that $M$ has units of length for the SSGR case
and units of inverse length for 2dDG. It is easy to
see that $N^{M}$ is dimensionless for both cases, and, therefore,
for both cases $\sf N$ has units of inverse length.
It turns out that this choice for the
lapse must be made in order to ensure that we retain fixation of
the boost rate $N'/\Lambda$ at
the bifurcation point as a feature of our
variational principle. In terms of
${\sf N}$ the boundary lapse is given by
\begin{equation}
\bar{N}^{2} = \left(\frac{\alpha}{yM}\right)^{2}
\left(\frac{2M}{\alpha\lambda}\right)^{4/y}
F {\sf N}^{2} - \lambda^{-2}
\exp\left[(y - 2)2 \Psi\right] F^{-1}\left(N^{\Psi}\right)^{2}\, .
\label{differentNbar}
\end{equation}
The fall-off conditions (\ref{top}) imply the following fall-off
conditions for the new variables:
\begin{eqnarray}
M(t,r) & = & {\textstyle \frac{1}{2}}
\alpha\lambda \exp\left[- y \Psi_{0}(t)\right]
+ M_{2} (t) r^{2} + O (r^{4}) \label{newfall}
\eqnum{\ref{newfall}a} \\
\Psi(t,r) & = & \Psi_{0}(t) + \Psi_{2} r^{2} + O(r^{4})
\eqnum{\ref{newfall}b} \\
P_{M}(t,r) & = & O(r) \eqnum{\ref{newfall}c} \\
P_{\Psi}(t,r) & = & O(r)\eqnum{\ref{newfall}d} \\
{\sf N}(t,r) & = & {\sf N}_{0}(t) + O(r^{2})\eqnum{\ref{newfall}e} \\
N^{\Psi}(t,r) & = & N_{2}^{\Psi}(t)r^{2} + O(r^{4})\, ,
\eqnum{\ref{newfall}f}
\addtocounter{equation}{1}
\end{eqnarray}
where
\begin{eqnarray}
\Psi_{0} & = & \Phi_{0} \label{newfall2} \eqnum{\ref{newfall2}a} \\
\Psi_{2} & = & \Phi_{2} \eqnum{\ref{newfall2}b} \\
M_{2} & = & - {\textstyle \frac{1}{2}} \alpha\lambda\Phi_{2}
\exp(- y\Phi_{0})\left[y + 4\lambda^{-2}(\Lambda_{0})^{-2}\Phi_{2}
\exp[(y-2)2\Phi_{0}]\right]
\eqnum{\ref{newfall2}c} \\
{\sf N}_{0} & = & - {\textstyle \frac{1}{4}} y \lambda^{2} N_{1}
\Lambda_{0}(\Phi_{2})^{-1}
\exp\left[(2 - y)2\Phi_{0}\right] \eqnum{\ref{newfall2}d} \\
N^{\Psi}_{2} & = & 2 \Phi_{2} N^{r}_{1} \eqnum{\ref{newfall2}e}\, .
\end{eqnarray}
We also have that
\begin{equation}
F = 4\lambda^{-2}(\Phi_{2})^{2}(\Lambda_{0})^{-2}\exp[(y - 2)2\Phi_{0}] r^{2}
+ O(r^{4})\, .
\end{equation}
For the SSGR case these fall-off results for the new variables
match those given in LW.
Equation (\ref{Lambdanaught}) implies that
\begin{equation}
\left. (N'/\Lambda)\right|_{r = 0} = - {\textstyle \frac{1}{4}} y \lambda^{2}
N_{1} \Lambda_{0}(\Phi_{2})^{-1} \exp\left[(2 - y)2\Phi_{0}
\right]\, ,\label{bstrate}
\end{equation}
which in turn gives ${\sf N}_{0} = N_{1}/\Lambda_{0}$.
Hence, we want to ensure that the
Lagrange multiplier ${\sf N}$ is fixed at the bifurcation point in
our new variational principle.
Let us now consider the new-variable version of the thermodynamical
action and show that it has the correct variational principle.
We could write down a general action which
covers both the SSGR and 2dDG cases, but the expression is a bit
unseemly. Therefore, let us examine both cases separately.
For the SSGR case we have
\begin{eqnarray}
{\cal S}_{\diamond}^{\it 1} & \equiv &
\int_{\cal M}{\rm d}^{2}x \left( P_{M}\dot{M}
+ P_{\sf R}\dot{\sf R} + 4{\sf N} M M'
- N^{\sf R} P_{\sf R}\right) \nonumber \\
& & + \left. \int_{t'}^{t''}
{\rm d}t\, 2 M^{2}{\sf N}\right|_{r = 0} + \int_{\bar{\cal T}_{o}}
{\rm d}t \left[ - \rho {\sf R}\dot{\sf R} - \bar{N}\bar{E}\right]\, ,
\label{LWaction}\end{eqnarray}
with the boundary lapse given by
\begin{equation}
\bar{N}^{2} = 16 M^{2}
F {\sf N}^{2} - F^{-1}\left(N^{\sf R}\right)^{2}\, .
\end{equation}
To get these expressions we have used the facts that $e^{-\Psi} = \lambda {\sf
R}$, $P_{\Psi} = - {\sf R} P_{\sf R}$, and $N^{\Psi} = - N^{\sf R}/{\sf R}$
(all in the notation of KVK and LW). Also, $\bar{E}$ and $\rho$ are still
given by (\ref{anotherEandJ}b) and (\ref{newrho}), respectively, but now
one must express them in terms of ${\sf R}$. This is precisely the action
considered in LW. For Witten's 2dDG model we find
\begin{eqnarray}
{\cal S}_{\diamond}^{\it 1} & \equiv &
\int_{\cal M}{\rm d}^{2}x \left[ P_{M}\dot{M}
+ P_{\Psi}\dot{\Psi} + \lambda^{-1}{\sf N} M'
- N^{\Psi} P_{\Psi}\right] \nonumber \\
& & \left.\int_{t'}^{t''}
{\rm d}t \lambda^{-1} M {\sf N}\right|_{r = 0} + \int_{\bar{\cal T}_{o}}
{\rm d}t \left[
\alpha e^{-2\Psi}\rho\dot{\Psi} - \bar{N}\bar{E}\right]\, ,
\label{dilatonthermo}
\end{eqnarray}
with
\begin{equation}
\bar{N}^{2} = \lambda^{-2}
F {\sf N}^{2} - \lambda^{-2} F^{-1}\left(N^{\Psi}\right)^{2}\, .
\label{CGHSWNbar}
\end{equation}
In contrast to the SSGR case, for 2dDG
the Lagrange multipler $N^{M}|_{r = 0}$ does, apart only
from a dimensionful constant, specify the boost rate of the $\Sigma$ normal
$u^{a}$ at the bifurcation point.
The variation of (\ref{LWaction})
has already been considered in LW,
so we will only consider the variation of (\ref{dilatonthermo}).
The equations of motion derived from (\ref{dilatonthermo}) are the
same as those given in (\ref{newEOM}),
except that now $\dot{P}_{M} = - \lambda^{-1} {\sf N}$.
Upon variation of the action (\ref{dilatonthermo}),
we find the boundary terms
\begin{eqnarray}
\left(\delta {\cal S}_{\diamond}^{\it 1}\right)_{\partial {\cal M}}
& = & \int^{t''}_{t'}{\rm d}r
\left(P_{M} \delta M + P_{\Psi}\delta\Psi\right)
+ \left.\int_{t'}^{t''}
{\rm d}t \lambda^{-1} M \delta {\sf N}\right|_{r = 0}\nonumber \\
& &
+ \int_{\bar{\cal T}_{o}}{\rm d}t\left( \bar{\Pi}_{\bar{N}} \delta \bar{N}
+ \bar{\Pi}_{\Psi} \delta \Psi + \bar{\Pi}_{M} \delta M\right) +
\left. \alpha e^{-2 \Psi}\rho
\delta\Psi\right|^{B_{o}''}_{B_{o}'}\, . \label{deltadiamond}
\end{eqnarray}
In the first integral $t'$ and $t''$ represent spacelike
slices, while in the second integral they are integration
parameters. The $\bar{\cal T}_{o}$ momenta in (\ref{deltadiamond})
are essentially the same as the outer-boundary ones
in (\ref{dagmomenta}), except that now $\bar{N}$ stands for
(\ref{CGHSWNbar}) and the momenta $\bar{\Pi}_{M} = \lambda^{-1} {\sf N}
+ (\alpha\lambda)^{-1} e^{y\Psi} F^{-1} \bar{N}\bar{E}$.
Like before, plugging in the explicit formula (\ref{CGHSWNbar})
for $\bar{N}^{2}$, one finds that $\bar{\Pi}_{M}$ vanishes
when the equation of motion $\dot{P}_{M} = - \lambda^{-1} {\sf N}$
holds. Therefore, $M$ need not be held fixed on $\bar{\cal T}_{o}$
in the variational principle associated with ${\cal S}_{\diamond}^{\it 1}$.
It is now straightforward to pass to the reduced form of the
thermodynamical action.
\section{Discussion}
We conclude with some comments concerning the possible
extension of the KVK and LW formalisms to other two-dimensional models
of gravity. Recently, important progress has been made
in the field of two-dimensional gravity with the realization
that a huge class of two-dimensional models can
be described within the framework of the so-called {\em
Poisson-sigma models} (PSM's)
of Kl\"{o}sch, Schaller, and Strobl.\cite{Schaller}
For all such models there exists an absolutely conserved
quantity $C$ (referred to as a {\em Casimir function}
in the Poisson-sigma model language) which is analogous to our
$M$ expression (\ref{invariant}), and recently Kummer and Widerin
have explored the relationship between the PSM $C$
and notions of quasilocal energy for such models.\cite{WW}
Many of our results, especially those concerning our general
treatment of quasilocal energy-momentum, seem to extend
to the general PSM formalism.
In particular, the absolutely conserved quantity $C$ can
be interpreted as a quasilocal boost invariant.\cite{WL}
Extension of the canonical-reduction method of KVK to PSM
theory also seems
possible, though several technical difficulties lie in the
way. For instance, one encounters an almost limitless
variety of singularity structures when considering the
set of all PSM's.\cite{Schaller} For SSGR the canonical
transformation of KVK is singular at the horizon. Similar technical
difficulties are likely to surface when applying the KVK method
to any two-dimensional model. Since the collection of all
possible Penrose diagrams obtainable from PSM's
is so large, it is questionable whether or
not a fully unified treatment for the canonical reduction of
all PSMs is possible. On the other hand,
the richness of singularity structures in PSM gravitation offers
what is perhaps a promising testing ground for gravitational thermodynamics.
The appropriate thermodynamical action, as expressed in the LW formalism,
would be a crucial ingredient in any study of PSM thermodynamics
via reduced canonical variables.
\section{Acknowledgments}
For discussions and helpful comments I thank H. Balasin,
J. D. Brown, W. Kummer, F. Schramm, S. Sinha, and J. W. York.
The bulk of this research has been supported by the Fonds zur
F\"{o}r\-der\-ung der wis\-sen\-schaft\-lich\-en For\-schung
(Lise Meitner Fellowship
M-00182-PHY). Also, some of this work was done
while visiting the University of North Carolina at Chapel Hill
with support provided by the National Science Foundation,
grant number PHY-9413207. Finally, this work was begun in
Pune at the Inter-University Centre for Astronomy \& Astrophysics,
with support from the Indo-US Exchange Programme and the
University Grants Commission of India.
\noindent
{\small NOTE ADDED:
After this project was completed, we learned of a new
paper\cite{Madhavan} by Varadarajan which also
treats the canonical reduction of two-dimensional pure
dilaton gravity in the manner of KVK. Moreover, this paper
also considers quantization.}
|
\section{Introduction}
In analysing the supersymmetric theories
one usually assumes the conservation of a discrete symmetry, called R
parity\cite{wein}, distinguishing matter fields from their superpartners.
The R parity violation could lead to phenomenologically interesting
effects, such as mixing of gauginos with leptons\cite{valleross} and Higgs
bosons with sleptons\cite{mas1,asj2}
(and hence breaking of the lepton number), non-zero
neutrino masses\cite{hall,asj2} and their decays\cite{RPMSW},
$Z^0$ decays to single charginos or neutralinos\cite{ROMA},
existence of a massless Goldstone boson called majoron\cite{CMP}
and many others.
In contrast to the R parity breaking in the fermion sector,
its consequences in the scalar sector are relatively less
explored. We present three examples of possible manifestations
of R-parity breaking in the Higgs sector:
high-energy mono-photon production at LEP\cite{MY3},
associated production of invisibly decaying Higgs bosons\cite{MY1}
and novel scalar decays in the MSSM with broken R parity\cite{MY2}.
\section{Single Photon Decays of the $Z^0$ and SUSY with Broken R-Parity}
Recently the OPAL collaboration
has published a high statistics single photon spectrum that shows
some excess of high energy photons above the expectations from g
the initial state radiation (ISR)\cite{opal}.
This excess could be explained in a class of SUSY
models with spontaneous violation of R parity,
\begin{figure}[htbp]
\begin{center}
\vspace*{-3mm}
\mbox{
\epsfig{figure=zfj1.eps,height=0.16\linewidth}
}
\end{center}
\vspace*{-5mm}
\caption{Diagrams contributing to the monochromatic
single photon emission $Z^0 \rightarrow \gamma J$
\label{fig:zfj1}}
\vspace*{-4mm}
\end{figure}
which predict the existence of new $Z^0$ decay modes
involving single ($Z^0 \rightarrow J \gamma$) or double
($Z^0 \rightarrow JJ \gamma$) invisible massless majoron emission.
Single majoron emission would produce
monochromatic photons emitted with energy $\frac{M_Z}{2}$.
In order to estimate the attainable values of Br($Z^0 \rightarrow J \gamma$)
we adopt as a model for the spontaneous R parity violation
the one proposed in\cite{MASI_pot3}. Diagrams contributing at the 1-loop
level to the process $Z^0 \rightarrow J \gamma$ are shown in
Fig.~\ref{fig:zfj1}. We obtained the explicit expression for
Br($Z^0\rightarrow\gamma J$) by
calculation of the 3-point Green functions of Fig.~\ref{fig:zfj1}.
We have varied model pa\-ra\-me\-ters, rejecting points
violating constraints imposed by existing
laboratory observations, cosmology and astrophysics.
We have found that the branching ratio for the 2-body $Z^0$ decay
to a single photon + missing momentum in the considered
model can reach $10^{-5}$, being within the sensitivity of LEP\cite{MY3}.
Double emission process $Z^0 \rightarrow JJ \gamma$ would give rise to a
continuous photon spectrum\footnote{
We do not consider here other possible mechanisms, like
LSP pair production followed by the radiative
decay of one of them $\chi^0 \rightarrow \gamma \nu$ and the invisible decay
of the other $\chi^0 \rightarrow J \nu$.}.
Imposing gauge and $CP$ invariance and Bose symmetry,
one can express the on-shell amplitude
in terms of a single form-factor $V_{0}$:
\begin{equation}
A (Z^0 \rightarrow JJ \gamma ) = V_{0}
[p^{\mu} (q_{1} + q_{2})^{\nu} - p (q_{1} + q_{2})g^{\mu \nu} ]
\epsilon_{\mu} (p+q_1+q_2) \epsilon_{\nu} (p)
\label{amplitude}
\end{equation}
As the simplest illustration we derive the
$\gamma$-spectrum following from eq.~(\ref{amplitude}) in the approximation of
constant $V_{0}$. We obtain
\begin{equation}
\frac{d \Gamma}{d E_{\gamma}} = \frac{|V_{0}|^{2} M_Z}{96 \pi^{3}}
E_{\gamma}^{3}~~~~~~~~~~~~0 \leq E_{\gamma} \leq \frac{M_Z}{2}
\end{equation}
The details of the shape of the $\gamma$ spectrum depend upon the specific
model and SUSY parameters choice, but its general characteristic is
determined by kinematics and is
different from the SM process $Z^0 \rightarrow \nu \overline{\nu} \gamma$.
Superimposed upon the continuous spectrum we have, in addition,
a spike at its endpoint, corresponding to the emission of the
monochromatic $\gamma $ in the $Z^0 \rightarrow \gamma J $ decay.
\section{Limits on Associated Production of
Invisibly Decaying Higgs Bosons from $Z^0$ Decays}
In a class of models with a spontaneously broken global
symmetry the CP-even Higgs boson(s) are expected to have sizeable invisible
decay modes to the majorons\cite{bj}. This decay
could contribute to the signal looked for at LEP -- two
acoplanar jets + missing momentum.
The existing analyses of the invisible Higgs search
have concentrated on a minimally extended SM Higgs sector with the
addition of a Higgs singlet and hence on the Bjorken process\cite{alfonso}.
We extend this analysis to include the model containing two
Higgs doublets and a singlet (carrying lepton number)\cite{MASI_pot3}
and the associated production of the CP-even and CP-odd Higgs
bosons\cite{MY1}.
For the illustration, we display the constraints
obtained for the published ALEPH data sample\cite{Aleph92}
based on a statistics of $1.23\times 10^6$ hadronic $Z^0$ events.
After spontaneous $SU(2) \times U(1) \times U(1)_L$
breaking, the Higgs sector of the considered model contains
3 massive CP-even scalars $H_i$,
a massive CP-odd scalar $A$ and
the massless majoron $J$. We assume that at LEP only the
lightest CP-even scalar $H_1 \equiv H$ can be produced.
The relevant couplings are:
\begin{eqnarray}
\label{HZZ3}
{\cal L}_{HZZ}&=&(\sqrt{2}G_F)^{1/2}M_Z^2\;\epsilon_B \;Z_{\mu}Z^{\mu} H\\
{\cal L}_{HAZ}&=&-{e\over \sin\theta_W\cos\theta_W}\;\epsilon_A\; Z_{\mu}H
\stackrel{\leftrightarrow}{\partial^\mu}A.
\end{eqnarray}
where $\epsilon_A^2,\epsilon_B^2\leq 1$.
The main $H$ decay modes are $b\overline b$, $JJ$ and $AA$
(if $m_H>2m_A$).
$HJJ$ coupling strength is unconstrained
and can be effectively parameterized by
$B$ = Br$(H \rightarrow JJ)$, $0\leq B \leq 1$\footnote{The parameterization of the
expected number of the invisible Higgs boson events in terms of
$m_H$, $m_A$, $\epsilon_A$, $\epsilon_B$ and $B$ is quite general
and not limited only to the model\cite{MASI_pot3}.}.
The decay $A\rightarrow JJJ$ does not exist since the (CP-allowed) couplings
$AJ^3$ or $AHJ$ vanish
at the tree level in our model.
\begin{figure}[htbp]
\vspace*{-7mm}
\begin{tabular}{p{0.468\linewidth}p{0.468\linewidth}}
\begin{center}
\mbox{
\epsfig{file=dp1.eps,width=0.95\linewidth,height=0.9\linewidth}}
\end{center} &
\begin{center}
\mbox{
\epsfig{file=dp3.eps,width=0.95\linewidth,height=0.9\linewidth}}
\end{center} \\
\vspace*{-12mm}
\caption{
Numbers of expected dijet + missing momentum events
after imposing ALEPH cuts.
\label{fig:dp1}
}&
\vspace*{-12mm}
\caption{
Limits on $\epsilon^2_{A}$ in the $m_A$-$m_H$ plane based on
$e^+ e^- \rightarrow HA \rightarrow JJ b\overline b$ channel.
We have assumed Br($H \rightarrow JJ$) = 100\%.
\label{fig:dp3}
} \\
\end{tabular}
\vspace*{-12mm}
\end{figure}
Since $A$ can decay only visibly ($B_A=$ Br$(A\rightarrow b \overline{b})
\approx 90\%$ for $m_A \geq 20$ GeV), one expects dijets +
missing momentum as a signal of the Higgs production.
This signal arises from three processes:
1) $Z^{0\star} \rightarrow \nu \overline{\nu}$,
$H \rightarrow b \overline{b}$,
2) $Z^{0\star} \rightarrow q \overline{q}$, $H \rightarrow JJ$
and 3) $H\rightarrow JJ$, $A\rightarrow b \overline{b}$.
For each process one has a sizeable missing momentum which is aligned
neither along the beam nor along the jets. For the SM
background missing momentum arises from i) jet fluctuation
(including escaping $\nu$ from $b,c$ and $\tau$ jets) in which case it
is aligned along the jets, and ii) ISR or
$e^+e^- \rightarrow (e^+e^-) \gamma\gamma$ process in
which case it is aligned along the beam direction. This enables one to
eliminate the SM background by a combination of kinematic cuts\cite{aleph1}.
We denote the number of expected signal events for the processes 1)-3), after
the cuts, by $N_{SM}$, $N_{JJ}$ and $N_A$ respectively, assuming no
suppression due to the mixing angles or branching fractions in each
case. The expected number of signal events, after incorporating
these effects, is given by
\begin{equation}
N_{2j}=\epsilon_B^2\left[ B N_{J}+(1-B) N_{SM}\right]+\epsilon_A^2
B_A B N_A
\end{equation}
As can be seen in Fig.~\ref{fig:dp1}, $N_A \gg N_{SM}, N_{JJ}$,
implying that, if not suppressed kinematically or by mixing angles,
associated production gives the strong limits
on $\epsilon_A^2$ (see Fig.~\ref{fig:dp3},
where we have assumed fully invisible $H$ decay).
The limit on $\epsilon_B^2$ is given by the Bjorken process and is
the same as obtained in\cite{alfonso}.
A decay mode independent limits on $\epsilon_A^2$,
$\epsilon_B^2$ can be obtained by varying $B$
from 0 to 1 and combining
the data on dijet + missing momentum with those from 4
and 6 $b$-jet searches. The overall limit is
dominated by visible channels, which give weaker constraints
on $\epsilon_A^2$, $\epsilon_B^2$ then those obtained for the invisible
$H$ decay.
\section{Novel Scalar Boson Decays in SUSY with Broken R-Parity}
The standard MSSM R parity conserving superpotential can be generalized
by adding the following R parity (and lepton number) violating terms:
\begin{equation} \label{wr}
W_R=\epsilon_{ab}\left[\lambda_{ijk}\hat{L}_i^a \hat{L}_j^b
\hat{E}_k^C + \lambda'_{ijk}\hat{L}_i^a \hat{Q}_j^b \hat{D}_k^C
+ \epsilon_i \hat{L}_i^a \hat{H}_2^b \right]
\end{equation}
We focus on the effect of the last term in \eq{wr},
assuming in addition $\epsilon_{1,2}\approx 0$.
The remaining $\epsilon_3\hat{L}_3 \hat{H}_2$
term induces a non-zero VEV for the tau sneutrino:
$\VEV{\tilde{\nu_{\tau}}} \equiv \frac{v_3}{\sqrt 2}$ and
leads to mixing of gauginos with leptons, Higgs scalars
with the tau sneutrino and hence
to the R parity violating Higgs bosons decay
modes.
All the elements of the various mixing matrices can be expressed
in terms of six independent parameters
which we choose as
$\tan \beta = \frac{v_2}{\sqrt{v_1^2 + v_3^2}}$,
$\mu$, $\epsilon_3$, $m_A^2$, the gaugino and soft sneutrino
mass parameters $M_2$ and $m_{L_3}$.
We have taken into account the following constraints on the model parameters:
\noindent 1) The major constraint on
$\epsilon_3$ and hence on R parity violating mixings comes
from bound on \hbox{$\nu_\tau$ } mass\cite{PDG94}, induced by the non-zero
$\epsilon_3$: $\hbox{$m_{\nu_\tau}$ }\leq 30$ MeV\footnote{see
comments in\cite{MY2} on possible cosmological constraints.}.
\noindent 2) Decay widths $\Gamma(Z^0\rightarrow \chi^0\chi^0,\chi^+\chi^-)$
should obey the LEP restrictions.
\noindent The decay $h\rightarrow \chi^0 \nu_\tau$ occurs either through the
sneutrino component of $h$,
\begin{equation}
h=a_{31} (\tilde{\nu}_\tau)_R + a_{21} (\phi_2)_R + a_{11} (\phi_1)_R
\label{a31}
\end{equation}
or through $\nu_{\tau}$ admixture in the LSP in the
$h \chi\chi$ vertex. The mixing $a_{31}$ appearing in~\eq{a31}
is of the order of
${\cal O}\left(\frac{\mu\epsilon_3}{m_h^2 -
m_{\tilde{\nu}_\tau}^2}\right)$ and
become large if the {\sl sneutrino} mass is close to the mass of the
relevant {\sl Higgs} boson.
The relative importance of the SUSY decay mode
$h \rightarrow \chi^0 \nu_\tau$ follows from the ratio:
\begin{figure}[htbp]
\vspace*{-8mm}
\begin{tabular}{p{0.468\linewidth}p{0.468\linewidth}}
\begin{center}
\mbox{
\epsfig{file=elh2.eps,width=\linewidth,height=\linewidth}}
\end{center} &
\begin{center}
\mbox{
\epsfig{file=elh5.eps,width=\linewidth,height=\linewidth}}
\end{center} \\
\vspace*{-16mm}
\caption{Br$(h \rightarrow \chi^0\nu_{\tau})$ as a function of the difference of
the physical $h$ and $\tilde{\nu}_{\tau}$ masses.
The \hbox{$\nu_\tau$ } mass labels the different curves.\label{fig:eps2}} &
\vspace*{-16mm}
\caption{Branching ratios for $\tilde{\nu}_{\tau}$ decays as a function of
its mass for $\tan\beta$=10,$M_2$=70
GeV,$\mu$=200$\,$GeV,$\epsilon_3$=1$\,$GeV,$m_A$=250$\,$GeV.\label{fig:eps5}}\\
\end{tabular}
\vspace*{-12mm}
\end{figure}
\begin{eqnarray}
\label{R}
R_0 = \frac{\Gamma(h\rightarrow \chi^0 \nu_\tau)}{\Gamma(h\rightarrow b \overline{b})}
\approx \frac{\tan^2 \theta_W}{2}\frac{M_W^2}{m_b^2}
\frac{ (1- m_{\chi}^2/M_H^2)^2 }{ (1-4 m_b^2/M_H^2)^{3/2} }
\frac{a_{31}^2}{a_{11}^2} \cos^2\beta \vert\xi\vert^2
\end{eqnarray}
where $\xi$ denotes the appropriate gaugino-LSP mixing element. $R_0$ and
also $R_+ = \frac{\Gamma(h\rightarrow \chi^\pm \tau^\mp)}{\Gamma(h\rightarrow b \overline{b})}$
can be estimated to be of the order of
${\cal O}\left({\epsilon_3^2\over m_b^2}\right)\times{\mu^2\over
(m_h-m_{\tilde\nu_{\tau}})^2}$
In Fig.~\ref{fig:eps2} we display the branching ratios for $h$
decays to LSP + \hbox{$\nu_\tau$ } as a function of the $h$-$\tilde{\nu}_{\tau}$
mass difference,
for a suitable choice of SUSY parameters. Clearly, for relatively small
$h$-$\tilde{\nu}_{\tau}$ mass differences of a few GeV, the supersymmetric
channel can dominate over the SM ones (similarly for $A\rightarrow \chi^0\nu_{\tau},
\chi^{\pm}\tau^{\mp}$ decays). Conversely, $\tilde{\nu}_{\tau}$ may decay
into R parity violating SM channels such as
$b \overline{b}$, $\tau^+ \tau^-$ or the invisible mode
$\nu \overline{\nu}$.
These decays are dominant when
the phase space for the R parity conserving channels
such as $\chi \nu$ is closed (see Fig.~\ref{fig:eps5} where
$m_{LSP} \approx 65$ GeV) and may be
non-negligible even above the LSP
threshold if $m_h\approx m_{\tilde{\nu}_{\tau}}$,
leading to a resonant enhancement of
$b \overline{b},\tau^+ \tau^-$ modes
(see small rise of Br($\tilde\nu_\tau \rightarrow b \overline b$) in
Fig.~\ref{fig:eps5}
for $m_{\tilde{\nu}_{\tau}} \approx m_h = 155$ GeV).
Finally, the R parity breaking terms
lead to the unstable LSP, which would decay inside the detector if
$\epsilon_3 \sim $ few GeV.
Folding the R parity violating Higgs boson decays
to the LSP with the standard decays of $Z^0$
one gets the signatures in $e^+e^-$ collisions which do not occur in the
SM, such as $\tau$$e$ or $\tau$$\mu$ pairs + missing momentum\cite{MY2}.
\section*{Acknowledgements}
\vspace{-2mm}
This work was supported in part by postdoctoral fellowship of the
Spanish MEC and by Alexander von Humboldt Stiftung.
\vspace{-3mm}
\section*{References}
|
\section{Introduction}
A problem of current interest is the ``reverse problem''
for protein folding: given a desired structure with
coordinates $\Gamma \equiv \{ r_1,\dots,r_N \}$, what is the
best choice of sequence $S$ to obtain that structure?
This problem is of great importance to
biotechnology as it facilitates the design of new drugs.
Progress towards a solution of this problem has been made\cite{yue&dill}, but
there is still no trustworthy algorithm to accomplish
this task.
To find the
optimum sequence it would seem necessary to apply
some minimization procedure over the space of all
possible sequences $S$. This had previously been
attempted\cite{SG,shakhnovich} by taking the Hamiltonian, which depends
both on $\Gamma$ and $S$ and minimizing it over $S$,
at fixed $\Gamma$. We will discuss this in more detail
below. As was recently argued\cite{us}, this is not guaranteed
to work, and in general is not an optimal design strategy.
Instead a new method was introduced that provides a
systematic approach to this problem. It was shown
that a minimization function exists though it
is more numerically intensive to evaluate than the
Hamiltonian. Applied to a system of block copolymers,
the method proved to be highly successful\cite{us}.
The purpose of this paper is to examine this method
for some simple models of much recent interest, the 27-mer
cube with ising interactions\cite{SG},
and the two dimensional HP model of Lau and Dill\cite{HP}.
It will be shown below that this new
minimization function works very well and gives results
far better than the minimization of the Hamiltonian attempted
previously\cite{SG}.
In designing proteins, three criteria are generally
considered to mark success.
First, the engineered molecule should have
the desired conformation when it is in its ground state.
Second, there should be no ground state degeneracy, so that the
molecule will always fold into the same conformation. Third,
there should be a large gap in the energy of the ground state
and the energies of the higher energy states so that the
molecule will stably sit in the ground state.
For some systems the method of minimizing the Hamiltonian can be
shown to always be successful in finding a sequence which has
a desired conformation as its ground state, as we will
examine below. But
sadly, we will see that the
sequences designed this way will usually fold up into thousands of other
structures, all with the identical energy.
Our new method gives sequences with a much lower
degeneracy, often producing unique ground states in the
desired conformation, and failing quite rarely.
In addition, we predict that it will
also result in sequences with a large gap in the energy
spectrum above the ground state.
This work is significant in that
it describes a general method to apply to any
model of protein folding and is
based on a systematic approach to the problem. Previous work\cite{SG}
does not appear to have recognized the need for a less
{\em ad hoc} approach, and instead blamed the nature
of the problem, rather than the faults in its solution.
For example, it has been pointed out\cite{whoops}
that within a given model there are ``good" desired
structures, structures that have sequences that will fold uniquely to
them. Structures that do not have this property are deemed ``bad".
When the method of minimizing the Hamiltonian failed to find unique
ground states, it was argued that it was because the
initial structures chosen were ``bad"
structures. As a result, it has been thought that more complicated
models were needed so that ground state degeneracies would be
reduced\cite{whoops},\cite{80-mer}, and that some means of
distinguishing between ``good" initial structures and ``bad" ones might be
necessary\cite{whoops}. While these considerations are valid,
our results show that the basic
problem of how to find the sequence with the lowest degeneracy that
folds into a given random conformation has not been addressed properly
in previous work. For a model where we can quantify
the success of this previous method, we find that most
``good" structures are misdesigned.
It is therefore important to fully understand the basic problem
of reverse folding before moving on to more complicated models. In
this letter we hope to clarify the basic logic behind the reverse folding
problem.
The outline for this letter is as follows. First the
results of this new method are summarized. Then the
approximate but numerically efficient scheme that we use is
briefly reviewed and related to recent work. In section
\ref{sec:results} this
method is applied numerically to the case of two
different molecular types on a $3 \times 3 \times 3$ lattice.
It is also compared with
the previous method of Shakhnovich and Gutin.
We then do a numerical comparison of the two dimensional
HP model with 16 monomers, where an even more detailed
comparison can be made. Finally the feasibility of
other extensions to the this work is discussed.
\section{Minimization Function}
\label{sec:minimization}
Here we summarize previous results\cite{us} in
which we obtained the minimization function
appropriate for use in the reverse folding problem.
Given a system with coordinates denoted by $\Gamma$ and chemical sequence
by $S$, the probability
that a sequence $S$ gives the desired structure $struct$ is
\begin{equation}
P(S|struct) \propto\exp ({-(F_{struct}(S)-F_o(S))/T})\equiv \exp ({-\Delta F/T})
\label{eq:minimize}
\end{equation}
where
\begin{equation}
\exp ({-F_o(S)/T}) \equiv \sum_{\Gamma} \exp ({-H_S(\Gamma)/T})
\end{equation}
\begin{equation}
\exp ({-F_{struct}(S)/T}) \equiv \sum_{\Gamma} \exp ({-(H_S(\Gamma)+V_{ext}(\Gamma))/T}) ~,
\end{equation}
with the sums being done over all possible structures.
$F_{struct}(S)$ is the free energy {\it pushed}
into a certain structure by the ``clamping potential", $V_{ext}(\Gamma)$, which
is chosen to force the molecule to have the
desired structure. So the optimum sequence is the one with the smallest
difference $\Delta F$ between the unrestricted free
energy and the free energy ``clamped" in the desired
structure.
This can be understood physically as follows. Suppose one desired
to create a protein of $N$ amino acids that folded into a
desired structure $struct$. One could imagine doing this in a brute force
way by creating all $20^N$ possible polypeptide chains all
maintained at equilibrium in a container. This is the ``sequence
space soup" of Chan and Dill\cite{soup}. One wishes to pick
out the molecule most closely resembling the desired structure.
To do this, one creates a kind of ``fish hook", that is a scaffolding
potential for which only molecules resembling $struct$ will
fit. This is the same as $V_{ext}$. $V_{ext}$, by design,
will predominantly
``catch" molecules of the
desired sort. In fact the probability of it containing a molecule
with sequence $S$ is precisely the same as eqn. (\ref{eq:minimize}).
It is also apparent that by minimizing the $\Delta F$ of
eqn. (\ref{eq:minimize}), one is not only finding a sequence
that will fold into the correct structure, but also the
sequence that has the highest probability of being in the
desired structure, at finite temperature. Therefore this
method satisfies the the third criterion stated above.
The above method has much in common with those used in
learning theory of ``neural networks''\cite{hertz}.
\section{Application to lattice models}
We look at two simplified lattice models with a self-avoiding chain and
apply our design algorithm to it. The first model has been studied much
recently in connection with the design problem\cite{SG}.
We use a $3\times 3 \times 3$ lattice and a 27-monomer chain so that it is
possible to directly enumerate the energies of all
conformations with a given sequence. Thus we can easily evaluate
the efficacy of our method.
The model involves sequences
$\{ \sigma_i\}$ of
two possible monomer types that are given values $\pm 1$, for chains of length
$N$. The monomer type values, $\{ \sigma_i\}$, plus the
positions of all the monomers, $\{r_i\}$, completely describe the state
of the chain. The energy is
\begin{equation}
E(\{\sigma_i\}, \{r_i\}) =
{1\over 2} \sum_{i,j}^N (B_0 + B\sigma_i\sigma_j)\Delta({\bf r}_i-{\bf r}_j)
\label{eq:model}
\end{equation}
We have taken $\Delta({\bf r}_i-{\bf r}_j) = 1$ if $\{r_i\}$ and $\{r_j\}$
are nearest neighbors and zero otherwise.
$B$ is negative so that the monomer types will segregate, giving
ferromagnetic ordering of the $\sigma$'s. $B_0$ has been taken to
have a large magnitude and is negative
so that there is a large attraction between all nearest neighbors,
causing the protein to collapse into a shape of minimal surface
area, in this case a cube. Thus the only conformations we
need consider are the internal arrangements of a chain packed
into a cube.
To find the sequence most likely to fold into some desired conformation,
we want to minimize $\Delta F$ as defined in
(\ref{eq:minimize}). Specializing to the model considered
here, the clamping potential is a delta function since we would
like to pick out one specific structure. If we call the coordinates of the
desired structure $\{r^0_i\}$, $\Delta F$ becomes
\begin{equation}
\Delta F = E(\{\sigma_i\}, \{r^0_i\}) - F_o(\{\sigma_i\})
\label{eq:correct}
\end{equation}
As described in previous work\cite{us}, we expand out $F(\{\sigma_i\})$,
keeping only the lowest order cumulant. Neglecting constant terms, this gives
\begin{equation}
F_o(\{\sigma_i\}) \approx {B\over 2} \sum_{i,j}^N
\sigma_i\sigma_j\langle\Delta({\bf r}_i-{\bf r}_j)\rangle ,
\label{eq:F_o}
\end{equation}
with the average done over all compact conformations with minimal
surface area. Note that this term gives an anti-ferromagnetic Ising
interaction. Thus we wish to minimize
\begin{equation}
\Delta F \approx {B\over 2} \sum_{i,j}^N
[\Delta({\bf r}_i-{\bf r}_j) - \langle\Delta({\bf r}_i-{\bf r}_j)\rangle]
\sigma_i\sigma_j
\label{eq:approx}
\end{equation}
as a function of the $\{ \sigma_i\}$.
The first work on this model, by Shakhnovich and Gutin, used as
stated earlier, energy minimization. To keep the chain from becoming
a homopolymer, they added the constraint of constant magnetization.
It has met with some degree of success and we have
previously attempted\cite{us} to
analyze why that should be in the systematic framework
that we have recently developed.
This method can be interpreted\cite{us} as
an approximation to eqn. (\ref{eq:approx}). If we approximate
$\langle\Delta({\bf r}_i-{\bf r}_j)\rangle$ to be a constant, then
$F_o$ is equivalent to an anti-ferromagnetic mean field interaction
proportional to the total magnetization squared. This has the effect
of preferring a total magnetization close to zero, and is therefore similar
to the constraint of a fixed magnetization used by Shakhnovich and Gutin.
It has further come to our attention that Shakhnovich and Gutin ignored
the interactions along the backbone\cite{private}.
It is clear that in designing sequences, the energy of interaction along
the backbone of the chain cannot matter, since it does not depend on the
conformation. This is therefore a sensible design procedure.
This is equivalent
to assuming that $\langle\Delta({\bf r}_i-{\bf r}_j)\rangle$ is unity, for
adjacent $i$ and $j$. For all other $i$ and $j$, it is the same as above,
similar to taking it to be a smaller constant. This is a
better approximation to $\langle\Delta({\bf r}_i-{\bf r}_j)\rangle$
but is still fairly crude.
Notice that the contribution along the backbone of the chain naturally
cancels in (\ref{eq:approx}); thus with our method there is no need to
ignore backbone interactions.
Shakhnovich and Gutin\cite{SG} have noted that not only do we want to design
sequences which have a low ground state degeneracy, but also we want
sequences which have a significant energy gap between the native state
and higher energy states\cite{SG}. This is necessary in order to get
stable ground state structures. After designing sequences by minimizing
the energy in sequence space, they noted that their designed sequences
tended to have more of an energy gap than randomly chosen sequences.
However, they did not decide to optimize this gap in their search of
sequence space. As noted above, minimizing the exact $\Delta F$ will give
maximally stable structures. Furthermore, in the above
approximation to $\Delta F$,
that is $\Delta F = E(\{r^0_i\}) - \langle E(\{r_i\})\rangle$, it can
also be seen to be similar to the requirement of a large energy gap.
$E(\{r^0_i\})$ is the minimum energy conformation and $\langle
E(\{r_i\})\rangle$ is the average over all conformations. Maximizing the
difference between the minimum and the average is likely to result in a
large energy gap.
\section{Numerical Results}
\label{sec:results}
For a given desired configuration, we minimized (\ref{eq:approx}) in
sequence space using simulated annealing in order to find the
sequence most likely to have the desired configuration as its ground state.
The second term of (\ref{eq:approx}) was calculated
by directly enumerating all possible configurations.
We carried out the above procedure
on 2066 randomly picked initial configurations out of the total
number of 103346 distinct configurations of the 27-mer cube. We then
folded the resulting molecules by minimizing the energy in configuration
space. It was found that each of the resulting molecules did indeed
have the desired configuration as its ground state.
In addition, the average
degeneracy of these ground states is only 3.37. Thus
the molecules we designed are quite likely to fold into the
desired conformations.
We now compare these results with results we have obtained
by a constrained minimization of the energy
instead of minimizing our new function $\Delta F$.
We performed the same simulated annealing procedure for the same
2066 given conformations, this time minimizing the energy (in sequence
space), keeping the total magnetization equal to one. This
last constraint is necessary to keep the chain from becoming a
homopolymer, and is in keeping with work done by Shakhnovich and Gutin\cite{SG}.
Although the sequences
found to minimize the energy can be proved to always
have the desired conformations
as ground states, the average degeneracy of these ``best" sequences is 1155,
averaged over 2066 random conformations! Clearly these
sequences are not likely to be found in the desired conformations.
We also wanted to compare our results with the method of minimizing
the energy in sequence space while ignoring interactions along the
backbone of the chain. As discussed above, this is quite similar
to minimizing $\Delta F$, though the results are not as good.
The sequences produced this
way were usually found to be ground states of the desired conformation, but
they have an average ground state degeneracy of 4.11. We also find that
16 of the conformations were misdesigned as they do not fold
up into the desired structure.
In addition, we examined the 2D HP model of Lau and Dill\cite{HP}. We
looked at 16-mers on an open 2D lattice with two types of monomers,
labeled hydrophobic (H) and polar (P), with an energy of -1 assigned to
pairs of hydrophobes that are nearest neighbors. HP and PP interactions
are both zero. In this model
the contribution to the energy from pairs along the backbone of the
chain is generally ignored.
We first folded up all possible sequences and from that found that
584 conformations were ``good" structures.
In other words, for each of these structures it was possible to
find at least one sequence which would fold into it
as a unique ground state.
We then went through all the ``good" structures
and tried to design each one, using the method of Shakhnovich
and Gutin, and by using our new method.
For the method of Shakhnovich and Gutin, 111 structures were
missed. That is, their method failed to design these sequences
to fold into the target structure. Of the remaining ones, only 123
folded correctly to a unique ground state.
This represents an overall success rate of 21\%.
The new method fared much better.
We employed an approximation analogous to eqn. (\ref{eq:F_o}) in
estimating $F_o$. For this we had to estimate the
monomer-monomer correlation function
$\langle\Delta({\bf r}_i-{\bf r}_j)\rangle$. We did this by
defining the average $\langle \dots \rangle$ over a set
of compact conformations. We defined a compact conformation as
having 7 or more contacts. Reducing this number to 5 contacts
makes little difference to our results. Only 22 out of 584 structures
were missed and 295 folded correctly to a unique ground state.
This is an overall success rate of over 50\%.
\section{Conclusions}
In conclusion, we have demonstrated for two simple models that we
have found a method far superior to that previously used in designing
sequences to fold to a desired structure. It is a
cumulant expansion approximation to
$\Delta F$, the difference between clamped and unclamped free energies.
It is superior to energy minimization in several ways.
First, it designs sequences that correctly fold into
the desired structure, more often than energy minimization.
Second, the sequences tend to give a lower ground state degeneracy.
Third, minimizing $\Delta F$, by construction should design maximally
stable conformations.
However the cumulant approximation used in this work is still
not perfect. As stated above for the HP model, our method
designs unique ground states in only 50\% of the conformations
that actually have unique ground states.
Further work on improving this method is still essential
in perfecting protein design.
One exciting new approach to the reverse folding problem
has recently been proposed for the HP model~\cite{UCSF}.
A novel minimization function has been proposed that is
quite different than our $\Delta F$. It also
involves adding an additional term to the energy,
but its form its quite unlike $F_o$. Further
work connecting these two approaches is in progress.
\section{Acknowledgments}
One of us (T.K.) wishes to thank Hemant Bokil
for useful discussions. J.M.D. would like to thank
Ken Dill and Hue Sun Chan for very valuable discussions.
This work is supported by NSF grant number DMR-9419362
and acknowledgment
is made to the Donors of the Petroleum Research Fund, administered
by the American Chemical Society for partial support of this research.
\newpage
|
\section{Introduction}
Asymmetric orbifolds \cite{NSV} belong to the efficient construction schemes
of 4D string models.
In constructing asymmetric orbifolds, we consider the left-right asymmetric
twists of the momentum lattices of toroidal compactifications.
Thus, we have in general no direct geometric interpretation of asymmetric
orbifolds.
However, some class of asymmetric orbifolds may possess the geometric
interpretation through the quantum equivalence to other compactifications.
In the previous paper \cite{Fermion}, we have studied the geometric
interpretation of asymmetric orbifolds as Narain's toroidal
compactifications~\cite{NSW} and found a simple condition for the geometric
interpretation of asymmetric orbifolds.
In this paper we construct the asymmetric orbifold models which are
based on the non-supersymmetric heterotic strings \cite{HETERO},
and investigate the geometric interpretation of such asymmetric orbifolds as
the toroidal compactification \cite{TORUS} of non-supersymmetric heterotic
strings.
Asymmetric orbifold models of non-supersymmetric heterotic strings have not
been studied before.
The main purpose of this paper is to extend the previous analysis by studying
the geometric interpretation of the {\em non-supersymmetric} asymmetric
orbifold models.
We classify the asymmetric orbifold models with standard
embeddings in a systematic way and give a list of the asymmetric orbifold
models which are geometrically interpreted as toroidal compactifications
of non-supersymmetric heterotic strings.
We shall study the condition for interpreting orbifold models as toroidally
compactified ones.
A crucial feature of ten-dimensional heterotic strings is the existence
of NSR fermions.
They will form an $SO(8)$ Ka\v{c}-Moody\ algebra in the light-cone gauge.
Since the compactification onto tori yields no effect on the NSR fermions,
the fermions on orbifolds should generate $SO(8)$ Ka\v{c}-Moody\ algabras if they are
interpreted as toroidal compactifications.
We define fermion currents as the current operators with conformal weight one
which generate $SO(N)$ algebra and which contain the $SO(2)$ Ka\v{c}-Moody\ algebra
generated by the space-time component of the fermions on orbifold models.
A necessary condition for interpreting orbifold models as toroidally
compactified ones is that the fermion currrents should generate $SO(8)$ Ka\v{c}-Moody\
algebras.
We shall show that the above condition is necessary and also sufficient
for our class of asymmetric orbifolds.
We should note that symmetric orbifolds cannot be interpreted as toroidal
compactifications since fermion currents on symmetric orbifolds are
constructed by the currents in the untwisted sector which are always smaller
than $SO(8)$.
Orbifold models must possess twist-untwist intertwining currents
\cite{CHDGM,ISST} in order to possess $SO(8)$ fermion currents.
Thereby, orbifold models must be asymmetric if they are interpreted as
toroidal compactifications.
The above asymmetric orbifold models with standard embeddings possess no
supersymmetry.
At first sight, any orbifold model constructed from non-supersymmetric
heterotic
strings seems to possess no supersymmetry.
However, this will turn out not to be true for generic asymmetric orbifold
models.
In fact, for asymmetric orbifold models of supersymmetric heterotic strings
supersymmetry can appear from the twisted sectors~\cite{ISST,I}.
We may expect that the similar mechanism still holds for some classes of the
asymmetric orbifolds constructed from non-supersymmetric heterotic strings.
In this paper, by studying non-standard embedding models, we shall show some
examples of {\em supersymmetric} asymmetric orbifold models
constructed from non-supersymmetric heterotic strings.
The organization of this paper is as follows.
In section 2, we explain the construction of the asymmetric orbifolds from
non-supersymmetric heterotic strings.
In section 3, we clarify a necessary and sufficient condition for the
asymmetric orbifold models to be interpreted as toroidally compactified
models.
In section 4, we classify the asymmetric orbifold models and investigate
fermion currents on them.
In section 5, we construct the examples of supersymmetric asymmetric
orbifolds of non-supersymmetric heterotic strings.
In section 6, we present our conclusion.
\section{Asymmetric orbifold models}
We shall recall the basic set up of the non-supersymmetric
ten-dimensional heterotic strings \cite{HETERO}.
The non-supersymmetric heterotic strings with rank 16 gauge groups
are specified by the left-moving 16-dimensional conjugacy classes
$\Gamma_{i}^{16,0}$ $(i = 1, \dots, 4)$.
The conjugacy class $\Gamma_{1}^{16,0}$ $(\Gamma_{2}^{16,0})$ is
connected to the NS sector states with the even (odd) G-parity.
The conjugacy class $\Gamma_{3}^{16,0}$ $(\Gamma_{4}^{16,0})$ is
connected to the R sector states with the positive (negative) chirality.
For definiteness, we will concentrate on the tachyon-free non-supersymmetric
heterotic strings with $SO(16) \times SO(16)$ gauge groups.
Then $\Gamma_{i}^{16,0}$ $(i = 1, \dots, 4)$ are given by
$\Gamma_{1}^{16,0} = (0, 0) \cup (c, c)$,
$\Gamma_{2}^{16,0} = (s, v) \cup (v, s)$,
$\Gamma_{3}^{16,0} = (v, v) \cup (s, s)$ and
$\Gamma_{4}^{16,0} = (c, 0) \cup (0, c)$,
where $0$, $v$, $s$ and $c$ are the adjoint, vector, spinor and conjugate
spinor conjugacy classes of $SO(16)$, respectively.
Almost arguments in this paper will be straightforwardly applied to the other
non-supersymmetric heterotic strings.
Let us now construct asymmetric ${\bf Z}_N$ -orbifolds of the non-supersymmetric
heterotic string.
We shall start with the toroidal compactification \cite{TORUS} of the
non-supersymmetric heterotic string defined by the conjugacy classes
$\Gamma_{i}^{16,0} \oplus \Gamma^{6,6}$ $(i = 1, \dots, 4)$, where
$\Gamma^{6,6}$ is a (6+6)-dimensional Lorentzian even self-dual lattice.
The left- and right-moving momentum $(p_L^I, p_L^i, p_R^i)$
$(I = 1, \dots, 16; i = 1, \dots, 6)$ lies on the conjugacy classes
$\Gamma_{i}^{16,0} \oplus \Gamma^{6,6}$.
The group element~$g$ which generates a cyclic group ${\bf Z}_N$\ is defined by the
following action on the string coordinates:
\begin{equation}
g: (X_L^I, X_L^i, X_R^i) \rightarrow (X_L^I + 2\pi v_L^I, U_L^{ij} X_L^j,
U_R^{ij} X_R^j),
\end{equation}
where $U_L$ and $U_R$ are rotation matrices which satisfy
$U_L^N = U_R^N = {\bf 1}$ and $v_L^I$ is a shift vector.
The rotation matrices $U_L$ and $U_R$ must be an automorphism of
$\Gamma^{6,6}$:
\begin{equation}
(U_L^{ij} p_L^j, U_R^{ij} p_R^j) \in \Gamma^{6,6} \ \ {\rm for \ all} \
(p_L^i,p_R^i) \in \Gamma^{6,6}.
\end{equation}
The action of the operator $g$ on the right-moving fermions is given by the
$U_R$ rotation.
Let $N_{\ell}$ be the minimum positive integer such that
$(g^{\ell})^{N_{\ell}} = 1$ in the $g^{\ell}$-twisted sector.
We shall denote the eigenvalues of $U_L^{\ell}$ and $U_R$ by
$ \{ e^{i 2\pi \zeta_{\ell}^a}, e^{-i 2\pi \zeta_{\ell}^a} ; a=1,2,3 \} $
and
$ \{ e^{i 2\pi \zeta_R^a}, e^{-i 2\pi \zeta_R^a}; a=1,2,3 \} $,
respectively.
Then we have the level matching condition for the one-loop modular invariance
for $N_{\ell}$ odd
\begin{equation}
N_{\ell}
\left[
\frac{1}{2} (\ell v_L^I)^2
+ \frac{1}{2} \sum_{a=1}^{3} \zeta_{\ell}^a (1-\zeta_{\ell}^a)
\right]
= 0 \ \ \bmod 1,
\end{equation}
\begin{equation}
N_{\ell} \sum_{a=1}^3 \ell \zeta_R^a = 0 \ \ \bmod 2;
\end{equation}
for $N_{\ell}$ even, in addition to the above conditions, we have
\begin{equation}
p_L^i ( U_L^{\ell\frac{N_{\ell}}{2}} )^{ij} p_L^j
- p_R^i ( U_R^{\ell\frac{N_{\ell}}{2}} )^{ij} p_R^j = 0
\ \ \bmod 2
\end{equation}
for all $(p_L^i,p_R^i) \in \Gamma^{6,6}$.
\section{Torus-orbifold equivalence of heterotic strings}
We now show that the necessary condition for the equivalence with torus
compactifications in terms of the fermion currents is also a sufficient
condition for our class of asymmetric orbifold models.
As we shall see in the next section, the ${\bf Z}_N$ -transformation of our
asymmetric orbifold model is an inner automorphism of the momentum lattice.
Then, the ${\bf Z}_N$ -transformation on the lattice $\Gamma^{6,6}$ is equivalent to
a shift \cite{NSV,ISST,DHVW,S}.
We shall use the bosonized representation of world-sheet
fermions.
The momentum $p_R^t$ $(t = 1, \dots, 4)$ of the fermions lies on the weight
lattice of $SO(8)$.
The momentum in the vector (adjoint) conjugacy class corresponds to the state
in the NS sector with even (odd) G-parity.
The momentum in the spinor (conjugate spinor) conjugacy class corresponds to
the state in the R sector with positive (negative) chirality.
The ${\bf Z}_N$ -transformation $g$ acts on the bosons as a shift, where the
shift vector is given by $v_R^t = (\zeta_R^a, 0)$.
Let us denote the conjugacy classes of $SO(2n)$ as $(i)_n$
$(i = 0, v, s, c)$.
The momentum
$(p_L^{\prime I}, p_L^{\prime i}, p_R^{\prime i}, p_R^{\prime t})$
$(I = 1, \dots, 16; i = 1, \dots, 6; t = 1, \dots, 4)$
in the $g^{\ell}$-sector
($\ell = 0$ for untwisted sector and $\ell = 1, \dots, N-1$ for twisted
sectors)
of the asymmetric orbifolds lies on the following lattice:
\begin{equation}
[
( \Gamma_{1}^{16,0} \oplus \Gamma^{6,6} \oplus (v)_4 )
\cup
( \Gamma_{2}^{16,0} \oplus \Gamma^{6,6} \oplus (0)_4 )
]
+ \ell (v_L^I, v_L^i, v_R^i, v_R^t),
\end{equation}
for NS sector, and
\begin{equation}
[
( \Gamma_{3}^{16,0} \oplus \Gamma^{6,6} \oplus (s)_4 )
\cup
( \Gamma_{4}^{16,0} \oplus \Gamma^{6,6} \oplus (c)_4 )
]
+ \ell (v_L^I, v_L^i, v_R^i, v_R^t),
\end{equation}
for R sector.
The operator $g$ in the $g^{\ell}$-sector
will be expressed as
\begin{equation}
g = \eta_{\ell} \exp [ i 2\pi (p_L^{\prime I} v_L^I + p_L^{\prime i}
v_L^i - p_R^{\prime i} v_R^i - p_R^{\prime t} v_R^t ) ],
\end{equation}
where $\eta_{\ell}$ is a constant phase and $(v_L^i, v_R^i)$ is a
shift vector which satisfy $N (v_L^i, v_R^i) \in \Gamma^{6,6}$.
The phase $\eta_{\ell}$ is determined from the modular transformations
\cite{ISST}:
\begin{equation}
\eta_{\ell}
= \exp \{ -i \pi \ell [ (v_L^I)^2 + (v_L^i)^2 - (v_R^i)^2 - (v_R^t)^2 ] \}.
\end{equation}
Every physical state in the $g^{\ell}$-sector must satisfy
the condition $g = 1$ because it must be invariant under the
${\bf Z}_N$ -transformation.
In order to examine Ka\v{c}-Moody\ algebras on heterotic string models, it
may be convenient to use the bosonic string map \cite{LSW} and
investigate Ka\v{c}-Moody\ algebras on the corresponding bosonic string
models.
We first decompose the momentum lattices of the asymmetric orbifolds with
respect to $SO(2)$ conjugacy classes to which the momentum $p_R^{\prime t}$
$( t=4 )$ belongs.
Then, preserving the modular transformation properties, the $SO(2)$ conjugacy
classes $(i)_1$ $(i = 0, v, s, c)$ are mapped to the $SO(10) \times E_8$
conjugacy classes $(i)_5 \oplus \Gamma^{0,8}$ $(i = 0, v, s, c)$ as follows:
$(0)_1 \rightarrow (v)_5 \oplus \Gamma^{0,8}$,
$(v)_1 \rightarrow (0)_5 \oplus \Gamma^{0,8}$,
$(s)_1 \rightarrow (s)_5 \oplus \Gamma^{0,8}$ and
$(c)_1 \rightarrow (c)_5 \oplus \Gamma^{0,8}$,
where $\Gamma^{0,8}$ is a root lattice of $E_8$.
After the bosonic string map, the momentum
$(p_L^{\prime I}, p_L^{\prime i}, p_R^{\prime i}, p_R^{\prime t})$
in the $g^{\ell}$-sector lies on the following lattice:
\begin{eqnarray}
&&
[
( \Gamma_{1}^{16,0} \oplus \Gamma^{6,6} \oplus (0)_8 \oplus \Gamma^{0,8} )
\cup
( \Gamma_{2}^{16,0} \oplus \Gamma^{6,6} \oplus (v)_8 \oplus \Gamma^{0,8} )
\\
&\cup&
( \Gamma_{3}^{16,0} \oplus \Gamma^{6,6} \oplus (s)_8 \oplus \Gamma^{0,8} )
\cup
( \Gamma_{4}^{16,0} \oplus \Gamma^{6,6} \oplus (c)_8 \oplus \Gamma^{0,8} )
]
+ \ell (v_L^I, v_L^i, v_R^i, v_R^t),
\nonumber
\end{eqnarray}
where $p_R^{\prime t}$ $( t=4, \dots, 8; 9, \dots, 16)$ are defined as the
momentum which belong to the $SO(10) \times E_8$ conjugacy classes and
$v_R^t = (\zeta_R^a, 0^5; 0^8)$.
Therefore, on the corresponding bosonic strings, the momentum
$(p_L^{\prime I}, p_L^{\prime i}, p_R^{\prime i}, p_R^{\prime t})$
of the physical states in the $g^{\ell}$-sector lies
on a $(22+22)$-dimensional Lorentzian even self-dual lattice
$\Gamma^{\prime 22,22}$, where
$(p_L^{\prime I}, p_L^{\prime i}, p_R^{\prime i}, p_R^{\prime t})$
lies on the above lattice and satisfies the physical state condition:
\begin{equation}
p_L^{\prime I} v_L^I + p_L^{\prime i} v_L^{\prime i}
- p_R^{\prime i} v_R^{\prime i} - p_R^{\prime t} v_R^t
- \frac{1}{2} \ell [ (v_L^I)^2 + (v_L^i)^2
- (v_R^i)^2 - (v_R^t)^2 ] = 0 \ \ \bmod 1.
\end{equation}
If the fermion currents on asymmetric orbifolds generate
$SO(8)$ Ka\v{c}-Moody\ algebras, then the corresponding $(22+22)$-dimensional
lattice $\Gamma^{\prime 22,22}$ is decomposed as follows:
\begin{equation}
\Gamma^{\prime 22,22} =
[
( \Gamma_{1}^{\prime 22,6} \oplus (0)_8 )
\cup
( \Gamma_{2}^{\prime 22,6} \oplus (v)_8 )
\cup
( \Gamma_{3}^{\prime 22,6} \oplus (s)_8 )
\cup
( \Gamma_{4}^{\prime 22,6} \oplus (c)_8 )
]
\oplus \Gamma^{0,8},
\end{equation}
where $\Gamma_{i}^{\prime 22,6}$ $(i = 1, \dots, 4)$ are
$(22+6)$-dimensional conjugacy classes.
This implies that, after reversing the bosonic string map,
we obtain the toroidal compactifications of the ten-dimensional
non-supersymmetric heterotic strings.
\section{Classification of asymmetric orbifold models}
Let us discuss the classification of asymmetric orbifold models.
Since we are investigating the geometric interpretation of asymmetric
orbifold models as toroidal compactifications, we classify the asymmetric
orbifolds with the right-moving twist-untwist intertwining currents.
We first consider the choice of the momentum lattices.
Unlike symmetric orbifolds models, the momentum lattices $\Gamma^{6,6}$ are
severely restricted by the left-right asymmetric automorphisms.
One of the known classes of such momentum lattices are given by
\begin{equation}
\Gamma^{6,6} = \{(p_L^i, p_R^i) \vert
p_L^i, p_R^i \in \Lambda_W \ {\rm and}\ p_L^i - p_R^i \in \Lambda_R\},
\end{equation}
where $\Lambda_W$ and $\Lambda_R$ are the weight and root lattices of a
simply-laced semisimple Lie algebra with the squared length of roots
normalized to two \cite{EN}.
The left- and right-moving rotation matrices $U_L$ and $U_R$ are taken to be
the Weyl group elements \cite{C} of the Lie algebra.
Then the matrices $U_L$ and $U_R$ always satisfy the condition
for the automorphism of $\Gamma^{6,6}$.
Next, we consider the standard embeddings in the gauge degrees of freedom.
Since the gauge group of the tachyon free non-supersymmetric strings is
$SO(16) \times SO(16)$, we can embed the shift $v_L^I$ in the $SO(6)$
subgroup of the first $SO(16)$.
Let us denote the eigenvalues of $U_L$ by $ \{ e^{i 2\pi \zeta_L^a},e^{-i
2\pi \zeta_L^a} ; a=1,2,3 \} $.
Then the shift vector $v_L^I$ is chosen as
\begin{equation}
v_L^I = (\zeta_L^a, 0^5; 0^8).
\end{equation}
By the above choice of the shift vector, the first equation of the level
matching conditions reduces to
\begin{equation}
N_{\ell} \sum_{a=1}^3 \ell \zeta_L^a = 0 \ \ \bmod 2.
\end{equation}
With these level matching conditions, we classify the modular invariant
asymmetric orbifold models.
To determine fermion currents on asymmetric orbifolds, we use
the bosonic string map.
We must investigate the full right-moving Ka\v{c}-Moody\ algebras of
asymmetric orbifolds.
This is because, unlike symmetric orbifold case \cite{HMKKKOOT},
there is no simple diagrammatical method for determining Ka\v{c}-Moody\
algebras if there exist twist-untwist intertwining currents.
The results of calculation are summarized in table 1.
We present in table 1 the right-moving Ka\v{c}-Moody\ algebras
and the fermion currents of the asymmetric orbifold models.
As we have seen above, the necessary and sufficient condition for the
asymmetric orbifold models to be interpreted as toroidal compactifications of
non-supersymmetric heterotic strings is that the fermion currents on the
asymmetric orbifolds should generate $SO(8)$ Ka\v{c}-Moody\ algebras.
In table 1, we see that many asymmetric orbifold models are geometrically
interpreted as toroidal compactifications of non-supersymmetric heterotic
strings.
\section{Supersymmetric asymmetric orbifold models}
We will present examples of the supersymmetric asymmetric orbifold models
constructed from non-supersymmetric heterotic strings.
We first consider the condition for obtaining massless gravitino states in
the
$g^{\ell}$-sector $(\ell = 0, 1, \dots, N-1)$ of asymmetric orbifold models.
The existence of a massless gravitino state will lead to the existence of a
supersymmetry.
Let us define the $g^{\ell}$-invariant sublattice $I_{\ell}$ of
$\Gamma^{6,6}$ by
\begin{equation}
I_{\ell} = \{ (p^i_L,p^i_R) \in {\Gamma^{6,6}}\vert
( (U_L^{\ell})^{ij}p_L^j, (U_R^{\ell})^{ij}p_R^j ) = ( p_L^i, p_R^i ) \}.
\end{equation}
The momentum $(p_L^i, p_R^i)$ of the $g^{\ell}$-twisted sector lies on the
lattice $I_{\ell}^{\ast} $, where $I_{\ell}^{\ast} $ is the dual
lattice of $I_{\ell}$.
The number of degeneracy of the ground states in the $g^{\ell}$-twisted
sector is given by
\begin{equation}
n_{\ell} =
\frac{
\sqrt{ \det^{\prime} ({\bf 1} - U_L^{\ell})
\det^{\prime} ({\bf 1} - U_R^{\ell}) }
}{
{\rm vol}(I_{\ell})
},
\end{equation}
where $\det^{\prime}$ is evaluated over the eigenvalues of $U_L^{\ell}$ and
$U_R^{\ell}$ not equal to one and ${\rm vol}(I_{\ell})$ is the volume of
the unit cell of the lattice $I_{\ell}$.
We shall denote the eigenvalues of $U_R^{\ell}$ by
$ \{ e^{i 2\pi {\bar \zeta}_{\ell}^a}, e^{-i 2\pi {\bar \zeta}_{\ell}^a};
a=1,2,3 \} $.
The mass formula in the $g^{\ell}$-sector is given by
\begin{eqnarray}
\frac{1}{8} m_L^2
& = &
\frac{1}{2} \sum_{I=1}^{16} (p_L^I + \ell v_L^I)^2
+ \frac{1}{2} \sum_{i=1}^{6} (p_L^i)^2
+ \frac{1}{2} \sum_{a=1}^{3} \zeta_{\ell}^a (1-\zeta_{\ell}^a)
+ N_L
- 1,
\\
\frac{1}{8} m_R^2
& = & \frac{1}{2} \sum_{i=1}^{6} (p_R^i)^2
+ \frac{1}{2} \sum_{t=1}^{4} (p_R^t + \ell v_R^t)^2
+ \frac{1}{2} \sum_{a=1}^{3} {\bar \zeta}_{\ell}^a (1-{\bar
\zeta}_{\ell}^a)
+ N_R
- \frac{1}{2},
\end{eqnarray}
where $(p_L^i, p_R^i) \in I_{\ell}^{\ast}$,
$(p_L^I, p_R^t) \in
( \Gamma_{1}^{16,0}\oplus (v)_4 )
\cup
( \Gamma_{2}^{16,0}\oplus (0)_4 )
\cup
( \Gamma_{3}^{16,0}\oplus (s)_4 )
\cup
( \Gamma_{4}^{16,0}\oplus (c)_4 )$, and
$N_L$ and $N_R$ are the number operators of oscillators.
In order to have massless spin 3/2 states, we must have
$p_R^{t = 4} = \pm 1/2$, $N_L=1$, $U_L^{\ell} = {\bf 1}$ and
$p_L^{\prime I} \equiv p_L^I + \ell v_L^I = 0$ for
$p_L^I \in \Gamma^{16,0}_{3} \cup \Gamma^{16,0}_{4}$,
where the condition $N_L=1$ must be satisfied by the oscillators of the
space-time coordinates.
It should be noted that
$p_L^{\prime I} = 0$ for
$p_L^I \in \Gamma^{16,0}_{3} \cup \Gamma^{16,0}_{4}$
never holds if we set
$\ell = 0$ (untwisted sector) or $v_L^I = 0$ (no embedding).
Let us denote the order of the left-moving shift (twist) as $N_S$
($N_L$).
Then, from the mass formula we can check that there is no massless gravitino
state for
the orbifold models with $N_S = N_L$.
Examples of the models satisfying such a condition are the orbifold models
with standard embeddings and symmetric orbifolds.
Thus, we have to consider the asymmetric orbifold models with
non-standard embeddings in order to obtain massless gravitino states.
We now construct examples of supersymmetric asymmetric orbifold models.
We start with a toroidal compactification of $SO(16)\times SO(16)$
non-supersymmetric heterotic strings, where the momentum lattice
$\Gamma^{6,6}$ is associated with a Lie algebra $(SU(3))^3$.
The left- and right-twists of asymmetric orbifolds are taken to be the
${\bf Z}_3$-twist matrix $U$ whose eigenvalues are given by
$\{e^{2 \pi i \zeta^a}, e^{-2 \pi i \zeta^a}; a = 1, 2, 3\}$ with
$\zeta^a = (1/3, 1/3, 2/3)$.
We take the shift vector to be $v_L^I \in \Gamma_{3}^{16,0}$.
For example, such shift vectors are given by
$v_L^I = (1, 0^7; 1, 0^7)$ and $v_L^I = ((1/2)^8; (1/2)^8)$.
The order of the shift is given by $N_S = 2$, and the level matching
conditions are satisfied since $(v_L^I)^2 = 0$ $\bmod$ $2$.
The first example is the asymmetric ${\bf Z}_6$-orbifold model with
$U_L = {\bf 1}$ and $U_R = U$.
Massless gravitino states may appear from $\ell = 1, 3, 5$ twisted sectors
since
the solution of $p_L^{\prime I} = 0$ exists for
$\ell = 1, 3, 5$ if we set $p_L^I \in \Gamma_{3}^{16,0}$.
In the $\ell = 1$ sector, from the massless condition we obtain
$p_R^t = (-1/2, -1/2, -1/2, -1/2)$.
The degeneracy of the ground states is given by $n_1 = 1$.
This state is physical since all massless states in the first twisted sector
are known to be physical \cite{FIQS}.
In the $\ell = 3$ sector, we obtain
$p_R^{\prime t} = (\pm 1/2, \pm 1/2, \pm 1/2, \pm 1/2)$,
where the number of minus signs should be even.
Since $U_L^3 = U_R^3 = {\bf 1}$, the $g$ operator in this sector will be
given by $g = \exp [ p_L^{\prime I} v_L^I - p_R^{\prime t} v_R^t
-\frac{1}{2} \ell ( (v_L^I)^2 - (v_R^t)^2) ]$.
Thus, physical (i.e. $g = 1$) states are given by
$p_R^{\prime t} = \pm (1/2, 1/2, -1/2, -1/2)$.
In the $\ell = 5$ sector, we obtain $p_R^t = (-3/2, -3/2, -7/2, 1/2)$.
The degeneracy of the ground states is given by $n_5 = 1$.
This state is physical since all massless states in the
$g^5$-twisted sector are physical.
The existence of these massless states will lead to the existence of two
gravitino states.
Therefore, in this model we have $N=2$ supersymmetry.
The second example is the asymmetric ${\bf Z}_6$-orbifold model with
$U_L = U$ and $U_R = {\bf 1}$.
{}From the mass formula, we see that gravitino states may appear from
the $\ell =3$ twisted sector.
In the $\ell = 3$ sector, we obtain
$p_R^{\prime t} = (\pm 1/2, \pm 1/2, \pm 1/2, \pm 1/2)$, where the number of
minus signs should be even.
The $g$ operator takes the value $g = 1$ for all such massless states.
Thus, the above states are all physical.
The existence of these massless states will lead to the existence of four
gravitino states.
Therefore, in this model we have $N=4$ supersymmetry.
This model will possess the geometric interpretation as a Narain's toroidal
compactification by the mechanism discussed in the previous
paper~\cite{Fermion}.
The last example is the asymmetric ${\bf Z}_6$-orbifold model with
$U_L = U_R = U$.
We see that gravitino states may appear from the $\ell =3$ twisted sector.
In the $\ell = 3$ sector, we obtain
$p_R^{\prime t} = (\pm 1/2, \pm 1/2, \pm 1/2, \pm 1/2)$, where the number of
minus signs should be even.
The $g$ operator will be given by
$g = \exp [ p_L^{\prime I} v_L^I - p_R^{\prime t} v_R^t
-\frac{1}{2} \ell ( (v_L^I)^2 - (v_R^t)^2) ]$.
Then physical states are given by
$p_R^{\prime t} = \pm (1/2, 1/2, -1/2, -1/2)$.
The existence of these massless states will lead to the existence of one
gravitino state.
Therefore, in this model we have $N=1$ supersymmetry.
\section{Conclusion}
In this paper we have constructed asymmetric orbifold models which are based
on non-supersymmetric heterotic strings.
We have shown that a simple condition in terms of fermion currents is
necessary and sufficient for our class of asymmetric orbifolds to be
interpreted as the toroidal compactifications of non-supersymmetric heterotic
strings.
We have made a systematic classification of the asymmetric ${\bf Z}_N$ -orbifold
models with standard embeddings and obtained many asymmetric orbifold models
which are geometrically interpreted as toroidal compactifications of
non-supersymmetric strings.
We have also discussed the supersymmetric asymmetric orbifold
models constructed from non-supersymmetric heterotic strings.
We have investigated the condition for obtaining supersymmetric ${\bf Z}_N$ -orbifold
models from non-supersymmetric heterotic strings
and have shown that in order to obtain supersymmetric models we must consider
asymmetric orbifold models with non-standard embeddings.
We have constructed the examples of the $N=1$, $N=2$ and $N=4$ asymmetric
orbifold models of non-supersymmetric heterotic strings.
It would be of interest to investigate other examples of such supersymmetric
models constructed from non-supersymmetric heterotic strings.
\bigskip
\begin{center}
{\large\bf Acknowledgements}
\end{center}
We would like to thank M. Sakamoto for reading the manuscript and useful
comments.
\newpage
|
\section{Introduction}\label{introduction}
In \cite{heindorf-shapiro}, a \Ba\ $A$ is said to have the Freese-Nation
property
(FN, for short) if there exists an FN-mapping
on $A$, i.e.\ a function $\mapping{f}{A}{[A]^{<\aleph_0}}$ \st\
\begin{assertion}{$(*)$}
\it if $a,b\in A$ satisfy $a\leq b$, then $a\leq c\leq b$ for some
$c\in f(a)\cap f(b)$.
\end{assertion}
This property is closely connected to the notion of freeness because of the
following
facts: \assertof{a} {\it every free \Ba\ $A$ has the FN}\/; to see
this, fix a subset $U$ of $A$ generating $A$ freely; then for $b\in A$, let
$u(b)$ be a finite subset of $U$ generating $b$ and $f(b)$ the finite
subalgebra of $A$ generated by $u(b)$. The Interpolation Theorem of
propositional logic
then tells us that $(*)$ holds for this $f$. Moreover,
we have: \assertof{b}
{\it if $A$ has the FN, then so has every retract of $A$}\/ (see
\Lemmaof{retract} below for a more general statement). From \assertof{a} and
\assertof{b}, it follows that: \assertof{c}
{\it every projective \Ba\ has the FN.}
Historically, the FN was first considered by R.\ Freese and
J.B.\ Nation in their paper \cite{freese-nation} which gives a
characterization of projective lattices. In particular, they proved that
every projective lattice has the FN.
The FN alone, however, is not equivalent to projectiveness, since, as
Heindorf proved in \cite{heindorf-shapiro}, a \Ba\ $A$ has the FN
if and only if $A$ is openly generated in the terminology given below (which
is also used in \cite{fuchino2}; in \cite{heindorf-shapiro} these
\Bas\ are called ``rc-filtered''). The notion of open generatedness
was introduced originally in a topological setting by \v{S}\v{c}epin
\cite{scepin}.
In the language of \Bas, a \Ba\ $A$ is said to be openly generated if
there exists a closed unbounded subset $\calC$ of $[A]^{\aleph_0}$ \st\
every $C\in\calC$ is a relatively complete subalgebra of $A$. \v{S}\v{c}epin
found examples of openly generated \Bas\ which are not projective.
In this paper, we continue the study of \Bas\ with the following weakening of
the Freese-Nation property, begun in \cite{heindorf-shapiro} or, to
some extent, already in \cite{scepin}: a \Ba\ $A$ is said to have
the {\em weak Freese-Nation property} ({\em WFN}\/ for short) if there is a
{\em WFN mapping on $A$}, that is, a mapping
$\mapping{f}{A}{[A]^{\leq\aleph_0}}$ satisfying the condition $(*)$ above.
We solve some open
problems from \cite{heindorf-shapiro} in Sections \ref{card-functions} and
\ref{Intalg}.
Clearly the WFN makes
perfect sense for arbitrary partial orderings and can be also generalized to
any uncountable cardinal $\kappa$: we say that a structure $A$ with a
distinguished partial ordering $\leq$ (we shall call such $A$ a
{\em partially ordered structure}\/) has the {\em $\kappa$-FN}\/ if there is a
{\em $\kappa$-FN mapping on $A$}, that is,
a mapping $\mapping{f}{A}{[A]^{<\kappa}}$ satisfying the condition $(*)$.
In particular, the FN is the $\aleph_0$-FN and the WFN is the $\aleph_1$-FN.
This generalization is also considered in the following sections.
The paper is organized as follows. In Section \ref{prelim}, we collect some
basic facts on the $\kappa$-FN and its connection to the
$\kappa$-embedding relation $A\leq_\kappa B$ of partially ordered structures.
In Section \ref{chars}, we give some conditions equivalent to the
$\kappa$-FN which are formulated in terms of elementary submodels, and
existence of winning strategies in certain infinitary games respectively.
The behavior of \Bas\ with the $\kappa$-FN \wrt\ the cardinal
functions of independence, length and cellularity is studied in Section
\ref{card-functions}. In Sections \ref{Intalg} through
\ref{Linftykappa-free}, we deal with the question
which members of the following classes of \Bas\ have
the WFN: interval algebras, power set algebras, complete \Bas\
and $L_{\infty\kappa}$-free \Bas.
Our notation is standard. For unexplained
notation and definitions on \Bas, the reader may
consult \cite{koppelbook} and \cite{monk}.
Some set theoretic notions and basic facts used here can be found in
\cite{greenbook} and/or \cite{multiple}.
The authors would like to thank L.\ Heindorf for drawing their attention to
the weak Freese-Nation property.
\section{$\kappa$-Freese-Nation property and
$\kappa$-embedding of partially
ordered structures}\label{prelim}
In this section, we shall look at some basic properties of partially ordered
structures with
the $\kappa$-FN.
In the following, $A$, $B$, $C$ etc.\ are always partially ordered
structures for
an arbitrary (but fixed) signature. Note that this setting includes the cases
that $A$, $B$, $C$ etc.\ are \assertof{a} \Bas\ (with their canonical
ordering) or \assertof{b} bare partially ordered sets without any additional
structure.
By the theorem of Heindorf mentioned above,
every openly generated \Ba\ has the WFN. But the class of \Bas\
with the WFN contains many
more \Bas. This can be seen already in the following:
\begin{Lemma}\label{aleph-1}
If $\cardof{A}\leq\kappa$ then $A$ has the $\kappa$-FN.
\end{Lemma}
\prf Let $A=\setof{b_\alpha}{\alpha<\kappa}$. The mapping
$\mapping{f}{A}{[A]^{<\kappa}}$ defined by
\[ f(b_\alpha)=\setof{b_\beta}{\beta\leq\alpha}
\]\noindent
for $\alpha<\kappa$, is a $\kappa$-FN mapping on $A$.\qedofLemma
\indentafterqed
For $A$, $B$ \st\ $A\leq B$ (i.e.\/ $A$ is a substructure of
$B$) and $b\in B$ we write:
\[
\begin{array}{l}
A\restr b =\setof{a\in A}{a\leq b},\medskip\\
A\upper b =\setof{a\in A}{a\geq b}.
\end{array}
\]\noindent
$A$ is a {\em$\kappa$-substructure of $B$} (or {\em$\kappa$-subalgebra of
$B$} in case of \Bas; notation: $A\leq_\kappa B$) if
$A\leq B$ and, for every $b\in B$, there are a cofinal subset $U$ of
$A\restr b$
and a coinitial subset $V$ of $A\upper b$ both of cardinality less than
$\kappa$.
For $\kappa=\aleph_1$ we say also that $A$ is a
{\em$\sigma$-substructure/subalgebra}
of $B$ and denote it by $A\leq_\sigma B$. For $\kappa=\aleph_0$,
a $\kappa$-substructure/subalgebra $A$ of $B$ is also called a {\em relatively
complete substructure/subalgebra}\/) of $B$ and this is denoted also by
$A\leq_\rc B$. Note that, if
$\leq$ is
lattice order on $A$, then $A\leq_\rc B$ holds if and only if,
for all $b\in B$,
$A\restr b$ has a cofinal subset $U$ and $A\upper b$ has a coinitial subset
$V$ consisting of a single element respectively. In this case these elements
are called the {\em lower} and the {\em upper projection of $b$ on $A$} and
denoted by
$p^B_A(b)$ and $q^B_A(b)$ respectively.
Note also that, for \Bas, to show that $A\leq_\kappa B$ holds, it is
enough to check that $A\restr b$ is $<\kappa$-generated for every $b\in B$,
by duality.
The following lemma can be proved easily:
\begin{Lemma}\label{trans}
\assert{a} If $\lambda\leq\kappa$ and $A\leq_\lambda B$ then
$A\leq_\kappa B$. In particular,
if $A\leq_\rc B$ then $A\leq_\sigma B$.\\
\assert{b} If $A\leq_\kappa C$ and $A\leq B\leq C$ then
$A\leq_\kappa B$.\\
\assert{c}
For a regular cardinal $\kappa$, if $A\leq_\kappa B$ and $B\leq_\kappa C$ then
$A\leq_\kappa C$.\qed
\end{Lemma}
\begin{Lemma}\label{sigma-embed}
\assert{a} For a regular cardinal $\kappa$, if $B$ has the $\kappa$-FN
and $A\leq_\kappa B$ then $A$ also
has the $\kappa$-FN.\smallskip\\
\assert{b} For a regular cardinal $\kappa$, if $A\leq_\kappa B$, $B$ has the
$\kappa$-FN and $f$ is a $\kappa$-FN
mapping on $A$, then there is a $\kappa$-FN mapping $\tilde{g}$ on $B$
extending $f$. \smallskip\\
\assert{c} If ${g}$ is a $\kappa$-FN mapping on $B$ and $C\leq B$ is
closed \wrt\ $g$ $($i.e.\ $g(c)\subseteq C$ holds for all $c\in C$\/$)$, then
$C\leq_\kappa B$.
\end{Lemma}
\prf
For \assertof{a} and \assertof{b},
let $\mapping{g}{B}{[B]^{<\kappa}}$ be a $\kappa$-FN mapping on $B$ and,
for each $b\in B$, let
$U(b)$ and $V(b)$ be \st\
$U(b)$ is a cofinal subset of $A\restr b$, $V(b)$ is a coinitial subset
of $A\upper b$ and $\cardof{U(b)},\cardof{V(b)}<\kappa$.\smallskip\\
\assertof{a}:
Let $f$ be the mapping on $A$ defined by
\[ f(a)=\bigcup\setof{U(b)}{b\in g(a)}.
\]\noindent
Since $\kappa$ is regular we have $f(a)\in[A]^{<\kappa}$ for every $a\in A$.
$f$ is a $\kappa$-FN mapping on $A$.
To see this let $a,a'\in A$ be \st\ $a\leq a'$. Then there is
$b\in g(a)\cap g(a')$ \st\ $a\leq b\leq a'$. Since $U(b)$ is cofinal
in $A\restr b$, there is $c\in U(b)$ ($\subseteq f(a)\cap f(a')$) \st\
$a\leq c$.
Since $c\leq b$ we also have $c\leq a'$. Note that in this proof we
only needed that one of $U(b)$ and $V(b)$ is of cardinality less than
$\kappa$ for every $b\in B$.
\smallskip\\
\assertof{b}: Let $\tilde g$ be the mapping on $B$ defined by
\[ \tilde{g}(b)=
\left\{\,
\begin{array}{@{}l@{}l}
f(b) &;\mbox{ if }b\in A,\\
g(b)\cup\bigcup\setof{f(c)}{c\in U(b)\cup V(b)}\qquad&;
\mbox{ otherwise.}
\end{array}
\right.
\]\noindent
Clearly $f\subseteq\tilde{g}$.
$\tilde{g}$ is a $\kappa$-FN mapping: since $\kappa$ is regular, we have
$\tilde{g}(b)\in[B]^{<\kappa}$ for every $b\in B$. Let $b,b'\in B$ be
\st\ $b\leq b'$. We
want to show that there is $c\in \tilde{g}(b)\cap\tilde{g}(b')$ \st\
$b\leq c\leq b'$. If $b$, $b'\in A$ or $b$, $b'\in B\setminus A$, this
follows immediately from the definition of $\tilde{g}$. Suppose that
$b\in A$ and $b'\in B\setminus A$. Then there is $d\in U(b')$ \st\
$b\leq d$. Hence there is $c\in f(b)\cap f(d)$ \st\ $b\leq c\leq d$ holds.
Since $f(b)=\tilde{g}(b)$ and $f(d)\subseteq g(b')$ by $d\in U(b')$, it
follows that $c\in \tilde{g}(b)\cap\tilde{g}(b')$ and $b\leq c\leq b'$. The
case, $b\in B\setminus A$ and $b'\in A$, can be treated similarly.
\smallskip\\
\assertof{c}: Let $C\leq B$ be closed \wrt\ $g$.
For $b\in B$, let $U=g(b)\cap(C\restr b)$ and $V=g(b)\cap(C\upper b)$. Then
clearly $U$ and $V$ are of cardinality $<\kappa$.
We show that $U$ is cofinal in $C\restr b$\,:
if $c\leq b$ for some $c\in C$ then there is
$e\in g(c)\cap g(b)$ \st\ $c\leq e\leq b$ holds. Since $g(c)\subseteq C$,
we have $e\in C\restr b$. Hence $e\in U$. Similarly we can also show that
$V$ is coinitial in $C\upper b$.
\qedofLemma\smallskip\\
As already mentioned in the introduction, the \Bas\ with the FN property are
exactly the openly generated \Bas\ (\cite{heindorf-shapiro}). Hence it
follows from the next lemma that, if $(B_\alpha)_{\alpha<\delta}$ is a
continuously increasing chain of openly generated \Bas\ \st\
$B_\alpha\leq_\sigma B_{\alpha+1}$ for every $\alpha<\delta$, then
$\bigcup_{\alpha<\delta}B_\alpha$ is also openly generated.
The original proof of this fact in \cite{scepin} employed very
complicated combinatorial arguments, while our proof below and also the
proof of the characterization of openly generated \Bas\ as those with the FN
property is quite elementary.
\begin{Lemma}\label{conti-chain}Suppose that $\kappa$ is a regular cardinal,
$\delta$ a limit ordinal and $(B_\alpha)_{\alpha\leq\delta}$ a
continuously increasing chain \st\
$B_\alpha\leq_\kappa B_{\alpha+1}$ for all $\alpha<\delta$. \smallskip Then
\ifvmode\\\fi
\assert{a} $B_\alpha\leq_\kappa B_\beta$ for every
$\alpha\leq\beta\leq\delta$. \smallskip\\
\assert{b}If $B_\alpha$ has the $\kappa$-FN
for every $\alpha<\delta$, then
$B_\delta$ also has the $\kappa$-FN.
\end{Lemma}
\prf \assertof{a}: By induction on $\beta$, using
\Lemmaof{trans},\,\assertof{c} for successor steps.\smallskip\\
\assertof{b}: By \Lemmaof{sigma-embed},\,\assertof{b}, we can construct a
continuously increasing sequence $(f_\alpha)_{\alpha<\delta}$ \st\
for each $\alpha<\delta$, $f_\alpha$ is a $\kappa$-FN mapping on $B_\alpha$.
$f_\delta=\bigcup_{\alpha<\delta}f_\alpha$ is then a $\kappa$-FN mapping on
$B_\delta$.
\qedofLemma
\begin{Lemma}\label{countable-union}
Suppose that $\mu<\kappa$, $\cf(\mu)<\cf(\kappa)$ and
$(B_\alpha)_{\alpha\in\mu}$ is an increasing sequence of
$\kappa$-substructures of $B$. Then
$\bigcup_{\alpha\in\mu}B_\alpha$ is also a $\kappa$-substructure of $B$.
\end{Lemma}
\prf
\Wolog\ we may assume that $\mu=\cf(\mu)$ holds.
For $b\in B$ let $U_\alpha(b)$ be a cofinal subset of $B_\alpha\restr b$ and
$V_\alpha(b)$ a coinitial subset of $B\upper b$ both of cardinality less
than $\kappa$. Then
$U(b)=\bigcup_{\alpha\in\mu}U_\alpha(b)$ is a cofinal subset of
$(\bigcup_{\alpha\in\mu}B_\alpha)\restr b$ and
$V(b)=\bigcup_{\alpha\in\mu}V_\alpha(b)$ is a coinitial subset of
$(\bigcup_{\alpha\in\mu}B_\alpha)\upper b$. Since $\mu<\cf(\kappa)$, we have
$\cardof{U(b)}, \cardof{V(b)}<\kappa$.
\qedofLemma
\begin{Lemma}\label{non-conti-chain}
Suppose that $\kappa$ is a regular cardinal,
$\delta$ a limit ordinal and $(A_\alpha)_{\alpha<\delta}$ an
increasing chain \st\
$A_\alpha\leq_\kappa A_{\alpha+1}$ for all
$\alpha<\delta$ and $A_\gamma=\bigcup_{\alpha<\gamma}A_\alpha$ for
all limit $\gamma<\delta$ with $\cf(\gamma)\geq\kappa$.
Let $A=\bigcup_{\alpha<\delta}A_\alpha$.
If $A_\alpha$ has the $\kappa$-FN for every $\alpha<\delta$, then
$A$ also has the $\kappa$-FN.
\end{Lemma}
\prf
Let $(B_\alpha)_{\alpha\leq\delta}$ be defined by:
\[ B_\alpha=
\left\{\,
\begin{array}{@{}ll}
A_\alpha &;\mbox{ if }\alpha\mbox{ is a successor or of cofinality }
\geq\kappa,\\
{\bigcup}_{\beta<\alpha}A_\beta &;\mbox{ otherwise.}
\end{array}\right.
\]\noindent
Then $(B_\alpha)_{\alpha\leq\delta}$ is continuously increasing,
$B_\delta=A$ and $B_\alpha\leq_{\kappa}B_{\alpha+1}$ for all $\alpha<\delta$:
for a limit $\alpha<\delta$ with $\cf(\alpha)<\kappa$, this follows
from \Lemmaof{countable-union}.
Hence, using \Lemmaof{conti-chain},\,\assertof{b}, we can show by induction
that $B_\alpha$ has the $\kappa$-FN for every $\alpha\leq\delta$.
\qedofLemma
\begin{Lemma}\label{retract}
Suppose that there are order preserving mappings $\mapping{i}{A}{B}$ and
$\mapping{j}{B}{A}$ \st\ $j\circ i=id_A$. If $B$ has the $\kappa$-FN, then
$A$ also has the $\kappa$-FN. In particular, for \Bas\ $A$, $B$, if
$A$ is a retract of $B$ and $B$ has the $\kappa$-FN, then $A$ also has the
$\kappa$-FN.
\end{Lemma}
\prf
Let $\mapping{g}{B}{[B]^{<\kappa}}$ be a $\kappa$-FN mapping on $B$ and
$f$ be the mapping on $A$ defined by
\[ f(a)=j[g(i(a))].
\]\noindent
We show that $f$ is a $\kappa$-FN mapping on $A$. Clearly
$f(a)\in[A]^{<\kappa}$ for every $a\in A$. Suppose that $a,a'\in A$ are
\st\ $a\leq a'$. Then we have $i(a)\leq i(a')$. Hence there is
$b\in g(i(a))\cap g(i(a'))$ \st\ $i(a)\leq b\leq i(a')$. It follows that
\[ a=j\circ i(a)\leq j(b)\leq j\circ i(a')=a'
\]\noindent
and $j(b)\in f(a)\cap f(a')$.
\qedofLemma
\section{Characterizations of partially ordered structures with the weak
Freese-Nation property}\label{chars}
For a partially ordered structure $B$, let us say that a regular cardinal
$\chi$ is sufficiently large if
the $n$'th power of $B$ ${\cal P}^n(B)$ is in $\calH_\chi$ for every
$n\in\omega$, where $\calH_\chi$ is the set of every sets of hereditary of
cardinality less than $\chi$.
\begin{Prop}\label{characterization}
For a regular $\kappa$ and a partially ordered structure $B$,
\tfae:\smallskip\\
\assert{1} $B$ has the $\kappa$-FN;\smallskip\\
\assert{2} For some, or equivalently, any sufficiently large $\chi$, if
$M\prec\calH_\chi=(\calH_\chi,\in)$ is \st\
$B\in M$, $\kappa\subseteq M$ and $\cardof{M}=\kappa$ then
$B\cap M\leq_\kappa B$
holds;\smallskip\\
\assert{3} $\setof{C\in[B]^{\kappa}}{C\leq_\kappa B}$ contains a club set;
\smallskip\\
\assert{4} There exists a \po\ $I=(I,\leq)$ and an indexed family
$(B_i)_{i\in I}$ of substructures of $B$ of cardinality $\kappa$ \st
\begin{subassertion}{}
\mbox{}\rassert{i
$\setof{B_i}{i\in I}$ is cofinal in $([B]^{\kappa},\subseteq)$,\\
\rassert{i{}i}%
$I$ is directed and for any $i$, $j\in I$, if $i\leq j$ then $B_i\leq B_j$,\\
\rassert{i{}i{}i}%
for every well-ordered $I'\subseteq I$ of cofinality $\leq\kappa$,
$i'=\sup I'$
exists and $B_{i'}=\bigcup_{i\in I'}B_i$ holds, and\\
\rassert{i{}v}%
$B_i\leq_\kappa B$ holds for every $i\in I$.
\end{subassertion}
\end{Prop}
\prf
\assertof{1}\implies\assertof{2}:
Let $f$ be a $\kappa$-FN mapping on $B$. Since $\chi$ is sufficiently large
for $B$, we
have $B$, $f\in\calH_\chi$. Let $M\prec H_\chi$ be \st\ $B\in M$,
$\kappa\subseteq M$ and
$\cardof{M}=\kappa$. Then
there is a $\kappa$-FN mapping $f'$ on $B$ in $M$. Clearly $B\cap M$ is
closed \wrt\
$f'$. Hence it follows by \Lemmaof{sigma-embed} that
$B\cap M\leq_\kappa B$. \smallskip\\
\assertof{2}\implies\assertof{3}:
Clear.\smallskip\\
\assertof{3}\implies\assertof{4}:
Let $I\subseteq\setof{C\in[B]^{\kappa}}{C\leq_\kappa B}$ be a club
subset of $[B]^{\kappa}$ with the substructure relation. For $A\in I$, let
$B_A=A$.
Then $(I,\leq)$ and
$(B_A)_{A\in I}$ satisfy the
conditions in \assertof{4}.\smallskip\\
\assertof{4}\implies\assertof{1}: we prove this in the following two claims.
Let $I$ and $(B_i)_{i\in I}$ be as in \assertof{4}.
For a directed $I'\subseteq I$, let $B_{I'}=\bigcup_{i\in I'}B_i$.
\begin{Claim}\label{cl0}
If $I'\subseteq I$ is directed, then $B_{I'}\leq_\kappa B$.
\end{Claim}
\prfofClaim
Otherwise there is $b\in B$ \st\ either $B_{I'}\restr b$ does not
have any cofinal subset of cardinality less than $\kappa$ or
$B_{I'}\upper b$ does
not have any coinitial subset of cardinality less than $\kappa$. For
simplicity, let
us assume the first case.
Then there exists an increasing sequence $(I_\alpha)_{\alpha<\kappa}$ of
directed subsets of $I'$ of cardinality less than $\kappa$ \st\
$B_{I_\alpha}\restr b$ is not cofinal in $B_{I_{\alpha+1}}\restr b$.
By \assertof{iii}, $i_\alpha=\sup I_\alpha$ exists and
$B_{i_\alpha}=B_{I_\alpha}$ holds for every $\alpha<\kappa$.
$(i_\alpha)_{\alpha<\kappa}$ is an increasing sequence in $I$. Hence,
again by \assertof{iii}, there exists $i^*=\sup_{\alpha<\kappa}i_\alpha$
and $B_{i^*}=\bigcup_{\alpha<\kappa}B_{i_\alpha}$. By \assertof{iv},
$B_{i^*}\leq_\kappa B$. But by the construction, $B_{i^*}\restr b$ cannot
have any cofinal subset of cardinality less than $\kappa$. This is a
contradiction.
\qedofClaim
\begin{Claim}\label{cl1}
If $I'\subseteq I$ is directed, then $B_{I'}$ has the $\kappa$-FN.
\end{Claim}
\prfofClaim
We prove the claim by induction on $\cardof{I'}$.
If $\cardof{I'}\leq\kappa$, we have $\cardof{B_{I'}}=\kappa$. Hence, by
\Lemmaof{aleph-1}, $B_{I'}$ has the $\kappa$-FN.
Assume that $\cardof{I'}=\lambda>\kappa$ and that we have
proved the claim for every directed
$I''\subseteq I$ with $\cardof{I''}<\lambda$.
Take a continuously increasing sequence $(I_\alpha)_{\alpha<\cf(\lambda)}$ of
directed subsets of $I'$ \st\ $\cardof{I_\alpha}<\lambda$ for every
$\alpha<\cf(\lambda)$ and $I'=\bigcup_{\alpha<\cf(\lambda)}I_\alpha$.
$(B_{I_\alpha})_{\alpha<\cf(\lambda)}$ is then a continuously increasing
sequence
of substructures of $B_{I'}$ and
$B_{I'}=\bigcup_{\alpha<\cf(\lambda)}B_{I_\alpha}$.
By the induction hypothesis, $B_{I_\alpha}$ has the $\kappa$-FN and,
by
\Claimof{cl0}, we have $B_{I_\alpha}\leq_\kappa B$. Hence, by
\Lemmaof{conti-chain},\assertof{b}, $B_{I'}$ has also the $\kappa$-FN.
\qedofClaim
\noindentafterqed
Now by applying \Claimabove\ to $I'=I$, we can conclude that
$B=B_I$ has the $\kappa$-FN.
\qedofProp%
\nc{\kappagame}[2]{\calG^{#1}(#2)}%
\nc{\kappalambdagame}[3]{\calG^{#1}_{#2}(#3)}%
\indentafterqed
Now we give yet another characterization of partially ordered structures
with the $\kappa$-FN
by means of a game. This characterization will be used later in
the proof of Propositions \ref{omega-2}, \ref{ccc-rc}, etc.
For a partially ordered structure\ $B$, let $\kappagame{\kappa}{B}$ be the
following game played by Players I and II:
in a play in $\kappagame{\kappa}{B}$, Players I and II choose
subsets
$X_\alpha$ and $Y_\alpha$ of $B$ of cardinality less than $\kappa$
alternately for
$\alpha<\kappa$ \st
\[
X_0\subseteq Y_0\subseteq X_1\subseteq Y_1\subseteq\cdots\subseteq X_\alpha
\subseteq Y_\alpha\subseteq\cdots\subseteq X_\beta\subseteq
Y_\beta\subseteq\cdots
\]\noindent
for $\alpha\leq\beta<\kappa$. So a play in $\kappagame{\kappa}{B}$
looks
like
\[
\begin{array}{l@{}l@{}@{}l@{}@{}l@{}l}
\mbox{\it Player I}\ &:\ \ &X_0,\ &X_1,\ \ldots,\ &X_\alpha,\ \ldots\medskip\\
\mbox{\it Player II}\ &:\ \ &Y_0,\ &Y_1,\ \ldots,\ &Y_\alpha,\ \ldots
\end{array}
\]\noindent
where $\alpha<\kappa$. Player II wins the play if
$\bigcup_{\alpha<\kappa}X_\alpha=\bigcup_{\alpha<\kappa}Y_\alpha$ is a
$\kappa$-substructure of $B$.
Let us call a strategy $\tau$ for Player II simple if, in $\tau$, each
$Y_\alpha$ is decided from the information of the set
$X_\alpha\subseteq B$ alone (i.e.\ also independent of $\alpha$).
For a sufficiently large $\chi$ (\wrt\ $B$), an elementary submodel $M$ of
$\calH_\chi$ is said to be $V_{\kappa}$-like if, either
$\kappa=\aleph_0$ and $M$ is countable, or there is an
increasing
sequence $(M_\alpha)_{\alpha<\kappa}$ of elementary submodels
of $M$ of cardinality less than $\kappa$ \st\
$M_\alpha\in M_{\alpha+1}$
for all
$\alpha<\kappa$ and $M=\bigcup_{\alpha<\kappa}M_\alpha$. If $M$ is
$V_{\kappa}$-like, we say that a sequence
$(M_\alpha)_{\alpha<\kappa}$ as above witnesses the $V_{\kappa}$-likeness of
$M$.
The notion of $V_\kappa$-like elementary submodels of $\calH_\chi$ is a
weakening of internally approachable elementary submodels introduced in
\cite{FMSh:240}.
An elementary submodel $M$ of $\calH_\chi$ is said to be internally
approachable if $M$ is the union of continuously increasing sequence
$(M_\alpha)_{\alpha<\kappa}$ of smaller elementary submodels \st\
$(M_\beta)_{\beta\leq\alpha}\in M_{\alpha+1}$ for every $\alpha<\kappa$.
The main reason of the use of $V_\kappa$-like elementary submodels here
instead of internally approachable ones is the following
\Lemmaof{V-kappa-like},\,\assertof{b} which seems to be false in general for
internally approachable elementary submodels.
\begin{Lemma}\label{V-kappa-like}
\assert{a}
If $M$ is a $V_{\kappa}$-like elementary submodel of $\calH_\chi$ \st\
$\kappa\in M$, then
$\kappa\subseteq M$ holds. Hence, if $x$ is of cardinality less or equal to
$\kappa$ and
$x\in M$ then we have $x\subseteq M$.
\medskip\\
\assert{b}
If $(N_\alpha)_{\alpha<\kappa}$ is an increasing sequence of
$V_{\kappa}$-like elementary submodels of $\calH_\chi$, then
$M=\bigcup_{\alpha<\kappa}N_\alpha$ is also a $V_{\kappa}$-like elementary
submodel of $\calH_\chi$.
\end{Lemma}
\prf
\assertof{a}:
Let $(M_\alpha)_{\alpha<\kappa}$ witness the $V_{\kappa}$-likeness of
$M$.
Assume that $\kappa\not\subseteq M$. Let
\[ \alpha_0=\min\setof{\alpha\in\kappa}{\alpha\not\in M}.
\]\noindent
Then we have $\alpha_0\subseteq M$. Let
\[ \alpha_1=\min\setof{\alpha\leq\kappa}{\alpha_0\leq\alpha,\,\alpha\in M}.
\]\noindent
Since $\alpha_0$ is of cardinality less than $\kappa$, there exists
$\alpha<\kappa$
\st\
$\alpha_0\subseteq M_\alpha$. Then we have $\alpha_1\in M_{\alpha+1}$.
Since $M_\alpha\in M_{\alpha+1}$,
$\alpha_0=\setof{\beta\in M_\alpha}{\beta<\alpha_1}$ is an element of
$M_{\alpha+1}\subseteq M$. This is a contradiction. Hence we have
$\kappa\subseteq M$.
If $x$ is of cardinality less or equal to $\kappa$ and $x\in M$, then there
is a
surjection $\mapping{f}{\kappa}{x}$ in $M$. Since $\kappa\subseteq M$,
it follows that $x=f[\kappa]\subseteq M$.
\medskip\\
\assertof{b}:
It is clear that $M$ is an elementary submodel of $\calH_\chi$. To prove
that $M$ is $V_{\kappa}$-like, let
$M=\setof{m_\xi}{\xi<\kappa}$ and,
for each $\alpha<\kappa$, let $(N_{\alpha,\beta})_{\beta<\kappa}$ be an
increasing sequence of elementary submodels of $\calH_\chi$ of cardinality
less than $\kappa$
witnessing the $V_{\kappa}$-likeness of $N_\alpha$. Since
$N_{\alpha,\beta}\in N_\alpha\subseteq M$ and
$\bigcup_{\beta<\kappa}N_{\alpha,\beta}=N_\alpha$,
we can choose
$\alpha_\xi$, $\beta_\xi<\kappa$ for $\xi<\kappa$ inductively
\st\medskip\\
\assert{a} $(N_{\alpha_\xi,\beta_\xi})_{\xi<\kappa}$ is an increasing
sequence,\smallskip\\
\assert{b} $N_{\alpha_\xi,\beta_\xi}\in N_{\alpha_{\xi+1},\beta_{\xi+1}}$
holds for every $\xi<\kappa$ and\smallskip\\
\assert{c} $m_\xi\in N_{\alpha_{\xi},\beta_{\xi}}$ for every
$\xi<\kappa$.\medskip\\
Then $(N_{\alpha_\xi,\beta_\xi})_{\xi<\kappa}$ witnesses the
$V_{\kappa}$-likeness of $M$.
\qedofLemma
\begin{Prop}\label{game}
For regular $\kappa$ and a partially ordered structure\ $B$,
\tfae:\smallskip\\
\assert{1} $B$ has the $\kappa$-FN;\smallskip\\
\assert{2} Player II has a simple winning strategy in
$\kappagame{\kappa}{B}$;\smallskip\\
\assert{3} For some, or equivalently any, sufficiently large $\chi$, if
$M\prec\calH_\chi$ is $V_{\kappa}$-like \st\ $B$, $\kappa\in M$,
then $B\cap M\leq_\kappa B$.
\end{Prop}
\prf
Assume that $\kappa$ is uncountable
(for $\kappa=\aleph_0$, the proof is easier than the following one and given
in \cite{appendix}).\medskip\\
\assertof{1}\implies\assertof{2}: Let $\mapping{f}{B}{[B]^{<\kappa}}$ be a
$\kappa$-FN mapping on $B$. Then Player II can win by the following strategy:
in the
$\alpha$'th move, Player II chooses $Y_\alpha$ so that
$X_\alpha\subseteq Y_\alpha$ and $Y_\alpha$ is a substructure of
$B$ of cardinality less than $\kappa$ closed under $f$. After $\kappa$ moves,
$\bigcup_{\alpha<\kappa}Y_\alpha$ is a substructure of $B$ closed under
$f$. Hence, by \Lemmaof{sigma-embed},\,\assertof{c}, it is a
$\kappa$-substructure of $B$.\medskip\\
\assertof{2}\implies\assertof{3}: Let $M$ be a $V_{\kappa}$-like
elementary submodel of $\calH_\chi$ \st\ $B$, $\kappa\in M$. We have to
show that
$B\cap M\leq_\kappa B$.
Let $(M_\alpha)_{\alpha<\kappa}$ witness the $V_{\kappa}$-likeness of
$M$. \Wolog\ we may assume that $B\in M_0$. By $M_0\prec\calH_\chi$, there
is a simple winning strategy
$\tau\in M_0$ for Player II in $\kappagame{\kappa}{B}$ (hence
$\tau\in M_\alpha$ for every $\alpha<\kappa$). Let
$(X_\alpha, Y_\alpha)_{\alpha<\kappa}$ be the play in
$\kappagame{\kappa}{B}$ \st\ at his $\alpha$'th move, Player I took
$B\cap M_{\xi_\alpha}$ for some $\xi_\alpha<\kappa$ and
Player II played always according to $\tau$. Such a game is possible since
if Player I chooses $B\cap M_{\xi_\alpha}$ at his $\alpha$'th move, then
$B\cap M_{\xi_\alpha}\in M_{\xi_\alpha+1}$. Hence Player II's move $Y_\alpha$
taken according to $\tau$ is also an element of $M_{\xi_\alpha+1}$. Since
$Y_\alpha$
is of cardinality less than $\kappa$, we have $Y_\alpha\subseteq M$ by
\Lemmaof{V-kappa-like},\,\assertof{a}. Hence there is some
$\xi_{\alpha+1}\geq\xi_\alpha$ \st\
$Y_\alpha\subseteq M_{\xi_{\alpha+1}}$.
Thus Player I may take $B\cap M_{\xi_{\alpha+1}}$ at his next move.
Now we have
$B\cap M=B\cap(\bigcup_{\alpha<\kappa}M_{\xi_\alpha})\leq_\kappa B$
since $\tau$ was a winning strategy of Player II.\medskip\\
\assertof{3}\implies\assertof{1}:
Let
\[
\begin{array}{@{}ll}
\calC=\setof{M}{&\cardof{M}=\kappa,\,M\mbox{ is a union of an increasing
sequence}\\
&\mbox{of }V_{\kappa}\mbox{-like elementary submodels of }
\calH_\chi\mbox{ \st\ }B,\,\kappa\in M}.
\end{array}
\]\noindent
Then it is easy to see that $\calC$ is club in $[\calH_\chi]^{\kappa}$.
Hence $\calC'=\setof{B\cap M}{M\in\calC}$ contains a club subset of
$[B]^{\kappa}$. By \Propof{characterization},\,\assertof{3}, if follows
from the claim below that $B$ has the $\kappa$-FN.
\iffalse
Note that, in the definition of $I$, we do not demand that the sequence of
$V_{\kappa}$-like submodels of $\calH_\chi$ is strictly increasing.
Hence every $V_{\kappa}$-like submodel of $\calH_\chi$ \st\ $B$,
$\kappa\in M$ is also
an element of
$I$. For $M$, $N\in I$, let $M< N$ \equivto\ $M\in N$.
By \Claimabove, we have $M\subseteq N$ if $M<N$ holds. It follows that $<$
is a transitive relation and hence partial ordering on $I$.
For $M\in I$, let
$B_M=B\cap M$. We show that $I$ and $(B_M)_{M\in I}$ satisfy
\assertof{i}~--~\assertof{iv} of \Propof{characterization},\,\assertof{4}.
\relax From this it follows, by \Propof{characterization}, that $B$ has the
$\kappa$-FN. \medskip\\
\assertof{i} {\it $\setof{B_M}{M\in I}$ is cofinal in
$([B]^{\kappa},\subseteq)$}\,:\smallskip\\
Let $X\in[B]^{\kappa}$, say
$X=\setof{b_\alpha}{\alpha<\kappa}$. By the downward L\"owenheim-Skolem
theorem, we can construct an increasing sequence
$(M_\alpha)_{\alpha<\kappa}$ of
elementary submodels of $\calH_\chi$ of cardinality less than $\kappa$
inductively so
that $B$, $\kappa\in M_0$ and
$M_\alpha$, $b_\alpha\in M_{\alpha+1}$ holds for every $\alpha<\kappa$.
Then $M=\bigcup_{\alpha<\kappa}M_\alpha$ is $V_{\kappa}$-like, $B$,
$\kappa\in M$
$X\subseteq M$. Hence we have $X\subseteq B_M.$\medskip\\
\assertof{ii} {\it$I$ is directed and for any $M$, $N\in I$, if $M\leq N$ then
$B_M\subseteq B_N$}:\smallskip\\
The directedness of $I$ can be proved as in \assertof{i}.
The second half of the assertion is trivial by the definition of $B_M$.
\medskip\\
\assertof{iii}
{\it for every well-ordered $I'\subseteq I$ of cofinality $\leq\kappa$,
$i'=\sup I'$
exists and $B_{i'}=\bigcup_{i\in I'}B_i$ holds\/}:\smallskip\\
Let $(M_\alpha)_{\alpha<\delta}$ be an increasing sequence of elements of
$I$ for $\delta\leq\kappa$. For each $\alpha<\delta$, let
$(M_{\alpha,\beta})_{\beta<\eta_\alpha}$ be an increasing sequence of
$V_{\kappa}$-like elementary submodels of $\calH_\chi$ \st\ $B$,
$\kappa\in M_{\alpha,0}$ for every $\alpha<\delta$ and
$M_\alpha=\bigcup_{\beta<\eta_\alpha}M_{\alpha,\beta}$. Since
$M_\alpha\in M_{\alpha+1}$ holds for every $\alpha<\delta$, there is
$\beta_\alpha<\eta_{\alpha+1}$ \st\ $M_\alpha\in M_{\alpha+1,\beta_\alpha}$
holds. By \Lemmaof{V-kappa-like},\,\assertof{a}, it follows that
$M_\alpha\subseteq M_{\alpha+1,\beta_\alpha}$. Hence
$\bigcup_{\alpha<\delta}M_\alpha=
\bigcup_{\alpha<\delta}M_{\alpha+1,\beta_\alpha}$. This shows that
$\bigcup_{\alpha<\delta}M_\alpha\in I$ holds.
Clearly $\bigcup_{\alpha<\delta}M_\alpha$ is the supremum of
$(M_\alpha)_{\alpha<\delta}$. The equation follows immediately from the
definition of $B_{M_\alpha}$.
\medskip\\
\assertof{iv} {\it $B_M\leq_\kappa B$ holds for every $M\in I$}\,:\smallskip\\
\f
\begin{Claim}
$B\cap M\leq_\kappa B$ holds for every $M\in \calC$.
\end{Claim}
\prfofClaim
If $M\in \calC$ is union of a sequence $(M_\alpha)_{\alpha<\rho}$ of
$V_{\kappa}$-like elementary submodels of $\calH_\chi$ for some
$\rho<\kappa$ \st\ $B$, $\kappa\in M_0$, then we
have
$B\cap M_\alpha\leq_\kappa B$ for every $\alpha<\rho$ by \assertof{3}. Hence
we have $B\cap M=\bigcup_{\alpha<\rho}(B\cap M_\alpha)\leq_\kappa B$ by
\Lemmaof{countable-union}. If $M\in \calC$ is the union
of a $\kappa$-chain of $V_{\kappa}$-like elementary submodels of
$\calH_\chi$, then if follows by \Lemmaof{V-kappa-like},\,\assertof{b}
that $M$ itself is
$V_{\kappa}$-like. Hence we have $B\cap M\leq_\kappa B$ by the assumption.
\qedofClaim
\qedofProp
\indentafterqed
Under $2^{<\kappa}=\kappa$, Propositions \ref{characterization} and
\ref{game} can be
yet improved. This is because of the following fact:
\begin{Lemma}\label{Vomega-1-like}
Assume that $\kappa$ is a regular cardinal \st\ $2^{<\kappa}=\kappa$.
Let $B$ be a partially ordered structure, $\chi$ be sufficiently
large for $B$ and $M\subseteq\calH_\chi$.
Then \tfae:\smallskip\\
\assert{1} $M$ is a $V_{\kappa}$-like elementary submodel of
$\calH_\chi$;\smallskip\\
\assert{2} $M\prec\calH_\chi$, $\cardof{M}=\kappa$, $B\in M$ and
$[M]^{<\kappa}\subseteq M$.
\end{Lemma}
\prf
For $\kappa=\aleph_0$, this is clear. Assume that $\kappa$ is uncountable.
\smallskip\\
\assertof{1}\implies\assertof{2}: Let $M$ be a $V_{\kappa}$-like
elementary submodel of $\calH_\chi$ and let
$(M_\alpha)_{\alpha<\kappa}$ be an increasing sequence of
elementary submodels of $\calH_\chi$ witnessing the
$V_{\kappa}$-likeness of $M$. It is enough to show that
$[M_\alpha]^{<\kappa}\subseteq M$ holds for every $\alpha<\kappa$. By
$M_\alpha\in M$, we have $[M_\alpha]^{<\kappa}\in M$. By
$2^{<\kappa}=\kappa$,
$[M_\alpha]^{<\kappa}$ has cardinality $\kappa$. Hence, by
\Lemmaof{V-kappa-like},\,\assertof{a}, it follows that
$[M_\alpha]^{\kappa}\subseteq M$.
\smallskip\\
\assertof{2}\implies\assertof{1}: Suppose that $M\prec\calH_\chi$,
$\cardof{M}=\kappa$, $B\in M$ and $[M]^{<\kappa}\subseteq M$. Let
$M=\setof{m_\alpha}{\alpha<\kappa}$. Then we can construct inductively an
increasing sequence $(M_\alpha)_{\alpha<\kappa}$ of elementary submodels of
$M$ of cardinality less than $\kappa$ \st\
$M_\alpha,m_\alpha\in M_{\alpha+1}$ for every
$\alpha<\kappa$.
This is possible since at $\alpha$'th step of the inductive construction, we
have that $M_\alpha$ is a subset of $M$ of cardinality less than $\kappa$. By
$[M]^{<\kappa}\subseteq M$, it follows that $M_\alpha\in M$. So by
the downward L\"owenheim-Skolem Theorem, we can take
$M_{\alpha+1}\prec M$ \st\ $\cardof{M_{\alpha+1}}<\kappa$ and
$M_\alpha,\,m_\alpha\in M_{\alpha+1}$.
At a limit $\gamma<\kappa$ we take $\bigcup_{\alpha<\gamma}M_\alpha$.
Then $(M_\alpha)_{\alpha<\kappa}$ witnesses the
$V_{\kappa}$-likeness of $M$.
\qedofLemm
\begin{Prop}\label{char-under-ch}
Assume that $\kappa$ is a regular cardinal \st\ $2^{<\kappa}=\kappa$.
Then for a partially ordered structure $B$,
the following are equivalent:\smallskip\\
\assert{1} $B$ has the $\kappa$-FN;\smallskip\\
\assert{2} For sufficiently large $\chi$ and for all $M\prec\calH_\chi$, if
$B\in M$, $\cardof{M}=\kappa$ and $[M]^{<\kappa}\subseteq M$, then
$B\cap M\leq_\kappa B$ holds;\smallskip\\
\assert{3} Player II has a winning strategy in $\kappagame{\kappa}{B}$.
\end{Prop}
\prf
\assertof{1}\equivto\assertof{2}: By \Lemmaof{Vomega-1-like} and
\Propof{game}. \smallskip\\
\assertof{1}\implies\assertof{3} follows from \Propof{game}.\smallskip\\
\assertof{3}\implies\assertof{2}: Let $M$ be as in \assertof{2} and let
$B\cap M=\setof{b_\alpha}{\alpha<\kappa}$. By \assertof{3}, there is a
winning strategy $\tau\in M$ of Player II in $\kappagame{\kappa}{B}$. Let
$(X_\alpha, Y_\alpha)_{\alpha<\kappa}$ be a play in
$\kappagame{\kappa}{B}$ \st\
Player I chooses $X_\alpha$ so that $\cardof{X_\alpha}<\kappa$ and
$b_\alpha\in X_\alpha$, and Player II played always according to
$\tau$. Such a
game is possible since, by $[M]^{<\kappa}\subseteq M$, at Player II's
$\alpha$'th innings, she has $(X_0,Y_0,\ldots,X_\alpha)\in M$. Hence her
move $Y_\alpha$ taken according to $\tau$ will be also an element of $M$.
Since $\cardof{Y_\alpha}<\kappa$,
$Y_\alpha$ is a subset of $M$. Now we have
$\bigcup_{\alpha<\kappa}Y_\alpha=\bigcup_{\alpha<\kappa}X_\alpha=B\cap M$.
Since $\tau$ was a winning strategy, we also have
$B\cap M=\bigcup_{\alpha<\kappa}X_\alpha\leq_\kappa B$.\smallskip\\
\qedofProp
\indentafterqed
We can also consider the following variant of the game
$\kappagame{\kappa}{B}$: for cardinals $\kappa$, $\lambda$ \st\
$\lambda\leq\kappa$ and
a partially ordered structure $B$, $\kappalambdagame{\kappa}{\lambda}{B}$ is
the game just like $\kappagame{\kappa}{B}$ except that
Player II wins in $\kappalambdagame{\kappa}{\lambda}{B}$ if and only if
$\bigcup_{\alpha<\kappa}X_\alpha\leq_\lambda B$.
As in Propositions \ref{characterization},\,\ref{game}, we can prove the
implication
\assertof{A}\implies\assertof{B}\implies\assertof{C}\implies\assertof{D} for
the following assertions for regular $\kappa$, $\lambda$. \bigskip\\
\assert{A} For every sufficiently large $\chi$ and $M\prec\calH_\chi$ of
cardinality $\kappa$ with $B\in M$, we have $B\cap M\leq_\lambda B$.\medskip\\
\assert{B} Player II has a simple winning strategy in
$\kappalambdagame{\kappa}{\lambda}{B}$.\medskip\\
\assert{C} For every sufficiently large $\chi$ and $M\prec\calH_\chi$, if
$M$ is $V_{\kappa}$-like then $B\cap M\leq_\lambda B$.\medskip\\
\assert{D} Player II has a winning strategy in
$\kappalambdagame{\kappa}{\lambda}{B}$.\medskip\\
\assert{E} For any sufficiently large $\chi$ and $M\prec\calH_\chi$ of
cardinality $\kappa$ \st\ $B\in M$ and $[M^{<\kappa}\subseteq M$,
$B\cap M\leq_\lambda B$ holds.\bigskip\\
For the implication \assertof{A}\implies\assertof{B}, we fix an expansion
of $\calH_\chi$ by Skolem functions. For
$x\subseteq\calH_\chi$, let $\tilde{h}(x)$ be the Skolem hull of $x$.
We may take the Skolem hull operation so that $B\in\tilde{h}(\emptyset)$
holds. Player II then
wins if she takes $Y_\alpha$ \st\ $X_\alpha\subseteq Y_\alpha$ and
$\tilde{h}(Y_\alpha)\cap B=Y_\alpha$ hold in each of her $\alpha$'th innings
for $\alpha<\kappa$.
By the same idea, we can also prove the equivalence of \assertof{B}
and \assertof{C}, if we allow Player II to remember her last move
in her simple winning strategy in \assertof{B}.
By \Lemmaof{Vomega-1-like}, we have
\assertof{C}\equivto\assertof{D}\equivto\assertof{E} under
$2^{<\kappa}=\kappa$.
\section{Cardinal functions on \Bas\ with the weak Freese-Nation property}%
\label{card-functions}
In \cite{heindorf-shapiro} it is shown that, for any openly generated \Ba\
(i.e., \Ba\ with the FN),
the cardinal functions (those studied in \cite{monk}, possibly except the
topological density $d$)
have the same value as for the free \Ba\ of the same cardinality,
as follows obviously from \Lemmaof{aleph-1}. Later we shall see some more
examples of \Bas\ with
the WFN which behave quite differently from free \Bas\ \wrt\ cardinal
functions. Nevertheless, there are some restrictions on the values of
cardinal functions on \Bas\ with the WFN.
\begin{Prop}\label{omega-2}
For every partially ordered structure $B$, if $\kappa^++1$ or $(\kappa^++1)^*$
is (order isomorphic) embeddable into $B$ then $B$ does not have the
$\kappa$-FN. In particular, for every \Ba\ with the $\kappa$-FN, we have
${\rm Depth}(B)\leq\kappa$.
\end{Prop}
\prf
Suppose that $\mapping{i}{\kappa^++1}{B}$ is an embedding (the case for
$(\kappa^++1)^*$ can be handled similarly). Let $\mapping{j}{B}{\kappa^++1}$ be
defined by
\[ j(b)=\sup\setof{\alpha}{i(\alpha)\leq b}
\]\noindent
for $b\in B$. Then $j$ is order preserving and $j\circ i=id_{\kappa^++1}$
holds.
Hence, by \Lemmaof{retract}, the following claim proves the proposition:
\begin{Claim}
$(\kappa^++1,\leq)$ does not have the $\kappa$-FN.
\end{Claim}
\prfofClaim
Player I wins a game in $\kappagame{\kappa}{\kappa^++1}$ if he chooses
$X_\alpha$ at his $\alpha$'th move \st\
$\sup X_\alpha\setminus\smallsetof{\kappa^+}
>\sup\bigcup_{\beta<\alpha}Y_\beta\setminus\smallsetof{\kappa^+}$ holds.
By \Propof{game}, it follows that $(\kappa^++1,\leq)$ does not have
the $\kappa$-FN. Note that for the implication
\assertof{1}\implies\assertof{2} in \Propof{game}
used here, we do not need the assumption of regularity of $\kappa$.
\qedofClaim
\qedofProp
\begin{Thm}\label{ind}
For a regular cardinal $\kappa$, if a \Ba\ $B$ has the
$\kappa$-FN, $\lambda=\lambda^{<\kappa}$ and $X\subseteq B$ is of
cardinality $>\lambda$ then there is an independent $Y\subseteq X$ of
cardinality $>\lambda$.
\end{Thm}
\prf
Essentially the same argument as the following one has been used in
\cite[\S 4]{Sh:92}.
Let $f$ be a $\kappa$-FN mapping on $B$. Let
$(a_\delta)_{\delta<\lambda^+}$ be a sequence of elements of $X$ \st,
letting $B_\delta$ be the closure of $\setof{a_\gamma}{\gamma<\delta}$ \wrt\
$f$ and the Boolean operations, $a_\delta\not\in B_\delta$ holds for every
$\delta<\lambda^+$.
By \Lemmaof{sigma-embed}, \assertof{c}, we have $B_\delta\leq_\kappa B$ for
every $\delta<\lambda^+$.
Let
$S=\setof{\delta<\lambda^+}{\cf(\delta)\geq\kappa}$.
For each $\delta\in S$, let $I_\delta$ and $J_\delta$ be
cofinal subsets of $B_\delta\restr a_\delta$ and $B_\delta\restr -a_\delta$
respectively, both of cardinality less than $\kappa$. Let
\[h(\delta)=\tuple{I_\delta, J_\delta}.
\]\noindent
By Fodor's lemma and $\lambda=\lambda^{<\kappa}$, there is a stationary
$T\subseteq S$ \st\ $h\restr T$ is
constant, say $h(\delta)=\tuple{I,J}$ for all $\delta\in T$.
Let
\[
\delta^*=\min\setof{\delta<\lambda^+}{I,\,J\subseteq B_\delta}.
\]
\Wolog\ we may assume that $\delta^*<\delta$ holds for every $\delta\in T$.
Let
\[ L=\setof{b\in B_{\delta^*}}{b\not\leq i+j\mbox{ for all }i\in I,\, j\in J}.
\]\noindent
Then we have\medskip\\
\assert{1} $1\in L$ (since, by $a_\delta\not\in B_{\delta^*}$ for any
$\delta\in T$, $I\cup J$ generates a proper ideal of $B_{\delta^*}$). In
particular we have $L\not=\emptyset$;\\
\assert{2} If $b\in L$ and $k\in I\cup J$ then $b\cdot-k\in L$. \medskip\\
Now, by \assertof{1} above, the following claim shows that
$\setof{a_\delta}{\delta\in T}$ is independent. Since
$\cardof{T}=\lambda^+$, this proves the theorem.
\begin{Claim}
If $b\in L$ and $p$ is an elementary product over
$a_{\delta_0}$,\ldots,$a_{\delta_{n-1}}$ for $\delta_i\in T$ \st\
$\delta_0<\cdots<\delta_{n-1}$
(i.e.\ $p$ is of the form
$(a_{\delta_0})^{\tau_0}\cdot\,\cdots\,\cdot(a_{\delta_{n-1}})^{\tau_{n-1}}$
for some $\tau_i\in 2$, $i<n$) then $b\cdot p\not= 0$.
Here, for a \Ba\
$B$, $b\in B$ and $i\in 2$, we define $(b)^i$ by:
\[ (b)^i=
\left\{\,
\begin{array}{@{}l@{}l}
b\quad&\mbox{; if }i=1,\\
-b\quad&\mbox{; if }i=0.
\end{array}
\right.
\]\noindent
\end{Claim}
\prfofClaim
By induction on $n$. For $n=0$, this is trivial since $0\not\in L$. Assume
that the claim holds for $n$. Let $\delta_0$,\ldots,$\delta_n\in T$ be \st\
$\delta_0<\cdots\delta_{n-1}<\delta_n$ and let $p$ be an arbitrary
elementary product over $a_{\delta_0}$,\ldots,$a_{\delta_{n-1}}$.
Let $b\in L$. By the
induction hypothesis, we have $b\cdot p\not=0$. We have to show that
$b\cdot p\cdot a_{\delta_n}\not=0$ and $b\cdot p\cdot-a_{\delta_n}\not=0$.
Toward a contradiction, assume that $b\cdot p\cdot a_{\delta_n}=0$ holds. Then
$b\cdot p\leq -a_{\delta_n}$. Since $b\cdot p\in B_{\delta_n}$, we can find
$j\in J$ \st\ $b\cdot p\leq j$. Hence $(b\cdot-j)\cdot p=0$. Since
$b\cdot-j\in L$ by \assertof{2} above, this is a contradiction to the
induction hypothesis. Similarly, from $b\cdot p\cdot-a_{\delta_n}=0$, it
follows that $(b\cdot-i)\cdot p=0$ for some $i\in I$ which again is a
contradiction to \assertof{2}.
\qedofClaim\\
\qedofThm
\indentafterqed
The next corollary gives a positive answer to a problem
by L.\ Heindorf.
\begin{Cor}\label{ind-ineq}
For a regular cardinal $\kappa$, if a \Ba\
$B$ has the $\kappa$-FN then $\cardof{B}\leq{\rm Ind}(B)^{<\kappa}$. In
particular,
for an openly generated \Ba\ $B$, $\cardof{B}= {\rm Ind}(B)$
holds (\cite{heindorf-shapiro}).
For
a \Ba\ $B$ with the WFN, we have
$\cardof{B}\leq{\rm Ind}(B)^{\aleph_0}$.
\end{Cor}
\prf
Assume that $B$ has the $\kappa$-FN but
$\cardof{B}>{\rm Ind}(B)^{<\kappa}$ holds. Then by
${\rm Ind}(B)^{<\kappa}=({\rm Ind}(B)^{<\kappa})^{<\kappa}$, we have
${\rm Ind}(B)>{\rm Ind}(B)^{<\kappa}$ by \Thmof{ind}. But this is impossible.
\qedofCor
\noindentafterqed
In \Corabove\ the equality is attained for every regular $\kappa$. For the
case of $\kappa=\aleph_1$,
the simplest example to see this would be $Intalg(\reals)$
(see \Propof{intalg} below).
The following corollary is an immediate consequence of \Thmof{ind}:
\begin{Cor}{\rm(\cite{heindorf-shapiro} for $\kappa\leq\aleph_1$)}\label{Lutz}
Let $\kappa$ be a regular cardinal. If a \Ba\ $B$ has the $\kappa$-FN, then
$c(B)\leq2^{<\kappa}$ and ${\rm Length}(B)\leq 2^{<\kappa}$. \qed
\end{Cor}
\section{Interval algebras and power set algebras } \label{Intalg}
\begin{Prop}\label{intalg}
\mbox{}\\
\assert{a} For $\rho\in\Ord$, $Intalg(\rho)$ has
the $\kappa$-FN if and only if $\rho<\kappa^+$.\smallskip\\
\assert{b}$Intalg(\reals)$ has the WFN.\smallskip\\
\assert{c} For a totally ordered set $X$, $Intalg(X)$ has the $\kappa$-FN
if and only if $X$ has the $\kappa$-FN.\smallskip\\
\assert{d} Assume that $\kappa$ is a regular cardinal. For a linearly
ordered set $X$, if $Intalg(X)$ (hence,
by \assertof{c}, also $X$) has the $\kappa$-FN then
$\cardof{X}\leq 2^{<\kappa}$.
\end{Prop}
\prf
For $b\in Intalg(X)$, let $ep(b)$ be the set of end points of
$b$, i.e.
\[
ep(b)=\setof{x_i}{i<2n}.
\]\noindent
where $b=\dot{\bigcup}_{j<n}[x_{2j},x_{2j+1})$ in the standard
representation. \medskip\\
\assertof{a}: For $\rho<\kappa^+$, $Intalg(\rho)$ has cardinality less or
equal to $\kappa$. Hence, by \Lemmaof{aleph-1}, $Intalg(\rho)$
has the $\kappa$-FN.
If $\rho\geq\kappa^+$, $Intalg(\rho)$ does not have the $\kappa$-FN
by \Propof{omega-2}.\smallskip\\
\assertof{b}:
For all $b\in Intalg(\reals)$, the mapping defined by
\[g(b)=\setof{c\in Intalg(\reals)}{ep(c)\subseteq\rationals
\cup ep(b)}.\]\noindent
is a WFN-mapping on $Intalg(\reals)$.\smallskip\\
\assertof{c}:
If
$\mapping{f}{X}{[X]^{<\kappa}}$ is a $\kappa$-FN mapping on $X$ then
$\mapping{g}{Intalg(X)}{[Intalg(X)]^{<\kappa}}$ defined by
\[ g(b)=\setof{c\in Intalg(X)}{ep(c)\subseteq\bigcup f[ep(b)]}
\]\noindent
is a $\kappa$-FN mapping on $Intalg(X)$. Conversely, if $g$ is a
$\kappa$-FN mapping on
$Intalg(X)$, then $\mapping{f}{X}{[X]^{<\kappa}}$ defined by
\[ f(x)=\bigcup\setof{ep(b)}{b\in g((-\infty,\,x))}
\]\noindent
is a $\kappa$-FN mapping on $X$.\medskip\\
\assertof{d}: We have ${\rm Ind}(Intalg(X))=\aleph_0$ (see e.g.\ Corollary
15.15 in \cite{koppelbook}) Hence, by \Corof{ind-ineq},
$\cardof{X}\leq\cardof{Intalg(X)}\leq 2^{<\kappa}$ holds if $X$ (or
equivalently $Intalg(X)$) has the $\kappa$-FN.
\qedofProp
\noindentafterqed
Note that \Lemmaabove,\,\assertof{c}
is not true for tree algebras: e.g., for any cardinal
$\lambda$ \st\ $\lambda>2^{\aleph_0}$, the tree $(\kappa,\emptyset)$
has the WFN but $Treealg((\kappa,\emptyset))$ does not by \Corof{Lutz}.
\par
For any sets $x$, $y$ we say that $x$ is a subset of $y$ modulo
$<\kappa$ (notation: $x\subseteq_{<\kappa}y$) if
$\cardof{x\setminus y}<\kappa$ holds. The following lemma is
well-known:
\begin{Lemma}\label{tower-in-omega-2}Suppose that $\kappa$ is a regular
uncountable
cardinal.\smallskip\\
\assert{a} If $(u_\alpha)_{\alpha<\kappa}$ is a sequence of non-stationary
subsets of $\kappa$ then there exists a non-stationary
$u\subseteq\kappa$ \st\ $u_\alpha\subseteq_{<\kappa}u$ holds for all
$\alpha<\kappa$. \smallskip\\
\assert{b} There exists a strictly $\subseteq_{<\kappa}$-increasing sequence
of elements of $\powersetof{\kappa}$ of order type $\kappa^+$.
\end{Lemma}
\prf
\assertof{a}: For each $\alpha<\kappa$, let
$c_\alpha\in\powersetof{\kappa}$ be a club subset of $\kappa$ \st\
$u_\alpha\cap c_\alpha=\emptyset$.
Let $(\beta_\alpha)_{\alpha<\kappa}$ be a strictly, continuously
increasing sequence in
$\kappa$ \st\ $\beta_\alpha\in\bigcap_{\delta<\alpha}c_\delta$ for every
$\alpha<\kappa$. Then
$\kappa\setminus\setof{\beta_\alpha}{\alpha<\kappa}$ is as
desired.\smallskip\\
\assertof{b}: We can construct a $\subseteq_{<\kappa}$-increasing sequence
$(u_\alpha)_{\alpha<\kappa}$ of elements of $\powersetof{\kappa}$
inductively so
that $u_\alpha$ is non-stationary for every $\alpha<\kappa$:
for a successor step let $u_{\alpha+1}$ be the union of $u_\alpha$ and any
non-stationary subset of $\kappa\setminus u_\alpha$ of cardinality less than
$\kappa$. For a limit $\delta<\kappa^+$ with
$\cf(\delta)=\lambda<\kappa$,
we choose increasing
$(\delta_\beta)_{\beta<\lambda}$ \st\
$\delta=\bigcup_{\beta<\lambda}\delta_\beta$, and let
$u_\delta=\bigcup_{\beta<\lambda}u_{\delta_\beta}$. For limit
$\delta<\kappa^+$ with $\cf(\delta)=\kappa$, we can take an appropriate
$u_\delta$ using \assertof{a}.
\qedofLemma
\begin{Prop}\label{P(omega-n)}Suppose that $\kappa$ is a regular uncountable
cardinal. Then\smallskip\\
\assert{a}$\powersetof{\kappa}/[\kappa]^{<\kappa}$ does not have the
$\kappa$-FN. \smallskip\\
\assert{b} $\powersetof{\kappa}$ does not have the $\kappa$-FN.
\end{Prop}
\prf
\assertof{a}: By \Lemmaof{tower-in-omega-2},\,\assertof{b},
$(\kappa^+,\,\leq)$ is
embeddable into $\powersetof{\kappa}/[\kappa]^{<\kappa}$. Hence,
by \Propof{omega-2}, $\powersetof{\kappa}/[\kappa]^{\kappa}$ does not have the
$\kappa$-FN. \smallskip\\
\assertof{b}:
Let $\chi$ be sufficiently large and let $M\prec\calH_\chi$ be
$V_{\kappa}$-like \st\ $\kappa\in M$ (and hence also
$\powersetof{\kappa}\in M$). Let
$(M_\alpha)_{\alpha<\kappa}$ be an increasing
sequence of elementary submodels of $M$ of cardinality less than $\kappa$
witnessing the
$V_{\kappa}$-likeness of $M$. We construct a
sequence
$(u_\alpha)_{\alpha<\kappa}$ of non-stationary subsets of $\kappa$
inductively \st\ $u_\alpha\in M_{\alpha+1}$,
$u\subseteq_{<\kappa}u_\alpha$
and $\cardof{u_\alpha\setminus u}=\kappa$ hold for every non-stationary
$u\in\powersetof{\kappa}\cap M_\alpha$.
This is possible since $M_\alpha\in M_{\alpha+1}$. By
\Lemmaof{tower-in-omega-2},\,\assertof{a}, there is a non-stationary
$u^*\in\powersetof{\kappa}$ \st\ $u_\alpha\subseteq_{<\kappa}u^*$
holds for every $\alpha<\kappa$. Clearly
$(\powersetof{\kappa}\cap M)\restr u^*$ is not generated by any subset of
cardinality less than $\kappa$.
Hence, by \Propof{game}, it follows that $\powersetof{\kappa}$ does not
have the $\kappa$-FN.
\qedofProp
\indentafterqed
By \Propof{P(omega-n)}, it follows that $\powersetof{\omega_1}$ does not
have the WFN. In contrast to this, the
statement ``$\powersetof{\omega}$ has the WFN'' is independent
from \ZFC\ or even from \ZFC\ $+$ $\neg$\CH.
For $x\in\powersetof{\omega}$, let us denote by $[x]$ the equivalence
class of $x$ modulo fin (the ideal of finite subsets of $\omega$).
\begin{Lemma}\label{fin}
$\powersetof{\omega}$ has the WFN if and only if
$\powersetof{\omega}/fin$ has the WFN.
\end{Lemma}
\prf
If $g$ is a WFN mapping on
$\powersetof{\omega}$, then
$\mapping{g'}{\powersetof{\omega}/fin}{%
[\powersetof{\omega}/fin]^{\leq\aleph_0}}$
defined by
\[ g'([x])
=\bigcup\setof{\setof{[y]}{y\in g(z)}}{z\in\powersetof{\omega},\,
[z]=[x]}
\]\noindent
for $x\in\powersetof{\omega}$, is a WFN mapping on $\powersetof{\omega}/fin$.
If $f$ is a WFN mapping on $\powersetof{\omega}/fin$, then
$\mapping{f'}{\powersetof{\omega}}{[\powersetof{\omega}]^{\leq\aleph_0}}$
defined by
\[ f'(x)=\setof{z}{[z]\in f([x])}
\]\noindent
for $x\in\powersetof{\omega}$, is a WFN mapping on $\powersetof{\omega}$.
\qedofLemma
\begin{Prop}\label{P(omega)}
\assert{a}{\rm(\CH)} $\powersetof{\omega}$ has the WFN.\smallskip\\
\assert{b} In the generic extension of a model of\/ \ZFC $+$ \CH\ by adding
less than $\aleph_\omega$ many
Cohen reals (by standard Cohen forcing),
$\powersetof{\omega}$ still has the WFN. In
particular the assertion
``\/$\MA(\mbox{\it Cohen})$ $+$ $\neg$\CH\ $+$
$\powersetof{\omega}$ has the WFN'' is consistent.
Here $\MA(\mbox{\it Cohen})$ stands for Martin's axiom restricted to
\pos\ of the form $\Fn(\kappa,2)$.
\smallskip\\
\assert{c} If ${\bf b}\geq\aleph_2$, then $\powersetof{\omega}$ does not
have the WFN.
\end{Prop}
\prf
\assertof{a}: Since $\cardof{\powersetof{\omega}}=\aleph_1$ under \CH,
the claim follows from \Lemmaof{aleph-1}.\smallskip\\
\assertof{b}:
This follows from \Thmof{cohen-model} below.
\iffalse
Let $V$ be a ground model of \ZFC\ $+$ \CH. By induction on
$n\in\omega$ we show that:
\begin{assertion}{$(*)_n$} If $G$ is $V$-generic over
$\Fn((\aleph_n)^V, 2)$, then, in $V[G]$, $\powersetof{\omega}^{V[G]}$ has the
WFN.
\end{assertion}
For $n\leq1$ this is clear since then $V[G]$ still satisfies the \CH. Suppose
that we have shown $(*)_m$ for every $m<n$. Let $G$ be $V$-generic over
$\Fn((\aleph_n)^V,2)$. In $V[G]$, let $B=\powersetof{\omega}^{V[G]}$ and
$A_\alpha=\powersetof{\omega}^{V[G_\alpha]}$ for $\alpha<(\aleph_n)^V$,
where $G_\alpha=G\cap\Fn(\alpha,2)$. Then it is easily seen that
$(A_\alpha)_{\alpha<(\aleph_n)^V}$ and $B$ satisfy the conditions
in \Lemmaof{non-conti-chain} (in $V[G]$). It follows that $B$ has the WFN
in $V[G]$.
\fi
\smallskip\\
\assertof{c}: By \Lemmaof{fin}, it is enough to show that
$\powersetof{\omega}/fin$ does not have the WFN. But under
${\bf b}\geq\aleph_2$, $(\omega_2,\leq)$ can be embedded into
$\powersetof{\omega}/fin$. Hence, by \Propof{omega-2},
$\powersetof{\omega}/fin$ does not have the WFN.
\qedofProp
\noindentafterqed
Note that, by \Propabove,\,\assertof{c}, the
statement ``$\powersetof{\omega}$ has the WFN'' is not consistent
with \MA({\it$\sigma$-centered\/}) $+$ $\neg$\CH.
\section{Complete \Bas}
\begin{Lemma}\label{P(kappa)embedded}
For \Bas\ $A$, $B$ and regular $\kappa$, if $A\leq B$, $A$ is
complete (but not necessarily a complete subalgebra of $B$) and $B$ has the
$\kappa$-FN, then $A$ also has the $\kappa$-FN.
\end{Lemma}
\prf
By Sikorski's Extension Theorem, there is a homomorphism $j$ from $B$ to
$A$ \st\ $j\restr A=id_A$. Hence the claim of the lemma follows from
\Lemmaof{retract}.
\qedofLemma
\begin{Thm}\label{complete-Ba}
\assert{a} If ${\bf b}\geq\aleph_2$, then no infinite complete \Ba\ has the
WFN.\smallskip\\
\assert{b}
For regular $\kappa$, no \cBa\ without the $\kappa$-cc has the
$\kappa$-FN.\smallskip\\
\assert{c} If $\kappa$ is regular and $2^{<\kappa}=\kappa$, then every
$\kappa$-cc \cBa\ has the $\kappa$-FN.
\end{Thm}
\prf
\assertof{a}: If $B$ is complete, then $\powersetof{\omega}$ is embeddable
into
$B$. Hence, by \Propof{P(omega)},\,\assertof{c} and \Lemmaof{P(kappa)embedded},
$B$ does not have the WFN.\smallskip\\
\assertof{b}: If $B$ is complete and not $\kappa$-cc, then
$\powersetof{\kappa}$ is embeddable into $B$. Hence, by
\Propof{P(omega-n)},\,\assertof{b} and \Lemmaof{P(kappa)embedded},
$B$ does not have the $\kappa$-FN.\smallskip\\
\assertof{c}: Assume that $B$ has the $\kappa$-cc.
Let $\chi$ be sufficiently large and let $M\prec\calH_\chi$ be \st\
$\cardof{M}=\kappa$, $B\in M$ and $[M]^{<\kappa}\subseteq M$ hold.
Then $B\cap M$ is a $\kappa$-complete subalgebra of $B$. By the $\kappa$-cc
of $B$ it follows that $B\cap M$ is complete subalgebra of $B$. Hence
$B\cap M\leq_\rc B$ holds.
By \Propof{char-under-ch} it follows that $B$ has the $\kappa$-FN.
\qedofThm
\indentafterqed
By \assertof{b} and \assertof{c} in \Thmabove, under \CH,
a complete \Ba\ $B$ has the WFN if and only if $B$ satisfies the
\ccc.
This assertion still holds partially in the model obtained by adding Cohen
reals to a model of \CH:
\begin{Thm}\label{cohen-model}
Suppose that $V\models{\ZFC+\CH}$. Let $\lambda$ be a cardinal in
$V$ \st\ $V\models{\lambda<\aleph_\omega}$. Let $P=\Fn(\lambda, 2)$ and let
$G$ be a $V$-generic filter $G$ over $P$. For $B\in V[G]$ \st\
$V[G]\models{B\xmbox{ \it is \ccc\ \cBa}}$, if either\smallskip\\
\assert{a} $V[G]\models{\cardof{B}<\aleph_\omega}$ or \\
\assert{b} there is a \Ba\ $A\in V$ \st\
$B=\overline{A}^{V[G]}$\smallskip\\
$($where $\overline{A}^{V[G]}$ denotes the completion
of $A$ in $V[G]$\relax$)$, then we have:
\[ V[G]\models{B \mbox{ has the WFN}}.
\]\noindent
\end{Thm}
For the proof of \Thmabove\ we need the following lemma.
\begin{Lemma}\label{Hilfslemma}
Let $V$ be a ground model and $P=\Fn(S,2)$ for some $S\in V$. Let $G$ be
$V$-generic over $P$ and $A\in V$ be \st\
\[ V\models{A\mbox{ is a \ccc\ \cBa}}.
\]\noindent
Then we have:\\
\assert{a} if $\calS\in V$ is \st\
\[
V\models{\calS\mbox{ is a }\sigma\mbox{-directed family of subsets of }S
\mbox{ and }\bigcup\calS=S}
\]\noindent
then $\overline{A}^{V[G]}=\bigcup\setof{\overline{A}^{V[G_T]}}{T\in\calS}$
where $G_T=G\cap \Fn(T,2)$;\smallskip\\
\assert{b} For a \Ba\ $B$ in $V[G], $ if $V[G]\models{A\leq B}$, then
$V[G]\models{ A\leq_\sigma B}$.
\end{Lemma}
\prf
\assertof{a}: If $b\in\overline{A}^{V[G]}$ then, by the \ccc\ of $A$, there
is a countable $X\subseteq A$ ($X\in V[G]$) %
\st\ $b=\Sigma^{\overline{A}^{V[G]}}X$.
Hence, by the \ccc\ of $\Fn(S,2)$ (in $V$), there is a name
$\dotb$ of $b$ in which only countably many elements of $\Fn(S,2)$
appear. Let $T\in\calS$ be \st\ every element of
$\Fn(S,2)$ appearing in $\dotb$ is contained in $\Fn(T,2)$.
Then $b\in V[G_T]$, hence $b\in \overline{A}^{V[G_T]}$.\smallskip\\
\assertof{b}:
It is enough to show that $V[G]\models{A\leq_\sigma\overline{A}^{V[G]}}$.
Let $b\in\overline{A}^{V[G]}$. Then, as in the proof
of \assertof{a}, there exists a countable $X\subseteq S$ (in $V$) %
\st\ $b\in V[G_X]$ where $G_X=G\cap\Fn(X,2)$. Let $\dotb$ be an
$\Fn(X,2)$ name of $b$. In $V[G]$, let
\[ I=\setof{\Sigma^A\setof{c\in A}{p\forces{\Fn(X,2)}{c\leq\dotb}}}{p\in G_X}.
\]\noindent
Then $I$ is countable and generates $A\restr b$.
\qedofLemma
\noindentafterqed
\prfof{\Thmof{cohen-model}}
First, let us consider the case \assertof{b}.
It is enough to show the following assertion for all \ccc\ \Bas\ $A$ in $V$
and for all $n\in\omega$:
\begin{assertion}{$(*)_n$}\it
If $H$ is $V$-generic over $\Fn(\aleph_n,2)$ then $\overline{A}^{V[H]}$ has
the WFN in $V[H]$.
\end{assertion}
For $n\leq 1$, $V[H]$ still satisfies the \CH. Hence, by
\Thmof{complete-Ba},\,\assertof{c}, $\overline{A}^{V[H]}$ has the WFN in
$V[H]$. Now suppose that $n>1$ and we have shown $(*)_m$ for all $m<n$. Let
$H$ be a $V$-generic filter over $\Fn(\aleph_n,2)$. For each
$\alpha<\aleph_n$, let $H_\alpha=H\cap\Fn(\alpha,2)$. Then by the induction
hypothesis and \Lemmaof{Hilfslemma},\,\assertof{a} and \assertof{b}, the
sequence $(\overline{A}^{V[H_\alpha]})_{\alpha<\aleph_n}$ satisfies the
conditions of \Lemmaof{non-conti-chain}. Hence
$\overline{A}^{V[H]}=\bigcup_{\alpha<\aleph_n}\overline{A}^{V[H_\alpha]}$ has
the WFN in $V[H]$. \medskip\par
Now suppose that $B$ is a \ccc\ \cBa\ in $V[G]$ \st\
$V[G]\models{\cardof{B}<\aleph_\omega}$.
\Wolog\ we may assume that the underlying set of $B$ is a cardinal
$\kappa<\aleph_\omega$. Let $\dot\leq$ be a
$P$-name of the partial ordering of $B$. In $V$, let
$u_{\alpha,\beta}\subseteq\setof{p\in P}{%
p\xmbox{ decides }\alpha\mathrel{\dot\leq}\beta}$ be maximally pairwise
disjoint for each $\alpha,\beta\in\kappa$ (more precisely the family
$(u_{\alpha,\beta})_{\alpha,\beta\in\kappa}$ should be taken in $V$).
Since $P$ satisfies the \ccc, we have
$\cardof{u_{\alpha,\beta}}\leq\aleph_0$. Further in $V$, let
\[
\begin{array}{@{}l@{}l}
I=\setof{(r,s)}{%
& r\in[\kappa]^{\aleph_1},\,s\in[\lambda]^{\aleph_1},\\
& \forces{P}{r\mbox{ is closed \wrt\ Boolean operations}}\\
& \mbox{ and }u_{\alpha,\beta}\subseteq\Fn(s,2)
\mbox{ for all }\alpha,\beta\in r
}.
\end{array}
\]\noindent
For $(r,s),\,(r',s')\in I$, let
\[ (r,s)\leq(r',s')\ \ \Leftrightarrow\ \ r\subseteq r'\mbox{ and }
s\subseteq s'.
\]\noindent
By the definition of $I$, we have $B\cap r\in V[G_s]$ for $(r,s)\in I$ where
$B\cap r$ is the subalgebra of $B$ with the underlying set $r$ and,
as before, $G_s=G\cap\Fn(s,2)$. Let
$B_{(r,s)}=\overline{B\cap r}^{V[G_s]}$ and let $\dotB_{(r,s)}$ be its
$\Fn(s,2)$-name.
Since $V[G_s]\Models{\CH}$, we have
$V[G_s]\models{\cardof{B_{(r,s)}}=\aleph_1}$ for every $(r,s)\in I$.
The rest of the proof is modeled after the proof of
\assertof{4}\implies\assertof{1} of \Propof{characterization}.
For directed $I'\subseteq I$, let $B_{I'}=\bigcup_{(r,s)\in I'}B_{(r,s)}$.
\addtocounter{Thm}{-1
\begin{Claim}
If $I'\subseteq I$ is \st\ $I'\in V$ and
$V\models{I'\xmbox{ is }\sigma\xmbox{-directed}}$, then
$V[G]\models{B_{I'}\leq_\sigma B}$.
\end{Claim}
\prfofClaim
Otherwise there is some $b\in B$ \st\ $B_{I'}\restr b$ is not countably
generated in $V[G]$. Let $\dotb$ be a $P$-name of $b$. In $V$, we can
construct an increasing sequence $(r_\alpha,s_\alpha)_{\alpha<\omega_1}$ of
elements of $I'$ \st\
\[ \forces{P}{\dotB_{(r_\alpha,s_\alpha)}\restr\dotb\mbox{ does not generate }
\dotB_{(r_{\alpha+1},s_{\alpha+1})}\restr\dotb}.
\]\noindent
Note that this is possible because of the \ccc\ of $P$. Let
$r^*=\bigcup_{\alpha<\omega_1}r_\alpha$ and
$s^*=\bigcup_{\alpha<\omega_1}r_\alpha$. Then $(r^*,s^*)\in I$. By an
argument similar to the proof of \Lemmaof{Hilfslemma},\,\assertof{a}, we can
show that $B_{(r^*,s^*)}=\bigcup_{\alpha<\omega_1}B_{(r_\alpha,s_\alpha)}$.
Hence, by the construction of $(r_\alpha,s_\alpha)$, $\alpha<\omega_1$, the
ideal $B_{(r^*,s^*)}\restr b$ is not countably generated. This is a
contradiction to \Lemmaof{Hilfslemma},\,\assertof{b}.
\qedofClaim\medskip\\
For $X\in\powersetof{\kappa}\cap V$ and $Y\in\powersetof{\lambda}\cap V$,
let
\[ I_{X,Y}=\setof{(r,s)\in I}{r\subseteq X,\,s\subseteq Y}.
\]\noindent
Note that $I_{X,Y}$ is $\sigma$-directed. By induction on
$\mu\leq\max(\kappa,\lambda)$, we can show easily that
\begin{Claim}
For $X\in\powersetof{\kappa}\cap V$ and $Y\in\powersetof{\lambda}\cap V$, if
$\cardof{X}$, $\cardof{Y}\leq\max(\kappa,\lambda)$ then $B_{I_{X,Y}}$ has
the WFN in $V[G]$. \qed
\end{Claim}
Since $I=I_{\kappa,\lambda}$, it follows from
\Lemmaof{Hilfslemma},\,\assertof{a} that $B_{I_{\kappa,\lambda}}=B_I=B$.
Hence, by \Claimabove, $B$ has the WFN in $V[G]$.
\qedof{\Thmof{cohen-model}}
\addtocounter{Thm}{1}
\begin{Cor}
The assertion ``every Cohen algebra has the WFN'' is consistent with
{\rm\MA({\it Cohen})} $+$ $\neg\CH$. \qed
\end{Cor}
\section{$L_{\infty\kappa}$-free \Bas}\label{Linftykappa-free}
A Boolean algebra $B$ is called {\em$L_{\infty\kappa}$-free}\/ if
$B\equiv_{L_{\infty\kappa}}\Fr\kappa$, i.e.\ if
$B$ is elementary equivalent to $\Fr\kappa$ in the infinitary logic
$L_{\infty\kappa}$. $B$ is
{\em$L_{\infty\kappa}$-projective}\/ if $B\equiv_{L_{\infty\kappa}}C$ for some
projective $C$. It is easily seen that if $B$ is
$L_{\infty\kappa}$-projective then $B\oplus\Fr\kappa$ is
$L_{\infty\kappa}$-free.
In \cite{FuKoTa} it is shown
that for every $\kappa$, there exists an $L_{\infty\aleph_1}$-free \Ba\ $B$
which does not satisfy the $\kappa$-cc. By \Corof{Lutz}, it follows that
\begin{Prop}
For any $\kappa$, there exists an $L_{\infty\aleph_1}$-free \Ba\
which does not have the $\kappa$-FN. \qed
\end{Prop}
On the other hand, we show in \Corof{L-infty-aleph-1} below that
every ccc $L_{\infty\aleph_1}$-free \Ba\ has the WFN.
Let us begin with the
following lemma:
\begin{Lemma}\label{union-of-rc}
If $B$ satisfies the \ccc\ and $(A_\alpha)_{\alpha<\kappa}$ is an
increasing sequence of relatively complete subalgebras of $B$ with
$\cf(\kappa)>\omega$ then $\bigcup_{\alpha<\kappa}A_\alpha\leq_\rc B$ holds.
\end{Lemma}
\prf
Suppose not. Then there is some $b\in B$ without projection on
$\bigcup_{\alpha<\kappa}A_\alpha$. Then, for
$\mu=\cf(\kappa)$, we can construct an increasing sequence of
ordinals
$(\gamma_\beta)_{\beta<\mu}$ \st\ $\gamma_\beta<\kappa$ and
$p^B_{A_{\gamma_\beta}}(b)<p^B_{A_{\gamma_{\beta+1}}}(b)$ holds for
every $\beta<\mu$. But this is impossible since $B$ satisfies the \ccc.
\qedofLemma
\begin{Prop}\label{ccc-rc}
If a \Ba\ $B$ satisfies the \ccc\ and
$\setof{C}{C\leq_\rc B,\,\cardof{C}=\aleph_0}$ is cofinal in
$[B]^{\aleph_0}$ then $B$ has the WFN.
\end{Prop}
\prf
By \Propof{game}, it is enough to show that Player II has a simple winning
strategy in $\kappagame{\omega_1}{B}$. By \Lemmaof{union-of-rc}, Player II
wins if he
takes $Y_\alpha\geq X_\alpha$ in his $\alpha$'th move \st\
$Y_\alpha\leq_\rc B$ holds (actually this is a simple winning strategy of
Player II in $\kappalambdagame{\omega_1}{\aleph_0}{B}$).
\qedofProp
\begin{Cor}\label{L-infty-aleph-1}
Every \ccc\ $L_{\infty\aleph_1}$-projective \Ba\ $B$ has the WFN.
\end{Cor}
\prf
Let $B$ be a \ccc\ $L_{\infty\aleph_1}$-projective \Ba.
The statement ``$\setof{C}{C\leq_\rc B,\,\cardof{C}=\aleph_0}$ is cofinal in
$[B]^{\aleph_0}$'' can be formulated in $L_{\infty\aleph_1}$ and is true in
any projective \Ba. Hence it is also true in $B$.
By \Propof{ccc-rc},
it follows that $B$ has the WFN.
\qedofCor
\indentafterqed
Since the \ccc\ is expressible in $L_{\infty\aleph_2}$ and it is true in any
projective \Ba, every $L_{\infty\aleph_2}$-projective \Ba\ satisfies the ccc.
Hence, by \Corabove, every $L_{\infty\aleph_2}$-projective \Ba\ has the WFN.
Under Axiom R we can obtain a stronger result.
Recall that Axiom R is the following statement:
\begin{assertion}{(Axiom R):}\it
For any $\lambda\geq\aleph_2$ with uncountable cofinality, stationary
$\calS\subseteq[\lambda]^{\aleph_0}$ and a
$\calT\subseteq[\lambda]^{\aleph_1}$ which is closed under union of
increasing chains of order type $\omega_1$, there exists an
$X\in\calT$ \st\ $\calS\cap[X]^{\aleph_0}$ is stationary in $[X]^{\aleph_0}$.
\end{assertion}
Axiom R follows from \MM\ (\cite{beaudoin}) but its consistency
with \CH\ can be also shown under a supercompact cardinal.
The following theorem was proved in \cite{fuchino2} (see also
\cite{appendix}):
\begin{Thm}
{\rm(Axiom R)} Every $L_{\infty\aleph_2}$-projective \Ba\ $B$ has the FN. \qed
\end{Thm}
The theorem above is not provable in \ZFC:
under $V=L$, we can obtain a counter-example to \Thmabove\ (\cite{appendix}).
In contrast to \Corof{L-infty-aleph-1}, we have the following:
\begin{Prop}
For any cardinal $\kappa$, there is a subalgebra of $\Fr\kappa^+$ without the
$\kappa$-FN.
\end{Prop}
\prf
The topological dual to the \Ba\ $B$ below is considered in Engelking
\cite{engelking} to show that there exists a non-projective subalgebra of a
free \Ba\ (in the language of topology, this means that there exists a
dyadic space which is not a Dugundji space).
Let $X$ be a set of cardinality $\kappa^+$. We shall show that there is a
subalgebra of $\Fr X$ without the WFN.
Let $U_1$ and $U_2$ be the ultrafilters of $\Fr X$ generated by $X$ and
$\setof{-x}{x\in X}$ respectively.
Let
\[ B=\setof{b\in\Fr X}{b\in U_1\,\Leftrightarrow\, b\in U_2}.
\]\noindent
Clearly $B$ is a subalgebra of $\Fr X$. We claim that $B$ does not have the
$\kappa$-FN.
For $Y\subseteq X$, let $B_Y=B\cap\Fr Y$.
\begin{Claim}
For every $Y\in[X]^{\kappa}$, $B_Y$
is not a $\kappa$-subalgebra of $B$.
\end{Claim}
\prfofClaim
Let $x_0\in Y$ and let $x_1$, $x_2$ be two distinct elements of $X\setminus Y$.
Let
\[ b=x_0+x_1+-x_2.
\]\noindent
Since
$b\in U_1$ and $b\in U_2$, we have $b\in B$. Let $I=B_Y\restr b$. We show
that $I$ is not $<\kappa$-generated: let $J$ be any subset of
$I$ of cardinality less than $\kappa$. For $c\in J$, we have $c\leq x_0$.
Since $x_0$ is not an element of
$B$, $c$ is strictly less than $x_0$. Let $Y'$ be a subset of
$Y$ of cardinality less than $\kappa$ \st\ $J\subseteq B_{Y'}$. Let $y_1$,
$y_2$ be
two distinct elements of
$Y\setminus Y'$. Let
\[ d=x_0\cdot y_1\cdot-y_2.
\]\noindent
Then we
have $d\not\in U_1$, $d\not\in U_2$ and $d\leq x_0$. Hence $d\in B_Y\restr b$.
But $d$ is incomparable with every non-zero element of $J$.
\qedofClaim
\noindentafterqed
Since there are club many $C\in[B]^{\kappa}$ of the form $C=B_Y$ for
$Y\in[X]^{\kappa}$, it follows that $B$ does not have the $\kappa$-FN by
\Propof{characterization}. Note that the implication
\assertof{1}\implies\assertof{3} in
\Propof{characterization} used here does not require the assumption of
regularity of $\kappa$.
\qedofProp
\indentafterqed
A subalgebra of an openly generated \Ba\ $B$ is openly generated
(i.e.\ has the FN) if and only if $B$ has the Bockstein separation
property (\cite{heindorf-shapiro}).
\begin{Problem}
Is there a \Ba\ with the Bockstein separation property but without the WFN?
\end{Problem}
In \cite{appendix}, the following partial answer to the problem is given:
{\em if $B$ is stable and satisfies the ccc and the Bockstein separation
property,
then $B$ has the WFN.} Here a \Ba\ $B$ is said to be stable if, for
any countable subset $X$ of $B$, only countably many types over $X$ are
realized in $B$.
|
\section*{Acknowledgements}
We are grateful to Z. Silagadze for correspondence on these matters and to J.
McGovern for critically reading the manuscript. This work is supported by the
EPSRC and PPARC.
|
\section*{\large\bf 1. Introduction}
\label{sec-intr}
Let us consider a spherically expanding nuclear system in the
metastable
nuclear fluid phase when it reaches the freeze-out at time $\tau_{fr}$.
Although at the freeze-out the fermionic degrees of freedom are frozen-out,
and internucleon collisions cease, softer long range nuclear interactions are
still effective, and represented by a nuclear mean field potential,
$U(\vec{r})$.
We will assume that the system, both before and after freeze-out
undergoes a spherical, scaling expansion. Such an expansion can be
represented by a four-velocity field, $u_\mu = x^\mu / \tau$, where
$\tau = \sqrt{ t^2 - x^2 - y^2 -z^2 }$, for the internal regions
of our expanding system (but not for the external surface).
This flow pattern is invariant under
Lorentz transformation, i.e., the points of the interior of our expanding
system are physically identical and indistinguishable from one another.
Consequently in the interior, in the Local Rest (LR) frame all thermodynamical
and fluid-dynamical quantities are equal, and from the point of view of
instabilities all internal points are equivalent. We will exploit these
symmetry features although in the calculations we will use a non-relativistic
approximation.
Let us assume that the nucleon phase space distribution
before, and at the freeze-out, $\tau_{fr}$, is a Fermi distribution:
\begin{equation}
f_0(\tau_{fr},\vec{r},\vec{p}) =
C \left\{ 1 +
\exp\left[\frac{[\vec{p} - m \vec{r}/\tau_{fr}]^2}{2mT_{fr}} -
\frac{\mu_{fr}}{T_{fr}}\right]
\right\}^{-1}
\label{e:fd}
\end{equation}
where this form assumes a correlation between the momentum and radial
distribution arising from radial expansion, $C=g/(2\pi \hbar)^3$ is the
normalization, and $g$ is the degeneracy of nucleons, so that
the proper (LR) density is
$n_0(\tau_{fr},\vec{r})=\int d^3p\ f_0(\tau_{fr},\vec{r},\vec{p})$.
We assume that in the interior of the collision zone the freeze-out
density is constant: $n_0(\tau_{fr},\vec{r}) = n_{fr}$.
In the center of the collision zone this is an
ideal Fermi distribution, while at finite radii,
$|\vec{r}\hspace*{2pt}|$, the distribution is
boosted (using non-relativistic, Galilei transformation), with a radially
directed and radially linearly increasing flow velocity of $\vec{v} =
\vec{r}/\tau$. We can also introduce the LR momentum:
$\vec{P}(\tau , \vec{r}\hspace*{2pt}) = \vec{p} - m \vec{r}/\tau$.
Furthermore, let us assume that after the freeze-out for $\tau>\tau_{fr}$,
the system expands homogeneously according to the
collisionless Vlasov equation. The distribution function, $f$, is the
solution of the Vlasov equation with a mean-field
potential, $U(\vec{r})$,
\begin{equation}
\frac{\partial f}{\partial \tau} + \frac{ \vec{p} }{ m }
\frac{\partial f}{\partial \vec{r} } -
\frac{\partial U}{\partial \vec{r}}\
\frac{\partial f}{\partial \vec{p}}\ = 0 .
\label{e:vl}
\end{equation}
In the special case of a homogeneous system, where the last term vanishes, for
such a free coasting expansion in the local rest frame is just obtained by
replacing $\vec{p}$ by $\tau \vec{p} / \tau_{fr}$ in Eq. (\ref{e:fd})
\cite{csm95,csmm95}:
\begin{equation}
f_0(\tau,\vec{r},\vec{p}) =
C \left\{ 1 +
\exp\left[\frac{[\vec{p} - m \vec{r}/\tau]^2}{2\ m\ T_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right]
\right\}^{-1}
\label{e:f0}
\end{equation}
where
$T_{eff} = T_{fr} (\tau_{fr}/\tau)^2$ and $\mu_{eff} =
\mu_{fr} (\tau_{fr}/\tau)^2$.
The condition, that the ratio of the chemical potential and the temperature
is constant during the expansion, in a usual thermodynamical system,
corresponds to an adiabatic process.
The density of the system changes with time in this inertial expansion
as $n_0(\tau) = n_{fr} (\tau_{fr}/\tau)^3$. This solution is valid starting
from the freeze-out, $\tau_{fr}$, and until inhomogeneities will spontaneously
develop in the system at some threshold time, $\tau_{th}$. Before this time
small perturbations will smooth out due to the mean field potential, while
after this threshold time the density dependent mean field will enhance
fluctuations. So after this threshold time the density will not be
homogeneous any more.
Note that this post freeze-out distribution, $f$,
is not a thermal equilibrium distribution
function, and the effective parameters, $T_{eff}$ and $\mu_{eff}$ are just
carrying the memory of the last equilibrium thermal parameters, $T_{fr}$ and
$\mu_{fr}$, but these are not the usual thermodynamical parameters. This can
be
easily seen if the expansion is not spherically symmetric
\cite{csm95}.
If we would have
a thermal expansion following $\tau_{fr}$, the Equation of State (EOS) would
determine the time dependence of the physical temperature in an adiabatic
expansion. Generally this would not coincide with $T_{eff} = T_{fr}
(\tau_{fr} /\tau)^2$,
only in the case of a large system, where the flow dominates the energy.
For example if the EOS is that of an ideal Stefan-Boltzmann
gas, ($\partial p /\partial e = c_0^2$, $c_0^2 = 1/3$ and $e = c T^4$), then
$T(\tau) = T_{fr} (\tau/\tau_{fr})^{-3c_0^2}$, which differs from $T_{eff}$.
We study the
stability of the system and the occurrence of instabilities arising from
the mean field. Such an instability may lead to a rapid multifragmentation
of our system.
\section*{\large\bf 2. Instabilities}
\label{sec-inst}
Let us consider a small perturbation in the expanding system.
The amplitude of this perturbation may grow, decrease or
oscillate depending on the conditions. In the presence of such a
perturbation the phase space distribution is:
$$
f(\tau,\vec{r},\vec{p}\ ) = f_0 + f_1 (\tau,\vec{r},\vec{p}\ ),
$$
with the normalization
$n (\tau,\vec{r}\ )=\int d^3p\ (f_0+f_1)$ and
$n_1(\tau,\vec{r}\ )=\int d^3p\ f_1 $,
where the unperturbed density $n_0$ is homogeneous, $\vec{\nabla} n_0 = 0$,
and $f_1$ should be a local spherical perturbation which is a solution of
the Vlasov equation
\begin{equation}
\frac{\partial f_0}{\partial \tau} +
\frac{\partial f_1}{\partial \tau} +
\frac{ \vec{p} }{ m } \left[
\frac{\partial f_0}{\partial \vec{r} } +
\frac{\partial f_1}{\partial \vec{r} }
\right ] -
\vec{\nabla} U\
\left[
\frac{\partial f_0}{\partial \vec{p}}\ +
\frac{\partial f_1}{\partial \vec{p}}\ \right] = 0 .
\label{e:vl3}
\end{equation}
The solution of the Vlasov equation is treated in details in
Ref.~\cite{kn:be88}. Here we separate the Vlasov
equation into two equations, one for $f_0$ and one for $f_1$.
Separating non-vanishing dominant zeroth order terms we get the equation
$$
\frac{\partial f_0}{\partial \tau} +
\frac{ \vec{p} }{ m } \frac{\partial f_0}{\partial \vec{r} } = 0 ,
$$
which is satisfied by $f_0$ as given in Eq. (\ref{e:f0}). The first order terms
yield the linearized equation
$$
\frac{\partial f_1}{\partial \tau} +
\frac{ \vec{p} }{ m } \frac{\partial f_1}{\partial \vec{r} } -
\vec{\nabla} U\
\frac{\partial f_0}{\partial \vec{p}}\ = 0 .
$$
We intend to find perturbations which grow, leading to instabilities of the
system. Some modes of growing perturbations may arise in thermal
surrounding, and their rate is determined by thermal and viscous damping
\cite{vv94,ck92,ckz93,tl80,lt73}.
These are usually slower, and so other faster processes may come into play
also. Here we intend to study non-thermal, growing perturbations, which
may occur after the thermal freeze-out only, but they can be faster
than the thermally damped processes \cite{csm95,np90,ck92,ckz93,cc94,aram}.
The growth
rate of such perturbations is determined by the long range nuclear mean
field potential, $U(\vec{r})$. Usually different non-thermal channels of
instability
open only when a given time is passed after the freeze-out at $\tau_{fr}$, and
we reach a threshold time, $\tau_{th}$.
There are two characteristic time scales in the system: (i) the longer
post freeze-out expansion between $\tau_{fr}$ and $\tau_{th}$, and (ii)
the rapid growth of instability which develops after $\tau_{th}$.
When studying the dynamics of rapidly growing instabilities we can
usually neglect the much slower dynamics of the post freeze-out
expansion.
Different configurations can and should be taken into account when studying
instabilities in a quenched (supercooled) system.
\subsection*{\large\sl 2.1. Plane wave perturbation}
The stability of the Vlasov equation against plane wave
perturbations was examined in detail by different groups
recently \cite{kn:rand,kn:mis}.
If we expand the perturbation, $f_1$, as
\begin{equation}
\label{e:p1}
f_1(\vec{r},\vec{p},t) = \sum_{k}\ f_{k}(p,t) {1 \over{
\sqrt{\Omega} }}\ e^{i\vec{k}\vec{r}} ,
\end{equation}
and search for the solution of $f_{k}(p,t)$ in the form
$$
f_{k}(p,t) = f_{{k,\omega}}(p)\ e^{i\omega t} \quad ,
$$
according to Ref. \cite{kn:rand} we get the dispersion relation
$\omega(k)$
\begin{equation}
\label{eq:om-rand}
1 = {\partial U(k) \over{ \partial n }} \int {d^3p \over{ (2\pi
\hbar)^3 }}\ {(\vec{k}\vec{p}\ )^2 \over{ (\vec{k}\vec{p}\ )^ - m^2\omega^2 }}
\ {\partial \tilde{f} \over{ \partial \epsilon }} \quad ,
\end{equation}
where $\tilde{f}$ is the static uniform solution, and $U(k)$ denotes
the Fourier component of the effective field $U(\vec{r}\ )$. The mode
corresponding to wavenumber $k$ will be
unstable, when the $\omega(k)$ frequency becomes imaginary. It was also
shown in Ref.~\cite{kn:rand}, that for zero temperature the condition of
instability can be written as
\begin{equation}
\label{e:inst}
{2\over 3} \epsilon_{F} + n {\partial U(k) \over{
\partial n }} < 0 \quad ,
\end{equation}
with $\epsilon_{F}$ Fermi kinetic energy. The expression
goes over for $k\to 0$ into the condition of
{\it mechanical instability} (i.e., the compressibility
becomes negative)
$$
\left( {\partial p \over{ \partial n }} \right)_{T=0} =
{\partial \over{ \partial n }} \left( n^2 {\partial \epsilon
\over{ \partial n }} \right) = {2\over 3} \epsilon_{F} + n
{\partial U(n) \over{ \partial n }} < 0 \quad ,
$$
where the potential $U(n)$ depends only on the homogeneous density.
In Ref.~\cite{kn:mis} the instability condition was examined for a one
dimensionally expanding ground state system. The acting force was a Skyrme
force with Yukawa surface term. The resulting instability condition reads as
Eq.~(\ref{e:inst}), where
$$
{\partial U(k) \over{ \partial n }} = -2 \beta
+ \gamma (\sigma + 1) (\sigma + 2) n_0^{\sigma} -
{4 \pi V_0 \over{ \mu^2 + k^2 }} \quad ,
$$
and $\beta$, $\gamma$, $\sigma$, $V_0$ and $\mu$ are the Skyrme and Yukawa
force parameters (see Eq.~(\ref{e:3p-1})).
In case of a uniformly expanding system the same condition
of instability for a plane wave perturbation can also be obtained
in a fashion similar to the spherical case discussed later,
see Appendix A. This approach leads to the following condition
\begin{equation}
1 = - \frac{2 \pi \tau_{fr} m C}{\tau}
\sqrt{2mT_{fr}}\ \frac{\partial U(k)}{\partial n}\
\int\limits_0^\infty
\frac{d y\ \sqrt{y} }{ (y+s) [1+ e^{y-\mu_{fr}/T_{fr}} ] }\quad ,
\label{e:dis1}
\end{equation}
where $s= m \kappa^2 \tau^4 /( 2 k^2 T_{fr} \tau_{fr}^4) $ ,
$\kappa = i \omega$ and
$\partial U(k)/\partial n$ is the same expression as in
Ref.~\cite{kn:mis}.
\subsection*{\large\sl 2.2. Spherical drop/bubble perturbations}
In the following we want to study more realistic perturbations
instead of plane waves.
We consider local spherical bubbles, since we believe these are the
first instabilities which start to grow
\cite{pnb92}. In general spherical perturbations minimize the
surface and surface energy, so these can be formed the earliest.
(On the other hand plane wave perturbations may grow more rapidly
at later stages with stronger driving forces due to the increased
surface.)
Consider a spherical drop with a surface density profile
exponentially decreasing characterized by the parameter $k$,
a central density, $n_{\rm c}$, and radius $R(\tau)$
\begin{equation}
f_1 = f_s
\left\{
\begin{array}{ll}
n_{\rm c} \exp\left[ - k\frac{\tau_{fr}}{\tau}
\left(r - R(\tau)\right) \right] \phantom{\frac{\tau_{fr}}{\tau}} &
r \geq R(\tau)\\
\phantom{-}& \\
n_{\rm c} & r < R(\tau)
\end{array} \right. ,
\label{e:ps5}
\end{equation}
where
$
R(\tau) = R^*\ \frac{\tau}{\tau_{th}}\ e^{\kappa (\tau-\tau_{th})}
$
is the $\tau$ dependence of the radius after the threshold time,
$\tau_{th}$ when the droplet becomes bigger than the critical radius
and it will be able to grow.
We are interested in the initial growth rate of the radius only, so that
in $R(\tau)$ the exponential term can be expanded into a power series for
small $\tau-\tau_{th}$, i.e.,
$$
R(\tau) \approx R^*\ \frac{\tau}{\tau_{th}}\ [1+ \kappa (\tau-\tau_{th})] .
$$
Inserting this approximation into Eq. (\ref{e:ps5}) we get for the
exterior part ($r > R(\tau)$) of the profile:
$$
f_1 = f_s n_{\rm c}
\exp\left( k R^* \frac{\tau_{fr}}{\tau_{th}} \right)\
\exp\left( - k \frac{\tau_{fr}}{\tau} r +
k R^* \frac{\tau_{fr}}{\tau_{th}} \kappa (\tau-\tau_{th}) \right) .
$$
Since we consider small perturbations only, where the linearization of
Eq.(~\ref{e:vl}) holds, this solution is valid only
for a short time after the instability starts to grow.
The dispersion relation will lead to a dynamical growth factor,
$\kappa$, depending both on $k$ and $R$.
(In the plane wave expansion of the perturbation studied in
Ref.~\cite{kn:rand,kn:mis} the opening of the channel of the instability was
indicated when $\omega$ became imaginary.
Thus perturbations preceding $\tau_{th}$ did not grow. Our approach is
basically equivalent to their one presented here, however, we wanted to
emphasize that the instability may grow only after $\tau_{th}$.)
Since $f_s(\tau,\vec{P})$ has a characteristic time dependence
on the slow scale (i) of the post freeze-out expansion, we assume that its
time-derivatives are negligible compared to other time-derivatives of
$f_1(\tau, \vec{r}, \vec{p}\ )$ corresponding to rapid inequilibrium processes
(ii) like\ \ $\exp[\kappa(\tau - \tau_{th})]$. Furthermore, we assume that
$f_s$ depends on the LR momentum, $\vec{P}$ only, i.e., it does not have any
other dependence on $\vec{r}$ other than what is included in $\vec{P}$.
We search for such solutions of the Vlasov equation,
$f_1(\tau,\vec{r},\vec{p}\hspace*{2pt})$ and $f_s(\tau,\vec{P})$. We assume
that
such solutions can be obtained only some time, $\tau$, after the freeze-out
at $\tau_{fr}$, i.e. at $\tau_{th} > \tau_{fr}$. Before this time thermal
processes and thermal damping is dominant which generally lead to
slower nucleation than post-freeze-out processes driven by the
background fields.
We can calculate the critical droplet radius, $R^*_{\rm crit}$. Droplets
smaller than $R^*_{\rm crit}$ tend to disappear, while droplets larger than
$R^*_{\rm crit}$ may
start to grow. Thus we will study the growth rate of critical size droplets.
The critical radius, $R^*_{\rm crit}$ is calculated in Appendix E.
The critical radius, $R^*_{\rm crit}$, should be evaluated when the channel of
instability opens at $\tau_{th}$, and the critical droplet just starts to
grow. In the 3-dimensional scaling expansion the critical radius scales with
the overall scaling, which leads to a quasi-static critical radius of
$R^*_{\rm crit} \tau/\tau_{th}$.
The critical radius, $R^*_{\rm crit}$, depends on the background
nucleon density at the
opening of the instability, $n_0(\tau_{th}) = n_{fr} (\tau_{fr}/\tau_{th})^3$,
or consequently on $\tau_{th}$, furthermore on the surface parameter, $k$,
on the central density, $n_{\rm c}$, and on the parameters of the interaction
potential. The total energy of the system can be simultaneously minimized
by varying $k$ and $n_{\rm c}$, and searching for an extremum as a function of
$R^*$.
As we mentioned the time-scale of the expansion is assumed to be slow
compared to the
time-scale of the instability, so $R^*$ can be considered as a time
independent constant when studying the growth of instability.
The perturbation (\ref{e:ps5}) satisfies the Vlasov equation both
for the exterior and interior region, if an averaged Yukawa potential is used
(see Appendix C).
The dispersion relation for such a spherical perturbation can be written as
\begin{equation}
1 = - \frac{2 \pi \tau_{fr} m C}{\tau} \sqrt{2mT_{fr}}\
\frac{\partial U}{\partial n}\
\int\limits_0^\infty
\frac{d y\ \sqrt{y} }{ (y-S) [1+ e^{y-\mu_{fr}/T_{fr}} ] }
\label{e:dis3}
\end{equation}
(see Appendix B), where $S= m \kappa^2 (R^*)^2 \tau^4
/( 2 T_{fr} \tau_{fr}^2 \tau_{th}^2) $ .
It is easy to see, that for $S=0$ without
the surface term this condition is equivalent to the isothermal mechanical
instability even for $T\neq 0$ (Appendix D), that is the region of dynamical
instability and of mechanical one coincide.
Although the direct $k$-dependence drops out of the dispersion
relation, but since $R^*$ depends on $k$ so the dispersion relation
is still applicable. It is interesting to mention that both
dispersion relations, (\ref{e:dis1},\ref{e:dis3}), yield
the same condition for evaluating the threshold time for the perturbation
which is just on the boundary to be able to grow, i.e., $s=S=0$.
The features of the potential and the Yukawa term in it are vital
in determining the properties of the static, critical droplet or bubble.
Physically we can consider two situations in the
course of final multifragmentation.
Depending on the beam energy, after the initial compression we reach
the most compressed and heated up state with a definite specific entropy.
This stage is then followed by an expansion, which is adiabatic
to a good approximation.
If the final specific entropy is smaller than the critical entropy of the
nuclear liquid-gas phase transition, the expansion will lead to a stretched
(or quenched) liquid state, with density $n_0$ below the
normal nuclear density, $n_N$. The instabilities will lead then to bubble
formation with an interior nuclear gas phase density, $n_1 + n_0 \approx
0.1-0.4 n_N$.
If on the other hand the final specific entropy exceeds the critical entropy,
the expansion will lead to a oversaturated (or quenched) nuclear vapor state,
with density $0.1-0.4 n_0$, below the critical nuclear density. The
instabilities will lead now to the condensation of a nuclear liquid droplet
with an interior nuclear density, $n_1 + n_0 \approx n_N$.
\bigskip
\section*{\large\bf 3. Condition for the instabilities to grow}
\label{sec-cond}
Equations~(\ref{e:dis1},\ref{e:dis3}) determine
the condition for $\kappa$ becoming real, but its sign remains undefined. From
equations~(\ref{e:p1},\ref{e:ps5}) it is easy to see, that the
amplitude of the perturbation will depend on the sign of $\kappa$: positive
$\kappa$-s will cause exponentially increasing perturbation, while
perturbations corresponding to negative $\kappa$ will be damped rapidly. To
determine the sign of the $\kappa$ we have to see, how the energy changes due
to the perturbation. If the configuration with the perturbation acquires
smaller total nuclear energy (the total energy without the flow) than the
unperturbed system can we speak about
growing
instabilities, that is flow can develop to take extra matter into the
perturbation.
In the following we consider density dependent Skyrme forces with an averaged
Yukawa term.
The total nuclear energy of the system can be written as
$$
E = E_{kin}
- \beta \int d^3r\ n^2(\vec{r}\ )
+ \gamma \int d^3r\ n^{\sigma+2}(\vec{r}\ ) -
\hspace*{1.5cm}
$$
\begin{equation}
V_0 \int d^3r_1 d^3r_2 \ n(\vec{r}_1) n(\vec{r}_2)
{e^{-\mu \mid \vec{r}_1 - \vec{r}_2 \mid}
\over{\mid \vec{r}_1 - \vec{r}_2 \mid} } ,
\label{e:3p-1}
\end{equation}
where the values of $\beta$, $\gamma$, $\sigma$, $V_0$ and $\mu$ are the same
as in Ref.~\cite{kn:mis} and summarized in Table 1.
\begin{table}[t]
\begin{center}
\begin{tabular}{||c|c|c|c|c|c||} \hline \hline \\[-4mm]
& $\beta$ (MeV fm$^3$) & $\gamma$ (MeV fm$^{3 ( \sigma + 1 )}$) & $\sigma$ &
$V_0$ (MeV fm) & $\mu$ (fm$^{-1}$) \\ \hline \hline
SOFT & 1051.76 & 1107.41 & 1/6 & 83.5 & 2 \\ \hline
HARD & 365.0 & 808.65 & 1 & 83.5 & 2 \\ \hline \hline
\end{tabular}
\end{center}
\caption{The parameterization of the SOFT and HARD EOS}
\end{table}
Let us consider a system which is initially
homogeneous and has constant density $n_0(\tau)=n_{fr} (\tau_{fr}/\tau)^3$.
Introducing now a small perturbation in the density in a way that the total
mass number has to be conserved, we obtain a density distribution
\begin{equation}
n(\tau,\vec{r}\ ) = n_0(\tau) + n_1(\tau,\vec{r}\ ) - {\Gamma \over{ \Omega }}
\label{e:3p-2}
\end{equation}
where $\Gamma=\int d^3r\ n_1(\tau,\vec{r}\hspace*{2pt})$, $\Omega$ is the
volume of the
system, and $n_1$ is assumed to be small compared to $n_0$.
If the initial configuration is such that the formation of a perturbation
may lead to a decrease of the energy such perturbations will appear
and grow spontaneously. This will lead to a multifragmentation of the system.
We consider here the energy of the system and not the Helmholtz free energy
because we are describing a post freeze-out situation when we do not have a
heath bath any more.
Substituting (\ref{e:3p-2}) into (\ref{e:3p-1}) and expanding in terms of
$n_1$ up to the second order, we evaluate the total energy of the perturbed
system. For large enough systems the terms containing $\Gamma^2/\Omega$ can
also be neglected. Thus the total nuclear energy can be written
as $E=E_0+\Delta E$,
where
\begin{equation}
{E_0\over A} = {E_{\rm {kin}}(n_0) \over A} - \left( \beta +
{4\pi V_0 \over{ \mu^2 }} \right) n_0 + \gamma n_0^2 \quad ,
\label{e:3p-3}
\end{equation}
and
$
A = \int d^3r\ n(\tau,\vec{r}) = \int d^3r\ n_0 = n_0 \Omega .
$
The change of the energy due to the formation of the perturbation
in the second order of $n_1$ (the first order terms cancel due to the mass
number conservation: Eq.~\ref{e:3p-2})
is then
\begin{eqnarray}
\Delta E
&=&
\Delta E_{\rm {kin}}
- \left( \beta
+ {4\pi V_0 \over{\mu^2 }}
- {1\over 2} (\sigma + 1) (\sigma + 2) \gamma n_0^{\sigma}
\right) \overline{N}_1
+ \Delta \varepsilon_{\rm {surf}} \overline{N}_1 ,
\nonumber \\
\Delta \varepsilon_{\rm {surf}}
&=&
\left[ {4\pi V_0 \over{ \mu^2 }} -
{V_0 \over{ \overline{N}_1 }}
\int d^3r_1 d^3r_2\ n_1(\vec{r}_1) n_1(\vec{r}_2)
{e^{-\mu \mid \vec{r}_1 - \vec{r}_2 \mid}
\over{\mid \vec{r}_1 - \vec{r}_2 \mid} }
\right] ,
\label{e:3p-4}
\end{eqnarray}
where
$ \overline{N}_1 = \int d^3r\ n_1^2(\vec{r}) $.
The term $\Delta \varepsilon_{\rm {surf}}$ is the surface correction due to
the Yukawa interaction.
The sign of $\Delta E$ will determine whether an instability
may increase or will be damped.
It is not immediately clear, whether the kinetic energy gives a contribution
to $\Delta E$ or not. In thermal systems at high temperature and low
density, where the exact value of the Fermi momentum is not too important we
can assume that $f = f_0 + f_1$, where $f_1 \approx [n_1(\tau,\vec{r})/n_0]\
f_0$, and the kinetic energy depends only linearly on $n_1$, that is a
density perturbation will not cause a change of total kinetic energy. For a
isotherm, degenerate system, where the kinetic energy is nearly proportional
to $n^{5/3}$, a perturbation may give a contribution to $\Delta E$. Here we
are studying a post freeze-out situation out of thermal equilibrium. Now the
kinetic energy depends on the form of our perturbed phase space distribution
function, $f_s(\tau,\vec{P})$, we choose or obtain. This may have different
characteristics depending on the density of our frozen-out system. Thus we
will examine both situations, the one without the kinetic energy contribution
($\Delta E_1$), and the one with ($\Delta E_2$).
\begin{figure}
\vspace*{-10mm}
\begin{center}
\leavevmode
\epsfxsize=12.6cm\epsfbox{fig-r.ps}
\end{center}
\vspace*{-10mm}
\caption{\small The dependence of the energy difference, $\Delta E$,
caused by the formation of a bubble of radius $R$, without the kinetic term
on the radius of the bubble
for two different diffuseness coefficients, $\alpha=
k \tau_{fr}/\tau$, and two equations of state.
The energy difference is evaluated for three post-freeze-out densities
$n_0/n_N = $ 0.3, 0.4 and 0.7. The maxima of the curves (if any) is at the
critical radius, $R^*$
}
\end{figure}
The surface correction, $\Delta \varepsilon_{surf}$, for the cases of
plane wave and spherical perturbations considered in Section 2 are:
\begin{eqnarray}
{\rm {Case\ A}}\ \quad & {4\pi V_0 \over { \mu^2 }} \left( 1 - {\mu^2 \over{
\mu^2 + k^2 }} \right)
\nonumber \\
{\rm {Case\ B}}\ \quad & ({4\pi V_0 /{ \mu^2 }})
\left[ 1 - g(R^*,k) \right]
\nonumber
\end{eqnarray}
where $g$ is a complicated function of $R^*$ and $k$ (see
Appendix C). For the used forces in Eq.~(\ref{e:3p-1}) in the case A
$\Delta E$ turns out to be as seen in Ref.~\cite{kn:mis}.
$$
\Delta E_{kin} - \left( \beta + {4 \pi V_0 \over{ k^2 + \mu^2 }}
- {(\sigma+1) (\sigma+2) \over 2 } \gamma n_0^{\sigma} \right)
\overline{N}_1 < 0 .
$$
We get a negative change in the energy only for $k < k_{\rm crit}$, where
the critical value of $k$ is
$$
k^2_{\rm crit} = -\mu^2 + {4 \pi V_0 \over{ {\Delta E_{kin} \over{
\overline{N}_1 }}
- \beta + {(\sigma+1) (\sigma+2) \over 2 } \gamma n_0^{\sigma} }}
$$
In case B
the required negativity of $\Delta E$ can be expressed in a more complicated
way.
\begin{figure}[t]
\vspace*{-10mm}
\begin{center}
\leavevmode
\epsfxsize=12.6cm\epsfbox{fig-rk.ps}
\end{center}
\vspace*{-10mm}
\caption{\small The same as Fig. 1 with kinetic energy term included}
\end{figure}
\section*{\large\bf 4. Results}
In the following we present here the results for a spherical droplet
perturbation considered in Section 2.2. The total energy of the system is given
in Eq.~(\ref{e:3p-1}) and the energy change due to the bubble formation in
Eq.~(\ref{e:3p-4}). The surface energy,
$\Delta\varepsilon_{\rm surf}$,
energy given in
Eq.~(\ref{e:as1}) depends strongly on $\alpha = k \tau_{fr}/\tau$ and on the
radius $R^*$ of the bubble. To see the effect of the bubble shape on the energy
change $\Delta E$ we give this change for different $\alpha$ and
density values as the function of the bubble radius.
In Fig.~1 we
assumed that the kinetic energy does not give any contribution in second order
to the energy, in Fig.~2 we considered a ground state Fermi kinetic energy
contribution. As one can see, the effect of the surface term is larger if we
have sharper surfaces (larger $\alpha$ values). As one expects, the effect of
the surface energy,
$\Delta\varepsilon_{\rm surf}$,
decreases for large radii.
As a first step we considered the solution of Eq.~(\ref{e:dis3}) for
$\kappa$=0, and compared the condition of the dynamical instabilities to grow
with that of the mechanical instability for infinite systems.
Without surface term the two condition are the same,
as was pointed out already in Ref.~\cite{kn:rand}. With the surface term
the instability region decreases, just as in Ref.~\cite{kn:mis}. In Fig.~3 we
give the $n$ -- $T$ curve of instability region for different $\alpha$ and
$R^*$ values.
With small supercooling first big droplets can nucleate and grow, then with
stronger supercooling smaller droplets can also be formed.
For big droplets ($R^*$ = 4.5 fm) the effect of the surface term
is almost negligible.
If we wait longer after the freeze-out in the expanding system, i.e., we
increase $\tau$ and thus having a smaller $\alpha$, $(\alpha=\kappa
\tau_{fr}/\tau)$ we have the possibility of instabilities earlier, with
smaller supercooling.
The calculations are done both for soft and hard equation of state.
One sees, that the effect of the surface term is more significant for the soft
equation of state.
As the next step we want to determine the break up of the system starting from
different freeze out densities times and temperatures. A reasonable freeze out
density should be in the range of $n_0=$0.08 -- 0.14 fm$^{-3}$.
In our calculation we
choose $n_{fr}$ = 0.1 fm$^{-3}$. Other freeze out densities can be considered
simply rescaling $\tau_{fr}$, $\mu_{fr}$ and $T_{fr}$ to keep the relation
$(n_0/n_{fr})^{2/3} = (T/T_{fr}) = (\tau_{fr}/\tau)^2$ and
$\mu_{fr}/T_{fr}$ constant (see the remarks after Eq.~(\ref{e:f0})).
The freeze out time
$\tau_{fr}$ defines the flow energy of the system as ${\displaystyle
E_{\rm flow}/A = \frac{3}{5} {R^2 \over{ \tau_{fr}^2 }} }$
for a system with radius $R$. We
assume, that the total excitation energy of the system is large enough to reach
the break up densities \cite{pw95}, so if that condition is fulfilled, the
freeze out time
is not defined in the model, and the time scale is not fixed.
We followed the paths along the trajectories in the $n$ -- $T$ plane
from different initial freeze out configurations.
The instability condition Eq.~(\ref{e:dis3}) is examined along the
trajectories, and the time of the solution for growing instability (break up)
can be expressed as ${\displaystyle \Delta \tau = \tau - \tau_{fr} =
\tau_{fr} \left[ \left( {n_{fr} \over n} \right)^{1/3} - 1 \right] }$
after the freeze out.
\begin{figure}[t]
\vspace*{-10mm}
\begin{center}
\leavevmode
\epsfxsize=12.6cm\epsfbox{fig-inst.ps}
\end{center}
\vspace*{-10mm}
\caption{\small The dynamical and mechanical instability regions.
The region of mechanical instability is bordered by the isothermal
spinodal, where the isotherm sound speed vanishes and becomes imaginary.
For large size (radius $R$) of the critical bubbles or for systems without
surface energy the dynamical instability region is the same as the mechanical
one.
The break up instability depends on the radius $R$
and on the stretching of the system parameterized by $\alpha$}
\end{figure}
In Fig.~4 we show the part of the trajectories
of the post-freeze-out expansion denoted by dotted line where
the instability condition is fulfilled and the energy change, $\Delta E_1$,
is negative. We found, that whenever condition~(\ref{e:dis3}) is
fulfilled, this energy change is always negative. The more strict condition,
using
$\Delta E_2$, however, excludes some part of the trajectories (solid line) at
temperatures above 10 MeV for the hard equation of state. For the soft equation
of state there is no such exclusion, nevertheless,
the instability region is smaller.
We give the results for the two equations of state and different parameters
for both.
For comparison we show the boundary of the isothermal mechanical instability
(which corresponds to the $R^* \to \infty$ situation). In the parameter region
we count as physical ($R^* \approx $ 2--3 fm, $k \approx $ 4 -- 6 fm$^{-1}$)
there are no significant changes on this parameters.
\begin{figure}
\vspace*{-10mm}
\begin{center}
\leavevmode
\epsfxsize=12.6cm\epsfbox{fig-traj.ps}
\end{center}
\vspace*{-10mm}
\caption{\small The trajectories of an adiabatic
post-freeze-out expansion
starting at the same, $n_{fr}=0.1$ fm$^{-3}$ freeze out
density. The different lines are originated from different freeze out
temperatures. The solid lines
correspond to the case where the energy difference with the kinetic term is
negative, the dotted line is the case where the energy difference without the
kinetic term is negative. The latter ones define the wider region.
The possibility of dynamical post-freeze-out instability opens only
after some penetration into the domain of the isothermal spinodal,
while thermally dominated homogeneous nucleation may start immediately,
although slowly
}
\end{figure}
Following the trajectory of
the post-freeze-out expansion starting from a given initial
$n_{fr},T_{fr}$ configuration the
instability condition Eq.~(\ref{e:dis3}) has solutions for different
$S$ or $\kappa$ values. For high densities and temperatures $S$ is negative,
that is there are no real solution for $\kappa$.
As the density (and correspondingly the temperature) decreases it
continuously becomes positive. The speed of growth of the instabilities is
determined by ${\displaystyle \kappa = {\tau_{fr} \over{ \tau }} {\tau_{th}
\over{ \tau }} \sqrt{ {2 T_{fr} S \over{ m (R^*)^2 }} } }$. Evaluating this
expression we assumed that $\tau_{th} / \tau \approx 1$. In Fig.~5
$\kappa$ is shown along given expansion trajectories
($k$=4 fm$^{-1}$ $R$= 2 fm,
$T_{fr}$ = 1 MeV, 16 MeV and 21 MeV) for the hard equation of state.
The speed $\kappa$, the instabilities are developing with, is changing,
first it increases as the nucleus
evolves to smaller densities, later it decreases back to zero.
The time scale of the expansion for the radius of the perturbations
from Fig.~5 gives $\approx$ 20 fm/c for the
$\kappa \approx $ 0.06. However, the growth rate of the instability from
Eq.~(\ref{e:ps5}) is a double exponential: $n_c\ exp\left( k \tau_{fr} /
\tau_{th}\ R^* e^{\kappa \Delta (\tau-\tau_{th})} \right)$, which is much
faster.
This region of the exponentially developing instabilities breaks up the
system.
\begin{figure}[t]
\vspace*{-10mm}
\begin{center}
\leavevmode
\epsfxsize=12.6cm\epsfbox{fig-om.ps}
\end{center}
\vspace*{-10mm}
\caption{\small The growth parameter, $\kappa$, along trajectories
of post-freeze-out expansion starting at $n_{fr}=$ 0.1 fm$^{-3}$ and
at the given freeze-out temperatures
}
\end{figure}
\section*{\large\bf Appendix A:
Dispersion relation for plane wave perturbation}
\label{sec-plane}
Let us consider a plane wave perturbation in another form than in
Ref. \cite{kn:rand} emphasizing our time scale
\begin{equation}
f_1(\tau, \vec{r}, \vec{p}) =
f_s(\tau, \vec{P})\
{\exp \left[ i \frac{\vec{k}\tau_{fr}}{\tau} \vec{r} +
\kappa (\tau-\tau_{th}) \right]} ,
\label{e:pp1}
\end{equation}
where it is taken into account, that the wave number, $\vec{k}$,
scales with $\tau$, as all other parameters of the flow. Thus
$k$ is a constant, independent of $\tau$.
Here $f_1$ is a phase space distribution which should
satisfy the Vlasov equation (\ref{e:vl}), and it represents a plane wave
with growing amplitude if $\kappa$ is a real positive number.
Since we assume small perturbations only, this solution is valid only
for a short time after the instability starts to grow.
(Frequently in similar studies the perturbation is studied in a form
containing an $\exp(i\omega \tau)$ term, and the opening of
the channel of instability is indicated when $\omega$ becomes imaginary.
Thus perturbations preceding $\tau_{th}$ will not grow. This approach is
basically equivalent to the one presented here, however, we wanted to
emphasize that the instability may grow only after $\tau_{th}$.)
As it was already mentioned we also assume that $f_s(\tau,\vec{P})$
has a characteristic time dependence
on the slow scale (i) of the post freeze-out expansion, so that its
time-derivatives are negligible compared to other time-derivatives of
$f_1(\tau, \vec{r}, \vec{p})$ corresponding to rapid inequilibrium processes
(ii) like $\exp[\kappa(\tau - \tau_{th})]$. Furthermore, we assume that
$f_s$ depends on the LR momentum, $\vec{P}$ only, i.e., it does not have any
other dependence on $\vec{r}$ other than what is included in $\vec{P}$.
We search for such solutions of the Vlasov equation,
$f_1(\tau,\vec{r},\vec{p})$ and $f_s(\tau,\vec{P})$. We assume that
such solutions can be obtained only some time, $\tau$, after the freeze-out
at $\tau_{fr}$, i.e. at $\tau_{th} > \tau_{fr}$.
Using the above form, (\ref{e:p1}), of perturbation the density change
arising from this perturbation is:
$$
n_1(\tau,\vec{r}) = n_s(\tau)\ \exp[i \frac{\vec{k}\tau_{fr}}{\tau}
\vec{r} + \kappa (\tau-\tau_{th})] .
$$
Inserting a plane wave perturbation
Eq. (\ref{e:p1})
into the Vlasov equation we obtain for
the force used in Eq.~(\ref{e:3p-1})
$$
f_s(\tau,\vec{P})\ \left[\kappa +
\frac{ i \vec{k}\tau_{fr} }{ m \tau }
\left( \vec{p} - m \vec{r} / \tau \right) \right]
- \hspace*{1.5cm}
$$
\begin{equation}
\frac{\partial U(k)}{\partial n}\
n_s(\tau) \frac{i \vec{k}\tau_{fr}}{\tau} \
\left( - \frac{f_0^2}{C} \right)
\frac{[\vec{p} - m \vec{r}/\tau]}{mT_{eff}}
\exp\left[\frac{[\vec{p} - m \vec{r}/\tau]^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right]
= 0 \quad ,
\label{e:vls}
\end{equation}
where
$$
\frac{\partial U(k)}{\partial n}\ = -2 \beta
+ \gamma (\sigma + 1) (\sigma + 2) \ n_0^{\sigma}
- {4 \pi V_0 \over{ \mu^2 + k^2 }}
$$
as in Ref.~\cite{kn:mis}.
Note that the momentum appears only inside expressions of the
LR momentum,
$\vec{p} - m \vec{r}/\tau$, thus we can integrate it out in the LR frame.
Thus from (\ref{e:vls}) we can express $f_s$, and integrating it
over the LR momentum, $\vec{P} = \vec{p} - m \vec{r} / \tau$,
we obtain $n_s(\tau)$, which then can be eliminated from
both sides yielding:
\begin{equation}
1 = - \frac{1}{T_{eff} C}\ \frac{\partial U(k)}{\partial n}\
\int d^3 P f_0^2(\vec{P})
\frac{i \tau_{fr} \vec{k} \vec{P} / \tau }
{ m \kappa + \frac{i \vec{k}\tau_{fr}}{\tau} \vec{P} }
\exp\left[\frac{\vec{P}^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right]
\label{e:vls2}
\end{equation}
We can separate the variable of the integral to a parallel, $P_{||}$, and
an orthogonal, $\vec{P}_\perp$, component with respect to $\vec{k}$,
and the integration over the components perpendicular to $k$ can be
performed. This yields
\begin{equation}
1 = - 2\pi m C\ \frac{\partial U(k)}{\partial n}
\hspace*{-1mm}
\int\limits_{-\infty}^\infty \hspace*{-1mm} d P_{||}
\frac{i k P_{||}\tau_{fr}/\tau}{m \kappa + i k P_{||} \tau_{fr} / \tau }
\left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right] \right\}^{-1} \hspace*{-4mm} .
\label{e:vls3}
\end{equation}
Only symmetric functions contribute to this integral, so we symmetrize it by
multiplying both the numerator and denominator by
$m \kappa - i k P_{||}\tau_{fr} / \tau $,
and then dropping the antisymmetric term
we end up having
\begin{equation}
1 = - 4\pi m C\ \frac{\partial U}{\partial n}
\hspace*{-1mm}
\int\limits_0^\infty \hspace*{-1mm} d P_{||}
\frac{P_{||}^2 }{ [ m \kappa \tau / (k\tau_{fr}) ]^2 + P_{||}^2 }
\left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right] \right\}^{-1} \hspace*{-4mm} .
\label{e:vls4}
\end{equation}
Introducing a new variable, $y = P_{||}^2 / (2 m T_{eff})$,
a straightforward calculation will lead to the dispersion relation in the
integral form:
\begin{equation}
1 = - 2 \pi m C \sqrt{2mT_{eff}} \ \frac{\partial U}{\partial n}\
\int\limits_0^\infty
\frac{d y\ \sqrt{y} }{ (y+s) [1+ e^{y-\mu_{fr}/T_{fr}} ] },
\label{e:adis1}
\end{equation}
where $s= m \kappa^2 \tau^2 /( 2 k^2 \tau_{fr}^2 T_{eff}) =
m \kappa^2 \tau^4 /( 2 k^2 T_{fr} \tau_{fr}^4) $ .
\section*{\large\bf Appendix B: Dispersion relation for spherical
perturbation}
\label{sec-yukpot}
For the sake of simplicity let us first study a spherical cusp
perturbation
centered around some
interior point $\vec{r}_c(\tau) = \vec{r}_0 \tau / \tau_{fr}$ of the type
\begin{equation}
f_1 = \exp\left[ - k\tau_{fr} |\vec{r} - \vec{r}_0 \tau / \tau_{fr} |/\tau
+ \kappa (\tau-\tau_{th}) \right]
f_s ,
\label{e:ps0}
\end{equation}
where the center of the perturbation moves along the scaling expansion, and
this
center was at point $\vec{r}_0$ at the time of the freeze-out.
As we discussed it in the introduction we can assume that $\vec{r}_0 = 0$,
so without loosing the generality of the assumption, $|\vec{r}-\vec{r}_0|
\longrightarrow |\vec{r}| = r $,
since the interior
points of the expanding nuclear system are equivalent. Thus
\begin{equation}
f_1 = \exp\left[ - k\tau_{fr} r/\tau +
\kappa (\tau-\tau_{th}) \right] f_s ,
\label{e:aps1}
\end{equation}
can serve to study the perturbation just as well.
Although this functional form of perturbation has a singularity in the center
this will not be essential for our study, and it could be removed by assuming
more complicated functional forms for the perturbation which would not show
such a singularity.
However, including the center of the perturbation, $\vec{r}_c$, explicitly
will allow us later to discuss interactions (e.g. fusion, repulsion, etc.)
of two (or more) elementary perturbations. Thus we will follow this somewhat
more complicated derivation although it is not necessary at the moment.
We will not explicitly define the form of $f_s$ at this stage, unlike in the
case of plane wave perturbations. Instead we will consider the integrals of
$f_1$ and $f_s$.
First the norms:
$$
n_1 = \int d^3p\ f_1 = \exp\left[ - k\frac{\tau_{fr}}{\tau} \left| \vec{r} -
\vec{r}_0 \frac{\tau}{\tau_{fr}}\right| + \kappa (\tau-\tau_{th})
\right] n_s ,
$$
and $n_s = \int d^3p f_s $, where we require that $n_s = n_s(\tau)$ does not
depend on $\vec{r}$. Second the projections orthogonal to $(\vec{r} -
\vec{r}_c)$, i.e.,
$$
g_1(\tau, \vec{r}, P_{||}) =
2\pi \int\limits_0^\infty P_\perp d P_\perp f_1 =
\exp\left[ - k\frac{\tau_{fr}}{\tau}
\left|\vec{r} -
\vec{r}_0 \frac{\tau}{\tau_{fr}} \right| +\kappa (\tau-\tau_{th}) \right] g_s .
$$
Here $g_s(\tau, P_{||}) = 2\pi \int P_\perp d P_\perp f_s $, where we require
that $g_s = g_s(\tau, P_{||})$ does not depend on $\vec{r}$. We chose our
coordinate system so that $P_{||}$ is parallel to $(\vec{r} - \vec{r}_c)$,
and $\vec{P}_\perp$ is orthogonal to it. These constraints are the required
implicit constraints on the choice of $f_s$.
Inserting Eq. (\ref{e:ps0}) into the Vlasov equation and integrating it
over $d^2 P_\perp$ we obtain
$$
g_s\
\left[\kappa
+
\frac{ k\tau_{fr} }{ m \tau } \vec{p}
\frac{\vec{r} - \vec{r}_0 \tau/\tau_{fr}}{|\vec{r} - \vec{r}_0 \tau/\tau_{fr}|}
-
\frac{ k\tau_{fr} }{ \tau^2 }
|\vec{r} - \vec{r}_0\tau/\tau_{fr}|
-
\frac{ k \vec{r}_0 }{ \tau }
\frac{\vec{r} - \vec{r}_0 \tau/\tau_{fr}}{|\vec{r} -
\vec{r}_0 \tau/\tau_{fr}|}
\right] -
$$
\begin{equation}
2\pi C n_s \frac{\partial U}{\partial n}\ \frac{k\tau_{fr}}{\tau} \
\frac{\vec{r} - \vec{r}_0 \tau/\tau_{fr}}{|\vec{r} - \vec{r}_0 \tau/\tau_{fr}|}
P_{||}\ \left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right] \right\}^{-1} = 0 ,
\label{e:vls7}
\end{equation}
where for an averaged Yukawa surface term $U$ is only the function of $n$.
Performing products in the first term and taking into account that
$
\vec{P}
\frac{\vec{r} - \vec{r}_0 \tau/\tau_{fr}}{|\vec{r} - \vec{r}_0 \tau/\tau_{fr}|}
$ $= $ $ P_{||}
$
we obtain
$$
g_s\
\left[\kappa
+
\frac{ k\tau_{fr} }{ m \tau } [ \vec{p} - m \vec{r} / \tau ]
\frac{\vec{r} - \vec{r}_0 \tau/\tau_{fr}}{|\vec{r} - \vec{r}_0 \tau/\tau_{fr}|}
\right] = \hspace*{2cm}
$$
$$
g_s\
\left[\kappa
+
\frac{ k\tau_{fr} }{ m \tau } P_{||}
\frac{|\vec{r} - \vec{r}_0 \tau/\tau_{fr}|}{|\vec{r} - \vec{r}_0
\tau/\tau_{fr}|}
\right]
=
g_s\
\left[\kappa
+
\frac{ k\tau_{fr} }{ m \tau } P_{||}
\right] ,
$$
so that Eq. (\ref{e:vls7}) takes the form
\begin{equation}
g_s
\left[\kappa + \frac{ k\tau_{fr} }{ m \tau } P_{||} \right] -
2\pi C n_s \frac{\partial U}{\partial n}\ \frac{k\tau_{fr}}{\tau} \
P_{||}
\left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{eff}}{T_{eff}}\right] \right\}^{-1}
\hspace*{-4mm} =0 .
\label{e:vlq6}
\end{equation}
We see that the position of the center of the perturbation, $\vec{r}_0$,
has dropped out. We indicated this symmetry already in the introduction
when we pointed out that in the spherical scaling expansion all interior
points are equivalent in the sense of their LR features. We would obviously
get the same result assuming $\vec{r}_0 \equiv 0$ from Eq. (\ref{e:ps0})
on in the course of this derivation.
Dividing both sides by
$(\kappa + \frac{ k\tau_{fr} }{ m \tau } P_{||} )$
and integrating over $d P_{||}$ leads to
\begin{equation}
1 = 2\pi C \frac{\partial U}{\partial n}\ \frac{k\tau_{fr}}{\tau}
\hspace*{-2mm}
\int\limits_{-\infty}^\infty
\hspace*{-2mm}
P_{||} d P_{||}
\left[\kappa + \frac{ k\tau_{fr} }{ m \tau } P_{||} \right]^{-1}
\left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{fr}}{T_{fr}}\right] \right\}^{-1} \hspace*{-4mm} .
\label{e:vlq7}
\end{equation}
Multiplying both the denominator and the numerator by
$\kappa - \frac{ k\tau_{fr} }{ m \tau } P_{||} $ and then dropping the
antisymmetric part, which does not contribute to the integral we obtain
\begin{equation}
1 = - 4\pi m C \frac{\partial U}{\partial n}
\hspace*{-1mm}
\int\limits_0^\infty
\hspace*{-1mm}
d P_{||} P_{||}^2
\left[\frac{ m^2 \tau^2 \kappa^2}{ k^2\tau_{fr}^2 } - P_{||}^2 \right]^{-1}
\left\{ 1+ \exp\left[\frac{P_{||}^2}{2mT_{eff}} -
\frac{\mu_{fr}}{T_{fr}}\right] \right\}^{-1} \hspace*{-4mm} .
\label{e:vlq8}
\end{equation}
The same way as we got the dispersion relation in the case of plane wave
perturbation from Eq. (\ref{e:vls4}) we get now the relation
\begin{equation}
1 = - 2 \pi m C \sqrt{2mT_{eff}}\ \frac{\partial U}{\partial n}\
\int\limits_0^\infty
\frac{d y\ \sqrt{y} }{ (y-s) [1+ e^{y-\mu_{fr}/T_{fr}} ] },
\label{e:adis2}
\end{equation}
where $s= m \kappa^2 \tau^2 /( 2 k^2 \tau_{fr}^2 T_{eff}) =
m \kappa^2 \tau^4 /( 2 k^2 T_{fr} \tau_{fr}^4) $ .
If we have a spherical droplet perturbation of a finite
radius described by Eq.~(\ref{e:ps5})
instead of a cusp, Eq.~(\ref{e:ps0}),
the dispersion relation can be obtained from
(\ref{e:adis2}) by making the transformation
$
\kappa \longrightarrow \kappa k R^* \tau_{fr} / \tau_{th}
$ arising from comparing the form of the two perturbations (\ref{e:ps0})
and (\ref{e:ps5}).
This leads to
exactly the same equation as the equation
above (\ref{e:adis2}) for the spherical cusp perturbations,
except that in place of $s$ we have $S = m \kappa^2 (R^*)^2 \tau^4
/( 2 T_{fr} \tau_{fr}^2 \tau_{th}^2) $ in the expression.
\section*{\large\bf Appendix C: Spherical droplet with Yukawa forces}
\label{app-yuk}
For the profile (\ref{e:ps5}) the Yukawa force can be written as follows
\begin{equation}
V_{\rm {Yuk}}(r) = {4\pi V_0 \over{ \mu^2 }} \left[
{\mu\over{r}} e^{-\mu R}
\sinh(\mu r)
\left( {R \over{ \mu }} -
{R \over{ \mu +\alpha }} + {1 \over{ \mu^2 }} -
{1 \over{ (\mu + \alpha)^2 }} \right)
- 1 \right]
\label{e:ay1}
\end{equation}
for $r < R(\tau)$, where $\alpha = k \tau_{fr} / \tau$, and
\begin{eqnarray}
V_{\rm {Yuk}}(r)
&=&
- 4\pi V_0 n_{\rm c} e^{-\alpha (r-R)} \left[
- {1 \over{ \alpha^2 - \mu^2 }} \right.
- {2\alpha \over{ r (\alpha^2 - \mu^2)^2 }}
\nonumber \\
&+&
{ e^{(\alpha - \mu) (r-R)} \over{ 2 r \mu }}
\left(
{R \over{ \mu }} - {1 \over{ \mu^2 }}
+ {R \over{ \alpha -\mu }} + {1 \over{ (\alpha - \mu)^2 }}
\right) \\
&+&
\left.
{e^{(\alpha - \mu) r}\over{2r\mu e^{(\alpha +\mu)R}}} \hspace*{-1pt}
\left({R \over{\mu}} +
{1 \over{\mu^2}}
- {R\over{\alpha +\mu}}
- {1 \over{(\alpha + \mu)^2}}
\right) \hspace*{-2pt}
\right]
\nonumber
\label{e:ay2}
\end{eqnarray}
for $r > R(\tau)$.
The Yukawa energy can be written as
\begin{eqnarray}
E_{\rm {Yuk}}
&=& \int d^3r\ V_{\rm {Yuk}}(\vec{r}) n_1(\vec{r})
\nonumber \\
&=& - {4\pi V_0 \over{ \mu^2 }}
n_{\rm c}^2 {4\pi \over 3}
\left[
R^3
- {3 \over 2} R^2
{\alpha^2 - \alpha \mu -\mu^2 \over{ \mu \alpha (\mu + \alpha) }}
\right.
\nonumber \\
&+&
{3 \over 2} R
\left(
- {2 \over{ \mu }} {1 \over{ (\mu + \alpha) }}
+ {1 \over{ (\mu + \alpha)^2 }}
+ {1 \over{ \alpha^2 }}
\right)
\nonumber \\
&+& {3 \over 2}\ \
\left(
{(2 \alpha + \mu) \mu \over{ 2 \alpha^3 (\mu + \alpha)^2 }}
- {2 \over{ \mu (\mu + \alpha)^2 }}\
+ {1 \over{ \mu^3 }}
\right) \\
&-&
\left.
{3 \over 2} \mu e^{-2 \mu R}
\left(
{R \alpha \over{ \mu (\mu + \alpha) }}
- {1 \over{ (\mu + \alpha)^2 }}
+ {1 \over{ \mu^2 }}
\right)^2
\right] \nonumber
\end{eqnarray}
One sees that substituting expression (\ref{e:ay1}) and (\ref{e:ay2}) into
the Vlasov equation, the perturbation (\ref{e:dis3}) is not a solution of it.
However, for sharply decreasing surfaces $\alpha R > 1$ and
we can consider instead of $V_{\rm Yuk}$ the
average of it.
That is, we use a
surface term, which is given as the average of the Yukawa interaction.
Instead of (\ref{e:ay2}) we introduce for $r > R(\tau)$ (see
Eq.~(\ref{e:3p-1}))
\begin{eqnarray}
\label{e:ayav}
U_{\rm surf} &=& U^0_{\rm surf} n_1(\vec{r})\ =
{4\pi V_0 \over{ \mu^2 }}\ \left[ 1 - \overline{V}_{\rm Yuk} \right] = \\
&=& {4\pi V_0 \over{ \mu^2 }}\ n_1(\vec{r})\
{\alpha \over{ \mu (\mu + \alpha) }}\
{R + {2\alpha + \mu \over{ 2\alpha (\alpha + \mu) }} \over{
R^2 + {R \over{ \alpha }} + {1 \over{ 2\alpha^2 }} }}
\nonumber
\end{eqnarray}
The surface energy term in Eq.~(\ref{e:3p-1}) can be written using
(\ref{e:ay2}) and neglecting the terms proportional to $e^{-2\mu R}$
(these terms are small) as
\begin{eqnarray}
\label{e:as1}
\Delta \varepsilon_{\rm surf} \hspace*{-2mm} &=& \hspace*{-2mm}
{4\pi V_0 \over{ \mu^2 }}\ - {E_{\rm Yuk} \over{ \overline{N}_1 }}
\nonumber \\
\hspace*{-2mm} &=& \hspace*{-2mm}
{1 \over{ \overline{N}_1 }} {4\pi V_0 \over{ \mu^2 }}\ \left[
{R^2 \alpha \over{ 2\mu (\mu + \alpha) }} + {R \over{ (\mu + \alpha)^2 }}
{2\alpha + \mu \over{ 2\mu }} + {4\alpha + \mu \over{
4\alpha (\mu + \alpha)^2 }} -{1 \over{ 2 \mu^3 }} \right]
\nonumber \\
\hspace*{-2mm} &\to& \hspace*{-2mm} {6\pi V_0 \over{ \mu^3 R }}\
{\alpha \over{ \mu + \alpha }}
\quad \mbox{for }\ \alpha R \gg 1\ .
\end{eqnarray}
\section*{\large\bf Appendix D: The mechanical instability region}
\label {sec-isoth}
The total energy density of the system can be written as
$e_{{pot}}(n)
+e_{F}$, with the potential energy density $e_{{pot}}(n)$ and the
kinetic (Fermi)
energy density $e_{F}$. The latter can be written as $e_{F}=const.\ T^{5/2}
F_{3/2}(\mu/T)$, using the integrals $F_{i/2}(\eta) = \int\limits_0^{\infty}\
dx
{x^{i/2} \over{ 1 + \exp(x-\eta) }}$. The density can be expressed as
$n=const.\ T^{3/2} F_{1/2}(\mu/T)$. Isotherm expansion, $dT=0$, leads to
the change of Fermi energy
$
de_{F} = 3 T {F_{1/2}(\mu/T) \over{ F_{-1/2}(\mu/T) }} dn
$
The pressure should be calculated from the free energy density
$f(T,n)=e-Ts$, with
the entropy density $s={5\over 3} e_{F}/T - n \mu/T$:
$
p = n^2 {\partial f/n \over{ \partial n }} = n
{\partial e_{{pot}}(n) \over{ \partial n }} -
e_{{pot}}(n) + {2\over 3} e_{F}
$
The region of mechanical instability where the derivative of the pressure above
with respect to the density at constant temperature is negative:
$
{d p\over{ d n }} = n {\partial^2 e_{{pot}}(n) \over{
\partial n^2 }} + 2 T {F_{1/2}(\mu/T) \over{ F_{-1/2}(\mu/T) }}
= n {\partial U \over{ \partial n }} + 2 T {F_{1/2}(\mu/T) \over{
F_{-1/2}(\mu/T) }}
$
The onset of the instability is determined then by
\begin{equation}
1 = - {1\over 2} c T^{1/2}
{\partial U \over{ \partial n }} F_{-1/2}(\mu/T) \quad , \quad
c = {g \over{ 4 \pi^2 }} \left( {2m \over { \hbar^2 }} \right)^{3/2}
\end{equation}
The condition of the dynamical instability derived in Appendix B is
$
1 = - {2\pi \tau_{fr} m \over{ \tau }} {g \over{ (2\pi \hbar )^3 }}
\sqrt{2 m T_{fr}}
{\partial U \over{ \partial n }} \int\limits_0^{\infty} {dy \sqrt{y}
\over{
(y-s) \left( 1 + \exp(y-\mu_{{eff}}/T_{{eff}}) \right) }}
$
substituting $\tau_{fr}/\tau=\left( T/T_{fr} \right)^{1/2}$ for the
critical mode ($s=0$ growing) one gets
\begin{equation}
1 = - {1\over 2} \left( {2m \over { \hbar^2 }} \right)^{3/2} {g \over{
4 \pi^2 }} T^{1/2}
{\partial U \over{ \partial n }}
F_{-1/2}(\mu_{{eff}}/T_{{eff}}) \quad ,
\end{equation}
which is the same as the condition for the isothermal spinodal.
\section*{\large\bf Appendix E: Critical radius}
Let us introduce the notation
$
\langle n_1^2 \rangle\ \frac{4\pi}{3}\ R^3 \equiv \overline{N}_1
$
and rewrite Eq.~(\ref{e:as1}) as
\begin{equation}
\Delta E_{\rm {surf}} =
{8\pi^2 V_0 \over{ \mu^2 }} {\alpha \langle n_1^2
\rangle \over{ R \mu (\mu +\alpha) }}
R^2 \equiv a(\tau) R^2 .
\label{e33d}
\end{equation}
Using Eq.~(\ref{e:3p-4}) the total energy change due to the formation of a
droplet of size $R$ can be cast in the form
$$
\Delta E (R) = \Delta E_{\rm {kin}}
- \underbrace{\left(
{\beta \over{3}}
+ {4\pi V_0 \over{3 \mu^2 }}
- { (\sigma+1) (\sigma+2) \over 2 } \gamma n_0^{\sigma}
\right)
4 \pi \langle n_1^2 \rangle }_{\equiv b} R^3
+ a R^2.
$$
Thus, assuming that the contribution of kinetic energy is independent of $R$
we obtain the critical radius from $\partial \Delta E / \partial R = 0$
$$
R^*_{\rm crit} = \frac{2 a(\tau)}{ 3 b} .
$$
At $R^*_{\rm crit}$ the function $\Delta E(R) $ has its maximum, thus
bubbles or droplets smaller than $R^*_{\rm crit}$ shrink while larger
than $R^*_{\rm crit}$
grow.
\section*{\large\bf Acknowledgments}
We thank J. Bondorf, I. Mishustin,
W. N\"orenberg
for stimulating discussions.
This work was supported in
part by the Norwegian Research Council (NFR), the Hungarian
Research Foundation OTKA and part by Contract N$^o$ ERB-CIPA-CT92-4023.
One of the authors (J.N) would like to express her thanks to Prof. W.
Greiner and
the University of Frankfurt for their kind hospitality, where part of this work
was done.
\section*{\large\bf Notes}
E-mail: L. P. Csernai: <EMAIL>, http://www.fi.uib.no/\verb+~+csernai;
Judit N\'e\-meth: <EMAIL>;
G. Papp: <EMAIL> http://www.gsi.de/\verb+~+papp.\\
$^\dagger$: Permanent address
|
\section{THE LENS MODEL}
In this letter we will limit the model analysis to gravitational lenses
described by singular isothermal elliptical potentials (Kormann,
Schneider \& Bartelmann 1994) which provide a sufficiently accurate
representation for our purposes. We adopt their definitions and
notations in this letter.
Since we had already developed software for 2-dimensional `disk $+$
bulge' decomposition of MDS galaxy images, we used the same procedure
with a slight modification to do `bulge $+$ gravitational lens'
decomposition of the observed light distribution. This procedure
iteratively converges simultaneously on the maximum likelihood model for
the lensing elliptical galaxy and the source, so as to produce the
observed lensing galaxy and the configuration of images from the lensed
source.
For the light distribution of the elliptical, we used the seven
parameters that we routinely use for the analysis of a galaxy image with
a single component: local sky, location (x,y), total magnitude,
half-light radius, axis ratio and position angle. The lensed source was
defined by four parameters: location (sx,sy) relative to the elliptical,
and intrinsic magnitude and half-light radius. Since the morphology of
the source was very poorly constrained, it was assumed to be bulge-like
with spherical symmetry. The lens is defined by the critical radius
($\xi_0$) and the axis ratio of the mass distribution. An optional
parameter is used to measure the difference in the orientation between
the mass and the observed light of the lens.
Given a set of model parameters, we generate 2-dimensional images for
the elliptical and the source galaxies. The expected configuration of
the lensed images is ray-traced by numerical integration. To improve
the accuracy of the numerical integration, the image is evaluated using
a pixel size smaller than the real one, such that the image fills a 64
pixel square array. The elliptical lens and the lensed source images are
then convolved with the adopted HST WFPC2 point spread function from
TinyTim (Krist 1994). The convolved image is spatially integrated to
the WFC pixel size of 0\farcs1 and is then compared with the observed
galaxy image. The likelihood function is defined as sum over all pixels
of the logarithm of the probability of the observed value, assuming that
it has a Gaussian error distribution with respect to the model. This is
similar to a weighted $\chi^2$, and is then minimized using a
quasi-newtonian method.
\section{MAXIMUM LIKELIHOOD ESTIMATES}
The results of the maximum likelihood estimates and errors are
summarized in Table 1 and the corresponding images are shown in Figure~1
for both of the galaxies analyzed. The rms errors are estimated from
the covariance matrix which is derived by inverting the Hessian at the
maximum of the likelihood function. Note that the model parameters
fitted independently to the F814W and F606W band images are very similar
and well within the statistical errors and expected variations due to
color. Since the fainter lensed images are blue and better resolved in
F606W, the error estimates are smaller in this filter, even though the
image exposure times in the F814W filter are 50\% longer.
The critical radii of HST14176+5226 and HST12531+2914 are 1\farcs5 and
0\farcs6 , in each case larger than the half light radii of 1\farcs2 and
0\farcs2 for these galaxies respectively. The lensed image components
are more distant from the centroid of the lensing elliptical galaxy
along the minor axis because the deflection is proportional to the
gradient of the potential. The ratio in the distance between components
at opposite ends of the cross is equal to the inverse axis ratio of the
potential. The distance of the intrinsic source from the centroid of the
lens needs to be less than about 0.15 of the critical radius for the
creation of a quadruple image. The impact parameters for these two
objects are about 0.08 and 0.09 of the critical radius.
In both cases the source has a half-light radius which is smaller than a
WFC 0\farcs1 pixel. However, the fact that the maximum likelihood
estimate converged to a small but finite intrinsic source half-light
radius appears to indicate that it is extended. We find such a value for
the half-light radius to be similar to those of field galaxies at the
magnitude of the lensed source (see Casertano \etal 1995, Im \etal
1995), i.e. within the range $0\farcs04-0\farcs40$.
\begin{figure}[bt!]
\psfig{file=fig1.eps,width=5.0in}
\caption{The observed light distribution of HST14176+5226, the
maximum likelihood potential according to the elliptical mass model, and
the resulting Mass/Light distribution, which is rounder (axis ratio=0.82) and
increases radially outwards. All the contours have been drawn at
constant 0.5 mag intervals, and units are 0\farcs1 WFC pixels. }
\end{figure}
A significantly better fit (99.9\%) is obtained for HST14176+5226 by
relaxing the constraint that the orientation of the lens mass is the
same as that of the observed light. The difference is practically
independent of the model used, and is 8\degpt5$\pm$0\degpt8 in F814W and
12\degpt9$\pm$0\degpt2 in F606W. Such a small rotation might be
expected from projection effects (see Franx, Illingworth \& de Zeeuw 1991).
As shown in Figure 2, the axis ratio
(0.68) of the light from the elliptical galaxy HST14176+5226 is
practically the same as that of the potential (0.74). The inferred
elliptical mass distribution is significantly flatter than this (0.40).
Using the observed V{$-$} I colors, apparent magnitudes and half-light
radii, we can estimate the redshift for each of the lens elliptical
galaxies to be $z(D_d)=0.7\pm 0.1$ (see, e.g., Connolly \etal 1995).
Using these redshifts, the luminosities of the elliptical galaxies can
be estimated for the plausible range of evolutionary and K-corrections
and the value of $\Omega$, and thus the velocity dispersion $\upsilon$
from the relation of Faber \& Jackson (1976). We find that the observed
critical radii $\xi_0=4\pi (\upsilon /c)^{2} D_d D_{ds}/D_s$, are more
probable for $\Omega < 1 $ for a source redshift smaller than $z(D_s) <
3 $ (Guhatakurta, Tyson \& Majewski 1990). For more detailed
discussions, see Im \etal (1995).
\section{CONCLUSIONS}
We have discovered two examples of `Einstein-cross' gravitational lenses
in HST survey data, one in the MDS data and one in the archived
Groth-Westphal GTO survey. The lens configurations show that the stars
responsible for the light distribution of these elliptical galaxies are
a trace population following the gravitational potential. The Mass to Light
ratio increases radially outwards.
These represent the first discoveries of lenses using the high
resolution of HST - indeed, apart from the exceptional original Einstein
cross discovered by Huchra \etal (1985), they represent the first
discoveries of field-galaxy gravitational lenses via the systematic
study of optical images. They are a new class of gravitational lens
candidates in which the cosmologically distant lens is a relatively
bright elliptical galaxy with well understood properties. If a
significant sample could be found and observed spectroscopically for
redshifts, they will be very useful cosmological probes.
These objects would have been very difficult discoveries from the ground
except under conditions of excellent seeing. Indeed, CCD frames taken at
the KPNO 4m prime focus (Connolly 1995) do not show the cross objects
flanking the elliptical galaxy HST14176+5226. We have not as yet
observed these galaxies spectroscopically. The redshift of the lensed
components in HST12531\M2914 is probably a challenging observation for
the Keck telescope in excellent seeing.
From the observed numbers of bright elliptical galaxies observed in the
GTO survey strip (300 to I=22), the numbers of faint objects in the
fields (8000 to I=26), and the expected cross-sections, we estimate
that we should find one quadruple lens in every $20-30$ WFPC2 fields
surveyed. The number that has been discovered so far is therefore
consistent with our expectations. An on-going systematic and careful
inspection of the MDS fields, looking very specifically for possible
gravitational lens candidates in the shallower MDS and GTO parallel
fields is in progress in order to expand the sample. As further MDS data
are taken in Cycle 5 and subsequent cycles, they will be examined for
similar spectacular lenses, and also for more common lenses consisting
of arcs or two or three components, to obtain a statistically
representative sample of HST gravitational lens candidates.
This paper is based on observations with the NASA/ESA Hubble Space
Telescope, obtained at the Space Telescope Science Institute, which is
operated by the Association of Universities for Research in Astronomy,
Inc., under NASA contract NAS5-26555. The Medium-Deep Survey is funded
by STScI grant GO2684. We gratefully acknowledge Lyman Neuschaefer for
help with the MDS pipeline and many helpful discussions, the anonymous
referee from many useful suggestions, and {\it et tu} Broadhurst.
|
\section{Introduction}
\label{sec:int}
Construction of chiral gauge theories is one of the long-standing problems
of lattice field theories.
Due to the fermion doubling problems, a naively discretized lattice fermion
field yields $2^d$ fermion particles, half of one chirality and half of the
other, so that the theory becomes non-chiral\cite{nielnino}.
Several lattice approaches have been proposed, but so far none of them have
been proven to work successfully.
Kaplan has proposed a new construction of lattice chiral gauge
theories via domain-wall models\cite{kaplan}.
Starting from a vector-like gauge theory in $2k+1$-dimensions
with a fermion mass term being a shape of a domain-wall in the
(extra) $2k+1$-th dimension,
he showed in the weak gauge coupling limit that
a massless chiral state arises as a zero mode
bound to the $2k$-dimensional domain-wall
while all the doublers have large masses of the lattice cut-off scale.
It has been also shown that the model works well for smooth back-ground gauge
fields\cite{aokihirose,jansen}.
Two simplified variants of the original Kaplan's domain-wall model
have been proposed: an ``overlap formula''\cite{overlap} and a
``waveguide model''\cite{waveg}. Gauge fields appeared in these variants
are $2k$-dimensional and are independent of the extra $2k+1$-th coordinate,
while those in the original model are $2k+1$-dimensional and depend on
the extra $2k+1$-th coordinate. These variants work successfully for
smooth back-ground gauge fields\cite{constchi,aokilevin}, as the original one
does. Non-perturbative investigations for these variants
seems easier than for the original model due to the simpler structure
of gauge fields.
However it has been reported\cite{waveg} that
the waveguide model in the weak gauge coupling limit
can not produce chiral zero modes needed
to construct chiral gauge theories.
In this limit, if gauge invariance were maintained,
pure gauge field configurations equivalent to the unity
by gauge transformation would dominate and gauge fields would become
smooth.
In the set-up of the waveguide model, however, $2k$-dimensional gauge
fields are non-zero only in the layers near domain-wall( waveguide ),
so that the gauge invariance is broken in the edge of the waveguide.
Therefore, even in the weak gauge coupling limit,
gauge fields are no more smooth and becomes very ``rough'',
due to the gauge degrees of freedom appeared to be dynamical in this edge.
As a result of the rough gauge dynamics,
a new chiral zero mode with the opposite chirality to the original zero mode
on the domain-wall appears in the edge, so that
the fermionic spectrum inside the waveguide becomes vector-like.
It has been claimed\cite{waveg} that this ''rough gauge'' problem also
exists in the overlap formula since the gauge invariance is broken
by the boundary condition at the infinity of the extra
dimension\cite{aokilevin,shamir}. Furthermore an equivalence between the
wave-guide model and the overlap formula has been pointed out for
the special case\cite{gswgol}. Although the claimed equivalence
has been challenged in ref.\cite{overwave}, it is still crucial
for the success of the overlap formula to solve
the ''rough gauge'' problem and to show the existence of a chiral zero mode
in the weak gauge coupling limit.
How about original Kaplan's model ?
In this model there are two inverse gauge coupling $\beta=1/g^2$ and
$\beta_s=1/g_s^2$, where $g$ is the coupling constant in (physical)
$2k$-dimensions and $g_s$ is the one in the (extra) $2k+1$-th dimension.
Very little are known about this model except $\beta_s=0$
case\cite{aoitnioshi,kornip} where the spectrum seems vector-like.
In the weak coupling limit, corresponding to the $g\rightarrow 0$ limit in this
model, all gauge fields in the physical dimensions can be gauged away, while
the gauge field in the extra dimension is still dynamical and its dynamics is
controlled by $\beta_s$. Instead of the gauge degrees of freedom
in the edge of the wave-guide, $2k+1$-th component of gauge fields represent
roughness of $2k$ dimensional gauge fields.
An important question is whether the chiral zero mode on the domain-wall
survives in the presence of this rough dynamics.
The dynamics of the gauge field in this limit is equivalent to
$2k$ dimensional scalar model with $L_s$ independent copies where $L_s$ is the
number of sites in the extra dimension. In general at large $\beta_s$ such a
system is in a ``broken phase'' where some global symmetry is spontaneously
broken, while at small $\beta_s$ the system is in a ``symmetric'' phase.
Therefore there exists a critical point $\beta_s^c$, and it is likely
that the phase transition at $\beta_s=\beta_s^c$ is continuous(second or higher
order).
The ''gauge field'' becomes rougher and rougher at smaller $\beta_s$.
Indeed we know that the zero modes disappears at $\beta_s=0$\cite{aoitnioshi},
while the zero mode exists at $\beta_s=\infty$ ( free case ).
So far we do not know the fate of the chiral zero mode in the intermediate
range of the coupling $\beta_s$. There are the following 3 possibilities:
(a) The chiral zero mode always exists except $\beta_s=0$. In this case
we may likely construct a lattice chiral gauge theory
in both broken ($\beta_s > \beta_s^c$) and symmetric($\beta_s < \beta_s^c$),
and the continuum limits may be taken at $\beta_s = \beta_s^c$.
This is the beset case for the domain-wall model.
(b) The chiral zero mode exists only in the broken
phase($\beta_s > \beta_s^c$). In this case we may construct a lattice chiral
gauge theory only in the broken phase via the domain-wall method.
This is unsatisfactory, since the chiral gauge theory in the symmetric phase,
which is an important theoretical basis for various models,
can not be described via the domain-wall method.
(c) No chiral zero mode survives except $\beta_s=\infty$. The original
Kaplan's model can not describe lattice chiral gauge theories at all.
It is very important to determine which possibility is indeed realized
in the domain-wall model.
Therefore, in this paper ,
in order to know the fate of the chiral zero mode,
we have carried out a numerical simulation of a domain-wall model
in $(2+1)$-dimension
with a quenched $U(1)$ gauge field in the $\beta=\infty$ limit.
Strictly speaking, there is no order parameter in a 2 dimensional
$U(1)$ model( XY model ). On a large but finite lattice, however, a behavior
of the 2 dimensional model is similar to the one of a 4 dimensional
scalar model. Thus, we hopefully think that
useful informations about the fate of the zero mode
can be obtained from such a toy model in (2+1)-dimensions.
In Sec.2 ,
we have defined our domain-wall model
with dynamical gauge fields.
We have calculated a fermion propagator
by using a kind of mean-field approximation,
to show that there is a critical value of the domain-wall mass
parameter above which the zero mode exist.
The value of the critical mass may depend on $\beta_s$,
which controls the dynamics of the gauge field.
In Sec.3 ,
we have calculated the fermion spectrum numerically
using quenched approximation at $\beta_s = 0.5,1.0.5.0$
and at various values of domain-wall masses.
We have found that at any value of three $\beta_s$
there always exists the range of domain-wall mass parameter
in which the chiral zero mode survives on the domain-wall.
Our conclusion and some discussions are given in Sec. 4.
\section{domain-wall model}
\subsection{Definition of the model}
We consider a vector gauge theory in $d=(2k+1)$-dimension
with a domain-wall mass term, which has a shape of a step function
in the coordinate of an extra dimension.
This domain-wall model is originally proposed by Kaplan\cite{kaplan},
and a fermionic part of the action
is reformulated by Narayanan-Neuberger\cite{naraneu},
in terms of a $2k$-dimensional theory.
The model is defined by the action
\begin{equation}
S = S_{G} + S_{F} ,
\end{equation}
where $S_{G}$ is the action of a dynamical gauge field , $S_{F}$ is the
fermionic action.
$S_{G}$ is given by
\begin{eqnarray}
S_{G} &=& \beta \sum_{n,\mu > \nu} \sum_{s} {\rm Re Tr} \left[
U_{\mu \nu}(n,s) \right] \nonumber \\
&+& \beta_{s} \sum_{n,\mu} \sum_{s} {\rm Re Tr} \left[
U_{\mu d}(n,s) \right] ,
\end{eqnarray}
where $\mu , \nu$ run from $1$ to $2 k$ ,
$n$ is a $2 k$-dimensional lattice point ,
and $s$ is a coordinate of an extra dimension.
$U_{\mu \nu}(n,s)$ is a $2 k$-dimensional plaquette
and $U_{\mu d}(n,s)$ is a plaquette
containing two link variables in the extra direction.
$\beta$ is the inverse gauge coupling for the plaquette $U_{\mu \nu}$
and $\beta_{s}$ is the one for the plaquette $U_{\mu d}$ .
In general , $\beta \neq \beta_{s}$ .
The fermion action $S_{F}$ on the Euclidean lattice,
in terms of the $2 k$-dimensional notation,
is given by
\widetext
\begin{eqnarray}
S_{F}& = & \frac{1}{2} \sum_{n \mu} \sum_s
\bar{\psi}_s(n) \gamma_\mu \left[ U_{s,\mu}(n) \psi_s(n + \mu)
- U_{s,\mu}^{\dag}(n - \mu) \psi_s(n - \mu) \right] \nonumber \\
&+& \sum_n \sum_{s,t} \bar{\psi}_s(n) \left[ M_0 P_R
+ M_0^{\dag} P_L \right] \psi_t(n) \nonumber \\
&+& \frac{1}{2} \sum_{n \mu} \sum_s
\bar{\psi}_s(n) \left[ U_{s,\mu}(n) \psi_s(n + \mu)
+ U_{s,\mu}^{\dag}(n - \mu) \psi_s(n - \mu) -2 \psi_s(n) \right]
\label{eqn:fermion}
\end{eqnarray}
\narrowtext
where $s , t$ are an extra coordinates ,
$P_{R/L} = \frac{1}{2} (1 \pm \gamma_{2k+1})$ ,
\begin{eqnarray}
&& ( M_0 )_{s,t} = U_{s,d}(n) \delta_{s + 1 , t} - a(s) \delta_{s,t} \\
&& ( M_0^{\dag} )_{s,t} = U_{s - 1 , d}^{\dag}(n) \delta_{s - 1 , t} - a(s)
\delta_{s,t} .
\end{eqnarray}
Here $U_{s,\mu}(n) , U_{s,d}(n)$ ($d=2k+1$) are link variables
connecting a site $(n,s)$ to $(n+\mu,s)$ or $(n,s+1)$, respectively,
Because of a periodic boundary condition in the extra dimension ,
$s , t$ run from $-L_{s}$ to $L_{s} - 1$ ,
and $a(s)$ is given by
\begin{eqnarray}
a(s) & = & 1 - m_0 \, {\rm{sign}}
\left[( s + \frac{1}{2} ) \, {\rm{sign}}( L_{s} - s - \frac{1}{2} ) \right]
\nonumber \\
& = & \left\{
\begin{array}{cc}
1 - m_0 & ( - \frac{1}{2} < s < L_{S} - \frac{1}{2} ) \\
1 + m_0 & ( - L_{s} - \frac{1}{2} < s < - \frac{1}{2} ) \, ,
\end{array}
\right.
\end{eqnarray}
where $m_{0}$ is the height of the domain-wall mass.
It is easy to check
that the above fermionic action is identical to the one
in $(2k+1)$-dimensions, proposed by Kaplan\cite{kaplan,naraneu}.
In weak coupling limit of both $\beta$ and $\beta_{s}$ ,
this model becomes free theory
and can be easily analyzed.
In free theory at $0 < m_{0} < 1$,
it has been shown that a desired chiral zero mode
appears on a domain-wall( $s=0$ plane ) without unwanted doublers.
Due to the periodic boundary condition in the extra dimension,
however , a zero mode of the opposite chirality to the one on the domain-wall
appears on the anti-domain-wall , $s = L_{s} - 1$ .
Overlap between two zero modes decreases exponentially at large $L_s$.
A free fermion propagator is easily calculated and
an effective action of a $(2+1)$-dimensional model
including the gauge anomaly and the Chern-Simons term can be obtained for
smooth background gauge fields\cite{aokihirose}.
Domain-wall models, however, have not been investigated yet
{\it non-perturbatively}.
Main question is whether the chiral zero mode survives
in the presence of rough gauge fields mentioned in the introduction.
To answer this question we will analyze
the fate of the chiral zero mode
in the weak coupling limit for $\beta$.
In this limit,
the gauge field action $S_{G}$ is reduced to
\begin{equation}
S_{G} = \beta_{s} \sum_{s} \sum_{n} {\rm Re Tr} \left[ V(n,s) V^{\dag}(n+\mu,s)
\right] ,
\label{eqn:gauge}
\end{equation}
where the link variable $U_{s,d}(n)$ in the extra direction
is regarded as a site variable $V(n,s)( = U_{s,d}(n))$.
This action is identical to the one of a $(d-1)$-dimensional spin model
and $s$ is regarded as an independent flavor.
The action eq.(\ref{eqn:gauge}) is invariant under
\begin{equation}
V(n,s) \longrightarrow g(s) V(n,s) g^{\dag}(s+1)\quad, \qquad
(g(s) \in G) ,
\label{eqn:symmetry}
\end{equation}
where $G$ is the gauge group of the original model.
Therefore the total symmetry of the model
is $G^{2 L_{s}}$, where $2L_s$, the size of the extra
dimension, is regarded as the number of independent flavors.
We use this (reduced) model for our numerical investigation.
\subsection{Mean field approximation for fermion propagators}
When the dynamical gauge fields are added even on the extra dimension only,
it is difficult to calculate the fermion propagator analytically.
Instead of calculating the fermion propagator {\it exactly} ,
we use a mean-field approximation
to see an effect of the dynamical gauge field qualitatively.
The mean-field approximation we adopt is
that the link variables are replaced as
\begin{equation}
V(n,s) \longrightarrow z ,
\end{equation}
where $z$ is a $(n,s)$-independent constant.
{}From eq.(\ref{eqn:fermion})
the fermion action in a $(d-1)$-dimensional momentum space becomes
\widetext
\begin{equation}
S_{F} \rightarrow
\sum_{s,t,p} \bar{\psi}_{s}(-p)
\left(\sum_{\mu} i \gamma_{\mu} \sin (p_{\mu}) \delta_{s,t}
+ \left[ M(z) P_{R} + M^{\dag}(z) P_{L} \right]_{s,t} \right)
\psi_{t}(p),
\end{equation}
\begin{equation}
(M(z))_{s,t} = (M_{0}(z))_{s,t} + \frac{\nabla(p)}{2} \delta_{s,t} ,
\qquad
(M^{\dag}(z))_{s,t} = (M^{\dag}_{0}(z))_{s,t} + \frac{\nabla(p)}{2}
\delta_{s,t} ,
\end{equation}
where
$ \nabla(p) \equiv \sum_{\mu=1}^{d-1} 2 ( \cos p_{\mu} - 1 ).$
Following Ref.\cite{aokihirose} it is easy to obtain
a mean field fermion propagator
on a finite lattice with the periodic boundary condition:
\begin{eqnarray}
G(p)_{s,t}
&=&
\left[ i \sum_\mu \gamma_\mu \bar{p}_\mu + M P_R + M^{\dag}
P_L \right]_{s,t}^{-1}
\nonumber \\
&=&
\left[ \left\{ \left(
- i \sum_\mu \gamma_\mu \bar{p}_\mu + M \right)
G_L(p)_{s,t} \right\} P_L
\right.
+ \left. \left\{ \left(
- i \sum_\mu \gamma_\mu \bar{p}_\mu + M^{\dag} \right)
G_R(p)_{s,t} \right\} P_R \right] ,
\label{eqn:propagator}
\end{eqnarray}
\narrowtext
\begin{equation}
G_L(p) = \frac{\displaystyle 1}{\displaystyle {\bar{p}^2 + M^{\dag} M}}
\quad , \quad
G_R(p) = \frac{\displaystyle 1}{\displaystyle {\bar{p}^2 + M M^{\dag}}} \quad
,
\label{eqn:fomal}
\end{equation}
with $\bar{p}_{\mu} = \sin (p_{\mu})$.
For large $L_{s}$ where we neglect terms of $O(e^{-c L_{s}})$ with $c > 0$,
$G_{L}$ and $G_{R}$ are given by
\widetext
\begin{equation}
\left[ G_L(p) \right]_{s,t} = \left\{
\begin{array}{ll}B e^{-\alpha_{+} |s - t|}
+ \left( A_L - B \right) e^{- \alpha_{+}(s + t)}
+ \left( A_R - B \right) e^{- \alpha_{+}(2L_{s} - s - t)} ,
& (s , t \geq 0) \\
A_L e^{- \alpha_{+} s + \alpha_{-} t}
+ A_R e^{- \alpha_{+}(L_{s} - s) - \alpha_{-}(L_{s} + t)} ,
& (s \geq 0 , t \leq 0) \\
A_L e^{ \alpha_{-} s - \alpha_{+} t}
+ A_R e^{- \alpha_{-}(L_{s} + s ) - \alpha_{+}(L_{s} - t)} ,
& (s \leq 0 , t \geq 0) \\
C e^{-\alpha_{-} |s - t|}
+ \left( A_L - C \right) e^{ \alpha_{-}(s + t)}
+ \left( A_R - C \right) e^{- \alpha_{-}(2L_{s} + s + t)},
& (s , t \leq 0)
\end{array}
\right.
\label{eqn:GL}
\end{equation}
\begin{equation}
\left[ G_R(p) \right]_{s,t} = \left\{
\begin{array}{ll}
B e^{-\alpha_{+} |s - t|}
+ \left( A_R - B \right) e^{- \alpha_{+}(s + t + 2)}
+ \left( A_L - B \right) e^{- \alpha_{+}(2L_{s} - s - t -2)} ,
& (s , t \geq -1) \\
A_R e^{- \alpha_{+}(s + 1) + \alpha_{-} (t + 1)}
+ A_L e^{- \alpha_{+}(L_{s} - s -1) - \alpha_{-}(L_{s} + t +1)} ,
& (s \geq -1 , t \leq -1) \\
A_R e^{ \alpha_{-} (s + 1) - \alpha_{+} (t + 1)}
+ A_L e^{- \alpha_{-}(L_{s} + s + 1) - \alpha_{+}(L_{s} - t -1)} ,
& (s \leq -1 , t \geq -1) \\
C e^{-\alpha_{-} |s - t|}
+ \left( A_R - C \right) e^{ \alpha_{-}(s + t +2)}
+ \left( A_L - C \right) e^{- \alpha_{-}(2L_{s} + s + t +2)} ,
& (s , t \leq -1)
\end{array}
\right.
\label{eqn:GR}
\end{equation}
\narrowtext
where
\begin{eqnarray}
&& a_{\pm} = z ( 1 - \frac{\nabla(p)}{2} \mp m_{0} ) = z b_{\pm} , \\
&& \alpha_{\pm} = {\rm{arccosh}}
\left[ \frac{\bar{p}^{2} + z^{2} + b_{\pm}^2}{2 z b_{\pm}} \right] , \\
&& A_L = \frac{1}{a_{+} e^{\alpha_{+}} - a_{-} e^{- \alpha_{-}}}, \,
A_R = \frac{1}{a_{-} e^{\alpha_{-}} - a_{+} e^{- \alpha_{+}}}, \\
&& B = \frac{1}{2 a_{+} \sinh \alpha_{+}} \quad , \quad
C = \frac{1}{2 a_{-} \sinh \alpha_{-}} .
\end{eqnarray}
Behaviors of $A_{R} , B$ and $C$ as $p\rightarrow 0$
are similar to the ones in free theory:
They have no singularity for all $z$
A behavior of $A_{L}$ is, however, different:
As $p \rightarrow 0$ $A_{L}$ behaves as
\begin{eqnarray}
A_{L} & \rightarrow &
\frac{\displaystyle 1}{\displaystyle [(1 - m_{0})^{2}] + O(p^{2})}, \quad
(0 < m_{0} < 1 - z), \\
& \rightarrow &
\frac{\displaystyle 4m_{0}^{2}-[(z^{2}-1)-m_{0}^{2}]^{2}}
{\displaystyle 4m_{0}z^{2}p^{2}}, \, \, (1 - z < m_{0} < 1) .
\end{eqnarray}
A critical value of the domain-wall mass
that separates a region with a zero mode
and a region without zero modes
is $m_{0}^c = 1 - z$.
Since $A_{L}$ term dominates for $1 - z < m_{0} < 1$
in the $G_{L}$ (eq.(\ref{eqn:GL}) ) and $G_{R}$ ( eq.(\ref{eqn:GR})),
a right-handed zero mode appears in the $s=0$ plane ,
and a left-handed zero mode in the $s=L_{s}-1$ plane.
For $0 < m_{0} < 1 - z$
the right- and left-handed fermions are massive in all $s$ planes.
Since the terms of $A_{L} , A_{R} , B$ and $C$ are almost same value
in this region of $m_{0}$,
a translational invariant term dominates
in $G_{L}$ and $G_{R}$, so that
the spectrum becomes vector-like.
If $z \rightarrow 1$ , the model becomes free theory.
The propagator obtained in this section agrees with the one
obtained in Ref.\cite{aokihirose}.
In the opposite limit that $z \rightarrow 0$ ,
since there is no hopping term to the neighboring layers,
this model becomes the one analyzed in Ref.\cite{aoitnioshi}
in the case of the strong coupling limit $\beta_{s} = 0$ ,
and in Ref.\cite{kornip} ,
in the case that $z$ is identified to the vacuum expectation value of the link
variables.
This consideration suggests that
the region where the zero modes exist
become smaller and smaller as $z$ $(1-z < m_{0} < 1)$ approaches zero.
What corresponds to $z$ ?
Boundary conditions $z$ satisfies are
$z=1$ at $\beta_s=\infty$ and $z=0$ at $\beta_s=0$.
The most naive candidate\cite{kornip} is
\begin{equation}
z= \langle V(n,s) \rangle .
\label{order}
\end{equation}
But this is not invariant under the symmetry (\ref{eqn:symmetry}).
The other choice invariant under (\ref{eqn:symmetry}) is
\begin{equation}
z^2 = \langle {\rm Tr Re} \{V(n,s) V^\dagger(n+\mu,s)\} \rangle .
\label{energy}
\end{equation}
If eq. (\ref{order}) is true, zero modes disappears in the symmetric phase,
where $\langle V(n,s) \rangle = 0$, while, for the case of eq. (\ref{energy}),
the zero modes always exist in both phases, since
$\langle {\rm Tr Re}\{ V(n,s) V^\dagger(n+\mu,s)\} \rangle$
is insensitive to which phase
we are in.
\section{numerical study of (2+1)-dimensional U(1) model}
\subsection{Method of numerical calculations}
In this section
we numerically study the domain-wall model in $(2+1)$-dimension
with a $U(1)$ dynamical gauge field in the extra dimension.
As seen from eq.(\ref{eqn:gauge}) ,
the gauge field action can be identified with a $2$-dimensional $U(1)$ spin
model (with $2 L_{s}$ copies).
In $(2+1)$-dimension,
$\gamma$-matrices are Pauli-matrices ,
$\sigma_{1} \, , \, \sigma_{2} \, , \, \sigma_{3}$.
Our numerical simulation has been carried out by the quenched approximation.
Configurations of $U(1)$ dynamical gauge field are generated
and fermion propagators are calculated on the configurations.
The obtained fermion propagators are gauge non-invariant in general
under the symmetry (\ref{eqn:symmetry}).
The fermion propagator $G(p)_{s,t}$ becomes ``invariant''
if and only if $s=t$.
Thus, we take the $s-s$ layer as propagating plane($\approx$
``physical space''),
and investigate the behavior of the fermion propagator in this layer.
To study the fermion spectrum,
we assume a form of eq.(\ref{eqn:propagator}) for
the fermion propagator, from which we extract $G_{L}$ and $G_{R}$.
We then obtain corresponding fermion masses
from $G_{L}^{-1}(p)$ and $G_{R}^{-1}(p)$
by fitting them linearly in $\bar{p}^{2}$, since,
from eq.(\ref{eqn:fomal}) :
\begin{eqnarray}
G_{L}^{-1} = \bar{p}^{2} + M^{\dag} M \rightarrow m_{f}^{2}
\quad , \quad (p \rightarrow 0) ,
\label{eqn:Lanalysis} \\
G_{R}^{-1} = \bar{p}^{2} + M M^{\dag} \rightarrow m_{f}^{2}
\quad , \quad (p \rightarrow 0) .
\label{eqn:Ranalysis}
\end{eqnarray}
We take the following setup for 2-dimensional momenta.
A periodic boundary condition is taken for the 1st direction
and the momentum in this direction is fixed on $p_{1} = 0$.
An anti-periodic boundary condition is taken for
the 2nd-direction
and the momentum in this direction is variable
such as $p_{2} = (2n+1) \pi / L \, , \, n = - L / 2 ,..., L / 2 - 1$.)
If $m_{f}^{2} = 0$ ,
we conclude that there is a zero mode, and if $m_{f}^{2} \neq 0$ ,
there is not.
\subsection{Simulation parameters}
Our simulation is performed in the quenched approximation on
$L^{2}\times 2 L_{s}$ lattices with $L = 16 , 24 , 32$ and $L_{s} = 16$.
The coordinate $s$ in the extra dimension runs $-16 < s < 15$.
Gauge configurations are generated by the 5-hit Metropolis algorithm
at $\beta_{s} =$ 0.5, 1.0, 5.0. For the thermalization first 1000 sweeps
are discarded.
The fermion propagators are calculated by the conjugate gradient method
on 50 configurations separated by at least 20 sweeps, except at $\beta_s=$5.0
on a $32^2\times 32$ lattice where the number of configurations are 11.
We take the domain-wall mass
$m_{0} =$ 0.7, 0.8, 0.9, 0.99 at $\beta_{s} =$ 0.5,
$m_{0} =$ 0.3, 0.4, 0.5, 0.6, 0.9 at $\beta_{s} =$ 1.0,
and $m_{0} =$ 0.1, 0.2, 0.3 at $\beta_{s} =$ 5.0 .
The boundary conditions in $1$st- and $3$rd(extra)-directions
are periodic and the one in $2$nd-direction is anti-periodic.
Wilson parameter $r$ has been set to $r = 1$.
The fermion propagators have been investigated at $s =$ 0, 8, 15.
These $s$ are the layers where we put sources.
The layer at $s=0$ is the domain-wall ,
at $s=15$ , the anti-domain-wall ,
and at $s=8$ , neither.
Errors are all estimated with the jack-knife method.
\subsection{Quenched phase structure}
As explained before
the gauge field action of our model is identical to
that of the $U(1)$ spin system in $2$-dimensions.
Therefore, there is a Kosterlitz-Thouless phase transition
and this system does not have an order parameter on the infinite lattice.
On the finite volume, however,
we take a vacuum expectation value of link variables as an order parameter
using rotation technique:
\begin{equation}
v = < |\frac{1}{L^{2}} \sum_{n} V(n,s)| >_{s} ,
\label{eqn:order}
\end{equation}
where $L$ is the lattice size of the $1,2$-dimension.
The defined vacuum expectation value $v$ above is zero
in the Kosterlitz-Thouless phase
but $v > 0$ in the spin-wave phase on the finite lattice.
(Increasing the lattice size, however , decreasing the value of $v$.
In the infinite lattice size, the value of $v$ is zero for all gauge coupling.)
Since we are interested in the dynamics of $4$-dimensional theories,
where the phase transition separates a symmetric phase from a broken phase ,
we have used
this $2$-dimensional system on large but finite volume
as a toy model of $4$-dimensional real world.
Therefore, in this letter,
we refer to the Kosterlitz-Thouless phase as the symmetric phase,
and to the spin-wave phase as the broken phase.
Fig. \ref{vev}(a) shows that,
on a $16^2\times 32$ lattice, $v$ behaves as if it was an order parameter.
{}From Fig. \ref{vev}(b)
we consider that the system is in the symmetric phase at $\beta_{s} = 0.5$,
while in the broken phase at $\beta_{s} =$ 1.0, 5.0 .
\subsection{Fermion spectrum in the broken phase}
At $\beta_{s} =$ 1.0 and 5.0,
the system is in broken phase.
Here we mainly discuss the result at $\beta_{s} = 1.0$ in detail.
We first consider the fermion spectrum on the layer at $s=0$.
Fig. \ref{prop} is a plot of the term corresponding to
$- \sin (p_{2})\cdot G_{L}$ and $- \sin (p_{2})\cdot G_{R}$
as a function of $p_{2}$ at $m_0$ =0.3 and 0.5.
( Note we always set $p_1=0$.)
This figure shows that, as $p_2$ goes to zero,
$G_{L}$ seems to diverge at $m_{0}$ =0.5 but stay finite at $m_{0}$ = 0.3,
while $G_{R}$ stays finite at both $m_0$.
Next let us show Fig. \ref{invprop}, which
is a plot of the $G_{L}^{-1}$ and $G_{R}^{-1}$ as a function of
$\bar{p}_2^{2} \equiv \sin^{2} (p_{2})$ at $m_0$ =0.3 and 0.5.
In the limit $p_{2} \rightarrow 0$,
$G_{R}^{-1}$ remains non-zero at both $m_{0}$,
while $G_{L}^{-1}$ vanishes at $m_{0} = 0.5$.
We obtain the value of $m_f^2$, which can be regarded as the mass
square in $2$-dimensional world,
by the linear fit in $\bar{p}_2^{2}$,
and plot $m_f$ as a function of $m_{0}$ in Fig. \ref{mfs0}.
The mass of right-handed fermion, obtained from $G_{L}^{-1}$,
becomes very small ( less than 0.1) at $m_{0}$ larger than $0.5$, so
we conclude that this critical value is $m_{0}^{c} \sim 0.5$.
Whenever the domain-wall mass is larger than this value ,
this model produces the right-handed chiral zero mode on the domain-wall
at $s=0$.
On the anti-domain-wall $(s=15)$,
on the other hand,
the mass of left-handed fermion becomes less than 0.1
at $m_{0}$ larger than the critical mass $m_{0}^{c} \sim 0.5$,
as seen in Fig. \ref{mfs15}.
It is noted
that chiralities between the zero modes on the domain-wall and the
anti-domain-wall
are opposite each other.
Finally Fig. \ref{mfs8} shows that, on $s=8$ ,
the layer in the middle between the domain-wall and the anti-domain-wall ,
both right-handed and left-handed fermions
stay heavy.
A similar result at $\beta_{s}=5.0$ on $s=0$ is given Fig. \ref{mfsb5}.
{}From these results above,
we conclude
that the domain-wall model with the dynamical gauge field
on the extra dimension ({\it {i.e.}}
the weak coupling limit of the original Kaplan's model)
can create the chiral zero mode on the domain-wall,
at least in the broken phase.
This suggests that the original Kaplan's model
has a great chance to work for the construction of
lattice chiral gauge theories
in the broken phase.
\subsection{Fermion spectrum in symmetric phase}
The system is in the symmetric phase at $\beta_{s} = 0.5$.
The fermion propagator is analyzed in the same way as in the broken phase.
However, for example on the $s=0$ layer,
$- \sin (p_{2})\cdot G_{L}$ and $- \sin (p_{2})\cdot G_{R}$
show similar behaviors on a $16^2\times 32$ lattice,
as seen in Fig. \ref{symprop}.
Smaller lattice sizes, stronger the similarity,
which makes analysis more difficult in the symmetric phase.
To see a difference between the right-handed and left-handed fermions,
we have to take larger lattice size such as $L =$ 24, 32.
In Fig. \ref{symmfs0}, we have plotted mass $m_f$ of both modes
at $s=0$ as a function of $m_0$.
Although a difference of masses
between the right-handed and the left-handed fermions
is very small, about $0.1$ or less at $m_{0}=0.99$,
this difference stays finite as we increase the spatial lattice size $L$
from 24 to 32.
Therefore we conclude
that the right-handed fermion becomes massless at $m_{0}$ larger than
$0.9$,
while the left-handed fermion stays massive at all $m_{0}$,
so that the fermion spectrum on the domain-wall is {\it chiral}.
In order to see that the difference of mass between the right and the left
is really a signal, not a statistical fluctuation,
we have plotted $m_f$ vs. $m_0$
in the case of putting a source at the anti-domain-wall $s=15$
in Fig. \ref{symmfs15}.
We observe, at $m_{0}=$0.99,
a massless fermion of the opposite chirality to the $s=0$ zero mode
and a finite difference of masses
between the right and the left, which
stays finite as we increase the spatial lattice size.
Furthermore, in the case of $s=8$,
the right-handed fermion and the left-handed fermion stay massive
at all $m_{0}$, as seen in Fig. \ref{symmfs8}
{}From these results above,
as the same case in the broken phase,
we conclude
that the original Kaplan's model
can create the chiral zero mode on the domain-wall
even in the symmetric phase.
\section{Conclusions and discussions}
Using the quenched approximation,
we have performed the numerical study of the domain-wall model
in (2+1)-dimensions
with the $U(1)$ dynamical gauge field on the extra dimension.
{}From this study we obtain the following results.
There exists the critical value of the domain-wall mass
separating the region with a chiral zero mode
and the region without it, both in the broken and the symmetric phases of
the gauge field.
At the domain-wall mass larger than its critical value
a zero mode with one chirality
exists on the domain-wall
and a zero modes with opposite chirality
on the anti-domain-wall,
and none in the middle between the domain-wall and the anti-domain-wall.
These results strongly suggest
that
it is possible to construct lattice chiral gauge theories
at all $\beta_{s}$ except for $\beta_{s} = 0$
via the domain-wall method, and continuum limits may possibly be taken
at the critical value of $\beta_s$ where the phase transition takes place.
In $(2+1)$-dimensions, however,
the gauge field in $\beta= \infty$ limit is special
since there is no order parameter and the phase transition is topological.
Thus, to make a definite conclusion on the construction of lattice chiral
gauge theories via the domain-wall method,
we must study realistic $(4+1)$-dimensional model
with $U(1)$ or $SU(N)$ gauge field in $\beta= \infty$ limit.
Such models in $\beta= \infty$ limit have a phase transition
characterized by an order parameter,
a vacuum expectation value of the link variables in the extra dimension.
Moreover, it is interesting and important to find
an appropriate correspondence
between the propagator obtained in the numerical simulation and
the mean field propagator with tuned parameter $z$.
So far it is not clear
what physical quantity is corresponding to $z$.
Since zero mode seems to exist even in the symmetric phase,
the correspondence (\ref{order}) is qualitatively incorrect.
On the other hand, we have found that the correspondence (\ref{energy})
fails to reproduce $m_0^c$ quantitatively.
Since mean-field approximations can not work well in
the lower-dimensions in general,
we must try to answer these questions
studying $(4+1)$-dimensional $U(1)$ or $SU(N)$ models.
\section*{Acknowledgements}
Numerical calculations for the present work have been carried out
at Center for Computational Physics, University of Tsukuba. This work
is supported in part by the Grants-in-Aid of the Ministry of
Education(Nos. 04NP0701, 06740199).
|
\section{\@startsection {section}{1}{\z@}{-4.2ex plus -1ex minus
-.2ex}{2.2ex plus .2ex}{\normalsize\bf}}
\def\subsection{\@startsection{subsection}{2}{\z@}{-2.2ex plus -1ex minus
-.2ex}{1.1ex plus .2ex}{\normalsize\bf}}
\def\subsubsection{\@startsection{subsubsection}{3}{\z@}{-2.2ex plus
-1ex minus -.2ex}{-1.2em}{\normalsize\it}}
\def\@arabic\c@section.{\@arabic\c@section.}
\def\thesection\@arabic\c@subsection.{\@arabic\c@section.\@arabic\c@subsection.}
\def\thesubsection\@arabic\c@subsubsection.{\thesection\@arabic\c@subsection.\@arabic\c@subsubsection.}
\def\@sect#1#2#3#4#5#6[#7]#8{\ifnum #2>\c@secnumdepth
\def\@svsec{}\else
\refstepcounter{#1}\edef\@svsec{\csname the#1\endcsname\hskip 1em }\fi
\@tempskipa #5\relax
\ifdim \@tempskipa>\z@
\begingroup #6\relax
\@hangfrom{\hskip #3\relax\@svsec}{\interlinepenalty \@M \sec@upcase{#8}\par}%
\endgroup
\csname #1mark\endcsname{#7}\addcontentsline
{toc}{#1}{\ifnum #2>\c@secnumdepth \else
\protect\numberline{\csname the#1\endcsname}\fi
#7}\else
\def\@svsechd{#6\hskip #3\@svsec #8\csname #1mark\endcsname
{#7}\addcontentsline
{toc}{#1}{\ifnum #2>\c@secnumdepth \else
\protect\numberline{\csname the#1\endcsname}\fi
#7}}\fi
\@xsect{#5}}
\def\@ssect#1#2#3#4#5{\@tempskipa #3\relax
\ifdim \@tempskipa>\z@
\begingroup #4\@hangfrom{\hskip #1}{\interlinepenalty \@M \sec@upcase{#5}\par}%
\endgroup
\else \def\@svsechd{#4\hskip #1\relax #5}\fi
\@xsect{#3}}
\def\@startsection{paragraph}{4}{1em{\@startsection{paragraph}{4}{1em}
{1ex plus .5ex minus .5ex}{-1em}{\bf}{\sec@upcase{Acknowledgments.}}}
\let\acknowledgements=\@startsection{paragraph}{4}{1em
\def\qanda@heading{Discussion}
\newif\if@firstquestion \@firstquestiontrue
\newenvironment{question}[1]{\if@firstquestion
\section*{\qanda@heading}\global\@firstquestionfalse\fi
\par\vskip 1ex
\noindent{\it#1\/}:}{\par}
\newenvironment{answer}[1]{\par\vskip 1ex
\noindent{\it#1\/}:}{\par}
\def\mathwithsecnums{
\@newctr{equation}[section]
\def\hbox{\normalsize\arabic{section}-\arabic{equation}}}{\hbox{\normalsize\arabic{section}-\arabic{equation}}}}
\def\section*{References{\section*{References}
\bgroup\parindent=0pt\parskip=.5ex
\def\relax{\par\hangindent=3em\hangafter=1}}
\def\refpar\egroup{\relax\egroup}
\def\section*{References{\section*{References}
\list{\null}{\leftmargin 3em\labelwidth 0pt\labelsep 0pt\itemindent -3em
\usecounter{enumi}}
\def\relax{\relax}
\def\hskip .11em plus .33em minus .07em{\hskip .11em plus .33em minus .07em}
\sloppy\clubpenalty4000\widowpenalty4000
\sfcode`\.=1000\relax}
\def\endlist{\endlist}
\def\@biblabel#1{\relax}
\def\@cite#1#2{#1\if@tempswa , #2\fi}
\def\relax\refpar{\relax\relax}
\def\@citex[#1]#2{\if@filesw\immediate\write\@auxout{\string\citation{#2}}\fi
\def\@citea{}\@cite{\@for\@citeb:=#2\do
{\@citea\def\@citea{,\penalty\@m\ }\@ifundefined
{b@\@citeb}{\@warning
{Citation `\@citeb' on page \thepage \space undefined}}%
{\csname b@\@citeb\endcsname}}}{#1}}
\let\jnl@style=\rm
\def\ref@jnl#1{{\jnl@style#1\/}}
\def\ref@jnl{AJ}{\ref@jnl{AJ}}
\def\ref@jnl{ARA\&A}{\ref@jnl{ARA\&A}}
\def\ref@jnl{ApJ}{\ref@jnl{ApJ}}
\def\ref@jnl{ApJ}{\ref@jnl{ApJ}}
\def\ref@jnl{ApJS}{\ref@jnl{ApJS}}
\def\ref@jnl{Appl.Optics}{\ref@jnl{Appl.Optics}}
\def\ref@jnl{Ap\&SS}{\ref@jnl{Ap\&SS}}
\def\ref@jnl{A\&A}{\ref@jnl{A\&A}}
\def\ref@jnl{A\&A~Rev.}{\ref@jnl{A\&A~Rev.}}
\def\ref@jnl{A\&AS}{\ref@jnl{A\&AS}}
\def\ref@jnl{AZh}{\ref@jnl{AZh}}
\def\ref@jnl{BAAS}{\ref@jnl{BAAS}}
\def\ref@jnl{JRASC}{\ref@jnl{JRASC}}
\def\ref@jnl{MmRAS}{\ref@jnl{MmRAS}}
\def\ref@jnl{MNRAS}{\ref@jnl{MNRAS}}
\def\ref@jnl{Phys.Rev.A}{\ref@jnl{Phys.Rev.A}}
\def\ref@jnl{Phys.Rev.B}{\ref@jnl{Phys.Rev.B}}
\def\ref@jnl{Phys.Rev.C}{\ref@jnl{Phys.Rev.C}}
\def\ref@jnl{Phys.Rev.D}{\ref@jnl{Phys.Rev.D}}
\def\ref@jnl{Phys.Rev.Lett}{\ref@jnl{Phys.Rev.Lett}}
\def\ref@jnl{PASP}{\ref@jnl{PASP}}
\def\ref@jnl{PASJ}{\ref@jnl{PASJ}}
\def\ref@jnl{QJRAS}{\ref@jnl{QJRAS}}
\def\ref@jnl{S\&T}{\ref@jnl{S\&T}}
\def\ref@jnl{Solar~Phys.}{\ref@jnl{Solar~Phys.}}
\def\ref@jnl{Soviet~Ast.}{\ref@jnl{Soviet~Ast.}}
\def\ref@jnl{Space~Sci.Rev.}{\ref@jnl{Space~Sci.Rev.}}
\def\ref@jnl{ZAp}{\ref@jnl{ZAp}}
\let\astap=\ref@jnl{A\&A}
\let\apjlett=\ref@jnl{ApJ}
\let\apjsupp=\ref@jnl{ApJS}
\def\hbox{$^\circ$}{\hbox{$^\circ$}}
\def\hbox{$\odot$}{\hbox{$\odot$}}
\def\hbox{$\oplus$}{\hbox{$\oplus$}}
\def\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}
\def\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}
\def\hbox{\rlap{$\sqcap$}$\sqcup$}{\hbox{\rlap{$\sqcap$}$\sqcup$}}
\def\hbox{$^\prime$}{\hbox{$^\prime$}}
\def\hbox{$^{\prime\prime}$}{\hbox{$^{\prime\prime}$}}
\def\hbox{$.\!\!^{\rm d}$}{\hbox{$.\!\!^{\rm d}$}}
\def\hbox{$.\!\!^{\rm h}$}{\hbox{$.\!\!^{\rm h}$}}
\def\hbox{$.\!\!^{\rm m}$}{\hbox{$.\!\!^{\rm m}$}}
\def\hbox{$.\!\!^{\rm s}$}{\hbox{$.\!\!^{\rm s}$}}
\def\hbox{$.\!\!^\circ$}{\hbox{$.\!\!^\circ$}}
\def\hbox{$.\mkern-4mu^\prime$}{\hbox{$.\mkern-4mu^\prime$}}
\def\hbox{$.\!\!^{\prime\prime}$}{\hbox{$.\!\!^{\prime\prime}$}}
\def\hbox{$.\!\!^{\scriptscriptstyle\rm p}$}{\hbox{$.\!\!^{\scriptscriptstyle\rm p}$}}
\def\hbox{$\mu$m}{\hbox{$\mu$m}}
\def\hbox{$\,^1\!/_2$}{\hbox{$\,^1\!/_2$}}
\def\hbox{$\,^1\!/_3$}{\hbox{$\,^1\!/_3$}}
\def\hbox{$\,^2\!/_3$}{\hbox{$\,^2\!/_3$}}
\def\hbox{$\,^1\!/_4$}{\hbox{$\,^1\!/_4$}}
\def\hbox{$\,^3\!/_4$}{\hbox{$\,^3\!/_4$}}
\def\hbox{$U\!BV\!R$}{\hbox{$U\!BV\!R$}}
\def\hbox{$U\!-\!B$}{\hbox{$U\!-\!B$}}
\def\hbox{$B\!-\!V$}{\hbox{$B\!-\!V$}}
\def\hbox{$V\!-\!R$}{\hbox{$V\!-\!R$}}
\def\hbox{$U\!-\!R$}{\hbox{$U\!-\!R$}}
\newcount\lecurrentfam
\def\LaTeX{\lecurrentfam=\the\fam \leavevmode L\raise.42ex
\hbox{$\fam\lecurrentfam\scriptstyle\kern-.3em A$}\kern-.15em\TeX}
\def\plotone#1{\centering \leavevmode
\epsfxsize=\textwidth \epsfbox{#1}}
\def\plottwo#1#2{\centering \leavevmode
\epsfxsize=.45\textwidth \epsfbox{#1} \hfil
\epsfxsize=.45\textwidth \epsfbox{#2}}
\def\plotfiddle#1#2#3#4#5#6#7{\centering \leavevmode
\vbox to#2{\rule{0pt}{#2}}
\special{psfile=#1 voffset=#7 hoffset=#6 vscale=#5 hscale=#4 angle=#3}}
\newif\if@finalstyle \@finalstylefalse
\if@finalstyle
\ps@myheadings
\let\ps@title=\ps@paspcstitle
\else
\ps@plain
\let\ps@title=\ps@plain
\fi
\ds@twoside
\makeatother
\setcounter{page}{1}
\begin{document}
\thispagestyle
\markright{\tiny\noindent To appear in {\bf Barred Galaxies}, IAU
Coll.~157, R.~Buta, B.G.~Elmegreen \& D.A.~Crocker
(eds.), ASP Series, (1996)}
\title{Stellar Dynamics and the 3D Structure of Bars}
\author{D. Pfenniger}
\affil{Geneva Observatory, CH-1290 Sauverny, Switzerland}
\begin{abstract}
Recent observational constraints restrict the strict applicability of stellar
dynamics in spirals to a few rotation periods. However, stellar dynamics
concepts such as periodic orbits are invaluable for understanding the various
dynamical processes occurring during much more periods. A distinction of two
instability types in stellar systems is pointed out, the first one being well
illustrated by the bar instability, and the second one by the bar bending
instability. In bars the third dimension brings essential dynamical effects
which modify the views about the history of bulges and the spiral secular
evolution. Bars may grow, bend, thicken, and dissolve into spheroidal bulges,
and spirals may evolve along the Hubble sequence in the sense Sd$\to$Sa. This
leads to a much more dynamical picture of isolated galaxies than imagined
before.
\end{abstract}
\keywords{barred galaxies, orbits}
\section{Applicability of Stellar Dynamics in Spirals}
Recent developments about spirals bring a quite different picture about their
physical state. The discovery of their large far-infrared (FIR) flux by IRAS
and COBE, comparable or sometimes superior to the optical one, and consistent
with the evidences that their optical parts are semi-transparent, means that a
substantial fraction of the stellar light is thermalized, and re-emitted by
dust in the FIR. This is {\it not negligible\/} with respect to the the
typical
power that large-scale dynamical processes (spiral arms, bars) can exchange,
the gravitational power (ratio of the gravitational energy to the dynamical
time). For a system in near virial equilibrium, $v^2 = GM/R$, we have $L_{\rm
grav}\approx\left(GM^2/R\right)\big/\left(R/v\right)=v^5/G$. Replacing $v$ by
the typical rotation speeds of spirals we find powers surprisingly close to
the the galaxy luminosities (e.g.~$10^{44}\,\rm erg\, s^{-1}$ for the Milky
Way). Since the IR Tully-Fisher relation ($L_{\rm H} \propto v^{4-5}$) is
nearly parallel to this equation, the match is close along the whole spiral
sequence (for more references on the subject see Pfenniger 1992).
This coincidence supports the proposition that a feedback mechanism relates
both dynamical instabilities and stellar activity. Essential is that although
a stellar population pours out energy over several Gyr, the first Myr of a
starburst is a quicker cooperative and intense reaction with respect to the
dynamical time $\tau_{\rm dyn}$. The idea of a feedback mechanism between
dynamics and star formation has been proposed several times (e.g.~Quirk 1972;
Kennicutt 1989). With a light thermalization power of the order of the
gravitational power, we just need to convert at an intermediate stage (between
light production and thermalization) a sizable fraction of this power into
{\it mechanical energy}. It is quite obvious in images of HII regions and
starbursts that a substantial mass of surrounding gas is set into coherent
motion by the massive star heating; this yields ultimately HI holes and
superbubbles. If a fraction $\gamma$ of the stellar power is converted into
mechanical power, the time-scale to change significantly the global binding
energy of the whole galaxy is $\tau_{\rm dyn}/\gamma$. If
$0.02\!<\!\gamma\!<\!1$ then the evolution time-scale associated with stellar
activity is shorter than the galaxy age; the galaxy as a whole must depend on
the stellar energy output. $\gamma$ is not well known but is estimated to be
in the range $\gamma\approx 0.1$ in starbursts (Leitherer \& Heckman 1995).
Such simple facts change fundamentally the way to see spirals. In the early
days of galactic dynamics, galaxies were viewed as essentially transparent
collections of stars, therefore the stellar energy output could be discounted,
exactly as the huge supernova neutrino flux. Also the large scale dynamical
instabilities such as bars were not understood, only the slow 2-body
relaxation was considered as a factor of evolution, then the dimension and
shape of galaxies had to be viewed as determined by the initial conditions of
formation. The subsequent evolution was viewed as so slow that stellar
dynamics concepts could be used for several Gyr. This led to absorb all the
dynamical effects in a rigid potential in many models, such as the stellar
population synthesis ones.
The conjunction of new elements on the energetics of spirals forces us to
modify the way to use stellar dynamics. Which of the concepts such as
relaxation, integral of motion, etc., are still relevant and what are their
new limits of applicability? If we take the Milky Way as a template of the
spirals, from observational data it is clear that the virial {\it gross
equilibrium}, at least in the optical disk, must be consistent with a
dominance of the bulk kinetic energy $E_{\rm kin}$ balancing gravitational
energy $E_{\rm grav}$:
\begin{equation}
{\textstyle\frac{1}{2}} \ddot I =
\underbrace{2 E_{\rm kin}}_{\sim4000\,{\rm eV\,cm^{-3}} \times V} -
\underbrace{|E_{\rm grav}|}_{> \sim 1000\,{\rm eV\,cm^{-3}} \times V} +
\underbrace{3 P_{\rm int}}_{\sim10\,\rm eV\,cm^{-3}} \!\!\!\!\!\!\!V \ -
\underbrace{3 P_{\rm ext}}_{<1\,\rm eV\,cm^{-3}} \!\!\!\!\!\!V,
\end{equation}
where $I$ is the moment of inertia inside the considered volume $V$. The
inner pressure $P_{\rm int}$ due to all the ISM components (gas, cosmic rays,
etc.) is a much too small energy reservoir to play any important role in the
virial equilibrium. The outer pressure $P_{\rm ext}$ due to intergalactic gas
and radiation is even more negligible. The only known large enough
{\it negative\/}
contribution to the virial balance susceptible to compensate the large
positive contribution of $E_{\rm kin}$ is gravitational, although the detected
mass is still insufficient. Thus dark matter must
be invoked, particularly in the outer disks of spirals.
Now the equilibrium is certainly imperfect as we have seen above with the
large visible and IR fluxes ($\sim 10^{44}\,\rm erg\,s^{-1}$) through the
interstellar medium. The quasi-static evolution of systems in near virial
equilibrium is to first order:
\begin{equation}
{\ddot I(t)} = \underbrace{{\ddot I(0)}}_{\sim 0} +
t\,\frac{d {\ddot I(0)}}{dt} + {\cal O}(t^2)
\approx 2 t \left[ 2 \dot E_{\rm kin} \!+\!
\dot E_{\rm grav} \!+\! 3 \dot P_{\rm int} V
\!-\! 3 \dot P_{\rm ext} V \right] + {\cal O}(t^2) .
\end{equation}
Hence, while an equilibrium needs similar interacting energies, the slowest
quasi-static evolution needs also similar interacting {\it powers\/} cancelling
each others. This is precisely what is suggested by FIR data on spirals.
Thus over time-scales of the order of $1-100\,\rm Myr$ a spiral may be
considered as in equilibrium at the largest scale because then the shortest
relevant evolution time-scale is the longer dynamical time. However, at
smaller scale ($\sim 1\,\rm kpc$) the local power imbalance may be large due
to either the energy output by massive stars, or the fast radiative cooling
leading to dense molecular clouds.
Over time-scales of the order of $0.1-10\,\rm Gyr$ the heating resulting from
stellar activity and gas cooling in average must mostly cancel. The feedback
mechanism regulating star formation at a constant {\it average\/} level via
dynamics is essential, otherwise one would expect the spiral as a whole either
to explode or to collapse rapidly. A tight feedback requires a minimum galaxy
size in order to reduce the fluctuations; for too small galaxies such as
dwarfs, a couple of OB stars in excess is already significant to disrupt the
galaxy: there one may find a reason for a minimum size for star forming
systems looking like spirals\footnote{An upper spiral velocity weakly
dependent on their mass follows from the maximum time-scale $\tau_{\rm max}$
to consume all the nuclear energy in galaxies with the feedback $L \approx
v^5/G$. Then $\tau_{\rm max} \ll 0.008\, G M c^2 / v^5$, so $v <
(0.008\,GMc^2/12\,{\rm Gyr})^{1/5} \approx 500^{+300}_{-200}\,\rm km\,s^{-1}$
for $M=10^{11\pm1}\,M_\odot$.}.
Over longer time-scales, say $1-10\,\rm Gyr$, the gas consumption by a
sustained star formation is substantial, and eventually the feedback weakens.
Furthermore, other dynamical processes, such as a bar instability or mergers,
can become relevant and may modify the conditions of star formation.
After such considerations the initial conditions of galaxies look secondary
for the present state. More important is the galactic micro-physics: star
formation and the ISM properties. Physically this is a much more comfortable
situation because nowadays we expect rather chaotic initial conditions of
formation as more realistic than the ordered collapses envisioned for decades.
In order to obtain galaxies with systematic properties the information must be
encoded, as for stars, within the matter instead of the initial conditions.
\section{Stellar Dynamics as a First Order Tool}\label{StDy}
Consequently, stellar dynamics seems now clearly incomplete for the
understanding of spirals over $10-15\,\rm Gyr$. However, for shorter
time-scales like a few rotational periods, since the dominating energies are
kinetic and gravitational, we can indeed approximate the full Boltzmann
equation (as applied either to stars or molecules)
\begin{equation}
\frac{\partial f}{\partial t} + \vec v \cdot \vec\nabla f
- \vec\nabla \Phi \cdot\frac{\partial f}{\partial\vec v} =
\left(\frac{\partial f}{\partial t}\right)_{\rm coll},
\end{equation}
by neglecting the right-hand side (rhs) collisional term. {\it Stellar
dynamics can thus still be applied, but over time-scales much shorter than
believed earlier}.
For the optical parts of galaxies where the stellar mass dominates, the
mean-free path of stars is large and furthermore often collective effects can
be neglected because the kinetic energy in non-systematic motion is large
enough to yield not too small Jeans' lengths. In sufficiently hot systems
collective effects are small at scale smaller than the Jeans' length, thus the
potential is replaced by a mean gravitational potential $\Phi$. In
such a mean-field approximation galactic dynamics reduces to describing the
possible orbits in the potential $\Phi$,
\begin{equation}
\ddot{\vec x} = - \vec\nabla \Phi(\vec x).
\label{Motion}
\end{equation}
This is a problem belonging to the well studied Hamiltonian mechanics. Let us
recall a few elementary general properties of such systems (e.g.~Arnold 1989),
useful to know before analyzing the phase space structure of galactic
potentials.
The first important concept is the one of {\bf orbit}, to distinguish from the
one of trajectory. An orbit is the whole subset of phase-space generated over
an {\it infinite time\/} in the past and in the future by an initial condition
in a dynamical system such as Eq.(\ref{Motion}). A {\bf trajectory} is the
subset of phase-space visited over a {\it finite\/} time by some initial point.
By construction an orbit is an {\it invariant\/} subset of phase space by the
motion, because all the points along an orbits generate trajectories within
it. Orbits are thus the most fundamental invariant blocks with which we can
build equilibrium models of stellar systems. Stellar systems may be seen as
made of orbits instead of stars.
There are several different types of orbits in 3D stellar systems which are
characterized by different {\it dimensions\/} $d$ in the 6-dimensional phase
space. Either they have $d=0$ if they don't move, they are fixed points, or
at least $d=1$ if they move, and at most $d=5$ in equilibrium systems
due to the energy integral. The only orbits with
non-vanishing phase space volumes are usually the quasi-periodic orbits with
$d=3$ and the chaotic orbits with fractal, non-integer $d$. The neighborhood
of fixed points and periodic orbits is either stable or unstable depending on
whether or not most of the neighboring trajectories remain close. In case of
instability the neighborhood is mostly made of chaotic orbits. These
properties are summarized as follow:
\begin{center}
\small
\begin{tabular}{lccc}
\tableline
\bf Orbit type & \bf Dimension $d$ & \bf Neighborhood & \bf Neighborhood \\
& & \bf if stable & \bf if unstable \\
\tableline
fixed point & 0 & periodic orbits & chaotic orbits \\
periodic & 1 & quasi-periodic orb. & chaotic orbits \\
quasi-periodic & 2, or 3 & quasi-periodic orb. & --- \\
chaotic & fractal: $1<d \leq 5$ & --- & chaotic orbits \\
\tableline
\end{tabular}
\end{center}
The chaotic orbit class is the class of the orbits not classifiable in the
other
classes (like the `peculiar' galaxies in the Hubble classification). It
includes different types of orbits which could be further sub-classified
according to their multifractal spectrum and fractal dimension.
As understood long ago by Poincar\'e, periodic orbits are certainly, after the
few fixed points, the most important class of orbits, not because many stars
are on them, but because when stable {\it they summarize the surrounding
phase-space}. The stable periodic orbits are always surrounded by concentric
quasi-periodic orbits, or tori. Those tori are much more numerous. This
structure of concentric tori surrounding the periodic orbits in Hamiltonian
systems is the fundamental property allowing to conceptualize phase-space by
its stable periodic orbits. This property is still insufficiently appreciated,
even by some dynamicists, motivating these elementary reminders.
Among the robust properties of orbits we note: 1) the shapes of the main
periodic orbits depend only on the potential {\it symmetries}, therefore we do
not need a precise description of the galaxy to know its orbital structure, 2)
the periodic orbits are insensitive to {\it low\/} spatial frequency
perturbations, so moderate low frequency symmetry breaking (such as a bar
bending out of the plane) do not change much the orbital structure, it only
deforms it slightly, and 3) the stable periodic orbits become {\it
attractors\/} by a weak dissipative perturbation. A small dissipation in fact
condenses the trajectories of the system toward its most fundamental periodic
orbits.
However, periodic orbits are fragile against {\it high\/} spatial frequency
perturbations anywhere along their path. For example, a small accumulation of
mass just at the galaxy center may profoundly change the stability of the
radial periodic orbits crossing the center, so the structure of the
surrounding phase space.
If we consider now all the neglected effects that should be taken into account
to reflect all the complexity of a real spiral, such as gas dynamics and star
formation, among all the possible structures of a pure stellar dynamical
representation {\it the main stable periodic orbits are certainly the ones that
survive the longest and are the best suited to describe evolution features}.
They become inevitable {\it concepts\/} to understand complex non-linear
dynamical processes. This reality can be experienced by those dealing with the
complexity of $N$-body simulations, also including gas. It is quite obvious
that, for example, when we think about a rotating disk we make a ``thought
economy'' by first considering only its circular orbits, even if no stars
follow exactly such orbits.
\section{Phase-Space Structure of Barred Galaxies}\label{PhSpStructure}
The main dynamical feature of a barred galaxy is shaped by its only general
isolating integral, the Jacobi integral, also called the ``energy'' in a
rotating frame of reference. The Hamiltonian in the rotating frame of the bar
reads
\begin{equation}
H = {\textstyle\frac{1}{2}}
(p_x^2+p_y^2+p_z^2) + \Phi(x,y,z) - \Omega_{\rm p}(x p_y - y p_x),
\end{equation}
where the momenta $p_x=\dot x-\Omega_{\rm p}y$, $p_y=\dot y+\Omega_{\rm p}x$,
and $p_z=\dot z$ are the velocity components in the instantaneous parallel
inertial frame. The zero-velocity surface (ZVS) in the rotating frame
\begin{equation}
H_0 = \Phi(x,y,z) - {\textstyle\frac{1}{2}}\Omega_{\rm p}^2(x^2+y^2),
\end{equation}
bounds motion in space at low ``energy'' inside a football shape (nearly like
a spheroid) inside corotation. Higher ``energy'' particles inside the bar
can escape first through tunnels in the ZVS near the end of the bar, and then
through the whole corotation circle at still higher ``energy''. Four of the
five fixed points, the Lagrangian points, are located at the corotation.
The closest simple analytical model of a barred galaxy is an axisymmetric
disk, in which the basic unperturbed orbits are the circular orbits in the
principal plane. The three orbital frequencies in the plane are the rotation
frequency $\Omega$, the radial and vertical epicyclic frequencies $\kappa$ and
$\nu$. These frequencies squared depend only on the potential local
derivatives at $z=0$ (see Binney \& Tremaine 1987, p.~121). Circular orbits
are then stable if $\Omega^2>0$, $\kappa^2>0$, and $\nu^2>0$.
If we consider the bar as a perturbation of the circular case, the circular
orbits are perturbed by the non-axisymmetric potential rotating at the
frequency $\Omega_{\rm p}$. The resulting transverse deviations $\xi$ from
the circular orbits (either radial or vertical) can be described by the Hill's
equation, thats is the equation of a harmonic oscillator the natural frequency
$\omega_0$ of which is periodically modulated:
\begin{equation}
\ddot \xi + \omega_0^2\left[1 + \epsilon (t)\right] \xi = 0,
\label{Hill}
\end{equation}
with $\epsilon(t)$ a small periodic function of frequency $\omega_{\rm per}$.
A general theorem (see e.g.~Arnold 1989) gives the {\it parametric
resonances\/}
conditions for arbitrarily {\it small\/} modulation $\epsilon(t)$:
\begin{equation}
\frac{\omega_0}{\omega_{\rm per}} =
\pm \frac{k}{2}, \quad k=0, 1, 2,\ldots\infty.
\end{equation}
With this theorem we can derive elegantly the resonance conditions, because we
do not need to know the precise form $\epsilon$ of the bar perturbation, but
only the perturbing frequency $\omega_{\rm per}$. To determine the
resonance widths one does need however to know the particular form of
$\epsilon$.
The radial and vertical epicyclic frequencies $\kappa$ and $\nu$ are the
natural oscillations frequencies around the circular orbits, and the bar
perturbing frequency is $2(\Omega \mp \Omega_{\rm p})$ for direct/retrograde
circular orbits. The factor 2 comes from the bi-symmetry of the bar (the
``number of arms"). Then we obtain the classical resonances when
\begin{equation}
\frac{\kappa }{ \Omega \mp \Omega_{\rm p}} = \pm m , \qquad
\frac{\nu }{ \Omega \mp \Omega_{\rm p}} = \pm n ,
\quad \qquad m,n = 0, 1, 2,\ldots\infty.
\end{equation}
The main direct resonances encountered in barred galaxies occur for
$m=n=\infty$ (corotation), $m=n=2$ (radial and vertical inner Lindblad
resonances), $m=-2$ (radial outer Lindblad resonance), and $m=n=4$ (radial and
vertical ultra-harmonic resonances). Around corotation we have the piling up
of an infinity of higher order resonances. The main purpose of recalling
these simple considerations is to stress that the treatments of radial and
vertical resonances are the same, there is no ground to neglect the vertical
resonances, particularly because bars extend in the central regions of the
galaxies which are not much thinner than large.
In addition, the knowledge of the resonances gives the approximate {\it
shapes\/} of the nearby orbits. For having elongated oval orbits we always
expect that the conditions are close to have a 2/1 Lindblad resonance, because
an oval is a 2/1 perturbation of a circle. Rectangular shapes are associated
with 4/1 resonances, etc. This applies in the vertical direction too, a 2/1
resonance is associated with 2/1 vertical oscillations out of the plane, etc.
Once the typical resonances in barred galaxies are known, it is simple to
guess how the main periodic orbits look like. Detailed numerical calculations
of orbits can help in quantifying precisely the orbit positions and shapes,
and also in finding higher order periodic orbits. Numerous works have been
made in this area allowing to grasp the main general features of the bar
dynamics (see references, e.g., in the review of Sellwood \& Wilkinson 1993).
The general properties are well understood in a descriptive way by several
numerical calculation works. Although the main periodic orbits families in the
plane and in 3D, including their shape, are well known by now, there is still
no simple but realistic analytical model of galactic bars.
Exactly similar considerations to circular orbits apply for radial periodic
orbits. A slight complication is that for the $z$-axis orbits in a rotating bar
the two transverse frequencies are coupled by rotation, which leads to a
generalizations of the Hill's equation and new possibilities of parametric
resonances via ``complex instability'' (Pfenniger 1987). But in non-rotating
bars, Hill's equation (\ref{Hill}) applies for each transverse component
of the three types of axial orbits, the perturbing frequency being the
oscillation frequency of the radial orbit. Then a 2/1 resonance leads to
banana shaped orbits (shaping the neighboring phase space) which favors a
bending instability in non-rotating ellipsoidal systems too.
For oblate axisymmetric disks with positive density (constraining the
frequencies by $2\Omega^2\! <\! \kappa^2+\nu^2$, and $\Omega\! =\! c\nu$,
where $c\!<\!1$ is the potential axis ratio) a central mass concentration {\it
always\/} produces low order radial {\it and vertical\/} resonances. Now as
long as the potential remains axisymmetric the width of the resonance is zero,
and these resonances are ineffective. But as soon as a triaxial deformation
exists, the resonance widths grow which fosters chaotic motion. Then
diffusion in the vertical direction is natural near the center of a galaxy,
especially if slow dissipative processes accrete mass near the bottom of the
potential. Then the idea that a bulge may naturally grow in a barred
potential is dynamically justified (Pfenniger 1984, 1985; Pfenniger \& Norman
1990).
The dissolution of a bar by a growing central mass concentration is an example
of possible qualitative prediction allowed by knowing the periodic orbits of
bars (Hasan \& Norman 1990; Pfenniger \& Norman 1990; Hasan et al.~1993).
Since the radial and vertical ILR radii increase rapidly when the central mass
concentration increases $R_{\rm ILR} \propto M_{\rm cm}^{2.8}$, and the
elongated orbits supporting the bar are replaced by orbits perpendicular to the
bar inside the growing ILR radius, the elongated shape of the bar is rapidly no
longer compatible with the growing central mass concentration. Orbit
calculations show that around $5-15\%$ percents of the disk mass as a central
mass concentration should be sufficient to destroy the bar. Since the vertical
ILR is typically close to the radial ILR, one expects an extended 3D diffusion
of the initial central part of the disk.
To conclude this Section, simple periodic orbits considerations show that
barred galaxies have numerous resonances and associated chaotic zones from
which we can {\it expect\/} instabilities. For studying these we need however
techniques able to take self-gravitation into account, such as $N$-body
methods.
\section{Types of Instabilities in Collisionless Gravitating Systems}
\label{InCoGR}
Many works have considered the problem of self-consistent stability with the
linearized Collisionless Boltzmann equation. Here we just point out how this
method may help in distinguishing two different sources of instabilities in
real stellar systems, simultaneously clarifying the limitations of the method.
Suppose we know a solution $f_0$, $\Phi_0$ ($\rho_0=\int{f_0 \,d^3v}$) of the
Collisionless Boltzmann and Poisson equations. Then, as usual, we want to
describe the evolution of small perturbations $\delta f$, $\delta\Phi$ around
the assumed known solution $f_0$, $\Phi_0$ in the linear approximation.
Linearizing the equations we obtain:
\begin{eqnarray}
\label{linCB}
& \overbrace{\frac{\partial \delta f}{\partial t} +
\vec v \cdot \vec\nabla\delta f-\vec\nabla\Phi_0 \cdot
\frac{\partial \delta f}{\partial \vec v}}^{
\mbox{\small orbital behavior, mixing of $\delta f$}}
\ \equiv\
{\displaystyle{\frac{{\rm D} \delta f }{{\rm D} t}} }
\quad=\quad
\overbrace{\vec\nabla\delta\Phi \cdot
\frac{\partial f_0}{\partial \vec v}}^{
\mbox{\small source of $\delta f$}} & ,\\[1mm]
& \vec\nabla\cdot\vec\nabla\delta\Phi \ =\ 4\pi G\delta\rho &.
\end{eqnarray}
As set, Eq.(\ref{linCB}) shows that the variations $\delta f$ depend on two
simple effects.
First, the rhs of Eq.(\ref{linCB}) shows that to modify the density of a volume
element moving around a given unperturbed orbit, a gradient of $f_0$ in the
velocity space is required. Furthermore, only the force fluctuation component
parallel to the velocity gradient contributes. Large density gradients in
velocity space favor instabilities parallel to the gradient (so radial in cold
rotating disks).
The second factor of effective instability is a little more subtle. The
left-hand side (lhs) term, $D\delta f/D t$, describes the Lagrangian derivative
in phase space of a fluctuating element along an unperturbed orbit. So, even if
the rhs vanishes, or is very small, the {\it form\/} of a compact element may
be rapidly modified into a threaded structure the fineness of which becomes
smaller than allowed by the various approximations. This occurs precisely
around unstable orbits in the unperturbed system, such as chaotic orbits. This
sensitive dependence means that the neglected terms (higher order terms,
collision and dissipative terms) become rapidly relevant for the real
system\footnote{The same occurs in many other problems: when the linear
operator $A$ of the problem $Ax = B$ has widely different {\it singular
values\/} then the problem is {\it ill-posed\/} and the solution $x$ is
ill-defined.}.
Thus in stellar systems which contain chaotic regions of phase space, so
typically around resonances, an initial compact volume element of $\delta f$
is rapidly distorted into very intermingled shapes, the slightest perturbation
is exponentially amplified, and non-linear considerations are very soon
necessary. In such a situation it is more natural to ascribe the resulting
instability not to the rhs source term, but to the lhs phase space structure.
On the other hand, when the orbital motion is regular, a compact initial
volume element remains compact much longer, the lhs is well-behaved and a
growing mode can only be attributed to the rhs term. In such cases
the instability can be ascribed to the full distribution function
$f_0$. Note that analytical models have usually a regular phase space, so the
problem appears preferentially in numerical models not biased against chaos.
In summary, in order to progress in the understanding of stellar systems and
their instabilities we can divide their analysis in two distinct parts:
\begin{enumerate}
\item
The understanding of the phase-space structure (i.e.~the orbits) of the
potential $\Phi_0$ independently of the velocity structure of $f_0$.
The resonance
regions and the chaotic zones are then susceptible to seed collective
instabilities and non-linear effects not describable with a linear theory.
\item
The understanding of the specific self-gravity effects associated with
particular $f_0$'s consistent with a given $\Phi_0$. One may generally expect
that sufficiently cold distribution functions should produce instabilities
not directly related to the previous aspect. For a regular phase-space the
linear mode analysis should work much better.
\end{enumerate}
As illustration, in spirals the bar instability comes mainly from a too cold
$f_0$, because phase space surrounding stable circular orbits is regular,
while the bending instability leading to peanut-shaped bars, discussed below,
is mainly due to a 2/1 vertical resonance and is little dependent on the
velocity structure of $f_0$.
\section{Smoothness Assumption}
An assumption which is most of the time not discussed is that $f$ should be a
differentiable function. In ordinary gases with molecules having short range
interactions and frequent collisions, any irregular distribution is rapidly
smoothed out locally by the ``molecular chaos''. The smoothing principle
comes then from the short relaxation time and the lack of long-range
interactions, which allows a fast local decorrelation and homogenization of
the particles.
In collisionless gravitating systems this rapid smoothing is far from being
obvious and demonstrated. As soon as instabilities occur, the long range of
gravitation correlates the fluctuating part $\delta f$ much more than in
ordinary gas, and since we lack of a smoothing principle, no good reason other
than commodity allows to assume that a differentiable $\delta f$ is a valid
assumption allowing its use in the variational differential equation.
In fact numerical experiments indicate rapid limitations of the linear theory.
For example Toomre \& Kalnajs (1991) have simulated a small portion of a
self-gravitating flat disk which is maintained slightly unstable. The results
show clearly that long range correlations are rapidly created and do persist;
a fractal scale invariant and dynamical state follows. The linearization
approach is of no use for describing such states. Another well documented
case of gravitational instability occurs in the numerous simulations of
expanding universes. The runs develop usually non-linear fractal structures
which cannot be derived from a linear study. In such a situation, analogous
to turbulence in fluids, the system has a sensitive dependence on small-scale
effects and perturbations. The collisional as well as weak dissipative
effects can be crucial. Therefore we should be cautious about drawing
conclusions without a thorough non-linear analysis. $N$-body methods are
presently essential to study non-linearities with the advantage that they
include {\it non-vanishing collisional effects}.
\section{Self-Consistent 3D Barred Galaxy Models}
In the early 80's it was generalized from a single peculiar edge-on S0 galaxy
with a small(!) bulge (NGC\,4762) that bars are generally flat ($a/c\sim10$)
(see Kormendy 1982). Theory says such flat bars are implausible to maintain
for a long time since they imply highly anisotropic velocity distribution, a
strong velocity gradient of $f_0$, and strong vertical resonances from the bar
potential.
In fact in 3D $N$-body simulations most of the initially flat bars thicken
rapidly and are subject to bending instabilities transverse to the plane
(Friedli \& Pfenniger 1990; Pfenniger \& Friedli 1991 (PF91); Raha et
al.~1991). The end-result of these bending instabilities are box- or
peanut-shaped bars (Miller \& Smith 1979): stable structures over several Gyr
resembling much the observed peanut-shaped bulges according to viewing angle
of the bar (Combes \& Sanders 1981; Combes et al.~1990; see also references in
Merrifield, this volume). The peanut-shaped bars look round when viewed
end-on.
While Raha et al.~have identified this instability with the ``fire-hose
instability'' (the out of the plane instability of an infinite homogeneous
thin sheet), the orbital analysis in PF91 shows clearly that the fire-hose
instability picture is a too rough analogy to predict the bending instability
main characteristics such as its size, the principal mode of bending, and the
instability threshold. The detailed analysis of the $N$-body bar run in PF91
(see also Pfenniger 1990) greatly improves the understanding of the
instability by considering not only 1) its simple morphological and kinematic
description, but also 2) its potential resonances, which give a first idea on
the orbits, further 3) its periodic and other orbits at different times, and
then 4) its distribution function changes. Finally the whole non-linear
evolution of the ensemble can be much better understood.
Non-linearities are essential all along the phenomenon: the instability starts
near the vertical 2/1 resonance, so is associated with this resonance, it
chooses to bend up or down randomly from the fluctuations, with precisely a
2/1 banana shaped mode, then it symmetrizes rapidly its vertical profile. The
instability saturates around 2/2/1 stable periodic orbits, which pre-exist,
accompany and survive the instability, explaining the final peanut shape. As
the instability proceeds, the vertical 2/1 resonance sweeps the bar
particles from low to high ``energy'', allowing most of them to leave the
plane up to heights and distances allowed by the ZVS. Thus, the bending
instability is a nice example of a gravitational instability {\it little
dependent on the velocity space structure of $f_0$}, it occurs also in bars
initially far from being flat. Clearly, pure orbital considerations, lacking
of the self-gravity, are insufficient to predict firmly the bending and time
evolution. But a linear mode analysis such as in Merritt \& Sellwood (1994)
is unable to describe the whole phenomenon too\footnote{In order to explain
this surprisingly fierce instability, these authors must finally distinguish
it from the fire-hose case and rely on giving an explanation in term of the
``oscillations of stars'' with a 2/1 frequency ratio. Of course the orbit
description is just the same in a systematic fashion. The reported strong
grid dependence in their numerical results seems natural once realized that
around a major resonance chaos causes a sensitive dependence not only on the
neglected non-linear terms, but also on the numerical technique; this is just
the signature of a so-called ``ill-posed'' problem mentioned above.}. It is
only the association of the two approaches together with the $N$-body
technique which allows a detailed understanding of the entire process.
As noted in Section \ref{PhSpStructure}, the knowledge of the orbital
structure in bars is also useful for predicting conditions of their
destruction. This can be tested with $N$-body runs, which confirm that indeed
a mass accumulation within the original ILR changes the orbital structure to
an extent that elongated bars are no longer possible, but only 1-3\% of
additional mass is required. Such an instability involves either a
dissipative factor in order to grow a central mass concentration (Friedli \&
Pfenniger 1991; Sellwood, this volume), or dynamical friction of dense
satellites (Pfenniger 1991, 1992, 1993). In this respect an accretion of
5-10\% of mass of satellites inside the bar region is able to destroy the bar
and form a much bigger bulge like the one in M104.
Other indications from $N$-body simulations that the dynamical picture that we
propose is essentially correct, are 1) the reshaping of an initial disk into
an exponential disk plus a steeper bulge-like profile in the bar region
results automatically from the dynamical effects of a bar (Hohl 1971), 2) in
such conditions the velocity ellipsoid tends to be anisotropic with an
exponential profile consistent with the Milky-Way observations (Lewis \&
Freeman 1989): $\sigma_R^2 \propto \exp(-R/h)$, with $\sigma_R> \sigma_{\phi}
>\sigma_z$ (PF91).
All these different kinds of dynamical evolutions give a living character to
the spirals, contrasting with older static views. It is then natural to assume
that bulges may grow secularly from their disk, which is consistent with the
observations that bulge stars are metal rich (Rich 1992). The general picture
of disk dynamical evolution is a sequence of barred and unbarred phases through
which the bulge size grows irreversibly. This is one of the several arguments
for proposing an evolution of the spirals along the Hubble sequence from Sd to
Sa (Pfenniger et al.~1994). It is not clear yet how bars can be recurrent,
because a central mass concentration once formed should prevent further bar
formation unless a large amount of cold and angular momentum rich material is
able to accrete quietly on the galaxy disk without heating it.
Finally, a general observed trend concerns the apparent near integrability of
stable stellar systems. The phase-space analysis of $N$-body runs, such as in
PF91, shows that the final most stable configurations are free of strong
chaotic orbits and have remarkably simple phase space structures reminiscent of
integrable systems, at least for the populated phase space regions. The
remaining ubiquitous weak chaos is indistinguishable from the particle noise.
This is consistent with the stability considerations given in
Sect.~\ref{InCoGR}: chaotic regions are likely to seed collective
instabilities.
\section{Final Remarks and Conclusions}\label{Conclusions}
The first order dynamics determining the shape and evolution of barred
galaxies is now well understood. The component decomposition of these
galaxies in bulge, bar, and disk is dynamically illusory since many stars
constantly switch from one ``component'' to the next and back. The gross
morphology of bars, their sense of evolution is well understood, and many
observed features like the exponential disks, the peanut-shaped bars, and the
SB0 profiles can be remarkably reproduced by numerical means.
Finer morphologies like rings, ansae, double bars are now at the limit of the
numerical resolution in fully self-consistent models.
For understanding galaxies the knowledge of the periodic orbits turns out to
be an invaluable tool to simplify the description of the dynamical processes,
and to develop an intuition allowing even correct guesses without computer!
Bars can form spontaneously from their disk, peanut-shaped bulges can grow
from their parent bar, and bars can fully dissolve in spheroidal bulges
through a central accumulation of mass, so secular dynamical evolution of
spirals appears as natural, with typical time-scales of the order of
$0.1-10\,\rm Gyr$ for changing significantly the spiral type.
\@startsection{paragraph}{4}{1em This work is supported by the Swiss FNRS.
|
\section{Introduction}
The problem of computing quantum corrections to a classical energy in a
complicated theory such as Einstein gravity, can be approached by performing
an analysis of the thermodynamical quantities that characterize the system
under consideration. This analysis can be done by means of the computation
of the free energy of the system at a given volume and temperature. Defining
the Euclidean action as
\begin{equation}
\label{aa1}\hat I\left[ g\right] =-\frac 1{16\pi G}\int_{{\cal M}}d_{}^4x%
\sqrt{g}R\left( g\right) -\frac 1{8\pi G}\int_{{\cal \partial M}}d_{}^3x%
\sqrt{h}K_i^i,
\end{equation}
where $R$ is the Ricci scalar of the metric $g_{\mu \nu }$ and $K_i^i$ is
the trace of the second fundamental form, one can compute quantum
corrections to the Euclidean action, of a fixed background geometry with the
appropriate boundary conditions. Since we wish to study quantum
fluctuactions with respect to the Schwarzschild geometry, there are two
types of boundary conditions that are related to the background under
consideration:
\begin{description}
\item[a)] Asymptotically Euclidean (AE),
\item[b)] Asymptotically Flat (AF).
\end{description}
An AE background metric is one in which the metric approaches the flat
metric on $R^4$ outside some compact set. The boundary at infinity is
topologically $S^3$. An AF background metric is one in which the metric
approaches the flat metric $R^3\times S^1$ outside some compact set. The
boundary of infinity is topologically $S^2\times S^1$. Anyway, we could
consider a different point of view based on the Hamiltonian approach. In
this framework one is able to deal with three dimensional fields
configurations separated out by the time variable. The advantage of the
Hamiltonian framework is that one can manage from the beginning with energy
fields configurations which give directly the quantum corrections to the
classical term. To do this, the first step is the separation of the three
dimensional space from the time by means of the ${\cal ADM}$ variables $\cite
{ADM}$. In terms of these variables the line element becomes
$$
ds_{}^2=g_{\mu \nu }\left( x\right) dx^\mu dx^\nu =-N_{}^2\left( dx^0\right)
^2+g_{ij}\left( N^idx^0+dx^i\right) \left( N^jdx^0+dx^j\right)
$$
\begin{equation}
\label{aa2}=\left( -N^2+N_iN^i\right) \left( dx^0\right)
^2+2N_jdx^0dx^j+g_{ij}dx^idx^j.
\end{equation}
$N$ is called the {\it lapse function}, while $N_i$ the {\it shift function}%
. The associated matrix representation of $g_{\mu \nu }$ is
\begin{equation}
\label{aa3}g_{\mu \nu }=\left(
\begin{array}{cc}
-N^2+N_iN^i & N_j \\
N_i & g_{ij}
\end{array}
\right) ,
\end{equation}
with the inverse given by
\begin{equation}
\label{aa3aa}g_{}^{\mu \nu }=\left(
\begin{array}{cc}
-\frac 1{N^2} & \frac{N_{}^j}{N^2} \\ \frac{N^i}{N^2} & g_{ij}-\frac{N_iN^i}{%
N^2}
\end{array}
\right) .
\end{equation}
Roman indices will be raised and lowered by the induced metric on the three
surface $x^0$. In terms of the ${\cal ADM}$ variables, the initial action
can be written as a sum of a ``{\it kinetic}'' and a ``{\it potential}''
term
\begin{equation}
\label{aa4}I=\frac 1{16\pi G}\int_{}dx_{}^0N\int dx_{}^3\text{ }^{\left(
3\right) }\sqrt{g}\left\{ \left( K_{ij}K^{ij}-K^2\right) +\text{ }^{\left(
3\right) }R\right\} ,
\end{equation}
where $K_{ij}^{}=\frac 1{2N}\left( N_{i|j}+N_{j|i}-g_{ij,0}\right) $ is
called the second fundamental form and ``%
\mbox{$\vert$}
'' means covariant differentiation with respect to the $3D$ gravitational
background, $K$ is the trace of the second fundamental form, $^3R$ is the
scalar curvature in $3D$, $^{\left( 3\right) }\sqrt{g}$ is the invariant of
the metric in $3D$. In this form the time derivative is isolated and it is
possible the computation of the conjugate momentum to $g_{ij}$, that is
\begin{equation}
\label{aa5}\pi ^{ij}=\frac{\delta I}{\delta \dot g_{ij}}=\left( -K^{ij}+%
\text{ }^{\left( 3\right) }g^{ij}K\right) \frac{\sqrt{g}}{16\pi G}.
\end{equation}
By a Legendre transformation we calculate the Hamiltonian
$$
H=\int d^3x\left\{ \left[ \pi ^{ij}\dot g_{ij}-\frac 1{16\pi G}\left[ \left(
K_{ij}K^{ij}-K^2\right) +\text{ }^{\left( 3\right) }R\right] N\text{ }%
^{\left( 3\right) }\sqrt{g}\right] \right\} =
$$
\begin{equation}
\label{aa6}%
\displaystyle \int
d^3x\left\{ N\left[ \frac{16\pi G}{^{\left( 3\right) }\sqrt{g}}\left( \pi
_{ij}\pi ^{ij}-\frac \pi 2^2\right) -\text{ }^{\left( 3\right) }\sqrt{g}%
\frac{^{\left( 3\right) }R}{16\pi G}\right] +N_i\left( 2\pi
_{|j}^{ij}\right) \right\} .
\end{equation}
The first term of $\left( \ref{aa6}\right) $ has a quadratic structure in
the momenta, suggesting, as a first approximation that we could compute
quantum corrections to the energy expanding $^3R$ in terms of the quantum
field fluctuations with respect to a given background, e.g. the
Schwarzschild background. Since we wish to understand the pure gravitational
vacuum, we neglect the matter fields and since in our approach only the
spatial part of the background comes into play our background is of the
wormhole type \cite{MTW}. Although the energy computation at quantum level
is very unclear because of the constraint coming from the lapse function we
will adopt the Hamiltonian approach the same and to this purpose a simple
framework will be shown in section \ref{p2a}. The rest of the paper is
structured as follows, in section \ref{p2} we analyze the orthogonal
decomposition of the Hamiltonian both in tangent and co-tangent space, in
section \ref{p3} we define the gaussian wave functional for gravity in
analogy with non-abelian gauge theories, in section \ref{p4} we give some of
the basic rules to perform the functional integration and we define the
Hamiltonian approximated up to second order, in section \ref{p5}, we analyze
the spin-2 operator or the operator acting on transverse traceless tensors,
only for positive values of $E^2$. We summarize and conclude in section \ref
{p6}.
\section{ The Hamiltonian on the slice}
\label{p2a}
After the introduction of ${\cal ADM}$
variables, we recall that the Hamiltonian is:
\begin{equation}
\label{a1}
\begin{array}{c}
H=\int d^3x(N{\cal H+}N_i{\cal H}^i)
\end{array}
\end{equation}
where
\begin{equation}
\label{a2}{\cal H}{\bf =}G_{ijkl}\pi ^{ij}\pi ^{kl}\left( \frac{l_p^2}{\sqrt{%
g}}\right) -\left( \frac{\sqrt{g}}{l_p^2}\right) \text{ }^{\left( 3\right)
}R\ \text{ (Super Hamiltonian)}
\end{equation}
and
\begin{equation}
\label{a3}{\cal H}^i=-2\pi _{|j}^{ij}\ \text{ (Super Momentum).}
\end{equation}
In $\left( \ref{a3}\right) $ the derivative is covariant with respect to the
$3D$ background field, $l_p^2$ is the usual Planck mass, and $G_{ijkl}$ is
the Wheeler-DeWitt (WDW) metric. If we look at $N$ and $N_i$ as fundamental
objects describing the correct variables, by variational principles we
obtain the usual constraint equations, that is
\begin{equation}
\label{a3aaa}
\begin{array}{c}
{\cal H}\text{ }=0\text{, }{\cal H}^i\text{ }=0\text{ Classical} \\ {\cal H}%
\Psi \text{ }=0\text{, }{\cal H}^i\text{ }\Psi =0\text{ Quantum } \\
\end{array}
\end{equation}
The usual interpretation of these equations is that they represent
constraints on the initial value problem or in other words they represent
gauge invariance with respect to time and gauge transformations.
Nevertheless we have a chance to define and computing energy if we restrict
on a given hypersurface, fixing the lapse function to a constant. Such a
gauge choice is the most appropriate for wormhole configurations of the
background geometry \cite{Garay} and in particular for the Schwarzschild
wormhole. By rescaling time intervals, we obtain\footnote{%
A different treatment, but close to our present approach, is based on the
separation between dynamical variables and embedding parameters and can be
found in Ref. \cite{Miller}}
\begin{equation}
\label{a4}N=1.
\end{equation}
Actually, this choice is compatible with the suspension constraint that one
can adopt in quantum cosmology to obtain a Schr\"odinger-like equation,
provided at the end of the calculation one assures that the gauge invariance
is restored \cite{Garay}\cite{J.J. Halliwell}. Then the Hamiltonian in the
time-like gauge is
\begin{equation}
\label{a5}H=\int d_{}^3x{\cal H}\text{ }=\int d_{}^3x\left[ G_{ijkl}\pi
^{ij}\pi ^{kl}\left( \frac{l_p^2}{\sqrt{g}}\right) -\left( \frac{\sqrt{g}}{%
l_p^2}\right) \text{ }^{\left( 3\right) }R\right]
\end{equation}
Since $\left( \ref{a5}\right) $ is valid on a ``{\it fixed}'' hypersurface,
to recover the original equation, i.e. $\left( \ref{a1}\right) $, the
correct procedure to perform will be a summation over all possible lapses;
this means that the constraint $\left( \ref{a3aaa}\right) $ (at classical or
at quantum level) will be restored after this summation.
\section{Ultralocal Metrics as a tool for decomposing tensor fields}
\label{p2}
Instead of performing calculations in the usual WDW metric we will use a
one-parameter family of supermetrics to disentangle gauge modes from
physical deformations. For this reason we require an orthogonal
decomposition for both $\pi _{ij\text{ }}^{}$and $h_{ij}^{}$, that is we
need a metric on the space of deformations, i.e. a quadratic form on the
tangent space at h. The condition of ultralocality, where $G_{ijkl}$ locally
depends on $g_{ij}^{}$ but not on its derivatives, could be taken as a good
condition for the functional measure, explicitly:
\begin{equation}
\label{a6}\left\langle h,k\right\rangle :=\int_{{\cal M}}^{}\sqrt{g}G_\alpha
^{ijkl}h_{ij}^{}\left( x\right) k_{kl}^{}\left( x\right) d_{}^3x,
\end{equation}
$$
\text{where}
$$
\begin{equation}
\label{a7}
\begin{array}{c}
G_\alpha ^{ijkl}=(g_{}^{ik}g_{}^{jl}+g_{}^{il}g_{}^{jk}-2\alpha
g_{}^{ij}g_{}^{kl}).
\end{array}
\end{equation}
The WDW metric, introduced in $\left( \ref{a2}\right) $, is just $\left( \ref
{a7}\right) $ with $\alpha =1$. The ``inverse'' metric is defined on
co-tangent space and it assumes the form
\begin{equation}
\label{a8}\left\langle p,q\right\rangle :=\int_{{\cal M}}^{}\sqrt{g}%
G_{ijkl}^\beta p_{}^{ij}\left( x\right) q_{}^{kl}\left( x\right) d_{}^3x,
\end{equation}
$$
\text{where}
$$
\begin{equation}
\label{a9}
\begin{array}{c}
G_{ijkl}^\beta =(g_{ik}^{}g_{jl}^{}+g_{il}^{}g_{jk}^{}-2\beta
g_{ij}^{}g_{kl}^{}). \\
\end{array}
\end{equation}
with $\alpha +\beta =3\alpha \beta $, so that
\begin{equation}
\label{a10}G_\beta ^{ijnm}G_{nmkl}^\beta =\frac 12\left( \delta _k^i\delta
_l^j+\delta _l^i\delta _k^j\right) .
\end{equation}
These are non-degenerate bilinear forms for $\alpha \neq \frac 1{3\text{ }}$%
, for $\alpha =\frac 1{3\text{ }}$ the metric is not invertible and becomes
a projector onto the tracefree subspace, while is positive definite for $%
\alpha <\frac 1{3\text{ }}$ and of mixed signature for $\alpha >\frac 1{3%
\text{ }}$ with infinitely many plus as well as minus signs.
We have now the desired decomposition on the tangent space of 3-metric
deformations $h_{ij}^{}$:
\begin{equation}
\label{a11}h_{ij}^{}=\frac 13hg_{ij}^{}+\left( L\xi \right)
_{ij}^{}+h_{ij}^{\bot },
\end{equation}
$$
\text{or, in matrix form,}
$$
\begin{equation}
\label{a12}h=\frac 13hg+\left( RangeL\right) +\left( KerL^{\dagger }\right)
,
\end{equation}
where the operator $L$ maps $\xi _i^{}$ into symmetric tracefree tensors,
according to\cite{MazMot}\cite{York}$,$
\begin{equation}
\label{a13}\left( L\xi \right) _{ij}^{}=\nabla _i^{}\xi _j^{}+\nabla
_j^{}\xi _i^{}-\frac 23g_{ij}^{}\left( \nabla \cdot \xi \right) .
\end{equation}
Consequently, the inversion of the metric $\left( \ref{a8}\right) \left(
\text{that is}\left( \ref{a9}\right) \right) ,$guarantees us the same
decomposition also in phase space (co-tangent space).
\section{The Gaussian Wave Functional}
\label{p3}
There are some reasons to introduce a gaussian wave functional for the
description of the vacuum state in gravity. Starting from the analogy
between nonabelian gauge theories and gravity we illustrate how gaussian
wave functional work in the former case. We define \cite{Kerman}
\begin{equation}
\label{b1}\Psi \left[ A_i^a\left( \overrightarrow{x}\right) \right] ={\cal N}%
\exp \left\{ -\frac 14\int d_{}^3xd_{}^3y\delta A_i^a\left( \overrightarrow{x%
}\right) G_{ij}^{-1ab}\left( \overrightarrow{x},\overrightarrow{y}\right)
\delta A_j^b\left( \overrightarrow{y}\right) \right\}
\end{equation}
where ${\cal N}$ is a normalization factor and where
\begin{equation}
\label{b2}\delta A_i^a\left( \overrightarrow{x}\right) =A_i^a\left(
\overrightarrow{x}\right) -\overline{A_i^a}\left( \overrightarrow{x}\right)
.
\end{equation}
In equation $\left( \ref{b2}\right) $, $\overline{A_i^a}\left(
\overrightarrow{x}\right) $ is a background field which can be treated as a
variational parameter together to the function $G_{ij}^{ab}\left(
\overrightarrow{x},\overrightarrow{y}\right) $in $\left( \ref{b1}\right) $.
From the definition in $\left( \ref{b1}\right) $ one finds the expectation
values
\begin{equation}
\label{b3}
\begin{array}{c}
\left\langle \Psi |A_i^a\left(
\overrightarrow{x}\right) |\Psi \right\rangle =\overline{A_i^a}\left(
\overrightarrow{x}\right) , \\ \\
\left\langle \Psi |A_i^a\left(
\overrightarrow{x}\right) A_j^b\left( \overrightarrow{y}\right) |\Psi
\right\rangle =\overline{A_i^a}\left( \overrightarrow{x}\right) \overline{%
A_j^b}\left( \overrightarrow{y}\right) +G_{ij}^{ab}\left( \overrightarrow{x},%
\overrightarrow{y}\right) , \\ \\
\left\langle \Psi |E_i^a\left(
\overrightarrow{x}\right) |\Psi \right\rangle =0, \\ \\
\left\langle \Psi |E_i^a\left(
\overrightarrow{x}\right) E_j^b\left( \overrightarrow{y}\right) |\Psi
\right\rangle =\frac 14G_{ij}^{-1ab}\left( \overrightarrow{x},%
\overrightarrow{y}\right) , \\ \\
\left\langle \Psi |B_i^a\left( \overrightarrow{x}\right) |\Psi \right\rangle
=\overline{B_i^a}\left( \overrightarrow{x}\right) +\frac 12\epsilon
_{ijk}^{}f^{abc}G_{ij}^{ab}\left( \overrightarrow{x},\overrightarrow{y}%
\right) ,
\end{array}
\end{equation}
where
\begin{equation}
\label{b4}E_i^a\left( \overrightarrow{x}\right) =-i\frac \delta {\delta
A_i^a\left( \overrightarrow{x}\right) },
\end{equation}
and
\begin{equation}
\label{b5}B_i^a\left( \overrightarrow{x}\right) =\epsilon _{ijk}\left\{
\nabla _jA_k^a\left( \overrightarrow{x}\right) +\frac 12f^{abc}A_i^a\left(
\overrightarrow{x}\right) A_j^b\left( \overrightarrow{y}\right) \right\} ,
\end{equation}
$\epsilon _{ijk}$ is the usual anti-symmetric tensor and $f^{abc}$ are the
structure constants of the gauge group, for ex. $SU\left( N\right) $. After
having experienced how the apparatus works on nonabelian gauge theories, we
define a ``Vacuum Trial State'' for gravity, and for this purpose we recall
the orthogonal decomposition $\left( \ref{a10}\right) $ to look at the
essential structure of the inner product between three-geometries%
$$
\left\langle h,h\right\rangle :=\int_{{\cal M}}^{}\sqrt{g}G_\alpha
^{ijkl}h_{ij}^{}\left( x\right) h_{kl}^{}\left( x\right) d_{}^3x=
$$
\begin{equation}
\label{b6}\int_{{\cal M}}^{}\sqrt{g}\left[ \left( \frac 13-\alpha \right)
h^2+\left( L\xi \right) ^{ij}\left( L\xi \right) _{ij}^{}+h_{}^{ij\bot
}h_{ij}^{\bot }\right]
\end{equation}
Previous formula leads us towards the definition of the ``{\it possible}''
trial wave functional for the gravitational ground state
\begin{equation}
\label{b7}\Psi _\alpha \left[ h_{ij}^{}\left( \overrightarrow{x}\right)
\right] ={\cal N}\exp \left\{ -\frac 1{4l_p^2}\left[ \left\langle
hK_{}^{-1}h\right\rangle _{x,y}^{\bot }+\left\langle \left( L\xi \right)
K_{}^{-1}\left( L\xi \right) \right\rangle _{x,y}^{\Vert }+\left\langle
hK_{}^{-1}h\right\rangle _{x,y}^{Trace}\right] \right\} ,
\end{equation}
or in other terms
\begin{equation}
\label{b8}\Psi _\alpha \left[ h_{ij}^{}\left( \overrightarrow{x}\right)
\right] ={\cal N}\Psi _\alpha \left[ h_{ij}^{\bot }\left( \overrightarrow{x}%
\right) \right] \Psi _\alpha \left[ \left( L\xi \right) _{ij}^{}\right] \Psi
_\alpha \left[ \frac 13g_{ij}^{}h\left( \overrightarrow{x}\right) \right] .
\end{equation}
In $\left( \ref{b7}\right) $ and in $\left( \ref{b8}\right) ,$ $h_{ij}^{\bot
}$ is the tracefree-transverse part of the $3D$ quantum field, $\left( L\xi
\right) _{ij}^{}$ is the longitudinal part and finally $h$ is the trace part
of the same field. The dependence of the functional by $\alpha $ will not be
discussed in this paper.
In $\left( \ref{b7}\right) $, $\left\langle \cdot ,\cdot \right\rangle
_{x,y}^{}$ denotes space integration and $K_{}^{-1}$ is the inverse
propagator. The main reason for a similar ``{\it Ansatz}'' comes not only
from $\left( \ref{b6}\right) $ but even to the observation that the momenta
quadratic part of the Hamiltonian decouples in the same way. Even if we had
to give up to $\left( \ref{b7}\right) $ from the beginning, making a more
general ``{\it Ansatz}'' about the vacuum wave functional (and for more
general we mean eqn. $\left( \ref{b1}\right) $ ) one would discover that the
kinetic part decouples in these three terms. For completeness, we give the
analogous expectation values for TT tensors. The other components satisfy
the same rules
\begin{equation}
\label{b9}
\begin{array}{c}
\left\langle \Psi |g_{ij}^{\bot }\left(
\overrightarrow{x}\right) |\Psi \right\rangle =\bar g_{ij}^{\bot }\left(
\overrightarrow{x}\right) , \\ \\
\left\langle \Psi |g_{ij}^{\bot }\left(
\overrightarrow{x}\right) g_{kl}^{\bot }\left( \overrightarrow{y}\right)
|\Psi \right\rangle =\bar g_{ij}^{\bot }\left( \overrightarrow{x}\right)
\bar g_{kl}^{\bot }\left( \overrightarrow{y}\right) +K_{ijkl}^{\bot }\left(
\overrightarrow{x},\overrightarrow{y}\right) , \\ \\
\left\langle \Psi |\pi _{ij}^{\bot }\left(
\overrightarrow{x}\right) |\Psi \right\rangle =0, \\ \\
\left\langle \Psi |\pi _{ij}^{\bot }\left(
\overrightarrow{x}\right) \pi _{kl}^{\bot }\left( \overrightarrow{y}\right)
|\Psi \right\rangle =\frac 14K_{ijkl}^{-1}\left( \overrightarrow{x},%
\overrightarrow{y}\right) , \\
\end{array}
\end{equation}
where $\pi _{ij}^{\bot }=-i\frac \delta {\delta h_{ij^{}}^{\bot }\left(
x\right) }$ is the representation for the TT momentum.
\section{Energy Density Calculation in Schr\"odinger Representation}
\label{p4}
To calculate the energy density associated to the trial functional, we need
to know the action of some basic operators on $\Psi \left[ h_{ij}^{}\left(
\overrightarrow{x}\right) \right] $. The action of the operator $h_{ij}^{}$
on $|\Psi \rangle =\Psi \left[ h_{ij}^{}\left( \overrightarrow{x}\right)
\right] $ is realized by
\begin{equation}
\label{c1}h_{ij}^{}\left( x\right) |\Psi \rangle =h_{ij}^{}\left( x\right)
\Psi \left[ h_{ij}^{}\left( \overrightarrow{x}\right) \right] .
\end{equation}
The action of the operator $\pi _{ij}^{}$ on $|\Psi \rangle $, in general, is
\begin{equation}
\label{c2}\pi _{ij}^{}\left( x\right) |\Psi \rangle =-i\frac \delta {\delta
h_{ij^{}}^{}\left( x\right) }\Psi \left[ h_{ij}^{}\left( \overrightarrow{x}%
\right) \right] .
\end{equation}
The inner product is defined by the functional integration:
\begin{equation}
\label{c3}\left\langle \Psi _1\mid \Psi _2\right\rangle =\int \left[ {\cal D}%
h_{ij}^{}\left( x\right) \right] \Psi _1^{*}\left\{ h_{ij}^{}\right\} \Psi
_2\left\{ h_{kl}^{}\right\} ,
\end{equation}
and the energy eigenstates satisfy the Schr\"odinger equation:
\begin{equation}
\label{c4}\int d_{}^3x{\cal H}\left\{ -i\frac \delta {\delta
h_{ij^{}}^{}\left( x\right) },h_{ij}\left( x\right) \right\} \Psi \left\{
h_{ij}^{}\left( \overrightarrow{x}\right) \right\} =E\Psi \left\{
h_{ij}^{}\left( \overrightarrow{x}\right) \right\} ,
\end{equation}
where ${\cal H}\left\{ -i\frac \delta {\delta h_{ij^{}}^{}\left( x\right) }%
,h_{ij}\left( x\right) \right\} $ is the Hamiltonian density. Instead of
solving $\left( \ref{c4}\right) $, which is of course impossible, we can
formulate the same problem by means of a variational principle. We demand
that
\begin{equation}
\label{c6}\frac{\left\langle \Psi _{}^{}\mid H\mid \Psi _{}\right\rangle }{%
\left\langle \Psi _{}^{}\mid \Psi _{}\right\rangle }=\frac{\int \left[ {\cal %
D}g_{ij}^{\bot }\left( x\right) \right] \int d_{}^3x\Psi _1^{*}\left\{
g_{ij}^{\bot }\right\} {\cal H}\Psi \left\{ g_{kl}^{\bot }\right\} }{\int
\left[ {\cal D}g_{ij}^{\bot }\left( x\right) \right] \mid \Psi \left\{
g_{ij}^{\bot }\right\} \mid _{}^2}
\end{equation}
be stationary against arbitrary variations of $\Psi \left\{ h_{ij}^{}\left(
\overrightarrow{x}\right) \right\} $. The form of $\left\langle \Psi
_{}^{}\mid H\mid \Psi _{}\right\rangle $ can be computed as follows. We
define normalized mean values by a straightforward modification of $\left(
\ref{b9}\right) $, i.e.
\begin{equation}
\label{c7}\bar g_{ij}^{\bot }\left( x\right) =\frac{\int \left[ {\cal D}%
g_{ij}^{\bot }\left( x\right) \right] \int d_{}^3xg_{ij}^{\bot }\left(
x\right) \mid \Psi \left\{ g_{ij}^{\bot }\left( x\right) \right\} \mid _{}^2%
}{\int \left[ {\cal D}g_{ij}^{\bot }\left( x\right) \right] \mid \Psi
\left\{ g_{ij}^{\bot }\right\} \mid _{}^2},
\end{equation}
\begin{equation}
\label{c8}\bar g_{ij}^{\bot }\left( x\right) \text{ }\bar g_{kl}^{\bot
}\left( x\right) +K_{ijkl^{}}^{\bot }\left( \overrightarrow{x},%
\overrightarrow{y}\right) =\frac{\int \left[ {\cal D}g_{ij}^{\bot }\left(
x\right) \right] \int d_{}^3xg_{ij}^{\bot }\left( x\right) g_{kl}^{\bot
}\left( y\right) \mid \Psi \left\{ g_{ij}^{\bot }\left( x\right) \right\}
\mid _{}^2}{\int \left[ {\cal D}g_{ij}^{\bot }\left( x\right) \right] \mid
\Psi \left\{ g_{ij}^{\bot }\right\} \mid _{}^2}.
\end{equation}
It follows that%
$$
\int \left[ {\cal D}h_{ij}^{\bot }\left( x\right) \right] \left(
g_{ij}^{\bot }\left( x\right) -\bar g_{ij}^{\bot }\left( x\right) \right)
\mid \Psi \left\{ g_{ij}^{\bot }\left( x\right) \right\} \mid _{}^2=0
$$
by translation invariance of the measure%
$$
\int \left[ {\cal D}h_{ij}^{\bot }\left( x\right) \right] h_{ij}^{\bot
}\left( x\right) \mid \Psi \left\{ g_{ij}^{\bot }\left( x\right) +\bar g%
_{ij}^{\bot }\left( x\right) \right\} \mid _{}^2=0
$$
\begin{equation}
\label{c9}\Longrightarrow \int \left[ {\cal D}h_{ij}^{\bot }\left( x\right)
\right] h_{ij}^{\bot }\left( x\right) \mid \Psi \left\{ h_{ij}^{\bot }\left(
x\right) \right\} \mid _{}^2=0,
\end{equation}
and $\left( \ref{c8}\right) $ becomes
\begin{equation}
\label{c10}
\begin{array}{c}
\int \left[
{\cal D}h_{ij}^{\bot }\left( x\right) \right] \int d_{}^3xh_{ij}^{\bot
}\left( x\right) h_{kl}^{\bot }\left( y\right) \mid \Psi \left\{
h_{ij}^{\bot }\left( x\right) \right\} \mid _{}^2= \\ \\
K_{ijkl^{}}^{\bot }\left( \overrightarrow{x},\overrightarrow{y}\right) \int
\left[ {\cal D}h_{ij}^{\bot }\left( x\right) \right] \mid \Psi \left\{
h_{ij}^{\bot }\right\} \mid _{}^2.
\end{array}
\end{equation}
Rather than applying the variational principle arbitrarily, the gaussian
{\it Ansatz} is made, according to which in the beginning of this calculus
one has to replace previous general formulas with
\begin{equation}
\label{c11}\Psi _\alpha \left[ h_{ij}^{}\left( \overrightarrow{x}\right)
\right] ={\cal N}\exp \left\{ -\frac 1{4l_p^2}\left\langle \left( g-%
\overline{g}\right) K_{}^{-1}\left( g-\overline{g}\right) \right\rangle
_{x,y}^{\bot }+\ldots \ldots \right\} .
\end{equation}
With this choice and with formulas $\left( \ref{c9},\ref{c10}\right) $, the
one loop-like Hamiltonian can be written as
\begin{equation}
\label{c12}H_{}^{\bot }=\frac 1{4l_p^2}\int_{{\cal M}}^{}d_{}^3x\sqrt{g}%
G_\alpha ^{ijkl}\left[ K_{}^{-1\bot }\left( x,x\right) _{ijkl}+\left(
\triangle _2^{}\right) _j^aK_{}^{\bot }\left( x,x\right) _{iakl}\right]
\end{equation}
where the first term in square brackets comes from the kinetic part and the
second comes from the expansion of the $^3R$ up to second order in such a
way to obtain a quantum harmonic oscillator equation type. the Green
function $K_{}^{\bot }\left( x,x\right) _{iakl}$ can be represented as
\begin{equation}
\label{ffi}K_{}^{\bot }\left( x,x\right) _{iakl}:=\sum_N\frac{h_{ia}^{\bot
}\left( x\right) h_{kl}^{\bot }\left( y\right) }{2\lambda _N\left( p\right) }%
,
\end{equation}
where $h_{ia}^{\bot }\left( x\right) $ are the eigenfunctions of $\triangle
_{2j}^a$ and $\lambda _N\left( p\right) $ are infinite variational
parameters. In formula $\left( \ref{c12}\right) $ we have written the Spin-2
contribution to the energy density alone; expressions like $\left( \ref{c12}%
\right) $ exist for Spin-1 and Spin-0 terms of ${\cal H}$.
\section{The Spectrum of the Spin-2 Operator and the evaluation of the
Energy Density}
\label{p5}
The Spin-2 operator is defined by:
\begin{equation}
\label{ff1}\triangle _2^{}:=-\triangle +2Ric
\end{equation}
$$
\text{or in components,}
$$
\begin{equation}
\label{ff2}\left( \triangle _2^{}\right) _j^a:=-\triangle \delta
_j^{a_{}^{}}+2R_j^a
\end{equation}
where $\triangle $ is the curved Laplacian (Laplace-Beltrami operator) on a
Schwarzschild background and
$R_{j\text{ }}^a$ is the mixed Ricci tensor whose components are:
\begin{equation}
\label{ff3}R_j^a=diag\left\{ \frac{-2m}{r_{}^3},\frac m{r_{}^3},\frac m{%
r_{}^3}\right\} ,
\end{equation}
where $2m=2MG$. This operator is similar to the Lichnerowicz operator
provided that we substitute the Riemann tensor with the Ricci tensor. In $%
\left( \ref{ff1}\right) $ or $\left( \ref{ff2}\right) $ Ricci tensor acts as
a potential on the space of TT tensors; for this reason we are led to study
the following eigenvalue equation
\begin{equation}
\label{ff4}\left( -\triangle \delta _j^{a_{}^{}}+2R_j^a\right)
h_a^i=E^2h_j^{i_{}^{}}
\end{equation}
where $E^2$ is the eigenvalue of the corresponding equation. In doing so, we
follow Regge and Wheeler in analyzing the equation into modes of definite
frequency, angular momentum and parity. In this paper we are interested to
positive $E^2$ and low lying states with $L=M=0$, where $L$ is the quantum
number corresponding to the square of angular momentum and $M$ is the
quantum number corresponding to the projection of the angular momentum on
the z-axis. For $L=0$, Regge-Wheeler decomposition \cite{Regge} shows that
there are no odd-parity perturbations at all, therefore:
\begin{equation}
\label{ff5}h_{ij}^{even}=diag\left[ H\left( r\right) \left( 1-\frac{2m}r%
\right) _{}^{-1},r^2K\left( r\right) ,r_{}^2\sin _{}^2\vartheta K\left(
r\right) \right] Y_{00}^{}\left( \vartheta ,\phi \right) .
\end{equation}
The representation $\left( \ref{ff5}\right) $ is very useful, because of the
decoupling of the components, in fact
\begin{equation}
\label{ff6}
\begin{array}{c}
-\triangle H\left( r\right) -
\frac{4m}{r_{}^3}H\left( r\right) =E^2H\left( r\right) \\ \\
-\triangle K\left( r\right) +
\frac{2m}{r_{}^3}K\left( r\right) =E^2K\left( r\right) \\ \\
-\triangle K\left( r\right) +\frac{2m}{r_{}^3}K\left( r\right) =E^2K\left(
r\right)
\end{array}
\end{equation}
The Laplacian in this particular geometry can be written as
\begin{equation}
\label{ff7}\triangle =\left( 1-\frac{2m}r\right) \frac{d_{}^2}{dr_{}^2}%
+\left( \frac{2r-3m}{r_{}^2}\right) \frac d{dr}.
\end{equation}
Defining reduced fields, such as:
\begin{equation}
\label{ff8}H\left( r\right) =\frac{h\left( r\right) }r;K\left( r\right) =%
\frac{k\left( r\right) }r,
\end{equation}
and changing variables to
\begin{equation}
\label{ff9}x=2m\left\{ \sqrt{\frac r{2m}}\sqrt{\frac r{2m}-1}+\ln \left(
\sqrt{\frac r{2m}}+\sqrt{\frac r{2m}-1}\right) \right\} ,
\end{equation}
the system $\left( \ref{ff6}\right) $ becomes
\begin{equation}
\label{ff10}
\begin{array}{c}
-
\frac{d_{}^2}{dx_{}^2}h\left( x\right) -V\left( x\right) h\left( x\right)
=E^2h\left( x\right) \\ \\
-
\frac{d_{}^2}{dx_{}^2}k\left( x\right) +V\left( x\right) k\left( x\right)
=E^2k\left( x\right) \\ \\
-\frac{d_{}^2}{dx_{}^2}k\left( x\right) +V\left( x\right) k\left( x\right)
=E^2k\left( x\right)
\end{array}
\end{equation}
where
\begin{equation}
\label{ff11}V\left( x\right) =\frac{3m}{r_{}^3}
\end{equation}
We note that the new variable is such that
\begin{equation}
\label{ff12}
\begin{array}{c}
x\simeq r
\text{ }r\longrightarrow \infty \text{ }V\left( x\right) \longrightarrow 0
\\ \\
x\simeq 0\text{ }r\longrightarrow r_0\text{ }V\left( x\right)
\longrightarrow \frac{3m}{\left( r_0\right) _{}^3}=const,
\end{array}
\end{equation}
where $r_0^{}$ is the wormhole radius, satisfying the condition $r_0^{}>2m$,
strictly. The solution of $\left( \ref{ff10}\right) $, in both cases (flat
and curved one) is a Bessel function and precisely the spherical Bessel
function of the first kind for the $L=0$ value of the angular momentum
\begin{equation}
\label{ff13}j_0^{}\left( pr\right) =p\sqrt{\frac 2\pi }\frac{\sin \left(
pr\right) }{pr}=\sqrt{\frac 2\pi }\frac{\sin \left( pr\right) }r
\end{equation}
The corresponding Green function for this problem will be
\begin{equation}
\label{ff14}K\left( x,y\right) =\frac{j_0^{}\left( px\right) j_0^{}\left(
py\right) }{2\lambda }\cdot \frac 1{4\pi }
\end{equation}
Substituting $\left( \ref{ff14}\right) $ in $\left( \ref{c12}\right) $ one
gets (after normalization in spin space and after a rescaling of the fields
in such a way to absorb $l_p^2$)
\begin{equation}
\label{ff15}E\left( m,\lambda \right) =\frac V{2\pi ^2}\sum_{i=1}^2\int_0^%
\infty dpp_{}^2\left[ \lambda _i^{}\left( p\right) +\frac{E_i^2\left(
p,m\right) }{\lambda _i^{}\left( p\right) }\right]
\end{equation}
where
\begin{equation}
\label{ff16}E_{1,2}^2\left( p,m\right) =p_{}^2\mp \frac{3m}{r_0^3},
\end{equation}
$\lambda _i^{}\left( p\right) $ are variational parameters corresponding to
the eigenvalues for a (graviton) spin-2 particle in an external field and $V$
is the volume of the system.
By minimizing $\left( \ref{ff15}\right) $ with respect to $\lambda
_i^{}\left( p\right) $ one obtains $\overline{\lambda }_i^{}\left( p\right)
=\left[ E_i^2\left( p,m\right) \right] ^{\frac 12}$ and
\begin{equation}
\label{ff17}E\left( m,\overline{\lambda }\right) =\frac V{2\pi ^2}%
\sum_{i=1}^2\int_0^\infty dp2\sqrt{E_i^2\left( p,m\right) }\text{ with }%
p_{}^2>\frac{3m}{r_0^3}
\end{equation}
The total energy in the presence of the background is
\begin{equation}
\label{ff18}E\left( m\right) =\frac V{2\pi ^2}\frac 12\int_0^\infty
dpp_{}^2\left( \sqrt{p_{}^2-c_{}^2}+\sqrt{p_{}^2+c_{}^2}\right) \text{ where
}c_{}^2=\frac{3m}{r_0^3}
\end{equation}
For flat space the calculation is essentially the same with the exception of
$c_{}^2=0$. Therefore the equivalent of $\left( \ref{ff18}\right) $ in flat
space is
\begin{equation}
\label{ff19}E\left( 0\right) =\frac V{2\pi ^2}\frac 12\int_0^\infty
dpp_{}^2\left( 2\sqrt{p^2}\right)
\end{equation}
Now, we are in position to perform the energy difference between $\left( \ref
{ff18}\right) $and $\left( \ref{ff19}\right) $. $\Delta E\left( m\right) $
up to second order in perturbations is
\begin{equation}
\label{ff20}\Delta E\left( m\right) =\frac V{2\pi ^2}\frac 12\int_0^\infty
dpp_{}^2\left[ \sqrt{p_{}^2-c_{}^2}+\sqrt{p_{}^2+c_{}^2}-2\sqrt{p^2}\right]
\end{equation}
We want to evaluate the $UV$ behaviour of $\left( \ref{ff20}\right) $,
therefore%
$$
\begin{array}{c}
\Delta E\left( m\right) =
\frac V{2\pi ^2}\frac 12\int_0^\infty dpp_{}^3\left[ \sqrt{1-\left( \frac cp%
\right) _{}^2}+\sqrt{1+\left( \frac cp\right) _{}^2}-2\right] \text{ } \\
\\
\text{becomes for }p_{}^2>>c_{}^2 \\ \\
\sim
\frac V{2\pi ^2}\frac 12\int_0^\infty dpp_{}^3\left[ 1-\frac 12\left( \frac c%
p\right) _{}^2-\frac 18\left( \frac cp\right) _{}^4+1+\frac 12\left( \frac cp%
\right) _{}^2-\frac 18\left( \frac cp\right) _{}^4-2\right] \\
\end{array}
$$
\begin{equation}
\label{ff21}=-\frac V{2\pi ^2}\frac{c_{}^4}8\int_0^\infty \frac{dp}p
\end{equation}
Introducing a cut-off one gets for the $UV$ limit\footnote{%
It is known that at one-loop level Gravity is renormalizable only in flat
space. In a dimensional regularization scheme its contribution to the action
is, on shell, proportional to the Euler character of the manifold that is
nonzero for the Schwarzschild instanton. Although in our approach we are
working with sections of the original manifold to deal with these
divergences one must introduce a regulator that indeed appears in the
contribution of the energy density.}
\begin{equation}
\label{ff22}\int_0^\infty \frac{dp}p\sim \int_0^{\frac \Lambda c}\frac{dx}x%
\sim \ln \left( \frac \Lambda c\right)
\end{equation}
and $\Delta E\left( m\right) $ for high momenta can be estimated by the
following expression
\begin{equation}
\label{ff23}\Delta E\left( m\right) \sim -\frac V{2\pi ^2}\frac{c_{}^4}{16}%
\ln \left( \frac{\Lambda _{}^2}{c_{}^2}\right) =-\frac V{2\pi ^2}\left(
\frac{3m}{r_0^3}\right) ^2\frac 1{16}\ln \left( \frac{r_0^3\Lambda _{}^2}{3m}%
\right) .
\end{equation}
At this point we can compute the total energy, namely the classical
contribution plus the quantum correction up to second order. Recalling the
definition of asymptotic energy for an asymptotically flat background, like
the Schwarzschild one
\begin{equation}
\label{ff24}E_{{\cal ADM}}^{}=\lim _{r\rightarrow \infty }\int_{{\cal %
\partial M}}^{}\sqrt{\hat g}\hat g^{ij}\left[ \hat g_{ik,j}-\hat g%
_{ij,k}\right] dS^k,
\end{equation}
where $\hat g_{ij}$ is the metric induced on a spacelike hypersurface ${\cal %
\partial M}$ which has a boundary at infinity like $S^2$, one gets,
\begin{equation}
\label{ff25}M-\frac V{2\pi ^2}\left( \frac{3m}{r_0^3}\right) ^2\frac 1{16}%
\ln \left( \frac{r_0^3\Lambda _{}^2}{3m}\right) =M-\frac V{2\pi ^2}\left(
\frac{3MG}{r_0^3}\right) ^2\frac 1{16}\ln \left( \frac{r_0^3\Lambda _{}^2}{%
3MG}\right)
\end{equation}
One can observe that
\begin{equation}
\label{ff26}\Delta E\left( m\right) \rightarrow \infty \text{ when }%
m\rightarrow 0\text{, for }r_0=2m=2GM
\end{equation}
and
\begin{equation}
\label{ff27}\Delta E\left( m\right) \rightarrow 0\text{ when }m\rightarrow 0%
\text{, for }r_0\neq 2m=2GM.
\end{equation}
{\bf Remark } We would like to explain the reasons that support the results
of formula $\left( \ref{ff23}\right) $. In that formula we introduced a
particular value of the radius, which behaves as a regulator with respect to
the horizon approach of the potential. The meaning of this particular value
is related to the necessity of explaining the dynamical origin of black hole
entropy by the entanglement entropy mechanism and by the so-called ``{\it %
brick wall model}'' \cite{t Hooft}. Indeed, the same mechanism is present
when one has to regularize entropy by imposing a kind of cut-off, that in
coordinate space means $r_0^{}>2m.$
\section{Summary and Conclusions}
\label{p6}
The trial wave functional approach, by means of Gaussian configurations, led
to possible calculations of quantum fluctuations of the gravitational field
around some fixed background geometry. In particular we have studied a
spherically symmetric background and by means of Birkhoff 's theorem we can
claim that our background is of the Schwarzschild type. Since we have
performed this analysis without any matter contribution and recalling the
definition of Ref. \cite{MTW} our result is valid for a Schwarzschild
wormhole. However this calculation apparatus is entirely based on the
possibility of explicitly breaking the invariance under reparametrisations,
expressed by the gauge fixing $\left( \ref{a4}\right) $, leading to the
conclusion that the final result seems depending on the foliation we choose
to work. For this reason to restore the invariance under reparametrisations,
we need to sum on every lapse function\footnote{%
A detailed version of this procedure will be studied in a future paper \cite
{Remo}}. With the gauge choice $\left( \ref{a4}\right) $, the problem of
defining a correct vacuum energy on every slice is well posed and the result
shows us an intrinsic energy depending only on the dynamics generated by
3-surfaces.
\section{Acknowledgments}
I wish to thank G. Esposito, V.Frolov, E. Gozzi, R. Parentani, D.L.
Rapoport, E. Recami for helpful discussion and P. Saurgnani who gave me the
technical support for the realization of this work. I also thank S. Liberati
and B. Jensen who suggested me how to justify the horizon approach.
|
\section{Introduction}
\indent
The last two years have seen the first observations of the
electromagnetic
penguins in $B$ decays by the CLEO collaboration.
These include the measurements of the exclusive decay rate,
$\BR (B \ra K^\star + \gamma) = (4.5 \pm 1.0\pm 0.9) \times 10^{-5}$ \cite{CLEOrare1}, and the
inclusive rate $\BR (B \ra X _{s} + \gamma) =(2.32 \pm 0.67) \times 10^{-4}$ \cite{CLEOrare2},
yielding $R(K^*/X_s) \equiv \Gamma (B \ra K^\star + \gamma )/\Gamma (B \ra X _{s} + \gamma) =0.19 \pm 0.09$.
In addition, the charged and neutral
$B$-meson decay rates are found equal within experimental measurements.
The inclusive decay rate is
in agreement with the predictions of
the standard model \cite{ag1,Ciuchini,Buras94}, with the rate
estimated as $\BR (B \ra X _{s} + \gamma) = (2.55 \pm 1.28) \times 10^{-4}$
assuming $\vert V_{ts}\vert/\vert V_{cb}\vert=1.0$ \cite{ag95}. Conversely, one can vary
the Cabibbo-Kobayashi-Maskawa
(CKM) matrix element ratio and determine it from $\BR (B \ra X _{s} + \gamma)$,
which yields $\vert V_{ts}\vert/\vert V_{cb}\vert=1.10 \pm 0.43$.
The ratio $R(K^*/X_s)$
is well explained by the QCD-sum-rule
based estimates of the recent vintage \cite{abs93,bksnsr} and by
wave-function models combined with vector meson dominance
(local parton-hadron duality) \cite{ag1}.
This and the near equality of the charged and neutral decay rates imply
that the observed radiative $B$ decays are dominated by the (common)
electromagnetic penguin (short distance) amplitudes and the
contributions from $B$-meson-specific diagrams (weak annihilation,
$W^\pm$-exchange) are small.
\indent
The photon energy spectrum in $B \ra X _{s} + \gamma$
yields information on the structure function of the photon in the
electromagnetic penguins \cite{neubertbsg,Bigietal,KS94,klp95}.
In specific models \cite{Alipiet,shifmangamma}, this information can be
transcribed in terms of non-perturbative
parameters, such as the $b$-quark mass and the kinetic energy of the
$b$ quark in $B$ hadron. These quantities can also be estimated in the
framework of the heavy quark effective theory
\cite{HQETpower} combined with QCD
sum rules \cite{bqmass,BB94}.
The results of a recent analysis of the CLEO data on $B \ra X _{s} + \gamma$ \cite{ag95}
are consistent with the values expected from such theoretical
considerations.
However, this agreement is presently not completely quantitative due to
imprecise data.
There is considerable interest in measuring the
CKM-suppressed radiative $B$ decays, such as $B \ra X _{d} + \gamma$ \cite{ag2}
and $B \rightarrow (\rho,\omega) + \gamma$ \cite{abs93}. A
determination of the CKM parameters from eventual measurements of these
decays requires careful treatment of the competing short-distance (SD) and
long-distance (LD) effects.
This problem can be formulated in terms of
model-independent correlation functions involving matrix elements of a few
dimension-6 operators in an effective theory.
Techniques, such as the QCD sum rules, can
then be invoked to estimate them. In
\cite{ab95,wyler95}, the leading
LD-effects in the exclusive decays $B \rightarrow (\rho, \omega) + \gamma$ are
calculated in terms of the weak annihilation amplitudes.
The largest such effects may show themselves in the charged $B^\pm$-decays,
$B^\pm \rightarrow \rho^\pm + \gamma$, contributing up to $O(15 \%)$ of the
corresponding SD-amplitudes; their influence in
the neutral $B$-decays is estimated to be much smaller. Hence,
there are good theoretical reasons to plead that the decays
$B^0 \rightarrow (\rho^0,\omega) + \gamma$ and
$B \rightarrow X_d + \gamma$ are well suited to determine the CKM parameters.
We take up these issues in this status report.
\section{Estimates of $\BR (B \ra X _{s} + \gamma)$ in the SM}
\label{sec:effham}
The framework that is used generally to discuss the decays $B \ra X _{s} + \gamma$
is that of an effective theory with
five quarks, obtained by integrating out the
heavier degrees of freedom,
which in the standard model are the top quark and the $W$-boson.
A complete set of dimension-6 operators relevant for the processes
\ $b \rightarrow s+ \gamma$ ~and \ $b \rightarrow s+ \gamma+ g$ ~is contained in the effective Hamiltonian
\begin{equation}
\label{heff}
H_{eff}(b \rightarrow s \gamma)
= - \frac{4 G_{F}}{\sqrt{2}} \, \lambda_{t} \, \sum_{j=1}^{8}
C_{j}(\mu) \, O_j(\mu) \quad ,
\end{equation}
where
$G_F$ is the Fermi constant
coupling constant,
$C_{j}(\mu) $ are the Wilson coefficients evaluated at the scale $\mu$,
and $\lambda_t=V_{tb}V_{ts}^*$ with $V_{ij}$ being the
CKM matrix elements. The overall multiplicative factor $\lambda_t$ follows
from the CKM unitarity and neglecting $\lambda_u$.
The operators $O_j$ read
\begin{eqnarray}
\label{operators}
O_1 &=& \left( \bar{c}_{L \beta} \gamma^\mu b_{L \alpha} \right) \,
\left( \bar{s}_{L \alpha} \gamma_\mu c_{L \beta} \right)\,, \nonumber \\
O_2 &=& \left( \bar{c}_{L \alpha} \gamma^\mu b_{L \alpha} \right) \,
\left( \bar{s}_{L \beta} \gamma_\mu c_{L \beta} \right) \,,\nonumber \\
O_3 &=& \left( \bar{s}_{L \alpha} \gamma^\mu b_{L \alpha} \right) \, \left[
\left( \bar{u}_{L \beta} \gamma_\mu u_{L \beta} \right) + ... +
\left( \bar{b}_{L \beta} \gamma_\mu b_{L \beta} \right) \right] \,,
\nonumber \\
O_4 &=& \left( \bar{s}_{L \alpha} \gamma^\mu b_{L \beta} \right) \, \left[
\left( \bar{u}_{L \beta} \gamma_\mu u_{L \alpha} \right) + ... +
\left( \bar{b}_{L \beta} \gamma_\mu b_{L \alpha} \right) \right] \,,
\nonumber \\
O_5 &=& \left( \bar{s}_{L \alpha} \gamma^\mu b_{L \alpha} \right) \, \left[
\left( \bar{u}_{R \beta} \gamma_\mu u_{R \beta} \right) + ... +
\left( \bar{b}_{R \beta} \gamma_\mu b_{R \beta} \right) \right] \,,
\nonumber \\
O_6 &=& \left( \bar{s}_{L \alpha} \gamma^\mu b_{L \beta} \right) \, \left[
\left( \bar{u}_{R \beta} \gamma_\mu u_{R \alpha} \right) + ... +
\left( \bar{b}_{R \beta} \gamma_\mu b_{R \alpha} \right) \right] \,,
\nonumber \\
O_7 &=& (e/16\pi^{2}) \, \bar{s}_{\alpha} \, \sigma^{\mu \nu}
\, (m_{b}(\mu) R + m_{s}(\mu) L) \, b_{\alpha} \ F_{\mu \nu} \,,
\nonumber \\
O_8 &=& (g_s/16\pi^{2}) \, \bar{s}_{\alpha} \, \sigma^{\mu \nu}
\, (m_{b}(\mu) R + m_{s}(\mu) L) \, (\lambda^A_{\alpha \beta}/2) \,b_{\beta}
\ G^A_{\mu \nu} \quad ,
\nonumber \\
\end{eqnarray}
where $e$ and $g_s$ are the electromagnetic and the strong
coupling constants, respectively. In the magnetic moment type
operators $O_7$ and $O_8$, $F_{\mu \nu}$ and $G^A_{\mu \nu}$
denote the electromagnetic and the gluonic field strength
tensors, respectively. The subscripts on the quark fields
$L\equiv (1-\gamma_5)/2$ and $R\equiv (1+\gamma_5)/2$
denote the left and right-handed projection operators, respectively.
QCD corrections to the decay rate for $b \rightarrow s \gamma$
bring in large logarithms of the form $\alpha_s^n(m_W) \, \log^m(m_b/M)$,
where $M=m_t$ or $m_W$ and $m \le n$ (with $n=0,1,2,...$).
Using the renormalization group equations the Wilson coefficient
can be calculated at the scale $\mu \approx m_b$ which
is the relevant scale for $B$ decays.
To leading logarithmic precision,
it is sufficient to know the leading order
anomalous dimension matrix and the matching
$C_i(\mu=m_W)$ to lowest order (i.e., without
QCD corrections) \cite{InamiLim}.
The $8 \times 8$ anomalous dimension matrix is given in
\cite{Ciuchini}, from where references to earlier calculations can also
be obtained, the
Wilson coefficients
are explicitly listed in
\cite{Buras94} and the
numerical values of these
coefficients being used here can be seen in \cite{ag95}.
\indent
It has become customary to calculate the
branching ratio for the radiative decay $B \ra X _{s} + \gamma$ in terms of the
semileptonic decay branching ratio ${\cal B} (B \rightarrow X\ell \nu_\ell)$
\begin{equation}
\label{brdef}
{\cal B} ( B \rightarrow X_{s} \gamma) = [\frac{\Gamma(B \rightarrow
X_{s} + \gamma)}{\Gamma_{sl}}]
\, R(m_b,\mu) \,{\cal B} (B \rightarrow X\ell \nu_\ell) \qquad ,
\end{equation}
where, in the approximation of including the leading-order QCD correction,
$\Gamma_{sl}$ is given by the expression
\begin{equation}
\label{widthsl}
\Gamma_{sl} = \frac{G_F^2 \, |V_{cb}|^2}{192 \pi^3} \, m_b^5 \,
g(m_c/m_b) \, ( 1-2/3 \frac{\alpha_s}{\pi} f(m_c/m_b))\quad .
\end{equation}
The phase space function $g(z)$
and the function $f(z)$ due to one-loop QCD corrections can be seen in
\cite{Alipiet}.
The radiative decay rate $\Gamma(B \rightarrow X_s + \gamma)$ are worked out
in \cite{ag1,ag95},
taking into account $O(\alpha_s)$ virtual and bremsstrahlung
corrections. In
calculating the matrix elements in these papers, the on-shell
subtraction prescription for the quark masses has been used.
Due to the explicit
factors of the running quark masses in the operators $O_7$ and
$O_8$, the $m_b^5$-factor contained in the decay rate $\Gamma(B \rightarrow X_s +
\gamma)$
is replaced by the following product
\begin{equation}
m_b^5 \longrightarrow m_b(pole)^3 ~m_b(\mu)^2 \quad ,
\end{equation}
where $m_b(pole)$ and $m_b(\mu)$ denote the pole mass and the $\overline
{\mbox{MS}}$-running
mass of the $b$ quark, respectively. Since, in the leading order
in $\alpha_s$, the semileptonic decay width $\Gamma_{sl}$ depends on the
product $m_b(pole)^5$, the ratio of the two decay widths brings in
the correction factor $R(m_b,\mu)$:
\begin{equation}
\label{rfactor}
R(m_b,\mu)=[m_b(\mu)/m_b(pole)]^2 \quad ,
\end{equation}
as also remarked in \cite{Buras94}. At the one-loop level, these masses are
related:
\begin{equation}
\frac{m_b(\mu)}{m_b(pole)}=[\frac{\alpha_s(\mu)}{\alpha_s(m_b)}]^{4/\beta_0}
[1-\frac{4}{3} \frac{\alpha_s(\mu)}{\pi}],
\label{mbrun}
\end{equation}
where $\beta_0=23/3$.
The parameters used in estimating
the inclusive rates for $\BR (B \ra X _{s} + \gamma)$ are
summarized in table \ref{tabparam}.
\begin{table}
\begin{center}
\begin{tabular}{| c | c | }
\hline
Parameter & Range\\
\hline \hline
$\overline{m_t}$ (GeV) & $170 \pm 11$\\
$\mu$ (GeV) & $5.0^{+5.0}_{-2.5}$ \\
$\Lambda_5$ (GeV) & $0.195 ^{+0.065}_{-0.05}$\\
${\cal B}(B \rightarrow X \ell \nu_\ell)$ & $(10.4 \pm 0.4)\%$\\
$m_c/m_b$ & $0.29 \pm 0.02$\\
$m_W$ (GeV) & $80.33$ \\
$\alpha_{\footnotesize{\mbox{QED}}}^{-1}$ & $ 130.0$\\
\hline
\end{tabular}
\end{center}
\caption{Values of the parameters used in estimating the branching
ratio $\BR (B \ra X _{s} + \gamma)$ in the standard model.}
\label{tabparam}
\end{table}
We now discuss
$\BR (B \ra X _{s} + \gamma)$ in the standard model and theoretical uncertainties on this
quantity \cite{ag95}.
\begin{itemize}
\item Scale dependence of the Wilson coefficients.
\end{itemize}
The largest theoretical
uncertainty stems from the scale dependence of the Wilson coefficients.
As derived explicitly in \cite{ag95},
the decay rate for $B \ra X _{s} + \gamma$
depends on seven of the eight Wilson coefficients in $H_{eff}(b \rightarrow s)$,
once one takes into account the bremsstrahlung corrections and is not
factored in terms of a single (effective) coefficient, namely $C_7^{eff}$,
that one encounters
for the two-body decays $b \rightarrow s + \gamma$ \cite{Ciuchini,Buras94}.
Numerical values of the two dominant effective coefficients, $C_7^{eff}$
and $C_8^{eff}$, as one varies $\mu$, the QCD scale $\Lambda_5$, and
the (running) top quark mass in the $\overline{\mbox{MS}}$-scheme
$\bar{m}_t(m_t)$
in the range given in table 1, are:
\begin{eqnarray}
C_7^{eff} &\equiv & C_7 -\frac{C_5}{3} - C_6 = -0.306 \pm 0.050,
\nonumber\\
C_8^{eff} &\equiv & C_8 + C_5 = -0.146 \pm 0.020 .
\label{c78eff}
\end{eqnarray}
This is the dominant theoretical error on
$\BR (B \ra X _{s} + \gamma)$, contributing about $\pm 35 \%$.
\begin{itemize}
\item
Scale-dependence of $m_b(\mu)$
in the operators $O_7$ and $O_8$.
\end{itemize}
This brings into fore the extra (scale-dependent) multiplicative factor
$R(m_b,\mu)$ for the branching ratio $\BR (B \ra X _{s} + \gamma)$, as
discussed above. Intrinsic uncertainties in the concept of the pole mass
due to infrared renormalons suggest that one should express all physical
results in terms of the running masses \cite{renormalons}.
This requires recalculating the
decay rate $\BR (B \ra X _{s} + \gamma)$ with the running masses, incorporating resummations
of the kind recently undertaken for the semileptonic $B$ decay rates
\cite{bslresum}. \begin{itemize}
\item Extrinsic errors in $\BR (B \ra X _{s} + \gamma)$
\end{itemize}
The next largest
error arises from the parameters which are extrinsic to the decay
$B \ra X _{s} + \gamma$ and have crept in due to
normalizing the
branching ratio $\BR (B \ra X _{s} + \gamma)$ in terms of ${\cal B} (B \rightarrow X \ell \nu_\ell)$.
The first of these extrinsic errors is related to the uncertainty
in the ratio $m_c/m_b$.
Using for the $b$ quark pole mass $m_b(pole)=4.8 \pm 0.15$ GeV
\cite{bqmass} and $m_b-m_c = 3. 40$ GeV \cite{HQET2}, one gets
$m_c/m_b=0.29 \pm 0.02$.
Taking into account the
experimental error of $\pm 4.1 \%$ on ${\cal B}(B \rightarrow X \ell \nu_\ell)$
\cite{Gibbons}, one estimates an extrinsic error of
$\pm 12 \%$ on $\BR (B \ra X _{s} + \gamma)$.
\indent
Assuming $|V_{ts}|/|V_{cb}=1$ \cite{PDG},
the branching
ratio ${\cal B}(B \rightarrow X_s \gamma)$ calculated
as a function of the top quark mass is shown in Fig. 1 \cite{ag95}.
\begin{figure}[htb]
\vspace{0.10in}
\centerline{
\epsfysize=3in
\rotate[r]{
\epsffile{aglettfig1.ps}
}
}
\vspace{0.08in}
\caption[]{
${\cal B}(B \rightarrow X_s \gamma)$ as a function of the $\overline{\mbox{MS}}$ top
quark mass. The
three solid lines correspond to the variation of the parameters
$\mu$ and $\Lambda_5$ as described in the text.
The experimental $(\pm 1\sigma)$-bounds
from CLEO \cite{CLEOrare2} are shown by the dashed lines (from
\protect\cite{ag95}). \label{fig:1}}
\end{figure}
For all three solid curves the quark mass ratio is fixed at
$m_c/m_b=0.29$. The top
solid curve is drawn for $\mu=2.5$ GeV and $\Lambda_5 = 0.260$ GeV.
The bottom solid curve is for $\mu=10$ GeV and $\Lambda_5 = 0.145$
GeV,
and the middle solid curve corresponds to the central values of
the input parameters in table 1.
Using $\overline{m}_t=(170 \pm 11)$ GeV, and adding the extrinsic error,
one obtains:
\begin{equation}
\BR (B \ra X _{s} + \gamma)=(2.55 \pm 1.28) \times 10^{-4} \,,
\end{equation}
to be compared with the
CLEO measurement $\BR (B \ra X _{s} + \gamma) = (2.32 \pm 0.67) \times 10^{-4}$.
The $(\pm 1\sigma)$-upper and -lower bound from the CLEO measurement
are shown in Fig. 1 by dashed lines. The agreement between SM
and experiment is good, given the large uncertainties on both.
In \cite{ag95}, the branching ratio $\BR (B \ra X _{s} + \gamma)$ has been calculated as a
function
of the CKM matrix element ratio squared $(|V_{ts}|/|V_{cb}|)^2$, varying
$\bar{m}_t$, $\mu$ and $\Lambda_5$ in the range specified in table 1.
Using the $(\pm 1 \sigma)$-experimental bounds on $\BR (B \ra X _{s} + \gamma)$,
one infers \cite{ag95}:
\begin{equation}
|V_{ts}|/|V_{cb}|=1.10 \pm 0.43 \,,
\end{equation}
which is consistent with the indirect constraints from the CKM
unitarity \cite{PDG} yielding $ |V_{ts}|/|V_{cb}| \simeq 1.0$
but imprecise. Further
improvements require reducing the perturbative scale($\mu$)-dependence of
the decay rate, which in turn implies calculations of the next-to-leading
order terms, and more accurate measurements.
\section{Photon Energy Spectrum in $B \ra X _{s} + \gamma$}
The two-body partonic process $b \rightarrow s \gamma$
yields a photon energy spectrum $1/\Gamma d \Gamma (b \rightarrow s
\gamma) = \delta(1-x)$, where
the scaled photon energy $x$ is defined as
$ E_\gamma = (m_b^2-m_s^2)/(2 \, m_b )\, x $;
$x$ then varies in the interval $[0,1]$.
Perturbative QCD corrections, such as
$b \rightarrow s \gamma + g$, give a characteristic bremsstrahlung spectrum
peaking near the end-points, $E_\gamma \rightarrow
E_\gamma ^{max}$ and $E_\gamma \rightarrow 0$, arising from the soft-gluon and
soft-photon configurations, respectively. As long as the $s$-quark mass
is non-zero, there is no collinear singularity in the spectrum.
Near the end-points, one
has to improve the spectrum obtained in fixed order perturbation
theory. This is usually done by isolating and exponentiating
the leading behaviour in $\alpha_{em}\alpha_s(\mu)^m
\log^n (1-x)$ and $\alpha_{em}\alpha_s(\mu)^m \log^n x$, with $m\leq n$,
where $\mu$ is a typical momentum in the decay $B \ra X _{s} + \gamma$. The running of
$\alpha_s$ is a non-leading effect, but as it is characteristic of QCD it
modifies the Sudakov-improved end-point photon energy spectrum \cite{KS94}
compared to its analogue in QED \cite{Sudakov}.
Away from the end-points, the photon energy
spectrum has to be calculated completely in a given order in $\alpha_s$ in
perturbation theory \cite{ag1,ag95}.
The complete photon energy spectrum in $B \ra X _{s} + \gamma$ is at present
not calculable in QCD from first principles. The situation is very much
analogous to that of other hadronic structure functions.
It has
been observed in a number of papers
\cite{neubertbsg,Bigietal,KS94}, that the x-moments
of the inclusive photon energy spectrum in $B \ra X _{s} + \gamma$ and those of the
lepton energy spectrum in the decay $B \rightarrow X_u \ell \nu_\ell$ are related.
Defining the moments as:
\begin{eqnarray}
{\cal M}_n(B \ra X _{s} + \gamma) &\equiv & \frac{1}{\Gamma} \int_{0}^{M_B/m_b} dx x^{n-1}
\frac{d \Gamma}{dx} \\ \nonumber
{\cal M}_n (B \rightarrow X_u \ell \nu_\ell) &\equiv& - \int_{0}^{M_B/m_b} dx x^n
\frac{d}{dx}\big(\frac{1}{\Gamma_\ell}\frac{d \Gamma_\ell}{dx}\big) \\
\nonumber
&=& \frac{n}{\Gamma_\ell} \int_{0}^{M_B/m_b} dx x^{n-1} \frac{d
\Gamma_\ell}{dx}~, \label{moments}
\end{eqnarray}
the ratios of the moments are free of non-perturbative complications.
The moments ${\cal M}_n$ have been worked out in
the leading non-trivial order in perturbation theory and
the results can be expressed as:
\begin{equation}
{\cal M}_n \sim 1 + \frac{\alpha_s}{2 \pi} C_F(A\log^2n + B \log n + \mbox{const.})
\end{equation}
where $C_F=4/3$,
the leading coefficient is universal with $A=-1$ \cite{Sudakov}, and the
non-leading coefficients are process dependent;
$B=7/2$ \cite{ag1} and $B
=31/6$ \cite{KJ},
for $B \ra X _{s} + \gamma$ and $B \rightarrow X_u \ell \nu_\ell$, respectively. Measurements of
the moments could eventually be used to relate the CKM matrix element
$V_{ts}$ and
$V_{ub}$. That this method will give competitive values for $V_{ub}$,
however, depends on whether or not the coefficient functions
in $\Gamma (B \ra X _{s} + \gamma)$ discussed
in the previous section are known to the desired level of
theoretical accuracy.
We shall leave such theoretically improved
comparisons for future Rencontres de Moriond and
confine ourselves
to the discussion of the present state-of-the-art
comparison of the measured photon energy spectrum in $B \ra X _{s} + \gamma$
with the perturbative QCD-improved treatment of the same.
The analysis that we discuss here \cite{ag1,ag95} treats the
non-perturbative effects
in terms of a $B$-meson wave function. In this model
\cite{Alipiet}, which admittedly is simplistic but not necessarily wrong,
the $b$ quark in $B$ hadron is assumed to have a Gaussian
distributed Fermi motion determined by a non-perturbative parameter, $p_F$,
\begin{equation}
\label{lett13}
\phi(p)= \frac {4}{\sqrt{\pi}{p_F}^3} \exp (\frac {-p^2}{{p_F}^2})
\quad , \quad p = |\vec{p}|
\end{equation}
with the wave function normalization
$ \int_0^\infty \, dp \, p^2 \, \phi(p) = 1.$
The photon energy spectrum from
the decay of the $B$-meson at rest is then given by
\begin{equation}
\label{lett15}
\frac{d\Gamma}{dE_\gamma}= \int_0^{p_{max}} \, dp \, p^2 \, \phi(p)
\frac {d\Gamma_b}{dE_\gamma}(W,p,E_\gamma) \quad ,
\end{equation}
where $p_{max}$ is the maximally allowed value of $p$ and
$ \frac{d\Gamma_b}{dE_\gamma}$
is the photon energy spectrum from the decay of the $b$-quark in
flight, having a momentum-dependent mass $W(p)$.
An analysis of the CLEO photon
energy spectrum has been undertaken in \cite{ag95} to determine
the non-perturbative parameters of this model, namely
$m_b(pole)$ and $p_F$.
The experimental
errors are still large and the fits result in relatively small $\chi^2$
values; the minimum, $\chi^2_{min}=0.038$, is obtained
for $p_F=450$ MeV and
$m_b(pole)=4.77$ GeV, in good agreement with theoretical estimates of
the same, namely
$m_b(pole)= 4.8\pm 0.15$ GeV \cite{bqmass} and
$p_F^2=\mu_\pi^2/2= 0.25 \pm 0.05$ GeV$^2$
obtained from the QCD sum rules \cite{BB94}.
In Fig. 2 we have plotted the photon energy spectrum normalized
to unit area in the interval between 1.95 GeV and 2.95 GeV for
the parameters which correspond to the minimum $\chi^2$ (solid curve)
and for another set of parameters that lies near the
$\chi^2$-boundary defined by $\chi^2=\chi^2_{min} +1$.
(dashed curve). Data from CLEO \cite{CLEOrare2} are also
shown. Further details of this analysis can be seen in \cite{ag95}.
\begin{figure}[htb]
\vspace{0.10in}
\centerline{
\epsfysize=3in
\rotate[r]{
\epsffile{aglettfig4a.ps}
}
}
\vspace{0.08in}
\caption[]{Comparison of the normalized photon energy distribution
using the
CLEO data \protect\cite{CLEOrare2} corrected for detector effects and
theoretical
distributions from \protect\cite{ag95} , both normalized to unit area in
the photon energy interval between 1.95 GeV and 2.95 GeV. The solid
curve corresponds to the values with the minimum $\chi^2$,
$(m_q,p_F)$=(0,450 MeV), and the dashed curve to the values
$(m_q,p_F)$=(300 MeV, 310 MeV).
\label{fig:4}}
\end{figure}
\section{Inclusive radiative decays $B \rightarrow X _{d} + \gamma$ }
\indent
The theoretical interest in the standard model in the
(CKM-suppressed) inclusive radiative decays
$B \rightarrow X _{d} + \gamma$\ lies in the first place in the
possibility of determining the CKM-Wolfenstein
parameters $\rho$ and $\eta$ \cite{Wolfenstein}.
The relevant region in the
decays $B \rightarrow X_d + \gamma$ is the end-point photon energy spectrum,
which has to be measured requiring that
the hadronic system $X_d$ recoiling against the
photon does not contain strange hadrons to suppress the large-$E_\gamma$
photons from the decay $B \ra X _{s} + \gamma$. Assuming that this is feasible,
one can determine
from the ratio of the decay rates
$\BR (B \ra X _{d} + \gamma)/\BR (B \ra X _{s} + \gamma)$ the CKM-Wolfenstein parameters.
This measurement was proposed in \cite{ag2}, where
the final-state spectra were also worked out.
\indent
In close analogy
with the $B \rightarrow X _{s} + \gamma$\ case discussed earlier,
the complete set of dimension-6 operators relevant for
the processes $b \rightarrow d \gamma$ and $b \rightarrow d \gamma g$
can be written as:
\begin{equation}
\label{heffd}
H_{eff}(b \rightarrow d)=
- \frac{4 G_{F}}{\sqrt{2}} \, \xi_{t} \, \sum_{j=1}^{8}
C_{j}(\mu) \, \hat{O}_{j}(\mu),\quad
\end{equation}
where $\xi_{j} = V_{jb} \, V_{jd}^{*}$ for $j=t,c,u$. The operators
$\hat{O}_j, ~j=1,2$, have implicit in them CKM factors. In the
Wolfenstein parametrization \cite{Wolfenstein}, one can express these
factors as :
$\xi_u = A \, \lambda^3 \, (\rho - i \eta),
{}~\xi_c = - A \, \lambda^3 ,
{}~\xi_t=-\xi_u - \xi_c$.
We note that all three CKM-angle-dependent quantities
$\xi_j$ are of the
same order of magnitude, $O(\lambda^3)$, where $\lambda =\sin \theta_C
\simeq 0.22$.
This is an important difference as compared to the effective
Hamiltonian ${\cal H}_{eff}(b \rightarrow s)$ written earlier,
in which case the
effective Hamiltonian
factorizes into an overall CKM factor $\lambda_t$.
For calculational ease, this difference
can be implemented by defining the operators $\hat{O}_1$ and $\hat{O}_2$
entering in $H_{eff}(b \rightarrow d)$ as follows \cite{ag2}:
\begin{eqnarray}
\label{basis}
&&\hat{O}_{1} =
-\frac{\xi_c}{\xi_t}(\bar{c}_{L \beta} \go{\mu} b_{L \alpha})
(\bar{d}_{L \alpha} \gu{\mu} c_{L \beta})
-\frac{\xi_u}{\xi_t}(\bar{u}_{L \beta} \go{\mu} b_{L \alpha})
(\bar{d}_{L \alpha} \gu{\mu} u_{L \beta}) ,\nonumber \\
&& \hat{O}_{2} =
-\frac{\xi_c}{\xi_t}(\bar{c}_{L \alpha} \go{\mu} b_{L \alpha})
(\bar{d}_{L \beta} \gu{\mu} c_{L \beta})
-\frac{\xi_u}{\xi_t}(\bar{u}_{L \alpha} \go{\mu}
b_{L \alpha}) (\bar{d}_{L \beta} \gu{\mu} u_{L \beta}) ,
\end{eqnarray}
and the rest of the operators $(\hat{O}_j;~j=3...8)$ are
defined like their
counterparts ${O}_j$ in $H_{eff}(b \rightarrow s)$, with the obvious replacement
$s \rightarrow d$. With this definition, the matching conditions $C_j(m_W)$
and the solutions
of the RG equations yielding $C_j(\mu)$ become
identical for the two operator basis $O_j$ and $\hat{O}_j$.
It has been explicitly checked in the $O(\alpha_s)$ calculations
of the decay rate and photon energy spectrum involving $b \rightarrow d g$ and
$b \rightarrow d g \gamma$ transitions that the limit $m_u \rightarrow 0$
for the decay rate $\Gamma(B \ra X _{d} + \gamma)$
exists \cite{ag2}. From this it
follows that, in the leading order QCD corrections, there are no
logarithms of the type $\alpha_s \log (m_u^2/m_c^2)$ \cite{Ricciardi}.
Some papers, estimating LD-contributions in radiative $B$ decays, seem
to contradict this by assuming light-quark contributions which have such
spurious log-dependence. There is no calculational basis for this
assumption.
In higher orders, such terms must be absorbed in the
non-perturbative functions.
On the other hand, as far as the dependence of the decay rate and spectra
on the external light quark masses is concerned,
one encounters logarithms of the
type $\alpha_{em}\alpha_s (1+(1-x)^2)/x \log (m_b^2/m_s^2)$
(for $b \rightarrow s g \gamma)$ and
$\alpha_{em}\alpha_s (1+(1-x)^2)/x \log (m_b^2/m_d^2)$ (for $b \rightarrow d g
\gamma$)
near the soft-photon
($x \rightarrow 0)$ region \cite{ag95}, which can, however, be exponentiated
\cite{klp95}.
The essential difference between $\Gamma (B \ra X _{s} + \gamma)$ and $\Gamma(B \ra X _{d} + \gamma)$
lies in the matrix elements of the first two operators $O_1$ and $O_2$
(in $H_{eff}(b \rightarrow s)$) and $\hat{O}_1$ and $\hat{O}_2$ (in $H_{eff}(b
\rightarrow d)$).
The derivation of the inclusive decay rate and the final-state distributions
in $B \rightarrow X _{d} + \gamma$\ otherwise
goes along very similar lines as for the decays $B \rightarrow X _{s} + \gamma$\ .
The branching ratio $\BR (B \ra X _{d} + \gamma)$
in the SM can be written as:
\begin{eqnarray}
\label{branstruc}
&& \BR (B \ra X _{d} + \gamma) = D_1 |\xi_t|^2 \nonumber \\
&& \{
1 - \frac{1-\rho}{(1-\rho)^2 + \eta^2} \, D_2
- \frac{\eta}{(1-\rho)^2 + \eta^2} \, D_3
+ \frac{D_4}{(1-\rho)^2 + \eta^2} \} , \quad
\end{eqnarray}
where the functions $D_i$ depend on the parameters listed in table 1. The
uncertainty on this branching ratio from the parametric dependence is very
similar to the one worked out for $\BR (B \ra X _{s} + \gamma)$.
For the central values of the parameters in table 1, one gets :
$D_1=0.21, ~D_2=0.17, ~D_3=0.03, ~D_4=0.10$.
To get the inclusive branching ratio
the CKM parameters $\rho$ and $\eta$ have to be constrained from the
unitarity fits. Taking the parameters from a recent fit, one gets
$ 5.0 \times 10^{-3} \leq \vert \xi_t \vert \leq 1.4 \times
10^{-2}$
(at 95\% C.L.) \cite{al95},
yielding an order of magnitude uncertainty in $\BR (B \ra X _{d} + \gamma)$ -
hence the interest in measuring it. Taking the central values of the
fit parameters
$A=0.8, \lambda=0.2205, \eta =0.34$ and $\rho=-0.07$ \cite{al95},
one gets $\BR (B \ra X _{d} + \gamma) = (1.7\pm 0.85) \times 10^{-5}$, which is
approximately a factor 10 -20
smaller than the CKM-allowed branching ratio $\BR (B \ra X _{s} + \gamma)$, measured by CLEO
\cite{CLEOrare1}.
\vspace*{3.0ex}
\section{ Estimates of ${\cal B}(B \rightarrow V + \gamma )$
and Constraints on the CKM Parameters $\rho$ and $\eta$}
\indent
Exclusive radiative
$B$ decays $B \rightarrow V + \gamma$, with $V=K^*,\rho,\omega$, are also
potentially
very interesting from the point of view of determining the CKM parameters
\cite{abs93}. The extraction of these parameters would, however, involve a
trustworthy
estimate of the SD- and LD-contributions in the decay amplitudes.
\par
The SD-contribution in the
exclusive decays $(B_u, B_d) \rightarrow (K^*,\rho) + \gamma$,
$B_d \rightarrow \omega + \gamma$ and the
corresponding $B_s$ decays, $B_s \rightarrow (\phi,K^*) + \gamma $,
involve the magnetic moment operator $O_7$ and the related one obtained
by the obvious change $s \rightarrow d$, $\hat{O}_7$.
The transition form factors governing the radiative $B$ decays
$B \rightarrow V + \gamma$ can be generically defined as:
\begin{equation}
\langle V,\lambda |\frac{1}{2} \bar \psi \sigma_{\mu\nu} q^\nu b
|B\rangle =
i \epsilon_{\mu\nu\rho\sigma} e^{(\lambda)}_\nu p^\rho_B p^\sigma_V
F_S^{B\rightarrow V}(0).
\label{defF}
\end{equation}
Here $V$ is a vector meson
with the polarization vector $e^{(\lambda)}$,
$V=\rho, \omega, K^*$ or $\phi$;
$B$ is a generic
$B$-meson $B_u, B_d$ or $B_s$, and $\psi$ stands for the
field of a light $u,d$ or $s$ quark. The vectors $p_B$, $p_V$ and
$q=p_B-p_V$
correspond to the 4-momenta of the initial $B$-meson and the
outgoing vector
meson and photon, respectively. In (\ref{defF}) the QCD
renormalization of the $\bar \psi \sigma_{\mu\nu} q^\nu b$ operator
is implied.
Keeping only the SD-contribution
leads to obvious relations among the exclusive
decay rates, exemplified here by the decay
rates for $(B_u,B_d) \rightarrow \rho + \gamma$ and $(B_u,B_d) \rightarrow K^* + \gamma$:
\begin{equation}
\frac{\Gamma ((B_u,B_d) \rightarrow \rho + \gamma)}
{\Gamma ((B_u,B_d) \rightarrow K^* + \gamma)}
= \frac{\vert \xi_t \vert^2}{\vert\lambda_t \vert ^2}
\frac{\vert F_S^{B \rightarrow \rho }(0)\vert^2}
{\vert F_S^{B \rightarrow K^* }(0)\vert^2} \Phi_{u,d}
=\kappa_{u,d}[\frac{\vert V_{td}\vert}{\vert V_{ts}\vert}]^2 \,,
\label{SMKR}
\end{equation}
where $\Phi_{u,d}$ is a phase-space factor which in all cases is close to 1
and $\kappa_{i} \equiv [F_S^{B_i \rightarrow \rho \gamma}/F_S^{B_i \rightarrow K^*
\gamma}]^2$ is the ratio of the (SD) form factors squared.
The transition form factors $F_s$
are model dependent. However, their ratios, i.e. $\kappa_i$,
should be more reliably calculable as they depend
essentially only on the SU(3)-breaking effects.
If the SD-amplitudes were the only contributions, the measurements of the
CKM-suppressed radiative decays $(B_u,B_d) \rightarrow \rho + \gamma ,
{}~B_d \rightarrow \omega + \gamma$ and $B_s \rightarrow K^* + \gamma$ could be
used in conjunction with the decays $(B_u,B_d) \rightarrow K^* + \gamma$ to
determine
the CKM parameters. The present experimental upper limits on the CKM ratio
$\vert V_{td}\vert/\vert V_{ts}\vert$ from radiative $B$ decays
are indeed based on this assumption, yielding \cite{cleotdul}:
\begin{equation}
\frac{\vert V_{td} \vert }{\vert V_{ts}} \vert \leq 0.75~,
\end{equation}
with a theoretical dispersion estimated in the range $0.64$ - $0.75$,
depending on the models used for the $SU(3)$ breaking effects
in the form factors \cite{abs93,SU3f2}.
The possibility of significant
LD- contributions in
radiative $B$ decays from the light quark intermediate states
has been raised in a number of papers
\cite{ldall}.
Their amplitudes necessarily involve other CKM matrix elements and hence the
simple factorization of the decay rates in terms of the CKM factors
involving $\vert V_{td}\vert$ and $\vert V_{ts}\vert$ no longer holds thereby
invalidating the relationships
given above. In what follows, we argue that the CKM-analysis of
charged $B$-decays,
$B^\pm \rightarrow \rho^\pm \gamma$, would require modifications due to the
LD-contributions but the corresponding analysis of the
neutral $B$-decays $B \rightarrow (\rho^0,\omega) \gamma$ remains essentially
unchanged.
The LD-contributions in $B \rightarrow V + \gamma$ are
induced by the matrix elements of the
four-Fermion operators $\hat{O}_1$ and $\hat{O}_2$ (likewise $O_1$ and
$O_2$). Estimates of these contributions
require non-perturbative methods.
This problem has been investigated recently
in \cite{ab95,wyler95} using
a technique
which treats the photon emission from the light quarks in a theoretically
consistent and model-independent way. This has been combined
with the light-cone QCD sum rule approach to calculate both the SD and LD
--- parity conserving and parity violating --- amplitudes
in the decays $B_{u,d} \rightarrow \rho(\omega) + \gamma$.
To illustrate this, we concentrate on the $B_u^\pm$ decays,
$B_u^\pm \rightarrow \rho^\pm + \gamma$ and take up the neutral $B$ decays
$B_d \rightarrow \rho (\omega) + \gamma$ at the end.
The LD-amplitude of the four-Fermion operators $\hat{O}_1$, $\hat{O}_2$
is dominated by the
contribution of the weak annihilation
of valence quarks in the $B$ meson. It is color-allowed for the
decays of charged $B^\pm$ mesons, as shown in fig. 3, where also the
tadpole diagram is shown, which, however, contributes only in the
presence of gluonic corrections, and hence neglected.
In the factorization approximation, one may write the dominant contribution
in the operator $\hat{O_2}$ (here $O^\prime_2$ is the part of $\hat{O_2}$
with the CKM factor $\xi_u/\xi_t$)
\begin{equation}\label{factor}
\langle \rho\gamma | O^\prime_2|B\rangle =
\langle \rho | \bar d \Gamma_\mu u|0\rangle
\langle \gamma | \bar u \Gamma^\mu b|B\rangle
+
\langle \rho\gamma | \bar d \Gamma_\mu u|0\rangle
\langle 0 | \bar u \Gamma^\mu b|B\rangle ~,
\end{equation}
and make use of the definitions of the decay constants
\begin{eqnarray}
\langle 0 | \bar u \Gamma_\mu b|B\rangle & =& i p_\mu f_B,
\nonumber\\
\langle \rho | \bar d \Gamma_\mu u|0\rangle &=&
\varepsilon^{(\rho)}_\mu m_\rho f_\rho,
\end{eqnarray}
to reduce the problem at hand to the calculation of simpler form factors
induced by vector and axial-vector currents.
\begin{figure}[htb]
\vspace{0.10in}
\vspace*{-3cm}
\centerline{
\epsfysize=6in
{
\epsffile{alibraunfig1.ps}
}
}
\vspace{0.08in}
\vspace*{-8cm}
\caption[]{
Weak annihilation contributions in $B_u \rightarrow \rho \gamma$
involving the operators $O^\prime_1$ and $O^\prime_2$ denoted by
$\bigotimes$ with the
photon emission from a) the loop containing the $b$ quark, b) the loop
containing the light quark, and c) the tadpole which contributes only
with additional gluonic corrections.
\label{fig:3}}
\end{figure}
The factorization approximation assumed in \cite{ab95,wyler95} has been
tested
(to some extent) in two-body and quasi-two body non-leptonic $B$ decays
involving the transitions $b \rightarrow c \bar{c} s$ and $b \rightarrow c \bar{u} d$.
It has not been tested experimentally in
radiative $B$ decays. From a theoretical point of view,
non-factorizable
contributions belong to either the $O(\alpha_s)$ (and higher order)
radiative corrections or to contributions of higher-twist operators to
the sum rules. Their inclusion should not change the conclusions
substantially.
The LD-amplitude in the decay $B_u \rightarrow \rho^\pm + \gamma$ can be written
in terms of the form factors $F_1^L$ and $F_2^L$,
\begin{eqnarray}\label{Along}
{\cal A}_{long} &=&
-\frac{e\,G_F}{\sqrt{2}} V_{ub}V_{ud}^\ast
\left( C_2+\frac{1}{N_c}C_1\right) m_\rho
\varepsilon^{(\gamma)}_\mu \varepsilon^{(\rho)}_\nu
\nonumber\\&&{}\times
\Big\{-i\Big[g_{\mu\nu}(q\cdot p)- p_\mu q_\nu\Big] \cdot 2 F_1^{L}(q^2)
+\epsilon_{\mu\nu\alpha\beta} p^\alpha q^\beta
\cdot 2 F_2^{L}(q^2)\Big\}\,.
\label{ratio2}
\end{eqnarray}
Again, one has to invoke a model to calculate the form factors. Estimates
from the light-cone QCD sum rules give
\cite{ab95}:
\begin{equation}\label{result}
F^L_1/F_S = 0.0125\pm 0.0010\,,\quad F^L_2/F_S = 0.0155\pm 0.0010 ~,
\end{equation}
where the errors correspond to the variation of the
Borel parameter in the QCD sum rules. Including other possible
uncertainties,
one expects an accuracy of the ratios in (\ref{result}) of order 20\%.
Since the parity-conserving and parity-violating amplitudes turn out
to be close to each other, $F_1^L\simeq F^L_2 \equiv F_L$,
the ratio of the LD- and the SD- contributions reduces to a number
\begin{equation}\label{ratio2p}
{\cal A}_{long}/{\cal A}_{short}=
R_{L/S}^{B_u\rightarrow\rho\gamma}
\cdot\frac{V_{ub}V_{ud}^\ast}{V_{tb}V_{td}^\ast} ~.
\end{equation}
Using $C_2=1.10$, $C_1=-0.235$, $C_7^{eff}=-0.306$ (at the scale $\mu=5$ GeV)
\cite{ag95} gives:
\begin{equation}\label{result2}
R_{L/S}^{B_u\rightarrow\rho\gamma} \equiv
\frac{4 \pi^2 m_\rho(C_2+C_1/N_c)}{m_b C_7^{eff}}
\cdot\frac{F_L^{B_u \rightarrow \rho \gamma}}{F_S^{B_u \rightarrow \rho \gamma}}=-0.30\pm
0.07 ~. \end{equation}
To get a ball-park estimate of the ratio
${\cal A}_{long}/{\cal A}_{short}$, we take the central values of
the CKM matrix elements,
$V_{ud}=0.9744\pm 0.0010$ \cite{PDG},
$|V_{td}|=(1.0\pm 0.2)\times 10^{-2}$,
$|V_{cb}|=0.039\pm 0.004$ and $|V_{ub}/V_{cb}|=0.08\pm 0.02$ \cite{al95},
yielding,
\begin{equation}
|{\cal A}_{long}/{\cal A}_{short}|^{B_u\rightarrow\rho\gamma}
= |R_{L/S}^{B_u\rightarrow\rho\gamma}|
\frac{|V_{ub}V_{ud}|}{|V_{td}V_{bt}|} \simeq 10\% ~.
\end{equation}
The analogous LD-contributions to the neutral $B$ decays
$B_d\rightarrow\rho\gamma $ and $B_d\rightarrow\omega\gamma $ are
expected to be much smaller, a point
that has also been noted in the context of the VMD and quark model
based estimates \cite{ldall}. In the present approach,
the corresponding form factors for the decays
$B_d \rightarrow \rho^0(\omega) \gamma$ are obtained from
the ones for the decay $B_u\rightarrow\rho^\pm \gamma$ discussed above by the
replacement of the light quark charges
$e_u\rightarrow e_d$, which gives the factor $-1/2$; in addition,
and more importantly, the
LD-contribution to the neutral $B$ decays
is colour-suppressed, which reflects itself
through the replacement of the factor
$a_1\equiv C_2+C_1/N_c$ in (\ref{ratio2}) by
$a_2\equiv C_1+C_2/N_c$. This yields for the ratio
\begin{equation}
\frac{R_{L/S}^{B_d\rightarrow\rho\gamma}}{R_{L/S}^{B_u\rightarrow\rho\gamma}}=
\frac{e_d a_2}{e_u a_1} \simeq -0.13 \pm 0.05 ,
\end{equation}
where the numbers are based on using
$a_2/a_1 = 0.27 \pm 0.10$ \cite{BHP93}. This would then yield at most
$R_{L/S}^{B_d\rightarrow\rho\gamma} \simeq R_{L/S}^{B_d\rightarrow\omega\gamma}=0.05$,
which in turn gives
${\cal A}_{long}^{B_d\rightarrow\rho\gamma}/{\cal A}_{short}^{B_d\rightarrow\rho\gamma}
\leq 0.02$. Even if this underestimates the LD-contribution by a factor
2, due to the approximations made in \cite{ab95,wyler95},
it is quite safe to neglect
the LD-contribution in the neutral $B$-meson radiative decays.
The ratio of the CKM-suppressed and CKM-allowed
decay rates for charged $B$ mesons
gets modified due to the LD contributions. Following \cite{GP95},
we ignore the LD-contributions in $\Gamma(B \rightarrow K^*\gamma)$. The ratio of
the decay rates in question can therefore be written as:
\begin{eqnarray}\label{ratio3}
\lefteqn{\frac{\Gamma(B_u\rightarrow \rho\gamma)}{\Gamma(B_u\rightarrow K^*\gamma)}
= \kappa_u \lambda^2[(1-\rho)^2+\eta^2]
}
\nonumber\\&&{}
\times\Bigg\{
1+2\cdot R_{L/S} V_{ud}\frac{\rho(1-\rho)-\eta^2}{(1-\rho)^2+\eta^2}
+(R_{L/S})^2 V_{ud}^2\frac{\rho^2+\eta^2}{(1-\rho)^2+\eta^2}\Bigg\}\,,
\end{eqnarray}
Using the central value from the estimates of the ratio of the
form factors squared
$\kappa_u=0.59 \pm 0.08$
\cite{abs93}, and the presently allowed range of the Wolfenstein parameters
$\rho$ and
$\eta$,
it is shown in \cite{ab95} that
the effect of the LD-contributions is modest but not negligible, introducing
an uncertainty
comparable to the $\sim 15\%$ uncertainty in the overall normalization
due to the $SU(3)$-breaking effects in the quantity $\kappa_u$.
\indent
Neutral $B$-meson radiative decays are less-prone to the LD-effects,
as argued above, and hence one expects that to a good approximation
the ratio of the decay rates for neutral $B$ meson obtained in the
approximation of SD-dominance remains valid \cite{abs93}:
\begin{equation}
\frac{\Gamma(B_d\rightarrow \rho\gamma,\omega\gamma)}{\Gamma(B\rightarrow K^*\gamma)}
= \kappa_d\lambda^2 [(1-\rho)^2+\eta^2]~.
\end{equation}
Here $\kappa_d$ represents the
SU(3)-breaking effects in the
form factor ratio squared.
It is a realistic hope that this relation is
theoretically (almost) on the same footing in the standard model
as the one for the ratio of the $B^0$-$\overline{B^0}$ mixing-induced
mass differences, which satisfies the relation \cite{al95}:
\begin{equation}
\frac{\Delta M_s}{\Delta M_d} = \kappa_{sd}
\left\vert \frac{V_{ts}}{V_{td}} \right\vert^2
= \kappa_{sd}\frac{1}{\lambda^2 [(1-\rho)^2+\eta^2]} ~.
\label{xratio}
\end{equation}
The
hadronic
uncertainty in this ratio
is in the SU(3)-breaking factor $\kappa_{sd}\equiv (f_{B_s}^2 \hat{B}_{B_s}/
f_{B_d}^2 \hat{B}_{B_d})$, which involves the pseudoscalar coupling
constants and the so-called bag constants. This quantity is
estimated as $\kappa_{sd}=1.35 \pm 0.25$ in the QCD sum
rules and lattice QCD approaches. (For details and references, see
\cite{al95}).
The present upper limit for the mass-difference ratio $\Delta M_s/\Delta M_d > 12.3
$ at 95 \% C.L. from the ALEPH data \cite{ALEPHxs} provides better
constraint on the CKM parameters, yielding $\vert V_{td}\vert/\vert V_{ts}\vert < 0.35$
than the corresponding constraints from the
rare radiative decays $B \rightarrow (\rho,\omega) + \gamma$, which give an upper
limit of 0.75 for the same CKM-ratio. We expect experimental sensitivity to
increase in both measurements,
reaching the level predicted for this ratio
in the standard model, $\vert V_{td}\vert/\vert V_{ts}\vert =0.24 \pm 0.05$ \cite{al95},
in the next several years in the ongoing experiments at CLEO, LEP and
Tevatron, and the forthcoming ones at the $B$ factories and HERA-B.
\indent
Finally, combining the estimates for the LD- and SD-form factors in
\cite{ab95} and
\cite{abs93}, respectively, and restricting the Wolfenstein
parameters in the range $-0.4 \leq \rho \leq 0.4$ and $ 0.2 \leq \eta
\leq 0.4$, as suggested by the CKM-fits \cite{al95}, we give the
following ranges for the absolute branching ratios:
\begin{eqnarray}\label{ratio4}
{\cal B}(B_u\rightarrow \rho\gamma)
&=& (1.9 \pm 1.6) \times 10^{-6} ~,
\nonumber\\
{\cal B}(B_d\rightarrow \rho\gamma) &\simeq& {\cal B}(B_d \rightarrow \omega \gamma)
= (0.85 \pm 0.65) \times 10^{-6} ~,
\end{eqnarray}
where we have used the experimental value for the branching ratio
${\cal B} (B \rightarrow K^* + \gamma) =(4.5 \pm 1.5 \pm 0.9) \times 10^{-5}$
\cite{CLEOrare1},
adding the errors in quadrature. The large error reflects the poor
knowledge of the CKM matrix elements and hence experimental determination
of these branching ratios will put rather stringent constraints on the
Summarizing the effect of the LD-contributions in
radiative $B$ decays, we note that they are
dominantly given by the annihilation diagrams. QCD sum-rule-based
estimates are very encouraging in that they lead to the conclusion
that such contributions are modest in exclusive radiative $B$ decays,
in particular in the neutral $B$-decays $B^0 \rightarrow (\rho^0,\omega) + \gamma$.
This estimate should be checked in other theoretically sound frameworks.
Of course, forthcoming data on specific $B$-meson decays will be able to
check this directly.
Presently available data suggest that
the contribution of annihilation diagrams in $B$ decays is not significant,
as seen through the near equality of the lifetimes
for the $B^\pm, ~B_d^0$ and $B_s^0$ mesons and the near equality of
the observed $B^\pm$ and $B^0$ radiative decay rates.
We have argued that this is very probably
also the case for the CKM-suppressed radiative decays, with $B^\pm \rightarrow
\rho^\pm \gamma$ modified by $O(20)\%$ from its SD-rate.
%
\noindent
{\bf Acknowledgements}:
I would like to thank Vladimir Braun, Christoph Greub, David London,
and Hubert Simma
for numerous helpful
discussions and valuable input. Informative discussions with Arkady
Vainshtein are also gratefully acknowledged.
|
\section{Introduction}
Statistical mechanics of randomly charged polymers, called
{\em polyampholytes} (PAs), is a challenging subject because it embodies an
interesting combination of long range interactions and randomness.
While the physics of homogeneous polymers has a reasonably firm
basis,\cite{rPolGen} considerably less is known about heteropolymers,
although the latter present an extremely rich problem of
biological significance.\cite{multi} In this presentation we
review some properties of PAs.
The spatial extent of a polymer is characterized by the critical exponent
$\nu$, which relates its radius of gyration (r.m.s. size) $R_g$ to the
number of monomers $N$, by the power law $R_g\propto N^\nu$. The polymer
will be called ``compact'' if $\nu=1/d$, where $d$ is the dimension of
the embedding space, and ``stretched'' if $\nu=1$.
The simplest model of PAs is as a flexible chain of $N$ monomers,
each of which
has a charge $\pm q_0$ selected from a well defined
ensemble of quenches. The polymer has a characteristic microscopic
length scale $a$, such as the range of excluded volume interactions,
or the nearest neighbor distance along the chain. The monomers of the
PA interact both via (short range) excluded volume interactions
and long range (unscreened) Coulomb interactions.
In the simplest ensemble of quenches, each monomer takes a
charge $q_i=\pm q_0$ independently of all the others; i.e.
$\overline{q_iq_j}=\delta_{ij}q_0^2$, where the overline
indicates averaging over quenches.
While the average excess charge $Q\equiv\sum_iq_i$ of such PAs is
zero, a ``typical''
sequence has Q of about $\pm Q_c$,
with $Q_c\equiv q_0N^{1/2}$. This statement, as well
as the definition of $Q_c$, is unrelated to the embedding
dimension $d$. However, the importance of charge fluctuations
(both for the overall polymer, or for large segments of it)
does depend on the space dimension.
The electrostatic energy of the excess charge,
spread over the characteristic size of an ideal polymer
($R_g\propto N^{1/2}$), grows as $Q^2/R_g^{d-2}\sim N^{(4-d)/2}$.
This simple dimensional argument shows that for $d>4$ weak
electrostatic interactions are irrelevant. (The excluded volume
effects are also irrelevant in $d>4$.)
Thus, in $d>4$ and at high temperatures, the PA behaves as an ideal polymer
with an entropy--dominated free energy of the order of $-NT$. However,
on lowering temperature it collapses into a dense state,
taking advantage of a condensation energy of the order of
$-Nq_0^2/a^{d-2}$.
We are primarily interested in the behavior of PAs in $d=3$.
We present both numerical and analytical evidence that the behavior
of PA in $d<4$ is extremely sensitive to the value of $Q$. In particular,
we show that for $Q<Q_c$ the PA assumes compact configurations, while for
$Q>Q_c$ it is in an expanded state. We suggest a ``necklace''
model which qualitatively describes the latter state.
\section{Properties of Polyampholytes}
\subsection{Neutral Heteropolymers: Short--Range {\em vs.} Long--Range
Interactions}
It is interesting to compare and contrast the behaviors of heteropolymers
with short and long range interactions. Consider a polymer represented
by a self--avoiding chain carrying a quenched sequence of random charges
which interact only at {\em short} distances, with similar charges repelling
each other and the opposite charges attracting.\cite{rsrim}
In $d=3$, at high
temperatures $T$, the presence of such short range interactions does not
influence the behavior of the polymer and it behaves as a self--avoiding
walk (SAW). If $Q$ is small, i.e. the positive and negative charges
are almost balanced, upon decrease of temperature, charges
of opposite sign are more likely to be in the vicinity of each other.
This introduces an effective attraction which increases in strength
with decreasing temperature, and at a certain temperature the polymer
collapses into a globule, as in a regular $\theta$--transition.
When the imbalance $Q$ between the two types of charges increases,
the transition temperature decreases, as demonstrated in
Fig.~\ref{fig:diagram}a, and for sufficiently large imbalance the transition
disappears altogether.\cite{rsrim} Since the behavior of the system varies very
gradually with increasing $Q/N$, it does not really matter whether the
ensemble of chains consists only of quenches with exactly vanishing $Q$,
or is obtained by independently choosing each
$q_i=\pm q_0$ with equal probability. In the latter case $Q$ may deviate
from zero by an amount of order $q_0N^{1/2}$; such fluctuations have a
negligible contribution to $Q/N$ for large $N$.
\begin{figure}
\vspace*{13pt}
\centerline{\epsfysize=2.5truein\epsffile{thetadiag.eps}
\hskip 1cm\epsfysize=2.5truein\epsffile{Fig1.eps}}
\centerline{\hskip 1.4cm (a) \hskip 6.3truecm (b) }
\fcaption{Phase diagrams for random polymers with (a) short--range,
and (b) Coulomb interactions, in $d=3$. Note the different scaling of the
excess charge $Q$ in the two figures.}
\label{fig:diagram}
\end{figure}
The long range Coulomb interactions in random PAs are always
relevant at $d<4$. An approximate treatment using a Debye--H\"uckel type
theory\cite{rHJ} leads to the conclusion that a PA minimizes its
free energy by assuming a configuration in which positive charges
are predominantly surrounded by negative charges, causing screening
as in a regular electrolyte. A detailed calculation suggests\cite{rHJ,KKL}
the following picture: At a finite $T$, the PA can be divided into
segments containing $n$ monomers each, where $n$ defines the length
scale at which electrostatic interactions become of order $T$, i.e.
$Q^2(n)/R_g(n)\sim q_0^2n/(an^{\nu_0})=T$ in $d=3$, where $\nu_0\approx 0.59$.
The overall configuration is then composed of $n$--mer blobs: Inside
the blob the PA behaves as an uncharged polymer, while the blobs form
a compact structure resembling an electrolyte. The resulting PA
is compact (i.e., has $R_g\sim N^{1/3}$) at all temperatures,
with a $T$--dependent density. On the other hand, scaling arguments\cite{rKK}
assume that the $N$--dependence of the $R_g$ of a random PA is
determined from the relation $Q^2/R_g\sim q_0^2N/R_g\approx T$,
leading to the conclusion that $R_g\sim N$ in $d=3$.
This apparent contradiction is resolved by noting\cite{KKL} that the
spatial configurations of PAs are extremely sensitive to the overall
charge $Q$. In particular, it can be shown\cite{KKL,KKMC}
that there is a critical
charge $Q_c=q_0N$, such that for $Q<Q_c$ the PAs are compact, while for
$Q>Q_c$ they are stretched. The diagram in
Fig.~\ref{fig:diagram}b indicates such separation between contracted
and expanded regimes of the PA. In the following subsections we shall
present different arguments and numerical results supporting this conclusion.
\subsection{Polyampholytes with Excess Charge: High Temperature Arguments}
For $d<4$, electrostatic interactions are relevant and the
high temperature phase is no longer a regular self--avoiding
walk. At high $T$, the behavior of the polymer can be
explored perturbatively. For the ensemble of uncorrelated
charges ($\overline{q_iq_j}=0$), the lowest order
($1/T$) correction to the quench--averaged $R_g^2$ vanishes.\cite{KKL}
However, restricting the ensemble to
yield fixed $Q$, slightly modifies the quench--averaged
charge--charge correlations. In
particular, the two--point correlation function becomes
$\overline{q_iq_j}=(Q^2-Q_c^2)/N^2$ for $i\ne j$.
This small (order of $1/N$) correction to the correlation
function causes a significant change in $R_g^2$ due to the
long range nature of the Coulomb interaction.
A $1/T$--expansion\cite{KKL} has a first order
term proportional to $Q-Q_c$. Thus the size of a PA tends to
decrease upon lowering temperature if $Q$ is less than
$Q_c$, and increases otherwise.
Confirming the $1/T$--expansion, a strong $Q$--dependence
of $R_g$ has been seen in Monte Carlo simulations:\cite{KKMC} It has been
shown\cite{KKL} that for $Q<Q_c$ the PA contracts with decreasing $T$,
while for $Q>Q_c$ it expands. Fig.~\ref{fig:config} depicts
low temperature equilibrium configurations of two PAs, one of them has $Q$ slightly
larger than $Q_c$ and is strongly stretched, while the other has
$Q=0$ and forms a globule. The two behaviors
are separated by the vertical line at the
top of the phase diagram in Fig.~\ref{fig:diagram}b. It should be
noted that the arguments leading to location of this line are
independent of $d$, as long as $d<4$.
\begin{figure}
\vspace*{13pt}
\centerline{\epsfysize=2.5truein\epsffile{Fig4d.epsi}
\hskip 4cm\epsfysize=2.5truein \epsffile{Fig4a.epsi}}
\centerline{\hskip 1.3cm (a) \hskip 8.5truecm (b) }
\fcaption{Low $T$ configuration of 64--monomer PAs with (a) $Q=12$,
and (b) $Q=0$. Dark and bright shades indicate opposite charges.}
\label{fig:config}
\end{figure}
\subsection{Polyampholytes with Excess Charge: Low Temperature Arguments}
Monte Carlo results indicate that PAs with $Q=0$ form
dense globules at low temperatures. We can use this observation
as a starting point
for the investigation of the dependence of low--$T$ shapes
on $Q$ in $d=3$. The energy (or rather the
quench--averaged free energy) of the PA is phenomenologically
related to its shape by
\begin{equation}\label{edefenerg}
E=-\epsilon_cN+\gamma S + {b Q^2\over R}.
\end{equation}
The first term is a condensation energy proportional to the
volume (assumed compact) and $\epsilon_c\sim q_0^2/a$,
the second term is proportional to the
surface area $S$ (with a surface tension $\gamma\sim q_0^2/a^3$),
while the
third term represents the long--range part of the electrostatic
energy due to an excess charge $Q$ ($b$ is a dimensionless
constant of order unity). The correctness of separating the energy
into the 3 terms appearing in Eq.~\ref{edefenerg} is not self--evident
in {\em random} systems. A justification for the assumptions
implicit in this equation is provided by an exact enumeration
study of the ground states of PAs defined on a lattice:\cite{KKenum}
Fig.~\ref{fig:spectrum} depicts the ground state energies of {\it all} 2080
possible quenches for 12--step (13--atom) chains.
(The horizontal axis represents an arbitrary numbering of the quenches.)
The energies are clearly separated into 7 bands, corresponding to
excess charges of $Q=1$, 3, 5, $\cdots$, 13. (There is only
one quench with $Q=13$.) While each band has a finite width, we
see that the energy of a PA can be determined rather accurately
by only specifying its net charge $Q$, while the additional details
of the quench appear only in the width of the band.
\begin{figure}
\vspace*{13pt}
\epsfysize=3.0truein\centerline{\epsffile{Fig3.eps}}
\fcaption{Ground state energies of (arbitrarily numbered) quenches.}
\label{fig:spectrum}
\end{figure}
The optimal shape is obtained by
minimizing the overall energy in Eq.~\ref{edefenerg}.
The first term is the same for
all compact shapes, while the competition between the surface
and electrostatic energies is controlled by the dimensionless
parameter
\begin{equation}
\alpha\equiv {Q^2\over 16\pi R^3\gamma}=
{Q^2\over 12V\gamma}\equiv {Q^2\over Q_R^2}\ .
\end{equation}
Here, $R$ and $V$ are the radius and volume of a spherical drop
of $N$ particles, and we have defined the {\it Rayleigh charge}
$Q_R$. We note that in the case of PAs\cite{KKMC,KKenum}
$Q_R\approx q_0N^{1/2}=Q_c$.
The behavior of a system described by Eq.~\ref{edefenerg}
has been analyzed in the past in the contexts of charged conducting
drops, as well as liquid drop models of atomic
nuclei. If one restricts the possible shapes of a globule
to spheroids, then one finds that for sufficiently large values of
$\alpha$ (of order one or more), the energy of the system can be
decreased by distorting the drop into a prolate spheroidal shape.
Both MC simulations\cite{KKMC} and experimental results\cite{rExT}
indicate that for $Q<Q_R$, the $R_g$ of PA at low temperature
is almost independent of $Q$, while
it increases extremely fast as a function of $Q$ for larger charges.
The transition between compact and expanded states with increasing
$Q$ is represented by the vertical line at the bottom of the diagram
in Fig.~\ref{fig:diagram}b. The presence of this line signifies the
instability of the spherical shape, but provides no indication of what
the stable state of the PA looks like. This question is taken up
in the following section. It should be stressed, that although both
high and low temperature arguments seem to provide a consistent picture
of transition between compact and extended states, the equality between
$Q_c$ and $Q_R$ depends on the space dimension. For $d<3$ ($d>3$)
$Q_R$ increases with $N$ slower (faster) than
$Q_c$, and therefore the high--$T$ criterion will not coincide
with the low--$T$ limit.
\section{The Necklace Model}
\subsection{Uniform Charges}
Although the ellipsoidal globule may have a lower energy than
the spherical one, the former is {\em not} the minimum energy
configuration of a charged drop: A uniformly charged
drop minimizes its energy by {\em splitting} into two equal,
infinitely removed, droplets for $\alpha>0.3$, and
the number of such droplets increases with $\alpha$. Obviously,
the PA must maintain its connectivity. We can
constrain the overall shape to remain singly
connected by linking the droplets via narrow tubes of total
length $L$ and diameter $a$. As long as $L a^2\ll R^3$, most
of the charge remains in the spheres. The total
electrostatic energy is proportional to $Q^2/L$,
while the surface energy cost grows as $\gamma a L$; equating
the two gives $L\propto Q$. For large $Q$, the PA will look
like a {\em necklace} of globules connected by narrow strands.
The configuration depicted in Fig.~\ref{fig:config}a indeed bears
some resemblance to such a necklace.
It should be noted that $R_g$ in this simplified discussion is
proportional to the typical charge $Q$, and therefore proportional
to $N^{1/2}$.
This serves as an indication that the quench--averaged
configurations are not compact. The picture has one important
shortcoming for PAs:
it assumes that the excess charge is uniformly distributed
along the chain and disregards the strong charge fluctuations.
It thus applies to such problems as an alternating sequence of
charges\cite{victor} to which some additional charge has been
added, or to a weakly charged homogeneous polymer
in which the attraction (and, therefore, the condensation
energy) is provided by a different mechanism, such as short range
attractions.
\subsection{Effects of Randomness}
We now illustrate the difficulties caused by randomness for
the case of an unrestricted PA. Since $\overline{Q^2}=q_0^2N$,
where the overline denotes an average over the ensemble of
all quenches, we have $\overline{\alpha}=1$. A uniformly charged
drop is unstable
to splitting already for $\alpha\approx0.3$, and thus a
typical random PA is expected to form several globules
connected by narrow tubes. Now consider splitting sequences
of $N$ monomers with total charge constrained by a particular
$\alpha$ into two equal subchains of charges $Q_1$ and $Q_2$.
It is easy to show that each segment has
$\overline{\alpha_{\rm subchain}}=(1+\alpha)/2$, while the mean
product of the charges is $\overline{Q_1Q_2}=q_0^2N(\alpha-1)/4$.
The subchains have, on average, values of $\alpha$ close to
unity. Also, for $\alpha=1$, the average value of the product
of charges vanishes. We thus have the paradoxical situation in
which most spherical shapes are unstable, while there is on
average no energetic gain in splitting the sphere into two
equal parts.
Therefore, charge inhomogeneities drastically modify the necklace
picture. The resulting PA is probably still composed of rather
compact globules connected by a (not necessarily linear)
network of tubes. The globules are selected preferentially from
segments of the chain that are approximately neutral (or at
least below the instability threshold), while the tubes are
from subsequences with larger than average excess charge.
It can be shown\cite{KE} that the probability
of finding a very large neutral segment in a random sequence
of charges is large, i.e. it is possible to build a configuration
consisting of one very large neutral globule with highly
charged ``tails'' sticking out of it.
While the necklace model provides a convenient starting picture
for the behavior of PAs, it does not encompass the entire complexity
of the problem. The probability distribution of the $R_g$ of
the ground states of PAs obtained from exact enumeration
study\cite{KKenum} has a large peak for small $R_g$,
indicating that most of the configurations are compact, as well as
a slowly decaying tail at large $R_g$. This tail
represents the expanded configurations, and determines the behavior of
the quench--averaged $R_g$. Indeed, for $N\le13$ we found\cite{KKenum}
that 80\% of all quenches can be classified as compact, i.e. their
$R_g$ increases as $N^{1/3}$, while the remaining 20\% have quench--averaged
$R_g$s increasing with $N$ at least as fast as that of a SAW.
We can summarize our current knowledge of the behavior of random PAs
as follows:
The exact enumeration and Monte Carlo studies confirm that random
PAs have (on the average) expanded spatial conformations, although
further work is needed to verify whether $\nu=1$. The necklace
model provides a useful qualitative view of the ground state configurations.
Further studies are needed to put the model on a more quantitative basis.
\section{Acknowledgements}
This work was supported by the US--Israel BSF grant
No. 92--00026, and by the NSF through grant No. DMR--94--00334 at MIT's
CMSE.
\section{References}
|
\partial{\partial}
\def1{1}
\def2{2}
\def4{4}
\def3{3}
\def5{5}
\def6{6}
\def7{7}
A novel approach to high energy scattering[1] in gauge theories was
recently suggested by Verlinde and Verlinde[2]: Their idea was to scale
the fields and coordinates in the classical action, to identify which pieces of
the scaled action controlled the behaviour of high energy scattering
with fixed (but larger than the confinement scale) momentum transfer.
Their pretty result suggested that the theory essentially became
a two-dimensional topological theory in this limit, making contact
with earlier ideas of Lipatov[3]. Now, massless quantum field
theories are {\it not} scale invariant in general,
so one does not really expect
this trivial scaling to be correct quantum mechanically. The question
is, how does the renormalized vertex function depend on anisotropic scale
changes? The literature on high energy scattering is enormous, but I have
not found a first-principles account of this issue.
Ever since the pioneering work of Stueckelberg and Petermann[4], it has
been understood that the renormalization group includes a great deal
more information than just the overall scale dependence of Green functions.
The invariance of physical quantities under changes of normalization
conditions can be used to relate couplings constants at different
momenta, for example. In the present note, I want to apply this
textbook technique to the problem of computing the change in the
coupling of three massless particles, if the three momenta are
scaled anisotropically, subject to momentum conservation. Anisotropic
renormalization group equations were considered by Robertson[5]---it is
instructive to compare the two approaches. I shall
do the calculation in the simplest possible, albeit non-perturbatively
unstable, setting, that of asymptotically
free $\phi^3$ theory in 6 dimensions. The gauge theory case is somewhat
more involved, since one has to work without the benefits of dimensional
regularization with minimal subtraction, and will be treated elsewhere.
Anisotropic gauge theory asymptotics on a lattice were considered in
Ref.~6, following the approach of
Karsch[7].
Green functions, or proper vertices, obtained by
using different regulators or different normalization schemes are related
by finite renormalizations. Thus the proper vertices obtained by
using dimensional regularization with minimal subtraction, $\Gamma^{(n)}_{dm},$
are related to those obtained by using the $R$-operation with some
normalization conditions (denoted collectively as $\Theta$),
$\Gamma^{(n)}_{R},$ by
$$\Gamma^{(n)}_{dm}(p_i;\mu,\hat g)= z^{-n/2}(\hat g,\Theta)
\Gamma^{(n)}_{R}(p_i;\mu\Theta,g),$$
where $g$ is a function of $\hat g$ and $\Theta.$
It is assumed that both sides are computed in perturbation theory,
so $\hat g$ and $g$ are small. Since $\Theta$ does not appear
on the left hand side of this equation, it follows that
$${\partial\over\partial\Theta}z^{-n/2}(\hat g,\Theta)
\Gamma^{(n)}_{R}(p_i;\mu\Theta,g)=0.$$
$z$ is computed by comparing $\Gamma^{(2)}$ in both
schemes. The change in the coupling under changes of $\Theta$ can then
be easily computed.
To be precise, the normalization conditions I will use are
$$\eqalign{\Gamma^{(2)}_R(p^2=0) &= 0\cr
{{\hbox{d}}\over{{\hbox{d}} p^2}}\Gamma^{(2)}_R(p^2=\mu^2) &=1\cr
\Gamma^{(3)}_R(p_i=\mu\theta_i) &= g,\cr}$$
as appropriate for a massless theory. The vectors $\theta_i$
can be used to determine the change in $g$ when momenta are varied, since
physical correlation functions are independent of $\theta$ up to a field
rescaling. Without any loss of generality (in Euclidean space), I assume
$\theta_1^2=\theta_3^2=1.$
It is of course unnecessary to use dimensional regularization
with minimal subtraction as the `reference' point.
I do not see a way to do this calculation without specifying
normalization conditions.
Actually, to the one-loop order I need, there is no need to compute $z.$ For
completeness I note that
$$\eqalign{
\Gamma^{(2)}_{dm}(p^2)= p^2\left(1-{g^2\over 2(4\pi)^3}C_{dm}\right)\left[1-
{g^2\over 12(4\pi)^3}\ln p^2/\mu^2\right],\cr
\Gamma^{(2)}_{R}(p^2)= p^2\left(1-{g^2\over12(4\pi)^3}\right)\left[1-
{g^2\over 12(4\pi)^3}\ln p^2/\mu^2\right],\cr }$$
where $C_{dm}$ is a constant. $z$ is obviously needed at higher
orders in the loop expansion.
It remains therefore to evaluate
$$\Gamma^{(3)}_{R}=g+ g^3\int {{{\hbox{d}}^6 q}\over (2\pi)^6}\left[{1\over
q^2(q+p_1)^2(q+p_1+p_2)^2} - {1\over
q^2(q+\mu\theta_1)^2(q+\mu\theta_1+\mu\theta_2)^2}\right].$$
The integrals are standard, and I find that for any parameter $\tau$
parametrizing the normalization conditions,
$${{\partial g}\over{\partial\tau}}
= \left({g\over 4\pi}\right)^3\int^1_0 {\hbox{d}} t
\left[-{1\over2 }{{\partial \ln w^2}\over{\partial\tau}} +
{{\partial b}\over{\partial\tau}}\left\{1-(1+b)\ln{1+b\over b}\right\}\right].$$
Here I have defined
$$\eqalign{w&\equiv \mu\left(t\theta_1 -(1-t)\theta_3\right)\cr
b&\equiv {t(1-t)(\theta_1+\theta_3)^2\over(t\theta_1-(1-t)\theta_3)^2}.\cr}$$
As a check, note that if $\tau=\mu,$ the only term contributing
is the first term, giving
$$\mu{{\partial g}\over{\partial\mu}}
= -\left({g\over 4\pi}\right)^3,$$
implying
$${1\over g^2}-{1\over g_0^2} = {1\over32\pi^3} \ln {\mu\over \mu_0}.$$
Suppose now that $\theta_1 \rightarrow \tau\theta_1.$ I now find
$$\eqalign{{{\partial g}\over{\partial\tau}}
&= \Bigg({g\over 4\pi}\Bigg)^3\mu\int^1_0 {\hbox{d}} t
\Bigg[-(1+2b){tw\cdot\theta_1\over w^2}\cr&+2t(1-t)
{\theta_1\cdot(\tau\theta_1+\theta_3)\over w^2} -2(1+b)\ln{1+b\over b}
\Big[t(1-t){\theta_1\cdot(\tau\theta_1+\theta_3)\over w^2}-b
{tw\cdot\theta_1\over w^2}\Big]\Bigg].\cr}$$
This is still not very transparent, but setting $\theta_1\cdot\theta_3=0$
gives
$$\tau{{\partial g}\over{\partial\tau}}
= \left({g\over 4\pi}\right)^3\int^1_0 {\hbox{d}} t
\left[2t-(3+2b)t^2-2\left(t(1-t)-bt^2\right)(1+b)\ln {1+b\over b}\right]
{\tau^2\mu^2\over w^2}.$$
Taking $\tau\uparrow\infty,$
this gives
$$\tau{{\partial g}\over{\partial\tau}}(\tau=\infty) = -\left({g\over 4\pi}\right)^3,
$$
reflecting the fact that in this limit, $\theta_3$ is set to zero, so changing
$\tau$ is the same as an overall scale change of the normalization point.
It is also possible to do the integral (still with $\theta_1\cdot\theta_3=0$)
explicitly at $\tau=1,$
$$\tau{{\partial g}\over{\partial\tau}}(\tau=1)=
-{1\over 2}\left({g\over 4\pi}\right)^3, $$
which exhibits the fact that the anisotropy dependence is non-trivial.
Write
$$\tau{{\partial g}\over{\partial\tau}}(\tau)=
-\left({g\over 4\pi}\right)^3 f(\tau).$$
$f(\tau)$ is monotonic and rises from $1\over 2$ at $\tau=1$ to $1$ at
$\tau=\infty,$ as shown in figure 1.
Integrating this equation, we get
$${1\over g(\tau)^2}-{1\over g(1)^2} = {1\over32\pi^3}\int^{\tau}_1
{{\hbox{d}}\varpi\over\varpi} f(\varpi)\equiv {1\over32\pi^3} G(\tau).$$
The function $G$ is shown in figure 2.
\vfill\eject
\centerline{}
\centerline{$\tau$}
\vskip-1truein
\centerline{\epsfbox{asbeta1.eps}}
\vskip-.8truein
\centerline{Fig. 1: $f(\tau)$ }
\vskip-.3truein
\centerline{\epsfbox{asbeta2.eps}}
\vskip-1truein
\centerline{$\tau$}
\smallskip
\centerline{Fig. 2: $G(\tau)$}
\vfill\eject
Consider now what happens if $\mu\rightarrow\mu_0\tau^{-\alpha}.$ The
function $f(\tau)$ is now replaced by $f(\tau)-\alpha,$ which is still
positive for $\tau\ge 1$ if $\alpha<{1\over 2}.$ In other words, in spite
of the fact that the theory is asymptotically free, so one expects that the
coupling will grow in the infrared, as long as the triangle formed by
the three momenta has increasing area, the coupling of the three particles
will decrease.
Since the correction from isotropic
scaling is 50\% at the point when the normalization point momenta
are of equal magnitude, the predicted anisotropy dependence (when extended
to gauge theories, of course) may be experimentally observable. The
calculation in the present work used Euclidean momenta, so there is no
direct comparison with high energy scattering data. The Verlinde
scaling made no reference to the signature of the spacetime metric, hence
should be testable within the present framework.
\bigskip
A conversation with D. Gross is gratefully acknowledged.
This work was supported in part by NSF Grant No. PHY90-21984.
\hfuzz=2pt
\bigskip
\centerline{References}
\bigskip
\item{1} For a recent review, see D. Kabat, {\sl Comments Nucl. Part.
Phys.} {\bf 20}, 325 (1992)
\item{2} H. Verlinde and E. Verlinde, {\it QCD at high energies and
two-dimensional field theory}, hep-th/9302104
\item{3} L.N. Lipatov, {\sl Nucl. Phys.} {\bf B365}, 614 (1991)
\item{4} E.C.G. Stueckelberg and A. Petermann, {\sl Helv. Phys.
Acta} {\bf 26} 499 (1953)
\item{5} D.G. Robertson, {\sl Int. J. Mod. Phys.} {\bf A6}, 3643 (1991)
\item{6} I.Ya.Aref'eva and I.V.Volovich, {\it Anisotropic Asymptotics
and High Energy Scattering}, hep-th/9412155
\item{7} F. Karsch, {\sl Nucl. Phys.} {\bf B205}, 285 (1982)
\bigskip
\bigskip
\end
|
\section{Introduction}
\label{sec:1}
Let $(r_n)$ denote the sequence of Rademacher functions.
For $1\leq p\leq2$, we say that a Banach space $X$ is of {\em
Rademacher type\/} $p$ if there exists a constant $c\geq0$ such that
$$
\bigg\|\sum_{k=1}^nx_kr_k\bigg| L_p\bigg\|\leq c\bigg( \sum_{k=1}^n
\|x_k\|^p\bigg)^{1/p}
$$
for all $n\in\Bbb N$ and $x_1,\dots,x_n\in X$.
Furthermore, for $1\leq p<2$, we say that $X$ is of {\em weak Rademacher
type\/} $p$ if there exists a constant $c\geq0$ such that
$$
\bigg\|\sum_{k=1}^nx_kr_k\bigg| L_2\bigg\|\leq c\,n^{1/p-1/2}
\bigg( \sum_{k=1}^n \|x_k\|^2\bigg)^{1/2}
$$
for all $n\in\Bbb N$ and $x_1,\dots,x_n\in X$.
Kahane's Inequality and the Cauchy--Schwarz Inequality tell us that
Rademacher type $p$ implies weak Rademacher type $p$.
Mascioni \cite{mas88} has shown a partial converse.
\begin{theorem}[Mascioni \cite{mas88}]\label{th:1}
If a Banach space $X$ is of weak Rademacher type $p$ for some $1\leq
p<2$ then it is of Rademacher type $r$ for all $r<p$.
\end{theorem}
A different concept can be introduced by using sequences of
$X$--valued martingale differences instead of Rademacher sequences.
Following Pisier \cite{pis86}, we say that a Banach space $X$ is of {\em
martingale type\/} $p$ ($1\leq p\leq2$) if there exists a constant
$c\geq0$ such that
$$
\bigg\|\sum_{k=1}^nd_k\bigg| L_p\bigg\|\leq c\bigg( \sum_{k=1}^n
\|d_k|L_p\|^p\bigg)^{1/p}
$$
for all $n\in\Bbb N$ and all $X$--valued martingale difference sequences
$d_1,\dots,d_n$.
Again, for $1\leq p<2$, we say that $X$ is of {\em weak martingale
type\/} $p$ if there exists a constant $c\geq0$ such that
$$
\bigg\|\sum_{k=1}^nd_k\bigg| L_2\bigg\|\leq c\,n^{1/p-1/2} \bigg(
\sum_{k=1}^n \|d_k|L_2\|^p\bigg)^{1/p}
$$
for all $n\in\Bbb N$ and all $X$--valued martingale difference sequences
$d_1,\dots,d_n$.
In \cite{pis75} Pisier showed a theorem similar to Theorem \ref{th:1}.
\begin{theorem}[Pisier \cite{pis75}]\label{th:2}
If a Banach space $X$ is of weak martingale type $p$ for some $1\leq
p<2$ then it is of martingale type $r$ for all $r<p$.
\end{theorem}
Denoting by $\chi_k^{(j)}$ the Haar functions, we see that the
sequence
$$
\bigg( \sum_{j=1}^{2^{k-1}}x_k^{(j)} \chi_k^{(j)} \bigg) \quad
k=1,\dots,n\quad (x_k^{(j)}) \subseteq X
$$
forms a sequence of $X$--valued martingale differences.
Restricting to those martingales (usually referred to as dyadic or
Walsh--Paley martingales) in the definition of martingale type, we can
define the apparently weaker concept of Haar type.
For $1\leq p\leq2$, we say that a Banach space $X$ is of {\em Haar
type\/} $p$ if there exists a constant $c\geq0$ such that
$$
\bigg\|\sum_{k=1}^n \sum_{j=1}^{2^{k-1}} x_k^{(j)}
\chi_k^{(j)}\bigg| L_p\bigg\|\leq c\bigg( \sum_{k=1}^n
\bigg\|\sum_{j=1}^{2^{k-1}}x_k^{(j)} \chi_k^{(j)}\bigg| L_p
\bigg\|^p\bigg)^{1/p}
$$
for all $n\in\Bbb N$ and $(x_k^{(j)})\subseteq X$.
\pagebreak[3]
Once more it was Pisier \cite{pis75} who showed that Haar type $p$ and
martingale type $p$ coincide.
\begin{theorem}[Pisier \cite{pis75}]
A Banach space $X$ is of martingale type $p$ if and only if it is of
Haar type~$p$.
\end{theorem}
Of course, the next step is to define weak Haar type.
For $1\leq p <2$, we say that a Banach space $X$ is of {\em weak Haar
type\/} $p$ if there exists a constant $c\geq0$ such that
$$
\bigg\|\sum_{k=1}^n \sum_{j=1}^{2^{k-1}} x_k^{(j)} \chi_k^{(j)}
\bigg| L_2\bigg\| \leq c\,n^{1/p-1/2}\bigg( \sum_{k=1}^n
\bigg\| \sum_{j=1}^{2^{k-1}} x_k^{(j)} \chi_k^{(j)} \bigg| L_2
\bigg\|^2 \bigg)^{1/2}
$$
for all $n\in\Bbb N$ and $(x_k^{(j)})\subseteq X$.
The main result of this paper is the following companion to
Theorems~\ref{th:1} and~\ref{th:2}.
\begin{theorem}\label{th1}
If a Banach space $X$ is of weak Haar type $p$ for some $1\leq p<2$
then it is of Haar type $r$ for all $r<p$.
\end{theorem}
This solves a problem in \cite{pis75}, where Pisier could show the
same conclusion under the stronger assumption that $X$ is of weak
martingale type $p$.
Let us now quickly review the contents of this article section by
section.
In Section \ref{sec:2}, we define the necessary concepts.
In Section \ref{sec:3}, we formulate and prove the main new ingredient
of the proof of the main theorem.
In Section \ref{sec:3a} we provide a lemma of combinatorial character
needed to prove a local variant of the main theorem.
In Section \ref{sec:4}, we show how to derive the main theorem from
the results in Section \ref{sec:3}.
Finally, in Section \ref{sec:final}, we consider some examples and
formulate some problems.
I am grateful to Albrecht Pietsch for his numerable hints and useful
remarks, which helped to smooth out the content of this paper.
\section{Definitions}
\label{sec:2}
For $k=1,2,\dots$ and $j=0,\pm1,\pm2,\dots$, we define the {\em Haar
functions} by
$$
\chi_k^{(j)}(t):=
\begin{cases}
+2^{(k-1)/2} & \text{ if $t\in\Delta_k^{(2j-1)}$,}\\
-2^{(k-1)/2} & \text{ if $t\in\Delta_k^{(2j)}$,}\\
0 & \text{ otherwise.}
\end{cases}
$$
Here
$$
\Delta_k^{(j)}:=\left[\tfrac{j-1}{2^k},\tfrac j{2^k}\right)
$$
are the {\em dyadic intervals}.
The following facts are obvious consequences of the definition of the
Haar functions
\begin{eqnarray}
\label{eq2a}
\chi_k^{(j)}(t-\tfrac1{2^{k-1}}) & = & \chi_k^{(j+1)}(t) \quad \text{ and }
\\
\chi_{k+1}^{(j)}(t) & = & \sqrt2\chi_k^{(j)}(2t).
\end{eqnarray}
We denote by
$$
\Bbb D\,:=\{(k,j)\;{:}\; k=1,2,\dots;j=1,\dots,2^{k-1}\}
$$
the {\em dyadic tree}.
For a finite subset $\Bbb F\,\subseteq\Bbb D\,$, we consider the orthonormal
system
$$
\eusm H(\Bbb F\,):=\{\chi_k^{(j)}\;{:}\; (k,j)\in\Bbb F\,\}\subseteq L_2[0,1).
$$
In particular, we will consider finite dyadic trees
$$
\Bbb D\,_m^n:=\{(k,j)\;{:}\; k=m,\dots,n;j=1,\dots,2^{k-1}\},
$$
where $m\le n$.
For $t\in[0,1)$, we also use the branches
$$
\Bbb B\,(t):=\{(k,j)\;{:}\; t\in\Delta_{k-1}^{(j)}\}=\{(k,j)\;{:}\;
\chi_k^{(j)}(t)\not=0 \}.
$$
These are exactly those points of the dyadic tree $\Bbb D\,$ which lie on
one of the infinite paths starting in the root $(1,1)$.
Figure \ref{fig:dyadic tree} shows a part of the set $\Bbb D\,$, where the
thicker dots stand for the first elements in any of the sets $\Bbb B\,(t)$
for $t\in[\frac3{16},\frac14)$.
\begin{figure}[ht]
\begin{center}
\leavevmode
\setlength{\unitlength}{0.018in}%
%
\begin{picture}(228,130)(96,560)
\thinlines
\put(210,570){\circle*{3}} \put(150,630){\circle*{3}}
\put(120,660){\circle*{3}} \put(135,675){\circle*{3}}
\put(143,683){\circle*{3}}
\put(210,570){\line( 1, 1){114}} \put(300,660){\line(-1, 1){ 24}}
\put(270,630){\line(-1, 1){ 54}} \put(240,660){\line( 1, 1){ 24}}
\put(210,570){\line(-1, 1){114}} \put(315,675){\line(-1, 1){ 9}}
\put(285,675){\line( 1, 1){ 9}} \put(255,675){\line(-1, 1){ 9}}
\put(225,675){\line( 1, 1){ 9}} \put(195,675){\line(-1, 1){ 9}}
\put(165,675){\line( 1, 1){ 9}} \put(135,675){\line(-1, 1){ 9}}
\put(105,675){\line( 1, 1){ 9}} \put(120,660){\line( 1, 1){ 24}}
\put(150,630){\line( 1, 1){ 54}} \put(180,660){\line(-1, 1){ 24}}
\thicklines
\put(210,570){\line(-1, 1){ 60}} \put(150,630){\line(-1, 1){ 30}}
\put(120,660){\line( 1, 1){ 15}} \put(135,675){\line( 1, 1){ 9}}
\thinlines
\put(199,555){%
\makebox(0,0)[lb]{$(1,1)$}}
\put(159,625){%
\makebox(0,0)[lb]{$(2,1)$}}
\put(129,655){%
\makebox(0,0)[lb]{$(3,1)$}}
\put(141,669){%
\makebox(0,0)[lb]{$(4,2)$}}
\put(135,690){%
\makebox(0,0)[lb]{$(5,4)$}}
\end{picture}
\end{center}
\caption{The dyadic tree $\Bbb D\,$}
\label{fig:dyadic tree}
\end{figure}
We define the following type ideal norms associated with the systems
$\eusm H(\Bbb F\,)$.
\begin{definition}
For $T\in\frak L(X,Y)$ and for a finite set $\Bbb F\,\subseteq\Bbb D\,$
denote by $\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,))$ the least constant $c\ge0$ such that
\begin{equation}
\left\| \left. \sum_{\Bbb F\,} Tx_k^{(j)} \chi_k^{(j)} \right|
L_2 \right\| \le c \left(\sum_{\Bbb F\,} \|x_k^{(j)}\|^2 \right)^{1/2}
\label{eq3}
\end{equation}
for all $\{x_k^{(j)}\;{:}\; (k,j)\in\Bbb F\,\}\subseteq X$.
We call the map
$$
\mbox{\boldmath$\tau$}(\eusm H(\Bbb F\,))\;{:}\; T\longrightarrow \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,))
$$
the {\em Haar type ideal norm} associated with the index set $\Bbb F\,$.
We write $\mbox{\boldmath$\tau$}(X|\eusm H(\Bbb F\,))$ instead of $\mbox{\boldmath$\tau$}(I_X|\eusm H(\Bbb F\,))$, where
$I_X$ is the identity map of the Banach space $X$.
\end{definition}
\section{Comparison of Haar type ideal norms}
\label{sec:3}
In this section we compare Haar type ideal norms associated with
different index sets $\Bbb F\,$.
The following proposition is an easy consequence of the facts
\eqref{eq2a} and
(2).
\begin{proposition} \label{prop2}
Let $n\ge m\ge1$.
Then for all $T\in\frak L(X,Y)$
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{n+1})) \le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_m^n)).
$$
\end{proposition}
{\sc \proofname:\ }
We split $\Bbb D\,_{m+1}^{n+1}$ into the two subtrees
\begin{eqnarray*}
\Bbb S_1 & := & \left\{ (k,j)\in\Bbb D\,_{m+1}^{n+1} \;{:}\; 1\le j \le 2^{k-2}
\right\}, \\
\Bbb S_2 & := & \left\{ (k,j)\in\Bbb D\,_{m+1}^{n+1} \;{:}\; 2^{k-2}+1 \le j \le
2^{k-1} \right\}.
\end{eqnarray*}
Note that for $0\le t<1/2$
$$
(k,j)\in\Bbb S_2\ \text{ implies }\ \chi_k^{(j)}(t)=0
$$
and for $1/2\le t<1$
$$
(k,j)\in\Bbb S_1\ \text{ implies }\ \chi_k^{(j)}(t)=0.
$$
Hence we may write
\pagebreak[3]
\begin{eqnarray*}
\lefteqn{\intop\limits_0^1 \Big\| \sum_{\Bbb D\,_{m+1}^{n+1}} Tx_k^{(j)} \chi_k^{(j)}(t)
\Big\|^2 dt = } \\
& = &
\intop\limits_0^{1/2} \Big\| \sum_{\Bbb D\,_{m+1}^{n+1}}
Tx_k^{(j)} \chi_k^{(j)}(t) \Big\|^2 dt + \intop\limits_{1/2}^1 \Big\|
\sum_{\Bbb D\,_{m+1}^{n+1}} Tx_k^{(j)} \chi_k^{(j)}(t) \Big\|^2 dt \\
& = &
\intop\limits_0^{1/2} \Big\| \sum_{\Bbb S_1} Tx_k^{(j)} \chi_k^{(j)}(t)
\Big\|^2 dt + \intop\limits_{1/2}^1 \Big\| \sum_{\Bbb S_2} Tx_k^{(j)}
\chi_k^{(j)}(t) \Big\|^2 dt.
\end{eqnarray*}
Substituting $s=2t$ and $h=k-1$ and taking into account formula
(2), the first integral can be estimated as follows
\begin{eqnarray*}
\intop\limits_0^{1/2} \Big\| \sum_{\Bbb S_1} Tx_k^{(j)} \chi_k^{(j)}(t)
\Big\|^2 dt
& = &
\frac12 \intop\limits_0^1 \Big\| \sum_{h=m}^n
\sum_{j=1}^{2^{h-1}} Tx_{h+1}^{(j)} \chi_{h+1}^{(j)} \big(\frac
s2\big) \Big\|^2 ds \\ & \le & \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_m^n))^2 \sum_{h=m}^n
\sum_{j=1}^{2^{h-1}} \|x_{h+1}^{(j)}\|^2 \\
& = &
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_m^n))^2 \sum_{\Bbb S_1} \|x_k^{(j)}\|^2.
\end{eqnarray*}
Replacing $t$ by $t+\frac12$ and $j$ by $j+2^{k-2}$, and using
\eqref{eq2a}, the estimate above passes into
$$
\intop\limits_{1/2}^1 \Big\| \sum_{\Bbb S_2} Tx_k^{(j)} \chi_k^{(j)}(t)
\Big\|^2 dt \le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_m^n))^2 \sum_{\Bbb S_2}
\|x_k^{(j)}\|^2.
$$
which implies the desired inequality.
$\Box$
By iterated application of Proposition \ref{prop2}, we obtain the
following result.
\begin{corollary}
Let $m,n\in\Bbb N$.
Then we have for all $T\in\frak L(X,Y)$ that
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{m+n}))\le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n)).
$$
\label{lem1}
\end{corollary}
The following transformation and their effects on the Haar functions
will play the crucial role in the proof of Proposition \ref{lem5}.
For $(h,i)\in\Bbb D\,$, we denote by $\phi_h^{(i)}$ the transformation of
$[0,1)$ that interchanges the intervals
$$
\Delta_{h+1}^{(4i-2)}\ \text{ and }\ \Delta_{h+1}^{(4i-1)}.
$$
More formally
\begin{equation}
\phi_h^{(i)}(t):=
\begin{cases}
t+\frac1{2^{h+1}} & \text{for $t\in\Delta_{h+1}^{(4i-2)}$} \\
t-\frac1{2^{h+1}} & \text{for $t\in\Delta_{h+1}^{(4i-1)}$} \\
t & \text{otherwise.}
\end{cases}
\label{eq9}
\end{equation}
Lemma \ref{lem2} describes the behavior of the Haar functions under
$\phi_h^{(i)}$.
What is $\chi_k^{(j)}\circ\phi_h^{(i)}$ for $(k,j)\in\Bbb D\,$\/?
It turns out that the most interesting cases are those if $(k,j)$
belongs to the fork $F_h^{(i)}:=\{(h,i),(h+1,2i-1),(h+1,2i)\}$.
\begin{figure}[h]
\begin{center}
\leavevmode
\setlength{\unitlength}{0.018in}%
%
\begin{picture}(228,150)(96,555)
\thinlines
\put(270,630){\circle*{3}} \put(210,570){\circle*{3}}
\put(150,630){\circle*{3}}
\put(210,570){\line( 1, 1){114}} \put(300,660){\line(-1, 1){ 24}}
\put(270,630){\line(-1, 1){ 54}} \put(240,660){\line( 1, 1){ 24}}
\put(210,570){\line(-1, 1){114}} \put(315,675){\line(-1, 1){ 9}}
\put(285,675){\line( 1, 1){ 9}} \put(135,675){\line(-1, 1){ 9}}
\put(105,675){\line( 1, 1){ 9}} \put(120,660){\line( 1, 1){ 24}}
\put(150,630){\line( 1, 1){ 54}} \put(180,660){\line(-1, 1){ 24}}
\thicklines
\put(210,570){\line( 1, 1){ 60}} \put(210,570){\line(-1, 1){ 60}}
\thinlines
\put(156,684){\line( 1, 0){ 48}} \put(159,681){\line( 1, 0){ 42}}
\put(162,678){\line( 1, 0){ 36}} \put(165,675){\line( 1, 0){ 30}}
\put(168,672){\line( 1, 0){ 24}} \put(171,669){\line( 1, 0){ 18}}
\put(174,666){\line( 1, 0){ 12}} \put(177,663){\line( 1, 0){ 6}}
\put(216,684){\line( 1, 0){ 48}} \put(219,681){\line( 1, 0){ 42}}
\put(222,678){\line( 1, 0){ 36}} \put(225,675){\line( 1, 0){ 30}}
\put(228,672){\line( 1, 0){ 24}} \put(231,669){\line( 1, 0){ 18}}
\put(234,666){\line( 1, 0){ 12}} \put(237,663){\line( 1, 0){ 6}}
\put(199,555){%
\makebox(0,0)[lb]{$(h,i)$}}
\put(159,625){%
\makebox(0,0)[lb]{$(h+1,2i-1)$}}
\put(277,625){%
\makebox(0,0)[lb]{$(h+1,2i)$}}
\put(174,690){%
\makebox(0,0)[lb]{$\Bbb S_1$}}
\put(234,690){%
\makebox(0,0)[lb]{$\Bbb S_2$}}
\end{picture}
\end{center}
\caption{Part of the dyadic tree $\Bbb D\,_1^n$}
\label{fig:1}
\end{figure}
If on the other hand the support of $\chi_k^{(j)}$ belongs to one of
the intervals $\Delta_{h+1}^{(4i-2)}$ or $\Delta_{h+1}^{(4i-1)}$ then
obviously $\chi_k^{(j)}\circ\phi_h^{(i)}$ is again a Haar function on
the same level $k$.
This is exactly the case if $(k,j)$ belongs to one of the subtrees
\begin{eqnarray*}
\Bbb S_1 & := &
\{(k,j)\;{:}\; \Delta_{k-1}^{(j)}\subseteq\Delta_{h+1}^{(4i-2)}\}, \\
\Bbb S_2 & := &
\{(k,j)\;{:}\; \Delta_{k-1}^{(j)}\subseteq\Delta_{h+1}^{(4i-1)}\}.
\end{eqnarray*}
In the remaining cases the Haar functions $\chi_k^{(j)}$ are invariant
under $\phi_h^{(i)}$.
Look at Figure \ref{fig:1} to see, how the different index sets
$\Bbb F\,_h^{(i)}$, $\Bbb S_1$ and $\Bbb S_2$ are related to each other.
\begin{lemma}
On the fork $\Bbb F\,_h^{(i)}$ the transformation $\phi_h^{(i)}$ acts as
follows
\begin{eqnarray*}
\chi_h^{(i)}\circ \phi_h^{(i)}
& = & \big(\chi_{h+1}^{(2i-1)}+\chi_{h+1}^{(2i)}\big)/\sqrt2, \\
\chi_{h+1}^{(2i-1)}\circ \phi_h^{(i)}
& = & \big( \sqrt2\chi_h^{(i)} + \chi_{h+1}^{(2i-1)} -
\chi_{h+1}^{(2i)} \big)/2, \\
\chi_{h+1}^{(2i)}\circ \phi_h^{(i)}
& = & \big( \sqrt2\chi_h^{(i)} - \chi_{h+1}^{(2i-1)} +
\chi_{h+1}^{(2i)} \big)/2.
\end{eqnarray*}
Moreover, for $(k,j)\in \Bbb S_1$ or $(k,j)\in\Bbb S_2$, we get
$$
\chi_k^{(j)}\circ \phi_h^{(i)}=
\begin{cases}
\chi_k^{(j+2^{k-h-2})} & \text{if $(k,j)\in\Bbb S_1$} \\
\chi_k^{(j-2^{k-h-2})} & \text{if $(k,j)\in\Bbb S_2$.}
\end{cases}
$$
The remaining Haar functions $\chi_k^{(j)}$ are invariant under
$\phi_h^{(i)}$.
\label{lem2}
\end{lemma}
{\sc \proofname:\ }
Looking at Figure \ref{fig:2}, we easily see that
\begin{eqnarray*}
\chi_h^{(i)}\circ \phi_h^{(i)}
& = & \big( \chi_{h+1}^{(2i-1)}+\chi_{h+1}^{(2i)} \big)/\sqrt2 \\
(\chi_{h+1}^{(2i-1)} + \chi_{h+1}^{(2i)}) \circ \phi_h^{(i)}
& = & \sqrt2 \chi_h^{(i)} \\
(\chi_{h+1}^{(2i-1)} - \chi_{h+1}^{(2i)}) \circ \phi_h^{(i)}
& = & \chi_{h+1}^{(2i-1)}-\chi_{h+1}^{(2i)}.
\end{eqnarray*}
\begin{figure}[h]
\begin{center}
\leavevmode
\unitlength=1.20mm
\linethickness{0.4pt}
\begin{picture}(70.00,34.00)(0.00,5.00)
\put(40.00,23.00){\oval(10.00,3.00)[t]}
\put(35.00,23.00){\vector(0,-1){1.00}}
\put(45.00,23.00){\vector(0,-1){1.00}}
\put(17.00,17.00){%
\makebox(0,0)[cc]{$\scriptstyle\Delta_{h+1}^{(4i-3)}$}}
\put(32.00,17.00){%
\makebox(0,0)[cc]{$\scriptstyle\Delta_{h+1}^{(4i-2)}$}}
\put(47.00,17.00){%
\makebox(0,0)[cc]{$\scriptstyle\Delta_{h+1}^{(4i-1)}$}}
\put(62.00,17.00){\makebox(0,0)[cc]{$\scriptstyle\Delta_{h+1}^{(4i)}$}}
\put(55.00,23.00){\makebox(0,0)[cc]{$\scriptstyle\Delta_h^{(2i)}$}}
\put(25.00,23.00){\makebox(0,0)[cc]{$\scriptstyle\Delta_h^{(2i-1)}$}}
\put(40.00,27.00){\makebox(0,0)[cc]{$\scriptstyle\phi_h^{(i)}$}}
\put(10.00,30.00){\line(1,0){30.00}}
\put(10.00,34.00){\line(1,0){15.00}}
\put(25.00,6.00){\line(1,0){15.00}}
\put(40.00,10.00){\line(1,0){30.00}}
\put(55.00,6.00){\line(1,0){15.00}}
\put(40.00,34.00){\line(1,0){15.00}}
\put(70.00,20.00){\line(-1,0){60.00}}
\put(10.00,21.00){\line(0,-1){2.00}}
\put(25.00,21.00){\line(0,-1){2.00}}
\put(40.00,21.00){\line(0,-1){2.00}}
\put(55.00,21.00){\line(0,-1){2.00}}
\put(70.00,21.00){\line(0,-1){2.00}}
\end{picture}
\end{center}
\caption{The graphs of the functions
$\chi_h^{(i)}$, $\chi_{h+1}^{(2i-1)}$ and $\chi_{h+1}^{(2i)}$}
\label{fig:2}
\end{figure}
Solving this system of equations, one gets the assertions for
$(k,j)\in\Bbb F\,_h^{(i)}$.
The other assertions follow easily from the definition of $\phi_h^{(i)}$.
$\Box$
The following concept turns out to be very useful in the proof of the
main theorem.
\begin{definition}
Let $\Bbb F\,\subseteq\Bbb D\,$ be a finite subset.
The {\em local height} of $\Bbb F\,$ is defined as the maximal number of
indices in $\Bbb F\,$, that are contained in one branch $\Bbb B\,(t)$.
That is, we let
$$
\mathop{\rm lh}\nolimits(\Bbb F\,):=\max_{t\in[0,1)}|\Bbb F\,\cap\Bbb B\,(t)|.
$$
\end{definition}
The next definition is due to the special behavior of the Haar
functions $\chi_k^{(j)}$ under $\phi_h^{(i)}$ for the indices $(k,j)$
in the fork $\Bbb F\,_h^{(i)}$.
\begin{definition}
For given $\Bbb F\,\subseteq\Bbb D\,$, we say that an index $(h,i)\in\Bbb F\,$ is
{\em $\Bbb F\,$--admissible}, if its successors $(h+1,2i-1)$ and $(h+1,2i)$ do
not belong to $\Bbb F\,$.
\end{definition}
For an $\Bbb F\,$--admissible index $(h,i)\in\Bbb F\,$, we assign to $\Bbb F\,$ another
index set $\Phi_h^{(i)}(\Bbb F\,)$ as follows.
We let
$$
(h+1,2i-1),(h+1,2i)\in\Phi_h^{(i)}(\Bbb F\,).
$$
Moreover, we let $(k,j^\ast)\in\Phi_h^{(i)}(\Bbb F\,)$ for all
$(k,j)\in\Bbb F\,\setminus\Bbb F\,_h^{(i)}$, where $j^\ast$ is determined by
$$
\chi_k^{(j)}\circ\phi_h^{(i)}=\chi_k^{(j^\ast)}.
$$
Note that by Lemma \ref{lem2} such a number $j^\ast$ exists.
The transformations $\Phi_h^{(i)}$ obviously enjoy the following properties.
\begin{lemma}
Let $(h,i)\in\Bbb F\,$ be $\Bbb F\,$--admissible.
Then
$$
|\Phi_h^{(i)}(\Bbb F\,)|=|\Bbb F\,|+1 \text{ and }
\mathop{\rm lh}\nolimits(\Phi_h^{(i)}(\Bbb F\,))=\mathop{\rm lh}\nolimits(\Bbb F\,).
$$
\end{lemma}
The next lemma clarifies the importance of the index set
$\Phi_h^{(i)}(\Bbb F\,)$.
\begin {lemma}
Let $(h,i)\in\Bbb F\,$ be $\Bbb F\,$--admissible.
If $f_{\Bbb F\,}$ is given by
$$
f_{\Bbb F\,}=\sum_{\Bbb F\,}x_k^{(j)}\chi_k^{(j)}\ \text{ with } x_k^{(j)}\in X,
$$
then there exist elements $y_k^{(j)}\in X$ such that
$$
f_{\Bbb F\,}\circ\phi_h^{(i)}=\sum_{\Phi_h^{(i)}(\Bbb F\,)}y_k^{(j)}\chi_k^{(j)}.
$$
Moreover, the $l_2$--sum of the families $(x_k^{(j)})$ and
$(y_k^{(j)})$ are the same,
$$
\sum_{\Bbb F\,}\|x_k^{(j)}\|^2=\sum_{\Phi_h^{(i)}(\Bbb F\,)}\|y_k^{(j)}\|^2.
$$
\label{lem4}
\end{lemma}
{\sc \proofname:\ }
Write $f_{\Bbb F\,}$ in the form
$$
f_{\Bbb F\,}=x_h^{(i)}\chi_h^{(i)}+\sum_{\Bbb F\,_0}x_k^{(j)}\chi_k^{(j)}
$$
with $\Bbb F\,_0=\Bbb F\,\setminus\{(h,i)\}$.
In view of Lemma \ref{lem2}
$$
f_{\Bbb F\,}\circ\phi_h^{(i)} =
\tfrac1{\sqrt2}x_h^{(i)}\chi_{h+1}^{(2i-1)} +
\tfrac1{\sqrt2}x_h^{(i)}\chi_{h+1}^{(2i)} +
\sum_{\Bbb F\,_0}x_k^{(j)}\chi_k^{(j^\ast)}.
$$
Again, $j^\ast$ is determined by
$$
\chi_k^{(j)}\circ\phi_h^{(i)}=\chi_k^{(j^\ast)}.
$$
Note that
$$
\sum_{\Bbb F\,_0}x_k^{(j)}\chi_k^{(j^\ast)} =
\sum_{\Phi_h^{(i)}(\Bbb F\,_0)}x_k^{(j^\ast)}\chi_k^{(j)}.
$$
Hence the elements $y_k^{(j)}$ are given by
\begin{eqnarray*}
& y_{h+1}^{(2i-1)}=\tfrac1{\sqrt2}x_h^{(i)},\
y_{h+1}^{(2i)}=\tfrac1{\sqrt2}x_h^{(i)}\ \text{and}\\
& y_k^{(j)}=x_h^{(j^\ast)}\ \text{ for }\ (k,j)\in\Phi_h^{(i)}(\Bbb F\,_0),
\end{eqnarray*}
\nopagebreak[4]
and their $l_2$--sum can easily be computed.
$\Box$
\pagebreak[3]
Note that an index set $\Bbb F\,$ of local height $n$ can be arbitrarily
large.
However, the following lemma shows that, using the transforms
$\Phi_h^{(i)}$, such an index set can be compressed so as to fit into
a certain dyadic tree $\Bbb D\,_{m+1}^{m+n}$.
The idea is to replace an admissible index $(h,i)\in\Bbb F\,$ by its
successors $(h+1,2i-1)$ and $(h+1,2i)$.
\begin{lemma}
Let $n:=\mathop{\rm lh}\nolimits(\Bbb F\,)$.
Then there exist $m\in\Bbb N$ and a finite sequence of transforms
$$
\Bbb F\,=:\Bbb F\,_0 \stackrel{\Phi_{h_1}^{(i_1)}}{\longrightarrow}
\Bbb F\,_1 \stackrel{\Phi_{h_2}^{(i_2)}}{\longrightarrow}
\dots \stackrel{\Phi_{h_N}^{(i_N)}}{\longrightarrow}
\Bbb F\,_N
$$
such that
$$
\Bbb F\,_N\subseteq\Bbb D\,_{m+1}^{m+n}.
$$
\label{lem3}
\end{lemma}
{\sc \proofname:\ }
Since $\Bbb F\,$ is finite, we can choose $m\in\Bbb N$ with
$\Bbb F\,\subseteq\Bbb D\,_1^{m+n}$.
As often as possible, we successively apply transforms
$\Phi_{h_l}^{(i_l)}$ for $\Bbb F\,_{l-1}$--admissible indices $(h_l,i_l)$
$$
\Bbb F\,_{l-1} \stackrel{\Phi_{h_l}^{(i_l)}}{\longrightarrow} \Bbb F\,_l.
$$
In order to guarantee that $\Bbb F\,_l$ is still contained in
$\Bbb D\,_{m+1}^{m+n}$, we always assume that $h_l<m+n$.
Since
$$
|\Bbb F\,_l|=|\Bbb F\,_{l-1}|+1\ \text{ and }\ |\Bbb F\,_l|\le|\Bbb D\,_1^{m+n}|,
$$
this process terminates after a finite number of steps.
Hence, the last index set $\Bbb F\,_N$ does not contain any
$\Bbb F\,_N$--admissible index $(h,i)$ with $h<m+n$.
Choose $h$ such that
$$
\Bbb F\,_N\subseteq\Bbb D\,_h^{m+n}\ \text{ and }\ \Bbb F\,_N\not\subseteq\Bbb D\,_{h+1}^{m+n}.
$$
\pagebreak[3]
If $h=m+n$, then
$\Bbb F\,_N\subseteq\Bbb D\,_{m+n}^{m+n}\subseteq\Bbb D\,_{m+1}^{m+n}$. Thus we are
done.
Otherwise, take any index $(h,j_0)\in\Bbb F\,_N$ on the lowest
level.
Since $h<m+n$ and $(h,j_0)$ is not $\Bbb F\,_N$--admissible, at least one
of its successors, say $(h+1,j_1)$, must belong to $\Bbb F\,_N$.
In this way, we find a sequence
$$
(h,j_0),\ (h+1,j_1),\dots,(m+n,j_{m+n-h})\in\Bbb F\,_N
$$
of length $m+n-h+1$, that belongs to some branch $\Bbb B\,(t)$.
Hence
$$
m+n-h+1\le n=\mathop{\rm lh}\nolimits(\Bbb F\,).
$$
Finally, we conclude from $m+1\le h$ that
$$
\Bbb F\,_N\subseteq\Bbb D\,_h^{m+n}\subseteq\Bbb D\,_{m+1}^{m+n}.
$$
$\Box$
\pagebreak[3]
We can now formulate the most interesting result of this section.
\begin{proposition}
Let $\Bbb F\,\subseteq\Bbb D\,$ be a finite subset of local height $n$.
Then there exists $m\in\Bbb N$ such that for all $T\in\frak L(X,Y)$
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,)) \le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{m+n})).
$$
\label{lem5}
\end{proposition}
{\sc \proofname:\ }
Consider the transforms
$\Phi_{h_1}^{(i_1)},\dots,\Phi_{h_N}^{(i_N)}$ constructed in the
previous Lemma.
Let $\phi_{h_1}^{(i_1)},\dots,\phi_{h_N}^{(i_N)}$ denote the
associated bijections defined in \eqref{eq9}.
Given
$$
f_{\Bbb F\,}=\sum_{\Bbb F\,}x_k^{(j)}\chi_k^{(j)},
$$
we let
$$
f_{\Bbb F\,_l}:=f_{\Bbb F\,_{l-1}}\circ\phi_{h_l}^{(i_l)}\text{ and }
f_{\Bbb F\,_0}:=f_{\Bbb F\,}.
$$
Write
$$
f_{\Bbb F\,_l}=\sum_{\Bbb F\,_l}x_k^{(j,l)}\chi_k^{(j)}.
$$
Since the transformations $\phi_h^{(i)}$ are measure preserving, we have
$$
\Big\|\sum_{\Bbb F\,}Tx_k^{(j)}\chi_k^{(j)}\Big|L_2\Big\| \! = \!
\|f_{\Bbb F\,\unskip_0}|L_2\| \! = \dots = \! \|f_{\Bbb F\,\unskip_N}|L_2\| \! = \!
\Big\|\sum_{\Bbb F\,\unskip_N}Tx_k^{(j,N)}\chi_k^{(j)}\Big|L_2\Big\|.
$$
Moreover, Lemma \ref{lem4} yields that
$$
\sum_{\Bbb F\,}\|x_k^{(j)}\|^2 = \sum_{\Bbb F\,_0}\|x_k^{(j,0)}\|^2 = \dots =
\sum_{\Bbb F\,_N}\|x_k^{(j,N)}\|^2.
$$
Since $\Bbb F\,_N\subseteq\Bbb D\,_{m+1}^{m+n}$, we have
$$
\Big\|\sum_{\Bbb F\,_N}Tx_k^{(j,N)}\chi_k^{(j)}\Big|L_2\Big\| \le
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{m+n}))
\left(\sum_{\Bbb F\,_N}\|x_k^{(j,N)}\|^2\right)^{1/2}.
$$
Hence
$$
\Big\|\sum_{\Bbb F\,}Tx_k^{(j)}\chi_k^{(j)}\Big|L_2\Big\| \le
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{m+n}))
\left(\sum_{\Bbb F\,}\|x_k^{(j)}\|^2\right)^{1/2}.
$$
$\Box$
We now summarize the results of this section.
\begin{theorem}
Let $\Bbb F\,\subseteq\Bbb D\,$ be a finite subset of local height $n$.
Then we have for all $T\in\frak L(X,Y)$ that
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,)) \le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n)).
$$
\label{th:comparision}
\end{theorem}
{\sc \proofname:\ }
The assertion follows immediately from Corollary \ref{lem1} and
Proposition \ref{lem5}.
$\Box$
\begin{remark}
Using the same methods as in the proof of Proposition \ref{lem5}, one can
show that
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,))=\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n))
$$
for all $T\in\frak L(X,Y)$, if $\Bbb F\,$ has {\em exact local height} $n$,
i.e.~if
$$
|\Bbb F\,\cap\Bbb B\,(t)|=n
$$
for all $t\in[0,1)$.
\end{remark}
As a consequence of the remark above, we get the following result.
\begin{corollary}
Let $m,n\in\Bbb N$.
Then we have for all $T\in\frak L(X,Y)$ that
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_{m+1}^{m+n})) = \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n)).
$$
\end{corollary}
\section{A combinatorial lemma}
\label{sec:3a}
In this section we provide a lemma, which is needed in the proof of
Theorem \ref{th:local}.
\begin{lemma}
Let $\Bbb F\,\subseteq\Bbb D\,_1^n$ be an index set of local height $l\le n$.
If $|\Bbb F\,|<2^l-1$ then there exists an index
$(k,j)\in\Bbb D\,_1^n\setminus\Bbb F\,$ such that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,\cup\{(k,j)\}) = l.
$$
\end{lemma}
{\sc \proofname:\ }
For $l=1$ the assertion is trivial.
Now assume that for $l-1$ the lemma is true.
To prove the lemma for $l$, we use induction over $n\ge l$.
If $n=l$ then $\Bbb D\,_1^n$ has exactly $2^n-1=2^l-1$ elements and each
subset of $\Bbb D\,_1^n$ has local height less than or equal to $l$.
Hence, we may take any element $(k,j)\in\Bbb D\,_1^n\setminus\Bbb F\,$.
Since $|\Bbb F\,|<2^l-1$, the latter set is certainly nonempty.
Now assume that $\Bbb F\,\subseteq\Bbb D\,_1^n$ and $n>l$.
Note that the subtrees
\begin{eqnarray*}
\Bbb S_1
& := &
\left\{(k,j)\in\Bbb D\,_2^n\;{:}\; 1\le j\le2^{k-2}\right\},\\
\Bbb S_2
& := &
\left\{(k,j)\in\Bbb D\,_2^n\;{:}\; 2^{k-2}+1\le j\le2^{k-1}\right\}
\end{eqnarray*}
can be canonically identified with $\Bbb D\,_1^{n-1}$.
Let
$$
\Bbb F\,_1:=\Bbb F\,\cap\Bbb S_1\ \text{ and }\ \Bbb F\,_2:=\Bbb F\,\cap\Bbb S_2
$$
and consider them as subsets of $\Bbb D\,_1^{n-1}$ via the identification
above.
If $(1,1)\notin\Bbb F\,$ then
$$
\mathop{\rm lh}\nolimits(\Bbb F\,_1)\le l\ \text{ and }\ |\Bbb F\,_1|\le|\Bbb F\,|<2^l-1.
$$
By the induction hypothesis, there exists
$(k,j)\in\Bbb S_1\setminus\Bbb F\,_1$ such that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,_1\cup\{(k,j)\})\le l.
$$
Since $(1,1)\notin\Bbb F\,$, this implies that also
$$
\mathop{\rm lh}\nolimits(\Bbb F\,\cup\{(k,j)\})\le l.
$$
If however $(1,1)\in\Bbb F\,$ then
$$
\mathop{\rm lh}\nolimits(\Bbb F\,_1)\le l-1\ \text{ and }\ \mathop{\rm lh}\nolimits(\Bbb F\,_2)\le l-1.
$$
Moreover, without loss of generality, we may assume that
$$
|\Bbb F\,_1|\le\frac{|\Bbb F\,-1|}2<2^{l-1}-1.
$$
Since we already know that the lemma is true for $l-1$, there exists
$(k,j)\in\Bbb D\,_1^{n-1}\setminus\Bbb F\,_1$ such that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,_1\cup\{(k,j)\})\le l-1.
$$
Since $(1,1)\in\Bbb F\,$, this implies that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,\cup\{(k,j)\})\le l.
$$
$\Box$
By repeated application of the previous lemma, one easily checks the
following statement.
\begin{corollary} \label{cor:combinatoric}
Let $\Bbb F\,\subseteq\Bbb D\,_1^n$ be an index set satisfying $\mathop{\rm lh}\nolimits(\Bbb F\,)\le
l\le n$.
If $|\Bbb F\,|<2^l-1$ then there exists an index set
$\Bbb F\,_0\subseteq\Bbb D\,_1^n\setminus\Bbb F\,$ of cardinality $2^l-1-|\Bbb F\,|$ such
that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,\cup\Bbb F\,_0) \le l.
$$
\end{corollary}
\section{Haar type and weak Haar type}
\label{sec:4}
First of all, let us introduce variants of the ideal norms $\mbox{\boldmath$\tau$}(\eusm
H(\Bbb D\,_1^n))$ defined in Section~\ref{sec:2}.
\begin{definition}
For $T\in\frak L(X,Y)$ and for $1\le p\le2$ denote by $\mbox{\boldmath$\tau$}_p(T|\eusm
H(\Bbb D\,_1^n))$ the least constant $c\ge0$ such that
$$
\bigg\| \sum_{\Bbb D\,_1^n} Tx_k^{(j)} \chi_k^{(j)} \bigg| L_p \bigg\|
\le c \left( \sum_{k=1}^n \bigg\| \sum_{j=1}^{2^{k-1}} x_k^{(j)}
\chi_k^{(j)} \bigg| L_p \bigg\|^p \right)^{1/p}
$$
for all $(x_k^{(j)}) \subseteq X$.
\end{definition}
Note that
$$
\mbox{\boldmath$\tau$}_2(\eusm H(\Bbb D\,_1^n)) = \mbox{\boldmath$\tau$}(\eusm H(\Bbb D\,_1^n)).
$$
The following theorem allows to easily determine the asymptotic
behavior of the ideal norms $\mbox{\boldmath$\tau$}_p(\eusm H(\Bbb D\,_1^n))$.
It is a refinement of a theorem of Pisier in \cite{pis75} and is due
to Gei\ss{} \cite{gei}.
\begin{theorem}
Let $n\in\Bbb N$ and $1\le p\le2$ be fixed.
For $T\in\frak L(X,Y)$ assume that
$$
\bigg\| \sum_{\Bbb D\,_1^n} Tx_k^{(j)} \chi_k^{(j)} \bigg| L_1 \bigg\|
\le c_n \, \left\| \left. \left( \sum_{k=1}^n \bigg\|
\sum_{j=1}^{2^{k-1}} x_k^{(j)} \chi_k^{(j)} \bigg\|^p \right) ^{1/p}
\right| L_\infty \right\|
$$
for all $(x_k^{(j)}) \subseteq X$.
Then it follows that
$$
\mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n))\le c\,c_n.
$$
Here $c$ is a universal constant not depending on $n$ nor $p$.
\label{th3}
\end{theorem}
The next definitions are motivated by the corresponding concepts in
the case of Rademacher functions.
\begin{definition}
For $1\leq p\leq2$, we say that an operator $T\in\frak L(X,Y)$ is of
{\em Haar type\/} $p$, if there exists a constant $c\geq0$ such that
$$
\mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n)) \leq c
$$
for all $n\in\Bbb N$.
For $1\leq p<2$, we say that an operator $T\in\frak L(X,Y)$ is of
{\em weak Haar type\/} $p$, if there exists a constant $c\geq0$ such
that
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n)) \leq c\,n^{1/p-1/2}
$$
for all $n\in\Bbb N$.
\end{definition}
\pagebreak[3]
\defRemarks{Remarks}
\begin{remark}
\begin{itemize}
\item
Theorem \ref{th3} ensures that for $p<2$ Haar type $p$ implies
weak Haar type $p$.
Indeed, note that
\begin{eqnarray*}
\lefteqn{\bigg\| \sum_{\Bbb D\,\unskip_1^n} Tx_k^{(j)}\chi_k^{(j)}
\bigg| L_1 \bigg\|
\le
\mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n))\, n^{1/p-1/2}\, \times} \\
& & \left\|\left. \bigg(\sum_{k=1}^n \bigg\| \sum_{j=1}^{2^{k-1}}
x_k^{(j)} \chi_k^{(j)} \bigg\| ^2 \bigg)^{1/2} \right| L_\infty
\right\|
\end{eqnarray*}
and hence by Theorem \ref{th3}
$$
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^n)) \le c\, \mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n))\,
n^{1/p-1/2}.
$$
\item
In the case $p=2$ the corresponding definition of weak Haar
type $2$ would obviously coincide with that of Haar type
$2$. That's why, we consider weak Haar type $p$ only for $p<2$.
\item
In Pisier's work \cite{pis75} it was shown that a Banach space
$X$ is of Haar type $p$ exactly if it admits an equivalent
$p$--smooth renorming. A similar statement also holds for
operators; see also \cite{BEA85} or the forthcoming book
\cite{PIW} for this connection.
\item
Pisier in \cite{pis86} also introduced a concept of martingale
type $p$, which again is equivalent to Haar type $p$. This follows
from the considerations in \cite{pis75}.
\end{itemize}
\end{remark}
We can now prove the main theorem of this article.
\begin{theorem}\label{th:main}
If an operator $T\in\frak L(X,Y)$ is of weak Haar type $p$ for some
$1\leq p<2$ then it is of Haar type $r$ for all $r<p$.
\end{theorem}
{\sc \proofname:\ }
The proof follows essentially that of Sublemma 3.1. in Pisier
\cite{pis75}.
The main new ingredient is the use of the results of Section
\ref{sec:3}.
Let $r<p$ and $(x_h^{(i)})\subseteq X$.
Set
$$
S_r:= \Bigg\| \left(\sum_{k=1}^n \bigg\| \sum_{j=1}^{2^{k-1}}
x_k^{(j)} \chi_k^{(j)} \bigg\|^r \right)^{1/r} \Bigg| L_\infty
\Bigg\|
$$
and for $l=1,2,\dots$ define
$$
\Bbb F\,_l:=\left\{ (k,j)\in\Bbb D\,_1^n \;{:}\; \frac {S_r}{2^{l/r}} < 2^{(k-1)/2}
\|x_k^{(j)}\| \le \frac {S_r}{2^{(l-1)/r}} \right\}.
$$
Then it follows that
\begin{equation}
\label{eq:partition}
\sum_{\Bbb D\,_1^n} x_k^{(j)} \chi_k^{(j)} = \sum_{l=1}^\infty \sum_{\Bbb F\,_l}
x_k^{(j)} \chi_k^{(j)}.
\end{equation}
Moreover the sets $\Bbb F\,_l$ have local height less than $2^l$.
To see this last fact, choose $t\in[0,1)$ and note that
\begin{eqnarray*}
S_r^r & \ge & \sum_{\Bbb D\,_1^n} \|x_k^{(j)} \chi_k^{(j)}(t) \|^r \\
& \ge & \sum_{\Bbb F\,_l\cap \Bbb B\,(t)} \| x_k^{(j)} \chi_k^{(j)}(t) \|^r >
\sum_{\Bbb F\,_l\cap \Bbb B\,(t)} \left( \frac {S_r}{2^{l/r}} \right)^r =
|\Bbb F\,_l\cap \Bbb B\,(t)|\, S_r^r \, 2^{-l}.
\end{eqnarray*}
This shows that
\begin{equation}
|\Bbb F\,_l\cap \Bbb B\,(t)|<2^l.
\label{eq:local_height}
\end{equation}
By Theorem \ref{th:comparision} and the definition of weak Haar type
$p$, there exists a constant $c$ such that
\begin{equation}
\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l)) \le \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb D\,_1^{2^l})) \le
c\,2^{l(1/p-1/2)}.
\label{eq:est_F_l}
\end{equation}
Since $r<p$ it follows that
$$
\sum_{l=1}^\infty 2^{l(1/p-1/r)} = c_{pr} < \infty.
$$
Hence \eqref{eq:est_F_l} implies
\begin{equation}
\sum_{l=1}^\infty \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l))\, 2^{l(1/2-1/r)} \le c \,
c_{pr}.
\label{eq:powersum_finite}
\end{equation}
Moreover we have
\begin{eqnarray}
\sum_{\Bbb F\,_l} \|x_k^{(j)}\|^2
& = &
\sum_{\Bbb F\,_l} \intop\limits_0^1 \|x_k^{(j)}\chi_k^{(j)}(t)\|^2 dt
=
\intop\limits_0^1
\sum_{\Bbb F\,_l\cap\Bbb B\,(t)} \|x_k^{(j)} \chi_k^{(j)}(t) \|^2 dt
\nonumber\\
& = &
\intop\limits_0^1\sum_{\Bbb F\,_l\cap\Bbb B\,(t)} 2^{k-1}\|x_k^{(j)}\|^2 dt \\
& \le &
\intop\limits_0^1 |\Bbb F\,_l\cap\Bbb B\,(t)| \left( \tfrac {S_r}{2^{(l-1)/r}}
\right)^2 dt
\le
2^{2/r} 2^{l(1-2/r)}S_r^2.\nonumber
\label{eq:estimat_sum_squares}
\end{eqnarray}
\pagebreak[3]
Now \eqref{eq:partition}, the definition of $\mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l))$,
\eqref{eq:estimat_sum_squares}, and \eqref{eq:powersum_finite} imply
\begin{eqnarray*}
\Big\| \sum_{\Bbb D\,_1^n} Tx_k^{(j)} \chi_k^{(j)} \Big| L_2 \Big\|
& \le &
\sum_{l=1}^\infty \Big\| \sum_{\Bbb F\,_l} Tx_k^{(j)} \chi_k^{(j)}
\Big| L_2 \Big\| \\
& \le &
\sum_{l=1}^\infty \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l)) \left(
\sum_{\Bbb F\,_l} \|x_k^{(j)} \|^2 \right)^{1/2} \\
& \le & 2^{1/r}
\sum_{l=1}^\infty \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,\unskip_l)) 2^{l(1/2-1/r)} S_r \le
2^{1/r} c \, c_{pr} \, S_r.
\end{eqnarray*}
Finally Theorem \ref{th3} gives
$$
\mbox{\boldmath$\tau$}_r(T|\eusm H(\Bbb D\,_1^n)) \le 2^{1/r}\,c \, c_{pr},
$$
which completes the proof.
$\Box$
Looking at this proof more closely and exploiting Corollary
\ref{cor:combinatoric}, we can even prove the following local variant
of the previous result:
\begin{theorem}\label{th:local}
If an operator $T\in\frak L(X,Y)$ is of weak Haar type $p$ for some
$1\leq p<2$ then there exists a constant $c\geq0$ such that
$$
\mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n)) \leq c\, (1+\log n)
$$
for all $n\in\Bbb N$.
\end{theorem}
{\sc \proofname:\ }
Let $(x_k^{(j)})\subseteq X$.
As in the previous proof define
$$
\Bbb F\,_l:=\left\{ (k,j)\in\Bbb D\,_1^n \;{:}\; \frac {S_p}{2^{l/p}} < 2^{(k-1)/2}
\|x_k^{(j)}\| \le \frac {S_p}{2^{(l-1)/p}} \right\}.
$$
Again \eqref{eq:partition} and \eqref{eq:local_height} hold.
Let $m$ be such that $2^m \leq n<2^{m+1}$.
We now use Corollary \ref{cor:combinatoric} to construct
modifications $\Bbb F\,_1',\dots,\Bbb F\,_{m+1}'$ of the sets $\Bbb F\,_l$ of maximal
cardinality.
This is done inductively.
If $|\Bbb F\,_1|\ge3$ then we let $\Bbb F\,_1':=\Bbb F\,_1$.
Otherwise, by Corollary~\ref{cor:combinatoric}, there exist a subset
$\Bbb F\,_1''\subseteq\Bbb D\,_1^n\setminus\Bbb F\,_1$ of cardinality $3-|\Bbb F\,_1|$ such
that $\mathop{\rm lh}\nolimits(\Bbb F\,_1\cup\Bbb F\,_1'')\le2$.
Defining $\Bbb F\,_1':=\Bbb F\,_1\cup\Bbb F\,_1''$, we get that
$$
\mathop{\rm lh}\nolimits(\Bbb F\,_1')\le2^1\ \text{ and }\ |\Bbb F\,_1'| \ge 2^{2^1}-1.
$$
Now assume that $l\le m$ and that $\Bbb F\,_1',\dots,\Bbb F\,_{l-1}'$ are
already defined such that
$$
\begin{array}{rlrl}
\mathop{\rm lh}\nolimits(\Bbb F\,_{l-1}') & \le 2^{l-1}, &
\sum_{k=1}^{l-1} |\Bbb F\,_k'| & \ge 2^{2^{l-1}}-1, \\
\Bbb F\,_{l-1}' & \subseteq \Bbb D\,_1^n \setminus
\bigcup_{k=1}^{l-2} \Bbb F\,_k, &
\bigcup_{k=1}^{l-1} \Bbb F\,_k & \subseteq \bigcup_{k=1}^{l-1} \Bbb F\,_k'.
\end{array}
$$
If
$$
\sum_{k=1}^{l-1}|\Bbb F\,_k'| + |\Bbb F\,_l| \ge 2^{2^l}-1
$$
then we let $\Bbb F\,_l':=\Bbb F\,_l$.
Otherwise, by Corollary \ref{cor:combinatoric} (note that $2^l\le
2^m \le n$), there exists a subset
$\Bbb F\,_l''\subseteq\Bbb D\,_1^n\setminus\Bbb F\,_l$ of cardinality
$2^{2^l}-1-|\Bbb F\,_l|$ such that $\mathop{\rm lh}\nolimits(\Bbb F\,_l\cup\Bbb F\,_l'')\le2^l$.
Defining
$$
\Bbb F\,_l':=(\Bbb F\,_l\cup\Bbb F\,_l'') \setminus \bigcup_{k=1}^{l-1}\Bbb F\,_k',
$$
we get that $\mathop{\rm lh}\nolimits(\Bbb F\,_l')\le2^l$, and
$$
\sum_{k=1}^l |\Bbb F\,_k'| \ge \sum_{k=1}^{l-1} |\Bbb F\,_k'| +
(|\Bbb F\,_l|+|\Bbb F\,_l''|) - \sum_{k=1}^{l-1}|\Bbb F\,_k'| = 2^{2^l}-1.
$$
Moreover, by construction
$$
\Bbb F\,_l'\subseteq \Bbb D\,_1^n \setminus \bigcup_{k=1}^{l-1} \Bbb F\,_k
\ \text{ and }\
\bigcup_{k=1}^l\Bbb F\,_k \subseteq \bigcup_{k=1}^l \Bbb F\,_k'.
$$
The last condition ensures that for $(k,j)\in\Bbb F\,_l'$ we have
$(k,j)\notin \bigcup_{h=1}^{l-1}\Bbb F\,_h$ and hence
$$
2^{(k-1)/2}\|x_k^{(j)}\|\le \frac {S_p}{2^{(l-1)/p}}.
$$
This construction yields a sequence $\Bbb F\,_1',\dots,\Bbb F\,_m'$ of length
$m$.
Finally, we let
$$
\Bbb F\,_{m+1}':=\Bbb D\,_1^n\setminus\bigcup_{l=1}^m \Bbb F\,_l'.
$$
Obviously $\mathop{\rm lh}\nolimits(\Bbb F\,_{m+1}') \le n < 2^{m+1}$.
Moreover, for $(k,j)\in\Bbb F\,_{m+1}'$,
$$
2^{(k-1)/2}\|x_k^{(j)}\|\le \frac {S_p}{2^{m/p}}.
$$
Since again $\mathop{\rm lh}\nolimits(\Bbb F\,_l')<2^l$ for $l=1,\dots,m+1$, we have
\eqref{eq:est_F_l} for $\Bbb F\,_l'$.
Moreover
$$
\sum_{l=1}^{m+1} \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l'))\,2^{l(1/2-1/p)} \le c\,
(m+1) \le c\, (1+\log n).
$$
Furthermore, Inequality \eqref{eq:estimat_sum_squares} holds
with $r$ replaced by $p$.
Now, since
$$
\bigcup_{l=1}^{m+1} \Bbb F\,_l' = \Bbb D\,_1^n,
$$
we get
\begin{eqnarray*}
\Big\| \sum_{\Bbb D\,_1^n} Tx_k^{(j)} \chi_k^{(j)} \Big| L_2 \Big\|
& \le &
\sum_{l=1}^{m+1} \Big\| \sum_{\Bbb F\,_l'} Tx_k^{(j)} \chi_k^{(j)}
\Big| L_2 \Big\|\\
& \le &
\sum_{l=1}^{m+1} \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l')) \left(
\sum_{\Bbb F\,_l'} \|x_k^{(j)} \|^2 \right)^{1/2} \\
& \le &
2^{1/p} \sum_{l=1}^{m+1} \mbox{\boldmath$\tau$}(T|\eusm H(\Bbb F\,_l')) 2^{l(1/2-1/p)} S_p\\
& \le &
2^{1/p} \, c\,(1+\log n)\, S_p.
\end{eqnarray*}
A glance at Theorem \ref{th3} completes the proof.
$\Box$
\section{Final remarks}
\label{sec:final}
To show that the notions of Haar type $p$ and weak Haar type $p$ do
not coincide, we provide the following example.
Let $1<p<2$ and $\sigma_k:=k^{-1/p'}$.
Consider the diagonal operator $D_s\;{:}\; l_1\to l_1$ associated with
the sequence $s=(\sigma_k)$, which is given by
$$
x=(\xi_k) \longmapsto D_sx=(\sigma_k\xi_k).
$$
Then
$$
\mbox{\boldmath$\tau$}(D_s|\eusm H(\Bbb D\,_1^n)) = \left( \sum_{k=1}^n k^{-2/p'}
\right)^{1/2} \le n^{1/p-1/2}
$$
and hence $D_s$ is of weak Haar type $p$.
On the other hand
$$
\mbox{\boldmath$\tau$}_p(D_s|\eusm H(\Bbb D\,_1^n)) = \left( \sum_{k=1}^n k^{-1}
\right)^{1/p'} \ge \frac 12(1+\log n)^{1/p'}
$$
and hence $D_s$ is not of Haar type $p$.
The tedious calculations are carried out in \cite{wen94}.
However, for {\em spaces\/} the situation seems to be more complicated.
\begin{problem}
Given $1<p<2$, does there exist a Banach space $X$ that is of weak
Haar type $p$ but not of Haar type $p$\/?
\end{problem}
In the Rademacher case, such examples are given by the well known
modifications of the Tsirelson space (see Tzafriri \cite{tza79}) which
are even Banach lattices.
It can also be shown, that for Banach lattices the concepts of
Rademacher and Haar type are the same.
However, it seems to be unknown, whether the same is true for the
corresponding weak properties.
The examples above also suggest that a stronger version of Theorem
\ref{th:main} is true.
\begin{problem}
If $T$ is of weak Haar type $p$, does there exist a constant
$c\ge0$ such that
$$
\mbox{\boldmath$\tau$}_p(T|\eusm H(\Bbb D\,_1^n)) \le c\, (1+\log n)^{1/p'}\/?
$$
\end{problem}
One more problem remains open.
In the introduction it was mentioned that martingale type $p$ and Haar
type $p$ are equivalent properties.
\begin{problem}
Is it true that weak martingale type $p$ and weak Haar type $p$ are
the same?
\end{problem}
|
\section{Introduction}
In this paper a new approach to the problem of constructing a
quantum theory of gravity in the cosmological context is proposed.
It is founded on results from four separate directions of
investigation,
which are:
1) A new point of view towards
the interpretation problem in quantum
cosmology\cite{louis,carloqm,forabner,lotc}, which
rejects the idea that a single quantum state, or a single Hilbert
space, can provide a complete description of a closed system like
the universe. Instead, the idea is to accept Bohr's original
proposal that the quantum state requires for its interpretation
a context in which we distinguish two subsystems
of the universe -the quantum system
and observer. However, we seek to relativize
this split, so that the boundary between the
part of the universe that is considered the system and that which
might be considered the observer may be chosen arbitrarily.
The idea is then that a quantum theory of cosmology is specified
by giving an assignment of a Hilbert space and algebra of
observables to every possible boundary that can be considered to
split the universe into two such subsystems. A quantum
state of the universe is then an assignment of a statistical
state to every one of these Hilbert spaces, subject to certain
conditions of consistency. Each of these states is interpreted to
contain the information that an observer on one side of each
boundary might have about the system of the other side.
This formulation then accepts the idea that each observer
can only have incomplete information about the universe,
so that the most complete description possible of the universe
is given by the whole collection of incomplete, but mutually
compatible quantum state descriptions of all the possible
observers. At the same time, the information of different
observers is, to some extent, different, so that there is no way,
in principle, to combine the descriptions of each observer to
construct a single quantum state that could give a complete
description of the whole universe.
This point of view, which has been developed in collaboration
with Louis Crane\cite{louis} and Carlo Rovelli\cite{carloqm} and
is discussed also
in papers by them, may
be called pluralistic quantum mechanics.
2) The Bekenstein bound\cite{bekenstein}, which requires, for
reasons
that will be reviewed in section 2, below, that the Hilbert
spaces associated
to timelike
boundaries of fixed spatial area
$A$ must have finite dimension proportional to
$exp(c A/l_{Planck}^2)$, with $c$ some fixed dimensionless
constant. This has recently led 'tHooft\cite{gerard-holographic}
and Susskind\cite{lenny-lorentz,lenny-bh} to
make the ``holographic hypothesis", according to which quantum
gravity must be understood to be constructed from field theories
on two dimensional surfaces, that describe what knowledge
an observer looking through the surface at the world on the other
side, might have. The question has then been raised as to
what specifies the surfaces on which these ``holographic
field theories" are to be defined. The answer pluralistic
quantum theory gives to this question is that the ``holographic"
description must be applicable to any possible surface, so that the
task of the theory is not to pick out certain surfaces but to provide
the conditions and relationships that hold between the theories
on the different surfaces.
3) Topological quantum field theory\cite{witten-tqft,atiyah,segal}
which, first of all, provides
examples of pluralistic quantum theories\cite{louis}, and, secondly,
gives us a set of finite dimensional
Hilbert spaces with which to realize the Bekenstein bound.
4) The loop representation approach to non-perturbative quantum
gravity \cite{lp1,lp2,carlo-review,ls-review,aa-review}, and,
in particular, the kinematical part of the theory,
associated with the description of spatially diffeomorphism invariant
states and observables in terms of spin networks
\cite{ls-review,weave,volume,renata,spinnet-us,ham1,spinnet-john,gangof5}.
In particular,
we may make use of the fact that the same mathematical structure-
spin networks\cite{roger-spinnet}- appears
as labels of diffeomorphism invariant states
in both the loop representation formulation of quantum gravity
and in topological quantum field
theory\cite{witten-tqft,qnet,louis2d3d}.
One of the motivations for the present work is a study
recently carried out by the author of
quantum general relativity in the presence of a particular set of
boundary conditions\cite{linking}. In this context,
the appearance of
the spin networks as labels of states in these two theories
was used to construct
state spaces that represent the algebra of observables of general
relativity on a timelike boundary out of the Hilbert spaces of
certain Chern-Simons theories. It was then natural to contemplate
the conjecture that the observables measured on the boundary
are sufficient to determine the quantum state of the system.
If this conjecture is true it means
that the whole physical state space for quantum general
relativity, in the presence of these boundary conditions,
can be constructed
from direct products of state spaces for Chern-Simons theory.
Further, the correspondence
between spin and area leads to the result that the state spaces
for quantum gravity constructed in this way satisfy the Bekenstein
bound. This is because they have the property that
once the metric of the boundary is measured, the
space of states accessible by measurements on the boundary is
reduced to a finite dimensional subspace, whose dimension
grew with the exponential of the area.
This work was carried out in a less general context than that
contemplated here, in which state spaces are to be associated with
any possible timelike boundary dividing the universe into two
parts, no matter what conditions are satisfied on it. However, its
results are part of the motivation for this proposal. Indeed, I will
argue below that general features of this construction, and in
particular the decomposition of state spaces for quantum gravity
in terms of the Hilbert spaces of topological quantum field theories,
may be realized in this more general context.
Another motivation for this proposal is to
try to reverse the usual strategy of constructing a quantum theory
of gravity and to take the point of view that, rather than
deriving that theory directly from the classical theory by following
some more or less well defined quantization procedure, the
goal is to formulate a set of postulates that define
the theory directly\cite{exp}. The inspiration for these postulates is certain
results and conjectures which have arisen in the course of
investigations of the quantization of classical gravitational
theories. The
hope is that there may be a set of such postulates that are sufficient to
found a theory which is well defined, has a meaningful physical
interpretation and at the same time allows both classical general
relativity and quantum field theory on fixed backgrounds
emerge as approximations in particular
limits\footnote{An important motivation for this
idea is the recent work of Jacobson\cite{ted-new}, where he does
exactly
this, by deriving classical general relativity from the connection
between area and information (essentially the Bekenstein
bound) and the laws of thermodynamics.}.
The reason for taking such an approach lies in the peculiar situation
that research in quantum gravity has led to, in which
several different
approaches have yielded results that we may hope will be
preserved in the final theory, while at the same time, no single
approach has led to the formulation of a complete theory.
Furthermore, there are, in each case, reasons to believe that
the successful results of each approach may not easily be
recovered in the others approaches. It may then be that what we
have in each case is a partial theory that describes successfully
some domain of quantum gravitational phenomena, but which does not
tell the whole story.
Here, I am referring mainly to three approaches: string theory,
quantum field theory in curved space-time and canonical
quantization of general relativity based on the merging of the
Ashtekar variables with the loop representation. For
example the latter approach has yielded a kinematical
description of quantum general relativity in terms of
(three dimensional) diffeomorphism
invariant states and observables
that has many appealing features. These include the
prediction of discrete spectra for areas and
volumes\cite{ls-review,weave,volume,renata,spinnet-us,gangof5}
the description of states in terms of spin networks, and the
suggestions of the existence of a linearized sector with a
natural Planck scale cutoff\cite{carlojuniche}.
Some of these results even turn out
to be derivable in a completely rigorous
mathematical framework\cite{gangof5}.
At the same time, despite very exciting progress in the last
year, this approach has yet to completely resolve the difficulties
concerning physical observables, time and the inner product which
plague it and, indeed, all canonical approaches to quantum
gravity\cite{timeproblem}.
As a result
of these difficulties, it is still not known whether the exact
physical states\cite{tedandi,lp1,jorgerodolfo}, which were
among the
first results of this approach,
are physically meaningful or useful.
This does not mean that
this approach is not useful. It may lead to a completely well
defined mathematical formalism which, however, suffers difficulties
of interpretation coming from the problems of extending quantum
theory to the cosmological case. It may
be possible in certain contexts to overcome some of these
difficulties by defining approximation procedures
which yield physically meaningful predictions in restricted
circumstances\cite{deiter,ham1,carlot,ham2}. Still,
unless the much discussed fundamental difficulties facing
such theories\cite{timeproblem} are overcome, we still may not
have a satisfactory quantum theory of gravity.
Another crucial issue for this kind of approach is the recovery of
Lorentz invariance in the appropriate limit.
In fact, the restoration of Lorentz invariance is an issue
for any quantum field theory one of whose parameters is
a length which serves
as an invariant
cutoff, for the following reason. An acceptable quantum theory of
gravity must have a limit in which the physics can be
described in terms of small excitations of a Lorentz invariant
vacuum.
But, the existence of a cutoff scale means that if an inertial
observer, in the presence of this vacuum, measures the
spectrum of graviton states he should see
no gravitons with wavelengths
smaller than $l_{cutoff}$. However, if the vacuum
is Lorentz invariant, a second inertial observer,
moving with a large $\gamma$ with respect to the first,
should see gravitons with all wavelengths longer than
$l_{cutoff}$ in their frame. But this means that they will see
gravitons that have wavelength $\gamma^{-1} l_{cutoff}$
in the first observers frame. We have an apparent contradiction.
Another way to see this problem is to ask whether Lorentz
invariance can be consistent with the requirement that the Hilbert
space describing the physics in any region of finite volume
must be finite dimensional, as is required by the Bekenstein
bound. At first sight it seems that the
answer must be no. In conventional quantum field theory it
is possible to construct wavepackets that describe quanta
moving with any peak wavelength $\lambda_0$, with
a spread $\delta \lambda_0 < L$ where $L >> \lambda_0$
is the linear
size of the region. Thus, the existence of finite volume
boundary conditions
does not prevent the theory from having states of arbitrarily
small wavelength. And, in a linear field theory, all these
states are orthogonal to each other, which means that there
are an infinite number of orthonormal states. Thus, it seems
that Lorentz invariance cannot be consistent with a
theory that has a finite number of degrees of freedom per
fixed spatial region.
It is then very impressive that there is one context in which
this problem has been definitely
solved, which is perturbative string
theory\cite{strings,he,attickwitten,stringdiscrete,lenny-lorentz,ks}.
The problem is solved there because the elementary excitations
are extended one dimensional objects. As is explained in detail
in \cite{lenny-lorentz,ks}, string theory is consistent
with Lorentz invariance in
spite of having a finite number of degrees of freedom per fixed
spatial
region because the strings, representing the small excitations
of the vacuum, can diffuse transversally as they are boosted
longitudinally.
A theory of extended objects is different in this respect
from a theory of
pointlike objects, because any extended excitation of
a theory that is both Lorentz invariant
and finite must, if boosted sufficiently
become effectively two dimensional, as its longitudinal extension
is contracted below the scale of the cutoff. This means that
extreme relativistic excitations may be naturally described as
excitations of a two dimensional field theory.
It is then very interesting that, as was shown by Klebanov
and Susskind\cite{ks}, continuum string theory can emerge from a
lattice field theory in which there is a cutoff in the
transverse directions by means of
a limit in which the lengths of the
strings diverge while the transverse cutoff remains fixed.
Given that string theory solves this problem and, more generally,
that the only known perturbative quantum field theories that
describe consistently the coupling of gravitons to themselves
and other fields are perturbative string
theories\cite{strings}, it seems that
any acceptable quantum theory of gravity, whatever its
ultimate formulation, is likely to reduce to a perturbative
string theory in the appropriate limit. It is, of course, possible
to imagine that non-perturbative quantum general relativity
will, for this reason, have some perturbative string theory as an
appropriate limit\cite{exp}. However, while this is a possibility
that must be explored, it's success is certainly not guaranteed by
what we know about the theory so far.
At the same time, string theory cannot be itself the whole theory
unless it has a nonperturbative formulation. While there have
been recently some very interesting hints, coming from the
very significant discovery of non-perturbative effects in string
theory\cite{newstringstuff},
it is still unclear to what extent string theory will have
to be modified to arrive at a completely non-perturbative
formulation of the theory. Thus, counting also quantum field
theory in curved space-time, our situation is that we have
interesting results, and in some cases even physical
predictions\cite{exp}, coming from three different starting
points, each of which, however, may not lead to a complete
theory. In this circumstance it may be prudent to try
to find a new starting point that combines what is useful
about each of these directions.
Another way to see that a new starting point may be needed is
that exactly to the extent that these three well explored approaches
are successful, they cast doubt on the physical
relevance of their starting points, which in each
case involve the quantization
of conventional classical field theories.
The
results of each of these three directions seem indeed to
indicate that there is a
natural Planck scale cutoff, below which physical degrees
of freedom cannot be resolved.
This suggests that, ultimately, a quantum theory of gravity
will not be formulated most simply as a theory of fields on
a differential manifold representing the idealized-and
apparently nonexistent- ``points" of space
and time\footnote{Of course, that space and/or time
must be fundamentally discrete is an old idea,
for a good review, see \cite{discrete}}.
To put this another way, the space of fields-the basic
configuration space of classical field theory-has
been replaced in the quantum theory by abstract Hilbert
spaces. At the same time, ordinary space, in these formulations,
remains classical, as it remains the label space for the field
observables. This perpetuates the idealization of arbitrarily
resolvable space-time points, that the results of string theory,
non-perturbative quantum gravity
and semiclassical quantum gravity (through the
Bekenstein bound) suggest we must give up.
The final motivation for the present approach comes from the
interpretational problem in quantum cosmology. Despite very
interesting proposals
\cite{everett,dwg,griffiths,omnes,gellmann-hartle,hartle,zeh,zurek,simon}
it may be said that
no proposal to interpret the conventional Hilbert space formulation
of quantum cosmology has so far
convincingly succeeded\cite{dowkerkent}. (I will
discuss this issue in some detail in section 4, below). This means
that even if the problem of physical observables could be solved
in the context of conventional Hamiltonian quantization, perhaps as
envisioned by Rovelli\cite{carlo-time} or recently by
Ashtekar, Lewandowski,
Marlof, Mour\~{a} and Thiemann\cite{gangof5}, the
resulting mathematical
theory would still not have an acceptable physical interpretation.
More precisely, as I will argue, while the theory might have
a fanciful interpretation in terms of the observations made by
some ``God" outside of the universe, this would be of no help
to connect the mathematical framework to
observations made by us observers who live in the universe.
For this reason also, it may be useful to contemplate
taking a new point of view according to which quantum cosmology
might have an interpretation strictly in terms of information that
might be held by observers inside the universe. This, as I will
describe, is another goal of the present proposal.
This paper is divided roughly into two parts, the first
of which is more general and motivational, while the
second is more focused and mathematical. The
next three sections motivate the idea of a pluralistic
formulation of quantum cosmology. Sections 2 and 3
describe the argument for, and the implications
of, the Bekenstein bound. Section 4 is devoted to the
problems of the interpretation of quantum cosmology,
and give an introduction to the main ideas of the approach
of Crane\cite{louis}, Rovelli\cite{carloqm} and
the author\cite{forabner,lotc} for a pluralistic
approach to quantum cosmology. The postulates of the
theory are then presented at the end of section 5.
Section 6 is devoted to the problem of time. I show that
this problem can be circumvented in pluralistic quantum cosmology
by, as it were, standing on their heads certain proposals concerning
the problem of time which are usually set in the conventional
formulations
in which a single
state and Hilbert space describe the universe. These proposals,
which fail in one way or another in that context, take on a
somewhat different aspect in the context of pluralistic quantum
cosmology. The problem of
the classical limit, and the related question of the timelike initial
date problem are the subject of section 7.
Sections 8 and 9 are more mathematical, and concern one attempt
to realize the postulates of pluralistic quantum cosmology
by building a theory on certain results of non-perturbative
quantum gravity and topological quantum field theory.
A representation for the connection and frame fields of the
Ashtekar formalism of general relativity, pulled back into an
arbitrary surface, is given in which the state space is constructed
from certain direct sums of representation spaces of quantum
Chern-Simons theory. This formalism incorporates a proposal for
solving the constraints of the theory, in the form given by
Reisenberger\cite{mikerconstraints}, quantum mechanically.
One consequence of this
formalism is that the Bekenstein bound is automatically satisfied.
The paper closes with a short summary of conclusions and open problems.
\section{The Bekenstein argument}
Consider a region of space, $\Sigma$,
the quantum dynamics of which we wish to study. We will
assume that the restriction of the system of interest to
$\Sigma$ is enforced by certain boundary conditions,
defined as conditions the fields must satisfy on the spatial
boundary of $\Sigma$, which we will denote $\cal S$.
Given these, we will assume we can
define a quantum theory which is described by
a Hilbert space ${\cal H}_\Sigma$, and an algebra of
physical observables ${\cal A}_\Sigma$. Among these
there is an important subalgebra, ${\cal A}_{boundary}$,
which consists of those physical observables which are
functionals only of fields on the boundary. Among these
we will assume are
the Hamiltonian, $H_\Sigma$ and the areas of regions
$\cal R$ of
the boundary $\cal S$, which I will denote $A[{\cal R}]$.
Recall that in general relativity the Hamiltonian
is, up to terms proportional to constraints,
defined as an integral on the boundary and is thus
an element of ${\cal A}_{boundary}$.
Since the system contains gravitation, we may assume that
among the spectrum of states are a set which correspond to
black holes. These are semiclassical statistical states, and
we will assume that their real statistical entropies
are given by the usual formulas, at least
in the semiclassical limit when their masses and areas are
large in Planck units.
The argument is simplest in the case that we assume that
the induced metric in $\cal S$ is the two sphere metric.
It proceeds by assuming that the region
$\Sigma$ can contain an object $\cal O$ whose complete
specification requires an amount of information
$I_{\cal O}$ which is larger
than
\begin{equation}
I_{\cal S}= {A[{\cal S} ] \over 4 l_{Pl}^2}
\end{equation}
which is of course the entropy of a black hole whose horizon just
fits inside of $\cal S$.
Let us assume that initially we know nothing about
$\cal O$, so that
$I_{\cal O}$ is a
measure of the entropy of the system.
However, with no other information we can conclude
that $\cal O$ is not a black hole,
as the largest information that could be contained in any black
hole in $\Sigma$ is $I_{\cal S}$. We may then argue, using the
Hoop theorem\cite{hoop} that the energy
contained within $\Sigma$ (as measured either by a quasilocal
energy on the surface or at infinity)
must be less than that in a black hole
whose horizon has area ${\cal A}[{\cal S}]$. But this being the case
we can now add energy to the system to bring it up to the mass
of that black hole, which has the result of transforming $\cal O$ into
the black hole whose horizon just fits inside the sphere $\cal S$.
This can be done by dropping quanta slowly into the black hole,
in a way that does not raise the entropy of its exterior. As a
result, once the black hole has formed we
know the entropy of the system, it is
$I_{\cal S}$. But we started with a system with
entropy $I_{\cal O}$, which we assumed is larger. Thus, we have
violated the second law of thermodynamics. The only way
to avoid this is if $I_{\cal O} < I_{\cal S}$.
We may remark that this argument employs a mixture of
classical, statistical and semiclassical reasoning. For example,
it assumes both that the hoop theorem from classical general
relativity applies, at least in the case of black hole masses
large in Planck units, to real, quantum black holes. One might
attempt to make a detailed argument that this must be the
case if the quantum theory is to have a good classical limit.
However worthy of a task, this will not be pursued here,
as it is unlikely that any such argument can be elevated to
establish the necessity, rather than plausibility of the
Bekenstein bound, in the absence
of a complete theory of quantum gravity. Perhaps it is
sufficient then just to note that in twenty years no plausible
counter argument to the Bekenstein bound has survived.
\section{Consequences of Bekenstein's bound}
The Bekenstein bound has profound consequences for
the question of how a quantum theory of gravity might be
formulated. If the entropy of
a quantum system is bounded above by $S_{max}$,
it means that that system
must have a finite dimensional state space, whose dimension
is given by
\begin{equation}
dim {\cal H}=e^{S_{max} Ln(2) } =e^{c A[{\cal S}] \over l_{Pl}^2}
\end{equation}
where $c$ is a fixed dimensionless constant.
This is true because a system whose entropy has an absolute
upper limit can, in principle, be completely specified by giving the
answers to a finite number of yes-no questions.
Thus, not only can a quantum theory of gravity not
be a conventional
quantum field theory, with an infinite number of degrees of
freedom per finite amount of spatial volume, it cannot even be
a conventional cutoff quantum field theory with a fixed finite
number of degrees of freedom per spatial volume. Instead,
we have a fixed number of degrees of freedom per unit area
of the surface of the region.
This has led 't Hooft\cite{gerard-holographic} and
Susskind\cite{lenny-lorentz,lenny-bh} to make the
{\it holographic hypothesis},
which is that a theory of quantum gravity
which describes a region bounded by a spatial surface $\cal S$
may be described by a quantum theory with a finite number
of degrees of freedom on the surface $\cal S$.
There are, however, two difficulties with the holographic hypothesis
which must be faced. The first is what kind of quantum theory
will live on the surfaces. It seems challenging to imagine
a kind of a field theory that could both provide a realistic
description of interacting gravitational and matter fields, but
is at the same time based on finite dimensional
Hilbert spaces. Even the harmonic oscillator has
an infinite dimensional Hilbert space. 't Hooft has explored the
possibility that such a theory might be constructed from
cellular automata. While this is an intriguing suggestion, in this
paper a different proposal will be explored, which is that these
theories are constructed, in a certain way to be described,
from topological quantum field theories.
The second question that must be confronted
is to what two dimensional
surfaces are we to associate the hypothetical holographic
field theories. Closely related to this is another worry.
Most of the arguments given for
the holographic hypothesis involve static situations. As a result, it is
not completely clear whether the boundaries on which
the field theories are to be defined should be in the general
case two dimensional or three dimensional.
To address these issues, we must
leave to one side for a moment the arguments of
Bekenstein, 't Hooft and Susskind and consider a different
set of problems, that seem at first sight to be completely
independent. These are the interpretational problems of
quantum cosmology.
\section{Pluralistic quantum cosmology}
In this section I motivate the main conceptual parts of the
proposal of this paper. I begin with a review of the
interpretational problems
that the new proposal is intended to address.
\subsection{Summary of the problem of the interpretation of
quantum cosmology}
The interpretational difficulties with
quantum cosmology arise
because the conventional interpretations of quantum theory
require that the quantum state description be applied only
to subsystems of the universe. The interpretation of the
theory requires the existence of things which are in the
universe but outside of the system described by the quantum
state, including the measuring instruments, the clocks that
give meaning to the Schrodinger evolution and the observers.
There is, of course, a long history of attempts to modify the
interpretation of the theory, while keeping the formalism
fixed, in order to overcome the dependence of the quantum
description on a split of the world into two parts, so that
quantum theory may be applied to the universe as a whole.
If I may briefly summarize the history of this story, it
began with the many worlds interpretation of
Everett\cite{everett},
which may be put in several different forms, some more
metaphysical\cite{dwg} and some more operational\cite{me-many}.
However, all forms
of the many worlds interpretation suffer from an ambiguity,
which is called the preferred basis
problem\cite{zurek,me-many}. The
connection
of the theory to actual observations requires the selection
of a preferred basis, corresponding to observables we use to
describe the real world. The theory itself seems to provide
no such choice. A number of interesting ideas have been
tried to solve this problem\cite{zurek,zeh},
but none were completely successful.
The last few years attention shifted towards one kind of proposal
to solve this problem,
which is the consistent or decoherent
histories program\cite{griffiths,omnes,gellmann-hartle,hartle}.
The basic ideas here are two: first to interpret
the theory in terms of sequences of projection operators, which
pick out a history of quantum mechanical observations, and
second, to give an interpretation only to measurements of
consistent or decoherent sets of such histories, which are those
for which the interference effects vanish, either absolutely or
``for all practical purposes". This program was based on the
observation that sets of alternative
histories that correspond to classical
motions do have the property of being consistent, at least to
an enormously good approximation.
However, this program is also facing a difficulty analogous
to the preferred basis problem because, as shown in a recent
paper by Dowker and Kent\cite{dowkerkent}, there are an
infinite number
of equally consistent sets of histories, most of which
do not correspond to anything that could
be described in classical language. Thus, while it is
true that there are sets of consistent histories that describe
evolution in terms of quasi-classical variables, these are not
distinguished by the formalism of the quantum theory. Thus,
a selection principle, external to the formalism of the theory,
seems to be needed to make a connection between the theory
and what is actually observed.
Very interesting ideas which relate this selection problem to the
existence of self-organized complexity have
been proposed\cite{gellmann-hartle,simon}, and
are very much worth exploring. However,
in light of this situation it may be
interesting to examine other approaches, that require the
formalism of quantum theory to be modified in order to
extend it to a theory of the whole universe.
\subsection{The basic idea of pluralistic quantum cosmology}
I would like to describe here an approach to such a modification,
that has evolved over a number of years through conversations
with Louis Crane and Carlo Rovelli.
I would like to introduce this by stepping back for a moment and
asking what the context of the problem is. The problem of extending
our physics from a description of isolated systems to a description
of the universe as a whole involves us in a number of profound
issues, one of which is the question of what an observable
in such a theory might be. Both abstract
argument\cite{leibniz,mach,pierce,me-ste}, and the
example of our one successful cosmological theory, which is
general relativity\cite{john-hole,julian-gr}, tell us that the
notion of what is observable
must be different in certain aspects in a cosmological theory
then it is in a theory of an isolated system. Indeed, general relativity
differs from Newtonian physics profoundly in its formulation of
observables, because it is based on a relational view which denies
the existence of any nondynamical background structure that
may give
meaning to observables, such as coordinates or an a priori notion
of time or spatial geometry. Instead, all observable quantities
are defined in terms of relationships between dynamical
degrees of freedom.
This relational conception of observables,
which has its origins in Leibniz's \cite{leibniz} and Mach's \cite{mach}
criticisms of Newton, is realized mathematically in terms of
invariance under active diffeomorphisms. As has been
discussed in detail, the so called ``problem of time" in
classical and quantum cosmology is no more and no less than
the expression of this situation with respect to the notion
of evolution: the solution is that time evolution can only
meaningfully be talked about in terms of relationships between
physical degrees of freedom\cite{carlo-time,julian-gr,julian-qc}.
The only exceptions to this are the
asymptotically flat case, or other cases in which boundary
conditions are imposed, in which case the diffeomorphism
invariance is broken by the conditions imposed on the boundary.
I would like to put this in the following way: the mathematical
structure of general relativity in the cosmological case
forbids any possibility of an
interpretation in terms of operations or measurements made
by an imaginary observer ``outside of the universe." If we
want to speak of an observer outside of the system we must
modify the mathematical structure of the theory in a way that
breaks the gauge invariance and so allows us to represent,
in a kind of idealized way, the possibility of measurements made
by an imaginary external observer.
We may then say that there should be a
problem with extending quantum
mechanics unmodified to the cosmological case, because the
theory does not satisfy the principle, just stated, that the
mathematical
structure of the theory should forbid the possibility of an
interpretation in terms of an observer outside of the system.
The mathematical
structure of quantum mechanics in fact does implicitly refer
to the existence of things outside the system. Among these
are
the clock, measuring instruments and observer that give
meaning to the expectation values and their time evolution.
It then follows that if we could find an interpretation of the
theory in terms of measurements made by observers inside
the quantum system, the theory would still be unsatisfactory,
because the theory would allow us to formulate an
interpretation
in terms of an observer outside of the system. This is true
by definition if we assume we can extend the formalism of
quantum mechanics to the cosmological case without modification,
because in that case the formal structures that are used in the
standard Bohr interpretation are still present. Whatever
other interpretation was offered, these structures could then be
used as the basis of an interpretation of quantum cosmology
in terms of an imaginary observer outside of the universe.
This is absurd, and it is my opinion that it is not enough just to
refrain from doing this. As in the case of general relativity,
we should require that {\it the
mathematical structure of quantum theory
must be modified in such a
way that there is no possibility of an interpretation of the formalism
in terms of an observer outside of the universe}.
I would like to raise this statement to the level of a
fundamental principle, which we may call the
{\it Principle of the absurdity of the possibility of an outside observer.}
I have discussed its
motivation and history in more detail elsewhere\cite{forabner,lotc,me-ste}.
Here I
would like to go on to sketch what modifications of quantum
theory might lead to a theory that satisfies it.
If we follow the example of general relativity we have a clue,
which is that the extension of the theory to the cosmological
case should correspond to the restoration of a gauge
symmetry, such that the standard theory, with fixed external
observers, is obtained by breaking it in some way.
What should this new gauge symmetry be? Given what
has been said, it is clear that
it should arise from treating the observer and the
quantum system on an equal footing, so that the split between
them can be made arbitrarily.
Thus, our goal is not to eliminate the observer, it is
instead, to relativize
him. We would like a formalism that allows us to divide the
universe arbitrarily into two parts, and call one part of it the
observer and the other the system. We would like there to
be something like a gauge symmetry, that expresses the
arbitrariness of the split. And, most importantly, to satisfy
the principle, we must do this in such a way that
it is impossible to construct a single state space that would
allow us the possibility of speaking in terms of a description
of the whole system by an external
observer.
There is a natural way to do this, which is to associate
the operator algebras and their representations in terms
of Hilbert spaces not to the universe as a whole, but to
every possible splitting of the universe into
subsystems\footnote{To my knowledge, this proposal was
first made by Crane\cite{louis}}.
Thus, for every way of splitting the universe into two parts
we will have an associated state space and observable algebra.
Unlike quantum mechanics, there will not be one preferred
split, every possible split will be allowed. Instead of conditions
that pick out which split is given an interpretation (analogous to
the preferred basis problem) we seek consistency conditions that
hold among the set of all state spaces and observable algebras.
These conditions will specify the ways in which the observations
made by each possible subsystem of the universe on the rest of it
may be consistent with each other.
Thus, our slogan is ``Not one state space and many worlds, but
one world, described consistently by many state spaces."
It is important for this point of view that we take the
statistical interpretation of the quantum state, according
to which the wavefunction is not something real, but is only
a representation of the information that the observer has
about the system\cite{statistical}. We may note
that the mixing of quantum
and ordinary thermal statistics
in gravitational fields, and the fact that they are mixed
up by transforming between inertial and noninertial reference
systems, strongly suggests that quantum statistics are
really ordinary statistics which arise due to some non-local physics,
and are distinguishable from conventional thermal effects only by
special observers in special situations\cite{me-mixing}. This point has
been discussed at length elsewhere, for the present, it is only
important to appreciate that there are reasons why the unification
of quantum theory and relativity might force us to accept the
statistical view of the
quantum state\footnote{Of course, there are claims of
the ``observability of the quantum
state". However, these seem to be in fact statements of the
observability of non-classical or
non-local quantum effects, which are undeniably
real.}.
Further, we may note that there are clear arguments against
the possibility of an observer, in either classical or quantum theory,
making a complete measurement of the state of any system
that includes them\cite{nonself}. These arguments reply on the
fact that a measurement requires a representation, within a
subsystem of the observer, of the information gained about
the system. To make a measurement of something is then
to gain some
information, which is represented in the state of some degrees
of freedom inside the observer. Thus, when an observer makes
an observation, its state must change, to record the information
gained. But if the observer is contained in the system, then
the state of the system changes consequently, as a result of
the measurement, from the measured state. The observer may
attempt to again measure its state to ascertain the change, but
again the state must change, so that the
result is an infinite regress.
Moreover, for a complete self-measurement he would need an
infinite number of degrees of freedom, because there is a problem
of infinite regress, each time he observes himself he needs a new
place to put the information gained specifying his previous state.
Thus, as soon as we accept the point of view that the quantum state
corresponds in some exact way with the information that an
observer may gain about the system
as the result of an observation of it, we must accept also that no
such description can give a complete specification of the whole
universe, assuming that it also contains the observer. The only
way to have a description which is both complete and
quantum mechanical is to use the fact that in fact the
universe is complex enough that there are many subsystems that
may be called observers, and to then postulate that the complete information
for describing the state of the universe is represented in a number
of Hilbert spaces, each of which represents the incomplete
information that
a particular observer may have about the
universe.
We might also note that taking the Hilbert space to correspond to
the interaction of an observer and a system brings the interpretation
of the formalism into complete correspondence with what we
actually do when we apply quantum theory to the real world. For
in real practice, physicists describe different isolated systems with
different Hilbert spaces and algebras of observables. One thing
that this interpretation accomplishes is thus to make the theory
correspond
directly to actual practice. We no longer need to
imagine that all the actual applications of quantum theory are
to a greater or lessor extent approximations to some ``real
quantum description" in terms of some ideal ``Hilbert space of
the universe."
\subsection{The measurement problem and the consensus condition}
If we accept the basic idea of pluralistic quantum cosmology,
we no longer have to answer
questions about which split or which basis or which set
of histories corresponds to reality. Instead, we have to answer
a new kind of question, which is what relationships may be
expected to hold between any two possible Hilbert space
descriptions, each giving an incomplete description of the world
arising from a different division of the universe into two
subsystems. This becomes the key question for the theory;
we will see, indeed, that the whole of the theory,
and indeed, even its dynamics, may be expressed in terms of
these relationships.
The first questions that must be asked about these relationships is
whether the existence of different incomplete descriptions of the
same system can be consistent with the
principles of quantum mechanics, and in particular with the
uncertainty principle. Further, does this point of view
shed any light on the measurement problem?
An argument of Rovelli suggests that the answer is yes\cite{carloqm}.
I would like now to review Rovelli's argument, as it suggests
the need for a certain consistency condition, which I
will call the {\it consensus condition.}
Suppose that there is a quantum system $A$ surrounded
by a two surface ${\cal S}_1$, which we may informally
call the ``box" that contains $A$.
Surrounding that box is second, ${\cal S}_2$,
which contains the
system $A$ and an observer, $O_1$. Outside of that is
a second observer, $O_2$. The measurements the
two observers,
$O_1$ and $O_2$ make are recorded in the states in two Hilbert
spaces we may call ${\cal H}_{{\cal S}_1}$ and
${\cal H}_{{\cal S}_2}$.
Now, let us assume that $O_1$ measures the
state of $A$, using a measuring instrument that
measures fields on the surface ${\cal S}_1$. The
result is described differently by the two observers,
using the two Hilbert spaces.
The first observer's instruments register a definite value,
$\lambda_1$
which means that she will project the state that describes
$A$ into a definite state $|1> \in {\cal H}_{{\cal S}_1}$.
The second observer measures the combined state of
$A$ and the measuring instruments and observer
$O_1$, and learns only that as a result of their interaction
they are in some correlated state,
$\sum_i |i>_1 \otimes |i>_2 \in {\cal H}_{{\cal S}_1}$
where the states $|i>_i$ are the quantum states of the
first observer in which she has observed the system
$A$ to have eigenvalue $\lambda_i$.
The point is that there is no contradiction between these two
records of the observation.
The measurement of the second
observer ascertains that the combined system of $A$ and
the first observer is in a correlated state, corresponding to a
measurement having been made. But he is unable to acertain what
the result of the measurement was. In the meantime, the
first observer, having made the measurement, has observed
a definite value $\lambda_1$, and has reduced her description
of the system $A$ to the corresponding eigenstate.
This example shows that it may be possible to have different
Hilbert spaces which describe the results of observations made
by different observers, in a way that allows the different states
to be different from each other, but also consistent with each
other. This example also shows us the way to
a condition of consistency
which must be imposed on the quantum states assigned to different
boundaries. The state $|i_o>$ on the inner surface seen by
the observer must not be precluded by the state
$\sum_i c_i |i>_1 \otimes |i>_2$ seen by the second observer,
i.e. the ampitude $c_{i_0}$ cannot be zero. We will call this
the {\it consensus condition}. Stated formally, it means that
\begin{equation}
Tr_1 \left [ \rho^1 [Tr_2 \rho^{1+2} ] \right ] \neq 0
\end{equation}
where $\rho^1$ is the statistical state the observer has that
represents his information about the system, $\rho^{1+2}$
is the statistical state describing the whole state of the
observer and system and $Tr_2$ means a trace over the
observers degrees of freedom, while $Tr_1$ means a trace
over the system's state space.
\subsection{Other consistency conditions}
There are two other consistency conditions that it is natural
to impose at this stage. The first is that if the surface
is composed of two disconnected parts, ${\cal S}_1$
and ${\cal S}_2$, we must have
\begin{equation}
{\cal H}_{{\cal S}_1 \cup {\cal S}_2 } =
{\cal H}_{{\cal S}_1 } \otimes { {\cal S}_2 }
\end{equation}
and, secondly, reversal of orientation is associated with
Hermitian conjugation, so that,
if $\bar{\cal S}$ is the same surface, with
orientation reversed,
\begin{equation}
{\cal H}_{\bar{\cal S} } = {\cal H}_{{\cal S} }^\dagger
\end{equation}
The states associated
with ``the observer" then correspond to linear functionals of
the states that correspond to ``the system", which is natural
given that the state space is a description of their interface.
A simple objection may be raised to this, which concerns
the case in which the area of the surface is small compard
to the size of the whole universe. In this case the
surface may contain a small observer looking at a large universe,
but might it also not surround a small system being studied with
all the resources available in a large universe?
Thus, in the limit that the surface shrinks down, shouldn't one
of its associated Hilbert spaces get very big, while the other
becomes very small? The answer to this is that were this to happen
it violates our {\it Principle of the impossibility
of an interpretation of a cosmological theory
by an observer outside of the universe.} We must keep in mind
that the state space associated to a boundary
describes the information that an observer
on one side may know about the part of the world in the other
side. As such, when the area of the surface decreases, the dimension
of the associated state space may decrease, representing either the
limited information capacity of the observer, in the case that the
``observer may be seen to be becoming small" or the decreasing
number of degrees of freedom of the ``system" in the case that it
may be becoming small.
However, while it may help to think this way, we should also
remember that there is no guarantee that the volume in one
region or the other is a good physical observable in quantum
gravity.
\section{Postulates of pluralistic quantum cosmology}
In the next section we will turn to the mathematical structure
which is available to implement these ideas. For the present
we may summarize the results of the reasoning up till this
point in a list of postulates, that we would like a quantum
theory of cosmology to satisfy.
{\bf 1) Definition of state spaces:}
A quantum cosmological theory associates to every
oriented three dimensional surface
$\Delta={\cal S} \times R$, with $\cal S$ a compact two surface
without boundary,
a Hilbert space ${\cal H}_{\Delta}^{QG}$, which is a representation
of an algebra of operators ${\cal A}_{\Delta}^{QG}$.
These assignments of Hilbert spaces satisfy the conditions (4)
and (5). Furthermore, for every cobordism $\cal M$, which
is a four manifold with boundaries
$\partial {\cal M}= \Delta \cup \Delta^\prime$ there is a linear
map,
${\cal M}_\Sigma : {\cal H}_{\Delta}^{QG}
\rightarrow {\cal H}_{\Delta^\prime}^{QG}$.
{\bf 2) Consensus condition:}
A {\it pluralistic quantum state of the universe} is an
assignment of a statistical state $\rho_{{\cal S}\times R}$ into
each ${\cal H}^{QG}_{{\cal S}\times R}$, subject to the
{\it consensus condition}, which in diffeomorphism invariant
langauge is expressed as follows. Let the surface $\cal S$
be of the form of two disconnected pieces
${\cal S}= {\cal S}_I \cup {\cal S}_{II}$. Then we know
from (4) that
\begin{equation}
{\cal H}^{QG}_{{\cal S}\times R}=
{\cal H}^{QG}_{{\cal S}_I \times R} \otimes
{\cal H}^{QG}_{{\cal S}_{II}\times R}.
\end{equation}
The consensus condition is then the requirement that
\begin{equation}
Tr_I \left ( \left [ Tr_{II} \rho_{I \cup II} \right ] \rho_I
\right ) \neq 0
\end{equation}
The interpretations of these objects are the following. For every
$\Delta $ we may construct a four dimensional region of spacetime
${\cal W}$ such that $\Delta \subset \partial {\cal W} $.
The region $\cal W$ may contain an
observer, together with clocks and measuring instruments that are
able to measure the values of fields induced on $\Delta$. The algebra
of observables ${\cal A}^{QG}_\Delta$ then must have a classical
limit which corresponds to an algebra of fields in a classical
spacetime induced on a surface $\Delta$ (see {\bf 4}, below.)
A state $\rho_{{\cal S} \times R}$ then corresponds to the information
that an observer in $\cal W$ may have about the fields induced on
$\Delta = {\cal S} \times R$. These, together with knowledge of
the constraints and dynamics, may then be used to infer
information about the values of fields in the region on ``the other side of
$\Delta$".
For example, if ${\cal S}= S^2 \times R$ whose interior
is ${\cal W}= S^3 \times R$, then we may say that an observer who lives
in the ``world-tube" ${\cal W}$ is able, by means of measurements of fields
on the boundary $S^2 \times R$, to gain information about the
universe external to the boundary. This information is then recorded
in the value of a statistical state $\rho_{S^2 \times R}$.
{\bf 3) No interpretation in terms of external observers:}
${\cal H}_{{\cal S} \times R}^{QG} \rightarrow C$, the complex
numbers, in the limit that
the two surface $\cal S$ shrinks to a point.
{\bf 4) Incorproration of quantum gravity:}
${\cal A}_{\Delta}$ must
have a subalgebra ${\cal A}_{\cal S}$, which includes observables
that measure the spacetime metric $g_{ab}$ and left handed
spin connection $A_A^{AB}$ pulled back into the
two surface ${\cal S}$. (These will be denoted $h_{\alpha \beta}$
and $a_{\alpha}^{AB}$ respectively.)
{\bf 5) The Bekenstein condition:}
Let ${\cal H}_{{\cal S}, h}$ be the subspace of
${\cal H}_{{\cal S} \times R}^{QG}$ containing eigenstates
of the two metric on $\cal S$ with eigenvalues denoted by
$h$.
Then, if $A(h)$ is the area of the two surface metric we
require,
\begin{equation}
dim ({\cal H}_{{\cal S}, h}) = e^{c A(h)/l_{Pl}^2}
\end{equation}
where $c$ is a fixed dimensionless constant.
Finally, we require an axiom of correspondence with linearized
quantum field theory, which may be formulated as follows,
{\bf 6) Correspondence with linearized QFT:}
Let us choose a cosmological constant $\Lambda$ (which is
assumed one of the paramters of the theory), an area $A_0$
and an ultraviolet cutoff $r$ such that
\begin{equation}
l_{Pl}^2 << r^2 << A_0 << \Lambda^{-1/2}
\end{equation}
Then, let ${\cal H}^{lin}_{\Lambda , A, r, \epsilon }$ be the
Hilbert space of
linearized gravitons (and other massless fields, if desired) restricted
to a region of
DeSitter spacetime with cosmological constant $\Lambda$ bounded
by a three boundary $S^2 \times R$, with the metric induced on
the two sphere being a two sphere metric with area $A$ by the
condition that the linearized fields vanish at the boundary. Further,
in the reference frame in which the boundary is static two conditions
are put on the spectrum. First, the energy
of the individual gravitons and other massless quanta
must be less than the inverse cutoff $r^{-1}$. Second, the total
energy (defined in terms of the quasi-local surface integral
$H=\int_{S^2} \kappa $) must be bounded by
\begin{equation}
E < \epsilon \sqrt{A}
\end{equation}
with epsilon much less than one. Let
${\cal A}^{lin}_{\Lambda , A, r, \epsilon }$ be an algebra of
observables associated with this space.
The meaning of this space and algebra is they define the linearized
limit of any quantum theory of gravity with small cosmological
constant, restricted to the interior of a static surface.
Two cutoffs are needed, one to limit the energy of any
single graviton to less than the Planck scale, the other to limit
the total energy to be much less than that at which gravitational
binding energy, and the collapse to a black hole, would be relevant.
For the quantum theory of cosmology to have an acceptable
linearized quantum field theory limit we require that here
exists a family of maps,
\begin{equation}
{\cal N}_{\Lambda , A, r, \epsilon } :
{\cal H}^{lin}_{\Lambda , A, r, \epsilon } \rightarrow
{\cal H}_{S^2 \times R}^{QG}
\end{equation}
\begin{equation}
{\cal N}_{\Lambda , A, r, \epsilon } :
{\cal A}^{lin}_{\Lambda , A, r, \epsilon } \rightarrow
{\cal A}^{QG}_{S^2 \times R}
\end{equation}
such that
\begin{equation}
<{\cal N} \circ \Psi |{\cal N} \circ
{\cal O} |{\cal N} \circ \Psi >^{QG} =
<\Psi |{\cal O} |\Psi >^{lin} + O(r/l_{Pl} ) + O(\epsilon ) +
O(l_{Pl}^2 /A )
\end{equation}
This guarantees that there is a subset of the physical states and
observables whose physical expectation values reproduce the
predictions of the linearized theory.
To proceed to investigate the reasonableness of these postulates
we must ask if it is consistent with the physical interpretation of
both general relativity and quantum theory, especially given the
special problems associated with the notion of time. This is the
subject of the next section.
\section{The problem of time}
One advantage of the proposal just made is that it may allow a new
possibility for the resolution of the problem of time, which has
plauged attempts so far to construct a quantum theory of
cosmology.
To introduce the basic idea, let me recall some of the proposals to
resolve the problem of time in quantum cosmology. One, due to
DeWitt\cite{bryce-time} and developed and
advocated by Rovelli\cite{carlo-time},
is to use the
Heisenberg picture and construct physical operators that
describe correlations between certain degrees of freedom,
associated to clocks and other degrees of freedom. In the
classical theory, it is certain that there are a large set of physical
observables (that is functions on the kinematical phase space
that commute with the constraints under the Poisson brackets)
which are what Rovelli calls ``evolving constants of motion."
It then may be that operators can be constructed in the quantum
theory that have the same meaning (although, of course there are
daunting technical problems in doing this for a realistic field theory
rather than for an integrable system.)
However, even if these technical problems could be overcome,
there is an important conceptual problem, which is how the
expectation values of these operators are to be correlated with
observers made by us observers inside the universe.
If we could imagine that there are observers outside of the
universe, who make observations of the universe, and are able
to many times prepare and measure the state of the whole
universe, they would
be able to verify whether or not the expectation values of these
``evolving constant of motion operators" in a particular physical
state are in accord with their observations. But, this is something
we, as observers in the universe, are unable to do.
If we insist on our principle that there should be no possibility
of speaking in terms of an observer outside of the universe, then
Rovelli's proposal, even if it works technically, would not lead
to predictions that could be checked by us observers inside
the universe.
The same situation holds in other proposals concerning the
problem of time in quantum cosmology such as the proposal
of Page and Wootters\cite{pagewooters},
and the proposal of \cite{julian-qc}. In the
first, one imagines that the Hilbert space is a direct
product
\begin{equation}
{\cal H}= {\cal H}^{clock} \otimes {\cal H}^{other}
\end{equation}
of a space representing a clock degree of freedom and
a space representing other degrees of freedom.
Physical quantum states states describe correlations
between certain degrees of freedom considered to be
clocks and the other degrees of freedom, coded in the
entanglement of the state, as in
\begin{equation}
|\Psi > = \sum_i c_i |t_i>^{clock} \otimes |\chi_i >^{other} .
\end{equation}
These correlations are understood to be induced by
the constraints, which impose the dynamics. The different
possible physical states then correspond to the different
sets of correlations that are possible between the clock
and the other degree of freedom, given the dynamics.
Again, we have the problem that the predictions of the existence
of such correlations could only be detected by an observer
outside the system. Even if we accept the many worlds
point of view, which Page and Wooters seem to do, an
observer inside the system can only verify the presence
of one branch, but they cannot give meaning to the coefficients
$c_i$ nor they can tell if the theory gives the right relative weights
to the different correlations.
Conversely, in the Barbour picture, the notion of a
single clock degree
of freedom is given up in favor of the more realistic picture that
time is measured from the coincidences among all the degrees of
freedom in a complex universe. Instead,
the ``wavefunction of the universe" is interpreted to give directly
the relative frequency for the occurance of instantaneous
configurations in a great collection of such configurations. The
collection of such ``moments" is taken to constitute reality; any
impression of time comes from the existence of memories, records
and what Barbour calls ``time capsules" in complex
instantaneous configurations.
The problem is again that only an observer outside of the universe
can give a meaning to these relative frequencies for ``moments"
in Barbour's conjectured reality.
Of course, all of these proposals allow that classical general
relativity coupled to quantum fields is
recovered if the state is appropriately ``semiclassical". But
while a necessary condition for an interpretation, this is not
sufficient, as a good quantum theory of cosmology should in
principle give predictions for all states, not only those that
are semiclassical. Otherwise, its predictions may not differ
from those of semiclassical gravity, perhaps together with
the addition of some particular boundary condition.
From the present point of view, the problem which all these
proposals have in common can be resolved
in the following way. The physical quantum state is now
associated to a three dimensional surface, which is understood
to represent the evolution in time of the two dimensional
boundary of a part of the universe. The description can
be completely invariant, with respect to diffeomorphisms both
in and inside of the surface. Thus, time must be understood in
terms of correlations between degrees of freedom inside the
surface, whether these are expressed in Barbour's, Page and
Wooter's or Rovelli's pictures. But there are in fact observers
outside of the system, they live just outside the boundary.
Furthermore, if we use the statistical interpretation of the
quantum state, then the statistics predicted from the
expectation values of operators are interpreted in terms of
the relative frequency probability for what correlations between
clocks and observers they will see if they observe the system
on their surface.
For example, if the state is of the form (15) then the $|c_i|^2$
are the probability that if they observe the system they will
find that the clock reads a time $t_i$ while the rest of the
system is in a state $|\chi_i >$ (assuming for simplicity the
trivial inner product.) They can in principle check the
predicted probabilities by
preparing the part of the universe inside the boundary many
times in the state $|\Psi>$ and repeating the experiment.
Suppose our observers
do the experiment in sequence many times, will
they record a sequence of times that are in cronological order?
Will the times they measure for the clock in the system be
synchronized with clocks they may carry outside of their
boundary? To answer such questions we cannot just rely
on the quantum description of the observers themselves.
Such questions can only be answered in the quantum theory
by enclosing the observers and the system inside a second
boundary ${\cal S}^\prime \times R$ and constructing a Hilbert
space to represent the combined system of the first system and
its observer. The state of this combined system then describes
the information that a second observer, outside the second
boundary can have about it. This observer can make observers
of the combined system of the first observer plus the first system,
and by doing so answer questions like these.
In this context as well, the consensus condition (3) is essential
to guarantee that the probabilities actually seen by the first
observer agree with the $|c_i|^2$ in the expansion of the state
seen by the second observer.
For, by the theorem of
Finkelstein\cite{finkelstein-prob} and Hartle\cite{hartle-prob}
it guarantees
that in the limit of a large number of observations, these two
probabilities will agree. Thus, in the present context, this
theorem plays an essential role, but one rather different than
in the many worlds interpretation. Rather than giving a meaning
to probability it guarantees that the probabilities for events
seen by the different observers are consistent with each other.
Thus, it may be said that this interpretation is giving meaning to
Bohr's statement that the split between the classical and quantum
world can be made arbitrarily\cite{bohr}. For it allows us to consider
a description that is given by specifying simultaneously the states
associated with all such possible splittings. It is very interesting
that to use this interpretation to resolve problems in quantum
cosmology such as the problem of time we must use some of the
results associated with the many worlds interpretation. However,
as in the case of the theorem of Finkelstein and Hartle, their role
is transformed from giving a meaning to an interpretation in
terms of one state and many worlds, to ensuring the consistency of
an interpretation in which many states describe many partial
observations of one world.
\section{The classical limit and the timelike initial data problem}
If the picture we are describing is to work, it must have a classical
limit which makes sense in terms of classical general relativity.
We must then ask whether there is a formulation of classical
general relativity in the spacially compact case in
which the theory may be described in terms
of fields induced on a timelike surface $\Delta={\cal S} \times R$.
This is equvalent to posing the
{\it timelike initial data problem} \cite{timedata}.
This is the question of whether there is data, subject to constraints,
that may be given on such a surface $\Delta$, such that the three
metric given on $\Delta$ has signature $(-,+,+)$ and such that
the data determines a solution to Einstein's equations, at least
in a neighborhood of the surface.
The timelike initial data problem, along with its cousin, the
null initial data problem, is not as well studied as the conventional
initial data problem in terms of spacelike surfaces. But, it has
often been observed that the Hamiltonian formalism does not
require that the surface on which the fields and momenta are
derfined be spacelike. Thus, at least
formally, one can construct a formalism for first order evolution
of data, subject to constraints, from a timelike initial data surface.
In this formalism, momenta correspond to derivatives off of the
surface, and there are constraints, which are of the same form
as the usual constraints, with certain changes of sign due to the
change in the signature of the induced three metric.
This formalism will be discussed in detail
elsewhere\cite{inprep}. For the
present, we need only know that at the formal level the
Hamiltonian formalism can be translated to a timelike
surface. It is also interesting to note that in this case
it is the diffeomorphism constraints
that induces the correlations between degrees of freedom on the
surface that might be used to measure time and other degrees of
freedom. The role of the Hamiltonian constraint is then to develop
the solution into the interior, and thus establish correlations between
fields on the surface and fields in the interior.
\section{Incorporating topological quantum field theory
and quantum gravity}
Now that we have seen that the problem of time, as well
as the classical limit, seem to pose no problems of principle,
we may turn to the question of trying to realize the
postulates made in section 5 in terms of a concrete
mathematical construction.
The proposal made above relies on an apparant coincidence,
which is that
both the Bekenstein
bound and the point of view about quantum cosmology proposed
in the section 4 lead to the same conclusion, which is that
quantum gravity, in the cosmological case, may
be defined in terms of state spaces
attached to three dimensional ``timelike"
surfaces which represent the
information that observers who measure fields induced on the
surface may have about the degrees of freedom
in its interior. Thus,
it is very natural to synthesize the two lines of development,
which is what we have done in formulating these postulates.
The question that must now be posed, however, is how
these postulates are to be realized in a concrete theory.
In the following sections I will present one proposal for
how to do this, which is based on a further
apparent coincidence.
In order to proceed the main question that must be answered
is how the finite dimensional state spaces, which are the
eigenspaces of the two metric operators are to be realized.
The hope to realize this and the other conditions comes from
an apparent coincidence of two results, one from non-perturbative
quantum gravity and one from topological quantum field theory.
The first is the discovery that in non-perturbative quantum gravity
the operators that measure areas of regions, $\cal R$ of
any two dimensional surface $\cal S$, which may be
denoted $A[{\cal R}]$, have discrete spectra.
To say this more precisely, the basic result is that
the quantum states of a diffeomorphism invariant
quantum field theory in a region $\Sigma$ are given by the
diffeomorphism classes of
the spin networks in $\Sigma$ \cite{volume,spinnet-us}.
The spin networks
are the eigenstates of $A[{\cal R}]$ and the corresponding
eigenvalues are given
by the spins of the edges of the network that intersect
the surface at points in the region $\cal R$. Thus,
if $|\Gamma >$ is the quantum state associated to the
spin network $\Gamma$ in $\Sigma$ and $y_\alpha^{\Gamma, {\cal R}}$
are the points at which the spin network meets the region
$\cal R$ of the boundary, and $j_\alpha$ are the spins of
the edges that meet
boundary, we have
\begin{equation}
\hat{\cal A}[{\cal R}] |\Gamma > = l_{Pl}^2 \sum_\alpha
\sqrt{j_\alpha (j_\alpha +1 )} |\Gamma >
\end{equation}
One way to measure the two metric, up to diffeomorphisms, is
to measure the $A[{\cal R}]$ for all the
${\cal R} \subset {\cal S}$. We see then that the two
metric can be considered to have a discrete spectrum.
The simultaneous eigenspaces of the $A[{\cal R}]$
are labled by punctures $y_\alpha$ marked by spins
$j_\alpha$ which label representations of $SU(2)$.
Thus, the eigenspaces ${\cal H}_{{\cal S}, h}^{QG}$, which
should have finite dimension, should more precisely be
labled ${\cal H}_{{\cal S}, y_\alpha , j_\alpha}^{QG}$.
As a result, there are no problems with continuum measures in
the statement of postulate {\bf 5}, so that we may write,
\begin{equation}
{\cal H}_{{\cal S} \times R}^{QG} = \oplus_n \oplus_{j_\alpha}
{\cal H}_{{\cal S}, y_\alpha , j_\alpha}^{QG}
\end{equation}
This is a step, but it does not yet explain to us how to
construct state spaces that are eigenspaces of these operators
that have finite dimension that grows exponentially with the
area. To see how to do this, we must turn to some basic
results of topological quantum field theory.
Before explaining how this problem is solved, though,
we must divert to discuss
a remarkable observation of Louis
Crane, which is that the axioms of topological quantum field
theory in three dimensions are exactly what is needed to
guarantee the consistency of a many-Hilbert space
reformulation of quantum theory of the type we have
been discussing\cite{louis}. This is because by the axioms of
Atiyah\cite{atiyah}, TQFT's automatically assign state spaces to boundaries,
so that the conditions (4) and (5) are satisfied. Moreover, the
posulate {\bf 3)} is {\it automatically} realized as a result
of the axioms.
For these and other reasons, described in his papers, Crane
has advocated the construction of a quantum theory of
gravity as an extension of topological
quantum field theory.
The proposal I will shortly make is indeed an attempt to
realize this program\footnote{Other attempts to realize
quantum gravity as an extension of TQFT are described
in\cite{louis,louisdavid,louisigor,baezdolan,rjl,jb}.}.
Furthermore, topological quantum field theory naturally
extends to the case of two dimensional
surfaces with points marked by
representations of groups\cite{louis2d3d} and spin networks play
a natural role in this formulation. Among the axioms of
$TQFT$ one finds that a three dimensional $TQFT$
associated to a quantum group $G$ is a functor that
assigns\cite{atiyah,louis}:
\begin{itemize}
\item to every two dimensional surface $\cal S$ with
$n$ punctures $y_\alpha$, $\alpha=1,...,n$ labled with
representations $j_\alpha$ of $G$ a finite dimensional
Hilbert space ${\cal H}^G_{{\cal S},j_\alpha}$.
\item to every three manifold $\Sigma$ with imbeded
$G$-spin network $\Gamma$, such that
$\partial \Sigma = {\cal S}$ and $\Gamma$ intersects
$\cal S$ in the $n$ labled points $y_\alpha$, with the
line intersecting $y_\alpha$ being labled by the same
representation $j_\alpha$, a state
$|\Gamma , \Sigma > \in {\cal H}^G_{{\cal S},j_\alpha}$.
\end{itemize}
Note that this is defined to be a functor from the category
of manifolds, with cobordims as maps to the category of
Hilbert spaces, with linear maps as the maps. This
guarantees that the conditions (4) and (5), as
well as postulate {\bf 3)} are automatically
satisfied.
Thus, it seems that topological quantum field theories
are exactly the objects we need to construct the
finite dimensional state spaces that represent
eigenspaces of ${\cal H}_{{\cal S}, h}$. More precisely,
if we use the coincidence that spin networks and points
marked with representations of a group appear in both
$TQFT$ and quantum gravity
we may then make the hypothesis that there is a
topological quantum field theory associated with $SU(2)$,
to which we may identify these spaces, so that
\begin{equation}
{\cal H}_{{\cal S}, y_\alpha , j_\alpha}^{QG}
= {\cal H}_{{\cal S}, y_\alpha , j_\alpha}^{TQFT}
\end{equation}
There is a natural candidate for this TQFT,
which is the Chern-Simons theory associated with the
group $SU(2)$. The variable there, which is an $SU(2)$
connection, is exactly what we need to represent
$a_\alpha^{AB}$, the pull back of the Ashtekar connection
to $\cal S$.
If we are to try to realize this hypothesis, however,
there are three
questions we must answer.
\begin{itemize}
\item Can we give an interpretation, in terms of
the parameters of quantum gravity, to the coupling constant,
or level, $k$ of the Chern-Simons theory? This involves two
related puzzles, first that it
seems to be ordinary $SU(2)$ spin networks that play
a role in non-perturbative quantum gravity, while it is the
q-deformed networks that are relevant for TQFT, and second that
the commutation relations of the two theories are different.
\item Can the constraints of quantum gravity be represented, or
its solutions expressed, in terms of the states of the
Chern-Simons theory?
\item Is the Bekenstein bound satisfied?
\end{itemize}
It turns out, as I will describe in the next section, that there
are natural answers to the first two questions and that,
furthermore, the answer to the third question is yes.
\section{A proposal}
From the side of quantum gravity, we seek a theory that provides
a representation for the observables that will be induced in
a three dimensional timelike surface, $\Delta = {\cal S} \times R$,
where $\cal S$ may be considered an intersection of $\Delta$
with a surface of constant ``time". We may then, to make a
connection
with the standard $3+1$ canonical formalism consider the algebra
of observables induced in the two surface $\cal S$.
These include the pull backs into $\cal S$ of the
curvature two form $F_{ab}^{AB}$ and the dual of the
densitized from field $E_{ab}^{AB} = \epsilon_{abc} \tilde{E}^{c AB}$
which I will denote by $f^{AB}$ and $e^{AB}$,
respectively\footnote{$A,B,...$ stand for
two component spinor indices. We will also choose conventions
in which $f^{AB}$ and $e^{AB}$ have dimensions of $length^{-2}$
and $length^o$, respectively.}. We may
note that these all commute with each other under the standard
Poisson
brackets of general relativity.
In ref. \cite{linking} I studied the case in which $\Delta$ was a
boundary
of spacetime, to which we needed to associate particular boundary
conditions
as well as a boundary term in the action. There I studied a
particular
condition called the ``self-dual" boundary condition, in which
\begin{equation}
f^{AB}= {2 \pi \over k G} e^{AB}
\end{equation}
where $k$, was required by gauge invariance to be an integer, and
was
as well related to the parameters of general relativity by
\begin{equation}
k = {6 \pi \over G^2 \Lambda} + \alpha
\end{equation}
where $G$ and $\Lambda$ are Newton's constant and the
cosmological constant,
respectively and $\alpha$ is a $CP$ violating parameter coming from
an
$\int F^{AB} \wedge F_{AB}$ term in the action.
This has the effect of inducing a topological quantum field theory
in the boundary, which was in fact $SU(2)_q$ Chern-Simon theory,
with
$k$ the level, or quantum deformation parameter. This is possible
because an additional effect of imposing the boundary condition (19)
is that the algebra of observables in $\cal S$ was deformed, so that
the connection one form $a_\alpha^{AB}$, which is the pull back of
the Ashtekar connection $A_a^{AB}$ into $\cal S$ satisfies,
\begin{equation}
\{ a^{AB}_\alpha (\sigma ) , a^{CD}_\beta (\sigma^\prime ) \} =
{2 \pi \over k} \epsilon_{\alpha \beta } \delta^2
(\sigma , \sigma^\prime )
(\epsilon^{AC}\epsilon^{BD} + \epsilon^{AD}\epsilon^{BC})
\end{equation}
which are in fact the Poisson brackets of Chern-Simons theory.
As a result, a Hilbert space that represents this
algebra of observables, subject to the condition (19)
is given, for each choice of the boundary $\cal S$ by
the direct sum,
\begin{equation}
{\cal H}^{gr}_{\cal S} = \oplus_{n=1}^\infty \oplus_{j_1 , ... , j_n}
\int d^2y_1 ... d^2 y_n
{\cal H}^{CS, k}_{{\cal S}, y_\alpha , j_\alpha}
\end{equation}
where $y_\alpha$ with $\alpha = 1, ...n$ are the positions of
$n$ punctures on the surface $\cal S$,
the $j_\alpha$,are $n$ labels of representations
of $SL(2)_q$ (or quantum spins) located at $n$ punctures on the
surface $\cal S$. By standard constructions in TQFT there is a finite
dimensional Hilbert space associated with each choice of
${\cal S},n , j_\alpha $ and $k$.
A question that was left open in \cite{linking} was whether the
diffeomorphisms of the two surface $\cal S$ can be imposed as a
gauge
symmetry. There seem to be technical obstacles to doing so, but
there is also reason to believe they may be overcome. In this case
the state space associated to the surface would simply be
\begin{equation}
{\cal H}^{gr}_{\cal S} = \oplus_{n=1}^\infty \oplus_{j_1 , ... , j_n}
{\cal H}^{CS, k}_{{\cal S} , j_\alpha}
\end{equation}
because, of course, the Hilbert space of the TQFT does not depend on
the positions of the punctures.
In reference \cite{linking} the conjecture was then made that the
algebra of observables on the boundary was sufficient to label the
quantum states of the interior. While this is not yet shown, there
is some evidence for it, which was discussed there. If this is the
case, then the physical state space of quantum gravity, in the
presence of the self-dual boundary conditions, would then have
the form (22).
In the case we are discussing here, the two
surface plays a somewhat different role. Rather then being a
boundary
of space, at which certain boundary conditions may be imposed, it is
now to be thought of as simply an arbitrary two dimensional surface
in
space. The question is whether any of the structure discovered for
this particular set of boundary conditions is at all relevant to this
case.
To investigate this question, let us ask what replaces the self-dual
condition (19) in the general case? We may recall that, in the case that
the field equations (or at least their pullback,
into $\cal S$) are satisfied, there is a relationship between the fields
$f^{AB}$ and $e^{AB}$ in $\cal S$, which is given by
\begin{equation}
f^{AB} = \left (\Psi^{ABCD} + {G \Lambda \over 3}
\epsilon^{AB} \epsilon^{CD} \right ) e_{CD}
\end{equation}
where $\Psi^{ABCD}$ is a totally symmetric spinor, representing the
spin-two degrees of freedom of the gravitational field. This
relationship can be seen both from the $CDJ$ formalism\cite{CDJ} and the
Newman-Penrose equations\cite{tedroger}. We may note that this equation
constitutes
in fact the general solution to the Hamiltonian and
diffeomorphism constraints, or equivalently the frame field
field equations, pulled back into any three surface that contains
$\cal S$.
This equation can also be understood to be an expression of the
hamiltonian and diffeomorphism constraints. As pointed out by
Reisenberger\cite{mikerconstraints},
one can express the constraints coming from
the CDJ action by the statement that there must exist a symmetric
spinor $\Psi^{ABCD}$ such that (23) is satisfied. This may have
some advantages over the standard, Ashtekar, form of the constraints,
as its derivation makes no assumption of the invertibility of the
metric.
How are we to represent this relationship quantum mechanically? It
is
clear that what the equation is saying is that in general the
two fields $e^{AB}$ and $f^{AB}$ are free to some extent to vary
independently, as long as they are related by (24). This means that
they can differ by the action of two terms, there is a contribution
to $f^{AB}$ which is proportional to $\Lambda$ times $e^{AB}$ and
there is a second term which is proportional to the action of a spin
two field.
To translate this quantum mechanically let us note that in the case
that the spin two field is absent they are strictly proprtional to
each other. In this case, how is the equation
(19) represented quantum mechanically?
In the loop representation, $e^{AB}$ will
be represented by a set of punctures, which are the intersections of
a spin
network $\Gamma$ with the two surface $\cal S$. These punctures
are
labled by representations of $SU(2)_q$, which are those carried by
the
lines of the spin network at the intersections. If
$f^{AB}$ is, in this
limiting case, to be proportional to $e^{AB}$, then it must also be
represented by the same set of punctures, labeled by the same
representations.
This is what happens when we impose the self-dual condition (19),
as described in \cite{linking}. The
result
is that, under these conditions, when we measure the metric of the
two
surface (and hence $e^{AB}$), we pick out an eigenstate of the
operators
that measure elements of area in the surface, and this picks out a set
of
punctures and representations. Because of the proportionality of
$e^{AB}$ and $f^{AB}$, the result is that this picks out a two
dimensional quantum field theory, which is $SU(2)_q$ Chern-Simons
theory with that choice of punctures and representations. The states
of this theory represent the freedom that the connection,
$a^{AB}_\alpha$,
still has, once the condition (19) is imposed.
Now let us consider the case that the spin two field $\Psi^{ABCD}$ is
non-zero, so that the self-dual condtion (19) is relaxed to (24). In
this case we may represent the situation as follows. The
representation
of the metric observables $e^{AB}$ should be the same, so these
will have eigenstates, associated with the intersections of spin
networks
with $\cal S$, which are hence labled by choices of
punctures, $y_\alpha$ and representations $j_\alpha$.
How are we to represent the degrees of freedom of the
connection $a^{AB}_\alpha$, now subject to the constrant (24)? Let
us represent this as before by the Hilbert space of $SU(2)_q$
Chern-Simons theory, with the same set of punctures as before.
This is in agreement with (24), which tells us that
in regions over which the integral of $e^{AB}$ vanishes,
the integral of $f^{AB}$ must vanish as well. However, when
it is allowed to be nonvanishing, let us
allow for the connection to vary independently by letting the spins,
which we will call $l_\alpha$ at these punctures vary.
Thus, the
degrees of freedom of the quantum theory will be punctures,
$y_\alpha$, each of which is labled by two spins, $j_\alpha$
and $l_\alpha$ corresponding to the separate metric and connection
degrees of freedom. However, the metric and connection are not
completely free to vary independently, because they are constrained
classically by (24). Since $\Psi^{ABCD}$ is a spin two field, this means
that quantum mechanically, each $l_\alpha$ must either be
proportional
to the corresponding $j_\alpha$, or must be in the decomposition of
the
multiplcation of $j_\alpha$ with the spin-two representation.
That is, the quantum mechanical version of (24) may be hypothesized
to be,
\begin{equation}
l_\alpha \in ( 2 \otimes j_\alpha , j_\alpha )
\end{equation}
(Of course $j_\alpha$ is already contained in the decomposition
of the product of itself with spin $2$, but for clarity, because it
comes from a separate term, we list it separately.)
The result is that the state space of quantum gravity, associated with
the two surface $\cal S$ now becomes,
\begin{equation}
{\cal H}^{QG}_{\cal S} = \oplus_n \oplus_{j_\alpha }
{\cal H}^{QG}_{{\cal S}, j_\alpha}
\end{equation}
where the ${\cal H}^{QG}_{{\cal S}, j_\alpha}$ are the eigenspaces of
${\cal A}[{\cal S}]$, the area of $\cal S$, with eigenvalues
$\sum_\alpha \sqrt{j_\alpha (j_\alpha +1 )}$ (neglecting corrections
in $1/k$ \cite{sethlee-qnet}.) Each of these eigenspaces
is then further decomposed,
\begin{equation}
{\cal H}^{QG}_{{\cal S}, j_\alpha}=
\oplus_{l_\alpha \in ({2} \otimes {j}_\alpha ,j_\alpha )}
{\cal H}^{CS, k}_{{\cal S} , l_\alpha}
\end{equation}
Having specified the state space of the theory we must now discuss
the role of the parameter $k$. We may first note that $k$ must
be considered to be a parameter of the theory as it comes into the
specification of the observables algebra, in (21). We can see from
that equation that $k^{-1}$ not vanishing indicates that there is
a kind of anomolie in the theory, so that the classical commutation
relations have been modified so that all components of $A^{AB}_a$
no longer commute. As is shown in \cite{linking} this turns out
to be related to the necessity to frame the loops that go into
the spin network constructions, so that a diffeomorphism
in $\cal S$ that has the effect of twisting a line of a spin
network (or rotating a puncture) has the effect of multiplying
the states of the theory in the spin network basis by a phase.
We may note that there are reasons to believe that this is a real
effect in the case of non-vanishing cosmological constant,
which comes from the attempts to define states in the
loop representation that would correspond to the loop transform of
the
Kodama state\cite{kodama}. Since the anomolie is proportional to
$k^{-1}$, it is consistent to take $k \approx \Lambda^{-1}$, as
equation (20) asserts. To fix the value of $k$, we can use the
arguments from \cite{linking}, which apply, however, to the case that
$\Psi^{ABCD}$ is constrained to vanish, yielding eq. (20).
Alternatively, we can use arguments coming from the attempts
to define the path integral involved in taking the loop transform
of the Kodama state, which lead to the same
conclusion\cite{ls-cs,BGP-cs}.
It would,
however,
be better to have a more general argument, independent of either a
restriction of the degrees of freedom at the surface, or the choice
of a particular state. Thus, at present, it is perhaps best to
take (20) over to
the present case as a working hypothesis.
Further, we may note that as (24) is the general solution to the
Hamiltonian
and diffeomorphism constraints, what we have expressed quantum
mechanically
through (25) is in fact a parametrization of the space of solutions
of the constraints. Further, the fact that the frame field has been
expressed in terms of spin networks (leading to the punctures)
means
that the Gauss's law constraint has been solved. We may then take
(27) to be a description of the physical state space of the theory.
Finally, we may note that the Bekenstein bound is in fact satisfied
in this case, at least in the limit of large $k$, or small
$G^2 \Lambda$. The bound will hold
(in the limit $k \rightarrow \infty$)
if there is a fixed constant
$c$ such that
\begin{equation}
\ln dim {\cal H}^{QG}_{{\cal S},j_\alpha} \leq c
\sum_\alpha \sqrt{j_\alpha (j_\alpha +1 )}
\end{equation}
for all choices of punctures and labels. However, in the same limit
the dimension of ${\cal H}^{QG}_{{\cal S},j_\alpha}$ is just the
product of all dimensions of
the allowed representations of $SU(2)$ at each
puncture\cite{witten-tqft}. Thus, we have,
\begin{equation}
\sum_\alpha \sum_{l\in (2 \times j_\alpha , j_\alpha )}(2l+1 )
\leq c
\sum_\alpha \sqrt{j_\alpha (j_\alpha +1 )}
\end{equation}
It is not difficult to discover that this will be true for all
$c \geq 28/\sqrt{3}$.
We have thus achieved our goal, which is to postulate a starting
point
for a quantum theory of gravity that is consistent with
the results we want to keep from previous investigations, including
the kinematical structure of the loop representation. Further,
we have the theory in a language which is suitable for pluralistic
quantum theory and, using the connection between spin networks,
as they represent eigenstates of the area operator in quantum
gravity,
and spin networks as they label states in topological quantum field
theory,
we have found a way to naturally implement the Bekenstein bound.
This
means that we have a field theory, associated to two dimensional
surfaces,
that implements the holographic hypothesis, and implements
naturally
the condition that the subspaces of the state space that are
eigenspaces of the metric of the boundary are finite dimensional,
with the dimension of the boundary growing exponentially with the
physical area.
\section{Conclusions}
It need hardly be said that the present theory is put forward
with all due reserve. What I have described here is only
a sketch of a research
program. There are several directions in which the proposals outlined
here might be pursued. The theory sketched in the previous section
needs to be developed by the introduction of additional observables,
correspoding to components of the metric, connections and curvatures
which are not tangent to the surfaces. It is possible, following
the results of section VIi of \cite{linking},
that this will enable the full tangle
algebra of Baez \cite{john-tangle} to be represented.
At the same time, one may try to sharpen the system of postulates
proposed here to a system of axioms that could be taken as the
foundations of a quantum theory of gravity. The question in this
case is whether a system of axioms can be found which allows
both classical general relativity and quantum field theory on
fixed backgrounds to be derived in the appropriate limits.
In this respect, a
very interesting possibility is that Jacobson's recent
construction\cite{ted-new}
can be used to derive the Einstein equations directly
from thermodynamics
together with a set of postulates about the fundamental
quantum theory, such as those proposed here.
According to Jacobson's construction all that
is needed to derive the Einstein's equations as the classical
limit of a quantum theory of gravity are three assumptions:
1) the laws of thermodynamics 2) the recovery of linearized
quantum field theory in restricted regions whose curvatures
are small in Planck units and 3) the connection between
entropy and surface area. The idea is that 2) may be guaranteed
by the last postulate while 3) is guaranteed by the Bekenstein
condition (8). We may expect that
the laws of thermodynamics are to be recovered
from the usual statistical assumptions, given that the use of
ordinary, relative frequency statistics is guaranteed by the
consensus condition. Whether this
can be carried out in detail is a very interesting question; if it can
then the conclusion would be that in a quantum theory of gravity
the fundamental axiom that introduces gravitation is the connection
between information and entropy, from which Newton's laws of
gravity would follow as consequences of the recovery of general
relativity in the classical limit. This reverses the conventional
point of view in which the law of gravitation, expressed in terms
of Einstein's equations is taken to be fundamental, with the
connection between information and surface area a consequence.
In this respect, the key result from canonical quantum gravity
that is being carried over into this context is the connection
between spin, labeling the spin networks that are the
kinematical quantum states, and area.
Finally, on the conceptual side, the proposal made here concerning
the interpretation of quantum theory has certain implications that
need to be investigated. As presented here, the quantum states
are taken to describe information held by, or available to,
certain observers. This certainly leaves open the possibility that
the quantum theory might turn out to be the statistical mechanics of
a more fundamental theory such as a non-local hidden
variables theory\cite{me-hidden}. For reasons that have been often
discussed, any departure from the physical postulates of the quantum
theory is likely to be cosmological in
origin\cite{forabner,lotc,me-hidden,spacetime}.
It may thus be hoped that, even if quantum mechanics, expressed in terms
of linear state spaces and operators, is not a fundamental theory, a
formulation of quantum theory that sensibly extends to a theory of
cosmology may be a stepping stone to a deeper theory.
\section*{ACKNOWLEDGEMENTS}
This work grew out of many years of collaboration and
conversations with Louis Crane and Carlo Rovelli. I am grateful to
Louis Crane for insisting on the usefullness of the categorical
point of view and
to Carlo Rovelli also for the suggestion that quantum gravity might
be formulated axiomatically. I must thank also David Alpert, Abhay
Ashtekar,
John Baez, Julian Barbour, Steve Carlip
Roumen Borissov, Mauro Carfora, Ted Jacobson,
Louis Kauffman,
Seth Major, Roger Penrose, Jorge Pullin, Michael Reisenberger,
Simon Saunders,
Chopin Soo, Leonard Susskind and
Gerard 't Hooft for conversations and
suggestions. Parts of this work were carried out during
visits to the Institute for Advanced Study in Princeton and SISSA
in Trieste, which allowed good conditions for working. A visit
to Lumine, for which I have to thank Daniel Kastler,
made possible very stimulating conversations with Louis Crane and
Carlo Rovelli. The opportunity to present
this proposal to the conference in Bielefeld on ``Quantum theory
without observers" was also most helpful.
This work
has been supported in part by the NSF under grant PHY-93-96246
to Pennsylvania State University.
|
\section*{Introduction}
\subsection*{ Motivation.}
The heuristic meaning of quantum group theory consists in the consideration
of transformations
that depend on noncommuting parameters.
The parameters serve as generators of a certain (noncommutative)
algebra, which is interpreted as the
``function algebra" of the quantum group. The group nature of the
transformations is formalized in the assumption that this algebra is endowed
with the Hopf structure.
In matrix case the Hopf structure is induced just by the standard matrix
multiplication and by taking matrix inverse.
But the transformations of the group nature (and the related algebraic
structures) are not the only possible. Quantum semigroups appear along with
quantum groups (at least as the intermediate product, see
~\cite{rtf},~\cite{dem2},~\cite{dem3}). Multivalued Hopf algebras were defined
by Buchstaber and Rees~\cite{buch2}. For Novikov's operator doubles (which
are not Hopf algebras) the ``Hopf type'' questions can be studied
~\cite{nov},~\cite{buch1},~\cite{buch3}. Non-standard diagonals appear in the
Odessky -- Feigin algebras~\cite{of}. All this formally exceeds the
boundaries of quantum groups but nevertheless is in coherence with the
heuristic thesis stated above.
In the present paper we introduce and study QUANTUM CATEGORIES and the
simplest but fundamental example, the QUANTIZATION OF THE CATEGORY OF
LINEAR (SUPER)SPACES.
This needs to be somehow explained since it is very common to treat
categories only as ``general nonsense''.
Quantum group theorists use
categories mostly in this manner. (Categories of
representations and their abstract generalizations like ``tensor'' or
``braided'' categories are good example.) But the philosophy of categories
in this paper is
quite different. We consider categories first of all as algebraic
structures, sets of arrows with the multiplication law.
Hence categories are an
immediate generalization of groups, for many reasons more useful
than semigroups. Actually, this structure appears quite often in very
classical situations. The category structure of such examples is
of course known but usually is not stressed.
(To mention just few.
Paths on a topological space form a category. Parallel transports
in fiber bundles and the homology ``local coefficients'' are examples of
its representations. Cobordism provide examples of categories with
``films'' as arrows. This includes ``histories'' or Lorentzian
cobordism, see~\cite{mis},~\cite{gib}. It is worth mentioning that the
coordinate changes in the fixed open domain $U\subset\mathbb R^n$ also form a
category, not a group. Its representations exactly correspond to
``geometrical objects''.) Recent research supplied new examples
of categories naturally appearing in the context very far from
``general nonsense''. For example, it was found that the representations of
classical groups, such as spinor representation, naturally extend to certain
categories of linear
relations~\cite{ner1}, ~\cite{ner2}. The study of Poisson geometry lead to
symplectic groupoids~\cite{kar}, which provide an example of ``continuous
categories''. We can mention the category of tangles~\cite{tur}, etc.
This list can be extended.
In the context of ``general nonsense'', the role of categories should be compared
to that of groups
in Klein's classification of geometries.
(This view was expressed by the founders of category theory themselves~\cite{em}.)
Hence the main thesis of a {\sl ``category program'' }\/
can be stated as follows.
{\it Categories are the most important generalization of groups; it is
necessary to study their actions, representations, deformations
and extensions,
cohomology} (see~\cite{bau},~\cite{wat}, ~\cite{qui}), {\it the analogues of
Lie theory}\/ etc.
In fact,
the research in these directions is going on, with various
motivations.
We can especially note ~\cite{ner2} and~\cite{kar}.
(The author himself was brought to the necessity of a program
like this through the reflections on some problems of infinite-dimensional
geometry connected with the topics of~\cite{v2}.)
The study of quantum categories initiated in the present paper
can be considered as one more step in the
realization of the ``category program''.
Inside
the quantum group theory the particular motivations for introducing
quantum categories are the following.
The same classical object can admit different ``quantum deformations'' (for
example, corresponding to different values of the deformation parameter).
Which
should be chosen ? The correct answer is that all of them should be
considered simultaneously, together with all possible transformations
between them. This inevitably leads to a quantum category or a quantum groupoid (if restricted to invertible transformations),
while quantum groups correspond to particular
choices of deformation.
Let us consider a simple example. It is new and can serve as
a very good illustration of the
basic ideas of this work.
\medskip
{\bf Example.} Take matrix elements of
$T_{\a \b }=\left(\begin{array}{cc} a&b \\
c&d
\end{array}\right)$
as generators of an associative algebra $M_{\a \b}$ and impose
the following commutation relations
\begin{equation}
\begin{aligned}
ab- \l q_\b ^{-1}ba&=0, \\
ac- \l q_\a ca&=0,\\
ad- q_\a q_\b ^{-1}da&=(\l - \l ^{-1})q_\a cb\\
bc -q_\a q_\b cb &=0,\\
bd -\l q_\a db&=0,\\
cd - \l q_\b ^{-1}dc&=0,
\end{aligned}
\end{equation}
where $\l \neq 0, \pm i$, \ $q_\a, q_\b \neq 0$.
(We shall treat $T_{\a \b}$
as the matrix of a homomorphism from the quantum deformation of the linear
space $\mathbb C^2$ with the parameter $q_\a$ to that with the parameter $q_\b$.
This is explained in the main text below.)
The category nature of the transformation $T_{\a \b}$ is reflected in the
fact that if we take the matrix
$T_{\a \b}$ suiting (1) and the matrix $T_{\b \c}$ suiting
(1) with $q_\b$ instead of $q_\a$ and $q_\c$ instead $q_\b$,
then the product
$T_{\a
\c}=T_{\a \b}T_{\b \c}$ also suits (1) with $q_\a$, $q_\c$ as the
parameters
(the matrix elements of the different matrices are assumed to be
commuting).
If
$q_\a=q_\b=\l=1$ then we obtain the usual algebra of functions on the matrix
semigroup
$\mathrm{M}\mathrm{a}\mathrm{t}(2)$.
Thus it is wise to think that after the quantization this semigroup
becomes a ``quantum category"
(with $q_\a$ as ``objects") and the group $\mathrm{G}\mathrm{L}(2)$
after the quantization
turns a ``quantum groupoid''
if one adds a formal inverse
$(\operatorname{det}_{\a \b}(T_{\a \b}))^{-1}$ to each of the algebras $M_{\a \b}$, where $\operatorname{det}_{\a \b}(T_{\a \b})$
stands for $ad-\l q_\a cb=ab-\l q_\b^{-1}bc$.
The algebras $M_{\a \b}$ possess the ``Poincare - Birkhoff - Witt'' property:
the ordered monomials in matrix entries form the additive bases.
For $q_\a=q_\b=1$ the algebra $M_{\a \a}$ reduces to
the function algebra on
the quantum group $\mathrm{G}\mathrm{L}(2)_q$ defined in~\cite{rtf},
with $q=\l$.
\bigskip
(The appearance of ``quantum categories" in such context can be related
to the thesis QUANTIZATION ELIMINATES DEGENERACY suggested by
Reshetikhin, Takhtajan, Faddeev \cite{rtf}.)
The results of this paper were obtained in November 1994 -- January 1995.
They were reported at the seminar on quantum groups under guidance of
J.~Bernstein, J.~Donin and S.~Shnider in Bar-Ilan University (Ramat-Gan,
Israel).
\bigskip
\subsection*{ Contents.} The paper is divided into three sections.
In Section 1 we
introduce generalized bialgebras which are in the same relation to
quantum categories as usual Hopf algebras are to quantum groups.
In Section 2 the quantum deformation of the category of vector superspaces
is constructed. We make use of the method of the ``universal coacting'' known
for
quantum groups \cite{man1},\cite{man2},\cite{dem2},\cite{dem3}. We
calculate the commutational relations and consider various examples
including the quantization of the dual space, the quantization of
bilinear forms etc.
In Section 3 we present a generalized
\mbox{$R$-matrix} form for the commutational relations on
the quantum category
$\mathsf{L}\mathsf{i}\mathsf{n}_q$ and establish a criterion for the
classical value of the dimension of the function algebra.
This is a technically hard theorem, which may be considered as the main
theorem of this paper. The connection with the ``Yang -- Baxter
structures'' on quantum linear spaces is established.
In the Appendix we give a general definition
of ``indexed bialgebras'' which include
both the bialgebras defined in Section 1 and those dual to them.
\subsection*{ Notation.}
Throughout the paper the standard tensor notation is used with the regard of
\mbox{$\mathbb Z_2$-grading}. (For the ``super'' notions that are used we
refer to
\cite{v1},\cite{v3}.) For some reason it is convenient to write the
coordinates of a vector as a row and those of a covector as a column,
and to consider their pairing in the form
$\langle v, v' \rangle,\, v \in V, \, v' \in V' $.
\subsection*{ Acknowledgements.}
I wish to thank V.M.~Buchstaber for the discussions of the results of this
paper and for supplying the texts of
~\cite{buch1},~\cite{buch2},~\cite{buch3}, E.E.~Demidov for supplying
the texts of~\cite{dem2} and~\cite{man1}, and O.M.~Khudaverdian for
the discussions of the role of categories in connection to mathematical
physics. I also wish to thank J.~Bernstein,
J.~Donin and S.~Shnider for the discussions in Tel-Aviv and
Bar-Ilan universities and for their hospitality in May--June 1995. I thank
J.~Donin and S.~Shnider for the text of their paper~\cite{don} and I thank
S.~Shnider for the copy of his remarkable book with S.~Sternberg~\cite{ss}.
\renewcommand{\hom}{\operatorname{\mathbf{H}\mathbf{o}\mathbf{m}}}
\newcommand{\operatorname{\mathbf{C}^\infty}}{\operatorname{\mathbf{C}^\infty}}
\newcommand{\operatorname{Mor}}{\operatorname{Mor}}
\newcommand{\operatorname{\mathbf{M}\mathbf{o}\mathbf{r}}}{\operatorname{\mathbf{M}\mathbf{o}\mathbf{r}}}
\newcommand{\operatorname{Ob}}{\operatorname{Ob}}
\newcommand{\mathsf{C}}{\mathsf{C}}
\section{Generalized bialgebras}
We are going to introduce a generalization of bialgebras as to make
possible the discussion of ``quantized categories", in the same way as
the usual bialgebras and Hopf algebras serve as the foundation of the
notion of a ``quantum group". (More general definition of an ``indexed
bialgebra" will be discussed in the appendix.)
Consider a set of indices $\L$ (``objects'').
{\bf Definition 1.1.} A set $A=(A_{\a \b}),\quad \a,\b\in\L$ of
linear spaces or modules over a commutative
ring
is said to be a {\it coalgebra with objects
}$\L$ if linear maps $\D_{\a \b \c} : A_{\a \c} \longrightarrow
A_{\a \b}\otimesA_{\b \c}$ are given, for all $\a,\b,\c\in\L$.
It is called a
{\it coassociative coalgebra} if the diagram
\begin{equation}
\begin{CD}
A_{\a \d} @>{\D_{\a\b\d}}>> A_{\a \b} \otimes A_{\b \d} \\
@V{\D_{\a \c \d}}VV @VV{1 \otimes \D_{\b \c \d}}V \\
A_{\a \c}\otimes A_{\c \d} @>>{\D_{\a \b \c} \otimes 1}> A_{\a \b}\otimes A_{\b \c}\otimes A_{\c \d}
\end{CD}
\end{equation}
is commutative for every $\a,\b,\c,\d\in\L$.
A set of linear maps $\e_\a:A_{\a \a}\longrightarrow R$
(here $R$ is the main field or ring) with the property that all
diagrams of the form
\begin{equation}
\begin{CD}
A_{\a \a}\otimesA_{\a \b} @<{\D_{\a\a\b}}<< A_{\a \b} @>{\D_{\a\b\b}}>> A_{\a \b}\otimesA_{\b \b} \\
@V{\e_\a\otimes1}VV @VVV @VV{1\otimes\e_\b}V \\
R\otimesA_{\a \b} @>>> A_{\a \b} @<<< A_{\a \b}\otimes R
\end{CD}
\end{equation}
must be commutative will be called a {\it counit} in $A$.
(Here $R\otimesA_{\a \b}\longrightarrowA_{\a \b}, A_{\a \b}\otimes R\longrightarrowA_{\a \b}$
are the natural isomorphisms and middle vertical arrow is the identity
map.)
\medskip
The definition of the {\it homomorphism } of coalgebras
$f:(A_{\a \b})_{\a,\b\in\L}\longrightarrow(B_{i,j})_{i,j\in I}$ is obvious.
(A map of sets $\a\mapsto i(\a)$ must be given.)
\medskip
{\bf Definition 1.2.} Let all modules be the associative
$R$-algebras with units and all maps $\D_{\a \b \c}$ and $\e_\a$ be the
$R$-linear algebra homomorphisms. Then we shall call $A$ a {\it
bialgebra with objects} $\L$.
In the sequel we shall simply say ``coalgebra'' and ``bialgebra'' having in
mind the given definitions.
\medskip
{\bf Example 1.1.} {\sl Functions on a category.} Let $\mathsf{C}$ be a small
category with the object set
$\operatorname{Ob}{\mathsf{C}}=\L$. Define $A_{\a \b}$ as
the function algebra on $\operatorname{Mor}(\b,\a)$. Then the composition induces
the comultiplication $\D_{\a \b \c}:A_{\a \c} \longrightarrow A_{\a \b}\otimesA_{\b \c}$ and the embedding of
the unit $1_\a\in
\operatorname{Mor}(\a,\a)$ induces the counit homomorphism $\e_\a:A_{\a \a} \longrightarrow R$.
(We assume the category and the class of functions in consideration to be
such that the Cartesian product of the morphism sets corresponds to a
tensor product of function algebras.)
We obtain the bialgebra with the object set
$\L$. Here all algebras $A_{\a \b}$ are commutative.
\medskip
{\bf Example 1.2.} {\sl Additive categories.} An additive category
$\mathsf{A}$ can be treated as ``ring with several objects''
\cite{mit}. The additive category axioms are exactly dual
(in the sense of reverting arrows)
to our definition of a coalgebra (over $\mathbb Z$,
coassociative and with a counit). An important case is
additive categories like $\mathbb Z \mathsf{C}$ or $R\mathsf{C}$, generated by the
morphisms
of an arbitrary small category $\mathsf{C}$ (an analogue of the group ring).
In this case
each module $A_{\a \b}=R\operatorname{Mor}_{\mathsf{C}}(\b,\a)$ is endowed with the comultiplication
$A_{\a \b} \longrightarrow A_{\a \b}\otimesA_{\a \b}$, defined on $a_i \in
\operatorname{Mor}_{\mathsf{C}}(\b,\a)$ by the formula $\D a_i=a_i\otimes\a_i$
(the same as for a group
ring). Having this additional structure in mind, we may say that the
category algebra is a ``bialgebra'', but not in the sense of the
definition above but of the ``dual'' definition (w.r.t. the reverse arrows).
The comultiplication is symmetric here.
Considering the dual modules
$A_{\a \b}'=\operatorname{Hom}(A_{\a \b},R)$ we (under certain assumptions) come to a bialgebra.
It coincides with the bialgebra of the previous example if, for example, all
sets $\operatorname{Mor}(\a,\b)$ are finite.
\medskip
{\bf Example 1.3.} ``{\it Supercategories}''. They can be considered in the
same fashion as Lie supergroups. (For instance one can obviously
define the supercategory
of $\mathbb Z_2$-graded vector spaces, which contains the general linear
supergroups,
and in the same way the (infinite-dimensional) supercategory of smooth
manifolds.)
If for a given supercategory we consider function algebras on the
supermanifolds
$\operatorname{\mathbf{M}\mathbf{o}\mathbf{r}}(\b,\a)$, then, with standard comments on tensor products, we again
obtain a bialgebra with the multiplication commutative in
$\mathbb Z_2$-graded sense.
\medskip
Thus categories and supercategories can be described by bialgebras with
commutative multiplication. It is natural to think that arbitrary
bialgebras (in our sense) without commutativity condition may be associated
with ``quantized categories''. Sure, it makes sense in the case a
classical (super)category is at hand and the ``quantum category'' in
consideration reduces to it for certain values of the ``parameters''
(in most broad sense) on which the construction depends.
\medskip
{\bf Definition 1.3.} By an {\it antipode} in a bialgebra $A=(A_{\a \b})$ we
shall mean a set of algebra antihomomorphisms
$S_{\a\b}:A_{\a \b} \longrightarrow A_{\b \a}$ such that
the diagram
{
\unitlength=1em
\newcommand{\larr}{\begin{picture}(0,0)
\put(-3,0.28){\vector(1,0){5.8}}
\put(-1,0.7){${\scriptstyle 1\otimes S_{\b\a}}$}
\end{picture}}
\newcommand{\rarr}{\begin{picture}(0,0)
\put(2.8,0.28){\vector(-1,0){5.8}}
\put(-1,0.7){${\scriptstyle S_{\b\a}\otimes 1}$}
\end{picture}}
\begin{equation}
\begin{CD}
A_{\a \b}\otimesA_{\b \a} @. \larr @. A_{\a \b}\otimesA_{\a \b} @. \rarr @. A_{\b \a}\otimesA_{\a \b} \\
@A{\D_{\a\b\a}}AA @. @VmVV @. @AA{\D_{\b\a\b}}A \\
A_{\a \a} @>>{\e_\a}> R @>>> A_{\a \b} @<<< R @<<{\e_\b}< A_{\b \b}
\end{CD}
\end{equation}
is commutative. (The unmarked arrows are inclusion of unit.)
}
Bialgebras with antipode correspond to
``quantum groupoids'', i.e. to quantum categories
in which ``all arrows are invertible''.
\medskip
In classical situation the notion of a ``concrete category''
is important (that is a subcategory of the category of sets).
A bialgebra $(A_{\a \b})$ is called {\it
concrete}, if the ``coaction'' on a given set of algebras $(A_{\a})$ is
provided.
That is the algebra homomorphisms $\d_{\a \b}:A_{\b} \longrightarrow A_{\a}\otimesA_{\a \b}$
that are compatible with
the comultiplication: $(\d_{\a \b}\otimes1)\circ\d_{\b \c}=(1\otimes\D_{\a \b \c})\circ\d_{\a \c}$
for all
$\a,\b,\c$.
In a similar way one can transfer to quantum categories the notions of
a covariant functor (a representation of a category), a dual category
etc. Note that in the following it will be convenient to consider both
``left'' and ``right'' coactions. Depending on it, the algebra
$A_{\a \b}$ can be interpreted as functions either on the arrows
$\a \longrightarrow \b$ or $\a\longleftarrow\b$.
\unitlength=1em
\newcommand{\underline{\operatorname{hom}}}{\underline{\operatorname{hom}}}
\newcommand{\underline{\operatorname{end}}}{\underline{\operatorname{end}}}
\newcommand{g^{AB}}{g^{AB}}
\newcommand{f_{KL}}{f_{KL}}
\newcommand{I_{V}}{I_{V}}
\newcommand{J_{V}}{J_{V}}
\newcommand{I_{W}}{I_{W}}
\newcommand{J_{W}}{J_{W}}
\newcommand{\strelka}
{\begin{picture}(1,1)\put(1,-1){\vector(3,-2){3.3}}\end{picture}}
\section{The construction of the quantum category ${\mathsf L\mathsf i\mathsf n}_q$. Examples.}
Let $A_{\a}$ be a set of algebras. Consider the following problem.
Consider sets of algebras $A_{\a \b}$, coacting on a given set of algebras
$Aa$, in a sense that
algebra homomorphisms $\d_{\a \b}:A_{\b}\longrightarrowA_{\a}\otimesA_{\a \b}$ are given
for all indices , and let us look for the universal (initial)
object among such coactions.
In other words for algebras $\tildeA_{\a \b}$ coacting in the same sense on $A_{\a}$,
there is a unique set of homomorphisms $A_{\a \b}\longrightarrow\tildeA_{\a \b}$ for which
the following diagram
\begin{equation}
\begin{CD}
A_{\b}\strelka@>>>A_{\a}\otimesA_{\a \b}\\
@. @VVV\\
{} @. A_{\a}\otimes\tildeA_{\a \b}
\end{CD}
\end{equation}
is commutative.
\medskip
{\bf Theorem 2.1.} {\it The universal set of algebras $A_{\a \b}$ (if exists)
is a bialgebra. The comultiplication and counit are canonically defined
by the compatibility with the coaction.}
{\bf Proof.} The iteration of the coaction is always defined,
hence for any three indices there exists a homomorphism
$A_{\c} \longrightarrowA_{\b}\otimesA_{\b \c}\longrightarrowA_{\a}\otimesA_{\a \b}\otimesA_{\b \c}$.
>From the universality we obtain homomorphisms
$A_{\a \c}\longrightarrowA_{\a \b}\otimesA_{\b \c}$. Similarly, the isomorphism
$A_{\a}\longrightarrowA_{\a}\otimes R$ yields a counit homomorphisms
$A_{\a \a}\longrightarrow R$. The coassociativity and the counit property are easily obtained
from the uniqueness part of the universality condition.
\medskip
It is quite obvious that we can similarly treat
two sets of algebras
$A_{\a},B_{\a}$ (or any number of them) demanding that $A_{\a \b}$
should coact on both. Or one can put additional constraints on the exact
outlook of the coactions. Then the similar universal coacting bialgebra
(if exists) will be
subject to more tight restrictions. The claim of the theorem still holds.
We shall apply this construction as follows.
Consider a set of
$\mathbb Z_2$-graded vector spaces (over a field $k$).
In parallel to each space $V$ we consider the space $V\P$, where
$\P$ stands for the parity reversion functor. Consider the dual spaces
(the spaces of linear functions). A canonical even isomorphism
\begin{equation}
V'\otimes V'\stackrel{\sim}{=} \P V' \otimes \P V', \label{p}
\end{equation}
takes the basis tensors $e^A\otimes e^B$ to $(e^A\otimes
e^B)^\P :=(-1)^{\tilde A}(\P e^A)\otimes(\P e^B)$. Let us fix a decompositipon of
$V'\otimes V'$ to two complementary subspaces
: $V'\otimes V'=I\oplus J$.
Consider the quadratic algebras
$(V',I):=T(V')/(I)$ ... $(\P V',J^\P):=T(\P V')/(J^\P)$.
(Quotient by the ideals generated by $I$ and $J$. By $^\P$ we denote
the isomorphism (2).) We call $(V',I)$ and $(\P V',J^\P)$ the
(polynomial) {\it function algebras on quantum superspaces }
$V_q$ and $V_q\P$.
Thus the definition of quantum superspace depends on our choice of the
decomposition
$I\oplus J=V'\otimes V'$, and the spaces $V_q$ and $V_q\P$
are defined simultaneously, not independently.
Denote by $x^A$ and $\x^A$
the image of basic linear functions $e^A$ and $\P e^A$ in the algebras
$(V',I)$ ... $(\P V',J^\P)$ respectively.
We shall call them the {\it coordinates} on quantum superspaces
$V_q$ ... $V_q\P$.
\medskip
{\bf Example 2.1.} Take $I$ spanned by
$e^A\otimes e^B-(-1)^{\tilde A\tilde B}e^B\otimes e^A$
(the basis of the skew-symmetric tensors), and take $J$ spanned by
$e^A\otimes e^B+(-1)^{\tilde A\tilde B}e^B\otimes e^A$
(the basis of the symmetric tensors). Then the relations in $(V',I)$ and $(\P V',J^\P)$
are plain commutativity relations: $x^A x^B=(-1)^{\tilde A \tilde B}x^B x^A$,
$\x^A\x^B=(-1)^{(\tilde A+1)(\tilde B+1)}\x^B\x^A$ (can be checked).
Thus quantum superspaces actually include classic ones.
\medskip
Do this for each $V$. We obtain a set of quantum superspaces
$V_q$, $V_q\P$, i.e. a set of algebras $(V',I), (\P V',J^\P)$
which are parametrized by triples: $(V,I,J)$. Denote $A_{\a}=(V',I), B_{\a}=(\P V',J^\P)$, where
$\a=(V,I,J)$, and consider a coaction
\begin{equation}
\begin{aligned}
\d:A_{\b}\longrightarrowA_{\a}\otimesM_{\a \b},\\
\d:B_{\b}\longrightarrowB_{\a}\otimesM_{\a \b}.
\end{aligned}
\end{equation}
The algebras $(V',I)$ and $(\P V',J^\P)$ inherit the $\mathbb Z$-grading from the tensor
algebra.
(Not to be confused with the parity.) We demand that the coaction must
preserve this grading (``the linearity condition''). Then
$\d(y^K)=x^A\otimest_A{}^K$,
$\d(\h^K)=\x^A\otimess_A{}^K $, where $x^A, y^K, \x^A, \h^K$ are coordinates on
$V_q, W_q, V_q\P, W_q\P$ respectively, and $t_A{}^K, s_A{}^K$ are certain elements
of the algebra
$M_{\a \b}$. We also demand that $t_A{}^K=s_A{}^K$ (``compatibily with the functor
$\P$'').
{
\renewcommand{\d_{\a \b}}{\d_A{}^B}
\renewcommand{\d_{\a \c}}{\d_A{}^C}
\newcommand{\d_B{}^D}{\d_B{}^D}
\medskip
{\bf Theorem 2.2.} {\it
For the set of pairs
$(V',I), (\P V',J^\P)$, with the given constraints (linearity and commuting with $\P$),
there exists the universal coacting.
This is a bialgebra $M=(M_{\a \b})$, where all $M_{\a \b}$ are quadratic algebras.
The choice of bases in $V$ and $W$ determines the generators
$t_A{}^K\inM_{\a \b}$, where $\a=(V,I_{V},J_{V}), \/\b=(W,I_{W},J_{W})$, and
\begin{equation}
\d(y^K)=x^A\otimest_A{}^K, \quad \d(\h^K)=\x^A\otimest_A{}^K.
\end{equation}
The quadratic relations in $M_{\a \b}$ are the following:
\begin{equation}
\begin{aligned}
(-1)^{\tilde B\tilde K}f^{AB}_{(J)}f_{KL}^{(I)}t_A{}^Kt_B{}^L=0,\\
(-1)^{\tilde B\tilde K}f^{AB}_{(I)}f_{KL}^{(J)}t_A{}^Kt_B{}^L=0, \label{2.5}
\end{aligned}
\end{equation}
where we denote by $f^{(I)}, f^{(J)}$ the basis tensors in $I\oplus
J=V'\otimes
V'$ (and the same for $W$) and by $f_{(I)}, f_{(J)}$ the corresponding
elements of the dual basis.
The comultiplication
$\D:M_{\a \c}\longrightarrowM_{\a \b}\otimesM_{\b \c}$
and the counit $\e:M_{\a \a}\longrightarrow k$ are defined by the standard formulas
\begin{align}
\D(t_A{}^S)&=t_A{}^K\otimest_K{}^S,\\
\e(t_A{}^B)&=\d_{\a \b}
\end{align}
}
{\bf Proof.}
{
\newcommand{\bar\delta}{\bar\delta}
We shall show that the relation~(\ref{2.5}) is valid in
any coacting
$A_{\a \b}$. Any coaction is defined by the formula~(\ref{4}), with some $t_A{}^K$.
That $\delta$ is homomorphic is equivalent to the condition
$\bar\delta(I_{W})\subsetI_{V}\otimesA_{\a \b}\subset(V'\otimes V')\otimesA_{\a \b}$ and
$\bar\delta(J_{W}^\P)\subsetJ_{V}^{\P}\otimesA_{\a \b}\subset(\P V'\otimes\P
V')\otimesA_{\a \b}$,
where $\bar\delta$ stands for the ``covering'' coaction on the free tensor
algebras,
which exists independently of the structure
of the algebra $A_{\a \b}$. Consider
first the conditions following from the
relation $(I)$. To avoid confusion with the elements of the tensor product
of two algebras we shall omit the symbol $\otimes$ for the tensors
on a given space
(so below $e_Ae_B:=e_A\otimes
e_B$ etc). Suppose $f=f_{KL} e^Ke^L\in I_{W}\subset W'\otimes W', \quad g=g^{AB}
e_Ae_B\in \operatorname{Ann}I_{V}\subset V\otimes V$. Then for any $f$ and $g$ it is
necessary that
$
0=\langle g,\bar\delta(f)\rangle=g^{AB}f_{KL}\langle
e_Ae_B,(e^C\otimest_C{}^K)(e^D\otimest_D{}^L)\rangle=g^{AB}f_{KL}(-1)^{\tilde D(\tilde C+\tilde K)}
\langle
e_Ae_B,e^Ce^D\otimest_C{}^Kt_D{}^L\rangle=g^{AB}f_{KL}(-1)^{\tilde D(\tilde C+\tilde K)}(-1)^
{\tilde B\tilde C}\d_{\a \c}\dbdt_C{}^Kt_D{}^L=(-1)^{\tilde B\tilde K}g^{AB}\fklt_A{}^Kt_B{}^L
$.
Thus the condition $\bar\delta(I_{W})\subsetI_{V}\otimesA_{\a \b}$ is equivalent to
\begin{equation}
(-1)^{\tilde B\tilde K}g^{AB}\fklt_A{}^Kt_B{}^L=0,\label{2.8}
\end{equation}
where as $f$ the basis elements of $I_{W}$ can be taken and as $g$
the basis elements of $\operatorname{Ann}I_{V}$.
Now we notice that the condition $\bar\delta(J_{W}^\P)\subsetJ_{V}^\P\otimesA_{\a \b}$
is equivalent to $\bar\delta(J_{W})\subsetJ_{V}\otimesA_{\a \b}$.
This follows from the fact that $\bar\delta$
commutes with the isomorphism~(\ref{p}), i.e.
$(\bar\delta f)^\P=\bar\delta(f^\P)$ for any $f\in
W'\otimes W'$. This immediately follows from
the equality $\bar\delta(\P
w')=\P\bar\delta(w')$ (the $\P$-symmetry of the coaction).
Thus it is proven that in any coacting
the relations~(\ref{2.8}) are valid, for $f\inI_{W}, \quad g\in\operatorname{Ann}I_{V}$
or $f\inJ_{W}, \quad g\in\operatorname{Ann}J_{W}$. Taking into account that $I$ and $J$ are
complementary one rewrite the relations as~(\ref{2.5}).
Now take the matrix entries $t_A{}^K$ as independent variables
and consider the associative algebras generated
by them with the defining relations~(\ref{2.8}).
The set of algebras obtained will be universal.
Indeed, for an arbitrary coacting
$A_{\a \b}$ change the notation of the matrix elements to
$s_A{}^K$, then for any homomorphism commuting with the coaction
by necessity $t_A{}^K\mapstos_A{}^K$. But this formula
defines the homomorphism uniquely
(since $t_A{}^K$ are generators)
and it is well-defined because of the relation proven above.
$\quad\square$
}
}
\bigskip
{\bf Remarks.} {\bf 1.} As follows from the proof, the universal coacting
for the pairs of algebras
$T(V')/(I), T(\P V')/(J^{\P})$ and for the pairs of algebras
$T(V')/(I), T(V')/(J)$ will be the same. Our preference is explained by the
fact that in the classical case (see above) both algebras
$T(V')/(I)$ and $T(\P V')/(J^{\P})$ are commutative (in $\mathbb Z_2$-graded sense),
while the algebra $T(V')/(J)$ is not commutative (odd generators
anticommute with even ones). The same alternative exists for the definition of
the exterior algebra, see~\cite{ber}.
\smallskip
{\bf 2.} If for each $V$ we fix a family of complementary subspaces
$I_k^V\subset V'\otimes V'$ and look for a universal coaction for the algebras
$T(V')/(I_k^V)$ (with the same form of the coaction on the generators,
$\bar\d(e^K)=e^A\otimest_A{}^K$, for all
$k$), then we shall obtain a bialgebra $M=(M_{\a \b})$, with the following
commutational relations for each of
$M_{\a \b}$ (matrix entries $t_A{}^K$ serve as generators):
\begin{equation}
(-1)^{\tilde B\tilde K}g^{AB}_{(k)}f_{KL}^{(k)}t_A{}^Kt_B{}^L=0,
\end{equation}
$f^{(k)}$ belongs to the basis of $I_k^W$, $g_{(k)}$ to the basis of
$\operatorname{Ann} I_k^V$. Here the
``numbers'' of objects are $\a=(V,I_1^V,\dots ,I_s^V)$ and
$\b=(W,I_1^W,\dots ,I_s^W)$.
\smallskip
{\bf 3.} The relations (\ref{2.5}) are linear independent.
It easy to calculate that for the classical values of
$\dim I_V$, $\dim J_V$, $\dim I_W$, $\dim J_V$
the dimension of the quadratic part of the ``quantum algebra'' $M_{\a \b}$
will be classical too. The question whether the same will be true for
the higher order terms is highly non-trivial. We discuss it in the next
section.
\smallskip
{\bf 4.} In quantum group theory the universal coaction method
was proposed by Yu.A.Kobyzev
(see~\cite{man1}). Originally a single quadratic algebra $A=T(V')/(I)$
was considered. This gives only ``half'' of the necessary commutational
relations on quantum (semi)group. This difficulty has been resolved in a
rather artificial manner by using the ``dual'' quadratic algebra $A^!$
(see the remark after Example 2.5 below). In particular that
implied that a single subspace $I\subset V'\otimes V'$ played the role of
the ``quantization parameter''. The application of families of
complementary subspaces is described in~\cite{dem2}, in purely even case.
(The functor $\P$ is not used in ~\cite{dem2}.)
It is worth noting that in
~\cite{man1} a quadratic algebra
$\underline{\operatorname{hom}}(A,B)=A^!\bullet B$ was introduced for each pair of
quadratic algebras $A$ and $B$, together with a sort of
``coproduct'' $\D:\underline{\operatorname{hom}}(A,C) \longrightarrow \underline{\operatorname{hom}}(B,C)\circ\underline{\operatorname{hom}}(A,B)$ (we refer
also to ~\cite{gel} for the notation). The algebras $\underline{\operatorname{hom}}$ are not so
interesting (they, and in particular $\underline{\operatorname{end}}(A)=\underline{\operatorname{hom}}(A,A)$,
indeed lack half of the relations, so they are strongly non-commutative).
But having in mind the natural homomorphism $A\circ B\longrightarrow A\otimes B$
~\cite{gel},~\cite{man1} they can be considered as an example of a quantum
category in our sense, or. more precisely, an examlple of the generalized
bialgebra (in the sense of Definition 1.2).
The interpetation given in ~\cite{man1} was completely different.
It was based on the ``internal $\operatorname{Hom}$'' formalism of the ``rigid tensor
categories'', which are just actual categories endowed with the additional
structure.
\bigskip
{\bf Definition 2.1.} The bialgebra $M=(M_{\a \b})$, defined in Theorem 2.2,
will be called the {\it algebra of {\em(polynomial)} functions on the
quantum category}
${\mathsf L\mathsf i\mathsf n}_q$.
\bigskip
{\bf Example 2.2.} Consider classical (not deformed)
spaces. For every $V$ the subspaces $\operatorname{Ann} I, \operatorname{Ann} J \subset
V\otimes V $ are spanned by the tensors $e^A\otimes
e^B+(-1)^{\tilde A\tilde B}e^B\otimes
e^A$ and $e^A\otimes e^B-(-1)^{\tilde A\tilde B}e^B\otimes e^A$. By Theorem 2.2
we obtain the following relations for the matrix elements of the
homomorphism from $V$ to $W$:
{
\newcommand{\d_A{}^{A'}}{\d_A{}^{A'}}
\renewcommand{\d_{\a \b}}{\d_A{}^{B'}}
\newcommand{\d_B{}^{B'}}{\d_B{}^{B'}}
\newcommand{\d_B{}^{A'}}{\d_B{}^{A'}}
\newcommand{\d_{K'}{}^{K}}{\d_{K'}{}^{K}}
\newcommand{\d_{L'}{}^L}{\d_{L'}{}^L}
\newcommand{\d_{K'}{}^L}{\d_{K'}{}^L}
\newcommand{\d_{L'}{}^K}{\d_{L'}{}^K}
\renewcommand{t_A{}^K}{t_{A'}{}^{K'}}
\renewcommand{t_B{}^L}{t_{B'}{}^{L'}}
\newcommand{\tilde {B'}}{\tilde {B'}}
\newcommand{\tilde {K'}}{\tilde {K'}}
\begin{eqnarray*}
(\d_A{}^{A'}\d_B{}^{B'}+(-1)^{\tilde A\tilde B}\d_B{}^{A'}\d_{\a \b})(\d_{K'}{}^{K}\d_{L'}{}^L-(-1)^{\tilde K\tilde L}\d_{K'}{}^L\d_{L'}{}^K)
(-1)^{\tilde {B'}\tilde {K'}}t_A{}^Kt_B{}^L=0, \\
(\d_A{}^{A'}\d_B{}^{B'}-(-1)^{\tilde A\tilde B}\d_B{}^{A'}\d_{\a \b})(\d_{K'}{}^{K}\d_{L'}{}^L+(-1)^{\tilde K\tilde L}\d_{K'}{}^L\d_{L'}{}^K)
(-1)^{\tilde {B'}\tilde {K'}}t_A{}^Kt_B{}^L=0.
\end{eqnarray*}
}
Then, after summation, we obtain:
$$
(-1)^{\tilde B\tilde K}t_A{}^Kt_B{}^L-(-1)^{\tilde K\tilde L+\tilde B\tilde L}\talt_B{}^K+(-1)^{\tilde A\tilde B+\tilde A\tilde K}t_B{}^K
t_A{}^L-(-1)^{\tilde A\tilde B+\tilde K\tilde L+\tilde A\tilde L}\tblt_A{}^K=0
$$
and a similar equality with the opposite signs
before the second and the third terms.
Summing and substracting these equalities, we get
$$
t_A{}^Kt_B{}^L-(-1)^{(\tilde A+\tilde K)(\tilde B+\tilde L)}\tblt_A{}^K=0,
$$
for any $A,B,K,L$, i.e. the usual commutativity condition.
\medskip
{\bf Example 2.3.} Take one-dimensional $k$ as $W$ with
$0$ and $k\stackrel{\sim}{=} k\otimes k$
as $I_W, J_W$ respectively. Then by Theorem 2.2 we get
\begin{equation}
g^{AB}_{(J)}t_A t_B=0
\end{equation}
as commutational relations for the coefficients of even {\it linear
forms} on $V_q, V_q\P$. Here $g_{(J)}\in \operatorname{Ann} J_V$. In the same way, for the
coefficients of {\it odd linear forms} $\theta_A, \quad \tilde
\theta_A=\tilde A+1$, we obtain the relations
\begin{equation}
(-1)^{\tilde B}g^{AB}_{(I)}\theta_A\theta_B=0,
\end{equation} where $g_{(I)}\in \operatorname{Ann} I_V$.
In other words, if one identifies
a quantum space with a triple $(V,I,J)$, then its {\it quantum dual space
} will be $(V',\operatorname{Ann} J, \operatorname{Ann} I)$. Pairings: $\langle x,t
\rangle=x^A\otimes t_A, \quad \langle x, \theta \rangle =x^A\otimes\theta_A$.
\medskip
{\bf Example 2.4.} From the previous example one can deduce the
commutational relations for the coefficients of
(even) {\it bilinear forms},
considered as homomorphisms to dual space (``lowering indices''):
\begin{equation}
\begin{aligned}
(-1)^{\tilde B\tilde C}g_{(I)}^{AC}g_{(J)}^{BD}t_{AB}t_{CD}=0,\\
(-1)^{\tilde B\tilde C}g_{(J)}^{AC}g_{(I)}^{BD}t_{AB}t_{CD}=0,
\end{aligned}
\end{equation}
where $\tilde t_{AB}=\tilde A +\tilde B, \quad g_{(I)}, g_{(J)}$ are tensors in $\operatorname{Ann} I,
\operatorname{Ann} J \subset V\otimes V$ respectively. The bilinear forms are:
$\langle x_1\mid T\mid x_2\rangle = x_1^A\otimes t_{AB}\otimes x_2^B$
and the similar for $\x_i$.
\medskip
{\bf Example 2.5.} Consider the quantum spaces with the following
commutational relations:
\begin{equation}
\begin{aligned}
x^A x^B-q^{AB}x^B x^A&=0,\\
\x^A\x^B+(-1)^{\tilde A+\tilde B}p^{AB}\x^B\x^A&=0.
\end{aligned} \label{sudb}
\end{equation}
Here $q^{AB}=(q^{BA})^{-1}$, $p^{AB}=(p^{BA})^{-1}$, $q^{AA}=p^{AA}=(-1)^{\tilde A}$.
(In the ``classical limit'' $q^{AB},p^{AB}\longrightarrow(-1)^{\tilde A\tilde B}$.)
Look for the relations in the function algebra
for the corresponding ``full subcategory''
of
${\mathsf L\mathsf i\mathsf n}_q$. Here the objects are triples
$(V,P_V,Q_V)$,
where
{\tolerance=500
${P_V=(p^{AB})=(p^{AB}_{(V)})}$, ${Q_V=(q^{AB})=(q^{AB}_{(V)})}$ are the matrices
of the parameters. The subspace $I$ is spanned by the tensors
${e^A\otimes e^B-q^{AB} e^B\otimes e^A}$, $J$ by the tensors
${e^A\otimes e^B+p^{AB} e^B\otimes e^A}$, $\operatorname{Ann} I$ by the tensors
${e_A\otimes e_B+\qban e_B\otimes e_A}$, and $\operatorname{Ann} J$ by the tensors
${e_A\otimes e_B-\pban e_B\otimes e_A}$.
As not to contradict the tensor notation, we have introduced here
the parameters with the lower indices:
$\qabn:=q^{AB}, \quad \pabn:=p^{AB}$.
The complementarity of $I$ and $J$
is equivalent to
\begin{equation}
q^{AB}+p^{AB}\neq0, \label{comp}
\end{equation}
(for any $V$). Similarly to Example 2.1 we obtain two relations
for the matrix elements of a
``linear map of the quantum space
$(V,P_V,Q_V)$ to the quantum space
$(W,P_W,Q_W)$'':
$$
\begin{aligned}
(-1)^{\tilde B\tilde K}t_A{}^Kt_B{}^L-q^{KL}(-1)^{\tilde B\tilde L}\talt_B{}^K+\qban((-1)^{\tilde A\tilde K}t_B{}^K
t_A{}^L-q^{KL}(-1)^{\tilde A\tilde L}\tblt_A{}^K)=0\\
(-1)^{\tilde B\tilde K}t_A{}^Kt_B{}^L+p^{KL}(-1)^{\tilde B\tilde L}\talt_B{}^K-\pban((-1)^{\tilde A\tilde K}t_B{}^K
t_A{}^L+p^{KL}(-1)^{\tilde A\tilde L}\tblt_A{}^K)=0
\notag
\end{aligned}
$$
Here $Q_V=(q^{AB}), P_V=(p^{AB}), Q_W=(q^{KL}), P_W=(p^{KL})$.
Provided \ref{comp}, these defining relations can be identically transformed
to the following final form:
\begin{multline}
t_A{}^Kt_B{}^L-\frac{\pban+\qban}{p^{LK}+q^{LK}}(-1)^{\tilde A\tilde L+\tilde B\tilde K}\tblt_A{}^K=\\
\frac{\pbanp^{LK}-\qbanq^{LK}}{p^{LK}+q^{LK}}(-1)^{(\tilde A+\tilde B)\tilde K}t_B{}^Kt_A{}^L \label{rel}
\end{multline}
for any $A, B, K, L$. Notice that for the elements of one column or of
one row
these relations reduce to
\begin{align}
t_A{}^Kt_B{}^K-\frac{\pban(1+(-1)^{\tilde K})+\qban(1-(-1)^{\tilde K})}{2(-1)^
{(\tilde A+\tilde B+1)\tilde K}}\tbkt_A{}^K=0, \label{rel1}\\
t_A{}^Kt_A{}^L-\frac{2(-1)^{(\tilde A)(\tilde K+\tilde L+1)}}{p^{LK}(1-(-1)^{\tilde A})+q^{LK}
(1+(-1)^{\tilde A})}\talt_A{}^K=0. \label{rel2}
\end{align}
\bigskip
{\bf Remarks. 1.} The particular case of (\ref{rel}) are the relations in
the function algebra on the
``multiparameter deformation'' of the general linear supergroup.
The original relations of the form (\ref{sudb}) for quantum linear spaces
(in the even case) together with the relevant quantization of $\operatorname{GL}\,(n)$
is due to Sudbery~\cite{sud} (see the surveys~\cite{dem2}, ~\cite{dem3}).
Before Manin~\cite{man2} introduced a multiparameter quantization
for supergroup
$\operatorname{GL}\,(n\mid m)$ starting from the relations similar to (\ref{sudb})
with $(q^{AB})^{-1}$ instead of $p^{AB}$ (in our notation).
There was a certain confusion in the papers~\cite{man1},
~\cite{man2}, ~\cite{dem1} between a vector space and its dual
and as a consequence the analogue of the second equation of (\ref{sudb})
was associated with the ``dual quadratic algebra''
$A^!$ while the coaction was taken as if it had been for the initial space.
The structure of the formulas (\ref{rel}-\ref{rel2}) is similar
to the relations for the standard (single-parameter) deformation
of the group $\operatorname{GL}\,(2)$ ~\cite{rtf}. The commutational relations
for the multiparameter deformations of $\operatorname{GL}\,(n\mid m)$ are commonly presented
in a more cumbersome form
~\cite{man1}, ~\cite{man2},
~\cite{dem1}, ~\cite{dem2},~\cite{dem3}.
{\bf 2.} In the paper~\cite{dem1} Demidov defined for two spaces $V$ and $W$
an algebra close to that defined above,
for $p^{AB}=(q^{AB})^{-1},
\quad p^{KL}=(q^{KL})^{-1}$, see the previous remark. But he never considered
a comultiplication
except for the case of a single space with fixed parameters
$q^{AB}$ (i.e. for ``quantum semigroups'').
\bigskip
{\bf Example 2.6.} For quantum space with the relations (\ref{sudb})
the commutational relations for the dual space (see Example 2.3)
are:
\begin{equation}
\begin{aligned}
t_A t_B-(\pabn)^{-1}t_B t_A=0,\\
\theta_A \theta_B-(-1)^{\tilde A+\tilde B}(\qabn)^{-1}\theta_B\theta_A=0.
\end{aligned}
\end{equation}
\medskip
{\bf Example 2.7.} Similarly for even bilinear forms (see Example 2.4):
{
\renewcommand{t_A{}^B}{t_{AB}}
\newcommand{t_{AD}}{t_{AD}}
\newcommand{t_{CD}}{t_{CD}}
\newcommand{t_{CB}}{t_{CB}}
\begin{multline}
t_A{}^Bt_{CD}-\frac{\pcan+\qcan}{\qbdn+\pbdn}(-1)^{\tilde A\tilde D+\tilde B\tilde C}\tcdt_A{}^B=\\
\frac{\pcan\qbdn-\qcan\pbdn}{\qbdn+\pbdn}(-1)^{(\tilde A+\tilde C)\tilde B}\tcbt_{AD}
\end{multline}
}
\bigskip
For convenience we shall also write down
the relations (\ref{sudb}--\ref{rel2}) for the subcategory of purely
even spaces
($V$ is purely even, $V\P$ is purely odd):
{
\renewcommand{q^{LK}}{q^{lk}}
\renewcommand{p^{LK}}{p^{lk}}
\renewcommand{q^{BA}}{q_{ba}}
\renewcommand{p^{BA}}{p_{ba}}
\renewcommand{t_B{}^L}{t_b{}^l}
\renewcommand{t_A{}^K}{t_a{}^k}
\renewcommand{t_B{}^K}{t_b{}^k}
\renewcommand{t_A{}^L}{t_a{}^l}
\begin{equation}
\begin{aligned}
x^a x^b-q^{AB} x^b x^a&=0,\\
\x^a\x^b+p^{AB} \x^b \x^a&=0, \label{evsudb}
\end{aligned}
\end{equation}
\begin{align}
t_A{}^Kt_B{}^L-\frac{p^{BA}+q^{BA}}{p^{LK}+q^{LK}}\tblt_A{}^K&=\frac{p^{BA}p^{LK}-q^{BA}q^{LK}}
{p^{LK}+q^{LK}}t_B{}^Kt_A{}^L,\\
t_A{}^Kt_B{}^K-p^{BA}\tbkt_A{}^K=&0,\\
t_A{}^Kt_A{}^L-(q^{LK})^{-1}\talt_A{}^K&=0,
\end{align}
}
\bigskip
Definition 2.1 makes sense for just a single vector space
$V$. Here we obtain
the ``full subcategory'' of ${\mathsf L\mathsf i\mathsf n}_q$ consisting of all
linear transformations between
various quantum deformations of $V$ defined by the decomposition $I\oplus
J=V'\otimes V'$.
This quantum category is the adequate quantization of the semigroup $\operatorname{End} V$.
We see that the quantum categories language is more to the point
here. It is more flexible than the language of quantum (semi)groups.
\medskip
{\bf Example 2.8.} Fix a space $V,\quad \dim V=2$ and let us change
the quantization parameters in (\ref{evsudb}). Suppose $p^{21}=p$,
$q^{21}=q$,$p=p_{\a}$, $q=q_\a$. We arrive to a quantum category,
with indices $\a, \b$ that number the parameters $p$ and $q$ as objects.
The commutational relations for the matrix elements
of a ``homomorphism from $(V,\a)$ to $(V,\b)$'' in the natural notation
are:
\begin{align}
ab-q_\b^{-1}ba&=0, \label{ab} \\
ac-p_{\a} ca&=0, \label{ac} \\
ad-\frac{p_{\a}+q_\a}{p_{\b}+q_\b}da&=\frac{p_{\a}p_{\b}-q_\aq_\b}{p_{\b}+q_\b}cb,
\label{ad} \\
bc-\frac{p_{\a}+q_\a}{p_{\b}+q_\b}\pbq_\b cb&=\frac{\paq_\b-q_\ap_{\b}}{p_{\b}+q_\b}da,
\label{bc}\\
bd-p_{\a} db&=0, \label{bd}\\
cd-q_\b^{-1} dc&=0. \label{cd}
\end{align}
From~(\ref{ad}) and~(\ref{bc}) the relation between $b$ and $c$ can be
expressed also as
\begin{equation}
bc-\frac{p_{\b}+q_\b}{p_{\a}+q_\a}\paq_\a
cb=\frac{\paq_\b-q_\ap_{\b}}{p_{\a}+q_\a}ad.
\end{equation}
For $\a=\b$ these relations reduce to the commutational relations on
the quantum group $\operatorname{GL}(2)$.
\medskip
{\bf Example 2.9} ({\sl the determinant on a quantum category}).
In the situation of the previous example
consider the ``area form'' $\x^1\x^2$.
Substituting
$\x_{\b}^k=\x_{\a}^a t_a{}^k$, we obtain
\begin{equation}
\d(\x_{\b}^1\x_{\b}^2)=\x_{\a}^1\x_{\a}^2 \otimes \operatorname{det}_{\a \b}(T).
\end{equation}
The factor denoted by $\operatorname{det}_{\a \b}(T)$ is called, by definition,
the {\it determinant} of the matrix $T=(t_a{}^k)$.
Here $\x_{\a}^a $ are the coordinates on $(\P V,\a)$ and $ \x_{\b}^k$
the coordinates
on $(\P V,\b) $. From the definition and the
commutational relations
we obtain the equivalent formulas:
\begin{equation}
\begin{split}
\operatorname{det}_{\a \b}(T)&= ad-p_{\a} cb\\
&=\frac{p_{\a}+q_\a}{p_{\b}+q_\b}(da-q_\b cb) \\
&=\frac{1+\paq_\a^{-1}}{p_{\b}+q_\b}(q_\b ad-bc) \\
&=p_{\b}^{-1}(p_{\a} da-bc).
\end{split}
\end{equation}
For any ``morphisms $\a\longrightarrow\b, \quad \b\longrightarrow\c$''
the identity
\begin{equation}
\operatorname{det}_{\a \c}(T_{\a \c})=\operatorname{det}_{\a \b}(T_{\a \b})\operatorname{det}_{\b \c}(T_{\b \c}),
\end{equation}
holds. Here
$T_{\a \c}=T_{\a \b}T_{\b \c}$. Thus
$\det$ defines a one-dimensional representation of the quantum category in
consideration.
Mind that the function $\operatorname{det}_{\a \b}$ is defined not uniquely but up to
the ``adding of a coboundary'', i.e. up to a factor
$f_\a f_\b^{-1}$, where $f_\a$ is a non-vanishing scalar function
of the parameters
$p_{\a}, q_\a$. (Due to the non-uniqueness of the basis form
$\x_{\a}^1\x_{\a}^2$). The subcategory with $\det$
cohomologous to zero is a correct
quantum analogue of the
group
$\operatorname{SL}$.
\newcommand{c_{AB}}{c_{AB}}
\newcommand{c_{KL}}{c_{KL}}
\newcommand{\e_{AB}}{\e_{AB}}
\newcommand{\e_{BA}}{\e_{BA}}
\newcommand{\e_{BC}}{\e_{BC}}
\newcommand{\e_{AC}}{\e_{AC}}
\newcommand{\e_{KL}}{\e_{KL}}
\newcommand{B_+}{B_+}
\newcommand{B_-}{B_-}
\newcommand{B_{NM}{}^{KL}}{B_{NM}{}^{KL}}
\newcommand{B_{AB}{}^{CD}}{B_{AB}{}^{CD}}
\newcommand{\Bod}{B^{12}}
\newcommand{\Bdt}{B^{23}}
\newcommand{\l_{\a}}{\l_{\a}}
\newcommand{\l_{\b}}{\l_{\b}}
\newcommand{\l_{k}}{\l_{k}}
\newcommand{\l_{j}}{\l_{j}}
\newcommand{\operatorname{sign}(B-A)}{\operatorname{sign}(B-A)}
\newcommand{\operatorname{sign}(A-B)}{\operatorname{sign}(A-B)}
\newcommand{\operatorname{sign}(L-K)}{\operatorname{sign}(L-K)}
\newcommand{I_k}{I_k}
\newcommand{I_l}{I_l}
\newcommand{I_k^V}{I_k^V}
\newcommand{I_k^W}{I_k^W}
\newcommand{P_k^W}{P_k^W}
\newcommand{P_k^V}{P_k^V}
\newcommand{P_j^W}{P_j^W}
\newcommand{P_i^V}{P_i^V}
\newcommand{P_k}{P_k}
\section{ The Poincare -- Birkhoff -- Witt property.
\mbox{$R$-matrix} formulation and the connection with the Yang -- Baxter
equation.}
Consider the quantum category ${\mathsf L\mathsf i\mathsf n}_q$ with triples
$(V,I.J)$, $I\oplus J={V'\otimes V'}$ as objects.
It is possible to consider even more general case
of the quantum category with objects like
${(V,I_1,\dots,I_s)}, \quad \bigoplus I_k={V'\otimes V'}$,
the number $s$ is fixed (see the remark after Theorem 2.2).
Denote it by ${\mathsf L\mathsf i\mathsf n}_q^{(s)}, \, s\geqslant 2$. Then
${\mathsf L\mathsf i\mathsf n}_q={\mathsf L\mathsf i\mathsf n}_q^{(2)}$.
\medskip
{\bf Theorem 3.1.} \ {\it Let the matrix entries of $T_{\a \b}=(t_A{}^K)$
generate the algebra
$M_{\a \b}$, which is the function algebra on homomorphisms from
$\a={(V,I_1^V,\dots,I_s^V)}$ to
$\b={(W,I_1^W,\dots,J_s^W)}$, in the quantum category ${\mathsf L\mathsf i\mathsf n}_q^{(s)}$. Let
$T_{\a \b}^1:=T_{\a \b}\otimes1,\quad T_{\a \b}^2:=1\otimesT_{\a \b}$.
Then the defining commutational relations for functions on
${\mathsf L\mathsf i\mathsf n}_q^{(s)}$ can be presented in the following
``$R$-matrix form'':
\begin{equation}
B_{\a}(T_{\a \b}^1T_{\a \b}^2)=(T_{\a \b}^1T_{\a \b}^2)B_{\b}, \label{1}
\end{equation}
where for each $\a$ the matrix $B_{\a}$
is the linear combination of the projectors
on the subspaces $I_k$ along the sum
$\bigoplus\limits_{l\neq k}^{}I_l$, where the coefficients $\l_{k}$
are pairwise
different and independent on
$\a$. }
\medskip
{
\renewcommand{\d}{\bar\delta}
\newcommand{\d_{jj}}{\d_{jj}}
\newcommand{\d_{kj}}{\d_{kj}}
{\bf Proof.} Consider the partitions of unity $1=\sum P_k^V$, $1=\sum P_k^W$
by the projectors
corresponding to the direct sum decompositions
$V'\otimes V'=\bigoplusI_k^V$, $ W'\otimes W'=\bigoplusI_k^W$. Let
$A_{\a,k}=T(V)/(I_k^V)$, $A_{\b,k}=T(W)/(I_k^W) $. The homorphisms $\delta:
A_{\b,k}\longrightarrow A_{\a,k}\otimesM_{\a \b}, \quad k=1,\dots,s$, (the coaction)
are covered by the map of the tensor algebras
$\d$, and the commutational relations in $M_{\a \b}$ are defined by the condition
$\d(I_k^W)\subset I_k^V\otimesM_{\a \b} $ for all $k$. Any linear map
$A:W'\otimes W'\longrightarrow V'\otimes V'\otimesM_{\a \b}$ can be uniquely decomposed as
$A=\sumP_i^V\! AP_j^W=\sum A_{ij}$. The condition above is equivalent
to that the operator
$\d$ is ``diagonal'' on the tensor square:
$\d=\sum\d_{jj}$. Let us introduce
$B_{\a}=\sum\l_{k}P_k^V, B_{\b}=\sum\l_{k}P_k^W$, where $\l_{k}\neq\l_{j}, \, k\neq j$. Then
the difference of the left- and right-hand sides of
~(\ref{1})
is
$B_{\b}\circ\d-\d\circB_{\a}=\sum\limits_{k,j}(\l_{k}-\l_{j})\d_{kj}$, which vanishes if
and only if
$\d_{kj}=0$ for $k\neq j$. That is just
the condition defining the commutational relations in
$M_{\a \b}$. $\quad\square$
}
\bigskip
In the index notation the relation (\ref{1})
can be rewritten as follows:
\begin{equation}
\tant_B{}^MB_{NM}{}^{KL}=\Babcdt_C{}^Kt_D{}^L, \label{2}
\end{equation}
where $B_{\a}=(B_{AB}{}^{CD}),\quad B_{\b}=(B_{NM}{}^{KL})$.
The claim of the theorem is analoguous to the
corresponding result for quantum groups,
where the relations are determined by
a single matrix $B$.
\medskip
For any quantum space $(V,I_1,\dots,I_s)$ the projectors $P_k$
can be expressed from the matrix $B_{\a}$ (``$R$-matrix'')
as the projections on its eigenspaces.
The commutational relations in the algebras
$A_k=T(V')/(I_k)$ are also expressed in terms of the operator $B_{\a}$.
Suppose one has the commutational relations in the $R$-matrix form.
It is common then to relate ``the Poincare -- Birkhoff -- Witt property''
(the equality of the graded dimension of function algebra on quantum space
or quantum group to that of the corresponding classical polynomial algebra)
with the identity
\begin{equation}
\Bod\Bdt\Bod=\Bdt\Bod\Bdt \label{YB}
\end{equation}
for the matrix $B$, where $\Bod=B\otimes1, \Bdt=1\otimes B$
(the Yang -- Baxter equation). Strictly speaking, no general statement exists
here and one can only say that a certain ``relation'' takes place.
(As Joseph Donin told me after this work was completed,
a sort of theory can be actually developped dealing with the quantum
algebra dimension in connection with the YB equation. This is the topic
of the recent paper~\cite{don} by him and Steve Shnider.)
Consider the case of ${\mathsf L\mathsf i\mathsf n}_q={\mathsf L\mathsf i\mathsf n}_q^{(2)}$, $I_1=I, I_2=J$.
We shall normalize the matrices
$B$ so as $B=P_1-\l P_2,\quad \l\neq-1$, and we shall look for $\l$ suiting
(\ref{YB}).
In general this is an overdetermined problem.
One can notice
(see, for example, ~\cite{dem2}), that if the solutions exist there
may be either exactly two of them:
$B_+=P_1-\l P_2$ and $B_-=P_1-\l^{-1}
P_2$, where $\l\neq1$, or just one: $B=P_1-P_2$, in the case the equation
(\ref{YB}) is satisfied for
$\l=1$. (In the last case an additional identity holds, $B^2=1$.)
Applying this to our situation, we obtain that if for two quantum spaces
$\a=(V,I_1^V,I_2^V)$ and $\b=(W,I_1^W,I_2^W)$ the normalized
matrices $B_{\a}$ and $B_{\b}$ suit the Yang -- Baxter equation, and we fix the
choice of $B_{\a}$ and $B_{\b}$,
then the commutational relations in $M_{\a \b}$
can be presented in the ``$R$-matrix form''(\ref{1}),(\ref{2}) with the
given $B_{\a}$ and $ B_{\b}$ if and only if
$\l_{\a}=\l_{\b}$ or $\l_{\a}=\l_{\b}^{-1}$.
\medskip
Let us turn now to problem of the graded dimension of the algebras $M_{\a \b}$.
We consider the subcategory with defining commutational relations for
quantum spaces of the form (\ref{sudb}).
Let $\a=(V,P_V,Q_V),
\quad \b=(W,P_W,Q_W)$ with the matrices of parameters $P_V=(p^{AB}), Q_V=(q^{AB}),
P_W=(p^{KL}), Q_W=(q^{KL})$. For arbitrary orderings of the tensor indices in
$V$ and $W$ let $\operatorname{sign}(B-A)$ equal $+1, 0, -1$ if $B>A,\
B=A,\ B<A$ respectively. In a similar way we define the function $\operatorname{sign}(L-K)$.
The matrix entries will be ordered
by rows: $t_A{}^K<t_B{}^L$, if $A<B$ or if $A=B,\quad K<L$. The {\it
ordered monomial\/} is defined as usual to be such that
the sequence of letters is not decreasing and any odd variable can appear
no more than once.
\medskip
{\bf Theorem 3.2.} \ {\it The algebra $M_{\a \b}$ is of classical dimension iff
the bases in $V$ and $W$ can be ordered in such a way that
the following equations hold
\begin{align}
p^{AB}&=q^{AB} c_\a^{\operatorname{sign}(B-A)},\label{a} \\
p^{KL}&=q^{KL} c_\b^{\operatorname{sign}(L-K)},
\label{4}
\end{align}
for some constants $c_\a,\/c_\b $, not equal to $0$ or $-1$, and either
\begin{equation}
c_\a=c_\b, \label{eq}
\end{equation}
or
\begin{equation}
c_\a=c_\b^{-1}, \label{inv}
\end{equation}
In this case the ordered monomials in matrix entries span the algebra $M_{\a \b}$.}
\medskip
{\bf Proof.} It actually consists of two parts.
First choose an arbitrary ordering of the tensor indices.
We shall show that the ordered monomials span the whole algebra
$M_{\a \b}$
for any values of the parameters $p^{AB}, q^{AB}, p^{KL},q^{KL}$.
Indeed, consider an arbitrary monomial in
$t_A{}^K$.
Consider the elements of the first row in it.
Let $\t$ stand for the number of the inversions with the elements
of other rows. That is the number of cases when the elements of the first
row appear to the right of the elements of other rows.
Take any neighbouring pair of elements
of the sort
${t_A{}^K t_1{}^L}, A>1 $.
By the commutational relations (\ref{rel})
it is possible to
substitute it by a linear combination
of the products $t_1{}^L t_A{}^K$ ... $t_1{}^Kt_A{}^L$,
and the number $\t$ will decrease by one.
Thus a finite iteration of this procedure
leads to a linear combination of the monomials
such that in each of them the elements of the first row
stand to the left of the product of the elements of the other rows.
We can apply the same method to these products, now taking the second row
instead of the first, etc.
(If at some step the elements of a certain row are absent, then we take the
next row.) As there are finite number of rows, after a finite steps we shall
obtain a linear combination of monomials with the correct order of the row
numbers. The elements inside the same row (they all are neighbours now)
are put in the correct order by the formulas (\ref{rel2}). Now the arbitrary
monomial from which we have started is expressed as linear combination of
the ordered mononials.
Such algorithm for a quantum deformation
of the supergroup $\operatorname{GL}\,(n\mid m)$ was described by Manin
~\cite{man2} with the reference to Yu.Kobyzev.
The second part of the proof deals with the linear independence of
the ordered monomials. This is much
more non-trivial problem. The so-called ``diamond lemma'' of the combinatorial
theory of rings
(see ~\cite{lat}, ~\cite{dem3}) implies that for quadratic algebras it is
necessary and sufficient to check the independence only
for the monomials of the third order.
(The lemma provides in this case the linear independence also for all
ordered monomials of the degree~ $\geqslant 3$.)
Let us look for the criterion of the independence for ordered cubic
monomials in
$t_A{}^K$.
The linear independence of ordered monomials is equivalent to the uniqueness
of the ``normal form'' of any arbitrary monomial, i.e. to the independence
``on path'' of the result of its expression
as a combination of the ordered monomials.
We need to finger all cubic monomials
with the broken order and for each of them to compare all the possible ways
to put them in the normal order.
Let ${a<b<c}$ be the arbitrary letters symbolizing
the matrix entries.
Then the cubic monomials with the broken order are the following:
$acb$, $bac$,
$bca$,
$cab$, $cba$ (three letters different) and $aba$, $ba^2 $ (two letters
different).
In the process of putting them into order
the new letters can arise.
Which ones (together with the ways of putting into order) actually
depends on the position of the elements
$a,b,c$ in the matrix $(t_A{}^K)$. It is most convenient
to draw pictures on which the matrix entries
are shown as the vertices of an orthogonal lattice.
By (\ref{rel}) the ``commutator closure'' of the given set of letters
consists of the vertices of the least sublattice containing this set.
Other letters cannot appear in the process of putting the monomial into
normal form.
For three pairwise different entries we obtain one picture with
six additional letters (if the initial elements stand on three different
rows) and two pictures with three additional letters (if initially we have
elements in two different rows).
For two different letters a picture with two additional letters appears.
(The pictures with the initial letters in one row do not include any
additional letters and they cannot produce any ambiguity.
The staightforward checking of all possible ways leading
to the sum of ordered monomials in each of these
$5\times 3+2=17$ cases shows that the non-uniqueness of the normal form
is apriori possible only for the monomials like
$cba$ (for all three variants of the position of the initial elements),
i.e. when branching of the algorithm takes place at the very first step.
(Have in mind that the conclusion is strongly based on the exact form of
the commutational relation.)
Happily no branching is possible here at the following steps.
In the most complex case with six additional letters two branches of the
algorithm look as follows:
$cba\longrightarrow{(bca,xya)}\longrightarrow{(bac,buv,xay,xtv)}\longrightarrow
{(abc,tsc,ubv,txv,axy,usy,txv,ubv)}$ and
$cba\longrightarrow{(cab,cts)}\longrightarrow{(acb,uvb,tcs,uys)}\longrightarrow
{(abc,axy,ubv,usy,tsc,txv,usy,ubv)} $.
Here all the arising monomials are successively written down.
In two other cases the branches are a bit shorter.
The final list of the ordered monomials is the same for both branches.
That means that the uniqueness of the normal form is equivalent
to the agreement of numeric factors in the similar terms.
The explicit calculation leads us to the 12 equations
for the parameters $p^{AB},q^{AB},p^{KL},q^{KL}$. Five of these equations happen to
be identities.
Let us give as example two of the non-trivial equations:
\begin{multline}
\frac{(\pabnq^{LN}-\qabnp^{LN})(\pacnp^{KN}-\qacnq^{KN})(\pbcn+\qbcn)}{(p^{KN}+q^{KN})
(p^{LN}+q^{LN})} +\\
\frac{(\pabn+\qabn)(\pacnp^{KL}-\qacnq^{KL})(\pbcnp^{LN}-\qbcnq^{LN})}{(p^{KL}+q^{KL})
(p^{LN}+q^{LN})}\\=
\frac{(\pabnp^{KL}-\qabnq^{KL})(\pacn+\qacn)(\pbcnp^{KN}-\qbcnq^{KN})}{(p^{KL}+q^{KL})
(p^{KN}+q^{KN})}, \label{7}
\end{multline}
if ${A<B<C}$, and
\begin{multline}
\frac{2(\pabn+\qabn)(\pabnp^{KN}-\qabnq^{KN})}
{(p^{KN}+q^{KN})(p^{NL}+q^{NL})(p^{LN}(1-(-1)^{\tilde B})+q^{LN}(1+(-1)^{\tilde B}))}=\\
\frac{2(\pabn+\qabn)(\pabnp^{KN}-\qabnq^{KN})}
{(p^{KN}+q^{KN})(p^{KL}+q^{KL})(p^{LK}(1-(-1)^{\tilde B})+q^{LK}(1+(-1)^{\tilde B}))}+\\
\frac{(\pabnp^{KL}-\qabnq^{KL})(\pabnp^{LN}-\qabnq^{LN})}
{(p^{KL}+q^{KL})(p^{LN}+q^{LN})}, \label{8}
\end{multline}
if $A<B, L<N$. To get the solution we shall make a reduction.
Substitute $K=N$ into ~(\ref{8}). Two terms cancel immediately
and we arrive to
$(\pabnp^{NL}-\qabnq^{NL})(\pabnp^{LN}-\qabnq^{LN})=0$,
which is equivalent to
\begin{equation}
\frac{\pabn}{\qabn}=\frac{p^{LN}}{q^{LN}} \quad\text{ or }\quad
\frac{\pabn}{\qabn}=\frac{q^{LN}}{p^{LN}} \label{9}
\end{equation}
(recall that $p^{LN}=(p^{NL})^{-1}, q^{LN}=(q^{NL})^{-1}$). Let $\pabn/\qabn=c_{AB},
\pkln/\qkln=c_{KL}$. The equations ~(\ref{9}) are equivalent to
\begin{equation}
c_{AB}+(c_{AB})^{-1}=c_{KL}+(c_{KL})^{-1} \label{10}
\end{equation}
for any $A<B,\quad K<L$. That means that both the right-hand side and the
left-hand side of~(\ref{10}) do not depend on indices and are equal to some
constant
$\m\neq 0$. Thus for all
$A,B \quad c_{AB}=c^{\e_{AB}}$, where $c: c+c^{-1}=\m$, $ \e_{AB}+\e_{BA}=0$,
$\e_{AB}=\pm 1$ if $A\neq B$, and the same for $c_{KL}=c^{\e_{KL}}$. Let $c\neq
1$. If the matrices $(\e_{AB}), (\e_{KL})$ do not possess the ``transitivity"
property
($\e_{AB}=1, \e_{BC}=1$ implies $\e_{AC}=1$), then the equations are not satisfied.
Indeed, substituting, for example, $\pabn=c\cdot\qabn$,
$\pbcn=c\cdot\qbcn$, $\pacn=c^{-1}\cdot\qacn$
into~(\ref{7})
and setting $K=L=N$, we obtain a contradiction (for $c\neq 1$).
It follows that if it is not true that $\e_{AB}=\operatorname{sign}(B-A), \e_{KL}=\operatorname{sign}(L-K)$
for some ordering of the bases
of $V$ and $W$ (perhaps different from the originally chosen),
then the dimension of the algebra $M_{\a \b}$ is less than the classical value.
(The ordered cubic monomials, w.r.t. the original ordering, will be linear
dependent and according to what was proven above they actually span the
subspace of all cubic functions.)
Assume now that this is true. We write the same equations on the
deformation parameters w.r.t. the new ordering.
To finish the proof it is
necessary to substitute in them the equations
~(\ref{4}), ~(\ref{eq}). Now the direct, rather laborous
check shows that all the seven equations are actually satisfied.
The case of $c=1$ is also contained here; it is indifferent to the choice
of ordering.
$\quad\square$
\bigskip
We actually proved that it is enough to consider the case
(\ref{eq}), for a suitable ordering. From the other point, it was known that
for a single space
$\a=(V,I_1,I_2)$ the possibility to choose the order of the basis in such a
way that the equation
(\ref{a}) holds for a certain $c_\a$ is equivalent both to the PBW property
for the algebra
$M_{\a \a}$ and to the Yang -- Baxter equation for
${B_\pm=P_1-(c_\a)^{\pm1}P_2}$ \ (see ~\cite{dem2},~\cite{dem3}).
The key point of our result is that the equality of ``quantum constants''
$c_\a=c_\b^{\pm1}$ for two spaces is exactly the criterion for the classical
value of the dimension of the algebra
$M_{\a \b}$ (the ``Poincare -- Birkhoff -- Witt'' for $M_{\a \b}$).
This criterion is in excellent agreement
with our heuristic consideration above, which was based on the Yang --
Baxter equation,
while there is no direct logical connection.
For $c_\a\neq1$ one can choose one of the two orderings compatible with
(\ref{a}) (direct or reverse),
i.e. to choose one of the two
``Yang--Baxter structures" $B_+$ or $B_-$ on $(V,I_1,I_2)$,
which could be viewed as a sort of ``orientation''. The statement of the
theorem takes it into account.
\smallskip
It is convenient to make a slight change of the notation.
Namely, instead of $c_\a$ we introduce $\l_\a: \/ c_\a=\l_\a^2$ and
rescale the parameters, $q^{AB}:=q^{AB}\l_\a^{\operatorname{sign}(A-B)}$.
Then the commutational relations will become more symmetric.
Fix now a number $\l\neq0,\pm i$. Define a quantum category ${\mathsf L\mathsf i\mathsf n}_q^+(\l)$,
with objects which are triples like $(V,Q,\pm1)$ with the following relations
\begin{equation}
\begin{aligned}
x^A x^B-q^{AB}\l^{\pm\operatorname{sign}(A-B)}x^B x^A=0,\\
\x^A\x^B+(-1)^{\tilde A+\tilde B}q^{AB}\l^{\mp\operatorname{sign}(A-B)}\x^B \x^A=0,
\end{aligned} \label{space}
\end{equation}
(the basis in $V$ is defined uniquely up to the scaling of the basis vectors).
Then the matrix entries of the ``homomorphisms from $(V,Q_V,\e)$ to
$(W,Q_W,\h)$'' are subject to the following commutational relations
implied by
(\ref{rel}-\ref{rel2}) ... (\ref{space}). For the elements of one row, if
$K<L$:
\begin{equation}
\begin{aligned}
t_A{}^Kt_A{}^L-q^{KL}\l^{-\h}\talt_A{}^K&=0, \text{ $\tilde A=0$,}\\
t_A{}^Kt_A{}^L+(-1)^{\tilde K+\tilde L}q^{KL}\l^{\h}\talt_A{}^K&=0, \text{ $\tilde A=1$}.
\end{aligned}
\end{equation}
For the elements of one column, if $A<B$:
\begin{equation}
\begin{aligned}
t_A{}^Kt_B{}^K-(\qabn)^{-1}\l^{-\e}\tbkt_A{}^K&=0, \text{ $\tilde K=0$,}\\
t_A{}^Kt_B{}^K+(-1)^{\tilde A+\tilde B}(\qabn)^{-1}\l^{\e}\tbkt_A{}^K&=0, \text{ $\tilde K=1$}.
\end{aligned}
\end{equation}
For the ``diagonal" elements ($A<B,\/ K<L$):
\begin{multline}
t_A{}^Kt_B{}^L-(\qabn)^{-1}q^{KL}(-1)^{\tilde A\tilde L+\tilde B\tilde K}\tblt_A{}^K=\\
(\qabn)^{-1}\frac{\l^{-\e-\h}-\l^{\e+\h}}{\l^{-\h}+\l^{\h}}
(-1)^{(\tilde A+\tilde B)\tilde K}t_B{}^Kt_A{}^L
\label{diag}
\end{multline}
For the ``antidiagonal" elements ($A<B, K>L$):
\begin{multline}
t_A{}^Kt_B{}^L-(\qabn)^{-1}q^{KL}(-1)^{\tilde A\tilde L+\tilde B\tilde K}\tblt_A{}^K=\\
(\qabn)^{-1}\frac{\l^{-\e+\h}-\l^{\e-\h}}{\l^{-\h}+\l^{\h}}
(-1)^{(\tilde A+\tilde B)\tilde K}t_B{}^Kt_A{}^L \label{adiag}
\end{multline}
Notice that for $\e=\h$ the right-hand side of (\ref{adiag}) vanishes, and
for
$\e=-\h$ so do the right-hand side of (\ref{diag}). The factor in the
right-hand side
of (\ref{diag}) or (\ref{adiag}) respectively reduces to
$\l^{-\e}-\l^{\e}$ (in both cases).
Now, if one will fix the classical space \ $V,\quad {\dim V=2}$\ and
change only the quantization parameters
$q^{AB}$, he will arrive to the example described in the introduction.
\newcommand{A_i}{A_i}
\newcommand{A_j}{A_j}
\newcommand{A_k}{A_k}
\newcommand{A_x}{A_x}
\newcommand{A_y}{A_y}
\newcommand{A_z}{A_z}
\newcommand{A_r}{A_r}
\newcommand{A_s}{A_s}
\newcommand{{m_{ijk}}}{{m_{ijk}}}
\newcommand{{m_{x\a r}}}{{m_{x\a r}}}
\newcommand{{m_{y\b s}}}{{m_{y\b s}}}
\newcommand{{\D_{xyz}}}{{\D_{xyz}}}
\newcommand{{\D_{krs}}}{{\D_{krs}}}
\newcommand{{\D_{ixy}}}{{\D_{ixy}}}
\newcommand{{\D_{j\a\b}}}{{\D_{j\a\b}}}
\section*{Appendix.}
The general definition of a {\it bialgebra $A=(A_i)$ indexed
by the set}
$I$ is as follows. We should fix the relations $P\subset I^3,\/Q\subset
I^3$ and define the linear maps
\begin{align}
{m_{ijk}}:A_i\otimesA_j\longrightarrowA_k & \text{ for all } (i,j,k)\in P, \notag\\
{\D_{xyz}}:A_x\longrightarrowA_y\otimesA_z & \text{ for all } (x,y,z)\in Q, \notag
\end{align}
such that the diagram
$$
\begin{CD}
A_i\otimesA_j @>{{m_{ijk}}}>> A_k @>{{\D_{krs}}}>> A_r \otimes A_s \\
@V{{\D_{ixy}}\otimes{\D_{j\a\b}}}VV @. @AA{{m_{x\a r}}\otimes{m_{y\b s}}}A \\
A_x\otimesA_y\otimesA_{\a}\otimesA_{\b} @. \longleftrightarrow @.
A_x\otimesA_{\a}\otimesA_y\otimesA_{\b}
\end{CD}
$$
is commutative when the maps in it make sense.
It is obvious that this definition
is self-dual: if ${(A_i,I,P,Q,m,\D)}$ is a bialgebra, then
${(A_i',I,Q,P,\D',m')}$ is also a bialgebra.
This definition contains in particular:
graded and almost graded algebras, coalgebras and bialgebras (in the usual
sense), function algebras on categories, additive categories,
function algebras on quantum categories,
algebras dual to them, etc.
Rather exotic
diagonals one can find, for example,
in the Odessky -- Feigin algebras~\cite{of}.
The possibility that need to be also considered
is that the indices
$i\in I$ may form not just a set but topological space or a smooth
manifold.
The natural farther step is to allow the indices
to be coordinates on a supermanifold or even a
``quantum manifold''.
In the context of this paper that means to allow to quantize the
quantization parameters themselves.
|
\section{Introduction}
For statistical systems in equilibrium or
near equilibrium critical phenomena arise around the second order
phase transition points. Due to the infinite spatial and time
correlation lengths there appear universality and scaling.
The universal behaviour of critical systems is characterized by
the critical exponents. The determination of critical exponents
has long been one of the main interests for both analytical
calculations and
numerical simulations.
Numerically critical exponents are usually measured by
generating the configurations in the
equilibrium with Monte Carlo methods. To obtain the critical
exponents from the finite size scaling,
Binder's method is widely accepted \cite{bin92,bin81}.
The dynamical exponent $z$ is
traditionally measured from the exponential decay of the
time correlation for finite systems in the long-time regime
\cite{wan91,wil85}.
As is well known, numerical simulations
near the critical point suffer from critical slowing
down. Much effort
has been made to circumvent this difficulty.
To study the static properties of the system,
some {\it non-local} algorithms, e.g., the cluster algorithm
\cite{swe87,wol89}, have proved to be very efficient compared to
the normal {\it local} algorithms. However, in this case
the original dynamic
universality class is altered by the non-locality of the algorithm.
Properties of the original local dynamics can not be obtained with
non-local algorithms.
In recent years the exploration of critical phenomena
has been broadened. Universality and scaling are also
discovered for systems
far from equilibrium. Better understanding has been achieved
of the critical relaxation process even up to the early time.
A representative example for such a process is
that the Ising model initially in a random state with
a small magnetization
is suddenly quenched to the critical temperature and
then evolves according to the dynamics of model A.
Janssen, Schaub and Schmittmann \cite{jan89} have
argued by an $\epsilon$-expansion up to two--loop order that,
besides the well known universal behaviour in the long-time
regime, there exists
another {\it universal} stage of the relaxation
{\it at early times}, the
so-called {\it critical initial slip}, which sets in right
after the
microscopic time scale $t_{mic}$.
The characteristic time scale for the critical initial slip
is $t_0 \sim m_0^{-z/x_0}$, where $m_0$ is the initial
magnetization and
$x_0$ is the dimension of it. It has been shown that $x_0$ is
a new
independent critical exponent for describing the critical
dynamic system.
The characteristic behaviour of the critical initial slip
is that, when a non-zero initial magnetization $m_0$ is generated,
due to the anomalous dimension of the operator $m_0$
the time dependent magnetization $M(t)$ undergoes a critical
initial increase
\begin{equation}
M(t) \sim m_0 \, t^\theta,
\label{cis}
\end{equation}
where $\theta$ is related to $x_0$ by $x_0 = \theta z + \beta/\nu$.
The exponent
$\theta$ has been measured with Monte Carlo simulation
for the Ising model
and the Potts model
both directly from the power law increase of the magnetization
in (\ref {cis}) \cite {li94,sch95} and
indirectly from the power law decay of the auto-correlation
\cite{hus89}-\nocite{hum91}\cite{men94}. The results are in good
agreement with those
from an $\epsilon$-expansion and the scaling relation is
confirmed.
In a previous paper \cite{li95}
we proposed to measure both the dynamic and static
exponents from the finite size scaling of the dynamic relaxation
at the early time. The idea is demonstrated for the 2-dimensional
Ising model. Since the measurement is carried out
from the beginning of the time evolution, the method is efficient
at least for the dynamic exponent $z$. Even though certain
aspects of this dynamic approach should still
be clarified, the results indicate a possible broad application
of the short-time dynamics
since the universal behaviour of the dynamics at early time
is found to be quite general
\cite{oer93}-\nocite{oer94}\nocite{bra94}\nocite{rit95}
\nocite{bra95}\cite{gra95}.
One of the purposes of the present paper is
to give a detailed and complete
analysis of the data briefly reported in \cite{li95}.
While in that letter we have extracted the exponents by the
optimal fit of two curves in a certain time interval
(``global'' fit), we propose here in addition a new approach by which
the critical exponents are obtained for each time step separately
(``local'' fit).
Furthermore the
simulation has been extended to a longer time interval in order
to confirm the stability of the measurements
in the time direction and to see how the scaling possibly
passes over to the long-time regime.
On the other hand, we may easily realize that
for the scaling
of the short-time dynamics a small initial magnetization is
important besides the short
initial correlation length.
This is because $m_0=0$ is a fixed point for the
renormalization group transformation.
However, there exits also another fixed point
corresponding to $m_0=1$.
Therefore one may like to know whether around that fixed point
universality and scaling are
also present or not.
Actually some trials have been made with Monte Carlo
simulation \cite{sta92,mun93}.
For a large enough lattice, one may expect a power
law decay of the magnetization
\begin{equation}
M(t) \sim t^{-\frac{\beta}{\nu z}}
\end{equation}
before the exponential decay starts. From this behaviour
the exponent $\beta/{(\nu z)}$ can be estimated.
However the results are not yet complete.
Therefore another purpose of this paper is
to present a systematic investigation
of the finite size scaling for the critical relaxation
starting from $m_0=1$. It will be shown
that scaling is observed in the early stage of the time
evolution and with the help of the finite size scaling
all the critical exponents $z$, $\beta$ and $\nu$
can be obtained from the lattices which are much smaller
that those in \cite{sta92,mun93}.
The following section~2 is devoted to
the critical relaxation with $m_0=0$ and Sect.~3 to $m_0=1$.
The final section contains some discussion.
\section{The Critical Relaxation \protect\newline with Zero
Initial Magnetization}
The Hamiltonian for the Ising model is
\begin{equation}
H=J \sum_{<ij>} S_i\ S_j\;,\qquad S_i=\pm 1\;,
\label{hami}
\end{equation}
with $<ij>$ representing nearest neighbours.
In the equilibrium the Ising model is exactly solvable.
The critical point locates at $J_c=\log(1+\surd 2)/2$,
and the exponents $\beta=1/8$ and $\nu=1$ are known.
In principle any type of the dynamics can be given to the system
to study the non-equilibrium evolution processes.
Unfortunately up to now none of them can be solved
exactly.
In this paper we consider only the dynamics of model A.
For the numerical simulation, typical examples are
the Monte Carlo Heat-Bath algorithm and the Metropolis
algorithm. For the analytical calculation, the Ising model
should be assumed to be described
by the $\lambda \phi ^4$ theory. Then the Langevin equation
can be introduced as a dynamic equation.
For the Langevin dynamic system the renormalization group method
may be applied to understand the critical behaviour
as universality and scaling.
For the critical relaxation with the initial condition
of a very short correlation and small magnetization,
Janssen, Schaub and Schmittmann \cite{jan89} have performed a
perturbative renormalization calculation
with an $\epsilon$-expansion
up to two--loop order. They have obtained the scaling relation
which is valid even
in the short-time regime, and all
the critical exponents including the new dynamic exponent
$\theta$ which governs the initial behaviour of the critical
relaxation.
Of special interest is here
the extension of the scaling form in Ref.~\cite{jan89} to
finite-size systems \cite{die93,li94}.
In accordance to the renormalization group analysis
for finite-size
systems,
after a microscopic time scale $t_{mic}$ we expect
a scaling relation to hold for the k-th moment of the
magnetization in the neighbourhood of the critical point
starting from the macroscopic short-time regime
\cite{pri84,jan89,zin89},
\begin{equation}
M^{(k)}(t,\tau,L,m_{0})=b^{k\beta/\nu}
M^{(k)}(b^{z}t,b^{-1/\nu}\tau,bL,
b^{-x_{0}}m_{0})
\label{e2}
\end{equation}
assuming that the initial correlation length is zero
and the initial magnetization $m_0$ is small enough.
Here $t$ is the dynamic evolution time,
$\tau=(T-T_{c})/T_{c}$ is the reduced temperature,
$L$ is the lattice size,
and $b$ is the spatial rescaling factor.
It has been discussed that under certain conditions
the effect of $m_0$ remains even in the long-time regime
of the critical relaxation \cite {rit95a}. This modifies the
traditional scaling relation where the effect of $m_0$
has usually been suppressed.
In this paper we are only interested in the measurement
of the well-known critical exponents $z$, $\beta$, and $\nu$.
To make the computation simpler and more efficient,
we set $m_{0}$ to its
fixed point $m_{0}=0$ in this section.
Therefore the exponent $x_{0}$
will not enter
the calculation. Furthermore, now the time scale
$t_{0}=m_0^{-z/x_0} \to \infty$, and the
critical initial slip gets most prominent in time direction
even though the magnetization itself will only fluctuate
around zero.
The initial state with $m_0=0$
is prepared by starting from a lattice with all spins equal, then
the spins of randomly chosen sites are switched,
until exactly half of the spins are opposite.
This initial state is updated with the Heat-Bath algorithm to
$300$ time steps for $L=8, 16, 32$, and to
$900$ time steps for $L=64$.
The average over 50\,000 samples of this kind
with independent initial configurations has been taken in each run,
and 8 runs are used to estimate the errors.
The critical value $J_c=0.4406$ has been used and, in order
to fix $1/\nu$ separately, we have repeated all simulations with
$J=0.4386$.
In each case the observables $|M(t)|$,
$ M^{(2)}(t)$ and $M^{(4)}(t)$
have been measured.
To determine $z$ {\it independently}, we introduce a
{\it time-dependent}
Binder cumulant
\begin{equation}
U(t,L) =1-\frac{ M^{(4)} }{ 3 (M^{(2)})^2}
\label{ez}
\end{equation}
Here the argument $\tau$ has been set to zero and skipped.
\
The simple finite size scaling relation
\begin{equation}
U(t,L)=U(t',L');\qquad t' = b^z t, \qquad L' = b L
\label{e4}
\end{equation}
for the cumulant
is easily deduced from Eq.(\ref{e2}).
The exponent $z$ can easily be obtained through
searching for a time scaling factor $b^{z}$ such that
the cumulants from two different lattices
in both sides in Eq.(\ref{e4}) collapse.
We call this global scaling fitting.
\begin{table}[h]\centering
$$
\begin{array}{|c|c|c|c|l|l|l|}
\hline
\mbox{Input} & \mbox{Lattice} & t'_{min} & t'_{max} &\quad z\ &
\quad 2\beta/\nu\ &\quad 1/\nu\ \\
\hline
\hline
&\ 8\leftrightarrow 16 & \ 10 & 300 & 2.100(2) &
0.2473(02) & 1.13(1) \\
\cline{5-7}
& 16\leftrightarrow 32 & \ 10 & & 2.149(2) &
0.2494(06) & 1.08(4) \\
\cline{2-7}
U & & \ 50 & 300 & 2.134(4) &
0.2510(11) & 1.00(5) \\
\cline{5-7}
& 32\leftrightarrow 64 & 200 & & 2.140(4) &
0.2488(11) & 1.03(4) \\
\cline{3-7}
M^2 & & \ 50 & 900 & 2.151(2) &
0.2531(08) & 1.02(2) \\
\cline{5-7}
& & 200 & & 2.153(2) &
0.2523(08) & 1.02(2) \\
\hline
\hline
\tilde U & 32\leftrightarrow 64 & \ 50 & 900 & 2.151(3) &
0.2515(11) & 1.03(2) \\
\cline{5-7}
|M| & & 200 & &2.152(3) &
0.2521(11) & 1.03(2) \\
\hline
\end{array}
$$
\caption{
\footnotesize
Results for $z$, $2\beta/\nu$ and $1/\nu$, respectively,
from the 2-dimensional Ising model with initial magnetization $m_0=0$.
The values are
obtained from a global scaling fit for two lattices.
}
\label{t1}
\end{table}
\begin{table}[t]\centering
$$
\begin{array}{|c|c|c|c|l|l|l|}
\hline
\mbox{Input} & \mbox{Lattice} & t'_{min} & t'_{max} &\quad z\ &
\quad 2\beta/\nu\ &\quad
1/\nu\ \\
\hline
\hline
&\ 8\leftrightarrow 16 & \ 10 & 300 & 2.333(4) &
0.2372(08) &
1.11(1) \\
\cline{2-7}
U & 16\leftrightarrow 32 & \ 10 & 300 & 2.143(2) &
0.2448(11) &
1.19(5) \\
\cline{2-7}
M^2 & 32\leftrightarrow 64 & \ 50 & 900 & 2.148(3) &
0.2510(16) &
1.07(2) \\
\cline{5-7}
& & 200 & & 2.155(2) &
0.2488(10) &
1.05(2) \\
\hline
\hline
\tilde U & 32\leftrightarrow 64 & \ 50 & 900 & 2.144(4) &
0.2541(28) &
1.07(2) \\
\cline{5-7}
|M| & & 200 & & 2.153(3) &
0.2501(14) &
1.05(2) \\
\hline
\end{array}
$$
\caption{
\footnotesize
Results for $z$, $2\beta/\nu$ and $1/\nu$, respectively,
from the $2$-dimensional Ising model with initial magnetization
$m_0=0$.
Values are from an
average in time direction with a local scaling fit.
}
\label{t2}
\end{table}
Here the cumulant $U(t,L)$ obtained from each of the 8 runs
for lattice size $L$
has been compared with each run for $L'=2L$. The best scaling
factor $2^{z}$ has
been estimated by the method of least squares.
Figure~\ref{ddU} shows the cumulants for $L=8$,$16$,$32$ and $64$ by
solid lines. The dots fitted to the lines for $L$ up to $32$ show the
results for $L'=2L$ rescaled in time
with the best fitting scaling factor $2^z$.
Only a selected number of 30 equidistant points has been plotted.
One sees the remarkable scaling collapse even in the short-time
regime. Compared with the figure shown
in the previous paper \cite{li95}
the evolution time for $L'=64$ has been extended up to
$t'_{max}=900$.
The average value for $z$ and
the error estimated from this procedure have been given in
Tab.~\ref{t1}
for different pairs of lattices using Eq. (\ref{e4}).
In the first steps of the time evolution, the
values of $M^{(2)}$ and $M^{(4)}$
are quite small.
A careful view at the data shows that
their accuracy and in particular the accuracy of the cumulant
$U(t,L)$ is not so good as for lager $t$. One may expect
that skipping the region of smaller t will give more reliable
results.
Therefore we have performed fits for
different time intervals $[t'_{min},t'_{max}]$.
The results in reference \cite{li95} corresponds to
$t'_{min}=1$ and $t'_{max}=300$. Fig.~\ref{ddU} is from
a fit with $t'_{min}=50$ and $t'_{max}=900$.
The longer evolution time of $t'_{max}=900$ for lattice
$L=64$ shows also the stability of the
scaling in the time direction. From the results
we can see that $z$ for larger time $t$ is slightly bigger
than that for smaller time $t$.
Later we will come back to this point.
With $z$ in hand, the scaling relation for the second moment
\begin{equation}
M^{(2)}(t,L)=b^{2\beta/\nu} M^{(2)}(t',L');
\qquad t'=b^{z}t \qquad L'=bL,
\label{e5}
\end{equation}
can be used to estimate the exponent $2\beta/\nu$ in a similar way.
The results have been included in Tab.~\ref{t1}.
The curves for
$M^{(2)}$ for the different lattice sizes and the corresponding
scaling fits
can be found in Fig.~\ref{ddM}. We have cut the time scale
at $t=220$ in order to show the data relevant for the
scaling fit more clearly.
Slightly more complicated appears the determination of
$1/\nu$. One can use the derivative with respect to $\tau$ either
of $U$ or of $M^{(2)}$. Here the latter gives
more stable results. From
\begin{equation}
\partial_\tau \ln M^{(2)}(t,\tau,L)|_{\tau=0}
=
b^{-{1/\nu}} \partial_{\tau'} \ln M^{(2)}(t',\tau',L')|_{\tau'=0}
\label{es}
\end{equation}
with $t'=b^{z}t$ and $L'=b L$, the exponent $1/\nu$ can
independently be
calculated. The derivative is approximated by
taking the difference of the
values for $M^{(2)}$ at $J=J_c=0.4406$ and $J=0.4386$, divided
by its value at $J_c$.
It is clear and can also be seen in Fig.~\ref{ddd} that the
result for this related quantity fluctuates more than that
for $U(t,L)$, in particular for small $t$.
The results of this calculation have also been included
in the last column of Tab.~\ref{t1}.
It is interesting to point out, see Tab.~\ref{t1}, that
the results from the scaling fit of $L=16$ and $L=32$ are already
quite good.
This is probably due to the fact that the spatial
correlation length in the short-time regime of the dynamic
evolution is very small
and therefore from small lattices one can already obtain
reasonable results.
We also like to
mention that the procedure of comparing each run for the small
lattice with each run for the big lattice may
underestimate the errors. Therefore the data including the
errors given in
Tab.~\ref{t1} sometimes do not cover the exact values.
The real errors may be a factor two bigger.
In order to have more rigorous understanding of the dynamic
scaling,
we alternatively present a {\it local} approach for estimating
the critical exponents, in contrast to the {\it global}
scaling fitting procedure discussed above.
For example, comparing the functions
$U(t,L)$ and
$U(t',2L)$, for each time step $t$ we search for
$t'$ such that $U(t,L)=U(t',2L)$,
and from the ratio
$t'/t=2^z$ the value of $z$ is obtained according to
Eq.(\ref{e4}). For the same time step $t$
this particular value of $z$
can be used to estimate $2\beta/\nu$
from Eq.(\ref{e5}) and $1/\nu$
from Eq.(\ref{es}). Then we obtained all the exponents
as functions of the time $t$.
The result is shown in Fig.~\ref{dd1_stp} for a pair
of lattices with $L=32$ and $L=64$. In order to guide the eye,
three horizontal lines $z=2.14$, $2\beta/\nu=0.25$, and $1/\nu=1$
are included,
the latter $2$ values being the exact results for the 2-dimensional
Ising model.
The figure shows clearly that the fluctuations
for smaller time $t$ are big. But for $z$ and in particular
for $2\beta/\nu$ the curve tends very nicely to a
horizontal line for $t\gsim30$.
The situation is less satisfactory for the curve of
$1/\nu$ where it shows still some fluctuations up to
a fairly late time $t$. The reason
may be either less statistics or from the approximation of
the differentiation by a difference.
The exponents $z$, $2\beta/\nu$ and $1/\nu$ can be obtained by
averaging over the time direction.
To show the effect of the
fluctuations for smaller $t$, one takes the average starting
from different initial times.
The values of the exponents obtained
in this way have been given in Tab.~\ref{t2}.
The errors have again been estimated
by a comparison of each run for $L$ and with each
run for $L'=2L$.
In Fig.~\ref{dd1_stp}, when $t\lsim30$ the exponent $z$ is somewhat
small.
This might be due to the effect of $t_{mic}$.
In some cases, e.g. in the measurement of $\theta$
\cite{li94,sch95,oka95} and $\beta/\nu$ in the next section,
this effect hardly shows up. However, in some other cases
it can remain till $t\sim20-30$ \cite{sch95,oka95}.
This kind of effect probably comes from the fact that the initial
magnetization density is not uniform enough.
As we have already mentioned before, from Fig.~\ref{dd1_stp}
one can see explicitly even for $t\gsim30$ that
the exponent $z$ slightly rises as time
evolves. Even though the reason is not clear, but it is
interesting that the tendency of $z$ to rise
does not affect the measurement
of the static exponents, especially
$2\beta/\nu$. They are quite stable.
On the other hand, even for $z$ itself,
$1\%$ of the fluctuation in the time direction should
also be not too bad.
It may be due to the finite size effect or some technical reasons.
The curves from a scaling fit of the
lattices $8\leftrightarrow16$ and $16\leftrightarrow32$
which are not shown in the figures here
look qualitatively the same as those in Fig.~\ref{dd1_stp}, but
the fluctuations for $1/\nu$ are somewhat larger.
The curve
for $z$ from the smaller lattice sizes $8\leftrightarrow16$, however,
rises
continuously for $t\gsim20$, thus showing that such a lattice size
is too small for this kind of analysis.
The result of an average of these values are also included in
Tab.~\ref{t2}.
\vspace{0.3cm}
To measure $z$, $2\beta/\nu$ and $1/\nu$,
instead of $U$, $M^{(2)}$ and $\partial_\tau M^{(2)}$ one
can also use $\tilde U$, $|M|$ and $\partial_\tau |M|$ with
\begin{equation}
\tilde U(t,L) =1-\frac{ M^{(2)} }{ |M|^2 }.
\label{ezs}
\end{equation}
{}From Eq.(\ref{e2}) one easily deduces that the relation (\ref{e4})
holds
for $U$ as well as for $\tilde U$, and (\ref{es}) holds as well for
$\partial_\tau \ln|M|$.
Only Eq.(\ref{e5}) is slightly modified to
\begin{equation}
|M(t,L)| = b^{\beta/\nu} |M(b^{z}t,L')|; \qquad L'=bL,
\label{e5s}
\end{equation}
We do not plot the curves here,
since they look very much the same as Figs.\ref{ddU}-\ref{ddd} for
the
global and local analysis.
For simplification
we have compared only the two lattice sizes $L=32$ and $L=64$,
where we have used the full time scale up to $t'_{max}=900$ for
$L=64$.
The results for the global fit
have been included in the lower part of
Tab.~\ref{t1}.
All the values in the table are remarkably
consistent.
The same holds for
the results from the local approach. They are shown in Tab.~\ref{t2}.
\section{The Critical Relaxation \protect\newline with
Initial Magnetization One}
In the previous section we have investigated
the finite size scaling of the critical relaxation of the Ising model
up to even the macroscopic short-time scale, starting from
a random state with zero initial magnetization,
i.e., a completely disordered state.
{}From a measurement of
the time evolution of the observables $|M(t,L)|$, $M^{(2)}(t,L)$,
and $M^{(4)}(t,L)$ together with the scaling relation (\ref{e2})
the critical exponents $z$, $2\beta/\nu$
and $1/\nu$ were obtained. They are in good agreement with the known
results. This is a strong support for the scaling relations
derived by Janssen, Schaub and Schmittmann \cite{jan89}.
At this stage one may ask whether there is also
a scaling relation for the critical relaxation starting from
a completely ordered state, i.e. with initial magnetization
$m_0=1$. It has been known for some time that, before the
exponential decay of the magnetization starts,
there exists a time regime where the magnetization
behaves non-linearly and decays according to a power law.
The question is only {\it when} such a scaling behaviour starts.
Some effort has been made in this direction \cite{sta92,mun93}
with Monte Carlo simulation.
The authors have
simulated the critical relaxation with an extreme big lattice
but only up to a quite short evolution time and
have estimated the exponent
$\beta/{\nu z}$ from the power law decay of M(t). However, the result
has not been so clear and also other exponents as $z$ and $1/\nu$
have not been obtained.
\begin{table}[h]\centering
$$
\begin{array}{|c|c|c|c|l|l|l|}
\hline
\mbox{Input} & \mbox{Lattice} & t'_{min} & t'_{max} &\quad z\ &
\quad 2\beta/\nu\ &\quad
1/\nu\ \\
\hline
\hline
& & \ 50 & 300 & 2.121(4) & 0.2489(4) &
1.03(2) \\
\cline{5-7}
U & 32\leftrightarrow 64 & 200 & &2.122(5) & 0.2491(5) &
1.02(2) \\
\cline{3-7}
M^2 & & \ 50 & 900 & 2.129(5) & 0.2503(5) &
1.04(2) \\
\cline{5-7}
& & 200 & & 2.129(5) & 0.2505(6) &
1.04(2) \\\cline{3-6}
\hline
\hline
\tilde U & 32\leftrightarrow 64 & \ 50 & 900& 2.140(5) & 0.2514(6)&
1.07(2) \\
\cline{5-7}
|M| & & 200 & & 2.141(5) & 0.2515(7) &
1.07(2) \\
\hline
\end{array}
$$
\caption{
\footnotesize
Results for $z$, $2\beta/\nu$ and $1/\nu$, respectively,
from the 2-dimensional Ising model with initial magnetization
$m_0=1$.
The values are
obtained from a global scaling fit for two lattices.
}
\label{t3}
\end{table}
In this section we study systematically the
scaling behaviour of the critical relaxation from a completely
ordered initial state, but in {\it finite systems },
following a procedure parallel to that discussed in the previous
section.
The advantage is that a not too big finite system allows
for more statistics and longer evolution time even though the
power law decay of the magnetization will not be perfect.
{}From our results we confirm that the scaling appears in a quite
early stage of the relaxation as in the case of $m_0=0$.
Here we will only present data
from the scaling collapse for lattice sizes $L=32$ and $L=64$.
Also we confine ourselves to 4 runs with 50\,000 updations each,
instead of 8 runs in the previous section.
The curves for the second moment of
the magnetization start with the value one
for $t=0$ and
decrease for later time.
This is seen in Fig.~\ref{ddMm1}.
A similar decrease is found for the
curves for the cumulants $U(t,L)$ in Fig.~\ref{ddUm1},
while $\partial_\tau \ln M^{(2)}(t,L)$ in
Fig~\ref{dddm1} shows a rising behaviour. In all three
figures points mark the values
for $L=64$ rescaled in time by the best fit values
of $z$, $2\beta/\nu$, and
$1/\nu$. Surprisingly here we also observe very nice dynamic scaling.
Table~\ref{t3}
shows the results of the global fitting procedure
up to $t'_{max}=300$ or
$t'_{max}=900$, respectively.
Since the values of $U$, $M^{(2)}$ and $\partial_\tau M^{(2)}$
for smaller $t$ are large now compared with those in the case of
$m_0=0$ in the previous section, the results show
less fluctuations for smaller time $t$.
Actually one may also expect that due to the unique
initial configuration less statistics is needed
to obtain stable results.
This is also supported
by Fig.~\ref{dd_m1_stp} from the local approach.
It is very interesting that the exponent $2\beta/\nu$
shows almost {\it invisible } fluctuations
in the whole time regime even up to the very beginning,
although
$z$ has some similar unstable behaviour as that in the case of
$m_0=0$. Especially in the first $30$ time steps its value is also
a bit small. The fact that in the first steps of $t$
the exponent $z$ is quite near to $2.0$ might indicate that
at the very beginning of the time evolution the system
is ``classical''.
Similarly as Tab.~\ref{t2} of the previous section,
Tab.~\ref{t4} gives the
averages over the time direction, starting
at different initial values $t'_{min}$.
\begin{table}[h]\centering
$$
\begin{array}{|c|c|c|l|l|l|}
\hline
\mbox{Input} & \mbox{Lattice} & t'_{min} &\quad z\ & \quad
2\beta/\nu\ &\quad
1/\nu\ \\
\hline
\hline
U & 32\leftrightarrow 64 & \ 50 & 2.122(7) & 0.2508(3) &
1.04(2) \\
\cline{3-6}
M^2
& & 200 & 2.122(8) & 0.2510(4) &
1.04(2) \\
\hline
\hline
\tilde U & 32\leftrightarrow 64 & \ 50 & 2.133(6) & 0.2516(4) &
1.06(2) \\
\cline{3-6}
|M| & & 200 & 2.134(7) & 0.2520(5) &
1.06(3) \\
\hline
\end{array}
$$
\caption{
\footnotesize
Results for $z$, $2\beta/\nu$ and $1/\nu$, respectively,
from the $2$-dimensional Ising model with initial magnetization
$m_0=1$.
Values are obtained from the average in the time direction
from $t'_{min}$ up to $t'_{max}=900$
with a local scaling fit.
}
\label{t4}
\end{table}
As in the previous section, we carry out also the analysis
with $\tilde U$ defined in (\ref{ezs}),
$|M|$ and $\partial_\tau|M|$.
The results have been included in the lower parts
of Tab.~\ref{t3} from the global fit,
or in Tab.~\ref{t4} from the local approach, respectively.
They again have a similar quality as those
reported above. In comparison with the results in
\cite{sta92,mun93,dam93}, the value of $z$ we obtained
is definitely smaller and near to the one from
the $\epsilon$-expansion and other traditional
measurements \cite{wil85,tan87}.
In case of $m_0=0$, the results for $z$ measured
from $U$ and $\tilde U$ are almost the same.
However, in case of $m_0=1$, the value measured
from $U$ is somewhat smaller than that measured from $\tilde U$.
In the construction of $U$ we have not subtracted the
odd moments. This may have some effect on the measurement of $z$.
As compared to those in the previous section for
the case of $m_0=0$, the results here are somewhat more stable.
This may really indicate that it is also promising
to measure the critical exponents from the critical
relaxation process starting from
an ordered state
even though some more theoretical arguments like that by
Janssen, Schaub and Schmittmann \cite{jan89} are still needed.
It is clear that in case of a random initial state
with $m_0=0$ all the observables
discussed start their evolution from zero and therefore
the fluctuations at the beginning of the time evolution
are natural bigger.
Besides this, the effect of the non-uniformity of the
magnetization density
in the practically
generated initial configurations is not completely
negligible for not too big lattices.
Another way to measure the critical exponents
is to study the critical relaxation
starting from an initial state with small but non-zero
initial magnetization \cite {sch95,oka95}.
However, the situation is not so clear
since the new dynamic exponent $\theta$
enters the calculation. Further investigation is needed.
Finally we plot the time evolution of the magnetization
with both double-log scale and semi-log scale in order to see
whether it has entered the regime of linear decay or not.
In Fig.~\ref{t3a} the straight line shows definitely
that the magnetization is still in the regime of
non-linear decay.
{}From the slope of $M(t)$ in double-log scale one can obtain the
exponent
$\beta/(\nu z)$. for each time $t$ we have measured it by a least
square fit
within a time interval $[t,t+50]$. In Fig.~\ref{f10}
instead of $\beta/(\nu z)$ the exponent $z$ is plotted vs. time using
the exact value $\beta/\nu=1/8$. Note that the time scale in
Fig.~\ref{f10} is different from that in
Fig.~\ref{dd_m1_stp}.
Figure~\ref{f10} shows that in the regime $150 \gsim t \gsim
t_{mic}\approx30$
the values for $z$ are rather consistent with those obtained before,
especially those measured from $\tilde U$ in Tab.~\ref{t3}.
However, the lattice size $L=64$ seems not to be big enough to
present
a rigorous power law behaviour in the whole time region.
\section{Discussion}
We numerically simulate the critical relaxation process of the
two-dimensional Ising model with the initial state
both completely disordered or completely ordered.
Based on the finite size scaling
for the dynamics at the early time,
both the static and dynamic critical exponents
are measured. To determine $z$ independently,
a time-dependent Binder cumulant is constructed.
The value of $z$ measured from the critical relaxation
from a completely ordered state is slightly smaller
than that from a completely disordered state.
The reason is not yet clear.
Taking
the average of the four measurements of $z$ from the global scaling
fit
of $U$ and $\tilde U$ within the time interval $[50,900]$ of $t'$
for both relaxation processes, see Tab.~\ref{t1} and Tab.~\ref{t3},
we conclude
\vspace*{0.3cm}
\centerline{$z = 2.143(5)$}
\vspace{0.2cm}
\noindent This should be compared with the existing numerical
results $z=2.13(8)$
from \cite{wil85}, and $z=2.14(5)$ from \cite{tan87},
and also with $z=2.126$ obtained from an
$\epsilon$-expansion in \cite{bau81}, even though some bigger
values
are also reported recently \cite{sta92,mun93,dam93}.
It is remarkable that from the short-time dynamics
one can not only efficiently measure the dynamic
exponent $z$, but also the static exponents.
Especially the quality of the exponent
$2\beta/\nu$ is very good.
All these results provide strong confirmation for the scaling
relation at the early time of the critical relaxation process.
Compared with the traditional methods,
the advantage of our {\it dynamic } Monte Carlo algorithm is
that the measurement is carried out in the beginning of the time
evolution rather than in the equilibrium where
critical slowing down is more severe. Therefore our method is
efficient. Compared with the
non-local algorithms, our dynamic algorithm can study
the properties of the original local dynamics.
On the other hand, it has recently been suggested that
the critical exponents
can also be measured from the {\it power law} behaviour
of the observables including the auto-correlation
in the macroscopic short-time regime in a large enough
lattice \cite {sch95,oka95}. Compared with that approach,
the advantage of estimating the exponents
from the dynamic finite size scaling
as reported in this paper is that one needs not too big lattices.
However, the result has
to be obtained by comparing two
lattices and longer time of the evolution for the bigger lattice
should
be carried out.
It is somewhat surprising that for the critical relaxation
from the completely ordered initial state
there exist also universality and scaling in such
an early stage of the time evolution. Further investigation
especially on a more general critical relaxation process
from an ordered state with initial magnetization $m_0$
smaller but near to one can be interesting. One may expect that a
new
dynamic exponent should be introduced in order to complete
the scaling relation.
\vspace{0.5cm}
{\bf Acknowledgement:} One of the authors (Z.B.L.)
is grateful to the Alexander von
Humboldt-Stiftung for a fellowship.
The authors would like to thank
K. Untch for the help in maintaining our Workstations.
\vspace{0.4cm}
|
\section{Introduction}
Some attention has been recently paid to the nonrelativistic symmetry of
$2+1$--dimensional space--time. The relevant Galilei group differs
significantly from its fourdimensiomal counterpart which makes the study
of its mathematical structure quite interesting. Moreover, one can expect
that such an analysis will appear helpful in understanding the propeties
of nonrelativistic systems that are effectivly confined to two spatial
dimensions. In papers \cite{1}, \cite{2} Bose considers the problem of
finding all central extensions of $2+1$--dimensional Galilei group (and
its Lie algebra) and constructs the relevant umitary projective
representations. However, we feel that not all interesting points were
exhausted there. It is the aim of the present short note to add some
futher remarks concerning the central extensions of Galilei group /
algebra in three dimensions. In section II we prove (actualey, the proof is
almost trivial) two theorems indicating that certain families of such
extensions are isomorphic and indicate how they can be used to find the
relevant Casimir operators and to simplify slighty the representation
theory. In section III the explanation is offered for the existence of
additional nontrivial cocycle in three dimensions. Namely, it is shown
that occurence of this cocycle is related to the Thomas precession
phenomenon for threedimensional Lorentz group. Finally, the
telegraphically short section IV contains some bibligraphical notes. This is
because we think the proper credit should be given to authors who obtained
the results contained in Ref. \cite{1}.
\section{\sloppy On central extensions of 2+1-di\-men\-sional
Galilean algebra and
group}
Let $M$, $N_{i}$, $H$ and $P_{i}$ be rotation, boost, time-- and
space-- translation generators, respectively. Bose \cite{1} has proven that
the vector space of central extensions of Lie algebra of Galilei group is
threedimensional. In the notation adopted above the extended algebra reads
\newpage
\begin{equation}
\left.
\begin{array}{ccl}
\left[H,P_{i}\right]&=&0 \\
\left[N_{i},H\right]&=&i P_{i} \\
\left[P_{i},P_{j}\right]&=&0 \\
\left[N_{i},N_{j}\right]&=&i k \varepsilon_{ij}
{\bf 1} \\
\left[M,P_{i}\right]&=&i \varepsilon_{ij} P_{j} \\
\left[N_{i},P_{j}\right]&=&i m \delta_{ij} {\bf 1} \\
\left[M,N_{i}\right]&=&i \varepsilon_{ij} N_{j} \\
\left[M,H\right]&=&i l {\bf 1}
\end{array}
\right\} \label{1}
\end{equation}
the extension being parametrized by three real numbers $m$, $k$, $l$; the
central element has been denoted by {\bf 1}.
Let us denote the above algebra by $g_{kml}$. The following suprisingly
simple result holds:\\
{\bf Theorem I}:\\
{\em Let $m \neq 0$, $l$ -- arbitrary but fixed. Then $g_{kml}$ are
isomorphic, as
Lie algebras, for all k}.\\
{\bf Proof}.\\
Redefine the basis as follows: $X' = X$, $X \neq N_{i}$, $N_{i}' = N_{i} +
\frac{k}{2 m} \varepsilon_{ij} P_{j}$ $\Box$\\
Let us point out that such an isomorphism does not necessarily imply
physical equivalence (cf. Ref. \cite{3}).
As an application we list all Casimir operators for arbitrary
$k$, $m$, $l$. It reads\\
(i) $l = 0$, $m \neq 0$, $k$---arbitrary
\[
\begin{array}{ccl}
C_{1} & = & H - \frac{1}{2 m} \vec{P}^2,\\
C_{2} & = & M - \frac{1}{m} \vec{N} \times \vec{P} - \frac{k}{m} H
\end{array}
\]
(ii) $l$--arbitrary, $m = 0$, $k = 0$\\
\[
\begin{array}{ccl}
C_{1}' & = & \vec{P}^2, \\
C_{2}' & = & \vec{N} \times \vec{P}
\end{array}
\]
(iii) $l$--arbitrary, $m = 0$, $k \neq 0$\\
\[
\begin{array}{c}
C_{1}'' = \vec{P}^2
\end{array}
\]
(iv) $l \neq 0$, $m \neq 0$, $k$--arbitrary -- none.\\
We are not going to give here the detailed proof but rather content
ourselves with few remarks. The case (i) is a straightforward
consequence of Theorem I and the analogy with fourdimensional case;
(ii) and (iii) are easily verified and only (iv) calls for some comments.
Let us put $k = 0$ which, by Theorem I, does not restrict the generality.
Let $C$ be the central element of $g_{0ml}$. It can be written in ``normal``
order as
\begin{equation}
C = \sum_{(\lambda),(\mu), \nu, \rho} c_{(\lambda)(\mu)\nu\rho}
\prod_{i=1}^{2} N_{i}^{\lambda_{i}} \prod_{i=1}^{2} P_{i}^{\mu_{i}}
H^{\nu} M^{\rho}. \label{2}
\end{equation}
Let $\rho_{max}(c)$ be the maximal power of $M$ on the right hand side of
eq.(\ref{2}). Assume that $\rho_{max}(c) > 0$; then
\begin{equation}
C'= C + i (l \rho_{max}(c))^{-1} \cdot \left[C,C_{1}\right] \cdot M = C
\label{3}
\end{equation}
while $\rho_{max}(c') \leq \rho_{max}(c) - 1$; therfore $\rho_{max}(c) =
0$. Now apply the same reasoning with $M$ replaced by $H$ and $C_{1}$ replaced
by $C_{2}$ (with $k=0$) to conclude that $\nu_{max}(c)=0$. We continue this
argument by taking $N_{i}$ and $P_{i}$ instead of $C_{i}$ to show that
$\lambda_{i\, max}(c)=0$ and $\mu_{i\, max}(c)=0$.
Let us now consider the central extensions of Galilei group. The algebras
$g_{km0}$ can be integrated to yield the central extensions $G_{km}$ of
Galilei group $G$ \cite{1} \cite{3}. They can be described as follows. Let
$(\tau, \vec{u}, \vec{v}, R)$ be an element of Galilei group with $\tau$,
$\vec{u}$, $\vec{v}$, $R$ being time translation, space traslation, boost and
rotation, respectively. Then the multiplication rule for $G_{km}$ reads
\[
(\zeta, \tau, \vec{u}, \vec{v}, R) \ast (\zeta', \tau', \vec{u'},
\vec{v'},R') =
\]
\begin{equation}
= (\zeta \zeta'\omega, \tau + \tau', \overrightarrow{R u'} + \vec{v}
\cdot \tau' + \vec{u}, \overrightarrow{R v'} + \vec{v}, R R') \label{4}
\end{equation}
where $\zeta \in {\Bbb C},\, |\zeta| = 1$ and non trivial cocycle is given by
\begin{equation}
\omega=\exp\left(-i m (\frac{\vec{v}^{2}}{2} \tau' + \vec{v} \cdot
\overrightarrow{R u'}
) - \frac{i k}{2} (\vec{v} \times \overrightarrow{R v'})\right) \label{5}
\end{equation}
We adopted here the results of Ref\cite{3}; the corresponding cocycle
differs by coboundary from the one given in Ref\cite{1}.
Theorem I has the following counterpart on the group level.\\
{\bf Theorem II.}\\
{\em Let $m \neq 0$; then all groups $G_{km}$ are isomorphic.}\\
{\bf Proof.}\\
Make the following change of parameters:
\[
u_{i} \rightarrow u_{i} + \frac{k}{2 m} \varepsilon_{ij} v_{j},
\]
the remaining parameters being unaffected. $\Box$\\
Again this result appears to be quite useful. In Ref.\cite{2} the induced
representations of $G_{km}$ have been found following Mackey's method.
However, when attempting to apply this method in straightforward way one
is faced with the following apparent difficulty: there seems to be no
convenient semidirect product structure for $G_{km}$. This difficulty was
overcome in Ref.\cite{2} by considering the extensions of Galilei group $G$
with the help of two central charges and selecting the appropriate
representations. However, in view of our Theorem II it is unnecessary: we
can always assume $k=0$ or $m=0$ and in both cases the semidirect
structure is transparent.
For $l \neq 0$, $g_{kml}$ can be integrated to the central extension $
\tilde{G}_{kml}$ of the universal covering $\tilde{G}$ of Galilei goup.
The relevant group multiplication rule reads
\[
(\zeta, \tau, \vec{u}, \vec{v}, \theta) \ast (\zeta', \tau',\vec{u}',
\vec{v}', \theta') =
\]
\begin{equation}
= (\zeta \zeta' \tilde{\omega}, \tau+\tau',
\overrightarrow{R(\theta) u'}+ \vec{v} \cdot \tau'+ \vec{u},
\overrightarrow{R(\theta) v'}+\vec{v}, \theta+\theta'); \label{6}
\end{equation}
here $\theta \in {\Bbb R}$,
\[
R(\theta) = \left(
\begin{array}{cc}
\cos \theta & \sin \theta \\
- \sin \theta & \cos \theta
\end{array}
\right)
\]
and
\begin{equation}
\tilde{\omega} = exp ( i\, l\, \theta\, \tau' - i\, m\,
(\frac{\vec{v}^2}{2} \tau' + \vec{v} \cdot
\overrightarrow{R(\theta) u'}) - \frac{i k}{2} (\vec{v} \times
\overrightarrow{R(\theta) v'})) \label{7}
\end{equation}
Theorem II applies here as well. Therefore, it seems that only the case
$m=0$, $l \neq 0$, $k \neq 0$ has to be treated in the way indicated in
Ref.\cite{2}.
\section{The origin of cocycles}
We would like to understand the origin of nontrivial cocycles on Galilei
group $G$. It is the more interesting that the relativistic counterpart of $G$
-- the Poincare group $P$ -- does not admit nontrivial cocycles. On the
other hand, $G$ can be obtained from $P$ by contraction procedure. It is
therefore desirable to offer some interpretation for emergence of such
cocycles in nonrelativistic limit. The following general picture can be
given \cite{4}. Let $\omega(g,g')$ be any cocycle on $P$; write
\begin{equation}
\omega(g,g')=\exp i \xi(g,g') \label{8}
\end{equation}
Now, $\omega(g,g')$ is necessarily trivial, i.e. there exists a function
$\zeta$ on $P$ such that
\begin{equation}
\xi(g,g') = (\delta \zeta)(g,g') \equiv \zeta(g g') - \zeta(g) -
\zeta(g'). \label{9}
\end{equation}
The exponent $\xi(g,g')$ gives rise to a nontrivial cocycle in the
nonrelativistic limit $c\rightarrow \infty$ provided it survives the
contraction while $\zeta(g)$ does not (typically, it diverges as
$c\rightarrow \infty$). To make this pictures more concrete let us
describe in some detail the contraction procedure. First, we write the
element of Poincare group in matrix form
\begin{equation}
\{\Lambda, a\} \rightarrow \left[
\begin{array}{c|c}
\displaystyle \Lambda^{\mu}_{\nu} & \displaystyle a^{\mu}\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right] =
\left[
\begin{array}{c|c}
\displaystyle \delta^{\mu}_{\nu} & \displaystyle a^{\mu}\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right]
\left[
\begin{array}{c|c}
\displaystyle \Lambda^{\mu}_{\nu} & \displaystyle 0\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right]. \label{10}
\end{equation}
The Lorentz matrix $[\Lambda^{\mu}_{\nu}]$ is further decomposed into pure
boosts and rotations
\begin{equation}
\Lambda={\cal L}(\vec{v}) \cdot {\cal R} \label{11}
\end{equation}
where
\begin{equation}
{\cal L}(\vec{v}) =
\left[
\begin{array}{c|c}
\displaystyle \gamma & \displaystyle \frac{\gamma v_{k}}{c}\\[1ex]
\hline
\displaystyle \frac{\gamma v_{i}}{c} & \displaystyle \delta_{ik} + (\gamma
- 1) \frac{v_{i} v_{k}}{\vec{v}^2}
\end{array}
\right], \gamma \equiv \left(1-\frac{\vec{v}^2}{c^2}\right)^{-\frac{1}{2}}
\label{12a}
\end{equation}
\begin{equation}
{\cal R}=
\left[
\begin{array}{c|c}
\displaystyle 1 & \displaystyle 0\\[1ex]
\hline
\displaystyle 0 & \displaystyle R
\end{array}
\right], R R^{T}=R^{T} R=I. \label{12b}
\end{equation}
So, finally
\begin{equation}
\left[
\begin{array}{c|c}
\displaystyle \Lambda & \displaystyle a\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right]=
\left[
\begin{array}{c|c}
\displaystyle I & \displaystyle a\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right]
\left[
\begin{array}{c|c}
\displaystyle {\cal L}(\vec{v}) & \displaystyle 0\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right]
\left[
\begin{array}{c|c}
\displaystyle {\cal R} & \displaystyle 0\\[1ex]
\hline
\displaystyle 0 & \displaystyle 1
\end{array}
\right] \label{13}
\end{equation}
The contraction limit is now performed by multiplying eq.(\ref{13}) by $X$
from the right and $X^{-1}$ from the left, where
\begin{equation}
X=\left[
\begin{array}{c|c}
\displaystyle c & \displaystyle 0\\[1ex]
\hline
\displaystyle 0 & \displaystyle I
\end{array}
\right], \label{14}
\end{equation}
taking the limit $c \rightarrow \infty$ and identifying: $a^{0} \rightarrow
c \tau$, $\vec{a} \rightarrow \vec{u}$.
Now, one can easily explain the emergence of standard cocycle related to the
mass of particle. Take
\[\zeta(\{\Lambda,a\}) = c a^{0}\]
in eq.(\ref{9}). Due to the identification $a^{0} = c \tau$, $\zeta$
diverges as $c^2$ in the contraction limit. However,
\[
\zeta(\{\Lambda,a\}, \{\Lambda',a'\}) = c ( \Lambda^{0}_{\mu} {a'}^{\mu} +
a^{0} ) - c a^{0} - c {a'}^{0} =
c ( \Lambda^{0}_{\mu} - \delta^{0}_{\mu} ) {a'}^{\mu} =
\]
\begin{equation}
= c^2(\gamma-1)\gamma' + \gamma v_{i} R_{ik} u'_{k}
\stackrel{\scriptstyle c\rightarrow\infty}{\displaystyle \longrightarrow}
\frac{\vec{v}^2}{2} \cdot \tau' + \vec{v} \cdot
\overrightarrow{R u'}. \label{16}
\end{equation}
This explanation works both for three and four dimensions.\\
To account for the second cocycle (related to the parameter $k$) let us note
that, in the case of threedimensional space--time the rotation matrix
appearing in the decomposition (\ref{11})--(\ref{12b}) of the Lorentz
matrix is an element of $SO(2)$ and is therefore characterized by one angle
$\theta$. We put
\begin{equation}
\zeta(\{\Lambda,a\}) = c^2 \theta(\Lambda). \label{17}
\end{equation}
Actually, $\theta$ is multivalued no $P$ (while singlevalued on $\tilde{P}$)
but this plays no role in what follows. Now, from eq.(\ref{9}) we get
\begin{equation}
\xi(\Lambda,\Lambda') = c^2 ((\theta(\Lambda \cdot \Lambda') -
\theta(\Lambda) - \theta(\Lambda')). \label{18}
\end{equation}
But
\begin{equation}
\theta(\Lambda \cdot \Lambda') = \theta(\Lambda) + \theta(\Lambda') +
\delta\theta(\Lambda,\Lambda') \label{19}
\end{equation}
where $\delta\theta(\Lambda,\Lambda')$ is $0(1/c^2)$ and is related to the
so called Thomas precession \cite{5}; its existence reflects the propety that
the composition of pure boosts is no longer a pure boost.\\
It follows from eqs.(\ref{17})--(\ref{19}) that $\xi$ survives the
$c\rightarrow\infty$ limit while $\zeta$ does not. To calculate
$\delta\theta$ we write
\[
\Lambda \cdot \Lambda' = ({\cal L}(\vec{v}){\cal R})({\cal L}(\vec{v'})
{\cal R}') =
{\cal L}(\vec{v})({\cal R}{\cal L}(\vec{v'}){\cal R}^{-1})({\cal R}{\cal
R'}) =
\]
\begin{equation}
= ({\cal L}(\vec{v}){\cal L}(\overrightarrow{R v'}))({\cal R}{\cal R}').
\label{20}
\end{equation}
The standard calculations (using eqs.(\ref{11}), (\ref{12a}), (\ref{12b}))
give
\begin{equation}
{\cal L}(\vec{v}){\cal L}(\overrightarrow{R v'}) = {\cal L}
(\overrightarrow{v''}){\cal R}(\delta\theta) \label{21}
\end{equation}
where the value of $\vec{v''}$ is there irrelevant while, in the limit
$c\rightarrow\infty$
\begin{equation}
\delta\theta = \frac{\vec{v} \times
\overrightarrow{R v'}}{2 c^2}. \label{22}
\end{equation}
By comparying eqs.(\ref{18}), (\ref{19}) and (\ref{22}) we get
\[
\xi = \frac{\vec{v} \times \overrightarrow{R v'}}{2}
\]
which gives the cocycle found previously.
\section{Bibliographical remarks}
We would like to conclude with the following bibligraphical remarks. The
central extensions of Lie algebra of threedimensional Galilei group were
found many years ago by Levy--Leblond \cite{6}. The corresponding cocycles
on Galilei group have been constructed by Grigore \cite{7}. In the same
paper Grigore has found the unitary projective representations of
$2+1$--dimensional Galilei group using the Mackey theory and exploiting the
trick (used again in Ref.\cite{2}) consisting in extending of Galilei group
with the help of two (three in the case of universal covering) central
charges. Grigore gave also a detailed discussion of projective
representations of $2+1$--dimensional Poincare group \cite{8}.
\newpage
|
\section{Introduction}
Most oxide surfaces react readily with water and become partially
covered with molecular H$_2$O or hydroxyl groups on exposure to
air under ambient conditions. This adsorption of water has important
consequences for surface processes such as catalysis and gas sensing.
It is widely accepted that molecular water will generally bind to
most oxide surfaces through the attraction of its electric dipole
to the ionic charges. However, the mechanisms by which dissociative
adsorption occurs are still very poorly understood. It has sometimes
been suggested that surface defects such as vacancies or steps are
essential to water dissociation, and there is no doubt that there are
some oxide surfaces for which this is true. For example, recent
calculations using both semi-empirical~\cite{gon95} and
first-principles~\cite{lan94,sca94}
techniques have shown that water dissociation is not energetically
favorable on the perfect MgO~(100) surface, but that it is
favorable at steps and corners. However, it is unlikely that
defects are essential for dissociative adsorption in general,
since oxide surfaces vary enormously
in their geometry and electronic structure.
The aim of this paper is to present first principles calculations
on the energetics of H$_2$O adsorption on the SnO$_2$ and
TiO$_2$~(100) surfaces. Both materials have the rutile crystal structure,
and the (110) surface is the most stable when the materials are
stoichiometric. The stoichiometric (110) surface is far from flat,
so we are dealing with a situation which is geometrically
very different from the MgO~(100) surface. Although they have the
same geometry, SnO$_2$ and TiO$_2$ are electronically different,
since Ti is a transition element and Sn is a group~IV element.
Our hope is therefore that by comparing with previous work on the
interaction of H$_2$O with MgO~(100) we can learn about the
influence both of geometry and of electronic structure.
Our choice of SnO$_2$ and TiO$_2$ is also motivated by their
technological importance. SnO$_2$ is widely used in gas-sensing
devices, and adsorbed water is known to have an important
effect on the adsorption of other molecules~\cite{tan89}. The interaction of
TiO$_2$ with water is important in photocatalytic applications~\cite{fuj72}.
A considerable experimental effort has gone into the study of H$_2$O
adsorption on TiO$_2$~(110), though a clear picture has yet to emerge.
Hydroxyl groups on the surface have been detected at 300~K in
ultraviolet photoelectron spectroscopy (UPS)
experiments by Henrich {\em et al.}~\cite{hen77}, and their
presence was later confirmed in synchrotron radiation studies
by Madey's group~\cite{kur89,pan92}. The amount of dissociated water
was less than a monolayer, and was weakly dependent on the oxygen
vacancy concentration before adsorption. The minor influence of
point defects and steps on the reactivity with water has also been
reported by Muryn {\em et al.}~\cite{mur91}. Experiments using
a combination of thermal desorption spectroscopy (TDS), x-ray
photoelectron spectroscopy (XPS) and work-function measurements
have recently been reported~\cite{hug94}. A three-peak desorption
spectrum was found,
with features at 170 and $\sim$275~K and a high-temperature tail
up to $\sim$375~K. The two low-temperature features appear to be
due to molecular water and the high-temperature tail to the reaction
of surface OH groups to form H$_2$O.
A widely accepted model is that of Kurtz {\em et al.}~\cite{kur89}, who
proposed
that adsorption on nearly perfect TiO$_2$~(110) occurs by molecular adsorption
of H$_2$O at 5-fold coordinated cation sites, followed by dissociation
to give OH$^-$ attached by its oxygen end to the cation, and H$^+$ bonded
to lattice oxygen to form a second type of hydroxyl group.
The presence of
at least two types of adsorbed OH$^-$ has been indicated by infrared
spectroscopy~\cite{mun71,jon71}.
The experimental situation for SnO$_2$ is rather similar, although
the experimental evidence is less extensive. Recent
experiments by Gercher and Cox~\cite{ger95} using a combination of
TDS and UPS show the presence of dissociatively adsorbed water
on SnO$_2$~(110). On the perfect stoichiometric
surface, three distinct TDS peaks are seen at $T$ = 200,
300 and 435~K. The 200~K peak is attributed entirely and the 300~K peak
mainly to molecular H$_2$O. The feature at 435~K is assigned to
disproportionation of surface OH$^-$ to form H$_2$O. As in the
case of TiO$_2$, surface defects do not appear to play a major role.
Few theoretical investigations of H$_2$O adsorption on rutile
surfaces have been reported. Jaycock and Waldsax~\cite{jay74} performed
calculations for the TiO$_2$ (110) and (100) surfaces using
an empirical interaction model, which predicted adsorption energies much
greater
than the experimental values. The dissociative adsorption of single molecules
on clusters embedded in a point-charge lattice has been studied using
the $X \alpha$ approach for TiO$_2$~(110) by Tsukada
{\em et al.}~\cite{tsu83}. The dissociative adsorption of single water
molecules
on TiO$_2$ as well as the case of full hydroxylation
have been studied using a semi-empirical technique by
Goniakowski {\em et al.}~\cite{gon93} and Goniakowski and Noguera~\cite{gon95}.
Both molecular and dissociative adsorption of water on TiO$_2$ have been
treated using a periodic Hartree-Fock approach by Fahmi and
Minot~\cite{fah94}, who find a strong preference of water to dissociate.
The above studies considered only very symmetrical
configurations of adsorbed species, and all found that both
molecular and dissociative adsorption are energetically favorable.
The only attempt to study relaxation of surface species to unsymmetrical
configurations has been the very recent semi-empirical study by
Bredow and Jug on the adsorption of H$_2$O on rutile and anatase TiO$_2$
surfaces~\cite{bre95}. To our knowledge no theoretical attempts to model
water adsorption on SnO$_2$ surfaces have been reported.
The present work is based on density functional theory (DFT) in the
pseudopotential approximation. This approach has been widely
used for oxides, including
MgO~\cite{vit92}, Li$_2$O~\cite{Ian92}, Al$_2$O$_3$~\cite{Ian93},
TiO$_2$~\cite{ram94a,ram94b}
and SnO$_2$~\cite{IanYY}, and is known to give accurate results
for the energetics of perfect crystals, lattice defects,
surfaces and molecular adsorption. We have recently reported a
detailed study of the stoichiometric and reduced SnO$_2$~(110)
surface using this approach~\cite{IanXX}.
However, since the energetics of dissociation is not generally
described very accurately within the standard local density
approximation (LDA), we include gradient corrections in the present work,
using the Becke-Perdew scheme~\cite{bec88,per86}.
We present here results
on the energetics and relaxed geometry of both molecularly
and dissociatively adsorbed water on SnO$_2$ and TiO$_2$~(110),
and we study both symmetrical and unsymmetrical modes of adsorption.
In addition, we report total and local electronic densities of
states, through which we attempt to make contact with recent
spectroscopic measurements.
\section{Techniques}
\subsection{General background}
The general principles of DFT and the pseudopotential method have
been described in the literature~\cite{hoh64,koh65,jon89,gil91,pay92}.
Within the pseudopotential approximation only
valence electrons are represented explicitly in the calculations, the
valence-core interaction being represented by non--local pseudopotentials
which are generated by first principles calculations on isolated atoms.
The calculations are performed using periodically repeating geometry, the
occupied orbitals being expanded in
a plane wave basis. This expansion includes all plane waves whose kinetic
energy $E_k = \hbar^2k^2/2m$ ($k$ the wavevector, $m$ the electron mass)
is less than the cut-off energy $E_{\rm cut}$, chosen so as to ensure
convergence with respect to the basis set.
In the present work the self-consistent ground state of the system was
determined using a band-by-band conjugate gradient technique to minimize
the total energy of the system with respect to the plane-wave coefficients.
Equilibrium positions of ions were determined by a steepest descent
method. The calculations were performed using the CETEP code~\cite{cla92}
(the parallel version of the serial code CASTEP~\cite{pay92}), running on
the 64-node Intel iPCS/860 machine at Daresbury Laboratory.
\subsection{Generalized Gradient Corrections}
Standard DFT calculations make use of the LDA, which treats the
electron density as locally uniform. Although this is highly
successful for many purposes, it does not give accurate results
for energy differences involving changes of bonding, which are of
interest here. In the last few years, methods have been developed
for improving the LDA through `generalized gradient' approximations
(GGA). It has been shown that these corrections lead to improvement
in the calculation of total energies of atoms and
molecules~\cite{per86,lan83,lan85,pw86,kut88,bos90,mly91},
cohesive energies~\cite{gar92,kon90,ort92}, and the energetics of
molecular adsorption and dissociation
on metal surfaces~\cite{whi94a,hu94,phi94,gun94}.
We have recently reported a comparison
of the influence of two popular GGA schemes -- those due to
Perdew and Wang~\cite{per86,pw86} and Becke and Perdew~\cite{bec88,per86}) --
on the bulk and surface properties of TiO$_2$ and SnO$_2$~\cite{unp}. We found
that the Becke-Perdew method gives better equilibrium lattice
parameters, while giving essentially the same surface properties as
Perdew-Wang, and we employ Becke-Perdew in the present work. We use
the technique of White and Bird~\cite{whi94} for incorporating
GGA within the DFT-pseudopotential calculation. In some places,
we shall also report LDA results for comparison; these were obtained
using the Ceperley-Alder (CA) exchange-correlation function~\cite{cep80}.
\subsection{Generation of pseudopotentials}
First-principles, norm-conserving pseudopotentials in Kleinman-Bylander
representation~\cite{kle82} were generated using the
optimization scheme of Lin {\em et al.}~\cite{lin93} in order to
reduce the required value of the plane-wave cut-off $E_{\rm cut}$.
The pseudopotentials used in GGA calculations were constructed in a
consistent way by including GGA in the generation scheme.
The Sn
pseudopotential was generated using the $5s^25p^2$ configuration for
$s$- and $p$-wave components, and the $5s^15p^{0.5}5d^{0.5}$
configuration for the $d$-wave. The core radii were equal to 2.1, 2.1 and
2.5~a.u. for the $s$, $p$ and $d$ components respectively.
The Ti pseudopotential was generated using
the $4s^{1.85}3d^2$ configuration for the $s$ and $d$ waves
and the $4s^14p^{0.5}3d^{0.5}$ configuration for the
$p$ wave, with core radii of 2.2, 1.5 and 2.4~a.u. for $s$, $p$ and $d$ waves
respectively.
The oxygen pseudopotential used in our LDA calculations
was generated
using the $2s^22p^4$ configuration for the $s$ and $p$ waves and the
$2s^22p^{2.5}3d^{0.5}$ configuration for the $d$ wave,
with a single core radius of 1.65~a.u. For the
gradient-corrected oxygen pseudopotential, we have used the single
configuration $2s^22p^{3.5}3d^{0.45}$ and the same core radius.
The use of a core radius of 1.65~a.u. means that there is an appreciable
overlap of the oxygen and metal core spheres in the SnO$_2$ and TiO$_2$
crystals, and in principle this could cause inaccuracies. However,
direct comparisons of the present results with our earlier work
on SnO$_2$~\cite{IanXX}, which employed an oxygen pseudopotential with the
smaller core radius of 1.25~a.u., show that any errors due to core overlap are
very small.
The calculations have been done using a plane wave cut-off of 600~eV for
SnO$_2$ and 1000~eV for TiO$_2$. Our tests show that with these cut-offs
the energy per unit cell is converged to within 0.2~eV.
\subsection{Densities of States}
Electronic densities of states (DOS) associated with the ground state were
calculated using the tetrahedron method~\cite{jep71,leh72}, with $k$-point
sampling corresponding to 750 tetrahedra in the whole
Brillouin zone. In addition, local densities of states (LDOS)
were calculated by taking contributions
only from chosen regions of real space -- in practice
we have used spheres (radius 1.5~\AA) centered on chosen
surface oxygen atoms. For presentation purposes,
we have broadened the calculated DOS and LDOS by Gaussians of width 0.5~eV.
In the ground state calculations the Brillouin zone sampling is performed
using the lowest order Monkhorst-Pack set of $k$ points~\cite{mon76},
as in our previous work on SnO$_2$~\cite{IanYY,IanXX}.
\section{Tests on SnO$_2$, TiO$_2$ and H$_2$O}
\subsection{Perfect crystals and (110) surfaces}
A more detailed discussion of the influence of gradient
corrections on the calculated parameters of the SnO$_2$
and TiO$_2$ perfect crystals and (110) surfaces is given in a
separate paper~\cite{unp}. Here we present the system used in the
calculations and recall the principal results obtained
using the Becke-Perdew GGA scheme.
\begin{figure}
\caption{ The unit cell of the rutile crystal structure. Cations are
represented by small black circles and oxygen atoms by large white circles.}
\end{figure}
The six-atom rutile unit cell of SnO$_2$ and TiO$_2$ is shown in Fig.~1.
The equilibrium structure has been determined by relaxation with
respect to the lattice parameters $a$ and $c$ and the
internal parameter $u$. The equilibrium values of these
parameters calculated using the Becke-Perdew form of GGA are given in Table~1.
\begin{table}
\caption{Comparison of theoretical and experimental \protect \cite{wyckoff}
values of lattice parameters $a$ and $c$ and the internal coordinate $u$ of
SnO$_2$ and TiO$_2$. The theoretical values are calculated using
the Becke-Perdew form of GGA.}
\begin{tabular}{lcccc}
\hline
\hline
&\multicolumn{2}{c}{SnO$_2$}&\multicolumn{2}{c}{TiO$_2$} \\
& calc. & expt. & calc. & expt. \\ \hline
$a$ (\AA)& 4.809 & 4.737 & 4.747 & 4.594 \\
$c$ (\AA)& 3.159 & 3.186 & 3.039 & 2.958 \\
$c/a$ (\AA)& 0.657 & 0.673 & 0.640 & 0.644 \\
$u$ (\AA)& 0.307 & 0.307 & 0.305 & 0.305 \\
\hline
\hline
\end{tabular}
\end{table}
The agreement of $a$ and $c$ with experiment is very
satisfactory for SnO$_2$ and acceptable for TiO$_2$;
the values of $u$ are excellent in all cases.
Our calculations on the stoichiometric (110) surface of
the materials have been done with the usual repeating slab geometry.
The rutile structure can be regarded as consisting of (110) planes of atoms
containing both metal (M) and oxygen (O) atoms, separated by planes
containing oxygen alone, so that the sequence of planes is
O - M$_2$O$_2$ - O - O - M$_2$O$_2$ - O etc. The entire crystal can
then be built up of the symmetrical 3-plane O - M$_2$O$_2$ - O units.
The slabs we use contain three of these units, and our repeating cell
contains 18 atoms (6 M and 12 O). The perfect (110) surface
consists of rows of bridging oxygens lying above a metal-oxygen
layer (see Fig.~2).
\begin{figure}
\caption{ Atomic structure of the clean (110) surface. 6- and 5-fold
coordinated cations are represented by small black circles and small black
circles with white centers respectively. In plane oxygen atoms
are represented by large white circles and bridging oxygens
by large gray circles.}
\end{figure}
The vacuum separating the slabs has been taken wide enough to
ensure that interactions between neighboring slabs are small
even when adsorbed H$_2$O and OH$^-$ are present.
The width we use corresponds to two O - M$_2$O$_2$ - O units, and
is such that planes of bridging oxygens on
the surfaces facing each other across the vacuum are
separated by about 6.8~\AA. This vacuum
width is somewhat greater then was used in our earlier
calculations on SnO$_2$~(110)~\cite{IanXX}.
(It will be convenient in the following
to specify slab thickness and vacuum width in terms of the equivalent
width of O - M$_2$O$_2$ - O units.)
The surface structure has been determined by relaxing the
entire system to equilibrium. As in our previous work on SnO$_2$~(110),
and the work of
Ramamoorthy {\em et al.} on TiO$_2$~(110)~\cite{ram94a}, we find
displacements of the surface atoms of order 0.1~\AA, with
5-fold and 6-fold coordinated metal atoms moving respectively into and
out of the surface, in-plane oxygens moving out and bridging
oxygens moving very little. The modifications of the bond lengths
between the surface atoms with respect to the perfect crystal for
Becke-Perdew GGA calculations are given in Table~2.
\begin{table}
\caption{Calculated bond length modifications on SnO$_2$~(110)
and TiO$_2$~(110) with respect to the bulk values for Becke-Perdew
GGA exchange-correlation. For atom indexing see Fig.~2}
\begin{tabular}{lcc}
\hline
\hline
&SnO$_2$&TiO$_2$ \\ \hline
O$_{\rm I}$ -- M$_{\rm I}$ & -3.8\% & -5.5\% \\
O$_{\rm IV}$ -- M$_{\rm II}$ & -4.2\% & -5.6\% \\
O$_{\rm II}$ -- M$_{\rm II}$ & -1.2\% & -1.2\% \\
O$_{\rm II}$ -- M$_{\rm I}$ & 2.9\% & 2.8\% \\
O$_{\rm III}$ -- M$_{\rm I}$ & 4.8\% & 4.5\% \\
\hline
\hline
\end{tabular}
\end{table}
We find that the relaxed surface energy of SnO$_2$~(110) is
1.16~Jm$^{-2}$ and 0.84~Jm$^{-2}$ for TiO$_2$~(110). These energies are,
as already disscused in Ref.~\cite{unp}, substantially lower than the
corresponding LDA results.
The electronic DOS and the LDOS on bridging oxygen calculated for
the valence band of the slab systems using the Becke-Perdew
form of GGA are shown in Fig.~3.
\begin{figure}
\caption{Calculated valence band DOS for
clean (110) surfaces of TiO$_2$ and SnO$_2$. The dashed line
represents the LDOS on the bridging oxygen site O$_I$. For the sake
of presentation the local contribution has been scaled by a factor of 5.}
\end{figure}
The main features of the calculated surface DOS of SnO$_2$
have already been discussed in our previous paper~\cite{IanXX}. We note
particularly the splitting off of a narrow band of states at
the top of the O(2s) band and the appearance of a
sharp peak at the top of the O(2p) valence band. Both these
features are associated with bridging oxygens. The features are
also present for TiO$_2$, although the splitting is less marked.
\subsection{H$_2$O and OH$^-$ molecules}
We have calculated the equilibrium geometry and electronic structure
of the H$_2$O and OH$^-$ molecules. Since our techniques require us
to use periodic boundary conditions, the calculations are actually
performed on periodic arrays of molecules, with the repeating cell
chosen to be a cube of side $R$. For H$_2$O, we find that the choice
$R$ = 7~\AA\ is large enough to render the interactions between periodic
images negligible. Matters are not so simple for the OH$^-$ molecule,
since it carries a net charge. To make mathematical sense of calculations
in which the repeating cell is charged, it is essential to introduce
a uniform compensating background, whose charge density is chosen so that
the net charge in the unit cell vanishes. This technique is well
established, and has often been discussed in the literature. However,
as discussed by Leslie and Gillan~\cite{les85},
the total energy still converges very
slowly with increasing cell size, and the leading term in its deviation
from the asymptotic value is $- \alpha / R$, where $\alpha$ is the
Madelung constant for the appropriate periodic array of point charges.
This leading
correction can therefore be subtracted exactly, and we are left with
a total energy which converges reasonably quickly to the value for the
molecule in free space. In our calculations on OH$^-$, we find that
the corrected total energy is converged to
better than 0.1~eV for $R$ = 13~\AA,
and the equilibrium bond length is converged well before this. Comparison
of the calculated and experimental structural parameters of
H$_2$O and OH$^-$ (Table~3) confirms that the molecules are described
accurately by the present methods.
\begin{table}
\caption{Comparison of theoretical and experimental values of bond
lengths d$_{O-H}$ and bond angle
$\protect \angle $H-O-H of H$_2$O and OH$^-$.
The theoretical values are calculated using the Ceperley-Alder
form of LDA~(CA), and the Becke-Perdew form of GGA~(BP).
Experimental values are taken from
Refs. \protect \cite{eisenberg,herzberg}}
\begin{tabular}{llcccccc}
\hline
\hline
&& \multicolumn{2}{c}{$d_{O-H}$ (\AA)}&\multicolumn{2}{c}{$\angle H-O-H$} \\
\hline
H$_2$O & CA & 0.977 & (2.1\%) & 104.8 & ( 0.3\%) \\
& BP & 0.975 & (1.9\%) & 103.6 & (-0.9\%) \\
& expt.& 0.957 & & 104.5 & \\ \hline
OH$^-$ & CA & 0.980 & (1.0\%) & & \\
& BP & 0.979 & (0.9\%) & & \\
&expt.& 0.970 & & & \\
\hline
\hline
\end{tabular}
\end{table}
Since we are concerned with dissociation of H$_2$O in this paper, we have
tested the influence of the GGA on its dissociation energy. The reaction
of interest is H$_2$O $\rightarrow$ OH$^-$ + H$^+$. To compare
our results with experimental data, we start from the experimental
dissociation energy for the reaction H$_2$O $\rightarrow$ OH + H, which
is 5.11~eV. We add to this the ionization energy of H (13.60~eV) and
we subtract the electron affinity of OH (1.83~eV), to obtain 16.88~eV.
Finally, addition of the zero-point vibrational energy
of H$_2$O (0.57~eV) and subtraction
of the corresponding quantity for OH (0.23~eV) gives us the
value 17.22~eV which can be compared with the calculations. The
calculated values with LDA and GGA are 16.6 and 16.7~eV, so that
even with gradient corrections there is a residual error of $\sim$~0.5~eV.
In order to compare with the electronic DOS reported later,
it is useful to note that the single-particle energies of the occupied
molecular orbitals (MO) of H$_2$O can be related to experimental measurements.
As usual, one must be cautious about comparing Kohn-Sham single particle
energies with spectroscopic energies, but it is known empirically that for
occupied states the comparison is usually justified. We therefore compare
the differences of our calculated MO energies with the corresponding
differences of measured ionization energies.
In the usual notation, the MO states of H$_2$O are
$2 a_1$, $1 b_2$, $3 a_1$ and $1 b_1$.
Our calculated separations of $2a_1 - 1b_2$,
$1b_2 - 3a_1$ and $3a_1 - 1b_1$ (experimental values from ref.~\cite{bal78}
in parentheses)
are 11.96~eV (13.6~eV), 3.64~eV (3.8~eV) and 2.12~eV (2.0~eV) respectively.
Our calculated
separations of $2\sigma - 3\sigma$ and $3\sigma - 1\pi$ levels in OH$^-$
are respectively 12.2~eV and 3.5~eV.
\section{Surface adsorption of H$_2$O}
We turn now to our calculations on dissociative and molecular
adsorption of H$_2$O on SnO$_2$ and TiO$_2$ (110). The calculations
are done using the same repeating-slab geometry used for the
bare surface, and we need to pay attention to the effects of slab
thickness and vacuum width. Another important technical
question is whether it is better to perform the calculations
with the particles adsorbed only on one surface of each slab
or on both surfaces. We refer to these as the one-sided and two-sided
geometries. There is a strong argument for working with
the symmetrical two-sided geometry in which the same particles are adsorbed on
opposite surfaces, because any possible dipole moment of the
repeated cell is then eliminated, and convergence of the adsorption
energy with increasing slab thickness is likely to be improved.
We have made tests which confirm that this is the case, and all our
calculations have therefore been made using the two-sided
geometry.
Our tests on the effects of slab thickness and vacuum
width were performed on the SnO$_2$ system in which
H$^+$ is adsorbed on top of every bridging oxygen and OH$^-$ is
adsorbed at every 5-fold Sn site (see Fig.~4a).
\begin{figure}
\caption{ Atomic structure of fully hydroxylated (110) surface of rutile
for a) SD, b) UD and c) SM adsorption geometries.
Atom symbols as in Fig.~2, with adsorbed oxygen atoms
represented by large white circles and hydrogens by small white circles.}
\end{figure}
In these tests, every
atom in the system is relaxed to its equilibrium position. Using
the two-sided slab geometry mentioned above, we find that increase
of the vacuum width from two to three O - Sn$_2$O$_2$ - O units changes the
adsorption energy by only 0.01~eV per water molecule, and increase of
the slab thickness from three to four units gives a change of 0.03~eV
per molecule. If the one-sided slab geometry is used, the changes
are nearly ten times as great. The results to be presented
have all been obtained with a slab thickness of three units and
a vacuum width of two units. We stress that in all the calculations
that follow the entire system is fully relaxed to equilibrium.
As noted in the Introduction, we cannot be sure in advance how H$_2$O
will prefer to adsorb. To study dissociative adsorption, we begin with
the symmetrical full-coverage case mentioned above (see Fig.~4a).
We then study symmetry-lowering distortions from this configuration.
A similar strategy is followed for the case of molecular adsorption.
The effect of going to lower coverage is than briefly examined for the
dissociative case.
\subsection{Symmetrical dissociative adsorption (SD)}
The dissociative adsorption energy is obtained from the fully relaxed total
energy of the slab system in which H$^+$ and OH$^-$ are adsorbed at both
surfaces. It is calculated in the natural way by subtracting
this energy per unit cell from the corresponding energy for the
bare slab plus the energy of a pair of H$_2$O molecules; the result is,
of course, divided by two, since two H$_2$O molecules (one on each
surface) are adsorbed per unit cell. As usual,
a positive adsorption energy means that the total energy decreases
when the molecule is adsorbed. In the symmetrical dissociative (SD)
case, our LDA calculations for SnO$_2$ and TiO$_2$ (110) yield
adsorption energies of 1.19 and 0.91~eV per H$_2$O respectively.
Inclusion of gradient corrections lowers the adsorption energies
considerably to 0.48 and 0.45~eV per H$_2$O. It is clear from this
that gradient corrections have an extremely important effect on the
calculated adsorption energies. A similar effect of lowering of the adsorption
energy by inclusion of GGA has already been reported for adsorption on metal
surfaces~\cite{hu94,phi94}, leading to a substantial
improvement with respect to experimental results.
We find that the dissociative adsorption of water causes the equilibrium
surface
structure to change significantly. The relaxations found for the clean surface
are greatly reduced and in some cases reversed. The adsorption causes
5- and 6-fold cations to
move respectively out of and into the surface, the in-plane oxygens
to move outward, and the bridging oxygens to move slightly outward, relative to
the relaxed clean surface.
Compared with free hydroxyl molecules, surface O--H bonds are
considerably shorter, the shortening being especially pronounced for
the OH group attached to the 5-fold surface cation. Similarly,
the bond between the OH group and the surface cation is
much shorter than the bulk O--cation bond, resembling rather the
inter-atomic distance in the corresponding diatomic molecule.
Details of inter-atomic bond length changes for
both materials (with respect their perfect crystal values), with and without
gradient corrections, are presented in Table~4.
The changes
of O--H bond lengths are given with respect to the free OH$^-$ molecule.
It is clear from the table that, in contrast to the adsorption energies,
the relaxed structure of the hydroxylated surface is little
affected by gradient corrections.
\begin{table}
\caption{Calculated bond length modifications on hydroxylated SnO$_2$~(110)
and TiO$_2$~(110) with respect to the bulk values for LDA (CA), and GGA (BP)
forms of exchange-correlation. The modifications of O--H bond lengths are
given with respect to the free OH$^-$ molecule. For atom indexing see Fig.~4a.}
\begin{tabular}{lcccc}
\hline
\hline
&\multicolumn{2}{c}{SnO$_2$}&\multicolumn{2}{c}{TiO$_2$} \\
& CA & BP & CA & BP \\ \hline
O$_{\rm I}$ -- M$_{\rm I}$ & 1.2\% & 1.5\% & 2.5\% & 2.7\% \\
O$_{\rm IV}$ -- M$_{\rm II}$ & 2.2\% & 3.4\% & 3.8\% & 4.3\% \\
O$_{\rm II}$ -- M$_{\rm II}$ & 1.1\% & 1.1\% & 1.4\% & 0.9\% \\
O$_{\rm II}$ -- M$_{\rm I}$ & -1.5\% & -1.4\% & -1.7\% & -1.0\% \\
O$_{\rm III}$ -- M$_{\rm I}$ & -1.5\% & -1.9\% & -2.7\% & -3.1\% \\
&&&&\\
O$_{\rm I}$ -- H$_{\rm I}$ & -0.8\% & -0.8\% & -1.6\% & -1.5\% \\
O$_{\rm A}$ -- H$_{\rm II}$ & -2.2\% & -2.2\% & -1.7\% & -2.0\% \\
O$_{\rm A}$ -- M$_{\rm II}$ & -5.5\% & -6.2\% & -6.9\% & -7.0\% \\
\hline
\hline
\end{tabular}
\end{table}
Figure~5a shows both total and local DOS
calculated for the hydroxylated (110) surface of SnO$_2$ and TiO$_2$.
Two oxygen sites were studied using the LDOS: the bridging
oxygen atom O$_{\rm I}$ and the oxygen of the adsorbed OH$^-$ group O$_{\rm
A}$.
Compared with clean surfaces, two main modifications are apparent:
\begin{figure}
\caption{Calculated valence band densities of states for
hydroxylated (110) surfaces of TiO$_2$ and SnO$_2$ in a) SD, b) UD
c) SM adsorption geometries. The dashed line
represents the LDOS on the bridging oxygen site O$_{\rm I}$, and the dotted
line
the LDOS on the adsorbed oxygen site O$_{\rm A}$. For the sake
of presentation the local contributions have been scaled by a factor of 5.}
\end{figure}
\begin{itemize}
\item the contribution due to the bridging oxygen atom, situated at
the top of the valence band for clean surfaces (and giving an
extra feature above the O(2s) band)
has been pushed towards lower energies, and hybridizes
more strongly with the bulk O(2p) band.
On the other hand, formation of the O--H$^+$ bond gives rise to sharp
bonding 3$\sigma$ states
below the O(2p) band and 2$\sigma$ states below the O(2s) band.
\item the adsorbed hydroxyl group gives a contribution to
the valence O(2p) band due to its occupied 1$\pi$ state. It
lies above the surface VBM in the case of SnO$_2$~(110) and
within the surface VB for TiO$_2$~(110). On the other hand,
bonding 3$\sigma$ states can be seen as a narrow peak below
the O(2p) band on TiO$_2$~(110), whereas on SnO$_2$~(110) they
lie within the O(2p) band.
For both materials, the 2$\sigma$ peak lies within O(2s) band.
Separations of $2\sigma - 3\sigma$ and $3\sigma - 1\pi$ are respectively
10.3~eV and 6.2~eV for SnO$_2$ and 11.4~eV and 5.3~eV for TiO$_2$.
\end{itemize}
It is clear from the results that, especially for the adsorption energy,
gradient corrections make
a substantial difference. All the remaining calculations are performed
with the Becke-Perdew GGA scheme only.
\subsection{Unsymmetrical dissociative adsorption (UD)}
The SD geometry described in the previous section is
not the most stable one, and a small displacement of the adsorbed atoms
from their fully symmetric positions makes the system relax to a
configuration of lower energy. We find that this unsymmetrical
dissociative (UD) configuration gives
an adsorption energy of 1.39~eV per H$_2$O for
SnO$_2$~(110) and 1.08~eV per H$_2$O for TiO$_2$~(110),
so that the breaking of symmetry yields a stabilization of
well over 0.5~eV.
The nature of the new relaxed configuration is shown in Fig.~4b.
The bridging (O$_{\rm I}$) and adsorbed (O$_{\rm A}$) oxygens
approach each other, and the proton (H$_{\rm I}$) attached to bridging oxygen
tilts towards O$_{\rm A}$, so as to form a
hydrogen bond O$_{\rm A}$--H$_{\rm I}$ of
length 1.81~\AA\ on SnO$_2$~(110) and 1.80~\AA\ on TiO$_2$~(110). The
separation between O$_{\rm A}$ and O$_{\rm I}$
oxygens is 2.78 and 2.77~\AA\ for
the two materials. These changes
induce a dilation of the O$_{\rm I}$--H$_{\rm I}$ bond (by
5.2\% for SnO$_2$ and 4.4\% for TiO$_2$) and smaller
dilations of bonds between oxygens and surface cations.
Relaxation to the unsymmetrical configuration causes noticeable changes to
the DOS and LDOS described in the previous section. We note (see Fig.~5b)
a significant broadening of the isolated peak due to non-bonding
$1\pi$ states of adsorbed OH$^-$. In both materials, the $3\sigma$
levels are shifted upwards and hybridize more strongly with the
valence band. Similar changes can be observed in the LDOS of the
O$_{\rm I}$--H$_{\rm I}$ group, with a marked upward shift of 3$\sigma$ into
the O(2$p$) band. The separations of the $2\sigma - 3\sigma$ and $3\sigma -
1\pi$ peaks become respectively 11.7~eV and 4.6~eV for SnO$_2$ and
13.3~eV and 3.7~eV for TiO$_2$.
\subsection{Symmetrical molecular adsorption (SM)}
We have considered the case of symmetrical molecular (SM) adsorption
in which the water molecule bonds by its oxygen to the
surface 5-fold coordinated cation and the plane of the molecule is
the (001) plane (Fig.~4c).
We find that for both materials adsorption in this geometry is
energetically favorable, the adsorption energies being
0.78~eV per H$_2$O for SnO$_2$ and 0.82~eV per H$_2$O for TiO$_2$. These
are considerably larger than the adsorption energies for
the SD geometry. (Recall that we are comparing
energies calculated with the GGA.)
We find that molecular adsorption of H$_2$O causes only minor changes to the
surface structure. Relative to the relaxed clean surface, the 5-fold
coordinated cation moves out by 0.05~\AA (SnO$_2$) and 0.03~\AA (TiO$_2$). The
bridging oxygen moves out somewhat less and the atoms of
the first M$_2$O$_2$
atomic plane move inwards. The size of the latter effect is nearly negligible
for SnO$_2$~(110) but the displacements are as
large as 0.1~\AA\ on TiO$_2$~(110).
There is also a significant deformation of the adsorbed molecule:
even though the O--H bond lengths
remain unchanged, the angle between the bonds is increased
by 10\%. This is mainly due to modification of the water 3a$_1$
orbital, because of its contribution to bonding to the surface cation.
The O-cation bond between the water molecule and the surfaces
is about 10\% longer than the corresponding bond in the bulk
crystal.
Densities of states for molecularly adsorbed water are
presented in Fig.~5c.
Compared with the DOS of the clean surface, the features attributed to the
bridging oxygens (additional peak above O(2s) and O(2p) bands)
remain practically unchanged. The modification due to
adsorbed H$_2$O can be seen as an isolated narrow peak
below the O(2s) and O(2p) bands, as well as a two-peak
contribution to the O(2p) band. These can be attributed
to 2a$_1$, 1b$_2$, 3a$_1$ and 1b$_1$ molecular states
respectively. The energy separations between peaks are
about 11.5/2.9/3.4~eV for SnO$_2$ and 11.5/3.4/3.2~eV for TiO$_2$,
which are quite
close to the free molecule results (12.0/3.6/2.1~eV),
with the biggest modification, as expected, being the
downward shift and a small splitting (for SnO$_2$~(110)) of the 3a$_1$ peak.
\subsection{Surface dissociation}
Of the geometries we have examined, the two most stable
are the UD and the SM configurations, with the difference of
adsorption energies being
$\sim$~0.6~eV for SnO$_2$ and 0.3~eV for TiO$_2$. In both
materials, dissociative adsorption is favored. It thus seems
possible that water adsorbed molecularly on the
surface can spontanously dissociate.
Whether or not this occurs
will depend crucially on the existence and height of the
energy barrier that has to be overcome when the
hydrogen bond is created and when
the proton migrates along it.
To investigate this problem, we have calculated the total energy
of the system for a number of configurations along the probable
dissociation path. In practice, we have chosen the reaction coordinate
to be the horizontal (in the surface plane) separation
of the 5-fold coordinated cation and the migrating proton.
The positions of all other atoms, as well as the vertical position
of 5-fold coordinated cation and of H$_{\rm I}$ have been relaxed
for each value of the reaction coordinate.
Results on the dependence of the adsorption energy
on the M$_{\rm I}$--H$_{\rm I}$
horizontal separation for both materials are displayed in Fig.~6.
\begin{figure}
\caption{Dependence of adsorption energy on the reaction coordinate
(see text) for water dissociation on SnO$_2$~(110) and TiO$_2$~(110).}
\end{figure}
For both materials, we find almost no energy barrier which would need to
be overcome on the passage from SM to UD adsorption geometries.
It is possible that there may be a very shallow minimum at the SM
geometry, but this would not be of any practical significance.
\subsection{Adsorption in the low adsorbate density limit}
In order to get insight into the dependence of adsorption
characteristics on the adsorbate density we have considered
the half-coverage case where there is one H$_2$O in every two surface
unit cells. Because of limitations on computer time, we have chosen a
single adsorption geometry, namely the
UD one, which is energetically the most
favorable in the high density limit.
The practical calculations were performed with a supercell
twice as big as the previous one, keeping however
a slab thickness of three O - M$_2$O$_2$ - O units and a vacuum width
equivalent to two units. The clean surface
calculations were repeated for this supercell in order
to ensure cancellation of errors between the clean and hydroxylated slabs.
The positions of all slab and adsorbate atoms were relaxed to equilibrium.
We find that the adsorption energy is greater for the half-coverage case,
namely 1.63~eV per H$_2$O for SnO$_2$~(110).
Compared with full coverage case, the atomic structure of the surface and
adsorbate are only slightly modified.
\section{Discussion}
Our calculations on the energies of different relaxed configurations
indicate that dissociative adsorption gives
the largest adsorption energy and that the most stable configuration
is unsymmetrical. The adsorption energies at full
coverage are 1.39 and 1.08~eV per water molecule for SnO$_2$ and TiO$_2$,
increasing to 1.63 for SnO$_2$ at half coverage.
It is important to note that dissociative adsorption gives the
strongest binding, even though we are dealing with the perfect
surface and no defects are involved.
This appears to be an entirely
geometrical effect, since our results are very similar for SnO$_2$ and
TiO$_2$, in spite of their different electronic structure. In fact, our
entire set of results both for molecular and for dissociative
adsorption, and for the instability of the molecularly adsorbed state, is
remarkably similar for the two materials. Their similarity may perhaps
be understood by noting that they differ mainly in their {\em unoccupied}
electronic states.
In the TDS experiments referred to in the Introduction \cite{hug94,ger95},
the high temperature features at 435~K (SnO$_2$) and 375~K (TiO$_2$) were
attributed to desorption from the dissociated state. Adsorption energies
can be deduced from these desorption temperatures following the
analysis of Redhead~\cite{red62}, which expresses the desorption energy rate as
the
product of an effective vibrational frequency and a Boltzman factor
$\exp(-E_{\rm ads}/kT)$, where $E_{\rm ads}$ is the adsorption energy.
Assuming a frequency of 10$^{13}$ s$^{-1}$, we find that this
analysis gives adsorption energies of 1.1 and 1.0~eV for SnO$_2$ and TiO$_2$
respectively.
Our calculated values are thus in fair agreement with experiment,
though they are systematically too high. It is likely that this discrepancy
is related to the fact that our DFT calculations -- even with gradient
corrections -- underestimate the energy of the reaction H$_2$O~$\rightarrow$
OH$^-$~+ H$^+$ (see sec. 3.2). The fact that the adsorption energy for
SnO$_2$ is greater than that for TiO$_2$ is correctly given by our
calculations.
The relation of our results to the two lower temperature peaks in the
TDS spectra depends on how one interprets these peaks. The peak at the
lowest temperature appears to arise from desorption of molecular
H$_2$O when more than a monolayer is present, i.e. desorption
of H$_2$O bound to other H$_2$O. Our calculations clearly have nothing
to say about this. The intermediate peak at 300~K (SnO$_2$) and
275~K (TiO$_2$) was also attributed mainly to desorption from the
molecularly bound state. Our calculations clearly suggest that molecular
H$_2$O directly bound to the (110) surface is unstable with respect
to dissociation.
It is relevant to note that the semi-empirical calculations
of Bredow and Jug~\cite{bre95} on the isolated H$_2$O molecule
on the TiO$_2$ (110) surface found a barrier separating the
molecular and dissociated states, but the height of this
barrier was only $\sim$~0.2~eV.
There are then two possible interpretations of the
experiments: either the water desorbed at the intermediate temperature is
not bound directly to the oxide, but is perhaps bound to hydroxyl groups;
or there is some more stable molecularly adsorbed state which we have not
examined. Both of these seem quite plausible. The second interpretation
could be probed by making a more extensive search for stable molecularly
adsorbed configurations, perhaps using dynamical simulations, and we
hope to return to this.
Our calculations have predicted an appreciable increase of dissociative
adsorption energy with decreasing coverage. This is consistent with
experimental findings that the coverage of dissociatively adsorbed
water on SnO$_2$ and TiO$_2$~(110) is less than a monolayer, and can be
attributed to the effect of repulsion between hydroxyl groups. The same
effect has been found in semi-empirical quantum calculations on the
dissociative adsorption of H$_2$O on TiO$_2$~\cite{gon95,gon93}.
Calculated local densities of states for the oxygen atom $O_{\rm A}$
in the adsorbed OH$^-$ group (geometries SD and UD) or
for the adsorbed water molecule (geometry SM) strongly resemble
the spectra of the corresponding free molecules. The principal modification
concerns the molecular orbital directly involved in formation of the bond
with the surface, $3\sigma$ for OH$^-$ and $3a_1$ for H$_2$O. In most
cases modifications consist of a downward shift (relative to the
free molecule) accompanied by a small splitting, and are consistent with
UPS findings on hydroxylated TiO$_2$~(110)~\cite{kur89}.
It is also worth noticing, that for the more stable unsymmetrical
geometry the peak related to nonbonding $1\pi$ or $1b_1$ is considerably
broadened or split by the formation of a hydrogen bond and hybridization with
the surface O(2p) band.
On the other hand adsorption of the proton on the bridging oxygen introduces
a strong modification to its LDOS. This is clearest for the symmetrical
dissociative adsorption case, where the additional electrostatic field
of adsorbed H$^+$ causes a substantial downward shift of the
totality of LDOS(O$_{\rm I}$) compared to the clean surface.
In addition a distinct peak due to bonding O--H states appears below
O(2p) band. The overall structure of the LDOS on bridging oxygen
for dissociative adsorption is quite similar
to the free molecule spectrum, although for the unsymmetrical geometry
broadening and splitting
of both features in the region of the O(2p) band become important.
Comparison of our results with experimental UPS spectra~\cite{kur89,ger95}
shows quite
good agreement. However it should be kept in mind that experimental spectra
taken at low temperatures tend to show molecularly adsorbed water which may
be in higher adsorbed layers, rather than being bound directly to the surface.
This could explain the much closer resemblance
of measured spectra to those of free molecules.
On the other hand, the high-temperature spectra showing a two-peak structure
which can be assigned to OH$^-$ molecular orbitals relate well to our
LDOS for adsorbed oxygen. A surprising feature of the experimental results
is the total absence of signal from OH groups formed on bridging oxygens, which
according to our calculations should also give a characteristic,
two-peak structure.
\section{Conclusions}
We have studied molecular and dissociative adsorption of water on SnO$_2$~(110)
and SnO$_2$~(110) by an {\em ab initio}, density functional approach.
We found that inclusion of gradient corrections to the LDA noticeably
reduces the adsorption energies, a tendency which has been found already
for surface adsorption on metals.
For both materials, in the full coverage regime, we find that both molecular
and dissociative adsorption are energetically favorable. However,
dissociative adsorpton has a substantially greater adsorption energy.
Investigation of a possible reaction path for the surface dissociation of water
shows the absence of any energy barrier.
We found an increase of adsorption energy in the half-coverage
regime, which indicates the existence of an effective repulsive interaction
between adsorbed species and suggests that the most stable adsorption
configurations belong to the low coverage limit.
Calculated valence band DOS spectra for molecular and dissociative
adsorption geometries show structures similar to these of free
H$_2$O and OH$^-$ molecules. Bonding to the surface introduces
small shifts and broadening of peaks in the region of the surface valence band.
\begin{ack}
The work of JG is supported by EPSRC grant GR/J34842. The major
calculations were performed on the Intel iPSC/860 parallel computer
at Daresbury Laboratory, and we are grateful for a generous
allocation of time on the machine. Analysis of the results was
performed using local hardware funded by EPSRC grant GR/J36266.
Assistance from L.~N.~Kantorovich, J.~M.~Holender and
J.~A.~White is also acknowledged.
\end{ack}
|
\subsection*{Introduction}
This paper presents a proof of the following theorem:
\bte
\label{main}
\footnote
{\rm\ It follows from an e-mail discussion between G.~Hjorth
and the author in May -- July 1995 that G.~Hjorth may have proved
equal or similar theorem independently.}
\ Let\/ $\mathbin{\relf{E}}$ be a\/ $\fs11$
equivalence on reals. Assume that\its
\begin{itemize}
\item[$(\dag)$] each real belongs to a ``virtual''
generic extension~\footnote
{\rm\ By a generic extension of some $M$ we always mean
a set generic extension via a forcing notion $P\in M.$
Here the extensions could be different for different reals.}
of the constructible universe~$\rbox{L}.$\its
\end{itemize}
Then at least one~\footnote
{\rm\ If all reals are constructible from one of them then
the statements are compatible.}
of the following two statements hold$:$\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{{\rm\hskip2pt(\Roman{enumi})\hskip2pt}}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{1} \hspace{-1\mathsurround}
$\mathbin{\relf{E}}$ admits a\/ $\fdh1$ reduction~\footnote
{\rm\ By $\fdh1$ we denote the class of all subsets of $\rbox{HC}$
(the family of all hereditarily countable sets) which are
$\id{}1$ in $\rbox{HC}$ by formulas which may contain
\underline{reals and countable ordinals} as parameters.}
to the equality on the set\/ $2^{<\om_1}$ of all countable
binary sequences$.$\its
\itla{2} \hspace{-1\mathsurround} $\mathbin{\relf{E}_0}\sqq\mathbin{\relf{E}}$ continuously$.$
\end{enumerate}
\ete
\subsubsection*{Remarks on the theorem}
By a {\it ``virtual'' generic extension $\rbox{L}$\/} we mean a
set generic extension, say, $\rbox{L}[G],$ which is not necessarily
an inner class in the basic universe $\rbox{V}$ (in other words,
$G\in\rbox{V}$ is not assumed).~\footnote
{\rm\ The assumption that a set $S\sq\ord$ belongs to a
``virtual'' set generic extension of $\rbox{L}$ can be adequately
formalized as follows: {\it there exists a Boolean valued extension
of $\rbox{L}[S]$ in which it is true that the universe is a set generic
extension of the constructible universe\/}, see Lemma~\ref{44}
below.}
Notice that the assumption $(\dag)$ of the theorem follows e.\ g.
from the hypothesis that the universe is a set generic extension
of $\rbox{L}.$ In fact the theorem remains true in a weaker assumption
that each real $x$ belongs to a ``virtual'' generic extension
of $\rbox{L}[z_0]$ for one and the same real $z_0$ which does not depend
on $x$.
We refer the reader to Harrington, Kechris, and Louveau~\cite{hkl}
on matters of the early history of ``Glimm -- Effros'' theorems ---
those of type: {\it each equivalence of certain class either
admits a reduction to equality or embeds\/ $\mathbin{\relf{E}_0}$} --- and relevant
problems in probability and the measure theory. (However
Section~\ref{ulm} contains the basic notation.)
The modern history of the topic began in the
paper \cite{hkl} where it is proved that each Borel equivalence
on reals either admits a Borel reduction to the equality on reals
or embeds $\mathbin{\relf{E}_0}.$ The proof is based on an advanced tool in
descriptive set theory, the
{\it Gandy -- Harrington topology\/} on reals, generated by
$\is11$ sets.
Hjorth and Kechris~\cite{hk} found that the case of $\fs11$
relations is much more complicated. Some examples have shown
that one cannot find a reasonable ``Glimm -- Effros'' result for
$\fs11$ relations simply taking a nonBorel reduction in \ref{1}
or discontinuous embedding in \ref{2}; it seemms that the equality
on {\it reals\/} rather than countable binary sequences in \ref{1}
does not match completely the nature of $\fs11$ relations.
Hjort and Kechris \cite{hk} suggested the adequate approach: one
has to take $2^{<\om_1}$ as the domain of the equality in
\ref{1}. (This approach is referred to as the {\it Ulm -- type
classification\/} in \cite{hk}, in connection with a
classification theorem of Ulm in algebra.)
On this way they proved that the dichotomy \ref{1} vs. \ref{2}
holds for each $\fs11$ equivalence relation on reals, in the
assumption of the ``sharps'' hypothesis (and the latter can be
dropped provided the $\fs11$ relation occasionally has only Borel
equivalence classes).
Theorem~\ref{main} of this paper establishes the same result (not
paying attention on the possible compatibility of \ref{1} and
\ref{2}) in the completely different than sharps assumption: each
real belongs to a generic extension of $\rbox{L}.$ Of course it is
the principal problem (we may refer to the list of open problems
in~\cite{hk}) to eliminate the ``forcing'' assumption and prove
the result in $\ZFC$.
One faces much more problems in higher projective classes. In
fact there exists a sort of upper bound for ``Glimm -- Effros''
theorems in $\ZFC.$ Indeed, in a nonwellfounded (of ``length''
$\om_1\times\mathord{{\sf Z}\hspace{-4.5pt}{\sf Z}},$ i. e. $\om_1$ successive copies of the integers)
iterated Sacks extension~\footnote
{\rm\ See Groszek~\cite{g94} or Kanovei~\cite{k-sacks} on matters
of nonwellfounded Sacks iterations.}
of $\rbox{L}$ the $\is12$ equivalence
$$
x\mathbin{\relf{E}} y\hspace{6mm}\hbox{iff}\hspace{6mm}\rbox{L}[x]=\rbox{L}[y]
$$
neither continuously embeds $\mathbin{\relf{E}_0}$ nor admits a real--ordinal
definable reduction to the equality on ${\skri P}(\kappa)$ for a
cardinal $\kappa$.
Thus the interest can be paid on classes $\fp11,$ $\fd12,$
$\fp12.$ One may expect that $\fd12$ relations admit a theorem
similar to Theorem~\ref{main}.~\footnote
{\rm\
G.~Hjorth informed the author that he had partial results in this
domain.}
More complicated relations can be investigated in strong
extensions of $\ZFC$ or in special models. Hjorth~\cite{h-det}
proved that in the assumption of ${\bf AD}$ and
$\rbox{V}=\rbox{L}[{\rm reals}]$ every equivalence on reals either admits
a reduction (here obviously a real--ordinal definable reduction)
to the equality on a set $2^\kappa,$ $\kappa\in\ord,$ or
continuously embeds $\mathbin{\relf{E}_0}.$ Kanovei~\cite{k-sm} proved even
a stronger result (reduction to the equality on $2^{<\om_1}$) in
Solovay model for $\ZF+\DC$.
\subsubsection*{The organization of the proof}
Theorem~\ref{main} is the main result of this paper. The proof is
arranged as follows.
First of all, we shall consider only the case when $\mathbin{\relf{E}}$ is a
lightface $\is11$ relation; if in fact $\mathbin{\relf{E}}$ is $\is11(z)$ in some
$z\in{\skri N}$ then this $z$ simply enters the reasoning in a uniform
way, not influenting substantially any of the arguments.
The splitting point between the statements \ref{1} and \ref{2}
of Theorem~\ref{main} is determined in Section~\ref{ulm}.
It occurs that we have \ref{1} in the assumption that
\begin{itemize}
\item[$(\ddag)$] each real $x$ belongs to a ``virtual'' \dd\la
collapsing generic extension of $\rbox{L}$ (for some ordinal $\la$)
in which $\mathbin{\relf{E}}$ is closed in a
topolody generated by $\rbox{OD}$ sets on the set ${\cD\cap{\sf Weak}_\la(\rbox{L})}$
of all reals \dd\la weak over $\rbox{L}.$ (We say that $x\in \cD$ is
{\it\dd\la weak over\/} $\rbox{L}$ iff it belongs to a \dd\al
collapsing extension of $\rbox{L}$ for some $\al<\la$.)
\end{itemize}
On the opposite side, we have \ref{2} provided the assumption
$(\ddag)$ fails.
Both sides of the proof depend on properties of reals in collapsing
extensions close to those of Solovay model. The facts we need are
reviewed in Section~\ref{clos}.
Section~\ref{ssif} proves assertion \ref{1} of Theorem~\ref{main}
assuming $(\ddag).$ The principal idea has a semblance of the
corresponding parts in \cite{hkl} and especially
\cite{hk}~\footnote
{\rm\ Yet we use a technique different from the approach of
\cite{hk}, completely avoiding any use of recursion theory.}
: in the assumption
of $(\ddag),$ each \dd\la weak over $\rbox{L}$ real in the relevant
``virtual'' \dd\la collapsing extension belongs to a set (one and
the same
for all \dd\mathbin{\relf{E}} equivalent reals) which admits a characterization
in terms of an element of $2^{<\om_1}.$ An absoluteness argument
allows to extend this fact to the universe of Theorem~\ref{main}.
Sections \ref{prod} and \ref{or} prove \ref{2} of
Theorem~\ref{main} in the assumption that $(\ddag)$ {\it fails\/}
(but $(\dag)$ still holds, as Theorem~\ref{main} assumes). In fact
is this case $\mathbin{\relf{E}}$ is {\it not\/} closed on the set
$\cD\cap{\sf Weak}_\la(\rbox{L})$ in a ``virtual'' \dd\la collapsing extension
of $\rbox{L}$ for some $\la.$ This suffices to see that $\mathbin{\relf{E}}$ embeds
$\mathbin{\relf{E}_0}$ continuously in the ``virtual'' universe; moreover, $\mathbin{\relf{E}}$
embeds $\mathbin{\relf{E}_0}$ in a certain special sense which can be expressed by
a $\is12$ formula (unlike the existence of an embedding in general
which needs $\is13$). We conclude that $\mathbin{\relf{E}}$ embeds $\mathbin{\relf{E}_0}$ in the
universe of Theorem~\ref{main} as well by Shoenfield.
The construction of the embedding of $\mathbin{\relf{E}_0}$ into $\mathbin{\relf{E}}$ follows the
principal idea of Harrington, Kechris, and Louveau~\cite{hkl}, yet
associated with another topology and arranged in a different way.
(In particular we do not play the strong Choquet game to define
the necessary sequence of open sets.) \vspace{4mm}
\noi
{\bf Important remark} \\[1mm]
It will be more convenient to consider $\cD=2^\om,$ the {\em
Cantor space\/}, rather than ${\skri N}=\om^\om,$ as the basic Polish
space for which Theorem~\ref{main} is being proved.
\newpage
\subsection{Approach to the proof of the main theorem}
\label{ulm}
First of all, we shall prove only the ``lightface'' case of the
theorem, so that $\mathbin{\relf{E}}$ will be supposed to be a $\is11$ equivalence
on reals. The case when $\mathbin{\relf{E}}$ is $\is11[z]$ for a real $z$ does not
differ much: the $z$ uniformly enters the reasoning.
By ``reals'' we shall understand points of the
{\it Cantor set\/} $\cD=2^\om$ rather than the {\it Baire space\/}
${\skri N}=\om^\om;$ this choice is implied by some technical reasons.
The purpose of this section is to describe how the two cases of
Theorem~\ref{main} will appear. This needs to recall some
definitions.
\subsubsection{Collapsing extensions}
\label{ce}
Let $\al$ be an ordinal. Then $\col\al$ is the forcing to collapse
$\al$ down to $\om.$ If $G\sq\col\al$ is \dd{\col\al}generic over
a transitive model $M$ ($M$ is a set or a class) then $f=\bigcup G$
is a function from $\om$ onto $\al,$ so that $\al$ is countable in
$M[G]=M[f].$ Functions $f:\om\,\lra\,\al$ obtained this way will be
called \dd{\col\al}{\em generic over\/ $M$.}
By {\em \dd\la collapse universe hypothesis\/}, \cuh\la{} in
brief, we shall mean the following assumption: $\rbox{V}=\rbox{L}[f_0]$
for a \dd{\col\la}generic over $\rbox{L}$ collapse function
$f_0\in\la^\om$.
By the assumption of Theorem~\ref{main}, each real $z$ belongs to
a ``virtual'' \dd{\col\la}generic extension of $\rbox{L},$ the
constructible universe, for some ordinal $\la.$ Such an extension
satisfies \cuh\la.
\begin{remark}\rmt\
\label{rr}
The extension is not necessarily supposed to be an inner class in
the universe of Theorem~\ref{main}, see Introduction.\qed
\erem
A set is \dd\la{\em weak over $M$} ($\la$ an ordinal in a model
$M$) iff it belongs to a ``virtual'' \hbox{\dd{\col\al}generic}
extension of $M$ for some $\al<\la.$
We define
$$
{\sf Weak}_\la(M)=\ans{x:x\,\hbox{ is \dd\la weak over }\,M}\,.
$$
In the assumption \cuh\la{}, reals in ${\sf Weak}_\la(\rbox{L})$ behave
approximately like all reals in Solovay model.
\subsubsection{The $\protect\rbox{OD}$ topology}
\label{top}
In $\ZFC,$ Let ${\cal T}$ be the topology generated on a given set
$X$ (for instance, $X=\cD=2^\om,$ the Cantor set) by all $\rbox{OD}$
subsets of $X.$ ${\cal T}^2$ is the product of two copies of ${\cal T},$
a topology on $\cD^2$.
This topology plays the same role in our consideration as the
Gandy -- Harrington topology in the proof of the classical
Glimm -- Effros theorem (for Borel relations) in
Harrington, Kechris, and Louveau~\cite{hkl}. In particular, it
has similar (although not completely similar: some special
\dd{\is11}details vanish) properties.
We define $\mathbin{\overline{\relf{E}}}$ to be the \dd{{\cal T}^2}closure of $\mathbin{\relf{E}}$ in $\cD^2.$
Thus ${x\mathbin{\not{\hspace{-2pt}\overline{\relf{E}}}} y}$ iff there exist $\rbox{OD}$ sets $X$ and $Y$ containing
resp. $x$ and $y$ and such that ${x'\mathbin{\not{\hspace{-2pt}\relf{E}}} y'}$ for all ${x'\in X,}$
${y'\in Y}.$ Obviously $X$ and $Y$ can be chosen as \dd\mathbin{\relf{E}}
invariant (simply replace them by their \dd\mathbin{\relf{E}} saturations), and
then $Y$ can be replaced by the complement of $X,$ so that
$$
x\mathbin{\overline{\relf{E}}} y\;\;\llra\;\;\forall\,X\;
[\,X\hbox{ is }\rbox{OD} \cj X\hbox{ is \dd\mathbin{\relf{E}} invariant}\;\,
\lra\;\,(x\in X\;\llra\;y\in X)\,]\,.
$$
Therefore $\mathbin{\overline{\relf{E}}}$ is an $\rbox{OD}$ equivalence on $\cD$.
\subsubsection{The cases}
\label{cases}
In \cite{hkl}, the two cases are determined by the equality
$\mathbin{\relf{E}}=\mathbin{\overline{\relf{E}}}:$ if it holds that $\mathbin{\relf{E}}$ admits a Borel reduction on
$\Da(\cD),$ otherwise $\mathbin{\relf{E}}$ embeds $\mathbin{\relf{E}_0}.$ Here the splitting
condition is a little bit more complicated. First of all, we
have to consider the equality in different universes. Second,
the essential domain of the equivalence is now a proper subset of
$\cD,$ the set of all weak reals.\vspace{2mm}
\noi
{\bf Case\ 1.}\ \ For each real $z,$ there exist an ordinal
$\la$ and a ``virtual'' \dd{\col\la}generic extension $V$ of the
constructible universe $\rbox{L}$ containing $z$ such that the
following is true in $V:$ $\mathbin{\relf{E}}$ coincides with $\mathbin{\overline{\relf{E}}}$ on
$\cD\cap{\sf Weak}_\la(\rbox{L})$ and $x$ is \dd\la weak over $\rbox{L}$.\vspace{2mm}
(Notice that, for a $\is11$ binary relation $\mathbin{\relf{E}},$ the assertion
that $\mathbin{\relf{E}}$ is an equivalence is $\ip12,$ therefore absolute for
all models with the same ordinals, in particular for $\rbox{L}$ and
all generic extensions of $\rbox{L}$.)\vspace{2mm}
\noi
{\bf Case\ 2.}\ \ Not Case 1.
\bte
\label{mt}
Suppose that each real belongs to a ``virtual'' generic extension
of\/ $\rbox{L}.$ Then, for the given\/ $\is11$ equivalence relation\/
$\mathbin{\relf{E}},$ we have\/ \its
\begin{itemize}
\item[--] assertion
\ref{1} of Theorem~\ref{main} in Case 1,\hfill and\hfill\its
\item[--] assertion \ref{2} of Theorem~\ref{main} in Case 2.
\end{itemize}
\ete
This is how Theorem~\ref{main} well be proved.
\newpage
\newcommand{{\underline S}}{{\underline S}}
\subsection{On collapsing extensions}
\label{clos}
In this section, we fix a limit constructible cardinal $\la.$ The
purpose is to establish some properties of \dd\la collapsing
generic extensions (= the universe under the hypothesis
\hbox{\cuh\la}). It will be shown that weak ponts (introduced in
Section~\ref{ulm}) behave approximately like all reals in
Solovay model.
\subsubsection{Basic properties}
\label{bp}
We recall that a set $S$ is \dd\la{\em weak over $M$} iff $S$
belongs to an \dd{\col\al}generic extension of the model $M$
for some $\al<\la$.
The hypothesis \cuh\la{} (the one which postulates that the
universe is a \dd\la generic extension of $\rbox{L}$) will be assumed
during the reasoning, but we shall not mind to specify \cuh\la{}
in all formulations of theorems.
\bpro
\label{col}
Assume\/ \cuh\la. Let\/ $S\sq \ord$ be\/ \dd\la weak over\/ $\rbox{L}.$
Then\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{{\arabic{enumi}}}
\def\theenumi{{\rm{\rmt\arabic{enumi}.}}.}
\itla{sm1}
The universe\/ $\rbox{V}$ is a\/ \dd{\col\la}generic extension of
$\rbox{L}[S]$.\its
\itla{sm2}
If\/ $\Phi$ is a sentence containing only sets in\/ $\rbox{L}[S]$ as
parameters then\/ $\La$ {\rm(}the empty sequence\/{\rm)} decides\/
$\Phi$ in the sense of\/ $\col\la$ as a forcing notion over
$\rbox{L}[S]$.\its
\itla{sm4}
If a set\/ $X\sq\rbox{L}[S]$ is\/ $\rbox{OD}[S]$ then\/ $X\in\rbox{L}[S]$.
\end{enumerate}
\epro
($\rbox{OD}[S]=S$\hspace{-1\mathsurround}{}--{\it ordinal definable\/}, that is, definable
by an \dd\in formula having $S$ and ordinals as parameters.)
The proof (a copy of the proof of Theorem 4.1 in Solovay~\cite{sol})
is based on several lemmas, including the following crucial lemma:
\ble
\label{44}
Suppose that\/ $P\in\rbox{L}$ is a p.o. set, and a set\/ $G\sq P$ is\/
\dd Pgeneric over\/ $\rbox{L}.$ Let\/ $S\in \rbox{L}[G],$ $S\sq\rbox{Ord}.$
Then there exists a set $\Sg\sq P,$ $\Sg\in \rbox{L}[S]$ such that\/
$G\sq\Sg$ and\/ $G$ is\/ \dd{\Sg}generic over\/ $\rbox{L}[S]$.
\ele
\noi{\bft Proof\hspace{2mm} }{}of the lemma. We extract the result from the proof of
Lemma 4.4 in \cite{sol}.
{\it We argue in $\rbox{L}[S]$}.
Let ${\underline S}$ be the name for $S$ in the language of the forcing $P$.
Define a sequence of sets
$A_\al\sq P\;\;(\al\in\ord)$ by induction on $\al$.\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{{\rm\hskip1pt(A\arabic{enumi})\hskip1pt}}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{aa1}\hspace{-1\mathsurround}
$p\in A_0$ iff either $\sg\in S$ but $p$ forces (in $\rbox{L}$ and in
the sense of $P$ as the notion of forcing) $\sg\not\in{\underline S},$ or
$\sg\not\in S$ but $p$ forces $\sg\in{\underline S}$ \ \ --- \ \ for some
$\sg\in\ord$.\its
\itla{aa2}\hspace{-1\mathsurround}
$p\in A_{\al+1}$ iff there exists a dense set $D\sq P,$ $D\in \rbox{L}$
such that every $q\in D$ satisfying $p\<q$ (means:
$q$ is stronger than $p$) belongs to $A_\al$.\its
\itla{aa3}
If $\al$ is a limit ordinal then $A_\al=\bigcup_{\ba<\al}A_\ba$.\its
\end{enumerate}
The following properties of these sets are easily verifiable
(see Solovay \cite{sol}): first, if\/ $p\in A_\al$ and\/
$p\< q\in P$ then\/ $q\in A_\al$, second,
if\/ $\ba<\al$ then\/ $A_\ba\sq A_\al$.
Since each $A_\al$ is a subset of $P,$ it follows that
$A_\da= A_{\da+1}$ for some ordinal $\da.$ We put
$\Sg=P\setminus A_\da.$ Thus $\Sg$ intends to be the set of all
conditions $p\in P$ which do not force something about ${\underline S}$
which contradicts the factual information about $S$.
We prove, following \cite{sol}, that $\Sg$ is as required. This
yields a pair of auxiliary facts.\vom
$(\Sg1)$ $G\sq\Sg$.\vom
\noi
Indeed assume on the contrary that
$G\cap A_\ga\not=\emptyset$ for some $\ga.$ Let $\ga$ be the least
such an ordinal. Clearly $\ga$ is not limit and $\ga\not=0;$ let
$\ga=\al+1.$ Let $p\in A_\ga\cap G.$ Since $G$ is generic,
definition~\ref{aa2} implies $G\cap A_\al\not=\emptyset,$
contradiction.\vom
$(\Sg2)$ {\it If\/ $D\in \rbox{L}$ is a dense subset of\/ $P$ then\/
$D\cap\Sg$ is a dense subset of\/ $\Sg$}.\vom
\noi
This is easy: if $p\in\Sg$ then $p\not\in A_{\da+1};$ hence
by~\ref{aa2} there exists $q\in D\setminus A_\da,$
$q\>p$.
We prove that $G$ is \dd\Sg generic
over $\rbox{L}[S].$ Let $D\in\rbox{L}[S]$ be a dense subset of $\Sg;$ we
have to check that $D\cap G\not=\emptyset.$ Suppose that
$D\cap G=\emptyset,$ and get a contradiction.
Since ${D\in\rbox{L}[S]},$ there exists an \dd\in formula $\Phi(x,y)$
containing only ordinals as parameters and such that $\Phi(S,y)$
holds in $\rbox{L}[S]$ iff $y=D$.
Let $\Psi(G')$ be the conjunction of the following formulas:\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{\hskip2pt(\arabic{enumi})\hskip2pt}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{1)}
\hspace{-1\mathsurround} $S'={\underline S}[G']$ (the interpretation of the ``term'' ${\underline S}$ via
$G'$)\hspace{1mm}{} is a set of ordinals, and there exists unique
$D'\in\rbox{L}[S']$ such that $\Phi(S',D')$ holds in $\rbox{L}[S']$;\its
\itla{2)}\hspace{-1\mathsurround}
$D'$ is a dense subset of $\Sg'$ where $\Sg'$ is the set obtained
by applying our definition of $\Sg$ within $\rbox{L}[S']$;\its
\itla{3)}\hspace{-1\mathsurround}
$D'\cap G'=\emptyset$.\its
\end{enumerate}
Then $\Psi(G)$ is true in $\rbox{L}[G]$ by our assumptions. Let
$p\in G$ force $\Psi$ over $\rbox{L}.$ Then $p\in\Sg$ by $(\Sg1).$ By
the density there exists $q\in D$ with $p\< q.$ We can consider
a \dd\Sg generic over $\rbox{L}[S]$ set $G'\sq\Sg$ containing $q.$
Then $G'$ is also \dd Pgeneric over $\rbox{L}$ by $(\Sg1).$
We observe that ${\underline S}[G']=S$ because $G'\sq\Sg.$ It follows that
$D'$ and $\Sg'$ (as is the description of $\Psi$) coinside with
resp. $D$ and $\Sg.$ In particular $q\in D'\cap G',$ a
contradiction because $p$ forces \ref{3)}.
\vspace{3mm}\qed
\noi{\bft Proof\hspace{2mm} }{}of the proposition.
{\em Item \ref{sm1}\/}. Lemma~\ref{44} (for $P=\col\la$)
implies that the universe
is a \dd\Sg generic extension of $\rbox{L}[S]$ for a certain tree
$\Sg\sq\col\la,$ $\Sg\in\rbox{L}[S].$ Notice that $\la$ is a cardinal
in $\rbox{L}[S]$ because $S$ is \dd\al weak over $\rbox{L}$ where $\al<\la;$
on the other hand, $\la$ is countable in the universe by \cuh\la.
It follows that there exists a condition $u\in G$ such that the
set of all \dd\la branching points of $\Sg$ is cofinal over $u$
in $\Sg.$ In other words, the set $\ans{v\in\Sg:u\sq v}$ includes
in $\rbox{L}[S]$ a cofinal subset order isomorphic to $\col\la$.
{\em Items \ref{sm2} and \ref{sm4}\/}. It suffices to
refer to item \ref{sm1} and argue as in the proofs of Lemma 3.5 and
Corollary 3.5 in \cite{sol} for $\rbox{L}[S]$ as the initial model.\qed
\subsubsection{Coding of reals and sets of reals in the model}
\label{coding}
We let $\dF\al(M)$ be the set of all \dd{\col\al}generic over $M$
functions $f\in \al^\om$.
The following definitions intend to introduce a useful coding
system for reals (i.\ e. points of $\cD=2^\om$ in this research)
and sets of reals in collapsing extensions.
Let $\al\in\ord.$ By $\cont\al$ we denote the set of all indexed
sets $=\ang{\al,\ang{_n:n\in\om}}$ -- the ``terms'' -- such that
${_n\sq\col\al}$ for each $n$.
We put $\cont{<\la}=\bigcup_{\al<\la}\cont\al$.
``Terms'' $\in\cont\al$ are used to code functions
$C:\al^\om\;\lra\;\cD=2^\om;$ namely, for every $f\in\al^\om$
we define $x=(f)\in\cD$ by: $x(n)=1$ iff $f\res m\in _n$
for some $m$.
Assume that $=\ang{\al,\ang{_n:n\in\om}}\in\cont\al,$
$u\in\col\al,$ $M$ arbitrary. We introduce the sets
${ u}(M)=\ans{(f):u\subset f\in\dF\al(M)}$ and
$(M)={ \La}(M)={\hbox{\hspace{2pt}\rmt ''}}\dF\al(M)$.
\bpro
\label{solMb}
Assume\/ \cuh\la. Let $S\sq\ord$ be\/ \dd\la weak over\/
$\rbox{L}.$ Then$:$\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{{\arabic{enumi}}}
\def\theenumi{{\rm{\rmt\arabic{enumi}.}}.}
\itla{sm6}
If\/ $\al<\la,$ $F\sq\dF\al(\rbox{L}[S])$ is\/ $\rbox{OD}[S],$ and\/
$f\in F,$ then there exists\/ $m\in\om$ such that each\/
$f'\in\dF\al(\rbox{L}[S])$ satisfying\/ $f'\res m= f\res m$ belongs
to\/ $F$.\its
\itla{sm5}
For each real\/ $x\in\cD\cap{\sf Weak}_\la(\rbox{L}[S]),$ there exist\/
$\al<\la,$ $\in\cont\al\cap\rbox{L}[S],$ and\/
$f\in\dF\al(\rbox{L}[S])$ such that\/ $x=(f)$.
\its
\itla{xl1}
Each\/ $\rbox{OD}[S]$ set $X\sq\cD\cap{\sf Weak}_\la(\rbox{L}[S])$ is a union of
sets of the form\/ $(\rbox{L}[S]),$ where $\in\cont{<\la}\cap\rbox{L}[S]$.
\its
\itla{xl2}
Suppose that\/ ${\in\cont\al\cap\rbox{L}[S],\;\;\al<\la,}$ and\/
${u\in\col\al}.$ Then every\/ $\rbox{OD}[S]$ set\/
$X\sq\linebreak[3]{{ u}(\rbox{L}[S])}$ is a union of sets of
the form\/ ${ v}(\rbox{L}[S]),$ where\/ $u\sq v\in\col\al$.
\end{enumerate}
\epro
\noi{\bft Proof\hspace{2mm} } {\em Item \ref{sm6}\/}. We observe that
$F=\ans{f'\in\al^\om:\Phi(S,f')}$ for an \dd\in formula $\Phi.$
Let $\Psi(S,f')$ denote the formula: ``$\La$ \dd{\col\la}forces
$\Phi(S,f')$ over the universe'', so that
$$
F=\ans{f'\in\al^\om:\Psi(S,f')\,\hbox{ is true in }\,\rbox{L}[S,f']}.
$$
by Proposition~\ref{col} (items \ref{sm1} and \ref{sm2}).
Therefore, since $f\in F\sq\dF\al[S],$ there exists $m\in\om$ such
that the restriction $u=f\res m$
\dd{\col\al}forces $\Psi(S,{\hat f})$ over $\rbox{L}[S]$
where $\hat f$ is the name of the \dd\al collapsing function.
{\em Item \ref{sm5}\/}. By the choice if $x,$ this real belongs to
a \dd{\col\al}generic extension of
$\rbox{L}[S].$ Thus $x\in\rbox{L}[S,f]$ where $f\in\dF\al(\rbox{L}[S]).$ Let
${\hat x}$ be the name of $x.$ It suffices to define
$_n=\ans{u\in\col\al:u\,\hbox{ forces }\,{\hat x}(n)=1}$
and take $=\ang{\al,\ang{_n:n\in\om}}$.
{\em Item \ref{xl1}\/}. Consider a real $x\in X.$ We use item 2 to
obtain
$\al<\la,$ $f\in\dF\al(\rbox{L}[S]),$ and\/ $\in\cont\al\cap\rbox{L}[S]$
such that\/ $x=(f).$ Then we apply item 1 to the $\rbox{OD}[S]$ set
$F=\ans{f'\in\dF\al[S]:(f')\in X}$ and the $f$ defined above.
This results in a condition $u=f\res m\in\col\la$ ($m\in\om$)
such that $x\in { u}[S]\sq X.$ Finally the set
${ u}[S]$ is equal to ${'}[S]$ for some other
$'\in \cont\al\cap\rbox{L}[S]$.
{\em Item \ref{xl2}\/}. Similar to the previous item.\qed
\newpage
\subsection{The case of closed relations: classifiable points}
\label{ssif}
In this section, we prove the ``case 1'' of Theorem~\ref{mt}.
Thus let $\mathbin{\relf{E}}$ be a $\is11$ equivalence relation.
\subsubsection{Classifiable points}
First of all, we introduce the notion of an \dd\mathbin{\relf{E}} classifiable
point.
As usual, $\rbox{HC}$ denotes the set of all hereditarily countable
sets. $\ish1$ will denote the collection of all subsets of $\rbox{HC}$
definable in $\rbox{HC}$ by a parameter-free $\is{}1$ formula. The class
$\iph1$ is understood the same way, and $\idh1=\ish1\cap\iph1$.
Let us fix a constructible $\idh1$ enumeration
$\cont{}\cap\rbox{L}=\ans{\tk\xi:\xi<\om_1}$ such that each
$\in\cont{}\cap\rbox{L}$ has uncountably many numbers
$\xi<\om_1$ satisfying $=\tk\xi.$
The following lemma gives a more special characterization for
$\mathbin{\overline{\relf{E}}},$ the \dd{{\cal T}^2}closure of $\mathbin{\relf{E}},$ based on this enumeration.
\ble
\label{har}
Assume\/ \cuh\la. Let\/ $x,\,y\in\cD\cap{\sf Weak}_\la(\rbox{L}).$ Then\/
${x\mathbin{\overline{\relf{E}}} y}$ if and only if for each\/ $\xi<\om_1$ we have\/
$x\in[{\tk\xi}(\rbox{L}_\xi)]_{\mathbin{\relf{E}}}\;\llra\;y\in[{\tk\xi}(\rbox{L}_\xi)]_{\mathbin{\relf{E}}}$.
\ele
\noi{\bft Proof\hspace{2mm} } The ``only if'' part follows from the fact that the sets
${\tk\xi}(\rbox{L}_\ga)$ are $\rbox{OD}.$ Let us prove the ``if'' direction.
Assume that ${x\mathbin{\not{\hspace{-2pt}\overline{\relf{E}}}} y}.$ There exists an $\rbox{OD}$ set $X$ such that
$x\in [X]_{\mathbin{\relf{E}}}$ but $y\not\in [X]_{\mathbin{\relf{E}}}.$ By
Proposition~\ref{solMb}, we obtain $x\in (\rbox{L})\sq [X]_{\mathbin{\relf{E}}},$
where
$=\ang{\al,\,\ang{_n:n\in\om}}\in\cont\al\cap\rbox{L},$ $\al<\la.$
Since $\la$ is a limit cardinal in $\rbox{L},$ there exists a
constructible cardinal $\ga,$ $\al<\ga<\la,$ such that
$\dF\al(\rbox{L})=\dF\al(\rbox{L}_\ga).$ Then
$'=\ang{\ga,\,\ang{_n:n\in\om}}$ is $\tk\xi$ for some
$\xi,$ $\ga\<\xi<\om_1.$
Then $(\rbox{L})={\tk\xi}(\rbox{L}_\xi)$.\qed
\vspace{4mm}
For each $x\in\cD,$ we define $\vpi_x\in 2^{\om_1}$ as
follows: $\vpi_x(\xi)=1$ iff $x\in [{\tk\xi}(\rbox{L}_\xi)]_{\mathbin{\relf{E}}}$.
\bdf
\label{psi}
We introduce the notion of a \hbox{\dd\mathbin{\relf{E}} c}lassifiable
point.
We let $T$ be the set of all triples $\ang{x,\psi,t}$ such that
$x\in\cD,$ $\psi\in 2^{<\om_1},$
$\in\cont\al\cap\rbox{L}_{\ga}[\psi],$ where
$\al<\ga=\rbox{dom}\,\psi<\om_1,$ and the following conditions \ref{aa}
through \ref{cc} are satisfied.
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{\hskip2pt{\rm (\alph{enumi})}\hskip2pt}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{aa} \hspace{-1\mathsurround}
$\rbox{L}_{\ga}[\psi]$ models $\ZFC^-$ (minus the Power Set axiom) so
that $\psi$ can occur as an extra class parameter in Replacement
and Separation.\its
\itla{bb}
It is true in $\rbox{L}_{\ga}[\psi]$ that $\ang{\La,\La}$
forces ${({\hat{f}})\mathbin{\relf{E}}({\hat{g}}})$ in the sense of
$\col\al\!\times\!\col\al$ as the forcing, where ${\hat{f}}$ and
${\hat{g}}$ are names for the generic functions in $\al^\om$.\its
\itla{xx}
For each $\xi<\ga,$ $\psi(\xi)=1$ iff
$x\in [{\tk\xi}(\rbox{L}_\xi)]_{\mathbin{\relf{E}}}$ --- so that
$\psi=\vpi_x\res\ga$.\its
\itla{cc}
\hspace{-1\mathsurround}
$x$ belongs to $[(\rbox{L}_\ga[\psi])]_{\mathbin{\relf{E}}}$.
\end{enumerate}
A point $x\in\cD$ is {\em \dd\mathbin{\relf{E}} classifiable} iff there
exist $\psi$ and $$ such that $\ang{x,\psi,}\in T$.\qed
\edf
The author learned from Hjorth and Kechris~\cite{hk} the
idea of forcing over countable models to get a $\id{}1$ reduction
function, the key idea of this definition.
\ble
\label{def}
$T_{\mathbin{\relf{E}}}$ is a\/ $\idh1$ set\/ {\rm(provided\/ $\mathbin{\relf{E}}$ is $\is11$)}.
\ele
\noi{\bft Proof\hspace{2mm} } Notice that conditions \ref{aa} and \ref{bb} in
Definition~\ref{psi} are $\idh1$ because they reflect truth
within $\rbox{L}_\ga[\psi]$ and the enumeration $\tk\xi$ was chosen
in $\idh1$.
Condition \ref{cc} is obviously $\ish1$ (provided $\mathbin{\relf{E}}$ is at least
$\is12$), so it remains to convert it also to
a $\iph1$ form. Notice that in the assumption of \ref{aa} and
\ref{bb}, the set $X=(\rbox{L}_{\ga}[\psi])$ consists of pairwise
\dd\mathbin{\relf{E}} equivalent points.
(Indeed, consider a pair of \dd{\col\al}generic over
$\rbox{L}_{\ga}[\psi]$ functions $f,\,g\in\al^\om$ (not necessarily a
{\it generic pair\/}). Let $h\in\al^\om$ be an \dd{\col\al}generic
over both $\rbox{L}_{\ga}[\psi,f]$ and $\rbox{L}_{\ga}[\psi,g]$ function.
Then, by \ref{bb}, ${(h)\mathbin{\relf{E}}(f)}$ holds in
$\rbox{L}_{\ga}[\psi,f,h],$ therefore in the universe by Shoenfield.
Similarly, ${(h)\mathbin{\relf{E}}(g)}.$ It follows that
${(f)\mathbin{\relf{E}}(g)},$ as required.)
Therefore \ref{cc} is equivalent to
the formula
$\forall\,y\in (\rbox{L}_{\ga}[\psi])\;(x\mathbin{\relf{E}} y)$ because
$(\rbox{L}_{\ga}[\psi])$ is not empty. This is
clearly $\iph1$ provided $\mathbin{\relf{E}}$ is at least $\ip12$.
Let us consider \ref{xx}. The right--hand side of the equivalence
``iff'' in \ref{xx} is $\is11$ with inserted $\idh1$ functions,
therefore $\idh1.$ It follows that \ref{xx} itself is
$\idh1$.~\footnote
{\rm\ Here we do not see how to weaken the assumption that
$\mathbin{\relf{E}}$ is $\is11;$ even if the relation is $\ip11,$ \ref{xx} becomes
$\idh2$.}
\qed
\subsubsection{The classification theorem}
\label{ct}
The following lemma will allow to define a $\idh1$ reduction of
the given $\is11$ equivalence relation $\mathbin{\relf{E}}$ to the equality on
$2^{<\om_1}$.
\ble
\label{key2}
In the assumption of Case 1 of Subsection~\ref{cases}, each
point\/ $x\in\cD$ is\/ \dd\mathbin{\relf{E}} classifiable.
\ele
\noi{\bft Proof\hspace{2mm} } Let $x\in\cD.$ By the assumption of Case 1, there exist an
ordinal $\la$ and a ``virtual'' \dd{\col\la}generic extension $V$
of the constructible universe $\rbox{L}$ containing $x$ such that $\mathbin{\relf{E}}$
coincides with $\mathbin{\overline{\relf{E}}}$ on $\cD\cap{\sf Weak}_\la(\rbox{L})$ in $V$ and $x$ is
\dd\la weak over $\rbox{L}$ in $V$.
Thus we have the two universes, $V$ and the universe of the lemma,
with one and the same class of ordinals. Since by Lemma~\ref{def}
``being \dd\mathbin{\relf{E}} classifiable'' is a $\ish1,$ therefore $\is12$
notion, it suffices to prove that $x$ is \dd\mathbin{\relf{E}} classifiable in the
``virtual'' universe $V$.
We observe that \cuh\la{} is true in $V$.
{\it We argue in $V$}.
Notice that $\vpi=\vpi_x$ is \dd\la weak over $\rbox{L}:$ indeed
$\vpi\in\rbox{L}[x]$ by Proposition~\ref{col} since $\vpi$ is $\rbox{OD}[x]$.
It follows that $[x]_{\mathbin{\relf{E}}}$ is $\rbox{OD}[\vpi]$ by Lemma~\ref{har},
because $\mathbin{\relf{E}}=\mathbin{\overline{\relf{E}}}$ on $\cD\cap{\sf Weak}_\la(\rbox{L}).$ Therefore by
Proposition~\ref{solMb}, $x\in(\rbox{L}[\vpi])\sq[x]_{\mathbin{\relf{E}}}$
for some $\in\cont\al\cap\rbox{L}[\vpi],$ $\al<\la$
The model $\rbox{L}_{\om_1}[\vpi]$ has an elementary submodel
$\rbox{L}_\ga[\psi],$ where $\ga<\om_1$ and $\psi=\vpi\res\ga,$
containing $$ and $\al.$ We prove that
$\ang{x,\psi,}\in T_{\mathbin{\relf{E}}}.$ Since conditions \ref{aa} and
\ref{xx} of Definition~\ref{psi} obviously hold
for $\rbox{L}_\ga[\psi],$ let us check requirements \ref{bb} and
\ref{cc}.\vspace{1mm}
{\em We check\/ \ref{bb}.}
Indeed otherwise there exist conditions $u,\,v\in\col\al$ such
that $\ang{u,v}$ forces ${{}({\hat{f}})\mathbin{\not{\hspace{-2pt}\relf{E}}} {}({\hat{g}})}$ in
$\rbox{L}_{\ga}[\psi]$ in the sense of $\col\al\!\times\!\col\al$ as the
notion of forcing. Then $\ang{u,v}$ also forces
${{}({\hat{f}})\mathbin{\not{\hspace{-2pt}\relf{E}}} {}({\hat{g}})}$ in $\rbox{L}_{\om_1}[\vpi]$.
Let us consider an
\dd{\col\al\!\times\!\col\al}generic over $\rbox{L}[\vpi]$ pair
$\ang{f,g}\in \al^\om\times\al^\om$ such that $u\subset f$ and
$v\subset g.$ Then both $y=(f)$ and $z=(g)$ belong to
$ (\rbox{L}[\vpi]),$ so ${y\mathbin{\relf{E}} z}$ because
$(\rbox{L}[\vpi])\sq [x]_{\mathbin{\relf{E}}}$.
On the other hand, ${y\mathbin{\relf{E}} z}$ is {\em false\/} in
$\rbox{L}_{\om_1}[\vpi,f,g],$ that is, in $\rbox{L}[\vpi,f,g],$ by the
forcing property of $\ang{u,v}.$ Therefore we have ${x\mathbin{\not{\hspace{-2pt}\relf{E}}} y}$
(in the ``virtual'' universe $V$) by Shoenfield,
contradiction.\vspace{1mm}
{\em We check\/ \ref{cc}.} Take any \dd{\col\al}generic
over $\rbox{L}[\vpi]$ function $f\in\al^\om.$ Then $y=(f)$ belongs
to $(\rbox{L}[\vpi]),$ hence ${y\mathbin{\relf{E}} x}.$ On the other hand, $f$ is
generic over $\rbox{L}_{\ga}[\psi]$.\vspace{1mm}
Thus $\ang{x,\psi,}\in T_{\mathbin{\relf{E}}}.$ This means that
$x$ is \dd\mathbin{\relf{E}} classifiable, as required.\qed
\bdf
\label{U}
Let $x\in\cD.$ It follows from Lemma~\ref{key2} that there
exists the least ordinal $\ga=\ga_x<\om_1$ such that
$T_{\mathbin{\relf{E}}}(x,\vpi_x\res\ga,)$ for some $.$ We put
$\psi_x=\vpi_x\res\ga$ and let $_x$ denote the least, in the
sense of the $\rbox{OD}[\psi_x]$ wellordering of $\rbox{L}_{\ga}[\psi_x],$
``term'' $\in\cont{}[\psi_x]\cap \rbox{L}_{\ga}[\psi_x]$ which
satisfies $T_{\mathbin{\relf{E}}}(x,\psi_x,).$ We put
$U(x)=\ang{\psi_x,_x}$.\qed
\edf
\ble
\label{inv}
If each\/ $x\in\cD$ is\/ \dd\mathbin{\relf{E}} classifiable then the map\/ $U$
is a\/ $\idh1$ reduction of\/ $\mathbin{\relf{E}}$ to equality.
\ele
\noi{\bft Proof\hspace{2mm} } First of all, $U$ is $\idh1$ by Lemma~\ref{def}.
If $x\mathbin{\relf{E}} y$ then $U(x)=U(y)$ because Definition~\ref{psi}
is \dd\mathbin{\relf{E}} invariant for $x$.
Let us prove the converse. Assume that $U(x)=U(y),$ that is, in
particular, $\psi_x=\psi_y=\psi\in 2^{<\om}$ and
${_x=_y=\in\cont\al[\psi]\cap\rbox{L}_{\ga}[\psi],}$
where $\al<\ga=\rbox{dom}\,\psi<\om_1$.
By \ref{cc} we have ${(f)\mathbin{\relf{E}} x}$ and ${(g)\mathbin{\relf{E}} y}$ for
some \dd{\col\al}generic over $\rbox{L}_{\ga}[\psi]$ functions
$f,\,g\in\al^\om.$ However $(f)\mathbin{\relf{E}} (g)$ (see the proof
of Lemma~\ref{def}).\qed
\begin{corollary}\TF\
\label{case1}
{\rm[\hspace{1pt}The classification theorem\hspace{1pt}]}\\[1mm]
In the assumption of Case 1 of Subsection~\ref{cases}, $\mathbin{\relf{E}}$
admits a\/ $\idh1$ reduction to the equality on $2^{<\om_1}$.
\end{corollary}
\noi{\bft Proof\hspace{2mm} } The range of the function $U$ can be covered by a subset
$R\sq \rbox{HC}$ (all pairs $\ang{\psi,}$ such that ...) which admits
a $1-1$ $\idh1$ correspondence with $2^{<\om_1}$.\qed\vspace{4mm}
This completes the proof of the ``case 1'' part of Theorem~\ref{mt}.
\newpage
\subsection{$\protect\rbox{OD}$ forcing}
\label{prod}
This section starts the proof of the ``Case 2'' part of
Theorem~\ref{mt}. At the beginning, we reduce the problem to a
more elementary form.
\subsubsection{Explanation}
\label{expl}
Thus let us suppose that each real $x$ belongs to a ``virtual''
generic extension of $\rbox{L},$ but the assumption of Case 1 in
Subsection~\ref{cases} fails.
This means the following. There exists a real $z\in\cD$ such that
for every ordinal $\la$
and a ``virtual'' \dd{\col\la}generic extension $V$ of the
constructible universe $\rbox{L}$ containing $z,$ the following is true
in $V:$ if $z$ is \dd\la weak over $\rbox{L}$ then $\mathbin{\relf{E}}$
{\it does not\/} coincide with $\mathbin{\overline{\relf{E}}}$ on $\cD\cap{\sf Weak}_\la(\rbox{L})$.
We know indeed that
$z$ belongs to a
``virtual'' generic extension of $\rbox{L}.$ Therefore there
exists a limit constructible cardinal $\la$ such that $z$ belongs
to a \dd{\col\la}generic extension $V$ of $\rbox{L}$ and $z$ is weak
in $V.$ (Simply take $\la$ sufficiently large.)
Let us fix $\la$ and $V$.
As the condensed matter of this reasoning, we obtain
\begin{itemize}
\item\hspace{-1\mathsurround}
$V$ is a ``virtual'' \dd{\col\la}generic extension of $\rbox{L},$
$\la$ is a limit cardinal in $\rbox{L},$ and $\mathbin{\relf{E}}\mathbin{\hspace{-0.8mm\mathbin{\overline{\relf{E}}}$
on $\cD\cap{\sf Weak}_\la(\rbox{L})$ in $V$.
\end{itemize}
This is the description of the starting position of the proof of
the ``Case 2'' part of Theorem~\ref{mt}. The aim is to see that in
this case $\mathbin{\relf{E}}$ continuously embeds $\mathbin{\relf{E}_0}$ in the universe of
Theorem~\ref{mt}.
The general plan will be first to prove that $\mathbin{\relf{E}}$ continuously
embeds $\mathbin{\relf{E}_0}$ {\it in the auxiliary ``virtual'' universe\/} $V,$
and second, to get the result in the universe of Theorem~\ref{mt}
by Shoenfield.
After a short examination, one can see a problem in this plan:
the existence of a continuous embedding $\mathbin{\relf{E}_0}$ into $\mathbin{\relf{E}}$ is in fact
a $\is13$ statement:
$$
\exists\,\hbox{ continuous }1-1\,\;U:\cD\,\lra\,\cD\;
\forall\,x,\,y\in\cD\;
\left[
\begin{array}{cccl}
x\mathbin{\relf{E}_0} y & \lra & U(x)\mathbin{\relf{E}} U(y), & \hbox{ and }\\[2mm]
x\mathbin{\not{\hspace{-2pt}\relf{E}_0}} y & \lra & U(x) \mathbin{\not{\hspace{-2pt}\relf{E}}} U(y) &
\eay
\right]
$$
The lower implication in the square brackets is $\ip11,$ which
would match the total $\is12,$ but the upper one is $\is11,$ so
that the total result is $\is13,$ worse than one needs for
Shoenfield.
\subsubsection{Special embeddings and proof of the ``Case 2''
part of Theorem~\protect\ref{mt}}
\label{ne}
To overcome this obstacle, we strengthen the upper implication
to convert it to a $\ip11$ (actually $\id11$) statement. We recall
that the $\is11$ set $\mathbin{\relf{E}}\sq\cD^2$ admits a partition
$\mathbin{\relf{E}}=\bigcup_{\al<\om_1}\mathbin{\relf{E}}_{\al}$ onto Borel sets $\mathbin{\relf{E}}_\al$ -- the
{\it constituents\/}, uniquely defined as soon as we have fixed a
$\ip01$ set $F\sq\cD^2\times{\skri N}$ which projects onto $\mathbin{\relf{E}}.$
\bdf
\label{nice}
A $1-1$ function $\phi:\cD\,\lra\,\cD$ is a {\it special
embedding\/} of $\mathbin{\relf{E}_0}$ into $\mathbin{\relf{E}}$ iff\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{\rm({\arabic{enumi}})}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{cp}
there exists an ordinal $\al<\om_1$ such that
$\ang{\phi(0^k\we 0\we z),\phi(0^k\we 1\we z)}\in\mathbin{\relf{E}}_\al$\\ for all
$z\in\cD$ and $k\in\om$, \hfill and \its
\item for all $x,\,y\in\cD,$ if $x\mathbin{\not{\hspace{-2pt}\relf{E}_0}} y$ then $\phi(x)\mathbin{\not{\hspace{-2pt}\relf{E}}}\phi(y)$.
\qed
\end{enumerate}
\edf
($0^k$ is the sequence of $k$ zeros.)
First of all, let us see that a special embedding is an embedding in
the usual sense. We have to prove that $x\mathbin{\relf{E}_0} y$ implies
$\phi(x)\mathbin{\relf{E}} \phi(y).$ We say that a pair of points $x,\,y\in\cD$
is a {\it neighbouring pair\/} iff there exist $k\in\om$ and
$z\in\cD$ such that $x=0^k\we 0\we z$ and $y=1^k\we 1\we z$ or vice
versa. Obviously a neighbouring pair is \dd\mathbin{\relf{E}_0} equivalent. Conversely,
if $x\mathbin{\relf{E}_0} y$ then $x$ and $y$ can be connected by a finite chain
of neighbouring pairs in $\cD.$ Therefore condition~\ref{cp} actually
suffices to guarantee that $x\mathbin{\relf{E}_0} y\,\lra\,\phi(x)\mathbin{\relf{E}}\phi(y)$.
Obviously the existence of a {\it special\/} embedding of $\mathbin{\relf{E}_0}$
into
$\mathbin{\relf{E}}$ is a $\is12$ property. Thus, by Shoenfield, to complete the
proof of the ``Case 2'' part of Theorem~\ref{mt}, it suffices to
prove the following theorem (and apply it in the auxiliary
``virtual'' universe $V$).
\bte
\label{mtv}
Assume\/ \cuh\la.
Let\/ $\mathbin{\relf{E}}$ be a\/ $\is11$ relation and\/ $\mathbin{\relf{E}}\mathbin{\hspace{-0.8mm\mathbin{\overline{\relf{E}}}$ on\/
$\cD\cap{\sf Weak}_\la(\rbox{L}).$ Then\/ $\mathbin{\relf{E}_0}$ admits a special continuous
embedding into\/ $\mathbin{\relf{E}}$.
\ete
This theorem is being proved in this and the next section. During
the course of the proof, we assume \cuh\la{} and fix a $\is11$
equivalence $\mathbin{\relf{E}}$ satisfying $\mathbin{\relf{E}}\mathbin{\hspace{-0.8mm\mathbin{\overline{\relf{E}}}$ on the set
$\cD\cap{\sf Weak}_\la(\rbox{L})$
(although the last assumption will not be used at the beginning).
In this section, we consider important interactions between $\mathbin{\relf{E}}$
and $\mathbin{\overline{\relf{E}}}.$ The next section defines the required embedding. This
will complete the proof of theorems~\ref{mtv} and \ref{mt}, and
Theorem~\ref{main} -- the main theorem.
\subsubsection{$\rbox{OD}$ topology and the forcing}
\label{tforc}
We recall that ${\cal T}$ be the topology generated by all $\rbox{OD}$
sets.
A set $X$ will be called \dd{\cal T}{\it separable\/} if the $\rbox{OD}$
power set ${{\skri P}^{\hbox{\tiny\rm OD}}(X)={\skri P}(X)\cap\rbox{OD}}$ has only
countably many different $\rbox{OD}$ subsets.
\ble
\label{dizl}
Assume \cuh\la. Let\/ $\al<\la$ and\/ $\in\cont\al\cap\rbox{L}.$ Each
set\/ $X=(\rbox{L})$ satisfying\/ $X\sq\cD\cap{\sf Weak}_\la(\rbox{L})$
is\/ \dd{\cal T} separable.
\ele
\noi{\bft Proof\hspace{2mm} } By Proposition~\ref{solMb} every $\rbox{OD}$ subset of $X$ is
uniquely determined by an $\rbox{OD}$ subset of $\col\al.$ Since each
$\rbox{OD}$ set $S\sq\col\al$ is constructible, we obtain an $\rbox{OD}$ map
$h:\al^+\,\hbox{ onto }\,{\skri P}^{\hbox{\tiny\rm OD}}(X),$ where $\al^+$ is the least
cardinal in $\rbox{L}$ bigger than $\al.$ Therefore ${\skri P}^{\hbox{\tiny\rm OD}}(X)$ has
\dd{\<\al^{++}}many $\rbox{OD}$ subsets. It remains to notice that
$\al^{++}<\la$ because $\la$ is a limit cardinal in $\rbox{L},$ but
$\la$ is countable in the universe.\qed\vspace{4mm}
Let $\mathord{{\rm X}\hspace{-7pt}{\rm X}}=\ans{X\sq\cD:X\,\hbox{ is }\,\rbox{OD}\;\hbox{ and nonempty}\,}$.
Let us consider $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ as a forcing notion (smaller sets are
stronger conditions) for generic extensions of $\rbox{L}.$ Of course
formally $\mathord{{\rm X}\hspace{-7pt}{\rm X}}\not\in\rbox{L},$ but $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ is $\rbox{OD}$ order isomorphic to
a partially ordered set in $\rbox{L}$. (Indeed it is known that there
exists an $\rbox{OD}$ map $\ell:$ ordinals onto the class of all $\rbox{OD}$
sets. Since $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ itself is $\rbox{OD},$ $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ is a 1--1 image of an
$\rbox{OD}$ set $\mathord{{\rm X}\hspace{-7pt}{\rm X}}'$ of ordinals via $\ell.$ By Proposition~\ref{col}
both $\mathord{{\rm X}\hspace{-7pt}{\rm X}}'$ and the \dd{\ell\hspace{0.5pt}}preimage of the order
on $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ belong to $\rbox{L}$.)
\pagebreak[3]
It also is true that a set $G\sq\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ is \dd\mathord{{\rm X}\hspace{-7pt}{\rm X}} generic over $\rbox{L}$
iff it nonempty intersects every dense $\rbox{OD}$ subset of $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$.
\begin{corollary}\TF\
\label{exis}
Assume \cuh\la. If\/ a set ${X\in\mathord{{\rm X}\hspace{-7pt}{\rm X}}}$ satisfies\/
$X\sq\cD\cap{\sf Weak}_\la(\rbox{L})$ then there exists a\/ \dd\mathord{{\rm X}\hspace{-7pt}{\rm X}} generic
over\/ $\rbox{L}$ set\/ $G\sq\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ containing $X$.
\end{corollary}
\noi{\bft Proof\hspace{2mm} } We can suppose, by Proposition~\ref{solMb}, that
$X=(\rbox{L})$ where $\in\cont\al\cap\rbox{L}$ and $\al<\la.$ Now
apply Lemma~\ref{dizl}.\qed
\ble
\label{choq-cor}
Assume \cuh\la. If\/ $G\sq\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ is a generic over\/ $\rbox{L}$ set
containing the set\/ $\cD\cap{\sf Weak}_\la(\rbox{L})$ then the intersection\/
$\bigcap G$ is a singleton $\ans{a}=\ans{a_G}$.
\ele
\noi{\bft Proof\hspace{2mm} } Assume that this is not the case. Let $\mathord{{\rm X}\hspace{-7pt}{\rm X}}'\in\rbox{L}$ be a
constructible p. o. set order isomorphic $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ via an $\rbox{OD}$
function $\ell:\mathord{{\rm X}\hspace{-7pt}{\rm X}}'\,\hbox{ onto }\,\mathord{{\rm X}\hspace{-7pt}{\rm X}}.$ Then $G'=\ell^{-1}(G)$ is
\dd{\mathord{{\rm X}\hspace{-7pt}{\rm X}}'}generic over $\rbox{L}.$ We assert that the statement that
$\bigcap G$ is not a singleton can be converted to a sentence
relativized to $\rbox{L}[G']$.
(Indeed, it follows from the reasoning in the proof of
Lemma~\ref{dizl} that $\rbox{L}[G']$ is in fact a \dd Pgeneric
extension of $\rbox{L}$ for a certain set $P\in\rbox{L},$ $P\sq\mathord{{\rm X}\hspace{-7pt}{\rm X}}'$ of
a cardinality $\al<\la$ in $\rbox{L}.$ The next \dd\rbox{L} cardinal
$\al^+$ is $<\la$ since $\la$ is a limit cardinal in $\rbox{L}.$
Therefore $G'$ belongs to a \dd{\col{\al^+}}generic extension of
$\rbox{L},$ so $G'$ is weak. Then by Proposition \ref{col} the universe
$\rbox{V}=\rbox{L}[f_0]$ is a \dd{\col\la}generic extension of $\rbox{L}[G'].$
This is enough to convert any statement about $G'$ in $\rbox{V}$ --
like the statement: $\bigcap \ell{\hbox{\hspace{2pt}\rmt ''}} G'$ is not a singleton --
to a sentence relativized to $\rbox{L}[G']$.)
Then there exists ${X\in \mathord{{\rm X}\hspace{-7pt}{\rm X}}},$ ${X\sq\cD\cap{\sf Weak}_\la(\rbox{L})},$ such
that $\bigcap G$ is not a singleton for {\em every\/} generic over
$\rbox{L}$ set $G\sq \mathord{{\rm X}\hspace{-7pt}{\rm X}}$ containing $X.$ We can assume that
$X=(\rbox{L}),$ where ${\in\cont\al\cap\rbox{L}},$ $\al<\la.$ Then $X$
is \dd{\cal T} separable; let $\ans{{\skri X}_n:n\in\om}$ be an enumeration
of all $\rbox{OD}$ dense subsets of ${\skri P}^{\hbox{\tiny\rm OD}}(X).$ Using
Proposition~\ref{col} (item~\ref{sm6}), we obtain an increasing
\dd{\col\al}generic over $\rbox{L}$ sequence
$u_0\sq u_1\sq u_2\sq...$ of $u_n\in\col\al$ such that
$X_n={\hspace{0.1em}{u_n}}(\rbox{L})\in{\skri X}_n.$ Obviously this
gives an \dd\mathord{{\rm X}\hspace{-7pt}{\rm X}} generic over $\rbox{L}$ set $G\sq\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ containing $X$
and all $X_n$.
Now let $f=\bigcup_{n\in\om}u_n;$ $f\in\al^\om$ and $f$ is
\dd{\col\al}generic over $\rbox{L}.$ Then $x=(f)\in X_n$ for all
$n,$ so $x\in\bigcap G.$ Since $\bigcap G$ obviously cannot
contain more than one point, it is a singleton, so we get a
contradiction with the choice of $X$.\qed\vspace{4mm}
Reals $a_G$ will be called \dd\rbox{OD}{\it generic over\/}
$\rbox{L}$
\subsubsection{The product forcing}
We recall that $\mathbin{\relf{E}}$ is assumed to be a $\is11$ equivalence on
$\cD;$ $\mathbin{\overline{\relf{E}}}$ is the closure of $\mathbin{\relf{E}}$ in
the topology ${\cal T}^2$ (the product of two copies of ${\cal T}$).
For a set ${P\sq\cD^2,}$ we put
${{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P=\ans{x:\exists\,y\;P(x,y)}}$ and
${{{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P=\ans{y:\exists\,x\;P(x,y)}.}$ Notice that if $P$ is
$\rbox{OD},$ so are ${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P$ and ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P$.
The classical reasoning in Harrington, Kechris, and
Louveau~\cite{hkl} plays on interactions between $\mathbin{\relf{E}}$ and $\mathbin{\overline{\relf{E}}}.$
In the forcing setting, we have to fix a restriction by $\mathbin{\overline{\relf{E}}}$
directly in the definition of the product forcing. Thus we
consider
$$
\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}=\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}(\mathbin{\overline{\relf{E}}})=\ans{P\sq\mathbin{\overline{\relf{E}}}:P\,\hbox{ is }\rbox{OD}\;\hbox{ and
nonempty and }\,P=({{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P\times {{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P)\cap\mathbin{\overline{\relf{E}}}}
$$
as a forcing notion. As above for $\mathord{{\rm X}\hspace{-7pt}{\rm X}},$ the fact that formally
$\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ does not belong to $\rbox{L}$ does not cause essential problems.
The following assertion connects $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ and $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$.
\vbox{
\bass
\label{proe}
Assume \cuh\la. Then\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{\rm{\arabic{enumi}}}
\item If\/ $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ then\/ ${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P$ and\/ ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P$ belong to
$\mathord{{\rm X}\hspace{-7pt}{\rm X}}$.\its
\item If\/ $X,\,Y\in\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ and\/ $P=(X\times Y)\cap \mathbin{\overline{\relf{E}}}\not=\emptyset$
then $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$.\its
\itla{i3}
If\/ $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}},$ $X\in\mathord{{\rm X}\hspace{-7pt}{\rm X}},$ $X\sq{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P,$ then
there exists\/ $Q\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}},$ $Q\sq P,$ such that $X={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} Q.$
Similarly for ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}}$.
\end{enumerate}
\end{assertion}
}\its\its
\noi{\bft Proof\hspace{2mm} } Set $Q=\ans{\ang{x,y}\in P:x\in X\cj y\mathbin{\overline{\relf{E}}} x}$ in
item~\ref{i3}.\qed\vspace{4mm}
A set $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ is {\em \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} separable} if
the set $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}_{\sq P}=\ans{Q\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}:Q\sq P}$ has only countably many
different $\rbox{OD}$ subsets.
\ble
\label{dizl2}
Assume \cuh\la. Let\/ $,\,'\in\cont{<\la}\cap\rbox{L}.$ Suppose
that the sets\/ $X=(\rbox{L})$ and\/ $Y={'}(\rbox{L})$ satisfy\/
$X\cup Y\sq\cD\cap{\sf Weak}_\la(\rbox{L}),$ and finally that\/
${P=(X\times Y)\cap\mathbin{\overline{\relf{E}}}}$
is nonempty. Then\/ $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ and\/ $P$ is\/ \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} separable.
\ele
\noi{\bft Proof\hspace{2mm} } $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ by Assertion~\ref{proe}. A proof of the
\dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} separability can be obtained by a minor modification of
the proof of Lemma~\ref{dizl}.\qed
\ble
\label{dp2oe}
Assume \cuh\la. Let\/ $G\sq \mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ be a\/ \dd{\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}}generic over\/
$\rbox{L}$ set containing\/ ${(\cD\cap{\sf Weak}_\la(\rbox{L}))^2\cap\mathbin{\overline{\relf{E}}}}.$
Then the intersection\/ $\bigcap G$ contains a single point\/
$\ang{a,b}$ where\/ $a$ and\/ $b$ are\/ \dd\rbox{OD} generic
over $\rbox{L}$ and $a\mathbin{\overline{\relf{E}}} b$.
\ele
\noi{\bft Proof\hspace{2mm} } By Assertion~\ref{proe}, both $G_1=\ans{{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P:P\in G}$ and
$G_2=\ans{{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P:P\in G}$ are \dd\rbox{OD} generic over $\rbox{L}$ subsets of
$\mathord{{\rm X}\hspace{-7pt}{\rm X}},$ so that there exist unique \dd\rbox{OD} generic over $\rbox{L}$ points
$a=a_{G_1}$ and $b=a_{G_2}.$ It remains to show that
$\ang{a,b}\in\mathbin{\overline{\relf{E}}}$.
Suppose not. There exists an \dd\mathbin{\relf{E}} invariant $\rbox{OD}$ set
$A$ such that we have $x\in A$ and $y\in B=\cD\setminus A.$ Then
$A\in G_1$ and $B\in G_2$ by the genericity. There exists a
condition $P\in G$ such that ${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P\sq A$ and ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} B\sq B,$
therefore ${P\sq (A\times B)\cap\mathbin{\overline{\relf{E}}}=\emptyset},$ which is
impossible.\qed\vspace{4mm}
Pairs $\ang{a,b}$ as in Lemma~\ref{dp2oe} will be called
\dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}{\it generic\/} and denoted by $\ang{a_G,b_G}$.
For sets $X$ and $Y$ and a binary relation ${}\,,$ let us write
${X{}Y}$ if and only if
$\forall\,x\in X\;\exists\,y\in Y\;(x{} y)$ \ and \
$\forall\,y\in Y\;\exists\,x\in X\;(x{} y)$.
\ble
\label{1for2}
Assume \cuh\la. Let\/ $P_0\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}},$ $P_0\sq(\cD\cap{\sf Weak}_\la(\rbox{L}))^2,$
points\/ $a,\,a'\in X_0={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P_0$ be\/
\dd\rbox{OD} generic over\/ $\rbox{L},$ and\/ ${a\mathbin{\overline{\relf{E}}} a'.}$ There exists a
point\/ $b$ such that both\/ $\ang{a,b}$ and $\ang{a',b}$ belong
to\/ $P_0$ and are\/ \dd{\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}}generic pairs.
\ele
\noi{\bft Proof\hspace{2mm} } By Lemma \ref{dizl2} and Proposition~\ref{solMb} there
exists a \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} separable set $P_1\sq P_0$ such that
$a\in X_1={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P_1.$ We put $Y_1={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P_1;$ then $X_1\mathbin{\overline{\relf{E}}} Y_1,$
and $P_1=(X_1\times Y_1)\cap\mathbin{\overline{\relf{E}}}$.
We let $P'=\ans{\ang{x,y}\in P_0:y\in Y_1}.$ Then $P'\in \mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$
and $P_1\sq P'\sq P_0.$ Furthermore $a'\in X'={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P'.$ (Indeed,
since ${a\in X_1}$ and ${X_1\mathbin{\overline{\relf{E}}} Y_1},$ there exists $y\in Y_1$
such that $a\mathbin{\overline{\relf{E}}} y;$ then $a'\mathbin{\overline{\relf{E}}} y$ as well because $a\mathbin{\overline{\relf{E}}} a',$
therefore $\ang{a',y}\in P'$.) By Lemma \ref{dizl2} and
Proposition~\ref{solMb}
there exists a \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} separable set $P'_1\sq P'$ such that
$a'\in X'_1={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P'_1.$ Then $Y'_1={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P'_1\sq Y_1$.
It follows from the choice of $P$ and $P'$ that $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ admits only
countably many different dense $\rbox{OD}$ sets below $P_1$ and below
$P'_1.$ Let $\ans{{\skri P}_n:n\in\om}$ and $\ans{{\skri P}'_n:n\in\om}$
be\pagebreak[3] enumerations of both families of dense
sets. We define sets $P_n,\,P'_n\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}\;\;(n\in\om)$
satisfying the following conditions:\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{(\roman{enumi})}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{i}
$a\in X_n={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P_n$ \ and \ $a'\in X'_n={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P'_n$;\its
\itla{ii}
$Y'_n={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P'_n \sq Y_n={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P_n$ \ and \ $Y_{n+1}\sq Y'_n$;\its
\itla{iii}
$P_{n+1}\sq P_n\,,\,$ $P'_{n+1}\sq P'_n\,,\,$
$P_n\in {\skri P}_{n-2}\,,\,$ and \ $P'_n\in {\skri P}'_{n-2}$.\its
\end{enumerate}
By \ref{iii} both sequences $\ans{P_n:n\in\om}$ and
$\ans{P'_n:n\in\om}$ are \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} generic over $\rbox{L},$ so by
Lemma~\ref{dp2oe} they result in two generic pairs,
$\ang{a,b}\in P_0$ and $\ang{a',b}\in P_0, $ having the first
terms equal to $a$ and $a'$ by \ref{i} and second terms equal to
each other by \ref{ii}. Thus
it suffices to conduct the construction of $P_n$ and $P'_n$.
The construction goes on by induction on $n$.
Assume that $P_n$ and $P'_n$ have been defined. We define
$P_{n+1}.$ By~\ref{ii} and Assertion~\ref{proe}, the set
${P=(X_n\times Y'_n)\cap\mathbin{\overline{\relf{E}}}\sq P_n}$ belongs to $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ and
$a\in X={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P.$ (Indeed, $\ang{a,b}\in P,$ where $b$ satisfies
$\ang{a',b}\in P'_n,$ because ${a\mathbin{\overline{\relf{E}}} a'}$.) However ${\skri P}_{n-1}$
is dense in $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ below $P\sq P_0;$ therefore
${{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} {\skri P}_{n-1}=\ans{{{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P':P'\in {\skri P}_{n-1}}}$ is dense in $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$
below\pagebreak[3]
$X={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P.$ Since $a$ is generic, we have $a\in {{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P'$ for
some $P'\in {\skri P}_{n-1},$ $P'\sq P.$ It\pagebreak[3]
remains to put $P_{n+1}=P',$
and then $X_{n+1}={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P_{n+1}$ and $Y_{n+1}={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P_{n+1}$.
\pagebreak[3]
After this, to define $P'_{n+1}$ we let
$P=(X'_n\times Y_{n+1})\cap\mathbin{\overline{\relf{E}}},$ etc.\qed
\subsubsection{The key set}
\label{second}
We recall that \cuh\la{} is assumed, $\mathbin{\relf{E}}$ is a $\is11$
equivalence on $\cD,$ and $\mathbin{\overline{\relf{E}}}$ is the \hbox{\dd{{\cal T}^2}closure}
of $\mathbin{\relf{E}}$ in $\cD^2.$ By the assumption of Theorem~\ref{mtv},
$\mathbin{\relf{E}}\mathbin{\hspace{-0.8mm\mathbin{\overline{\relf{E}}}$ on $\cD\cap{\sf Weak}_\la(\rbox{L}).$ This means that there
exist \dd\mathbin{\overline{\relf{E}}} classes of elements of $\cD\cap{\sf Weak}_\la(\rbox{L})$ which
include more than one \dd\mathbin{\relf{E}} class. We define
the union of all those \dd\mathbin{\overline{\relf{E}}} classes,
$$
H=\ans{x\in\cD\cap{\sf Weak}_\la(\rbox{L}):\exists\,y\in
\cD\cap{\sf Weak}_\la(\rbox{L})\;(x\mathbin{\overline{\relf{E}}} y\cj x\mathbin{\not{\hspace{-2pt}\relf{E}}} y)}\,.
$$
Obviously $H$ is $\rbox{OD},$ nonempty, and \dd\mathbin{\relf{E}} invariant {\em inside\/}
$\cD\cap{\sf Weak}_\la(\rbox{L}),$ and moreover $H'=H^2\cap\mathbin{\overline{\relf{E}}}\not=\emptyset,$
so that in particular $H'\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ by Assertion~\ref{proe}.
\ble
\label{noE}
Assume \cuh\la. If\/ $a,b\in H$ and\/ $\ang{a,b}$ is\/
\dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} generic over $\rbox{L}$ then ${a\mathbin{\not{\hspace{-2pt}\relf{E}}} b}\hspace{1.5pt}.$
\ele
\noi{\bft Proof\hspace{2mm} } Otherwise there exists a set $P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}},$ $P\sq H\times H$ such
that $a\mathbin{\relf{E}} b$ holds for {\it all\/} \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} generic $\ang{a,b}\in P.$
We conclude that then $a\mathbin{\overline{\relf{E}}} a'\;\lra\;a\mathbin{\relf{E}} a'$ for all \dd\rbox{OD} generic
points $a,\,a'\in X={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P;$ indeed, take $b$ such that both
$\ang{a,b}\in P$ and $\ang{a',b}\in P$ are \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} generic,
by Lemma~\ref{1for2}. In other words the relations $\mathbin{\relf{E}}$ and
$\mathbin{\overline{\relf{E}}}$ coincide on the set
${Y=\ans{x\in X:x\,\hbox{ is \dd\rbox{OD} generic over }\,\rbox{L}}\in\mathord{{\rm X}\hspace{-7pt}{\rm X}}.}$
($Y$ is nonempty by corollaries \ref{exis} and \ref{choq-cor}.)
Moreover, $\mathbin{\relf{E}}$ and $\mathbin{\overline{\relf{E}}}$ coincide on the set
${Z=[Y]_{\mathbin{\relf{E}}}\cap\cD\cap{\sf Weak}_\la(\rbox{L}).}$ Indeed if $z,\,z'\in Z,$
${z\mathbin{\overline{\relf{E}}} z'},$
then let ${y,\,y'\in Y}$ satisfy ${z\mathbin{\relf{E}} y}$ and ${z'\mathbin{\relf{E}} y'}.$
Then ${y\mathbin{\overline{\relf{E}}} y'},$ therefore ${y\mathbin{\relf{E}} y'},$ which implies $z\mathbin{\relf{E}} z'.$
We conclude that $Y\cap H=\emptyset$.
(Indeed, suppose that $x\in Y\cap H.$ Then by definition there
exists $y\in\cD\cap{\sf Weak}_\la(\rbox{L})$
such that ${x\mathbin{\overline{\relf{E}}} y}$ but ${x\mathbin{\not{\hspace{-2pt}\relf{E}}} y}.$ Then ${y\not\in Z}$ because
$\mathbin{\relf{E}}$ and $\mathbin{\overline{\relf{E}}}$ coincide on $Z.$ Thus the pair $\ang{x,y}$ belongs
to the $\rbox{OD}$ set $P=Y\times [(\cD\cap{\sf Weak}_\la(\rbox{L}))\setminus Z].$
Notice that $P$
does not intersect $\mathbin{\relf{E}}$ by definition of $Z.$ Therefore
$\ang{x,y}$ cannot belong to the closure $\mathbin{\overline{\relf{E}}}$ of $\mathbin{\relf{E}},$
contradiction.)
But $\emptyset\not=Y\sq X\sq H,$ contradiction.\qed\vspace{4mm}
Lemma~\ref{noE} is a counterpart of the proposition in
Harrington, Kechris, Louveau~\cite{hkl} that $\mathbin{\relf{E}}\res H$ is
meager in $\mathbin{\overline{\relf{E}}}\res H.$ But in fact the main content of this
argument in~\cite{hkl} was implicitly taken by Lemma~\ref{1for2}.
\ble
\label{E}
Assume \cuh\la. Let\/ $X,\,Y\sq H$ be nonempty\/ $\rbox{OD}$ sets
and\/ ${X\mathbin{\overline{\relf{E}}} Y}.$ There
exist nonempty\/ $\rbox{OD}$ sets\/ $X'\sq X$ and\/ $Y'\sq Y$ such
that\/ $X'\cap Y'=\emptyset$ but still\/ $X'\mathbin{\overline{\relf{E}}} Y'$.
\ele
\noi{\bft Proof\hspace{2mm} } There exist points $x_0\in X$ and
$y_0\in Y$ such that $x_0\not= y_0$ but ${x_0\mathbin{\overline{\relf{E}}} y_0}.$
(Otherwise $X=Y,$ and $\mathbin{\overline{\relf{E}}}$ is the equality on $X,$ which is
impossible, see the previous proof.) Let $U$ and $V$ be disjoint
Baire intervals in $\cD$ containing resp. $x_0$ and $y_0.$
The sets $X'= X\cap U \cap [Y\cap V]_{\mathbin{\overline{\relf{E}}}}$ and
$Y'= Y\cap V \cap [X\cap U]_{\mathbin{\overline{\relf{E}}}}$ are as required.\qed
\newpage
\subsection{Embedding $\protect\mathbin{\relf{E}_0}$ into $\protect\mathbin{\relf{E}}$}
\label{or}
In this section we end the proof of Theorem~\ref{mtv}. Thus
we prove, assuming \cuh\la{} and $\mathbin{\relf{E}}\mathbin{\hspace{-0.8mm\mathbin{\overline{\relf{E}}}$ on
$\cD\cap{\sf Weak}_\la(\rbox{L}),$ that $\mathbin{\relf{E}}$ embeds $\mathbin{\relf{E}_0}$ via a
continuous special (see Definition~\ref{nice}) embedding.
\subsubsection{The embedding}
\label{embed}
By the assumption the set $H$ of Subsection~\ref{second} is
nonempty; obviously $H$ is $\rbox{OD}.$ By lemmas \ref{dizl},
\ref{dizl2}, and Proposition~\ref{solMb} there exists a nonempty
\dd{\cal T} separable $\rbox{OD}$ set $X_0\sq H$ such that the set
${P_0=(X_0\times X_0)\cap\mathbin{\overline{\relf{E}}}}$ belongs to $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ and is \dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}
separable. We observe that
${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P_0={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P_0=X_0\sq H\sq\cD\cap{\sf Weak}_\la(\rbox{L})$.
We define a family of sets $X_u\;\;(u\in 2^{<\om})$ satisfying\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{(\alph{enumi})}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{a}
$X_u\sq X_0,$ $X_u$ is nonempty and $\rbox{OD},$ and $X_{u\we i}\sq X_u,$
for all $u$ and $i$.\its
\end{enumerate}
In addition to the sets $X_u,$ we shall define relations
${uv}\sq\cD^2$ for {\em some} pairs $\ang{u,v},$ to provide
important interconnections between branches in $2^{<\om}$.
Let $u,\,v\in 2^n.$ We say that $\ang{u,v}$ is a {\em neighbouring
pair\/} iff $u=0^k\we 0\we r$ and $v=0^k\we 1\we r$ for some $k<n$
($0^k$ is the sequence of $k$ terms each equal to $0$) and some
$r\in 2^{n-k-1}$ (possibly $k=n-1,$ that is, $r=\La$).
Thus we define sets ${uv}\sq X_u\times X_v$ for all neighbouring pairs
$\ang{u,v},$ so that the following requirements \ref{b} and
\ref{d} will be satisfied.\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{(\alph{enumi})}
\def\theenumi{{\rmt\arabic{enumi}.}}
\setcounter{enumi}{1}
\itla{b} \hspace{-1\mathsurround}
${uv}$ is $\rbox{OD},$ ${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} {uv}=X_u,$ ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} {uv}=X_v,$ and
${u\we i\,,\,v\we i}\sq {uv}$ for every neighbouring pair
$\ang{u,v}$ and each $i\in\ans{0,1}$.\its
\itla{d}
For any $k,$ the set $ k={0^k\we 0\,,\,0^k\we 1}$ is
\dd{\cal T} separable, and $ k\sq \mathbin{\relf{E}}_\al$ for some ordinal
$\al=\al(k)<\om_1$.\its
\end{enumerate}
Notice that if $\ang{u,v}$ is neighbouring then
$\ang{u\we i,v\we i}$ is neighbouring, but $\ang{u\we i,v\we j}$
is not neighbouring for $i\not=j$ (unless $u=v=0^k$ for some $k$).
It follows that $X_u {uv} X_v,$ therefore
$X_u\mathbin{\relf{E}} X_v,$ for all neighbouring pairs $u,\,v.$~\footnote
{\ We recall that $X{}Y$ means that
$\forall\,x\in X\;\exists\,y\in Y\;(x{} y)$ and
$\forall\,y\in Y\;\exists\,x\in X\;(x{} y)$.}
\begin{remark}\rmt\
\label{newrem}
Every pair of $u,\,v\in 2^n$ can be tied in $2^n$ by a
finite chain of neighbouring pairs. It follows that
${X_u\mathbin{\relf{E}} X_v}$ and ${X_u\mathbin{\overline{\relf{E}}} X_v}$ hold for {\em all} pairs
$u,\,v\in 2^n$.\qed
\erem
Three more requirements will concern genericity.
Let $\ans{{\skri X}_n:n\in\om}$ be a fixed (not necessarily $\rbox{OD}$)
enumeration of all dense in $\mathord{{\rm X}\hspace{-7pt}{\rm X}}$ below $X_0$ subsets of $\mathord{{\rm X}\hspace{-7pt}{\rm X}}.$
Let $\ans{{\skri P}_n:n\in\om}$ be a fixed enumeration of all dense in
$\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ below $P_0$ subsets of $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}.$ It is assumed that
${\skri X}_{n+1}\sq{\skri X}_n$ and ${\skri P}_{n+1}\sq{\skri P}_n.$ Note that
${{\skri X}'=\ans{P\in\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}: P\sq P_0\cj {{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P\cap{{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P=\emptyset}}$
is dense in $\mathord{{\rm I}\hspace{-2.5pt}{\rm P}}$ below $P_0$ by Lemma~\ref{E}, so we can suppose
in addition that ${\skri P}_0={\skri X}'$.
In general, for any \dd{\cal T} separable set $S$ let
$\ans{{\skri X}_n(S):n\in\om}$ be a fixed enumeration of all dense
subsets in the algebra ${\skri P}^{\hbox{\tiny\rm OD}}(S)\setminus\ans{\emptyset}$ such
that ${\skri X}_{n+1}(S)\sq{\skri X}_n(S)$.
We now formulate the three additional requirements.
\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{({\rm g}\arabic{enumi})}
\def\theenumi{{\rmt\arabic{enumi}.}}
\itla{g1} \hspace{-1\mathsurround}
$X_u\in {\skri X}_n$ whenever $u\in 2^n$.\its
\itla{g2}
If $u,\,v\in 2^n$ and $u(n\hspace{-1pt}-\hspace{-1pt} 1)\not=v(n\hspace{-1pt}-\hspace{-1pt} 1)$ (that is, the
last terms of $u,\,v$ are different), then
$P_{uv}=(X_u\times X_v)\cap\mathbin{\overline{\relf{E}}}\in {\skri P}_n$.\its
\itla{g3}
If $\ang{u,v}=\ang{0^k\we 0\we r,0^k\we 1\we r}\in (2^n)^2$
then ${uv}\in {\skri X}_n( k)$.\its
\end{enumerate}
In particular \ref{g1} implies by Corollary~\ref{choq-cor} that
for any $a\in 2^\om$ the intersection
$\bigcap_{n\in\om}X_{a\res n}$ contains a single point, denoted
by $\phi(a),$ which is \dd\rbox{OD} generic over $\rbox{L},$ and the map
$\phi$ is continuous in the Polish sense.
\bass
\label{embe}
Assume\/ \cuh\la.
$\phi$ is a special continuous 1--1 embedding\/ $\mathbin{\relf{E}_0}$ to $\mathbin{\relf{E}}$.
\end{assertion}
\noi{\bft Proof\hspace{2mm} }
Let us prove that $\phi$ is 1--1. Suppose that
${a\not=b\in 2^\om.}$ Then ${a(n\hspace{-1pt}-\hspace{-1pt} 1)\not=b(n\hspace{-1pt}-\hspace{-1pt} 1)}$ for
some $n.$ Let ${u=a\res n},$ ${v=b\res n},$
so that we have $x=\phi(a)\in X_u$ and $y=\phi(b)\in X_v.$ But
then the set ${P=(X_u\times X_v)\cap \mathbin{\overline{\relf{E}}}}$ belongs to ${\skri P}_n$ by
\ref{g2}, therefore to ${\skri P}_0.$ This implies
$X_u\cap X_v=\emptyset$ by definition of ${\skri P}_0,$
hence $\phi(a)\not=\phi(b)$ as required.
Furthermore if $a\mathbin{\not{\hspace{-2pt}\relf{E}_0}} b$ (which means that $a(k)\not=b(k)$ for
infinitely many numbers $k$) then $\ang{\phi(a),\phi(b)}$ is
\dd\mathord{{\rm I}\hspace{-2.5pt}{\rm P}} generic by \ref{g2}, so $\phi(a)\mathbin{\not{\hspace{-2pt}\relf{E}}} \phi(b)$ by
Lemma~\ref{noE}.
Let us finally verify that
$\ang{\phi(0^k\we 0\we c),\phi(0^k\we 1\we c)}\in\mathbin{\relf{E}}_\al$ for all
$c\in\cD$ and $k\in \om,$ where $\al=\sup_k\al(k)<\om_1.$
The sequence of sets
$W_m={0^k\we 0\we c\res m\,,\,0^k\we 1\we c\res m}\;\;\,(m\in\om)$
is then generic over $\rbox{L}$ by \ref{g3} in the sense of the forcing
${\skri P}^{\hbox{\tiny\rm OD}}( k)\setminus\ans{\emptyset}$ (we recall that
$ k={0^k\we 0\,,\,0^k\we 1}$), which is simply a copy of $\mathord{{\rm X}\hspace{-7pt}{\rm X}},$
so that by Corollary~\ref{choq-cor} the intersection of all sets
$W_m$ is a singleton. Obviously the singleton can be only equal to
$\ang{\phi(0^k\we 0\we c)\,,\,\phi(0^k\we 1\we c)}.$ We conclude
that $\phi(0^k\we 0\we c)\mathbin{\relf{E}}_\al \phi(0^k\we 1\we c),$ as
required.\qed
\subsubsection{Two preliminary lemmas}
Thus the theorem is reduced to the construction of sets $X_u$
and ${uv}.$ Before the construction starts, we prove a couple
of important lemmas.
\ble
\label{impo}
Assume\/ \cuh\la.
Let\/ $X,\,Y\sq\cD\cap{\sf Weak}_\la(\rbox{L})$ be\/ $\rbox{OD}$ sets such that\/
$(X\times Y)\cap\mathbin{\relf{E}}$ is nonempty. Then\/ $(X\times Y)\cap\mathbin{\relf{E}}$ contains a
weak over\/ $\rbox{L}$ point $\ang{x,y}$.
\ele
\noi{\bft Proof\hspace{2mm} }
First of all, by Proposition~\ref{solMb} we can
assume that $X={}(\rbox{L})$ and $Y={'}(\rbox{L}),$ where $$ and
$t'$ belong to some $\cont\al\cap\rbox{L},$ $\al<\la.$ Then, since
$\la$ is a limit \dd\rbox{L} cardinal, we have $X={}(\rbox{L}_\ba)$ and
$Y={'}(\rbox{L}_\ba)$ for a suitable $\ba,$ $\al\<\ba<\la.$ Take
an arbitrary \dd{\col\ba}generic over $\rbox{L}$ function
$f\in\ba^\om.$ Then the statement ${(X\times Y)\cap\mathbin{\relf{E}}\not=\emptyset}$ turns
out to be a $\is11$ formula with reals in $\rbox{L}[f]$ (those coding
$f,\;,\;'$) as parameters.
Notice that all sets in $\rbox{L}[f]$ are weak over $\rbox{L},$ so it
remains to apply Shoenfield.\qed
\ble
\label{comb}
Assume\/ \cuh\la.
Let\/ $n\in\om,$ and\/ $X_u$ be a nonempty\/ $\rbox{OD}$ set for each\/
$u\in 2^n.$ Assume that an\/ $\rbox{OD}$ set\/ ${uv}\sq {\skri N}^2$ is
given for every neighbouring pair of\/ $u,\,v\in 2^n$ so that
$X_u {uv} X_v$.\its
\begin{enumerate}
\def{\rmt\arabic{enumi}.}{{\rm\arabic{enumi}.}}
\def\theenumi{{\rmt\arabic{enumi}.}}
\item If\/ $u_0\in 2^n$ and\/ $X'\sq X_{u_0}$ is\/ $\rbox{OD}$ and
nonempty then there exists a system of\/ $\rbox{OD}$ nonempty sets\/
$Y_u\sq X_u\;\;(u\in 2^n)$ such that\/ $Y_u {uv} Y_v$ holds for
all neighbouring pairs\/ $u,\,v,$ and in addition $Y_{u_0}=X'$.\its
\item
Suppose that\/ $u_0,\,v_0\in 2^n$ is a neighbouring pair and
nonempty\/ $\hspace{-1pt}\rbox{OD}\hspace{-1pt}$
sets\/ ${X'\sq X_{u_0}}$ and $X''\sq X_{v_0}$ satisfy\/
$X' {u_0v_0} X''.$ Then there exists a system of\/ $\rbox{OD}$
nonempty sets\/ ${Y_u\sq X_u}$ $(u\in 2^n)$ such that\/
${Y_u {uv} Y_v}$ holds for all neighbouring pairs\/ $u,v,$ and
in addition\/ $Y_{u_0}=X',\,\;Y_{v_0}=X''.$
\end{enumerate}
\ele
\noi{\bft Proof\hspace{2mm} } Notice that 1 follows from 2. Indeed take arbitrary $v_0$
such that either $\ang{u_0,v_0}$ or $\ang{v_0,u_0}$ is neighbouring,
and put respectively
${X''=\ans{y\in X_{v_0}: \exists\,x\in X'\;(x {u_0v_0} y)}},$ or
${X''=\ans{y\in X_{v_0}: \exists\,x\in X'\;(y {v_0u_0} x)}}$.
To prove item 2, we use induction on $n.$
For $n=1$ --- then $u_0=\ang{0}$ and $v_0=\ang{1}$ ---
we take $Y_{u_0}=Y'$ and $Y_{v_0}=Y''$.
The step. We prove the lemma for $n+1$ provided it has been proved
for $n;\,\,n\>1.$ The principal idea is to divide $2^{n+1}$ on two
copies of $2^n,$ minimally connected by neighbouring pairs, and handle
them more or less separately using the induction hypothesis. The
two ``copies'' are $U_0=\ans{s\we 0:s\in 2^n}$ and
$U_1=\ans{s\we 1:s\in 2^n}$.
The only neighbouring pair that connects $U_0$ and $U_1$ is the pair
of ${\hat u}=0^n\we 0$ and ${\hat v}=0^n\we 1.$ If in fact $u_0={\hat u}$
and $v_0={\hat v}$ then we apply the induction hypothesis (item~1)
independently for the families ${\ans{X_u:u\in U_0}}$ and
${\ans{X_u:u\in U_1}}$ and the given sets ${X'\sq X_{u_0}}$ and
${X''\sq X_{v_0}.}$ Assembling the results, we get nonempty $\rbox{OD}$
sets ${Y_u\sq X_u\,\;(u\in 2^{n+1})}$ such that ${Y_u {uv} Y_v}$
for all neighbouring pairs\/ $u,\,v,$ perhaps with the exception of the
pair of $u=u_0={\hat u}$ and $v=v_0={\hat v},$ and in addition
$Y_{u_0}=X'$ and $Y_{v_0}=X''.$
Thus finally $Y_{\hat u} {{\hat u}{\hat v}}Y_{\hat v}$ by the choice of $X'$ and
$Y'$.
It remains to consider the case when both $u_0$ and $v_0$ belong
to one and the same domain, say to $U_0.$ Then we first apply the
induction hypothesis (item 2) to the family
${\ans{X_u:u\in U_0}}$ and the sets ${X'\sq X_{u_0}}$ and
${X''\sq X_{v_0}.}$ This results in a system of nonempty $\rbox{OD}$
sets ${Y_u\sq X_u\;\,(u\in U_0);}$ in particular we get an $\rbox{OD}$
nonempty set $Y_{\hat u}\sq X_{\hat u}.$ We put
${Y_{\hat v}=\ans{y\in X_{\hat v}:\exists\,x\in Y_{\hat u}\,
(x{{\hat u}{\hat v}}y)},}$ so that ${Y_{\hat u}{{\hat u}{\hat v}}Y_{\hat v},}$
and apply the induction hypothesis (item 1) to the family
${\ans{X_u:u\in U_1}}$ and the set $Y_{\hat v}\sq X_{\hat v}$.\qed
\subsubsection{The construction}
We put $X_\La=X_0.$
Now assume that the sets $X_s\,\;(s\in 2^n)$
and relations ${st}$ for all neighbouring pairs of $s,\,t\in 2^{\<n}$
have been defined, and expand the construction at level $n+1.$
We first put $A_{s\we i}=X_s$ for all $s\in 2^n$ and
$i\in\ans{0,1}.$ We also define ${uv}={st}$ for any neighbouring
pair of $u=s\we i,\,\,v=t\we i$ in $2^{n+1}$ other than the pair
${\hat u}=0^n\we 0,\,\,{\hat v}=0^n\we 1.$ For the latter one (notice
that $A_{{\hat u}}=A_{{\hat v}}=X_{0^n}$) we put ${{\hat u}{\hat v}}=\mathbin{\overline{\relf{E}}},$
so that $A_u{uv} A_v$ holds for all neighbouring pairs of
$u,\,v\in 2^{n+1}$ including the pair $\ang{{\hat u},{\hat v}}$.
The sets $A_u$ and relations ${uv}$ will be reduced in several
steps to meet requirements \ref{a}, \ref{b}, \ref{d} and
\ref{g1}, \ref{g2}, \ref{g3} of Subsection~\ref{embed}.\vom
{\em Part 1}. After $2^{n+1}$ steps of the procedure of
Lemma~\ref{comb} (item 1) we obtain a system of nonempty $\rbox{OD}$
sets $B_u\sq A_u\;\,(u\in 2^{n+1})$ such that still
$B_u{uv} B_v$ for all neighbouring pairs $u,\,v$ in $2^{n+1},$ but
$B_u\in {\skri X}_{n+1}$ for all $u.$ Thus \ref{g1} is fixed.\vom
{\em Part 2}. To fix \ref{g2}, consider an arbitrary pair of
$u_0=s_0\we 0,$ $v_0=t_0\we 1,$ where $s_0,\,t_0\in 2^n.$ By
Remark~\ref{newrem} and density of the set ${\skri P}_{n+1}$ there
exist nonempty $\rbox{OD}$ sets $B'\sq B_{u_0}$ and $B''\sq B_{v_0}$
such that ${P=(B'\times B'')\cap\mathbin{\overline{\relf{E}}}\in {\skri P}_{n+1}}$ and
${{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} P=B',$ ${{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} P=B'',$ so in particular ${B'\mathbin{\overline{\relf{E}}} B''}.$ Now we
apply Lemma~\ref{comb} (item 1) separately for the two systems
of sets,
${\ans{B_{s\we 0}:s\in 2^n}}$ and ${\ans{B_{t\we 1}:t\in 2^n}}$
(compare with the proof of Lemma~\ref{comb}~!), and the sets
$B'\sq B_{s_0\we 0},$ $B''\sq B_{t_0\we 1}$ respectively.
This results in a system of nonempty
$\rbox{OD}$ sets ${B'_u\sq B_u}$ ${(u\in 2^{n+1})}$ satisfying
${B'_{u_0}=B'}$ and ${B'_{v_0}=B'',}$ so that we have
${(B'_{u_0}\times B'_{v_0})\cap\mathbin{\overline{\relf{E}}}\in {\skri P}_{n+1},}$ and still
$B'_u{uv} B'_v$ for all neighbouring pairs $u,\,v\in 2^{n+1},$
perhaps with the exception of the pair of
${\hat u}=0^n\we 0,\,\,{\hat v}=0^n\we 1,$ which is the only one that
connects the two domains. To handle this exceptional pair,
note that ${B'_{{\hat u}} \mathbin{\overline{\relf{E}}} B'_{u_0}}$ and ${B'_{{\hat v}} \mathbin{\overline{\relf{E}}} B'_{v_0}}$
(Remark~\ref{newrem} is applied to each of the two domains),
so that ${B'_{\hat u}\mathbin{\overline{\relf{E}}} B'_{\hat v}}$ since ${B'\mathbin{\overline{\relf{E}}} B''}.$ Finally
we observe that ${{\hat u}{\hat v}}$ is so far equal to $\mathbin{\overline{\relf{E}}}$.
After $2^{n+1}$ steps (the number of pairs $u_0,\,v_0$ to be
considered) we get a system of nonempty $\rbox{OD}$ sets
$C_u\sq B_u\;\,(u\in 2^{n+1})$ such that
$(C_u\times C_v)\cap\mathbin{\overline{\relf{E}}}\in {\skri P}_{n+1}$ whenever $u(n)\not=v(n),$
and still $C_u{uv} C_v$ for all neighbouring pairs
$u,\,v\in 2^{n+1}.$ Thus \ref{g2} is fixed.\vom
{\em Part 3}. We fix \ref{d} for the exceptional neighbouring
pair of ${{\hat u}=0^n\we 0},$ ${{\hat v}=0^n\we 1}.$ Since $\mathbin{\relf{E}}$ is
\dd{{\cal T}^2}dense in $\mathbin{\overline{\relf{E}}},$ and ${C_{\hat u}\mathbin{\overline{\relf{E}}} C_{\hat v},}$ the set
${{}=(C_{\hat u} \times C_{\hat v})\cap\mathbin{\relf{E}}}$ is nonempty. We observe that
the $\rbox{OD}$ set
$$
{{}}'=\ans{\ang{x,y}\in {{}}:\ang{x,y}\,\hbox{ is weak over }
\,\rbox{L}}
$$
is nonempty, too, by Lemma~\ref{impo}. Then, since
${{}}'\sq{}\sq\mathbin{\relf{E}},$ the intersection
${{}}''={}'\cap\mathbin{\relf{E}}_\al$ is nonempty for some $\al<\om_1.$
($\mathbin{\relf{E}}_\al$ is the \dd\al th constituent of the \dd{\is11}set
$\mathbin{\relf{E}}$.) Finally some nonempty $\rbox{OD}$
set ${}\sq {{}}''$ is \dd{\cal T} separable by Lemma~\ref{dizl}.
Consider the $\rbox{OD}$ sets $C'={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}} {}\,\,(\sq C_{\hat u})$ and
$C''={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}} {}\,\,(\sq C_{\hat v});$ obviously $C'{} C'',$ so that
$C'{{\hat u}{\hat v}} C''.$ (We recall that at the moment
${{\hat u}{\hat v}}=\mathbin{\overline{\relf{E}}}.$) Using Lemma~\ref{comb} (item 2) again, we
obtain a system of nonempty $\rbox{OD}$ sets
$Y_u\sq C_u\;\,(u\in 2^{n+1})$ such that still $Y_u{uv} Y_v$
for all neighbouring pairs $u,\,v$ in $2^{n+1},$ and $Y_{\hat u}=C',$
$Y_{\hat v}=C''.$ We re--define ${{\hat u}{\hat v}}$ by ${{\hat u}{\hat v}}={}$
(then ${{\hat u}{\hat v}}\sq \mathbin{\relf{E}}_\al$),
but this keeps $Y_{\hat u}{{\hat u}{\hat v}} Y_{\hat v}$.\vom
{\em Part 4}. We fix \ref{g3}. Consider a neighbouring pair
$u_0,\,v_0$ in $2^{n+1}.$
Then we have $u_0=0^k\we 0\we r,$ $v_0=0^k\we 1\we r$ for
some ${k\<n}$ and ${r\in 2^{n-k}}.$ It follows that
${\Ip{}={u_0v_0}\cap(Y_{u_0}\times Y_{v_0})}$ is a nonempty
(since ${Y_{u_0}{u_0v_0} Y_{v_0}}$) $\rbox{OD}$ subset of
$ k={0^k\we 0\,,\,0^k\we 1}$ by the construction. Let
${}\sq \Ip{}$ be a nonempty $\rbox{OD}$ set in ${\skri X}_{n+1}( k).$ We
now define $Y'={{\bbb\rbox{pr}\bbb}_1\hspace{1.5pt}}{}$ and $Y''={{\bbb\rbox{pr}\bbb}_2\hspace{1.5pt}}{}$ (then ${Y'{}Y''}$
and ${Y'{u_0v_0}Y''}$) and run Lemma~\ref{comb} (item 2) for the
system of sets $Y_u\;\,(u\in2^{n+1})$ and the sets
${Y'\sq Y_{u_0}},$ ${Y''\sq Y_{v_0}}$. After this define
the ``new'' ${u_0v_0}$ by ${u_0v_0}={}$.
Do this consequtively for all neighbouring pairs; the finally
obtained sets -- let them be $X_u\,\;(u\in 2^{n+1})$ -- are as
required. The final relations ${uv}\;\,(u,\,v\in 2^{n+1})$ can
be obtained as the restrictions of sets ${uv}$ to
$X_u\times X_v$.\vom
This ends the construction.
\vspace{4mm}
This also ends the proof of theorems \ref{mtv} and \ref{mt},
and Theorem~\ref{main} (the main theorem), see
Subsection~\ref{ne}.\qed
\newpage
|
\section{INTRODUCTION}
Turbulence and magnetic fields are both topics of morbid curiousity
in astrophysics. In that context they are usually seen as poorly
understood, undoubtedly real phenomena that can be used as part of
an explanation of last resort, i.e. when all calculable models have
been disproven. Consequently both these phenomena, together and
separately, have been used
in constructing models of angular momentum transport in accretion
disks, another process of indisputable reality whose nature is
obscure. Notwithstanding this troubled history, I will present in
this paper a summary of my recent work on magnetic fields in turbulent
media and explore its implications for accretion disks.
I have three basic reasons for pressing ahead with such an unpromising
topic. First, the structure of magnetic fields in accretion disks
determines the rate at which magnetic flux is lost from the disk.
This implies that one can get a variety of rates depending on one's
model for the magnetic field structure, but it also implies that any
physically well motivated model can have interesting, and possibly
unique, implications. Second, magnetic field instabilities represent
a mechanism guaranteed to move angular momentum outward and matter
inward (Balbus \& Hawley 1991), which is not true for many of the instabilities
suggested as the basis for dissipation in accretion disks. It is
therefore critical to explore the nature of the turbulence resulting from
these instabilities. Third, vertical magnetic fields, entrained in accretion
disks, are widely believed to be responsible for driving violent outflows,
especially jets, from a wide variety of accretion disks. The radial
transport of vertical magnetic flux can't be understood without examining
the nature of magnetic fields in turbulent disks, and the global structure
of accretion disks fields hinges on this issue.
Before I begin I need to summarize the relevant features of accretion
disks. The single most important point is that they are luminous
due to the conversion of orbital energy into heat. This implies an
outward flux of angular momentum. Such a flux would follow from the
existence of some local viscosity, but it would have to be much larger
than the viscosity implied by microscopic processes. The usual solution
is to invoke an effective viscosity due to collective processes which
is $\nu\equiv\alpha h c_s$, where $h$ is the disk height, $c_s$ is the
sound speed, and $\alpha$ is an arbitrary constant of order unity
(Shakura and Sunyaev 1973). For a thin disk with no self-gravity, i.e. a
`Keplerian
disk', we have a rotational frequency $\Omega(r)\propto r^{-3/2}$,
a disk height $h\sim c_s/\Omega$, and, by definition, $h\ll r$.
In this case the differential rotation in the disk plus the assumed
effective viscosity leads to an inward flux of mass given by
\begin{equation}
\dot M\sim\alpha \Sigma h^2\Omega,
\end{equation}
and a radiative flux from the disk surface of
\begin{equation}
{}F_{radiative}\sim \dot M\Omega^2.
\end{equation}
Attempts to model the outbursts for dwarf novae and X-ray transients have
led to the conclusion that $\alpha$ is probably not a constant, but
a function of local conditions (Cannizzo 1994 and references therein).
Assuming that $\alpha$
goes as $(h/r)^n$, where $n$ is a constant of order unity, gives a
reasonable fit to the observations.
In \S II of this paper I will summarize my recent work on the distribution
of magnetic fields in a turbulent medium. In \S III I will discuss the
implications of this work for magnetic buoyancy in disks and how this
leads to a direction connection between dynamo growth rates in disks and
the appropriate value of $\alpha$. In \S IV I will draw some general
conclusions and point the way towards future progress on this topic.
\section{MAGNETIC FIELDS IN TURBULENT FLUIDS}
The material in this section is a synopsis of Vishniac (1995a).
The basic feature of this model is that the magnetic field in
a high $\beta$ fluid, i.e. one in which the magnetic field pressure
is small compared to other sources of pressure, is spatially intermittent.
Most of the magnetic flux is contained in flux tubes, whose radii are
much smaller than the scale of curvature for the field. This is
not a novel suggestion. In fact, it is about what one would guess from
examining the magnetic field in the photosphere of the Sun. It does
raise the question of how such flux tubes form. Why should the
magnetic field and the gas spontaneously separate from one another?
The mechanism I have proposed is a process I call turbulent pumping.
If we consider an isolated flux tube in a turbulent medium, then
as long as the flux tube is flexible enough to respond to the hydrodynamic
forces exerted by the surrounding fluid then it will undergo stretching
at a rate roughly equal to the shearing rate on the scale of curvature
of the flux tube. This lowers the linear density of matter in the
flux tube at the same rate. By the time the flux tube length has
doubled it will be twisted by the surrounding flow in such a way that
it will intersect itself, or a whole set of neighboring flux tubes.
This will result in formation of a set of closed loops which will shrink
down to dissipative scales and vanish, thereby maintain a constant
flux tube length in the turbulent fluid. In this way matter is removed
from the magnetic flux tubes at a rate which is dependent only on
the properties of the turbulent medium and not at all on the specific
resistivity of the fluid. In a stationary state this loss of matter
from the flux tubes is balanced by ohmic diffusion of the charged
particles onto the field lines, a process which becomes extremely
slow as the conductivity of the fluid increases to astronomical values.
The consequence is the appearance of flux tubes whose internal gas density
can be far below that of the surrounding fluid. In the Sun
flux tubes will be largely evacuated only near the top of the solar
convection zone, whereas in accretion disks flux tubes will be almost
empty whenever the disks are largely ionized.
This process of turbulent pumping depends on several conditions. First,
there can be little or no turbulent diffusion of matter into the
flux tubes. Otherwise the mismatch between the mass loss driven
by collective processes (flux tube stretching and the creation of
closed loops) and mass loading driven by ohmic diffusion will disappear.
Preliminary work indicates that this condition will be satisfied
whenever the Alfv\'en speed in the flux tubes is significantly
greater than the turbulent velocity outside. In other words, this
condition is satisfied self-consistently when turbulent pumping is
effective. The transition from a diffuse field to one contained in
flux tubes is not yet understood. Second, reconnection must be
efficient, in the sense of allowing the magnetic field to rearrange
its topology in less than an eddy turn over time. Once again, this
condition is met self-consistently in the flux tube model. This
result assumes the Sweet-Parker rate for reconnection, which is generally
considered to the slowest reasonable estimate for reconnection rates.
Third, turbulent pumping relies on the notion that closed loops whose
radii are less than a typical eddy size will tend to shrink to dissipative
scales quickly, thereby unloading their entrained mass into the surrounding
plasma. This also follows from the physics implicit in the flux tube
description of the magnetic field, although the loops tend to shred
as they shrink, thereby creating a more diffuse, but weaker and less
organized, component to the magnetic field.
What does this pumping lead to? If the resistivity is large enough that
particles can diffuse to the center of a flux tube in less than one
eddy turnover time, then the flux tube radius is
\begin{equation}
r_t\approx \left({\eta\over kV_l}\right)^{1/2},
\end{equation}
where $\eta$ is the resistivity, $V_l$ is the turbulent velocity on
the scale $l$ on which the flux tubes are bent, and $k$ is the corresponding
wavenumber, $k=2\pi/l$. In this limit the magnetic pressure in a flux
tube has a gaussian profile. When the resistivity is sufficiently small
the flux tubes will become largely empty, i.e. the magnetic field inside
the flux tube, $B_t$, is given by
\begin{equation}
{B_t^2\over 8\pi}=P,
\end{equation}
where $P$ is the pressure of the surrounding fluid. In this limit each
flux tube will have a skin depth of width $(\eta/kV_l)^{1/2}$ surrounding
a hollow core.
None of this tells us to how decide what $r_t$ or $l$ should be in a given
situation. For this we need to understand the forces between flux tubes, since
these forces will determine the distribution of flux elements in the turbulent
fluid. Since in this picture the magnetic flux is confined to the interiors
of the flux tubes the forces between them will be hydrodynamic and stem from
the turbulent wakes created as the fluid moves past the semi-rigid flux tubes.
The most obvious effect (cf. Parker 1979 \S8.9) is the attraction between flux
tubes when one lies
downstream from the other. This is simply due to the fact that the downstream
flux tube feels a reduced turbulent drag since the momentum flux around it
is reduced by $\rho V_l^2 r_t/w(r)$, where $\rho$ is the fluid density, $V_l$
is
the fluid velocity relative to the flux tubes, $r_t$ is the typical flux tube
radius, and $w(r)$ is the width of the turbulent wake at a distance $r$
downstream
from the leading flux tube. This attraction is analogous to mock gravity, in
that
it stems from the ability of neighboring tubes to block statistically isotropic
repulsive forces in the environment. It is less well known that one expects
neighboring flux tubes whose separation is more or less perpendicular to the
ambient flow to repel one another. This is known experimentally (Zdravkovich
1977,
Gu, Sun, He, \& Zhang 1993)
since no adequate analytical treatment of the near-field turbulent flow is
available. Unlike the shielding effect this repulsion depends critically on
the
nature of the flux tube wakes. When the wakes are purely laminar and stable
there is an attractive force, which is the basis for previous claims that
flux tubes embedded in a turbulent flow always attract one another (Parker 1979
\S8.9).
The transition to an unstable wake, capable of producing repulsion, occurs when
\begin{equation}
{V_l r_t\over \nu}> \sim \pi^3.
\end{equation}
In other words, when the Reynolds number {\it on the scale of the flux tube
radius}
exceeds a critical value which lies in the range of 30 to 40. When this
criterion
is not satisfied flux tubes will aggregate. At higher Reynolds numbers the
magnetic
field will be broadly distributed through the fluid in the form of discrete
flux tubes
that maintain their separate identities through a rough equilibrium between
attractive
and repulsive interactions. This qualitative difference is significant, since
this
criterion is almost always satisfied in astrophysical objects, and not (yet)
satisfied in
numerical simulations.
Neglecting viscosity, which is reasonable in stars and accretion disks, I have
used
these points to construct a simple model of the magnetic field structure in a
turbulent conducting fluid. A detailed discussion is given in Vishniac
(1995a).
Here I simply quote the relevant results. First, if the magnetic field energy
density is
comparable to, or less than, the turbulent energy density, there exists some
scale
$l$ such that
\begin{equation}
{B_t^2\over l}\sim \rho {V_l^2\over r_t}.
\label{eq:tense}
\end{equation}
The left hand side of this equation is the force per volume in the flux tube
exerted by magnetic tension. The right hand side is the force per volume
exerted
by turbulent drag from the ambient fluid. Their rough equality defines the
scale
$l$ as the scale of curvature for a typical flux tube. In typical turbulent
cascade $V_l^2l$ is a sharply increasing function of $l$. Consequently, on
scales
larger than $l$ the magnetic field is almost passively advected. On smaller
scales the flux tubes are almost rigid.
Second, in order to balance the time-averaged attractive and repulsive forces
between flux tubes, it is necessary to suppose that on all scales less than $l$
the number of flux tubes within a radius $r$ of a given flux tube, $N_r$,
satisfies the condition
\begin{equation}
N_r r_t\sim r.
\end{equation}
This defines a fractal distribution of dimension one which extends from the
flux tube radius up to the scale $l$.
I can combine these results to get an expression for the average magnetic
field energy density. This average is well-defined only on scales larger than
$l$. On that scale I obtain
\begin{equation}
\langle B^2\rangle \sim N_l B_t^2 \left({r_t\over l}\right)^2,
\end{equation}
or
\begin{equation}
\langle B^2\rangle \sim B_t^2 {r_t\over l}.
\end{equation}
Combining this with equation (\ref{eq:tense}) we obtain
\begin{equation}
\langle B^2\rangle \sim \rho V_l^2.
\end{equation}
In other words, the scale of curvature for the flux tubes is the
scale of equipartition between the mean square magnetic field and the
average turbulent energy density.
\section{BUOYANCY AND DYNAMOS IN DISKS}
Given this specific model for the structure of a magnetic field in a turbulent
medium it is possible to discuss the systematic motion of a magnetic field
in an accretion disk. I begin by noting that the rate at which a flux
tube will rise due to buoyant forces is given by the balance between
turbulent drag and the buoyant acceleration. For a flux tube this gives
\begin{equation}
\Delta\rho (\pi r_t^2) g\sim \rho V_b V_l r_t,
\end{equation}
where $\Delta \rho$ is the density deficit inside the flux tube, $g$ is the
local
gravitational acceleration, and $V_b$ is the buoyant velocity. The left
hand side of this equation is the buoyant force per unit length. The right
hand side is the turbulent drag, assuming that $V_b\ll V_l$. In what
follows I will assume that the magnetic field is in equipartition with
the turbulence and write $V_T$ instead of $V_l$. Assuming that the flux
tube is small enough to be in good thermal contact with the surrounding
medium, which is usually reasonable, the fractional density deficit
$\Delta\rho/\rho$
is roughly the ratio of the magnetic pressure in the flux tube to the ambient
pressure. This implies that
\begin{equation}
\left({B_t^2\over 8\pi P}\right)\rho (\pi r_t^2) g\sim \rho V_b V_l r_t,
\end{equation}
or
\begin{equation}
V_b\sim\left({\rho g\over P}\right) L_T V_T\sim {L_T\over l_p} V_T,
\end{equation}
where $l_p$ is the local pressure scale height and I have used the condition
that the
radius of curvature of the
magnetic field lines is $L_T$. Note that if the magnetic field energy is
below equipartition then I need to replace $L_T V_T$ with the appropriate
$l V_l$, implying a slower buoyant rise. For stellar convective turbulence
$l_p\sim L_T$ and magnetic flux will rise at substantial fraction of the local
turbulent velocity once the magnetic field reaches equipartition with the
turbulence.
This result cannot be extended to accretion disks. There the most plausible
source of turbulence is magnetic field instability first described by
Velikhov (1959) (see also Chandrasekhar 1961) and applied to accretion disks by
Balbus and Hawley (1991). Here I invoke the description of the saturated
state of this instability for a large scale azimuthal field embedded in an
accretion disk given in Vishniac and Diamond (1992). The dominant eddies
will be those characterizing the fastest growing mode, for reasons explained
in that paper. The instability will saturate in a turbulent state
characterized
by a typical turbulent velocity comparable to the Alfv\'en speed, i.e.
$V_T\sim V_A$. The eddy size will be $L_T\sim V_A/\Omega$, where $\Omega$
is the local rotational frequency. This gives rise to an effective
viscosity and diffusion coefficient of order $L_TV_T\sim V_A^2/\Omega$
or $(V_A/c_s)^2 h^2\Omega$, where $h$ is the disk thickness and I have
used the relationship $c_s\sim h\Omega$, which applies to thin,
non-self-gravitating
accretion disks. The dimensionless viscosity of the disk, due to magnetic
field stresses, is just $(V_A/c_s)^2$.
What does this imply about flux tubes in accretion disks? It can be shown
(Vishniac 1995b) that flux tubes in hot disks are in the ideal fluid regime,
i.e. they are almost completely empty. In this case
\begin{equation}
r_t\sim L_T\left({V_T\over c_s}\right)^2\sim \left({V_A\over c_s}\right)^3
h\sim
\alpha^{3/2} h.
\end{equation}
The buoyant velocity is
\begin{equation}
V_b\sim {L_T\over l_p} V_T\sim {V_A/\Omega\over h} V_A\sim {V_A^2\over
c_s}\sim\alpha c_s.
\end{equation}
Since the loss rate for magnetic flux is just $V_b/h\sim\alpha \Omega$ this
implies that the
azimuthal magnetic field escapes from the disk at a rate which is comparable to
the
thermal relaxation rate for an optically thick disk. I note in passing that
this
flux loss rate is smaller, by a factor of $V_A/c_s$, than estimates based on
the Parker
instability, which is normally taken to imply a buoyant velocity $\sim V_A$.
The reason for this discrepancy is that the Parker instability is strongly
suppressed by the Balbus-Hawley instability (Vishniac and Diamond 1992).
In a stationary state the magnetic field must be regenerated by some dynamo
process
so that the dynamo growth rate balances the buoyant flux losses, i.e.
\begin{equation}
\Gamma_{dynamo}\sim {V_b\over h}\sim \alpha\Omega,
\end{equation}
or
\begin{equation}
\alpha\sim {\Gamma_{dynamo}\over \Omega}.
\end{equation}
When radiation pressure is large and electron scattering dominates the opacity,
a situation normally encountered in the inner regions of AGN disks, the flux
tube
properties change significantly. In this case the magnetic pressure in the
flux tubes is limited by the ambient gas pressure, since the ambient photons
can
diffuse into the flux tubes on a very short time scale. This does not affect
the efficiency of turbulent pumping, with the consequence that the flux tubes
are larger than one would expect in the ideal fluid regime, but are still
evacuated.
This gives a modified expression for the flux tube radius, i.e.
\begin{equation}
r_t\sim L_T\left({V_T\over c_{s,gas}}\right)^2\sim h\left({V_A\over
c_s}\right)^3
{P\over P_{gas}}\sim\alpha^{3/2} h{P\over P_{gas}}.
\end{equation}
This leads to an enhanced buoyancy so that
\begin{equation}
V_b\sim g\left({\rho V_T^2 L_T\over P_{gas}}\right) {1\over V_T}\sim{P\over
P_{gas}}
\alpha c_s.
\end{equation}
Consequently, for a given dynamo growth rate, balancing magnetic flux
generation
with buoyant losses I get
\begin{equation}
\alpha\sim\left({\Gamma_{dynamo}\over\Omega}\right)\left({P_{gas}\over
P}\right).
\end{equation}
In other words, as long as the dynamo growth rate is independent of the
magnetic energy
density the implied disk viscosity will scale with the gas pressure, rather
than the
total pressure. Of course, this may imply that other angular momentum
transport mechanisms,
normally dominated by magnetic stresses, become important.
What are some possible disk dynamos?
The equilibrium state of the azimuthal magnetic field is determined by the
balance between
dynamo activity and vertical buoyancy. However, one can also imagine that a
typical
accretion disk will have a large scale vertical magnetic field, if for no other
reason
than the fact that such a field is likely to be accreted along with matter
added to
the outer edge of the disk. Clearly vertical buoyancy is irrelevant to the
evolution
of this field. Moreover, since the field lines cross the disk in concentrated
flux
tubes, and spread out above and below the disk, the tension due to strong
bending
of the external field lines is negligible. Nevertheless, there are
two effects which will tend to move vertical field lines
outward. First, if the field lines are bent radial by some total angle
$2\theta$
as they cross the disk, then turbulent diffusion through the disk will tend to
combine radial field lines of opposite polarity, moving the point at which the
field lines cross the disk outward at a rate of roughly $\alpha c_s\tan\theta$
(Van Ballegooijen 1989). If the magnetic field curvature is determined only by
large
scale stresses than $\theta\sim h/r$ and this velocity is comparable to the
inward flow of matter in the disk. Consequently it will be difficult to
determine the direction of drift for the magnetic field. Of course, if the
concentration of magnetic field in the inner disk increases, then the global
stresses will increase and the disk will stop accreting vertical flux
regardless.
In addition, if the field is responsible for driving a wind or jet (cf. Shu et
al.
1994 and references therein) then it will tend to bend sharply near the
disk which will move the field outward.
Second, the flux tubes containing the vertical flux will be subjected to
radial buoyancy forces (Parker \& Vishniac 1995). Each flux tube will have an
associated
energy due to its displacement of matter and associated pressure. This energy
has
two (usually) comparable parts. The first is due to the surrounding pressure
and is
roughly equal to $\Delta P L \pi r_t^2$, where $L$ is the length of the flux
tube, and
$\Delta P$ pressure contributed by the magnetic field in the flux tube. The
other
is due to the displacement of matter which could otherwise settle to the
disk midplane. This term is of order $\Delta \rho (h\Omega)^2 L\pi r_t^2$,
where
$\Delta\rho$ is the density deficit in the flux tube and is of order $\rho$
under
normal circumstances in a hot disk. For a gas pressure dominated disk
the two are comparable and roughly equal to $B_t^2r_t^2 L$. For a radiation
pressure
dominated disk the gravitational term dominates is larger by a factor of
$P/P_{gas}$.
One would get the same effect by replacing the $\Delta P$ in the pressure
contribution
with $P$ rather than $P_{gas}$.
If the energy associated with a flux tube is $U_t$ then the consequent radial
drift
velocity is obtained by equating the turbulent drag with the radial gradient of
$U_t$, or
\begin{equation} \rho V_TV_b r_t \sim -{\partial_r U_t\over L}\sim -P
r_t^2\partial_r (\ln U_t).
\end{equation}
Consequently,
\begin{equation}
V_b \sim -\alpha c_s h {P\over P_{gas}}\partial_r (\ln U_t),
\end{equation}
which for $\partial_r(\ln U_t)\sim r^{-1}$ implies a radial drift velocity
which
is larger than the inward accretion velocity by a factor of $P/P_{gas}$. The
direction of the drift depends on the sign of $\partial_r(\ln U_t)$. Since
a single flux tube can break apart, or combine, in the course of its radial
drift,
we need to evaluate this derivative under the constraint that the magnetic flux
remains fixed, or that the area goes as $P_{gas}^{1/2}$. The length of the
flux tube will exceed $h$, since each tube actually crosses the disk in a
random
walk. We will assume that $L\propto h\alpha^{-1/2}$. Subject to these
constraints
we find that
\begin{equation}
V_b \sim -\alpha c_s h {P\over P_{gas}}{1\over2}\partial_r \left({P\dot M
c_s\over P_{gas}\alpha}
\right).
\end{equation}
If $\dot M$ is constant then this will almost give a strong outward buoyancy to
the flux
tubes, which will clearly dominate over accretion when $P\gg P_{gas}$. In the
event
that the inner disk is unstable and $\dot M$ varies with $r$ there will still
be a
averaged outward flow. The evidence for magnetic activity in accretion disks,
especially
in AGN, must be read as evidence for large scale dynamo activity in these
disks.
What are the prospects for a reasonable theory of dynamo activity in disks?
There
is an extensive literature on this topic, too extensive to summarize here. My
own
view is that there are only a few processes which we can be reasonably sure
exist and
which may dominate in real accretion disks. All of them rely on the notion
that
shearing of a radial field provides for efficient generation of a large scale
azimuthal
field. The tricky step is to understand how the azimuthal field component
regenerates
the radial field. One of the most popular notions is that magnetic buoyancy
produces
turbulent motions with a preferred helicity, which close the cycle by twisting
the azimuthal field lines into radial field lines (Galeev, Rosner, \&
Vaiana 1979). However, this relies on
using the Parker instability, which has a large azimuthal wavenumber. Since
modes
with a large $k_\theta$ and a slow growth rate will be suppressed by the
Balbus-Hawley
instability, this mode is not expected to exist in real accretion disks
(Vishniac
and Diamond 1992), nor is it seen in numerical simulations (Brandenburg et al.
1995).
Another idea is that internal waves, excited near the outer edge of the disk
via
tidal forcing (Goodman 1993), will fill the disk and drive a dynamo with a
growth
rate $\sim (H/r)^{3/2}$ (Vishniac, Jin, and Diamond 1990, Vishniac and Diamond
1992).
This implies a dimensionless viscosity of $\sim (H/r)^{3/2}(P_{gas}/P)$.
{}Finally,
Balbus and Hawley (1991) have claimed that the Balbus-Hawley instability will
lead to a local turbulent dynamo which will saturate near equipartition between
the field and ambient pressure, leading to an $\alpha$ of order unity.
Something
like this is seen in numerical simulations (e.g. Brandenburg et al. 1995),
although
it would appear to be inconsistent with phenomenological work on accretion
disks
(cf. Cannizzo 1994 and references therein). On
the other hand, the numerical results are also consistent with the idea that
the
Balbus-Hawley instability is driving an incoherent dynamo according to mean
field
theory, leading to a saturation which depends on the geometry of the
computational
box (Vishniac and Brandenburg 1995). This theory predicts an effective
$\alpha$
of order $(H/r)^2(P_{gas}/P)^6$. The steeper scaling with $H/r$ and strong
suppression
in radiation pressure dominated environments implies that while this process
may play a role in real disks, it will only dominate in disks not subject to
significant elliptical distortions at large radii and free of significant
radiation pressure.
\section{CONCLUSIONS AND FUTURE PROSPECTS}
I can summarize my results as follows. First, I have proposed a simple model
of flux tube
formation that is consistent with solar observations. Second, in this picture
there
is a failure of `flux-freezing' to adequately describe the macroscopic motions
of the
field and fluid. In general there will be some significant relative motion
between
the flux tubes and the fluid. This suggests the possibility of mean-field
dynamos.
Third, direct numerical simulation of astrophysically realistic driven MHD
turbulence
is not currently possible. Turbulence driven by magnetic field instabilities
can
be simulated to obtain qualitative results, but such simulations will be
quantitatively unreliable. Fourth, accretion disks might be able to move
magnetic field lines inward, but only if their radial bending angle across the
disk
is of order $h/r$ or less. Even in this case radial buoyancy may move them
outward. Fifth, magnetic viscosity in AGN disks couples primarily to gas
pressure,
not radiation pressure. This may imply that the viscosity due to hydrodynamic
effects, e.g. internal wave breaking, dominates. Finally, given the relatively
high rate of vertical and radial buoyant magnetic flux losses, evidence for
continued
magnetic activity is disks (and stars) can only be explained by the presence of
some dynamo mechanism.
It is somewhat disappointing that the MHD turbulence model used here implies
that
numerical simulations of astrophysical turbulence are not physically realistic.
{}Fortunately, this same theory does predict scaling behavior and approximate
saturation
values for magnetic fields in the viscous regime, which is currently accessible
to
numerical simulations. Somewhat fragmentary results from current work seem to
show the predicted behavior of the magnetic field as a function of Reynolds
number
(Vishniac 1995a). Future tests of the theory in this regime should give us
some
confidence in its application to astrophysics, or allow us to discard it in
favor
of some alternative description. However, until we are in position to do more
realistic MHD simulations it will not be possible to replace the approximate
results sketched here with quantitative estimates.
|
\section{Introduction}
Heavy quark production in high energy hadronic collisions constitutes a
benchmark process for the study of perturbative QCD and an important tool to
explore flavour physics.
\begin{itemize}
\item $b$ quarks are produced in abundance in hadronic
collisions, and will eventually allow ultimate tests of CP
violation in the $b$ system, as well as studies of rare $b$ decays with
branching ratio levels of the order of $10^{-10}$. A
detailed understanding of the
production properties at future machines (such as the LHC) is therefore of
the utmost importance.
\item The comparison of the current experimental data with the predictions of
QCD provides a necessary check that the ingredients entering the evaluation of
hadronic processes (partonic distribution functions and higher order
corrections) are under control and can be used to evaluate the rates for more
exotic phenomena or to extrapolate the calculations to even higher energies.
\item Accurate studies of the production properties of the $top$ quark
rely on a solid understanding of the QCD dynamics, in order to
isolate possible contributions from new phenomena.
\item Measurements of heavy quark production in fixed target experiments are
dominated by data at low \mbox{$p_{\sss \rm T}$}, a region where non-perturbative effects are not
negligible. The comparison of data with the expectations of perturbative QCD
offers the possibility to explore some interesting features of
non-perturbative hadronic dynamics \cite{fmnr}.
\item At HERA, $c$ and $b$ quarks are largely produced via photon-gluon fusion,
therefore providing a direct probe on the gluon density of the proton,
complementary to the information extracted from the measurement of structure
functions. First data have already become available, and have been shown at
this Conference.
\item Production of quarkonium states, in addition to providing yet another
interesting framework for the study of the boundary between perturbative and
non-perturbative QCD, is important in view of the use of exclusive
charmonium decays of $b$ hadrons for the detection of CP violation phenomena.
\end{itemize}
In this presentation we review the current status of theoretical
calculations, and discuss the implications of the most recent experimental
measurements of $b$ quarks and charmonium states
performed at the Tevatron $p\bar p$ collider.
\section{Open Flavour Production: Theory Overview}
To start with, we briefly report on the current status of the theoretical
calculations. A distinction must be made between calculations performed at a
complete but fixed order in perturbation theory (PT), and those performed by
resumming classes of potentially large logarithmic contributions that arise at
any order in PT. The exact matrix elements squared for heavy quark production
in hadronic collisions are fully known up to the ${\cal O}(\mbox{$\alpha_s$}^3)$, for both
real and virtual processes. These matrix elements have been used to evaluate at
the next-to-leading order (NLO) the total production cross section \cite{btot},
single-particle inclusive distributions \cite{bpt}\ and two-particle inclusive
distributions (a.k.a. correlations) \cite{mnr}.
Three classes of large logarithms can appear in the perturbative expansion for
heavy quark production:
\begin{enumerate}
\item $[\mbox{$\alpha_s$}\log(S/m_Q^2)]^n \sim [\mbox{$\alpha_s$}\log(1/x_{\rm Bj})]^n$ terms, where $S$ is
the hadronic CM energy squared. These small-$x$ effects are possibly relevant
for the production of charm or bottom quarks at the current energies, while
they should have no effect on the determination of the top quark cross section,
given the large $t$ mass. Several theoretical studies have been performed
\cite{smallx}, and the indications are that \mbox{$b$}\ production cross sections
should not increase by more than 30--50\% at the Tevatron energy because of
these effects.
\item $[\mbox{$\alpha_s$}\log(m_Q/p^T_{QQ})]^n$ terms, where $p^T_{QQ}$ is the transverse
momentum of the heavy quark pair. These contributions come from the multiple
emission of initial-state soft gluons, similarly to standard Drell Yan
corrections. These corrections have been studied in detail in the case of top
production, where the effect is potentially large due to the heavy top mass
\cite{laenen}. They are not relevant to the total cross
section of \mbox{$b$}\ quarks, but affect the kinematical distributions of pairs
produced just above threshold \cite{berger2}, or in regions at the edge of
phase space, such as $\Delta\phi=\pi$.
\item $[\mbox{$\alpha_s$}\log(p_T/m_Q)]^n$ terms, where $p_T$ is the transverse momentum of
the heavy quark. These terms arise from multiple collinear gluons emitted by
a heavy quark produced at large transverse momentum, or from the
branching of gluons into heavy quark pairs. Again these corrections are
not expected to affect the total production rates, but will contribute to the
large-\mbox{$p_{\sss \rm T}$}\ distributions of $c$ and $b$ quarks. No effect is expected for the
top at current energies. These logarithms can be resummed using a fragmentation
function formalism \cite{greco},
with a significant improvement in the stability w.r.t. scale changes
for $\mbox{$p_{\sss \rm T}$}>50$ GeV.
\end{enumerate}
\section{Single Inclusive Bottom Production}
\begin{figure}[t]
\centerline{\psfig{figure=ratio_dflm.ps,width=10cm,angle=90}}
\ccaption{}{Ratio of data and theory for the integrated $b$ \mbox{$p_{\sss \rm T}$}\ distribution
at
UA1, CDF and D0. Theory evaluated at NLO \cite{bpt}\ using DFLM parton
densities \cite{dflm}\ ($\lambdamsb=173$ MeV), $m_b=4.75$ GeV,
$\muf=\mur=\sqrt{m^2+p_T^2}$. \label{ratio-dflm}}
\end{figure}
The status of $b$ production at hadron colliders has been quite puzzling for
some time. Data collected by UA1~\cite{ua1b} at the CERN
S$p\bar p$S collider ($\sqrt S$ = 630
GeV) were in good agreement with theoretical expectations based on the NLO QCD
calculations \cite{bpt}.
On the contrary, the first measurements performed at 1.8
TeV by the CDF \cite{cdfpre93} experiment at the Fermilab collider showed a
significant discrepancy with the same calculation.
The new data presented at this Conference \cite{paulini}\
by the two Fermilab experiments, CDF and D0, allow us to draw a more complete
picture of the situation.
\begin{figure}[t]
\centerline{\psfig{figure=ratio_mrsa.ps,width=10cm,angle=90}}
\ccaption{}{Ratio of data and theory for the integrated $b$ \mbox{$p_{\sss \rm T}$}\ distribution
at
UA1, CDF and D0. Theory evaluated at NLO \cite{bpt}\ using MRSA parton
densities \cite{mrsa}, $\lambdamsb = 300$ MeV, $m_b = 4.5$ GeV,
\muf = \mur = $\sqrt{m^2+p_T^2}/2$. \label{ratio-mrsa}}
\end{figure}
We present all three sets of data from UA1, CDF and D0 in a single
plot, containing the ratio of the measurements to the theory, for a uniform
choice of parameters entering the theoretical calculation. In
fig.~\ref{ratio-dflm}, we choose the same theoretical prediction as was
available at
the time of the UA1 measurements, namely the central set of the DFLM
\cite{dflm}\ parton distributions ($\lambdamsb=173$ MeV), $m_b=4.75$ GeV and
renormalization/factorization scales equal to the transverse mass of the $b$
quark, $\muf=\mur=\sqrt{m^2+p_T^2}\equiv \muo $. Depending on whether we use
the D0 or the CDF data as representative of the $b$ cross section at 1.8 TeV,
we can draw two different conclusions. The plot shows clearly that the ratio
{\em data/theory} is the same, at UA1 and at D0,
to within less than 10\%.
While larger than 1, this ratio can be reduced by selecting different input
parameters, still in the range of acceptable values. For example,
fig.~\ref{ratio-mrsa}\ shows the same distributions with the theory
evaluated using the more recent set of parton densities MRSA \cite{mrsa},
$m_b=4.5$ GeV, $\muf=\mur=\muo/2$ and $\lambdamsb=300$ MeV, a value
close to the LEP measurement of \mbox{$\alpha_s$}. With this choice of parameters the
agreement between NLO QCD and data is perfect, both at 630 and at 1800 GeV. The
data from CDF are in good agreement with the theory shape, but are about
30--40\% higher in
normalization relative to the D0 ones. If one were to choose CDF data as
representative of the Tevatron rate, the conclusion would be that the $b$
production cross section grows between 630 and 1800 GeV by a factor of 40\%
more than NLO QCD predicts. An effect of such a size would be in agreement with
the evaluation of the small-$x$ effects mentioned earlier. The only conclusions
we can therefore draw from the current data are that:
\begin{enumerate}
\item the comparison of data and NLO predictions for the $b$ production at 630
and 1800 GeV favours small values of the $b$ mass and values of \mbox{$\alpha_s$}\ consistent
with the LEP measurement;
\item the relative difference in the data/theory ratio at 630 and 1800~GeV is
at most 40\%, value obtained using the CDF data. This is consistent with the
estimated effect of small-$x$ higher order corrections;
\begin{figure}[t]
\begin{center}
\psfig{figure=cdfpsp.ps,width=10cm,angle=90,clip=}
\ccaption{}{\label{fpspdata}\small Inclusive prompt $\mbox{$\psi^\prime$}$ $p_T$ distribution.
CDF data versus theory. We show the contribution of the different sources.
Dotted lines: LO production in the CSM; dashed lines: gluon and charm
fragmentation in the CSM; solid line: gluon fragmentation in the colour octet
mechanism.}
\end{center}
\end{figure}
\item there is a residual 30--40\% discrepancy in absolute normalization
between
the CDF and the D0 results, that will need to be resolved before additional
progress can be made in interpreting the data.
\end{enumerate}
\section{Charmonium production}
The production of heavy quarkonium states in high energy processes has recently
attracted a lot of theoretical and experimental interest.
Detailed measurements of differential cross sections for
production of \mbox{$\psi$}, \mbox{$\psi^\prime$}\ and $\chi$ states have recently become available
[14--18], and have been reviewed at this
Conference \cite{sansoni}.
Theoretical models for production have existed for several years (see
ref.~\cite{schuler} for a comprehensive review and references).
The comparison of these models with the most recent data
has shown dramatic discrepancies, the most striking one (theory predicting
a factor of 50 fewer prompt \mbox{$\psi^\prime$}\ than measured by CDF
\cite{cdf94}) having become known as the ``CDF anomaly''.
Attempts to explain the features of these data have recently led
to a deeper theoretical understanding of the mechanisms of
quarkonium production. I will briefly summarize here this progress (for a more
complete review, see ref.~\cite{mlmppbar}).
\begin{figure}[t]
\begin{center}
\psfig{figure=cdfchi.ps,width=10cm,angle=90,clip=}
\ccaption{}{
\label{fchi}\small Inclusive $p_T$ distribution of $\mbox{$\psi$}$'s from
$\chi_c$ production and decay.
}
\end{center}
\end{figure}
The production of quarkonium at large \mbox{$p_{\sss \rm T}$}\ is a phenomenon with two
different time scales: the shorter time scale corresponds to the generation of
the heavy quark pair, the longer one to the binding of the pair.
In all models, it is assumed that the details of the bound state
formation can be absorbed into some non-perturbative parameter, directly
related to the value of the non-relativistic wave function at the origin. Where
the models differ is in specifying how the heavy quark pair, produced by the
hard scattering in a generic colour and angular momentum state, evolves into a
state with the right quantum numbers to form the desired hadron.
In the first QCD-based model, the so-called colour singlet model (CSM,
\cite{csm}), it was assumed that the quark pair is produced in a colour singlet
state with the right angular momentum already during the hard collision. It is
easy to show that the production of quarkonium in the CSM
is heavily suppressed at large \mbox{$p_{\sss \rm T}$}, mainly because it is difficult to hold the
bound state together when this is probed at the large virtualities typical of a
high-\mbox{$p_{\sss \rm T}$}\ phenomenon \cite{mlmppbar}. In other words, in the naive CSM,
production of quarkonium is a higher-twist effect, highly suppressed by a
form-factor-like damping as soon as \mbox{$p_{\sss \rm T}$}\ becomes larger than the mass of
the state.
This prediction has been recently disproved by the data.
The dotted line in fig.~\ref{fpspdata}, for example, shows the prediction of
the
CSM model for \mbox{$\psi^\prime$}\ production at the Tevatron, compared to the CDF data
\cite{cdf94}. Not only is the overall normalization of the theory curve
significantly lower than the data, but also the shape is much steeper than
observed.
It has been pointed out recently \cite{bygluon}\ that higher order
contributions in \mbox{$\alpha_s$}\ can dominate production at large \mbox{$p_{\sss \rm T}$}.
The process responsible for these contributions is the splitting of a
large-\mbox{$p_{\sss \rm T}$}\ gluon into a $Q\bar Q$ pair, which then evolves into a colour
singlet
state by emission of one or more perturbative gluons. The additional powers of
\mbox{$\alpha_s$}\ required for this process are largely compensated by the absence of a
form factor suppression. It turns out, in fact, that these terms are of order
$[\mbox{$\alpha_s$} \times (\mbox{$p_{\sss \rm T}$}/m)^2]^n$ relative to the LO diagrams ($n=1$ in the case of
$\chi$ production, and $n=2$ in the case of \mbox{$\psi$}\ or \mbox{$\psi^\prime$}), and become
dominant as soon as \mbox{$p_{\sss \rm T}$}\ is slightly larger than $m$, the quarkonium mass.
The effect of these {\em fragmentation} contributions is shown by the dashed
line in fig.~\ref{fpspdata}: the \mbox{$p_{\sss \rm T}$}\ shape is now correct, although the total
rate is still low by more than an order of magnitude. A similar behaviour is
observed in the production of \mbox{$\psi$}'s. On the contrary, the
predictions for $\chi$ production, as shown in fig.~\ref{fchi}, agree with
the data both in shape and in rate \cite{pheno}.
A possible solution to this remaining discrepancy in the \mbox{$\psi$}\ and \mbox{$\psi^\prime$}\
sector was recently suggested by Braaten and Fleming \cite{bf}. Their proposal
is based on the observation that, contrary to the case of the gluon
fragmentation into \mbox{$\psi$}\ states, where the emitted gluons are both hard,
the fragmentation process into $\chi$ states is dominated by emission of a soft
gluon. A large infrared logarithm enhances the fragmentation rate of a gluon
into $\chi$'s relative to that into \mbox{$\psi$}'s.
This logarithm signals the presence in the $\chi$ state
of a large component made by a $c \bar c$ pair in a colour octet $^3S_1$
state, accompanied by an on-shell gluon \cite{bbl}. This component has a
non-zero overlap with the $c\bar c$ state produced by the splitting of the
large-\mbox{$p_{\sss \rm T}$}\ gluon. Braaten and Fleming suggested that a similar
colour octet $^3S_1$ component might be present in the relativistic expansion
of the \mbox{$\psi$}\ and \mbox{$\psi^\prime$}\ wave function. The work of ref.~\cite{bbl}\ indicates
that such a component should have an amplitude of order $v^2$ relative
to the leading order, colour singlet component. Therefore,
the transition of a hard gluon into a \mbox{$\psi$}, via coupling to the $^3S_1$ colour
octet component of the \mbox{$\psi$}\ wave function, would be of order $v^4/\mbox{$\alpha_s$}^2$
relative to the standard fragmentation function of the CSM. A detailed
evaluation of the transition amplitudes, \cite{bf}, shows that the ratio of the
two contributions is actually $\sim 25 \pi^2 {\cal O}(v^4)/\mbox{$\alpha_s$}^2$, a number
large enough to explain the factor of 50 discrepancy between the data and the
predictions of the CSM model.
\begin{figure}[t]
\centerline{\psfig{figure=ptinc.ps,width=10cm,angle=90}}
\caption[]{ \label{toppt}
Inclusive \mbox{$p_{\sss \rm T}$}\ distribution of the top quark.
}
\end{figure}
The contribution of this colour octet production to the \mbox{$\psi^\prime$}\ distribution is
shown as a solid line in fig.~\ref{fpspdata}. Here the new non-perturbative
parameter \mbox{$\langle {\cal O}_8^{\psi'}(^3S_1) \rangle$}, {\em i.e.\/} the value of the overlap squared between
the $c\bar c$ colour octet state from gluon splitting and the \mbox{$\psi^\prime$}\ wave
function, was derived from a fit to the data. Its numerical value
has the right order of magnitude expected from the
$v^2$ suppression, consistently with what was suggested earlier. Similar
results can be obtained for the production of \mbox{$\psi$}'s \cite{cgmp}.
The model of quarkonium production via the colour octet mechanisms is now the
subject of intense work, and we will soon have available a more
complete picture of its implications, including the phenomenology of
$\Upsilon$ and charmonium in fixed target experiments.
\section{Top quark production}
\begin{figure}[t]
\centerline{\psfig{figure=mass.ps,width=10cm,angle=90}}
\caption{ \label{tmass}
Invariant mass distribution of the $t\bar{t}$ pair.
}
\end{figure}
Now that the existence of the $top$ quark has been firmly established
via its detection in hadronic collisions \cite{toptev}, experimental
studies will focus on the determination of its properties. In particular, the
measurement of its mass and of the production cross section and distributions
will certainly be among the first studies of interest.
The production properties, should they display anomalies, could point to
the existence of exotic phenomena \cite{lane-et-al}. We present here some
kinematical distributions \cite{fmnr-top}\ that are of potential interest for
these comparisons.
The inclusive \mbox{$p_{\sss \rm T}$}\ distribution is sensitive to channels such as $W g \rightarrow t
\bar b$ \cite{wg}, which are found to contribute with a small cross section,
predominantly at low \mbox{$p_{\sss \rm T}$}.
The invariant mass of the pair is an obvious probe of the possible existence of
strongly coupled exotic resonances, such as technimesons \cite{lane-et-al}.
The transverse momentum of the pair is an indication of
the emission of hard hadronic jets in addition to those coming from the top
decays. The presence of these additional jets generates potentially large
combinatorial backgrounds to the reconstruction of the top mass peak
from the decay products \cite{cdf1}.
An accurate understanding of these backgrounds is
very important for a precise measurement of the top mass.
All the results we show were obtained using the NLO QCD matrix
elements \cite{bpt,mnr}, MRSA parton densities \cite{mrsa}, $m_{top}=176$ GeV.
Since most experimental studies are performed using shower Monte Carlo event
generators to simulate the behaviour of the events in the
detectors, we will compare our NLO results with those we obtained with HERWIG
\cite{herwig}. This is important in order to assess the reliability of the
theoretical inputs used by the experiments.
Throughout our plots, we rescale the HERWIG calculations by the perturbative
$K$ factor given by the ratio of the NLO to LO results. The $K$
factor is of the order of 1.3 for all choices of parameters.
\begin{figure}[t]
\centerline{\psfig{figure=ptpair.ps,width=10cm,angle=90}}
\caption[]{ \label{fptpair}
Transverse momentum distribution of the $t\bar{t}$ pair.
}
\end{figure}
As an example of a single inclusive quantity, we show the top \mbox{$p_{\sss \rm T}$}\ distribution
in fig.~\ref{toppt}. The solid line corresponds to the NLO result obtained
using $\mur=\muf=\muo$. The square points correspond to the HERWIG prediction,
rescaled by a $K$ factor equal to 1.34. The curves obtained by changing \mur\
and \muf\ by a factor 0.5 to 2 show an overall normalization change by
approximately $\pm 10\%$. No change in the shape is observed. The agreement
between the NLO and HERWIG results, and the stability under scale changes,
indicate that the prediction for the \mbox{$p_{\sss \rm T}$}\ distribution of top quarks is solid.
Aside from an overall small change in normalization, this result is not
affected by the inclusion of higher order soft gluon emission, \cite{smith}.
Similar conclusions \cite{fmnr-top}\ can be drawn for the rapidity
distribution, and for the distribution in invariant mass of the top pair, shown
in fig.~\ref{tmass}.
Those distributions which are trivial at leading order,
\mbox{$\Delta\phi$}\ and \mbox{$p^{t \bar t}_{\scriptscriptstyle T}$}, are on the contrary most
sensitive to multiple gluon emission from the initial state.
This is because even small perturbations
can smear a distribution that at leading order is represented by a delta
function, as is the case for the \mbox{$p^{t \bar t}_{\scriptscriptstyle T}$}\ and \mbox{$\Delta\phi$}\ ones.
The largest effect is
observed in \mbox{$p^{t \bar t}_{\scriptscriptstyle T}$}, fig.~\ref{fptpair}, where
we include the NLO curves relative to the
three choices of scales, $\mur=\muf=\muo$ (solid), \muo/2 (dots) and 2\muo
(dashes). Contrary to the previous cases, significant differences in shape
arise here among the three choices in the small-\mbox{$p_{\sss \rm T}$}\ region.
The HERWIG result (normalized to the area of the solid curve) is also shown.
The NLO and the HERWIG distributions assume the same shape only
for \mbox{$p^{t \bar t}_{\scriptscriptstyle T}$}\ larger than approximately 20 GeV.
We conclude that an accurate
description of the region \mbox{$p^{t \bar t}_{\scriptscriptstyle T}$}$<20$ GeV requires the resummation of leading
soft and collinear logarithms, as implemented in appropriate shower
Monte Carlo programs.
Studies of the angular correlations between bremstrahlung gluon jets and the
$b$-jets from the decay of the top quarks have been performed in
ref.~\cite{orr}. These authors found some important discrepancies between the
results obtained from a fixed order perturbative calculation and from HERWIG. A
full understanding of the origin of these discrepancies and a detailed study
of their possible impact on the combinatorial background to the reconstruction
of the top mass are in progress.
|
\section*{1) Introduction}
The photon is the simplest of all bosons. It is the gauge boson of QED, which
implies that it is massless and structureless (i.e., pointlike). Predictions
for $\gamma e$ interactions can be made with impressive accuracy, in some
cases (e.g., $g-2$ \cite{1}) to better than one part in $10^8$. Indeed, for
many physicists the great quantitative success of QED is one of the strongest
arguments in favour of Quantum Field Theories in general.
At first glance it might therefore be surprising, perhaps even a bit
embarassing, that many reactions involving (quasi--)real photons are much less
well understood, both theoretically and experimentally. However, a moment's
reflection will show that this is really not astonishing at all. The
uncertainty principle tells us that for a short period of time a photon can
fluctuate into a pair of charged particles. Fluctuations into virtual
two--lepton states are well understood, and are in fact a crucial ingredient
of the quantitative success of QED. Fluctuations into quark--antiquark pairs
are much more problematic, however. Whenever the lifetime of the virtual
state exceeds about $10^{-25}$ sec, corresponding to a characteristic energy
or momentum scale of about 1 GeV or less, the (virtual) \mbox{$q \bar{q}$}\ pair has
sufficient time to evolve into a complicated hadronic state that cannot be
described by perturbative methods only. Even if the lifetime is shorter, i.e.
the energy--momentum scale is larger, hard gluon emission and related
processes complicate the picture substantially.
The understanding of such virtual hadronic states becomes particularly
important when they are ``kicked on the mass shell" by an interaction of the
photon. The most thoroughly studied reactions of this type involve
interactions of a real and a virtual photon (e.g. in $e \gamma$ scattering);
of two real photons (\mbox{$\gamma \gamma $}\ scattering at \mbox{$e^+ e^- $}\ colliders); and of a real
photon with a hadron (e.g. \mbox{$\gamma p$}\ scattering). All these reactions allow to
probe some aspects of the hadronic nature of the photon, and we will discuss
them in turn.
Why is it important to study such reactions? There are at least two answers.
First, we can use data taken at present colliders to sharpen our predictions
for (background) processes at future colliders. We will see that in some cases
this may contribute significantly to the assessment of the potential of future
colliders; this is true for high--energy linear \mbox{$e^+ e^- $}\ colliders that are now
being discussed, and especially for the so--called \mbox{$\gamma \gamma $}\ colliders. Secondly,
while the hadronic structure of the photon certainly has nonperturbative
aspects, we expect the photon to be a simpler system than any real hadron,
like the proton. After all, no matter how complicated it is, the hadronic
structure of the photon certainly originated from the $\gamma \mbox{$q \bar{q}$}$ vertex;
if this vertex vanished, the photon would not have any ``hadronic nature" to
speak of. No such (in principle) simple starting point can be defined for
protons, constituent quarks themselves being highly complicated objects. One
can therefore expect photonic reactions to be particularly well suited to
study both perturbative and nonperturbative aspects of strong interactions,
and everything in between. We should mention right away that the present
theoretical description of, say, \mbox{$\gamma p$}\ scattering is, if anything, even more
complex than that of \mbox{$ \bar{p}p$}\ scattering; however, we feel that this does not
refute our argument that \mbox{$\gamma p$}\ collisions are simpler at some fundamental
level. At the very least reactions probing the hadronic nature of the photon
give us an additional handle on most aspects of perturbative and
nonperturbative QCD that are of relevance for collider physics.
The remainder of this article is organized as follows. We start in Sec.~2 with
a discussion of the scattering of a highly virtual (probing) photon off an
almost real (target) photon, $\gamma^* \gamma$ or $e \gamma$ scattering. These
were the first photonic reactions for which predictions were made in the
framework of the quark parton model (QPM) \cite{2} and within QCD \cite{3}.
$e \gamma$ scattering was also among the first of the ``hard'' photonic
reactions, which can at least partly be described by perturbation theory, to
be studied experimentally \cite{4}. Finally, the conceptual simplicity of
these processes makes them ideal for introducing the notion of photon
structure functions and the parton content of the photon.
The commissioning of the $ep$ collider HERA has opened a new era in the
experimental study of hard $\gamma p$ collisions, increasing the available
\mbox{$\gamma p$}\ centre--of--mass energy by about a factor of ten compared to previous
experiments. In Sec.~3 we attempt to cover both the recent intense theoretical
activity \cite{5} that has been triggered by this prospect, and the relevant
experimental data.
Lately there has also been much progress in the field of hard scattering of
two quasi--real photons. An important milestone was set \cite{6} by the AMY
collaboration at the TRISTAN collider, who for the first time analyzed their
data using an essentially complete leading order QCD event generator. Previous
generators had omitted important classes of diagrams, and had therefore not
been able to reproduce data taken at the older PEP and PETRA colliders.
Recent developments in this area are summarized in Sec.~4.
The emphasis in Secs.~2 to 4 is on hard processes, which are amenable to a
perturbative treatment. In Sec.~5 we loosen this restriction and comment on
issues that go beyond standard perturbative QCD. In particular, we discuss
\mbox{$\gamma p$}\ and \mbox{$\gamma \gamma $}\ cross--sections, and related quantities, in the minijet
model \cite{7}. This model has been introduced to describe purely hadronic
($pp, \ \mbox{$ \bar{p}p$}$) reactions; some modifications are necessary \cite{8}
before it can be used for \mbox{$\gamma p$}\ and \mbox{$\gamma \gamma $}\ scattering. Finally, Sec.~6
contains some concluding remarks, as well as a list of open problems and
experimental challenges.
This field is still very much in flux. Hardly a week goes by without a new
experimental study of some reaction that probes the hadronic structure of the
photon, a new or refined calculation of a relevant cross--section for given
photon structure, and/or a new model for or parametrization of this structure.
We nevertheless think it worthwhile to summarize the present status, partly
to celebrate what has already been achieved, but mostly to highlight open
questions and how to address them. We hope that this review, which updates
ref.\cite{9}, will be of some use both for those who have already worked on
some aspects of the hadronic nature of the photon, and for those who
contemplate starting such work.
\setcounter{footnote}{0}
\section*{2) Photon Structure Functions}
In this section we introduce the concept of photon structure functions. To
this end we first discuss in Sec.~2a deep--inelastic \mbox{$e \gamma $}\ scattering, which
is theoretically very clean, being fully inclusive; it is thus well suited to
serve as the defining process for photon structure functions and the parton
content of the photon. For reasons of space we have to be rather brief here;
we refer the reader to ref.\cite{9a} for a more pedagogical introduction to
photon structure functions. In Sec.~2b we then describe existing
parametrizations of the photonic parton distribution functions.
\subsection*{2a) Deep--inelastic \mbox{$e \gamma $}\ Scattering}
Studies of deep--inelastic electron nucleon scattering,
\begin{equation} \label{e2.1}
e N \rightarrow e X,
\end{equation}
where $X$ is any hadronic system and the squared four momentum transfer $Q^2
\equiv - q^2 \geq 1$ GeV$^2$, have contributed much to our understanding of
strong interactions. The pioneering SLAC experiments \cite{10} played a key
role in the development of both the quark parton model (QPM) \cite{11}, and,
due to the observation of scaling violations and the concept of asymptotic
freedom \cite{12}, of QCD. Since then our understanding of QCD in general and
the structure of the proton in particular has improved greatly. The most
recent progress in this area has come from the analysis of data taken at the
$ep$ collider HERA; see ref.\cite{13} for a study of the impact of these data
on the determination of the parton content of the proton.
Formally deep--inelastic \mbox{$e \gamma $}\ scattering is quite similar to $ep$
scattering. The basic kinematics is explained in Fig.~1. The differential
cross section can be written in terms of the scaling variables $x \equiv Q^2 /
(2 p \cdot q)$ and $y \equiv Q^2 / (s x)$, where \mbox{$\sqrt{s}$}\ is the total available
centre--of--mass (cms) energy:
\begin{equation} \label{e2.2}
\frac {d^2 \sigma (e \gamma \rightarrow e X)} {dx dy} =
\frac {2 \pi \alpha^2_{\rm em} s} {Q^4}
\left\{ \left[ 1 + \left( 1-y \right)^2 \right] \mbox{$ F_2^\gamma $}(x,Q^2) - y^2
F_L^{\gamma}(x,Q^2) \right\};
\end{equation}
this expression is completely analogous to the equation defining the protonic
structure functions $F_2$ and $F_L$ in terms of the differential
cross--section for $ep$ scattering via the exchange of a virtual
photon.\footnote{The $Z$ exchange contribution to eq.(\ref{e2.2}) is
negligible for $Q^2$ values that can be achieved in the foreseeable future.}
The special significance \cite{2} of \mbox{$e \gamma $}\ scattering lies in the fact that,
while (at present) the $x-$dependence of the nucleonic structure functions can
only be parametrized from data, the structure functions appearing in
eq.(\ref{e2.2}) can be {\em computed} in the QPM from the diagram shown in
Fig.~2a:
\begin{equation} \label{e2.3}
F^{\gamma,{\rm QPM}}_2(x,Q^2) = \frac {6 \mbox{$\alpha_{\rm em}$}}{\pi} x \sum_q e_q^4 \left\{
\left[ x^2 + \left( 1-x \right)^2 \right] \log \frac {W^2} {m_q^2} + 8 x
\left( 1-x \right) - 1 \right\},
\end{equation}
where we have introduced the squared cms energy of the $\gamma^* \gamma$
system
\begin{equation} \label{e2.4}
W^2 = Q^2 \left( \frac {1}{x} - 1 \right).
\end{equation}
The sum in eq.(\ref{e2.3}) runs over all quark flavours, and $e_q$ is the
electric charge of quark $q$ in units of the proton charge. An experimental
test of this equation was thought to not only allow to confirm the existence
of pointlike quarks, but also to measure their charges through the $e_q^4$
factor; both were topics of interest in the early 1970's when the study of
\mbox{$e \gamma $}\ scattering was first proposed \cite{2}.
Unfortunately, eq.(\ref{e2.3}) depends on the quark masses $m_q$. If this
ansatz is to describe data \cite{14} even approximately, one has to use
constituent quark masses of a few hundred MeV here; constituent quarks are not
very well defined in field theory. Moreover, we now know that QPM predictions
can be modified substantially by QCD effects. In case of proton structure
functions these lead to, among other things, scaling violations and a nonzero
$F_L$. In case of \mbox{$e \gamma $}\ scattering, QCD corrections are described by the kind
of diagrams shown in Figs.~2b,c. Diagrams of the type 2b leave the flavour
structure unchanged and are therefore part of the (flavour) nonsinglet
contribution to \mbox{$ F_2^\gamma $}, while diagrams with several disconnected quark lines,
as in Fig.~2c, contribute to the (flavour) singlet part of \mbox{$ F_2^\gamma $}.
The interest in photon structure functions received a boost in 1977, when
Witten showed \cite{3} that such diagrams can be computed exactly, at least
in the so--called ``asymptotic" limit of infinite $Q^2$. Including
next--to--leading order (NLO) corrections \cite{15}, the result can be
written as
\begin{equation} \label{e2.5}
F_2^{\gamma,{\rm asymp}}(x,Q^2) = \mbox{$\alpha_{\rm em}$} \left[ \frac {1} {\alpha_s(Q^2)}
a(x) + b(x) \right],
\end{equation}
where $a$ and $b$ are {\em calculable} functions of $x$. The absolute
normalization of this ``aymptotic" solution is therefore given uniquely by
$\alpha_s(Q^2)$, i.e. by the value of the QCD scale parameter $\Lambda_{\rm
QCD}$. It was therefore hoped that eq.(\ref{e2.5}) might be exploited for a
very precise measurement of $\Lambda_{\rm QCD}$.
Unfortunately this no longer appears feasible. One problem is that, in order
to derive eq.(\ref{e2.5}), one has to neglect terms of the form
${\displaystyle \left( \frac {\alpha_s(Q^2)} {\alpha_s(Q_0^2)} \right)^P}$,
where $Q_0^2$ is some input scale (see below). Neglecting such terms is
formally justified {\em if} $\alpha_s(Q^2) \ll \alpha_s(Q_0^2)$ {\em} and $P$
is positive. Unfortunately the first inequality is usually not satisfied at
experimentally accessible values of $Q^2$, assuming $Q_0^2$ is chosen in the
region of applicability of perturbative QCD, i.e. $\alpha_s(Q_0^2)/\pi \ll 1$.
Worse yet, $P$ can be zero or even negative! In this case ignoring such terms
is obviously a bad approximation. Indeed, one finds that eq.(\ref{e2.5})
contains divergencies as $x \rightarrow 0$ \cite{3,15}:
\begin{equation} \label{e2.6}
a(x) \sim x^{-0.59}, \ \ \ \ \ b(x) \sim x^{-1}.
\end{equation}
The coefficient of the $1/x$ pole in $b$ is {\em negative}; eq.(\ref{e2.5})
therefore predicts negative counting rates at small $x$. Notice that the
divergence is worse in the NLO contribution $b$ than in the LO term $a$. It
can be shown \cite{16} that this trend continues in yet higher orders, i.e.
the ``asymptotic" prediction for \mbox{$ F_2^\gamma $}\ rapidly becomes more and more
divergent for $x \rightarrow 0$ as more higher order corrections are included:
$\mbox{$ F_2^\gamma $} \sim x^{-4.3}$ in 3rd order, $\sim x^{-25.6}$ in 4th order, and so on.
Clearly the ``asymptotic" solution is not a very useful concept, having a
violently divergent perturbative expansion.\footnote{Notice that in the same
``asymptotic" limit, nucleonic structure functions collapse to a
$\delta-$function at $x=0$. While this is formally correct for infinite $Q^2$,
it gives obviously a poor description of the true proton structure at {\em
any} finite $Q^2$.}
The worst divergencies in $F_2^{\gamma,{\rm asymp}}$ occur in the singlet
sector, i.e. originate from diagrams of the type shown in Fig.~2c. It has been
speculated \cite{17} that this hints at a nonperturbative solution. For
example, if the invariant mass of the lowest \mbox{$q \bar{q}$}\ pair in Fig.~2c is small,
they might form a bound state. Traditionally, however, nonperturbative
contributions have been estimated using the vector dominance model (VDM)
\cite{18}, from the diagrams shown in Fig.~2d: The target photon undergoes a
transition into a nearly on--shell vector meson $(\rho, \ \omega, \ \phi,
\dots)$, so that \mbox{$e \gamma $}\ scattering is ``really" $e \rho, \ e \omega, \dots$
scattering, which should look qualitatively like $ep$ scattering. In
particular, the contribution of Fig.~2d should itself be well--behaved, i.e.
non--singular; it {\em cannot} cancel the divergencies of the ``asymptotic"
solution.\footnote{Of course, one can always {\em define} \mbox{$ F_2^\gamma $}\ to be the
sum of the (divergent) ``asymptotic" solution and a (divergent)
``nonperturbative contribution". However, nothing is gained by this as long
as one cannot at least estimate the nonperturbative contribution. Our
argument shows that the VDM is of no help here.}
Indeed, the existence of the contribution shown in Fig.~2d demonstrates that
we cannot hope to compute $\mbox{$ F_2^\gamma $}(x,Q^2)$ from perturbation theory alone.
Moreover, even if we assume that the VDM correctly describes {\em all}
nonperturbative contributions to \mbox{$ F_2^\gamma $}, it seems essentially impossible to
estimate them without making further assumptions. The problem is that the
vector mesons $\rho, \ \omega, \ \phi, \dots$ are much too short--lived to
allow an independent measurement of their parton distribution
functions.\footnote{Assuming that such distributions can even be defined for
resonance states; e.g., should a $\rho$ be treated as a \mbox{$q \bar{q}$}\ resonance with
two valence quarks, or a $\pi \pi$ resonance with four valence quarks?} It
therefore seems to us that the only meaningful approach is that suggested by
Gl\"uck and Reya \cite{19}. That is, one {\em formally} sums the contributions
from Figs.~2a--d into the single diagram of Fig.~2e, where we have introduced
quark densities in the photon \mbox{$ q_i^{\gamma} (x,Q^2) $}\ such that (in LO)
\begin{equation} \label{e2.7}
\mbox{$ F_2^\gamma $}(x,Q^2) = 2 x \sum_i e^2_{q_i} \mbox{$ q_i^{\gamma} (x,Q^2) $},
\end{equation}
where the sum runs over flavours, $e_{q_i}$ is the electric charge of quark
$q_i$ in units of the proton charge, and the factor of 2 takes care or
antiquarks. This is merely a definition. In the approach of ref.\cite{19}
one does not attempt to compute the absolute size of the quark densities
inside the photon. Rather, one introduces input distribution functions
$q^{\gamma}_{i,0}(x) \equiv q_i^{\gamma}(x,Q_0^2)$ at some scale $Q_0^2$. This
scale is in principle arbitrary, as long as $\alpha_s(Q_0^2)$ is sufficiently
small to allow for a meaningful perturbative expansion. In practice, $Q_0^2$
is usually chosen as the {\em smallest} value where this criterion is assumed
to be satisfied. We will come back to this point in the following
subsection.
Given these input distributions, the photonic parton densities, and thus
\mbox{$ F_2^\gamma $}, at different values of $Q^2$ can be computed using the inhomogeneous
evolution equations. In LO, they read \cite{3,20}:
\begin{subequations} \label{e2.8} \begin{eqalignno}
\frac { d q^{\gamma}_{\rm NS} (x,Q^2)} { d \log Q^2} &= \frac {\mbox{$\alpha_{\rm em}$}}{2 \pi}
k^{\gamma}_{\rm NS}(x) + \frac {\alpha_s(Q^2)} {2 \pi} \left( P^0_{qq} \otimes
q^{\gamma}_{\rm NS} \right) (x,Q^2) ; \label{e2.8a} \\
\frac { d \Sigma^{\gamma} (x,Q^2)} { d \log Q^2} &= \frac {\mbox{$\alpha_{\rm em}$}}{2 \pi}
k^{\gamma}_{\Sigma}(x) + \frac {\alpha_s(Q^2)} {2 \pi} \left[ \left( P^0_{qq}
\otimes \Sigma^{\gamma} \right) (x,Q^2) + \left( P^0_{qG} \otimes G^{\gamma}
\right) (x,Q^2) \right] ; \label{e2.8b} \\
\frac { d G^{\gamma} (x,Q^2)} { d \log Q^2} &= \frac {\alpha_s(Q^2)} {2 \pi}
\left[ \left( P^0_{Gq}
\otimes \Sigma^{\gamma} \right) (x,Q^2) + \left( P^0_{GG} \otimes G^{\gamma}
\right) (x,Q^2) \right], \label{e2.8c}
\end{eqalignno} \end{subequations}
where we have used the notation
\begin{equation} \label{e2.9}
\left( P \otimes q \right) (x,Q^2) \equiv \int_x^1 \frac {dy}{y} P(y)
q(\frac{x}{y},Q^2).
\end{equation}
The $P^0_{ij}$ are the usual (LO) $j \rightarrow i$ splitting functions \cite{21}.
The inhomogeneous terms $k_i^{\gamma}$ describe $\gamma \rightarrow \mbox{$q \bar{q}$}$ splitting,
i.e. the diagram of Fig.~2a; for one quark flavour, one has
\begin{equation} \label{e2.1n}
k_{q_i}^{\gamma} (x) = 3 e^2_{q_i} \left[ x^2 + (1-x)^2 \right].
\end{equation}
The $k_i^{\gamma}$ of eqs.(\ref{e2.8}) follow from eq.(\ref{e2.1n}) by taking
appropriate sums or differences of quark flavours. Eq.(\ref{e2.8a}) describes
the evolution of the nonsinglet distributions (differences of quark
densities), i.e. resums only diagrams of the type shown in Fig.~2b, while
eqs.(\ref{e2.8b},\ref{e2.8c}) describe the evolution of the singlet sector
($\Sigma^{\gamma} \equiv \sum_i q_i^{\gamma} + \bar{q}_i^{\gamma}$), which
includes diagrams of the kind shown in Fig.~2c. Notice that this necessitates
the introduction of a gluon density inside the photon \mbox{$ G^{\gamma}(x,Q^2)$}, with its
corresponding input distribution $G_0^{\gamma}(x) \equiv G^{\gamma}(x,Q_0^2)$.
It is crucial to note that, given non--singular input distributions, the
solutions of eqs.(\ref{e2.8}) will also remain \cite{19} well--behaved at all
finite values of $Q^2$. This is true both in LO and in NLO \cite{22}. On the
other hand, by introducing a priori unknown input distributions, one clearly
abandons the hope to make an absolute prediction of $\mbox{$ F_2^\gamma $}(x,Q^2)$ in terms
of $\Lambda_{\rm QCD}$ alone. The solutions of eqs.(\ref{e2.8}) still show an
approximately linear growth with $\log Q^2$; in this sense eq.(\ref{e2.5})
remains approximately correct, but the functions $a$ and $b$ now do depend
weakly on $Q^2$ (approximately like $\log \log Q^2$), and the $x-$dependence
of $b$ is {\em not} computable. In fact, only the nonleading $Q^2$ dependence
(contained in the functions $a$ and $b$), which corresponds formally to the
scaling violations in the case of $ep$ scattering\footnote{Recall that in case
of \mbox{$e \gamma $}\ scattering, the QPM also predicts a logarithmic growth of \mbox{$ F_2^\gamma $}\
with $Q^2$, see eq.(\ref{e2.3}).}, can in this approach be used to determine
$\Lambda_{\rm QCD}$; a change of the normalization $1/\alpha_s$ multiplying
the first term can always be compensated by adding a constant to the input
distributions. Notice that {\em no} momentum sum rule applies for the parton
densities in the photon as defined here. The reason is that these densities
are all of first order in the fine structure constant \mbox{$\alpha_{\rm em}$}. Even a relatively
large change in these densities can therefore always be compensated by a small
change of the ${\cal O}(\alpha^0_{\rm em})$ term in the decomposition of the
physical photon, which is simply the ``bare" photon [with distribution
function $\delta(1-x)$]. Formally this would manifest itself by the addition
of ${\cal O}(\alpha^2_{\rm em})$ contributions to the inhomogeneous terms in
eqs.(\ref{e2.8}), which are numerically negligible.\footnote{It is sometimes
argued that one can still use the ``asymptotic" NLO prediction for \mbox{$ F_2^\gamma $}\ to
determine $\Lambda_{\rm QCD}$, if one sticks to the region of large $x$ where
the influence of the $x \rightarrow 0$ pole is supposed to be weak. However, this
rests on the {\em assumption} that the nonperturbative contributions to
\mbox{$ F_2^\gamma $}\ are small at large $x$, and the {\em second assumption} that the terms
that regularize the pole \cite{23} vanish as $x \rightarrow 1$. The only way to test
these assumptions seems to be to compare the value of $\Lambda_{\rm QCD}$
extracted in this way with other measurements \cite{24} of $\alpha_s$. In our
opinion this shows that a measurement of \mbox{$ F_2^\gamma $}\ at fixed $Q^2$ {\em cannot}
be used to determine $\Lambda_{\rm QCD}$ unambiguously. Recall also that the
perturbative expansion of the ``asymptotic" prediction in yet higher orders in
QCD is extremely problematic.}
Before discussing our present knowledge of and parametrizations for the
parton densities in the photon, we briefly address a few issues related to the
calculation of \mbox{$ F_2^\gamma $}. As mentioned above, eqs.(\ref{e2.7}),(\ref{e2.8})
have been extended to NLO quite early, although a mistake in the two--loop
$\gamma \rightarrow G$ splitting function was found \cite{25} only fairly recently. A
full NLO treatment of massive quarks is now also available \cite{26} for both
\mbox{$ F_2^\gamma $}\ and $F_L^{\gamma}$. A first treatment of small$-x$ effects in the
photon structure functions, i.e. $\log 1/x$ resummation and parton
recombination, has been presented in ref.\cite{27}; however, the predicted
steep increase of \mbox{$ F_2^\gamma $}\ at small $x$ has not been observed experimentally
\cite{28}. Finally, nonperturbative contributions to \mbox{$ F_2^\gamma $}\ are expected to
be greatly suppressed if the target photon is also far off--shell. One can
therefore derive unambiguous QCD predictions \cite{29} in the region $Q^2 \gg
P^2 \gg \Lambda^2$, where the first strong inequality has been imposed to
allow for a meaningful definition of structure functions.\footnote{If $Q^2
\simeq P^2$, the use of fixed--order perturbation theory is more appropriate,
since it includes terms $\propto (P^2/Q^2)^n$, which are not included in the
usual structure functions.} However, it has recently been pointed out
\cite{30} that nonperturbative effects might survive longer than previously
expected; an unambiguous prediction would then only be possible for very large
$P^2$, and even larger $Q^2$, where the cross--section is very small.
\setcounter{footnote}{0}
\subsection*{2b) Parametrizations of Photonic Parton Densities}
As discussed in the previous subsection, the $Q^2$ evolution of the photonic
parton densities $\mbox{$\vec q^{\gamma} $}(x,Q^2) \equiv (q_i^{\gamma}, G^{\gamma})(x,Q^2)$ is
uniquely determined by perturbative QCD, eqs.(\ref{e2.8}) and their NLO
extension. However, given the flaws of the ``asymptotic" prediction for \mbox{$\vec q^{\gamma} $},
one has to specify input distributions \mbox{$\vec q^{\gamma}_0 $}\ at a fixed $Q^2 = Q_0^2$. The
situation is therefore in principle quite similar to the case of nucleonic
parton densities. In practice, however, the determination of these input
distributions is much more difficult in case of the photon, for a variety of
reasons.
To begin with, no momentum sum rule applies for \mbox{$\vec q^{\gamma}_0 $}, as discussed above. This
means that it will be difficult to derive reliable information on
$G_0^{\gamma}$ from measurements of $F_2^{\gamma}$ alone: in LO, the gluon
density only enters via the (subleading) $Q^2$ evolution in $F_2^{\gamma}$. In
contrast, analyses of $eN$ scattering revealed very early that gluons must
carry about 50\% of the proton's momentum, thereby fixing the overall scale of
the gluon density in the nucleon. We will see below that existing
parametrizations for $G^{\gamma}$ still differ by sizable factors over the
entire $x$ range unless $Q^2$ is very large.
Secondly, so far deep--inelastic $e \gamma$ scattering could only be studied
at \mbox{$e^+ e^- $}\ colliders, where the target photon is itself radiated off one of the
incoming leptons. Its spectrum is given by the well--known
Weizs\"acker--Williams function \cite{31}
\begin{equation} \label{e2.10}
\mbox{$f_{\gamma|e}$}(z) = \frac {\mbox{$\alpha_{\rm em}$}} {2 \pi z} \left\{ \left[ 1 + \left( 1-z \right)^2
\right] \log \frac {2 E^2 \left( 1 - z \right)^2 \left( 1 -
\cos \theta_{\rm max} \right) } { m_e^2 z^2 } - 2 \left( 1-z \right) \right\},
\end{equation}
where $E$ is the electron beam energy and $zE$ the energy of the target
photon. In eq.(\ref{e2.10}) we have assumed that there is no experimentally
imposed lower bound on the virtuality $p^2 \equiv -P^2$ of the target photon,
so that $P^2_{\rm min} = m_e^2 z^2 / (1-z)$; however, we have introduced an
upper limit $\theta_{\rm max}$ on the scattering angle of the electron
emitting the target photon, in order to allow for antitagging.
Eq.(\ref{e2.10}) implies that the cross section from the measurement of which
$F_2^{\gamma}$ is to be determined is of order $d \sigma / d Q^2 \sim
\alpha^4_{\rm em}/ \left( \pi Q^4 \right) \log \left( E/m_e \right)$, see
eq.(\ref{e2.2}). [Recall that $F_2^{\gamma}$ is itself ${\cal O}(\mbox{$\alpha_{\rm em}$})$.]
The event rate is therefore quite small; the most recent measurements
\cite{14,28,32} typically have around 1,000 events at $Q^2 \simeq 5$ GeV$^2$,
and the statistics rapidly gets worse at higher $Q^2$. This is to be compared
with millions of events in, for example, deep--inelastic $\nu N$ scattering
\cite{33}.
Another problem is that the $e^{\pm}$ emitting the target photon is usually
not detected, since it emerges at too small an angle. This means that the
energy of the target photon, and hence the Bjorken variable $x$, can {\em only}
be determined from the hadronic system. All existing analyses try to determine
$x$ from the invariant mass $W$, using eq.(\ref{e2.4}). This is problematic,
since at least some of the produced hadrons usually also escape undetected,
e.g. in the beam pipes; the measured value of $W$, usually denoted by $W_{\rm
vis}$, is therefore generally {\em smaller} than the true value. (It can
exceed the true $W$ due to the finite energy resolution of real--world
detectors.) One has to correct for this by ``unfolding" the measured $W_{\rm
vis}$ distribution to arrive at the true $W$ (or $x$) distribution. However,
in order to do this one has to model the hadronic system $X$. In other words,
a ``measurement" of $F_2^{\gamma}$ only seems possible if we already know many
details about multi--hadron production in $\gamma^* \gamma$ scattering!
In practice the situation is not quite as bad, since one can check the
assumptions made by comparing various distributions predicted by the Monte
Carlo with real data. An iterative procedure can then be used to arrive at a
model for $X$ that should allow to do the unfolding reliably. However, one
should be aware that this might lead to large uncertainties at the boundaries
of the accessible range of $x$ values. The reason is that in the event
selection one usually imposes upper and lower cuts on $W_{\rm vis}$,
corresponding to small and large $x$, respectively. Relatively minor changes
of the model for $X$ can therefore significantly change the predicted
efficiency for accepting events with $x$ close to one of the kinematic
boundaries. The region of large $x$ (small $W$) is in any case plagued by
higher twist uncertainties (e.g., production of resonances); however, the
region of small $x$ is very interesting, since here $F_2^{\gamma}$ is
dominated by sea quarks whose density is closely related to the gluon
distribution. It has been shown explicitly \cite{34} that different
ans\"atze for $X$ can lead to quite different ``measurements" of
$F_2^{\gamma}$ at small $x$.
Most recent analyses use the ``FKP model" of refs.\cite{35} as starting point
of the unfolding procedure. Here $F_2^{\gamma}$ is split into ``soft" and
``hard" components, depending on the virtuality of the quark exchanged in the
$t-$ or $u-$channel, see figs.~2a--c. The cut--off parameter $t_0$ separating
these two regions is to be fitted from the data. Contributions with $|t| \leq
t_0$ are modelled using VDM ideas; in practice this means that a scaling
($Q^2-$independent) ansatz is used, fitted from low$-Q^2$ data \cite{14}.
Contributions with $|t| > t_0$ can be computed from an evolution equation in
$\log |t|$, with the boundary condition that this ``pointlike" part vanish at
$t = t_0$. This procedure is formally equivalent to the one suggested by
Gl\"uck and Reya \cite{19}, {\em if} we identify $Q_0^2 = t_0$ and impose the
upper limit $|t|_{\rm max} = Q^2$ for the $t-$evolution. However, as presently
used \cite{36,28,32}, the unfolding procedure has several weaknesses:
\begin{itemize}
\item It uses a parametrization \cite{36} of the ``pointlike" part of
$F_2^{\gamma}$ which includes some terms which are of next--to--leading order
in a formal operator product expansion; however, not all such terms are
included. In particular, it uses the actual kinematical maximum for the
$t-$evolution, $|t|_{\rm max} \simeq Q^2/x$; as a result, the predicted
$F_2^{\gamma}$ tends to be larger at small and median $x$ than what one would
get from the usual $Q^2-$evolution.\footnote{Very roughly, one replaces the
overall growth $\propto \log Q^2$ by $\log \left( Q^2/x \right)$.} On the
other hand, this expression ignores sea quarks, i.e. $g \rightarrow \mbox{$q \bar{q}$}$ splitting;
it therefore underestimates the true result for very small $x$.
\item The $Q^2$ (or $t$) evolution of the ``hadronic part" is ignored; at high
$Q^2$, this over--estimates the soft contribution to $F_2^{\gamma}$ at median
and large $x$, and underestimates it at small $x$.
\item While this procedure treats the effect of gluon radiation on the shape
of $F_2^{\gamma}$ (approximately) correctly, it does {\em not} include any
parton showering in the MC event generator. Rather, the generator produces
\mbox{$q \bar{q}$}\ pairs whose $p_T$ distribution in the $\gamma^* \gamma$ cms is either
exponential (for the ``hadronic part") or follows the QPM prediction (for the
``pointlike part"). This \mbox{$q \bar{q}$}\ pair is then fed into a string--based
fragmentation program. While string fragmentation can mimick the effects of
parton showering to some extent, it is not able to produce additional jets.
It is therefore not surprising that a recent analysis \cite{32} found larger
jet rates in $e \gamma$ scattering than the MC predicted.
\end{itemize}
Some of these points have also been raised in ref.\cite{36a}. We would like to
emphasize that none of these problems is intrinsic to the formalism of
ref.\cite{35}. Moreover, the input model used for the unfolding always turned
out to be consistent within errors with the extracted $F_2^{\gamma}$, if $t_0$
is chosen in the vicinity of 0.5 GeV$^2$.
The point of this lengthy discussion is not to criticize our experimental
colleagues. Rather, we hope that it might serve as a starting point for
further work by people familiar with the design of MC event generators, which
we are not. However, this discussion shows that present data on $F_2^{\gamma}$
have to be taken with a grain of salt. In particular, the shortcomings listed
above are usually {\em not} addressed in the estimates of the systematic error
due to the unfolding procedure; this error only includes things like the
choice of binning \cite{28}. This might help to explain the apparent
discrepancy between different data sets \cite{14}. Fortunately, new ideas
for improved unfolding algorithms \cite{36b} are now under investigation; this
should facilitate the measurement of $F_2^{\gamma}$ at small $x$, especially
at high energy (LEP2).
In spite of this, measurements of $F_2^{\gamma}$ probably still provide the
most reliable constraints on the input distributions $\mbox{$\vec q^{\gamma}_0 $}(x)$; they are
certainly the only data that have been taken into account when constructing
existing parametrizations of \mbox{$ \vec q^{\gamma} (x,Q^2) $}. In the remainder of this subsection we
will briefly describe these parametrizations.
The simplest and oldest parametrizations \cite{37,37a} are based on the
``asymptotic" LO prediction \cite{3,37b}. While the theoretical basis for this
prediction is weak, as described in the previous subsection, their simplicity
makes these parametrizations useful for first estimates of reaction rates, as
long as one stays away from very small $x$. Refs.\cite{37,37a} only give
parametrizations for $N_f=4$ active flavours. More recently, Gordon and
Storrow \cite{38} provided different, more accurate fits for $N_f = 3, \ 4$
and $5$.
All other parametrizations involve some amount of data fitting. However, due
to the rather large experimental errors of data on $F_2^{\gamma}$, additional
{\em assumptions} always had to be made. In particular, quark and antiquark
distributions (of the same flavour) are always assumed to be identical, which
guarantees that the photon carries no flavour. This assumption is eminently
reasonable. The $\gamma \mbox{$q \bar{q}$}$ vertex treats quarks and antiquarks
symmetrically, and we do not know of any effect that could destroy this
symmetry.
The \underbar{DG parametrization} \cite{39} was the first to start from input
distributions. At the time of this fit, only a single measurement of
$F_2^{\gamma}$ at fixed $Q^2\simeq 5.3$ GeV$^2$ existed. In order to determine
three, in principle independent, input distributions (for nonsinglet and
singlet quarks as well as gluons), two assumptions were made: All input quark
densities were assumed to be proportional to the squared quark charges, i.e.
$\mbox{$ u^{\gamma}$} = 4 \mbox{$ d^{\gamma}$} = 4 \mbox{$ s^{\gamma}$}$ at $Q_0^2 = 1$ GeV$^2$; and the gluon input was
generated purely radiatively, i.e.
\begin{equation} \label{e2.11}
G^{\gamma}_{0, {\rm DG}} = \frac {2} {\beta_0} \Sigma^{\gamma}_{0, {\rm DG}}
\otimes P_{Gq},
\end{equation}
where $\beta_0$ = 9 is the coefficient of the 1--loop QCD $\beta$ function.
This parametrization only exists in LO. Moreover, it treats flavour thresholds
by introducing three independent sets of distributions for $N_f=3, \ 4, \ 5$,
so that the transition across a threshold is not automatically smooth.
Nevertheless it continues to describe data reasonably well, although a
combined fit would probably give a higher $\chi^2$ than for more recent
parametrizations.
The \underbar{LAC parametrizations} \cite{40} are based on a much larger data set.
The main point of these fits was to demonstrate that data on $F_2^{\gamma}$
constrain \mbox{$ G^{\gamma}$}\ very poorly. In particular, they allow a very hard gluon,
with $x \mbox{$ G^{\gamma}$}$ having a maximum at $x \simeq 0.9$ (LAC3), as well as very soft
gluon distributiuons, with $x\mbox{$ G^{\gamma}$}$ rising very steeply at low $x$ (LAC1, LAC2).
The LAC parametrizations only exist for $N_f=4$ massless flavours and in LO.
{\em No} assumptions about the relative sizes of the four input quark
densities were made in the fit. LAC3 has been clearly excluded by data on jet
production in $ep$ scattering (Sec.~3a) as well as in real \mbox{$\gamma \gamma $}\ scattering
(Sec.~4a); the experimental status of LAC1,2 is less clear.
The recent \underbar{WHIT parametrizations} \cite{41} follow a similar philosophy
as LAC, at least regarding the gluon input; however, their choices for
$G_0^{\gamma}$ are much less extreme. In the WHIT1,2,3 parametrizations,
gluons carry about half as much of the photon's momentum as quarks do (at the
input scale $Q_0^2 = 4$ GeV$^2$), while in WHIT4,5,6 gluons and quarks carry
about the same momentum fraction. These two groups of fits also have slightly
different valence quark densities.\footnote{The definition of ``valence" and
``sea" quarks used here differs from the more common ``nonsinglet" and
"singlet" distributions. The $Q^2$ evolution of the valence density is
independent of \mbox{$ G^{\gamma}$}, i.e. obeys eq.(\ref{e2.8a}), while a non--zero sea
quark density is produced only through $g \rightarrow \mbox{$q \bar{q}$}$ splitting. In the
absence of mass effects, the valence distributions are proportional to the
squared quark charges, while the sea distributions are independent of the
quark charge.} The input gluon density is assumed to have the simple shape $x
G_0^\gamma \propto \left( 1-x \right)^{c_g}$, with $c_g=3$ for WHIT1,4; $c_g=9$
for WHIT2,5; and $c_g=15$ for WHIT3,6. Recall that the normalization is
adjusted such that $\int x G_0^\gamma(x) dx$ is constant within each group of
parametrizations; a larger $c_g$ therefore also means a larger \mbox{$ G^{\gamma}$}\ at small
$x$. The input distributions for the sea quarks are computed from the cross
section for $\gamma^* g \rightarrow \mbox{$q \bar{q}$}$, regularized by light quark masses $m=0.5$
GeV. The WHIT parametrizations only exist in LO, but great care has been taken
to treat the ($x-$dependent) charm threshold correctly. This is much more
important here than for nucleonic parton densities, since the photon very
rapidly develops an ``intrinsic charm" component from $\gamma \rightarrow \mbox{$c \bar{c}$}$
splitting.
The \underbar{GRV parametrization} \cite{42} is the first NLO fit of \mbox{$\vec q^{\gamma} $}; a LO
version is also available. This parametrization is based on the same
``dynamical" philosophy as the earlier fits of protonic \cite{43} and pionic
\cite{44} parton densities by the same authors. The idea is to start from a
very simple input at a very low $Q_0^2$ (0.25 GeV$^2$ in LO, 0.3 GeV$^2$ in
NLO); this scale is assumed to be the same for $p, \ \pi$ and $\gamma$
targets. The observed, more complex structure is then generated dynamically by
the evolution equations. In case of the proton, only valence quarks were
originally assumed to be present at scale $Q_0^2$. While the gluon density
does evolve fast enough to carry approximately half the proton's momentum at
$Q^2$ of a few GeV$^2$, it was found to be too soft in shape. The ansatz
therefore had to be modified \cite{45} by introducing ``valence--like" gluon
and even sea quark densities already at the input scale, thereby giving up
much of the original simplicity of the idea. Their pionic input distributions
also include a ``valence--like" gluon density, which is in fact strictly
proportional to the valence quark density, but no sea quarks at scale $Q_0^2$.
In the photonic case,
\begin{equation} \label{e2.12}
\vec{q}^{\gamma}_{0,{\rm GRV}}(x) = \kappa \frac {4 \pi \mbox{$\alpha_{\rm em}$}} {f^2_{\rho}}
\vec{q}^{\pi}_{0,{\rm GRV}}(x).
\end{equation}
This input is motivated from VDM ideas, where $f_{\rho}^2$ determines the
$\gamma \rightarrow \rho$ transition probability ($f^2_{\rho}/4\pi \simeq 2.2$), and
$\kappa$ has been introduced to describe contributions from heavier vector
mesons ($\omega, \phi, \dots$). In fact, $\kappa$ is the {\em only} free
parameter in this ansatz; it was determined to be $\kappa=2$ (1.6) in LO (NLO).
However, it should be clear that eq.(\ref{e2.12}) is an {\em assumption} that
has to be tested experimentally; in particular, it is not obvious that the
parton densities in a pion resemble those in a vector meson, or that QCD is
applicable at such small input scales.\footnote{It has been shown previously
\cite{45a} that an ansatz like eq.(\ref{e2.12}) cannot be brought into
agreement with the data if one insists on $Q_0^2 \geq 1$ GeV$^2$. One way out
\cite{27} would be to multiply the resulting $F_2^{\gamma}$ with something
like $Q^2/(Q^2 + \mu^2)$, but this goes beyond the leading--twist partonic
contributions.} On the positive side, the GRV parametrization ensures a smooth
onset of the charm density, using an $x-$independent threshold. Moreover, care
has been taken to split the NLO parametrizations for \mbox{$\vec q^{\gamma} $}\ into LO and NLO
pieces; only the former should be multiplied with NLO pieces of hard cross
sections.\footnote{This distinction is more important in the photonic case,
since possible $\log ( 1-x)$ divergencies are {\em not} always regularized by
parton densities falling like a power of $1-x$. However, even in the photonic
case this problem is greatly ameliorated if one uses the ``DIS$_{\gamma}$"
scheme introduced \cite{46} by the same authors.}
The \underbar{GS parametrizations} \cite{38} were developed shortly after GRV, but
follow a quite different strategy. Problems with low input scales \cite{38a}
are avoided by choosing $Q_0^2 = 5.3$ GeV$^2$. This is certainly in the
perturbative region, but necessitates a rather complicated ansatz for the
input distributions:
\begin{equation} \label{e2.13}
\vec{q}^{\gamma}_{0,{\rm GS}}(x) = \kappa \frac {4 \pi \mbox{$\alpha_{\rm em}$}} {f^2_{\rho}}
\vec{q}^{\pi}_0(x,Q_0^2) + \vec{q}^{\gamma}_{\rm QPM}(x,Q_0^2).
\end{equation}
The free parameters in the fit are the momentum fractions carried by gluons
and sea--quarks in the pion, the parameter $\kappa$, and the light quark
masses. In the GS2 parametrization, $G_0^{\gamma}$ is assumed to come entirely
from the first term in eq.(\ref{e2.13}), while in GS1 the second term also
contributes via radiation, see eq.(\ref{e2.11}). The fit gives $m_u = m_d =
0.29$ GeV and $\kappa = 1.96$. These values are not unreasonable. However, the
ansatz (\ref{e2.13}) might be somewhat suspect: It holds in the perturbative
region of $Q^2$, but its form is not invariant under the evolution equations.
On the other hand, for practical purposes it includes sufficiently many free
parameters to allow a decent description of data on $F_2^{\gamma}$. This
parametrization is now being updated \cite{38b}. The new version uses a
slightly reduced input scale $Q_0^2=3$ GeV$^2$, and for the first time
includes data on jet production in two--photon collisions (see Sec.~4a) in the
fit; unfortunately this still does not allow to pin down \mbox{$G^{\gamma}$}\ with any
precision.
The \underbar{AGF parametrization} \cite{47} is (in its ``standard" form) quite
similar to GRV. In particular, they also assume that at a low input scale
$Q_0^2 = 0.25$ GeV$^2$ the photonic parton densities are described by the VDM.
There are some differences, however. First, AGF point out that in NLO the
input densities are scheme--dependent, if physical quantities like
$F_2^{\gamma}$ are to be scheme--independent. GRV use their ansatz
(\ref{e2.12}) in their DIS$_{\gamma}$ scheme, since it is perturbatively more
stable than the more commonly used $\overline{\rm MS}$ scheme. AGF point out
that this treatment includes certain process--dependent terms in \mbox{$\vec q^{\gamma} $};
they therefore prefer to use the $\overline{\rm MS}$ scheme, and define their
input distributions to be the difference of a (regular) ``VDM" term and a
process--independent term containing a $\log(1-x)$ divergence, so that
$F_2^{\gamma}(x,Q_0^2)$ is well--behaved in the limit $x \rightarrow 1$. Secondly,
they include $\rho-\omega-\phi$ interference when specifying the ``VDM" input,
so that
\begin{eqalignno} \label{e2.14}
u^{\gamma}_{0,{\rm AGF}}(x) &= K \mbox{$\alpha_{\rm em}$} \left[ \frac{4}{9}
u^{\pi}_{\rm valence}(x,Q_0^2) + \frac{2}{3} u^{\pi}_{\rm sea}(x,Q_0^2)
\right]; \nonumber \\
d^{\gamma}_{0,{\rm AGF}}(x) = s^{\gamma}_{0,{\rm AGF}}(x) & =
K \mbox{$\alpha_{\rm em}$} \left[ \frac{1}{9} u^{\pi}_{\rm valence}(x,Q_0^2)
+ \frac{2}{3} u^{\pi}_{\rm sea}(x,Q_0^2) \right]; \nonumber \\
G^{\gamma}_{0,{\rm AGF}}(x) &= K \mbox{$\alpha_{\rm em}$} \frac{2}{3} G^{\pi}(x,Q_0^2),
\end{eqalignno}
where the pion distribution functions are taken from ref.\cite{48}. (Recall
that in NLO the input quark densities have to be modified by a subtraction
term.) Perhaps the for practical purposes most important difference from the
GRV parametrization is that the coefficient $K$ is left free, i.e. separate
fits are provided for the ``anomalous" (or ``pointlike") and
``nonperturbative" contributions to \mbox{$\vec q^{\gamma} $}, allowing the user to specify the
absolute normalization (although not the shape) of the latter.
Finally, two of the \underbar{SaS parametrizations} \cite{36a} are based on a
similar philosophy as the GRV and AGF parametrizations, by assuming that at a
low $Q_0 \simeq 0.6$ GeV the perturbative component vanishes (SaS1). However,
while the normalization of the nonperturbative contribution is taken from
the VDM (including $\rho, \ \omega$ and $\phi$ contributions with fixed
normalization), the shapes of the quark and gluon distributions are fitted
from data. Although the SaS parametrizations are available in LO only, the
authors attempt to estimate the scheme dependence (formally an NLO effect)
providing a parametrization (SaS1M) where the non leading--log part of the QPM
prediction for $F_2^{\gamma}$ has been added to eq.(\ref{e2.7}), while SaS1D
is based on eq.(\ref{e2.7}) alone. There are also two parametrizations with
$Q_0=2$ GeV; however, in this case the normalization of the fitted ``soft"
contribution had to be left free, and its shape is much harder than one what
expects from hadronic parton densities. Again two sets of parametrizations are
available, using different schemes (SaS2D, SaS2M).
The SaS1 sets preferred by the authors are quite similar to AGF; the real
significance of ref.\cite{36a} is that it carefully describes the properties
of the hadronic state $X$ for both the hadronic and ``anomalous"
contributions, as needed for a full event characterization. We will come back
to this aspect of their work in Sec.~5.
In Fig.~3 we compare various LO parametrizations of $F_2^{\gamma}$ at
$Q^2=15$ GeV$^2$ with recent data taken by the OPAL \cite{28} and TOPAZ
\cite{32} collaborations; present data are not able to distinguish between LO
and NLO fits. In order to allow for a meaningful comparison, we have added a
charm contribution to the OPAL data, as estimated from the QPM; this
contribution had been subtracted in their analysis. We have used the DG and GRV
parametrizations with $N_f=3$ flavours, since their parametrizations of
$c^\gamma$ are meant to be used only if $\log Q^2 / m_c^2 \gg 1$; the charm
contribution has again been estimated from the QPM.\footnote{We have ignored
the small contribution \cite{26} from $\gamma^* g \rightarrow \mbox{$c \bar{c}$}$ in this figure.}
As discussed earlier, WHIT provides a parametrization of \mbox{$ c^{\gamma}$}\ that includes
the correct kinematical threshold, while LAC treat the charm as massless at
all $Q^2$.
We see that most parametrizations give quite similar results for
$F_2^{\gamma}$ over most of the relevant $x-$range; the exception is LAC1,
which exceeds the other parametrizations both at large and at very small $x$.
It should be noted that the data points represent averages over the
respective $x$ bins; the lowest bin starts at $x=0.006 \ (0.02)$ for the OPAL
(TOPAZ) data. The first OPAL point is therefore in conflict \cite{34} with the
LAC1 prediction. Unfortunately there is also some discrepancy between the
TOPAZ and OPAL data at low $x$. As discussed above, one is sensitive to the
unfolding procedure here; for this reason, WHIT chose not to use these (and
similar) points in their fit. (The other fits predate the data shown in
Fig.~3.) This ambiguity in present low$-x$ data is to be regretted, since
in principle these data have the potential to discriminate between different
ans\"atze for $G_0^{\gamma}$. This can most clearly be seen by comparing the
curves for WHIT4 (long dashed) and WHIT6 (long--short dashed), which have the
{\em same} valence quark input, and even the same $\int x G_0^{\gamma} dx$:
WHIT4 has a harder gluon input distribution, and therefore predicts a larger
$F_2^{\gamma}$ at $x \simeq 0.1$; WHIT6 has many more soft gluons, and
therefore a very rapid increase of $F_2^{\gamma}$ for $x \leq 0.05$, not
unlike LAC1. Finally, we should mention that the GS, AGF and SaS
parametrizations also reproduce these data quite well.
Discriminating between these parametrizations would be much easier if one
could measure the gluon density directly. This is demonstrated in Fig.~4a,
where we show results for $x \mbox{$ G^{\gamma}$}$ at the same value of $Q^2$; we have
chosen the same LO parametrizations as in Fig.~3, and included the LAC3
parametrization with its extremely hard gluon density. Note that, for
example, WHIT4 and WHIT6 now differ by a factor of 5 for $x$ around 0.3.
The gluon distribution of WHIT6 is rather similar in shape to the one of LAC1,
but significantly smaller in magnitude. Indeed, in all three LAC
parametrizations, gluons carry significantly more momentum than quarks for
$Q^2 \leq 20$ GeV$^2$; this is counter--intuitive \cite{38}, since in known
hadrons, and hence presumably in a VMD--like low$-Q^2$ photon, gluons and
quarks carry about equal momentum fractions, while at very high $Q^2$ the
inhomogeneous evolution equations (\ref{e2.8}) predict that quarks in the
photon carry about three times more momentum than gluons. Notice finally that
GRV predicts a relatively flat gluon distribution. This results partly from the
low value of the input scale $Q_0^2=0.25$ GeV$^2$, compared to 1 GeV$^2$ for
DG and 4 GeV$^2$ for WHIT and LAC1; a larger $Q^2/Q_0^2$ allows for more
radiation of relatively hard gluons off large$-x$ quarks. Recall also that
their (pionic) input distribution includes a valence--like (hard) gluon
density.
Finally, Fig.~4b shows that some parametrizations also differ substantially in
the flavour structure. Both DG and WHIT assume $\mbox{$ d^{\gamma}$}=\mbox{$ s^{\gamma}$}$ at all $Q^2$.
Moreoever, DG assumes $q^{\gamma}_{i,0} \propto e_{q_i}^2$ for the entire
input quark distribution, while WHIT assumes this only for the valence input.
Nevertheless the smaller value of $Q_0^2$ assumed by DG leads to a strangeness
content quite similar to that predicted by WHIT: At very low $x$, sea quarks
dominate, which have $\mbox{$ u^{\gamma}$}=\mbox{$ d^{\gamma}$}=\mbox{$ s^{\gamma}$}$, so that $\mbox{$ s^{\gamma}$}/(\mbox{$ u^{\gamma}$}+\mbox{$ d^{\gamma}$}) \simeq
1/2$; at high $x$, $q^{\gamma}_i \propto e^2_{q_i}$, so that $\mbox{$ s^{\gamma}$}/(\mbox{$ u^{\gamma}$}+\mbox{$ d^{\gamma}$})
\simeq 1/5$.\footnote{The ratio of 4:1 between large$-x$ (valence) $u$ and
$d$ quark densities is implicit to the structure of the WHIT fits, while in
DG, \mbox{$ u^{\gamma}$}\ and \mbox{$ d^{\gamma}$}\ are parametrized independently; this explains the $\sim 2\%$
deviation between the two parametrizations shown in Fig.~4b at large $x$.}
Since sea quarks are produced by gluon splitting, the transition between the
sea--dominated and valence--dominated regions depends on \mbox{$ G^{\gamma}$}.
In contrast, GRV assumes $u_0^{\gamma}=d_0^{\gamma}$ and $s_0^{\gamma}=0$ at
the input scale; for given $F_2^{\gamma}$, the former assumption increases
$\mbox{$ u^{\gamma}$}+\mbox{$ d^{\gamma}$}$ and the latter reduces \mbox{$ s^{\gamma}$}, compared to the ansatz $q_i^{\gamma}
\propto e^2_{q_i}$. This explains the smallness of the strangeness content
of the photon predicted by GRV, which persists to surprisingly large values
of $x$. It is worth mentioning that AGF, which is otherwise quite similar to
GRV [at least for the standard normalization of the nonperturbative
contribution, $K=1$ in eqs.(\ref{e2.14})], also assumes the input valence
quark distributions to be proportional to the squared quark charges; its
predictions for $\mbox{$ s^{\gamma}$}/(\mbox{$ u^{\gamma}$}+\mbox{$ d^{\gamma}$})$ are therefore quite similar to those of the
WHIT group. GS falls between WHIT and GRV, since here a small but nonzero
$s_0^{\gamma}$ is assumed, from the pionic sea quarks as well as the QPM part
in eq.(\ref{e2.13}). On the other hand, LAC treats all four quark input
distributions as completely independent quantities, without imposing any
constraints between them. This results in the erratic behaviour of the
strangeness content depicted in Fig.~4b; the ratio even exceeds unity for
$x \simeq 0.1$. Clearly LAC should therefore not be used when the flavour
structure of the photon is important, e.g. in $W$ and $Z$ boson production at
HERA.
\setcounter{footnote}{0}
\section*{3) Resolved Photon Processes in \mbox{$\gamma p$}\ Scattering}
In this section we discuss \mbox{$\gamma p$}\ scattering reactions that are sensitive to
the hadronic structure of the photon. Most of our numerical results will be
for the $ep$ collider HERA, with appropriate (anti--)tagging conditions for
the outgoing electron, in order to make sure that the virtuality of the
exchanged photon is small; we will also make a few comments on fixed target
\mbox{$\gamma p$}\ scattering. We follow the terminology of ref.\cite{49} in
distinguishing between {\em direct} and {\em resolved photon} contributions to
a given process; the existence of these physically distinct contributions had
already been emphasized in ref.\cite{37b}. The former are defined as reactions
where the incident photon participates directly in the relevant hard
scattering process; some examples are shown in Fig.~5a. In contrast, in
resolved photon contributions the incoming photon takes part in the hard
scattering via one of its constituents, a quark or gluon, as shown in Fig.~5b.
Notice that in direct processes the entire photon energy goes into the hard
partonic final state, while in resolved photon processes only a fraction
$x_{\gamma}$ of the photon energy is available for the hard scattering. This
also leads to different topologies for the two classes of contributions. In
LO, the direct contribution to jet production will give rise to two high$-p_T$
jets and the remnant jet from the proton, see Fig.~6a; there will be little or
no hadronic activity in the direction of the incident photon. In contrast,
resolved photon contributions are characterized by a second spectator jet,
which is formed when a coloured parton is ``taken out" of the photon; this
photonic remnant or spectator jet will usually go approximately in the
direction of the incident photon, which coincides with the electron direction
at HERA. This additional jet in principle allows distinction between direct
and resolved photon contributions on an event--by--event basis.
Occasionally one sees the erroneous statement in the literature that resolved
photon contributions are NLO corrections to the corresponding direct process.
However, by counting powers of coupling constants in Figs.~5a and 5b one can
easily convince oneself that this is not the case. The direct contribution to
di--jet production is obviously of order $\mbox{$\alpha_{\rm em}$} \mbox{$\alpha_s$}$. The resolved photon
contribution is of order $\mbox{$\vec q^{\gamma} $} \cdot \alpha^2_s$, where \mbox{$\vec q^{\gamma} $}\ stands for any
photonic parton density. We have seen early in Sec.~2 that these densities are
of first order in \mbox{$\alpha_{\rm em}$}. Moreover, we also saw that they grow logarithmically
with the relevant scale $Q^2$ of the given process; in leading--log summed
perturbation theory this $\log Q^2$ has to be counted as a factor $1/\mbox{$\alpha_s$}$,
see eq.(\ref{e2.5}).\footnote{Recall that this expression remains
approximately correct even beyond the ``asymptotic" approximation, if we allow
a weak $Q^2$ dependence of the functions $a$ and $b$.} Altogether the photonic
parton densities therefore have to be counted as ${\cal O} \left( \mbox{$\alpha_{\rm em}$} /
\mbox{$\alpha_s$} \right)$, so that $\mbox{$\vec q^{\gamma} $} \cdot \alpha^2_s$ is again of order $\mbox{$\alpha_{\rm em}$}
\mbox{$\alpha_s$}$, just like the direct contribution. It should be emphasized that LO
QCD is {\em always} leading log summed, i.e. $\mbox{$\alpha_s$}(Q^2) \cdot \log Q^2$ is
always counted as ${\cal O}(1)$, not ${\cal O}(\mbox{$\alpha_s$})$. To give one example,
the resummation of such leading logarithms leads to the scaling violations
of nucleonic structure functions; it should be clear that one is {\em not}
performing an NLO analysis by simply using $Q^2-$dependent parton densities
when estimating \mbox{$p \bar{p}$}\ cross--sections.
Part of the confusion is caused by the fact that in NLO, direct and resolved
photon contributions mix. This can, e.g., be seen from the Feynman diagram
of Fig.~7, which shows an NLO contribution to direct jet production. If no
restrictions are imposed on the transverse momentum of the outgoing
antiquark, it could go in the photon direction and thus form a ``remnant
jet". In fact, the total contribution from the diagram of Fig.~7 to the
inclusive jet (pair) cross--section will be dominated by configurations where
the transverse momentum of the antiquark is small, due to the $1/t$ pole
associated with the exchanged quark. The crucial point is that the
contribution from this pole has already been included in the resolved LO $q q'$
scattering process, where the exchanged quark is treated as being on--shell.
In order to avoid double--counting, the contribution of the $1/t$ pole
therefore has to be subtracted from the NLO contribution of Fig.~7; in other
words, the corresponding collinear divergence is absorbed in the quark
distribution function in the photon. We emphasize that this treatment is
completely analogous to the calculation of NLO corrections to jet production
in \mbox{$p \bar{p}$}\ scattering. For example, the incident photon in Fig.~7 could be
replaced by a gluon coming from the $\bar{p}$. The above argument then tells
us that $qq'$ and $gq$ scattering mix in NLO, i.e. a part of the NLO
contribution from $gq$ scattering has to be absorbed in $q q'$ scattering;
nevertheless nobody would consider one to be an NLO correction to the
other.\footnote{The diagram of Fig.~7 also has a divergence when the exchanged
gluon goes on--shell; this is absorbed in the gluon density in the proton,
i.e. in direct $\gamma g$ scattering. Finally, in principle both the exchanged
quark and the gluon can be (nearly) on--shell. However, in this case none of
the final state partons has large transverse momentum; this configuration
therefore only contributes to soft processes, which cannot be treated
perturbatively.}
We do not attempt to split resolved photon contributions into those coming
from the ``anomalous", ``pointlike" or ``perturbative" part of \mbox{$\vec q^{\gamma} $}\ and
those due to the ``hadronic" or ``nonperturbative" part. We have seen in the
previous section that it is not easy to separate these parts consistently;
indeed, it should be clear that in reality there is a smooth transition from
the one to the other. Nevertheless we will see later (in Sec.~5) that the
existence of the ``pointlike" contribution may have some impact on overall
event characteristics.
After these preliminaries, we are ready to discuss various hard \mbox{$\gamma p$}\
reactions. We start with jet production in Sec.~3a, where both NLO
calculations and high--energy data from the $ep$ collider HERA are available.
Open heavy quark production (Sec.~3b) has also been treated in NLO, and first
HERA data have started to appear. Direct photon production (Sec.~3c) also also
been treated in NLO, but no HERA data have yet been published. We then discuss
the production of \mbox{$J/\psi$}\ mesons in Sec.~3d, and lepton--pair (Drell--Yan)
production in Sec.~3e.
\setcounter{footnote}{0}
\subsection*{3a) Jet Production in \mbox{$\gamma p$}\ Collisions}
The production of high$-p_T$ jets offers the largest cross--section of all
hard \mbox{$\gamma p$}\ scattering reactions. It was therefore the first such process for
which resolved photon contributions were calculated \cite{50}, and also among
the first reactions to be studied experimentally at HERA \cite{51,52}.
Moreover, many properties of direct and resolved photon contributions to jet
production carry over to the production of heavy quarks and direct photons, to
be discussed in subsequent subsections. We therefore treat jet production in
somewhat more detail than other photoproduction processes.
Feynman diagrams contributing to jet production in \mbox{$\gamma p$}\ scattering are
sketched in Fig.~5. In leading order (LO), direct contributions (Fig.~5a) come
from either the ``QCD Compton" process (left diagram), or from photon--gluon
fusion (right diagram); notice that, unlike DIS, this latter process is
sensitive to the gluon content of the proton already in LO. The resolved
photon contributions (Fig.~5b) involve the same matrix elements for $2
\rightarrow 2$ QCD scattering processes that appear in calculations of jet
production at purely hadronic (\mbox{$p \bar{p}$}\ or $pp$) scattering. Note that, in
contrast to the direct processes, many of these QCD scattering processes can
proceed via the exchange of a gluon in the $t-$ or $u-$channel; processes like
$gg \rightarrow \mbox{$q \bar{q}$}$ that only proceed via the exchange of a quark in the $t-$ or
$u-$channel, and/or the exchange of a gluon in the $s-$channel, contribute
only little to the inclusive jet cross--section.
In LO, the cross--section for the electroproduction of two (partonic) jets
with transverse momentum $p_T$ and (pseudo)rapidities\footnote{There is no
difference between the partonic rapidity and pseudorapidity in LO.} $\eta_1,
\ \eta_2$ can be written as:
\begin{equation} \label{e3.1}
\frac {d^3 \sigma (e p \rightarrow e j_1 j_2 X)} {d p_T d \eta_1 d \eta_2} = 2 p_T x_e
x_p \sum_{i,j,k,l} f_{i|e}(x_e) f_{j|p}(x_p) \frac {d \hat{\sigma}_{ij \rightarrow kl}
(\hat{s},\hat{t},\hat{u})} {d \hat{t}}.
\end{equation}
Here, $i$ stands for a photon, quark or gluon, and $j, \ k, \ l$ stand for a
quark or gluon. If $i=\gamma$ (direct contribution), the function $f_{i|e}$ is
just the Weizs\"acker--Williams photon flux (\ref{e2.10}); otherwise it is
given by
\begin{equation} \label{e3.2}
f_{i|e}(x_e) = \int_{x_e}^1 \frac {dz}{z} \mbox{$f_{\gamma|e}$}(z) f_{i|\gamma} \left( \frac
{x_e}{z} \right) .
\end{equation}
The resolved photon contribution to the cross--section (\ref{e3.1}) therefore
only depends on the product $x_e$ of the fraction $z$ of the electron energy
carried by the incident photon, and the fraction \mbox{$x_{\gamma}$}\ ($=x_e/z$) of the
photon energy carried by the parton in the photon. Of course, $f_{i|\gamma}$
and $f_{j|p}$ are nothing but the parton densities in the photon and proton,
respectively.
The (pseudo)rapidities of the jets are related to these Bjorken variables
by:
\begin{subequations} \label{e3.3} \begin{eqalignno}
x_p &= \frac {1}{2} x_T \sqrt{ \frac{E_e}{E_p} } \left( e^{\eta_1} +
e^{\eta_2} \right) ; \label{e3.3a} \\
x_e &= \frac{1}{2} x_T \sqrt{ \frac{E_p}{E_e} } \left( e^{-\eta_1} +
e^{-\eta_2} \right) \label{e3.3b},
\end{eqalignno} \end{subequations}
with $x_T = 2 p_T / \mbox{$\sqrt{s}$}$, where \mbox{$\sqrt{s}$}\ is the $ep$ centre--of--mass energy; note
that in our convention positive rapidities correspond to the direction of the
proton. Finally, the subprocess cross--sections $\hat{\sigma}$ for the direct
\cite{53,50} and resolved photon \cite{54} contributions depend on the
Mandelstam variables describing the hard partonic scattering, with $\hat{s} =
x_e x_p s$ and
\begin{equation} \label{e3.4}
\hat{t} = - \frac{\hat{s}}{2} \left( 1 \pm \sqrt{ 1 - \frac {4 p_T^2}
{\hat{s}} } \right);
\end{equation}
both solutions in eq.(\ref{e3.4}) have to be included when evaluating
eq.(\ref{e3.1}), which can be accomplished by simply symmetrizing the
subprocess cross--sections under $\hat{t} \leftrightarrow \hat{u}$.
Direct and resolved photon contributions to the photoproduction of jets were
first compared by Owens \cite{50} in 1979 for fixed target energies, where
the resolved photon contributions were found to be subdominant, except at very
small $p_T$. Although a few theoretical analyses \cite{55a,55} of the
photoproduction of jets appeared in the first half of the 1980's, the
importance of resolved photon contributions was fully appreciated only in
1987, when it was realized \cite{49,56b,56} that at HERA energies, they could
exceed the direct contribution by as much as a factor of ten at $p_T \simeq 5$
GeV, and remained dominant out to $p_T \simeq 35$ to 40 GeV.
An update \cite{56a} of this result is shown in Fig.~8, where the ratio of
resolved photon and direct contributions to the single jet inclusive
cross--section\footnote{Recall that by definition, events with two accepted
jets count twice here.} is plotted for the nominal HERA energy $\mbox{$\sqrt{s}$} = 314$
GeV. Unlike in ref.\cite{56}, we have imposed some acceptance cuts, taken from
a recent ZEUS analysis \cite{57}: The jet has to fall in the pseudorapidity
range $-1 \leq \eta_{\rm jet} \leq 2$, and the ``antitag" requirement that
the outgoing electron is not seen in the main detector implies that the photon
virtuality $Q^2 \leq 4$ GeV$^2$. Most results in Fig.~8 have been obtained
using the MRSD-' parametrization \cite{58a} for the parton densities in the
proton; comparison between the solid and the dotted curve, which is for the
MRSD0' parametrization, shows that even the pre--HERA uncertainty of nucleon
densities only leads to an uncertainty of a few percent in the ratio of
Fig.~8, except at very small $p_T$. HERA data on deep--inelastic scattering
have since then improved our knowledge of the nucleon structure considerably;
the impact of these data on parametrizations of $G^p(x)$ is still under
investigation \cite{60}, but it is already clear that very soon the
uncertainty from the parton densities in the nucleon will be negligible
compared to the differences between predictions based on various
parametrization of \mbox{$\vec q^{\gamma} $}. In particular, HERA data clearly favour MRSD-' over
MRSD0'; we therefore take the former as our standard choice.
We see that implementation of the acceptance cuts reduces the region where
resolved photon contributions are dominant to $p_T \leq 25$ to 30 GeV; this is
mostly due to the upper limit on $\eta_{\rm jet}$, which reduces the resolved
photon contribution much more than the direct one (see below). Nevertheless,
the former still exceeds the latter by a factor between 5 and 11 at $p_T = 5$
GeV; at present our lack of knowledge of \mbox{$\vec q^{\gamma} $}\ does not allow us to predict
this ratio more precisely. This dominance of resolved photon contributions at
small $p_T$ can partly be explained by the fact that they get contributions
from gluon exchange in the $t-$channel, see Fig.~5b; this enhances the squared
matrix elements for resolved photon processes by a factor $\hat{s}/|\hat{t}| >
2$, compared to those for direct processes. Colour factors generally also
favour the former over the latter. Finally, we saw in Figs.~4 that for small
\mbox{$x_{\gamma}$}, the parton densities in the photon can actually exceed $1/\mbox{$\alpha_{\rm em}$}$
substantially; for small $x_e$, which contribute only at small $p_T$, the
integral in eq.(\ref{e3.2}) can therefore enhance resolved photon
contributions even further.\footnote{Recall that the factor of \mbox{$\alpha_{\rm em}$}\ contained
in \mbox{$\vec q^{\gamma} $}\ is cancelled in the ratio by the explicit factor of \mbox{$\alpha_{\rm em}$}\ appearing
in the $\hat{\sigma}$ for direct processes.} On the other hand, this
convolution integral decreases more rapidly with increasing $x_e$ than the
photon flux factor \mbox{$f_{\gamma|e}$}\ does; this explains the more rapid decrease of
resolved photon contributions with increasing $p_T$, and thus the shape of the
curves in Fig.~8.
Obviously parametrizations with sizable \mbox{$G^{\gamma}$}\ (LAC1, WHIT4) predict a
considerably larger resolved photon contribution at small $p_T$ than those
with smaller \mbox{$G^{\gamma}$}\ [WHIT1, WHIT3; DG, GRV, AGF and GS2 (not shown) also belong
to this class]. Note that the prediction from the WHIT1 parametrization
exceeds that using WHIT3 even at the smallest $p_T$ shown; this indicates
that the $p_T$ distribution integrated over jet rapidities is not sensitive to
very small \mbox{$x_{\gamma}$}, where the WHIT3 gluon density exceeds that of WHIT1, see
Fig.~4b. Finally, we should warn the reader that additional experimental cuts
can change the ratio of direct and resolved photon contributions
substantially. This is demonstrated by the dot--dashed curve, which has been
obtained with the same parametrizations of parton densities in the proton and
photon as the solid curve, but where we have demanded that the outgoing
electron be detectable in the ZEUS luminosity monitor; this implies $Q^2 <
0.01$ GeV$^2$ and, more importantly, $0.2 \leq z \leq 0.75$. The lower cut on
the scaled photon energy $z$ greatly reduces direct contributions at low
$p_T$, while the upper cut reduces resolved photon contributions at high $p_T$
more than direct ones; as a result, the $p_T-$dependence of the ratio of the
two contributions becomes considerably steeper.
In Fig.~9 we further split the resolved photon (9a) and direct (9b)
contributions depending on whether the two high$-p_T$ partons in the final
states are of two quarks, two gluons, or a quark and a gluon. Since the
contributions from $\mbox{$q \bar{q}$} \rightarrow gg$ and $gg \rightarrow \mbox{$q \bar{q}$}$ are very small, the
curves in Fig.~9a can also be read as coming from $qq, \ gg$ and $qg$ initial
states, while in Fig.~9b the $qg$ and $qq$ final states come from $\gamma q$
and $\gamma g$ initial states, respectively. We have used the WHIT1
parametrization for \mbox{$\vec q^{\gamma} $}, and applied the same acceptance cuts as in Fig.~8.
WHIT1 assumes a relatively small \mbox{$G^{\gamma}$}; as a result, the $gg$ final state is
dominant only for $p_T \leq 5$ GeV. However, even the WHIT4 parametrization,
which assumes a two times larger input distribution $G_0^{\gamma}$, predicts
the cross--section for the $gg$ final state to be well below that for the $qg$
final state for $p_T > 10$ GeV. Note that in the majority of $qg$ events the
quark comes from the photon and the gluon from the proton; this is partly
because the photon is assumed to be relatively poor in gluons, and partly
because the cut $\eta_{\rm jet} \leq 2$ removes many events with a gluon in
the photon in the initial state, as discussed below. We saw in sec.~2 that the
quark densities in the photon are much better known than \mbox{$G^{\gamma}$}, except at small
\mbox{$x_{\gamma}$}. Together with the result of Fig.~9a, this explains the rapid
convergence of the curves in Fig.~8, although some difference between the LAC1
and WHIT predictions persists even at large $p_T$ (see also
Fig.~3\footnote{Notice, however, that all quark flavours contribute equally to
the jet cross--section, while contributions to \mbox{$ F_2^\gamma $}\ are weighted by the
squared charge. Two parametrizations can therefore have very similar \mbox{$ F_2^\gamma $}\
and yet lead to different predictions for quark--initiated jet production at
HERA.}). Finally, the comparison of Figs.~9a and 8 shows that two quark final
states dominate resolved photon contributions only for values of $p_T$ where
the total jet cross section is already dominated by direct contributions.
As expected from Fig.~8, the direct contributions shown in Fig.~9b have a
considerably flatter transverse momentum spectrum than the resolved photon
contributions. Unfortunately, photon--gluon fusion (the two quark final state)
dominates the direct contribution only at relatively small $p_T$; this is
because the gluon density in the proton is softer (i.e., decreases more
rapidly with increasing $x_p$) than the valence quark distributions. Since at
small $p_T$ the inclusive jet cross section is dominated by resolved photon
contributions, a direct study of photon--gluon fusion, which might allow to
further constrain the gluon density in the proton even at rather small $x_p$
\cite{58}, will be difficult unless the resolved photon contribution can be
suppressed by additional cuts.
As shown in Figs.~10, the (pseudo)rapidity distribution of the jet can be
used to help disentangle direct and resolved photon contributions. In Fig.~10a
we have used the present HERA energy $\mbox{$\sqrt{s}$}= 296$ GeV, and applied the same cuts
on the parton level that the ZEUS collaboration applied \cite{57} on their
reconstructed jets. Unfortunately the $p_T$ cut chosen is still too low to
allow a direct comparison between partonic and jet cross sections, since a
substantial part of the jets still comes from the ``underlying event" (beam
fragments, initial state radiation, and possibly multiple interactions
producing minijets \cite{59,210}; see sec.~5); these effects are expected to
become more important as one approaches the proton beam direction, which
corresponds to $\eta = + \infty$ in our convention, and can thus distort the
jet rapidity distribution compared to the parton--level distribution shown in
Figs.~10. We nevertheless expect that the {\em differences} between the curves
shown in these figures will not be washed out by the underlying event.
We see that in this single--differential cross section the direct contribution
is always subdominant; however, one can enhance its relative importance by
requiring $\eta_{\rm jet} < 0$. Of more interest for us is the opposite region
of positive, sizable $\eta_{\rm jet}$, which is sensitive to small $x_e$, see
eq.(\ref{e3.3b}), and thus allows to probe the parton densities in the photon
at small \mbox{$x_{\gamma}$}, see eq.(\ref{e3.2}). Indeed, the figure shows that in this
region one is most sensitive to differences between the various
parametrizations of \mbox{$\vec q^{\gamma} $}. We also observe again that these differences are
much larger than even the pre--HERA uncertainty from nucleonic structure
functions.
We have already noted above that the jet rapidities only depend on the product
$x_e = z \cdot \mbox{$x_{\gamma}$}$; events with large $\eta_{\rm jet}$ can come from
soft partons in hard photons (small \mbox{$x_{\gamma}$}, large $z$), but also from hard
partons in soft photons (large \mbox{$x_{\gamma}$}, small $z$). Fortunately, ZEUS has
demonstrated \cite{57} the ability to (approximately) determine $z$ purely
from the longitudinal momenta of particles in the main detector, without
having to detect the outgoing electron in the luminosity monitor, which would
reduce the accepted event rate by about a factor of four. This should allow
one to increase the lower cut on $z$ from 0.2 to 0.5, which enhances the
relative importance of contributions with small \mbox{$x_{\gamma}$}, so that the differences
between predictions based on different ans\"atze for \mbox{$\vec q^{\gamma} $}\ become larger,
as illustrated in Fig.~10b. (H1 prefers to only use photoproduction events
where the outgoing electron is tagged in the small--angle detector; $z$ can
then be determined from its energy.) Although this stronger cut on $z$ reduces
the resolved photon contribution by slightly more than a factor of two for
$\eta_{\rm jet} \geq 1$, it should still enhance the discriminative power of
this measurement, given that the ZEUS data sample \cite{57} with the looser
cuts contains almost 20,000 events with reconstructed jets.
The integration over the rapidity of the second jet in single--jet inclusive
cross sections leads to a substantial spread in $x_e$ and $x_p$, see
eqs.(\ref{e3.3})\footnote{The integration over $p_T$ is less important here,
since most events will have $p_T \simeq p_{T, {\rm cut}}$ anyway.}; in order
to further increase the sensitivity of the jet rate to the region of small
\mbox{$x_{\gamma}$}\ in general and the gluon density in the photon in particular, one
therefore has to study more differential cross--sections. In Figs.~11a,b we
show predictions for the triple--differential di--jet cross section, as given
by eq.(\ref{e3.1}), for $\mbox{$\sqrt{s}$}=314 \ {\rm GeV}, \ p_T = 10$ GeV and $\eta_1 =
\eta_2 \equiv \eta$. As in Fig.~9a we display resolved photon contributions
with different final states separately, as predicted from the WHIT1
parametrization. In Fig.~11a we have applied the antitag cut $Q^2 \leq 4$
GeV$^2$ on the virtuality of the photon, but we have not restricted the
allowed range of the scaled photon energy $z$. We see that the direct
contribution now dominates at $\eta < 0$, and remains sizable even for $\eta =
2$; in the region of negative $\eta$ it is mostly due to photon--gluon fusion.
Note that our choice $\eta_1 = \eta_2$ implies $\hat{t}=\hat{u} = - \hat{s} /
2$, which minimizes the dynamical enhancement factor $\hat{s} / |\hat{t}|$ of
most resolved photon contributions. Together with the slightly larger value of
$p_T$ this explains why direct contributions appear more prominent in Fig.~10a
than in Fig.~9a.
Contributions from the gluon in the photon peak at $\eta \simeq +2$. The
reason is that a gluon usually only carries a rather small fraction of the
photon energy, see Fig.~4a; for given $\hat{s} \geq 4 p_T^2$, a large
contribution to the total energy in the hard sub--process then has to come
from the proton, leading to a rather strong boost of the high$-p_T$ partons in
the proton direction. The quarks in the photon can be substantially more
energetic, and therefore already contribute at $\eta \simeq 0$.
Although according to the WHIT1 parametrization, the contribution from gluons
in the photon is clearly enhanced at large $\eta$, at $\eta=2$ slightly more
than half of the total cross--section still comes from quarks in the photon or
direct processes. As before, the sensitivity to the region of small \mbox{$x_{\gamma}$}\ can
be further enhanced by imposing cuts on the scaled photon energy $z$. This is
illustrated in Fig.~11b, where we have required $0.3 \leq z \leq 0.8$. The
lower limit has been chosen such that the direct contribution vanishes
identically for $\eta > 0$, see eq.(\ref{e3.3b}). (Recall that $x_e=z$ for
direct contributions, while $x_e = z \mbox{$x_{\gamma}$}$ otherwise.) Moreover, the relative
importance of contributions from gluons in the photon is clearly enhanced
compared to Fig.~11a; in particular, the $gg$ final state is now dominant for
$\eta \simeq 2$, where only $\sim 20 \%$ of the total cross section is
predicted to come from quarks in the photon. Recall that \mbox{$G^{\gamma}$}\ might well be
larger than assumed in the WHIT1 parametrization, so that the contributions
from gluons in the photon might be even larger. It should therefore be
possible to derive stringent constraints on \mbox{$G^{\gamma}$}\ even using jets with
relatively large $p_T$, by focussing on events with large rapidity and large
$z$. Of course, the sensitivity to resolved photon contributions involving
soft partons in the photon is even larger \cite{60a} if one can measure a
cross--section for fixed $z$; however, this will need a rather large event
sample.
If $z$ and both jet rapidities have been measured, one can in principle
reconstruct both $x_p$ and \mbox{$x_{\gamma}$}, using eqs.(\ref{e3.3}). In practice, HERA
experiments use an estimator for \mbox{$x_{\gamma}$}\ to separate the events into ``direct''
and ``resolved'' samples; this estimator reduces to \mbox{$x_{\gamma}$}\ in LO QCD, but only
uses the measured rapidities and transverse energies of the jets, and can thus
also be computed in NLO (where it will differ from the partonic \mbox{$x_{\gamma}$}). The
ZEUS collaboration \cite{61} has used 284 reconstructed di--jet events from
the first run of HERA to show that the \mbox{$x_{\gamma}$}\ distribution has a peak near
$\mbox{$x_{\gamma}$} = 1$, as expected from direct contributions. The H1 collaboration
\cite{62} has gone one step further and subtracted the contribution from
quarks in the photon, estimated using the GRV parametrization. They find
evidence for a non--vanishing contribution from gluons in the photon only for
$\mbox{$x_{\gamma}$} \leq 0.2$, thereby ruling out the LAC3 parametrization (see Fig.~4a).
H1 reconstructs \mbox{$G^{\gamma}$}\ in the range $0.02 \leq \mbox{$x_{\gamma}$} \leq 0.2$ at an effective
scale $\mu^2 \simeq 60$ GeV$^2$; the result is in good agreement with GRV, but
disfavours LAC1. However, a fairly sophisticated Monte Carlo analysis was
necessary to extract \mbox{$G^{\gamma}$}\ even using the LO formalism. For example, we saw in
the discussion of Fig.~10a that parts of the ``underlying event" contribute to
the reconstructed jets, which obscures the relation between hard partons and
jets. The H1 analysis \cite{62} depends on such details quite sensitively. It
might therefore be somewhat premature to exclude the LAC1 parametrization on
the basis of this evidence alone.
The ZEUS collaboration has also published \cite{62b} an updated jet analysis,
based on 12,000 reconstructed di--jet events with $E_T($jet$) \geq 6$ GeV.
They study $d \sigma / d \overline{\eta}$ for 4,000 events with $|\eta_1 -
\eta_2| \leq 0.5$, where $\overline{\eta}$ is the average pseudorapidity of
the two jets; this quantity is closely related to the double differential
cross section $d^2 \sigma / (d \eta_1 d \eta_2)$ at $\eta_1 = \eta_2$. They
find reasonable agreement between their data and LO QCD predictions in the
``direct" sample (events with large measured \mbox{$x_{\gamma}$}), but most parametrizations
of \mbox{${q^{\gamma} } $}\ give too low predictions for the ``resolved" sample. This again
indicates that at this rather low value of $E_T$ the ``underlying event" plays
quite an important role; as we will discuss in more detail in Sec.~5, this
aspect is not well described by present standard QCD Monte Carlo event
generators.
\setcounter{footnote}{0}
Direct and resolved photon contributions are also expected to have different
distributions in $\cos \!\theta^*$, where $\theta^*$ is the cms scattering
angle \cite{62a}. Due to diagrams where a gluon is exchanged in the $t-$ or
$u-$channel, resolved photon contributions are more strongly peaked at small
$\theta^*$ than direct contributions. The increasing importance of the
latter over the former at higher $p_T$ means that the $\cos \! \theta^*$
distribution of di--jet events with large $p_T$, will be flatter than at low
$p_T$.
As remarked earlier, the most obvious distinction between direct and resolved
photon events is that only the latter contain a photonic spectator (or
remnant) jet, see Fig.~6. In Fig.~12 we show the average energy of this jet in
the lab frame, as predicted from the WHIT1 parametrization; on the parton
level, this energy is simply given by $E_e \cdot z \cdot (1-\mbox{$x_{\gamma}$})$. For
negative $\eta$ (of the high$-p_T$ jets), the spectator jet is rather soft,
since both $z$ and \mbox{$x_{\gamma}$}\ have to be large, so that $1-\mbox{$x_{\gamma}$}$ is quite small.
For $\eta \simeq 0$, events with a gluon from the photon usually have quite
large $z$, but moderate \mbox{$x_{\gamma}$}, yielding a high spectator jet energy; it
declines at large $\eta$ since the average $z$ becomes smaller.
Quark--initiated events typically have considerably larger \mbox{$x_{\gamma}$}\ and hence a
softer photonic spectator jet. Obviously the average spectator jet energy will
increase (decrease) if a lower (upper) cut on $z$ is applied.
In principle the results of Fig.~12 offer another possibility to enhance the
contributions from gluons in the photon, by requiring the presence of an
energetic remnant jet in the electron direction. However, in practice the
energy of this jet cannot be measured very accurately, since some part of it
will usually be lost in the beam pipe. It should nevertheless be emphasized
that the first analysis of jet data \cite{51} taken during the HERA pilot run
found substantial energy deposition in the backward calorimeter even if all
high$-p_T$ jets have positive rapidities; this can be understood only if the
photonic remnant jet is included in the MC simulation. This jet has recently
been studied in more detail by the ZEUS collaboration \cite{62c}; we will
discuss their results in Sec.~5.
Recently new data from the Fermilab fixed target photoproduction experiment
E683 have been published \cite{63}. It uses a tagged photon beam with mean lab
energy of 260 GeV, giving a mean \mbox{$\sqrt{s}$}\ of slightly over 20 GeV. A Monte Carlo
analysis suggest that their di--jet sample, required to have two reconstructed
jets with average $p_T > 4$ GeV, gets approximately equal contributions from
direct and resolved photon contributions, in agreement with theoretical
expectations. However, unlike at HERA, no direct experimental evidence for the
existence of resolved photon contributions could be established (other than
the overall event rate). In particular, the hadronic energy flow in the very
forward direction was found to be quite similar for direct and resolved photon
events, and even for di--jet events produced from a pion beam; this somewhat
counter--intuitive result can be explained in terms of jet fluctuations.
So far all our predictions have been computed in LO in QCD. As well known, the
overall normalization of such predictions is uncertain, since in the leading
log approximation used here, one cannot with certainty determine the values of
the factorization scales in the parton distribution functions and the
renormalization scale appearing in the running QCD coupling constant. These
scales have no physical significance; predictions would be independent of them
if all orders of perturbation theory could be summed. One therefore expects
reduced scale dependence already in next--to--leading order (NLO). Moreover,
many quantities can be predicted meaningfully only if one allows at least
three high$-p_T$ partons in the final state; these include the dependence of
jet cross--sections on the jet definition, and the distribution in the
transverse opening angle between the two (hardest) jets in events with (at
least) two jets.
The first step towards a full NLO calculation of jet photoproduction was taken
in 1980 with a calculation \cite{64} of direct $2 \rightarrow 3$ cross
sections (e.g. $\gamma q \rightarrow qgg$ and $\gamma g \rightarrow \mbox{$q \bar{q}$}
g$). A first complete NLO calculation of the direct contribution, including
virtual (1--loop) corrections, was performed \cite{65} in 1986. These results
were applied to jet production at fixed target energies in \cite{66}, and at
HERA energies in \cite{67}. An NLO prediction of resolved photon contributions
to jet production become possible only after corrections to the hard partonic
QCD cross--sections had been computed \cite{68}. These results were applied to
jet production from resolved photons at HERA in ref.\cite{69}. Finally, in
refs.\cite{70,71,72,73}, complete NLO calculations for single--jet inclusive
jet cross--sections were presented, including both direct and resolved photon
contributions.
Typical results are presented in Fig.~13, adapted from B\"odeker et al.
\cite{72}. We show the scale dependence of the predicted jet cross--section
at HERA for $E_T=25$ GeV and $\eta_{\rm jet} = 1.5$. Note that $E_T$ and $p_T$
are in general no longer identical in NLO, since now a jet might be made up of
two partons. In Fig.~13 two partons have been merged into a single jet if
$\Delta R \equiv \sqrt{ \left( \Delta \eta \right)^2 + \left( \Delta \phi
\right)^2 } \leq 0.7$.
The solid and short dashed curves have been obtained by setting all three
scales (the renormalization scale, and the factorization scales for the
photonic and nucleonic parton densities) equal to each other. As expected, the
LO prediction exhibits a much stronger scale dependence than the NLO result.
It is worth noting, however, that for the ``natural" choice $\mu = p_T$ (or
$E_T$ in NLO), LO and NLO predictions almost coincide. Generally the
difference between LO and NLO predictions was found to be quite small at
HERA energies if $\mu = p_T$ has been chosen and jets are defined with
$\Delta R$ between about 0.7 and 1.\footnote{There appears to be some
discrepancy between the results of ref.\cite{71} and the earlier calculations
\cite{69,70}; this is now being sorted out (M. Greco, private communication).}
This means that the results of Figs.~8--12 should not be affected too much by
NLO corrections.
As already discussed in the beginning of this section, in NLO the distinction
between direct and NLO contributions is blurred. Diagrams like the one shown
in Fig.~7 contribute to direct jet production in NLO, but they also contain
a logarithmically divergent piece which has already been included in the LO
resolved photon contribution; this piece therefore has to be subtracted from
the NLO direct contribution. This subtraction term grows logarithmically with
the photonic factorization scale $M_{\gamma}$, which is also the scale
appearing in the photonic parton distribution functions. In NLO the direct
contribution (dotted curve) therefore decreases with increasing $M_{\gamma}$,
while the resolved photon contribution (dot--dashed curve) increases. The sum
of the two (long dashed) is nearly independent of $M_{\gamma}$; the dependence
does not cancel completely since in NLO the subtraction term is exactly
proportional to $\log M_{\gamma}$, while the photonic parton densities only
increase approximately like $\log M_{\gamma}$, as discussed in Sec.~2.
We saw above that di--jet cross sections that are differential in both jet
rapidities are more powerful discriminators of photonic parton densities than
single--jet inclusive cross sections. Unfortunately, in conventional NLO
calculations going from single--jet to di--jet cross sections introduces
considerable complications;\footnote{This step should be much easier using the
Monte Carlo method of ref.\cite{66}.} as a result, only the direct production
of jet pairs has been treated in NLO to date \cite{74}. The urgent need for a
full calculation is emphasized by the recent ZEUS data \cite{62b}, which have
the potential to discriminate between different parametrizations of photonic
parton densities once the cross section can be predicted with some confidence.
Recently the calculation of jet rates at HERA has been further refined by
including two additional effects. First, eq.(\ref{e3.3a}) shows that (direct)
jet production at negative rapidities probes the proton structure at quite low
values of $x_p$; for $p_T \leq 10$ GeV, partons with $x_p$ as low as $10^{-3}$
contribute. Small$-x$ effects may then become important, and one may have to
use \cite{75} the so--called $k_T$ factorization \cite{76}. Secondly, if one
defines the photoproduction sample with a rather moderate (no--tag) cut on the
outgoing electron, one includes contributions where the photon virtuality
$Q^2$ may not be entirely negligible; recall, for example, that the ZEUS cuts
\cite{57} include events with $Q^2$ up to 4 GeV$^2$. On the other hand, when
using the simple Weizs\"acker--Williams approximation (\ref{e3.2}), one
assumes $Q^2$ to be small compared to all other scales in the problem. This is
still a good approximation for the direct contribution in this case, but the
parton densities in the photon become suppressed \cite{29,77} once $Q^2 >
\Lambda^2_{\rm QCD}$. The simple factorization (\ref{e3.2}) then breaks down,
but one can still define a ``parton density in the electron", which will
depend on the experimental cut on $Q^2$ \cite{78}. For the ZEUS cuts, this
suppression only amounts to a few percent.
Finally, as mentioned in the discussion of Fig.~10, so far jet production at
HERA could only be investigated experimentally at rather moderate transverse
momenta, where the ``underlying event" can still contribute significantly to
the reconstructed jets. The influence of the underlying event might be less
problematic when one studies the production of high$-p_T$ particles
\cite{79,65} rather than jets. On the other hand, one now has to specify not
only parton distribution functions, but also fragmentation functions, before
a prediction can be made. Also, the cross--section falls off very rapidly with
$p_T$, forcing one to work at scales of only a few GeV where the convergence
of the perturbative expansion might be rather slow. Moreover, one either has
to experimentally identify different particle species (chiefly pions and
kaons, if only charged particles are counted), or make assumptions about the
relative abundances of these species. Recent measurements of the
cross--section for the production of charged particles with $p_T \geq 1.5$
GeV, by both the H1 and ZEUS collaborations \cite{80}, find good agreement
with a theoretical NLO prediction \cite{81} as far as the $p_T$ spectrum is
concerned; the pseudorapidity distribution measured by H1 is less well
described, but the experimental errors do not allow to draw a definite
conclusion at this point.
\setcounter{footnote}{0}
\subsection*{3b) The Photoproduction of Heavy Quarks}
The production of heavy quarks offers two theoretical advantages over the
production of light partons (jets) discussed in the previous subsection.
First, their large mass $m_Q \gg \mbox{$\Lambda_{\rm QCD}$}$ ensures that QCD perturbation theory
is applicable in all of phase space, although nonperturbative corrections
$\propto \left( \mbox{$\Lambda_{\rm QCD}$} / m_Q \right)^{n\geq 1}$ might not be negligible for
charm quarks. In particular, the total cross--section without any cuts could
now be predicted with some reliability if the values of certain parameters
($m_Q, \ \mbox{$\Lambda_{\rm QCD}$}$ and the parton distribution functions) were known precisely.
Secondly, at least in leading order the number of contributing partonic
processes is much smaller, making heavy quark production easier to analyze.
Specifically, in LO only photon--gluon fusion contributes to \mbox{$Q \bar{Q}$}\
production from direct photons, while the relevant resolved photon processes
are $gg$ fusion and light \mbox{$q \bar{q}$}\ annihilation.\footnote{It can be argued that
at very high transverse momentum, $p_T \gg m_Q$, $\alpha_s \log p_T/m_Q$
should be counted as ${\cal O}(1)$, rather than ${\cal O}(\mbox{$\alpha_s$})$. In this
case the ``excitation" processes $Qg \rightarrow Qg$ and $Qq \rightarrow Qq$ also contribute
in LO, since the $Q-$quark density in the photon grows logarithmically with
the hard scale of the process. In these reactions the heavy quark jet is
balanced by a light quark or gluon jet, while in \mbox{$Q \bar{Q}$}\ creation events two
heavy quark jets occur with equal and opposite $p_T$. These ``flavour
excitation" contributions are now under study \cite{82}.}
In LO, the \mbox{$Q \bar{Q}$}\ production cross--section in $ep$ scattering is still
given by eqs.(\ref{e3.1})--(\ref{e3.3}). However, $\eta_{1,2}$ now have to be
interpreted as true rapidities, which differ from the pseudorapidity for
massive particles; moreover, $x_T$ in eqs.(\ref{e3.3}) is now given by $2
\sqrt{ p_T^2 + m_Q^2 } / \mbox{$ \sqrt{s} $ }$, and eq.(\ref{e3.4}) has to be replaced by
\begin{equation} \label{e3.5}
\hat{t} = \frac {\hat{s}} {2} \left[ \frac {2 m_Q^2} {\hat{s}} - 1 \pm
\sqrt{ 1 - \frac {4 \left( m_Q^2 + p_T^2 \right)} {\hat{s}} } \right].
\end{equation}
The partonic cross--sections $\hat{\sigma}$ for $\gamma g \rightarrow \mbox{$Q \bar{Q}$}$ and
$gg,\mbox{$q \bar{q}$} \rightarrow \mbox{$Q \bar{Q}$}$ can be found in refs.\cite{83} and \cite{84,84a},
respectively.
Resolved photon contributions to heavy quark production were first treated in
ref.\cite{39}, for the case of the top quark, whose mass was then believed to
be in the vicinity of 40 GeV. We now know \cite{85} that the top quark is too
heavy to be produced at HERA, but \mbox{$c \bar{c}$}\ and \mbox{$b \bar{b}$}\ pairs will be produced
copiously. It was pointed out in refs.\cite{49,86} that (for most
parametrizations of \mbox{${q^{\gamma} } $}) resolved photon contributions to the total
cross--sections are subdominant, but not negligible; e.g., they amount to
$\sim 20\%$ for the DG parametrization. Resolved photon contributions to the
production of heavy quark pairs are therefore considerably less important than
for high$-p_T$ jet production. The reason is that now the resolved photon
processes also only involve gluon exchange in the $s-$channel or (heavy) quark
exchange in the $t-$ or $u-$channel; there is no enhancement factor
$\hat{s}/|\hat{t}|$, unlike for jet production. Moreover, the two
sub--processes involving the parton content of the photon have rather small
colour factors, again in contrast to the matrix elements appearing in resolved
photon contributions to jet production. These analyses also showed that $gg$
fusion is predicted to dominate over \mbox{$q \bar{q}$}\ annihilation, by a factor of 10
(3.5) in case of the DG parametrization and \mbox{$c \bar{c}$}\ (\mbox{$b \bar{b}$}) production. The
resolved photon contribution therefore offers a good opportunity to constrain
\mbox{$G^{\gamma}$}, while (in LO) the direct contribution is proportional to $G^p$.
We saw in the previous subsection that at HERA, contributions from gluons
in the photon are most important at sizable, positive rapidities, but are
suppressed at negative rapidity. The rapidity distribution therefore offers
a good handle for separating the two contributions to \mbox{$Q \bar{Q}$}\ production. In
Fig.~14 we show the $p_T$ spectrum of $c$ and $b$ quarks at central rapidity,
$y_1 = y_2 = 0$. We see that the resolved photon contribution (difference
between solid and dashed curves) only amounts to 10--15\% at small $p_T$, and
becomes entirely negligible for $p_T > 5$ GeV. In fact, as shown in
ref.\cite{56}, the resolved photon contribution to heavy quark production at
HERA can always be suppressed to an insignificant level by requiring (at
least) {\em one} of the two heavy quarks to emerge at $y<0$.\footnote{This
result does not hold for the LAC3 parametrization; fortunately, this
parametrization is excluded by other data, as discussed in secs.~3a and 4a.}
Note that we have chosen the WHIT4 parametrization in Fig.~14, which is
characterized by a rather large and hard \mbox{$G^{\gamma}$}; it predicts that resolved
photon contributions amount to more than 20\% (30\%) of the total \mbox{$c \bar{c}$}\
(\mbox{$b \bar{b}$}) cross--section at HERA. This contribution is more important for
\mbox{$b \bar{b}$}\ production, since the direct contribution is suppressed by the small
charge of the $b$ quark; for parametrizations with rather hard \mbox{$G^{\gamma}$}\ (WHIT1,4,
DG, GS2, GRV) this suppression is stronger than the relative reduction of the
resolved photon contributon with increasing $m_Q$, which is caused by the
additional convolution (\ref{e3.2}) with the gluon density in the photon.
Since resolved photon contributions to \mbox{$Q \bar{Q}$}\ production are insignificant at
high $p_T$ (unless their importance is enhanced by specific cuts, as discussed
below), the ratio of \mbox{$c \bar{c}$}\ to \mbox{$b \bar{b}$}\ cross--sections in the region $p_T
\geq 10$ GeV simply reflects the ratio of their squared charges. This relative
suppression of the \mbox{$b \bar{b}$}\ cross--section means that $b-$tagging at HERA will
be significantly more difficult than it is at \mbox{$p \bar{p}$}\ colliders, where \mbox{$b \bar{b}$}\
and \mbox{$c \bar{c}$}\ cross--sections become equal at high $p_T$. At \mbox{$p \bar{p}$}\ colliders
the harder fragmentation function of $b-$flavoured hadrons \cite{87} means
that inclusive high$-p_T$ muon production is dominated by \mbox{$b \bar{b}$}\ events; this
will not be true at HERA, however, except at very high $p_T$ where the
cross--section is quite small. Charm quarks will also be a serious background
to $b-$tagging by micro--vertex detectors at HERA.
As mentioned earlier, the resolved photon contribution is concentrated at
positive rapidities. Just as in the case of jet production it can be isolated
by a cut on the scaled incident photon energy $z$; e.g., requiring $z>0.3$ at
$p_T=10$ GeV removes all direct contributions with $y_1=y_2>0$. Fig.~15 shows
that the remaining resolved photon contribution is indeed very sensitive to
the gluon content of the photon. Even according to the WHIT1 parametrization
the \mbox{$q \bar{q}$}\ annihilation contribution (short dashed) is considerably below the
one from $gg$ fusion (long dashed). The WHIT4 parametrization therefore
predicts a considerably larger cross--section, but the shape of the rapidity
distribution is similar to that predicted from WHIT1. In contrast, the LAC1
prediction differs in both normalization and shape. However, we should warn
the reader that without the cut on $z$, even at $y_1 = y_2 = 2$ the direct
contribution would be at least ten times larger than the resolved one; the
experimental implementation of this cut therefore has to be very efficient. It
might even be necessary to require the presence of a photonic remnant jet to
extract the resolved photon contribution; recall that this jet is expected to
be quite energetic in events that originate from the gluon content of the
photon. Recall also that there will be a large contribution from ``charm
excitation" \cite{82} if only one of the two high$-p_T$ jets is tagged as a
heavy quark. Finally, the cross--section shown in Fig.~15 is quite small, even
though we have not yet required any specific charm signal (e.g., a hard muon
or reconstructed $D^*$ meson). Clearly HERA will have to accumulate
significantly more data than the present 6 pb$^{-1}$ (as of the end of 1994)
to measure such triple--differential cross--sections even at lower $p_T$.
The predictions shown in Figs.~14 and 15 were computed in LO. NLO calculations
of the photoproduction of heavy quarks exist \cite{86,88,89}; unlike for jet
production, even the fully differential cross--section is available in NLO
\cite{90}. It has been demonstrated \cite{91} that the direct contribution can
be extracted reliably also in NLO, i.e. the sensitivity to the gluon content
of the proton is not degraded. In view of their smaller size, extraction of
the resolved photon contribution might prove more difficult. In particular, in
direct events with an additional hard parton in the final state ($\gamma g \rightarrow
\mbox{$Q \bar{Q}$} g, \ \gamma q \rightarrow \mbox{$Q \bar{Q}$} q$) the heavy quarks can occur at large
positive rapidity even after a cut on $z$ has been applied; one may have to
veto the occurence of additional high$-p_T$ jets, and/or require the two
highest $p_T$ jets to be back--to--back in the transverse plane, in order to
efficiently suppress such backgrounds to the cross--section shown in Fig.~15.
The considerable body of data on photoproduction of charm at fixed target
energies ($\mbox{$ \sqrt{s} $ } \leq 20$ GeV) is well described by NLO QCD calculations
\cite{92}. However, at these low energies the resolved photon contribution is
quite small; it is significant only on the backward direction (opposite to the
incident photon) \cite{90}, where the experimental acceptance is poor.
Very recently, first data on charm production at HERA have become available.
The ZEUS collaboration \cite{93} searched for fully reconstructed $D^{\pm*}$
mesons. They observe a signal of $48 \pm 11$ events within the acceptance
region $p_T(D^*) > 1.5$ GeV, $|\eta(D^*)| < 1.5$; this corresponds to
$\sigma(ep \rightarrow D^{\pm*} X) = (32 \pm 7 \ {}^{+4}_{-7})$ nb at $\mbox{$ \sqrt{s} $ }=296.7$ GeV
with $Q^2 \leq 4$ GeV$^2$ and $0.15 \leq z \leq 0.86$. They then attempt to
estimate the {\em total} \mbox{$c \bar{c}$}\ cross--section from this measurement; this,
however, sensitively depends on the extrapolation of the cross--section into
kinematic regions where it has not been measured, which introduces a strong
dependence of the ``measured" \mbox{$c \bar{c}$}\ cross--section on the assumed parton
distribution functions in the proton and photon as well as on $m_c$. This not
only greatly increases the quoted (systematic) error; even the central value
depends on these assumptions. It does not make much sense to compare this
``measured" cross--section with different theoretical predictions, since
``measurement" and ``prediction" depend on the same quantities! Notice also
that the prediction for the total \mbox{$c \bar{c}$}\ cross section suffers from large
uncertainties \cite{95}. Since $\mbox{$\alpha_s$}(m_c)$ is still quite large, the
perturbative expansion only converges slowly, which manifests itself in a
rather strong dependence of even the NLO prediction on the factorization and
renormalization scales. Moreover, the prediction is very sensitive to $m_c$,
decreasing by more than a factor of three when $m_c$ is increased from 1.2 to
1.8 GeV. Finally, small$-x$ effects might be sizable \cite{95,95a}. All these
sources of theoretical uncertainties are reduced once we require the charm
quarks, or their fragmentation and decay products, to have significant
transverse momentum. One should therefore directly compare QCD predictions for
the cross--section in the experimentally accessible region with the data.
The same remarks also apply to the as yet preliminary analysis of charm
production by the H1 collaboration \cite{94}, which is based on events with a
hard muon. They find 484 events where at least one muon satisfies $p_T(\mu) >
1.5$ GeV and $30^{\circ} \le \theta(\mu) \le 130^{\circ}$; some 280 of these
events are expected to contain fake muons, or muons from $\pi$ and $K$ decays.
This gives an accepted cross--section $\sigma(ep \rightarrow \mu^{\pm} X) = (2.03
\pm 0.43 \pm 0.7)$ nb; about 95\% of this signal is expected to come from
\mbox{$c \bar{c}$}\ events, the rest coming from \mbox{$b \bar{b}$}\ production.
We attempted to reproduce the cross--sections measured by the ZEUS and H1
collaborations with a parton--level MC generator based on LO QCD expressions.
We take the renormalization and factorizations scales to be $\sqrt{m_c^2 +
p_T^2}$ and $m_c=1.6$ GeV; as mentioned earlier, the $p_T$ cuts greatly
reduce the sensitivity to $m_c$. However, these cuts also introduce an
additional difficulty in the theoretical treatment. They are sufficiently high
so that fragmentation effects will play a role. On the other hand, $p_T$
cannot safely be assumed to be much larger than $m_c$ here, so factorizing the
result into a hard production cross section and a fragmentation function may
not yet be a good approximation. We therefore ran our MC programs with two
different options, using the standard Peterson et al. fragmentation functions
\cite{87} or no fragmentation at all. In the former case we also include
contributions from the charm in the photon ($qc \rightarrow qc$ and $gc \rightarrow
gc$);\footnote{There is also a very small contribution of this kind from the
charm in the proton, which we neglect.} after all, the use of both
fragmentation and structure functions rests on the factorization theorem, so
it seems reasonable to treat them symmetrically.
In our comparison with the ZEUS result we always include a factor of 0.26,
which is \cite{93} the probability for a charm quark to fragment into a
charged $D^*$ meson. We find \cite{82} that if we leave out both fragmentation
and the contribution from the charm in the photon, we can reproduce the
experimental cross section only if we assume a gluon distribution that
increases rapidly at small $x$; MRSD-' works well, while the prediction of
MRSD0' is too low by nearly a factor of two for all reasonable choices of
momentum scale. Due to the rapidity cut, the contribution from resolved photon
processes (chiefly $gg$ fusion) only amounts to 20\% or less. On the other
hand, if we include fragmentation effects in the standard way, the average
$p_T$ of the charm quarks in accepted events increases by nearly a factor of
two, thereby reducing the sensitivity to very small $x$; moreover, the total
cross section is now actually dominated by contributions involving the charm
content in the photon, chiefly $cg$ scattering. As a result, we can reproduce
the experimental cross section using either MRSD0' or MRSD-' partons in the
proton, and LAC1 or WHIT partons for the photon; the DG parametrization now
predicts a far too large cross section, since it assumes $c^{\gamma} =
u^{\gamma}$, which is manifestly a bad approximation at these rather low
momentum scales.
The H1 sample is less sensitive to the gluon in the proton at small $x$,
due to both the $p_T(\mu)$ cut (which leads to a considerably higher mean
transverse momentum for accepted charm quarks than in the ZEUS data sample),
and the requirement that $\theta(\mu) \le 130^{\circ}$. The data therefore do
not even allow to discriminate between MRSD0' and MRSD-' if we ignore both
fragmentation and the charm content of the photon. In this case the resolved
photon contribution is quite small, so that any (reasonable) combination of
photon and proton structure functions is in agreement with the data, yielding
a LO cross section of about 1.4 to 2.4 nb. If we include both charm
fragmentation and the charm content of the photon, the sensitivity to soft
gluons in the proton is reduced even further, and predictions using MRSD0'
differ from those using MRSD-' by less than 0.1 nb. The LAC1 and WHIT
parametrizations yield a predicted cross--section of about 1.5 nb, very close
to the MRSD0' prediction without fragmentation and without the contribution
from charm in the photon. The DG parametrization gives a prediction of about
2.9 nb; as already discussed, this parametrization over--estimates the charm
content of the photon, but even this high number is not inconsistent with the
experimental result.
\setcounter{footnote}{0}
\subsection*{3c) Direct Photon Production in \mbox{$\gamma p$}\ Scattering}
Interest in the ``deep--inelastic Compton process" $\mbox{$\gamma p$} \rightarrow \gamma X$ dates
back to the early days of the quark--parton model \cite{96}; among other
things, the cross--section for the simplest contributing sub--process, $\gamma
q \rightarrow \gamma q$, depends on the fourth power of the quark charge, and thus
allows an independent determination of these charges. However, very soon after
the introduction of the concept of the parton content of the photon it was
realized \cite{37b} that other subprocesses contribute to the production of
direct photons\footnote{As opposed to photons stemming from the decay of
hadrons, mostly $\pi^0$ and $\eta$ mesons.} at high $p_T$: Not only resolved
photon processes like $\mbox{$q \bar{q}$} \rightarrow g \gamma$ and $gq \rightarrow q \gamma$ have to be
included in a consistent LO QCD treatment, but also fragmentation processes,
where a high$-p_T$ parton fragments into a hard photon and an additional
(nearly collinear) jet; note that the $q \rightarrow \gamma$ and $g \rightarrow \gamma$
fragmentation functions \cite{37a} are ${\cal O}(\mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$})$, just like the
parton densities in the photon. A full LO treatment therefore has to include
direct processes like $\gamma q \rightarrow g q \rightarrow g q \gamma$ and resolved photon
processes like $q q' \rightarrow q q' \rightarrow q q' \gamma$; these processes involve the
same partonic matrix elements that appear in direct and resolved photon
contributions to the production of high$-p_T$ jets. Hence direct photon
production is in general actually more complicated to analyze theoretically
than jet production, contrary to statements often found in the literature. On
the other hand, photons are easier to study experimentally than jets; in
particular, their energy can be measured considerably more precisely.
In LO the triple differential cross--section for the production of a direct
photon in processes that do not involve parton $\rightarrow$ photon fragmentation is
still given by eq.(\ref{e3.1}); the cross--section for fragmentation processes
contain a convolution with a fragmentation function:
\begin{equation} \label{e3.2n}
\frac{d^3 \sigma_{\rm frag}(ep \rightarrow e \gamma j X)}{d p_T d \eta_\gamma d \eta_j}
= \sum_{i,j,k,l} \int \frac {d z'}{z'} f_{i|e}(x_e)
f_{j|p}(x_p) D_{k \rightarrow \gamma}(z')
\frac { d \hat{\sigma}_{ij \rightarrow kl} } {d \hat{t}}.
\end{equation}
Eqs.(\ref{e3.3}),(\ref{e3.4}) also still apply here, but for the fragmentation
contribution one has to replace $p_T$ by $p_T' = p_T/z'$. Moreover, unlike the
case of di--jet production, the final state particles are now distinguishable.
Note that eqs.(\ref{e3.3}) allow a two--fold ambiguity for $\eta_{1,2}$ in
terms of the Bjorken$-x$ variables; taking $\eta_1 \equiv \eta_\gamma$, one
has
\begin{equation} \label{e3.1n}
\eta_\gamma = \ln \left[ \sqrt{ \frac{E_p}{E_e} } \frac {x_p}{x_T} \left(
1 \pm \sqrt{ 1 - \frac {x_T^2} {x_e x_p} } \right) \right].
\end{equation}
In case of jet production one can arbitrarily fix the sign in
eq.(\ref{e3.1n}), since this only corresponds to the definition of which of
the two partons gives ``jet 1". However, in case of direct photon production
the sign in eq.(\ref{e3.1n}) is correlated with the choice of sign in
$\hat{t}$, eq.(\ref{e3.4}). In particular, if $\hat{t}$ is defined as the
momentum transfer from the incident photon, or parton in the incident photon,
to the final photon, or parton fragmenting into the final photon, taking a $+$
sign in eq.(\ref{e3.4}) means $|\hat{t}| > \hat{s}/2$, which implies that one
has to take the $+$ sign in eq.(\ref{e3.1n}); recall that we define the proton
direction has having positive rapidity.
The first quantitative estimates of cross--sections for the production of
direct photons in \mbox{$\gamma p$}\ scattering, including all sub--processes listed
above, were presented in refs.\cite{97}, using very simple ans\"atze for
parton distribution and fragmentation functions. The first partial NLO
analysis for fixed--target energies was published in ref.\cite{37a}, where
only NLO corrections to $\gamma q \rightarrow \gamma q$, as well as the direct NLO
process $\gamma g \rightarrow \mbox{$q \bar{q}$} \gamma$, were included; all other contributions
were treated in LO. This was justified by the observation that at these (low)
energies a rather modest $p_T$ cut on the outgoing photon suffices to greatly
suppress contributions involving photonic parton densities and/or
fragmentation functions. Notice that only processes involving both the parton
content of the photon and parton to photon fragmentation can proceed via gluon
exchange in the $t-$ or $u-$channel; one needs to convolute the cross section
with {\em two} additional functions, compared to simple $\gamma q \rightarrow \gamma
q$ scattering, before the $\hat{s}/|\hat{t}|$ enhancement characteristic of
these gluon exchange processes becomes available. It is therefore not
surprising that resolved photon contributions are not quite as important in
the case of high$-p_T$ direct photon production as they are for jet
production.
The analysis of ref.\cite{37a} was repeated independently, as well as extended
to HERA energies, in ref.\cite{98}. However, here the resolved photon
contributions involving fragmentation (which have comparatively soft $p_T$
spectra) were omitted, while the contribution from $\gamma g \rightarrow \gamma g$
(via a box diagram \cite{99}) was included; this last contribution was shown
to be significant in certain regions of phase space (see below), even though
it is formally of next--to--next--to--leading order (NNLO).
In contrast, the emphasis of refs.\cite{100} was on processes involving the
parton content of the photon. In particular, it was shown that a clear signal
from the gluon content of the photon (largely from $gq \rightarrow \gamma q$) can be
extracted from a sample of events with {\em fixed energy} of the incident
photon. Due to the large Lorentz boost from the parton--parton cms to the lab
frame these photons can be quite energetic if they emerge at large rapidity,
even at rather small $p_T$.\footnote{The same is true for very forward jets,
of course. However, it should be significantly easier to detect a photon at
small angle than to reconstruct a jet just a few degrees away from the proton
beam.}
A further step towards a full NLO analysis was taken in ref.\cite{101}, where
corrections to the resolved photon contributions were included; however, all
fragmentation contributions were still treated in LO. Notice that inclusion of
NLO corrections to non--fragmentation contributions are necessary to reduce
the dependence on the scale appearing in the fragmentation functions. In NLO,
fragmentation and non--fragmentation contributions mix, just like direct and
resolved photon processes do. For example, $g q \rightarrow g q \gamma$ is an NLO
correction to $g q \rightarrow q \gamma$, but also part of the LO contribution
involving $q \rightarrow \gamma$ fragmentation {\em if} the final state $q$ and
$\gamma$ are (nearly) collinear. In order to avoid double counting, this
collinear (divergent) contribution therefore has to be subtracted from the NLO
correction. This subtraction term is proportional to the logarithm of the
scale appearing in the fragmentation function. Since the subtraction term
(obviously) appears with a negative sign in the final result it largely
cancels the dependence on this ``fragmentation scale"; the cancellation is not
perfect, since the fragmentation function resums all orders of this leading
logarithm, while the subtraction term does not. On the other hand, since the
fragmentation contributions have only been treated in LO in ref.\cite{101},
a rather strong dependence on the renormalization scale appearing in \mbox{$\alpha_s$}\
remains.
In Fig.~16 we show some LO predictions for the rapidity dependence of the
direct photon cross--section, adapted from numerical results of
ref.\cite{101}. We see that the ``Born" cross--section (from $\gamma q \rightarrow
\gamma q$) peaks at negative rapidity (in the direction of the incoming
photon); even though the $1/\hat{u}$ pole of the hard matrix element favours
configurations where the final photon is emitted in the proton direction,
negative rapidities are favoured since they probe the proton at small $x$,
where the (sea) quark densities increase quickly. In the resolved photon
process $g^{\gamma} q^p \rightarrow \gamma q$ the final photon is also preferentially
emitted in the proton direction; moreover, at large positive rapidity one
becomes sensitive to the gluon density in the photon at small $x$, where it is
(presumably) large. As a result, contributions $\propto \mbox{$G^{\gamma}$}$ are expected to
dominate at large, positive $\eta_{\gamma}$, as first noted in refs.\cite{100}.
Notice that resolved photon contributions involving parton to photon
fragmentation have not been included in Fig.~16; however, we already know from
the discussion of Sec.~3a that at large positive rapidity they are also
dominated by contributions $\propto \mbox{$G^{\gamma}$}$. Finally, the box contribution,
$\gamma g \rightarrow \gamma g$, is important only at very negative $\eta_{\gamma}$,
where one probes the gluon density in the proton at small $x$.
The results of Fig.~16 have been computed using LO expressions. While NLO
corrections are significant (e.g., they change the rapidity distribution of
the direct contributons \cite{37a,98}), they do not change the conclusion that
direct photon production offers a good handle for constraining \mbox{$G^{\gamma}$}. More
recently this conclusion has been challenged on different grounds
\cite{102,103}: The integral over the photon spectrum (\ref{e2.10}) tends to
smear out the distributions, which in Fig.~16 were shown for a fixed energy
(10 GeV) of the incident photon. In particular, direct contributions from
rather soft initial photons also populate the region of positive rapidity.
These papers also complete the NLO calculations by including corrections to
the fragmentation processes, but these corrections have little bearing on the
question whether direct photon production is useful for constraining the gluon
content of the photon.
Ref.\cite{103} introduces two additional refinements. It makes use of an NLO
parametrization \cite{104} of the parton to photon fragmentation functions,
which updates the ``asymptotic" expressions of ref.\cite{37a}. More
importantly, this analysis for the first time introduces an isolation
requirement. At fixed target energies, backgrounds from $\pi^0$ and $\eta$
decays can be suppressed quite reliably, since they contain two photons with
(usually) substantial opening angle \cite{104a}. At higher energy, and higher
$p_T^{\gamma}$, this opening angle becomes much smaller, making it more
difficult to detect both photons individually. At \mbox{$p \bar{p}$}\ colliders a direct
photon signal could therefore only be detected \cite{105} if the photons were
isolated, i.e. after events were discarded if more than some maximal (small)
amount of energy was found in a cone around the photon. Such a cut reduces the
fragmentation contribution considerably; in particular, the contribution from
$g \rightarrow \gamma$ fragmentation becomes very small, since here most of the energy
usually goes into the accompanying jet rather than the photon. Nevertheless,
at hadron colliders contributions from $q \rightarrow \gamma$ fragmentation can still
be significant \cite{106}. Strictly speaking, the necessity to impose an
isolation cut at HERA has not yet been demonstrated. Indeed, the only
background study we are aware of \cite{107} reaches the conclusion that even
without isolation cut the background in the most interesting region of
positive rapidity can be suppressed to the 25\% level, which might be
tolerable. However, this study ignores resolved photon contributions to the
background (they are included for the signal); we saw in Sec.~3a that these
contributions will increase the total jet cross--section (and hence also the
cross--section for the production of high$-p_T$ particles) by a large factor
at positive rapidity. In the absence of a more complete background study we
are therefore inclined to believe that an isolation cut will indeed be
necessary at HERA.
In Fig.~17, which has been adapted from numerical results of ref.\cite{103},
we show the rapidity dependence of $\sigma(ep \rightarrow e \gamma X)$ after requiring
that the hadronic energy in a cone $\delta \equiv \sqrt{ (\Delta \eta)^2 +
(\Delta \phi)^2}/\cosh \eta_\gamma = 0.4$ around the outgoing photon be less
than 10\% of the energy of that photon. No cut on the energy of the incident
photon has been imposed. We see that, except at very large $\eta_\gamma$, the
cross--section is dominated by the direct contributions and resolved photon
contributions involving the quark content of the photon only. In particular,
in sharp contrast to the ``Born" cross--section of Fig.~16, after integration
over the incident photon spectrum the direct contribution has a very flat
rapidity distribution. The authors concluded that, while one might be able to
further constrain the quark densities in the photon from this measurement (see
the difference between the dotted curve, obtained from the GS parametrization,
and the solid one, which is for the GRV parametrization), there is very little
sensitivity to the gluon content of the photon.
However, we saw in Sec.~3a that even a rather modest lower cut on the energy
of the incident photon greatly suppresses direct photon contributions at
positive rapidity; see e.g. Fig.~10. Such a cut would also reduce
contributions $\propto q_i^{\gamma}$ substantially, but would have much less
effect on contributions $\propto \mbox{$G^{\gamma}$}$. The sensitivity to the gluon content
of the photon would presumably be enhanced even more if the rapidity of the
jet balancing the direct photon can be measured as well
\cite{102}\footnote{Unfortunately, no NLO calculation for $\gamma+$jet
production exists as yet.}. We therefore believe that the conclusions of
refs.\cite{102,103} are probably too pessimistic. On the other hand, HERA will
have to accumulate a large body of data before a substantial number of direct
photon events with known (fixed) energy of the incident photon will become
available; only then can predictions like those shown in Fig.~16 be compared
with experiment.
So far no HERA data on direct photon production have been published. Data from
a fixed target photoproduction experiment exist \cite{104a}; they agree with
theoretical predictions \cite{107a}, but the energy is too low to extract a
signal for the resolved photon contribution.
Before closing this subsection we briefly mention some work on closely
related topics. In ref.\cite{108} the cross--section for $c+\gamma$ production
has been computed (ignoring fragmentation contributions, however), the main
motivation being that this might allow to constrain the charm content of the
photon. In ref.\cite{109} it has been suggested that one might learn something
about the parton content of polarized photons from the study of direct photon
production. Indeed, longitudinally polarized electron beams should become
available at HERA in a few years; part of this polarization will be passed on
to photons emitted from these electrons, if they carry a substantial fraction
of the electron's energy. However, one would also have to measure the
polarization of the {\em outgoing} photon, which seems quite difficult.
Finally, ref.\cite{110} is a first study of the production of {\em two}
direct photons. The cross--section is substantially smaller (by a factor
$\propto e_q^2 \mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$}$) than the one for single direct photon
production. Apart from fragmentation contributions, only the resolved photon
process $\mbox{$q \bar{q}$} \rightarrow \mbox{$\gamma \gamma $}$ contributes in LO; processes like $\gamma q \rightarrow q
\mbox{$\gamma \gamma $}$ are formally of NLO, since $q^{\gamma}_i \propto \mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$}$, and are
indeed found to be subdominant numerically. Due to the large charge factor,
this process might be useful for probing the up--quark density in the photon
at small $x$. If $p_T^{\gamma}$ values as low as 3 GeV are experimentally
accessible, one might even be able to extract the contribution from the
box diagram, $gg \rightarrow \mbox{$\gamma \gamma $}$; since it is proportional to the gluon density in
the photon, it should be concentrated at larger rapidities than the
contribution from \mbox{$q \bar{q}$}\ annihilation. This close analogue of the famous
light--by--light scattering process \cite{110a} has yet to be studied
experimentally.
\setcounter{footnote}{0}
\subsection*{3d) \mbox{$J/\psi$}\ Production in \mbox{$\gamma p$}\ Scattering}
The inelastic production of \mbox{$J/\psi$}\ mesons in (virtual or real) \mbox{$\gamma p$}\
collisions has long been regarded as one of the cleanest methods to constrain
the shape of the gluon distribution in the proton \cite{111}. ``Inelastic"
here means that the quantity
\begin{equation} \label{e3.6}
Z \equiv \frac {p_p \cdot p_{J/\psi}} {p_p \cdot p_{\gamma}}
\end{equation}
is significantly below unity. In the proton rest frame, eq.(\ref{e3.6})
reduces to $Z = E_{J/\psi}/E_{\gamma}$; $Z=1$ therefore means that the
incident photon transmits its entire energy to the \mbox{$J/\psi$}\ meson. The
cross--section for inelastic \mbox{$J/\psi$}\ production is nowadays usually computed
using the ``colour singlet" model of ref.\cite{112}, see Fig.~18. In this
model one computes the cross--section for $\gamma g \rightarrow \mbox{$c \bar{c}$} g$, and
projects out the contribution where the \mbox{$c \bar{c}$}\ system is in a
colour--singlet, $s-$wave, $J=1$ state. One further assumes that the relative
momentum of the $c$ and $\bar{c}$ in the \mbox{$J/\psi$}\ are negligible; the matrix
element for \mbox{$J/\psi$}\ production is then proportional to the wave function at the
origin $\Psi(0)$, which can be determined from the leptonic decay width of
\mbox{$J/\psi$}.
This model seems to describe the $Z$ distribution of fixed target data better
\cite{113} than the alternative ``dual model" \cite{114,84}. The agreement in
the high$-Z$ region becomes even better if one introduces a small but
nonvanishing relative momentum between the $c$ and $\bar{c}$ quarks
\cite{115}. Until recently the overall normalization of the cross--section was
not well reproduced by theoretical calculations, i.e. a sizable ``K--factor"
had to be introduced. Fortunately NLO corrections to the direct diagram of
Fig.~18 have recently been calculated \cite{116}. Together with the NLO
corrections to the leptonic decay width \cite{117} they give an overall
inelastic cross--section in agreement with fixed target photoproduction
experiments; however, the photoproduction cross--section extracted from fixed
target leptoproduction experiments seems to be somewhat higher \cite{113}. The
calculation of ref.\cite{116} predicts that the K--factor should be smaller at
high (HERA) energies.
As usual, resolved photon contributions are small at the energies of fixed
target experiments, but they could be quite important at HERA \cite{118,119}.
In addition to the process $gg \rightarrow \mbox{$J/\psi$} g$, one also has to consider $gg \rightarrow
\chi_c$ and $gg \rightarrow \chi_c g$, with subsequent decay of the heavier $\chi_c$
states into a \mbox{$J/\psi$}\ and a (soft) photon \cite{120}; the latter process has
to be considered if one imposes a cut on the transverse momentum of the \mbox{$J/\psi$}.
For $p_T(\mbox{$J/\psi$}) > 3$ to 5 GeV, the contribution from $b \rightarrow \mbox{$J/\psi$}$ decays
becomes important, and eventually even dominates the direct contribution
\cite{121}. There might also be sizable contributions from $cg \rightarrow \mbox{$J/\psi$} c$
\cite{122}. However, the typical momentum scale in most \mbox{$J/\psi$}\ events is too
small to reliably use charm distribution functions in the photon or proton;
we therefore neglect this contribution here.
As usual, resolved photon contributions are characterized by the presence of
the photonic remnant jet, and by a rapidity distribution that peaks at large,
positive values. In addition, direct and resolved photon contributions have
very different $Z$ distributions. The direct contribution is peaked at $Z
\simeq 1$, which corresponds to a small energy of the outgoing gluon. In
contrast, for fixed $x_{\gamma}$ the resolved photon contribution peaks at
$Z \simeq x_{\gamma}$; after convolution with the photon spectrum
(\ref{e2.10}) this leads to a $Z$ distribution that quickly rises with
decreasing $Z$.
This is shown in Fig.~19, which we adapted from numerical results of
ref.\cite{123}. Since only a mild cut on the $p_T$ of the \mbox{$J/\psi$}\ meson has
been applied, the contribution from $b$ decays is relatively small and has
been neglected. LO expressions have been used everywhere; neither NLO nor
non--relativistic corrections to the resolved photon contribution are as yet
known. The B1 parametrization of ref.\cite{124} has been used for the gluon
density in the proton, and DG \cite{39} for the photon. Finally, although the
cross--section has not been multiplied with any branching ratio, it is
clear that only the leptonic decays $\mbox{$J/\psi$} \rightarrow \mbox{$e^+ e^- $}, \mu^+ \mu^-$ are
detectable at HERA; the combined branching ratio for these modes is about
12\%. The results of Fig.~19 have been obtained under the condition that both
leptons can be reconstructed by the ZEUS tracking system. Unfortunately this
removes events at large rapidity, which greatly reduces the resolved photon
contribution. Including the leptonic branching ratio, it only amounts to about
7 pb after integration over $Z$, compared to a direct contribution of about
110 pb.\footnote{Recall that these are LO predictions, and hence somewhat
uncertain; the cross--sections also depend on the gluon densities in the
photon and proton, of course.} Requiring $Z > 0.2$ leaves a very pure direct
sample, which should allow to determine the shape of the gluon density in
the proton for $2 \cdot 10^{-4} \leq x \leq 0.1$ \cite{123}.
Unfortunately the extraction of \mbox{$G^{\gamma}$}\ seems to be much less straightforward.
The simulation of ref.\cite{123} indicated that measurement errors would
smear out the direct contribution into the region of small $Z$, totally
obscuring the resolved photon contribution. It remains to be seen whether an
unfolding procedure, and/or the application of additional cuts (e.g., tagging
of the photon remnant jet) will improve the situation sufficiently to allow to
extract new information on \mbox{$G^{\gamma}$}\ from \mbox{$J/\psi$}\ production at
HERA.\footnote{Previous analyses \cite{119,60a}, which had led to very
optimistic conclusions, had used much milder acceptance cuts on the leptons,
or none at all.}
Both H1 and ZEUS have published first results \cite{125} on \mbox{$J/\psi$}\
production. However, in both analyses events were rejected if any particle in
addition to the two leptons resulting from \mbox{$J/\psi$}\ decay was observed. This
cut removes most if not all inelastic contributions, but is sensitive to
elastic or diffractive \mbox{$J/\psi$}\ production. This allows to test Pomeron--based
models \cite{123,115,125a}, but teaches us nothing about the parton content of
the photon. Very recently, ZEUS has announced \cite{125b} preliminary results
on inelastic \mbox{$J/\psi$}\ production. However, they required $Z > 0.2$, which again
removes most of the resolved photon contribution.
Finally, we briefly mention associate $\mbox{$J/\psi$} + \gamma$ production. As first
pointed out by Fletcher et al. \cite{119}, in LO only the resolved process
$gg \rightarrow \mbox{$J/\psi$} + \gamma$ contributes, the direct contribution being forbidden
by colour conservation. In principle this process therefore allows a clean
determination of (the shape of) \mbox{$G^{\gamma}$}\ \cite{126}. Unfortunately the
cross--section at HERA is quite small, very roughly of order 0.1 to 1 pb after
acceptance cuts and multiplication with the leptonic branching ratio.
\setcounter{footnote}{0}
\subsection*{3e) Production of Lepton Pairs in \mbox{$\gamma p$}\ Collisions}
In this subsection we discuss the production of lepton pairs, either due to
the exchange of a virtual photon or from the decay of an on--shell $W$ or $Z$
boson. The corresponding cross--sections are quite small even at HERA
energies, so we will be brief here.
In the theoretical treatment of these reactions one has to distinguish the
cross--section integrated over the transverse momentum of the lepton pair (not
to be confused with the $p_T$ of the individual leptons) from the
cross--section for the production of a high$-p_T$ lepton pair. In the former
case, the only LO contribution comes from the resolved photon process $\mbox{$q \bar{q}$}
\rightarrow l^+l^-$, which produces a lepton pair with vanishing transverse momentum.
The corresponding $ep$ cross--section is ${\cal O}(\alpha_{\rm em}^3/
\alpha_s)$ since, as emphasized repeatedly, the parton distribution functions
in the photon are ${\cal O}(\mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$})$. In contrast, both the direct
process $\gamma q \rightarrow l^+ l^- q$ and the resolved photon process $g q \rightarrow l^+
l^- q$ contribute in LO to the production of a lepton pair with sizable $p_T$.
The corresponding $ep$ cross--sections are both ${\cal O}(\alpha_{\rm em}^3)$.
This distinction has been well understood in existing treatments of the
``Drell--Yan" production of $l^+ l^-$ pairs via the exchange of a virtual
photon; see ref.\cite{118} for a LO estimate, and \cite{89,127,128} for NLO
calculations. Compared to direct photon production this process has the
theoretical advantage that one does not have to worry about fragmentation
contributions. Moreover, backgrounds are low, and the final state can be
reconstructed cleanly even at rather low $p_T(l)$, allowing one to probe quite
small $x$ values \cite{128}. As usual, the region of positive rapidity
corresponds to small $x_{\gamma}$ and moderate $x_p$, while negative
rapidities correspond to large $x_{\gamma}$ and very small $x_p$. The
production of lepton pairs at positive rapidity and sizable $p_T$ is sensitive
to the gluon content of the photon \cite{127}. Unfortunately the expected
event rates are quite low: ${\displaystyle \frac{ d \sigma (ep \rightarrow l^+ l^- X)}
{d M_{l^+l^-}} \simeq 1}$ pb/GeV at $M_{l^+l^-}=4$ GeV after integration over
the transverse momentum of the pair and before any acceptance cuts have been
imposed \cite{128}.
In contrast, the fact that the {\em total} photoproduction cross--section for
$W$ and $Z$ bosons is, in LO, given by the resolved photon contribution {\em
only} has not been appreciated in the existing literature.\footnote{The total
$W$ and $Z$ cross--sections in $ep$ collisions also get contributions where
the heavy gauge boson is radiated off the electron line; in case of $Z$
production at HERA this contributes roughly 50\% of the total cross--section
\cite{129}.} The resolved photon contribution has first been estimated in
ref.\cite{130}, but here this contribution was considered to be
an addition to the direct process even at vanishing $p_T$ of the heavy gauge
boson. This is not correct, since the direct contribution has a $u-$channel
singularity which has to be absorbed into the resolved photon contribution;
simply adding both contributions implies double--counting. While current
calculations \cite{131,129} of the total $W$ and $Z$ photoproduction
cross--sections are in our view not entirely satisfactory, since they mix
LO and NLO contributions in an ill--controlled manner, they should predict
the production of high$-p_T$ gauge bosons quite accurately.\footnote{In
principle the resolved photon contribution from $gq \rightarrow W q'$ should be
included in a full LO treatment, but it is strongly suppressed at HERA
energies due to phase space constraints.} This high$-p_T$ region is sensitive
to the form of the $W^+W^- \gamma$ vertex \cite{131}. The total
cross--sections for $W^+$ and $W^-$ production at HERA amount to approximately
0.5 pb each. This cross--section is to be divided by another factor of five
if only the clean $e \nu_e$ and $\mu \nu_{\mu}$ final states are observable;
indeed, a first study \cite{132} has concluded that the detection of
hadronically decaying $W$ and $Z$ bosons at HERA is quite challenging,
although perhaps not impossible. Finally, we mention that recently the H1
collaboration has announced \cite{133} observation of one event that can be
interpreted as the production of a leptonically decaying $W$ boson at high
$p_T$; within the SM this interpretation is quite unlikely, due to the
smallness of the predicted cross--section, but all other interpretations seem
even less likely. Clearly much more data have to be analyzed before any
definite conclusion can be drawn.
\setcounter{footnote}{0}
\section*{4) Real \mbox{$\gamma \gamma $}\ Scattering at \mbox{$e^+ e^- $}\ Colliders}
In this section we discuss hard processes with two (quasi--)real photons in
the initial state and a hadronic final state. At present, and in the near
future, such reactions can only be studied at \mbox{$e^+ e^- $}\ storage rings, where the
effective photon flux is given by eq.(\ref{e2.10}). At future linear \mbox{$e^+ e^- $}\
colliders the photon spectrum might receive large additional contributions
from ``beamstrahlung" \cite{134}, which is emitted when a particle is
accelerated in the field produced by the opposite bunch; beamstrahlung photons
are exactly on--shell. The produced spectrum sensitively depends on the size
and shape of the electron and positron bunches; see ref.\cite{134a} for a
handy parametrization of the spectrum in terms of a few machine parameters.
Finally, in recent years the possibility has been discussed to convert a
linear \mbox{$e^+ e^- $}\ collider into a ``\mbox{$\gamma \gamma $}\ collider" by scattering laser photons
off the incident $e^{\pm}$ beams \cite{135}. The achievable luminosity is
predicted to be comparable to the (geometrical) \mbox{$e^+ e^- $}\ luminosity prior to
conversion of the beams; in contrast, at existing \mbox{$e^+ e^- $}\ storage rings one has
${\cal L}_{\gamma \gamma} \sim \left( \frac {\mbox{$\alpha_{\rm em}$}}{\pi} \ln \frac {s} {m_e^2}
\right)^2 \sim 10^{-3} {\cal L}_{ee}$. The photon spectrum of such a photon
collider again depends on the geometrical set--up, and also on the
polarization of the incident electron and laser photon beams. Most of the
results presented here will be for present and near--future colliders; these
are obviously of more immediate interest, and already allow to illustrate most
physics principles. For comparison we will also give a few results for the
more ``futuristic" colliders.
Since we now have two photons in the initial state, we have to distinguish
three physically distinct event classes \cite{37b}, see Fig.~20. Following the
notation of ref.\cite{136} we call reactions where the entire energy of both
photons goes into the hard subprocess ``direct", see Fig.~20a. If only one of
the photons couples in a pointlike manner while the other participates via
its quark and gluon content, as in Fig.~20b, the process is called ``once
resolved" (``1--res" for short). Finally, processes where both photons are
resolved into their partonic constituents are called ``twice resolved"
(``2--res"); an example is shown in Fig.~20c. Recall that each resolved
photon produces a remnant or ``spectator" jet, which goes approximately in the
direction of the incident $e^{\pm}$ beams; the three event classes are
therefore characterized by having zero, one or two of these photonic remnant
jets. Since the parton distribution functions in the photon are ${\cal
O}(\mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$})$, all three contributions are of the same order in coupling
constants, and have to be treated on the same footing \cite{37b}.
We saw in the previous section that direct and resolved photon contributions
to photoproduction processes mix in NLO QCD. Similarly, the three event
classes contributing to real \mbox{$\gamma \gamma $}\ scattering mix once higher--order QCD
corrections are included. Parts of the NLO direct (1--res) contributions have
already been included in the LO 1--res (2--res) terms; these parts therefore
have to be subtracted from the NLO contributions. Note that mixing between
direct and twice resolved contributions only occurs in NNLO in QCD.
Following our preceding discussion of \mbox{$\gamma p$}\ scattering, we discuss different
final states in separate sub--sections. Jet production is treated in Sec.~4a,
open heavy flavour production in Sec.~4b, \mbox{$J/\psi$}\ production in Sec.~4c, and
direct photon production in Sec.~4d.
\setcounter{footnote}{0}
\subsection*{4a) Jet Production in \mbox{$\gamma \gamma $}\ Collisions}
As in case of \mbox{$\gamma p$}\ scattering, the production of jets offers the largest
cross--section of all hard \mbox{$\gamma \gamma $}\ collisions that lead to hadronic final
states. In LO, the cross--section can be written as [see eq.(\ref{e3.1})]:
\begin{equation} \label{e4.1}
\frac{d^3 \sigma(\mbox{$e^+ e^- $} \rightarrow \mbox{$e^+ e^- $} j_1 j_2)} {d p_T d \eta_1 d \eta_2} = 2 p_T
x_1 x_2 \sum_{i,j,k,l} f_{i|e} (x_1) f_{j|e} (x_2)
\frac { d \hat{\sigma}_{ij \rightarrow kl} ( \hat{s}, \hat{t}, \hat{u} )} {d \hat{t}}.
\end{equation}
If $i$ or $j$ is a quark or gluon, $f_{i|e}$ can again be obtained by
convoluting the photon flux function \mbox{$f_{\gamma|e}$}\ with the quark or gluon density in
the photon, see eq.(\ref{e3.2}). However, antitag requirements at current
\mbox{$e^+ e^- $}\ collider experiments often allow larger photon virtualities than one
has in photoproduction events at HERA; at the same time, the scale $|\hat{t}|$
or $p_T^2$ of the hard process is usually smaller. This means that the
suppression of resolved photon contributions due to the reduced parton
content of virtual photons is usually larger at \mbox{$e^+ e^- $}\ colliders than at HERA.
In our numerical estimates to be presented below we have included this effect
using the formalism of ref.\cite{78}.
The scaling variables $x_{1,2}$ in eq.(\ref{e4.1}) are related to the jet
(pseudo)rapidities $\eta_{1,2}$ by:
\begin{subequations} \label{e4.2} \begin{eqalignno}
x_1 &= \frac {x_T}{2} \left( e^{\eta_1} + e^{\eta_2} \right); \label{e4.2a}
\\
x_2 &= \frac {x_T}{2} \left( e^{-\eta_1} + e^{-\eta_2} \right), \label{e4.2b}
\end{eqalignno} \end{subequations}
with $x_T = 2 p_T / \mbox{$\sqrt{s}$}$ as before. The Mandelstam variable $\hat{s}$ of the
hard scattering sub--process is again given by $\hat{s} = x_1 x_2 s$, and
$\hat{t}$ is as in eq.(\ref{e3.4}). Most of the sub--processes contributing
to jet production in \mbox{$\gamma \gamma $}\ collisions also contribute to jet production at
HERA. In particular, if both $i$ and $j$ are partons (twice--resolved
contribution), $\hat{\sigma}$ is the same as in resolved \mbox{$\gamma p$}\ collisions,
and the case where either $i$ or $j$ is a parton while the other is a photon
(single--resolved contribution) corresponds to the direct contribution at
HERA. Finally, if $i=j=\gamma$ (direct contribution), then $k=q,\ l=\bar{q}$
and one has
\begin{equation} \label{e4.3}
\frac{ d \hat{\sigma} (\mbox{$\gamma \gamma $} \rightarrow \mbox{$q \bar{q}$})} {d \hat{t}} = 3 e_q^4
\frac {2 \pi \alpha^2_{\rm em}} {\hat{s}^2} \left( \frac {\hat{t}} {\hat{u}} +
\frac {\hat{u}} {\hat{t}} \right),
\end{equation}
where $e_q$ is the electric charge of quark $q$ in units of the proton charge,
and the factor of 3 is due to colour.
The first quantitative estimate of high$-p_T$ jet production in (quasi-)real
\mbox{$\gamma \gamma $}\ scattering has been presented by Brodsky et al. \cite{137}. However,
gluon--initiated processes were omitted, and a rather crude parametrization
for the quark densities in the photon was used.\footnote{Ref.\cite{137} also
finds quite large higher twist contributions to the production of high$-p_T$
mesons and jets. It was shown later \cite{138} that the ``constituent
interchange model" used in ref.\cite{137} over--estimates the normalization
of these terms by as much as a factor of a thousand.} Gluon--initiated
contributions have been included in refs.\cite{139}, but again very simple
parametrizations for the parton densities in the photon were used. In spite of
their shortcomings, these early studies clearly demonstrated that resolved
photon contributions are quite important, and often even dominant, if $x_T
\leq 0.2$. Notice that jet production from two--photon collisions at present
\mbox{$e^+ e^- $}\ colliders cannot be studied experimentally if $x_T \geq 0.4$; the
cross--section becomes too small, and annihilation backgrounds too large.
Hence resolved photon contributions were predicted to be sizable for at least
half the experimentally accessible range of $x_T$.
Nevertheless data \cite{14} on multi--hadron production from \mbox{$\gamma \gamma $}\ collisions
taken at the PEP and PETRA storage rings were usually only compared to the
direct (QPM) contribution.\footnote{The PLUTO collaboration \cite{140}
attempted to include 1--res $qg$ or 2--res $qq$ final states in their
analysis. However, they made many simplifying assumptions, some of which are
incorrect. In particular, they assumed that the cross--sections for these
resolved photon contributions drops like the square of the \mbox{$\gamma \gamma $}\ invariant
mass $W$; in contrast, for fixed $p_T$ QCD predicts these cross--sections to
increase with $W$. Moreover, PLUTO did not attempt to include 1--res and
2--res contributions simultaneously.} Not surprisingly, a significant excess
of data over MC prediction was observed. Given that the importance of resolved
photon contributions had been emphasized quite early on \cite{137,139}, it is
surprising that it took ten years before the origin of this excess of data
over QPM prediction was clarified experimentally.
In ref.\cite{136} it was pointed out that the characteristics of these excess
events at least qualitatively agreed with those expected from resolved photon
contributions: Their $p_T$ spectrum is softer than that of the direct
contribution, just as in case of photoproduction events. In addition, resolved
photon events are less two--jet like, i.e. have smaller thrust $T$, than
direct events; in fact, the PLUTO collaboration had shown \cite{140} that the
excess could be removed by requiring $T \geq 0.9$. Ref.\cite{136} also
contains the prediction that, for fixed $p_T$, resolved photon events should
rapidly become more important as the beam energy is increased. The reason is
that raising \mbox{$\sqrt{s}$}\ decreases $x_T$, and hence $x_{1,2}$ in eqs.(\ref{e4.2});
due to the additional convolution with photonic parton distribution functions,
the quark and gluon density in the electron is always considerably softer,
i.e. grows faster with decreasing $x$, than the photon flux in the electron.
Resolved photon contributions were therefore expected to play an even more
important role at TRISTAN and LEP than at PEP and PETRA. This was confirmed
soon afterwards by the AMY collaboration \cite{6}, who showed that their data
on multi--hadron production in antitagged \mbox{$\gamma \gamma $}\ collisions could be
reproduced by their QCD--based Monte Carlo program, in both shape and
normalization, if and only if it included the full set of resolved photon
contributions, including those with a gluon in the initial state. This was the
first unambiguous observation of resolved photon interactions other than in
deep--inelastic $e \gamma$ scattering, and the first direct experimental
evidence for a nonzero gluon density in the photon.
Early experimental analyses of multi--hadron production in \mbox{$\gamma \gamma $}\ collisions
\cite{14}, including the AMY analysis \cite{6}, were not amenable to direct
comparison with theoretical calculations, since the experiments defined a
``jet" as a thrust hemisphere. The production of actual, reconstructed jets
(using a cone algorithm) has been studied only quite recently, first by the
TOPAZ collaboration \cite{141} and then by AMY \cite{142}. This is an
important development, since it allows a much more direct comparison between
theory and experiment.
In the meantime theoretical estimates are also becoming more sophisticated, by
including NLO corrections. This process was actually already startd in 1979
with two calculations \cite{143} of $\mbox{$\gamma \gamma $} \rightarrow gg$ via a quark box diagram.
NLO corrections to the direct process $\mbox{$\gamma \gamma $} \rightarrow \mbox{$q \bar{q}$}$ (for massless quarks)
were calculated soon afterwards \cite{144}. However, the result contains
collinear divergencies, which have to be absorbed in the 1--res contribution;
in NLO it therefore makes little sense to consider the direct contribution in
isolation. This was recognized in ref.\cite{145}, which deals with the
production of high$-p_T$ hadrons. Here the direct contribution was treated in
NLO (including the contribution from $\mbox{$\gamma \gamma $} \rightarrow gg$, which is formally NNLO),
but at the time resolved photon contributions could only be included in
leading order.\footnote{Notice that one only needs the LO 1--res contribution
from $\gamma q$ scattering in order to absorb the collinear divergencies of
the NLO direct term. The procedure of ref.\cite{145} should therefore be quite
accurate at high $x_T$, where the direct contribution dominates.} This
shortcoming could be remedied only after the NLO corrections to the hard
partonic sub--process had been calculated \cite{68}. A full NLO calculation of
the single--jet inclusive cross--section has become available only recently
\cite{146}; it finds only modest NLO corrections to the total cross--section
($\leq 25\%$, for a jet cone size $\Delta R = 1$ used by both TOPAZ and AMY,
and renormalization and factorization scale $\mu = p_T$), but the relative
weights of direct, single-- and double--resolved contributions can change by
larger amounts.\footnote{However, there seems to be a discrepancy between the
LO result of ref.\cite{146} and our calculation. In particular, Aurenche et
al. find that 2--res processes still contribute about 25\% to the jet
cross--section measured by TOPAZ at $p_T =7.5$ GeV, while in our calculation
it amounts to less than 10\% at this rather large value of $x_T$ (unless we
use the LAC3 parametrization).} Finally, in ref.\cite{147} the 2--res
contribution has been studied in NLO, with emphasis on constraining the gluon
content of the photon.
Since NLO corrections appear to be quite modest, and since many more LO
parametrizations of parton densities in the photon exist, we will only present
results from LO analyses here. As mentioned earlier, we do include the
suppression of resolved photon contributions due to the finite virtuality of
the photons, using the simplest ans\"atze in ref.\cite{78}. (Quark and gluon
densities have to be treated separately.) We remind the reader that, unlike
ref.\cite{146}, these ans\"atze do not assume a power--like suppression of the
``hadronic" contribution to the parton densities of virtual photons; they
might therefore slightly under--estimate the (in any case only modest) size of
the correction.
In Fig.~21 we compare our LO calculation of the single--jet inclusive
cross--sections as measured by the TOPAZ (a) and AMY (b) collaborations at
TRISTAN ($\mbox{$\sqrt{s}$}=58$ GeV). TOPAZ requires the jet to have pseudorapidity
$|\eta_{\rm jet}| \leq 0.7$, and rejects an event if it contains an electron
or positron at an angle $\theta > 3.2^{\circ}$ relative to the beam pipe and
with energy $E > 0.25 E_{\rm beam} = 7.2$ GeV. AMY accepts jets out to
$|\eta_{\rm jet}| = 1.0$; moreover, it can antitag an event only if it
contains an electron or positron with $E > 0.25 E_{\rm beam}$ at an angle
$\theta > 14.1^{\circ}$.
We see from Fig.~21 that the DG, LAC1 and WHIT1 parametrizations reproduce
the TOPAZ data at $p_T \leq 4.5$ GeV quite well, and also fit the AMY data
over the entire $p_T$ range. Notice, however, that we had to use $N_f=4$
active flavours already at $p_T({\rm jet}) = 2.5$ GeV in order to achieve this
agreement. This is not really justifiable for the DG and LAC parametrizations,
which treat the charm quark as massless; on the other hand, WHIT explicitly
includes some charm mass effects, which leads to a greatly reduced
contribution from charm in the photon, in agreement with expectations. (It
makes up for the shortfall by slightly larger light quark and gluon
densities.) The GRV parametrization seems to fall below the data at low $p_T$;
this is mostly because we have used $\Lambda_{\rm QCD}(N_f=3) = 0.4$ GeV for
DG and WHIT, as compared to 0.2 GeV for GRV and LAC, as is implicit in these
parametrizations. NLO corrections have been found to be most important at low
$p_T$ \cite{146}; they might very well bring the GRV prediction into agreement
with the data. In contrast, the WHIT4 prediction lies above the data at low
$p_T$, since it assumes a quite large and hard gluon density in the photon.
This discrepancy might be reduced if the ``hadronic" piece of the parton
density of virtual photons is suppressed by a power of the virtuality, but it
seems unlikely that this will restore full agreement with the data. All the
other WHIT parametrizations seem acceptable, however.
Fig.~21 shows that the LAC3 parametrization clearly over--estimates the
cross--section, while the direct contribution by itself falls well below the
data. Notice that the resolved photon contributions remain non--negligible out
to the highest $p_T$ where data exist, although their importance clearly
diminishes with increasing $p_T$, as discussed above. Finally, we observe that
our calculation falls somewhat below the TOPAZ data for $p_T > 4.5$ GeV. This
is disturbing, since in principle our neglect of the charm mass should be
better justified if $p_T^2 \gg m_c^2$. Notice, however, that our calculation
agrees very nicely with the AMY data. This hints at an experimental
discrepancy between these two data sets. Indeed, from the increased acceptance
in rapidity as well as the much looser antitag requirements used by AMY, one
expects that the cross--section measured by AMY should exceed that measured by
TOPAZ by a factor of about 1.7 to 1.8; here we have made use of the fact
\cite{141,142} that $d \sigma / d \eta_{\rm jet}$ is quite flat over the
observed range, and have used eq.(\ref{e2.10}) to estimate the effect of the
tagging criteria on the photon flux. On the other hand, the cross--section
measured by AMY only exceeds that reported by TOPAZ by a factor of $1.36 \pm
0.26$ ($1.29 \pm 0.20$) at $p_T = 4.55$ (6.5) GeV, where we have assumed all
errors to be independent. Although this procedure probably over--estimates the
true error, since part of the systematics should be common to both
experiments, this comparison indicates that the discrepancy in any one $p_T$
bin is not very significant statistically. Due to bin--to--bin correlations
produced by the unfolding procedure, it is difficult to combine different bins
to arrive at an overall significance of the discrepancy. At the lowest $p_T$
bin, the ratio of the two experimental cross--sections is 1.55, which agrees
very well with expectations.\footnote{The effect of the looser antitag
requirement of AMY is smaller at low $p_T$, due to the dynamical upper bound
$P^2 \leq p_T^2$ that has to be imposed on the virtuality of the exchanged
photons in order to meaningfully speak of parton or photon densities in the
electron.} Moreover, both collaborations also publish di--jet cross--sections,
although with somewhat larger relative errors; our LO calculation reproduces
both these measurements over the entire $p_T$ range.
\setcounter{footnote}{0}
Note that the published cross--sections are actually partonic cross--sections,
which have been obtained by unfolding the observed $p_T$(jet) and $\eta$(jet)
distributions using a LO MC program. We will argue in Sec.~5 that such a
procedure might be more reliable at TRISTAN than at HERA, since events with
multiple partonic interactions should not pose much of a problem here.
Nevertheless this procedure does produce some dependence on the details of the
Monte Carlo program, e.g. via the predicted jet reconstruction efficiency.
It might also be dangerous to directly compare these extracted cross--sections
with NLO calculations, since the higher order corrections seem to change
\cite{146} the relative weights of the three classes of contributions compared
to what has been assumed in the LO MC.\footnote{It is not obvious how the
``theoretical" definition of the three classes of contributions used in
ref.\cite{146} relates to a more experimental definition relevant for event
reconstruction, which could e.g. be based on the presence of absence of
(remnant) jets close to the beam pipes. For example, the relative weights of
the three classes of contributions in the NLO calculation strongly depend on
the factorization scale, which is an unphysical parameter. Unfortunately it is
quite difficult to write an MC generator based on a full NLO calculation; no
such generator for jet events produced in hadronic or photonic collisions
exists as yet.} Both experiments also give the measured rate of jet events
as a function of $p_T$; however, at these rather low transverse momenta the
jet reconstruction efficiency is still rather small and $p_T-$dependent. A
fair amount of MC work is therefore (unfortunately) necessary before a
comparison with theoretical predictions can be made.
The published data are based on an integrated luminosity of about 90 pb$^{-1}$
for TOPAZ, and 27 pb$^{-1}$ for AMY. The total available data samples are more
than three times larger than this. Notice that the error bars in Fig.~21 are
already quite small; if the slight discrepancy between AMY and TOPAZ can be
resolved, the full data set should therefore allow a quite precise comparison
between theory and experiment.
The larger data sets should also allow to investigate more differential
distributions, using events with two reconstructed jets. As an example we show
in Fig.~22 the triple--differential cross--section for events with two jets
with equal (pseudo)rapidity $\eta_1 = \eta_2 \equiv \eta$. This distribution
is quite flat for the direct contribution, at least over the range covered by
TRISTAN detectors. For parametrizations with a relatively hard gluon density
(WHIT1, DG) the total single--resolved contribution also depends only weakly
on $\eta$; for the given choice of $p_T$, the decrease of the contributions
$\propto q_i^{\gamma}$ ($qg$ final state) is more or less balanced by the
increase of those $\propto \mbox{$G^{\gamma}$}$ (\mbox{$q \bar{q}$}\ final state). Increasing $\eta$
means increasing $x_1$ but decreasing $x_2$ in eqs.(\ref{e4.2}). Since the
gluon content in the electron is always quite strongly peaked at small $x$
while the photon flux function \mbox{$f_{\gamma|e}$}\ is relatively hard, the increase of
$f_{G|e}$ more than compensates the decrease of \mbox{$f_{\gamma|e}$}\ as $\eta$ is increased.
(Note that $f_{q|e}$ increases much more slowly with decreasing $x$ than
$f_{G|e}$ does.) If the gluon density in the photon is large but strongly
peaked at small $x$ (WHIT6, LAC1), the rise of the cross--section for
$\gamma g \rightarrow \mbox{$q \bar{q}$}$ even leads to a total 1--res contribution that peaks
at $\eta \simeq 2$, rather than at $\eta = 0$.
Turning to the twice--resolved contributions shown in Fig.~22b, we observe
that the cross--sections for processes with two gluons in the final state,
which most of the time also have two gluons in the initial state, is quite
strongly peaked at $\eta=0$, even for the comparatively hard gluon density of
WHIT1. The same parametrization predicts the cross--section from $qg \rightarrow qg$
scattering to remain quite flat for $|\eta| \leq 1$, since the decrease of
$f_{q|e}(x_1)$ is compensated by the increase of $f_{G|e}(x_2)$.\footnote{Of
course, there is also a contribution $\propto f_{q|e}(x_2) f_{G|e}(x_1)$, but
it decreases very quickly with increasing, positive $\eta$.} Nevertheless,
parametrizations with relatively hard gluon density still predict a
significant decrease of the total 2--res cross--section as $\eta$ is increased
from 0 to 1. In contrast, parametrizations with a large and soft gluon content
(WHIT6, LAC1) predict the 2--res contribution to rise with increasing $\eta$,
reaching a maximum at $\eta \simeq 1.3$. This increase is entirely due to the
$qg$ final state, where the increase of $f_{G|e}$ now over--compensates the
decrease of $f_{q|e}$.
As discussed in the next subsection, the TOPAZ collaboration has proven
capable of detecting photon remnant jets with an efficiency of about 70\%;
this ability was already implicit in their observation \cite{141} of
considerable energy flow at small angles relative to the beam pipes. They are
now beginning to exploit this capability also in the jet analysis \cite{148}.
This shows that the 1--res, 2--res and direct contributions can indeed be
studied separately. On the other hand, Fig.~22 shows that even the sum over
all contributions should allow to discriminate between some existing
parametrizations of the parton content of the photon. In particular, at $p_T =
3$ GeV, parametrizations with a large but soft gluon density predict a flat or
even slowly rising jet cross--section as $\eta$ is increased away from zero,
in contrast to parametrizations with a small or hard gluon density.
Gluon--initiated processes are expected to contribute more than 50\% of the
total 2--res cross--section, and still some 25 to 35\% of the total di--jet
cross--section at small rapidity. This fraction, and hence the difference
between predictions using different parametrizations, will be smaller (larger)
for $p_T > (<) 3$ GeV.
Experiments at the LEP storage ring should also be able to contribute
significantly to our understanding of (almost) real \mbox{$\gamma \gamma $}\ collisions. At LEP1
the vicinity of the $Z$ peak means that multi--jet annihilation events are a
much more severe background than at TRISTAN; however, a parton--level
investigation \cite{149} concluded that this should not be much of a problem
as long as the invariant mass $M_{jj}$ of the two high$-p_T$ jet system is
below 15 to 20 GeV. Indeed, some experimental results have already been
published. The ALEPH collaboration repeated \cite{150} the AMY analysis
\cite{6} of multi--hadron production. Although some of the details differ, as
will be discussed in Sec.~5, ALEPH also finds that the description of the data
is greatly improved once resolved photon interactions are included in the
Monte Carlo model.
DELPHI \cite{151} reached similar conclusions regarding ``minimum bias"
multi--hadron production. In addition, they analyzed a subsample of events
containing two jets with $|\eta| \leq 1.0$ and $p_T > 1.75$ GeV, where the
jets were defined using a cluster algorithm. At these small transverse momenta
the jet reconstruction efficiency is rather low [$\sigma(2-{\rm jet}) \sim 0.1
\sigma($parton)], which introduces a strong dependence on the details of the
MC program. DELPHI finds that DG falls below the data for $p_T \leq 3$ GeV,
while the LAC1 and GS parametrizations work well. Note that their event sample
is mostly sensitive to direct events, as well as resolved photon events with
large $x$, i.e. soft remnant jets, since they require $W_{\rm vis} < 13$ GeV
and $E_{\rm vis} < 20$ GeV; the calculation of these quantities includes
information from the small--angle taggers, which cover angles down to
2.5$^{\circ}$ and should therefore see parts of the remnant jets. Fig.~3 shows
that the DG parametrization does indeed fall well below LAC1 for large $x$ and
small $Q^2$.
More recently, DELPHI has published \cite{152} a study of hadronic \mbox{$\gamma \gamma $}\
events where either the $e^+$ or the $e^-$ is tagged at a very small angle,
corresponding to a photon virtuality $P^2 \sim 0.1$ GeV$^2$. Since even
stronger cuts were imposed on the hadronic system than in the first DELPHI
analysis, the data sample was quite small (491 events); nevertheless it was
sufficient to once again prove the necessity to include resolved photon
contributions if the data are to be described by the MC program. This small
angle tagging technique holds much promise, since the tag helps to reconstruct
the kinematics of \mbox{$\gamma \gamma $}\ events by determining the energy of one of the
two photons. If (as at HERA) the energy of the second photon can be determined
using the Jacquet--Blondel method, a measurement of the jet rapidities in
di--jet events would allow to directly reconstruct the Bjorken$-x$ variables
of the partons in the photons. If, as TOPAZ results indicate, the DELPHI
small angle detectors (not to be confused with the very small angle taggers
used for the electron tagging) can be used for detecting the presence or
absence of remnant jets, DELPHI (and similar detectors) should be able to
perform quite detailed analyses of hard two--photon reactions. In order to
fully exploit this potential, the strong upper limits on the visible energy
and invariant mass used in refs.\cite{151,152} should be relaxed; this should
certainly be possible at LEP2 energies, where the annihilation background is
much smaller.
The LEP energy is expected to soon be increased to $\mbox{$\sqrt{s}$} \simeq 180$ GeV; LEP
experiments will then have the unique opportunity to study two--photon
cross--sections over a wide range of energies. As illustrated in Fig.~23, the
energy dependence can be quite strong. In this figure we show the
triple--differential cross--section at $p_T=5$ GeV as a function of $\eta_1 =
\eta_2 \equiv \eta$, imposing an antitag condition similar to that used by
ALEPH \cite{150}. For $\mbox{$\sqrt{s}$} = 90$ GeV, $x_T$ is similar to the value used in
Fig.~22, which leads to a similar shape of the pseudorapidity distribution;
the kinks at $\eta \simeq 1.5$ occur because the ALEPH antitag becomes
ineffective if the outgoing electron has less than half the beam energy, i.e
for scaled photon energy $z > 0.5$. Raising \mbox{$\sqrt{s}$}\ from 90 to 180 GeV increases
the direct contribution at $\eta = 0$ only by a factor of 1.45, while the
1--res and 2--res contributions increase by factors of 2.0 and 3.2,
respectively. This again demonstrates the strong dependence of the
cross--section for resolved photon processes on the available phase space, or,
equivalently, on the Bjorken$-x$ variables of eqs.(\ref{e4.2}). Note that the
single--jet inclusive cross--section at central rapdity grows even faster with
energy, since the kinematical integration limits for the rapidity of the
second jet also increase with \mbox{$\sqrt{s}$}. Finally, we remind the reader that the
shape and normalization of the solid curves in Fig.~22 could be quite
different if the photon has a large but soft gluon component; this would also
result in an even more rapid increase of the cross--section for resolved
photon events with \mbox{$\sqrt{s}$}.
LEP will almost certainly be the highest energy \mbox{$e^+ e^- $}\ storage ring ever. In
order to reach even higher energies, one will need to build a linear collider.
In such a device each bunch can most likely only be used once; the repitition
rate (number of bunch collisions per second) will therefore almost certainly
be smaller than at LEP. At the same time the total luminosity must increase
like the square of the beam energy in order to maintain a roughly constant
rate of \mbox{$e^+ e^- $}\ annihilation events. The luminosity per bunch crossing will
therefore have to be much larger than at existing storage rings, forcing one
to use very dense bunches. The correspondingly large charge density gives rise
to strong electromagnetic fields. When particles inside one bunch enter the
field produced by the opposite bunch, they will be accelerated, and will
therefore radiate real photons. This is known as ``beamstrahlung" \cite{134}.
The flux of these beamstrahlung photons depends quite sensitively on the
design characteristics of the collider; this is not surprising, since this
radiation is due to the field of the entire bunch, not due to individual
\mbox{$e^+ e^- $}\ collisions (unlike the only quasi--real ``bremsstrahlung" photons we
have dealt with up to now). Beamstrahlung can be reduced by using flat beams,
and by splitting large bunches into ``trains" of smaller ones. On the other
hand, for a given class of designs, beamstrahlung increases quite rapidly with
increasing beam energy.
This is ilustrated in Figs.~24, which show the single--jet inclusive
cross--section as a function of $p_T$, where we have accepted jets with
$|\eta| \leq 2.0$. We have used the WHIT1 parametrization, and imposed the
antitag requirement $\theta < 10^{\circ}$ when computing the bremsstrahlung
contribution to the total photon flux. The beamstrahlung spectrum has been
calculated using the analytical expressions of ref.\cite{134a}, for the JLC
design as specified at the 1993 international linear collider conference
\cite{153}. Designs for linear colliders are still evolving; the results of
Fig.~24 should therefore be considered as indicative only.
At $\mbox{$\sqrt{s}$}=0.5$ TeV (Fig.~24a) the beamstrahlung spectrum is considerably softer
than the equivalent bremsstrahlung spectrum (\ref{e2.10}). This enhances the
relative importance of the direct contribution, since the cross--section for
resolved photon contributions increases with the two--photon invariant mass
$W$ while that for the direct contribution decreases. For $p_T > 100$ GeV, the
total cross--section is dominated by directly interacting bremsstrahlung
photons. Notice that this design leads to a luminosity of about 50 fb$^{-1}$
per year; the two--photon cross--section should therefore remain measureable
for jets with $p_T$ well above 100 GeV. Moreover, one expects of the order
$10^8$ events per year with a jet with $p_T > 5$ GeV. This sounds like a large
number, but corresponds to a trigger rate of 10 Hz or less, which should be
easily manageable.
Increasing \mbox{$\sqrt{s}$}\ to 1.0 TeV (Fig.~24b) greatly increases the flux of
beamstrahlung photons, and also makes it harder. This enhances the relative
importance of resolved photon contributions, as can be seen from the $x_T$
value where direct and resolved photon cross--sections are equal. Notice also
that now beamstrahlung increases the cross--section by about a factor of 5.5
even at $p_T = 200$ GeV. Finally, comparing Figs.~24a and b, we see that the
jet cross--section increases by a factor of about 7 (20) for $p_T = 5$ (100)
GeV; without beamstrahlung, the corresponding factors would ``only" have
been 3 and 5, respectively. We remind the reader that these results depend on
the specific machine design; see refs.\cite{154,154a} for further discussions
of this point.
As mentioned in the beginning of this section, one might be able to convert a
linear \mbox{$e^+ e^- $}\ collider into a \mbox{$\gamma \gamma $}\ collider by back--scattering laser
photons off the $e^+$ and $e^-$ beams \cite{135}. The spectrum and luminosity
of such a collider depend quite strongly on details such as the polarization
of the laser photons and incident electrons beams. The production of jets at
such a collider has first been discussed in ref.\cite{155}. In Fig.~25 we
present a result for the simple case of unpolarized beams and small distance
between the conversion and interaction points. This leads to a photon spectrum
which peaks at $z=0.828$, where it also cuts off; the two--photon luminosity
is then quite flat over a wide range of $W$. Most modifications that have been
discussed in the literature \cite{135} give even harder photon spectra. By
comparing Fig.~25 with Fig.~24a we see that, as expected, the much harder
photon spectrum of the \mbox{$\gamma \gamma $}\ collider has greatly increased the relative
importance of resolved photon contributions. Moreover, the hard photon
spectrum allows to efficiently access soft partons in the photon via 1--res
contributions at large rapidity. As a result, the cross--section increases
with $\eta$ even for parametrizations with rather modest gluon content, like
WHIT1; this is to be contrasted with the situation at present and future
\mbox{$e^+ e^- $}\ colliders, see Figs.~22 and 23. This also explains the large difference
between the predictions from the LAC1 and WHIT1 parametrizations close to the
kinematical maximum of $\eta$, inspite of the rather large value of $x_T$.
Finally, we should warn the reader that from LEP2 energies onwards, multiple
interactions could substantially increase the true jet cross-section, compared
to the simple parton--level estimates of Figs.~23 to 25; we already saw in
Sec.~3a that this phenomenon seems to play an important role in jet
production from resolved photons at HERA. This will be discussed in more
detail in Sec.~5.
\setcounter{footnote}{0}
\subsection*{4b) Heavy Quark Production in \mbox{$\gamma \gamma $}\ Collisions}
Apart from the production of jets discussed in the previous subsection, the
production of heavy quarks is the only hard QCD process that has been
studied experimentally in two--photon collisions. As discussed in Sec.~3b, the
main advantage of heavy \mbox{$Q \bar{Q}$}\ pair production is that perturbative QCD
should be applicable over the entire phase space. However, as we already saw
in the discussion of the photoproduction of heavy quarks, predictions for
total \mbox{$c \bar{c}$}\ production rates are at present still quite uncertain, due to
unknown higher order corrections (which lead to a strong scale dependence) and
the uncertainty in the value of $m_c$ to be used here.
Progress in the theoretical treatment of heavy quark production in two--photon
processes has been relatively slow. The first complete LO calculation,
including resolved photon processes, was only performed in 1989 \cite{136}.
The contribution from twice resolved processes was found to be very small at
TRISTAN energies, but the 1--res contribution (from $\gamma g \rightarrow \mbox{$c \bar{c}$}$) is
quite sizable. Further LO predictions, for newer parton densities in the
photon, were published in \cite{41}. A full NLO analysis of the direct and
1--res contributions has been performed in ref.\cite{156}.
In Fig.~26 we show updated predictions for the total \mbox{$c \bar{c}$}\ cross--section in
the PETRA to LEP2 energy range. We have included NLO corrections to the direct
contribution, using a simple parametrization.\footnote{Note that here direct
and 1--res contributions remain well--defined even in NLO, since there is no
LO 1--res contribution from the quarks in the photon.} In ref.\cite{156} the
direct contribution to the total cross--section is written as
\begin{equation} \label{e4.4}
\sigma_{\rm dir}(\mbox{$\gamma \gamma $} \rightarrow \mbox{$Q \bar{Q}$} (g)) = \frac {\alpha^2_{\rm em} e_Q^4}
{m_Q^2} \left( c^{(0)}_{\gamma \gamma}
+ 4 \pi \alpha_s c^{(1)}_{\gamma \gamma} \right).
\end{equation}
Here $c^{(0)}_{\gamma \gamma}$ describes the well--known \cite{157}
tree--level (QPM) prediction, while $c^{(1)}_{\gamma \gamma}$ can be
parametrized as:
\begin{eqalignno} \label{e4.5}
c^{(1)}_{\gamma \gamma} &= \frac{\pi}{2} - \sqrt{r} \left( \frac{5}{\pi}
- \frac{\pi}{4} \right), \ \ \ \ \ r < 2.637 \nonumber \\
& = 0.35 r^{-0.3}, \hspace*{2.44cm} r \geq 2.637
\end{eqalignno}
where $r = W^2_{\gamma \gamma} / (4 m_Q^2) - 1$. This parametrization is
exact at threshold $r \rightarrow 0$, and describes the full NLO result to better than
10\% accuracy for all $r \leq 100$. On the other hand, the predictions for the
1--res contribution shown in Fig.~26 have been computed in LO only. One reason
is that here NLO corrections are considerably smaller than for the direct
contribution \cite{156}, largely because the threshold (``Sommerfeld")
corrections are negative for a colour--octet \mbox{$Q \bar{Q}$}\ state. Moreover, the
uncertainty of the prediction is much larger than for the direct
cross--section, making it less important to include 10\% corrections.
In Fig.~26 this uncertainty is given by the width of the bands defined by two
curves with the same pattern. Following ref.\cite{136} we have in all cases
required $\mbox{$W_{\gamma\gamma}$} > 2 m_D = 3.74$ GeV, but the ``dynamical" charm quark mass
appearing in the expressions for the hard cross--sections is less well
defined. In case of the direct cross--section we have varied $m_c$ between 1.3
GeV (upper dotted curve) and 1.6 GeV (lower dotted curve). Another uncertainty
arises from the choice of scale in \mbox{$\alpha_s$}\ and (for the 1--res contribution)
in \mbox{$G^{\gamma}$}. In Fig.~26 we have estimated this uncertainty by varying this scale
between $\mbox{$M_{c \bar{c}}$}/4$ (upper curve) to \mbox{$M_{c \bar{c}}$}\ (lower curve). In case of the direct
contribution the combined uncertainty only amounts to slightly over 20\%,
almost independently of \mbox{$\sqrt{s}$}. Note, however, that other authors \cite{95,156}
prefer to use an even wider range of values for $m_c$.
Unfortunately the uncertainty for the prediction of the 1--res contribution is
considerably larger than this. One reason is that now \mbox{$\alpha_s$}\ already appears
in the tree--level cross--section, leading to a stronger dependence on the
renormalization scale; there is also a factorization scale dependence in this
case. As usual, the inclusion of NLO corrections should reduce these scale
uncertainties. However, the biggest uncertainty comes from the choice of
$m_c$, and of the minimal allowed \mbox{$M_{c \bar{c}}$}. In case of the direct contribution one
always has $\mbox{$W_{\gamma\gamma}$} = \mbox{$M_{c \bar{c}}$}$, at least in leading order. On the other hand, 1--res
events have $\mbox{$W_{\gamma\gamma}$} > \mbox{$M_{c \bar{c}}$}$; it is then not clear whether one has to require
$\mbox{$M_{c \bar{c}}$} > 2 m_D$ in order to describe open charm production, or whether it is
sufficient to have $\mbox{$W_{\gamma\gamma}$} > 2 m_D$. If $\mbox{$M_{c \bar{c}}$} < 2 m_D < \mbox{$W_{\gamma\gamma}$}$, some energy has
to be transferred from the remnant jet to the ``hard" \mbox{$c \bar{c}$}\ pair; it has to
be remembered that in any case colour needs to be exchanged between these two
systems in the hadronization step. It seems unlikely to us that this soft
energy exchange can exceed 1 GeV, however. We have therefore used $m_c = 1.4$
GeV, $\mbox{$M_{c \bar{c}}$} > 2 m_c$ and scale $\mu = \mbox{$M_{c \bar{c}}$}/4$ in order to estimate the upper
limit of the 1--res contribution, and $m_c = 1.6$ GeV, $\mbox{$M_{c \bar{c}}$} > 2 m_D$ and $\mu
= \mbox{$M_{c \bar{c}}$}$ for the lower limit.
Fig.~26 shows that this uncertainty makes it impossible to distinguish between
the DG and WHIT1 parametrizations based on the total \mbox{$c \bar{c}$}\ production rate
alone. Note that the uncertainty is larger for LAC1, which has a very soft
gluon density and therefore reacts very sensitively to changes of the lower
bound of \mbox{$M_{c \bar{c}}$}. Even for the more conservative parametrizations, the
uncertainty amounts to roughly a factor of two. Nevertheless, some sensitivity
to the parton densities does remain; we will come back to this point shortly.
Finally, we note that even at LEP2 the 2--res contribution amounts to at most
5\% of the 1--res one \cite{156}; we have therefore not shown it in this
figure.
In the last few years several experimental studies of charm production in
\mbox{$\gamma \gamma $}\ collisions have been published. The first analysis, by the JADE
collaboration \cite{158}, looked for fully reconstructed charged $D^*$ mesons
in single--tag events. They reported a considerable excess over QPM
predictions, but this has not been confirmed by the TASSO \cite{159} or
TPC/2$\gamma$ \cite{160} collaborations, who took data at similar energies.
TASSO also measured the cross section for exclusive $D^0 \overline{D^0}$
production. The total \mbox{$c \bar{c}$}\ production cross--sections derived by these two
experiments agree with expectations \cite{156}.
The first study of charm production in two photon collisions at TRISTAN, by
the TOPAZ collaboration \cite{161}, also searched for fully reconstructed
$D^*$ mesons. Unfortunately the reconstruction efficiency is quite low,
leading to rather poor statistics (a few dozen events per experiment). TOPAZ
therefore also published a study \cite{162} based on $D^{\pm*} \rightarrow \pi_s^{\pm}
D^0$ decays, where only the soft pion needs to be detected; here the signal
after background subtraction consists of $372 \pm 54$ events. VENUS
\cite{163} and TOPAZ \cite{164} have also published analyses of hadronic
two--photon events containing an electron or positron (electron--inclusive
analysis); after subtracting backgrounds (mostly from Dalitz decays and photon
conversions), these experiments extract a charm signal with ${\cal O}(100)$
events each. Finally, very recently ALEPH presented \cite{165} results of an
analysis based on 33 fully reconstructed $D^{\pm*}$ mesons.
Unfortunately these experimental results are somewhat contradictory. All
TRISTAN experiments find some excess of events with high $p_T$; e.g., TOPAZ
\cite{162} reports a 2.9 $\sigma$ excess of events with $p_T(D^*) \geq 3.6$
GeV. This is actually not all that surprising, given that we had to include
contributions from the charm in the photon in order to reproduce the
production of central jets with $p_T \geq 2.5$ GeV; such ``charm excitation"
contributions are not included in present MC programs.\footnote{The NLO
correction to the direct process should describe $\gamma c \rightarrow g c$ scattering
accurately at these energies; however, it is not clear whether the
parametrized form of the NLO corrections used by TOPAZ \cite{161,162} treats
such contributions properly. Moreover, 2--res excitation contributions are not
included at all.} On the other hand, ALEPH \cite{165} finds a cross--section
for the production of charged $D^*$ mesons with $p_T \geq 2.0$ GeV that agrees
with the {\em lower} range of predictions; no excess is visible here.
Most of these experimental analyses are only sensitive to charmed hadrons with
significant transverse momentum. The notable exception is the TOPAZ inclusive
electron analysis \cite{164}; electrons with momentum as low as 400 MeV are
accepted, so that even charm quarks at rest can contribute to the signal.
Moreover, TOPAZ used their forward calorimeter, which covers angles down to
3.2$^{\circ}$ with respect to the beam pipe, to look for the presence of
photon remnant jets in the event; their MC predicts a jet tagging efficiency
of $73\pm2\%$. This allows them to study direct and resolved photon
contributions separately. The direct contribution agrees roughly with the
upper range of NLO predictions. The resolved photon contribution agrees with
NLO predictions using the LAC1 parametrization, while the DG prediction seems
to be at least two standard deviations below the data. This demonstrates that
this kind of analysis can lead to significant constraints on the gluon density
in the photon.
In order to compare experimental results on charm production with theoretical
calculations, one has to model the quark to hadron transition. This is more
difficult here than in the more familiar high--energy \mbox{$e^+ e^- $}\ annihilation
events, since for the low values of \mbox{$M_{c \bar{c}}$}\ or $p_T(c)$ relevant for present
two--photon data the effect of hard gluon radiation can {\em not} be absorbed
into fragmentation functions; this is only possible if gluons are
predominantly emitted collinear to the quarks, which requires $p_T \gg m_c$.
In this context it is interesting to note that the CLEO collaboration
\cite{166}, working at values of \mbox{$M_{c \bar{c}}$}\ only slightly above the upper range
probed by present two--photon experiments, found that popular fragmentation
functions could describe the observed $D^*$ meson spectrum {\em only} if the
emission of hard gluons was allowed explicitly. In particular, it was not
possible to describe the spectrum by simply convoluting the tree--level
\mbox{$c \bar{c}$}\ cross--section with the Peterson et al. fragmentation function
\cite{87}. Other fragmentation functions gave a better fit, but only if the
fragmentation parameters were chosen differently from those used for data at
higher energy; this is not surprising, since fragmentation functions are scale
dependent, just like parton distribution functions. It is therefore
encouraging to note that an NLO event generator for heavy quark production in
two photon collisions, written by J. Zunft, is now publicly available
\cite{167}.\footnote{This generator only allows the emission of a single hard
gluon; the generator used by CLEO to model \mbox{$e^+ e^- $}\ annihilation events included
$2 \rightarrow 4$ matrix elements.} Once hard gluon radiation has been included
explicitly, the fragmentation function only has to include nonperturbative
effects, which should indeed factorize to good approximation. Studies like
that by the TOPAZ collaboration \cite{164}, which are sensitive to charmed
hadrons at rest, can play an important role in testing this formalism, since
by definition fragmentation functions can only change the spectrum, but not
the total cross--section. Once the fragmentation process has been fully
understood, direct \mbox{$c \bar{c}$}\ pair production might be the best way to measure
the value of $m_c$ to be used in perturbative QCD calculations, since the
scale uncertainty is rather small here. A good knowledge of $m_c$ would help
to sharpen predictions for 1--res \mbox{$c \bar{c}$}\ production, as well as photo-- and
hadro--production of charm.
Fig.~26 shows that the total \mbox{$c \bar{c}$}\ pair production cross--section is
expected to grow quite rapidly with energy. However, at higher energies it
might be necessary to impose quite stringent cuts in order to extract a charm
signal; this will reduce the detectable cross--section significantly. For
example, in ref.\cite{156} it was found that requiring one of the charm quarks
to have rapidity $|y| \leq 1.7$ and $p_T \geq 5$ GeV reduces the
cross--section by almost a factor of 50. This still leaves us with at least
5,000 \mbox{$c \bar{c}$}\ events in 500 pb$^{-1}$ of data; however, at this point we have
not yet required anything that would actually identify these events as being
due to charm, e.g. a hard lepton or a reconstructed $D^*$ meson.
At a 500 GeV \mbox{$e^+ e^- $}\ linear collider even the DG parametrization predicts
\cite{154a,167a} the total \mbox{$c \bar{c}$}\ pair cross--section to reach a value
between 5 and 50 nb, depending on the amount of beamstrahlung generated.
However, requiring the event to contain at least one muon with rapidity $|y|
\leq 2$ and $p_T(\mu) \geq 5$ GeV reduces this cross--section by at least a
factor of 2,000.
As in case of jet production, cross--sections for the production of heavy
quarks can be boosted considerably by converting an \mbox{$e^+ e^- $}\ collider into a
\mbox{$\gamma \gamma $}\ collider by means of laser backscattering. However, at such a collider
QCD processes will likely be regarded primarily as backgrounds to ``new
physics" signals. In particular, one advantage of a \mbox{$\gamma \gamma $}\ collider is that it
can produce a Higgs boson as an $s-$channel resonance. If $m_H < 150$ GeV,
this Higgs boson will predominantly decay into into a \mbox{$b \bar{b}$}\ pair \cite{168}.
The direct (tree--level) background from $\mbox{$\gamma \gamma $} \rightarrow \mbox{$b \bar{b}$}$ can be reduced
significantly by chosing the two incident photons to be (predominantly) in
a $J_z = 0$ state \cite{169}. There is still a direct background from $\mbox{$\gamma \gamma $}
\rightarrow \mbox{$b \bar{b}$} g$, but this appears to be manageable \cite{170}. However, at least
if one uses a collider with a broad photon spectrum, which allows to search a
wide range of Higgs masses simultaneously, the dominant background will
actually come from resolved photon events. This has been pointed out in
ref.\cite{171}, where the 1--res $\gamma g \rightarrow \mbox{$b \bar{b}$}$ contribution is
studied. As emphasized in ref.\cite{172}, the exact size of this background
will also depend on the extent to which a gluon in a polarized photon is
itself polarized.
The to date most complete study of resolved photon backgrounds to the Higgs
signal at a broad--band \mbox{$\gamma \gamma $}\ collider has recently been performed by
Baillargeon et al. \cite{173}. They ignore polarization effects on the parton
densities in the photon (which amounts to the assumption $\Delta \mbox{$G^{\gamma}$} = 0$),
but include all possible processes, including $b$ and $c$ excitation
contributions; these are very large in the relevant kinematical region. They
include production of $c$ quarks, as well as light partons, in order to
estimate the effects of imperfect $b-$tagging.
An example is shown in Fig.~27, which shows the signal for a 120 GeV Standard
Model Higgs boson \cite{168} (assuming a \mbox{$b \bar{b}$}\ invariant mass resolution of
5 GeV), as well as various backgrounds. The electron beams have 175 GeV
energy, and 90\% polarization; 100\% polarized laser beams have been assumed.
This suppresses the direct backgrounds by about an order of magnitude.
Moreover, a $p_T$ cut of 30 GeV has been applied. We see that the largest
contribution to the single$-b$ inclusive cross--section comes from 2--res $b$
excitation processes (labelled as ``$bX_{2-res}$"). Depending on details of
the $b-$tagging efficiency, it is therefroe often advantageous to require both
high$-p_T$ jets to be tagged as $b$ quarks, or at least as heavy flavours.
Another conclusion of ref.\cite{173} is that Higgs searches at a \mbox{$\gamma \gamma $}\
collider become easier with increasing Higgs mass and {\em de}creasing beam
energy. The reason is that, as we have seen several times already,
cross--sections for resolved photon processes increase rapidly with \mbox{$\sqrt{s}$}, but
decrease equally quickly when the invariant mass of the hard system is
increased for fixed \mbox{$\sqrt{s}$}. In particular, for the case shown in Fig.~27, even
with a $b-$tagging efficiency comparable to present LEP detectors, a 7
$\sigma$ Higgs signal could be extracted from 10 fb$^{-1}$ of data, which
roughly corresponds to one year's running; for the same parameters, one could
at best hope to extract a 3.5 $\sigma$ signal at a 500 GeV collider. Since
one here probes the photon at very high momentum scales these conclusions do
not depend very sensitively on the parametrization chosen \cite{173}.
Nevertheless, Fig.~27 clearly shows that a good understanding of heavy quark
production in resolved photon interactions is mandatory if we ever want to
look for Higgs bosons with mass below 150 GeV at a \mbox{$\gamma \gamma $}\ collider.
\setcounter{footnote}{0}
\subsection*{4c) \mbox{$J/\psi$}\ Production in \mbox{$\gamma \gamma $}\ Collisions}
\mbox{$J/\psi$}\ production in principle offers a clean method for constraining \mbox{$G^{\gamma}$}\ in
two photon collisions \cite{136}. At least in the framework of the colour
singlet model \cite{112} and to leading order in \mbox{$\alpha_s$}, only resolved photon
processes can contribute to $\mbox{$\gamma \gamma $} \rightarrow \mbox{$J/\psi$} +$hadrons. Moreover, the 2--res
contribution is expected to be very small, just as in case of open heavy
flavour production discussed in the previous subsection. In LO the
cross--section is therefore essentially proportional to \mbox{$G^{\gamma}$}; the hard
sub--process, $\gamma g \rightarrow \mbox{$J/\psi$} + g$, is the same as in direct inelastic
photoproduction of \mbox{$J/\psi$}\ mesons, see Sec.~3d. As mentioned in that
subsection, NLO corrections to \mbox{$J/\psi$}\ production in $\gamma g$ fusion have
recently been computed \cite{116}. We use the results of this calculation to
update our previous LO predictions \cite{136} for \mbox{$J/\psi$}\ production at TRISTAN
energies, and also extend the calculation into the LEP energy range.
In order to simplify this calculation, we again use a parametrized form of the
NLO corrections. According to ref.\cite{116} the total cross--section for
$\gamma g \rightarrow \mbox{$J/\psi$} + X$, integrated over the region $Z \leq 0.9$ where $Z$ is
the inelasticity defined in eq.(\ref{e3.6}), can be written as:\footnote{In
ref.\cite{116} the factor $8/m^3_{J/\psi}$ has been written as $1/m^3_c$.
However, in the framework of the colour singlet model it seems more natural to
us to have the partonic cross--section scale like $1/m_c^2$, rather than
$1/m^5_c$. In the latter case the uncertainty related to the value of $m_c$
discussed below would obviously be significantly larger.}
\begin{equation} \label{e4.6}
\hat{\sigma}(\gamma g \rightarrow \mbox{$J/\psi$} +X) = \frac {\mbox{$\alpha_{\rm em}$} \alpha^2_s} {m_c^2}
\frac { 8 e_c^2 |\psi(0)|^2 } {m^3_{J/\psi}} \left\{ c^{(0)}(r) +
4 \pi \mbox{$\alpha_s$} \left[ c^{(1)}(r) + \bar{c}^{(1)}(r) \log \frac {\mu^2}{m_c^2}
\right] \right\},
\end{equation}
where $r = \hat{s}/m^2_{J/\psi} - 1$. The wave function at the origin
$|\psi(0)|^2$ can be derived from the measured leptonic partial width
\cite{117,116}. The function $c^{(0)}(r)$ describes the tree--level
cross--section \cite{112}; we do not impose any cut on $Z$ for this part,
since it remains finite in the limit $Z \rightarrow 1$, unlike the NLO correction,
which diverges in this limit. These corrections are given by the two functions
$c^{(1)}(r), \ \bar{c}^{(1)}(r)$, which can be parametrized as:
\begin{subequations} \label{e4.7} \begin{eqalignno}
c^{(1)}(r) &= \sqrt{r-0.1} \left[ \frac{ (r-0.8)(r-6.2) } {0.85 + 1.1r -0.1
r^2 + 2.7 r^{2.5}} - \frac {1} {1.22 -0.90 r + 1.14r^{1.5}} \right];
\label{e4.7a} \\
\bar{c}^{(1)}(r) &= (r-0.1)^{0.7} \frac {2.6-r} {1.4 - 1.2r + 3 r^{1.7}}.
\label{e4.7b}
\end{eqalignno} \end{subequations}
Note that the NLO corrections (\ref{e4.7}) remain finite as $r \rightarrow \infty$,
while the tree--level cross--section drops like $1/r^2$ in this limit.
In Fig.~28 we show results of a partial NLO calculation of \mbox{$J/\psi$}\ production
at current \mbox{$e^+ e^- $}\ colliders based on eqs.(\ref{e4.6}) and (\ref{e4.7}). We did
not include the small NLO contribution from $\gamma q$ scattering \cite{116};
moreover, the NLO direct contribution from $\mbox{$\gamma \gamma $} \rightarrow \mbox{$J/\psi$} + gg$ is not yet
available. Note also that $\psi'$ production with subsequent $\psi' \rightarrow \mbox{$J/\psi$}$
decay should increase the total \mbox{$J/\psi$}\ yield by another 15\% or so. The
results of Fig.~28 are for a no--tag situation, and include the suppression of
\mbox{$G^{\gamma}$}\ due to the virtuality of the photon \cite{78}. We checked that our
parametrization of the NLO cross--section shows significantly reduced scale
dependence, as expected, at least as long as the scale $\mu^2 \geq 1.5 m_c^2$;
the cross--section drops quickly for even smaller $\mu$. The same behaviour
has been observed in \cite{116} for the case of photoproduction of \mbox{$J/\psi$}. We
therefore estimate the scale dependence by varying $\mu^2$ between $m_c^2$
(lower curves) and $2 m_c^2$ (upper curves); the cross--section slowly
decreases again for larger values of $\mu$. In addition, we have varied $m_c$
in eq.(\ref{e4.6}) from 1.6 GeV (lower curves) to 1.3 GeV
(upper curves). Since we are now explicity producing a colour singlet state
already in the hard scattering process we have always required the $\gamma g$
cms energy to exceed $m_{J/\psi}$, even if $2m_c < m_{J/\psi}$.\footnote{In
ref.\cite{116} $r$ has also been written as a function of $m_c^2$, rather
than a function of $m^2_{J/\psi}$. However, this results in negative
cross--sections for $m_c > m_{J/\psi}/2$. It seems to us that the $m_c$
dependence of the $c$ functions should be treated on a par with corrections to
the non--relativistic approximation used in the colour singlet model.}
Fig.~28 shows that the uncertainty estimated in this way is still quite
sizable; most of this uncertainty, however, is due to the variation of the
factor $1/m_c^2$ in eq.(\ref{e4.6}). Notice also that the LO estimate (dotted
curve, for scale $\mu^2 = 2 m_c^2$ and $m_c$ = $m_{J/\psi}/2$) is quite close
to the lower range of NLO predictions. This indicates that the perturbative
result is relatively stable. Recall also that ref.\cite{116} finds good
agreement of their prediction with low--energy photoproduction data, although
leptoproduction data favour larger cross--sections. However, there might be
sizable corrections to the non--relativistic approximation intrinsic to the
colour singlet model \cite{115}; this has not been included in our estimate of
the theoretical uncertainty.
We see that the total expected cross--section for \mbox{$J/\psi$}\ production at TRISTAN
lies in the range from about 0.4 to 2.5 pb, depending on the parametrization
of \mbox{$G^{\gamma}$}. This corresponds to several hundred events per experiment in the
total TRISTAN data sample. However, one will almost certainly have to demand
that the \mbox{$J/\psi$}\ meson decays into an \mbox{$e^+ e^- $}\ or $\mu^+ \mu^-$ pair; this
reduces the rate by about a factor of 7.5 even before any acceptance cuts have
been applied. The cross--section increases quite rapidly with energy, possibly
exceeding 10 pb at LEP2; however, the acceptance at these higher energies
might also be poorer. A detailed MC study is necessary to decide whether the
study of \mbox{$J/\psi$}\ production in two photon events at current colliders is
feasible.
As usual, the cross--section at future linear \mbox{$e^+ e^- $}\ colliders is expected to
be even larger, partly due to contributions from beamstrahlung photons. Since
the invariant mass of the produced final state is quite small, the
cross--section is now largest for designs giving a large number of rather soft
photons. For example, at the 500 GeV TESLA collider (we again use the 1993
design \cite{153}) one expects a total \mbox{$J/\psi$}\ production cross--section
between 0.3 and 2.0 $\mu$b, or some $10^7$ events per year. At the same
collider the $\Upsilon(1s)$ cross--section is expected to lie between 0.25 and
0.85 pb, which still gives around 10,000 events per year. It is at present not
clear what fraction of these events would be detectable in the considerably
``dirtier" environment of such linear colliders.
\setcounter{footnote}{0}
\subsection*{4d) Direct Photon Production in \mbox{$\gamma \gamma $}\ Scattering}
The production of a direct photon in the collision of two (quasi--)real
photons, $\mbox{$\gamma \gamma $} \rightarrow \gamma \ X$, is another reaction that in LO only receives
contributions from resolved photon processes: The direct $2 \rightarrow 3$ process
$\mbox{$\gamma \gamma $} \rightarrow \mbox{$q \bar{q}$} \gamma$ is ${\cal O}(\alpha^3_{\rm em})$, and thus of
next--to--leading order compared to the 1--res contribution from $\gamma q
\rightarrow \gamma q$, as well as the 2--res contributions from $g q \rightarrow \gamma q$ and
$\mbox{$q \bar{q}$} \rightarrow g \gamma$, which are formally ${\cal O}(\alpha^3_{\rm
em}/\mbox{$\alpha_s$})$; the relevant hard scattering cross--sections can be found in
ref.\cite{173a}. There are additional LO contributions involving parton $\rightarrow$
photon fragmentation; as discussed in Sec.~3c, the corresponding fragmentation
functions are ${\cal O}(\mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$})$, which compensates for the relative
factor of $\mbox{$\alpha_s$}/\mbox{$\alpha_{\rm em}$}$ that appears in the hard sub--process cross--sections
for these contributions. However, these fragmentation contributions will in
general lead to softer photons, which are accompanied by a jet.
The only existing calculation \cite{173a} for this process omitted the
fragmentation contributions; this roughly corresponds to imposing stringent
isolation requirements for the direct photon. Ref.\cite{173a} uses LO
expressions; an NLO calculation is now under way \cite{38b}.
In Fig.~29 we show updated LO estimates\footnote{As discussed in Sec.~3c, care
has to be taken when computing the rapidities in this case. This has not been
described properly in ref.\cite{173a}, although the numerical results of this
paper are correct.}. We have required $p_T^{\gamma} \geq 1.5$ GeV, since QCD
perturbation theory would be very unreliable for even softer photons. We have
also applied the acceptance cut $|\eta_{\gamma}| \leq 1$ on the
(pseudo)rapidity of the emitted photon. In case of the WHIT1 parametrization,
2--res and 1--res contributions are shown separately, together with the sum;
for all other parametrizations only the total prediction is given. We see that
WHIT1 predicts the 2--res contribution to be very small in the PEP to TRISTAN
energy range, and to remain sub--dominant even at the highest LEP2 energy. On
the other hand, according to the LAC1 parametrization the 2--res contribution
should be dominant for $\mbox{$\sqrt{s}$} > 130$ GeV, due to the rapid rise of the
cross--section for $gq \rightarrow \gamma q$, which profits from the steep gluon
density assumed in this parametrization. In contrast, the GRV prediction for
the 2--res contribution at $\mbox{$\sqrt{s}$} =200$ GeV is about a factor of 1.6 below the
WHIT1 result, largely due to the smaller value of $\Lambda_{\rm QCD}$ that has
been assumed for GRV; note that we are probing QCD at rather small momentum
scales, where \mbox{$\alpha_s$}\ depends quite sensitively on $\Lambda_{\rm QCD}$.
The similarity of the various predictions shown in Fig.~29 is therefore
somewhat misleading; clearly one could gain more information if the 2--res and
1--res contributions can be separated experimentally, e.g. by remnant jet
tagging. Due to the large charge factor, the 1--res contribution mostly probes
the $u-$quark density in the photon. Its measurement might therefore yield
valuable information about the flavour structure of the photon at rather low
momentum scales, where nonperturbative contributions to the photonic parton
densities are still important. Whenever the 2--res contribution is sizable, it
is dominated by $gq$ scattering; it could therefore give us an additional
handle on the gluon content of the photon.
Unfortunately the cross--section is not large even if we allow $p_T^{\gamma}$
to be as small as 1.5 GeV; the NLO calculation now being performed \cite{38b}
will show how well QCD perturbation theory converges at such low momentum
scales.\footnote{When calculating the LAC1 prediction shown in Fig.~29 we had
to ``freeze" the parton densities, i.e. use $\min(4 \ {\rm GeV}^2,
(p_T^{\gamma})^2)$ as momentum scale, since this parametrization produces
negative parton densities if used at scales below the input scale $Q_0^2=4$
GeV$^2$.} Furthermore, the Bjorken$-x$ of the partons (or photon) ``in" the
electron can only be reconstructed if the rapidity of the jet balancing the
photon is also known. However, requiring $|\eta_{\rm jet}| \leq 1$ reduces the
cross--section by another factor of 1.7 (2.0) at $\mbox{$\sqrt{s}$} = 30$ (200) GeV;
moreover, the reconstruction efficiency for such a soft jet might be quite
low. On the positive side, Fig.~29 shows that each TRISTAN experiment should
have seen about 100 \mbox{$\gamma \gamma $}\ events with an isolated hard central photon in the
final state; at LEP2, several hundred such events should be accumulated by
each experiment. They are well worth looking for.
\setcounter{footnote}{0}
\section*{5) Beyond Perturbation Theory}
In this section we will address several topics that go beyond QCD perturbation
theory. We will be rather brief here, due to lack of both space and expertise
on our part. Nevertheless, we feel that we should at least comment on these
issues. For one thing, when comparing perturbative QCD calculations with data,
one nearly always has to correct for nonperturbative effects, usually with
the aid of Monte Carlo event generators; we have already commented on the MC
dependence of certain results in the preceding sections. Secondly, as remarked
in the Introduction, the study of reactions involving real photons might
provide us with new insight into the interplay of soft and hard QCD; given
that perturbative QCD is by now generally accepted as the theory of hard
hadronic interactions, much future research in strong interactions will
presumably be focussed on this transition region.
nonperturbative effects clearly play an important role in so--called minimum
bias events. This name derives from the (idealized) definition that in a
minimum bias event sample, each inelastic scattering event should be included,
without any experimental (trigger) bias. This is very hard to achieve in
practice, in particular for final states with low particle multiplicity and
small energy deposition in the detector. Measuring total inelastic
cross--sections is therefore not easy.
One usually distinguishes three classes of events in $\mbox{$\gamma p$}$ scattering. In
\underbar{quasi--elastic} events, $\mbox{$\gamma p$} \rightarrow Vp$, the proton remains intact while
the photon is transformed into a single vector meson $V$; usually only $\rho,
\ \omega, \ \phi$, and occasionally \mbox{$J/\psi$}, are included here. In
\underbar{diffractive} events either the photon or the proton (or both) gets broken
up into several hadrons, but no colour is exchanged between the two systems.
In such events there is usually a gap in rapidity space between the two
diffractive systems (in double diffractive events), or between the hadronic
system and the single $p$ or $V$ (in single diffraction). In contrast, in
\underbar{non--diffractive} events colour is thought to be exchanged between the
photon and the proton. Both get broken up (in case of direct interactions, the
photon is absorbed), and soft particles usually fill rapidity space more or
less uniformly.
All these components have to be modelled accurately for a complete
understanding of minimum bias events, and in particular if one wants to
extract the total \mbox{$\gamma p$}\ cross--section from the observed event rate. Both
HERA experiments have performed \cite{174,175} such measurements.
H1 \cite{175} has so far only published an analysis of 1992 data, while ZEUS
has also published \cite{174} results based on the much larger 1993 sample.
Both experiments use different MC models to describe the various event
classes, and then try to fit the relative weights to the data. For example,
H1 determines that about 26\% of all events are quasi--elastic or diffractive;
their result for the total \mbox{$\gamma p$}\ cross--section at $\mbox{$\sqrt{s}$}=195$ GeV
is\footnote{We have increased the published number by 8\%, since H1 used an
inaccurate expression for \mbox{$f_{\gamma|e}$}\ when converting $ep$ into $\mbox{$\gamma p$}$
cross--sections} $\sigma^{\rm tot}_{\gamma p}({\rm H1}) = (171 \pm 7 \pm 22)
\ \mu$b, where statistical and systematic errors are listed separately. The
1993 ZEUS data at $\mbox{$\sqrt{s}$} \simeq 180$ GeV favour a slightly larger quasi--elastic
$+$ diffractive event fraction ($\simeq 35\%$), but a lower total
cross--section: $\sigma^{\rm tot}_{\gamma p}({\rm ZEUS}) = (143 \pm 4 \pm 17)
\ \mu$b. Very recently, H1 announced \cite{176} preliminary results from a
special 1994 run with a minimum bias trigger. Their analysis indicates an even
larger quasi--elastic $+$ diffractive event fraction (around 40\%), and
$\sigma^{\rm tot}_{\gamma p}({\rm H1, \ prelim.}) = (172 \pm 3 \pm 10) \
\mu$b.
Note that the systematical uncertainties of these measurements are
significantly larger than the statistical errors. This indicates that our
understanding of minimum bias photoproduction events still needs to be
improved. Note also that the best description of non--diffractive events found
by ZEUS \cite{174}, based on a superposition of ``soft" and ``semi--hard"
events (see below), still has $\chi^2/d.o.f. > 2$. The detection efficiencies,
and hence the extracted value of $\sigma^{\rm tot}_{\gamma p}$, determined
from the same MC program may therefore also not yet be entirely reliable.
So far little effort has been made to understand diffractive events in the
framework of QCD. A possible connection \cite{177a} might be derived from the
observation of jets in single--diffractive \mbox{$p \bar{p}$}\ events by the UA8
collaboration \cite{177}, as well as the recent observation of di--jet events
with rapidity gap between the jets by the D0 collaboration \cite{178}. Very
recently the ZEUS collaboration announced \cite{179} preliminary results
indicating the presence of such di--jet events with gap between the jets at
HERA; ZEUS interprets them as resolved photon events with large \mbox{$x_{\gamma}$}.
Much more effort has been devoted to the understanding of non--diffractive
inelastic reactions. This is necessary even for the study of hard processes,
since the ``underlying event" in events with high$-p_T$ jets is closely
related to minimum bias events without any (obvious) hard interactions.
nonperturbative effects again play an important role in the description of
such events; however, it by now seems quite likely that semi--hard QCD
interactions, with partonic transverse momenta of order 1 to 2 GeV, are also
important here. Such interactions are said to lead to ``minijets": Even though
a perturbative (hard) scattering took place, the resulting ``jet" might be too
soft to pass experimental jet identification cuts.
The idea that such minijets might play an important role in minimum bias
hadronic physics dates back to 1973, when Cline et al. \cite{7} proposed that
the rapid increase of the inclusive jet cross--section with energy might be
the root cause of the observed increase of total hadronic cross--sections.
This jet cross--section is given by
\begin{equation} \label{e5.1}
\sigma_{ab}^{\rm jet} (s) = \int_{p_{T,{\rm min}}}^{\mbox{$\sqrt{s}$}/2} d p_T
\int_{4 p_T^2/s}^1 d x_1 \int_{4 p_T^2/(x_1 s)}^1 d x_2 \sum_{i,j,k,l}
f_{i|a}(x_1) f_{j|b}(x_2) \frac { d \hat{\sigma}_{ij \rightarrow kl}(\hat{s})}
{d p_T},
\end{equation}
where subscripts $a$ and $b$ denote particles ($\gamma, \ p, \dots$) and
$i, \ j, \ k, \ l$ are partons. $\hat{s} = x_1 x_2 s$ as usual, and
$\hat{\sigma}$ are hard partonic scattering cross--sections. Note that
$d \hat{\sigma} / d p_T \propto p_T^{-3}$; the cross--section defined in
eq.(\ref{e5.1}) therefore depends very sensitively on \mbox{$ p_{T,{\rm min}} $}, which is
supposed to parametrize the transition from perturbative to nonperturbative
QCD.
If $\mbox{$\sqrt{s}$} \gg \mbox{$ p_{T,{\rm min}} $}$, the integral in eq.(\ref{e5.1}) receives its dominant
contribution from $x_{1,2} \ll 1$. The relevant parton densities can then be
approximated by a simple power law, $f \propto x^{-J}$. In case of $pp$ or
\mbox{$p \bar{p}$}\ scattering, $a=b$ and the cross--section asymptotically scales like
\cite{180}
\begin{equation} \label{e5.2}
\sigma^{\rm jet} \propto \frac {1} {p^2_{T,{\rm min}}} \left( \frac {s}
{4 p^2_{T,{\rm min}}} \right)^{J-1} \log \frac {s} {4 p^2_{T,{\rm min}}},
\end{equation}
if $J>1$. For $J \simeq 1.3$, as measured by HERA, the jet cross--section
will therefore grow much faster than the total \mbox{$p \bar{p}$}\ cross--section, which
only grows $\propto \log^2 s$ (Froissart bound \cite{181}), or,
phenomenologically \cite{182} for $\mbox{$\sqrt{s}$} \leq 2$ TeV, $\propto s^{0.08}$.
Eventually the jet cross--section (\ref{e5.1}) will therefore exceed the total
\mbox{$p \bar{p}$}\ cross--section.
This apparent paradox is solved by the observation that, by definition,
inclusive cross--sections include a multiplicity factor. Since a hard partonic
scattering always produces a pair of (mini--)jets, we can write
\begin{equation} \label{e5.3}
\sigma_{ab}^{\rm jet} = \langle n_{\rm jet \ pair} \rangle
\sigma_{ab}^{\rm inel},
\end{equation}
where $\langle n_{\rm jet \ pair} \rangle$ is the average number of
(mini--)jet pairs per inelastic collision. $\sigma_{ab}^{\rm jet} >
\sigma_{ab}^{\rm inel}$ then implies $\langle n_{\rm jet \ pair} \rangle > 1$,
which means that, on average, each inelastic event contains more than one hard
partonic scatter. The simplest possible assumption about these multiple
partonic interactions is that they occur completely independently of each
other, in which case $n_{\rm jet \ pair}$ obeys a Poisson distribution. At a
slightly higher level of sophistication, one assumes these interactions to
be independent only at fixed impact parameter $b$; indeed, it seems natural
to assume that events with small $b$ usually have larger $n_{\rm jet \
pair}$. This leads to the eikonal formalism \cite{183}, where one writes the
cross--section as an integral over the two--dimensional impact parameter $b$:
\begin{equation} \label{e5.4}
\sigma_{ab}^{\rm inel}(s) = P_{ab}^{\rm had} \int d^2 b \left[ 1 - \exp \left(
- \frac { A_{ab}(b) \chi_{ab}(s) } { P_{ab}^{\rm had} } \right) \right].
\end{equation}
Here $P_{ab}^{\rm had}$ is the probability that both initial particles
are in a hadronic state when they interact (see below), and $A_{ab}$ describes
the transverse overlap of the partons. In realistic analyses \cite{184} it is
necessary to introduce both nonperturbative (soft) and hard contributions to
the eikonal $\chi_{ab}$:
\begin{equation} \label{e5.5}
\chi_{ab}(s) = \sigma_{ab}^{\rm soft}(s) + \sigma_{ab}^{\rm jet} (s),
\end{equation}
where $\sigma_{ab}^{\rm jet}$ is given by eq.(\ref{e5.1}). The connection to
the Poisson distribution then becomes evident from the identity
\begin{eqalignno} \label{e5.6}
\sigma_{ab}^{\rm inel} &= P_{ab}^{\rm had} \int d^2 b \left[ 1 - \exp \left(
- \frac { A_{ab}(b) \sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} } \right)
\right]
\nonumber \\
&+ P_{ab}^{\rm had} \int d^2 b \exp \left( - \frac { A_{ab}(b)
\sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} } \right) \left[ 1 - \exp \left(
-\frac { A_{ab}(b) \sigma_{ab}^{\rm soft} } {P_{ab}^{\rm had}} \right) \right]
\nonumber \\
&= P_{ab}^{\rm had} \sum_{n=1}^{\infty} \int d^2 b \exp \left(
-\frac { A_{ab}(b) \sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} } \right)
\cdot \left( \frac { A_{ab}(b) \sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} }
\right)^n \frac {1} {n!} \ \ + \cdots,
\end{eqalignno}
where the dots stand for the second line in eq.(\ref{e5.6}). Each term in the
sum corresponds to the cross--section for events with exactly $n$ hard
partonic scatters; the second line gives the cross--section for events without
any hard scatter [probability $= \exp \left( - \frac { A_{ab}(b)
\sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} } \right)$, for fixed impact
parameter $b$], but with a soft interaction. Note that the inclusive jet
cross--section (\ref{e5.1}) can be recovered from eq.(\ref{e5.6}) by
introducing a multiplicity factor $n$ in the sum:
\begin{eqalignno} \label{e5.7}
\sigma_{ab}^{\rm jet} &= P_{ab}^{\rm had} \sum_{n=1}^{\infty} n \cdot
\int d^2 b \exp \left( -\frac { A_{ab}(b) \sigma_{ab}^{\rm jet} }
{ P_{ab}^{\rm had} } \right) \cdot \left( \frac { A_{ab}(b)
\sigma_{ab}^{\rm jet} } { P_{ab}^{\rm had} } \right)^n \frac {1} {n!}
\nonumber \\
&= \int d^2b A_{ab}(b) \sigma_{ab}^{\rm jet},
\end{eqalignno}
since the function $A_{ab}$ describing the transverse overlap of the partons
in the two projectiles is by definition normalized such that $\int d^2 b
A_{ab}(b) = 1$.
We have already emphasized that, for fixed impact parameter, the ansatz
(\ref{e5.4}) assumes hard scatters to occur independently of each other. This
cannot be strictly true, since energy conservation implies that the maximal
Bjorken$-x$ for the partons in the second hard interaction is less than 1, but
it might still be a good approximation at high energies. A second assumption
is that the dependence of the parton densities on Bjorken$-x$ and on the
impact parameter $b$ factorizes; only in this case can the dependence on $b$
be parametrized by a single function $A_{ab}(b)$. At present we are not (yet?)
able to derive these properties from first principles. The ansatz (\ref{e5.4})
therefore goes beyond perturbation theory, even though it postulates that
perturbative interactions play an important role in minimum bias events; it
clearly needs to be checked against experimental data to assess its validity.
It has been shown \cite{184} that eq.(\ref{e5.4}) with $P_{p \bar{p}}^{\rm
had} = 1$ can be brought into agreement with data on total \mbox{$p \bar{p}$}\
cross--sections, with \mbox{$ p_{T,{\rm min}} $}\ around 1.5 GeV. In these calculations $A_{p
\bar{p}}(b)$ is usually computed from the Fourier transform of the charge form
factor of the proton; this amounts to the assumption that colour charges track
the distribution of electric charge in the nucleon. Moreover, the soft
cross--section is usually parametrized as
\begin{equation} \label{e5.8}
\sigma_{p \bar{p}}^{\rm soft} (s) = \sigma_0 + \frac {\sigma_1} {\mbox{$\sqrt{s}$}},
\end{equation}
where $\sigma_{0,1}$ are constants. Including \mbox{$ p_{T,{\rm min}} $}, one therefore has three
new free paramters (beyond those describing high$-p_T$ jet production) to fit
total and inelastic cross--sections (the elastic cross--section can be
calculated in this model using the optical theorem); the success of such a fit
is not entirely trivial. In particular, the rise of the total cross--section
appears very naturally here, although the rate of increase is determined by
\mbox{$ p_{T,{\rm min}} $}, and is thus fitted rather than predicted. These fits do have some
weaknesses, however. They basically ignore the existence of diffractive
events; more exactly, if the diffractive cross--section is supposed to be part
of $\sigma^{\rm soft}$, eq.(\ref{e5.8}), then eq.(\ref{e5.6}) leads to the
prediction that the fraction of diffractive events (which are part of the
second term, without hard scatters) decreases quickly with energy. A remedy
within the framework of the minijet model is possible \cite{185}, but only at
the cost of introducing additional free parameters. Secondly, existing
analyses typically assume that parton densities only increase slowly with
decreasing $x$, at least at scale $\mu^2 \simeq p^2_{T,\rm{min}}$; HERA DIS
data favour a much steeper behaviour \cite{13}. It is not clear whether the
faster increase of $\sigma^{\rm jet}$ predicted by such parton densities can
be compensated by re--fitting the free parameters of the model. We should also
keep in mind that total cross--section data can just as well be fitted in a
more conventional Pomeron picture \cite{182}.
Fortunately there is more direct evidence that an ansatz like eq.(\ref{e5.4})
can describe some features of hadronic interactions. Using eq.(\ref{e5.6}),
and the standard machinery to describe parton $\rightarrow$ hadron transitions, the
idea of having multiple partonic interactions within a single \mbox{$p \bar{p}$}\ reaction
can be built into an event generator \cite{186,187,188}. In ref.\cite{186} it
was shown that many features of \mbox{$p \bar{p}$}\ scattering as seen at the CERN SpS can
be understood in such a picture. These authors did not try to fit total
cross--section data; instead, \mbox{$ p_{T,{\rm min}} $}\ was determined from the measured
multiplicity distribution. It is encouraging that this again leads to a value
around 1.5 GeV. Multiple interactions then naturally lead to the observed
broad (non--Poissonian) multiplicity distribution. The perhaps greatest
success of this ansatz is that it reproduces the ``jet pedestal" effect. It
had been observed that events with hard jets ($E_T > 15$ GeV or so) also
contain more hadronic activity far away from the jet cores (in the
``pedestal") than minimum bias events do. Eq.(\ref{e5.4}) predicts such a
behaviour, since events with hard jets are likely to have small impact
parameter $b$, which significantly enhances the probability of having
additional semi--hard interactions (minijets) in the same events. Further
qualitative and quantitative successes of this ansatz are described in
refs.\cite{186,187}. However, it should be noted that these analyses again
assume a rather flat small$-x$ behaviour of the parton distribution functions.
The most direct confirmation of eq.(\ref{e5.4}) would obviously be the
observation of events with (at least) two independent jet pairs due to
multiple interactions. These jets should be pairwise back--to--back in the
transverse plane; the transverse opening angle should have a flat distribution
for these events. Both properties are quite distinct from QCD $2 \rightarrow 4$ events
(usually called ``double bremsstrahlung"). The cross--section for the
production of at least two independent jet pairs can be computed by inserting
the multiplicity factor $\left( \begin{array}{c} n \\ 2 \end{array} \right)$
into the sum in eq.(\ref{e5.6}), and summing all $n \geq 2$:
\begin{eqalignno} \label{e5.9}
\sigma_{ab}(\geq 2 \ {\rm jet \ pairs}) &= \frac {1} {2 P_{ab}^{\rm had}}
\left[ \sigma_{ab}^{\rm jet}(s) \right]^2 \int d^2b A^2_{ab}(b)
\nonumber \\
&\equiv \frac{1}{2} \frac{ \left[ \sigma_{ab}^{\rm jet} (s) \right]^2 }
{\sigma_{\rm eff}}.
\end{eqalignno}
Energy conservation will again necessitate a slight modification of this
simple expression.
The first experimental search for multiple interactions was performed by the
AFS collaboration \cite{189} at the CERN ISR ($pp$ collisions at $\mbox{$\sqrt{s}$}=63$
GeV). They required $\sum E_T({\rm jet}) > 28$ GeV, which implies $\sum_i x_i
\geq 0.45$. It is doubtful whether multiple interactions at these rather large
Bjorken$-x$ can be treated as independent; indeed, implementing energy
conservation here reduces the expected event rate by about a factor of 4.
Moreover, no complete calculation of the QCD $2 \rightarrow 4$ background processes
was available at the time, so that this background had to be modelled using a
leading logarithmic approximation in a region of phase space where no large
logarithms occur. The large signal reported by the AFS collaboration,
corresponding to $\sigma_{\rm eff} \simeq 5$ mb, therefore should be taken
with a grain of salt.
More recently the UA2 collaboration \cite{190} found some indication for the
existence of multiple interactions, but due to its rather small statistical
significance ($\sim 3$ standard deviations) they prefer to only quote the
lower bound $\sigma_{\rm eff} > 8.3$ mb. Finally, the CDF collaboration
\cite{191} found a 2.7 $\sigma$ signal, corresponding to $\sigma_{\rm eff}
\simeq 12$ mb, in the range expected by model calculations.
In the first theoretical studies \cite{192} of minijet production in \mbox{$\gamma p$}\
collisions, eq.(\ref{e5.4}) with $P^{\rm had}_{\gamma p} = 1$ was assumed to
hold also for the case of photoproduction. In this case eikonalization effects
are very small even at HERA energies, and a rather rapid increase of the total
cross--section was predicted. However, as first pointed out by Collins and
Ladinsky \cite{8}, $P^{\rm had}_{\gamma p}$ should be ${\cal O}(\mbox{$\alpha_{\rm em}$})$. This
enhances the exponent in eq.(\ref{e5.4}) by a factor of order $1/\mbox{$\alpha_{\rm em}$}$, so
that eikonalization does become relevant at HERA energies if $\mbox{$ p_{T,{\rm min}} $} \leq 2.5$
GeV. The necessity to have $P^{\rm had}_{\gamma p} \sim {\cal O}(\mbox{$\alpha_{\rm em}$})$ can
most easily be seen from eq.(\ref{e5.6}), once we recognize that
$\sigma_{\gamma p}^{\rm jet}$ is ${\cal O}(\mbox{$\alpha_{\rm em}$})$; if $P^{\rm had}_{\gamma p} =
1$, the production of additional minijet pairs would therefore be suppressed
by powers of \mbox{$\alpha_{\rm em}$}, even though only strong interactions are involved once the
photon has been transformed into a hadronic state. Notice also that $P^{\rm
had}_{ab}$ cancels out in eq.(\ref{e5.7}).
In ref.\cite{8}, $P^{\rm had}_{\gamma p}$ was taken to be the $\gamma \rightarrow
\rho$ transition probability $\simeq 1/300$. In ref.\cite{193} it was
suggested to instead estimate it as the momentum fraction carried by partons
in the photon at scale $\mu^2 \simeq p^2_{T,{\rm min}}$; this gives slightly
larger values $P^{\rm had}_{\gamma p} \simeq 1/200$. One can (roughly)
reproduce the HERA measurements with either number.
One might argue that $P^{\rm had}_{\gamma p}$ should really be of order
$\mbox{$\alpha_{\rm em}$}/\mbox{$\alpha_s$}$; after all, in eq.(\ref{e5.6}) one wants the production of a
second jet pair to be suppressed by a factor $\alpha_s^2$, not just a single
power of \mbox{$\alpha_s$}.\footnote{Recall that even the resolved photon contribution
to $\sigma_{\gamma p}^{\rm jet}$ is ${\cal O} (\mbox{$\alpha_{\rm em}$} \mbox{$\alpha_s$})$, since the
parton densities in the photon are ${\cal O} (\mbox{$\alpha_{\rm em}$} /\mbox{$\alpha_s$})$.} However, it
should be recognized that $P^{\rm had}_{\gamma p}$ cannot be discussed
independently of $A_{\gamma p}(b)$. In refs.\cite{193,194,195} $A_{\gamma
p}(b)$ was computed from the Fourier transform of the pion form factor; it is
not at all clear whether this describes the transverse distribution of partons
in a vector meson properly, and it is certainly not applicable to the
perturbative (``anomalous") component of the photon structure functions.
Mathematically, eq.(\ref{e5.4}) only depends on the combination $A_{ab}/P^{\rm
had}_{ab}$; one can see this by using the substitution $b' = b \cdot
\sqrt{P^{\rm had}_{ab}}$:
\begin{equation} \label{e5.10}
\sigma_{ab}^{\rm inel}(s) = \int d^2 b' \left[ 1 - \exp \left(
- A'_{ab}(b') \chi_{ab}(s) \right) \right],
\end{equation}
with $A_{ab}'(b') = A_{ab}\left(b'/\sqrt{P_{ab}^{\rm had}}\right) /
P_{ab}^{\rm had}$. Note that $\int d^2 A(b) = 1$ implies $\int d^2 b' A'(b') =
1$ as well; nevertheless $b'$ cannot be interpreted as the physical impact
parameter. Eq.(\ref{e5.10}) shows that one can always compensate an increase
in $P_{ab}^{\rm had}$ by a steeper decrease of $A_{ab}(b)$ at large $b$, which
implies an increase of $A(b)$ at small $b$. One can therefore simply fix
$P^{\rm had}_{\gamma p}$ to some value and fit $A_{\gamma p}(b)$ to data, or
vice versa; this approach was followed in ref.\cite{195}.
However, in case of photons the ansatz (\ref{e5.4}) might in any case be too
simple. As already discussed in the Introduction, the hadronic structure of
the photon is generally assumed to have both perturbative and
nonperturbative components. The corresponding contributions to the parton
densities have quite different $x$ and $Q^2$ dependence. It therefore makes
sense to split the cross--section (\ref{e5.4}) into (at least) two terms,
characterized by different values of $P^{\rm had}$, different parton densities
[and hence different $\sigma^{\rm jet}$, see eq.(\ref{e5.1})], and presumably
also different overlap functions $A(b)$. Yet another term has to be introduced
to describe direct interactions; these cannot be eikonalized, since here the
photon is ``used up" after the first interaction.
Two different ans\"atze of this kind have been proposed so far. Both Honjo et
al. \cite{196} and Schuler and Sj\"ostrand \cite{197,36a} split the hadronic
photon into a discrete sum over vector mesons, and a perturbative
(``anomalous") contribution; symbolically
\begin{equation} \label{e5.11}
| \gamma \rangle = | \gamma \rangle_{\rm bare} + \sum_{\rho,\omega,\phi}
\frac {e}{f_V} |V \rangle + \sqrt{P_{q \bar q}} | \mbox{$q \bar{q}$} \rangle,
\end{equation}
where $| \gamma \rangle_{\rm bare}$ is the ``direct" photon\footnote{Strictly
speaking the coefficient of the first term in eq.(\ref{e5.11}) should differ
from unity by an amount of order \mbox{$\alpha_{\rm em}$}; this effect can safely be ignored.},
$e/f_V$ are the $\gamma \rightarrow V$ transition amplitudes, $|V \rangle$ is a vector
meson state, and $| \mbox{$q \bar{q}$} \rangle$ is a state that develops from a hard
$\gamma \rightarrow \mbox{$q \bar{q}$}$ splitting. The sum in eq.(\ref{e5.11}) is assumed to be
incoherent, so that each term also contributes separately to the parton
densities and to the total \mbox{$\gamma p$}\ cross--section. In refs.\cite{196,197} the
vector meson contributions to the parton content of the photon were described
in terms of pion structure functions, while in ref.\cite{36a} the shapes of
these contributions were fitted from \mbox{$ F_2^\gamma $}\ data (SaS parametrization; see
Sec.~2b). The perturbative contribution to eq.(\ref{e5.11}) is really a
continuum of states; its contribution to the parton densities can be written
as \cite{36a}
\begin{equation} \label{e5.12}
f_{i|\gamma}^{\rm pert} (x,\mu^2) = \frac {\mbox{$\alpha_{\rm em}$}}{\pi} \sum_q
\int_{k_0^2}^{\mu^2} \frac {d k^2}{k^2} f_{i|q \bar q} (x,\mu^2, k^2).
\end{equation}
Note that the perturbative $\gamma \rightarrow \mbox{$q \bar{q}$}$ transition has been factored
out here; it gives rise to the factor $dk^2/k^2$. The ``state distributions"
$f_{i|q \bar q}$ therefore obey {\em homogeneous} evolution equations in the
factorizatioon scale $\mu^2$, with the (leading order) boundary conditions
\cite{36a}
\begin{subequations} \label{e5.13} \begin{eqalignno}
f_{q|q \bar q} (x,k^2,k^2) &= f_{\bar{q}|q \bar q} (x,k^2,k^2) = \frac{3}{2}
\left[ x^2 + (1-x)^2 \right]; \label{e5.13a} \\
f_{q'|q \bar q} (x,k^2,k^2) &= f_{G|q \bar q} (x,k^2,k^2) = 0. \label{e5.13b}
\end{eqalignno} \end{subequations}
At this point the treatment of refs.\cite{196} and \cite{197,36a} diverges.
Schuler and Sj\"ostrand do not attempt to predict the total \mbox{$\gamma p$}\
cross--section. Rather, the emphasis of their work is on the properties of
photoproduction events, both minimum bias events and events containing hard
jets. They therefore assume a parametrization of the total cross--section as
in ref.\cite{182}. The overall normalization of the nonperturbative
contributions to the photon structure functions, given by the $f_V$, is fixed
by low energy data. This also determines the contribution of this component to
the total cross--section, since the $Vp$ cross--sections are assumed to be
identical to the $\pi p$ cross--section. Moreover, they assume that the
perturbative component (\ref{e5.12}) only contributes to hard scattering
events, which are very rare at $\sqrt{s(\mbox{$\gamma p$})} \simeq 10$ GeV. However, their
choice of the $f_V$ gives a contribution to $\sigma^{\rm tot}_{\gamma p}$ which
falls about 20\% below the data at these low energies. The difference then has
to come from direct interactions. This forces them to include direct $\mbox{$\gamma p$}$
scattering with partonic transverse momentum as low as 0.5 GeV.\footnote{In
ref.\cite{197} Schuler and Sj\"ostrand modify the proton structure functions
for scales $\mu^2 < 5$ GeV$^2$ and/or small $x$, in order to enforce a smooth
transition between DIS and photoproduction. However, in ref.\cite{36a} they
use standard leading twist QCD to describe \mbox{$ F_2^\gamma $}\ for momentum scales down to
about 1 GeV$^2$.} The same value is also used for the cut--off parameter $k_0$
in eq.(\ref{e5.12}); this is in accordance with ref.\cite{196}, as well as
(approximately) with the GRV \cite{42} and AGF \cite{47} parametrization.
\setcounter{footnote}{0}
Ref.\cite{197} devotes much attention to the description of quasi--elastic and
diffractive events, which are entirely due to contributions from the
$|V \rangle$ states. These states also contribute to soft and semi--hard
(minijet) non--diffractive \mbox{$\gamma p$}\ interactions, where the cut--off $\mbox{$ p_{T,{\rm min}} $}
\simeq 1.3$ GeV at $\mbox{$\sqrt{s}$}=200$ GeV is fixed from multiplicity distributions of
\mbox{$p \bar{p}$}\ events at the same energy. Schuler and Sj\"ostrand allow for multiple
interactions in this sector, but assume them to be completely independent of
each other; in contrast, the ansatz (\ref{e5.4}) assumes independent
scattering {\em only} at fixed impact parameter $b$. We have seen above that
the correlation between the presence of hard jets and small $b$ offered a
natural explanation of the jet pedestal effect. Assuming completely
independent interactions means that every partonic interaction occurs with
probability
\begin{equation} \label{e5.14}
P_{\rm jet}(p_T) = \frac {1} { \sigma^{\rm nd}_{Vp} } \frac
{d \sigma^{\rm jet}_{Vp} } {d p_T},
\end{equation}
where the superscript ``nd" stands for non--diffractive. Finally, in this
model no multiple interactions are allowed to originate from the perturbative
contribution (\ref{e5.12}) to the photon structure functions. In order to
maintain the assumed $s^{0.08}$ behaviour of the total \mbox{$\gamma p$}\ cross--section
it then becomes necessary to increase \mbox{$ p_{T,{\rm min}} $}\ (in this sector only!) linearly
with \mbox{$\sqrt{s}$}, so that $\mbox{$ p_{T,{\rm min}} $} \simeq 2.2$ GeV for HERA ($\sqrt{s(\mbox{$\gamma p$})} \simeq
200$ GeV).\footnote{In an alternative version of this model, \mbox{$ p_{T,{\rm min}} $}\ is held
fixed in the perturbative sector, while the $f_V$ are assumed to decrease with
energy. Clearly this can at best be a temporary solution, since eventually the
minijet contribution from the perturbative sector alone will exceed the
assumed total \mbox{$\gamma p$}\ cross--section unless it is unitarized in some way.
Schuler and Sj\"ostrand therefore favour the variant of their model with
constant $f_V$. They also discuss a few other versions, not of all of which
are meant to be potentially realistic.}
The great advantage of the model of refs.\cite{197,36a} is that is comes
pre--packaged in the successful LUND/PYTHIA Monte Carlo event generator
\cite{198}. This allows to test many aspects of the model against data from
HERA, as well as from experiments at lower energies. On the positive side, not
only does the total \mbox{$\gamma p$}\ cross--section seem to follow the ``universal"
$s^{0.08}$ behaviour, but the relative sizes of the contributions from
quasi--elastic, diffractive, and non--diffractive events also match more or
less the model predictions. The recent observation by the ZEUS collaboration
\cite{62c} that the transverse momentum $k_T$ of the photon remnant jet in
events with at least two high$-p_T$ jets is much harder than the Gaussian with
width $\sim 0.4$ GeV characteristic for hadronic remnant jets can also be
regarded as a success of this model; indeed, it has been predicted more than
ten years ago \cite{200} that the perturbative contribution should have a
$k_T$ distribution $\propto d k_T^2/k_T^2$ for large $k_T$, which is directly
related to the ansatz (\ref{e5.12}). Moreover, the observation \cite{80} that
in the region below about 1 GeV, the shape of the $p_T-$distribution of
charged particles at HERA closely matches that of \mbox{$p \bar{p}$}\ events at the same
energy indicates the presence of a component in the hadronic photon that
behaves similarly to other hadrons. At higher $p_T$, the spectrum becomes
harder, in agreement with perturbative calculations \cite{81} that include
direct contributions as well as contributions from the perturbative part of
the photon structure.
There also appear to be some problems with this model in its present form,
however. We already mentioned that the properties of the ZEUS minimum bias
sample \cite{174} cannot be described properly by PYTHIA. Moreover, when
trying to extract partonic cross--sections (and, ultimately, \mbox{$G^{\gamma}$}) from the
measured di--jet cross--section, the H1 collaboration observed \cite{60} that
switching on multiple partonic interactions in PYTHIA is not sufficient to
describe the energy flow in resolved photon events.\footnote{Since including
multiple interactions improved agreement with the data, H1 implicitly assumed
that the remaining difference is due to even more partonic interactions; since
these produce a jet pedestal that is uniform in $\phi$, they can then simply
subtract this pedestal from the measured $E_T$ of the jet to arrive at the
partonic $p_T$ (up to showering and fragmentation effects, which are
presumably described adequately by the model). However, this treatment would
not give the correct answer if the additional energy flow was (partly) due to,
e.g., enhanced initial state radiation, which is not uniform in $\phi$.} It is
at present not clear whether this problem can be solved by replacing the
simple ansatz (\ref{e5.14}) by something like eq.(\ref{e5.4}) (for the
contribution from the $|V \rangle$ states), or whether a more substantial
modification of the model will be necessary.
Finally, we find the introduction of three different scales that are meant to
separate perturbative and nonperturbative interactions not very appealing.
One might argue that \mbox{$ p_{T,{\rm min}} $}\ ought to be larger for the pertubative $| \mbox{$q \bar{q}$}
\rangle$ component than for the $|V \rangle$ components, since the state
that develops from a hard $\gamma \rightarrow \mbox{$q \bar{q}$}$ splitting may have a smaller
transverse size (as we will see shortly, this is assumed in ref.\cite{196}),
so that a larger momentum transfer becomes necessary to resolve individual
colour charges. (A similar argument is used in ref.\cite{195} for the photon
as a whole.) However, this does not explain why $k_0$, which is also the
minimal $p_T$ of direct interactions, is so much smaller than even the value
of \mbox{$ p_{T,{\rm min}} $}\ used for the semi--hard interactions of the $|V \rangle$
components. Alternatively, one might argue that quark exchange (which
dominates direct interactions and, obviously, $\gamma \rightarrow \mbox{$q \bar{q}$}$ splitting)
is perturbative down to significantly smaller momentum scales than gluon
exchange (which dominates minijet production in all resolved photon
interactions). This could explain why $k_0$ is smaller than \mbox{$ p_{T,{\rm min}} $}, but the
difference between the \mbox{$ p_{T,{\rm min}} $}\ values used for the $|V \rangle$ and $|\mbox{$q \bar{q}$}
\rangle$ components cannot be explained by such an argument. The different
energy dependence of the two \mbox{$ p_{T,{\rm min}} $}\ values is also difficult to understand
intuitively.
In this respect the model of ref.\cite{196} seems somewhat more appealing.
Here a similar value for the cut--off $k_0$ in eq.(\ref{e5.12}) is used, but
all minjet production, from direct photons as well as the different classes
of resolved photons, is assumed to be regularized by a single, energy
independent parameter \mbox{$ p_{T,{\rm min}} $}. A fit to low energy data, which show a small
but significant increase of the total \mbox{$\gamma p$}\ cross section between 10 and
about 18 GeV (the highest energy that had been reached in fixed target
experiments), gives $\mbox{$ p_{T,{\rm min}} $} \simeq 1.5$ GeV. Since this is significantly
larger than $k_0$, Honjo et al. also allow the $|\mbox{$q \bar{q}$} \rangle$ states to
have nonperturbative interactions, which are assumed to scale $\propto
1/k^2$; the idea here is that the ``hardness" $k^2$ of the $\gamma \rightarrow \mbox{$q \bar{q}$}$
splitting determines the transverse size of the state that develops from this
splitting. Minijet production from this state is also eikonalized, where the
typical transverse size [which determines $A(b)$] again scales like $1/k^2$;
that is, eikonalization \`a la eq.(\ref{e5.4}) occurs for each value of $k$
independently. As written in ref.\cite{196} the model predicts a total \mbox{$\gamma p$}\
cross--section at $\mbox{$\sqrt{s}$} = 200$ GeV between 190 and 250 $\mu$b, the lower end of
which is compatible at least with the H1 measurement \cite{175}; presumably
the agreement could be improved if one allowed the $x$ and/or $b$ dependence
of the partons in the $|V \rangle$ states to differ from those in the pion.
However, given the assumptions involved, we do not think that total
cross--section data can be used to constrain parton distribution functions.
Note that this model does predict a jet pedestal effect; it might even be
stonger than in \mbox{$p \bar{p}$}\ collisions, since for the perturbative contribution to
the photon structure functions a very hard interaction not only implies small
$b$, but also (at least on average) a larger value of $k^2$ and hence a
narrower $A(b)$, which increases the chances for additional semi--hard
interactions. Unfortunately this model has not yet been built into an event
generator.
We finally mention two slightly different approaches to \mbox{$\gamma p$}\ scattering. In
ref.\cite{201} it was observed that, at least in a single--component model of
the photon\footnote{A single component description was used here since in
ref.\cite{201a} the perturbative component of the photon structure functions
had been found to be too small to have much impact on minijet production.
However, comparison with the explicit calculation of ref.\cite{36a} shows that
the simple estimate of ref.\cite{201a}, which used the double scaling limit of
QCD, greatly under--estimates the perturbative contribution to \mbox{$G^{\gamma}$}\ at small
$x$ and scale $\mu^2 \simeq p^2_{T,{\rm min}}$.}, where $P_{\gamma p}^{\rm
had}$ and $A_{\gamma p}(b)$ are determined from ``canonical" VDM ideas, it is
difficult to simultaneously describe the rise of $\sigma^{\rm tot}_{\gamma p}$
between $\mbox{$\sqrt{s}$}=10$ and 18 GeV, and the rather small value of the total
cross--section measured at HERA (or at least the smaller ZEUS result
\cite{174}). These authors therefore allowed the {\em soft} contribution to
the cross--section to grow $\propto s^{0.058}$, similar to the behaviour found
earlier in ref.\cite{202}. Adding a minijet contribution estimated using the
DG parametrization with $\mbox{$ p_{T,{\rm min}} $}=3.0$ GeV (where eikonalization effects are
still quite small for HERA energies) then gives $\sigma^{\rm tot}_{\gamma
p}(\mbox{$\sqrt{s}$}=200 \ {\rm GeV}) \simeq 160 \ \mu$b. A similar approach is taken in
ref.\cite{203}. This work attempts a comprehensive description of \mbox{$\gamma p$}\
scattering, including quasi--elastic and diffractive events, in the framework
of the Dual Parton Model \cite{204}; it is now being implemented in the PHOJET
event generator. Since the rise of the soft cross--section is now also fitted
from data, in these models one can no longer claim to explain the rise of
total cross--sections in terms of minijets. Moreover, $\mbox{$ p_{T,{\rm min}} $} \simeq 3.0$ GeV
seems rather large to us; there is evidence from \mbox{$J/\psi$}\ decays, DIS, and
\mbox{$e^+ e^- $}\ annihilation that perturbative QCD is applicable at lower momentum
scales. This is also supported by analyses of \mbox{$\gamma \gamma $}\ data, to which we turn
next.
Once the description of the photon has been fixed, e.g. by fitting the free
parameters of models like those of refs.\cite{196,197,203} from \mbox{$\gamma p$}\ data,
the properties of \mbox{$\gamma \gamma $}\ interactions can in principle be predicted
unambiguously \cite{205}. However, two--photon experiments have needed
comprehensive event generators well before the recent sophisticated models of
\mbox{$\gamma p$}\ scattering were developed. We already discussed in Sec.~2b that the
determination of Bjorken$-x$ in deep--inelastic $e \gamma$ scattering
necessitates a model of the hadronic final state; in Sec.~4 we mentioned that
the event generators that were used to model quasi--real \mbox{$\gamma \gamma $}\ scattering
prior to the pioneering AMY study \cite{6} did not include resolved photon
processes at all.
The situation has certainly improved a great deal since then; however, there
are still some problems. At present most experiments use the same basic ansatz
to describe real \mbox{$\gamma \gamma $}\ scattering, which contains three separate
contributions. Soft interactions, characterized by an exponential $p_T$
spectrum, are modelled using VDM ideas.\footnote{This ``VDM component" is not
to be confused with the nonperturbative contribution to photon structure
functions, which is often also estimated from the VDM. As emphasized in
refs.\cite{197,205}, these ``VDM partons" also participate in hard resolved
photon interactions.} The second component is estimated from the QPM. At high
$p_T$ it coincides with the direct contribution introduced in Sec.~4, but the
$p_T \rightarrow 0$ divergence is here regularized by constituent quark masses
(typically $m_u=m_d \simeq 300$ MeV, $m_s \simeq 500$ MeV), rather than by a
momentum cut--off. This QPM contribution therefore extends to (partonic)
$p_T=0$, and is hence not entirely perturbative.\footnote{Recall that in the
model of refs.\cite{197,205} direct interactions with partonic $p_T$ as low as
500 MeV are allowed; in practice this should give a similar contribution as
the QPM prediction using constituent quark masses.} In ref.\cite{206} it has
been shown that data on total \mbox{$\gamma \gamma $}\ cross--sections and \mbox{$ F_2^\gamma $}\ can be fitted
from these two components only. However, the description of multi--hadron and
jet production in \mbox{$\gamma \gamma $}\ collisions is only possible if one also introduces
single and double resolved contributions, which form the third component of
current \mbox{$\gamma \gamma $}\ event generators.
So far all currently used generators agree. There are some differences in the
details, however. For example, the TRISTAN experiments \cite{6,141,142}, as
well as DELPHI \cite{151,152} use parameter values for the description of the
soft (``VDM") component that have been determined by experiments at lower
energies. This might be dangerous, since these earlier analyses did not allow
for resolved photon interactions; there is a strong correlation between
the assumed size of the soft component and the parameter \mbox{$ p_{T,{\rm min}} $}, which
essentially fixes the size of the resolved photon contribution to the minimum
bias event sample. The ALEPH collaboration \cite{150} therefore preferred to
fit both \mbox{$ p_{T,{\rm min}} $}\ and the parameters describing the soft contribution from
their own data. This might explain the rather large value of \mbox{$ p_{T,{\rm min}} $}, 2.5 GeV,
obtained by ALEPH for the DG parametrization, compared to 1.45 GeV for DELPHI
and 1.6 to 2.0 GeV for AMY and TOPAZ. The expression for the total \mbox{$\gamma \gamma $}\
cross--section derived by ALEPH also differs significantly from conventional
VDM expectations. It should be noted that the jet cross--sections derived by
TRISTAN experiments \cite{141,142} are based only on high$-p_T$ events; these
analyses are therefore not sensitive to the assumed value of \mbox{$ p_{T,{\rm min}} $}.
Furthermore, the experimental definition (trigger conditions) of the ``minimum
bias" sample differs quite significantly between the various experiments,
with ALEPH having the loosest cuts and largest visible cross--section, while
AMY uses quite stringent requirements. Of course, a complete event generator
should still give the same values (within errors) for its free parameters even
when quite different data samples are used for the fit. The discrepancies
between the values of \mbox{$ p_{T,{\rm min}} $}\ determined by different experiments therefore
indicates that (some) present generators are not complete.
One problem of these generators is that they do not include initial and final
state radiation (parton showers). ALEPH states \cite{150} that generators like
PYTHIA \cite{198} and HERWIG \cite{207} with parton showering switched on did
not describe the data well. However, QCD tells us that such radiation must
exist at some level. It might, for example, change the $p_T-$dependence of the
partonic cross--sections extracted by TOPAZ \cite{141} and AMY \cite{142},
although the string fragmentation scheme used in these analyses may mimick
some of these effects; a preliminary TOPAZ study \cite{148} finds that their
MC program gives a reasonable, but not perfect description of the distribution
of the transverse opening angle $\phi$ between the jets in di--jet events.
Note that initial state radiation should be treated differently for photons
than for real hadrons \cite{197,208}, due to the presence of the hard $\gamma
\mbox{$q \bar{q}$}$ vertex. Whenever one reaches this vertex in the standard backward
evolution \cite{209} of initial state radiation, the shower has to be
terminated, even if the parton's virtuality is still quite high at this point.
This feature is included in the latest versions of PYTHIA and HERWIG, at least
in an average ($x-$independent) sense.
\setcounter{footnote}{0}
Another shortcoming of present \mbox{$\gamma \gamma $}\ event generators is that they do not
allow for the hard transverse momentum distribution of the remnant jets that
has been predicted by theory \cite{200} and seen experimentally by ZEUS
\cite{62c}. This seems to have little impact on the $p_T-$spectrum of the hard
jets \cite{62c}, but might change the predicted efficiency for detecting
remnant jets, e.g. in the TOPAZ detector \cite{210a,141,164}.
Finally, these event generators at present do not allow for multiple partonic
interactions. Even if only the contribution from the nonperturbative part
of photon structure functions is eikonalized, multiple interaction effects
will play a significant role from LEP energies onward \cite{210}. In a
single--component model of the photon, eq.(\ref{e5.4}) with $P^{\rm
had}_{\gamma \gamma} \sim \alpha^2_{\rm em}$, four jet events due to multiple
parton scattering might already be detectable at TRISTAN\footnote{Even in this
case they are not expected to significantly affect the properties of events
with one or two high$-p_T$ jets; they do therefore not invalidate the
analyses of refs.\cite{141,142}}, and ought to become a prominent feature of
two--photon events at LEP2 energies \cite{211}. The study of \mbox{$\gamma \gamma $}\ collisions
at the highest available energies should therefore tell us how to unitarize
the contribution to minijet production from the perturbative (``anomalous")
component of the photon structure.
Apart from its intrinsic interest, a good understanding of minimum bias \mbox{$\gamma \gamma $}\
events is necessary for a realistic evaluation of possible hadronic
backgrounds at future linear \mbox{$e^+ e^- $}\ colliders. In ref.\cite{212} it was
pointed out that in designs with strong beamstrahlung, several hadronic \mbox{$\gamma \gamma $}\
collisions might occur in each bunch crossing, thereby effectively giving rise
to an underlying event. This conclusion has been criticized \cite{213} on the
grounds that the minijet contribution had not been eikonalized in
ref.\cite{212}. However, in refs.\cite{154,214} is has been shown that
eikonalization does not significantly change predictions for the next
generation of linear colliders, planned to operate at $\mbox{$\sqrt{s}$} \simeq 500$ GeV.
The ``hadron crisis" has not been solved by improved calculations, but by
improved machine designs, which reduce beamstrahlung by employing a larger
number of flat bunches. It should be emphasized that hadronic backgrounds are
not the only, and perhaps not even the most severe, problem of designs with
high beamstrahlung; they also suffer from large backgrounds from soft \mbox{$e^+ e^- $}\
pairs \cite{215}, as well as from a poorly defined beam energy, which
complicates the study of thresholds \cite{216}.
So far only one Monte Carlo study of the properties of these soft hadronic
backgrounds has been published, by Chen et al. \cite{214}. They assume that
the total \mbox{$\gamma \gamma $}\ cross--section tracks the measured \mbox{$p \bar{p}$}\ cross--section at
high energies. It should be pointed out that at present even the value of the
total \mbox{$\gamma \gamma $}\ cross--section at low energies is only poorly determined
experimentally. Chen et al. then modelled minimum bias events using ISAJET
\cite{217}, which contains a parametrization of the underlying event as
determined in $pp$ and \mbox{$p \bar{p}$}\ collisions. Using the same model for \mbox{$\gamma \gamma $}\
collisions may not be a bad first approximation, given the similarities of
some features of HERA and \mbox{$p \bar{p}$}\ events discussed earlier. However, Chen et
al. did not attempt to compare their model with actual data. They also used a
rather large value for \mbox{$ p_{T,{\rm min}} $}, 3.6 GeV, this being (approximately) the
smallest partonic $p_T$ leading to observable jets at the UA1 experiment
\cite{218} at the SpS. However, TRISTAN experiments have demonstrated
\cite{141,142,148} that much softer jets can be reconstructed in \mbox{$\gamma \gamma $}\
collisions. The analysis by Chen et al., which uses the simplified
unitarization scheme (\ref{e5.14}), therefore probably under--estimates the
effect of minijets on event characteristics.\footnote{As noted above, ISAJET
simply fits the properties of the underlying event in \mbox{$p \bar{p}$}\ collisions,
including effects that can be explained in terms of minijets. However, it is
known that \mbox{$\gamma p$}\ events are more ``jetty" than \mbox{$p \bar{p}$}\ events, and hard
partonic reactions are expected \cite{205} to play an even more prominent role
in \mbox{$\gamma \gamma $}\ scattering.} Chen et al. conclude that the average \mbox{$\gamma \gamma $}\ event only
deposits about 3.5 GeV of hadronic energy in the central detector ($|\cos
\theta | \leq 0.9$) at a typical design for a 500 GeV \mbox{$e^+ e^- $}\ collider. It
would be interesting to repeat their analysis, using an event generator that
has been tuned to describe real \mbox{$\gamma \gamma $}\ data.
Finally, we briefly mention that eikonalization effects \cite{213} do
certainly play an important role if the next \mbox{$e^+ e^- $}\ collider is converted into
a \mbox{$\gamma \gamma $}\ collider, which greatly increases the average \mbox{$\gamma \gamma $}\ cms energy. The
analysis of ref.\cite{214} indicates that even here one can stay well below
one hadronic event per bunch crossing, at least at a 500 GeV collider; the
problem rapidly gets worse at higher energies, mostly due to the need to
increase the luminosity $\propto s$ in order to achieve a constant rate of
hard events. Since at those colliders \mbox{$\gamma \gamma $}\ scattering occurs at energies
well beyond the reach of current colliders, and in view of the incomplete
description even of these accessible events, any prediction of event
properties at future \mbox{$\gamma \gamma $}\ colliders should be taken with a grain of salt.
\setcounter{footnote}{0}
\section*{6) Outlook}
In this article we have reviewed the present knowledge of resolved photon
interactions. Great progress has been made in the last few years, both
experimentally and theoretically. We now know for a fact that partons ``in"
(quasi--)real photons play an important role in both \mbox{$\gamma p$}\ and \mbox{$\gamma \gamma $}\
interactions. Theory tells us that the hard $\gamma \mbox{$q \bar{q}$}$ vertex should lead
to (a component of the) photonic parton densities that have a much harder
$x-$dependence than the more familiar nucleonic parton distribution functions,
and which should grow approximately linearly with the logarithm of the
momentum scale at which the photon is being probed. Both properties have been
confirmed experimentally, at least for the quark densities. Studies of
photonic remnant jets also indicate that resolved photons do not always act
like ordinary hadrons.
At the same time many properties of minimum bias photoproduction and two
photon events do resemble those of purely hadronic collisions at similar
energies. Moreover, the total \mbox{$\gamma p$}\ cross--section seems to show
approximately the same energy dependence as total hadronic cross--sections.
We therefore have good evidence for the existence of both perturbative and
nonperturbative contributions to the hadronic structure of the photon.
Qualitatively speaking this summarizes the present level of understanding of
resolved photons. In the preceding sections we have reviewed the various
pieces of information that contributed to this understanding; we have also
described attempts to cast this understanding in a quantitative form. It seems
rather futile to try and summarize this summary. We therefore decided to
conclude this article with a list of open problems, which might become the
foci of future work.
The most pressing problems in \underbar{deep--inelastic $e \gamma$ scattering}
(Sec.~2) have to do with the description of the final state:
\begin{itemize}
\item
The measurement of Bjorken$-x$ has traditionally relied on a reconstruction of
the invariant mass $W$ of the hadronic final state from the measured quantity
$W_{\rm vis}$. An improvement of this procedure has recently been suggested
\cite{36b}. It might also be possible to determine the energy of the target
photon using the Jacquet--Blondel method, as is done by HERA experiments.
Detailed Monte Carlo work, using realistic detector resolutions and
acceptances, is needed to decide how well these ideas work in practice.
\item
There are some discrepancies between existing data on \mbox{$ F_2^\gamma $}. Improved
reconstruction methods might be of some help here. It is not clear whether
older PEP and PETRA data can be re--analyzed in this way, but new measurements
and/or new analyses of \mbox{$ F_2^\gamma $}\ at small $x_\gamma$ might help to resolve the
discrepancy between recent results by the TOPAZ and OPAL collaborations. Such
studies have the potential to discriminate between existing parametrizations
of photonic parton densities.
\end{itemize}
A list of open problems in \underbar{hard \mbox{$\gamma p$}\ scattering} (Sec.~3) includes:
\begin{itemize}
\item
No complete NLO treatment of di--jet production exists. Recall that one needs
to measure the rapidities of both high$-p_T$ partons/jets in a hard event in
order to reconstruct the Bjorken$-x$ variables.
\item
The measured jet cross--sections should be extended both in rapidity and in
$p_T$. The former increases the sensitivity to the interesting region of small
\mbox{$x_{\gamma}$}, while the latter should allow to test theory cleanly, since a detailed
understanding of the underlying event (see below) is less crucial at high
$p_T$, and differences between parametrizations of photonic parton densities
are small at large $x_\gamma$ and large momentum scale.
\item
It might be interesting to try to correlate properties of the photonic remnant
jet with those of the high$-p_T$ jets. In the usual ``nonperturbative +
anomalous" description of the hadronic photon, the nonperturbative component
should always have a remnant jet with very small $k_T$; this component is also
characterized by soft parton densities. In this picture one therefore expects
nontrivial correlations between \mbox{$x_{\gamma}$}\ and $k_T$.
\item
Studies of heavy flavour production hold great potential. We do not think it
very interesting to try to derive total cross--sections from measurements
covering only a limited region of phase space, which contains only a small
fraction of all produced heavy quarks. It might be more fruitful to attempt to
extract the resolved photon contribution, which is sensitive to the as yet
poorly constrained gluon density in the photon. At high $p_T$, ``excitation"
contributions from the charm in the photon have to be taken into account. An
important open problem is the fragmentation of rather soft (low$-p_T$) charm
quarks, which contribute most to the total charm cross--section. Theoretical
predictions are more reliable for $b$ production, but it might be difficult to
find a clean signal.
\item
The production of direct photons is by now quite well understood, although an
NLO calculation of photon + jet production would certainly be welcome.
Realistic background studies are also needed, but can presumably only be
performed by members of HERA experiments.
\item
In spite of the recent NLO calculation \cite{116} of direct \mbox{$J/\psi$}\ production
in the colour singlet model, much needs to be done here: The resolved photon
contribution is only known to leading order in the colour singlet model.
Nothing is known about the contribution from the colour octet component of the
wave function of the \mbox{$J/\psi$}\ \cite{219}, which is accessible to resolved
photons already in LO. In addition, there are contributions at high $p_T$
coming from charm and gluon fragmentation \cite{220}.
\item
It is important to test our understanding of the hadronic photon in as many
different channels as possible. The production of Drell--Yan lepton pairs,
two photon final states, and associate $\mbox{$J/\psi$} + \gamma$ final states are all
plagued by rather small cross--sections, but this should at least partly be
compensated by the cleanliness of the final states.
\end{itemize}
We should emphasize that many of the measurements proposed here need more
luminosity than HERA has delivered so far. Fortunately new data are being
collected at an ever accelerating pace.
Some open problems in \underbar{hard \mbox{$\gamma \gamma $}\ scattering} (Sec.~4) are:
\begin{itemize}
\item
An NLO calculation of di--jet production is still lacking.
\item
There is some discrepancy in the high$-p_T$ end between the partonic
cross--sections published by AMY and TOPAZ. Also, it is not clear how these
cross--sections, which have been extracted with the aid of leading order event
generators, can be compared to NLO calculations.
\item
More detailed tests of theory should be possible if the kinematic
reconstruction of events with high$-p_T$ jets can be improved. One possibility
might be to determine the Bjorken$-x$ variables using small angle electron
taggers and/or the Jacquet--Blondel method. A study of the energy distribution
of the remnant jet(s), and possible correlations with the hard jet(s), might
also prove rewarding. All these measurements are probably quite difficult,
however.
\item
Measurements of the cross--sections for the production of high$-p_T$ particles
should be relatively straightforward; they test similar aspects of the theory
as studies of jet final states do.
\item
The excess of high$-p_T$ charm events seen by TRISTAN experiments might well
be explainable in terms of charm excitation from the photon; in the
kinematical regime that is being probed by these experiments, this should be
well described by the NLO event generator that has become available recently.
Recall, however, that ALEPH does not see any excess.
\item
\mbox{$J/\psi$}\ and direct photon production has so far only been treated in LO. No
data on these final states exists as yet.
\end{itemize}
Finally, some of the most challenging problems go \underbar{beyond perturbation
theory} (Sec.~5):
\begin{itemize}
\item
We have to understand the energy flow in resolved photoproduction events. This
is crucial for the comparison of measured jet rates with theoretical
(partonic) calculations, especially for presently accessible values of $p_T$.
\item
The modelling of minimum bias photoproduction events needs to be improved.
Among other things, this is necessary for an accurate determination of total
\mbox{$\gamma p$}\ cross--sections.
\item
The total hadronic \mbox{$\gamma \gamma $}\ cross--section is at present only determined very
poorly. An accurate measurement can probably only be made if at least one of
the outgoing electrons is tagged at a small angle. This is important for the
assessment of hadronic backgrounds at future linear \mbox{$e^+ e^- $}\ colliders, which
may play a role in the decision which of several competing designs should be
pursued.
\item
The role of multiple partonic interactions needs to be clarified. They will
probably (help to) explain the energy flow in hard \mbox{$\gamma p$}\ events mentioned
above. Theoretically least understood is the treatment of contributions from
the perturbative component of photon structure functions. Studies of
multi--jet production in high energy \mbox{$\gamma \gamma $}\ collisions might yield important
clues here. This also has ramifications for a theory (as opposed to
parametrization) of total hadronic cross--sections.
\item
\mbox{$\gamma \gamma $}\ event generators in present use are known to be incomplete: They do not
include parton showers, and assume too soft a $k_T$ distribution of the
remnant jets. At higher energies multiple interactions might also affect the
properties of the underlying event substantially. Finally, some
standardization might help to decide whether putative differences in published
results signal real discrepancies in the data, or are merely a reflection of
different MC generators.
\item
It is important to test the (potentially) complete event generators for \mbox{$\gamma p$}\
and \mbox{$\gamma \gamma $}\ scattering that are now being developed in as many different ways
as possible. Not only should things like the energy flow, multiplicity
distributions, particle $p_T$ spectra, and strangeness production be
investigated, there might also be interesting correlations between some of
these quantities.
\end{itemize}
This list is almost certainly not complete. Moreover, new questions are likely
to arise as soon as current problems are solved. Much needs to be done before
we can say with confidence that we understand the structure of light.
\subsection*{Acknowledgements}
We greatly benefitted from numerous discussions with many people working in
this field. In particular, we wish to thank P. Aurenche, G. B\'elanger, F.
Borzumati, F. Boudjema, J. Butterworth, R. Enomoto, J. Forshaw, L. Gordon, M.
Greco, H. Hayashii, M. Iwasaki, D. Miller, T. Nozaki, G. Schuler, M. Seymour,
T. Sj\"ostrand and J. Storrow for their replies to our questions.
The work of M.D. was supported in part by the U.S. Department of Energy under
grant No. DE-FG02-95ER40896, by the Wisconsin Research Committee with funds
granted by the Wisconsin Alumni Research Foundation, as well as by a grant
from the Deutsche Forschungsgemeinschaft under the Heisenberg program. R.M.G.
wishes to acknowledge research grant no. 03(0745)/94/EMR-II, from the Council
of Scientific and Industrial Research.
\clearpage
|
\section{Introduction}
\label{section-introduction}
\par Measurements of $b$ production in $p{\overline p}$ collisions
provide quantitative test of perturbative QCD. Single integral $b$
cross section measurements at $\sqrt{s}$ = 1.8 TeV have been
systematically higher than predictions from next-to-leading order
(NLO) QCD calculations~\cite{cdf_bmu,D0-mu,FMNR}. These cross
section measurements, from inclusive $b \rightarrow$ lepton decays
and exclusive $B$ meson decays ($B^+\rightarrow$J/$\psi K^+$),
use the kinematical relationship between the decay product
(e.g, the lepton) and the $b$ quark spectra to obtain the production
cross section integrated over a rapidity range $\mid y \mid <$ 1 and
a $p_{\rm T}$ range from a threshold $p_{\rm T}^{min}$ to infinity. Single
differential $B$ meson cross section measurements~\cite{CDF_Bmes} are
also systematically higher than the NLO prediction.
\par Semi-differential $b - \overline b$ cross sections give further
information on the underlying QCD production mechanisms by exploring
the kinematical correlations between the two $b$ quarks. Comparison
of NLO predictions with experimental measurements can give
information on whether higher order corrections serve as a scale
factor to the NLO prediction or change the production distributions.
As future high precision $B$ decay measurements at hadron colliders
($e.g.$, CP violation studies in $B^0\rightarrow$J/$\psi
K^0_s$~\cite{snowmass_beta}) may depend upon efficient identification
of the decay products of both $b$ quarks, understanding of the
correlated cross sections is necessary.
\par This paper describes measurements of $\mu - \overline b$ correlated
cross sections as a function of the jet transverse energy
(d$\sigma$/d${\rm E}_{\rm T}$, where ${\rm E}_{\rm T} = E \times sin \theta$) and transverse
momentum (d$\sigma$/d$p_{\rm T}$) of the $\overline b$ and as a function of the
azimuthal separation (d$\sigma$/d$\delta\phi$) between the muon and
$\overline b$ jet, for $p_{\rm T}^{\mu}> 9$ GeV/c, $|\eta^{\mu}|<$ 0.6,
${\rm E}_{\rm T}^{\overline b}>$ 10 GeV/c, $|\eta^{\overline b}|<$ 1.5. The data are $15.08
\pm 0.54$ $\rm{pb^{-1}}$ of $p\overline{p}$ collisions at $\sqrt{s}
= 1.8$ TeV collected with the CDF detector between August, 1992 and
May, 1993. We make use of two features of $B$ hadrons to separate
them from the large jet backgrounds at 1.8 TeV: the high branching
fraction into muons ($\approx$ 10\%~\cite{PDG}) and the relatively
long lifetime ($\approx$ 1.5 picoseconds~\cite{PDG}). The advent of
precision silicon microstrip detectors, with hit resolutions
approaching 15 $\mu$m, provides the ability to efficiently identify
the hadronic decays of $B$ hadrons as well as the semi-leptonic
decays.
\par We use the identification of a high transverse momentum muon as
the initial signature of the presence of $b$ quarks. In
$p\overline{p}$ collisions, high transverse momentum muons come from
the production and decay of heavy quarks ($c,b,t$), vector bosons
($W,Z^\circ$), and light mesons ($\pi,K$). Additional identification
techniques are necessary to convert a $\mu -$ jet cross section into
a $\mu - \overline b$ cross section.
\par For these measurements, the first $b$ is identified from a
semi-leptonic decay muon and the other $b$ (referred to for
simplicity as the $\overline b$, though we do not perform explicit flavor
identification for either $b$) is identified by using precision track
reconstruction in jets to measure the displaced particles from
$\overline b$ decay. Jets are identified as clusters of energy in the
calorimeter~\cite{jet_papers}. In this paper, a jet energy (or jet
transverse energy) refers to the measured energy in the cluster. A
procedure to simultaneously unfold the effects of detector response
and resolution is used to translate the results from $\overline b$ jets to
$\overline b$ quarks.
\par It should be noted that we have chosen to report the
measurements as differential $\mu - \overline b$ cross sections rather than
$b - \overline b$ cross sections in order to facilitate comparison to
calculations of the production cross sections. The process of
converting a muon cross section to a quark cross section includes
systematic uncertainties~\cite{cdf_bmu} with strong dependence on
both production, fragmentation, and decay models. By presenting $\mu
- \overline b$ cross sections, we facilitate the future comparison of the
experimental results to different models, since the data results and
uncertainties are not tied to specific models.
\par Section~\ref{section-detector} describes the detector systems
used for muon and $\overline b$ jet identification.
Section~\ref{section-dataset} contains descriptions of the muon and
jet identification requirements. The $\overline b$ jet counting is
discussed in section~\ref{section-correlated}. In
section~\ref{section-acceptance}, the muon and $\overline b$ jet
identification efficiencies and acceptances are described. The cross
section results, the calculation of additional physics backgrounds,
and jet to quark unfolding are discussed in
section~\ref{section-results}. Section~\ref{section-conclusions}
closes with a discussion of the experimental results.
\section{Detector Description}
\label{section-detector}
\par The CDF has been described in detail elsewhere~\cite{NIM_book}.
The analysis presented in this paper depends on the tracking and muon
systems for triggering and selection, while identification of
hadronic jets uses the information from the calorimeter elements.
\subsection{Tracking and Muon Systems}
\par This analysis uses the silicon vertex detector
(SVX)~\cite{SVXNIM}, the vertex drift chamber (VTX) and the central
tracking chamber (CTC)~\cite{CTCNIM} for charged particle tracking.
These are all located in a 1.4 T solenoidal magnetic field. The SVX
consists of 4 layers of silicon-strip detectors with $r-\phi$
readout, including pulse height information, with a total active
length of 51 cm in the range $-27.3 < z < 27.3$
cm~\cite{coordinates}. The pitch between readout strips is 60 $\mu$m
on the inner 3 layers and 55 $\mu$m on the outermost layer. A single
point spatial resolution of 13 $\mu$m has been obtained. The first
measurement plane is located 2.9 cm from the interaction point,
leading to an impact parameter resolution of $\approx$ 15 $\mu$m for
tracks with transverse momentum, $p_{\rm T}$, greater than 5 GeV/c. The
VTX is a time projection chamber providing information out to a
radius of 22 cm and $\mid \eta \mid <$ 3.5. The VTX is used to
measure the $p{\overline p}$ interaction vertex ($z_0$) along the
$z$ axis with a resolution of 1 mm. The CTC is a cylindrical drift
chamber containing 84 layers, which are grouped into alternating
axial and stereo superlayers containing 12 and 6 wires respectively,
covering the radial range from 28 cm to 132 cm. The momentum
resolution of the CTC is $\delta$$p_{\rm T}$/$p_{\rm T}$ = 0.002~$\times~p_{\rm T}$
for isolated tracks (where $p_{\rm T}$ is in GeV/c). For tracks found in
both the SVX and CTC, the momentum resolution improves to
0.0009~$\times~p_{\rm T} \oplus 0.0066$ (where $p_{\rm T}$ is in GeV/c).
\par The muon system consists of two detector elements. The Central
Muon system (CMU)~\cite{CMUNIM}, which consists of four layers of
limited streamer chambers located at a radius of 384 cm, behind
$\approx$ 5 absorption lengths of material, provides muon
identification for the pseudorapidity range $|\eta| < $0.6. This
$\eta$ region is further instrumented by the Central Muon upgrade
system (CMP)~\cite{CMP}, which is a set of four chambers located
after $\approx$ 8 absorption lengths of material. Approximately 84\%
of the solid angle of $|\eta|\leq$0.6 is covered by the CMU, 63\% by
the CMP, and 53\% by both. Muon transverse momentum is measured with
the charged tracking systems and has the tracking resolutions
described above. CMU (and CMP) segments are defined as a set of 2 or
more hits along radially aligned wires.
\subsection{Calorimeter Systems}
\par This analysis uses the CDF central and plug calorimeters, which
are segmented into separate electromagnetic and hadronic
compartments. In all cases, the absorber in the electromagnetic
compartment is lead, and in the hadronic compartment, iron. The
central region subtends the range $|\eta|< $ 1.1 and spans 2$\pi$ in
azimuthal coverage, with scintillator as the active medium. The plug
region subtends the range $1.1 < |\eta| < 2.4$ with gas proportional
chambers as the active media, again with 2$\pi$ azimuthal coverage.
The calorimeters have resolutions that range from 13.7\%/$\sqrt{{\rm
E}_{\rm T}} \oplus 2\%$ for the central electromagnetic to
106\%/$\sqrt{{\rm E}_{\rm T}} \oplus 6\%$ for the plug
hadronic~\cite{TOPPRD}.
\subsection{Trigger System}
\par CDF uses a three-level trigger system~\cite{TRIGGERNIM}. Each
level is a logical OR of a number of triggers designed to select
events with electrons, muons or jets. The analysis presented in this
paper uses only the muon trigger path. Section~\ref{section-dataset}
includes a description of the trigger efficiencies for muons.
\par The Level 1 central muon trigger requires a pair of hits on
radially aligned wires in the CMU system. The $p_{\rm T}$ of the track
segment is measured using the arrival times of the drift electrons at
the wires to determine the deflection angle due to the magnetic
field. The trigger requires that the segment have $p_{\rm T}>$ 6 GeV/c,
with at least two confirming hits in the projecting CMP chambers.
\par The Level 2 trigger includes information from a list of $r-\phi$
tracks found by the central fast tracker (CFT)~\cite{CFTNIM}, a
hardware track processor which uses fast timing information from the
CTC as input. The CFT momentum resolution is
$\deltap_{\rm T}$/$p_{\rm T}~\approx 0.035 \times p_{\rm T}$, with a plateau efficiency
of 91.3$\pm$0.3\% for tracks with $p_{\rm T}$ above 12 GeV/c. The CMU
chamber segment is required to match a CFT track with $p_{\rm T} >$ 9.2
GeV/c within 5$^\circ$ in the $\phi$ coordinate.
\par The Level 3 trigger makes use of a slightly modified version of
the offline software reconstruction algorithms, including full 3
dimensional track reconstruction. The CMU segment is required to
match a CTC track with $p_{\rm T}>$ 7.5 GeV/c, extrapolated to the chamber
radius, within 10 cm in $r-\phi$. Confirming CMP hits are required.
\section{Dataset Selection} \label{section-dataset}
\par Beginning with the sample of muon triggered events, we select
events with both a well identified muon candidate and a minimum
transverse energy jet. A primary vertex is found by a weighted fit
of the VTX $z_0$ vertex position and SVX tracks. An iterative search
removes tracks with large impact parameters (the distance of closest
approach in the $r - \phi$ plane) from the fit. Since the $\overline b$ jet
identification technique (described in
section~\ref{section-correlated}) depends upon the precision track
reconstruction in the SVX, we require the event primary vertex $\mid
z_0 \mid < 30$ cm. In this section, we discuss the identification
variables, efficiency, and geometric acceptance for muon and $\overline b$
jet candidates. Table~\ref{table-efficiencies} contains a summary of
the muon efficiency and acceptance results and
table~\ref{table-bbar_jet_acceptance} contains a summary of the
$\overline b$ jet identification and acceptance results.
\subsection{Muon Identification}
\par Muons are identified as a well matched coincidence between a
track in the CTC and segments in both the CMU and CMP muon systems.
The CTC track is required to have $p_{\rm T} > 9$ GeV/c and point back to
within 5 cm in $z$ of the found primary vertex. The measured track
is extrapolated to the muon chambers and is required to match the
muon chamber track segment position to $<~3\sigma$ in the transverse
direction (for both CMU and CMP) and $<\sqrt{12}\sigma$ in the
longitudinal direction (for CMU). In all cases, $\sigma$ includes
the contributions from smearing due to multiple scattering in the
absorber and the muon chamber resolution. We require that the track
be found in the SVX.
\par There are 144097 events passing all the muon requirements in
this data sample. In the case where there is more than 1 identified
muon in an event, we take the highest $p_{\rm T}$ muon as the $b$ candidate
muon. The fraction of muons from $b$ decay is measured to be
approximately 40\%~\cite{TOPPRD}, with a fraction from charm decays
of approximately 20\%. Figure~\ref{fig-muon_pt_spectrum} shows the
transverse momentum spectrum for the muons in this dataset. The
flattening of the slope at high $p_{\rm T}$ is due to muons from
electroweak boson decay.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_pt_mu_spectrum.ps}
\caption{The $\mu$ $p_{\rm T}$ spectrum for the 9 GeV/c sample. There are
144097 events, with 80 having $p_{\rm T} >$ 60 GeV/c. The enhancement
above 25 GeV/c is due to the presence of muons from W and Z boson
decays.}
\label{fig-muon_pt_spectrum}
\end{figure}
\subsection{Jet Identification}
\par Jets are identified in the CDF calorimeter systems using a fixed
cone (in $\eta - \phi$ space) algorithm. A detailed description of
the algorithm can be found in reference~\cite{jet_papers}. For this
analysis, we use a cone radius of 0.4. We require that jets have
transverse energy, ${\rm E}_{\rm T}$ = E $\times$ sin~$\theta$ (where $E$ is the
total energy in the cone), greater than 10 GeV, and $\mid \eta \mid \leq$ 1.5.
There are 50154 events passing the muon and jet ${\rm E}_{\rm T}$ requirements.
We use tracking techniques to identify $\overline b$ jets, so the
pseudorapidity range is restricted to the region with tracking
coverage. All jet energies in this paper are measured energies, not
including corrections for known detector effects({\em e.g.},
calorimeter non-linearities). An unsmearing procedure, described in
section~\ref{section-results}, is used to convert measured jet ${\rm E}_{\rm T}$
distributions to parton momentum distributions.
\par We associate SVX tracks to a jet by requiring that the track be
within the cone of 0.4 around the jet axis. To remove tracks
consistent with photon conversions and $K_S$ or $\Lambda$ decays
originating from the primary vertex, we require that the impact
parameter, $d$, be less than 0.15 cm. In addition, track pairs
consistent with $K_S \rightarrow \pi^+\pi^-$ or $\Lambda \rightarrow
p\pi$ decays are removed. We select jets with two or more well
measured tracks~\cite{TOPPRD}, $p_{\rm T} >$ 1 GeV/c, with positive impact
parameters. The impact parameter sign is defined to be $+1$ for
tracks where the point of closest approach to the primary vertex lies
in the same hemisphere as the jet direction, and $-1$ otherwise.
for
\par We require that the distance, $\Delta$R, in $\eta - \phi$ space
between the muon and the jet axis be greater than 1.0. There are
16842 events passing all the muon and jet requirements. The
$\Delta$R separation is chosen so that tracks clustered around the
jet axis are separated from the muon direction, in order to have
physical separation of the $b$ and $\overline b$ decay products. As there
may be more than one jet in an event passing these requirements, we
select the jet with the lowest jet probability (defined in
section~\ref{section-correlated}), so as to have a unique combination
of $\mu$ --- jet in each event.
\section{$\overline b$ Jet Counting}
\label{section-correlated}
\par The $\overline b$ jet is not identified on an event-by-event basis,
but instead by fitting for the number of $\overline b$ jets present in the
sample. For each jet, we combine the impact parameter information
for tracks in the jet cone into one number which describes the
probability that the given collection of tracks has no decay products
from long lived particles. In a $\overline b$ jet, there will be a
significant number of tracks from the $B$ hadron decay, and hence the
probability for a $\overline b$ jet will be much less than 1.
\subsection{The Jet Probability Algorithm}
\par The $\overline b$ jet identification makes use of a probability
algorithm~\cite{ALEPH} which compares track impact parameters to
measured resolution functions in order to calculate for each jet a
probability that there are no long lived particles in the jet cone.
This probability is uniformly distributed for light quark or gluon
jets (we refer to these jets as prompt jets), but is very low for
jets with displaced vertices from heavy flavor decay. We now
briefly describe the transformation from the track impact parameters
to the jet probability measure.
\par The track impact parameter significance is defined as the value
of the impact parameter divided by the uncertainty in that quantity,
which includes both the measured uncertainties from the track and
primary vertex reconstruction.
Figure~\ref{figure-resolution_function} shows the distribution of
impact parameter significance ($s_0 = d/\sigma$) from a sample of
jets taken with a 50 GeV jet trigger~\cite{TOPPRD}, overlayed with a
fitted function. The tails of the distribution come from a
combination of non-Gaussian effects and true long lived particles.
Using a combination of data and Monte Carlo simulation of heavy
flavor decays, we estimate approximately 30\% of the tracks with
$\mid s_0 \mid >$ 3.0 are from the decay products of long lived
particles, which is consistent with the excess in the positive $s$
side of the distribution. The negative side of the fitted function,
$R(s)$, is used to map the impact parameter significance $s_0$ to a
track probability measure:
\begin{equation}
P(s_0) = \frac{\int_{-\infty}^{-|s_0|}R(s)ds}
{\int_{-\infty}^{0}R(s)ds}.
\end{equation}
\noindent The track probability is a measure of the probability of
getting a track with impact parameter significance greater than $s$.
The function $R(s)$ can be defined for both Monte Carlo simulated
datasets and the jet dataset. The mapping of the resolution function
to the track probability distribution removes differences in the
resolution between the simulated detector performance and the true
detector performance and creates a variable which is consistent
between the two datasets.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_res_function.ps}
\caption{A sample track resolution function, including fits to both
positive and negative signed impact parameters. The function is fit
to 2 gaussians plus two exponentials, one for the positive side and
one for the negative side. The excess on the positive side is
attributable to long lived particles in the sample.}
\label{figure-resolution_function}
\end{figure}
\par The jet probability is then calculated from the independent
track probabilities as: \begin{equation} P_{jet} =
\Pi\sum_{k=0}^{N-1}\frac{(-\ln\Pi)^k}{k!}, \end{equation} where
\begin{equation} \Pi = P_1 P_2 \cdots P_N \end{equation} \noindent is
the product of the individual probabilities of the selected tracks.
For the rest of this paper, when the track selection requirements
pick tracks with negative signed impact parameters, we will refer to
the measure as the ``negative jet probability''. When the track
selection requirements pick tracks with positive signed impact
parameters, we will refer to the measure as $P_{jet}$.
\par Figure~\ref{figure-jet_prob_samples} shows the distribution of
the negative jet probability in the 50 GeV sample. Since this
distribution reflects the smearing of the impact parameter
significance distribution due to resolution effects, we expect that
the $P_{jet}$ distribution for prompt (light quark and gluon) jets to
be similar. Simulated jets containing heavy flavor decays show
distinct differences from this distribution, peaking at low values of
$P_{jet}$. In figure~\ref{fig-smoothed_shapes}, we show the
distributions of log$_{10}(P_{jet})$ for $\overline b$, charm, and prompt
jets.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_jet_prob_neg.ps}
\caption{The negative jet probability spectrum, calculated using
tracks with negative signed impact parameters, in a sample of 50 GeV
jets.}
\label{figure-jet_prob_samples}
\end{figure}
\par We have found that the $P_{jet}$ shape for heavy flavor jets is
affected by the number of tracks used in the calculation of $P_{jet}$
which are also used in the primary vertex fit. The turnover visible
in the $\overline b$ and charm distributions around -3 in log$_{10}(P_{jet})$
is a combination of the vertex requirements ($d/\sigma <$ 3 for
tracks in the fit) and the $\overline b$ and charm lifetimes. $\overline b$ and
charm jets are affected differently, due to differences in lifetime
and decay multiplicities.
\subsection{$\overline b$ Jet Fit Technique}
\par We use a binned maximum likelihood fit to distinguish the
$\overline b$, $c$, and prompt jet contributions in the sample. For a
binned likelihood fit, we find that log$_{10}(P_{jet})$ shows stronger
differentiation between $\overline b$, $c$, and prompt jets (see
figure~\ref{fig-smoothed_shapes}) than $P_{jet}$ and use this variable
in the fitting algorithm. We fit over the range -10 --- 0 in
log$_{10}(P_{jet})$, where the $\overline b$, $c$, and prompt contributions
are constrained to be positive. No other constraints are included in
the fit. We model the prompt jets with an exponential distribution,
since a logarithm transforms a uniform distribution to an
exponential distribution.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_input_shapes.ps}
\caption{The log$_{10}(P_{jet})$ distributions used as inputs to the
fitting program. The $b$ and $c$ shapes are smoothed versions of
Monte Carlo distributions, while the primary shape is an exponential
function. The three distributions are normalized to equal area and
shown on the same vertical scale.}
\label{fig-smoothed_shapes}
\end{figure}
\par We have explored the effect of different Monte Carlo samples to
construct the input shape used in the fit. Using different input
$\overline b$ jet Monte Carlo samples (see
section~\ref{section-acceptance}) compared to the test distribution
shows a 5\% change in the fit fractions. Changing the average
$\overline b$ lifetime by 6\%~\cite{b_lifetime,lifetime_choice} changed the
fit fraction by 3\%. We include a 5.8\% systematic uncertainty to
our fit results to account for systematic uncertainties in the
fitting procedure and uncertainty on the $\overline b$ lifetime.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_fit_results.ps}
\caption{For all jets (${\rm E}_{\rm T} >$ 10 GeV) in the $\mu$ sample, we show
the data distribution overlayed with the fit results. There are two
events in the data with log$_{10}(P_{jet}) <$ -10. Statistical errors
on the data and the fit results are included. The fit results model
the data well over the entire range of the fit.}
\label{fig-sample_fit}
\end{figure}
\par In figure~\ref{fig-sample_fit}, we show the distribution of
log$_{10}(P_{jet})$ for all jets, ${\rm E}_{\rm T} >$ 10 GeV, in the muon sample,
overlayed with the fit results. In this sample, the fit finds 2484
$\pm$ 94 $\overline b$ jets, 1988 $\pm$ 175 $c$ jets, and 12368 $\pm$ 157
prompt jets for a total of 16840. There are 16842 events in the data
sample. Figure~\ref{fig-fit_comparisons} shows three comparisons of
the data and fit results, showing the bin-by-bin difference in the
results, the bin-by-bin difference divided by the errors, and the
distribution of the difference divided by the errors. In these
distributions, the errors are the statistical errors in the data
points. We do not include any error on the Monte Carlo shapes. From
these distributions, we can see that the inputs model the data well.
The difference divided by the errors has a mean of 0.04 and RMS of
0.95.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_shape_comparison.ps}
\caption{Various comparisons of the data distribution and the fit
results. We show (a) the bin by bin difference between the data and
the fit results, (b) the bin-by-bin difference scaled to the errors,
and (c) the distribution of the difference scaled to the errors, with
mean 0.04 and RMS = 0.95. In all cases, the errors are the
statistical error in the data points and the fitted results.}
\label{fig-fit_comparisons}
\end{figure}
\par For the semi-differential measurements, we do an independent fit
of the log$_{10}(P_{jet})$ distribution and then correct for the
acceptance in each ${\rm E}_{\rm T}$ or $\delta\phi$ bin.
Table~\ref{table-fit_results} contains a summary of the number of
total jets and the number of $\overline b$ jets in each ${\rm E}_{\rm T}$ and
$\delta\phi$ bin considered.
\begin{table}[tb]
\begin{center}
\begin{tabular}{|c|c|c|} \hline
${\rm E}_{\rm T}$ Range & Number of Jets & Estimated Number \\
& & of $\overline b$ Jets \\ \hline\hline
10 --- 15 & 5174 & 547 $\pm$ 49 \\
15 --- 20 & 3818 & 618 $\pm$ 47 \\
20 --- 25 & 2563 & 453 $\pm$ 39 \\
25 --- 30 & 1698 & 278 $\pm$ 30 \\
30 --- 40 & 1921 & 327 $\pm$ 33 \\
40 --- 50 & 819 & 140 $\pm$ 20 \\
50 --- 100 & 849 & 107 $\pm$ 19 \\ \hline \hline
$\delta\phi$ Range & & \\ \hline\hline
0---$\frac{\pi}{8}$ & 43 & 4.8 $+5.5 - 4.8$ \\
$\frac{\pi}{8}$ --- $\frac{\pi}{4}$ & 83 & 25.0 $\pm$ 8.6 \\
$\frac{\pi}{4}$ --- $\frac{3\pi}{8}$ & 230 & 54.7 $\pm$ 13.3 \\
$\frac{3\pi}{8}$ --- $\frac{\pi}{2}$ & 336 & 78.2 $\pm$ 15.9 \\
$\frac{\pi}{2}$ --- $\frac{5\pi}{8}$ & 519 & 105. $\pm$ 18.5 \\
$\frac{5\pi}{8}$ --- $\frac{3\pi}{4}$ & 1008 & 160. $\pm$ 25. \\
$\frac{3\pi}{4}$ --- $\frac{7\pi}{8}$ & 3229 & 461. $\pm$ 42. \\
$\frac{7\pi}{8}$ --- $\pi$ & 11394 & 1593. $\pm$ 75. \\
\hline
\end{tabular}
\end{center}
\caption{$\overline b$ fit results as a function of jet ${\rm E}_{\rm T}$ and
$\delta\phi$ between the muon and $\overline b$ jet. We have not included
a common systematic uncertainty of 5.8\%.}
\label{table-fit_results}
\end{table}
\section{Acceptance and Efficiency}
\label{section-acceptance}
\subsection{Muon Requirements}
\par The muon geometric acceptance is the fraction of events with a
muon in the good fiducial region of the CMU and CMP chambers,
starting from a sample where the muon has $p_{\rm T} >$ 9 GeV/c and $\mid
\eta \mid <$ 0.6. Note that this term is only a geometric acceptance
and does not include kinematical cuts on the muon.
\par The geometric acceptance is studied with a $b\rightarrow\mu$
Monte Carlo generator (which includes the sequential decays
$b\rightarrow c \mu$), with the input spectra coming from the next to
leading order calculation of $b - \overline b$ production by Mangano,
Nason, and Ridolfi (MNR)~\cite{MNR}. The input spectra use the MRSD0
structure functions~\cite{MRSD0} and renormalization scale $\mu_0 =
\sqrt{m_b^2 + (p_{\rm T}^{b 2} + p_{\rm T}^{\overline b 2}2)/2}$, with $m_b$ = 4.75
GeV/$c^2$. This generator produces $b$ quarks and $B$ hadrons, using
the Peterson fragmentation form~\cite{Peterson} with
$\epsilon=0.006~\pm~0.002$~\cite{fragmentation}. $B$ hadrons are
decayed according to the CLEO Monte Carlo program, QQ~\cite{CLEOMC}.
We select events with a $b \rightarrow \mu$ decay, with muon $p_{\rm T} >$
9 GeV/c and $\mid \eta \mid <$ 0.6.
\par For these studies, event vertices are distributed along the $z$
axis as a Gaussian with mean = -1.4 cm and $\sigma$ = 26.65
cm~\cite{Badgett_thesis}, which is a good approximation to the
average conditions seen in the data. The muons are propagated to the
CMU and CMP chamber radii, including the effects of the central
magnetic field and multiple scattering. The acceptance is then
defined as the fraction of muons which are in the good fiducial area
of both the CMU and CMP chambers and is found to be 53.0 $\pm$ 0.3\%
(statistical), independent of variations of the $\epsilon$ parameter
from 0.004 to 0.008.
\par The muon trigger and selection depends significantly upon the
track reconstruction efficiency in the CTC. We have defined our
efficiencies to be multiplicative, so that we can measure them
independently. In this section, the efficiencies of the individual
selection requirements, and methods of measuring them, are
described.
\par The trigger efficiency is measured using independently
triggered samples for each level of the system, where the efficiency
is expressed as a function of the muon $p_{\rm T}$.
Figure~\ref{fig-trigger_efficiency} shows the efficiency curves for
the 3 levels of the trigger system. The efficiency curves are then
convoluted with the $p_{\rm T}$ spectrum of the muons, to extract the
efficiency for a muon with $p_{\rm T} >$ 9 GeV/$c$. This convolution is
done independently for the differential ${\rm E}_{\rm T}$ cross section bins (see
table~\ref{table-trigger_efficiency}), since the muon $p_{\rm T}$ spectrum
may depend upon the transverse momentum distribution of the $\overline b$
jet recoiling against the $b \rightarrow \mu$ decay. For $\overline b$
jets with ${\rm E}_{\rm T} >$ 10 GeV, the combined L1, L2, and L3 trigger
efficiency is measured to be 83.0 $\pm$ 1.7 \%.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_trigger_curves.ps}
\caption{The trigger efficiency curves for the 3 levels of the
trigger system. The trigger efficiency is the product of the three
curves, convoluted with the $\mu~p_{\rm T}$ spectrum.}
\label{fig-trigger_efficiency}
\end{figure}
\par The vertex requirement, $\mid z_0 \mid < 30$ cm, is studied in a
minimum bias trigger dataset, comparing the vertex distribution to
the predicted shape, including the measured longitudinal
distribution of the proton and anti-proton bunches and the effects of
the accelerator $\beta$ function~\cite{Badgett_thesis}. The
efficiency is found to be 74.2 $\pm$ 2.1 \%, where the uncertainty
comes from uncertainty in the measured beam longitudinal
distributions and $\beta$ function.
\par The track finding efficiency in the CTC is a function of the
density of charged particles. By embedding Monte Carlo simulated
track hits into data samples, we quantify the probability of finding
the Monte Carlo simulated track as a function of the relative
density of CTC hits. The quantified probability is convoluted with
the hit density distribution for the muon sample. The track finding
efficiency is measured to be 96 $\pm$ 1.7 \%, where the uncertainty
represents the change in the result using different parametrizations
of the probability curve vs hit density.
\par The combined $\chi^2$ matching efficiency is measured in a
$J/\psi \rightarrow \mu^+ \mu^-$ sample identified by tracking and
mass requirements and is found to be $98.7 \pm 0.2$ \%. The muon
segment reconstruction efficiency is found to $98.1 \pm 0.3$ \%,
resulting in a combined efficiency of $96.8 \pm 0.4$ \%.
\par The track finding efficiency in the SVX is studied in the 9
GeV/$c$ muon sample, requiring the CTC track to extrapolate to a
good SVX fiducial region. The efficiency is found to be 90 $\pm$
1\%, where the uncertainty is the statistical error only.
\begin{table}[tb]
\begin{center}
\begin{tabular}{|c|c|} \hline
${\rm E}_{\rm T}$ bin & Trigger Efficiency \\ \hline \hline
10 --- 15 GeV & 82.6 $\pm$ 1.7 \% \\
15 --- 20 GeV & 83.0 $\pm$ 1.7 \% \\
20 --- 25 GeV & 83.4 $\pm$ 1.7 \% \\
25 --- 30 GeV & 83.6 $\pm$ 1.7 \% \\
30 --- 40 GeV & 83.8 $\pm$ 1.7 \% \\
40 --- 50 GeV & 83.9 $\pm$ 1.7 \% \\
50 --- 100 GeV & 83.7 $\pm$ 1.7 \% \\ \hline
All ${\rm E}_{\rm T}$ & 83.0 $\pm$ 1.7 \% \\
\hline
\end{tabular}
\end{center}
\caption{The muon trigger efficiency for each jet ${\rm E}_{\rm T}$ bin. A
common 2\% uncertainty is assigned to each bin.}
\label{table-trigger_efficiency}
\end{table}
\begin{table}[tb]
\begin{center}
\begin{tabular}{|c|c|} \hline
Geometric Acceptance & 53.0 $\pm$ 0.3 \% \\ \hline
CTC Track Finding & 96.0 $\pm$ 1.7 \% \\
Matching Efficiency & 96.8 $\pm$ 0.4 \% \\
Z Vertex Requirements & 74.2 $\pm$ 2.1 \% \\
SVX Track Finding & 90 $\pm$ 1 \% \\ \hline
Combined Acceptance & \\
and Efficiency & 32.9 $\pm$ 1.1 \% \\ \hline
\end{tabular}
\end{center}
\caption{Summary of muon acceptance and efficiency numbers. The
trigger efficiency is applied on a bin by bin basis for the jet ${\rm E}_{\rm T}$
measurement.}
\label{table-efficiencies}
\end{table}
\subsection{$\overline b$ Jet Requirements}
\par The $\overline b$ jet acceptance combines the fiducial acceptance of
the SVX and the CTC, the track reconstruction efficiency, and
fragmentation effects and the $\Delta$R separation requirement.
These tracking and $\Delta$R effects are studied separately, with a
full simulation used for the combination of the track requirements
and fiducial acceptance, while a MNR based $\mu-\overline b$ model is
used for the $\Delta$R acceptance. The ${\overline b}$ jet
acceptance is calculated separately as a function of the jet ${\rm E}_{\rm T}$
and azimuthal opening angle between the muon and the jet.
\par Monte Carlo samples for $b$ and $c$ quarks are produced using
ISAJET version 6.43~\cite{ISAJET}. The CLEO Monte Carlo
program~\cite{CLEOMC} is used to model the decay of $B$ hadrons. $b$
quarks produced using the HERWIG Monte Carlo~\cite{herwig} and PYTHIA
Monte Carlo~\cite{pythia} programs are also used for systematic
studies. The ISAJET and PYTHIA samples used the Peterson form as the
fragmentation model, with $\epsilon =$ 0.006 $\pm$ 0.002. While
none of these generators use a NLO calculation of $b$ production, the
$\eta$ distribution of the quarks agrees well with the NLO
calculation. For tracking efficiency studies, events with a muon with
$p_{\rm T} >$ 8 GeV are passed through the full CDF simulation and
reconstruction package. The simulation used an average $b$ lifetime
of $c\tau$ = 420 $\mu$m~\cite{lifetime_choice}.
\par The track acceptance represents the fraction of $\overline b$ quarks,
${\rm E}_{\rm T} >$ 10 GeV, $|\eta| <$ 1.5 which produce jets with at least 2
good tracks inside a cone of 0.4 around the jet axis, where there is
also a $b$ quark which decays to a muon with $p_{\rm T} >$ 9 GeV within the
CMU-CMP acceptance. The average track acceptance for the
${\overline b}$ is 51.4 $\pm$ 0.8\%. It ranges from 45.7 $\pm$ 1.1\%
(statistical error only) for 10 $<$ ${\rm E}_{\rm T}$ $<$ 15 GeV to 64.8 $\pm$
2.6\% for 50 $<$ ${\rm E}_{\rm T}$ $<$ 100 GeV.
\begin{table*}[t]
\begin{center}
\begin{tabular}{|c|c|c|} \hline
${\rm E}_{\rm T}$ Range & Track Acceptance & $\Delta$R Acceptance \\ \hline\hline
10 --- 15 & 45.7 $\pm$ 1.1 $\pm$ 2.3 \%
& 86.9 $\pm$ 1.0 $^{+1.4}_{-1.6}$ \% \\
15 --- 20 & 55.9 $\pm$ 1.7 $\pm$ 2.8 \%
& 88.2 $\pm$ 1.5 $^{+1.7}_{-1.9}$ \% \\
20 --- 25 & 58.1 $\pm$ 2.5 $\pm$ 2.9 \%
& 88.3 $\pm$ 2.0 $^{+2.2}_{-2.0}$ \% \\
25 --- 30 & 61.3 $\pm$ 3.5 $\pm$ 3.1 \%
& 88.3 $\pm$ 2.3 $^{+3.0}_{-3.5}$ \% \\
30 --- 40 & 61.7 $\pm$ 3.8 $\pm$ 3.1 \%
& 87.9 $\pm$ 3.4 $^{+3.6}_{-5.4}$ \% \\
40 --- 50 & 64.8 $\pm$ 2.6 $\pm$ 3.2 \%
& 87.1 $\pm$ 3.5 $^{+4.2}_{-5.1}$ \% \\
50 --- 100 & 65.0 $\pm$ 2.6 $\pm$ 3.3 \%
& 85.5 $\pm$ 3.7 $^{+5.2}_{-1.9}$ \% \\
\hline \hline
$\delta\phi$ Range (radians) & & \\ \hline\hline
0---$\frac{\pi}{8}$ & 46.3 $\pm$ 1.4 $\pm$ 2.6 \%
& 6.9 $\pm$ 0.03 $^{+0.3}_{-0.2}$ \%\\
$\frac{\pi}{8}$ --- $\frac{\pi}{4}$
& 47.3 $\pm$ 1.4 $\pm$ 2.6 \% & 20.8 $\pm$ 0.2 $^{+2.1}_{-0.3}$ \%\\
$\frac{\pi}{4}$ --- $\frac{3\pi}{8}$
& 51.4 $\pm$ 0.8 $\pm$ 2.6 \% & 74.7 $\pm$ 0.9 $^{+6.0}_{-0.0}$ \%\\
$\frac{3\pi}{8}$ --- $\pi$
& 51.4 $\pm$ 0.8 $\pm$ 2.6 \% & 100 \% \\ \hline
\end{tabular}
\end{center}
\caption{$\mu - \overline b$ track and $\Delta$R acceptance as a function
of jet ${\rm E}_{\rm T}$ and $\delta\phi$ (statistical and systematic
uncertainties). There is a common (relative) systematic uncertainty
of 5\% in the tracking efficiency. For $\delta\phi >$ 1 radian, the
$\Delta$R acceptance is 100\% by definition.}
\label{table-bbar_jet_acceptance}
\end{table*}
\par We have compared the values for the ${\overline b}$ track
acceptance from ISAJET samples to the acceptance from HERWIG samples.
The acceptance agrees within the statistical error in the samples as
a function of ${\rm E}_{\rm T}$, differing at the 5\% level. We include this
variation as an additional systematic uncertainty on the track
acceptance. Comparisons of inclusive jet track acceptances from an
ISAJET sample and from data show reasonable agreement.
\par For the calculation of the $\Delta$R acceptance, we have used a
model based on the MNR calculation~\cite{MNR}. This calculation can
be used to give exact ${\cal O}(\alpha_s^3)$ results in situations
where kinematical cuts have been applied at the parton level. We
have made additions to the calculation to model the $\mu - \overline b$
differential cross sections.
\par The MNR calculation~\cite{MNR} produces the vectors $p^b$,
$p^{\overline b}$, and $p^{gluon}$ with appropriate weights. We include
additional weighting for the following:
\begin{itemize}
\item Probability of $p_{\rm T}^{\mu} >$ 9 GeV/c for given $p_{\rm T}^b$, ${\cal P}(p_{\rm T}^{\mu},\pt^b)$
\item Probability of ${\rm E}_{\rm T}^{\overline b}$ jet in a given ${\rm E}_{\rm T}$ bin for given
$p_{\rm T}^{\overline b}$, ${\cal P}({\rm E}_{\rm T}^{\bbar},\pt^{\bbar})$
\end{itemize}
\par ${\cal P}(p_{\rm T}^{\mu},\pt^b)$ is defined as the fraction of $b$ quarks, with given
$p_{\rm T}^b$, which decay into muons with $p_{\rm T}^{\mu} >$ 9 GeV/c. We use
the $b\rightarrow\mu$ Monte Carlo generator described above to derive
this function, using B$(b\rightarrow\mu)$ = 0.103 $\pm$
0.005~\cite{PDG}. Since the probability is defined as a function of
$p_{\rm T}^b$, the exact shape of the $p_{\rm T}^b$ distribution does not enter
into the result. Figure~\ref{fig-pt_mu_probability} shows the value
${\cal P}(p_{\rm T}^{\mu},\pt^b)$ as a function of $p_{\rm T}^b$. The three curves are for
different values of the Peterson $\epsilon$ parameter used in the
fragmentation model. In addition to this probability weighting, we
also smear the $b$ quark direction in pseudo-rapidity and azimuth.
The smearing is based on the results from the $b\rightarrow\mu$ Monte
Carlo generator.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_pt_mu_probability.ps}
\caption{The probability of a $b\rightarrow\mu$ decay, with
$p_{\rm T}^{\mu} >$ 9 GeV/c, as a function of $p_{\rm T}^b$. We have included
the branching fraction $B(b\rightarrow\mu) = 0.103$. The curves
represent three choices of the Peterson $\epsilon$ parameter used in
the fragmentation process.}
\label{fig-pt_mu_probability}
\end{figure}
\par ${\cal P}({\rm E}_{\rm T}^{\bbar},\pt^{\bbar})$ is defined as the probability that a $\overline b$ quark,
with given $p_{\rm T}^{\overline b}$, would produce a jet with given
${\rm E}_{\rm T}^{\overline b}$. Using the methods outlined in
section~\ref{section-results}, we have a binned probability
distribution in ${\rm E}_{\rm T}$ for each $p_{\rm T}^{\overline b}$. Since the measured jet
${\rm E}_{\rm T}$ integrates over a range in pseudo-rapidity and azimuth (a cone
of radius 0.4), we approximate this clustering effect by clustering
partons (adding the $\overline b$ and the gluon momenta vectorially)
within the same cone size. For the rest of this paper, when we
discuss the ${\rm E}_{\rm T}^{\overline b}$ or $p_{\rm T}^{\overline b}$ theory distributions, it
means the clustered partons ${\rm E}_{\rm T}$ or $p_{\rm T}$.
\par We use a renormalization and factorization scale $\mu_0 =
\sqrt{m_b^2 + (p_{\rm T}^{b 2} + p_{\rm T}^{\overline b 2})/2}$, MRSA structure
functions~\cite{MRSA}, and $m_b$ = 4.75 GeV/$c^2$. Applying the
additional weights and the appropriate kinematical cuts ($\mid
\eta^{\mu} \mid <$ 0.6 and $\mid \eta^{\overline b} \mid <$ 1.5), we obtain
the calculated d$\sigma$/d${\rm E}_{\rm T}^{\overline b}$ and
d$\sigma$/d$\delta\phi(\mu-\overline b)$ distributions. We create the same
distributions with the requirement that the muon and $\overline b$ be
separated by $\Delta$R $>$ 1 and do a bin by bin comparison of the
calculated cross sections to define the $\Delta$R acceptance. We
have varied the renormalization scale, $b$ quark mass, and parton
distribution functions used in the MNR calculation to estimate the
systematic uncertainties in the $\Delta$R acceptance.
Table~\ref{table-bbar_jet_acceptance} shows the bin by bin values
used in the differential cross section measurements.
\section{Cross Section Results}
\label{section-results}
\par The cross section results are presented as $\mu - \overline b$ cross
sections. Since we have not specifically done flavor identification,
there is an additional factor of $1/2$ in the calculation of the
cross sections. For the semi-differential measurements, we do an
independent fit of the log$_{10}(P_{jet})$ distribution and then
correct for the acceptance in each ${\rm E}_{\rm T}$ or $\delta\phi$ bin. With
the number of $\overline b$ jets from table~\ref{table-fit_results}, the
bin by bin trigger efficiencies from
table~\ref{table-trigger_efficiency}, the combined muon acceptance
and efficiency from table~\ref{table-efficiencies}, and the $\overline b$
track and $\Delta$R acceptances from
table~\ref{table-bbar_jet_acceptance}, we calculate the the cross
section in each ${\rm E}_{\rm T}$ and $\delta\phi$ bin considered. The sum of the
7 ${\rm E}_{\rm T}$ bins is 614.4 $\pm$ 63.0 pb and the sum of the 8 $\delta\phi$
bins is 633.0 $\pm$ 70.6 pb. The results are summarized in
table~\ref{table-final_results}.
\begin{table}[tb]
\begin{center}
\begin{tabular}{|c||c|} \hline
${\rm E}_{\rm T}$ Range & Cross Section (pb) \\ \hline\hline
10 --- 15 & 168.1$^{+15.8}_{-15.8}$ \\
15 --- 20 & 152.2$^{+12.7}_{-12.8}$ \\
20 --- 25 & 106.6$^{+10.5}_{-10.5}$ \\
25 --- 30 & 61.93$^{+7.79}_{-7.79}$ \\
30 --- 40 & 72.53$^{+8.96}_{-9.43}$ \\
40 --- 50 & 29.81$^{+4.60}_{-4.68}$ \\
50 --- 100 & 23.13$^{+4.38}_{-4.23}$ \\ \hline \hline
$\delta\phi$ Range & \\ \hline\hline
0---$\frac{\pi}{8}$ & 18.36$^{+24.38}_{-18.36}$ \\
$\frac{\pi}{8}$ --- $\frac{\pi}{4}$ & 30.86$^{+18.63}_{-10.91}$ \\
$\frac{\pi}{4}$ --- $\frac{3\pi}{8}$ & 17.30$^{+4.60}_{-4.21}$ \\
$\frac{3\pi}{8}$ --- $\frac{\pi}{2}$ & 18.48 $\pm$ 3.76 \\
$\frac{\pi}{2}$ --- $\frac{5\pi}{8}$ & 24.81 $\pm$ 4.37 \\
$\frac{5\pi}{8}$ --- $\frac{3\pi}{4}$ & 37.81 $\pm$ 5.91 \\
$\frac{3\pi}{4}$ --- $\frac{7\pi}{8}$ & 108.9 $\pm$ 9.93 \\
$\frac{7\pi}{8}$ --- $\pi$ & 376.4 $\pm$ 17.7 \\
\hline
\end{tabular}
\end{center}
\caption{$\mu - \overline b$ cross sections as a function of jet ${\rm E}_{\rm T}$ and
$\delta\phi$ between the muon and $\overline b$ jet. We have not included
a common systematic uncertainty of 9.3\% in the results. Physics
backgrounds have not been subtracted at this stage.}
\label{table-final_results}
\end{table}
\subsection{Physics Backgrounds}
\par There are backgrounds which need to be included before comparing
to theoretical predictions on $b - \overline b$ production, since there are
additional sources of $\mu - \overline b$ production. Specifically, the
decay products of light mesons ($\pi$, $K$) produced in association
with $b - \overline b$ pairs or heavy particles (e.g, the $Z^\circ$ boson,
top quark production) can give a similar signature.
\par A contribution to the sample occurs when the identified muon is
not coming from a $b$ quark decay but instead from the decay of a
light meson ($\pi$ or $K$) or charm quark. In the inclusive muon
sample, the $b$ fraction is measured to be approximately
40\%~\cite{TOPPRD}, with a charm fraction of approximately 20\% and
the remaining 40\% from the decay of light mesons. Since jets from
gluons are the dominant production process in this jet ${\rm E}_{\rm T}$ range,
we assume that the light mesons come predominantly from gluon jets.
With the further assumption that the gluon splitting to $b\overline b$
probability is approximately 1.5\%~\cite{gluon_splittin}, we estimate
that in 0.6\% ($0.015 \times 0.4$) of the muon events we correctly
identify the $\overline b$ but the muon is from a light meson decay. The
case where the identified muon comes from the decay of a charm
particle can be estimated in a similar manner. With the same
assumptions about the gluon splitting to heavy quark probability
(1.5\%), a measured charm fraction of 20\%, and that approximately
75\% of charm quarks are produced via gluon splitting, we estimate
that in 0.2\% ($0.015 \times 0.75 \times 0.2$) of the muon events we
correctly identify the $\overline b$ but the muon is from a charm particle
decay.
\par With an identified fraction of 40\% $b$ muons and 50\% of the
produced $b$'s from gluon splitting~\cite{gluon_splittin}, in 20\% of
the muon events we correctly identify the $\overline b$ and the muon from
the $b$ decay. Combining these calculations yields a fractional
background in the $\mu - \overline b$ cross section of 0.04 ($ = (0.006 +
0.002) / 0.20$). We assume that this background has the same shape
as the signal and reduce the cross sections by a constant 4.0 $\pm$
2.0\% (the uncertainty is taken as half the change).
\par We have used the PYTHIA Monte Carlo program to generate $Z^\circ
\rightarrow b \overline b$ events, and the CLEO Monte Carlo program for the
decay of the resulting $B$ hadrons. We normalize the production
cross section to measured CDF cross section of $Z^\circ \rightarrow e^+
e^-$~\cite{PDG,my_result}, and apply the same $\mu$ and $\overline b$ jet
requirements as presented in section~\ref{section-dataset}. The
predicted cross section remaining after these requirements is 3.6
$\pm$ 0.28 pb, where the uncertainty includes the relative
normalization to the dielectron decay mode, the $b \rightarrow \mu$
branching fraction, and acceptance uncertainties.
Table~\ref{table-z_b_bbar} shows the contributions from this process
in the same ${\rm E}_{\rm T}$ and $\delta\phi$ bins as in
table~\ref{table-final_results}.
\begin{table}[tb]
\begin{center}
\begin{tabular}{|c|c|} \hline
${\rm E}_{\rm T}$ Range & Cross Section (pb) \\
& Statistical Uncertainty only \\ \hline \hline
10 --- 15 & 0.43 $\pm$ 0.06 \\
15 --- 20 & 0.75 $\pm$ 0.08 \\
20 --- 25 & 0.82 $\pm$ 0.09 \\
25 --- 30 & 0.60 $\pm$ 0.07 \\
30 --- 40 & 0.87 $\pm$ 0.10 \\
40 --- 50 & 0.12 $\pm$ 0.02 \\
50 --- 100 & 0.015 $\pm$ 0.008 \\
\hline \hline
$\delta\phi$ Range & \\ \hline\hline
0---$\frac{\pi}{8}$ & 0 \\
$\frac{\pi}{8}$ --- $\frac{\pi}{4}$ & 0 \\
$\frac{\pi}{4}$ --- $\frac{3\pi}{8}$ & 0.015 $\pm$ 0.014 \\
$\frac{3\pi}{8}$ --- $\frac{\pi}{2}$ & 0.031 $\pm$ 0.012 \\
$\frac{\pi}{2}$ --- $\frac{5\pi}{8}$ & 0.036 $\pm$ 0.013 \\
$\frac{5\pi}{8}$ --- $\frac{3\pi}{4}$ & 0.11 $\pm$ 0.024 \\
$\frac{3\pi}{4}$ --- $\frac{7\pi}{8}$ & 0.53 $\pm$ 0.05 \\
$\frac{7\pi}{8}$ --- $\pi$ & 2.88 $\pm$ 0.12 \\ \hline
\end{tabular}
\end{center}
\caption{Contributions from $Z^\circ \rightarrow \mu \overline b$ to the
cross section as a function jet ${\rm E}_{\rm T}$ and $\delta\phi$ between the
$\mu$ and $\overline b$ jet. There is an addition 8.0\% uncertainty in the
overall normalization.}
\label{table-z_b_bbar}
\end{table}
\par Top quark production and decay can also contribute to the $\mu -
\overline b$ cross sections. The CDF measurement of the total top cross
section is $6.8^{+3.6}_{-2.4}$ pb~\cite{TOPPRL95}. However, once we
account for branching fractions and acceptance criteria, the total
cross section from this process is less than 1 pb and will not be
considered further.
\subsection{Jet Unsmearing Procedure}
\par The cross sections measured above depend upon the selection of
jets with ${\rm E}_{\rm T} >$ 10 GeV, and in the case of the d$\sigma$/d${\rm E}_{\rm T}$
distribution, depend upon the binning of the distribution. Jets
coming from $\overline b$ quarks with transverse momentum $p_{\rm T}^{\overline b}$
will contribute to more than one bin in the measured distribution,
due to the combined effects of calorimeter energy response,
calorimeter energy resolution, and quark fragmentation. An
unsmearing procedure has been developed at CDF to account for these
effects.
\par We use Monte Carlo produced samples to define the expected jet
${\rm E}_{\rm T}$ response distribution for a given quark $p_{\rm T}$. An iterative
procedure is used to correct the measured cross sections. The quark
$p_{\rm T}$ distribution is described by a smooth function and smeared with
the simulation derived ${\rm E}_{\rm T}$ response functions. The input
distribution is adjusted until the smeared distribution matches the
measured distribution. We then perform a simultaneous unfolding of
the measured jet ${\rm E}_{\rm T}$ spectrum to the parton $p_{\rm T}$ spectrum to
account for energy loss and resolution. This unfolding corrects both
the cross section and ${\rm E}_{\rm T}$ ($p_{\rm T}$) axes.
\subsubsection{Response Functions}
\par The calorimeter single particle response in the range 0.5 to 227
GeV has been determined from both test beam data and isolated tracks
from collider data. A Monte Carlo simulation incorporating the
calorimeter response and the ISAJET, HERWIG, and PYTHIA samples is
used to determine a response function for $\overline b$ jets in the ${\rm E}_{\rm T}$
range 5 to 150 GeV, including energy loss, resolution, and jet
finding efficiency effects. For each $p_{\rm T}$, the response function
represents the probability distribution for measuring a particular
value of ${\rm E}_{\rm T}$. These response functions are convoluted with the
expected $\overline b ~ p_{\rm T}$ distributions, creating an expected ${\rm E}_{\rm T}$
distribution.
\subsubsection{Unsmearing}
\par The input $\overline b$ distribution comes from the $\mu - \overline b$
model described in section~\ref{section-acceptance}, where we have
required a muon with $p_{\rm T}>$ 9 GeV/c. We have parametrized the
distribution with a multi-quadric function and varied a scale
parameter until the smeared distribution matches the measured
distribution. Figure~\ref{fig-bbar_pt_distribution} shows the best
match $\overline b$ $p_{\rm T}$ distribution, overlayed with the smeared
distribution. Table~\ref{table-unsmeared} shows the unfolding
effects on the cross section and transverse momentum. Note that the
unsmearing procedure introduces correlated systematic uncertainties
in the bins.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_pt_smearing.ps}
\caption{The best match $p_{\rm T}^{\overline b}$ distribution, overlayed
with the smeared distribution (dashed) and the data ${\rm E}_{\rm T}$
measurement. The process is reversed to take the data ${\rm E}_{\rm T}$
distribution to a $p_{\rm T}$ distribution.}
\label{fig-bbar_pt_distribution}
\end{figure}
\begin{table*}[t]
\begin{center}
\begin{tabular}{|c|c|c||c|c|} \hline
Jet ${\rm E}_{\rm T}$ Bin & Mean Jet ${\rm E}_{\rm T}$ & $\sigma$ (pb/GeV)
& Mean ${\overline b}~p_{\rm T}$ & $\sigma$ (pb/GeV/c) \\
\hline \hline
10 --- 15 & 12.38 & 32.20 & 25.28 & 27.66 \\
15 --- 20 & 17.35 & 29.07 & 30.67 & 24.62 \\
20 --- 25 & 22.30 & 20.30 & 35.99 & 18.78 \\
25 --- 30 & 27.34 & 11.77 & 41.20 & 12.52 \\
30 --- 40 & 34.31 & 6.88 & 48.38 & 7.13 \\
40 --- 50 & 44.36 & 2.85 & 59.00 & 3.05 \\
50 --- 100 & 63.19 & 0.44 & 79.18 & 0.57 \\
\hline
\end{tabular}
\end{center}
\caption{Smeared and unsmeared means and cross sections for the 7
bins in the differential $p_{\rm T}$ measurement. The cross sections are
after background subraction and are presented here without
uncertainties. Note that the unsmearing procedure introduces
correlated uncertainties in the bins.}
\label{table-unsmeared}
\end{table*}
\subsubsection{Systematic Uncertainties}
\par Systematic uncertainties in the smearing procedure arise from
uncertainties in the knowledge of the calorimeter energy scale, the
calorimeter resolution, the jet finding efficiency, the $\overline b$ quark
fragmentation, and the effects of the underlying event in defining
the jet energy. The parameters in the smearing procedure are
adjusted to account for these uncertainties, the input distribution
is smeared, and the difference between the standard smeared
distribution and the new smeared distribution is used to estimate the
bin by bin systematic uncertainties. The uncertainties are added in
quadrature to extract a total systematic uncertainty.
Table~\ref{table-all_systematics} contains the bin by bin systematic
uncertainties.
\begin{table*}[t]
\begin{center}
\begin{tabular}{|c|c|c|c|} \hline
Variation & 10 --- 15 GeV ${\rm E}_{\rm T}$ & 15 --- 20 GeV ${\rm E}_{\rm T}$
& 20 --- 25 GeV ${\rm E}_{\rm T}$ \\
\hline \hline
Energy Scale & + 7.2\% - 4.6\%
& + 4.7\% - 3.5\% & + 9.1\% - 7.3\% \\
Underlying event & + 0.2\% - 0.2\%
& + 0.1\% - 0.2\% & + 0.2\% - 0.2\% \\
Calorimeter Resolution & + 4.4\% - 4.2\%
& + 2.6\% - 2.5\% & + 4.1\% - 4.1\% \\
Jet Finding & $\pm$ 2.6\%
& $\pm$ 0.7\% & $\pm$ 1.0\% \\
$b$ Fragmentation & + 1.0\% & - 4.0\% & - 4.7\% \\
\hline
Total & + 8.9\% - 6.7\%
& + 5.4\% - 5.9\% & +10.0\% - 9.6\%\\
\hline \hline
& 25 --- 30 GeV ${\rm E}_{\rm T}$ & 30 --- 40 GeV ${\rm E}_{\rm T}$ & 40 --- 50 GeV ${\rm E}_{\rm T}$ \\
\hline \hline
Energy Scale & +12.5\% -10.2\%
& +16.5\% -13.4\% & +20.7\% -16.5\% \\
Underlying event & + 0.3\% - 0.3\%
& + 0.4\% - 0.4\% & + 0.4\% - 0.4\% \\
Calorimeter Resolution & + 3.5\% - 3.5\%
& + 0.9\% - 0.9\% & + 4.5\% - 4.5\% \\
Jet Finding & $\pm$ 0.3\%
& $\pm$ 0.0\% & $\pm$ 0.0\% \\
$b$ Fragmentation & - 4.4\% & - 3.4\% & + 1.6\% \\
\hline
Total & +13.0\% -11.6\%
& +16.5\% -13.9\% & +21.2\% -17.2\%\\
\hline \hline
& 50 --- 100 GeV ${\rm E}_{\rm T}$ & & \\
\hline \hline
Energy Scale & +27.8\% -21.3\% & & \\
Underlying event & + 0.4\% - 0.4\% & & \\
Calorimeter Resolution & +12.7\% -12.7\% & & \\
Jet Finding & $\pm$ 0.0\% & & \\
$b$ Fragmentation & + 1.0\% & & \\
\hline
Total & +30.6\% -24.8\% & & \\
\hline
\end{tabular}
\end{center}
\caption{Systematic uncertainties for each bin in the $\mu -
{\overline b}$ differential jet ${\rm E}_{\rm T}$ distribution. There are bin to
bin correlations for each systematic variation.}
\label{table-all_systematics}
\end{table*}
\subsubsection{$\overline b$ Jet $p_{\rm T}^{min}$ Definition}
\par For future comparisons to theoretical predictions on overall
normalization, we need to define a $p_{\rm T}^{min}$ threshold for the
recoiling $\overline b$ quark. The standard definition is to take the
$p_{\rm T}$ value where $>$90\% of all decays pass the kinematic cuts. In
this case, we need to find the point where $>$ 90\% of all jets have
E$_T>$ 10 GeV. We begin with the $\overline b$ $p_{\rm T}$ spectrum shown in
figure~\ref{fig-bbar_pt_distribution} and apply the resolution
smearing to this distribution. We weight each bin in the $p_{\rm T}$
spectrum by the probability that a $\overline b$ quark with that $p_{\rm T}$
would give a jet with E$_T>$ 10 GeV. Integrating the resulting
weighted distribution gives a 90\% $p_{\rm T}^{min}$ value of 20.7 GeV/c
for the $\overline b$ jet.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_et_spectrum.ps}
\caption{The differential ${\rm E}_{\rm T}$ cross section, for
$p_{\rm T}^{\mu}>~9$~GeV/c, $|\eta^{\mu}|<$~0.6, ${\rm E}_{\rm T}^{\overline b}>$~10~GeV,
$|\eta^{\overline b}|<$~1.5, compared to theoretical predictions. The data
points have a common systematic of $\pm$ 9.5\%. The common
uncertainty in the theory points comes from the muonic branching
fraction and fragmentation model. The theory points do include
uncertainties from the smearing procedure. }
\label{fig-et_comparison}
\end{figure}
\subsection{Comparison with NLO QCD}
\par In figure~\ref{fig-et_comparison}, we show a comparison of the
differential jet ${\rm E}_{\rm T}$ cross section,
\begin{eqnarray}
\frac{{\rm d}\sigma}{{\rm dE}_{\rm T}^{\overline b}}(p_{\rm T}^{\mu}> 9~{\rm GeV/c},
|\eta^b|<1,
|\eta^{\overline b}| < 1.5, {\rm E}_{\rm T}^{\overline b} > 10~{\rm GeV}) \nonumber
\end{eqnarray}
\noindent to a prediction from the $\mu - \overline b$ model discussed in
section~\ref{section-dataset}. There is a 9.5\% common uncertainty
in the measured points, coming from the jet probability fit (5.8\%),
the $\overline b$ jet tracking efficiency (5\%), the muon acceptance and
identification efficiencies (3.9\%), the luminosity normalization
(3.6\%), and the remaining background subtraction (2\%). This common
uncertainty is displayed separately. The uncertainty in the model
prediction represents the uncertainty from the muonic branching
fraction (5\%)~\cite{PDG}, the acceptance of the muon $p_{\rm T}$ cut from
variations in the fragmentation model (5\%), which are common to all
points, and the uncertainties associated with $p_{\rm T}$ to ${\rm E}_{\rm T}$
smearing. The data has an integral value of 586. $\pm$ 61.8 pb,
while the model predicts an integral value of 383.5 $\pm$ 5.9 pb.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_pt_spectrum.ps}
\caption{The differential $p_t$ cross section, for
$p_{\rm T}^{\mu}>~9$~GeV/c, $|\eta^{\mu}|<$~0.6, $|\eta^{\overline b}|<$~1.5,
compared to theoretical predictions. The data points have a common
systematic of $\pm$ 9.5\% and there are correlated systematic
uncertainties. The uncertainty in the theory curves comes from the
muonic branching fraction and fragmentation model. }
\label{fig-pt_comparison}
\end{figure}
\par In figure~\ref{fig-pt_comparison}, we show the unsmeared
differential jet $p_{\rm T}$ cross section,
\begin{eqnarray}
\frac{{\rm d}\sigma}{{\rm d}p_{\rm T}^{\overline b}}(p_{\rm T}^{\mu}> 9~{\rm GeV/c},
|\eta^{\mu}| <0.6, |\eta^{\overline b}| < 1.5) \nonumber
\end{eqnarray}
\noindent compared to the $\overline b~p_{\rm T}$ prediction from the $\mu -
\overline b$ model, where we have included systematic uncertainties
associated with the resolution smearing on the measured points.
Again, the common normalization uncertainties are displayed
separately.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_dphi_spectrum.ps}
\caption{The differential $\delta\phi$ cross sections, for
$p_{\rm T}^{\mu}>~9$~GeV/c, $|\eta^{\mu}|<$~0.6, ${\rm E}_{\rm T}^{\overline b}>$~10~GeV,
$|\eta^{\overline b}|<$~1.5 compared to theoretical predictions. The data
points have a common systematic of $\pm$ 9.5\%. The uncertainty in
the theory curves comes from the muonic branching fraction and
fragmentation model. }
\label{fig-phi_comparison}
\end{figure}
\par In figure~\ref{fig-phi_comparison}, we show a comparison of the
differential $\delta\phi(\mu - \overline b)$ cross section,
\begin{eqnarray}
\frac{{\rm d}\sigma}{{\rm d}\delta\phi^{\mu-\overline b}}(p_{\rm T}^{\mu}>9~{\rm GeV/c},
|\eta^{\mu}|<0.6,{\rm E}_{\rm T}^{\overline b}>10~{\rm GeV},|\eta^{\overline b}| < 1.5)
\nonumber
\end{eqnarray}
\noindent to the predictions from the $\mu - \overline b$ model. The
uncertainty in the theoretical prediction represents the uncertainty
in the muonic branching fraction and fragmentation model only.
\par While we find qualitative agreement in shape between the
measured distributions and model predictions, there are some
differences. To investigate in more detail, we present in
figure~\ref{fig-data_minus_qcd} the experimental results minus the
model prediction, scaled to the model prediction for the ${\rm E}_{\rm T}$,
$p_{\rm T}$, and $\delta\phi$ distributions. The ${\rm E}_{\rm T} (p_{\rm T}$)
distributions have similar shapes for ${\rm E}_{\rm T} (p_{\rm T}) >$ 20 GeV( 35
GeV/c), but different normalizations. At lower values of ${\rm E}_{\rm T}$
($p_{\rm T}$), the measurements and predictions are in agreement. The data
$\delta\phi$ distribution is somewhat broader than the model
predictions, with enhancement in the region $\pi/4$ to $3\pi/4$, as
well as being at consistently higher values. We have also shown how
the model prediction changes with change of the renormalization and
factorization scale, by plotting the prediction for scale $\mu_0/2$
minus the prediction for $\mu_0$, scaled to the prediction for
$\mu_0$. The integral cross section increases by 7\%, with very
little change as a function of ${\rm E}_{\rm T}$ or $p_{\rm T}$. In the $\delta\phi$
distribution, the $\mu_0/2$ prediction is uniformly larger than the
$\mu_0$ prediction, except for the region $\delta\phi \approx \pi$.
\begin{figure}
\epsfysize=7in
\epsffile[0 90 594 684]{figure_data_minus_qcd.ps}
\caption{For the (a) d$\sigma$/d${\rm E}_{\rm T}$, (b) d$\sigma$/d$p_{\rm T}$, and (c)
d$\sigma$/d$\delta\phi$ distributions, we plot the difference between
the data measurement (filled circles) and the model prediction,
scaled to the model prediction. There is a common systematic
uncertainty of 9.5\% in all the points, which has not been included
in the error bar. The open circles are the model prediction for
renormalization scale of $\mu_0/2$ minus the model prediction for
$\mu_0$, scaled to the model prediction for $\mu_0$.}
\label{fig-data_minus_qcd}
\end{figure}
\par Recent work has shown that the addition of an intrinsic $k_{\rm T}$
kick to a next-to-leading order QCD calculation improves the
agreement between measurements and predictions for both direct photon
production~\cite{CTEQ_kt} and charm production~\cite{FMNR}. We have
investigated the effects of additional intrinsic $k_{\rm T}$ in the $\mu -
\overline b$ model. We use a Gaussian distribution with mean 0 and
adjustable width to model the magnitude of the kick, with a random
azimuthal direction. With widths of 2 - 4 GeV/c, we find that the
dominant effects occur for $\delta\phi <$ 1 radian. The cross section
for $\delta\phi <$ 1 is predicted to change by approximately 7\% with
a width of 4 GeV/c. With the current statistical uncertainties at
small $\delta\phi$ (ranging from 25 - 100 \%), we are unable to
distinguish effects at that level. Similarly, the dominant effect in
the $p_{\rm T}^{\overline b}$ distribution occurs in regions where we have no
sensitivity ($p_{\rm T}^{\overline b} <$ 20 GeV/c). We conclude that the
addition of intrinsic $k_{\rm T}$ with width of 4 GeV/c does not account
for the difference between the model prediction and the measurement.
\section{Conclusions}
\label{section-conclusions}
\par We have presented results on the semi-differential $\mu -
\overline b$ cross sections as a function of the $\overline b$ jet transverse
energy (d$\sigma$/d${\rm E}_{\rm T}$), $\overline b$ transverse momentum
(d$\sigma$/d$p_{\rm T}$), and the azimuthal opening angle between the muon
and the $\overline b$ jet (d$\sigma$/d$\delta\phi$). These results are
based on precision track reconstruction in jets. The effects of
detector response and resolution have been unfolded to translate the
results from $\overline b$ jets to $\overline b$ quarks. We have compared these
results to a model based on a full NLO QCD calculation~\cite{MNR}.
We have investigated the effects an additional intrinsic $k_{\rm T}$ and
find that it cannot account for the difference between the
measurements and the model prediction. Unlike previous CDF
measurements~\cite{cdf_bmu,CDF_Bmes}, a normalization change alone
does not account for the differences between this measurement and the
model prediction.
\par We thank the Fermilab staff and the technical staffs of the
participating institutions for their vital contributions. This work
was supported by the U.S. Department of Energy and National Science
Foundation; the Italian Istituto Nazionale di Fisica Nucleare; the
Ministry of Education, Science and Culture of Japan; the Natural
Sciences and Engineering Research Council of Canada; the National
Science Council of the Republic of China; the A. P. Sloan Foundation;
and the Alexander von Humboldt-Stiftung.
|
\section{1. Introduction}
The molecular gas in the Galactic Center region
is generally confined in a thin dense layer of a few tens
of pc thickness composing the nuclear molecular disk,
while a small amount of out-of-plane gas is present
extending up to $\sim 50$ pc above the disk
(see Paper I and the literature cited therein).
In paper I (Sofue 1995) we have reanalyzed the longitude-velocity $(l, V)$
diagrams of the $^{13}$CO line data of Bally et al (1987, 1988),
and have shown the presence molecular arms and ring structure within the
nuclear disk.
Besides the disk components, the central 200 pc region is characterized
by the ``200-pc expanding molecular ring'', which exhibits high
positive- and negative-velocity features consisting of an `ellipse' in the
$(l, V)$\ diagram (Scoville 1972: Kaifu et al 1972, 1974).
It has been suggested that this structure is
vertically extended for about 100 pc perpendicularly to the galactic plane,
and its cylindrical structure has been pointed out (Sofue 1989).
Alternatively, it might be a part of a ``parallelogram'' in the
$(l, V)$\ plot (Binney et al 1991), or
this feature might be an inner part
of an expanding tilted disk of 2 kpc radius (Liszt and Burton 1978).
In this paper, we revisit this prominent feature
and investigate its vertical structure based on $(b, V)$\ diagrams
of the $^{13}{\rm CO}~(J=1-0)$ and CS lines observed by Bally et al (1987)
with the Bell Telephone 7-m off-set Cassegrain telescope (BTL survey).
The details of the data and analysis method are given in Paper I (Sofue 1995).
\section{2. Dumbbell-Shaped Expanding Molecular Shell}
\sub{2.1. Expanding Ring in Longitude-Velocity Diagrams}
The 200-pc expanding molecular ring
is recognized in the $(l, V)$\ diagrams as the high-velocity
arcs at $V_{\rm LSR} \sim \pm 150-200$ km s$^{-1}$, which composes a tilted
ellipse (Bally et al 1978, 1988).
It is known that the vertical extent of this feature is as large
as $\sim 100$ pc (Sofue 1989).
The vertical extent is significantly greater than that of the nuclear disk
component which contains 85\% of the total emission in the diagram.
Since the ``expanding'' component seen in the $(l, V)$\ diagrams
amounts to only 15\% of the total emission in the central $\pm 1^\circ$ region,
it is not a major disk component.
The positive-velocity side of the expanding ring is most clearly visible at
$b\sim 4'$, while the negative-velocity side is most clearly seen
at $b\sim -15'$.
In order to see how the ellipse shows up across its densest parts,
we have combined $(l, V)$\ plot at $b=4'$ for positive velocity side and that
at $b=-15'$ for the negative velocity side (Fig. 1).
In the figure the expanding ellipse appears as an almost symmetric
feature around the nucleus.
\centerline{-- Fig. 1 --
\sub{2.2. Latitude-Velocity $(b, V)$\ Diagrams}
In order to investigate the vertical structure of these features, we
make use of $(b, V)$\ diagrams.
Fig. 2 shows typical $(b, V)$\ plots averaged in 10$'$ interval of longitude,
where local components have been subtracted (see Paper I for the analysis).
The expanding ring feature is now
recognized as arched features at $V_{\rm LSR} \sim \pm 150 - 180$ km s$^{-1}$.
These arcs are largely extended in the $b$ direction over
$\sim \pm 20'$ (50 pc),
and are distinguished from the massive disk component which has a
smaller scale height of $\sim \pm 5-10'$.
The arcs are all convex with respect to the center of the diagrams.
The arcs appear to compose an ellipse, which is slightly squeezed near
the galactic plane, representing a dumbbell-shaped expanding
shell feature in the $(b, V)$\ space with its neck (equator) at $b=-3'$.
This equator coincides with the plane of the
major molecular disk crossing Sgr A.
\centerline{-- Fig. 2 --}
The velocity width of the shell (velocity difference between the
positive and negative-most edges of the arcs in the $(b, V)$\ plot)
appears to vary with longitude.
The width attains maximum at around $l\sim 0^\circ$, and decreases with
the distance from the center.
Although the present data do not cover the regions at
$|l|>1^\circ$, the decrease of the width indicates that the
tangential edges of the expanding
shell are located at $l=\pm 1^\circ.2~(\pm 180$ pc).
Since the higher-velocity part of an expanding feature in a position-velocity
plots correspond to a larger-radius part, the $(b, V)$\ ellipse (or dumbbell)
features can be most naturally attributed to a cross section of an expanding
spheroidal (or dumbbell-shaped) surface (shell), which we call the
dumbbell-shaped expanding molecular shell (EMS).
{}From the $(l, V)$\ diagrams, the EMS is rotating at about 70 km s$^{-1}$\ around the
nucleus with the negative-velocity part being in the near-side.
This rotation speed is much smaller than the galactic rotation velocity,
and may indicate that the gas has been carried from the inner region.
\sub{2.3. Asymmetry about the Galactic Plane,
and Anti-correlation with the Disk Component}
The dumbbell-shape in the $(b, V)$\ space has its equator (neck) at $b=-3'$.
which coincides with the plane of the major molecular disk.
However, the $(b, V)$\ arcs are not symmetric with respect to the equator:
The lower-$b$ counterpart is missing at $V_{\rm LSR}>0$ km s$^{-1}$, while the
upper-$b$ counterpart is missing at $V_{\rm LSR}<0$ km s$^{-1}$.
It is remarkable that the latitude of the missing part of the $(b, V)$\ arcs
coincide with the latitude where the major disk component is most clearly
seen.
This can be also recognized in $(l, V)$\ diagrams:
The expanding feature at negative $V_{\rm LSR}$
is missing in the $(l, V)$\ diagrams at positive latitudes where the major
molecular arm (Arm I in Paper I) is most prominent.
On the other hand, the feature with positive $V_{\rm LSR}$\ is missing at negative
latitudes where Arm II is most prominent.
Fig. 3 shows the relation of the Arms and missing parts of
the expanding molecular features.
In paper I, we interpreted that Arms I and II compose a ring of radius 120 pc
around the Galactic Center, and they are located at opposite
halves of the ring.
The anti-correlation of the expanding shell features with the
Arms can be understood, if the near side (negative velocity side)
of the expanding shell has been interrupted by Arm I which is
in front of the nucleus.
On the other hand, the far side of the shell has been interrupted
by Arm II which is on the far side of the nucleus.
\centerline{-- Fig. 3 --}
\sub{2.4. Parameters for the EMS}
The mass of the whole EMS can be obtained by using a wider-area survey
of Bally et al (1988).
The $^{13}$CO intensity toward $l\sim 1^\circ.2$ is roughly $\sim 50 $ K km s$^{-1}$.
We use a recent value of the $^{12}CO$ to H$_2$ conversion factor at
the galactic center,
$0.92\times 10^{20}$ H$_2$ cm$^{-2}$/K km s$^{-1}$ (Arimoto et al 1995),
and assuming the $^{12}$CO to $^{13}$CO temperature ratio to be 6.2
(see Paper I).
{}From the ($V_{\rm LSR}-l$) diagram we calculated an averaged intensity in
$-0^\circ.6 \le b \le +0^\circ.6$,
and estimated the molecular (H$_2$\ + He gas)
mass of the EMS as $ M\sim 0.8 \times 10^7 M_{\odot \hskip-5.2pt \bullet}$.
Since the expansion velocity is about 160 km s$^{-1}$, this leads to a total kinetic
energy of expansion of $ \sim 2 \times10^{54}$ ergs.
Table 1 summarizes the derived parameters for the molecular shell.
\centerline{-- Table 1 --}
\sub{2.5. Comparison with Other Data}
A wider view of the expanding-ring feature can be seen in the low-resolution
$(l, V)$\ diagram.
Two vertically extended ridges at $l=\pm 1^\circ.2$
are also recognized in the low-resolution integrated intensity maps (Bally et
al 1988).
In Fig. 4 we compare this global feature with other data as
reproduced from Sofue (1989).
Fig. 4a is a low-resolution $(l, V)$\ diagram of the CO line emission.
Fig. 4b shows the CO intensities integrated within
$-100 \le V_{\rm LSR} \le -20$ km s$^{-1}$\ for
$l \le 0^\circ$ and $20 \le V_{\rm LSR} \le 100$ km s$^{-1}$\ for $l>0^\circ$, respectively,
as reproduced from Bally et al (1988).
This map represents `cross sections' of the tangential parts of the EMS,
which comprise vertical spurs.
The spur at $l=1^\circ.2$ is shifted toward the positive latitude side
and can be traced from $b\sim-0^\circ.2$ to $+0^\circ.4$, whereas the spur at
$l=-1^\circ.2$ toward negative latitude, suggesting that the expanding
shell is tilted.
\centerline{-- Fig. 4 --}
Figs. 4c to 4e compare the expanding molecular feature
with the 10 GHz (Handa et al 1987) and 2.7 GHz (Reich et al 1984)
radio continuum maps as well as the IRAS 60 $\mu$m map (Beichman et al 1985).
Radio continuum maps reveal a number of spurs which emerge from the galactic
plane.
Near the tangential direction of the EMS at $l \sim 1^\circ.2$,
a spur is visible extending from the galactic plane to
$(l=1^\circ.2,~b=0^\circ.4)$, which appears to be convex with respect to the
Galactic center in a consistent sense with the EMS.
The spurs are also identified on the FIR (far-infrared) emission map of
IRAS at 60 $\mu$m (Beichman et al 1985), indicating that the EMS contains
warm dust.
The observed brightness of the thermal radio emission is about 100 mK toward
the spur associated with the EMS. If we assume an optically thin
ionized gas of $T_{\rm e}\sim 10^4$ K and $T_B\sim 0.1$ K at $\nu=10$ GHz,
we obtain an emission measure of $EM \sim 4\times 10^3 ~{\rm pc~cm}^{-6}$.
We then obtain a thermal electron density $n_{\rm e}\sim6~{\rm cm}^{-3}$, and
the total mass of HII gas is derived to be $\sim 4\times 10^5M_\odot$ for the
same geometry as the EMS.
Then the thermal energy involved in the EMS is of the order of $\sim 10^{51}$
ergs.
A hot plasma of $ \sim10^8$ K has been observed in the Galactic center
in the iron line at 6.7 keV with the Ginga Satellite
(Koyama et al 1989).
The emission region is extended with a size of $1^\circ.8$ (250 pc) about the
galactic center, and the observed size of
the hot plasma region fits the interior of the EMS.
Considering the angular resolution of the detector and data-sampling grids,
we may suggest that the plasma is surrounded by the EMS.
However, the time scale of the expansion of the hot plasma
($\sim$ radius/sound velocity $ \sim 100$ pc/1000 km s$^{-1}$ $\sim 10^5$ yr)
is much shorter than the expansion time scale of the molecular feature.
Therefore, the origin of the hot plasma would not be directly related to
the origin of the EMS, but the molecular gas could only act to confine the
plasma in the spheroidal, or possibly cylindrical, space.
Moreover, the thermal energy contained by the hot plasma
($ \sim3 \times 10^{53}$ ergs; Koyama et al 1989; Sofue 1989) is much smaller
than the kinetic energy of the EMS, which implies that the hot plasma
cannot be a driving power for the EMS formation.
\section{3. Discussion}
We have shown that the so-called expanding molecular ring
in the galactic center is a part of a dumbbell-shaped expanding shell
with its equator (neck) in the dense molecular disk.
The shell is squeezed near its equator probably due to interruption by
the dense disk gas near the galactic plane.
The positive latitude and negative-velocity side (near side of the nucleus)
of the EMS appears to be interrupted by the near-side Arm I, while
the negative-latitude and positive-velocity side (far side) of the EMS
is interrupted by the far-side Arm II.
These structures suggest that the EMS is an expanding molecular
spheroid interacting with the dense molecular disk which comprises
ring and arms.
Fig. 5 illustrates the proposed dumbbell-shaped
spheroid in the $(l, b, V_{\rm LSR})$ space.
This may be transposed to a three-dimensional expanding spheroid
in the $(x, z, y)$ space, where $x$, $z$, and $y$ are axes along the
galactic plane, rotation axis, and depth in the line-of-sight direction,
respectively.
The expanding molecular gas is observed to be distributed in the
intermediate latitude regions of the spheroid.
The gas in the equatorial and polar regions are missing.
The asymmetry of the distribution of gas in the EMS
with respect to the galactic plane
would be somehow related to the fact that the major molecular disk
containing Arms I and II is tilted by 5$^\circ$\ from the galactic plane.
\centerline{-- Fig. 5 --}
The total kinetic energy of the shell is approximately $2\times 10^{54}$
ergs, which is equivalent to energy released by
$\sim 10^3$ type II SN explosions.
It is likely that the Galactic center experienced a phase of intense
star formation in the past $\sim 10^6$ years.
Suppose that a coherent star formation containing some $10^3$ SN explosions
(mini burst) lasted for a period short enough compared to the EMS's life
time, e.g., within $\sim 10^5$ yr at the central part of the nuclear disk
some $10^6$ yr ago.
Then, subsequent SN explosions would have formed many shocked shells.
Soon after the explosions, the shock fronts would have accumulated on
a single large shell expanding into the disk and halo.
Such a shocked shell would become deformed by the disk-to-halo density
gradient, and would easily attain a dumbbell shape with the neck
in the disk plane.
Simulations of a shock front evolution in a disk-halo
system have shown that an initially spherical shocked shell becomes
elongated in the direction of the halo and then attain a dumbbell shape
(Sofue 1984, 1994).
If the gravity, which is stronger in the direction
perpendicular to the galactic plane, as well as the fact that the disk gas
before repulsion was rotating with the centrifugal force,
are taken into account, the oblateness of the EMS could be accounted for.
Finally, we comment on another model of the expanding molecular feature.
The closed orbit model based on the ``parallelogram" in the
$(l, V)$\ plot, as proposed by Binney et al (1991),
would be an alternative promising possibility to explain the present
expanding molecular features in the $(l, V)$\ plot.
If the model is successful in reproducing the rigid rotation features of
the disk component, which
shares 85\% mass of the total emission in the $(l, V)$\ plot (Paper I) of the
presently discussed area, this model would be attractive, since most of
the $(l, V)$\ features can be understood under a unified theoretical
scheme of gas dynamics in a central bar potential.
It is also an open question about this model, if it can
explain the three dimensional structure as pointed out in this paper.
\v\v
{\bf Acknowledgement}: The author would like to express his sincere thanks
to Dr. John Bally for making him available with the molecular line data
in a machine-readable format.
\section{References}
\parskip=0pt
\def\hangindent=1pc \hangafter=1 \no{\hangindent=1pc \hangafter=1 \noindent}
\hangindent=1pc \hangafter=1 \no Arimoto, N., Sofue, Y., Tsujimoto, T. 1995, submitted to AA.
\hangindent=1pc \hangafter=1 \no{Bally, J., Stark, A.A., Wilson, R.W., and Henkel, C. 1987, ApJ Suppl 65,
13.}
\hangindent=1pc \hangafter=1 \no{Bally, J., Stark, A.A., Wilson, R.W., and Henkel, C. 1988, ApJ 324, 223.}
\hangindent=1pc \hangafter=1 \no{Beichman, C.A., Neugebauer, G., Habing, H.J., Clegg, P.E., and
Chester, T.J. (ed.) 1985, IRAS Explanatory Supplement, JPL D-1855
(Jet Propulsion Laboratory, Pasadena).}
\hangindent=1pc \hangafter=1 \no Binney, J.J., Gerhard, O.E., Stark, A.A., Bally, J., Uchida, K.I., 1991
MNRAS 252, 210.
\hangindent=1pc \hangafter=1 \no{Handa, T., Sofue, Y., Nakai, N. Inoue, M., and Hirabayashi, H. 1987,
PASJ 39, 709.}
\hangindent=1pc \hangafter=1 \no{Kaifu, N., Iguchi, T., and Kato, T. 1974, PASJ 26, 117.}
\hangindent=1pc \hangafter=1 \no{Kaifu, N., Kato, T., and Iguchi, T. 1972, Nature 238, 105.}
\hangindent=1pc \hangafter=1 \no{Koyama, K., Awaki, H., Kunieda, H., Takano,S., Tawara, S., Yamanuchi, S.,
Hatsukade, I., and Nagase, F. 1989, Nature 339, 603.}
\hangindent=1pc \hangafter=1 \no{Reich, W., F{\"u}rst, E., Steffen, P., Reif, K., and Haslam, C.G.T. 1984,
AA Suppl 58, 197.}
\hangindent=1pc \hangafter=1 \no{Scoville, N.Z. 1972, ApJ 175, L127.}
\hangindent=1pc \hangafter=1 \no{Sofue, Y. 1984, PASJ 36, 539.
\hangindent=1pc \hangafter=1 \no Sofue, Y. 1989, Ap. Let. Com. 28, 1
\hangindent=1pc \hangafter=1 \no Sofue, Y. 1994, ApJ.L. 431, L91.
\hangindent=1pc \hangafter=1 \no Sofue, Y. 1995, PASJ, in press (Paper I).
\hangindent=1pc \hangafter=1 \no{Sofue, Y., and Handa, T. 1984, Nature 310, 568.}
\vfil\break
\settabs 2 \columns
\noindent Table 1: Dumbbell-shaped expanding molecular shell$^*$.
\vskip 2mm
\hrule
\vskip 2mm
\+ Radius\dotfill & $1^\circ.2$=180 ($\pm 5$) pc \cr
\+ Vertical extent\dotfill & $\sim \pm 50$ pc \cr
\+ Thickness of the shell \dotfill & $\sim15$ pc \cr
\+ Rotation velocity\dotfill & 70 ($\pm 10$) km s$^{-1}$ \cr
\+ Expansion velocity\dotfill & 160 ($\pm 5$) km s$^{-1}$ \cr
\+ Molecular gas mass\dotfill & $ \sim 0.8\times 10^7 M_{\odot \hskip-5.2pt \bullet} $ \cr
\+ Kinetic energy\dotfill & $ \sim 2 \times10^{54}$ ergs \cr
\vskip 2mm
\hrule
\noindent $*$ The distance to the galactic center is taken to be 8.5 kpc.
\noindent $\dagger$ 10-keV ($1.2\times10^8$ K) thin hot plasma detected by
Koyama et al (1989).
\vfil\break
\section{Figure Captions}
\v\v
Fig. 1: A composite $(l, V)$\ diagram by combining data at $b=+4'$ for the upper
half ($V_{\rm LSR}>0$ km s$^{-1}$) and that at $b=-15'$ for the lower half ($V_{\rm LSR}<0$ km s$^{-1}$).
The map has been smoothed by a Gaussian function with a half-width
$1' \times 2.7$ km s$^{-1}$.
This diagram shows the expanding ring feature at its clearest part.
Contours are in unit of K $T^*_{\rm A}$\ at levels
$0.1\times$(1, 2, 3, 4, 5, 6, 8, 12, 16, 20, 25).
The local and foreground CO emissions have been subtracted by applying
the ``pressing method'' to the BTL cube data (Bally et al 1987)
(see Paper I for the procedure).
\v\v
Fig. 2:
(a) Typical $(b, V)$\ diagrams after subtraction of local/foreground emission.
The expanding molecular shell can be fitted by a dumbbell-shape
as illustrated by the full lines with its neck (equator) at $b=-3'$.
Contours are in unit of K $T^*_{\rm A}$\ at levels
$0.1\times$(1, 2, 3, ..., 9, 10, 12, 14, 16, 18, 20, 25, 30, 35, 40).
(b) All $(b, V)$\ diagrams at every 5$'$ longitude interval in grey scale.
Each diagram has the same $b$ and $V_{\rm LSR}$\ coverage as in (a).
The longitude decreases from the left-bottom at $l=+1^\circ$
to the right, toward top, and ends at the right-top panel at $l=-1^\circ$.
\v\v
Fig. 3: $(l, V)$\ diagrams at $b=+2$ to $+4'$ and at $b=-10$ to $-8'$.
This diagrams show an anti-correlation of the disk components Arms I and
II with the missing parts of the expanding features at positive and
negative latitude sides, respectively.
\v\v
Fig. 4: Comparison in the same scale of $^{13}$CO$(J=1-0)$ data with
radio and FIR maps (reproduced from Sofue 1989):
(a) The $(l, V)$\ diagram of the $^{13}$CO gas (reproduced from Bally et al 1988).
(b) CO intensity distribution integrated in the velocity ranges from 20 to
100 km s$^{-1}$\ for $l \ge 0^\circ$, and from $-20$ to $-200$ km s$^{-1}$\ for $ l \le 0^\circ$
(Bally et al 1988).
(c) A radio continuum map at 2.7 GHz (Reich et al
1984). The background emission of scale sizes greater than $ 0^\circ.4$
has been subtracted, and the map is smoothed to $6'$ resolution.
(d) The same as (c) but at 10 GHz (Handa et al 1987).
(e) The same as (c) but at 60 $\mu$m made from the IRAS data (Beichmann
1985).
\v\v
Fig. 5: Schematic view of the dumbbell-shaped shell
in the $(l,b,V_{\rm LSR})$ space, which may be transposed to a similar
shaped three-dimensional expanding spheroid, which is rotating at
70 km s$^{-1}$ and expanding at 160 km s$^{-1}$.
The molecular gas is distributed in the intermediate-latitude regions
of the spheroid, and is missing in the equatorial and polar regions.
\bye
|
\section{Introduction}
Topology-changing transitions play an important role in the electroweak
theory since they are accompanied by baryon-number violating processes
\cite{thooft}.
At zero temperature these processes can only appear as
true quantum effects since the energy of the sphaleron configuration
is a barrier separating two gauge equivalent vacua \cite{sphaleron}.
At temperatures far above the sphaleron energy one would expect
that this barrier height,
as well as the concept of vacuum to vacuum amplitude,
are irrelevant for the transition rate.
Unfortunately there presently is no way to calculate
analytically the transition rate at these high temperatures.
{}From a numerical point of view the situation is not any better if we
insist on a fully quantum treatment, since the only
non-perturbative tool available is lattice gauge theories
in the imaginary time formalism. Real time transition rates
are not easily extracted in this formalism.
The object of our study is the topological charge
\begin{formula}{1.1}
B(t) = \frac{1}{32 \pi^2}\int^t_0 \int d^3x\,
F^a_{\mu\nu}\tilde{F}^{a \mu\nu}= N_{{\rm cs}}(t)-N_{{\rm cs}}(0),
\end{formula}
where the Chern-Simons number
\begin{formula}{1.1b}
N_{{\rm cs}}(t) \equiv \frac{1}{32\pi^2} \,\epsilon_{ijk} \int d^3x
\left(F^a_{ij} A^a_k - \frac{1}{3} \epsilon^{abc} A^a_iA^b_jA^c_k\right).
\end{formula}
The theory has a periodic (with period 1) potential in the
$N_{\rm cs}$ direction. This periodicity is due to the symmetry under large
gauge transformations, changing $N_{\rm cs}$ by an integer. There also are
dynamical processes whose result for a field configuration is the same as a
large gauge transformation. These processes have an integer topological
charge. Due to nonlinear interactions of $N_{\rm cs}$ with other degrees
of freedom of the theory it is expected that for large times $t$ $B(t)$ is
a random walk in a periodic $N_{\rm cs}$ potential
\begin{formula}{1.1x}
\left\langle B^2(t)\right\rangle_T=\Gamma Vt,
\end{formula}
where $V$ is the space volume,
by $\left\langle\right\rangle_T$ we mean average over the canonical ensemble, and $\Gamma$, the
diffusion constant per unit volume, is the transition rate we are interested in
\cite{khls}.
At high temperatures processes with an integer topological charge occur
predominantly by classically allowed thermal activation rather than by quantum
tunneling. Some time
ago it was suggested \cite{grs} to exploit this property and to determine
$\Gamma$ entirely
within the classical theory. This is done by solving {\it classical} equations
of motion for the fields and averaging the resulting $B^2(t)$ over the
{\it classical} canonical ensemble. Behind this approach is the expectation
that the topology-changing transitions mostly involve field fluctuations with
long wavelengths and large magnitudes (hence many quanta). Such fluctuations
are well described by classical statistical mechanics. An argument usually
given to support this picture is as follows. Consider a classically allowed
process with $B(t)=1$, whereby the field configuration evolves from the
vicinity of one minimum in the $N_{\rm cs}$ potential to the vicinity of
another
one. If a typical linear space-time size of the process is $r$, the system
will cross an energy barrier of height $E\sim 1/\alpha r$, where
$\alpha=g^2/4\pi$. While large $r$ are favored energetically,
at high temperatures $r$ is unlikely to exceed the inverse
of the magnetic screening mass $m_{\rm mag}\sim\alpha T$. We then expect that
processes with $r\sim 1/\alpha T$ give the most important contribution to
$\Gamma$. In the electroweak
theory, where $\alpha_{\rm w}\approx 1/30$, these $r$ are much larger than the
thermal
wavelength, {\it i.e.,} fluctuations of size $r$ can be treated classically.
In order to estimate the transition rate, the Boltzmann factor associated with
the barrier, $\exp(-E/T)\sim 1$, is divided by the space-time volume $r^4$
of the process, to give \cite{am}
\begin{formula}{1.1y}
\Gamma=\kappa (\alpha T)^4,\label{esti}\end{formula}
where $\kappa$ is a dimensionless constant.
This estimate should be valid at temperatures far above the electroweak
transition
scale. We also expect that at such high temperatures the mass scales of the
Higgs sector are irrelevant, and that $\Gamma$ can be reliably obtained from
a pure Yang-Mills theory.
The classical approximation offers an enormous simplification in computing
real-time correlation functions. However,
it is well known that classical field theory at finite temperature suffers
from the Rayleigh-Jeans ultraviolet divergence. This means that, upon cutoff
regularization, classical thermodynamic average of a generic observable
will be sensitive to thermal fluctuations at the cutoff scale $\Lambda$
\cite{bms}.
The classical rate $\Gamma$, if it is to be trusted, must not show such
sensitivity. This is indeed what we expect: considerations leading to
the estimate (\ref{esti}) may be repeated for a purely classical theory which
also possesses a magnetic mass $m_{\rm mag}\sim\alpha T$. In the quantum
case, we required separation between the magnetic and the thermal scales:
$\alpha T\ll T$. Likewise, in the classical theory we should require separation
between the magnetic and the cutoff scales: $\alpha T\ll \Lambda$. These
two conditions make it plausible that $\Gamma$ is not strongly affected by either
quantum corrections or the classical cutoff.
On the other hand,
it is not clear how the $\Gamma$ independence of $\Lambda$ is technically
possible: after
all, $\dot B(t)$, the topological charge per unit time, is an ultraviolet
divergent quantity whose classical standard deviation is given perturbatively
by
\begin{formula}{1.3b}
\left\langle (\dot B )^2 \right\rangle_T =
\left(\frac{1}{8\pi^2}\right)^2
\left\langle \left(\int d^3x E^a_i(x)B^a_i(A(x)) \right)^2 \right\rangle_T \sim
T^2 V \L^3,
\end{formula}
where $E^a_i$ and $B^a_i$ are, respectively, color electric and magnetic
fields,
and $V$ is the space volume.
One then naturally expects the Rayleigh-Jeans divergency to show up in
$\left\langle B^2(t)\right\rangle_T$ itself.
The key point, confirmed by the numerical simulations in the following, is that
the UV divergency shows up in $\left\langle B^2(t)\right\rangle_T$ in a way that does not affect
$\Gamma$. To see how this could be anticipated, consider the structure of
$\left\langle B^2(t)\right\rangle_T$ in some more detail. The classical motion of $N_{\rm cs}$ can
be thought of as consisting of two pieces: ``true'' thermal fluctuations
and a random walk between gauge equivalent vacua.
If the
theory did not have an infinite series of degenerate vacua, $N_{{\rm cs}}$ would
simply fluctuate around zero, as it does in an Abelian theory.
The properties of these thermal fluctuations can be studied perturbatively.
As discussed in the Appendix,
in an Abelian theory, where no random walk of $N_{\rm cs}$ is possible,
$\left\langle B^2(t)\right\rangle_T$ is bounded
\begin{formula}{1.3d}
\left\langle B^2(t)\right\rangle_T^{{\rm Abelian}}\leq
2\left\langle (N_{{\rm cs}})^2 \right\rangle_T^{{\rm Abelian}}
\end{formula}
and at large $t$ fluctuates around the average value
\begin{formula}{1.3dd}
\nu\equiv\left\langle (N_{{\rm cs}})^2 \right\rangle_T^{{\rm Abelian}}\approx 0.00228
{{\left(\a T\right)^2}\over a} V,
\end{formula}
diverging, as expected, linearly with the lattice cutoff $\L=1/a$.
In the nonabelian
theory the random walk of $N_{\rm cs}$ is superimposed with this thermal
background. A reasonable approximation to the nonabelian $\left\langle B^2(t)\right\rangle_T$
is then
\begin{formula}{1.3e}
\left\langle B^2(t) \right\rangle_T = c\nu + \Gamma V t,
\end{formula}
where $c\approx 3$, assuming that thermal
fluctuations of $N_{\rm cs}$ are exactly those for three independent copies
of an Abelian theory.
If we let $\L \to \infty$ for fixed $t$ the r.h.s. of (\ref{1.3e}) diverges.
On the other hand, for a fixed cutoff \rf{1.3e} is well approximated by
(\ref{1.1x}) for large enough $t$ and allows unambiguously to extract $\Gamma$.
We will show that
numerical simulations, using a classical thermal ensemble
with a lattice cutoff $\L = 1/a$, point
to a $\Gamma$ independent of $a$ for sufficiently small $a$.
On dimensional grounds this means that \rf{1.1y} holds classically in the
continuum limit, with $\kappa$ of \rf{1.1y} being
a {\it non-perturbative constant of the classical
theory}. For the gauge group $SU(2)$ we find $\kappa = 1.09 \pm 0.04$.
This is the central result of our work.
The ultraviolet insensitivity of $\Gamma$ extracted from \rf{1.3e}
might be linked to the topological
aspect of the special field tensor combination
$F_{\mu\nu}^a \tilde{F}^{a \mu\nu}$ in non-abelian field theories.
Replacing $F_{\mu\nu}^a \tilde{F}^{a \mu\nu}$ by a generic bilinear of field
tensors with zero average would yield a UV divergent transition rate.
We illustrate this point by measuring in addition the
quantity $\eta(t)$ whose diffusion rate is related to the shear viscosity of
the gluon plasma \cite{wyld}:
\begin{formula}{1.5}
\eta(t) = \int_0^t dt \int d^3 x \left(T_{11}-T_{22}\right),
\end{formula}
where $T_{ij}$ are components of the energy-momentum tensor.
\section{The model and the numerical method}
The philosophy behind the classical measurements of real-time quantities
has been presented in detail
in a number of articles \cite{jan,alex1}.
Here we only concentrate on the essential technical points.
We work in the temporal gauge. The classical dynamics of $SU(2)$ Yang-Mills
theory on the lattice is given by Kogut-Susskind Hamiltonian
\begin{formula}{2.1}
H={1\over 2}\sum_l E_l^\alpha E_l^\alpha
+ \sum_\Box\left(1-{1\over 2}{\rm Tr}U_\Box\right).
\end{formula}
The variable set consists of SU(2) matrices $U_l$ and lattice analogs of
electric field $E_l$, both residing on the links of a cubic lattice.
We shall also use notation $l=j,n$ for a link emanating from site $j$
in positive direction $n$.
The second term in (\ref{2.1}) is the standard plaquette term, representing the
color magnetic energy. Every variable $v$ evolves in time according to its
canonical equation $\dot v=\{H,v\}$, derived using Poisson brackets between
the independent variables:
\begin{formula}{2.2}
\{E^\alpha_l,E^\beta_{l'}\}=2\delta_{ll'}\epsilon^{\alpha\beta\gamma}E^\gamma_l;
\ \ \{E^\alpha_l,U_{l'}\}=-i\delta_{ll'}U_l\sigma^\alpha;
\ \ \{U_l,U_{l'}\}=0
\end{formula}
(no summation over repeated $l$; $\sigma^\alpha$ are the Pauli matrices).
The time evolution is subject to the set of three Gauss' constraints per site
\begin{formula}{2.3}
\sum_n\left(\overline{E}^\alpha_{j,n}-E^\alpha_{j-n,n}\right)=0,
\end{formula}
where
$$\overline{E}^\alpha_{j,n}\equiv {1\over 2}E^\beta_{j,n}{\rm Tr}
\left(\sigma^\alpha U_{j,n}\sigma^\beta U_{j,n}^\dagger\right).$$
These constraints
are to be imposed on initial conditions. We use the leapfrog algorithm to
integrate the equations of motion, performing {\it exact} integration of the
$E$
equations for fixed $U$ and of the $U$ equations for fixed $E$. In this way
the Gauss' laws are obeyed exactly, independently of the time step, and the
only
source of spurious static charge is the limited computer accuracy.
The increment of Chern-Simons number is computed as in (\ref{1.1}) with
the topological charge per unit time approximated by
\begin{formula}{2.4}
\dot N_{\rm cs}={i\over{32\pi^2}}\sum_{j,n}\left(\overline{E}^\alpha_{j,n}
+E^\alpha_{j-n,n}\right)
\sum_{\Box_{j,n}}{\rm Tr}\left(U_{\Box_{j,n}}\sigma^\alpha\right),
\end{formula}
where $\Box_{j,n}$ denotes plaquettes in the plane perpendicular to
direction $n$,
running counterclockwise in that plane, and beginning and ending at site $j$.
Our purpose is to study the average of $B^2(t)$ ({\it cf} (\ref{1.1})) over the
thermal ensemble of initial field configurations. Due to the presence of
first-class constraints (Gauss' laws) this point is nontrivial. The initial
configurations must satisfy the constraints and be thermally distributed in the
reduced phase space of physical, gauge-invariant degrees of freedom. This goal
is achieved by using the constraint-respecting Langevin algorithm. The set
of Langevin equations reads ($\beta$ is the inverse temperature in lattice
units)
\begin{formula}{2.5}
\dot U_{j,k}=\{H, U_{j,k}\}+\left(\Gamma_{j,mn}(t)-\beta\{T_{j,mn},H\}\right)
\{T_{j,mn},U_{j,k}\}; \ \ \dot E^\alpha_l=\{H,E^\alpha_l\},
\end{formula}
where, for fixed $j$, $T_{j,mn}\equiv \sqrt{\gamma}\overline{E}^\alpha_{j,m}
\overline{E}^\alpha_{j,n}$ (assuming for definitness $m\leq n$),
and $\Gamma_{j,mn}(t)$ is a random variable with zero average and correlation
$$\langle\Gamma_{j,mn}(t)\Gamma_{j',m'n'}(t')\rangle=2\delta(t-t')\delta_{jj'}
\delta_{mn,m'n'}.$$
We are free to choose the friction coefficient $\gamma>0$. This freedom can be
used to optimize the algorithm performance. Evolution generated by (\ref{2.5})
is not to be confused with real-time evolution according to canonical equations
of motion: the sole purpose of (\ref{2.5}) is to bring the system in question
to thermal equilibrium at temperature $1/\beta$.
The set of equations (\ref{2.5}) is integrated numerically using a
modified leapfrog
algorithm, which obeys the Gauss' laws as accurately as a computer arithmetic
allows.
For a more detailed description of the algorithm we refer to \cite{alex}.
Results of simulation must eventually be converted from lattice units to
the physical ones in the continuum. To do so it is convenient to parametrize
the link matrices as $U_l=\exp(i a\sigma^\alpha A^\alpha_l/2)$ where $a$ is the
lattice spacing and $A^\alpha_l$ is the gauge potential. It follows from the
equation of motion for $U_l$ that $E^\alpha_l=i{\rm Tr}(\sigma^\alpha
U^\dagger_l\dot U_l)/2$. Therefore for small $a$ (\ref{2.1}) takes form
\begin{formula}{2.6}
H\approx {a^2\over 8}\sum_{j,n}\left(\dot A^\alpha_{j,n}\dot A^\alpha_{j,n}
+a^2 B^\alpha_{j,n}B^\alpha_{j,n}\right),
\end{formula}
where $B^\alpha_{j,n}$ is a chromo-magnetic field. Comparing (\ref{2.6}) to the
standard continuum Hamiltonian
$$ H_C={1\over{2 g^2}}\int d^3x \left(\dot A^\alpha_i\dot A^\alpha_i +
B^\alpha_i B^\alpha_i\right)$$
we find the correspondence between the time and the temperature expressed in
lattice ($L$) and continuum ($C$) units:
\begin{formula}{2.7}
t_C=at_L; \ \ T_C={4\over{g^2 a}} T_L. \end{formula}
Taking into account that $N_{\rm cs}$ itself is a dimensionless quantity,
\rf{1.1y} on the lattice takes form $\Gamma_L=\kappa (T_L/\pi)^4$, where the
lattice rate $\Gamma_L$ is per unit cell.
\section{Results}
We performed series of simulations using methods described in the previous
section for three values of the inverse lattice temperature $\beta=10,12,14$
on lattices of equal size $L_L$ and periodic boundary conditions in all
spatial directions. The value of $L_L$ ranged between 8 and 32. Between 20
and 60 statistically independent initial configurations were produced for
every $\beta,L_L$ pair (60 for the largest volumes, used for the estimate of
$\kappa$). At low temperatures considered the system is only weakly nonlinear,
and we expect an average energy per physical degree of freedom to be close to
$1/2\beta$. This is indeed what we observe: the average chromo-electric
energy per site was found to be $3/\beta$ with a three-digit accuracy.
Our measurement results for the topological charge density are also close to
the perturbative estimate \rf{5.3}.
Every independent initial configuration was let evolve in real time for
5000 time units. We used the integration time step of 0.05 for the classical
equations of motion. From our experience in earlier work \cite{jan} we know
that
reducing the time step beyond this value has negligible impact on the
diffusion of Chern-Simons number.
During the real-time evolution the value of $B(t)$, determined by
numerically integrating the lattice topological charge per unit time,
was recorded with intervals between 0.5 and 1 time unit. In the same way,
we recorded the time history of $\eta(t)$. Thus our results for
$\langle B^2(t)\rangle_T$, discussed in the following, represent
averaging both over the canonical ensemble of initial configurations and within
every individual real-time trajectory. The same averaging method was used
in earlier work on a low-dimensional model \cite{alex1}.
A typical plot of $\langle B^2(t)\rangle_T$ as a function of the lag $t$
is shown in Figures \ref{ill} and \ref{ill1}. As expected, for very short lags
($t\leq 1$)
$\langle B^2(t)\rangle_T$ is mostly accounted for by the perturbative result
obtained from \rf{5.2}. For larger lags a random walk
of $N_{\rm cs}$ sets in, and $\langle B^2(t)\rangle_T$ rapidly approaches a
linear asymptote. Thus, for large enough $t$, $\langle B^2(t)\rangle_T$ can
be fitted to the form (\ref{1.3e}). The range of $t$ suitable for this
linear fit is bounded from below by the initial transient behavior, and from
above by the growing statistical errors in our determination of
$\langle B^2(t)\rangle_T$. With the temperatures and lattice sizes used, we
found the optimal range to be approximately $40<t<200$. As Figures \ref{ill}
and \ref{ill1} indicate, at these lags
$\langle B^2(t)\rangle_T$ is dominated by the nonperturbative effects.
The central values of
$\langle B^2(t)\rangle_T$ for larger $t$ are consistent with linear
dependence on $t$.
\begin{figure}
\setlength{\unitlength}{0.240900pt}
\ifx\plotpoint\undefined\newsavebox{\plotpoint}\fi
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\begin{picture}(1800,1080)(0,0)
\font\gnuplot=cmr10 at 10pt
\gnuplot
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,113){\makebox(0,0)[r]{0}}
\put(1716.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,231.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,231){\makebox(0,0)[r]{0.5}}
\put(1716.0,231.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,349.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,349){\makebox(0,0)[r]{1}}
\put(1716.0,349.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,467.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,467){\makebox(0,0)[r]{1.5}}
\put(1716.0,467.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,585.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,585){\makebox(0,0)[r]{2}}
\put(1716.0,585.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,703.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,703){\makebox(0,0)[r]{2.5}}
\put(1716.0,703.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,821.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,821){\makebox(0,0)[r]{3}}
\put(1716.0,821.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,939.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,939){\makebox(0,0)[r]{3.5}}
\put(1716.0,939.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,1057){\makebox(0,0)[r]{4}}
\put(1716.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220,68){\makebox(0,0){0}}
\put(220.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(599.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(599,68){\makebox(0,0){50}}
\put(599.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978,68){\makebox(0,0){100}}
\put(978.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1357.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1357,68){\makebox(0,0){150}}
\put(1357.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736,68){\makebox(0,0){200}}
\put(1736.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,1057.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(45,585){\makebox(0,0){$\langle B^2(t)\rangle$}}
\put(978,23){\makebox(0,0){{\large t}}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220,113){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(258,178){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(296,204){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(334,227){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(372,248){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(410,269){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(447,289){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(485,310){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(523,330){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(561,351){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(599,372){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(637,392){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(675,412){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(713,431){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(751,451){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(789,472){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(826,492){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(864,512){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(902,533){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(940,553){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(978,573){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1016,594){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1054,616){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1092,637){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1130,659){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1168,680){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1205,702){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1243,723){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1281,744){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1319,766){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1357,788){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1395,809){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1433,830){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1471,851){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1509,873){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1547,894){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1584,915){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1622,936){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1660,956){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1698,978){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1736,1000){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(220,113){\usebox{\plotpoint}}
\put(210.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(210.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(258.0,178.0){\usebox{\plotpoint}}
\put(248.0,178.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(248.0,179.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(296.0,204.0){\usebox{\plotpoint}}
\put(286.0,204.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(286.0,205.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(334.0,226.0){\rule[-0.200pt]{0.400pt}{0.482pt}}
\put(324.0,226.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(324.0,228.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(372.0,247.0){\rule[-0.200pt]{0.400pt}{0.482pt}}
\put(362.0,247.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(362.0,249.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(410.0,267.0){\rule[-0.200pt]{0.400pt}{0.723pt}}
\put(400.0,267.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(400.0,270.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(447.0,287.0){\rule[-0.200pt]{0.400pt}{0.964pt}}
\put(437.0,287.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(437.0,291.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(485.0,308.0){\rule[-0.200pt]{0.400pt}{0.964pt}}
\put(475.0,308.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(475.0,312.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(523.0,328.0){\rule[-0.200pt]{0.400pt}{1.204pt}}
\put(513.0,328.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(513.0,333.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(561.0,348.0){\rule[-0.200pt]{0.400pt}{1.686pt}}
\put(551.0,348.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(551.0,355.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(599.0,368.0){\rule[-0.200pt]{0.400pt}{1.686pt}}
\put(589.0,368.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(589.0,375.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(637.0,388.0){\rule[-0.200pt]{0.400pt}{1.927pt}}
\put(627.0,388.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(627.0,396.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(675.0,407.0){\rule[-0.200pt]{0.400pt}{2.409pt}}
\put(665.0,407.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(665.0,417.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(713.0,426.0){\rule[-0.200pt]{0.400pt}{2.650pt}}
\put(703.0,426.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(703.0,437.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(751.0,445.0){\rule[-0.200pt]{0.400pt}{2.891pt}}
\put(741.0,445.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(741.0,457.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(789.0,465.0){\rule[-0.200pt]{0.400pt}{3.132pt}}
\put(779.0,465.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(779.0,478.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(826.0,485.0){\rule[-0.200pt]{0.400pt}{3.373pt}}
\put(816.0,485.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(816.0,499.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(864.0,504.0){\rule[-0.200pt]{0.400pt}{3.854pt}}
\put(854.0,504.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(854.0,520.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(902.0,524.0){\rule[-0.200pt]{0.400pt}{4.095pt}}
\put(892.0,524.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(892.0,541.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(940.0,544.0){\rule[-0.200pt]{0.400pt}{4.336pt}}
\put(930.0,544.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(930.0,562.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(978.0,564.0){\rule[-0.200pt]{0.400pt}{4.577pt}}
\put(968.0,564.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(968.0,583.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1016.0,584.0){\rule[-0.200pt]{0.400pt}{5.059pt}}
\put(1006.0,584.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1006.0,605.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1054.0,604.0){\rule[-0.200pt]{0.400pt}{5.541pt}}
\put(1044.0,604.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1044.0,627.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1092.0,625.0){\rule[-0.200pt]{0.400pt}{6.022pt}}
\put(1082.0,625.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1082.0,650.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1130.0,646.0){\rule[-0.200pt]{0.400pt}{6.263pt}}
\put(1120.0,646.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1120.0,672.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1168.0,666.0){\rule[-0.200pt]{0.400pt}{6.986pt}}
\put(1158.0,666.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1158.0,695.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1205.0,686.0){\rule[-0.200pt]{0.400pt}{7.468pt}}
\put(1195.0,686.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1195.0,717.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1243.0,707.0){\rule[-0.200pt]{0.400pt}{7.709pt}}
\put(1233.0,707.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1233.0,739.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1281.0,727.0){\rule[-0.200pt]{0.400pt}{8.431pt}}
\put(1271.0,727.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1271.0,762.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1319.0,748.0){\rule[-0.200pt]{0.400pt}{8.913pt}}
\put(1309.0,748.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1309.0,785.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1357.0,768.0){\rule[-0.200pt]{0.400pt}{9.395pt}}
\put(1347.0,768.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1347.0,807.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1395.0,788.0){\rule[-0.200pt]{0.400pt}{10.118pt}}
\put(1385.0,788.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1385.0,830.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1433.0,808.0){\rule[-0.200pt]{0.400pt}{10.600pt}}
\put(1423.0,808.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1423.0,852.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1471.0,828.0){\rule[-0.200pt]{0.400pt}{11.081pt}}
\put(1461.0,828.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1461.0,874.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1509.0,848.0){\rule[-0.200pt]{0.400pt}{11.804pt}}
\put(1499.0,848.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1499.0,897.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1547.0,869.0){\rule[-0.200pt]{0.400pt}{12.045pt}}
\put(1537.0,869.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1537.0,919.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1584.0,888.0){\rule[-0.200pt]{0.400pt}{12.768pt}}
\put(1574.0,888.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1574.0,941.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1622.0,908.0){\rule[-0.200pt]{0.400pt}{13.249pt}}
\put(1612.0,908.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1612.0,963.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1660.0,928.0){\rule[-0.200pt]{0.400pt}{13.731pt}}
\put(1650.0,928.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1650.0,985.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1698.0,948.0){\rule[-0.200pt]{0.400pt}{14.454pt}}
\put(1688.0,948.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1688.0,1008.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1736.0,968.0){\rule[-0.200pt]{0.400pt}{15.177pt}}
\put(1726.0,968.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1726.0,1031.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\end{picture}
\caption{Diffusion of $N_{\rm cs}$: mean square of $B(t)$ as a function of
the lag $t$ for $\beta=12$, $L_L=32$.}
\label{ill}
\end{figure}
\begin{figure}
\setlength{\unitlength}{0.240900pt}
\ifx\plotpoint\undefined\newsavebox{\plotpoint}\fi
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\begin{picture}(1800,1080)(0,0)
\font\gnuplot=cmr10 at 10pt
\gnuplot
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,113){\makebox(0,0)[r]{0}}
\put(1716.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,248.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,248){\makebox(0,0)[r]{0.1}}
\put(1716.0,248.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,383.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,383){\makebox(0,0)[r]{0.2}}
\put(1716.0,383.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,518.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,518){\makebox(0,0)[r]{0.3}}
\put(1716.0,518.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,652.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,652){\makebox(0,0)[r]{0.4}}
\put(1716.0,652.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,787.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,787){\makebox(0,0)[r]{0.5}}
\put(1716.0,787.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,922.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,922){\makebox(0,0)[r]{0.6}}
\put(1716.0,922.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,1057){\makebox(0,0)[r]{0.7}}
\put(1716.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220,68){\makebox(0,0){0}}
\put(220.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(523.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(523,68){\makebox(0,0){5}}
\put(523.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(826.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(826,68){\makebox(0,0){10}}
\put(826.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1130.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1130,68){\makebox(0,0){15}}
\put(1130.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1433.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1433,68){\makebox(0,0){20}}
\put(1433.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736,68){\makebox(0,0){25}}
\put(1736.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,1057.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(45,585){\makebox(0,0){$\langle B^2(t)\rangle$}}
\put(978,23){\makebox(0,0){{\large t}}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220,113){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(281,346){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(341,372){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(402,415){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(463,450){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(523,485){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(584,518){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(644,549){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(705,578){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(766,606){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(826,632){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(887,659){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(948,685){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1008,711){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1069,735){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1130,760){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1190,784){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1251,809){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1312,833){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1372,857){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1433,881){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1493,904){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1554,929){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1615,951){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1675,975){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1736,999){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(220,113){\usebox{\plotpoint}}
\put(210.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(210.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(281.0,345.0){\rule[-0.200pt]{0.400pt}{0.482pt}}
\put(271.0,345.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(271.0,347.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(341.0,372.0){\usebox{\plotpoint}}
\put(331.0,372.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(331.0,373.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(402.0,414.0){\rule[-0.200pt]{0.400pt}{0.482pt}}
\put(392.0,414.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(392.0,416.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(463.0,449.0){\rule[-0.200pt]{0.400pt}{0.482pt}}
\put(453.0,449.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(453.0,451.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(523.0,484.0){\rule[-0.200pt]{0.400pt}{0.723pt}}
\put(513.0,484.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(513.0,487.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(584.0,516.0){\rule[-0.200pt]{0.400pt}{0.723pt}}
\put(574.0,516.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(574.0,519.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(644.0,547.0){\rule[-0.200pt]{0.400pt}{0.964pt}}
\put(634.0,547.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(634.0,551.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(705.0,576.0){\rule[-0.200pt]{0.400pt}{0.964pt}}
\put(695.0,576.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(695.0,580.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(766.0,604.0){\rule[-0.200pt]{0.400pt}{0.964pt}}
\put(756.0,604.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(756.0,608.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(826.0,630.0){\rule[-0.200pt]{0.400pt}{1.204pt}}
\put(816.0,630.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(816.0,635.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(887.0,656.0){\rule[-0.200pt]{0.400pt}{1.445pt}}
\put(877.0,656.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(877.0,662.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(948.0,682.0){\rule[-0.200pt]{0.400pt}{1.445pt}}
\put(938.0,682.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(938.0,688.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1008.0,708.0){\rule[-0.200pt]{0.400pt}{1.445pt}}
\put(998.0,708.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(998.0,714.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1069.0,731.0){\rule[-0.200pt]{0.400pt}{1.927pt}}
\put(1059.0,731.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1059.0,739.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1130.0,756.0){\rule[-0.200pt]{0.400pt}{1.927pt}}
\put(1120.0,756.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1120.0,764.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1190.0,779.0){\rule[-0.200pt]{0.400pt}{2.168pt}}
\put(1180.0,779.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1180.0,788.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1251.0,804.0){\rule[-0.200pt]{0.400pt}{2.168pt}}
\put(1241.0,804.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1241.0,813.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1312.0,828.0){\rule[-0.200pt]{0.400pt}{2.409pt}}
\put(1302.0,828.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1302.0,838.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1372.0,851.0){\rule[-0.200pt]{0.400pt}{2.650pt}}
\put(1362.0,851.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1362.0,862.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1433.0,875.0){\rule[-0.200pt]{0.400pt}{2.650pt}}
\put(1423.0,875.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1423.0,886.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1493.0,898.0){\rule[-0.200pt]{0.400pt}{2.891pt}}
\put(1483.0,898.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1483.0,910.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1554.0,922.0){\rule[-0.200pt]{0.400pt}{3.132pt}}
\put(1544.0,922.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1544.0,935.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1615.0,945.0){\rule[-0.200pt]{0.400pt}{3.132pt}}
\put(1605.0,945.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1605.0,958.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1675.0,968.0){\rule[-0.200pt]{0.400pt}{3.373pt}}
\put(1665.0,968.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1665.0,982.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1736.0,991.0){\rule[-0.200pt]{0.400pt}{3.613pt}}
\put(1726.0,991.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1726.0,1006.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\sbox{\plotpoint}{\rule[-0.400pt]{0.800pt}{0.800pt}}%
\put(220,113){\usebox{\plotpoint}}
\multiput(221.41,113.00)(0.508,1.201){23}{\rule{0.122pt}{2.067pt}}
\multiput(218.34,113.00)(15.000,30.711){2}{\rule{0.800pt}{1.033pt}}
\multiput(236.41,148.00)(0.508,2.794){23}{\rule{0.122pt}{4.467pt}}
\multiput(233.34,148.00)(15.000,70.729){2}{\rule{0.800pt}{2.233pt}}
\multiput(251.41,228.00)(0.508,2.687){23}{\rule{0.122pt}{4.307pt}}
\multiput(248.34,228.00)(15.000,68.061){2}{\rule{0.800pt}{2.153pt}}
\multiput(266.41,305.00)(0.507,1.220){25}{\rule{0.122pt}{2.100pt}}
\multiput(263.34,305.00)(16.000,33.641){2}{\rule{0.800pt}{1.050pt}}
\put(281,340.84){\rule{3.614pt}{0.800pt}}
\multiput(281.00,341.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\multiput(296.00,340.09)(0.493,-0.508){23}{\rule{1.000pt}{0.122pt}}
\multiput(296.00,340.34)(12.924,-15.000){2}{\rule{0.500pt}{0.800pt}}
\multiput(311.00,325.08)(0.863,-0.516){11}{\rule{1.533pt}{0.124pt}}
\multiput(311.00,325.34)(11.817,-9.000){2}{\rule{0.767pt}{0.800pt}}
\put(326,317.34){\rule{3.614pt}{0.800pt}}
\multiput(326.00,316.34)(7.500,2.000){2}{\rule{1.807pt}{0.800pt}}
\multiput(341.00,321.38)(2.104,0.560){3}{\rule{2.600pt}{0.135pt}}
\multiput(341.00,318.34)(9.604,5.000){2}{\rule{1.300pt}{0.800pt}}
\put(356,324.84){\rule{3.854pt}{0.800pt}}
\multiput(356.00,323.34)(8.000,3.000){2}{\rule{1.927pt}{0.800pt}}
\put(372,325.34){\rule{3.614pt}{0.800pt}}
\multiput(372.00,326.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(387,323.34){\rule{3.614pt}{0.800pt}}
\multiput(387.00,324.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(402,321.84){\rule{3.614pt}{0.800pt}}
\multiput(402.00,322.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(432,321.84){\rule{3.614pt}{0.800pt}}
\multiput(432.00,321.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(417.0,323.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(478,321.84){\rule{3.614pt}{0.800pt}}
\multiput(478.00,322.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(493,320.84){\rule{3.614pt}{0.800pt}}
\multiput(493.00,321.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(447.0,324.0){\rule[-0.400pt]{7.468pt}{0.800pt}}
\put(554,319.84){\rule{3.614pt}{0.800pt}}
\multiput(554.00,320.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(508.0,322.0){\rule[-0.400pt]{11.081pt}{0.800pt}}
\put(584,318.84){\rule{3.614pt}{0.800pt}}
\multiput(584.00,319.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(599,317.84){\rule{3.614pt}{0.800pt}}
\multiput(599.00,318.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(569.0,321.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(644,316.84){\rule{3.854pt}{0.800pt}}
\multiput(644.00,317.34)(8.000,-1.000){2}{\rule{1.927pt}{0.800pt}}
\put(614.0,319.0){\rule[-0.400pt]{7.227pt}{0.800pt}}
\put(675,315.84){\rule{3.614pt}{0.800pt}}
\multiput(675.00,316.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(690,314.84){\rule{3.614pt}{0.800pt}}
\multiput(690.00,315.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(660.0,318.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(720,313.84){\rule{3.614pt}{0.800pt}}
\multiput(720.00,314.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(705.0,316.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(751,312.84){\rule{3.614pt}{0.800pt}}
\multiput(751.00,313.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(766,311.84){\rule{3.614pt}{0.800pt}}
\multiput(766.00,312.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(735.0,315.0){\rule[-0.400pt]{3.854pt}{0.800pt}}
\put(796,310.84){\rule{3.614pt}{0.800pt}}
\multiput(796.00,311.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(811,309.84){\rule{3.614pt}{0.800pt}}
\multiput(811.00,310.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(781.0,313.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(842,308.84){\rule{3.614pt}{0.800pt}}
\multiput(842.00,309.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(857,307.84){\rule{3.614pt}{0.800pt}}
\multiput(857.00,308.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(872,306.84){\rule{3.614pt}{0.800pt}}
\multiput(872.00,307.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(887,305.84){\rule{3.614pt}{0.800pt}}
\multiput(887.00,306.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(826.0,311.0){\rule[-0.400pt]{3.854pt}{0.800pt}}
\put(917,304.84){\rule{3.854pt}{0.800pt}}
\multiput(917.00,305.34)(8.000,-1.000){2}{\rule{1.927pt}{0.800pt}}
\put(933,303.84){\rule{3.614pt}{0.800pt}}
\multiput(933.00,304.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(948,302.84){\rule{3.614pt}{0.800pt}}
\multiput(948.00,303.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(963,301.84){\rule{3.614pt}{0.800pt}}
\multiput(963.00,302.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(978,300.84){\rule{3.614pt}{0.800pt}}
\multiput(978.00,301.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(993,299.84){\rule{3.614pt}{0.800pt}}
\multiput(993.00,300.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1008,298.84){\rule{3.614pt}{0.800pt}}
\multiput(1008.00,299.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(902.0,307.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(1039,297.84){\rule{3.614pt}{0.800pt}}
\multiput(1039.00,298.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1054,296.84){\rule{3.614pt}{0.800pt}}
\multiput(1054.00,297.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1069,295.84){\rule{3.614pt}{0.800pt}}
\multiput(1069.00,296.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1084,294.84){\rule{3.614pt}{0.800pt}}
\multiput(1084.00,295.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1023.0,300.0){\rule[-0.400pt]{3.854pt}{0.800pt}}
\put(1130,294.84){\rule{3.614pt}{0.800pt}}
\multiput(1130.00,294.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1145,295.84){\rule{3.614pt}{0.800pt}}
\multiput(1145.00,295.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1160,296.84){\rule{3.614pt}{0.800pt}}
\multiput(1160.00,296.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1175,298.84){\rule{3.614pt}{0.800pt}}
\multiput(1175.00,297.34)(7.500,3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1190,301.84){\rule{3.614pt}{0.800pt}}
\multiput(1190.00,300.34)(7.500,3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1205,304.84){\rule{3.854pt}{0.800pt}}
\multiput(1205.00,303.34)(8.000,3.000){2}{\rule{1.927pt}{0.800pt}}
\put(1221,308.34){\rule{3.200pt}{0.800pt}}
\multiput(1221.00,306.34)(8.358,4.000){2}{\rule{1.600pt}{0.800pt}}
\put(1236,312.34){\rule{3.200pt}{0.800pt}}
\multiput(1236.00,310.34)(8.358,4.000){2}{\rule{1.600pt}{0.800pt}}
\put(1251,315.84){\rule{3.614pt}{0.800pt}}
\multiput(1251.00,314.34)(7.500,3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1266,318.34){\rule{3.614pt}{0.800pt}}
\multiput(1266.00,317.34)(7.500,2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1281,319.84){\rule{3.614pt}{0.800pt}}
\multiput(1281.00,319.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1296,319.84){\rule{3.854pt}{0.800pt}}
\multiput(1296.00,320.34)(8.000,-1.000){2}{\rule{1.927pt}{0.800pt}}
\put(1312,318.34){\rule{3.614pt}{0.800pt}}
\multiput(1312.00,319.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1327,315.34){\rule{3.200pt}{0.800pt}}
\multiput(1327.00,317.34)(8.358,-4.000){2}{\rule{1.600pt}{0.800pt}}
\put(1342,311.34){\rule{3.200pt}{0.800pt}}
\multiput(1342.00,313.34)(8.358,-4.000){2}{\rule{1.600pt}{0.800pt}}
\put(1357,307.34){\rule{3.200pt}{0.800pt}}
\multiput(1357.00,309.34)(8.358,-4.000){2}{\rule{1.600pt}{0.800pt}}
\put(1372,303.84){\rule{3.614pt}{0.800pt}}
\multiput(1372.00,305.34)(7.500,-3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1387,301.34){\rule{3.614pt}{0.800pt}}
\multiput(1387.00,302.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1099.0,296.0){\rule[-0.400pt]{7.468pt}{0.800pt}}
\put(1433,300.84){\rule{3.614pt}{0.800pt}}
\multiput(1433.00,300.34)(7.500,1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1402.0,302.0){\rule[-0.400pt]{7.468pt}{0.800pt}}
\put(1463,300.84){\rule{3.614pt}{0.800pt}}
\multiput(1463.00,301.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1478,299.34){\rule{3.614pt}{0.800pt}}
\multiput(1478.00,300.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1493,297.34){\rule{3.854pt}{0.800pt}}
\multiput(1493.00,298.34)(8.000,-2.000){2}{\rule{1.927pt}{0.800pt}}
\put(1509,295.34){\rule{3.614pt}{0.800pt}}
\multiput(1509.00,296.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1524,293.84){\rule{3.614pt}{0.800pt}}
\multiput(1524.00,294.34)(7.500,-1.000){2}{\rule{1.807pt}{0.800pt}}
\put(1448.0,303.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\put(1554,294.84){\rule{3.614pt}{0.800pt}}
\multiput(1554.00,293.34)(7.500,3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1569,298.34){\rule{3.200pt}{0.800pt}}
\multiput(1569.00,296.34)(8.358,4.000){2}{\rule{1.600pt}{0.800pt}}
\multiput(1584.00,303.39)(1.579,0.536){5}{\rule{2.333pt}{0.129pt}}
\multiput(1584.00,300.34)(11.157,6.000){2}{\rule{1.167pt}{0.800pt}}
\multiput(1600.00,309.39)(1.467,0.536){5}{\rule{2.200pt}{0.129pt}}
\multiput(1600.00,306.34)(10.434,6.000){2}{\rule{1.100pt}{0.800pt}}
\multiput(1615.00,315.40)(1.176,0.526){7}{\rule{1.914pt}{0.127pt}}
\multiput(1615.00,312.34)(11.027,7.000){2}{\rule{0.957pt}{0.800pt}}
\multiput(1630.00,322.39)(1.467,0.536){5}{\rule{2.200pt}{0.129pt}}
\multiput(1630.00,319.34)(10.434,6.000){2}{\rule{1.100pt}{0.800pt}}
\multiput(1645.00,328.38)(2.104,0.560){3}{\rule{2.600pt}{0.135pt}}
\multiput(1645.00,325.34)(9.604,5.000){2}{\rule{1.300pt}{0.800pt}}
\put(1660,331.84){\rule{3.614pt}{0.800pt}}
\multiput(1660.00,330.34)(7.500,3.000){2}{\rule{1.807pt}{0.800pt}}
\put(1675,333.84){\rule{3.854pt}{0.800pt}}
\multiput(1675.00,333.34)(8.000,1.000){2}{\rule{1.927pt}{0.800pt}}
\put(1691,333.34){\rule{3.614pt}{0.800pt}}
\multiput(1691.00,334.34)(7.500,-2.000){2}{\rule{1.807pt}{0.800pt}}
\put(1706,330.34){\rule{3.200pt}{0.800pt}}
\multiput(1706.00,332.34)(8.358,-4.000){2}{\rule{1.600pt}{0.800pt}}
\multiput(1721.00,328.08)(1.176,-0.526){7}{\rule{1.914pt}{0.127pt}}
\multiput(1721.00,328.34)(11.027,-7.000){2}{\rule{0.957pt}{0.800pt}}
\put(1539.0,295.0){\rule[-0.400pt]{3.613pt}{0.800pt}}
\end{picture}
\caption{An enhanced region of $0\leq t\leq 25$ of Figure 1. The perturbative
estimate of $\left\langle B^2(t)\right\rangle_T$ is shown by the solid curve.}
\label{ill1}
\end{figure}
We determined the coefficient $\kappa$ of \rf{1.1y} by performing a correlated
fit of $\langle B^2(t)\rangle_T$ to the form (\ref{1.3e}) for data points
within the optimal range. As Figure~\ref{fs} shows, the values
of $\kappa$ for different
combinations of $\beta$ and $L_L$ depend, to a good approximation, on a
dimensionless ratio $L_L/\beta$ (note, in particular, the equality of
$\kappa$ values for $\beta=12,L_L=12$ and for $\beta=14,L_L=14$).
We conclude that in the range of temperatures
considered $\kappa$ is nearly insensitive to the lattice spacing, {\it i.e.,}
close to a finite continuum limit. On the other hand, $\kappa$ exhibits strong
finite-size effects and grows rapidly for $L_L/\beta\leq 2$. This is in
agreement with our expectation that the diffusion of $N_{\rm cs}$ is dominated
by large-scale structures whose linear size is of order $\beta$. Once the
lattice is too small to accommodate objects of this size, the rate rapidly
decreases\footnote{That the sphaleron-like transitions are related
to large scale objects has been verified directly by ``freezing'' the
time evolution in the neighborhood of $N_{cs} =n+1/2$. One finds energy
profiles corresponding to extended object \cite{af},
in contrast to the situation when one is far from $N_{cs} =n+1/2$.
In addition one could check that level crossing occurred as expected
from continuum physics.}. For large $L_L/\beta>2$ $\kappa$ reaches a constant
value.
In fact, the length scale at which $\kappa$ saturates can be naturally interpreted
in terms of magnetic mass, $m_m=0.466 g^2 T$ in physical units \cite{karsch}:
$\kappa$ stops growing as soon as the lattice size exceeds twice the inverse
magnetic mass (the factor of 2 relating the two lengths is to be expected
with periodic boundary conditions). Our estimate,
based on $\beta=12,L_L=32$ is $\kappa=1.09\pm 0.04$. Estimates from other
large-volume measurements are consistent with this one within the error bars.
\begin{figure}
\setlength{\unitlength}{0.240900pt}
\ifx\plotpoint\undefined\newsavebox{\plotpoint}\fi
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\begin{picture}(1800,1080)(0,0)
\font\gnuplot=cmr10 at 10pt
\gnuplot
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\put(220.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,113){\makebox(0,0)[r]{0.1}}
\put(1716.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,199.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,199){\makebox(0,0)[r]{0.2}}
\put(1716.0,199.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,285.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,285){\makebox(0,0)[r]{0.3}}
\put(1716.0,285.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,370.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,370){\makebox(0,0)[r]{0.4}}
\put(1716.0,370.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,456.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,456){\makebox(0,0)[r]{0.5}}
\put(1716.0,456.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,542.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,542){\makebox(0,0)[r]{0.6}}
\put(1716.0,542.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,628.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,628){\makebox(0,0)[r]{0.7}}
\put(1716.0,628.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,714.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,714){\makebox(0,0)[r]{0.8}}
\put(1716.0,714.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,800.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,800){\makebox(0,0)[r]{0.9}}
\put(1716.0,800.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,885.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,885){\makebox(0,0)[r]{1}}
\put(1716.0,885.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,971.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,971){\makebox(0,0)[r]{1.1}}
\put(1716.0,971.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,1057){\makebox(0,0)[r]{1.2}}
\put(1716.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220,68){\makebox(0,0){0.5}}
\put(220.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(473.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(473,68){\makebox(0,0){1}}
\put(473.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(725.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(725,68){\makebox(0,0){1.5}}
\put(725.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978,68){\makebox(0,0){2}}
\put(978.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1231.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1231,68){\makebox(0,0){2.5}}
\put(1231.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1483.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1483,68){\makebox(0,0){3}}
\put(1483.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736,68){\makebox(0,0){3.5}}
\put(1736.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,1057.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(45,585){\makebox(0,0){$\kappa$}}
\put(978,23){\makebox(0,0){$L_L/\beta$}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(1584,936){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1382,964){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1029,984){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(675,744){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(574,563){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1584.0,903.0){\rule[-0.200pt]{0.400pt}{15.899pt}}
\put(1574.0,903.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1574.0,969.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1382.0,929.0){\rule[-0.200pt]{0.400pt}{16.622pt}}
\put(1372.0,929.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1372.0,998.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1029.0,949.0){\rule[-0.200pt]{0.400pt}{16.622pt}}
\put(1019.0,949.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1019.0,1018.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(675.0,717.0){\rule[-0.200pt]{0.400pt}{12.768pt}}
\put(665.0,717.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(665.0,770.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(574.0,545.0){\rule[-0.200pt]{0.400pt}{8.431pt}}
\put(564.0,545.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(564.0,580.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\sbox{\plotpoint}{\rule[-0.400pt]{0.800pt}{0.800pt}}%
\put(1315,959){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(1146,966){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(852,954){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(473,260){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(304,171){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(1315.0,929.0){\rule[-0.400pt]{0.800pt}{14.454pt}}
\put(1305.0,929.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1305.0,989.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1146.0,940.0){\rule[-0.400pt]{0.800pt}{12.286pt}}
\put(1136.0,940.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1136.0,991.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(852.0,926.0){\rule[-0.400pt]{0.800pt}{13.490pt}}
\put(842.0,926.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(842.0,982.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(473.0,249.0){\rule[-0.400pt]{0.800pt}{5.300pt}}
\put(463.0,249.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(463.0,271.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(304.0,166.0){\rule[-0.400pt]{0.800pt}{2.409pt}}
\put(294.0,166.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(294.0,176.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\sbox{\plotpoint}{\rule[-0.500pt]{1.000pt}{1.000pt}}%
\sbox{\plotpoint}{\rule[-0.600pt]{1.200pt}{1.200pt}}%
\put(1122,968){\makebox(0,0){$\triangle$}}
\put(834,913){\makebox(0,0){$\triangle$}}
\put(473,245){\makebox(0,0){$\triangle$}}
\put(1122.0,939.0){\rule[-0.600pt]{1.200pt}{14.213pt}}
\put(1112.0,939.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(1112.0,998.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(834.0,875.0){\rule[-0.600pt]{1.200pt}{18.549pt}}
\put(824.0,875.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(824.0,952.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(473.0,229.0){\rule[-0.600pt]{1.200pt}{7.950pt}}
\put(463.0,229.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(463.0,262.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\end{picture}
\caption{The rate prefactor $\kappa$ dependence on the dimensionless parameter
$L_L/\beta$. The data points are for $\beta=10$ (diamonds), 12 (squares), and
14 (triangles).}
\label{fs}
\end{figure}
Our results for $\kappa$ are in sharp contrast with those for the shear viscosity,
extracted from the diffusion rate of $\eta(t)$. In the chosen range of
temperatures $\eta(t)$ behaves as an underdamped random walk, and determination
of the corresponding diffusion constant requires knowledge of
$\langle\eta^2(t)\rangle_T$ at lags $t$ of the order of 1000 (as compared to
$t\leq 200$ for the Chern-Simons number), making it a difficult task.
As Figure~\ref{visc} shows, the corresponding diffusion rate $\Gamma_\eta$
shows no marked temperature or finite-size dependence. The absence of finite
size effects for $\Gamma_\eta$ indicates that, unlike $\kappa$, this quantity is
not predominantly sensitive to long-wavelength modes of the system.
Independence
of $\Gamma_\eta$ of the temperature means,
on dimensional grounds, that $\Gamma_\eta$, and hence the classical shear
viscosity,
diverges as $a^{-4}$, where $a$ is the lattice spacing. This cutoff
dependence of transport coefficients is found in a similar situation in
condensed-matter physics, for phonons at temperatures far above
the Debye temperature \cite{berman}.
\begin{figure}
\setlength{\unitlength}{0.240900pt}
\ifx\plotpoint\undefined\newsavebox{\plotpoint}\fi
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\begin{picture}(1800,1080)(0,0)
\font\gnuplot=cmr10 at 10pt
\gnuplot
\sbox{\plotpoint}{\rule[-0.200pt]{0.400pt}{0.400pt}}%
\put(220.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,113){\makebox(0,0)[r]{0.12}}
\put(1716.0,113.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,207.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,207){\makebox(0,0)[r]{0.13}}
\put(1716.0,207.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,302.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,302){\makebox(0,0)[r]{0.14}}
\put(1716.0,302.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,396.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,396){\makebox(0,0)[r]{0.15}}
\put(1716.0,396.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,491.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,491){\makebox(0,0)[r]{0.16}}
\put(1716.0,491.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,585.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,585){\makebox(0,0)[r]{0.17}}
\put(1716.0,585.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,679.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,679){\makebox(0,0)[r]{0.18}}
\put(1716.0,679.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,774.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,774){\makebox(0,0)[r]{0.19}}
\put(1716.0,774.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,868.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,868){\makebox(0,0)[r]{0.2}}
\put(1716.0,868.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,963.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,963){\makebox(0,0)[r]{0.21}}
\put(1716.0,963.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(198,1057){\makebox(0,0)[r]{0.22}}
\put(1716.0,1057.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220,68){\makebox(0,0){0.5}}
\put(220.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(473.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(473,68){\makebox(0,0){1}}
\put(473.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(725.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(725,68){\makebox(0,0){1.5}}
\put(725.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(978,68){\makebox(0,0){2}}
\put(978.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1231.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1231,68){\makebox(0,0){2.5}}
\put(1231.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1483.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1483,68){\makebox(0,0){3}}
\put(1483.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(1736,68){\makebox(0,0){3.5}}
\put(1736.0,1037.0){\rule[-0.200pt]{0.400pt}{4.818pt}}
\put(220.0,113.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(1736.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(220.0,1057.0){\rule[-0.200pt]{365.204pt}{0.400pt}}
\put(45,585){\makebox(0,0){$\Gamma_\eta$}}
\put(978,23){\makebox(0,0){$L_L/\beta$}}
\put(220.0,113.0){\rule[-0.200pt]{0.400pt}{227.410pt}}
\put(574,646){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(675,778){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1029,669){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1382,389){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(1584,721){\raisebox{-.8pt}{\makebox(0,0){$\Diamond$}}}
\put(574.0,535.0){\rule[-0.200pt]{0.400pt}{53.721pt}}
\put(564.0,535.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(564.0,758.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(675.0,675.0){\rule[-0.200pt]{0.400pt}{49.866pt}}
\put(665.0,675.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(665.0,882.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1029.0,568.0){\rule[-0.200pt]{0.400pt}{48.903pt}}
\put(1019.0,568.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1019.0,771.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1382.0,306.0){\rule[-0.200pt]{0.400pt}{39.989pt}}
\put(1372.0,306.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1372.0,472.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1584.0,576.0){\rule[-0.200pt]{0.400pt}{69.620pt}}
\put(1574.0,576.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\put(1574.0,865.0){\rule[-0.200pt]{4.818pt}{0.400pt}}
\sbox{\plotpoint}{\rule[-0.400pt]{0.800pt}{0.800pt}}%
\put(304,674){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(473,728){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(852,663){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(1146,912){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(1315,609){\raisebox{-.8pt}{\makebox(0,0){$\Box$}}}
\put(304.0,571.0){\rule[-0.400pt]{0.800pt}{49.625pt}}
\put(294.0,571.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(294.0,777.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(473.0,591.0){\rule[-0.400pt]{0.800pt}{65.766pt}}
\put(463.0,591.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(463.0,864.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(852.0,525.0){\rule[-0.400pt]{0.800pt}{66.729pt}}
\put(842.0,525.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(842.0,802.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1146.0,806.0){\rule[-0.400pt]{0.800pt}{50.830pt}}
\put(1136.0,806.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1136.0,1017.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1315.0,531.0){\rule[-0.400pt]{0.800pt}{37.821pt}}
\put(1305.0,531.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\put(1305.0,688.0){\rule[-0.400pt]{4.818pt}{0.800pt}}
\sbox{\plotpoint}{\rule[-0.500pt]{1.000pt}{1.000pt}}%
\sbox{\plotpoint}{\rule[-0.600pt]{1.200pt}{1.200pt}}%
\put(473,725){\makebox(0,0){$\triangle$}}
\put(834,304){\makebox(0,0){$\triangle$}}
\put(1122,597){\makebox(0,0){$\triangle$}}
\put(473.0,626.0){\rule[-0.600pt]{1.200pt}{47.939pt}}
\put(463.0,626.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(463.0,825.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(834.0,193.0){\rule[-0.600pt]{1.200pt}{53.721pt}}
\put(824.0,193.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(824.0,416.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(1122.0,490.0){\rule[-0.600pt]{1.200pt}{51.553pt}}
\put(1112.0,490.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\put(1112.0,704.0){\rule[-0.600pt]{4.818pt}{1.200pt}}
\end{picture}
\caption{Diffusion constant per unit volume $\Gamma_\eta$ of $\eta(t)$ plotted
against the dimensionless parameter $L_L/\beta$.
The data points are for $\beta=10$ (diamonds), 12 (squares), and
14 (triangles).}
\label{visc}
\end{figure}
\section{Discussion}
We have seen that $\kappa \approx 1$ in a classical $SU(2)$ gauge theory,
and that we consequently have an observable (the diffusion rate
between gauge equivalent non-Abelian vacua) which is
not plagued by the generic ultraviolet thermal
fluctuations of classical field theory. In addition we expect that
the result obtained describes the high temperature limit of the
classical field theory corresponding to the standard model.
While this result is interesting by itself, it is of even more
interest to understand to what extent it is a good approximation
to the full quantum theory. In the Introduction we gave
qualitative arguments in favor of reasonable agreement between the
classical result for the rate and the full quantum one, essentially reproducing
the considerations which originally led Grigoriev, Rubakov
and Shaposhnikov to suggest the classical real time method.
Unfortunately, not much is known at present about the size and functional form
of quantum corrections to dynamical, real-time quantities.
The best we can offer is an educated guess based on a much better understood
relation between the static properties of the quantum theory and of the
effective dimensinally reduced one \cite{bielefeld,fks}.
The full thermal path integral,
where the bosonic fields obey periodic boundary conditions in the
imaginary time direction,
reduces in the high temperature limit to a partition function of a
three-dimensional statistical mechanical problem.
This partition function inherits the $UV$ thermal cutoff
from the underlying quantum field theory. If a lattice regularization
is used in the quantum theory, the spatial part of this lattice will provide
the regularization of the thermal three-dimensional theory.
It is known that dimensional reduction is not completely equivalent to the
classical approximation.
The correspondence between the two is conveniently described in terms
of the field component $A_0 (x)$ which is the Lagrange multiplier
of the Gauss' law. The latter may be imposed with the help of a
$\delta$-function factor in the classical statistical weight. Alternatively,
one can introduce $A_0 (x)$ and integrate out the color electric fields. The
classical partition function then takes form
\begin{formula}{4.1}
Z = \int {\cal D} A_i {\cal D} A_0 \; e^{-S_{{\rm eff}} (A_i,A_0)},
\end{formula}
where
\begin{formula}{4.2}
S_{{\rm eff}} (A) =\frac{1}{4g^2 T}\int d^3 x\; \left[ F^a_{ij}F^a_{ij}
+ (D_i^{ab} A_0^b)^2\right].
\end{formula}
On the other hand, the dimensionally reduced partition function has the same
form as \rf{4.1}, but this time the effective action includes Higgs potential
terms for $A_0$ and reads
\begin{formula}{4.3}
S_{{\rm eff}} (A) = \frac{1}{4g^2(T) T}\int d^3 x \;\left[F^a_{ij}F^a_{ij}
+ (D_i^{ab} A_0^b)^2 +\mu(T) (A^a_0)^2 + c g(T)^2 ((A_0^a)^2)^2 +\cdots
\right],
\end{formula}
where the parameters $g(T)$ and $\mu(T)$ are calculable functions
of the temperature, $g(T)$ being the running, temperature
dependent coupling constant, unlike in the purely classical case.
The important point is that, despite the apparent inequivalence of the two
partition functions,
we do not expect any qualitative difference
between the dimensionally reduced theory given by \rf{4.3} and the classical
one given by \rf{4.2}, since they both are in the confined
phase. In particular, both theories should give a magnetic screening
mass of order $g^2 T$. It therefore seems likely that the classical theory
is able to generate the approximatively correct non-perturbative
behavior at the magnetic-mass scale, and that the replacement $\a_w \to\a_w
(T)$
in the classical expression \rf{1.1y} yields a reasonable transition rate
estimate in the full theory.
\section*{Acknowledgments}
We thank L. McLerran and M.E. Shaposhnikov for useful comments.
This work was supported by the Danish Research Council under contract
No. 9500713. The bulk of the simulations was perfomed on the Cray-C92 and
the SP2 supercomputers at UNI-C.
\section*{Appendix}
In this Appendix we briefly describe thermal properties of
Chern-Simons number in the noncompact classical U(1) theory whose lattice
Hamiltonian is
\begin{formula}{5.1}
H\equiv {1\over 2}\sum_{j,n}E^2_{j,n}
+{1\over 4}\sum_{j,m\neq n}\left(\theta_{j,n}+\theta_{j+n,m}-
\theta_{j+m,n}-\theta_{j,m}\right)^2,\end{formula}
where, in the usual lattice notation, $j$ are sites and $m,n$ are positive
directions on a cubic lattice, and $\theta$ are lattice gauge potentials,
whose conjugate momenta $E_{j,n}$ are the electric fields.
Note that upon substitution $U_{j,n}=\exp(i\sigma^\alpha\theta^\alpha_{j,n})$
and linearization \rf{2.1} reduces to \rf{5.1} summed over
color indices $\alpha$. Imposing the Coulomb gauge
$$\sum_n\left(\theta_{j,n}-\theta_{j-n,n}\right)=0,$$
we eliminate longitudinal degrees of freedom (the electric field is transversal
by the Gauss' law). In momentum representation \rf{5.1} becomes
$$H={1\over 2}\sum_p \left[{\cal E}^i_p {\cal E}^i_{-p}
+4\sum_m\sin^2\left({p_m\over 2}\right)a^i_p a^i_{-p}\right],$$
where the Cartesian components of the momentum are $p_m=2\pi l_m/L_L$ with
integer
$1-L_L/2\leq l_m\leq L_L/2$ on an $L_L^3$ lattice with periodic boundary
conditions,
$a^i_{-p}=(a^i_p)^*$ is a photon field with momentum $p$ and polarization $i$,
and ${\cal E}^i_p=({\cal E}^i_{-p})^*$ are electric fields in momentum space.
It is easy to see that a mode with momentum $p$ has an eigenfrequency
$$\omega_p=2\sqrt{\sum_m\sin^2\left({p_m\over 2}\right)}.$$
In the same notation the Chern-Simons number is
\begin{eqnarray}
N_{\rm cs}&=&{1\over{32\pi^2}}\epsilon_{lmn}\sum_j\left(\theta_{j-n,n}+
\theta_{j,n}
\right)\left(\theta_{j-l+m,l}+\theta_{j+m,l}-\theta_{j-m,l}-\theta_{j-l-m,l}
\right)\nonumber\\
&=&{1\over{4\pi^2i}}\epsilon_{lmn}\sum_p {\rm e}^{i(p_l-p_n)/2}
\cos\left({p_l\over 2}\right)\cos\left({p_n\over 2}\right)
\sin(p_m)e^i(p)_n e^j(-p)_l a_p^i a_{-p}^j,\label{nlat}\end{eqnarray}
where the transverse polarization vectors $e^i(p)$ on the lattice have the
properties
$$e^i(-p)=\left(e^{i}(p)\right)^*; \ \ e^i(p)_n e^j(-p)_n=\delta^{ij};
\ \ e^i(p)_l e^i(-p)_n=\delta_{ln}-{\rm e}^{{i\over 2}(p_l-p_n)}
{{\sin\left({p_l\over 2}\right)\sin\left({p_n\over 2}\right)}
\over{\sum_r\sin^2\left({p_r\over 2}\right)}}.$$
Using these properties, and performing a straightforward Gaussian integration
with the thermal weight $\exp(-\beta H)$, one obtains the autocorrelation
function
\begin{formula}{5.2}
\left\langle N_{\rm cs}(t)N_{\rm cs}(0)\right\rangle_T=
{1\over{4\pi^4\beta^2}}\sum_p{{\cos^2\left(\omega_pt\right)}\over{\omega^2_p}}
\prod_l\cos^2\left({p_l\over 2}\right). \end{formula}
This expression can be explicitly evaluated for any finite lattice. The thermal
width $\nu$ of Abelian $N_{\rm cs}$ is given by \rf{5.2} at $t=0$. In the
limit $L_L\to\infty$ the sum in \rf{5.2} may be replaced by an integral, and
$\nu$ becomes an extensive quantity: $\nu\approx 6.930\times
10^{-4}\beta^{-2}L_L^3$. We used this estimate, together with the conversion
rule \rf{2.7}, to obtain \rf{1.3dd}. Note that the long-time behavior of
\rf{5.2} depends on the order in which the thermodynamic $L_L\to\infty$
and the $t\to\infty$ limits are taken: if the latter is taken first (which is
the case of practical interest here) on a finite-size lattice, \rf{5.2} will
at long times fluctuate around $\nu$ and not approach zero,
as can be verified by taking time average of \rf{5.2} over a long time
interval.
Another useful Abelian quantity is the thermal width of $\partial_t N_{\rm
cs}$.
Taking time derivative of (\ref{nlat}) and averaging over the thermal ensemble
yields
\begin{formula}{5.3}
\left\langle \dot N^2_{\rm cs}\right\rangle_T={1\over{2\pi^4\beta^2}}\sum_p
\prod_l\cos^2\left({p_l\over 2}\right) \to {L_L^3\over{16\pi^4\beta^2}},\end{formula}
where in the last step the large-volume limit was taken.
|
\section{#1}\indent}
\renewcommand{\theequation}{\thesection.\arabic{equation}}
\textwidth 159mm
\textheight 220mm
\begin{document}
\topmargin 0pt
\oddsidemargin 5mm
\def\bbox{{\,\lower0.9pt\vbox{\hrule \hbox{\vrule height 0.2 cm
\hskip 0.2 cm \vrule height 0.2 cm}\hrule}\,}}
\newcommand{\AP}[1]{Ann.\ Phys.\ {\bf #1}}
\newcommand{\NP}[1]{Nucl.\ Phys.\ {\bf #1}}
\newcommand{\PL}[1]{Phys.\ Lett.\ {\bf #1}}
\newcommand{\CMP}[1]{Comm.\ Math.\ Phys.\ {\bf #1}}
\newcommand{\PR}[1]{Phys.\ Rev.\ {\bf #1}}
\newcommand{\PRL}[1]{Phys.\ Rev.\ Lett.\ {\bf #1}}
\newcommand{\PTP}[1]{Prog.\ Theor.\ Phys.\ {\bf #1}}
\newcommand{\PTPS}[1]{Prog.\ Theor.\ Phys.\ Suppl.\ {\bf #1}}
\newcommand{\MPL}[1]{Mod.\ Phys.\ Lett.\ {\bf #1}}
\newcommand{\IJMP}[1]{Int.\ Jour.\ Mod.\ Phys.\ {\bf #1}}
\newcommand{\CQG}[1]{Class.\ Quant.\ Grav.\ {\bf #1}}
\newcommand{\PRep}[1]{Phys.\ Rep.\ {\bf #1}}
\newcommand{\RMP}[1]{Rev.\ Mod.\ Phys.{\bf #1}}
\newcommand{\frac{1}{2}}{\frac{1}{2}}
\newcommand{\partial}{\partial}
\newcommand{\frac{dz}{2\pi i}}{\frac{dz}{2\pi i}}
\newcommand{\rangle}{\rangle}
\newcommand{\langle}{\langle}
\newcommand{\nonumber \\}{\nonumber \\}
\newcommand{\vs}[1]{\vspace{#1 mm}}
\def\alpha{\alpha}
\def\beta{\beta}
\def\gamma{\gamma}
\def\Gamma{\Gamma}
\def\delta{\delta}
\def\Delta{\Delta}
\def\epsilon{\epsilon}
\def\varepsilon{\varepsilon}
\def\zeta{\zeta}
\def\theta{\theta}
\def\vartheta{\vartheta}
\def\iota{\iota}
\def\rho{\rho}
\def\varrho{\varrho}
\def\kappa{\kappa}
\def\lambda{\lambda}
\def\Lambda{\Lambda}
\def\omega{\omega}
\def\Omega{\Omega}
\def\sigma{\sigma}
\def\Sigma{\Sigma}
\def\varphi{\varphi}
\def\av#1{\langle#1\rangle}
\def\partial{\partial}
\def\nabla{\nabla}
\def\hat g{\hat g}
\def\underline{\underline}
\def\overline{\overline}
\begin{titlepage}
\setcounter{page}{0}
\begin{flushright}
COLO-HEP-363 \\
hep-th/9508053 \\
August 1995
\end{flushright}
\vspace{5 mm}
\begin{center}
{\large Remarks on supersymmetric effective actions and supersymmetry
breaking}
\vspace{10 mm}
{\large S. P. de Alwis\footnote{e-mail: <EMAIL>}
}\\
{\em Department of Physics, Box 390,
University of Colorado, Boulder, CO 80309}\\
\vspace{5 mm}
\end{center}
\vspace{10 mm}
\centerline{{\bf{Abstract}}}
We discuss some issues related to the definition of different
effective actions,
in connection with the work on supersymmetric theories by Seiberg and
collaborators. We also comment on the possibility of extending this
work to
broken supersymmetric theories.
\end{titlepage}
\newpage
\renewcommand{\thefootnote}{\arabic{footnote}}
\setcounter{footnote}{0}
\setcounter{equation}{0}
\vs{5}
\sect{Introduction}
Recent work by Seiberg and collaborators \cite{seione} on exact
results in
supersymmetric (SUSY) field theories depends crucially on the
existence of the
so-called Wilson effective action. In contrast to the usual one
particle
irreducible (1PI) effective action, this action, since it has an
explicit
infra-red cutoff, is free from holomorphic anomalies\footnote{This
point seems
to have been highlighted first by Shifman and Vainshtein \cite{sv}} .
This
enables one to make effective use of the power of holomorphy to
constrain the
form of the superpotential that is generated by quantum effects, even
at the
non-perturbative level. As a consequence of this analysis many
remarkable
properties of these field theories (such as the existence of
electro-magnetic
duality in non-abelian theories) have been discovered. It is
therefore of some
importance to discuss the precise (non-perturbative) definition of
the Wilson
effective action in the context of these theories. In this paper we
first
discuss problems associated with defining such an action in
continuum gauge
theories. Next (ignoring the above problem) we investigate how such
actions
may be defined for composite fields (mesons, and baryons). In the
fourth
section we discuss questions related to singularities at the origin
of moduli
space. We conclude with an evaluation of the prospects for extending
these
methods to broken SUSY theories.
\sect{Defining a Wilson effective action}
The Wilson effective action is obtained by integrating out short
distance
degrees of freedom of a microscopic theory valid at some short
distance scale,
down to some low energy scale $\Lambda$. It is then an effective field
theory for
discussing physics below the scale $\Lambda$. A precise definition was
originally
given in the context of
lattice gauge theory \cite{wilson}\footnote{Recently an explicitly
construction
of this has been given \cite{tom}.}. Unfortunately lattice methods
cannot be
easily extended to supersymmetric theories and one would like to have
a
continuum formulation of the effective action. The first step towards
this was
taken by Polchinski \cite{pol} who derived a continuum form of the
Wilson
renormalization group equation for scalar $\phi^4$ field theory and
used it to
prove the renormalizability of the theory. The method depended on
explicitly
introducing a
momentum space cutoff in the kinetic term of the action, and this
enables one
to derive a differential equation for the Wilson action of the form,
\begin{equation}\label{rgeqn}
\Lambda{d\over d\Lambda}S_W (\phi,\Lambda ) = F(S_W,\Lambda)
\end{equation}
The important point about this equation is not so much that it
enables one to
give a proof of the perturbative renormalization of the theory that
is much
simpler than the usual one, but that it gives a non-perturbative
definition of
the theory. In other words regardless of the
fact that the derivation of \ref{rgeqn} depended on the existence of
a
perturbative formulation of the theory (in so far as one assumes a
separation
between a kinetic term and an interaction), the equation gives a
unique
non-perturbative definition of an action at any (``final") scale
$\Lambda$, given
the microscopic action at some (``initial") scale $\Lambda_0$. Thus if we
could
derive such an equation
for SUSY gauge theories we would have a satisfactory object for which
the non-
perturbative arguments of Seiberg and collaborators will apply.
Unfortunately matters are not so simple as soon as one goes beyond
scalar field
theory. The problem arises when one tries to define a regulator for
gauge
theory in order to derive the analog of (\ref{rgeqn}). The
introduction of
higher covariant derivatives by themselves does not help since they
do
not
regularize one loop terms. Thus one is forced to introduce an
additional
regulator (typically Pauli-Villars) to deal with those. Such a
procedure which
regularizes all diagrams (including those in which there are
divergent one-loop
subgraphs) has been given by Warr \cite{warr} and has been extended
to SUSY and
super gravity theories in \cite{kl}. However apart from the rather
baroque
nature of this scheme, it suffers from one serious problem with
regard to
deriving an equation of the form \ref{rgeqn}. The use of
Pauli-Villars
regulators requires the introduction of preregulators in order to
define each
diagram separately, before the cancellation between diagrams
involving physical
fields and P-V fields takes place. This is done in a fashion that
manifestly
breaks the gauge invariance. This preregulator must then be first
sent to
infinity. \footnote{It is claimed that at the end of the day one has
a theory
which satisfies the Slavnov-Taylor identities.} The problem is that
the
procedure, even if it yields
a well-defined gauge (or BRST) invariant perturbation series, cannot
be used to
derive the analog of \ref{rgeqn}, since it is essentially a diagram
by diagram
method that is inextricably linked to perturbation theory. It is
thus
important to devise an alternative regularization that enables one to
derive an
RG equation.
In the last few years there has been quite a bit of activity on this
problem
\cite{wet, bon, becchi, ell}. However none of these attempts seem to
be
completely satisfactory. In the first part of this work we will
derive (in
somewhat simpler fashion and using a different regularization method)
a flow
equation very similar to one derived in \cite{wet}, \cite{ell}.
This will
then enable us to elucidate the problems that are involved in
defining a
Wilsonian action.
A 1PI effective action which is invariant under background field
gauge
transformations may be defined by the equation \footnote{To keep the
notation
simple we will use the notation of scalar field theory. When relevant
we will
discuss the additional complications of gauge theory such as gauge
fixing and
ghosts. }
\begin{eqnarray}\label{1pi}
e^{-\Gamma (\phi_c)}&=&\int [d\phi ' ]_{\epsilon}e^{-I[\phi_c+\phi ' ]-\phi '.
{\delta\Gamma\over\delta\phi_c}}\nonumber \\
&=&\int [d\phi ' ]_{\epsilon}e^{-[I[\phi _c]+\phi '.K[\phi_c].\phi
'+I_1[\phi_c,\phi
'+J.\phi ' ]}|\nonumber \\
&=&e^{-I[\phi _c]}e^{-{1\over 2}tr\ln{K[\phi_c]}-I_1[\phi
_c,-{\delta\over\delta
J}]}e^{{1\over 4}J.K^{-1}.J}|.
\end{eqnarray}
$I_1$ is at least cubic in the field $\phi'$ and the bar at the end
of the
equation is an instruction to evaluate the expression with
\begin{equation}\label{j}
J=-{\delta\Gamma\over\delta\phi_c}+{\delta I\over\delta\phi_c}\equiv\bar J.
\end{equation}
The subscript $\epsilon$ on the measure is an instruction to cut off the
functional
integral at an ultraviolet cutoff $\Lambda_o^2=\epsilon^{-1}$ and the dot
product implies
integration over space time. In the case of a gauge theory we need to
add
(background covariant) gauge fixing terms and the corresponding FP
term. Since
every term in the functional integral is back ground gauge invariant
so is the
effective action. {\it In addition the BRST invariance of the
functional
integrand and measure implies that $\Gamma [\phi_c]$ is independent of
the gauge
fixing.}
In order to define a Wilson effective action $\Gamma_L[\phi_c]$ we have
to
introduce an infra-red cut off $L =\Lambda^{-2}$. A convenient way of
doing this
while preserving the background gauge invariance, is to modify the
background field propagator $K[\phi_c ]^{-1}$ using the Schwinger
proper time
representation, \begin{equation}\label{prop}
K^{-1}_L[\phi_c]=\int_0^L dt e^{-tK[\phi_c]},~~\ln K_L = \int_0^L
{dt\over t}
e^{-tK[\phi_c]}.
\end{equation}
The Wilson action $\Gamma_L$ is then defined by replacing $K\rightarrow
K_L$ in
(\ref{1pi}). In a gauge theory this would include the contribution
from the
gauge fixing term and so K would explicitly depend on the gauge
fixing
parameter $\xi$ and there would also be a ghost contribution. From
(\ref{1pi})
(with $K\rightarrow K_L$) and (\ref{prop}) we see that
\begin{equation}\label{}
\Gamma_{L\rightarrow\infty} =\Gamma,~~ \Gamma_{L\rightarrow 0} =I[\phi_c],
\end{equation}
(ignoring an infinite constant in the second equation).
To get a flow equation in this background field formalism we need
first to
define an object $W_L[\phi_c, J]$ from equation (\ref{1pi}) by
removing the
restriction $J=\bar J$ (\ref{j}) and of course replacing $K$ by
$K_L$. Then
differentiating this equation with respect to $\ln L$ and
replacing $\phi '\rightarrow -{\delta\over\delta J}$ we easily obtain
($\dot{}=L{d\over d L} $)
\begin{equation}\label{w}
\dot{W}_L[\phi_c,J]=tr \dot{K}_L.\left[ {\delta W_L\over\delta J}{\delta
W_L\over\delta
J}-{\delta\over \delta J}{\delta W\over\delta J} \right].
\end{equation}
Since $K_L$ is linearly divergent as $L\rightarrow 0$ the initial
condition has
to be imposed at $L=\epsilon <<|\phi_c |^{-2}$ and we need to keep the
$O(\epsilon )$ term
so that we have (from the second line
of equation (\ref{1pi}))
\begin{equation}\label{initial}
W_{\epsilon} =I[\phi_c]+{1\over 2}tr\ln K_{\epsilon}[\phi_c]-{1\over
4}J.K^{-1}_{\epsilon}[\phi_c].J
\end{equation}
The equation then gives us $W_L$. Now the Wilsonian action $\Gamma_L$ is
obtained
from this by
putting $J=\bar J$ but since $\bar J$ is defined in terms of $\Gamma_L$
(see
\ref{}) this only gives an implicit definition. To get an explicit
version we
need to Legendre transform with respect to $J$. Putting ${\delta
W_L\over\delta J}
=<\phi '>_J\equiv \phi_c' $ and
\begin{equation}\label{j}
\bar\Gamma [\phi_c,\phi '_c]=W_L[\phi_c ,J]-J.\phi_c'
\end{equation}
one immediately has from \ref{w} (using $
\dot{\bar\Gamma}
=\dot W,~{\delta^2W_L\over\delta J\delta J}=-\left[
{\delta\bar\Gamma_L\over\delta\phi_c'\phi_c'}
\right]^{-1} $) the flow equation,
\begin{equation}\label{evolution}
\dot{\bar\Gamma}_L[\phi_c,\phi_c']=tr \dot K_L[\phi_c]\left\{
\phi_c'\phi_c'+\left[
{\delta\bar\Gamma_L\over\delta\phi_c'\phi_c'} \right]^{-1} \right\},
\end{equation}
and the initial condition (\ref{initial}) becomes,
\begin{equation}\label{}
\Gamma_{\epsilon}(\phi_c,\phi_c') = I[\phi_c] +\frac{1}{2} tr\ln K_{\epsilon}[\phi_c]+
{1\over
4}\phi_c'.K_{\epsilon}.\phi_c'
\end{equation}
In the limit $L\rightarrow\infty$ we have $\phi_c'=<\phi '>_J=<\phi
>-<\phi_c>=0$ and $\bar \Gamma[\phi_c,0 ]=\Gamma[\phi_c]$ the 1PI effective
action.
Hence we may define the Wilson effective action as
\begin{equation}\label{wil}
\Gamma_L[\phi ]= \bar\Gamma_L[\phi_c,0]
\end{equation}
Thus the equation gives us a Wilson effective action at a low scale
$\Lambda
=L^{-1/2}$ given an ``initial" action at a high scale $\Lambda_0
=\epsilon^{-1/2}$.
For a gauge theory the above derivations go through essentially
unchanged
except for the fact
that K now involves terms from the gauge fixing and also from ghosts.
The
effective action obtained this way is background gauge invariant.
However the
introduction of the cutoff $L$ violates the BRST invariance of the
functional
integral.\footnote{I wish to thank Joe Polchinski for raising this
issue.} The
question is, given the background gauge invariance of the effective
action,
does this matter? Through its dependence on $K$ the right hand side
of the
evolution equation is explicitly dependent on $\xi$ the gauge fixing
parameter.
With generic initial data we would evolve into an effective action
that is
$\xi$ dependent. On the other hand in the limit $L\rightarrow\infty$
the
functional integral representation implies that BRST invariance is
restored and
the 1PI effective action is
independent of $\xi$. It is hard to understand how this comes about
from the
evolution equation, unless one fine tunes the initial conditions at
the
physical cutoff in a $\xi$ dependent manner . This seems to be
rather an
unsatisfactory state of affairs. Unfortunately we do not know of any
procedure
for resolving this. Nevertheless we will ignore this gauge fixing
dependence in
the rest of the paper.
\sect{Effective action for composite fields}
As constructed above the Wilson action is not suitable for a
description of the
physics below
the confinement scale in a theory such as QCD, where one expects the
effective
low energy fields to be mesons and baryons. We would need an action
that is
expressed in terms of these
gauge invariant degrees of freedom such as the chiral Lagrangian. So
one may
define (in a schematic though obvious notation), a generating
functional for
mesons in the following way\footnote{The extension to baryons can be
done in an
analogous manner.}.
\begin{eqnarray}\label{wilson1}
e^{iW[J]}&=&\int
e^{i\Gamma_{\epsilon}[Q,\tilde{Q},V]+iJ.Q\tilde{Q}}[dQd\tilde{Q}dV]\int
[dM]\delta (M-Q\tilde{Q})\nonumber \\
&=&\int [dM] e^{i\bar\Gamma_{\epsilon}+iJ.M}.
\end{eqnarray}
In the above $Q,\tilde Q$ are quark and anti-quark fields,
$V$ represents the gauge fields and $M$ is a meson field.
Also $G_{\epsilon}$ is the Wilson action valid below the scale
$\epsilon^{-1}$ (so
it should be identified with $\Lambda^2$ of the previous section) and the
functional
integral is effectively cutoff above that scale. An effective action
for mesons
is then
given by,
\begin{equation}\label{}
e^{i\bar\Gamma_{\epsilon}[M]}=\int [dQd\tilde{Q}dV]e^{iS_{\epsilon}[Q,\tilde{Q},V]}\delta
(M-Q\tilde{Q}).
\end{equation}
It is important to note that $\bar\Gamma_{\epsilon}[M]$ inherits the
ultra-violet cutoff
so that in calculating meson processes from it one is only supposed
to
integrate up to the scale $\epsilon^{-1}$. Also it is clear that $\bar\Gamma$
has the same
global symmetry as $S$. It is of course expected that
$M$ will in general acquire a non-zero expectation value and the
action that
describes the degrees of freedom in that vacuum will only exhibit the
diagonal
$ SU(N_f)\times U(1)$ symmetry. It should also be pointed out here
that
$\bar\Gamma$ is a Wilsonian action and is not the same as the 2PI
effective action
\cite{cjt} which is defined by
\begin{equation}\label{}
\Gamma [M_c] =(W[J]-J.M_c)_{M_c={\delta W\over\delta J}}\simeq
\bar\Gamma_{\epsilon}[M_c]+O(\hbar ),
\end{equation}
the last step following from the saddle point evaluation of the
functional
integral \ref{wilson1}.
In the SUSY case the action $\bar\Gamma$ will satisfy in particular the
requirements of holomorphy, so the arguments of \cite{seione} as
well as
that of earlier work \cite{vy},\cite{ads},\cite{amati}, should
apply to this.
In particular this would mean not only that for $N_f>N_c $, $\det
M=0$ but also
that (for $N_f\ge N_c $ the classical constraints (for instance $\det
M =
B\tilde B$) are satisfied. Now while the former result is one that is
used in
\cite{ads} to argue that no superpotential is generated for this
case, the
latter is at variance with the statement in \cite{seitwo} that the
quantum
moduli space is modified from the classical space, so that for
instance in the
$N_f=N_c$
case the constraint is changed to $\det M- B\tilde{B}=\Lambda^{2N_c}$. In
order to
resolve this it seems that one has to redefine the Wilson action for
mesons
(and baryons). To this end let us first write, (again we explicitly
indicate
only the meson dependence)
\begin{equation}\label{J}
e^{-W[J]}=\int e^{-S_{L}(Q,\tilde{Q}, V)-J.(Q\tilde{Q})}
\end{equation}
The functional integral in the above is taken down to some low
(non-zero )
scale $L^{-\frac{1}{2}}$
and we define the expectation value in the presence of this cutoff
by,
\begin{equation}\label{exp}
{\delta W[J]\over\delta J}=<Q\tilde{Q}>_{L ,J}\equiv M_{L, c}
\end{equation}
This expression agrees with the actual vacuum expectation value of
the theory
when we let $L\rightarrow\infty$. Now we define the Wilson effective
action for
the mesons by the Legendre transform.
\begin{equation}\label{wilson2}
\Gamma_L(M)=(W[J]-J.M)
\end{equation}
where $J$ is replaced by solving \ref{exp} and we have the equation
of motion
${\delta\Gamma\over\delta M }=J$.\footnote{\it Henceforth we drop the subscripts
on $M$,
but this object should not be confused with the meson field
introduced earlier,
which was restricted to be $Q\tilde{Q}$.} This is a Wilsonian action
for
finite $L$. To get the true 1MI (2QI) action one needs to take the
limit
$L\rightarrow\infty$ and we have then a functional of the true vacuum
expectation value\footnote{An interpretation of the effective action
of
\cite{vy} involving the composite field $S\simeq W^2$ where $W$ is
the gauge
superfield, as a 2PI effective action has been given earlier in
\cite{fer}.}.
This will suffer from infra-red problems, but the
Wilsonian
version defined above has an explicit infra-red cutoff and is
therefore
expected to satisfy all the holomorphicity restrictions. The
arguments of
Seiberg et al will therefore apply to this object. Furthermore we
see from \ref{exp} that quantum effects will prevent us from
concluding that
$\det M=0$ for $3N_c>N_f>N_c$ and hence we may write down a
superpotential
\begin{equation}\label{w}
W={(\det M)^{1\over N_f-N_c}\over \Lambda_s^{3Nc-N_f\over
N_f-N_c}}f(L\Lambda_s^2)
\end{equation}
even in this case ($\Lambda_s$ is the SQCD scale). It should be noted that
this
interpretation of $M$ would appear to resolve the
apparent conflict between \cite{ads} and \cite{vy}.
As pointed out in \cite{seitwo} however there are other problems with
such an
effective action. One is that this superpotential is singular at the
origin\footnote{The superpotential for $N_f<N_c$ is also singular at
the origin
but in that case the potential is minimized away from the origin - at
infinity
in fact, whereas in the present case it is singular - though finite,
at the
point where its minimum lies.}. The other point is that the masssless
degrees
of freedom at the origin of moduli space do not match the quarks and
gluons
of the original formulation. Equation (\ref{w}) should not therefore
be
considered as an alternative to the discussion of Seiberg for $N_f\ge
N_c$,
nevertheless it is amusing to consider some of its consequences,
especially
since it can be written down once one adopts (3.4) - (3.6) in
contrast to
(3.2) as the definition of the effective action. We will have more to
say about
the
origin of
moduli space later, but for the time being we
will consider the theory with mass terms for mesons (and baryons) so
that the
minimum is away from the origin. In this case one can still use the
above (and
its generalizations including baryons) to caculate expressions for
the vacuum
expectation values. As pointed out in \cite{seione} one can consider
such
terms (for instance a term $m^i_{\hat i}M_i^{\hat i}$) as arising
from
a
coupling to a background field (or external source) and then the
modification
of the field $M=<Q\tilde{Q}>_{L}$ in the presence of this source is
given by
the (analog of) the standard equation for the effective action $\Gamma$
namely
${\delta\Gamma\over\delta\phi_c}=J$ . Thus the new expectation value is
determined by,
\begin{equation}\label{}
{\delta W\over\delta M^{\tilde{i}}_i} = m_{\tilde{i}}^i
\end{equation}
with $W$ given by (\ref{w}). This yields (upon taking the limit
$L\rightarrow\infty$ since in the presence of a non-singular mass
matrix $[m]$
there are no infra red singularities ) the result of Amati et al
\cite{amati}
that was used in \cite{seitwo}.
\begin{equation}\label{Mexp}
M^{\tilde i}_{c~i} = <Q_i\tilde{Q}^{\tilde i}>=\Lambda^{3Nc-N_f\over
N_c-N_f}(\det
[m])^{1\over N_c}[m^{-1}]^{\tilde i}_i
\end{equation}
\footnote{In the above we have put ${f(\infty)\over
Nc-Nf}^{N_c-N_f\over N_c}$
equal to unity. This amounts to a convention for the definition of
$\Lambda$.} As
observed by Seiberg \cite{seitwo} taking $m\rightarrow 0$ in
different
directions one can generate any value for the meson expectation
values of the
massless theory.
Note that even though (\ref{w}) is not defined for $N_f=N_c$ one can
take this
limit in (\ref{Mexp}) and we get $M=\Lambda^2(\det [m])^{1\over
N_c}[m^{-1}]$ so
that $\det M =\Lambda^{2N_c} $
When $N_f>N_c$ we can also define
$B_{L,c}^{f_{N_c+1}...f_{N_f}}=
\epsilon^{f_1...f_{N_f}}<Q_{f_1}...Q_{f_{N_c}}>_L$
where the RHS is defined as in (\ref{J}, \ref{exp}). Then one can for
instance
generate terms of the form $W\sim\Lambda^{\#}B_c^{N_f}\tilde{B}_c^{N_f}$
and by
adding source terms of the form $bB$ we can generate non-zero
expectation
values for $B$. The $N_f=N_c$ case is again a limiting one and as
with the
mesons we argue that even though the superpotential at this point is
not
defined one can still first compute the expectation values
(in the presence of the sources) and then take the limit
$N_f\rightarrow N_c$.
The cases $N_f=N_c$ and $N_f=N_c+1$ were discussed extensively in
\cite{seitwo}. The arguments depend
crucially on the assumption that the constraint on the quantum
moduli space is
changed (in the $N_f=N_c$ case) from the classical one $\det
M-B\tilde{B}=0$,
to the relation $\det M-B\tilde{B}=\Lambda_s^{2N_c}$. As pointed out
earlier this
is consistent only with the interpretation of the action given in
(\ref{wilson2}). It should also be noted that
although with (\ref{wilson1}) we are supposed to integrate down from
the
infra-red cutoff to get the 1MI (and 1BI) actions, with
(\ref{wilson2}) one
simply takes the limit $L\rightarrow\infty$. In other words with the
latter,
one is not supposed to compute loops to
get the 1PI (or 2PI etc.) actions whereas with the former ( the
chiral
lagrangian for QCD for instance), one computes loops.
The most interesting results of the analysis of Seiberg and
collaborators
relate to the
discussion of the theory at the origin of moduli space. This is the
point at
which superpotential is singular and it is argued that the
description in
terms of $M, B$ becomes invalid for $N_f>N_c+1$. Below we will
discuss some
questions related to the origin of moduli space.
\sect{The origin of moduli space}
Firstly it should be pointed out that since
$M=<Q\tilde{Q}>=\left[{\delta\Gamma_{1PI}\over \delta Q\delta\tilde{Q}}\right]^{-1}$
the point
$M=0$ is a place where the quark inverse propagator blows up.
\footnote{It
should be noted that we are always working with a ultra violet cutoff
so that
products of fields at the same point are well defined.} It is thus
possible
(as already observed in \cite{seitwo}) that this point is infinitely
far away
in moduli space. Let us examine these issues more closely.
To begin the discussion let us consider the abelian Higgs model. The
model has
an Abelian gauge field $A_{\mu}$ and a complex scalar field $\phi$
with the
invariance under the $U(1)$ gauge transformations $A_{\mu}\rightarrow
A_{\mu}+\partial_{\mu}\chi $ and $\phi\rightarrow e^{ig\chi}\phi$. The
degrees of
freedom (DOF) in the model are counted as follows: $A_0$ is a
Lagrange
parameter ($\Pi_0=0$) leaving $2\times (3+2)$ phase space dimensions
for the
three spatial gauge fields and two scalar fields. The Gauss law
constraint
${\bf \nabla . E}=j(\phi,\Pi_{\phi})$ and the gauge freedom ${\bf
A\rightarrow
A+\nabla }\chi $ then reduce the
dimension of phase space by two leaving four as the number of DOF.
This may be
interpreted as two for the (massless) gauge boson and two for the
complex
scalar. Now in this theory it is usually argued that when the
potential for
$\phi$ has a non-trivial minimum
($ V'(\phi_0\ne 0)=0$), the gauge symmetry is spontaneously broken;
the resulting Goldstone mode is ``eaten" by the gauge field which
thus
acquires a mass and hence has three degrees of freedom with one more
coming
from real part of the Higgs field. This at least is the classsical
argument.
Quantum mechanically the argument is somewhat more subtle. The
problem is that
in a quantum gauge theory one can never really ``break" the gauge
invariance.
The reason is that (at least for compact gauge groups) Elitzur's
theorem
\cite{elit} tells us that the
vacuum expectation value of any local order parameter (which does not
have a
gauge singlet component so that its integral over the gauge group
vanishes)
such as $<\phi>$ is zero.\footnote{Strictly speaking this theorem
has been
proved only for a Euclidean lattice gauge theory. In a Minkowski
formulation it
would follow from the Gauss law constraint provided that a
non-perturbative
gauge invariant regularization exists. } Thus if we construct a
gauge
invariant (1PI or Wilson)
effective action $\Gamma [A_c,\phi_c]$ the solution of
${\delta\Gamma[A_c,\phi_c]\over
\delta\phi_c }=0$ is $<\phi >=0$ which seems to indicate that there is
no
spontaneous symmetry breaking. However it is possible to reformulate
the theory
in terms of gauge invariant variables\footnote{The fact that Higgs
theories can
be expressed in terms of physical gauge invariant variables is an old
story and
was discussed in connection with the absence of a phase transition
between
Higgs and confinement phases in theories with fundamental Higgs in
\cite{banks}. }. One makes the {\it field redefinitions}
\footnote{This
procedure is usually called gauge fixing to the unitary gauge.
However it is
more appropriate to call it a field redefinition. The discussion
later, of the
measure should clarify the difference.} $\phi =e^{ig\theta}\rho
{}~(\rho
>0),~~A_{\mu}=G_{\mu}+\partial_{\mu}\theta$.
Under the gauge transformation $\theta\rightarrow \theta+\epsilon$; but the fields
$\rho ,
{}~G$ are gauge invariant. Because of gauge invariance the $\theta$
dependence drops out and the classical Lagrangian becomes,
\begin{equation}\label{}
L=-{1\over 4}(\partial_{\mu}G_{\nu}-\partial_{\nu}G_{\mu})^2-{1\over
2}(\partial_{\mu}\rho
)^2-{1\over 2}g^2\rho^2G_{\mu}^2-V(\rho).
\end{equation}
Obviously when expressed in terms of the physical variables, there is
no gauge
invariance in the Lagrangian. It is instructive to count the degrees
of freedom
(DOF) in this formulation. Again $G_0$ is Lagrange multiplier and so
the phase
space variables are $(\rho ,\Pi_{\rho}),~({\bf G,\Pi =E)}$. These
variables are
unconstrained (the Gauss law which is now ${\bf\nabla .E}=g^2G_0\rho
$
involves the Lagrange multiplier) and hence the number of degrees of
freedom
are again four, but with one coming from the scalar and three from
the gauge
field. It should be stressed that this has nothing to do with the
shape of the
potential $V$. What does depend on the latter is the count of
linearized DOF.
Unless $V(\rho )$ has a non-trivial minimum this count will not give
the
correct number of DOF.
What happens in the quantum theory? Here we need to consider the
measure. Under
our change of variables we get $[dA][d\phi_R][d\phi_I] =[dG][d\theta
][\rho d\rho
]$. Since the action is independent of $\theta$ the first factor just
gives the
volume of the gauge group. The last factor exhibits the typical
singularity of
polar coordinates and is not in a suitable form for perturbation
theory. A
further change of variable $\rho\rightarrow\chi =\rho^2$ gives a
linear measure
but now the problem reappears in the kinetic term which becomes
${1\over
8\chi}(\partial\chi)^2$.The metric on ``moduli" space appears to be
singular though
the distance from any point to the origin is in this case is actually
finite
(of course here the singularity is just a coordinate singularity
because the
space is one dimensional). A perturbative evaluation of the
functional integral
requires the existence of a saddle point $\chi_0\ne 0$. Nevertheless
the
expression eqn. (\ref{1pi}) will remain well defined even if (when
the external
source is set to zero) $\chi_c=<\chi >=0$.
These arguments are easily generalized to non-Abelian Higgs theories.
Let us
first consider the case of a Higgs field ${\bf\phi}$ in the
fundamental
representation of the gauge group ${\cal G} = SU(N)$.
Under gauge transformation $\bf\phi\rightarrow g\phi$ and ${\bf
A\rightarrow
gAg^{-1}+gdg^{-1}}$ where ${g\bf }\epsilon {\cal G}$. Again we can redefine
the
fields by putting
$ {\bf \phi =\Omega\rho }$ (with $\rho =[\rho,0...0]^T$ and
${\bf\Omega}\epsilon\cal G$ with
$\rho$ a real and positive field) and $\bf A = \Omega (\partial + G)\Omega^{-1}$.
The gauge
transformations are $\Omega\rightarrow g\Omega ~\rho\rightarrow\rho,~
G\rightarrow G$,
and the Lagrangian becomes
\begin{equation}\label{}
L=-{1\over 4}tr F^{G~2}-{1\over 2}(D^G\rho )^2-V(\rho ).
\end{equation}
In these variables the gauge invariance has been reduced from
$SU(N)$ to
$SU(N-1)$. This fact is again independent of the nature of $V$. Let
us now look
at the measure. The original gauge invariant metric on field space is
$\int
d^Dx (\delta\phi^{\dag}\delta\phi +tr\delta A^2)$. Using
$\Omega^{\dag}\Omega = \Omega\O^{\dag}=1$ and the reality of $\rho$, we find that
this
becomes$\int d^Dx(\rho^{\dag}\delta\Omega^{\dag}\delta\Omega\rho +(\delta\rho )^2+(\delta
G^2))$. Hence
the measure becomes $\prod_{x} [d(G/H)(x)\rho(x)d\rho\prod_{\mu\alpha}\delta
B_{\mu}^a
$, where the first factor is a volume element on $ SU(N)/SU(N-1)$
which will
just give a factor of the volume of the coset in the functional
integral. Again
as in the Abelian case we have the quantum theory expressed entirely
in terms
of variables which display a smaller gauge invariance than the
original ones
but as in the earlier case there is no perturbative description
around
$\rho=0$.
By repeating the process when there are $M$ complex scalar fields in
the
fundamental representation we can express the quantum theory (i.e.
the
classical lagrangian as well as the measure) completely as a SU(N-M)
invariant
theory for $M<N-1$ and as a theory with no gauge invariance for $M\ge
N-1$.
Again this fact has nothing to do with the nature of the classical
potential.
Consider now a scalar Higgs field in the adjoint representation. We
may do the
field redefinition $\Phi=\Omega D\Omega^{-1}$ and $\bf A = \Omega (\partial +
G)\Omega^{-1}$, where
$\Omega\epsilon{\cal G}$ and $D=diag [\lambda_1,\ldots ,\lambda_N]$, ($\lambda_i$ being the
(real)
eigenvalue fields of the Hermitian matrix $\Phi$). As before $\Omega$
disappears
from the Lagrangian but now we are left with
a $U(1)^{N-1}$ symmetry (except on a set of measure zero in the
field space
where the symmetry is larger). The metric on field space may now be
written as
\begin{equation}\label{metadj}
\int d^Dx (tr\delta\Phi^2+tr\delta A^2)=\int d^Dx \{tr\delta D^2+2tr[\Omega^{-1}\delta\Omega
D
)D-D^2(\Omega^{-1}\delta\Omega) ]+tr\delta G^2\}
\end{equation}
{}From this we would again expect a measure of the form $\sim Vol
[SU(N)/U(1)^{N-1}]\prod_x D\delta D\delta G $, though it is not clear to us
how to
establish this precisely.
Again one expects that because of the singularity at $D=0$ that
perturbation
theory about this point is invalid. Nevertheless the complete
regulated (for example on
a lattice) functional integral expressed as a theory with reduced
gauge
invariance is perfectly well defined.
Let us now turn to SUSY theories. Consider first $N=1$ SQCD and for
simplicity
we will take the gauge group to be $SU(2)$ with two doublet chiral
super
fields, $\Phi,~\tilde{\Phi}$. The Lagrangian takes the form (in
standard
superfield notation) ,
\begin{equation}\label{sqcd}
L={1\over 4\pi}Im\int d^2\theta\tau tr W^{\alpha}W_{\alpha}+\int d^4\theta
\Phi^{\dag}e^V\Phi +\tilde{\Phi}e^-V^T\tilde{\Phi}+\int d^2\theta
W[\Phi,\tilde{\Phi}]+h.c..
\end{equation}
Under a gauge transformation, we have $\Phi\rightarrow e^{i\Lambda}\Phi,
e^V\rightarrow e^{-i\Lambda^{\dag}}e^{V}e^{i\Lambda}$ and the gauge invariant
measure is
obtained from the metric,
\begin{equation}\label{susymet}
||\delta V||^2+||\delta\Phi ||^2 +...=\int d^xd^4\theta tr [(e^{-V}\delta e^{V})^2
+\delta\Phi^{\dag}e^V\delta\Phi+...]
\end{equation}
(the ellipses are for the $\tilde{\Phi}$ terms). This is a highly
non-linear
metric and so is the associated metric. (In SUSY perturbation theory
the
measure may be replaced by a linear approximation to it). Let us now
try to do
the (analogs of the) field redefinitions that we
performed in non SUSY theories. Thus we put
$\Phi=e^{-i\omega}\underline\rho,
\tilde{\Phi}=e^{-i\omega}\tilde{\underline\rho}$ where the first factor
in these
expressions is a chiral superfield valued in the gauge group and
$\underline\rho =[\rho,0]^T $. Also we put
$e^U=e^{i\omega^{\dag}}e^Ve^{-i\omega}$. The
gauge transformation is now $e^{-i\omega}\rightarrow e^{-i\Lambda}e^{-i\omega}$
but $\omega$
disappears from the Lagrangian and we have a classical theory
expressed
entirely in terms of gauge invariant variables. Note that $U$ is no
longer in
Wess Zumino gauge, but (for each generator of the group) the 4
massless degrees
of freedom of the vector multiplet combine with the 4 DOF of the
chiral
multiplet $\omega$ to give the 8 DOF of a massive multiplet. This is just
the Higgs
effect in a SUSY theory and again as in the case of the non-SUSY
Higgs theory
discussed above the count of the DOF of the classical non-linear
evolution
equations have nothing to do with the nature of the potential.
Unfortunately in
the SUSY case it is not easy to show that the volume of the gauge
group factors
out. It may be recalled that in the non-supersymmetric case the
metric became
diagonal because we of the reality of the (gauge invariant) Higgs
field $\rho$.
In the SUSY case this is necessarily chiral and we cannot make the
same
argument. Nevertheless it is still reasonable to assume that the
qualitative
features of the previous discussion (for example the statement that
perturbation theory around
$\rho =0$ is singular) may still be expected to be valid. In any case
the
potential is given by
$V=G^{i,\bar j}{\partial W\over \partial\phi^i}{\partial \bar{W}\over
\partial\phi^{\bar{j}}}$
where $G$ is the metric on moduli space and it is invariant under
holomorphic
field redefinitions and so there are additional singularities (i.e.
apart from
the singularities of $W$), only if there is a true singularity of the
metric as
opposed to a coordinate singularity. Our point here is that the
singularity at
the origin coming from the measure appears to be of the latter kind
and hence
should not have any physical effect (in contrast to those of the
superpotential
discussed by Seiberg \cite{seione}).
Next we make a remark about $N=2$ SUSY theories. In these theories
there is a
$N=1$ chiral superfield in the adjoint representation, and hence as
in the
corresponding case of the non-SUSY theory discussed above, (modulo
the caveat
about measures in the previous case) the $SU(N)$ theory is equivalent
to a
$U(1)^{N-1}$ theory except that there is no perturbative description
around
$D=0$ (here the D is defined by the SUSY generalization of the non
SUSY
version given in the paragraph above (\ref{metadj}). As explained in
\cite{sw}
one needs in that case to go to the dual theory, which brings in the
issues of
monopole condensation and the dual Meissner effect. It appears to us
however
that this effect, while it has a gauge invariant formulation in these
$N=2$
theories is not easily taken over as an explanation of confinement in
QCD where
there is no adjoint scalar field to break the symmetry to $U(1)^2$.
The
attempts to gauge fix to the maximal Abelian gauge \cite{'thooft} for
instance
by introducing a term $G^2$ (where $G$ is the gluon field strength)
is
tantamount to introducing a factor $0\over 0$ into the path integral
(since the
integral over the gauge group of any non-trivial irreducible
representation of
the group is zero) and is thus bound to yield ambiguous results.
\sect{Supersymmetry breaking}
Finally let us make some remarks on supersymmetry breaking. We will
combine the
old observations of Giradello and Grisaru \cite{gg} on soft
breaking terms,
the recent work of Seiberg and Intrilligator \cite{seione}, and those
of this
paper, to investigate to what extent non-perturbative statements can
be made in
the presence of SUSY breaking.
Consider the SQCD action (\ref{sqcd}) together with the following
terms
coupling local gauge invariant composite fields to external
sources.
\begin{eqnarray}\label{break}
S_B &=& \int d^4 x \{
J_TQ^{\dag}e^VQ|_D+J_{\tilde{T}}\tilde{Q}^{\dag}e^{-V^T}\tilde{Q}|_D+
\nonumber \\
&+&(J_SW_{\alpha}W^{\alpha}+J_MQ\tilde{Q}+J_BQ^{N_c}+
J_{\tilde{B}}\tilde{Q}^{N_c}+h.c.)|_F\}
\end{eqnarray}
In the above $J_{T(\tilde{T})}$ are real super fields while the
other sources
are chiral superfields. Also for simplicity flavor symmetry
invariance is
preserved.
The computation of correlation functions of gauge invariant composite
fields in
the SUSY theory
are obtained by differentiating the generating functional $W[J]$,
defined in
(\ref{J}), with respect to the sources and then setting them equal to
zero. The
correlation functions of (softly) broken theory on the other hand
may be
obtained by the same generating functional by doing the functional
differentiations and then setting\footnote{In using the method of
reference
\cite{gg} we find it unnecessary to include a global SUSY breaking
(O'Raffeartaigh) sector as in \cite{evans}. This is just as well
since in
realistic models one expects the soft breaking to be generated from
the
breaking of local SUSY at the Planck scale.}
\begin{equation}\label{susybreak}
J_T=m^2_T\theta\t\bar\theta\bar\theta ,~J_S=m_g\theta\t ,~J_M=m_M\theta\t,~etc.
\end{equation}
The equations
\begin{eqnarray}\label{}
{\delta W\over \delta J_T}&=&<Q^{\dag}e^VQ>_J\equiv T,~{\delta W\over \delta
J_S}=<W_{\alpha}W^{\alpha}>_J\equiv S,\nonumber \\{\delta W\over \delta
J_M}&=&<Q\tilde{Q}>\equiv M,
{}~{\delta W\over \delta J_{B}}=<Q^{N_c}>\equiv B
\end{eqnarray}
(together with two more for anti-quarks and anti-baryons) define
effective
fields in terms of which an effective action for mesons and baryons,
and
glueballs can be defined. Indeed
if $W$ has been defined in the presence of an infra-red cutoff as in
(\ref{J})
then we can again define a Wilsonian effective action $\Gamma
[T.\tilde{T}, S, M,
B,\tilde{B}]$ as in (\ref{wilson2}) such that the effective fields (
in the
presence of the source superfields ) are solutions of the
equations,
\begin{equation}\label{}
{\delta \Gamma\over \delta T}=J_T,~{\delta \Gamma\over \delta S }=J_S, {\delta \Gamma\over \delta M}=J_M,
{}~{\delta
\Gamma\over \delta B}=J_B,~etc..
\end{equation}
The effective action in the presence of SUSY breaking can then be
written as,
\begin{equation}\label{}
\bar\Gamma
[T,...\tilde{B},J]=\Gamma[T,...\tilde{B}]+J_T.T|_D+...+(J_{\tilde{B}}.
\tilde{B}|_F+h.c.)
\end{equation}
with $J$ being replaced by the values given in (\ref{susybreak}).
When the
supersymmetry breaking parameters ($m_T,...m_{\tilde{B}}$) are set to
zero and
$T$ is replaced by using the equation of motion ${\delta\Gamma\over\delta T}=0 $
giving
$T=aM^{\dag}M+b B^{\dag}B+etc. $,
the
effective action should reduce to those derived in the literature.
However in
the presence of the D-term SUSY breaking involving the real
superfield $T$
very little can be said about the SUSY broken theory. This is
because in
solving for T one now has to use ${\delta\Gamma\over\delta T}=J_T $ and without
knowing
the dependence of $\Gamma$
on $T$ it is not possible to extract the dependence of the SUSY
broken action
on the symmetry breaking parameters $M_t,m_{\tilde{T}}$. Of course
one may
consider pure F type symmetry breaking and then one would be able to
calculate
precisely the modifications of the results of Seiberg and
collaborators.
However as is well known such breaking is not phenomenologically
acceptable
since for instance they predict squark (and slepton) masses which are
smaller
than those of quarks (and leptons). One can also assume that the
SUSY
breaking parameters are small and then the qualitative features of
the SUSY
limit are preserved \cite{aharony}. Unfortunately in the real world
the SUSY
breaking scale is much larger than the QCD scale. Of course in any
case,
since there is no chiral scalar superfield description of the proton
for
instance
\footnote{the color antisymmetrization will cause $uud$ to vanish
when each is
a scalar
super field.}
there is no smooth limit from the effective theory of the mesons and
baryons
defined by Seiberg \cite{seione} \footnote{For instance with
$N_f=N_c$ where
the number
of massless constituent degrees of freedom in the SUSY limit is
matched by the
set of $M$'s
and $B$'s defined in \cite{seione}. Perhaps this implies that the
proton
composite operator becomes massive.} to an effective description in
terms of
protons, neutrons, pions etc of strongly coupled QCD. A
description in
terms of the observed baryons and mesons, of the limit where the
superpartners decouple, which is smoothly obtained from the SUSY
limit, would
seem to require additional operators in the original SUSY theory.
This fact may
have important implications for understanding supersymmetry breaking.
Note added: While preparing this work for publication we came across
\cite{evans2} in which aspects of supersymmetry breaking are
discussed.
{\it Acknowledgements}
We are grateful to R. Leigh and N. Seiberg for very instructive
lectures on
exact results in SUSY field theories and for sharing their insight
into the
subject. We also wish to thank S. Balakrishnan, E. Brezin, M. Dine,
S.
Giddings, A. Hasenfratz, R. Kallosh, D. MacIntyre, K. Mahanthappa,
M.
Shifman, L. Thorlacius, and especially, T. DeGrand, J. Polchinski
and L.
Susskind, for discussions. I also wish to thank N. Evans,
S.D.H. Hsu, M. Schwetz, S.B. Selipsky, and F. Quevedo for comments
on the
manuscript.
This work is
partially supported by the
Department of Energy contract No. DE-FG02-91-ER-40672.
|
\section{Introduction}
Leading order (LO) QCD predicts photon interactions to have a two-component
nature. In direct photon processes the whole of the photon takes part in the
hard subprocess with a parton from the proton whereas in resolved photon
processes, the photon acts as a source of partons and one of these enters the
hard subprocess (see Figure \ref{fig:diresfig}). Both LO processes are
characterized by having two outgoing partons of large transverse energy.
Previous studies of dijet photoproduction at HERA have shown that both classes
of process are evident for the case of quasi-real photons (those of negligible
virtuality $P^2$) \cite{prevpap}.
The parton content of photons is neither well constrained theoretically nor
well known experimentally, particularly for photons with small but non-zero
virtualities. Various theoretical predictions exist for the behaviour of the
photon structure as a function of the photon virtuality \cite{theory}. The
general expectation is that the contribution to the dijet cross section of
resolved photon processes should decrease relative to the contribution from
direct photon processes ({\it i.e.} that the partonic content of the photon is
suppressed) as the virtuality of the photon increases.
\begin{figure}[hbt]
\begin{center}\mbox{\epsfig{file=glasdiag.eps,width= 12cm}}\end{center}
\caption{Diagrams showing a) direct and b) resolved photon processes. In both
cases the photon of virtuality $P^2$ carries a fraction $y$ of the positron
momentum.}
\label{fig:diresfig}
\end{figure}
\section{The Data}
The data used in this analysis were collected during the 1994 run when HERA
collided 27.5 GeV positrons with 820 GeV protons. The ZEUS detector is described
elsewhere\cite{standard}. A tungsten-silicon sampling calorimeter was installed
for the 1994 running period. This 'beampipe' calorimeter tagged positrons
scattered through small angles (17 - 35 mrad) and gave a sample of events with
photon virtualities in the range $0.1 < P^{2} < 0.55 $ GeV$^2$. A sample of
events with quasi-real photons ($P^2 < 0.02$ GeV$^2$ with a median of $10^{-5}$
GeV$^2$) was obtained by requiring that the scattered positron be detected in
the downstream luminosity calorimeter\cite{lumi}. Jets were found in the main
uranium-scintillator\cite{main} calorimeter using a cone algorithm \cite{cone}
in $\eta$ - $\phi$ space, where ${\phi}$ is azimuth and the pseudo rapidity
${\eta} = -{\ln}(\tan({\theta}/2))$, ${\theta}$ being defined with respect to
the proton direction. Only those events with two or more jets of transverse
energy $E_T^{jet} > 4$ GeV in the range $-1.125 < \eta^{jet} < 1.875 $ were
selected as dijet events.
To reduce contamination from beam gas and deep inelastic scattering events, a
cut of $ 0.15 < y_{JB}=\sum(E-p_{z})/2E_{e} < 0.70 $ was applied, where the sum
is over calorimeter cells with deposits of total and longitudinal energy $E$ and
$p_z$ respectively and where $E_e$ is the positron beam energy. In
photoproduction, $y$ is the fraction of the positron energy carried by the
photon. These cuts left a sample of 375 events with virtual photons and a sample
of 14181 events with quasi-real photons, corresponding to respective integrated
luminosities of 2.07 pb$^{-1}$ and 2.19 pb$^{-1}$.
\section{Results}
For each dijet event the fraction of the photon momentum manifest in the two
highest $E_T$ jets, $x_{\gamma}^{obs}$ was calculated. $x_{\gamma}^{obs}$ is
defined by
\[
x_{\gamma}^{obs} = \frac{ \sum_{jets}E_T^{jet}e^{-\eta^{jet}}}{2yE_e}
\]
We measure $E_T^{jet}$ and ${\eta}^{jet}$ using the raw calorimeter energies and
use the Jaquet-Blondel method, $y_{JB}$ above, to measure $y$. No corrections
are made for detector effects.
Uncorrected distributions of $x_{\gamma}^{obs}$ are shown in figure
\ref{fig:xgfig} a) for events with virtual photons and figure \ref{fig:xgfig} b)
for events with quasi-real photons. Events at high $x_{\gamma}^{obs}$ are
associated with direct photon processes while those at low $x_{\gamma}^{obs}$
are associated with resolved photon processes. It is clear that both classes of
event are present in both $P^2$ ranges. To quantify the relative contributions
of these classes of event, we have calculated the ratio $N_{res}/N_{dir}$,
defined as the number of events at low $x_{\gamma}^{obs}$ ($x_{\gamma}^{obs} <
0.75$) divided by the number of events at high $x_{\gamma}^{obs}$
($x_{\gamma}^{obs} > 0.75$). Figure \ref{fig:resdirfig} shows $N_{res}/N_{dir}$
as a function of $P^2$. $N_{res}/N_{dir}$ is independent of $y_{JB}$ for the
sample of events with quasi-real photons passing all the cuts applied. This
implies that the difference in the $x_{\gamma}^{obs}$ distributions for the two
samples is not due to the differing $y_{JB}$ acceptances of the two positron
detectors.
The acceptance corrections that will eventually be applied to these data are
only weakly dependent on $x_{\gamma}^{obs}$ and we therefore expect the
corrections to $N_{res}/N_{dir}$ to be small. More work is required to
understand fully these corrections however. In conclusion we find the
preliminary result that for events with photons of virtuality $0.1 < P^2 < 0.55$
GeV$^2$ there is a contribution from resolved photon processes. The size of this
contribution relative to that from direct photon processes seems to decrease
with increasing photon virtuality.
\begin{figure}[hbt]
\begin{center}\mbox{\epsfig{file=xgamsglas.eps,width= 11cm}}\end{center}
\caption{Uncorrected $x_{\gamma}^{obs}$ distributions for a) virtual and b)
quasi-real photons}
\label{fig:xgfig}
\end{figure}
\begin{figure}[hbt]
\begin{center}\mbox{\epsfig{file=resdirglas.eps,width= 9cm}}\end{center}
\caption{The uncorrected ratio $N_{res}/N_{dir}$ as a function of photon
virtuality $P^2$ (GeV$^2$)}
\label{fig:resdirfig}
\end{figure}
\setcounter{secnumdepth}{0}
|
\section{Introduction}
\label{Sect1}
\vspace{10pt}
Even though the Standard Model (SM) remains unchallenged by
an impressive body of precision electroweak measurements,
over the years the general consensus has been
that there must exist new physics lurking
behind the horizon of the related problems of the origin of mass
and the chirality structure of the electroweak interactions.
Any embodiment of such new physics would manifest itself trough
deviations from the predictions of the SM for specific
observables and production mechanims, be it at low, intermediate
or high energies.
In the case of semi-leptonic weak interactions at low energies,
it is essentially only the radioactive nucleus which
is available as a laboratory for the search of new physics
through $\beta$-decay\cite{DQ}.
In particular, it has recently been emphasized\cite{Quin} that when
compared to asymmetry measurements,
the {\em relative\/}
longitudinal polarisation of $\beta$ particles emitted
in directions parallel
or antiparallel to the polarisation vector of an oriented nucleus
presents an enhanced sensitivity to a hypothetical
right-handed charged current contribution.
Indeed, one among
other attractive extensions of the SM is
obtained by enlarging the electroweak sector of the latter model
to the gauge group
$SU(2)_L\times SU(2)_R\times U(1)_{B-L}$
in the context of so-called
Left-Right Symmetric Models (LRSM)\footnote{For reviews and references
to the original literature, see for example Refs.\cite{LRSM,Herczeg,Lang},}.
In such a case, the new
physics to be found beyond the SM stems from the existence of additional
charged and neutral gauge $W'$ and $Z'$
and physical Higgs bosons, as well as a host of additional
phenomenological parameters leading to massive neutrinos
and the ensuing leptonic flavour mixing, to new sources of CP violation
o\-ri\-gi\-na\-ting in the Higgs as well as in the Yukawa sectors,
and to other physical effects of interest (see for example
Refs.\cite{Herczeg,Lang}). In spite of their great appeal,
such extensions do not provide an understanding for the origin of mass
nor for the chirality structure of the fundamental electroweak
interactions---now enlarged to include right-handed ones---, even though
parity invariance may {\em effectively\/} be restored in
certain classes of such models for
processes characterised by momenta transfers much larger than
the masses of the new gauge bosons $W'$ and $Z'$. Nevertheless,
the structure of LRSM typically arises as a low-energy effective theory
for most grand unified theories---supersymmetric or not---whose
gauge symmetry breaking pattern includes the gauge
group $SO(10)$.
In this note, the sensitivity offered by the relative longitudinal
polarisation measurement mentioned above is considered\cite{Quin}
in the case of allowed $\beta$-decays. Specific results
are established for mirror nuclei in the context of {\em general\/}
LRSM, following the discussion
of Ref.\cite{Oscar} which analysed the sensitivity offered by
{\em asymmetry\/} measurements in the context of so-called
Manifest Left-Right Symmetric Models (MLRSM) which are constructed such that
at tree-level gauge coupling constants and flavour mixing matrices be identical
for both chirality sectors of the theory.
In fact, two such polarisation-asymmetry correlation experiments have already
yielded results in the cases of the allowed pure Gamow-Teller (GT)
decays of $^{107}$In\cite{107In} and $^{12}$N\cite{12N},
for which the sen\-si\-ti\-vi\-ty to possible right-handed
charged current contributions must {\em a priori\/} be among the largest
attainable, owing to the Gamow-Teller character of these transitions
and their large asymmetry parameter.
The outline of the note is as follows. Sect.\ref{Sect2} addresses the
longitudinal polarisation-asymmetry correlation measurement in the general case
of allowed $\beta$-decays assuming only vector and axial contributions
to the charged current interaction. In Sect.\ref{Sect3}, these
results are particularised to {\em general\/}
Left-Right Symmetric Models.
The case of mirror nuclei is then
analysed in Sect.\ref{Sect4}, in order to identify possible
attractive candidates for this type of measurements beyond the
cases of allowed pure GT decays. This is done
on general grounds, without paying attention to the necessary
considerations concerning the technical
feasibility either of such experiments or of the required degree
of nuclear polarisation, which would have to be addressed
on a case by case basis. Further results
of interest are presented in
an Appendix, while the note ends with some general conclusions.
\section{Allowed $\beta$-decay and $(V,A)$ charged currents}
\label{Sect2}
The expressions for physical observables of relevance
to allowed $\beta$-decay in the
case of the most general four-fermi effective interaction are
available from Ref.\cite{JTW}. In this note, {\em only vector $(V)$ and
axial $(A)$ a priori complex four-fermi
coupling coefficients are considered\/}, associated
to pure vector and axial charged current contributions to $\beta$-decay.
{\em Under this specific restriction\/} and ignoring recoil
order corrections\footnote{In the case of
super-allowed decays such as those of mirror nuclei, these corrections
are small and such an approximation is certainly justified
in the present context which aims at identifying potential
candidates for relative longitudinal polarisation-asymmetry
correlation experiments.},
when only the $\beta^{\mp}$ particle is
observed the corresponding distribution is then given by the spin density
matrix\cite{JTW}
\begin{displaymath}
\frac{d^3W}{dE\,d^2\Omega}\ =\ W_0(E)\,\xi\,\Bigg\{\ 1\,+\,\beta J A
(\hat{p}.\hat{J})\ +
\end{displaymath}
\begin{equation}
+\ \vec{\sigma}.\Bigg[\beta G\hat{p}+J\frac{\gamma_z}{\gamma}N'\hat{J}+
J(1-\frac{\gamma_z}{\gamma})N'(\hat{p}.\hat{J})\hat{p}
\mp J\frac{\alpha Z}{\gamma} A (\hat{J}\times\hat{p})\Bigg]\Bigg\}\ \ ,
\label{eq:Dist1}
\end{equation}
with the following notation. The parameter $(-1\le J\le +1)$
stands for the degree
of nuclear polarisation of the oriented nucleus, while the normalised
vector $\hat{J}$ represents the direction of nuclear polarisation.
Of course, $\beta$ is the velocity of the emitted $\beta^{\mp}$
particle---normalised to the speed of light in vacuum---, with $\gamma$
the associated relativistic dilatation factor $(\gamma=1/\sqrt{1-\beta^2})$,
while $E$, $\hat{p}$ and $\vec{\sigma}$
are the $\beta^{\mp}$ particle total energy, normalised
momentum and polarisation (spin) quantum operator, respectively. Coulomb
corrections\cite{JTW} induce a dependence on the fine structure constant
$\alpha$, while $\gamma_z$ is defined as $(\gamma_z=\sqrt{1-(\alpha Z)^2}\,)$
where $Z$ is the charge of the {\em daughter\/} nucleus. The coefficients
$W_0(E)$, $\xi$, $A$, $G$ and $N'$ are given in the Appendix for
$\beta$-transitions of initial nuclear spin $J$ and final
nuclear spin $J'$ in terms
of the effective four-fermi complex coupling coefficients $C_V$,
$C_V'$, $C_A$ and $C_A'$ introduced in Ref.\cite{JTW}.
Finally, the choice of sign in front of the last term in the above expression
corresponds to whether it is an electron or a positron which is emitted;
throughout the note, whenever such a choice is indicated, the upper
sign always corresponds to electron emission, and the lower sign to positron
emission.
Since the purpose of the present analysis is the identification of
attractive candidates for polarisation-asymmetry correlation
experiments besides allowed pure GT decays, let us consider the ideal
situation in which $\beta^{\mp}$ particles of a specific
energy $E$---or velocity $\beta$---and initial momentum direction $\hat{p}$
are observed and for which it is {\em exactly\/}
their {\em longitudinal
polarisation\/} which is measured. All other things being equal,
dif\-fe\-rent nuclei may then be compared on an equal footing under
these identical {\em idealised\/} experimental conditions.
Obviously on a practical level,
other considerations would also have to be addressed, such as the
finite energy and angular acceptance of any experimental set-up
leading to additional contributions from transverse spin components as well,
the spin precession in magnetic fields of spectrometers or polarimeters,
and more importantly, the
feasibility of the production and of the polarisation of a given
nucleus. As mentioned in the Introduction, such issues are not
tackled in this note, since they can only be considered
on a case by case
basis once a potential candidate is identified.
Under the stated idealised conditions,
and given a nucleus of polarisation $J$, the
experimental asymmetry of the $\beta$-decay distribution is of the form
\begin{equation}
N_\beta(J)\ =\ N_0\,\Big[\,1+\beta J A (\hat{p}.\hat{J})\,\Big]\ \ ,
\end{equation}
where $N_0$ represents the normalised source activity.
Similarly, from the spin density matrix in (\ref{eq:Dist1}),
the {\em longitudinal\/} polarisation of the produced $\beta^{\mp}$
particle is then given by the expression
\begin{equation}
P_L(J)\ =\ \frac{\beta G + J N' (\hat{p}.\hat{J})}
{1+\beta J A (\hat{p}.\hat{J})}\ \ \ .
\end{equation}
Following the suggestion of Ref.\cite{Quin}, let us consider
{\em relative\/} measurements of such observables for different
values of the degree $J$ of the
polarisation of the oriented nucleus, without changing the
direction $\hat{J}$ of its orientation. The obvious advantage of such
measurements is that they are much less sensitive to systematic
effects than are absolute measurements of asymmetries or
polarisations. Therefore, given two different degrees
$J_1$ and $J_2$ of nuclear polarisation
and {\em identical\/} normalised source
activities $N_0$, the relative {\em experimental\/} asymmetry
$A_{\rm exp}(J_2,J_1)$ may be defined according to
\begin{equation}
1\ -\ A_{\rm exp}(J_2,J_1)\ =\ \frac{N_\beta(J_2)}{N_\beta(J_1)}\ =\
\frac{1+\beta J_2 A (\hat{p}.\hat{J})}
{1+\beta J_1 A (\hat{p}.\hat{J})}\ \ \ ,
\end{equation}
while the relative longitudinal polarisation is simply
\begin{equation}
R(J_2,J_1)\ =\ \frac{P_L(J_2)}{P_L(J_1)}\ =\
\frac{1}{1-A_{\rm exp}(J_2,J_1)}\ \frac{\beta G + J_2 N' (\hat{p}.\hat{J})}
{\beta G + J_1 N' (\hat{p}.\hat{J})}\ \ \ .
\end{equation}
Since the purpose of the present discussion is the identification
of potential attractive candidates for this type of measurement,
all other things being equal, it proves simpler to consider henceforth
only two specific situations with regards to nuclear polarisation.
The first is obtained when the reference nuclear polarisation $J_1$
is vanishing, $(J_1=0)$, and when $(J_2=-J)$. The second\cite{Quin}
corresponds to a situation in which
the $\beta^{\mp}$ longitudinal polarisation is considered for opposite
directions of nuclear polarisation, namely for $(J_2=-J_1=-J)$.
In the first instance, the experimental asymmetry is given by
\begin{equation}
A_{\rm exp}(-J,0)\ =\ \beta\,J\,A\,(\hat{p}.\hat{J})\ \ \ ,
\label{eq:Aexp1}
\end{equation}
which upon substitution in the expression for the relative
longitudinal polarisation $R(-J,0)$ leads to the result,
\begin{equation}
R(-J,0)\ =\ \frac{1}{\beta^2}\,\frac{1}{1-A_{\rm exp}(-J,0)}\,
\Bigg[\,\beta^2\ -\ A_{\rm exp}(-J,0)\,
\frac{\xi\,(\xi N')}{(\xi A)\,(\xi G)}\,\Bigg]\ \ \ .
\label{eq:R(-J,0)}
\end{equation}
Similarly in the second instance when $\left(\,(J_2,J_1)=(-J,J)\,\right)$,
one has
\begin{equation}
A_{\rm exp}(-J,J)\ =\ \frac{2\,\beta\,J\,A\,(\hat{p}.\hat{J})}
{1\,+\,\beta\,J\,A\,(\hat{p}.\hat{J})}\ \ \ ,
\label{eq:Aexp2}
\end{equation}
or equivalently,
\begin{equation}
\beta\,J\,A\,(\hat{p}.\hat{J})\ =\ \frac{A_{\rm exp}(-J,J)}
{2\,-\,A_{\rm exp}(-J,J)}\ \ \ .
\end{equation}
When substituted in the definition for $R(-J,J)$, one then
derives the expression for the relative longitudinal polarisation
in this case,
\begin{equation}
R(-J,J)\ =\ \frac{1}{1-A_{\rm exp}(-J,J)}\,
\Bigg[\,1\ -\ 2\ \frac{1}{1\,+\,\beta^2\,\frac{2-A_{\rm exp}(-J,J)}
{A_{\rm exp}(-J,J)}\,\frac{(\xi A)(\xi G)}{\xi (\xi N')}}\,\Bigg]\ \ \ .
\label{eq:R(-J,J)}
\end{equation}
Note that it is the same
quantity
\begin{equation}
\frac{\xi\,(\xi N')}{(\xi A)\,(\xi G)}\ \ \ ,
\label{eq:ratio}
\end{equation}
depending on the underlying physics, which appears in
the expressions for $R(-J,0)$ and $R(-J,J)$. Since
this combination of parameters takes the value unity in the SM
(see the Appendix), any deviation
of the ratio (\ref{eq:ratio}) from unity must stem from some new
physics\footnote{Note that this remark does not account for possible
recoil order corrections to the relative longitudinal
polarisation, which are ignored in this note.}
beyond the SM. Let us thus introduce the quantity
$\Delta$ defined by\footnote{The numerical factor of a quarter is for later
convenience.}
\begin{equation}
\Delta\ \equiv\ \frac{1}{4}\,\Big[\,
\frac{\xi\,(\xi N')}{(\xi A)\,(\xi G)}\ -\ 1\,\Big]\ =\
\frac{1}{4}\,\frac{\xi(\xi N')-(\xi A)(\xi G)}{(\xi A)(\xi G)}\ \ \ ,
\end{equation}
which thus characterises the new physics beyond the SM which
may be probed through
relative longitudinal polarisation-asymmetry correlation experiments.
In order to assess the sensitivity of these measurements
to such new physics, it
is necessary to compare the values taken by the relative
polarisations $R(-J,0)$ and $R(-J,J)$ to their values
obtained simply by setting $(\Delta=0)$ in (\ref{eq:R(-J,0)})
and (\ref{eq:R(-J,J)}) without modification of the
{\em experimental\/} asymmetries $A_{\rm exp}$.
This leads to, respectively,
\begin{equation}
R_0(-J,0)\ =\ \frac{1}{\beta^2}\,\frac{\beta^2\,-\,A_{\rm exp}(-J,0)}
{1\,-\,A_{\rm exp}(-J,0)}\ \ \ ,
\end{equation}
and
\begin{equation}
R_0(-J,J)\ =\ \frac{1}{1\,-\,A_{\rm exp}(-J,J)}\
\frac{\beta^2\Big[2-A_{\rm exp}(-J,J)\Big]\,-\,
A_{\rm exp}(-J,J)}
{\beta^2\Big[2-A_{\rm exp}(-J,J)\Big]\,+\,A_{\rm exp}(-J,J)}\ \ \ .
\end{equation}
Let us emphasize that these expressions {\em are not
those which one may derive in the SM\/} for the
relative longitudinal polarisations,
since they are obtained simply by setting only the quantity
$\Delta$ to zero in the results for $R(-J,0)$ and $R(-J,J)$
without assuming that the asymmetry parameter $A$ is given
by its value $A_0$ in the SM.
Indeed, the expressions for $R_0(-J,0)$
and $R_0(-J,J)$ still involve the {\em experimental\/} asymmetries
$A_{\rm exp}(-J,0)$ and $A_{\rm exp}(-J,J)$
which may differ from the values predicted by the SM
if new physics does contribute to the asymmetry parameter $A$,
as made explicit in (\ref{eq:Aexp1}) and (\ref{eq:Aexp2}).
Nevertheless, it is obviously useful\cite{Quin} to express
the relative longitudinal polarisation in terms of the {\em directly
observable\/} experimental asymmetries $A_{\rm exp}(-J,0)$
and $A_{\rm exp}(-J,J)$.
Given the quantities $R_0(-J,0)$ and $R_0(-J,J)$, any genuine physical
deviation from the SM contributing to $R(-J,0)$ and $R(-J,J)$ would
be manifested through a value different from unity for the ratios,
\begin{equation}
\frac{R(-J,0)}{R_0(-J,0)}\ \ \ ,\ \ \
\frac{R(-J,J)}{R_0(-J,J)}\ \ \ .
\end{equation}
In terms of the quantities introduced in the forthcoming definitions,
it is straightforward to establish the following {\em exact\/}
results, valid independently of whether the parameter $\Delta$
is small in comparison to unity or not. One finds,
\begin{equation}
\frac{R(-J,0)}{R_0(-J,0)}\ =\ 1\ -\ k(-J,0)\,\Delta\ \ \ ,
\end{equation}
with the factor $k(-J,0)$ given by
\begin{equation}
k(-J,0)\ =\ 4\,\frac{A_{\rm exp}(-J,0)}
{\beta^2\,-\,A_{\rm exp}(-J,0)}\ \ \ .
\end{equation}
Similarly for $R(-J,J)$, one has
\begin{equation}
\frac{R(-J,J)}{R_0(-J,J)}\ =\ 1\ -\ k(-J,J)\,
\frac{\Delta}{1+4\,\frac{A_{\rm exp}(-J,J)}{\beta^2
\left[2-A_{\rm exp}(-J,J)\right]
+ A_{\rm exp}(-J,J)}\,\Delta}\ \ \ ,
\label{eq:RR02}
\end{equation}
with the factor $k(-J,J)$ defined by
\begin{equation}
k(-J,J)\ =\ 8\ \frac{\beta^2\,A_{\rm exp}(-J,J)\,
\Big[2-A_{\rm exp}(-J,J)\Big]}
{\beta^4\,\Big[2-A_{\rm exp}(-J,J)\Big]^2-A^2_{\rm exp}(-J,J)}\ \ \ .
\end{equation}
Note that under the considered idealised situation in which it is
exactly the longitudinal $\beta^{\mp}$ polarisation which is measured,
both the values of $R_0(-J,0)$ or $R_0(-J,J)$ and of the
factors $k(-J,0)$ or $k(-J,J)$
only depend\cite{Quin}
on the corresponding
{\em observed experimental\/} asymmetry $A_{\rm exp}$
and on the value of $\beta^2$.
Moreover, the factors $k(-J,0)$ and $k(-J,J)$
offer an enhanced\cite{Quin}
sensitivity to any deviation from the value unity expected in the SM
for the corresponding ratio $R/R_0$ of relative longitudinal polarisations.
Indeed, the factor appropriate to the first case, namely $k(-J,0)$,
diverges\footnote{This divergence does not entail a loss of
physical significance of the results, but follows simply from the fact
that the quantity $R_0(-J,0)$ vanishes as the experimental
asymmetry $A_{\rm exp}(-J,0)$ approaches the value $A_{\rm exp}^{(0)}(-J,0)$,
while at the same time the product $R_0(-J,0)\,k(-J,0)$ remains
finite as it should since $R(-J,0)$ is finite under all circumstances.
In other words, the significance of the divergence is that
when the experimental asymmetry $A_{\rm exp}(-J,0)$ is
optimised at the value $A^{(0)}_{\rm exp}(-J,0)$, the
contribution of $(-R_0(-J,0)\,k(-J,0)\,\Delta)$
to $R(-J,0)$ becomes increasingly larger than that of $R_0(-J,0)$.
Of course, the same comments also apply to the case of $R(-J,J)$
which follows.}
as the experimental asymmetry $A_{\rm exp}(-J,0)$ approches
the value
\begin{equation}
A_{\rm exp}^{(0)}(-J,0)\ =\ \beta^2\ \ .
\end{equation}
Similarly, the enhancement factor appropriate
to the second case, namely $k(-J,J)$,
diverges as the experimental asymmetry $A_{\rm exp}(-J,J)$
approches the value
\begin{equation}
A_{\rm exp}^{(0)}(-J,J)\ =\ \frac{2\,\beta^2}{1+\beta^2}\ \ ,
\end{equation}
while the quantity multiplied by $k(-J,J)$ in (\ref{eq:RR02}) then
reduces to
\begin{equation}
\frac{\Delta}{1+2\Delta}\ \ \ .
\end{equation}
Note that in either case, the optimal experimental asymmetry
$A^{(0)}_{\rm exp}$ corresponds to a degree of nuclear
polarisation $J$ and a choice of $\beta$ such that
\begin{equation}
A\,J\,(\hat{p}.\hat{J})\ =\ \beta\ \ \ .
\label{eq:bJA}
\end{equation}
In other words, for a given nucleus, namely a given asymmetry
parameter $A$, the optimal sensitivity
to a possible contribution
from right-handed currents is achieved for values of the
{\em effective\/} degree of nuclear polarisation,
\begin{equation}
{\cal P}\ =\ |J(\hat{p}.\hat{J})|\ \ \ ,
\end{equation}
and of the $\beta^\mp$ particle velocity $\beta$ such that
\begin{equation}
|A\,{\cal P}|\ =\ \beta\ \ \ ,
\label{eq:optimal}
\end{equation}
the choice of sign for $J(\hat{p}.\hat{J})$
being such that the experimental asymmetries
$A_{\rm exp}$ as defined in this note be positive.
These conclusions correspond to the advertised\cite{Quin} sensitivity of this
type of measurement to right-handed currents: the closer the experimental
asymmetry $A_{\rm exp}$ to $A_{\rm exp}^{(0)}$, namely the closer
the choice of values of $(\beta,{\cal P})$ to the optimal situation
such that $(\beta=|A{\cal P}|)$, the
larger the sensitivity to a possible
deviation $(\Delta\ne 0)$ from the SM. Since values of $\beta$
for which the decay count rate is the largest
are typically close to the maximal value of unity\footnote{Small,
albeit vanishing values of $\beta$ are of course possible also,
in which case the optimal sensitivity is achieved for small
effective nuclear polarisations ${\cal P}$. However, $\beta^{\mp}$ count
rates decrease as $\beta$ approaches zero, thus leading to a loss
in statistics for any precision measurement.}, clearly the best
sensitivity requires both the largest possible effective degree of
nuclear polarisation ${\cal P}$ and the largest possible
asymmetry parameter $|A|$. In particular, since the asymmetry
parameter $|A|$ for allowed pure GT transitions of nuclear spin sequence
$(J'=J-1)$ is maximal in the SM
(see the Appendix), such $\beta$-decays
are certainly among the best candidates to probe for right-handed
currents through longitudinal polarisation-asymmetry correlation experiments.
This is the case for example
for the $^{107}$In\cite{107In} and $^{12}$N\cite{12N} nuclei.
Given the expressions for $\xi$, $A$, $G$ and $N'$
listed in the Appendix, the quantity $\Delta$ may easily be
related to the underlying effective coupling coefficients
$C_V$, $C_V'$, $C_A$ and $C_A'$. It proves useful to introduce
the notation,
\begin{equation}
a_L\ =\ M_F^2\,|C_V+C_V'|^2\ +\ M^2_{GT}\,|C_A+C_A'|^2\ \ \ ,
\label{eq:aL}
\end{equation}
\vspace{3pt}
\begin{equation}
a_R\ =\ M_F^2\,|C_V-C_V'|^2\ +\ M^2_{GT}\,|C_A-C_A'|^2\ \ \ ,
\label{eq:aR}
\end{equation}
\vspace{3pt}
\begin{equation}
b_L\ =\ \mp M^2_{GT}\,\lambda_{J'J}\,|C_A+C_A'|^2\ -\
2\delta_{J'J} M_F M_{GT}\,\sqrt{\frac{J}{J+1}}\,{\rm Re}
\left((C_V+C_V')(C_A^*+{C_A'}^*)\right)\ \ \ ,
\label{eq:bL}
\end{equation}
\begin{equation}
b_R\ =\ \mp M^2_{GT}\,\lambda_{J'J}\,|C_A-C_A'|^2\ -\
2\delta_{J'J} M_F M_{GT}\,\sqrt{\frac{J}{J+1}}\,{\rm Re}
\left((C_V-C_V')(C_A^*-{C_A'}^*)\right)\ \ \ ,
\label{eq:bR}
\end{equation}
in terms of which one then derives the {\em exact\/}
expression
\begin{equation}
\Delta\ =\ \frac{1}{2}\,\frac{a_L\,b_R\ +\ a_R\,b_L}
{(a_L\,-\,a_R)(b_L\,-\,b_R)}\ \ \ .
\label{eq:4Delta}
\end{equation}
Since the numerator of the result in (\ref{eq:4Delta}) involves
precisely the differences $(C_V-C_V')$ and $(C_A-C_A')$
which vanish for purely left-handed couplings as is the
case in the SM, it is
clear that the quantity $\Delta$---probed through
relative longitudinal polarisation-asymmetry correlation measurements---is
indeed sensitive to
right-handed charged current contributions to allowed $\beta$-decay.
Note also that in terms of the quantities $a_L$, $a_R$, $b_L$ and $b_R$
introduced above, the asymmetry parameter $A$ reads,
\begin{equation}
A\ =\ \frac{b_L\,-\,b_R}{a_L\,+\,a_R}\ \ \ .
\end{equation}
When deviations of the coefficients $C_V$, $C_V'$, $C_A$
and $C_A'$ from their values in the SM are small, it is justified to
consider a first order expansion of $\Delta$ in the quantities
$a_R$ and $b_R$, leading to
\begin{equation}
\Delta\ \simeq\ \frac{1}{2}\,\Bigg[\,\frac{a_R}{(a_L)_0}\ +\
\frac{b_R}{(b_L)_0}\,\Bigg]\ \ \ ,
\end{equation}
where $(a_L)_0$ and $(b_L)_0$ are the values of $a_L$ and $b_L$
in the SM, respectively,
\begin{equation}
(a_L)_0\ =\ 4\,|C_V^{(0)}|^2\,M_F^2\,\Big[\,1+\lambda^2\,\Big]\ \ \ ,
\end{equation}
and
\begin{equation}
(b_L)_0\ =\ 4\,|C_V^{(0)}|^2\,M_F^2\,\Big[\,
\mp\,\lambda^2\,\lambda_{J'J}\ -\ 2\,\delta_{J'J}\lambda\,
\sqrt{\frac{J}{J+1}}\,\Big]\ \ \ .
\end{equation}
Here, $C_V^{(0)}$ is the value of the coefficient $C_V$ in the
SM, while $\lambda$ is the ratio
\begin{equation}
\lambda\ =\ \frac{g_A}{g_V}\,\frac{M_{GT}}{M_F}\ \ \ ,
\label{eq:lambda}
\end{equation}
$g_V$ and $g_A$ being the nucleon vector and axial couplings,
respectively (see the
Appendix for further details).
Another situation of particular interest is that of
allowed pure GT transitions, for which one simply finds,
\begin{equation}
A_{|_{GT}}\ =\ {A_0}_{|_{GT}}\,
\frac{|C_A+C_A'|^2\,-\,|C_A-C_A'|^2}
{|C_A+C_A'|^2\,+\,|C_A-C_A'|^2}\ \ \ ,
\label{eq:AGT}
\end{equation}
${A_0}_{|_{GT}}$ being the asymmetry parameter for allowed pure GT
decays in the SM (see the Appendix), as well as
\begin{equation}
\Delta_{|_{GT}}\ =\ \frac{1}{4}\,\Bigg[
\left(\frac{{A_0}_{|_{GT}}}{A_{|_{GT}}}\right)^2\,-\,1\Bigg]\ =\
\frac{|C_A-C_A'|^2}{|C_A+C_A'|^2}\,
\frac{1}{\Big[1-\frac{|C_A-C_A'|^2}{|C_A+C_A'|^2}\Big]^2}\ \ \ .
\end{equation}
These expressions are independent of the nucleus involved
and of the specifics of the
underlying new physics which would be leading to $(\Delta_{|_{GT}}\ne 0)$.
Thus in such a case, the parameter $\Delta$ indeed provides a {\em direct\/}
measure for right-handed current contributions to the coupling of
the leptonic charged current to the hadronic axial
charged current in nuclear $\beta$-decay. On the other hand, note that
even though the quantity $\Delta$ is in fact related
to the asymmetry parameter $A$ in the case of allowed pure GT transitions,
a measurement of the relative longitudinal polarisation-asymmetry
correlation is potentially far more sensitive
to contributions from right-handed
currents than is a measurement of the asymmetry parameter $A$
itself. Indeed,
the former sensitivity to the ratio
$\Big(|C_A-C_A'|^2/|C_A+C_A'|^2\Big)$
is characterised by the enhancement factors
$k(-J,0)$ and $k(-J,J)$ which {\em a priori\/} may reach quite large
values by appropriate choices of $\beta$ and of the effective degree
of nuclear polarisation ${\cal P}$. On the other hand,
the sensitivity
of the asymmetry parameter $A$ to the same ratio
is essentially characterised by a
fixed enhancement factor of two only, as follows from (\ref{eq:AGT}).
Before concluding this general discussion, let us address one last
issue. As was already pointed out previously,
since in most cases the value of $\beta$ is not much different
from unity in the energy domain where the $\beta^\mp$ count
rate is maximal, for a given effective degree of nuclear polarisation
${\cal P}$ the enhancement factors $k(-J,0)$ and $k(-J,J)$
are the largest for experimental asymmetries $A_{\rm exp}(-J,0)$
and $A_{\rm exp}(-J,J)$ as
close to the value unity as possible. However, this also implies
that the count rate of $\beta^{\mp}$ particles associated to the
direction of nuclear polarisation for which the sensitivity to $\Delta$ is
the largest, is also the smaller the closer the experimental
asymmetry to the value unity. Indeed, this count rate
at the optimal sensitivity such that $(|A{\cal P}|=\beta)$
is proportional to $(1-\beta^2)$. Nevertheless, it is possible to show
that the loss in count rate is compensated for by the
gain in sensitivity; for either configuration of nuclear
polarisations considered in this note, the figure of merit characterising the
precision with which a deviation from the value unity for the
ratio $R/R_0$ may be established experimentally
is indeed optimal for
the previously given value $A_{\rm exp}^{(0)}$ of the
experimental asymmetry $A_{\rm exp}$ at which the corresponding
enhancement factor $k$ is the largest.
\section{General Left-Right Symmetric Models}
\label{Sect3}
The results of the previous section are valid quite generally for allowed
decays,
since the only assumptions made so far are that the
effective Hamiltonian for $\beta$-decay receives contributions
from vector and axial
currents only, with arbitrary complex coupling coefficients,
and that recoil order corrections to relative
longitudinal polarisations are negligible.
Let us now particularise the discussion to {\em general\/} Left-Right
Symmetric Models, based on the gauge group
$SU(2)_L\times SU(2)_R\times U(1)_{B-L}$ in the electroweak sector.
In as far as semi-leptonic charged weak interactions are
concerned, contributions to the $\beta$-decay process in such models
follow from charged gauge boson and Higgs exchanges. However,
charged Higgs exchanges---when contributing\footnote{Charged Higgs
exchanges do not contribute in the case of allowed pure GT
transitions.}---shall be ignored in
the present discussion, assuming that they are suppressed
through small coupling constants and large masses.
This effectively
leaves only charged gauge boson exchanges,
namely those of the ordinary gauge boson $W$ of mass\cite{PDG}
\begin{equation}
M_1\ =\ 80.22\,\pm\,0.26\ {\rm GeV}/c^2\ \ ,
\end{equation}
and of the hypothetical heavy charged gauge boson $W'$ of mass $M_2$.
Thus, since scalar and pseudoscalar charged Higgs exchange contributions are
ignored, within the framework of LRSM $\beta$-decay processes
indeed receive contributions from vector and axial couplings only,
namely from left- and right-handed fermionic gauge currents.
However, the propagating gauge bosons $W$ and $W'$ are not necessarily those
which couple to fermions of definite chirality. Indeed, the physical
charged gauge bosons $W$ and $W'$ and the charged gauge bosons associated
to the underlying gauge group $SU(2)_L\times SU(2)_R$ which thus
couple to currents of specific chirality, are related to
one another through the mixing matrix\cite{Herczeg},
\begin{equation}
\begin{array}{c c l}
W_L^+ & = & \cos\zeta\,W_1^+\ +\ \sin\zeta\,W_2^+\ \ \ ,\\ \\
W_R^+ & = & e^{i\omega}\Big[\,-\sin\zeta\,W_1^+\ +\
\cos\zeta\,W_2^+\,\Big]\ \ \ ,
\end{array}
\end{equation}
or equivalently
\begin{equation}
\begin{array}{c c l}
W_1^+ & = & \cos\zeta\,W_L^+\ -\ e^{-i\omega}\sin\zeta\,W_R^+\ \ \ ,\\ \\
W_2^+ & = & \sin\zeta\,W_L^+\ +\ e^{-i\omega}\cos\zeta\,W_R^+\ \ \ .
\end{array}
\end{equation}
Here, $W_L$ and $W_R$ denote the fundamental gauge bosons
coupling to the fermionic currents of left- and right-handed chirality,
respectively, while $W_1$ and $W_2$ denote the physical mass eigenstate
gauge bosons of masses $M_1$ and $M_2$, respectively.
The parameter $\zeta$ is a mixing angle\footnote{Our choice
of sign for $\zeta$ is opposite to that made in Ref.\cite{Lang}
but agrees with that of Ref.\cite{Herczeg}.} for charged gauge
bosons, while the parameter $\omega$ determines a CP violating
phase originating from complex vacuum expectation values in the
Higgs sector. These quantities are constrained
phenomenologically in certain classes of LRSM\cite{Herczeg,Lang}.
In addition, the coupling strength of the gauge bosons $W_L$ and $W_R$ to the
fundamental fermions is specified by the gauge coupling constants
$g_L$ and $g_R$, respectively. Again,
the ratio $g_R/g_L$ is constrained
phenomenologically\cite{Cvetic}.
Finally, the coupling of the charged gauge bosons $W_L$ and $W_R$
to quarks and leptons also involves different Cabibbo-Kobayashi-Maskawa (CKM)
flavour mixing matrices in each chirality sector. Since in the
hadronic sector, only the up and down quarks couple to the
$\beta$-decay process, the relevant CKM matrix elements are
denoted
\begin{equation}
V^L_{ud}\ \ \ ,\ \ \ V^R_{ud}\ \ \ ,
\end{equation}
for the left- and right-handed sectors, respectively.
Similarly in the leptonic sector, {\em a priori\/} the emitted
electron or positron may be produced together
with a mass eigenstate neutrino $\nu_i$,
with an amplitude determined by leptonic
CKM matrix elements denoted as
\begin{equation}
U^L_{ie}\ \ \ ,\ \ \ U^R_{ie}\ \ \ .
\end{equation}
Generally, the ratios
\begin{equation}
v_{ud}\ =\ \frac{V^R_{ud}}{V^L_{ud}}\ \ \ ,\ \ \
v_{ie}\ =\ \frac{U^R_{ie}}{U^L_{ie}}\ \ \ ,\ \ \
\label{eq:comb1}
\end{equation}
are arbitrary complex numbers, related to the underlying complex Yukawa
couplings and Higgs vacuum expectation values, thus
potentially leading to new
CP violating processes in their own right. Indeed,
even though it is always possible by an appropriate choice of
phases of the fermionic fields to fix the quark as well as the
leptonic CKM matrix elements $V^{(L,R)}_{ud}$ and $U^{(L,R)}_{ie}$
either in the left- or in the right-handed sectors
to be real---as is the case for the ordinary Cabibbo angle---, this is
not possible for both sectors simultaneously. Assuming that either
$v_{ud}$ or $v_{ie}$ or both be real, would imply particular restrictions
on the class of LRSM being considered. Here again,
there exist\cite{Herczeg,Lang}
certain phenomenological constraints on these quark and lepton CKM matrix
elements.
The above description thus provides the complete set of parameters
required for the application of the discussion of Sect.\ref{Sect2}
to {\em general\/} LRSM, no assumption nor approximation
whatsoever as to the definition of such models
being implied at this stage.
It proves useful to introduce the following combinations of these
quantities,
\begin{equation}
t\ =\ \tan\zeta\ \ \ ,\ \ \
\delta\ =\ \frac{M^2_1}{M^2_2}\ \ \ ,\ \ \
r\ =\ \frac{g_R}{g_L}\ \ ,
\label{eq:comb2}
\end{equation}
\begin{equation}
v_u\ =\ \frac{|V^R_{ud}|^2}{|V^L_{ud}|^2}\ =\ |v_{ud}|^2\ \ \ ,\ \ \
v_e\ =\ \frac{\sum_i'|U^R_{ie}|^2}{\sum_i'|U^L_{ie}|^2}\ \ \ ,
\label{eq:comb3}
\end{equation}
and finally
\begin{equation}
\eta_0\ =\ \frac{1}{M^2_1}\left(\frac{g_L^2}{8}\right)\,\cos^2\zeta\ =\
\frac{1}{M^2_1}\left(\frac{g_L^2}{8}\right)\,\frac{1}{1+t^2}\ \ \ .
\label{eq:comb4}
\end{equation}
In particular, the symbol $\sum_i'$
appearing in the definition
of the quantity $v_e$ in a notation to be detailed
presently stems from the following fact.
Since the neutrino produced in the $\beta$-decay process is not
observed, any measurement involves a sum over all neutrino mass
eigenstates whose production is not forbidden kinematically.
Therefore, assuming that all neutrinos produced in the process
have a mass sufficiently small in order not to induce
a significant distortion of the $\beta^{\mp}$ energy distribution,
one need only sum the
corresponding decay spectra over all such mass eigenstate neutrinos
$\nu_i$ without accounting for a modification
in phase-space factors. In other words, the symbol $\sum_i'$ stands
for a sum over all neutrinos $\nu_i$ whose production is
not kinematically suppressed. In particular, note that when all
mass eigenstate neutrinos
do participate in the process, each of the sums
\begin{equation}
{\sum_i}'|U^L_{ie}|^2\ \ \ ,\ \ \
{\sum_i}'|U^R_{ie}|^2\ \ \ ,
\end{equation}
then reduces to the value unity, owing to the unitarity of the
corresponding leptonic CKM flavour mixing matrix\cite{Herczeg},
in which case one simply has $(v_e=1)$.
So-called Manifest Left-Right Symmetric Models (MLRSM) are those
LRSM such that the gauge coupling constants $g_R$ and $g_L$
and the quark and lepton CKM matrices in the left- and right-handed
sectors are equal, and such that
the CP violating phase
$\omega$ vanishes. Namely, these MLRSM are such that except for the different
masses for the charged $W$ and $W'$ and neutral $Z$ and $Z'$ gauge bosons,
the sectors of left- and right-handed chirality are indistinguishable,
and no CP violation originates in these models except for the
two ordinary Kobayashi-Maskawa phases appearing in
quark and lepton CKM matrices
associated to three generations of quarks and leptons.
In particular, parity invariance is then effectively restored
in these MLRSM for
processes of high momenta transfers.
The present purely experimental lower limit on the mass $M_2$
of a charged heavy gauge boson $W'$ in the context of these MLRSM
is\cite{CDF},
\begin{equation}
M_2\ >\ 652\ {\rm GeV}/c^2\ \ \ (95\%\ {\rm C.L.})\ \ .
\end{equation}
However, let us emphasize here again that the restrictions leading
to MLRSM are not assumed
in this note; our discussion applies to the most general LRSM possible.
Given this description of weak charged current interactions in
general LRSM,
it is straightforward to determine the corresponding
four-fermi effective Hamiltonian\cite{Herczeg,Lang}
at the quark-lepton level relevant to $\beta$-decay, which
is thus of the form,
\pagebreak
\begin{displaymath}
H^{\rm quark}_{\rm eff}\ ={\hspace{380pt}}
\end{displaymath}
\begin{equation}
\begin{array}{r l}
=&\eta_{LL}\overline{\psi}_u\gamma_\mu(1-\gamma_5)\psi_d\,
\overline{\psi}_e\gamma^\mu(1-\gamma_5)\psi_{\nu_e}\ +\
\eta_{LR}\overline{\psi}_u\gamma_\mu(1-\gamma_5)\psi_d\,
\overline{\psi}_e\gamma^\mu(1+\gamma_5)\psi_{\nu_e}\ + \\ \\
+&\eta_{RL}\overline{\psi}_u\gamma_\mu(1+\gamma_5)\psi_d\,
\overline{\psi}_e\gamma^\mu(1-\gamma_5)\psi_{\nu_e}+
\eta_{RR}\overline{\psi}_u\gamma_\mu(1+\gamma_5)\psi_d\,
\overline{\psi}_e\gamma^\mu(1+\gamma_5)\psi_{\nu_e}\ \ ,
\end{array}
\end{equation}
$\psi$ denoting the ordinary Dirac spinors for quarks and leptons.
The coefficients $\eta_{LL}$, $\eta_{LR}$, $\eta_{RL}$ and
$\eta_{RR}$ are given by
\begin{equation}
\begin{array}{r c l}
\eta_{LL}&=&\ \,\eta_0\,v_{LL}\,\Big(1+\delta t^2\Big)\ \ \ ,\\ \\
\eta_{LR}&=&-\,\eta_0\,v_{LR}\,r t\,e^{-i\omega}\Big(1-\delta\Big)\ \ \ ,\\ \\
\eta_{RL}&=&-\,\eta_0\,v_{RL}\,r t\,e^{i\omega}\Big(1-\delta\Big)\ \ \ ,\\ \\
\eta_{RR}&=&\ \,\eta_0\,v_{RR}\,r^2\Big(t^2+\delta\Big)\ \ \ ,
\end{array}
\end{equation}
with the notation
\begin{equation}
v_{LL}=V^L_{ud}\,{U^L_{ie}}^*\ \ ,\ \
v_{LR}=V^L_{ud}\,{U^R_{ie}}^*\ \ ,\ \
v_{RL}=V^R_{ud}\,{U^L_{ie}}^*\ \ ,\ \
v_{RR}=V^R_{ud}\,{U^R_{ie}}^*\ \ \ .
\end{equation}
In the particular case of MLRSM, these expressions agree of course
with those derived in Refs.\cite{Beg,Hol}.
At the nucleon level, the effective four-fermi Hamiltonian is
defined in terms of the couplings $C_V$, $C_V'$, $C_A$ and $C_A'$
introduced in the Appendix. Given the relations above, in LRSM these
coupling coefficients are thus determined to be
\begin{equation}
\begin{array}{r c l}
C_V &=&g_V\,\Big[\,\ \eta_{LL}\,+\,\eta_{LR}\,
+\,\eta_{RL}\,+\,\eta_{RR}\,\Big]\ \ ,\\ \\
C_V'&=&g_V\,\Big[\,\ \eta_{LL}\,-\,\eta_{LR}\,
+\,\eta_{RL}\,-\,\eta_{RR}\,\Big]\ \ ,\\ \\
C_A &=&g_A\,\Big[\,\ \eta_{LL}\,-\,\eta_{LR}\,
-\,\eta_{RL}\,+\,\eta_{RR}\,\Big]\ \ ,\\ \\
C_A'&=&g_A\,\Big[\,\ \eta_{LL}\,+\,\eta_{LR}\,
-\,\eta_{RL}\,-\,\eta_{RR}\,\Big]\ \ .\\ \\
\end{array}
\label{eq:coeffLRSM}
\end{equation}
With the help of the results listed in the Appendix, it is
then possible to compute the expression of
any observable of interest. In particular,
the asymmetry parameter $A$ and the quantity
$\Delta$ as defined in (\ref{eq:4Delta}) are given by, respectively,
\begin{equation}
A\ =\ \frac{\mp\lambda^2\lambda_{J'J}\,\Big[\,Z_+-X_+\,\Big]\ -\
2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}\,\Big[\,T+Y\,\Big]}
{\Big[\,Z_-+X_-\,\Big]\ +\ \lambda^2\,\Big[\,Z_++X_+\,\Big]}\ \ \ ,
\label{eq:ALRSM}
\end{equation}
and
\begin{displaymath}
\Delta\ =\ \frac{1}{2}\,\frac{1}
{\Big[(Z_- - X_-)+\lambda^2(Z_+ - X_+)\Big]\,
\Big[\mp\lambda^2\lambda_{J'J}(Z_+ - X_+)-2\delta_{J'J}\lambda
\sqrt{\frac{J}{J+1}}\,(T+Y)\Big]}\times
\end{displaymath}
\begin{equation}
\begin{array}{c l}
\times & \Bigg\{\,\mp\lambda^2\lambda_{J'J}
\Big[(X_+ Z_- +X_- Z_+)+2\lambda^2 X_+ Z_+\Big]\ -\\ \\
&\ -\ 2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}\,
\Big[(X_- T - Z_- Y)+\lambda^2(X_+ T - Z_+ Y)\Big]\Bigg\}\ \ \ .
\end{array}
\label{eq:DeltaLRSM}
\end{equation}
In these expressions, the parameter $\lambda$ is defined
in (\ref{eq:lambda}),
while the quantities $X_{\pm}$, $Y$, $Z_{\pm}$ and $T$ are
given by,
\begin{equation}
X_+\ =\ v_e r^2 t^2 (1-\delta)^2 + v_u v_e r^4 (t^2+\delta)^2 +
2\,{\rm Re}\left(v_{ud}e^{i\omega}\right) v_e r^3 t (1-\delta)
(t^2+\delta)\ \ \ ,
\label{eq:Xplus}
\end{equation}
\begin{equation}
X_-\ =\ v_e r^2 t^2 (1-\delta)^2 + v_u v_e r^4 (t^2+\delta)^2 -
2\,{\rm Re}\left(v_{ud}e^{i\omega}\right) v_e r^3 t (1-\delta)
(t^2+\delta)\ \ \ ,
\label{eq:Xminus}
\end{equation}
\begin{equation}
Y\ =\ v_e r^2 t^2 (1-\delta)^2 - v_u v_e r^4 (t^2+\delta)^2\ \ \ ,
\label{eq:Y}
\end{equation}
\begin{equation}
Z_+\ =\ (1+t^2\delta)^2 + v_u r^2 t^2 (1-\delta)^2 +
2\,{\rm Re}\left(v_{ud}e^{i\omega}\right) rt (1-\delta) (1+t^2\delta)\ \ ,
\label{eq:Zplus}
\end{equation}
\begin{equation}
Z_-\ =\ (1+t^2\delta)^2 + v_u r^2 t^2 (1-\delta)^2 -
2\,{\rm Re}\left(v_{ud}e^{i\omega}\right) rt (1-\delta) (1+t^2\delta)\ \ ,
\label{eq:Zminus}
\end{equation}
\begin{equation}
T\ =\ (1+t^2\delta)^2 - v_u r^2 t^2 (1-\delta)^2\ \ \ .
\label{eq:T}
\end{equation}
In particular for allowed pure GT transitions, one simply has
\begin{equation}
A_{|_{GT}}\ =\ {A_0}_{|_{GT}}\,\frac{Z_+-X_+}{Z_++X_+}\ \ \ ,\ \ \
\Delta_{|_{GT}}\ =\ \frac{1}{4}\,\Bigg[
\left(\frac{{A_0}_{|_{GT}}}{A_{|_{GT}}}\right)^2\,-\,1\Bigg]\ =\
\frac{X_+ Z_+}{\Big[Z_+ - X_+\Big]^2}\ \ \ ,
\end{equation}
${A_0}_{|_{GT}}$ being the asymmetry parameter in the SM for allowed pure GT
decays, given in (\ref{eq:Appendix.A0GT}) in the Appendix.
Another particular case of interest is obtained when the mixing
angle $\zeta$ vanishes identically, $(\zeta=0)$, for which one finds,
\begin{equation}
A_{|_{\zeta=0}}\ =\ A_0\ \frac{1-v_u v_e \, r^4\, \delta^2}
{1+v_u v_e \, r^4\, \delta^2}\ \ \ ,\ \ \
\Delta_{|_{\zeta=0}}\ =\ \frac{v_u v_e\,r^4\,\delta^2}
{\Big[\,1-v_u v_e\,r^4\,\delta^2\,\Big]^2}\ \ \ ,
\label{eq:Azeta0}
\end{equation}
$A_0$ being the asymmetry parameter in the SM
for arbitrary allowed $\beta$-decays,
given in (\ref{eq:Appendix.A0}) in the Appendix.
Note that in this particular case, the following identity happens
to be satisfied, independently of whether the allowed transition
is pure GT or not,
\begin{equation}
\Delta_{|_{\zeta=0}}\ =\ \frac{1}{4}\,\Bigg[
\left(\frac{A_0}{A_{|_{\zeta=0}}}\right)^2\,-\,1\,\Bigg]\ \ \ .
\end{equation}
Let us recall that these results, and in particular
the general ones in (\ref{eq:ALRSM}) and (\ref{eq:DeltaLRSM}),
are of application in the {\em most general\/}
LRSM possible, and do not assume that the parameters $\delta$
nor $(\tan\zeta)$ are small compared to the value unity. The expressions
given in (\ref{eq:ALRSM}) and (\ref{eq:DeltaLRSM})
for $A$ and $\Delta$ are {\em exact\/}, no approximation
whatsoever being implied at this stage (except for the fact that
recoil order corrections and possible charged Higgs contributions to the ratio
$R/R_0$ are neglected in the present discussion).
In order to gain some more insight into these general results, let us now
consider them in the particular case that no CP violation originates either
from the parameter $\omega$ nor from the ratio $v_{ud}$. Namely,
let us assume that the former parameter takes one of the
two values $(\omega=0)$ or $(\omega=\pi)$, and that the ratio
$v_{ud}$ is a real number, in which case
\begin{equation}
v_{ud}\,e^{i\omega}\ =\ \epsilon\,\sqrt{v_u}\ \ \ ,\ \ \
\epsilon\,=\,\pm\,1\ \ .
\end{equation}
Under these circumstances, it proves useful to define the quantities
\begin{equation}
x\ =\ \sqrt{v_u v_e}\, r^2 (\delta+t^2)\ -\
\epsilon\,\sqrt{v_e}\, r t\, (1-\delta)\ \ ,
\end{equation}
and
\begin{equation}
y\ =\ \sqrt{v_u v_e}\, r^2 (\delta+t^2)\ +\
\epsilon\,\sqrt{v_e}\, r t\, (1-\delta)\ \ ,
\end{equation}
as well as
\begin{equation}
\overline{x}\ =\ \delta t^2\ -\ \epsilon\,\sqrt{v_u}\, r t\, (1-\delta)\ \ ,
\end{equation}
and
\begin{equation}
\overline{y}\ =\ \delta t^2\ +\ \epsilon\,\sqrt{v_u}\, r t\, (1-\delta)\ \ .
\end{equation}
Indeed, one then observes that\footnote{The upper-script (CP) stands
for the fact that the expressions in the remainder of this section
are valid only when no CP violation originates either from $\omega$
nor from $v_{ud}$.}
\begin{equation}
X^{({\rm CP})}_+\ =\ y^2\ \ \ ,\ \ \
X^{({\rm CP})}_-\ =\ x^2\ \ \ ,\ \ \
Y^{({\rm CP})}\ =\ -\,x y\ \ ,
\end{equation}
as well as
\begin{equation}
Z^{({\rm CP})}_+\ =\ \left(1+\overline{y}\right)^2\ \ \ ,\ \ \
Z^{({\rm CP})}_-\ =\ \left(1+\overline{x}\right)^2\ \ \ ,\ \ \
T^{({\rm CP})}\ =\ \left(1+\overline{x}\right)\left(1+\overline{y}\right)\ \ \
{}.
\end{equation}
Therefore, under the assumptions stated above concerning $\omega$ and
$v_{ud}$, the asymmetry parameter $A$ and the quantity $\Delta$ reduce to,
respectively,
\begin{equation}
A^{({\rm CP})}\ =\ \frac{\mp\lambda^2\lambda_{J'J}
\left[(1+\overline{y})^2-y^2\right]\ -\ 2\delta_{J'J}\lambda
\sqrt{\frac{J}{J+1}}\left[(1+\overline{x})(1+\overline{y})-xy\right]}
{\left[(1+\overline{x})^2+x^2\right]\ +\ \lambda^2
\left[(1+\overline{y})^2+y^2\right]}\ \ \ ,
\end{equation}
and
\begin{displaymath}
\Delta^{({\rm CP})}\ =\ \frac{1}{2}\,
\frac{1}{\Big[\,(1+\overline{x})^2-x^2+\lambda^2
\Big((1+\overline{y})^2-y^2\Big)\Big]}\times
\end{displaymath}
\begin{displaymath}
\times\,\frac{1}{\Big[\,\mp \lambda^2\lambda_{J'J}
\Big((1+\overline{y})^2-y^2\Big)-2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}\,
\Big((1+\overline{x})(1+\overline{y})-xy\Big)\Big]}\times
\end{displaymath}
\begin{equation}
\begin{array}{c l}
\times & \Bigg\{\,\mp\lambda^2\lambda_{J'J}\Big[\,
x^2(1+\overline{y})^2+y^2(1+\overline{x})^2+2\lambda^2 y^2
(1+\overline{y})^2\,\Big]\ -\\
&\ -\ 2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}\,
\Big[\,x(1+\overline{y})+y(1+\overline{x})\,\Big]\,
\Big[\,x(1+\overline{x})+\lambda^2 y(1+\overline{y})\,\Big]\,\Bigg\}\ \ \ .
\end{array}
\end{equation}
In particular for allowed pure GT transitions, these results simplify to
\begin{equation}
A^{({\rm CP})}_{|_{GT}}\ =\ {A_0}_{|_{GT}}\ \frac{(1+\overline{y})^2-y^2}
{(1+\overline{y})^2+y^2}\ \ \ ,
\end{equation}
as well as
\begin{equation}
\Delta^{({\rm CP})}_{|_{GT}}\ =\ \frac{y^2(1+\overline{y})^2}
{\Big[(1+\overline{y})^2-y^2\Big]^2}\ =\ \frac{1}{4}\,\Bigg[\,
\left(\frac{{A_0}_{|_{GT}}}{A^{({\rm CP})}_{|_{GT}}}\right)^2\,-\,1\,
\Bigg]\ \ \ ,
\end{equation}
${A_0}_{|_{GT}}$ being the value of the asymmetry parameter in the SM
for allowed pure GT $\beta$-decays.
The above expressions generalise those obtained in Ref.\cite{Quin}
in the case of MLRSM for which $(r=1)$, $(v_u=1=v_e)$ and $(\omega=0)$ and
in the limit that both $\delta$ and $(t=\tan\zeta)$ are much
smaller than unity.
In contradistinction, the results derived in this
note are valid for {\em arbitrary\/}
LRSM parameters, independently of such or any other approximations
(except for recoil order corrections and possible charged Higgs contributions
to the ratio
$R/R_0$ which are ignored in the present discussion).
Nevertheless, to conclude let us consider
the limit in which both $\delta$ and
$(t=\tan\zeta)$ are indeed much smaller than unity,
still under the assumption that the parameter
$\left(v_{ud}e^{i\omega}\right)$
is real. Given the definitions,
\begin{equation}
\tilde{\delta}\ =\ \sqrt{v_u v_e}\, r^2 \delta\ \ \ ,\ \ \
\tilde{t}\ =\ \epsilon\sqrt{v_e}\, r t\ \ \ ,
\label{eq:tildedeltat}
\end{equation}
to first order in the quantities $\delta$ and $(t=\tan\zeta)$,
one then finds
\begin{equation}
x\ \simeq\ \tilde{\delta}\ -\ \tilde{t}\ \ \ ,\ \ \
y\ \simeq\ \tilde{\delta}\ +\ \tilde{t}\ \ \ ,\ \ \
\label{eq:tildexy}
\end{equation}
as well as
\begin{equation}
\overline{x}\ \simeq\ -\sqrt{\frac{v_u}{v_e}}\,\tilde{t}\ \ \ ,\ \ \
\overline{y}\ \simeq\ +\sqrt{\frac{v_u}{v_e}}\,\tilde{t}\ \ \ .
\label{eq:tildexybar}
\end{equation}
In other words, within the approximation that
$(\delta\ll 1)$ and $(\tan\zeta\ll 1)$, both $A^{({\rm CP})}$ and
$\Delta^{({\rm CP})}$
are determined by quadratic expressions in terms of the parameters
$\tilde{\delta}$ and $\tilde{t}$.
One then has,
\begin{equation}
A^{({\rm CP})}_{|_{\delta,t\ll 1}}\ \simeq\ \frac{\mp\lambda^2\lambda_{J'J}
\left[(1+\overline{y})^2-y^2\right]\ -\ 2\delta_{J'J}\lambda
\sqrt{\frac{J}{J+1}}
\left[(1+\overline{x})(1+\overline{y})-xy\right]}
{\left[(1+\overline{x})^2+x^2\right]\ +\ \lambda^2
\left[(1+\overline{y})^2+y^2\right]}\ \ \ ,
\label{eq:Alinearised}
\end{equation}
as well as
\begin{equation}
\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}\ \simeq\ \frac{1}{2}\,\Bigg\{
\frac{\mp \lambda^2\lambda_{J'J} y^2\ -\
2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}\,xy}
{\mp \lambda^2\lambda_{J'J}\ -\
2\delta_{J'J}\lambda\sqrt{\frac{J}{J+1}}}\ +\
\frac{x^2+\lambda^2 y^2}{1+\lambda^2}\Bigg\}\ \ \ ,
\label{eq:Deltalinearised}
\end{equation}
with $x$, $y$, $\overline{x}$ and $\overline{y}$
now given in (\ref{eq:tildexy}) and (\ref{eq:tildexybar}), provided that
$(\delta\ll 1)$ and $(\tan\zeta\ll 1)$. Within these approximations
and under the assumptions that $(\omega=0)$ and that $v_{ud}$ is
real, the expression (\ref{eq:Deltalinearised})
for $\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$
in terms of the parameters $x$ and $y$
coincides precisely with the one following from
Ref.\cite{Quin} within the context of MLRSM. In other words,
given the approximations $(\delta\ll 1)$ and $(\tan\zeta\ll 1)$
and the restriction that $\left(v_{ud}e^{i\omega}\right)$ is
real, the result obtained\cite{Quin}
for $\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$
in the context of MLRSM remains
valid for general LRSM provided the parameters $\delta$
and $(t=\tan\zeta)$ are simply replaced by the parameters $\tilde{\delta}$
and $\tilde{t}$ defined in (\ref{eq:tildedeltat}), respectively.
The quadratic form obtained in (\ref{eq:Alinearised})
for the asymmetry parameter $A^{({\rm CP})}_{|_{\delta,t\ll 1}}$
in terms of the parameters $(v_u/v_e)$, $\tilde{\delta}$ and $\tilde{t}$
has been analysed in Ref.\cite{Oscar} already, albeit in the context
of MLRSM, namely when $(v_u=1=v_e)$,
$(\tilde{\delta}=\delta)$ and $(\tilde{t}=t)$.
For the quantity $\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$
in (\ref{eq:Deltalinearised}),
the corresponding quadratic form is characterised by the relation
\begin{equation}
\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}\ \simeq\
\tilde{\delta}^2\ +\ 2\,\Delta_{\tilde{t}\tilde{\delta}}\,
\tilde{t}\tilde{\delta}\ +\ \Delta_{\tilde{t}\tilde{t}}\,\tilde{t}^2\ \ ,
\label{eq:quadraticDelta}
\end{equation}
with coefficients $\Delta_{\tilde{t}\tilde{\delta}}$ and
$\Delta_{\tilde{t}\tilde{t}}$ which may easily be determined
from (\ref{eq:Deltalinearised}). Therefore, any experimental
upper limit or value $\Delta_0$ for
the quantity $\Delta$ would determine an elliptic
or hyperbolic (exclusion) contour in the plane $(\tilde{t},\tilde{\delta})$,
under the conditions for which the result in (\ref{eq:Deltalinearised})
is applicable.
In particular, the slope of this contour at $(\tilde{t}=0)$,
corresponding to a vanishing mixing angle $(\zeta=0)$, is simply
given by the coefficient $(-\Delta_{\tilde{t}\tilde{\delta}})$, namely,
\begin{equation}
\left(\frac{d\tilde{\delta}}{d\tilde{t}}\right)_{|_{\zeta=0}}\ =\ -\
\Delta_{\tilde{t}\tilde{\delta}}\ \ .
\label{eq:slope1}
\end{equation}
On the other hand, the hyperbolic or elliptic character of the
contour plot is simply determined from the sign of the quantity
\begin{equation}
\Delta_{\tilde{t}\tilde{t}}\ -\
\Delta_{\tilde{t}\tilde{\delta}}^2\ \ .
\end{equation}
When this sign is positive, the contour is elliptic;
when it is negative, the contour is hyperbolic; and when this
quantity vanishes identically, the contour is a straight line.
The latter instance applies in particular to allowed pure GT decays,
since $\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$ then reduces precisely to
$\left(y^2\simeq (\,\tilde{\delta}+\tilde{t}\,)^2\right)$.
\section{The case for mirror nuclei}
\label{Sect4}
Let us now apply the results of the previous general developments
to specific super-allowed $\beta$-decays, namely mirror nuclei.
The interest of this choice
lies with the fact that the initial and daughter nuclei
being members of a single isospin multiplet,
one may be quite confident in the determination
of recoil order corrections\cite{Holstein}
to the ratio $R/R_0$
for such nuclei using both experimental information and theoretical
considerations such as CVC and PCAC.
The list of mirror nuclei considered here\cite{Oscar}
is given in Table \ref{Table1},
with in the second column the spin sequence $(J,J')$.
Since for mirror nuclei one has $(J'=J)$, only the common value
of $J$ is indicated. Table \ref{Table1} also includes the two
allowed pure GT transitions of $^{107}$In and $^{12}$N, for
comparison. The third and fourth columns of the Table
give the values for the quantities $\lambda$ and $A_0$ introduced
previously, as evalued\footnote{The data for the neutron
are from Ref.\cite{PDG}, while a change of sign for the parameter
$\lambda$\cite{Budick} as compared to the one given in Ref.\cite{Oscar},
and the ensuing modification of the value for
the asymmetry
parameter $A_0$, are effected in the case of $^3$H.
The sign of $\lambda$ relative to that of the neutron
for all cases listed in Table~\ref{Table1} agrees with the
results given in Ref.\cite{Raman} on basis of the shell model.}
in Ref.\cite{Oscar}.
The fifth column gives the end-point total
energy $E_0$, thus including the electron (positron) rest-mass $m$,
while the sixth column lists the corresponding value $\beta_0$ of
the velocity of the $\beta^{\mp}$ particle.
The meaning of the last two columns of Table \ref{Table1} is
detailed below. Note that
except for the first two entries,
namely the neutron and $^3$H,
all nuclei listed in Table~\ref{Table1}
decay by positron emission.
In order to assess the potentiality of a given nucleus as to its
sensitivity to right-handed current contributions through a
relative longitudinal polarisation-asymmetry correlation measurement, given
the ideal experimental conditions assumed in this note---namely a
measurement of exactly the longitudinal polarisation at a specific
energy $E$ and momentum direction $\hat{p}$
of the $\beta^{\mp}$ particle---, the
following strategy is applied. The ratio $R/R_0$
is determined experimentally with a certain precision, and is
established either not to differ from the value unity by more than
that precision $\epsilon_0$, or to differ from the value unity
by a value $\epsilon_0$. In other words, in either case one may
write\footnote{In the case of the ratio $R(-J,J)/R_0(-J,J)$,
strictly speaking
the expression in (\ref{eq:limit1}) in fact ignores an additional
correction factor dependent on $\Delta$, $\beta$ and
$A_{\rm exp}(-J,J)$, given in (\ref{eq:RR02}). However, for the purpose
of the discussion of the present section, any correction brought about
by that factor may safely be ignored, since $\Delta$ is certainly
small in comparison to the value unity.}
\begin{equation}
\mid\, 1\ -\ \frac{R}{R_0}\,\mid\ =\ \mid\,k\,\Delta\,\mid\
\le\ \epsilon_0\ \ \ ,
\label{eq:limit1}
\end{equation}
where $k$ is one of the enhancement factors $k(-J,0)$ or $k(-J,J)$
depending on whether it is $R(-J,0)/R_0(-J,0)$ or $R(-J,J)/R_0(-J,J)$
which is measured. In the general case,
the quantity $\Delta$ given in (\ref{eq:DeltaLRSM})
is a rather complicated
function of the LRSM parameters $\delta$, $t$,
${\rm Re}\left(v_{ud}e^{i\omega}\right)$, $v_u$, $v_e$ and $r$, and as such
a characterisation of the potentiality of a given nucleus in terms
of the quantity $\Delta$ may not be very indicative of the new
physics it would imply. Rather, it seems more efficient to characterise
this potentiality in terms of the mass range for the $W'$ mass $M_2$
one may hope to reach with this type of experiment. For this purpose,
it is appropriate to consider the expression of the quantity $\Delta$
for a vanishing mixing angle $\zeta$, namely,
\begin{equation}
\Delta_{|_{\zeta=0}}\ =\ \frac{\tilde{\delta}^2}
{\Big[\,1-\tilde{\delta}^2\,\Big]^2}\ \ \ ,\ \ \
\tilde{\delta}^2\ =\ v_u v_e\,r^4\,\delta^2\ =\
v_u v_e\,r^4\,\frac{M_1^4}{M_2^4}\ \ ,
\end{equation}
which is valid in the most general LRSM possible (see (\ref{eq:Azeta0})\,).
Since $\tilde{\delta}^2$ is positive, one always has,
\begin{equation}
\tilde{\delta}^2\ \le\ \frac{\tilde{\delta}^2}
{\Big[\,1-\tilde{\delta}^2\,\Big]^2}\ =\ \Delta_{|{\zeta=0}}\ \ \ ,
\end{equation}
which upon substitution of the upper bound or value for
$\Delta$ given in (\ref{eq:limit1}), leads to
the following lower limit on $M_2$,
\begin{equation}
M_2\ \ge\,|r|\,\left(v_u v_e\right)^{1/4}\,M_{\rm min}\ \ \ ,
\label{eq:limit2}
\end{equation}
where the mass-reach $M_{\rm min}$ is defined as,
\begin{equation}
M_{\rm min}\ =\ \left(\frac{|k|}{\epsilon_0}\right)^{1/4}\ M_1\ \ ,
\end{equation}
$(M_1=80.22\ {\rm GeV}/c^2)$ being the mass of the ordinary gauge boson $W$.
Therefore, the potentiality of a given nucleus is to be characterised
in terms of the quantity $M_{\rm min}$, evaluated for either of the
two measurements considered here, namely corresponding
to the ratios $R(-J,0)/R_0(-J,0)$ or
$R(-J,J)/R_0(-J,J)$. Note that $M_{\rm min}$ represents precisely the lower
bound to be obtained for $M_2$ in the context of MLRSM, in the
limit that $\delta$ is much smaller than the value unity. For general
LRSM however, this is not the case and the potential lower bound on $M_2$
is obtained by multiplying the mass-reach $M_{\rm min}$ by the
factor $\Big(|r|(v_u v_e)^{1/4}\Big)$.
Clearly, the evaluation of the mass-reach $M_{\rm min}$
requires the value of the enhancement factor $k$,
which in turn needs the values for the $\beta^{\mp}$ particle velocity
$\beta$ and the experimental asymmetry $A_{\rm exp}$. The choice of value
for $\beta$ is conditioned by the necessity of high statistics measurements,
which makes it is preferable to work at the maximum of the energy distribution
$W_0(E)$. When ignoring the Coulomb correction
represented by the Fermi function $F(\pm Z,E)$, this maximum
is reached at an energy $E_{\rm max}$ whose value is given
in (\ref{eq:Emax}) of the Appendix
in terms of the end-point total energy $E_0$.
The values of $E_{\rm max}$ as well as the associated value for
the velocity $\beta_{\rm max}$ are listed in the last two columns
of Table \ref{Table1}. Thus, $\beta_{\rm max}$ is the value at which
we choose to evaluate the enhancement factors $k(-J,0)$ and $k(-J,J)$.
Accounting for Coulomb corrections through Fermi's function
$F(\pm Z,E)$ would not modify the value for $\beta_{\rm max}$
significantly, since $\beta$-decay energy spectra are typically rather
smooth around their maximum and the effect of the
shift in $\beta_{\rm max}$ due to Fermi's function is small
for small values of $Z$, as is the case for the nuclei considered
here. In any case, from the practical
experimental point of view, this issue is rather academic since
any energy acceptance is always of finite resolution.
As shown in Sect.\ref{Sect2}, the enhancement factors $k(-J,0)$
and $k(-J,J)$ are the largest when the experimental asymmetries
$A_{\rm exp}(-J,0)$ and $A_{\rm exp}(-J,J)$
approach the values $\beta^2$ and $2\beta^2/(1+\beta^2)$,
respectively.
Given the value $\beta_{\rm max}$ chosen here, these optimal
experimental asymmetries $A^{(0)}_{\rm exp}$
are listed in the last two columns of
Table \ref{Table2}. Note that with the exception of $^3$H, these values
are rather large and thus imply the requirement of the largest
degree of nuclear polarisation possible, as is indeed to be expected.
The other data in Table \ref{Table2} give
on the one hand, the
coefficients $\Delta_{\tilde{t}\tilde{\delta}}$ and
$\Delta_{\tilde{t}\tilde{t}}$ defining the quadratic form
in (\ref{eq:quadraticDelta}) which determines the quantity
$\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$
for values of $\delta$ and $(t=\tan\zeta)$ small compared
to the value unity and when $(v_{ud}e^{i\omega})$ is real,
and on the other hand, the type of
curve so obtained, ``E", ``H" and ``L" standing for
an elliptic, hyperbolic or linear curve, respectively.
In particular, the coefficient $(-\Delta_{\tilde{t}\tilde{\delta}})$
determines the slope $(d\tilde{\delta}/d\tilde{t})_{|_{\zeta=0}}$
at $(\zeta=0)$ of the dependence $\tilde{\delta}(\tilde{t})$
determined by $\Delta^{({\rm CP})}_{|_{\delta,t\ll 1}}$
when both $\delta$ and $(t=\tan\zeta)$
are small in comparison to the value unity;
this slope thus characterises the sensitivity
of the measurement of $\Delta$ to values of $\zeta$ different from zero
when $\delta$ is small compared to unity.
Incidentally, note that the data listed in Tables \ref{Table1}
and \ref{Table2}
are independent both of the experimental precision $\epsilon_0$ and
of the experimental asymmetries $A_{\rm exp}(-J,0)$ and
$A_{\rm exp}(-J,J)$---namely
the attainable degree of nuclear polarisation $J$---, to which
we now turn.
The experimental asymmetries
$A_{\rm exp}(-J,0)$ and $A_{\rm exp}(-J,J)$
are determined in terms of the quantity
$\Big(\beta\,J\,A\,\left(\hat{p}.\hat{J}\right)\,\Big)$.
Besides the quantity $\beta$ whose value has now been specified
to be $\beta_{\rm max}$,
one also requires the asymmetry parameter $A$ and the {\em effective\/}
degree of nuclear polarisation
$\left({\cal P}=|J\,(\hat{p}.\hat{J})|\right)$.
For the sake of the present evaluation, it is obviously justified to
approximate the former quantity by the asymmetry parameter $A_0$
in the SM, which is listed\cite{Oscar} in Table~\ref{Table1}. Indeed,
even though this approximation ignores possible right-handed contributions,
the latter are certainly small and may effectively be accounted
for through a small rescaling of the degree of nuclear
polarisation $J$.
Finally, the effective degree of nuclear polarisation is characterised
by the quantity $\Big({\cal P}=|J(\hat{p}.\hat{J})|\Big)$,
with the relative directions of $\hat{J}$ and $\hat{p}$
chosen such that the experimental
asymmetries $A_{\rm exp}(-J,0)$ and $A_{\rm exp}(-J,J)$ as defined
in this note are {\em positive\/}.
For the comparison of the potentiality offered by the nuclei considered
here, the following values for the experimental precision
$\epsilon_0$ on the measurement of $R/R_0$ and for the effective
degree of nuclear polarisation ${\cal P}$ are used,
\begin{equation}
\epsilon_0\ =\ 0.01\ \ \ ,\ \ \
{\cal P}\ =\ 0.80\ \ \ ,
\end{equation}
corresponding in fact to rather stringent experimental
requirements and achievements.
The corresponding results for the enhancement factor $k$,
the experimental asymmetry $A_{\rm exp}$ and the mass-reach
$M_{\rm min}$ are listed in Table \ref{Table3}, for both types
of configurations of nuclear polarisation
considered in this note.
The values of $M_{\min}$ in Table \ref{Table3} reveal that among
mirror nuclei, those with the best prospects with regards to our
purpose are $^{17}$F, $^{41}$Sc and $^{25}$Al, in order of
decreasing potentiality. These nuclei are also those for which
the enhancement factors are the largest, and for which the
asymmetry parameter $A_0$ is closest to the maximal value
of unity attained for allowed pure GT transitions. Indeed,
these three mirror nuclei compete well with the two
examples of the latter type of decay, namely $^{107}$In and
$^{12}$N. In fact, there are three factors---related to
one another---which concur
to explain the distinguished role of these three mirror nuclei:
a large asymmetry parameter $A_0$, allowing an experimental
asymmetry $A_{\rm exp}$ close to its optimal value $A_{\rm exp}^{(0)}$,
hence a large enhancement factor.
In spite of the large effective degree of nuclear polarisation
$({\cal P}=0.80)$ assumed here, experimental asymmetries
$A_{\rm exp}$ are still less than their optimal values
$A_{\rm exp}^{(0)}$ for all nuclei considered, given the
present choice for $\beta$, namely $(\beta=\beta_{\rm max})$.
The values for $A^{(0)}_{\rm exp}$ are quite close to unity---except
for $^3$H---simply because the values for $\beta_{\rm max}$
are also quite close to the maximal value of unity.
Indeed, as was pointed out in Sect.\ref{Sect2}, the optimal sensitivity
is achieved for values of $\beta$ and of the effective degree of
nuclear polarisation ${\cal P}$ such that $\Big(\beta=|A{\cal P}|\Big)$,
which is not possible for asymmetry parameters $|A|$ less than unity
and an effective nuclear polarisation $({\cal P}=0.80)$, once the value
of $\beta$ is specified to be $\beta_{\rm max}$.
Therefore for all nuclei considered, given this value
for $\beta$, an increased sensitivity
to right-handed charged currents requires a larger enhancement
factor, hence a larger effective degree of nuclear polarisation ${\cal P}$.
Independently of the technical feasibility of the production
and polarisation of the mirror nuclei $^{17}$F, $^{41}$Sc
and $^{25}$Al---as well as of $^{107}$In\cite{107In} and
$^{12}$N\cite{12N}---with
sufficient yields and degrees of polarisation, it thus appears that
a mass-reach $M_{\rm min}$ of the order of $600$ GeV/c$^2$
is the ultimate limit attainable
using relative longitudinal polarisation-asymmetry correlation measurements
in allowed nuclear $\beta$-decays. Note that all these cases correspond
to positron emitters, for which well established precision
polarimetry techniques are readily available\cite{107In,12N}.
However, let us remark that this conclusion which is established on quite
general grounds leaves open two possible types of loopholes. On the
one hand, there exist specific mirror nuclei for which
efficient polarisation techniques are becoming available, possibly
reaching the ideal degree of polarisation of 100\%. The above analysis
has then to be reconsidered separately for such particular cases.
On the other hand, but then at the cost of a loss of statistics,
there also remains the possibility to work at
values of $\beta$ smaller than $\beta_{\rm max}$, in order to
reach more easily the optimal sensitivity attained when $(\beta=|A{\cal P}|)$
given a certain effective degree of nuclear polarisation ${\cal P}$
achieved in practice.
The first possibility is realised for example in the cases of
$^{21}$Na\cite{Freedman} and
$^{37}$K\cite{Deutsch}. To illustrate the point,
Table~\ref{Table4} lists the same information
as Table~\ref{Table3} for a precision $(\epsilon_0=0.01)$ but
for an effective degree of nuclear
polarisation taking the maximal value possible $({\cal P}=1.00)$.
Note that $^{17}$F, $^{41}$Sc and $^{25}$Al then still remain the favorite
cases among mirror nuclei, but now with a different order of interest.
This is due to the choice of the $\beta^{\mp}$ particle energy at $E_{\rm
max}$,
which is such that for the first two cases the experimental asymmetries
$A_{\rm exp}$ are {\em larger\/} than their optimal values $A^{(0)}_{\rm exp}$.
Indeed, the enhancement factors $k(-J,0)$ and $k(-J,J)$ are now negative
for $^{17}$F and $^{41}$Sc, whereas those for $^{25}$Al remain positive
but have become large. The same applies also to the pure GT transitions
of $^{107}$In and $^{12}$N. In other words, in these specific cases,
a choice of $\beta$ slightly less than $\beta_{\rm max}$ may lead
to large enhancement factors indeed, at no significant loss
in statistics. Nevertheless, this assumes the maximal possible
effective degree of nuclear polarisation $({\cal P}=1.00)$, quite a unique
experimental circumstance. For example, even though the case of $^{12}$N
may appear from Table~\ref{Table4}
to be the most attractive with a mass-reach of
1.7~TeV/c$^2$, an effective nuclear polarisation
of $({\cal P}\simeq 0.15)$ only is obtained\cite{12N} in practice.
Therefore,
given a technically achievable effective degree of nuclear polarisation
${\cal P}$ for a specific mirror nucleus,
the other avenue open towards large enhancement factors is to consider working
at values of $\beta$ lower than $\beta_{\rm max}$.
The ensuing loss in statistics has then to be weighted
against the possibly important gain in sensitivity, but such
an evaluation is possible only on a case by case basis in contradistinction
to the general considerations of this note. However, as was remarked
previously, the $\beta$ particle count rate for the direction of
nuclear polarisation offering the greatest sensitivity to right-handed
currents is proportional to $(1-\beta^2)$ when the optimal choice
of values for $(\beta,{\cal P})$ such that $(\beta=|A{\cal P}|)$ is made.
Thus, even though the overall statistics may decrease by choosing
a value for $\beta$ smaller than $\beta_{\rm max}$ in order
to achieve a sensitivity closer to the optimal situation,
the relative statistics measured
for the direction of nuclear polarisation most sensitive to the
sought-for physical effect will increase.
One particular case which is to be distinguished from that point of view
is that of $^3$H, with a maximum value of the $\beta^-$ particle
velocity at $(\beta_0=0.2626)$, in spite of the rather small
asymmetry parameter of $(A_0=-0.09405)$. Since the corresponding value
of $(\beta_{\rm max}=0.1208)$ is quite small, working at a value of $\beta$
such that the optimal configuration $(\beta=|A{\cal P}|)$ is achieved
should not lead to any significant loss of statistics, provided
an effective degree of nuclear polarisation of at least $({\cal P}=0.80)$
can be achieved. For example, given an
effective nuclear polarisation $(0.80\le{\cal P}\le 1.00)$, the
optimal choice for $\beta$ lies in the interval
$(0.0752\le\beta\le 0.0941)$,
which is not much less than the value $(\beta_{\rm max}=0.1208)$.
Under such circumstances, given sufficient energy
resolution\footnote{A poor energy
resolution dilutes enhancement factors which otherwise could reach
very large values.}, quite large enhancement factors may be
expected, possibly opening up a mass-reach in the TeV/c$^2$ region.
Nevertheless, such instances of relative longitudinal polarisation-asymmetry
correlation measurements can be assessed on a case by case basis only.
\section{Conclusion}
\label{Sect5}
In this note, the sensitivity\cite{Quin}
of relative polarisation-asymmetry correlation $\beta$-decay experiments
to charged weak current interactions of right-handed chirality
is con\-si\-de\-red independently of any specific model for physics
beyond the Standard Model of the electroweak interactions.
Starting with the general results of Ref.\cite{JTW}
based on a four-fermi effective Hamiltonian for allowed $\beta$-decay
including arbitrary complex vector and axial coefficients
$C_V$, $C_V'$, $C_A$ and $C_A'$ only, and ignoring recoil
order corrections expected to be small for super-allowed
decays, it is shown that this class
of experiments is directly sensitive to physics beyond the SM
through a certain combination
of these four parameters which is
characterised by a single quantity $\Delta$
given in (\ref{eq:4Delta}).
A non vanishing value for $\Delta$ would establish the existence
of charged right-handed currents and thus
of new physics beyond the SM.
These general considerations are then developed further in the
specific case of so-called Left-Right Symmetric Models in their
most general form possible, assuming only that possible
charged Higgs contributions are negligible.
Which combinations of the fundamental
parameters of such LRSM are probed through the class of measurements
mentioned above is made explicit, in particular in terms
of the asymmetry parameter $A$ and the quantity $\Delta$ in
(\ref{eq:ALRSM}) and (\ref{eq:DeltaLRSM}),
respectively. These expressions, which do not
involve any simplifying restriction nor approximation whatsoever,
are also considered for restricted classes of LRSM in which no
CP violation originates from a lack of
complete complex phase alignment between
the two sectors of opposite chiralities in such theories, neither
in the Higgs nor in the Yukawa sectors.
In particular, the as\-so\-cia\-ted
results generalise those obtained previously\cite{Quin}
in the context of so-called Manifest Left-Right Symmetric Models---which
provide but one type of a very restricted class of LRSM---under
the approximation that both the ratio of
the squared masses of light to heavy charged
gauge bosons $W$ and $W'$ as well as their mixing angle be much
smaller than the value unity.
These general results are then applied specifically to the case
of mirror nuclei, which offer the advantage that recoil order corrections
are more amenable to sufficiently precise evaluation than for
other instances of allowed $\beta$-decay, since initial
and daughter nuclei then belong to the same isospin multiplet.
The potentiality of these mirror nuclei as to the sensitivity
to contributions from charged currents of right-handed chirality
is then characterised in terms of the mass-reach---the
precise technical meaning of this notion in the context
of general LRSM is defined in Sect.\ref{Sect4}---for
the hypothetical $W'$ charged gauge boson which may be achievable
by using each of these nuclei,
given a certain experimental precision and
degree of nuclear polarisation. The analysis es\-ta\-bli\-shes that among
mirror decays, the cases of $^{17}$F, $^{41}$Sc and $^{25}$Al,
in order of decreasing interest,
certainly offer the best prospects, which are comparable to those
achieved by on-going
experiments using the allowed pure Gamow-Teller transitions of
$^{107}$In\cite{107In} and $^{12}$N\cite{12N}.
Indeed, allowed pure Gamow-Teller decays are expected
to provide the best mass-reach possible owing to their
maximal asymmetry parameter.
The analysis is performed on quite general grounds, not paying
attention to specific circumstances which may apply
to a given particular nucleus, nor to the technical feasibility
of the production and polarisation of these nuclei.
In fact, the sensitivity to right-handed
charged current contributions of the type of experiment considered
here may become quite large for some special cases, by appropriatedly
choosing to work at a specific value of the
$\beta^\mp$ particle energy, given an achievable effective
degree of nuclear polarisation. It is then not excluded that
some particular mirror nucleus presents the potential to extend
the mass-reach of relative longitudinal polarisation-asymmetry
correlation measurements into the TeV/c$^2$ region. One such
instance which may be worth pursuing further could be that of
the $\beta^-$-decay of $^3$H, owing to the rather low value
of the end-point energy in that case.
This conclusion also opens the prospect that for specific
values of the $\beta^\mp$ particle energy and of the nuclear polarisation,
other observables
in the $\beta$-decay of mirror nuclei
offer a similarly large sensitivity to other
couplings appearing\cite{JTW} in the effective Hamiltonian
for nuclear $\beta$-decay, including for example scalar
or tensor contributions, as well as time reversal violating
effects\cite{Deutsch}. Such possibilities certainly deserve to be
investigated in detail, along lines similar to those developed here.
\section*{Acknowledgement}
Drs. O. Naviliat-Cuncic and
N. Severijns are gratefully acknowledged for useful information
concerning the value of $\lambda$
in the case of $^3$H.
\pagebreak
\section*{Appendix}
\label{Appendix}
In the case of vector ($V$) and axial ($A$) contributions only,
as assumed in this note, the
general four-fermi effective Hamiltonian considered in Ref.\cite{JTW}
at the nucleon level is of the form
\begin{equation}
H^{\rm nucleon}_{\rm eff}\ =\ \overline{\psi}_{p}\gamma_\mu\psi_n\,
\overline{\psi}_e\,\Big(C_V\gamma^\mu-{C_V}'
\gamma^\mu\gamma_5\Big)\psi_{\nu_e}
\ -\ \overline{\psi}_p\gamma_\mu\gamma_5\psi_n\,
\overline{\psi}_e\,\Big(C_A\gamma^\mu\gamma_5-{C_A}'\gamma^\mu\Big)\psi_{\nu_e}
\ \ ,
\label{eq:Heff1}
\end{equation}
where $C_V$, $C_V'$, $C_A$ and $C_A'$ are {\em a priori\/} arbitrary complex
coefficients, and $\psi_p$, $\psi_n$, $\psi_e$ and $\psi_{\nu_e}$
represent Dirac spinors for the proton, the neutron,
the electron and the neutrino
of electronic flavour, respectively. Our phase conventions are as follows.
The chirality operator $\gamma_5$ is defined with a sign such that
left-handed couplings are of the form $\gamma_\mu(1-\gamma_5)$,
which is the choice opposite to that taken in Ref.\cite{JTW}.
To account for that difference, changes of sign have been included in
the expression in (\ref{eq:Heff1}) in such a way that the coefficients
$C_V$, $C_V'$, $C_A$ and $C_A'$ are as defined in Ref.\cite{JTW}.
The quantities $W_0(E)$, $\xi$, $A$, $G$ and $N'$ appearing in
(\ref{eq:Dist1}) are then given by the following expressions\cite{JTW},
\vspace{10pt}
\begin{equation}
\xi=M_F^2\Big[\,|C_V|^2+|C_V'|^2\,\Big]\ +\
M_{GT}^2\Big[\,|C_A|^2+|C_A'|^2\,\Big]\ \ ,
\end{equation}
\vspace{10pt}
\begin{equation}
\begin{array}{r c l}
\xi A&=&M_{GT}^2\,\lambda_{J'J}\Big[\,\mp 2\,{\rm Re}
\left(C_A {C_A'}^*\right)\,\Big]\ +\ \\ \\
&+& \delta_{J'J}M_F M_{GT}\,\sqrt{\frac{J}{J+1}}\,
\Big[\,-2\,{\rm Re}\left(C_V{C_A'}^*+C_V'{C_A}^*\right)\,\Big]\ \ ,
\end{array}
\end{equation}
\vspace{10pt}
\begin{equation}
\xi G=M_F^2\,\Big[\,\mp\,2\,{\rm Re}
\left(C_V{C_V'}^*\right)\,\Big]\ +\ M_{GT}^2\,
\Big[\,\mp\,2\,{\rm Re}\left(C_A{C_A'}^*\right)\,\Big]\ \ ,
\end{equation}
\vspace{10pt}
\begin{equation}
\begin{array}{r c l}
\xi N'&=&M_{GT}^2\,\lambda_{J'J}\,\Big[\,
|C_A|^2+|C_A'|^2\,\Big]\ +\ \\ \\
&+& 2\delta_{J'J}\,M_F M_{GT}\,
\sqrt{\frac{J}{J+1}}\,\Big[\,\pm{\rm Re}
\left(C_V{C_A}^*+C_V'{C_A'}^*\right)\,\Big]\ \ ,
\end{array}
\end{equation}
\vspace{10pt}
and finally
\begin{equation}
W_0(E)\ =\ \frac{1}{(2\pi)^4}\,p\,E(E_0-E)^2\,F(\pm Z,E)\ \ .
\end{equation}
Here, $E_0$ is the $\beta$-spectrum end-point total energy, $p$ and $E$
the momentum and total energy of the $\beta^{\mp}$ particle,
respectively, and $F(\pm Z,E)$
Fermi's function for Coulomb corrections, $Z$ being the charge of
the {\em daughter\/} nucleus. For an allowed transition from an initial
state of nuclear spin $J$ to a final state of nuclear spin $J'$, the quantity
$\lambda_{J'J}$ is defined by\cite{JTW}
\begin{equation}
\lambda_{J'J}\ =\ \left\{\begin{array}{c l}
1 & J\rightarrow J'=J-1 \\ \\
\frac{1}{J+1} & J\rightarrow J'=J \\ \\
-\frac{J}{J+1} & J\rightarrow J'=J+1
\end{array}\right.\ \ .
\end{equation}
And finally, $M_F$ and $M_{GT}$ are the Fermi and Gamow-Teller nucleon matrix
elements, respectively.
Ignoring the factor $F(\pm Z,E)$, it is possible to show that
the function $W_0(E)$ reaches its maximal value for a total energy
$E_{\rm max}$ given by
\begin{equation}
\frac{E_{\rm max}}{E_0}\ =\ \frac{1}{6}\ +\ \rho_E\,
\sin\Bigg\{\,\frac{1}{3}\arcsin
\Bigg[\frac{1}{\rho_E^3}
\left(\frac{1}{2}\left(\frac{m}{E_0}\right)^2-\frac{1}{27}\right)\Bigg]\
+\ \frac{2\pi}{3}\,\Bigg\}\ \ \ ,
\label{eq:Emax}
\end{equation}
where
\begin{equation}
\rho_E\ =\ \sqrt{\left(\frac{m}{E_0}\right)^2\,+\,\frac{1}{9}}\ \ \ ,
\end{equation}
$m$ being of course the electron (positron) mass.
Note that in terms of the quantities $a_L$, $a_R$, $b_L$ and
$b_R$ introduced in (\ref{eq:aL}) to (\ref{eq:bR}) of Sect.\ref{Sect2},
one may also write
\begin{equation}
\xi\ =\ \frac{1}{2}\,\left(a_L\,+\,a_R\right)\ \ \ ,
\end{equation}
\begin{equation}
\xi A\ =\ \frac{1}{2}\,\left(b_L\,-\,b_R\right)\ \ \ ,
\end{equation}
\begin{equation}
\xi G\ =\ \mp\,\frac{1}{2}\,\left(a_L\,-\,a_R\right)\ \ \ ,
\end{equation}
\begin{equation}
\xi N'\ =\ \mp\,\frac{1}{2}\,\left(b_L\,+\,b_R\right)\ \ \ .
\end{equation}
Moreover, in the particular case of allowed pure GT transitions,
one observes that
\begin{equation}
(\xi G)_{|_{GT}}\ =\ \frac{1}{\lambda_{J'J}}\,(\xi A)_{|_{GT}}\ \ \ ,\ \ \
(\xi N')_{|_{GT}}\ =\ \lambda_{J'J}\,\xi_{|_{GT}}\ \ \ ,
\end{equation}
thus showing that in this instance only the matrix element $\xi_{|_{GT}}$
and the asymmetry parameter $A_{|_{GT}}$ are relevant to the description
of the decay. In particular, relative measurements as those considered
in this note are then only dependent on the asymmetry parameter $A_{|_{GT}}$
in the case of allowed pure GT transitions.
In the Standard Model, the coefficients
$C_V$, $C_V'$, $C_A$ and $C_A'$ are simply
determined up to a common factor $C_V^0$ by the relations,
\begin{equation}
C_V'\ =\ C_V\ \ \ ,\ \ \
C_A'\ =\ C_A\ \ \ ,
\end{equation}
together with
\begin{equation}
C_V\ =\ C^0_V\ \ \ ,\ \ \
C_A\ =\ \frac{g_A}{g_V}\,C^0_V\ \ \ ,
\end{equation}
$g_V$ and $g_A$ being the standard vector and axial couplings of
nucleon $\beta$-decay, respectively, such that\cite{PDG}
\begin{equation}
\frac{g_A}{g_V}\ =\ -1.2573\pm 0.0028\ \ .
\end{equation}
Given the definitions
\begin{equation}
\rho\ =\ \frac{g_A}{g_V}\,<\,0\ \ ,\ \ \
\lambda\ =\ \frac{g_A}{g_V}\,\frac{M_{GT}}{M_F}\ =\
\rho\, \frac{M_{GT}}{M_F}\ \ \ ,
\end{equation}
one then obtains,
\begin{equation}
\xi_0=2\,|C^0_V|^2\,\Big[\,M_F^2+\rho^2 M_{GT}^2\,\Big]=
2\,|C^0_V|^2\,M_F^2\,\Big[\,1+\lambda^2\,\Big]\ \ \ ,
\end{equation}
\vspace{10pt}
\begin{equation}
\begin{array}{c c l}
(\xi A)_0&=&2\,|C^0_V|^2\,\Big[\,\mp\rho^2 M_{GT}^2\lambda_{J'J}\ -\
2\,\delta_{J'J}\,\rho\,M_F M_{GT}\,\sqrt{\frac{J}{J+1}}\,\Big]\\ \\
& = & 2\,|C^0_V|^2\,M_F^2\,\Big[\,\mp\lambda^2\lambda_{J'J}\ -\
2\,\delta_{J'J}\lambda\,\sqrt{\frac{J}{J+1}}\,\Big]\ \ ,
\end{array}
\end{equation}
\vspace{10pt}
\begin{equation}
(\xi G)_0=\mp\,\xi_0\ \ \ ,
\end{equation}
and finally
\begin{equation}
(\xi N')_0=\mp\,(\xi A)_0\ \ \ .
\end{equation}
Thus for example, the asymmetry parameter $A_0$ in the SM
simply reduces to
\begin{equation}
A_0\ =\ \frac{1}{1+\lambda^2}\,
\Big[\,\mp\lambda^2\lambda_{J'J}\ -\
2\,\delta_{J'J}\lambda\,\sqrt{\frac{J}{J+1}}\,\Big]\ \ .
\label{eq:Appendix.A0}
\end{equation}
In particular,
for allowed pure GT transitions this result becomes
\begin{equation}
{A_0}_{|_{GT}}\ =\ \mp\lambda_{J'J}\ \ \ ,
\label{eq:Appendix.A0GT}
\end{equation}
which is thus maximal only for transitions such that $(J'=J-1)$,
\begin{equation}
{A_0}_{|_{GT}}\ =\ \mp\, 1\ \ \ ,\ \ \ J'=J-1\ \ .
\end{equation}
This is the case for example for $^{107}$In and $^{12}$N.
In the case of LRSM, the coefficients $C_V$, $C_V'$, $C_A$
and $C_A'$ are given in (\ref{eq:coeffLRSM}) in terms of the
combinations of fundamental parameters of LRSM defined
in (\ref{eq:comb1}) to (\ref{eq:comb4}). In order to list
the expressions required for the evaluation of the
parameters $\xi$, $A$, $G$ and $N'$, let us also introduce
the notation
\begin{equation}
C^2_N\ =\ 2\,g_V^2\,\eta_0^2|V^L_{ud}|^2\,{\sum_i}'|U^L_{ie}|^2\ \ \ ,
\end{equation}
and refer to the relations (\ref{eq:Xplus}) to
(\ref{eq:T}) in Sect.\ref{Sect3}
for the definition of the other quantities appearing in
the expressions which follow.
One then obtains,
\begin{equation}
|C_V|^2+|C_V'|^2=C^2_N\Bigg[\,Z_-\,+\,X_-\,\Bigg]\ \ \ ,
\end{equation}
\begin{equation}
|C_A|^2+|C_A'|^2=\rho^2\,C^2_N\Bigg[\,Z_+\,+\,X_+\,\Bigg]\ \ \ ,
\end{equation}
\begin{equation}
2\,{\rm Re}\Bigg(C_V {C_V'}^*\Bigg)=C^2_N\Bigg[\,Z_-\,-\,X_-\,\Bigg]\ \ \ ,
\end{equation}
\begin{equation}
2\,{\rm Re}\Bigg(C_A {C_A'}^*\Bigg)=\rho^2\,
C^2_N\Bigg[\,Z_+\,-\,X_+\,\Bigg]\ \ \ ,
\end{equation}
\begin{equation}
{\rm Re}\Bigg(C_V {C_A'}^*+C_V'{C_A}^*\Bigg)\,=\,\rho\,
C^2_N\Bigg[\,T\,+\,Y\,\Bigg]\ \ \ ,
\end{equation}
\begin{equation}
{\rm Re}\Bigg(C_V C_A^*+C_V'{C_A'}^*\Bigg)\,=\,\rho\,
C^2_N\Bigg[\,T\,-\,Y\,\Bigg]\ \ \ ,
\end{equation}
\vspace{10pt}
\begin{equation}
|C_V+C_V'|^2=2\,C^2_N\,Z_-\ \ \ ,
\end{equation}
\begin{equation}
|C_V-C_V'|^2=2\,C^2_N\,X_-\ \ \ ,
\end{equation}
\begin{equation}
|C_A+C_A'|^2=2\,\rho^2\,C^2_N\,Z_+\ \ \ ,
\end{equation}
\begin{equation}
|C_A-C_A'|^2=2\,\rho^2\,C^2_N\,X_+\ \ \ ,
\end{equation}
\begin{equation}
{\rm Re}\Bigg(C_V+C_V'\Bigg)\Bigg({C_A}^*+{C_A'}^*\Bigg)=2\,\rho\,
C^2_N\,T\ \ \ ,
\end{equation}
\begin{equation}
{\rm Re}\Bigg(C_V-C_V'\Bigg)\Bigg({C_A}^*-{C_A'}^*\Bigg)=-\,2\,\rho\,
C^2_N\,Y\ \ \ .
\end{equation}
\clearpage
\newpage
|
\section{Introduction}
In this paper we deal with the problem of the evolution of two replicas of a
Boolean system (cellular automaton) that evolve stochastically under the
same realization of the noise. The system is defined on a regular lattice of
$L$ sites and evolves in discrete time steps. We limit the explicit analysis
to one dimensional systems, but the results can be extended to higher
dimensions.
Let us indicate the time with the index $t=1,\dots ,\infty $ and the space
with $i=0,1,\dots ,L-1$. The state variables $\sigma (i,t)$ can assume the
values 0 or 1 (Boolean variables). The evolution of $\sigma (i,t)$ is given
by probabilistic transition rules and depends on a small number of
neigboring sites; in its simpler form, $\sigma (i,t)$ depends only on the
state of the two nearest neigbors. In this case one can consider the
space-time lattice as a tilted square lattice.
In order to simplify the notation, we write $\sigma _{+}=\sigma (i+1,t)$, $%
\sigma _{-}=\sigma (i-1,t)$, $\sigma ^{\prime }=\sigma (i,t+1)$. The
evolution rule can be synthetically written as
\[
\sigma ^{\prime }=f(\sigma _{-},\sigma _{+}).
\]
Since the number of possible values of the couple $(\sigma _{-},\sigma _{+})$
is four, the function $f$ is usually specified by giving the four transition
probabilities $\tau (\sigma _{-},\sigma _{+}\rightarrow 1)$ from each
possible configuration to one:
\begin{equation}
\begin{tabular}{lll}
$\tau (0,0\rightarrow 1)$&$=$ & $p_0$ \\
$\tau (0,1\rightarrow 1)$&$=$ & $p_1$ \\
$\tau (1,0\rightarrow 1)$&$=$ & $p_2$ \\
$\tau (1,1\rightarrow 1)$&$=$ & $p_3$%
\end{tabular}
\label{transprob}
\end{equation}
The normalization condition gives $\tau (\sigma _{-},\sigma _{+}\rightarrow
0)=1-\tau (\sigma _{-},\sigma _{+}\rightarrow 1)$.
All sites of the lattice are generally updated synchronously.
Except for the case of deterministic cellular automata, for which the
transition probabilities are either zero or one, we do not expect strong
differences between parallel and sequential updating.
This schematization naturally arise in the modelization of several systems
(contact processes), in physical and biological investigations. It has been
introduced by E. Domany and W. Kinzel\thinspace \cite{domany:DK,kinzel:DK},
and can be considered the prototype for all local stochastic processes. For
a short review of the applicability of this model, see references~\cite
{grassberger:damage} and \cite{dickman:numerical}.
In the thermodynamical limit, the Domany-Kinzel (DK) model exhibits a phase
transition from an ordered to a disordered phase for $p_0=0$. The ordered
configuration is $\sigma (i)=0$ for all $i$ (adsorbing state). The order
parameter is the asymptotic density $m=\lim_{t\rightarrow \infty
}\lim_{L\rightarrow \infty }m(t,L)$, where
\[
m(t,L)=\frac 1L\sum_{i=0}^{L-1}\sigma (i,t).
\]
In the following, we refer to the critical surface $m=0$ and all its
intersections with planes in the $p_j$ space with the symbol $\alpha $
(see the figures).
This transition has been studied mainly for the symmetric case $p_1=p_2$.
Except for a phenomenological renormalization study to which we collaborated
\cite{bagnoli:PRG}, the transition line has been found numerically to belong
to the universality class of directed percolation, which is a particular
case of the model. The disagreement for the renormalization group results
can originate from finite-size effects. For the asymmetric case $p_1\neq p_2$%
, it has been claimed \cite{martins:DKnewPhase} that the phase transition
belongs to a different universality class (mean field).
The existence of an adsorbing state is a non-equilibrium feature of the
model, allowing the presence of a phase transition also in a one-dimensional
(spatial) system. It is shown is section\thinspace \ref{section:extensions}
that in the DK model there can be two adsorbing states ($\sigma(i)=0$ and $%
\sigma(i)=1$), related by a simple transformation of the transition
probabilities. The two transition lines met at the point $M$ ($p_1=1/2$, $%
p_3=1$). This point corresponds to the problem of a random walk in one
dimension, and thus exhibit mean field exponents.
A powerful tool for the investigation of this kind of models is the study of
damage spreading. One considers two replicas $\sigma $ and $\eta $ of the
same model, with different initial conditions (they can be completely
uncorrelated or differ only in some sites). The two replicas evolve under
the same realization of the stochasticity. The difference at site $i$ and at
time $t$ between the two configurations is given by
\[
h(i,t) =\sigma (i,t)\oplus \eta (i,t);
\]
where the symbol $\oplus $ represents the sum modulus two (eXclusive OR,
XOR). Since we use Boolean variables ($a,b\in \{0,1\}$) one can interpret
the exclusive or as $a\oplus b=a+b-2ab$. When mixing XOR and AND
(represented as a multiplication), one can use the algebraic rules for the
sum and the multiplication.
The order parameter for the damage spreading transition is the asymptotic
Hamming distance $H=\lim_{t\rightarrow \infty }\lim_{L\rightarrow \infty
}H(t,L)$ defined as
\[
H(t,L) =\frac 1L\sum_{i=0}^{L-1}h(i,t),
\]
using the usual sum.
The critical
surface $H=0$ and its intersection are indicated with the symbol
$\gamma $.
In the DK model, numerical and analytical investigations \cite
{martins:DKnewPhase,kohring:DKrevisited, zabende:DKphaseDiagram,
rieger:DKreentrant, tome:DKspreading, grassberger:damage} indicated the
existence of a damage spreading phase.
The damage phase transition can be thought as an ergodicity breaking
transition: in the phase where the damage disappears, all initial conditions
asymptotically follow a trajectory that does not depend on the initial
conditions, but only on the realization of the noise. The Hamming distance
can be easily related to the overlap between the configurations.
The critical exponents for the density and the damage transitions in the
plane $p_1=p_2$ and $p_0=0$ are numerically the same\thinspace \cite
{martins:DKnewPhase,grassberger:damage}. It has been guessed\thinspace \cite
{janssen:guess,grassberger:guess} that all continuous transition from an
adsorbing to and active state belong to the universality class of the DK
model (and thus of directed percolation), and that the same universality
class should include all damage spreading transitions\thinspace \cite
{grassberger:damage}.
Here we want to investigate the connection between the density phase
transition and the damage phase transition in the DK model. We have to
carefully describe the dynamics of the model: the position of the transition
line depends on the way in which the randomness is implemented in the actual
simulations. In section \ref{section:notations} we introduce the formalism
that allows an exact description of how randomness is implemented in the
model. We are thus able to write down the evolution equation for the spins,
and to obtain the evolution equation for the distance between two replicas.
The structure of the latter equation corresponds to the DK model with $p_0=0$%
. We conclude that the universality class of damage spreading is, at least
for this simple case, that of directed percolation. In section \ref
{section:phaseDiagram} we obtain the phase diagram of the DK model by
mapping the transition line for the density to the transition line for the
damage by means of mean field approximations. In section \ref
{section:extensions} we show that one can infer the existence of a phase
transition for the damage also in cases for which there is no phase
transition for the density, and that there are two disjoint regions in the
parameters space for the damage spreading. Finally, conclusions and open
questions are drawn in the last section.
\section{The damage spreading transition}
\label{section:notations} Let us start from a simple example, the dilution
of rule 90 (in Wolfram's notation\thinspace \cite{wolfram:ca}) that will
also serve to fix the notation. Rule 90 is a deterministic rule that evolves
according with
\[
\sigma ^{\prime }=\sigma _{-}\oplus \sigma _{+}.
\]
The transition probabilities for the diluted rule 90 are
\[
\begin{tabular}{lll}
$\tau (0,0\rightarrow 1)$&$=$&$0$ \\
$\tau (0,1\rightarrow 1)$&$=$&$p$ \\
$\tau (1,0\rightarrow 1)$&$=$&$p$ \\
$\tau (1,1\rightarrow 1)$&$=$&$0,$%
\end{tabular}
\]
$p$ being the control parameter of the model.
In order to apply rule 90 for a fraction $p$ of sites, and rule 0 (all
configurations give 0) for the rest, one usually extracts a random
number $r=r(i,t)$ for each site and at each time step and chooses the
application of rule 90 or rule 0 according with $r<p$ or $r\ge p$ resp.
We can easily write the explicit expression for this rule by means of the
function $[\cdot ]$, assuming that $[\mbox{\it logical proposition\/}]$
takes the value 1 if {\it logical proposition\/} is true, and 0 otherwise
(this interpretation of logical propositions is the standard one in C
language). Finally, we have for the diluted rule 90
\begin{equation}
\sigma ^{\prime }=[r<p]\left( \sigma _{-}\oplus \sigma _{+}\right) .
\label{rule90}
\end{equation}
One can also think of having all random numbers $r(i,t)$ extracted before
the simulation and attached to the sites of the space-time lattice even
though they are not always used. The random numbers are thus similar to a
space-time quenched (disordered) field.
Once given the set of random numbers, the evolution is completely
deterministic, and the evolution function depends on the lattice position
(spatial and temporal) via the random numbers $r(i,t)$. One can
alternatively define the model stating that deterministic functions are
randomly distributed on the space-time lattice according to a certain
probability distribution. This description is very reminiscent of the
Kauffman model \cite{kauffman:model}.
The damage spreading can be considered a measure of the stability of the set
of possible trajectories, averaging over the realizations of the noise. The
original definition of Lyapunov exponent is a measure of the instantaneous
effects of a vanishing perturbation. Since the state variables of cellular
automata assume only integer values, one has to extend the definition to a
finite initial distance (and to finite time steps), thus taking into account
the possibility of non-linear effects. For cellular automata, the smallest
initial perturbation corresponds to a difference of only one site between
the two replicas. The short-time effects of a (vanishing) perturbation
define the analogous of the derivatives for a continuous system \cite
{bagnoli:derivatives}. The study of the equivalent of the usual (linear)
Lyapunov exponent for deterministic cellular automata allows a
classification of the rules according with the trend of the damage \cite
{bagnoli:lyapunov}. The general problem of damage spreading can thus be
considered equivalent to the study of the non-linear Lyapunov exponent (i.e.
finite initial distance and finite evolution times) for space-time
disordered cellular automata.
Using the concept of Boolean derivatives \cite{bagnoli:derivatives}, we
develop a Boolean function $f(a,b)$ as
\[
f(a,b)=f_0\oplus f_1a\oplus f_2b\oplus f_3ab,
\]
where the Taylor coefficients are
\begin{eqnarray*}
f_0 &=&f(0,0) \\
f_1 &=&f(0,1)\oplus f(0,0) \\
f_2 &=&f(1,0)\oplus f(0,0) \\
f_3 &=&f(1,1)\oplus f(0,1)\oplus f(1,0)\oplus f(0,0).
\end{eqnarray*}
One can verify the previous expression by enumerating all the possible
values of $a$ and $b$.
Using the bracket $[\cdot ]$ notation, the transition probabilities (\ref
{transprob}) correspond to
\[
\begin{tabular}{lll}
$f(0,0)$&$=$&$[r_0<p_0]$ \\
$f(0,1)$&$=$&$[r_1<p_1]$ \\
$f(1,0)$&$=$&$[r_2<p_2]$ \\
$f(1,1)$&$=$&$[r_3<p_3],$%
\end{tabular}
\]
where the random numbers $r_j(i,t)$ belongs to the interval $[0,1)$ and
constitute the quenched random field. We neglect to indicate the spatial and
temporal indices for simplicity.
The Taylor coefficients become
\begin{eqnarray*}
f_0 &=&[r_0<p_0] \\
f_1 &=&[r_1<p_1]\oplus [r_0<p_0] \\
f_2 &=&[r_2<p_2]\oplus [r_0<p_0] \\
f_3 &=&[r_3<p_3]\oplus [r_2<p_2]\oplus [r_1<p_1]\oplus [r_0<p_0].
\end{eqnarray*}
In the following we shall assume $p_1=p_2$ and $r_1=r_2$ so that $f_1 = f_2$
and
\begin{eqnarray*}
f_3 &=& [r_3<p_3]\oplus [r_0<p_0].
\end{eqnarray*}
The correlations among the random numbers $r_j$ (at same space-time
position) affect the position of the critical line for the damage ($\gamma $%
), as already pointed out also by P. Grassberger\thinspace \cite
{grassberger:damage} and E. Domany\thinspace \cite{domany:private}, but not
the position of the transition line for the density ($\alpha $). Only a
careful description of how the randomness is implemented in the model
completely specify the problem of damage spreading. In principle one could
study the case of generic correlations among these random numbers. Here we
consider only two cases: either all $r_j$ are independent (case {\it i},
transition line $\gamma _i$) or they are all identical (case {\it ii},
transition line $\gamma _{ii}$).
The evolution equation for the single site variable $\sigma =\sigma (i,t)$
is
\begin{eqnarray}
\sigma ^{\prime }&=&[r_0<p_0]\oplus
\left( [r_1<p_1]\oplus [r_0<p_0]\right)
(\sigma _{-} \oplus \sigma _{+})\oplus \nonumber \\
&&\left( [r_3<p_3]\oplus [r_0<p_0]\right) \sigma _{-}\sigma _{+}. \label{DK}
\end{eqnarray}
We can substitute the evolution equation for the replica $\eta =\sigma
\oplus h$, with the evolution equation for the damage $h=\sigma \oplus \eta $%
, obtaining
\begin{eqnarray}
h^{\prime } &=&\left( [r_1<p_1]\oplus [r_0<p_0]\oplus \left( [r_3<p_3]\oplus
[r_0<p_0]\right) \sigma _{+}\right) h_{-}\oplus \nonumber \\
&&\left( \lbrack r_1<p_1]\oplus [r_0<p_0]\oplus \left( [r_3<p_3]\oplus
[r_0<p_0]\right) \sigma _{-}\right) h_{+}\oplus \nonumber \\
&&\left( [r_3<p_3]\oplus [r_0<p_0]\right) h_{-}h_{+}. \label{DKd}
\end{eqnarray}
This equation has the same structure of the evolution equation of the
original model (\ref{DK}) if in this latter we set $p_0=0$. Remembering that
only for this value of $p_0$ the DK model exhibits a phase transition, we
have a strong argument for the correspondence between directed percolation
and damage spreading transitions. However, also in the symmetric case $%
p_1=p_2$ and $r_1=r_2$, the evolution equation of $h$ is symmetric only in
average, and one has to take into consideration the correlations between $%
\sigma _{-}$ and $\sigma _{+}$. As discussed before, these correlations can
be included in the definition of the DK model, which specify only the
transition probabilities. It remains to be proved that all these versions do
belong to the same universality class.
Let us now consider the case $p_0=0$. Previous numerical investigations
showed that on this plane the two curves $\alpha$ and $\gamma$ meet at the
point $Q = (\sim 0.81,0)$. Inserting the value $p_3=p_0=0$ in the equation (%
\ref{DKd}), we see that the evolution law for $h$ is the same of that for $%
\sigma $, and so both transitions coincide on this line. This corresponds
also to the dilution of rule 90.
Since the rest of the $\gamma $ curve lies away from the density transition
line, the correlations among sites decay rapidly in time and space. This
allow us to use a mean field approximation. In the simplest form, we replace
$\sigma (i,t)$ with a random bit that assumes the value one with probability
$m$. With this assumption the equation (\ref{DKd}) becomes
\begin{eqnarray}
h^{\prime } &=&([r_1<p_1]\oplus [r_3<p_3][r_4<m])h_{-}\oplus \nonumber \\
&&([r_1<p_1]\oplus [r_3<p_3][r_5<m])h_{+}\oplus [r_3<p_3]h_{-}h_{+};
\label{h}
\end{eqnarray}
where $r_4$ and $r_5$ are independent random numbers. This is a strong
approximation, both because of correlations and because the same $\sigma
(i,t)$ is shared by $h(i-1,t+1)$ and $h(i+1,t+1)$. Nevertheless, we can
assume this equation as a starting point in our derivation of the phase
diagram.
We now want to remap this model onto the original DK model, assuming that
the asymmetry ($r_4\neq r_5$), that in average vanishes, does not strongly
affect the transition.
The remapped transition probabilities $\tilde p$ are
\[
\begin{tabular}{lllll}
$\tilde \tau (0,0\rightarrow 1)$&$=$&$\tilde p_0$&$=$&$0$ \\
$\tilde \tau (0,1\rightarrow 1)$&$=$&$\tilde p_1$&$=$&$\pi \left(
[r_1<p_1]\oplus [r_3<p_3][r_5<m]\right)$ \\
$\tilde \tau (1,0\rightarrow 1)$&$=$&$\tilde p_1$&$=$&$\pi \left(
[r_1<p_1]\oplus [r_3<p_3][r_4<m]\right)$ \\
$\tilde \tau (1,1\rightarrow 1)$&$=$&$\tilde p_3$&$=$&$\pi \left(
[r_3<p_3]([r_4<m]\oplus [r_5<m]\oplus 1)\right),$%
\end{tabular}
\]
where $\pi (f(r))=\int_0^1drf(r)$ is the probability that the Boolean
function $f$ of the random number $r$ takes the value one.
For case {\it i} ($r_1\ne r_3$), we have
\begin{equation}
\begin{tabular}{lll}
$\tilde p_1$&$=$&$ p_1+p_3m-2p_1p_3m$ \\
$\tilde p_3$&$=$&$p_3(1-2m(1-m)),$
\end{tabular}
\label{casei}
\end{equation}
while for case {\it ii} ($r_1 = r_3$)
\begin{equation}
\begin{tabular}{lll}
$\tilde p_1$&$=$&$m|p_1-p_3|+(1-m)p_1;$ \\
$\tilde p_3$&$=$&$p_3(1-2m(1-m)).$
\end{tabular}
\label{caseii}
\end{equation}
Since $\gamma $ lies in the $p_1>p_3$ region, one has for case {\it ii}
\[
\tilde p_1=p_1-mp_3.
\]
Notice that for $p_3=0$ or for $m=0$ the two curves $\gamma _i$ and
$\gamma_{ii}$ coincide, as already noticed numerically by Grassberger \cite
{grassberger:damage}.
Given a certain point $(p_1,p_3)$, it belongs to the damage transition line $%
\gamma $ ($H(p_1,p_3)=0$) if the point $(\tilde p_1,\tilde p_3)$ belongs to
the density transition line $\alpha $ ($m(p_1,p_3)=0$). In order to draw the
phase diagram for the Hamming distance, one has to know the value of the
density $m$ in all the parameter space, and in particular the position of $%
\alpha $. Unfortunately, we do not have a simple expression for these
quantities; in the next section we use some approximation in order to draw a
rough phase diagram. However, we are able to demonstrate that $\alpha $ and $%
\gamma $ are tangent at point $Q$.
The slope $q$ of the normal to $\alpha $ at $Q$ can be given as
\[
q=\left. \frac{\partial m}{\partial p_1}\right/ \left. \frac{\partial m}{\ {%
\partial p_3}}\right| _Q.
\]
Considering that $\gamma \equiv H(p_1,p_3)=0\equiv m(\tilde
p_1(p_1,p_3),\tilde p_3(p_1,p_3))=0$, the partial derivatives of $H$ are
given by
\begin{eqnarray*}
\frac{\partial H}{\partial p_1} &=&\frac{\partial m}{\partial \tilde p_1}%
\frac{\partial \tilde p_1}{\partial p_1}+\frac{\partial m}{\partial \tilde
p_3}\frac{\partial \tilde p_3}{\partial p_1}; \\
\frac{\partial H}{\partial p_3} &=&\frac{\partial m}{\partial \tilde p_1}%
\frac{\partial \tilde p_1}{\partial p_3}+\frac{\partial m}{\partial \tilde
p_3}\frac{\partial \tilde p_3}{\partial p_3}.
\end{eqnarray*}
One has to take into account that $\tilde p_j$ depends on $p_i$ both
directly and via $m$. Inserting the relations \ref{casei} or \ref{caseii}
and considering that at point $Q$, $m=p_3=0$ one obtains that
\[
q^{\prime }=\left. \frac{\partial H}{\partial p_1}\right/ \left. \frac{%
\partial H}{{\partial p_3}\ }\right| _Q=q.
\]
Since we know from numerical experiments and from all mean field
approximation beyond the very first one that the slope of $\alpha $ at $Q$
is negative in the $(p_1,p_3)$ plane, the tangency of $\gamma $ to $\alpha $
implies a reentrant behavior for the damage transition curve, as observed in
reference~\cite{rieger:DKreentrant}.
\section{The phase diagram}
\label{section:phaseDiagram} The problem of sketching an approximate phase
diagram for the damage in an analytical way has been dealed with by several
authors \cite{kohring:DKrevisited, rieger:DKreentrant, tome:DKspreading}.
Since any equation for the damage depends on the behavior of one replica,
there are two sources of errors to be controlled: the approximations for the
evolution of one replica and that for the difference (or for the other
replica). As a consequence, all approximation schemes proposed so far
require large efforts for a poor result. Our method is able to exploit the
knowledge of the density phase to study the damage phase transition. There
are several methods that rapidly converge to a good approximation of $\alpha
$; to our knowledge the best ones are the phenomenological renormalization
group \cite{bagnoli:PRG} and the cluster approximation (local structure)
\cite{gutowitz:localStructure} improved by finite-size scaling. This latter
method can also give a good approximation of the behavior of $m(p_0,p_1,p_3)$
at any point.
Since here we are not interested in numerical competitions, we use the
high-quality data for the density transition line from reference \cite
{grassberger:damage} combined with a first order mean field approximation
for the density. The $\alpha $ curve has been approximated in the $(p_1,p_3)$
plane by a $5^{\text{th}}$ order polynomial
\begin{equation}
p_3=\sum_{i=0}^5a_ip_1^i. \label{alpha}
\end{equation}
The simplest mean field approximation for the asymptotic density $m$ gives
\[
m=\frac{1-2p_1}{p_3-2p_1}.
\]
By using these approximations one obtains from equations (\ref{casei}) or (%
\ref{caseii}) the curves reported in figure~\ref{figure:phaseDiagram},
together with the presently best numerical results \cite{grassberger:damage}%
.The main source of error is that in the mean field approximation the $\alpha $
curve does not corresponds to the zero of the density. This is particularly
evident in the absence of reentrancy of curves $\gamma _1$ and $\gamma _2$.
Nevertheless even this rough approximation is able to reproduce
qualitatively the phase diagram and to exhibit the influence on the damage
critical line of the different implementations of randomness. Notice that
the $\gamma $ curve from reference \cite{grassberger:damage} corresponds to
the implementation of equation~(\ref{casei}).
\section{The $p_0>0$ case}
\label{section:extensions} The DK model with arbitrary $p_0$ includes all
one dimensional symmetric cellular automaton model or spin system with two
inputs. We can represent each possible model as a point in the three
dimensional unit cube parametrized by $p_0$, $p_1$, $p_3$. The general form
of the transition probabilities from equation (\ref{DKd}) is
\begin{equation}
\begin{tabular}{lll}
$\tilde p_1$&$=$&$p_1+(1-2p_1)(mp_3+(1-m)p_0);$ \\
$\tilde p_3$&$=$&$(p_0+p_3-2p_0p_3)(1-2m+2m^2).$%
\end{tabular}
\label{p0g0}
\end{equation}
There is a trivial transformation of the original DK model with $p_0=0$. One
can revert ($0\leftrightarrow 1$) all the spins before and after the
application of the rule. The new transition probabilities $p_i^{\prime }$
are
\begin{eqnarray*}
p_0^{\prime } &=&1-p_3; \\
p_1^{\prime } &=&1-p_1; \\
p_3^{\prime } &=&1-p_0.
\end{eqnarray*}
The critical plane $p_0=$ $0$ maps to $p_3=1$, and the adsorbing state is
now the configuration in which all spins are one. We indicate with the
symbol $\alpha ^{\prime }$ the critical curve obtained by this
transformation. The point $Q$ is mapped to the point $Q^{\prime }=(1,$ $\sim
0.2,1)$. The parameter cube and the critical curves are reported in figure~%
\ref{figure:cube}. This mapping suggests the presence of a damaged zone near
the corner $(1,0,1)$.
In order to study the position of the critical surfaces for the damage, we
numerically solved equation (\ref{p0g0}) combined with the expression (\ref
{alpha}) of the critical line $\alpha $ in the very simple approximation $%
m=0.5$. The results are reported in figure \ref{figure:cube}. Direct
numerical simulations qualitatively agree with this picture.
The one dimensional Ising model in zero field with heath bath dynamics can
also be expressed with this formalism.
The local field $g=g_i$ for the one dimensional Ising model is
\[
g=K\left( (2\sigma _{-}-1)+(2\sigma _{+}-1)\right) ,
\]
where $K=\beta J=J/k_{\text{B}}T$ is the rescaled coupling constant and $%
\sigma =0,1$ the site variables (spin). The local field $g$ can assume the
values $-2K$, $0$, $2K$.
For the heath bath dynamics, $\sigma'$ takes the value one with probability
$p$ given by
\[
p=\frac 1{1+\exp (-2g)}.
\]
The transition probabilities are
\begin{eqnarray*}
p_0 &=&\frac \xi {1+\xi }; \\
p_1 &=&\frac 12; \\
p_3 &=&\frac 1{1+\xi },
\end{eqnarray*}
where $\xi =\exp (-4K)$. Notice that $p_3=1-p_0$; for $T>0$, $p_0<1/2$,
while for negative temperatures $p_0>1/2$. The point $p_0=p_3=1/2$
corresponds to infinite $T$.
The evolution equation for the site variable is
\begin{eqnarray*}
\sigma ^{\prime } &=&[r<p_0]\oplus ([r<p_1]\oplus [r<p_0])(\sigma _{-}\oplus
\sigma _{+})\oplus \\
&&([r<p_3]\oplus [r<p_0])\sigma _{-}\sigma _{+};
\end{eqnarray*}
where usually all Taylor coefficients depend on the same random number $%
r=r(i,t)$. The existence line $\omega _{+}$ for the Ising model with $T>0$, $%
p_1=1/2$, $p_3=1-p_0$, intersects the critical line $\alpha $ at $%
M=(0,1/2,1) $. The existence line $\omega _{-}$ for $T<0$ ends at $M^{\prime
}=(1,1/2,0)$. The point $R=(1/2,1/2,1/2)$ corresponds to $T=\infty $ (see
figure \ref{figure:cube}).
The evolution equation for the Hamming distance $h$ is equivalent to
equation (\ref{DKd}), with all $r_j$ equal to $r$. Taking into account the
correlations induced by the random numbers, and that the magnetization is $%
1/2$ except at the critical point, one obtains
\begin{eqnarray*}
\tilde p_1 &=&\frac{1-\xi }{2(1+\xi )}; \\
\tilde p_3 &=&\frac{1-\xi }{1+\xi },
\end{eqnarray*}
i.e. the line $\chi \equiv p_3=2p_2$, $p_0=0$ for positive or negative
temperatures. The line $\chi $ intersects $\alpha $ at point $M$ for $%
T=0^{\pm }$, confirming that the symmetry breaking transition for the Ising
model occurs at zero temperature.
\section{Conclusions and perspectives}
In this work we presented a formalism that allows the careful description of
Boolean algorithms for stochastic cellular automata (including spin system
like the Ising model). Using this formalism we were able to derive the exact
equation for the evolution of a damage between two replicas that evolve
under the same realization of the noise. Using a mean field hypothesis, we
gave strong indications that the critical line for the damage phase
transition belongs to the same universality class of that for the density in
the DK model, and thus to the directed percolation universality class. We
mapped the density critical line to the damage critical line, obtaining the
regions in the parameter space of a general symmetric cellular automaton
where the replica symmetry breaking is to be expected. Our predictions are
qualitatively confirmed by numerical simulations.
Several questions remain to be answered. Among others: is it possible to
obtain similar results starting from a field description? How does the phase
diagram for more general (asymmetric, three-input, etc.) cellular automata
look like?
\subparagraph{Acknowledgments}
We want to acknowledge fruitful discussions with
P. Grassberger, R. Livi, J. Parrondo,
R. Rechtman and S. Ruffo. We thank P. Grassberger for having
gracefully furnished his high-quality data for the curve $\alpha $ and $%
\gamma $. This work was started during a stage in Mexico with the support of
CNR and CONACYT. Part of the work was performed in Germany with the support
of {\em programma Vigoni} of the {\em Conferenza permanente dei rettori
delle universit\`a italiane}. Finally, the work was terminated during the
workshop {\em Chaos and Complexity} as ISI-Villa Gualino under the CE
contract n. ERBCHBGCT930295.
|
\section*{\large\bf 1. Introduction}
In the field of relativistic heavy-ion collisions
the analysis of Bose-Einstein correlations \cite{HBT56} has attracted much
attention recently. The general hope is to extract
information about the size of a source radiating mesons by studying
their two-particle correlation function \cite{NA44,NA49}.
These correlations are typical quantum effects,
hence quantum field theory is a proper framework
to describe the problem theoretically. Such efforts have been
undertaken with great success for many years
\cite{KP72,GKW79,P84,DS92}, with a big emphasis on
the quantum properties of mesons that were
emitted from a distribution (in space and time) of classical currents.
The merger of these theoretical considerations with
space-time distributions of quasi-particles generated
by simulation codes for relativistic heavy ion collisions
has led to predictions of correlation radii that are
in rough agreement with experimental data.
The problem of a free quantum field radiating from
a classical current is exactly solvable \cite[pp.438]{IZ80}, also
at finite temperature \cite{h89coh}.
However, in the original field where Bose-Einstein correlations have been used
to measure source radii, i.e., in the Hanbury-Brown--Twiss analysis
of star''light'', as well as in the analysis of
relativistic heavy-ion collisions, the radiation source is a
localized {\em thermal\/} distribution. Such a state is
a non-equilibrium state, hence somewhat difficult to handle theoretically.
As was pointed out above,
the usual way to circumvent this problem is the approximation of
a non-equilibrium state as the superposition of an infinite
number of classical currents, with a certain current distribution
in space and time (mostly assumed to be gaussian) \cite{GKW79}.
From the viewpoint of quantum field theory this approximation
is somewhat unsatisfactory, if not questionable: True {\em non-equilibrium\/}
effects on the propagating particles are neglected, only
{\em local\/} equilibrium effects are maintained.
Since also in relativistic heavy-ion collisions the
radiating system is off equilibrium, one question arises immediately:
Are there non-equilibrium effects on
the quantum Bose-Einstein correlation ? A secondary
question is, whether the correlation effects generated by
a distribution of classical currents and those generated by a thermal
source are equivalent or not.
To address these questions we consider a stationary temperature
distribution with limited spatial extension, and calculate the two-particle
correlation function of bosons emitted thermally from such a distribution.
In other words, we are investigating correlations in the
``glow'' from a hot spot -- in close analogy to the physical
situation present in astrophysics, but on the quantum level.
To this end we
formulate the two-particle correlation problem in a field
theoretical method suited to handle non-equilibrium states.
However, since the solution of the full problem is beyond our
capabilities, we restrict ourselves to the same order of accuracy
that is reached in standard transport theory: Our correlation function
incorporates non-equilibrium effects beyond the local
equilibrium function, but only up to first order in the
gradients of the ``temperature'' distribution of this system.
Furthermore, the discussion is limited to a static non-equilibrium
system.
In another sense we treat the correlation problem
more consistently than in standard transport theory: We
take into account a nonzero spectral width for the particles we consider.
This is necessary, because at nonzero temperature {\em every\/}
excitation acquires a finite lifetime due to collisions
with the medium \cite{L88}. We describe this finite lifetime by attributing
a certain spectral width $\gamma>0$ {\em inside the hot spot\/}
also to asymptotically stable particles.
For strongly interacting particles, like e.g. pions, we may assume
that such a spectral width is due to the coupling
to $\Delta_{33}$-resonances.
Quantum field theory for non-equilibrium states comes in two flavors:
The Schwinger-Keldysh method \cite{SKF} and thermo field dynamics (TFD,
see \cite{h94rep,Ubook} for references to the original work).
For the purpose of the present paper, we prefer the latter method:
The problem
of an inhomogeneous temperature distribution has been solved
explicitly in TFD up to first order in the temperature gradients
\cite{h94rep}. This solution includes a nontrivial spectral
function of the quantum field under consideration. It employs
a perturbative expansion in terms of {\em generalized free fields\/}
with continuous mass spectrum \cite{L88}.
The paper is organized as follows. The next section contains a
brief introduction to the formalism
of thermo field dynamics for spatially inhomogeneous systems.
In section 3 we derive expressions for the two-boson
correlation function in non-equilibrium states as well as
its local equilibrium approximation. Section 4 contains a study
of the effect that is exerted on the correlation function by a
nonzero spectral width, section 5 contains an example with
semi-realistic pion spectral function in hot nuclear matter.
Finally we draw some conclusions and discuss the experimental
relevance.
\section*{\large\bf 2. Outline of the method}
In ''ordinary'' quantum mechanics,
a statistical state of a quantum system is described by a statistical
operator (or density matrix) $W$, and the measurement of
an observable will yield the average
\begin{equation}\label{av1}
\left\langle \vphantom{\int} {\cal E}(t,\vec{x}) \right\rangle =
\frac{\mbox{Tr}\left[ {\cal E}(t,\vec{x})\;W \right]}{
\mbox{Tr}\left[ W \right]}
\;,\end{equation}
where the trace is taken over the Hilbert space of the quantum system
and ${\cal E}$ is the hermitean operator associated with the
observable. In thermo field dynamics (TFD), the calculation
of this trace is simplified to the calculation of a matrix element
\begin{equation}\label{av3}
\left\langle \vphantom{\int} {\cal E}(t,\vec{x}) \right\rangle =
\frac{ (\mkern-4mu( 1 |\mkern-2.5mu| \;{\cal E}(t,\vec{x}) \;
|\mkern-2.5mu| W)\mkern-4mu)}{
(\mkern-4mu( 1 |\mkern-2.5mu| W)\mkern-4mu)}
\;,\end{equation}
with ''left'' and ''right'' statistical state defined in terms of
the two different commuting representations (see refs.
\cite{Ubook,h94rep} for details).
In this state, we consider a complex, scalar boson field
describing spinless charged excitations in a statistical
system not too far from equilibrium. In the spirit of the
first remark, one could think of this field as describing
positive and negative pions in nuclear matter.
According to the reasoning above,
this ``thermal'' boson field is described by two field
operators $\phi_x$, $\widetilde{\phi}_x$ and their adjoints
$\phi_x^\star$, $\widetilde{\phi}_x^\star$, with canonical
commutation relations
\begin{eqnarray}
\left[\phi(t,\vec{x}) , \partial_t
\phi^\star(t,\vec{x}^\prime)\right] & = &
\mathrm{i}\delta^3(\vec{x}-\vec{x}^\prime) \nonumber \\
\left[\widetilde{\phi}(t,\vec{x}) ,
\partial_t\widetilde{\phi}^\star(t,\vec{x}^\prime)\right] & =-&
\mathrm{i}\delta^3(\vec{x}-\vec{x}^\prime)
\;\end{eqnarray}
but commuting with each other.
These two fields may be combined in a statistical doublet,
see ref. \cite{h94rep,h93trans} for details.
The free as well as the interacting scalar field
can be expanded into momentum eigenmodes
\begin{eqnarray}\label{bf1}
\phi_x & = &
\int\!\! \frac{d^3\vec{k}}{\sqrt{(2\pi)^3}}
\left( a^\dagger_{k-}(t)\,{\mathrm e}^{-\mathrm{i}\vec{k}\vec{x}} +
a_{k+}(t)\, {\mathrm e}^{ \mathrm{i}\vec{k}\vec{x}}\right)
\nonumber \\
\widetilde{\phi}_x & = &
\int\!\! \frac{d^3\vec{k}}{\sqrt{(2\pi)^3}}
\left( \widetilde{a}^\dagger_{k-}(t)\,{\mathrm e}^{\mathrm{i}\vec{k}\vec{x}} +
\widetilde{a}_{k+}(t) \,{\mathrm e}^{-\mathrm{i}\vec{k}\vec{x}}
\right)
\;.\end{eqnarray}
$\vec{k}$ is the three-momentum of the modes, therefore in this notation
$a^\dagger_{k-}(t)$ creates a negatively charged excitation with momentum
$\vec{k}$, while $a_{k+}(t)$ annihilates a positive charge.
Henceforth the two different charges are distinguished
by an additional index $l=\pm$ whenever possible.
For the free case the commutation relations of the $a$-operators
at different times are simple, while they are unknown for the
interacting fields. However, we want to go only one step
beyond the free field case, i.e., we
approximate the fully interacting quantum fields
by {\em generalized free fields\/} \cite{L88}. In this
formulation, the operators $a$,$\widetilde{a}$ do not excite
stable on-shell pions. Rather, they are
obtained as an integral over more general operators
$\xi$, $\widetilde{\xi}$ with a continuous energy parameter $E$
\cite{NUY92,h93trans}:
\begin{eqnarray}\label{bbg4}\nonumber
\left({\array{r} a_{kl}(t)\\
\widetilde{a}^\dagger_{kl}(t)\endarray}\right)\;=
& \int\limits_0^\infty\!\!dE\,\int\!\!d^3\vec{q}
\;{\cal A}^{1/2}_l(E,\vec{k})\,
\left({\cal B}^{-1}_l(E,\vec{q},\vec{k})\right)^\star
\,\left({\array{r}\xi_{Eql}\\
\widetilde{\xi}^\#_{Eql}\endarray}\right)
\,{\mathrm e}^{-\mathrm{i} Et} \nonumber \\
\left({\array{r}a^\dagger_{kl}(t)\\
-\widetilde{a}_{kl}(t)\endarray}\right)^T\;=
& \int\limits_0^\infty\!\!dE\,\int\!\!d^3\vec{q}
\;{\cal A}^{1/2}_l(E,\vec{k})\,
\left({\array{r}\xi^\#_{Eql}\\
-\widetilde{\xi}_{Eql}\endarray}\right)^T\,
{\cal B}_l(E,\vec{q},\vec{k})
\,{\mathrm e}^{\mathrm{i} E t}
\;,\end{eqnarray}
where ${\cal B}$ is a $2\times 2$ matrix,
the weight functions ${\cal A}_l(E,\vec{k})$ are positive and have
support only for positive energies, their normalization is
\begin{equation}\label{norm}
\int\limits_0^\infty\!\!dE\,E\, {\cal A}_l(E,\vec{k}) =\frac{1}{2}
\;\;\;\;\;\;
\int\limits_0^\infty\!\!dE \, {\cal A}_l(E,\vec{k}) =Z_{kl}
\;.\end{equation}
The principles of this expansion have been discussed in ref. \cite{L88},
its generalization to non-equilibrium states was introduced
in ref. \cite{h94rep}. For equilibrium states the combination
\begin{equation}\label{spec}
{\cal A}(E,\vec{k}) = {\cal A}_+(E,\vec{k})\Theta(E) -
{\cal A}_-(-E,-\vec{k})\Theta(-E)
\;\end{equation}
is the spectral function of the
field $\phi_x$
and the limit of free particles with mass $m$ is recovered when
\begin{equation}\label{fb}
{\cal A}(E,\vec{k}) \longrightarrow
\mbox{sign}(E)\,
\delta(E^2 -\vec{k}^2 -m^2)
=\mbox{sign}(E)\,
\delta(E^2 -\omega_k^2)
\;.\end{equation}
For non-equilibrium systems, the existence
of a spectral decomposition cannot be guaranteed \cite{Ubook}.
We may expect however, that close to equilibrium the field
properties do not change very
much. Thus, with this formalism we study a
quantum system under the influence of small gradients in the
temperature, with {\em local\/}
spectral function ${\cal A}(E,\vec{k})$. Corrections to such
a picture only occur in second order of temperature gradients
\cite{h93trans,h94rep}.
A thorough discussion of the $2\times 2$ Bogoliubov matrices was
carried out in ref. \cite{h94rep}. For the purpose of the present paper,
we simply state their explicit form as
\begin{equation}\label{gb}
{\cal B}_l(E,\vec{q},\vec{k}) = \left( { \array{lr}
\left(\delta^3(\vec{q}-\vec{k}) + N_l(E,\vec{q},\vec{k})\right)
\;\;\;& -N_l(E,\vec{q},\vec{k}) \\
-\delta^3(\vec{q}-\vec{k}) & \delta^3(\vec{q}-\vec{k})
\endarray} \right)
\;,\end{equation}
where $N(E,\vec{q},\vec{k})$ is the Fourier transform of a space-local
Bose-Einstein distribution function
\begin{eqnarray}\nonumber \label{nloc}
N_l(E,\vec{q},\vec{k})& = & \frac{1}{(2\pi)^3}\;
\int\!\!d^3\vec{z}\,{\mathrm e}^{-\mathrm{i}
(\vec{q}-\vec{k})\vec{z} }\,n_l(E,\vec{z})\\
n_l(E,\vec{z}) & = &
{\displaystyle \frac{1}{{\mathrm e}^{\beta(\vec{z}) (E-\mu_l(\vec{z}))}-1}}
\;.\end{eqnarray}
Here we have assumed a distribution function that only depends on the
energy parameter $E$ and on the space coordinate $z$. The expansion
allows for a generalization of this, to more general distribution functions
depending also on the momentum $(\vec{q}+\vec{k})/2$.
We have argued, that our ansatz for the fields gives rise to a local
spectral function. A moving particle however feels an influence also of
the {\em gradients\/} of this local equilibrium distribution.
Consequently also the propagator for the fields we consider is correct beyond
a local equilibrium situation, to be precise it is correct to first order
in the gradients of $n_l(E,\vec{z})$.
To complete the brief description of the TFD formalism, we specify the
commutation relation of the various operators in our expressions.
The $\xi$-operators have commutation relations
\begin{equation}\label{difc}
\left[\xi_{Ekl},\xi^\#_{E^\prime k^\prime l^\prime}\right]=
\delta_{ll^\prime}\,
\delta(E-E^\prime)\,
\delta^3(\vec{k}-\vec{k}^\prime)
\;.\end{equation}
Similar relations hold for the $\widetilde{\xi}$ operators,
all other commutators vanish, see \cite{L88}.
It follows from these definitions, that
\begin{eqnarray}\label{co}
\left[a_{kl}(t),a^\dagger_{k^\prime l^\prime}(t)\right]&=&
Z_{kl}\,
\delta_{ll^\prime}\,
\delta^3(\vec{k}-\vec{k}^\prime)\nonumber \\
\left[\widetilde{a}_{kl}(t),
\widetilde{a}^\dagger_{k^\prime l^\prime}(t)\right]&=&
Z_{kl}\,
\delta_{ll^\prime}\,
\delta^3(\vec{k}-\vec{k}^\prime)
\; \end{eqnarray}
are the equal-time commutation relations for the $a$, $\widetilde{a}$
operators.
The $\xi$, $\widetilde{\xi}$ operators act on the
''left'' and ''right'' statistical state according to
\begin{equation}\label{tscc}
\xi_{Ekl}|\mkern-2.5mu| W )\mkern-4mu) = 0, \;\;
\widetilde{\xi}_{Ekl}|\mkern-2.5mu| W )\mkern-4mu) =0,\;\;
(\mkern-4mu( 1|\mkern-2.5mu| \xi^\#_{Ekl} = 0 ,\;\;
(\mkern-4mu( 1|\mkern-2.5mu| \widetilde{\xi}^\#_{Ekl} =
0\;\;\;\forall\,E,\vec{k},l=\pm1
\;. \end{equation}
With these rules, all bilinear expectation values can be
calculated exactly. Higher correlation functions have
a perturbative expansion in the spectral function.
\section*{\large\bf 3. Two-particle correlation function}
Of the higher correlation functions,
we are interested in the two-particle correlation function,
which is the probability to find
in the system a pair of pions with momenta $p$ and $q$.
For the non-equilibrium system we are considering, this
correlation function is
\begin{equation} \label{cfc}
c_{ll^\prime}(\vec{p},\vec{q})=\frac{\left\langle \vphantom{\int}
a^\dagger_{pl}(t)
a^\dagger_{ql^\prime}(t)
a_{ql^\prime}(t) a_{pl}(t)\right\rangle}{
\left\langle \vphantom{\int}a^\dagger_{pl}(t) a_{pl}(t)\right\rangle\;
\left\langle \vphantom{\int}a^\dagger_{ql^\prime}(t) a_{ql^\prime}(t)
\right\rangle}
=1+\delta_{ll^\prime}
\frac{{\cal F}(\vec{p},\vec{q}){\cal F}(\vec{q},\vec{p})
}{{\cal F}(\vec{p},\vec{p}) {\cal F}(\vec{q},\vec{q})}
\;.\end{equation}
For simplicity, we abbreviate the mean momentum of this pair by
$\vec{Q}= (\vec{q}+\vec{p})/2$. The function
${\cal F}(\vec{q},\vec{p})$ is calculated using the
standard rules of thermo field dynamics given above. One
obtains
\begin{equation} \label{cfcf}
{\cal F}(\vec{p},\vec{q}) =
\int\limits_{0}^{\infty}\!dE\int\!d^3\vec{z}\,
\left({\cal A}_l(E,\vec{p})
{\cal A}_l(E,\vec{q})\right)^{\frac{1}{2}}\,
\mbox{e}^{{\mathrm i}(\vec{p}-\vec{q})\vec{z}}\,
n_{l}(E,\vec{z})
\;, \end{equation}
where one may also insert a $\vec{z}$-dependent spectral function
without violating the accuracy to first order in the gradients.
How these gradients enter the above expressions may be seen when
performing an expansion of ${\cal A}$ around the mean momentum $\vec{Q}$.
\begin{eqnarray} \label{cfc2} \nonumber
{\cal F}(\vec{p},\vec{q}) & = & {\cal F}^0(\vec{p},\vec{q}) \\
\nonumber
&+&
\int\limits_{0}^{\infty}\!dE\int\!d^3\vec{z}\,
\mbox{e}^{{\mathrm i}(\vec{p}-\vec{q})\vec{z}}\,\left(
{\mathrm i} \nabla_{\vec{Q}}{\cal A}_l(E,\vec{Q},\vec{z})\,
\nabla_{\vec{z}} n_{l}(E,\vec{z})\right)\\
&+&{\cal O}(\nabla^2_z n)
\;.\end{eqnarray}
Here, the lowest order term
\begin{equation}\label{cfc3}
{\cal F}^0(\vec{p},\vec{q}) =
\int\limits_{0}^{\infty}\!dE\int\!d^3\vec{z}\,
\mbox{e}^{{\mathrm i}(\vec{q}-\vec{p})\vec{z}}\,
{\cal A}_l(E,\vec{Q},\vec{z})\,
n_{l}(E,\vec{z})
\;\end{equation}
is the {\em local equilibrium\/} contribution to the
${\cal F}$ we have obtained above.
For a possible generalization, i.e., to explicitly
momentum dependent $n_l$, it is worthwhile to note
that the gradient term in (\ref{cfc2}) is just one half of the
Poisson bracket of ${\cal A}$ and $n$
\cite{h94rep}. Furthermore, we find that
$
{\cal F}(\vec{p},\vec{p})={\cal F}^0(\vec{p},\vec{p})
$,
i.e., the denominator of the correlation function is not affected by the
gradient expansion.
We therefore obtain as the
{\em local equilibrium two-particle correlation function\/} the expression
\begin{equation}\label{cfcs}
c_{ll^\prime}^{\mbox{\small loc}}(\vec{p},\vec{q})
=1+\delta_{ll^\prime}
\frac{{\cal F}^0(\vec{p},\vec{q}){\cal F}^0(\vec{q},\vec{p})
}{{\cal F}^0(\vec{p},\vec{p}) {\cal F}^0(\vec{q},\vec{q})}
\;.\end{equation}
However, the full {\em non-equilibrium correlation function\/}
$c_{ll^\prime}(\vec{p},\vec{q})$ is the one measured experimentally.
The exact correspondence between the local equilibrium result and
other calculations of the correlator
\cite{KP72,GKW79,P84,DS92} may be found when inserting the
free spectral function from (\ref{fb}):
\begin{equation}\label{cfc4}
c_{ll^\prime}^{\mbox{\small free}}(\vec{p},\vec{q})
=1+\delta_{ll^\prime}
\frac{\displaystyle \left|\int\!d^3\vec{z}\,
\mbox{e}^{{\mathrm i}(\vec{q}-\vec{p})\vec{z}}\,
n_{l}(\omega_{\vec{Q}},\vec{z})\right|^2}{\displaystyle
\left(\int\!d^3\vec{z}\,
n_{l}(\omega_{\vec{p}},\vec{z})\right)
\left(\int\!d^3\vec{z^\prime}\,
n_{l}(\omega_{\vec{q}},\vec{z^\prime})\right)}
\;.\end{equation}
Obviously, one may not insert the free spectral function
into equation (\ref{cfcf}) for ${\cal F}$. This is
{\em not\/} a flaw of the derivation, but suggests -- as expected --
that the limit of zero spectral width at finite temperature is ill-defined
\cite{h94spec}.
\section*{\large\bf 4. Simple spectral function}
In this section we study the difference between the local equilibrium
correlation function (\ref{cfcs})
and the non-equilibrium result (\ref{cfcf}) in more detail.
To this end we calculate the correlation functions
with a simple parameterization of a boson (pion) spectral function,
\begin{equation}\label{sfs}
{\cal A}_l(E,\vec{p})=\frac{2E\gamma}{\pi}
\frac{1}{(E^{2}-\Omega_{p}^{2})^{2}+4E^{2}\gamma^{2}}
\end{equation}
where $\Omega_{p}=\sqrt{m_{\pi}^{2}+\vec{p}^{2}+\gamma^{2}}$ and
$m_\pi$ = 140 MeV. To
gain information about the {\em maximal\/} influence exerted
by the occurrence of a nonzero spectral width, we use
an energy and momentum independent $\gamma$ equal for both charges.
The temperature distribution is taken as a radially symmetric
gaussian,
\begin{equation} \label{tempd}
T(\vec{z})=T(r) = T_0 \exp\left(-\frac{r^{2}}{2R_0^{2}}\right)
\;,\end{equation}
with chemical potential $\mu=0$ and $R_0$ = 5 fm.
The local equilibrium pion distribution
for a given momentum $\vec{k}$ is obtained by folding $n(E,\vec{z})$
with the
spectral function. Hence, the mean radius of the {\em particle\/}
distribution function acquires a $\gamma$-dependence. We define
the rms radius {\em orthogonal\/} to the direction of $\vec{k}$ as
\begin{equation}\label{rav}\nonumber
R_{\mbox{\small rms}} = \sqrt{\frac{I_2}{I_0}} \;\;\;\;\;\;\;
I_j = \int\limits_0^\infty \!\!dr\,r^j\,\int\limits_0^\infty
\!\!dE\,{\cal A}(E,\vec{k})\,\left(\mbox{e}^{E/T(r)}-1\right)^{-1}
\;.\end{equation}
Note, that $R_{\mbox{\small rms}}$ is {\em not\/} the 3-dimensional
rms radius of the distribution function (which would be $I_4/I_2$).
Rather, $R_{\mbox{\small rms}}$ is half the product of angular
diameter and distance between detector and source.
A constant temperature over a sphere of radius $R_0$
would yield an $R_{\mbox{\small rms}} = R_0/\sqrt{3}$, while
its 3-D rms radius is $R_0 \sqrt{3/5}$.
In fig. \ref{fig1} we have plotted the correlation functions
for two different constant values of the parameter $\gamma$.
Clearly, for $\gamma$=50 MeV the correlation function
$c_{ll^\prime}(\vec{p},\vec{q})$ agrees quite well with the
local equilibrium result, and very closely resembles a gaussian.
However, for smaller $\gamma$=5 MeV the non-equilibrium correlation function
is much narrower in momentum space than the local equilibrium result,
it also deviates from a gaussian form. Nevertheless we may
approximate it by such a simple functional form in order to extract
quantitative information, i.e.,
\begin{equation}\label{gau}
c_{ll^\prime}(\vec{p},\vec{q}) \approx
1 + \exp\left( -R^2 (\vec{p}-\vec{q})^2 \right)
\;\end{equation}
and similarly for $c_{ll^\prime}^{\mbox{\small loc}}(\vec{p},\vec{q})$
with parameter $R_{\mbox{\small loc}}$.
In figure \ref{fig2} we show the two fit parameters $R$ and
$R_{\mbox{\small loc}}$ as function of $\gamma$, together
with the $\gamma$-dependent rms radius of the particle distribution.
We find that the {\em measured\/} correlation radius $R$
is always larger than $R_{\mbox{\small loc}}$, with a minimum
reached at $\gamma\approx 26.7$ MeV.
The plot may be divided in two regions, with a boundary at
$
2\gamma R_{\mbox{\small rms}}[\gamma] = 1 \Leftrightarrow
\gamma\approx 30.2$ MeV. For these two regions we find
\begin{equation}
{\array{lllll}
2\gamma R_{\mbox{\small rms}} \;\ll\; 1\;&\;\Rightarrow\;&\;
R \approx R_{\mbox{\small loc}} +
1/\gamma \; &\;\gg\; R_{\mbox{\small rms}}\;&\; >
\;R_{\mbox{\small loc}}\\
2\gamma R_{\mbox{\small rms}}\;\gg\; 1\;&\;\Rightarrow &\;
R \approx R_{\mbox{\small loc}}
&\stackrel{>}{\sim}\; R_{\mbox{\small rms}}& >\;
\;R_{\mbox{\small loc}}
\;.\endarray}
\end{equation}
For larger $\gamma$, the small differences between $R$,
$R_{\mbox{\small loc}}$ and $R_{\mbox{\small rms}}$ may be attributed to our
use of a gaussian temperature distribution:
$n(T(r))$ is not strictly gaussian, only in the (unphysical) limit
$\gamma\rightarrow\infty$ one reaches
$R=R_{\mbox{\small loc}}=R_{\mbox{\small rms}}=R_0/\sqrt{2}$
The interpretation of this result is straightforward:
A finite lifetime
or nonzero spectral width $\gamma>0$ of the bosons {\em inside the
source\/} is essential,
if one wants to infer the {\em thermal source radius\/} $R_0$
from correlation measurements. To be more precise,
only for $2\gamma R_{\mbox{\small rms}} \ge 1$
the correlation function measures
the mean diameter of the particle distribution function $n(E,\vec{z})$.
This result is in agreement with our view of the {\em relaxation
process\/} of a non-equilibrium distribution function:
The relaxation rate is, to
lowest order, given by the spectral width of the particle \cite{h94rep}.
Consequently, {\em zero\/} $\gamma$ corresponds to a system
without dissipation. In such a system, only quantum coherence effects
exist, and thus the correlation function approaches
the ``quantum limit''
\begin{equation}\label{qaml}
\lim_{\gamma\rightarrow 0}
c_{ll^\prime}(\vec{p},\vec{q}) = 1 +
\delta_{ll^\prime} \delta_{\vec{p}\,\vec{q}}
\;.\end{equation}
For this case,
the correlation radius obtained by a gaussian fit becomes infinite.
We may also view, for a given energy, $1/\gamma$ as a
measure for the spatial size of the pion ``wave packet'', which must
be smaller than the object to be resolved. In other words, the
mean free path of the bosons must not exceed the object size to produce
correlations.
\section*{\large\bf 5. Semi-realistic spectral function}\label{real}
To get a more realistic result for the non-equilibrium two-pion correlation
function measured in the thermal radiation from a hot spot
in nuclear matter, we use the the spectral function derived
in \cite{h94rep,h93pion}. It includes the coupling of pions to
$\Delta_{33}$-resonances in nuclear matter, which are taken to
have a constant spectral width $\Gamma$ by themselves.
It was argued in ref. \cite{h94rep}, that using such an approximate
spectral width for the $\Delta_{33}$ resonances constitutes
the only way to achieve an analytical solution of the
$\Delta$-hole polarization problem in nuclear matter. An even
more realistic energy-momentum dependent spectral width for
the $\Delta_{33}$ resonance can be treated only in a fully self-consistent
numerical treatment involving dispersion integral techniques.
To first order in such a constant $\Gamma$, the pionic spectral
function is
\begin{equation}\label{asys}
{\cal A}(E,\vec{k},\vec{z}) =
\frac{\vec{k}^2C(\vec{z})}{\pi}\,\frac{\Gamma\,E\,\omega_\Delta}{
(E^2-\omega^{\prime\,2}_+)^2\,
(E^2-\omega^{\prime\,2}_-)^2 + \Gamma^2\,E^2\,(E^2-E^2_\pi)^2}
\;.\end{equation}
Note, that this function is coordinate dependent.
The energies in the denominator are
\begin{equation}\label{omf2}
\omega^{\prime\, 2}_\pm = \frac{1}{2}\left(E_{N\Delta}^2
+(\Gamma/2)^2+E_\pi^2 \pm
\sqrt{ \left(E_{N\Delta}^2+(\Gamma/2)^2
-E_\pi^2\right)^2 + 4\vec{k}^2 C(\vec{z}) \omega_\Delta}
\right)
\;,\end{equation}
with functions
\begin{eqnarray}\nonumber\label{odf2}
\omega_\Delta & = & E_\Delta(\vec{k}) -M_N
= \sqrt{\vec{k}^2+M_\Delta^2}-M_N\\
\nonumber
C(\vec{z}) & = &
\frac{8}{9}\left(\frac{f^\pi_{N\Delta}}{m_\pi}\right)^2\,
\left( \rho^0_N(\vec{z})-\frac{1}{4}\rho^0_\Delta(\vec{z})\right)\\
E_{N\Delta}(\vec{k}) & = & \vphantom{\int}\sqrt{
\omega_\Delta(\vec{k})
\left( \omega_\Delta(\vec{k}) + g^\prime\,C(\vec{z})\right)}
\;.\end{eqnarray}
\begin{table}[t]
\begin{center}
\begin{tabular}{ccccccc} \hline
~$f^\pi_{N\Delta}$~ & ~$g^\prime$~ & ~$m_\pi$~ & ~$M_N$~ &
~$M_\Delta$~ & ~$\Gamma$~ & $\rho_0$\\ \hline
2 & ~0.5~ & ~0.14 GeV~& ~0.938 GeV~ &
~1.232 GeV~ & ~0.12 GeV~& 0.155 fm ${}^{-3}$ \\ \hline
\end{tabular}
\end{center}
\caption{Coupling constants and masses used in the calculations of
this work.}
{\small
The value of $g^\prime$ was chosen to allow for a direct
comparison with simulation codes for heavy-ion collisions,
a more realistic value to describe
pion scattering data would be $g^\prime\approx 1/3$.}\\ \hrule
\end{table}
The baryon number in each small volume and hence the baryon density is a
constant parameter of the calculations,
for free particles
with bare ''on-shell'' energies
\begin{equation}
E_N(\vec{p}) = \sqrt{\vec{p}^2 + M_N^2}\;\;\;\mbox{and}\;\;\;
E_\Delta(\vec{p}) = \sqrt{\vec{p}^2 + M_\Delta^2}
\;\end{equation}
this baryon density is obtained as
\begin{eqnarray}\nonumber\label{rb0}
\rho_b^0& =& \rho^0_N(\vec{z}) + \rho^0_\Delta(\vec{z}) \\
& =& 4\,\int\!\!\frac{d^3\vec{p}}{(2\pi)^3}\,n_N(E_N(\vec{p}),\vec{z})
+16\,\int\!\!\frac{d^3\vec{p}}{(2\pi)^3}\,
n_\Delta(E_\Delta(\vec{p}),\vec{z})
\;.\end{eqnarray}
The distribution functions are taken as local Fermi-Dirac functions
\begin{equation}
n_{N,\Delta}(E,\vec{z}) =
\frac{1}{{\mathrm e}^{(E-\mu_{N,\Delta})/T(\vec{z})}+1}
\;,\end{equation}
with temperature $T(\vec{z})$.
As temperature distribution we use the same as
in the previous section, eq. (\ref{tempd}), but with different central
temperatures $T_0$.
\begin{table}[b]
\hrule
\begin{center}
\begin{tabular}{llllll}
$T$ & $(\vec{p}+\vec{q})/2$ & $R$ & $R_{\mbox{\small loc}}$ &
$R_{\mbox{\small rms}}$ & $R/R_{\mbox{\small rms}}$ \\ \hline
100 MeV & 100 MeV & 3.99 fm & 2.89 fm & 3.14 fm &1.27 \\
160 MeV & 100 MeV & 4.86 fm & 3.23 fm & 3.51 fm &1.38 \\ \hline
100 MeV & 350 MeV & 3.23 fm & 3.16 fm & 3.34 fm &0.97 \\
160 MeV & 350 MeV & 3.57 fm & 3.49 fm & 3.73 fm &0.96 \\ \hline
\end{tabular}
\end{center}
\caption{Correlation radii obtained by gaussian fit to
the correlation function with semi-realistic spectral function.}
\label{tab2}
\end{table}
Since the temperature depends on the spatial coordinate,
fixed baryon density implies that the ``baryochemical''
composition of the hot spot changes with coordinate $\vec{z}$.
In the center of the hot spot baryons are, to a large extent,
present in the form of $\Delta$-resonances. Outside the
hot region baryons are ``only'' nucleons -- and as shown before, the
pion properties there do not influence our calculation.
In fig. \ref{fig3}, we have plotted the correlation function
for a pair momentum of $\vec{Q}=(\vec{p}+\vec{q})/2 =$ 100 MeV
at the two values $T_0$= 100 MeV and $T_0$ = 160 MeV.
In this momentum region, the spectral width of the pion is small
due to its pseudo-vector coupling with baryons. This we
assume to be a general feature of pions in nuclear matter,
although one may argue about the exact value of $\gamma$.
We find, that at higher temperature the non-equilibrium correlation
function (which we had assumed to be the one measured experimentally)
is much narrower in momentum space than the local equilibrium
function.
The local equilibrium correlation function
however is close to the the rms-radius of the thermal source.
The correlation radii obtained by a gaussian fit to the
non-equilibrium as well as the local equilibrium distribution are given
in table \ref{tab2}. Following these results we conclude, that the
effect we propose is absent at higher momentum of the pions,
where the p-wave coupling to nuclear matter is big enough to
give it a sufficiently large spectral width for local
equilibration. In the low momentum region, the
{\em measured\/} correlation radius
overestimates the source size by as much as 30 - 40 \%.
\section*{\large\bf 6. Conclusions}
Before we draw a final conclusion from our work, we must emphasize that
it is still too early to use our result for the correlation function in
a direct comparison to experimental data.
For any realistic situation, we certainly have to take into account
also the partially {\em coherent\/} production of pions:
A substantial fraction of
pions arriving in a detector stems from the free-space decay
of $\Delta_{33}$ resonances, thus forcing
$c_{ll^\prime}(\vec{p},\vec{p})<2$.
Furthermore it seems worthwhile to note that our
derivation does not contradict existing
work on the correlation function. In
an early semiclassical treatment, an infinite lifetime
(equivalent to $\gamma\equiv0$)
of an excitation was found to produce a correlation function $c \equiv 1$
\cite{KP72}, whereas we find an infinitely narrow peak in this
``quantum limit'', see eq. (\ref{qaml}.
Another paper based on the Schwinger-Keldysh method also finds
that the total production rate of particles is proportional to
an energy integral over the off-diagonal self-energy components,
i.e., to the integral over $\gamma\cdot n$ in our formalism \cite{DS92}.
However, due to their explicit quasi-particle approximation
the authors of ref. \cite{DS92} lose the consistent treatment
of the non-equilibrium effects
to first order in the temperature gradients.
Let us briefly remark on the physical situation
present in {\em stars\/} emitting photons:
For these we may assume a thermal width of the photon which
is very small, e.g., $\gamma\approx \alpha T\approx 0.5/137$ eV.
However, the source radius $R_{\mbox{\small rms}}$ of a star is, in general,
so big that the condition $2\gamma R_{\mbox{\small rms}} \gg 1$
is always satisfied. Hence our results do {\em not\/} affect correlation radii
measured for astronomical objects.
Having in-lined our calculation with existing work, we may now
carefully conclude the following physical effect relevant
for relativistic heavy-ion collisions:
When pion pairs with a relatively low
momentum are created in hot nuclear matter, their
p-wave interaction with the surrounding medium is small.
Hence, in their movement outwards from the hot zone they do
not have a sufficiently short mean free path to be locally
equilibrated. Thus, their correlation length is dominated by
their mean free path ($\approx 1/\gamma$) rather than
by the thermal source radius.
As we have shown, this leads to a correlation function
which is {\em narrower\/} in momentum space than expected
in a local equilibrium situation: Compare the solid and the
dashed curves in fig. \ref{fig3}. Consequently, measuring
the correlation function of low momentum pions in
a non-equilibrium state
{\em overestimates\/} the source radius. For a semi-realistic
pion spectral function and a pair momentum of 100 MeV, we find
this effect to be as large as 30 -- 40\%, depending strongly
on the actual source size.
Turning to experimental results of NA44, we learn that
correlation measurements of pions and K-mesons emanating from
relativistic heavy-ion collisions yield comparable fitted
correlation radii of 3--4 fm \cite{NA44}. As we have shown, it
might be premature to conclude from these measurements
that the {\em source size\/} for both mesons is similar: One
would have to compare the mean free path of both particle species before
such a conclusion.
More recently the experiment NA49 has measured pion correlation
functions in central Pb+Pb collisions at 33 TeV. Resulting was
a correlation radius of 6-7 fm, as compared to 4.5 fm
in S+Au collisions \cite{NA49}. It is a particularity of our
semi-realistic pion spectral function that it is narrower for
higher temperature and low momenta. Physically, the first
effect stems from the higher $\Delta_{33}$ abundance in the hotter system:
$\Delta$-particle/nucleon-hole pairs are less easily polarized.
The second effect is due to the p-wave coupling of pions to nucleons,
as was pointed out above.
Consequently, the effective $\gamma$ of the pions drops with
increasing temperature -- and one measures a higher correlation radius.
If we presume this effect, which was deduced for {\em equilibrated\/}
matter, also to hold in the highly non-equilibrium situation
present in the experiments, we can state that the higher correlation
radius of the NA49 data might indicate a higher temperature
of the reaction zone rather than its bigger size.
However, this interpretation depends crucially on the
momentum range where the pions are measured.
The uncertainties in this statement indicate
the importance of calculations for mesonic spectral functions
in hot nuclear matter.
The effect of narrowing the correlation function
might be even more pronounced due to the short time-scales
of relativistic heavy-ion collisions, which may suppress the spectral
width ($\simeq$ collision rate) of pions even more. Indications for
such a behavior are obtained in simulation codes, where at high enough
collision energies one may reproduce experimental data with
{\em free\/} particle collision rates. Thus, instead of speaking about
``temperature'', which always is connected with the notion of a partial
equilibrium, we may also reformulate our conclusion even more
sharply: Measured correlation functions become narrower due
to non-equilibrium effects in statistically radiating sources.
A bigger correlation radius therefore is an indication, that
a system of colliding heavy-ion collisions is farther from
local equilibrium in the collision zone.
Finally we point out that in common calculations
of the correlation function one has to introduce ad-hoc
random phases between several classical sources and then obtains
only the local equilibrium correlator
$c^{\mbox{\small loc}}_{ll^\prime}(\vec{p},\vec{q})$. Instead, we
relied on a proper field theoretical treatment which incorporates
{\em non-equilibrium effects} correctly up to first order in
gradients of the temperature. We found, that the non-equilibrium character
of the system must be taken serious when calculating the correlation
function -- which answers the secondary question
we asked in the introduction.
\subsection*{\normalsize\bf Coda}
A large part of the work of E.Wigner was dedicated to the study
of symmetries in physical systems. The fact, that particles
in a thermal state must have a nontrivial spectral function
rather than a sharp ''mass-shell constraint'', is connected
to the breaking of such a symmetry: A thermal state has a
preferred rest frame and therefore violates the Lorentz invariance
of the usual field theoretical ground state \cite{BS75}.
|
\section{Introduction}
\subsection{Classical laws and dissipative chaos.}
Recently, Gell-Mann and Hartle, among others \cite{GMHart,DowkHall,Brun},
have studied the problem of classical laws arising from quantum theory in the
light of the {\it decoherence formalism}.
In this approach, one considers possible
histories of a given system to which probabilities can be assigned that obey
classical probability sum rules. In order for histories to {\it decohere}
in this way, it is usually necessary to coarse-grain the description of the
system, by giving the values of its variables only at certain times, or
averaged over certain intervals, or by neglecting certain variables and
retaining others, or a combination of all of these.
They have found that within this formalism it is possible to define in a very
rigorous way the classical equation of motion based only on the underlying
quantum theory. In doing so, both dissipation and noise typically appear,
arising as a consequence of coarse-graining over neglected degrees of
freedom. This takes advantage of the well-studied phenomenon of
environmentally induced decoherence \cite{Zurek}.
In addition to casting new light on the problem of how classical
laws of physics arise, this provides an unparalleled tool for studying
quantum mechanical systems with dissipation, and seeing how this alters
their behavior from the more usual Hamiltonian behavior of closed systems.
One area which can profitably be treated this way is dissipative chaos.
There has been an enormous amount of work done on ``Quantum Chaos,''
i.e., quantizing nonintegrable Hamiltonians which classically exhibit
chaotic behavior. This has turned up beautiful connections between
classical chaotic behavior and their quantum quasiperiodic equivalents.
But very little has been done in looking at the quantum versions of systems
which classically exhibit dissipative chaos,
or on looking at their classical
limit \cite{Graham,ZhengSavage}. Classical dissipative chaos is
qualitatively very different from Hamiltonian chaos, and one would
expect their quantum equivalents to reflect this difference, but
this has not been widely investigated.
Indeed, even very extensive treatments
of quantum chaos rarely deal with dissipative systems at all
\cite{Gutzwiller}.
There are a number of reasons for this. The first is that dissipation is
difficult to treat in normal quantum mechanics.
The usual Schr\"odinger equation
is only valid for closed systems without friction. Open systems in general,
and dissipation in particular, can be handled using the influence functional
approach of Feynman and Vernon \cite{FeynVern}; this has been done in the
case of Brownian motion by Caldeira and Leggett \cite{CaldLegg} among others.
This approach has not been widely used, though, until recently,
as it involves considerable
conceptual and mathematical baggage \cite{Zurek}. Also, the types of
behavior of most interest to those who study quantum Hamiltonian chaos
involve the coherent evolution of the wave function, with its attendant
complicated structure (e.g., the ``scarring'' of energy eigenfunctions
about classical periodic orbits, the statistics of energy level spacing).
The presence of strong damping wipes out this coherent structure.
Chirikov et al. have summed up the usual attitude towards quantum dissipative
chaos: ``In what follows we will discuss only Hamiltonian (nondissipative)
systems, considering them to be the more fundamental ones. Phenomenological
friction is but a crude approximation of the molecular Hamiltonian chaos
which is inevitably related to some noise according to the
fluctuation-dissipation theorem.'' And further, they divide the problem
of quantum chaos into two parts, the quantum dynamics of the wave function
in isolation, and the results of measurement ``with its unavoidable
statistical effect of the irreversible $\psi$ collapse which is a sort of
inevitable noise.'' \cite{Chiretal}
While this is undeniably true, most systems are {\it not} isolated, and
so it is perhaps useful to consider systems for which dissipation is
important. Dissipative chaotic systems may not be fundamental, but they
are nevertheless interesting.
Decoherence is an appropriate formalism in which to
study them \cite{Brun2}.
In the rest of this section I give a brief introduction to the decoherent
histories formalism of Gell-Mann and Hartle. Then in section 2 I derive
a model for a quantum forced, damped nonlinear oscillator, following
the usual system/environment coarse-graining. In section 3 I discuss the
classical properties of the forced, damped Duffing oscillator, and
describe some of the properties of dissipative chaos which it exhibits.
In section 4 I treat the quantum version of this problem, and show
how one can make close contact with the classical theory using the
decoherent histories formalism. In section 5 I illustrate this with a
numerical example, and in section 6 I summarize my conclusions.
\subsection{Decoherent histories.}
In the formalism of decoherent histories, systems are described by a
set of exclusive and exhaustive histories $\{\alpha\}$, which can be
thought of as different possibilities for the system's evolution. While
there is a vast range of possible sets of histories to choose from, these
sets are restricted by the {\it decoherence condition}
\begin{equation}
D[\alpha,\alpha'] = p_\alpha \delta_{\alpha\alpha'},
\end{equation}
where $D[\alpha,\alpha']$ is the {\it decoherence functional} and
$p_\alpha$ is the probability of history $\alpha$ occurring. This
decoherence condition restricts one to histories which obey the usual
probability sum rules; these histories do not interfere with each other.
To make this more explicit, in ordinary nonrelativistic quantum mechanics
one can specify a history by enumerating a complete set of orthogonal
projection operators $\{P^i_{\alpha_i}(t_i)\}$ at a sequence of times
$t_i$. A single history is then given by choosing one projection operator
at each time. This is equivalent to enumerating a set of possible assertions
about the system at a sequence of times, and having each history be a
string of such assertions. One can define a {\it history operator}
\begin{equation}
C_\alpha = P^n_{\alpha_n}(t_n)
\cdots P^2_{\alpha_2}(t_2) P^1_{\alpha_1}(t_1),
\end{equation}
where $\alpha$ is a shorthand for the choices $\alpha_i$ at times $t_i$.
The decoherence functional is then
\begin{equation}
D[\alpha,\alpha'] =
{\rm Tr} \bigl\{ C_\alpha \rho C^\dagger_{\alpha'} \bigr\}.
\label{DecoherenceFunctional0}
\end{equation}
The density operator $\rho$ is the system's initial condition.
As a rule, it is impossible for very fine-grained histories to decohere;
thus, considerable coarse-graining is required. One very common
coarse-graining used to study decoherence in systems with many
degrees of freedom is to completely trace out certain freedoms (the
``environment'') while leaving others completely fine-grained (the
``disinguished subsystem''). This was first studied by Feynman and Vernon
\cite{FeynVern} and applied to decoherence by Zurek \cite{Zurek} among
others. We will initially be considering this type of coarse-graining.
\section{The Model}
The particular model we will study is based on earlier work on decoherence
in systems with dissipation \cite{Zurek,GMHart,Brun,Brun2}.
In this model we will divide our system into a distinguished variable $x$,
termed the {\it system variable}, and a set of {\it reservoir variables}
$\{ Q_k \}$ which we will trace over.
This system and reservoir will have a total action
\begin{equation}
S[x(t),{\bf Q}(t)] = S_{\rm sys}[x(t)] + S_{\rm res}[{\bf Q}(t)]
+ \int_{t_0}^{t_f} V_{\rm int}(x(t),{\bf Q}(t)) dt,
\end{equation}
where the system variable will be treated as a particle moving in a potential
\begin{equation}
S_{\rm sys}[x(t)] = \int_{t_0}^{t_f} \biggl( {M\over2}{\dot x}^2(t) -
U(x(t)) \biggr) dt,
\end{equation}
the reservoir is approximated as a collection of harmonic oscillators
\begin{equation}
S_{\rm res}[{\bf Q}(t)] = {m\over2} \sum_k \int_{t_0}^{t_f} \biggl(
{\dot Q}_k(t)^2 - \omega_k^2 Q_k(t)^2 \biggr) dt,
\end{equation}
and the interaction is linear in $x$ and {\bf Q}:
\begin{equation}
V_{\rm int}(x,{\bf Q}) = - x \sum_k \gamma_k Q_k.
\end{equation}
We will make the additional assumption that the initial density matrix of
the system and reservoir factors, and that the reservoir is initially in a
thermal state. Then $\rho_{\rm total}(x,{\bf Q}; x^\prime, {\bf Q}^\prime) =
\chi(x; x^\prime) \psi_0({\bf Q}; {\bf Q}^\prime)$, where $\psi_0 = \rho_T$
is a thermal density operator at temperature $T$.
The {\it decoherence functional} in this coarse-graining is then
\begin{equation}
D[x^\prime(t),x(t)] = \exp{i\over\hbar}\biggl\{ S_{\rm sys}[x^\prime(t)]
- S_{\rm sys}[x(t)] + W[x^\prime(t),x(t)] \biggr\} \chi(x_0; x^\prime_0).
\label{DecoherenceFunctional}
\end{equation}
$W[x^\prime(t),x(t)]$ is the {\it influence phase}, which includes the
collective effects of the traced-over reservoir degrees of freedom. As
shown by Caldeira and Leggett \cite{CaldLegg}, this functional is
\begin{eqnarray}
&& W[x^\prime(t),x(t)] = \sum_k {{i\gamma_k^2}\over{m\omega_k}}
\coth(\hbar\omega_k/kT) \nonumber\\
&& \times \int_{t_0}^{t_f} dt
\int_{t_0}^{t_f} ds\ \cos(\omega_k(t-s)) (x^\prime(t) - x(t))
(x^\prime(s) - x(s)) \nonumber\\
&& - {{\gamma_k^2}\over{2m\omega_k}} \int_{t_0}^{t_f} dt \int_{t_0}^t ds\
\sin(\omega_k(t-s)) (x^\prime(t) - x(t)) (x^\prime(s) + x(s)).
\end{eqnarray}
We can now switch to new variables
$X = (x + x^\prime)/2$ and $\xi = x - x^\prime$.
In these variables
\begin{eqnarray}
&& S_{\rm sys}[x^\prime(t)] - S_{\rm sys}[x(t)] = \nonumber\\
&& \int_{t_0}^{t_f} \xi(t) \biggl(
- M{\ddot X}(t) - {{dU}\over{dt}}(X(t)) \biggr) dt
- \xi_0 M{\dot X}_0 + O(\xi^3),
\end{eqnarray}
which as we see contains the Euler-Lagrange equations. We also go to a
continuum of oscillator frequencies with a Debye distribution, in which the
discrete sums become integrals over a weighting function $g(\omega) =
\eta \omega^2 \exp(-\omega/\Omega)$, where $\Omega$ is a large cutoff
frequency, so that $1/\Omega << t_f - t_0$. In this limit,
the influence phase becomes
\begin{eqnarray}
&& W[X(t),\xi(t)] = \int_{t_0}^{t_f} \xi(t) \biggl( - M \Lambda X(t)
- 2 \Gamma {\dot X}(t) \biggr) dt \nonumber\\
&& + i K \int_{t_0}^{t_f} \xi^2(t) dt + O(\Omega^{-2}).
\end{eqnarray}
where $\Lambda = \eta\Omega/m$, $\Gamma = \pi\eta/4mM$,
and $K = 4M\Gamma kT/\hbar$. The
$\Lambda$ term has the form of a linear force; it can be absorbed
into the system action by going to an effective potential
\begin{equation}
U_{\rm eff}(X) = U(X) + M \Lambda X^2/2.
\end{equation}
The $\Gamma$ term has the form of a dissipation.
The imaginary term is of particular interest.
It suppresses $D[X(t),\xi(t)]$ when $\xi \ne 0$. Since $\xi \ne 0$
corresponds to the ``off-diagonal'' terms of the decoherence functional
$(x(t) \ne x^\prime(t))$, the suppression of these terms results in
approximate decoherence of this set of histories.
This suppression of off-diagonal terms is
clearly related to the presence of noise \cite{Zurek,GMHart}.
The kernel of this term can be identified with
the two-time correlation function of
a stochastic driving force $F(t)$ in the
quasiclassical limit.
This correlation function is
\begin{equation}
\langle F(t) F(s) \rangle = \hbar K \delta(t-s)
\end{equation}
in the continuum case, with $\langle F(t) \rangle = 0$. So in the
quasiclassical limit this system obeys the classical equation of motion
\begin{equation}
{\ddot x} + {1\over M}{dU_{\rm eff}\over{dx}}(x) + 2\Gamma{\dot x}
= F(t)/M.
\label{EOM1}
\end{equation}
Instead of taking the reservoir to be in a thermal state initially, we
can take it to be in a {\it displaced} thermal state,
\begin{equation}
\rho_{\rm DT} = {\hat D}(q(\omega),p(\omega)) \rho_T
{\hat D}(q(\omega),p(\omega))^\dagger,
\end{equation}
${\hat D}(q(\omega),p(\omega))$ is the
coherent state displacement operator:
\[
{\hat D}(q(\omega),p(\omega)) | 0 \rangle = | q(\omega),p(\omega) \rangle,
\]
where $| q(\omega),p(\omega) \rangle$ is the coherent state centered on
$(q(\omega),p(\omega))$ at frequency $\omega$.
If we take $q(\omega)$ to be sharply peaked around a certain
frequency, $q(\omega) = q \delta(\omega - \omega_0)$,
and $p(\omega) = 0$, then in the
quasiclassical limit the above equation of motion (\ref{EOM1}) gains an
additional term
\begin{equation}
{\ddot x} + {1\over M}{dU_{\rm eff}\over{dx}}(x) + 2\Gamma{\dot x}
= q \cos(\omega_0 t) + F(t)/M.
\label{EOM2}
\end{equation}
This is exactly the form for a nonlinear oscillator with damping and a
periodic driving force, with additional noise.
In a truly classical system,
$F(t)$ would vanish as $T\rightarrow0$,
but in the quantum theory noise is always present, even at absolute zero.
One can think of it as arising from the
zero-point oscillations of the reservoir oscillators. At low temperatures,
however, the two-time correlation function of the noise is
highly {\it nonlocal in time}. At $T = 0$,
\begin{equation}
{\rm Re}\ W[X(t),\xi(t)] \sim \int_{t_0}^{t_f} dt \int_{t_0}^{t_f} ds\
\xi(t)\xi(s)/(t-s)^2.
\end{equation}
Correlations in the noise persist for all times. Because of this form of the
kernel, doing exact (or even numerical) calculations in the low-temperature
limit is extremely difficult. This is why the high $T$ limit is generally
used.
\section{The Classical Forced Damped Duffing Oscillator}
We are interested in finding quantum equivalents to classical systems
which exhibit dissipative chaos. While many such systems (e.g., fluid
mechanics) have no easily realizable quantum limit, there are some
which can be readily quantized as shown in section 2. These are the
nonlinear oscillators with damping and driving.
One much-studied classical nonlinear oscillator is the Duffing oscillator,
characterized by a two-welled polynomial potential,
\begin{equation}
U(x) = {x^4\over4} - {x^2\over2}.
\end{equation}
With forcing and damping, this gives an equation of motion
\begin{equation}
{\ddot x} + (x - x^3) + 2\Gamma{\dot x}
= q \cos(\omega_0 t),
\label{EOM3}
\end{equation}
where we take $M=1$. This system is chaotic for certain values of $q$,
$\Gamma$, and $\omega_0$ \cite{GuckHolmes}. For example, $q=0.3$,
$\Gamma=0.125$ and $\omega_0 = 1.0$ is a common choice.
Since this system has explicit time dependence,
its phase space is three-dimensional, $(x,p,t)$. It is
common to discretize the dynamics by taking a {\it Poincar\'e section},
considering only the points on a surface of constant phase $(x_i, p_i)$ at
times $t_i = 2\pi i/\omega_0$. The continuous dynamics defines a {\it map}
${\bf f}$:
\begin{equation}
(x_i, p_i) \rightarrow (x_{i+1}, p_{i+1}) = {\bf f} (x_i, p_i).
\label{poincare_map}
\end{equation}
If the oscillator is non-chaotic, there is a stable attracting fixed point
or group of periodic points to which the $(x_i, p_i)$ quickly tend.
These points
correspond to a periodic orbit of the continuous dynamics. When the
oscillator becomes chaotic, the stable set becomes a {\it strange
attractor}, a fractal structure with non-periodic behavior. There are, in
addition, an infinite number of unstable fixed points and periodic points.
(See figure 1.)
We can also look at the classical dynamics from the point of view of
probability measures $P(x,p)$ on phase space. The map ${\bf f}$ of
points in phase space induces a map on probability measures
\begin{equation}
P_i(x,p) \rightarrow P_{i+1}(x,p) = \int dx^\prime dp^\prime\
\delta( (x,p) - {\bf f}(x^\prime,p^\prime) ) P_i(x^\prime, p^\prime).
\label{ProbabilityMeasure}
\end{equation}
By means of this sort of map we can readily make contact with the quantum
theory.
Of particular interest are
{\it invariant} probability measures $P_{\rm inv}$.
There are many of these, most corresponding to unstable fixed points and
periodic points of the map ${\bf f}$. It is possible to eliminate these
unstable solutions by including a small amount of noise in the equation of
motion (\ref{EOM3}). This effectively broadens the delta functions in
(\ref{ProbabilityMeasure}) into peaks of finite width $\epsilon$, and
eliminates all unstable solutions, leaving a single unique $P_{\rm inv}$
corresponding to the strange attractor. Classically, we can then allow
the noise to go to zero, and look at $P_{\rm inv}$ in the zero-noise
limit. In that limit, the invariant probability measure becomes a
generalized function with substructure at all length scales.
It is a fractal.
\section{Decoherent Histories, Quantum Maps, and Probability}
At best, the functional described in (\ref{DecoherenceFunctional}) can only
be approximately decoherent. Clearly, off-diagonal terms will not vanish
for sufficiently small $|\xi(t)|$. More coarse-graining is needed in the
description of $x(t)$ and $x^\prime(t)$. Also, specifying a value, even an
approximate value, of $x(t)$ for all times $t$ is an extreme fine graining.
It is more common to instead specify $x$ at a series of discrete times $t_i$.
Thus, instead of a complete trajectory $x(t)$ one gives only a series of
$x$ values $\{x_i\}$. Coarse-graining in position as well, one could divide
up the range of $x$ into finite non-zero intervals $\Delta_j^i$,
where $j$ is an index specifying which interval $x$ fell in at time $t_i$.
A history would now be a series of indices $\{\alpha_i\}$, specifying that
$x$ fell in the interval $\Delta_{\alpha_i}^i$ at time $t_i$.
Note that to achieve decoherence, these times $t_i$ cannot be too close
together; they must generally be separated by at least the {\it
decoherence time} \cite{GMHart}. For high temperature systems this
is typically quite short, of the order $\hbar^2/2M\Gamma kT d^2$,
where $d$ is the size of the intervals.
Such a coarse-graining gives us a new decoherence functional:
\begin{equation}
D[\alpha,\alpha^\prime] = \int_\alpha \delta x
\int_{\alpha^\prime} \delta x^\prime \
D[x(t),x^\prime(t)],
\label{DecoherenceFunctional2}
\end{equation}
where the limits specify integration only over those paths which pass
through the series of intervals $\Delta^i_{\alpha_i}$ at the times $t_i$.
The probability of a given history $\alpha$ is of course given by the
diagonal terms of this functional. Since the original decoherence functional
given by (\ref{DecoherenceFunctional}) has an exponent quadratic in $\xi$,
the path integrals over $\xi$ can be carried out; we then let $\alpha =
\alpha^\prime$ and get
\begin{equation}
p(\alpha) = \sqrt{2\pi\over K} \int_\alpha \delta X\ \exp\biggl\{
- {1\over{K\hbar}} \int_{t_0} e^2(t) dt \biggr\} w(X_0,M{\dot X}_0 ),
\label{Probability}
\end{equation}
where $e(t) = M {\ddot X} + (dU_{\rm eff}/dt)(X) + 2M\Gamma{\dot X}
-q\cos(\omega_0 t)$ is the right-hand side of the equation of motion
\cite{Brun2}.
{}From this we see that the probability will by peaked about histories which
approximately obey the classical equation of motion $e(t) = 0$, more and
more sharply as we approach the classical limit where $M$ is large.
The $w(X_0, M{\dot X}_0)$ is the initial Wigner distribution of the system.
The Wigner distribution is defined in terms of the density matrix:
\begin{equation}
w(X,p) = {1\over\pi} \int e^{- i\xi p/\hbar} \chi(X+\xi/2; X-\xi/2)\ d\xi.
\end{equation}
The distribution behaves very
similarly to a classical probability distribution
on phase space, except that $w(X,p)$ can be locally negative (though it must
sum to 1 over all of phase space, and be non-negative on average over regions
with volumes larger than $\hbar$). The expectation values of functions of
$X$ and $p$ can be calculated by averaging them over phase space using
$w(X,p)$ as a weighting function,
though there is usually some ambiguity about
the ordering of operators. As one goes to the classical limit, on scales
large compared to $\hbar$, this ambiguity becomes unimportant.
An interesting way of looking at this system is in terms of the evolution of
the Wigner distribution with time.
If we consider surfaces of constant phase,
as in the classical case, we can define a {\it quantum map},
\begin{eqnarray}
&& w_i \rightarrow w_{i+1} = {\bf T} w_i, \nonumber\\
&& w_{i+1}(X_1,p_1) = \int dX_0 \int dp_0\ T(X_1,p_1; X_0, p_0) w(X_0,p_0).
\end{eqnarray}
The transition matrix ${\bf T}$ is defined by the path integral
\begin{eqnarray}
&& T(X_1,p_1; X_0,p_0) = {1\over\pi} \int d\xi_0 d\xi_1
e^{ -i (\xi_1 p_1 - \xi_0 p_0) } \nonumber\\
&& \times \int \delta X \delta \xi\
\exp {i\over\hbar} \biggl\{ S_{\rm sys}[X(t) + \xi(t)/2] \nonumber\\
&& - S_{\rm sys}[X(t) - \xi(t)/2] + W[X(t),\xi(t)] \biggr\},
\end{eqnarray}
\begin{eqnarray}
&& = {1\over\pi} \int d\xi_0 d\xi_1
e^{ -i (\xi_1 p_1 - \xi_0 p_0) } \nonumber\\
&& \times \int \delta X\
\exp \biggl\{ - {1\over{\hbar K}} \int_{t_i}^{t_{i+1}} e^2(t) dt
+ i ( \xi_1 M {\dot X}_1 - \xi_0 M {\dot X}_0 ) \biggr\}, \\
&& = 4 \pi \int \delta X\ \delta(p_0 - M{\dot X}_0) \delta(p_1 - M{\dot X}_1)
\exp \biggl\{ - {1\over{\hbar K}}
\int_0^{2\pi/\omega_0} e^2(t) dt \biggr\}.
\nonumber
\label{TransitionMatrix}
\end{eqnarray}
This evolution strongly resembles the classical evolution of probability
measures induced by the phase-space map, and in the classical limit we expect
$w(X,p)$ to evolve towards an invariant distribution $w_{\rm inv}(X,p)$ which
closely resembles the classical invariant measure $P_{\rm inv}(x,p)$.
Graham \cite{Graham} has demonstrated this sort of behavior in his work
on the quantum Lorenz model, which, though very different in approach
from this paper, may nevertheless be indicative;
and the author's own numerical
simulations \cite{Brun3} seem to bear this out
(though of course one would not
expect numerical simulations to exhibit unstable alternative solutions).
In the quantum case,
it is impossible for the noise to ever truly vanish. Even
at absolute zero, zero-point fluctuations remain that prevent $w(X,p)$ from
becoming a true fractal. Though the invariant distribution may strongly
resemble $P_{\rm inv}$ for a wide range of scales, there is always some
scale at which the quantum noise ``smears out'' $w_{\rm inv}(X,p)$.
While these maps on Wigner distributions make contact with the classical
theory, ideally we would like to find some quantum analog to
(\ref{poincare_map}), i.e., a description in terms of individual histories,
rather than probability distributions. To do this, let us consider
yet another coarse graining. Consider the decoherence functional
(\ref{DecoherenceFunctional0}), where we take the sequence of times
to be those corresponding to the surface of section $t_i = 2\pi i$,
and let the projection operators $P_{q,p}$ be onto localized cells
of phase space centered at $(q,p)$. While there are no true projections
onto cells of phase space, there are approximate projectors which
can be used to get approximate decoherence \cite{Halliwell,Brun2}.
For example, simple coherent state projections $|q,p\rangle\langle q,p|$
can be used.
A history can then be specified by a series of points $\{q_i,p_i\}$ at
the times $t_i$, and the decoherence functional calculated
\begin{equation}
D[\{q_i,p_i\},\{q_i',p_i'\}] =
{\rm Tr} \bigl\{ P_{q_n,p_n} {\bf T}( \cdots
{\bf T}(P_{q_1,p_1} \rho_0 P_{q_1',p_1'} \cdots ) P_{q_n',p_n'} \bigr\}.
\end{equation}
Here we have taken {\bf T} to be the transition matrix on density operators
rather than Wigner distributions; it is simple to go from one representation
to the other. This is a {\it quantum surface of section}. At each time
$t_i$ the system is localized in a cell in phase space centered on
$(q_i,p_i)$, and probabilities can be assigned to each possible next point
$(q_{i+1},p_{i+1})$. This differs, of course, from the classical case
where the evolution is deterministic; but from (\ref{Probability}) it
is clear that these histories will be peaked about the classical evolution
in the quasiclassical limit.
This is shown explicitly by the numerical example in the next section.
\section{Numerical simulation and quantum state diffusion}
While the decoherent histories formalism has great interpretational
power, it is not very convenient for numerical simulation. Enumerating
all the possible histories and calculating the elements of the decoherence
functional is a daunting task. What one would like is a method of
generating the histories with the correct probabilities without having
to solve the full master equation at every step.
Recently, it has been shown that the theory of quantum state diffusion
provides just such a technique. Quantum state diffusion is one of
the so-called {\it quantum trajectory} methods, in which the master
equation evolution is unravelled into quantum trajectories
of individual states. These states obey a nonlinear stochastic differential
equation which in the mean reproduces the master equation. Because
one need deal only with a single quantum state at a time, this is
very suitable for numerical calculation \cite{QSD}.
Diosi, Gisin, Halliwell and Percival \cite{DGHP}
have shown that these individual
quantum trajectories correspond to a set of approximately decoherent
histories. In the case of a dissipative interaction, these correspond
to histories of systems localized into small cells in phase space,
and their probabilities match those given by decoherent histories.
Thus it is ideal for the sort of problem we are interested in.
For further details see the references. More work on the connections
between decoherent histories and quantum state diffusion, and their
application to dissipative chaos, is currently underway \cite{Brun3}.
In figure 2 we see one such trajectory, generated in the quasiclassical
limit (where $\hbar = 10^{-4}$). One can see that this is very
close to the classical limit, but with additional ``smearing'' due to
the presence of noise. This smearing sets a lower cutoff scale to the
substructure of the strange attractor. As one continues to go to the
classical limit, more and more substructure appears, and the noise
becomes less and less important.
Note that in this chaotic system we expect the trajectory to evenly
sample the ``invariant'' Wigner distribution $w_{\rm inv}$ over time.
{}From the distribution of points we see that this is indeed very close
to the structure of the classical strange attractor in figure 1.
\section{Conclusions}
As we have seen, it is possible to use the decoherence formalism to study
at least some systems exhibiting classically chaotic behavior, and to
do so in a way which includes dissipation in a simple and natural fashion.
Though the model treated here is very much a special case, intended only
to illustrate the basic ideas of the theory, it is remarkable how many
details can be brought out and studied by its means.
Certainly these techniques should work for any kind of nonlinear oscillator,
or for multivariable extensions of them. It might well be possible to
treat systems of experimental interest, arising in fields such as quantum
optics. Some work on such systems has already been done by other workers
\cite{Savage}.
Using the usual master equation formalism it is possible to draw a
close connection between the classical theory of probability measures
and the quantum Wigner distribution. But with decoherent histories,
one can also find a quantum analog to individual chaotic orbits,
such as the {\it quantum surface of section} defined in section 4.
One can then argue analytically that these quantum histories become
more and more closely peaked about the classical equations of motion
as one goes to the classical limit; and this correspondence can also
be demonstrated numerically.
Further analytical
study may yield better results for the probabilities and decoherence of
phase space histories. And it may be fruitful to explore what equivalents
there are in the quantum case to classical quantities such as Lyapunov
exponents, fractal dimension, and Kolmogorov entropy.
This theory should amply reward further study, both analytical
and numerical.
\section*{Acknowledgements}
First and foremost I would like to acknowledge the guidance of my
advisor, Murray Gell-Mann, and many stimulating discussions with
my Seth Lloyd, Juan Pablo Paz, Wojciech Zurek, and Jim Hartle. Considerable
help on phase space histories was provided by Jonathan Halliwell, and
I am indebted to Ian Percival and Nicolas Gisin for my knowledge of
quantum state diffusion theory. This work was supported in part by
DOE Grant No. DE-FG03-92-ER40701.
|
\subsection{ Summary }
There are several points to recall before discussing the results found
in Sections 5 and 6.
All the numerical codes we have designed for the present work
have been duplicated, independently by two of us, in order to avoid
computational errors.
None of the steps of the entire test procedure
depends on the Hubble constant $H_0$.
The test is completely independent of any absolute magnitude or
distance scaling,
since it only relies on volume ratios and on flux ratios.
So, the only parameters which do depend on the choice of $H_0$ are
the characteristic evolution time and the quasar absolute
luminosities such as $\Phi_{*}$, the normalization of the
luminosity function given in Paper II.
Schade and Hartwick 1994, on the contrary, need an explicit
value of $\Phi_{*}$, thus a choice of $H_0$. They also make use
of the same functional form and
parameters for the luminosity function in all cosmological models. Even
if the values used are within error bars for all models, they
are better suited to one or more peculiar model, which is
likely to bias this test.
Our results on the cosmological parameters only apply {\it under
specific hypotheses concerning quasar evolution}.
Such hypotheses are difficult to avoid and the hypotheses made
here are quite widely encountered. The present paper is
entirely devoted to the case for Luminosity Evolution, which has been
favoured by recent studies (Boyle et al. 1988, 1990).
There is a way to test the adopted evolution laws: applying the
$<V/V_{max}>$ test in various redshift bins, and comparing the values
found for $k_L$. Such a procedure however requires quite a large
sample. Applying the test in the frame of Density Evolution
will be the subject of a forthcoming paper.
Quite interesting are the results of cross-tests of the two
standard functional forms: samples issued from a simulation
with power law (PWLE) and tested with exponential (LEXP)
lead to similar probability maps in the
$(\Omega _{mat},\Lambda )$ plane, which makes
us rather confident on their generality. Hopefully, such similarities
may be expected from cross-tests between Density and Luminosity
Evolutions, which will follow in a forthcoming paper.
The test seems to be insensitive to the cosmological constant. In fact
we showed it is sensitive to a certain function $f(\Omega_{mat},
\Lambda)$ which depends on the redshift depth of the catalogue.
$f(\Omega_{mat},\Lambda)=\Omega_{mat}$ for a catalogue with a low
redshift limitation $(z_2 \simeq 2.2)$. $f(\Omega_{mat},\Lambda)$
tends to the curvature $\Omega_{mat}+\Lambda$ for a deep catalogue.
By this way, with at least two different catalogues, it is possible to
determine $\Lambda$. The recent LBQS catalogue
of Hewett {\it et al} (1995), with more than 1000 QSOs,
may be useful for this purpose.
\subsection{ Limits on $\Omega _{mat}$, and the status of $\Lambda$}
As already underlined, the constraints brought in by the cosmological
test described in this paper hold mainly on $\Omega _{mat}$, and
to a lesser extent on the curvature term $\Omega _k$; see Table 2.
A low $\Omega _{mat}$ is rejected, and we exclude a flat Universe
without a cosmological constant. This implies introducing a non zero
$\Lambda$ in the standard inflation models ($\Omega_{tot}=1$).
This is in the sense of the recent determinations of the Hubble
parameter which give too low an age for the Universe if $\Lambda=0$,
compared to the age of the oldest stars only (Jaffe, 1995).
Let us compare our limits on $\Omega _{mat}$ to what is found
in the current litterature, arising from tests of different nature
and/or on different scales:
The limits coming from velocity fields in voids are
obtained without assumptions regarding either the galaxy biasing, $\Lambda $,
or the fluctuation statistics (Dekel and Rees 1994).
White {\it et al.} (1993) express $\Omega _{mat}$ in terms of
$f_b$, the baryon fraction in clusters of galaxies:
$\Omega _{mat}< 0.3 \left({\Omega _b \over 0.06} \right) \left({0.2 \over
f_b } \right) $, {\it if} the ratio of baryonic to total matter in
clusters is representative of the universe.
Accounting for the limits on
the baryon fraction deduced from
X-Ray observations of the Coma cluster,
$ f_b = 0.009 + 0.05 h^{-3/2} $, and on the limits on
$\Omega _b$ from standard Big-Bang Nucleosynthesis:
$0.009 h^{-2}< \Omega _{ baryon}< 0.02 h^{-2}$,
we finally get:
$\Omega _{mat}< 0.4 {h^{-1/2} \over 1+ 0.18 h^{3/2}}$.
Comparing our results with two recent papers is of particular interest:
Bernardeau {\it et al.} (1995) fix $\Omega >0.3$
at the 2 $\sigma $ level and favour $\Omega =1$ from the skewness of the
cosmic velocity field divergence. Meanwhile
Hudson {\it et al.} (1995), by comparing the distribution of optical
galaxies with the POTENT mass distribution, conclude that
$\beta =\Omega ^{0.6}/b=0.74 \pm 0.13 \; (1 \sigma ) $,
favouring either
$\Omega =1, \; b=1.35 \pm 0.23 $ or $b=1, \; \Omega =0.6 \pm 0.18$.
Cole {\it et al.} (1995) describe similar work from IRAS
redshift surveys and find that $\beta =0.5 \pm 0.2 $,
implying
$\Omega =1, \; b\simeq 2 $ or $b=1, \; \Omega =0.35 $.
This last two results are
surprinsingly close to ours: $ \Omega =0.5 \pm 0.2 \; (2 \, \sigma)$
(PWLE) or $ \Omega =0.55 \pm 0.3 \; (2 \, \sigma)$ (LEXP).
Perhaps we can conclude that the bias parameter $b$ is of order unity,
but this still appears to be premature!
All these results are summarized in Table 4.
\begin{table*}
\caption{ Limits on $\Omega _{mat}$ obtained in various recent
tests }
\halign{ \hfil # \hfil & \hfil #\hfil &\hfil
#\hfil &\hfil #\hfil & \quad \hfil # \hfill\quad \cr
\noalign{\hrule\medskip}
\multispan5 {\it (Sp. H. = spiral haloes, CoG=Clusters
of Galaxies, $f_b$=baryon fraction, VT=Virial
Theorem, CVT=Cosmic VT)} \cr
\noalign{\medskip}
\noalign{\hrule\medskip}
&&&&\cr
Test & $\Omega $& confidence level & $\Lambda $& reference \cr
&&&&\cr
Galaxies:&&&&\cr
&&&&\cr
Sp - H. & $0.1 < \Omega < 1$&$1.5 \sigma $ & $\Lambda =$ 0 &
Zaritsky and White (1993)\cr
&&&&\cr
\noalign{\smallskip\hrule\smallskip}
Clusters:&&&&\cr
&&&&\cr
CoG: VT &$ \Omega h_{75}^{-1} = 0.2 \pm 0.1 $ && $\forall \Lambda $&\cr
CoG: Giant Arcs & $ \Omega h_{75}^{-1}= 0.2 \pm 0.1 $ &&$\Lambda $=0& Fort and Mellier (1994) \cr
CoG: Weak Shear &$ \Omega h_{75}^{-1}= 1 \pm 0.5 $ &&$\Lambda $=0& Bonnet {\it et al. } (1994) \cr
&&&& Fahlman {\it et al. } (1994) \cr
CoG: $f_b$ &$\Omega h_{75}^{1/2} <0.5 $ &&$\Lambda $=0& White {\it et al. } (1993) \cr
\noalign{\smallskip\hrule\smallskip}
&&&&\cr
Velocity Fields:&&&&\cr
&&&&\cr
CVT-CfA1 & $ \Omega = 0.5 \pm 0.1 $ &&$\forall\Lambda $&\cr
POTENT--IPDF & $0.3 < \Omega $&$5 \sigma $ &$\forall\Lambda $& Dekel (1994)\cr
POTENT--VOIDS &0.3 $< \Omega $ & $ 2.4 \sigma $ &$\forall\Lambda $&
Dekel and Rees (1993) \cr
POTENT--SKEW & 0.3 $< \Omega $ & $ 2\sigma $ & $ \forall\Lambda $ & Bernardeau {\it
et al. } (1995) \cr
\noalign{\smallskip\hrule\smallskip}
&&&&\cr
QSOs:&&&&\cr
&&&&\cr
spatial distribution& $ \Omega = 0.4 \pm $0.2 &1 $ \sigma $ &
$ \Lambda =$ 0 & Deng et al. (1994)\cr
Loh--Spillar & $\Omega = 0.2\pm $ 0.1 & 1 $\sigma $ & $ \Lambda $ =0&\cr
& $ \Omega = 0.4\pm 0.1 $ &1 $ \sigma $ & $ \Lambda =1-\Omega $ & Schade
et al. (1994)\cr
This Work: & $ \Omega = 0.5 \pm 0.2 $ & $ 2\sigma $ &
$ \Lambda<$ 1.4 & PWLE\cr
& $ \Omega = 0.5 \pm 0.3 $ & $2\sigma $ & $ \Lambda<1.4 $ & LEXP \cr
&&&&\cr
\noalign{\medskip\hrule}}
\end{table*}
The test described in the present paper, performed with QSOs,
measures the cosmological
parameters on a scale of several $Gpc$. It agrees well
with most of the tests given in Table 4. The only exception
concerns the results on the centers of clusters of galaxies (both
the Virial Theorem and the Giant Arcs) which
have been operated on smaller scales, typical of structures which have
detached from the general expansion, then collapsed and dissipated.
Furthermore, these
tests rely on the assumption of a universal Mass to Light
ratio, and it is not surprising
to find smaller $\Omega _{mat}$ on smaller scales.
\bigskip
\subsection{ Optimal strategy for quasar detection aimed at cosmology }
We combined the results of the last two series of simulations to look
for the most efficient use of telescope time aimed at cosmological
tests based on quasars.
We saw that for the same precision on the cosmological parameters, say
$\Omega _{tot}$, about 10 times less QSOs are needed if their limiting
redshift is 3.5 instead of 2.2.
Let us compare two possible observational strategies for quasar
detection, for example with
the FUEGOS multifiber spectrograph to be build for the VLT at ESO:
first, programme 1, a classical preselection of UVX candidates brighter
than $B=24$ with some limit in $U-B$, and second, programme 2,
consisting off the
spectroscopy of all stellar or compact objects brighter than $R=23$ (
Mathez {\it et al}, 1995a).
A selection of $I<22.5$ could be
preferred.
A preselection of UVX candidates (or simply through the blue
filter, see Glazebrook {\it et al.} 1995) will be efficient
with a QSO/candidates ratio around 1/3. The drawback, however, is
the redshift limitation inherent in the blue selection.
In programme 2, the drawback is that the $QSO/candidate \, \,
ratio$ is far (to what extent ?) lower.
The question is wether this drawback may
be compensated by the higher
limiting redshift, which would be higher than 2.2.
The last source of error is that we used a simulated catalogue with an
evolution extrapolated from low redshifts to $z=3.5$.
Although the evolution law at these
high redshifts is unknown, it is probably different from the evolution at low
redshift.
The two strategies are compared in Table 4, note however
that many input quantities in Table 4 are only first guesses.
The surface density $\Sigma =300 \, deg.^{-2}$ of QSOs
brighter than $B=24$ is one of the most
secure inputs (to within a factor of 2) since convergent
results have been found by
Glazebrook {\it et al.} (1995) and Koo, Kron and Cudworth (1986).
The QSO/candidate ratio expected in programme 2 was taken as 1/30
to compare the two programmes. We think that it is rather conservative
for a selection through the red filter and is likely to allow a higher
limiting redshift. The quasar evolution at redshifts above 2.5 is
also questionable. It seems to reverse (Hook {\it et al.} 1995),
but it is not easy to ascertain because of the different observational
approaches below and above this redshift.
The surface density of stars to 23rd magnitude has
been estimated from the model of Robin (1989). As for the surface density of
quasars to $R=23$, Table 4 gives a conservative estimate since no
quasar selection has ever been made in $R$. Recent results by Webster
(1994) seem to indicate that red quasars are not rare, and that
the surface density of red or infrared QSOS could be far higher.
With the 80 fibres of FUEGOS, both programmes may be expected to have
similar
global efficiencies for the determination of $\Omega _{mat}$.
However they would lead to different scientific outputs: a
complete sample of 10,000 QSOs in the first case, allowing excellent
statistics for quasar studies, and a better
idea of the QSO evolution at high redshift, i.e. presumably constraints
on galaxy formation, in the second one.
And a good model of high--z evolution is what we need
to determine $\Lambda$ from this test.
\begin{table}
\caption{ Optimal strategy for detection of quasars with FUEGOS with
80 fibres (FF= FUEGOS FIELD = 1/7 sq. deg.)}
\halign{ \hfil # \hfill & \quad #\hfill & \quad #\hfill & \quad #
\hfill \cr
\noalign{\hrule\medskip}
& $U-B<0.2$ & No Preselection \cr
$m_{lim}$ & B=24 & R=23 \cr
$\Sigma _{cand}\,(deg^{-2}$)& 1050 & 7500 \cr
$\Sigma _{QSO} \,(deg^{-2}$)& 300 & 250 \cr
QSO/candidate & 1/3.5 & 1/30 \cr
candidate/FF & 150 & 1100 \cr
QSO/FF & 50 & 35 \cr
exposures/FF \, (80 \, fb) & 2 & 14 \cr
$N_{QSO} \, (\sigma _{\Omega _k}=0.5)$ & 5000& 400\cr
survey area $(deg^2)$ & 17& 1.6 \cr
FF & 100 & 12 \cr
total exposures & 200 & 170 \cr
\noalign{\medskip\hrule}}
\end{table}
\subsection{The spirit of the test}
Our working frame is the Friedmann-Lema\^ \i tre cosmologies.
The QSO Luminosity Evolution is described, as usual,
by a given functional form depending on a single evolution parameter.
Our null hypothesis is that the QSOs are uniformly distributed in
volume, and that we deal with a sample complete in magnitude.
To properly account for the well known observational biases present
in complete QSO samples at low and high redshifts,
one has to work in a restricted redshift range.
The variable which is uniformly distributed is not $V/V_{max}$, but
a ratio $x$ defined in Section 4.
The sample is binned according to $M_0$, the absolute magnitude
defined at the epoch $z=0$ and at the
frequency where observations are done.
Since in the frame of the correct model (evolution and cosmology) the
$x$ distribution does not depend on the absolute magnitude,
each bin must have a uniform $x$ distribution. A correlation
between $V/V_{max}$ and absolute magnitude is one of the most sensitive
pieces of evidence for a wrong choice of cosmology and/or
evolution (see Fig. 1 and Bigot and Triay 1991).
Fig. 1 shows the positions of QSOs in the $(M_0,V/V_{max})$
plane for three values of the evolution parameter.
A study of the $V/V_{max}$ distribution
is clearly less sensitive in the whole sample than in luminosity bins.
\begin{figure}
\picplace{19cm}
\caption{The basic idea of the test: Boyle's et al. (1990)
QSOs are represented in the (absolute magnitude-$V/V_{max}$) plane.
The model is power law luminosity evolution
(PWLE), $\Omega=1, \Lambda=0$. The values of the evolution parameter are
$k_L=0., 3., 6.$, respectively. The upper and lower plots show a strong
correlation, corresponding to a bad combination of evolution and
cosmological model.
In the middle, there is no correlation whatever the magnitude.
It can be seen that the {\it global} $<V/V_{max}>$ may be 1/2, while
the distribution is far from uniform in the high-- and low--luminosity
bins.}
\end{figure}
Another interesting piece of information is the evolution of individual $x$
versus the cosmology and the evolution characteristic parameters (see Fig. 2).
We expect that a cosmological test using the $x$ values might be mostly
sensitive in the regions of parameter space where $x$ vary strongly. As we see
in Fig. 2, two asymptotic regions, toward $\Lambda <0$ and $\Omega_{mat}>1$, are
observed, which show that the test will be insensitive to the cosmology in
these regions.
\begin{figure}
\picplace{8cm}
\caption{$V/V_{max}$ ratio versus cosmological parameters for
15 QSOs randomly chosen in the Boyle's {\it et al.} (1990) sample.
Top left, $\Lambda=0$, top right, $\Lambda=1$, bottom left,
$\Omega_{mat}=1$, bottom right, $\Omega_{mat}=2$.
The evolution parameter always corresponds to the maximum likelihood.
Discontinuities in the derivative correspond to change of status from
magnitude-- to redshift--limited or conversely. Noise in the bottom
curves is due to computational difficulties. In the two asymptotic regimes
discussed in text (towards large $\Omega $ and small $\Lambda $),
the test has little or no sensitivity, while the test is quite efficient at
low $\Omega $.}
\end{figure}
\subsection{ Luminosity Evolution}
Pure Luminosity Evolution has been introduced in a phenomenological way,
without physical support. It corresponds to the case where the fraction
of active galaxies is constant versus redshift, but not their luminosity.
Although it is a statistical evolution of the whole population,
the usual simplifying hypothesis on luminosity evolution is that {\it all}
quasar absolute luminosities $L$ do evolve in parallel according to the law
(Mathez 1976, Marshall et al. 1984, Boyle et al. 1987):
\begin{equation}
L\left( t(z) \right) = L( t_0) \times e\left(t(z)\right),
\end{equation}
where $e(z)$ has either the 'power law' form in the PWLE case:
\begin{equation}
e(z) = (1+z)^{k_L},
\end{equation}
or the 'exponential' form in the LEXP case:
\begin{eqnarray}
e(z)&=&exp \left[\left(t_0-t(z)\right)/\tau \right]\nonumber\cr
&=&exp \left[k_L H_0\left(t_0-t(z)\right)\right].
\end{eqnarray}
$t(z)$ is the cosmological epoch corresponding to redshift $z, t_0 =
t(z=0)$, $\tau =1/(k_L H_0) $ is the characteristic evolution time,
and $k_L$, the 'evolution strength' in the standard notation.
\subsection{Evolutionary tracks in the Hubble plane}
Several pairs of independent variables may be used to
study the quasar distribution: redshift, volume, and comoving volume on
one hand and apparent magnitude, flux density, intrinsic luminosity
on the other hand. Here we prefer to work in the Hubble plane,
apparent magnitude--redshift ($m,z$), because it corresponds to observables. No
hypothese either on cosmology or on evolution are needed to represent
quasars in this plane.
Assume that we have a quasar sample complete
in the magnitude range $[m_1,m_2]$, in a given sky area.
The fainter limit $m_2$ is obvious. The bright limit $m_1$
may arise from photometric saturation (either in the photometric or in
the spectroscopic data).
Quasars of the sample thus lie inside what we call 'box $\cal B$'
in the following, i.e. the domain
$[m_1,m_2]\otimes [z_1,z_2]$ of the Hubble plane:
we further restrict ourselves to some range
of absolute magnitude $[M_0^1,M_0^2]$, arbitrary provided that
$[M_0^1,M_0^2]\otimes[z_1,z_2]$ is included in box $\cal B$.
According to Eq. 2, all quasars follow parallel
evolutionary tracks in the Hubble plane.
Now consider a given quasar in box $\cal B$.
Its evolutionary track will go along the
$m(z)$ curve in the Hubble plane, described by the Mattig
relation:
\begin{equation}
m(z)=M_0-5+2.5log_{10}\left( d_L(z)^2 \times K(z) /e(z) \right)
\end{equation}
where $M_0$ is the absolute magnitude at epoch $z=0$,
$d_L(z)$ is the luminosity distance and $K(z)$ is the
K--correction: $K(z)=(1+z)^{\alpha -1}$ with $\alpha =0.5$ as spectral
index. Note that in the case of a strong luminosity evolution, the $m(z)$
relation is no longer monotonic, and can even exhibit several
local maxima in some extreme cases.
It is necessary to exclude for any volume computation, all
the redshift ranges where $m>m_2$ or $m<m_1$ (see the details in Paper II).
The stronger the evolution, the smaller will
be the slope of the evolutionary track. It can become negative if evolution
is such that apparent magnitude passes through a maximum.
Fig. 3 shows one illustration of the Mattig relation when the evolution
is stronger than the cosmological extinction.
\begin{figure}
\picplace{8cm}
\caption{Various Mattig functions with different absolute magnitudes,
show parallel evolutionary tracks in the Hubble plane.
Between the separators $\cal S_1$ and $\cal S_2$, all QSOs are
redshift--limited.
The sign $+$ on the bottom line denotes a QSO whose
redshift $z$ is limited in the range $[z_{min},z_{max}]$ because of
the bright magnitude limitation $m<m_1$.}
\end{figure}
\subsection{Observational cutoffs at low and high redshift}
As has been known for a long time, the redshift
distribution of most large complete QSO samples is biased at low and high
redshift. These cutoffs are
important to account for, since they govern the locus of both
extremities of evolutionary tracks in the Hubble plane.
At low redshift, it is easier to distinguish the
host galaxy of QSOs, so many of them {\it do not} have a stellar
appearance and are missed in quasar samples.
We will restrict our analysis to redshifts z$>$0.3.
Searching for ultraviolet excess (UVX) stellar objects
is the most efficient way to detect
quasar candidates, so the largest complete
samples consist of UVX quasars.
At high redshift, quasars no longer have any
ultraviolet excess, since
$Ly_{\alpha }$, the dominant emission line, passes to the blue
filter at $z\simeq 2.3$.
So most of the large quasar samples are in fact limited to $0.3<z<2.2$.
To our knowledge, Kassiola and
Mathez (1991) and Crawford (1995) are the first authors to explicitly
mention that using maximum or minimum redshifts outside the range $[0.3,2.2]$
for UVX QSOs is erroneous.
In Paper II, the joint distribution of redshift and magnitude
is examined in the frame of PLE, and it is shown how to compute the
individual ratios $x$ which are uniformly distributed over $[0,1]$ for
quasars in any arbitrary bin of absolute magnitude.
In the next Section, we will take advantage of these properties to derive a
cosmological test.
\subsection{ Choosing a grid of cosmological models}
Most of the previous cosmological tests have been
performed under the assumption $\Lambda =0$, with either a low
density parameter $\Omega _{mat}$ or a flat universe $\Omega _{mat}=1$
($\Omega _{mat}$ is the {\it total} density
parameter, including baryonic and non-baryonic matters).
Some tests however adopt a non zero cosmological constant
but impose a null curvature by the condition $\Omega _{mat}+\Lambda =1$.
Present ideas on the density parameter are synthetized in Table 4.
To explore a large range of cosmological models, we choose the range
$[0,3]$ for $\Omega_{mat}$, and $[-1,2]$ for $\Lambda $, varying these
parameters independently, by steps of 0.2. However we exclude from the
test all cosmologies without an initial singularity (see Carroll {\it et
al.} (1992) for details).
It results in 16$\times $16-17=239 cosmological models.
Absolute magnitudes are irrelevant in the test, only the
rank of these magnitudes is important, which is
independent of $H_0$. Our test, based on the statistics of individual
$V/V_{max}$ ratios, is {\it totally independent} of the Hubble
parameter.
\subsection{Associating a confidence level to each cosmological model }
We work in the $(absolute \; magnitude - V/V_{max})$
plane. We choose a functional form of the evolution before performing
the test.
Our test consists in ensuring a maximum likelihood to the null
hypothesis of {\it uniform distributions} of the $x$ ratio
defined in Section 4 {\it jointly in all bins of
absolute magnitude}. We have to ensure this {\it without any hypothesis on the
functional form of the luminosity function}.
Let ${\cal M}$ be the current cosmological model $(\Omega_{mat}, \Lambda)$.
To assign a probability to ${\cal M}$, we proceed as follows:
once a value of the evolution parameter $k_L$ is chosen, each ratio $x$
is computed.
The sample is ordered by increasing absolute
luminosity and divided into $n$ luminosity
bins containing equal numbers of quasars. This is to minimize
the noise that would be caused by very different numbers
of quasars in each bin.
In each luminosity bin we calculate the Kolmogorov-Smirnov probability
that the corresponding $x$ ratios come from a uniform
distribution (see Press et al. 1992). Let
$P_{KS}^i({\cal M},k_L)$ be this probability for the luminosity bin
$i$. These bins are statistically independent, so are the
probabilities, thus we can compute the likelihood function of
the total sample of $x$:
\begin{equation}
{\cal L} ({\cal M},k_L)=\prod_{i=1}^{N} P_{KS}^i({\cal M},k_L)
\end{equation}
This likelihood is used to calculate an entropy:
\begin{equation}
{\cal E}=-2log({\cal L})
\end{equation}
For each model ${\cal M}$, we determine the parameter
$k_L$ giving the lowest value of the entropy ${\cal E}_{low} (\cal M)$.
${\cal E}_{low} (\cal M)$ is defined as the entropy of the cosmological
model $\cal M$. This operation is repeated for the 239
cosmological models to get an entropy map. At the end, the entropy is
${\cal E}_{low}({\cal M})={\cal E}_{min}+\Delta {\cal E}({\cal M})$,
where ${\cal E}_{min}$ is the lowest entropy found in the
cosmological grid. $\Delta {\cal E}$ follows a $\chi^2$
statistic with three degrees of freedom (Lampton {\it et al.} 1976),
since three parameters are determined. For three parameters, the confidence
intervals around the
best model $\cal M$ are given by $\Delta {\cal E}=6.25, 7.80, 11.30,
16.20$ for the $90\%, 95\%, 99\%, 99.9\%$ confidence level
respectively.
\subsection{Choice of binning }
The absence of {\it an analytical form
of the luminosity function} has the drawback of preventing the use of a 2D
Kolmogorov--Smirnov test in the $(M_0,V/V_{max})$ plane.
So we are obliged to work in $n$ luminosity bins and an
arbitrary binning is substituted for an arbitrary analytic
luminosity function. It is easier however to explore all the
possible binnings than to explore all the possible analytic expressions!
Moreover, the binning procedure is well suited to the method,
which applies whatever the choice of the bin $[M_0^1,M_0^2]$.
In principle, the choice of the binning is not constrained. However,
two practical limitations arise:
$\bullet $ As noted by Lampton {\it et al} (1976),
we should not proceed to
reject some values of the cosmological parameters from
confidence levels without
a preliminary good fit cosmological model.
This is equivalent to requiring that the
lowest entropy ${\cal E}_{min}$ of the most probable cosmological model
not exceed $n-3$ by much too much. This
problem arises with too large $n$.
$\bullet $ The $x$ ratio is calculated from the fictitious
redshift displacement
of a QSO according to the evolution law.
However it must not be forgotten that the evolution law has only a
statistical sense, and that it applies only to a QSO {\it population}.
If the number of QSOs per bin is too small (i.e. if $n$ is too large),
the statistical sense of the evolution law is lost and the
probability map becomes noisy. We choose a bin
number which keeps the smooth appearance of the probability map.
The method is validated by the analysis of synthetic
quasar catalogues which are described in Paper II.
By trial and error based on
varying the size of the simulated QSO sample (see Section 5, Fig. 7), we
found that the best choice from the point of view of the minimum entropy
is to take approximately between 10 and 20 bins.
Since a large part of the work is made with these catalogues,
we discuss them in Section 5.
\subsection{Behaviour of $x$ versus the parameters}
In Paper II are introduced the various quantities necessary for
computing the individual $x$ ratios in a complete QSO sample. Two
curves in the Hubble
plane, the bright and the faint separators ${\cal S}_1$ and ${\cal
S}_2$, are of particular importance. These curves, shown in
Fig. 3, delimit the domain ${\cal D}$, lying between them, and the
domain ${\cal D}^*$ which is its complementary with respect to box
$\cal B$, or
$[m_1,m_2]\otimes [z_1,z_2]$. All quasars in domain ${\cal D}$
(here after $\cal D$ quasars) are redshift--limited, while all $\cal
D^*$ quasars are magnitude--limited.
According to Paper II, the $x$ ratio is uniformly distributed
in the range $[0,1]$
for quasars in any absolute magnitude range $[M_0^1,M_0^2]$.
The $x$ ratio takes quite a simple form for $\cal D$ quasars:
\begin{equation}
x \;=\; {V-V_1 \over V_2 - V_1 }
\end{equation}
where the volumes $V_i=V(z_i)\;(i=1,2)$, the comoving volumes out to
redshift $z_i$, are determined only from the cosmological
model. For $\cal D^*$ quasars, on the contrary, there are
many different cases according to the respective values
of $z_1$ and $z_2$ and of the redshifts where the
magnitude equals one of the limiting magnitudes. An
example is given in
Fig. 3. Consider the QSO denoted by $+$. If
we exclude from the volume
computation the two redshift
ranges where $m(z)<m_1$, then the $x$ ratio for this QSO is:
\begin{equation}
x \;=\; {V-V_{min} \over V_{max}-V_{min} },
\end{equation}
It should be noticed that $k_L$ does
not enter Eq. 6. $k_L$ concerns $\cal D$ quasars only inasmuch
it governs the repartition of quasars
with respect to the separators.
On the contrary, $V_{min}$ and/or
$V_{max}$ which enter the equivalent of Eq. 6 for $\cal D^*$
quasars, do depend {\it explicitely} on $k_L$.
As evidenced in Fig. 2, the run of individual $x$ ratios
versus cosmological parameters varies substantially from quasar to
quasar, although $k_L$ is fitted to the maximum likelihood in each
cosmological model of Fig. 2. Large variations and even inversion of slope
correspond to the transition from one class to the other
(magnitude-- or redshift--limited). As $\Omega _{mat}$
increases at fixed $\Lambda $, and as $\Lambda $ decreases at
fixed $\Omega _{mat}$, the curves tend towards horizontal
asymptotes so it becomes impossible to test between various
low $\Lambda $ or high $\Omega _{mat}$ universes.
However we will introduce in the following an alternative method based on
the synthetic catalogues analysis to avoid this problem.
\subsection{ Towards efficient cosmological tests }
Our method is far more demanding than a {\it
global application} of the $<V/V_{max}>$ test.
We have several distinct sub-samples (the $N$ absolute magnitude
bins), which are independent in
the sense that they do {\it not} contain the same objects and their
$x$ ratios do not behave similarly, neither with respect to evolution
strength, nor with respect to cosmological parameters.
The $x$-uniformity puts in $N$ constraints instead of a single one:
in some cosmological models, whatever the value of $k_L$,
all the distributions of $x$ ratios have too low a probability to be uniform
{\it simultaneously} for all the absolute magnitude bins.
This is exactly what is needed
to reject some cosmological models as incompatible with a given
hypothesis on quasar evolution.
One shortcoming of this test is that the evolution hypothesis itself
may be wrong, although it is quite widely used. However the values of
individual $x$ ratios for $\cal D$ quasars do not depend on the form and
the value of the luminosity evolution and only depend on the cosmology.
{\it If} there the proportion of $\cal D$ quasars is sufficient in current
samples then we will find some common results, even when varying
the various evolution hypotheses. We will see in Section 6 that this is
roughly the case.
There are various ways to perform the test:
the redshift range $[z_1,z_2]$ may be divided into bins.
There is presumably an optimal way to operate this redshift binning:
since most of the effects of the cosmological constant
are noticeable at high
redshift, say $z>1$, the essential contribution of the range $[0.3,1]$
is likely to increase the noise. On the other hand the effects of
varying $\Omega _{mat}$ are already noticeable at lower redshift
(Carroll {\it et al}, 1992), so that
one can expect to get constraints on $\Omega _{mat}$ from the low redshift
bin and constraints on $\Lambda $ from the high redshift bin.
There is probably an optimal compromise for the cutoff
to be found somewhere around $z=1$. In Section 6.1 we use this idea to
apply the test to low redshift ($0.3<z<1.5$) and high redshift ($1.5<z<2.2$)
samples of the Boyle QSO sample.
\subsection{Construction of Monte Carlo catalogues}
Before applying the test on real data, we checked its efficiency on
simulated samples. This
is the only way to vary the cosmological parameters, or to search for the
effect of the sample size, or of various biases in the redshift
distribution.
We generated Monte-Carlo samples of 400 objects for
each of the cosmological models of the grid
defined above, and we analysed the simulated samples exactly as we did
with Boyle's one. To simulate the catalogues, the procedures are the
following:
$\bullet $ choose a model (cosmology + evolution functional form)
$\bullet $ determine the function $\Phi (M_0)$
from the Boyle {\it et al.} global luminosity function.
$\bullet $ given a limiting magnitude and an absolute luminosity,
compute the available volume $V_a(M_0)$
$\bullet $ sort in luminosity according to
the observed probability distribution function
of luminosities in a sample, that is the
product $\Phi (M_0) \times V_a(M_0)$
$\bullet $ given $V_a(M_0)$, sort in redshift ensuring homogeneous spatial
distribution.
The last two steps are repeated once per object in the sample.
All the details are described in Paper II.
\subsection{ Application of the test to simulated catalogues}
In this Section we give the results of the test on various simulated
catalogues by varying different parameters and by introducing some bias to
understand how the test works.
We choose to represent the contours at $90, 95, 99, 99.9 \%$ confidence
levels which are the minimum confidence levels to reject some models.
PWLE refers to power law luminosity evolution, and LEXP to exponential
luminosity evolution. The limiting magnitude of the simulated catalogues
is always $m_{lim}=21$. No bright limit is set, since it complicates
the computations and does not change the test.
\subsection{ Comparing parent cosmologies }
Figs. 4 give the probability maps obtained in the
$(\Omega _{mat}, \Lambda )$ plane for 5 different parent cosmological
models. All maps correspond to samples of 400 objects with the
same Monte--Carlo seed. All catalogues have been generated under
the PWLE assumption, nevertheless varying the functional form of the
evolution does not modify significantly the form of the contour maps.
A large part of the parameter space is rejected with the models at low
$\Omega_{mat}$. This is expected from the Fig.2 because the
stronger variations of the $x$ ratios are precisely for low $\Omega_{mat}$
cosmologies. Our test will be able to discriminate between low and
high $\Omega_{mat}$ values. A general characteristic of the contour maps
is the tendency to be drawn toward negative values of $\Lambda$. This too is
expected from Fig.2 where we note the asymptotic regime of the $x$
ratios toward these values. Our test is not able to discriminate between
different values of $\Lambda$ with only 400 QSOs.
\subsection{ Binning}
The effect of binning is shown in Figs. 5.1 to 5.6.
First, with a single bin, the test reduces to the global $<V/V_{max}>$,
and no model at all may be rejected. Only the models with no Big--Bang
(up--left of the map) are rejected because they are not computed!
As expected, increasing the binning decreases the allowed range in
parameter space, until the
map becomes noisy and our method to find the
minimum entropy no longer applies.
\subsection{ The seed }
In order not to superimpose the effects of varying the seed and
cosmological effects, all the previous figures correspond to
Monte--Carlo catalogues drawn with the same initial seed.
To illustrate the effects of the statistical fluctuations, we show in
Figs. 6 the analyses of ten catalogues of the same
size and parent cosmology, which correspond to ten different
random seeds. The result is that varying the seed changes the form of
the probability map, but does not reject the parent cosmological model.
\subsection{Sample size and limiting redshift}
Of course the larger the sample, the smaller is the probability map at
given confidence level and at fixed higher limiting redshift $z_2$.
This effect is illustrated in Figs. 7.1 to 7.4.
Despite the large number of QSOs in the
sample, the test is not able to determine $\Lambda$. The fact that the
test is not very sensitive to $\Lambda$ comes from the value of
$z_2=2.2$ which is not deep enough.
Conversely, the higher $z_2$, the more efficient is the test versus
cosmological parameters. The effect of increasing $z_2$ at fixed $n$
is shown on Figs. 8.1 to 8.3.
In fact, these figures show that a high--z limited
catalogue is very sensitive to
the curvature: on Fig.8.3 the contour probability
follows very well the iso--curvature line of the model,
with only 400 QSOs in the sample!
To illustrate these effects, we give in Table 1
the ranges corresponding to the 2
$\sigma $ (95 \%) confidence level for $\Omega _{mat}$ and $\Omega _k$
respectively.
However these results are quite doubtful beyond $z=2.2$ where
we did not modify our models, while it
seems that evolution could reverse (see e.g. Hook {\it et al.} 1995).
Fortunately, the evolution governs only the maximum redshift of
magnitude--limited quasars,
which are a minority (remember Section 4 and see Paper II).
\begin{table}
\caption{ 95 \% confidence level limits read in Figs. 7 and 8
for $\Omega _{mat}$ and $\Omega _k$ for a parent
cosmology $\Omega _{mat}$ =0.6, $\Omega _k$ = -0.2 }
\halign{ \hfil # \hfill & \quad #\hfill & \quad #\hfill & \quad # &
\quad # \hfill \cr
\noalign{\hrule\medskip}
N &$z_1$ &$z_2$& $\Omega _{mat}=0.60$ & $\Omega _{k}=-0.20$ \cr
&&&&\cr
400&0.3&2.2&$\Omega _{mat}=0.70\pm 0.50$&$\Omega _{k}=-0.15\pm 0.85$\cr
400&0.3&3.5&$\Omega _{mat}=0.70\pm 0.50$&$\Omega _{k}=-0.05\pm 0.55$\cr
400&0.3&5.0&$\Omega _{mat}=0.80\pm 0.6$&$\Omega _{k}=-0.05\pm 0.45$\cr
5000&0.3&2.2&$\Omega _{mat}=0.75\pm 0.35$&$\Omega _{k}=-0.1\pm 0.5$\cr
\noalign{\medskip\hrule}}
\end{table}
We combined the last two results to look for the most
efficient use of telescope time dedicated to cosmological tests based on
quasars (see Section 7).
\begin{figure*}
\picplace{20cm}
\caption{Contour probability maps varying the parent cosmological model of Monte--Carlo catalogues. Significance levels are 0.1, 1.0 5.0 and 10 \%.
All five models are constructed under PWLE with 400 QSOs and the redshift range [0.3,2.2]. They are analysed either under PWLE (left column) or under LEXP (right column). From top to bottom, the models are: ($\Omega_{mat}=0.0, \Lambda=0.0, k_L=2.95$); ($\Omega_{mat}=1.0, \Lambda=0.0, k_L=3.60$);($\Omega_{mat}=0.0, \Lambda=1.0, k_L=1.91$); ($\Omega_{mat}=1.0, \Lambda=1.0, k_L=3.82$);($\Omega_{mat}=2.5, \Lambda=1.0, k_L=4.12$). Note the strong difference between the probability maps for the low $(<0.2)$ and high ($>0.2)$ $\Omega_{mat}$.}
\end{figure*}
\begin{figure*}
\picplace{23cm}
\caption{Same as Fig. 4, varying the binning.
A PWLE simulated catalogue
($\Omega_{mat}=\Lambda=0.6, k_L=3.47$) of 400
QSOs, with a redshift range of [0.3,2.2] is analysed under PWLE hypothesis,
according to different binnings. From left to right and top to bottom,
the number of QSOs per bin is 400, 100, 50, 40, 27, 16.}
\end{figure*}
\begin{figure*}
\picplace{23cm}
\caption{Same as Fig. 4, varying the seed.
Ten different PWLE simulated
catalogues in the
same cosmological model ($\Omega_{mat}=0.6 ,\; \Lambda=0.6,\; k_L=3.47$)
are analysed under PWLE hypothesis.
All catalogues containe 400 QSOs in the redshift range [0.3,2.2], only the
random number seed is different in each catalogue.}
\end{figure*}
\begin{figure*}
\picplace{23cm}
\caption{Same as Fig. 4, varying the catalogue size $n$.
Four PWLE simulated
catalogues in the same model ($\Omega=\Lambda=0.6, k_L=3.47$), the same
redshift range [0.3,2.2] are analysed under PWLE hypothesis. The values of
$n$ are 400,1000,3000,5000 QSOs from top to bottom and from left to
right.}
\end{figure*}
\begin{figure*}
\picplace{23cm}
\caption{Same as Fig. 4, varying the upper redshift limitation
of the catalogue. Three PWLE
simulated catalogues with 400 QSOs, in the same model
($\Omega=\Lambda=0.6, k_L=3.47$) are analysed under PWLE hypothesis. The
lower redshift is always $z_1=0.3$, and the three values of the upper
redshift are $z_2=2.2, 3.5, 5.0$.}
\end{figure*}
\begin{figure*}
\picplace{20cm}
\caption{Effects of the redshift and magnitude biases. The unbiased
catalogue contains 546 QSOs, PWLE, in the model $\Omega=\Lambda=0.6,
k_L=3.47$, and the redshift range [0.3,2.2]. All analyses are done under
PWLE hypothesis. The first
map (Fig. 9.1) is the analyse of the unbiased simulated catalog.
In the top right map, $30 \%$ of the QSOs in the redshift range [0.5,0.9]
are subtracted. In the middle left map $30 \%$ of the QSOs in the redshift
range [1.5,1.8] are subtracted. In the fourth map the last two
redshift biases are present. In the fifth map only a gaussian noise in
magnitude, of mean 0.0
and deviation 0.01, and a second component of mean 0.2 and deviation 0.2
has been done. The last map contains all the previous biases.}
\end{figure*}
\subsection{Biasing }
\subsubsection{Biasing due to lensing}
At given redshift,
a presumably small, but still unknown, fraction of QSOs
are amplified while the other ones are deamplified.
The probability that a (lens) galaxy lies closer to the line-of-sight
of a given quasar than
some threshold sufficient to induce gravitational
amplification increases with increasing quasar redshift.
As amplified quasars enter preferentially magnitude-limited
samples, the result of lensing is to mimic intrinsic luminosity
evolution. Unfortunately, we do not know the respective contributions of
intrinsic and extrinsic biasing to the observed evolution. As
the evolution law
we use is purely phenomenological, and not the prediction
of e.g. a physical model of the
emission in the environment of a massive black hole, it can fit
the effects of lensing as well - or as badly - as those of
an intrinsic luminosity evolution.
\subsubsection{Clustering}
Our test relies on the null hypothesis of
uniform spatial distribution, which is in principle not
compatible with clustering, if any. In fact it
is negligible here since our working scale is of the order
of the Gpc, and only a weak clustering scale of 40-100~Mpc seems to be
detected (Clowes and Campusano 1991, Shanks {\it et al}, 1994,
Georgantopoulos {\it et al}, 1994).
\subsubsection{Biasing in $z$}
In order to be as realistic as possible,
samples with a deficiency of objects in various redshift
ranges have been used. According to Boyle {\it et al.}
(1987; see also
V\'eron 1983), the combination of the run of color versus
redshift with a pre-selection based on UV excess results in
a defficiency around 30\% of objects in the redshift
range $0.5<z<0.9$. We analysed samples with such a random
depletion in this range. We also studied the effects of a
similar defficiency in the range $1.5<z<1.8$.
Figs. 9.2, 9.4, and 9.6 show the effect of this defficiency, which clearly
results in a general enlargement of the probability map. Fortunately we have
not detected a strong shift of the contour probabilities.
\subsubsection{Biasing in magnitude}
The photometric biasing has been analysed. We introduced a gaussian
noise of mean 0.00 and standard deviation 0.01 mag. To mimic QSO
variability, a second component of mean 0.2 and standard
deviation 0.2 has been added (Giallongo {\it et al.} 1991).
In Fig. 9.3 there is no redshift bias, but the two photometric
biases are included, and Fig. 9.5 has
been drawn with the two redshift biases plus the two photometric biases.
The absence of strong differences
with the non--biased map (Fig. 9.1) shows that the $x$ ratios of the
$\cal D$ QSOs do not depend on their magnitude.
Fig. 9.5 shows a map enlarged from Fig.9.1, which is the expected
effect caused by any bias.
\subsection{ Application to real quasars}
We now apply the test to the sample of 400 UVX quasars
of Boyle {\it et al.} (1990). This sample is based on
fiber FOCAP spectroscopy
of 1409 UVX objects selected through COSMOS photometry
in eight high-latitude Schmidt fields. The limiting $B$ magnitudes and
$U-B$ color indices are given in Boyle {\it et al.}'s Table 2 (1990).
Boyle {\it et al.} (1988) show that the best fit to the data in this
sample is
obtained with Pure Luminosity Evolution (PLE) models, far better than
PDE. The favoured functional form for LE seems to
be a power law (PWLE) rather than an exponential (LEXP).
Both evolution models depend on a single parameter $k_L$ (see Section
2). Fig. 10 gives the probability map obtained with PWLE and LEXP, which
appear to be essentially similar.
Restricting ourselves to the $\Lambda >0$ half plane, we discuss the
results at the 95 \% confidence level:
1) Regarding the matter term, we find, from Fig. 10.a,b that
$0.3<\Omega _{mat}<1.3$ (PWLE) and $0.2<\Omega _{mat}<2.$ (LEXP).
The minimum entropy in both models is the same and the corresponding
probability is $52\%$. More interestingly,
the two functional forms (power law and exponential)
lead to similar probability maps, which was expected from the analysis
of the simulated catalogues.
2) In Fig. 10.c,d,e,f the two probability maps obtained at low-
($0.3<z<1.5$) and high- ($1.5<z<2.2$) redshift respectively are compared.
The latter range does not allow any rejection among
cosmological models, although the 'cosmological signal' is believed
to increase with redshift. This is due to the
narrowness of the redshift range which does not allow one to discriminate
cosmologies, and to the poor sample size (only
112 high--z QSOs). The low-redshift map leads
to the following constraint: $\Omega_{mat}< 0.7 $ (PWLE) and
$\Omega_{mat}< 0.9$ (LEXP). As expected from simulated catalogues, at
low redshift range, the test is not sensitive to $\Lambda$.
3) Concerning the space curvature, the limits are the following. With
QSOs over the entire redshift range $[0.3,2.2]$ we read from Fig. 10.a,b:
$-1.1<\Omega_k <0.7 $ (PWLE) and $-1.1<\Omega_k <0.8$ (LEXP).
In the redshift range $[0.3,1.5]$ we find from Fig. 10.c,d,e,f:
$-0.7<\Omega_k <1.$ (PWLE), and $-0.8<\Omega_k <1.$.
The constraint on the cosmological constant is only $\Lambda < 1.4$ for
both evolution models.
It does not make sense to multiply the probability map of the
low--z sample with that obtained from the entire sample because the
samples are not independent. However we can restrict the space parameter
to the intersection of the space parameter given from these two samples.
The results are summarized in Table 2. The $1 \sigma$ confidence levels
are given in Table 3.
\begin{table}
\caption{ 95 \% confidence limits on $\Omega _{mat}$, $\Omega _k$, and
$q_0$ from low--z and all QSOs together}
\halign{ \hfil # \hfill & \qquad \hfill #\hfill &\qquad \hfill
#\hfill \cr
\noalign{\hrule\medskip}
\multispan3 \hfill $0<\Lambda <1.4$ \hfill \cr
\noalign{\medskip}
\noalign{\hrule\medskip}
&PWLE&LEXP\cr
$\Omega _{mat}$ &$0.5 \pm 0.2$&$0.55 \pm 0.35$\cr
$\Omega _k$ &$0.0 \pm 0.7$&$0.0 \pm 0.8$\cr
$q_0$ &$-0.25 \pm 1.0$&$-0.2 \pm 1.3$\cr
\noalign{\medskip\hrule}}
\end{table}
The Luminosity Functions derived from the sample
and the estimation of the characteristic evolution times in best
models are given in Paper II.
\begin{table}
\caption{ 68 \% confidence limits on $\Omega _{mat}$, $\Omega _k$, and
$q_0$ from low--z and all QSOs separately}
\halign{ \hfil # \hfill & \qquad \hfill #\hfill &\qquad \hfill
#\hfill \cr
\noalign{\hrule\medskip}
\multispan3 \hfill $0.3< z <2.2$ \hfill \cr
\noalign{\medskip}
\noalign{\hrule\medskip}
&PWLE&LEXP\cr
$\Omega _{mat}$ &$0.75 \pm 0.35$&$0.55 \pm 0.25$\cr
$\Omega _k$ &$-0.05 \pm 0.65$&$-0.15 \pm 0.55$\cr
$q_0$ &$0.1 \pm 1.2$&$-0.3 \pm 0.9$\cr
\noalign{\hrule\medskip}
\multispan3 \hfill $0.3< z <1.5$ \hfill \cr
\noalign{\medskip}
\noalign{\hrule\medskip}
&PWLE&LEXP\cr
$\Omega _{mat}$ &$0.2 \pm 0.2$&$0.25 \pm 0.25$\cr
$\Omega _k$ &$0.35 \pm 0.65$&$0.25 \pm 0.75$\cr
$q_0$ &$-0.35 \pm 1.0$&$-0.4 \pm 1.1$\cr
\noalign{\medskip\hrule}}
\end{table}
\begin{figure*}
\picplace{20cm}
\caption{ Contour probability maps (0.1, 1.0, 5.0
and 10 \%
significance level from outside to center) of the Boyle catalog. Left
(right) column is the PWLE (LEXP) analysis. At the top:
the whole redshift range $[0.3,2.2]$, in the middle:
the restricted low--redshift range $[0.3,1.5]$, and in the bottom: the
high--redshift range $[1.5,2.2]$. The latter range is too small to give
useful limitations.}
\end{figure*}
\section{Introduction}
\input intro.tex
\section{ Distribution of Maximum Volumes in a Complete Sample}
\input part2.tex
\section{ Maximum Volume and Minimum Entropy }
\input part3.tex
\section{How to compute $V/V_{ \, m\, a\, x \, }$ }
\input part4.tex
\section{ Monte--Carlo Simulations }
\input part5.tex
\section{Real QSO Samples}
\input part6.tex
\section{Summary and Discussion}
\input conclu.tex
\paragraph{Acknowledgements}
It is a pleasure to acknowledge B. Fort, P.Y. Longaretti,
S. Andreon, J. Bartlett, J.F. Le Borgne, L. Nottale
R. Pell\'o, JP Picat, G. Soucail, S. Collin--Souffrin for
all the enlighting discussions we had on QSOs and cosmology,
and especially T. Bridge for a careful reading of the manuscript.
L.V.W. thanks the french MESR for grant 93135.
This work was supported by grants from the french CNRS (GdR Cosmologie),
from the
European Community (Human Capital and Mobility ERBCHRXCT920001).
\section{References}
\input ref.tex
\end{document}
|
\section{Introduction}
\label{introduction}
\setcounter{page}{2}
\setlength{\parskip}{\baselineskip}
Microscopic nuclear structure models try to explain the excitation energies
and other experimental spectroscopic information starting from a realistic
effective nucleon-nucleon interaction. An example is the interacting shell
model. It is generally acknowledged to be the most fundamental theory of a
nucleus \cite{BRU77} \cite{MCG80} \cite{BRO88}. But since the number of
A-nucleon configurations to be diagonalized increases rapidly with the
number of single particle states, its application is limited to rather small
basis systems, like the $sd$ shell \cite{SCH87b}. Thus for many problems
the configuration spaces have to be truncated drastically. How to select
a numerically feasible number of physical relevant configurations is the
central question of all nuclear structure models which use the shell model
as the conceptual basis.
\par
One prescription for such a truncation is provided by the mean field approaches
like the Hartree-Fock \cite{HAR28} \cite{FOC30} or the Hartree-Fock-Bogoliubov
(HFB) \cite{BOG58} \cite{RIN80} theory. They use the variational principle to
describe the ground state of a nucleus by one single Slater determinant.
However, usually in this determinant all physical symmetries are broken. Thus
one has to use projection techniques to find the component with the desired
quantum numbers in order to obtain physical states. Moreover, in order to
obtain the energetically lowest solution for a given set of quantum numbers,
one has to perform the projection on the good symmetries before the mean field
is determined by variation.
\par
Following this idea, a whole hierarchy of symmetry conserving mean field
theories and their extensions into multi-configuration mixing methods have
been proposed by some of our group a few years ago \cite{SCH84a}. They are
known as the VAMPIR ({\bf V}ariation {\bf A}fter {\bf M}ean field
{\bf P}rojection {\bf I}n {\bf R}ealistic model spaces) and the MONSTER
({\bf MO}del for handling many {\bf N}umber- and {\bf S}pin-projected
{\bf T}wo-quasiparticle {\bf E}xcitations with {\bf R}ealistic interactions
and model spaces) approaches.
\par
The models of the VAMPIR family, like VAMPIR \cite{SCH84c}, EXCITED VAMPIR
\cite{SCH86a}, FED (FEw Determinant) VAMPIR and EXCITED FED VAMPIR
\cite{SCH89b}, can be used to describe the lowest few states of a given
symmetry, irrespective of their particular structure. Each state is obtained
by variational calculations. So, e.g., the VAMPIR model approximates the
lowest state of certain quantum numbers by just a single symmetry-projected HFB
vacuum. The underlying HFB transformation is determined by variation after
the projection on the desired quantum numbers. Correlating configurations
as well as excited states can be obtained by chains of similar variational
calculations.
\par
On the other hand, the MONSTER approaches are constructed for the description
of complete excitation spectra with respect to a one-body transition operator
\cite{SCH87b} \cite{SCH84b}. Thus they are well suited, e.g., for the
description of giant resonances. States which are reached by a one-body
operator are of similar structure as the initial ground or yrast state.
Therefore they can be described by expanding the nuclear wavefunction around
a symmetry-projected reference vacuum, which may be either a usual HFB, or,
e.g., a VAMPIR solution for the ground or yrast state. This is done in the
MONSTER approach. More explicitly, the MONSTER configuration space is built
out of the symmetry-projected vacuum and all symmetry-projected
two-quasiparticle excitations defined on it. The excited states are
then obtained by diagonalizing the residual interaction between these
configurations.
\par
In all realistic applications up to now, these models have been simplified
by imposing symmetry restrictions on the underlying HFB transformations.
In the first VAMPIR calculations only real, time-reversal invariant and
axially symmetric HFB transformations, which neither mix proton and neutron
states nor states of different parity, were admitted \cite{SCH84c}
\cite{SCH86a}. With this {\sl real} VAMPIR approach, as it is called, only
states in even-even nuclei with even spin and positive parity can be described.
If a MONSTER calculation is based on such a {\sl real} VAMPIR transformation,
states with different symmetries are described entirely by the
two-quasiparticle components of the wavefunction. Some years ago, this VAMPIR
approach has been improved by using HFB transformations, which are essentially
complex and allow proton-neutron and parity mixing \cite{SCH87a}. Only time
reversal and axiality are still imposed on the transformation. This
{\sl complex} VAMPIR approach can describe states of arbitrary spin
parity in even-even and odd-odd nuclei. It considers many more nucleon
correlations \cite{ZHE89}.
\par
Recently the MONSTER approach has been generalized for the use of such
{\sl complex} VAMPIR transformations \cite{BEN95a}. It was tested for
two light even-even $sd$ shell nuclei. The results were encouraging. The
performance of the {\sl complex} MONSTER description for heavier
even-even nuclei, for odd-odd and odd-mass nuclei, however, remained an
open question. In the present paper we try to answer this question by
a systematic investigation.
\par
For this purpose the {\sl complex} MONSTER model was applied to
several even-even, odd-odd and odd mass systems. The calculations were
performed again in an $sd$ shell basis, in order to be able to compare
with the exact shell model results. The considered nuclei of each type are
covering the range from very light systems to the ones with the biggest
possible numbers of shell model configurations in the $sd$ shell. The results
are compared to those of the older more restricted {\sl real} MONSTER approach,
too. It is demonstrated that the {\sl complex} MONSTER approach approximates
the shell model results very well and shows a clear improvement on the older
{\sl real} approach.
\par
In the next section the main features of the MONSTER on VAMPIR model are
summarized and the consequences of the symmetry restrictions for the
HFB transformation are shortly explained. In section (\ref{application})
the results of the application are presented and discussed. Finally
conclusions and an outlook are given in (\ref{summary}).
\section{The Model}
\label{model}
The detailed formulation of the MONSTER on VAMPIR approach has been given
elsewhere \cite{BEN95a}. Therefore only the main ideas of the model are
sketched here.
\subsection{The VAMPIR model}
\label{vampirmodel}
A finite model space is defined by a $D$-dimensional set of orthonormal
single particle wave functions, which are eigenstates of some spherically
symmetric one-body potential, e.g. the harmonic oscillator. In second
quantization the corresponding creation and annihilation operators are
denoted by $ \{ c_i^{\dagger}, c_k^{\dagger}, \ldots \}_{D} $ and
$ \{ c_i, c_k, \ldots \}_{D} $, respectively. The indices $i$, $k$
summarize the quantum numbers of a state. It is assumed that
the effective many body Hamiltonian appropriate for this model space
is known and can be represented by a sum of only one- and two-body terms
\begin{equation}
\label{hamiltonian}
\hat{H} = \hat{T} + \hat{V} .
\end{equation}
The one-body part $\hat{T}$ contains the matrix elements of the kinetic energy
(or, if an inert core is assumed, the single particle energies) and the
two-body part $\hat{V}$ the matrix elements of the effective interaction.
\par
Starting from the chosen particle basis, the Hartree-Fock-Bogoliubov (HFB)
transformation \cite{RIN80} is used to define the corresponding quasiparticle
basis with its quasiparticle creation and annihilation operators,
$ \{ a_{\alpha}^{\dagger}, a_{\beta}^{\dagger}, \ldots \}_{D} $ and
$ \{ a_{\alpha}, a_{\beta}, \ldots \}_{D} $. $\alpha$,~$\beta$ enumerate the
quasiparticle states. The HFB transformation is the most general linear
transformation which ensures that the Fermion anti-commutation relations are
fulfilled by the quasiparticle operators. In matrix notation it is given by
\begin{equation}
\label{hfbtransformation}
\left(
\begin{array}{c} a^{\dagger}(F) \\ a(F) \end{array}
\right)
=
F \left(
\begin{array}{c} c^{\dagger} \\ c \end{array}
\right)
\end{equation}
with $F$ being a unitary $2D\!\times\! 2D$ dimensional matrix.
The vacuum $|F\rangle$ for a set of quasiparticles obtained by the HFB
transformation $F$ is defined in the usual way by the request that application
of any quasiparticle annihilation operator onto it should yield zero,
\begin{equation}
\label{vacuum}
a_{\alpha}(F) |F\rangle = 0 \quad \forall \: \alpha =1,\ldots,D .
\end{equation}
An $n$-quasiparticle state with respect to the vacuum is defined by
\begin{equation}
\label{n-qpstate}
| F \{ a^{\dagger} \}_n \rangle =
\left( \prod _{\alpha = 1} ^n a^{\dagger}_{\alpha}(F) \right)
| F \rangle \quad \mbox{for} \quad n=1 ,\ldots ,D .
\end{equation}
\par
The vacuum as well as the $n$-quasiparticle states contain components of many
nucleon configurations with different angular momenta, different angular
momentum $z$-components, both parities and various proton and neutron numbers
\cite{MAN75}. In order to get physical states which are characterized by good
quantum numbers, projection techniques have to be used for selecting the
components with the desired symmetries. The corresponding projection operator
$\hat{\Theta}^{AT_zI^{\pi}}_{MK}$ is a product of projection operators,
\begin{equation}
\label{symmetryprojector}
\hat{\Theta} ^{AT_zI^{\pi}}_{MK} \equiv
\hat{P}(IM;K) \hat{P}(2T_z) \hat{P}(A) \hat{P}(\pi) ,
\end{equation}
where the first projects on good angular momentum $I$
with $z$-component $M$, the next ones on good isospin $z$-component $T_z$,
good mass number $A$ and finally on good parity $\pi$. Explicit expressions
for them can be found in \cite{SCH87a}. All projectors apart from
$\hat{P}(\pi)$ are integral operators.
\par
Physical configurations are obtained by applying the operator
$\hat{\Theta}^{AT_zI^{\pi}}_{MK}$ on the above introduced quasiparticle
configurations. Since any of the $n$-quasiparticle configurations
$| F \{ a^{\dagger} \}_n \rangle $ built on the vacuum $|F\rangle$
can be looked upon at the same time as a vacuum for another HFB transformation
$F^{\prime}$, it is enough to consider only HFB vacua in the following.
Thus, a physical configuration with a good symmetry $S$, where $S$ represents
the quantum numbers $AT_zI^{\pi}$, is given by
\begin{equation}
\label{projectedvacuum}
|F;SM\rangle =
\sum_{K=-I}^{I}
\hat{\Theta}^S_{MK} |F\rangle f_K^S .
\end{equation}
The sum over all intrinsic angular momentum $z$-components $K$ has to be taken
in order to avoid a dependency of the projected wave function on the
orientation of the intrinsic reference frame \cite{SCH84a}.
\par
In the VAMPIR approach one symmetry-projected vacuum (\ref{projectedvacuum})
is used to describe the energetically lowest state of a given symmetry $S$.
The configuration mixing degrees of freedom $f_K^S$ and the underlying HFB
transformation $F$ are determined by variation
\begin{equation}
\label{vampirvariation}
\delta E^S \equiv
\delta \frac{\langle F;SM | \hat{H} | F;SM \rangle}
{\langle F;SM | F;SM \rangle}
= 0 .
\end{equation}
This variation leads to three sets of equations which have to be solved
self-con\-sis\-tent\-ly \cite{SCH87a}. One obtains the optimal description of
the considered yrast state which can be achieved by using only a single
determinant. Generally this VAMPIR approach yields already a rather good
approximation to the actual state \cite{ZHE89}.
\par
A similar approach, the FED VAMPIR, uses a linear combination of a few
symmetry-projected vacua to describe one yrast state. Here, the correlating
configurations, which contribute to one state, are searched for by successive
variational calculations \cite{SCH89b}. Furthermore, by introducing
orthogonality constraints, the VAMPIR and the FED VAMPIR procedures can be
easily extended to the description of excited states. These EXCITED VAMPIR
\cite{SCH87a} and EXCITED FED VAMPIR \cite{SCH89b} models can describe states
of any arbitrary structure. But since each state is searched for by variation,
a numerical application is quite time consuming. Thus the VAMPIR models are
especially well suited for the description of the lowest few states of a
given symmetry $S$.
\par
For problems which require a complete excitation spectrum with respect to
some transition operator, it is better to follow another avenue.
Since here only specific configurations are needed, one can describe them
by expanding the nuclear wave functions around a suitable VAMPIR vacuum.
This is explained in the next section.
\subsection{The MONSTER on VAMPIR approach}
\label{monsteronvampir}
For states with symmetry $S$ the configuration space is chosen to consist
of the symmetry-projected vacuum
\begin{equation}
\label{monsterconfigurationspace2a}
|F;SMK\rangle \equiv \hat{\Theta}^S_{MK} |F\rangle
\end{equation}
and the symmetry-projected two-quasiparticle states with respect to it
\begin{equation}
\label{monsterconfigurationspace2b}
|F\alpha\beta;SMK\rangle
\equiv
\hat{\Theta}^S_{MK} a^{\dagger}_{\alpha}(F) a^{\dagger}_{\beta}(F)
|F\rangle .
\end{equation}
The underlying HFB transformation $F$ is fixed in a preceding VAMPIR
calculation for an yrast state, which not necessarily needs to have just
the same symmetry $S$. In many cases it is sufficient to use the HFB
transformation obtained for the ground state of the considered nucleus.
\par
A general wavefunction for excited states is then given by
\begin{equation}
\label{monsterwavefunction}
|\psi_i(F);SM\rangle
= \left\{ \sum_{K=-I,\ldots ,+I} |F;SMK\rangle g^S_{0K;i}
+ \sum_{{\alpha < \beta} \atop K=-I,\ldots ,+I}
|F\alpha\beta ; SMK \rangle g^S_{\alpha\beta K;i} \right\} .
\end{equation}
The expansion coefficients $g^S$ are obtained by diagonalizing the effective
Hamiltonian in the space of the nonorthogonal configurations
(\ref{monsterconfigurationspace2a}, \ref{monsterconfigurationspace2b}).
\par
The above choice of the model space restricts the MONSTER on VAMPIR
approach to excited states with a structure similar to that of the
underlying projected vacuum. This is here desired, since we are interested
in states predominantly populated by a transition from the vacuum
via a one-body operator. Furthermore it should be mentioned
that the MONSTER approach provides an approximate elimination of spurious
center-of-mass excitations, which can be introduced by the configuration
mixing if more than one major oscillator shell is taken as particle basis
(for details see \cite{BEN95a}).
\subsection[Symmetry restrictions]
{Symmetry restrictions imposed on the \\
HFB transformation}
\label{restrictions}
If the most general HFB transformations are allowed, symmetry-projected vacua
of type (\ref{projectedvacuum}) can describe arbitrary states in arbitrary
nuclei \cite{SCH87a}. Up to now, however, this has not been achieved in any
numerical implementation. Instead, for the existing numerical realizations
of the VAMPIR approaches, certain symmetry requirements were imposed on the
underlying HFB transformations.
\par
In the first VAMPIR calculations \cite{SCH84c} axial symmetry and time-reversal
invariance were assumed, parity and proton-neutron mixing were neglected and
only real HFB transformation coefficients were admitted. As a consequence of
these approximations the {\sl real} VAMPIR approach was only suitable for
positive parity states with even angular momenta in doubly even nuclei.
Performing MONSTER calculations on top of such a {\sl real} VAMPIR solution
these limitations are removed. In the two-quasiparticle approximation not
only states with arbitrary spin-parity in doubly even but also in doubly odd
systems become accessible. However, for the description of odd spin and/or
negative parity states in doubly even nuclei one has to use mean fields
obtained for different spin and/or parity values than the considered one. And
for doubly odd systems one even has to rely on mean fields for neighbouring
systems. Considering as configuration space for the MONSTER calculation
all symmetry-projected one-quasiparticle states
\begin{equation}
\label{1qpmonsterconfigurationspace}
|F\alpha;SMK\rangle
\equiv
\hat{\Theta}^S_{MK} a^{\dagger}_{\alpha}(F) |F\rangle ,
\end{equation}
arbitrary states in odd systems become approachable, too.
Here again, one has to build the one-quasiparticle spectrum on
the mean field determined for a neighbouring doubly even nucleus.
\par
In the more recent implementations of the VAMPIR calculations parity as well
as proton-neutron mixing are taken into account and essentially complex HFB
transformations are admitted. In this {\sl complex} VAMPIR approach only
axial symmetry and time reversal invariance are kept \cite{SCH87a}.
This introduces many more correlations into the projected vacua than
the older {\sl real} calculations. Now states of arbitrary spin-parity
in both doubly even and doubly odd systems are accessible. The {\sl complex}
vacua contain all possible two nucleon couplings, but some
four-, six- and more nucleon couplings are missing \cite{ZHE89}. Only recently
the MONSTER approach could be extended to such {\sl complex} VAMPIR vacua
\cite{BEN95a}. Here, in addition to doubly even nuclei, also the calculation
for doubly odd nuclei can be based on a HFB transformation determined for
just the considered system itself. And the structures which are missing in
the {\sl complex} VAMPIR vacuum can be
introduced by the two-quasiparticle admixtures. However, for the description
of odd systems one still has to use the one-quasiparticle approximation
like in the older {\sl real} approach. But now aside from the mean
fields of neighbouring even-even nuclei also the ones of neighbouring
odd-odd nuclei can be chosen to construct the one-quasiparticle
states~(\ref{1qpmonsterconfigurationspace}).
\par
Generally, since the underlying HFB transformation is much more general
and the configuration space is much larger, the {\sl complex} MONSTER
is expected to yield a better description of nuclear spectra
than the {\sl real} one.
\section{A systematic application to {\it sd} shell nuclei }
\label{application}
In \cite{BEN95a} we presented as a first test of the new {\sl complex} MONSTER
approach an application to $^{20}$Ne and $^{22}$Ne. Considering the $sd$ shell
as model space, it was shown that the model is able to reproduce the
shell model spectrum of $^{20}$Ne exactly. For $^{22}$Ne it yielded an
excellent approximation to the shell model result. A clear improvement on
the older {\sl real} approach was found. However, the two investigated nuclei
are even-even systems. Moreover, they are quite light~: for both nuclei
the number of shell model states with a certain spin is comparable to
the number of {\sl complex} MONSTER configurations. How does the approach
perform for heavier nuclei, where the number of states is much larger than
the number of {\sl complex} MONSTER configurations ? Furthermore, which
kind of agreement is obtained for odd-odd nuclei and for odd mass systems,
where the simpler one-quasiparticle description is used ? These questions are
adressed in the present investigation.
\par
Again the $1s0d$ shell was chosen as single particle basis in
order to be able to compare with exact shell model configuration mixing
calculations. The single particle energies
$\epsilon\,(d{\,5/2}) = -4.15\mbox{ MeV}$,
$\epsilon\,(s{1/2}) = -3.28\mbox{ MeV}$, and
$\epsilon\,(d\,{3/2}) = +0.93\mbox{ MeV}$ for both protons and neutrons
\cite{AJZ86} and the effective two-body residual interaction
(the mass-dependent version of the Chung and Wildenthal force \cite{WIL83}
($\hat{V}(A) = \hat{V}(18) \times (18/A)^{\alpha}$) with $\alpha = 1/3$
instead of $\alpha = 0.3$) are the same as in our earlier calculations.
This force is generally accepted to be ``the standard'' force for
the $sd$ model space. The results are compared in addition to those
obtained by the more restricted {\sl real} MONSTER approach.
\subsection{Even-even nuclei}
\label{eveneven}
As examples for ``heavier'' even-even $sd$-shell nuclei, we calculated the
spectra of $^{24}_{12}$Mg, $^{26}_{12}$Mg, $^{28}_{14}$Si and $^{30}_{14}$Si
within the {\sl complex} and the {\sl real} MONSTER approaches.
For each MONSTER calculation the underlying HFB transformation was determined
by a preceding VAMPIR calculation for the $0^+$-ground state. This HFB
transformation F($0^+$) has then been used for states of all spins in
one nucleus.
\par
The energies obtained by the {\sl complex} and the more restricted {\sl real}
MONSTER have been compared to those from an exact shell model
diagonalization. The result of this investigation is summarized in table~1.
The average deviation of the energies of the yrast states with spin $0^+$
to $5^+$ from the shell model energies is presented for each nucleus. The
deviation is given in keV and in percent of the shell model ground
state energy of the nucleus referred to. The exact ground state energy is also
displayed. For comparison the corresponding values for the nuclei $^{20}$Ne
and $^{22}$Ne of our first application are listed here, too. The table
shows furthermore the minimum and maximum number of shell model configurations
per spin for each nucleus and the corresponding spin value. In addition the
number of shell model states per spin averaged over the states with spin $0^+$
to $5^+$ is given. It should be compared to the number of MONSTER
configurations available in the $sd$-shell. As explained in more detail
in \cite{BEN95a} the {\sl complex} MONSTER provides 57 configurations for
the description of spin $0^+$ states, 151 for $1^+$, 223 for $2^+$, 259
for $3^+$, 275 for $4^+$, and always 277 for all higher spin states. In the
more restricted {\sl real} MONSTER approach less configurations per spin
are available~: 21 for the $I=0$ states, 30 for $I=1$, 61 for $I=2$,
56 for $I=3$, always 60 for $I=5,7,9,11,13$, and always 73 for the
$I=4,6,8,10,12,14$ states, respectively.
\par
As expected for both MONSTER approaches the mean deviation of the yrast
state energies from the exact shell model result, averaged for spins $0^+$
to $5^+$, does increase with the number of actually existing shell model
states. For the {\sl complex} MONSTER the largest deviation measured
relatively to the shell model ground state energy occurs for $^{26}$Mg
and amounts to 1.43\% . For $^{20}$Ne the {\sl complex} calculation reproduced
the exact result. In the {\sl real} case the largest average deviation,
2.42 \% , is found for $^{24}$Mg and the smallest, 0.87\% , for $^{22}$Ne.
For all nuclei the {\sl complex} MONSTER approach yields a clear improvement
on the {\sl real} one. The largest energy gain could be achieved in
$^{24}$Mg, where the averaged deviation of the {\sl real} MONSTER is 3.8
times larger than the one of {\sl complex} MONSTER. The smallest amount
of energy was won for $^{26}$Mg, where the average deviation of the {\sl real}
approach is by a factor of 1.3 larger compared to the {\sl complex} one.
For all even-even nuclei the {\sl complex} MONSTER approach yields an
excellent approximation to the exact calculation.
\par
As an example, in fig.~1, the spectra of $^{28}$Si as obtained by the
{\sl complex} and the {\sl real} MONSTER are shown up to $\sim$~14~MeV
above the shell model ground state energy. For comparison the energies
resulting from the shell model diagonalization are plotted up to $\sim$~12~MeV
excitation energy. It can be seen that the {\sl complex} MONSTER reproduces
the shell model spectrum well and shows a clear improvement on the
{\sl real} MONSTER~: the states are considerably more bound and
many more shell model states can be described.
\par
The nucleus $^{28}$Si was investigated in more detail, since it has the
largest number of states of all $sd$-shell nuclei. There exist in total
93710 states. More precisely, there are 3372 $I^{\pi}\!\!=\!0^+$ states,
9216 $I^{\pi}\!\!=\!1^+$, 13562 $I^{\pi}\!\!=\!2^+$, 15385 $I^{\pi}\!\!=\!3^+$,
15089 $I^{\pi}\!\!=\!4^+$, 12876 $I^{\pi}\!\!=\!5^+$ and
9900 $I^{\pi}\!\!=\!6^+$ states. For the higher spins the number of states
per spin continues to decrease; there is just one state with the highest
possible spin $I^{\pi}\!\!=\!14^+$. In fig.~2 the yrast state energies as
obtained by various models of the VAMPIR and MONSTER family are compared
for the spins $0^+$ to $6^+$. The energies from the exact shell model
calculation are given in the leftmost column. Then, from left to right,
the energies calculated by the {\sl complex} MONSTER, the {\sl complex}
VAMPIR, the {\sl real} MONSTER, and the {\sl real} VAMPIR are shown. Note
that the {\sl real} VAMPIR approach can only describe even spin states.
\par
The average deviation of the energies obtained by the
{\sl complex} MONSTER approach from the shell model, now including the
$6^+$ yrast state, is about 1085~keV or 0.79\% of the shell model
ground state energy of 137.495~MeV. The deviation for the even spin states
is on average somewhat smaller, namely 840~keV, than the one for
the odd spin states, which amounts to 1412~keV. This was found also
for the other investigated even-even nuclei, apart from $^{24}$Mg, where
the difference is insignificant. There the average deviation of the odd
spin states was only 51~keV larger than for the even spin states.
{\sl Real} MONSTER gives for $^{28}$Si an average deviation of 1746~keV,
which amounts to 1.27\% of the shell model ground state energy.
\par
As it is to be expected, the {\sl complex} MONSTER yrast states are more
bound than the ones calculated by {\sl complex} VAMPIR, except for the
$0^+$ VAMPIR state, which is stable with respect to any projected
two-quasiparticle excitations of the same symmetry. Obviously the odd spin
states $1^+$, $3^+$ and $5^+$ are not well described by the {\sl complex}
VAMPIR. When the difference between the binding energies obtained by the
two models is so pronounced, it can be inferred that these state are dominated
by the above mentioned ``missing couplings'', which are not accessible within
the {\sl complex} VAMPIR model. This was also found in Ref.~\cite{BEN95a},
where the odd spin yrast states of $^{22}$Ne were rather poorly described by
the {\sl complex} VAMPIR approach.
\par
The differences in the binding energies of {\sl real} MONSTER and {\sl real}
VAMPIR are negligibly small. The even spin {\sl real} MONSTER yrast states,
$2^+$, $4^+$ and $6^+$, are on average only 56~keV more bound. The
$0^+$ state does not mix with the projected two-quasiparticle states
of the same symmetry in the {\sl real} case, too.
\subsection{Odd-odd nuclei}
\label{oddodd}
In order to study the quality of the {\sl complex} MONSTER approach for
applications to odd-odd nuclei, we calculated the energy spectra of
$^{20}_{\ 9}$F, $^{22}_{11}$Na, $^{24}_{11}$Na, $^{26}_{13}$Al,
$^{28}_{13}$Al and $^{30}_{15}$P. For each nucleus the {\sl complex} MONSTER
configuration space was built on the HFB transformation, which yielded the
energetically lowest VAMPIR solution in just this nucleus, no matter which
spin it corresponded to. The spectra have also been calculated with {\sl real}
MONSTER. In this case one has to take the HFB transformation of a neighbouring
even-even nucleus to construct the MONSTER configuration space. Naturally, one
has several possibilities~: for an odd-odd nucleus with Z protons and N neutrons
one can take the even-even neighbours with the proton and neutron numbers
(Z-1,N-1), (Z-1,N+1), (Z+1,N-1), and (Z+1,N+1). It is a priori not clear,
which choice may be the best. Therefore we looked in each neighbour
for the HFB transformation, which belongs to the $0^+$ ground state
VAMPIR solution and used each for a subsequent {\sl real} MONSTER calculation.
\par
The energies obtained by both MONSTER approaches have been compared
to the ones of a full shell model configuration mixing calculation.
To get a quantitative measure, for each nucleus as already for
the even-even nuclei, the average deviation of the energies of the yrast
states with spins $0^+$ to $5^+$ from the shell model energies has been
calculated. The resulting deviations are shown in table~2, again in keV and in
percent of the shell model ground state energy. Besides the energy
the table displays the spin of the shell model ground state and, for
{\sl complex} MONSTER the spin of the VAMPIR solution which was used. For
{\sl real} MONSTER always the result for that spectrum which yielded the best
agreement with the shell model is listed in the table. The corresponding
neighbouring nucleus is indicated. The {\sl real} MONSTER spectra for
one odd-odd nucleus, obtained from the HFB transformations of the various
possible neighbours, differ often quite much. So, e.g., for $^{22}$Na
the average deviation of the yrast states $0^+$ to $5^+$ is 4.85\% if the
HFB transformation for the $0^+$ VAMPIR solution of $^{22}$Ne is used,
and 1.47\% for the one of $^{20}$Ne.
\par
Table~2 also shows the minimum and maximum number of shell model
configurations per spin and the number per spin averaged for spins
$0^+$ to $5^+$. Since protons and neutrons are mixed by the {\sl complex} HFB
transformation, the same configurations are used for the calculation of
even-even and odd-odd nuclei in a {\sl complex} MONSTER calculation. Thus the
number of configurations contributing to one spin given in section
(\ref{eveneven}) is valid here, too. For the {\sl real} MONSTER with no
proton-neutron mixing the situation is different compared to the even-even
case. For the description of odd-odd nuclei in $sd$ shell there are
14 configurations available for the $0^+$ states, 39 for $1^+$, 57 for $2^+$,
67 for $3^+$, 71 for $4^+$ and always 72 for the states with spins
$5^+, \ldots 13^+$. These numbers are comparable in size to the ones of the
even-even case.
\par
As already for the even-even nuclei, also for the odd-odd ones an increase
of the average deviation with the number of shell model states per spin is
observed. For $^{20}$F the whole shell model spectrum is exactly reproduced by
{\sl complex} MONSTER using any {\sl complex} HFB transformation $F$. This is
to be expected, since $^{20}$F belongs to the same isospin multiplet like
$^{20}$Ne. $^{20}$Ne encloses states of all isospins $T\! =\! 0, 1, 2$,
$^{20}$F only the ones with isospins $T\! =\! 1$ and $2$. This provides
another stringent test for the numerics. The largest deviation relative to
the shell model ground state energy was found for $^{24}$Na~: 2.21\%.
$^{24}$Na is a typical example for which one should better choose different
HFB transformations for states with different spins, instead of using one
transformation for all states. If on uses the HFB transformation of the $0^+$
VAMPIR solution for the low spin states $0^+$ to $3^+$ and the transformation
F($4^+$) for the higher spins, the average deviation of the yrast states
$0^+$ to $5^+$ can be reduced to 995~keV or 1.26\% of the shell model ground
state energy. The possibility to choose different HFB transformations for
different spins should always be considered, since we found that the
{\sl complex} MONSTER energies are quite sensitive on the underlying HFB
transformation in contrast to the {\sl real} case \cite{BEN95a}.
\par
The average deviations of the {\sl real} MONSTER are always bigger than
for the {\sl complex} approach, apart from $^{24}$Na, where both are of
similar size. But this could be improved by using several HFB
transformations in the {\sl complex} case. As far as the higher excited
states are concerned, the {\sl complex} MONSTER is always superior to the
{\sl real} one, since it is able to describe many more states.
\par
Compared to the even-even case the deviations obtained by {\sl complex}
MONSTER are somewhat bigger for each mass number. It is known that the
structure of odd-odd nuclei is generally more complicated. More types of
couplings contribute significantly. This seems to be reflected in our results.
But also for the odd-odd case the exact spectra are over all rather
well approximated by the MONSTER approach.
\par
In fig.~3 the energy spectra of $^{22}$Na calculated with the {\sl complex}
and the {\sl real} MONSTER are plotted up to spin 7 and an energy of $-52$~MeV.
For comparison the lowest shell model energies for each spin are presented.
In the {\sl real} case, MONSTER is built on the HFB transformation determined
for the $0^+$ VAMPIR solution in $^{20}$Ne. This gave the best agreement with
the shell model. The most bound {\sl complex} VAMPIR solution in this
nucleus is the one for the $1^+$ yrast state. {\sl Complex} MONSTER on top of
the corresponding HFB transformation corrects then the sequence of the yrast
states to the one of the exact calculation, where the ground state has spin
$3^+$. The largest deviation of a {\sl complex} MONSTER yrast state energy
from the shell model value occurs for the $0^+$ yrast state~: 362~keV,
the smallest for $7^+$~: 83~keV. {\sl Complex} MONSTER shows
also for the higher lying states an excellent agreement with the energies
of the exact shell model and yields a significant improvement on the
{\sl real} MONSTER.
\par
Finally, as an example for the heavier odd-odd nuclei, fig.~4 displays the
calculated energies of $^{30}$P. In this figure two {\sl real} MONSTER
calculations are presented, the ones yielding the best and the worst agreement
with the result of the exact diagonalization. The former has been got by using
the HFB transformation for the $0^+$ VAMPIR solution in $^{28}$Si, the latter
by the HFB transformation for the $0^+$ VAMPIR solution in $^{32}$S.
The lowest {\sl complex} VAMPIR solution in $^{30}$P was found to be the
one with spin $0^+$. Its HFB transformation was used in the {\sl complex}
calculation. The {\sl complex} and the {\sl real} MONSTER results are
plotted up to $\sim$11~MeV above the shell model ground state for spins
$0^+$ to $5^+$. For each spin the lowest energies of the exact
diagonalization are given. {\sl Complex} MONSTER reproduces the general
trends of the spectrum rather well. The largest deviation of the yrast
state energies is found for spin $2^+$ and amounts to 1579~keV or 1.01\% of
the shell model ground state energy, the smallest,
1157~keV or 0.74\%, for spin $0^+$. The {\sl complex} calculation shows a
clear improvement on both {\sl real} ones, where much less states are
described.
\subsection{Odd mass nuclei}
\label{oddmass}
As described in section (\ref{restrictions}), in the case of odd mass nuclei
the {\sl complex} MONSTER spectrum has to be built on a HFB transformation
determined for a neighbouring even mass nucleus. We proceeded here in
the following way. We calculated the shell model spectra of several of those
odd mass nuclei, which are neighbouring the even mass nuclei discussed
in sections (\ref{eveneven}) and (\ref{oddodd}). Then we used the HFB
transformations, which we had already determined, to built up the MONSTER
one-quasiparticle configuration spaces. Accordingly, for the mass parameter
of the Chung-Wildenthal force, we took always the mass of the even mass
nucleus of which the HFB transformation was used. The energy spectra of
$^{21}_{10}$Ne, $^{23}_{11}$Na, $^{25}_{12}$Mg, $^{27}_{13}$Al and
$^{29}_{14}$Si have been calculated. For $^{21}$Ne the HFB transformations
of the deepest VAMPIR solutions in $^{22}$Ne and $^{22}$Na have been used. The
other odd mass nuclei with mass number A have been calculated always in four
different ways~: using the HFB transformation for the deepest VAMPIR solution
of the even-even neighbour with mass A$-$1, the one of the even-even neighbour
with mass A$+$1 and the ones of the odd-odd neighbours with A$-$1 and A$+$1.
Thus, e.g., $^{29}$Si has been calculated using the mean fields of $^{28}$Si,
$^{30}$Si, $^{28}$Al and $^{30}$P. In the case of {\sl real} MONSTER only
the mean fields of the two neighbouring even-even nuclei are available.
Both have been used for the {\sl real} calculations.
\par
The MONSTER configuration space for the calculation of odd mass nuclei is
much smaller than the one for the even mass nuclei, because there
are much less one-quasiparticle than two-quasiparticle excitations.
For {\sl complex} MONSTER in the $sd$ shell there are 12 configurations
available for states with spin $I\! =\! 1/2$, 20 for $I\! =\! 3/2$ and
always 24 for states with higher spins. The {\sl real} MONSTER provides even
only 3 configurations for $I\! =\! 1/2$ states, 5 for $I\! =\! 3/2$, and always
6 for all higher spin states.
\par
Let us first turn our attention to $^{21}$Ne. It is the lightest odd mass
nucleus which we have calculated. For this nucleus there are in total 1935
shell model configurations. More precisely, there are 199 $I\! =\! 1/2$,
341 $I\! =\! 3/2$, 400 $I\! =\! 5/2$, 368 $I\! =\! 7/2$, 287 $I\! =\! 9/2$,
183 $I\! =\! 11/2$, 100 $I\! =\! 13/2$, 41 $I\! =\! 15/2$, 14 $I\! =\! 17/2$,
and 2 $I\! =\! 19/2$ states. Fig.~5 shows for each spin the lowest energies
resulting from the complete shell model diagonalization up to spin
$I\! =\! 11/2$. In addition the results of two {\sl complex} MONSTER
calculations are displayed up to an energy of $-38$~MeV. One calculation
was done by using the mean field of the $0^+$ ground state for $^{22}$Ne, the
other is built on the mean field of the deepest VAMPIR solution for $^{22}$Na,
i.e., the $1^+$ solution. In fig.~5 also the result of a {\sl real} MONSTER
calculation is plotted for comparison. It is built on the HFB transformation
determined for the $0^+$ ground state of $^{22}$Ne by a {\sl real} VAMPIR
calculation. The obtained agreement between both {\sl complex} MONSTER
results and the shell model is very good. For the calculation built on
$^{22}$Ne the average deviation of the yrast state energies from the exact
ones is 490~keV. That is 1.03\% of the shell model ground state energy,
which is $-47.753$~MeV. The largest deviation occurs for the $11/2$ state
and amounts to 664~keV or 1.39\% of the shell model ground state energy,
the smallest for spin $5/2$, and is 256~keV or 0.54\%. The calculation using
the mean field of $^{22}$Na gives an average deviation of 463~keV or 0.97\%
measured relative to the shell model ground state energy. The largest
deviation, 653~keV or 1.37\%, was found for the $1/2$ state, the smallest,
308~keV or 0.64\%, for the $7/2$ yrast state. Thus, the mean field of the
$1^+$ yrast state of $^{22}$Na is on average slightly better suited for
the description of the yrast states of $^{21}$Ne than that of the $0^+$
ground state of $^{22}$Ne. Except for the spin $1/2$ and $5/2$ yrast states,
it yields deeper bound states. Also the excited states obtained by this
calculation display the bigger binding energies. The {\sl real} MONSTER
gives for the yrast state energies an average deviation of 576~keV or
1.21\% from the shell model ones. The $1/2$ state shows the largest
deviation~: 704~keV or 1.47\%, the $9/2$ state the smallest~:
467~keV or 0.98\%. For most of the yrast states both {\sl complex} MONSTER
calculations yield more binding than the {\sl real} ones, as it is expected.
However, e.g., for the $9/2$ yrast state the {\sl real} MONSTER yielded
the deepest energy. The {\sl complex} results, built on the mean fields of
$^{22}$Na and $^{22}$Ne, are higher by 16~keV and 114~keV, respectively.
This indicates that in the {\sl complex} case, the mean field determined
for a low spin may not be well suited for the description of high spin states.
\par
As example for an heavier odd mass nucleus, we present the energy
spectra of $^{29}$Si obtained by using the mean fields of the mass 28 systems.
The number of shell model states in this nucleus is huge compared to the
number of MONSTER configurations available. In total this nucleus has
80112 shell model states. 5638 of these states have spin $I\! =\! 1/2$,
10176 $I\! =\! 3/2$, 12877 $I\! =\! 5/2$, 13450 $I\! =\! 7/2$,
12240 $I\! =\! 9/2$, 9835 $I\! =\! 11/2$, 7053 $I\! =\! 13/2$,
4469 $I\! =\! 15/2$, 2502 $I\! =\! 17/2$, 1197 $I\! =\! 19/2$,
485 $I\! =\! 21/2$, 152 $I\! =\! 23/2$, 35 $I\! =\! 25/2$,
and 3 are spin $I\! =\! 27/2$ states. For each spin up to spin $9/2$ the
lowest energies resulting from the exact shell model calculation are shown
in fig.~6. The obtained MONSTER spectra are plotted up to an energy of
$\sim$ 10~MeV above the shell model ground state energy of $-147.020$~MeV.
One {\sl complex} MONSTER spectrum is obtained by using the HFB transformation
of the VAMPIR solution for the $0^+$ ground state in $^{28}$Si, the other
uses the transformation of the $0^+$ VAMPIR solution in $^{28}$Al. The
calculation on top of the mean field of the even-even nucleus shows an
average deviation of the yrast state energies from the exact ones of 1373~keV
or 0.93\% in terms of percentage of the shell model ground state energy.
The yrast state with spin $9/2$ shows the largest deviation, 1575~keV or
1.07\%, the $7/2$ yrast state the smallest, 1291~keV or 0.88\%.
The average deviation obtained by using the mean field of the odd-odd nucleus
is somewhat bigger, namely 1912~keV or 1.30\%. The largest and smallest
deviations, which occur, are 2172~keV or 1.48\% and 1607~keV or 1.09\%,
respectively. The general trend of the shell model spectrum is reproduced
by both {\sl complex} calculations. In fig.~6 in addition also a {\sl real}
MONSTER spectrum is shown. It was obtained by using the HFB transformation
of the $0^+$ VAMPIR solution in $^{28}$Si. This spectrum yielded a better
agreement with the shell model result than the {\sl real} calculation on top
of the mean field for $^{30}$Si. For the displayed {\sl real} MONSTER spectrum
the average deviation of the yrast state energies from the exact ones is
1973~keV or 1.34\%. The largest deviation occurs for the $1/2$ yrast state,
2286~keV or 1.55\%, the smallest for $7/2$, 1640~keV or 1.12\%. The
{\sl complex} MONSTER calculation on top of the mean field of
$^{28}$Si shows a clear improvement over this best {\sl real} result.
\par
For all the considered odd mass nuclei, apart from $^{21}$Ne, we found that
by using the HFB transformation of a neighbouring even-even nucleus, a better
agreement between the {\sl complex} MONSTER result and the shell model could
be achieved than by using the mean field of an odd-odd neighbour. In general
the odd mass spectra on top of a mean field of a neighbouring even-even
nucleus approximated the shell model results well (more details can be found
in \cite{BEN95c}). The only exceptions were the spectrum of $^{23}$Na obtained
by using the HFB transformation of the $0^+$ VAMPIR solution for $^{22}$Ne
and the spectrum of $^{25}$Mg with the transformation for $^{26}$Mg. These two
calculations gave a rather poor agreement with the exact result. In both cases
it could be much improved by using the HFB transformation of $^{24}$Mg.
Thus, in the odd mass case the choice of the neighbour can be crucial.
One should always try several ones to find the mean field, of which the
structure is best suited for the odd mass nucleus under consideration.
\par
Finally we investigated for the odd mass case the dependence of the
{\sl complex} MONSTER energies on the spin of the mean field determined by
the preceding VAMPIR calculation. For $^{21}$Ne we built always the full
MONSTER spectrum on each of the yrast solutions of $^{22}$Ne obtained by
{\sl complex} VAMPIR. The same was done for {\sl real} MONSTER. Naturally
here only mean fields for even spin states are available.
\par
Fig.~7 shows a typical result of this investigation. The energies of the
five lowest $3/2$ states of $^{21}$Ne are plotted in dependence on
the spin used in the preceding VAMPIR calculation for $^{22}$Ne.
For comparison, the leftmost column presents the energies of the
three lowest $3/2$ shell model states. Solid lines denote the results of the
{\sl complex} MONSTER, dotted ones those of the {\sl real} counterpart.
It can be seen that {\sl real} MONSTER displays only a weak dependence on
the spin of the underlying HFB transformations. For the {\sl complex}
MONSTER this is also true if the VAMPIR transformations of the low spin
mean fields, $0^+$, $2^+$ and $4^+$, are considered. The mean fields of the
higher spin states are insufficient for the description of the $3/2$ states.
The poor description of the $3/2$ states using the VAMPIR transformations
of the low odd spins, $1^+$, $3^+$ and $5^+$, is due to the fact that
those yrast states are dominated by structures which are missing in a
{\sl complex} VAMPIR calculation \cite{BEN95a}.
\par
The same behaviour as for the $3/2$ states was also found for most of the
spins of $^{21}$Ne. Only for the very high spins, where the number of MONSTER
configuration is of similar size as the number of shell model states, the
dependence becomes small. When the number of MONSTER configurations is bigger
than the number of shell model states, the resulting energies are independent
of the underlying transformation.
\par
This investigation confirms what we have found already for the even mass case
\cite{BEN95a}. The {\sl complex} HFB transformation shows a much
stronger dependence on the spin of the underlying VAMPIR calculation than the
{\sl real} one. It means that the structure of the {\sl complex} mean field
differs more from spin to spin. The choice of the spin of the underlying
VAMPIR calculation is important in the {\sl complex} description of odd mass
nuclei, too.
\section{Summary}
\label{summary}
A recently developed nuclear structure approach, the {\sl complex} MONSTER,
has been studied. The approach is the newest member of a group of models,
which are based on variational methods. This group can be divided into two
subgroups, one consisting of the VAMPIR models, the other of the MONSTER
approaches. Both are designed for large scale nuclear structure calculations.
\par
The models of the VAMPIR family use symmetry-projected HFB quasiparticle
vacua as test wavefunctions. The underlying HFB transformations are
determined by variation. By construction, these models can only be used to
describe the lowest few states of a certain symmetry. In the first
numerically realized VAMPIR models the HFB transformations were rather
restricted~: time-reversal invariance and axial symmetry were required,
parity and proton-neutron mixing were neglected and only {\sl real} HFB
transformation coefficients were allowed. In the newest models most
restrictions are removed and essentially {\sl complex} HFB transformations
are allowed. Only time reversal and axiality are kept.
\par
In the MONSTER approaches the nuclear wavefunctions are expanded around
a VAMPIR solution for the ground or an yrast state. The spectrum of excited
states is obtained by diagonalizing the Hamiltonian in the space of the
VAMPIR solution and all symmetry-projected two-quasiparticle configurations
with respect to it. These models are therefore suited for problems,
where a complete set of excitations with respect to a particular transition
operator is needed. Till lately the MONSTER approach was restricted
to the use of {\sl real} VAMPIR solutions. In a recent work it was generalized.
Now {\sl complex} HFB transformations can be used to build up the configuration
space. Then more nucleon correlations are considered already in the mean field
and the configuration space becomes much larger. Consequently a more
detailed description of nuclear spectra is expected by this {\sl complex}
MONSTER approach. Furthermore one can use HFB transformations determined
for the particular system under consideration not only for even-even systems,
as in the {\sl real} approach, but also for odd-odd systems. Only for
odd mass nuclei, which are described in the one-quasiparticle approximation,
one still has to use the mean field of a neighbouring nucleus, which now may
be even-even or odd-odd. To get insight into the ability of this {\sl complex}
approach a systematic investigation was performed in this work.
\par
We presented first results of an application of the {\sl complex} MONSTER
approach to several nuclei in $sd$ shell. This basis, which is rather tiny
for a MONSTER calculation, has been chosen to enable a comparison with complete
shell model configuration mixing calculations. The investigated even-even
and odd-odd nuclei have masses between A$=$20 and 30, the odd mass nuclei
reach from A$=$21 to 29. Thus, quite light systems with only a few valence
nucleons are considered as well as heavier systems, up to $^{28}$Si with
the biggest number of shell model configurations in $sd$ shell.
\par
For the even-even nuclei the {\sl complex} MONSTER yielded an excellent
agreement with the exact shell model approach. The largest mean deviation
of the yrast state energies from the shell model result, averaged over
the states with spins $0$ to $5$, was found for $^{26}$Mg and amounts only
to 1.43\% of its shell model ground states energy. Concerning the odd-odd
systems it was demonstrated that {\sl complex} MONSTER reproduced the
shell model spectrum of $^{20}$F exactly. $^{24}$Na showed the largest
average deviation of the energies of the yrast states with spins $0$ to $5$
from the shell model ones, namely 2.21\% of the ground state energy.
The average deviations of the odd-odd systems were slightly bigger than the
ones of the even-even systems with the same mass number. But in general
the {\sl complex} MONSTER yielded a very good approximation to the shell
model result also for odd-odd nuclei. For both even-even and odd-odd nuclei
a clear improvement on the previous {\sl real} approach was proved.
\par
For odd mass nuclei it was found that the description depends much on the
underlying mean field. By choosing a suitable mean field a good agreement
with the shell model could be achieved for all investigated nuclei. As
examples the results of $^{21}$Ne and $^{29}$Si were presented. For $^{21}$Ne,
using the HFB transformation of the $1^+$ VAMPIR solution in $^{22}$Na, the
deviation of the yrast states energies, averaged over spins $1/2$ to $11/2$,
was 0.97\% of its shell model ground state energy. For $^{29}$Si,
with the mean field determined for the ground state of $^{28}$Si, the average
deviation of the yrast states energies for spins $1/2$ to $9/2$ amounted to
0.93\% relative to the ground state energy. Again the {\sl complex}
approach achieved better agreement with the shell model than the
{\sl real} one.
\par
Already the previous {\sl real} MONSTER has been applied successfully to the
description of nuclear structure phenomena in various mass regions, e.g.,~in
the mass 130 region \cite{HAM85} \cite{HAM86}. The present investigation has
clearly shown that the new {\sl complex} approach is even superior. So we
expect that the {\sl complex} MONSTER approach will develop as a more
powerful tool for nuclear structure studies in large model spaces.
\par
Up to now only the energies calculated by the different models have been
compared. It will be interesting to look also for other observables,
like transition strengths up to high excitation energies, and to analyze
the wavefunctions. This is planned for the future. Finally we would like to
mention that the description of nuclear wavefunctions by VAMPIR and
MONSTER could be improved by removing all restrictions imposed on the
HFB transformation. The development of such approaches allowing the most
general HFB transformations is in progress, too.
\\[2.5\parskip]
We thank Prof.~Dr.~Herbert M\"uther for performing the shell model
calculations with the Glasgow code. Furthermore this work was partly
supported by the Gra\-du\-ier\-ten\-kol\-leg ``Struktur und Wechselwirkung
von Hadronen und Kernen'' (DFG, Mu 705/3).
\newpage
|
\section{Introduction}
The aim of this article is to revise and update the profile of the
Cabibbo-Kobayashi-Maskawa (CKM) matrix reported earlier by us \cite{AL94},
in particular the CKM unitarity triangle and the CP asymmetries in $B$
decays, which are the principal objects of interest in experiments at
present and forthcoming $B$ facilities. In performing this update, we
include the improvements reported in a number of measurements of the
lifetime, mixing ratio, and the CKM matrix elements $\vert V_{cb}\vert$ and $\vert
V_{ub}/V_{cb} \vert$ from $B$ decays, as well as the top quark mass. On the
theoretical side, we mention the improved estimates of the power
corrections in the analysis of the exclusive semileptonic decay $B \to D^*
\ell \nu_\ell$ in the context of the heavy quark effective theory (HQET)
\cite{neubert95,shifman95}, and the calculation of the missing part of the
next-to-leading order calculations in the analysis of the CP-violating
quantity $\vert\epsilon\vert$ \cite{HN95}. We note here the changes that we have made
in the input to our present analysis compared to that reported by us in
Ref.~\cite{AL94}:
\begin{itemize}
\item
The top quark (pole) mass $m_t =174 \pm 16$ GeV, measured earlier by the
CDF collaboration \cite{CDFmtold}, is now replaced by improved measurements
by the same collaboration \cite{CDFmtnew} and by D0 \cite{D0mtnew},
yielding the present world average $m_t =180 \pm 11 $ GeV
\cite{Belletini95}. This leads to the running top quark mass in the
$\overline{MS}$ scheme, $\overline{m_t} (m_t)= 170 \pm 11 $ GeV
\cite{mtmsbar}.
\item
A new and improved measurement of the quantity ${\cal F}(1) \vert V_{cb}
\vert$ in the decays $B \to D^* \ell \nu_\ell$, using methods based on the
heavy quark effective theory (HQET), has been reported by the ALEPH
collaboration \cite{ALEPHVcb95}. This is lower than their previous number
\cite{ALEPHVcb94}, as well as the corresponding numbers from the CLEO
\cite{CLEOVcb94} and ARGUS \cite{ARGUSVcb94} analyses. Likewise, new
measurements are reported by the DELPHI collaboration \cite{DELPHIVcb95}.
In the meantime, estimates for the quantity ${\cal F}(1) \equiv \xi(1)
\eta_A$ have undergone some revision in both the QCD perturbative part
$\eta_A$ and power corrections to the Isgur-Wise function at the symmetry
point $\xi(1)$. We use the value ${\cal F}(1) = 0.91\pm 0.04$, obtained
recently by Neubert \cite{neubert95}, and which is in good agreement with
the estimates of Shifman et al. \cite{shifman95,suv94}. Taking into account
the updated experimental and theoretical input, we obtain $\vert V_{cb}
\vert =0.0388 \pm 0.0036$. The central value for this matrix element has
come down compared to the value $\vert V_{cb} \vert =0.041 \pm 0.006$ used
by us previously, and the error on this quantity is now smaller, about $\pm
9\%$.
\item
Until recently, the knowledge of the CKM matrix element ratio $\vert
V_{ub}/V_{cb} \vert$ was based on the analysis of the end-point lepton
energy spectrum in semileptonic $B$ decays \cite{Patterson}, which is quite
model-dependent. We had used a value $\vert V_{ub}/V_{cb} \vert = 0.08 \pm
0.03$ to take this model dependence into account. In the meantime,
the measurement of the exclusive semileptonic decays $B \to (\pi, \rho) \ell
\nu_\ell$ has been reported by the CLEO collaboration \cite{Thorn95}.
The matrix element ratio so determined is also model-dependent due to the
decay form factors. However, this set of data permits a discrimination
among a number of models, all of which were previously allowed from the
inclusive decay analysis. The convolution of the two methods reduces the
theoretical uncertainty somewhat. We use a value $\vert V_{ub}/V_{cb} \vert
= 0.08 \pm 0.02$, which is a fair reflection of the underlying present
theoretical dispersion on this ratio.
\item
In the analysis of the CP-violating quantity $\vert\epsilon\vert$, the perturbative
renormalizations of the various pieces in the $\vert \Delta S \vert =2$
Hamiltonian from the intermediate charm and top quark are required
\cite{Burastop}. While the next-to-leading order results for the quantities
$\hat{\eta}_{cc}$ and $\hat{\eta}_{tt}$ have been known for some time and
were used in our previous analysis, the next-to-leading order calculation
of the quantity $\hat{\eta}_{ct}$ has been completed only recently
\cite{HN95}. We use the improved calculation of $\hat{\eta}_{ct}$ in
performing the CKM fits presented here.
\item
The measurements of the $B_d^0$-$\overline{B_d^0}$ mass difference $\Delta M_d$ have
become quite precise, with the present world average being $\Delta M_d
=0.465 \pm 0.024~(ps)^{-1}$ \cite{Wells}. The present lower limit on the
$B_s^0$-$\overline{B_s^0}$ mass difference has improved slightly, $\Delta M_s >
6.1~(ps)^{-1}$ at 95 \% C.L., assuming for the probability of the
fragmentation of a $b$ quark into a $B_s$ meson a value $f_s=12\%$
\cite{ALEPHxs}, yielding $\Delta M_s/\Delta M_d
> 12.3$ at $95 \%$ C.L. \cite{Wells}.
\end{itemize}
All of these improvements warrant an updated fit of the CKM parameters.
As in our previous analysis, we consider two types of fits. In Fit 1, we
assume particular fixed values for the theoretical hadronic quantities. The
allowed ranges for the CKM parameters are derived from the (Gaussian)
errors on experimental measurements only. In Fit 2, we assign a central
value plus an error (treated as Gaussian) to the theoretical quantities. In
the resulting fits, we combine the experimental and theoretical errors in
quadrature. For both fits we calculate the allowed region in CKM parameter
space at 95\% C.L. We also estimate the SM prediction for the $B_s^0$-${\overline{B_s^0}}$\
mixing parameter, $x_s$, and show how the ALEPH limit of $\Delta M_s /\Delta M_d >
12.3$ (95\% C.L.) constrains the CKM parameter space. We give the present
$95\%$ C.L. upper and lower bounds on the matrix element ratio $\vert
V_{td}/V_{ts} \vert$, as well as the allowed (correlated) values of the CKM
matrix elements $|V_{td}|$ and $|V_{ub}|$.
We also present the corresponding allowed ranges for the CP-violating
phases that will be measured in $B$ decays, characterized by $\sin 2\beta$,
$\sin 2\alpha$ and $\sin^2\gamma$. These can be measured directly through
rate asymmetries in the decays $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}} \to J/\psi K_S$, $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}} \to \pi^+
\pi^-$, and $\hbox{$B_s$\kern-1.4em\raise1.4ex\hbox{\barp}}\ \to D_s^\pm K^\mp$ (or $B^\pm \to \hbox{$D$\kern-1.1em\raise1.4ex\hbox{\barp}}\ K^\pm$),
respectively. We also give the allowed domains for two of the angles,
$(\sin 2\alpha,\sin 2\beta)$. Finally, we briefly discuss the role of
penguins (strong and electroweak) in extracting the CP phases $\alpha$ and
$\gamma$ from the measurements of various CP-asymmetries.
This paper is organized as follows. In Section 2, we present our update of
the CKM matrix, concentrating especially on the matrix element $\vert V_{cb}\vert$
which, thanks to the progress in HQET and experiments, is now well under
control. The constraints that follow from $\vert V_{ub}/V_{cb} \vert$,
$\vert\epsilon\vert$ and $\Delta M_d$ on the CKM parameters are also discussed here.
Section 3 contains the results of our fits. These results are summarized in
terms of the allowed domains of the unitarity triangle, which are displayed
in several figures and tables. In Section 4, we discuss the impact of the
recent lower limit on the ratio $\Delta M_s/\Delta M_d$ reported by the
ALEPH collaboration on the CKM parameters and estimate the expected range
of the mixing ratio $x_s$ in the SM based on our fits. Here we also present
the allowed 95\% C.L. range for $\vert V_{td}/V_{ts} \vert$. In Section 5
we discuss the predictions for the CP asymmetries in the neutral $B$ meson
sector and calculate the correlations for the CP violating asymmetries
proportional to $\sin 2\alpha$, $\sin 2 \beta$ and $\sin^2 \gamma$. We also
review some of the possible theoretical uncertainties in extracting these
CP-phases due to the presence of strong and/or electroweak penguins. We
present here the allowed values of the CKM matrix elements $|V_{td}|$ and
$|V_{ub}|$. Section 6 contains a summary and an outlook for improving the
profile of the CKM unitarity triangle.
\section{An Update of the CKM Matrix}
In updating the CKM matrix elements, we make use of the Wolfenstein
parametrization \cite{Wolfenstein}, which follows from the observation that
the elements of this matrix exhibit a hierarchy in terms of $\lambda$, the
Cabibbo angle. In this parametrization the CKM matrix can be written
approximately as
\begin{equation}
V_{CKM} \simeq \left(\matrix{
1-{1\over 2}\lambda^2 & \lambda
& A\lambda^3 \left( \rho - i\eta \right) \cr
-\lambda ( 1 + i A^2 \lambda^4 \eta )
& 1-{1\over 2}\lambda^2 & A\lambda^2 \cr
A\lambda^3\left(1 - \rho - i \eta\right) & -A\lambda^2 & 1 \cr}\right)~.
\label{CKM}
\end{equation}
In this section we shall discuss those quantities which constrain these CKM
parameters, pointing out the significant changes in the determination of
$\lambda$, $A$, $\rho$ and $\eta$.
We recall that $\vert V_{us}\vert$ has been extracted with good accuracy
from $K\to\pi e\nu$ and hyperon decays \cite{PDG} to be
\begin{equation}
\vert V_{us}\vert=\lambda=0.2205\pm 0.0018~.
\end{equation}
This agrees quite well with the determination of $V_{ud}\simeq 1-{1\over
2}\lambda^2$ from $\beta$-decay,
\begin{equation}
\vert V_{ud}\vert=0.9744\pm 0.0010~.
\end{equation}
The parameter $A$ is related to the CKM matrix element $V_{cb}$, which can
be obtained from semileptonic decays of $B$ mesons. We shall restrict
ourselves to the methods based on HQET to calculate the exclusive and
inclusive semileptonic decay rates. In the heavy quark limit it has been
observed that all hadronic form factors in the semileptonic decays $B \to
(D,D^*) \ell \nu_\ell$ can be expressed in terms of a single function, the
Isgur-Wise function \cite{Wisgur}. It has been shown that the HQET-based
method works best for $B\to D^*l\nu$ decays, since these are unaffected by
$1/m_Q$ corrections \cite{Luke,Boyd,Neubert}. This method has been used by
the ALEPH, ARGUS, CLEO and DELPHI collaborations to determine $\xi (1)
\vert V_{cb}\vert$ and the slope of the Isgur-Wise function.
Using HQET, the differential decay rate in $B \to D^* \ell \nu_\ell$ is
\begin{eqnarray}
\frac{d\Gamma (B \to D^* \ell \bar{\nu})}{d\omega }
&=& \frac{G_F^2}{48 \pi^3} (m_B-m_{D^*})^2 m_{D^*}^3 \eta_{A}^2
\sqrt{\omega^2-1} (\omega + 1)^2 \\ \nonumber
&~& ~~~~~~~~~~~~~~\times [ 1+ \frac{4 \omega}{\omega + 1}
\frac{1-2\omega r + r^2}{(1-r)^2}] \vert V_{cb}\vert ^2 \xi^2(\omega) ~,
\label{bdstara1}
\end{eqnarray}
where $r=m_{D^*}/m_B$, $\omega=v\cdot v'$ ($v$ and $v'$ are the
four-velocities of the $B$ and $D^*$ meson, respectively), and $\eta_{A}$
is the short-distance correction to the axial vector form factor. In the
leading logarithmic approximation, this was calculated by Shifman and
Voloshin some time ago -- the so-called hybrid anomalous dimension
\cite{hybrid}. In the absence of any power corrections, $\xi (\omega=1)=1$.
The size of the $O(1/m_b^2)$ and $O(1/m_c^2)$ corrections to the Isgur-Wise
function, $\xi (\omega)$, and partial next-to-leading order
corrections to $\eta_A$ have received a great deal of theoretical
attention, and the state of the art has been
summarized recently by Neubert \cite{neubert95} and Shifman
\cite{shifman95}. Following them, we take:
\begin{eqnarray}
\label{neubertxiold}
\xi (1) &=& 1+ \delta (1/m^2)= 0.945 \pm 0.025 ~, \nonumber \\
\eta_{A} &=& 0.965 \pm 0.020 ~.
\end{eqnarray}
This gives the range \cite{neubert95}:
\begin{equation}
{\cal F}(1)=0.91 \pm 0.04~.
\label{alxi}
\end{equation}
The present experimental input from the exclusive
semileptonic channels is based on the data by
CLEO \cite{CLEOVcb94}, ALEPH
\cite{ALEPHVcb95}, ARGUS \cite{ARGUSVcb94}, and DELPHI \cite{DELPHIVcb95}:
\begin{eqnarray}
\label{Vcbf195}
\vert V_{cb}\vert \cdot {\cal F}(1)
&=& 0.0351 \pm 0.0019 \pm 0.0020 ~~~[\mbox{CLEO}], \nonumber \\
&=& 0.0314 \pm 0.0023 \pm 0.0025 ~~~[\mbox{ALEPH}], \nonumber \\
&=& 0.0388 \pm 0.0043 \pm 0.0025 ~~~[\mbox{ARGUS}], \nonumber \\
&=& 0.0374 \pm 0.0021 \pm 0.0034 ~~~[\mbox{DELPHI}],
\end{eqnarray}
where the first error is statistical and the second systematic.
The ARGUS number has been updated by taking
into
account the updated lifetimes for the $B^0$ and $B^\pm$ mesons
\cite{komamiya95}.
The statistically weighted average of these numbers is:
\begin{equation}
\vert V_{cb}\vert \cdot {\cal F}(1) = 0.0353 \pm 0.0018 ~,
\end{equation}
which, using ${\cal F}(1)$ from Eq.~(\ref{alxi}), gives the following
value:
\begin{equation}
\vert V_{cb} \vert= 0.0388 \pm 0.0019 ~(expt) \pm 0.0017 ~(th).
\label{Vcbhqet95}
\end{equation}
Combining the errors linearly gives $\vert V_{cb} \vert = 0.0388 \pm
0.0036$. This is in good agreement with the value $\vert V_{cb}\vert = 0.037
^{+0.003}_{-0.002}$ obtained from the exclusive decay $B \to D^* \ell
\nu_\ell$, using a dispersion relation approach \cite{BGL95}.
Likewise, the value of $\vert V_{cb}
\vert$ obtained from the inclusive semileptonic $B$
decays using HQET is quite compatible
with the above determination \cite{Patterson}:
\begin{equation}
\vert V_{cb} \vert= 0.040 \pm 0.0010 ~(expt) \pm 0.005 ~(th) ~.
\end{equation}
In the fits below we shall use
$ \vert V_{cb} \vert = 0.0388 \pm 0.0036$, yielding
\begin{equation}
A = 0.80 \pm 0.075~.
\label{Avalue}
\end{equation}
The other two CKM parameters $\rho$ and $\eta$ are constrained by the
measurements of $\vert V_{ub}/V_{cb}\vert$, $\vert\epsilon\vert$ (the CP-violating
parameter in the kaon system), $x_d$ ($B_d^0$-${\overline{B_d^0}}$\ mixing) and (in principle)
$\epsilon^\prime/\epsilon$ ($\Delta S=1$ CP-violation in the kaon system).
We shall not discuss the constraints from $\epsilon^\prime/\epsilon$, due
to the various experimental and theoretical uncertainties surrounding it at
present, but take up the rest in turn and present fits in which the allowed
region of $\rho$ and $\eta$ is shown.
Up to now, $\vert V_{ub}/V_{cb}\vert$ was obtained by looking at the
endpoint of the inclusive lepton spectrum in semileptonic $B$ decays.
Unfortunately, there still exists quite a bit of model dependence in the
interpretation of the inclusive data by themselves. As mentioned earlier, a
recent new input to this quantity is provided by the measurements of the
exclusive semileptonic decays $B \to (\pi, \rho) \ell \nu_\ell$. The ratios
of the exclusive semileptonic branching ratios provide some discrimination
among the various models \cite{Thorn95}. In particular, models such as that
of Isgur et al.~\cite{ISGW}, which give values in excess of 3 for the
ratio of the decay widths $\Gamma (B^0 \to \rho^-\ell^+ \nu)/\Gamma (B^0
\to \pi^- \ell^+ \nu)$, are disfavoured by the CLEO data. The disfavoured
models are also those which introduce a larger theoretical dispersion in
the interpretation of the inclusive $B \to X_u \ell \nu_\ell$ and exclusive
decay data in terms of the ratio $\vert V_{ub}/V_{cb}\vert$. Excluding them
from further consideration, measurements in both the inclusive and
exclusive modes are compatible with
\begin{equation}
\left\vert \frac{V_{ub}}{V_{cb}} \right\vert = 0.08\pm 0.02~.
\label{vubvcbn}
\end{equation}
This gives
\begin{equation}
\sqrt{\rho^2 + \eta^2} = 0.36 \pm 0.08~.
\end{equation}
With the measurements of the form factors in semileptonic decays $B \to
(\pi,\rho,\omega) \ell \nu_\ell$, one should be able to further constrain
the models, thereby reducing the present theoretical uncertainty on this
quantity.
The experimental value of $\vert\epsilon\vert$ is \cite{PDG}
\begin{equation}
\vert\epsilon\vert = (2.26\pm 0.02)\times 10^{-3}~.
\end{equation}
Theoretically, $\vert\epsilon\vert$ is essentially proportional to the imaginary part
of the box diagram for $K^0$-${\overline{K^0}}$\ mixing and is given by \cite{Burasetal}
\begin{eqnarray}
\vert\epsilon\vert &=& \frac{G_F^2f_K^2M_KM_W^2}{6\sqrt{2}\pi^2\Delta M_K}
\hat{B}_K\left(A^2\lambda^6\eta\right)
\bigl(y_c\left\{\hat{\eta}_{ct}f_3(y_c,y_t)-\hat{\eta}_{cc}\right\}
\nonumber \\
&~& ~~~~~~~~~~~~~~+ ~\hat{\eta}_{tt}y_tf_2(y_t)A^2\lambda^4(1-\rho)\bigr),
\label{eps}
\end{eqnarray}
where $y_i\equiv m_i^2/M_W^2$, and the functions $f_2$ and $f_3$ can be
found in Ref.~\cite{AL94}. Here, the $\hat{\eta}_i$ are QCD correction
factors, of which $\hat{\eta}_{cc}$ \cite{HN94} and $\hat{\eta}_{tt}$
\cite{etaB} were calculated some time ago to next-to-leading order, and
$\hat{\eta}_{ct}$ was known only to leading order \cite{Burastop,Flynn}.
Recently, this last renormalization constant was also calculated to
next-to-leading order \cite{HN95}. We use the following values for the
renormalization-scale-invariant coefficients: $\hat{\eta}_{cc}\simeq 1.32
$, $\hat{\eta}_{tt}\simeq 0.57$, $\hat{\eta}_{ct}\simeq 0.47 $, calculated
for $\hat{m}_c= 1.3$ GeV and the NLO QCD parameter
$\Lambda_{\overline{MS}}=310$ MeV in Ref.~\cite{HN95}.
The final parameter in the expression for $\vert\epsilon\vert$ is the
renormalization-scale independent parameter $\hat{B}_K$, which represents
our ignorance of the hadronic matrix element $\langle K^0 \vert
{({\overline{d}}\gamma^\mu (1-\gamma_5)s)}^2 \vert
{\overline{K^0}}\rangle$. The evaluation of this matrix element has been
the subject of much work. The earlier results are summarized in
Ref.~\cite{AL92}.
In our first set of fits, we consider specific values in the range 0.4 to
1.0 for $\hat{B}_K$. As we shall see, for $\hat{B}_K = 0.4$ a very poor fit
to the data is obtained, so that such small values are quite disfavoured.
In Fit 2, we assign a central value plus an error to $\hat{B}_K$. As in our
previous analysis \cite{AL94}, we consider two ranges for $\hat{B}_K$:
\begin{equation}
\hat{B}_K = 0.8 \pm 0.2 ~,
\label{BKrange1}
\end{equation}
which reflects the estimates of this quantity in lattice QCD
\cite{Shigemitsu,bklattice}, or
\begin{equation}
\hat{B}_K = 0.6 \pm 0.2 ~,
\label{BKrange2}
\end{equation}
which overlaps with the values suggested by chiral perturbation theory
\cite{Pich94}. As we will see, there is not an enormous difference in the
results for the two ranges.
We now turn to $B_d^0$-${\overline{B_d^0}}$\ mixing. The present world average of $x_d\equiv
\Delta M_d/\Gamma_d$, which is a measure of this mixing, is \cite{Wells}
\begin{equation}
x_d = 0.73 \pm 0.04~,
\label{xdvalue}
\end{equation}
which is based on time-integrated measurements which directly measure
$x_d$, and on time-dependent measurements which measure the mass difference
$\Delta M_d$ directly. This is then converted to $x_d$ using the $B_d^0$
lifetime, which is known very precisely $(\tau(B_d)=1.57 \pm 0.05~ps$).
{}From a theoretical point of view it is better to use the mass difference
$\Delta M_d$, as it liberates one from the errors on the lifetime
measurement. In fact, the present precision on $\Delta M_d$, pioneered by
time-dependent techniques at LEP, is quite competitive with the precision
on $x_d$. The LEP average for $\Delta M_d$ has been combined with that
derived from time-integrated measurements yielding the present world
average \cite{Wells}
\begin{equation}
\Delta M_d = 0.465 \pm 0.024~(ps)^{-1} ~.
\label{deltamd}
\end{equation}
We shall use this number instead of $x_d$, which has been the usual
practice to date \cite{Burastop,AL92,Pich94,BLO94}.
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&\cr
& Parameter && Value & \cr
height2pt&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&\cr
& $\lambda$ && $0.2205$ & \cr
& $\vert V_{cb} \vert $ && $0.0388 \pm 0.0036$ & \cr
& $\vert V_{ub} / V_{cb} \vert$ && $0.08 \pm 0.02$ & \cr
& $\vert\epsilon\vert$ && $(2.26 \pm 0.02) \times 10^{-3}$ & \cr
& $\Delta M_d$ && $(0.465 \pm 0.024)~(ps)^{-1}$ & \cr
& $\tau(B_d)$ && $(1.57 \pm 0.05)~(ps)$ & \cr
& $\overline{m_t}(m_t(pole))$ && $(170 \pm 11)$ GeV & \cr
& $\hat{\eta}_B$ && $0.55$ & \cr
& $\hat{\eta}_{cc} $ && $1.32$ & \cr
& $\hat{\eta}_{ct} $ && $0.47$ & \cr
& $\hat{\eta}_{tt} $ && $0.57$ & \cr
& $\hat{B}_K$ && $0.8 \pm 0.2$ & \cr
& $\hat{B}_B$ && $1.0 \pm 0.2$ & \cr
& $f_{B_d}$ && $180 \pm 50$ MeV & \cr
height2pt&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{Parameters used in the CKM fits. Values of the hadronic quantities
$f_{B_d}$, $\hat{B}_{B_d}$ and $\hat{B}_K$ shown are motivated by the lattice
QCD results. In Fit 1, specific values of these hadronic quantities are
chosen, while in Fit 2, they are allowed to vary over the given ranges. (In
Fit 2, for comparison we also consider the range $\hat{B}_K = 0.6 \pm 0.2$,
which is motivated by chiral perturbation theory and QCD sum rules.)}
\label{tabfit}
\end{table}
The mass difference $\Delta M_d$ is calculated from the $B_d^0$-${\overline{B_d^0}}$\ box
diagram. Unlike the kaon system, where the contributions of both the $c$-
and the $t$-quarks in the loop were important, this diagram is dominated by
$t$-quark exchange:
\begin{equation}
\label{bdmixing}
\Delta M_d = \frac{G_F^2}{6\pi^2}M_W^2M_B\left(f^2_{B_d}\hat{B}_{B_d}\right)\hat{\eta}_B y_t
f_2(y_t) \vert V_{td}^*V_{tb}\vert^2~, \label{xd}
\end{equation}
where, using Eq.~\ref{CKM}, $\vert V_{td}^*V_{tb}\vert^2=
A^2\lambda^{6}\left[\left(1-\rho\right)^2+\eta^2\right]$. Here,
$\hat{\eta}_B$ is the QCD correction. In Ref.~\cite{etaB}, this correction
is analyzed including the effects of a heavy $t$-quark. It is found that
$\hat{\eta}_B$ depends sensitively on the definition of the $t$-quark mass,
and that, strictly speaking, only the product $\hat{\eta}_B(y_t)f_2(y_t)$
is free of this dependence. In the fits presented here we use the value
$\hat{\eta}_B=0.55$, calculated in the $\overline{MS}$ scheme, following
Ref.~\cite{etaB}. Consistency requires that the top quark mass be rescaled
from its pole (mass) value of $m_t =180 \pm 11$ GeV to the value
$\overline{m_t}(m_t(pole))$ in the $\overline{MS}$ scheme, which is
typically about 10 GeV smaller \cite{mtmsbar}.
For the $B$ system, the hadronic uncertainty is given by $f^2_{B_d}\hat{B}_{B_d}$, analogous
to $\hat{B}_K$ in the kaon system, except that in this case, also $f_{B_d}$ is
not measured. In our fits, we will take ranges for $f^2_{B_d}\hat{B}_{B_d}$ and
$\hat{B}_{B_d}$ which are compatible with results from both lattice-QCD and
QCD sum rules \cite{Shigemitsu,Narison,Sommer94}:
\begin{eqnarray}
f_{B_d} &=& 180 \pm 50 ~\mbox{MeV}~, \nonumber \\
\hat{B}_{B_d} &=& 1.0 \pm 0.2 ~.
\label{FBrange}
\end{eqnarray}
In Table \ref{tabfit}, we summarize all input quantities to our fits, of
which seven quantities ($\vert V_{cb}\vert$, $\vert V_{ub}/V_{cb} \vert$, $\Delta M_d$,
$\tau(B_d)$, ${\overline{m_t}}$, $\hat{\eta}_{cc}$, $\hat{\eta}_{ct}$)
have changed compared to
their values used in our previous fit \cite{AL94}.
\section{The Unitarity Triangle}
The allowed region in $\rho$-$\eta$ space can be displayed quite elegantly
using the so-called unitarity triangle. The unitarity of the CKM matrix
leads to the following relation:
\begin{equation}
V_{ud} V_{ub}^* + V_{cd} V_{cb}^* + V_{td} V_{tb}^* = 0~.
\end{equation}
Using the form of the CKM matrix in Eq.~\ref{CKM}, this can be recast as
\begin{equation}
\frac{V_{ub}^*}{\lambda V_{cb}} + \frac{V_{td}}{\lambda V_{cb}} = 1~,
\end{equation}
which is a triangle relation in the complex plane (i.e.\ $\rho$-$\eta$
space), illustrated in Fig.~\ref{triangle}. Thus, allowed values of $\rho$
and $\eta$ translate into allowed shapes of the unitarity triangle.
\begin{figure}
\centerline{\psfig{figure=rhoeta.ps,height=5.0cm,angle=90}}
\caption{The unitarity triangle. The angles $\alpha$, $\beta$ and $\gamma$
can be measured via CP violation in the $B$ system.}
\label{triangle}
\end{figure}
In order to find the allowed unitarity triangles, the computer program
MINUIT is used to fit the CKM parameters $A$, $\rho$ and $\eta$ to the
experimental values of $\vert V_{cb}\vert$, $\vert V_{ub}/V_{cb}\vert$, $\vert\epsilon\vert$ and
$x_d$. Since $\lambda$ is very well measured, we have fixed it to its
central value given above. As discussed in the introduction, we present
here two types of fits:
\begin{itemize}
\item
Fit 1: the ``experimental fit.'' Here, only the experimentally measured
numbers are used as inputs to the fit with Gaussian errors; the coupling
constants $f_{B_d} \sqrt{\hat{B}_{B_d}}$ and $\hat{B}_K$ are given fixed
values.
\item
Fit 2: the ``combined fit.'' Here, both the experimental and theoretical
numbers are used as inputs assuming Gaussian errors for the theoretical
quantities.
\end{itemize}
We first discuss the ``experimental fit" (Fit 1). The goal here is to
restrict the allowed range of the parameters ($\rho,\eta)$ for given values
of the coupling constants $f_{B_d} \sqrt{\hat{B}_{B_d}}$ and $\hat{B}_K$.
For each value of $\hat{B}_K$ and $f_{B_d}\sqrt{\hat{B}_{B_d}}$, the CKM
parameters $A$, $\rho$ and $\eta$ are fit to the experimental numbers given
in Table \ref{tabfit} and the $\chi^2$ is calculated.
First, we fix $\hat{B}_K = 0.8$, and vary $f_{B_d}\sqrt{\hat{B}_{B_d}}$ in
the range 130 MeV to 230 MeV. The fits are presented as an allowed region
in $\rho$-$\eta$ space at 95\% C.L. ($\chi^2 = \chi^2_{min} + 6.0$). The
results are shown in Fig.~\ref{rhoeta1}. As we pass from
Fig.~\ref{rhoeta1}(a) to Fig.~\ref{rhoeta1}(e), the unitarity triangles
represented by these graphs become more and more obtuse. Even more striking
than this, however, is the fact that the range of possibilities for these
triangles is quite large. There are two things to be learned from this.
First, our knowledge of the unitarity triangle is at present rather poor.
This will be seen even more clearly when we present the results of Fit 2.
Second, unless our knowledge of hadronic matrix elements improves
considerably, measurements of $\vert\epsilon\vert$ and $x_d$, no matter how precise,
will not help much in further constraining the unitarity triangle. This is
why measurements of CP-violating rate asymmetries in the $B$ system are so
important \cite{BCPasym,AKL94}. Being largely independent of theoretical
uncertainties, they will allow us to accurately pin down the unitarity
triangle. With this knowledge, we could deduce the correct values of
$\hat{B}_K$ and $f_{B_d}\sqrt{\hat{B}_{B_d}}$, and thus rule out or confirm
different theoretical approaches to calculating these hadronic quantities.
\begin{figure}
\vskip -2.4truein
\centerline{\epsfxsize 7.0 truein \epsfbox {figures.ps}}
\vskip -1.0truein
\caption{Allowed region in $\rho$-$\eta$ space, from a fit to the
experimental values given in Table \protect{\ref{tabfit}}. We have fixed
$\hat{B}_K=0.8$ and vary the coupling constant product
$f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$ as indicated on the figures. The solid
line represents the region with $\chi^2=\chi_{min}^2+6$ corresponding to
the 95\% C.L.\ region. The triangles show the best fit.}
\label{rhoeta1}
\end{figure}
Despite the large allowed region in the $\rho$-$\eta$ plane, certain values
of $\hat{B}_K$ and $f_{B_d}\sqrt{\hat{B}_{B_d}}$ are disfavoured since they
do not provide a good fit to the data. For example, fixing $\hat{B}_K=1.0$,
we can use the fitting program to provide the minimum $\chi^2$ for various
values of $f_{B_d}\sqrt{\hat{B}_{B_d}}$. The results are shown in Table
\ref{tabbk1}, along with the best fit values of $(\rho,\eta)$. Since we
have two variables ($\rho$ and $\eta$), we use $\chi^2_{min}<2.0$ as our
``good fit" criterion, and we see that $f_{B_d} \sqrt{\hat{B}_{B_d}} < 130$
MeV and $f_{B_d} \sqrt{\hat{B}_{B_d}} > 270$ MeV give poor fits to the
existing data. Note also that the $\chi^2$ distribution has two minima, at
around $f_{B_d} \sqrt{\hat{B}_{B_d}} = 150$ and 230 MeV. We do not consider
this terribly significant, since the surrounding values of $f_{B_d}
\sqrt{\hat{B}_{B_d}}$ also yield good fits to the data. The very small
values of $\chi^2_{min}$ depend sensitively on the central values of the
various experimental quantities -- if these values move around a little
bit, the values of $f_{B_d} \sqrt{\hat{B}_{B_d}}$ which give the minimum
$\chi^2$ values will move around as well. In Tables \ref{tabbk8} and
\ref{tabbk6}, we present similar analyses, but for $\hat{B}_K=0.8$ and
$0.6$, respectively. From these tables we see that the lower limit on
$f_{B_d} \sqrt{\hat{B}_{B_d}}$ remains fairly constant, at around 130 MeV,
but the upper limit depends quite strongly on the value of $\hat{B}_K$
chosen. Specifically, for $\hat{B}_K =0.8$ and $0.6$, the maximum allowed
value of $f_{B_d} \sqrt{\hat{B}_{B_d}}$ is about 240 and 210 MeV,
respectively.
In Table \ref{tabbk4}, we present the $\chi^2$ values as a function of
$f_{B_d} \sqrt{\hat{B}_{B_d}}$ for $\hat{B}_K=0.4$, which is not favoured
by lattice calculations or QCD sum rules. What is striking is that, over
the entire range of $f_{B_d} \sqrt{\hat{B}_{B_d}}$, the minimum $\chi^2$ is
always greater than 2. This indicates that the data strongly disfavour
$\hat{B}_K \leq 0.4$ solutions.
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $f_{B_d}\sqrt{\hat{B}_{B_d}}$ (MeV) && $(\rho,\eta)$ && $\chi^2_{min}$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $120$ && $(-0.43,~0.13)$ && $2.89$ & \cr
& $130$ && $(-0.38,~0.16)$ && $1.16$ & \cr
& $140$ && $(-0.34,~0.18)$ && $0.3$ & \cr
& $150$ && $(-0.30,~0.21)$ && $9.2 \times 10^{-3}$ & \cr
& $160$ && $(-0.25,~0.24)$ && $0.07$ & \cr
& $170$ && $(-0.20,~0.27)$ && $0.29$ & \cr
& $180$ && $(-0.14,~0.29)$ && $0.52$ & \cr
& $190$ && $(-0.07,~0.31)$ && $0.64$ & \cr
& $200$ && $(0.0,~0.31)$ && $0.59$ & \cr
& $210$ && $(0.07,~0.31)$ && $0.38$ & \cr
& $220$ && $(0.13,~0.31)$ && $0.13$ & \cr
& $230$ && $(0.19,~0.3)$ && $3.8 \times 10^{-3}$ & \cr
& $240$ && $(0.23,~0.31)$ && $0.07$ & \cr
& $250$ && $(0.27,~0.31)$ && $0.36$ & \cr
& $260$ && $(0.31,~0.31)$ && $0.87$ & \cr
& $270$ && $(0.34,~0.31)$ && $1.59$ & \cr
& $280$ && $(0.38,~0.32)$ && $2.51$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{The ``best values'' of the CKM parameters $(\rho,\eta)$ as a
function of the coupling constant $f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$,
obtained by a minimum $\chi^2$ fit to the experimental data, including the
renormalized value of $m_t=170 \pm 11$ GeV. We fix $\hat{B}_K=1.0$. The
resulting minimum $\chi^2$ values from the MINUIT fits are also given.}
\label{tabbk1}
\end{table}
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $f_{B_d}\sqrt{\hat{B}_{B_d}}$ (MeV) && $(\rho,\eta)$ && $\chi^2_{min}$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $120$ && $(-0.42,~0.16)$ && $3.04$ & \cr
& $130$ && $(-0.37,~0.19)$ && $1.32$ & \cr
& $140$ && $(-0.32,~0.23)$ && $0.43$ & \cr
& $150$ && $(-0.27,~0.26)$ && $0.07$ & \cr
& $160$ && $(-0.22,~0.29)$ && $1.4 \times 10^{-3}$ & \cr
& $170$ && $(-0.15,~0.32)$ && $0.05$ & \cr
& $180$ && $(-0.08,~0.34)$ && $0.09$ & \cr
& $190$ && $(-0.01,~0.35)$ && $0.06$ & \cr
& $200$ && $(0.06,~0.35)$ && $0.01$ & \cr
& $210$ && $(0.13,~0.35)$ && $0.02$ & \cr
& $220$ && $(0.18,~0.35)$ && $0.2$ & \cr
& $230$ && $(0.23,~0.35)$ && $0.61$ & \cr
& $240$ && $(0.28,~0.35)$ && $1.29$ & \cr
& $250$ && $(0.32,~0.35)$ && $2.22$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{The ``best values'' of the CKM parameters $(\rho,\eta)$ as a
function of the coupling constant $f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$,
obtained by a minimum $\chi^2$ fit to the experimental data, including the
renormalized value of $m_t=170 \pm 11$ GeV. We fix $\hat{B}_K=0.8$. The
resulting minimum $\chi^2$ values from the MINUIT fits are also given.}
\label{tabbk8}
\end{table}
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $f_{B_d}\sqrt{\hat{B}_{B_d}}$ (MeV) && $(\rho,\eta)$ && $\chi^2_{min}$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $120$ && $(-0.4,~0.21)$ && $3.39$ & \cr
& $130$ && $(-0.35,~0.25)$ && $1.7$ & \cr
& $140$ && $(-0.29,~0.29)$ && $0.78$ & \cr
& $150$ && $(-0.22,~0.33)$ && $0.35$ & \cr
& $160$ && $(-0.15,~0.36)$ && $0.18$ & \cr
& $170$ && $(-0.07,~0.38)$ && $0.16$ & \cr
& $180$ && $(0.01,~0.39)$ && $0.24$ & \cr
& $190$ && $(0.08,~0.4)$ && $0.48$ & \cr
& $200$ && $(0.15,~0.4)$ && $0.96$ & \cr
& $210$ && $(0.21,~0.4)$ && $1.73$ & \cr
& $220$ && $(0.26,~0.4)$ && $2.85$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{The ``best values'' of the CKM parameters $(\rho,\eta)$ as a
function of the coupling constant $f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$,
obtained by a minimum $\chi^2$ fit to the experimental data, including the
renormalized value of $m_t=170 \pm 11$ GeV. We fix $\hat{B}_K=0.6$. The
resulting minimum $\chi^2$ values from the MINUIT fits are also given.}
\label{tabbk6}
\end{table}
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $f_{B_d}\sqrt{\hat{B}_{B_d}}$ (MeV) && $(\rho,\eta)$ && $\chi^2_{min}$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&\cr
& $130$ && $(-0.28,~0.35)$ && $2.95$ & \cr
& $140$ && $(-0.2,~0.39)$ && $2.22$ & \cr
& $150$ && $(-0.11,~0.43)$ && $2.04$ & \cr
& $160$ && $(-0.02,~0.45)$ && $2.32$ & \cr
& $170$ && $(0.06,~0.46)$ && $3.07$ & \cr
height2pt&\omit&&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{The ``best values'' of the CKM parameters $(\rho,\eta)$ as a
function of the coupling constant $f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$,
obtained by a minimum $\chi^2$ fit to the experimental data, including the
renormalized value of $m_t=170 \pm 11$ GeV. We fix $\hat{B}_K=0.4$. The
resulting minimum $\chi^2$ values from the MINUIT fits are also given.}
\label{tabbk4}
\end{table}
We now discuss the ``combined fit" (Fit 2). Since the coupling constants
are not known and the best we have are estimates given in the ranges in
Eqs.~(\ref{BKrange1}) and (\ref{FBrange}), a reasonable profile of the
unitarity triangle at present can be obtained by letting the coupling
constants vary in these ranges. The resulting CKM triangle region is shown
in Fig.~\ref{rhoeta2}. As is clear from this figure, the allowed region is
rather large at present. The
preferred values obtained from the ``combined fit" are
\begin{equation}
(\rho,\eta) = (-0.07,0.34) ~~~(\mbox{with}~\chi^2 = 6.6\times 10^{-2})~.
\end{equation}
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {totalbk8.ps}}
\vskip -1.4truein
\caption{Allowed region in $\rho$-$\eta$ space, from a simultaneous fit to
both the experimental and theoretical quantities given in Table
\protect{\ref{tabfit}}. The theoretical errors are treated as Gaussian for
this fit. The solid line represents the region with $\chi^2=\chi_{min}^2+6$
corresponding to the 95\% C.L.\ region. The triangle shows the best fit.}
\label{rhoeta2}
\end{figure}
For comparison, we also show the allowed region in the $(\rho,\eta)$ plane
for the case in which $\hat{B}_K = 0.6 \pm 0.2$ [Eq.~(\ref{BKrange2})],
which is more favoured by chiral perturbation theory and QCD sum rules. The
CKM triangle region is shown in Fig.~\ref{rhoeta3}. Clearly, there is not
much difference between this figure and Fig.~\ref{rhoeta2}. The preferred
values obtained from this fit are
\begin{equation}
(\rho,\eta) = (-0.05,0.37) ~~~(\mbox{with}~\chi^2 = 0.1)~.
\end{equation}
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {totalbk6.ps}}
\vskip -1.4truein
\caption{Allowed region in $\rho$-$\eta$ space, from a simultaneous fit to
both the experimental and theoretical quantities given in Table
\protect{\ref{tabfit}}, except that we take $\hat{B}_K = 0.6 \pm 0.2$. The
theoretical errors are treated as Gaussian for this fit. The solid line
represents the region with $\chi^2=\chi_{min}^2+6$ corresponding to the
95\% C.L.\ region. The triangle shows the best fit.}
\label{rhoeta3}
\end{figure}
\section{$x_s$ and the Unitarity Triangle}
Mixing in the $B_s^0$-${\overline{B_s^0}}$\ system is quite similar to that in the $B_d^0$-${\overline{B_d^0}}$\
system. The $B_s^0$-${\overline{B_s^0}}$\ box diagram is again dominated by $t$-quark exchange,
and the mass difference between the mass eigenstates $\Delta M_s$ is given by a
formula analogous to that of Eq.~(\ref{xd}):
\begin{equation}
\Delta M_s = \frac{G_F^2}{6\pi^2}M_W^2M_{B_s}\left(f^2_{B_s}\hat{B}_{B_s}\right)
\hat{\eta}_{B_s} y_t f_2(y_t) \vert V_{ts}^*V_{tb}\vert^2~.
\label{xs}
\end{equation}
Using the fact that $\vert V_{cb}\vert=\vert V_{ts}\vert$ (Eq.~\ref{CKM}),
it is clear that one of the sides of the unitarity triangle, $\vert
V_{td}/\lambda V_{cb}\vert$, can be obtained from the ratio of $\Delta M_d$ and
$\Delta M_s$,
\begin{equation}
\frac{\Delta M_s}{\Delta M_d} =
\frac{\hat{\eta}_{B_s}M_{B_s}\left(f^2_{B_s}\hat{B}_{B_s}\right)}
{\hat{\eta}_{B_d}M_{B_d}\left(f^2_{B_d}\hat{B}_{B_d}\right)}
\left\vert \frac{V_{ts}}{V_{td}} \right\vert^2.
\label{xratio}
\end{equation}
All dependence on the $t$-quark mass drops out, leaving the square of the
ratio of CKM matrix elements, multiplied by a factor which reflects
$SU(3)_{\rm flavour}$ breaking effects. The only real uncertainty in this
factor is the ratio of hadronic matrix elements. Whether or not $x_s$ can
be used to help constrain the unitarity triangle will depend crucially on
the theoretical status of the ratio $f^2_{B_s}\hat{B}_{B_s}/f^2_{B_d}\hat{B}_{B_d}$. In what follows, we will
take $\xi_s \equiv (f_{B_s} \sqrt{\hat{B}_{B_s}}) / (f_{B_d}
\sqrt{\hat{B}_{B_d}}) = (1.16 \pm 0.1)$, consistent with both lattice-QCD
\cite{Shigemitsu} and QCD sum rules \cite{Narison}. (The SU(3)-breaking
factor in $\Delta M_s/\Delta M_d$ is $\xi_s^2$.)
The mass and lifetime of the $B_s$ meson have now been measured at LEP and
Tevatron and their present values are $M_{B_s}=5370.0 \pm 2.0$ MeV and
$\tau(B_s)= 1.58 \pm 0.10 ~ps$ \cite{komamiya95}. We expect
the QCD correction factor $\hat{\eta}_{B_s}$ to be equal to its $B_d$
counterpart, i.e.\ $\hat{\eta}_{B_s} =0.55$. The main uncertainty in $x_s$
(or, equivalently, $\Delta M_s$) is now $f^2_{B_s}\hat{B}_{B_s}$. Using the determination of $A$
given previously, $\tau_{B_s}= 1.58 \pm 0.10~(ps)$ and $\overline{m_t}=171
\pm 11$ GeV, we obtain
\begin{eqnarray}
\Delta M_s &=& \left(13.1 \pm 2.8\right)\frac{f^2_{B_s}\hat{B}_{B_s}}{(230~\mbox{MeV})^2}
{}~(ps)^{-1}~, \nonumber \\
x_s &=& \left(20.7 \pm 4.5\right)\frac{f^2_{B_s}\hat{B}_{B_s}}{(230~\mbox{MeV})^2}~.
\end{eqnarray}
The choice $f_{B_s}\sqrt{\hat{B}_{B_s}}= 230$ MeV corresponds to the
central value given by the lattice-QCD estimates, and with this our fits
give $x_s \simeq 20$ as the preferred value in the SM. Allowing the
coefficient to vary by $\pm 2\sigma$, and
taking the central value for
$f_{B_s}\sqrt{\hat{B}_{B_s}}$, this gives
\begin{eqnarray}
11.7 &\leq & x_s \leq 29.7~, \nonumber\\
7.5 ~(ps)^{-1} &\leq & \Delta M_s \leq 18.7 ~(ps)^{-1}~.
\label{bestxs}
\end{eqnarray}
It is difficult to ascribe a confidence level to this range due to the
dependence on the unknown coupling constant factor. All one can say is that
the standard model predicts large values for $x_s$, most of which are above
the present experimental limit $x_s > 8.8$ (equivalently $\Delta M_s > 6.1
{}~(ps)^{-1}$) \cite{ALEPHxs}.
An alternative estimate of $\Delta M_s$ (or $x_s$) can also be obtained by
using the relation in Eq.~(\ref{xratio}). Two quantities are required.
First, we need the CKM ratio $\vert V_{ts}/V_{td} \vert$. In
Fig.~\ref{vtdts} we show the allowed values (at 95\% C.L.) of the inverse
of this ratio as a function of $f_{B_d}\sqrt{\hat{B}_{B_d}}$, for
$\hat{B}_K=0.8\pm 0.2$. From this one gets
\begin{equation}
2.8 \leq \left\vert {V_{ts} \over V_{td}} \right\vert \leq 7.6~.
\end{equation}
The second ingredient is the SU(3)-breaking factor which we take to be
$\xi_s = 1.16 \pm 0.1$, or $1.1 \le \xi_s^2 \le 1.6$.
The result of the CKM fit can
therefore be expressed as a $95\%$ C.L. range:
\begin{equation}
10.5 \left(\frac{\xi_s}{1.16}\right)^2
~\leq ~\frac{\Delta M_s}{\Delta M_d} ~\leq ~
77.7 \left(\frac{\xi_s}{1.16}\right)^2 ~.
\end{equation}
Again, it is difficult to assign a true confidence level to $\Delta M_s/\Delta M_d$
due to the dependence on $\xi_s$. The large allowed range reflects our poor
knowledge of the matrix element ratio $\vert V_{ts}/V_{td} \vert$, which
shows that this method is not particularly advantageous at present for the
determination of the range for $\Delta M_s$.
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {vtdvts.ps}}
\vskip -1.0truein
\caption{Allowed values of the CKM matrix element ratio $\vert
V_{td}/V_{ts} \vert$ as a function of the coupling constant product
$f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$, for $\hat{B}_K=0.8\pm 0.2$. The solid
line corresponds to the best fit values and the dotted curves correspond to
the maximum and minimum allowed values at 95 \% C.L.}
\label{vtdts}
\end{figure}
The ALEPH lower bound $\Delta M_s > 6.1~(ps)^{-1}$ (95\% C.L.) \cite{ALEPHxs}
and the present world average $\Delta M_d = (0.465 \pm 0.024)~(ps)^{-1}$ can
be used to put a bound on the ratio $\Delta M_s/\Delta M_d$. The lower limit on
$\Delta M_s$ is correlated with the value of $f_s$, the fraction of $b$ quark
fragmenting into $B_s$ meson, as shown in the ALEPH analysis
\cite{ALEPHxs}. The value obtained from the measurement of the quantity
$f_s BR(B_s \to D_s \ell \nu_\ell)$ is $f_s=(12 \pm 3) \%$. The
time-integrated mixing ratios $\bar{\chi}$ and $\chi_d$, assuming maximal
mixing in the $B_s$-$\overline{B_s}$ system $\chi_s =0.5$, give $f_s=(9\pm
2)\%$. The weighted average of these numbers is $f_s=(10 \pm 2)\%$
\cite{Wells}. With $f_s=10 \%$, one gets $\Delta M_s > 5.6~(ps)^{-1}$ at 95\%
C.L., yielding $\Delta M_s/\Delta M_d > 11.3$ at 95\% C.L. Assuming, however,
$f_s=12\%$ gives $\Delta M_s > 6.1~(ps)^{-1}$, yielding $\Delta M_s/\Delta M_d >
12.3$ at 95\% C.L. We will use this latter number.
The 95\% confidence limit on $\Delta M_s/\Delta M_d$ can be turned into a bound on
the CKM parameter space $(\rho,\eta)$ by choosing a value for the
SU(3)-breaking parameter $\xi_s^2$. We assume three representative values:
$\xi_s^2 = 1.1$, $1.35$ and $1.6$, and display the resulting constraints in
Fig.~\ref{xslimit}. From this graph we see that the ALEPH bound marginally
restricts the allowed $\rho$-$\eta$ region for small values of $\xi_s^2$,
but does not provide any useful bounds for larger values.
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {totalbk8xs.ps}}
\vskip -1.4truein
\caption{Further constraints in $\rho$-$\eta$ space from the ALEPH bound
on $\Delta M_s$. The bounds are presented for 3 choices of the SU(3)-breaking
parameter: $\xi_s^2 = 1.1$ (dotted line), $1.35$ (dashed line) and $1.6$
(solid line). In all cases, the region to the left of the curve is ruled
out.}
\label{xslimit}
\end{figure}
Summarizing the discussion on $x_s$, we note that the lattice-QCD-inspired
estimate $f_{B_s} \sqrt{\hat{B}_{B_s}} \simeq 230$ MeV and the CKM fit
predict that $x_s$ lies between 12 and 30, with a central value around 20.
The upper and lower bounds and the central value scale as
$(f_{B_s}\sqrt{\hat{B}_{B_s}}/230 ~\mbox{MeV})^2$. The present constraints
from the lower bound on $x_s$ on the CKM parameters are marginal but this
would change with improved data. In particular, one expects to reach a
sensitivity $x_s \simeq 15$ (or $\Delta M_s \simeq 10~ps^{-1})$ at LEP
combining all data and
tagging techniques \cite{Wells},
which would be in the ball-park estimate for this
quantity in the SM presented here.
Of course, an actual measurement of $x_s$
would be very helpful in further constraining the CKM parameter space.
\section{CP Violation in the $B$ System}
It is expected that the $B$ system will exhibit large CP-violating effects,
characterized by nonzero values of the angles $\alpha$, $\beta$ and
$\gamma$ in the unitarity triangle (Fig.~\ref{triangle}) \cite{BCPasym}.
The most promising method to measure CP violation is to look for an
asymmetry between $\Gamma(B^0\to f)$ and $\Gamma({\overline{B^0}}\to f)$,
where $f$ is a CP eigenstate. If only one weak amplitude contributes to the
decay, the CKM phases can be extracted cleanly (i.e.\ with no hadronic
uncertainties). Thus, $\sin 2\alpha$, $\sin 2\beta$ and $\sin 2\gamma$ can
in principle be measured in $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}} \to \pi^+ \pi^-$, $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}}\to J/\psi
K_S$ and $\hbox{$B_s$\kern-1.4em\raise1.4ex\hbox{\barp}}\to\rho K_S$, respectively.
Unfortunately, the situation is not that simple. In all of the above cases,
in addition to the tree contribution, there is an additional amplitude due
to penguin diagrams \cite{penguins}. In general, this will introduce some
hadronic uncertainty into an otherwise clean measurement of the CKM phases.
In the case of $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}}\to J/\psi K_S$, the penguins do not cause any
problems, since the weak phase of the penguin is the same as that of the
tree contribution. Thus, the CP asymmetry in this decay still measures
$\sin 2\beta$.
For $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}} \to \pi^+ \pi^-$, however, although the penguin is expected to
be small with respect to the tree diagram, it will still introduce a
theoretical uncertainty into the extraction of $\alpha$. Fortunately, this
uncertainty can be removed by the use of isospin \cite{isospin}. The key
observation is that the $I=2$ component of the $B\to\pi\pi$ amplitude is
pure tree (i.e., it has no penguin contribution) and therefore has a
well-defined CKM phase. By measuring the rates for $B^+\to\pi^+\pi^0$,
$B^0\to\pi^+\pi^-$ and $B^0\to\pi^0\pi^0$, as well as their CP-conjugate
counterparts, it is possible to isolate the $I=2$ component and obtain
$\alpha$ with no theoretical uncertainty. Thus, even in the presence of
penguin diagrams, $\sin 2\alpha$ can in principle be extracted from the
decays $B\to\pi\pi$. It must be admitted, however, that this isospin
program is ambitious experimentally. If it cannot be carried out, the error
induced on $\sin 2\alpha$ is of order $|P/T|$, where $P$ ($T$) represents
the penguin (tree) diagram. The ratio $|P/T|$ is difficult to estimate --
it is dominated by hadronic physics. However, one ingredient is the ratio
of the CKM elements of the two contributions: $|V_{tb}^* V_{td} / V_{ub}^*
V_{ud} | \simeq |V_{td}/V_{ub}|$. From our fits, we have determined the
allowed values of $|V_{td}|$ as a function of $|V_{ub}|$. This is shown in
Fig.~\ref{vtdvub} for the ``combined fit". The allowed range for the ratio
of these CKM matrix elements is
\begin{equation}
1.2 \leq \left\vert {V_{td}\over V_{ub}} \right\vert \leq 5.8 ~,
\end{equation}
with the central value close to 3.
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {vtdvub.ps}}
\vskip -1.4truein
\caption{Allowed region of the CKM matrix elements $|V_{td}|$ and
$|V_{ub}|$ resulting from the ``combined fit" of the data for the ranges
for $f_{B_d}\protect\sqrt{\hat{B}_{B_d}} $ and $\hat{B}_K$ given in the text.}
\label{vtdvub}
\end{figure}
It is $\hbox{$B_s$\kern-1.4em\raise1.4ex\hbox{\barp}}\to\rho K_S$ which is most affected by penguins. In fact,
the penguin contribution is probably larger in this process than the tree
contribution. This decay is clearly not dominated by one weak (tree)
amplitude, and thus cannot be used as a clean probe of the angle $\gamma$.
Instead, two other methods have been devised, not involving CP-eigenstate
final states. The CP asymmetry in the decay $\hbox{$B_s$\kern-1.4em\raise1.4ex\hbox{\barp}}\to D_s^\pm K^\mp$ can
be used to extract $\sin^2 \gamma$ \cite{ADK}. Similarly, the CP asymmetry
in $B^\pm\toD^0_{\sss CP} K^\pm$ also measures $\sin^2 \gamma$ \cite{growyler}.
Here, $D^0_{\sss CP}$ is a $D^0$ or ${\overline{D^0}}$ which is identified in a CP-eigenstate
mode (e.g.\ $\pi^+\pi^-$, $K^+K^-$, ...).
These CP-violating asymmetries can be expressed straightforwardly in terms
of the CKM parameters $\rho$ and $\eta$. The 95\% C.L.\ constraints on
$\rho$ and $\eta$ found previously can be used to predict the ranges of
$\sin 2\alpha$, $\sin 2\beta$ and $\sin^2 \gamma$ allowed in the standard
model. The allowed ranges which correspond to each of the figures in
Fig.~\ref{rhoeta1}, obtained from Fit 1, are found in Table \ref{cpasym1}.
In this table we have assumed that the angle $\beta$ is measured in
$\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}}\to J/\Psi K_S$, and have therefore included the extra minus sign
due to the CP of the final state.
\begin{table}
\hfil
\vbox{\offinterlineskip
\halign{&\vrule#&
\strut\quad#\hfil\quad\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&&\omit&\cr
& $f_{B_d}\sqrt{\hat{B}_{B_d}}$ (MeV) && $\sin 2\alpha$ &&
$\sin 2\beta$ && $\sin^2 \gamma$ & \cr
height2pt&\omit&&\omit&&\omit&&\omit&\cr
\noalign{\hrule}
height2pt&\omit&&\omit&&\omit&&\omit&\cr
& $130$ && 0.46 -- 0.88 && 0.21 -- 0.37 && 0.12 -- 0.39 & \cr
& $155$ && 0.75 -- 1.0 && 0.31 -- 0.56 && 0.34 -- 0.92 & \cr
& $180$ && $-$0.59 -- 1.0 && 0.42 -- 0.73 && 0.68 -- 1.0 & \cr
& $205$ && $-$0.96 -- 0.92 && 0.49 -- 0.86 && 0.37 -- 1.0 & \cr
& $230$ && $-$0.98 -- 0.6 && 0.57 -- 0.93 && 0.28 -- 0.97 & \cr
height2pt&\omit&&\omit&&\omit&&\omit&\cr
\noalign{\hrule}}}
\caption{The allowed ranges for the CP asymmetries $\sin 2\alpha$, $\sin
2\beta$ and $\sin^2 \gamma$, corresponding to the constraints on $\rho$ and
$\eta$ shown in Fig.~\protect\ref{rhoeta1}. Values of the coupling constant
$f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$ are stated. We fix $\hat{B}_K=0.8$. The
range for $\sin 2\beta$ includes an additional minus sign due to the CP of
the final state $J/\Psi K_S$.}
\label{cpasym1}
\end{table}
Since the CP asymmetries all depend on $\rho$ and $\eta$, the ranges for
$\sin 2\alpha$, $\sin 2\beta$ and $\sin^2 \gamma$ shown in Table
\ref{cpasym1} are correlated. That is, not all values in the ranges are
allowed simultaneously. We illustrate this in Fig.~\ref{alphabeta1},
corresponding to the ``experimental fit" (Fit 1), by showing the region in
$\sin 2\alpha$-$\sin 2\beta$ space allowed by the data, for various values
of $f_{B_d}\sqrt{\hat{B}_{B_d}}$. Given a value for
$f_{B_d}\sqrt{\hat{B}_{B_d}}$, the CP asymmetries are fairly constrained.
However, since there is still considerable uncertainty in the values of the
coupling constants, a more reliable profile of the CP asymmetries at
present is given by our ``combined fit" (Fit 2), where we convolute the
present theoretical and experimental values in their allowed ranges. The
resulting correlation is shown in Fig.~\ref{alphabeta2}. From this figure
one sees that the smallest value of $\sin 2\beta$ occurs in a small region
of parameter space around $\sin 2\alpha\simeq 0.4$-0.6. Excluding this
small tail, one expects the CP-asymmetry in $\hbox{$B_d$\kern-1.4em\raise1.4ex\hbox{\barp}}\to J/\Psi K_S$ to be
at least 30\%.
\begin{figure}
\vskip -2.4truein
\centerline{\epsfxsize 7.0 truein \epsfbox {figures3.ps}}
\vskip -1.0truein
\caption{Allowed region of the CP asymmetries $\sin 2\alpha$ and $\sin
2\beta$ resulting from the ``experimental fit" of the data for different
values of the coupling constant $f_{B_d}\protect\sqrt{\hat{B}_{B_d}}$
indicated on the figures a) -- e). We fix $\hat{B}_K=0.8$.}
\label{alphabeta1}
\end{figure}
\begin{figure}
\vskip -1.0truein
\centerline{\epsfxsize 3.5 truein \epsfbox {cptotbk8.ps}}
\vskip -1.4truein
\caption{Allowed region of the CP asymmetries $\sin 2\alpha$ and $\sin
2\beta$ resulting from the ``combined fit" of the data for the ranges for
$f_{B_d}\protect\sqrt{\hat{B}_{B_d}} $ and $\hat{B}_K$ given in the text.}
\label{alphabeta2}
\end{figure}
It may be difficult to extract $\gamma$ using the techniques described
above. First, since $\hbox{$B_s$\kern-1.4em\raise1.4ex\hbox{\barp}}\to D_s^\pm K^\mp$ involves the decay of $B_s$
mesons, such measurements must be done at hadron colliders. At present, it
is still debatable whether this will be possible. Second, the method of
using $B^\pm\toD^0_{\sss CP} K^\pm$ to obtain $\gamma$ requires measuring the rate
for $B^+ \to D^0 K^+$. This latter process has an expected branching ratio
of $\rly< O(10^{-6})$, so this too will be hard.
Recently, a new method to measure $\gamma$ was proposed \cite{PRL}. Using a
flavour SU(3) symmetry, along with the neglect of exchange- and
annihilation-type diagrams, it was suggested that $\gamma$ could be found
by measuring rates for the decays $B^+ \to \pi^0 K^+$, $B^+ \to \pi^+ K^0$,
$B^+ \to \pi^+ \pi^0$, and their charge-conjugate processes. The $\pi K$
final states have both $I=1/2$ and $I=3/2$ components. The crucial
ingredient is that the gluon-mediated penguin diagram contributes only to
the $I=1/2$ final state. Thus, a linear combination of the $B^+ \to \pi^0
K^+$ and $B^+ \to \pi^+ K^0$ amplitudes, corresponding to $I = 3/2$ in the
$\pi K$ system, can be related via flavour SU(3) to the purely $I = 2$
amplitude in $B^+ \to \pi^+ \pi^0$, permitting the construction of an
amplitude triangle. The difference in the phase of the $B^+ \to \pi^+
\pi^0$ side and that of the corresponding triangle for $B^-$ decays was
found to be $2 \gamma$. SU(3) breaking can be taken into account by
including a factor $f_K/f_\pi$ in relating $B\to\pi\pi$ decays to the
$B\to\pi K$ decays \cite{GHLR}.
The key assumption is that the penguin is mediated by gluon exchange.
However, there are also electroweak contributions to the penguins
\cite{EWPs}. These electroweak penguins (EWP's) are not constrained to be
isosinglets. Thus, in the presence of EWP's, there is no longer a triangle
relation $B\to \pi K$ and $B\to\pi\pi$ amplitudes \cite{DH}. Indeed,
electroweak penguins can, in principle, even invalidate the isospin
analysis in $B\to \pi\pi$, since the $I=2$ amplitude will include a
contribution from EWP's, and hence will no longer have a well-defined weak
CKM phase. However, theoretical estimates \cite{DH,GHLREWP} show that
electroweak penguins are expected to be relatively unimportant for
$B\to\pi\pi$.
The question of the size of EWP's has therefore become a rather interesting
question, and a number of papers have recently appeared discussing this
issue \cite{EWPsize}. These include both theoretical predictions, as well
as ways of isolating EWP's experimentally. The general consensus is that
EWP's are large enough to invalidate the method of Ref.~\cite{PRL} for
obtaining $\gamma$. However, two new methods making use of the flavour
SU(3) symmetry, and which do not have any problems with electroweak
penguins, have been suggested. Both are rather complicated, making use of
the isospin quadrangle relation among $B\to\pi K$ decays, as well as
$B^+\to\pi^+\pi^0$ plus an additional decay: $B_s \to \eta\pi^0$ in one
case \cite{GHLREWP}, $B^+ \to \eta K^+$ in the other \cite{DH2}. Although
electroweak penguins do not cause problems, SU(3)-breaking effects which
cannot be parametrized simply as a ratio of decay constants are likely to
introduce errors of about 25\% into both methods. It is clear that this is
a subject of great interest at the moment, and work will no doubt continue.
\section{Summary and Outlook}
We summarize our results:
\smallskip
(i) We have presented an update of the CKM unitarity triangle using the
theoretical and experimental improvements in the following quantities:
$\vert V_{cb}\vert$, $\vert V_{ub}/V_{cb} \vert$, $\Delta M_d$, $\tau(B_d)$,
${\overline{m_t}}$,
$\hat{\eta}_{cc}$, $\hat{\eta}_{ct}$. The fits can be used to exclude
extreme values of the pseudoscalar coupling constants, with the range
$130~\mbox{MeV} \leq f_{B_d} \sqrt{\hat{B}_{B_d}} \leq 270~\mbox{MeV}$
still allowed for $\hat{B}_K=1$. The lower limit of this range is quite
$\hat{B}_K$-independent, but the upper limit is strongly correlated with
the value chosen for $\hat{B}_K$. For example, for $\hat{B}_K=0.8$ and
$0.6$, $f_{B_d} \sqrt{\hat{B}_{B_d}} \leq 240$ and 210 MeV, respectively,
is required for a good fit. The solutions for $\hat{B}_K = 0.8 \pm 0.2$
are slightly favoured by the data as compared to the lower values. These
numbers are in very comfortable agreement with QCD-based estimates from sum
rules and lattice techniques. The statistical significance of the fit is,
however, not good enough to determine the coupling constant more precisely.
We note that $\hat{B}_K \leq 0.4$ is strongly disfavoured by the data,
since the quality of fit for such values is very poor.
\smallskip
(ii) The newest experimental and theoretical numbers restrict the allowed
CKM unitarity triangle in the $(\rho,\eta)$-space somewhat more than
before. However, the present uncertainties are still enormous -- despite
the new, more accurate experimental data, our knowledge of the unitarity
triangle is still poor. This underscores the importance of measuring
CP-violating rate asymmetries in the $B$ system. Such asymmetries are
largely independent of theoretical hadronic uncertainties, so that their
measurement will allow us to accurately pin down the parameters of the CKM
matrix. Furthermore, unless our knowledge of the pseudoscalar coupling
constants improves considerably, better measurements of such quantities as
$x_d$ will not help much in constraining the unitarity triangle. On this
point, help may come from the experimental front. It may be possible to
measure the parameter $f_{B_d}$, using isospin symmetry, via the
charged-current decay $B_u^\pm\to\tau^\pm \nu_\tau$. With $\vert V_{ub}/V_{cb}
\vert =0.08 \pm 0.02$ and $f_{B_d}=180\pm 50~{\rm MeV}$, one gets a branching
ratio $BR(B_u^\pm\to\tau^\pm\nu_\tau)=(1.5$--$14.0)\times 10^{-5}$, with
a central value of $5.2\times 10^{-5}$. This
lies in the range of the future LEP and asymmetric $B$-factory experiments,
though at LEP the rate $Z \to B_c X \to \tau^\pm \nu_\tau X$ could be just
as large as $Z \to B^\pm X \to \tau^\pm \nu_\tau X$. Along the same lines,
the prospects for measuring $(f_{B_d},f_{B_s})$ in the FCNC leptonic and photonic
decays of $B_d^0 $ and $B_s^0$ hadrons, $(B_d^0,B_s^0)\to\mu^+\mu^-, (B_d^0,B_s^0) \to
\gamma\gamma$ in future $B$ physics facilities are not entirely dismal
\cite{ALIINT94}.
\smallskip
(iii) We have determined bounds on the ratio $\vert V_{td}/V_{ts} \vert$
from our fits. For $130~\mbox{MeV} \leq f_{B_d} \sqrt{\hat{B}_{B_d}} \leq
270~\mbox{MeV}$, i.e.\ in the entire allowed domain, at 95 \% C.L. we find
\begin{equation}
0.13 \leq \left\vert {V_{td} \over V_{ts}} \right\vert \leq 0.35~.
\end{equation}
The upper bound from our analysis is more restrictive than the current
experimental upper limit following from the CKM-suppressed radiative
penguin decays $BR(B \to \omega + \gamma )$ and $BR(B \to \rho + \gamma )$,
which at present yield at 90\% C.L. \cite{cleotdul}
\begin{equation}
\left\vert {V_{td} \over V_{ts}} \right\vert \leq 0.64 - 0.75~,
\end{equation}
depending on the model used for the SU(3)-breaking in the relevant form
factors \cite{SU3ff}. Long-distance effects in the decay $B^\pm \to
\rho^\pm + \gamma$ may introduce theoretical uncertainties comparable to
those in the SU(3)-breaking part but the corresponding effects in the
decays $B^0 \to (\rho^0,\omega) +\gamma$ are expected to be very small
\cite{AB95}. Furthermore, the upper bound is now as good as that obtained
from unitarity, which gives $0.08 \leq \vert V_{td}/V_{ts} \vert \leq
0.36$, but the lower bound from our fit is more restrictive.
\smallskip
(iv) Using the measured value of $m_t$, we find
\begin{equation}
x_s = \left(20.7 \pm 4.5\right)\frac{f^2_{B_s}\hat{B}_{B_s}}{(230 ~\mbox{MeV})^2}~.
\end{equation}
Taking $f_{B_s}\sqrt{\hat{B}_{B_s}}= 230$ (the central value of lattice-QCD
estimates), and allowing the coefficient to vary by $\pm 2\sigma$, this
gives
\begin{equation}
11.7 \leq x_s \leq 29.7~.
\end{equation}
No reliable confidence level can be assigned to this range -- all that one
can conclude is that the SM predicts large values for $x_s$, most of which
lie above the ALEPH 95\% C.L. lower limit of $x_s > 8.8$.
\smallskip
(v) The ranges for the CP-violating rate asymmetries parametrized by $\sin
2\alpha$, $\sin 2\beta$ and and $\sin^2 \gamma$ are determined at 95\% C.L.
to be
\begin{eqnarray}
&~& -1.0 \leq \sin 2\alpha \le 1.0~, \nonumber \\
&~& 0.21 \leq \sin 2\beta \le 0.93~, \\
&~& 0.12 \leq \sin^2 \gamma \le 1.0~. \nonumber
\end{eqnarray}
(For $\sin 2\alpha < 0.4$, we find $\sin 2\beta \ge 0.3$.) Electroweak
penguins may play a significant role in some methods of extracting
$\gamma$. Their magnitude, relative to the tree contribution, is therefore
of some importance. One factor in determining this relative size is the
ratio of CKM matrix elements $\vert V_{td}/V_{ub} \vert$. We find
\begin{equation}
1.2 \leq \left\vert {V_{td}\over V_{ub}} \right\vert \leq 5.8 ~.
\end{equation}
\bigskip
\noindent
{\bf Acknowledgements}:
\bigskip
We thank Viladimir Braun, Roger Forty, Christoph Greub, Matthias Neubert,
Olivier Schneider, Vivek Sharma, Tomasz Skwarnicki, Ed Thorndike, Pippa
Wells and Sau Lan Wu for very helpful discussions. A.A. thanks the
hospitality of the organizing committee for the 6$^{th}$ symposium on heavy
flavour physics in Pisa.
|
\section{Introduction}
The \6He nucleus is an example of a three-cluster system whose lowest
threshold ($\alpha+n+n$) is a three-particle one. Such systems have a number
of remarkable properties determined mainly by two factors---the Pauli
principle and the character of the potential energy dependence due to the
three-body structure.
The Pauli principle imposes essential restrictions on the wave function of a
system allowing it to contain only the components antisymmetric with respect
to nucleon permutation. Thus, if we are using the expansion over the
harmonic-oscillator basis, we must solve the problem of excluding the states
forbidden by the Pauli principle \cite{ho-basis}. This, in particular, leads
to the fact that the simplest wave function of the $0^+$-state of \6He
obtained with the translation-invariant shell model (it coincides with the
lowest oscillator-basis function) is a superposition of states with
three-particle hypermomentum $K=0$ ($\sim5$\%) and $K=2$ ($\sim95\%$). This
function generates an infinite set of states differing by a number of
oscillator quanta and having almost the same ratio of $K=0$ and $K=2$
components. The approximation using only this set of functions can be
appropriately called the $K_{\min}$-approximation. Other basis states contain
components with $K>2$ whose weight is small, as shown in \cite{k-harm}.
Notice that for $2^+$-states, wave functions of the $K_{\min}$-approximation
contain only $K=2$ components, while for the $1^-$-states they are the
superpositions of $K=1$ (${\gtrsim 25\%}$) and $K=3$ (${\lesssim 75\%}$)
components.
The second characteristic feature of systems under consideration is the slow
decrease of their potential energy for large values of the hyperradius $\rho$.
As known (see for example \cite{rho-cube}), in three-body systems, the
potential energy has the asymptotical behaviour ${\rm const}/\rho^3$ as $\rho
\to \infty$, even if the two-body forces acting between any two of constituent
particles are short-range. This fact leads to two important results. First,
the boundary of a system is extremely loose and the rms radius is
significantly larger compared with that of those neighbouring nuclei whose
three-particle threshold is much higher than the binary one. Second, the
phase shift of the three-body scattering rises sharply at low energies (it is
proportional to $\sqrt{E}$ as $E\to0$) which directly affects the behaviour of
the electromagnetic transitions matrix elements between the ground state of
\6He and its continuous spectrum.
The aim of the present paper is to give a detailed quantitative analysis of
general regularities in \6He structure, based on a simple model taking into
account the main features of a loosely-bound system with the lowest
three-particle threshold.
\section{The Model}\indent
The allowed basis functions for states with zero angular momentum
$\varphi_{n}(L=0)$ of the $K_{\min}$-approximation have the following form
\be
\varphi_n(L=0) = A_0(n)\,\varphi_{n0}(\rho)\,u_0(\omega) +
A_2(n)\,\varphi_{n2}(\rho)\,u_2(\omega),
\end{eqnarray}
where the superposition coefficients are
\be
A_0(n) = \sqrt{4(n+1)\over 29n+104},~~
A_2(n) = \sqrt{25(n+4)\over 29n+104},
\end{eqnarray}
$u_K(\omega)$ are the hyperspherical functions,
\be
\varphi_{nK}(\rho) = N_{nk} \rho^K e^{-\rho^2/2} L^{K+2}_n(\rho^2)
\end{eqnarray}
are the normalized hyperradial harmonic-oscillator basis functions,
$L^{\alpha}_n(\rho^2)$ are the Laguerre polynomials,
$N_{nk}$ are the normalizing coefficients, $n = 0,1, \ldots$
Therefore, the wave function of the $L=0$ state in the
$K_{\min}$-approximation is reduced to the expansion
\be
\Psi(L=0) = \sum_{n=0}^{\infty} C_n \varphi_n(L=0).
\label{3}
\end{eqnarray}
For all $n$, the contribution of $K=2$ states is significantly greater than
that of the $K=0$ states
\be
0.86 < A_2^2(n) < 0.96;~~ A_0^2(n) = 1 - A_2^2(n),
\nonumber
\end{eqnarray}
hence, in a simple model, the contribution of $K=0$ states can be neglected.
Then, for the expansion (\ref{3}) the following relations hold
\be
\Psi(L=0) = \sum_{n=0}^{\infty} C_n \varphi_{n2}(\rho)\,u_2(\rho) =
\Phi(\rho)\, u_2(\Omega) = {\chi(\rho) \over \rho^{5/2}} u_2(\Omega).
\end{eqnarray}
The function
$\chi(\rho)$ should satisfy the one-dimensional Schr\"odinger equation
\be
{\hbar^2\over2m}\left[
-{d^2\chi\over d\rho^2} +
\left(K+{3\over2}\right)\left(K+{5\over2}\right){\chi\over \rho^2}
\right] + V(\rho)\,\chi = E\chi,
\label{5}
\end{eqnarray}
where $m$ is the nucleon mass, $K$ is the hypermomentum ($K=2$ for $L=0$),
$V(\rho)$ is the effective three-body potential.
The equation (\ref{5}) is a starting point for the subsequent discussion.
It was solved by a numerical integration over $\rho$ from zero to a certain
sufficiently large cutoff radius $\rho_{\max}$.
The effective potential is modelled by a function
\be
V(\rho) = {V_0 \over {\displaystyle 1 + \left({\rho\over a}\right)^3}},
\label{6}
\end{eqnarray}
having correct asymptotics and without a singularity at $\rho=0$. To reproduce
the states with zero angular momentum, we have chosen the parameters $V_0$ and
$a$ such that the equation (\ref{5}) would give experimental values for the
\6He binding energy and the rms radius, $E = -{\cal E} = -0.97$ MeV,
$\sqrt{\langle r^2 \rangle} = 2.57$~Fm. The appropriate values of the parameters are
\be
V_0 = -87~ {\rm MeV},~~ a = 3.073~ {\rm Fm}.
\end{eqnarray}
With these values we have calculated the wave function of the \6He ground
state and $0^+$-states of its continuous spectrum.
\section{The Ground State of \6He}
The wave function of the ground state of \6He together with the effective
potential $V(\rho)$ are presented in Fig.~1. The horizontal line below the
$\rho$-axis corresponds to the ground state energy, its intersection with the
potential energy curve marks the classical turning point. The vertical line
separates the values of $\rho$ less than $\sqrt{\langle\rho^2\rangle} =
5.59$~Fm. The wave function falls as $\rho^{K+5/2}$ for small $\rho$,
due to the strong kinematical barrier, while for large $\rho$, the long-range
character of the potential $V(\rho)$ leads to the slow decrease of the wave
function and, respectively, to a significant diffuseness of the
boundary of a nuclear system.
As known, loosely-bound binary systems with a short-range potential can be
rather well approximated outside the potential range by the exponential
function
\be
\chi(\rho) \simeq
\sqrt{2\kappa}\exp(-\kappa r),~~~~\kappa = \sqrt{{2m\over\hbar^2}{\cal E}\,},
\label{8}
\end{eqnarray}
where ${\cal E}$ is the bound state energy, $r$ is the radial variable of the
binary channel. The formal criterion of validity of such an approximation (the
zero-range approximation) is the smallness of a ratio of the potential range
and the cluster radii to the system radius $1/\sqrt{2\kappa}$ expressed in
terms of binding energy. It seems useful to compare the exact wave function
obtained after the solution of Eq.~(\ref{5}) with the approximation given by
Eq.~(\ref{8}) (for $\kappa \approx 0.22$ Fm$^{-1}$, dashed line in Fig.~1). It
can be seen that the nucleon system is much more loose than it is predicted by
the wave function (\ref{8}). The reason for that is not only that the function
$\chi(\rho)$ at small $\rho$ behaves as $\rho^{K+5/2}$ as $\rho\to 0$, but
first of all that the effective potential energy decreases slower than the
exponential, while the formula (\ref{8}) is obtained by supposing that the
potential is short-range. As a result, the estimates of the square of
hyperradius mean value $\sqrt{\langle\rho^2\rangle}$ based on (\ref{8}) are
about three times less than the experimental value.
Of course, for sufficiently large $\rho$ the function $\chi(\rho)$ has the
asymptotics
\be
\chi(\rho) \simeq C\sqrt{\kappa\rho}\,K_4(\kappa\rho)
\simeq C \exp(-\kappa\rho),
\label{10}
\end{eqnarray}
where
$K_4(\rho)$ is the Macdonald function, and the value of the coefficient $C$
(it is related to the so-called nuclear vertex constants, see \cite{vertex}) is
considerably less than that of $\sqrt{2\kappa}$
\be
C \approx 0.12~ {\rm Fm}^{-1/2} < \sqrt{2\kappa}
\approx 0.66~ {\rm Fm}^{-1/2}.
\end{eqnarray}
This asymptotics is denoted by short-dashed line in Fig.~1.
\section{0$^+$-States of the Continuous Spectrum}
Of the special interest are the solutions of Eq.~(\ref{5}) for the continuous
spectrum for relatively low over-threshold energies. In particular, attention
must be paid to the question of the behaviour of the three-to-three
scattering phase $\delta$ as a function of energy in a potential field with
the asymptotical behaviour ${\rm const}/\!\rho^3$ and the powerful kinematic
barrier $(\hbar^2/2m)(63/4\rho^2)$ corresponding to $K=2$. According to the
estimates \cite{LK-63,Landau}, at low energies, the phase shift, as is also
demonstrated by our calculations, is proportional to $k$ (or $\sqrt{E}$)
\be
\delta \simeq Ak + \ldots\:;~~~~ A = 3.95~{\rm Fm}.
\end{eqnarray}
The low-energy values of $\tan\delta$ calculated with different values of the
cutoff radius $\rho_{\max}$ are presented as functions of $k$ in Fig.~2. As
seen, $\rho_{\max}$ should be rather large, at least 1000~Fm for
energies below 10 keV\@. The obtained value of $A$ gives a rather reasonable
prediction for the ground-state energy of \6He
\be
{\cal E}_{\rm approx.} = {\hbar^2\over2m}\,{1\over A^2} \approx 1.3~{\rm MeV},
{}~~~({\rm cf.}~{\cal E}_{\rm exp.} = 0.97~{\rm MeV}).
\end{eqnarray}
The three-body phase shift $\delta$ calculated with different $\rho_{\max}$ as
functions of energy $E$ in an interval up to 5 MeV are presented in Fig.~3. As
seen, for reliable calculations for medium-energy region ($\sim 1$~MeV)
$\rho_{\max}$ may be taken about 50~Fm and only for the lowest energies it
must be increased. The phase shift rises steeply from zero to values
exceeding $90^{\circ}$ for $E\approx2.5~{\rm MeV}$ and then slowly goes down.
When the phase shift is near $90^{\circ}$ the first maximum of the wave
function moves closer to zero so that the matrix element of the isoscalar
transition from the ground state into the continuous spectrum increases (see
Fig.~4). This matrix element reaches its maximal value for $E\approx 1.3$~MeV
which could directly affect the electrodisintegration cross-section of \6He.
\section{2$^+$-States of the Continuous Spectrum}
The $K_{\min}$-approximation satisfying the Pauli principle for the
$2^+$-states of \6He contain only the hyperspherical function with
$K=2$. Therefore, to obtain the wave functions $\chi(\rho)$ of $2^+$-states,
we again turn to the equation (\ref{5}) with $K=2$ but with the other
parameters of the potential $V(\rho)$.
The \6He nucleus has no $2^+$ bound state but it has a $2^+$-resonance at the
energy of 0.822$\pm$0.025 MeV with the width 0.113$\pm$0.020 MeV \cite{Ajz}.
The energy and the width of the resonance can be reproduced with the following
parameters of potential:
\be
V_0 = -92~ {\rm MeV},~~ a = 2.834~{\rm Fm}.
\label{17}
\end{eqnarray}
The wave function $\chi(\rho)$ of the resonance state and the potential energy
with the parameters (\ref{17}) are presented in Fig.~5. For small $\rho$ the
resonance wave function behaves similarly to the wave function of the ground
state (dashed line in Fig.~5) but for larger $\rho$, in the asymptotical
region it, as it should be, oscillates.
The scattering phase shift in the $2^+$-state obtained by solving the equation
(\ref{5}) with the new potential parameters is presented in Fig.~6. At the
energy of about 1 MeV the phase shift has a typical behaviour for a resonance
region.
We have also calculated the matrix element of the operator of the isoscalar
$E2$ transition from the ground state of \6He to the $2^+$-states of its
continuous spectrum. The dependence of this matrix element upon the energy is
presented in Fig.~7. The narrow peak observed for the energy about $0.8$~MeV
again demonstrates the small width of $2^+$ resonance and the large value of
the matrix element.
\section{Conclusion}
The simple three-cluster model based on the phenomenological long-range
potential with the Pauli principle taken into account has allowed us to reveal
a number of regularities both for the weakly-coupled ground state of \6He and
for the states of its continuous spectrum for the relatively low (up to a few
MeV) energies. The large value of the rms radius in the ground
state can be explained by the considerable diffuseness of the wave function
caused by the slowly-decreasing effective potential. The asymptotic region,
where the wave function decreases exponentially begins only at the hyperradius
values greater than 15--20~Fm.
The phase shift of elastic scattering ($3\to 3$) is proportional to
$k$ or $\sqrt{E}$ at low energies which leads to the sharp maximum of the
matrix element of the isoscalar transition from the ground state to the
$0^+$-states of the continuous spectrum. Finally, our calculations predict a
considerable enhancing of the probability of the radiative capture of two
neutrons by the alpha-particle for the energy corresponding to the
$2^+$-resonance of \6He.
\medskip
We thank Dr.\ A. A. Korsheninnikov for stimulating discussions.
|
\section{\@startsection {section}{1}{\z@}{-3.5ex plus -1ex minus
-.2ex}{2.3ex plus .2ex}{\large\bf\centering}}
\def\subsection{\@startsection{subsection}{2}{\z@}{-3.25ex plus%
-1ex minus -.2ex}{1.5ex plus .2ex}{\bf}}
\def\subsubsection{\@startsection{subsubsection}{3}{\z@}{-3.25ex plus%
-1ex minus -.2ex}{1.5ex plus .2ex}{\sl}}
\def\vphantom{\vrule height 3ex depth 0pt}{\vphantom{\vrule height 3ex depth 0pt}}
\def\vphantom{\vrule height 0pt depth 2ex}{\vphantom{\vrule height 0pt depth 2ex}}
\gdef\@publabel{\hfil}
\gdef\@pubdate{\null}
\gdef\@pubnumber{\null}
\gdef\@author{\null}
\gdef\@title{\null}
\gdef\@abstract{\null}
\long\def\pubdate#1{\gdef\@pubdate{#1}}
\long\def\pubnumber#1{\gdef\@pubnumber{#1}}
\long\def\publabel#1{\gdef\@publabel{#1}}
\long\def\author#1{\gdef\@author{#1}}
\long\def\title#1{\gdef\@title{#1}}
\long\def\abstract#1{\gdef\@abstract{#1}}
\def\titlerelax{
\let\maketitle\relax
\let\settitleparameters\relax
\let\consolidatetitle\relax
\let\inittitlepage\relax
\let\finishtitlepage\relax
\let\titlepagecontents\relax
\let\multithanks\relax
\let\titlebaselines\relax
\let\@makepub\relax
\let\@maketitle\relax
\let\@makeauthor\relax
\let\@makeabstract\relax
\let\@maketitlenote\relax
\let\thanks\relax
\let\titlerelax\relax}
\def\gdef\@titlenote{
{\gdef\@titlenote{}
\gdef\@abstract{}
\gdef\@author{}
\gdef\@title{}
\gdef\@pubdate{}\gdef\@pubnumber{}\gdef\@publabel{}
\gdef\@dpublabel{}
}
\def\@makepub{\vbox to \z@{\hbox to \textwidth{\hfill
\@publabel \hfill
\llap{\parbox[t]{0.33\textwidth}{\raggedleft\@pubnumber}}}%
\vss}}
\def\@maketitle{\vskip 60pt \begin{center}
{\LARGE \@title \par}
\end{center}}
\def\@makeauthor{{%
\def\smallskip {\normalsize \rm and\smallskip }{\smallskip {\normalsize \rm and\smallskip }}
\def\medskip {\normalsize \rm and\\}\medskip{\medskip {\normalsize \rm and\\}\medskip}
\long\def\address##1{{\def\smallskip {\normalsize \rm and\smallskip }{\\smallskip {\normalsize \rm and\smallskip }\\}\medskip
{\small \it \\##1\\}
}}
{\centering
\vskip 3em
\large \lineskip .75em
\@author}
\par}}
\def\@makedate{\vskip 1.5em
{\raggedright \small \noindent\@pubdate \par}}
\def\@makeabstract{\vskip 1.5em
{\small
\begin{center}
{\bf ABSTRACT\vspace{-.5em}\vspace{0pt}}
\end{center}
\quotation \@abstract \endquotation}}
\def\maketitle{\titlepage
\let\footnotesize\small \setcounter{page}{0}
\@makepub
\vfil
\@maketitle
\@makeauthor
\vfil
\@makeabstract
\@thanks
\vfil
\@makedate
\if@restonecol\twocolumn \else \eject \fi
\titlerelax \gdef\@titlenote{
\setcounter{footnote}{0}
}
\newcommand{\newcommand}{\newcommand}
\newcommand{\ol}{\overline}
\newcommand{\ul}{\underline}
\newcommand{\ve}{\varepsilon}
\newcommand{\gf}{\bar{g}}
\newcommand{\hf}{\bar{h}}
\newcommand{\Wf}{\ol{W}}
\newcommand{\ga}{g}
\newcommand{\si}{\sigma}
\newcommand{\h}{h^{\vee}}
\newcommand{\kh}{k+\h}
\newcommand{\alphav}{\alpha^{\vee}}
\newcommand{\lambdab}{\bar{\lambda}}
\newcommand{\rhob}{\bar{\rho}}
\newcommand{\mub}{\bar{\mu}}
\newcommand{\xmax}{\ol{x}}
\newcommand{\xmin}{\ul{x}}
\newcommand{\ymax}{\ol{y}}
\newcommand{\ymin}{\ul{y}}
\newcommand{\wmax}{\ol{w}}
\newcommand{\wmin}{\ul{w}}
\newcommand{\zmax}{\ol{z}}
\newcommand{\zmin}{\ul{z}}
\newcommand{\ha}{h}
\newcommand{\na}{n}
\newcommand{\Da}{\Delta}
\newcommand{\Qa}{Q}
\newcommand{\Qf}{\ol{Q}}
\newcommand{\Wa}{W}
\newcommand{\ch}{\mbox{ch}\,}
\newcommand{\W}{{\cal W}}
\newcommand{\la}{\langle}
\newcommand{\ra}{\rangle}
\newtheorem{theorem}{Theorem}
\newtheorem{conjecture}{Conjecture}
\newtheorem{lemma}{Lemma}
\newcommand{\be}{\begin{equation}
\addtolength{\abovedisplayskip}{\extraspaces}
\addtolength{\belowdisplayskip}{\extraspaces}
\addtolength{\abovedisplayshortskip}{\extraspace}
\addtolength{\belowdisplayshortskip}{\extraspace}}
\newcommand{\ee}{\end{equation}}
\newcommand{\bea}{\begin{eqnarray}
\addtolength{\abovedisplayskip}{\extraspaces}
\addtolength{\belowdisplayskip}{\extraspaces}
\addtolength{\abovedisplayshortskip}{\extraspace}
\addtolength{\belowdisplayshortskip}{\extraspace}}
\newcommand{\eea}{\end{eqnarray}}
\newcommand{\beas}{\begin{eqnarray*}
\addtolength{\abovedisplayskip}{\extraspaces}
\addtolength{\belowdisplayskip}{\extraspaces}
\addtolength{\abovedisplayshortskip}{\extraspace}
\addtolength{\belowdisplayshortskip}{\extraspace}}
\newcommand{\eeas}{\end{eqnarray*}}
\addtolength{\baselineskip}{.4mm}
\newcommand{{\it i.e}.\ }{{\it i.e}.\ }
\newcommand{{\it e.g}.\ }{{\it e.g}.\ }
\setlength{\parskip}{3mm}
\setlength{\parindent}{0mm}
\newlength{\extraspace}
\setlength{\extraspace}{.5mm}
\newlength{\extraspaces}
\setlength{\extraspaces}{2.5mm}
\begin{document}
\pubnumber{BONN-TH-95-14 \\ULB-TH-95/09}
\pubdate{}
\title{The Kazhdan-Lusztig conjecture for $\W$-algebras}
\author{KOOS DE VOS
\address{
Physikalisches Institut der Universit\"at Bonn\\ Nussallee 12, 53115 Bonn, Germany}
\medskip {\normalsize \rm and\\}\medskip
PETER VAN DRIEL
\address{
Service de Physique Th\'eorique\\
Universit\'e Libre de Bruxelles, Campus Plaine, C.\ P.\ 225\\
Boulevard du Triomphe, B-1050 Bruxelles, Belgium}
}
\abstract{
The main result in this paper is the character formula for arbitrary
irreducible highest weight modules of $\W$ algebras. The key
ingredient is the functor provided by quantum Hamiltonian reduction,
that constructs the $\W$ algebras from affine Kac-Moody algebras and
in a similar fashion $\W$
modules from KM modules. Assuming certain properties of this functor,
the $\W$ characters are subsequently derived from the Kazhdan-Lusztig
conjecture for KM algebras. The result can be formulated in terms of a
double coset of the Weyl group of the KM algebra: the Hasse diagrams
give the embedding diagrams of the Verma modules and the
Kazhdan-Lusztig polynomials give the multiplicities in the characters.
}
\maketitle
\section{Introduction}
$\W$ algebras were introduced more than a decade ago as (higher spin)
extensions of the Virasoro algebra in the context of two dimensional
conformal field theory~\cite{Zam}. Analogous to the Virasoro algebra, one
expects that the representation theory of $\W$ algebras plays a
crucial role in applications such as in conformal field theories with
$\W$ symmetry, and in theories where the $\W$ symmetry is gauged ($\W$
strings and $\W$ gravity), see~\cite{BS,PH} for reviews. For these applications,
the relevant representations are highest weight modules. A basic goal
is therefore to describe the irreducible modules, and more
specifically to compute their characters. \\
There exists a general approach to find the irreducible characters from
the characters of Verma modules. Any Verma module $M(x)$ can be
decomposed into irreducible highest weight modules $L(y)$ (local
composition series). This gives rise to character formulas of the form
\be\label{eq:ML}
\ch M(x) = \sum_y m_{xy} \, \ch L(y),
\ee
where $m$ is a matrix whose entries $m_{xy}$ count the number of times
that $L(y)$ appears in the decomposition of $M(x)$. Doing this for all
$M(x)$ such that $m$ can be inverted gives
\be\label{eq:LM}
\ch L(x) = \sum_y m^{-1}_{xy} \, \ch M(y).
\ee
The characters of Verma modules are in general easy to compute, hence the
computation of the characters of the irreducible modules boils down to
determining the multiplicities $m_{xy}$. \\
This general programme has been applied successfully to
the Virasoro algebra~\cite{FF}. The key ingredient there is that
every submodule of a Verma module is a sum of Verma modules.
Since there is at most one embedding between Verma modules this implies
that the multiplicities $m_{xy}$ are 0 or 1, and the irreducible
characters follow directly from the embedding pattern of the Verma
modules. These embedding patterns are completely classified, and
consequently for the Virasoro algebra, the characters of {\em all}
irreducible highest weight modules are known. \\
For $\W$ algebras the submodule structure of Verma modules is much
more complicated: in general submodules are not sums of Verma modules.
Therefore the embedding patterns of the Verma modules do not determine
the irreducible characters. This is directly related to the occurence
of multiplicities $m_{xy}\!>\!1$.\\
There are of course also other approaches. For instance, for the
$\W_N$ minimal models, the irreducible character $\ch L(x)$ (for $x$
inside the Kac-table) has been determined directly, using free field
methods~\cite{FL}. In terms of the multiplicities this amounts to
having computed a single row of $m^{-1}$. It does not appear to be
possible to apply these methods to compute the other rows, which is
necessary to determine the characters of all irreducible highest
weight modules ({\it i.e}.\ also for $x$ outside or on the boundary of the
Kac-table). In a way, the results of~\cite{FL} for $\W_N$ algebras
amount to having the $\W$ analogue of the Weyl-Kac character formula
for affine Kac-Moody algebras.\\
For affine KM algebras the characters are known beyond the Weyl-Kac
character formula. For $\kh\neq 0$ the programme described above
has been fully completed. The result can be summarized as follows: \\
(1) The weights $y$ appearing in the decomposition~(\ref{eq:ML}) are
determined by a subgroup of the affine Weyl group, and the
associated Bruhat ordering (the Kac-Kazhdan condition~\cite{KK})\\
(2) The multiplicities $m_{xy}$ are given in terms of the
Kazhdan-Lusztig polynomials associated to the affine Weyl group (the
Kazhdan-Lusztig conjecture~\cite{KL1,DGK})\\
The main ingredient in the proof of (1) is the Jantzen filtration,
whereas (2) has been proven using the intersection cohomology of Schubert
varieties (only for integral weights, for other weights it is still a
conjecture). Neither of these concepts seems to have been worked out
for $\W$ algebras. \\
It is now interesting to note that $\W$ algebras and KM algebras are
intimately related. In particular, a large class of $\W$ algebras can
be obtained from affine KM algebras by (quantum) Hamiltonian
reduction, where one imposes certain constraints on the KM
generators~(see \cite{Fea} for a review). In this way a $\W$ algebra
can be constructed for every embedding of $sl_2$ into the simple Lie
algebra underlying the affine KM algebra~\cite{BTV}. The quantum
construction naturally allows for a BRST formulation, in which the
$\W$ algebra arises as the BRST cohomology of a complex involving the
KM algebra~\cite{FeFr,FKW,BT}. Of course, given an $sl_2$ embedding, one can also compute the
cohomology of a KM module. By construction, the result will be a
module of the corresponding $\W$ algebra. Thus, one obtains in a
natural way a functor from KM modules to $\W$ modules. The
action of this `reduction functor' is in general hard to compute.
In~\cite{FKW}, the action on (resolutions of) admissible KM modules
was computed for principal $sl_2$ embeddings, assuming certain
properties of the reduction functor. This way the characters of the
$\W_N$ minimal models are recovered. \\
The main new idea in this paper is to apply the reduction functor to
`arbitrary' KM modules, to find the analogues of the general results (1)
and (2) for $\W$ algebras. The result is a natural generalization of
the KL conjecture to $\W$ algebras associated to arbitrary $sl_2$
embeddings. We show how this `KL conjecture for $\W$
algebras' can be derived from the KL conjecture for KM algebras,
assuming similar properties as in~\cite{FKW} of the reduction functor.
These assumptions are motivated by the results~\cite{DV}
for finite $\W$ algebras. The upshot is that {\em all} irreducible
characters for such $\W$ algebras are thereby determined. We verified
the conjecture for a nontrivial set of $\W_3$ modules. \\
The setup of this paper is as follows. In section 2, we review the
representation theory and KL conjectures of affine KM algebras,
including a discussion of the translation functor that serves as a
helpful analogy with the reduction functor. Then in section 3, after
some remarks on the representation theory of general $\W$ algebras, we
present the main result of this paper in section 3.2, the KL
conjecture for $\W$ algebras. We also give an idea of how it can be
derived using the reduction functor. Several applications of the conjecture
are discussed in section 4. Properties of Coxeter groups and their KL
polynomials are given in the appendix.
\section{The Kazhdan-Lusztig conjectures for affine Kac-Moody algebras}
We first present a collection of results concerning affine KM algebras
and their highest weight modules, leading to the KL
conjectures. Virtually everything stated here can be found somewhere
in the mathematical literature on the subject, or can be concluded
directly from it. We have avoided a rigorous presentation, but instead
focussed on the line of thought, and made clear what is
well-established and what is conjectured.
For background and explanations on KM algebras and the structure of
the highest weight modules we refer to~\cite{Jant,Kac}, and
for Weyl groups and KL polynomials to~\cite{Hu}.
\subsection{Composition series and character formulae}
Let $\ga$ be an affine KM algebra, and fix a triangular
decomposition $\ga=\na_+\oplus \ha\oplus \na_-$ in positive root
generators, Cartan subalgebra and negative root generators. A singular
vector $v_{\lambda}$ is an eigenvector of the generators of the Cartan
subalgebra $\ha$ with weight $\lambda\in\ha^*$, and is annihilated by
the positive root generators. A highest weight module is a module that is generated from
a singular vector, the highest weight vector, by the action of the negative root generators.
There are two important examples of highest weight modules. The first
is the Verma module $M(\lambda)$, which is uniquely defined by the
property that it is generated freely from $v_{\lambda}$. The second is the quotient of
$M(\lambda)$ by its maximal proper submodule, which gives the unique
irreducible highest weight module $L(\lambda)$. \\
Highest weight modules themselves are special examples of modules in the
so-called category ${\cal O}$~\cite{BGG,DGK}. In general this category consists
of modules $V$ which have a weight space decomposition
\be
\label{eq:weightdeco}
V = \oplus_{\mu \leq \lambda} V_{\mu} \;,
\ee
where the $\mu$'s satisfy $\mu \leq \lambda$ for $\lambda$ in some
finite subset of $h^*$ (recall that $\mu\leq\lambda$ iff
$\lambda-\mu$ is on the positive root lattice $Q_+$ of $\ga$) and $\dim
V_{\mu}<\infty$. The category ${\cal O}$ contains highest weight
modules, tensor products, submodules, quotients, etc.\\
For every module $V$ in ${\cal O}$, one can define a (formal) character
$\ch V$,
\be\label{eq:charv}
\ch V = \sum_{\mu} \dim V_{\mu} \: e^{\mu} \;,
\ee
where the formal exponentials satisfy $e^{\lambda} e^{\mu} =
e^{\lambda + \mu}$ and $e^0=1$. The character of the Verma module
$M(\lambda)$ is given by
\be
\label{eq:vchar}
\ch M(\lambda) = e^{\lambda} \sum_{\gamma \in \Qa_+} P(\gamma)
e^{-\gamma} = e^{\lambda} \prod_{\alpha \in \Da_+}
(1-e^{-\alpha})^{-\dim \ga_{\alpha}} \;,
\ee
where $\ga_{\alpha}$ is the root space of root $\alpha$, $\Delta_+$ is
the set of positive roots and $P(\gamma)$ is the (generalized) Kostant
partition function.
One of the central problems of representation theory is to find the characters of the
irreducible highest weight modules $L(\lambda)$. The strategy is to relate
these to the explicit characters of Verma modules~(\ref{eq:vchar}). This
is possible due to the following general structure theorem, which
also illustrates that ${\cal O}$ is natural in the context of highest
weight modules (in particular, the $L(\lambda)$'s are the only
irreducibles in ${\cal O}$). Every module $V$ in the category
$\cal{O}$ has a local composition series at any weight $\lambda$ of
$V$. A local composition series for $V$ at $\lambda$ is a sequence of submodules of
$V$, $V=V_0 \supset V_1 \supset \ldots \supset V_{n-1} \supset V_n =
0$, such that either $V_i/V_{i+1} \cong L(\mu)$ for some $\mu
\geq \lambda$, or $(V_i/V_{i+1})_{\mu} = 0$ for all $\mu \geq
\lambda$.
One denotes by $[V\!:\!L(\mu)]$ the number of
times that $L(\mu)$ appears in the local composition series of $V$ at
$\lambda$, it is called the multiplicity of $L(\mu)$ in $V$. It is
independent of the particular sequence of submodules one chooses. We stress
that $[V\!:\!L(\mu)]$ does not count the number of singular vectors at
weight $\mu$ in $V$: the statement that $V_i/V_{i+1}
\cong L(\mu)$ only requires that there is a vector $v_{\mu}$ which is
singular in the quotient $V_i/V_{i+1}$ but not necessarily singular
in $V_i$, let alone $V$. A vector that is singular in a quotient of
submodules is called primitive, and the corresponding weight is called
a primitive weight. Obviously, a singular vector is also primitive,
but it is important to realize that there are also other types of
primitive vectors. We also stress that the multiplicities $[M:L]$ can
be larger than $1$, contrary to what was initially thought based on
the known trivial multiplicities of the simple Lie algebras
$\gf=\ol{sl}_2,\ol{sl}_3$ and the affine KM algebra $g=sl_2$.
\\
At the level of characters, the local composition series implies
that~(\ref{eq:charv}) is given by a sum over the irreducible
characters: $\ch V = \sum_{\mu} [V\!:\!L(\mu)] \: \ch L(\mu)$ where
the sum runs over the weights of $V$ (of course, only the primitive
weights give a non-vanishing contribution). This applies in particular
to Verma modules, leading to
\be
\label{eq:verdeco}
\ch M(\lambda) =
\sum_{\mu\leq\lambda} [M(\lambda)\!:\!L(\mu)] \: \ch L(\mu)\;.
\ee
Note that the composition series starts with
$[M(\lambda)\!:\!L(\lambda)]=1$, since $\dim M(\lambda)_{\lambda}=1$.
Ordering the set of weights $\mu \leq \lambda$ as
$\lambda=\mu_0$, $\mu_1$, $\mu_2$, \ldots such that $j \geq i$
whenever $\mu_j \leq \mu_i$, one has the
following set of equations,
$$
\ch M(\mu_i) =
\sum_{\mu_j \leq \mu_i} [M(\mu_i)\!:\!L(\mu_j)] \: \ch L(\mu_j)\;.
$$
The matrix $[M(\mu_i)\!:\!L(\mu_j)]$, called the Jantzen matrix (for
$\lambda$), is upper triangular with ones on the main diagonal.
Therefore, it can be inverted. Denoting the inverse matrix
elements by $(L(\mu_i):M(\mu_j))$ (which are possibly negative
integers), one finds
$$
\ch L(\mu_i) = \sum_{\mu_j \leq \mu_i} (L(\mu_i):M(\mu_j)) \: \ch M(\mu_j).
$$
In conclusion, from~(\ref{eq:verdeco}) one finds the character formula
\be
\label{eq:irchar}
\ch L(\lambda) =
\sum_{\mu \leq \lambda} (L(\lambda)\!:\!M(\mu)) \: \ch M(\mu) \;.
\ee
Here $\ch M(\mu)$ is given through~(\ref{eq:vchar}). Computing $\ch
L(\lambda)$ boils down to computing the numbers
$(L(\lambda)\!:\!M(\mu))$ for all $\mu \leq \lambda$, or equivalently,
the Jantzen matrix $[M(\mu_i)\!:\!L(\mu_j)]$ for $\lambda$.
\subsection{The Kac-Kazhdan conditions}
The first step in determining the multiplicities $[M(\lambda)\!:\!L(\mu)]$
is to find all pairs $\lambda,\mu$ such that $[M(\lambda)\!:\!L(\mu)] \neq
0$. The general solution to this problem has been given by Kac and
Kazhdan~\cite{KK}, using the generalized Casimir of $\ga$ and the
Jantzen-filtration of Verma modules~\cite{Jant}. For the
purposes of this paper it is sufficient to consider only weights
$\lambda$ with $\la\lambda+\rho,\delta\ra=\kh\neq 0$. In that
case the result of~\cite{KK} can be rephrased in terms of properties of the
affine Weyl group~\cite{Kum}.\\
The affine Weyl group $W$ is a Coxeter group, generated by the simple
reflections $s_i$ where $s_i(\lambda)=\lambda-\la\lambda,\alphav_i \ra
\alpha_i$ are the reflections in the simple roots $\alpha_i$ of $\ga$.
Arbitrary elements $w\in W$ correspond to expressions
$w=s_{i_1}s_{i_2}\ldots s_{i_k}$. The minimal number of simple
reflections needed to generate $w$ is called the length $\ell(w)$ of
$w$. An expression of minimal length is called reduced. If $w,w' \in
W$ are two reduced expressions, then we denote $w<w'$ if the reduced
expression for $w$ can be obtained by dropping simple reflections from
a reduced expression for $w'$. The resulting relation $w\leq w'$ is a
partial ordering of $W$, called the Bruhat ordering.
An important ingredient in what follows is the subgroup
$\Wa_{\lambda}\subset\Wa$: it is the group generated by reflections
$r_{\hat{\alpha}}$ with $\hat{\alpha} \in \Delta_{\lambda,+}^{re} = \{
\alpha \in \Delta_+^{re} | \la\lambda + \rho, \alphav \ra \in {\bf Z}
\}$. Clearly, only if $\lambda$ is integral, $W_{\lambda}=W$ (recall
that $\rho\in h^*$ satisfies $\la\rho,\alphav_i\ra=1$), otherwise
$W_{\lambda}$ will be a proper subgroup of $W$ (which for affine $W$
may be isomorphic to $W$). It can be shown that $\Wa_{\lambda}$ is
again a Weyl group, it is generated by simple reflections
$\hat{s}_i=r_{\hat{\alpha}_i}$ (simple in $W_{\lambda}$) where
$\hat{\alpha}_i$ are the simple roots of the rootsystem
$\Delta_{\lambda,+}^{re}$. The length function on $W_{\lambda}$ is
denoted $\ell_{\lambda}(w)$. Obviously, the relation between
$\lambda$ and $\Wa_{\lambda}$ is many-to-one, for instance
$\Wa_{\lambda}=\Wa_{\lambda+\mu}$ for arbitrary integral weight $\mu$.
In fact, up to isomorphisms, there is only a finite number of
$\Wa_{\lambda}$~\cite{Dy}. \\
The groups $\Wa_{\lambda}$ organize the non-vanishing multiplicities
in the following way: the primitive weights of $M(\lambda)$ are
on the shifted Weyl orbit $\Wa_{\lambda}.\lambda$, where
\be\label{eq:shift}
w.\lambda\equiv w(\lambda+\rho)-\rho,
\ee
and vice versa: only the lower weights (with respect to Bruhat
ordering) on the orbit are primitive weights of $M(\lambda)$. In the
remainder of this section we describe this in more detail.
First consider $\kh>0$. Then every orbit $\Wa_{\mu}.\mu$ has
precisely one maximal element $\lambda$, the dominant weight, such that
$w.\lambda \leq \lambda$ for all $w \in \Wa_{\mu}$. Using~(\ref{eq:shift})
it is easy to see that such a dominant weight $\lambda$ is characterised by
\be
\la \lambda + \rho, \hat{\alpha}^{\vee}_i \ra \geq 0.
\ee
Clearly, there is a one-to-one correspondence between dominant weights
$\lambda$ and orbits $W_{\lambda}.\lambda$.
There may not be a one-to-one correspondence between
elements of $\Wa_{\lambda}$ and the weights on the orbit
$\Wa_{\lambda}.\lambda$. This happens precisely if there is a subgroup
$W_{\lambda}^0$ of $\Wa_{\lambda}$ which leaves $\lambda$ invariant.
$W_{\lambda}^0$ is a finite parabolic subgroup of $W_{\lambda}$, generated by the simple reflections
$r_{\hat{\alpha}_i}$ with $\hat{\alpha}_i$ satisfying $\la\lambda
+ \rho, \hat{\alpha}^{\vee}_i \ra = 0$.
A dominant weight is called regular if $W_{\lambda}^0$ is trivial, and
it is called singular otherwise. Thus, weights on the orbit $\Wa_{\lambda}.\lambda$ of a dominant
weight are in one-to-one correpondence with elements of the coset
\be\label{eq:scos}
\Wa_{\lambda}/W^0_{\lambda},
\ee
{\it i.e}.\ any weight $\mu$ can be written uniquely as $\mu=w.\lambda$
with $\lambda$ dominant and $w\in\Wa_{\lambda}/W^0_{\lambda}$. This
coset will be crucial in what follows: in particular {\it the multiplicities depend on
$\lambda$ only through the coset $\Wa_{\lambda}/W^0_{\lambda}$!}\\
Denote $M_w=M(w.\lambda)$ and $L_w=L(w.\lambda)$, then the Kac-Kazhdan
condition for $\kh>0$ can be described as follows
\be\label{eq:kkbruhp}
[M_w\!:\!L_{w'}] \neq 0 \quad \mbox{iff} \quad w \leq w'
\quad \mbox{with $w,w' \in \Wa_{\lambda}/W^0_{\lambda}$}.
\ee
Here, the ordering on the coset $\Wa_{\lambda}/W^0_{\lambda}$ is induced
from the Bruhat ordering on $\Wa_{\lambda}$:
\be
w\leq w'\quad \mbox{with}\, w,w'\in \Wa_{\lambda}/W^0_{\lambda}\qquad \mbox{iff}\qquad
\ul{w}\leq\ul{w}'\quad \mbox{with} \,\ul{w},\ul{w}'\in \Wa_{\lambda}.
\ee
(here $\ul{w}$ is the minimal coset representative of $w$ in the
coset, defined through $\ell(\ul{w}s)>\ell(\ul{w})$ for all $s\in W^0_{\lambda}$.
Of course we could also have chosen the maximal representatives $\ol{w}$ which have
$\ell(\ol{w}s)<\ell(\ol{w})$ for all $s\in W^0_{\lambda}$).
For the character formulas~(\ref{eq:verdeco}) and~(\ref{eq:irchar})
the Kac-Kazhdan result implies the following. First of all, the sum
over the weight space in~(\ref{eq:verdeco}) reduces to a sum over
$w'\in\Wa_{\lambda}/\Wa_{\lambda}^0$
\be \label{eq:verdeco1}
\ch M_w =
\sum_{w'\geq w} [M_w\!:\!L_{w'}] \: \ch L_{w'}.
\ee
Secondly, using transitivity of the Bruhat order this can be inverted
\be \label{eq:irchar1}
\ch L_w =
\sum_{w'\geq w} (L_w\!:\!M_{w'}) \: \ch M_{w'}.
\ee
Unlike the sum in~(\ref{eq:verdeco1}) not all terms in this sum have to be
nonvanishing.
For weights with $\kh<0$ the result can be rephrased analogously. We
note that weights with $\kh<0$ are the image of weights with $\kh>0$
under the shifted inversion
\be
\si.\lambda = -\lambda - 2\rho.
\ee
Clearly, $\Wa_{\si.\lambda} = \Wa_{\lambda}$, so $\si$ is also a one-to-one map
between the orbits on either side (orbits always belong to one side only
as $\Wa$ leaves $\kh$ invariant).
Since $\si$ reverses the order of weights, every orbit now will have a
minimal weight, called anti-dominant, which is of the form $\si.\lambda$
with $\lambda$ dominant. In terms of these anti-dominant weights
one has the analogue of~(\ref{eq:kkbruhp}) describing the
full KK condition for $\kh<0$
\be\label{eq:kkbruhn}
[M_w\!:\!L_{w'}] \neq 0 \quad \mbox{iff} \quad w \geq
w' \quad \mbox{with $w,w' \in \Wa_{\lambda}/W^0_{\lambda}$}.
\ee
Thus one finds the same character formulas~(\ref{eq:verdeco1})
and~(\ref{eq:irchar1}) but with the sum over $w'\leq w$.
\subsection{Embeddings of Verma modules}
In the previous section we have discussed the role of the cosets
$\Wa_{\lambda}/W^0_{\lambda}$ in finding the primitive weights of a
Verma module $M(\lambda)$. In this section we discuss how the same cosets
also describe the embeddings between Verma modules. \\
This is based on the property of KM Verma modules that at every primitive
weight there is at least one singular vector~\cite{KK}. Since a singular
vector $v_{\mu}$ in a Verma module $M(\lambda)$ give rise to a
homomorphism $M(\mu) \hookrightarrow M(\lambda)$ (embedding)
between Verma modules, this statement implies that there is a
homomorphism iff the multiplicity $[M(\lambda)\!:\!L(\mu)]$ is
nonvanishing. Hence
\be\label{eq:emb}
M_{w'} \hookrightarrow M_w \quad \mbox{iff} \quad w \leq w' \quad \mbox{with $w,w' \in \Wa_{\lambda}/W^0_{\lambda}$}.
\ee
In other words: the diagram representing the embeddings of the Verma
modules is given by the Hasse diagram of the coset
$\Wa_{\lambda}/W_{\lambda}^0$: the vertices of this diagram are the
elements of the coset and the links between the vertices connect the
adjacent elements (two coset elements $x,y$ are called adjacent if
there is no third coset element $z$ such that $x<z<y$). Since one can
classify the Hasse diagrams, this gives a classification of embedding
diagrams.
In fact, if $\kh\neq 0$ the relation between embeddings and the Hasse
diagram is even stronger, because in that case there is at most 1
singular vector at every primitive weight. This implies that the
homomorphism $M(\mu) \hookrightarrow M(\lambda)$ is unique, or
\be\label{eq:dimhom}
\dim\mbox{Hom}(M(\mu),M(\lambda))\leq 1.
\ee
This can be argued as follows. If there is a sequence
$M(\mu_1)\hookrightarrow M(\mu_2)\hookrightarrow M(\mu_3)$ of
homomorphisms, the embedding property implies that
\be\label{eq:mulincr}
\dim\mbox{Hom}(M(\mu_1),M(\mu_3))\geq\dim\mbox{Hom}(M(\mu_2),M(\mu_3)).
\ee
For $\kh<0$ any Verma module contains always a lowest primitive
weight (the anti-dominant weight). At this weight, there is precisely one
singular vector (because any two embedded Verma modules
necessarily overlap). This immediately implies~(\ref{eq:dimhom}). \\
For $\kh>0$ there is no lowest primitive weight. In that
case~(\ref{eq:dimhom}) follows from the result for $\kh<0$ through the
`reflection principle' of semi-infinite homology~\cite{FLZ},
\be
\mbox{Hom}(M(\mu),M(\lambda)) \simeq \mbox{Hom}(M(\sigma.\lambda),M(\sigma.\mu)).
\ee
\subsection{Jantzens translation functor}
In this section we discuss how the multiplicities
for {\sl arbitrary} dominant weights follow from
the multiplicities for {\sl regular} dominant weights. The idea is to use Jantzens translation
functor~\cite{Jant,DGK} to map modules with regular weights (trivial
$\Wa^0_{\lambda}$) to modules with singular weights (nontrivial
$\Wa^0_{\lambda}$). The reason for highlighting this ingredient here
is the striking similarity between this derivation and the derivation of $\W$
multiplicities from KM multiplicities using the reduction functor in
section 3.2. \\
Let $\lambda'$ be a singular dominant weight, and let $\lambda$ be a
regular dominant weight such that $\lambda-\lambda'$ is an integral
weight. Clearly, $W_{\lambda}=W_{\lambda'}$, but $W_{\lambda}^0$ is
trivial whereas $W_{\lambda'}^0$ is not. The tensorproduct with the
irreducible module associated with $\lambda'-\lambda$ gives rise to an
exact functor~\cite{Jant,DGK,KW1} (the translation functor) that maps
\be\label{eq:map1}
M(w.\lambda)\stackrel{t}{\mapsto} M(w.\lambda').
\ee
To obtain the action of the translation functor on irreducible
modules, observe that for Verma modules $M(w'.\lambda)
\hookrightarrow M(w.\lambda)$ with $w,w'$ in the same coset, the
functor maps the quotient $M(w.\lambda)\backslash M(w'.\lambda)$
(which contains $L(w.\lambda')$) to zero, so it immediately follows that
\be\label{eq:map2}
L(w.\lambda)\stackrel{t}{\mapsto} L(w.\lambda')\delta_{\wmax,w}.
\ee
(with $\wmax$ is the maximal representative of $w$ in the coset
$W_{\lambda}/W_{\lambda'}^0$). The maps~(\ref{eq:map1})
and~(\ref{eq:map2}) determine the multiplicities for singular weights
from the multiplicities of the regular weights:
\be
[M(w.\lambda'):L(w'.\lambda')]=[M(\wmax.\lambda):L(\wmax'.\lambda)].
\ee
Another useful application of the translation functor is the
computation of the character of the irreducible module $L_e$ for
regular dominant weights $\lambda$ (without having to determine the
full Jantzen matrix). For such weights namely, the sum
in~(\ref{eq:irchar1}) runs over all the elements of
$W_{\lambda}$. Applying~(\ref{eq:map1}) and~(\ref{eq:map2}) to it
for a translation chosen such that $W_{\lambda'}^0$ contains just one
reflection, gives that the coefficients are given by
$\ve_w=(-1)^{\ell_{\lambda}(w)}$~\cite{KW1}, hence
\be\label{eq:kwchar}
\ch L_e = \sum_{w \in W_{\lambda}} \ve_w \ch M_w.
\ee
This is the generalisation of the Weyl-Kac formula~\cite{Kac} to
arbitrary regular dominant weights. The same trick cannot be applied to
obtain arbitrary characters ({\it i.e}.\ $\ch L_w$ or for $\lambda$ singular).
It is this particular character formula (for admissible $\lambda$)
that forms the starting point of~\cite{FKW} for generalization to $\W$ algebras.
\subsection{The KL conjectures}
\label{kl}
Now we are ready to describe the final step, {\it i.e}.\ to give the
Kahdan-Lusztig formula for the multiplicities. In~\cite{KL1}, Kazhdan
and Lusztig defined a set of polynomials $P_{x,y}(q)$, labelled by
pairs of elements $x,y$ for an arbitrary Coxeter group $W$, and
depending on a single variable $q$. For details and properties about
the definition of these polynomials see the appendix, important for us
is that they can be computed explicitly from a recursion relation
(see~(\ref{eq:recrel}))
\be\label{eq:recrel2}
P_{x,ys}=q^{1-c}P_{xs,y}+q^cP_{x,y}-q\sum_{\stackrel{x\leq
z<y}{zs<z}} P_{x,z}\slash{\!\!\!\!P}_{z,y}.
\ee
The simple reflection $s$ is chosen such that $y<ys$, such that
the polynomials $P_{x,y}$ are expressed in terms of polynomials
$P_{x',y'}$ with $\ell(y')<\ell(y)$. \\
Similarly, one defines a set of inverse polynomials $Q_{x,y}(q)$
through
\be
\sum_{x \leq z \leq w}
P_{x,z}(q) Q_{z,y}(q)\ve_z\ve_y = \delta_{x,y}
\ee
which can also be computed directly from a recursion relation (see~(\ref{eq:qrecrel}))
\be\label{eq:qrecrel2}
Q_{x,ys} = c Q_{xs,y} + (-q)^c Q_{x,y} + c q \sum_{\stackrel{x <
z \leq y}{zs>z}} \slash{\!\!\!\!Q}_{x,z} Q_{z,y}
\ee
Analogously, one may also associate KL polynomials $P^I,Q^I$ to a
coset $W/W_I$ for $W_I$ a parabolic subgroup of $W$. If $W_I$ is
finite these are related to the KL polynomials on $W$ as follows
\be
\label{eq:oneside}
P^{I}_{x,y} = P_{\xmax,\ymax}, \qquad Q^{I}_{x,y} = Q_{\xmin,\ymin}
\ee
Here $\zmin$ and $\zmax$ are the minimal and maximal representatives
of $z$ in the coset $[z]$. In general, the polynomials $P^I$ and $Q^I$
are not each others inverse. The inverse polynomials of $P^I,Q^I$
are denoted $\tilde{Q}^I,\tilde{P}^I$, they are defined through
\be
\sum_{x \leq z \leq y} \tilde{Q}^{I}_{x,z} P^{I}_{z,y} = \sum_{x \leq
z \leq y} Q^{I}_{x,z} \tilde{P}^{I}_{z,y} = \delta_{x,y}.
\ee
They can also be expressed in terms of the polynomials on $W$
\be
\tilde{P}^{I}_{x,y} = \sum_{z \in [x]} P_{z,\ymin} \ve_z
\ve_{\ymin},\qquad \tilde{Q}^{I}_{x,y} = \sum_{z \in [y]} Q_{\xmax,z}
\ve_{\xmax} \ve_z
\ee
The KL conjectures relate the multiplicities in the character formulas to
the value of these polynomials at $q=1$. Let $\lambda$ be a dominant
weight with coset $\Wa_{\lambda}/\Wa_{\lambda}^0$, $P_{w,w'}$ the KL
polynomials for $W_{\lambda}$ and $Q_{w,w'}$ the associated inverse KL
polynomials. Then the multiplicities are given by~\cite{KL1,DGK,Jant,Lus1}
\be
\label{eq:klkmd}
\begin{array}{lll}
\kh>0 : \qquad&
[M_w\!:\!L_{w'}] = P^I_{w,w'}(1), \quad&
(L_w\!:\!M_{w'}) = \tilde{Q}^I_{w,w'}(1) \\ {}
\kh<0 :\qquad &
[M_w\!:\!L_{w'}] = Q^I_{w',w}(1), \quad&
(L_w\!:\!M_{w'}) = \tilde{P}^I_{w',w}(1)
\end{array}
\ee
(the superscript $I$ refers to the subgroup $W^0_{\lambda}$)
These conjectures have been proven for integral weights,
in~\cite{iwp} for $\kh>0$, and~\cite{iwn} for $\kh<0$. It is not
inconceivable that the conjectures for $\kh>0$ are related to the
conjecture for $\kh<0$ through the semi-infinite cohomology of affine
KM algebras. \\
The conjectures naturally fit in a circle of ideas generally referred
to as Kazhdan-Lusztig theory. This theory interrelates many different
problems, such as the classification of primitive ideals in enveloping
algebras, the computation of the multiplicities in composition series
and the intersection cohomology of Schubert varieties (see~\cite{shi}
for an overview). It applies in particular to simple Lie algebras,
affine KM algebras and quantum groups. In the next section we show
that it also applies to $\W$ algebras.
\section{The KL conjectures for $\W$ algebras}
Compared to the situation for affine Kac-Moody algebras, relatively
little is known about the representation theory of $\W$ algebras. The
fact that a classification of such algebras is still lacking makes it
harder to give a general approach to this problem. We claim, however,
that for the class of $\W$ algebras obtained through hamiltonian
reduction of affine KM algebras, the analogue of most results
described in the previous section exists. In particular, we formulate
the KL conjecture for such $\W$ algebras.
\subsection{Some generalities on $\W$ algebras and modules}
Let us first consider a general $\W$ algebra, generated by the modes of
a finite set of quasiprimary fields (for a precise definition see~\cite{BS}).
The $\W$ algebra will have a CSA $h$, {\it i.e}.\ a maximal abelian subalgebra of
the zero modes. Unlike for KM algebras, the adjoint action of $h$
on the generators of the $\W$ algebra is in general not diagonalizable
({\it e.g}.\ the zero-mode $W_0$ of the spin 3 field of $\W_3$); therefore the
`triangular' decomposition of $\W=\W_+\oplus h\oplus\W_-$ in positive
root generators, Cartan subalgebra and negative root generators is
given with respect to a subalgebra $h'\subset h$
\be
\W = \oplus_{a'} \W_{-a'} \oplus h \oplus_{a'} \W_{a'},
\ee
where $a'$ runs over the set of positive roots $\Delta'_+$. By assumption
$\W_{-a'} \cong \W_{a'}$ as vector spaces, paired by an involutive map
$\sigma : \W_{-a'} \rightarrow \W_{a'}$.\\
The set-up of representation theory is similar to that of affine KM
algebras, in the following sense. A singular vector $v_a$ is an
eigenvector of the generators of $h$ with weight $a \in h^*$, and
$v_a$ is annihilated by all positive root generators. A highest weight
module $V$ is generated from $v_a$ by the action of the negative root
generators.
Similarly, one introduces a category ${\cal O}$, which consists of
modules $V$ which have a weight space decomposition into
a direct sum of weight spaces of the subalgebra $h'$,
\be \label{eq:wdeco}
V = \oplus_{b' \leq a'} V_{b'},
\ee
where the sum is over weights $b'$ satisfying $b' \leq a'$ for $a'$ in
some finite subset of $h'^*$, and $\dim V_{b'} < \infty$ (note that
$b' \leq a'$ iff $a' - b'$ is on the positive root lattice $\Qa'_+$ of $\ha'$). \\
The category ${\cal O}$ again contains Verma modules $M(a)$, irreducible quotients
$L(a)$, submodules, etc (but no tensor products as in general the tensor product of
two $\W$ modules is not a $\W$ module).\\
For every module $V$ in ${\cal O}$ one
can define a (formal) character $\ch V$,
\be
\ch V = \sum_{b'} \dim V_{b'} e^{b'}.
\ee
The Verma module $M(a)$ has character formula
\be
\ch M(a) = e^{a'} \sum_{b' \in \Qa'_+} P(b') e^{-b'} = e^{a'}\prod_{b' \in
\Delta'_+} (1-e^{-b'})^{-\dim \W_{b'}},
\ee
where $P(b')$ is some generalized Kostant partition function. \\
The finite dimensionality of the weight spaces $V_{b'}$ implies that
the action of the generators of the CSA $h$ outside $h'$ is reasonably
well-behaved: every weight space $V_{b'}$ can be decomposed into a
finite number of Jordan blocks $U_b$,
\be
V_{b'} = \oplus_b U_b.
\ee
This implies that one can make local composition series in ${\cal O}$,
where the irreducible quotients are again the highest weight modules
$L(b)$, occurring with multiplicities $[V\!:\!L(b)]$. This leads to
character formulas
$
\ch V = \sum_b [V\!:\!L(b)] \ch L(b),
$
where of course $b$ can only appear in the sum if $b'$ is a weight of
$V$. This applies in particular to a Verma module $M(a)$, leading to
\be\label{eq:wverrep}
\ch M(a) = \sum_{b' \leq a'} [M(a)\!:\!L(b)] \ch L(b),
\ee
where clearly $[M(a)\!:\!L(a)] = 1$.
Once again, this character formula can be inverted, such that the
characters of irreducible modules can be expressed in characters of Verma modules
\be\label{eq:wirrep}
\ch L(a) = \sum_{b' \leq a'} (L(a)\!:\!M(b)) \ch M(b).
\ee
To conclude: also for $\W$ algebras, the general strategy to find
character formulas is to compute the multiplicities $[M(a)\!:\!L(b)]$.
This is what we will do in the next section.
\subsection{$\W$ modules from $sl_2$ reductions}
A large class of $\W$ algebras can be obtained through a procedure of
(quantum) Hamiltonian reduction of affine KM algebras~\cite{Fea}. A
particularly nice set of reductions are those related to $sl_2$
embeddings~\cite{BTV}. For every $sl_2$ embedding into the simple Lie
algebra underlying the untwisted affine KM algebra, one can define a
BRST complex such that the associated cohomology is non-vanishing only
in the zero-th term. This cohomology is a $\W$
algebra~\cite{FeFr,FKW,BT}. \\
Similarly, on the level of the representation theory, the cohomology
of a complex associated to a KM module gives a $\W$ module. This
defines a functor from the category of KM modules to the category of
$\W$ modules. We assume the following properties of this reduction
functor~\cite{FKW,DV}:
(1)
the cohomology of the BRST complex associated to the KM module is
non-vanishing only in the zero-th term,
(2)
KM Verma modules $M(\lambda)$ are mapped to $\W$ Verma modules
$M(a(\lambda))$, and
(3)
a local composition series of a KM Verma module is mapped to a local composition series of the
corresponding $\W$ Verma module.
From these assumptions it immediately follows that, when acting on KM
irreducible modules, the reduction functor maps
\be\label{eq:par}
L(\lambda)\rightarrow L(a(\lambda)) \quad\mbox{or}\quad L(\lambda)\rightarrow 0,
\ee
where at least one of the maps is to be non-trivial.
If one knows which $L(\lambda)$ have vanishing or
non-vanishing cohomology, then the multiplicities of $\W$ Verma modules
are determined. The main result of this paper is an explicit
formula for these multiplicities, in terms of KL polynomials
associated to a double coset which is completely fixed by the
reduction data. Note that the reduction only gives rise to a $\W$
algebra for $\kh\neq 0$, {\it i.e}.\ precisely those weights for which the
KM multiplicities are given by the KL conjecture. This implies that one
has the complete KL conjecture for this class of $\W$ algebras, so
that the characters of all irreducible highest weight $\W$ modules are known.
Let us explain how this should work. Associated to the particular
$sl_2$-reduction is a regular subalgebra $g_r$ of the
finite-dimensional simple Lie algebra $\gf$ underlying the affine
Kac-Moody algebra $\ga$~\cite{DV}. The $sl_2$ subalgebra is principally embedded
into $g_r$. This embedding determines a set of constraints which can
be chosen in such a way that they involve only positive roots. This is
necessary to get non-vanishing cohomology from KM Verma modules. In
explicit examples it is possible to verify that this cohomology is
given by a Verma module of the corresponding $\W$
algebra~\cite{FKW,DV}. We assume that this holds in general.
From the results of~\cite{DV} we expect that the parametrization
$a(\lambda)$ of the $\W$ weight is invariant under the shifted action
of the Weyl group $W^r$ of $g_r$ (which is a finite parabolic subgroup
of $\Wa$). More precisely
\be\label{eq:inva}
a(w.\lambda) = a(\lambda) \quad \mbox{iff} \quad w\in W^r,
\ee
so there is a one-to-one correspondence between the $\W$-weights and
the invariants of the Weyl group $W^r$.
Using this parametrization we will from now on
denote Verma modules and irreducible modules for the $\W$ algebra by
$M^r(\lambda)$ and $L^r(\lambda)$ with $\lambda$ a weight of $\ga$. Up
to $W^r$ invariance, the labelling by $\ga$ weights fixes the $\W$ weights
uniquely. \\
Let $\lambda$ be a dominant weight. From the existence of the
composition series it follows that the set of primitive weights in a
$\W$ Vermas module $M^r(\lambda)$ is contained in the orbit of the double coset
\be\label{eq:dcos}
W^r_{\lambda} \backslash \Wa_{\lambda}/W^0_{\lambda},
\ee
where $W^r_{\lambda}= W^r\cap W_{\lambda}$.
From the embedding property~(\ref{eq:emb}) of KM Verma modules it now
follows that for each weight on this orbit there is an embedding of
$\W$ Verma modules, thus there is a one-to-one correspondence between
primitive weights and weights on the orbit of the double
coset~(\ref{eq:dcos}).
\\
It is instructive to note the analogy with the translation functor
discussed in section 2.4: the translation functor maps regular KM Verma
modules $M(\lambda)$ to arbitrary KM Verma modules $M(\lambda')$, such
that the relevant cosets $W_{\lambda}$ are mapped to
$W_{\lambda'}/W_{\lambda'}^0$. Similarly, the reduction functor maps
arbitrary KM Verma modules $M(\lambda)$ to arbitrary $\W$ Verma modules
$M^r(\lambda)$, such that the relevant cosets
$W_{\lambda}/W_{\lambda}^0$ are mapped to $W_{\lambda}^r\backslash
W_{\lambda}/W_{\lambda}^0$. Indeed, the derivation of the $\W$
multiplicities from KM multiplicities from this point on goes completely
analogous to the derivation in section 2.4.
\\
The irreducible $\W$ module $L^r(\mu)$ may arise only as
the cohomology of the KM modules $L(w.\mu)$ with $w\in W^r$.
Oviously, the cohomology of the associated KM Verma modules
$M(w.\mu)$ are identical. Therefore, the cohomology of $L(\mu)$ must
vanish when there is a $w\in W^r_{\mu}$ such
that $M(w.\mu)\subset M(\mu)$ with $w.\mu\neq\mu$.
On every $W^r_{\mu}$ orbit of $\mu$, only the lowest weight
contributes therefore. \\
It follows that the reduction functor maps
\be\label{eq:maps}
M_w \rightarrow M^r_w
,\qquad
L_w \rightarrow L^r_w\delta_{w,\wmax}.
\ee
where again $M^r_w=M^r(w.\lambda),L^r_w=L^r(w.\lambda)$ and
$\wmax$ is the maximal representative of $w$ in the double
coset~(\ref{eq:dcos}).\\
Thus we observe that again the way to associate KL polynomials with
the double coset~(\ref{eq:dcos}) is to take maximal representatives.
To summarize, consider the $\W$ algebra associated with the regular
subalgebra $g_r$. Let $\lambda$ be a dominant weight, and let $w,w'
\in W^r_{\lambda} \backslash \Wa_{\lambda}/W^0_{\lambda}$.
Denote the double coset of $w$ by $[w]$, the minimal representatives
by $\wmin$ and the maximal representative by $\wmax$, and define the
following polynomials
\be\label{eq:twoside}
\begin{array}{ll}
P^{IJ}_{w,w'} = P_{\wmax,\wmax'},\qquad&
Q^{IJ}_{w,w'} = Q_{\wmin,\wmin'}.\\
\tilde{P}^{IJ}_{w,w'} = \sum_{x \in [w]} P_{x,\wmin'} \ve_x \ve_{\wmin'}, \qquad&
\tilde{Q}^{IJ}_{w,w'} = \sum_{x \in [w']} Q_{\wmax,x} \ve_{\wmax} \ve_x.
\end{array}
\ee
\begin{conjecture}[KL conjecture for $\W$ algebras]
The multiplicities in Verma modules are given by the KL
polynomials associated with the double coset~(\ref{eq:dcos})
\be
\label{eq:klw1}
\begin{array}{lll}
\kh>0:\qquad &
[M^r_w:L^r_{w'}] = P^{IJ}_{w,w'}(1),\quad&
(L^r_w:M^r_{w'}) = \tilde{Q}^{IJ}_{w,w'}(1).\\
\kh<0:\qquad &
[M^r_w:L^r_{w'}] = Q^{IJ}_{w',w}(1),\quad&
(L^r_w:M^r_{w'}) = \tilde{P}^{IJ}_{w',w}(1).
\end{array}
\ee
\end{conjecture}
Hence the character formulas for irreducible $\W$ modules is given by
\be
\label{eq:charfor1}
\begin{array}{ll}
\kh>0:\quad\ch L^r_w=\sum_{w'\geq w} \tilde{Q}^{IJ}_{w,w'}(1) \ch M^r_{w'}.\\
\kh<0:\quad\ch L^r_w=\sum_{w'\leq w} \tilde{P}^{IJ}_{w',w}(1) \ch M^r_{w'}.
\end{array}
\ee
\begin{conjecture}
The embedding diagram of Verma modules corresponds to the Hasse diagram of the double
coset~(\ref{eq:dcos})
\be\label{eq:dihow}
\left.\begin{array}{c}
\kh>0:\quad M^r_{w'}\hookrightarrow M^r_w \\
\kh<0:\quad M^r_w\hookrightarrow M^r_{w'}\end{array}\right\}
\quad \mbox{iff}
\quad w \leq w' \quad \mbox{with $w,w' \in W^r_{\lambda} \backslash
\Wa_{\lambda}/W^0_{\lambda}$}.
\ee
\end{conjecture}
Moreover, we expect that also for $\W$ algebras there is just one
singular vector at any given weight. The KL conjecture supports this
as follows. For $\kh<0$ there is an anti-dominant weight, so here the
proof is identical to the case discussed in section 2.3. Barring a
reflection principle for $\W$ algebras, a general proof for $\kh>0$ is
lacking. However, in the examples we studied, the polynomials appear
to have the property that at arbitrary length $\ell(w)$ one can always
find a $w$ such that polynomial $P_{e,w}=1$. This provides an upperbound for the
number of singular vectors at that weight and consequently also at
every primitive weight $w'.\lambda$ for any $w'\leq w$
(see~(\ref{eq:mulincr})). So we expect that also for $\W$ algebras,
$\dim\mbox{Hom}((M^r(\lambda),M^r(\mu))\leq 1$.
\section{Examples and Applications}
In this section we discuss some examples that on the one hand
provide evidence for the validity of the conjectures, and on the other
are an illustration of their effectiveness for
actual computations. In particular the (explicit) calculation of $\W$
characters for irreducible highest weight modules is now reduced to
combinatorics on the Weyl group of $\ga$. For simplicity we restrict
to $g=sl_N$. If $\lambda=\sum_{i=0}^l\lambda_i\Lambda_i$ is a dominant
weight (where $\Lambda_i$ are the fundamental weights of $sl_N$,
$\la\Lambda_i,\alphav_j\ra=\delta_{ij}$), then the level
$k=\sum_{i=0}^l\lambda_i$, $P^k_+$ is the set of dominant integral
weights of level $k$, and $P^k_{++}$ are the regular weights in
$P_+^k$. The finite part $\lambdab\in\ol{h}^*$ is
$\lambdab=\sum_{i=1}^l\lambda_i\Lambda_i$ and finally $h^{\vee}=N$.
\subsection{Comparison with known results}
The first check is provided by the Virasoro algebra, which is the quantum
Hamiltonian reduction of the affine KM algebra $g=sl_2$. In that case,
it is a straightforward exercise to show that the conjectures agree
with the results of Feigin and Fuchs: (1) the
embedding diagrams are classified by the double cosets of the
reflection subgroups of the affine Weyl group $\hat{a_1}$ (see Table~\ref{FFtab})
and (2) the multiplicities in the characters are given by the
corresponding KL polynomials: $P_{x,y}=Q_{x,y}=1$ for all $x\leq y$.
{\sl
\begin{table}[hbt]
\begin{center}
\begin{tabular}{c|c}
Feigin-Fuchs & coset \\ \hline
$I$ & trivial \\ \hline
$II_{\pm}$ & $a_1$ \\
$II^0$ & $a_1/a_1$ \\ \hline
$III_{\pm}$ & $\hat{a}_1$ \\
$III_{\pm}^0$ & $\hat{a}_1/a_1$ \\
$III_{\pm}^{00}$ & $a_1 \backslash \hat{a}_1/a_1$
\end{tabular}
\end{center}
\caption{\sl classification of embedding patterns of the Virasoro
algebra.}\label{FFtab}
\end{table}
}
A second check is provided by the $\W_N$ minimal models, which are the
quantum Hamiltonian reduction of the affine KM algebra $g=sl_N$ with
respect to the principal $sl_2$ embedding. Consider the dominant weights $\lambda$ with
$W_{\lambda}$ isomorphic to $W$~\cite{KW1}
\be\label{eq:adm}
\lambda+\rho = w(\Lambda^+ - t \Lambda^-),
\ee
where $t=p/p'$ ($p,p'$ relative prime integers), $\Lambda^+\in P_{++}^{p}$,
$\Lambda^-\in P_{++}^{p'-1}$ and $w$ an arbitrary element of the Weyl
group $\Wf$ of $\gf$. The simple roots of $\Delta^{re}_{\lambda,+}$
are given by $\hat{\alpha}_i = w(\alpha_i) + \Lambda^-_i \delta$. Let
$\hat{s}_i=r_{\hat{\alpha}_i}$, then
\begin{itemize}
\item[1.]
$W_{\lambda}$ is generated by the simple reflections $\hat{s}_i$
\item[2.]
$W_{\lambda}^0$ is generated by the $\hat{s}_i$ for which $\Lambda^+_i=0$,
\item[3.]
$W_{\lambda}^r$ is generated by the $\hat{s}_i$ for which $\Lambda^-_i=0$ and
$\alpha_i$ is a simple root of $g_r$.
\end{itemize}
The $\W_N$ minimal models arise from dominant weights~(\ref{eq:adm})
which have trivial $W_{\lambda}^0$ and $W_{\lambda}^r$, hence
$\Lambda^+-\rho \in P_{++}^{p-N}$ and $\Lambda^- -\rho\in P_{++}^{p'-N}$.
The multiplicities $Q_{e,w}$ for these regular dominant are easily read off
from the recusion relation~(\ref{eq:qrecrel2}) for $x=e$, since in
that case $Q_{e,ys} = Q_{e,y}$ for all $sy>y$ so it easily follows
that $Q_{e,w}=1$.
This reproduces the character formulas of~\cite{FL,FKW} for the
$\W_N$ minimal models. Similarly, admissible modules for arbitrary $\W$
algebras can be obtained. As should be clear from above, the only difference
will be in the domain of $\Lambda^-$.\\
New testcases for weights $\lambda$ with $W_{\lambda}\cong W$ arise
when one considers nontrivial subgroups $W_{\lambda}^0$ and/or
$W_{\lambda}^r$ and non-dominant highest weights. In the next section
we will do this for the case of the $\W_3$ algebra.
\subsection{Classification of $\W_3$ modules}
The Zamolodchikov algebra $\W_3$~\cite{Zam} is the quantum Hamiltonian
reduction of the affine KM algebra $sl_3$ with respect to the
principal $sl_2$ subalgebra~\cite{BO}. The $W_3$ weights are
$(h,w,c)$, the eigenvalues of the zero modes $L_0,W_0$ and $c$
respectively. The parametrization of the $\W_3$ weights in terms of the $sl_3$ weight
$\lambda$ that follows from the BRST construction is
\be\label{eq:w3norm}
h=\frac{1}{2t}|\lambdab+\rhob|^2 + \frac{c-2}{24},\quad
w=\frac{1}{27t(t-1)}(\lambdab+\rhob,\Lambda_1)
(\lambdab+\rhob,\Lambda_2)
(\lambdab+\rhob,\Lambda_1-\Lambda_2),
\ee
with $c=50-24t-24/t$ for $t=k+3$. The character of a $\W_3$ Verma
module is given by $\ch M^r(\lambda)=q^h \eta(q)^2$ (note that if
$t=1$, the parametrization (\ref{eq:w3norm}) is singular, in that case
one replaces $w\rightarrow (t-1) w$). \\
Up to isomorphism, there are two nontrivial parabolic subgroups of the
affine Weyl group $\hat{a}_2$ of $sl_3$, namely $a_1\simeq Z_2$ and
$a_2\simeq D_3$. This gives
rise to 9 inequivalent double cosets (we eliminated the
invariance of the $\W_3$ weights under $(t,\lambdab) \rightarrow (1/t,-\lambdab/t)$, which
interchanges $W_{\lambda}^0$ and $W_{\lambda}^r$), see table 2.
For each of these double cosets, we computed the multiplicities
$P^{IJ}_{w,w'}(1)$ (for the first $15$ elements) from the KL
polynomials $P_{w,w'}(q)$ of $sl_3$.
Together with the associated Hasse diagrams, they are given in
tables 5-13. For completeness, we also give the KL polynomials
$P_{w,w'}(q)$ and $Q_{w,w'}(q)$ for $sl_3$, which we computed up till
$\ell(w)=15$, tables 3~\footnote{We thank M.\ Goresky, who first
computed this table for us.} and 4. \\
To check if the polynomials and Hasse diagrams correspond with
multiplicities and embedding diagrams of $\W$ Verma modules, we
subsequently calculated (parts of) the irreducible characters and
embedding patterns directly on the Verma module.\\
This goes as follows. Starting from a highest weight vector $v_a$
(eigenvector of $L_0,W_0,c$ with eigenvalues $h,w,c$, and annihilated
by the $L_n,W_n$ for $n>0$) a basis $M(a)_{h+N}$ of the
Verma module at depth N is constructed. The rank of the innerproduct matrix (Shapovalov form)
at depth N gives the dimension of the irreducible character at depth
$N$, and the eigenvectors of $W_0$ in the kernel of $L_1,L_2$ and $W_1$ gives the singular vectors. \\
In practice, even at modest depth these calculations require a lot of computer
time: with the Mathematica routines we had available, the singular
vectors could generically be determined up to depth 9, and the
characters up to depth 6\footnote{to appreciate the effectiveness
of the KL conjecture: applying table 3 to the vacuum module
for $c=0$ the characters are already determined beyond depth 200,
where the Verma module has of the order of $10^{20}$ states.}
To verify the predictions of the
conjectures, we selected a dominant weight with every coset such that
the singular vectors in the associated Verma module occur at the
lowest possible levels, see table 2. For every coset we computed
the characters of the first 15 submodules, and reconstructed the
embedding patterns. We found complete agreement with the KL conjecture.
\begin{table}[bht]\label{tab3}
\begin{center}
\begin{tabular}{cccrrrrrc}\hline
$W^r_{\lambda}$&$W^0_{\lambda}$ &$t$&$\Lambda^+$&$\Lambda^-$& $c$ &
$h$ & $w$ & {\small Table}\\ \hline
$a_2$&$a_2$ &1 & (1, 0, 0) & (1, 0, 0) & 2 & 0 & $0$ &5 \\
$a_2$&$a_2'$&1 & (0, 1, 0) & (1, 0, 0) & 2 &$\frac{1}{3}$& $\frac{2}{27}$&6 \\
$a_2$&$a_1$ &2 & (1, 1, 0) & (1, 0, 0) &$-10$&$-\frac{1}{3}$& $\frac{1}{27}$&7\\
$a_2$&$a_1'$&2 & (0, 1, 1) & (1, 0, 0) &$-10$& 0&0&8\\
$a_2$& -- &3 & (1, 1, 1) & (1, 0, 0) &$-30$&$ -1$& 0&9\\
$a_1$&$a_1$ &3/2 & (2, 1, 0) & (1, 1, 0) &$-2 $&$-\frac{1}{9}$& $-\frac{1}{81}$&10\\
$a_1$&$a_1'$&3/2 & (2, 0, 1) & (1, 1, 0) &$-2 $&$\frac{2}{9}$& $\frac{10}{81}$&11\\
$a_1$& -- &3/2 & (1, 1, 1) & (1, 1, 0) &$-2 $& 0& 0&12\\
-- & -- &4/3 & (2, 1, 1) & (1, 1, 1) & 0 & 0 & 0&13\\ \hline
\end{tabular}
\caption{\sl Classification of $\W_3$ modules with $W_{\lambda}$
isomorphic to $\hat{a}_2$.}
\end{center}
\end{table}
\subsection{Closed character formulae}
The practical upshot of the KL conjectures is that characters of $\W$
algebras can be computed using only the combinatorics of double cosets
of affine Weyl groups. In general however, the word problem posed by the
recursion relation~(\ref{eq:qrecrel2}) is too complicated to solve in closed form.
Only in certain special cases, one does get a closed
expression for character formulas, as in the case of the Virasoro
algebra ($Q_{x,y}=1$ for $x\leq y$) and the $\W_N$ minimal models
($Q_{e,y}=1$). We end this section by
discussing two more examples where a closed formula can be obtained:
cosets of type $\Wf\backslash \Wa$ and of type $\Wf\backslash \Wa/\Wf$
\subsubsection{The coset $\Wf\backslash \Wa$}
Consider a coset of the form $\Wf\backslash \Wa$, with
$\Wa$ the affine Weyl group and $\Wf$ the corresponding finite Weyl
group. These cosets correspond to modules on the boundary of
the Kac-table (type $III^0$ of the Virasoro), where the characters are
given by finite sums over Verma characters. \\
For the $\W_3$ algebra $W=\hat{a}_2$ and $\ol{W}=a_2$. The KL
polynomials $P^{IJ}_{x,y}$ and Hasse diagram of $a_2\backslash\hat{a}_2$ is given in
table 9. In that case it can be shown that there are 2
different character formulas, depending only on the length
of $\wmin$ (the minimal representative of $w$ in the coset).
If the length $\ell(\wmin)$ is odd, there are precisely 2 adjacent
elements of $\wmin$ of length $\ell(\wmin)+1$, of the form $\wmin.j$ and $\wmin.k$
(where $j$ and $k$ are distinct simple reflections). Let $a_2$ denote the $a_2$ generated
by $j$ and $k$. Then the character reads
\be\label{eq:sglhex}
\ch L(w.\lambda) = \sum_{x\in a_2} \epsilon_x \ch M(w.x.\lambda)
\ee
If the length $\ell(\wmin)$ is even, there are at most 3 adjacent
elements of $\wmin$ of length $\ell(\wmin)+1$, and there is precisely one of the form
$\wmin.i$ (for $i$ a simple reflection).
Then we find
\be\label{eq:dblhex}
\ch L(w.\lambda) = \sum_{x\in a_2} \epsilon_x ( \ch M(w.i.x.i.\lambda)
- \ch M(w.i.x.\lambda))
\ee
where again the $a_2$ is generated by $j,k$ ($i,j,k$ are distinct simple reflections).
It appears that these two formulas are related to the generic
decomposition patterns of Weyl modules, studied in~\cite{Lus4}. Similarly,
there is 1 formula in the case of $A_1$, 4 different formulas for
$B_2$ and 12 for $G_2$. \\
These character formulas apply in particular to the (p,1) topological minimal
models. The admissible weights~(\ref{eq:adm}) in that case are
integral, hence $W^r_{\lambda}=\Wf$. The regular integral weights
\be\label{eq:primfix}
(\lambda+\rho,\alphav_i)\neq 0\bmod p
\ee
are on the orbit of the coset $\ol{W}\backslash W$, with
dominant weights $\lambda\in P^{p}_{++}$. For $\W_3$
therefore, provided $p\geq 3$,~(\ref{eq:sglhex}) and (\ref{eq:dblhex})
give the character formulas for all the regular weights.
We observe that~(\ref{eq:primfix}) is exactly the condition for the
physical states in (p,1) topological minimal matter coupled to
$W$-gravity~\cite{LWS}. An interesting open question is whether the
non-trivial multiplicities indicate the presence of extra physical states.
The character formulas~(\ref{eq:sglhex}) and (\ref{eq:dblhex}) apply in
other cases as well, for instance (i) for regular integral weights on the boundary
of the $(p,p')$ Kac table, (ii) for weights with $\Wa_{\lambda}^0=\Wf$
and $\Wa_{\lambda}^r$ trivial (interchanging left and right
multiplication) and (iii) for the other $g=sl_3$ related $\W$ algebras
($W_3^{(2)}$ and $sl_3$ itself).
\subsubsection{The coset $\Wf\backslash \Wa/ \Wf$}
Consider a coset of the form $\Wf\backslash \Wa/\Wf$, with
$\Wa$ an affine Weyl group and $\Wf$ a subgroup isomorphic to the finite Weyl
group $\Wf$. These cosets correspond to modules in a corner of the
Kac-table (type $III^{00}$ of the Virasoro), where again the
characters are given by finite sums over Verma characters, but in
addition they are now grouped in $\gf$ multiplets.
Specifically, the multiplicities $P^{IJ}_{x,y}(1)$ are related to the
dimension of weight spaces in finite dimensional modules of
the simple Lie algebra $\gf$~\cite{Lus3}. The correspondence is as follows. \\
First consider the case where the embedding of the left- and right
subgroups is the same and given by $\Wf$ (depending on $\ga$ there may
be more ways to embed $\Wf$ in $\Wa$). Since $\Wa=\Wf\cdot \Qf$
(semi-direct product) and in $\Qf$ there is a unique dominant element
$\alpha$ on each $\Wf$ orbit, it follows that there is a one-to-one
correspondence between coset elements $w\in\Wf\backslash \Wa/ \Wf$ and dominant
roots $\bar{\alpha}_{w}\in P_+\cap\ol{Q}$. More generally, if the
embeddings are chosen differently, this correspondence is between
coset elements $w\in\Wf\backslash \Wa/ \Wf'$ and dominant weights
$\lambdab\in P_+\cap\ol{Q}+\Lambda_i$ where
$\Lambda_i$ is the fundamental weight that determines the embedding:
$\Wf'$ is generated by the set of simple reflections $s_j$ for
$j\neq i$.
Then the result of~\cite{Lus3} states that
\be\label{eq:lus1}
P_{w,w'}^{IJ}(1)=\dim L(\lambdab_{w'})_{\lambdab_w}
\ee
where $L(\lambdab)_{\mub}$ is the weight space of weight $\mub$ in the
(finite dimensional) irreducible $\gf$ module of highest
weight $\lambdab$. \\
This result applies in particular to the $\W_N$ algebras with
$c=\mbox{rank}({\gf})$, {\it i.e}.\ with $\kh=1$. From~(\ref{eq:adm}) it
follows that the dominant weight is determined by a level 1 weight,
{\it i.e}.\ $\lambda+\rho=\Lambda_i$. $W_{\lambda}^r=\Wf$ and
$W_{\lambda}^0=\Wf'$. Using the correspondence described
above, to every primitive weight $\mu=w.\lambda$ one
associates the dominant weight $\lambdab_w=\mub+\rhob$. Then the sum
$w'\geq w$ can be rewritten as follows
\be\label{eq:karc2}
\ch M^r(\mu) = \sum_{\mub'\geq\mub} \dim
L(\mub'+\rhob)_{\mub+\rhob} \ch L^r(\mu').
\ee
Inverting this (using the basis transformation from Verma modules
$M(\mub)$ to singlets $e^{\mub}$) gives
\be
\label{eq:dblhexc2}
\ch L^r(\mu) = \sum_{x\in\Wf} \epsilon_x \ch M^r(\mu+\rho-x(\rho))
\ee
This character formula was first proposed in~\cite{MMMO}, where it was
obtained as a limit of the characters of the $\W_N$ minimal
models~\cite{FL}. The inverse formula~(\ref{eq:karc2}) was obtained
for $\W_3$ in~\cite{BMP}, using an explicit construction of the
singular vectors in the Fock space and comparison of the characters for each side.
Again, this character formula applies more generally, in particular to
the weights in a corner of the $(p,p')$ Kac table (in that case one
replaces $\rho\rightarrow p\rho$ everywhere on the RHS
of~(\ref{eq:dblhexc2}))
\section{Concluding remarks}
In this paper we have formulated the KL conjecture for $\W$ algebras
associated with arbitrary $sl_2$ reductions. The result can be
described in terms of a double coset $W_{\lambda}^r\backslash
W_{\lambda}/W_{\lambda}^0$: the Hasse diagram gives the embedding
diagram of the Verma modules, and the KL polynomials give the
multiplicities in the characters.
The conjectures also apply to finite $\W$ algebras, which
are the Hamiltonian reduction of simple Lie algebras $\gf$~\cite{BT}.
In that case one simply takes $W$ to be the Weyl group $\Wf$ of $\gf$.
The character formulas for this class of algebras are given in~\cite{DV}
(for regular integral weights only) and the results agree
completely.
We remark that the conjecture is also a useful tool to analyse the
structure of the Verma modules in more detail. This is particularly
important if one attempts to construct a resolution of the irreducible
modules by Verma modules. The physical motivation for doing so is the
application to $\W$ gravity/strings: it is much simpler to compute the
(string) BRST cohomology on Verma modules than on irreducible modules.
The problem is that it is not always possible to find a resolution by
Verma modules. For instance, for the $\W_3$ string at $c=2$ it is found by
explicit construction~\cite{BMP} that there is no resolution by Verma
modules, but instead one is forced to introduce generalized Verma
modules. This is directly linked to the existence of primitive vectors
that are pseudo-singular rather than singular (they are not an
eigenvector of $W_0$). For such an analysis it is convenient to have
the data presented by the KL conjecture. This way for instance, one
can easily show that the character~(\ref{eq:dblhex}) of the (p,1) topological
minimal models for $\W_3$ cannot be reproduced by a resolution by
(generalized) Verma modules. This is due to the occurence of subsingular
vectors~\cite{DV}. It would be very interesting to know what type of
modules are needed to build such a resolution, since these modules are
going to carry the cohomology of the topological $\W_3$ string.
Work on this is in progress.
\renewcommand{\theequation}{A.\arabic{equation}}
\setcounter{equation}{0}
\section*{Appendices}
\subsection*{KL polynomials on Coxeter groups}
In this appendix, we summarize the definition and some of the properties of KL
polynomials $P_{x,y}$ for a Coxeter group $W$, for details
see~\cite{KL1,Hu}.
The starting point is the Hecke algebra ${\cal H}$ with generators
$T_y$ (one for each $y \in W$) and defining relations
\bea\label{eq:hecke}
T_x T_{y} = T_{xy} & \mbox{if} & \ell(xy) = \ell(x) + \ell(y) \\
(T_s + 1)(T_s-q) = 0 & \mbox{if} & s \in S
\eea
where $S$ is the set of simple reflections that generate $W$. The
elements $T_y$ are invertible in ${\cal H}$, and one can write
\be\label{eq:hecke-inverse}
T_{y^{-1}}^{-1} = \sum_{x \leq y} \ve_x \ve_y R_{x,y}(q) q^{-\ell(y)} T_x
\ee
where $\ve_y = (-1)^{\ell(y)}$, and $R_{x,y}(q)$ is a polynomial
in $q$ of degree $\ell(y)-\ell(x)$ for $x\leq y$, uniquely defined
by~(\ref{eq:hecke-inverse}). The map $\imath$ defined by
\be\label{eq:hecke-auto}
\imath (q) = q^{-1},\qquad \imath (T_y) = T^{-1}_{y^{-1}}
\ee
is an automorphism of ${\cal H}$. The KL polynomials are associated
with the invariants of $\imath$. For any pair $x \leq y$ in $W$, there is a
uniquely defined polynomial $P_{x,y}$ of degree $\leq (\ell(y)
-\ell(x)-1)/2$ if $x < y$, and $P_{x,x}=1$, such that
\be
C_y = \sum_{x \leq y} \ve_x \ve_y q^{\ell(y)/2-\ell(x)}P_{x,y}(q^{-1})T_x
\ee
satisfies
\be
\imath(C_y) = C_y
\ee
for all $y \in W$. Equivalently, the $P$-polynomials satisfy
\be
q^{\ell(y)-\ell(x)} P_{x,y}(q^{-1}) = \sum_{x \leq z \leq y} R_{x,z}
P_{z,y}(q) \quad \mbox{for all $x \leq y$}
\ee
From this, one can extract a recursion relation (expressing the
polynomials $P_{x,y}$ in terms of the polynomials $P_{x',y'}$ with
$y'<y$). Namely, for $ys>y$ one has~\cite{KL1}
\be\label{eq:recrel}
P_{x,ys}=q^{1-c}P_{xs,y}+q^cP_{x,y}-q\sum_{\stackrel{x\leq
z<y}{zs<z}} P_{x,z}\slash{\!\!\!\!P}_{z,y}
\ee
Here $c=1$ if $xs<x$ and $0$ otherwise, and $\slash{\!\!\!\!P}_{z,y}$ is the
term in $P_{z,y}$ of (maximal) degree $\frac{1}{2}(\ell(y)-\ell(z)-1)$.
The initial values of the recursion relation are
$P_{x,e}(q)=\delta_{x,e}$. This implies in particular that
$P_{x,y}(q)=0$ unless $x\leq y$. From~(\ref{eq:recrel}) it also
follows that $P_{x,y}(0)=1$ if $x\leq y$. In the case crystallographic
Coxeter groups (which includes (affine) Weyl groups)
the coefficients of $P_{x,y}$ give the dimensions of stalks of
cohomology sheaves of the intersection cohomology complexes associated
to Schubert varieties~\cite{KL2}. This implies in particular that
these coefficients are nonnegative integers.
Similarly, if $ys<y$ it can be shown that $C_y T_s = -C_y$ which
implies that
\be\label{eq:pshift}
P_{x,y} = P_{xs,y} \quad \mbox{for $x \leq y$ and $ys<y$}
\ee
For finite Coxeter groups (where there is a unique longest element
$w_0$) it easily follows that
\be
P_{x,w_0}=1
\ee
\subsection*{Inverse KL polynomials on Coxeter groups}
The KL polynomials $P_{y,w}$ form an upper triangular
matrix with 1's on the main diagonal, which naturally can be inverted.
Thus, one can define for each $x \leq y$ in $W$ a polynomial $Q_{x,y}$
such that
\be\label{eq:invdef}
\sum_{x \leq z \leq y} \ve_x \ve_z P_{x,z} Q_{z,y} =
\delta_{x,y} \quad \mbox{for all $x \leq y$}
\ee
It is clear that $Q_{x,x}(q)=1$ and that $Q_{x,y}(q)$ has degree $\leq
(\ell(y)-\ell(x)-1)/2$ for $x<y$, and $\slash{\!\!\!\!Q}_{x,y} =
\slash{\!\!\!\!P}_{x,y}$. It also follows that
\be
q^{\ell(y)-\ell(x)} Q_{x,y}(q^{-1}) = \sum_{x \leq z \leq y} Q_{x,z}
R_{z,y}(q) \quad \mbox{for all $x \leq y$}
\ee
The $Q$-polynomials are associated to invariants of $\imath$. Define
elements $S_x,D_x$ of ${\cal H}^*$ by
\be
\langle S_x, \imath(T_y) \rangle = \langle D_x , C_y \rangle = \delta_{x,y^{-1}}
\ee
and let $\langle \imath(u), h \rangle = \imath( \langle u, \imath(h)
\rangle)$. It follows that $\imath (D_x) = D_x$ and
\be
D_x = \sum_{x \geq y} q^{\ell(x)/2 - \ell(y)} Q_{x,y} S_y
\ee
with $Q_{x,y}$ the inverse polynomials~(\ref{eq:invdef}). From the
right ${\cal H}$-action on ${\cal H}^*$ given by
$\langle u \cdot h,h' \rangle = \langle u, h h' \rangle$
one can conclude that $D_x \cdot T_s = q D_x$ if $xs>x$. This implies
\be\label{eq:qshift}
Q_{x,ys} = Q_{x,y} \quad \mbox{for $x \leq y$ and $xs>x$}
\ee
From this it easily follows that
\be\label{eq:nulfix}
Q_{e,y} = 1 \quad \mbox{for all $y \in W$}
\ee
The analogue of~(\ref{eq:recrel}) for the $Q$-polynomials is (for $xs<x$)
\be\label{eq:qoldrec}
Q_{xs,y}=q^{1-c}Q_{x,ys}+q^cQ_{x,y}-q\sum_{\stackrel{x <
z \leq y}{zs>z}} \slash{\!\!\!\!Q}_{x,z} Q_{z,y}
\ee
with $c=1$ for $ys>y$ and $c=0$ for $ys<y$.
In this form~(\ref{eq:qoldrec}) cannot be used to solve for the $Q$-polynomials
(at least not in the case of infinite Coxeter groups). But,
combining~(\ref{eq:qoldrec}) and~(\ref{eq:qshift}) for $ys>y$, one
gets a useful relation (expressing $Q_{z,w}$ in terms of $Q_{z',w'}$
with $w'<w$)
\be\label{eq:qrecrel}
Q_{x,ys} = c Q_{xs,y} + (-q)^c Q_{x,y} + c q \sum_{\stackrel{x <
z \leq y}{zs>z}} \slash{\!\!\!\!Q}_{x,z} Q_{z,y}
\ee
with $c=1$ for $xs<x$ and $c=0$ for $xs>x$.
In the case of (affine) Weyl groups the coefficients of $Q_{x,y}$ are again
nonnegative integers. \\
In the case of a finite Coxeter group, the KL polynomials $P$ are
related to the inverse polynomials $Q$ through
\be
Q_{x,y} = P_{w_0y,w_0x}
\ee
On the other hand: if the group is not finite, $P$ and $Q$ do not
appear to be related in any such way.
\subsection*{KL polynomials on cosets}
Similar as for ordinary Coxeter groups $W$, one can also construct KL
polynomials on cosets $W/W_I$ (or $W_I\backslash W$), where
$W_I$ is a parabolic subgroup of $W$~\cite{Deo}. In particular one
finds that again there are recursion relation for the associated
polynomials $P^I_{x,y}$ and $Q^I_{x,y}$. If the subgroup $W_I$ is
finite (which is sufficient for our purposes) the polynomials
$P^I,Q^I$ and their inverses $\tilde{Q}^I,\tilde{P}^I$ defined
through\footnote{Note that this notation is slighly different from the
notation used in~\cite{DV}.}
\be
\sum_{x \leq z \leq y} \tilde{Q}^{I}_{x,z} P^{I}_{z,y} = \sum_{x \leq
z \leq y} Q^{I}_{x,z} \tilde{P}^{I}_{z,y} = \delta_{x,y}
\ee
can be expressed in terms of the KL polynomials of $W$
\be\label{eq:oneside2}
\begin{array}{ll}
P^{I}_{x,y} = P_{\xmax,\ymax} &
Q^{I}_{x,y} = Q_{\xmin,\ymin} \\
\tilde{P}^{I}_{x,y} = \sum_{z \in [x]} P_{z,\ymin} \ve_z \ve_{\ymin}\qquad &
\tilde{Q}^{I}_{x,y} = \sum_{z \in [y]} Q_{\xmax,z} \ve_{\xmax} \ve_z
\end{array}
\ee
Here $\zmin$ is the minimal- and
$\zmax$ is the maximal representative of the coset $[z]$ of $z$.
However, the cosets that play a role in this paper are two-sided cosets
$W_I\backslash W / W_J$, with respect to parabolic subgroups
$W_I$ and $W_J$. There does not appear to be an abstract
set-up for double-sided cosets. In particular, the partial ordering on
these cosets is more complicated than in the case of one-sided cosets
(for example, the length of adjacent elements may differ by more than
1). So instead of defining KL polynomials through a recursuion
relation we take~(\ref{eq:oneside2}) as our starting point, {\it i.e}.\ we
define
\be
\begin{array}{ll}
P^{IJ}_{w,w'} = P_{\wmax,\wmax'} &
Q^{IJ}_{w,w'} = Q_{\wmin,\wmin'} \\
\tilde{P}^{IJ}_{w,w'} = \sum_{x \in [w]} P_{x,\wmin'} \ve_x \ve_{\wmin'} \qquad &
\tilde{Q}^{IJ}_{w,w'} = \sum_{x \in [w']} Q_{\wmax,x} \ve_{\wmax} \ve_x
\end{array}
\ee
where the polynomials $\tilde{P}^{IJ},\tilde{Q}^{IJ}$ are again the
inverse polynomials
\be
\sum_{x \leq z \leq y} \tilde{Q}^{IJ}_{x,z} P^{IJ}_{z,y} = \sum_{x
\leq z \leq y} Q^{IJ}_{x,z} \tilde{P}^{IJ}_{z,y} = \delta_{x,y}
\ee
\section{Tables}
Table 3:
{\sl
KL polynomials $P_{x,y}$ for the Weyl group $\hat{a}_2$ of the affine KM
algebra $g=sl_3$, up to $\ell(y)=15$. To find $P_{x,y}$ for arbitrary
pairs $x,y$ (with $\ell(y)\leq 15$) use that\\
1. $P_{x,y}=0$ unless $x\leq y$, \\
2. $P_{x,y}=P_{x^{-1},y^{-1}}$, \\
3. $P_{x,y}=P_{\tau(x),\tau(y)}$ with $\tau$ an automorphism of the
Dynkin diagram, \\
4. $P_{x,y}=P_{x',y}$ for $x\leq x'$ and $P_{x',y}(1)$ maximal, \\
5. if 1-4 does not apply then $P_{x,y}=1$.\\
So, given a pair $x,y$ with $x\leq y$, one first fixes i,j,k and an
order ({\it i.e}.\ reading from left-to-right or from right-to-left) such that
$y$ is in the table. Secondly, one searches for an $x'$ (in the fixed
order) in the table such that $x\leq x'$ and $P_{x',y}(1)$ is maximal;
then $P_{x,y}=P_{x',y}$. If either of the two steps fail then
$P_{x,y}=1$.
}\\[2mm]
Table 4:
{\sl
Inverse KL polynomials $Q_{x,y}$ for the Weyl group $\hat{a}_2$ of the affine KM
algebra $g=sl_3$, up to $\ell(x)=14$. To find $Q_{x,y}$ for arbitrary
pairs $x,y$ (with $\ell(x)\leq 14$) use the rules of table 3. with x,y
interchanged and the ordering reversed.
}
\newpage
\epsfxsize = 14 cm
\epsfbox[100 -100 500 800]{t1.ps}
\newpage
\epsfxsize = 13.5 cm
\epsfbox[100 -100 500 800]{t2.ps}
\newpage
\epsfxsize = 14 cm
\epsfbox[50 -100 450 800]{t3.ps}
\newpage
\epsfxsize = 14 cm
\epsfbox[50 -100 450 800]{t4.ps}
\newpage
\epsfxsize = 14 cm
\epsfbox[50 -100 450 800]{t5.ps}
\newpage
\hoffset = 0 cm
\section*{Acknowledgements}
Discussions with P.\ Bouwknegt, W.\ Lerche, K.\ Pilch, W.\ Soergel, G. Zuckerman
and especially M.\ Goresky are gratefully acknowledged. KdV thanks the
Dept. of Applied Mathematics in Durham for hospitality and support. \\
KdV is supported by the EC Human Capital and Mobility Program.
|
\section{Introduction} \label{sec-intro}
In spite of great progress in the study of the comic x-ray background (CXB)
radiation achieved since its discovery more than 30 years ago (Giacconi et
al. 1962) its origin remains unclear. In the soft x-ray band, deep ROSAT
source counts show that the population of faint x-ray sources (of them some
70\% are QSOs, Boyle et al. 1993) is able to account for the main portion
(if not 100\%) of the soft x-ray (1-2 keV) background, while the
contribution of a truly diffuse component (such as the hot intergalactic
gas) must be less than 25\% at 90\% confidence (Hasinger et al. 1993).
At harder x-rays, 3-60 keV, sensitive imaging surveys have not yet been
performed, and the main efforts have been concentrated on explaining the CXB
spectrum which is remarkably well fit by the thermal bremsstrahlung model
with $kT=40$ keV (Marshall et al. 1980). The COBE limit on the
Comptonization parameter excludes the possibility of a uniform hot medium
producing more than a few percent of the CXB in the harder x-ray band
(Mather et al. 1990). Hence, most likely the hard x-ray background is also
composed by unresolved compact sources (see the discussion of the ``spectral
coincidence'' and the contribution of point sources to the CXB in Giacconi
\& Zamorani 1987).
The main difficulty with the discrete source explanation of the hard x-ray
background has been that there is no known source population whose spectra
resemble that of the CXB. Clusters of galaxies, which are as numerous as
AGNs at high fluxes in the standard x-ray band (Piccinotti et al. 1982),
typically have thermal bremsstrahlung spectra with temperatures of about 6
keV (David et al. 1993), and exhibit negative cosmological evolution (e.g.,
Gioia et al. 1990). Hence, they cannot be significant contributors to the
broad-band x-ray background. Population of AGN's which dominate the soft
x-ray source counts and exhibit positive cosmological evolution have 2--20
keV spectra significantly steeper than the x-ray background. HEAO
A-1/EXOSAT/Ginga observations of low luminosity AGNs (mostly Seyfert 1
galaxies) revealed a ``canonical'' $\alpha=0.7$ spectrum (Mushotzky 1984),
to be compared with the energy index of 0.4 for the CXB in this energy band.
Both soft and hard x-ray observations of high luminosity AGNs (mostly
quasars) suggested that there was no universal power law for this class of
sources. Observed spectral indices are distributed from $\alpha \sim 0$ to
$\alpha \sim 2.0$, with the mean value of about 1 in the soft x-rays and
0.8-0.9 in the 2-10 keV x-ray band (Williams et al. 1992, Comastri et al.
1992, and Wilkes \& Elvis 1987). The average spectral index of quasars also
corresponds to a spectrum which is significantly softer than that of the
CXB, which was used to argue that their contribution to the hard x-ray
background could not exceed $\sim 10$\% (Fabian, Canizares, \& Barcons
1989).
In this paper we show that it is still possible to make a significant
portion of the hard x-ray background from QSOs, i.e. that the observed
composite spectrum of quasars can be as hard as the CXB spectrum. The basic
idea relies on the observed broad spread of QSO spectral indices. X-rays
observed from QSOs at energy $E$ are emitted at higher energies, $(1+z)E$.
Therefore, distant objects with flat spectra appear brighter than those with
steep spectra, and hence, it follows that the mean spectrum hardens with
increasing redshift even though no true spectral evolution occurs. A similar
effect arises if soft x-ray luminosities do not correlate with spectral
properties. The absence of such a correlation implies that hard spectrum
sources should be on average more luminous in the 2-10 keV x-ray band than
the steep spectrum sources. Futhermore, the broad distribution of spectral
indices provides a natural explanation for the mismatch between the results
of soft and hard x-ray source counts found from the comparison of
Einstein/ROSAT and HEAO A-1/Ginga surveys. Since hard sources are brighter
at high redshifts than soft ones, a difference in number-flux relations
determined at, for example, 1 keV and 10 keV is expected, with larger
surface number density at the hard energy band.
In section \ref{sec-chatter} we argue that it is necessary to account for
the contribution of QSOs in source models of the x-ray background. Models of
the QSO population (spectra, luminosity function, and the cosmological
evolution) used in the paper are described in section \ref{sec-models}.
Basic equations used for calculations of the composite QSO spectrum and the
number-flux relations in different energy bands are derived in section
\ref{sec-formulas}. Section \ref{sec-results} presents the results of
calculations. The main results of the paper are briefly discussed in section
\ref{sec-discussion}.
\section {QSOs, Seyfert Is, and the CXB} \label{sec-chatter}
The difficulties with the explanation of the CXB spectrum by AGNs arise from
the assumption that their spectra are simple power laws over a broad energy
range. Schwartz \& Tucker (1988) have shown that if AGN spectra flatten with
increasing energy up to some energy cutoff, these difficulties can be
overcome and it is possible to match the CXB spectrum by the composite
spectrum of unresolved distant AGNs. In fact, the spectral flattening was
observed in Ginga spectra of Seyfert 1 galaxies and was interpreted as
either the partial covering by cold material of a power law x-ray spectrum
or as reprocessed emission from cold material (Piro et al. 1989, Matsuoka et
al. 1990, Pounds et al. 1990). GRANAT and OSSE observations of a sample of
Seyfert 1s revealed that their spectra do steepen again at energies of about
50-100 keV (Jourdain et al. 1992, Johnson et al. 1994). Thus the observed
spectra of Seyfert 1s are quite close to what is required by the model of
Schwartz \& Tucker (1988) -- the source spectra flatten above $\sim 10$ keV
and they steepen again at about 100 keV. This stimulated Zdziarski et al.
(1993) to claim that it is AGNs with Seyfert~1--like spectra which comprise
the 2--100 keV x-ray background. In fact, using the cosmological evolution
model of Boyle et al. (1993) and the model spectrum determined from
Ginga/OSSE observations of Seyfert 1s Zdziarski et al. (1993) obtained an
excellent fit to the CXB spectrum in the 2-100 keV energy range.
However, the explanation of the CXB by Seyfert 1 galaxies may fail because,
as discussed by Giacconi \& Zamorani (1987), the contribution of known
classes of steep spectrum sources must also be included in modeling the CXB
spectrum such as quasars and clusters of galaxies. Giacconi \& Zamorani
showed that if the contribution of the $\alpha \sim 0.7-1$ sources to the
soft x-ray background amounts to at least 50\% of the CXB, the residual CXB
spectrum is much flatter than the $kT=40$ keV spectrum. Therefore, it may
be difficult to fit the residual CXB spectrum by spectra of Seyfert 1
galaxies.
Deep, soft surveys show that the contribution of QSO's to the CXB is in fact
significant. ROSAT deep surveys show that point-like faint sources
contribute the bulk of the 2 keV CXB, with the directly resolved fraction
being about 60\%. From fluctuation analyses, point sources contribute about
75\% of the CXB and a reasonable extrapolation of the number-flux relation
to even fainter fluxes can account for 100\% of the soft x-ray background
(Hasinger et al. 1993). In a deep ROSAT survey, Boyle et al. (1993)
identify 64\% of the extragalactic x-ray sources with QSOs and this number
can be as high as $\sim 80$\%. Hence, it seems very likely that QSOs
contribute not less 50\% of the 2 keV CXB, and consequently their
contribution cannot be overlooked in any model of the CXB.
QSO spectra are generally less well studied than those of Seyfert 1
galaxies. Both soft and hard x-ray observations show that spectral indices
of quasars are distributed over a broad range of values, from $\alpha \sim
0$ to $\alpha \sim 2$, and the mean spectral index in the 2-10 keV band is
0.8-0.9 (Comastri et al. 1992, Williams et al. 1992). In their review of
x-ray observations of AGNs, Mushotzky, Done \& Pounds (1993) discuss the
difference between 2-10 keV spectra of Seyfert 1 galaxies and quasars.
Ginga spectra show that the radio-loud QSOs do not require additional
reflection components and, apart from 3C273 and PHL 1657, show no deviations
from a simple power law over the 2--20 keV energy band (Williams et al.
1992). The radio-quiet QSO spectra do show deviations from the power law but
the steeper underlying spectrum ($\alpha \approx 1.0$) implies a real
difference between radio-quiet QSOs and Seyfert 1 galaxies. We conclude
that QSOs are different from Seyfert Is from the point of view of x-ray
spectra and show less evidence for spectral flattening above 10 keV.
Therefore, it is necessary to subtract the contribution of QSOs to the x-ray
background before fitting the CXB spectrum by spectra of Seyfert Is. If we
assume that QSOs all have the $\alpha=0.8-0.9$ spectrum and contribute about
50\% to the 2 keV background, the spectrum of the residual background will
be much harder than the $kT=40$ keV spectrum (Giacconi \& Zamorani 1987),
which will be difficult to fit even with spectra of Seyfert Is with
significant reflection contributions.
\section{Models of The Source Population} \label{sec-models}
We consider four models of the source population to reconstruct the CXB
spectrum and to illustrate properties of the reconstruction -- ``realistic''
(model {\bf A}), ``reduced realistic'' ({\bf B}), ``pure'' ({\bf C}), and
``common'' ({\bf D}). Model spectra are single power laws modified at the
soft and the hard energies to describe a soft excess and a high energy
cutoff:
\begin{eqnarray}
F(E) & = & \left\{\begin{array}{ll}
a_sE^{-\alpha_s} & \mbox{$E<E_s$}\\
a_mE^{-\alpha_m} & \mbox{$E_s<E<E_h$}\\
a_hE^{-\alpha_h} & \mbox{$E>E_h$}
\end{array}
\right.
\end{eqnarray}
In the standard x-ray band (2-10 keV) we assume that sources have power law
spectra, with no additional features such as strong emission lines,
absorption edges, or a reflection component, and have spectral indices
distributed according to a Gaussian distribution with the mean value of 0.8
and the dispersion of 0.35. ``Unphysical'' energy spectra with negative
spectral indices are prohibited in the model by a cutoff of the distribution
at zero. The choice of parameters is based on the results of {\em EXOSAT}\/
and {\em Ginga}\/ measurements of spectra of subsamples of soft x-ray
selected high luminosity AGN (Comastri et al. 1992, Williams et al. 1992).
Both sets exhibit a steep 2-10 keV mean spectral index (0.81 for {\em
Ginga}\/ and 0.89 for {\em EXOSAT}\/ data), with a wide range of spectral
indices, distributed from 0.08 to 1.26, with an intrinsic dispersion of
about 0.3. There is evidence of a soft excess in QSO energy spectra below
an energy of a few keV (Wilkes \& Elvis 1987, Comastri et al. 1992, Fiore
et al. 1994), manifesting itself as a gradual steepening of source spectra
in softer energy bands. Hence, we introduce the soft excess to the model
spectra as an additional power law below 1 keV, with the mean soft x-ray
spectral index of 1.4. The choice of a mean soft x-ray spectral index is
motivated by ROSAT measurements of bright QSO spectra in the energy band of
0.1-2.4 keV (Laor et al. 1994).
The bulk of sources in the model have spectral indices less than 1, which
yields infinite broad band luminosities. Therefore, the break in spectra at
high energies is required. We introduce such a break at around 100 keV by
steepening the spectral index up to the value of 2.0. The cutoff energy of
100 keV is chosen so that it is possible to reproduce the CXB spectrum up to
$\sim 100$ keV, rather that this choice is motivated by the observations of
QSO spectra.
Parameters of the soft excess and the high energy cutoff are distributed
according to a Gaussian distribution with a dispersion equal to 0.3 of the
mean value. For the high energy cutoff, the extremely low values of the
break energy (below 30 keV) and flat spectra (with the hard spectral index
less than the main spectral index) are prohibited.
The four models differ by the distribution of spectral parameters. In the
``realistic'' model {\bf A} all the spectral parameters are characterized by
Gaussians. The ``reduced realistic'' model {\bf B} is used to show how the
range in the soft excess and the high energy cutoff parameters influence the
composite spectrum; the soft excess parameters are held fixed at a value of
1.4 for the spectral index and 1 keV for the break energy, and the high
energy cutoff parameters at values of 2.0 for the spectral index and 100 keV
for the break energy. The ``pure'' model {\bf C} is used to demonstrate the
effects of the soft excess and the high energy cutoff and is characterized
by single power law spectra over the whole energy range with a distribution
of spectral indices analagous to that in the model {\bf A}. Finally, in the
``common'' model {\bf D} all parameters are held fixed which corresponds to
the assumption of an identical spectrum for all sources.
We assume further that there is no correlation between the spectral
properties and the luminosity in the soft x-ray band. This would of course
imply the existence of a negative correlation between the luminosity in the
harder x-ray band (2-10 keV) and the spectral index, i.e. hard sources
must be on average more luminous in this band. Existing x-ray data do not
enable one to explore the existence of such a correlation with high
confidence, since the 2-10 keV x-ray spectra are measured only for the
several dozens of the brightest quasars. However, Williams et al. (1992),
found a negative correlation between the x-ray luminosity and 2-10 keV
spectral index at more than 90\% significance in their sample of 13
high-luminosity AGNs, while Wilkes \& Elvis (1987) find no correlation
between the x-ray luminosities and spectral indices measured in the 0.3-3.5
keV band. Finally, recent ROSAT measurements indicate that quasar intensities
are much less scattered at around 0.3 keV than at higher energies (Laor et
al. 1994).
For the model luminosity function we use the QSO soft x-ray luminosity
function derived by Boyle et al. (1993) from the joint analysis of the deep
ROSAT survey, and the EMSS luminosity function. At $z=0$, this luminosity
function can be represented as two power laws, with a break at
$L^*_X=10^{43.9}$ erg s$^{-1}$ in the 0.3-3.5 keV energy band,
$\Phi(L_X)\propto L_X^{-3.4}$ above the break, and $\Phi(L_X)\propto
L_X^{-1.7}$ below the break. Boyle et al. (1993) also derive the
cosmological evolution of QSOs which can be described as pure luminosity
evolution, $L_X \propto (1+z)^K$, with $K=2.8$ for $\Omega=0$ and $K=2.5$
for $\Omega=1$ universe (Boyle et al. 1993). For $\Omega=1$ a ``cut off''
in the evolution is required at around $z=2$. We use the same model of
cosmological evolution for the values of $\Omega$ ranging from $\Omega=0$
and $\Omega=1$ in our calculations, with $K=2.7$, and cutoff of evolution at
$z=3.0$.
It must be emphasized that no spectral evolution is present in calculations.
This means that, by assumption, the distribution of spectral parameters is
the same at every redshift.
\section{Basic Equations} \label{sec-formulas}
If at the source rest-frame the source spectrum is given as $A\,S_0(E)$, and
the source lies
at a redshift $z$, then the observed energy flux density (flux per unit
energy interval) is:
\begin{eqnarray} \label{eq-flux}
f(E) & = & \frac{A\,S_0(E(1+z))}{4\pi(a_0 r)^2 (1+z)},
\end{eqnarray}
where $r(z,\Omega)$ is the angular size distance,
\begin{equation} \label{eq-angsize}
a_0\, r(z,\Omega)\, (H_0/c) = \frac{2}{\Omega^2(1+z)}\;\{\Omega z +
(\Omega-2)\,[\sqrt{1+\Omega z} -1]\}
\end{equation}
(Mattig 1958). For power law spectra,
$S_0(E)=E^{-\alpha}$, eq. \ref{eq-flux} becomes:
\begin{eqnarray}
f(E) & = & \frac{A\,E^{-\alpha}\,(1+z)^{-\alpha-1}}{4\pi(a_0r)^2}.
\end{eqnarray}
If sources are distributed homogenously in space and there is no density
evolution, the number of sources per unit redshift increment is
\begin{eqnarray} \label{eq-dn}
\frac{dN}{dz} & = & n_0 \;\frac{4\pi c^3}{H_0^3}\; z^2 \frac {4[\Omega z +
(\Omega-2)(\sqrt{1+\Omega z}-1)]^2} {\Omega^4 z^2 (1+z)^3\sqrt{1+\Omega z}}
\;\; =\;\; n_0 \;\frac{4\pi c^3}{H_0^3}\; z^2\; \xi(z,\Omega)
\end{eqnarray}
where $n_0$ is the volume number density at z=0 (Mattig 1958).
Einstein and ROSAT observations indicate that the observed properties of the
high luminosity AGN population require a cosmological evolution which can
be described in terms of pure luminosity evolution, i.e. the luminosity
scales with redshift as $L_x(z)=L_x(0)\,C(z)$, with $C(z)=(1+z)^K$ (Boyle et
al. 1993). For this type of evolution the $z$-dependence of the luminosity
function is:
\begin{eqnarray} \label{eq-lfunctionz}
\Phi(L_x,z) & = & \frac{1}{C(z)}\;\Phi\left(\frac{L_x}{C(z)},0\right)
\end{eqnarray}
Available x-ray data on the luminosity function and its evolution is derived
from observations in the 0.3-3.5 keV energy band. For simplicity, we assume
that the the same relations apply to 1 keV luminosities; $L_1$ is denoted as
the 1 keV luminosity in the source rest-frame.
Using equations \ref{eq-flux}-\ref{eq-lfunctionz} we calculate the composite
energy spectrum of the source population, and the results of source counts
as a function of energy band.
The spectral parameter $\alpha$ is used to
designate different spectral shapes. For a class of power-law spectra
$\alpha$ is simply the power-law index; for more complicated spectral shapes
$\alpha$ becomes a vector of parameters, and integrals over the range of
$\alpha$ indicate integrals over different spectral shapes. The assumed
absence of any correlation of spectral properties with redshift or with
soft x-ray luminosity (section \ref{sec-models}) means that the density
distribution of the spectral
parameter, $\alpha$, is a function of $\alpha$ only, i.e. the probability
that $\alpha$ is within the interval $(\alpha,\alpha+d\alpha)$ is
\begin{eqnarray}
dP & = & W(\alpha)\,d\alpha
\end{eqnarray}
Different spectral shapes are denoted as $S_0(E,\alpha)$, with this function
normalized to unity at 1 keV, so that the energy flux density at a
particular energy $E_0$ is given by $L_1 S_0(E_0,\alpha)$, where $L_1$ is
the 1~keV source luminosity.
\subsection {Composite Spectrum}
In this section we calculate the composite spectrum of sources uniformly
distributed in space, with spectral properties described above, and pure
luminosity evolution.
Let us first find the mean energy spectrum of sources located at a redshift
$z$. This can be evaluated by integration of equation \ref{eq-flux} over the
appropriate range of luminosities and spectral shapes:
\begin{eqnarray} \label{eq-cxbspec-z1}
F_{z}(E) & = & \int\limits_0^\infty \Phi(L_1,z)\,dL_1\;\; L_1
\int\limits_{\alpha}W(\alpha)\,d\alpha\;
\frac{S_0(E(1+z),\alpha)}{4\pi\,(a_0r)^2\;(1+z)}
\end{eqnarray}
For pure luminosity evolution the luminosity function changes with redshift
following eq.\ref{eq-lfunctionz}, and the evaluation of the integral over
$L_1$ in eq.\ref{eq-cxbspec-z1} yields:
\begin{eqnarray} \label{eq-cxbspec-z}
F_{z}(E) & = & \overline{L}_1\;\frac{C(z)}
{4\pi\,(a_0r)^2\;(1+z)}\;
\int\limits_{\alpha}W(\alpha)\,d\alpha\;
S_0(E(1+z),\alpha)
\end{eqnarray}
where $\overline{L}_1$ is the mean present day soft x-ray luminosity and
$C(z)$ is the cosmological luminosity evolution function.
It is easy to show that equation \ref{eq-cxbspec-z} implies that the
composite spectrum of sources at a given $z$ hardens with increasing
redshift. Let the source spectra be described as a simple power law,
i.e. $S_0(E,\alpha)=E^{-\alpha}$. Then equation \ref{eq-cxbspec-z} becomes
\begin{eqnarray}
F_{z}(E) & \propto & \int\limits_{\alpha} W(\alpha)\,d\alpha\, E^{-\alpha}
\,(1+z)^{-\alpha}
\end{eqnarray}
The spectral index of the composite spectrum (as a function of energy) is
the logarithmic derivative of the spectral flux:
\begin{eqnarray}
\overline{\alpha} & = & - \frac {d\,\log F_z(E)}{d\,\log E}
= - \frac{E}{F_z(E)}\;\frac{d\,F_z(E)}{d\,E} \nonumber \\
& = & \frac
{
\int\limits_{\alpha} \alpha\,W(\alpha)\,E^{-\alpha}\,(1+z)^{-\alpha}\,d\alpha
}
{
\int\limits_{\alpha} W(\alpha)\,E^{-\alpha}\,(1+z)^{-\alpha}\,d\alpha
},
\end{eqnarray}
i.e. the initial distribution of spectral indices, $W(\alpha)$, must be
replaced by $W(\alpha)\,E^{-\alpha}\,(1+z)^{-\alpha}$ when one observes
sources at a redshift $z$ and an energy $E$. Thus, the distribution is
weighted towards flatter spectra at higher energies and redshifts.
On multiplying eq.\ref{eq-cxbspec-z} by $dN$, the number of sources in the
redshift interval $(z,z+dz)$ (eq. \ref{eq-dn}), and integrating over
redshift, we find the expression for the composite spectrum of the source
population:
\begin{eqnarray} \label{eq-cxbspec}
F(E) & = & \left(\frac{c}{H_0}\right)^3\; n_0\, \overline{L}_1 \;
\int\limits_0^\infty dz \;
\frac{\xi(z,\Omega)\; z^2}{(a_0r)^2\;(1+z)}\; C(z) \;
\int\limits_{\alpha} W(\alpha)\,d\alpha\, S_0(E(1+z),\alpha).
\end{eqnarray}
This equation is used in section \ref{sec-results-spec} for calculations of
the composite spectrum of QSOs, using models {\bf A}-{\bf D} for the spectral
parameter distributions.
\subsection {Source counts at different energies}
The 1 keV luminosity of a source whose energy flux density observed at the
energy $E_0$ is $f$ can be calculated (cf. eq\ref{eq-flux}) as:
\begin{eqnarray} \label{eq-L1}
L_1 & = & \frac{4\pi\;(ar_0)^2\;(1+z)}{S_0(E_0(1+z),\alpha)}\; f
\end{eqnarray}
The fraction of sources of a given spectrum, at a given redshift
yielding an observed flux greater than $f$ is calculated by integrating the
luminosity function:
\begin{eqnarray}
n^\prime & = &
\int\limits^\infty_l \Phi(L,z)\,dL
\end{eqnarray}
where the lower limit, $l$ is the limiting luminosity for a given flux,
$f$:
\begin{eqnarray}
l & = & \frac{4\pi\;(a_0r)^2\;(1+z)}{S_0(E_0(1+z),\alpha)}\; f.
\end{eqnarray}
For pure luminosity evolution the redshift dependence of the luminosity
function is given by eq.\ref{eq-lfunctionz}, and the above equation
becomes:
\begin{eqnarray} \label{eq-n'}
n^\prime & = & \int\limits^\infty_{l/C(z)}
\Phi(L,0)\,dL
\end{eqnarray}
The number-flux relation is obtained by multiplying equation \ref{eq-n'} by
the number of sources in the redshift interval (eq. \ref{eq-dn}) and
integrating over redshift:
\begin{eqnarray} \label{eq-counts}
N(>f) & = & n_0 \;\frac{4\pi c^3}{H_0^3}\; \int\limits_0^\infty dz \,
\xi (z,\Omega)\; z^2\, C(z) \;
\int\limits_{\alpha} W(\alpha)\, d\alpha
\int\limits^\infty_{l/C(z)}
\Phi(L,0)\,dL
\end{eqnarray}
To compare results of source counts at different energies it is necessary to
express these fluxes in terms of a common energy band. It is common practice
to do this using the mean source spectrum (known or assumed). Let us denote
$S(E,\alpha_0)$ as the mean source spectrum (i.e. corresponding to some
mean value of the spectral parameters). The corresponding transformation of
the observed energy flux density $f(E)$ at energy $E$ to flux, $F$, in the
energy band ($E_1,E_2$) is:
\begin{eqnarray} \label{eq-fluxtransform}
F & = & f(E) \;\;\frac{\int\limits_{E_1}^{E_2} S(E,\alpha_0)\, dE}
{S(E,\alpha_0)}
\end{eqnarray}
Equations \ref{eq-counts} and \ref{eq-fluxtransform} can be used to calculate
difference between the soft and hard x-ray source counts, for example,
at 1 and 10 keV.
\section{Results} \label{sec-results}
In this section we present results of calculations of the CXB spectrum and
the number-flux relation in different energy bands for four models of
spectra distribution (Section \ref{sec-models}) and several values of the
density parameter, $\Omega$.
No qualitative difference is found in results of calculations performed for
different values of the cosmological density parameter, $\Omega$, ranging
from 0 to 1. When presenting the results we will simply point out the trend
of changing the derived relationships for $\Omega$ increasing from 0 to 1
and will not emphasize the dependence of the results on the density
parameter.
\subsection{Spectrum} \label{sec-results-spec}
Composite spectra of the source populations derived for models {\bf A}
through {\bf D} are shown in Fig.\ref{fig-spectrum}. These spectra are
obtained by integrating (eq.\ref{eq-cxbspec}) over redshifts from 0 to 10.
The assumption that sources are still bright at very high redshifts is of
course uncertain. However, we use the model for cosmological evolution in
which the growth of luminosity ceases at a redshift of 3 (section
\ref{sec-models}), and redshift effects rapidly make the contribution
of sources at higher redshifts negligible (cf. eq.\ref{eq-cxbspec-z}).
Models with a broad distribution of spectral parameters and the high energy
cutoff ({\bf A} and {\bf B}) perfectly reproduce the CXB spectrum from $\sim
1$ to $\sim 60-100$ keV. Model {\bf C}, without the high energy cutoff,
predicts further flattening of the CXB spectrum above $\sim 20$ keV, but
yields the correct slope of the spectrum in the 2-10 keV energy band. The
``standard'' model {\bf D}, without any scattering of spectral parameters,
reproduces the well-known result -- the average QSO/AGN spectrum is too
steep to fit the CXB spectrum in the energy range from 2-10 keV. Below
$\sim 2$ keV all models predict a steepening of the CXB spectrum, which is
not related to the presence of a soft excess in individual spectra. For
example, model spectrum {\bf C}, without any soft excess lies very close
to spectra {\bf A} and {\bf B} (with soft excess below 1 keV) above 0.7-1
keV.
In the harder x-ray band, above $\sim 20$ keV, a high energy cutoff is
required in source spectra to fit the CXB spectrum. For an $\Omega=0$
Universe, in the context of the present model, the break must be at around
100 keV; this value gradually decreases to 70 keV for $\Omega=1$ (see trend
of model spectra with increasing $\Omega$ indicated by the arrow in Fig.
\ref{fig-spectrum}). The broad distribution of the high energy spectral
index (model {\bf A}) reproduces the flattening of the CXB at $\geq 300$
keV. However, there is a lack of spectral data on QSO's at such high
energies. Not only are the particular values of the cutoff energies
unknown, but the very existence of breaks in QSO spectra is not very well
established (Mushotzky, Done, \& Pounds 1993). For this reason, we avoid
adjusting the spectral shapes at high energies to reproduce the detailed
shape of the CXB spectrum.
\subsection{Source counts}
We calculated the source number-flux relations at three different energies,
1, 5, and 10 keV, by Monte-Carlo integration of eq.\ref{eq-counts}. Source
fluxes at these energies were transformed to 2-10 keV fluxes by means of
eq.\ref{eq-fluxtransform}, using the average source spectrum (i.e. the
spectrum defined by mean values of parameters in the appropriate model).
The results are presented in Fig. \ref{fig-counts}. Models {\bf A} and {\bf
B}, which are distinguished by Gaussian distribution of the soft excess
parameters, yield essentially the same results, and therefore, only model
{\bf B} is presented in Fig. \ref{fig-counts}.
Models with broad distributions of spectral parameters predict a significant
difference between the results of hard and soft x-ray source counts. In the
EMSS flux range, $10^{-13}-10^{-12}$ ergs~s$^{-1}$~cm$^{-2}$, there are approximately three
times more sources observed from the same population at 10 keV than at the
corresponding 1 keV flux. We emphasize that this effect does not arise from
the presence of any spectral features (like a soft excess). Model {\bf C},
having a simple power law spectrum for individual sources, predicts almost
the same difference in hard and soft x-ray source counts, as does model {\bf
B}, with the soft excess. As can be expected, model {\bf D} predicts
number-flux relations to be the same at all energies. The agreement is
partially distorted if the soft spectral component which we assumed to be
present below 1 keV starts at higher energies. In this case, the soft excess
in spectra of faint sources is shifted to low energies. As a result, when
using the $z=0$ average spectrum for flux transformations, the 2-10 keV flux
is underestimated, which causes an artificial decrease in the soft x-ray
source counts relative to hard x-ray counts. Calculations show however, that
in the EMSS flux range, $10^{-13}-10^{-12}$ ergs~s$^{-1}$~cm$^{-2}$, the difference is less
than a factor of 1.3 if a soft excess starts at energies of about 3 keV.
\section {Discussion} \label{sec-discussion}
In this paper we have shown the importance of the distribution of spectral
parameters in calculations of the contribution of any source population to
the x-ray background. Of known classes of potential contributors to the CXB,
quasars demonstrate probably the broadest distribution of spectral
parameters, with 2-10 keV power law indices ranging from $\sim 0$ to $\sim
1.5-2$. Using models of QSO spectra distributed closely to what is observed,
we have shown that the composite spectrum of faint QSOs can be significantly
harder than that given by the average value, $\sim 0.8$, of the spectral
index. In fact, the composite spectrum of the QSO population closely mimics
the 0.5-100 keV CXB spectrum, if one assumes the existence of a high energy
cutoff in QSO spectra at around 100 keV. Another important effect which is
readily reproduced, when source spectra span a broad range, is the
difference between the results of hard and soft x-ray source counts. The
same population of sources yields about three times more sources at around
10 keV than at the corresponding 1 keV flux.
The key point of our spectral distribution models is that the soft x-ray
luminosity, at around 1 keV, does not correlate with other spectral
properties of sources. This means that hard sources are brighter at higher
energies than soft sources, and this actually accounts for approximately
half the effect of hardening required to reproduce the CXB spectrum;
redshift effects serve to enhance the effect. As was discussed in Section
\ref {sec-models}, available x-ray observations can neither prove nor
disaprove this assumption, and this is obviously the weak point of the
described model. Although the existence of a characteristic energy at which
quasar luminosities and spectra are not correlated does not seem
unreasonable, its particular value of 1 keV was chosen partially for the
sake of simplicity and partially because of the wish to reproduce the CXB
spectrum as well as possible. At the time of this writing it was generally
accepted that the CXB intensity in the soft x-ray band is in excess over the
$E^{-0.4}$ extrapolation from higher energies. Note that models {\bf A}
through {\bf C} spectra show $\sim 40-50$\% excess over the $E^{-0.4}$
spectrum in the $1-2$ keV energy band, which is in agreement with ROSAT
measurements of the CXB spectrum (Hasinger 1992). Later ASCA measurements
(Gendreau et al. 1994), on the contrary, show that CXB follows the
$E^{-0.4}$ spectrum down to at least 1 keV. Our model still can fit this
spectrum (without the ``soft excess''), provided that quasar luminosities
are uncorrelated with spectra at lower energies, $0.1-0.3$ keV, as is
suggested by Laor et al. (1994).
This paper assumes that the high energy break in QSO spectra occurs at
around 100 keV. With this value of the break energy it is possible to
reproduce the CXB spectrum in the very broad energy range from 0.5 to about
100 keV, if the cosmological density parameter, $\Omega$ is zero. To
reproduce the CXB spectrum in $\Omega=1$ Universe a break energy of about 70
keV is required, in good agreement with the break energy of 50-100 keV
observed by OSSE in low luminosity AGNs, mostly Seyfert 1s (Johnson et al.
1994).
Summarizing, the paper shows that quasars still may be considered as the
main contributors to the broad-band x-ray background. It also shows the
importance of relaxing the assumption that all source have the same
spectrum when the source population is characterized by significant
distribution of spectral shapes.
\begin {acknowledgements}
The author would like to thank R.Sunyaev for his friendly criticism. Special
thanks to W.Forman for his great help in work on the manuscript.
\end {acknowledgements}
\section*{References}
\beginrefs
Boyle B.J., Griffiths, R.E., Shanks, T., Steward G.C., and Georgantopoulous,
I. 1993, MNRAS, 260, 49.
Comastri, A., Setti, G., Zamorani, G., Elvis, M., Giommi, P., Wilkes, J., \&
McDowell, J., 1992, ApJ, 384, 62.
David, L.P., Slyz, A., Jones, C., Forman, W., Vrtilek, S.D., Arnaud, K.A.,
1993, ApJ, 412, 479
Fabian, A.C., Canizares, C.R., \& Barcons, X. 1989, MNRAS, 239, 15.
Fiore, F., Elvis, M., Siemiginowska, A., Wilkes, B.J. \& McDowell, J.C. 1994
ApJ in press.
Gendreau, K., et al. 1994, ``New Horizons of X-ray Astronomy -- First
Results from ASCA'', preprint.
Giacconi, R., Gursky, H., Paolini, F., and Rossi, B. 1962, Phys. Rev.
Letters, 9, 439.
Giacconi, R., \& Zamorani, G. 1987, ApJ, 313, 20.
Gioia, I.M., Henry, J.P., Maccacaro, T., Morris, S.L., \& Stocke, J.T. 1990,
ApJ, 356, L35.
Hasinger, G. 1992, in: The X-ray Background, X. Barcons \& A.C. Fabian eds.
(Cambridge University Press), p.1.
Hasinger, G., Burg, R., Giacconi, R., Hartner, G., Schmidt, M., Tr\"umper,
J., \& Zamorani, G., 1993, A\&A, 275, 1.
Johnson, W.N. et al., 1994, in the AIP Conference Proceedings 304, eds. C.
Fitchel, N. Gehrels, J. Norris, 515.
Jourdain, E., et al. 1992, A\&A, 256, L38.
Laor, A., Fiore, F., Elvis, M., Wilkes, B., \& McDowell, J.C., 1994, ApJ,
435, 611.
Marshall, F. et al. 1980. ApJ, 235, 4.
Mather, J. et al. 1990, ApJ, 354, L37.
Matsuoka, M., Piro, L., Yamauchi, M., and Murakami, T. 1990, ApJ,
361, 440.
Mattig, W. 1958, Astron.Nachr. 284, 109.
Mushotzky, R.F., 1984, Adv. Space. Res., 3, 10.
Mushotzky, R.F., Done, C., \& Pounds, K. 1993, Ann. Rev. Astr. \& Ap., 31,
717.
Nandra, K., \& Pounds, K.A., 1994, MNRAS, 1994, 268, 405.
Piccionotti, G., Mushotzky, R.F., Boldt, E.A., Holt, S.S., Marshall, F.E.,
Serlemitsos, P.J., and Shafer, R.A. 1982, ApJ, 253, 485.
Piro, L., Matsuoka, M., and Yamauchi, M. 1989, Proc. 23rd ESLAB
Symposium on ``Two Topics in X-ray Astronomy'', ESA SP-296, 819.
Pounds, K., Nandra, K., Stewart, G., George, I., and Fabian A. 1990,
Nature, 344, 132.
Schwartz, D. and Tucker, W. 1988, ApJ, 332, 157.
Wilkes, B. and Elvis, M. 1987, ApJ, 323, 243.
Williams, O.R. et al., 1992, ApJ, 389, 157.
\endgroup\vfill\eject
\clearpage
\begin{figure}
\epsfysize=6.5in \epsffile{spectra.ps}
\caption{
The composite spectrum calculated for models {\bf A}-{\bf D}, and the x-ray
background spectrum. The absolute normalization of model spectra is
arbitrary. All models with the scattering of spectral parameters ({\bf
A}-{\bf C}) adequately describe the CXB spectrum from $\sim 2$ to more than
10 keV. Models with the high energy cutoff at around 100 keV ({\bf A} and
{\bf B}) succeed in describing the CXB spectrum up to 60-100 keV. Model {\bf
D}, without any scattering of the spectral parameters, reproduces the well
known result: the mean spectrum of QSOs is too steep to fit the CXB spectrum
in the 2-10 keV energy band. The model spectra deviate from the hard x-ray
background extrapolation at energies below 2-2.5 keV. Note that the models
with and without the soft excess in source spectra ({\bf A} \& {\bf B}, and
{\bf C}, respectively) yield almost the same spectra above $\sim 1$ keV.
The arrow shows the trend of model spectra when the density parameter,
$\Omega$, increases. }
\label{fig-spectrum}
\end{figure}
\begin{figure}
\epsfysize=6.5in \epsffile{counts.ps}
\caption{
Source number-flux relation at three energies, 1, 5, and 10 keV, calculated
for different models of source spectra distribution. Normalizations are
arbitrary. For models with scatter of spectral parameters ({\bf B} and {\bf
C}) hard x-ray source counts yield more than two times more sources than the
observations performed at 1 keV; in the EMSS flux range, $10^{-13}-10^{-12}$
ergs~s$^{-1}$~cm$^{-2}$, the difference is approximately a factor of three for both models
with and without the soft excess ({\bf B} and {\bf C}, respectively).
Model {\bf D}, with the soft excess and without any scattering of spectral
parameters, predicts perfect agreement between soft and hard x-ray source
counts.}
\label{fig-counts}
\end{figure}
\end{document}
|
\section{Introduction}
\label{intro}
Topological effects play an important role in the dynamics of
asymptotically free field theories. In QCD instantons may be
responsible for breaking the axial symmetry resolving the so--called
U(1) problem \cite{WEINBERG}. In a large $N_c$ limit the
topological susceptibility relates the masses of the pseudo--scalars
$\eta$, $\eta'$ and $K$ \cite{WITTEN}.
The topological susceptibility $\chi_t$ may be defined as the infinite
volume limit of
\begin{equation}
\chi^{V}_t = \frac{\langle Q^2\rangle}{V},
\end{equation}
where $Q$ is the topological charge and $V$ is the space--time volume.
In the two--dimensional O(3) non--linear $\sigma$--model it is a
dimension two quantity that vanishes to all orders in the weak
coupling expansion. From the perturbative renormalization group (RG)
it is expected to scale according to the two loop $\beta$ function
\begin{equation} \label{eq-scaling}
\chi_t \propto \beta^2 \exp(- 4 \pi \beta), \qquad
( \beta \rightarrow \infty).
\end{equation}
It is a non--trivial task to recover the correct continuum results from
lattice Monte Carlo simulations. A lattice topological charge
definition is needed which returns even for large fluctuations
reliable results.
A `geometric' definition proposed by Berg and L\"uscher \cite{BERG1}
is based on adding up the area of spherical triangles which are
defined by the spin vectors in an elementary plaquette. As the
contributions from all plaquettes are summed up, the internal
space --- the sphere described by the spin variables --- is covered and
if periodic boundary conditions are used one obtains an integer charge
signifying the number of times this sphere is `wrapped'. The
topological susceptibility evaluated with this charge definition (and
the standard action) completely failed to scale
\cite{BERG1,LUSCHER1,BERG2}. The reason was ascribed to special
configurations called `dislocations' \cite{BERG2}, which are dominant
in the statistical average. Dislocations are non--zero charged
configurations whose contributions to the topological charge come
entirely from small localized regions where they become `singular'.
If the minimal action of dislocations is smaller than the continuum
value of a one--instanton configuration (i.e. $4\pi$) then
dislocations will dominate the path integral and spoil the scaling
behaviour, eq.~(\ref{eq-scaling}) \cite{LUSCHER1,BERG2}.
Another definition goes back to DiVecchia et al.\ \cite{DIVECCHIA} ---
for a recent discussion including the fixed point (FP) action see
ref.~\cite{DELIA}. It is called field theoretical or plaquette
definition and uses a `naive' discretization of the continuum charge
operator. This prescription does not yield integer values and to
obtain continuum results renormalization factors are needed. For
large $\beta$ these factors can be determined perturbatively, but for
intermediate $\beta$ one has to use non--perturbative techniques
\cite{TEPER1,DIGIACOMO,TEPER2,MICHAEL}. Results obtained with the
field theoretical charge indicate for the susceptibility a behaviour
consistent with scaling
\cite{DIGIACOMO,MICHAEL,FARCHIONI1}.
A serious problem in these approaches is the role of the lattice
artifacts, sensitive both to the form of the lattice action and the
choice of the topological charge. A recent work \cite{HASENFRATZ1}
suggests to use the FP action of a renormalization group transformation
to study topological effects. In particular, an important feature is that
the FP action has scale invariant instanton solutions (with an action
value exactly $4\pi$), and hence --- as will be discussed in this
paper --- one can define a topological charge with no lattice
defects. In ref.~\cite{HASENFRATZ1} and here the O(3) $\sigma$--model
is considered, but the methods apply to other asymptotically free
theories as well. The SU(3) gauge theory has been studied in
refs.~\cite{DEGRAND1,DEGRAND2}. A subsequent paper by one of us
\cite{BURKHALTER}, will deal with the application of these ideas to
$\mbox{CP}^{\mbox{\rm \tiny N-1}}$ models and in particular to the
$\mbox{CP}^{3}$ model. Some of our results were already
presented in ref.~\cite{BLATTER}.
The paper is organized as follows: First we review some results
derived in ref.~\cite{HASENFRATZ1} and define the FP field operator.
This is followed by a closer look at instantons in the continuum, in a
finite periodic volume and finally on a lattice using the FP
action. We then define the FP topological charge and present some
numerical results on classical solutions. In the last section we
analyze the topological susceptibility evaluated in a Monte Carlo
simulation. After a brief description of the methods used, we present
the results which are followed by a conclusion and an outlook.
\section{RG results at the classical level}
\subsection{Review of the RG transformation and its fixed point}
\label{review}
Let us briefly summarize some RG results which were developed in a
previous paper \cite{HASENFRATZ1}. For a detailed discussion we refer
the reader to this paper.
We consider the O(3) non--linear $\sigma$--model in two--dimensional
Euclidean space defined on a square lattice. The partition function
reads as follows
\begin{equation}
Z = \int\! \! \mbox{D} {\rm \bf S} \;
e^{ -\beta {\cal A}( \hbox{\scriptsize \bf S} ) }.
\end{equation}
Here $ \mbox{D} {\bf S} $ is the O(3) invariant measure
\begin{equation}
\mbox{D} {\bf S} = \prod_n d^{3} \mbox{S}_n \; \delta( {\bf S}_n^2 - 1)
\end{equation}
and $\beta \mbox{$\cal A$}({\bf S}) $ is a regularization of the continuum action
\begin{equation}
\beta {\cal A}_{\mbox{\tiny cont}}({\bf S}) =
{ \frac{\beta}{2} } \int \! d^2 \! x \,
\partial_{\mu} {\bf S}(x) \, \partial_{\mu} {\bf S}(x),
\qquad \mbox{where} \qquad {\bf S}^2(x) = 1.
\end{equation}
We perform exact RG transformations by a Kadanoff type of blocking,
i.e.\ we divide the lattice into $2 \times 2$ blocks labeled by
indices $n_{\mbox{\tiny B}}$. To each block we define a block spin variable
${\bf R}_{n_{\mbox{\tiny B}}}$ which is some mean of the spin variables ${\bf S}_n$ in
the block. The block spins ${\bf R}_{n_{\mbox{\tiny B}}}$ form a lattice
whose spacing is twice as large as the original one. An effective
action is defined by integration over the original lattice:
\begin{equation}
e^{-\beta' {\cal A}'( \mbox{\scriptsize \bf R} )} = \int\! \!
\mbox{D} {\bf S} \, e^{ -\beta \left [ {\cal A}(\mbox{\scriptsize \bf S})
+ {\cal T}(\mbox{\scriptsize \bf R},\mbox{\scriptsize \bf S} ) \right
] },
\end{equation}
where ${\cal T}$ is the kernel of the RG transformation and its
normalization
\begin{equation}
\int\! \! \mbox{D} {\bf R} \, e^{ -\beta
{\cal T}(\mbox{\scriptsize \bf R},\mbox{\scriptsize \bf S} ) } = 1
\end{equation}
ensures the invariance of the partition function under this
transformation. In the classical limit $(\beta \rightarrow \infty)$
the path integral is dominated by its saddle point:
\begin{equation}
{\cal A}'({\bf R}) = \min_{\{ \mbox{\scriptsize \bf S} \} }
\left \{ {\cal A}({\bf S}) + {\cal T}({\bf R},{\bf S}) \right\}.
\end{equation}
The transformation kernel used in ref.~\cite{HASENFRATZ1} has in the
limit $\beta \rightarrow \infty$ the simple form:
\begin{equation}
{\cal T}({\bf R},{\bf S}) = \kappa \sum_{n_{\mbox{\tiny B}}} \left (
\left | \sum_{n \in n_{\mbox{\tiny B}}}{\bf S}_n \right | - {\bf R}_{n_{\mbox{\tiny B}}} \cdot
\sum_{n\in n_{\mbox{\tiny B}}}{\bf S}_n \right ).
\end{equation}
Here $\kappa$ is a free positive parameter of the RG transformation,
which is tuned to make the FP action as compact as possible. As
indicated by the free field theory in one dimension, the choice
$\kappa = 2$ gives the most short--ranged FP action. A fixed point of
the transformation satisfies the equation
\begin{equation}
\label{FP-equation}
{\cal A}_{\mbox{\tiny FP}}({\bf R}) = \min_{\{ \mbox{\scriptsize \bf S} \} }
\left \{ {\cal A}_{\mbox{\tiny FP}}({\bf S}) + {\cal T}({\bf R},{\bf S}) \right \}.
\end{equation}
This equation --- called FP equation --- fixes for arbitrary
configurations $\{{\bf R}\}$ the value of the FP action. Starting from a
lattice regularization of the continuum action repeated RG
transformations will drive the effective action to its fixed point.
This takes
on the form of a minimization in a multigrid of lattice
configurations:
\begin{equation} \label{iFP-equation}
\mbox{$\cal A$}^{(k)}({\bf R}) = \min_{\{ \mbox{\scriptsize \bf
S}^{(1)},\mbox{\scriptsize \bf S}^{(2)},\ldots, \mbox{\scriptsize \bf
S}^{(k)} \}} \left \{ \mbox{$\cal A$}^{(0)}({\bf S}^{(k)}) +
\trkern({\bf S}^{(k)},{\bf S}^{(k-1)}) + \ldots + \trkern({\bf R},{\bf S}^{(1)})
\right \}.
\end{equation}
\begin{figure}[htb]
\begin{center}
\vskip -5mm
\leavevmode
\epsfxsize=90mm
\epsfbox{levels.eps}
\vskip 5mm
\caption{A multigrid is obtained by iterating the FP equation.}
\label{levels}
\end{center}
\end{figure}
On each successive level --- see figure~\ref{levels} --- the spin
configurations become smoother, not only because the lattice spacing
is halved, but also because the minimization tends to smooth out the
fluctuations around a solution to the equations of motion. Hence one
may choose for the action $\mbox{$\cal A$}^{(0)}({\bf S}^{(k)})$ on the finest
configuration $\{{\bf S}^{(k)}\}$ any lattice discretization of the
continuum action. The FP action $\mbox{$\cal A_{\mbox{\tiny FP}}$}$ is then obtained as the
limit of $k \rightarrow \infty$ of $\mbox{$\cal A$}^{(k)}({\bf R})$. For practical
purposes, however, only a few levels are needed and starting from the
standard action on the lowest level the FP value is reached soon.
\subsection{Parametrization of the FP action}
\label{paraction}
In principle the above multigrid approach can be used to evaluate the
FP action for arbitrary configurations to any precision desired. For
practical calculations, however, a parametrization of the FP action is
needed. In ref.~\cite{HASENFRATZ1} a parametrization has been
obtained by fitting the known values of the action for $\sim 500$
configurations. That parametrization represented well the FP action on
those configurations. However, to control the topological effects
better, we decided to improve the parametrization further by
including some small size topological solutions in the fitting
procedure. We used several two--instanton solutions of the lattice FP
action. While improving the fit for these instanton solutions, the
new parametrization does not affect the quality of the fit for the
previous configurations.
The resulting couplings are given in table~\ref{newcouplings} --- together
with a graphical notation of the corresponding operators. Let us explain
here again the meaning of this notation. The
parametrization of the action has the form
\begin{equation}
{\cal A}_{\mbox{\tiny FP}}({\bf S}) = \sum \; {\rm coupling \; \times \;
products \; of} \; \frac{1}{2}\vartheta_{n_i,n_j}^2\,,
\end{equation}
where $\vartheta_{n_i,n_j}$ is the angle between the two spins
${\bf S}_{n_i}$ and ${\bf S}_{n_j}$. Two dots connected with a line
\begin{picture}(22,6)(0,0)
\put(2,3){\circle*{3.5}}
\put(20,3){\circle*{3.5}}
\put(2,3){\line(1,0){18}}
\end{picture}
represent a factor $\frac{1}{2}\vartheta_{n_i,n_j}^2$ in the
action and the positions of the dots represent the lattice sites $n_i$
and $n_j$ respectively. Double, triple connected dots stand for the
square, cube of the above factors. The operator, finally, is the
product of all the factors $\frac{1}{2}\vartheta_{n_i,n_j}^2$
as indicated by the lines in the figure. The quadratic
and quartic couplings \# 1,2,4,5,7,10,16 and 19 are determined
analytically \cite{HASENFRATZ1}, the others were determined with a
numerical fitting procedure with the new instanton configurations
added.
\begin{table}
\setlength{\unitlength}{0.5mm}
\begin{tabular}{r cr c cr c cr}
\# & type & coupling & ~~~ & type & coupling
& ~~~ & type & coupling \\
1 &
\begin{picture}(20,10)(0,3)
\put(5,5){\circle*{2}}
\put(15,5){\circle*{2}}
\put(5,5){\line(1,0){10}}
\end{picture} & $0.61884$ & &
\begin{picture}(20,10)(0,3)
\put(5,5){\circle*{2}}
\put(15,5){\circle*{2}}
\put(5,4.5){\line(1,0){10}}
\put(5,5.5){\line(1,0){10}}
\end{picture} & $-0.04957$ & &
\begin{picture}(20,10)(0,3)
\put(5,5){\circle*{2}}
\put(15,5){\circle*{2}}
\put(5,4.1){\line(1,0){10}}
\put(5,5){\line(1,0){10}}
\put(5,5.9){\line(1,0){10}}
\end{picture} & $-0.00932$ \\
4 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,1){10}}
\end{picture} & $0.19058$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.5){\line(1,1){10}}
\put(5,1.5){\line(1,1){10}}
\end{picture} & $-0.02212$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.1){\line(1,1){10}}
\put(5,1.9){\line(1,1){10}}
\put(5,1){\line(1,1){10}}
\end{picture} & $-0.00746$ \\
7 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,1){\line(1,1){10}}
\end{picture} & $0.01881$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(5,1){\line(1,1){10}}
\end{picture} & $-0.00180$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,0.5){\line(1,1){10}}
\put(5,1.5){\line(1,1){10}}
\end{picture} & $0.00658$ \\
10 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $0.02155$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $0.00536$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(14.5,1){\line(0,1){10}}
\put(15.5,1){\line(0,1){10}}
\end{picture} & $-0.00081$ \\
13 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,1){\line(1,1){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $0.00941$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(5,1){\line(1,1){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $0.00488$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,0.5){\line(1,1){10}}
\put(5,1.5){\line(1,1){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $-0.00225$ \\
16 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,1){\line(1,1){10}}
\put(5,11){\line(1,-1){10}}
\end{picture} & $0.01209$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,0.5){\line(1,1){10}}
\put(5,1.5){\line(1,1){10}}
\put(5,11){\line(1,-1){10}}
\end{picture} & $0.00534$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,0.5){\line(1,1){10}}
\put(5,1.5){\line(1,1){10}}
\put(5,10.5){\line(1,-1){10}}
\put(5,11.5){\line(1,-1){10}}
\end{picture} & $0.00066$ \\
19 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,11){\line(1,0){10}}
\end{picture} & $-0.00258$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(5,11){\line(1,0){10}}
\end{picture} & $-0.00173$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,0.5){\line(1,0){10}}
\put(5,1.5){\line(1,0){10}}
\put(5,10.5){\line(1,0){10}}
\put(5,11.5){\line(1,0){10}}
\end{picture} & $0.00146$ \\
22 &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,11){\line(1,0){10}}
\put(5,1){\line(0,1){10}}
\end{picture} & $-0.01040$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,11){\line(1,0){10}}
\put(4.5,1){\line(0,1){10}}
\put(5.5,1){\line(0,1){10}}
\end{picture} & $-0.00218$ & &
\begin{picture}(20,14)(0,5)
\put(5,1){\circle*{2}}
\put(15,1){\circle*{2}}
\put(15,11){\circle*{2}}
\put(5,11){\circle*{2}}
\put(5,1){\line(1,0){10}}
\put(5,11){\line(1,0){10}}
\put(5,1){\line(0,1){10}}
\put(15,1){\line(0,1){10}}
\end{picture} & $0.02720$ \\
\end{tabular}
\caption{Couplings used for the parametrization of the FP action
including instanton configurations. The graphical notation is
explained in the text.}
\label{newcouplings}
\end{table}
\subsection{The fixed point field}
\label{parfield}
As we shall see in section~\ref{tcharge},
FP operators are closely related to the FP field.
The FP field is the fine field ${\bf S}^{(k)}$ in the
multigrid solution of the iterated FP equation~(\ref{iFP-equation}) as
$k$ goes to infinity. If the functional dependence of the solution on
the first fine level ${\bf S}^{(1)}$ on ${\bf R}$ is known, the FP
field can be evaluated by iteration. Below we construct the operator
${\bf S}^{(1)}={\bf S}^{(1)}({\bf R})$. (In the limit $k \rightarrow \infty$ the
solution $\{ {\bf S}^{(1)} \}$ of the iterated FP equation is identical to
the solution $\{ {\bf S} \}$ of the FP equation.)
For smooth fields a quadratic approximation of the FP equation
can be made which can be solved analytically.
Consider a smooth configuration $\{{\bf R}\}$
where the spins fluctuate around the first axis:
\begin{equation} \label{13}
{\bf R}_{n_{\mbox{\tiny B}}}=\left( {\sqrt{1-\vec{\chi}_{n_{\mbox{\tiny B}}}^{\,2}}
\atop \vec{\chi}_{n_{\mbox{\tiny B}}}} \right),
\end{equation}
where $\vec{\chi}_{n_{\mbox{\tiny B}}}$ has $2$ components, and
$|\vec{\chi}_{n_{\mbox{\tiny B}}}|\ll 1$.
For the minimizing fine field we can make the ansatz
\begin{equation} \label{14}
{\bf S}_{n}^{(1)}=\left( {\sqrt{1-\vec{\pi}_{n}^{\,2}}
\atop \vec{\pi}_{n}} \right).
\end{equation}
Inserted in the FP equation, the above expansion leads in leading
order to the free field case. This was solved by Bell and Wilson
\cite{BELL} --- for a brief review see for instance
ref.~\cite{HASENFRATZ1} --- here we only report the relation between
$\vec{\pi}_n$ and $\vec{\chi}_{n_{\mbox{\tiny B}}}$:
\begin{equation} \label{15}
\vec{\pi}_n = \sum_{n_{\mbox{\tiny B}}} \alpha (n, n_{\mbox{\tiny B}})
\vec{\chi}_{n_{\mbox{\tiny B}}}.
\end{equation}
Here $\alpha$ is given by
\begin{equation}
\alpha (n, n_{\mbox{\tiny B}}) =
\int_0^{2\pi} \frac{d^2q}{(2\pi)^2}
e^{- i q ( n - 2 n_{\mbox{\tiny B}} ) }
\frac{1}{4} \frac{\tilde{\rho}_{\mbox{\tiny FP}}(2 q)}
{\tilde{\rho}_{\mbox{\tiny FP}}( q)}
\prod_{j = 1}^2 \frac{1 - e^{- 2 i q_j } }
{1 - e^{- i q_j } }
\end{equation}
where $\tilde{\rho}_{\mbox{\tiny FP}} (q)$ is the coefficient in the
quadratic part of the FP action (given by the free field case). By
iterating eq.~(\ref{15}) one can obtain the FP field operator in the free
field case. The form is very similar to the above result, but with
slightly modified parameters \cite{DEGRAND1}.
On smooth configurations $\{{\bf R}\}$ eqs.~(\ref{13}--\ref{15}) give
a good approximation. However we want to evaluate the fine field
using a parametrization that performs well not only for smooth
configurations $\{{\bf R}\}$, but also for general ones.
Our experience with the parametrization of the FP action suggests
the ansatz
\begin{equation}
{\bf S}_n = {\cal N} \left [\sum_{n_{\mbox{\tiny B}}} \alpha(n, n_{\mbox{\tiny B}})
{\bf R}_{n_{\mbox{\tiny B}}} + \!\!\! \sum_{n_{\mbox{\tiny B}} \atop
m_{\mbox{\tiny B}}^{},m_{{\mbox{\tiny B}}}'} \!\!\! \beta(n,n_{\mbox{\tiny
B}},m_{{\mbox{\tiny B}}}^{},m_{\mbox{\tiny B}}') \frac{1}{2}
\vartheta^2_{m_{\mbox{\tiny B}}^{},m_{\mbox{\tiny B}}'} {\bf R}_{n_{\mbox{\tiny B}}} \right ],
\label{parametfield}
\end{equation}
where $\cal N$ is a normalizing factor which ensures ${\bf S}_n^2=1$. Like
for the action it proved to be useful to use instead of the scalar
product $(1 - {\bf R}_{m_{\mbox{\tiny B}}^{}}{\bf R}_{m_{\mbox{\tiny B}}'})$
between the coarse spins at sites $m_{\mbox{\tiny B}}$ and $m_{\mbox{\tiny B}}'$ respectively,
the angle $\frac{1}{2}\vartheta^2_{m_{\mbox{\tiny B}}^{},m_{\mbox{\tiny B}}'}$.
In order to
determine the coefficients $\beta$, we numerically minimized the FP
equation~(\ref{FP-equation}) using 60 configurations with lattice size
$5$ as input and stored the resulting fine lattices. The coefficients
were then determined by minimizing the difference between the
minimized fine spins and the parametrization (\ref{parametfield}).
The numerical values of the coefficients $\alpha$ and $\beta$ are
given in table~\ref{alphabeta} together with a symbolic notation of the
corresponding operators. We chose a set of 23 operators mainly because
of their compactness. \#~1--6 are the analytically determined
coefficients $\alpha$ and \#~7--23 are the numerically determined
coefficients $\beta$. The meaning of the graphical notation of
the operators is the following: The dashed lines represent a $3\times
3$ section of the coarse lattice grid. The cross
\begin{picture}(8,6)(0,0)
\put(4,3){\line(1,0){3.5}}
\put(4,3){\line(0,1){3.5}}
\put(4,3){\line(-1,0){3.5}}
\put(4,3){\line(0,-1){3.5}}
\end{picture}
inbetween indicates the position $n$ of the fine spin ${\bf S}_n$ in
equation~(\ref{parametfield}). The little square
\begin{picture}(8,6)(0,0)
\put(2,1){\framebox(4,4)}
\end{picture}
denotes the position $n_{\mbox{\tiny B}}$ of the coarse spin
${\bf R}_{n_{\mbox{\tiny B}}}\,$. The two connected dots
\begin{picture}(22,6)(0,0)
\put(2,3){\circle*{2.5}}
\put(20,3){\circle*{2.5}}
\put(2,3){\line(1,0){18}}
\end{picture}
are the positions $m_{\mbox{\tiny B}}$ and $m_{\mbox{\tiny B}}'$ of
the spins whose angle $\frac{1}{2} \vartheta^2_{ m_{\mbox{\tiny
B}}^{}\, m_{\mbox{\tiny B}}'}$ enters into the parametrization.
Graphs obtained by trivial symmetry transformations are not drawn
separately.
\begin{table}\centering
\setlength{\unitlength}{0.5mm}
\begin{tabular}{|c c c| c c c| c c c|}
\hline
\# & type & coeff & \# & type & coeff
& \# & type & coeff \\
\hline
1 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,9){\framebox(2,2)}
\end{picture} &
$0.59497$
&
2 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\end{picture} &
$0.15621$
&
3 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,-1){\framebox(2,2)}
\end{picture} &
$0.08300$ \\
& & & & & & & & \\
\hline
4 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,19){\framebox(2,2)}
\end{picture} &
$0.00942$
&
5 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,19){\framebox(2,2)}
\end{picture} &
$-0.00171$
&
6 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(19,19){\framebox(2,2)}
\end{picture} &
$-0.00668$ \\
& & & & & & & & \\
\hline
7 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,9){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,10){\line(1,0){10}}
\end{picture} &
$-0.01228$
&
8 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,9){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(1,1){10}}
\end{picture} &
$-0.02004$
&
9 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,9){\framebox(2,2)}
\put(10,0){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,10){\line(1,-1){10}}
\end{picture} &
$-0.03832$ \\
& & & & & & & & \\
\hline
10 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,9){\framebox(2,2)}
\put(0,10){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(0,1){10}}
\end{picture} &
$-0.05095$
&
11 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,10){\line(1,0){10}}
\end{picture} &
$0.01475$
&
12 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(10,0){\circle*{1.5}}
\put(10,0){\line(0,1){10}}
\end{picture} &
$-0.00586$ \\
& & & & & & & & \\
\hline
13 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(1,1){10}}
\end{picture} &
$-0.00303$
&
14 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(10,0){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,10){\line(1,-1){10}}
\end{picture} &
$-0.00123$
&
15 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(0,10){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(0,1){10}}
\end{picture} &
$-0.00118$ \\
& & & & & & & & \\
\hline
16 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,9){\framebox(2,2)}
\put(10,0){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(1,0){10}}
\end{picture} &
$-0.01596$
&
17 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,-1){\framebox(2,2)}
\put(0,10){\circle*{1.5}}
\put(10,10){\circle*{1.5}}
\put(0,10){\line(1,0){10}}
\end{picture} &
$-0.00090$
&
18 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,-1){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(0,0){\circle*{1.5}}
\put(0,0){\line(1,1){10}}
\end{picture} &
$0.00140$ \\
& & & & & & & & \\
\hline
19 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,-1){\framebox(2,2)}
\put(10,0){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,10){\line(1,-1){10}}
\end{picture} &
$-0.00128$
&
20 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,-1){\framebox(2,2)}
\put(0,0){\circle*{1.5}}
\put(0,10){\circle*{1.5}}
\put(0,0){\line(0,1){10}}
\end{picture} &
$0.00447$
&
21 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(-1,19){\framebox(2,2)}
\put(0,10){\circle*{1.5}}
\put(10,10){\circle*{1.5}}
\put(0,10){\line(1,0){10}}
\end{picture} &
$0.00199$ \\
& & & & & & & & \\
\hline
22 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(9,19){\framebox(2,2)}
\put(0,10){\circle*{1.5}}
\put(10,10){\circle*{1.5}}
\put(0,10){\line(1,0){10}}
\end{picture} &
$0.00304$
&
23 &
\begin{picture}(20,20)(0,4)
\put(0,0){\dashbox{1}(10,10)}
\put(0,10){\dashbox{1}(10,10)}
\put(10,0){\dashbox{1}(10,10)}
\put(10,10){\dashbox{1}(10,10)}
\put(7.5,7.5){\line(1,0){1.5}}
\put(7.5,7.5){\line(0,1){1.5}}
\put(7.5,7.5){\line(-1,0){1.5}}
\put(7.5,7.5){\line(0,-1){1.5}}
\put(19,19){\framebox(2,2)}
\put(10,10){\circle*{1.5}}
\put(20,20){\circle*{1.5}}
\put(10,10){\line(1,1){10}}
\end{picture} &
$-0.00015$
&
& & \\
& & & & & & & & \\
\hline
\end{tabular}
\caption{Coefficients of the parametrization of the fine field.}
\label{alphabeta}
\end{table}
\section{Instantons}
\label{instanti}
\subsection{Instantons on the lattice}
\label{instanto}
In infinite volume continuum field theory topology is a well defined
concept. Field configurations can be classified in topological
sectors according to a `winding number' or topological charge
\cite{BELAVIN2}. In a lattice formulation this concept breaks
down. When discretizing a theory, continuity in coordinate and
internal space is lost. On the other hand topology is based on
continuous transformations of mappings which are separable in
classes. In a discretized theory every field configuration can be
continuously transformed into any other. If lattice configurations
are sufficiently smooth, an unambiguous topological charge may be
assigned. Conversely, for field configurations containing large
fluctuations an interpolation is not unique and a charge definition
becomes ambiguous.
An additional problem arises due to the discretization of the
continuum action. While the continuum action possesses scale invariant
instanton solutions, this is generally not true for discretized
actions. In particular, starting with non--zero charged configurations
L\"uscher \cite{LUSCHER1} found that one can continuously lower the
standard lattice action to zero by a local minimization in the spin
variables.
In reference \cite{HASENFRATZ1} it was suggested to investigate the
above problems with the aid of renormalization group methods and it
was shown that the FP action has scale invariant instanton solutions.
In this section we continue along these lines and
construct instanton solutions of the FP action. These in turn can be
used to study the performance of a proposed improved topological
charge definition. But first let us review some basic facts about
instantons in the continuum \cite{BELAVIN2}.
In an infinite volume configurations with a finite action play a
special role: At `infinity' all spin variables point in the same
direction and the space $\reell^2$ can be compactified by stereographic
projection into a sphere $S^2$. A finite action configuration is thus
a mapping of a `coordinate' sphere onto the internal sphere
${\bf S}_n^2=1$. Such mappings can be classified by homotopy classes, with
an integer number --- the topological charge $Q$ --- characterizing the
sectors. While configurations from the same topological sector can be
continuously deformed into each other, this is not true for
configurations with a different charge. The charge $Q$ is the number
of times the internal sphere is wrapped as the coordinate sphere is
traversed. It may be defined as the integral
\begin{equation}
Q= \frac{1}{8\pi}\int \! d^2 \! x \, \epsilon_{\mu\nu} \,
{\bf S} \cdot ( \partial_{\mu} {\bf S} \times \partial_{\nu} {\bf S} ).
\end{equation}
and it is related to the action by the inequality
\begin{equation}
\mbox{$\cal A$} \geq 4\pi\, |Q|.
\label{actinequality}
\end{equation}
If for a given configuration the equality is satisfied, the
configuration minimizes the action and is therefore a solution of the
equations of motion.
Let us now turn to the theory in a finite volume. By demanding
periodic boundary conditions
\begin{equation}
{\bf S}(x_1 + Lm , x_2 + L n) = {\bf S}(x_1,x_2), \qquad \mbox{where}
\qquad m, \, n \in \ganz,
\end{equation}
we define the theory on a square torus of size $L$. In a finite
volume every field configuration has a finite action and due to the
periodic boundary conditions an integer topological charge $Q$
associated with it.
We can now explicitly construct pure instanton or pure anti--instanton
configurations with an action $\mbox{$\cal A$}=4\pi|Q|$ \cite{RICHARD}. We
use the plane coordinates defined by the stereographic projection to
describe the solutions:
\begin{eqnarray}
{\bf S}_i & = & \frac{ 2 \, u_i}{ 1 + |u|^2} \hspace{1cm} i = 1,\, 2,
\nonumber \\ & & \\ {\bf S}_3 & = & \frac{1 - |u|^2}{1 + |u|^2}, \nonumber
\end{eqnarray}
where $u = u_1 + i u_2$ .
The instanton solutions at the boundary of equation~(\ref{actinequality})
satisfy the
Cauchy--Riemann equations for $u$ being an analytic function in
$z = x_1 + i x_2$:
\begin{equation}
(\partial_1 + i \partial_2) u = 0.
\end{equation}
The solutions are doubly--periodic meromorphic functions called
elliptic functions \cite{RICHARD}.
They can be written as
\begin{equation}
u = c \prod_{i = 1}^{k} \frac{ \sigma(z-a_i) }{\sigma(z-b_i)},
\quad {\rm with} \quad \sum_{i=1}^{k} a_i = \sum_{i=1}^{k} b_i.
\end{equation}
Here the integer $k$ is the topological charge of the solution and
$c$, $a_1,\ldots,a_k$ and $b_1,\ldots,b_k$ are complex numbers.
$\sigma(z)$ is the Weierstrass $\sigma$--function with half--periods
$\omega = L / 2$ and $\omega'= iL / 2 $. Using Cauchy's theorem one
can show that there are no solutions with topological charge equal to
one \cite{RICHARD}. Hence we are forced to construct charge two
instanton solutions. Specifically, we set $k=2$ and $c=1$.
A reasonable definition for the instanton size is
\begin{equation}
\rho = \mbox{$\scriptstyle \frac{1}{2}$} \min\{|a_1-b_1|,|a_1-b_2|\}.
\label{size}
\end{equation}
Let us now turn to the construction of instanton solutions on the
lattice. As was pointed out in ref.~\cite{HASENFRATZ1} the FP action
allows scale invariant instanton solutions. Since we use this fact to
construct instantons on the lattice, it is appropriate to repeat the
statement:
If a given configuration $\{{\bf R}\}$ satisfies the equations of motion
for the FP action $\mbox{$\cal A_{\mbox{\tiny FP}}$}$ and it is a minimum, then the solution
$\{ {\bf S} \}$ of the FP equation~(\ref{FP-equation}) satisfies
the equations of motion also. Moreover, both configurations yield the
same value for the action.
The proof is quite simple: If $\{{\bf R}\}$ is at a local minimum of the
FP action \mbox{$\cal A_{\mbox{\tiny FP}}$}, variations with respect to $\{{\bf R}\}$ will
vanish:
\begin{equation}
\frac{\delta\mbox{$\cal A_{\mbox{\tiny FP}}$}({\bf R})}{\delta{\bf R}_{n_{\mbox{\tiny B}}}} =
\kappa \left ( - \sum_{n \in n_{\mbox{\tiny B}}} {\bf S}_n + {\bf R}_{n_{\mbox{\tiny B}}}
\left ( {\bf R}_{n_{\mbox{\tiny B}}} \cdot \sum_{n \in n_{\mbox{\tiny B}}} {\bf S}_n \right )
\right ) = 0.
\label{variation}
\end{equation}
Here ${\bf S}={\bf S}({\bf R})$ is the solution of the FP equation with $\{ {\bf R} \}$
as coarse input configuration.
Since $\{ {\bf R} \}$ is a local minimum eq.~(\ref{variation}) implies
\begin{equation}
{\bf R}_{n_{\mbox{\tiny B}}} = \frac{\sum_{n \in n_{\mbox{\tiny B}}}{\bf S}_n}
{| \sum_{n \in n_{\mbox{\tiny B}}}{\bf S}_n | }.
\label{blocking}
\end{equation}
Consequently, we have
\begin{equation}
\trkern({\bf R},{\bf S}) = 0.
\end{equation}
Since the transformation kernel $\trkern({\bf R},{\bf S}) \ge 0$,
the configuration $\{ {\bf S} \}$ gives its minimum for fixed $\{ {\bf R} \}$.
Because the configuration $\{{\bf S} \}$ minimizes the right hand side
of the FP equation, it minimizes the FP action $\mbox{$\cal A_{\mbox{\tiny FP}}$}({\bf S})$ separately.
Therefore it is a solution of the equations of motion.
Since $\trkern({\bf R},{\bf S}) = 0$, both actions have the same value:
$\mbox{$\cal A_{\mbox{\tiny FP}}$}({\bf R}) = \mbox{$\cal A_{\mbox{\tiny FP}}$}({\bf S})$. This concludes the proof.
The reverse of the above statement is in general not true for arbitrary
configurations. Although a fine configuration which is a solution of
the FP equations of motion and which is used to construct a coarse
configuration by means of the above blocking
equation~(\ref{blocking}), locally minimizes the right hand side of
the FP equation, this minimum does not necessarily coincide with the
absolute one. This in fact prevents the existence of arbitrary small
instanton solutions.
Using the above ideas it is clear how to construct instanton solutions of
\mbox{$\cal A_{\mbox{\tiny FP}}$}. We naively discretize the continuum instanton solution on a
very fine lattice with spacing $a_0=2^{-k}a$. After performing $k$
blocking steps as defined by equation~(\ref{blocking}), we obtain a
configuration $\{{\bf R}\}$ on a lattice with spacing $a$. In the limit $
k \rightarrow \infty $ we recover the continuum solution, which is of
course a local minimum of the continuum action. Since all the
successive blocked configurations minimize the transformation kernels
in the iterated FP equation, the configuration $\{ {\bf R}\}$ is a good
candidate for a lattice solution of the FP action\footnote{ It will be
a solution, unless the size of the instantons is too small with
respect to the lattice spacing.}. We may solve the FP equation for
$\{{\bf R}\}$ to check whether it is still a solution. If $\{ {\bf R} \}$ is
a solution, then the multigrid minimization procedure should lead to
the same configurations on finer lattices as those which were used in
constructing $\{ {\bf R} \}$ by blocking.
\subsection{Definition of the topological charge on the lattice}
\label{tcharge}
In the following we define a topological charge operator based on the
multigrid solution of the FP action. We evaluate the FP topological
charge by means of the solutions of the FP
equation~(\ref{FP-equation}). Under a RG transformation an operator
${ \cal O}( {\bf S})$ transforms into ${\cal O}'({\bf R})$ on the coarse
lattice as
\begin{equation}
{\cal O}'({\bf R}) e^{-\beta' {\cal A}'(\mbox{\scriptsize\bf R})} = \int\!
\! \mbox{D} {\bf S} \;
{ \cal O}( {\bf S}) e^{ -\beta [ {\cal A}(\mbox{\scriptsize \bf
S}) + {\cal T}(\mbox{\scriptsize \bf R},\mbox{\scriptsize \bf S} ) ]
},
\end{equation}
In the limit $\beta \rightarrow \infty$ the path integral on the right
hand side is approximated by its saddle point and we obtain
\begin{equation}
{\cal O}'({\bf R}) = {\cal O}({\bf S}({\bf R})),
\end{equation}
where the spin configuration $\{{\bf S}({\bf R})\}$ is the solution of the FP
equation~(\ref{FP-equation}). Repeated application of this
transformation will single out the operator with largest eigenvalue.
Since the topological charge is expected to be a marginal operator, we
may obtain it as the limit
\begin{equation}
Q_{\mbox{\rm \tiny FP}}({\bf R}) = \lim_{k \rightarrow \infty}
Q({\bf S}^{(k)}({\bf R})).
\end{equation}
Here $Q$ is some standard lattice charge definition and $\{ {\bf S}^{(k)}
\}$ is the solution of the iterated FP equation~(\ref{iFP-equation})
on the lowest level in a $k$ level multigrid (see figure~\ref{levels}).
In other words, the FP topological charge is a standard topological charge
evaluated on the FP field.
Note that ${\bf S}^{(k)}({\bf R})$ becomes increasingly smooth as $k$ grows:
first, because the corresponding lattice spacing $a_0=2^{-k} a$
decreases, secondly, because ${\bf S}^{(k)}({\bf R})$ becomes almost a solution
to the equations of motion. Consequently, any sensible definition of the
topological charge can be used on these configurations, the final result
will not depend on this choice. Nevertheless, it is more convenient
to use the geometric definition since it is stable against small
variations of the field and hence $k$, the number of levels in the
multigrid minimization could be kept small with this definition.
For a review of the geometric definition we refer the reader to
ref.~\cite{BERG1}.
One can easily show that with this definition of the topological
charge there are no dangerous dislocations. More precisely,
one has
\begin{equation} \label{31}
{\cal A}_{\mbox{\rm \tiny FP}}({\bf R}) \geq 4 \pi \, | Q_{\mbox{\rm \tiny
FP}}({\bf R}) |,
\end{equation}
for arbitrary configuration $\{ {\bf R} \}$.
The corresponding statement is true in the continuum, hence it is also
true for ${\bf S}^{(k)}({\bf R})$ for $k\to\infty$. Eq.~(\ref{31}) follows
then by observing that the contribution of the ${\cal T}$ terms in
eq.~(\ref{FP-equation}) is non--negative.
We are now ready to discuss the numerical
aspects of classical solutions.
\subsection{Classical numerical results}
\label{classres}
Following the above program we naively discretize two--instanton
solutions on the torus of various sizes on very fine lattices.
We found that four to five blocking steps are sufficient to make any
lattice artifacts of the original discretization negligible.
On the finally blocked configurations we can measure several
quantities. On the coarse configuration itself we measure the standard
action and the parametrization of the FP action presented in
section~\ref{paraction}. Performing a minimization on a multigrid with
three finer levels we measure the exact FP action and on the finest
level the FP charge. Using the instanton radius given by
eq.~(\ref{size}), we get a parameter which characterizes well the
breakdown of the blocking to obtain instanton solutions on coarse
lattices: In figure~\ref{rhovar} it can be clearly seen that below an
instanton radius of $\rho \mathrel{\mathpalette\@sim<} 0.7 a$ the configurations are no
longer instanton solutions and we shall say that the instanton falls
through the lattice. It is gratifying to see that the FP charge
immediately falls off to zero, as the FP action drops below the
continuum value. Furthermore, figure~\ref{rhovar} demonstrates how
well the parametrization for the FP action is suited for instanton
configurations. The deviation from the exact value is quite small, in
particular the parametrized FP action is only marginally smaller than
the continuum value in the region above the point where the instanton
falls through the lattice. In contrast, the values of the standard
action are quite different from the continuum ones. These numerical
findings support the above statement, that there are no dangerous
dislocations present when using the FP action together with the FP
topological charge.
\begin{figure}[htb]
\begin{center}
\leavevmode
\epsfxsize=110mm
\epsfbox{rhovar.eps}
\vskip -5mm
\caption{Actions and charge of instantons with radii of the order of one
lattice spacing}
\label{rhovar}
\end{center}
\end{figure}
In numerical simulations we use a parametrized form of the
FP action and the parametrization in eq.~(\ref{parametfield}) for the
FP field instead of the time consuming minimization procedure. One
then is interested if the use of these two parametrizations has an
influence on the existence of dislocations. The curves referring to
the FP action and FP charge in fig.~\ref{rhovar} have a non--analytic
break at $\rho / a \simeq 0.7$. Our parametrization does not fully
reproduce this behaviour, and so one might expect that observable
deviations will occur in this instanton region. We systematically
searched for minimal action configurations with a parametrized action
lower than the continuum value in the $Q_{\mbox{\tiny FP}}^{\rm
par.}=2$ sector. We found a minimal action of $1.84\cdot 4\pi$
(compare this with the value $0.93\cdot 4\pi$ we found using the
standard action and geometric charge). This value can be ascribed to
the not exactly accurate parametrization of the FP field. If we
actually solve the FP equation (\ref{FP-equation}) for this
`dislocation configuration' we get the correct charge
$Q=0$. Nevertheless, as one has to use parametrizations for Monte
Carlo simulations, such configurations could be dangerous. On the
other hand, the search for dislocations reveals the weakest point of
the parametrization which performs very well in other cases
(c.f. figure~\ref{rhovar}). What actually counts is not how the
parametrization works for some configurations that were specially
sought for their bad performance, but how well it performs for
configurations in thermal equilibrium occurring in a Monte Carlo
simulation. The results of a test of this performance --- presented in
section~\ref{results} --- shows that indeed there is no problem.
We also searched for the minimal action configuration in the $Q=1$
sector and did not find a configuration with an action below
the continuum action. This is not astonishing as there are no
one--instanton solutions on a torus.
\section{Topological Susceptibility}
\label{topological}
If the topological susceptibility is a well defined physical quantity
that is renormalization group invariant, then one expects that it
scales like a $\rm (mass)^2$ in the continuum limit. One additionally
measures a second quantity, e.g. the correlation length $\xi$ and
builds the dimensionless product $\chi_t \, \xi^2$ which should go to
a constant in the limit $\xi \rightarrow \infty$. Earlier
Monte Carlo calculations do not show convincingly whether this is the
case. Furthermore, perturbative considerations indicate that in
the O(3) model there might be a problem with the topological
susceptibility in the continuum limit.
One may calculate the contribution of instantons in the continuum
using a semiclassical expansion. The probability density to find an
instanton with topological charge $Q = 1$ and size $\rho \ll 1/
\Lambda$ is \cite{JEVICKI,SCHWAB}:
\begin{equation}
P_1 \sim \frac{d\rho}{\rho},
\label{scaleint}
\end{equation}
where $\Lambda$ is the scale parameter of the model. Using the
renormalization group one can show, that eq.~(\ref{scaleint}) is exact
as $\rho \rightarrow 0$ \cite{LUSCHER1} assuming the small instantons
form a dilute gas. The infrared divergence at $\rho = 0$ in
eq.~(\ref{scaleint}) indicates, however, that this assumption is not
true: small instantons are not suppressed, but contribute strongly to
the susceptibility. Such a dominance of small instantons is also
indicated by numerical studies, trying to determine the instanton size
distribution with different methods
\cite{MICHAEL,FARCHIONI1}.
On the lattice, however, not the whole range of the instanton radius
is probed. The lattice cuts the contribution of small instantons
because there is a smallest possible radius before the instanton falls
through the lattice (c.f. figure~\ref{rhovar}). Hence the measured
topological susceptibility is always finite. It is not excluded,
however, that it raises boundlessly with increasing correlation length
as the perturbative considerations suggest.
\subsection{Numerical Results}
\label{results}
Using the perfect lattice action and the perfect charge we performed
extensive Monte Carlo simulations at correlation lengths in the range
$\xi \in (2 - 60)$.
In order to avoid finite size effects we kept the ratio $L/\xi\approx 6$
constant. The correlation length was obtained from the long distance
fall off of the zero momentum correlation function.
We determined the topological charge using both the geometric
definition and the definition of the FP charge given in
section~\ref{tcharge}. For the measurement of the FP charge we used the
geometric definition of the charge on a finer lattice of the multigrid
with the Monte Carlo generated lattice as coarsest level.
To determine the configuration ${\bf S}({\bf R})$ on the fine lattice one can
either minimize the FP equation (which is very time consuming) or use
the parametrization of the dependence on ${\bf R}$ given in
section~\ref{parfield}. We denote the corresponding charges $Q_{\rm
coarse}$ for the geometric charge, $Q_{\rm par.\,1.\,level}$
for the charge measured on the first finer level using the
parametrization of the fine field etc.
For two $\beta$ values we compared the results of using the
parametrization on a finer level and of minimizing on a multigrid.
As is shown in table~\ref{test}, the
results were found to be consistent within the statistical errors.
This confirms, that the parametrization
performs well for configurations occurring in a Monte Carlo simulation
(c.f. the discussion at the end of section~\ref{classres}).
In order to test if the result of going to a lower level is already stable,
we also calculated the charge for a few $\beta$ values
on the second finer lattice. As also can be seen in table~\ref{test}, the
values on the first and the second finer level were found to be
consistent within the statistical errors.
Thus it is sufficient to calculate the fine field only
on the first finer level. This is not
astonishing as the maximal angle between two neighbouring spins halves
as one goes one step down to a finer level. So even on the first finer
level the maximal possible angle between two neighbouring spins is
$90^{\circ}$ and there is practically no ambiguity left for the
topological charge.
\begin{table}\centering
\begin{tabular}{|*{6}{r@{}l|}}
\hline
\multicolumn{2}{|c|}{$\beta$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm coarse} \rangle $} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm par.\, 1.\, level} \rangle$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm par.\, 2.\, level} \rangle$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm min.\, 1.\, level} \rangle$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm min.\, 2.\, level} \rangle$} \\
\hline
0.&6 & 2.&56(5) & 1.&96(4) & 1.&93(4) & & & & \\
0.&7 & 3.&39(4) & 2.&48(3) & 2.&45(3) & 2.&45(3) & 2.&45(3) \\
0.&85 & 5.&62(11) & 4.&31(9) & 4.&30(9) & & & & \\
1.&0 & 6.&38(8) & 4.&86(6) & 4.&82(6) & & & & \\
1.&0 & 6.&33(6) & 4.&78(5) & & & 4.&75(5) & & \\
\hline
\end{tabular}
\caption{Results of test MC Simulations in the O(3) model indicating that it
is sufficient to measure the charge only on the first finer parametrized
level. Minimizing or going to a lower level yields the same result.}
\label{test}
\end{table}
In table~\ref{mcresults} we report the results of the simulations for
the correlation length $\xi$, the geometric charge $\langle Q_{\rm
coarse}^2 \rangle$ and the FP charge $\langle Q^2_{\rm 1. level}
\rangle$ evaluated on the first parametrized finer level. Using the
geometric charge as well as the FP charge, we build the dimensionless
quantity $\chi^t\cdot \xi^2$ of topological
susceptibility and correlation length.
\begin{table}\centering
\begin{tabular}{|r@{.}l|r|*{5}{r@{.}l|}}
\hline
\multicolumn{2}{|c|}{$\beta$} & L & \multicolumn{2}{c|}{$\xi$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm coarse} \rangle $} &
\multicolumn{2}{|c|}{$\chi^t_{\rm coarse} \cdot \xi^2$} &
\multicolumn{2}{|c|}{$\langle Q^2_{\rm 1. level} \rangle$} &
\multicolumn{2}{|c|}{$\chi^t_{\rm 1. level} \cdot \xi^2$} \\
\hline
0&51 & 10 & 1&6049(42) & 1&87(1) & 0&0482(4) & 1&329(8) & 0&0342(3) \\
0&6 & 14 & 2&1960(46) & 2&57(2) & 0&0631(6) & 1&90(2) & 0&0467(4) \\
0&685 & 20 & 3&012(14) & 3&57(3) & 0&0810(10) & 2&65(2) & 0&0601(7) \\
0&7 & 20 & 3&186(15) & 3&36(3) & 0&0852(11) & 2&52(2) & 0&0640(8) \\
0&85 & 40 & 6&057(17) & 5&73(4) & 0&1314(12) & 4&38(3) & 0&1004(9) \\
1&0 & 70 & 12&156(34) & 6&36(5) & 0&1918(19) & 4&80(4) & 0&1448(15) \\
1&1 & 130 & 20&397(86) & 10&04(12) & 0&2472(36) & 7&69(9) & 0&1893(27) \\
1&2 & 180 & 34&44(30) & 8&23(14) & 0&3013(73) & 6&11(10) & 0&2237(53) \\
1&3 & 340 & 58&06(37) & 12&01(29) & 0&3502(96) & 8&94(22) & 0&2607(72) \\
\hline
\end{tabular}
\caption{Results of MC Simulations with the FP action.}
\label{mcresults}
\end{table}
Figure~\ref{noscaling} shows the results for the topological
susceptibility. Clearly, no scaling is seen even at correlation
lengths as large as 60. Both curves, the one for the geometric charge
and the one for the FP charge, are rising and no flattening occurs at
the largest correlation lengths. There is a significant difference
between the topological susceptibility built with the geometric charge
and the one with the FP charge.
The value of the geometric charge lies several standard deviations
above the value of the FP charge. Furthermore, the difference is slowly
growing with increasing correlation length.
\begin{figure}[htb]
\begin{center}
\leavevmode
\epsfxsize=110mm
\epsfbox{noscaling.eps}
\vskip -5mm
\caption{Results of Monte Carlo measurements of the topological
susceptibility at different correlation lengths.}
\label{noscaling}
\end{center}
\end{figure}
\section{Conclusion and Outlook}
\label{discussion}
The definition of the FP topological charge is based on the FP field
operator. The FP field can be evaluated to any precision desired by
solving the iterated FP equation on a $k$ level multigrid. As was
demonstrated by fig.~\ref{rhovar} and can be shown analytically for
sufficient large multigrids the FP charge together with the FP action
has no lattice defects whatsoever.
For use in MC simulations however a parametrization of the action and
the charge are needed. Instead of parametrizing the charge directly,
we have parametrized the solution of the FP equation which is to be
iterated to obtain the FP field. The accuracy of the parametrizations
have been rechecked in MC simulations.
The partition function of the lattice $\sigma$--model is --- as
indicated by a semiclassical approximation \cite{JEVICKI} ---
dominated by small sized topological excitations. These unphysical
fluctuations are the cause for the divergence which is seen in
fig.~\ref{noscaling}.
Although we studied the $\sigma$--model in this paper, the methods
derived are quite general and can be applied to other asymptotically
free theories such as $\mbox{CP}^{\mbox{\rm \tiny N-1}}$ models or
SU(N) gauge models. For instance, in $\mbox{CP}^{\mbox{\rm \tiny
N-1}}$ models a lowest order semiclassical approximation estimates for
instanton contributions \cite{JEVICKI,SCHWAB}
\begin{equation}
I \sim \int \, d\rho \, \rho^{\mbox{\tiny N}-3}.
\end{equation}
If $\mbox{N} \geq 4$, the contribution of short distance topological
excitations is small relative to that of the physical ones and we
expect to see a scaling of the topological susceptibility according to
the perturbative RG. In fact in the $\mbox{CP}^{3}$ model, which is
studied by one of us
\cite{BURKHALTER}, the topological susceptibility already exhibits the
expected scaling behaviour.
\noindent{\large \bf Acknowledgments }
\noindent We wish to thank T.~DeGrand, A.~Hasenfratz, A.~Papa and
U.~Wiese for useful discussions.
|
\section{Introduction}
In equilibrium, the superconducting Meissner phase of a bulk type-I
superconducting sample exists below a thermodynamic critical field $H_c$
(or below $H_{c1}$ in a type-II superconductor); above this field the
sample reverts to the normal phase (or the flux lattice phase in a
type-II superconductor). However, because the phase transition
in both cases is first-order it is possible to superheat the Meissner
phase and delay the transition to fields well above $H_c$ or $H_{c1}$.
This superheated, metastable Meissner phase is
eventually destroyed at a maximum superheating field $H_{\rm sh}$.
Understanding the origin and stability of the superheated state is the
first step in providing a complete description of the time-dependent
collapse of the Meissner phase, which is important for many applications
of type-I superconductors. For instance, there have been recent
proposals to use superheated type-I superconductors as detectors
for elementary particles, so that the sample acts as a superconducting
``bubble chamber'' \cite{detectors}. The passage of a sufficiently energetic
particle through the sample would initiate the transition to the
normal state. Measuring the superheating
field also provides one of the few methods of experimentally
determining the Ginzburg-Landau parameter $\kappa$ in type-I
superconductors \cite{burger}.
The precise value of the maximum superheating field may depend upon
extrinsic factors such as defects in the sample and sample preparation
and geometry. If these effects can be minimized
then the limit of superheating is determined by the boundaries and geometry
of the sample. The simplest
and most widely studied geometry is a superconducting half-space with
a magnetic field applied parallel to the surface of the superconductor,
which is the geometry considered in the remainder
of this paper. To model the superconductor we will use the
Ginzburg-Landau (GL) equations, which provide an accurate description of
the surface behavior provided the coherence length $\xi$ is large compared
to microscopic length scales \cite{degennes}.
Previous studies of superheating in type-I superconductors have used a variety
of heuristic methods to determine the behavior of the GL equations
near the surface. Ginzburg \cite{ginzburg58} inferred the leading
$\kappa^{-1/2}$ dependence of the superheating field from the form of the
Ginzburg-Landau equations. Apparently unaware of
Ginzburg's work, the Orsay group \cite{orsay} used a variational argument
to show that $H_{\rm sh}/H_c \approx 2^{-1/4} \kappa^{-1/2}$.
By combining an ingenious guess for the behavior of the superconducting
order parameter near the surface with a variational calculation,
Parr \cite{parr76} was able to calculate the next order correction to
the Orsay group's result. In addition to this analytical work there
has also been a great deal of numerical
work on solving the GL equations in small-$\kappa$ limit
\cite{matricon67,kramer68,fink69,bolley}, which is reviewed in
Ref.~\cite{fink}. The numerical results appear to confirm at least
some of the analytical work, although admittedly over a somewhat
restricted range of $\kappa$. One deficiency common to all of the
previous analytical approaches is an {\it ad-hoc} construction
of approximate solutions of the GL equations, leaving us without a
procedure for systematically improving upon these approximations.
In addition, the issue of the {\it stability} of the solutions in
the small-$\kappa$ limit seems not to have been addressed
rigorously (for one attempt see \cite{galaiko68}).
In this paper we re-examine the problem of superheating in type-I
superconductors by using the method of matched asymptotic expansions
\cite{bender,vandyke} to solve the GL equations in the small-$\kappa$ limit.
This method was originally developed to treat
boundary layer problems in fluid mechanics \cite{vandyke} in a controlled
and systematic fashion, and
is particularly well suited to the superheating problem, as all of
the technical difficulties arise due to a ``boundary layer'' at the surface.
Using this method we can calculate the superheating
field in the small-$\kappa$ limit as an asymptotic expansion in
powers of $\kappa^{1/2}$, construct
uniform asymptotic expansions (i.e., expansions valid for all $x$ as
$\kappa\rightarrow 0$) for the order parameter and magnetic field,
determine the stability of the solutions, and treat
nonlocal electrodynamic effects. Where appropriate we compare our
asymptotic results against numerical solutions of the GL equations,
and we generally find excellent agreement.
We have chosen to present our results in detail, as the methods are probably
unfamiliar to most physicists.
Matched asymptotic expansions have recently been used by Chapman
\cite{chapman95} to study superheating in the large-$\kappa$ limit, and our
results complement his work. This paper, taken together with Chapman's work,
demonstrates that these perturbation
methods can provide a powerful calculational tool for solving problems
in inhomogeneous or nonequilibrium superconductivity.
The remainder of this paper is organized as follows. In Sec.~II we describe
our numerical methods for solving the GL equations. In Sec.~III we
develop the method of matched asymptotic expansions for the solution
of the GL equations in the small-$\kappa$ limit, and determine the
first three terms in the expansion of $H_{\rm sh}$ in powers of
$\kappa^{1/2}$. In addition, we construct uniform expansions for the
order parameter and magnetic field and compare them against our
asymptotic results. In Sec.~IV we examine the second variation of the
GL free energy, $\delta^2 {\cal F}$, in order to determine the stability
of our solutions. This is done for both one and two-dimensional
perturbations. In Sec.~V the method is generalized to treat
nonlocal electrodynamics. Section~VII compares Chapman's asymptotic
expansion for $H_{\rm sh}$ for large-$\kappa$ with our numerical results,
and we find remarkably good agreement. Finally, Sec.~VIII is a summary
and discussion of our results.
\section{Numerical methods}
The GL free energy of a superconducting sample occupying the
half-space $x>0$ is
\begin{equation}
{\cal F}[f,q] = \int_{x>0}\:d^3r\left[\frac{1}{\kappa^2}({\bf\nabla} f)^2+
\frac{1}{2}(1-f^2)^2+f^2 {\bf q}^2+({\bf H}_a-{\bf\nabla\times q})^2\right]
\label{energy}
\end{equation}
where $\kappa$ is the GL parameter, $f$ is the amplitude of
the superconducting order parameter, ${\bf q}$ is the gauge-invariant
vector potential (${\bf h}={\bf\nabla\times q}$), and ${\bf H}_a$ is the
applied magnetic field. The lengths are in units
of the penetration depth $\lambda$ and fields are in units of $\sqrt{2}H_c$.
Minimizing this expression with respect to both $f$ and ${\bf q}$ results
in the GL equations. In one dimension, with $f=f(x)$ and
${\bf q}=(0,q(x),0)$, these equations are
\begin{equation}
{1\over \kappa^2} f'' - q^2 f + f - f^3 = 0,
\label{GL1}
\end{equation}
\begin{equation}
q'' - f^2 q = 0 ,
\label{GL2}
\end{equation}
\begin{equation}
h = q'.
\label{GL3}
\end{equation}
The task at hand is to solve these equations numerically for a
superconducting half-space and to find the largest possible applied field
($H_{\rm sh}$) which permits a superconducting solution. To
insure that no current passes through the boundary at $x=0$ and that
the sample is totally superconducting infinitely far from the surface,
we impose the boundary conditions
\begin{equation}
f'(0) = 0, \qquad f(x)\rightarrow 1\ \ {\rm as} \ x\rightarrow \infty.
\label{BC1}
\end{equation}
Since the field at the surface must equal the applied field $H_a$, and the
field infinitely far from the surface must equal 0, we impose the boundary
conditions
\begin{equation}
h(0) = H_a, \qquad q(x) \rightarrow 0 \ \ {\rm as} \ x\rightarrow \infty.
\label{BC2}
\end{equation}
For $\kappa\rightarrow0$, we rescale the equations as $x'=\kappa x$ making
the new unit of length the correlation length~$\xi$. Since $\xi\gg\lambda$
in this limit, a numerical solution over a domain much larger than $\xi$
would insure that the regions of rapid change for $f$ and $h$ would be
included. (For small $\kappa$, we find that solving for $x'<500$ is
sufficient.) In the large $\kappa$ limit, we use the rescaled equations
again, but we increase the size of the domain depending on the value of
$\kappa$. (The equations must be solved for domains as large as $x'<10^4$
for values of $\kappa\sim10^3$.)
The equations can be solved using the relaxation method \cite{press}.
By replacing these ordinary differential equations with finite
difference equations, one can start with a guess to the solution and
iterate using a multi-dimensional Newton's method until it relaxes
to the true solution. In order to more accurately pick up the detail near
the boundary, we choose a grid of discrete points with a higher density
near $x=0$. In particular we choose a density which roughly varies as
the inverse of the distance from the boundary. (For low $\kappa$ our
density, in units of mesh points per coherence length, varies approximately
from $10^7$ near the boundary to $10^3$ at the farthest point from the
boundary, while for high $\kappa$ it varies from $10^5$ to $10^{-2}$.)
$H_{\rm sh}$ can be found in the following way. For a given value of
$\kappa$ an initial guess is made where there is no applied field and the
sample is completely superconducting ($f\equiv1, q\equiv0, h\equiv0$).
The field $H_a$ is then ``turned up'' in small increments. For each value
of $H_a$ a solution is sought using the result from the previous lower
field solution as an initial guess. Eventually a maximum value for $H_a$
is reached, above which one of two things happens: our algorithm fails to
converge to a solution or it converges to the normal (nonsuperconducting
solution). This maximum value of $H_a$ is the numerical result for
$H_{\rm sh}$.
Using this algorithm, $H_{\rm sh}(\kappa)$ can be found for a wide range of
$\kappa$'s. Each run (for a given $\kappa$) takes about 60 cpu minutes
on an IBM RS 6000/370. We find it sufficient for the purposes of this paper
to deal with superheating field values for $10^{-3}<\kappa<10^3$.
\section{Asymptotic expansions for small-$\kappa$}
In this section we will develop an asymptotic expansion for the superheating
field for small-$\kappa$, using the method of matched asymptotic
expansions \cite{bender,vandyke}.
For small-$\kappa$ the dominant length scale is the coherence length $\xi$,
so it is natural to have $\xi$ serve as our unit of length. This is
achieved by rescaling $x$ by $\kappa$, introducing a new dimensionless
coordinate $x'=\kappa x$. The resulting GL equations in these
``outer variables'' are
\begin{equation}
f'' - q^2 f + f - f^3 = 0,
\label{GL4}
\end{equation}
\begin{equation}
\kappa^2 q'' - f^2 q = 0 ,
\label{GL5}
\end{equation}
\begin{equation}
h = \kappa q',
\label{GL6}
\end{equation}
with the primes now denoting differentiations with respect to $x'$.
{\it Outer solution.} In order to obtain the outer solutions expand
$f$, $q$, and $h$ in powers of $\kappa$:
\begin{equation}
f = f_0 + \kappa f_1 + \kappa^2 f_2 + \ldots,
\label{expand1}
\end{equation}
\begin{equation}
q = q_0 + \kappa q_1 + \kappa^2 q_2 + \ldots,
\label{expand2}
\end{equation}
\begin{equation}
h = h_0 + \kappa h_1 + \kappa^2 h_2 + \ldots.
\label{expand3}
\end{equation}
Substituting into Eqs.~(\ref{GL4})-(\ref{GL6}), at $O(1)$ we have
\begin{equation}
f_0'' - q_0^2 f_0 + f_0 - f_0^3=0,
\label{outer1}
\end{equation}
\begin{equation}
- f_0^2 q_0 = 0.
\label{outer2}
\end{equation}
Since we want $f\rightarrow 1$ as $x'\rightarrow \infty$, the only possible
solution to Eq.~(\ref{outer2}) is $q_0 = 0$. We can then immediately
integrate Eq.~(\ref{outer1}),
\begin{equation}
f_0(x') = \tanh \left( {x' + x_0}\over \sqrt{2} \right),
\label{outer3}
\end{equation}
with $x_0$ an arbitrary constant. To $O(\kappa)$, the outer
equations are
\begin{equation}
f_1'' - 2q_0f_0q_1 - q_0^2f_1 + f_1 - 3f_0^2 f_1 = 0,
\label{outer4}
\end{equation}
\begin{equation}
- f_0^2 q_1 - 2 f_0q_0 f_1 = 0,
\label{outer5}
\end{equation}
\begin{equation}
h_0 = 0.
\label{outer6}
\end{equation}
Once again, the only solution to Eq.~(\ref{outer5}) is $q_1 = 0$;
substituting this into Eq.~(\ref{outer4}), we find $f_1 =C_1 f_0'$,
with $C_1$ a constant:
\begin{equation}
f_1 = {C_1\over \sqrt{2}} {\rm sech}^2\left( { x'+x_0 \over \sqrt{2}} \right).
\label{outer7}
\end{equation}
We can continue in this manner; at every order $q_n= 0$, $h_n=0$,
and $f_n = C_n f^{(n)}_0$, with the $C_n$'s constants which are determined
by matching onto the inner solution.
{\it Inner solution.} The outer solution breaks down within a boundary
layer of $O(\kappa)$ near the surface. This suggests introducing a rescaled
inner coordinate $X=x'/\kappa$, so that $X=O(1)$ within the boundary layer.
It is also possible to rescale $f$ and $q$, with the hope that this
will lead to a tractable inner problem. Such a rescaling must lead to
a successful matching of the inner and outer solutions; i.e., the inner
solutions as $X\rightarrow \infty$ must match onto the outer solutions as
$x'\rightarrow 0$. Since $f_0(0) = \tanh (x_0/\sqrt{2})$, then assuming
that $x_0\neq0$ we have $f_0(0)= O(1)$, indicating
that the order parameter should not be rescaled in the inner region;
therefore we set $f(x') = F(X)$ in the inner region. However, from the
outer solution for the vector potential we see that the only constraint
on $q(X)$ in the inner region is that $q(X)\rightarrow 0 $ as
$X\rightarrow\infty$ (presumably exponentially). Therefore, we
are free to rescale $q$ by $\kappa$ in the inner region, hopefully in
a way which simplifies the inner equations. One possibility is
$q(x')=\kappa^{-\alpha} Q(X)$; substituting this into the GL equations,
Eqs.~(\ref{GL4})-(\ref{GL6}), we see that unless $2\alpha$ is an
integer, fractional powers of $\kappa$ will be introduced into the inner
equations, contradicting our expansion of $f$ and $q$ in integer powers of
$\kappa$ in the outer region. Therefore, the simplest assumption is that
$\alpha=1/2$, leading to the following choice for the inner variables:
\begin{equation}
x'= \kappa X,\quad f(x') = F(X),\quad q(x') = \kappa^{-1/2} Q(X), \quad
h(x') = H(X).
\label{inner1}
\end{equation}
In these variables Eqs.~(\ref{GL4})-(\ref{GL6}) become
\begin{equation}
F'' - \kappa Q^2 F + \kappa^2 (F - F^3) = 0,
\label{inner2}
\end{equation}
\begin{equation}
Q'' - F^2 Q = 0,
\label{inner3}
\end{equation}
\begin{equation}
\kappa^{1/2} H = Q',
\label{inner4}
\end{equation}
where now the primes denote differentiation with respect to $X$.
The boundary conditions are
\begin{equation}
F'(0)=0, \qquad H(0) = H_a.
\label{inner4a}
\end{equation}
The next step is to expand the inner solutions in powers of $\kappa$:
\begin{equation}
F = F_0 + \kappa F_1 + \kappa^2 F_2 + \ldots,
\label{inner5}
\end{equation}
\begin{equation}
Q= Q_0 + \kappa Q_1 + \kappa^2 Q_2 + \ldots,
\label{inner6}
\end{equation}
\begin{equation}
H = \kappa^{-1/2} H_0 + \kappa^{1/2} H_1 + \ldots.
\label{inner7}
\end{equation}
Note that there is no term of $O(1)$ in the expansion for $H$, since
we would be unable to match such a term to the outer solution.
Using the boundary condition $H(0)=H_a$ leads to
\begin{equation}
H_a = \kappa^{-1/2} H_0(0) + \kappa^{1/2} H_1(0) + \ \ldots.
\label{inner7a}
\end{equation}
Substituting these expansions into Eqs.~(\ref{inner2})-(\ref{inner4}),
at $O(1)$ we obtain
\begin{equation}
F_0'' = 0, \qquad Q_0'' - F_0^2 Q_0 = 0, \qquad H_0 = Q_0'.
\label{inner9}
\end{equation}
Solving these equations subject to the boundary conditions (\ref{inner4a})
(we also need $Q_0\rightarrow 0 $ as $x\rightarrow \infty$ in order to
match onto the outer solution), we obtain
\begin{equation}
F_{0}(X) = A_0, \qquad Q_0(X) = B_0 e^{-A_0 X},
\qquad H_0(0) = -A_0 B_0,
\label{inner10}
\end{equation}
with $A_0$ and $B_0$ constants. In what follows we will assign $F_n(0) = A_n$
and $Q_n(0)=B_n$ for notational simplicity. The $O(\kappa)$ equations are
\begin{equation}
F_1'' = Q_0^2 F_0, \qquad Q_1'' - F_0^2 Q_1 = 2 F_0 Q_0 F_1,
\qquad H_1 = Q_1'.
\label{inner11}
\end{equation}
Solving with the boundary condition $F_{1}'(0)=0$, we obtain
\begin{equation}
F_1(X) = A_1 + { B_0^2 \over 4 A_0}\left[ 2 A_0 X + e^{-2 A_0 X}
- 1\right],
\label{inner12}
\end{equation}
\begin{eqnarray}
Q_1(X) &=& e^{-A_0 X} \left\{ B_1 - {B_0^3\over 16 A_0^2} \biggl[
1 - e^{-2A_0 X} \right. \nonumber \\
& & \qquad \left. + 16 {A_0^2 A_1 \over B_0^2} X
+ 4 A_0^2 X^2 \biggr] \right\},
\label{inner13}
\end{eqnarray}
\begin{equation}
H_1(0)= -{1\over 8} {B_0^3 \over A_0} - A_0B_1 - A_1 B_0.
\label{inner13a}
\end{equation}
Finally, to $O(\kappa^2)$ we have for $F_2$
\begin{equation}
F_2'' = - F_0 + F_0^3 + 2Q_0Q_1F_0 + Q_0^2F_1,
\label{next1}
\end{equation}
the solution of which (with $F_2'(0) = 0$) is
\begin{eqnarray}
F_2(X) &=& {17 \over 128}{B_0^4\over A_0^3} + {1\over 4} {B_0^2 A_1 \over
A_0^2}
- {1\over 2} {B_0 B_1 \over A_0} + A_2 + \left( B_0B_1 - {3\over 32}
{B_0^4 \over A_0^2} \right) X - {1\over 2} A_0(1-A_0^2) X^2 \nonumber \\
& & \quad + \left[ {1\over2} {B_0 B_1 \over A_0}
- {1\over 4}{B_0^2 A_1\over A_0^2}- {5\over 32}{B_0^4\over A_0^3}
- \left( {1\over 8}{B_0^4\over A_0^2} +{1\over 2} {B_0^2 A_1\over
A_0}\right)X
- {1\over 8} {B_0^4\over A_0} X^2\right] e^{-2A_0 X} \nonumber \\
& & \qquad + {3\over 128} {B_0^4 \over A_0^3} e^{-4A_0 X}.
\label{inner14}
\end{eqnarray}
The expression for $Q_2$ is even more unwieldy, and is not needed in
what follows.
{\it Matching.} To determine the various integration constants which have
been introduced we must match the inner solution to the outer solution.
Since the outer solution for $q$ is simply $q=0$, and all of our
inner solutions decay exponentially for large $X$, the matching is
automatically satisfied for $q$, as well as for the magnetic field $h$.
To match the inner and outer solutions
for the order parameter, we are guided by the {\it van Dyke matching principle}
\cite{vandyke}, which states that the $m$ term inner expansion of the
$n$ term outer solution should match onto the $n$ term outer expansion of the
$m$ term inner solution. In our case we will take $m=3$ and $n=2$.
Therefore, write the two term outer solution
$f_0(x') + \kappa f_1(x')$ in terms of the inner
variable $X$, and expand for small $\kappa$, keeping the first three terms in
the expansion in powers of $\kappa$:
\begin{eqnarray}
f_0(\kappa X) + \kappa f_1(\kappa X) & \sim&
\tanh \left({x_0\over \sqrt{2}}\right)
+ \kappa\, {\rm sech}^2 \left({x_0 \over \sqrt{2}}\right)
{1\over \sqrt{2}}\left[ C_1 + X \right] \nonumber \\
& & \qquad + \kappa^2 {\rm sech}^2 \left({x_0 \over \sqrt{2}}\right)
\tanh \left({x_0\over \sqrt{2}}\right)\left[- C_1 X
- {X^2\over 2}\right].
\label{match1}
\end{eqnarray}
Next, write the three term inner solution
$F_0(X) + \kappa F_1(X)+\kappa^2 F_2(X)$ in terms
of the outer variable $x'$, and expand for small $\kappa$, this time keeping
the first two terms of the expansion:
\begin{eqnarray}
F_0(x'/\kappa) + \kappa F_1(x'/\kappa) + \kappa^2 F_2(x'/\kappa) & \sim&
A_0 + {B_0^2 \over 2} x' - {1\over 2} A_0(1-A_0^2) x'^2 \nonumber \\
& & \qquad+ \kappa \left[ A_1-{B_0^2 \over 4 A_0} + \left( B_0 B_1 - {3\over
32}
{B_0^4 \over A_0^2}\right)x' \right].
\label{match2}
\end{eqnarray}
By writing both expressions in terms of $x'$, and equating the various
coefficients of $x'$ and $\kappa$, we see that the expansions do indeed
match if we choose
\begin{equation}
A_0 = \tanh\left({x_0\over \sqrt{2}}\right),
\label{match3}
\end{equation}
\begin{equation}
\qquad B_0 =- 2^{1/4} {\rm sech}\left({x_0 \over \sqrt{2}}\right)
= -2^{1/4} ( 1- A_0^2)^{1/2},
\label{match4}
\end{equation}
\begin{equation}
A_1 = {B_0^2 \over 4A_0} + {\rm sech}^2\left({x_0 \over \sqrt{2}}\right)
{C_1 \over \sqrt{2}}
= {\sqrt{2}\over 4} {1 - A_0^2 \over A_0} + (1-A_0^2) {C_1\over \sqrt{2}},
\label{match5}
\end{equation}
\begin{equation}
B_1 = {3\over 32} {B_0^3 \over A_0^2} - {\sqrt{2} A_0(1-A_0^2) \over B_0}
{C_1\over \sqrt{2}}.
\label{match6}
\end{equation}
Eliminating $C_1$,
\begin{equation}
B_1 = -{\sqrt{2} A_0 A_1 \over B_0} + {3\over 32} {B_0^3\over A_0^2}
+ {1\over 2} {1-A_0^2 \over B_0}.
\label{match7}
\end{equation}
Substituting into our expressions for $H_0(0)$ and $H_1(0)$ from
Eqs.~(\ref{inner10}) and (\ref{inner13a}), we obtain
\begin{equation}
H_0(0) = 2^{1/4} A_0 ( 1- A_0^2)^{1/2},
\label{match8}
\end{equation}
\begin{equation}
H_1(0) = {2^{3/4} \over 64} {(2A_0^2 +14)(1-A_0^2)^{1/2} \over A_0}
- {2^{1/4}(2A_0^2 -1) \over (1-A_0^2)^{1/2}} A_1.
\label{match9}
\end{equation}
In order to calculate the superheating field (or, more correctly, the
{\it maximum} superheating field), we need to maximize $H_0(0)$ and
$H_1(0)$ with respect to $A_0$ and $A_1$.
Maximizing $H_0(0)$ with respect to $A_0$, we find that the maximum occurs at
$A_0^* = 1/\sqrt{2}$, $B_0^* = -2^{-1/4}$, so that $H_0 (0) = 2^{-3/4}$.
Substituting this
result into $H_1(0)$, we find the surprising result that the coefficient
of $A_1$ is zero, and $H_1(0) = 2^{3/4} 15/64$.
Our superheating field is then
\begin{equation}
H_{\rm sh} = 2^{-3/4}\kappa^{-1/2} \left[ 1 + {15 \sqrt{2} \over 32} \kappa
+ O(\kappa^2) \right].
\label{final1}
\end{equation}
In order to determine $A_1$ we need to proceed to a higher order
calculation. The method is the same as before, although the algebra
quickly becomes tedious; we have used the
computer algebra system {\it Maple V} to organize the expansion.
The results from a six term inner expansion are summarized in
Table~\ref{table1}. Including the next
order term in the expansion in the superheating field, we have
\begin{equation}
H_{\rm sh} = 2^{-3/4}\kappa^{-1/2} \left[ 1 + {15 \sqrt{2} \over 32} \kappa
- {325\over 1024} \kappa^2 + O(\kappa^3) \right].
\label{final2}
\end{equation}
The first term is exactly the result obtained by the Orsay group
\cite{ginzburg58,orsay}, who used a variational argument to obtain their
result. The second term is identical to the result
obtained by Parr \cite{parr76}. Parr combined an inspired guess for the
behavior of the order parameter near the surface with a variational
calculation in order to obtain his result. It is interesting to note
that our result for $A_1$ also agrees with Parr's result. The advantage
of the method of matched asymptotic expansions is that we can make this
expansion systematic, and therefore in principle carry out this expansion
as far as we wish. The third term in Eq.~(\ref{final2}) is one of the
new results of this paper; the fourth and fifth terms are included in
Table~\ref{table1}. With the five-term expansion for $H_{\rm sh}$
it is possible to employ resummation techniques to improve the
expansion. For instance, the $[2,2]$ Pad\'e approximant \cite{bender} is
\begin{equation}
H_{\rm sh}^{\rm Pade} = 2^{-3/4}\kappa^{-1/2}
{ 1 + 5.4447812\,\kappa + 4.2181012\,\kappa^2 \over
1 + 4.7818686\, \kappa + 1.3655230\, \kappa^2}.
\label{pade}
\end{equation}
In Fig.~\ref{Hshplot} we compare the numerically calculated superheating
field against the one, two, and three term asymptotic expansions.
The one term (i.e., the Orsay group) result never seems particularly
accurate. There is a marked improvement with the two term expansion, with
the three term expansion offering only a modest additional improvement.
The $[2,2]$ Pad\'e approximant agrees with the numerical data to within about
1\% all the way to $\kappa=1$.
{\it Uniform solutions.} From the inner and outer expansions it is
possible to construct {\it uniform} solutions, which are asymptotically
correct for all $x$ as $\kappa\rightarrow 0$. To do this we simply add
the inner and outer solutions of a given order, which guarantees the correct
behavior in the outer region as well as in the boundary layer. However,
this would produce a result which was $2 f_{\rm match}$ in the
matching region, so we need to subtract $f_{\rm match}$ in order to obtain
the correct behavior in this region. As an example, we will construct the
2-term uniform solution for the order parameter. Adding the two-term
outer solution, $f_0(x')+ \kappa f_1(x')$, to the two-term inner solution,
$F_0(X) + \kappa F_1(X)$, subtracting the solution in the matching region,
which is $1/\sqrt{2} + (\sqrt{2}/4)\kappa X - (15/32)\kappa$, and writing
the entire combination in terms of the original variable $x$ (which is the
same as $X$), we obtain
\begin{equation}
f_{\rm unif,2}(x) = \tanh\left( {\kappa x + x_0 \over \sqrt{2}}\right)
- {15\over 16} \kappa\, {\rm sech}^2\left( {\kappa x + x_0 \over
\sqrt{2}}\right)
+ {\kappa \over 4} e^{-\sqrt{2} x}.
\label{unif}
\end{equation}
As $x\rightarrow \infty$, $f_{\rm unif,2}(x)\rightarrow 1$; also,
$f_{\rm unif,2}(0) = 1/\sqrt{2} - (7/32)\kappa$, as we expect. However,
$f_{\rm unif,2}'(0) = (15/64)\kappa^2$, so that the zero-derivative boundary
condition is only satisfied to $O(\kappa)$.
In Fig.~\ref{matching_fig} and Fig.~\ref{matching_fig2} we compare the
numerically calculated order parameter and magnetic field with the
two term outer solutions and the three term inner solutions.
The agreement is quite good for $\kappa=0.1$, with deviations appearing
at $\kappa=0.5$. These figures also illustrate the existence of a
matching region where the inner and outer solutions overlap; this
region grows as $\kappa\rightarrow 0$.
Lastly, we show in Fig.~\ref{uniform_fig} how the two term uniform expansion
constructed earlier supplies a uniform approximation to the order parameter and
magnetic field over the whole region.
\section{Stability analysis of the solutions}
Having obtained an asymptotic expansion for the superheating field
$H_{\rm sh}$ in powers of
$\kappa^{1/2}$, we now examine the stability of the solution with respect
to infinitesimal perturbations by studying the second variation
of the free energy, $\delta^2 {\cal F}$. Perturbations with
$\delta^2 {\cal F}>0$ correspond to stable solutions, while those
with $\delta^2 {\cal F}<0$ correspond to unstable solutions. We will again use
the method of matched asymptotics to solve for the eigenfunctions of the
linear stability operator. We first determine the stability in the
simpler one-dimensional situation then we discuss the two-dimensional case.
\subsection{Stability with respect to one-dimensional perturbations}
If we perturb the extremal solution $(f,q)$ of the GL equations by allowing
$f(x)\rightarrow f(x)+\tilde{f}(x)$ and
$q(x)\rightarrow q(x)+\tilde{q}(x)$, then the
second variation of the free energy functional is
\begin{equation}
\delta^2 {\cal F} =
\int_0^{\infty}dx\left[\frac{1}{\kappa^2}\tilde{f}'^2+(3f^2+q^2-1)
\tilde{f}^2+4fq\tilde{f}\tilde{q}+f^2\tilde{q}^2+
\tilde{q}'^2\right].
\label{stab1}
\end{equation}
The boundary conditions on $\tilde{f}$ and $\tilde{q}$ should be chosen
so as to not perturb $f$ and $h$ at the surface, so that
\begin{equation}
\tilde{f}'(0) = \tilde{q}'(0)=0, \qquad
\tilde{f}(\infty)=\tilde{q}(\infty)=0.
\label{stab_boundary}
\end{equation}
We can then integrate Eq.~(\ref{stab1}) by parts to obtain
\begin{equation}
\delta^2 {\cal F} = \int_0^{\infty}dx\left[\tilde{f}\left(
-\frac{1}{\kappa^2}\frac{d^2}{dx^2} +
q^2+3f^2-1\right)\tilde{f} +
\tilde{q}\left(-\frac{d^2}{dx^2}+
f^2\right)\tilde{q}+4qf\tilde{q}\tilde{f}\right].
\end{equation}
This quadratic form can be conveniently written as
\begin{equation}
\delta^2{\cal F} = \int_0^{\infty}dx\, (\tilde{f},\tilde{q})\hat{L}_1
\left(\begin{array}{c}\tilde{f}\\ \tilde{q}\end{array}\right)
\end{equation}
where $\hat{L}_1$ is the self-adjoint linear operator
\begin{equation}
\hat{L}_1\left(\begin{array}{c}\tilde{f}\\ \tilde{q}\end{array}\right) =
\left(\begin{array}{cc}
-\frac{1}{\kappa^2}\frac{d^2}{dx^2}+q^2+3f^2-1 & 2fq \\
2fq & -\frac{d^2}{dx^2} +f^2\end{array}\right)
\left(\begin{array}{c}\tilde{f}\\ \tilde{q}\end{array}\right).
\end{equation}
In order to analyze the stability, expand $\tilde{f}$ and $\tilde{q}$ as
\begin{equation}
\left(\begin{array}{c}\tilde{f}\\ \tilde{q}\end{array}\right) =
\sum_{n} c_n \left(\begin{array}{c}\tilde{f}_n\\ \tilde{q}_n\end{array}\right),
\label{eigen1}
\end{equation}
where the $c_n$'s are real constants, and $(\tilde{f}_n,\tilde{q}_n)$
is a normalized eigenfunction of $\hat{L}_1$ with eigenvalue $E_n$:
\begin{equation}
\hat{L}_1\left(\begin{array}{c}\tilde{f}_n\\ \tilde{q}_n\end{array}\right) =
E_n\left(\begin{array}{c}\tilde{f}_n\\ \tilde{q}_n\end{array}\right).
\end{equation}
Then
\begin{equation}
\delta^2{\cal F} = \sum_n E_n c_n^2.
\end{equation}
The second variation $\delta^2 {\cal F}$ ceases to be positive-definite
when the lowest eigenvalue first becomes negative, indicating that the
corresponding solutions $(f,q)$ of the GL equations are unstable.
Therefore the entire issue of the stability of the solutions has
been reduced to finding the eigenvalue spectrum of the linear stability
operator $\hat{L}_1$, which in the $\kappa\rightarrow 0$ limit can
be studied using matched asymptotic expansions.
{\it Outer solution.} The outer equations for $(\tilde{f},\tilde{q})$
are rescaled with
$x'=\kappa x$ as before to yield (we will drop the subscript $n$
for notational convenience)
\begin{equation}
-\tilde{f}''+(3f^2+q^2-1)\tilde{f}+2fq\tilde{q} = E\tilde{f},
\label{2d_perturb1}
\end{equation}
\begin{equation}
-\kappa^2\tilde{q}''+ f^2\tilde{q} + 2fq\tilde{f} = E\tilde{q}.
\label{2d_perturb2}
\end{equation}
Expanding $\tilde{f}$, $\tilde{q}$, and $E$ in powers of $\kappa$, and
recalling that $q=0$ to all orders in $\kappa$ in the outer region,
we have at leading order
\begin{equation}
-\tilde{f}_0''+(3f_0^2-1)\tilde{f}_0 = E_0\tilde{f}_0,
\label{perturb1}
\end{equation}
where $f_0 = \tanh\left(\frac{x'+x_0}{\sqrt{2}}\right)$.
By changing variables to $y = \tanh\left(\frac{x'+x_0}{\sqrt{2}}\right)$
we see that the solution of Eq.~(\ref{perturb1}) is the associated
Legendre function of the first kind:
\begin{equation}
\tilde{f}_0(x') = c_0 P_2^{\mu}\left[\tanh\left(\frac{x'+x_0}{\sqrt{2}}
\right)\right],
\label{legendre}
\end{equation}
where $\mu = -\sqrt{2(2-E_0)}$ and $c_0$ is a constant.
The leading order solution for $\tilde{q}$ is $\tilde{q}_0=0$.
{\it Inner solution.} To obtain the inner equations, we rescale as in
Eq.~(\ref{inner1}), with the perturbations rescaled as
\begin{equation}
\tilde{f}(x') = \tilde{F}(X),\quad \tilde{q}(x') = \kappa^{-1/2}\tilde{Q}(X),
\label{perturb2}
\end{equation}
such that
\begin{eqnarray}
-\frac{1}{\kappa^2}\tilde{F}''+
(3F^2+\frac{1}{\kappa}Q^2-1)\tilde{F}+\frac{1}{\kappa}2 FQ
\tilde{Q} = E\tilde{F}, \\
-\tilde{Q}''+ F^2\tilde{Q} + 2FQ\tilde{F} = E\tilde{Q}.
\end{eqnarray}
To leading order, $\tilde{F_0}'' = 0$, so that $\tilde{F_0} = a_0$, with
$a_0$ a constant. The leading order equation for
the variation in $Q$ is
\begin{equation}
-\tilde{Q}_0'' + 2F_0Q_0\tilde{F}_0+(F_0^2-E_0)\tilde{Q}_0 = 0.
\end{equation}
The solution which satisfies the boundary condition $\tilde{Q}'(0)=0$ is
\begin{equation}
\tilde{Q}_0(X) =
\frac{2a_0A_0B_0}{E_0}\left(e^{-A_0X}-\frac{A_0}{\sqrt{A_0^2-E_0}}
e^{-\sqrt{A_0^2-E_0}X}\right).
\end{equation}
At $O(\kappa)$ we find
\begin{equation}
\tilde{F}_1'' = Q_0^2\tilde{F}_0+2F_0Q_0\tilde{Q}_0,
\end{equation}
with the solution
\begin{equation}
\tilde{F}_1(X) =\displaystyle
a_1+ a_0B_0^2\left[\frac{E_0+4A_0^2}{4A_0^2E_0}e^{-2A_0X}-
\frac{4A_0^2}{E_0}\frac{e^{-(A_0+\sqrt{A_0^2-E_0})X}}
{(A_0+\sqrt{A_0^2-E_0})^2\sqrt{A_0^2-E_0}}\right]
\qquad\atop\quad\displaystyle
+a_0B_0^2\left[\frac{E_0+4A_0^2}{2A_0E_0}-
\frac{4A_0^3/E_0}{(A_0+\sqrt{A_0^2-E_0})\sqrt{A_0^2-E_0}}\right]X.
\end{equation}
We now have enough terms in the inner and outer region for a nontrivial match.
{\it Matching.} We complete the matching of the inner and outer perturbations
to obtain the eigenvalue, $E_0$. Performing a two term inner expansion of
the one term outer solution, we have
\begin{equation}
\tilde{f}_0(\kappa X) \sim c_0 \left[ P_2^{\mu}(A_0) +
\frac{1}{\sqrt{2}}\mbox{sech}^2(x_0/\sqrt{2})
\frac{dP_2^{\mu}(A_0)}{d A_0 } \kappa X \right],
\end{equation}
where we have used $\tanh(x_0/\sqrt{2}) = A_0$.
Next, the one term outer expansion of the two term inner solution is
\begin{equation}
\tilde{F}_0(x'/\kappa) + \kappa \tilde{F}_1(x'/\kappa) \sim
a_0+\frac{a_0 2^{1/2} (1-A_0^2)}{E_0}\left[
\frac{E_0+4A_0^2}{2A_0}
-\frac{4A_0^3}{(A_0+\sqrt{A_0^2-E_0})\sqrt{A_0^2-E_0}}\right] x',
\end{equation}
where we have used $B_0 = -2^{1/4}(1-A_0^2)^{1/2}$.
Matching the two expansions using the van Dyke matching principle yields
\begin{equation}
c_0 = \frac{a_0}{P_2^{\mu}(A_0)},
\end{equation}
\begin{equation}
\frac{1}{P_2^{\mu}(A_0)}\frac{dP_2^{\mu}(A_0)}{dA_0} =
\frac{2}{E_0}\left[
\frac{E_0+4A_0^2}{2A_0}
-\frac{4A_0^3}{(A_0+\sqrt{A_0^2-E_0})\sqrt{A_0^2-E_0}}\right].
\label{metastable}
\end{equation}
The last equation is a rather complicated implicit equation for the
eigenvalue $E_0(A_0)$, which generally must be solved numerically.
However, when $A_0=1/\sqrt{2}$ we find $E_0=0$, corresponding to the
critical case, with $E>0$ for $A_0>1/\sqrt{2}$. The numerical evaluation
of Eq.~(\ref{metastable}) is shown in Fig.~\ref{stable_plot}. Therefore, we
see that our maximum superheating field (at lowest order) corresponds to
the limit of metastability for these one-dimensional perturbations.
In Fig.~\ref{nose} we show $A_0$ as a function of the lowest order
magnetic field at the surface, $H_0$, from Eq.~(\ref{match8}).
The stability analysis of this section shows that only the upper
branch of this double valued function corresponds to solutions which
are locally stable, with the field at the ``nose'' being the superheating
field.
\subsection{Stability with respect to two-dimensional perturbations}
We next turn to the stability of the solutions with respect to two
dimensional perturbations. If we perturb the extremal solution $(f,{\bf q})$
of the GL equations by allowing
$f\rightarrow f+\delta f$ and ${\bf q}\rightarrow {\bf q}+\delta {\bf q}$,
then the second variation of the free energy functional is
\begin{equation}
\delta^2 {\cal F} = \int dx\,dy \left[\frac{1}{\kappa^2}
({\bf \nabla}\delta f)^2+
4 f (\delta f) {\bf q}\cdot\delta {\bf q}+f^2(\delta {\bf q})^2+
(3 f^2+{\bf q}^2-1)(\delta f)^2 + ({\bf\nabla}\times\delta {\bf q})^2\right]
\label{2d}
\end{equation}
(we neglect perturbations along the $z$-direction).
Expanding in Fourier modes with respect to $y$ \cite{kramer68},
\begin{equation}
\delta f(x,y) = \tilde{f}(x)\cos ky, \quad
\delta q_x(x,y) = \tilde{q}_x(x)\sin ky,\quad
\delta q_y (x,y) = \tilde{q}_y(x)\cos ky,
\end{equation}
substituting into Eq.~(\ref{2d}), recalling that ${\bf q} = (0,q(x),0)$,
and integrating over $y$, we obtain (up to a multiplicative constant)
\begin{equation}
\delta^2{\cal F} =
\int_0^{\infty}dx\left[\frac{1}{\kappa^2}\tilde{f}'^2+(3f^2+q^2
+\frac{1}{\kappa^2}k^2-1)
\tilde{f}^2+4fq\tilde{f}\tilde{q}_y+f^2(\tilde{q}_x^2+\tilde{q}_y^2)+
(\tilde{q}_y'-k\tilde{q}_x)^2\right].
\end{equation}
By integrating by parts and using the boundary conditions,
Eq.~(\ref{stab_boundary}), we can cast this functional into the form
\begin{equation}
\delta^2{\cal F} = \int_0^{\infty}dx\, (\tilde{f},\tilde{q}_y,\tilde{q}_x)
\hat{L}_2 \left(\begin{array}{c}\tilde{f}\\ \tilde{q}_y\\
\tilde{q}_x\end{array}
\right),
\end{equation}
where the self-adjoint linear operator $\hat{L}_2$ is given by
\begin{equation}
\hat{L}_2\left(\begin{array}{c}\tilde{f}\\ \tilde{q}_y\\ \tilde{q}_x
\end{array}\right) =
\left(\begin{array}{ccc}
-\frac{1}{\kappa^2}\frac{d^2}{dx^2}+q^2+3f^2+ k^2/\kappa^2-1 & 2fq & 0\\
2fq & -\frac{d^2}{dx^2} +f^2 & - k \frac{d}{dx}\\
0 & k \frac{d}{dx} & f^2 + k^2 \end{array}\right)
\left(\begin{array}{c}\tilde{f}\\ \tilde{q}_y \\ \tilde{q}_x
\end{array}\right).
\end{equation}
As in the previous section, we want to determine the eigenvalue spectrum of
this operator. We are primarily interested in the effects of long-wavelength
perturbations (i.e., $k\rightarrow 0$), so we rescale $k$ as $k=\kappa k'$.
Then the eigenvalue
equations in terms of the outer coordinate $x'=\kappa x$ are (dropping the
prime on $k$ from now on)
\begin{equation}
-\tilde{f}''+(3f^2+q^2-1+k^2)\tilde{f}+2fq\tilde{q} = E\tilde{f},
\label{2d1}
\end{equation}
\begin{equation}
- \kappa^2 \tilde{q}_y'' + f^2 \tilde{q}_y + 2 fq\tilde{f}
- \kappa^2 k \tilde{q}_x' = E\tilde{q}_y,
\label{2d2}
\end{equation}
\begin{equation}
\kappa^2 k \tilde{q}_y' + (f^2 + \kappa^2 k^2) \tilde{q}_x = E \tilde{q}_x.
\label{2d3}
\end{equation}
By using the last equation we may eliminate $\tilde{q}_x$ from Eq.~(\ref{2d2}),
which becomes
\begin{equation}
- \kappa^2 {d \over dx}\left[ {f^2 - E \over f^2 + \kappa^2 k^2 - E}
\tilde{q}_y'\right] + f^2\tilde{q}_y + 2fq\tilde{f} = E \tilde{q}_y.
\label{2d4}
\end{equation}
For $k=0$ Eqs.~(\ref{2d1}) and (\ref{2d4}) reduce to the one-dimensional
perturbation equations of the last section, Eqs.~(\ref{2d_perturb1})
and (\ref{2d_perturb2}); for $E=0$ they reduce to
the Euler-Lagrange equations derived by Kramer \cite{kramer68}.
The perturbation equations (\ref{2d1}) and (\ref{2d4}) may be solved
by the method of matched asymptotic expansions, just as in the
one-dimensional case. The derivation of the eigenvalue condition
is essentially identical, with the final result that
\begin{equation}
\frac{1}{P_2^{\mu}(A_0)}\frac{dP_2^{\mu}(A_0)}{dA_0} =
\frac{2}{E_0}\left[
\frac{E_0+4A_0^2}{2A_0}
-\frac{4A_0^3}{(A_0+\sqrt{A_0^2-E_0})\sqrt{A_0^2-E_0}}\right],
\end{equation}
where now $\mu = - \sqrt{2(2+E_0-k^2)}$. The eigenvalue $E_0(k)$ is plotted
in Fig. \ref{stable_plot2} for several different values of $A_0$.
For $A_0>1/\sqrt{2}$, $E_0(k)>0$ for all $k$, while for $A_0<1/\sqrt{2}$
there exists a band of long-wavelength perturbations for which
$E_0(k)<0$. In all cases the most unstable modes are at $k=0$, i.e.,
the one-dimensional perturbations are the least stable. This is
in contrast to the large-$\kappa$ limit, where the most unstable mode
occurs for $k\neq 0$ \cite{galaiko66,kramer68,chapman95}.
\section{Nonlocal effects as $\kappa \rightarrow 0$}
In the previous sections we have studied superheating in type-I
superconductors starting from the conventional GL
equations, which assume a {\it local} relationship between the current
and the vector potential. However, in very clean type-I superconductors
{\it nonlocal} effects are often important (in the {\it Pippard} limit;
see Ref.~\cite{tinkham}). We can model these effects by replacing the
second GL equation, Eq.~(\ref{GL5}), by a nonlocal equation
of the form
\begin{equation}
\kappa^2 q'' - \int_{0}^{\infty} K(x - x') f^2(x') q(x')\, dx' = 0,
\label{non1}
\end{equation}
where $K(x-x')$ is a kernel whose Fourier transform $K(k)$ behaves as
\begin{eqnarray}
K(k) = \left\{
\begin{array}{ll}
\lambda^2/\lambda_{L}^2\qquad {\rm (local\ limit)}; \\
a/|k| \qquad {\rm (extreme\ anomalous\ limit)}, \\
\end{array}
\right.
\label{limits}
\end{eqnarray}
with $\lambda_L$ the London penetration depth and $a$ a constant
\cite{tinkham}. For $\lambda\approx \lambda_L$ we recover the local limit
considered in the previous sections of this paper. It is still possible
to calculate the
superheating field in this nonlocal limit using the method of matched
asymptotic expansions. Indeed, the prescription is the same as for the
local case discussed above; we only need to solve a slightly more
complicated inner problem. In this section we will calculate the
leading order superheating field in the nonlocal limit, in order to
further illustrate the power and flexibility of our method.
{\it Outer solution.} The outer solution is the same as before; the
vector potential is zero to all orders, and the first two terms in the
expansion for the order parameter are given by Eqs.~(\ref{outer3}) and
(\ref{outer7}).
{\it Inner solution.} In the inner region we rescale the variables as
in Eq.~(\ref{inner1}). In terms of these variables Eq.~(\ref{non1})
becomes
\begin{equation}
Q'' - \int_{0}^{\infty} K(X - X') F^2(X') Q(X')\, dX' = 0.
\label{non2}
\end{equation}
We need to solve this equation, along with the first GL
equation, Eq.~(\ref{inner2}), perturbatively in $\kappa$. Expanding
$F$ and $Q$ as in Eqs.~(\ref{inner5})--(\ref{inner7}), we obtain
$F_0(X)=A_0$, as before, and
\begin{equation}
Q_0'' - A_0^2 \int_{0}^{\infty} K(X-X') Q_0 (X')\, dX' = 0.
\label{non3}
\end{equation}
This is an integral equation of the Wiener-Hopf type \cite{integral}.
To solve, we Fourier transform, introducing
\begin{equation}
Q_{+}(k) = \int_{0}^{\infty} Q_0(X) e^{ikX}\, dX.
\label{wh1}
\end{equation}
After Fourier transforming the integral equation, we perform a
Wiener-Hopf factorization \cite{integral}, with the result that
\begin{equation}
Q_{+}(k) = B_0 { i e^{i\varphi(k)} \over [ k^2 + A_0^2 K(k)]^{1/2}},
\label{wh2}
\end{equation}
where $B_0 = Q_0(0)$ is a constant, and
\begin{equation}
\varphi(k) = {k\over \pi} \int_{0}^{\infty} \ln\left[ { k^2 + A_0^2 K(x)
\over x^2 + A_0^2 K(k) }\right] {1\over x^2 - k^2} \, dx.
\label{wh3}
\end{equation}
The Fourier transform can be inverted once a particular form for $K(k)$ is
specified (although this is unnecessary for the calculation of the
superheating field; see below). The magnetic field at this order is
$H_0(0)=-A_0 B_0$ as before.
Proceeding to the next order, we have
\begin{equation}
F_1'' = A_0 Q_0^2.
\label{non4}
\end{equation}
By applying the boundary condition $F_1'(0)=0$, we find the general solution
\begin{equation}
F_1(X) = A_1 + A_0\left[ X\int_{0}^{X} Q_0^2(y)\, dy
- \int_{0}^{X} y Q_0^2(y)\, dy\right],
\label{non5}
\end{equation}
with $A_1 = F_1(0)$ another constant.
The equation for $Q_1(X)$ is a rather messy inhomogeneous Wiener-Hopf
integral equation. Fortunately, its solution is not needed for the
leading order calculation of the superheating field.
{\it Matching.} We now turn to the matching of the inner and outer solutions.
The two term inner expansion of the one term outer solution is
\begin{equation}
f_0(\kappa X) \sim \tanh\left( {x_0 \over \sqrt{2}}\right)
+ {\kappa \over \sqrt{2}} {\rm sech}^2\left( {x_0 \over \sqrt{2}}\right)X.
\label{non6}
\end{equation}
The one term outer expansion of the two term inner solution is
\begin{equation}
F_0(x'/\kappa) + \kappa F_1(x'/\kappa) \sim A_0 + A_0 \left( \int_{0}^{\infty}
Q_0^2(y) \, dy \right) x' .
\label{non7}
\end{equation}
By using the van Dyke matching principle we find
$A_0 = \tanh(x_0/\sqrt{2})$ as before, and
\begin{equation}
A_0 \int_0^{\infty} Q_0^2(y)\, dy = {1\over \sqrt{2}} (1-A_0^2).
\label{non8}
\end{equation}
We can use Parseval's identity to express the left hand side of
Eq.~(\ref{non8}) in terms of $|Q_{+}(k)|^2$, and then use
Eq.~(\ref{wh2}) to finally arrive at
\begin{equation}
{A_0 B_0^2\over 2\pi} \int_{-\infty}^{\infty} {1\over k^2 + A_0^2 K(k)}\, dk
= {1\over \sqrt{2}} (1-A_0^2).
\label{non9}
\end{equation}
To calculate the superheating field we use Eq.~(\ref{non9}) to express
$B_0$ as a function of $A_0$; we then substitute
this result into $H_0(0) = -A_0 B_0$ and maximize with respect to
$A_0$ in order to determine the lowest order superheating field.
In the local limit, $K(k) = \lambda^2/\lambda_L^2$,
and we obtain $A_0^* = 1/\sqrt{2}$ and
$H_0(0)= 2^{-3/4} (\lambda_L/\lambda)$, which is the same as our previous
result when $\lambda\approx \lambda_L$. In the extreme anomalous limit
$K(k) = a/|k|$, with $a=(3\pi/4) (\lambda^2 \xi /\lambda^2_L \xi_0)$
in the Pippard theory \cite{tinkham}, where $\xi_0$ is the zero temperature
coherence length. Performing the integral, we find
\begin{equation}
B_0 = - 3^{3/4} 2^{-3/4} a^{1/6} A_0^{-1/6} (1-A_0^2)^{1/2},
\label{non10}
\end{equation}
so that
\begin{equation}
H_0(0) = 3^{3/4} 2^{-3/4} a^{1/6} A_0^{5/6} (1-A_0^2)^{1/2}.
\label{non11}
\end{equation}
The maximum occurs at $A_0^* = \sqrt{5/11}$, so that
$H_0(0) = 0.721\, a^{1/6}$. Therefore, the superheating field is
\begin{equation}
H_{\rm sh} = 0.721\, a^{1/6} \kappa^{-1/2} + O(\kappa^{1/2})
\qquad {(\rm extreme\ anomalous\ limit)}.
\label{non12}
\end{equation}
The same result has been obtained by Smith {\it et al.} \cite {smith70}
using an approximation for the order parameter along with a variational
calculation (in the spirit of method used by the Orsay group \cite{orsay}).
The advantage of our method is that it can be systematically improved upon.
Although we have not checked the stability in the extreme anomalous limit,
the procedure should be entirely analogous to that of the previous section.
\section{Large-$\kappa$ results}
So far we have used the method of matched asymptotics to solve the
GL equations in the small-$\kappa$ limit. Chapman \cite{chapman95}
has recently used the same method to treat the one dimensional
GL equations in the high-$\kappa$ limit.
His final result for the superheating field is
\begin{equation}
H_{\rm sh} = \frac{1}{\sqrt{2}}+C\kappa^{-4/3} + O(\kappa^{-6/3})
\label{chapsh1}
\end{equation}
where the constant $C$ is determined from the solution of the second
Painlev\`e transcendent; a numerical evaluation yields $C=0.326$
\cite{dolgert}. The first term was originally derived by Ginzburg
\cite{ginzburg58}, and the second term with the unusual dependence upon
$\kappa$ is the new term. As seen in Fig~\ref{chapman} the asymptotic
and numerical results agree very well. It turns out, however, that the
calculated $H_{\rm sh}$ is not actually the
superheating field, since the one dimensional solution in the large-$\kappa$
limit is unstable with respect to two-dimensional perturbations
\cite{galaiko66,kramer68,chapman95}; these instabilities occur at
at the smaller field $H_{\rm sh}^{2D} = \sqrt{5}/3\sqrt{2}=0.527$.
This situation is quite different from that of the small- $\kappa$ limit
in which our stability calculation (Section IV) found the limit of
stability to be right at $H_{\rm sh}$.
\section{Discussion}
In this paper, the one-dimensional GL equations are solved analytically
and numerically for a semi-infinite superconducting sample in the
small-$\kappa$ limit in order to determine the maximum superheating field
$H_{\rm sh}$. We have used the method of matched asymptotic expansions
to construct for the first time a systematic perturbative solution of
the Ginzburg-Landau equations, the results of which agree quite closely with
our numerical solutions. The same method has been used to
determine the stability of these solutions with respect to both one-
and two-dimensional infinitesimal fluctuations; our analysis shows
that two dimensional fluctuations do not lead to any additional
destabilizing effects, in contrast to the situation in the large-$\kappa$
limit. With little modification this method can also be adapted to treat
nonlocal electrodynamic effects. Finally, our numerical results for
large-$\kappa$ compare well with Chapman's asymptotic analysis of this
regime. Taken collectively, our results demonstrate the effectiveness
of the method of matched asymptotic expansions for dealing with
boundary layer problems in the theory of superconductivity. We
hope that others will find useful applications of the methods
developed in this paper in treating inhomogeneous superconductors.
\acknowledgments
We would like to thank Dr. Chung-Yu Mou for helpful discussions,
Dr. S. J. Chapman for useful correspondence, and Dr. C. Bolley for
sending us her preprints and reprints on superheating
in superconductors. This work was supported by NSF Grant DMR 92-23586,
as well as by the Alfred P. Sloan Foundation (ATD).
|
\section*{Introduction}
\def\roughly#1{\raise.3ex\hbox{$#1$\kern-.75em\lower1ex\hbox{$\sim$}}}
Up to now, supersymmetry has been a theorist's dream and an
experimentalist's nightmare. On the one hand, theorists tend
to like supersymmetry because it provides a beautiful
mathematical structure which can be used to stabilize
the mass hierarchy against radiative corrections. On
the other hand, many experimentalists despise the subject
because supersymmetric predictions always seem to lie just
out of reach.
At present, direct searches for supersymmetry are just shots
in the dark because current accelerators do not have the power
to explore significant regions of the parameter
space. As we will see, this will soon change, but for now,
direct searches do not restrict the theory in any
important way.
Precision electroweak measurements also reveal very little
about supersymmetry. The technical reason for this is
that supersymmetry decouples from all standard-model electroweak
observables. For example, the supersymmetric and standard-model
values of $S$ and $T$ are related as follows \cite{Peskin}
\begin{eqnarray}
S_{\rm SUSY} &\ =\ & S_{\rm SM}\ +\ {\cal O}(M_W/M_S)^2 \nonumber \\
T_{\rm SUSY} &\ =\ & T_{\rm SM}\ +\ {\cal O}(M_W/M_S)^2 \nonumber\ ,
\end{eqnarray}
where $M_W$ is the mass of the $W$, and $M_S$ denotes the scale
of the supersymmetric spectrum. Theorists can simply raise
$M_S$ and bring supersymmetry into complete accord with
standard-model predictions.
Fortunately, the next generation of accelerators, including the
Fermilab Main Injector, LEP 200, and a possible higher-luminosity
Tevatron, will open a new era in supersymmetric particle searches.
These accelerators will -- for the first time -- begin to probe
significant regions of the supersymmetric parameter space. And
with the advent of the LHC, the search for supersymmetry will
finally cover most -- if not all -- of the parameter space that
is relevant for weak-scale supersymmetry.
In this talk we will examine the supersymmetric parameter space.
We will impose supersymmetric unification and show the preferred range
for the supersymmetric particle masses. We will also discuss two more
specialized points: the constraints on the supersymmetric spectrum
that follow from supersymmetric unification, as well as the one-loop
constraints on the mass of the lightest Higgs boson.
\section*{The Supersymmetric Spectrum}
Supersymmetry stabilizes the hierarchy $M_W/M_P \simeq 10^{-17}$
against quadratically-divergent radiative corrections. However, it
does so at a tremendous cost: a doubling of the particle spectrum.
In supersymmetric theories, all bosons are paired with fermions
and vice versa. For the case at hand, each of the standard-model
particles is accompanied by a supersymmetric partner with the same
SU(3) $\times$ SU(2) $\times$ U(1) quantum
numbers. If supersymmetry is not broken, these superparticles are
degenerate in mass with their original standard-model partners.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig1.eps,height=3.0in}}
\caption{The gauge couplings unify in the minimal supersymmetric
standard model.}
}
\label{figlabel}
\end{figure}
The fact that such particles have not been observed tells us that
supersymmetry must be broken -- and that the breaking must be soft.
In other words, the supersymmetry breaking must not reintroduce
destabilizing quadratic divergences. The various types of soft
supersymmetry breaking have been studied extensively \cite{Girardello}.
One finds over 50 new parameters:
\begin{itemize}
\item 3 gaugino masses: $M^{(a)}_{1/2}$
\item 23 scalar masses: $M^2_{0\ ij^*}$
\item 27 trilinear scalar couplings: $A_{0\ ijk}$
\item 1 bilinear scalar coupling: $B \mu$.
\end{itemize}
\noindent
(plus phases). Naturalness requires that each of the dimensionful
couplings be less than about a TeV, but that is all we know. The most
general softly-broken supersymmetric theory has an enormous parameter
space -- and theorists are prepared to use every bit of it to evade
experimental limits!
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig2.eps,height=3.0in}}
\caption{The soft supersymmetry-breaking masses can be arranged to
unify in the supersymmetric standard model. Here $M_0 = 300$ GeV
and $M_{1/2} = 100$ GeV. The solid lines denote squark masses and
the dotted lines sleptons. The dashed lines are gaugino masses,
while the dot-dashed line marks the mass of the Higgs.}
}
\label{figlabel}
\end{figure}
Motivated by the successful unification of the gauge couplings in the
supersymmetric standard model (Fig.~1), it is not unreasonable to assume
that the soft breakings unify as well. In this case the parameter space
reduces substantially, to include
\begin{itemize}
\item 1 universal gaugino mass: $M_{1/2}$
\item 1 universal scalar mass: $M^2_0$
\item 1 universal trilinear scalar coupling: $A_0$
\item 1 bilinear scalar coupling: $B \mu$.
\end{itemize}
\noindent
As usual in a unified theory, these parameters are fixed at the
unification scale $M_{GUT}$. Their values in the low-energy
theory are determined by the renormalization group, as shown
in Fig.~2.
At the unification scale, the unification assumption forbids
electroweak symmetry breaking because all scalar masses -- including
that of the Higgs -- have the common value $M^2_0$. However,
top-quark loops decrease the Higgs mass. Therefore, in the
renormalization group evolution, they drive down the mass of
the Higgs. If the top Yukawa coupling is sufficiently
large (corresponding to a top mass of about 150 -- 200 GeV),
the mass squared goes negative and triggers electroweak symmetry
breaking. It is remarkable that this radiative mechanism for
symmetry breaking \cite{Ibanez}, first proposed in 1983, is
based on a top-quark mass that is in complete agreement with
current experiments!
In what follows, we present expectations for the supersymmetric
spectrum based on this unification scenario. For simplicity, we
set $A_0 = 0$, and we trade $B$ for the value of $\tan\beta =
v_d/v_u$, evaluated at the scale $M_Z$. In our figures we take
$\alpha_s(M_Z) = 0.12$, $M_t = 175$ GeV, and the supersymmetric
Higgs mass parameter $\mu > 0$. We find the values of various
supersymmetric masses as a function of $M_0$ and $M_{1/2}$, for
two values of $\tan\beta$. (In the figures, all masses are
one-loop pole masses.)
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig3.eps,height=2.75in}}
\caption{The mass of the lightest supersymmetric particle,
$\chi^0_1$, for $\mu > 0$, $A_0 = 0$, $\alpha_s(M_Z) =
0.12$ and $M_t = 175$ GeV. The shaded region is forbidden
by experimental and theoretical constraints. Most of the
supersymmetric parameter space is still open.}
}
\label{figlabel}
\end{figure}
In Fig.~3 we show mass contours for the lightest superparticle,
$\chi^0_1$. The $\chi^0_1$ is neutral, and because of a global
symmetry called $R$-parity, is assumed to be stable. In the figure,
the shaded areas represent forbidden regions of parameter space, either
because of present experimental limits or because of theoretical
constraints such as the cosmological requirement that the lightest
(stable) superparticle be neutral, or the phenomenological constraint
that electroweak symmetry be broken, but not color.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig4.eps,height=2.75in}}
\caption{The mass of the up squark (solid line) and the
gluino (dashed line), for $\mu > 0$, $A_0 = 0$, $\alpha_s(M_Z)
= 0.12$ and $M_t = 175$ GeV. Our parameter space
corresponds to $M_{\tilde q}\ \roughly{<}\ 1$ TeV.}
}
\label{figlabel}
\end{figure}
In Fig.~4 we show contours for the (up) squark and the gluino
masses. (The masses of the up, down, charm and strange squarks
are almost degenerate.) From the plot we see that our parameter
space covers squark masses up to about 1 TeV. This is the range of
interest if supersymmetry is to stabilize the weak-scale hierarchy.
(The rule of thumb is that $M_{\tilde g} \sim 3 M_{1/2}$
and $M^2_{\tilde q} \sim M^2_0 + 4 M^2_{1/2}$.)
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig5.eps,height=2.75in}}
\caption{The mass of the lightest chargino, $\chi^\pm_1$,
(solid line) and lightest Higgs, $h$, (dashed line), for $\mu
> 0$, $A_0 = 0$, $\alpha_s(M_Z) = 0.12$ and $M_t = 175$ GeV.
The Higgs mass $M_h\ \roughly{<}\ 120$ GeV over our
parameter space.}
}
\label{figlabel}
\end{figure}
In Fig.~5 we plot contours for the masses of the lightest Higgs
scalar, $h$, and the lightest chargino, $\chi^\pm_1$. We see
that $M_{\chi^\pm} \sim M_{1/2}$, and that for our parameter
space, the maximum Higgs mass is about 120 GeV. (For completeness,
we note that the slepton masses are approximately $M_{\tilde \ell}
\sim M_0$.)
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig6.eps,height=2.75in}}
\caption{The mass of the charged Higgs, $H^\pm$,
(solid line) and lightest stop, $\tilde t$, (dashed line),
for $\mu > 0$, $A_0 = 0$, $\alpha_s(M_Z) = 0.12$ and $M_t =
175$ GeV. The decays $t \rightarrow
\tilde t \tilde \chi^0_1$ and $t \rightarrow H^+ b$ are
kinematically forbidden over most of the parameter space.}
}
\label{figlabel}
\end{figure}
Finally, in Fig.~6 we show contours for the
lightest stop squark, $\tilde t$, and charged Higgs,
$H^\pm$. From the figure we see that the decays $t \rightarrow
\tilde t \tilde \chi^0_1$ and $t \rightarrow H^+ b$ are kinematically
forbidden over most of the parameter space. (The stop can be
lighter for $A_0 \ne 0$, but a very light stop requires a fine
tuning of the parameters.)
These figures can be used to illustrate the supersymmetry reach
of a given accelerator. For example, LEP 200 has a mass reach
of about $\sqrt s - 100$ GeV for a supersymmetric Higgs particle,
and $\sqrt s /2$ for a chargino. Therefore Fig.~5 shows that
LEP 200 has an excellent chance of discovering the lightest
supersymmetric Higgs and a reasonable possibility of finding the
lightest chargino.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig7.eps,height=3.0in}}
\caption{The one-loop $W$-boson pole mass, for $\tan\beta = 2$,
$A_0 = 0$, $\alpha_s(M_Z) = 0.12$, $M_t = 175$ GeV, and two values
of $M_0$. The dashed lines correspond to the standard-model
$W$ masses, assuming the same Higgs masses as in the
supersymmetric cases. Note that the supersymmetry effects
decouple for large $M_{1/2}$.}
}
\label{figlabel}
\end{figure}
The Tevatron's discovery potential is more model-dependent, and
varies considerably with the Tevatron luminosity. For an integrated
luminosity between 200 pb${}^{-1}$ and 25 fb${}^{-1}$, the gluino
discovery reach is in the range of 300 -- 400 GeV. Likewise, the
chargino/neutralino reach varies between 150 -- 250 GeV in the
trilepton decay channel, $\chi^+_1 \chi^0_2 \rightarrow \ell^+
\ell^- \ell^{\prime +}$ plus missing energy \cite{Baer}. From
Figs.~4 and 5 we see that an upgraded Tevatron would begin
to cover a significant amount of the supersymmetric parameter
space.
\section*{Radiative Corrections}
In the introduction, we argued that precision electroweak measurements
cannot be used to restrict the allowed regions of $M_0$ and $M_{1/2}$
because supersymmetric effects decouple from electroweak
observables. This can be seen explicitly in Fig.~7, where we plot
the one-loop $W$-boson pole mass against $M_{1/2}$, for two values
of $M_0$. We see that the supersymmetric $W$ mass is indistinguishable
from the standard-model mass for $M_{1/2}\ \roughly{>}\ 300$ GeV. The
figure also shows that a measurement uncertainty for $M_W$ of 40 MeV
gives a sensitivity to the region $M_{1/2}\ \roughly{<}\ 150$ GeV for
$M_0 \simeq 100$ GeV. But this is just the region that will be
probed directly by LEP 200 and by the Tevatron with the Main
Injector!
In the context of supersymmetric unification, however, precision
measurements can play an important role in restricting the
supersymmetric parameter
space. To see this, we recall that in a unified model, $\alpha_s
(M_Z)$ can be predicted as a function of $\alpha_1(M_Z)$, $\alpha_2
(M_Z)$, and the weak- and unification-scale thresholds. For the
case at hand, we determine $\alpha_1(M_Z)$ and $\alpha_2
(M_Z)$ from the electroweak observables $\alpha_{EM}$, $G_F$ and
$M_Z$. We then calculate the weak-scale thresholds using the
minimal supersymmetric standard model, and take the
unification-scale thresholds to be those of a particular unified
model. In this way we can compute $\alpha_s(M_Z)$ as a function
of the parameters $M_0$ and $M_{1/2}$ in any unified model.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig8.eps,height=3.0in}}
\caption{Contours of $\alpha_s(M_Z)$ in the $M_0$, $M_{1/2}$
plane with $m_t=175$ GeV, $\tan\beta=2$ and $A_0=0$.}
}
\label{figlabel}
\end{figure}
In Fig.~8 we show the prediction for $\alpha_s(M_Z)$ in the
absence of unification-scale thresholds. We see that $\alpha_s(M_Z)$
is generally much larger than the experimental value of $\alpha_s(M_Z)
= 0.117 \pm 0.005$. Indeed, we see that $\alpha_s(M_Z) \le 0.127$
requires $M_0\ \roughly{>}\ 1$ TeV or $M_{1/2}\ \roughly{>}\ 500$
GeV \cite{Bagger}.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig9.eps,height=3.0in}}
\caption{The light shaded regions indicate the allowed
values of the gauge coupling threshold correction
$\epsilon_g$ in the minimal and missing-doublet SU(5)
models. The dark shaded region indicates the range of
$\epsilon_g$ necessary to obtain $\alpha_s(M_Z) =
0.117 \pm 0.01$.}
}
\label{figlabel}
\end{figure}
In Fig.~9 we show the effects of unification-scale thresholds.
We parametrize these thresholds by $\epsilon_g$, and illustrate
the allowed values in the minimal and missing-doublet SU(5) models,
together with the threshold required to give $\alpha_s(M_Z) = 0.117
\pm 0.01$. From the figure we see that minimal SU(5) requires
$M_0\ \roughly{>}\ 1$ TeV, which leads to squark masses of more than 1
TeV.
Weak-scale radiative corrections are also important
in determining the Higgs mass. As is well-known,
in supersymmetric models the tree-level Higgs mass is determined by
gauge couplings, and is bounded from above by $M_Z$. For heavy
top, this value receives significant radiative corrections, and for
$M_t \simeq 175$ GeV, the bound increases to about 120 GeV.
Experimentally, this is a very interesting number because it is
almost within reach of LEP 200. Theoretically, $M_h \simeq 120$
GeV is interesting as well, because it is approximately the {\it lower}
bound for the Higgs mass in the ordinary, nonsupersymmetric standard
model. In the standard model, top-quark loops give a negative
logarithmically-divergent $\phi^4$ contribution to the effective
potential, and this contribution can destabilize the vacuum. If
we require that the standard model hold all the way
to the Planck scale, so that the cutoff $\Lambda \simeq M_P$,
then the Higgs mass must be more than about 120 GeV \cite{Quiros}.
\begin{figure}
\vbox{
\centerline{\psfig{figure=Fig10.eps,height=2.8in}}
\caption{The maximum one-loop Higgs mass in the minimal
supersymmetric standard model, and the minimum Higgs mass
in the ordinary standard model, as a function of the
top-quark mass (After Ref.~(7)).}
}
\label{figlabel}
\end{figure}
This is illustrated in Fig.~10, where we plot the allowed
Higgs masses as a function of $M_t$. From the figure we see
that the maximum mass increases with $M_t$ in supersymmetric
standard model, as does the minimum mass in the ordinary
standard model. The curves have different slopes, and cross
at $M_t \simeq 175$ GeV. These curves indicate that
if the Higgs is discovered at LEP 200, either supersymmetry
is right, or that there must be other new physics below the
Planck scale!
\section*{Conclusions}
In this talk we have seen that supersymmetry searches are about to
enter an important new era. With LEP 200 and the Main Injector, they
will begin to probe large regions of the supersymmetric parameter
space. With luck, supersymmetry might even be found before the
advent of the LHC!
I would like to thank my collaborators Konstantin Matchev, Renjie
Zhang, and especially Damien Pierce for sharing their insights
on the supersymmetric standard model. This work was supported
by the U.S. National Science Foundation under grant NSF-PHY-9404057.
|
\section{Overview}
There are four questions I would like to address in reviewing the
long-term goals of charm decay studies:
{\em why, how, where and when?}
{\em Why?}
\noindent There exists a triple motivation for a detailed
analysis of charm decays: (i) it provides novel probes of QCD,
(ii) sharpens our tools for dealing with beauty
decays and (iii) represents one of the more promising avenues to
finding the hoped-for `unexpected', namely the intervention of New
Physics; this would most clearly be realized
through the observation of CP asymmetries
\cite{BURDMAN,HQ94}.
{\em How?}
\noindent A comprehensive program is required with its
cornerstones being the measurements of (a) lifetimes,
(b) semileptonic branching ratios,
(c) nonleptonic branching ratios and
(d) lepton spectra in exclusive as well as
inclusive semileptonic charm decays.
{\em Where?}
\noindent Three different experimental environments
have been considered
for high-statistics charm decay studies, namely
$(\alpha )$ $B$ factories, $(\beta )$ photo- and hadro-production
and $(\gamma )$ a $\tau$-charm factory.
{\em When?}
\noindent There are two benchmark dates in evaluating the
impact of a $\tau$-charm factory, namely
\noindent (A) 1998 - 2000: the next round of fixed target
experiments at FNAL will be finished by then; the data on
charm decays from CLEO will have reached a new dimension statistically
and systematically; BABAR and BELLE will commence data taking;
\noindent (B) $\sim$ 2005: the analysis of charm and beauty
decays at the $B$ factories should have reached a fully
mature level; new initiatives like CHARM2000 will have had a run;
the LHC with its dedicated program on beauty physics will hopefully
start up.
My talk will be organized as follows: in Sect.2 I will sketch the
relevant theoretical framework for charm decays; in Sect.3 I list the
database that is required -- or at least desired -- for a deeper
understanding to come about; in Sect.4 I review various
experimental stages before presenting an outlook in Sect.5.
\section{Theoretical Framework}
\subsection{The Landscape}
Four second-generation theoretical technologies have emerged
for treating the impact of the strong interactions on heavy-flavour
decays:
\noindent $\bullet$ {\em QCD Sum Rules} are employed for
describing exclusive semileptonic and non-leptonic
decays of hadrons with strangeness, charm and beauty.
\noindent $\bullet$ {\em Lattice Simulations of QCD} have
worked their way down in distance scale to deal with
exclusive semileptonic and non-leptonic
charm decays, albeit only in the quenched approximation;
quite significant jumps are required before beauty decays
can be tackled and one can go beyond
the quenched approximation.
\noindent $\bullet$ {\em Heavy Quark Effective
Theory} (HQET) on the other hand deals with exclusive
semileptonic decays of beauty hadrons; its applicability
to charm decays is dubious.
\noindent $\bullet$ {\em $1/m_Q$ Expansions} are distinct
from the other technologies in that they deal with
inclusive transitions, semileptonic as well as non-leptonic ones.
Similarly to HQET they apply best to beauty decays while
their numerical usefulness in charm decays is a priori uncertain.
There is one point of particular relevance for our discussion here.
Charm decays occupy a special place in the theoretical landscape:
they represent common ground for all the QCD methods listed
above, albeit sometimes only at the extreme range of their
applicability \footnote{The non-perturbative contributions can be
expressed through an expansion in powers of
$\mu _{had}/m_Q$; for
charm one has $\mu _{had}/m_Q \sim 0.4$. It is at least smaller than
unity, but not by a large factor.}. Comparing the predictions
from the various theoretical technologies, which emphasize
complementary aspects of QCD, with detailed and comprehensive
data has a two-fold benefit:
\noindent $\bullet$ It provides valuable insights into the inner
workings of QCD in general, and on the interplay between
perturbative and non-perturbative effects in particular.
\noindent $\bullet$ These findings can be extrapolated to
beauty decays and applied there with
{\em tested} confidence.
\subsection{Inclusive Transitions}
\subsubsection{Total Rates}
Total rates for a heavy-flavour hadron $H_Q$ to decay
into an inclusive final state $f$ can be expressed
through an expansion in powers of $1/m_Q$ \cite{SV}; through order
$1/m_Q^3$ one obtains the following master equation\cite{BUV}:
$$\Gamma (H_Q\rightarrow f)=\frac{G_F^2m_Q^5}{192\pi ^3}|KM|^2
\left[ c_3^f\matel{H_Q}{\bar QQ}{H_Q}+
c_5^f\frac{
\matel{H_Q}{\bar Qi\sigma \cdot G Q}{H_Q}}{m_Q^2}+ \right.
$$
$$\left. +\sum _i c_{6,i}^f\frac{\matel{H_Q}
{(\bar Q\Gamma _iq)(\bar q\Gamma _iQ)}{H_Q}}
{m_Q^3} + {\cal O}(1/m_Q^4)\right] \eqno(1)$$
where the dimensionless coefficients $c_i^f$ depend on the
parton level
characteristics of $f$ (such as the ratios of the final-state quark
masses
to $m_Q$); $KM$ denotes the appropriate combination of KM
parameters,
and $\sigma \cdot G = \sigma _{\mu \nu}G_{\mu \nu}$
with $G_{\mu \nu}$ being the gluonic field strength tensor. The last
term
in eq.(1)
implies also the summation over the four-fermion operators with
different light flavours $q$.
It is through the quantities
$\matel{H_Q}{O_i}{H_Q}$ that the dependence on the {\em decaying
hadron} $H_Q$, and
on
non-perturbative forces in general, enters; they reflect the
fact that the weak decay of the heavy quark $Q$ does not proceed
in empty space, but within a cloud of light degrees of
freedom -- (anti)quarks and gluons -- with which $Q$ and
its decay products can interact strongly. The practical usefulness
of the $1/m_Q$ expansion rests on our ability to determine the
size of these matrix elements.
Through order $1/m_Q^3$ there are three types of expectation
values that determine the non-perturbative corrections:
\noindent (i) The leading term can be expanded further:
$$\matel{H_Q}{\bar QQ}{H_Q} = 1-
\frac{\aver{(\vec p_Q)^2}_{H_Q}}{2m_Q^2}+
\frac{\aver{\mu _G^2}_{H_Q}}{2m_Q^2} + {\cal O}(1/m_Q^3)
\eqno(2)$$
where $\aver{(\vec p_Q)^2}_{H_Q}\equiv
\matel{H_Q}{\bar Q(i\vec D)^2Q}{H_Q}$ denotes the average
kinetic energy of the quark $Q$ moving inside the hadron
$H_Q$ and $\aver{\mu _G^2}_{H_Q}\equiv
\matel{H_Q}{\bar Q\frac{i}{2}\sigma \cdot G Q}{H_Q}$. The first term on the
right-hand-side of eq.(2) represents the naive
spectator ansatz.
\noindent (ii) The quantity $\aver{\mu _G^2}_{P_Q}$ is known
from the meson hyperfine splitting:
$$ \aver{\mu _G^2}_{P_Q}\simeq
\frac{3}{4} (M_{V_Q}^2-M_{P_Q}^2)\; , \eqno(3)$$
where $P_Q$ and $V_Q$ denote the pseudoscalar and vector mesons,
respectively. Therefore
$$\aver{\mu _G^2}_B\simeq 0.37\, GeV \; \; , \; \;
\aver{\mu _G^2}_D\simeq 0.41\, GeV \eqno(4a)$$
For $\Lambda _Q$ and $\Xi _Q$ baryons one has
instead
$$\aver{\mu _G^2}_{\Lambda _Q, \Xi _Q} \simeq 0 \eqno(4b)$$
since the light diquark system inside $\Lambda _Q$ and $\Xi _Q$
carries no spin, whereas
$$\aver{\mu _G^2}_{\Omega _Q} \neq 0 \eqno(4c)$$
For $\aver{(\vec p_Q)^2}_{H_Q}$
there exists an estimate from a QCD
sum rules analysis\cite{QCDSR}
$$\aver{(\vec p_b)^2}_B
\simeq 0.5\, \pm \, 0.1\, GeV \eqno(5a)$$
and one can expect one from
lattice QCD in the foreseeable future. We do have a
model-independant lower
bound on it\cite{OPTICAL}:
$$\aver{(\vec p_b)^2}_B \geq
0.37\, \pm \, 0.1\, GeV \; .\eqno(5b)$$
The
{\em difference} in the kinetic energy of Q inside
baryons and mesons can be related to the masses of charm and
beauty hadrons:
$$ \langle (\vec p_Q)^2\rangle _{\Lambda _Q}-
\langle (\vec p_Q)^2\rangle _{P_Q} \simeq
\frac{2m_bm_c}{m_b-m_c}\cdot
\{ [\langle M_B\rangle -M_{\Lambda _b}]-
[\langle M_D\rangle -M_{\Lambda _c}] \}
\eqno(5c)$$
where $\langle M_{B,D}\rangle$ denote the `spin averaged' meson
masses: $ \langle M_B\rangle \equiv \frac{1}{4}(M_B+3M_{B^*})$
and likewise for $\langle M_D\rangle$.
Using data
one finds: $\langle (\vec p_Q)^2\rangle _{\Lambda _Q}-
\langle (\vec p_Q)^2\rangle _{P_Q}= -(0.07 \pm 0.20)(GeV)^2$;
i.e., the present measurement of $M_{\Lambda _b}$
is not yet sufficiently accurate, but this will change
in the foreseeable future.
\noindent (iii) The expectation values for the four-quark operators
taken between {\em meson} states can be expressed in terms of a
single quantity, namely the decay constant:
$$\matel{H_Q(p)}{\bar Q_L \gamma _{\mu}q_L)
(\bar q_L \gamma _{\nu}Q_L)}{H_Q(p)}\simeq
\frac{1}{4} f^2_{H_Q}p_{\mu}p_{\nu} \eqno(6)$$
where factorization has been assumed. The theoretical
expectations for the decay constants are \cite{GUIDO}
$$f_D \simeq 200\, \pm 30\, MeV\; \; \; , \; \; \;
f_B \simeq 180 \, \pm 30\, MeV \eqno(7a)$$
$$f_{D_s}/f_D \simeq 1.15 - 1.2\; \; \; , \; \; \;
f_{B_s}/f_B \simeq 1.15 - 1.2 \eqno(7b)$$
The size of the expectation values taken between
{\em baryonic} states are quite uncertain at present. There
exists more than one relevant contraction, and for the time
being quark model estimates provide us with the only guidance!
I will return to this point when discussing predictions of
baryon lifetimes.
To illustrate the method I give the semileptonic and
non-leptonic widths for charm hadrons through order $1/m_c^2$:
$$\Gamma _{SL}(H_c) \simeq \Gamma _0
\matel{H_c}{\bar cc}{H_c}\cdot
\left( 1- \frac{2\aver {\mu _G^2}_{H_c}}{m_c^2} +
{\cal O}(1/m_c^3)\right) \eqno(8a)$$
$$\Gamma _{NL}(H_c) \simeq \Gamma _0 N_C
\matel{H_c}{\bar cc}{H_c}\cdot \left[ A_0
\left( 1- \frac{2\aver {\mu _G^2}_{H_c}}{m_c^2}\right) -
4A_2 \frac{2\aver {\mu _G^2}_{H_c}}{m_c^2} +
{\cal O}(1/m_c^3)\right] \eqno(8b)$$
$$\Gamma _0 = \frac{G_F^2m_c^5}{192\pi ^3} |V(cs)|^2\; ,
\eqno(8c)$$
where $A_{0,2}$ denote perturbative QCD corrections; I have
ignored here the small phase space correction due to
$m_s^2/m_c^2 \neq 0$. The following results can be read off
from eqs.(8):
\noindent $\bullet$ For $m_c \rightarrow \infty$ the parton model spectator
expression is recovered.
\noindent $\bullet$ $\Gamma _{SL}$ as well as $\Gamma _{NL}$ receive
non-perturbative corrections, as does $BR_{SL}(H_c)$; the latter
quantity is lowered since the last term in eq.(8b)
enhances $\Gamma _{NL}$ ($A_2 < 0$ !)\cite{BUV}.
\noindent $\bullet$ $\Gamma _{SL}$ is {\em not} universal for all
charm hadrons $H_c$ once ${\cal O}(1/m_c^2)$ contributions
are included. One actually finds
$$\Gamma _{SL}(D)\; / \; \Gamma _{SL}(\Lambda _c) \; / \;
\Gamma _{SL}(\Omega _c) \; \sim \; 1\; / \; 1.5 \; / \; 1.2 \; ,
\eqno(9)$$
where the difference in the $\Lambda _c$ and $\Omega _c$
widths to this order are due to eqs.(4b,c);
i.e., the ratio of semileptonic branching ratios does
{\em not} coincide
with the ratio of lifetimes when comparing mesons and
baryons.
\noindent $\bullet$ The widths for $D$ and $\Lambda _c$ decays
differ already in order $1/m_c^2$.
\noindent $\bullet$ The differentiation between the widths for
$D^0$, $D^+$ and $D_s$ mesons occurs at order $1/m_c^3$ as
expressed in eq.(1), but left out in eq.(8). \footnote{$SU(3)_{Fl}$
symmetry is very much observed in the expectation values
$\aver{\mu _G^2}_{D,D_s}$ and $\aver{(\vec p_c^2)}_{D,D_s}$.}
The underlying pattern can be expressed as follows:
$$\Gamma (\Lambda _Q) = \Gamma (P_Q^0) = \Gamma (P_Q^+)
+ {\cal O}(1/m_Q^2) \eqno(10a)$$
$$\Gamma (\Lambda _Q) > \Gamma (P_Q^0) \simeq \Gamma (P_Q^+)
+ {\cal O}(1/m_Q^3) \eqno(10b)$$
$$\Gamma (\Lambda _Q) > \Gamma (P_Q^0) > \Gamma (P_Q^+)
+ {\cal O}(1/m_Q^4) \eqno(10c)$$
i.e., one predicts $\tau (\Lambda _b) < \tau (B_d) < \tau (B^-)$
as well as $\tau (\Lambda _c) < \tau (D^0) < \tau (D^+)$.
Yet keeping in mind that the $1/m_Q$ expansion is at
best a semi-quantitative tool
for $m_Q = m_c$ ($\mu _{had}/m_c \sim 0.4$!) one cannot
expect to make precise predictions for the charm lifetime ratios.
On the other hand -- and that is a central point of my message --
measurements of the lifetime ratios for all weakly decaying
charm hadrons can be used with great profit to disentangle
the various contributions in charm decays. Such an anatomy
will then pave the way for more reliable predictions on beauty
decays. To say it differently: a comprehensive study of charm decays
can be harnessed as Nature's microscope onto the
numerically smaller effects in beauty decays.
In Tables \ref{TABLE1} and \ref{TABLE2}
I juxtapose the theoretical expectations and
predictions on charm and beauty lifetime ratios
\cite{MIRAGE,DS,STONE2,BARYONS1,REPORT} with present
data \cite{MALVEZZI,SHARMA}.
\begin{table}
\begin{tabular} {|l|l|l|}
\hline
Observable &QCD ($1/m_c$ expansion) &Data \\
\hline
\hline
$\tau (D^+)/\tau (D^0)$ & $\sim 2 \; \; $
[for $f_D \simeq$ 200 MeV]
&$2.547 \pm 0.043$ \\
&(mainly due to {\em destructive} interference) & \\
\hline
$\tau (D_s)/\tau (D^0)$ &$1\pm$ few$\times 0.01$
& $ 1.125\pm 0.042$ \\
\hline
$\tau (\Lambda _c)/\tau (D^0)$&$\sim 0.5\; ^*$ &
$0.51\pm 0.05$\\
\hline
$\tau (\Xi ^+ _c)/\tau (\Lambda _c)$&$\sim 1.3\; ^*$ &
$1.75\pm 0.36$\\
\hline
$\tau (\Xi ^+ _c)/\tau (\Xi ^0 _c)$&$\sim 2.8\; ^*$ &
$3.57\pm 0.91$\\
\hline
$\tau (\Xi ^+ _c)/\tau (\Omega _c)$&$\sim 4\; ^* $&
$3.9 \pm 1.7$\\
\hline
\end{tabular}
\centering
\caption{QCD Predictions for Charm Lifetimes}
\label{TABLE1}
\end{table}
\begin{table}
\begin{tabular} {|l|l|l|}
\hline
Observable &QCD ($1/m_b$ expansion) &Data \\
\hline
\hline
$\tau (B^-)/\tau (B_d)$ & $1+
0.05(f_B/200\, \,\mbox{MeV} )^2
[1\pm {\cal O}(10\%)]>1$ &$1.04 \pm 0.05$ \\
&(mainly due to {\em destructive} interference) & \\
\hline
$\bar \tau (B_s)/\tau (B_d)$ &$1\pm {\cal O}(0.01)$
& $ 0.98\pm 0.08$ \\
\hline
$\tau (\Lambda _b)/\tau (B_d)$&$\sim 0.9\; ^*$ & $0.76\pm 0.06$
\\
\hline
\end{tabular}
\centering
\caption{QCD Predictions for Beauty Lifetimes}
\label{TABLE2}
\end{table}
As mentioned before, at present one has to rely on quark models
to estimate the size of the relevant {\em baryonic} expectation values.
Thus there is a model dependance in the predictions on
{\em baryon} lifetimes -- in contrast to the case with meson
lifetimes. This is indicated in the tables by an asterisk.
The general agreement with the data is remarkable, in particular
for the charm system, where the expansion parameter
is not much smaller than unity. A few more detailed comments are in order:
\noindent $\bullet$ The $D^+$-$D^0$ lifetime difference is driven
mainly by a destructive interference \cite{PI} with `Weak
Annihilation' (WA) contributing not more than 10 - 20\%.
Within the accuracy of the expansion, the
data are reproduced.
\noindent $\bullet$ The $D_s$-$D^0$ lifetime ratio can be treated
with better theoretical accuracy, namely of order a few percent.
The observed near equality of $\tau (D_s)$ and $\tau (D^0)$
represents rather direct evidence for the reduced weight of WA in
charm meson decays\cite{DS}.
\noindent $\bullet$ Even the expectations on the charm baryon
lifetimes reproduce the data which is quite remarkable since there
are constructive as well as destructive contributions to baryon
lifetimes. It has to be noted though that the present measurements
suffer from large uncertainties.
\noindent $\bullet$ However the prediction on
$\tau (\Lambda _b)/\tau (B_d)$ appears to be in
serious (though not yet conclusive) disagreement with the data. The
details of what went into that prediction can be found in
ref.\cite{REPORT}; here I want to state only the following conclusion.
If
$\tau (B_d)$ indeed exceeds $\tau (\Lambda _b)$ by 25 - 30 \%,
then a `theoretical price' has to be paid. It strongly suggests that
the present agreement between theoretical
expectations and data on charm baryon lifetimes is largely
accidental and most likely would not survive in the face of more
precise measurements! To state it in a more constructive
manner: more precise measurements on charm baryon
lifetimes would then allow to isolate the source of the discrepancy
between prediction and observation.
As already said, on general grounds one does not predict the
semileptonic widths to be the same for all charm hadrons --
apart from $\Gamma _{SL}(D^0)\simeq \Gamma _{SL}(D^+)$,
which is protected by isospin invariance and Cabibbo suppression.
A priori there could be a sizeable difference in
$\Gamma _{SL}(D^0)$ vs. $\Gamma _{SL}(D_s)$ due to a
WA contribution to the latter observable. It is then a non-trivial
prediction that those two quantities largely coincide:
$$ 1\pm \sim \, {\rm few}\, \% = \frac{\tau (D_s)}{\tau (D^0)}
\simeq \frac{BR_{SL}(D_s)}{BR_{SL}(D^0)} \simeq
1\pm \sim 10\% \eqno(11)$$
On the other hand a sizeable difference is expected in
the semileptonic widths of baryons and mesons which is
expressed as follows:
$$BR_{SL}(\Lambda _c) > BR_{SL}(D^0)\cdot
\frac{\tau (\Lambda _c)}{\tau (D^0)} \simeq
0.5 \cdot BR_{SL}(D^0) \eqno(12)$$
\subsubsection{Lepton Spectra}
A detailed study of the lepton spectra in inclusive semileptonic
decays of $D^0$, $D^+$ and $D_s$ mesons is highly desirable.
One expects \cite{WA} sizeable differences between the
energy spectra in $D^0$ and $D_s$ and to a lesser degree also in
$D^+$ decays. For there is a WA process that
is Cabibbo allowed [forbidden] for $D_s$ [$D^+$] mesons where
the hadrons in the final state emerge from (double) gluon
emission of the initial anti-quark line. These differences will
show up mainly in the endpoint region. An analogous
complication is expected for semileptonic $B$ decays:
hadronization affects the spectra in
the endpoint region differently in
$B_d$ and $B^-$ transitions. This creates a systematic uncertainty
in the value extracted from inclusive decays that cannot
be evaluated reliably unless
\noindent $\bullet$ one separates $B_d$ and $B^-$ decays or
\noindent $\bullet$ measures the corresponding effects for $D$ mesons
and extrapolates to $B$ mesons through a $1/m_Q$ expansion.
\subsection{Exclusive Charm Decays}
\subsubsection{Leptonic and Semileptonic Channels}
Measuring $BR(D^+,D_s \rightarrow \mu \nu ,\, \tau \nu )$ with
{\em good} accuracy represents a high priority goal since it
allows to extract the decay constants $f_D$ and $f_{D_s}$.
There exists considerable intrinsic interest in the value of these
hadronic parameters; in addition -- and maybe more
urgently from a phenomenological perspective -- once
$f_D$ and $f_{D_s}$ have been well measured, one can
confidently extrapolate to the beauty sector and predict
$f_B$ and $f_{B_s}$.
As discussed in detail in El'Khadra's talk
\cite{AIDA} a host of theoretical
tools can be brought to bear on exclusive semileptonic
charm decays: QCD sum rules, Lattice QCD and HQET in addition
to quark models. Confronting their predictions with
comprehensive measurements of the relevant hadronic
form factors and their dependance on the momentum transfer
will provide us with valuable insights into the inner
workings of QCD; it also will be of great benefit in quantitatively
understanding exclusive semileptonic $B$ decays.
\subsubsection{Nonleptonic Modes}
Most important is a general caveat: the relationship
between the pattern in exclusive modes and in inclusive
transitions is quite tenuous. The former in contrast to the
latter are very sensitive to the dramatic behaviour of QCD
in the infrared regime; there exist relatively straightforward
examples \cite{MIRAGE} showing that while {\em individual}
exclusive rates get enhanced or decreased significantly
by the strong
interactions, these effects average out to a large degree
in the sum. No
theoretical tools have been developed yet that can master
these complexities and at present one can employ only models
of uncertain reliability. Nevertheless there is a valid motivation
behind such `phenomenological engineering', in particular
when applied to two-body modes. For it allows us -- once
sufficiently many branching ratios have been well measured --
to extract the size of the contributing isospin amplitudes and
their phase shifts \cite{BUCELLA}. This provides information
that is essential for
designing a strategy for CP studies and for interpreting
its outcome.
\subsubsection{Radiative Decays}
While in the Standard Model there is no (short-distance)
penguin operator generating $D\rightarrow \gamma + X,\,
\gamma K^*,\, \gamma \rho/\omega $ transitions, long distance
dynamics can. One should note that even the inclusive rate
receives contributions from a non-local (though
higher-dimensional) operator. Thus the radiative branching
ratios cannot be predicted in a reliable fashion. Yet measuring
$BR(D\rightarrow \gamma K^*,\, \gamma \rho /\omega)$ helps us
in two ways:
On the one hand one can again extrapolate
to the beauty system and obtain a reliable estimate for the
impact of long-distance dynamics on the corresponding modes
$B \rightarrow \gamma K^*,\, \gamma \rho /\omega $. This is important
for any attempt to extract $|V(ub)|$ from these radiative $B$
decays. On the other hand one has opened up a new
window onto New Physics; for it can manifest itself by
producing a significant deviation of the ratio
$BR(D\rightarrow \gamma \rho /\omega )/BR(D\rightarrow \gamma K^*)$
from $\tan ^2(\theta _c)$.
\section{Required/Desired Database}
The preceding discussion should have made it clear that
even without aiming at possible manifestations of New
Physics the need for further data on charm decays has
not diminished; the advances in our
theoretical understanding actually allow us to define
more precisely the kind of future measurement one needs
for further progress. I will briefly sketch them.
While there is no need from theory to measure the $D^+$, $D^0$ and
$\Lambda _c$ lifetimes more precisely, it would be quite
useful to determine $\tau (D_s)$ to within 1\%. Clearly the
greatest need for improvement exists for the $\Xi _c$ and
$\Omega _c$ lifetimes. A 5-10\% accuracy in $\tau (\Xi _c^0)$,
$\tau (\Xi _c^+)$ and $\tau (\Omega _c)$ would enable us to
extract the size of the relevant baryonic matrix elements.
The benchmark to aim for in $BR(D^+,D_s \rightarrow \mu \nu , \tau \nu )$
is a 10\% accuracy allowing to extract the decay constants to within
5\%.
For practical reasons one wants to know the {\em absolute}
branching ratios for charm hadrons to within a few percent.
Such information which is sorely missing for $D_s$, $\Lambda _c$
and $\Xi _c$ decays
\cite{ROUDEAU} is needed, among other things, for
proper charm counting in $B$ decays and for determining the
absolute values of $BR(B_s \rightarrow l \nu D_s^{(*)})$,
$BR(B \rightarrow D \bar D_s^{(*)})$ and $BR(B_s \rightarrow D_s^{(*)} \bar D_s^{(*)})$.
Likewise one wants to know the inclusive rates for
$\Lambda _c \rightarrow \Xi +X_s$, $\Xi _c \rightarrow \Xi +X$ etc. Our
ignorance here constitutes a major bottle neck in
many studies, like using $l\Xi$ correlations to
distinguish between $\Lambda _b$ and $\Xi _b$ decays.
It is also important to know the absolute branching ratios for the
inclusive transitions $D_s \rightarrow l+X$, $\Lambda _c \rightarrow l +X$,
$\Xi _c \rightarrow l +X$ to complement the information obtained from
the lifetimes. A detailed analysis of the lepton spectra in
$D,D_s \rightarrow l + X$ (and also in $\Lambda _c \rightarrow l+X$) would be
of great theoretical help when extracting $|V(ub)|$ from the
endpoint region in $B\rightarrow l +X$ decays.
The dependance of the hadronic form factors in
exclusive semileptonic charm decays on the momentum transfer
has to be measured directly and the analysis has to be extended
to include also channels like $D^+, D_s \rightarrow l \nu \eta / \eta ^{\prime}$.
The data base for the program of `theoretical engineering' in two-body modes
referred to above has to be completed by analysing final states
containing (multi)neutrals.
Enough statistics has to be accumulated to study doubly Cabibbo
suppressed decays in detail.
The radiative channels $D,D_s \rightarrow \gamma K^*, \gamma \rho / \omega$
have to be searched for in a dedicated manner.
Finally it would be quite useful to remeasure the reaction
$e^+ e^- \rightarrow D \bar{D} +X$ for $E_{c.m.} \sim 5-6$ GeV. Old
SPEAR data suggest an enhancement there; if true, it would point
to rather virulent final state interactions in that interval.
That region happpens to be the one that is also probed in
$B\rightarrow D \bar D_{(s)}$; such effects would have an obvious impact
on the CP phenomenology in those $B$ decays.
\section{Experimental Stage}
The measurements listed above fall into three categories:
\noindent (A) The lifetimes can be measured by fixed target
experiments (or at $B$ factories).
\noindent (B) Some of the measurements might not be
impossible in hadronic collisions or at a $B$ factory,
but certainly represent a stiff challenge
there \cite{WISS}. Determining $BR(D^+,D_s \rightarrow \mu \nu )$
with a 20\% accuracy
presumably belongs into that category, as do observing
non-leptonic decays with multi-neutrals in the final state and
extracting
{\em absolute} branching ratios for $D_s$ mesons and the charmed baryons.
A $\tau$-charm factory on the other hand offers the cleanest
measurements \cite{ROUDEAU}.
\noindent (C) There are finally measurements that
presumably will
remain in the sole domain of a $\tau$-charm factory. Among them
are: studies of the lepton spectra in $D/D_s/\Lambda _c \rightarrow l + X$;
the semileptonic branching ratios for the various charm hadrons
and the identification of genuine
radiative $D$ decays which requires the efficient rejection of
nonleptonic modes like $D\rightarrow K^* \pi ^0 \rightarrow K^*\gamma [\gamma ]$, i.e.
where one photon escapes detection; this can probably be achieved
only by making use of the excellent energy resolution available due
to beam-energy constraints at a threshold machine.
\section{Outlook}
Let me start out with some very general statements which I then
relate to the purpose of our meeting. A comprehensive program on
Heavy Flavour Physics is essential in any serious quest to unveil
Nature's Grand Design. Detailed studies of charm decays have to form
an integral part of such a program. In an `ideal' or at least
`optimal' world a $\tau$-charm factory plays a central role in such
studies. This factory is justified by its unique capabilities to advance
our understanding of QCD,
and it would run with good
luminosity in the energy range 3 GeV $\leq \sqrt{s} \leq$ 6 GeV;
its discovery potential for New Physics
would serve as the `icing on the cake'.
The `ideal' world is defined as one where a $\tau$-charm factory would
be running by now; in an `optimal' world it would start delivering data
by the end of the millenium, like the asymmetric $B$ factories.
Not surprisingly, our world is not ideal and probably not optimal.
There is still an excellent motivation for a first-class
$\tau$-charm factory starting up later, yet the emphasize will shift
somewhat. High precision studies of $\tau$ and charmonium physics
would represent the superb primary justification. On the other
hand, the battle lines for open charm physics might be re-drawn.
I expect that our experimental colleagues will devise some ingenious
new methods for obtaining at least decent measurements of
absolute branching ratios for charmed baryons etc. For cost and
time (also running time) reasons it might make more sense then to
limit the effective energy range of the machine to
3 GeV $\leq \sqrt{s} \leq $ 4.4 GeV and
concentrate on the core part of open charm physics, namely the
weak decays of $D$ and $D_s$ mesons with a two-fold
purpose: to
perform measurements of absolute branching ratios, semileptonic
decays and their lepton spectra that cannot be
made in other experimental environments
-- and to follow up on tantalizing hints
for the intervention of New Physics that might have surfaced in the
meantime!
{\bf Acknowledgements:} I am grateful to T.D. Lee for sharing
his insights and his enthusiasm with us. It was a
stimulating meeting nicely organized by J. Repond. This work was
supported by the National Science Foundation under
grant number PHY 92-13313. I also thank the Institute for
Nuclear Theory at the University of Washington for its
hospitality and the Department of Energy for partial
support during the write-up of this manuscript.
|
\section{ Introduction}
Let us call the
$n$-dimensional oscillator algebra ${\bf osc}(n)$
Lie algebra with
$(2n+1)$ basic generators $T^{\alpha},T^{\overline\alpha}$ and $T$
which obey the following commutation relations
$$
[T^{\alpha},T^{\overline\beta}]=[T^\alpha, T^\beta]=
[T^{\overline\alpha},T^{\overline\beta}]=0,
$$
\begin{equation} \label{te1}
[T^{\alpha},T\hskip 1pt ]=-i\hskip 1pt T^{\alpha}, \qquad
[T^{\overline\alpha},T\hskip 1pt ]=i\hskip 1pt T^{\overline\alpha},
\qquad \alpha=1,...,n.
\end{equation}
Let ${\cal A}_m$ be the manifold of Lie algebra structures in
an $m$-dimensional vector space $V_m$. The curve $ \ell (t) $
in ${\cal A}_m$ passing through the algebra
$\ell =\ell (0)$ is called the deformation of the Lie algebra
$\ell$.
It is well known that deformation of one-dimensional
oscillator algebra play an important role in various
considerations. They are
closely related with many interesting physical and
mathematical objects, for example, anti-de Sitter quantum mechanics
[{\bf 4,6}], symplectic geometry of one-dimensional complex disk
[{\bf 5,7,14}] and quantization of spinning particle [{\bf 12}].
This paper is devoted to quantization of the mechanical systems
for which the Hamiltonian vector fields of observables form the
deformation of $n$-dimensional oscilator algebra. Because of this fact
these systems can be considered as "deformations" of
the harmonic oscillator. The paper consists of
three sections. In the first section we give short review of the
geometric
quantization procedure. In the second section is
constructed the set of above-mentioned
mechanical systems which are realized at the
classical level in the form of K\"ahler symplectic manifolds
of constant holomorphic curvature [{\bf 9}]. Such mechanical
systems
are quantized later with the help of the geometric quantization
approach [{\bf 8,12}]. As it is known the K\"ahler manifold of
constant holomorphic curvature is $H$-projectively flat,
i.e. it admits $H$-projective mappings on the flat space. In
the third section of the paper we discuss the quantization of
more general K\"ahler manifolds (not necessarily of constant
holomorphic curvature) admitting $H$-projective holomorphic
mappings.
\\
\section{ Preliminaries}
To start with, we recall some relevant
facts about geometric
quantization procedure [{\bf 8,12}]. Let $(M,\omega)$ be a symplectic
manifold. According to Dirac {\it quantization} is the linear
map
$Q:f\rightarrow \hat f$ of Poisson (sub)algebra $C^ {\infty} (M)$ into
the set of operators in some (pre)Hilbert space
${\cal H}$ possessing the properties:
\begin{enumerate}
\item $\hat 1=1$;
\item $\widehat {\{ f,g\}}_h = \frac{i}{h}(\hat f \hat g - \hat g \hat
f)$;
\item $\hat {\overline f} = (\hat f)^{*}$;
\item for some complete set of functions
$f_1 ,..., f_n$ the operators ${\hat f}_1 ,...,{\hat f}_n$ also
form the complete set,
\end{enumerate}
where bar is the complex conjugation, star
denotes the conjugation of the operator and $h= 2\pi\hbar$
is the Planck constant.
The linear map ${\cal P}:f \rightarrow \check f$ possessing the
first
three properties is called {\it prequantization}. For the case
$M=T^{\ast}M$, $\omega = d \alpha $ prequantization was constructed
by Koopman, Van Hove and Segal. It has the form
\begin{equation} \label{te2}
{\cal P}f=\check f = f-i \hbar V(f) - \alpha (V(f)),
\end{equation}
where vector field
$V(f)$ is the Hamiltonian vector field of the function $f \in
C^{\infty} (M)$ defined by the condition
$$
V(f)\hskip 1pt \rfloor \hskip 1pt \omega = -df,
$$
where $\rfloor$ is the internal product. In local coordinates
$x^i$, $i=1,...,2n$ from here we have
\begin{equation} \label{te3}
V(f) = \omega ^{ij}\partial_j f\partial_i .
\end{equation}
Let ${\cal L}$ be the Hermitian line bundle with connection $D$
and $D$-invariant Hermitian structure $<,>$. Recall that
$D$-invariance means that for each pair of sections $\lambda$ and
$\mu$ of ${\cal L}$ and each real vector field $X$ on $M$ holds
\begin{equation} \label{te4}
X<\lambda ,\mu >=<D_X \lambda ,\mu>+<\lambda,D_X \mu>.
\end{equation}
Let $(x,U)$ is local coordinate system on $M$. If $\mu_0$ is
nonvanishing section of ${\cal L}$ over $U$, then we can identify the
space $\Gamma ({\cal L}, U)$ of section with $C^{\infty}(M)$ by
the
formula $C^{\infty}(U)\owns \varphi \leftrightarrow \varphi \mu_0 \in
\Gamma
({\cal L}, U)$. The operator $D_X$ in this case takes the next
form
\begin{equation} \label{te5}
D_X \varphi = X\varphi -i\hbar^{-1}\alpha(X)\varphi,
\end{equation}
where 1-form $\alpha$ is given by the relation
\begin{equation} \label{te6}
D_X \mu_0 = -i\hbar^{-1}\alpha(X)\mu_0 .
\end{equation}
By comparing (\ref{te2}) and (\ref{te6}) we have the prequantization
formula of
Souriau-Kostant
\begin{equation} \label{te7}
\check f = f-i\hbar D_{V(f)}.
\end{equation}
The curvature form $\Omega$ of $D$ is defined by the identity
$$
\Omega(X,Y) = \frac{1}{2\pi i}([D_X,D_Y] -
D_{[X,Y]}),
$$
and locally we have
\begin{equation} \label{te8}
\Omega = h^{-1} d\alpha.
\end{equation}
{\bf Theorem 1} [{\bf 8}]. {\it The Souriau-Kostant for\-mu\-la}
(6) {\it defines
the pre\-qu\-an\-ti\-za\-ti\-on
if and only if the curvature form $\Omega$
coincides with} $h^{-1}\omega$.
\\
The construction of Hilbert space ${\cal H}$ in
geometric quantization procedure
essentially involves the choice of the {\it polarization} that is
the involutive Lagrange distribution $F$
in $TM\otimes_{\bf R}{\bf C}$.
The polarization $F$ is called {\it K\"ahler} if
$F\cap\overline F=\emptyset$, and the Hermitian form $b(X,Y)=
i\hskip 1pt \omega (X,\overline Y)$ is positively defined for $X\in F$.
If the polarization $F$ on $TM\otimes {\bf C}$ is chosen then
the Hilbert space ${\cal H}$ consists of the sections
$\lambda$ of ${\cal L}$ which are covariantly constant along $F$
\begin{equation} \label{te9}
D_X \lambda=0,\quad\lambda\in{\cal L},\quad X \in F.
\end{equation}
It is said that the function $f \in C^{\infty}(M)$ preserves the
polarization $F$
if the flow $V(f)$ of $f$ obeys the condition
\begin{equation} \label{te10}
[V(f),X^{\alpha}]=a[f]^{\alpha}_{\beta}X^{\beta},
\end{equation}
where $a[f]^{\alpha}_{\beta}$ are some smooth functions on $M$
and vector
fields $X^{\alpha}$, $\alpha = 1,...,n$ span $F$.
Let us consider the particular case when the polarization $F$ is
spanned by the complex Hamiltonian vector fields. In this case the
functions preserving
polarization can be quantized with the help of the next
formula [{\bf 12}]:
\begin{equation} \label{te11}
Qf=-i\hbar D_{V(f)}+f- \frac{i\hbar}{2}a [f],
\end{equation}
where $a[f]=\Sigma^{n}_{\alpha=1}a[f]^{\alpha}_{\alpha}$.
In general situation $f$ it does not preserve the po\-la\-ri\-za\-ti\-
on
and for qu\-an\-ti\-zing $f$ one have to use the
{\it Blat\-tner-Kos\-tant-Stern\-berg (BKS) kernel} which connects
representations for different po\-la\-ri\-za\-ti\-ons
[{\bf 8,12}]. It is easy
to see that ever in the case of an $n$-dimensional oscillator
the flow of the Ha\-mil\-tonian
\begin{equation} \label{te12}
H=\frac{1}{2}\sum\limits^{n}_{\alpha =1}((p^{\alpha})^{2}
+ (q^{\alpha})^{2})
\end{equation}
does not preserve the polarization spanned by the Hamiltonian
vector fields of both
position $q^{\alpha}$ and momentum $p^{\alpha}$ variables.
Therefore we must use BKS kernel to quantize $H$. However, when
we introduce the complex coordinates
$$
z^{\alpha} = \frac{1}{\sqrt{2}}(p^{\alpha} + iq^{\alpha}),
\qquad
z^{\overline\alpha} = \frac{1}{\sqrt{2}}(p^{\alpha} - iq^{\alpha})
$$
and K\"ahler polarization spanned by the vector fields
$V(z^{\alpha})$ we can see that $H=\Sigma z^{\alpha}
z^{\overline\alpha}$ preserves the polarization
and one can use (\ref{te11}) to quantize $H$. As the result they
obtain
the differential operator $\hat H$
\begin{equation} \label{te14}
\hat H= Q_{\cal FB}\hskip 2pt H =
\hbar (z^{\alpha}\frac{\partial}{\partial z^\alpha}+
\frac{n}{2})
\end{equation}
on the space ${\cal O}(U)$ of holomorphic functions on
$U\in {\bf C}^{n}$ which we denote $Q_{\cal FB}$ because the
corresponding rep\-re\-sen\-ta\-tion is called Fock-Barg\-
mann
rep\-re\-sen\-ta\-tion.
In the considered case the rep\-re\-sen\-ta\-tion space ${\cal H}$
consists of the sections of ${\cal L}$ which have the form
$\psi (z) \mu_{0}$, where $\mu_{0}$ is a nonvanishing
section of ${\cal L}$ and $\psi (z)\in {\cal O}(U)$.
\\
\section{ Generalized Fock-Bargmann representation}
Now we consider the mechanical
system whose quantization is connected with generalization of
Fock-Bargmann representation. Let $(M,\omega)$ be
$2n$-dimensional K\"ahler manifold with fundamental
form $- \omega$ and positively definite K\"ahler metric $g$
to be given in local complex coordinates $(z^{\alpha},
z^{\overline\alpha})$ by the formula
\begin{equation} \label{te15}
-\omega= -\omega^{\alpha\overline\beta}
dz^\alpha \land dz^{\overline\beta}= -i\partial_{\alpha\overline\beta}\Phi\hskip
1pt
dz^\alpha \land dz^{\overline\beta},
\end{equation}
\begin{equation} \label{te16}
g=g_{\alpha\overline\beta}dz^\alpha dz^{\overline\beta}=
\partial_{\alpha\overline{\beta}}\Phi \hskip 1pt dz^{\alpha}
dz^{\overline{\beta}},
\end{equation}
where $\Phi$ is the K\"ahler potential. As the 2-form
$\omega$ is closed and nondegenerate, it defines the symplectic
structure on $M$ and we can consider $(M,\omega)$ as
symplectic manifold and the phase space of some mechanical
system. The classical observables [{\bf 8}] of such system form Lie
algebra $C^\infty(M)$ with respect to Poisson brackets.
Let as define the K\"ahler polarization $F$ on $TM \otimes
{\bf C}$ in the form
\begin{equation} \label{te17}
F=\{ X \in TM\otimes {\bf C} \hskip 2pt\vert
X=\xi_{\alpha}V(z^{\alpha}),
\xi_{\alpha}\in C^\infty (M) \},
\end{equation}
where
$V(z^{\alpha})=\omega^{\overline\sigma\alpha}\partial_{\overline\sigma}$
is given by (\ref{te3}). By (\ref{te10}) the function $f$ preserves
the
polarization $F$ if and only if
\begin{equation} \label{te18}
[V(f),V(z^{\alpha})]=a[f]^{\alpha}_{\mu}V(z^{\mu})
\end{equation}
where according to (\ref{te3}) and (\ref{te15})
$V(f)=\omega^{\mu\overline\nu}
(\partial_{\overline\nu}f\partial_{\mu} - \partial_{\mu}f\partial_
{\overline\nu})$, whence
$$
- \omega^{\mu\overline\nu}\partial_{\overline\nu} f
\partial_{\mu}\omega^{\alpha\overline\sigma}\partial_{\overline\sigma}
+ \omega^{\mu\overline\nu}\partial_{\mu}f
\partial_{\overline\nu} \omega^{\alpha\overline\sigma}\partial_{\overline
\sigma}
+ \omega^{\alpha\overline\sigma} \partial_{\overline\sigma}
\omega^{\mu\overline\nu} \partial_{\overline\nu} f
\partial_{\mu}
+
$$
$$
\omega^ {\alpha\overline\sigma} \partial_{\overline\nu}\partial_{\overline\sigma}
f \omega^{\mu\overline\nu} \partial_{\mu}
- \omega^{\alpha\overline\sigma}\partial_{\overline\sigma}
\omega^{\mu\overline\nu} \partial_\mu f \partial_{\overline\nu}
- \omega^{\alpha\overline\sigma}\partial_{\overline\sigma}\partial_\mu f
\omega^{\mu\overline\nu} \partial_{\overline\nu} =
- a[f]^\alpha_\mu \omega^{\mu \overline\sigma}\partial_{\overline\sigma}.
$$
Equating the corresponding components of vector fields in left and
right parts of the last relation we find
\begin{equation} \label{te20}
\partial_{\overline\sigma}\omega^{\mu\overline\nu}\partial_{\overline\nu} f +
\omega^{\mu\overline\nu}\partial_{\overline\nu}\partial_{\overline\sigma}f
\partial_\nu=0.
\end{equation}
Differentiating the equation
$\omega^{\nu\overline\mu}\omega_{\overline\mu\rho} =\delta^\nu_\rho$
we obtain the identity
$\omega^{\nu\overline\mu}\partial_{\overline\gamma}\omega_{\overline\mu\rho} =
-\partial_{\overline\gamma}\omega^{\nu\overline\mu}\omega_{\overline\mu\rho}$.
After this (\ref{te20}) takes the form
\begin{equation} \label{te21}
\nabla_X \nabla_Y f = 0, \qquad X, Y \in F
\end{equation}
where $\nabla$ denotes the covariant derivation with respect
to K\"ahler metric $g$.
By Theorem 1 we find from (\ref{te8})
$\omega=d\alpha$ and from (\ref{te15})
\begin{equation} \label{te22}
\alpha=-i\hskip 1pt \partial_\alpha\Phi \hskip 1pt dz^\alpha
\end{equation}
modulo to the exact 1-form $d\beta$.
If we choose both $\mu$ and $\lambda$ in (\ref{te4})
equal to nonvanishing
section $\mu_0$, then using (\ref{te6}) we obtain
$$
X<\mu_0,\mu_0>=
i\hbar^{-1}(\alpha(X)-\overline{\alpha(X)})<\mu_0,\mu_0>.
$$
Evaluating this formula on the vector fields
$\partial_\alpha$, $\partial_{\overline\alpha}$, $\alpha=1,...,n$
we find with the help of (\ref{te22})
\begin{equation} \label{te25}
<\mu_0,\mu_0>= exp(- \hbar^{-1}\Phi)
\end{equation}
up to constant multiplier which we omitted.
Now we determine the sections of Hermitian line bundle ${\cal L}$
which
form the representation space ${\cal H}$. Being covariantly
constant with respect to $D_{X\in F}$ these sections must obey the
equation (\ref{te9}):
$$
D_{V(z^\alpha)} \mu=0, \quad \mu\in\Gamma({\cal L}).
$$
{}From here we find with the help of (\ref{te5}), (\ref{te6}),
(\ref{te17}) and (\ref{te22})
that $\mu =\psi (z)\mu_0$, where $\psi (z)$ is holomorphic
function on $U \subset {\bf C}^n$.
If $M$ is contractible then using the Hermitian structure
$<,>$ in ${\cal L}$ we can define the
scalar product in ${\cal H}$ by the formula [{\bf 8,12}]
\begin{equation} \label{te26}
(\mu_1,\mu_2) = \int \psi_1 (z) \overline{ \psi_2 (z)} <\mu_0,\mu_0>
\omega^n,
\end{equation}
where $\mu_1=\psi_1(z)\mu_0$, $\mu_2=\psi_2(z)\mu_0$
and $\omega^n$ is n-th
external degree of $\omega$. Using (\ref{te25}) we find
\begin{equation} \label{te27}
(\mu_1,\mu_2) = \int \psi_1 (z)\overline { \psi_2 (z)}exp(-\hbar^{-1}\Phi)
\hskip 2pt \omega^n.
\end{equation}
In this case the representation Hilbert space associated with
polarization $F$ given by (\ref{te17}) can be identified with
Fock space $L_2^{hol}(U,dm)$ of holomorphic functions
on $U\subset{\bf C}^n$
quadratically integrable with the measure $dm=$ exp $(-
\hbar^{-1}\Phi)\hskip 1pt \omega^n$.
Let us consider the K\"ahler space ${\cal K}_{2n}$ of
constant
holomorphic curvature $k$ (see for example [{\bf 9}]).
As it is known the space ${\cal K}_{2n}$ is
isometric to the projective space ${\bf CP}^n$ for $k>0$, to the disk
$D^R_n= \{z\in{\bf C}^n\vert z\overline z<R\}$ for $k<0$ and to ${\bf C}^n$
for $k=0$. The metric of the space ${\cal K}_{2n}$ in the local
complex coordinates is
\begin{equation} \label{te28}
ds^2= 2g_{\alpha\overline\beta}dz^\alpha dz^{\overline\beta},
\qquad g_{\alpha\overline\beta} =\partial_{\alpha\overline\beta}\Phi=
(A\delta_{\alpha\beta} -\frac{k}{4}z^{\overline\alpha}z^\beta)A^{-2},
\end{equation}
$$
\Phi=\frac{4}{k}ln\hskip 1pt A,\quad A=1+\frac{k}{4}
\Sigma z^\nu z^{\overline\nu}.
$$
The curve $x(t)$ on K\"ahler manifold $M$ is called $H$-{\it planar}
(or {\it holomorphical planar}) [{\bf 11}] if
it obeys the following equation
$$
\nabla_{\chi} \chi = a(t)\chi + b(t) J(\chi), \qquad
\chi\equiv \dot x
$$
where $a(t)$, $b(t)$ are some real-valued functions and
$J$ is complex structure operator in $TM$.
Let $M$ and $M'$ be two K\"ahler manifolds.
The mapping $f:M\to M'$ is called $H$-projective (see for
example [{\bf 13}]) if it
transforms $H$-planar
curves of $M$ into $H$-planar curves of $M'$.
The contravariant components and nonvanishing Christoffel
symbols of the metric (\ref{te28}) in local complex coordinates
are given by the formula
$$
g^{\alpha\overline\beta}= (\delta^{\alpha\beta}
+\frac{k}{4}z^\alpha z^{\overline\beta})A^2 \equiv
-i\omega^{\alpha\overline\beta},\qquad
\Gamma^\alpha_{\beta\gamma}=
-\frac{k}{4}A^{-1}(z^{\overline\alpha}\delta^\beta_\gamma
+z^{\overline\beta}\delta^\alpha_\gamma)=
\overline{\Gamma^{\overline\alpha}_{\overline\beta\overline\gamma}}.
$$
{}From here it is follows that considered metrics are $H$-projectively
flat. Now we find from (\ref{te21})
\begin{equation} \label{te29}
\partial_{\mu\overline\nu} f +2A^{-1}\partial_{(\overline\mu}A
\partial_{\overline\nu)}f=0.
\end{equation}
After substitution $f= W A^{-1}$ in this equation it takes the form
$$
\partial_{\overline\mu\overline\nu} W=0
$$
whence
$$
W=u_{\overline\alpha}(z)z^{\overline\alpha} + v(z),
$$
where $u_{\overline\alpha}$ and $v$ are arbitrary holomorphic functions.
In [7] the system of observables
\begin{equation} \label{te32}
\tilde H =\frac{1+z\overline z} {1-z\overline z},\quad N=\frac{z}{1-z\overline z},
\qquad \overline N=\frac{\overline z}{1-z\overline z}
\end{equation}
was considered when quantizing 1-dimensional harmonic oscilator.
The Ha\-mil\-to\-ni\-an vec\-tor fields $V(\tilde H)$, $V(N)$
and $V(\overline N)$ form the basis of holomorphic isometries Lie algebra
in the space ${\cal K}_2$ of holomorphic curvature $k=-4$
(see for example [{\bf 14}]). We use the next system of observables
\begin{equation} \label{te33}
H =\frac{\Sigma z^\nu z^{\overline\nu}}{A},\quad (u_{\overline\alpha} =z^\alpha,
\quad v=0),
\end{equation}
\begin{equation} \label{te34}
N^\beta =\frac{ z^\beta}{A},\quad (u_{\overline\alpha} =0,
\quad v=z^\beta),
\end{equation}
\begin{equation} \label{te35}
N^{\overline\beta} =\frac{z^{\overline\beta}}{A},\quad(u_{\overline\alpha}=
\delta_\alpha^\beta,\quad v=0).
\end{equation}
One can easily check that $H$, $N^\alpha$ and
$N^{\overline\alpha}$ are the
solutions of equation (\ref{te29}).
The Hamiltonian vector fields of this functions define
infinitesimal isometries in ${\cal K}_{2n}$ and $H$-projective
transformations in the flat K\"ahler manifold ${\bf C}^n$. Note
that
these isometries do not form Lie algebra. In 1-dimensional case
$N^1$, $N^{\overline 1}$ coincide with $N,\overline N$ from (24) and $H$ can
be obtained from $\tilde H$ by the linear substitution. The
using of
$H$ is more preferable from the point of view of the limit transition
to the flat space ($k=0$). In the limit $k\to 0$
we obtain $H=\Sigma\hskip
1pt z^\nu z^{\overline\nu}$, i.e. the Hamiltonian of harmonic oscillator
(\ref{te12})
written in complex coordinates.
The Hamiltonian vector fields of the functions $H,N^\alpha$ and
$N^{\overline\alpha}$ have the form
$$
T \equiv V(H)= \omega^{\mu\overline\nu}(\partial_{\overline\nu}H \partial_\mu
-\partial_\nu H \partial_{\overline\mu})= i\hskip 1pt(z^\nu\partial_\nu
-z^{\overline\nu}\partial_{\overline\nu}),
$$
\begin{equation} \label{te36}
T^\alpha \equiv V(N^\alpha) =-i\hskip 1pt (\frac{k}{4}z^\alpha
z\nu\partial_\nu+\partial_{\overline\alpha}),
\end{equation}
$$
T^{\overline\alpha} \equiv V(N^{\overline\alpha}) =i\hskip 1pt
(\frac{k}{4}z^{\overline\alpha}
z{\overline\nu}\partial_{\overline\nu}+\partial_\alpha),
$$
Using this formulae we can calculate the commutators of the vector
fields $T$, $T^\alpha$ and $T^{\overline\alpha}$
\begin{equation} \label{te37}
[T^\alpha,T^\beta]=0,\qquad
[T^\alpha,T^{\overline\beta}]= i\hskip 1pt \frac{k}{4}(\delta^\alpha_\beta
T+
T^{\alpha\overline\beta}),
\end{equation}
$$
[T^\alpha,T\hskip 1pt]= -i\hskip 1pt T^\alpha,
\qquad [T^{\overline\alpha},T \hskip 1pt]= -iT^{\overline\alpha},
$$
where
$T^{\alpha\overline\beta}=$ $V(z^\alpha z^{\overline\beta }A^{-1})=$
$i \: (z^\alpha\partial_\beta- z^{\overline\beta}\partial_{\overline\alpha})$.
The generators $T$, $T^\alpha$ and $T^{\overline\alpha}$ do not form a
Lie
algebra but if we join to them the generator $T^{\alpha\overline\beta}$
then we obtain
\begin{equation} \label{te38}
[T^\alpha,T^{\beta\overline\gamma}]= i\delta^\alpha_\gamma
T^\beta,\qquad
[T^{\overline\alpha},T^{\beta\overline\gamma}]=-i\delta^{\alpha}_{\beta}
T^{\overline\gamma},
\end{equation}
$$
[T^{\alpha\overline\beta},T^{\gamma\overline\nu}]=i(\delta^\gamma_\beta
T^{\alpha\overline\nu}-\delta^\nu_\alpha T^{\gamma\overline\beta}).
$$
Because $T^\alpha,$ $T^{\overline\alpha},$ $T^{\alpha\overline\beta}$ are
linearily independent and $T= \Sigma\hskip 1pt T^{\alpha\overline\alpha}$
from (29), (30) it follows that $T^\alpha,\quad T^{\overline\alpha}$ and
$T^{\alpha\overline\beta}$ form the basis of $n(n+4)$-dimensional (over
${\bf
R}$) Lie algebra $ \ell (k)$ which is the Lie algebra of
infinitesimal isometries
of the space ${\cal K}_{2n}$ preserving the
complex structure.
The Poisson brackets $\{f,g\}=
\omega^{\alpha\overline\beta}(\partial_\alpha f \partial_{\overline\beta} g-
\partial_{\overline\beta}f \partial_\alpha g)$ of the functions
$H^{\alpha\overline\beta}=z^\alpha z^{\overline\beta}A^{-1}$, $N^\alpha=
z^\alpha A^{-1}$ and $N^{\overline\beta}= z^{\overline\beta}A^{-1}$ are
$$
\{N^\alpha,N^\beta\}=0,\qquad
\{N^\alpha,N^{\overline\beta}\}= i\frac{k}{4}(\delta^\alpha_\beta H+
N^{\alpha\overline\beta})-i\delta^\alpha_\beta,
$$
$$
\{N^\alpha,N^{\beta\overline\gamma}\}= i\hskip 1pt \delta^\alpha_\gamma
N^\beta,
\qquad
\{N^{\overline\alpha},N^{\beta\overline\gamma}\}= -i\hskip 1pt
\delta^\alpha_\beta N^{\overline\gamma},
$$
$$
\{ N^{\alpha\overline\beta},N^{\gamma\overline\nu}\}=
i(\delta^\alpha_\nu
N^{\gamma\overline\beta} -\delta^\gamma_\beta N^{\alpha\overline\nu}).
$$
Note that if we take the limit $k\to 0$ (\ref{te37}), (\ref{te38})
turns to
$$
[T^\alpha,T^\beta]=0, \qquad
[T^\alpha,T^{\overline\beta}]=0,
$$
\begin{equation} \label{te40}
[T^\alpha,T^{\beta\overline\gamma}]= i\hskip 1pt \delta^\alpha_\gamma
T^\beta,
\qquad [T^{\overline\alpha},T^{\beta\overline\gamma}]= -i\hskip 1pt
\delta^\alpha_\beta
T^{\overline\gamma},
\end{equation}
$$
[T^{\alpha\overline\beta},T^{\gamma\overline\nu}]=
i\hskip 1pt (\delta^\alpha_\nu
T^{\gamma\overline\beta} -\delta^\gamma_\beta T^{\alpha\overline\nu}).
$$
and define the $n(n+4)$-dimensional Lie algebra $\ell (0)$.
The curve $\ell (k)$ in the manifold of the
$n(n+4)$-dimensional Lie algebra structures is the deformation of
algebra $\ell (0)$ defined by the commutation relations
(\ref{te40}) and
containing the $n$-dimensional harmonic oscillator algebra ${\bf
osc}(n)$ as the Lie subalgebra. That is why we can consider the
mechanical system with the phase space ${\cal K}_{2n}$, symplectic
form $\omega$ and the observables $H$, $N^\alpha$ and $N^{\overline\alpha}$
as
the "deformation" of classical $n$-dimensional harmonic oscillator.
Now we quantize the classical mechanical systems
obtained in the preceding sections using the polarization $F$ defined
by (\ref{te17}). We calculate now $a[f]= \Sigma a[f]^\nu_\nu$ (see
(\ref{te11}))
for $f=H$, $N^\alpha$ and $N^{\overline\alpha}$. Substituting in
(\ref{te10})
$V(z^\alpha)= \omega^{\overline\nu\alpha}\partial_{\overline\nu}$ instead of
$X^\alpha$ and $T$, $T^\alpha,T^{\overline\alpha}$ instead of $V(f)$ we
find
using formulae (\ref{te27}) and (\ref{te36})
$$
[T,V(z^\beta)]= iV(z^\alpha),
$$
$$
[T^\alpha,V(z^\beta)]= -i\hskip 1pt
\frac{k}{4}(z^\alpha\delta^\beta_\nu
+z^\beta \delta^\alpha_\nu)V(z^\nu),
$$
$$
[T^{\overline\alpha},V(z^\beta)]=0
$$
whence
$$
a[N^\alpha]=-i \frac{k}{4}z^\alpha (n+1), \qquad a[N]= 0,\qquad
a[H]=in.
$$
After this from (\ref{te33})-(\ref{te35})
using (\ref{te5}), (\ref{te22}) and (\ref{te26}) we obtain
the following expressions for differential
operators in $L^{hol}_2 (U,dm)$ which are the quantizations of the
observables $H$, $N^\alpha$ and $N^{\overline\alpha}$
\begin{equation} \label{te45}
{\cal Q}H \equiv \hat H= \hbar(z^\nu \partial_\nu
\psi+\frac{n}{2}\psi),
\end{equation}
$$
{\cal Q}N^\alpha \equiv \hat N^\alpha =-\hbar\frac{k}{4}z^\alpha
z^\nu\partial_\nu \psi+ z^\alpha(1-\frac{\hbar k}{8}(n+1))
\psi,
$$
$$
{\cal Q}N^{\overline\alpha} \equiv \hat N^{\overline\alpha}
=\hbar\partial_\alpha \psi.
$$
Let $B:{\cal H}\to{\cal H}$ be the selfadjoint operator in Hilbert
space ${\cal H}$. The set $\sigma (B)= \{\rho \in {\bf R}\vert\hskip
1pt
\exists \hskip 1pt
\mu_\rho \in{\cal H}: \hskip 1pt B\mu_\rho= \rho\mu_\rho \}$
is called the
{\it spectrum} of the operator $B$. The number $\rho\in \sigma (B)
\subset
{\bf R}$ and section $\mu_\rho \in {\cal H}$
are called the {\it eigenvalue} of $B$ and {\it eigenstate}
with eigenvalue $\rho$.
Let us consider the eigenstate $\psi_E$ of the operator ${\cal Q}H$
with the eigenvalue $E$. Equation (\ref{te45}) yields
$$
\hbar (z^\nu\partial_\nu\psi_E +\frac{n}{2}\psi_E)= E\psi_E
$$
which is equivalent to
$$
z^\nu\partial_\nu\psi_E= (E\hbar^{-1}- \frac{n}{2})\psi_E
$$
whence $\psi_E$ is a homogeneous function of $z$ of degree $l=
(E\hbar^{-1}-
\frac{n}{2})$. Since $\psi_E$ is holomorphic it follows that $l$
is non-negative integer, so that the spectrum of ${\cal Q} H$ is
given by
$$
E_l= (l+ \frac{n}{2})\hbar,\quad l \in \{0\}\cup {\bf N}
$$
and coincides with the spectrum of the $n$-dimensional harmonic
oscillator Hamiltonian (\ref{te12}) (see for example [{\bf 12}]).
\\
\section{ Quantization of K\"ahler manifolds
\newline
admitting $H$-projective mappings}
In this section we consider the quantization of K\"ahler
spaces admitting $H$-projective mappings onto another K\"ahler
spaces. Let
$(M,\omega)$ and $(M^\prime, \omega ')$ be two K\"ahler
manifolds with fundamental forms $-\omega$ and $-\omega'$.
Let
$\varrho:M \to M'$ be $H$-projective mapping, it is well known
that $H$-projective mapping preserves the
complex structure. Therefore we can choose the local
complex chart $(z^\alpha,z^{\overline\alpha},U)$ in $M$ such that for each
point $p \in U$ with coordinates $(z^\alpha,z^{\overline\alpha})$ its
image $\varrho (p) \in \varrho (U)$ has the same coordinates.
The
necessary and sufficient condition for the mapping $\varrho:M \to
M'$
to be $H$-projective is expressed with the following equation [{\bf
13}]
\begin{equation} \label{te49}
b_{\alpha\overline\beta;\gamma}= 2 \phi'_\alpha g_{\overline\beta\gamma},
\end{equation}
where
$$
b_{\alpha\overline\beta}= e^{2 \phi}g'^{\overline\mu\nu} g_{\overline\mu\alpha}
g_{\nu\overline\beta}, \qquad b_{\alpha\beta}= b_{\overline\alpha\overline\beta}= 0,
$$
$$
\phi'_{\alpha}= \partial_\alpha \phi' =\partial_\mu \phi
e^{2 \phi}g'^{\mu\overline\nu} g_{\alpha\overline\nu},
$$
$$
J^i_k= J'^i_k, \qquad J^\mu_\nu= -J^{\overline\mu}_{\overline\nu}
=i\delta^\mu_\nu,\qquad J^\mu_{\overline\nu}= J^{\overline\mu}_\nu= 0,
$$
$\phi$ is some function on $U$ and semicolon denotes the
covariant derivation with respect to $g$.
Because of the positive definiteness of $g$ we can define the complex
frame $\{Z_A,Z_{\overline A}\}$ which is adapted for the Hermitian
structure
of {\it M} [{\bf 10}]. Then for the frame components of $g$ we have
$$
g_{A \overline B}= \delta^A_B.
$$
Transformations, preserving this form of $g$, belong to the
unitary
group $U(n)$ for each point $p \in M$. With the help of such
transformations we can choose the frame $\{Y_A,Y_{\overline B}\}$, so that
\begin{equation} \label{te51}
g_{A\overline B}= \delta ^A_B,
\qquad b_{A\overline B}= \lambda_A \delta^A_B,
\end{equation}
where $\lambda_A={\overline{\lambda}}_A$ are the roots of
the
$\lambda$-matrix $(b- \lambda g)$. Written in the frame (see for
example [{\bf 1,2}]) (\ref{te21}) and (\ref{te49}) have the form
\begin{equation} \label{te52}
Y_{\overline A} \hskip 2pt Y_{\overline B}f- \sum\limits_S
\gamma_{\overline B S\overline A} Y_S f= 0,
\end{equation}
\begin{equation} \label{te53}
\delta_{AB}Y_A\lambda_A +\sum\limits_S (\gamma_{\overline S A C}\lambda_S
\delta_{SB}+ \gamma_{S\overline B C}\lambda_S \delta_{SA})= 2\delta_{CB}Y_A
\phi',
\end{equation}
where $\gamma_{\overline S A C}$ $(\gamma_{ S A C}= \gamma_{\overline S \overline A
C})=0$ are Ricci rotation coefficients of the frame.
\vskip 2pt
Let $\{\theta^A, \theta^{\overline A}\},$ $\theta^A,\theta^{\overline A}\in
T^\ast M$ be the coframe dual to the frame $\{Y^A,Y^{\overline A}\}$. Then
the connection form $\alpha$ in the Hermitian line bundle ${\cal L}$
(see \S {\bf 1}) can be written in the following form
$$
\alpha= -iY_A \Phi \theta^A
$$
as in \S {\bf 2}. From (\ref{te15}) and (\ref{te51}) it follows
$\omega_{A\overline B} =\omega^{A\overline B}= i\delta^A_B$. Then for
$Y_A= \xi^\mu_A \partial_\mu$ we obtain
$$
V(z^\alpha)= -i\hskip 1pt \Sigma\hskip 1pt \xi^\alpha_B\hskip 1pt
Y_{\overline B}.
$$
If the function $F \in C^\infty (U)$ preserves polarization $F$ it
obeys the condition (\ref{te18}) which we can write as the form
$$
[V(f),V(z^\alpha)]= a[f]^\alpha_\mu V(z^\mu)= \sum\limits_{A,B}
(Y_{\overline A}f Y_A \xi^\alpha_B- Y_A f Y_{\overline A} \xi^\alpha_B+
\xi^\alpha_A (Y_{\overline A} Y_B f))Y_{\overline B},
$$
and here we find
$$a[f]= i\zeta^A_\alpha \sum\limits_B (Y_{\overline B}fY_B \xi^\alpha_A
-Y_B fY_{\overline B}\xi^\alpha_A+ \xi^\alpha_B Y_{\overline B}Y_A
f), $$
where $\zeta^B_\alpha$ are components of the inverse to
$(\xi^\alpha_B)$ matrix:
$\zeta^A_\mu \xi^\nu_A= \delta^\nu_\mu$.
At last we can evaluate the differential operator ${\cal Q}f$ in
$L^{hol}_2 (U,dm)$ for function $f\in C^\infty (U)$ obeying
(\ref{te52})
$$
({\cal Q})f \psi\equiv\hat f \psi= \hbar \sum\limits_A Y_{\overline A}fY_A
\psi- Y_A\Phi Y_{\overline A}f \psi+ f\psi- \frac{i\hbar}{2}a[f]\psi.
$$
To obtain concrete results we have to use specific expressions for
Ricci rotation coefficients and frame vector fields. In particular,
for
4-dimensional K\"ahler manifold admitting $H$-projective mappings
there
are two possibilities
$$
1)\quad\lambda_1= \lambda_2,\qquad 2) \quad \lambda_1 \neq
\lambda_2.
$$
In the first case we have $b= \lambda g$ and, hence, $g'=\mu g$
where $\mu= (\lambda e^{2 \phi})^{-1}$. Then (\ref{te53})
implies $\mu= const$, and
$H$-projective mappings are only rescalings of the metric.
The second case is more interesting. If $\lambda_1 \neq\lambda_2$ then
from (\ref{te53}) it follows [{\bf 3}]
$$
Y_1 \lambda_1= Y_{\overline 1}\lambda_2= Y_2\lambda_1= Y_{\overline
2}\lambda_1= 0,\qquad \gamma_{1\overline 2 2 }=Y_1 \ln | \lambda_1-
\lambda_2 |,
$$
$$
\gamma_{\overline 1 2 1 }=-Y_2 \ln |\lambda_1- \lambda_2|, \qquad
\gamma_{\overline 2 1 1 }=\gamma_{\overline 1 2 2 }.
$$
\bigskip
This work was partially supported by grant 1749
of International Science Foundation and grant RFFI-94-01-01118-a of
Rus\-sian Foun\-da\-tion for Fun\-da\-men\-tal In\-ves\-ti\-ga\-ti\-
ons.
\newpage
\centerline{\bf References}
\medskip
\noindent
[{\bf 1}] A.V. Aminova: On skew-orthonormal frame and parallel
symmetric
bilinear form on Riemannian manifolds,
{\it Tensor, N.S.},{\bf 45} (1987), 1-13.
\noindent
[{\bf 2}] A.V. Aminova: Pseudo-Riemannian manifolds with common
geodesics,
{\it Uspekhi Matematicheskikh Nauk}, {\bf 48} (1993), 107-159.
\noindent
[{\bf 3}] A.V. Aminova and D.A. Kalinin: $H$-pro\-jec\-ti\-vely
equi\-va\-lent
four-di\-men\-si\-onal Rieman\-nian connections,
{\it IZV. VUZ. Matem.}, No.8 (1994), 11-21.
\noindent
[{\bf 4}] R. Balbinot, A.El Gradechi, J.-P. Gazeau
and B.Giorgini: Phase space
for quantum elementary systems in anti-de Sitter and Minkowski
spacetimes,
{\it J. Phys. A},{\bf 25} (1992), 1185-1210.
\noindent
[{\bf 5}] F.A. Berezin: General conception of quantization,
{\it Comm. Math. Phys.}, {\bf 40} (1975), 153-174.
\noindent
[{\bf 6}] J.-P. Gazeau and V.Hussin: Poicar\'e contraction of
$SU(1,1)$ Fock-Bargmann
structure, {\it J.Phys. A}, {\bf 25} (1992), 1549-1573.
\noindent
[{\bf 7}] J.-P. Gazeau and J. Renaud: Lie algorithm for
an interacting $SU(1,1)$
elementary systems and its contraction,
{\it Preprint Universit\'e Paris VII,
PAR-LPTM-92}.
\noindent
[{\bf 8}] A.A. Kirillov: Geometric quantization. In:"Mo\-dern prob\-
lems of
mathematics" (Itogi Nauki i Tekniki),
{\bf 4}, VINITI SSSR, Moscow, 1985, 141-204 {\sloppy
}
\noindent
[{\bf 9}] S. Kobayashi and K. Nomizu: Foundation of differential
geometry,
V. II, Intersci. Publ., N.Y., 1969.
\noindent
[{\bf 10}] A. Lichnerowicz: Theorie globale des connexiones et des
groupes
d'ho\-lo\-no\-mie, Cremonese, Roma. 1955.
\noindent
[{\bf 11}] T. Otsuki and Y. Tashiro: On curves in Kahlerian spaces,
{\it Math. J. Okayama Univ.}, {\bf 4}(1954), 57-78.
\noindent
[{\bf 12}] J. Sniatycki: Geometric quantization and quantum
mechanics,
Springer, Berlin etc, 1980.
\noindent
[{\bf 13}] N.S. Sinyukov: Geodesic mappings of Riemannian spaces,
{\it Moscow, Nauka}, 1979.
\noindent
[{\bf 14}] J.M. Tuynman: Quantization. Towards a comparison
between
methods, {\it J. Math. Phys.}, {\bf 28}(1987). 2829-2840.
\end{document}
|
\section*{Acknowledgments}
This work was supported by Deutsche Forschungsgemeinschaft under grant SFB-375
and the EC programs SC1-CT91-0729 and SC1-CT92-0789
|
\section*{Abstract}
\else \small
\begin{center}
{\bf ABSTRACT}
\end{center}
\quotation
\fi}
\newif\iffn\fnfalse
\@ifundefined{reset@font}{\let\reset@font\empty}{}
\long\def\@footnotetext#1{\insert\footins{\reset@font\footnotesize
\interlinepenalty\interfootnotelinepenalty
\splittopskip\footnotesep
\splitmaxdepth \dp\strutbox \floatingpenalty \@MM
\hsize\columnwidth \@parboxrestore
\edef\@currentlabel{\csname p@footnote\endcsname\@thefnmark}\@makefntext
{\rule{\z@}{\footnotesep}\ignorespaces
\fntrue#1\fnfalse\strut}}}
\makeatother
\ifamsf
\newfont{\bf}{msbm10 scaled\magstep2}
\newfont{\bbbfont}{msbm10 scaled\magstep1}
\newfont{\smallbbbfont}{msbm8}
\newfont{\tinybbbfont}{msbm6}
\newfont{\smallfootbbbfont}{msbm7}
\newfont{\tinyfootbbbfont}{msbm5}
\fi
\ifscrf
\newfont{\scrfont}{rsfs10 scaled\magstep1}
\newfont{\smallscrfont}{rsfs7}
\newfont{\tinyscrfont}{rsfs7}
\newfont{\smallfootscrfont}{rsfs7}
\newfont{\tinyfootscrfont}{rsfs7}
\fi
\ifamsf
\newcommand{\bf}[1]{\iffn
\mathchoice{\mbox{\footbbbfont #1}}{\mbox{\footbbbfont #1}}
{\mbox{\smallfootbbbfont #1}}{\mbox{\tinyfootbbbfont #1}}\else
\mathchoice{\mbox{\bbbfont #1}}{\mbox{\bbbfont #1}}
{\mbox{\smallbbbfont #1}}{\mbox{\tinybbbfont #1}}\fi}
\else
\def\bf{\bf}
\def\bf{\bf}
\fi
\ifscrf
\newcommand{\cal}[1]{\iffn
\mathchoice{\mbox{\footscrfont #1}}{\mbox{\footscrfont #1}}
{\mbox{\smallfootscrfont #1}}{\mbox{\tinyfootscrfont #1}}\else
\mathchoice{\mbox{\scrfont #1}}{\mbox{\scrfont #1}}
{\mbox{\smallscrfont #1}}{\mbox{\tinyscrfont #1}}\fi}
\else
\def\cal{\cal}
\fi
\def\operatorname#1{\mathop{\rm #1}\nolimits}
\def{\Bbb C}{{\bf C}}
\def{\cal F}{{\cal F}}
\def{\cal O}{{\cal O}}
\def{\Bbb P}{{\bf P}}
\def{\Bbb Q}{{\bf Q}}
\def{\Bbb R}{{\bf R}}
\def{\Bbb Z}{{\bf Z}}
\def\operatorname{Aut}{\operatorname{Aut}}
\def\mathop{\widetilde{\rm Aut}}\nolimits{\mathop{\widetilde{\rm Aut}}\nolimits}
\def\operatorname{Hom}{\operatorname{Hom}}
\def\operatorname{Ker}{\operatorname{Ker}}
\def\ |\ {\ |\ }
\def\operatorname{Spec}{\operatorname{Spec}}
\def\operatorname{Area}{\operatorname{Area}}
\def\operatorname{Vol}{\operatorname{Vol}}
\def\operatorname{gen}{\operatorname{gen}}
\def\operatorname{div}{\operatorname{div}}
\def\operatorname{Div}{\operatorname{Div}}
\def\operatorname{WDiv}{\operatorname{WDiv}}
\def\operatorname{ad}{\operatorname{ad}}
\def\operatorname{tr}{\operatorname{tr}}
\def\operatorname{CPL}{\operatorname{CPL}}
\def\operatorname{cpl}{\operatorname{cpl}}
\def\operatorname{Im}{\operatorname{Im}}
\def\operatorname{Re}{\operatorname{Re}}
\def\operatorname{Gr}{\operatorname{Gr}}
\def\operatorname{rank}{\operatorname{rank}}
\def\opeq#1{\advance\lineskip#1 \advance\baselineskip#1
\advance\lineskiplimit#1}
\def\eqalignsq#1{\null\,\vcenter{\opeq{2.5\jot}\mathsurround=0pt
\everycr={}\tabskip=0pt\offinterlineskip
\halign{\strut\hfil$\displaystyle{##}$&$\displaystyle{{}##}$\hfil
\crcr#1\crcr}}\,\null}
\def\eqalign#1{\null\,\vcenter{\opeq{2.5\jot}\mathsurround=0pt
\everycr={}\tabskip=0pt
\halign{\strut\hfil$\displaystyle{##}$&$\displaystyle{{}##}$\hfil
\crcr#1\crcr}}\,\null}
\def$\sigma$-model{$\sigma$-model}
\defnon-linear \sm{non-linear $\sigma$-model}
\def\sm\ measure{$\sigma$-model\ measure}
\defCalabi-Yau{Calabi-Yau}
\defLandau-Ginzburg{Landau-Ginzburg}
\def{\Scr R}{{\cal R}}
\def{\Scr M}{{\cal M}}
\def{\Scr A}{{\cal A}}
\def{\Scr B}{{\cal B}}
\def{\Scr K}{{\cal K}}
\def{\Scr D}{{\cal D}}
\def{\Scr T}{{\cal T}}
\defS_{\hbox{\scriptsize LG}}{S_{\hbox{\scriptsize LG}}}
\def\cM{{\Scr M}}
\def{\hfuzz=100cm\hbox to 0pt{$\;\overline{\phantom{X}}$}\cM}{{\hfuzz=100cm\hbox to 0pt{$\;\overline{\phantom{X}}$}{\Scr M}}}
\def{\hfuzz=100cm\hbox to 0pt{$\;\overline{\phantom{X}}$}\cD}{{\hfuzz=100cm\hbox to 0pt{$\;\overline{\phantom{X}}$}{\Scr D}}}
\defalgebraic measure{algebraic measure}
\def\ff#1#2{{\textstyle\frac{#1}{#2}}}
\def{\cal F}#1#2{{}_{#1}F_{#2}}
\defV_{\Delta}{V_{\Delta}}
\defV_\delta{V_\delta}
\def$R\leftrightarrow1/R${$R\leftrightarrow1/R$}
\def\normalord#1{\mathord{:}#1\mathord{:}}
\def\Gep#1{#1_{\hbox{\scriptsize Gep}}}
\def\th#1{$#1^{\hbox{\scriptsize\it th}}$}
\def\cMs#1{{\Scr M}_{\hbox{\scriptsize #1}}}
\ifamsf
\def.{\mathbin{\mbox{\bbbfont\char"6E}}}
\def.{\mathbin{\mbox{\bbbfont\char"6F}}}
\else
\def.{.}
\def.{.}
\fi
\begin{document}
\setcounter{page}0
\title{\large\bf Some Relationships Between\\ Dualities in String Theory\\[10mm]
\insert\footins{\hbox to\hsize{\footnotesize
Talk given at ``S-Duality and Mirror Symmetry'', Trieste June 1995.\hfil}}}
\author{
Paul S. Aspinwall\\[0.7cm]
\normalsize F.R.~Newman Lab.~of Nuclear Studies,\\
\normalsize Cornell University,\\
\normalsize Ithaca, NY 14853\\[10mm]
}
{\hfuzz=10cm\maketitle}
\def\large{\large}
\def\large\bf{\large\bf}
\vskip 1.5cm
\vskip 1cm
\begin{abstract}
Some relationships between string theories and eleven-dimensional
supergravity are discussed and reviewed. We see how some relationships
can be derived from others. The cases of $N=2$ supersymmetry in nine
dimensions and $N=4$ supersymmetry in four dimensions are discussed in
some detail. The latter case leads to consideration of quotients of a
K3 surface times a torus and to a possible peculiar relationship
between eleven-dimensional supergravity and the heterotic strings in
ten dimensions.
\end{abstract}
\vfil\break
\section{Introduction} \label{s:intro}
Recent ideas concerning duality in string theory (such as
\cite{Sen:4d,HT:unity,W:dyn}) have given hope to gaining insights
into some non-perturbative form of string theory. Given the current
status of string theory it is not easy to see how to prove such
statements about duality. Rather, one can take the attitude that such
dualities could be used, in part, as a defining property of string
theory.
Given the many dualities that have been proposed, if we want to
understand how to formulate a new form of string theory, it is
important to know which dualities can be derived from the others.
In particular we appear to have many forms of dualities relating
theories with $N$ supersymmetries in $d$-dimensional flat space-time
for various values of $N$ and $d$. In this talk I will give some simple
ideas on how to formulate relationships
between dualities by concentrating on the cases $N=2,d=9$ and
$N=4,d=4$.
Much of this talk is not original and draws particularly
heavily from \cite{W:dyn} and the later sections are based on the
collaborative work of \cite{AM:Ud}.
Many aspects of section \ref{s:9d} were discussed in \cite{BHO:d=9}
although not in quite the same way as here. The way that the
$U$-duality group is built up in section \ref{s:U} is very similar in
spirit to the work of \cite{Sen:3d}.
It is hoped that the simple
examples explained below show how the dualities can be directly
related to each other in some contexts to build up a rather intricate
picture.
In section \ref{s:dual} we will review the basic dualities used in
this talk. In section \ref{s:9d} we do a ``warm-up'' exercise for the
later sections. In section \ref{s:4d} we have an overview of four
dimensional theories and in section \ref{s:U} we look at the simplest
example of an $N=4$ theory in four dimensions.
\section{Dualities} \label{s:dual}
``Duality'' is a much over-used word in the context we wish to use it
and we need to refine our definitions somewhat. Firstly let us discuss
$U$-duality as discussed in \cite{HT:unity}. Consider a particular
string theory. Such a theory will have some deformations (e.g., ``truly
marginal operators'' in the language of conformal field theory) which
will allow us to smoothly reach other string theories. Let us use
${\Scr M}$ to denote the moduli space of such theories. To avoid
complications we will allow ${\Scr M}$ to include boundary points a finite
distance away, but we will
not allow ourselves to pass through the boundary to other theories by
processes such as the one described in \cite{GMS:con}. In simple cases
one expects the moduli space to appear naturally in the form
\begin{equation}
{\Scr M} = U\backslash {\Scr T},
\end{equation}
where ${\Scr T}\,$ is some smooth domain and $U$ is some discrete group
acting upon it. ${\Scr T}\,$ is some generalized notion of a
Teichm\"uller space and $U$ is the group of $U$-dualities. We divide
by discrete groups from the left as ${\Scr T}\,$ will typically be a
right-coset as we will see later.
In general one can expect $U$ to be generated by 3 subsets defined
roughly as follows:
\begin{enumerate}
\item $C$-dualities: (This is not conventional notation.) These are
equivalences in ${\Scr T}\,$ coming from the classical modular group. That is, if
we can associate our string theory with some geometry, the classical
moduli space of the geometric object will be $C\backslash{\Scr T}$. The
canonical example is $Sl(2,{\Bbb Z})$ for the moduli space of complex
structures on a 2-torus.
\item $T$-dualities: These are further identifications within ${\Scr T}\,$
due to the conformal field theories associated with two different
geometries being isomorphic. The canonical example is $R\leftrightarrow1/R$\
duality. In some conventions $T$ is a group that contains $C$.
\item $S$-dualities: These are further identifications due to the
effective quantum field theories associated to the string target space
for two apparantly different models being isomorphic. The canonical example
is strong-weak string-coupling duality.
\end{enumerate}
It is generally hoped that the full group $U$ is generated completely
by the elements of $C$, $T$ and $S$.
Together with the notion of $U$-dualities we also have the concept of
equivalences between theories which, at first sight, are qualitatively
different. We list the ones needed in this talk below.
\begin{enumerate}
\item String-string duality. The type IIA superstring compactified on
a K3 surface is equivalent to the heterotic string compactified on a
4-torus. We will denote this relationship by
\begin{equation}
({\rm IIA} \to {\rm K3}) \cong ({\rm Het} \to T^4).
\end{equation}
This notion goes back as far as \cite{Sei:K3} but has been developed
subsequently in many other references. The strongly coupled type II
string corresponds to the weakly coupled heterotic string.
\item 11-dimensional supergravity as a string theory
\cite{T:11d,W:dyn}.
\begin{equation}
({\rm 11d} \to S^1) \cong {\rm IIA}. \label{eq:11d}
\end{equation}
In this case the string coupling of the type II string becomes larger
as the radius of the $S^1$ becomes larger.
\item Type II equivalences \cite{DHS:IIAB,DLP:IIAB}.
\begin{equation}
({\rm IIA} \to S^1) \cong ({\rm IIB} \to S^1).
\end{equation}
In this case there is an $R\leftrightarrow1/R$\ relationship between the two $S^1$s.
\item Heterotic equivalences \cite{N:torus,Gins:torus}.
\begin{equation}
({\rm Het}_{E_8\times E_8} \to S^1)\cong ({\rm Het}_{SO(32)}\to
S^1).
\end{equation}
In this case the two 10-dimensional heterotic strings are different
limits in the space $O(1,17)/O(17)$.
\end{enumerate}
It is fairly clear what is meant by each of the above equivalences
with the exception of that for equation (\ref{eq:11d}). Are we really
meant to believe that eleven-dimensional supergravity on a circle
is entirely equivalent to string theory? The answer to this question
is probably no. In \cite{W:dyn} this equivalence was more
conservatively given as that between low-energy effective actions. We
should be aware of this uncertainty whenever eleven-dimensional
supergravity is mentioned below.
\section{Nine Dimensions} \label{s:9d}
We may now try to mix the ideas of $U$-duality and equivalences from
the previous section. Consider the case of eleven-dimensional
supergravity compactified on a 2-torus, $T^2$. From the last section
we therefore have
\begin{equation}
\eqalign{({\rm 11d}\to T^2) &\cong \left[({\rm 11d}\to S^1)
\to S^1\right]\cr
&\cong ({\rm IIA}\to S^1)\cr
&\cong ({\rm IIB}\to S^1),\cr}
\end{equation}
thus relating eleven-dimensional supergravity to the IIB superstring.
Now let $T^2$ be given by a fundamental region in ${\Bbb R}^2$ in the usual
way of the form of a rectangle with sides $r_1$ and $r_2$. The
starting point for our space of theories is thus a quadrant of
${\Bbb R}^2$. Since the interchange of $r_1$ and $r_2$ clearly has no effect
on the underlying theory, we should divide out by this
interchange. This leads to an infinite triangle as shown in figure
\ref{fig:a}.
\iffigs
\begin{figure}
\centerline{\epsfxsize=9cm\epsfbox{trl2-fa.ps}}
\caption{A slice of the space of theories in 9 dimensions.}
\label{fig:a}
\end{figure}
\fi
A generic point in this space corresponds to a
nine-dimensional theory. When both radii go to infinity we obtain the
eleven-dimensional theory. Consider the case when $r_1$ is finite and
$r_2$ is infinite. This gives us the correspondence with the IIA
theory as explained in \cite{W:dyn}. Let us denote the IIA string
coupling by $\lambda_A=\exp(\phi_A)$, where $\phi$ is the string
dilaton. We then have
\begin{equation}
\lambda_A = r_1^{\frac32}.
\end{equation}
Thus the bottom left corner of figure \ref{fig:a} is the type IIA
string in ten dimensions at zero string coupling.
An important point of \cite{W:dyn} is that the ten space-time
dimensions as seen by eleven-dimensional supergravity compactified on
a circle are not quite the same ten space-time dimensions as seen by
the type IIA superstring. They are related by a rescaling. This means
that when $r_2$ is finite, it is not the radius of the circle on which
the type IIA string is compactified. Denoting this latter radius by $r_A$ we
have
\begin{equation}
r_A = r_1^{\frac12}r_2.
\end{equation}
Now consider the type IIB interpretation. The effective field theories
for the type IIA and IIB theories show that the respective dilatons
must shift when the $R\leftrightarrow1/R$\ transformation of \cite{DHS:IIAB,DLP:IIAB}
is performed. This means that we calculate the string coupling of the
type IIB superstring as
\begin{equation}
\eqalign{
\lambda_B &= \lambda_A.r_A^{-1}\cr
&= r_1r_2^{-1}.\cr}
\end{equation}
Consider now the bottom right corner of figure \ref{fig:a}. This now
corresponds to the type IIB string in 10 dimensions but the coupling
is not defined. We should really do a real blow-up at this point do
get the correct moduli space. It is easy to see that the symmetry of
the eleven-dimensional supergravity picture that exchanged the radii
$r_1$ and $r_2$ now translates into the IIB superstring as
\begin{equation}
\lambda_B \leftrightarrow 1/\lambda_B.
\end{equation}
That is, we have obtained $S$-duality for the IIB string.
Actually, we have not analyzed the complete moduli space. The moduli
space of the torus should also allow the angle between the vectors of
length $r_1$ and $r_2$ to vary. This gives the well-known result that
the moduli space is actually the upper half plane divided by
$Sl(2,{\Bbb Z})$. In the language of the type IIB superstring, this extra
degree of freedom comes from the expectation value of the axion. Thus,
the $Sl(2,{\Bbb Z})$ modular invariance of the torus on which
eleven-dimensional supergravity was compactified can be used to ``deduce''
$Sl(2,{\Bbb Z})$ $S$-duality for the IIB string.
This $S$-duality for the type IIB string was conjectured in
\cite{HT:unity}. We see here that this conjecture is not
independent of the others in the previous section.
\section{Four Dimensional Theories} \label{s:4d}
Let us consider obtaining four-dimensional theories by compactifying
the known supersymmetric ten-dimensional string theories and
eleven-dimensional
supergravity over manifolds of six and seven dimensions
respectively. The number of supersymmetries in four dimensions can be
found by counting the number of covariantly constant spinors on the six
and seven-dimensional manifolds. This in turn depends purely on the
holonomy group of the compact manifold. In table \ref{t:list} we list
the number of supersymmetries in four dimensions for each higher
dimensional theory.
\begin{table}
\def\st{\rule[-1.5ex]{0em}{4ex}}
\centerline{\begin{tabular}{|c|c|c|c|}
\hline
\st $N$ & 11d & II & Het \\
\st &Hol,$X$ &Hol,$X$ &Hol,$X$ \\
\hline
\st 8& 1,$T^7$& 1,$T^6$& -- \\
\st 4& $SU(2)$, K3$\times T^3$& $SU(2)$, K3$\times T^2$& 1, $T^6$\\
\st 2& $SU(3)$, CY$\times S^1$& $SU(3)$, CY& $SU(2)$, K3$\times T^2$\\
\st 1& $G_2$, Joyce& --& $SU(3)$, CY\\
\hline
\end{tabular}}
\caption{Four-dimensional theories obtained by compactification.}
\label{t:list}
\end{table}
This table requires some discussion. Firstly we only list possible
geometric compactifications. By using more asymmetric methods, other
models can be built such as an $N=1$ theory built from the type II
string (see, for example, \cite{LNS:asym}). Each of the
entries in the table gives the holonomy of the compact space $X$,
followed by an example of such a space where CY stands for a
Calabi-Yau manifold. A ``Joyce Manifold'' is that of the type
discovered in \cite{Joyce:G2}.
It is tempting to conjecture that for each of the rows in table
\ref{t:list} there is some equivalence between each of the
entries. For the $N=8$ this follows immediately from the conjectured
equivalence between eleven-dimensional supergravity and the type IIA
string by compactifying further on $T^6$. Similarly the $N=4$ row
follows from equivalences mentioned earlier. Analysis on the $N=2$ row
was begun in \cite{KV:N=2,FHSV:N=2}.
In some cases one can classify all the possibilities for $X$ given the
holonomy (see theorem 10.8 of \cite{Sal:hol}).
For the rest of this section we will analyze the case of obtaining
$N=4$ supersymmetry from the type II string where this classification
may be done. Any 6-dimensional manifold with holonomy $SU(2)$ must be
of the form $({\rm K3}\times T^2)/G$ where $G$ acts freely. Any
element $g\in G$ can be decomposed into an automorphism, $g_1$, of the
K3 surface and an automorphism, $g_2$, of the torus. To retain $SU(2)$
holonomy, these automorphisms must preserve the holomorphic 2-form and
1-form respectively. Such a $g_1$ necessarily has fixed points and so
$g_2$ must act freely. Clearly then, if $g$ is nontrivial, $g_2$ acts
by a translation on the torus.
We can list all possibilities for the group $G$ in this case. Since
any nontrivial element of $G$ must be fixed-point free, the associated
$g_2$ must be nontrivial. Thus $G$ is faithfully represented by
translations in $T^2$. It then follows that $G$ must be of the form
${\Bbb Z}_m$ or ${\Bbb Z}_m\times{\Bbb Z}_n$ for integers $m,n$. Without loss of
generality we may assume that any element of $G$ acts nontrivially on
the K3 surface. From Nikulin's work \cite{Nik:K3aut} we can then list
the possibilities for $G$. This is done in table \ref{t:K3aut}. $M$ is
the rank of the maximal sublattice of $H^2({\rm K3},{\Bbb Z})$ that
transforms nontrivially under $G$.
\begin{table}[b]
\centerline{\begin{tabular}{|c||c|c|c|c|c|c|c|c|c|c|c|c|}
\hline
$G$&${\Bbb Z}_2$&${\Bbb Z}_2\times{\Bbb Z}_2$&${\Bbb Z}_2\times{\Bbb Z}_4$&${\Bbb Z}_2\times{\Bbb Z}_6$
&${\Bbb Z}_3$&${\Bbb Z}_3\times{\Bbb Z}_3$
&${\Bbb Z}_4$&${\Bbb Z}_4\times{\Bbb Z}_4$&${\Bbb Z}_5$&${\Bbb Z}_6$&${\Bbb Z}_7$&${\Bbb Z}_8$\\
\hline
$M$&8&12&16&18&12&16&14&18&16&16&18&18\\
\hline
\end{tabular}}
\caption{Possible quotienting groups.}
\label{t:K3aut}
\end{table}
Given a type II string compactified on a manifold $X_G=({\rm K3}\times
T^2)/G$, it is natural to ask if this is equivalent to some orbifold
of the heterotic string compactified on $T^6$. That is, can we divide
the $N=4$ row in table \ref{t:list} by $G$ and maintain equivalences?
This appears to be the
case. From \cite{AM:K3p} we expect to identify the lattice of total
cohomology $H^*({\rm K3},{\Bbb Z})$ with the even-self dual lattice defining
the heterotic string compactified on $T^4$. Thus the action of $G$ on
$H^*({\rm K3},{\Bbb Z})$ gives us a candidate for an asymmetric orbifold of
the heterotic string. This appears to correspond precisely to the
models studied
in \cite{CHL:bigN,CP:ao}. This point has been investigated further
in \cite{CL:KT/G}.
Two points are worth mentioning. Firstly, the asymmetric orbifolds of
\cite{CP:ao} should provide more examples of heterotic strings than
we have listed here. This is because we have restricted our attention
on the type II side to geometric quotients.
Other models based on type II strings are possible, such as those of
\cite{FK:fc}.
Secondly when one does an
asymmetric orbifold it is important to check that the level-matching
conditions of \cite{Vafa:tor} are satisfied. Given the values of $M$
in table \ref{t:K3aut} we can check whether this is so for all our
examples. The answer is yes and, at least at first sight, this appears
to be remarkable. This should be contrasted to cases where $N=4$
supersymmetry is broken such as \cite{FHSV:N=2}.
\section{$U$-duality} \label{s:U}
In this section we will analyze the moduli space of the $N=4$ theories
as discussed in the last section. We will focus on the case of
K3$\times T^2$ but it should be easy to extend this analysis to the
quotients. A conjecture for the form of this moduli space was made in
\cite{HT:unity} by making some assumptions about the soliton
spectrum. Here we will be able to rederive this result without making
any direct reference to solitons but rather using the equivalences we
already listed earlier. One can argue that these equivalences rest on
details of the soliton spectrum and so what we are doing in this
section may be completely equivalent to \cite{HT:unity}. Anyway the
analysis below clearly shows
the interrelation between such conjectures. This argument first
appeared in \cite{AM:Ud} and the reader is referred there for
details. The basic idea will be that we will take the moduli space of
theories and try to identify a boundary. This process is not unique
and the different boundaries will correspond to different
interpretations of the theory. The mathematical principles of this
process are in \cite{BS:corner} but the reader is also referred to
\cite{W:dyn,AM:Ud} for a simpler treatment.
By general arguments from supergravity \cite{deRoo:}, the general form
of the Teichm\"uller space for $N=4$ theories in four dimensions is
\begin{equation}
{\Scr T} \cong \frac{O(6,k)}{O(6)\times O(k)} \times \frac{Sl(2)}{U(1)},
\end{equation}
where $k$ is the number of $N=4$ matter supermultiplets. For ease of
notation, when a coset is written $\ff{a}{b}$, the action is assumed
to be from the right. In this case
$k=22$ (or $k=22-M$ for the cases listed in table \ref{t:K3aut}). To
form the moduli space we need to quotient by some group $U$.
{}From the conjecture concerning the rows of table \ref{t:list}, any
point in ${\Scr M} = U\backslash{\Scr T}$ can be thought of as a
compactification of eleven-dimensional supergravity, the type IIA or
IIB superstring, or the two heterotic strings. Thus each point has
five interpretations (compared to the three interpretations in figure
\ref{fig:a}). Given any one of these five interpretations we should be
able to find part of the moduli space we already understand
nonperturbatively.
Let us begin with the type IIB string. We will assume that we
understand this theory in the weak-coupling limit, i.e., when
$\lambda_B\to0$. In this case we
should just recover the Teichm\"uller space for conformal field theories with
target space K3$\times T^2$ \cite{Sei:K3} together with directions in
the moduli
space for deformations of the axion and 48 fields from the R-R sector.
Thus we expect to be able to find a boundary of the form
\begin{equation}
\partial_{\lambda_B\to0}{\Scr T}\;\cong\;
\frac{O(4,20)}{O(4)\times O(20)}\times\frac{Sl(2)}{U(1)}
\times\frac{Sl(2)}{U(1)}\times{\Bbb R}^{49},
\label{eq:BB}
\end{equation}
for the Teichm\"uller space. This is indeed the case following methods
explained in \cite{W:dyn,AM:Ud}. Now since we know that
$O(4,20;{\Bbb Z})\times Sl(2,{\Bbb Z})\times Sl(2,{\Bbb Z})$ acts on the above boundary,
it must also be a subgroup of $U$.
We also know about another limit of this IIB theory. If we rescale the
$T^2$ part so that its area becomes infinite then we are left with a
type IIB string compactified on a K3 surface. The Teichm\"uller space
we are left with should be that for IIB strings on a K3 surface
\cite{W:dyn} together with the complex structure and $B$-field for
$T^2$ and the remaining R-R moduli. This is of the form
\begin{equation}
\partial_{T^2\to\infty}{\Scr T}\;\cong\;
\frac{O(5,21)}{O(5)\times O(21)}\times\frac{Sl(2)}{U(1)}
\times{\Bbb R}^{26},
\end{equation}
which can also be found as a boundary. In this case we show that
$O(5,21;{\Bbb Z})\times Sl(2,{\Bbb Z})\subset U$. The relationship between the
boundaries tells us the way these two subgroups fit together within
$U$. Using methods such as those in \cite{AM:K3p} or \cite{Giv:rep}
one can then show that
\begin{equation}
U \supseteq O(6,22;{\Bbb Z}) \times Sl(2,{\Bbb Z}),
\label{eq:U}
\end{equation}
and that the equality must be satisfied if ${\Scr M}$ is Hausdorff
\cite{AM:K3p}.
No we have found $U$ we can interpret ${\Scr M}$ in terms of the other
strings. Firstly we can find the IIA string interpretation. Going to
the weak-coupling string where we really understand what we are doing,
this theory is mirror to the IIB theory. The action of the mirror
map on (\ref{eq:BB}) is within the $O(4,20)$ factor but it exchanges
the two $Sl(2)$ factors. Thus, the only noticeable effect is on the
r\^ole of the $Sl(2,{\Bbb Z})$ factor in (\ref{eq:U}). For the IIB string
this factor came from the complex structure of $T^2$ whereas now it
acts as a $T$-duality on
the radius and $B$-field as in \cite{DVV:torus}.
Now consider the heterotic string. In this case we know that weakly
coupled string has a Teichm\"uller space of $O(6,22)/(O(6)\times
O(22))\times {\Bbb R}$ thanks to \cite{N:torus}. This is easy to fit into
the required moduli space. In this case the only extra information
coming from the $U$-duality group is the $Sl(2,{\Bbb Z})$ factor which forms
the $S$-duality group.
Thus we see that for the type IIA, type IIB and heterotic string the
r\^ole of the $Sl(2,{\Bbb Z})$ group is of a $T$, $C$ and $S$ duality
respectively. This ``triality of dualities'' generalizes the work of
\cite{Duff:S} (and was independently investigated in \cite{DLR:tri}
and recently in \cite{Kal:tri}).
Lastly we need to fit the interpretation of the four-dimensional model
as a compactification of eleven-dimensional supergravity into the
picture. Eleven-dimensional supergravity does not have any weak
coupling limit since the coupling in the action is fixed. However, we
can take the large radius limit. Actually, there are many large radius
limits which appear to be qualitatively different.
In \cite{W:dyn} it was argued that eleven-dimensional supergravity
compactified on a K3 surface was equivalent to the heterotic string
compactified on a 3-torus. Therefore, we should be able to identify
eleven-dimensional supergravity compactified on K3$\times T^3$. Sure
enough, one of the boundaries of the space is
\begin{equation}
\partial_{R_1\to\infty}{\Scr T}\;\cong\;
\frac{O(3,19)}{O(3)\times O(19)}\times\frac{Sl(3)}{SO(3)}
\times\frac{Sl(2)}{U(1)}\times{\Bbb R}^{69},
\end{equation}
where the first factor can be recognized as the Teichm\"uller space of
Ricci-flat metrics on a K3 surface of fixed volume and the second
factor as the Teichm\"uller space of flat metrics of xied volume on
$T^3$. $R_1$ is
some parameter such that the limit $R_1\to\infty$ takes the volume of
both the K3 surface and the 3-torus to infinity. The other degree of
freedom for the two volumes is in the $Sl(2)/U(1)$ factor. It would
thus appear that the $Sl(2,{\Bbb Z})$ factor in the $U$-duality group has
yet another meaning for the eleven-dimensional supergravity picture.
It is interesting to
note that as soon as we interpret our moduli space in terms of
eleven-dimensional supergravity we recognize factors in the boundary
which correspond {\em classical\/} moduli spaces rather than the
moduli spaces of conformal field theories we were seeing earlier. This
is intimately connected with the fact that a conformal field theory
moduli space has a classical moduli space on its boundary (from the
$\alpha^\prime\to0$ limit).
Another boundary of the moduli space can be written as
\begin{equation}
\partial_{R_2\to\infty}{\Scr T}\;\cong\;
\frac{O(2,18)}{O(2)\times O(18)}\times\frac{Sl(4)}{SO(4)}
\times\ldots
\end{equation}
The second factor is clearly the space of metrics on $T^4$ but what is
the first factor? It can be written as part of the boundary of the
space of Ricci-flat metrics on a K3 surface and so is the space of
some kind of singular K3. One natural interpretation \cite{AM:Ud} is
that the K3 surface has collapsed in itself and is now like a 3
dimensional object. We denote this as a ``squashed K3''. Thus, our
four-dimensional theory has an interpretation as eleven-dimensional
supergravity compactified on a squashed K3 times $T^4$ (both being at
large radius in the above limit).
Continuing this line of argument we can should be able to squash the
K3 surface down to a 2-dimensional and then a 1-dimensional
object. Actually there are two natural 1-dimensional limits, $\Xi_1$
and $\Xi_2$ depending on how we decompose the moduli space with
respect to the lattice structure preserved by $O(4,20;{\Bbb Z})$. This is
tied to the fact that there are two even self-dual lattices in 16
dimensions. We now see that our four-dimensional theory can be thought
of as eleven-dimensional supergravity compactified on $\Xi_i\times
T^6$. We already knew however that it could also be thought of as the
heterotic string compactified on $T^6$.
The classical moduli space of
the $T^6$ in the eleven-dimensional picture embeds nicely into the
stringy moduli space of the $T^6$ in the heterotic picture and so it
is tempting to ``cancel'' the two $T^6$'s against each other and make
the bold assertion that the heterotic string in ten dimensions is
equivalent to eleven-dimensional supergravity compactified on
$\Xi_i$. Clearly the two choices of $\Xi_i$ should give the $E_8\times
E_8$ string and the $SO(32)$ string.
Whether this statement only makes sense in some delicate limit or
whether we can really directly analyze compactification on $\Xi_i$
remains unclear. In particular, we have not yet calculated what shape
the $\Xi_i$'s are. Clearly neither is a circle since we already know
this should lead to the type IIA string. Thus, if they exist, they
must be some more complicated 1-skeleton object. Clearly some degree
of complexity is required from them since they contain the information
about the gauge group of the heterotic string.
This prediction of some relation between eleven-dimensional
supergravity and heterotic strings arises in this section in a way
very similar to the way $Sl(2,{\Bbb Z})$ $S$-dulaity for the type IIB string arose in
section \ref{s:9d}. We see that often a full understanding of
the moduli space of a given theory can tell us interesting things
about the relationships between the associated higher-dimensional theories.
\section*{Note added}
Since this talk was presented further constructions, in some ways
similar to the orbifolds presented
here and in \cite{FHSV:N=2}, have been presented in
\cite{SS:pairs,SV:pairs,VW:pairs,HLS:pairs}. In general the $N<4$ case appears
to be more subtle than the version discussed above. Eleven-dimensional
supergravity compactified on manifolds with $G_2$ holonomy has
recently been discussed in \cite{PT:G2,Ach:G2}. The recent paper
\cite{Sch:d=9} has some overlap with section \ref{s:9d}.
\section*{Acknowledgements}
It is a pleasure to thank D.~Morrison for collaboration on some of the
work presented here. I also thank S.~Chaudhuri,
B.~Greene, D.~Joyce, W.~Lerche and R.~Plesser for useful conversations.
The work of the author is supported by a grant from the National
Science Foundation.
|
\section{Introduction}
One of the major open problems in theoretical physics is how to
construct a consistent theory of quantum gravity. This long-standing
issue has been approached in various different ways.
Up to now it is not clear which of those approaches, ranging as far
as from non-perturbative canonical quantum gravity
\pcite{Ashtekar} to string theory, is most adequate. Also it may be that
they all provide complementary, but legitimate viewpoints \pcite{Au}.
In all of these approaches some control over the classical theory
seems indispensable. In a path integral, e.g., the leading
contributions come from the local extremals of the action; or in the
canonical approach the space of observables is correlated directly to
the space of classical solutions modulo gauge transformations
(provided always that there are no anomalies). In addition to this the
relation between different approaches to quantum gravity might be
illuminated through considerations on the classical level. However, in
comparison to the vast infinity of possible solutions to the Einstein
equations only a negligible number of (exact) solutions is known and
the space of classical solutions (modulo gauge transformations) is
maybe even less seizable than the corresponding quantum theory.
The situation changes drastically, if one regards gravity models in
lower dimensions. Within the recent decade such models have attracted
considerable interest -- in three (cf., e.g., \pcite{Wit,Car}) as well
as in two spacetime dimensions (cf., e.g., \pcite{viele}). In the
present treatise we will restrict ourselves to two spacetime
dimensions with Minkowski signature. The claim of this and the
following papers is that in the case of Lagrangians describing
gravity without additional matter fields (a dilaton field is not
considered as matter in this context) there is complete control over
the space of classical solutions as well as on the space of quantum
states. In Parts II,III \pcite{II} of the present series we will provide
a classification of all global solutions (modulo gauge
transformations) of more or less all available gravitational models in
1+1 dimensions without matter couplings! We allow for all
possible topologies of the spacetime manifold; and indeed there are
models with solutions with basically arbitrarily complicated topologies.
In Part IV \pcite{IV}, on the other hand, we will construct all
quantum states of the considered two-dimensional models in a
Hamiltonian approach \`a la Dirac.
What can we hope to learn from this? Is not the situation in two
dimensions just too far away from Einstein gravity in four
dimensions? In part this is true, certainly, but the point is that
there {\em are} questions which are not sensitive to the dimensional
simplification. One example is the issue of time (cf., e.g.,
\pcite{Isham}), which arises in any diffeomorphism invariant theory.
Another such question could be the interplay between a path
integral, a canonical Hamiltonian approach, and the topology of
spacetime. Irrespective of the dimension of
spacetime a Hamiltonian treatment implies a restriction to topologies
of the form $\S \times \dR$, where $\S$ is some (generically
space-like) hypersurface. (In two dimensions obviously $\S$ is either
a circle or a real line). In this way one excludes, e.g., the
interesting topic of topology changes. In a path integral, on the
other hand, there is no restriction to particular topologies. Now, in
two-dimensional Yang-Mills (YM) theories people invented some cut and paste
technique so as to infer transition amplitudes of non-trivial
spacetime topologies from the knowledge of the Hilbert space
associated to cylindrical spacetimes \pcite{YMWit}. It is realistic to
hope that something similar will be possible in the two-dimensional
gravity theories. The result might provide a first idea of what may be
expected of the somewhat analogous issue in four-dimensional
canonical and path integral quantum gravity.
But there are also interesting technical issues that may be
investigated: One of these is the role which degenerate metrics play
in a canonical framework, another one is the construction of an
inner product in a Dirac approach to a quantum theory,
a third one
a comparison of quantum theories for Minkowskian and Euclidean
signatures of the (quantized) metric. All of
these issues, taken up in Part IV of this work, are of current
interest in 4D quantum gravity \pcite{Ashtekar2}.
\medskip
The purpose of the present paper is to set the stage for a thorough
investigation of 1+1 gravity on the classical and quantum level. In
Section 2 we introduce the models under study. They comprise all
generalized dilaton theories \pcite{Banks}, where, in its reformulated
version \pcite{Kunst}, we allow also for nontrivial torsion
\pcite{LNP,Mod}. In this way one captures theories such as
$R^2$-gravity with \pcite{KV} and without \pcite{R2} torsion or
spherically reduced 4D gravity \pcite{Hajetal}; but it should be
stressed that the class of considered theories is much more general.
Not included in the present treatise is axially reduced 4D gravity
\pcite{Nicolai}.
In Section 4 we further deal with dynamically
coupled Yang-Mills gauge fields. A coupling to fermion or scalar
matter fields (besides the dilaton), on the other hand, is mentioned
briefly only, but cf.\ also \pcite{CGHS,Zwie} for instance.
\hfill The breakthrough in the analysis of 2+1 gravity (with
Lagrangian\break
$\int d^3x \sqrt{|\det g|} \, R$) came about with its
identification as a Chern-Simons gauge theory of the 2+1
dimensional Poincare group \pcite{Wit}. Similarly, in 1+1
dimensions there are two models that could be identified with
standard gauge theories: The first of these is
the Jackiw-Teitelboim (JT) model of 1+1
deSitter gravity with Lagrangian \pcite{JT} \begin{equation}
L^{JT}=-\frac12\int_M d^2x\sqrt{-\det g} \; \varphi \; (R-\L)\,
, \el JT where $R$ is the Levi-Civita curvature scalar of the
metric $g$ and $\varphi$ is a Lagrange multiplier field
enforcing the field equation $R = \L \equiv const$. Rewriting
this action in an Einstein-Cartan formulation, it is found to
coincide with a YM gauge theory of the BF-type
\pcite{Isler} (where the gauge group is the universal covering
group of $SO(2,1)$ \pcite{versus}). Analogously, ordinary dilaton
theory \pcite{BH} may be reformulated as a BF-theory of the 1+1
dimensional Poincare group, if one promotes its cosmological
constant to a dynamical field which becomes constant on-shell
\pcite{Jackiw}. In all of these cases the respective group
structure greatly facilitated classical and especially quantum
considerations.
Within the class of theories considered in this paper the JT- and the
dilaton model are very particular (and comparatively simple). They
are, e.g., the only ones which, in a Hamiltonian formulation, allow
for global phase space coordinates such that the structure functions
in the constraint algebra become constants (cf.\ Part IV). Still some
of their features, such as a local degrees of freedom count, hold for
more complicated 2D gravity theories, too. In a way the situation reminds
one a bit of the Ashtekar formulation of 3+1 gravity with its similarities
but also its differences to a 4D YM theory. The question arises:
Given, say, spherically reduced gravity or the already
somewhat more complicated $R^2$-theory, or even a generalized
dilaton theory defined by a potential function $V(\cdot)$ or
$W(\cdot,\cdot)$ (cf.\ Sec.\ 2 below): Is
there some kind of gauge theory formulation for them, which is similarly
helpful in the determination of the space of quantum states or
the space of classical solutions modulo gauge transformations?
In other words: Is there some unifying approach to all of
these 2D gravity theories generalizing in a nonlinear way main features of
YM-type gauge theories?
Indeed, this question may be answered in the affirmative, as is
demonstrated at the beginning of Section 3. The key feature will be
the identification of Poisson brackets on an appropriate {\em target}
space associated with any generalized dilaton theory. The resulting
point of view not only allows for a unified treatment of
gravity-Yang-Mills systems in two dimensions, it also provides tools
for their classical and quantum analysis, which are hardly accessible
otherwise. This will be demonstrated first when employing the
formalism to solve the field equations in a particularly
efficient manner in the remainder of Section 3. There we will provide
the general local solution to the field equations of the general
model in the vicinity of arbitrary spacetime points. The extension of
this to the case of dynamically coupled YM-fields is
taken up in Section 4, finally. Section 4 includes also a brief
summary of the results of this first part as well as a short outlook
on Parts II--IV.
\section{Models of 1+1 Dimensional Gravity}
\plabel} \def\pl{\label{Models}
\hfill It is well-known that in two dimensions the Einstein-Hilbert term,\\
$\int \sqrt{|\det g|} R d^2x$, does not provide a useful action for field
equations, as it is a `boundary term'. However, a natural approach to
find a gravity action in two dimensions is to dimensionally reduce the
four-dimensional Einstein-Hilbert action. Implementing, e.g.,
spherical symmetry, by plugging
\begin{equation} (ds^2)_{(4)}=g_{\mu \nu}(x^\m) dx^\m dx^\n -\Phi^2 \,
(d \vartheta^2 + sin^2\vartheta d \varphi^2),\quad \m, \n \in \{0,1\}, \end{equation}
into the four-dimensional action and
integrating over the angle coordinates
$\vartheta$ and $\varphi$,
one obtains the two-dimensional action \pcite{Hajetal}
\begin{equation} L^{spher}= \int_M d^2x \sqrt{-\det g}
\, \, \left[{1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 4} \Phi^2 R(g) +
\2 g^{\m\n} \6_\m \Phi\6_\n \Phi - {1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 2}\right] . \el spher
Here $R(g)$ denotes the Ricci scalar of the
Levi-Civita connection of the (two-dimensional) metric $g$
and $\det g \equiv \det g_{\m \n}$.
Note that the field $\Phi(x^\m)$ is restricted to positive values by
definition.\footnote{This
could be avoided by introducing a new
field variable proportional to $\ln \Phi$.}
As a consistency check one may verify that
the resulting field equations of this effective two-dimensional
action provide solutions to the four-dimensional
Einstein equations, and these are nothing but the Schwarzschild
solutions, parametrized, according to
Birkhoffs theorem, by the Schwarzschild mass $m$:
\begin{equation} g=\left(1- {2m \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} r}\right)(dt)^2 -\left(1- {2m \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} r}\right)^{-1}
(dr)^2 \, .\el gspher
The currently most popular action for a two-dimensional gravity theory is,
however, the CGHS-model \pcite{CGHS}
\begin{eqnarray}
L^{CGHS}(g,\phi,f_i) & =& L^{dil}(g,\phi) + L^{Mat}(f_i,g) \, \, ,
\plabel} \def\pl{\label{CGHS} \\
L^{dil}(g,\phi)& =& \int_M d^2x \sqrt{-\det g} \, \exp (-2 \phi) \, \,
\left[R+4 g^{\m\n} \6_\m \phi\6_\n \phi -\L \right] \plabel} \def\pl{\label{Dil} \\
L^{Mat}(f_i,g) &=& \int_M d^2x \sqrt{-\det g} \,\sum_{i=1}^N \,
g^{\m\n} \6_\m f_i \6_\n f_i
\, \, . \plabel} \def\pl{\label{Mat}
\end{eqnarray}
The first part of this action, $L^{dil}$, is the so-called `string inspired'
or dilaton gravity action \pcite{BH} ($\phi$ corresponds to the dilaton
field in string theory),
the second part, $L^{Mat}$, is the standard kinetic term for $N$
scalar fields $f_i$. The vacuum solutions ($f_i \equiv 0$) of
(\ref }%\def\pr{(\pref{CGHS}) are of a similar form as (\ref }%\def\pr{(\pref{gspher}) (with identical
Penrose diagrams). Moreover, the classical
model can be solved completely also when the $f_i$ are present
\pcite{CGHS}. The CGHS-model thus opened
the possibility to discuss, e.g., the Hawking effect \pcite{Hawking}
in a simplified two-dimensional framework \pcite{BHreview}. Also, motivated by
the classical solvability of this model, one may hope for
an exact quantum treatment of \rz CGHS \pcite{Zwie}, allowing to test the
semiclassical considerations leading to the Hawking effect.
By means of the field redefinition $\Phi := 2 \sqrt{2}
\exp(- \phi), \, \Phi >0$,
the gravitational or dilaton part of \rz CGHS may be put into the form
\begin{equation} L^{dil}(g,\Phi) =
\int_M d^2x \sqrt{-\det g} \, \, \left[\mbox{${1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 8}$}
\Phi^2 R+ \mbox{$\2$} g^{\m\n} \6_\m \Phi\6_\n \Phi -
\mbox{${1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 8}$} \Phi^2 \L \right] \, . \el Dil2
Now \rz Dil and \rz spher are found to show much similarity.
Both of the examples given above are a special case of the general
action
\begin{equation} L^{gdil} (g, \Phi)= \int_M d^2x \sqrt{-\det g} \,\, \left[D(\Phi) R(g) +
\mbox{$\2$} g^{\m\n} \6_\m \Phi\6_\n \Phi - U(\Phi) \right]
\, . \el gDil
This action, which we will call generalized dilaton action,
was suggested first in \pcite{Banks}. It is the most general
diffeomorphism invariant action yielding second order differential
equations for the metric $g$ and a scalar dilaton field
$\Phi$. In the following we will
restrict ourselves to the case that $D$ has an inverse function
$D^{-1}$ everywhere on its
domain of definition. Also, for simplicity, we assume
that $D$, $D^{-1}$, and $U$ are $C^\infty$.
With these assumptions we may use\footnote{At this point the
chosen nomenclature might appear bizarre. In the
sequel, however, $X^3$ will turn out to serve as the third target
space coordinate of a useful $\sigma$-model formulation of (\ref }%\def\pr{(\pref{gDil}).}
\begin{equation} X^3 := D(\Phi) \el X3
as a new field variable instead of $\Phi$.
Introducing \pcite{Kunst} instead of $g$
\begin{equation} \widetilde g := \exp[ \rho(\Phi)] \, g \; \; , \;\;\; \rho =
\mbox{$\2$} \int^\Phi {du \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} dD(u)/du} + const \, \, ,
\el gtilde
moreover, the action \rz gDil takes the simplified form
\begin{equation} L^{gdil} (\widetilde g,X^3) = \int_M d^2x \sqrt{-\det \widetilde g} \,\,
[X^3 R(\widetilde g) -V(X^3)] \,\, .\el Ltilde
Here we have put
$V(z) := (U / \exp \rho)\left(D^{-1}(z)\right)$ and
different constants chosen for the definition of $\rho$ rescale only
the potential $V$.\footnote{Mainly we use only one symbol for a
function or functional, if it is represented in
different coordinates (cf., e.g., (\ref }%\def\pr{(\pref{Dil}) and (\ref }%\def\pr{(\pref{Dil2})).
To avoid misinterpretations we haven't done so in the case of $V$.}
Note that this field-dependent conformal transformation allowed to get rid
of the kinetic term for the dilaton field.
In the case of $L^{dil}$ or $L^{CGHS}$, respectively,
the transformation from $\phi$, $g$ to $X^3$, $\widetilde g$
proves specifically powerful \pcite{Ver}: With an appropriate
choice of $const$ in $\rho$ one obtains $\exp \rho(\Phi)=D(\Phi),
\, U(\Phi) =\L D(\Phi)$ and thus $V(X^3)=\L=const.$
Since, moreover, $L^{Mat}$ is
invariant under conformal transformations, the action \rz CGHS becomes
\begin{equation} L^{CGHS}(\widetilde g, X^3, f_i) = \int_M d^2x \sqrt{-\det \widetilde g} \,\,
[X^3 R(\widetilde g) -\L] + L^{Mat}(f_i, \widetilde g) \, \, .
\el CGHStilde
In this formulation the classical solvability of the CGHS-model is
most obvious: The variation with respect to $X^3$ yields $R(\widetilde
g)=0$. This implies that up to diffeomorphisms the metric $\widetilde g$
is Minkowskian. Thus the field equations resulting from the variation
with respect to the $f_i$ reduce to the ones of $N$ massless scalar fields
in Minkowski space. One then is left only to realize that due to the
diffeomorphism invariance only one of the three field equations $\delta
L^{CGHS}/\delta \widetilde g_{\m\n}(x) =0$ is independent \pcite{Sundermeyer}
and that this one may be solved always for the Lagrange multiplier field
$X^3$ locally. Still, the
representation of $L^{CGHS}$ in the form \rz CGHStilde
does not imply that the scalar fields
$f_i$ and the original metric $g$ decouple completely. Rather one should
compare it to the introduction of normal coordinates for coupled harmonic
oscillators. To trace the coupling explicitly, one notices that the
transition from $\widetilde g$ to $g$ involves $X^3$, which in turn is coupled
directly (via $\widetilde g$ in (\ref }%\def\pr{(\pref{CGHStilde})) to the scalar fields $f_i$.
Let us represent $L^{gdil}$ in first order form.
For this purpose we switch to the Cartan formulation
of a gravity theory, implementing the zero-torsion condition
by means of Lagrange multiplier fields $X^\pm$:
\begin{equation} L^{gdil} (e^a,\o,X^i) = -2 \int_M \,\,
X_a De^a + X^3 d\o + {V(X^3) \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 2} \, \varepsilon \,\,, \el PSX3
with
\begin{equation} \widetilde g = 2 e^-e^+ \equiv e^- \otimes e^+ + e^+ \otimes e^-
\, \, \el tildevielbein
and
\begin{equation} De^a \equiv de^a + \varepsilon^a{}_b \o \wedge e^b \, , \quad
a \in \{-,+\} \, , \qquad
\varepsilon \equiv e^- \wedge e^+ \, \, ,
\el torsion
such that $e^\pm$ is the zweibein in a light cone basis of the
frame bundle, $\o$ (or $\o^a{}_b \equiv
\varepsilon^a{}_b \o$) is the Lorentz or spin connection,
and $\varepsilon_{-+}= +1$. Here we have used $\varepsilon R=-2 d \o$.
We derived \rz PSX3 from the general action \rz gDil (for the case
that \rz X3 is a diffeomorphism). In this way the zweibein and spin
connection are interpreted as quantities corresponding to the
auxiliary metric $\widetilde g$, which in turn is related to the `true'
metric via (\ref }%\def\pr{(\pref{gtilde}). In the following we will argue that (\ref }%\def\pr{(\pref{PSX3}),
or its generalization
\begin{eqnarray} L^{grav} &=& \int_M \,\,
X_a De^a + X^3 d\o + W((X)^2,X^3) \, \varepsilon \,\, , \plabel} \def\pl{\label{grav}
\\
(X)^2 &\equiv& X_aX^a \equiv 2X^-X^+ \, , \end{eqnarray}
may be regarded also as a gravity theory with metric
\begin{equation} g = 2 e^-e^+ \, \, .\el vielbein
First we note that $L^{grav}$ is invariant with respect to the
standard gravity symmetries, which are diffeomorphisms and local frame
rotations. Second, more or less by construction the action \rz grav
is in first order form. It is not difficult to see then that the $X^i, \, i
= +,-,3$, are precisely the generalized momenta canonically conjugate to
the one-components
of the zweibein and the spin connection (the corresponding
zero-components serve as
Lagrange multipliers for the constraints of the theory, cf.\ also
Part IV). Obviously
there is a need
for momenta in any first order formulation of a gravity theory, so
the $X^i$ appear very natural from this point of view.
Last but not least, \rz grav may be seen to yield further already
accepted models of 2D gravity for some specific choices of the
potential $W$. Let us choose, e.g., $2W[(X)^2,X^3]=V(X^3)=
{(X^3)}^\g \L$ where $\L \neq 0$ and
$\g$ are some real constants. For $\g =1$ we immediately recognize the good
old Jackiw-Teitelboim model of two-dimensional deSitter gravity \pcite{JT}.
For $\g \neq 1$, on the other hand, we may eliminate the field $X^3$ by means
of its equation of motion. Implementing, furthermore, the zero-torsion
constraint by hand again, the resulting Lagrangian is found to be of the form
$L \propto \int \sqrt{- \det g} R^{\g/(\g -1)}$. To obtain an integer exponent
$n$ for $R$, one sets $\g = n/(n-1)$. These are the purely geometrical
Lagrangians for higher derivative gravity proposed in \pcite{R2}.
So, e.g., the potential $W^{R^2}:= -(X^3)^2 +\L$ yields the Lagrangian
\begin{equation} L^{R^2}= \int_M d^2x \sqrt{-\det g} \, (R^2/16 + \L) \, \el R2
of two-dimensional $R^2$-gravity.
Similarly, the potential
\begin{equation} W^{KV}= - \a (X)^2/2 - (X^3)^2 +\L/\a^2 \el WKV
leads, upon elimination of the $X$-coordinates, to
\begin{equation} L^{KV} = \int [-{1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 4} d\o \wedge \ast d\o - {1\over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 2\a} De^a \wedge \ast De_a
+ {\L \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a^2} \e] \, , \plabel} \def\pl{\label{KV} \end{equation}
proposed first in \pcite{KV}.
This Lagrangian is the most general (diffeomorphism and frame
invariant) Lagrangian yielding second
order differential equations for
zweibein and spin connection. It is noteworthy that, in contrast to
four dimensions \pcite{vanNieu}, it contains only three terms.
Here one allowed for nontrivial torsion.
All torsion-free theories described by \rz grav have a potential
$W$ that is independent of
$(X)^2$, or, equivalently, by Lagrangians of the form (\ref }%\def\pr{(\pref{gDil}) (with
$\widetilde g \rightarrow g$).
Before we close this section, let us return to the case of spherical
symmetry (\ref }%\def\pr{(\pref{spher}). An appropriate choice of
the integration constant in (\ref }%\def\pr{(\pref{gtilde}) yields $g = \widetilde g /\sqrt{X^3}$
and the potential $W$ becomes $W=V/2=1/4\sqrt{X^3}$ in this case. Thus, the
space of solutions to \rz spher will be reproduced from \rz grav with this
potential, if, according to \ry tildevielbein , $g := 2e^-e^+/\sqrt{X^3}$.
Alternatively, as we will find in the following section, the positive mass
solutions (\ref }%\def\pr{(\pref{gspher}) may be described also by (\ref }%\def\pr{(\pref{grav}) with potential
$W=1/(X^3)^2$, {\em if\/} we use the simpler identification (\ref }%\def\pr{(\pref{vielbein}),
$g:= 2e^-e^+$.
In this section we have shown that \rz grav is a universal action
for gravity theories in two dimensions. In the following section we will
find it to be a special case of a $\sigma$-model defined by a Poisson
structure on a target space, an observation that allows to solve
the theory in an elegant and efficient manner.
\section{The Local Solutions of the Field Equations}
\plabel} \def\pl{\label{local}
With the notational convention \begin{equation} A_- \equiv e_- \equiv e^+ \, ,
\,\,\, A_+ \equiv e_+ \equiv e^- \, , \,\,\, A_3 \equiv \o \, \el
identi we can rewrite the action \rz grav up to a boundary term as \begin{equation}
L= \int_M A_i \wedge dX^i + \2 {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}(X(x)) A_i \wedge A_j \, \el PS
with
\begin{equation} \left({\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}\right)(X) = \left( \begin{array}{lll}
0 & -W & -X^- \\
W & \phantom{-}0 & \phantom{-}X^+\\
X^- & -X^+ & \phantom{-}0
\end{array} \right) \,\quad \, i,j
\in \{ -,+,3 \} \, , \el P where, as before, $W$ is a function of
$(X)^2\equiv 2X^-X^+$ and $X^3$. The first decisive observation is
that in this form the action is not only covariant with respect to
diffeomorphisms on the spacetime manifold $M$, but also with respect
to diffeomorphisms on the space of values of the fields $X^i$, i.e.\
on a `target space' $N=\dR^3$; we only have to define the
transformation of the $A_i$ and of ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}$ as those of one-forms
and bivectors on $N$, respectively. (The term $A_i \wedge dX^i$ in \rz
PS may then be interpreted as the pullback of a one-one-form $A=A_{\m
i}dx^{\m} \wedge dX^i$ on $M \times N$ under the map of $M$ into the
space of fields and likewise the second term in \rz PS as the pullback
of the twofold contraction of $A$ with the two-tensor ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}= (1/2)
{\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij} \partial_i \wedge \partial_j$ on $N$, cf.\
\pcite{proce,Brief}). The second decisive observation is that the
matrix ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$ obeys the following identity\footnote{The study of
actions of the form \rz PS where ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}$ satisfies \rz Jacobi has
been proposed also in \pcite{Ikeda}. However, the implications of
the identity \ry Jacobi , recapitulated in what follows, have been
realized only in \pcite{LNP,Mod,proce,Brief}.}
\begin{equation} { \partial {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij} \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \partial X^l} {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{lk} + cycl.(ijk) =0 \,\,.
\el Jacobi
It establishes that ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$
is a Poisson structure on $N=\dR^3$. To see this one defines
\begin{equation} \{F,G\}_N = {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}(X) { \6 F(X) \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 X^i }
{ \6 G(X) \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 X^j }
\, \el Poi
for any two functions $F$ and $G$ on $N$; the Jacobi identity for the
Poisson brackets $\{ \cdot , \cdot \}_N$ on $N$ is then found to be
equivalent to (\ref }%\def\pr{(\pref{Jacobi}).
Vice versa, the Leibniz rule and antisymmetry of
Poisson brackets ensures that they
can be written in the form \rz Poi with a skew-symmetric bivector ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$
on $N$. Thus we see that (\ref }%\def\pr{(\pref{PS}), and therefore also (\ref }%\def\pr{(\pref{grav}), may
be interpreted as a $\sigma$-model, where the world sheet $M$ is
the spacetime manifold and the target space $N$, which in the present case
equals $\dR^3$, is a Poisson manifold \pcite{Mad,Buchneu}.
Note, however, that in our case the Poisson tensor \rz P is
degenerate necessarily,
as it is skew-symmetric and $N$ is three-dimensional here. Obviously at
points of $N$ where
\begin{equation} X^-=X^+=W=0 \,
\el crit
${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$ has rank zero -- we will call these points `critical' further on
--, everywhere else it has rank two.
Furthermore (as a consequence of the
Jacobi identity for ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$) in the neighbourhood of generic (i.e.\
non-critical) points there exists a foliation
of $N$ into two-dimensional submanifolds $\CS$, which are integral
manifolds of the set of Hamiltonian vectorfields (we will label them with
the coordinate function $\widetilde X^{1}$).
Clearly, these leaves $\CS$ are symplectic (the restriction of ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$ onto
them is nondegenerate),
and one may choose coordinates $\widetilde X^2,\widetilde X^3$ such that on each of the
leaves ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}|_{T^\ast \CS}$ (or its inverse) is in Darboux-form. In such an
adapted coordinate system $\widetilde X^i$, which we will call
Casimir-Darboux (CD) coordinate system\footnote{In some textbooks (cf., e.g.,
\cite{Buchneu}) such coordinates are called
simply `Darboux coordinates'; however, we prefer the more suggestive term
above.}
${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C} \in \L^2(TN)$ takes the simple form
${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}= {\6 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 \widetilde X^2 } \wedge {\6 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 \widetilde X^3 }$.
The notation of \rz PS allows to derive and depict
the gravity field equations in a concise manner:
\begin{eqnarray} dX^i + {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij} A_j&=&0 \plabel} \def\pl{\label{eom1a}\\
dA_i + \2 {\6 {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{lm} \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 X^i} A_l \wedge A_m &=&0 \, . \plabel} \def\pl{\label{eom1b} \end{eqnarray}
But what is more important, in order to solve these equations of
motion, the considerations above suggest the
use of CD coordinates on the {\em target space} $N$.
As $L$ is written in an $N$-covariant manner, the field
equations will still have the form (\ref }%\def\pr{(\pref{eom1a}, \ref{eom1b}),
only now ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$ is in Casimir-Darboux form. Explicitly this reads
\begin{eqnarray} d \widetilde X^1 =0 \; , && d A_{\widetilde 1}=0 \plabel} \def\pl{\label{eom2a} \\
A_{\widetilde 2} = d \widetilde X^3 \; , && A_{\widetilde 3}
= -d \widetilde X^2 \; , \plabel} \def\pl{\label{eom2b} \end{eqnarray}
while the remaining two field equations $d A_{\widetilde 2} =d A_{\widetilde 3}=0$
are redundant obviously. In this form the solution of the
field equations becomes a triviality:
Locally (\ref }%\def\pr{(\pref{eom2a}) is equivalent to
\begin{equation} \widetilde X^1= \mbox{const ,} \quad A_{\widetilde 1}= df \el Loesung
where $f$ is some arbitrary function on $M$, while (\ref }%\def\pr{(\pref{eom2b}) determines
$A_{\widetilde 2}$ and $A_{\widetilde 3}$ in terms of the otherwise unrestricted
functions $\widetilde X^2$, $\widetilde X^3$.
Now we have to transform this solution back to the gravity variables
\rz identi only. Let us do this for the torsion-free case $W=V(X^3)/2$
first, Eqs.\ (\ref }%\def\pr{(\pref{PSX3}, \ref{Ltilde},\ref{gDil}).
Here
\begin{equation} \widetilde X^i := \left({\2}\Big[(X)^2\!-\!\!\int\limits^{\;\;\;X^3}\!\! V(z)
dz\Big] \;,\; \ln|X^+| \;,\; X^3\right) \el Xtildeneu
forms a CD coordinate system on $N$ on patches with $X^+ \neq 0$.
To verify this we merely have to check $\{\widetilde X^1, \cdot \}_N = 0$ and
$\{\widetilde X^2,\widetilde X^3 \}_N=1$, using the definition (\ref }%\def\pr{(\pref{P},\ref{Poi}) of
the brackets.
{}From $A_i= {\6 \widetilde X^j \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \6 X^i} A_{\widetilde j}$
we then infer
\begin{equation} e^+ \equiv A_- = X^+ A_{\widetilde 1} \, , \quad e^- \equiv A_+ = {1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4}
X^+ } A_{\widetilde 2} + X^- A_{\widetilde 1} ,. \el relationneu
By means of (\ref }%\def\pr{(\pref{eom2b}, \ref{Loesung}) the metric $g=2e^+e^-$ thus
becomes:
\begin{equation} g=2 dX^3 df + (X)^2 \, df df \, ,
\el gX3neu
where $(X)^2 = \int^{X^3} \, V(z) dz + const=: h(X^3)$ according to
(\ref }%\def\pr{(\pref{Xtildeneu}) and (\ref }%\def\pr{(\pref{Loesung}). Using $X^3$ and $f$ as coordinates
$x^0$ and $x^1$ on $M$, this may be rewritten as \begin{equation} g = 2
dx^0dx^1 + h(x^0) dx^1 dx^1 \el gh
with the function $h$ as defined above. In the case of $R^2$-gravity
(\ref }%\def\pr{(\pref{R2}), e.g., $h^{R^2}= -\mbox{${2\over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 3}$} (x^0)^3 + 2 \L x^0 +C$,
where $C$ denotes the integration constant. In (\ref }%\def\pr{(\pref{gh}) $h$ depends
on one integration constant only, which is a specific
function of the total mass (at least in cases where the latter may be
defined in a sensible way), reobtaining what
has been called `generalized Birkhoff theorem' in various special
cases (cf.\ \pcite{R2,Katanaev}).
Maybe at this point it is worth mentioning that the coordinate
transformation \begin{equation} r := x^0 \, , \;\; t := x^1 + \int^{x^0} {dz \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4}
h(z) } \, , \el rtKoord well-defined wherever $h \neq 0$, brings the
generalized Eddington-Finkelstein form of the metric, Eq.\ (\ref }%\def\pr{(\pref{gh}),
into the `Schwarzschild form' \begin{equation} g= h(r)(dt)^2 -{1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} h(r)} (dr)^2 \,
, \el gSSneu with the same function $h$. This confirms also that the
Schwarzschild case (with positive $m$) may be described by (\ref }%\def\pr{(\pref{grav})
also with the identification $g=2e^+e^-$: The potential $W=V/2 =
1/(X^3)^2$ yields $h(r)=C-1/r$, $C=const$, from which one finds
$m=C^{-(3/2)}$, after rescaling coordinates according to $r \to
\sqrt{C} \, r$ and $t \to t/\sqrt{C}$ ($C>0$). In fact, we learn that
given {\em any\/} metric $g$ in (effectively) $1+1$ dimensions with (at
least) one Killing field, \rz grav with $g=2e^+e^-$ and $2W := V(X^3)
= h'(X^3)$ will provide an action which has $g$ within its space of
solutions.\footnote{This holds because any metric with a Killing field
$v$ may be brought into the form (\ref }%\def\pr{(\pref{gh}) locally. Coordinate
independently $h$ may be characterized as the norm squared of $v$ as
a function of an affine parameter along a null-line, furthermore.
For more details cf.\ Part II.} We remark, finally, that in the
present case of a torsion-free connection the Ricci scalar is just:
\begin{equation} R = h''(x^0)\,. \el R
For a reformulated dilaton theory (\ref }%\def\pr{(\pref{gDil}, \ref{Ltilde}) Eq.\ \rz
gh gives $\widetilde g$ only, which we will write as $\widetilde g=2d\widetilde x^0 dx^1 +
\widetilde h(\widetilde x^0)(dx^1)^2$ with $\widetilde h (\widetilde x^0) \equiv \int^{\widetilde x^0} \,
V(z) dz + const$. {}From Eqs.\ (\ref }%\def\pr{(\pref{X3},\ref{gtilde}) we have
$g=\exp[-\r\left(D^{-1}(\widetilde x^0)\right)]\widetilde g$. Here the coordinate
transformation
\begin{equation} x^0(\widetilde x^0):=\int^{\widetilde x^0} \exp[-\r\left(D^{-1}(x)\right)] dx \end{equation}
brings $g$ into the form (\ref }%\def\pr{(\pref{gh}) again, where now
$h(x^0)=e^{-\r(\widetilde x^0(x^0))}\widetilde h(\widetilde x^0(x^0))$. Let us illustrate
this by means of the dilaton theory (\ref }%\def\pr{(\pref{Dil2}).
There $\widetilde h^{dil}(\widetilde x^0) = \L \widetilde x^0 + C$, $C$ denoting the
integration constant, and
$\widetilde x^0 = X^3 = D(\Phi)={1\over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 8}\Phi^2=\exp(-2\phi) \in \dR^+$.
As $\r=\ln \widetilde x^0$, Eq.\ \rz gtilde becomes $g= \widetilde g / \widetilde x^0$.
The above coordinate transformation $x^0 = \ln\widetilde x^0$
yields
\begin{equation} h^{dil}(x^0) = \L + C \exp(-x^0) \el hDil
with $x^0 \in \dR$.
For (\ref }%\def\pr{(\pref{spher}) the analogous procedure yields
$h^{SS}(x^0) = 1 + 2C / x^0$, leading to the identification
$m=-C$ in this case (cf.\ Eqs.\ (\ref }%\def\pr{(\pref{gSSneu}) and (\ref }%\def\pr{(\pref{gspher})).
The transition from $\widetilde g$ to $g$, although
conformal, may have important implications on the global structure of
the resulting theory, namely if due to a divergent conformal factor
the domain of $g$ is only part of the maximally extended domain of $\widetilde g$.
For instance, in the dilaton theory the maximal extension of $\widetilde g$ is
Minkowski space with its diamond-like Penrose-diagram, whereas the Penrose
diagram of $g$, found by studying the universal coverings of the local charts
obtained above (cf., e.g., Part II), is of
Schwarzschild-type. Although $X^3 = \widetilde x^0$ was defined for positive
values only, $\widetilde g$
remains well-behaved for $\widetilde x^0 \in \dR$
and may be extended to that values. The conformal factor in the
relation between $g$ and $\widetilde g$, $1/\widetilde x^0$, on the other hand,
blows up at $\widetilde x^0 =0$. Correspondingly $R(g)$ is seen to
diverge at $\widetilde x^0 \equiv \exp x^0 = 0$, cf.\ Eqs.\
(\ref }%\def\pr{(\pref{R},\ref{hDil})
and $g$ cannot be extended in the same way as $\widetilde g$. As a result
the shapes of the Penrose diagrams are different.
Such a behavior is generic also in the
case that \rz X3 maps $\Phi \in \dR$ to a part of $\dR$ only, say,
e.g., to $(a,\infty)$ with an increasing $D$. The conformal
exponent $\r$ \rz gtilde will diverge at $\widetilde x^0=a$, as
$D$ was required to be a diffeomorphism, and for potentials $U$ that do
not diverge too rapidly for $\phi \to - \infty$ the Penrose diagrams
of $\widetilde g$ and $g$ differ.
\medskip
Let us now discuss the solution to the field equations for a general
potential $W$ in (\ref }%\def\pr{(\pref{grav}),
using this opportunity to present also a more
systematic construction of a CD-coordinate system.
By definition a Casimir coordinate $\widetilde X^1$ is characterized
by the equation ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}(d \widetilde X^1, \cdot)=0$ $\Leftrightarrow
(\6 \widetilde X^1 / \6 X^i) {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij}=0$.
For $j=3$ the latter implies that $\widetilde X^1$ has to be a Lorentz
invariant function of $X^\pm$, i.e.
\begin{equation} \widetilde X^1=\2 C\left[(X)^2,X^3\right] \el tildeX1
for some two-argument function $C=C(u,v)$.
For $j=\pm$ we then obtain
\begin{equation} 2 W(u,v) C,_u + C,_v =0 \, \el CDiff
where the comma denotes differentiation with respect to the corresponding
argument of $C$. As \rz CDiff is a first order differential equation, it
may be solved for any given potential $W$ locally, illustrating
the general feature of a local foliation of Poisson manifolds for the special
case \ry P . An important consequence of \rz CDiff is the relation $C,_u
\neq 0$. This follows as on the target space (!) we have $dC \neq 0$
(by definition of a target space coordinate function), and, according to
Eq.\ \ry CDiff , $C,_u =0$ at some point implies that there
also $C,_v=0$ and thus $dC=0$. Certainly, as $C,_u$ is a function on $N$,
in contrast to $dC$ it remains non-zero also upon restriction to the
submanifolds $C=const$. We may verify this also explicitly
at the torsion-free example: (\ref }%\def\pr{(\pref{Xtildeneu}) yields $C,_u(u,v) \equiv 1$.
Using the method of characteristics, \rz CDiff may be reduced to an ordinary
first order differential equation: We may express the lines of constant
values of the function $C$ in the form
\begin{equation} {du \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} dv}=2W(u,v) \, .\el Diff
The constant of integration of this equation is a
function of $C$ in general; however, as clearly any function of a Casimir is
a Casimir again, we may just identify the integration constant with $C$.
To illustrate this, we choose
\begin{equation} 2W(u,v) := V(v) + T(v) u \, \, . \el T
We then obtain from \rz Diff
\begin{equation} u = \left[ \int^v V(z) \exp \left(-\int^z T(y) dy \right) dz +
const(C) \right]
\, \exp \left(\int^v T(x) dx \right) \, , \el u
where the lower boundaries in the integrations over $T$ coincide.
Upon the choice $const(C) :=C$ \rz u gives
\begin{equation} C (u,v) = u \, \exp \left(-\int^v T(x) dx \right) -
\int^v V(z) \exp \left(-\int^z T(y) dy \right) dz \, . \el C
The integrations on the right-hand side should be understood as definite
integrals with somehow fixed, $C$-independent lower boundaries. Different
choices for these boundaries rescale $C$ linearly.
Let us specialize \rz C to some cases of particular interest.
In the case $T \equiv 0$, describing torsionless gravity and
discussed already above, it gives
\begin{equation} C= (X)^2 - \int^{X^3} V(z) dz \, , \el CV
in coincidence with the first entry of (\ref }%\def\pr{(\pref{Xtildeneu}).
For the Katanaev-Volovich (KV) model of 2D gravity with torsion
(\ref }%\def\pr{(\pref{KV}), as a second example, \rz C and \rz WKV
yield upon appropriate choices for the constants of integration
and a rescaling by $\a^3$
\begin{equation} C^{KV}= -2 \a^2 \exp (\a X^3) \left( W^{KV} +{2 X^3 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a} -
{2 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a^2} \right) \, . \el CKV
Here $C,_u (u,v) = \a^3 \exp (\a v)$.
The remaining task is to find Darboux coordinates. On patches with either
$X^+ \neq 0$ or $X^- \neq 0$ this is a triviality almost: According to
the defining relations (\ref }%\def\pr{(\pref{P}) of the Poisson brackets we have
$\{X^\pm,X^3\}_N=\pm X^\pm$, so obviously $\pm \ln |X^\pm|$ and
$X^3$ are conjugates. Altogether therefore
\begin{equation} \widetilde X^i := (\mbox{$\2$}C,\pm \ln|X^\pm|,X^3) \, \el Xtilde
forms a CD coordinate system on regions of $N$ with $X^\pm \neq 0$,
respectively.
So, now we just have to repeat the steps following Eq.\ (\ref }%\def\pr{(\pref{Xtildeneu}) in
the more general setting of a potential $W(u,v)$. However, as we do
not want to restrict ourselves to potentials e.g.\ purely linear in
$u$, Eq.\ (\ref }%\def\pr{(\pref{T}), we do not know $C$ in explicit form. Still it is
nice to find that also in the case of a completely general Lagrangian
(\ref }%\def\pr{(\pref{grav}) the metric takes the form (\ref }%\def\pr{(\pref{gh}) locally. Moreover,
$h$ may be determined in terms of the Casimir function $C(u,v)$ and
the Killing field $\6/\6_1$ will be shown to be a symmetry direction
of all of the solution\footnote{This is not a triviality as, e.g., for
nonvanishing torsion the connection $\o$ is not determined (up to
Lorentz transformations) by the metric $g$ already.}. For the sake
of brevity we display the calculation for both of the sets
(\ref }%\def\pr{(\pref{Xtilde}) of CD-coordinates simultaneously. This means that in the
following we restrict our attention to local solutions on $M$ that map
into regions of $N$ with $X^+ \neq 0$ {\em or} $X^- \neq 0$.
Inserting into the analogue of Eqs.\ (\ref }%\def\pr{(\pref{relationneu}) the solutions
(\ref }%\def\pr{(\pref{eom2b}, \ref{Loesung}) and re-expressing thereby the $\widetilde X^i$ in
terms of the original fields $X^i$ again, (\ref }%\def\pr{(\pref{Xtilde}), we obtain
(the upper/lower signs being valid for the charts $X^+$ resp.\ $X^-\neq 0$)
\begin{eqnarray}
e^\pm \equiv A_\mp &=& C,_u X^\pm df \nonumber \\
e^\mp \equiv A_\pm &=& \pm {1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} X^\pm } dX^3 + C,_u X^\mp df \nonumber \\
\o \,\,\equiv A_3 \, &=& \mp d\ln |X^\pm| - C,_u W df \, .
\label{relation}
\end{eqnarray}
where in the last line we used (\ref }%\def\pr{(\pref{CDiff}).
For the metric $g=2e^+e^-$
this yields
\begin{equation}
g=\pm 2C,_u \, dX^3 df + \bar h(X^3,C) \, df df \,,
\label{gX3}
\end{equation}
with \begin{equation} \bar h(X^3,C) := C,_u{}^2 \cdot (X)^2 \; .\el hbar
\hfill Here $(X)^2$ is a function of $X^3$ and $C$ by inverting
the field equation\\
$C((X)^2,X^3)=const$, whereas $C,_u$ in \rz hbar
and \rz gX3 is, more explicitly, $C,_u \left((X)^2(X^3,C),X^3\right)$.
Note that according to the context $C$
either denotes a function of $X^i$ or the constant which it equals due to
the first equation (\ref }%\def\pr{(\pref{eom2a}).
Now again we want to fix a gauge. {}From \rz gX3 we learn that
$C,_u dX^3 \wedge df \neq 0$, because otherwise $\det g = 0$. This
implies that we may choose $C,_u dX^3$ and $df$ as coordinate
differentials on $M$. Let us therefore fix the diffeomorphism invariance
of the underlying gravity theory by setting
\begin{eqnarray} \int^{X^3} C,_u[(X)^2(z,C),z] dz &:=& x^0 \, , \plabel} \def\pl{\label{x0} \\
f &:=& \pm x^1 \, .\plabel} \def\pl{\label{x1} \end{eqnarray} In this gauge $g$ is seen to take the form
(\ref }%\def\pr{(\pref{gh}) again with
\begin{equation} h(x^0) = \bar
h(X^3(x^0),C) \, , \el h where $X^3(x^0)$ denotes the inverse of
(\ref }%\def\pr{(\pref{x0}).
The local Lorentz invariance may be fixed by means of
\begin{equation} X^\pm := \pm 1 \, ,\el Lorentz
finally. Besides (\ref }%\def\pr{(\pref{Lorentz}, \ref{x0})
the complete set of fields then takes the form
\begin{equation}
e^\pm = C,_u dx^1 \, , \quad e^\mp = {dx^0 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} C,_u} + \mbox{$\2$} (X)^2 \,
C,_u dx^1 \, , \quad \o = \mp W \, C,_u dx^1 \, ,
\label{inKarte}
\end{equation}
and $X^\mp= \pm \mbox{$\2$} (X)^2$.
Here again $(X)^2$, $C,_u$, and $W$
depend on $X^3(x^0)$ and $C=const$ only, $C$ being the only
integration constant left in the local solutions.
In the torsion-free case, considered already above,
$C,_u=1$, $X^3(x^0)=x^0$, and, cf.\ Eqs.\ (\ref }%\def\pr{(\pref{hbar}, \ref{h}, \ref{CV}),
$(X)^2=h(x^0)=\int^{x^0}V(z)dz +C$, thus reproducing the results
obtained there. For the KV-model (\ref }%\def\pr{(\pref{KV}), as an example for a theory
with torsion, on the other hand,
the above formulas yield
$x^0 = \a^2 \exp (\a X^3)$, $x^0 \in \dR^+$, and
\begin{equation} h^{KV}(x^0) = {1 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a} \left\{C x^0 - 2 (x^0)^2 \,
\left[(\ln x^0-1)^2+1-
\Lambda\right] \right\} \, \el hKV
for instance. Certainly \rz R does not hold any more, instead we find
$R = -{4 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a} \ln \left({x^0 \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} \a^2} \right)$. Here this is obtained most
easily by concluding $R = -4X^3$ from (the Hodge dual of) the three-component
of (\ref }%\def\pr{(\pref{eom1b}).
In the above we captured the solutions within regions of $M$ where
either $X^+ \neq 0$ or $X^-\neq 0$. Clearly, in regions where $(X)^2
=2X^+X^-\neq 0$ the two charts (\ref }%\def\pr{(\pref{inKarte}) must be related to each
other by a gauge transformation, i.e., up to a Lorentz transformation,
by a diffeomorphism. However, one of these two charts extends smoothly
into regions with only $X^+\neq 0$ (but possibly with zeros of $X^-$),
and $(+\leftrightarrow -)$ for the other chart.\footnote{In Part II
regions with $(X)^2\neq 0$ will be called `sectors', while patches
with merely $X^+\neq 0$ {\em or} $X^-\neq 0$, generically containing
several sectors, will be our `building blocks' for the global extension.}
In this way the above
mentioned diffeomorphism may serve as a gluing diffeomorphism,
allowing to extend the generically just local solution (\ref }%\def\pr{(\pref{gh})
to one that applies wherever $X^+$ and
$X^-$ do not vanish simultaneously.
Let us remark here, furthermore, that the two representatives
(\ref }%\def\pr{(\pref{inKarte}) are mapped into each other by \begin{equation} e^+
\longleftrightarrow e^- \; ,\quad \omega \longleftrightarrow -\omega
\; ,\quad X^+ \longleftrightarrow -X^- \; ,\quad X^3
\longleftrightarrow X^3 \, .\label{5} \el Strafo This transformation
reverses the sign of the action integral (\ref }%\def\pr{(\pref{grav}) only and therefore
does not affect the equations of motion. {}From (\ref }%\def\pr{(\pref{5}) it is obvious
that the gluing diffeomorphism (cf.\ above) maps one set of null
lines onto the respective other one, leaving the form (\ref }%\def\pr{(\pref{gh}) of
$g=2e^+e^-$ unchanged. It corresponds to a discrete symmetry of
(\ref }%\def\pr{(\pref{gh}) (called `flip' in Part II), which is independent from the
continuous one generated by $\6/\6_1$.
Further details shall be provided in Part II.
We are left with finding the local shape of the solutions in the
vicinity of points on $M$ that map to $X^+=X^-=0$. Here we have to
distinguish between two qualitatively different cases: First,
$C=C_{crit}\equiv C(0,X^3_{crit})$, where $X^3_{crit}$ has been chosen
to denote the zeros of $W(0,X^3)$, and second, $C \neq C_{crit}$.
The special role of the critical values of $C$ is due to the fact that
the Poisson structure ${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}$ vanishes precisely at the points $X^+=X^-=0$
on the two-surfaces $C=C_{crit}$, cf.\ Eq.\ \ry crit .
We start treating the non-critical case $C \neq C_{crit}$: For this,
one could attempt to construct a CD coordinate system valid in a
neighborhood of a (non-critical) point $X^+=X^-=0$. At least in the
torsionless case this may be done in an explicit way, but while e.g.\
for ordinary dilaton gravity ($W\equiv const$) Darboux coordinates
are provided already by rescaling merely $X^+$ and $X^-$ (since
$\{X^+,X^-\}_N=W$), in the more general case the formulas are somewhat
cumbersome. We therefore follow a somewhat less systematic, but
simpler route here: {}From the three-component of \rz eom1a it is
straightforward to infer that points with $X^+=X^-=0$ are saddle
points of $X^3$. This suggests to replace the gauge conditions
(\ref }%\def\pr{(\pref{x0},\ref{x1}) by an ansatz of the form \begin{eqnarray} \int^{X^3}
C,_u[(X)^2(z,C),z] dz &:=& x y + a \, , \plabel} \def\pl{\label{x0neu}
\\ f &:=& b \ln x \, ,\plabel} \def\pl{\label{x1neu} \end{eqnarray} where $a$ and $b$ are constants
to be determined below. As a first justification of
(\ref }%\def\pr{(\pref{x0neu},\ref{x1neu}) we find $C,_u dX^3 \wedge df$ to be finite and
non-vanishing on $(x,y) \in \dR^2$. Implementing the above conditions
in \ry gX3 , the metric becomes \begin{equation} g=2b \, dx dy + b {2xy + b \, h(xy
+ a) \over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} x^2 } \, (dx)^2 \, , \el saddle where $h$ is the function
defined in \ry h . For generic values of $a$ and $b$ \rz saddle is
singular at $x=0$. However, the choice \begin{equation} a := h^{-1}(0) \qquad b:= -
2/h'\left(h^{-1}(0)\right) \el ab is seen readily to yield a smooth
$g$!
The singularity of the gauge choice \rz x1neu at $x=0$ was devised
such that it compensated precisely the singularity of the CD
coordinate system at $X^\pm=0$, used to derive \ry gX3 .
Indeed, the lines of vanishing $x$ or $y$
in \rz saddle may be seen to correspond to lines of
vanishing $X^+$ and $X^-$, respectively. Also, they are Killing
horizons: According to (\ref }%\def\pr{(\pref{hbar}) zeros of $(X)^2$ coincide
precisely with the zeros of $h(x^0)$, indicating that the
Killing-field $\6/\6x^1$ (in charts (\ref }%\def\pr{(\pref{gh})) becomes null on those lines
(cf.\ also \ry gSSneu ).
The charts \rz saddle provide a simple alternative to a generalized
Kruskal extension (cf.\ Part II). For Schwarzschild, $h(r)=1-2m/r$, it
is a global chart (as $h^{-1}(0)$ is single-valued), in the more
complicated Reissner-Nordstr\oe m case $h(r)=1-2m/r+q^2/r^2$ the
constant $a$ may take one of the two values $r_{\pm}=m \pm
\sqrt{m^2-q^2}$ and \rz saddle provides a local chart in the vicinity
of the respective value of $r=x^0$. A generalization of the right-hand
sides of \rz x0neu and \rz x1neu to $F(x) y + G(x)$ and $\int^x
dz/F(z)$, respectively, with appropriate functions $F$ and $G$, allows
even for global charts of (two-dimensional) spacetimes of the form
$\dR\, \times$ `null-lines'. For more details on this confer
\pcite{rnletter}, where also a more systematic approach to these
charts is presented, illustrating the considerations by Schwarzschild
and Reissner-Nordstr\oe m.
For \rz saddle to exist it is decisive that the corresponding zero $a$
of $h$ is simple, cf.\ the second Eq.\ \ry ab . For non-critical
values of $C$, e.g., all zeros of $h$ are simple. This is
particularly obvious for torsionless theories, where $h=2X^+X^-$ and
$h'(X^3)=V(X^3)$, but holds also in general. If $C \in \{C_{crit}\}$,
on the other hand, there exist zeros of $h$ of higher degree. Then
the spacetime manifolds $M$ with varying $X^i$ do not contain the
critical points $X^i=(0,0,X^3_{crit})$. This may be seen in two
different ways: First, studying extremals running towards such a
point, one finds the point to be infinitely far away, cf.\ Part II.
Second, from the field equations point of view: Taking successive
derivatives $\6/\6 x^\m$ of the Eqs.\ (\ref }%\def\pr{(\pref{eom1a}) and evaluating them
at the critical points, we find \begin{equation} X^- \equiv 0 \, ,\;\; X^+ \equiv 0
\, ,\;\; X^3 \equiv X^3_{crit}=\hbox{const.} \el momdeSitter
\rz momdeSitter corresponds to additional, separate solutions of
the field equations not treated before. Actually, they come as
no surprise. More or less by definition the critical points \rz
crit of the target space constitute zero dimensional
symplectic leaves. It is a general feature of Poisson
$\s$-models, verified here explicitly in \rz momdeSitter and
the first Eq.\ of (\ref }%\def\pr{(\pref{Loesung}), that the image $X(x)$ of the
map from the worldsheet or space time $M$ into the target space
$N$ has to lie entirely within a symplectic leaf $\CS \subset
N$.
The remaining field equations (\ref }%\def\pr{(\pref{eom1b}), which are, more explicitly,
\begin{equation} De^a =0 \, ,\,\, d\omega = -W,_v(0,X^3_{crit}) \e\,, \el deSitter
show that the solutions \rz momdeSitter have vanishing torsion and
constant curvature all over $M$. The metric for such a solution can
be brought into the form (\ref }%\def\pr{(\pref{gh}), too, with $h(x^0)=
W,_v(0,X^3_{crit}) \cdot [(x^0)^2 +1]$. This in turn determines the
zweibein and spin connection up to Lorentz transformations.
\section{Summary and Extension to Gravity-Yang-Mills}
We demonstrated that any of the 2D gravity models introduced in
Section \ref{Models} is of Poisson $\s$-form \ry PS . Exploiting some
fundamental facts of Poisson structures, namely the local existence of
what we called Casimir-Darboux coordinates, the field equations
reduced to (\ref }%\def\pr{(\pref{eom2a},\ref{eom2b}), the solution of which is
immediate. The relation between the original field variables and the
transformed ones, Eq.\ \ry relation , provided the general solution in
terms of the metric then.\footnote{The employed method may be viewed
as a perfected generalization of what has been done previously
in ordinary dilaton theory \pcite{Ver} or the KV-model \pcite{Solo};
it is, however, not inspired by that works, but self-evident from
the Poisson $\s$ point of view. Let us note on this occasion that
with {\em appropriate} gauge conditions the field equations of \rz
grav may be solved in a maybe less elegant, but almost as
straightforward manner, too \pcite{rnletter}
(cf.\ also \cite{Kummerneu}).} With the choice of a
gauge the latter took the form \ry gh , where $h$ was parametrized by
a single meaningful constant (the value of the Casimir-function $C$).
\rz gh provided the local solution on strips with either $X^+ \neq 0$
or $X^- \neq 0$. For non-critical values of the Casimir constant $C$
(guaranteeing $h'|_{h=0} \neq 0$) the metric could be brought into the
form (\ref }%\def\pr{(\pref{saddle},\ref{ab}) in the vicinity of points $X^+=X^-=0$. For
critical values of $C$, finally, we obtained the deSitter solutions
\rz deSitter in addition to the solutions \rz gh (which in this case
may not be extended to points of simultaneous zeros of $X^+$ and
$X^-$).
The Poisson $\s$-model formulation \rz PS of 2D gravity theories
provides the proper generalization of Yang-Mills (YM) gauge theories
advocated in the introduction. Actually \rz PS is able to describe 2D
YM-theories with arbitrary gauge group. Identifying the target space
$N$ of a Poisson $\s$-model with (the dual of) some Lie algebra with
structure constants $f^{ij}{}_k$ and defining $P^{ij} := f^{ij}{}_k
X^k$, the action \rz PS is seen to become \begin{equation} \int X^i F_i \el BF
after a partial integration, where $F= dA + A \wedge A$ is the
standard Lie algebra valued YM-curvature two-form. The local
symmetries of the BF-YM action \rz BF are the standard ones: $A \to
g^{-1} A g + g^{-1} d g, \, X \to g^{-1} X g$. The symmetries of the
general model \rz PS are a straightforward, nonlinear generalization
of this: \begin{equation} \d_\ep X^i =
\ep_j(x) {\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ji} \, , \; \d_\ep A_i = d\ep_i + {{\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{jk}}_{,i} A_j
\ep_k \, . \el symmetries These symmetries are the Lagrangian
analogues of what is generated by the constraints in a Hamiltonian
formulation (cf.\ Part IV). In the present context of \ry grav ,
where $N=\dR^3$ with Poisson bracket \ry P , the symmetry
transformations \rz symmetries are equivalent to diffeomorphisms and
local Lorentz transformations on-shell. This equivalence holds only
under the assumption of non-degenerate metrics $g=2A_+A_-$, however, a
feature shared also by the Ashtekar formulation of 3+1 gravity. We
will see in Part IV (cf.\ also \pcite{versus}) how this seemingly
irrelevant restriction will lead to different factor spaces (even if
chosen representatives of gauge equivalence classes are restricted to
non-degenerate metrics $g$).
With the addition of one more term not spoiling the symmetries \rz
symmetries the model \rz PS is capable also of describing YM actions
$\int F \wedge \ast F$ \pcite{Mod,proce} or even G/G gauged WZW theories
\pcite{Anton} (cf.\ \pcite{Brief} for a pedagogical exposition). Of
more interest for our present intentions are, however, dynamically
coupled 2D gravity-YM-systems. Allowing for a dilaton- and
also $(X)^2$-dependent coupling constant $\a((X)^2,X^3)$, the action
for such a system has the form \begin{eqnarray} L^{gravYM} &=& L^{grav} + L^{YM}
\label{GYM1} \\
L^{YM}&=& \int {1\over } \def\1{\vec } \def\2{{1\over2}} \def\4{{1\over4} 4 \a((X)^2,X^3)} tr (F \wedge \ast F) \, ,
\label{GYM2} \end{eqnarray}
where the trace is taken in some matrix representation of the chosen
Lie algebra and $\ast$ is the Hodge dual operation with respect to the
dynamical metric $g=2e^+e^-$ of $L^{grav}$, Eq.\ \ry grav . The
special case with an abelian YM-part (gauge group $U(1)$) and with
$\a$ and $W$ depending on $X^3$ only has received some attention in
\pcite{KunstU(1)} recently (but cf.\ also \pcite{Solo,PRD}). Let us
show in the following that the general combined system \rz GYM1 is of
Poisson $\s$-form again! As a byproduct many of the results of
\pcite{KunstU(1)} may be obtained as a lemma to the general theory of
Poisson $\s$-models (including an exact Dirac quantization, cf.\ Part
IV as well as \pcite{PRD,LNP,Mod}).
To begin with we bring \rz GYM2 into
first order form: \begin{equation} L^{YM} \sim {L^{YM}}'=
\int \left( E^i F_i + \a((X)^2,X^3) E^i E_i \,
\varepsilon \right) \, , \el equiYM
where the indices $i$ are raised and lowered by means of the Killing
metric and $\varepsilon \equiv e^- \wedge e^+$.
The equivalence of $L^{YM}$ with ${L^{YM}}'$ is seen by integrating out
the `electrical' fields $E$ (either on the path integral level or just
by implementing the equations of motion for the $E_i$ back into the action,
in complete analogy to how we obtained \rz KV from \ry grav ).
To avoid notational confusion let us rename
$X^\pm, X^3$ into $\varphi^\pm, \phi$ and denote the YM-connection by
a small letter $a$.
Then $L^{grav} + {L^{YM}}'$ reads
\begin{equation} {L^{gravYM}}'= \int \varphi^a De_a + \phi \, d \o + E^j F_j +
\left[W((\varphi)^2,\phi)+ \a((\varphi)^2,\phi) \, E^j E_j\right]
\varepsilon \, , \el GYMprime
where $F_j \equiv d a_j + f^{kl}{}_j a_k \wedge a_l$ and
the indices $a$ run over $+$ and $-$ while the indices $j$ run from
1 to $n$, $n$ being the dimension of the chosen Lie group.
After partial integrations (dropping the corresponding surface
terms) and the identifications
\begin{equation} X^i := (\varphi^a,\phi,E^j) \; , \quad A_i := (e_a,\o,a_j)\,, \el coor
\rz GYMprime becomes of Poisson $\s$-form \rz PS on an
$n+3$-dimensional target space with Poisson brackets:
\begin{eqnarray} & \{\varphi^+,\varphi^-\}= W + \a E^j E_j \; , \quad
\{\varphi^\pm,\phi\}=\pm \varphi^\pm \; , & \nonumber \\
& \{\varphi^\pm,E^j\}=
0=\{\phi,E^j\}\; , & \nonumber \\
& \{E^j,E^k\}=f^{jk}{}_lE^l \; . & \label{PoiGYM} \end{eqnarray}
As any Poisson tensor also the one defined in \rz PoiGYM (note
${\cal P}}\def\CS{{\cal S}}\def\C{{\cal C}^{ij} \equiv \{X^i,X^j\}$) allows for Casimir-Darboux coordinates
in the neighbourhood of generic points. If the rank of the chosen Lie
algebra is $r$ then the rank of the Poisson tensor is $n-r+2$ and there
will be $r+1$ such Casimir coordinates. Correspondingly there will
be $r+1$ field equations \rz eom2a and $n-r+2$ field equations \ry
eom2b . In the CD-coordinates the symmetries \rz symmetries take a
very simple form and it is a triviality to realize that again the
local solutions are parametrized by a number of integration constants
which coincides with the number of independent Casimir functions.
Note that in the present context \rz symmetries entail
diffeomorphisms, Lorentz transformations, and non-abelian gauge
transformations all at once. So, more or less without performing any
calculation, we obtain the result that the local solutions are
parametrized by $r+1$ constants now. In the case of \pcite{KunstU(1)}
$n=1, r=1$, and $r+1=2$. (Here we ignored additional exceptional
solutions of the type \ry momdeSitter , corresponding to maps into
lower dimensional symplectic leaves).
The Poisson structure \rz PoiGYM has a very particular form: First the
$E^j$ span an $n$-dimensional Poisson submanifold of $N=\dR^{n+3}$.
Second, the Poisson brackets between the gravity coordinates
$(\varphi^a,\phi)$ and the coordinates $E^i$ of this submanifold
vanish. And, last but not least, the Poisson brackets between the
`gravity'-coordinates close among themselves up to a {\em Casimir}
function of the $E$-submanifold. With this observation it is a
triviality to infer the local form of $g$ of $L^{gravYM}$ from the
results of Section 3: On-shell $E^jE_j$ is some constant $C^{E}$. So
in all of the formulas of Section 3 we merely have to replace $W$ by
$W+C^{E} \a$. Thus the metric $g$ again takes the form \rz gh locally,
where now $h$ is parametrized by {\em two} constants $C^{E}$ and
$C^{grav}$ (`generalized Birkhoff theorem for 2D gravity-YM-systems').
For torsionless theories, e.g., \begin{equation} h(x^0)=\int^{x^0}_c V(z) dz + \2
C^{E} \int^{x^0}_c \alpha(z) dz + C^{grav} \, , \end{equation} where $c$ is some
fixed constant (and again $2W(u,v)=V(v)$).
The addition of a YM-part still rendered a theory with finitely many
`physical' degrees of freedom. There is at least one generalization of
the gravity action \rz grav that yields a (classically)
solvable theory with an {\em infinite} number of degrees of freedom.
This is obtained when coupling fermions of one chirality to \ry grav
\ \pcite{Heiko}, as was observed first in \pcite{FermionKummer}
in the context of the KV-model \ry KV .
On the level of local considerations the models introduced in Section
\ref{Models} all look quite alike. They all have a one-parameter
family of solutions of the form \rz gh (in the neighbourhood of
generic points). As will be seen in Parts II,III of this work, this changes
drastically, if one turns to global considerations. The richness and
complexity of the global solutions is encoded in the kind in which the
target space $N=\dR^3$ foliates (more precisely `stratifies') into
symplectic leaves. E.g., a potential $V$ in \rz grav with many zeros
will lead to a topologically complicated stratification of $N$ into
two- and zero-dimensional symplectic submanifolds. Correspondingly, as
we will find in Part III, there will exist smooth solutions on
space-times $M$ with relatively complicated topologies. If, on the
other hand, $V$ has no zeros (such as in the JT- or the ordinary
dilaton theory), the foliation (stratification) of $N$ is quite simple
and the most complicated topologies of smooth space-time
solutions $M$ occuring are cylinders and M\"obius strips.
On the global level the solutions will be parametrized no longer by a
single parameter only (or, in the context of \ry GYM1 , no longer by
$r+1$ parameters). Instead there will be $m+1$ (resp.\ $(m+1)$
times $(r+1)$) such continuous parameters labeling them (in
addition to further discrete parameters), where $m$ is the number of
independent non-contractible loops on $M$. Here it should be noted
that given some particular model (via specifying $V$ or $W$) there
will be solutions with different topologies of $M$. Consequently the
solution space (defined as the space of all possible globally smooth
solutions modulo gauge transformations) of the chosen model will have
different components differing in dimension.\footnote{Not to be mixed
up with the dimension of $M$, which certainly always remains two. It
is the fundamental group of a possible $M$ which determines the
dimension of the respective component of the solution space.} As at
present there is no classification of the solution space of a general
Poisson $\s$-model yet, in Parts II,III we will approach this issue in our
specific gravity context from a more traditional, pedestrian point of
view. First we will construct the universal coverings to the local
solutions studied in this paper. Thereafter we then investigate the
possible smooth factor solutions, keeping track of all arising
parameters.
In Part IV, finally, we will turn to the quantum theory of \rz grav
or \rz GYM1 and compare the result to the classical solution space.
It is in this quantum regime where the full power of the Poisson
$\s$-formulation will become particularly apparent.
\section*{Acknowledgement}
We are grateful to H.D.\ Conradi and W.\ Kummer
for discussions and remarks concerning the manuscript.
This work has been supported in part by the Austrian Fonds zur F\"orderung
der wissenschaftlichen Forschung (Project P10221-PHY).
|
\section{Introduction}
\setcounter{equation}{0}
The understanding and description of irreversible evolution of macroscopic
systems, presumably towards equilibrium states is the aim of kinetic theory.
This is done by formulating appropriate kinetic equations giving the time
evolution of the state of the system which loosely speaking, is assumed to be
some probability measure on the space of variables describing the system,
which henceforth will be called its phase-space. Observables are assumed to
be well-behaved phase-space functions and experimental results refer to
expectation values obtained via a bilinear (or sesquilinear) form on the
cartesian product of the spaces of states and observables. Thus at the
very heart of kinetic theory a probabilistic view point is rooted, the
precise interpretation of which is often a matter of debate.
Theoretically speaking, kinetic equations are derived by following two
different procedures: \begin{enumerate}
\item[(i)] stochastic arguments, based on some presumably plausible
assumptions on the behaviour of a large number of microscopic events
characterising the system.
\item[(ii)] application of more or less systematic approximation schemes
on the exact microscopic dynamics of the system under consideration.
\end{enumerate}
Typical examples for (i) are the Fokker-Planck and Kramers equations
(e.g.\ \cite{1} ch.VIII). The first refers to the probability
distribution function (p.d.f.) of the velocity of a heavy Brownian particle
suspended in an equilibrium medium of light particles, the second one, to
the p.d.f.\ of its phase-space position under the presence of an external
field. The basic assumptions for their derivation are that the microscopic
dynamics of the Brownian particle are governed by Langevin's equation and
that its p.d.f.\ is a Markovian diffusion process (see e.g.\ \cite{2}
ch.II, \cite{2a}).
This procedure usually leads to {\em model} equations
that are easier to study than the {\em assumed} more fundamental equations
following from exact dynamics by (ii). Mathematically speaking, this
procedure essentially {\em replaces exact dynamics by stochastic differential
equations} which in turn imply evolution equations for the p.d.f.\ of the
system (e.g.\ \cite{3} ch.4, \cite{4} ch.9).
Though interesting and important from both the mathematical and physical
point of view, it seems that one has to know somehow the kinetic equation
he wants to derive and modify accordingly the corresponding dynamics.
Moreover the modification is not easily interpreted physically
(see the discussion following (\ref{3.1}) below). That is, it seems that
there is no general prescription of how this modification has to be carried
out. In this paper, starting from mathematically totally different concepts
we do arrive at results that may be interpreted stochastically, at the same
time giving hints on the nature of such a general prescription
(cf.\ the discussion following (\ref{3.1}) below).
Typical examples for (ii) are the Boltzmann equation for dilute classical
gases (or its quantum weak-coupling analogue, Pauli's equation) or the
Landau and Balescu-Lenard equations for neutral plasmas, as they are derived
from Liouville's (or von Neumann's) equations by using either iteration
schemes and projection operator methods, or by using its equivalent,
the BBGKY hierarchy of equations truncated on the basis of physical
considerations.\footnote{See e.g.\ \cite{5} \S\S30, 41, \cite{6} ch.20,
\cite{7} ch.IX, \cite{8} \S2.4, \cite{9} \S\S4.3, 4.6, \cite{10}
{\S}VI.}
Although many specific equations can be derived by mathematically
satisfactory (or even rigorous) methods in some particular limit of
appropriate parameters of the system (for a survey see \cite{32}), it is
true that {\em any approximation scheme, leading to satisfactory kinetic
equations for particular classes of systems, runs into trouble as long as
one tries to extend it to other systems and/or higher order of
approximation} (\cite{11}, \cite{12} \S5). For instance the linearized
Landau equation follows for spatially homogeneous plasmas as a 2nd order
approximation in the plasma parameter to the Liouville equation (plus
some additional assumptions which need not be discussed here). Any effort
to find its generalization either in higher order of approximation and/or
for inhomogeneous systems, runs into difficulties (e.g.\ equations
violating the positivity of p.d.f or having no $H$-theorem, are obtained
--- see e.g.\ \cite{13} in connection with \cite{12} \S5), or involves
highly ad hoc steps which are sometimes hidden in the calculations
(see e.g.\ \cite{14} in connection with \S2 below).
Although one may be tempted to accept that there is no reason to expect that
equations in an approximate theory share all properties of the corresponding
equations in the exact theory, in kinetic theory things are more complicated
since the exact theory in this case is simply (classical or quantum) dynamics;
however as it stands, microscopic dynamics of a system with a very large
number of degrees of freedom is useless as a macroscopic theory since it does
not incorporate irreversible evolution and its relation to a theory dealing
with macroscopically defined quantities is remote, or at least not
straightforward. Moreover from what has been said above, the situation is
even worse, since it often happens that equations obtained at a lower level
of approximation (e.g.\ with respect to expansion in some parameter)
exhibit the correct properties, which however disappear in any higher
level of approximation!\footnote{See e.g.\ equations following from expansion
of
Liouville's equation or the so-called generalized master equation \cite{9}
\S2.3, \cite{11} \S2, \cite{12} \S5, \cite{26} in connection with the last
paragraph of the next section. See also \cite{33} for expansions of the
Boltzmann equation.}
In our opinion this is an inevitable consequence of the philosophy underlying
kinetic theory, namely that {\em irreversible evolution} is simply an
approximation to the exact (classical or quantum) {\em reversible dynamics}.
To put it differently, the {\em aim} of kinetic theory (description of
macroscopic irreversible evolution) seems not to be {\em compatible} with
its {\em assumed fundamental laws} (reversible microscopic dynamics), at
least in a straightforward manner. Therefore it seems necessary that a
fundamentally different approach to kinetic theory may not be worthless!
\section{Mathematical assumptions underlying kinetic theory
and their implications.}
\setcounter{equation}{0}
As already mentioned, at the very root of kinetic theory lies a
probabilistic interpretation. Using this as a motivation and in order to
clarify the difficulties noticed in the previous paragraph, we here describe
the basic assumptions for a kinetic equation to be in principle, acceptable,
and conditions under which an explicit general form can be obtained.
We restrict the discussion to {\em linear, autonomous} kinetic equations,
which include kinetic equations for the important class of {\em open systems}.
To be more specific, we consider {\em classical} systems, though similar
results are known for quantum systems as well (\cite{18}, \cite{15}
Theorem 4.2).
We assume that \begin{enumerate}
\item[(i)] The phase-space $M$ is a locally compact, Hausdorff topological
space (e.g. the phase-space of a Hamiltonian system).
\item[(ii)] Observables $A$ are in $C(M,{\ibb C})$ the space of complex-valued
continuous functions on $M$, having a finite limit at infinity.
\item[(iii)] States $\ell$ are positive linear functionals on the
observables, their values $\ell(A)$ giving expectations. Since positivity
of $\ell$, implies its boundedness in the supremum norm (e.g. \cite{20}
p.106--107), states belong to $C^\ast(M,{\ibb C})$, the Banach dual with
respect to this norm, which is the space of (regular) complex Borel
measures on $M$.
\item[(iv)] Kinetic equations for the observables have well-posed initial
value problem, i.e.\ uniqueness and continuous dependence of solutions on
the initial data hold. Moreover, expectation values are continuous in time.
These imply that solutions of a kinetic equation define strongly
continuous semigroups of linear operators on the space of observables and
that the corresponding adjoint equation defines such a semigroup on the
state space.\footnote{Strictly speaking, strong continuity of the latter
holds on a smaller subspace, which however uniquelly defines the adjoint
semigroup (\cite{21} Theorem 1.36).}
\item[(v)] The adjoint semigroup conserves positivity and normalization
of the states, i.e.\ initial probability measures retain their character
for all positive times.
\end{enumerate}
These plausible assumptions imply that the solutions of a kinetic equation
for the observables define a Markov semigroup, i.e.\ a strongly continuous,
positivity and normalization preserving one-parameter semigroup of operators
on $C(M,{\ibb C})$ \cite{26}. The terminology stems from the fact that such
semigroups are in one to one correspondence with (time-homogeneous)
stochastically continuous Markov processes described by a transition
probability distribution $p(t,x,E)$, which for each $t,x$ is a regural
probability Borel measure on $M$ (see e.g.\ \cite{22} p.399, \cite{23}
Theorem 2.1, for outline of a proof). If we further {\em assume} that
$M$ is an $n$-dimensional differential manifold, the generator of the
semigroup is defined on $C^2$-functions and that a Lindenberg's type
condition holds,
\begin{eqnarray*}
\lim_{t\to 0^+}{p(t,x,E)\over t}=\chi_E(x)\qquad\qquad
\mbox{uniformly in $x$,}
\end{eqnarray*}
where $\chi_E$ is the characteristic function of $E$, then it can be proved
that the kinetic equation for observables (i.e.\ essentially the generator
of the corresponding semigroup) has the form\footnote{See \cite{24}
Theorem 5.3 for outline of a proof, and \cite{25} Theorem XIII.53 for a
partial generalization --- the L\'{e}vy-Khinchine formula.}
\begin{eqnarray}
{\partial A\over\partial t}=\alpha^{ij}(x)\partial_i\partial_j A+ a^i(x)\partial_i A\;,\label{2.1}
\end{eqnarray}
where $\alpha^{ij}$, $a^i$ are continuous, $\alpha^{ij}$ is a non-negative
definite matrix function and the summation convention has been used, as
it will be done in the sequel.\footnote{For detailed proofs and more precise
formulation of the various conditions see \cite{26}.}
Actually $\alpha^{ij}, a^i$ are related to the diffusion and drift
coefficients associated with the corresponding Markov process, given by
the first two moments of $p$.
In many approaches in kinetic theory which lead to satisfactory kinetic
equations of the form (\ref{2.1}), the following condition holds
\begin{eqnarray} a^i=b^i+X_H^i\;, \label{2.2} \end{eqnarray}
where $X_H^i$ is a {\em Hamiltonian} vector field. Moreover for {\em open}
(classical or quantum) systems with Hamiltonian $H$, in interaction with
``baths'' in canonical equilibrium, the corresponding Hamiltonian function
is an integral of the unperturbed motion of the open system, and therefore
has in general the form $-(H+F)$ with
\begin{eqnarray} \{ F,H \}=0\;,\label{2.3} \end{eqnarray}
$\{\,,\}$ being the Poisson bracket, or the operator commutator, and the minus
sign gives the correct Hamiltonian equation when the system does not
interact with the bath.
Moreover
\begin{eqnarray} \partial_j\alpha^{ij}-b^i=\beta\alpha^{ij}\partial_j H\;, \label{2.4}\end{eqnarray}
where $\beta$ is proportional to the inverse temperature of the
bath.\footnote{See e.g.\ \cite{12} Proposition 4.1, \cite{27} eq.(2.19),
\cite{5} p.190, \cite{17} eqs(III.26) and (III.16) together with (III.19),
\cite{31} eq.(5.8).} These conditions will be used in section 4.
Substitution of (\ref{2.2})--(\ref{2.4}) in (\ref{2.1}) gives
\begin{eqnarray}
{\partial A\over \partial t}=-\{H+F,A\}+\partial_i(\alpha^{ij}\partial_j A)-
\beta\alpha^{ij}\partial_j H\partial_i A\;.\label{2.5}
\end{eqnarray}
The essential conclusion drawn from the above discussion is (in a somewhat
nonrigorous language) that {\em linear autonomous kinetic equations for
classical systems, which are differential equations are necessarily at
most of the 2nd order with nonnegative-definite leading coefficient and
vanishing zeroth order term (cf.\ eq.(\ref{2.1}))}\,.\footnote{A corresponding
result is also known for quantum systems \cite{18}, \cite{16}, \cite{17}.
The special case of the Kramers-Moyal expansion of the linearized
Boltzmann equation has been considered in \cite{33}.}
Since kinetic equations following from microscopic dynamics (see \S1)
are often differential equations, the above result, severely restricts their
form. In particular all methods based on power expansions of the solution of
Liouville's equation with respect to some appropriate parameter, usually
lead to unacceptable results since each approximation step increases
the order of the differential operator by one.\footnote{See e.g.\ \cite{9}
\S2.4 eq.(2.199), \cite{1} p.215, {\S}IX.6 particularly p.280.}
In the rest of this paper, we will present a different point of view,
motivated by the discussion in the next section.
\section{Non-commutative geometry and stochastic calculus.}
\setcounter{equation}{0}
The usual approach to kinetic theory (method (ii) in Section 1) follows the
scheme
\begin{eqnarray*}
\begin{array}{ccc}
\mbox{Microscopic (Hamiltonian dynamics)} & & \\
+ & \Longrightarrow & \mbox{Kinetic equations} \\
\mbox{Some systematic approximation scheme} & &
\end{array}
\end{eqnarray*}
On the other hand, method (i) in Section 1, involves no systematic
approximation scheme but a modification of microscopic dynamics which
become now stochastic differential equations. In our opinion, the weak
point in this case is that there is no general method of how to choose the
microscopic stochastic differential equations. That is, it seems necessary
that one somehow knows the kinetic equation he wants to derive and then
writes down the corresponding stochastic equation. However it is not always
clear how to interpret the latter.
For instance it would be desirable to be able to derive by method (i) kinetic
equations obtained from microscopic Hamiltonian dynamics. However, the
latter in general involve operators with derivatives in the $q$-coordinates
(see e.g.\ \cite{11}, \cite{12}, \cite{27}, \cite{31}). And it is a standard
fact that stochastic differentil equations in phase-space involving Wiener
processes $\vec{X}_t$ of the form
\begin{eqnarray}
\left(\begin{array}{c} d\vec{q} \\
d\vec{p} \end{array}\right) =
\left(\begin{array}{c}
\vec{A}_q(\vec{q},\vec{p})\, dt+{\bf F}_q(\vec{q},\vec{p})\cdot d\vec{X}^q_t\\
\vec{A}_p(\vec{q},\vec{p})\, dt+{\bf F}_p(\vec{q},\vec{p})\cdot d\vec{X}^p_t
\end{array}\right)\;, \label{3.1}
\end{eqnarray}
imply such derivatives in the corresponding kinetic equation if ${\bf F}_q
\neq 0$ (see e.g.\ \cite{3} \S4.3.3). However in many cases, one would
like to interpret $d\vec{q}/dt$ simply as a velocity, all dynamics being
incorporated in $d\vec{p}/dt$ (e.g. the velocity of a Brownian particle),
in which case the above stochastic equation (\ref{3.1}) is not easily
interpreted physically.
In the next sections we outline {\em another approach} in which {\em the
Hamiltonian character of microscopic dynamics is retained, but instead of
approximation schemes, we make the fundamental assumption that observables
are now defined on a manifold with noncommutative geometrical structure}\,.
As it will be explained, this may be interpreted as a stochastic dynamical
structure, though it is not known if this is necessary.
Nevertheless, the formulation developped in the next section is closely
related to what may be called a differential geometric approach to
stochastic calculus\footnote{See e.g.\ \cite{36}, \cite{37} particularly
ch.VI, compare also with the approach in \cite{38}.} though this relation
will be further explored in another paper.
The {\em mathematical} motivation for introducing noncommutative geometrical
structure, stems from the fact that in the one-dimensional version of
(\ref{3.1}), $X_t$ being a Wiener process (i.e.\ Langevin's equation)
the usual associative product between differential forms spanned by $dt$,
$dX_t$ and functions of $t$, $X_t$, implies that stochastic differentiation
does not obey Leibniz's rule for the product of two functions, due to the
appearance of 2nd derivatives in It\^o's formula for the differential of
such functions (\cite{28} \S2, particularly eq.(2.4)).However it is possible
to define a modified {\em noncommutative} but still associative product
between functions and differential forms such that Leibniz's rule holds
(\cite{28} section 3). This product induces a noncommutative differential
calculus on the ordinary algebra of functions of $t$, $X_t$ (\cite{29}
section 2) via the basic commutation relations between functions and 1-forms
\begin{eqnarray} \begin{array}{r}
\lbrack \d t,t \rbrack =\lbrack \d t,X_t\rbrack =0\;, \\
\lbrack\d X_t,X_t \rbrack =2\gamma\,\d t\;, \end{array}\label{3.2}
\end{eqnarray}
$\gamma$ being the diffusion constant appearing in the usual Fokker-Planck
equation obtained from Langevin's equation (\cite{28} eq.(3.16)).
Eq.(\ref{3.2}) is readily generalized to a multi-dimensional Wiener process:
\begin{eqnarray} \begin{array}{r}
\lbrack\d t,t \rbrack =\lbrack\d t,X^i_t\rbrack =0\;, \\
\lbrack\d X^i_t,X^j_t \rbrack =b^{ij}\,\d t\;,\end{array} \label{3.3}
\end{eqnarray}
where $b^{ij}$ is a symmetric bilinear form on the space of
1-forms.\footnote{This is implied by the fact that since $\d$ satisfies
Leibniz's rule, $\lbrack\d X^i_t,X^j_t \rbrack = \lbrack\d X^j_t,X^i_t \rbrack$ which
in fact shows that this commutator depends only on $\d X^i_t$, $\d X^j_t$
--- see \cite{29} \S3.} Therefore (\ref{3.3}) can be rewritten more generally
as
\begin{eqnarray} \lbrack \d f, g\rbrack = b(\d f,\d g)\,\d t\;, \label{3.4}
\end{eqnarray}
for functions $f,g$ of $\vec{X}_t,t$ and a symmetric, bilinear form with
components $b^{ij}$ in the ``coordinates'' $X^i_t$ and such that $\d t$ lies
in its kernel, i.e.\ $b^{tt}=b^{it}=0$.
In view of the above discussion, we consider in the next section a
$(2n+1)$-dimensional manifold $M\times{\Ibb R}$ and a differential calculus
on a subalgebra ${\cal A}$ of the algebra of complex-valued functions on
$M\times{\Ibb R}$ satisfying (\ref{3.4}), and outline {\em the formulation of
extended Hamiltonian dynamics as symplectic geometry on $M\times{\Ibb R}$}.
The basic result is that {\em Liouville's equation for observables turns out
to be of the form (\ref{2.1}) with conditions (\ref{2.2}), (\ref{2.4})
having a simple geometrical meaning. Therefore it may be interpreted as a
kinetic equation on the space of observables corresponding to a classical open
system with phase-space $M$}.
The construction is coordinate-independent, and presupposes the definition of
such fundamental concepts as vector fields, linear connections, symplectic
structure and antisymmetric wedge product of forms on the differential
calculus defined on ${\cal A}$ by (\ref{3.3}), in close analogy with the
corresponding concepts of the ordinary differential geometry.\footnote{For
details on the systematic definition and presentation of general results
in noncommutative geometry on a commutative algebra, see \cite{29},
particularly \S\S2,3.} The derivation is {\em formal} in the sense that no
systematic study of the representation theory of (\ref{3.3}) is made. Its
already mentioned relation with stochastic calculus is a possibility,
but it is not clear if others exist. A preliminary discussion of this problem
is given in section 5.
To make the presentation as transparent as possible, detailed calculations
will be given in a subsequent paper in which tensor analysis for the
corresponding noncommutative differential calculus is developped
systematically. In fact this work constitutes only a first step towards
a systematic formulation of kinetic theory as Hamiltonian (symplectic)
dynamics in phase-space equipped with a noncommutative geometrical
structure.
\section{Noncommutative symplectic geometry}
\setcounter{equation}{0}
Let $M$ be a $2n$-dimensional manifold and ${\cal A}$ the algebra of smooth
functions on $M\times{\Ibb R}$. The coordinate on ${\Ibb R}$ will be denoted by
$t$. Let $\~{\O}$ be the universal differential envelope of ${\cal A}$, i.e.\
$\~{\O}$ is a $Z\!\!\!Z$-graded algebra $\~{\O}=\bigoplus_{r\inZ\!\!\!Z}\~{\O}^r$
with $\~{\O}^r=\{0\}$ for $r<0$ and $\~{\O}^0={\cal A}$. Then there exists a linear
mapping $\~{d}:\~{\O}\to\~{\O}$ of grade 1, which satisfies
\begin{enumerate}
\item[(i)] $\~{d} 1=0$, where 1 is the constant function with value 1,
\item[(ii)] $\~{d}$ satisfies the graded Leibniz rule i.e.\
$\~{d}(\psi\psi')=(\~{d}\psi)\psi'+(-1)^r\psi(\~{d}\psi')$, for
$\psi\in\~{\O}^r$,
\item[(iii)] $\~{d}^2=0$ on all of $\~{\O}$ and
\item[(iv)] ${\cal A}$ and $\~{d}{\cal A}$ generate $\~{\O}$.
\end{enumerate}
The universal differential envelope $(\~{\O},\~{d})$ of ${\cal A}$ can be
realized as follows (see \cite{39}): think of $\phi\in\~{\O}^r$ as a function
on
$(M\times{\Ibb R})^{r+1}$, where for $f\in\~{\O}^0$ and $a\in M\times{\Ibb R}$
$f(a)$ is the value of $f$ as an element of ${\cal A}$ on $a$ and
for $a_0,\ldots,a_{r+1}\in M\times{\Ibb R}$ and $\phi\in\~{\O}^r$ we set
\begin{eqnarray}
(\~{d}\phi)(a_0,\ldots,a_{r+1}):=\sum_{k=0}^{r+1}
(-1)^k\phi(a_0,\ldots,a_{k-1},a_{k+1},\ldots,a_{r+1})\;.\nonumber
\end{eqnarray}
Furthermore for $\phi\in\~{\O}^r$, $\psi\in\~{\O}^s$ and $a_0,\ldots
a_{r+s}\in M\times{\Ibb R}$ we set
\begin{eqnarray}
(\phi\psi)(a_0,\ldots,a_{r+s}):=[\phi(a_0,\ldots,a_r)]
[\psi(a_r,\ldots,a_{r+s})]\;, \nonumber
\end{eqnarray}
for any non negative integers $r,s$. According to these rules
\begin{eqnarray}
(f\~{d} g\,h)(a,b)=f(a)[g(b)-g(a)] h(b)\;,\label{4.1}
\end{eqnarray}
and hence $\~{d} f\,g\neq g\~{d} f$. On the ${\cal A}$-bimodule of 1-forms
$\~{\O}^1$ we define a new product $\~{\bullet}:\~{\O}^1\times\~{\O}^1
\to\~{\O}^1$ as follows: for $\alpha,\beta\in\~{\O}^1$ and
$a,b\in M\times{\Ibb R}$ we set
\begin{eqnarray*}
(\alpha\~{\bullet}\beta)(a,b):=\alpha(a,b)\,\beta(a,b)\;.
\end{eqnarray*}
Note that
\begin{eqnarray*}
(f_1\alpha f_2)\~{\bullet}(g_1\beta g_2)=f_1g_1(\alpha\~{\bullet}\beta)f_2g_2\;,
\end{eqnarray*}
and $[\~{d} f,g]=\~{d} f\~{\bullet}\~{d} g$.
The universality of $(\~{\O},\~{\d})$ is expressed by the property that, if
$(\O,\d)$ is any differential algebra on ${\cal A}$, then there is a
graded-algebra homomorphism $\pi:\~{\O}\to\O$ of grade 0 such that
$\pi|\~{\O}^0=\mbox{id}_{\cal A}$ and $\d\circ\pi=\pi\circ\~{\d}$
(cf.\ \cite{29} \S3.1).
Let $\~{b}:\~{\O}^1\times\~{\O}^1\to{\cal A}$ be a symmetric left-right
${\cal A}$-bilinear form i.e.\ $\~{b}(f_1\alpha f_2,g_1\beta g_2)=
f_1g_1\,\~{b}(\alpha,\beta)\,f_2g_2$ and assume that $\~{d}t$ lies in
the kernel of $\~{b}$.
Let also ${\cal I}$ denote the differential ideal of $\~{\O}$ generated by
$\alpha\~{\bullet}\beta-\~{\d}t\,\~{b}(\alpha,\beta)$, then we set
$\O:=\~{\O}/{\cal I}$ and $\pi:\~{\O}\to\O$ for the canonical projection.
Since ${\cal I}$ is differential the operator $\d:\O\to\O$ given by
$\d=\pi\circ\~{\d}$ is well defined, and because $\~{\d}t$
lies in the kernel of $\~{b}$, a symmetric left-right ${\cal A}$-bilinear form
$b:\O\times\O\to{\cal A}$ is uniquelly defined by $b\circ(\pi\times\pi)=\~{b}$.
Now set
\begin{eqnarray} \d f\bullet\d g:=[\d f,g] \;, \label{4.2} \end{eqnarray}
and extend by left-right ${\cal A}$-bilinearity, then it is easy to see that
$\pi(\~{\alpha}\~{\bullet}\~{\beta})=\alpha\bullet\beta$, where
$\alpha=\pi(\~{\alpha})$ and similarly for $\beta$. Obviously we have
\begin{eqnarray} \alpha\bullet\beta=\d t\,b(\alpha,\beta)\;. \label{4.3}\end{eqnarray}
This is but a special case of the general procedure used to relate
$(\~{\O},\~{\d})$ to any other differential calculus $(\O,\d)$ via
an ${\cal A}$-bimodule homomorphism $\pi$, with ${\cal I}=\ker\,\pi$
and induce $\bullet$ on $\O^1$ by $\~{\bullet}$ on $\~{\O}^1$
(cf.\ \cite{29} \S 3.2).
Let $\xi^i,\,i=1,\ldots,2n$ be local coordinates on $M$ then the elements
of ${\cal A}$ can be written locally as functions of $t,\,\xi^i,\,i=1,\ldots,2n$
(see also section 5). If we set $b^{ij}:=b(\d\xi^i,\d\xi^j)$ we find the
commutation relations --- note that $b(\d t,\d\xi^i)=b(\d\xi^i,\d t)=0$ ---
\begin{eqnarray}
\lbrack \d\xi^i,\xi^j \rbrack & = & \d t b^{ij}\;, \label{4.4}\\
\lbrack \d t,t \rbrack = \lbrack \d t, \xi^i \rbrack = \lbrack \d\xi^i, t \rbrack & = & 0\;.
\label{4.5}
\end{eqnarray}
These are special cases of
\begin{eqnarray} [\d f, g] = \d t \, b(\d f,\d g)\;. \label{4.6}\end{eqnarray}
By (\ref{4.2}), (\ref{4.5}) we get for any $f,g,h\in{\cal A}$ that
\begin{eqnarray} \d f\bullet\d g\bullet\d h=0\;. \label{4.7}\end{eqnarray}
Applying $\d$ on a product of two functions and using (\ref{4.6}) we obtain
\begin{eqnarray} \d(fg)=(\d f)g + (\d g)f - \d t\,b(\d f,\d g)\;. \label{4.8} \end{eqnarray}
Considering $\O^1$ as a right ${\cal A}$ module the dual module ${\cal X}$ is a
left ${\cal A}$ module. We write $\langle X,\alpha\rangle$ for the duality
contraction. If we define
\begin{eqnarray} Xf:=\langle X,\d f\rangle \;, \label{4.9}\end{eqnarray}
then we obtain from (\ref{4.8})
\begin{eqnarray} X(fg)=g(Xf)+f(Xg)- b(\d f,\d g) (Xt)\;. \label{4.10}\end{eqnarray}
The elements of ${\cal X}$ will be called {\em vector fields}. It can be proved
that as a left ${\cal A}$ module ${\cal X}$ is free with basis
given by $\hat{\partial}_t,\,\partial_1,\ldots,\partial_{2n}$, where
\begin{eqnarray}\begin{array}{c}
\partial_i:=\displaystyle{\partial\over\partial\xi^i}\;,\\
\hat{\partial}_t:=\partial_t-{1\over2}b^{ij}\partial_i\partial_j\;.
\end{array}\label{4.11}
\end{eqnarray}
Thus for $X\in{\cal X}$ we have
\begin{eqnarray} X=\hat{X}^t\,\hat{\partial}_t + X^i\,\partial_i\;, \label{4.12}\end{eqnarray}
with $X^i:=(X\xi^i)$, $\hat{X}^t:=(Xt)$. More generally it can be proved
that for any differential calculus on a differential manifold, satisfying
(\ref{4.7}) vector fields, i.e.\ elements of ${\cal X}$, are second order
differential operators without constant term, like (\ref{4.12}). Thus the
name 2nd order calculus is justified in this case (cf.\ section 5).
As a further consequence $\O^1$ is
free with dual basis $\d t,\,\d\xi^i,\,i=1,\ldots,2n$, and hence
\begin{eqnarray} \d f=\d t\,\hat{\partial}_t f+\d\xi^i\,\partial_i f\;.\label{4.13} \end{eqnarray}
Using the bilinear form $b$ we define a linear mapping from $\O^1$ to ${\cal X}$,
$\alpha\mapsto\alpha^b$ by
\begin{eqnarray*} \langle \alpha^b,\beta\rangle:=b(\alpha,\beta)\;. \end{eqnarray*}
Note that $(\d t)^b=0$ and $(\d\xi^i)^b=b^{ij}\partial_j$.
Relations for forms of higher grade are obtained by applying $\d$ on
equations (\ref{4.4}), (\ref{4.5}). We find
\begin{eqnarray} \begin{array}{c}
\d\xi^i \d \xi^j + \d\xi^j \d\xi^i = \d t \d b^{ij}\;, \\
\d t \d t =0\;, \quad\qquad\d\xi^i \d t + \d t \d\xi^i=0\;.
\end{array}\label{4.14}
\end{eqnarray}
These are special cases of
\begin{eqnarray} \d f\d g+\d g\d f= \d t\,\d b(\d f,\d g)\;, \label{4.15} \end{eqnarray}
which is obtained by application of $\d$ on (\ref{4.6}).
{\em From (\ref{4.6}) and (\ref{4.15}) follows that all deviations of the
present differential calculus from the classical differential calculus
are proportional to $\d t$. Therefore and by the second equation in
(\ref{4.14}) it is also clear that for forms $\d t\phi$, for any $\phi\in\O$
all calculations proceed classically.} This will help us to proceed more
rapidly in what follows.
We extend the $\bullet$ product to act between any 1-form $\alpha$ and an
arbitrary form $\phi$ by using the ``insert'' opeartor $\ins$ of ordinary
exterior calculus
\begin{eqnarray} \alpha\bullet\phi:=\d t\,\alpha^b\ins\phi\;. \label{4.16} \end{eqnarray}
On the right hand side everything is as in the ordinary differential calculus
because of the presence of $\d t$ . It is not difficult to see that
(cf. (\ref{4.2}))
\begin{eqnarray*} \d f\bullet\phi =[\phi,f]\;. \end{eqnarray*}
For a one form $\alpha$ the combination $\alpha\bullet$ acts as a derivation
of the product of differential forms, i.e.
\begin{eqnarray*} \alpha\bullet(\phi\psi)=(\alpha\bullet\phi)\psi+\phi(\alpha\bullet\psi)\;.\end{eqnarray*}
The elements $u$ of ${\cal X}$ which vanish on $t$, i.e.\ $u(t)=0$ are derivations
of ${\cal A}$ and define a left ${\cal A}$ submodule ${\cal X}_1$ of ${\cal X}$. With every
$u\in{\cal X}_1$ we associate mappings $D_u:\O\to\O$ defined up to terms
lying in $\d t\,\O$ through the following relations
\begin{eqnarray} D_u(\phi\psi)=(D_u\phi)\psi+\phi(D_u\psi)\pmod{\d t}\;,\label{4.17}\end{eqnarray}
\begin{eqnarray} D_u\,f:=u(f)\;,\label{4.18}\end{eqnarray}
and
\begin{eqnarray} D_u\d t=0\pmod{\d t}\;. \label{4.19}\end{eqnarray}
We write $D_i$ for $D_{\partial_i}$ and we set
\begin{eqnarray} D_i\d\xi^j =-\d\xi^k\,\Gamma^j{}_{ki}\pmod{\d t}\;, \label{4.20}\end{eqnarray}
for the coefficients of the connection. For an $r$-form $\phi$
with
\begin{eqnarray*}
\phi={1\over r!}\phi_{i_1\cdots i_r}\d\xi^{i_1}\cdots\d\xi^{i_r}
\pmod{\d t}\;,
\end{eqnarray*}
we find
\begin{eqnarray}
D_k\phi={1\over r!}\nabla_k\phi_{i_1\cdots i_r}\d\xi^{i_1}\cdots\d\xi^{i_r}
\pmod{\d t}\;, \label{4.21}
\end{eqnarray}
where
\begin{eqnarray}
\nabla_k\phi_{i_1\cdots i_r}:=\partial_k\phi_{i_1\cdots i_r}-
\Gamma^j{}_{ki_1}\phi_{j\cdots i_r}-\cdots-
\Gamma^j{}_{ki_r}\phi_{i_1\cdots j}\;.\label{4.22}
\end{eqnarray}
Extending these definitions as usual to tensor products we obtain
\begin{eqnarray}
\nabla_i b^{jk}:=\partial_i b^{jk}+b^{\ell(j}\Gamma^{k)}{}_{\ell i};. \label{4.23}
\end{eqnarray}
In the following {\em we demand the connection to be torsion free i.e.}
$\Gamma^i{}_{[jk]}=0$, from which follows
\begin{eqnarray} \d\phi =\d\xi^i\,D_i\phi\pmod{\d t}\;, \label{4.24} \end{eqnarray}
and {\em $b$-compatible}, that is $\nabla_ib^{jk}=0$.
It should be emphasized here that in the context of the present
noncommutative differential calculus it is possible to develop systematically
tensor analysis so that the introduction of the above mentioned concepts
of a connection and covariant derivative (cf.\ eqs(\ref{4.17})--(\ref{4.20}),
(\ref{4.21})--(\ref{4.23})) is made perfectly rigorous. However, this would
lead us far away from our task to develop symplectic geometry and Hamiltonian
dynamics and therefore it will be presented in a subsequent paper.
Let us also remark here that in the old-fashioned index notation these
differential
geometric tools are first introduced in \cite{40}.
With the aid of these mappings and the $\bullet$ we define a new product in
$\O$
\begin{eqnarray} \phi\wedge\psi:=\phi\psi+(D_i\phi)(\d\xi^i\bullet\psi)\;.\label{4.25} \end{eqnarray}
It is easy to check that $\wedge$ is right ${\cal A}$ linear in both factors, i.e.
\begin{eqnarray} (\phi f)\wedge(\psi g)=(\phi\wedge\psi)fg\;.\label{4.26} \end{eqnarray}
Furthermore it can be shown that the product is associative and as we shall
see below also graded commutative, i.e.\ for $\phi\in\O^r$
and $\psi\in\O^s$
\begin{eqnarray*} \phi\wedge\psi=(-1)^{rs}\psi\wedge\phi\;. \end{eqnarray*}
An easy consequence of the above definition is
\begin{eqnarray*}
\phi\wedge f=\phi f\;,\qquad\qquad
f\wedge\phi=f\phi+\d f\bullet\phi=\phi f\;.
\end{eqnarray*}
We define now an operator
\begin{eqnarray} \mbox{D} f:=\d f+{1\over 2}\d\xi^i\bullet D_i\d f\;,\label{4.27} \end{eqnarray}
motivated by the fact that it satisfies the usual Leibniz rule
\begin{eqnarray*} \mbox{D}(fg)=(\mbox{D} f)g+(\mbox{D} g)f\;\end{eqnarray*}
Note that because of this property the 1-forms
\begin{eqnarray} \mbox{D}\xi^i:=\d\xi^i-{1\over2}\d t \Gamma^i\;,\label{4.28} \end{eqnarray}
transform right-covariantly under a change of coordinates $\xi'{}^j
=\xi'{}^j(\xi)$, i.e.\ we have $\mbox{D}'\xi'{}^j=\mbox{D}\xi^i\,(\partial_i\xi'{}^j)$.
Here we have set $\Gamma^i:=b^{jk}\Gamma^i{}_{jk}$. Clearly $\d t,\,\mbox{D}\xi^1,
\ldots,\mbox{D}\xi^{2n}$ form a basis of $\O^1$ with
\begin{eqnarray*} \mbox{D} f=\mbox{D}\xi^i(\partial_i f)+\d t(\partial_t f)\;,\end{eqnarray*}
and
\begin{eqnarray} \d f & = & \mbox{D} f - {1\over2}\d t \,b^{ij}\nabla_i\partial_j f\;,\nonumber\\
& = & \mbox{D}\xi^i(\partial_i f)+\d\xi^i (\tilde{\partial}_t f)\;, \label{4.29}\end{eqnarray}
where
\begin{eqnarray} \tilde{\partial}_t f:=\partial_t f -{1\over 2} b^{ij}\nabla_i\partial_j f\;.\label{4.30}\end{eqnarray}
The vector fields $\tilde{\partial}_t,\,\partial_1,\ldots,\partial_{2n}$ form a basis of ${\cal X}$ dual
to the above basis of $\O^1$. From the definitions we find
\begin{eqnarray}
\mbox{D}\xi^i\mbox{D}\xi^j=\d\xi^i\d\xi^j+{1\over2}\d t \d\xi^{[i}\Gamma^{j]}\;,
\label{4.31}\end{eqnarray}
\begin{eqnarray}
\mbox{D}\xi^i\wedge\mbox{D}\xi^j=\mbox{D}\xi^i\mbox{D}\xi^j+\d t\d\xi^k\,
b^{\ell j}\Gamma^i{}_{k\ell}\;,\label{4.32}
\end{eqnarray}
and using (\ref{4.14}) we obtain
\begin{eqnarray} \begin{array}{c}
\mbox{D}\xi^i\wedge\mbox{D}\xi^j+\mbox{D}\xi^j\wedge\mbox{D}\xi^i=0\;,\\
\d t\wedge\mbox{D}\xi^i=\d t\d\xi^i\;,\qquad \mbox{D}\xi^i\wedge\d t=\d\xi^i\d t\;,
\qquad \d t\wedge \d t =\d t\d t=0\;.\end{array}\label{4.33}
\end{eqnarray}
It is now easy to see using (\ref{4.26}) and (\ref{4.33}), that $\wedge$ is
antisymmetric. With the aid of the curvature of the connection $\Gamma$,
\begin{eqnarray*}
R^i{}_{jk\ell}:=\partial_k\Gamma^i{}_{j\ell}-\partial_\ell\Gamma^i{}_{jk}
+\Gamma^i{}_{mk}\Gamma^m{}_{j\ell}-\Gamma^i{}_{m\ell}\Gamma^m{}_{jk}\;,
\end{eqnarray*}
and the curvature 2-form
\begin{eqnarray*} \O^i{}_j:={1\over2} R^i{}_{jk\ell}\d\xi^k\d\xi^\ell\pmod{\d t}\;, \end{eqnarray*}
we can prove the following usefull formulae
\begin{eqnarray*}
D_{[i}D_{j]}\phi=-R^k{}_{\ell ij}\d\xi^\ell(\partial_k\ins\phi)\pmod{\d t}\;,
\end{eqnarray*}
\begin{eqnarray*}
\d D_i\phi-D_i\d\phi=-\O^j{}_i(\partial_j\ins\phi)+
\d\xi^j\Gamma^k{}_{ij}D_k\phi\pmod{\d t}\;,
\end{eqnarray*}
and
\begin{eqnarray*}
\d(\d\xi^i\bullet\phi)-\d\xi^i\bullet\d\phi=-\d t\,b^{ij}D_j\phi-
\d\xi^j\Gamma^i{}_{jk}(\d\xi^k\bullet\phi)\;.
\end{eqnarray*}
Using these one can prove (see also \cite{28} \S7) for $\phi\in\O^r$ and
$\psi\in\O$
\begin{eqnarray}
\d(\phi\wedge\psi)=(\d\phi)\wedge\psi+(-1)^r\phi\wedge(\d\psi)+
\O^j{}_i(\partial_j\ins\phi)(\d\xi^i\bullet\psi)+
\d t\,b^{ij}(D_i\phi)(D_j\psi)\;.\label{4.34}
\end{eqnarray}
A special case of this formula is
\begin{eqnarray}
\d(\phi f)=(\d\phi)f+(-1)^r\phi\wedge\d f+\d t\,b^{ij}(D_i\phi)(\partial_j f)\;.
\label{4.35}\end{eqnarray}
For a 1-form $\alpha=\mbox{D}\xi^i\alpha_i+\d t \alpha_t$ we find using (\ref{4.34})
\begin{eqnarray*}
\d\alpha={1\over2}\mbox{D}\xi^i\wedge\mbox{D}\xi^j\,\partial_{[i}\alpha_{j]}+\d t\d\xi^i
\left[\partial_t\alpha_i-\partial_i\alpha_t-{1\over2}b^{jk}(\nabla_j\nabla_k\alpha_i
+R^\ell{}_{jki}\alpha_\ell)\right]\;.
\end{eqnarray*}
To connect the above results to symplectic geometry and Hamiltonian dynamics
we need the exterior derivative of a 2-form $\omega$. In general
\begin{eqnarray*}
\omega={1\over2}\mbox{D}\xi^i\wedge\mbox{D}\xi^j\,\omega_{ij}+\d t\d\xi^i\omega_i\;,
\end{eqnarray*}
(see (\ref{4.33}) and the remark before (\ref{4.29})). Then a lengthy
calculation gives
\begin{eqnarray}
\d\omega &=& {1\over3!}\mbox{D}\xi^i\wedge\mbox{D}\xi^j\wedge\mbox{D}\xi^k\,\left[{1\over2}\partial_{[i}
\omega_{ik]}\right]+ \nonumber\\
&&\!\!{1\over2}\d t\d\xi^i\d\xi^j\left[\partial_t\omega_{ij}-
\partial_{[i}\omega_{j]}-{1\over2}b^{k\ell}(\nabla_k\nabla_\ell\omega_{ij}-
R^m{}_{k\ell[i}\omega_{j]m}-R^m{}_{kij}\omega_{\ell m})\right]\:.
\label{4.36}
\end{eqnarray}
Therefore a 2-form $\omega$ is closed, if
\begin{eqnarray} \partial_{[i}\omega_{jk]}=0\;, \label{4.37} \end{eqnarray}
and
\begin{eqnarray}
\partial_t\omega_{ij}-\partial_{[i}\omega_{j]}-{1\over2}b^{k\ell}
(\nabla_k\nabla_\ell\omega_{ij}-R^m{}_{k\ell[i}\omega_{j]m}-
R^m{}_{kij}\omega_{\ell m})=0\;.\label{4.38}
\end{eqnarray}
If we now assume as in ordinary symplectic mechanics, that
$\partial_t\omega_{ij}=0$ then it can be proved that (\ref{4.37}), (\ref{4.38})
imply that $\omega_i$ satisfies
\begin{eqnarray}
\partial_{[i}(\omega_{j]}+{1\over2}b^{k\ell}\nabla_{|k}\omega_{\ell|j]})=0\;,
\label{4.39}\end{eqnarray}
where only $i,j$ are antisymmetrized. Therefore $\omega$ takes the form
\begin{eqnarray*}
\omega={1\over2}\mbox{D}\xi^i\wedge\mbox{D}\xi^j\,\omega_{ij}
+\d t\d\xi^i(\partial_i H-{1\over2}b^{k\ell}\nabla_{ k}\omega_{\ell i})\;,
\end{eqnarray*}
for some function $H$. On the other hand (\ref{4.37}) is the condition for
$\omega_{ij}$ to be closed in the ordinary exterior calculus, hence by
Darboux's theorem it can be brought locally to the form
\begin{eqnarray*}
\omega_{ij}=J_{ij}\;,\qquad
J=(J_{ij})=\left(\begin{array}{cc} 0 & I \\ -I & 0 \end{array}\right)\;,
\qquad I=(\delta_{\alpha\beta})\;,\;\; \alpha,\beta=1,\ldots,n\;,
\end{eqnarray*}
where $J$ is the symplectic form in canonical coordinates. According to the
interpretation of $\xi^i$ in section 5 this can also be assumed in the present
context. Therefore, to give our results a more familiar form we write
$\omega$ in this form
\begin{eqnarray}
\omega={1\over2}\mbox{D}\xi^i\wedge\mbox{D}\xi^j\,J_{ij}
+\d t\d\xi^i(\partial_i H-{1\over2}b^{k\ell}\nabla_{ k}J_{\ell i})\;.\label{4.40}
\end{eqnarray}
However the subsequent calculations do not make any use of the fact that
$\xi^i$ are taken here to be canonical coordinates.
We already used the ``insert'' operator $\ins$ in the sense of the classical
differential calculus. It is necessary to extend its definition in the present
context, since it is needed in the calculation of Hamiltonian vector fields.
For $X\in{\cal X}$, $f\in{\cal A}$ it is natural to put $X\ins f=0$ and as in the
ordinary exterior calculus we set
\begin{eqnarray} \begin{array}{c}
X\ins\alpha := \< X,\alpha\>\;,\\
X\ins(\phi\wedge\psi) := (X\ins\phi)\wedge\psi+(-1)^r\phi\wedge(X\ins\psi)\;,
\end{array}\label{4.41}
\end{eqnarray}
for $\alpha\in\O^1$, $\phi\in\O^r$ and $\psi\in\O$.
If $X=X^i\partial_i+X^t\tilde{\partial}_t$ in the coordinate system in which $\omega$ has the
form (\ref{4.40}), we have $X\ins\mbox{D}\xi^i=X^i$ and $X\ins\d t=X^t$.
Therefore we find that
\begin{eqnarray}
X\ins\omega=\mbox{D}\xi^i\left[-J_{ij}X^j+X^t(\partial_i H+F_i)\right]+
\d t X^i(\partial_i H +F_i)\;,\label{4.42}
\end{eqnarray}
\begin{eqnarray} F_i:=-(1/2)b^{jk}\nabla_j J_{ki}\;. \label{4.43}\end{eqnarray}
In order to write $\omega$ in the form (\ref{4.40}), it is necessary
that it has maximal rank, which is $2n$ since $M\times{\Ibb R}$ is
odd-dimensional. Therefore it has a 1-dimensional kernel given by the relation
\begin{eqnarray} X\ins\omega=0\;.\label{4.44} \end{eqnarray}
In ordinary extended Hamiltonian mechanics $H$ is the
Hamiltonian\footnote{Notice that by (\ref{4.31}), (\ref{4.32}),
eq.(\ref{4.40}) is $\omega=(1/2)\d\xi^i\d\xi^j J_{ij}+\d t \d\xi^i\partial_i H$,
strongly reminding conventional extended Hamiltonian dynamics.} and
the kernel is identified by definition with the space of Hamiltonian
vector fields $X$.
Equations (\ref{4.42}), (\ref{4.44}) give
\begin{eqnarray}
J_{ij} X^j=X^t(\partial_i H+F_i)\;,\qquad\qquad X^i(\partial_i H+F_i)=0\;. \label{4.45}
\end{eqnarray}
Setting $J^{ij}:=J_{ij}$ we have $J^{ik}J_{jk}=\delta^i_j$ hence the first
equation gives
\begin{eqnarray*} X^i=-X^t J^{ij}(\partial_jH+F_j)\;, \end{eqnarray*}
and consequently the 2nd equation is identically satisfied. Therefore the
Hamiltonian vector field defined by $H$ is given by
\begin{eqnarray}
X=X^t\left( (\partial_iH+F_i)J^{ij}\partial_j+(\partial_t-{1\over2}b^{ij}\nabla_i\partial_j)\right)
\;. \label{4.46}
\end{eqnarray}
As in ordinary extended Hamiltonian dynamics, the equation of motion for an
observable $A$, i.e.\ for $A\in{\cal A}$ takes the form $X\,A=0$, which gives
\begin{eqnarray}
\partial_t A=- \left[ \{H,A\}+F_i J^{ij}\partial_j A \right]+{1\over2}
\partial_i(b^{ij}\partial_j A)+{1\over2}b^{ij}\Gamma^k{}_{ki}\partial_j A\;.\label{4.47}
\end{eqnarray}
The {\em Hamiltonian equation} is now identical with the general kinetic
equation (\ref{2.5}) for a classical open system in interaction with a
large bath at canonical equilibrium, provided that
\newcounter{saveeqn}
\setcounter{saveeqn}{\value{equation}}
\stepcounter{saveeqn}
\setcounter{equation}{0}
\renewcommand{\theequation} {\arabic{section}.\arabic{saveeqn}\alph{equation}}
\begin{eqnarray}
\alpha^{ij} & = & {1\over2}b^{ij}\;, \label{4.48a}\\
F_i & = & \partial_i F\;, \label{4.48b} \\
\Gamma^k{}_{ki} & = & -\beta\partial_i H\;, \label{4.48c}
\end{eqnarray}
\setcounter{equation}{\value{saveeqn}}
\renewcommand{\theequation} {\arabic{section}.\arabic{equation}}
with $\{F,H\}=0$.
Condition (\ref{4.48b}) is equivalent to $\partial_{[i}F_{j]}=0$ which by
(\ref{4.38}), (\ref{4.39}), (\ref{4.43}) takes the form
\begin{eqnarray}
b^{k\ell}(\nabla_k\nabla_\ell J_{ij}-R^m{}_{k\ell[i}J_{j]m}-
R^m{}_{kij}J_{\ell m})=0\;.\label{4.49}
\end{eqnarray}
When $b^{ij}$ is nondegenerate, this is equivalent to the condition that the
{\em symplectic form $J_{ij}$ is harmonic with respect to the Laplace-Beltrami
operator of $b$ } (see e.g.\ \cite{34} p.3). Equation (\ref{4.48c}) is
equivalent to the condition that the canonical measure
\begin{eqnarray}\epsilon=e^{-\beta H}\d\xi^1\cdots\d\xi^{2n}\pmod{\d t}\;,\label{4.50}\end{eqnarray}
is covariantly constant $D_i\epsilon=0$ (see e.g.\ \cite{35} p.215).
The above results can now be summarized by saying that if the phase-space
of a classical open system $\Sigma$ is endowed with a noncommutative
geometrical structure (\ref{4.4}), (\ref{4.5}), because of its interaction
with a bath at canonical equilibrium, then the corresponding {\em Hamiltonian}
evolution of observables is identical to that given by conventional kinetic
theory {\em if} the symplectic form is ``harmonic'' with respect to the
``metric'' connection defined by $b^{ij}$, and the canonical
(Maxwell-Boltzmann) measure defined on $\Sigma$ at the bath temperature is
covariantly constant. Condition (\ref{2.3}) is not expected to follow from
the procedure followed so far, unless a precise relation of (\ref{4.4}),
(\ref{4.5}) to conventional dynamics is somehow made plausible. This will
be considered in another paper.
\section{Discussion}
\setcounter{equation}{0}
In the previous section we have given geometric conditions so that Hamiltonian
dynamics in the context of noncommutative differential calculus defined on
$M\times{\Ibb R}$ by (\ref{4.4})-(\ref{4.6}) can be interpreted physically as
kinetic theory of classical open systems interacting with a large bath at
canonical equilibrium.
As already remarked at the end of section 3 the presentation so far is
formal in the sense that the nature of the algebra ${\cal A}$ and its corresponding
coordinate representation in terms of $\xi^i$ has not been specified. Here we
discuss these questions further, but it should be emphasized that this is not
done rigorously. Actually much remains to be done for the complete
clarification of the problems addressed in this section.
At the beginning of section 4 we remarked that a 1-form $\alpha$ in the
universal differential envelope of the algebra of functions on a set $N$ is
a function $\alpha:N\times N\to{\ibb C}$ obtained by the obvious extension of
(\ref{4.1}), that is of
\begin{eqnarray} (f\~{\d}g\,h)(a,b)=f(a)[g(b)-g(a)] h(b)\;.\label{5.1}\end{eqnarray}
For the $\~{\bullet}$ we can show that (\ref{5.1}) implies
\begin{eqnarray}
(\~{\d}f_1\~{\bullet}\cdots\~{\bullet}\~{\d}f_r)(a,b)=
(f_1(b)-f_1(a))\cdots(f_r(b)-f_r(a))\;.\label{5.2}
\end{eqnarray}
Relations like $\alpha\~{\bullet}\beta-\~{\d}t\,\~{b}(\alpha,\beta)=0$
are in general incompatible with the above prescription of evaluating
differential forms. Therefore if we do impose such relations (i.e.\ pass
from $(\~{\O},\~{\d},\~{\bullet})$ to $(\O,\d,\bullet)$ as it is outlined at the
beginning of section 4) {\em and} at the same time we still want to
retain somehow an interpretation of $\bullet$, similar to that given by
(\ref{5.2}) then the elements of $\O^1$ cannot be functions on the whole of
$N\times N$. In fact such relations induce some structure on $N\times N$ by
grouping together points of $N$ which may be considered as neighbouring.
This is best illustrated by giving some examples.
Take $N:={\Ibb R}$ the real line. Let $x$ be the coordinate function on $N$ and
impose the relation $\d x\bullet\d x -\d x=0$ (cf.\ \cite{41}). If we want to
keep the interpretation of one forms as functions on some set $N_1$, this
cannot be the whole of $N\times N$. It is obvious that $N_1$ must be that
subset of $N\times N$ on which the imposed condition is satisfied identically.
If $(a,b)\in N_1 \subset N\times N$ then since $x(a)=a$ we find
\begin{eqnarray*}
0=(\d x\bullet\d x-\d x)(a,b)=(b-a)(b-a-1)\;.
\end{eqnarray*}
Hence in order for $(a,b)$ to be an element of $N_1$ either $b=a$ or
$b=a+1$. Hence $N_1=\{(a,a),(a,a+1)|\,a\in{\Ibb R}\}$. This set gives a structure
on the set $N$ by specifying which of its points are to be considered as
neighbouring. Obviously the above condition specifies a discrete structure
on ${\Ibb R}$. The possibility to evaluate a one form on $(a,b)$ with $a$ and
$b$ not neighbours is still given, if $b-a=m\inZ\!\!\!Z$, and corresponds to the
``integral''
\begin{eqnarray*} \int^b_a\alpha:=\sum_{k=1}^m \alpha(a+k-1,a+k)\;.\end{eqnarray*}
Applying the same reasoning to the relation $\d f\bullet\d g=0$ with
smooth functions $f,g:{\Ibb R}\to{\ibb C}$, then for $a,b\in{\Ibb R}$ we find
\begin{eqnarray*}
(\d f\bullet\d g)(a,b)=\left(f(b)-f(a)\right)\left(g(b)-g(a)\right)=
f'(x_1)g'(x_2)(b-a)^2=0\;,
\end{eqnarray*}
where we have used the mean value theorem. Since this condition must hold
for all smooth functions we must have $\epsilon:=b-a$ and $\epsilon^2=0$.
Now this relation gives something trivial since $\epsilon$ must identically
vanish. But the relation $\d f\bullet\d g=[\d f,g]=0$ holds in the usual
differential
calculus and consequently it cannot be trivial. In fact one may interprete
relation $\epsilon^2=0$ as saying, that $\epsilon$ is an infinitesimal
of first order. In this sense ${\Ibb R}$ is again structured since now
$N_1:=\{(a,a),(a,a+\epsilon)|\,a\in{\Ibb R}\}$. For arbitrary $a<b\in{\Ibb R}$ the
integral is defined by
\begin{eqnarray} \int^b_a\alpha := \lim\sum_{k=1}^m\alpha(x_{k-1},x_k)\;,\label{5.3}\end{eqnarray}
where this is obtained by taking the limit of vanishing width of the partition
$a=x_0<x_1<\cdots<x_m=b$ of $[a,b]$. Now the relation $\d f\bullet\d g=0$
integrated over $[a,b]$ for arbitrary $a<b\in{\Ibb R}$ gives
\begin{eqnarray}
0=\int^b_a \d f\bullet\d g=\lim\sum_{k=1}^m
\left(f(x_k)-f(x_{k-1})\right)\left(g(x_k)-g(x_{k-1})\right)\;,\label{5.4}
\end{eqnarray}
which can be expressed by saying that the ``quadratic variation''
of functions must vanish. This is true {\em if} $f,g$ are of bounded variation,
a condition which is {\em necessary} for the Riemann-Stieltjes integral
$\int^b_a g\d f$ to exist.
For a second order calculus on smooth functions of one variable $\xi$,
parametrizing $N:={\Ibb R}$, we have by definition (section 4, eq.(\ref{4.7}))
that $\d f\bullet\d g\bullet\d h=0$, hence by applying $\d\xi\bullet\d\xi\bullet\d\xi=0$ on
$(a,b)\in{\Ibb R}^2$ we get $(\xi(b)-\xi(a))^3=0$. Thus $\xi(b)-\xi(a)$ is an
infinitesimal of {\em second} order. Therefore $N_1=\{(a,a),(a,a+\epsilon),
(a,a+\epsilon^2)|\,a\in\xi^{-1}({\Ibb R})\}$, in this case and consequently for
given $a\in N$ we can move away from $a$ in two ways, either to $a+\epsilon$
or to $a+\epsilon^2$. In this sense then $N$ becomes structured and can be
considered as 2-dimensional.\footnote{The equality $N={\Ibb R}$ is only
set-theoretic.}
For $a<b\in{\Ibb R}$ we define formally an integral as in (\ref{5.2}). Applying
this on $\d\xi\bullet\d\xi\bullet\d\xi=0$ we obtain
\begin{eqnarray} \lim \sum_{k=1}^m\left(\xi(x_k)-\xi(x_{k-1})\right)^3=0\;,\label{5.5}\end{eqnarray}
where again the limit is obtained as in (\ref{5.3}). This relation can be
expressed by saying that the ``cubic variation'' of $\xi$ must vanish. It
is perhaps of independent mathematical interest to find equivalent
characterizations of such functions.
Since the quadratic variation of $\xi$ does not vanish in general, $\xi$
cannot be the usual coordinate function of ${\Ibb R}$. Consequently we have ${\Ibb R}$
as a differentiable manifold but with a differential structure which is
not the standard one. If we set $\d t:=(1/b)\d\xi\bullet\d\xi$ with some constant
$b$ and $t$ a function on ${\Ibb R}$, then by (\ref{2.5}) $\d t\bullet\d t=0$ and
hence $t$ is of bounded variation and can be taken to be the coordinate
function on some copy of ${\Ibb R}$. This additional coordinate $t$ realizes
somehow the fact that $N$ is 2-dimensional.
In the light of the above remark, if $M={\Ibb R}^{2n}$, it seems that $\xi^i$
should be interpreted as local coordinates defined on ${\Ibb R}^{2n}$ by an atlas
$\hat{\cal U}$ not compatible with the usual one giving the standard
differential structure of ${\Ibb R}^{2n}$; that is there are charts in
$\hat{\cal U}$ not $C^k$-related to the identity mapping of ${\Ibb R}^{2n}$,
for some $k>0$. Equivalently we may say that the identity mapping of
${\Ibb R}^{2n}$, does not belong to $\hat{\cal U}$.
More generally, for a $2n$-dimensional differentiable manifold $M$ the above
discussion implies that if $\xi$ is a local chart of $M$ in ${\Ibb R}^{2n}$ and
$\hat{\xi}$ a local chart of ${\Ibb R}^{2n}$ from $\hat{\cal U}$ the functions
$f:M\to{\ibb C}$ belonging to ${\cal A}$, are smooth functions of $\xi$ but are not
smooth, not even of bounded variation, as functions of $\hat{\xi}$,
that is $f\circ\xi^{-1}$ are smooth, but $f\circ(\hat{\xi}\circ\xi)^{-1}$
are not.
The whole discussion in this section reminds us strongly of stochastic
calculus on manifolds developped in the context of semimartingale theory,
in particular when stochastic terms are given in terms of Wiener processes
(see e.g.\ \cite{36}, \cite{37}). In fact there are many results in
stochastic calculus having an exact analogue in the formalism developped
here. As examples compare (\ref{4.8}) with (4) in \cite{37} p.134 and
their properties, or elements of ${\cal X}$ eq.(\ref{4.10}) with the
characterization of 2nd order fields in \cite{37} Lemma 6.1. Moreover
our relation of the ``metric connection'' with the drift term in the general
Fokker-Planck type kinetic equation (\ref{4.43}) suggests a close relation
with the interpretation in stochastic calculus of a connection on a manifold
as a mapping giving the ``drift'' of a 2nd order vector field
(\cite{36} p.258--259).In fact it seems possible --- and it will be examined
elsewhere --- that our present model of noncommutative geometry can be
realized in the context of stochastic calclulus. However, whether this is the
only possibility remains an interesting, but to our knowledge, still unsolved
problem.
\vskip.5cm
\noindent
{\bf Acknowledgment.} We would like to thank D.~Ellinas and
F.~M\"uller-Hoissen for stimulating discussions and critical
comments, and H.~Goenner for his interest in this work.
\small
|
\section{Introduction}
\begin{notation}
\label{notation}
\begin{enumerate}
\item In the present paper all cardinals are infinite so we will not repeat
this additional demand. Cardinals will be denoted by $\lambda$, $\mu$,
$\theta$ (with possible indexes) while ordinal numbers will be called
$\alpha$, $\beta$, $\zeta$, $\xi$, $\varepsilon$, $i$, $j$. Usually $\delta$
will stand for a limit ordinal (we may forget to repeat this assumption).
\item Sequences of ordinals will be called $\eta$, $\nu$, $\rho$ (with
possible indexes). For sequences $\eta_1,\eta_2$ their longest common initial
segment is denoted by $\eta_1\wedge\eta_2$. The length of the sequence $\eta$
is ${\rm lg}\/(\eta)$.
\item Ideals are supposed to be proper and contain all singletons. For a limit
ordinal $\delta$ the ideal of bounded subsets of $\delta$ is denoted by
$J^{{\rm bd}}_\delta$. If $I$ is an ideal on a set $X$ then $I^+$ is the family of
$I$-large sets, i.e.
\[a\in I^+\quad\mbox{ if and only if }\quad a\subseteq X\ \&\ a\notin I\]
and $I^c$ is the dual filter of sets with the complements in $I$.
\end{enumerate}
\end{notation}
\begin{notation}
\label{morenotation}
\begin{enumerate}
\item In a Boolean algebra we denote the Boolean operations by $\cap$ (and
$\bigcap$\/), $\cup$ (and $\bigcup$), $-$. The distinguished elements are
${\bf 0}$ and ${\bf 1}$. In the cases which may be confusing we will add indexes to
underline in which Boolean algebra the operation (the element) is considered,
but generally we will not do it.
\item For a Boolean algebra ${\Bbb B}$ and an element $x\in{\Bbb B}$ we denote:
\[x^0=x\quad\mbox{ and }\quad x^1=-x.\]
\item The free product of Boolean algebras ${\Bbb B}_1$, ${\Bbb B}_2$ is denoted by
${\Bbb B}_1*{\Bbb B}_2$. We will use $\bigstar$ to denote the free product of a family
of Boolean algebras. For $\bigstar_{i<\sigma} {\Bbb B}_i$ (wlog ${\Bbb B}_i$
pairwise disjoint) is the Boolean algebra generated by the formal
intersection $\bar b= \bigcap\limits_{i<\sigma} b_i$, $b_i\in {\Bbb B}$
freely except $\tau(\bar b^1, \ldots, \bar b^n)=0_{{\Bbb B}}$ iff some
$u\in \sigma$ finite non empty for simplicity
$$
\bigstar_{i\in u} {\Bbb B}_i \models \tau(\bigcap_{i\in u} b^1_i\ldots)\neq
0
$$
\end{enumerate}
\end{notation}
\begin{definition}
\begin{enumerate}
\item A Boolean algebra ${\Bbb B}$ satisfies the $\lambda$-cc if there is no
family $\F\subseteq{\Bbb B}^+\stackrel{\rm def}{=}{\Bbb B}\setminus\{{\bf 0}\}$ such
that $|\F|=\lambda$ and any two members of $\F$ are disjoint (i.e. their meet
in ${\Bbb B}$ is ${\bf 0}$).
\item The cellularity of the algebra ${\Bbb B}$ is
\[{\rm c}({\Bbb B})=\sup\{|\F|: \F\subseteq{\Bbb B}^+\ \&\ (\forall x,y\in\F)(x\neq
y\ \Rightarrow x\cap y={\bf 0})\},\]
\[{\rm c}^+({\Bbb B})=\sup\{|\F|^+: \F\subseteq{\Bbb B}^+\ \&\ (\forall x,y\in\F)(x\neq
y\ \Rightarrow x\cap y={\bf 0})\}.\]
\item For a topological space $(X,\tau)$:
\[\begin{array}{ll}
{\rm c}(X,\tau)=\sup\{|{\cal U}|:&{\cal U}\mbox{ is a family of pairwise disjoint}\\
\ &\mbox{nonempty open sets}\}.\\
\end{array}\]
\end{enumerate}
\end{definition}
The problem can be posed in each of the three ways ({\em $\lambda$-cc} is the
way of forcing, {\em the cellularity of Boolean algebras} is the approach of
Boolean algebraists, and {\em the cellularity of a topological space} is the
way of general topologists). It is well known that the three are equivalent,
though (1) makes the attainment problem more explicite. We use the second
approach.
A stronger property then $\lambda$-cc is the $\lambda$-Knaster property. This
property behaves nicely in free products -- it is productive. We will use it
in our construction.
\begin{definition}
\label{knaster}
A Boolean algebra ${\Bbb B}$ has the $\lambda$-Knaster property if for every
sequence $\langle z_\varepsilon: \varepsilon<\lambda\rangle\subseteq{\Bbb B}^+$
there is $A\in [\lambda]^{\textstyle \lambda}$ such that
\[\varepsilon_1,\varepsilon_2\in A\ \ \ \Rightarrow\ \ \ z_{\varepsilon_1}\cap
z_{\varepsilon_2}\neq{\bf 0}.\] \end{definition}
We are interested in the behaviour of the cellularity of Boolean algebras when
the free product of them is considered.
\begin{thema}
\label{thema}
When, for Boolean algebras ${\Bbb B}_1$, ${\Bbb B}_2$
\[{\rm c}^+({\Bbb B}_1)\leq\lambda_1\ \&\ {\rm c}^+({\Bbb B}_2)\leq\lambda_2\ \ \
\Rightarrow\ \ \ {\rm c}^+({\Bbb B}_1*{\Bbb B}_2)\leq\lambda_1+\lambda_2 ?\]
\end{thema}
There are a lot of results about it, particularly if $\lambda_1=\lambda_2$
(see \cite{Sh:g} or \cite{M1}, more \cite{Sh:572}). It is well know
that if
\[(\lambda_1^+ +\lambda_2^+)\longrightarrow (\lambda_1^+,\lambda_2^+)^2\]
then the answer is ``yes''. These are exactly the cases for which ``yes''
answer is known. Under {\bf GCH} the only problem which remained open was the
one presented below:
\begin{pws}
\label{pwa}
\ \\
{\bf (posed by D. Monk as Problem 1 in \cite{M1}, \cite{M2} under ${\rm GCH}$)}\\
Are there Boolean algebras ${\Bbb B}_1$, ${\Bbb B}_2$ and cardinals $\mu,\theta$ such
that
\begin{enumerate}
\item $\lambda_1=\mu$ is singular, $\mu>\lambda_2=\theta>\cf(\mu)$ and
\item ${\rm c}({\Bbb B}_1)=\mu$, ${\rm c}({\Bbb B}_2)\leq\theta$ but ${\rm c}({\Bbb B}_1*{\Bbb B}_2)>\mu$?
\end{enumerate}
\end{pws}
We will answer this question proving in particular the following result (see
\ref{main}):
\begin{quotation}
If {\em $\mu$ is a strong limit singular cardinal,
$\theta=(2^{\cf(\mu)})^+$, $2^\mu=\mu^+$}
then {\em there are Boolean algebras ${\Bbb B}_1,{\Bbb B}_2$ such that
${\rm c}({\Bbb B}_1)=\mu$,
${\rm c}({\Bbb B}_2)<\theta$ but ${\rm c}({\Bbb B}_1*{\Bbb B}_2)=\mu^+$.
}
\end{quotation}
Later we deal with better results by refining the method.
\begin{remark}
{\em
On products of many Boolean algebras and square bracket arrows see
\cite{Sh:345}, 1.2A, 1.3B.
If $\lambda\longrightarrow [\mu]^2$, $[\tau<\sigma\ \Rightarrow\
2^\tau<\theta]$, the cardinals $\theta,\sigma$ are possibly finite,
${\Bbb B}_i$ (for $i<\theta$) are Boolean algebras such that for each $j<\theta$
the free product $\mathop{\bigstar}\limits_{i\in\theta\setminus\{j\}}{\Bbb B}_i$
satisfies the $\mu$-cc {\em then} the algebra ${\Bbb B}=\mathop{\bigstar}
\limits_{i<\theta}{\Bbb B}_i$ satisfies the $\lambda$-cc.
}
\end{remark}
[Why? if $\langle a^\zeta_i: i<\theta\rangle\in
\prod\limits_{i<\theta}{{\Bbb B}}^+_i$ for $\zeta<\lambda$ such that for
every $\zeta<\xi<\lambda$ for some $i$, ${\Bbb B}_i \models$
``$a^\zeta_i\cap a^\xi_i={\bf 0}$'', let $\i=\i(\zeta, \xi)$, and we can
find $A\in [\lambda]^\mu$, $j<\theta$ such that for $\zeta<\xi$ from
$A$, $\i(\zeta, \xi)=j$, so $\langle a^\zeta_i: i<\theta, i\neq
i^*\rangle$ for $\zeta\in A$ exemplifies $\bigstar_{i\in
\theta\setminus\{i^*\}}{{\Bbb B}}_i$ fails the $\mu-{\rm c.c.}$. We can deal also
with ultraproducts and other products similarly.]
\section{Preliminaries: products of ideals}
\begin{notation}
\label{quantifiers}
For an ideal $J$ on $\delta$ the quantifier $(\forall^J i<\delta)$ means ``for
all $i<\delta$ except a set from the ideal'', i.e.
\[(\forall^J i<\delta)\varphi(i)\ \equiv\ \{i<\delta: \neg\varphi(i)\}\in J.\]
The dual quantifier $(\exists^J i<\delta)$ means ``for a $J$-positive set of
$i<\delta$''.
\end{notation}
\begin{proposition}
\label{prefub}
Assume that $\lambda^{0}>\lambda^{1}>\ldots>\lambda^{n-1}$ are cardinals,
$I^\ell$ are ideals on $\lambda^\ell$ (for $\ell<n$) and
$B\subseteq\prod\limits_{\ell<n}\lambda^\ell$. Further suppose that
\begin{description}
\item[$(\alpha)$] $(\exists^{I^{0}}\gamma_{0})\ldots (\exists^{I^{n-1}}
\gamma_{n-1})(\langle\gamma_\ell:\ell<n\rangle\in B)$
\item[$(\beta)$] the ideal $I^\ell$ is $(2^{\lambda^{\ell+1}})^+$-complete
(for $\ell+1<n$).
\end{description}
{\em Then} there are sets $X_\ell\subseteq\lambda^\ell$, $X_\ell\notin I^\ell$
such that $\prod\limits_{\ell<n} X_\ell\subseteq B$.
\noindent [Note that this translates the situation to arity 1; it is a kind of
polarized $(1,\ldots,1)$-partitions with ideals.]
\end{proposition}
\Proof We show it by induction on $n$. Define
\[\begin{array}{ll}
E_0\stackrel{\rm def}{=}&\{(\gamma',\gamma''): \gamma',\gamma''<\lambda^0
\quad\mbox{ and}\\
\ &\mbox{for all }\gamma_1<\lambda^1,\ldots,\gamma_{n-1}<\lambda^{n-1}\ \mbox{
we have}\\
\ &(\langle\gamma',\gamma_1,\ldots,\gamma_{n-1}\rangle\in B\ \ \
\Leftrightarrow\ \ \ \langle\gamma'',\gamma_1,\ldots,\gamma_{n-1}\rangle
\in B)\}.
\end{array}\]
Clearly $E_0$ is an equivalence relation on $\lambda^0$ with $\leq
2^{\prod_{0<m<n}\lambda^m}=2^{\lambda^1}$ equivalence classes. Hence the set
\[A_0\stackrel{\rm def}{=}\bigcup\{A: A\mbox{ is an }E_0\mbox{-equivalence
class}, A\in I^0\}\]
is in the ideal $I^0$. Let
\[A_0^*\stackrel{\rm def}{=}\{\gamma_0<\lambda^0:(\exists^{I^{1}}\gamma_{1})
\ldots (\exists^{I^{n-1}}\gamma_{n-1})(\langle\gamma_0,\gamma_1,\ldots,
\gamma_{n-1}\rangle\in B).\]
The assumption $(\alpha)$ implies that $A_0^*\notin I^0$ and hence we may
choose $\gamma^*_0\in A_0^*\setminus A_0$. Let
\[B_1\stackrel{\rm def}{=}\{\bar{\gamma}\in\prod_{k=1}^{n-1}\lambda^k:
\langle\gamma^*_0 \rangle\hat{\ }\bar{\gamma}\in B\}.\]
Since $\gamma^*_0\in A_0^*$ we are sure that
\[(\exists^{I^{1}}\gamma_{1})\ldots (\exists^{I^{n-1}}\gamma_{n-1})(\langle
\gamma_1,\ldots, \gamma_{n-1}\rangle\in B_1).\]
Hence we may apply the inductive hypothesis for $n-1$ and $B_1$ and we find
sets $X_1\in (I^1)^+,\ldots,X_{n-1}\in (I^{n-1})^+$ such that
$\prod\limits_{\ell=1}^{n-1}X_\ell\subseteq B_1$, so then
\[(\forall\gamma_1\in X_1)\ldots(\forall\gamma_{n-1}\in X_{n-1})(\langle
\gamma^*_0,\gamma_1,\ldots,\gamma_{n-1}\rangle\in B).\]
Take $X_0$ to be the $E_0$-equivalence class of $\gamma^*_0$ (so $X_0\in
(I^0)^+$ as $\gamma_0^*\notin A_0$). By the definition of the relation $E_0$
and the choice of the sets $X_\ell$ we have that for each $\gamma_0\in X_0$
\[(\forall\gamma_1\in X_1)\ldots(\forall\gamma_{n-1}\in X_{n-1})(\langle
\gamma_0,\gamma_1,\ldots,\gamma_{n-1}\rangle\in B)\]
what means that $\prod\limits_{\ell<n}X_\ell\subseteq B$. The proposition is
proved. \QED
\begin{proposition}
\label{weakfub}
Assume that $\lambda_0>\lambda_1>\ldots>\lambda_{n-1}\geq\sigma$ are
cardinals, $I_\ell$ are ideals on $\lambda_\ell$ (for $\ell<n$) and
$B\subseteq\prod\limits_{\ell<n}\lambda_\ell$. Further suppose that
\begin{description}
\item[$(\alpha)$] $(\exists^{I_0}\gamma_0)\ldots(\exists^{I_{n-1}}
\gamma_{n-1})(\langle\gamma_\ell:\ell<n\rangle\in B)$
\item[$(\beta)$] for each $\ell<n-1$ the ideal $I_\ell$ is
$((\lambda_{\ell+1})^\sigma)^+$-complete, $[\lambda_{n-1}]^{<\sigma}\subseteq
I_{n-1}$.
\end{description}
Then there are sets $X_\ell\in [\lambda_\ell]^\sigma$ such that
$\prod\limits_{\ell<n} X_\ell\subseteq B$.
\end{proposition}
\Proof The proof is by induction on $n$. If $n=1$ then there is nothing to do
as $I_{n-1}$ contains all subsets of $\lambda_{n-1}$ of size
$<\sigma$ and $\lambda_{n_1}\geq \sigma$ so every $A\in I^+_{n_1}$ has
cardinality $\geq \sigma$.
\noindent Let $n>1$ and let
\[a_0\stackrel{\rm def}{=}\{\gamma\in\lambda_0: (\exists^{I_1}\gamma_1)\ldots
(\exists^{I_{n-1}}\gamma_{n-1})(\langle\gamma,\gamma_1,\ldots,\gamma_{n-1}
\rangle\in B)\}.\]
By our assumptions we know that $a_0\in (I_0)^+$. For each $\gamma\in
a_0$ we may apply the inductive hypothesis to the set
\[B_\gamma\stackrel{\rm def}{=}\{\langle\gamma_1,\ldots,\gamma_{n-1}\rangle\in
\prod_{0<\ell<n}\lambda_\ell:\langle\gamma,\gamma_1,\ldots,\gamma_{n-1}
\rangle\in B\}\]
and we get sets $X^\gamma_1\in [\lambda_1]^\sigma,\ldots,X^\gamma_{n-1}\in
[\lambda_{n-1}]^\sigma$ such that
\[\prod\limits_{0<\ell<n}X^\gamma_\ell\subseteq B_\gamma.\]
There is at most $(\lambda_1)^\sigma$ possible sequences $\langle X^\gamma_1,
\ldots,X^\gamma_{n-1}\rangle$, the ideal $I_0$ is
$((\lambda_1)^\sigma)^+$--complete so for some sequence $\langle X_1,\ldots,
X_{n-1}\rangle$ and a set $a^*\subseteq a_0$, $a^*\in (I_0)^+$ we have
\[(\forall\gamma\in a^*)(X^\gamma_1=X_1\ \&\ \ldots\ \&\ X_{n-1}^\gamma=
X_{n-1}).\]
Choose $X_0\in [a^*]^\sigma$ (remember that $I_0$ contains singletons and it
is complete enough to make sure that $\sigma\leq |a^*|$). Clearly
$\prod\limits_{\ell<n} X_\ell\subseteq B$. \QED
\begin{remark}
{\em
We can use $\sigma_0\geq \sigma_1\geq \ldots\geq\sigma_{n-1}$, $I_\ell$
is $(\lambda^{\sigma_{\ell+1}}_{\ell+1})^+$-complete,
$[\lambda_\ell]^{<\sigma_\ell}\subseteq I_\ell$.
}
\end{remark}
\begin{proposition}
\label{fubini}
Assume that $n<\omega$ and $\lambda^m_\ell$, $\chi^m_\ell$, $P^m_\ell$,
$I^m_\ell$, $I^m$ and $B$ are such that for $\ell,m\leq n$:
\begin{description}
\item[$(\alpha)$] $I^m_\ell$ is a $\chi^m_\ell$-complete ideal on
$\lambda^m_\ell$ (for $\ell,m\leq n$)
\item[$(\beta)$] $P^m_\ell\subseteq\P(\lambda^m_\ell)$ is a family dense in
$(I^m_\ell)^+$ in the sense that:
\[(\forall X\in (I^m_\ell)^+)(\exists a\in P^m_\ell)(a\subseteq X)\]
\item[$(\gamma)$] $I^m=\{X\!\subseteq\!\prod\limits_{\ell\leq n}
\lambda^m_\ell: \neg(\exists^{I^m_0}\gamma_0)\ldots(\exists^{I^m_n}\gamma_n)
(\langle\gamma_0,\ldots,\gamma_n\rangle\in X)\}$
\noindent [thus $I^m$ is the ideal on $\prod\limits_{\ell\leq
n}\lambda^m_\ell$ such that the dual filter $(I^m)^c$ is the Fubini product of
filters $(I^m_0)^c,\ldots,(I^m_n)^c$]
\item[$(\delta)$] $\chi^m_{n-m}>\sum\limits_{\ell=m+1}^{n}(|P^\ell_{n-\ell}|+
\sum\limits_{k=0}^{n-\ell}\lambda^\ell_k)$
\item[$(\varepsilon)$] $B\subseteq\prod\limits_{m\leq n}\prod\limits_{\ell\leq
n}\lambda^m_\ell$ is a set satisfying
\[(\exists^{I^0}\eta_0)(\exists^{I^1}\eta_1)\ldots (\exists^{I^n}\eta_n)
(\langle\eta_0,\eta_1,\ldots,\eta_n\rangle\in B).\]
\end{description}
{\em Then} there are sets $X_0,\ldots,X_n$ such that for $m\leq n$:
\begin{description}
\item[(a)] $X_m\subseteq\prod\limits_{\ell\leq n-m}\lambda^m_\ell$
\item[(b)] if $\eta,\nu\in X_m$, $\eta\neq\nu$ then
\begin{description}
\item[(i)\ ] $\eta\rest (n-m)=\nu\rest (n-m)$
\item[(ii) ] $\eta(n-m)\neq\nu (n-m)$
\end{description}
\item[(c)] $\{\eta(n-m):\eta\in X_m\}\in P^m_{n-m}$\ \ \ \ \ \ \ \ \ \ and
\item[(d)] for each $\langle\eta_0,\ldots,\eta_n\rangle\in\prod\limits_{m\leq
n} X_m$ there is $\langle\eta^*_0,\ldots,\eta^*_n\rangle\in B$ such that
$(\forall m\leq n)(\eta_m\trianglelefteq\eta^*_m)$.
\end{description}
\end{proposition}
\noindent{\bf Remark}~\ref{fubini}.A:
\begin{enumerate}
\item Note that the sets $X_m$ in the assertion of \ref{fubini} may be thought
of as sets of the form $X_m=\{\nu_m\hat{\ }\langle\alpha\rangle: \alpha\in
a_m\}$ for some $\nu_m\in\prod\limits_{\ell<n-m}\lambda^m_\ell$ and $a_m\in
P^m_{n-m}$.
\item We will apply this proposition with
$\lambda^m_\ell=\lambda_\ell$, $I^m_\ell=I_\ell$ and\\
$\lambda_\ell>\chi_\ell>\sum_{k<\ell}\lambda_k$.
\item In the assumption $(\delta)$ of \ref{fubini} we may have that the last
sum on the right hand side of the inequality ranges from $k=0$ to $n-\ell-1$.
We did not formulate that assumption in this way as with $n-\ell$ there it is
easier to handle the induction step and this change is not important for our
applications.
\item In the assertion {\bf (d)} of \ref{fubini} we can make $\eta^*_\ell$
depending on $\langle\eta_0,\ldots,\eta_\ell\rangle$ only.
\end{enumerate}
\medskip
\Proof The proof is by induction on $n$. For $n=0$ there is nothing to do. Let
us describe the induction step.
Suppose $0<n<\omega$ and $\lambda^m_\ell$, $\chi^m_\ell$, $P^m_\ell$,
$I^m_\ell$, $I^m$ (for $\ell,m\leq n$) and $B$ satisfy the assumptions
$(\alpha)$--$(\varepsilon)$. Let
\[B^*\stackrel{\rm def}{=}\{\langle\eta_0,\eta_1\rest n,\ldots,\eta_n\rest
n\rangle: \eta_m\in\prod_{\ell\leq n}\lambda^m_\ell\mbox{ (for $m\leq n$) and
} \langle\eta_0,\eta_1,\ldots,\eta_n\rangle\in B\}\]
and for $\eta_0\in\prod\limits_{\ell\leq n}\lambda^0_\ell$ let
\[B^*_{\eta_0}\stackrel{\rm def}{=}\{\langle\nu_1,\ldots,\nu_n\rangle\in
\prod_{m=1}^n\prod_{\ell=0}^{n-1}\lambda^m_\ell: \langle\eta_0,\nu_1,\ldots,
\nu_n\rangle\in B^*\}.\]
Let $J^m$ (for $1\leq m\leq n$) be the ideal on $\prod\limits_{\ell=0}^{n-1}
\lambda^m_\ell$ coming from the ideals $I^m_\ell$, i.e. a set $X\subseteq
\prod\limits_{\ell<n}\lambda^m_\ell$ is in $J^m$ if and only if
\[\neg (\exists^{I^m_0}\gamma_0)\ldots(\exists^{I^m_{n-1}}\gamma_{n-1})
(\langle\gamma_0,\ldots,\gamma_{n-1}\rangle\in X).\]
Let us call the set $B^*_{\eta_0}$ {\em big} if
\[(\exists^{J^1}\nu_1)\ldots(\exists^{J^n}\nu_n)(\langle\nu_1,\ldots,\nu_n
\rangle\in B^*_{\eta_0}).\]
We may write more explicitely what the bigness means: the above condition is
equivalent to
\[\begin{array}{ll}
\ &(\exists^{I^1_0}\gamma^1_0)\ldots (\exists^{I^1_{n-1}}\gamma^1_{n-1})
\ldots\\
\ldots &(\exists^{I^n_0}\gamma^n_0)\ldots (\exists^{I^n_{n-1}}\gamma^n_{n-1})
(\langle\langle\gamma^1_0,\ldots,\gamma^1_{n-1}\rangle,\ldots \langle
\gamma^n_0,\ldots,\gamma^n_{n-1}\rangle \rangle\in B^*_{\eta_0})\\
\end{array}\]
which means
\[\begin{array}{ll}
\ &(\exists^{I^1_0}\gamma^1_0)\ldots\ldots(\exists^{I^n_{n-1}}\gamma^n_{n-1})\\
\ &(\exists\gamma^1_n)\ldots(\exists\gamma^n_n)(\langle\eta_0,\langle
\gamma^1_0,\ldots,\gamma^1_n\rangle,\ldots,\langle\gamma^n_0,\ldots,
\gamma^n_n\rangle\rangle\in B).\\
\end{array}\]
By the assumptions $(\gamma)$ and $(\varepsilon)$ we know that
\[\begin{array}{ll}
\ &(\exists^{I^0_0}\gamma^0_0)\ldots(\exists^{I^0_n}\gamma^0_n)
(\exists^{I^1_0}\gamma^1_0)\ldots(\exists^{I^1_n}\gamma^1_n)\ldots\\
\ldots &(\exists^{I^n_0}\gamma^n_0)\ldots(\exists^{I^n_{n}}\gamma^n_{n})
(\langle\langle\gamma_0^0,\ldots,\gamma^0_n\rangle,\langle
\gamma^1_0,\ldots,\gamma^1_n\rangle,\ldots,\langle\gamma^n_0,\ldots,
\gamma^n_n\rangle\rangle\in B).\\
\end{array}\]
Obviously any quantifier $(\exists^{I^m_\ell}\gamma^m_\ell)$ above may be
replaced by $(\exists\gamma^m_\ell)$ and then ``moved'' right as much as we
want.
Consequently we get
\[\begin{array}{l}
(\exists\gamma^0_0)\ldots(\exists\gamma^0_{n-1})(\exists^{I^0_n}
\gamma^0_n)(\exists^{I^1_0}\gamma^1_0)\ldots(\exists^{I^1_{n-1}}
\gamma^1_{n-1})\ldots\ldots(\exists^{I^n_0}\gamma^n_0)\ldots
(\exists^{I^n_{n-1}}\gamma^n_{n-1})\\
(\exists\gamma^1_n)\ldots(\exists\gamma^n_n)(\langle\langle\gamma_0^0,
\ldots,\gamma^0_n\rangle,\langle\gamma^1_0,\ldots,\gamma^1_n\rangle,\ldots,
\langle\gamma^n_0,\ldots,\gamma^n_n\rangle\rangle\in B)\\
\end{array}\]
which means that
\[(\exists\gamma^0_0)\ldots(\exists\gamma^0_{n-1})(\exists^{I^0_n}
\gamma^0_n)(B^*_{\langle\gamma^0_0,\ldots,\gamma^n_n\rangle}\mbox{ is big}).\]
Hence we find $\gamma^0_0,\ldots,\gamma^0_{n-1}$ and a set $a\in (I^0_n)^+$
such that
\[(\forall\gamma\in a)(B^*_{\langle\gamma^0_0,\ldots,\gamma^n_n\rangle}\mbox{
is big}).\]
Note that the assumptions of the proposition are such that if we know that
$B^*_{\eta_0}$ is big then we may apply the inductive hypothesis to
\[\lambda^m_\ell,\chi^m_\ell,P^m_\ell,I^m_\ell, J^m\ \mbox{ (for $1\leq m\leq
n$, $\ell\leq n-1$) and } B^*_{\eta_0}.\]
Consequently for each $\gamma\in a$ we find sets $X^\gamma_1,\ldots,
X^\gamma_n$ such that for $1\leq m\leq n$:
\begin{description}
\item[(a)$^*$] $X^\gamma_m\subseteq\prod\limits_{\ell\leq n-m}\lambda^m_\ell$
\item[(b)$^*$] if $\eta,\nu\in X^\gamma_m$, $\eta\neq\nu$ then
\begin{description}
\item[(i)\ ] $\eta\rest (n-m)=\nu\rest (n-m)$
\item[(ii) ] $\eta(n-m)\neq\nu (n-m)$
\end{description}
\item[(c)$^*$] $\{\eta(n-m):\eta\in X^\gamma_m\}\in P^m_{n-m}$\ \ \ \ \ \ \ \
\ \ and
\item[(d)$^*$] for all $\langle\eta_0,\ldots,\eta_n\rangle\in\prod
\limits_{m\leq n} X^\gamma_m$ we have
\[(\exists\langle\eta^*_0,\ldots,\eta^*_n\rangle
\in B^*_{\langle\gamma^0_0,\ldots,\gamma^0_{n-1},\gamma\rangle})(\forall
1\leq m\leq n)(\nu_m\trianglelefteq\nu^*_m).\]
\end{description}
Now we may ask how mane possibilities for $X^\gamma_m$ do we have: not too
many. If we fix the common initial segment (see {\bf (b)$^*$}) the only
freedom we have is in choosing an element of $P^m_{n-m}$ (see {\bf (c)$^*$}).
Consequently there are at most $|P^m_{n-m}|+\sum\limits_{\ell\leq n-m}
\lambda^m_\ell$ possible values for $X^\gamma_m$ and hence there are at most
\[\sum_{m=1}^n(|P^m_{n-m}|+\sum_{\ell\leq n-m}\lambda^m_\ell)<\chi^0_n\]
possible values for the sequence $\langle
X^\gamma_1,\ldots,X^\gamma_n\rangle$. Since the ideal $I^0_n$ is
$\chi^0_n$-complete we find a sequence $\langle X_1,\ldots,X_n\rangle$ and a
set $b\subseteq a$, $b\in (I^0_n)^+$ such that
\[(\forall\gamma\in b)(\langle X^\gamma_1,\ldots,X^\gamma_n\rangle = \langle
X_1,\ldots, X_n\rangle).\]
Next choose $b^0_n\in P^0_n$ such that $b^0_n\subseteq b$ and put
\[X_0=\{\langle\gamma^0_0,\ldots,\gamma^0_{n-1},\gamma\rangle: \gamma\in
b^0_n\}.\]
Now it is a routine to check that the sets $X_0,X_1,\ldots,X_n$ are as
required (i.e. they satisfy clauses {\bf (a)}--{\bf (d)}). \QED
\section{Cofinal sequences in trees}
\begin{notation}
\label{evenmorenot}
\begin{enumerate}
\item For a tree $T\subseteq {}^{\delta{>}}\mu$ the set of $\delta$-branches
through $T$ is
\[{\lim}_\delta(T)\stackrel{\rm def}{=}\{\eta\in{}^\delta\mu: (\forall\alpha
<\delta)(\eta\rest\alpha\in T)\}.\]
The $i$-th level (for $i<\delta$) of the tree $T$ is
\[T_i\stackrel{\rm def}{=} T\cap {}^i\mu\]
and $T_{{<}i}\stackrel{\rm def}{=}\bigcup\limits_{j<i}T_j$.
If $\eta\in T$ then the set of immediate successors of $\eta$ in $T$ is
\[\suc_T\stackrel{\rm def}{=}\{\nu\in T: \eta\vartriangleleft\nu\ \&\
{\rm lg}\/(\nu)={\rm lg}\/(\eta)+1\}.\]
We shall not distinguish strictly between $\suc_T(\eta)$ and $\{\alpha:
\eta\hat{\ }\langle\alpha\rangle \in T\}$.
\end{enumerate}
\end{notation}
\begin{definition}
\label{cofinal}
\begin{enumerate}
\item ${\cal K}_{\mu,\delta}$ is the family of all pairs $(T,\bar{\lambda})$ such
that $T\subseteq{}^{\delta{>}}\mu$ is a tree with $\delta$ levels and
$\bar{\lambda}=\langle\lambda_\eta:\eta\in T\rangle$ is a sequence of
cardinals such that for each $\eta\in T$ we have $\suc_T(\eta)=\lambda_\eta$
(compare the previous remark about not distinguishing $\suc_T(\eta)$ and
$\{\alpha:\eta\hat{\ }\langle\alpha\rangle\in T\}$).
\item For a limit ordinal $\delta$ and a cardinal $\mu$ we let
\[\begin{array}{ll}
{\cal K}^{{\rm id}}_{\mu,\delta}\stackrel{\rm def}{=}\{(T,\bar{\lambda},\bar{I}):
&(T,\bar{\lambda})\in {\cal K}_{\mu,\delta}, \ \bar{I}=\langle I_\eta:\eta\in
T\rangle\\
\ &\mbox{each }I_\eta\mbox{ is an ideal on }\lambda_\eta=\suc_T(\eta)\}.\\
\end{array}\]
Let $(T,\bar{\lambda},\bar{I})\in{\cal K}_{\mu,\delta}^{{\rm id}}$ and let $J$ be an
ideal on $\delta$ (including $J^{{\rm bd}}_\delta$ if we do not say otherwise).
Further let
$\bar{\eta}=\langle\eta_\alpha:\alpha<\lambda\rangle\subseteq{\lim}_\delta(T)$
be a sequence of $\delta$-branches through $T$.
\item We say that $\bar{\eta}$ is {\em $J$-cofinal in} $(T,\bar{\lambda},
\bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every sequence $\bar{A}{=}\langle A_\eta\!:\eta\in T\rangle
\in\!\!\prod\limits_{\eta\in T} I_\eta$ there is $\alpha^*{<}\lambda$ such
that
\[\alpha^*\leq\alpha<\lambda\ \ \ \Rightarrow\ \ \ (\forall^J i<\delta)
(\eta_\alpha\rest (i+1)\notin A_{\eta_\alpha\rest i}).\]
\end{description}
\item If $I$ is an ideal on $\lambda$ then we say that
$(\bar{\eta},I)$ is {\em a $J$-cofinal pair} for $(T,\bar{\lambda},
\bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every sequence $\bar{A}=\langle A_\eta:\eta\in T\rangle\in
\prod\limits_{\eta\in T} I_\eta$ there is $A\in I$ such that
\[\alpha\in\lambda\setminus A\ \ \ \Rightarrow\ \ \ (\forall^J i<\delta)
(\eta_\alpha\rest (i+1)\notin A_{\eta_\alpha\rest i}).\]
\end{description}
\item The sequence $\bar{\eta}$ is {\em strongly $J$-cofinal in}
$(T,\bar{\lambda}, \bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every $n<\omega$ and functions $F_0,\ldots,F_{n}$ there is
$\alpha^*<\lambda$ such that
\noindent {\em if} $m\leq n$, $\alpha_0<\ldots<\alpha_n<\lambda$,
$\alpha^*\leq\alpha_m$
\noindent {\em then} the set
\[\hspace{-1.3cm}\begin{array}{ll}
\{i<\delta:&\mbox{(i)\ \ } (\forall \ell<m)(\lambda_{\eta_{\alpha_\ell}\rest
i}< \lambda_{\eta_{\alpha_m}\rest i})\mbox{ and}\\
\ &\mbox{(ii) }F_m(\eta_{\alpha_0}\rest (i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i ,\ldots,
\eta_{\alpha_n}\rest i)\in I_{\eta_{\alpha_m}\rest i}\\
\ &\ \ \mbox{(and well defined) but }\\
\ &\ \ \eta_{\alpha_m}\rest(i{+}1)\in F_m(\eta_{\alpha_0}\rest
(i{+}1),\ldots, \eta_{\alpha_{m-1}}\rest (i{+}1), \eta_{\alpha_m}\rest
i,\ldots, \eta_{\alpha_n}\rest i)\}\\
\end{array}\]
is in the ideal $J$.
\end{description}
[Note: in {\bf (b)} above we may have $\alpha^*<\alpha_0$, this causes no real
change.]
\item The sequence $\bar{\eta}$ is {\em stronger $J$-cofinal in}
$(T,\bar{\lambda}, \bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every $n<\omega$ and functions $F_0,\ldots,F_{n}$ there
is $\alpha^*<\lambda$ such that
\noindent {\em if} $m\leq n$, $\alpha_0<\ldots<\alpha_n<\lambda$,
$\alpha^*\leq\alpha_m$
\noindent {\em then} the set
\[\hspace{-1.3cm}\begin{array}{ll}
\{i<\delta:&\mbox{(ii) }F_m(\eta_{\alpha_0}\rest (i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i ,\ldots,
\eta_{\alpha_n}\rest i)\in I_{\eta_{\alpha_m}\rest i}\\
\ &\ \ \mbox{(and well defined) but }\\
\ &\ \ \eta_{\alpha_m}\rest(i{+}1)\in F_m(\eta_{\alpha_0}\rest
(i{+}1),\ldots, \eta_{\alpha_{m-1}}\rest (i{+}1), \eta_{\alpha_m}\rest
i,\ldots,\eta_{\alpha_n}\rest i)\}
\end{array}\]
is in the ideal $J$.
\end{description}
\item The sequence $\bar{\eta}$ is {\em strongest $J$-cofinal in}
$(T,\bar{\lambda}, \bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every $n<\omega$ and functions $F_0,\dots,F_n$ there is
$\alpha^*<\lambda$ such that
\noindent {\em if} $m\leq n$, $\alpha_0<\ldots<\alpha_n<\lambda$,
$\alpha^*\leq\alpha_m$
\noindent {\em then} the set
\[\hspace{-1.8cm}\begin{array}{ll}
\{i<\delta:&\mbox{(i')\ \ } (\exists \ell<m)(\lambda_{\eta_{\alpha_\ell}\rest
i}\geq \lambda_{\eta_{\alpha_m}\rest i})\mbox{ or}\\
\ &\mbox{(ii) }F_m(\eta_{\alpha_0}\rest (i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i ,\ldots,
\eta_{\alpha_n}\rest i)\in I_{\eta_{\alpha_m}\rest i}\\
\ &\ \ \mbox{(and well defined) but }\\
\ &\ \ \eta_{\alpha_m}\rest(i{+}1)\in F_m(\eta_{\alpha_0}\rest
(i{+}1),\ldots, \eta_{\alpha_{m-1}}\rest (i{+}1), \eta_{\alpha_m}\rest
i,\ldots, \eta_{\alpha_n}\rest i)\}\\
\end{array}\]
is in the ideal $J$.
\item The sequence $\bar \theta$ is {\em big$^*$ $J$-cofinal} in
$(T, \bar \lambda, \bar I)$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq \eta_\beta$ for distinct $\alpha$,
$\beta<\lambda$
\item[(b)] for every $\eta$ and functions $F_0, \ldots, F_n$ there is
$\alpha^*$ such that
\noindent if $\alpha_0<\ldots <\alpha_n$ and $\alpha^*\leq \alpha_m$ then for
$m\leq n$ the set
\[\hspace{-1.8cm}\begin{array}{ll}
\{i<\delta: & \mbox{ if }\nu_\ell=\left\{
\begin{array}{ll}
\eta_{\alpha_\ell}\rest (i{+}1) & \mbox{ if }
\lambda_{\eta_{\alpha_\ell}\restriction i}=
\lambda_{\eta_{\alpha-m}\rest i}\ \mbox{ or}\\
\ &\lambda_{\eta_{\alpha_\ell}\rest i}=
\lambda_{\eta_{\alpha_n}\rest i}\ \mbox{ and }\
\eta_{\alpha_\ell}(i)< \eta_{\alpha_m}(i)\\
\eta_{\alpha_\ell}\restriction i & \mbox{ if not}\\
\end{array}
\right.\\
\ &\mbox{ then we have }\\
\ &\eta_{\alpha_m}(i)\in F_m(\nu_\ell)\in I_{\eta_{\alpha_m}\restriction
i}\}\\
\end{array}\]
is in the ideal $J$.
\end{description}
\item We define ``strongly$^*$ $J$-cofinal'', ``stronger$^*$ $J$-cofinal'' and
``strongest$^*$ big $J$-cofinal'' in almost the same way, replacing the
requirement that $\alpha^*\leq \alpha_m$ in 5{\bf (b)}, 6{\bf (b)}, 7{\bf (b)}
above (respectively) by $\alpha^*\leq\alpha_0$.
\end{description}
\end{enumerate}
\end{definition}
\begin{remark}
\label{impcof}
{\em
\begin{enumerate}
\item Note that ``strongest $J$-cofinal'' implies ``stronger $J$-cofinal''
and this implies ``strongly $J$-cofinal''. ``Stronger $J$-cofinal''
implies ``$J$-cofinal''. Also ``bigger'' $\Rightarrow$ ``big''
$\Rightarrow$ ``cofinal'', ``big'' $\Rightarrow$ ``strongly''.
\item The different notions of ``strong $J$-cofinality'' (the conditions {\em
(i)} and {\em (i')}) are to allow us to carry some diagonalization arguments.
\item The difference between ``strongly $J$-cofinal'' and ``strongly$^*$
$J$-cofinal'' etc is, in our context, immaterial. we may in all places in this
paper replace the respective notion with its version with ``$*$'' and no harm
will be done.
\end{enumerate}
}
\end{remark}
\begin{remark}
{\em
\begin{enumerate}
\item Remind {\bf pcf}:
An important case is when $\langle\lambda_i: i<\delta\rangle$ is an
increasing sequence of regular cardinals, $\lambda_i>\prod\limits_{j<i}
\lambda_j$, $\lambda_\eta=\lambda_{{\rm lg}\/(\eta)}$,
$I_\eta=J^{{\rm bd}}_{\lambda_\eta}$ and $\lambda=
{\rm tcf}(\prod\limits_{i<\delta}\lambda_i/J)$.
\item Moreover we are interested in more complicated $I_\eta$'s (as in
\cite{Sh:430}, section 5), connected to our problem, so ``the existence of the
true cofinality'' is less clear. But the assumption $2^\mu=\mu^+$ will rescue
us.
\item There are natural stronger demands of cofinality since here we are not
interested just in $x_\alpha$'s but also in Boolean combinations. Thus
naturally we are interested in behaviours of large sets of $n$-tuples, see
\ref{super}.
\end{enumerate}
}
\end{remark}
\begin{proposition}
\label{strimpstr}
Suppose that $(T,\bar{\lambda},\bar{I})\in{\cal K}^{{\rm id}}_{\mu,\delta}$,
$\bar{\eta}=\langle\eta_\alpha:\alpha<\lambda\rangle\subseteq{\lim}_\delta
(T)$ and $J$ is an ideal on $\delta$, $J\supseteq J^{{\rm bd}}_\delta$.
\begin{enumerate}
\item Assume that
\begin{description}
\item[$(\circledcirc)$] if $\alpha<\beta<\lambda$ then $(\forall^J i<\delta)
(\lambda_{\eta_\alpha\rest i}<\lambda_{\eta_\beta\rest i})$.
\end{description}
Then the following are equivalent
\[\mbox{``}\bar{\eta}\mbox{ is strongly $J$-cofinal for }(T,\bar{\lambda},
\bar{I})\mbox{''}\]
\[\mbox{``}\bar{\eta}\mbox{ is stronger $J$-cofinal for }(T,\bar{\lambda},
\bar{I})\mbox{''}\]
\[\mbox{``}\bar{\eta}\mbox{ is strongest $J$-cofinal for }(T,\bar{\lambda},
\bar{I})\mbox{''.}\]
\[\mbox{``}\bar{\eta}\mbox{ is big $J$-cofinal for }(T,\bar{\lambda},
\bar{I})\mbox{''.}\]
\item If $I_\nu\supseteq J^{{\rm bd}}_{\lambda_\nu}$ and $\lambda_\nu
=\lambda_{{\rm lg}\/(\nu)}$ for each $\nu\in T$ and the
sequence $\bar{\eta}$ is stronger $J$-cofinal for $(T,\bar{\lambda},\bar{I})$
then for some $\alpha^*<\lambda$ the sequence $\langle\eta_\alpha:\alpha^*\leq
\alpha<\lambda\rangle$ is $<_J$-increasing.
\item If $\eta\in T_i \Rightarrow \lambda_\eta=\lambda_i$ and $\bar
\eta <_J$-increasing in $\prod\limits_{i<\delta}$ {\em then}
``big'' is equivalent to ``stronger''. \QED
\end{enumerate}
\end{proposition}
\begin{proposition}
\label{getcofinal}
Suppose that
\begin{enumerate}
\item $\langle\lambda_i: i<\delta\rangle$ is an increasing sequence of regular
cardinals, $\delta<\lambda_0$ is a limit ordinal,
\item $T=\bigcup\limits_{i<\delta}\prod\limits_{j<i}\lambda_j$,
$I_\eta=I_{{\rm lg}\/(\eta)}=J^{{\rm bd}}_{\lambda_{{\rm lg}\/(\eta)}}$,
$\lambda_\eta=\lambda_{{\rm lg}\/(\eta)}$,
\item $J$ is an ideal on $\delta$,
$\lambda={\rm tcf}(\prod\limits_{i<\delta}\lambda_i/J)$ and it is examplified
by a sequence $\bar{\eta}=\langle\eta_\alpha:\alpha<\lambda\rangle\subseteq
\prod\limits_{i<\delta}\lambda_i$
\item for each $i<\delta$
\[|\{\eta_\alpha\rest i: \alpha<\lambda\}|<\lambda_i\]
(so e.g. $\lambda_i>\prod\limits_{j<i}\lambda_j$ suffices).
\end{enumerate}
Then the sequence $\bar{\eta}$ is $J$-cofinal in $(T,\bar{\lambda},\bar{I})$.
\end{proposition}
\Proof First note that our assumptions imply that each ideal
$I_\eta=I_{{\rm lg}\/(\eta)}$ is $|\{\eta_\alpha\rest {\rm lg}\/(\eta):
\alpha<\lambda\}|^+$-complete. Hence for
each sequence $\bar{A}=\langle A_\eta:\eta\in T\rangle\in
\prod\limits_{\eta\in T} I_\eta$ and $i<\delta$ the set
\[A_i\stackrel{\rm def}{=}\bigcup\{A_{\eta_\alpha\rest i}: \alpha<\lambda\}\]
is in the the ideal $I_i$, i.e. it is bounded in $\lambda_i$ (for $i<\delta$).
(We should remind here our convention which says in this case that we do not
distinguish $\lambda_i$ and $\suc_T(\eta)$ if ${\rm lg}\/(\eta)=i$,
see \ref{evenmorenot}.) Take $\eta^*\in\prod\limits_{i<\delta}\lambda_i$
such that for each $i<\delta$ we have $A_i\subseteq\eta^*(i)$. As the sequence
$\bar{\eta}$ realizes the true cofinality of
$\prod\limits_{i<\delta}\lambda_i/J$ we find $\alpha^*<\lambda$ such that
\[\alpha^*\leq\alpha<\lambda\ \ \ \Rightarrow\ \ \ \{i<\delta: \eta_\alpha(i)
<\eta^*(i)\}\in J\]
which allows us to finish the proof. \QED
\medskip
It follows from the above proposition that the notion of $J$-cofinal sequences
is not empty. Of course it is better to have ``strongly (or even: stronger)
$J$-cofinal'' sequences $\bar{\eta}$. So it is nice to have that sometimes the
weaker notion implies the stronger one.
\begin{proposition}
\label{cofimpstrong}
Assume that $\delta$ is a limit ordinal, $\mu$ is a cardinal,
$(T,\bar{\lambda},\bar{I})\in{\cal K}^{{\rm id}}_{\mu,\delta}$. Let $J$ be an ideal on
$\delta$ such that $J\supseteq J^{{\rm bd}}_\delta$ (which is our standard
hypothesis). Further suppose that
\begin{description}
\item[$(\circledast)$] if $\eta\in T_i$ then the ideal $I_\eta$ is
$(|T_i|+\sum\{\lambda_\nu\!: \nu{\in} T_i\ \&\ \lambda_\nu{<}\lambda_\eta
\})^+$--complete.
\end{description}
{\em Then} each $J$-cofinal sequence $\bar{\eta}$ for
$(T,\bar{\lambda},\bar{I})$ is strongly $J$-cofinal for
$(T,\bar{\lambda},\bar{I})$.
If in addition $\eta\neq\nu\in T_i\ \Rightarrow\ \lambda_\eta\neq\lambda_\nu$
then $\bar \eta$ is big $J$-cofinal for $(T,\bar{\lambda},\bar{I})$. Also if
in addition
\[\eta\in T_i\ \ \Rightarrow\ \ (\exists^{!1}\nu\in
T_i)(\lambda_\nu=\lambda_\eta)\ \vee\ [(\exists^{\leq \lambda_\eta}
\nu\in T_i)(\lambda_\nu =\lambda_\eta)\ \&\ I_\eta\mbox{ normal}]\]
then $\bar \eta$ is big $J$-cofinal.
\end{proposition}
\Proof Let $n<\omega$ and $F_0,\ldots,F_n$ be $(n+1)$-place functions. First
we define a sequence $\bar{A}=\langle A_\eta: \eta\in T\rangle$. For $m\leq n$
and a sequence $\langle\eta_m,\ldots,\eta_n\rangle\subseteq T_i$ we put
\[\begin{array}{ll}
A^m_{\langle\eta_m,\ldots,\eta_n\rangle}{=}\bigcup\{& F_m(\nu_0,\ldots,
\nu_{m-1},\eta_m,\ldots,\eta_n):\nu_0,\ldots,\nu_{m-1}\in T_{i+1},\\
\ &\quad\quad(\nu_0,\ldots,\nu_{m-1},\eta_m,\ldots,\eta_n)\!\in\!\dom(F),\\
\ &\quad\quad\lambda_{\nu_0\rest i}<\lambda_\eta,\ldots,\lambda_{\nu_{m-1}
\rest i}<\lambda_{\eta_m}\\
\ &\quad\quad\mbox{and }F(\nu_0,\ldots,\nu_{m-1},\eta_m,\ldots,\eta_n)\in
I_{\eta_m}\}
\end{array}\]
and next for $\eta\in T_i$ let
\[A_\eta=\bigcup\{A^m_{\langle\eta,\eta_{m+1},\ldots,\eta_n\rangle}: m\leq n\
\&\ \eta_{m+1},\ldots,\eta_n\in T_i\}.\]
Note that the assumption $(\circledast)$ was set up so that
$A^m_{\langle\eta_m,\ldots,\eta_n\rangle}\in I_{\eta_m}$ and the sets $A_\eta$
are in $I_\eta$ (for $\eta\in T$).
By the $J$-cofinality of $\bar{\eta}$, for some $\alpha^*<\lambda$ we have
\[\alpha^*\leq\alpha<\lambda\ \ \ \Rightarrow\ \ \ (\forall^J i<\delta)
(\eta_\alpha\rest(i+1)\notin A_{\eta_\alpha\rest i}).\]
We are going to prove that this $\alpha^*$ is as required in the definition of
strongly $J$-cofinal sequences. So suppose that $m\leq n$, $\alpha_0<\ldots
<\alpha_n<\lambda$ and $\alpha^*\leq\alpha_m$. By the choice of $\alpha^*$ we
have that the set $A\stackrel{\rm def}{=}\{i<\delta: \eta_{\alpha_m}
\rest(i+1)\in A_{\eta_{\alpha_m}\rest i}\}$ is in the ideal $J$. But if
$i<\delta$ is such
that
\begin{quotation}
\noindent $(\forall \ell<m)(\lambda_{\eta_{\alpha_\ell}\rest i}<
\lambda_{\eta_{\alpha_m}\rest i})$ and
\noindent $F(\eta_{\alpha_0}\rest(i+1),\ldots,
\eta_{\alpha_{m-1}}\rest (i+1),\eta_{\alpha_m}\rest i,\ldots,\eta_{\alpha_n}
\rest i)\in I_{\eta_{\alpha_m}\rest i}$ but
\noindent $\eta_{\alpha_m}\rest (i+1)\in F(\eta_{\alpha_0}\rest(i+1),\ldots,
\eta_{\alpha_{m-1}}\rest (i+1),\eta_{\alpha_m}\rest i,\ldots,\eta_{\alpha_n}
\rest i)$
\end{quotation}
then clearly $\eta_{\alpha_m}\rest (i+1)\in A^m_{\langle\eta_{\alpha_m}\rest
i,\ldots,\eta_{\alpha_n}\rest i\rangle}$ and so $i\in A$. This finishes the
proof.
The ``big'' verion should be clear too. \QED
\begin{proposition}
\label{stronger}
Assume that $\mu$ is a strong limit uncountable cardinal and
$\langle\mu_i:i<\delta \rangle$ is an increasing sequence of cardinals with
limit $\mu$. Further suppose that $(T,\bar{\lambda},\bar{I})\in
{\cal K}^{{\rm id}}_{\mu,\delta}$, $|T_i|\leq\mu_i$ (for $i<\delta$), $\lambda_\eta<\mu$
and each $I_\eta$ is $\mu_{{\rm lg}\/(\eta)}^+$-complete and contains all singletons
(for $\eta\in T$). Finally assume
$2^\mu=\mu^+$ and let $J$ be an ideal on $\delta$, $J\supseteq
J^{{\rm bd}}_\delta$.
\noindent Then there exists a stronger $J$-cofinal sequence $\bar{\eta}$ for
$(T,\bar{\lambda},\bar{I})$ of the length $\mu^+$ (even for
$J=J^{{\rm bd}}_\delta$).
\noindent We can get ``big'' if
$$
\rho\neq \eta\in T_i\ \&\ \lambda_\rho=\lambda_\eta \Rightarrow
(\exists^{\leq \lambda_\eta}\nu\in T_i)(\lambda_\nu=\lambda_\eta)\ \&\
I_\eta\mbox{ normal.}
$$
\end{proposition}
\Proof This is a straight diagonal argument. Put
\[\begin{array}{ll}
Y\stackrel{\rm def}{=}\{\langle F_0,\ldots,F_n\rangle: &n<\omega \mbox{ and
each }F_\ell\mbox{ is a function with}\\
\ &\dom(F)\subseteq T^{n+1},
{\rm rng}(F)\subseteq\bigcup\limits_{\eta\in T} I_\eta\}.\\
\end{array}\]
Since $|Y|=\mu^\mu=\mu^+$ (remember that $\mu$ is strong limit and
$\lambda_\eta<\mu$ for $\eta\in T$) we may choose an enumeration $Y=\{\langle
F^\xi_0,\ldots,F^\xi_{n_\xi}\rangle: \xi<\mu^+\}$. For each $\zeta<\mu^+$
choose an increasing sequence $\langle\A^\zeta_i: i<\delta\rangle$ such that
$|\A^\zeta_i|\leq\mu_i$ and $\zeta=\bigcup\limits_{i<\delta}\A^\zeta_i$. Now
we choose by induction on $\zeta<\mu^+$ branches $\eta_\zeta$ such that for
each $\zeta$ the restriction $\eta_\zeta\rest i$ is defined by induction on
$\zeta$ as follows.
\noindent If $i=0$ or $i$ is limit then there is nothing to do.
\noindent Suppose now that we have defined $\eta_\zeta\rest i$ and $\eta_\xi$
for $\xi<\zeta$. We find $\eta_\zeta(i)$ such that
\begin{description}
\item[$(\alpha)$] $\eta_\zeta(i)\in\lambda_{\eta_\zeta\rest i}$
\item[$(\beta)$] if $\varepsilon\in\A^\zeta_i$, $m\leq n_\varepsilon$,
$\alpha_0,\ldots,\alpha_{m-1}\in \A^\zeta_i$ (hence $\alpha_\ell<\zeta$ so
$\eta_{\alpha_\ell}$ are defined already), $\nu_{m+1},\ldots,\nu_n\in T_i$ and
\[\hspace{-1cm}F^\varepsilon_m(\eta_{\alpha_0}\rest (i+1),\ldots,
\eta_{\alpha_{m-1}}\rest (i+1), \eta_\zeta\rest i, \nu_{m+1},\ldots,\nu_n) \in
I_{\eta_\zeta\rest i}\quad\mbox{and well defined}\]
then
\[\eta_\zeta\rest (i+1)\notin F_m^\varepsilon(\eta_{\alpha_0}\rest
(i+1),\ldots, \eta_{\alpha_{m-1}}\rest (i+1),\eta_\zeta\rest i,
\nu_{m+1},\ldots,\nu_n)\]
\item[$(\gamma)$] $\eta_\zeta\rest(i+1)\notin\{\eta_\varepsilon\rest (i+1):
\varepsilon\in\A^\zeta_i\}$.
\end{description}
{\em Why it is possible?\/} Note that there is $\leq\aleph_0+|\A^\zeta_i|+
|\A^\zeta_i|^{\textstyle {<}\aleph_0}+|T_i|\leq\mu_i$ negative demands
and each of them says that $\eta_\zeta\rest (i+1)$ is not in some set from
$I_{\eta_\zeta\rest i}$ (remember that we have assumed that the ideals
$I_{\eta_\zeta\rest i}$ contain singletons). Consequently using the
completeness of the ideal we may satisfy the requirements
$(\alpha)$--$(\gamma)$ above.
Now of course $\eta_\zeta\in{\lim}_\delta(T)$. Moreover if
$\varepsilon<\zeta<\mu^+$ then $(\exists i<\delta)(\varepsilon\in \A^\zeta_i)$
which implies $(\exists i<\delta)(\eta_\varepsilon\rest (i+1)\neq
\eta_\zeta\rest (i+1))$. Consequently
\[\varepsilon<\zeta<\mu^+\ \ \ \Rightarrow\ \ \
\eta_\varepsilon\neq\eta_\zeta.\]
Checking the demand {\bf (b)} of ``stronger $J$-cofinal'' is straightforward:
for functions $F_0,\ldots,F_n$ (and $n\in\omega$) take $\varepsilon$ such that
\[\langle F_0,\ldots,F_n\rangle = \langle F^\varepsilon_0,\ldots,
F^\varepsilon_{n_\varepsilon}\rangle\]
and put $\alpha^*=\varepsilon+1$. Suppose now that $m\leq n$,
$\alpha_0<\ldots<\alpha_n<\lambda$, $\alpha^*\leq\alpha_m$. Let $i^*<\delta$
be such that for $i>i^*$ we have
\[\varepsilon,\alpha_0,\ldots,\alpha_{m-1}\in \A^{\alpha_m}_i.\]
Then by the choice of $\eta_{\alpha_m}\rest(i+1)$ we have that for each
$i>i^*$
\smallskip
if $F_m^\varepsilon(\eta_{\alpha_0}\rest (i+1),\ldots,
\eta_{\alpha_{m-1}}\rest (i+1),\eta_\zeta\rest i, \eta_{\alpha_{m+1}}\rest i,
\ldots,\eta_{\alpha_n}\rest i)\in I_{\eta_{\alpha_m}\rest i}$
then $\eta_{\alpha_m}\rest i\notin F_m^\varepsilon(\eta_{\alpha_0}\rest
(i+1),\ldots, \eta_{\alpha_{m-1}}\rest (i+1),\eta_\zeta\rest i,
\eta_{\alpha_{m+1}}\rest i, \ldots,\eta_{\alpha_n}\rest i)$.
\smallskip
\noindent This finishes the proof. \QED
\begin{remark}
{\em
The proof above can be carried for functions $F$ which depend on
$(\eta_{\alpha_0},\ldots,\eta_{\alpha_{m-1}},\eta_{\alpha_m}\rest i,\ldots,
\eta_{\alpha_n}\rest i)$. This will be natural later.
}
\end{remark}
Let us note that if the ideals $I_\eta$ are sufficiently complete then
$J$-cofinal sequences cannot be too short.
\begin{proposition}
Suppose that $(T,\bar{\lambda},\bar{I})\in{\cal K}^{{\rm id}}_{\mu,\delta}$ is such that
for each $\eta\in T_i$, $i<\delta$ the ideal $I_\eta$ is
$(\kappa_i)^+$--complete (enough if $[\lambda_\eta]^{\kappa_i}
\subseteq I_\eta$). Let $J\supseteq J^{{\rm bd}}_\delta$ be an ideal on $\delta$
and let $\bar{\eta}=\langle\eta_\alpha:\alpha<\delta^*\rangle$ be a
$J$-cofinal sequence for $(T,\bar{\lambda},\bar{I})$. Then
\[\delta^*>\lim\sup_J \kappa_i\]
and consequently
\[\cf(\delta^*)>\lim\sup_J \kappa_i.\]
\end{proposition}
\Proof Fix an enumeration $\delta^*=\{\alpha_\varepsilon: \varepsilon<
|\delta^*|\}$ and for $\alpha<\delta^*$ let $\zeta(\alpha)$ be the unique
$\zeta$ such that $\alpha=\alpha_\zeta$.
\noindent For $\eta\in T_i$, $i<\delta$ put
\[A_\eta\stackrel{\rm def}{=}\{\nu\in\suc_T(\eta): (\exists\varepsilon\leq
\kappa_i)(\nu\vartriangleleft\eta_\varepsilon)\}.\]
Clearly $|A_\eta|\leq \kappa_i$ and hence $A_\eta\in I_\eta$. Apply the
$J$-cofinality of $\bar{\eta}$ to the sequence $\bar{A}=\langle A_\eta:
\eta\in T\rangle$. Thus there is $\alpha^*<\delta^*$ such that for each
$\alpha\in [\alpha^*,\delta^*)$ we have
\[(\forall^J i<\delta)(\eta_\alpha\rest (i+1)\notin A_{\eta_\alpha\rest i})\]
and hence
\[(\forall^J i<\delta)(\zeta(\alpha)>\kappa_i)\]
and consequently
\[\zeta(\alpha)\geq \lim\sup_J \kappa_i.\]
Hence we conclude that $|\delta^*|>\lim\sup_J \kappa_i$.
\noindent For the part ``consequently'' of the proposition note that if
$\langle\eta_\alpha: \alpha<\delta^*\rangle$ is $J$-cofinal (in
$(T,\bar{\lambda},\bar{I})$) and $A\subseteq\delta^*$ is cofinal in $\delta^*$
then $\langle\eta_\alpha: \alpha\in A\rangle$ is $J$-cofinal too. \QED
\begin{remark}
{\em
\begin{enumerate}
\item So if we have a $J$-cofinal sequence of the length $\delta^*$ then we
also have one of the length $\cf(\delta^*)$. Thus assuming regularity of the
length is natural.
\item Moreover the assumption that the length of the sequence is above
$|\delta|+|T|$ is very natural and in most cases it will follow from the
$J$-cofinality (and completeness assumptions). However we will try to state
this condition in the assumptions whenever it is used in the proof (even if
it can be concluded from the other assumptions).
\end{enumerate}
}
\end{remark}
\section{Getting $(\kappa,{\rm not}\lambda)$-Knaster algebras}
\begin{proposition}
\label{algide}
Let $\lambda,\sigma$ be cardinals such that $(\forall\alpha<\sigma)
(2^{|\alpha|}<\lambda)$, $\sigma$ is regular. Then there are a Boolean algebra
${\Bbb B}$, a sequence $\langle y_\alpha:\alpha<\lambda\rangle\subseteq{\Bbb B}^+$ and
an ideal $I$ on $\lambda$ such that
\begin{description}
\item[(a)] if $X\subseteq\lambda$, $X\notin I$ then $(\exists
\alpha,\beta\in X)({\Bbb B}\models y_\alpha\cap y_\beta={\bf 0})$
\item[(b)] the ideal $I$ is $\sigma$-complete
\item[(c)] the algebra ${\Bbb B}$ satisfies the $\mu$-Knaster condition for any
regular uncountable $\mu$ (really ${\Bbb B}$ is free).
\end{description}
\end{proposition}
\Proof Let ${\Bbb B}$ be the Boolean algebra freely generated by
$\{z_\alpha:\alpha<\lambda\}$ (so the demand {\bf (c)} is satisfied). Let
$A=\{(\alpha,\beta):\alpha<\beta<\lambda\}$ and
$y_{(\alpha,\beta)}=z_\alpha-z_\beta(\neq{\bf 0})$ (for $(\alpha,\beta)\in A$).
The ideal $I$ of subsets of $A$ defined by
\begin{quotation}
\noindent a set $X\subseteq A$ is in $I$ if and only if
\noindent there are $\zeta<\sigma$, $X_\varepsilon\subseteq A$ (for
$\varepsilon<\zeta$) such that $X\subseteq\bigcup\limits_{\varepsilon<\zeta}
X_\varepsilon$ and for every $\varepsilon<\zeta$ no two
$y_{(\alpha_1,\beta_1)}, y_{(\alpha_2,\beta_2)}\in X_\varepsilon$ are disjoint
in ${\Bbb B}$.
\end{quotation}
First note that
\begin{claim}
$A\notin I$.
\end{claim}
\noindent{\em Why?}\ \ \ If not then we have witnesses $\zeta<\sigma$ and
$X_\varepsilon$ (for $\varepsilon<\zeta$) for it.
So $A=\bigcup\limits_{\varepsilon<\zeta}X_\varepsilon$ and hence for
$(\alpha,\beta)\in A$ we have $\varepsilon(\alpha,\beta)$ such that
$y_{(\alpha,\beta)}\in X_{\varepsilon(\alpha,\beta)}$. So
$\varepsilon(\cdot,\cdot)$ is actually a function from $[\lambda]^2$ to
$\zeta<\sigma$. By the Erd\"os--Rado theorem we find $\alpha<\beta<\gamma<
\lambda$ such that $\varepsilon(\alpha,\beta)=\varepsilon(\beta,\gamma)$. But
\[y_{(\alpha,\beta)}\cap y_{(\beta,\gamma)}=(z_\alpha-z_\beta)\cap
(z_\beta-z_\gamma)={\bf 0}\]
so $(\alpha,\beta)$, $(\beta,\gamma)$ cannot be in the same $X_\varepsilon$
-- a contradiction.
\medskip
To finish the proof note that $I$ is $\sigma$-complete (as $\sigma$ is
regular), if $X\notin I$ then, by the definition of $I$, there are two
disjoint elements in $\{y_{(\alpha,\beta)}: (\alpha,\beta)\in X\}$. Finally
$|A|=\lambda$. \QED
\begin{definition}
\label{well1}
\begin{description}
\item[(a)] A pair $({\Bbb B},\bar{y})$ is called {\em a $\lambda$-marked Boolean
algebra} if ${\Bbb B}$ is a Boolean algebra and $\bar{y}=\langle
y_\alpha:\alpha<\lambda\rangle$ is a sequence of non-zero elements of ${\Bbb B}$.
\item[(b)] A triple $({\Bbb B},\bar{y},I)$ is called a {\em $(\lambda,\chi)$-well
marked Boolean algebra} if $({\Bbb B},\bar{y})$ is a $\lambda$-marked Boolean
algebra, $\chi$ is a regular cardinal and $I$ is a (proper) $\chi$-complete
ideal on $\lambda$ such that
\[\{A\subseteq\lambda: (\forall \alpha,\beta\in A)({\Bbb B}\models y_\alpha\cap
y_\beta\neq {\bf 0})\}\subseteq I.\]
By {\em $\lambda$-well marked Boolean algebra} we will mean
$(\lambda,\aleph_0)$-well marked one. As in the above situation $\lambda$ can
be read from $\bar{y}$ (as $\lambda={\rm lg}\/(\bar{y})$) we may omit it and then we
may speak just about well marked Boolean algebras.
\end{description}
\end{definition}
\begin{remark}
\label{well2}
{\em
Thus proposition \ref{algide} says that if $\lambda,\sigma$ are regular
cardinals and
\[(\forall \alpha<\sigma)(2^{|\alpha|}<\lambda)\]
then there exists a $(\lambda,\sigma)$-well marked Boolean algebra
$({\Bbb B},\bar{y},I)$ such that ${\Bbb B}$ satisfies the $\kappa$-Knaster
property for every $\kappa$.
}
\end{remark}
\begin{definition}
\label{constructor}
\begin{description}
\item[(a)] For cardinals $\mu$ and $\lambda$ and a limit ordinal $\delta$,
{\em a $(\delta,\mu,\lambda)$-constructor} is a system
\[{\cal C}=(T, \bar{\lambda}, \bar{\eta},\langle ({\Bbb B}_\eta,\bar{y}_\eta):\eta\in
T\rangle)\]
such that
\begin{enumerate}
\item $(T,\bar{\lambda})\in {\cal K}_{\mu,\delta}$,
\item $\bar{\eta}= \langle \eta_i: i\in \lambda\rangle$ where
$\eta_i\in\lim_\delta(T)$ (for $i<\lambda$) are distinct $\delta$-branches
through $T$ and
\item for each $\eta\in T$:\ \ \ $({\Bbb B}_\eta,\bar{y}_\eta)$ is a
$\lambda_\eta$-marked Boolean algebra, i.e. $\bar{y}_\eta=\langle
y_{\eta\hat{\ }\langle\alpha\rangle}:\alpha< \lambda_\eta\rangle\subseteq
{\Bbb B}_\eta^+$ (usually this will be an enumeration of ${\Bbb B}_\eta^+$).
\end{enumerate}
\item[(b)] Let ${\cal C}$ be a constructor (as above). We define Boolean algebras
${\Bbb B}_2={\Bbb B}^{{\rm red}}={\Bbb B}^{{\rm red}}({\cal C})$ and ${\Bbb B}_1={\Bbb B}^{{\rm green}}={\Bbb B}^{{\rm green}}({\cal C})$
by:
\medskip
\noindent ${\Bbb B}^{{\rm red}}$ is the Boolean algebra freely generated by
$\{x_i:i<\lambda\}$ {\em except that}
\begin{quotation}
\noindent {\em if} $i_0,\ldots,i_{n-1}<\lambda$, $\nu=\eta_{i_0}\rest
\zeta =\eta_{i_1}\rest\zeta=\ldots=\eta_{i_{n-1}}\rest\zeta$ and
${\Bbb B}_\nu\models\bigcap\limits_{\ell<n} y_{\eta_{i_\ell}\rest(\zeta+1)}
={\bf 0}$
\noindent {\em then} $\bigcap\limits_{\ell<n} x_{i_\ell}={\bf 0}$
\end{quotation}
[Note: we may demand that the sequence $\langle\eta_{i_\ell}(\zeta):\ell<n
\rangle$ is strictly increasing, this will cause no difference.]
\medskip
\noindent ${\Bbb B}^{{\rm green}}$ is the Boolean algebra freely generated by
$\{x_i:i<\lambda\}$ {\em except that}
\begin{quotation}
\noindent {\em if} $\nu=\eta_i\rest\zeta =\eta_j\rest\zeta$,
$\eta_i(\zeta)\neq\eta_j(\zeta)$ and
\noindent ${\Bbb B}_\nu\models y_{\eta_i\rest(\zeta+1)}\cap
y_{\eta_j\rest(\zeta+1)}\neq{\bf 0}$
\noindent {\em then} $x_i\cap x_j={\bf 0}$.
\end{quotation}
\end{description}
\end{definition}
\begin{remark}
{\em
\begin{enumerate}
\item The equations for the green case can look strange but they have to be
dual to the ones of the red case.
\item ``Freely generated except $\ldots$'' means that a Boolean combination is
non-zero except when some (finitely many) conditions implies it. For this it
is enough to look at elements of the form
\[x_{i_0}^{{\frak t}_0}\cap\ldots \cap x_{i_{n-1}}^{{\frak t}_{n-1}}\]
where ${\frak t}_\ell\in\{0,1\}$.
\item Working in the free product ${\Bbb B}^{{\rm red}}*{\Bbb B}^{{\rm green}}$ we will use
the same notation for elements (e.g. generators) of ${\Bbb B}^{{\rm red}}$ as for
elements of ${\Bbb B}^{{\rm green}}$. Thus $x_i$ may stay either for the respective
generator in ${\Bbb B}^{{\rm red}}$ or ${\Bbb B}^{{\rm green}}$. We hope that this will not be
confusing, as one can easily decide in which algebra the element is considered
from the place of it (if $x\in{\Bbb B}^{{\rm red}}$, $y\in{\Bbb B}^{{\rm green}}$ then $(x,y)$
will stay for the element $x\cap_{{\Bbb B}^{{\rm red}}*{\Bbb B}^{{\rm green}}} y\in{\Bbb B}^{{\rm red}}*
{\Bbb B}^{{\rm green}}$). In particular we may write $(x_i,x_i)$ for an element
which could be denoted $x_i^{{\rm red}}\cap x_i^{{\rm green}}$.
\end{enumerate}
}
\end{remark}
\begin{remark}
{\em
If the pair $({\Bbb B}^{{\rm red}},{\Bbb B}^{{\rm green}})$ is a counterexample with the free
product ${\Bbb B}^{{\rm red}} *{\Bbb B}^{{\rm green}}$ failing the $\lambda$-cc but each of the
algebras satisfying that condition then each of the algebras fails the
$\lambda$-Knaster condition. But ${\Bbb B}^{{\rm red}}$ is supposed to have $\kappa$-cc
($\kappa$ smaller than $\lambda$). This is known to restrict $\lambda$.
}
\end{remark}
\begin{proposition}
\label{notknaster}
Assume that ${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta):
\eta\in T\rangle)$ is a $(\delta,\mu,\lambda)$-constructor and $J\supseteq
J^{{\rm bd}}_\delta$ is an ideal on $\delta$ such that
\begin{description}
\item[(a)] $\bar{\eta}=\langle \eta_i: i\in T\rangle$ is $J$-cofinal for
$(T,\bar{\lambda},\bar{I})$
\item[(b)] if $X\in I^+_\eta$ then
\[(\exists\alpha,\beta\in X)({\Bbb B}_\eta\models y_{\eta\hat{\
}\langle\alpha\rangle}\cap y_{\eta\hat{\ }\langle\beta\rangle}={\bf 0}).\]
\end{description}
Then the sequence $\langle x^{{\rm red}}_\alpha:\alpha<\lambda\rangle$ examplifies
that ${\Bbb B}^{{\rm red}}({\cal C})$ fails the $\lambda$-Knaster condition.
\end{proposition}
\noindent{\bf Explanation:}\ \ \ The above proposition is not just something
in the direction of Problem~\ref{pwa}. The tuple $({\Bbb B}^{{\rm red}},\bar{x},
J^{{\rm bd}}_\lambda)$ is like $({\Bbb B}_\eta,\bar{y}_\eta, I_\eta)$ but
$J^{{\rm bd}}_\lambda$ is nicer than ideals given by previous results. Using such
objects makes building examples for Problem~\ref{pwa} much easier.
\medskip
\Proof It is enough to show that
\begin{quotation}
\noindent for each $Y\in [\lambda]^{\textstyle\lambda}$ one can find
$\varepsilon,\zeta\in Y$ such that
\[{\Bbb B}_{\eta_\varepsilon\rest i}\models y_{\eta_\varepsilon\rest (i+1)}\cap
y_{\eta_\zeta\rest (i+1)}={\bf 0}\]
where $i={\rm lg}\/(\eta_\varepsilon\wedge\eta_\zeta)$.
\end{quotation}
For this, for each $\nu\in T$ we put
\[A_\nu\stackrel{\rm def}{=}\{\alpha<\lambda_\nu: (\exists\varepsilon\in Y)
(\nu\hat{\ }\langle\alpha\rangle\vartriangleleft\eta_\varepsilon)\}.\]
\begin{claim}
There is $\nu\in T$ such that $A_\nu\notin I_\nu$.
\end{claim}
\noindent{\em Why?}\ \ \ First note that by the definition of $A_\nu$, for
each $\varepsilon\in Y$ we have
\[(\forall i<\delta)(\eta_\varepsilon \hat{\ }\langle i\rangle \in
A_{\eta_\varepsilon\rest i}).\]
Now, if we had that $A_\nu\in I_\nu$ for all $\nu\in T$ then we could apply
the assumption that $\bar{\eta}$ is $J$-cofinal for
$(T,\bar{\lambda},\bar{I})$ to the sequence $\langle A_\nu: \nu\in T\rangle$.
Thus we would find $\alpha^*<\lambda$ such that
\[\alpha^*\leq\alpha<\lambda\ \ \ \Rightarrow\ \ \ \{i<\delta: \eta_\alpha(i)
\notin A_{\eta_\alpha \rest i}\}\in J\]
which contradicts our previous remark (remember $|Y|=\lambda$). The claim is
proved.
\medskip
Due to the claim we find $\nu\in T$ such that $A_\nu\notin I_\nu$. By the part
{\bf (b)} of our assumptions we find $\alpha,\beta\in A_\nu$ such that
\[{\Bbb B}_\nu\models y_{\nu\hat{\ }\langle\alpha\rangle}\cap y_{\nu\hat{\
}\langle\beta\rangle}={\bf 0}.\]
Choose $\varepsilon,\zeta\in Y$ such that $\nu\hat{\ }\langle\alpha\rangle
\vartriangleleft\eta_\varepsilon$, $\nu\hat{\ }\langle\beta\rangle
\vartriangleleft\eta_\zeta$ (see the definition of $A_\nu$). Then
$\nu=\eta_\varepsilon\wedge\eta_\zeta$ and
\[{\Bbb B}_\nu\models y_{\eta_\varepsilon\rest (i+1)}\cap y_{\eta_\zeta\rest
(i+1)}={\bf 0}\]
(where $i={\rm lg}\/(\nu)$), finishing the proof of the proposition. \QED
\begin{lemma}
\label{consknaster}
Let ${\cal C}=(T, \bar{\lambda}, \bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta):
\eta\in T\rangle)$ be a $(\delta,\mu,\lambda)$-constructor such that
\begin{description}
\item[$(\bigstar)$] the Boolean algebras ${\Bbb B}_\eta$ satisfy the
$(2^{|\delta|})^+$--Knaster condition.
\end{description}
{\em Then} the Boolean algebra ${\Bbb B}^{{\rm red}}({\cal C})$ satisfies the
$(2^{|\delta|})^+$--Knaster condition.
\noindent In fact we may replace $(2^{|\delta|})^+$ above by any regular
cardinal $\theta$ such that
\[(\forall\alpha<\theta)(|\alpha|^{|\delta|}<\theta).\]
\noindent To get that ${\Bbb B}^{{\rm red}}({\cal C})$ satisfies the $(2^{|\delta|})^+$--cc it
is enough if instead of $(\bigstar)$ we assume
\begin{description}
\item[$(\bigstar\bigstar)$] every free product of finitely many of the Boolean
algebras ${\Bbb B}_\eta$ satisfies the $(2^{|\delta|})^+$--cc.
\end{description}
\end{lemma}
\noindent{\bf Remark}: \ \ \ 1. Usually we will have $\delta=\cf(\mu)$.
\noindent 2. Later we will get more (e.g. $|\delta|^+$-Knaster if
$(T,\bar{\eta})$ is hereditarily free, see \ref{free}, \ref{morekna}).
\medskip
\Proof Let $\theta=(2^{|\delta|})^+$ and assume $(\bigstar)$ (the other cases
have the same proofs). Suppose that
$z_\varepsilon\in{\Bbb B}^{{\rm red}}\setminus\{{\bf 0}\}$ (for $\varepsilon<\theta$). We
start with a series of reductions which we describe fully here but later, in
similar situations, we will state what is the result of the procedure only.
\medskip
\noindent{\bf Standard cleaning}:\ \ \ Each $z_\varepsilon$ is a Boolean
combination of some generators $x_{i_0},\ldots,x_{i_{n-1}}$. But, as we want
to find a subsequence with non-zero intersections, we may replace
$z_\varepsilon$ by any non-zero $z\leq z_\varepsilon$. Consequently we may
assume that each $z_\varepsilon$ is an intersection of some generators or
their complements. Further, as $\cf(\theta)=\theta>\aleph_0$ we may assume that
the number of generators needed for this representation does not depend on
$\varepsilon$ and is equal to, say, $n^*$. Thus we have two functions
\[i:\theta\times n^*\longrightarrow\lambda\quad\mbox{ and }\quad
{\frak t}:\theta\times n^*\longrightarrow 2\] such that for each
$\varepsilon<\theta$: \[z_\varepsilon=\bigcap_{\ell<n^*}
(x_{i(\varepsilon,\ell)})^{{\frak t}(\varepsilon,\ell)}\] and there is no repetition
in $\langle i(\varepsilon,\ell):\ell<n^*\rangle$. Moreover we may assume that
${\frak t}(\varepsilon,\ell)$ does not depend on $\varepsilon$, i.e.
${\frak t}(\varepsilon,\ell)={\frak t}(\ell)$. Due to the $\Delta$-lemma for finite sets
we may assume that $\langle \langle i(\varepsilon,\ell):\ell<n^*\rangle:
\varepsilon<\theta\rangle$ is a $\Delta$-system of sequences, i.e.:
\begin{description} \item[$(*)_1$]
$i(\varepsilon,\ell_1)=i(\varepsilon,\ell_2)\ \ \Rightarrow\ \ \ell_1=\ell_2$\
\ \ \ and \item[$(*)_2$] for some $w\subseteq n^*$ we have
\[\hspace{-0.7cm}(\exists\varepsilon_1\!<\!\varepsilon_2<\theta)
(i(\varepsilon_1,\ell)\!=\! i(\varepsilon_2,\ell))\ \mbox{ iff }\
(\forall\varepsilon_1,\varepsilon_2\!<\!\theta)(i(\varepsilon_1,\ell)\!=\!
i(\varepsilon_2,\ell))\ \mbox{ iff }\ \ell\in w\] \end{description} Now note
that, by the definition of the algebra ${\Bbb B}^{{\rm red}}$, \begin{description}
\item[$(*)_3$] $z_{\varepsilon_1}\cap z_{\varepsilon_2}={\bf 0}$\ \ \ \ if and
only if
\[\hspace{-1cm}\bigcap\{x^{{\frak t}(\ell)}_{i(\varepsilon_1,\ell)}:
\ell<n^*, {\frak t}(\ell)=0\}
\cap\bigcap\{x^{{\frak t}(\ell)}_{i(\varepsilon_2,\ell)}: \ell<n^*,
{\frak t}(\ell)=0\}={\bf 0}.\] \end{description} Consequently we may
assume that
\[(\forall\ell<n^*)(\forall\varepsilon<\theta)({\frak t}(\ell)=0).\]
\medskip
\noindent{\bf Explanation of what we are going to do now:}\ \ \ We want to
replace the sequence $\langle z_\varepsilon:\varepsilon<\theta\rangle$ by a
large subsequence such that the places of splitting between two branches used
in two different $z_\varepsilon$'s will be uniform. Then we will be able to
translate our $\theta$--cc problem to the one on the algebras ${\Bbb B}_\eta$.
\medskip
Let
\[A_\varepsilon\stackrel{\rm def}{=}\{\nu\in {}^{\delta{>}}\mu: (\exists
j<\varepsilon)(\exists \ell<n^*)(\nu\vartriangleleft\eta_{i(j,\ell)})\}\]
and let $B_\varepsilon$ be the closure of $A_\varepsilon$:
\[\begin{array}{ll}
B_\varepsilon\stackrel{\rm def}{=}\{\rho\in {}^{\delta{\geq}}\mu: &\rho\in
A_\varepsilon \mbox{ or } {\rm lg}\/(\rho)\mbox{ is a limit ordinal and}\\
\ &(\forall\zeta<{\rm lg}\/(\rho))(\rho\rest\zeta\in A_\varepsilon)\}\\
\end{array}\]
Note that $|A_\varepsilon|\leq |\varepsilon|\cdot |\delta|$ and hence
$|B_\varepsilon|\leq |A_\varepsilon|^{{\leq}|\delta|}<\theta$. Next we define
(for $\varepsilon<\theta$, $\ell<n^*$):
\[\zeta(\varepsilon,\ell)\stackrel{\rm def}{=}\sup\{\zeta<\delta:
\eta_{i(\varepsilon,\ell)} \rest\zeta\in B_\varepsilon\}.\]
Thus $\zeta(\varepsilon,\ell)\leq{\rm lg}\/(\eta_{i(\varepsilon,\ell)})=\delta$. Let
$\S=\{\varepsilon<\theta:\cf(\varepsilon)>|\delta|\}$. For each
$\varepsilon\in\S$ we neccessarily have
\[\eta_{i(\varepsilon,\ell)}\rest\zeta(\varepsilon,\ell)\in
B_\varepsilon\quad\mbox{ and }\quad B_\varepsilon=\bigcup_{\xi<\varepsilon}
B_\xi\]
(remember that $\cf(\varepsilon)>|\delta|$ and for limit $\varepsilon$ we have
$A_\varepsilon=\bigcup\limits_{\xi<\varepsilon} A_\xi$) and hence
\[\eta_{i(\varepsilon,\ell)}\rest\zeta(\varepsilon,\ell)\in
B_{\xi(\varepsilon,\ell)},\quad\quad\mbox{ for some }\xi(\varepsilon,\ell)<
\varepsilon.\]
Let $\xi(\varepsilon)=\max\{\xi(\varepsilon,\ell):\ell<n^*\}$. By the Fodor
lemma we find $\xi^*<\theta$ such that the set
\[\S_1\stackrel{\rm def}{=}\{\varepsilon\in\S: \xi(\varepsilon)=\xi^*\}\]
is stationary. Thus $\eta_{i(\varepsilon,\ell)}\rest
\zeta(\varepsilon,\ell)\in B_{\xi^*}$ for each $\varepsilon\in\S_1$,
$\ell<n^*$. Since $|B_{\xi^*}|,|\delta|<\theta$ we find
$\nu_0,\ldots,\nu_{n^*-1}\in B_{\xi^*}$ and $\alpha(\ell_1,\ell_2)\leq\delta$
(for $\ell_1\leq\ell_2<n^*$) such that the set
\[\begin{array}{ll}
\S_2\stackrel{\rm def}{=}\{\varepsilon\in\S_1:& (\forall
\ell< n^*)(\eta_{i(\varepsilon,\ell)}{\rest}\zeta(\varepsilon,\ell)
=\nu_\ell)\ \&\\
\ &\&\ (\forall\ell_1\leq\ell_2<n^*)({\rm lg}\/(\eta_{i(\varepsilon,\ell_1)}\wedge
\eta_{i(\varepsilon,\ell_2)})=\alpha(\ell_1,\ell_2))\}\\
\end{array}\]
is stationary. Further, applying the $\Delta$-lemma we find a set $\S_3\in
[\S_2]^{\textstyle \theta}$ such that
\[\{\langle\eta_{i(\varepsilon,\ell)}({\rm lg}\/(\nu_\ell)): \ell<n^*\rangle:
\varepsilon\in\S_3\}\]
forms a $\Delta$-system of sequences.
For $\varepsilon\in\S_3$ and $\nu\in T$ denote
\[b^\varepsilon_\nu\stackrel{\rm def}{=}\bigcap\{y_{\eta_{i(\varepsilon,
\ell)}\rest ({\rm lg}\/(\nu)+1)}: \ell<n^*, \nu\vartriangleleft\eta_{i
(\varepsilon,\ell)}\}\in{\Bbb B}_\nu.\]
\begin{claim}
For each $\varepsilon\in\S_3$, $\nu\in T$ the element
$b^\varepsilon_\nu$ (of the algebra ${\Bbb B}_\nu$) is non-zero.
\end{claim}
\noindent{\em Why?}\ \ \ Because of the definition of ${\Bbb B}^{{\rm red}}$ and the
fact that $z_\varepsilon\neq{\bf 0}$:
\[b^\varepsilon_\nu={\bf 0}\ \ \ \Rightarrow\ \ \
\bigcap\{x_{\eta_{i(\varepsilon,\ell)}}: \ell<n^*, \nu\vartriangleleft\eta_{i
(\varepsilon,\ell)}\}={\bf 0} \ \ \ \Rightarrow\ \ \ z_\varepsilon={\bf 0}.\]
\medskip
Since for each $\ell<n^*$ the algebra ${\Bbb B}_{\nu_\ell}$ satisfies the
$\theta$--Knaster property we find a set $\S_4\in [\S_3]^{\textstyle \theta}$
such that for each $\ell<n^*$ and $\varepsilon_1,\varepsilon_2\in\S_4$ we have
\[\varepsilon_1\neq\varepsilon_2\ \ \ \Rightarrow\ \ \
b^{\varepsilon_1}_{\nu_\ell}\cap b^{\varepsilon_2}_{\nu_\ell}\neq{\bf 0}
\quad\mbox{ in }{\Bbb B}_{\nu_\ell}.\]
Now we may finish by proving the following claim.
\begin{claim}
For each $\varepsilon_1,\varepsilon_2\in\S_4$
\[{\Bbb B}^{{\rm red}}\models z_{\varepsilon_1}\cap z_{\varepsilon_2}\neq{\bf 0}.\]
\end{claim}
\noindent{\em Why?}\ \ \ Since $z_{\varepsilon_1}\cap z_{\varepsilon_2}$ is
just the intersection of generators it is enough to show that (remember the
definition of ${\Bbb B}^{{\rm red}}$):
\begin{description}
\item[$(\otimes)$] for each $\varepsilon_1,\varepsilon_2\in\S_4$ and for every
$\nu\in T$
\[{\Bbb B}_\nu\models\bigcap\{y_{\eta_i\rest({\rm lg}\/(\nu)+1)}:
i\in\{i(\varepsilon_1,\ell), i(\varepsilon_2,\ell): \ell<n^*\}
\mbox{ and } \nu\vartriangleleft\eta_i\}\neq{\bf 0}.\]
\end{description}
If $\nu=\nu_\ell$, $\ell<n^*$ then the intersection is
$b^{\varepsilon_1}_{\nu_\ell}\cap b^{\varepsilon_2}_{\nu_\ell}$ which by the
choice of the set $\S_4$ is not zero. So suppose that $\nu\notin\{\nu_\ell:
\ell<n^*\}$. Put
\[u_\nu\stackrel{\rm def}{=}\{i: \nu\vartriangleleft\eta_i\mbox{ and for some
}\ell<n^* \mbox{ either } i=i(\varepsilon_1,\ell) \mbox{ or }
i=i(\varepsilon_2,\ell)\}.\]
If
\[\{\eta_i({\rm lg}\/(\nu)): i\in u_\nu\}\subseteq\{\eta_{i(\varepsilon_2,\ell)}
({\rm lg}\/(\nu)):\ell<n^*\ \&\ \nu\vartriangleleft\eta_{i(\varepsilon_2,\ell)}\}\]
then we are done as $b^{\varepsilon_2}_\nu\neq{\bf 0}$. So there is $\ell_1<n^*$
such that $\nu\vartriangleleft\eta_{i(\varepsilon_1,\ell_1)}$ and
\[\eta_{i(\varepsilon_1,\ell_1)}\rest
({\rm lg}\/(\nu)+1)\notin\{\eta_{i(\varepsilon_2,\ell)}\rest ({\rm lg}\/(\nu)+1):
\ell<n^*\ \&\ \nu\vartriangleleft\eta_{i(\varepsilon_2,\ell)}\}.\]
Similarly we may assume that there is $\ell_2<n^*$ such that
$\nu\vartriangleleft\eta_{i(\varepsilon_2,\ell_2)}$ and
\[\eta_{i(\varepsilon_2,\ell_2)}\rest
({\rm lg}\/(\nu)+1)\notin\{\eta_{i(\varepsilon_1,\ell)}\rest ({\rm lg}\/(\nu)+1):
\ell<n^*\ \&\ \nu\vartriangleleft\eta_{i(\varepsilon_1,\ell)}\}.\]
Because of the symmetry we may assume that $\varepsilon_1<\varepsilon_2$. Then
\[\nu=\eta_{i(\varepsilon_2,\ell_2)}\rest{\rm lg}\/(\nu)\in
A_{\varepsilon_1+1}\subseteq B_{\varepsilon_2}\]
and hence $\zeta(\varepsilon_2,\ell_2)\geq{\rm lg}\/(\nu)$. By the choice of $\S_2$
(remember $\varepsilon_1,\varepsilon_2\in\S_4\subseteq\S_2$), we get
$\nu\trianglelefteq\nu_{\ell_2}$. But we have assumed that
$\nu\neq\nu_{\ell_2}$, so $\nu\vartriangleleft\nu_{\ell_2}$. Hence (once again
due to $\varepsilon_1,\varepsilon_2\in\S_2$)
\[\eta_{i(\varepsilon_2,\ell_2)}\rest ({\rm lg}\/(\nu)+1)=
\eta_{i(\varepsilon_1,\ell_2)}\rest ({\rm lg}\/(\nu)+1)=\nu_{\ell_2}\rest
({\rm lg}\/(\nu)+1)\]
which contradicts the choice of $\ell_2$.
The claim and so the lemma are proved. \QED
\begin{remark}
{\em
We can strengthen ``$\theta$-Knaster'' in the assumption and conclusion of
\ref{consknaster} in various ways. For example we may have that ``intersection
of any $n$ members of the final set is non-zero''.
}
\end{remark}
\begin{definition}
Let $({\Bbb B},\bar{y})$ be a $\lambda$-marked Boolean algebra,
$\kappa\leq\lambda$. We say that
\begin{enumerate}
\item $({\Bbb B},\bar{y})$ satisfies the $\kappa$-Knaster property if ${\Bbb B}$
satisfies the definition of the $\kappa$-Knaster property (see \ref{knaster})
with restriction to subsequences of $\bar{y}$.
\item $({\Bbb B},\bar{y})$ is $(\kappa,{\rm not}\lambda)$--Knaster if
\begin{description}
\item[(a)] the algebra ${\Bbb B}$ has the $\kappa$-Knaster property but
\item[(b)] the sequence $\bar{y}$ witnesses that the $\lambda$-Knaster
property fails for ${\Bbb B}$.
\end{description}
\end{enumerate}
\end{definition}
\begin{conclusion}
\label{knacon}
Assume that $\mu$ is a strong limit singular cardinal, $\lambda=2^\mu=\mu^+$
and $\theta=(2^{\cf(\mu)})^+$.
\noindent{\em Then} there exists a $\lambda$-marked Boolean algebra
$({\Bbb B},\bar{y})$ which is $(\theta,{\rm not}\lambda)$--Knaster.
\end{conclusion}
\Proof Choose cardinals $\mu_i^0,\mu_i<\mu$ (for $i<\cf(\mu)$) such that
\begin{description}
\item[$(\alpha)$] $\cf(\mu)<\mu^0_0$
\item[$(\beta)$] $\prod\limits_{j<i}\mu_j<\mu^0_i$, $\mu_i=(2^{\mu^0_i})^+$
\item[$(\gamma)$] the sequences $\langle\mu_i: i<\cf(\mu)\rangle$,
$\langle\mu_i^0: i<\cf(\mu)\rangle$ are increasing cofinal in $\mu$
\end{description}
(possible as $\mu$ is strong limit singular). By proposition \ref{algide} we
find $\mu_i$-marked Boolean algebras $({\Bbb B}_i,\bar{y}^i)$ and
$(\mu^0_i)^+$-complete ideals $I_i$ on $\mu_i$ (for $i<\delta$) such that
\begin{description}
\item[(a)] if $X\subseteq\mu_i$, $X\notin I_i$ then $(\exists\alpha,\beta \in
X)({\Bbb B}_i\models y^i_\alpha\cap y^i_\beta={\bf 0})$
\item[(b)] the algebra ${\Bbb B}_i$ has the $(2^{\cf(\mu)})^+$--Knaster property.
\end{description}
Let $T=\bigcup\limits_{i<\cf(\mu)}\prod\limits_{j<i}\mu_j$ and for $\nu\in
T_i$ ($i<\cf(\mu)$) let $I_\nu=I_i$, ${\Bbb B}_\nu={\Bbb B}_i$, $\bar{y}_\nu=\bar{y}^i$
and $\lambda_\nu=\mu_i$. Now we may apply proposition \ref{stronger} to
$\mu$, $\langle\mu^0_i: i<\cf(\mu)\rangle$ and $(T,\bar{\lambda},\bar{I})$ to
find a stronger $J^{{\rm bd}}_{\cf(\mu)}$-cofinal sequence $\bar{\eta}$ for
$(T,\bar{\lambda},\bar{I})$ of the length $\lambda$. Consider the
$(\cf(\mu),\mu,\lambda)$-constructor ${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle
({\Bbb B}_\nu,\bar{y}_\nu): \nu\in T\rangle)$. By {\bf (b)} above we may apply
lemma \ref{consknaster} to get that the algebra ${\Bbb B}^{{\rm red}}({\cal C})$
satisfies the $(2^{\cf(\mu)})^+$--Knaster condition. Finally we use
proposition \ref{notknaster} (and {\bf (a)} above) to conclude that
$({\Bbb B}^{{\rm red}}({\cal C}),\langle x^{{\rm red}}_\alpha:\alpha<\lambda\rangle)$ is
$(\theta,{\rm not}\lambda)$--Knaster. \QED
\begin{proposition}
\label{clospcf}
Assume that:\\
$\kappa$ is a regular cardinal such that
$(\forall\alpha<\kappa)(|\alpha|^{|\delta|}<\kappa)$, $\bar{\lambda}=
\langle\lambda_i: i<\delta\rangle$ is an increasing sequence of regular
cardinals such that $\kappa\leq\lambda_0$,
$\prod\limits_{j<i}\lambda_j<\lambda_i$ (or just $\max{\rm pcf}\/\{\lambda_j: j<i\}<
\lambda_i$) for $i<\delta$ and $\lambda\in{\rm pcf}\/\{\lambda_i: i<\delta\}$.
Further suppose that for each $i<\delta$ there exists a $\lambda_i$-marked
Boolean algebra which is $(\kappa,{\rm not}\lambda_i)$--Knaster.
\noindent Then there exists a $\lambda$-marked Boolean algebra which is
$(\kappa,{\rm not}\lambda)$--Knaster.
\end{proposition}
\Proof If $\lambda=\lambda_i$ for some $i<\delta$ then there is nothing to do.
If $\lambda<\lambda_i$ for some $i<\delta$ then let $\alpha<\delta$ be the
maximal limit ordinal such that $(\forall i<\alpha)(\lambda_i<\lambda)$ (it
neccessarily exists) . Now we may replace $\langle_i:i<\delta\rangle$ by
$\langle \lambda_i: i<\alpha\rangle$. Thus we may assume that $(\forall
i<\delta)(\lambda_i<\lambda)$. Further we may assume that
\[\lambda=\max{\rm pcf}\/\{\lambda_i: i<\delta\}\]
(by \cite{Sh:g}, I, 1.8). Now, due to \cite{Sh:g}, II, 3.5, p.65, we find a
sequence $\bar{\eta}\subseteq\prod\limits_{i<\delta}\lambda_i$ and an ideal
$J$ on $\delta$ such that
\begin{enumerate}
\item $J\supseteq J^{{\rm bd}}_\delta$ and $\lambda={\rm tcf}(\prod\limits_{i<\delta}
\lambda_i/J)$
\noindent (naturally: $J=\{a\subseteq\delta: \max{\rm pcf}\/\{\lambda_i: i\in a\}<
\lambda\}$)
\item $\bar{\eta}=\langle\eta_\varepsilon:\varepsilon<\lambda\rangle$ is
${<}_J$-increasing cofinal in $\prod\limits_{i<\delta}\lambda_i/J$
\item for each $i<\delta$
\[|\{\eta_\varepsilon\rest i: \varepsilon<\lambda\}|<\lambda_i.\]
\end{enumerate}
Let $T=\bigcup\limits_{i<\delta}\prod\limits_{j<i}\lambda_j$ and for
$\nu\in T_i$ ($i<\delta$) let $\lambda_\nu=\lambda_i$,
$I_\nu=J^{{\rm bd}}_{\lambda_i}$.
\noindent It follows from the choice of $\bar{\eta}, J$ above and our
assumptions that we may apply proposition \ref{getcofinal} and hence
$\bar{\eta}$ is $J$-cofinal for $(T,\bar{\lambda},\bar{I})$. For $\nu\in T$
let $({\Bbb B}_\nu,\bar{y}_\nu)$ be a $\lambda_\nu$-marked $(\kappa,{\rm
not}\lambda_\nu)$--Knaster Boolean algebra (exists by our assumptions). Now we
may finish using \ref{consknaster} and \ref{notknaster} for
${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta):\eta\in
T\rangle)$, $\bar{I}$ and $J$ (note the assumption {\bf (b)} of
\ref{notknaster} is satisfied as $I_\eta=J^{{\rm bd}}_{\lambda_\eta}$; remember the
choice of $({\Bbb B}_\eta,\bar{y}_\eta)$). \QED
\begin{remark}
{\em
Note that from cardinal arithmetic hypothesis $\cf(\mu)=\chi$, $\chi^{{<}\chi}
<\chi<\mu$, $\mu^+=\lambda<2^\chi$ alone we cannot hope to build a
counterexample. This is because of section 4 of \cite{Sh:93}, particularly
lemma 4.13 there. It was shown in that paper that if $\chi^{{<}\chi}<\chi_1=
\chi_1^{{<}\chi_1}$ then there is a $\chi^+$-cc $\chi$-complete forcing notion
${\Bbb P}$ of size $\chi_1$ such that
\[\begin{array}{ll}
\forces_{{\Bbb P}}&\mbox{``if }|{\Bbb B}|<\chi_1, {\Bbb B}\models \chi\mbox{-cc}\\
\ &\mbox{then }{\Bbb B}^+\mbox{ is the union of }\chi\mbox{ ultrafilters''}.\\
\end{array}\]
More on this see in section 8.
\noindent So the centrality of $\lambda\in{\rm Reg}\cap (\mu,2^\mu]$, $\mu$ strong
limit singular, is very natural.
}
\end{remark}
\section{The main result}
\begin{proposition}
\label{notcc}
Suppose that ${\cal C}$ is a $(\delta,\mu,\lambda)$-constructor. Then the free
product ${\Bbb B}^{{\rm red}}({\cal C})*{\Bbb B}^{{\rm green}}({\cal C})$ fails the $\lambda$-cc (so
${\rm c}({\Bbb B}^{{\rm red}}({\cal C})*{\Bbb B}^{{\rm green}}({\cal C}))\geq\lambda$).
\end{proposition}
\Proof Look at the elements $(x_i,x_i)\in{\Bbb B}^{{\rm red}}*{\Bbb B}^{{\rm green}}$ for
$i<\lambda$. It follows directly from the definition of the algebras that for
each $i<j<\lambda$:
\[\mbox{either }\quad {\Bbb B}^{{\rm red}}\models x_i^{{\rm red}}\cap x_j^{{\rm red}}={\bf 0}
\quad\mbox{ or }\quad {\Bbb B}^{{\rm green}}\models x_i^{{\rm green}}\cap x_j^{{\rm green}}={\bf 0}.\]
Consequently the sequence $\langle (x_i,x_i):i<\lambda\rangle$ witnesses the
assertion of the proposition. \QED
\begin{proposition}
\label{finite}
Suppose that $n<\omega$ and for $\ell\leq n$:
\begin{enumerate}
\item $\chi_\ell,\lambda_\ell$ are regular cardinals, $\chi_\ell<\lambda_\ell
<\chi_{\ell+1}$
\item $({\Bbb B}_\ell,\bar{y}_\ell, I_\ell)$ is a $(\lambda_\ell,\chi_\ell)$-well
marked Boolean algebra (see definition \ref{well1}), $\bar{y}_\ell= \langle
y^\ell_i: i<\lambda_\ell\rangle$
\item ${\Bbb B}$ is the Boolean algebra freely generated by $\{y_\eta:
\eta\in\prod\limits_{\ell\leq n}\lambda_\ell\}$ {\em except that}
\begin{quotation}
\noindent {\em if} $\eta_{i_0},\ldots,\eta_{i_{k-1}}\in\prod\limits_{\ell\leq
n}\lambda_\ell$, $\eta_{i_0}\rest \ell=\eta_{i_1}\rest\ell=\ldots=
\eta_{i_{n-1}}\rest\ell$ and\\
${\Bbb B}_\ell\models\bigcap\limits_{m<k} y^\ell_{\eta_{i_m}(\ell)}={\bf 0}$
\noindent {\em then} $\bigcap\limits_{m<k} y_{\eta_{i_m}}={\bf 0}$.
\end{quotation}
[Compare to the definition of the algebras ${\Bbb B}^{{\rm red}}({\cal C})$.]
\item $I=\{B\subseteq\prod\limits_{\ell\leq n}\lambda_\ell: \neg
(\exists^{I_0}\gamma_0)\ldots(\exists^{I_n}\gamma_n)(\langle\gamma_0,\ldots,
\gamma_n\rangle\in B)\}$.
\end{enumerate}
{\em Then:}
\begin{description}
\item[(a)] if all the algebras ${\Bbb B}_\ell$ (for $\ell\leq n$) satisfy the
$\theta$-Knaster property, $\theta$ is a regular uncountable cardinal
then ${\Bbb B}$ has the $\theta$-Knaster property;
\item[(b)] $I$ is a $\chi_0$-complete ideal on $\prod\limits_{\ell\leq
n}\lambda_i$;
\item[(c)] if $Y\subseteq (\prod\limits_{\ell\leq n}\lambda_\ell)^n$ is such
that
\[(\exists^{I}\eta_0)\ldots(\exists^{I}\eta_n)(\langle\eta_0,\ldots,
\eta_n\rangle\in Y)\]
then there are $\langle\eta_0^\prime,\ldots,\eta_n^\prime\rangle,
\langle\eta_0^{\prime\prime},\ldots,\eta_n^{\prime\prime}\rangle \in Y$ such
that for all $\ell\leq n$
\[{\Bbb B}\models y_{\eta^\prime_\ell}\cap y_{\eta^{\prime\prime}_\ell}={\bf 0}.\]
\end{description}
\end{proposition}
\Proof {\bf (a)}\ \ \ The proof that the algebra ${\Bbb B}$ satisfies
$\theta$-Knaster condition is exactly the same as that of \ref{consknaster}
(actually it is a special case of that).
\noindent{\bf (b)}\ \ \ Should be clear.
\noindent{\bf (c)}\ \ \ For $\ell,m\leq n$ put
\[\chi^m_\ell=\chi_\ell, \lambda^m_\ell=\lambda_\ell,\ I^m_\ell= I_\ell,\
P^m_\ell= \{\{\alpha,\beta\}\subseteq\lambda_\ell: {\Bbb B}_\ell\models
y^\ell_\alpha\cap y^\ell_\beta={\bf 0}\}, B=Y.\]
It is easy to check that the assumptions of proposition \ref{fubini} are
satisfied. Applying it we find sets $X_0,\ldots,X_n$ satisfying the respective
versions of clauses {\bf (a)}--{\bf (d)} there. Note that our choice of the
sets $P^m_\ell$ and clauses {\bf (b)}, {\bf (c)} of \ref{fubini} imply that
\begin{quotation}
\noindent $X_m=\{\nu_m^\prime,\nu_m^{\prime\prime}\}\subseteq\prod
\limits_{\ell\leq n-m}\lambda_\ell$
\noindent $\nu_m^\prime\rest (n-m)=\nu_m^{\prime\prime}\rest (n-m)$
\noindent ${\Bbb B}_{n-m}\models y^{n-m}_{\nu_m^{\prime}(n-m)}\cap
y^{n-m}_{\nu_m^{\prime\prime}(n-m)}={\bf 0}$
\end{quotation}
Look at the sequences $\langle\nu^\prime_0,\ldots,\nu^\prime_n\rangle$,
$\langle\nu^{\prime\prime}_0,\ldots,\nu^{\prime\prime}_n\rangle$. By the
clause {\bf (d)} of \ref{fubini} we find $\langle\eta^\prime_0,\ldots,
\eta^\prime_m\rangle,\langle\nu^{\prime\prime}_0,\ldots,
\nu^{\prime\prime}_n\rangle\in Y$ such that for each $m\leq n$
\[\nu_m^\prime\trianglelefteq\eta^\prime_m,\quad \nu_m^{\prime\prime}
\trianglelefteq\eta^{\prime\prime}_m.\]
Now, the properties of $\nu_m^\prime$, $\nu^{\prime\prime}_m$ and the
definition of the algebra ${\Bbb B}$ imply that for each $m\leq n$:
\[{\Bbb B}\models y_{\eta^\prime_m}\cap y_{\eta_m^{\prime\prime}}={\bf 0},\]
finishing the proof. \QED
\begin{lemma}
\label{green}
Assume that $\lambda$ is a regular cardinal, $|\delta|<\lambda$, $J$ is an
ideal on $\delta$ extending $J^{{\rm bd}}_\delta$,
${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta): \eta\in
T\rangle)$ is a $(\delta,\mu,\lambda)$-constructor and $\bar{I}$ is such that
$(T,\bar{\lambda},\bar{I})\in {\cal K}^{{\rm id}}_{\delta,\mu}$. Suppose that
$\bar{\eta}=\langle \eta_\alpha: \alpha<\lambda\rangle$ is a stronger
(or big) $J$-cofinal in $(T,\bar{\lambda},\bar{I})$ sequence such that
\[(\forall i<\delta)(|\{\eta_\alpha\rest i: \alpha<\lambda\}|<\lambda).\]
Further assume that
\begin{description}
\item[$(\circleddash)$] for every $n<\omega$ for a $J$-positive set of
$i<\delta$ we have:
\begin{quotation}
\noindent {\em if} $\eta_0,\ldots,\eta_n\in T_i$ are pairwise distinct and the
set $Y\subseteq\prod\limits_{\ell\leq n}\lambda_{\eta_\ell}$ is such that
\[(\exists^{I_{\eta_0}}\gamma_0)\ldots(\exists^{I_{\eta_n}}\gamma_n)(\langle
\gamma_0,\ldots,\gamma_n\rangle\in Y)\]
\noindent{\em then} for some $\gamma_\ell^\prime,\gamma_\ell^{\prime\prime}<
\lambda_{\eta_\ell}$ (for $\ell\leq n$) we have
\[\langle\gamma_\ell^\prime:\ell\leq n\rangle, \langle\gamma_\ell^{\prime
\prime}:\ell\leq n\rangle\in Y\quad\mbox{ and for all }\ell\leq n\]
\[{\Bbb B}_{\eta_{\ell}}\models y_{\eta_\ell\hat{\ }\langle\gamma^\prime_\ell
\rangle}\cap y_{\eta_\ell\hat{\ }\langle\gamma^{\prime\prime}_\ell\rangle}=
{\bf 0}.\]
\end{quotation}
\end{description}
{\em Then} the Boolean algebra ${\Bbb B}^{{\rm green}}({\cal C})$ satisfies $\lambda$-cc.
\end{lemma}
\Proof Suppose that $\langle z_\alpha: \alpha<\lambda\rangle\subseteq
{\Bbb B}^{{\rm green}}\setminus\{{\bf 0}\}$. By the standard cleaning (compare the first part
of the proof of \ref{consknaster}) we may assume that there are $n^*\in\omega$
and a function $\varepsilon:\lambda\times n^*\longrightarrow\lambda$ such that
\begin{enumerate}
\item $z_\alpha=\bigcap\limits_{\ell<n^*}x_{\varepsilon(\alpha,\ell)}$ (in
${\Bbb B}^{{\rm green}}$)
\item $\varepsilon(\alpha,0)<\varepsilon(\alpha,1)<\ldots<
\varepsilon(\alpha,n^*-1)$
\item $\langle\langle\varepsilon(\alpha,\ell):\ell<n^*\rangle:
\alpha<\lambda\rangle$ forms a $\Delta$-system of sequences with the kernel
$m^*$, i.e. $(\forall \ell<m^*)(\varepsilon(\alpha,\ell)=\varepsilon(\ell))$
and
\[(\forall \ell\in [m^*,n^*))(\forall\alpha<\lambda)(\varepsilon(\alpha,\ell)
\notin\{\varepsilon(\beta,k):(\beta,k)\neq (\alpha,\ell)\})\]
\item there is $i^*<\delta$ such that for each $\alpha<\lambda$ there is no
repetition in the sequence $\langle\eta_{\varepsilon(\alpha,\ell)}\rest i^*:
\ell<n^*\rangle$.
\end{enumerate}
Since $|\{\eta_\alpha\rest i: \alpha<\lambda\}|<\lambda$ (for $i<\delta$) and
$|\delta|<\lambda$ we may additionally require that
\begin{description}
\item[$\hat{(*)}$] for each $i<\delta$, for every $\alpha<\lambda$ we have
\[(\exists^\lambda \beta<\lambda)(\forall \ell<n^*)(\eta_{\varepsilon(\alpha,
\ell)}\rest (i+1)=\eta_{\varepsilon(\beta,\ell)}\rest (i+1))\]
and
\item[$\hat{(**)}$] for each $\alpha<\beta<\lambda$, $\ell<n^*$
\[\eta_{\varepsilon(\alpha,\ell)}\rest i^*=\eta_{\varepsilon(\beta,\ell)}
\rest i^*.\]
\end{description}
\medskip
\noindent{\bf Remark}: \ \ \ \ Note that the claim below is like an
$(n^*-m^*)$-place version of \ref{notknaster}. Having an $(n^*-m^*)$-ary
version is extra for the construction but it also costs.
\begin{claim}
\label{claimgreen}
{\em Assume that:}
\noindent ${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta):
\eta\in T\rangle$ is a $(\delta,\mu,\lambda)$-constructor, $\lambda$ a regular
cardinal, $\delta<\lambda$, $\bar{I}$ is such that
$(T,\bar{\lambda},\bar{I})\in {\cal K}^{{\rm id}}_{\delta,\mu}$, $J$ is an ideal on
$\delta$ extending $J^{{\rm bd}}_\delta$ and the sequence $\bar{\eta}$ is stronger
$J$-cofinal in $(T,\bar{\lambda},\bar{I})$.\\
Further suppose that $\varepsilon:\lambda\times n^*\longrightarrow$, $m^*,n^*$
and $i^*<\delta$ are as above (after the reduction, but the property
$\hat{(**)}$ is not needed).
\noindent {\em Then}
\begin{description}
\item[$(\boxtimes)$] for every large enough $\alpha<\lambda$ the set:
\[\hspace{-1cm}\begin{array}{l}
Z_\alpha\stackrel{\rm def}{=}\{i<\delta:
\neg(\exists^{I_{\eta_{\varepsilon(\alpha,m^*)}\rest i}}
\gamma_{m^*})(\exists^{I_{\eta_{\varepsilon(\alpha,m^*+1)}\rest i}}
\gamma_{m^*+1})\ldots\\
\ \ldots(\exists^{I_{\eta_{\varepsilon(\alpha,n^*-1)}\rest i}}
\gamma_{n^*-1})(\exists^\lambda \beta)(\forall \ell\in [m^*,n^*))
(\eta_{\varepsilon(\beta,\ell)}\rest (i{+}1)=(\eta_{\varepsilon(\alpha,\ell)}
\rest i)\hat{\ }\langle\gamma_\ell\rangle)\}\\
\end{array}
\]
is in the ideal $J$.
\end{description}
\end{claim}
\noindent{\em Why?}\ \ \ For each $i<\delta$, $i\geq i^*$ and distinct
sequences $\nu_{m^*},\ldots,\nu_{n^*-1}\in T_i$ define
\[\begin{array}{ll}
B_{\langle\nu_\ell:\ell\in [m^*,n^*)\rangle}\stackrel{\rm def}{=}
\{\bar{\gamma}: &\bar{\gamma}=\langle\gamma_\ell: \ell\in [m^*,n^*)\rangle
\mbox{ and}\\
\ &\mbox{for arbitrarly large }\alpha<\lambda\mbox{ for all }m^*\leq\ell<n^*\\
\ &\nu_\ell\hat{\ }\langle\gamma_\ell\rangle\vartriangleleft
\eta_{\varepsilon(\alpha,\ell)}\}.\\
\end{array}\]
We will call a sequence $\langle\nu_\ell: \ell\in [m^*,n^*)\rangle$ {\em a
success} if
\[(\exists^{I_{\nu_{m^*}}}\gamma_{m^*})\ldots(\exists^{I_{\nu_{n^*-1}}}
\gamma_{n^*-1})(\langle\gamma_\ell: \ell\in [m^*,n^*)\rangle\in
B_{\langle\nu_\ell\in [m^*,n^*)\rangle}).\]
Using this notion we may reformulate $(\boxtimes)$ (which we have to prove) to
\begin{description}
\item[$(\boxtimes^*)$] for every large enough $\alpha<\lambda$, for
$J$-majority of $i<\delta$, $i>i^*$ the sequence
$\langle\eta_{\varepsilon(\alpha,\ell)} \rest i: \ell\in [m^*,n^*)\rangle$ is
a success.
\end{description}
\noindent To show $(\boxtimes^*)$ note that if a sequence
$\langle\nu_\ell:\ell\in [m^*,n^*)\rangle$ is not a success then there are
functions $f^k_{\langle\nu_\ell: \ell\in [m^*,n^*)\rangle}$ (for $m^*\leq
k<n^*$) such that
\[f^k_{\langle\nu_\ell: \ell\in [m^*,n^*)\rangle}:\prod_{\ell=m^*}^{k-1}
\lambda_{\nu_\ell}\ \longrightarrow\ I_{\nu_k}\quad\mbox{ and}\]
\[\begin{array}{ll}
\mbox{if }&\langle\gamma_\ell:\ell\in [m^*,n^*)\rangle\in B_{\langle\nu_\ell:
\ell\in [m^*,n^*)\rangle}\\
\mbox{then }&(\exists k\in [m^*,n^*))(\gamma_k\in f^k_{\langle\nu_\ell:\ell\in
[m^*,n^*)\rangle}(\gamma_{m^*},\ldots,\gamma_{k-1})).\\
\end{array}\]
If $\langle\nu_\ell: \ell\in [m^*,n^*)\rangle$ is a success then we declare
that $f^k_{\langle\nu_\ell: \ell\in [m^*,n^*)\rangle}$ is constantly equal to
$\emptyset$.
\noindent Now we may finish the proof of the claim applying clause {\bf (b)} of
definition \ref{cofinal}(5) to $n^*-1$ and functions $F_0,\ldots,F_{n^*-1}$
such that for $k\in [m^*,n^*)$
\[F_k(\nu_0\hat{\ }\langle\gamma_0\rangle,\ldots, \nu_{k-1}\hat{\ }\langle
\gamma_{k-1}\rangle,\nu_k,\ldots,\nu_{n^*-1}\rangle)=
f^k_{\langle\nu_\ell: \ell\in [m^*,n^*)\rangle}(\gamma_{m^*},\ldots,
\gamma_{k-1}).\]
This gives us a suitable $\alpha^*<\lambda$. Suppose $\varepsilon(\alpha,m^*)
\geq\alpha^*$. Then for $J$-majority of $i<\delta$ for each $k\in [m^*,n^*)$
we have
\medskip
if
\[F_m(\eta_{\varepsilon(\alpha,0)}\rest (i+1),\ldots,
\eta_{\varepsilon(\alpha,k-1)}\rest (i+1),\eta_{\varepsilon(\alpha,k)}\rest i,
\ldots,\eta_{\varepsilon(\alpha,n^*-1)}\rest i )\in
I_{\eta_{\varepsilon(\alpha,k)\rest i}}\]
then
\[\eta_{\varepsilon(\alpha,k)}\rest (i+1)\notin
F_m(\eta_{\varepsilon(\alpha,0)}\rest (i+1),\ldots,
\eta_{\varepsilon(\alpha,k-1)}\rest (i+1),\eta_{\varepsilon(\alpha,k)}\rest i
,\ldots,\eta_{\varepsilon(\alpha,n^*-1)}\rest i ).\]
\smallskip
\noindent But the choice of the functions $F_k$ implies that thus for
$J$-majority of $i<\delta$, for each $k\in [m^*,n^*)$
\[\eta_{\varepsilon(\alpha,k)}(i)\notin
f^k_{\langle\eta_{\varepsilon(\alpha,\ell)}\rest i: \ell\in [m^*,n^*)\rangle}
(\eta_{\varepsilon(\alpha,m^*)}(i),\ldots,\eta_{\varepsilon(\alpha,
k-1)}(i)).\]
Now the definition of the function $f^k_{\langle\nu_\ell: \ell\in
[m^*,n^*)\rangle}$ works:
if for some relevant $i<\delta$ above the sequence
$\langle\eta_{\varepsilon(\alpha,\ell)}\rest i: \ell\in [m^*,n^*)\rangle$ is
not a success then $\langle\eta_{\varepsilon(\alpha,\ell)}(i): \ell\in
[m^*,n^*)\rangle\notin B_{\langle\eta_{\varepsilon(\alpha,\ell)}\rest i:
\ell\in [m^*,n^*)\rangle}$ and this contradicts $\hat{*}$ before.
The claim is proved.
\medskip
Let $\alpha^*$ be such that for each $\alpha\geq\alpha^*$ we have $Z_\alpha\in
J$. Choose $i\in\delta\setminus Z_{\alpha^*}$ such that the clause
$(\circleddash)$ applies for $n^*-m^*$ and $i$. Let
\[Y\stackrel{\rm def}{=}\{\langle\gamma_{m^*},\ldots,\gamma_{n^*-1}\rangle\!:
(\exists^\lambda\beta)(\forall \ell{\in}[m^*,n^*))(\eta_{\varepsilon
(\beta,\ell)}\rest (i{+}1){=} (\eta_{\varepsilon(\alpha^*,\ell)}\rest i)\hat{\
}\langle\gamma_\ell\rangle)\}.\]
The definition of $Z_{\alpha^*}$ (and the choice of $i$) imply that the
assumption $(\circleddash)$ applies to the set $Y$ and we get
$\gamma_\ell^\prime,\gamma_\ell^{\prime\prime}<
\lambda_{\eta_{\varepsilon(\alpha^*,\ell)}\rest i}$ (for $m^*\leq\ell<n^*$)
such that
\[\langle\gamma^\prime_\ell:m^*\leq\ell<n^*\rangle,\langle
\gamma^{\prime\prime}_\ell:m^*\leq\ell<n^*\rangle\in Y\quad\mbox{ and}\]
\[{\Bbb B}_{\eta_{\varepsilon(\alpha^*,\ell)}\rest i}\models y_{\eta_{\varepsilon(
\alpha^*,\ell)}\rest i\hat{\ }\langle\gamma^\prime_\ell\rangle} \cap
y_{\eta_{\varepsilon(\alpha^*,\ell)}\rest i\hat{\ }\langle
\gamma^{\prime\prime}_\ell\rangle}={\bf 0}\quad\mbox{for }m^*\leq\ell< n^*.\]
Now choose $\alpha<\beta<\lambda$ such that for $m^*\leq\ell<n^*$
\[\eta_{\varepsilon(\alpha^*,\ell)}\rest i\hat{\ }\langle\gamma_\ell^\prime
\rangle = \eta_{\varepsilon(\alpha,\ell)}\rest (i+1),\quad
\eta_{\varepsilon(\alpha^*,\ell)}\rest i\hat{\ }\langle
\gamma_\ell^{\prime\prime}\rangle=\eta_{\varepsilon(\beta,\ell)}\rest (i+1)\]
(possible by the choice of $Y$ and $\gamma^\prime_\ell,
\gamma^{\prime\prime}_\ell$). The definition of the algebra ${\Bbb B}^{{\rm green}}({\cal C})$
and the choice of $\gamma^\prime_\ell,\gamma^{\prime\prime}_\ell$ imply that
for $m^*\leq\ell<n^*$
\[{\Bbb B}^{{\rm green}}({\cal C})\models x_{\varepsilon(\alpha,\ell)}\cap
x_{\varepsilon(\beta,\ell)}\neq {\bf 0}.\]
If $\ell\neq m$ then
\[{\Bbb B}^{{\rm green}}({\cal C})\models x_{\varepsilon(\alpha,\ell)}\cap
x_{\varepsilon(\beta,m)}\neq {\bf 0}\]
by the conditions $\hat{(**)}$ and 4) of the preliminary cleaning (and the
definition of ${\Bbb B}^{{\rm green}}({\cal C})$, remember $z_\alpha\neq{\bf 0}$). Finally,
remembering that $\varepsilon(\alpha,\ell)=\varepsilon(\beta,\ell)$ for
$\ell<m^*$, $z_\alpha\neq{\bf 0}$ and $z_\beta\neq{\bf 0}$, we may conclude that
\[{\Bbb B}^{{\rm green}}({\cal C})\models \bigcap_{\ell<n^*} x_{\varepsilon(\alpha,\ell)}\cap
\bigcap_{\ell<n^*} x_{\varepsilon(\beta,\ell)}\neq {\bf 0}\]
finishing the proof. \QED
\begin{theorem}
\label{main}
{\em If} $\mu$ is a strong limit singular cardinal, $\lambda\stackrel{\rm
def}{=}2^\mu=\mu^+$
\noindent {\em then} there are Boolean algebras ${\Bbb B}_1,{\Bbb B}_2$ such that
the algebra ${\Bbb B}_1$ satisfies the $\lambda$-cc, the algebra ${\Bbb B}_2$ has the
$(2^{\cf(\mu)})^+$--Knaster property but the free product ${\Bbb B}_1*{\Bbb B}_2$ does
not satisfy the $\lambda$-cc.
\end{theorem}
\Proof Let $\delta=\cf(\mu)$ and let $h:\delta\longrightarrow\omega$ be a
function such that
\[(\forall n\in\omega)(\exists^\delta i)(h(i)=n).\]
Choose an increasing sequence $\langle\mu_i:i<\delta\rangle$ of regular
cardinals such that $\mu=\sum\limits_{i<\delta}\mu_i$. Next, by induction on
$i<\delta$ choose $\lambda_i$, $\chi_i$, $({\Bbb B}_i,\bar{y}_i)$ and $I_i$ such
that
\begin{enumerate}
\item $\lambda_i,\chi_i$ are regular cardinals below $\mu$
\item $\lambda_i>\chi_i\geq\prod\limits_{j<i}\lambda_j + \mu_i$
\item $I_i$ is a $\chi_i^+$-complete ideal on $\lambda_i$ (containing all
singletons)
\item $({\Bbb B}_i,\bar{y}_i)$ is a $\lambda_i$-marked Boolean algebra such that
\noindent {\em if} $n=h(i)$ and the set $Y\subseteq (\lambda_i)^{n+1}$ is such
that
\[(\exists^{I_i}\gamma_0)\ldots(\exists^{I_i}\gamma_n)(\langle
\gamma_0,\ldots,\gamma_n\rangle\in Y)\]
\noindent{\em then} for some $\gamma_\ell^\prime,\gamma_\ell^{\prime\prime}<
\lambda_i$ (for $\ell\leq n$) we have
\[\langle\gamma_\ell^\prime:\ell\leq n\rangle, \langle\gamma_\ell^{\prime
\prime}:\ell\leq n\rangle\in Y\quad\mbox{ and for all }\ell\leq n\]
\[{\Bbb B}_{i}\models y^i_{\gamma^\prime_\ell}\cap y^i_{\gamma^{\prime
\prime}_\ell}={\bf 0}.\]
\item Each algebra ${\Bbb B}_i$ satisfies the $(2^{|\delta|})^+$--Knaster
condition.
\end{enumerate}
Arriving at the stage $i$ of the construction first we put $\chi_i=
(\prod\limits_{j<i}\lambda_j+\mu_i)^+$. Next we define inductively
$\chi_{i,k},\lambda_{i,k}$ for $k\leq h(i)$ such that
\[\chi_{i,0}=\chi_i,\quad \lambda_{i,k}=(2^{\chi_{i,k}})^+,\quad
\chi_{i,k+1}=(\lambda_{i,k})^+.\]
By \ref{algide}, for each $k\leq h(i)$ we find a
$(\lambda_{i,k},\chi_{i,k}^+)$--well marked Boolean algebra
$({\Bbb B}_{i,k},\bar{y}_{i,k},I_{i,k})$ such that ${\Bbb B}_{i,k}$ has the
$(2^\delta)^+$--Knaster property (compare \ref{well2}). Let
$\lambda_i=\lambda_{i,h(i)}$. Proposition \ref{finite} applied to
$\langle({\Bbb B}_{i,k},\bar{y}_{i,k},I_{i,k}): k\leq h(i)\rangle$ provides a
$\lambda_i$-marked Boolean algebra $({\Bbb B}_i,\bar{y}_i)$ and a
$\chi_i^+$-complete ideal $I_i$ on $\lambda_i$ such that the requirements 4,5
above are satisfied.
\medskip
Now put $T=\bigcup\limits_{j<\delta}\prod\limits_{i<j}\lambda_i$ and for
$\eta\in T$:
\[{\Bbb B}_\eta={\Bbb B}_{{\rm lg}\/(\eta)},\ \bar{y}_\eta=\bar{y}_{{\rm lg}\/(\eta)},\
I_\eta=I_{{\rm lg}\/(\eta)}.\]
By \ref{stronger} we find a stronger $J^{{\rm bd}}_\delta$-cofinal sequence
$\bar{\eta}=\langle \eta_\alpha: \alpha<\lambda\rangle$ for
$(T,\bar{\lambda},\bar{I})$. Take the $(\delta,\mu,\mu^+)$-conctructor ${\cal C}$
determined by these parameters. Look at the algebras ${\Bbb B}_2={\Bbb B}^{{\rm red}}({\cal C})$,
${\Bbb B}_1={\Bbb B}^{{\rm green}}({\cal C})$. Applying \ref{notcc} we get that ${\Bbb B}_1 *{\Bbb B}_2$
fails the $\lambda$-cc. The choice of the function $h$ and the requirement 4
above allow us to apply \ref{green} to conclude that the algebra ${\Bbb B}_2$
satisfies $\lambda$-cc. Finally, by \ref{consknaster}, we have that ${\Bbb B}_1$
has the $(2^\delta)^+$--Knaster property. \QED
\begin{remark}
{\em
\begin{enumerate}
\item We shall later give results not using $2^\mu=\mu^+$ but still not in ZFC
\item Applying the methods of \cite{Sh:576} one can the consistency of:
{\em for some $\mu$ strong limit singular there is no example for
$\lambda=\mu^+$}.
\item If we want ``for no regular $\lambda\in [\mu,2^\mu]$'' more is needed, we
expect the consistency, but it is harder (not speaking of ``for all $\mu$'')
\item Remark 1) above shows that $2^\mu>\mu^+$ is not enough for the negative
result.
\end{enumerate}
}
\end{remark}
\section{Toward improvements}
\begin{definition}
\label{super}
Let $(T,\bar{\lambda},\bar{I})\in{\cal K}_{\mu,\delta}^{{\rm id}}$ and let $J$ be an
ideal on $\delta$ (including $J^{{\rm bd}}_\delta$, as usual). We say that a
sequence $\bar{\eta}=\langle\eta_\alpha: \alpha<\lambda\rangle$ of
$\delta$-branches through $T$ is {\em super $J$-cofinal for}
$(T,\bar{\lambda},\bar{I})$ if
\begin{description}
\item[(a)] $\eta_\alpha\neq\eta_\beta$ for distinct $\alpha,\beta<\lambda$
\item[(b)] for every function $F$ there is $\alpha^*<\lambda$ such that
\noindent {\em if} $\alpha_0<\ldots<\alpha_n<\lambda$, $\alpha^*\leq\alpha_n$
\noindent {\em then} the set
\[\hspace{-1.3cm}\begin{array}{ll}
\{i<\delta:&\mbox{(ii)$^*$ }F(\eta_{\alpha_0},\ldots,\eta_{\alpha_{n-1}},
\eta_{\alpha_n}\rest i)\in I_{\eta_{\alpha_n}\rest i}\\
\ &\ \ \mbox{(and well defined) but }\\
\ &\ \ \eta_{\alpha_n}\rest(i{+}1)\in F(\eta_{\alpha_0},\ldots,
\eta_{\alpha_{n-1}},\eta_{\alpha_n}\rest i)\}\\
\end{array}\]
is in the ideal $J$.
\end{description}
\end{definition}
\begin{remark}
{\em
\begin{enumerate}
\item The main difference between the definition of super $J$-cofinal sequence
and those in \ref{cofinal} is the fact that here the values of the function
$F$ depend on $\eta_{\alpha_\ell}$ (for $\ell<n$), not on the restrictions of
these sequences as it was in earlier notions.
\item ``super$^*$ $J$--cofinal'' is defined by adding
``$\alpha^*\leq\alpha_0$'' (compare \ref{cofinal}(10)).
\end{enumerate}
}
\end{remark}
\begin{proposition}
\label{supimpstr}
Suppose that $(T,\bar{\lambda},\bar{I})\in {\cal K}^{{\rm id}}_{\mu,\delta}$ is such that
for each $\nu\in T_i$, $i<\delta$ the ideal $I_\nu$ is $|T_i|^+$-complete. Let
$J\supseteq J^{{\rm bd}}_\delta$ be an ideal on $\delta$. Then every super
$J$-cofinal sequence is stronger$^*$ $J$-cofinal.
\end{proposition}
\Proof Assume that $\bar{\eta}=\langle\eta_\alpha:\alpha<\lambda\rangle
\subseteq\lim_\delta(T)$ is super $J$-cofinal for $(T,\bar{\lambda},\bar{I})$.
Let $n<\omega$ and let $F_0,\ldots,F_{n-1}$ be functions. For each $\ell\leq
n$ we define an $(\ell+1)$--place function $F^*_\ell$ such that
if $\alpha_0<\alpha_1<\ldots<\alpha_{\ell-1}<\lambda$, $\rho\in
T_i$, $i<\delta$
then
\[\begin{array}{ll}
F^*_\ell(\eta_{\alpha_0},\ldots,\eta_{\alpha_{\ell-1}},\rho)=
&\bigcup\{F_\ell(\eta_{\alpha_0}\rest(i{+}1),\ldots,\eta_{\alpha_{\ell-1}}\rest
(i{+}1), \rho,\nu_{\ell+1},\ldots,\nu_n):\\
\ &\nu_{\ell+1},\ldots,\nu_n\in T_i\ \&\\
\ &F_\ell(\eta_{\alpha_0}\rest(i{+}1),\ldots,\eta_{\alpha_{\ell-1}}\rest
(i{+}1), \rho,\nu_{\ell+1},\ldots,\nu_n)\in I_\rho\\
\ &\mbox{(and well defined)}\}.\\
\end{array}\]
As the ideals $I_\rho$ (for $\rho\in T_i$) are $|T_i|^+$-complete we know that
\[F^*_\ell(\eta_{\alpha_0},\ldots,\eta_{\alpha_{\ell-1}},\rho)\in I_\rho.\]
Applying \ref{super}(b) to the functions $F^*_\ell$ ($\ell<n$) we choose
$\alpha^*_\ell<\lambda$ such that
if $\alpha_0<\ldots<\alpha_\ell<\lambda$, $\alpha^*_\ell\leq\alpha_\ell$
then the set
\[\begin{array}{ll}
B^*_\ell\stackrel{\rm def}{=}\{i<\delta: &F^*_\ell(\eta_{\alpha_0},\ldots,
\eta_{\alpha_{\ell-1}},\eta_{\alpha_\ell}\rest i)\in I_{\eta_{\alpha_\ell}
\rest i}\ \mbox{ but}\\
\ &\eta_{\alpha_\ell}\rest (i+1)\in F^*_\ell(\eta_{\alpha_0},\ldots,
\eta_{\alpha_{\ell-1}},\eta_{\alpha_\ell}\rest i)\}\\
\end{array}\]
is in the ideal $J$.
\noindent Put $\alpha^*=\max\{\alpha^*_\ell:\ell\leq n\}$. We want to show
that this $\alpha^*$ works for the condition \ref{cofinal}(6)(b) (version for
``stronger$^*$''). So suppose that $m\leq n$, $\alpha^*\leq\alpha_0<\alpha_1<
\ldots<\alpha_n<\lambda$. Let
\[\begin{array}{ll}
B_m\stackrel{\rm def}{=}&\{i<\delta: F_m(\eta_{\alpha_0}\rest(i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i,\ldots,\eta_{\alpha_n}
\rest i)\in I_{\eta_{\alpha_m}\rest i}\ \&\\
\ &\eta_{\alpha_m}\rest (i{+}1)\in F_m(\eta_{\alpha_0}\rest(i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i,\ldots,\eta_{\alpha_n}
\rest i)\}.\\
\end{array}\]
Note that if $i\in B_m$ then, as $\alpha^*_m\leq\alpha^*\leq\alpha_m$,
\[\begin{array}{rl}
\eta_{\alpha_m}\rest (i{+}1)\in &F_m(\eta_{\alpha_0}\rest(i{+}1),\ldots,
\eta_{\alpha_{m-1}}\rest (i{+}1),\eta_{\alpha_m}\rest i,\ldots,\eta_{\alpha_n}
\rest i)\subseteq\\
\subseteq &F^*_m(\eta_{\alpha_0},\ldots,\eta_{\alpha_{m-1}},\eta_{\alpha_m}
\rest i)\in I_{\eta_{\alpha_m}\rest i}.
\end{array}\]
Hence we conclude that $B_m\subseteq B^*_m$ and therefore $B_m\in J$,
what finishes the proof of the proposition. \QED
\begin{proposition}
\label{supimpfub}
Assume that $(T,\bar{\lambda},\bar{I})\in{\cal K}_{\mu,\delta}^{{\rm id}}$, each ideal
$I_\eta$ (for $\eta\in T_i$, $i<\delta$) is $(|\delta|+|T_i|)^+$-complete and
$J\supseteq J^{{\rm bd}}_\delta$ is an ideal on $\delta$. Further suppose that a
sequence $\bar{\eta}=\langle\eta_\alpha:\alpha<\lambda\rangle$ is super
$J$-cofinal for $(T,\bar{\lambda},\bar{I})$, $\lambda$ is a regular cardinal
greater than $|T|$ and a sequence $\langle\alpha_{\varepsilon,\ell}:
\varepsilon<\lambda, \ell<n\rangle\subseteq \lambda$ is with no repitition and
such that
\[\alpha_{\varepsilon,0}<\alpha_{\varepsilon,1}<\ldots<\alpha_{\varepsilon,
n-1}\quad\quad\mbox{ for all }\varepsilon<\lambda.\]
{\em Then} for every $\varepsilon<\lambda$ large enough there is $a\in J$ such
that
\begin{description}
\item[$(\boxdot)$] if $i_\ell\in\delta\setminus a$ (for $\ell<n$), $i_0\geq
i_1\geq\ldots\geq i_{n-1}$ then
\[\hspace{-1cm}\begin{array}{l}
(\exists^{I_{\eta_{\alpha_{\varepsilon,0}}\rest i_0}}\gamma_0)\ldots
(\exists^{I_{\eta_{\alpha_{\varepsilon,n-1}}\rest i_{n-1}}}\gamma_{n-1})\\
(\exists^\lambda \zeta<\lambda)(\forall\ell<n)(\eta_{\alpha_{\zeta,\ell}}\rest
(i_\ell+1) = \eta_{\alpha_{\varepsilon,\ell}}\rest i_\ell\hat{\
}\langle\gamma_\ell\rangle).\\
\end{array}\]
\end{description}
\end{proposition}
\Proof This is very similar to claim \ref{claimgreen}. First choose
$\varepsilon_0<\lambda$ such that for each $\varepsilon\in
[\varepsilon_0,\lambda)$ and for every $i_0,\ldots,i_{n-1}<\delta$ we have
\[(\exists^\lambda \zeta<\lambda)(\forall\ell<n)(\eta_{\alpha_{\zeta,\ell}}
\rest(i_\ell+1)=\eta_{\alpha_{\varepsilon,\ell}}\rest (i_\ell+1))\]
(possible as $|T|<\cf(\lambda)=\lambda$).
Now, for $\bar{\imath}=\langle i_\ell: \ell<n\rangle\subseteq\delta$ and
$\bar{\nu}=\langle\nu_\ell: \ell<n\rangle$ such that $i_0\geq i_1\geq\ldots
\geq i_{n-1}$, $\nu_\ell\in T_{i_\ell}$ and $k<n$ we define a function
$f^k_{\bar{\imath},\bar{\nu}}: \prod\limits_{\ell<k} \lambda_{\nu_\ell}
\longrightarrow I_{\nu_k}$ (with a convention that
$f^0_{\bar{\imath},\bar{\nu}}$ is supposed to be a $0$-place function, i.e. a
constant) as follows
Let
\[B_{\bar{\imath},\bar{\nu}}\stackrel{\rm def}{=}\{\langle\gamma_\ell:
\ell<n\rangle\in\prod_{\ell<n}\lambda_{\nu_\ell}: (\exists^\lambda \zeta<
\lambda)(\forall \ell<n)(\eta_{\alpha_{\zeta,\ell}}\rest (i_\ell +1)=\nu_\ell
\hat{\ }\langle\gamma_\ell\rangle)\}.\]
If
\begin{description}
\item[$(\blacklozenge_{\bar{\imath},\bar{\nu}})$] $\neg(\exists^{I_{\nu_0}}
\gamma_0)\ldots(\exists^{I_{\nu_{n-1}}}\gamma_{n-1})(\langle\gamma_0,\ldots,
\gamma_{n-1}\rangle\in B_{\bar{\imath},\bar{\nu}})$
\end{description}
then $f^0_{\bar{\imath},\bar{\nu}},\ldots,f^{n-1}_{\bar{\imath},\bar{\nu}}$
are such that
\begin{description}
\item[$(\lozenge)$] if $\langle\gamma_0,\ldots,\gamma_{n-1}\rangle\in
B_{\bar{\imath},\bar{\nu}}$
then $(\exists k<n)(\gamma_k\in f^k_{\bar{\imath},\bar{\nu}}(\gamma_0,\ldots,
\gamma_{k-1}))$.
\end{description}
Otherwise (i.e. if not $(\blacklozenge_{\bar{\imath},\bar{\nu}})$ the
functions $f^k_{\bar{\imath},\bar{\nu}}$ are constantly equal to $\emptyset$
(for $k<n$).
Next, for $k<n$, choose functions $F_k$ such that if $\eta_0,\ldots,\eta_k\in
\lim_\delta(T)$, $i<\delta$ then
\[\begin{array}{ll}
F_k(\eta_0,\ldots,\eta_{k-1},\eta_k\rest i)= &\ \\
\bigcup\{f^k_{\bar{\imath},\bar{\nu}}(\eta_0(i_0),\ldots,\eta_{k-1}(i_{k-1})):
&\bar{\imath}=\langle i_\ell:\ell<n\rangle,\ \bar{\nu}=\langle\nu_\ell:
\ell<n\rangle,\\
\ &\delta>i_0\geq\ldots\geq i_k=i\geq i_{k+1}\geq\ldots\geq i_{n-1}\\
\ &\nu_\ell=\eta_\ell\rest i_\ell\ \mbox{ for }\ell\leq k\ \mbox{ and}\\
\ &\nu_\ell\in T_{i_\ell}\ \mbox{ for }k<\ell<n\}.
\end{array}\]
Note that $F_k(\eta_0,\ldots,\eta_{k-1},\eta_k\rest i)$ is a union of at most
$|\delta|+|T_i|$ sets from the ideal $I_{\eta_k\rest i}$ and hence
$F_k(\eta_0,\ldots,\eta_{k-1},\eta_k\rest i)\in I_{\eta_k\rest i}$ (for each
$\eta_0,\ldots,\eta_k\in \lim_\delta(T)$, $i<\delta$). Thus, using the super
$J$-cofinality of $\bar{\eta}$ we find $\alpha^*<\lambda$ such that
\begin{quotation}
\noindent if $\alpha^*\leq\alpha_<\ldots<\alpha_n<\lambda$
\noindent then the set
\[\{i<\delta: (\exists k<n)(\eta_{\alpha_k}(i)\in F_k(\eta_{\alpha_0},\ldots,
\eta_{\alpha_{k-1}},\eta_{\alpha_k}))\}\]
is in the ideal $J$.
\end{quotation}
Let $\varepsilon_1>\varepsilon_0$ be such that for every $\varepsilon\in
[\varepsilon_1,\lambda)$ we have $\alpha^*<\alpha_{\varepsilon,0}<\ldots<
\alpha_{\varepsilon,n-1}$.
Suppose now that $\varepsilon_1<\varepsilon<\lambda$. By the choice of
$\alpha^*$ we know that the set
\[a\stackrel{\rm def}{=}\{i<\delta: (\exists\ell<n)(\eta_{\alpha_{\varepsilon,
\ell}}(i)\in F_\ell(\eta_{\alpha_{\varepsilon,0}},\ldots\eta_{
\alpha_{\varepsilon,\ell-1}},\eta_{\alpha_{\varepsilon,\ell}}\rest i))\}\]
is in the ideal $J$. We are going to show that the assertion $(\boxdot)$ holds
for $\varepsilon$ and $a$.
\smallskip
\noindent Suppose that $\bar{\imath}=\langle
i_\ell:\ell<n\rangle\subseteq\delta\setminus a$, $i_0\geq i_1\geq\ldots\geq
i_{n-1}$. Let $\bar{\nu}=\langle \nu_\ell: \ell<n\rangle$,
$\nu_\ell=\eta_{\alpha_{\varepsilon,\ell}}\rest i_\ell$. If the condition
$(\blacklozenge_{\bar{\imath},\bar{\nu}})$ fails then we are done. So assume
that it holds true. By the choice of the set $a$ (and $\alpha^*$) we have
\[(\forall\ell<n)(\eta_{\alpha_{\varepsilon,\ell}}(i_\ell)\notin F_\ell
(\eta_{\alpha_{\varepsilon,0}},\ldots,\eta_{\alpha_{\varepsilon,\ell-1}},
\eta_{\alpha_{\varepsilon,\ell}}\rest i_l))\]
what, by the definition of $F_\ell$, implies that
\[(\forall\ell<n)(\eta_{\alpha_{\varepsilon,\ell}}(i_\ell)\notin
f^\ell_{\bar{\imath},\bar{\nu}}(\eta_{\alpha_{\varepsilon,0}}(i_0),\ldots,
\eta_{\alpha_{\varepsilon,\ell-1}}(i_{\ell-1}))).\]
By $(\lozenge)$ we conclude that
\[\langle\eta_{\alpha_{\varepsilon,0}}(i_0),\ldots,\eta_{\alpha_{\varepsilon,
n-1}}(i_{n-1})\rangle\notin B_{\bar{\imath},\bar{\nu}}\]
and hence, by the definition of $B_{\bar{\imath},\bar{\nu}}$,
\[\neg(\exists^\lambda \zeta)(\forall\ell<n)(\eta_{\alpha_{\zeta,\ell}}\rest
(i_\ell+1) =\eta_{\alpha_{\varepsilon,\ell}}\rest (i_\ell))\]
which contradicts the choice of $\varepsilon_0$ (remember $\varepsilon\geq
\varepsilon_1>\varepsilon_0$). \QED
\begin{definition}
We say that a $\lambda$-marked Boolean algebra $({\Bbb B},\bar{y})$ {\em has
character $n$} if
\begin{quotation}
\noindent for every finite set $u\in [\lambda]^{{<}\omega}$ such that
${\Bbb B}\models\bigcap\limits_{\alpha\in u} y_\alpha={\bf 0}$ there exist a subset
$v\subseteq u$ of size $|v|\leq n$ such that
${\Bbb B}\models\bigcap\limits_{\alpha\in v} y_\alpha={\bf 0}$.
\end{quotation}
\end{definition}
\begin{proposition}
If a $\lambda$-marked Boolean algebra $({\Bbb B},\bar{y})$ is $(\theta,{\rm
not}\lambda)$-Knaster (or other examples considered in the present paper) and
$({\Bbb B},\bar{y})$ has character 2 then without loss of generality
$({\Bbb B},\bar{y})$ is determined by a colouring on $\lambda$:
\begin{quotation}
\noindent if $c:[\lambda]^2\longrightarrow 2$ is such that
\[c(\{\alpha,\beta\})=0\quad\mbox{ \rm iff }\quad{\Bbb B}\models y_\alpha\cap
y_\beta={\bf 0}\]
\noindent then the algebra ${\Bbb B}$ is freely generated by $\{y_\alpha:
\alpha<\lambda\}$ except that
\begin{quotation}
{\em if} $c(\{\alpha,\beta\})=0$
{\em then} $y_\alpha\cap y_\beta=0$. \QED
\end{quotation}
\end{quotation}
\end{proposition}
\begin{remark}
{\em
These are nice examples.
}
\end{remark}
\begin{proposition}
In all our results (like: \ref{algide} or \ref{consknaster}), the marked
Boolean algebra $({\Bbb B},\bar{y})$ which we get is actually of character 2 as
long as any $({\Bbb B}_\eta,\bar{y}_\eta)$ appearing in the assumptions (if any) is
like that.
\noindent Then automatically the $\theta$--Knaster property of the marked
Boolean algebra $({\Bbb B},\bar{y})$ implies a stronger condition:
\noindent if $Z\in [{\rm lg}\/(\bar{y})]^\theta$ then there is a set $Y\in
[Z]^\theta$ such that $\{y_i: i\in Y\}$ generates a filter in ${\Bbb B}$. \QED
\end{proposition}
\begin{proposition}
\label{ultfub}
Let $(T,\bar{\lambda},\bar{I})\in{\cal K}_{\mu,\delta}^{{\rm id}}$ be such that for each
$\eta\in T$ the filter $(I_\eta)^c$ (dual to $I_\eta$) is an ultrafilter on
$\suc_T(\eta)$, and let $J$ be an ideal on $\delta$ (extending
$J^{{\rm bd}}_\delta$). {\em If:}
\begin{description}
\item[(a)] ${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta):
\eta\in T\rangle)$ is a $(\delta,\mu,\lambda)$-constructor, the sequence
$\bar{\eta}$ is stronger $J$-cofinal for $(T,\bar{\lambda},\bar{I})$,
$|T|<\cf(\lambda)=\lambda$,
\item[(b)] the sequence $\langle\alpha_{\varepsilon,\ell}:\varepsilon<\lambda,
\ell<n\rangle\subseteq\lambda$ is with no repetition,
\item[(c)] for each distinct $\eta,\nu\in T$ either the ideal $I_\eta$ is
$(2^{\lambda_\nu})^+$--complete (which, of course, implies $\lambda_\eta>
2^{\lambda_\nu}$) or the ideal $I_\nu$ is $(2^{\lambda_\eta})^+$--complete (it
is enough if this holds true for $\eta,\nu$ such that ${\rm lg}\/(\eta)={\rm lg}\/(\nu)$
\end{description}
{\em then} for every large enough $\varepsilon<\lambda$ for $J$-almost all
$i<\delta$ there are sets $X_\ell\in (I_{\eta_{\alpha_{\varepsilon,\ell}}\rest
i})^+$ (for $\ell<n$) such that
\[(\forall\gamma_0\in X_0)\ldots(\forall\gamma_{n-1}\in X_{n-1})
(\exists^\lambda\zeta<\lambda)(\forall\ell<n)
(\eta_{\alpha_{\varepsilon,\ell}}\rest i\hat{\ }\langle\gamma_\ell\rangle
\vartriangleleft\eta_{\alpha_{\zeta,\ell}}).\]
\end{proposition}
\noindent{\bf Remark~\ref{ultfub}.A}\ \ \ We can replace {\em stronger} by
{\em big} and then omit being an ultrafilter.
\medskip
\Proof First note that we may slightly re-enumerate are sequence $\langle
\alpha_{\varepsilon,\ell}: \varepsilon<\lambda,\ell<n\rangle$ and we may
assume that for each $\varepsilon<\lambda$
\[\alpha_{\varepsilon,0}<\alpha_{\varepsilon,1}<\ldots<\alpha_{\varepsilon,
n-1}.\]
Now, since $|T|<\cf(\lambda)=\lambda$ we may apply claim \ref{claimgreen} to
\[\langle\langle\alpha_{\varepsilon,\ell}:\ell<n\rangle:\varepsilon_0\leq
\varepsilon<\lambda\rangle\]
(we need to take $\varepsilon_0$ large enough to get the condition $\hat{(*)}$
of the proof of \ref{green}). Consequently we may conclude that there is
$\varepsilon_1<\lambda$ such that for every $\varepsilon\in
[\varepsilon_1,\lambda)$
\begin{description}
\item[$(\boxtimes_\varepsilon)$] for $J$-majority of $i<\delta$ we have
\[\hspace{-0.7cm} (\exists^{I_{\eta_{\alpha_{\varepsilon,0}}\rest i}}\gamma_0)
\ldots (\exists^{I_{\eta_{\alpha_{\varepsilon,n-1}}\rest i}}\gamma_{n-1})
(\exists^\lambda\zeta{<}\lambda)(\forall\ell{<}n)(\eta_{\alpha_{\zeta,\ell}}
\rest (i{+}1) =\eta_{\alpha_{\varepsilon,\ell}}\rest i\hat{\
}\langle\gamma_\ell \rangle).\]
\end{description}
Now we would like to apply \ref{prefub}. We can not do this directly as we do
not know if the cardinals $\lambda_{\eta_{\varepsilon,\ell}\rest i}$ are
decresing (with $\ell$). However the following claim helps us.
\begin{claim}
\label{transp}
Suppose that $\lambda_0<\lambda_1$ are cardinals and $I_0,I_1$ are maximal
ideals on $\lambda_0,\lambda_1$ respectively. Assume that the ideal $I_1$ is
$(\lambda_0)^+$--complete and $\varphi(x,y)$ is a formula. Then
\[(\exists^{I_0}\gamma_0)(\exists^{I_1}\gamma_1)\varphi(\gamma_0,\gamma_1)\ \
\ \Rightarrow\ \ \ (\exists^{I_1}\gamma_1)(\exists^{I_0}\gamma_0)
\varphi(\gamma_0,\gamma_1).\]
\end{claim}
\noindent{\em Why?}\ \ \ First note that if $I$ is a maximal ideal then the
quantifiers $\exists^I$ and $\forall^I$ are equivalent. Suppose now that
\[(\exists^{I_0}\gamma_0)(\exists^{I_1}\gamma_1)\varphi(\gamma_0,\gamma_1).\]
This implies (as $I_0,I_1$ are maximal) that
\[(\forall^{I_0}\gamma_0)(\forall^{I_1}\gamma_1)\varphi(\gamma_0,\gamma_1).\]
Thus we have a set $a\in I_0$ and for each $\gamma\in\lambda_0\setminus a$ we
have a set $b_\gamma\in I_1$ such that
\[(\forall\gamma_0\in\lambda_0\setminus a)(\forall\gamma_1\in\lambda_1
\setminus b_{\gamma_0})\varphi(\gamma_0,\gamma_1).\]
Let $b=\bigcup\limits_{\gamma\in\lambda_0\setminus a} b_\gamma$. As $I_1$ is
$(\lambda_0)^+$--complete the set $b$ is in $I_1$. Clearly
\[(\forall\gamma_1\in\lambda_1\setminus b)(\forall\gamma_0\in\lambda\setminus
a)\varphi(\gamma_0,\gamma_1)\]
which implies $(\exists^{I_1}\gamma_1)(\exists^{I_0}\gamma_0)\varphi
(\gamma_0,\gamma_1)$, finishing the proof of the claim.
\medskip
Now fix $\varepsilon>\varepsilon_1$ ($\varepsilon_1$ as chosen earlier). Take
$i^*<\delta$ such that the elements of $\langle\eta_{\alpha_{\varepsilon,
\ell}}\rest i:\ell<n\rangle$ are pairwise distinct. Suppose that $i\in
[i^*,\delta)$ is such that the formula of $(\boxtimes_\varepsilon)$ holds
true. Let $\{k_\ell: \ell<n\}$ be an enumeration of $n$ such that
\[\lambda_{\eta_{\alpha_{\varepsilon,k_0}}\rest i}>
\lambda_{\eta_{\alpha_{\varepsilon,k_1}}\rest i}>\ldots>
\lambda_{\eta_{\alpha_{\varepsilon,k_{n-1}}}\rest i}.\]
(Note that by the assumption {\bf (c)} we know that all the
$\lambda_{\eta_{\alpha_{\varepsilon,k_\ell}}\rest i}$ are distinct, remember
the choice of $i^*$.) Applying claim \ref{transp} we conclude that
\[(\exists^{I_{\eta_{\alpha_{\varepsilon,k_0}}\rest i}}\gamma_{k_0})\ldots
(\exists^{I_{\eta_{\alpha_{\varepsilon,k_{n-1}}}\rest i}}\gamma_{k_{n-1}})
(\exists^\lambda\zeta{<}\lambda)(\forall\ell{<}n)(\eta_{\alpha_{\varepsilon,
\ell}} \rest i\hat{\ }\langle\gamma_\ell\rangle=\eta_{\alpha_{\zeta,\ell}}\rest
(i{+}1)).\]
But now we are able to use \ref{prefub} and we get that there are sets
$X_{k_\ell}\subseteq\lambda_{\eta_{\alpha_{\varepsilon,k_\ell}}\rest i}$,
$X_{k_\ell}\notin I_{\eta_{\alpha_{\varepsilon,k_\ell}}\rest i}$ (for
$\ell<n$) such that
\[\prod_{\ell<n}X_\ell\subseteq\{\langle\gamma_0,\ldots,\gamma_{n-1}
\rangle: (\exists^\lambda\zeta{<}\lambda)(\forall\ell{<}n)
(\eta_{\alpha_{\varepsilon,\ell}}\rest i\hat{\ }\langle\gamma_\ell\rangle=
\eta_{\alpha_{\zeta,\ell}}\rest (i{+}1))\}\]
what is exactly what we need. \QED
\medskip
If we assume less completeness of the ideals $I_\eta$ in \ref{ultfub} then
still we may say something.
\begin{proposition}
Let $\langle\sigma_i: i<\delta\rangle$ be a sequence of cardinals. Suppose
that $T,\bar{\lambda},\bar{I},\bar{\eta},J,\lambda,\mu,\delta$ and
$\langle\alpha_{\varepsilon,\ell}:\varepsilon<\lambda,\ell<n\rangle$ are as in
\ref{ultfub} but with condition {\bf (c)} replaced by
\begin{description}
\item[(c)$^-_{\langle\sigma_i: i<\delta\rangle}$] if $\eta,\nu\in T_i$,
$\eta\neq\nu$, $i<\delta$ then either the ideal $I_\eta$ is
$((\lambda_\nu)^{\sigma_i})^+$--complete or the ideal $I_\nu$ is
$((\lambda_\eta)^{\sigma_i})^+$--complete.
\end{description}
Then for every large enough $\varepsilon<\lambda$ for $J$-almost all
$i<\delta$ there are sets $X_\ell\in
[\lambda_{\eta_{\alpha_{\varepsilon,\ell}} \rest i}]^{\sigma_i}$ (for
$\ell<n$) such that
\[(\forall\gamma_0\in X_0)\ldots(\forall\gamma_{n-1}X_{n-1})(\exists^\lambda
\zeta<\lambda)(\forall\ell<n)(\eta_{\alpha_{\varepsilon,\ell}}\rest i\hat{\ }
\langle\gamma_\ell\rangle\vartriangleleft\eta_{\alpha_{\zeta,\ell}}).\]
\end{proposition}
\Proof The proof goes exactly as the one of \ref{ultfub}, but instead of
\ref{prefub} we use \ref{weakfub}. \QED
\begin{remark}
{\em
\begin{enumerate}
\item Note that in the situation as in \ref{ultfub}, we usually have that
``$J$--cofinal'' implies ``stronger $J$--cofinal'' (see \ref{cofimpstrong},
\ref{strimpstr}).
\item The first assumption of \ref{ultfub} (ultrafilters) coupled with our
normal completeness demands is a very heavy condition, but it has rewards.
\item A natural context here is when $\langle\mu_i: i\leq\kappa\rangle$ is a
strictly increasing continues sequence of cardinals such that each $\mu_{i+1}$
is compact and $\mu=\mu_\kappa$. Then every $\mu_{i+1}$-complete filter can be
extended to an $\mu_{i+1}$-complete ultrafilter. Moreover $2^\mu=\mu^+$
follows by Solovay \cite{So74}.
{\em If} for some function $f$ from cardinals to cardinals, for each $\chi$
there is an algebra ${\Bbb B}_\chi$ of cardinality $f(\chi)$ which cannot be
decomposed into $\leq\mu$ sets $X_i$ each with some property ${\bf
Pr}({\Bbb B}_\chi, X_i)$ and if each $\mu_i$ if $f$-inaccessible
\noindent {\em then} we can find $T,\bar{I},\bar{\lambda}$ as in \ref{ultfub}
and such that $\eta\in T_i\ \ \Rightarrow\ \ \mu_i<\chi_\eta<\lambda_\eta<
\mu_{i+1}$ and for $\eta\in T_i$ there is an algebra ${\Bbb B}_\eta$ with universe
$\lambda_\eta$ and the ideal $I_\eta$ is $\chi_\eta$--complete,
\[\mbox{if } X\subseteq {\Bbb B}_\eta\mbox{ and } {\bf Pr}({\Bbb B}_\eta,X)\mbox{ then }
X\in I_\eta\]
(compare \ref{algide}) and $\lambda_\eta<\lambda_\nu\ \ \Rightarrow\ \
(2^{\lambda_\eta})^+<\chi_\nu$. Now choosing cofinal $\bar{\eta}$ we may
proceed as in earlier arguments.
\item It seems to be good for building nice examples, however we did not find
the right question yet.
\item Central to our proofs is an assumption that
\[\mbox{``}\langle\alpha_{\zeta,\ell}: \zeta<\lambda,\ell<n\rangle\subseteq
\lambda\mbox{ is a sequence with no repetition'',}\]
i.e. we deal with $\lambda$ disjoint $n$-tuples. This is natural as the
examples constructed here are generated from $\{x_i: i<\lambda\}$ by finitary
functions. One may ask what happens if we admit functions with, say,
$\aleph_0$ places? We can still try to get for $\mu$ as above that:
\begin{description}
\item[$(\boxtimes)$] there is $h:[\mu^+]^2\longrightarrow 2$ such that
{\em if} $\langle u_\varepsilon:\varepsilon<\lambda\rangle$ are pairwise
disjoint, $u_\varepsilon=\{\alpha_{\varepsilon,\ell}:\ell<\ell^*\}$ is the
increasing (with $\ell$) enumeration, $\ell^*<\mu$ ($\ell^*$ infinite), for a
sequence $\langle\nu_\ell:\ell<\ell^*\rangle\subseteq T_i$
\[\hspace{-1cm}\begin{array}{l}
B_{\langle\nu_\ell:\ell<\ell^*\rangle}\stackrel{\rm def}{=}\\
\{\langle\eta_{\alpha_{\varepsilon,\ell}}(i): \ell<\ell^*\rangle:
(\exists^\lambda\zeta<\lambda)(\forall\ell<\ell^*)(\eta_{\alpha_{\varepsilon,
\ell}}\rest(i+1)=\eta_{\alpha_{\zeta,\ell}}\rest (i+1))\},\\
\end{array}\]
for some $i^*<\delta$ there are no repetitions in $\langle
\eta_{\alpha_{\varepsilon,\ell}}\rest i^*: \ell<\ell^*\rangle$ and $h\rest
[u_\varepsilon]^2\equiv 1$ (for each $\varepsilon<\lambda$)
{\em then} there are $\alpha<\beta$ (really a large set of these) such that
\[h\rest [u_\alpha\cup u_\beta]^2\equiv 1.\]
\end{description}
The point is that we can deal with functions with infinitely many variables.
Looking at previous proofs, ``in stronger'' we can get (for $\mu$ strong
limit singular etc):
\begin{quotation}
\noindent for $\alpha$ large enough
\noindent for $i<\delta=\cf(\mu)$ large enough
\noindent $\ldots\ldots\ldots$
\noindent we can defeat
\[\hspace{-0.5cm}(\ldots\ldots(\forall^{I_{\eta_{\alpha_{\varepsilon,\ell}}
\rest i}}\gamma_\ell)\ldots\ldots)(\langle\gamma_\ell: \ell<\ell^*\rangle\in
B_{\langle\eta_{\alpha_{\varepsilon,\ell}}\rest i: \ell<\ell^*\rangle})\]
but the duality of quantifiers fails, so the conclusion is only that
\[\hspace{-1cm}(\forall^J i<\delta)[\neg(\ldots
(\forall^{I_{\eta_{\alpha_{\varepsilon,\ell}}\rest i}}
\gamma_\ell)\ldots)_{\ell<\ell^*}(\langle \eta_{\alpha_{\varepsilon,\ell}}(i):
\ell<\ell^*\rangle\notin B_{\langle\eta_{\alpha_{\varepsilon,\ell}}\rest i:
\ell<\ell^*\rangle})].\]
\end{quotation}
\item (no ultrafilters) If $I\supseteq J^{{\rm bd}}_\eta$, $\delta$ is a regular
cardinal, $\lambda_\eta=\lambda_{{\rm lg}\/(\eta)}$ and for each $u\in [T_i]^{{<}
|\delta|\chi}$, $i<\delta$ the free product $\mathop{\bigstar}\limits_{\eta\in
u}{\Bbb B}_\eta$ satisfies the $\lambda$-cc then we can show that the algebra
${\Bbb B}^{{\rm red}}_{<\chi}$ satisfies the $\lambda$-cc too,
where for a cardinal $\chi$ the algebra ${\Bbb B}^{{\rm red}}_{<\chi}$ is the Boolean
algebra freely generated by
\[\hspace{-0.5cm}\begin{array}{ll}
\{\bigcap\limits_{\alpha\in u} x_\alpha^{{\frak t}(\alpha)}:& {\frak t}:u\longrightarrow
2, u\in [\lambda]^{<\delta}, h\rest [u\cap{\frak t}^{-1}[1]]^2\equiv 1\mbox{ and}\\
\ &|u|<\chi\mbox{ and}\\
\ &(\exists i<\delta)(\mbox{the mapping }\alpha\mapsto\eta_\alpha(i)\mbox{ is
one-to-one (for $\alpha\in u$)})\\
\ &(\exists i<\delta)(\exists\alpha\in u)(\forall j\in (i,\delta))(\forall
\beta\in u)(f_\alpha(j)\leq f_\beta(j))\}.\\
\end{array}\]
[Note that if $\chi\leq\cf(\delta)$ simpler.]
\end{enumerate}
}
\end{remark}
\[*\quad*\quad*\quad*\quad*\]
Now we will deal with an additional demand that the algebra ${\Bbb B}^{{\rm red}}$
satisfies $|\delta|^+$-cc (or even has the $|\delta|^+$--Knaster property).
Note that the demand of $|\delta|$-cc does not seem to be reasonable: if
every $\bar{y}_\eta$ has two disjoint members (and every node $t\in T$ is an
initial segment of a branch through $T$) then we can find $\delta$
branches which, if in $\{\eta_\alpha: \alpha<\lambda\}$, give $\delta$
pairwise disjoint elements. {\em Moreover}:
\begin{quotation}
\noindent for each $\nu\in T_\ell$ let $A_\nu=\{\eta_\alpha(i):
\eta_\alpha\rest i=\nu\}$ and
\[a_\alpha=\{i<\delta: (\exists \beta\in A_{\eta_\alpha\rest i})
({\Bbb B}_{\eta_\alpha\rest i}\models y_{\eta_\alpha(i)}\cap y_\beta={\bf 0})\}.\]
So if ${\Bbb B}^{{\rm red}}\models\sigma$-cc then
$(\forall\alpha<\lambda)(|a_\alpha|<\sigma)$.
\end{quotation}
\begin{definition}
\label{free}
Let $(T,\bar{\lambda})\in {\cal K}_{\mu,\delta}$ and let $\bar{\eta}=\langle
\eta_\alpha:\alpha<\lambda\rangle\subseteq{\lim}_\delta(T)$. We say that
$\bar{\eta}$ is {\em hereditary $\theta$-free} if for every $Y\in
[\lambda]^\theta$ there are $Z\in [Y]^\theta$ and $i<\delta$ such that
\[(\forall \alpha,\beta\in Z)(\alpha\neq\beta\ \ \ \Rightarrow\ \ \
[\eta_\alpha\rest i=\eta_\beta\rest i\ \&\ \eta_\alpha(i)\neq\eta_\beta(i)]).\]
\end{definition}
\begin{proposition}
\label{morekna}
Assume that ${\cal C}=(T,\bar{\lambda},\bar{\eta},\langle ({\Bbb B}_\eta,\bar{y}_\eta):
\eta\in T\rangle)$ is a $(\delta,\mu,\lambda)$-constructor.
If $\bar{\eta}$ is hereditary $\theta$-free, each algebra ${\Bbb B}_\eta$ has
the $\theta$-Knaster property and $\theta$ is regular then the algebra
${\Bbb B}^{{\rm red}}({\cal C})$ has the $\theta$-Knaster property.
\end{proposition}
\Proof The same as for \ref{consknaster}. Note that the proof there
shows actually that,
\begin{quotation}
\noindent if $(\forall\alpha<\theta)(|\alpha|^{|\delta|}<\theta=\cf(\theta))$
\noindent then $\bar{\eta}$ is $\theta$-hereditary free. \QED
\end{quotation}
\begin{proposition}
\label{getfree}
Assume that $(T,\bar{\lambda})\in {\cal K}_{\mu,\delta}$, $\bar{\eta}=\langle
\eta_\alpha:\alpha<\lambda\rangle\subseteq{\lim}_\delta(T)$, $\lambda$ is a
regular cardinal. Further suppose that
\begin{description}
\item[(a)] $(\forall\alpha<\theta)(|\alpha|^{{<}\delta}<\theta =\cf(\theta))$,
$\delta<\theta$, $J$ is an ideal on $\delta$ extending $J^{{\rm bd}}_\delta$ and
\item[(b)] the sequence $\bar{\eta}$ is $<_J$-increasing and one of the
following conditions is satisfied:
\begin{description}
\item[$(\alpha)$] $\bar{\eta}$ is $<_J$-cofinal in $\prod\limits_{i<\delta}
\lambda_i/J$, $\lambda_i$ are regular cardinals above $\theta$ (at least for
$J$-majority of $i<\delta$), $\{\alpha<\lambda: \cf(\alpha)=\theta\}\in
I[\lambda]$ and $\lambda_\eta=\lambda_{{\rm lg}\/(\eta)}$;
\item[$(\beta)$] there are a sequence $\langle C_\alpha:\alpha<\lambda\rangle$
of subsets of $\lambda$, a closed unbounded subset $E$ of $\lambda$ and
$i^*<\delta$ such that
\begin{enumerate}
\item $C_\alpha\subseteq\alpha$, ${\rm otp}(C_\alpha)\leq\theta$,
\item if $\beta\in C_\alpha$ then $C_\beta=C_\alpha\cap\beta$ and $\eta_\beta
\rest [i^*,\delta)<\eta_\alpha\rest [i^*,\delta)$,
\item if $\alpha\in E$ and $\cf(\alpha)=\theta$ then $\alpha=\sup(C_\alpha)$.
\end{enumerate}
\end{description}
\end{description}
Then there is $A\in [\lambda]^\lambda$ such that the restriction $\bar{\eta}
\rest A$ is $\theta$-hereditary free.
\end{proposition}
\Proof First let us assume that the case $(\beta)$ of the clause {\bf (b)} of
the assumptions holds.
\begin{claim}
\label{claimfree}
Suppose that $Y\in [E]^\theta$. Then
\begin{enumerate}
\item $(\exists Z\in [Y]^\theta)(\exists i^\otimes)(\mbox{the sequence
}\langle f_{\beta_\varepsilon}(i^\otimes): \varepsilon\in Z\rangle\mbox{ is
strictly increasing})$.
\item If additionally $J=J^{{\rm bd}}_\delta$ then
\[\hspace{-0.8cm}(\exists Z\in [Y]^\theta)(\exists i^\otimes<\delta)(\mbox{the
sequence }\langle\eta_\beta\rest [i^\otimes,\delta):\beta\in Z\rangle\mbox{ is
strictly increasing}).\]
\end{enumerate}
\end{claim}
\noindent{\em Why?}\ \ \ Suppose $Y\in [E]^\theta$. Without loss of generality
we may assume that ${\rm otp}(Y)=\theta$. Let $\alpha=\sup(Y)$. So $\alpha\in E$,
$\cf(\alpha)=\theta$ and hence $C_\alpha$ is unbounded in $\alpha$. Let
$C_\alpha=\langle\alpha_\varepsilon:\varepsilon<\theta\rangle$ be the
increasing enumeration. Clearly the set
\[A\stackrel{\rm def}{=}\{\varepsilon<\theta:[\alpha_\varepsilon,
\alpha_{\varepsilon+1})\cap Y\neq\emptyset\}\]
is unbounded in $\theta$. For $\varepsilon\in A$ choose $\beta_\varepsilon\in
[\alpha_\varepsilon, \alpha_{\varepsilon+1})\cap Y$. Then
\[(\exists a_\varepsilon\in J)(\eta_{\alpha_\varepsilon}\rest (\delta\setminus
a_\varepsilon)\leq \eta_{\beta_\varepsilon}\rest (\delta\setminus
a_\varepsilon)<\eta_{\alpha_{\varepsilon+1}}\rest (\delta\setminus
a_\varepsilon)).\]
Now choose $i_\varepsilon\in\delta\setminus a_\varepsilon$, $i_\varepsilon>
i^*$ and find $B\in [A]^\theta$ such that
\[\varepsilon\in B\ \ \ \Rightarrow\ \ \ i_\varepsilon=i^\otimes.\]
Easily, by the assumption $(\beta)(2)$, this $i^\otimes$ and
$Z=\{\beta_\varepsilon: \varepsilon\in B\}$ are as required in
\ref{claimfree}(1).
\noindent If additionally we know that $J=J^{{\rm bd}}_\delta$ then for some $B\in
[A]^\theta$ we have
\[(\exists i^\otimes\in [i^*,\delta))(\varepsilon\in B\ \ \Rightarrow\ \
a_\varepsilon\subseteq i^\otimes)\]
and hence the sequence $\langle f_{\beta_\varepsilon}\rest [i^\otimes,\delta):
\varepsilon\in B\rangle$ is as required in \ref{claimfree}(2) (remember
$(\beta)(2)$).
\medskip
But now, using $i^\otimes$ given by \ref{claimfree} we may deal with the
sequence $\langle f_{\beta_\varepsilon}\rest (i^\otimes +1): \varepsilon\in
B\rangle$ and using the old proof (see \ref{consknaster}) on the tree
$\bigcup\limits_{i\leq i^\otimes} T_i$ (note that we may apply the assumption
{\bf (a)} to arguments like there) we may get the desired conclusion. This
finishes the case when $(\beta)$ of {\bf (b)} holds true.
\bigskip
\noindent Now, assume that the case $(\alpha)$ of the clause {\bf (b)} of
the assumptions holds. We reduce this case to the previous one (using
cofinality).
\noindent Take $\bar{C}$, $E$ witnessing that the set $\{\alpha<\lambda:
\cf(\alpha)=\theta\}$ is in $I[\lambda]$ and build a $<_J$-increasing sequence
$\bar{\eta}'=\langle\eta_\alpha^\prime:\alpha<\lambda\rangle\subseteq
\prod\limits_{i<\delta}\lambda_i$ such that $\eta_\alpha^\prime>\eta_\alpha$
and $\bar{\eta}'$ satisfies the clause $(\beta)$ of {\bf (b)} for $\bar{C}$,
$E$. [The construction of $\eta_\alpha^\prime$ is by induction on
$\alpha<\lambda$. Suppose that we have defined $\eta_\beta^\prime$ for
$\beta<\alpha$. Now, at the stage $\alpha$ of the construction, we first
choose $\eta^0_\alpha\in\prod\limits_{i<\delta} \lambda_i$ such that
\[(\forall\beta<\alpha)(\eta_\beta^\prime<_J \eta^0_\alpha).\]
This is possible since the condition $(\alpha)$ implies that $\lambda=
{\rm tcf}(\prod_{i<\delta}\lambda_i/J)$, $\alpha<\lambda$. Now we put for
$i<\delta$:
\[\eta_\alpha^\prime(i)=\max\big\{\eta_\alpha^0(i),\eta_\alpha(i)+1,
\sup\{\eta_\gamma^\prime(i)+1: \gamma\in C_\alpha\}\big\}.\]
Now one can check that this $\bar{\eta}^\prime$ is as required.]
\smallskip
Now we use the fact that $\bar{\eta}$ is cofinal. The set
\[E'=\{\gamma\in E:(\forall\alpha<\gamma)(\exists\beta<\gamma)
(\eta_\alpha^\prime<_J \eta_\beta)\}\]
is a club of $\lambda$. Look at $\bar{\eta}\rest E'$. Suppose that $Y\in
[E']^\theta$. Without loss of generality we may assume that ${\rm otp}(Y)=\theta$
and let $\alpha=\sup(Y)$. By induction on $\varepsilon<\theta$ choose
$\alpha_\varepsilon<\beta_\varepsilon<\gamma_\varepsilon$ such that
\begin{quotation}
$\beta_\varepsilon\in Y$, $\alpha_\varepsilon\in C_\alpha$,
$\gamma_\varepsilon\in C_\alpha$, $\eta_{\alpha_\varepsilon}^\prime<_J
\eta_{\beta_\varepsilon}<_J \eta_{\gamma_\varepsilon}^\prime$ and
if $\zeta<\varepsilon$ then $\gamma_\zeta<\alpha_\varepsilon$.
\end{quotation}
Next choose $i_\varepsilon>i^*$ such that
\[\eta_{\alpha_\varepsilon}^\prime(i_\varepsilon)< \eta_{\beta_\varepsilon}
(i_\varepsilon)<\eta_{\gamma_\varepsilon}^\prime(i_\varepsilon).\]
We may assume that $i_\varepsilon=i^\otimes$ for all $\varepsilon<\theta$.
Now, as $\bar{\eta}^\prime$ obeys $\bar{C}$, we have
\[\zeta<\varepsilon\ \ \ \Rightarrow\ \ \ \eta_{\gamma_\zeta}^\prime
(i^\otimes)<\eta_{\alpha_\varepsilon}^\prime (i^\otimes)\]
and hence we conclude that the sequence $\langle\eta_{\beta_\varepsilon}
(i^\otimes): \varepsilon<\theta\rangle$ is strictly increasing.
Now we may finish the proof like earlier. \QED
\begin{conclusion}
\label{moreconclusion}
If $\mu$ is a strong limit singular cardinal, $2^\mu=\mu^+=\lambda$ and
$\neg(\exists 0^{\#})$ or at least
\[\{\delta<\mu^+: \cf(\delta)=(2^{<\cf(\mu)})^+\}\in I[\lambda]\]
then there is a $(\cf(\mu),\mu,\lambda)$-constructor ${\cal C}$ such that
the algebra ${\Bbb B}^{{\rm red}}({\cal C})$ has the $(2^{<\cf(\mu)})^+$--Knaster property,
its counterpart ${\Bbb B}^{{\rm green}}({\cal C})$ is $\lambda$--cc and the free product is
not $\lambda$--cc.
\noindent [Note that if {\rm GCH} holds then $(2^{<\cf(\mu)})^+=(\cf(\mu))^+$
so the problem is closed then.]
\end{conclusion}
\Proof Like \ref{main} using \ref{getfree}, \ref{morekna} instead of
\ref{stronger}, \ref{consknaster}. \QED
\section{The use of pcf}
Assuming that $2^{{<}\kappa}$ is much larger than $\kappa=\cf(\kappa)$
($=\cf(\mu)<\mu$) we may still want to have examples with the $(\kappa^+,{\rm
not}\lambda)$--Knaster property and the non-multiplicativity. Here
\ref{moreconclusion} does not help if GCH holds on an end segment of the
cardinals (and $\neg(\exists 0^\#)$). We try to remedy this.
It is done inductively. So \ref{startind} uses $\cf(\mu)=\aleph_0$ just to
start the induction. We can phrase (a part of) it without this assumption but
in applications we use it for $\cf(\mu)=\aleph_0$. Also \ref{startind}(b)
really needs this condition (otherwise we would have to assume that
$(\forall\alpha<\theta)(|\alpha|^{{<}\delta}<\mu)$). This result says that, if
$\cf(\mu)=\aleph_0$, then we have gotten the $\theta$--Knaster property for
{\em every} regular cardinal $\theta\in\mu\setminus \kappa^+$.
\begin{definition}
\begin{enumerate}
\item By ${\cal K}_{{\rm wmk}}$ we will denote the class of all tuples $(\theta,\lambda,
\chi,J)$ such that $\theta<\lambda$, $\chi$ are regular cardinals, $J$ is a
$\chi$--complete ideal on $\lambda$ and there is a $(\lambda,\chi)$--well
marked Boolean algebra $({\Bbb B},\bar{y},J)$ (see \ref{well1}) such that the
algebra ${\Bbb B}$ satisfies the $\theta$--Knaster property (${\rm wmk}$ stays for
``{\bf w}ell {\bf m}arked {\bf K}naster'').
When we write $(\theta,\lambda)\in{\cal K}_{{\rm wmk}}$ we really mean $(\theta,\lambda,
\lambda,J^{{\rm bd}}_\lambda)\in {\cal K}_{{\rm wmk}}$ (what means just that there exists a
$(\theta,{\rm \lambda})$--Knaster marked Boolean algebra).
\item By ${\cal K}_{{\rm smk}}$ (${\rm smk}$ is for ``{\bf s}equence {\bf m}arked {\bf
K}naster'') we will denote the class of all triples $(\theta,\lambda,\chi)$ of
cardinals such that $\theta<\lambda$ are regular and there is a sequence
$\langle({\Bbb B}_\alpha,\bar{y}^\alpha):\alpha<\chi \rangle$ of $\lambda$--marked
Boolean algebras such that (for $\alpha<\chi$) the algebras ${\Bbb B}_\alpha$ have
the $\theta$--Knaster property, $\bar{y}^\alpha=\langle y^\alpha_i:
i<\lambda\rangle$ and
\begin{quotation}
\noindent {\em if} $n<\omega$, $\alpha_0<\ldots<\alpha_{n-1}<\chi$ and
$\beta_{\varepsilon,\ell}<\lambda$ for $\varepsilon<\lambda$, $\ell<n$ are
such that $(\forall\varepsilon_1<\varepsilon_2<\lambda)(\forall \ell<n)
(\beta_{\varepsilon_1,\ell}<\beta_{\varepsilon_2,\ell})$
\noindent{\em then} there are $\varepsilon_1<\varepsilon_2<\lambda$ such that
\[\ell<n\ \ \ \Rightarrow\ \ \ {\Bbb B}_{\alpha_\ell}\models\mbox{``}
y^{\alpha_\ell}_{\beta_{\varepsilon_1,\ell}}\cap
y^{\alpha_\ell}_{\beta_{\varepsilon_2,\ell}}={\bf 0}\mbox{''}.\]
\end{quotation}
\end{enumerate}
\end{definition}
\begin{remark}
{\em
\begin{enumerate}
\item On some closure properties of ${\cal K}_{{\rm wmk}}^\theta\stackrel{\rm def}{=}
\{\lambda: (\theta,\lambda)\in{\cal K}_{{\rm wmk}}\}$ under pcf see \ref{clospcf}: if
$\lambda_i\in{\cal K}^\theta_{{\rm wmk}}$ (for $i<\delta$), $\lambda_i>\max{\rm pcf}\/
\{\lambda_j: j<i\}$ and $\lambda\in{\rm pcf}\/\{\lambda_i: i<\delta\}$ and
$(\forall\alpha<\theta)(|\alpha|^{|\delta|}<\theta)$ {\em then}
$\lambda\in{\cal K}^\theta_{{\rm wmk}}$.
\item We can replace $\theta$ by a set $\Theta$ of such cardinals, no real
difference. And thus we may consider the class ${\cal K}^*_{{\rm wmk}}$ of all tuples
$(\Theta,\lambda,\chi,J)$ such that there exists a $(\lambda,\chi)$--well
marked Boolean algebra $({\Bbb B},\bar{y},J)$ with
\[(\forall\theta\in\Theta)({\Bbb B}\mbox{ satisfies the }\theta\mbox{--Knaster
property}).\]
\end{enumerate}
}
\end{remark}
\begin{proposition}
\label{startind}
Assume that $\mu$ is a strong limit singular cardinal, $\aleph_0=\cf(\mu)<\mu$
and $\lambda=2^\mu=\mu^+$.
\begin{description}
\item[(a)] If $(\forall\alpha<\theta)(|\alpha|^{\cf(\mu)}<\theta=\cf(\theta)<
\lambda)$
then $(\theta,\lambda)\in{\cal K}_{{\rm wmk}}$. Moreover $(\theta,\lambda,2^\lambda)\in
{\cal K}_{{\rm smk}}$.
\item[(b)] If $\cf(\mu)<\theta=\cf(\theta)<\mu$ and $\{\alpha<\lambda:
\cf(\alpha)=\theta\}\in I[\lambda]$
then $(\theta,\lambda)\in {\cal K}_{{\rm wmk}}$. Moreover $(\theta,\lambda,2^\lambda)\in
{\cal K}_{{\rm smk}}$.
\end{description}
\end{proposition}
\Proof This is similar to previous proofs and the first parts of
\ref{startind}{\bf (a)}, {\bf (b)} follow from what we have done already:
\ref{startind}{\bf (a)} is an obvious modification of \ref{knacon};
\ref{startind} is similar, but based on \ref{morekna}, \ref{getfree} (and
\ref{stronger}, \ref{notknaster}) (see below). What we actually have to prove
are the ``moreover'' parts. We will sketch the proof of it for clause {\bf
(b)} only, modifying the proof of \ref{main}.
As in \ref{main} we choose a function $h:\cf(\mu)\longrightarrow\omega$ such
that for each $n\in\omega$ the preimage $h^{-1}[\{n\}]$ is unbounded (in
$\cf(\mu)$). Next we take an increasing equence $\langle\mu_i:i<\cf(\mu)
\rangle$ of regular cardinals such that $\mu=\sum\limits_{i<\delta}\mu_i$.
Finally (like in \ref{main}) we construct $\lambda_i$, $\chi_i$,
$({\Bbb B}_i,\bar{y}_i)$ and $I_i$ such that for $i<\cf(\mu)$:
\begin{enumerate}
\item $\lambda_i,\chi_i<\mu$ are regular cardinals,
\item $\lambda_i>\chi_i\geq\prod\limits_{j<i}\lambda_j + \mu_i$,
$\chi_0> \theta + \mu_0$,
\item $I_i$ is a $\chi_i^+$-complete ideal on $\lambda_i$,
\item $({\Bbb B}_i,\bar{y}_i)$ is a $\lambda_i$-marked Boolean algebra such that
\noindent {\em if} $n=h(i)$ and the set $Y\subseteq (\lambda_i)^{n+1}$ is such
that
\[(\exists^{I_i}\gamma_0)\ldots(\exists^{I_i}\gamma_n)(\langle
\gamma_0,\ldots,\gamma_n\rangle\in Y)\]
\noindent{\em then} for some $\gamma_\ell^\prime,\gamma_\ell^{\prime\prime}<
\lambda_i$ (for $\ell\leq n$) we have
\[\langle\gamma_\ell^\prime:\ell\leq n\rangle, \langle\gamma_\ell^{\prime
\prime}:\ell\leq n\rangle\in Y\quad\mbox{ and for all }\ell\leq n\]
\[{\Bbb B}_{i}\models y^i_{\gamma^\prime_\ell}\cap y^i_{\gamma^{\prime
\prime}_\ell}={\bf 0},\]
\item each algebra ${\Bbb B}_i$ satisfies the $\theta$--Knaster condition,
\item for $\xi<\lambda_i$ the set $[\xi,\lambda_i)$ is not in the ideal $I_i$.
\end{enumerate}
Note that the last requirement is new here. Though we cannot demand that the
ideals $I_i$ extend $I^{{\rm bd}}_{\lambda_i}$, the condition (6) above is
satisfied in our standard construction. Note that the ideal from \ref{algide}
has this property if $\lambda$ there is regular. Moreover it is preserved when
the (finite) products of ideals (as in \ref{finite}) are considered. Also, if
$I$ is an ideal on $\lambda$, $A_0\in I$ is such that $|\lambda\setminus A_0|$
is minimal and $A_1\in I^+$ is such that $|A_1|$ is minimal then we can use
either $I\rest A_0$ or $I\rest A_1$. All relevant information is preserved
then (in the first case the condition (6) holds in the second $J^{{\rm bd}}_\lambda
\subseteq I$ -- under suitable renaming).
\noindent Now we put $T=\bigcup\limits_{i<\delta}\prod\limits_{j<i}\lambda_j$,
${\Bbb B}_\eta={\Bbb B}_{{\rm lg}\/(\eta)}$, $\bar{y}_\eta=\bar{y}_{{\rm lg}\/(\eta)}$, $I_\eta=
I_{{\rm lg}\/(\eta)}$. Applying \ref{stronger} we find a stronger
$J^{{\rm bd}}_\delta$-cofinal sequence $\bar{\eta}=\langle \eta_\alpha:
\alpha<\lambda\rangle$ for $(T,\bar{\lambda},\bar{I})$. Due to the requirement
(6) above we may additionally demand that $\bar{\eta}$ is
$<_{J^{{\rm bd}}_{\cf(\mu)}}$--increasing cofinal in $\prod\limits_{i<\cf(\mu)}
\lambda_i/J^{{\rm bd}}_{\cf(\mu)}$. Let $\langle B_\xi: \xi< 2^\lambda\rangle$ be a
sequence of pairwise almost disjoint elements of $[\lambda]^\lambda$ (i.e.
$|B_\xi\cap B_\zeta|<\lambda$ for distinct $\xi,\zeta<2^\lambda$). For each
$\xi< 2^\lambda$ we may apply \ref{getfree} (the version of {\bf
(b)}($\alpha$)) to the sequence $\langle\eta_\alpha:\alpha\in B_\xi\rangle$
and we find $A_\xi\in [B_\xi]^\lambda$ such that each sequence $\langle
\eta_\alpha:\alpha\in A_\xi\rangle$ is $\theta$--hereditary free. Let
\[{\Bbb B}^*_\xi={\Bbb B}^{{\rm red}}(T,\bar{\lambda},\langle\eta_\alpha: \alpha\in
A_\xi\rangle,\langle({\Bbb B}_\eta,\bar{y}_\eta): \eta\in T\rangle),\quad\quad
\bar{x}_\xi= \langle x^{{\rm red}}_{\alpha}: \alpha\in A_\xi\rangle.\]
Of course each ${\Bbb B}^*_\xi$ is a subalgebra of ${\Bbb B}^{{\rm red}}(T,\bar{\lambda},
\bar{\eta},\langle({\Bbb B}_\eta,\bar{y}_\eta): \eta\in T\rangle)$ (generated by
$\bar{x}_\xi$). By \ref{morekna} and \ref{notknaster} we know that the marked
Boolean algebras $({\Bbb B}^*_\xi, \bar{x}_\xi)$ are $(\theta,{\rm
not}\lambda)$--Knaster. To show that they witness $(\theta,\lambda,2^\lambda)
\in {\cal K}_{{\rm smk}}$ suppose that $n<\omega$, $\xi_0,\ldots,\xi_{n-1}< 2^\lambda$,
$\beta_{\varepsilon,\ell}<\lambda$ (for $\varepsilon<\lambda$, $\ell<n$) are
such that
\[(\forall\varepsilon_1<\varepsilon_2<\lambda)(\forall\ell<n)
(\beta_{\varepsilon_1,\ell}< \beta_{\varepsilon_2,\ell})\]
and of course $\{\beta_{\varepsilon,\ell}: \varepsilon<\lambda\}\subseteq
A_{\xi_\ell}$. Since $A_{\xi_\ell}$ are almost disjoint we may assume that
\[(\forall\varepsilon_1,\varepsilon_2<\lambda)(\forall\ell_1<\ell_2<n)
(\beta_{\varepsilon_1,\ell_1}\neq \beta_{\varepsilon_2,\ell_2}).\]
Further we may assume that we have $i^*<\cf(\mu)$ such that for each
$\varepsilon<\lambda$ the sequences $\eta_{\beta_{\varepsilon,\ell}}\rest i^*$
for $\ell< n$ are pairwise distinct.
\noindent By the choice of $\bar{\eta}$, $T$, $\bar{\lambda}$ etc we may apply
\ref{claimgreen} and conclude that for all sufficiently large
$\varepsilon<\lambda$ the set
\[\begin{array}{r}
Z_\varepsilon=\{i<\cf(\mu): \neg(\exists^{I_{\eta_{\beta_{\varepsilon,0}}\rest
i}}\gamma_{0})\ldots(\exists^{I_{\eta_{\beta_{\varepsilon,n-1}}\rest i}}
\gamma_{n-1})(\exists^\lambda\zeta)(\forall \ell<n)\\
(\eta_{\beta_{\varepsilon,\ell}}\rest (i+1)=(\eta_{\beta_{\varepsilon,\ell}}
\rest i)\hat{\ }\langle\gamma_\ell\rangle)\}\\
\end{array}
\]
is in the ideal $J^{{\rm bd}}_{\cf(\mu)}$. Take one such $\varepsilon$. Choosing
$i\in \cf(\mu)\setminus Z_\varepsilon$, $i>i^*$ such that $h(i)=n$ we may
follow exactly as in the last part of the proof of \ref{green} and we find
$\varepsilon_0,\varepsilon_1<\lambda$ such that for each $\ell<n$
\begin{quotation}
\noindent $\eta_{\beta_{\varepsilon_0,\ell}}\rest i=
\eta_{\beta_{\varepsilon_1,\ell}}\rest i$ but
\noindent ${\Bbb B}_{\eta_{\beta_{\varepsilon_0,\ell}}\rest i}\models
y_{\eta_{\beta_{\varepsilon_0,\ell}}\rest (i+1)} \cap
y_{\eta_{\beta_{\varepsilon_1,\ell}}\rest (i+1)}={\bf 0}$
\end{quotation}
which implies that
\[(\forall\ell<n)({\Bbb B}^*_{\xi_\ell}\models x^{{\rm red}}_{\beta_{\varepsilon_0,\ell}}
\cap x^{{\rm red}}_{\beta_{\varepsilon_1,\ell}}={\bf 0}).\]
\QED
\begin{proposition}
\label{contind}
Assume that
\begin{description}
\item[(a)] $\langle\lambda_i: i<\delta\rangle$ is an increasing sequence of
regular cardinals such that $\delta<\lambda_0$,
$\lambda_i>\max{\rm pcf}\/\{\lambda_j: j<i\}$ (the last is our natural assumption);
\item[(b)] $\aleph_0<\theta=\cf(\theta)<\bigcup\limits_{i<\delta}\lambda_i$
(naturally we assume just $\cf(\theta)=\theta<\lambda_0$)
\item[(c)] $\lambda=\max{\rm pcf}\/\{\lambda_i: i<\delta\}$
\item[(d)] $(\theta,\lambda_i,\max{\rm pcf}\/\{\lambda_j:j<i\})\in{\cal K}_{{\rm smk}}$
\item[(e)] for each $\tau\in\{\lambda\}\cup\bigcup\limits_{\alpha<\delta}
{\rm pcf}\/\{\lambda_i: i<\alpha\}$ we have
\[\{\xi<\tau: \cf(\xi)=\theta\}\in I[\tau]\]
or at least
\begin{quotation}
\noindent for some $\bar{f}^\tau=\langle f^\tau_\varepsilon: \varepsilon<
\tau\rangle$, $<_{J_{=\tau}}$--increasing cofinal in $\prod\limits_{i<\alpha}
\lambda_i/J_{=\tau}$ we have:
\[\gamma<\tau\ \&\ \cf(\gamma)=\theta\ \ \ \Rightarrow\ \ \
f^\tau_\gamma\mbox{ is good in }\bar{f}^\tau\]
\end{quotation}
(see \cite{Sh:345a}, \cite{Sh:355}, section 1 (see 1.6(1) and then
\cite{MgSh:204}),
\item[(f)] $|{\rm pcf}\/\{\lambda_i:i<\delta\}|<\theta$ or at least for each
$\alpha<\delta$ we have $|{\rm pcf}\/\{\lambda_i: i<\alpha\}|<\theta$.
\end{description}
{\em Then} $(\theta,\lambda)\in{\cal K}_{{\rm wmk}}$. Moreover
$(\theta,\lambda,\chi)\in{\cal K}_{{\rm smk}}$ provided there is an almost disjoint
family of size $\chi$ in $[\lambda]^\lambda$. We may get algebras ${\Bbb B}^{{\rm red}}$,
${\Bbb B}^{{\rm green}}$ as in main constructions such that
\[{\Bbb B}^{{\rm red}}\models\theta\mbox{-Knaster},\quad{\Bbb B}^{{\rm green}}\models\lambda
\mbox{-cc}\quad\mbox{ and }{\Bbb B}^{{\rm red}}*{\Bbb B}^{{\rm green}}\models\neg\lambda
\mbox{-cc.}\]
\end{proposition}
\noindent{\bf Remark~\ref{contind}.A:}\ \ \ Continues also the proof of 3.5 of
\cite{Sh:g}.
\medskip
\Proof The main difficulty of the proof will be to construct a hereditary
$\theta$--free $<_{J_{{<}\lambda}}$--increasing sequence $\bar{\eta}=\langle
\eta_\alpha:\alpha<\lambda\rangle \subseteq \prod\limits_{i<\delta}\lambda_i$.
This is done in the claim below. For the notation used there let us note that
if $\alpha\leq\delta$ is a limit ordinal, $\tau\in {\rm pcf}\/\{\lambda_i:
i<\alpha\}$ then $J_{=\tau}[\{\lambda_i: i<\alpha\}]=J^\alpha_\tau$ is the
ideal on $\alpha$ generated by
\[J_{{<}\tau}[\{\lambda_i: i<\alpha\}]\cup \{\alpha\setminus {\frak b}_\tau
[\{\lambda_i: i<\alpha\}]\}.\]
So in particular ${\rm tcf}(\prod\limits_{\i<\alpha}\lambda_i/J^\alpha_\tau)=\tau$.
\begin{claim}
There exists a tree $T\subseteq \bigcup\limits_{i<\delta}\prod\limits_{j<i}
\lambda_j$ such that $\lim_\delta(T)$ is $\theta$--hereditary free (and
$<_{J_{{<}\lambda}}$--cofinal).\\
Moreover for each $\alpha<\delta$ the size of $T_\alpha$ is
$\leq\max{\rm pcf}\/\{\lambda_i: i<\alpha\}$.
\end{claim}
\noindent{\em Why?}\ \ \ For a limit ordinal $\alpha\leq\delta$ and $\tau\in
{\rm pcf}\/\{\lambda_i: i\leq\alpha\}$ (if $\alpha=\delta$ then $\tau=\lambda$)
choose a $<_{J^\alpha_\tau}$--increasing sequence $\bar{f}^{\alpha,\tau}=
\langle f^{\alpha,\tau}_\zeta: \zeta<\tau\rangle\subseteq\prod
\limits_{i<\alpha} \lambda_i$ cofinal in $\prod\limits_{i<\alpha}\lambda_i
/J^\alpha_\tau$ and such that
\begin{description}
\item[$(\tilde{\otimes})$] {\em if} $\zeta<\tau$, $\cf(\zeta)=\theta$
{\em then} for some unbounded set $Y_\zeta\subseteq \zeta$ (for simplicity
consisting of successor ordinals) and a sequence $\bar{s}^\tau=\langle
s^\tau_\xi: \xi\in Y_\zeta\rangle\subseteq J^\alpha_\tau$ we have
\[[\xi_1,\xi_2\in Y_\zeta\ \&\ \xi_1<\xi_2\ \&\ i\in\alpha\setminus
(s^\tau_{\xi_1}\cup s^\tau_{\xi_2})]\quad \Rightarrow\quad
f^{\alpha,\tau}_{\xi_1}(i) < f^{\alpha,\tau}_{\xi_2}(i).\]
\end{description}
[Why we can demand $(\tilde{\otimes})$? If in the assumption {\bf (e)} the
first part is satisfied then we follow similarly to the proof of
\ref{getfree}, compare \cite{Sh:355}, 1.5A, 1.6, pp 51--52. If we are in the
case of ``at least'' then this is exactly the meaning of goodness.]
Further we may demand that the sequence $\bar{f}^{\alpha,\tau}$ is
$^b$continuous:
\begin{description}
\item[$(\tilde{\oplus})$] {\em if} $|\delta|<\cf(\zeta)<\lambda_0$,
$\zeta<\tau$
{\em then} $f^{\alpha,\tau}_\zeta(i)=\min\{\bigcup\limits_{\xi\in C}
f^{\alpha,\tau}_\xi(i): C$ is a club of $\zeta\}$
\end{description}
[compare the proof of 3.4 of \cite{Sh:345a}, pp 25--26].
For a limit ordinal $\alpha<\delta$ we define
\[\begin{array}{ll}
T^0_\alpha=\{f{\in}\prod\limits_{i<\alpha}\lambda_i: &\mbox{\bf (a) }f=
\max\{f^{\alpha,\tau_\ell}_{\zeta_\ell}: \ell{<}n\}\mbox{ for some}\\
\ &\quad n<\omega, \tau_\ell\in{\rm pcf}\/\{\lambda_i: i<\alpha\},
\zeta_\ell<\tau_\ell\\
\ &\mbox{{\bf (b)} for every }\tau{\in}{\rm pcf}\/\{\lambda_i: i{<}\alpha\}\mbox{ if
} \tau=\lambda\mbox{ or }\alpha<\delta\mbox{ then}\\
\ &\mbox{there is } \zeta_f(\tau){<}\tau\mbox{ such that}\\
\ &\quad f^{\alpha,\tau}_{\zeta_f(\tau)}\leq f\ \&\
f^{\alpha,\tau}_{\zeta_f(\tau)} = f\mbox{ mod } J^\alpha_\tau\}.\\
\end{array}\]
(Note that if $\alpha=\delta$ then there is only one value of $\tau_\ell,\tau$
which we consider here: $\lambda$.) Let
$T'\subseteq\bigcup\limits_{i\leq\delta}\prod_{j<i}\lambda_j$ be a tree such
that for $\gamma\leq\delta$:
\[T_\gamma^\prime=\{f\in\prod_{i<\gamma}\lambda_i: f\rest\alpha\in
T^0_\alpha\mbox{ for each limit }\alpha\leq\gamma\}.\]
Let
\[A=\{\zeta<\lambda: (\exists
f\in\prod_{i\leq\delta}\lambda_i)[f^{\delta,\lambda}_\zeta\leq f\ \&\
f^{\delta,\lambda}_\zeta=f \mbox{ mod}\; J^\alpha_\lambda\ \&\ (\forall i\leq
\delta)(f\rest i\in T^\prime_\alpha)]\}\]
and for each $\zeta\in A$ let $f^*_\zeta$ be a function witnessing it. Now let
$T\subseteq\bigcup\limits_{i\leq\delta}\prod_{j<i}\lambda_j$ be a tree such
that $T_\delta=\{f^*_\zeta: \zeta\in A\}$.
\noindent By the definition, $T$ is a tree, but maybe it does not have enough
levels? Let $\chi$ be a large enough regular cardinal. Take an increasing
continuous sequence $\langle N_i: i\leq \theta\rangle$ of elementary submodels
of $({\cal H}(\chi),{\in},{<}^*)$ such that
\begin{quotation}
$|N_i|=\Upsilon=\theta+|{\rm pcf}\/\{\lambda_\alpha: \alpha<\delta\}|<\lambda_0$,
$\Upsilon+1\subseteq N_i\in N_{i+1}$,
all relevant things are in $N_0$.
\end{quotation}
Now we define $f^*\in\prod\limits_{\alpha<\delta}\lambda_\alpha$ by
\[f^*(\alpha)=\sup(\bigcup_{i<\theta}N_i\cap\lambda_\alpha).\]
Similarly as in \cite{Sh:355}, pp 63--65, one proves that $f^*\rest\alpha\in
T^0_\alpha$ for each limit $\alpha\leq\delta$. Hence for some $\zeta$ we have
$f^*= f^{\delta,\lambda}_\zeta\ \mbox{mod}\; J^\delta_\lambda$ and thus
$\zeta\in A$. Consequently $A$ is unbounded in $\lambda$.
By induction on $\alpha\leq\delta$ we prove that
\begin{description}
\item[$(\tilde{\circledcirc})$] {\em if} $f_\zeta\in T_\alpha$ (for
$\zeta<\theta$) are pairwise distinct
{\em then} there are $Z\in [\theta]^\theta$ and $j<\alpha$ such that
\[(\forall \zeta_0,\zeta_1\in Z)(\zeta_0\neq\zeta_1\ \Rightarrow\
[f_{\zeta_0}\rest j=f_{\zeta_1}\rest j\ \&\ f_{\zeta_0}(j)\neq f_{\zeta_1}(j)]
).\]
\end{description}
If $\alpha$ is a non-limit ordinal then this is trivial. So suppose that
$\alpha$ is limit, $\alpha<\delta$. Then for some
$\tau_{\zeta,\ell}\in{\rm pcf}\/\{\lambda_i: i<\alpha\}$,
$\xi_{\zeta,\ell}<\tau_{\zeta,\ell}$, $n_\zeta<\omega$ (for $\zeta<\theta$,
$\ell<n_\zeta$) we have
\[f_\zeta=\max\{f^{\alpha,\tau_{\zeta,\ell}}_{\xi_{\zeta,\ell}}: \ell<
n_\zeta\}.\]
As $\theta>|{\rm pcf}\/\{\lambda_\beta: \beta<\alpha\}|$ we may assume that
$n_\zeta=n^*$, $\tau_{\zeta,\ell}=\tau_\ell$ and for each $\ell<n^*$ the
sequence $\langle\xi_{\zeta,\ell}: \zeta<\theta\rangle$ is either constant or
strictly increasing. Now, the second case has to occur for some $\ell$ and we
may follow similarly to \ref{claimfree} and then apply the inductive
hypothesis. We are left with the case $\alpha=\delta$. So let
$f_\zeta=f^*_{\beta_\zeta}$ for $\zeta<\delta$ and we continue as before (with
$\lambda$ for $\tau_\ell$).
\noindent This ends the proof of the claim (note that the arguments showing
that all the $T^0_\alpha$ are not empty prove actually that the tree $T$ has
enough branches to satisfy our additional requirements).
\medskip
Now let $T$ be a tree is in the claim above. Let
$\bar{\eta}=\langle\eta_\alpha: \alpha<\lambda\rangle\subseteq\lim_\delta(T)$
be the enumeration of $\{f^*_\zeta: \zeta\in A\}$ such that $\bar{\eta}$ is
$<_{J_{<\lambda}}$--increasing cofinal in
$\prod\limits_{i<\delta}\lambda_i/J_{<\lambda}$. By the assumption {\bf (d)}
for each $\eta\in T$ we find a marked Boolean algebra
$({\Bbb B}_\eta,\bar{y}_\eta)$ such that for every $i<\delta$ the sequence $\langle
({\Bbb B}_\eta,\bar{y}_\eta): \eta\in T_i\rangle$ witnesses that
$(\theta,\lambda_i, |T_i|)\in{\cal K}_{{\rm smk}}$. These parameters determine a
$(\delta,\mu,\lambda)$--constructor ${\cal C}$, so we have the respective Boolean
algebra ${\Bbb B}^{{\rm red}}({\cal C})$ (and its counterpart ${\Bbb B}^{{\rm green}}({\cal C})$). To show
that they have the required properties we follow exactly the proof that
$(\theta,\lambda,\chi)\in{\cal K}_{{\rm smk}}$, so we will present this only.
First note that by \ref{morekna} the algebra ${\Bbb B}^{{\rm red}}({\cal C})$ has the
$\theta$--Knaster property. Now, let $\langle A_\zeta: \zeta<\chi\rangle
\subseteq [\lambda]^\lambda$ be such that
\[\zeta_1<\zeta_2<\chi\ \ \ \Rightarrow\ \ \ |A_{\zeta_1}\cap A_{\zeta_2}| <
\lambda.\]
Let $\bar{x}_\zeta=\langle x^{{\rm red}}_\xi: \xi\in A_\zeta\rangle$ and let
${\Bbb B}_\zeta$ be the subalgebra of ${\Bbb B}^{{\rm red}}({\cal C})$ generated by
$\bar{x}_\zeta$. We want to show that the sequence $\langle
({\Bbb B}_\zeta,\bar{x}_\zeta): \zeta<\chi\rangle$ witnesses
$(\theta,\lambda,\chi)\in {\cal K}_{{\rm smk}}$. For this suppose that $\zeta_0<\ldots<
\zeta_{n-1}<\chi$, $n<\omega$ and $\beta_{\varepsilon,\ell}\in A_{\zeta,\ell}$
are increasing with $\varepsilon$ (for $\varepsilon<\lambda$, $\ell<n$) and
without loss of generality with no repetition. We may assume that
\[(\forall\ell<n)(\forall\varepsilon<\lambda)(\beta_{\varepsilon,\ell}\notin
\bigcup_{m\neq\ell} A_{\zeta_m}).\]
Further we may assume that for some $i^*<\delta$ and pairwise distinct
$\eta_\ell\in T_{i^*}$ (for $\ell<n$) we have
\[(\forall\varepsilon<\lambda)(\forall\ell<n)(\eta_{\beta_{\varepsilon,\ell}}
\rest i^*=\eta_\ell).\]
Now we take $i\in [i^*,\delta)$ such that
\[(\forall\gamma<\lambda_i)(\exists^\lambda \varepsilon<\lambda)(\forall
\ell<n)(\eta_{\beta_{\varepsilon,\ell}}(i)>\gamma)\]
(remember that each $\langle\eta_{\beta_{\varepsilon,\ell}}:
\varepsilon<\lambda\rangle$ is $<_{J_{<\lambda}}$--cofinal). Since $|T_i|<
\lambda_i$ we can find $\nu_0,\ldots,\nu_{n-1}\in T_i$ such that
$\eta_\ell\trianglelefteq\nu_\ell$ and
\[(\forall\gamma<\lambda_i)(\exists^\lambda \varepsilon<\lambda)(\forall
\ell<n)(\eta_{\beta_{\varepsilon,\ell}}\rest i=\nu_\ell\ \&\
\eta_{\beta_{\varepsilon,\ell}}(i)>\gamma).\]
Consequently we may choose a sequence $\langle\langle \gamma_{\xi,\ell}:\ell<
n\rangle: \xi<\lambda_i\rangle\subseteq\lambda_i$ such that
$\xi<\gamma_{\xi,\ell}$ and
\[(\forall\xi<\lambda_i)(\exists^\lambda\varepsilon<\lambda)(\forall \ell<n)
(\eta_{\beta_{\varepsilon,\ell}}\rest (i+1)=\nu_\ell\hat{\ }\langle
\gamma_{\xi,\ell}\rangle).\]
Now we use the choice of $({\Bbb B}_{\nu_\ell},\bar{y}_{\nu_\ell})$ (witnessing
$(\theta,\lambda_i,|T_i|)\in {\cal K}_{{\rm smk}}$) and we find $\xi_1<\xi_2<\lambda_i$
such that
\[(\forall\ell<n)({\Bbb B}_{\nu_\ell}\models y^{\nu_\ell}_{\gamma_{\xi_1,\ell}}
\cap y^{\nu_\ell}_{\gamma_{\xi_2,\ell}}={\bf 0}),\]
which allows us to find $\varepsilon_1<\varepsilon_2<\lambda$ such that for
each $\ell<n$ the intersection $x_{\beta_{\varepsilon_1,\ell}}\cap
x_{\beta_{\varepsilon_2,\ell}}$ is ${\bf 0}$. \QED
\begin{conclusion}
If $\langle\mu_i: i\leq\kappa\rangle$ is a strictly increasing continuous
sequence of strong limit singular cardinals such that $\kappa<\mu_0$,
$2^{\mu_i}=\mu_i^+$, $\kappa<\theta=\cf(\theta)<\mu_0$ and
\[\{\alpha<\mu^+_i: \cf(\alpha)=\theta\}\in I[\mu_i^+]\]
{\em then} $(\theta,\mu_\kappa^+)\in{\cal K}_{{\rm wmk}}$ and we may construct the
respective Boolean algebras ${\Bbb B}^{{\rm red}}$, ${\Bbb B}^{{\rm green}}$. \QED
\end{conclusion}
\begin{proposition}
Suppose that we have Boolean algebras ${\Bbb B}^{{\rm red}}$, ${\Bbb B}^{{\rm green}}$ such that
\begin{quotation}
\noindent ${\Bbb B}^{{\rm red}}$ satisfies the $\theta$--Knaster condition
\noindent for each $n<\omega$ the free product $({\Bbb B}^{{\rm green}})^n$ satisfies the
$\lambda$--cc
\noindent the free product ${\Bbb B}^{{\rm red}}*{\Bbb B}^{{\rm green}}$ fails the $\lambda$-cc.
\end{quotation}
Then $(\theta,\lambda,\chi)\in{\cal K}_{{\rm smk}}$, where $\chi=\lambda^+$ (or even if
$\chi$ is such that there is an almost disjoint family ${\cal A}\subseteq
[\lambda]^\lambda$ of size $\chi$).
\end{proposition}
\Proof We have $y_\alpha\in({\Bbb B}^{{\rm red}})^+$ and $z_\alpha\in({\Bbb B}^{{\rm green}})^+$
for $\alpha<\lambda$ such that if $\alpha<\beta<\lambda$ then
\[\mbox{either }\quad{\Bbb B}^{{\rm red}} \models y_\alpha\cap y_\beta={\bf 0}\quad\mbox{ or
}\quad{\Bbb B}^{{\rm green}}\models z_\alpha\cap z_\beta={\bf 0}.\]
Let $A_\zeta\in [\lambda]^\lambda$ (for $\zeta<\chi$) be pairwise
almost disjoint sets. We want to show that the sequence
\[\langle({\Bbb B}^{{\rm red}},\bar{y}\rest A_\zeta): \zeta<\chi\rangle\]
is a witness for $(\theta,\lambda,\chi)\in{\cal K}_{{\rm smk}}$. So we are given
$\zeta_0<\zeta_1<\ldots<\zeta_{n-1}<\chi$ and sequences
$\langle\alpha_{\varepsilon,\ell}: \varepsilon<\lambda\rangle\subseteq
A_{\zeta_\ell}$. Then, for some $\varepsilon^*<\lambda$ we have
\[\varepsilon^*\leq\varepsilon<\lambda\ \ \ \Rightarrow\ \ \
\alpha_{\varepsilon,\ell}\notin \bigcup_{m\neq \ell} A_{\zeta,m}.\]
We should find $\varepsilon_1<\varepsilon_2$ such that for all $\ell<n$
\[{\Bbb B}^{{\rm red}}\models y_{\alpha_{\varepsilon_1,\ell}}\cap
y_{\alpha_{\varepsilon_2,\ell}}={\bf 0}.\]
For this it is enough to find $\varepsilon^*<\varepsilon_1<\varepsilon_2$ such
that for $\ell<n$
\[{\Bbb B}^{{\rm green}}\models z_{\alpha_{\varepsilon_1,\ell}}\cap
z_{\alpha_{\varepsilon_2,\ell}}\neq{\bf 0}.\]
But this we easily get from the fact that the free product $({\Bbb B}^{{\rm green}})^n$
satisfies the $\lambda$-cc. \QED
\begin{comment}
The proofs that the algebra ${\Bbb B}^{{\rm green}}$ satisfies the $\lambda$-cc (see
\ref{green}, \ref{contind}) give that actually for each $n<\omega$ the product
$({\Bbb B}^{{\rm green}})^n$ satisfies $\lambda$-cc. So it is reasonable to add it
(though not needed originally).
\end{comment}
\begin{comment}
The ``$\bar{\eta}$ is (strong{-}) $J$-cofinal for $(T,\bar{\lambda},\bar{I})$''
has easy consequences for the existence of colourings.
\end{comment}
\begin{remark}
{\em
For $\mu$ strong limit singular we may sometimes get a cofinal sequence of
length $\lambda\in (\mu,2^\mu]$ without $2^\mu=\mu^+$. By \cite{Sh:430},
section 5,
\noindent{\em if}:
\begin{description}
\item[(a)] $I_i$ is a $\chi_i$--complete, $|I_i|=\tau_i$, $\chi_i$ regular
\item[(b)] $\chi_i\leq\tau_i\leq(\chi_i)^{+n^*}$, $n^*<\omega$
\item[(c)] ${\rm tcf}(\prod\limits_{i<\delta}(\chi_i)^{+\ell}/J)=\lambda$ for each
$\ell\leq n^*$
\end{description}
{\em then}
\begin{description}
\item[$(\alpha)$] there is a cofinal sequence in
$\prod\limits_{i<\delta}({\cal P}(\lambda_i)/I_i)/J$, because
\item[$(\beta)$] it has the true cofinality.
\end{description}
So if for arbitrarily large $\chi$, $2^{\chi}=\chi^+$, $2^{\chi^+}=\chi^{++}$
then we have the ideal we want and maybe the ${\rm pcf}\/$ condition holds. so
combining this and \ref{proexa} below we get that there may be an example of
our kind not because of GCH reasons, but still requiring some cardinal
arithmetic assumptions.
}
\end{remark}
\begin{proposition}
\label{proexa}
Suppose that $\langle\lambda_i: i<\delta\rangle$ is a strictly increasing
sequence of regular cardinals, $I_i$ is a $(\prod\limits_{j<i}
\lambda_j)^+$--complete ideal on $\lambda_i$ (so
$\prod\limits_{j<i}\lambda_j<\lambda_i$) and $({\Bbb B}_i,\bar{y}_i,I_i)$
is a $\lambda_i$--well marked Boolean algebra (for $i<\delta$).
\begin{enumerate}
\item Assume that $\prod\limits_{i<\delta}(I_i,{\subseteq})/J$ has true
cofinality $\lambda$. {\em Then} there exists a $(\theta,{\rm
not}\lambda)$--Knaster marked Boolean algebra.
\item Suppose in addition that $h:\delta\longrightarrow\omega$ is a function
such that
\[(\forall n<\omega)(h^{-1}[\{n\}]\in J^+)\]
and $I^{[h(i)]}_i$ are the product ideals on $(\lambda_i)^n$ (for
$i<\delta$):
\[I^{[h(i)]}_i\stackrel{\rm def}{=}\{B{\subseteq}(\lambda_i)^n:
\neg(\exists^{I_i}\gamma_0)\ldots(\exists^{I_i}\gamma_{h(i)-1})(\langle
\gamma_\ell: \ell< h(i)\rangle\in B).\]
Assume that
\[\lambda={\rm tcf}(\prod_{i<\delta}(I^{[h(i)]}_i,{\subseteq})/J)\]
and that the $({\Bbb B}_i,\bar{y}_i,I_i)$ satisfy the following requirement:
\begin{description}
\item[$(\tilde{*})_{h(i)}$] if $B\subseteq (\dom(\bar{y}_i))^{h(i)}$ is such
that
\[(\exists^{I_i}\gamma_0)\ldots(\exists^{I_i}\gamma_{h(i)})(\langle\gamma_\ell:
\ell\leq h(i)\rangle\in B)\]
then there are $\gamma_\ell^\prime,\gamma_\ell^{\prime\prime}<\lambda_i$ (for
$\ell\leq h(i)$) such that for each $\ell$
\[{\Bbb B}_i\models y_{i,\gamma_\ell^\prime}\cap y_{i,\gamma_\ell^{\prime\prime}}=
{\bf 0}.\]
\end{description}
{\em Then} we can conclude that
$((2^{|\delta|})^+,\lambda,\lambda^+)\in{\cal K}_{{\rm smk}}$ and we have a pair of
algebras $({\Bbb B}^{{\rm red}},{\Bbb B}^{{\rm green}})$ as in main theorem \ref{main}.
\end{enumerate}
\end{proposition}
\Proof The main point here is that with our assumptions in hands we may
construct a sequence $\langle\eta_\alpha: \alpha<\lambda\rangle \subseteq
\prod\limits_{i<\delta}\lambda_i$ which is quite stronger $J$--cofinal:
it satisfies the requirement of \ref{cofinal}(6){\bf (b)} weakened to the
demand that the set there is not in the dual filter $J^c$. Of course this is
still enough to carry our proofs and we may use such a sequence to build the
right examples.
\noindent{\em 1).}\ \ \ Let $\langle\langle A^\alpha_i:i<\delta\rangle:
\alpha<\lambda\rangle$ witness the true cofinality. By induction on
$\alpha<\lambda$ choose $\gamma_\alpha<\lambda$ and $\eta_\alpha\in
\prod\limits_{i<\alpha}\lambda_i$ such that
\begin{quotation}
\noindent $\langle\{\eta_\beta(i)\}: i<\delta\rangle\in\prod\limits_{i<\delta}
I_i$,
\noindent if $\beta<\alpha$ then $\gamma_\beta<\gamma_\alpha$ and $(\forall^J
i)(\eta_\beta(i)\in A^{\gamma_\alpha}_i)$ and
\noindent $\eta_\alpha(i)\notin A^{\gamma_\alpha}_i$.
\end{quotation}
For $\alpha=0$ or $\alpha$ limit, first choose $\gamma_\alpha=
\sup\{\gamma_{\alpha_1}+1: \alpha_1<\alpha\}$ and then choose $\eta_\alpha(i)$
by induction on $i$.\\
For $\alpha=\alpha_1+1$ first note that
\[\langle\{\eta_{\alpha_1}(i)\}:i<\delta\in\prod\limits_{i<\delta} I_i.\]
Hence for some $\gamma^0_\alpha<\lambda$ we have
\[(\forall^J i)(\eta_{\alpha_1}(i)\in A^{\gamma_\alpha}_i).\]
Let $\gamma_\alpha=\max\{\gamma_{\alpha_1},\gamma^0_\alpha\}$. Now choose
$\eta_\alpha(i)$ by induction on $i$.
As $I_i$ is $|T_i|^+$--complete, clearly $\langle\eta_\alpha:
\alpha<\lambda\rangle$ is $J$--cofinal for $(T,J,\bar{I})$ and
\ref{notknaster}, \ref{consknaster} give the conclusion.
\noindent{\em 2).}\ \ \ The construction of $\bar{\eta}$ is in a sense similar
to the one in the proof of \ref{stronger}, but we use our cofinality
assumptions. We have a cofinal sequence in
$\prod_{i<\delta}(I^{[h(i)]}_i,{\subseteq})/J$:
\[\langle\langle A^\alpha_i: i<\delta\rangle: \alpha<\lambda\rangle.\]
For each $A^\alpha_i$ we have ``Skolem functions'' $f^\alpha_{i,\ell}$ for
$\ell< h(i)$ (like in the proofs of \ref{claimgreen}, \ref{supimpfub}).
We define $\eta_\alpha$ by induction on $\alpha<\lambda$.
In the exclusion list we put all substitutions by
$\eta_{\gamma_0}\rest i,\ldots,\eta_{\gamma_{\ell-1}}\rest i$ for
$\gamma_k<\alpha$ to $f^\alpha_{i,\ell}$: each time we obtain a
set in the ideal $I_i$ and a member $\bar{A}$ of $\prod\limits_{i<\delta} I_i$
such that if $(\forall^J i)(\eta(i)\notin A_i)$, $\eta\in\prod
\limits_{i<\delta} \lambda_i$ then $\eta$ satisfies the demand. Eventually we
have $|\alpha|^{<\omega}$ such elements of $\prod\limits_{i<\delta} I_i$. Let
them be $\{\bar{B}^{\alpha,\xi}: \xi\leq |\alpha|+\aleph_0\}$. Then for some
$\gamma_\alpha$
\[(\forall \xi<|\alpha|+\aleph_0)(\forall^J i<\delta)(B^{\alpha,\xi}_i
\subseteq A^{\gamma_\alpha}_i)\]
and similarly
\[(\forall \beta<\alpha)(\forall^J i<\delta)(\eta_\beta(i)\in
A^{\alpha_i}_i).\]
Choose $\eta_\alpha\in\prod\limits_{i<\delta}(\lambda_i\setminus
A^{\gamma_\alpha}_i)$. \QED
\begin{remark}
{\em
One of the main tools used in this section are (variants of) the following
observation:
\begin{quotation}
{\em if} $({\Bbb B},\bar{y})$ is a $\lambda$-marked Boolean algebra such that ${\Bbb B}$
is $\theta$--Knaster and if $\varepsilon(\alpha,\ell)<\lambda$ (for
$\alpha<\lambda$, $\ell<n$) are pairwise distinct then for some
$\alpha<\beta<\lambda$, for each $\ell<n$ we have ${\Bbb B}\models
y_{\varepsilon(\alpha,\ell)}\cap y_{\varepsilon(\beta,\ell)}={\bf 0}$
{\em then} $(\theta,\lambda,\lambda^+)\in{\cal K}_{{\rm smk}}$.
\end{quotation}
}
\end{remark}
\begin{conrem}
{\em
If $\mu$ is a strong limit singular cardinal,
$\cf(\mu)\leq\theta=\cf(\theta)<\mu$
then, by the methods of \cite{Sh:576}, one may get consistency of
\begin{quotation}
if an algebra ${\Bbb B}$ satisfies the $\theta$-ccc
then it satisfies the $\mu^+$-Knaster condition.
\end{quotation}
}
\end{conrem}
One may formulate the following question now:
\begin{pms}
\label{pms}
Suppose that ${\Bbb B}$ is a Boolean algebra satisfying the $\theta$-cc and
$\lambda$ is a regular cardinal between $\mu^+$ and $(2^\mu)^+$.
Does ${\Bbb B}$ satisfy the $\lambda$-Knaster condition?
\end{pms}
There a reasonable amount of information on consistency of the negative answer
in the next section, though \ref{pms} is not fully answered there. But a real
problem is the following.
\begin{problem}
Assume $\lambda=\mu^+$, $\cf(\mu)=\theta$ and $\mu$ is a strong limit
cardinal. Suppose that an algebra ${\Bbb B}_0$ satisfies the $\lambda$-cc and an
algebra ${\Bbb B}_1$ satisfies the $\theta^+$-cc.
Does the free product ${\Bbb B}_0*{\Bbb B}_1$ satisfy the $\lambda$-cc? (is this
consistent? see \ref{moreconclusion}).
\end{problem}
\begin{problem}
Is it consistent that
\begin{quotation}
\noindent each Boolean algebra with the $\aleph_1$--Knaster property has the
$\lambda$--Knaster property for every regular (uncountable) cardinal
$\lambda$?
\end{quotation}
\end{problem}
\section{Some consistency results}
We had seen that without inner models with large cardinals we have a complete
picture, e.g.:
\begin{description}
\item[$(\aleph)$] if $\theta=\cf(\theta)>\aleph_0$, ${\Bbb B}$ is a Boolean algebra
satisfying the $\theta$--cc and $\lambda$ is a regular cardinal such that
\[(\forall\tau<\lambda)(\tau^{{<}\theta}<\theta)\]
{\em then} the algebra ${\Bbb B}$ satisfies the $\lambda$-Knaster condition.
\item[$(\beth)$] if $\theta=\cf(\theta)>\aleph_0$, $\theta<\mu=\mu^{{<}\mu}<
\lambda=\cf(\lambda)<\chi=\chi^{\lambda}$
\noindent{\em then} there is a $\mu^+$-cc $\mu$-complete forcing notion ${\Bbb P}$ of
size $\chi$ such that
\[\forces_{{\Bbb P}}\mbox{``the }\theta\mbox{--cc implies the }\lambda\mbox{--Knaster
property''.}\]
Moreover
\item[$(\beth)^+$] if $\mu=\mu^{<\theta}<\lambda=\cf(\lambda)\leq 2^{\mu}$
then the $\theta$--cc implies the $\lambda$--Knaster property.
\item[$(\gimel)$] if $\theta=\cf(\theta)<\mu$, $\mu$ is a strong limit
singular cardinal, $\cf(\mu)=\theta$
\noindent{\em then} the $\theta^+$--cc {\em does not} imply the
$\mu^+$--Knaster property (and even we have the product example).
\end{description}
In $(\gimel)$, if we allow $(2^\theta)$-cc we may get even better conclusion.
In this section we want to show, under a large cardinals hypothesis, the
consistency of failure.
\begin{proposition}
Assume that $\kappa$ is a supercompact cardinal,
$\kappa<\lambda=\cf(\lambda)$. Let ${\Bbb B}$ be a Boolean algebra which does not
have the $\lambda$--Knaster property. {\em Then}
\[(\exists\theta)(\aleph_0<\theta=\cf(\theta)<\kappa\ \&\ {\Bbb B}\mbox{ does not
have the }\theta\mbox{--Knaster property}).\]
\end{proposition}
\Proof Since $\kappa$ is supercompact, for every second order formula $\psi$:
\begin{quotation}
\noindent if $M\models\psi$
\noindent then for some $N\prec M$, $|N|<\kappa$, $N\models \psi$
\end{quotation}
(see Kanamori and Magidor, \cite{KnMg78}). \QED
\begin{proposition}
\begin{enumerate}
\item If $\aleph_0<\lambda_0<\lambda_1$ are regular cardinals such that
\begin{description}
\item[$(*)_{\lambda_0,\lambda_1}$] for every $x\in {\cal H}(\lambda_1^+)$ there is
$N\prec({\cal H}(\lambda_1^+), {\in})$ such that $x\in N$ and $N\cong
({\cal H}(\lambda_0^+), {\in})$
\end{description}
{\em then} if a Boolean algebra ${\Bbb B}$ has the $\lambda_0$--Knaster property
then it has $\lambda_1$--Knaster property (and ${\Bbb B}\models\lambda_0$--cc
implies ${\Bbb B}\models\lambda_1$--cc).
\item The condition $(*)_{\lambda_0,\lambda_1}$ above holds if for some
$\kappa_0,\kappa_1$, $\kappa_0<\lambda_0$, $\kappa_1<\lambda_1$ we have:
\begin{description}
\item[$(\oplus)$] there is an elementary embedding $j:\bV\longrightarrow M$
with the critical point $\kappa_0$ and such that $j(\kappa_0)=\kappa_1$,
$j(\lambda_0)=\lambda_1$ and $M^{\lambda_1}\subseteq M$.
\end{description}
\item If $\kappa_0$ is a $2$-huge cardinal (or actually less) and e.g.
$\lambda_0=\kappa_0^{+\omega+1}$ then for some
$\lambda_1=\kappa_1^{+\omega+1}$ the condition $(\oplus)$ above holds (we can
assume GCH).
\end{enumerate}
\end{proposition}
\Proof Just check. \QED
\begin{proposition}
\label{consist}
Assume that
\[\bV\models\mbox{`` GCH}+\mbox{ there is 2-huge cardinal}>\theta=\cf(\theta)
\mbox{''}\]
(can think of $\theta=\aleph_0$). Then there is a $\theta$--complete forcing
notion ${\Bbb P}$ such that in $\bV^{{\Bbb P}}$:
\begin{description}
\item[(a)] GCH holds
\item[(b)] if a Boolean algebra ${\Bbb B}$ has the $\theta^+$--Knaster property
then it has the $\theta^{+\theta+1}$--Knaster property
\noindent (note that if $\aleph_\theta>\theta$ then
$\theta^{+\theta+1}=\aleph_{\theta+1}$).
\end{description}
\end{proposition}
\Proof Similar to \cite{LMSh:198}. \QED
\medskip
Chasing arrows what we use is
\begin{proposition}
If $\bV\models$GCH (for simplicity), $\theta=\cf(\theta)=\cf(\mu)<\mu$, a
Boolean algebra ${\Bbb B}$ does not satisfy the $\mu^+$--Knaster condition and
${\Bbb Q}=\Levy(\theta,\mu)$
\noindent then $\bV^{{\Bbb Q}}\models$``${\Bbb B}$ does not have the $\theta^+$--Knaster
property''. \QED
\end{proposition}
\section{More on getting the Knaster property}
Our aim here is to get a ${\rm ZFC}$ result (under reasonable cardinal arithmetic
assumptions) which implies that our looking for $(\kappa,{\rm
not}\lambda)$--Knaster marked Boolean algebras near strong limit singular is
natural. Bellow we discuss the relevant background. The proof relays on ${\rm pcf}\/$
theory (but only by quoting a simply stated theorem) and seems to be a good
example of the applicability of ${\rm pcf}\/$.
\begin{theorem}
\label{th8.1}
Assume $\mu= \mu^{<\beth_\omega}$.
\begin{enumerate}
\item If a Boolean algebra ${\Bbb B}$ of cardinality $\leq 2^\mu$ satisfies the
$\aleph_1$--cc then ${\Bbb B}$ is $\mu$-linked (see below).
\item If ${\Bbb B}$ is a Boolean algebra satisfying the $\aleph_1$--cc then ${\Bbb B}$
has the $\lambda$-Knaster property {\em for every} regular cardinal
$\lambda\in (\mu,2^\mu].$
\end{enumerate}
\end{theorem}
Where
\begin{definition}
\label{def8.2}
\begin{enumerate}
\item A Boolean algebra ${\Bbb B}$ is $\mu$-linked if ${\Bbb B}\setminus\{{\bf 0}\}$ is
the union of $\leq \mu$ sets of pairwise compatible elements.
\item A Boolean algebra ${\Bbb B}$ is $\mu$-centered if ${\Bbb B}\setminus \{{\bf 0}\}$
is the union of $\leq \mu$ filters.
\end{enumerate}
\end{definition}
Of course we can replace the $\aleph_1$--cc, $\beth_\omega$ by the
$\kappa$--cc, $\beth_\omega(\kappa)$ (see more later). The proof is self
contained except relayence on a theorem quoted from \cite{Sh:460}.
Let us review some background. By \cite{Sh:92}, 3.1, if ${\Bbb B}$ is a
$\kappa$-cc Boolean algebra of cardinality $\mu^+$ and $\mu=\mu^{<\kappa}$
then ${\Bbb B}$ is $\mu$-centered. The proof did not work for ${\Bbb B}$ of cardinality
$\mu^{++}$ even if $2^\mu\geq \mu^{++}$ by \cite{Sh:126}, point being we
consider three elements. But if $\mu=\mu^{<\mu} < \lambda^{<\lambda}$, for
some $\mu^+$--cc $\mu$-complete forcing notion ${\Bbb P}$ of cardinality $\lambda$,
in $\bV^{{\Bbb P}}$:
\begin{quotation}
\noindent if ${\Bbb B}$ is a $\mu-{\rm c.c.}$ Boolean algebra of cardinality $<\lambda$
the ${\Bbb B}$ is $\mu$-centered
\end{quotation}
(follows from an appropriate axiom). Juhasz, Hajnal and Szentmiklossy
\cite{HaJuSz} continue this restricting themselves to $\mu$-linked. Then proof
can be carried for $\mu^{++}$, and they continue by induction. However as in
not few cases, the problem was for $\lambda^+$, when $\cf(\lambda)=\aleph_0$
so they assume
\begin{description}
\item{$\otimes$}\ \ \ \ \ if $\lambda\in (\mu, 2^\mu)$,
$\cf(\lambda)=\aleph_0$ then $\lambda=\lambda^{\aleph_0}$ and
$\square_\lambda$
\end{description}
(on the square see Jensen \cite{Jn}). This implies that if we start with
$\bV={\bf L}$ and force, then the assumption ($\otimes$) holds, so it is a
reasonable assumption. Also they prove the consistency of the failure of the
conclusion when $\otimes$ fails relaying on \cite{HJSh:249} (on a set system +
graph constructed there) and on colouring of graphs (section 2 of
\cite{HaJuSz}), possibly $2^{\aleph_0}=\aleph_1$,
$2^{\aleph_1}=\aleph_{\omega+1}$, $|{\Bbb B}|=2^{\aleph_1}$, ${\Bbb B}$ satisfies the
$\aleph_1$--cc but is not $\aleph_1$-linked, only $\aleph_2$-linked.
This gives the impression of essentially closing the issue, and so I
would have certainly thought some years ago, but this is not the case,
examplifying the danger of looking at specific cases. In fact, as we shall
note in the end, their consistency result is best possible under our knowledge
of relevant forcing methods. They use \cite{HJSh:249} to have ``many very
disjoint sets''(i.e. $\langle X_\alpha: \alpha\in S\rangle$, $S\subseteq
\{\delta< \aleph_{\omega+1}: \cf(\delta)=\aleph_1\}$, $X_\alpha \subseteq
\alpha=\sup X_\alpha$, and $\alpha\neq \beta \Rightarrow X_\alpha\cap X_\beta$
finite).
On ${\rm pcf}\/$ see \cite{Sh:g}. Now, \cite{Sh:460} has half jokingly a strong
claim of proving ${\rm GCH}$ under reasonable reinterpretation. In particular
\cite{Sh:460} says there cannot be many strongly almost disjoint quite
large sets, so this blocks reasonable extensions of \cite{HaJuSz}. Now the
main theorem of \cite{Sh:460} enables us to carry the induction on
$\lambda\in (\mu, 2^{\mu}]$ as in \cite{Sh:92}, 3.1, \cite{HaJuSz}, 3.x.
\begin{proposition}
\label{claim8.3}
Suppose that:
\begin{description}
\item[(a)] $\lambda>\theta = \cf(\theta)\geq \kappa =\cf(\kappa)>
\aleph_0$
\item[(b)] there are a club $E$ of $\lambda$ and a sequence $\bar {\P}
=\langle {\P}_\alpha: \alpha\in E\rangle$ (with $\alpha\in E \Rightarrow
|\alpha| \mid \alpha$) such that
\begin{description}
\item[(i)] ${\P}_\alpha \subseteq [\alpha]^{<\kappa}$,
$|{\P}_\alpha|\leq |\alpha|$ and $\bar {\P}$ is increasing continuous,
\item[(ii)] if $X\subseteq \lambda$ has order type $\theta$, then for
some increasing $\langle \gamma_\varepsilon: \varepsilon< \kappa\rangle$
we have $\gamma_\varepsilon \in X$ and for each $\varepsilon< \kappa$
for some $\zeta\in (\varepsilon, \kappa)$ and $\alpha\leq
\min(E\setminus \gamma_\varepsilon)$ we have $\{ \gamma_\zeta:
\zeta<\varepsilon\} \in {\P}_\alpha$,
\end{description}
\item[(c)] ${\Bbb B}$ is a Boolean algebra satisfying the $\kappa$--cc,
$|{\Bbb B}|=\lambda$.
\end{description}
{\em Then} we can find a Boolean algebra ${{\Bbb B}}'$ and a sequence
$\langle{{\Bbb B}}'_\alpha: \alpha\in E\rangle$ of subalgebras of ${\Bbb B}'$ such that
\begin{description}
\item[$(\alpha)$] ${\Bbb B}\subseteq {{\Bbb B}}' \subseteq {{\Bbb B}}^{{\rm com}}$ (the
completion)
\item[$(\beta)$] ${{\Bbb B}}'=\bigcup\limits_{\alpha\in E} {{\Bbb B}}'_\alpha$,
$|{{\Bbb B}}'_\alpha|\leq |\alpha|\neq \aleph_0$, $\langle{\Bbb B}'_\alpha:\alpha\in
E\rangle$ increasing continuous in $\alpha$
\item[$(\gamma)$] if $\alpha\in E$, $x\in {{\Bbb B}}'\setminus \{{\bf 0}\}$
then for some $Y\subseteq {{\Bbb B}}'_\alpha\setminus \{{\bf 0}\}$, $|Y|<\theta$ we
have
\begin{description}
\item[\ ] if $y\in Y$ then $y\cap x=0_{{{\Bbb B}}'}$ and
\item[\ ] if $z\in {{\Bbb B}}'_\alpha$ is such that $z\cap x= 0_{{{\Bbb B}}'}$ then
$z\leq \sup Y'\in {{\Bbb B}}'_\alpha$ for some $Y'\in [Y]^{<\kappa}$
\end{description}
\item[$(\delta)$] if either $(*)_1$ or $(*)_2$ (see below) holds then we can
add
\begin{description}
\item[\ ] $Y$ generates the ideal $\{z\in {{\Bbb B}}'_\alpha: z\cap
x={\bf 0}_{{{\Bbb B}}'}\}$
\end{description}
where
\begin{description}
\item[$(*)_1$] $(\forall \varepsilon<\theta)(|\varepsilon|^{<\kappa}<\theta)$
\item[$(*)_2$] in {\bf (b)} we can add:
if $X\subseteq \alpha$, $|X|<|\alpha|$ then for some $\tau$, $\tau^{<\kappa}
<\theta$ and $h: X \rightarrow \tau$ we have: if $Y\subseteq X$,
$h\restriction Y$ is constant then $Y\in {\P}_\alpha$.
\end{description}
\end{description}
\end{proposition}
\Proof Let $\chi$ be a large enough regular cardinal. Let
${\Bbb B}=\{x_\varepsilon: \varepsilon<\lambda\}$, let ${{\Bbb B}}^{{\rm com}}$ be the
completion of ${\Bbb B}$. We choose by induction on $\alpha\in E$ an elementary
submodel $N_\alpha$ of $(H(\chi), \in, <^*_\chi)$ of cardinality $|\alpha|$,
increasing continuous in $\alpha$, such that ${\Bbb B}$, $\langle x_\varepsilon:
\varepsilon< \lambda\rangle$, ${{\Bbb B}}^{\rm com}$, $\bar {\P}$, $\lambda$, $\theta$,
$\kappa$ belong to $N_0$ and $\langle N_\zeta: \zeta\leq
\varepsilon\rangle \in N_{\varepsilon+1}$.
Note: if $\alpha\in {\rm nacc} (E)$ then $\alpha\in N_\alpha$ hence
${\P}_\alpha\subseteq N_\alpha$.
Let
\[{{\Bbb B}}'_\alpha\stackrel{\rm def}{=} N_\alpha \cap {{\Bbb B}}^{\rm com},\qquad
{{\Bbb B}}'=\bigcup\limits_{\alpha\in E}{{\Bbb B}}'_\alpha.\]
We define by induction on $\alpha\in E$ a one-to-one function $g_\alpha$ from
${{\Bbb B}}'_\alpha$ onto $\alpha$ such that
\[\beta\in \alpha\cap E\ \ \Rightarrow\ \ g_\beta \subseteq g_\alpha,\mbox{
and } g_\alpha\mbox{ is the }<^*_\chi\mbox{-first such }g,\]
so $g_\alpha\in N_{\min(E\setminus (\alpha+1))}$. Let $g=
\bigcup\limits_{\alpha\in E} g_\alpha$. Thus $g$ is a one-to-one function
from ${{\Bbb B}}'$ onto $\lambda$. In the conclusion clauses $(\alpha)$, $(\beta)$
should be clear and let us prove clause $(\gamma)$. So let $\alpha\in E$,
$x\in {{\Bbb B}}'\setminus \{{\bf 0}\}$. We define $J=\{z\in {{\Bbb B}}'_\alpha:
{{\Bbb B}}'\models$ ``$z\cap x = {\bf 0}$''$\}$. Then $J$ is an ideal of
${{\Bbb B}}'_\alpha$. We now try to choose by induction on $\varepsilon<\theta$,
elements $y_\varepsilon\in J$ such that
\begin{description}
\item[(i)] $y_\varepsilon$ is a member of $J\setminus \{{\bf 0}_{{\Bbb B}}\}$
\item[(ii)] there is no $u\in [\varepsilon]^{<\kappa}$ such that $y_\alpha\leq
\sup\limits_{\zeta\in u} y_\zeta \in {{\Bbb B}}'_\alpha$ ($\sup$ - in the
complete Boolean algebra ${{\Bbb B}}^{{\rm com}}$)
\item[(iii)] under {\bf (i)} + {\bf (ii)}, $g(y_\varepsilon)$ ($<\lambda$) is
minimal (hence under {\bf (i)} + {\bf (ii)}, $\beta_\varepsilon\stackrel{\rm
def}{=}\min\{ \beta\leq \alpha: y_\varepsilon \in {{\Bbb B}}'_\beta\}$ is minimal).
\end{description}
If we are stuck for some $\varepsilon<\theta$, then for every $y\in J$ the
condition {\bf (ii)} fails (note that {\bf (iii)} does not change at this
point) i.e. there is a respective set $u$. So suppose $y_\varepsilon$
is defined for $\varepsilon<\theta$. Clearly
\[\zeta<\varepsilon\ \ \Rightarrow\ \ g(y_\zeta)< g(y_\varepsilon)\]
and hence $\zeta<\varepsilon<\theta \Rightarrow \beta_\zeta \leq
\beta_\varepsilon$, and $\zeta<\varepsilon \Rightarrow y_\zeta \neq
y_\varepsilon$. Now apply clause {\bf (b)(ii)} of the assumption to the set
$X=\{\gamma_\varepsilon: \varepsilon< \theta\}$ to get a contradiction. \QED
\begin{proposition}
\label{claim8.4}
Suppose that
\begin{description}
\item[(a)] $\lambda>\theta=\cf(\theta)\geq\kappa=\cf(\kappa)>\aleph_0$, and
$\mu=\mu^{<\theta}\leq \lambda\leq 2^{\mu}$ and
$(\forall \lambda<\theta)[|\lambda|^{<\kappa}<\theta]$,
\item[(b)] as in \ref{claim8.3} and either $(*)_1$ or $(*)_2$ of clause
$(\delta)$ of \ref{claim8.3},
\item[(c)] ${\Bbb B}$ is a $\kappa$-cc Boolean algebra of cardinality $\lambda$,
\item[(d)] every subalgebra ${{\Bbb B}}'\subseteq {{\Bbb B}}^{\rm com}$ of cardinality
$<\lambda$ is $\mu$-linked (see definition \ref{def8.2}(1)).
\end{description}
{\em Then} ${\Bbb B}$ is $\mu$-linked.
\end{proposition}
\Proof Let $\langle {{\Bbb B}}'_\alpha: \alpha \in E\rangle$ was as in the
conclusion of \ref{claim8.3}. Without loss of generality we may assume that
the set of elements ${{\Bbb B}}'_\alpha$ is $\alpha$. Let for $\alpha\in E$,
$h_\alpha: {{\Bbb B}}'_\alpha \setminus \{{\bf 0}\} \rightarrow \mu$ be such that:
\[h_\alpha(x_1) = h_\alpha(x_2)\ \ \Rightarrow\ \ x_1 \cap x_2 \neq
{\bf 0}_{{\Bbb B}}.\]
For each $x\in {{\Bbb B}}'\setminus{\Bbb B}'_{\min(E)}$ let $\alpha(x)=\max\{\alpha\in
E: x\notin {{\Bbb B}}'_\alpha\}$ (well defined as ${{\Bbb B}}'=\bigcup\limits_{\alpha\in
E}{{\Bbb B}}'_\alpha$ and $\langle {{\Bbb B}}'_\alpha: \alpha\in E\rangle$ is increasing
continuous), and let $Y_{x, \alpha}\subseteq {{\Bbb B}}'_\alpha$ be such that
$|Y_{x,\alpha}|<\theta$ and
\[Y_x\subseteq J_x\stackrel{\rm def}{=}\{y\in {{\Bbb B}}'_\alpha: y\cap
x={\bf 0}_{{\Bbb B}}\}\qquad \mbox{ and}\]
$Y_x$ is cofinal in $J_{x}$ ($Y_x$ exists by \ref{claim8.3}, see clause
$(\delta)$).
Let us define $u^0_x=\{ 0, \alpha(x)\}$ and $Y^0_x$ the subalgebra of
${{\Bbb B}}'$ generated by $\{x\}$, and $u^{n+1}_x= u^n_x \cup \{\alpha(y):
y\in Y^n_x\}$ and $Y^{n+1}_x$ be the subalgebra of ${{\Bbb B}}'$ generated by
\[Y^n_x\cup\bigcup\{Y_{x_1, \alpha}: x_1\in Y^n_x \mbox{ and }\alpha\in
u^n_x\}.\]
Finally let $Y^\omega_x= \bigcup\limits_{n<\omega} Y^n_x$. As $\theta$ is
regular, $|Y^n_x|<\theta$ and as in addition $\theta$ is uncountable,
$|Y^\omega_x|<\theta$. Let $u_x=\{\alpha(y): y\in Y^\omega_x\}$. We can find
$A_\zeta\subseteq {{\Bbb B}}'\setminus \{{\bf 0}\}$ for $\zeta< \mu$ such that
${{\Bbb B}}'\setminus \{{\bf 0}\}= \bigcup\limits_{\zeta< \mu} A_\zeta$ and
\begin{description}
\item[$(\tilde{\circledast})$] if $x_1$, $x_2\in A_\zeta$, then there are
one-to-one functions $f:Y^\omega_{x_1}\stackrel{\rm onto}{\longrightarrow}
Y^\omega_{x_2}$ and $g:u_{x_1}\stackrel{\rm onto}{\longrightarrow}u_{x_2}$
such that:
\begin{description}
\item[(i)] $f$, $g$ preserve the order,
\item[(ii)] $f(x_1)=x_2$ and if $y\in Y^\omega_{x_1}$ then
$g(\alpha(y))=\alpha(f(y))$,
\item[(iii)] if $\alpha\in u_{x_1}$, $y\in {{\Bbb B}}'_\alpha\cap
Y^\omega_{x_1}$ then $h_\alpha(x_1)= h_{g(\alpha)}(f(x_1))$
\item[(iv)] $f$ is an isomorphism (of Boolean algebras)
\item[(v)] $g$ is the identity on $u_{x_1}\cap u_{x_2}$
\item[(vii)] $f$ is the identity on $Y^\omega_{x_1}\cap Y^\omega_{x_2}$
\end{description}
\end{description}
(Why? By \cite{EK} or use $\langle\eta_x: x\in {{\Bbb B}}'\rangle$, $\eta_x\in
{}^{\mu}2$ with no repetitions.)
So it is enough to prove:
\[x_1, x_2\in A_\zeta \Rightarrow x_1\cap x_2 \neq 0_{{\Bbb B}}.\]
Let $D_1$ be an ultrafilter of $Y^\omega_{x_1}$ to which $x_1$ belongs,
$D_2=: \{ f(y): y\in Y^\omega_{x_2}\}$ (an ultrafilter on
$Y^\omega_{x_2}$ to which $x_2$ belongs). It suffices to prove that for each
$\alpha\in E$, $D_1\cap {{\Bbb B}}_\alpha$, $D_2\cap {{\Bbb B}}_\alpha$ generate non
trivial filters on ${{\Bbb B}}_\alpha$. We do it by induction on $\alpha$ (note if
$\alpha\leq \beta$ this holds for $\alpha$ provided it holds for $\beta$).
If $\alpha\in u_{x_1}\cap u_{x_2}$ use clause {\bf (iii)} of
$(\tilde{\circledast})$ -- note that this includes the case when $\alpha=0$.
For $\alpha\in {\rm acc}(E)$ it follows by the finiteness of the condition. In the
remaining case $\beta=\sup(E\cap\alpha)<\alpha$ and if $Y^\omega_{x_1}\cap
{{\Bbb B}}'_\alpha\subseteq{{\Bbb B}}'_\beta$, $Y^\omega_{x_2}\cap {{\Bbb B}}'_\alpha
\subseteq {{\Bbb B}}'_\beta$ this is trivial. So by symmetry we may assume that
$\alpha\in u_{x_1}\setminus u_{x_2}$ and use the definition of $Y_y$ for $y\in
B_\alpha\cap Y^\omega_{x_1}\setminus {{\Bbb B}}_\beta$. \QED
\begin{proposition}
\label{claim8.5}
Assume $\mu=\mu^{<\beth_\omega(\kappa)}$. Then for every $\lambda\in (\mu,
2^\mu]$ of cardinality $>\mu$, for every large enough regular $\theta<
\beth_\omega(\kappa)$ clause (b) of \ref{claim8.3} holds.
\end{proposition}
\Proof By \cite{Sh:460}, for every $\tau\in [\mu, \lambda)$ for some
$\theta_\tau< \beth_\omega(\kappa)$, we have:
\begin{description}
\item[$(\tilde{\circleddash})$] there is $\P={\P}_\tau\subseteq [\tau]^{<
\beth_\omega(\kappa)}$ closed under subsets such that $|\P| \leq \tau$
and every $X\in [\tau]^{<\beth_\omega(\kappa)}$ is the union of $<\theta_\tau$
members of members of ${\P}_\tau$.
\end{description}
Now as $\cf(\lambda)>\mu$ for some $n<\omega$, the set
\[\Theta=\{\tau: \mu<\tau<\lambda, \theta_\tau\leq \beth_\omega(\kappa)\}\]
is an unbounded subset of ${\rm Card}\cap (\mu, \lambda)$. Let $\theta\geq
(\beth_{n+1}(\kappa))$ be regular. Now choose $E$ club of $\lambda$ such
that $\alpha\in {\rm nacc}(E) \Rightarrow |\alpha|\in \Theta$ and choose
${\P}_\alpha\subseteq [\alpha]^{<\kappa}$ increasing continuous with
$\alpha\in E$, such that for $\alpha\in {\rm nacc}(E)$, for every $X\in
[\alpha]^\theta$, for some $h: X \longrightarrow \beth_n(\kappa)$, if
$Y\subseteq X$, $|Y|<\kappa$ and $h\restriction Y$ constant then $Y\in
{\P}_\alpha$.
Now suppose $X\subseteq \lambda$, ${\rm otp}(X)=\theta$, so let
$X=\{\gamma_\varepsilon: \varepsilon<\theta\}$, $\beta_\varepsilon$ increasing
with $\varepsilon$; let $\beta_\varepsilon=\min\{\alpha\in E:
\gamma_\varepsilon< \beta\}$, so $\zeta<\varepsilon \Rightarrow
\beta_\zeta\leq \beta_\varepsilon$ and $\beta_\varepsilon \in {\rm nacc}(E)$
and there is $h_\varepsilon: \{\zeta: \zeta<\varepsilon\} \rightarrow
\beth_n(\kappa)$ such that for every $j< \beth_n(\kappa)$,
\[u\in [\varepsilon]^{<\kappa}\ \&\ (h\restriction u\mbox{ constant})\ \
\Rightarrow\ \ \{ \gamma_\zeta: \zeta\in u\}\in {\P}_{\beta_\varepsilon}.\]
Applying the Erd\"os--Rado theorem (i.e. $\theta \rightarrow
(\beth_n(\kappa)^+)^2_{\beth_n(\kappa)}$) we get the desired result (the proof
is an overkill). \QED
\begin{mainconc}
\label{conc8.6}
Suppose that $\kappa$ is a regular uncountable cardinal,
$\mu=\mu^{\beth_\omega(\kappa)}$ and ${\Bbb B}$ is a Boolean algebra satisfying
the $\kappa$-cc.
\begin{enumerate}
\item If $|{{\Bbb B}}|\leq 2^\mu$ then ${\Bbb B}$ is $\mu$-linked
\item If $\lambda$ is regular $\in (\mu, 2^\mu]$ {\em then} ${\Bbb B}$
satisfies the $\lambda$-Knaster condition.
\end{enumerate}
\end{mainconc}
\Proof 1) We prove this by induction on $\lambda=|{\Bbb B}|$. If $|{\Bbb B}|\leq \mu$
this is trivial and if $\cf(|{\Bbb B}|)\leq \mu$ this follows easily by
the induction hypothesis. In other cases by \ref{claim8.5}, for some
$\theta^*< \beth_\omega(\kappa)$ for every regular $\theta\in (\theta^*,
\beth_\omega(\kappa))$, clause (b) of \ref{claim8.3} holds. Choose
$\theta=(\theta^{\aleph_0})^{++}$, so for this $\theta$ both clause (b)
of \ref{claim8.3} and $(*)_1$ of clause $(\delta)$ of \ref{claim8.3}
hold. Thus by claim \ref{claim8.4} we can prove the desired conclusion for
$\lambda=|{\Bbb B}|$.
\medskip
\noindent 2) Follows from part 1). \QED
\begin{proposition}
\label{claim8.7}
\begin{enumerate}
\item In \ref{conc8.6} we can replace the assumption
$\mu=\mu^{\beth_\omega(\kappa)}$ by $\mu=\mu^{<\tau}$ if
\begin{description}
\item[$\otimes$] for every $\lambda\in (\mu, 2^\mu)$ of cardinality
$>\mu$ for some $\theta=\cf(\theta)\geq \kappa$ we have: clause (b) of
\ref{claim8.3} and $(*)_2$ of clause $(\delta)$ of \ref{claim8.3} hold.
\end{description}
\item If $\lambda^*\in (\mu, 2^\lambda)$ and we want to have the
conclusion of \ref{conc8.6}(1) with $|{\Bbb B}|=\lambda^*$ and
\ref{conc8.6}(2) for $\lambda^*$-Knaster only {\em then} it suffice to
restrict ourselves in $\otimes$ to $\lambda\leq \lambda^*$. \QED
\end{enumerate}
\end{proposition}
\begin{proposition}
\label{fact8.8}
In \ref{claim8.3}, if $(\forall \varepsilon<
\theta)[|\varepsilon|^{<\kappa}<\theta]$ then we can weaken clause (ii)
of assumption (b) to
\begin{description}
\item[(ii)$'$] if $X\subseteq \lambda$ has order type $\theta$ then for
some $\langle \gamma_\varepsilon: \varepsilon< \kappa\rangle$ we have:
$\gamma_\varepsilon\in X$ and
\[(\forall \varepsilon<\kappa)(\exists \gamma\in X)(\exists\alpha\leq
\min(E\setminus \gamma))(\{\gamma_\zeta: \zeta< \varepsilon\}\in
{\P}_\alpha).\]
\end{description}
\end{proposition}
\Proof Let $X=\{j_\varepsilon: \varepsilon<\theta\}$ strictly increasing
with $\varepsilon$, and let $\beta_\varepsilon=\min(E\setminus
(j_\varepsilon+1))$, so $\zeta<\varepsilon \Rightarrow \beta_\zeta\leq
\beta_\varepsilon$. Let
\begin{eqnarray}
e\stackrel{\rm def}{=}\{\varepsilon < \theta: &\ & \varepsilon \mbox{ is a
limit ordinal and } \nonumber\\
\ &\ & \mbox{if }\varepsilon_1< \varepsilon\mbox{ and }u\in
[\varepsilon_1]^{<\kappa}\mbox{ and }\{j_\xi: \xi\in u]\in
\bigcup\limits_{\zeta< \theta}{\P}_{\beta_\zeta} \nonumber\\
\ &\ & \mbox{then }\{j_\varepsilon: \varepsilon\in u\}\in
\bigcup\limits_{\zeta< \varepsilon}{\P}_{\beta_\varepsilon}\}. \nonumber
\end{eqnarray}
Now $e$ is a club of $\theta$ as ($\theta$ is regular and) $(\forall
\varepsilon<\theta)[|\varepsilon|^{<\kappa}<\theta]$. So we can apply
clause {\bf (ii)}$'$ to $X'=:\{j_\varepsilon: \varepsilon \in e\}$, and get a
subset $\{\gamma_\varepsilon: \varepsilon< \kappa\}$ as there, it is as
required in clause {\bf (ii)}. \QED
\begin{proposition}
\label{claim8.9}
\begin{enumerate}
\item Assume $\lambda>\theta=\cf(\theta)\geq
\kappa=\cf(\kappa)>\aleph_0$. Then a sufficient condition for clause (b) +
$(\delta)(*)_1$ of claim \ref{claim8.3} is
\begin{description}
\item[$(\tilde{\otimes})$\ \ (a)] $\lambda>\theta=\cf(\theta)$
\item[\qquad\ (b)] for arbitrarily large $\alpha<\lambda$ for some regular
$\tau< \theta$ and $\lambda'< \lambda$, for every ${\frak a}\subseteq {\rm Reg}
\cap |\alpha|\setminus \theta$ for some $\langle {\frak b}_\varepsilon:
\varepsilon< \varepsilon^*<\tau\rangle$ we have ${\frak a}=
\bigcup\limits_{\varepsilon< \varepsilon^*} {\frak b}_\varepsilon$ and
$[{\frak b}_\varepsilon]^{<\kappa} \subseteq J_{\leq \lambda'}[{\frak a}]$
for every $\varepsilon< \varepsilon^*$.
\item[\qquad\ (c)] $(\forall \varepsilon<\theta)[|\varepsilon|^{<\kappa}<
\theta]$ or for every $\lambda'\in [\mu, \lambda]$, $\square_{\{\delta<
\lambda': \cf(\delta)=\theta\}}$
\end{description}
\item Assume $\mu> \theta\geq \kappa=\cf(\kappa)> \aleph_0$, a
sufficient condition for clause (b) of \ref{claim8.3} to hold is:
for every $\lambda\in [\mu, 2^\mu]$ of cofinality $>\mu$ for some $\theta'\leq
\theta$, $\otimes_1$ holds (with $\theta'$ instead $\theta$).
\end{enumerate}
\end{proposition}
\Proof 1) By \cite{Sh:430}, \cite{Sh:420}, 2.6 or \cite{Sh:513}.
2) Follows. \QED
\begin{remark}
{\em
So it is still possible that (assuming CH for simplicity)
\begin{description}
\item[$\otimes$] if $\mu=\mu^{\aleph_1}$, ${\Bbb B}$ is ${\rm c.c.c.}$ Boolean
algebra, $|{\Bbb B}|\leq 2^\mu$ then ${\Bbb B}$ is $\mu$-linked.
\end{description}
On the required assumption see \cite{Sh:410}, Hyp. 6.1(x).
\noindent Note that the assumptions of the form $\lambda\in I[\lambda]$ if
added save us a little on ${\rm pcf}\/$ hyp. (we mention it only in 8.x). But if we
are interested in the $[\kappa-{\rm c.c.} \Rightarrow \lambda$-Knaster], it can
be waived.
}
\end{remark}
\nocite{Ju}
\nocite{M1}
\nocite{M2}
\nocite{MgSh:204}
\nocite{Sh:93}
\nocite{Sh:345a}
\nocite{Sh:355}
\nocite{Sh:430}
\nocite{Sh:481}
\nocite{Sh:572}
\nocite{Sh:g}
\bibliographystyle{literal-unsrt}
|
\section{#1}\setcounter{equation}{0}}
\newcommand{{\cal A}}{{\cal A}}
\newcommand{{\cal D}}{{\cal D}}
\newcommand{{\cal O}}{{\cal O}}
\newcommand{{\bf x}}{{\bf x}}
\newcommand{{\bf \hat{r}}}{{\bf \hat{r}}}
\font\sansf=cmss12
\newcommand{\hbox{\sansf P}}{\hbox{\sansf P}}
\newcommand{\hbox{\sansf T}}{\hbox{\sansf T}}
\newcommand{\fbox{\phantom{${\scriptstyle *}$}}}{\fbox{\phantom{${\scriptstyle *}$}}}
\newcommand{\vartheta}{\vartheta}
\newcommand{\varepsilon}{\varepsilon}
\begin{titlepage}
\renewcommand{\thefootnote}{\fnsymbol{footnote}}
\rightline{BUTP-95/31}
\vspace{0.1in}
\LARGE
\center{Quantum Field Theory on Certain\\ Non-Globally Hyperbolic
Spacetimes}
\Large
\vspace{0.2in}
\center{C.J. Fewster\footnote{E-mail address: <EMAIL>}
and A. Higuchi\footnote{E-mail address: <EMAIL>}}
\vspace{0.2in}
\large
\center{\em Institut f\"{u}r theoretische Physik, Universit\"{a}t Bern \\
Sidlerstrasse 5, CH-3012 Bern, Switzerland}
\vspace{0.2in}
\center{\today}
\vspace{0.2in}
\begin{abstract}
We study real linear scalar field theory on two simple
non-globally hyperbolic spacetimes containing closed timelike curves
within the framework proposed by Kay for algebraic
quantum field theory on non-globally hyperbolic spacetimes. In this
context, a spacetime $(M,g)$ is said to be `F-quantum compatible'
with a field theory if it admits a $*$-algebra of local observables
for that theory which satisfies a locality condition known as
`F-locality'. Kay's proposal is that, in formulating algebraic
quantum field theory on $(M,g)$, F-locality should be imposed as
a necessary condition on the $*$-algebra of observables.
The spacetimes studied are the 2- and 4-dimensional spacelike
cylinders (Mink\-owski space quotiented by a timelike translation).
Kay has shown that the 4-dimensional spacelike cylinder is F-quantum
compatible with massless fields. We prove that it is also F-quantum
compatible with massive fields and prove the F-quantum compatibility
of the 2-dimensional spacelike cylinder with both massive and
massless fields. In each case, F-quantum compatibility is proved
by constructing a suitable F-local algebra.
\vspace{0.3truecm}
{\noindent {\bf PACS Numbers} 04.62.+v}
\end{abstract}
\setcounter{footnote}{0}
\renewcommand{\thefootnote}{\arabic{footnote}}
\end{titlepage}
\sect{Introduction}
The theory of quantum fields in curved spacetime has been developed
for the most part in the context of globally hyperbolic Lorentzian
spacetimes. These spacetimes are distinguished by the fact that they
contain Cauchy surfaces\footnote{Following \cite{Kay}, we define a
Cauchy surface in a Lorentzian spacetime to be a smooth spacelike
surface intersected precisely once by each inextendible causal
(time-like or null) curve contained in the spacetime. A spacetime is
said to be globally hyperbolic if and only if it contains a Cauchy
surface.}, which entails the global existence and uniqueness of
certain fundamental solutions to the Klein-Gordon and Dirac equations
(in particular the advanced-minus-retarded fundamental solution) which
play a key r\^ole in the quantisation of these theories.
There has, however, been recent interest in quantum field theory on
spacetimes containing closed timelike curves (CTCs), which do not
possess Cauchy surfaces and for which the usual methodology must be
modified. There are various reasons for this interest. For example,
because such spacetimes are a `source of tension' between quantum
theory and general relativity, one would expect that they might be
associated with or forbidden by non-trivial quantum gravitational
effects (cf. Hawking's Chronology Protection Conjecture
\cite{Hawk}). Even disregarding back-reaction, the formulation of
quantum field theory in fixed background spacetimes containing CTCs
raises many interesting conceptual problems stemming from the lack of
a global Cauchy surface. As an example, we mention the fact that the
evolution of interacting quantum fields from the far past of an
isolated region of CTCs to its far future turns out to be
non-unitary~\cite{Boul,FPS1,Pol}, which has led to a debate on how
this can be interpreted \cite{Hawk2} or alternatively, how unitarity
may be restored \cite{Arley,FW}.
In the presence of such conceptual problems, there are clearly
advantages in adopting a mathematically
rigorous approach to quantum field theory in curved spacetime,
such as that offered by the algebraic approach, even though this
restricts us (at least for the moment) to the study of linear
(non-self-interacting) field theories. The algebraic approach
has also been
developed largely in the globally hyperbolic case (see~\cite{Kay,KW}
for detailed reviews). Recently, however, Kay~\cite{Kay} and
Yurtsever~\cite{Yurt}
have independently proposed suitable generalisations of this approach
to the
non-globally hyperbolic case. Here, we will work within the framework
developed by Kay. In this proposal, the aim is to construct
a $*$-algebra ${\cal A}(M,g)$ whose elements are interpreted as
polynomials in quantum fields smeared by smooth test functions compactly
supported on a given spacetime $(M,g)$, and such that
${\cal A}(M,g)$ possesses the property of {\em F-locality}. This
condition, which will be reviewed in Section~\ref{sect:Alg},
essentially requires that every point in $M$ should have a globally
hyperbolic neighbourhood $N$ such that the {\em induced algebra}
${\cal A}(M,g;N)$ (i.e., the subalgebra of ${\cal A}(M,g)$ consisting
of polynomials of fields smeared by test functions supported
in $N$) should coincide with the {\em intrinsic algebra}
${\cal A}(N,g|_{N})$, obtained by regarding $(N,g|_{N})$ as a
spacetime in its own right and following the normal procedure
for globally hyperbolic spacetimes with some choice of time
orientation on $N$. More precise definitions will be given in
Section~\ref{sect:Alg}. If a spacetime admits an F-local $*$-algebra
corresponding to a given field theory, the spacetime is said to be
{\em F-quantum compatible} with this field theory. Kay argues
that non-F-quantum compatible spacetimes would not appear as
classical approximations to states in quantum gravity. The proposal
of Yurtsever~\cite{Yurt} differs from that of Kay in that it
constructs an algebra which generally fails to be F-local, except
in the case of `quantum benign' spacetimes~\cite{Yurt}.
Returning to Kay's proposal, it is clearly important to determine
which spacetimes are
F-quantum compatible with given field theories and which are not.
Three results due to Kay \cite{Kay} shed some light on this issue
for both massive and massless real linear scalar field theory.
Firstly, all globally hyperbolic spacetimes are F-quantum compatible
with these theories;
moreover, the usual minimal algebra on these spacetimes
is F-local. Secondly, {\em any} subspacetime (of zero co-dimension)
of a globally hyperbolic spacetime is also F-quantum compatible with
these field theories. Thirdly, non-time orientable spacetimes are
not F-quantum compatible with either field theory.
As well as these results, Kay discussed two examples of spacetimes
with closed time-like curves. The first, 2-dimensional Misner space
-- in which a region of closed time-like curves `develops' -- was
shown to be non-F-quantum compatible with massless scalar field
theory. The second, the 4-dimensional `spacelike cylinder' (i.e.,
Minkowski space quotiented by a fixed time translation) is F-quantum
compatible with massless real linear scalar field theory. However,
because (in both cases) the assumption of zero mass plays an
essential r\^ole in his argument, Kay left the massive case as an
open question.
In the present paper, we resolve this question for the four
dimensional spacelike cylinder, which we show to be F-quantum
compatible with massive real linear scalar field theory. We also
demonstrate the F-quantum compatibility of the 2-dimensional
spacelike cylinder with both massive and massless
fields.\footnote{The massless case was also known to Kay (private
communication).} Our method in each of these cases is essentially to
find a global bi-solution $\widetilde{\Delta}(x_1;x_2)$ to the
appropriate Klein-Gordon equation which plays the r\^ole of the
advanced-minus-retarded fundamental solution, and which agrees with
the Minkowski space fundamental solution for $x_1$ and $x_2$
sufficiently close together. Once this is found, it is easy to
construct an F-local $*$-algebra for the corresponding real linear
scalar field theory. Our construction is not unique: it turns out that there
are many F-local algebras on the spacetimes we consider. As yet, it
is not clear whether or not these different algebras correspond to
different physics. To resolve this issue, one would have to study the
physically `nice' states (say, quasi-free and Hadamard) on these
algebras, as suggested by Kay \cite{Kay}.
\sect{Algebraic Quantum Field Theory and F-locality}
\label{sect:Alg}
We start by briefly reviewing Kay's proposal \cite{Kay} for the algebraic
approach for real quantum field theory for the Klein-Gordon equation
\begin{equation}
(\fbox{\phantom{${\scriptstyle *}$}}_g +m^2)\phi =0
\label{eq:Feq}
\end{equation}
on non-globally hyperbolic, time orientable spacetimes $(M,g)$. This
proposal is developed for the version of algebraic field theory in
which the basic object is a $*$-algebra whose elements are
interpreted as polynomials in smeared quantum fields. (Yurtsever
adopts an alternative viewpoint, in which one considers the Weyl
algebra associated with a suitable symplectic space of classical
solutions -- see, e.g., \cite{KW} -- in his approach to quantisation
on non-globally hyperbolic spacetimes \cite{Yurt}.) In the following,
$C_0^\infty(M)$ denotes the space of smooth, real-valued functions
compactly supported on $M$.
We begin by defining a {\em pre-field algebra} to be a $*$-algebra
with identity, $\leavevmode\hbox{\small1\kern-3.8pt\normalsize1}$, of polynomials over $\relax{\hbox{$\inbar\kern-.3em{\rm C}$}}$ in abstract objects
$\phi(f)$ labelled by $f\inC_0^\infty(M)$ such that the following hold:
\begin{list}{(Q\arabic{enumii})}{\usecounter{enumii}}
\item Hermiticity: $(\phi(f))^*=\phi(f)$ for all $f\inC_0^\infty(M)$
\item Linearity: $\phi(\lambda_1f_1+\lambda_2f_2)=\lambda_1\phi(f_1)+
\lambda_2\phi(f_2)$ for all $\lambda_i\in{\rm I\! R}$, $f_i\inC_0^\infty(M)$
\item Field Equation: $\phi((\fbox{\phantom{${\scriptstyle *}$}}_g+m^2)f)=0$ for all $f\inC_0^\infty(M)$.
\end{list}
The motivation for this definition is the interpretation of $\phi(f)$
as the smearing by $f$ of a Hermitian weak solution $\phi(x)$
to~(\ref{eq:Feq}), i.e.,
\begin{equation}
\hbox{`}\phi(f) = \int_M \phi(x) f(x) d\eta \hbox{'},
\end{equation}
where $d\eta=|\det g_{ab}|^{1/2}d^nx$ is the volume element. We emphasise
the heuristic nature of this interpretation: there is no quantum field
$\phi(x)$ as such underlying this approach.
There is a natural mapping between any two pre-field algebras on
$(M,g)$ which identifies elements corresponding to the same
polynomial in the smeared fields. If this mapping is an isomorphism,
we say that the pre-field algebras are {\em naturally isomorphic}.
In general, a pre-field algebra does not represent a quantised field
theory. If $(M,g)$ is globally hyperbolic, there exists a unique
advanced-minus-retarded fundamental solution $\Delta(x_1;x_2)$
to~(\ref{eq:Feq}) (see, e.g., \cite{Lich}) and quantisation is
effected by supplementing relations (Q1--3) with
\begin{list}{(Q4)}{}
\item Covariant Commutation Relations:
$[\phi(f_1),\phi(f_2)] = i\Delta(f_1,f_2)\leavevmode\hbox{\small1\kern-3.8pt\normalsize1}$ for all $f_i\inC_0^\infty(M)$,
\end{list}
where $\Delta(f_1,f_2)$ is the smeared version of $\Delta(x_1;x_2)$
defined by
\begin{equation}
\Delta(f_1,f_2) = \int_{M\times M} \Delta(x_1;x_2) f(x_1)f(x_2)
d\eta(x_1) d\eta(x_2).
\end{equation}
We will refer to a pre-field algebra satisfying (Q4) as a field
algebra. Such an algebra may be constructed as described
in~\cite{Kay}, by first forming the free $*$-algebra generated (over
$\relax{\hbox{$\inbar\kern-.3em{\rm C}$}}$ and with $\leavevmode\hbox{\small1\kern-3.8pt\normalsize1}$) by the abstract objects $\{\phi(f)\mid
f\inC_0^\infty(M)\}$ and then quotienting by the relations (Q1--4). We will
refer to the algebra ${\cal A}(M,g)$ generated in this manner as the
{\em usual field algebra} on $(M,g)$, when $(M,g)$ is globally hyperbolic.
To be precise, the relations (Q1--4) generate a {\em congruence in
algebra}\footnote{A congruence in algebra is a linear equivalence
relation which respects the algebraic operations. Thus in our case,
if the polynomials $P$ and $Q$ are congruent to $P^\prime$ and
$Q^\prime$ respectively, then the product $PQ$ is congruent to
$P^\prime Q^\prime$.} on the algebra of polynomials in the $\phi(f)$
and their adjoints, such that two polynomials are congruent if one
can be manipulated into the form of the other using (Q1--4) and the
usual rules of algebra. The quotient algebra ${\cal A}(M,g)$ is the
algebra of congruence classes of polynomials. We will abuse notation
by denoting a polynomial and its congruence class in the same way.
For the most part, we will also suppress the mention of congruence
classes.
For general non-globally hyperbolic manifolds $(M,g)$, there is no
advanced-minus-retarded fundamental solution and so one cannot define
a field algebra in the above manner, although there may exist
non-trivial pre-field algebras. Clearly, a criterion is required to
select those pre-field algebras which correspond to some reasonable
notion of quantum field theory. Kay proposed in~\cite{Kay} that this
should be done as follows. Suppose ${\cal A}(M,g)$ is a pre-field algebra
on a time orientable manifold $(M,g)$ and let $N$ be a globally
hyperbolic subspacetime of $M$. There are two $*$-algebras naturally
associated with $N$. Firstly, there is the {\em induced} $*$-algebra
${\cal A}(M,g;N)$, namely, the subalgebra of ${\cal A}(M,g)$ consisting of
polynomials in the $\phi(f_i)$ with all $f_i$ supported in $N$, along
with the identity. ${\cal A}(M,g;N)$ is easily seen to be a pre-field
algebra on $(N,g|_N)$. Secondly, because $(N,g|_N)$ may be regarded
as a globally hyperbolic spacetime in its own right, there is the
{\em intrinsic} $*$-algebra, ${\cal A}(N,g|_N)$ which is the usual field
algebra on $(N,g|_N)$. We require that quantisation in each region
$N$ is carried out with respect to a time orientation induced from a
fixed global time orientation on $(M,g)$.\footnote{In~\cite{Kay}, Kay
initially left open the possibility of the time orientation differing
in different regions $N$. However, he then showed that time
orientability was a necessary condition for F-quantum compatibility
of a manifold: we have essentially incorporated this result into our
discussion.} The induced and intrinsic algebras are generally
different (see~\cite{Kay} for explicit examples) because the
commutator in ${\cal A}(N,g|_N)$ is determined by the
advanced-minus-retarded fundamental solution intrinsic to $(N,g|_N)$
(and therefore depends on the causal structure of $(N,g|_N)$), whilst
that in ${\cal A}(M,g;N)$ is induced from ${\cal A}(M,g)$ and may not respect
the causal structure of $(N,g|_N)$. However, Kay's proposal is that,
as a {\em necessary} condition for ${\cal A}(M,g)$ to be a reasonable
$*$-algebra for quantum field theory on $(M,g)$, these algebras
should agree if $N$ is `sufficiently small'. More precisely, this is
codified by the notion of F-locality \cite{Kay}:
\begin{Def}[The F-locality Condition]
Let $(M,g)$ be a (not necessarily globally hyperbolic) time
orientable spacetime and suppose ${\cal A}(M,g)$ is a pre-field algebra
for real linear scalar field theory on $(M,g)$. ${\cal A}(M,g)$ is said
to be {\em F-local} if each point $p$ of $M$ has a globally
hyperbolic neighbourhood $N$ for which the induced ${\cal A}(M,g;N)$ and
intrinsic ${\cal A}(N,g|_N)$ $*$-algebras are naturally isomorphic. In
addition, $(M,g)$ is said to be {\em F-quantum compatible} if it
admits a pre-field algebra satisfying F-locality.
\end{Def}
In~\cite{Kay}, Kay established various consequences of this
definition, including the result that the usual field algebra on a
globally hyperbolic spacetime is F-local. Thus F-locality provides a
generalisation of the usual quantisation procedure. It is important
to note that, even for F-local algebras, it is not the case that the
induced and intrinsic algebras coincide for {\em any} subspacetime
$N$ -- see \cite{Kay} for discussion.
We now describe the technique we will use to prove the F-quantum
compatibility of the spacelike cylinder spacetimes in
Section~\ref{sect:exam}. This is accomplished by constructing a
global bi-solution $\widetilde{\Delta}(x_1;x_2)$ to the Klein-Gordon
equation which is equal to the intrinsic advanced-minus-retarded
fundamental solution on sufficiently small globally hyperbolic
subspacetimes (i.e., for $x_1$ and $x_2$ sufficiently close
together). We then use $\widetilde{\Delta}$ to define an F-local
$*$-algebra.
More formally, we define:
\begin{Def} \label{Def:Famr}
Let $(M,g)$ be a (not necessarily globally hyperbolic) time orientable
spacetime, and suppose that $\widetilde{\Delta}(x_1;x_2)$ is an
antisymmetric, real-valued
global bi-solution to the homogenous Klein-Gordon equation on $M\times M$.
We call $\widetilde{\Delta}$ an {\em F-local advanced-minus-retarded
fundamental solution} on $(M,g)$ if every point $p\in M$ has
a globally hyperbolic neighbourhood $N$ such that the restriction
of $\widetilde{\Delta}$ to $N\times N$ agrees with the intrinsic
advanced-minus-retarded fundamental solution on $(N,g|_N)$.
\end{Def}
Here, the property of antisymmetry (i.e.,
$\widetilde{\Delta}(x_1;x_2)=- \widetilde{\Delta}(x_2;x_1)$) is
required because we would like to regard $\widetilde{\Delta}$
(suitably smeared) as $-i$ times the commutator of smeared fields.
With this definition, the following is then almost immediate:
\begin{Thm}
If a spacetime $(M,g)$ admits an F-local advanced-minus-retarded
fundamental solution $\widetilde{\Delta}$ to the Klein-Gordon
equation, then it is F-quantum compatible with real linear scalar
field theory. Moreover, the algebra ${\cal A}(M,g)$ defined as the
free $*$-algebra generated by the
$\phi(f)$ for $f\inC_0^\infty(M)$ quotiented by the relations (Q1)-(Q3) and
\begin{list}{(Q$4^\prime$)}{}
\item Modified CCRs:
$[\phi(f_1),\phi(f_2)] = i\widetilde{\Delta}(f_1,f_2)\leavevmode\hbox{\small1\kern-3.8pt\normalsize1}$
for all $f_i\inC_0^\infty(M)$
\end{list}
is F-local, where $\widetilde{\Delta}(f_1,f_2)$ is the smeared
version of $\widetilde{\Delta}(x_1;x_2)$.
\end{Thm}
{\em Proof:} Let $p$ be an arbitrary point in $M$ and let $N$ be
a globally hyperbolic neighbourhood of the type guaranteed by
Definition~\ref{Def:Famr}.
Then the F-locality of $\widetilde{\Delta}$ implies that the natural
mapping between ${\cal A}(N,g|_N)$ and ${\cal A}(M,g;N)$ is an isomorphism,
because the relations (Q1--3) and (Q$4^\prime$) for $(M,g)$ agree with
the relations (Q1--4) for $(N,g|_N)$ on polynomials involving only
test functions supported in $N$.~$\vrule height 1.5ex width 1.2ex depth -.1ex $
Thus it suffices to exhibit an F-local advanced-minus-retarded
fundamental solution on a spacetime $(M,g)$ in order to conclude its
F-quantum compatibility and to construct an F-local algebra there.
In addition to F-locality, we will require that our $*$-algebra ${\cal A}(M,g)$
should be globally covariant with respect to the isometries of the
spacetime (cf.~\cite{Dimock}). That is, we require there to be
a representation $\kappa\mapsto\alpha_\kappa$ of the isometry
group of $(M,g)$ as automorphisms on ${\cal A}(M,g)$ such that
$\alpha_\kappa(\phi(f)) = \phi(f\circ\kappa)$ for each $f\inC_0^\infty(M)$
and isometry $\kappa$.
This places constraints on the F-local advanced-minus-retarded
fundamental solution $\widetilde{\Delta}(x_1;x_2)$, because we have
\begin{equation}
i\widetilde{\Delta}(f\circ\kappa,g\circ\kappa)\leavevmode\hbox{\small1\kern-3.8pt\normalsize1} = \alpha_\kappa(
i\widetilde{\Delta}(f,g)\leavevmode\hbox{\small1\kern-3.8pt\normalsize1})
\end{equation}
for all $f,g\inC_0^\infty(M)$. Thus we require $
\widetilde{\Delta}(\kappa(x_1);\kappa(x_2))= \pm
\widetilde{\Delta}(x_1;x_2)$ for each isometry $\kappa$, with the
sign depending on whether
$\alpha_\kappa$ is a linear ($+$) or anti-linear ($-$) automorphism.
\sect{Quantum Field Theory on Spacelike Cylinders}
\label{sect:exam}
In this section, we discuss the F-quantum compatibility of the 2- and
4-dimensional spacelike cylinders, which are obtained as quotients of
2- or 4-dimensional flat Minkowski space by a fixed time-like
translation. We write coordinates for these
spacetimes in the form $(t,{\bf x})$ and, in the 2-dimensional case,
write the single spatial coordinate as $z$.
The spacelike cylinder spacetime is then defined to be the quotient
of Minkowski space by the fixed time translation $t\rightarrow t+T$
for some $T>0$.
In the following, it will be useful to
have an explicit class of globally hyperbolic neighbourhoods
available: accordingly, we define a {\em diamond of size $d$} about
position $(t,{\bf x})$ to be the region
\begin{equation}
N_d=\{(t^\prime,{\bf x}^\prime)\mid~
|t^\prime-t|+|{\bf x}^\prime-{\bf x}|<d \}.
\end{equation}
In our example spacetimes every point has a globally
hyperbolic diamond neighbourhood of any size less than $T/2$. If a
diamond neighbourhood in one of the spacelike cylinders is
globally hyperbolic, its intrinsic advanced-minus-retarded fundamental
solution is just the restriction of that for 2- or 4-dimensional
Minkowski space as appropriate.
\subsection{The 2-dimensional Spacelike Cylinder}
\label{sect:2dsc}
We now show that the 2-dimensional spacelike cylinder spacetime is
F-quantum compatible with both massless and massive real linear
scalar field theory, by exhibiting suitable F-local
advanced-minus-retarded fundamental solutions.
As mentioned above, we will require global covariance with respect to
the isometries of the spacetime: these are generated by the group of
spacetime
translations and the two discrete symmetries of parity and time
reversal. The translations form a continuous group of isometries and
must therefore be implemented by linear automorphisms at the
algebraic level, whilst parity and time reversal are implemented as
usual by linear and anti-linear automorphisms respectively (locally,
this is actually a consequence of F-locality). In consequence,
$\widetilde{\Delta}(x_1;x_2)$ can be written as a function of
$x_1-x_2$ alone (by translational invariance):
$\widetilde{\Delta}(x_1;x_2) = \widetilde{\Delta}(t_1-t_2,z_1-z_2)$,
where $\widetilde{\Delta}(t,z)$ is even in $z$ and odd in $t$.
We now prove the F-quantum compatibility of the 2-dimensional
spacelike cylinder, treating the massless and massive cases
separately.
{\noindent\bf Massless Klein-Gordon Theory}
It is helpful to begin by recalling the
definition of the advanced-minus-retarded fundamental solution to
the massless Klein-Gordon equation on 2-dimensional Minkowski space.
This is given by $\Delta(t_1,z_1;t_2,z_2)=\Delta(t_1-t_2,z_1-z_2)$,
where $\Delta(t,z)$ solves
\begin{equation}
\left(\frac{\partial^2}{\partial t^2} -
\frac{\partial^2}{\partial z^2} \right) \Delta(t,z) = 0
\label{eq:KG2}
\end{equation}
with initial data
\begin{equation}
\Delta(0,z) = 0, \qquad \left.\frac{\partial\Delta}{\partial t}
(t,z)\right|_{t=0}=-\delta(z).
\end{equation}
As is well known, the solution is
\begin{equation}
\Delta(t,z) = -\frac{1}{2} \varepsilon(t)\vartheta(t^2-z^2),
\label{eq:amrf1}
\end{equation}
where $\vartheta(x)$ is the Heaviside function, and $\varepsilon(x)$ is the sign
of $x$. Thus $\Delta(t,z)$ is equal to $-\frac{1}{2}$ in the future
light-cone of the origin, $\frac{1}{2}$ in the past light-cone and
vanishes elsewhere.
To construct an F-local advanced-minus-retarded fundamental solution
on the spacelike cylinder, we regard it as the strip
$|t|<T/2$ in Minkowski space and impose periodic boundary conditions
on all fields at $\pm T/2$. We seek a solution $\widetilde{\Delta}$
to~(\ref{eq:KG2}) which is periodic in time with period $T$ and
which agrees with $\Delta(t,z)$ for $t,z$ sufficiently close to the
origin. There are many possibilities for $\widetilde{\Delta}$
(see the corresponding discussion for the massive case); we
give the simplest, which is
\begin{equation}
\widetilde{\Delta}(t,z) = \sum_{n=-\infty}^\infty
(-1)^n \Delta(t, z-nT/2).
\end{equation}
The sum converges (because only finitely many summands are non-zero
at any given $(t,z)$) and its values
are displayed in Figure 1.
Note that $\widetilde{\Delta}$ is
periodic in both $t$ and $z$ with period $T$, so it suffices to
show the values for $|t|,|z|<T/2$. $\widetilde{\Delta}$ is clearly
the solution to~(\ref{eq:KG2}) with initial data
\begin{equation}
\widetilde{\Delta}(0,z) = 0,
\qquad \left.\frac{\partial\widetilde{\Delta}}{\partial t}
(t,z)\right|_{t=0}=\sum_{n=-\infty}^\infty (-1)^{n+1}\delta(z-nT/2).
\end{equation}
One may also regard $\widetilde{\Delta}(t,z)$ as the result
of `wrapping' the Minkowski space fundamental solution $\Delta$ round
a spacelike cylinder of period $T/2$ with {\em anti-periodic}
boundary conditions. It is interesting to compare this with Kay's proof
of F-quantum
compatibility for massless fields on the 4-dimensional spacelike
cylinder, which proceeds by wrapping the Minkowski fundamental
solution round the cylinder of period $T$ with periodic boundary
conditions.
Finally, it is easy to see from Fig. 1 that
$\widetilde{\Delta}(t_1,z_1;t_2,z_2)
=
\widetilde{\Delta}(t_1-t_2,z_1-z_2)$ agrees with
$\Delta(t_1,z_1;t_2,z_2)$ within any diamond neighbourhood of size
less than $T/4$ about any point $p$ in the spacetime. (Note that
if two points lie in a diamond region of size $d$ of some point,
then their difference lies in a diamond region of size $2d$ about
the origin.) Thus $\widetilde{\Delta}$ is an F-local
advanced-minus-retarded fundamental solution on the two dimensional
spacelike cylinder.
As an immediate corollary, due to the periodicity of $\widetilde{\Delta}$
in both $t$ and $z$, we also conclude F-quantum compatibility of
the 2-torus with equal periods with massless field theory.
\begin{figure}
\setlength{\unitlength}{0.0075in}%
\center{\begin{picture}(440,378)(120,320)
\thicklines
\put(160,680){\line( 1, 0){320}}
\put(160,360){\line( 1, 0){320}}
\thinlines
\put(160,360){\line( 1, 1){320}}
\put(160,680){\line( 1,-1){320}}
\put(320,360){\line( 1, 1){180}}
\put(320,360){\line(-1, 1){180}}
\put(320,680){\line( 1,-1){180}}
\put(320,680){\line(-1,-1){180}}
\put(160,360){\line(-1, 1){20}}
\put(160,680){\line(-1,-1){20}}
\put(480,360){\line( 1, 1){20}}
\put(480,680){\line( 1,-1){20}}
\put(310,600){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$-\frac{1}{2}$}}}
\put(240,520){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(400,520){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(460,600){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$+\frac{1}{2}$}}}
\put(240,640){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(400,640){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(240,380){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(400,380){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$0$}}}
\put(160,600){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$+\frac{1}{2}$}}}
\put(160,440){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$-\frac{1}{2}$}}}
\put(320,440){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$+\frac{1}{2}$}}}
\put(460,440){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$-\frac{1}{2}$}}}
\put(320,320){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$z=0$}}}
\put(480,320){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$z=T/2$}}}
\put(160,320){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$z=-T/2$}}}
\put(560,520){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$t=0$}}}
\put(560,360){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$t=-T/2$}}}
\put(560,680){\makebox(0,0)[c]{\raisebox{0pt}[0pt][0pt]{$t=T/2$}}}
\end{picture}}
\caption{The values taken by the function $\widetilde{\Delta}(t,z)$ on the
spacelike cylinder.}
\end{figure}
\vspace{0.2truecm}
{\noindent\bf Massive Klein-Gordon Theory}
In this case, we seek a global solution $\widetilde{\Delta}(t,z)$ to
the massive Klein-Gordon equation on the spacelike cylinder, which
agrees with the usual advanced-minus-retarded fundamental solution on
2-dimensional Minkowski space
\begin{equation}
\Delta(t,z) = -\frac{1}{2}\varepsilon(t)\vartheta(t^2-z^2)J_0(m(t^2-z^2)^{1/2})
\end{equation}
for $t$ and $z$ sufficiently close to the origin. Here, $J_0(z)$ is
the Bessel function of order zero. As mentioned above, we will
require $\widetilde{\Delta}(t,z)$ to be even in $z$ and odd in $t$.
The key to constructing such a function is that the initial
value problem {\em is}
well-posed for the Klein-Gordon equation
on the spacelike cylinder, when one specifies initial data
\begin{equation}
f(t) = \phi(t,0) \quad {\rm and}\quad g(t) = \left.
\frac{\partial\phi}{\partial z}(t,z)\right|_{z=0}
\label{eq:IVP}
\end{equation}
on the circle $z=0$, where $f(t)$ and $g(t)$ are periodic with
period $T$. (Essentially, we are just turning the spacelike
cylinder on its end to obtain the globally hyperbolic `timelike
cylinder', with $t$ regarded as a periodically identified
`spatial' coordinate and $z$ regarded as `time'.)
We now set $g(t)$ to be identically zero (reflecting the fact that we
want $\widetilde{\Delta}(t,z)$ to be even in $z$), and require that
$f(t)$ should coincide with
$\Delta(t,0)=-\frac{1}{2}\varepsilon(t)J_0(m|t|)$ for $|t|<T/2-\epsilon$ for
some $\epsilon>0$. Outside this region, we can choose $f(t)$
arbitrarily, subject to the requirements that it be real-valued, odd,
smooth and obey periodic boundary conditions at $t=\pm T/2$. From
elementary properties of hyperbolic equations, it follows that the
solution $\widetilde{\Delta}(t,z)$ to the massive Klein-Gordon
equation with this initial data agrees with $\Delta(t,z)$ in the
diamond of size $T/2-\epsilon$ about the origin. Thus
$\widetilde{\Delta}(x_1;x_2)= \widetilde{\Delta}(t_1-t_2,z_1-z_2)$ is
an F-local advanced-minus-retarded fundamental solution for the
massive Klein-Gordon equation on the spacelike cylinder.
There is clearly a substantial non-uniqueness in our construction,
because $f(t)$ can be chosen freely (amongst smooth, odd functions)
for $T/2-\epsilon<|t|<T/2$ (at $t=T/2$, $f$ must vanish, along
with all its even derivatives). Thus there are very many F-local
algebras on this spacetime. However, in general, the corresponding
$\widetilde{\Delta}(t,z)$ functions are not bounded as
$|z|\rightarrow\infty$ for each $t$, owing to the presence of
exponentially growing modes in the solution to the initial value
problem~(\ref{eq:IVP}).\footnote{This is the penalty for switching
the interpretations of $t$ and $z$: the `mass' becomes imaginary!}
To see this explicitly, note that mode solutions to
the massive Klein-Gordon equation
\begin{equation}
\left(\frac{\partial^2}{\partial t^2} -
\frac{\partial^2}{\partial z^2} +m^2\right) \phi(t,z) = 0
\label{eq:KG2m}
\end{equation}
with periodic time dependence $e^{ 2\pi int/T}$ are
$\phi_{n,\pm}(t,z)\propto e^{2\pi int/T \pm i k_n z}$, where
$k_n$ is determined by
\begin{equation}
k_n = \left( \frac{4\pi^2n^2}{T^2} - m^2 \right)^{1/2}.
\end{equation}
Accordingly, for any $m>0$ there is at least one mode (solution
with imaginary $k_n$) which grows exponentially as $|z|\rightarrow\infty$.
Presumably, one should exclude solutions with unbounded
$\widetilde{\Delta}(t,z)$. One can arrange this by using the freedom
in $f(t)$ to
ensure that it possesses no harmonics corresponding to growing modes.
Defining
\begin{equation}
\ell_n(f) = \int_{-T/2}^{T/2} e^{2\pi i nt/T} f(t) dt
\end{equation}
for $n\in\hbox{\sansf Z\kern-0.4em Z}$, this is equivalent to requiring $\ell_n(f) = 0$
for $|n|\le mT/2\pi$. We note that the $n=0$ harmonic vanishes in any
case because $f(t)$ is odd; thus for $mT<2\pi$ (small cylinders or
light particles) growing modes are automatically absent.
More generally, because $f(t)$ is chosen to be real-valued and odd,
we have
$\ell_{-n}(f) = -\ell_n(f)=\ell_n(f)^*$, and so growing modes are
excluded
provided that $\ell_{n}(f) = 0$ for $1\le n\le mT/2\pi$. This
can be arranged by choosing $f(t)$ to have the form
\begin{equation}
f(t) = -\frac{1}{2}\varepsilon(t)J_0(m|t|)\chi(t) +
\sum_{1\le k \le mT/2\pi} \lambda_k h_k(t),
\end{equation}
with $\lambda_k\in{\rm I\! R}$,
where $\chi(t)$ is smooth, even, compactly supported in
$|t|<T/2-\epsilon/2$, and equal to unity for $|t|<T/2-\epsilon$, and
the $h_k(t)$ are smooth odd functions compactly supported in
$T/2-\epsilon/2<|t|<T/2$ such that $\ell_j(h_k)=0$ for
$0\le j<k$ and $\ell_k(h_k)\not =0$. The functions $\chi$ and $h_k$
are also required to be real-valued.
To construct suitable $h_k$, we define the function spaces
\begin{equation}
{\cal D}_n = \left\{ \varphi\in{\cal D}_0 \mid \ell_j(\varphi)
=0~{\rm for}~j=1,2,\ldots,k\right\}
\end{equation}
where ${\cal D}_0$ is the space of smooth odd real-valued functions
compactly supported
in $T/2-\epsilon/2<|t|<T/2$. For each $k$, $h_k$ is chosen to be a
representative of any non-zero equivalence class in the quotient
${\cal D}_{k-1}/{\cal D}_k$, which is non-trivial because $\ell_1,\ldots,\ell_k$
are linearly independent distributions on ${\cal D}_0$. The $h_k$ clearly
possess the required properties and the $\lambda_k$ may therefore
be chosen to ensure that $\ell_n(f)=0$ for each $|n|\le mT/2\pi$ as
required. Reality of the $\lambda_k$ follows from the fact that
$\ell_n(\varphi)$ is purely imaginary for odd real-valued $\varphi$.
One could, of course, specify a distinguished $\widetilde{\Delta}$
by letting $\epsilon$ shrink to zero, thereby putting all the
ambiguity in $f(t)$ into a distributional contribution supported at
the point $t=\pm T/2$. The distinguished $\widetilde{\Delta}$ would
be obtained by requiring this distribution to be as `regular' as
possible, but nonetheless singular enough to ensure the absence of
growing modes. In fact, the F-local advanced-minus-retarded
solution identified in the massless
case (where there are in fact no growing modes) can be regarded as a
(trivial) example of this procedure.
\subsection{The 4-dimensional Spacelike Cylinder}
As we have mentioned, it has already been shown by Kay \cite{Kay}
that the 4-dimensional spacelike cylinder is F-quantum compatible
with massless scalar fields. Here, we demonstrate that the same is
true for the massive case.
The isometries of the 4-dimensional spacelike cylinder are generated
by the group of spacetime translations, the group of rotations, and
the discrete symmetries of parity and time reversal. Of these
generators, only time reversal is represented by an anti-linear
automorphism. The requirement of covariance with respect to these
isometries therefore entails that $\widetilde{\Delta}(x_1;x_2)=
\widetilde{\Delta}(t_1-t_2,|{\bf x}_1-{\bf x}_2|)$, where
$\widetilde{\Delta}(t,r)$ is odd in $t$.
In addition, we require that $\widetilde{\Delta}(x_1;x_2)$ should
agree with the advanced-minus-retarded fundamental solution for
4-dimensional Minkowski space given by $\Delta(x_1;x_2) =
\Delta(t_1-t_2,|{\bf x}_1-{\bf x}_2|)$, where
\begin{equation}
\Delta(t,r) =m\varepsilon(t)\vartheta(t^2-r^2)
\frac{J_1(m(t^2-r^2)^{1/2})}{4\pi(t^2-r^2)^{1/2}} -
\frac{\varepsilon(t)}{2\pi}
\delta(t^2-r^{2}).
\end{equation}
The essential point here is that spherical symmetry has reduced
this to an effectively 2-dimensional problem. Writing
$\widetilde{\Delta}(t,r)=u(t,r)/r$,
$u(t,r)$ satisfies
\begin{equation}
\left(\frac{\partial^2}{\partial t^2} -
\frac{\partial^2}{\partial r^2} +m^2\right) u(t,r) = 0,
\label{eq:KG4m}
\end{equation}
and we may proceed much as in Section~\ref{sect:2dsc}, by specifying
initial data on the circle $r=0$, defined by
\begin{equation}
f(t) = u(t,0) \quad{\rm and}\quad
g(t) = \left.
\frac{\partial u}{\partial r}(t,r)\right|_{r=0} .
\end{equation}
We impose data with $f(t)$ identically zero and $g(t)$ given by
\begin{equation}
g(t) = m\varepsilon(t)\frac{J_1(mt)}{4\pi t} + \frac{\delta^\prime(t)}{2\pi}
\quad {\rm for}\quad |t|<T/2-\epsilon
\end{equation}
for some $\epsilon>0$. For $t$ outside this region, $g(t)$ is chosen
to be smooth, odd and to obey periodic boundary conditions at
$t=\pm T/2$. We also use some of this freedom to ensure the absence of
tachyonic modes in the same manner as before.
These data agree with the corresponding quantities obtained from
the flat spacetime fundamental solution $\Delta(t,r)$ for
$|t|<T/2-\epsilon$ as may be seen by a simple distributional
calculation, which rests on the following easily established result:
Let $\psi_r(t)=r\varepsilon(t)\delta(t^2-r^2)$
be a family of distributions labelled by $r$. Then
\begin{list}{(\alph{enumii})}{\usecounter{enumii}}
\item $\psi_r(t) = \frac{1}{2}(\delta(t-r)-\delta(t+r))
\rightarrow 0$ as $r\rightarrow 0^+$
\item $(\partial \psi_r/\partial r)(t) =
-\frac{1}{2}(\delta^\prime(t-r)+\delta^\prime(t+r)) \rightarrow -
\delta^\prime(t)$ as $r\rightarrow 0^+$.
\end{list}
Thus, the solution $\widetilde{\Delta}(t,r)$ obtained from this
initial data agrees with $\Delta(t,r)$ for $|t|+r<T/2-\epsilon$. Thus
$\widetilde{\Delta}(x_1;x_2)=
\widetilde{\Delta}(t_1-t_2,|{\bf x}_1-{\bf x}_2|)$
agrees with $\Delta(x_1;x_2)$ when $x_1,x_2$ lie in a diamond
neighbourhood of size $T/4-\epsilon/2$ of any given point, so
$\widetilde{\Delta}$ is an F-local advanced-minus-retarded
fundamental solution to the massive Klein-Gordon equation on the
4-dimensional spacelike cylinder. This spacetime is therefore
F-quantum compatible with massive real linear scalar field theory.
As in the 2-dimensional case, our construction is highly non-unique:
similar comments to those made in that case also apply here.
We also note that by setting $m=0$ in the above, we obtain a new
proof of the F-quantum compatibility of this spacetime with
massless field theory.
\section{Conclusion}
We have studied the 2- and 4-dimensional spacelike cylinders and have
shown that the 2-dimensional cylinder is F-quantum compatible with
both massive and massless fields, and that the 4-dimensional cylinder
is F-quantum compatible with massive fields. This complements the
earlier result of Kay on the F-quantum compatibility of the
4-dimensional spacelike cylinder with massless fields and resolves
the question raised in~\cite{Kay} as to whether or not this
spacetime is F-quantum compatible with massive fields. These results
help to strengthen the claim that F-locality provides a good basis
for quantum field theory on non-globally hyperbolic spacetimes.
Our construction exhibits a considerable degree of non-uniqueness. A
similar phenomenon was noted by Kay for spacetimes which can be
isometrically embedded in a globally hyperbolic spacetime: here we
have shown that this can happen even for spacetimes which cannot be
embedded in this way. Non-uniqueness might present a serious problem
if it transpired that different F-local algebras could correspond to
different physics. To investigate this, it would be necessary
to move beyond the construction of the algebra of observables, to the
study of physically reasonable states on the algebra. If necessary,
it might be possible to remove this non-uniqueness by imposing further
conditions on the algebra of observables, or on the class of allowed
states, as mentioned in~\cite{Kay}.
Finally, our treatment of the spacelike cylinders relied heavily on
the special property that they may be turned on end to yield globally
hyperbolic spacetimes. More generally, it is of interest to determine
what conditions a spacetime must satisfy in order to admit an F-local
advanced-minus-retarded fundamental solution. This appears to be
closely related to whether or not a spacetime is `classically
benign'~\cite{Yurt2} with respect to weak solutions of the
Klein-Gordon equation. We note that (almost by definition) a
translationally invariant classically benign spacetime admits an
F-local advanced-minus-retarded fundamental solution, and is
therefore F-quantum compatible. More generally, one needs to
generalise the notion of classical benignity to treat bi-solutions.
Progress in the classification of such manifolds would be of
considerable interest in the context of F-locality.
{\noindent\bf Acknowledgments:} We are grateful to Bernard Kay, Petr
H\'aj\'{\i}{\v c}ek and Peter Minkowski for useful conversations and
comments. CJF thanks the Royal Society and we both thank the
Schweizerischer Nationalfonds for financial support.
\newpage
|
\section{ Introduction }
\label{sec:1}
The occurrence of spontaneous chiral symmetry breaking in strongly
coupled QED is one of the most challenging issues of
non-perturbative quantum field theory nowadays. In his pioneering
work, Miranski reported a regime of cutoff independence
(CI-regime)\footnote{We do not use the notion of a ``scaling limit''
in order to avoid confusion with the term in solid state physics,
where it corresponds to the case of exact scaling (zero masses).}
corresponding to a phase of
QED with spontaneous broken chiral symmetry~\cite{mir85}. His results
are based on a study of the Dyson-Schwinger equation for the electron
self-energy, where electron loop corrections to the photon propagator
are neglected (quenched approximation). In the quenched approximation,
no renormalization of the electric charge is required, and
spontaneous symmetry breakdown occurs, if the bare charge exceeds
a critical value~\cite{mir85,bar86,co89}. Subsequently, it was
argued by Bardeen, Leung and Love that a Nambu-Jona-Lasinio type
interaction~\cite{nam61}, must be included in order to render
the renormalization of the Dyson-Schwinger equation consistent
with the known renormalization properties of the theory~\cite{bar86}.
The results of the quenched
approximation are caught into question, since vacuum polarization
effects might enforce the renormalized coupling to vanish in the
the infinite cutoff limit~\cite{landau} (triviality). First
investigations of this
effect were done by Kondo by parametrizing the effect of the vacuum
polarization~\cite{ko89}. Further insight in the issue of triviality
was gained by the important work of Rakow~\cite{rakow}. He numerically
solved the coupled set of (bare) Dyson-Schwinger equations for the
electron self-energy and the photon propagator, and therefore included
vacuum polarization effects self-consistently. He finds a second
order chiral phase transition and zero renormalized charge at the
critical point~\cite{rakow}. This seems to rule out the quenched
approximation, since the quenched approximation predicts an
interacting infinite cutoff limit.
Parallel to the studies of the truncated Dyson-Schwinger equations,
extensive lattice investigations were performed in order to
clarify the triviality problem of QED~\cite{latt94,illinois,desy}.
Lattice simulations are not bounded by any approximation, but
suffer from a small correlation length compared with the intrinsic
fermionic energy scale due to the finite lattice sizes. In order to
overcome the difficulty of a small correlation length, one fits
the lattice data to two Ans\"atze of the equation of state
(the bare electron mass as function of the fermionic condensate).
One of these Ans\"atze favors an interacting infinite cutoff
limit~\cite{illinois},
whereas the other Ansatz is compatible with triviality~\cite{desy}.
The present status is that the lattice data are not conclusive enough in
order to distinguish between the two equations of state~\cite{latt94}.
In this paper, we further develop the results of~\cite{rakow} and study the
coupled set of renormalized Dyson-Schwinger equations for the electron
self-energy and the photon propagator. The tower of Dyson-Schwinger
equations is truncated by using the tree level electron-photon-vertex
as Ansatz for the renormalized three point function. In order to introduce
our notation and to make the renormalization procedure transparent, we
first derive the renormalized Dyson-Schwinger equations (section \ref{sec:2}).
All Ans\"atze are made for renormalized quantities.
This will later turn out to be crucial in order to compare the full
results with those of the quenched approximation.
We first concentrate on the case of a massive electron (section \ref{sec:3}).
A cutoff independence is found for sufficiently small values of the cutoff.
The theory is called to be {\it weakly renormalizable}.
In contrast, the quenched approximation always predicts a CI-regime
with no upper limit on the cutoff. Although the quenched approximation
is incapable to predict the correct CI-behavior, it provides
good results for physical quantities in the CI-regime (subsection
\ref{sec:3.2}). We then focus onto the chiral limit (section \ref{sec:4}).
A logarithmic cutoff dependence is seen as well in the electron
self-energy as in the vacuum polarization.
The problem of the existence of a CI-regime characterized
by a spontaneous breakdown of chiral symmetry is still unsolved.
\section{ Renormalized Dyson-Schwinger equations }
\label{sec:2}
The generating functional for {\it bare} Green's functions of QED
in Euclidean space is given by the functional integral
\begin{eqnarray}
W[j_\mu ^B, \eta ^B, \bar{\eta }^B ](e_B,m_B) &=& \int
{\cal D} A_\mu \; {\cal D} q \; {\cal D} \bar{q} \;
\exp \biggl\{ - \int d^4 x \; \Bigl[ L_0(x)
\label{eq:1} \\
&-& \bar{\eta }^B(x) q(x) \, - \, \bar{q}(x) \eta ^B (x) \, - \,
j_\mu ^B (x) A_\mu (x) \Bigr] \biggr\} \; ,
\nonumber
\end{eqnarray}
\begin{equation}
L_0 (x) \;=\; \frac{1}{4e_B^2} F_{\mu \nu }[A](x) F_{\mu \nu }[A](x) \, + \,
\bar{q}(x) ( i\partial\kern-.5em\slash +im_B )q(x) \, + \, \bar{q}(x) A\kern-.5em\slash (x) q(x) \; ,
\label{eq:2}
\end{equation}
where $F_{\mu \nu }[A](x) = \partial _\mu A_\nu (x)- \partial _\nu
A_\mu (x)$ is the field strength tensor, and $e_B$, $m_B$ represent the
bare electric charge and the bare electron mass respectively.
At zero external sources, the generating functional $W[0,0,0](m^B, e^B)$
is invariant under U(1) gauge transformations of the integration
variables, i.e.
\begin{equation}
q(x) \rightarrow \exp \{ i \alpha (x) \} \, q(x) \; , \hbox to 1 true cm {\hfill }
A_\mu (x) \rightarrow A_\mu (x) \, + \, \partial _\mu \alpha (x) \; .
\label{eq:3}
\end{equation}
Bare Green's functions are divergent implying the theory must be
regularized and renormalized in order to make sense.
Renormalization is performed by absorbing the divergences into
the renormalization constants $Z_{2,3,e,m}$, which relate
the bare sources and the bare parameters to the renormalized ones, i.e.
\begin{eqnarray}
\eta ^B(x) &=& Z_2^{-1/2} \, \eta (x) \; , \; \; \;
\bar{\eta }^B(x) \; = \; Z_2^{-1/2} \, \bar{\eta }(x) \; , \; \; \;
j_\mu ^B(x) \; = \; Z_3^{-1/2} \, Z_e^{-1} \, j_\mu (x) \; ,
\label{eq:4} \\
e_B &=& Z_e \, e_R \; , \; \; \; m_B \; = \; Z_m \, m_R \; .
\nonumber
\end{eqnarray}
The generating functional for {\it renormalized } Green's functions
is obtained from $W[j_\mu ^B, \eta ^B, \bar{\eta }^B ](e_B,m_B)$ by
replacing the bare sources and bare parameters by the
renormalized sources and the renormalized parameters respectively, i.e.
\begin{equation}
W_R[j_\mu , \eta , \bar{\eta } ](e_R,m_R) \; = \;
W[Z_3^{-1/2} Z_e^{-1} j_\mu , Z_2^{-1/2} \eta , Z_2^{-1/2} \bar{\eta }]
(Z_e e_R, Z_m m_R) \; .
\label{eq:5}
\end{equation}
The renormalized Green's functions are obtained by taking functional
derivatives of $W_R$ (\ref{eq:5}) with respect to the corresponding
external sources. For example, the renormalized electron propagator
$S_R$ and the renormalized photon propagator $D_{\mu \nu }^R$ are
given by
\begin{equation}
S_R(x,y) \; = \; \frac{ \delta \ln W_R[j_\mu , \eta , \bar{\eta } ] }{
\delta \eta (x) \, \delta \bar{\eta }(y) }
\vert _{j, \eta , \bar{\eta } = 0 } \; , \; \; \;
D_{\mu \nu }^R (x,y) \; = \; \frac{ \delta \ln W_R[j_\mu , \eta ,
\bar{\eta } ] }{ \delta j_\mu (x) \, \delta j_\nu (y) }
\vert _{j, \eta , \bar{\eta } = 0 } \; .
\label{eq:5a}
\end{equation}
After a change of the integration variables, the generating functional
(\ref{eq:5}) can be cast into
\begin{eqnarray}
W_R[j_\mu , \eta , \bar{\eta } ](e_R,m_R) &=& \int
{\cal D} A_\mu \; {\cal D} q \; {\cal D} \bar{q} \;
\exp \biggl\{ - \int d^4 x \; \Bigl[ L(x)
\label{eq:6} \\
&-& \bar{\eta }(x) q(x) \, - \, \bar{q}(x) \eta (x) \, - \,
j_\mu (x) A_\mu (x) \Bigr] \biggr\} \; ,
\nonumber
\end{eqnarray}
\begin{eqnarray}
L (x) &=& \frac{Z_3}{4e_R^2} \, F_{\mu \nu }[A](x) F_{\mu \nu }[A](x)
\, + \, Z_2 \, \bar{q}(x) i\partial\kern-.5em\slash q(x) \, + \, Z_0 \, im_R \bar{q}(x) q(x)
\label{eq:7} \\
& + & Z_1 \, \bar{q}(x) A\kern-.5em\slash (x) q(x) \; ,
\end{eqnarray}
where we have introduced
\begin{equation}
Z_1 \; = \; Z_2 \, Z_3^{1/2} \, Z_e \; , \hbox to 1 true cm {\hfill }
Z_0 \; = \; Z_2 \, Z_m \; .
\label{eq:8}
\end{equation}
The generating functional $\Gamma $ for renormalized
one-particle irreducible functions (in the following vertex functions)
is obtained from $W_R$ (\ref{eq:5}) by a Legendre transformation, i.e.
\begin{equation}
\Gamma [{\cal A}_\mu , \psi , \bar{\psi }] \; = \;
\ln W_{R}[j_\mu , \eta, \bar{\eta }] \; - \; \int d^4x \; \left[
j_\mu (x) {\cal A}_\mu (x) + \bar{\eta }(x) \psi (x) +
\bar{\psi }(x) \eta (x) \right] \; ,
\label{eq:9}
\end{equation}
where the external sources are implicitly related to the fields
${\cal A}$, $\psi $, $\bar{\psi }$ by
\begin{equation}
{\cal A}_\mu (x) = \frac{ \delta \ln W_R [j_\mu ,\eta , \bar{\eta } ] }{
\delta j_\mu (x) } \; , \; \;
\psi (x) = \frac{ \delta \ln W_R [j_\mu ,\eta , \bar{\eta } ] }{
\delta \bar{\eta } (x) } \; , \; \;
\bar{\psi } (x) = \frac{ \delta \ln W_R [j_\mu ,\eta , \bar{\eta } ] }{
\delta \eta (x) } \; .
\label{eq:10}
\end{equation}
Below we will use the renormalized electron photon vertex, which is
defined by
\begin{equation}
\Lambda _\mu (x,y,z) \; = \; \frac{ \delta \Gamma [{\cal A}_\mu , \psi ,
\bar{\psi }] }{ \delta {\cal A}_\mu (x) \, \delta \psi (y) \,
\delta \bar{\psi }(z) } \vert _{{\cal A}_\mu , \psi , \bar{\psi } = 0} \; .
\label{eq:11}
\end{equation}
Exploiting the fact that the functional integral $W_R [j_\mu ,\eta ,
\bar{\eta } ]$ (\ref{eq:6}) is not changed by a shift of the integration
variables $A_\mu $, $q$, $\bar{q} $ generates the {\it renormalized }
Dyson-Schwinger equations~\cite{itz80}. In particular, one finds
\begin{eqnarray}
\left( S_R(x,y) \right) ^{-1} &=& \left[ Z_2 \, i \partial\kern-.5em\slash
+ Z_0 \, im_R \right] \, \delta (x-y)
\label{eq:12} \\
&-& Z_1 \int d^4z \; d^4w \; \gamma _\mu D_{\mu \nu }^R (x,z) S_R(x,w)
\Lambda _\nu (z,w,y) \; ,
\nonumber \\
\delta _{\mu \nu } \delta (x-y) &=& \frac{Z_3}{e_R^2}
\left( - \partial ^2 \delta _{\mu \alpha } + \partial _\mu
\partial _\alpha \right) D_{\alpha \nu }^R (x,y)
\label{eq:13} \\
&+& Z_1 \int d^4z \; d^4w \; d^4v \; \hbox{tr} \left\{ \gamma _\mu S_R(x,z)
\Lambda _\alpha (v,z,w) S_R(w,x) \right\} D_{\alpha \nu }^R (v,y) \, .
\nonumber
\end{eqnarray}
Throughout this paper, we will truncate the tower of Dyson-Schwinger
equations by making an Ansatz for the renormalized vertex function
$\Lambda _\nu $. In this case, the renormalized electron propagator and
the renormalized photon propagator can be obtained by solving
the coupled system (\ref{eq:12},\ref{eq:13}).
In order to investigate the consistency of the Ansatz, one studies
the Ward-Takahashi identities~\cite{itz80}, which provide
relations among Green's functions induced by gauge invariance.
One first realizes that $Z_1=Z_2$ must hold in the renormalized
Lagrangian (\ref{eq:7}) in order to preserve the invariance under the
transformation (\ref{eq:3}).
Combining (\ref{eq:8}) with (\ref{eq:4}), one finds that
the electric charge is renormalized by the photon wave function
renormalization constant, i.e. $e_R= Z_3 ^{1/2} e_B$. If the bare
charge $e_B$ acquires a finite value in the limit of large cutoff
(as suggested by the quenched approximation), and if
further this limit enforces $Z_3$ to vanish, then the theory
is only consistent with a zero renormalized electric charge. This scenario
is referred to as triviality of QED in the literature.
Exploring the fact that the generating
functional $W_R$ (\ref{eq:6}) is invariant under small gauge
rotations (\ref{eq:3}), one obtains
\begin{equation}
\partial _\mu ^{(x)} \, \Lambda _\mu (x,y,z) \; = \;
i S_R^{-1}(y,x) \, \delta (x-z) \; - \; i S_R^{-1}(x,z) \, \delta (x-y) \; .
\label{eq:14}
\end{equation}
Once the coupled system (\ref{eq:12},\ref{eq:13}) was solved for
a particular choice of the renormalized vertex $\Lambda _\mu $,
one inserts the solution for $S_R$ into (\ref{eq:14}) in order to
check the accuracy of the Ansatz for the vertex function $\Lambda _\mu $.
\medskip
In the following, we will work in Landau gauge and set $Z_1=Z_2$ as
imposed by gauge invariance. We will study the Ansatz
\begin{equation}
\Lambda _\mu (z,x,y) \; = \; \gamma _\mu \; \delta (z-x) \;
\delta (z-y) \;
\label{eq:15}
\end{equation}
for the renormalized vertex function, which is the tree level
electron-photon vertex. Recently, the full one-loop QED vertex was
obtained~\cite{kiz95}. The results might provide the structure
for a more general ansatz than (\ref{eq:15}).
Note that the choice of the renormalized
vertex is the only freedom we have. There is no further possibility to
argue in favor of a four fermion interaction first introduced in~\cite{bar86}.
Note also that a four fermion interaction arises and can be naturally
incorporated in renormalized Dyson-Schwinger equations in the dual
formulation of QED~\cite{sugamoto}.
The coupled
Dyson-Schwinger equations are then solved by parametrizing the
renormalized electron propagator and the renormalized photon
propagator in momentum space by
\begin{equation}
\widetilde{S}_R(p) \; = \; \frac{1}{ F(p^2) p\kern-.5em\slash \, + \, i
\Sigma (p^2) } \; , \hbox to 1 true cm {\hfill }
\widetilde{D}^{R}_{\mu \nu }(p) \; = \; \frac{ 4 \pi D_R(p^2) }{ p^2 } \;
\left( \delta _{\mu \nu } \, - \, \frac{ p_\mu p_\nu }{p^2} \right) \, .
\label{eq:16}
\end{equation}
Up to the occurrence of the renormalization constants $Z_{1,3}$,
the derivation yields the same results as reported in~\cite{rakow}.
The first Dyson-Schwinger equation is straightforwardly reduced to
two equations to determine $F(p^2)$ and $\Sigma (p^2) $. Using a
sharp O(4)-invariant cutoff $\Lambda $ to regularize the momentum
integration, these equations are
\begin{eqnarray}
\Sigma (p^2) &=& m_0 \, + \, Z_1 \frac{3}{ 2 \pi ^2 } \int _0^{\Lambda ^2 }
dk^2 \; k^2 \; \int _0^\pi d\theta \; \sin ^2 \theta \;
\frac{ \Sigma (k^2) }{ s(k^2) } \, \frac{ D_R(q^2) }{ q^2 } \; ,
\label{eq:17} \\
F(p^2) &=& Z_1 \, \Biggl[ 1 \; + \; \frac{1}{\pi ^2 p }
\label{eq:18} \\
&& \times \int _0^{\Lambda } dk \; k^4 \;
\int _0^\pi d\theta \; \sin ^2 \theta \; \frac{ F (k^2) }{ s(k^2) }
\frac{ D_R(q^2) }{q^4} \left\{ 3 q^2 \cos \theta - 2 k p \sin ^2 \theta
\right\} \, \Biggr] \; ,
\nonumber
\end{eqnarray}
where we have introduced $m_0 = Z_m m_R $, $s(k^2) =
F(k^2) k^2 + \Sigma (k^2) $ and $q^2 = k^2 + p^2 -2 k p \cos \theta $.
The equation for the photon propagator (\ref{eq:13}) needs more thoughts.
The vacuum polarization $\Pi _{\mu \nu }(p)$ due to the electron loop
is transverse, i.e. $p_\mu \, \Pi _{\mu \nu }(p) =0$, if the
regularization prescription does not violate gauge invariance.
Unfortunately, the sharp O(4)-invariant cutoff prescription spoils gauge
invariance implying that the vacuum polarization $\Pi _{\mu \nu }(p)$
acquires spurious terms proportional to $\delta _{\mu \nu }$, which
are, in addition, quadratic divergent. On the other hand, regularization
schemes consistent with gauge invariance are very time consuming when
solving the coupled equations (\ref{eq:12},\ref{eq:13}) numerically.
In order to circumvent this problem, one starts calculating
$\Pi _{\mu \nu }(p)$ using a regularization scheme which respects gauge
invariance (e.g.~Schwinger proper time, Pauli-Villars regularization).
The result for $\Pi _{\mu \nu }(p)$ is the transverse projector times a
function depending on $p^2$. In this function, one first introduces a second
regularization which corresponds to the O(4)-invariant cutoff scheme, and
then removes the regulator of the scheme which is compatible with gauge
invariance. The final result allows a fast numerical treatment
and is consistent with the transversality of the vacuum polarization
$\Pi _{\mu \nu }(p) $. In the limit of a large cutoff,
all regularization schemes
which respect the symmetries of the model are supposed to yield the
same result for physical quantities.
In agreement with the result stated in~\cite{rakow},
this lengthy procedure yields
\begin{eqnarray}
\frac{1}{D_R(p^2)} &=& \frac{Z_3}{\alpha _R} \; - \; Z_1 \frac{2}{3 \pi ^2}
\label{eq:19} \\
&& \times \int _0 ^{\Lambda ^2 } dk^2 \; k^2 \, \int _0^\pi d\theta \;
\sin ^2 \theta \; \frac{ F(q_+^2) \, F(q_-^2) }{
s(q_+^2) \, s(q_-^2) } \, \left\{ \frac{ k^2}{p^2} ( 8 \cos ^2 \theta
-2 ) - 3/2 \right\} \; ,
\nonumber
\end{eqnarray}
where $\alpha _R = e_R^2/4\pi $ and $q_{\pm} = (k \pm p/2)^2$.
\section{ QED with massive electrons }
\label{sec:3}
In order to calculate the renormalization constants $Z_{1,3}$ and $m_0$
as function of the cutoff, we impose renormalization conditions. In order
to fix $Z_1$, we demand that the inverse renormalized electron propagator
has the canonical kinetic term, i.e. $F(0)=1$. Our conventions imply
that the residue of the renormalized photon propagator $D_R(0)$ can be
identified with $\alpha _R$. Providing a value for $\alpha _R$
determines the constant $Z_3$. Finally, we set the scale of the
electron self energy $\Sigma (p^2)$ by the constraint $\Sigma (0)
=\Sigma _0$, which yields a condition to calculate $m_0$.
Although $m_0$ might tend to zero for infinite cutoff, the renormalized
(current) mass $m_R$ is non-zero. The behavior of $m_0$ in the quenched
approximation might serve as an illustrative example~\cite{bar86,co89}, i.e.
\begin{equation}
m_0 \; = \; \left( \frac{\mu }{\Lambda } \right)^w \; m_R(\mu) \; ,
\hbox to 2 true cm {\hfil with \hfil }
w=1-\sqrt{1- \frac{3e^2}{4\pi ^2 } } \; ,
\end{equation}
where $\mu $ is an arbitrary renormalization point.
Since $m_0$ will turn out to be different from zero for any fixed value
of the cutoff, the renormalized (current) mass $m_R$ is also non-zero
implying that the above set of renormalization conditions corresponds to
the case of a massive electron. The chiral symmetry is explicitly broken.
\subsection{ Numerical results }
\label{sec:3.1}
The set of coupled equations (\ref{eq:17},\ref{eq:18},\ref{eq:19})
for the functions $F(p^2)$, $\Sigma (p^2)$, $D_R(p^2)$ was
numerically solved by iteration. The integrals were done by
Simpson's integration with a step-size control. A CI-regime
is found for sufficiently small cutoff. A representative result
is shown in figure~1 for $\alpha _R = D_R(0) = 0.35$. The function
$F(p^2)$ is weakly $p$-dependent. The maximum deviation from one
occurs for $p^2=\Lambda ^2$ and is $0.04776$ for $\ln \Lambda ^2
/ \Sigma _0^2 =16$. The self-energy
roughly stays constant for $\ln p^2/ \Sigma _0^2 \stackrel{<}{\sim } 4 $
and smoothly decays afterwards. The vacuum polarization also stays
constant up to approximately the same momentum and then decays
according to
\begin{equation}
\frac{1}{D_R(p^2)} \; = \; c_1 (\alpha _R) \; - \;
c_2(\alpha _R) \, \ln \Lambda ^2/ p^2 \; ,
\hbox to 2cm {\hfil for \hfil }
(\ln p^2 / \Sigma _0^2 > 4 ) \; ,
\label{eq:20}
\end{equation}
where $c_1(0.35) \approx 2.97 $ and where $c_2 \approx 1/\pi ^2$ for
a wide range of renormalized couplings. The renormalization constants
are given in the table below.
\begin{center}
\begin{tabular}{|c|c|c|c|} \hline
$\ln \Lambda ^2 / \Sigma _0^2 $ & $Z_1$ & $Z_3$ & $m_0/\Sigma _0$ \\ \hline
7 & 1.0097 & 0.75537 & 0.49437 \\
10 & 1.0183 & 0.64295 & 0.33425 \\
13 & 1.0305 & 0.52927 & 0.20643 \\
16 & 1.0497 & 0.41307 & 0.10936 \\ \hline
\end{tabular}
\end{center}
The renormalization group flow is shown in figure~\ref{fig:2} for
$\alpha _R = 0.35 $. The vertex renormalization constant $Z_1$
(which is identical to the wave function renormalization constant $Z_2$
of the electron) stays close to one. The photon wave function
renormalization constant almost decays logarithmically.
The behavior of the bare mass $m_0$ corresponds to a powerlaw
decay with logarithmic corrections.
In order to check the accuracy
of the Ansatz (\ref{eq:15}) for the vertex, we study the Ward
identity (\ref{eq:14}) in momentum space and at zero photon momentum, i.e.
\begin{equation}
\widetilde{\Lambda }_{\mu }(0,p) \; = \; \frac{ \partial }{
\partial _\mu } S^{-1}(p) \; .
\label{eq:20a}
\end{equation}
Using the parametrization (\ref{eq:16}) of the electron propagator and
neglecting derivatives of the functions $F(p^2)$ and $\Sigma (p^2)$, the
Ward identity (\ref{eq:20a}) becomes
\begin{equation}
\widetilde{\Lambda }_{\mu }(0,p) \; = \; F(p^2) \, \gamma _\mu \; .
\label{eq:20b}
\end{equation}
The deviation of $F(p^2)$ from unity measures the violation
of the ward identity due to the Ansatz (\ref{eq:15}) for the vertex.
The numerical result shows that this violation is small, since
the deviation of $F(p^2)$ from unity is always beyond 5\% for
the values of the cutoff used in table 1.
Increasing the cutoff for fixed renormalized coupling $\alpha _R$, one
observes a critical upper limit $\Lambda _C$ of the cutoff.
Beyond this critical
value, no solution of the set of equations
(\ref{eq:16},\ref{eq:17},\ref{eq:18}) was found. This result can be
anticipated from the asymptotic behavior of the vacuum polarization
(\ref{eq:20}). For a fixed (positive) value of the renormalized
coupling $\alpha _R =D_R(0)$, equation (\ref{eq:20}) cannot be
satisfied for an arbitrarily large cutoff $\Lambda $.
Figure~3 shows the relation between the maximal possible
renormalized coupling and the critical cutoff.
The results indicate that in the limit $\Lambda \rightarrow \infty $
a solution of the coupled
Dyson-Schwinger equations only exists for $\alpha _R \rightarrow 0$.
This phenomenon is referred to as triviality of massive QED.
Note, however, that for values of the cutoff smaller than the critical
value the electron self-energy is cutoff independent. The theory is called
weakly renormalizable. For moderate values of the renormalized
coupling, the theory allows for a large value of the critical cutoff
and therefore for a cutoff many times bigger than the typical
energy scale set by the mass of the electron. This implies that
weakly renormalizable QED is perfectly compatible with the observations
in nature. The lack of a cutoff independence at high momentum does no harm,
since QED
in the real world is embedded in the Weinberg-Salam model at high energies.
\subsection{ The quenched approximation }
\label{sec:3.2}
Neglecting fermion loop effects in the photon propagator is called
quenched approximation. In our formulation, this corresponds to
replacing (\ref{eq:19}) by
\begin{equation}
\frac{1}{D_R(p^2)} \; = \; \frac{Z_3}{\alpha _R} \; .
\label{eq:21}
\end{equation}
The vacuum polarization $D_R(p^2)$ is momentum independent, and from
the renormalization conditions (see beginning of section \ref{sec:3})
one immediately obtains $Z_3=1$. For a constant vacuum polarization
$D_R$, the angle integral in (\ref{eq:18}) can be done explicitly and
yields zero. This implies that $F(p^2)=Z_1$ is also constant, and
taking into account the renormalization conditions, we have $Z_1=1$.
The only remaining non-trivial equation is the integral equation for
the electron self-energy (\ref{eq:17}). For standard QED (massless photon),
this integral equation can be transformed into a non-linear
differential equation with appropriate boundary
conditions~\cite{mir85,bar86,ko89,co89}.
In the table below, we compare the
renormalization constants of the full and the quenched approach
for $\alpha _R=0.35$ and a cutoff in the CI-region
$\ln \Lambda ^2 /\Sigma _0^2 = 16$.
\begin{center}
\begin{tabular}{|c|c|c|c|} \hline
& $Z_1$ & $Z_3$ & $m_0/\Sigma _0$ \\ \hline
quenched & 1 & 1 & 0.23172 \\
full & 1.0497 & 0.41307 & 0.10936 \\ \hline
\end{tabular}
\end{center}
Large deviations are found. Note, however, that renormalization constants
are not physical observables, and the quenched approximation might
improve when physical quantities are studied. Figure~\ref{fig:4}
shows the electron self-energy in the quenched approximation in
comparison with the full result.
One finds that the quenched approximation yields good results for the
electron self-energy at least for small momentum $p$. This behavior
can be understood by a scaling argument. Due to cutoff independence,
the momentum
dependence of the electron-self-energy for a small momentum $p$
and for cutoff slightly below the upper critical value is also obtained
in a calculation with a cutoff far beyond the critical cutoff and
momentum $p$ comparable to this cutoff (see figure 1).
At small cutoff, however, the bare electron mass $m_0$ is large
(see table 1), and
the electron loop contributing to the inverse vacuum polarization
in (\ref{eq:19}) is negligible, hence the quenched approximation
is good.
If the cutoff exceeds the upper critical limit, a solution
of the full equations ceases to exist, whereas the quenched
approximation still predicts a solution. This implies that
polarization effects are important to address the CI-behavior
of the model.
We conclude that within the quenched approximation one cannot decide
whether a CI-regime exists or not. If a CI-regime exists,
the quenched approximation may be a good approximation for
physical observables, although non-observables (e.g.\ renormalization
constants) might turn out to be completely different.
\section{ CI-violation in chiral symmetric QED }
\label{sec:4}
In the following, we will study the chiral limit $m_R=0$ (implying
$m_0 \equiv 0$) of QED, which requires a new set of renormalization
conditions. Again we demand that the residue of the electron
propagator is one, i.e.\ $F(0)=1$, which fixes $Z_1$.
We are interested in a phase with spontaneously broken
chiral symmetry. We therefore insist on $\Sigma (0)= \Sigma _0$,
which must now be accomplished by a choice of $Z_3$ since $m_0$
is now identical zero. All renormalization constants are fixed.
The renormalized coupling $\alpha _R = D_R(0)$ will be self-consistently
calculated.
We numerically solved the coupled equations
(\ref{eq:17},\ref{eq:18},\ref{eq:19}) for this set of renormalization
conditions. The result for the electron self-energy $\Sigma (p^2)$
and the vacuum polarization $D_R(p^2)$ is shown in figure~5,
which should be compared with figure 1. We find a logarithmic
dependence on the cutoff. Although the qualitative behavior of
$1/D_R(p^2)$ is qualitatively the same as in the case of massive
electron, the plateau at small $p^2$ does not stabilize, but
continuously increases with increasing cutoff. One might think that
this CI-violation in the self-energy is due to the change of the
plateau of $D_R(p^2)$ when the cutoff is varied.
One might therefore be tempted to parametrize
the vacuum polarization by
\begin{equation}
\frac{1}{D_R(p^2)} \; = \; c_3 (\alpha _R) \; - \;
c_4(\alpha _R) \, \ln \Lambda ^2/ p^2 \; ,
\hbox to 2cm {\hfil for \hfil }
\Lambda ^2 \rightarrow \infty
\label{eq:22}
\end{equation}
and to search for a CI-regime by solving the reduced set of
equations (\ref{eq:17},\ref{eq:18}). However, it turns out that
the CI-violation in the electron self-energy qualitatively
remains the same.
Two possible explanations of this cutoff dependence are immediate:
the theory is trivial in a sense that the limit $\Lambda \rightarrow
\infty $ is only compatible with $\alpha _R=0$.
In this case, the interaction is not strong enough to spontaneously
break chiral symmetry, and one cannot keep the renormalization
condition $\Sigma (0)=\Sigma _0 \not= 0$.
A second possible explanation is that in the chiral limit
the solutions are sensitive to the infrared behavior of the integrals
in (\ref{eq:17},\ref{eq:18},\ref{eq:19}). The momentum independent
Ansatz for the renormalized
vertex (\ref{eq:15}) might be to crude for the chiral limit and induces
the logarithmic CI-violation. An improved Ansatz for the vertex
function might provide a CI-regime with a spontaneous broken
symmetry. None of these two explanations can be favored without
further studies beyond the Ansatz (\ref{eq:15}) for the vertex
studied here.
\section{Conclusions}
\label{sec:5}
We numerically studied the coupled set of renormalized
Dyson-Schwinger equations for the electron and the photon propagator
using a tree level electron-photon-vertex as an Ansatz for the
renormalized vertex function. We imposed the constraint $Z_2=Z_1$
between the electron wave function renormalization constant $Z_2$
and the vertex renormalization constant $Z_1$ as required by gauge
invariance. The violation of the Ward identity is smaller than 5\%.
In the case of a massive electron (explicitly broken chiral symmetry)
and fixed renormalized coupling $\alpha _R$, a regime of cutoff
independence (CI-regime) was found, if the cutoff is below an upper
critical value. The theory is weakly renormalizable.
For larger values of the cutoff than the critical value,
no solution of the Dyson-Schwinger equations exists.
The results indicate that the infinite cutoff limit is only compatible with
a zero renormalized charge. In the CI-regime, the quenched
approximation yields reasonable results for the electron self-energy,
although the values of the renormalization constants completely
differ from the full result.
In the case of chiral symmetry (zero renormalized mass of the electron)
a logarithmic dependence on the cutoff was found. The question,
whether this cutoff dependence is due to the crude Ansatz for
the renormalized vertex or induced by triviality, cannot be answered
at the present stage of investigations. The question, whether
a CI-regime with spontaneously broken chiral symmetry exists in
strongly coupled QED, cannot be answered from the viewpoint of
the truncated Dyson-Schwinger equations supplied with tree level
vertex.
\bigskip
{\bf Acknowledgements: }
I thank J.-B.~Zuber for helpful discussions.
\begin {thebibliography}{sch90}
\bibitem{mir85}{ P. I. Fomin, V. P. Gusynin, V. A. Miransky,
Yu. A. Sitenko, Riv. Nuovo Cimento {\bf 6} (1983)1.
V. A. Miransky, Nuovo Cimento {\bf 90A} (1985) 149. }
\bibitem{bar86}{ C. N. Leung, S. T. Love, W. A. Bardeen,
Nucl. Phys. {\bf B273} (1986) 649.
W. A. Bardeen, C. N. Leung, S. T. Love, Phys. Rev. Lett.
{\bf 56} (1986) 1230, Nucl. Phys. {\bf B273} (1986) 649. }
\bibitem{co89}{ A. Cohen, H. Georgi, Nucl. Phys. {\bf B314} (1989) 7. }
\bibitem{nam61}{ Y. Nambu, G. Jona-Lasinio, Phys. Rev. {\bf 122}
(1961) 345, Phys. Rev. {\bf 124} (1961) 246.
D. Ebert, H. Reinhardt, Nucl. Phys. {\bf B271} (1986) 188.
S. P. Klevansky, Rev. Mod. Phys. {\bf 64} (1992) 649. }
\bibitem {landau} { L. D. Landau, I. Pomeranchuk,
Sov. Phys. JETP {\bf 29} (1955) 772.
E. Fradkin, Zh. JETP {\bf 28} (1955) 750.
M. Gell-Mann, F.\ Low, Phys. Rev. {\bf 95} (1954) 1300. }
\bibitem {ko89} {K.-I. Kondo, Phys. Lett. {\bf B226} (1989) 329,
Nucl. Phys. {\bf B351} (1991) 259. }
\bibitem {rakow} {P. E. L. Rakow, Nucl. Phys. {\bf B356} (1991) 27. }
\bibitem {latt94} { Lattice 94, Proc. XIIth Int. Symposium on Lattice
Field Theory, eds. F. Karsch et al., (Bielefeld, Germany, 1994)
Nucl. Phys. B (Proc. Suppl.) {\bf 42} (1995).}
\bibitem {illinois} { J. B. Kogut, E. Dagotto, A. Kocic,
Phys. Rev. Lett. {\bf 60} (1988) 772.
E. Dagotto, A. Kocic, J. B. Kogut,
Phys. Rev. Lett. {\bf 61} (1988) 2416, Nucl. Phys.
{\bf B317} (1989) 253, Nucl. Phys. {\bf B317} (1989) 271.
S. Hands, J. B. Kogut, E. Dagotto,
Nucl. Phys. {\bf B333} (1990) 551.
A. Kocic, J. B. Kogut, M.-P. Lombardo,
Nucl. Phys. {\bf B398} (1993) 376.}
\bibitem {desy} { M. G\"ockeler, R. Horsley, E. Laermann, P.E.L. Rakow,
G. Schierholz, R. Sommer, U.-J. Wiese, Nucl. Phys. {\bf B334} (1990) 527,
Phys. Lett. {\bf B251} (1990) 567, Erratum {\bf B256} (1991) 562.
M. G\"ockeler, R. Horsley, P.E.L. Rakow,
G. Schierholz, R. Sommer, Nucl. Phys. {\bf B371} (1992) 713.
M. G\"ockeler, R. Horsley, V. Linke, P.E.L. Rakow,
G. Schierholz, H. St\"uben, Nucl. Phys. (Proc. Suppl.)
{\bf B42} (1995) 660. }
\bibitem{itz80}{ C. Itzykson, J.-B. Zuber, ``Quantum field theory''
McGraw-Hill {\bf 1980}, S.~Pokorski, ``Gauge Field Theories'',
Cambridge-University-Press {\bf 1987}. }
\bibitem{kiz95}{ A. Kizilers\"u, M. Reenders, M. R. Pennigton,
Phys. Rev. {\bf D52} (1995) 1242. }
\bibitem {sugamoto}{ A. Sugamoto, Phys. Rev. {\bf D19} (1979) 1820.
M. Chemtob, K. Langfeld, ``Chiral symmetry breaking in strongly
coupled scalar QED'', Saclay preprint T95/072, hep-ph/9506341. }
\end{thebibliography}
\newpage
\centerline{ \bf \large Figure captions }
\vspace{2 true cm }
\vspace{.5 cm}
Figure 1: CI-behavior: the electron self-energy $\Sigma (p^2)$ and the
vacuum polarization $D_R$ as function of the
momentum $p^2$ for several values of the cutoff $\Lambda $ in units
of $\Sigma _0 = \Sigma (0)$.
\setcounter{figure}{1}
\begin{figure}[h]
\vspace{.5cm}
\caption{ The renormalization group flow for $\alpha _R = 0.35$;
$Z_1(\Lambda )$ short dashed line, $Z_3(\Lambda )$ long dashed line,
$m_0(\Lambda )/\Sigma _0 $ solid line. }
\label{fig:2}
\end{figure}
\vspace{.5 cm}
Figure 3: The maximal possible renormalized coupling $\alpha _R$
as function of the critical cutoff $\Lambda _c$ (left).
The ratio $Z_3/\alpha _R$ at the critical cutoff (right).
\vspace{.5 true cm }
\setcounter{figure}{3}
\begin{figure}[h]
\vspace{.5cm}
\caption{ The electron self-energy $\Sigma (p^2)$ in the quenched
approximation compared with the full result for $\alpha _R=0.35$
and $\Lambda ^2 = 8.9 \, 10^{6} \, \Sigma _0 $. }
\label{fig:4}
\end{figure}
\vspace{.5 cm}
Figure 5: CI-violations:
the electron self-energy $\Sigma (p^2)$ and the
vacuum polarization $D_R$ as function of the
momentum $p^2$ for several values of the cutoff $\Lambda $ in units
of $\Sigma _0 = \Sigma (0)$ in the chiral limit $m_R=0$.
\end{document}
|
\section{Introduction}
In the present paper we propose a
heterogeneous reaction-diffusion model for the hydration and
setting of cement. The proposed model is based on the
experimental observation
that cement hydration can be described by a
dissolution/precipitation mechanism \cite{b0,b3,b8,b16,b12}.
The elementary aspects of the cement hydration/setting process
on a mesoscopic length scale can be characterized as follows.
At the initial stage cement particles or powder
(in our case tricalcium silicate, $C_3S \equiv Ca_3SiO_5$)
are mixed well with the solvent (water).
Rapidly after mixing the {\em dissolution} reaction
of the cement particles starts. Its principal reaction
products are ions which are mobile and
may {\em diffuse} into the bulk of the solvent
(in our case $Ca^{2+}$ , $OH^{-}$ and $H_2SiO_4^{2-}$ ions). At this stage the ion
concentrations in the bulk of the solvent are very low and, as
a consequence, one finds strong ion fluxes from the dissolution
front away into the solvent. However, for a given temperature and
pressure the ion concentrations can not take arbitrary high values.
Rather, the ion concentrations are bounded by finite
solubility products above which solid phases start to {\em precipitate}
from the solution. There are two associated precipitation reactions:
a) the precipitation of calcium hydro-silicate sometimes referred as
`cement gel'
($C_{1.5}SH_{2.5} \equiv (CaO)_{1.5}\,(SiO_2)\,(H_2O)_{2.5}$)
and b) the precipitation of calcium hydroxide or `Portlandite'
($CH \equiv Ca(OH)_2 $)\cite{b8}.
While the growth of the cement gel is the
basis for the whole cement binding process, the growth of Portlandite
mainly happens in order to compensate for the accumulation of
$Ca^{2+}$ and $OH^{-}$ ions in solution (see below).
The process of cement dissolution, ion transport,
and cement gel/Portlandite precipitation is usually referred as
`cement hydration'.
This process is to a high extent heterogeneous in the sense, that
starting from initially random cement particle positions
dissolution and precipitation reactions change the
physical (boundary) conditions for the ion transports themselves.
This is due to the fact
that the diffusion within the solid phases can be neglected
relative to the solvent one.
Coupled chemical dissolution/precipitation reactions
are well known since a long time in geology and geochemistry.
As such kind of processes are fundamental for water/mineral systems
some effort has been undertaken to modellize
such systems \cite{b5,b14}. The approaches differ methodologically
and focus on different physical and chemical aspects.
Ref.~\ref{b5} considers the ion transport problem
for a single, solubility controlled, dissolution/precipitation reaction
employing a one dimensional cellular automaton approach.
A chemically more detailed approach of water/mineral interactions
includes considerations on the influence of
nucleation, electro chemistry and temperature
on chemical reaction rates, however,
without considering a transport equation\cite{b14}.
Recently, a relatively simple stochastic cellular automaton model
for cement hydration/setting has been proposed \cite{b6}.
Therein it has been tacitly assumed that ion transport is not relevant
for the hydration process above a certain mesoscopic
length scale (`pixel size').
Furthermore, in this model hydrate `pixels' perform a random walk
until they touch another solid `pixel', where they then
stick \cite{b6}.
In contrast to this
we position our deterministic model into the `opposite' direction:
mass transport happens due to diffusion of ions, and the hydrates
are regarded as immobile.
In Sec.~\ref{model} we give a detailed description of the employed
model. We consider heterogeneity aspects in Sec.~\ref{model-hetero},
chemical aspects in Sec.~\ref{model-reactions} and transport aspects
in Sec.~\ref{model-transport}.
Sec.~\ref{results} contains first results.
We present some images of calculated cement micro
structures, which
comprehensively demonstrate basic features and capabilities of the
present approach.
We consider parametric plots for the mean ion
concentrations, known in cement literature as `kinetic path
approach' and compare our findings with experimental
results \cite{b1}.
Following this we investigate the variation of the maximum average silica
concentration in solution for various chemical reaction rate
constants.
Finally we present curves for the hydration advancement in
time for two different sets of reaction rate constants.
In Sec.~\ref{conclusions} we summarize and give some
ideas about future work.
\section{The Model}\label{model}
In the following we describe the proposed
reaction diffusion model for the hydration and setting of
cement in more detail.
We will first discuss how to quantitatively describe
the heterogeneities and outline some general features of
the employed model.
Following this we will summarize the dissolution/precipitation
reactions and the employed reaction rate laws.
Finally some physical aspects of the ion transport in the solvent are
discussed. The later one represents the {\em physical coupling}
between the dissolution and precipitation reactions.
\subsection{Heterogeneity aspects}\label{model-hetero}
We consider a discrete reaction-diffusion model in space
and time. The physicochemical system can be regarded as
being composed of `sufficient' small volume elements $V=\Delta x^3$
with $\Delta x$ typically between $10^{-6}m$ and $10^{-4}m$.
The initial configuration is determined from a digitized micro graph
image as follows:
the micro graph is immediately taken
after mixing at time $t_0$,
one can assume to find only the solid phase $C_3S$
and water. By measuring the occupied volumes $V_i^{(\alpha)}$
of the phase $\alpha$, ($\alpha = aq, C_3S$), in each cell $i$ using
an image processing system, one determines the initial distribution
of mole numbers $n_i^{(\alpha)}(t_0)$ according to
$n_i^{(\alpha)}(t_0)= V_i^{(\alpha)}(t_0)/v^{(\alpha)}$.
The $v^{(\alpha)}$ are known molecular volumes for room temperature,
see Table I.
In general, such a micro graph shows cement particles of
different sizes immersed in water.
Because each cell volume is completely filled with solid
and solvent phases, one can calculate the solvent volume
of cell $i$ {\em at any time step}
from the corresponding solid volume(s),
\begin{equation}\label{solvent-volume}
V_i^{(aq)}(t) = V - \sum_{\alpha\ne aq} n_i^{(\alpha)}(t)
v^{(\alpha)}.
\end{equation}
This is a very equation, because ion concentrations are
usually calculated with respect to the actual solvent volume
$V_i^{(aq)}(t)$.
The physical constraints of Eq.(\ref{solvent-volume}) are that
the chemical reactions must be {\em sufficiently slow}
in comparison to the solvent flow. Furthermore, it is
tacitly assumed that the whole system is connected to an external water
reservoir in such a way that the solid phases do not hinder the
flow.
In a given volume element
the dissolution/precipitation reactions will, in general, not
instantaneously go to completion. The solvent volume
in each cell will rather increase/decrease
continuously, defining a solvent distribution field,
\begin{equation}\label{porosity-field}
\varepsilon_i^{(aq)}(t)=V_i^{(aq)}(t)/V.
\end{equation}
The solid phase volume fractions $\varepsilon_i^{(\alpha)}(t)$ are
defined analogously.\\
We believe in fact that the volume fractions
$\varepsilon_i^{(\alpha)}(t)$, reflecting the
actual solvent/solid distribution in the system,
are sufficient to characterize most of
the aspects of the hydration process on a
macroscopic length scale.\\
Though we do not have information about these distributions
on a length scale smaller than $\Delta x$ we will assume
that the reactants are homogeneously distributed in each
volume element. The volume fractions may then be interpreted
as probabilities to find
at a random position within cell $i$
the reactant $\alpha$.
\subsection{Chemical aspects}\label{model-reactions}
In the following we will summarize the
considered dissolution/precipitation
reactions and the employed reaction rate laws.
During the course of the simulation
it may happen that the reaction flux,
as calculated from the kinetic equations, does exceed the available
amounts of chemical reactants in a given cell.
We therefore check at each time step and for each chemical reaction
if there exists a limiting reactant. If this happens, we define the
reaction flux through the amount of the available limiting reactant.
\subsubsection{$C_3S$ dissolution}
The $C_3S$ dissolution reaction is a spontaneous and exothermic
{\em surface reaction}
which happens at the $C_3S$-solvent interface(s) \cite{b8}.
It can be considered as irreversible,
\begin{equation}\label{dissolution-reaction}
C_3S_{(s)} + 3H_2O_{(\ell)}
\stackrel{k_1}{\rightharpoonup} 3Ca^{2+}_{(aq)} +
4OH^{-}_{(aq)} + H_2SiO_{4(aq)}^{2-}.
\end{equation}
Here $k_1$ denotes an appropriate surface reaction
constant in units of $mol\,m^{-2}\,s^{-1}$.\\
Chemical reactions must be considered as local, i.e.,
each volume element is acting as a small and independent
chemical reactor as long as no transport occurs. However,
for a pronounced dissolution reaction in a given cell
one has also to consider the possible reaction amounts
originating from the interfaces with its neighboring cells.
This can be done by allowing cement dissolution in cell $i$ through the
electro chemicalsolvent of that cell and through a fraction $\gamma$ of the
solvent of the neighboring cells $j$,
\begin{equation}\label{dissolution-kinetics}
\nu_1^{(\alpha)} \, d\xi_{i,1}(t)=
\nu_1^{(\alpha)}\, k_1\, \Delta t\, \Delta x^2\, \bigl(
p_{ii}
+\gamma \sum_{j=nn(i)}
(p_{ij} - p_{ii})\bigr).
\end{equation}
Here $\nu_1^{(\alpha)}$ denote the stoichiometric numbers of
species $\alpha$ in reaction (\ref{dissolution-reaction}), see
Table I.
The changes
in mole numbers of species $\alpha$ in cell $i$
in a time intervall $\Delta t$ due to this reaction are
denoted by $\nu_1^{(\alpha)} \, d\xi_{i,1}(t)$.
The sum on the right hand has to be taken over all
next nearest neighboring cells of cell $i$, as indicated by
$j=nn(i)$.
The first term $p_{ii}=\varepsilon_i^{(C_3S)}\varepsilon_i^{(aq)}$
in Eq.(\ref{dissolution-kinetics}) describes the
dissolution within cell $i$ ($\gamma = 0$) due to a `typical' reaction
interface. It can be understood
as a probability to find the two reactants,
$C_3S$ and water, in contact at an arbitrary
chosen point within cell $i$.
The `chemically active interface' between cell $i$ and $j$
is given by $p_{ij}\Delta x^2=
\Delta x^2 \varepsilon_i^{(C_3S)}\varepsilon_j^{(aq)}$.
The constant $\gamma$ controls the degree of dissolution due to
this interface(s). We have used a value $\gamma =1/8$ in our
simulations, yielding comparable reaction rates from inner
and outer cell surfaces.
The determination of the reaction interfaces
in terms of cement and water volume fractions is similar to
the degree of surface coverage used in
Langmuir absorption theory \cite{b11}.
However, the absorption of water
on the cement's surface and the desorption of ions
from this surface are not necessarily the rate determining steps.
In fact, the dissolution reaction (\ref{dissolution-reaction})
involves various physical and subprocesses
\cite{b12}
which will not be considered here.
\subsubsection{$CSH$ dissolution/precipitation}
The precipitation of $C_{1.5}SH_{2.5}$ is an endothermic reaction
which can be considered as
reversible \cite{b8},
\begin{eqnarray}\label{csh-precipitation}
C_{1.5}SH_{2.5(s)} &\stackrel{k_2}{\rightleftharpoons} &H_2SiO_{4(aq)}^{2-} \nonumber\\
&+& 1.5Ca^{2+}_{(aq)} + OH^{-}_{(aq)} +H_2O_{(\ell)},
\end{eqnarray}
having a forward (dissolution) rate constant $k_2$
(in units of $mol\,m^{-2}\,s^{-1}$).
We note, that the $CSH$ precipitation, i.e., the backward reaction
in Eq.(\ref{csh-precipitation}) is the main reaction in this
balance equation.
The reactants and products are considered to have a
fixed stoichiometry.
However, it is known from experimental
data that the stoichiometry of $C_xS_yH_z$
can be rather
variable (`solid solution') \cite{b3}.
We will not include these complications
into the model. Instead we consider a fixed and typical
calcium/silica ratio equal to $1.5$.
In equilibrium the $C_{1.5}SH_{2.5}$
dissolution/precipitation reactions are controlled
by an empirical value of solubility, giving the maximum amount
of solid one can dissolve in aqueous solution at a given
temperature (room temperature) and pressure (normal pressure).
Employing this empirical solubility constant one can directly
determine the equilibrium solubility product,
$S_2^{equi}=S_{2i}\vert^{equi}=
(Ca^{2+})_i^{1.5}(OH^{-})_i(H_2SiO_4^{2-})_i\vert^{equi}
\approx 1.2\times 10^{-9}\,mol^{3.5}\,liter^{-3.5}$, comp. Table I.
The square brackets denote
the ion concentrations with respect to the
available solvent, i.e.,
$(Ca^{2+})_i=n_i^{(Ca^{2+})}/V_i^{(aq)}$ etc..
When in a given cell the ion product
is larger than the solubility product, locally precipitation happens.
If it is lower, $C_{1.5}SH_{2.5}$ becomes dissolved.
The $CSH$ dissolution is only marginal and for this reason we will
employ for the rate of dissolution a simpler expression than
for the $C_3S$ dissolution reaction.
We assume that the dissolution reaction in cell $i$ is proportional
to the typical
reaction interface $\Delta x^2 \varepsilon_i^{(aq)}$ with
the dissolution constant $k_2$ as the constant of
proportionality.
The precipitation reaction is assumed to be proportional to the
reaction interface and proportional
to the ion product $S_{2i}$, as defined above.
The rate constant of the precipitation reaction
can be eliminated, by use of the equilibrium condition, $d\xi_{i,2}=0$.
We have employed the following rate equation,
\begin{eqnarray}\label{precipitation-kinetics1}
\nu_2^{(\alpha)} \, d\xi_{i,2}(t) =
\nu_2^{(\alpha)}\, k_2\, \Delta t\, \Delta x^2\, \varepsilon_i^{(aq)}
\bigl(1- \frac{S_{2i}}{S_2^{equi}}\bigr).
\end{eqnarray}
The stoichiometric numbers again are denoted by $\nu_2^{(\alpha)}$,
see Table I, and the change in mole numbers of
species $\alpha$ due to this reaction is
$\nu_2^{(\alpha)} \, d\xi_{i,2}(t)$.
\subsubsection{CH dissolution/precipitation}
The precipitation of calcium hydroxide accompanies
the $CSH$ precipitation.
This is because the $CSH$ precipitation does not consume
the ions in the same proportions as they are released due to $C_3S$
dissolution.
The non-reacted $Ca^{2+}$ and $OH^{-}$ ions soon begin to
accumulate in solution, until the solubility
of $CH$ is exceeded. However, the precipitation of $CH$ and $CSH$
are, in general, not simultaneous because the corresponding
solubilities are very different, comp. Table I.
The dissolution/precipitation of $CH$
can be considered as reversible,
\begin{equation}\label{ch-precipitation}
CH_{(s)} \stackrel{k_3}{\rightleftharpoons} Ca^{2+}_{(aq)} + 2 OH^{-}_{(aq)} ,
\end{equation}
having a dissolution constant $k_3$ ($mol\,m^{-2}\,s^{-1}$).
The corresponding solubility product for $CH$ is defined as
$S_3^{equi}=S_{3i}\vert^{equi}=
(Ca^{2+})_i(OH^{-})_i^2\vert^{equi}
\approx 3.3\times 10^{-6}\, mol^3\, liter^{-3}$
which is about three orders of
magnitude larger than $S_2^{equi}$, see Table I,
\begin{equation}\label{precipitation-kinetics2}
\nu_3^{(\alpha)} \, d\xi_{i,3}(t) =
\nu_3^{(\alpha)}\, k_3\, \Delta t\, \Delta x^2\,
\varepsilon_i^{(aq)}
\bigl(1- \frac{S_{3i}}{S_3^{equi}}\bigr).
\end{equation}
In all other respects the $CH$ reaction is treated analogously to
the $CSH$ dissolution/precipitation reaction.
\subsubsection{Chemical shrinkage}
One can write down from Eqs.(\ref{dissolution-reaction}),
(\ref{csh-precipitation}) and (\ref{ch-precipitation}) the
net reaction for the cement hydration process,
\begin{equation}\label{net-reaction}
C_3S_{(s)} + 4H_2O_{(\ell)}
\rightharpoonup C_{1.5}SH_{2.5(s)} + 1.5 CH_{(s)}.
\end{equation}
Calculating for the above equation the occupied volumes for one mole
one finds for the left side approximately $144\, cm^3$ and for the
hydrate products $130\, cm^3$, comp. Table I.
Hence, the reaction products occupy
a volume around 10\% smaller than the reactants (including solvent).
This effect, which
is typical for hydraulic binders, is termed `chemical shrinkage' or
Le Chatelier effect. It is responsible for the transition of cement
suspensions or pastes having a finite viscosity and zero elastic
moduli to rigid hydrated cement having an infinite viscosity and
finite elastic moduli. Wherever topologically possible water flow
will try to compensate for the loss of volume. However, in the
course of an experimental cement setting and hardening (rigidification)
process available
solvent flow paths may vanish and volume loss becomes partially
compensated building up mechanical deformations (stresses)
within the solid phase agglomerate. In turn these stresses influence
the reaction rates (which in general are pressure dependent) in a
complex manner \cite{b16}.
The appearence of voids (pores) on the submicron scale
is experimentally also well established \cite{b0}.
We will not modellize such complex problems as mechanical stresses,
solvent cavitation and solvent flow
due to chemical shrinkage in our present approach.
Instead we will assume throughout all calculations that the solvent
is able to balance the chemical shrinkage for all time steps and
for all volume elements according to Eq.(\ref{solvent-volume}).
This assumption is equivalent to the
introduction of local source terms for water, i.e., the hydrating
system is connected to an external water reservoir in an
appropriate way.
We would like to point out that
Eq.(\ref{net-reaction}) is an overall reaction which per definition
only holds for an isolated system. On the other hand one
cannot treat the volume elements as chemically isolated,
simply because of the ion fluxes going through the elements.
This and the different values for the $CSH$ and $CH$ solubilities
imply in general locally non-congruent precipitation reactions.
As a consequence one cannot simply `replace' one
dissolved unit volume of $C_3S$
by 1.7 unit volumes $CSH$ and 0.6 unit volumes $CH$ as
has been proposed in \cite{b6}.
\subsection{Transport aspects}\label{model-transport}
It is known from experiments that ion transport due to
diffusion is fundamental
in cement chemistry, because of very strong ion concentration
gradients close to the dissolving $C_3S$-solvent interface(s).
In general one
should consider the convective diffusion equation, however,
this would imply to solve the
hydrodynamic equations.
In the present paper we consider the
usual diffusion equation with local, appropriately
defined transport coefficients. It is in the
context of the investigated problem useful to consider only the ion
diffusion within the solvent. For convenience we document in
Table I the diffusion constants $D^{(\alpha)}$
as calculated from
the electric mobilities at room temperature and normal pressure
for infinite dilution. The solids and the
solvent's diffusion constants are set to zero \cite{b13}.
This allows a relatively compact notation for the equations
of continuity,
\begin{eqnarray}\label{continuity-equation}
\Delta n_i^{(\alpha)}&=&
\frac{\Delta t}{\Delta x^2}V
\sum_{j=nn(i)}
\bigl(
D_{ij}^{(\alpha)}c_j^{(\alpha)}
-D_{ij}^{(\alpha)}c_i^{(\alpha)}
\bigr) \nonumber \\
&+& \sum_{k=1}^3 \nu_k^{(\alpha)}\, d\xi_{i,k}.
\end{eqnarray}
The second term on the right side of Eq.(\ref{continuity-equation})
represents the
already known source term due to the chemical reactions
Eqs.(\ref{dissolution-kinetics}), (\ref{precipitation-kinetics1})
and (\ref{precipitation-kinetics2}).
The $D_{ij}^{(\alpha)}$ are the transport coefficients, that depend
on the solvent volume fractions (porosities) of the cells involved
in the diffusion process,
$D_{ij}^{(\alpha)}=D^{(\alpha)}\varepsilon_i^{(aq)}\varepsilon_j^{(aq)}$.
The ion concentrations are taken with respect to the actual
solvent volume, i.e., $c_i^{(\alpha)}=n_i^{(\alpha)}/V_i^{(aq)}$.
We have employed periodic boundary conditions
in all calculations.
The general formulation of the ion transport problem
in cement chemistry
involves various subproblems which we will not consider at this
stage of modelization, as there are the transport due to
\subsubsection{Heat conductance}
The
cement dissolution reaction is strongly exothermic.
Analogous to the concentration gradients one finds near the
various dissolution fronts strong temperature gradients,
leading to non isothermal solvent flow fields, with its complications.
The heat redistribution is probably also of direct
importance for the dissolution/precipitation
reactions, because their reaction constants
as well as the corresponding solubility products
usually depend very sensitively on the temperature. We will
neglect this possible effects in our approach assuming an
overall constant room temperature, $T_0=298$ Kelvin.
\subsubsection{Electrostatic interactions}
Because one is dealing with the transport of electrically
charged particles, one must consider, in general, Poisson's equation
for the electrostatic potential in order to determine the
resulting ion mobilities due to their electric field.
We will neglect in our simulations
all electrostatic interactions. Furthermore,
all chemical rate calculations are done
in terms of molar concentrations and not in terms of
`chemical activities' (which is equivalent in the limit
of infinite dilution). Hence we will treat the ions in solution
as uncharged particles throughout all calculations.
\subsubsection{Solvent flow}
This is
possibly of importance for the overall hydration process,
because the solvent has to move during the
dissolution/precipitation reactions accordingly, in order to
compensate for the loss/gain of solid volume,
creating advective ion fluxes. For an infinite, plane cement-water
interface one might argue that the solvent flow is always
perpendicular to the interface. Under this particular conditions
it can be shown by scaling arguments that
advective ion transport is
unimportant relative to the diffusion transport, because of
the relatively low molecular volume of cement, comp. Table I.
However, in a finite heterogeneous system solvent flows tangential
with respect to the interfaces are also expected and the argument
given above is limited.
In principle one has to consider the equations of
hydrodynamics for this purpose. However, this moving boundary
problem is complex and we will not consider it in this form here.
Instead we will determine the spatial and temporal
solvent distribution $V_i^{(aq)}(t)$ according to
Eq.(\ref{solvent-volume}).
\section{Results}\label{results}
{}From a general point of view cement hydration can be regarded
as a heterogeneous (non equilibrium)
solid phase transformation forming from an anhydrous solid
phase two hydrated solid phases\cite{b17}.
In contrast to known solid-solid phase transformations for example
in metallic alloys, cement hydration bases strongly on the
presence of a solvent phase (mostly water).
The solvent controlls the transformation in a twofold sense: it is
directly part of the chemical reactions and on the other hand
it controlls the ion transport to a very large extent.
The complexity of the problem and the features of the present
approach are most convenient illustrated by Fig.~\ref{fig:images}.
Therein we show the solid volume fractions of anhydrous
cement, $\varepsilon^{(C_3S)}$ (left side), and of hydrated cement,
$\varepsilon^{(CSH)}$ (right side), for three different hydration times
(a) $t=1500\,s$, (b) $t=2500\,s$ and (c) $t=5000\,s$. The occupied
volume fractions are color coded ranging from blue (0-20\%),
cyan (20-40\%), green (40-60\%) over yellow-orange (60-80\%) to
red (80-100\%).
The calculated micrographs Fig.~\ref{fig:images}a-c
show, as expected, spatially inhomogeneous
nucleation and growth of $CSH$ hydrate.
The hydrate precipitation and the associated growth
of hydrate surface layers surrounding dissolving
cement particles is relatively slow.
However, in Fig.~\ref{fig:images}b one
especially observes that the
rate of $CSH$ precipitation is spatially
strongly varying (notice the red and cyan
spots in the right image). This reflects the spatial fluctuations
in chemical reactivities (local $C_3S$ surface/volume ratio), i.e.,
the reactivity of a `fjord' is higher
than that of a big `lake'.
Supported by experimental observations it has been argued that
cement particle surfaces close to each other may indeed act as
very strong inhomogeneities leading to localized $CSH$ nucleation
forming `bridges' between adjacent particles \cite{b9}.
Our micrographs confirm this picture. We note, that the proposed
model does not include any (auto-)catalytic effect of $CSH$ on
its growth yet, compare Eq.~(\ref{precipitation-kinetics1}).
This point is intensively studied at the moment.
As the hydrate precipitation/dissolution reaction rates
depend on the local supersaturation values, the interesting question
arises how global characteristic points of the cement hydration
kinetics can be defined.
In a recent experimental work it has been proposed to characterize
different kinetic regimes occuring during the $C_3S$ hydration
by means of parametric plots of `typical'
calcium versus silicate concentrations as estimated from
calorimetric or conductimetric measurements
(`kinetic path approach')\cite{b1}. The experiments have been
conducted employing stirred diluted $C_3S$ suspensions having a
water/cement weight ratio between $10$ and $50$.
However, to obtain typical quantities of such diluted suspensions
employing numerical simulations one would have to consider very
large systems. Instead we will compare the experimental suspension
data with numerical cement paste data having a water/cement ratio
close to the practically important value $0.5$.
For convenience we reproduce in Fig.~\ref{fig:fig1}
some of the experimental data $(\diamond)$ of Ref.~\ref{b1}.
Initially one finds an `inductive'
cement dissolution period (period $AB$ in Fig.~\ref{fig:fig1}).
During this period the bulk solution is everywhere undersaturated
with respect to $CSH$ and $CH$, the concentration gradients
close to the cement particles are very high, and ions
rapidly distribute into the bulk solvent. We have
observed in our simulations that $CSH$ is already formed during
the `induction' period, however, to a very low extent. Because
of the initially very high concentration gradients
the ion products exceed the
$CSH$ solubility only in a very thin layer surrounding the cement
particles.
As the ions cummulate in solution the thickness of this
precipitating layer increases due to the broadening of the
ion distributions.
The precipitation counteracts
the further increase in silica concentration
both by chemical reaction and
by decreasing the cement particles surface permeability.
In result the silica concentration passes through a maximum
(point $B$ in Fig.~\ref{fig:fig1}), which might be interpreted as a
{\em global measure} for the onset of $CSH$ precipitation.
During the period $BC$ the silica concentration
decreases, while the calcium and hydroxyl concentrations continue
to increase as the $CSH$ precipitation consumes only a fraction of
these ions, see Eqs.~(\ref{dissolution-reaction})
and (\ref{csh-precipitation}).
Typically after a couple of hours
the solution becomes oversaturated with respect to calcium
hydroxide $CH$ (point $C$ in Fig.~\ref{fig:fig1}).
Both hydrates
precipitate very slowly
(experimentally several weeks between point $C$ and $D$),
lowering both calcium
and hydroxyl concentrations. Point $D$ corresponds to the
state after infinite hydration time terminating the shown curve.
The progress in time is indicated by arrows. Our numerical results
$(+)$ are in reasonable agreement with the experimental data $(\diamond)$.
We have observed that calculated positions and values for point $C$
depend a) on the employed initial water/cement weight ratio
and b) on the three reaction rate constants $k_1$, $k_2$ and $k_3$
in a rather complex manner.
Contrary to this point $B$ is found to depend only slightly
on the {\em relative} reaction rate constant $k_1/k_2$ over
four orders of magnitude, see Fig.~\ref{fig:fig3}a.
A consequence of the much smaller solubility of $CSH$ as compared to
the second hydrate $CH$ is that in the period $AB$ only {\em two}
chemical reaction are operative. Reaction Eq.~(\ref{ch-precipitation})
is inoperative.
For $k_1/k_2$ ranging between $0.1$ and $10$ the maximum silica
concentrations take constant values of about $1\,mmol/liter$ while
for $k_1/k_2=10^3$ we find $\approx 3\,mmol/liter$.
We have also performed simulations for similar $k_1/k_2$ values employing
different {\em absolute} reaction rate constants, see
Fig.~\ref{fig:fig3}a. We find a
relatively good data collapse over four orders of magnitude.
This result is important insofar as the reaction rate constants have
not been experimentally measured yet. In Fig.~\ref{fig:fig3}b we plot
the momentaneous calcium concentration
versus the silica concentration at point $B$ for various reaction
rate constants. The data are centered around a straight line of slope
$1/3$. Comparison of this result with the stoichiometric coefficients in
Eq.~(\ref{dissolution-reaction}) shows that the silica/calcium
ratio of ions is determined by the dissolution reaction.
The characterisation of the
temporal advancement of the hydration process is an important
problem. We show in Fig.~\ref{fig:fig4}
some preliminary results for the time dependence of the mean volume
fraction $\langle \varepsilon^{(CSH)}\rangle$
of the cementious hydrate $CSH$.
For `moderate' reaction constants $(\diamond)$ the hydration is
found to be relatively slow, i.e., after $70\, hours$
of hydration we find only $10\%$ $CSH$.
This value is approximately two times
lower than the experimental one observed by NMR\cite{dobs}.
We note that both
the hydration curve $(\diamond)$ in Fig.~\ref{fig:fig4}
and the parametric curve $(+)$ in Fig.~\ref{fig:fig1}
belong to the same simulation.
Apparently the hydration advancement for `moderate' reaction rates
is nearly constant in time, see curve $(\diamond)$ in Fig.~\ref{fig:fig4}.
For comparison we also show in Fig.~\ref{fig:fig4} a hydration
curve for `fast' hydration $(+)$. After a few minutes
of `induction' period
the hydration rapidly accelerates
and goes already after $1\,hour$ to completion (remaining $C_3S$ less
than $1\;\%$). This process is much to fast and it leads to typical
ion concentrations of $mol/liter$ which are unrealistic high.
As long as the reaction rate constants have not been estimated from
experiments yet one main difficulty in the modellization of cement
hydration consists in finding appropriate values for the rate constants,
{\em which do not contradict} experimental measurements
of mean ion concentrations and hydration advancement. However, the
experimental results belong to hydration in space while
our calculations correspond to two dimensional hydration.
This point is being currently investigated.
\section{Conclusions}\label{conclusions}
We have presented a general, heterogeneous reaction-diffusion model
for solid phase transformation
due to chemical dissolution/precipitation
reactions. The model has focussed on the
important industrial problem of Portland cement hydration though it is
more generally applicable to water/mineral systems.
We have tried to develop an `open' approach based on physical and
chemical considerations (mainly the laws of mass conservation and mass
action). The model includes in its present form
on a coarse grained
length and time scale the full spatial
distribution of solid and liquid phases,
the three main chemical dissolution/precipitation
reactions, and the transport of ions due to diffusion.
The proposed approach naturally allows to include for a varity
of more or less important phenomena depending on imposed
conditions such as solvent flow effects, electrochemical effects,
exothermic effects including heat conduction, pressure effects,
inert filler effects etc..
We have tried to incorporate a reasonable amount of
specific information about the cement hydration into the
investigated model such as
the stoichiometry and kinetics of Portland cement dissolution, the
precipitation/dissolution of the main cementious hydrate (CSH) and
of Portlandite (CH), their approximate solubilities and
molecular volumes, approximate values for the ion
diffusivities in aqueous solution, and initial conditions for water
immersed cement particles (sizes and spatial positions).
The presented results demonstrate considerable richness and complexity
of the cement hydration phenomenon close to controlled
experimental situations.
We have presented some calculated cement micro structures as they
evolve in time. Nucleation and growth of hydrates is found to be
strongly heterogeneous in agreement with experimental observations.
The problem of `autocatalytic' effects on the precipitation processes needs
further investigation.
The presented parametric plots for the average
concentrations of ions in solution
are in qualitative agreement with recent indirect experimental
observations made for stirred cement suspensions\cite{b1}.
The calculated maximum silica concentrations and their
corresponding calcium
concentrations are in reasonable quantitative agreement with experimental
values.
For high dissolution and low precipitation rate constants
we find unacceptable high ion concentrations of order $mol/liter$.
One way to circumvent this problem would be the introduction
of a {\em finite} solubility for $C_3S$ in
Eq.~(\ref{dissolution-kinetics}), however, such a solubility constant
has not been experimentally determined yet.
Furthermore we have investigated the variations of the maximum silica
concentrations for
various reaction rate constants.
The concentrations are found to vary only
slightly with the relative reaction rate constant $k_1/k_2$
over four orders of magnitude.
It would be very interesting to see how the second characteristic
hydration point (point $C$ in Fig.~\ref{fig:fig1}) depend
on the initial water/cement ratio and on the employed reaction
rate constants.
Finally we presented first results on overall hydration curves.
This curves are not very realistic yet. It would be very helpfull
to have experimental order of magnitude estimates for the three
(unknown) reaction rate constants.
For future work we are planning to conduct calculations
in three dimensions with improved kinetic equations.
\section*{Acknowledgments}
We would like to acknowledge stimulating and interesting discussions
with Ch. Vernet, H. Van Damme, S. Schwarzer, A. Nonat and D. Damidot.
F.T. would also like to acknowledge financial support from
CNRS, from GDR project 'Physique des
Milieux H\'et\'erog\`enes Complexes' and from Lafarge Coppee Recherche.
|
\section{\@startsection {section}{1}{\z@}{-3.5ex plus -1ex minus
-.2ex}{2.3ex plus .2ex}{\normalsize\bf}}
\def\subsection{\@startsection{subsection}{2}{\z@}{-3.25ex plus -1ex minus
-.2ex}{1.5ex plus .2ex}{\normalsize\bf}}
\def\@cite#1#2{${}^{\mbox{\scriptsize#1\if@tempswa , #2\fi}}$}
\def\thebibliography#1{\section*{References\markboth
{REFERENCES}{REFERENCES}}\list
{\arabic{enumi}.}{\settowidth\labelwidth{[#1]}\leftmargin\labelwidth
\advance\leftmargin\labelsep
\usecounter{enumi}}
\def\hskip .11em plus .33em minus -.07em{\hskip .11em plus .33em minus -.07em}
\sloppy
\sfcode`\.=1000\relax}
\let\endthebibliography=\endlist
\catcode `\@=12
\begin{document}
{\bf ON HAMILTONIAN AND QUANTUM DYNAMICS OF MASSLESS
PARTICLES.}\vspace{1.3cm}\\
\noindent
\hspace*{1in}
\begin{minipage}{13cm} Andreas Bette, \vspace{0.1cm}\\
Stockholm University, \vspace{0.1cm}\\
Department of Physics, \vspace{0.1cm}\\
Box 6730, \vspace{0.1cm}\\
S-113 85 Stockholm, Sweden.\vspace{0.1cm}\\
fax: +46-8347817 att. Andreas Bette, \vspace{0.1cm}\\
e-mail: $<$ab@vanosf.physto.se$>$
\end{minipage}
\vspace*{0.5cm}
\begin{abstract}
\noindent
A short review of special relativistic dynamics describing a particle
acted upon by an arbitrary conservative external force is presented.
If the mass of the particle is zero and the force is central then the
equations of motion turn out to be completely integrable. A well-known
result\cite{shm}.
\vskip 8pt
Hamiltonian flows on the twistor phase space \bf T \rm are constructed
which, for conservative forces and value of the helicity equal to zero,
reproduce equations of motion of the classical massless particle. For
helicities different from zero the same hamiltonian flows produce
equations of motion showing a curious "Zitterbewegung" like behaviour.
\vskip 8pt
A canonical Poincar{\'e} covariant quantization procedure on \bf T \rm
is suggested. One simple example describing a spinning and massless
3-D quantum mechanical harmonic oscillator is analysed in some detail.
\end{abstract}
\vfill
\eject
\section*{\hspace{-4mm}\hspace{2mm}INTRODUCTION AND NOTATION.}
It is possible that elementary massive particles such as electron,
proton, neutron etc. should be regarded as bound states of a finite
number of massless and spinning interacting constituents.
\vskip 10pt
In this article a mathematical formalism with its roots in the Twistor
Theory of Penrose\cite{prmc} is investigated. The physical force in
the model is external (and conservative). Therefore its future
physical application (if any) aims only at a new phenomenological
attempt to understand where the masses of elementary particles come
from.
\vskip 10pt
The author does not claim that the model (as it now stands) describes
any known physical system. He wishes just to show that there are some
concrete uninvestigated possibilities hidden in the mathematics of
Twistors. The virtue of such models (when fully developed) is the
simplicity and economy of thought they provide.
\vskip 10pt
Earlier, exploring the idea of instantaneous relativistic action at a
distance\cite{wfw} in the phase space of twistors we have
shown\cite{ab1} how a free massive and spinning particle may be
thought of as a relativistic rigid rotator (endowed with intrinsic
spin) composed of two massless spinning parts. Instantaneous refers
to the rest frame defined by the total time-like four-momentum of the
rigid rotator the two massless and spinning particles happen to
define.
\vskip 10pt
By continuing these ideas and by taking a larger number of such
massless constituents more complicated closed massive and spinnig
systems may be constructed\cite{ab2}.
\vskip 10pt
However, with the increasing number of massless parts, calculations,
although st- raightforward, become quite cumbersome.
\vskip 10pt
This fact triggered the work presented in this paper. In order to get
an idea of how a closed system, composed of a large finite number of
massless spinning mutually interacting particles, might behave we
investigate dynamics of just one massless spinning particle acted upon
by an external conservative force. The latter may be thought of as an
effective force coming from an inertial "source" defined by the total
freely moving composite system.
\vskip 10pt
In other words we assume that there exists a special inertial frame in
Minkowski space (the rest frame of the "infinitely" heavy "source" of
the force) to which the massless particle is bounded.
\vskip 10pt
One of the shortcomings of our approach is that there is nothing in
the formalism which tells us how to choose the external force (or
equivalently the corresponding hamiltonian) in order to describe a
physical system. Future investigations will perhaps show how the
external force should be chosen in concrete physical situations.
\vskip 10pt
The work is organized as follows:
\vskip 10pt
In the next section we review the relativistic Newton's
second law of dynamics with emphasis on the massless particle case.
It is demonstrated again that when a conservative and central force acts
on a massless particle then its relativistic equations of motion
are completely integrable\cite{shm}.
\vskip 10pt
In the third section an Hamiltonian mechanics, reproducing equations
of motion of a spinless and massless particle, is formulated on the
phase space of twistors. Further, the approach is generalized to be
valid for non-vanishing values of the helicity.
\vskip 10pt
Finally, in the fourth section a canonical quantization is performed.
One relatively simple example, representing an analog of the 3-D
harmonic oscillator, is studied in some detail.
\vskip 10pt
Latin letters with lower case latin indices denote four-vectors and
four-tensors. Latin letters with lower case greek indices within
brackets denote three-vectors. In section $1$, $2$ the usual
three-vector notation (with a line over a letter) will also be used.
In section $2$ and $3$ a bar (not a line) over a letter or over an
expression denotes complex conjugation. Lower case greek letters with
upper case latin indices (either primed or unprimed) denote spinors.
Upper case latin letters with lower case greek indices denote
twistors. The physical units are so chosen that $c={\hbar}=1$. The
signature of the metric $g_{ij}$ in Minkowski space is taken to be
$+---$. The fully antisymmetric alternating four-tensor will be
denoted by $\eta_{ijkl}$. The fully antisymmetric alternating
three-tensor will be denoted by $\epsilon_{(\alpha)(\beta)(\gamma)}$.
The usual summation convention over repeted indices will be assumed
throughout.
\section{\hspace{-4mm}\hspace{2mm}A SHORT REVIEW OF RELATIVISTIC PARTICLE
DYNAMICS.}
In this paper we intend to achieve two goals. The first is to
describe relativistic classical dynamics of a massless spinning
particle acted upon by an external conservative force in terms of
canonical flows on the twistor phase space. The second is to
formulate the corresponding relativistic quantum dynamics and examine
the case when the force is chosen to be of the 3-D harmonic oscillator
type.
\vskip 10pt
To define the context we first review the special relativistic version
of Newton's second law of dynamics in general and its massless limit
in particular.
\vskip 10pt
For a massive particle, Newton's second law of dynamics
may be written in the following Poincar{\'e} covariant form:
\begin{equation} {dY^{i} \over d{\tau}} = {P^{i} \over m}, \end{equation}
\begin{equation}{dP_{i} \over d{\tau}} = F_{i} \qquad \qquad where \qquad
\qquad
F^{i}P_{i}=0.\end{equation}
\vskip 10pt
$Y^{i}$ is the four-position of the particle in Minkowski space,
$P_{i}$ denotes its four- momentum, $\tau$ its proper time, $m$ its
rest mass and $F_{i}$ represents the so called four-force acting on
the particle. Suppose that the world-line of an inertial frame is
given by:
\begin{equation}X^{i} = X_{0}^{i} + (t-t_{0})t^{i},\end{equation}
\vskip 10pt
\noindent
where $X^{i}$ denotes its four-position, $t^{i}$ its constant
time-like four-velocity, $t$ its proper time and where $X_{0}^{i}$
represents a constant four-vector starting from some arbitrarily
chosen origin in Minkowski space and ending at a point (an event) on
the world-line of the inertial frame where we have put the value of
its proper time to be equal to $t_{0}$.
\vskip 15pt
Then, a space-like four-vector $r^{i}$, which represents the
particle's instantaneous \it centre of energy \rm
with respect to the inertial frame defined by (1.3), is given by:
\begin{equation}r^{i}:= R^{i} - (R^{k}t_{k})t^{i},\end{equation}
\vskip 10pt
\noindent
where
\begin{equation}R^{i} = Y^{i} - X^{i}.\end{equation}
\vskip 10pt
\noindent
Using (1.3) it yields:
\begin{equation}r^{i} = (Y^{i} - X_{0}^{i}) -
[(Y^{k}-X_{0}^{k})t_{k}]t^{i}.\end{equation}
\vskip 10pt
Let us also introduce a space-like four-vector $f_{k}$ which fulfils:
\begin{equation} f_{k}t^{k} = 0, \end{equation}
\vskip 10pt
\noindent
and represents the physical three-force, exerted on the particle under
consideration, as "measured" by an observer following the world-line of the
inertial frame, where
\begin{equation}-{f_{n}P^{n} \over P_{k}t^{k}},\end{equation}
\vskip 10pt
\noindent
represents the work done by this force. The four-force in (1.2) may
now be split into two components\cite{rw}. One along the four-velocity $t^i$
and another projected onto the space-like three-plane orthogonal to
$t^i$:
\begin{equation}F_{i}={P_{k}t^{k} \over m} f_{i} - {f_{k}P^{k} \over m}
t_{i}.\end{equation}
\vskip 15pt
The equations of motion in (1.1) and (1.2) may then be rewritten as
follows:
\begin{equation}{\dot r}^{i} = {P^{i}\over P^{k}t_{k}} - t^{i},\end{equation}
\begin{equation}{\dot P}_{i} - ({\dot P}_{k}t^{k})t_{i} = f_{i}
\qquad and \qquad {\dot P}_{i}{P}^{i}=0.\end{equation}
\vskip 10pt
The dot over a letter denotes differentiation with respect
to the proper time in the inertial frame defined by (1.3). If $m$
differs from zero this is just a reformulation of the equations given
in (1.1) and (1.2). However, in contrast to (1.1) and (1.2), the
equations in (1.10) and (1.11) are also valid for $m = 0$.
The physical three force represented by $f_{k}$ may
depend functionally on $r^{i}$ i.e.:
\begin{equation}f_{k} = f_{k}(r^{i}).\end{equation}
\vskip 10pt
The \it centre of energy \rm space-like four-vector $r^{i}$ depends on the
location of the inertial frame and also on the location of the
particle in Minkowski space. Therefore the assumption in (1.12)
implies that the inertial frame, with the world-line given by (1.3),
is not arbitrary but constitutes a source producing the force
acting on the particle.
In the inertial frame of the source we have:
\begin{equation}t^{i}=(1, {\overline 0}),\end{equation}
\begin{equation}r^{i}=(0, {\overline r}),\end{equation}
\begin{equation}f^{i}(r^{k})=(0, {\overline f}(\overline r)),\end{equation}
\vskip 10pt
\noindent
where we have used the familiar three-vector notation.
The equations of motion in (1.10) and (1.11) now read:
\begin{equation}{\dot {\overline r}} = {{\overline p} \over E},\end{equation}
\begin{equation}{\dot {\overline p}} = {\overline f}({\overline
r}).\end{equation}
In addition we also have:
\begin{equation}m^{2}:=P^{i}P_{i} = E^{2} - {\vert {\overline p} \vert}^{2} =
E^{2} - p^{2}=constant.\end{equation}
\vskip 15pt
If the force ${\overline f}({\overline r})$ is conservative then one
has:
\begin{equation}{\overline f} = - {\nabla U}({\overline r}),\end{equation}
\vskip 10pt
\noindent
where $U({\overline r})$ represents the potential energy of the particle.
{}From (1.16)-(1.19) we obtain in the usual manner that:
\begin{equation}{\dot E} = - [{\dot {\overline r}} {\cdot } {\nabla
U({\overline r})}],\end{equation}
\vskip 10pt
\noindent
which implies the energy conservation law:
\begin{equation}H := E + U({\overline r}) = constant.\end{equation}
\vskip 10pt
The constant $H$ represents the total energy of the particle.
Differentiating (1.16), using (1.17), (1.20) and (1.21) yields:
\begin{equation}(H-U){\ddot {\overline r}} + {\nabla U({\overline r}}) - {\dot
{\overline r}}
[{\dot {\overline r}}{\cdot }{\nabla U({\overline r})}] = 0.\end{equation}
\vskip 10pt
Again,
if $m$ differs from zero and the non-relativistic condition is
fulfilled i.e.:
\begin{equation}{\vert {\dot {\overline r}} \vert} << 1,\end{equation}
\noindent
then one has that:
\begin{equation}(H-U) {\simeq } m,\end{equation}
\vskip 10pt
\noindent
which implies that the equations of
motion in (1.22) above acquire (as they of course
should) the familiar Newtonian form:
\begin{equation}m{\ddot {\overline r}} + {\nabla U({\overline r}}) {\simeq }
0.\end{equation}
\vskip 15pt
\it
Note that the relativistic non-linear equation in (1.22) is also valid
for massless particles.
\rm
\vskip 15 pt
To proceed further we assume that the force is also central i.e. that:
\begin{equation}U({\overline r}) = U({\vert {\overline r} \vert}) =
U(r).\end{equation}
\vskip 10pt
{}From (1.16), (1.17), (1.21) and (1.26) we then obtain (as in the
non-relativistic case) that the orbital angular momentum
does not change in time:
\begin{equation}{\overline L} := {\overline r} {\times } {\overline p} =
[H-U(r)] [{\overline r} {\times } {\dot {\overline r}}] =
constant,\end{equation}
\vskip 10pt
\noindent
implying plane particle motion.
${\times }$ denotes the usual
vector product in the three dimensional space (Lorentz) "perpendicular"
to the time-like direction given by $t^{i}$.
\vskip 10 pt
Choosing the space origin at the location of the
observer who follows the world-line of the inertial frame
and the $z$-axis along the three-vector ${\overline L}$ the equations
of motion, in the familiar polar coordinates on the
spatial plane perpendicular to the $z$-axis, read:
\begin{equation}{\ddot \rho} - {U'\over (H-U)}({\dot \rho}^{2} - 1)-
{L^{2} \over {{(H-U)}^{2}\rho^{3}}} = 0,\end{equation}
\begin{equation}{\dot \phi}={L \over {(H-U) \rho^{2}}} \qquad \qquad \qquad
\qquad
{\dot z} = z = 0,\end{equation}
\vskip 10pt
\noindent
where
\begin{equation}L:={{\vert {\overline L} \vert}},\end{equation}
\vskip 10pt
\noindent
and where
\begin{equation}U':={dU \over d{\rho}}.\end{equation}
\vskip 10pt
In the massless case, which is our main concern here, we have that:
\begin{equation}{\vert {\overline p} \vert}=E,\end{equation}
\vskip 10pt
\noindent
which, by the use of (1.16), implies:
\begin{equation}{{\vert {\dot {\overline r}} \vert}}=1.\end{equation}
\vskip 10pt
In the introduced polar coordinates this yields:
\begin{equation}{\dot {\rho}}^{2} + {\rho}^{2} {\dot \phi}^{2} =
1,\end{equation}
\vskip 10pt
\noindent
which inserted into the first part of (1.29) reduces the equations of
motion to a simple expression:
\begin{equation}{\dot {\rho}}^{2} = 1 - {L^{2} \over {(H-U)^{2}
{\rho}^{2}}}.\end{equation}
\vskip 10pt
The equation above may be integrated giving $t$ as a function of $\rho$:
\begin{equation}t={\pm }{{\int}
{d\rho \over {\sqrt {1 - {L^{2} \over {(H-U)^{2} \rho^{2}}}}}}},\end{equation}
\vskip 10pt
\noindent
and for the polar angle $\phi$ as a function of $\rho$ one obtains:
\begin{equation}{\phi}={\pm }L{{\int} {d\rho \over {(H-U)\rho^{2} {\sqrt {1 -
{L^{2} \over
{(H-U)^{2} \rho^{2}}}}}}}}.\end{equation}
\vskip 15pt
\it
The equations of motion, describing massless particles acted upon by
central conservative forces, are completely integrable\cite{shm}.
\rm
\vskip 15pt
The above results may be understood in terms of symplectic (phase
space) geometry. The natural symplectic structure
on the cotangent bundle $T^*M$ (eight dimensions) of the Minkowski
space $M$ is given by:
\begin{equation} \Omega:=dP_{i}\wedge dY^{i}. \end{equation}
\vskip 10pt
The Lorentz covariant and four-translation invariant coordinates of
the momentum four-vector $P_{i}$ and the Poincar{\'e} covariant
coordinates of the position four-vector $Y^i$ regarded as functions on
$T^*M$ are canonically conjugated to each other. The ten Poincar{\'e}
covariant functions on $T^{*}M$ given by:
\begin{equation} P_{i} \qquad \qquad and \qquad \qquad M^{ij}:=2Y^{[i}P^{j]}
\end{equation}
\vskip 10pt
\noindent
define the so called momentum mapping for the action of the
Poincar{\'e} group on $T^{*}M$. In other words the algebra of their
Poisson brackets represents the commutation relations of the
Poincar{\'e } algebra.
\vskip 10pt
The particle's instantaneous position relative to the inertial source,
which is a uniformly moving point in $M$, resides on space-like planes
given by:
\begin{equation} Y^{i}t_{i}=0. \end{equation}
\vskip 10pt
If $M^{ij}$ is taken about this uniformly moving point in $M$ then the
position of the particle's \it centre of energy \rm is given by:
\begin{equation} r^{i}={M^{ij}t_{j} \over P^{m}t_{m}}. \end{equation}
\vskip 10pt
The equations of motion in (1.22) arise as a consequence of the
canonical flow in $T^{*}M$ (canonical with respect to the symplectic
structure $\Omega$ in (1.38)) generated by $H$ in (1.21) treated as a
real valued function on $T^{*}M$.
\vskip 10pt
Note that the flow generated by the mass squared function defined
in (1.18) commutes with the flow generated by the function $H$.
$m^{2}$ is therefore a constant of the physical motion.
\vskip 10pt
In the next section we are going to show that, for a massless particle,
equations of motion in (1.22) may be regarded in a completely
different way. They will arise as a consequence of the canonical flow
generated by $H$ in (1.21) treated as a real valued function on the
twistor phase space.
\section{\hspace{-4mm}\hspace{2mm}TWISTOR PHASE SPACE FORMULATION FOR $m=0$.}
The pair $(X^{A{A^\prime}}_{0}, \ t^{A{A^\prime}})$ (here we adopt
Penrose's abstract index notation\cite{pr1}) defining the straight
world-line (of the inertial source producing the external force acting
on the massless particle) in (1.3) may be represented by two fixed
intersecting null-twistors $V^{\alpha}$, $W^{\alpha}$ in \bf T \rm (and their
twistor complex conjugates ${\bar V}_{\alpha}$, ${\bar W}_{\alpha}$)
i.e. twistors which fulfil:
\begin{equation}V^{\alpha}\bar V_{\alpha} = W^{\alpha}{\bar W}_{\alpha} =
V^{\alpha}{\bar W}_{\alpha} = 0,\end{equation}
\begin{equation}I_{\alpha \beta}V^{\alpha}W_{\beta} = {1 \over
\sqrt2},\end{equation}
\vskip 10pt
\noindent
where $I_{\alpha \beta}$ is the infinity twistor\cite{prmc,prrw}. Using the
spinor
representation this yields:
\begin{equation}V^{\alpha} = (\sigma^{A},\ \alpha_{A^\prime}),
\qquad \qquad \qquad
{\bar V}_{\alpha} = ({\bar \alpha_{A}},\ {\bar
\sigma}^{A^\prime}),\end{equation}
\begin{equation}W^{\alpha} = (\varsigma^{A},\ \beta_{A^\prime}),
\qquad \qquad \qquad
{\bar W}_{\alpha} = ({\bar \beta_{A}},\ {\bar
\varsigma}^{A^\prime}).\end{equation}
\vskip 10pt
Therefore the relation in (2.2) may be written as:
\begin{equation}I_{\alpha \beta}V^{\alpha}W^{\beta} =
\alpha^{A^\prime}\beta_{A^\prime}=
{1 \over \sqrt2},\end{equation}
\vskip 10pt
\noindent
while $X^{A{A^\prime}}_{0}$ is explicitly given by (see Ref. 2):
\begin{equation}X^{A{A^\prime}}_{0}:=i{\sqrt2}(\sigma^{A} \beta^{A^\prime}
- \varsigma^{A} \alpha^{A^\prime})=
-i{\sqrt2}({\bar \sigma}^{A^\prime}{\bar \beta}^{A} -
{\bar \varsigma}^{A^\prime}{\bar \alpha}^{A}),\end{equation}
\vskip 10pt
\noindent
and $t^{A{A^\prime}}$ by:
\begin{equation}t^{A{A^\prime}}:={\alpha^{A^\prime}{\bar \alpha}^{A} +
\beta^{A^\prime}{\bar \beta}^{A}}.\end{equation}
\vskip 10pt
The two fixed twistors $V$ and $W$ also define three space-like
directions rigidly attached to the inertial frame defined in (1.3):
\begin{equation}u^{a}_{(1)} \ {\equiv}\ u^{a}\ {\equiv}\ u^{A{A^\prime}}:=
\alpha^{A^\prime}{\bar \alpha}^{A} -
\beta^{A^\prime}{\bar \beta}^{A},\end{equation}
\begin{equation}u^{a}_{(2)} \ {\equiv}\ v^{a}\ {\equiv}\
v^{A{A^\prime}}:=\alpha^{A^\prime}{\bar \beta}^{A} +
\beta^{A^\prime}{\bar \alpha}^{A},\end{equation}
\begin{equation}u^{a}_{(3)}\ {\equiv}\ w^{a}\ {\equiv}\
w^{A{A^\prime}}=:i(\alpha^{A^\prime}{\bar \beta}^{A} -
\beta^{A^\prime}{\bar \alpha}^{A}).\end{equation}
\vskip 10pt
We may call the direction defined by $u^{a}$ the $z$-axis direction,
the direction defined by $v^{a}$ the $x$-axis direction and the
direction defined by $w^{a}$ the $y$-axis direction.
\vskip 10pt
In effect $u^{a}_{(\alpha)}$ and $t^{a}$ form a non-rotating fixed
tetrad rigidly attached to the inertial source.
\vskip 10pt
Coordinates of a variable point in \bf T \rm (i.e. a twistor) $Z^{\alpha}$ and
coordinates of its (twistor) complex conjugate point ${\bar Z}_{\alpha}$
will be represented by two variable Weyl spinors and their conjugates:
\begin{equation}Z^{\alpha} = (\omega^{A},\ \pi_{A^\prime}),
\qquad \qquad
{\bar Z}_{\alpha} = ( {\bar \pi_{A}},\ {\bar
\omega}^{A^\prime}).\end{equation}
\vskip 10pt
This spinor representation of a point in \bf T \rm is very convenient because
it shows explicitly how the Poincar{\'e} group acts on \bf T. \rm Coordinates
of the two spinors represented by $\pi_{A^\prime}$ and $\omega^{A}$
are covariant with respect to the (identity connected part of the)
Lorentz group while four-translations $T^{a}$ act only on the
"$\omega$" spinor parts of the twistor
$Z$ and its (twistor) complex conjugate $\bar Z$
according to the
following simple rule\cite{prmc}:
\begin{equation}{\breve {\omega}}^{A}={\omega}^{A} +
iT^{A{A^\prime}}{\pi}_{A^\prime},
\qquad \qquad \qquad
{{\bar {\breve {\omega}}}}^{A^\prime}=
{{\bar {\omega}}}^{A^\prime} - iT^{A{A^\prime}}{\bar \pi}_{A}.\end{equation}
\vskip 10pt
(Do not confuse $(\alpha)$ indices which label the three mutually
orthogonal physical space directions with $\alpha$ indices which label
twistors.)
\vskip 15 pt
The natural SU(2,2) invariant symplectic structure\cite{pr2,prrw,hl} on \bf T
\rm
defines the following canonical Poisson bracket relations:
\begin{equation}\{Z^{\alpha},\ {\bar Z}_{\beta}\} =
i{\delta}^{\alpha}_{\beta},
\qquad \qquad \qquad
\{Z^{\alpha},\ Z^{\beta}\} = \{{\bar Z}_{\alpha},\ {\bar Z}_{\beta}\} =
0,\end{equation}
\vskip 10pt
\noindent
which in terms of spinor variables reads:
\begin{equation}\{\omega^{A},\ {\bar \pi_{B}}\} =
i{\delta}^{A}_{B},
\qquad \qquad \qquad
\{\pi_{B^\prime},\ {\bar \omega}^{A^\prime}\} =
i{\delta}^{A^\prime}_{B^\prime},\end{equation}
\begin{equation}\{\omega^{A},\ \omega^{B}\} = \{\omega^{A},\ \pi_{A^\prime}\}=
\{\pi_{A^\prime},\ \pi_{B^\prime}\} =
\{\pi_{A^\prime},\ {\bar \pi}_{B}\} = 0,\end{equation}
\begin{equation}\{{\bar \omega}^{A^\prime},\ {\bar \omega}^{B^\prime}\}=
\{{\bar \omega}^{A^\prime},\ {\bar \pi_{A}}\} =
\{{\bar \pi_{A}},\ {\bar \pi_{B}}\} =
\{{\omega}^{A},\ {\bar \omega}^{B^\prime}\} = 0.\end{equation}
\vskip 15pt
The linear
four-momentum $P_{a}$ and the angular four-momentum
${ M}_{ab} = - { M}_{ba}$
of a massless particle may be regarded as given\cite{prmc}
by the following set of Poincar{\'e} covariant functions on \bf T: \rm
\begin{equation}P_{a}:= \pi_{A^\prime}{\bar \pi}_A,\end{equation}
\begin{equation}{ M}_{ab}:= i{\bar
\omega}_{(A^\prime}{\pi}_{B^{\prime})}{\epsilon}_{AB}-i{\omega}_{(A}{\bar
\pi}_{B)} {\epsilon}_{A^{\prime}B^{\prime}}.\end{equation}
\vskip 10pt
The canonical Poincar{\'e} covariant
Poisson brackets in (2.13)-(2.16) imply that $P_{a}$ and
${ M}_{ab}$ in (2.17)-(2.18) fulfil the Poisson bracket
relations of the Poincar{\'e} algebra\cite{hl}:
\begin{equation}\{P_{a},\ P_{b}\}=0,\end{equation}
\begin{equation}\{{ M}_{ab},\ P_{c}\}=2g_{c[a}P_{b]},\end{equation}
\begin{equation}\{{ M}_{ab},\ { M}_{cd}\}=
2(g_{c[a}{ M}_{b]d}) + g_{d[b}{ M}_{a]c}).\end{equation}
\vskip 10pt
The Poisson bracket relations in (2.19) - (2.21) define the momentum
mapping for the action of the Poincar{\'e} group on \bf T. \rm
\vskip 10pt
The Pauli-Luba{\'n}ski spin
four-vector of a massless particle:
\begin{equation}S^{a}:={1 \over 2}{\eta}^{abcd}P_{b}{ M}_{cd}\end{equation}
\vskip 10pt
\noindent
reduces itself to a simple expression\cite{prmc}
(use spinor representation\cite{prrw}
of ${\eta}^{abcd}$ to prove it):
\begin{equation}S^{a}=sP^{a},\end{equation}
\vskip 10pt
\noindent
where
\begin{equation}s = {1 \over2}
(Z^{\alpha}\bar Z_{\alpha}) =
{1 \over2}({\omega^{A}}{\bar \pi_{A}} +
{\pi_{A^\prime}}{\bar \omega^{A^\prime}}).\end{equation}
\vskip 10pt
The kinetic energy of the massless particle, in the special inertial
frame defined by the source, will be defined by the following
function on \bf T. \rm
\begin{equation}E := \pi_{C^{\prime}}{\bar
\pi}_{C}t^{C{C^\prime}},\end{equation}
\vskip 10pt
\noindent
while the linear three-momentum will be given by:
\begin{equation}p_{(\alpha)} := -\pi_{C^{\prime}}{\bar
\pi}_{C}u^{C{C^\prime}}_{(\alpha)}.\end{equation}
\vskip 10pt
If ${M}_{ab}$ is taken about the inertial source producing the force
(i.e. about a uniformly moving point in Minkowski space $M$) then the
functions representing angular momentum of the massless particle about
this source, are given by:
\begin{equation}J_{(\alpha)} := {1 \over 2}\epsilon_{(\alpha)(\gamma)(\delta)}
{ M}_{ab}u_{(\gamma)}^{a}u_{(\delta)}^{b},\end{equation}
\vskip 10pt
\noindent
where
\begin{equation}\epsilon_{(\alpha)(\beta)(\gamma)}:=\eta_{abcd}t^{a}u^{b}_{(\alpha)}
u^{c}_{(\beta)}u^{d}_{(\gamma)}.\end{equation}
\vskip 10pt
\noindent
while the three functions representing the position of the \it centre
of energy \rm relative to the inertial source are given by:
\begin{equation}{ y}_{(\alpha)}:=-{{ M}_{ab}t^{b}u^{a}_{(\alpha)}
\over P_{c}t^{c}}.\end{equation}
\vskip 10pt
Explicitly for the three components of the total angular momentum
it yields:
\begin{equation}J_{z}=J_{(1)}=-[{1 \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \beta}^{B} +
{\bar \beta}_{A}{\bar \alpha}^{B}){\bar \pi}_{B}\omega^{A}
+ {1 \over \sqrt{2}}
(\alpha_{A^\prime}\beta^{B^\prime} +
\beta_{A^\prime}\alpha^{B^\prime})\pi_{B^\prime}{\bar \omega}^{A^\prime}],
\end{equation}
\vskip 10pt
\begin{equation}J_{x}=J_{(2)}={1 \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \alpha}^{B}-{\bar \beta}_{A}{\bar \beta}^{B})
{\bar \pi}_{B}\omega^{A}
+{1 \over \sqrt{2}}
(\alpha_{A^\prime}\alpha^{B^\prime}-\beta_{A^\prime}\beta^{B^\prime})
\pi_{B^\prime}{\bar \omega}^{A^\prime},\end{equation}
\vskip 10pt
\begin{equation}J_{y}=J_{(3)}={-i \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \alpha}^{B} + {\bar \beta}_{A}{\bar \beta}^{B})
{\bar \pi}_{B}\omega^{A}
+{i \over \sqrt{2}}
(\alpha_{A^\prime}\alpha^{B^\prime} + \beta_{A^\prime}\beta^{B^\prime})
\pi_{B^\prime}{\bar \omega}^{A^\prime},\end{equation}
\vskip 10pt
\noindent
and for the three components of the position vector of the \it centre
of energy \rm one obtains:
\begin{equation}{z}=y_{(1)}=
{-i \over \pi_{C^{\prime}}{\bar \pi}_{C}t^{C{C^\prime}}}
[{1 \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \beta}^{B} +
{\bar \beta}_{A}{\bar \alpha}^{B}){\bar \pi}_{B}\omega^{A}
-{1 \over \sqrt{2}}
(\alpha_{A^\prime}\beta^{B^\prime} +
\beta_{A^\prime}\alpha^{B^\prime})\pi_{B^\prime}{\bar \omega}^{A^\prime}],
\end{equation}
\vskip 10pt
\begin{equation}{x}=y_{(2)}=
{-i \over \pi_{C^{\prime}}{\bar \pi}_{C}t^{C{C^\prime}}}
[{1 \over \sqrt{2}}
({\bar \beta}_{A}{\bar \beta}^{B}-{\bar \alpha}_{A}{\bar \alpha}^{B})
{\bar \pi}_{B}\omega^{A}
-{1 \over \sqrt{2}}
(\beta_{A^\prime}\beta^{B^\prime}-\alpha_{A^\prime}\alpha^{B^\prime})
\pi_{B^\prime}{\bar \omega}^{A^\prime}],\end{equation}
\vskip 10pt
\begin{equation}{y}=y_{(3)}=
{1 \over \pi_{C^{\prime}}{\bar \pi}_{C}t^{C{C^\prime}}}
[{1 \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \alpha}^{B} + {\bar \beta}_{A}{\bar \beta}^{B})
{\bar \pi}_{B}\omega^{A}
+{1 \over \sqrt{2}}
(\alpha_{A^\prime}\alpha^{B^\prime} + \beta_{A^\prime}\beta^{B^\prime})
\pi_{B^\prime}{\bar \omega}^{A^\prime}].\end{equation}
\vskip 10pt
The helicity $s$ in (2.24) is a conformal scalar and thereby also a
Poincar{\'e} scalar function on \bf T. \rm Therefore the function $s$ Poisson
commutes with all the functions introduced in (2.30)-(2.35).
\vskip 10pt
Using (2.19)-(2.21) one obtains following Poisson bracket relations:
\begin{equation}\{{ y}_{(\alpha)},\ { y}_{(\beta)}\} =
{{s \epsilon_{(\alpha)(\beta)(\gamma)}p_{(\gamma)}} \over E^3},\end{equation}
\begin{equation}\{p_{(\beta)},\ { y}_{(\alpha)}\} =
\delta_{(\alpha)(\beta)},\end{equation}
\begin{equation}\{E,\ { y}_{(\alpha)}\} = {p_{(\alpha)} \over
E},\end{equation}
\begin{equation}\{E,\ p_{(\beta)}\} = \{E,\ J_{(\alpha)}\} = 0,\end{equation}
\begin{equation}\{J_{(\alpha)}, \ J_{(\beta)}\}=
\epsilon_{(\alpha)(\beta)(\gamma)}J_{(\gamma)},\end{equation}
\begin{equation}\{J_{(\alpha)}, \ { y}_{(\beta)}\} =
\epsilon_{(\alpha)(\beta)(\gamma)}
{ y}_{(\gamma)},\end{equation}
\begin{equation}\{J_{(\alpha)}, \ p_{(\beta)}\} =
\epsilon_{(\alpha)(\beta)(\gamma)}
p_{(\gamma)}.\end{equation}
\vskip 10pt
The above commutation relations are quite reasonable from the physical
point of view. Apart from (2.36) they are what one should expect.
\vskip 10pt
The energy $H$ in (1.21) (for $m=0$) may now be treated as a function
on \bf T. \rm The canonical flow which $H$ generates on \bf T \rm is then
explicitly
given by:
\begin{equation}\dot \omega^{A} = \{H, \ \omega^{A}\} = -i{{\partial H} \over
{\partial
{{\bar \pi}_{A}}}},\end{equation}
\begin{equation}{\dot \pi}_{B^\prime}= \{H, \ \pi_{B^\prime}\} =
- i {{\partial H} \over {\partial {\bar \omega}^{B^\prime}}}.\end{equation}
\vskip 10pt
For functions describing physical variables in (2.27)-(2.35)
it yields:
\begin{equation}{\dot { y}}_{(\alpha)} =\{H,\ { y}_{(\alpha)}\} =
\{E+U,\ { y}_{(\alpha)}\}= {p_{(\alpha)} \over E}-
{{s \epsilon_{(\alpha)(\beta)(\gamma)}p_{(\gamma)}} \over E^3}
{{\partial U} \over {\partial { y}_{(\beta)}}} ,\end{equation}
\begin{equation}{\dot p}_{(\alpha)} = \{E+U,\ p_{(\alpha)}\} = \{U,\
p_{(\alpha)}\}
= {{\partial U} \over {\partial { y}_{(\beta)}}}
\{{ y}_{(\beta)},\ p_{(\alpha)}\}
= - {{\partial U} \over {\partial {y}_{(\alpha)}}},\end{equation}
\vskip 10pt
\noindent
which for $s=0$ implies the equations of motion in (1.22). For $s$ not
being equal to zero the above description is a generalization of the
massless phase space dynamics. Due to (2.36) the velocity of the \it
centre of energy \rm and the velocity of the massless spinning
particle cease to define the same physical quantity. This indicates
the extended nature of the massless spinning particle. One possible
interpretation of this Zitterbewegung-like behaviour is presented in
Ref. 11. Another is that, when external forces are acting on a spinning
massless particle, the direction of motion of the centre of its \it
entire \rm kinetic energy (including energy generated by the helicity)
simply ceases to be parallell with the (null-) direction of motion of
the centre of its \it translational \rm kinetic energy.
Relative to the rest frame of the source the velocity of the centre of
\it entire \rm kinetic energy of the interacting massless spinning
particle exceeds the velocity of light. The velocity of the centre
of its translational kinetic energy remains on the other hand always
equal to the velocity of light. The two velocities coincide when the
massless spinning particle is moving freely.
\vskip 10pt
Using arguments having their origins in symplectic geometry we will
now demonstrate that for central forces, irrespective whether the
helicity is equal to zero or not, equations in (2.45)-(2.46) are
completely integrable.
\vskip 10pt
First we note that on the 8-D twistor phase space \bf T \rm the energy
function $H$ Poisson commutes with the helicity function $s$. For a
fixed value of $s$ the energy function $H$ may therefore be regarded
as a function on the 6-D phase space obtained as a reduction of \bf T \rm
by the function $s$. In other words, points in the reduced 6-D phase
space are represented by these curves on \bf T \rm of the hamiltonian
flow generated by $s$ which lie on the shell-surface given by a fixed
value of the function $s$.
\vskip 10pt
In the same way components of the total angular momentum $J_{(\alpha)}$
which Poisson commute with the helicity function $s$ on \bf T \rm
may be regarded as functions on the reduced 6-D phase space.
For central forces they also commute with the energy function $H$.
So they also commute with $H$ on the reduced 6-D phase space.
\vskip 10pt
Any of the components of the total angular momentum $J_{(\alpha)}$ say
$J_{(1)}$, the square of the total angular momentum
$J^{2}:=J_{(\alpha)}J_{(\alpha)}$ and $H$ are mutually Poisson
commuting functions on the 6-D reduced phase space
\vskip 10pt
Reduction of the 6-D phase space by the two mutually Poisson commuting
functions $J_{(1)}$ and $J^{2}$ produces, for each fixed value of these
functions, a 2-D phase space. The energy function $H$ may now be regarded
as a function on this 2-D phase space. But
equations of motion on a 2-D phase space
are always completely integrable.
\vskip 10pt
The somewhat unexpected result\cite{shm} reproduced in the
introduction, proving the complete integrability of the equations of
motion of a massless and spinless particle moving under the action
of a conservative and central force, thus turns out to be a special
case of the general feature as described above.
\vskip 10pt
Consequently, for external central forces, there exists a general solution of
the equations of motion in (2.45)-(2.46). For $s=0$ this solution
reduces itself to that presented in (1.36)-(1.37). To find the general
solution by means of quadratures we proceed as follows. In the rest
frame of the source we reintroduce the standard vector notation.
\vskip 10pt
We recall that the translational kinetic energy of the massless
spinning particle is denoted by:
\begin{equation}
E=\vert {\overline p} \vert,
\end{equation}
\vskip 10pt
\noindent
the distance from the source to the centre of the
entire kinetic energy of the particle by:
\begin{equation}
r=\vert {\overline r} \vert=\sqrt{{\rho}^{2}+z^{2}}
\end{equation}
\vskip 10pt
\noindent
($\rho$ and $z$ are plane polar coordinates of the corresponding
position vector) and the total energy of the massless and spinning
particle by:
\begin{equation}
H=E+U(r).
\end{equation}
\vskip 10pt
For later convenience we introduce a function defined by:
\begin{equation}
f(r):=-r{dU \over dr},
\end{equation}
\vskip 10pt
\noindent
while we also note that in standard vector notation one has:
\begin{equation}
{\overline J}=
{\overline r} \times {\overline p} + {s{\overline p} \over E}
\end{equation}
\vskip 10pt
\noindent
where ${\overline J}$ is the total angular momentum vector of the
massless spinning particle and where ${\overline p}$ is its linear momentum
vector.
\vskip 10pt
In three vector notation the equations of motion in (2.45)-(2.46)
now read:
\begin{equation}
{\dot {\overline p}}={f(r) \over r^{2}}{\overline r}
\end{equation}
\begin{equation}
{\dot {\overline r}}={{\overline p} \over E}
- {s \over E^{3}}({\overline r} \times {\overline p})
{f(r) \over r^{2}}
\end{equation}
\vskip 10pt
Standard calculations also give:
\begin{equation}
({\overline r} \cdot {\overline p})^{2}=E^{2}r^{2}+s^{2}-J^{2}.
\end{equation}
\vskip 10pt
Choosing the direction of the constant total angular momentum
along the positive direction of the $z$ axis implies:
\begin{equation}
{\overline J}=
J{\overline e}_{z}.
\end{equation}
\vskip 10pt
\noindent
(2.55), (2.51) and (2.54) then yield:
\begin{equation}
z=\pm {rs \over J} \sqrt{1+{s^{2}-J^{2} \over E^{2}r^{2}}}.
\end{equation}
\vskip 10pt
{}From (2.52) follows that:
\begin{equation}
{\dot E}=
{f(r)({{\overline r} \cdot {\overline p}}) \over r^{2}E},
\end{equation}
\vskip 10pt
\noindent
while from (2.49) follows that:
\begin{equation}
{\dot E}=
-{dU \over dr}{\dot r}.
\end{equation}
\vskip 10pt
Using (2.50), (2.54), (2.57), (2.58) we finally obtain:
\begin{equation}
{\dot r}=\sqrt{1+{s^{2}-J^{2} \over E^{2}r^{2}}}.
\end{equation}
\vskip 10pt
For $s=0$ the above results imply that $z=0$ and that (2.59) is
equivalent to (1.36).
\vskip 10pt
Using (2.51), (2.55) one obtains:
\begin{equation}
p_{\varphi}={{E^{2} \rho J} \over {s^{2} + r^{2} E^{2}}},
\end{equation}
\begin{equation}
p_{z}={{EJ(s^{2} +z^{2}E^{2})} \over {s(s^{2} + r^{2} E^{2})}},
\end{equation}
\begin{equation}
p_{\rho}={{E^{3} J z \rho} \over {s(s^{2} + r^{2} E^{2})}}.
\end{equation}
\vskip 10pt
The above results and the equation of motion in (2.53) yield:
\begin{equation}
{\dot \varphi}={EJ \over {s^{2}+r^{2}E^{2}}}(1+{s^{2}f(r) \over r^{2}E^{2}})
\end{equation}
\vskip 10pt
\noindent
which for $s=0$ imply (1.37).
\vskip 10pt
For specific choices of the potential energy in (2.49) the above
general equations may easily be integrated and plotted by the use of
modern computer programs such as e.g. Maple V, Release 3. The
concrete results of such calculations we hope to be able to present in
a future publication.
\section{\hspace{-4mm}\hspace{2mm}MASSLESS RELATIVISTIC QUANTUM DYNAMICS.}
In this section a quantization of the twistor phase space dynamics is
suggested. The energy eigenvalue equation corresponding to the
potential of the 3-D harmonic oscillator is studied in some detail.
\vskip 10 pt
A non-standard, as opposed to the standard procedure introduced by
Penrose\cite{prmc}, canonical twistor quantization is obtained by
means of a natural prescription {\'a} la Dirac\cite{dpam1,dpam2}
given by:
\begin{equation}{\hat \omega}^{A} :=
- {{\partial} \over {\partial {\bar \pi}_{A}}}, \qquad \qquad
{\hat {\bar \omega}}^{A^\prime} := {{\partial} \over {\partial
{\pi}_{A^\prime}}},\end{equation}
\begin{equation}{\hat {\bar \pi}}_{A} := {\bar \pi}_{A}, \qquad \qquad
{\hat {\pi}}_{A^\prime} := {\pi}_{A^\prime}.\end{equation}
\vskip 10pt
The Poisson brackets relations in (2.14) - (2.16) will hereby be
replaced by the corresponding commutators turning the classical
twistor phase space dynamics of a massless particle into its quantum
mechanical analog.
\vskip 10pt
So by the use of (3.1)-(3.2) the linear four-momentum functions in
(2.17), the angular four-momentum functions in (2.18), the helicity
function in (2.24) turn into the corresponding operators:
\begin{equation}{\hat P}_{a}:={\bar \pi}_{A}{\pi}_{A^\prime},\end{equation}
\begin{equation} {\hat M}^{ab}:=
i{\pi}^{({A^\prime}}
{{\partial} \over {\partial
{\pi}_{{B^\prime})}}}\epsilon^{AB}
+
i{\bar \pi}^{({A}}{{\partial} \over {\partial {\bar \pi}_{{B)}}}}
\epsilon^{{A^\prime}{B^\prime}},
\end{equation}
\begin{equation}{\hat s}: =
-{1 \over 2}
({\bar \pi}_{A}{\partial \over \partial {\bar \pi}_{A}}-
\pi_{A^\prime}{\partial \over \partial \pi_{A^\prime}}).\end{equation}
\vskip 10pt
The Poisson bracket relations in (2.19)-(2.21) ensure that operators
in (3.3) and (3.4) obey commutation relations of the Poincar{\'e}
algebra.
In this sense the quantization suggested in (3.1)-(3.2) is
Poincar{\'e} covariant. It is not conformally covariant, however.
Besides,
the helicity operator in (3.5)
commutes with all the operators
in (3.3) and (3.4).
\vskip 10pt
Reassuming, the helicity function $s$ in (2.24), the three
functions representing components of the angular momentum in
(2.30)-(2.32), the three functions representing components of the
position vector of the \it centre of entire energy \rm in (2.33)-(2.34) turn
into linear differential operators acting on the infinitely
dimensional space $\Gamma$ of complex valued functions defined on the
(two complex dimensional) configuration space $\Pi$ spanned by Weyl
spinors (multiplicative operators in (3.2) "acting" on $\Gamma$
itself). It is obvious from (3.3) that the functions in (2.25) -
(2.26) representing the energy and the linear three-momentum of the
massless particle turn into multiplicative operators "acting" on
$\Gamma$. The Hamiltonian function $H$ in (1.21) becomes a (possibly
non-linear and non-local) differential operator acting on $\Gamma$.
(For free particles $H$ is simply the kinetic energy in (2.25) i.e. a
multiplicative operator).
\vskip 10pt
A Poincar{\'e} invariant scalar product on the space of complex valued
functions on $\Pi$ will, tentatively, be defined as:
\begin{equation}<g_{1}\vert g_{2}>:=
\int
[{\bar g}_{1}({\bar \pi}_{B}, \ \pi_{B^\prime})
g_{2}(\pi_{B^\prime}, \ {\bar \pi}_{B})]
d\pi^{A^\prime}\wedge
d\pi_{A^\prime} \wedge d{\bar \pi}^{A} \wedge d{\bar \pi}_{A}.\end{equation}
\vskip 10pt
The subspace ${\aleph}$ of $\Gamma$ consisting of functions having
finite norm (with respect to the scalar product in (3.6)) defines a
Hilbert space of quantum states of a massless particle.
\vskip 10pt
Note that, provided state functions vanish at infinity i.e. vanish for
infinite values of the kinetic energy (see 3.9), the operators in
(3.3) and (3.4) are hermitian with respect to the scalar product in
(3.6). Our tentative choice of the scalar product is based on this
fact.
\vskip 10pt
The helicity operator in (3.5) may be regarded as "hermitian" with
respect to the scalar product in (3.6) if the two functions $g_{1}$
and $g_{2}$ in (3.6) represent eigenstates corresponding to the same
eigenvalue of the helicity operator in (3.5).
\vskip 10pt
A point in the four-dimensional configuration space $\Pi$ has a
relatively clear physical interpretation. Three of its coordinates
combine to give components of the linear null four-momentum (the pole)
while the fourth coordinate represents the phase (the
flag) of the Weyl spinor.
\vskip 10pt
This implies that instead of $\pi_{A^\prime}$ and ${\bar \pi}_{A}$ we
may choose more physical coordinates on ${\Pi}$ to label its points.
These physical coordinates are $E=p$ - the kinetic energy of the massless
quantum particle as "observed" in the inertial frame of the source
moving along the world-line in (1.3), $\varphi$ -, $\theta$ - the
angles denoting the direction of motion of the particle in this frame
and $\psi$ - the angle representing the phase variable in this frame.
The correspondence between the two types of coordinates is given by:
\begin{equation}
{\bar \alpha}^{A}{\bar \pi}_{A} =
\pm {\sqrt E} e^{{i\varphi \over 2}}
e^{-i\psi}\sin{\theta \over 2},\qquad \qquad
\alpha^{B^\prime}\pi_{B^\prime}=
\pm {\sqrt E} e^{-{i\varphi \over 2}}
e^{i\psi}\sin{\theta \over 2},\end{equation}
\begin{equation}{\bar \beta}^{B}{\bar \pi}_{B}=
\pm {\sqrt E} e^{-{i\varphi \over 2}}
e^{-i\psi}\cos{\theta \over 2}, \qquad \qquad
\beta^{A^\prime}\pi_{A^\prime}=
\pm {\sqrt E} e^{{i\varphi \over 2}}
e^{i\psi}\cos{\theta \over 2},\end{equation}
\vskip 10pt
\noindent
and inversely by:
\begin{equation}
E=({\alpha}^{B^\prime}{\bar \alpha}^{B} +
{\beta}^{B^\prime}{\bar \beta}^{B}){\pi}_{B^\prime} {\bar \pi}_{B},
\end{equation}
\begin{equation}e^{4i\psi}=
{({\alpha}^{A^\prime}{\pi}_{A^\prime})
({\beta}^{B^\prime}\pi_{B^\prime})
\over
({\bar \beta}^{A}{\bar \pi}_{A})({\bar \alpha}^{B}{\bar \pi}_{B})},
\end{equation}
\begin{equation}
e^{2i\varphi}=
{{p_{(2)} + ip_{(3)}} \over {p_{(2)} - ip_{(3)}}}=
{({\bar \alpha}^{A}{\bar \pi}_{A})
(\beta^{A^\prime}\pi_{A^\prime})
\over
({\bar \beta}^{B}{\bar \pi}_{B})
(\alpha^{B^\prime}\pi_{B^\prime})},\end{equation}
\begin{equation}\cos\theta= {p_{(1)} \over E}=
{(\beta^{A^\prime}{\bar \beta}^{A} -
\alpha^{A^\prime}{\bar \alpha}^{A})
\pi_{A^\prime} {\bar \pi}_{A}
\over
(
\alpha^{B^\prime}{\bar \alpha}^{B} +
\beta^{B^\prime}{\bar \beta}^{B}
)
\pi_{B^\prime} {\bar \pi}_{B}}.\end{equation}
\vskip 10pt
{}From (3.4) and (2.27), (2.30)-(2.32) it follows that the hermitian
differential operators corresponding to the three components of the
the total angular momentum $J_{(1)}$, $J_{(2)}$, $J_{(3)}$ are given
by:
\begin{equation}{\hat J}_{(1)}=
{1 \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \beta}^{B}
+
{\bar \alpha}^{B}{\bar \beta}_{A}){\bar \pi}_{B}
{\partial \over \partial {\bar \pi}_{A}}
- {1 \over \sqrt{2}}
(\alpha_{A^\prime}\beta^{B^\prime}
+
\beta_{A^\prime}\alpha^{B^\prime})\pi_{B^\prime}
{\partial \over \partial \pi_{A^\prime}},\end{equation}
\begin{equation}{\hat J}_{(2)}=
{1 \over \sqrt{2}}
({\bar \beta}_{A}{\bar \beta}^{B}-{\bar \alpha}_{A}{\bar \alpha}^{B})
{\bar \pi}_{B} {\partial \over \partial {\bar \pi}_{A}}
- {1 \over \sqrt{2}}
(\beta_{A^\prime}\beta^{B^\prime}-\alpha_{A^\prime}\alpha^{B^\prime})
\pi_{B^\prime}{\partial \over \partial \pi_{A^\prime}},\end{equation}
\begin{equation}
{\hat J}_{(3)}=
{i \over \sqrt{2}}
({\bar \alpha}_{A}{\bar \alpha}^{B} + {\bar \beta}_{A}{\bar \beta}^{B})
{\bar \pi}_{B} {\partial \over \partial {\bar \pi}_{A}}
+{i \over \sqrt{2}}
(\alpha_{A^\prime}\alpha^{B^\prime} + \beta_{A^\prime}\beta^{B^\prime})
\pi_{B^\prime}{\partial \over \partial \pi_{A^\prime}}.
\end{equation}
\vskip 10pt
The hermitian differential operator representing the square of the
absolute value of the total angular momentum:
\begin{equation}
{\hat J}^{2}:=
{\hat J}_{(1)}{\hat J}_{(1)}+
{\hat J}_{(2)}{\hat J}_{(2)}+
{\hat J}_{(3)}{\hat J}_{(3)}
\end{equation}
\vskip 10pt
\noindent
may, as well-known, be written as:
\begin{equation}
{\hat J}^{2}=({\hat J}_{(2)} + i{\hat J}_{(3)})
({\hat J}_{(2)} - i {\hat J}_{(3)}) +{\hat J}_{(1)}{\hat J}_{(1)}
-{\hat J}_{(1)}.
\end{equation}
\vskip 10pt
For the massless harmonic oscillator the classical (total energy)
Hamilton function on \bf T \rm is given by:
\begin{equation}H=E+{k \over 2}r^{2}
\qquad \qquad
r^{2}:=
y_{(\alpha)}y_{(\alpha)}
\end{equation}
\vskip 10pt
\noindent
where $y_{(1)}$, $y_{(2)}$, $y_{(3)}$ are functions
on \bf T \rm in accordance with
(2.33)-(2.35). Using the quantization prescription in (3.1) and (3.2)
(and the principle of normal ordering) these functions may easily be turned
into differential operators acting on $\Gamma$. However,
from (2.54) we already know
that on \bf T \rm one has:
\begin{equation}
r^{2}=y_{(\alpha)}y_{(\alpha)}=
{{J^{2}-s^{2} + {({ y}_{(\alpha)}p_{(\alpha)})}^{2}}
\over E^{2}}.\end{equation}
\vskip 10pt
Direct calculations also show that in terms of twistor variables we have:
\begin{equation}
-i{\kappa}:=y_{(\alpha)}p_{(\alpha)}=
{i \over 2} {({\bar \pi}_{A}
{\omega}^{A} - \pi_{A^\prime} {{\bar \omega}}^{A^\prime})}.
\end{equation}
\vskip 10pt
Therefore normal ordering of terms yields:
\begin{equation}{\hat \kappa}:=
{1 \over 2}
{({\bar \pi}_{A}
{\partial \over \partial {\bar \pi}_{A}}
+
\pi_{A^\prime}
{\partial \over \partial \pi_{A^\prime}}
+4)}\end{equation}
\vskip 10pt
\noindent
which implies that:
\begin{equation}[{\hat \kappa}, \ \pi_{A^\prime}{\bar \pi}_{A}]=
\pi_{A^\prime}{\bar \pi}_{A}.\end{equation}
\begin{equation}[{\hat \kappa}, \ t^{AA^\prime}\pi_{A^\prime}{\bar \pi}_{A}]=
[{\hat \kappa}, \ E]=t^{AA^\prime}
\pi_{A^\prime}{\bar \pi}_{A}=E.\end{equation}
\vskip 10pt
\noindent
where by square brackets we denote the commutator.
\vskip 10pt
(3.20)-(3.21),
normal ordering of terms and quantization prescription in
(3.1)-(3.2) imply that the hermitian differential
operator corresponding to the function $r^{2}$ in (3.18)-(3.19) is
given by:
\begin{equation}{\hat r}^{2}
:=
{1 \over E^{2}}
({\hat J}^{2}-{\hat s}^{2} -
{\hat \kappa}^{2}).\end{equation}
\vskip 10pt
For the quantum mechanical harmonic oscillator the Hamilton
(total energy) hermitian differential operator acting on $\Gamma$ is
represented by:
\begin{equation}{\hat H}:=E+{k \over 2}{\hat r}^{2}.\end{equation}
\vskip 10pt
Common eigenfunctions and eigenvalues of the maximal set of mutually
commuting hermitian differential operators $\hat s$, ${\hat J}_{(1)}$,
${\hat J}^{2}$, ${\hat H}$ may now be determined, in the usual
fashion, by the following set of differential equations:
\begin{equation}{\hat s}f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
= sf(\pi_{A^{\prime}}, \ {\bar \pi}_{A}),\end{equation}
\begin{equation}{\hat J}_{(1)}f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
= mf(\pi_{A^{\prime}}, \ {\bar \pi}_{A}),\end{equation}
\begin{equation}{\hat J}^{2}f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
= J(J+1)f(\pi_{A^{\prime}}, \ {\bar \pi}_{A}),\end{equation}
\begin{equation}{\hat H}f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
=\epsilon f(\pi_{A^{\prime}}, \ {\bar \pi}_{A}).\end{equation}
\vskip 10pt
We will be looking for a set of solutions of the above equations in the form:
\begin{equation}
f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
=f_{\sigma}(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
Y_{ln}(\theta, \ \varphi)
f_{\epsilon}(E)
\end{equation}
\vskip 10pt
\noindent
where $f_{\sigma}(\pi_{A^{\prime}}, \ {\bar \pi}_{A})$ is any of the
following four holomorphic (either in ${\bar \pi}_{A}$ or in
$\pi_{A^{\prime}}$) simple expressions:
$$({\bar \alpha}^{A}{\bar \pi}_{A})^{-\sigma} \qquad
({\bar \beta}^{A}{\bar \pi}_{A})^{-\sigma} \qquad
(\alpha^{A^\prime}\pi_{A^\prime})^{-\sigma} \qquad
(\beta^{A^\prime}\pi_{A^\prime})^{-\sigma}.$$
\vskip 10pt
The analysis in the sequel does not depend on which of these four
options we assume so the second one is adopted for definitness. We
just note that complex conjugation "creates" a massless particle with
reversed eigenvalue of the helicity operator while the two
remaining options correspond to the different choices of
the "direction" of the helicity relative to the "direction" of
the orbital magnetic quantum number $n$ (see 3.33).
\vskip 10pt
$\sigma$ is a positive integer (say) and $Y_{ln}(\theta, \
\varphi)$ represent the usual spherical harmonics where $\varphi$ and
$\theta$ are functions of the spinor variables as defined in
(3.11)-(3.12). Spherical harmonics are thus homogenous functions of
degree $0$ in spinor variables. Besides for any $f_{\epsilon}(E)$
where $E$ is given by (3.9) and $\hat s$ by (3.5) one has :
\begin{equation}
{\hat s}f_{\epsilon}(E)=0.
\end{equation}
\vskip 10pt
Therefore we obtain:
$${\hat s}
f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})
= {\hat s}[({\bar \beta}^{A}{\bar \pi}_{A})^{\sigma}
Y_{ln}(\theta, \ \varphi)f_{\epsilon}(E)]
=
Y_{ln}(\theta, \ \varphi)f_{\epsilon}(E)
{\hat s}({\bar \beta}^{A}{\bar \pi}_{A})^{\sigma}
=$$
\begin{equation}={\sigma \over 2}
({\bar \beta}^{A}{\bar \pi}_{A})^{\sigma}
Y_{ln}(\theta, \ \varphi)f_{\epsilon}(E)
=s f(\pi_{A^{\prime}}, \ {\bar \pi}_{A})\end{equation}
\vskip 10pt
\noindent
which shows that eigenvalues of the helicity operator are given by
$s={\sigma \over 2}$.
\vskip 10pt
Simple calculations also show that:
$${\hat J}_{(1)}
f(\pi_{A^{\prime}}, \ {\bar \pi}_{A}) =
Y_{ln}(\theta, \ \varphi)f_{\epsilon}(E)
{\hat J}_{(1)}({\bar \beta}^{A}{\bar \pi}_{A})^{-\sigma}
+
({\bar \beta}^{A}{\bar \pi}_{A})^{-\sigma}
f_{\epsilon}(E)
{\hat J}_{(1)}Y_{ln}(\theta, \ \varphi)$$
\begin{equation}=({\sigma \over 2} + n)
({\bar \beta}^{A}{\bar \pi}_{A})^{-\sigma}
Y_{ln}(\theta, \ \varphi)f_{\epsilon}(E)
= m f(\pi_{A^{\prime}}, \ {\bar \pi}_{A}).\end{equation}
\vskip 10pt
The eigenvalues of ${\hat J}_{(1)}$ are thus given by half-integer or
integer numbers:
\begin{equation}m = {\sigma \over 2} + n.\end{equation}
\vskip 10pt
{}From standard considerations (see e.g. Ref. 14) it now follows
that, for a given eigenvalue of the operator ${\hat J}^{2}$
the quantum numbers $m$ are bounded by certain upper and lower limits.
If we denote the upper limit by a positive number $J$, then
one obtains\cite{ll}
that the eigenvalue of ${\hat J}^{2}$ is $J(J+1)$. Note that
the positive number $J=l+{\sigma \over 2}$, where $l$ is the upper limit
(for a given eigenvalues of ${\hat J}^{2}$ and of $\hat s$)
of $n$ in (3.34), may assume half-integer or integer values.
\vskip 10pt
Finally we use the eigenvalue equation in (3.29) which, given quantum
numbers $s$, $m$ and $J$, determines the eigenfunctions $f_{\epsilon}$
and the corresponding eigenvalues of the total energy operator $\hat
H$ in (3.25). When written out explicitly the equation in (3.29)
reduces itself to an ordinary differential equation of second order:
\begin{equation}-{d^{2}f_{\epsilon} \over dE^{2}}
-{5 \over E}{df_{\epsilon} \over dE}
+({a \over E^{2}} + b E)f_{\epsilon} = \lambda f_{\epsilon},\end{equation}
\vskip 10pt
\noindent
where
\begin{equation}a=J(J+1) - ({\sigma^{2} \over 2} + 2\sigma +12),
\qquad b={2 \over k}, \qquad
\lambda={2\epsilon \over k}.\end{equation}
\vskip 10pt
The energy levels $\lambda$ and the corresponding eigenfunctions
$f_{\epsilon}$ may
be calculated explicitly on a computer using
standard methods e.g. the Ritz method\cite{asd}. The already
mentioned computer program Maple V, Release 3 may be of great value
here. The results of such explicit calculations we hope to be able to
present at a later occasion.
\section{\hspace{-4mm}\hspace{2mm}CONCLUSIONS AND REMARKS.}
If "elementary" particles such as e.g. electron, proton, neutron etc.
may be regarded as bound states of a finite number of massless
spinning parts then twistor theory combined with the idea of
relativistic action at a distance should provide a very powerful tool
for construction of such models.
\vskip 10pt
There arises a possibility to use a twistor phase space formulation.
Such a reducible phase space may then be thought of as a direct
product of a finite number of copies of an elementary twistor phase
space \bf T. \rm Dynamics will then be generated by an appropriately
chosen Poincar{\'e} scalar hamiltonian function.
\vskip 10pt
In this paper we therefore emphasized particle aspects of Penrose's
twistor formalism as opposed to the standard treatments where field
aspects are at the front.
\vskip 10pt
The suggested non-standard quantization in the previous section of
this note corresponds to the real polarization of the twistor phase
space as described by Wood- house\cite{woo}. Choosing this non-standard
quantization we however loose some of the results of conventional
twistor theory such as the twistor description of massless free fields
in terms of holomorphic sheaf cohomology, the scalar product on such
fields, geometrization of the concept of positive frequency of the
field and the relationship between conformal curvature and the twistor
"position" (twistor variables) and "momentum" (complex conjugates of
the twistor variables) operators\cite{pr3}.
\vskip 10pt
What we gain is that the real dimension of the relativistic
configuration space of a massless spinning particle is one half of the
real dimension of the configuration space obtained by means of the
conventional holomorphic twistor quantization\cite{prmc}. Further, the
configuration space obtained in our paper has a clear physical
interpretation. Wave functions on such a configuration space define
quantum states in (the "square root" of) the linear momentum
representation. However in our opinion the most important gain is the
fact that using our formulation we are able to treat \it interacting
\rm massless spinning particles (not fields) forming a closed composite
bound system.
\vskip 10pt
Using ideas presented in this paper and in\cite{ab1} it would be
interesting to investigate a fully relativistic closed system forming
a massive and spinning particle composed of three or four directly interacting
massless and spinning parts.
\vskip 10pt
Similar in spirit, attempts to remodel the physics of elementary
particles, have been made before by Hughston\cite{hl},
Popovich\cite{pa} and Perj{\`e}s\cite{pz}. These authors make use of
conventional twistor quantization and as primary objects regard \it
free \rm fields which are then represented by elements of the
holomorphic sheaf cohomology group of an appropriate twistor space.
\vskip 10pt
Finally we note that commutation relation in (2.36) may have an
experimental implication. The uncertainty principle following from
(2.36) predicts that position of a massless and spinning particle can
never be measured exactly. No sharp value of its position vector
exists.
\vskip 10pt
The speculations presented in this note are inconclusive with respect
to their physical significance. Some calculations are in progress in
order to find phenomenological support for the presented ideas.
\vskip 10pt
Nevertheless, our attempts seem to comply with the Twistor
Programme announced by Penrose\cite{penr}.
|
\partial{\partial}
\def\rightarrow{\rightarrow}
\begin{document}
\vspace{20pt}
\title{ DYNAMICAL GENERATION OF LIGHT FERMIONS}
\vspace{25pt}
\author{PANKAJ JAIN \footnote {e-mail: <EMAIL>}}
\address{Physics Department, I.I.T., Kanpur, India 208016}
\maketitle
\vspace{35pt}
\begin{abstract}
We show that composite fermions with masses much smaller than the scale of
confinement arise naturally in certain models which admit dynamical
breakdown of chiral symmetry.
The models are such that to leading order some of the fermions
remain massless but pick up small dynamical masses at subleading order.
\end{abstract}
\bigskip
\noindent
{\bf 1. Introduction}
\bigskip
The existence of fermions with masses much smaller than the scale of
electroweak symmetry breaking might be the result of an approximate chiral
symmetry of the underlying model. There exist several composite models
\cite{comp1}
based on this idea, but in most cases in the absence of fundamental scalars
some of the fermions remain exactly massless
and remaining fermions pick up masses of the order of the scale of confinement
of the underlying strong dynamics, which may be equal to or larger than the
scale of
electro-weak symmetry breaking. Generation of nonvanishing masses for light
fermions in a dynamical framework has proven to be a very difficult problem
in all of the popular scenarios including technicolor \cite{techni},
top condensate \cite{top} as well as models in which fermions and/or
electroweak bosons are composite \cite{comp1}.
In the present paper we display some situations
in which some of the fermions dynamically acquire
very small but non zero masses.
\bigskip
\noindent
{\bf 2. Light fermions in a large N chiral model}
\bigskip
Dynamical light fermions can arise if the fermion representation is such
that to leading order fermion condensate is prevented from forming. We have
in mind some nonabelian gauge group and by leading order we mean leading order
in either the loop expansion or the $1/N$ expansion \cite{largen}
or a small gauge
coupling parameter. In the present section we will confine ourselves to the
cases in which $1/N$ is the small parameter and will discuss
the generalizations
in the later sections. The possibility of a systematic
loop expansion within a dynamical framework is very interesting although there
is no evidence that this is a reliable expansion scheme.
To give a simple example
we consider a SU(N)$_1\times$SU(N)$_2\times$SU(2)$_1\times$SU(2)$_2$
model with N large. The small parameter
in this example is $1/N$.
We introduce
fermions in following representation of the
SU(N)$_1\times$SU(N)$_2\times$SU(2)$_1\times$SU(2)$_2$ gauge group.
$$a_L\rightarrow (N,1,1,1),\ \ \ \ a_R\rightarrow (1,N,1,1)$$
$$b_L\rightarrow (N,1,1,1),\ \ \ \ b_R\rightarrow (1,N,1,1)$$
$$c_L\rightarrow (N^*,1,2,1),\ \ \ \ c_R\rightarrow (1,N^*,1,2)$$
The representation has been chosen such that the interaction is free of
anomalies. We have not specified the transformation of these fermion
representations under electroweak and color interactions. Generalization
of the present model to include electroweak and color interactions will
be discussed in a separate publication.
The two SU(2) groups are assumed to be strong which will prevent the
condensation of the fermions $a$ or $b$ with the fermion $c$.
The strong SU(2) groups continue to have asymptotic freedom as long as $N<11$.
As we discuss later
this model is expected to be confining because of its non-abelian nature and
the
gauge symmetry breaking is small.
We assume the scale of confinement of the two SU(N) groups
to be $\Lambda_{\rm conf}$.
The reason we are attracted to such SU(N)$_1\times$SU(N)$_2$ type models is
that the scale of chiral
symmetry breaking will be considerably suppressed compared to the scale of
confinement. Because of the presence
of the strong SU(2) interactions, the fermion $a_L$ can only condense with
either $a_R$ or $b_R$. We assume that it condenses only with one of these two
right handed particles and the horizontal symmetry which rotates $a$ into $b$
remains unbroken. Based on our experience with QCD we expect this to be true.
In any case we can always assign different electric charges to these two
fermions to assure that $a_L$ condenses only with $a_R$.
To analyze the pattern of mass generation we note that
all possible diagrams that can convert a left
handed fermion to a right handed fermion contain atleast one internal fermion
loop and are therefore suppressed by one power of N.
All of these graphs
will therefore vanish as $N\rightarrow \infty$ and chiral symmetry will
remain unbroken. In arriving at this conclusion we have assumed that
the dynamically generated fermion mass decreases as the value of the
effective coupling decreases. Based on model calculations this is generally
expected to be the case for QCD type vectorial theories. For example,
the one loop analysis of Schwinger-Dyson shows that as $\alpha_s$, treated as
a constant, decreases
the dynamically generated mass also decreases and eventually vanishes as the
coupling goes below a critical value \cite{dsb}. If instead a renormalization
group improved expression is used for the effective coupling, such that the
coupling continues to rise with decrease in momentum, then there is
no critical point but the dynamical mass continues to decrease
with the decrease
in the value of effective coupling at some scale.
We will assume this also to be the case for the present model.
The Schwinger-Dyson (SD)
equation including only the leading order contribution
in the loop expansion of CJT effective action \cite{cjt},
which can lead to chiral symmetry breaking, is shown is Fig. 1.
To see how the $1/N$ suppression
factor effects the scale of chiral symmetry breaking we use a simple model
for the nonabelian gauge coupling. We assume that the coupling has the form
$$\alpha \approx 1/{\rm log}\bigg(q^2/\Lambda^2_{\rm conf} + 1\bigg)$$
The reason for this choice \cite{richardson} is simply that it interpolates
between the
correct asymptotic behavior and the popular infrared behavior \cite{infra}
since it
leads to $1/q^4$ momentum dependence for the hypergluon propagator. We take
the scale at which this coupling becomes equal to 1 to be the scale of
chiral symmetry breaking. For the chosen behavior the coupling becomes
equal to 1 at $q^2/\Lambda^2_{\rm conf} = 1.7$. This is roughly the
scale of chiral
symmetry breaking if the theory is vectorial. However in the present case
the effective coupling is a factor of $N$ smaller. To get the
scale of chiral symmetry breaking in this case we set $\alpha/N=1$ to get
$q^2/\Lambda^2_{\rm conf} = {\rm exp}(1/N) -1$. For $N$ of the order of 10
this yields
$q^2/\Lambda^2_{\rm conf} = 0.1$, which shows a significant suppression factor.
This model calculation atleast shows that for N large enough the scale of
chiral symmetry is much smaller than the scale of confinement. We point out
that if N is arbitrarily large but not infinite than in the model
discussed above chiral symmetry breaking will take place. It is not clear,
however, if this
is true in reality because of our lack of knowledge of the behavior
of nonabelian theories at low energies and the effective coupling may
not increase monotonically with decrease in momentum.
In any event there may exist a
large range of values of N for which the chiral symmetry still takes place.
We assume this to be the case.
This theory will behave very
different from QCD in which case the scale of confinement is roughly the
same as the scale of chiral symmetry breaking. In particular, in the present
case to leading order we may simply ignore chiral symmetry breaking. Indeed
in the limit as N$\rightarrow\infty$, there is no breakdown of chiral
symmetry.
This implies that to leading order all fermions will be massless.
This conclusion holds not only for the elementary confined fermions but also
for the composite fermions which are bound states of the type
N$\times$N$\times$N$\times$..... . In the left-right theory under consideration
there will infact exist several such states, the simplest
one being made of N
left handed or N right handed fermions. More complicated composite fermions
may also be formed by including one or more hypergluons along with the
N fermions. In the infinite N limit these are the only type of fermions
allowed.
One cannot, for example, have a bound state containing some left handed and
some right handed fermions since there is no binding between left and
right handed fermions in this limit.
We next consider the finite N corrections which will link the two
composite fermions discussed above. In order to convert a left handed composite
fermion to a right handed composite fermion we need to convert all the N
fundamental left handed fermions into right handed fermions. This coupling of
left and right handed composite fermions is shown in figure 2. Coversion of
each of these fundamental left handed fermion into a right handed fermion is
suppressed by one power of the ratio of the scale of chiral symmetry breaking
and the confinement scale. We call this suppression factor $\xi$. The
conversion of N elementary left handed fermions into right handed fermions
will therefore be suppressed
by $\xi^N$. This shows that the mass of these composite particles will be
extremely small. The logic used above to get this suppression factor is
very different from the usual intuition one has about bound states. However
even the usual intuition applied to the present case shows that the suppression
factor has to be very large. We are forming very tightly bound states of
fermions which have mass much smaller than the scale of confinement. The
binding
energy is necessarily very large, and results in very small bound state mass.
This argument, however, does not tell us whether the lightest state
is a fermion or a boson. The systematic large N expansion gives a very
good indication that the fermion has to be the lightest state.
The boson states,
which have a group theoretic structure N$\times \bar{\rm N}$,
are much heavier since they do not have the suppression factor $\xi^N$. The
mass of these states in the large N limit is independent of N and therefore
these states will have masses much smaller than the confinement scale but
not as small as the fermion masses.
We note that because of finite N corrections the gauge symmetry
SU(N)$_1\times$SU(N)$_2$ will be broken to SU(N)$_{hc}$, where $hc$ stands
for hypercolor. However since this breaking is subleading the massive gauge
particles will also be very light. Furthermore since the breaking is
negligible to leading order, the theory is confining and all the physical
states must be singlets under both SU(N)$_1$ and under SU(N)$_2$.
The model described in this paper naturally generates very tightly bound
composite fermions
which are much lighter than the scale of confinement. Generalizations of
this model to include color and electroweak interactions are currently
under consideration and will be described in a separate publication.
\bigskip
\noindent
{\bf 3. Acknowledgement}
\bigskip
I would like to thank Chris Hill,
Doug McKay, Herman Munczek, John Ralston, Joseph
Schechter and
especially Kimball Milton for useful discussions. This work was
supported in part by the U.S. DOE contract
no. DE-FG05-91ER-40636.
|
\section{Introduction}
\label{introduction}
The Dyson-Schwinger equations (DSEs) provide a useful, semi-phenomenological
tool for the study of QCD. These coupled integral equations relate the
n-point (Schwinger) functions of QCD to each other. They provide a
nonperturbative, Poincar\'e invariant framework that enables one to correlate
hadronic observables through the properties of the Schwinger functions of the
elementary excitations in QCD; i.e., the Schwinger functions of quarks and
gluons. (Quark and gluon propagators (2-point functions) are examples of
such Schwinger functions.) This makes it particularly suitable for
addressing questions such as confinement and dynamical chiral symmetry
breaking and also hadronic spectroscopy and interactions. This approach is
reviewed in Ref.~\cite{review} and has recently been applied to the study of
$\pi$-$\pi$ scattering\cite{RCSI94}, the electromagnetic pion form
factor\cite{CDRpion}, $\rho$-$\omega$ mixing\cite{rhoomega} and the anomalous
$\gamma^\ast\pi^0\rightarrow \gamma$-transition form factor.\cite{gpig}
It is possible to obtain information about such Schwinger functions via a
numerical simulation of a lattice-QCD action.\cite{Lattice,Lattice2,Lattice3}
However, in addition to the usual problems associated with identifying and
establishing the existence of the continuum limit, and recovering the global
symmetries of QCD, this also requires gauge fixing on the spacetime lattice.
Gauge fixing eliminates a number of gauge-equivalent gauge-field
configurations, thereby leading to poorer statistics. It does not eliminate
all such configurations, however. One is left with Gribov copies; i.e.,
gauge configurations in the gauge-fixed ensemble that are not distinct but
are related by topologically nontrivial gauge
transformations.\cite{gribov,rivers,Z94} This entails an overcounting problem
in the evaluation of gauge-fixed correlation functions. Present studies are
encouraging, having established that this approach to the calculation of
gauge-fixed QCD Schwinger functions is feasible.\cite{Lattice2} However, the
problems identified above entail that they are currently qualitatively and
quantitatively unreliable.
Presently the most reliable estimates of the behaviour of quark and gluon
Schwinger functions are obtained in DSE studies. The DSEs are a tower of
coupled equations and a solution is only tractable if this tower is truncated.
Truncation procedures that preserve the global symmetries of QCD are easy to
construct and implement. This has not yet been accomplished for the local
symmetry in QCD, however, progress is being made following the realisation of
the importance of the nonperturbative structure of the fermion--gauge-boson
vertex.\cite{review,BC80,CP90,BR93,DMR94,BP94} This introduces an uncertainty
in the infrared; i.e., for $k^2< 1-2$~GeV$^2$. However, this uncertainty is
merely quantitative. There is general agreement on the qualitative features
of the quark and gluon 2-point Schwinger functions; i.e., 1) that the gluon
2-point function is significantly enhanced at small
spacelike-$k^2$\cite{review,BBZ,Atkin83,BP89} and that this entails an
enhancement of the momentum-dependent quark
mass-function\cite{review,MN83,PCR89,MM89,WKR91,Smek}; and 2) that for $k^2>
1-2$~GeV$^2$ the two-loop, renormalisation group improved, perturbative
results are quantitatively reliable.
Some phenomenological DSE studies have employed a parametrisation of the
2-point quark Schwinger function based on these results; for example,
Refs.~\cite{RCSI94,CDRpion,rhoomega,gpig}. Such studies are
phenomenologically efficacious. However, they involve the addition of new
parameters when applied to systems involving other than $u$ and $d$ quarks.
The introduction of new parameters is unnecessary when the propagator for a
quark of a given flavour is obtained directly from a quark DSE whose kernel
is determined by the 2-point gluon Schwinger function and the quark-gluon
vertex. This procedure correlates the propagators for quarks of different
flavours via the parameters in the gluon 2-point function. There have been
studies that employ this approach; for example Refs.~\cite{PCR89,Kugo,JM93}.
However, it is computationally more intensive and the studies therefore
addressed the calculation of a smaller class of observables. The present
study is a first step in extending this latter approach.
Herein we employ a one parameter model gluon propagator (gluon 2-point
Schwinger function), motivated by the results of
Refs.~\cite{BBZ,Atkin83,BP89}, in a calculation of a range of $\pi$- and
$\rho$-meson observables. The one parameter is a mass scale that can be
interpreted as marking the transition between the perturbative and
nonperturbative domains. This model gluon propagator provides the kernel for
a quark DSE, which is solved to obtain the quark propagator (quark 2-point
Schwinger function) for real-$p^2\in (-\infty,\infty)$. These two Schwinger
functions provide the kernel of the meson Bethe-Salpeter equation (BSE),
whose solution yields the meson mass and Bethe-Salpeter amplitude, which is a
necessary element in the calculation of decay constants and scattering
lengths, for example. The single mass parameter determines all of these
Schwinger functions and is varied to obtain a good fit to a range of
calculated $\pi$ observables. This illustrates the utility and economy of
the approach.
In studying the pion BSE we derive a mass formula for the pion, which
involves only the vacuum, dressed quark propagator, valid to all orders in
$m_R$, the renormalised current quark mass. Our numerical studies show that
this formula provides an excellent estimate of the mass that is obtained by
actually solving the BSE.
The model gluon propagator is discussed in Sec.~\ref{secprop} and the quark
DSE in Sec.~\ref{secdse}. The pion mass formula is presented in
Sec.~\ref{secmassform}. Our numerical results are discussed in
Sec.~\ref{secres} and we summarise and conclude in Sec.~\ref{secconc}.
\section{Model gluon propagator}
\label{secprop}
In Euclidean metric\cite{Euclidean} the Landau gauge gluon propagator is
\begin{equation}
g^2 D_{\mu\nu}(k) =
\left(\delta_{\mu\nu} - \frac{k_\mu k_\nu}{k^2}\right)
\frac{g^2}{k^2[1+\Pi(k^2)]} \label{glprp1}
\end{equation}
where $\Pi(k^2)$ is the gluon vacuum polarisation. Setting
\mbox{${\cal Z}_1^{gh} = {\cal Z}_3^{gh}$}, where ${\cal Z}_1^{gh}$ is the
renormalisation constant for the ghost-gluon vertex and ${\cal Z}_3^{gh}$
that for the ghost wave function, then
\begin{equation}
\Delta(k^2) \equiv \frac{g^2}{1+\Pi(k^2)}
\end{equation}
satisfies the same renormalisation group equation as the QCD running coupling
constant, $\alpha(k^2)$,\cite{UBG80} and hence
\begin{equation}
\label{glprp2}
\left(g^2 D_{\mu\nu}(k)\right)_R =
\left(\delta_{\mu\nu} - \frac{k_\mu k_\nu}{k^2}\right)
\frac{4\pi\alpha(k^2)}{k^2}~.
\end{equation}
This is sometimes described as the ``Abelian approximation'' because it
entails the QED-like Ward identity $Z_1 = Z_2$, where $Z_1$ is the
quark-gluon vertex renormalisation constant and $Z_2$ is the quark
wavefunction renormalisation constant.\cite{review}
The two-loop renormalisation group expression for the running coupling
constant only receives small ($\sim 10$\%) corrections from higher orders for
spacelike-$k^2>1$~GeV$^2$ and hence can be said to provide an accurate
representation on this domain. For $k^2<1$~GeV$^2$, however, $\alpha(k^2)$
is not known and can only be calculated nonperturbatively. The current
status of such studies is summarised in Ref.~\cite{review} and, as remarked
in Sec.~\ref{introduction}, gluon-DSE studies agree on the qualitative
behaviour of $\alpha(k^2)$ at small-$k^2$. Present phenomenological
quark-DSE studies rely on an Ansatz for $\alpha(k^2<1~{\rm GeV}^2)$ motivated
by these gluon-DSE studies.
Herein we consider a parametrisation suggested by the Landau gauge studies of
Ref.~\cite{BP89}, which revealed a strong enhancement in the gluon propagator
at small spacelike-$k^2$ ($< 1$~GeV$^2$) that could be described by an
integrable singularity. We employ the one parameter form:
\begin{equation}
\label{delk}
\Delta(k^2) = 4 \, \pi^2 \, d\,
\left[ 4\,\pi^2\,m_t^2\,\delta^{4}(k)
\, + \,
\frac{1-{\rm e}^{(-k^2/[4 m_t^2])}}{k^2}\right]~,
\end{equation}
where $d=12/(33-2N_f)$, with $N_f=3$ the number of light flavours. The first
term in Eq.~(\ref{delk}) provides an integrable, infrared
singularity\cite{MN83}, which generates long-range effects associated with
confinement, and the second ensures that the propagator has the correct large
spacelike-$k^2$ behaviour, up to $\ln[k^2]$-corrections. A form similar to
this has been used by other authors\cite{PCR89,MM89,WKR91,Smek} with 1-loop
logarithmic corrections included in the second term. We neglect these terms
as a simple expedient to ensure that our gluon propagator does not have a
Lehmann representation and may therefore be interpreted as describing a
confined particle; i.e., an elementary field with which there is no
associated asymptotic state.\cite{review,RWK92}
Since ours is a model gluon propagator there is no reason why the
coefficients of the two terms in Eq.~(\ref{delk}) should be related in the
particular fashion we have chosen. However, consider
\begin{equation}
\label{delx}
\Delta(x^2) \equiv \int \frac{d^4k}{(2\pi )^4}
{\rm e}^{i k\cdot x} \Delta(k^2) \nonumber \\
= d\,\left[ m_t^2
+ \frac{1}{x^2}\,{\rm e}^{-x^2\,m_t^2} \right]~.
\end{equation}
It is clear from this that with our choice of the ratio of these coefficients
the effects of $\delta^4(k)$ in Eq.~(\ref{delk}) are completely cancelled at
small $x^2$; i.e.,
\begin{equation}
\label{delsmx}
\Delta(x^2) \stackrel{m_t^2 x^2< 1}{\simeq}
\frac{d}{x^2} + {\rm O}(x^2)~,
\end{equation}
which is the form expected from QCD (again neglecting logarithmic-corrections).
One can therefore interpret $m_t$ as the mass scale in our model that marks
the transition from the perturbative to the nonperturbative regime. Herein
$m_t$ is varied to provide a best fit to a range of calculated pion
observables. [See Eq.~(\ref{params}) and the associated discussion.]
\section{quark self energy}
\label{secdse}
In Euclidean metric\cite{Euclidean} the DSE for the quark propagator is
\begin{equation}
\label{dserenorm}
S^{-1}(p)=Z_2(i\gamma \cdot p+ m_0)+\Sigma '(p),\label{s^-1}
\end{equation}
where
\begin{equation}
\label{sep}
\Sigma '(p)\equiv Z_1\int ^{\Lambda }\frac{d^4k}{(2\pi
)^4}\,\case{4}{3}\,g^2\,
D_{\mu \nu }(p-k)\gamma _{\mu }S(k)\Gamma_{\nu }(p,k)~,
\end{equation}
with $\Gamma_\mu(p,k)$ the quark-gluon vertex, is the regularised self
energy, which can be decomposed as
\begin{equation}
\Sigma'(p) = i\,\gamma\cdot p \left(A'(p^2) - 1\right) + B'(p^2)~.
\end{equation}
The inverse of the renormalised quark propagator is
\begin{equation}
S^{-1}(p) = i\,\gamma\cdot p + \Sigma(p) =
i\,\gamma\cdot p A(p^2) + B(p^2)~.
\label{sinvrenorm}
\end{equation}
Herein the prime denotes regularised quantities and unprimed quantities are
fully renormalised.
We employ a subtractive renormalisation scheme, requiring that, at a
spacelike renormalisation point, $\mu^2$,
\begin{equation}
\label{bcrenorm}
S^{-1}(p)|_{p^2=\mu ^2}=i\gamma \cdot p +m_R,
\end{equation}
with $m_R$ the renormalised {\it current } quark mass.
In this scheme, the wavefunction and mass renormalisation constants are
given by
\begin{equation}
Z_2\equiv 2 - A'(\mu ^2,\Lambda ^2) \; \; \; \mbox{and} \; \; \;
m_R\equiv Z_2\,m_0(\Lambda ^2)+B'(\mu ^2,\Lambda ^2)~, \label{wfmren}
\end{equation}
respectively, and the renormalised self energies are therefore obtained from
\begin{eqnarray}
A(p^2,\mu^2) & = & 1 + A'(p^2,\Lambda^2) - A'(\mu^2,\Lambda^2)~, \\
B(p^2,\mu^2) & = & m_R(\mu^2) + B'(p^2,\Lambda^2) - B'(\mu^2,\Lambda^2)~.
\label{rense}
\end{eqnarray}
In this scheme, $A(\mu ^2)=1$ and $B(\mu ^2)=m_R(\mu^2)$. (In the following
we often write $m_R(\mu^2)$ as simply $m_R$, in which case the $\mu^2$
dependence is implicit.)
The renormalised axial-vector Ward identity is
\begin{equation}
\label{awi}
(p-q)_{\mu }i\,\Gamma _{\mu }^5(p,q)
= S^{-1}(p)\,\gamma _5 +\gamma _5\,S^{-1}(q)
- 2m_R\,\Gamma ^5(p,q)~.
\end{equation}
The composite operators $\Gamma _{\mu }^5$ and $\Gamma ^5$ are renormalised
such that, at $p^2=\mu^2=q^2$,
\mbox{$\Gamma _{\mu }^5(p,q)= \gamma_\mu\gamma_5$} and
\mbox{$\Gamma^5(p,q)=\gamma_5$}.
The chiral limit is identified as the limit in which the renormalised
axial-vector current is conserved; i.e, with the limit $m_R(\mu^2)\rightarrow
0$.
\subsection{Analysis of the large-$p^2$ behaviour of the quark propagator}
At large spacelike-$k^2$ and $p^2$ one may replace the gluon propagator and
the quark-gluon vertex by their asymptotic forms:
\begin{equation}
\Delta(k^2) \rightarrow \frac{1}{k^2}\; \; \; \mbox{and} \; \; \;
\Gamma_\mu(p,k) \rightarrow \gamma_\mu~. \label{asforms}
\end{equation}
In this limit $A(p^2)\equiv 1$ and $B(p^2)$ is the solution of
\begin{equation}
B(x) = Z_2m_0+\frac{\lambda}{4}\,\int_0^{\Lambda^2}\,dy\,y\,
\left(\frac{1}{x}\,\theta(x-y) + \frac{1}{y}\theta(y-x)\right)\,
\frac{B(y)}{y + B^2(y)},\label{asdse}
\end{equation}
where $x=p^2$, $y=k^2$ and $\lambda= 4\,Z_1\,d$.
For $x$ such that $B(x)^2 \ll x$; i.e., for $x \ge \mu^2$, this integral
equation is equivalent to the differential equation
\begin{equation}
\label{bdifeq}
\frac{d}{dx}\left( x^2\,\frac{d}{dx}\,B(x)\right)
+ \frac{\lambda}{4}\,B(x) = 0~,
\end{equation}
subject to the boundary condition
\begin{equation}
\label{bndconmu}
B(\mu^2) = m_R
\end{equation}
or
\begin{equation}
\left.\left( \frac{d}{dx}\left[ x B(x)\right] \right)
\right|_{x=\Lambda^2} = Z_2\,m_0~.\label{bndconUV}
\end{equation}
Under the change of variables $x=\mu ^2{\rm exp}(2 z )$, Eq.~(\ref{asdse})
becomes
\begin{equation}
\ddot{B}(z )+2\dot{B}(z )+\lambda B(z)=0~,\label{dhoeq}
\end{equation}
which is the equation of motion for a damped harmonic oscillator. One has
critical damping for $\lambda =\lambda _C=1$ and this yields the critical
coupling for dynamical chiral symmetry breaking; i.e., in the absence of the
first term in Eq.~(\ref{delk}), the model would still exhibit dynamical
chiral symmetry breaking for $\lambda>1$. This behaviour has been observed
in QED\cite{FK76} and phenomenological models of QCD without an
infrared-singular model gluon propagator.\cite{H84,AJ88,RM90}
The solution of Eq.~(\ref{bdifeq}) consistent with Eq.~(\ref{bndconmu}) is
\begin{equation}
\label{basympt}
B(z) = \kappa\,{\rm e}^{-z}\,\cos\left(z\,\sqrt{\lambda-1} + \phi\right)~,
\end{equation}
with
\begin{equation}
\kappa\,\cos\phi = m_R~.
\end{equation}
In the chiral limit $m_R=0$ and hence $\phi=\pi/2$. In general $\kappa$ is
only
determined in a complete solution of the integral equation.
The boundary conditions in Eqs.~(\ref{bndconmu}) and (\ref{bndconUV}) are
equivalent: a given value of $m_R$ entails a given value of $Z_2\,m_0$ and
vice-versa. In fact, for finite $\Lambda$, $m_R=0$ generally entails
$Z_2\,m_0\neq 0$. It follows from Eq.~(\ref{basympt}), however, that for any
finite value of $m_R$
\begin{equation}
\lim_{\Lambda^2\rightarrow \infty}\,Z_2(\mu^2,\Lambda^2)\,m_0(\Lambda^2)
= 0~.
\end{equation}
Equation~(\ref{basympt}) indicates that the renormalised mass function will
exhibit damped oscillations about zero for $p^2>\mu^2$, a feature we observed
in our numerical solutions, which were well described by Eq.~(\ref{basympt})
on $p^2\in [\mu^2,\Lambda^2]$. With the exception of Ref.~\cite{HW95}, other
DSE studies implicitly use $\mu=\Lambda$ and hence the oscillations are not
observed. The oscillations were observed in Ref.~\cite{HW95}, which
addresses in detail the nonperturbative renormalisation of the fermion DSE in
QED.
\subsection{Additional remarks}
The ``Abelian approximation'' entails that $Z_1 = Z_2$ in
Eqs.~(\ref{dserenorm}) and (\ref{sep}). We make this identification
hereafter.
In the numerical studies described below we employed the rainbow
approximation:
\begin{equation}
\label{rainbow}
\Gamma_\mu(p,k)=\gamma_\mu~.
\end{equation}
This is a quantitatively reliable approximation in Landau gauge. (This is
not the case in other gauges). For example, in studies of the critical
coupling for dynamical chiral symmetry breaking, a comparison of the results
obtained using this approximation\cite{FK76} with those obtained using more
realistic vertex Ans\"atze\cite{review,BC80,DMR94,BP94,Atkin93} shows this
approximation to be accurate to 5\%. The improvements to this approximation
are qualitatively important\cite{review,BC80,DMR94,BP94,Atkin93}, being
crucial to the restoration of multiplicative renormalisability and gauge
covariance. However, herein a quantitatively reliable calculation scheme is
sufficient and this is provided by Eq.~(\ref{rainbow}) in Landau gauge.
\section{A Pion Mass Formula}
\label{secmassform}
The unrenormalised BSE for the pion in generalised-ladder approximation is,
with unrenormalised n-point functions beyond denoted by $\tilde{\cdot}$,
\begin{equation}
\label{bse}
\tilde{\Gamma}_\pi(p;P) + \int\,\frac{d^4q}{(2\pi )^4}\,
\case{4}{3}\,g^2\,\tilde{D}_{\mu\nu}(p-q)\,\gamma_\mu\,
\tilde{S}(q+\case{1}{2}P)\,\tilde{\Gamma}_\pi(q;P)\,
\tilde{S}(q-\case{1}{2}P)\,
\gamma_\nu = 0~,
\end{equation}
where $P=p_1+p_2$ is the total momentum and $p=(p_1-p_2)/2$ the relative
momentum of the $\bar q$-$q$ pair.
For the pion it is a good approximation\cite{JM93,sepetal} to write
\begin{equation}
\label{pileading}
\tilde{\Gamma}_\pi(p;P) = \gamma_5 \tilde{F}(p^2,P^2)~,
\end{equation}
in the sense that $\tilde{\Gamma}_\pi(p;P)$ is a general pseudoscalar
$4\times 4$ matrix and the right-hand-side is, pointwise, a good
approximation to it and the inclusion of the other allowed Dirac amplitudes
alters the mass eigenvalue by $< 1$~\%. With this approximation
Eq.~(\ref{bse}) becomes~$[C_2(R) = (N_c^2-1)/(2N_c) = 4/3~{\rm for}~N_c=3]$
\begin{equation}
\label{bseconv}
8\,N_c\,\tilde{F}(p^2,P^2) = 3\,C_2(R)\,\int\,\frac{d^4q}{(2\pi )^4}\,
\tilde{\Delta}(p-q)\,\tilde{H}(q;P)
\end{equation}
with
\begin{equation}
\tilde{H}(p;P) = 8\,N_c\,\left(p_+ \cdot p_
- \tilde{\sigma}_V^+ \tilde{\sigma}_V^-
+ \tilde{\sigma}_S^+ \tilde{\sigma}_S^- \right)\,
\tilde{F}(p^2,P^2)~,
\end{equation}
where we have defined $p_\pm = p \pm P/2$,
\begin{eqnarray}
\tilde{\sigma}_V^\pm = \frac{\tilde{A}(p_\pm^2)}
{p_\pm^2\,\tilde{A}(p_\pm^2)^2 + \tilde{B}(p_\pm^2)^2}
\;\; & \mbox{and} & \;\;
\tilde{\sigma}_S^\pm = \frac{\tilde{B}(p_\pm^2)}
{p_\pm^2\,\tilde{A}(p_\pm^2)^2 + \tilde{B}(p_\pm^2)^2}~.
\end{eqnarray}
Equation~(\ref{bseconv}) is a convolution in four dimensions and can be
rewritten as
\begin{equation}
\label{bse3}
0= 8\,N_c\,\tilde{F}_P(x) - C_2(R)\,3\,\tilde{\Delta}(x)\,\tilde{H}_P(x)
\end{equation}
with $\tilde{H}_P(x)$ the Fourier transform, with respect to $p$, of
$\tilde{H}(p;P)$.
Multiplying the right-hand-side of Eq.~(\ref{bse3}) by
$(\tilde{F}_P(-x)/[3 C_2(R)\tilde{\Delta}(x)])$ one can construct
\begin{equation}
\Pi_\pi(P) \equiv \int\,d^4x\,\left( \frac{8 N_c}{3 C_2(R)}
\frac{\tilde{F}_P(-x)\,\tilde{F}_P(x)}{\tilde{\Delta}(x)}
- \tilde{F}_P(-x)\,\tilde{H}_P(x) \right)~.
\end{equation}
In the auxiliary-field bosonisation of the Global Colour-symmetry
Model\cite{review,CR85} the effective action contains the term
\begin{equation}
\int d^4x\,d^4y\,\pi^i(x)\,\Pi_\pi(x-y)\,\pi^i(y)~,
\end{equation}
with $\pi^i(x)$ a local field variable identified with the pion field. One
sees from this that $\Pi_\pi(P)$ plays the role of the inverse propagator for
the composite pion field. Further, at the solution of the BSE,
$P^2=-m_\pi^2$, Eq.~(\ref{bse3}) is satisfied and hence
\begin{equation}
\Pi_\pi(P^2= - m_\pi^2) = 0~.
\end{equation}
It has been shown\cite{DS79} that for $m_0=0$ the unrenormalised
BSE has a massless, $P^2=0$, solution with
\begin{equation}
\label{Goldthma}
\tilde{F}_P(x) = \tilde{F}_{P=0}(x) = \tilde{B}_{m_0=0}(x)~,
\end{equation}
which is the manifestation of Goldstone's theorem in the DSE approach. Using
this as an approximation for $P^2= -m_\pi^2\neq 0$, via the unrenormalised DSE:
\begin{equation}
\tilde{B}_{m_0=0}(x)= 3\,C_2(R)\,
\tilde{\Delta}(x)\,\tilde{\sigma}_S^{m_0=0}(x)~,
\end{equation}
one obtains
\begin{equation}
\Pi_\pi(P) \approx \int\,d^4x\,\tilde{B}_{m_0=0}(x)\,\left(
8 N_c\,\tilde{\sigma}_S^{m_0=0}(x) - \tilde{H}_P(x) \right)
\equiv \bar \Pi_\pi(P)~.
\end{equation}
This is manifestly invariant under renormalisation and hence one may write
\begin{equation}
\label{Pirenorm}
\bar\Pi_\pi(P) = \int\,d^4x\,B_{m_R=0}(x)\left(
8 N_c\,\sigma_S^{m_R=0}(x) - H_P(x;m_R) \right)~,
\end{equation}
with every quantity on the right-hand-side renormalised ($\sigma_S$ and $H$
have the same form but with unrenormalised quantities replaced by
renormalised ones) and evaluated with $m_R\neq 0$ unless otherwise specified.
As remarked above, $\Pi(P^2=-m_\pi^2)=0$ at the solution of the BSE.
Equation~(\ref{Pirenorm}) therefore allows one to obtain a simple pion mass
formula derived from the generalised-ladder approximation to the BSE and
expressed solely in terms of the massless and massive renormalised, vacuum,
dressed quark propagators.
For the pion (because $m_\pi^2 \simeq 0$) it is a good approximation to write
\begin{equation}
\label{massform}
\bar\Pi_\pi(P) \approx \bar\Pi_\pi(0) + P^2\,N_\pi^2
\end{equation}
where
\begin{eqnarray}
\lefteqn{N_\pi^2 = \left(\frac{d}{dP^2}\bar\Pi_\pi(P^2)\right)_{P^2=0}=}\\
& &
\frac{N_c}{8\pi^2}\int_0^{\Lambda^2}\,ds\,s\,B_{m_R=0}(s)^2\,
\left( \sigma_{V}^2 - 2 \left[\sigma_S\sigma_S' + s
\sigma_{V}\sigma_{V}'\right]
- s \left[\sigma_S\sigma_S''- \left(\sigma_S'\right)^2\right]
- s^2 \left[\sigma_V\sigma_V''- \left(\sigma_V'\right)^2\right]\right)~,
\nonumber
\end{eqnarray}
with the primes denoting differentiation with respect to $s=p^2$ and
$\sigma_V$ and $\sigma_S$ evaluated at $m_R$. This is just the conventional,
generalised-ladder approximation Bethe-Salpeter amplitude normalisation
constant, calculated neglecting small ($\sim$ 2\%) O$(m_\pi^2)$ corrections.
We note that if $A(p^2)\equiv 1$, $N_\pi=f_\pi$. In general, the
approximation $N_\pi\approx f_\pi$ is accurate to within 10\% and the
difference is a measure of the error introduced by the approximation of
Eq.~(\protect\ref{pileading}).\cite{review} (Also see Table.~\ref{tabobs}.)
Equation~(\ref{massform}) yields the explicit pion mass formula\cite{RTC}
\begin{equation}
\label{expmass}
m_\pi^2\,N_\pi^2 =
\frac{N_c}{2\pi^2}\,\int_0^{\Lambda^2}\,ds\,s\,
\frac{B_{m_R=0}(s)}{B_{m_R\neq0}(s)}
\left(B_{m_R\neq0}(s)\,\sigma_S^{m_R=0}(s)
- B_{m_R=0}(s)\,\sigma_S^{m_R\neq 0}(s)\right)~.
\end{equation}
One notes immediately that, for a given value of $m_R$, $m_\pi^2 \rightarrow
\mbox{constant} < \infty$ as $N_c\rightarrow \infty$ and that, for arbitrary
$N_c$, $m_\pi^2\rightarrow 0$ as \mbox{$m_R\rightarrow 0$}. Further, if the
DSE is solved with a quark-gluon vertex that ensures multiplicative
renormalisability then $m_\pi^2$ is a renormalisation point invariant and the
result is independent of the cutoff $\Lambda^2$. The integral on the
right-hand-side of Eq.(\ref{expmass}) is convergent in the limit
$\Lambda^2\rightarrow\infty$.
{}From Eq.~(\ref{expmass}) one can recover what is sometimes called the
Gell-Mann-Oakes-Renner relation in the form:
\begin{equation}
\label{GMOR}
m_\pi^2\,f_\pi^2 =
- \,m_R^{\mu^2} \, \langle \bar q q \rangle_{\rm vac}^{\mu^2}~,
\end{equation}
where
\begin{equation}
\label{convqbq}
-\langle \bar q q \rangle_{\rm vac}^{\mu^2}
= \frac{N_c}{2\pi^2}\,\int_0^{\Lambda^2}\,ds\,s\,\sigma_S^{m_R=0}(s)~,
\end{equation}
which is the customary definition of the vacuum condensate. However, in
terms of the nonperturbatively dressed quark propagator, equality between the
integrands requires the following {\it ad hoc} and mutually incompatible
``approximations'': $\forall s$,
\begin{mathletters}
\begin{equation}
\label{approxa}
B_{m_R=0}(s) \approx B_{m_R\neq0}(s)~;
\end{equation}
\begin{equation}
\sigma_S^{m_R=0}(s) \approx\sigma_S^{m_R\neq 0}(s)~;
\end{equation}
\begin{equation}
\label{approxc}
B_{m_R\neq0}(s)\approx m_R + B_{m_R=0}(s)~,
\end{equation}
\end{mathletters}
which yields Eq.~(\ref{GMOR}) when one makes the additional approximation
$N_\pi\approx f_\pi$, discussed above. That these are bad ``approximations''
is clear; for example, Eq.~(\ref{approxa}) has the effect of replacing a
convergence factor in the integrand by unity and it is incompatible with
Eq.~(\ref{approxc}). As elucidated in Ref.~\cite{RCP88}, Eq.~(\ref{GMOR})
can only be obtained if the (renormalised) current quark mass is treated
strictly as a perturbation. The inadequacy of Eqs.~(\ref{GMOR}) and
(\ref{convqbq}) is only exposed by a careful treatment of the Dyson-Schwinger
and Bethe-Salpeter equations.
We emphasise that Eq.~(\ref{expmass}) is completely consistent with the
general arguments of Ref.~\cite{GMOR}. It is derived from the generalised
ladder BSE and measures the expectation value of the explicit chiral symmetry
breaking term in the pion state under the approximation that
Eq.~(\ref{Goldthma}) is valid for $P^2\neq 0$, which is why the
right-hand-side involves only vacuum quantities: massless and massive,
renormalised, vacuum, dressed quark propagators.
We demonstrate below that Eq.~(\ref{expmass}) provides an extremely accurate
estimate of the pion mass obtained by solving the pion BSE in
generalised-ladder approximation. (See Eq.~(\ref{masslinear}) and
Table~\ref{tabobs}.)
\subsection{Solving the pion Bethe-Salpeter equation.}
\label{secbse}
In our numerical studies we are interested in the subtractively renormalised
Bethe-Salpeter amplitude, $F(p;P)$. This is defined in terms of the
regularised
amplitude $F'(p;P)$ via
\begin{equation}
F(p;P) \equiv F'(p;P)-F'(\mu,P)~,
\end{equation}
which, in generalised ladder approximation, is obtained as the solution of
\begin{eqnarray}
\label{bsepi}
F'(p;P) & = &
Z_2\,3\,C_2(R)\,\int^\Lambda\,\bar d^4q\,\Delta(p-q)\,
\left(q_+ \cdot q_- \,\sigma_V^+ \sigma_V^-
+ \sigma_S^+ \sigma_S^- \right)\,F(q;P)~.
\end{eqnarray}
It is clear that all corrections to free-field behaviour vanish at the
renormalisation point; i.e., $\left.F(p;P)\right|_{p^2=\mu^2}=0$.
Upon comparison with the DSE for $B(p^2)$ in
Sec.~\ref{secdse}, it is clear that in the chiral limit $(m_R=0)$ one has
\begin{equation}
\label{goldthm}
F(p;P)=B_{m_R=0}(p)~;
\end{equation}
i.e., that Goldstone's theorem is manifest.\cite{DS79}
One may solve Eq.~(\ref{bsepi}) numerically by introducing an eigenvalue,
$\lambda(P^2)$, on the right-hand-side. This yields an equation that has a
solution at every value of $P^2$. The equation can then be solved repeatedly
until that $P^2$ is found for which $\lambda(P^2)=1$.
The eigenvalue and eigenvector are determined by employing the Tschebyshev
decomposition
\begin{equation}
F(p;P) = \sum_{i=1}^\infty\,F_i(p^2,P^2)\,U_i(\cos\beta)
\end{equation}
and solving for the Tschebyshev moments of $F(p;P)$, which are obtained via
\begin{equation}
F_i(p^2,P^2) = \case{2}{\pi}\,\int_0^\pi\,d\beta\,\sin^2\beta\,
U_i(\cos\beta)\,F(p,P)~.
\end{equation}
In practice we only keep the lowest moment $F_0(p^2,P^2)$; neglecting the
coupling to the higher moments. This is a very good approximation for the
pion.\cite{JM93}
For an on-shell pion $P^2=-m_\pi^2$ and hence the right-hand-side of
Eq.~(\ref{bsepi}) samples the quark propagator at complex values of its
argument. To avoid solving the quark DSE off the real-$p^2$ axis we expanded
\mbox{$(q_+\cdot q_-\, \sigma_V^+ \sigma_V^- + \sigma_S^+ \sigma_S^-)$}
to O$(P^2)$ and solved the resulting equation, which involves derivatives of
the propagator at real-$p^2\geq 0$.
\section{Numerical Results and Phenomenology}
\label{secres}
We have two parameters: the mass scale $m_t$ in the gluon propagator, which
marks the transition point between the perturbative and nonperturbative
domains, Eq.~(\ref{delsmx}); and $m_R$, the renormalised current quark mass.
We varied these parameters in order to obtain the best $\chi^2$-fit to the
pion observables: $m_\pi$ [calculated using Eq.~(\ref{expmass})], the weak
pion decay constant\cite{review}
\begin{equation}
f_\pi = \frac{N_c}{4\pi^2}\int_0^{\Lambda^2}\,
ds\,s\,\case{1}{N_\pi}\,F_0(s,P^2)\,
\left[\sigma_V\sigma_S
+ \case{1}{2}s\left(\sigma_V'\sigma_S
- \sigma_V\sigma_S'\right)\right]~,
\end{equation}
$r_\pi$ and the $\pi$-$\pi$ scattering lengths: $a_0^0$, $a_0^2$, $a_1^1$,
$a_2^0$, expressions for which are given in Ref.~\cite{RCSI94}.
At each pair of parameter values the quark DSE was
solved numerically with $\mu = 48$~fm$^{-1}=9.47$~GeV, which is large enough
to be in the purely perturbative domain, and $\Lambda= 2^{18}$~fm$^{-1}\sim
5461\mu$. The results were almost independent of the cutoff; doubling it
leading only to a 3\% change in $f_\pi$, for example. Our results would have
been completely independent of $\Lambda$ if we had employed a vertex that
preserves multiplicative renormalisability. This observation provides a
quantitative measure of the violation of multiplicative renormalisability
when the rainbow approximation is used in Landau gauge. It is significantly
worse in other gauges. As remarked above, rainbow approximation entails a
loss of gauge covariance. Our experience suggests that our results would
change by no more than 10\% if we had used a dressed fermion--gauge-boson
vertex that ensured gauge covariance of the fermion
propagator.\cite{review,DMR94,BP94,Atkin93}
The formulae for the observables were then evaluated using the solution
obtained and the approximation that Eq.~(\ref{goldthm}) is valid for $m_R\neq
0$. After obtaining the optimal values of the parameters we recalculated the
observables using the pion Bethe-Salpeter amplitude calculated as described
in Sec.~\ref{secbse}. We found numerically that
\begin{equation}
\label{feqb}
F_0(p^2;P^2) \approx B_{m_R=0}(p^2)~.
\end{equation}
The best $\chi^2$-fit was obtained with
\begin{equation}
\label{params}
\begin{array}{lcl}
m_t = 0.69~\mbox{GeV} &\;\;\mbox{and}\;\; & m_R = 1.1~\mbox{MeV}~.
\end{array}
\end{equation}
We also carried out an extended $\chi^2$-fit where the ratio of the
coefficients of the two terms in Eq.~(\ref{delk}) was allowed to vary. In
this case the best $\chi^2$ was obtained with the value of $m_t$ in
Eq.~(\ref{params}) and a ratio that agreed with that in Eq.~(\ref{delk}) to
within 2\%. The data therefore requires both terms in the propagator and the
cancellation of long-range effects described in Eq.~(\ref{delsmx}).
The observables calculated with these parameter values are presented in
Table~\ref{tabobs}. One observes immediately that our one parameter model
for the gluon propagator provides a good description of low energy pion
observables. This improves upon the results of
Refs.~\cite{RCSI94,CDRpion,rhoomega,gpig}, in which the quark propagator was
parametrised and illustrates the connection, suggested in these articles,
that may be made between hadronic observables and the quark-quark
interaction.
We have made a direct comparison on the spacelike-$p^2$ axis of the numerical
solutions for $\sigma_V$ and $\sigma_S$ obtained herein with the parametrised
forms used in Ref.~\cite{CDRpion}. The agreement in form and magnitude is
very good, which suggests that the one parameter model gluon propagator will
also provide a good description of hadronic form factors.
One observes that the mass formula in Eq.~(\ref{expmass}) yields an accurate
estimate of the mass obtained by solving the pion BSE. We find that, with
parameters of Eq.~(\ref{params}), the right-hand-side of Eq.~(\ref{expmass})
is well described by
\begin{equation}
\label{masslinear}
m_\pi^2 N_\pi^2 = 2\,(0.45)^3\,m_R + (2.6)^2\,m_R^2 + 150\, m_R^3
\end{equation}
on the range $m_R\in [0,0.02]$~GeV, from which one may infer a value of
\mbox{$\langle \bar q q\rangle_{\mu} = -(0.45~{\rm GeV})^3$}. At the value of
$m_R$ in Eq.~(\ref{params}) the term linear in $m_R$ contributes almost 96\%
of the total. We see, therefore, that Eq.~(\ref{expmass}) entails $m_\pi^2
\propto m_R$, for small $m_R$, but that the constant of proportionality is not
given by the usual definition of the vacuum quark condensate,
Eq.~(\ref{convqbq}).
Our one parameter model for the gluon propagator explicitly {\it excludes}
the $\ln[k^2]$-corrections associated with the anomalous dimensions in QCD.
It is therefore inappropriate to directly compare $m_R(\mu)$ in
Eq.~(\ref{params}) with the QCD evolution of the commonly quoted value of
$m_{\mu=1 {\rm GeV}}\approx 7.5~{\rm MeV} $.~\cite{PDG94} (This entails that
the same is true of \mbox{$\langle \bar q q\rangle_{\mu}$}.) We note that
replacing $(\pi d)/k^2$ by $\alpha_S^{\rm two-loop}(k^2)/k^2$ in
Eq.~(\ref{delk}) would lead to a suppression of the tail of the quark mass
function, thereby requiring a larger value of $m_R$ to reproduce the pion
mass and a commensurate change in $m_t$. This represents a quantitative
improvement of our model but would not change its qualitative features.
\subsection{$\rho$-meson observables.}
\label{secrho}
We have employed our model gluon propagator in a preliminary study of
$\rho$-meson properties.
The regularised, generalised ladder approximation to the $\rho$-meson BSE is
\begin{equation}
\label{bserho}
F_\rho'(p;P) = Z_2\,3\,C_2(R)\,\int^\Lambda\,g^2\,D_{\mu\nu}(p-q)\,
\case{1}{12}{\rm tr}\left[\gamma_\alpha i\gamma_\mu\,S(q_+)\,
iT_\alpha(P)\,S(q_-)\gamma_\nu\right]\,F_\rho(q;P)~,
\end{equation}
where $[T_\mu(P) = \gamma_\mu + \gamma\cdot P P_\mu/m_\rho^2]$. The
subtractively renormalised amplitude is given by
\mbox{$F_\rho(p;P)=F_\rho'(p;P) - F_\rho'(\mu;P)$}. We neglected
the other Dirac-structures allowed in the vector-meson Bethe-Salpeter
amplitude. For the $\rho$-meson the error introduced by this truncation is
approximately 10\%.\cite{sepetal} The $\rho$- and $\omega$-mesons are
degenerate at this level of approximation. As for the pion, we project this
equation onto the lowest Tschebyshev moment and solve for $F_0(p^2,P^2)$,
neglecting the coupling to the higher moments. This is a good approximation
for the $\rho$-meson.\cite{JM93}
In this preliminary study we have only solved the quark DSE at real-$p^2$.
For an on-shell $\rho$-meson $P^2<0$ and hence Eq.~(\ref{bserho}) samples the
quark propagator at complex values of $p^2$. To obtain an approximate
solution of Eq.~(\ref{bserho}), without solving the quark DSE at
complex-$p^2$, we introduced an eigenvalue, $\lambda(P^2)$, on the
right-hand-side of Eq.~(\ref{bserho}) and solved this equation at spacelike
values of $P^2$, thereby obtaining $\lambda(P^2>0)$. For $0<P^2<10~{\rm
fm}^{-2}$ the results could be described by the quadratic (in $P^2$):
\begin{equation}
\lambda(P^2) = 0.44 - 0.021\,P^2 + 0.000076\,P^4
\end{equation}
with a standard-deviation of $0.000044$. We compared this with both linear
and cubic fitting forms: it provides a smaller standard-deviation than the
linear form and is monotonic, whereas the cubic is not. The value of $P^2$
for which this algebraic form of $\lambda(P^2)=1$ provides the mass estimate
presented in Table~\ref{tabobs}.
The calculated $\rho$-meson Bethe-Salpeter amplitude is much narrower in
momentum space than that of the pion, in agreement with the results of
Ref.\cite{JM93}.
The calculation of $g_{\rho\pi\pi}$ proceeds in a similar manner. In
generalised impulse approximation the $\rho\pi\pi$ coupling can be expressed
in terms of a nonlocal coupling functional, $N_\mu(p,q)$, which is discussed
in Ref.~\cite{PRC87}. This expression is used to evaluate
$g_{\rho\pi\pi}(P^2)$ on $0<P^2<10~{\rm fm}^{-2}$. The results were fitted
and extrapolated to the calculated mass-shell point. The best fit was
obtained with:
\begin{equation}
g_{\rho\pi\pi}(P^2)= 1.15 - 0.076\,P^2 + 0.0013\,P^4 - 0.000022 P^6~,
\end{equation}
giving a standard-deviation of 0.00023. This form provides a smaller value
of the standard-deviation than either a linear or quadratic form and is
monotonic whereas the quartic is not. The value obtained at the calculated
on-mass-shell point is given in Table~\ref{tabobs}.
These calculations are only a first step. They serve merely to indicate that
our one parameter model gluon propagator, which was fitted to pion
observables, can reasonably be expected to provide a good description of
other observables too.
\subsection{Confinement.}
We have also solved the quark DSE for real-$p^2<0$. There is no singularity
on the real-$p^2$ axis. The solution therefore does not have a Lehmann
representation and hence may be interpreted as describing a confined
particle.
A plot of $1/[p^2 + M(p^2)^2]$, which for a free particle would have a pole
at the mass-shell point, has a broad resonance-like peak centred on
$p^2\approx -(0.55)$~GeV$^2$. This admits an interpretation as the
``constituent-quark-mass'' in our model.
The form of our solution is suggestive of a pair of complex conjugate poles
or branch points with timelike real parts and large magnitude imaginary
parts. We have made no attempt to confirm this. A thorough study must
identify whether this structure is an artifact of the rainbow approximation,
which is known to be associated with unexpected behaviour of the fermion
propagator in the complex plane\cite{AB79,MH92,SC92,Maris,Stainsby} that is
modified when the vertex is dressed.\cite{BRW92}
\section{Summary and conclusions}
\label{secconc}
Using a confining, one parameter model form for the gluon propagator,
Eq.~(\ref{delk}), which incorporates the essence of the solution of
realistic, approximate gluon Dyson-Schwinger equations (DSEs), we solved the
renormalised, rainbow approximation quark DSE and subsequently the
renormalised, generalised ladder approximation $\pi$- and $\rho$-meson
Bethe-Salpeter equations (BSEs). We varied the parameter in the gluon
propagator, $m_t$, which is a mass scale that marks the transition between
the perturbative and nonperturbative domains, and the renormalised current
quark mass and obtained a good description of a range of $\pi$- and
$\rho$-meson observables. The value of $m_t$ was not known {\it a priori}.
Good agreement with the data {\em required} $m_t\sim 700$~MeV, which
corresponds to a length of $\sim 0.3$~fm.
In studying the pion BSE we were led to a mass formula
for the pion, Eq.~(\ref{expmass}), expressed solely in terms of the massive
and massless quark propagators. This formula provides a very accurate
estimate of the pion mass. It is valid to all orders in $m_R$, the
renormalised current quark mass, and for $m_R<20$~MeV the nonlinear terms
provide a contribution of no more than $\sim$10\%.
We obtained numerical solutions of the quark DSE on the timelike-$p^2$ axis,
which showed the quark propagator to have no singularity on the real-$p^2$
axis in our model. We found evidence to suggest that, as a function of
$p^2$, the quark propagator has a pair of complex conjugate poles or branch
points with timelike real parts and large imaginary parts. Such a propagator
does not have a Lehmann representation and admits the interpretation of
describing a confined particle.
Our study illustrates the manner in which the DSEs can be used to develop a
semi-phenomenological approach to QCD that incorporates the perturbative,
large spacelike-$k^2$ behaviour known from renormalisation group studies and,
via an economical parametrisation, extrapolates this into the
nonperturbative, small spacelike-$k^2$ domain. This efficacious,
nonperturbative approach allows for the correlation of a large range of
observables via very few parameters, which it may be possible to relate to
the fundamental parameters of QCD.
\acknowledgements
The preparation of this manuscript has benefited from useful discussions with
A. Bender, F. T. Hawes and A. G. Williams. This work was supported by the US
Department of Energy, Nuclear Physics Division; that of MRF under contract
No. DE-FG06-90ER40561 and that of CDR under contract No. W-31-109-ENG-38.
The calculations described herein were carried out using a grant of computer
time and the resources of the National Energy Research Supercomputer Center.
|
\section{Introduction}
The minimal supersymmetric standard model (MSSM) \cite{revs} has a
Higgs sector consisting of two doublets $H_1$ and $H_2$ coupling to the
down-type quarks and charged leptons, and to the up-type quarks
respectively. The particle content, other than Goldstone bosons, is
then two CP-even states $h$, $H$; one CP-odd state $A$; and a charged
scalar $H^{\pm}$. The requirements of supersymmetry and gauge
invariance then constrain the quartic Higgs coupling in terms of the
gauge couplings, and we are left with only two free parameters to
describe the whole Higgs sector. These are conventionally taken to be
$m_A$, the mass of the CP-odd state, and $\tan\beta=\nu_2/\nu_1$ where
$\nu_i=\langle H_i^0\rangle$, and $\nu_1^2+\nu_2^2=\nu^2=(174{\rm
GeV})^2$. It is then straightforward to derive the formulae $m_h^2\le
M_Z$ and $m_{H^{\pm}}\ge M_W$ where the first of these in particular is
greatly affected by radiative corrections\cite{higgsRC}. Thus the
MSSM is quite constrained and leads to the usual LEP2 Higgs discovery
limits in the $m_A-\tan\beta$ plane.
If the Higgs sector of the MSSM is extended by the addition of further
particle content then much of this predictivity is lost. This occurs
for two reasons. Firstly, the constraint on the quartic couplings in
the Higgs sector is not purely a result of supersymmetry but also an
artefact of having only doublets in the Higgs sector, making it
impossible to introduce extra Yukawa couplings through the Higgs
superpotential in a way consistent with gauge invariance. Increasing
the particle content and including such Yukawa couplings destroys both
the upper bound on $m_h$ and the lower bound on $m_{H^{\pm}}$. Secondly, the
extra states which we can introduce will mix with the states present in
the MSSM, and can alter both their couplings and their masses.
For example, in the next-to-MSSM (NMSSM) where there is an extra gauge
singlet state $N$ which only couples through the Higgs superpotential,
and where there are only trilinear terms in the superpotential
\cite{NMSSM}, there are now three CP-even neutral Higgs scalars and two
CP-odd neutral Higgs scalars, due to the real and imaginary components
of the additional singlet scalar. In this model the lightest CP-even
neutral scalar may be significantly heavier than in the MSSM, and to
make matters worse this lightest CP-even scalar may have diluted
couplings to the Z boson due to the admixture of singlet component
\cite{NMSSM,dilution}. However, it is possible to derive a bound on the
lightest CP-even state, both in this and in more general models
\cite{NMSSMbound,boundRC,dilution}. Furthermore, in the limit that the
lightest CP-even Higgs boson is completely decoupled (and hence we
might think that the bound does not tell us anything about states which
could be detected), the bound applies instead to the second lightest
CP-even Higgs boson, while for light states which are merely weakly
coupled one can derive precise bounds on their heavier partners. These
become closer to the bound on the lightest as the singlet component of
the lightest becomes greater \cite{Kam}.
Our intention in this paper is to discuss how much of the parameter
space of such a general SUSY model can be covered by searches at LEP2,
comparing them closely with the corresponding results for the MSSM. We
shall consider the most general possible model with a gauge singlet and
show that LEP2 can discover CP-even Higgs states in any part of the
$m_A-\tan\beta$ plane, and can exclude significant parts of this plane
even with the most general possible mixing.
We begin with a short review of the MSSM and its simplest extensions
and how they affect the mass matrices. In section 4, we discuss the
conventional bound on the lightest CP-even state in this model. In
section 5, a number of search strategies are discussed, and we consider
how effective the various MSSM searches are in an extended model.
Finally, in section 6 we present results applicable to LEP2. Section 7
is the conclusion.
\section{The MSSM}
Here the superpotential is:
\begin{equation}
W_{MSSM}=-\mu H_1H_2+\ldots
\end{equation}
where $H_1H_2=H_1^0H_2^0-H_1^-H_2^+$.
The superpotential leads to the tree-level Higgs potential:
\begin{eqnarray}
V_{MSSM} & = & m_{1}^2|H_1|^2 +m_{2}^2|H_2|^2 +
m_{12}^2(H_1H_2+H.c.) \nonumber \\
& + & \frac{1}{8}(g^2+g'^2)(|H_1|^2-|H_2|^2)+
\frac{g^2}{2}|H_1^{\ast}H_2|^2
\end{eqnarray}
where $m_{1}^2=m_{H_1}^2+\mu^2$, $m_{2}^2=m_{H_2}^2+\mu^2$,
and $m_{12}^2=-B\mu$ where $B$ and $m_{H_i}^2$ are the usual soft
SUSY-breaking terms.
The tree-level CP-odd mass-squared matrix is
\begin{equation}
M_{A}^2=
\left(\begin{array}{cc}
m_{12}^2t_{\beta}& m_{12}^2 \\
m_{12}^2 & m_{12}^2/t_{\beta}
\end{array} \right)
\label{CP-odd}
\end{equation}
$M_A^2$ is diagonalised by
$UM_{A}^2U^{\dagger}={\rm diag}(m_A^2,0)$ where $U$ is given by
\begin{equation}
U=\left(\begin{array}{cc}
s_{\beta} & c_{\beta} \\
-c_{\beta} & s_{\beta}
\end{array} \right)
\label{U}
\end{equation}
and $m_A$ is given by $m_A^2=m_{12}^2/(s_{\beta}c_{\beta})$ where
$s_{\beta}=\sin\beta$, $c_{\beta}=\cos\beta$ and $t_{\beta}=\tan\beta$.
The tree-level charged Higgs mass squared matrix is
\begin{equation}
M_{H^{\pm}}^2=
\left(\begin{array}{cc}
M_{W}^2s_{\beta}^2 + m_{12}^2t_{\beta}
& M_{W}^2s_{\beta}c_{\beta} + m_{12}^2 \\
M_{W}^2s_{\beta}c_{\beta} + m_{12}^2
& M_{W}^2c_{\beta}^2 + m_{12}^2/t_{\beta}
\end{array} \right)
\label{Hpm}
\end{equation}
$M_{H^{\pm}}^2$ is diagonalised by
$UM_{H^{\pm}}^2U^{\dagger}={\rm diag}(m_{H^{\pm}}^2,0)$ where
$U$ is as before and
$m_{H^{\pm}}$ is given by
$m_{H^{\pm}}^2=M_W^2+m_A^2$.
The tree-level CP-even Higgs mass-squared matrix is,
after substituting the tree-level expression for $m_{12}^2$
in terms of $m_A^2$,
\begin{equation}
M^2=
\left(\begin{array}{cc}
M_{Z}^2c_{\beta}^2 + m_{A}^2s_{\beta}^2
& -(M_{Z}^2+m_{A}^2)s_{\beta}c_{\beta} \\
- (M_{Z}^2+m_A^2)s_{\beta}c_{\beta}
& M_{Z}^2s_{\beta}^2 + m_{A}^2c_{\beta}^2
\end{array} \right)
\label{CP-even}
\end{equation}
$M^2$ is diagonalised by
$VM^2V^{\dagger}={\rm diag}(m_{H}^2,m_{h}^2)$ where
\begin{equation}
V=
\left(\begin{array}{cc}
c_{\alpha} & s_{\alpha} \\
-s_{\alpha} & c_{\alpha}
\end{array} \right)
\label{V}
\end{equation}
where $-\pi/2\leq \alpha \leq 0$ and $m_{h}^2,m_{H}^2$
are given by
\begin{equation}
m_{H,h}^2=\frac{1}{2}(m_A^2+M_Z^2)
\pm \frac{1}{2} \sqrt{(m_A^2+M_Z^2)^2-4m_A^2M_Z^2c_{2\beta}^2}
\label{hmass}
\end{equation}
and the two angles are related by:
\begin{equation}
\tan 2\alpha =\tan 2\beta \frac{(m_A^2+M_Z^2)}{(m_A^2-M_Z^2)}
\end{equation}
In the above conventions, the ZZh coupling has an additional
factor of $\sin (\beta -\alpha)$ relative to the standard model,
and the ZZH coupling has a factor of $\cos (\beta - \alpha )$.
The ZhA coupling is proportional to $\cos (\beta - \alpha )$.
It is interesting to observe that if we do not diagonalise $M^2$ but
instead rotate to a primed basis where the second Higgs doublet does
not have any VEV, then the 11 element of the matrix no longer contains
any $m_A$ dependence. Since the smallest eigenvalue of a real symmetric
matrix is less than or equal to the smallest diagonal term, this gives
a useful upper bound on the lightest CP-even Higgs mass. This rotation
is given by $WM^2W^{\dagger}=M'^2$ where
\begin{equation}
W=
\left(\begin{array}{cc}
c_{\beta} & s_{\beta} \\
-s_{\beta} & c_{\beta}
\end{array} \right)
\label{W}
\end{equation}
and the bound is then given by
\begin{equation}
m_h^2\leq M^{\prime 2}_{11}=M_Z^2\cos^22\beta.
\end{equation}
Starting from the primed basis
the CP-even Higgs matrix is diagonalised by
\begin{equation}
XM'^2X^{\dagger}={\rm diag}(m_{H}^2,m_{h}^2)
\end{equation}
where $X=VW^{\dagger}$.
Note that in this basis the additional factor in the ZZh coupling
relative to the standard model of $\sin(\beta-\alpha)$ is simply
$X_{21}$, and similarly the ZZH factor of $\cos(\beta-\alpha)$
is just $X_{11}$.
\section{The MSSM plus a Singlet}
Now let us consider a very general Higgs superpotential of the form
\begin{equation}
W=W_{MSSM}+\lambda NH_1H_2+f(N)
\end{equation}
where $f$ is an arbitrary holomorphic function of $N$, the singlet
field, and $H_1$ and $H_2$ are the usual Higgs doublets coupling to
down and up quarks respectively. Since the singlet does
not have gauge couplings we may then write
\begin{eqnarray}
V &=&V_{MSSM}+\lambda^2 \bigl |H_1H_2 \bigr |^2 \cr
&& +\Bigl(\lambda (N+\bar N)\mu +\lambda^2|N|^2 \Bigr)
\Bigl ( |H_1|^2+|H_2|^2 \Bigr ) \cr
&& +\Bigl ( \lambda \frac{\partial \bar f}{\partial \bar N}
- \lambda N A_\lambda \Bigr ) H_1H_2 +h.c.
+\cdots
\end{eqnarray}
where the ellipsis indicates terms with no dependence on $H_1$ or
$H_2$. Given that $V_{MSSM}$ includes terms $\mu^2(|H_1|^2+|H_2|^2)$
and $B\mu H_1H_2$ it is then clear that we can account for the
effects of all the singlet dependent terms on the upper 2$\times$2
block of the mass matrices (CP-odd, CP-even, and charged) purely by
redefinitions of $B\rightarrow B'$ and
$\mu\rightarrow \mu'$ in terms of the VEV of the singlet
$<N>=x$ and
by including the effects of the $\lambda$ dependent quartic term.
We further note that we could obtain identical results with an arbitrary
number of singlets, since we can always rotate so that only one of them
couples to $H_1H_2$ and the remainder simply complicate the form of $f$
and of the soft terms, which we are essentially regarding as arbitrary.
With one singlet, the most general superpotential is
\begin{equation}
W =-\mu H_1H_2 + \lambda NH_1H_2 - \frac{k}{3}N^3
+ \frac{1}{2}\mu^{\prime}N^2 + \mu^{\prime\prime}N
\label{Wgeneral}
\end{equation}
In the limit that $\mu,\mu^{\prime}\mu^{\prime\prime}\rightarrow 0$, the
above superpotential reduces to that of the NMSSM\cite{NMSSM}, while if
$N$ is removed it reduces to that of the MSSM.
The mass matrix for the CP-odd scalars
in the basis $(H_1,H_2,N)$ is a complicated
3-dimensional generalisation of $M_A^2$ in the MSSM.
With only minimal risk of ambiguity we shall use the same notation
for such generalisations as for the MSSM.
The tree-level CP-odd mass squared matrix becomes :
\begin{equation}
M_{A}^2=
\left(\begin{array}{ccc}
m_{12}'^2t_{\beta}& m_{12}'^2 & . \\
m_{12}'^2 & m_{12}'^2/t_{\beta} & . \\
. & . & .
\end{array} \right)
\label{CP-odd2}
\end{equation}
where $m_{12}'^2= -B'\mu'$.
The CP-odd matrix has its Goldstone modes isolated by
\begin{equation}
UM_{A}^2U^{\dagger} \equiv
\left(\begin{array}{ccc}
0 & 0 & 0 \\
0 & m_A'^2 & . \\
0 & . & .
\end{array} \right)
\end{equation}
where $U$ is given by
\begin{equation}
U=
\left(\begin{array}{ccc}
c_{\beta} & -s_{\beta} & 0 \\
s_{\beta} & c_{\beta} & 0 \\
0 & 0 & 1
\end{array} \right)
\label{U2}
\end{equation}
Clearly $m_A'$ is the analogue of $m_A$ of the MSSM, and
$m_A'^2=m_{12}'^2/(s_{\beta}c_{\beta})$. The entries in these matrices
represented by dots are complicated functions of soft terms and
parameters from $f$ and the extended soft potential, and since $f$ is
arbitrary they are essentially unconstrained. We define the matrix
$V_A$ which diagonalises $M_A^2$ as follows
\begin{equation}
(V_AU)M_{A}^2(V_AU)^{\dagger} = {\rm diag}(0,m_{A_1}^2,m_{A_2}^2)
\end{equation}
where we have taken $m_{A_1}<m_{A_2}$ and
$V_A$ can be represented in terms of one mixing angle $\gamma$
\begin{equation}
V_A=
\left(\begin{array}{ccc}
1 & 0 & 0 \\
0 & c_{\gamma} & s_{\gamma} \\
0 & -s_{\gamma} & c_{\gamma}
\end{array} \right)
\label{gamdef}
\end{equation}
As regards the charged Higgs sector,
the singlets obviously cannot mix with charged scalars, and we find
that (at tree-level) the mass of the charged Higgs is given by
\begin{equation}
m_{H^{\pm}}^2=m_A'^2+M_W^2-\lambda^2\nu^2
\end{equation}
We can immediately obtain the MSSM results by simply setting
$\lambda=0$ and removing all of the singlet terms from the mass
matrices. Clearly a non-zero $\lambda$ tends to reduce the
charged scalar masses which can be arbitrarily small.
The CP-even mass squared matrix is now a 3$\times$3 matrix in the
basis $(H_1,H_2,N)$ which may be written as
\begin{equation}
M^2=
\left(\begin{array}{ccc}
M_Z^2c_{\beta}^2+m_A'^2s_{\beta}^2 &
-(M_Z^2+m_A'^2-2\lambda^2\nu^2)s_{\beta}c_{\beta} & . \\
-(M_Z^2+m_A'^2-2\lambda^2\nu^2)s_{\beta}c_{\beta} &
M_Z^2s_{\beta}^2+m_A'^2c_{\beta}^2 & . \\
. & . & .
\end{array} \right)
\label{CPevenmass}
\end{equation}
where, as before, the dots indicate complicated entries.
As in the MSSM the CP-even Higgs matrix is diagonalised by
\begin{equation}
VM^2V^{\dagger}={\rm diag}(m_{h_1}^2,m_{h_2}^2,m_{h_3}^2)
\label{diag3}
\end{equation}
where
$V$ is some complicated 3$\times$3 matrix.
Note that we have now chosen to order the
mass eigenstates as
\begin{equation}
m_{h_1}<m_{h_2}<m_{h_3}
\label{ordering}
\end{equation}
so that the definition of $V$ here
does {\em not} reduce to that of $V$ defined
earlier where the eigenvalues are conventionally ordered oppositely.
For example in the MSSM if we had required that the
CP-even Higgs matrix were diagonalised as
$VM^2V^{\dagger}={\rm diag}(m_{h}^2,m_{H}^2)$
then $V$ would have to include a re-ordering of the mass eigenstates,
and would have been given by
\begin{equation}
V =
\left(\begin{array}{cc}
-s_{\alpha} & c_{\alpha} \\
-c_{\alpha} & -s_{\alpha}
\end{array} \right)
\end{equation}
where $\alpha$ is the angle defined earlier in the MSSM.
It is to this form that our generalised $V$ must
reduce in the MSSM limit.
As in the MSSM we observe that if we do not diagonalise
$M^2$ but instead rotate to a basis where the second
Higgs doublet does not have any VEV, then the 11 element
of the matrix gives a useful upper bound on the lightest
CP-even Higgs mass \cite{NMSSMbound}. This rotation is given by
$WM^2W^{\dagger}=M'^2$ where
\begin{equation}
W=
\left(\begin{array}{ccc}
c_{\beta} & s_{\beta} & 0 \\
-s_{\beta} & c_{\beta} & 0 \\
0 & 0 & 1
\end{array} \right)
\label{W2}
\end{equation}
and the bound is given by
\begin{equation}
m_{h_1}^2 \leq M'^2_{11}=
M_Z^2\cos^2 2\beta + \lambda^2\nu^2\sin^22\beta
\label{bound1}
\end{equation}
The reason why this bound is useful is simply that
it has no ${m'_A}^2$ dependence.
For example a less useful bound is,
\begin{equation}
m_{h_1}^2 \leq M'^2_{22}=
(M_Z^2- \lambda^2\nu^2)\sin^22\beta + {m'_A}^2
\label{bound1lessuseful}
\end{equation}
Starting from the primed basis
the CP-even Higgs matrix is diagonalised by
\begin{equation}
XM'^2X^{\dagger}={\rm diag}(m_{h_1}^2,m_{h_2}^2,m_{h_3}^2)
\label{XX}
\end{equation}
where $X=VW^{\dagger}$.
We shall define the relative couplings
$R_i\equiv R_{ZZh_i}$ as the $ZZh_i$ coupling in units of the standard model
$ZZh$ coupling, and similarly we shall define a $Zh_iA_j$ coupling
factor $R_{Zh_iA_j}$.
For example $R_{ZZh_1}$ is a generalisation of $\sin (\beta - \alpha)$
and the $R_{Zh_1A_i}$ are generalisations of $\cos (\beta - \alpha)$
in the MSSM. In our notation we find, using the results of
Ellis {\em et al} \cite{NMSSM},
\begin{eqnarray}
R_{i} & = & \cos \beta V_{i1} + \sin \beta V_{i2} \nonumber \\
& = & (VW^{\dagger})_{i1}=X_{i1}.
\label{ZZh}
\end{eqnarray}
The $Zh_iA_j$ coupling factorises
into a CP-even factor $S_i$ and a CP-odd factor $P_j$\cite{NMSSM}
\begin{equation}
\label{ZhA}
R_{Zh_iA_j}=S_i P_{j}
\end{equation}
where
\begin{eqnarray}
S_i & = & -V_{i1}\sin \beta + V_{i2}\cos \beta \nonumber \\
& = & (WV^{\dagger})_{2i} = X_{i2}
\label{CP-evenfactor}
\end{eqnarray}
and
\begin{equation}
P_j = (V_AU)_{(j+1)2}\cos \beta +(V_AU)_{(j+1)1}\sin \beta
\label{CP-oddfactor}
\end{equation}
Eq.(\ref{CP-oddfactor}) implies the simple intuitive results
\begin{equation}
P_1=\cos \gamma, \ \ P_2=-\sin \gamma
\label{P}
\end{equation}
where $\gamma$ defined in Eq.(\ref{gamdef})
is the angle which controls the amount of singlet
mixing in the CP-odd sector.
\section{Implementing the Bound on $h_1$}
In this section we shall consider the absolute upper bound
on the mass of the lightest CP-even state in this model,
using Eq.(\ref{bound1}) plus radiative corrections.
For notational ease we shall first define
\begin{equation}
\Lambda^2 \equiv M'^2_{11}.
\label{Lambda}
\end{equation}
Thus the bound is simply
\begin{equation}
m_{h_1}^2 \leq \Lambda^2
\end{equation}
Clearly $\Lambda^2$ is a function of $\tan \beta$ and $\lambda$,
$\Lambda^2(\tan \beta,\lambda)$,
and to find the absolute upper bound we must maximise this
function so that
\begin{equation}
m_{h_1}^2 \leq \Lambda_{max}^2
\end{equation}
where $\Lambda_{max}^2$ is the maximum value of
$\Lambda^2(\tan \beta$ and $\lambda)$,
\begin{equation}
\Lambda_{max}^2 \equiv max(\Lambda^2(\tan \beta,\lambda)).
\end{equation}
Our task in this section is therefore to find
$\Lambda_{max}^2$ in the presence of radiative corrections.
It is well known that radiative corrections, which we have so far
ignored, drastically affect the bound \cite{boundRC}.
In order to deal with these radiative corrections many techniques have
been proposed \cite{higgsRC}. Here we shall follow the method proposed
in ref.\cite{ceqw}, which we shall briefly review. According to this
method a scale $M_{susy}$ is defined by $M_{susy}^2=(m_{\tilde
t_1}^2+m_{\tilde t_2}^2)/2$ where $m_{\tilde t_i}$ are the stop mass
eigenvalues, and the couplings $h_t$ and $h_b$ found at this scale.
Here the squarks are integrated out, leaving an effective Higgs
potential involving Yukawa couplings with boundary conditions which
differ from the tree-level ones by some finite corrections
\cite{higgsRC}. Then the Yukawa couplings (including $h_t$) are run
from $M_{susy}$ down to $m_t$. Below $m_t$ the Yukawa couplings are
approximately constant, and so the Higgs potential may be minimised at
this scale. The effects of the RG running have been estimated
analytically, leading to fully analytic results for the Higgs masses
which agree well with more elaborate methods \cite{ceqw}.
For $\lambda=0$ (as in the MSSM) the bound is obviously largest for
$\cos^2(2\beta)=1$. Given the restricted range $\pi/4 \leq \beta \leq
\pi/2$ this implies that the bound is maximised for $\beta = \pi/2$,
$\cos 2\beta =-1$, corresponding to $\tan \beta = \infty$ (or in
practice its maximum allowed value). However, for sufficiently large
$\lambda$ the tree-level bound will be maximised for $\tan\beta$ equal
to its minimum value. In order to obtain an absolute upper bound, we
must thus derive an upper limit on $\lambda$, which will turn out to be
a function of $\tan\beta$. This can be done by demanding that all of
the Yukawa couplings remain perturbative up to the GUT scale
\cite{NMSSMbound,boundRC,dilution}.
We have used the newly discovered mass of the top quark \cite{CDF} and
the two loop SUSY RG equations in the NMSSM \cite{epic} to derive an
upper bound on $\lambda$ (defined at $m_t$, and calculated by requiring
all Yukawa couplings $h_t$, $h_b$ and $\lambda$ to remain perturbative
up to a scale of $10^{16}$GeV), which we show as a function of
$\tan\beta$ in Figure 1. Results are shown for various values of $m_t$
and $\alpha_3(M_Z)$, and it is likely that the errors from these two
measurements will greatly dominate any other uncertainties of our
calculation. The main features of this graph are that $\lambda$ has an
upper bound of around 0.5 to 0.7 for intermediate $\tan\beta$, while
for large (small) $\tan\beta$ the triviality constraints on $h_b$
($h_t$) force this upper bound very rapidly to zero.
In Figure 1, we have assumed that the effect of any other Yukawa
couplings in the singlet sector of the superpotential is negligible,
but in fact such couplings can have a significant impact, as shown in
Figure 2, where we fix $\tan\beta$ and explore the dependence of the
upper bound on $\lambda$ as a function of the coupling $k$ defined in
Eq.~(\ref{Wgeneral}) for a model with only one singlet. The most
important feature of this figure is that it clearly displays how adding
extra Yukawa couplings will always {\em reduce} the bound on $\lambda$,
since these will always appear with the same sign in the RG equations.
The reason why $\lambda$ falls so rapidly to zero above a certain value
of $k$ is that here $k$ is approaching its triviality limit, while for
small $k$ its impact on the upper bound of $\lambda$ is negligible.
In Figures 3a to 3c, we plot the bound $\Lambda$ as a function of
$\tan\beta$, $M_{susy}$, and $m_t$ respectively. For each of these
figures, we set the squark soft masses squared $m_Q^2$, $m_T^2$, and
$m_B^2$ to be equal, and we allow the trilinear squark--Higgs coupling
$A_t$, $\lambda$ and, except in Figure 3a, $\tan\beta$ to take on the
values which maximise the bound. In Figures 3a,3c we select $M_{susy}$
as defined previously to be 1TeV, and in Figures 3a and 3b we select
$m_t$=175GeV. We display bounds for both the MSSM (lower dashed line)
and its extension with a singlet. Note that for extreme values of
$\tan\beta$ the maximum allowed value of $\lambda$ falls rapidly to
zero, and so the two bounds are then the same, while the same happens
for very large $m_t$ since here $h_t$ is again close to triviality.
Thus the large value of $m_t$ preferred by $CDF$ \cite{CDF} means that
the two bounds are now within around 8GeV. However, it is important to
realise that one of the main problems for LEP2 will be that for the
MSSM small $\tan\beta$ always gives a light CP-even Higgs, while this
need no longer be true in a non-minimal extension, thus seriously
damaging the prospects for particle searches.
\section{LEP2 Search Strategies}
\subsection{$Z\rightarrow Zh_i$ and General Bounds on CP-even Higgses}
Recently upper bounds on all three neutral CP-even Higgs scalars in the
general extension of the MSSM with a single gauge singlet superfield
have been derived \cite{Kam}. The basic observation which follows from
Eq.(\ref{XX}) and the definition $M'^2_{11}\equiv \Lambda^2$ is:
\footnote{Note that $X$ is real if the Higgs sector conserves CP.}
\begin{eqnarray}
\Lambda^2&=& X_{11}^2m_{h_1}^2+
X_{21}^2m_{h_2}^2+ X_{31}^2m_{h_3}^2 \cr
&=&R_1^2m_{h_1}^2+R_2^2m_{h_2}^2+R_3^2m_{h_3}^2
\label{basic}
\end{eqnarray}
where we have used the fact that $R_i=X_{i1}$.
Eq.(\ref{basic}) together with Eq.(\ref{ordering})
clearly implies
\begin{equation}
\Lambda^2\geq (R_1^2+R_2^2+R_3^2)m_{h_1}^2 \nonumber
\end{equation}
and given that
\begin{equation}
R_1^2+R_2^2+R_3^2=1
\label{unitarity}
\end{equation}
by the unitarity of $X$,
we find the usual bound in Eq.(\ref{bound1}),
\begin{equation}
m_{h_1}^2\leq \Lambda^2
\end{equation}
Using a similar argument,
a bound on the $h_2$ mass may be obtained from
Eq.(\ref{basic}) :
\begin{equation}
m_{h_2}^2 \leq \frac{\Lambda^2-R_1^2m_{h_1}^2}{1-R_1^2}
\label{bound2}
\end{equation}
$m_{h_3}^2$ is given from Eq.(\ref{basic}) by
\begin{equation}
m_{h_3}^2 =
\frac{\Lambda^2-R_1^2m_{h_1}^2-R_2^2m_{h_2}^2}
{1-R_1^2-R_2^2}
\label{equality3}
\end{equation}
from which we can extract a bound
\begin{equation}
m_{h_3}^2 \leq
\frac{\Lambda^2-(R_1^2+R_2^2)m_{h_1}^2}{1-R_1^2-R_2^2}
\label{bound3}
\end{equation}
As we shall see, Eqs.(\ref{bound2}) and (\ref{bound3}) are useful upper
bounds on $m_{h_2}$ and $m_{h_3}$ in terms of $m_{h_1}^2$, $R_1$, $R_2$
and $\Lambda^2\equiv M'^2_{11}$ (which is just the bound on
$m_{h_1}^2$). These bounds tightly constrain the spectrum in terms of
the couplings and hence will allow us to study the reach of colliders
when the mixing parameters take on arbitrary values.
The above general bounds are useful
since $h_1$ may be light but very weakly
coupled. In the MSSM this can happen when $\sin (\beta - \alpha)$
becomes small. In the present model $R_1$ can also be small because
$h_1$ may contain a large singlet component. In this case
it does not matter how light $h_1$ is since its couplings may be
so weak that it can never be detected. At first sight this seems to be
a catastrophe for LEP2 where in the MSSM $h_1$ is usually the easiest
state to find. However, if $R_1\approx 0$ we may simply ignore $h_1$
and concentrate on $h_2$ which then becomes the lightest physically
coupled CP-even state. Moreover it can be seen from Eq.(\ref{bound2})
that if $R_1\rightarrow 0$ then $h_2$ must satisfy
\begin{equation}
m_{h_2}^2 \leq \Lambda^2
\label{bound22}
\end{equation}
which is just the bound on $m_{h_1}$. In other words, if the lightest
CP-even state is essentially invisible, then the second lightest CP-even
state must be quite light, and may be within the reach of LEP2.
Similarly if both $R_1$ and $R_2$ are $\approx 0$ then Eq.(\ref{equality3})
implies that
\begin{equation} m_{h_3}^2=\Lambda^2
\label{bound33}
\end{equation}
The above restrictions on the Higgs masses
and couplings imply that a collider of a
given energy and integrated luminosity will be able to exclude values
of $\Lambda$ below which
it is impossible for all the Higgses to have simultaneously
escaped detection.
We shall discuss this further in Section 5.3.
\subsection{Exclusion Plots in the $R^2-m_h$ Plane at LEP}
In this sub-section we we shall be concerned with exclusion plots in
the $R^2-m_h$ plane such as those that have already been obtained at
LEP1 and whose possible form at LEP2 has been predicted, where
$R^2\equiv \sin^2(\beta - \alpha)$ in the MSSM. LEP1 has not discovered
a CP-even Higgs boson \cite{LEPex}, and this non-observation has
enabled a $95\%$ CL exclusion limit to be extracted on the value of the
square of the relative $ZZh$ coupling $R^2$, as a function of the
CP-even Higgs mass $m_h$ extending out to $m_h=65$ GeV\cite{mRLEP1}.
Similar exclusion plots at LEP2 will extend the Higgs mass range to
$m_h>65$ GeV leading to possible exclusion curves whose precise shape
will depend on the energy and integrated luminosity of
LEP2\cite{janot}. Three different sets of LEP2 machine parameters have
been considered corresponding to energies and integrated luminosities
per experiment of \cite{janot}
\begin{eqnarray}
\sqrt{s} & = & 175 \ {\rm GeV},\ \ \int{\cal{L}}=150
\ {\rm pb}^{-1} \nonumber \\
\sqrt{s} & = & 192 \ {\rm GeV},\ \ \int{\cal{L}}=150
\ {\rm pb}^{-1} \nonumber \\
\sqrt{s} & = & 205 \ {\rm GeV},\ \ \int{\cal{L}}=300
\ {\rm pb}^{-1} .
\label{LEP2}
\end{eqnarray}
The existing LEP1 \cite{mRLEP1} and anticipated LEP2 \cite{janot}
exclusion plots are combined in Fig.4a, for the three different LEP2
scenarios, and assuming a 100\% $h_1$ branching fraction into
$\bar{b}b$. The very steep rise of the contours means that, if a
standard model-like Higgs boson of a certain fixed mass can be produced
at LEP (corresponding to $R^2=1$) then reducing $R_{ZZh}^2$ has little
effect on its visibility until $R_{ZZh}^2$ becomes quite small. The
relatively flat LEP1 regions of the contours correspond to minimum
threshold values of $R_{ZZh}^2$ below which nothing can be seen for any
value of the mass.
The above exclusion plots in Fig.4a are the result of a sophisticated
Monte Carlo simulation of the LEP detectors. It is interesting to
observe that these contours roughly correspond to the value of the
coupling $R^2$ which would yield 50 Higgs events of a given mass. The
plot of contours of 50 $h$ events as a function of $m_{h}$ and $R^2$
for the three different LEP2 machine scenarios is shown for comparison
in Fig.4b. Note that this simple parameterisation of the exclusion
plots of Fig.4a fails badly for $m_h \approx M_Z$, and also the actual
exclusion limits from LEP1 are significantly stronger than our LEP2
approximation for small Higgs masses. Nevertheless Fig.4b provides a
useful caricature of the exclusion plots in Fig.4a.
Finally note that the exclusion plots in Fig.4 may be interpreted for
each of the three CP-even Higgs bosons in this model taken separately.
In other words the disallowed region to the upper and left of the
lines will exclude values of ${R_1}^2$ for a given $m_{h_1}$, and
similarly for the other two CP-even Higgs bosons.
\subsection{Excluded Values of $\Lambda$}
Although there are a large number of parameters in this model, and
three CP-even Higgs bosons, it proves possible to exclude values of
$\Lambda$ smaller than a certain amount, where $\Lambda$ is defined as
above to be the upper bound on the lightest CP-even Higgs boson, and
is to be regarded as a function of $\lambda, \tan \beta$ and all the
other parameters which enter the calculation of radiative corrections
such as the top mass, the squark masses, and so on. If LEP2 does not
discover any CP-even Higgs bosons then exclusion plots such as those in
Fig.4a may be produced. We now show that such exclusion plots may be
used to place an excluded lower limit on the value of $\Lambda$ in this
model. If the excluded lower limit on $\Lambda$ reaches the theoretical
upper limit of about 146 GeV (dependent on the top mass and SUSY
spectrum as discussed above) then the model is excluded. For example, a
linear collider of energy 300 GeV should be able to exclude the
model \cite{Kam}. We now describe how LEP may be used to exclude values
of $\Lambda$ in this model.
Clearly a specified value of $\Lambda$ is consistent with many sets of
values of the parameters $m_{h_i}$, $R_i$ subject to the bounds
discussed in section 5.1. In general some of these sets of $m_{h_i}$,
$R_i$ will lead to one or more CP-even Higgs boson in the excluded
(upper left) part of Fig.4a and some sets will lead to Higgs bosons
only in the allowed region. This means that, for a given $\Lambda$, one
or more CP-even Higgs bosons may or may not be discovered at LEP2,
depending on the values of the other (complicated unknown) parameters
in the model. According to our discussion in Section 5.1, it is clear
that as $\Lambda$ is reduced, more and more of the allowed sets of
$m_{h_i}$, $R_i$ will move into the excluded part of Fig.4a. As
$\Lambda$ is reduced below some critical value all the sets of values
of $m_{h_i},R_i$ will eventually fall into the excluded region of
Fig.4a. This critical value of $\Lambda$ is the maximum value of
$\Lambda$ excluded by LEP. This implies that the maximum excluded value
of $\Lambda$ is determined by whether the ``worst case'' ({\em i.e.}
hardest to see experimentally) values of $m_{h_i},R_i$ consistent with
this value of $\Lambda$ lie within the allowed region or not.
In order to determine the ``worst case'' parameters we first fix
$\Lambda$ at some specified value (less than its maximum as determined
in section 4) and then we scan over values of $m_{h_1}$ from zero up to
$\Lambda$. For each value of $m_{h_1}$ we set $R_1$ equal to the
maximum allowed value as shown in Fig.4a (or 1 if this would require
$R_1>1$). For $\Lambda$, $m_{h_1}$ and $R_1$ fixed as above, we then
scan over $m_{h_2}$ from $m_{h_1}$ up to its upper bound which is now
fixed according to Eq.(\ref{bound2}). For each value of $m_{h_2}$ we
set $R_2$ equal to its maximum allowed value as before. For $\Lambda$,
$m_{h_1}$, $m_{h_2}$, $R_1$ and $R_2$ fixed as above, $m_{h_3}$ and
$R_3$ are now completely specified by Eqs.(\ref{unitarity}) and
(\ref{equality3}). If $R_3$ is larger than its maximum allowed value
(according to Fig.4b), then the ``worst case'' with the given values of
$m_{h_1}$ and $m_{h_2}$ is excluded. If all the ``worst case'' points
in the $m_{h_1}$, $m_{h_2}$ scan turn out to be excluded then we
conclude that this value of $\Lambda$ is excluded by LEP.
According to Eq.(\ref{bound1}), the value of $\Lambda^2\equiv
{M'_{11}}^2$ is a function of $\lambda$ and $\tan \beta$, plus
parameters which enter in the radiative corrections. For a given value
of $\lambda$, and given radiative correction parameters, the exclusion
limit on $\Lambda$ may be interpreted as an exclusion limit on $\tan
\beta$ in this model, independently of $m'_A$. Thus the present
analysis will yield a horizontal exclusion line in the $m'_A-\tan
\beta$ plane, as we shall see in Section 6.
If the observability criterion above is replaced by the
approximate exclusion limit obtained from
50 higgs events regardless of branching
fractions, as in Fig.4b, then the above algorithm can be solved
analytically. In this simple case,
the ``worst case'' for LEP2 to discover would be one
where the largest of the three cross-sections $\sigma_i$ was minimised,
where
\begin{equation}
\sigma_i= \sigma (e^+e^- \rightarrow Zh_i)=
\sigma_{SM} (e^+e^- \rightarrow Zh)|_{m_h=m_{h_i}}R_i^2
\end{equation}
It is not hard to show that this
will occur when $\sigma_1=\sigma_2=\sigma_3$, and hence when
$\sum_i\sigma_{SM}(m_{h_i})R_i^2$
is minimised.
Using the constraints in Eqs.(\ref{basic}) and (\ref{unitarity}),
together with the analytical form for the
mass dependence of the tree-level cross-section,
\begin{equation}
\sigma_{SM}(m_{h_i})\sim \lambda^{1/2}(\lambda+12sm_Z^2)
\end{equation}
where $\lambda=(s+(M_Z+m_h)^2)(s-(M_Z-m_h)^2)$, it is then
straightforward to prove that this will always be minimised when all
three masses are degenerate and $R_i^2=1/3$.
This simple result leads to the contour plot of
$\Lambda=\sqrt{M^{\prime 2}_{11}}$ in the integrated luminosity-energy
plane shown in Fig.5. \footnote{Note that this is {\it total}
luminosity, not luminosity per experiment as elsewhere in this paper.}
Along each of the contours the energy and integrated
luminosity correspond to at least one CP-even Higgs boson being
produced with a yield of 50 events, for any allowed choice of $m_{h_1}$
and $m_{h_2}$. Thus this contour plot is an estimate of the values of
$\Lambda$ which may be excluded for different machine parameters. For
example, LEP2 with an energy of 205 GeV will place a limit
$\Lambda>100$ GeV, depending on the integrated luminosity. For this
energy there is a rapid increase in the reach of $\Lambda$ from 10-100
GeV as the luminosity is increased from 100-400 pb${}^{-1}$, followed
by a very slow increase if the luminosity is increased beyond this.
To summarise, the worst case parameters are when all the three CP-even
Higgs bosons have a mass equal to $m_{h_i}=\Lambda$ and equally
suppressed couplings ${R_i}^2=1/3$. In fact since in this case there
are three Higgs bosons to discover the cross-section will be three
times larger than for each Higgs boson taken separately. Nevertheless,
it is clear that for sets of parameters when the Higgs boson masses
are not degenerate and the Higgs bosons must be considered separately
the worst case will never be worse than that just described. The
excluded value of $\Lambda$ is therefore simply determined by the
following rule of thumb: consider a single Higgs boson with a coupling
equal to $R^2=1/3$, and find its maximum excluded mass for a given set
of machine parameters, then equate this mass with the maximun excluded
value of $\Lambda$.
According to this rule of thumb, LEP1 already places a limit on
$\Lambda$ of $\Lambda>59$ GeV, which is just equal to the mass limit
for a CP-even Higgs boson with its ZZh coupling suppressed by
$R^2=1/3$. Note that because of the steep rises of the function plotted
in Fig.4a, the excluded values of $\Lambda$ are not far from the
present excluded value of SM Higgs boson mass of about 65 GeV.
Similarly we find that LEP2 will yield the exclusion limits
$\Lambda>\Lambda_{min}=81,93,105$ GeV, for the three sets of LEP2
machine parameters in Eq.(\ref{LEP2}), respectively.
The above rule of thumb is easily extended to the more general case of
$n$ singlets being added to the MSSM, {\em i.e.} the (M+$n$)SSM. In such
a model the worst case will correspond to $2+n$ CP-even Higgs bosons
each having a mass equal to $\Lambda$, and each having a coupling
suppression relative to the SM of $R^2=1/(2+n)$. The excluded value of
$\Lambda$ is determined from the exclusion plot of a Higgs boson with
a coupling equal to $R^2=1/(2+n)$. For example for $n=1,2,3$ we find
that $R^2=1/3,1/4,1/5$ corresponding to the LEP1 excluded values given
from Fig.4a of about $\Lambda>59,58,57.5$ GeV, respectively. We find it
remarkable that such strong limits on $\Lambda$ can be placed on models
with several singlets.
We emphasise that this argument is only approximate, and may become
unreliable when more realistic Higgs exclusion data as in Fig.4a are
used, rather than the simple analytic calculations of how large $R_i^2$
may become as a function of $m_{h_i}^2$ (as in Fig4b).
Finally note that if Higgs bosons are discovered then this does not
mean, in the context of this model, that $\Lambda$ is small. One may
have a large $\Lambda$ (as large as 146 GeV ) and still be lucky enough
to find that there is a visible CP-even Higgs boson corresponding to
some fortunate values of parameters. All we have shown is how the
non-observation of Higgs bosons enables a firm lower limit to be placed
on $\Lambda$, corresponding to a ``worst case'' situation. In general,
for a given $\Lambda$, the true situation will be easier than this,
leading to the possibility of Higgs discovery at LEP2 even for very
large $\Lambda$.
\subsection{$Z\to hA$}
In section 3 it was seen that the $Zh_iA_j$ couplings factorise into
a factor from the CP-odd matrix multiplied by a factor from the CP-even
matrix \cite{NMSSM} given by
\begin{eqnarray}
R_{Zh_iA_1}&=&S_i\cos\gamma \cr
R_{Zh_iA_2}&=&-S_i\sin\gamma.
\end{eqnarray}
Using equation (\ref{XX}) we find
\begin{eqnarray}
M^{\prime 2}_{22}&=& X_{12}^2m_{h_1}^2
+ X_{22}^2m_{h_2}^2
+ X_{32}^2m_{h_3}^2 \cr
&=&S_1^2m_{h_1}^2+S_2^2m_{h_2}^2+S_3^2m_{h_3}^2
\label{basicA}
\end{eqnarray}
where we have used the fact that $S_i=X_{i2}$.
Again
\begin{equation}
S_1^2+S_2^2+S_3^2=1
\label{unitarityA}
\end{equation}
from the unitarity of $X$.
$M^{\prime 2}_{22}$ just corresponds to the less useful bound
in Eq.(\ref{bound1lessuseful}).
Eq.(\ref{basicA}) is of course virtually identical to Eq.(\ref{basic}),
and hence we may immmediately write down analagous
bounds which relate the masses of the lightest CP-even states
to their $ZhA$ couplings :
\begin{eqnarray}
m_{h_1}^2 &\leq & M_{22}^{\prime 2}\cr
m_{h_2}^2 &\leq &\frac{M_{22}^{\prime 2}-S_1^2m_{h_1}^2}{1-S_1^2}\cr
m_{h_3}^2 &\leq &
\frac{M_{22}^{\prime 2}-(S_1^2+S_2^2)m_{h_1}^2}{1-S_1^2-S_2^2}
\label{newbounds}
\end{eqnarray}
It is trivial to prove that
\begin{equation}
m_{A}^{\prime 2}=m_{A_1}^2\cos^2\gamma+m_{A_2}^2\sin^2\gamma
\label{basicA2}
\end{equation}
which is similar to Eqs.(\ref{basic}) and (\ref{basicA}).
Eq.(\ref{basicA2}) implies the bound
\begin{equation}
{m_{A_1}}^2\leq {m'_A}^2
\end{equation}
so that ${m'_A}^2$ is just the upper bound on the
lightest CP-odd Higgs boson in this model.
When $A_1$ is weakly coupled ($\cos \gamma \approx 0$),
Eq.(\ref{basicA2}) implies that
${m_{A_2}}^2\approx {m'_A}^2$.
Thus if the lighter CP-odd
state is weakly coupled,
corresponding to its having a large singlet component,
then the heavier CP-odd state
(which essentially corresponds to the MSSM CP-odd state) sits at the bound,
in complete analogy with Eq.(\ref{equality3}).
We now discuss how LEP limits on the non-observation of $Z\rightarrow
hA$ may be used to exclude values of ${m'_A}^2$ and $M_{22}^{\prime 2}$
smaller than a certain amount. The following discussion obviously
parallels that of Section 5.3. As before, the ``worst case'' for $Z\to
hA$ production will provide the best exclusion limit on ${m'_A}^2$ and
$M_{22}^{\prime 2}$. In order to determine this worst case we simply
repeat the algorithm in Section 5.3, but with the $R_i$ couplings
replaced by the $S_i$ couplings in Eq.(\ref{newbounds}). However in the
present case we need to generalise this procedure to include the CP-odd
parameters $m_{A_1}$ and $\gamma$. Using the exclusion criterion of
requiring 50 $hA$ events for exclusion, it is readily seen that such a
procedure enables the maximum excluded values of ${m'_A}^2$ and
$M_{22}^{\prime 2}$ to be obtained.
Similarly to the case of Section 5.3, we find the worst case to
correspond to the CP-odd parameters $m_{A_1}=m_{A_2}=m'_A$ and
$\gamma=\pi/4$, and the CP-even parameters $S_1=S_2=S_3=1/\sqrt 3$ and
$m_{h_1}=m_{h_2}=m_{h_3}=M'_{22}$. This corresponds to
${R_{Zh_iA_j}}^2=1/6$ in each case. The excluded values of ${m'_A}^2$
and $M_{22}^{\prime 2}$ may be found by using the approximation of 50
$hA$ events with $h$ having a mass equal to $M'_{22}$, $A$ having a
mass equal to ${m'_A}$, and the coupling ${R_{ZhA}}^2=1/6$, similarly
to the $Z\to Zh$ case discussed above.
As before, the above rule of thumb is easily extended to the more
general case of $n$ singlets being added to the MSSM, {\em i.e.} the
(M+$n$)SSM. In such a model the worst case of $hA$ production will
correspond to $n+2$ CP-even Higgs bosons each having a mass equal to
$\sqrt{M^{\prime 2}_{22}}$, and each having a coupling factor of
$S^2=1/(2+n)$, plus $n+1$ CP-odd Higgs bosons each having a mass equal
to $m'_A$, and each having a coupling factor of $P^2=1/(1+n)$. As
before, when the states are non-degenerate, there must be at least one
pair $h_i$ and $A_j$ which have a cross-section larger than this. For
example for $n=1,2,3$, the couplings are ${R_{ZhA}}^2=1/6,1/12,1/20$.
We expect this simple approximation to be rather more robust for the
$Z\to hA$ process than for $Z\to Zh$, because of the more rapid
fall-off of the cross-section with increasing masses for the $hA$
process than for $Zh$. Furthermore, the larger number of parameters
here make a more elaborate analysis, such as we shall later perform for
$Z\to Zh$, more difficult. Thus we shall always use this approximate
method when we present our results for $Z\to hA$.
We note that of course ${m'_A}^2$ and $M_{22}^{\prime 2}$ are not
independent parameters. Using Eq.(\ref{bound1lessuseful}) the value of
$M_{22}^{\prime 2}$ may be related to ${m'_A}^2$ and, for a given
$\lambda$ and $\tan \beta$, $M_{22}^{\prime 2}$ may be eliminated.
Thus, for a fixed value of $\lambda$, we shall present exclusion plots
in the $m'_A-\tan \beta$ plane from the non-observation of
$Z\rightarrow hA$.
\subsection{Charged Higgs Detection}
As discussed above, the $Z\to hA$ and $Z\to Zh$ searches are less
powerful in extended models because the cross-section can be greatly
reduced. We now turn to the last feasible search at LEP2, that for
charged Higgs production. One feature of the model which was discussed
earlier was that the charged Higgs can be lighter than in the MSSM when
singlets are included; furthermore, its couplings cannot be suppressed
by singlet mixing. Hence the charged Higgs signal, which in the MSSM is
completely dominated by $Z\to hA$, is now far more important.
Charged Higgs discovery is complicated, since the rate is strongly
dependent not only on the Higgs mass, which must not be too close to
the $W$ or $Z$ mass to allow elimination of background, but also on its
branching fractions to $cs$ and $\tau\nu$ which are determined by
$\tan\beta$ \cite{charged}. In addition, any LEP2 discovery region may
be dominated by study of top quark decays at the TeVatron \cite{topch}.
For the purposes of this paper we shall adopt the rather optimistic
view that LEP2 will have sufficient luminosity so that the kinematic
limit may be approached.
\section{Exclusion Limits in the $m'_A$-$\tan\beta$ Plane}
In this section we shall present our results as excluded regions in the
$m'_A-\tan \beta$ plane, which is familiar from similar studies in the
MSSM, and should simplify the comparison of the (M+1)SSM with the MSSM.
Let us first briefly summarise the LEP2 search strategies we have so
far introduced in Section 5.
For the process $Z\rightarrow h_iA_j$ (discussed in section 5.4) we
have shown how LEP2 can be used to place exclusion limits on ${m'_A}^2$
and $M_{22}^{\prime 2}$. For a fixed value of $\lambda$ (and radiative
correction parameters) these limits may be interpreted as excluded
regions in the $m'_A-\tan \beta$ plane. We show both the worst case
mixing scenario and the simple scenario where there is no singlet
mixing. These are calculated using the simple approximation that
50 $Z\to hA$ events will be sufficient to ensure a discovery.
Similarly, for the processes $Z\rightarrow Zh_i$ (discussed in sections
5.1-5.3) we have shown how LEP2 can be used to place an exclusion limit
on the value of $\Lambda$. For a fixed value of $\lambda$ (and
radiative correction parameters) this can be interpreted as an
exclusion limit on $\tan\beta$, independently of $m'_A$ ({\em i.e.} a
horizontal exclusion line in the $m'_A$-$\tan\beta$ plane.) However, as
we shall see, the resulting excluded region is not very large, so
we shall resort to a more powerful technique as discussed below.
This new technique exploits the fact that the upper 2$\times$2 block of
the CP-even mass squared matrix in Eq.(\ref{CPevenmass}) is completely
specified (for fixed $\lambda$) in the $m'_A$-$\tan\beta$ plane.
However, unlike the MSSM, the CP-even spectrum is not completely
specified since it depends on three remaining unknown real parameters
associated with singlet mixing ({\em i.e.} the dots in
Eq.(\ref{CPevenmass}).) Each choice of these unconstrained terms then
completely specifies the parameters $m_{h_i},R_i$, and we can test to
see if the resulting Higgs spectrum is excluded or allowed. We then
scan over all possible choices (which we parametrise as $m_{h_1}$,
$R_1$ and one other mixing angle) and if the resulting spectrum can
always be excluded by LEP2, then we conclude that this point in the
$m'_A$-$\tan\beta$ plane (for fixed $\lambda$) can be excluded.
Since we have presented two separate algorithms for generating
$Z\rightarrow Zh_i$ contours, it is worthwhile here mentioning some of
the advantages and disadvantages of the different techniques. The
method of scanning over the whole of parameter space is naturally
better than the algorithm based on the exclusion limit of $\Lambda$, in
the sense that it gives a better reach in the plane. This is clear from
the fact that we are considering a general 3$\times$3 mass matrix which
has the upper 2$\times$2 block given by our position in the
$m'_A-\tan\beta$ plane, while the bound algorithm in section 5.3 allows
any mass matrix which has the correct 11 component in a particular
basis (which is why the exclusion contours derived in this way do not
have any $m_A$ dependence). On the other hand, the scanning technique
is very CPU-intensive, and would rapidly become even more so if we
allowed extra singlet states. In addition, it is hard to understand
these results analytically, whereas the simple arguments in section
5.3 based around Eq.(\ref{basic}) are much more straightforward. Hence
both techniques are worth considering.
\subsection{$\lambda=0$}
In figures 6-8 we consider the impact of mixing only, with $\lambda$
set equal to zero. This may not be as unreasonably optimistic as it
sounds, since recent GUT scale analyses \cite{ulrich,epic} have
concluded that very small $\lambda$ is preferred with universal soft
parameters. In each case we show the charged Higgs kinematic limit as
a dot-dashed line. The two $Z\to hA$ exclusion contours are represented
by dashed lines and drawn with the assumption of worst case mixing
(left) and of no singlet mixing (right). The three $Z\to Zh$ exclusion
contours are represented by solid lines, and shown for the case of no
singlet mixing (uppermost line), the full scan over parameter space
(middle line), and for the bound algorithm (lower horizontal line)
discussed in detail in section 5.3. The no singlet mixing lines with
$\lambda=0$ simply correspond to the MSSM.
Figure 6 shows the exclusion contours for $\sqrt s=175$GeV, and
integrated luminosity of 150pb${}^{-1}$ per experiment. Figure 7 is
identical but with $\sqrt s=192$GeV, while Figure 8 has $\sqrt
s=205$GeV and integrated luminosity of 300pb${}^{-1}$ per experiment.
All these figures have $m_t=175$GeV, degenerate squarks at 1TeV, and
$\alpha_3(M_Z)=0.12$.
In each of these figures it is clear that the charged Higgs kinematic
limit discovery line (which, as mentioned above, is rather
over-optimistic as a discovery process, but is nevertheless indicative
of where this process will become important) is completely superseded
by the $Z\to hA$ line, although to a rather less extreme degree than in
the MSSM. The inclusion of singlets reduces the excluded region for
both $Z\to hA$ and $Z\to Zh$ quite substantially relative to the MSSM,
but still leaves reasonably large areas covered. Simply using the bound
algorithm however gives rather poorer reach.
\subsection{$\lambda>0$}
To show how large $\lambda$ affects our results, in Figure 9 we show a
figure with $\sqrt s=192$GeV, and integrated luminosity of
150pb${}^{-1}$ per experiment, but with $\lambda=0.5$. Here the charged
line is becoming quite competitive with the $Z\to hA$ line, and so a
charged Higgs search may well be the most practical one at LEP2. The
$Z\to hA$ contours are hardly changed from those in the $\lambda=0$
case except for very small $\tan\beta$.
The most significant impact is however on the $Z\to Zh$ lines. Here
even the no mixing scenario has very little reach in the
$m_A-\tan\beta$ plane because, as is clear from Eq.(\ref{bound1}),
small $\tan\beta$ no longer implies a light CP-even state. With singlet
mixing, no part of the plane can be covered at all.
\section{Conclusion}
The addition of extra singlets to the MSSM greatly complicates the
model and renders far harder the main experimental searches. However
it is possible to exclude this model by placing an experimental lower
bound $\Lambda_{min}$ on the value of the parameter $\Lambda$ which is
just the upper bound on the lightest CP-even Higgs boson mass. In the
limit that the lightest CP-even Higgs boson is very weakly coupled,
this bound applies to the second lightest CP-even Higggs boson. Thus
for a fixed value of $\Lambda$ the entire CP-even Higgs boson spectrum
is constrained, resulting in experimentally excluded values of
$\Lambda$ associated with the ``worst case'' scenarios discussed in
section 5.3.
We have seen that LEP1 already finds $\Lambda>\Lambda_{min}=59$ GeV,
and LEP2 will set limits of $\Lambda_{min}\approx 81,93,105$ GeV for
three different levels of operation. The theoretical upper bound on
$\Lambda$ is $\Lambda_{max} \approx 146$ GeV, depending on the details
of the squark spectrum and on the top mass. Once $\Lambda_{min}$
becomes greater than or equal to $\Lambda_{max}$ then the model will be
excluded. We have generalised this procedure to the case of an
arbitrary number of extra singlets. Thus our first conclusion is that
it will be possible to exclude a version of the MSSM containing
additional singlets.
The effects of the additional singlet can be thought of as reducing the
Higgs sector of the model to the MSSM with two additional complicating
factors : extra singlet states which can mix in an arbitrary way with
the usual neutral Higgs states altering the masses and diluting the
couplings of the mass eigenstates; and an extra Higgs sector quartic
coupling $\lambda$ which changes the mass matrices even in the absence
of singlet mixing.
We have systematically studied the case of singlet mixing with
$\lambda=0$, with the primary conclusion that, while the inclusion of a
singlet can significantly complicate matters for searches at LEP2, it
is still possible to cover a significant amount of the $m_A'-\tan\beta$
plane using the usual $Z\to Zh$ and $Z\to hA$ searches. The main
difference in strategy between the searches in the MSSM and in models
with singlets is that more luminosity is needed to cover the same area
of the plane, since states can be more weakly coupled than in the MSSM.
A second difference is that, because singlets cannot mix with charged
states, the charged Higgs signal cannot be degraded and so is more
important than in the MSSM where it is completely dominated by $Z\to
hA$.
The effect on non-zero $\lambda$ is more troublesome. For large
$\lambda$ the mass bounds in the MSSM are markedly increased,
particularly at small $\tan\beta$ which is of course the region where
LEP2 normally has the best reach. This largely wrecks the prospect of
excluding large parts of the $m_A'-\tan\beta$ plane through the process
$Z\to Zh$, and has some impact on $Z\to hA$. However, we note that
large $\lambda$ also reduces the charged Higgs mass substantially,
making its discovery easier.
We conclude on a positive note by pointing out that for much of this
paper we have assumed worst case mixing scenarios, and considered how
it might be possible to rule out regions of parameter space in a
consistent way for arbitrary parameters from the singlet sector. For
the MSSM, specifying a point in the $m_A-\tan\beta$ plane completely
specifies all the masses and mixings, and so points outside the
discovery contours cannot be discovered; however with singlets it is
possible that, if we are lucky enough, {\em any} point in the
$m_A'-\tan\beta$ plane could lead to a discovery, since it is possible
to construct singlet mixing parameters which will reduce the mass of a
heavy CP-even state while leaving its coupling reasonably large. Hence
we can argue that despite the rather negative impact of the mixing on
the LEP2 exclusion contours, a SUSY Higgs discovery may be no less
likely in a model with singlets than in the MSSM.
\begin{center}
{\bf Acknowledgements}
\end{center}
We would like to thank the LEP2 Higgs working group and particularly
M.~Carena, U.~Ellwanger, H.~Haber, C.~Wagner, and P.~Zerwas for many
stimulating discussions and for encouraging us finally to do some work
on this topic. We are also very grateful to P.~Janot and A.~Sopczak for
sharing their data and expertise.
|
\section{Introduction}
One of the major challenges in the density functional theory
\cite{HK64,KS65,DG90,sah} is to improve the treatment of the exchange
and correlation energy terms in inhomogeneous systems. Among the most
vigorously pursued schemes is the inclusion of generalized gradient
corrections to the local density approximation (LDA) \cite{sah,per}.
One of the persistent problems related to the current formulations of
LDA is the fact that the single--particle energies obtained in these
methods have no direct relationship with the actual single--particle
spectrum of the systems under consideration (with the exception of the
ionization energies though). In particular, the gaps in semiconductors
are severely underestimated \cite{gap}. It has been argued \cite{gc}
that the origin of this discrepancy is the very nature of the local
approximation to the density functional theory. This is one reason why,
following an earlier argument \cite{LM83}, the electronic properties of
several compounds were recently computed in an exact treatment of the
exchange energy \cite{exc}. An improvement to the gap problem in
semiconductors seems to be provided by the so called self--interaction
method \cite{sic}.
Here we have chosen to explore another line of inquiry, inspired by the
so called optimized effective potential (OEP) treatment of the exchange
energy. This method was first introduced in atomic physics
\cite{SH55,TS76} and has been increasingly revisited lately
\cite{last}. This recent activity generated some significant
developments, for instance, a time dependent extension of the OEP
approach \cite{tdlda}. So far, most of the applications of the OEP have
been devoted to the exchange only functionals.
As in the OEP method, we shall consider the total energy of a many
electron system $E_{tot}$ in the Hartree--Fock approximation only
\begin{eqnarray}
\label{totalE}
E_{tot}
&=& \frac{\hbar ^2}{2 m_0} \sum_h \sum_{\sigma } \int \! d\bbox{r} \,
\bbox{\nabla}\Psi_h^\ast(\bbox{r},\sigma )
\cdot \bbox{\nabla}\Psi _h(\bbox{r}, \sigma )
+ e^2 \sum_{i<k }\frac{Z_i \,Z_k}{|\bbox{R}_i-\bbox{R}_k|}
\nonumber \\
&-& e^2\sum_{i=1}^N \int d\bbox{r} \, \frac{Z_i\rho(\bbox{r},\bbox{r})}
{ |\bbox{R}_i-\bbox{r}| }+
\frac{e^2}{2} \int\! d\bbox{r}\!\int\! d\bbox{r}^{\prime} \,
\frac{\rho(\bbox{r},\bbox{r})\rho(\bbox{r}^{\prime},
\bbox{r}^{\prime})}{|\bbox{r}-\bbox{r}^{\prime}|}
\nonumber \\
&-& \frac{e^2}{2} \sum_{\sigma }\int \! d\bbox{r}\!\int\! d\bbox{r}^{\prime}
\,\frac{\rho_{\sigma }(\bbox{r},\bbox{r}^{\prime}) \,
\rho_{\sigma }(\bbox{r}^{\prime},\bbox{r})}
{ |\bbox{r}-\bbox{r}^{\prime}| } \;,
\end{eqnarray}
where $m_0$ is the electron mass, $e$ is its charge, $\bbox{R}_i$ gives
the position of nuclei with charge $Z_i e$, $\Psi_h(\bbox{r}, \sigma)$
the single--particle electron wave functions and $\sigma$ the spin
variables. We shall use throughout this work the index $h$ for labelling
the hole (occupied) states and $p$ for the particle (unoccupied)
states. The indices $h$ and $p$ will stand for the corresponding
quantum numbers of the single--particle states. The single--particle
density is given by
\begin{equation}
\label{density}
\rho_{\sigma }(\bbox{r},\bbox{r}^{\prime})= \sum_h\Psi_h(\bbox{r},\sigma)
\Psi _h^*(\bbox{r}^{\prime},\sigma ), \quad
\rho(\bbox{r},\bbox{r}^{\prime})= \sum_{\sigma}\rho _{\sigma }(\bbox{r},
\bbox{r}^{\prime})\;.
\end{equation}
For the sake of simplicity we shall suppress the spin variables in the
following formulas.
The standard approach to self--consistently minimize $E_{tot}$ is to
solve the Hartree--Fock equations. As customary, those are obtained by
varying $E_{tot}$ with respect to $\Psi_h^\ast(\bbox{r})$ keeping the
single--particle wave functions normalized, leading to
\begin{equation}
\label{eqHF}
H_{\mathrm{HF}} \Psi_h ({\bbox{r}}) \equiv
\left(-{\frac{\hbar^2}{2m_0}}\nabla^2 +
V_{\mathrm{dir}}({\bbox{r}})\right)\!\Psi_h(\bbox{r}) -
e^2\sum_{h'} \int \! d \bbox{r}^\prime \,
\frac{\Psi_{h'}^*(\bbox{r}^\prime) \Psi_{h'}(\bbox{r})}
{| \bbox{r} - \bbox{r}^\prime|}\,
\Psi_h(\bbox{r}^\prime) = \varepsilon_h \Psi_h ({\bbox{r}}) \; .
\end{equation}
The local (direct) part of the potential is given by
\begin{equation}
\label{dirterm}
V_{\mathrm{dir}}(\bbox{r}) = V_{\mathrm{ions}}(\bbox{r}) + e^2\!
\int \! d\bbox{r}^\prime \,
\frac{\rho(\bbox{r}^\prime)}{| \bbox{r} - \bbox{r}^\prime|} \; .
\end{equation}
In the OEP approximation the single--particle wave functions are
the solutions of a local Schr\"odinger equation
\begin{equation}
\label{eqOEP}
-\frac{\hbar^2}{2 m_0}\nabla^2 \Psi_h(\bbox{r})+
V_{\mathrm{OEP}}(\bbox{r})\Psi_h(\bbox{r})
= \varepsilon_h \Psi_h(\bbox{r})
\end{equation}
and the potential $V_{\mathrm{OEP}}(\bbox{r})$ is determined so as
to minimize the total energy of the system
\begin{equation}
\mbox{min} \mbox{\Large \{} E_{tot}(\left\{V_{\mathrm{OEP}}
(\bbox{r})\right\}) \mbox{\Large \}}
\;\; \Longrightarrow \;\;
\frac{\delta E_{tot}}{\delta V_{\mathrm{OEP}}(\bbox{r})} = 0 \;.
\end{equation}
At this point a comment about the well known Slater approximation for
the exchange energy is in order. We shall call {\sl Slater} the
approximation in which the single--particle wave functions obtained in
the LDA with exchange only (LDAX) are used to compute $E_{tot}$
according to Eq.(\ref{totalE}). Thus, it should be fairly obvious that
total energy estimates in these three methods satisfy the following
relation
\begin{equation}
E_{tot}^{\mathrm{HF}} < E_{tot}^{\mathrm{OEP}}<
E_{tot}^{\mathrm{Slater}}\;.
\end{equation}
In the same spirit, notice that one cannot vouch for the value of
$E_{tot}$ computed in LDAX ({\sl i.e.} consistently using the Slater
prescription for the exchange energy to calculate $E_{tot}$) to be
either an upper or lower bound estimate for the total energy.
This paper is organized as follows: In the next section we present a
generalized local approximation (GLA) and explain the naturalness of
introducing a coordinate dependent effective mass. In Section III we
describe two strategies for implementing the GLA and discuss in
detail the case of spherically symmetric systems. In Section IV we
present results for metallic clusters in the jellium approximation.
There we compare different approximation schemes. We comment on some
further possible extensions of GLA in Section V and present our
conclusions in Section VI.
\section{The generalized local approximation}
In this section we shall present an extension of the OEP method, which
we shall refer to as the generalized local approximation (GLA). This
method is partially inspired by early attempts to treat the exchange
term in a systematic way, put forward mostly in nuclear physics
\cite{FL57,PB62,Aus65,Bul88}. The basic idea of the GLA is to replace
the OEP Schr\"odinger equation (\ref{eqOEP}) by a generalized local
Schr\"odinger equation with a coordinate dependent effective mass
$m_{\mathrm{eff}}(\bbox{r})=m_0 \,\mu(\bbox{r})$
\begin{equation}
\label{localH}
-\frac{\hbar^2}{2 m_0}\bbox{\nabla}\frac{1}{\mu(\bbox{r})}
\bbox{\nabla} \Psi_h(\bbox{r}) + V(\bbox{r}) \Psi_h(\bbox{r})=
\varepsilon _h \Psi_h(\bbox{r})\;.
\end{equation}
The GLA local potential $ V(\bbox{r})$ and the effective mass $\mu
(\bbox{r})$ are now determined by following set of equations
\begin{equation}
\label{minGLA}
\mbox{min} \mbox{\Large \{} E_{tot}(\{V(\bbox{r}),\mu (\bbox{r})\})
\mbox{\Large \}} \quad
\Longrightarrow \quad
\frac{\delta E_{tot}}{\delta V(\bbox{r})} = 0 \;\;\; \mbox{and} \;\;\;
\frac{\delta E_{tot}}{\delta \mu (\bbox{r})} = 0 \;.
\end{equation}
To motivate that a coordinate dependent effective mass is a natural
ansatz, we shall invoke two different arguments. The first one, formal
and general in nature, is essentially a summary of a more comprehensive
reasoning presented in Ref.\cite{Bul88}. The second argument is more
physical, but will be presented by putting in perspective the different
approximation schemes for the particular case of a jellium model for
metallic clusters.
Let us start by writing the Hartree--Fock equations (\ref{eqHF}) in the
form
\begin{equation}
\label{newnonloc}
-\frac{\hbar^2}{2m_0}\nabla^2 \Psi_h(\bbox{r}) +
\int \! d \bbox{r}^\prime \, U(\bbox{r},\bbox{r}^\prime) \,
\Psi_h(\bbox{r}^\prime) = \varepsilon_h \Psi_h ({\bbox{r}}) \; ,
\end{equation}
where the kernel $U(\bbox{r},\bbox{r}^\prime)$ contains the exchange
potential which we are interested in, plus a direct term
$V_{\mathrm{dir}}(\bbox{r})\delta (\bbox{r}-\bbox{r}^{\prime})$ as given
by Eq. (\ref{dirterm}). For convenience, we define the new space
coordinates
\begin{equation}
\bbox{x} = \frac{1}{2}(\bbox{r} + \bbox{r}^\prime)
\quad \mbox{and} \quad
\bbox{s} = \bbox{r}^\prime - \bbox{r}
\end{equation}
and change accordingly the kernel to
\begin{equation}
\widetilde U(\bbox{x},\bbox{s}) = U(\bbox{r},\bbox{r}^\prime) \; .
\end{equation}
An approximate way to obtain a local equivalent of (\ref{newnonloc}) is
to expand the wave functions and the diagonal part of the kernel
$\widetilde U$ in a Taylor series in $\bbox{s}$, retaining terms up to
the second order in $\bbox{s}$
\begin{equation}
\Psi_h(\bbox{r}^\prime) \approx \Psi_h(\bbox{r}) +
\bbox{\nabla _r} \Psi_h(\bbox{r}) \cdot \bbox{s} +
\frac{1}{2}\sum_{i,j} \frac{\partial^2 \Psi_h(\bbox{r})}
{\partial r_i \,\partial r_j} s_i s_j
\end{equation}
and the kernel itself
\begin{equation}
\widetilde U(\bbox{r}+\frac{1}{2}\bbox{s},\bbox{s}) \approx
\widetilde U(\bbox{r},\bbox{s}) +
\frac{1}{2} \bbox{\nabla _r}
\widetilde U(\bbox{r},\bbox{s}) \cdot \bbox{s}
+ \frac{1}{8} \sum_{i,j}
\frac{\partial^2 \widetilde U(\bbox{r},\bbox{s})}
{\partial r_i \,\partial r_j} s_i s_j \;,
\end{equation}
where $i$ and $j$ label the axis and $r_i$ is the component of the
vector $\bbox{r}$ along the axis $i$. Direct insertion of the
approximate $\Psi_h(\bbox{r}^\prime)$ and $\widetilde
U(\bbox{r}+\frac{1}{2}\bbox{s},\bbox{s})$ into Eq.~(\ref{newnonloc})
gives an equation of the very same structure as Eq.~(\ref{localH}).
Furthermore, we can explicitly write for the optimized potential
\begin{equation}
\label{FLVloc}
V(\bbox{r}) = \int\! d \bbox{s} \,\widetilde U (\bbox{r}, \bbox{s})
+ \frac{1}{8} \sum_{i,j} \int \! d \bbox{s}\,
\frac{\partial^2 \widetilde U(\bbox{r},\bbox{s})}
{\partial r_i \,\partial r_j} s_i s_j
\end{equation}
and for the effective mass
\begin{equation}
\label{FLmloc}
\frac{1}{\mu(\bbox{r})} = 1 - \frac{m_0}{\hbar^2}
\sum_{i,j} \int \!d\bbox{s}\,\widetilde U(\bbox{r},\bbox{s})\,
s_i s_j \;.
\end{equation}
This set of equations is a good local approximation to the exchange
term provided that the range of non--locality $s_0$ in $\widetilde
U(\bbox{r},\bbox{s})$ is smaller than the local wave length, or in
general, smaller than the typical bulk characteristic length of the
direct part, {\sl i.e.} $s_0 \ll 2\pi/k_F$, where $k_F$ is the Fermi
wave number. This approximation scheme was put forward by Frahn and
Lemmer \cite{FL57}. A comprehensive overview on this and similar
approaches is presented in Ref. \cite{Bul88}.
One might be tempted to use Eq.(\ref{FLVloc}) and (\ref{FLmloc}) to
construct $V(\bbox{r})$ and $\mu(\bbox{r})$. The problem is that often
we cannot assume $k_F s_0$ to be small, particularly at surfaces.
Unfortunately, a well defined systematic approximate scheme to construct
$V(\bbox{r})$ and $\mu(\bbox{r})$ is only known for non--local
potentials in one--dimension \cite{Bul88}. Even for the simple
spherically symmetric three dimensional case a satisfactory solution can
be found only in some particular cases. In spite of this, the discussion
presented above and the insight provided by Refs. \cite{FL57,PB62,Bul88}
motivated us to consider the variational ansatz in Eq.~(\ref{localH}) as
a very natural one.
Another formal reason to introduce an effective electron mass in
addition to an optimized local potential was pointed out by Austern
\cite{Aus65}. The local and non--local wave functions differ in one
rather subtle aspect. A node of a local wave function coincides always
with an inflection point of that wave function, as one can easily see
from the radial Schr\"odinger equation, if $\phi (r) =0 $ then $\phi ''
(r)=0$ as well. This is no longer true for non--local wave functions
and the only simple way to remedy this functional difference between
local and non--local wave functions is the introduction of a position
dependent effective mass \cite{PB62}. The exact Hartree--Fock single
particle wave functions, being solutions of a non--local equation, do
not have the property that nodes correspond to inflection points. This
is one reason why the trivial exact local effective potentials
corresponding to wave functions with nodes have usually pole
singularities. The remnants of these poles can be seen, for example, in
the local effective potentials determined by Talman and Shadwick
\cite{TS76}.
Based on these considerations one should expect that
\begin{equation}
E_{tot}^{\mathrm{HF}} < E_{tot}^{\mathrm{GLA}}
< E_{tot}^{\mathrm{OEP}} < E_{tot}^{\mathrm{Slater}}\; ,
\end{equation}
since the GLA variational ansatz is more flexible and more suited for
the exchange term than OEP.
In the next sections we shall exemplify how the GLA method can be
implemented on alkali atomic clusters \cite{Bra93}. Presently we shall
treat the ionic background in the well known jellium approximation and
consider only spherical closed shell clusters, when only the valence
electrons are explicitly taken into account. One reason for choosing
this system is that the electronic density of an alkali cluster is
almost constant in the interior of the cluster. Inhomogeneities of the
electronic distribution arise only because of the presence of a
surface. The surface induces the natural falloff of the
electronic density outside of the cluster, as well as Friedel--like
oscillations
\cite{Bra93}. The simplicity of the jellium approximation and the
spherical symmetry of the closed shell clusters allows us to more
easily single
out the effects originating from inhomogeneities and the role of
an effective mass.
One of the appealing features of the OEP method for finite Coulomb
systems is the correct asymptotic behaviour of the local potential,
namely that for $r \rightarrow \infty$, $V(\bbox{r})\rightarrow -e^2/r$.
This feature is also characteristic of the GLA approach. Besides
providing a better estimate for the total energy, which in itself might
not be really a significant gain, the GLA method has an additional desirable
feature: It generates a better approximation to the single--particle
spectrum than the LDA and the OEP methods.
Let us consider, for example, the Na$_{92}$ cluster in various
approximations. For more details see Ref. \cite{MGJ95} and Section IV.
In the LDAX the width of the occupied band is $\Delta \varepsilon
_{\mathrm{LDAX}} = 2.55 $ eV, while in the Hartree--Fock approximation
$\Delta \varepsilon _{\mathrm{HF}} = 5.27 $ eV. The Fermi gas estimate
gives $\Delta \varepsilon_{\mathrm{Fg}} = p_F^2 /2m_0$, where $p_F$ is
the Fermi momentum. For a Wigner Seitz radius $r_s = 4$ a.u., which is
approximately the value for bulk sodium, $\Delta
\varepsilon_{\mathrm{Fg}} = 3.13$ eV. The introduction of the
self--interaction correction (SIC) in the exchange energy increases LDA
band width from 2.55 eV to 2.94 eV, which is closer to the Fermi gas
estimate. This is an indication that the observed smaller occupied band
width in clusters than in the bulk is due to electronic spill--out
effects. The rather big discrepancy between the Hartree--Fock value and
the Fermi gas estimate can be naturally attributed to an electron
effective mass arising from the non--local Fock potential \cite{FL57}. A
naive estimate gives $m_{\mathrm{eff}}/m_0= \Delta
\varepsilon_{\mathrm{Fg}}/ \Delta \varepsilon _{\mathrm{HF}} \approx
0.6$. This value is actually very close to what we determine for this
cluster. The reason for this is that in the cluster interior the electronic
density is to a fair approximation equal to the jellium model bulk
constant density. The occurrence of an electronic effective mass,
different form the bare one, can be also interpreted as an energy
dependence of the electron self--energy \cite{FW71}.
The trivial exact local potentials, for nodeless wave functions,
determined from the exact solution of the Hartree--Fock equations for
Na$_{92}$ system show that $V_{1s}^{HF}\approx -9.12$ eV and
$V_{1h}^{HF}\approx -6.40 $ eV \cite{MGJ95}. These values can be
interpreted either as an energy or as an angular momentum dependence of
the trivial exact local effective potential. An angular momentum
dependence of the effective local potential is in principle present and
shall be discussed below. We claim, however, that the potential depth
difference between $V_{1s}^{\mathrm{HF}}$ and $V_{1h}^{\mathrm{HF}}$
can be mainly accounted for by the energy dependence of the local
effective potential. In other words, the effect can be interpreted by an
electron effective mass smaller than the bare electron mass. In this
model, for Na$_{92}$ the last occupied level $1h$ ($\varepsilon
_{1h}=-3.38$ eV) is very close in energy to the $3s$ level ($\varepsilon
_{3s}=-3.50$ eV), while the lowest occupied level $1s$ has the energy
$\varepsilon _{1s}=-8.65$ eV. Had one attributed the energy difference
$\varepsilon _{1h}- \varepsilon_{1s}=$ 5.27 eV to an angular momentum
dependence of the local effective potential, one would have great
difficulty in explaining why the energy difference $\varepsilon _{3s}-
\varepsilon_{1s}=$ 5.15 eV is almost as big and not much closer to the
Fermi gas estimate $\varepsilon_{\mathrm{Fg}}= 3.13$ eV.
Before concluding this section, it is worthwhile to mention that in
nuclear
physics another approximation beyond the Slater prescription for the
exchange energy has been suggested. The method, called density matrix
expansion (DME), proved to be very successful for short range nuclear
forces between fermions \cite{nv}. In spirit, the DME method is a
generalization of the traditional Slater approximation to inhomogeneous
systems. The resulting local self--consistent equations for the
single--particle wave functions are similar in structure to
Eq.(\ref{localH}). The difference between DME and GLA is that in DME
$\mu(\bbox{r})$ and $V(\bbox{r})$ are self--consistently obtained from
the eigenstates. Unfortunately, when applied to Coulomb systems the
ensuing DME equations have inherent instabilities \cite{pgr}.
\section{Numerical Implementation}
The basic novel idea of the present work is entirely contained in
Eq.(\ref{localH}), as discussed in the previous section. Nonetheless, in
order to have a useful generalized local approximation one needs an
efficient minimization algorithm for the total energy $E_{tot}$. This
section is devoted to the discussion of two different strategies
conceived to determine the best local potential and effective mass in
the GLA approximation.
\subsection{Explicit parameterization}
In this first approach we shall represent the potential $V(\bbox{r})$
and effective mass $\mu (\bbox{r})$ as functions determined by a set of
parameters $\{ a_k \}$. Hence, the problem of finding the best local
potential is equivalent to finding the minimum of the functional
$E_{tot}(\{a_k \})$. It is obvious that the quality of the GLA depends
on the choice of the variational ansatz. In this section we shall first
present the method without specifying any parameterization. We then
proceed showing how this method is implemented for a spherically
symmetric system. The numerical results shall be discussed in section
IV.
Recalling Eq.(\ref{totalE}) the partial derivatives of the total energy
with respect to $\{a_k\}$ can be written as
\begin{equation}
\label{delEda}
\frac{\partial E_{tot}}{\partial a_k} =
\sum_h \int\! d\bbox{r} \,\frac{\partial \Psi_h^\ast (\bbox{r})}
{\partial a_k}\, H_{\mathrm{HF}} \Psi_h(\bbox{r}) + {\rm c.c.} \; ,
\end{equation}
where $\partial \Psi_h ({\bbox{r}})/\partial a_k$ is the solution of
the equation
\begin{equation}
\label{HGLAeq}
(\hat{H}_{\mathrm{GLA}} - \varepsilon_h)\frac{\partial \Psi_h ({\bbox{r}})}
{\partial a_k} +
\left ( \frac{ \partial \hat{H}_{\mathrm{GLA}} }{\partial a_k}
-\frac{\partial \varepsilon _h}{\partial a_k} \right ) \!
\Psi_h ({\bbox{r}}) = 0
\end{equation}
where $\hat{H}_{\mathrm{GLA}}$ is the GLA Schr\"odinger operator. We
stress that the conditions
\begin{equation}
V(\bbox{r}) \rightarrow 0 \;\;\; {\mathrm{and}} \;\;\;
\mu(\bbox{r}) \rightarrow 1 \qquad {\mathrm{for}} \;\;\;
r \rightarrow \infty
\end{equation}
are satisfied throughout.
Since the systems we consider in this paper are spherically symmetric
one can write $V(\bbox{r})$ and $\mu (\bbox{r})$ as functions of the
radial coordinate $r$ only. Thus, a suitable variational ansatz reads
\begin{eqnarray}
V(r) = \sum_{k=-N}^N
\widetilde V_k \exp \left [\frac{(r-r_k)^2}{2a^2}\right ]
\qquad \mbox{and} \qquad
\mu(r) = \sum_{k=-N}^N
\widetilde \mu_k \exp \left [ \frac{(r-r_k)^2}{2a^2}\right ] \;,
\end{eqnarray}
where $r_k=a k$. The sum over $k$ encompasses negative values in order
to guarantee the correct behaviour near the origin, namely $V^\prime
(0) = 0$ and $\mu^\prime (0) = 0$ \cite{Ama77}, with $\widetilde V_k =
\widetilde V_{-k}$ and $\widetilde \mu_k = \widetilde \mu_{-k}$.
Working with this ansatz implies that $\{a_k\} \equiv (\{\widetilde
V_k\}, \{\widetilde\mu_k\})$.
The method is implemented as follows: For a given starting $V(r)$ and
$\mu(r)$ in the form $(\{\widetilde V_k\},\{\widetilde\mu_k\})$ we
solve Eq.(\ref{localH}). New amplitudes $\{\widetilde V_k\}$ and
$\{\widetilde \mu_k \}$ are obtained by using the gradients given by
Eq. (\ref{delEda}) in a suitable minimization algorithm. In particular,
we used the molecular dynamics method described in Ref. \cite{BK} and a
simplified simulated annealing procedure in order to reach the condition
$\partial E_{tot}/\partial a_k \approx 0$ and thus minimize $E_{tot}$
as given by Eq.(\ref{totalE}).
\subsection{Unconstrained minimization}
In this second approach we have applied the steepest descent method to
find directly in coordinate representation the optimized local
potential and the effective mass. At a first glance this approach is
more attractive than the previous one, since it does not rely on a good
variational ansatz. On the other hand, for practical use, an explicit
parameterization allows one to reduce the space of minimization
variables and obtain very efficiently reasonable solutions. However, if
the system in question does exhibit a specific symmetry this method is
particularly easy to implement. In the remaining of this section we
discuss the particular case of spherical symmetric systems. We believe
that this approach can also be implemented for other situations.
In order to solve Eq.(\ref{localH}) numerically, it is convenient to
represent the single--particle wave functions as
\begin{equation}
\Psi_i (\bbox{r}) \equiv \psi_i(r) \,Y_{l_i m_i}(\bbox{\hat{r}})=
\frac{\sqrt{\mu (r)}}{r}\phi_i(r) \,Y_{l_i m_i}(\bbox{\hat{r}})
\end{equation}
where the index $i$ labels states throughout the spectrum of the
$H_{\mathrm{GLA}}$. After some straightforward manipulations,
Eq.(\ref{localH}) can be written as
\begin{equation}
\label{Hphi}
-\phi^{\prime\prime}_i (r)+\left ( U(r)+\frac{l_i(l_i+1)}{r^2}-
\mu (r) \,\epsilon_i\right ) \!\phi_i(r) = 0
\end{equation}
where the energy was rescaled as $\epsilon_i =
{2m_0/\hbar^2}\varepsilon_i$ and the potential $U(r)$ is given by
\begin{eqnarray}
U(r) = \frac{2m_0}{\hbar^2} \mu (r)\, V(r) +
\left \{ \frac{3}{4}\left [\frac{\mu '(r)}{\mu (r)}\right ]^2
-\frac{1}{2}\frac{\mu ''(r)}{\mu (r)}-\frac{\mu '(r)}{r\mu (r)}
\right \} \; .
\end{eqnarray}
The normalization and completeness relations for the single--particle
wave functions $\phi_i (r)$ read
\begin{eqnarray}
\int \! dr \, \phi _k ^*(r)\mu (r) \phi _l(r) &=&
\langle \phi _k | \mu | \phi _l \rangle
= \delta _{kl}\; ,
\nonumber \\
\sum _k \phi _k ^*(r') \mu (r) \phi_k(r) & = & \delta (r-r')\; .
\end{eqnarray}
where $\sum_k$ includes an integration over the continuous (unbound)
spectrum of $H_{\mathrm{GLA}}$. The functional variation of the total
energy (\ref{totalE}) can be brought to the following form
\begin{equation}
\label{delEm2}
\delta E_{tot} =
\int\! dr \left \{
\delta U(r) \sum _h (2l_h+1)\phi_h(r)\chi_h(r)-
\delta \mu (r)\sum _h (2l_h+1)\epsilon_h\phi_h(r)\chi_h(r) \right\}
+ c.c. \; .
\end{equation}
The auxiliary functions $\chi_h(r)$ are solutions of the following set
of equations
\begin{equation}
\label{chiequation}
-\chi^{\prime\prime}_h (r) +\left [ U(r)+\frac{l_h(l_h+1)}{r^2}-
\mu (r) \epsilon _h\right ] \chi _h(r) =
r\sqrt{\mu (r)}[1-\rho ]_{l_h}\,h_{\mathrm{HF}}\psi_h (r)\;,
\end{equation}
where $\rho$ is the single--particle density matrix as in
Eq.(\ref{density}), and $h_{\mathrm{HF}}$ refers to the radial part of
the Hartree--Fock hamiltonian, $H_{\mathrm{HF}}$. On the right hand side
of Eq.(\ref{chiequation}), $[1-\rho ]$ is the projection operator
outside of the occupied single--particle space and $[1-\rho ]_{l_h}$
its radial component corresponding to the angular momentum $l_h$. In
Eq.(\ref{delEm2}) the summation is over the occupied (hole) states $h$
only. The auxiliary single--particle functions $\chi_h(r)$ are
orthogonal to the occupied single--particle states
\begin{equation}
\langle \phi _k | \mu | \chi_l \rangle =
\int\! dr \, \phi_k^\ast(r) \mu(r) \chi_l(r) =
\delta_{kl}\;,
\end{equation}
for all occupied states $k$ and $l$. It is well known that an equivalent
formulation of the Hartree--Fock equations is
\begin{equation}
\label{HFcondition}
[1-\rho ]H_{\mathrm{HF}}\rho = 0 \,,
\;\;\; {\mathrm{or}} \;\;\;
[H_{\mathrm{HF}}]_{ph}=[H_{\mathrm{HF}}]_{hp}=0
\;\;\; {\mathrm{or}} \;\;\;
(1-\rho)H_{\mathrm{HF}}\Psi _h(\bbox{r})= 0\;,
\end{equation}
{\sl i.e.} that all the particle--hole matrix elements of the
Hartree--Fock single--particle Hamiltonian vanish. Consequently, if the
single--particle wave functions $\phi_h(\bbox{r})$ are the solutions of
the Hartree--Fock equations (\ref{eqHF}), then $\chi_h
(r)\equiv 0$. One should not construe this as a statement that the
Hartree--Fock minimum can be reached exactly in this way though, barring
a pure coincidence. According to Eqs. (\ref{eqHF}) and (\ref{localH})
the single--particle wave functions have to be solutions of both GLA and
HF equations at the same time, which in general is very unlikely. The
best one can hope for is that (\ref{HFcondition}) will be rather well
satisfied for some $U(r)$ and $\mu(r)$. Since the total energy is
bounded from below the existence of an optimized local potential and an
effective mass is certain. However, the uniqueness of a minimum, {\sl
i.e.}, the existence of only one global minimum, is not guaranteed. In
principle one has the same problem in the exact HF case also. The
vanishing of $\delta E_{tot}$ upon variation of the local potential
$U(r)$ and effective mass $\mu(r)$ leads to the following equations
\begin{eqnarray}
\sum _h (2l_h+1)\phi_h(r) \chi _h(r) &= &0 \; ,
\\
\sum _h (2l_h+1)\epsilon_h \phi _h(r)\chi _h(r) &=& 0\; .
\end{eqnarray}
In the case when $\mu (r)\equiv 1$, Eq. (30) can be readily rewritten
in the form of the equation for OEP derived by Talman and Shadwick
\cite{TS76}. To implement the steepest descent method one has to change
the local optimized potential and the effective mass according to the
simple rules
\begin{eqnarray}
U(r) \rightarrow U(r)+\delta U(r)
\quad \mbox{and} \quad
\mu (r) \rightarrow \mu (r)+\delta \mu(r)
\end{eqnarray}
where
\begin{eqnarray}
\delta U(r) &=& -2 \lambda \sum _h (2l_h+1)\phi _h(r)\chi_h(r)
\nonumber \\
\delta \mu (r) &=&2 \lambda \sum _h (2l_h+1)\epsilon _h \phi _h(r)
\chi_h(r) \;.
\end{eqnarray}
The step size $\lambda$ has to be gauged with a certain care, so that
Eq. (\ref{localH}) leads to new single--particle wave functions
corresponding to a lower total energy $E_{tot}$. It is worth mentioning
that the corrections to $U(r)$ and $\mu (r)$ introduced above satisfy
the following constraints
\begin{eqnarray}
\int \! dr \,\mu(r) \,\delta U(r)= 0 \qquad \mbox{and} \qquad
\int \! dr \,\mu(r) \,\delta \mu(r) = 0\;,
\end{eqnarray}
which in particular imply that one cannot change the {\sl real}
effective local potential $V(r)$ by a constant only, {\sl i.e.} $V(r)
\rightarrow V(r)+\mbox{const}$ is not a possible change within this
scheme.
\section{Results}
In our numerical study we used the jellium model to illustrate and
compare the previously discussed approximation schemes. The
model is defined by replacing $V_{\mathrm{ions}}(\bbox{r})$ as appears in
Eq.(\ref{dirterm}), by the potential given by an uniform positive
background charge density. We analyzed the following alkali clusters:
Na$_{40}$, Na$_{92}$, Na$_{138}$ and Na$_{196}$. For those,
corresponding to electronic magic numbers, the spherical jellium
approximation
provides remarkably good results for the ground state properties and
optical response \cite{Bra93}. The spherical geometry and the bulk
Wigner--Seitz radius for sodium, $r_s=4$ a.u., determine the model
entirely. In this section we examine the results of different schemes to
solve the problem posed by Eq.(\ref{totalE}) in the jellium
approximation, namely Hartree--Fock, OEP, GLA and Slater.
Since the discussed schemes are approximations to the exchange
potential, we first need to solve the jellium Hartree--Fock problem to
have a standard to compare with. For this purpose we wrote a code that
uses the same kind of algorithm as Ref.\cite{MGJ95}. Apparently the
method used by Hansen and Nishioka \cite{HN93} is more accurate, but
since we had no problems with obtaining converged results for cluster
sizes up to Na$_{196}$, we did not improve further our code. As for the
Slater functional, we used a method very similar to the one described in
Ref. \cite{Ber90}, but with a series of refinements to ensure
a higher numerical accuracy and increase the speed of computation.
To obtain the
solution of Eq.(\ref{localH}) and (\ref{minGLA}) we typically used the
explicit parameterization method with $N=20, \cdots, 30$, depending on
cluster size and $a=1$ \AA . By starting with a reasonable guess for
$\mu(r)$ and $V(r)$ we were always able to find a value for $E_{tot}$
very close to the Hartree--Fock value (see Tables I-III). Using the
unconstrained minimization we could only marginally improve the minima
obtained by explicit parameterization. In order to get a feeling of how
well the GLA method works for different electronic densities we have
varied the jellium density by a factor of two in both directions, {\sl
i.e.} smaller and higher densities (3 a.u. $<r_s<$ 5 a.u.). Again the
results were always in very good agreement with the Hartree--Fock ones.
We should mention, however, that for alkali clusters in the jellium
model the GLA has likely several very close lying energy minima and
often the corresponding $U(r)$ and $\mu (r)$ differ quite considerably
from each other. Furthermore, even for the states we have identified, we
can only claim that our numerical solutions are very close to the
actual minimum. The direct minimization procedures we have used, meet
with considerable numerical accuracy problems close to a minimum and
one can hardly improve on the quality of an approximate solution. The
total energy of alkali clusters, as of any many--electron system as
well, comes as a result of a strong cancellations of different
contributions. Thus, a numerically accurate solution (relative accuracy
better than $\approx 10^{-5}$) is very difficult to obtain.
In Tables I--III we show the kinetic, Hartree, exchange, electron--ion
and total energies for various sodium clusters in Slater, OEP, GLA and
Hartree--Fock approximations. As one can see from these tables the GLA
results are very close to the Hartree-Fock minimum and are of better
quality than the OEP and Slater results. In Fig. 1 we compare the
electron density profiles obtained in OEP, GLA and Hartree--Fock
methods. Again, even though our solution might not exactly correspond to
a local minimum, it is much closer to Hartree--Fock than the OEP
solution. The local effective potential $V(r)$ for the GLA and OEP cases
are displayed in Fig. 2 and in Fig. 3 we show the renormalized
effective potential $U(r)$, see Eq. (23), and the effective mass $\mu
(r)$. We have not imposed the ``correct'' asymptotic behavior $-e^2/r$
on the local potential as was done, for example, by Talman and Shadwick
\cite{TS76}. Since at radii a few \AA $\;$ larger than the jellium
radius the electronic density is very small, any change of the
potential or of the effective mass in this region has little effect on
the hole single--particle wave functions. This is the origin of possible
unexpected features in the local effective potential, such as
non--monotonic behavior beyond the jellium radius. If unphysical
characteristics appear in the local potential for large $r$ at the
early minimization stages, they are very difficult to correct. Such
unwanted features can only significantly alter the particle (unoccupied)
states. Whenever we encountered such problems, after obtaining an
approximate local minimum in the GLA, we proceeded as follows: Using the
parameterized minimization method we constrained the amplitudes
$\{\widetilde \mu_k\}$ and $\{\widetilde V_k\}$ for gaussians inside
the jellium radius to be fixed, varying only the others. Although these
corrections have a negligible effect on $E_{tot}$, the effective
potential tends slowly to acquire the correct form for large $r$.
The single--particle spectrum we have obtained for the occupied states
is in very good agreement with the HF one. The corresponding
single--particle spectra for the occupied states in either LDAX, Slater
or OEP methods is much more compressed, as is shown in Fig. 4. These
differences arise because the effective mass in GLA is smaller than
$m_0$ inside the cluster ($m_{\mathrm{eff}}\approx 0.6 m_0$), as we
have alluded to in Section II.
The unoccupied single--particle states show very interesting features.
In the Hartree--Fock approximation the level density of the unoccupied
states is too small and the gap at the Fermi level is too large. The
reason is well known: the particle states in this approximation are
computed in the field of a complete screened positive charge. On the
other hand, in the OEP method the level density of the unoccupied
states is large and the gap at the Fermi level is small. The main reason
for it is that the $V_{\mathrm{OEP}}(\bbox{r})$ exhibits the correct
asymptotic behaviour as $r\rightarrow \infty$. This causes the major
difference between the OEP and the LDAX single--particle spectra. In
LDAX one observes an overall upward shift of the spectrum, mainly due
to the too sharp fall--off of the potential. One can reconcile almost
perfectly the HF and the GLA particle spectra by rescaling the
Hartree--Fock potential by a factor $(N_e-1)/N_e$, where $N_e$ is the
total number of electrons. In this way we force by hand the HF potential
to have the ``correct'' $-e^2/r$ asymptotic behaviour for particle
states. Notice that the hole states automatically have the correct
asymptotic behaviour built in. The pleasant feature of the GLA
approach is that one obtains not only a correct band width for the
occupied states, but also a correct asymptotic
behaviour for the effective potential of the particle states.
As a result the gap at the Fermi level is larger
in the GLA method than in the OEP method and smaller than in the HF
approximation.
\section{Further generalizations}
The GLA we have presented here is not yet the most general one. As we
have mentioned in the introduction, one can extend it by considering an
arbitrary angular momentum dependence of either the effective local
potential or of the effective mass. (In general, an arbitrary angular
dependence of the effective local potential leads strictly speaking to
a non--local potential.) For example, one can consider a GLA
Schr\"odinger equation of the following type
\begin{equation}
-\frac{\hbar ^2}{2 m_0}\bbox{\nabla}
\stackrel{\leftrightarrow}{\bbox{M}}
(\bbox{r})\bbox{\nabla}
\Psi _h(\bbox{r}) + V(\bbox{r}) \Psi
_h(\bbox{r})=\varepsilon _h \Psi _h(\bbox{r})\;.
\end{equation}
where $\stackrel{\leftrightarrow}{\bbox{M}}(\bbox{r})$ is a symmetric
tensor of rank two. One can chose this tensor in an appropriate manner
such that the Schr\"odinger equation still preserves the spherical
symmetry. For example, in the case of spherical systems one can choose
\begin{equation}
\left [ \stackrel{\leftrightarrow}{\bbox{M}}(\bbox{r})\right ]
_{ij} = \frac{1}{\mu (r)}\delta _{ij} +
\frac{b(r)}{\mu (r)}(\delta _{ij}-\bbox{\hat{r}}_i\bbox{\hat{r}}_j)\;,
\end{equation}
where $i$ and $j$ are axis labels; and rewrite Eq. (\ref{Hphi}) as
\begin{equation}
-\phi^{\prime\prime}_i (r)+ \left [ U(r)+
\frac{l_i(l_i+1)}{r^2}\left(1+b(r)\right)- \mu(r)\epsilon_i\right ]
\phi_i(r) = 0\;.
\end{equation}
In this case one can interpret the appearance of a tensor inverse
effective mass as an angular momentum dependence of the effective
local potential $U(r)$. This ambiguity occurs only in the case of
spherical symmetry. A simple analysis of the HF single--particle
spectra of the clusters discussed in the previous section suggests that
such a correction exists. The fact that the splitting between
consecutive $s$--levels is smaller than between $p$--levels, which in
its turn is smaller than the splitting between $d$--levels and so forth,
gives room for such speculation. On the other hand, the total energy
and the electronic density distribution of the alkali clusters we have
considered here is very little affected by this type of correction. We
have investigated to some extent this possibility as well, but not very
thoroughly.
One can consider some further generalizations, {\sl e.g.} to introduce
a pseudovector component (or equivalently an antisymmetric component of
$\stackrel{\leftrightarrow}{\bbox{M}}(\bbox{r})$) of the inverse
effective mass and/or also the introduction of a vector effective
potential as well and still have a local Schr\"odinger equation which
is a
partial differential equation of at most second order. For example, a
natural candidate for a pseudovector is the spin density. One can expect
that in spin unsaturated systems a term with this symmetry might arise
not only in the local effective potential but also in the inverse
effective mass as well. One has to keep in mind that this type of
correction violates time--reversal symmetry (which is violated in these
systems anyway) and consider it with care. Whether there will be a
reason to introduce a vector effective potential as well is still
unclear at the present moment. The presence of an effective vector
potential might signal also the existence of nonvanishing {\sl
currents} in the ground state, since in such a case the GLA Hamiltonian
would not automatically be invariant under the transformation $\bbox{p}
\rightarrow -\bbox{p}$, which will occur for example if the system is in
an external magnetic field.
\section{Conclusions}
Our analysis suggests a possible way to generalize the popular LDA
treatment of Coulomb systems by enlarging the class of considered
functionals. There is no {\sl a priori} or fundamental reason why one
should consider a total energy functional of the Kohn--Sham type only
and not allow for terms, in which the inverse mass is replaced, {\sl
e.g.}, by a density dependent function. One can find typical examples
of such functionals in nuclear physics \cite{nv}. By considering such
type of generalized local energy density functionals one can improve
the quality of the description of inhomogeneous systems (total energy,
electronic density distribution) and at the same time achieve a
significant improvement of the single--particle spectrum as well, which
is a long standing unsolved issue in LDA. We would like to remind the
reader of one potential problem with using the present approach in
infinite homogeneous systems. It is well known that the electron
self--energy in the HF approximation for a pure Coulomb interaction
between electrons has a logarithmic singularity at the Fermi level
\cite{FW71}. In principle that should prevent us from introducing an
effective mass the way we have described here. On the other hand the
Coulomb interaction between electrons is strongly renormalized in
medium, see {\sl e.g.} Ref. \cite{KSK94}, and this fact in particular
leads to Coulomb screening. (In its simplest classical form that is the
Thomas--Fermi screening at large distances.) This should be sufficient
to remove such singularities of the electron self--energy and thus lends
support to a GLA approach.
Financial support from the National Science Foundation and the
Department of Energy is greatly appreciated.
|
\section{INTRODUCTION}
Quantum Brownian motion (QBM) models provide a paradigm of
open quantum systems that has been very useful in quantum
measurement theory [1], quantum optics [2] and decoherence [3-5]. One of the
advantages of the QBM models is that they are reasonably simple, yet
sufficiently complex to manifest many important features of
realistic physical processes.
Central to the study of QBM is the master equation for the reduced
density operator of the Brownian particle, derived by tracing out
the environment in the evolution equation for the combined system
plus environment. A variety of such derivation have been given
[6-9]. The most general is that of Hu, Paz and Zhang [10,11], who used
path integral techniques and in particular, the Feynman-Vernon
influence functional.
The purpose of this paper is to provide an alternative and
elementary derivation of the Hu-Paz-Zhang master equation for QBM,
by tracing the evolution equation for the Wigner function of the
whole system.
\section{MASTER EQUATION FOR QUANTUM BROWNIAN MOTION}
The system we considered is a harmonic oscillator
with mass $M$ and bare frequency $\Omega$, in interaction with a
thermal bath consisting of a set of harmonic
oscillators with mass $m_n$ and natural frequency $\omega_n$. The
Hamiltonian of the system plus environment is given by
\begin{equation} H={ p^2\over{2M}}+ {1\over2} M\Omega^2 q^2+
\sum_n\left({p^2_n\over
{2m_n}}+ {1\over2}m_n\omega^2_n q^2_n\right)+q\sum_n C_nq_n \> ,\end{equation}
where $q, p$ and $ q_n,p_n$ are the coordinates and momenta of the Brownian
particle and oscillators, respectively, and $C_n$ are coupling constants.
The state of the combined system (1) is most completely described by a density
matrix $\rho (q,q_i;q',q'_i,t)$ where $q_i$ denotes $(q_1,... q_N)$, and
$\rho$ evolves according to
\begin{equation}
\dot\rho = -{i\over\hbar}[H\,, \rho] \>.\end{equation}
The state of the Brownian particle is described the reduced density
matrix, defined by tracing over the environment,
\begin{equation}
\rho_r(q,q',t)=\int \prod_n (dq_ndq'_n\,\delta(q_n-q_n'))\rho
(q,q_i;q',q'_i,t) \>.
\end{equation}
The equation of time evolution for the reduced density matrix is
called the master equation. For a general environment,
Hu, Paz, and Zhang [10] derived the following master equation by using path
integral techniques:
\begin{eqnarray}
i\hbar{ \partial\rho_r\over \partial t} &=&
- {\hbar^2\over 2M}\left({\partial^2\rho_r \over \partial
q^2} - {\partial^2\rho_r \over \partial q'^2 }\right) +
{1\over 2}M\Omega^2(q^2-q'^2)
\rho_r\nonumber \\
& & + {1\over 2}M\delta\Omega^2(t)(q^2-q'^2)\rho_r\nonumber \\
& & - i\hbar\Gamma(t)
(q-q')\left( {\partial\rho_r\over \partial q} -{\partial\rho_r \over \partial
q'}\right) \nonumber \\
& & - iM\Gamma(t)h(t)(q-q')^2\rho_r\nonumber \\
& & +\hbar \Gamma(t)f(t)(q-q')
\left({\partial\rho_r \over \partial q}+{\partial\rho_r\over\partial q'}\right)
\>.
\end{eqnarray}
The explicit form of the coefficients of the above equation will be
given later on.
The coefficient $\delta\Omega^2(t)$ is the frequency shift term, the
coefficients $\Gamma(t)$ is the ``quantum dissipative'' term, and the
coefficients
$\Gamma(t)h(t), \Gamma(t)f(t)$ are
``quantum diffusion'' terms. Generally, these coefficients are time dependent
and of quite complicated behaviour.
We find it convenient to use the Wigner function of the
reduced density matrix,
\begin{equation}\tilde W(q,p,t)={1\over {2\pi}}\int du\>e^{{iu p
/\hbar}}\rho_r\left(q-{u\over 2},q+{u\over 2},t\right)\>.
\end{equation}
Taking the Wigner transform
of (4), we obtain{\footnote {We believe Eq. (2.48) in Ref. [10]
contains some incorrect numerical factors.}}
\begin{eqnarray}
{\partial \tilde W\over \partial t}= &-&{1\over M} p{\partial\tilde W \over
\partial q } +M[\Omega^2 + \delta
\Omega^2(t)]q{\partial\tilde W \over \partial p}\nonumber \\
&+& 2\Gamma(t){\partial (p\tilde W)\over \partial p}
+ \hbar M\Gamma(t)h(t){\partial^2\tilde W \over \partial
p^2} \nonumber \\
&+& \hbar \Gamma(t)f(t) {\partial^2\tilde W\over {\partial q \partial
p}}\>.
\end{eqnarray}
The inverse transformation of (5) is given by
\begin{equation}
\rho_r(q,q',t)=\int dp\> e^{-{ip(q-q')/\hbar}}\tilde W\left({{q+q'}\over
2},p,t\right).
\end{equation}
Our strategy for deriving the master equation (4) is to derive the
Fokker-Planck type equation (6) from the Wigner equation for the
total system. The master equation can be obtained from the Wigner
equation for the system by using the transformation (7).
We shall make the following two assumptions:
(1) The system and the environment are initially uncorrelated, {\it {ie}}. the
initial Wigner function factors
\begin{equation}W_0 (q,p;q_i,p_i)=W^s_0(q,p) W^b_0(q_i,p_i)\>,\end{equation}
where $W^s_0$ and $W^b_0$ are the Wigner functions of the system and the
bath, respectively, at $t=0$.
(2) The heat bath is initially in a thermal equilibrium state at
temperature $T=(k_B\beta)^{-1}$.
This means that the initial Wigner function of bath is of Gaussian form,
\begin{eqnarray}
W^b_0&=&\prod_n W^b_{n0}\nonumber \\
&=&\prod_n N_n{\rm exp}\{-{2\over \omega_n\hbar}\tanh({1\over
2}\hbar \omega_n \beta)H_n\}\>,
\end{eqnarray}
where $H_n$ is the Hamiltonian of the $n$-th oscillator in the bath,
\begin{equation}
H_n={p^2_n\over 2m_n}+{1\over 2}m_n\omega_n^2q_n^2 \>.
\end{equation}
In addition, one easily see that the initial moments of the bath are
\begin{eqnarray}
& & \langle q_n(0)\rangle = \langle p_n(0) \rangle =0 \>,\\
& &\langle q_n(0)q_m(0)\rangle =0\,\> ({\rm if}\> m \neq n)\>,\\
& &\langle p_n(0)p_m(0)\rangle =0\,\> ({\rm if}\> m \neq n)\>,\\
& &\langle q_n(0)p_m(0)+ p_m(0)q_n(0)\rangle =0 \>,
\end{eqnarray}
and
\begin{eqnarray}
\langle q^2_n(0)\rangle &
= & {\hbar\over 2m_n\omega_n}
\coth({1\over 2}\hbar\omega_n\beta)\> ,\nonumber \\
\langle p^2_n(0)\rangle & =
& {1\over 2}\hbar m_n\omega_n
\coth({1\over 2}\hbar\omega_n\beta) \>.
\end{eqnarray}
For the QBM problem described by (1) and (2), the Wigner function of the
combined system plus environment satisfies
\begin{eqnarray}
{\partial W \over \partial t}= & - & {p\over M}{\partial W\over
\partial q} + M\Omega^2 q{\partial W \over \partial p} \nonumber \\
& + & \sum_n \left (-{p_n \over
m_n}{\partial W \over \partial q_n}+m_n\omega_n^2 q_n{\partial W \over
\partial p_n}\right ) \nonumber \\
& + & \sum_nC_n\left (q_n{\partial
W\over \partial p}+q{\partial W \over \partial p_n}\right ) .\end{eqnarray}
By integrating over the bath variables on the both sides of the above
equation , one obtains
\begin{equation}{{\partial {\tilde W}}\over \partial t}=-{p\over M}{{\partial
{\tilde W}} \over \partial q}+M\Omega^2 q{{\partial {\tilde W}} \over
\partial p} + \sum_n C_n\int \prod_i dq_idp_i q_n{\partial
W\over\partial p} \>,\end{equation}
where $ {\tilde W}(q,p)$ is the reduced Wigner function and it follows
from Eq. (3) that
\begin{equation}{\tilde W }(q,p)=\int^{+\infty}_{-\infty} \prod_i dq_i
dp_i W(q,p;q_i,p_i)\> .
\end{equation}
This definition is equivalent to Eqs. (3) and (5). The first two
terms on the right-hand side of the Eq. (17) give rise to
the standard evolution equation of the system. The last term contains
all the
information about the behaviour of the system in the presence of
interaction with environment.
In what follows, we shall demonstrate that the quantity
\begin{equation}G(q,p)=\sum_n C_n \int \prod_i dq_i dp_i q_n W\end{equation}
appearing (differentiated with respect to $p$) in (17) can
be expressed in terms of $\tilde W $and its derivatives.
To this end, we first perform Fourier
transform of $G(q,p)$
\begin{eqnarray}
G(k,k')&=&\int dqdp\,e^{ikq+ik'p}G(q,p)\nonumber \\
&=&\sum_nC_n\int
dqdp\prod_idq_idp_i\>q_n\,e^{ikq+ik'p}\,W(q,p;q_i,p_i)\>.
\end{eqnarray}
It is well known that $q(t),p(t)$ and $ q_n(t) , p_n(t)$ are related
to the classical evolution of their initial values $ q(0),p(0)$
and $q_n(0), p_n(0)$ through a
canonical transformation:
\begin{equation}{\bf z}(t)= U(t){\bf z}(0)\>,\end{equation}
where
$${\bf z}(t) =(q(t),q_1(t)...q_N(t);p(t),p_1(t),...p_N(t))\>. $$
Since the Hamiltonian (1) is quadratic, the Eq. (16) has the same
form as the classical Liouville equation, so the solution of
Eq. (16) is of the form,
\begin{equation}W_t({\bf z})=W_0(U^{-1}(t)\bf z) \>.\end{equation}
Changing the integration variables into their initial values by
this canonical transformation, we obtain
\begin{eqnarray}
G(k,k') &=& \int dq(0)dp(0)\prod_i dq_i(0)dp_i(0)\nonumber \\
& & \times\left[fq(0)+gp(0)
+\sum_n(f_nq_n(0)
+g_np_n(0))\right]\nonumber \\
& & \times {\rm exp}\left[ik\left(\alpha q(0)+\beta
p(0))+\sum_n(a_nq_n(0)+b_np_n(0))\right)\right]\nonumber \\
& & \times {\rm exp}\left[ik'M\left(\dot \alpha
q(0)+\dot\beta p(0)+
\sum_n(\dot a_n q_n(0)+\dot b_n
p_n(0))\right)\right]\nonumber \\
& & \times W^s_0(q(0),p(0))W^b_0(q_i(0),p_i(0)\>.
\end{eqnarray}
Here the coefficients $f,g, f_n, g_n, \alpha, \beta, a_n, b_n$ are time
dependent. Their explicit values are not required.
Similarly, the Fourier transform of the reduced Wigner function is
\begin{eqnarray}
\tilde W(k,k')&=&\int dqdp\,e^{ikq+ik'p}\tilde W(q,p)\nonumber \\
&=& \int dq(0)dp(0)\prod_i dq_i(0)dp_i(0)\nonumber \\
& & \times {\rm exp}\left[ik\left(\alpha q(0)+\beta
p(0))+\sum_n(a_nq_n(0)+b_np_n(0))\right)\right]\nonumber \\
& & \times {\rm exp}\left[ik'M\left(\dot \alpha
q(0)+\dot\beta p(0)+
\sum_n(\dot a_n q_n(0)+\dot b_n
p_n(0))\right)\right]\nonumber \\
& & \times W^s_0(q(0),p(0))W^b_0(q_i(0),p_i(0))\>.
\end{eqnarray}
Now compare $G(k,k')$ and $\tilde W(k,k')$. They differ by the terms
linear in $q(0), p(0), q_n(0), p_n(0)$ in the preexponential factor
in $G(k,k')$. Consider the factors $f_nq_n(0)$ and $g_np_n(0)$ in
$G(k,k')$. Since they multiply $W^b_0(q_i(0),p_i(0))$, and since
$W^b_0(q_i(0),p_i(0))$ is Gaussian in $q_n(0), p_n(0)$, the terms
$f_nq_n(0)W_0^b$ and $g_np_n(0)W^b_0$ may be replaced by terms of
the form ${\partial W^b_0 / \partial q_n(0)}, {\partial
W^b_0/\partial p_n(0)}$ up to time dependent factors. An integration
by parts then may be performed, and these factors are then
effectively replaced by multiplicative factors of $k, k'$.
Similarly, the factors $fq(0), gp(0)$ in the prefactor in $G(k,k')$ may
be replaced by $\partial /\partial k, \partial /\partial k'$ (plus
some more factors of $k$ and $k'$). Hence, it is readily seen that
$G(k,k')$ is a linear combination of terms of the form $k, k',
\partial /\partial k, \partial /\partial k'$ operating on $\tilde
W(k,k')$, with time dependent coefficients.
Inverting the Fourier transform, it follows that
\begin{equation}G=A(t)q\tilde W + B(t)p\tilde W + C(t){\partial\tilde
W\over\partial q} + D(t){\partial\tilde
W \over \partial p}\> .\end{equation}
for some coefficients $A(t), B(t), C(t), D(t)$ to be determined. This
result immediately leads to the general form Wigner equation :
\begin{eqnarray}
{\partial\tilde W \over\partial t}= & -
& {p\over M}{\partial\tilde W\over \partial
q} + M\Omega^2 q{\partial\tilde W \over \partial p}
+A(t)q{\partial\tilde W\over \partial p} \nonumber \\
& + & B(t){\partial(p\tilde W) \over
\partial p} + C(t){\partial^2 \tilde W\over {\partial p\partial q}}
+ D(t){\partial^2 \tilde W\over{\partial p^2}}\>.
\end{eqnarray}
\section{DETERMINATION OF THE COEFFICIENTS(GENERAL CASE)}
Having found the functional form of the Wigner equation (26) of
the Brownian particle, the next step is to determine the coefficients
in the equation. Undoubtedly, there is more than one way to do this.
Here we shall choose a way which is both mathematically simple
and physically heuristic. Towards this direction, let us
consider the time evolution of the expectation values of the
system variables: $ q, p, q^2, p^2 $ and ${1\over 2}(pq+qp)$.
By using Eq. (16), we have
\begin{eqnarray}
& &{d\over dt}\langle q\rangle = {1\over M}\langle p \rangle \>,\\
& &{d\over dt}\langle p\rangle = -M\Omega^2 \langle q\rangle -
\sum_n C_n\langle q_n \rangle \>,\\
& &{d\over dt}\langle q^2 \rangle = {1\over M}\langle pq+qp \rangle \>, \\
& &{d\over dt}\langle p^2 \rangle = -M\Omega^2 \langle
pq+qp \rangle
-2\sum_n C_n \langle pq_n \rangle \>,\\
& &{d\over dt}\langle pq+qp \rangle = {2\over M}\langle p^2 \rangle - 2M
\Omega^2 \langle q^2 \rangle - 2\sum_n C_n \langle qq_n \rangle \>.
\end{eqnarray}
Similarly, using Eq. (26) yields
\begin{eqnarray}
& &{d\over dt}\langle q\rangle = {1\over M}\langle p \rangle \>,\\
& &{d\over dt}\langle p\rangle = -(M\Omega^2 + A)\langle q\rangle -
B\langle p\rangle \>,\\
& &{d\over dt} \langle q^2 \rangle = {1\over M}\langle pq+qp\rangle \>,\\
& &{d\over dt} \langle p^2 \rangle = -(M\Omega^2 +
A)\langle pq+qp \rangle - 2B\langle p^2 \rangle + 2D \>,\\
& &{d\over dt}\langle pq +qp\rangle =
{2\over M}\langle p^2 \rangle -2(M\Omega^2 +A)
\langle q^2 \rangle - B\langle pq+qp \rangle + 2C \>.
\end{eqnarray}
Since the evolution equations of the expectation values are confined to the
system variables, the above two sets of equations must be
identical.
Now by comparing (28) with (33) we see that
\begin{equation}
\sum_nC_n\langle q_n\rangle =A\langle q\rangle + B\langle p\rangle \>.
\end{equation}
Similarly, by comparing (31) with (36), (30) with (35), respectively, we
get
\begin{eqnarray}
\sum_nC_n\langle qq_n \rangle &=& A\langle q^2\rangle +
{B\over 2}\langle qp+pq\rangle -C \>,\\
\sum_nC_n\langle pq_n \rangle &=& {A\over 2}\langle pq+qp \rangle +
B\langle p^2 \rangle - D \>.
\end{eqnarray}
The coefficients $A, B, C, D $ may now be determined from (37)-(39)
by regarding the expectation values $ \langle q \rangle , \langle
q_nq\rangle$ etc. as expectation values of Heisenberg picture
operators, and by solving the operator equation of motion. For
simplicity, we still use ordinary notation to represent an operator
without adding a hat on it.
The solution to the equation of motion may be written,
\begin{eqnarray}
q_n(t)&=&\alpha_nq(t)+\beta_n p(t)
+\sum_m(a_{nm}q_m(0)+b_{nm}p_m(0)) \>,\\
q(t) &=&\alpha q(0)+\beta p(0)
+\sum_n(a_nq_n(0)+b_np_n(0)) \>,
\end{eqnarray}
for some time-dependent coefficients $\alpha_n, \beta_n, a_{nm},
b_{nm}, \alpha, \beta, a_n, b_n$.
Note that $q_n(t)$ has been expressed in terms of the final, not
initial values of $q, p$.
By substituting Eq. (40) into Eq. (37), keeping (11) in mind, and
comparing the two sides of the resulting equation, we have
\begin{equation}
A=\sum_nC_n\alpha_n\>,\>\> B=\sum_nC_n\beta_n \>.
\end{equation}
Similarly, substituting (40) and (41) into (38) and (39), respectively,
we get
\begin{equation}
C=-\sum_{mn}C_n(a_{nm}a_m\langle q^2_m(0)\rangle +b_{nm}b_m\langle
p^2_m(0)\rangle )\>,
\end{equation}
\begin{equation}
D=-M\sum_{mn}C_n(a_{nm}\dot a_m\langle q^2_m(0)\rangle +b_{nm}\dot b_m\langle
p^2_m(0)\rangle)\>.
\end{equation}
Here we have made use of $p=M\dot q$. The coefficients $A, B, C, D$
are therefore completely determined by solving the equation of motion.
We now do this explicitly.
We have
\begin{equation} \ddot q(t)+ \Omega^2 q(t)=-{1\over M}\sum_n C_n
q_n(t) \>,\end{equation}
\begin{equation}\ddot q_n(t)+ \omega^2_n q_n(t)=-{C_n \over
m_n}q(t) \>.\end{equation}
The solution to Eq. (46) is as follows:
\begin{eqnarray}
q_n(t)&=& q_n(0)\cos(\omega_nt)+{p_n(0)\over m_n}{\sin(\omega_n t)\over
\omega_n}\nonumber \\
& & -C_n\int^t_0 ds\,{\sin[\omega_n(t-s)]\over
\omega_n}{q(s)\over m_n}\>.
\end{eqnarray}
Combining (45) and (47) gives
\begin{equation}
\ddot q(t) + \Omega^2 q(t) +{2\over
M}\int_0^td\tau\eta(t-s)q(s)={f(t)\over M}\>,
\end{equation}
where
\begin{equation}
f(t)=-\sum_nC_n\left(q_n(0)\cos(\omega_nt) + {p_n(0)\over
m_n}{\sin(\omega_nt) \over \omega_n}\right).
\end{equation}
The kernel $\eta(s)$ is defined as
\begin{equation}
\eta(s) ={d\over ds}\gamma(s)\>,
\end{equation}
where
\begin{equation}
\gamma(s)=\int^{+\infty}_0d\omega\,{I(\omega)\over \omega}\cos(\omega s) \>.
\end{equation}
Here $I(\omega)$ is the spectral density of the environment:
\begin{equation}
I(\omega)=\sum_n\delta(\omega-\omega_n){C_n^2\over 2m_n\omega_n}\>.
\end{equation}
In order to get the expressions (40) and (41), we solve equation (48)
with
the following two different initial conditions:
\begin{equation}
q(s=0)=q(0)\>,\,\> \dot q(s=0)={p(0)\over M}\>.
\end{equation}
and
\begin{equation}
q(s=t)=q(t)\>,\,\> \dot q(s=t)={p(t)\over M}\>.
\end{equation}
where $t$ is any given time point. In doing so, we consider the
elementary functions $u_i(s) (i=1,2)$ introduced by
Hu, Paz and Zhang [10] which satisfy the following
homogeneous integro-differential equation
\begin{equation}{\ddot\Sigma}(s)+\Omega^2\Sigma(s)+{2\over M}\int_0
^s\,d\lambda\eta(s-\lambda)\Sigma(\lambda)=0
\end{equation}
with the boundary conditions:
\begin{equation}u_1(s=0)=1\>,\,\, u_1(s=t)=0\> ,
\end{equation}
and
\begin{equation}u_2(s=0)=0\>,\,\, u_2(s=t)=1 \>.
\end{equation}
The solution to equation (55) with the initial condition (53) is
obtained as the linear combination of $u_1,u_2$,
\begin{equation}
w(s)=\left(u_1(s)-{\dot u_1(0)\over \dot
u_2(0)}u_2(s)\right)q(0)+{u_2(s)\over \dot u_2(0)}\,{p(0)\over M}\>.
\end{equation}
The solution to equation (48) with the homogeneous initial conditions
can be formally written as
\begin{equation}
\tilde w(s)={1\over M}\int^s_0 d\tau G_1(s,\tau)f(\tau) \>.
\end{equation}
Where $G_1(s_1,s_2)$ is the Green function which can be constructed in
terms of $u_i(i=1,2)$:
\[G_1(s_1,s_2)= \left\{\begin{array}{ll}
{u_1(s_2)u_2(s_1)-u_2(s_2)u_1(s_1)\over
u_1(s_2)\dot u_2(s_2)-\dot u_1(s_2)u_2(s_2)}
&\mbox{if $s_1>s_2$}\\
0 &\mbox{otherwise.}
\end{array}
\right. \]
Note that $G_1(s_1,s_2)$ as the function of $s_1$
satisfies equation (55) with
\begin{equation}
G_1(s_1=s_2,s_2)= 0\>,\> \> {d\over ds_1}G_1(s_1=s_2,s_2)=1\>.
\end{equation}
Then the solution to the equation (48) with initial condition (53) reads
\begin{equation}
q(s)= w(s) + \tilde w(s),
\end{equation}
explicitly,
\begin{eqnarray}
q(s)&=&\left(u_1(s)-{\dot u_1(0)\over \dot
u_2(0)}u_2(s)\right)q(0)+
{u_2(s)\over \dot u_2(0)}\,p(0) \nonumber \\
& &-\sum_n{C_n\over M}\int^s_0 d\tau\,G_1(s,\tau)\cos(\omega_n\tau)
q_n(0)\nonumber \\
& & - \sum_n{C_n\over M}\int^s_0d\tau\, G_1(s,\tau){\sin(\omega_n\tau)\over
\omega_n}{p_n(0)\over m_n} .
\end{eqnarray}
It can be shown that the solution to the homogeneous equation (55)
with the initial conditions (54) is
\begin{equation}
u(s)=\left(u_2(s)-{\dot u_2(t)\over \dot u_1(t)}u_1(s)\right )q(t)+{u_1(s)\over
\dot u_1(t)}\,{p(t)\over M}\>,
\end{equation}
and
\begin{equation}
\tilde u(s) ={1\over M} \int^s_t d\tau G_2(s,\tau)f(\tau),\>(s\leq t)
\end{equation}
is the solution to the inhomogeneous equation (48) with the homogeneous
initial conditions
\begin{equation}
\tilde u(t)=0\>,\> \dot{\tilde u}(t)=0\>,
\end{equation}
where Green function $G_2(s_1,s_2)$ is
\[G_2(s_1,s_2)=\left\{\begin{array}{ll}
{u_1(s_2)u_2(s_1)-u_2(s_2)u_1(s_1)\over
u_1(s_2)\dot u_2(s_2)-\dot u_1(s_2)u_2(s_2)}
&\mbox{if $s_2>s_1$}\\
0 &\mbox{otherwise.}
\end{array}
\right. \]
Hence, we get the solution to Eq. (48) with the initial conditions
(54)
\begin{eqnarray}
q(s)&=& u(s)+\tilde u(s)\nonumber \\
&=& \left(u_2(s)-{\dot u_2(t)\over \dot u_1(t)}u_1(s)\right)q(t)+
{u_1(s)\over \dot u_1(t)}\,{p(t)\over M}\nonumber \\
& & + \sum_n{C_n\over M}\int^t_sd\tau G_2(s,\tau)\cos(\omega_n\tau)\,
q_n(0)\nonumber \\
& & +\sum_n{C_n\over M}\int^t_sd\tau G_2(s,\tau){\sin(\omega_n\tau)\over
\omega_n}{p_n(0)\over m_n} \>.
\end{eqnarray}
Substituting (66) into (47), one obtains
\begin{eqnarray}
q_n(t)&=&-{C_n\over
m_n\omega_n}\int^t_0ds\,\sin[\omega_n(t-s)]\left(u_2(s)-
{\dot u_2(t)\over \dot u_1(t)}u_1(s)\right )q(t)\nonumber \\
& &-{C_n\over m_n\omega_n}\int^t_0ds\,\sin[\omega_n(t-s)]{u_1(s)\over
\dot u_1(t)}\,{p(t)\over M}\nonumber \\
& &+q_n(0)\cos(\omega_nt)+
{p_n(0)\over m_n}{\sin(\omega_nt)\over\omega_n}\nonumber \\
& &-{1\over M} \sum_m{C_nC_m\over m_n\omega_n}\int^t_0ds\int^t_sd\tau
\sin[\omega_n(t-s)]G_2(s,\tau)\cos(\omega_m
\tau)\,q_m(0)\nonumber \\
& &-{1\over M} \sum_m{C_nC_m\over m_n\omega_n}\int^t_0ds\int^t_sd\tau
\sin[\omega_n(t-s)]\,G_2(s,\tau){\sin(\omega_m\tau)\over
m_m\omega_m}\>p_m(0) \>.
\end{eqnarray}
By using (42) we immediately arrive at
\begin{eqnarray}
A(t) &=&-\sum_n{C^2_n\over m_n\omega_n}\int^t_0ds\,\sin[\omega_n(t-s)]
\left(u_2(s)-{\dot u_2(t)\over \dot u_1(t)}u_1(s)\right )\>, \\
B(t) &=&- {1\over M}\sum_n{C^2_n\over
m_n\omega_n}\int^t_0ds\,\sin[\omega_n(t-s)]{u_1(s)\over\dot u_1(t)}\>
{}.
\end{eqnarray}
Furthermore, $A,B$ can be written as
\begin{eqnarray}
A(t) &=& 2\int^t_0ds\eta(t-s)u_2(s)-2{\dot u_2(t)\over \dot u_1(t)}
\int^t_0ds\eta(t-s)u_1(s)\>,\\
B(t) &=& {2\over M\dot u_1(t)}\int^t_0ds\eta(t-s)u_1(s)\>.
\end{eqnarray}
{}From (62), the momentum of the Brownian particle is then
\begin{eqnarray}
p(s)&=&M\dot q(s)\nonumber \\
&=&\left(\dot u_1(s)-{\dot u_1(0)\over \dot u_2(0)}\dot
u_2(s)\right)\,Mq(0)+{\dot u_2(s)\over \dot
u_2(0)}\,p(0)\nonumber \\
& &-\sum_n\int^s_0d\tau
G'_1(s,\tau)\cos(\omega_n\tau)\,q_n(0)\nonumber \\
& &-\sum_n\int^s_0d\tau
G'_1(s,\tau){\sin(\omega_n\tau)\over \omega_n}\,{p_n(0)\over m_n}\>.
\end{eqnarray}
Here ``prime'' stands for derivative with respect to the first variable
of $G_1(s,\tau)$.
With these results, It can be easily shown that
\begin{eqnarray}
C(t)&=&{\hbar\over M}\int^t_0\,d\lambda G_1(t,\lambda)\nu(t-\lambda)\nonumber
\\
& & -{2\hbar\over M^2}\int^t_0ds\int^t_sd\tau \int^t_0d\lambda
\eta(t-s)G_1(t,\lambda)G_2(s,\tau)\nu(\tau-\lambda)\>,
\end{eqnarray}
and
\begin{eqnarray}
D(t)&=&\hbar\int^t_0\,d\lambda G'_1(t,\lambda)\nu(t-\lambda)\nonumber \\
& & -{2\hbar\over M}\int^t_0ds\int^t_sd\tau \int^t_0d\lambda
\eta(t-s)G'_1(t,\lambda)G_2(s,\tau)\nu(\tau-\lambda)\>.
\end{eqnarray}
where $\nu(s)$ is defined as
\begin{equation}
\nu(s)=\int^{+\infty}_0d\omega I(\omega)\coth({1\over
2}\hbar\omega\beta)\cos(\omega
s)\>.
\end{equation}
It is seen that the coefficients $A(t), B(t), C(t), D(t)$ are dependent only on
the
kernels $\eta(s)$ and $\nu(s)$ and the initial state of the bath, not
dependent on the initial state of the system. Once the spectral density of the
environment is given, the elementary functions $u_i (i=1,2)$ can be
solved from equation (55), the Green functions $G_i (i=1,2)$ are then obtained.
Thus the coefficients of master equation can be
determined.
\section{PARTICULAR CASES}
In this section, we will consider some special cases. Let us at
first treat a special case in which we assume that the interaction
between the system and environment is weak, so the $C_n$ are small.
In this case, the coefficients are of simple forms, and the
determination of these coefficients is very simple and
straightforward. We shall work out these coefficients directly using
the method in the last section, rather than the general formulae.
The solution to Eq. (46) may be written as
\begin{eqnarray}
q_n(t)& = & q_n(0) \cos (\omega_nt) +
{p_n(0)\over m_n}{\sin (\omega_nt) \over \omega_n}
\nonumber \\
&-& {C_n\over m_n} \int^t_0 dt'{\sin[\omega_n(t-t')] \over \omega_n}
\cos[\Omega(t'-t)]\,q(t)\nonumber \\
&-& {C_n\over m_n} \int^t_0 dt'{\sin[\omega_n(t-t')] \over \omega_n}
{\sin[\Omega(t'-t)]\over \Omega }\,{p(t)\over M} + O(C_n^2) \>.
\end{eqnarray}
Using Eq. (37) and ignoring terms with higher than the second
order of $C_n$, we get
\begin{eqnarray}
\sum_nC_n \langle q_n(t) \rangle & = & \lbrace \sum_n
-{C_n^2\over m_n} \int^t_0 dt'
{\sin[\omega_n(t- t')]\over \omega_n}
\cos[\Omega (t'-t)] \rbrace \langle q(t) \rangle \nonumber \\
& + & \lbrace {1\over M}\sum_n -{C_n^2\over m_n}
\int^t_0 dt' {\sin [\omega (t-t')] \over
\omega_n}{\sin [\Omega (t'-t)]\over \Omega} \rbrace \langle p(t) \rangle \>.
\end{eqnarray}
Then we immediately get
\begin{equation} A(t)=2 \int^t_0 ds \eta (s) \cos (\Omega s) \>,
\end{equation}
\begin{equation}B(t)=-{2\over M\Omega}\int^t_0 ds \eta (s) \sin(\Omega s)\>
.\end{equation}
We next evaluate $\sum_n C_n \langle q(t)q_n(t) \rangle $ and $\sum_n
C_n\langle p(t)q_n(t)\rangle$. After a few
manipulations, we arrive at the expressions
\begin{equation}C(t)=- \sum_nC_n\lbrace \langle q(t)q_n(0)\rangle \cos
(\omega_nt) + \langle q(t)p_n(0)\rangle{\sin(\omega_nt)\over
m_n\omega_n}
\rbrace \>,
\end{equation}
and
\begin{equation}D(t)=- \sum_nC_n \lbrace \langle p(t)q_n(0)\rangle\cos
(\omega_nt)+ \langle p(t)p_n(0)\rangle {\sin(\omega_nt) \over
m_n\omega_n}\rbrace \>.
\end{equation}
To calculate $C(t)$ and $D(t)$, we need to expand $q(t)$ up to the
second order of $C_n$:
\begin{eqnarray}
q(t)& = & q(0)\cos(\Omega t) + {p(0)\over M}{\sin(\Omega t)\over \Omega}
\nonumber \\
&-& \sum_n{C_n\over M}\int^t_0ds{\sin[\Omega (t-s)]\over
\Omega}\cos(\omega_ns)\>q_n(0)
\nonumber \\
&-& \sum_n{C_n\over M}\int^t_0ds{\sin[\Omega(t-s)]\over
\Omega}{\sin(\omega_ns)
\over \omega_n}\>{p_n(0)\over m_n} \nonumber \\
& + & O(C_n^2) \> .
\end{eqnarray}
The expansion of $p(t)$ is easily obtained from that of $q(t)$,
\begin{equation}p(t)=M\dot q(t) .\end{equation}
With these results it is easy to compute $C(t)$ and $D(t)$:
\begin{equation}C(t)={\hbar\over
M\Omega}\int^t_0ds\nu(s)\sin(\Omega s) ,\end{equation}
\begin{equation}D(t)=\hbar\int^t_0ds\nu(s)\cos(\Omega s) .\end{equation}
This simple example exhibits the time dependency of the coefficients
of the master equation in a general environment. Eqs. (78), (79),
(84), (85) are in agreement with Hu, Paz and Zhang [10].
As another example, we briefly discuss the purly Ohmic case in the
Fokker-Planck limit (a particular high temperature limit), which has
been extensively discussed in the literature [6,10]. In this case one has
\begin{eqnarray}
\eta(s-s')&=&M\gamma\delta'(s-s')\\
\nu(s-s') &=&{2M\gamma k_BT\over \hbar}\delta(s-s')
\end{eqnarray}
Then the equation (55) reduces to
\begin{equation}
\ddot u(s)+\Omega_{\rm ren}^2u(s)+\gamma\dot u(s)=-2\gamma\delta(s)u(0)
\end{equation}
where $\Omega^2_{\rm ren}=\Omega^2 -2\gamma\delta(0)$. After solving this
equation, a few calculations give
\begin{eqnarray}
A(t)&=&-2M\gamma\delta(0),\\
B(t)&=&2\gamma,\\
C(t)&=&0,\\
D(t)&=&2M\gamma k_BT,
\end{eqnarray}
then the Wigner equation reads:
\begin{equation}
{\partial \over \partial t}\tilde W =-{p\over M}{\partial\tilde W
\over \partial
q}+M\Omega^2_{\rm ren}q{\partial\tilde W \over \partial
p}+2\gamma{\partial \tilde W\over \partial p} +2M\gamma
k_BT{\partial^2\tilde W\over
\partial p^2}\>.
\end{equation}
In this regime, the coefficients of this Wigner equation are
constants.
\section*{Acknowledgements}
We are grateful to Roland Omn\`{e}s for suggesting this method of
deriving the master equation, and encouraging us to pursue it.
|
\section{Introduction}
One of the more interesting results of the study of quantum field
theory
in curved spacetime is the fact that the expectation value of the
trace
of the stress tensor of a conformally coupled field does not vanish.
It
has an anomaly. This trace or conformal anomaly, as it is known, was
first
noticed by Capper and Duff \cite{3} using a dimensional
regularization scheme.
Since then many other regularization procedures have been used and
when used
correctly lead to same result \cite{4}.
Unfortunately, as anyone who has ever calculated this trace
anomaly knows, the computations required are rather lengthy and
certainly
less than illuminating. On the other hand, if one has a particle
interpretation the problem can be handled more simply. This fact was
first exploited by Massacand and Schmid \cite{1}. In this paper we
adapt their
method to a computation in 1+1 dimensions using only the following
two
inputs.
\noindent 1) The frame components of the stress tensor at a given
point are,
for two frames based at this point, related by a Lorentz
transformation.
\noindent 2) The vacuum expectation value of the energy momentum
density
(relative to a given frame) should vanish. Thus, the vacuum can have
pressure,
but no energy or momentum.
In general there would remain the vexing question, ``Which
vacuum?" The answer we propose is to use the coordinate independent
definition of Capri and Roy.
In section II we give a brief review of this construction of
the
vacuum and apply the result to a calculation of the vacuum
expectation
value of the trace of the stress tensor in section III. Our
conclusions are
set out in section IV.
\section{Coordinate independent definition of time and vacuum}
In a globally hyperbolic spacetime one can choose a foliation based
solely on geodesics. Thus, given a timelike (unit) vector
$N_{\mu}(P_0)$ at
the
point $P_0$ one establishes a frame (zweibein) at $P_0$ with
components:
\begin{eqnarray}
e^{\mu\hat{0}} &=& N^\mu(P_0) \nonumber \\
e^{\mu\hat{1}} &=& p^\mu(P_0)
\label{1}
\end{eqnarray}
where $p^\mu(P_0)$ is a unit vector orthogonal to $N^\mu(P_0)$ at
$P_0$. The
spacelike hypersurface (line) consiting of the geodesic through $P_0$
with
tangent vector $p^\mu(P_0)$ defines the surface $t=0$. The ``time"
$t$
corresponding to an arbitrary point $P$ is the distance along a
geodesic
$P_1-P$ which intersects the line $t=0$ orthogonally at some point
$P_1$ .
The geodesic distance $P_0-P_1$ along the line $t=0$ yields the space
coordinate $x$. These geodesic normal coordinates prove to be very
useful since
in these coordinates the metric becomes
\begin{equation}
ds^2 = dt^2 - \alpha^2(t,x) dx^2
\label{2}
\end{equation}
where
\begin{eqnarray}
\alpha(0,0) &=& 1 \nonumber \\
\frac{\partial\alpha}{\partial t}|_{t=0} = \frac{\partial
\alpha}{\partial x}|_{t=0} = &0& =
\frac{\partial^2\alpha}{\partial t \partial x}|_{P_0} =
\frac{\partial^2\alpha}{\partial x^2}|_{P_0}
\label{3}
\end{eqnarray}
Also,
\begin{equation}
\frac{2}{\alpha}\frac{ \partial^2 \alpha}{\partial t^2} = R
\label{4}
\end{equation}
where $R$ is the curvature scalar.
The field equations in these coordinates, for a massless scalar field
read:
\begin{eqnarray}
\frac{1}{\sqrt{g}}\partial_\mu(
\sqrt{g}g^{\mu \nu}\partial_\nu)\phi&=&0 \nonumber \\
\frac{\partial^2 \phi}{\partial t^2 } +
\frac{\dot{\alpha}}{\alpha}\frac{\partial \phi }{\partial t}
+ \frac{\alpha'}{\alpha^3}\frac{\partial \phi }{\partial x }
- \frac{1}{\alpha^2}\frac{\partial^2 \phi }{\partial x^2 } &=&0.
\label{5}
\end{eqnarray}
Here,
\begin{equation}
\dot{\alpha}=\frac{\partial \alpha }{\partial t}
\ \ \ {\rm and}\ \ \ \alpha'=\frac{\partial \alpha }{\partial x}.
\end{equation}
The positive frequency modes $\phi$ of this field are obtained by
solving
these field equations with the two initial conditions
\begin{equation}
1)\ \ \ \ \ \ \ \ \ \ \phi_{p,\epsilon}(0,x) =
\frac{1}{\sqrt{4\pi p}}
\exp(- i p \epsilon x)\ \ \ \ \ \ p>0.
\label{6}
\end{equation}
Here we have
\begin{eqnarray}
\epsilon &=& +1\ \ \ {\rm corresponds\ \ to\ \ left\ \
travelling \ \
waves}\nonumber \\
\epsilon &=& -1\ \ \ {\rm corresponds\ \ to\ \ right\ \
travelling\ \
waves. }
\label{7}
\end{eqnarray}
and
\begin{equation}
2)\ \ \ \ \ \ \ \ \ \ i \frac{\partial
\phi_{p,\epsilon}}{\partial
t}|_{t=0} = p \phi_{p,\epsilon}|_{t=0}.
\label{8}
\end{equation}
A useful Ansatz to implement these initial conditions is:
\begin{equation}
\phi_{p,\epsilon}(t,x) = \frac{1}{\sqrt{4\pi p}} \exp(-i p
f_{\epsilon}(t,x))
\label{9}
\end{equation}
where $f_\epsilon$ is real.
Equation (\ref{5}) then yields that
\begin{equation}
\frac{\partial f_\epsilon}{\partial t} = \frac{\epsilon}{\alpha}
\frac{\partial f_\epsilon}{\partial x}
\label{10}
\end{equation}
The initial conditions become
\begin{equation}
f_\epsilon(0,x) = \epsilon x
\label{11}
\end{equation}
and near $t = 0$
\begin{equation}
f_\epsilon(t,x) \approx t + \epsilon x
\label{12}
\end{equation}
The quantized field is now given by
\begin{equation}
\Psi(t,x) = \sum_{\epsilon=\pm 1}
\int_{0}^\infty d(\epsilon p)
\left(
\phi_{p,\epsilon}(t,x)a_{p,\epsilon} +
\phi_{p,\epsilon}^\ast(t,x)a_{p,\epsilon}^{\dagger}
\right)
\label{13}
\end{equation}
with the vacuum defined by
\begin{equation}
a_{p,\epsilon} |0> = 0 .
\label{14}
\end{equation}
These modes have been normalized such that
\begin{eqnarray}
(\phi_{p,\epsilon} ,\phi_{q,\epsilon}) &=& i\epsilon
\int_{-\infty}^{\infty}dx\sqrt{g}\left( \phi^{\ast}_{p,\epsilon}(t,x)
\stackrel{\leftrightarrow}{\partial_{t}}
\phi_{q,\epsilon}(t,x)
\right)\nonumber \\
&=& \frac{p+q}{4\pi\sqrt{pq}}\int_{-\infty}^{\infty}dx \epsilon
\alpha
\frac{\partial f_\epsilon }{\partial
t}\exp(i(p-q)f_\epsilon(t,x))\nonumber \\
&=& \frac{p+q}{4\pi\sqrt{pq}}\int_{-\infty}^{\infty}dx \frac{\partial
f_\epsilon }{\partial x}\exp(i(p-q)f_\epsilon(t,x))\nonumber \\
&=& \frac{p+q}{4\pi\sqrt{pq}} 2\pi \delta(p-q) \nonumber \\
&=& \delta (p-q)
\label{15}
\end{eqnarray}
\section{The trace anomaly}
We begin with two ``observers" with tangents to their world lines
given by
\begin{equation}
N^\mu(P_0) = (1,0)\ \ \ {\rm and}\ \ \ \bar{N}^\mu(P_0)
= (\cosh(\chi),\frac{\sinh(\chi)}{\bar{\alpha}})
\label{16}
\end{equation}
The corresponding frames are:
\begin{eqnarray}
e^{\mu \hat{0}} = (1,0)\ \ &\ &\ \ e^{\mu \hat{1}} = (0,
\frac{1}{\alpha})
\label{17} \\
\bar{e}^{\mu \hat{0}} =
(\cosh(\chi),\frac{\sinh(\chi)}{\bar{\alpha}})
\ \ &\ &\ \ \bar{e}^{\mu \hat{1}} =
(\sinh(\chi),\frac{\cosh(\chi)}{\bar{\alpha}})
\label{18}
\end{eqnarray}
Corresponding to this the metric has the two forms
\begin{equation}
ds^2 = dt^2 - \alpha^2(t,x) dx^2 = d\bar{t}^2 -
\bar{\alpha}^2(\bar{t},\bar{x})
d\bar{x}^2
\label{19}
\end{equation}
We can solve for the positive frequency modes in the barred as well
as in
the unbarred coordinates to obtain the corresponding quantized fields
$\bar{\Psi}(\bar{t},\bar{x})$ and $\Psi(t,x)$. Their respective sets
of
annihilation and
creation operators are
$(\bar{a}_{p,\epsilon} , \bar{a}^\dagger_{p,\epsilon})$ and
$(a_{p,\epsilon}
,a^\dagger_{p,\epsilon})$ .
At $P_0$, the point with coordinates $(0,0)$ in both
coordinate systems the two fields coincide.
Corresponding to these two fields we
have their respective Fock space vacuums $|\bar{0}>$ , $|0>$
defined by
\begin{equation}
\bar{a}_{p,\epsilon} |\bar{0}> = 0\ \ \ ,\ \ \ a_{p,\epsilon} |0> =
0
\label{20}
\end{equation}
Any bilinear expression in the field operators which, for physical
reasons, should have vanishing vacuum expectation value is defined by
normal ordering with respect to its own vacuum. Thus since we expect
the
vacuum to be the state of zero energy and momentum density we require
that
\begin{equation}
<\bar{0}|:\bar{T}^{\hat{0}\hat{\mu}}:|\bar{0}> = 0
\label{20.5}
\end{equation}
and
\begin{equation}
<0|:T^{\hat{0}\hat{\mu}}:|0> = 0 ,
\label{21}
\end{equation}
where,
\begin{eqnarray}
T^{\hat{\alpha}\hat{\beta}} &=&
e^{\mu\hat{\alpha}}e^{\nu\hat{\beta}}T_{\mu\nu}
\nonumber \\
\bar{T}^{\hat{\alpha}\hat{\beta}} &=&
\bar{e}^{\mu\hat{\alpha}}\bar{e}^{\nu\hat{\beta}}\bar{T}_{\mu\nu}.
\end{eqnarray}
Furthermore, since the barred and unbarred frames $\bar{e}^{\mu
\hat{\alpha}}$
,
$e^{\mu \hat{\alpha}}$ are related by a Lorentz transformation
\begin{equation}
\Lambda^{\hat{\beta}}_{ \ \hat{\alpha}} = \pmatrix{
\cosh(\chi) & \sinh(\chi) \cr
\sinh(\chi) & \cosh(\chi) \cr}
\label{22}
\end{equation}
we have that at $P_0$
\begin{equation}
:T^{\hat{\alpha} \hat{\beta}}:|_{P_0} =
\Lambda^{\hat{\alpha}}_{\ \hat{\gamma}}
\Lambda^{\hat{\beta}}_{ \ \hat{\delta}}
:\bar{T}^{\hat{\gamma}\hat{\delta}}:|_{P_0}
\label{23}
\end{equation}
so that in particular
\begin{equation}
:T^{\hat{0}\hat{0}}:|_{P_0} =
\cosh^2(\chi):\bar{T}^{\hat{0}\hat{0}}:|_{P_0} +
2\cosh(\chi)\sinh(\chi):\bar{T}^{\hat{0}\hat{1}}:|_{P_0} +
\sinh^2(\chi):\bar{T}^{\hat{1}\hat{1}}:|_{P_0} .
\label{24}
\end{equation}
Taking the vacuum expectation value with respect to the barred vacuum
of
this equation, and using (\ref{20.5}) we have
\begin{equation}
<\bar{0}|:T^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}> = \sinh^2(\chi)
<\bar{0}|:\bar{T}^{\hat{1}\hat{1}}:|_{P_0}|\bar{0}>
\label{25}
\end{equation}
Since $<\bar{0}|:\bar{T}^{\hat{0}\hat{0}}:|\bar{0}> = 0$ we find
that the vacuum expectation value of the trace is:
\begin{equation}
<\bar{0}|\eta_{\hat{\alpha}\hat{\beta}} :
\bar{T}^{\hat{\alpha}\hat{\beta}}:|_{P_0}|\bar{0}> = -
<\bar{0}|:\bar{T}^{\hat{1}\hat{1}}:|_{P_0}|\bar{0}> = -
\frac{1}{\sinh^2(\chi)}
<\bar{0}|:T^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}>
\label{26}
\end{equation}
To evaluate this expression we have to take the term
$:T^{\hat{0}\hat{0}}:|_{P_0}$ which has
been normal ordered with respect to the vacuum $|0>$ , rewrite it in
terms
of the operators
$(\bar{a}_{p,\epsilon},\bar{a}^\dagger_{p,\epsilon})$ and
commute the
terms so that the resulting expression is normal ordered with
respect to the
vacuum $|\bar{0}>$ . To do this we write out the term
$:T^{\hat{0}\hat{0}}:|_{P_0}$
explicitly. A simplification due to the use of equation (\ref{10})
occurs so
that only time derivatives of the field operators appear. Also since
$\Psi(t,x)=\bar{\Psi}(\bar{t},\bar{x})$ we may
write
\begin{eqnarray}
:T^{\hat{0}\hat{0}}:|_{P_0} &=& :\frac{\partial \Psi}{\partial t}
\frac{\partial
\Psi}{\partial t}:|_{P_0}
= :\frac{\partial \bar{\Psi}}{\partial t} \frac{\partial
\Psi}{\partial
t}:|_{P_0} \nonumber \\
&=& \sum_{\epsilon=\pm 1} \int d(\epsilon p)
[\frac{\partial \bar{\Psi}}{\partial t}
\frac{\partial\phi_{p,\epsilon}}{\partial t} a_{p,\epsilon} +
\frac{\partial\phi^{\ast}_{p,\epsilon}}{\partial t}
a^\dagger_{p,\epsilon} \frac{\partial \bar{\Psi}}{\partial
t}]|_{P_0}.
\label{27}
\end{eqnarray}
To simplify the notation we drop the $|_{P_0}$ , but keep in mind
that these
equations only apply at the point $P_0$. Also we only evaluate this
expression for a fixed $\epsilon$. Thus,
\begin{eqnarray}
:T_\epsilon^{\hat{0}\hat{0}}: = \int_{0}^{\infty} d(\epsilon p)
\int_{0}^{\infty} d(\epsilon q) \left[ \right. & &
(\frac{\partial \bar{\phi}_{q,\epsilon}}{\partial
t}\bar{a}_{q,\epsilon}+
\frac{\partial \bar{\phi}^{\ast}_{q,\epsilon}}{\partial
t}\bar{a}^{\dagger}_{q,\epsilon})a_{p,\epsilon}\frac{\partial
\phi_{p,\epsilon}}{\partial t} \nonumber \\
&+& \left. \frac{\partial \phi^{\ast}_{p,\epsilon}}{\partial
t}a^{\dagger}_{p,\epsilon}
(\frac{\partial \bar{\phi}_{q,\epsilon}}{\partial
t}\bar{a}_{q,\epsilon}+
\frac{\partial \bar{\phi}^{\ast}_{q,\epsilon}}{\partial
t}\bar{a}^{\dagger}_{q,\epsilon})\right]
\label{28}
\end{eqnarray}
The operators $(a_{p,\epsilon} , a^\dagger_{p,\epsilon})$ are related
to the
barred
operators $(\bar{a}_{k,\epsilon} ,\bar{a}^\dagger_{k,\epsilon})$ by
a
Bogolubov
transformation
\begin{equation}
a_{k,\epsilon} = \int d(\epsilon q)(\alpha_{k,q} \bar{a}_{q,\epsilon}
+
\beta^{\ast}_{k,q} \bar{a}^\dagger_{q,\epsilon})
\label{29}
\end{equation}
where
\begin{eqnarray}
\alpha_{k,q} &=& (\phi_{k,\epsilon} , \bar{\phi}_{q,\epsilon})
\nonumber \\
\beta_{k,q} &=& (\phi^{\ast}_{k,\epsilon} , \bar{\phi}_{q,\epsilon})
\label{30}
\end{eqnarray}
In our evaluation of the vacuum expectation value, the only term of
interest is the c-number term that results from the commutator
\begin{equation}
\bar{a}_{q,\epsilon'} \bar{a}^\dagger_{k,\epsilon} =
\bar{a}^\dagger_{k,\epsilon} \bar{a}_{q,\epsilon'} +
\delta_{\epsilon,\epsilon'} \delta(k-q)
\end{equation}
Thus, we get
\begin{equation}
<\bar{0}|:T_\epsilon^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}> = \int
d(\epsilon p)
d(\epsilon q) [\frac{\partial \bar{\phi}_{q,\epsilon} }{\partial
t}
\frac{\partial \phi_{p,\epsilon}}{\partial t} \beta^\ast_{p,q} +
c.c. ]
\label{31}
\end{equation}
These terms are evaluated by replacing $\beta$ by its expression
(\ref{30}) and
interchanging the order of integration to first do the momentum
integrals.
In doing so the only regularization required is to define an integral
of
the form
\begin{equation}
\int_0^\infty dx x\exp(ixp)
\label{32}
\end{equation}
This is accomplished by replacing $p$ by $p + i \delta$ . No further
regularizations are needed. Further details of such a calculation
are in the appendix as well as the paper by Massacand and Schmid
\cite{1} and yield a Schwarz derivative. The
final result is:
\begin{equation}
<\bar{0}|:T_\epsilon^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}> =
\frac{1}{24\pi}
\frac{\frac{\partial^3\bar{f}_{\epsilon}}{\partial x^3}|_{P_0}}{\frac{\partial
\bar{f}_{\epsilon}}{\partial x}|_{P_0}}
\label{33}
\end{equation}
So we only have to evaluate these terms. Now,
\begin{equation}
\frac{\partial \bar{f}_{\epsilon}}{\partial x} = \frac{\partial
\bar{f}_{\epsilon}}{\partial \bar{x}}
\frac{\partial \bar{x}}{\partial x} + \frac{\partial
\bar{f}_{\epsilon}}{\partial \bar{t}}
\frac{\partial \bar{t}}{\partial x}
\label{34}
\end{equation}
and as initial conditions at $P_0$ we have
\begin{equation}
\frac{\partial \bar{x}}{\partial x}|_{P_0} = \cosh(\chi)\ \ \ \ ,
\ \ \
\ \frac{\partial \bar{t}}{\partial x}|_{P_0} = \sinh(\chi)
\label{35}
\end{equation}
Furthermore, we also have that
\begin{equation}
\bar{f}_\epsilon(0,\bar{x}) = \epsilon \bar{x}\ \ \ \ , \ \ \ \
\frac{\partial\bar{f}_{\epsilon}}{\partial \bar{t}} |_{P_0} = 1 \ \ \
\ , \ \ \ \
\frac{\partial \bar{f}_{\epsilon}}{\partial \bar{x}}|_{P_0} = \epsilon
\bar{\alpha}\frac{\partial \bar{f}_{\epsilon}}{\partial \bar{t}}|_{P_0} =
\epsilon
\label{35.5}
\end{equation}
since $\bar{\alpha}|_{P_0} = 1$ .
Also, as we stated earlier,
\begin{eqnarray}
\frac{\partial \bar{\alpha}}{\partial \bar{t}}|_{P_0} =
\frac{\partial
\bar{\alpha}}{\partial \bar{x}}|_{P_0} &=& 0
\label{36} \\
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}\partial
\bar{x}}|_{P_0} =
\frac{\partial^2\bar{\alpha}}{\partial \bar{x}^2}|_{P_0} &=& 0
\label{37}
\end{eqnarray}
and
\begin{equation}
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2}|_{P_0} =
\frac{R}{2} .
\label{38}
\end{equation}
By repeatedly using the barred version of equation (\ref{10}) ,
namely
\begin{equation}
\frac{\partial \bar{f}_{\epsilon}}{\partial \bar{x}} = \epsilon \bar{\alpha}
\frac{\partial \bar{f}_{\epsilon}}{\partial \bar{t}}
\label{39}
\end{equation}
as well as
(\ref{34}), (\ref{35}) and (\ref{36}) we find:
\begin{equation}
\frac{\partial^2\bar{f}_{\epsilon}}{\partial \bar{t}^2}|_{P_0} =
\frac{\partial^2\bar{f}_{\epsilon}}{\partial \bar{t}\partial \bar{x}}|_{P_0} =
\frac{\partial^2\bar{f}_{\epsilon}}{\partial \bar{x}^2}|_{P_0} = 0
\label{40}
\end{equation}
as well as
\begin{equation}
\frac{\partial^3\bar{f}_{\epsilon}}{\partial \bar{x}^3}|_{P_0} = \epsilon
\frac{\partial^3\bar{f}_{\epsilon}}{\partial \bar{t}^3}|_{P_0} + \epsilon
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2} \frac{\partial
\bar{f}_{\epsilon}}{\partial \bar{t}}|_{P_0} =0
\label{41}
\end{equation}
Thus we arrive at the result that
\begin{equation}
\frac{\partial^3\bar{f}_{\epsilon}}{\partial \bar{t}^3}|_{P_0} = -
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2}|_{P_0} = -
\frac{R}{2}|_{P_0}
\label{42}
\end{equation}
This result now allows us to obtain that
\begin{eqnarray}
\frac{\partial^3\bar{f}_{\epsilon}}{\partial x^3}|_{P_0} &=&
[\frac{\partial^3\bar{t}}{\partial x^3} + \epsilon \bar{\alpha}
\frac{\partial^3\bar{x}}{\partial x^3} +
\epsilon \frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2}
\frac{\partial
\bar{x}}{\partial x} (\frac{\partial \bar{t}}{\partial x})^2]
\frac{\partial
\bar{f}_{\epsilon}}{\partial \bar{t}}|_{P_0} +
(\frac{\partial \bar{t}}{\partial x} +\epsilon \frac{\partial
\bar{x}}{\partial
x}) \frac{\partial^3\bar{f}_{\epsilon}}{\partial \bar{t}^3} (\frac{\partial
\bar{t}}{\partial x})^2|_{P_0} \nonumber \\
&=& -\frac{R}{2} \sinh^3(\chi) +
\frac{\partial^3\bar{t}}{\partial x^3} +
\epsilon \frac{\partial^3\bar{x}}{\partial x^3}
\label{43}
\end{eqnarray}
To evaluate the last two terms in this expression we use the fact
that
$(t,x)$ as well as $(\bar{t},\bar{x})$ satisfy the geodesic
equations, but have
different initial data on the spacelike geodesic that passes through
$P_0$.
These initial data are:
\begin{eqnarray}
\frac{d x}{ds}|_{P_0} = 1 \ \ \ \ &,& \ \ \ \
\frac{dt}{ds}|_{P_0} = 0
\label{44} \\
\frac{d\bar{x}}{ds}|_{P_0} = \cosh(\chi) \ \ \ \ &,& \ \ \ \
\frac{d\bar{t}}{ds}|_{P_0} = \sinh(\chi)
\label{45}
\end{eqnarray}
The geodesic equations read:
\begin{eqnarray}
\frac{d^2t}{ds^2} &=& - \alpha \frac{\partial \alpha}{\partial t}
(\frac{dx}{ds})^2 , \nonumber \\
\frac{d^2x}{ds^2} &=& - \frac{2}{\alpha}
\frac{\partial \alpha}{\partial t}\frac{dx}{ds} \frac{dt}{ds} -
\frac{1}{\alpha} \frac{\partial \alpha}{\partial x} (\frac{dx}{ds})^2
\label{46}
\end{eqnarray}
\begin{eqnarray}
\frac{d^2\bar{t}}{ds^2} &=& - \bar{\alpha} \frac{\partial
\bar{\alpha}
}{\partial \bar{t}} (\frac{d\bar{x}}{ds})^2 , \nonumber \\
\frac{d^2\bar{x}}{ds^2} &=& - \frac{2}{\bar{\alpha}}
\frac{\partial \bar{\alpha}}{\partial \bar{t}}\frac{d\bar{x}}{ds}
\frac{d\bar{t}}{ds} - \frac{1}{\bar{\alpha}} \frac{\partial
\bar{\alpha}}{\partial \bar{x}} (\frac{d\bar{x}}{ds})^2
\label{47}
\end{eqnarray}
By differentiating these equations as well as using (\ref{35}) we
find that
\begin{equation}
\frac{\partial^2\bar{t}}{\partial x^2}|_{P_0} =
\frac{\partial^2\bar{x}}{\partial x^2}|_{P_0} = 0
\label{48}
\end{equation}
and
\begin{equation}
\frac{\partial^3\bar{t}}{\partial x^3}|_{P_0} = -
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2}|_{P_0} \sinh(\chi)
\cosh^2(\chi) = - \frac{R}{2}
\sinh(\chi) \cosh^2(\chi)
\label{49}
\end{equation}
\begin{equation}
\frac{\partial^3\bar{x}}{\partial x^3}|_{P_0} = - 2
\frac{\partial^2\bar{\alpha}}{\partial \bar{t}^2}|_{P_0} \cosh(\chi)
\sinh^2(\chi) = - R
\cosh(\chi) \sinh^2(\chi).
\label{50}
\end{equation}
Combining these results we obtain that
\begin{equation}
<\bar{0}|:T_{\epsilon}^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}> =
-\frac{R}{48\pi}
\epsilon \exp(\epsilon \chi) \sinh(\chi)
\label{51}
\end{equation}
Adding the results for both values of $\epsilon$ we obtain
\begin{equation}
<\bar{0}|:T^{\hat{0}\hat{0}}:|_{P_0}|\bar{0}> = -\frac{R}{24\pi}
\sinh^2(\chi)
\label{52}
\end{equation}
Inserting this into equation (\ref{25}) we finally obtain the vacuum
expectation
value of the trace of the stress-energy tensor, namely
$\frac{R}{24\pi}$ .
\section{Conclusion}
For the case of a conformally coupled massless scalar field in 1+1
dimensions it is much simpler to evaluate the trace anomaly using a
particle picture than to avoid this. The only regularization
required is
very simple, but it must be this very simple regularization that
suffices
to break the conformal symmetry and thus give a non-zero result for
the
vacuum expectation value of the trace.
\acknowledgments
AZC would like to thank the Theoretical Physics Institute of the
University of Innsbruck, as well as the Max-Planck-Inst. fuer Physik;
Werner Heisenberg Inst. for their hospitality. He also acknowledges
support from the Natural Sciences and Engineering Research Council of
Canada (NSERC) as well as the Alexander von Humboldt Stiftung.
|
\section{Introduction}
Anderson's suggestion\cite{Anderson_1987_1196} that the copper-oxygen
planes of the high-temperature superconductors\cite{Bednorz_1986_189}
are strongly correlated systems has sparked renewed interest in the
two-dimensional Hubbard model. Much of our understanding of the
strong-coupling limit of the model, and the related $t$-$J$ model, has
been obtained by numerical work (reviewed in
Ref.~\onlinecite{Dagotto_1994_763}). Although the single-hole
properties have been studied extensively, exact-diagonalization
studies of small systems are hindered by large finite-size effects in
the parameter region of interest, and Monte-Carlo studies of larger
systems are hindered by the minus-sign problem; other methods have
also been used, but there is still no general agreement on these
properties, particularly for $t/J$ values in the physical region. For
this reason, we have studied the single-hole properties using the
hopping basis of Trugman\cite{Trugman_1988_1597,Trugman_1990_892} and
compared them with results obtained by other methods.
In the limit $U\gg t$, the Hubbard Hamiltonian can be approximated by
the strong-coupling Hamiltonian\cite{Hirsch_1985_1317} $H_{sc} =
H_{t\mbox{\small -}J} + H_3$; this differs from the $t$-$J$
Hamiltonian $H_{t\mbox{\small -}J}$ (which has its own justification)
by the three-site terms in $H_3$:
\begin{eqnarray}
\label{t-J_Hamiltonian_eq}
H_{t\mbox{\small -}J} &=&
-t \sum_{i,\delta,\sigma}
c_{i+\delta,\sigma}^\dagger c_{i,\sigma}^{\phantom\dagger}
+ J \sum_{\langle ij\rangle}({\bf S}_i \cdot {\bf S}_j
- \textstyle {1\over 4} n_i n_j)\ , \\
\label{three_site_term_eq}
H_3 &=& -{{J}\over{4}}\sum_{i,\sigma}\sum_{\delta,\delta^\prime}
(c_{i+\delta,\sigma}^\dagger c_{i,-\sigma}^\dagger
c_{i,-\sigma}^{\phantom\dagger}c_{i+\delta^\prime,\sigma}^{\phantom\dagger}
\nonumber \\
& & \qquad\qquad\ \ -c_{i+\delta,-\sigma}^\dagger c_{i,\sigma}^\dagger
c_{i,-\sigma}^{\phantom\dagger}c_{i+\delta^\prime,\sigma}^{\phantom\dagger})\ ;
\end{eqnarray}
here sites $i+\delta$ and $i+\delta^\prime$ are distinct nearest
neighbors of site $i$, $\langle ij \rangle$ are nearest-neighbor
pairs, and $J = 4t^2/U$. The $t$-$J$ and strong-coupling Hamiltonians
operate in the reduced Hilbert space with no doubly occupied sites;
this restriction is implicit in the above. Validity of the
strong-coupling approximation requires $U\gg t$; the parameter range
believed appropriate to the high-temperature superconductors is
$2<t/J<10$, or $8<U/t<40$. We present results for the $t$-$J$ model in
the region $0.1 \leq t/J \leq 10$ and for the strong-coupling model in
the region $1 \leq t/J \leq 10$.
The single-hole properties in the $t$-$J$ and strong-coupling models
have been studied previously, the first having received more
attention. Methods include exact-diagonalization studies of small
lattices\cite{Kaxiras_1988_656,Dagotto_1989_6721,Dagotto_1990_9049,%
Bonca_1989_7074,Elser_1990_6715,Roder_1991_6284,Fehske_1991_8473,%
Elrick_1993_6004}, studies of infinite lattices using a restricted
basis set\cite{Trugman_1988_1597,Trugman_1990_892,Inoue_1990_2110,%
Inoue_1990_3467,Inoue_1994_6213,Macready_1991_5166}, Monte-Carlo
methods\cite{Boninsegni_1991_10353,Boninsegni_1992_560,Boninsegni_1992_4877,%
Dagotto_1991_8705,Giamarchi_1993_2775,Barnes_1993_11247,Dagotto_1994_728},
and other
methods\cite{Schmitt-Rink_1988_2793,Kane_1989_6880,Marsiglio_1991_10882,%
Martinez_1991_317,Liu_1991_2414,Liu_1992_2425,Liu_1995_3156,%
Sushkov_1992_303,Sushkov_1992_199}. Properties discussed include the
ground-state energy, the bandwidth, the dispersion, the band masses,
the nearest-neighbor spin-spin correlations and the spectral function.
As well, there is an extensive literature on the Hubbard model itself,
including recent finite-temperature Monte-Carlo
results\cite{Bulut_1994_705,Bulut_1994_748,Bulut_1994_7215}.
This paper studies the one-hole properties on an infinite lattice,
using a restricted basis set (in effect a variational method).
Section \ref{hopping_basis} describes the basis, and Sections
\ref{dispersion}-\ref{correlations} give results for the bandwidth,
the dispersion, the band masses, and the nearest-neighbor spin-spin
correlations respectively. For both models, the band minimum is at
${\bf k}=(\pi/2,\pi/2)$ and the maximum at ${\bf k}=(0,0)$ for the
$t/J$ values investigated. The bandwidth is approximately $2J$ at
large $t/J$, in agreement with loop-expansion
results\cite{Marsiglio_1991_10882,Martinez_1991_317,Liu_1995_3156} and
in disagreement with variational Monte-Carlo
results\cite{Boninsegni_1992_4877}. At large $t/J$, the effects of
three-site terms on the bandwidth are well described by first-order
perturbation theory using the $t$-$J$ ground-state wavefunction; that
is, the three-site terms appear to have little effect on the
ground-state wavefunction at large $t/J$. The band mass parallel to
the zone face is much larger than the perpendicular mass. The
spin-spin correlations near the hole are reduced relative to the
starting state, but remain antiferromagnetic.
\section{Hopping Basis}
\label{hopping_basis}
We study a system of $N-1$ electrons on a square lattice of $N$ sites
with periodic boundary conditions; the Hilbert space is restricted to
the $S_z=1/2$ sector with no doubly occupied sites. We use the same
basis for both models, namely the hopping
basis\cite{Trugman_1988_1597,Trugman_1990_892} which has been used
previously\cite{Trugman_1988_1597,Trugman_1990_892,Inoue_1990_2110,%
Inoue_1990_3467,Inoue_1994_6213,Macready_1991_5166}. This method
allows the study of infinite systems (eliminating finite-size
effects), but only certain properties, like the bandwidth and the band
masses, can be studied.
In zeroth order, the basis (denoted $B_0$) consists of a single state
(denoted $|cN\rangle$), the N\'eel state with a missing down-spin
electron. Higher-order bases are generated by repeatedly applying the
$t$ term in the Hamiltonian (which hops the hole to a nearest-neighbor
site). The first-order basis $B_1$ contains the $|cN\rangle$ state
plus the four states generated by hopping the hole. The $n$-th order
basis $B_n$ consists of the states in the basis $B_{n-1}$ plus those
generated by applying the hopping operator to the states in the
difference $B_{n-1}-B_{n-2}$. The basis size (values are given in
Table \ref{basissize_table}) grows exponentially with order. The
hopping basis, which emphasizes states differing from the $|cN\rangle$
state only near the hole, cannot give a good value of the ground-state
energy (because, for example, it does not generate spin interchanges
far from the hole in reasonable order); the expectation is that it
describes well properties like the dispersion and the nearest-neighbor
spin-spin correlations near the hole.
We have used the bases from $B_6$ to $B_{13}$ for most quantities,
going to such large bases because some properties were still changing
significantly; even with basis $B_{13}$ ($\sim2\times10^6$ states),
however, some properties are incompletely converged. Various
extrapolation schemes were considered but judged unreliable, and so we
usually present values for the three largest bases to provide an
estimate of the error due to the truncation of the basis.
The system size ($16\times16$; the lattice constant $a$ is unity) is
effectively infinite since there are no paths which wrap around the
system even in 13-th order. Since the hole moves in an
antiferromagnetic background, the Brillouin zone is reduced to the
square formed by the points $(\pm\pi,0)$ and $(0,\pm\pi)$. The
symmetries of the lattice reduce the independent part of the Brillouin
zone to the triangle with corners at $(0,0)$, $(\pi,0)$, and
$(\pi/2,\pi/2)$, denoted $\bf\Gamma$, $\bf M$, and $\bf S$
respectively. Each state $|n\rangle$ in the basis is a Bloch state,
an eigenstate of the translation operator corresponding to an allowed
value of the momentum. For each basis, and each value of the momentum
{\bf k}, the lowest eigenvalue and eigenvector were found using a
conjugate-gradient method to minimize the function
$\langle\Psi|H|\Psi\rangle/\langle\Psi|\Psi\rangle$ with respect to
the expansion coefficients in $|\Psi\rangle=a_n|n\rangle$; this method
is reported to converge more rapidly than others commonly
used\cite{Nightingale_1993_7696}, but gives the eigenvector to only
single precision. Where necessary, the eigenvector was improved by a
Lanczos method.
The dispersion (in the energy as a function of {\bf k}) results from
several processes. The Trugman
paths\cite{Trugman_1988_1597,Trugman_1990_892} translate the hole to a
next-nearest-neighbor site or a third-nearest-neighbor site on the
same sub-lattice, restoring the original configuration. In the
lowest-order path, the hole hops six times around the smallest square
to a next-nearest-neighbor site; as a result, matrix elements like
$\langle B_2 | c_{i+\delta,\sigma}^\dagger
c_{i,\sigma}^{\phantom\dagger} | B_3\rangle$ are momentum-dependent.
Momentum dependence can also arise from the $J$ term in $H$; for
example, the basis $B_2$ contains states with the hole translated by
$2a$ and a pair of flipped spins, and so matrix elements like $\langle
B_0|{\bf S}_i\cdot{\bf S}_j|B_2\rangle$ depend on {\bf k}. The
results show odd-even effects in the order of the basis; as the basis
size increases, Trugman paths of higher order, and also states
differing from the starting state by nearest-neighbor spin
interchanges, are generated.
Related bases were also studied, in an effort to determine which
states are important for the hole properties. The hopping basis can
be described symbolically as $B_n=\sum_{k=0}^n h^k |cN\rangle$ where
$h$ is the hole hopping operator. We define also operators ${\cal
S}_8$, ${\cal S}_{12}$ and ${\cal S}_{20}$; the first scrambles the 8
spins at distances $a$ and $\sqrt2 a$ from the hole (giving 70 states
when operating on the $|cN\rangle$ state), the second these spins plus
the four at distance $2a$, and the third the 20 spins inside a
$5\times 5$ square minus the four corner sites. If hole properties
like the bandwidth are determined primarily by configurations which
differ from the $|cN\rangle$ state only near the hole, then the bases
$\sum_k h^k{\cal S}_m|cN\rangle$, or (likely better) ${\cal
S}_m\sum_kh^k|cN\rangle$, should converge more rapidly than the
hopping basis; we find the opposite: when the bandwidth is plotted
against the inverse of the log of the basis size, these modified bases
behave like the hopping basis, except that properties are shifted
toward larger basis sizes. We considered also two other bases, both
of which reduce the importance of string states (in which the hole
wanders without looping): (i) the basis $\sum_kM_m h^k|cN\rangle$
where the operator $M_m$ removes states in which the Manhattan
displacement ($|x|+|y|$) of the hole relative to its initial position
is greater than $ma$, and (ii) the basis $\sum_{k=0}^\infty (I_n h)^k
|cN\rangle$, where the operator $I_n$ removes states with more than
$n$ ``bad bonds'' (that is, it filters states according to their Ising
energy relative to the $|cN\rangle$ state; the limit $\infty$ means
that the hop-filter combination is applied until the basis no longer
grows, for given $n$). Neither the Manhattan nor the Ising filters
improved the convergence. We conclude from these numerical
experiments that the single-hole properties are determined not so much
by the spin configurations near the hole as by loop and string paths.
It appears that the hopping basis, whether in its original form or in
the modified forms we have investigated, is capable of only limited
accuracy even if carried to very high order.
\section{Bandwidth}
\label{bandwidth}
Because the lattice is effectively infinite, the lowest energy can be
found for any $\bf k$. For both models, we found $E({\bf k})$ at 81
independent $\bf k$ values of the form $(2\pi n/L, 2\pi m/L)$ with $n$
and $m$ integers and $L = 32$, for $t/J$ values in the range $0.1\leq
t/J\leq10$ for the $t$-$J$ model and in the range $1\leq t/J\leq10$
for the strong-coupling model (for which the lower values of $t/J$ are
of little interest).
For the $t$-$J$ model, the energy is a minimum at ${\bf k}={\bf S}$
(and a maximum at ${\bf\Gamma}$) for $0.1\leq t/J\leq10$, for all
bases used ($B_6$ to $B_{13}$), in agreement with all previous work.
For the strong-coupling model, the energy is also a minimum at ${\bf
k}={\bf S}$ (and a maximum at ${\bf\Gamma}$) for all $t/J$ in the
range $1.0\leq t/J\leq10$, but only for the largest bases at small
$t/J$; this result disagrees with predictions (based on fits to
exact-diagonalization results for small
systems\cite{Fehske_1991_8473}) that the minimum is at ${\bf M}$ for
$t/J \leq 5$. For the smaller bases, particularly for the smaller
values of $t/J$, the minimum can be at ${\bf M}$ or elsewhere along
the zone face; for example, the minimum is at ${\bf S}$ only in 11-th
order and higher for $t/J=1$.
Figure~\ref{bw_ours_figure} plots the bandwidth
$W=E({\bf\Gamma})-E({\bf S})$ for both models as found using the bases
$B_{11}$, $B_{12}$, and $B_{13}$. The convergence is good for the
$t$-$J$ model at all $t/J$ investigated; it is moderately good for the
strong-coupling model at larger $t/J$, but worsens at smaller $t/J$.
The $t$-$J$ bandwidth is approximately $t$ for $t/J<2$ and $2J$ for
$t/J>2$, but decreases weakly at large $t/J$. The strong-coupling
bandwidth is also about $2J$ (though about 20\% larger) and also
decreases as $t/J$ increases. The hopping-basis results are
incompletely converged, however; the bandwidth is still increasing
with basis size, and the trend is greater at larger $t/J$. It is
possible then that the slight decrease which we find is due to the
finite size of the hopping basis.
Figure~\ref{bw_comp_figure} compares our values for the $t$-$J$
bandwidth with those obtained by other methods; major differences
occur in the physical region $t/J>2$. The hopping-basis results agree
best with loop-expansion
results\cite{Marsiglio_1991_10882,Martinez_1991_317,Liu_1995_3156}
and poorly with variational Monte-Carlo
results\cite{Boninsegni_1992_4877} (for unknown reasons); the
$4\times4$ exact-diagonalization results\cite{Elrick_1993_6004} at
large $t/J$ are unreliable due to finite-size effects. Our results at
large $t/J$ are qualitatively consistent with the mean-field
result\cite{Schrieffer_1989_11663} $W\approx4J$ for strong coupling.
From Figure 1, the normalized bandwidth difference
$(W_{sc}-W_{t\mbox{\small -}J})/J$ is almost independent of $t/J$ for
$t/J\agt4$. Since $(H_{sc} - H_{t\mbox{\small -}J})/J=H_3/J$ has no
explicit dependence on $t$ or $J$, this suggests treating the
three-site terms as a perturbation to the $t$-$J$ model.
The error in the first-order result for the bandwidth difference
$\Delta W_1= \Delta E_1({\bf\Gamma}) - \Delta E_1({\bf M})$, where
$\Delta E_1({\bf k}) = \langle\Psi_{t-J}|H_3|\Psi_{t-J}\rangle({\bf k})$,
is less than 2\% at $t/J=10$ and $t/J = 8$, but is much larger at smaller
$t/J$ (52\% at $t/J=$ 4).
Of course the estimate for the strong-coupling bandwidth itself is much
better (errors are 0.3\%, 0.3\%, and 11\% at $t/J=$ 10, 8, and 4).
It appears then that the three-site terms can be treated in first order
for $t/J\agt6$.
Further investigation revealed that the first-order estimates of the
energy at $\bf S$ are excellent; $(\langle H_{sc}
\rangle_{t\mbox{\small -}J} - E_{sc})/W_{sc}$ is $0.1\%$, $0.09\%$,
$0.06\%$ and $0.04\%$ at $t/J=10$, 8, 4, and 1 respectively; the
corresponding values at $\bf\Gamma$ are $0.4\%$, $0.4\%$, $11\%$ and
$41\%$. For unknown reasons, at intermediate $t/J$ values the
three-site terms appear to affect the $\bf\Gamma$ ground state
strongly and the ${\bf M}$ ground state very weakly.
\section{Dispersion}
\label{dispersion}
The Fourier coefficients $a_{lm}$ defined by
\begin{equation}
\label{dispersion_equation}
E({\bf k}) = \sum_{l,m=0}^{L/2} a_{lm} \cos{l k_x} \cos{m k_y}
\end{equation}
are easily obtained by inversion from the energy as a function of
${\bf k}$. The symmetries of the lattice give $a_{lm} = a_{ml}$, and
$a_{lm} = 0$ for $l+m$ odd. The independent coefficients are then the
81 $a_{lm}$ with $0\leq l\leq16$, $0\leq m\leq l$, and $l+m$ even.
The coefficient $a_{00}$ depends strongly on the order of the basis,
as more states important for the ground-state energy are generated; it
affects none of our results since we look only at quantities (like the
dispersion) which depend on energy differences.
Of the other coefficients, $a_{11}$ and $a_{20}$ (both positive) are
the largest, with the ratio $a_{20}/a_{11}$ less than about 0.6 for
both models for the range of $t/J$ values studied. The remaining
coefficients are less than about $0.1a_{11}$ in magnitude for both
models at the $t/J$ values studied. Figures \ref{a11_figure} and
\ref{a20_figure} plot the two leading coefficients as functions of
$t/J$ for the two models. The convergence is of course qualitatively
the same as for the bandwidth, good for the $t$-$J$ model at all $t/J$
and for the strong-coupling model for $t/J\agt4$, but increasingly
poor for the latter with decreasing $t/J$.
At large $t/J$, the values $a_{20}/J$ are almost independent of $t/J$,
whereas the coefficients $a_{11}/J$ decrease with increasing $t/J$.
The strong-coupling coefficients are larger than the $t$-$J$
coefficients, reflecting the enhanced mobility due to the three-site
terms. Also, at larger $t/J$, the difference
$(a_{20}/J)_{sc}-(a_{20}/J)_{t\mbox{\small -}J}$ for the two models is
almost independent of $t/J$, as is the difference in the values of
$a_{11}/J$, for the reason discussed in Section \ref{bandwidth}.
Figures \ref{a11_figure} and \ref{a20_figure} also plot other
results\cite{Marsiglio_1991_10882,Martinez_1991_317} for the $t$-$J$
Fourier coefficients; the agreement is as expected from Section
\ref{bandwidth}. Recent Monte-Carlo
results\cite{Dagotto_1994_728,Giamarchi_1993_2775}, available only at
$t/J = 2.5$, are about 25\% higher than ours.
\section{Band Masses}
\label{bandmass}
The band masses at the band minimum, which is at $\bf S$ for both
models in the region $1\leq t/J\leq10$, are defined in terms of the
second derivatives of $E({\bf k})$ with respect to ${\bf k}$:
\begin{equation}
m_{\mu\nu} = \hbar^2 \left(\frac{\partial^2 E({\bf k})}
{\partial k_\mu\partial k_\nu}\right)^{-1}.
\end{equation}
The masses were obtained by calculating $E({\bf k})$ at additional
points near $\bf S$ and using finite-difference approximations for the
derivatives. Figures~\ref{mper_figure} and \ref{mpar_figure} give
results for the masses perpendicular and parallel to the zone face
respectively, in units of the bare mass $m_0=\hbar^2/2t$. The
parallel mass is much larger than the perpendicular mass, as found
previously\cite{Inoue_1990_2110,Macready_1991_5166,Fehske_1991_8473,%
Martinez_1991_317,Marsiglio_1991_10882,Dagotto_1994_728}.
The perpendicular mass is well converged for both models. For the
$t$-$J$ model, $m_\bot/m_0$ is almost linear in $t/J$ at large $t/J$,
but flattens out at small $t/J$. For the strong-coupling model,
$m_\bot/m_0$ is almost proportional to $t/J$; the smaller effective
mass reflects again the increased hole mobility relative to that in
the $t$-$J$ model.
The parallel mass is much more poorly converged, especially at smaller
$t/J$; even at $t/J=10$ (the most favorable value), the masses change
by over $5\%$ between the bases $B_{12}$ and $B_{13}$. The poor
convergence results because the energies are nearly independent of
${\bf k}$ (the mass is large). For large $t/J$, though, it appears
that $m_\parallel/m_0$ increases only weakly with $t/J$ for both
models and that the two models have the same parallel mass.
Figures~\ref{mper_figure} and \ref{mpar_figure} also give the results
from Ref. \onlinecite{Martinez_1991_317}, derived from their
dispersion results (Table II of Ref. \onlinecite{Martinez_1991_317})
using the free mass $m_0=\hbar^2/2t$, rather than the effective masses
of their Table III. The difference
is due in part to a genuinely different dispersion, but part of it
arises because they used only two components in the Fourier expansion
(the parallel mass, being large, is particularly sensitive to small
changes in the energy).
\section{Spin-spin Correlations}
\label{correlations}
Figures \ref{corr_tj_figure} and \ref{corr_sc_figure} show the
nearest-neighbor spin-spin correlation $\langle {\bf S}_i \cdot {\bf
S}_j \rangle$ for pairs of sites $i$ and $j$ near the hole, for the
$t$-$J$ model and strong-coupling model respectively. The momentum is
${\bf k} = {\bf S}$ (the band minimum), $t/J = 2.5$, and the basis is
$B_{13}$. In the units of $\hbar^2 = 1$ used, the spin-spin
correlation is -0.75 for a singlet pair of spins, $0.25$ for a triplet
pair, and -0.25 for a N\'eel pair. The correlations are
antiferromagnetic, and moderately less than in the starting state.
The ``cigar'' polaron in Figures 7 and 8 is well known from other
studies\cite{Su_1988_9904,Dagotto_1989_6721,%
Elser_1990_6715,Inoue_1990_3467}.
\acknowledgments
This research was supported by the Natural Sciences and Engineering
Research Council of Canada. Computations were done on a Kendall
Square Research 1-32 computer provided by University of Toronto
Instructional and Research Computing; we are grateful to UTIRC staff
for aid in making efficient use of the parallel architecture.
|
\section{Introduction}
{}~~~~Over a decade ago \cite{A}, there began efforts to utilize supersymmetric
models to construct the successor to the standard model. These efforts
received
a further boost with the realization \cite{B} that such theories seem naturally
to
occur as the low-energy limit of four dimensional superstring and heterotic
string theories. A brief survey of the literature would lead one to believe
that there are no unresolved issues in how 4D, N = 1 superfields occur in
this limit. In fact, there are a number of {\it {assumptions}} that are most
often {\underline {not}} {\underline {even}} stated in presentations of the
low-energy action (purportedly {\it {derived}} from superstrings) upon
which most model building is based and phenomenology elucidated. One of
these assumptions is that the spin-0 and spin-1/2 fields that are derived
from the spectrum of string theory necessarily are described by 4D, N = 1
chiral superfields. It is not generally recognized that this is just an
assumption. The reason why this {\underline {is}} an assumption lies in the
fact that there exist little recognized alternative 4D, N = 1 superfields,
the non-minimal scalar multiplet \cite{C} being one, that contain exactly
the same on-shell spectrum as the usual chiral multiplet. We named such
off-shell representations of 4D, N = 1 supersymmetry ``variant
representations.''
Although the non-minimal multiplet has exactly the same on-shell spectrum as
the chiral multiplet, it contains a very different set of auxiliary fields.
As we pointed out previously, the non-minimal scalar multiplet can appear
as an alternate to the usual N = 1 K\" ahler non-linear $\sigma$-models and
as well interact with the usual chiral multiplets \cite{D}. Among these latter
interactions there is a curious result that if a non-minimal scalar multiplet
gains a mass, it can {\underline {only}} do so in tandem with a chiral
multiplet!
In other words, this mechanism provides a natural explanation for the
occurrence
of Dirac spinors within the context of 4D, N = 1 supersymmetric models.
In most discussions of supersymmetric theories, the issue of auxiliary
fields is treated in a cavalier fashion. One would think that there is no
essentially important role played by auxiliary fields.
Nothing could be further from the truth. One reason this attitude prevails
is that there have been few demonstrations of just what dynamical consequences
exist when the off-shell spectrum of two multiplets with the same on-shell
spectrum are compared. A place where such differences can be shown to have
demonstrable consequences is non-linear $\sigma$-models. Similarly differences
can also be observed in higher derivative theories. Typically, what occurs
is that the fields that are usually considered ``auxiliary'' can become
propagating. Under this circumstance, clearly the structure of the auxiliary
fields is important. Higher derivative theories are typically characteristic
of effective field theories. Among these, perhaps the most important is the
low-energy effective Lagrangian ${\cal L}_{eff}$ of QCD. It is known that
leading terms of this theory are described by a chiral SU(3) $\otimes$ SU(3)
non-linear $\sigma$-model. Another term of ${\cal L}_{eff}$ is the
Wess-Zumino-Novikov-Witten term (WZNW) \cite{E}.
Along these lines there has been some discussion of what is the structure of
the 4D, N = 1 supersymmetric extension of the WZNW term \cite{F}. It is the
purpose of this note to show that the introduction of non-minimal scalar
multiplets, to describe some of the spin-0 and spin-1/2 fields in the
effective action, opens an alternate formulation of the 4D, N = 1
supersymmetric
WZNW term. This result highlights the importance of auxiliary fields. Our
result
also provides the most striking evidence to date that the assumption that
{\underline {only}} chiral superfields describe the matter seen in Nature is
incorrect.
\section{Chiral and Non-minimal Multiplet WZNW \newline
Theory}
{}~~~~Almost every researcher who has investigated four dimensional
N = 1 supersymmetry is aware of the chiral scalar or Wess-Zumino multiplet
\cite{G}. The multiplet is described by a chiral superfield $\Phi$ ($
{\Bar D}_{\dot \alpha} \Phi = 0$). The component fields are defined by (we use
{\it {Superspace}} conventions \cite{D})
$$
A ~\equiv~ \Phi \, | ~~~~,~~~~ \psi_{\alpha} ~\equiv~ {D}_{\alpha} \Phi \, | ~~~~,~~~~
F ~\equiv~ {D}^2 \Phi \, | ~~~~,
\eqno(2.1) $$
and appear in the usual linear action as
$$
{\cal S}_{WZ} ~=~ \int d^4 x d^2 \theta d^2 {\Bar \theta} ~ {\Bar \Phi} \, \Phi ~=~
\int d^4 x ~ \Big[ \, - (\pa^{\underline a} {\Bar A} \, ) (
\pa_{\underline a} A \, ) ~-~ i {\Bar \psi}^{\dot \alpha} \pa_{\underline a}
{\psi}^{\alpha} ~+~ {\Bar F} F \, \Big] ~~~~.
\eqno(2.2) $$
The non-minimal scalar multiplet is described by a complex linear superfield
$\Sigma$ (subject to the constraint $ {\Bar D}^2 \Sigma = 0$). The component fields are
defined by
$$ \eqalign{
B ~\equiv~ \Sigma \, | ~~~&,~~~~ {\Bar \zeta}_{\dot \alpha} ~\equiv~ {\Bar D}_{\dot
\alpha} \Sigma \, | ~~~~, \cr
{\rho}_{\alpha} ~\equiv~ {D}_{\alpha} \Sigma \, | ~~~&,~~~~ H ~\equiv~
{D}^2 \Sigma \, | ~~~~,\cr
p_{\underline a} ~\equiv~ {\Bar D}_{\dot \alpha} D_{\alpha} \Sigma \, | ~~~&,~~~~
{\Bar \beta}_{\dot \alpha} ~\equiv~ \frac 12 D^{\alpha} {\Bar D}_{\dot \alpha} D_{\alpha} \Sigma
\, | ~~~~, }
\eqno(2.3) $$
and appear in the usual linear action as
$$ \eqalign{
{\cal S}_{NM} &=~ -~ \int d^4 x d^2 \theta d^2 {\Bar \theta} ~ {\Bar \Sigma} \, \Sigma \cr
&=~ \int d^4 x ~ \Big[ \, - (\pa^{\underline a} {\Bar B} \, ) (
\pa_{\underline a} B \, ) ~-~ i {\Bar \zeta}^{\dot \alpha} \pa_{\underline a}
{\zeta}^{\alpha} ~-~ {\Bar H} H \cr
&{~~~~~~~~~~~~~~~}\, ~+~ 2 \, {\Bar p}^{\underline a} p_{\underline a} ~+~
{\beta}^{\alpha} {\rho}_{ \alpha} ~+~ {\Bar \beta}^{\dot \alpha} {\Bar \rho}_{\dot \alpha}
\, \Big] ~~~~. }
\eqno(2.4) $$
As is apparent from the last result above, only $B$ and ${\zeta}^{\alpha}$ are the
propagating fields among the off-shell 12 + 12 (bosons + fermions) degrees of
freedom of the non-minimal scalar multiplet.
At this point we recall for the reader results in 2D, N = 2 superfield theory
\cite{G1}. Within this class of theories, it is known that there are two
{\it {distinct}} minimal scalar multiplets, chiral multiplets and twisted
chiral multiplets \cite{H}. The superfield form of the linear kinetic term
for the twisted chiral multiplet has a minus sign in comparison to that of
the chiral multiplet. We see exactly the same behavior above for the 4D
chiral and non-minimal superfield actions. In a 2D, N = 2 non-linear
$\sigma$-model theory with manifest supersymmetry, the only known way to
introduce torsion requires the simultaneous presence of both chiral and
twisted chiral superfields. In 4D, the analog of the 2D torsion term is
provided by the WZNW term. Thus, it is natural to suggest that we should
be able to introduce a 4D, N = 1 supersymmetric WZNW term by utilizing
both chiral and non-minimal multiplets.
The starting point in the implementation of this proposal is to note that
the condition that $\Sigma$ is a complex linear superfield (i.e. ${\Bar D}^2 \Sigma
= 0$) necessarily implies that the quantity ${\Bar D}_{\dot \alpha} \Sigma $ is a
chiral superfield and can therefore lead to a supersymmetric invariant in
an F-term! So we introduce a number of chiral superfields $\Phi^{\rm I}$
along with an equal number of non-minimal scalar superfields $\Sigma^{\rm I}$
where ${\rm I} = 1,..., m$. We also require the existence of a fourth order
tensor that is a function of the chiral superfields. We denote this tensor
by ${\cal J}_{\rm I \, J \, K \, L} (\Phi)$. It follows that the term
below is a supersymmetric invariant
$$
{\cal S}_{WZNW} ~=~ \frac 14 \, \int d^4 x \, d^2 \theta ~ {\cal J}_{\rm I \,
J \, K \, L} (\Phi) ({\Bar D}^{\dot \alpha} \Sigma^{\rm I} \, ) \, ({\Bar D}^{\dot
\beta} \Sigma^{\rm J} \, ) \, (\pa^{\gamma} {}_{\dot \alpha} \Phi^{\rm K} \,) \, ( \pa_{\gamma
\dot \beta} \Phi^{\rm L} \,) ~+~ {\rm h.}{\rm c.} ~~~~.
\eqno(2.5) $$
Let us note that the most general non-linear $\sigma$-model term involving these
superfields takes the form,
$$
{\cal S}_{\sigma} ~=~ \int d^4 x d^2 \theta d^2 {\Bar \theta} ~ {\Hat \Omega} (\, \Phi, {\Bar
\Phi} ; \Sigma, {\Bar \Sigma} \, ) ~~~~.
\eqno(2.6) $$
The function ${\Hat \Omega}$ is similar to a K\" ahler potential. However, as
shown in the the latter work of \cite{D}, the metric for the space for which
$\Sigma^{\rm I}$ provides coordinates is not of the form of a K\" ahler metric.
In fact, the metric for the $\Sigma^{\rm I}$-space is {\underline {not}} even of
hermitian form in general. We thus have a counter-example to the well known
folklore that 4D, N = 1 supersymmetric non-linear $\sigma$-models necessarily
describe K\" ahler manifolds. (The global description of the complex space
described solely by $\Sigma$-coordinates has never been given.) Note that one
special choice\footnote{The choice of this function is ultimately done to
produce the best phenomenological fit.} of the function ${\Hat \Omega}$ is given
by a fibration in which the $\Sigma$-coordinates are fibers over a space with
the $\Phi$-coordinates. Such a space is described by
$$
{\Hat \Omega} ~=~ \frac 12 \, [ \, g_{\rm {I\, J}}(\Phi) ~+~ {\Bar g}_{\rm {J\, I}}
({\Bar \Phi}) \,] \, [~ {\Bar \Phi}^{\rm I} {\Phi}^{\rm J} ~-~
{\Bar \Sigma}^{\rm I} {\Sigma}^{\rm J} ~] ~~~~,
\eqno(2.7) $$
in terms of a holomorphic function $g_{\rm {I\, J}}(\Phi)$ (for one choice
of this function see appendix A). In the limit where we set the non-minimal
multiplets to zero, we see that the chiral superfields have a special
K\" ahler geometry\footnote{The first appearance of this type of geometry
is in \cite{G1}.}. The limiting K\" ahler potential $K(\Phi, {\Bar \Phi})$
can be written in the form $K(\Phi, {\Bar \Phi}) = \frac 12 [ {\Bar \Phi}^{
\rm I} g_{\rm {I\, J}}(\Phi) {\Phi}^{\rm J} + {\Phi}^{\rm J} {\Bar g}_{\rm
{I\, J}} ({\Bar \Phi}) {\Bar \Phi}^{\rm I}] $ in order to make the special
K\" ahler geometry for the chiral superfields manifest. Defining ${\Tilde
\Phi}_{\rm I} \equiv g_{\rm {I\, J}}(\Phi) {\Phi}^{\rm J}$, we can re-write
$K(\Phi,
{\Bar \Phi})$ in the form $K(\Phi, {\Bar \Phi}) = \frac 12 [{\Bar \Phi}^{\rm
I} {\Tilde \Phi}_{\rm I} + {\Phi}^{\rm I} {\Tilde {\Bar \Phi}}_{\rm I}]$ in
order to make contact with the recent work on exact results for N = 2
supersymmetric Yang-Mills effective actions \cite{I}. This form also makes
obvious the presence of the duality pairs $({\Phi}^{\rm I} , \, {\Tilde
\Phi}_{\rm I})$ that are related by elliptic curves.
Let us offer another interpretation of (2.5), (2.7) and (2.8) below. Within
the confines of 2D, N = 2 superconformal field theory, there have been found
to exist (c,c) rings and (a,c) rings. The former correspond to functions of
chiral multiplets while the latter correspond to twisted chiral multiplets.
The interesting point is that (a,c) rings were discovered much later than
(c,c) rings. This discovery of these distinct supersymmetry representations
in the spectrum of the theory occurred even though they were not put in as
elementary representations on the 2D world sheet. This example shows
us that non-perturbatively supersymmetric systems are able to generate
states that are distinct supersymmetry representations from the elementary
states. On the other hand, if (a,c) rings are included at the elementary level,
then 2D, N = 2 superconformal field theories can possess an additional
symmetry, i.e. mirror symmetry. This suggests that the non-minimal scalar
multiplet may be generated non-perturbatively in 4D and if they are included
in the underlying supersymmetric renormalizable QCD theory, it may possess a
larger symmetry group.
Thus, we should be able to embed the QCD low-energy effective
action into a supersymmetric action of the form
$$
{\cal S}_{eff} ~=~ {\cal S}_{\sigma} ~+~ {\cal S}_{WZNW} ~~~~.
\eqno(2.8) $$
In the next section we will look at the component formulation that follows
from the proposal above. However, in closing this section, we note that our
proposed description of the 4D, N = 1 supersymmetric QCD low-energy effective
action with WZNW term is the {\underline {first}} that is consistent with
holomorphy \cite{I}, the concept that holomorphic functions determine
the effective action. In fact, we gave the {\underline {first}} demonstration
\cite{G1} that the 4D, N = 2 supersymmetric Yang-Mills action is classically
determined by holomorphic functions. Recently, major advances have occurred
in understanding the 4D, N = 2 supersymmetric Yang-Mills effective action
due to the presence of holomorphy and proposals have been made that it
should play a role in increasing our understanding of the 4D, N = 1
supersymmetric Yang-Mills effective action.
\section{Embedding ${\cal L}_{eff} (QCD) $ in a 4D, N = 1 \newline
${}$ Supersymmetric Theory}
{}~~~~The calculation of the component results follows using the by now
well established projection technique. We find ${\cal S}_{WZNW}$ leads to
$$ \eqalign{
& \frac 14 \, \int d^4 x d^2 \theta ~ {\cal J}_{\rm I \, J \, K \, L} (
\Phi) ({\Bar D}^{\dot \alpha} \Sigma^{\rm I} \, ) ({\Bar D}^{\dot \beta} \Sigma^{\rm J} \,
) (\pa^{\gamma}{}_{\dot \alpha} \Phi^{\rm K} \,) ( \pa_{\gamma \dot \beta} \Phi^{\rm L} \,)
{~~~~~~~~~~} \cr
&=~ \frac 14 \, \int d^4 x \Big[ \, - {\cal J}_{\rm I \, J \, K \, L} (A)
\, ( i 2 \pa^{\alpha \dot \alpha} B^{\rm I} ~-~ p^{\alpha \dot \alpha ~ \rm I} \, )
( i 2 \pa_{\alpha}{}^{\dot \beta} B^{\rm J} ~-~ p_{\alpha} {}^{\dot \beta ~ \rm J} \, )
(\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,) ( \pa_{\gamma \dot \beta} A^{\rm L} \,) \cr
& {~~~~~~~~~~~~~~~} +~ 2 {\cal J}_{\rm I \, J \, K \, L} (A)\, ( i \pa_{\alpha}
{}^{\dot \alpha} \rho^{\alpha \, {\rm I}} ~-~ \beta^{\dot \alpha \, {\rm I}} \,)
\,{\Bar \zeta}^{\dot \beta \, {\rm J}} (\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,)
( \pa_{\gamma \dot \beta} A^{\rm L} \,) \cr
& {~~~~~~~~~~~~~~~} +~ {\cal J}_{\rm I \, J \, K \, L} (A) \,{\Bar
\zeta}^{\dot
\alpha \, {\rm I}}{\Bar \zeta}^{\dot \beta \, {\rm J}} [\, (\pa^{\gamma}{}_{\dot \alpha}
\psi^{\alpha
\rm K} \,) ( \pa_{\gamma \dot \beta} \psi_{\alpha}{}^{\rm L} \,) ~+~ 2
(\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,) ( \pa_{\gamma \dot \beta} F^{\rm L}
\,) \,] \cr
& {~~~~~~~~~~~~~~~} +~ 4 {\cal J}_{\rm I \, J \, K \, L} (A) \,
( i 2 \pa^{\alpha \dot \alpha} B^{\rm I} ~-~ p^{\alpha \dot \alpha ~ \rm I} \, ) \,
{\Bar \zeta}^{\dot \beta \, {\rm J}} (\pa^{\gamma}{}_{\dot \alpha} \psi_{\alpha}^{
\rm K} \,) ( \pa_{\gamma \dot \beta} A^{\rm L} \,) \cr
& {~~~~~~~~~~~~~~~} +~ 2 {\cal J}_{\rm I \, J \, K \, L \, , {\rm M}} (A) \,
\psi^{\alpha \rm M} \, {\Bar \zeta}^{\dot \alpha \, {\rm I}}{\Bar \zeta}^{\dot \beta \,
{\rm J}} (\pa^{\gamma}{}_{\dot \alpha} \psi_{\alpha}^{\rm K} \,) ( \pa_{\gamma \dot \beta}
A^{\rm L} \,) \cr
& {~~~~~~~~~~~~~~~} -~ 2 {\cal J}_{\rm I \, J \, K \, L \, , {\rm M}} (A) \,
\psi^{\alpha \rm M} ( i 2 \pa^{\alpha \dot \alpha} B^{\rm I} ~-~ p^{\alpha \dot \alpha ~ \rm I}
\, ) \, {\Bar \zeta}^{\dot \beta \, {\rm J}} (\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,)
( \pa_{\gamma \dot \beta} A^{\rm L} \,) \cr
& {~~~~~~~~~~~~~~~} +~ {\cal J}_{\rm I \, J \, K \, L \, , {\rm M}} (A) \,
F^{\rm M} \,{\Bar \zeta}^{\dot \alpha \, {\rm I}}{\Bar \zeta}^{\dot \beta \,
{\rm J}} (\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,) ( \pa_{\gamma \dot \beta} A^{\rm L}
\,) ~ \Big]
{}~~~~. }
\eqno(3.1) $$
As can be seen, only the first line of the rhs consists of purely bosonic
terms. Let us focus our analysis by only considering these terms.
It is our first observation that if we set the auxiliary field $p_{
\underline a}$ to zero, then the purely bosonic terms collapse to
$$ \eqalign{ \frac 14 \,
\int d^4 x \, d^2 \theta ~ &{\cal J}_{\rm I \, J \, K \, L} (\Phi) \,
({\Bar D}^{\dot \alpha} \Sigma^{\rm I} \, ) \, ({\Bar D}^{\dot \beta} \Sigma^{\rm J} \, ) \,
(\pa^{\gamma} {}_{\dot \alpha} \Phi^{\rm K} \,) \, ( \pa_{\gamma \dot \beta} \Phi^{\rm L}
\,) |_{phys.\, fields}{~~~~~~~~~~} \cr
&=~ \int d^4 x \Big[ \, {\cal J}_{\rm I \, J \, K \, L} (A) \, (
\pa^{\alpha \dot \alpha} B^{\rm I} \, ) \, ( \pa_{\alpha}{}^{\dot \beta} B^{\rm J} \, ) \,
(\pa^{\gamma}{}_{\dot \alpha} A^{\rm K} \,) \, ( \pa_{\gamma \dot \beta} A^{\rm L} \,) \,
\Big] \cr
&=~ \int d^4 x \Big[ \, {\cal J}_{\rm I \, J \, K \, L} (A) \,
{\rm P}^{\underline a \underline b \underline c \underline d} \,
( \pa_{\underline a} B^{\rm I} \, ) \, ( \pa_{\underline b} B^{\rm J} \, )
\, (\pa_{\underline c} A^{\rm K} \,) \, ( \pa_{\underline d} A^{\rm L} \,)
\, \Big] ~~~~. }
\eqno(3.2) $$
where ${\rm P}^{\underline a \underline b \underline c \underline d}
\equiv [ \eta^{\underline a [ \underline c} \eta^{\underline d ]
\underline b} ~+~ i \epsilon^{\underline a \underline b \underline c
\underline d} ] $. Up until this point, we have not made any assumption
regarding the explicit form of ${\cal J}_{\rm I \, J \, K \, L} (A)$. We
could easily choose it to be the (4,0) form that is defined in the
non-supersymmetric component WZNW action (see appendix A). However, (3.2)
has the consequence that it can describe both the WZNW term as well as
the Skyrme term. In the following we we simply concentrate on the WZNW
term and thus we choose ${\cal J}_{\rm I \, J \, K \, L} (A)$ to be
define by (A.6)\footnote{A few minor modifications are required in the
supersymmetric case.}. Since $p_{\underline a}$ actually has a more
complicated equation of motion that depends on the leading term of the
effective action, its elimination will produce other higher order
interactions. However, their presence does not disturb our present results.
These and a number of other details will be discussed in a future work.
Now we want the component pion fields that are contained in our QCD
superfield WZNW term of (2.8) to agree precisely the non-supersymmetric
QCD effective action (see (A.7)). This will be the case if the following
identifications are made,
$$\Phi | ~=~ {\cal A} (x) ~+~ i \, [ \, \Pi (x) ~+~ \Theta (x) \,] ~~~,~~~
\Sigma | ~=~ {\cal B} (x) ~+~ i \, [ \, \Pi (x) ~-~ \Theta (x) \,] ~~~.
\eqno(3.3) $$
where $\Pi (x)$ is the pion octet. Thus, we see that the pion superfield
is a linear mixture of chiral and complex linear superfields.
This is analogous to the fact that a Dirac field in a supersymmetric theory
can only occur as a linear combination of basic superfields. We are thus
motivated to define the super-pion superfield by
$$ \Pi ~\equiv~ -i \frac 14 \Big[ \, \Phi ~+~ \Sigma ~-~
{\Bar \Phi} ~-~ {\Bar \Sigma} \, \Big] ~~~~.
\eqno(3.4) $$
By the same token we see that in a manifestly supersymmetric world, in
addition to the super-pion, there are mirror super-pions defined by
$$ \Theta ~\equiv~ -i \frac 14 \Big[ \, \Phi ~-~ \Sigma ~-~
{\Bar \Phi} ~+~ {\Bar \Sigma} \, \Big] ~~~~.
\eqno(3.5) $$
There are also parity doubles of these fields that are most conveniently
defined by
$$ {\cal A} ~\equiv~ \frac 12 \Big[ \, \Phi ~+~ {\Bar \Phi} \, \Big] ~~~~,
{}~~~~ {\cal B} ~\equiv~ \frac 12 \Big[ \, \Sigma ~+~ {\Bar \Sigma} \, \Big] ~~~~.
\eqno(3.6) $$
Similarly, applying various spinor derivatives to these superfields
produce the spin-1/2 pionino SU(3) multiplet and their parity doubles.
Here we have some ambiquity. We have enough spinor components to form
a Dirac pionino SU(3) multiplet or two Majorana pionino SU(3) multiplets.
In the former case, the pionini are isomorphic to the baryon octet that
contains the proton!
The leading term in (2.8) will also contain exactly the leading term
of (A.7) if we identify the function $g_{\rm I \, J}$\footnote{Once again a
few minor modifications are required in the supersymmetric case.} that appears
in (2.7) with that defined in (A.4). Thus, we find that there is a very simple
embedding of ${\cal L}_{eff} (QCD)$ into our superfield theory.
\section{ Conclusion}
{}~~~~At this point, it is useful to compare our new suggestion for a 4D,
N = 1 supersymmetric extension of the WZNW terms to those that exist in
the prior literature. The relevant work occurred in reference \cite{F}.
There it was proposed that the 4D, N = 1 supersymmetric extension of the
WZNW term is of the form
$${\cal S}_{WZNW} ~=~ \int d^4 x d^2 \theta d^2 {\Bar \theta} ~ \Big[ \, \beta_{\rm
{ I \,J {\bar {\rm K}}}} ( D^{\alpha} \Phi^{\rm I}\,) ( \pa_{\alpha \dot \beta} \Phi^{
\rm J} \,) ( {\Bar D}^{\dot \beta} {\Bar \Phi}^{\bar {\rm K}} \,) ~+~ {\rm h.}
{\rm c.} ~\Big] ~~~~.
\eqno(4.1) $$
If we compare our results to the older ones, several features are apparent.
Foremost, the previous result utilizes an action that is integrated over the
full superspace. (This means for example that all of the chiral superfields
contained in (4.1) could be replaced by complex linear superfields and we
would then obtain another WZNW-type term.) In particular, the quantity
$\beta_{\rm { I \,J {\bar {\rm K}}}}$ is {\underline {not}} holomorphic.
Our choice need only be integrated over a chiral subspace due to its chirality
(i.e. holomorphicity). At the level of component fields, the differences
are simply tremendous! Our suggestion contains many fewer terms. At most four
fermion but not six fermion terms appear in our construction in contrast to
(4.1). Finally, there are terms in (4.1) that are quartic in temporal
derivatives of bosonic fields. In our proposal no such terms of this high
order in temporal derivatives appear. This last point is rather telling.
It is certainly true that the non-supersymmetric WZNW terms contains no more
than first order temporal derivatives.
Ordinary 4D, N = 1 chiral and non-minimal multiplets possess an uncanny
resemblance to 2D, N = 2 chiral and twisted chiral multiplets. This
naturally raises the question of whether there might exist some 4D, N = 1
analog to mirror symmetry. We could formally define a 4D mirror operator
that sends chiral multiplets into non-minimal multiplets and vice-versa.
There are important differences, however. Off-shell chiral and non-minimal
multiplets do not possess the same number of degrees of freedom. So there
are some issues that require additional study. Finally, we believe that
our result regarding the simple embedding of ${\cal L}_{eff} (QCD)$ should
act as a warning that the sole use of chiral multiplets to describe
matter is not always wise. We re-emphasize the cautionary note we made
along these lines previously in the second work of
\cite{D}.
The problem our presentation demonstrates has its ultimate cause
in our lack of mastery of string theory. As
presently formulated, we simply do not possess a direct (i.e.
without making any assumptions) way to derive from string theory the
off-shell superfields that presumably emerge in its low-energy
limit. For some time, we have believed that it is quite likely
that non-minimal scalar multiplets must be involved in this limit.
Our reason for this belief is that it appears likely that the
4D, N = 2 low-energy limit of string theory contains at least some
non-minimal scalar multiplets! The only known off-shell formulation
of 4D, N = 2 hypermultiplets \cite{ST} contains 4D, N = 1
non-minimal scalar multiplets. Finally, it is interesting to ponder
further WZNW extensions to 4D, N = 2
supersymmetry. The recent advances \cite{I} are silent on
the 4D, N = 2 WZNW term. Here we would like to know
if the {\it {two}} distinct 4D, N = 2 hypermultiplets (\cite{ST} and
\cite{FY}) play roles analogous to that of the 4D, N = 1 chiral
and non-minmal multiplets in the 4D, N = 1 WZNW term.
This latest result together with the ``natural Dirac mass'' associated
with a pairing of a chiral superfield together with a complex linear
superfield (i.e. $(\Phi^{\rm I} , \, \Sigma^{\rm I} \, )$) seems to be
hinting that there is something truly fundamental but not understood
occurring. As we noted previously, the current generation of
supersymmetric phenomenological models totally ignores the possibility
that ordinary matter may contain such pairings. We can well imagine
scenarios in which one chiral
part of a Dirac particle is assigned to chiral superfields and the
other chiral part of the same Dirac particle is assigned to complex
linear superfields. This might well serve as an intrinsic reason
why chiral asymmetry occurs in supersymmetric extensions of the
standard model and as well could easily provide the long sought
use of supersymmetry to protect the vanishing masses of neutrini.
Indeed, if supersymmetry is observed in Nature this could make an
attractive explanation for why handedness matters in our universe!
$$\eqalign{ ~~ &{~} \cr &{~} \cr &{~} \cr &{~} \cr
&{~} \cr &{~} \cr &{~} \cr &{~} \cr &{~} \cr &{~} \cr } $$
\noindent
{\bf {Acknowledgment; }} \newline \noindent
I wish to thank Ms. Lubna Rana for useful discussions.
\newpage \noindent
{\bf {Appendix A: Brief Review of ${\cal L}_{eff}(QCD)$ }}
In this very brief appendix we simply gather together the basic
facts concerning the low-energy effective action for QCD. We begin
with a definition of the SU(3) pion octet
$$ \frac 1{f_{\pi}} \Pi ~\equiv~
\frac 1{ f_{\pi}} \Pi^i \lambda_i ~=~ \frac 1{ f_{\pi}} \left(\begin{array}{ccc}
{}~\frac{\pi^0}{\sqrt 2} ~+~ \frac{\eta}{\sqrt 6} & ~~\pi^+ & ~~K^+ \\
{}~\pi^- & ~~-\, \frac{\pi^0}{\sqrt 2} ~+~ \frac{\eta}{\sqrt 6} & ~~K^0\\
{}~K^- & ~~{\Bar K}^0 & ~~ - \eta \sqrt {\frac 23} \\
\end{array}\right) ~~~~.
\eqno(A.1) $$
Here $\lambda_1 , \,..., \lambda_8$ are the Gell-Mann SU(3) matrices. Further $f_{\pi
}$ is the weak pion coupling constant\footnote{It should be noted that we
differ from Witten's convention of this parameter by a factor of two.}
with the dimensions of mass. Group elements are formed by writing $U(\Pi)
= \exp [\, i {f_{\pi}}^{-1}\Pi \,]$. We define left ($ L_m {}^i (\Pi)$)
and right ($ R_m {}^i (\Pi)$) Maurer-Cartan forms by the equations
$$
U^{-1} \pa_{\underline a} U ~=~ i {f_{\pi}}^{-1} ( \, \pa_{\underline a} \Pi^m
\,)~ L_m {}^i (\Pi) \, \lambda_i ~~~~,~~~~ (\, \pa_{\underline a} U \,) U^{-1} ~=~
i {f_{\pi}}^{-1} ( \, \pa_{\underline a} \Pi^m \,) ~R_m {}^i
(\Pi) \, \lambda_i ~~~~.
\eqno(A.2) $$
These definitions allow $L_m {}^i (\Pi)$ and $R_m {}^i (\Pi)$ to be calculated
as power series in $\Pi^i$ from \cite{ACG}
$$\eqalign{
L_m {}^i (\Pi) &\equiv ~ (C_2)^{-1} {\rm {Tr}} \Big[\, T^i \Big(
\frac { 1 ~-~ e^{-\Delta} }{\Delta} \Big) T_m \Big] ~~~, \cr
R_m {}^i (\Pi) &\equiv ~ (C_2)^{-1} {\rm {Tr}} \Big[\, T^i \Big(
\frac { e^{\Delta} ~-~ 1 }{\Delta} \Big) T_m \Big] ~~~, }
\eqno(A.3) $$
where $\Delta T_m \equiv i {f_{\pi}}^{-1} [ \Pi \, , \, T_m ]$, $\Delta^2 T_m = \Delta \Delta
T_m$, etc. and the constant $C_2$ is determined so that $ L_m {}^i (0) = R_m
{}^i
(0) = \delta_m {}^i$. As a consequence we see
$$ {\cal S}_{\sigma} ~=~ \frac {f_{\pi}^2}{2 C_2} \, \int d^4 x ~
{\rm {Tr}} [\, ( \pa^{\underline a} U \,) ~ (\pa_{\underline a} U ^{-1} \,) \,]
{}~=~ \frac 12 \, \int d^4 x ~ \, g_{m \, n} \, (\Pi) ~ ( \pa^{\underline
a} \Pi^m \,) ~ (\pa_{\underline a} \Pi^n \,) ~~~,
\eqno(A.4) $$
where $g_{m \, n} = \delta_{i \, j} \, L_m {}^i \, L_n {}^j = \delta_{i \, j} \, R_m
{}^i \, R_n {}^j$.
The remaining well known term in the QCD effective action is described by the
WZNW term. We follow Witten \cite{E} who, using the Vainberg technique
\cite{V},
showed that with an appropriate normalization this term possesses an
integer quantized coefficient, $N_C$. Using a real parameter $y$ that
takes on values between 0 and 1, we define an extended group element $\Hat U$
through the relation $\Hat U = \exp [\, i y f_{\pi}^{-1} \Pi \,]$.
In terms of the extended group element, the WZNW term is given by
$$ \eqalign{ {~~~~~~~~}
{\cal S}_{WZNW} &=~ - i N_C \, [ \, 2 {~}_{\dot {~}} 5! \, ]^{-1}
\int d^4 x \, \int_0^1 d y ~ {\rm {Tr}} \Big[ \, ( {\Hat U}^{-1} \pa_y {\Hat U}
\,) ~ {\Hat {\cal W}}_4 \, \Big] ~~~~, \cr
{\Hat {\cal W}}_4 &=~ \epsilon^{{\underline a}{\underline b}{\underline c}{\underline
d}} \, (\pa_{\underline a} {\Hat U} ^{-1} \,) \, (\pa_{\underline b} {\Hat U}
\,) \, (\pa_{\underline c} {\Hat U}^{-1} \,) \, (\pa_{\underline d} {\Hat U}
\,) \, ~~~~. } \eqno(A.5) $$
or more directly using the elements of the pion octet this just becomes
$$ \eqalign{ {~~~~}
{\cal S}_{WZNW} &=~ \int d^4 x \, \epsilon^{{\underline a}{\underline b}{\underline
c}
{\underline d}}{\cal J}_{m \, n \, r \, s} (\Pi)
(\pa_{\underline a} \Pi^m \,) \, (\pa_{\underline b} \Pi^n \,) \,
(\pa_{\underline c} \Pi^r \,) \, (\pa_{\underline d} \Pi^s \,) ~~~~, \cr
{\cal J}_{m \, n \, r \, s} (\Pi) &=~ - \, {N_C} [ \, 8 {~}_{\dot {~}} 5!
f_{\pi}^5 \, ]^{-1} f_{a \, b}{}^k \, f_{c \, d}{}^l {\rm {Tr}} \Big[ \,
\lambda_k \lambda_l \lambda_h \, \Big] \, \int_0^1 d y ~ y^4 \, \Pi^e {\Hat L}_e {}^h
{\Hat L}_m {}^a {\Hat L}_n {}^b {\Hat L}_r {}^c {\Hat L}_s {}^d ~~~~. }
\eqno(A.6) $$
where ${\Hat L}_m {}^i \equiv L_m {}^i (y \Pi)$. Also $f_{a \, b}{}^k$ denotes
the structure constants of the group defined by $ [\lambda_a , \lambda_b ] = i f_{a \, b}
{}^k \lambda_k $. The effective QCD Lagrangian is simply given
$$ {\cal S}_{eff} ~=~ {\cal S}_{\sigma} ~+~ {\cal S}_{WZNW}
\eqno(A.7) $$
with $ {\cal S}_{\sigma}$ defined in (A.3) and $ {\cal S}_{WZNW}$ defined in (A.6).
\noindent
{\bf {Appendix B: Manifest Supersymmetric Formulation of
Kazama-Suzuki \newline ${~~~~~~~~~~~~~~~~~}$ Models and New
(2,2) Superstrings}}
In heterotic string theory, one of the well known N = 2 compactification
techniques is give by Kazama-Suzuki models \cite{KS}. An erstwhile
mystery has been, ``How does one find a superfield formulation of
Kazama-Suzuki models?'' Up until now no one has been able to provide
an answer. We now wish to suggest that the missing ingredient seems
to have been the use of the non-minimal scalar multiplet reduced from
4D, N = 1 superspace down to 2D, N = 2 superspace. The reduction itself
is trivial if we introduce the 2D, N = 2 supercovariant derivatives
$D_{\alpha}$ and their conjugates ${\Bar D}_{\alpha}$ which satisfy
$$
\Big[ \, D_{\alpha} , ~ D_{\beta} \, \Big\} ~=~ 0 ~~~,~~~
\Big[ \, {\Bar D}_{\alpha} , ~ {\Bar D}_{\beta} \, \Big\} ~=~ 0 ~~~,~~~
\Big[ \, {D}_{\alpha} , ~ {\Bar D}_{\beta} \, \Big\} ~=~ i (\gamma^c)_{\alpha \beta}
\pa_c ~~~~.
\eqno(B.1) $$
The 2D, N = 2 non-minimal multiplet is now defined by ${\Bar D}^{\alpha}
{\Bar D}_{\alpha} \Sigma = 0$. The component fields are defined with a few very
slight modifications (below we use the chiral components)
$$
\eqalign{ {~~~~}
B &\equiv~ \Sigma \, | ~~~,~~~ {\Bar \zeta}_{ \pm} ~\equiv~ {\Bar D}_{\pm} \Sigma
\, | ~~~, ~~~
{\rho}_{\pm} ~\equiv~ {D}_{\pm} \Sigma \, | ~~~,~~~ H ~\equiv~ - i
\, {D}_+ {D}_- \Sigma \, | ~~~,\cr
u &\equiv~ - i \, {\Bar D}_+ D_- \Sigma \, | ~~~,~~~
v ~\equiv~ - i \,{\Bar D}_- D_+ \Sigma \, | ~~~, ~~~
p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} ~\equiv~ - i \, {\Bar D}_+ D_+ \Sigma \, | ~~~,~~~\cr
p_{{\scriptsize{\dvm}}~~} &\equiv~ - i \, {\Bar D}_- D_- \Sigma \, | ~~~, ~~~
{\Bar \beta}_{\pm} ~\equiv~ - i \, D_+ {\Bar D}_{\pm} D_- \Sigma
\, | ~~~~. }
\eqno(B.2) $$
The complex quantities $u$ and $v$ are the extra components arising
from the dimensional reduction of $p_{\underline a}$ from 4D.
The 2D supersymmetry variations take the forms
$$
\eqalign{ {~~~~~~~~~~~~~~~~}
\delta_Q \, B &=~ {\Bar \epsilon}^+ {\Bar \zeta}_+ ~+~ {\Bar \epsilon}^- {\Bar \zeta}_-
{}~+~ \epsilon^+ \rho_+ ~+~ \epsilon^- \rho_- ~~~~, \cr
\delta_Q \, {\Bar \zeta}_+ &=~ i\, \epsilon^+ (\, \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} B ~-~ p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }}
\,) ~-~ i \ \epsilon^- u ~~~~, \cr
\delta_Q \, {\Bar \zeta}_- &=~ ~-~ i \, \epsilon^+ v ~+~ i\, \epsilon^- (\, \pa_{{\scriptsize{\dvm}}~~}
B ~-~ p_{{\scriptsize{\dvm}}~~} \,) ~~~~, \cr
\delta_Q \, \rho_+ &=~ - i \, \epsilon^- H ~+~ i\, {\Bar \epsilon}^+ p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }}
{}~+~ i\, {\Bar \epsilon}^- v ~~~~, \cr
\delta_Q \, \rho_- &=~ i \, \epsilon^+ H ~+~ i\, {\Bar \epsilon}^+ u
{}~+~ i\, {\Bar \epsilon}^- p_{{\scriptsize{\dvm}}~~} ~~~~, \cr
\delta_Q \, u &=~ \epsilon^+ \beta_+ ~-~ {\Bar \epsilon}^- \pa_{{\scriptsize{\dvm}}~~} {\Bar \zeta}_+ ~~~~, \cr
\delta_Q \, v &=~ \epsilon^- (\, \pa_{{\scriptsize{\dvm}}~~} \rho_+ ~-~ \beta_- \,) ~-~ {\Bar \epsilon}^+
\pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} {\Bar \zeta}_- ~~~~, \cr
\delta_Q \, H &=~ -\, i \, {\Bar \epsilon}^+ (\, \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} \rho_- ~-~ \beta_+
\,) ~-~ {\Bar \epsilon}^- \beta_- ~~~~, \cr
\delta_Q \, p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} &=~ \epsilon^+ \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} \rho_+ ~+~ \epsilon^- (\, \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }}
\rho_- ~-~ \beta_+ \,) ~+~ {\Bar \epsilon}^- \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} {\Bar \zeta}_- ~~~~, \cr
\delta_Q \, p_{{\scriptsize{\dvm}}~~} &=~ \epsilon^+ \beta_- ~+~ \epsilon^- \pa_{{\scriptsize{\dvm}}~~} \rho_- ~+~ {\Bar \epsilon}^+
\pa_{{\scriptsize{\dvm}}~~} {\Bar \zeta}_+ ~~~~, \cr
\delta_Q \, \beta_+ &=~ i\, {\Bar \epsilon}^+ \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} u ~+~ i {\Bar \epsilon}^- \pa_{{\scriptsize{\dvm}}~~}
(\, \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} B ~-~ p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} \,) ~~~~, \cr
\delta_Q \, \beta_- &=~ - i\, {\epsilon}^- \pa_{{\scriptsize{\dvm}}~~} H ~-~ i {\Bar \epsilon}^+ (\, \pa_{{\scriptsize{\dvm}}~~}
\pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} B ~-~ \pa_{{\scriptsize{\dvm}}~~} p_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} ~-~ \pa_{{{ =\hskip-3.75pt{\lin}}\hskip3.75pt }} p_{{\scriptsize{\dvm}}~~} \,)
{}~~~~. }
\eqno(B.3) $$
Finally for the superfield action that should act as the starting point
for the 2D (2,2) Kazama-Suzuki models we propose
$$\eqalign{ {~~~~}
{\cal S}_{KS} ~=~ &\int d^2 \sigma \, d^2 \zeta \, d^2 {\Bar \zeta} ~ {\Hat \Omega} (
{\Phi}, \, {\Bar \Phi}; {\Sigma} , \, {\Bar \Sigma} \,) \cr
& ~+~ \Big[ \int d^2 \sigma \, d^2 \zeta ~ {\cal J}_{\rm {I \, J}}
(\Phi) \, ({\Bar D}_+ \Sigma^{\rm I} \,) \, ({\Bar D}_- \Sigma^{\rm J} \,) ~+~
{\rm {h. \, c.}} ~+~ ... ~ \Big]
{}~~~~. }
\eqno(B.4) $$
The terms in the ellipsis represent the introduction of world sheet
2D, N = 1 gauge superfields for the $H$ sub-group in the K-S constructions.
In (B.4) the potential $ {\Hat \Omega} ( {\Phi}, \, {\Bar \Phi}; {\Sigma} , \,
{\Bar \Sigma} \,)$ is most likely given by (2.7) with $g_{\rm { I \, J}}$
constructed from the Maurer-Cartan forms as in (A.4) and ${\cal J}_{
\rm {I \, J}}(\Phi)$ is given by
$$
{\cal J}_{\rm { I \, J}}(\Phi) ~=~ - \, c_0 f_{\rm { K \, L \, M}}
\, \int_0^1 d y ~ y^2 \, \Phi^{\rm N} {\Hat L}_{\rm N} {}^{\rm K}
{\Hat L}_{\rm I} {}^{\rm L } {\Hat L}_{\rm J} {}^{\rm M} ~~~~.
\eqno(B.5) $$
Here the Maurer-Cartan forms are defined in terms of the chiral superfields
and the group is arbitrary. However, the final arbiter that determines
these functions is 2D, N = 2 superconformal invariance. This is a topic
to be studied in the future. Thus, we see for every compact group,
there exist a way to construct a 2D, N = 2 action that possesses manifest
supersymmetry. Let us emphasize that (2.7) is an explicit construction
that associates with every group manifold with metric $g_{\rm {I \, J}}$
(constructed from the group Maurer-Cartan forms) a special K\" ahler geometry
with a metric whose potential is given by (2.7). To our knowledge this is
the first observation relating group manifolds to special Kahler geometry
in this manner. It will be of interest to see if the condition of quantum
superconformal invariance acts as a restriction to the choices considered by
Kazama and Suzuki. Finally, we note that there must exists twisted versions
of the action of (B.5). That is the chiral superfields in (B.5) can be
replaced by twisted chiral superfields, if simultaneously we replace the
complex linear superfields by twisted complex linear superfields, $\Xi$,
(i.e. $\Sigma \to \Xi$ where $\Xi$ satisfies $ {\Bar D}_+ D_- \Xi = 0$).
Let us be explicit, we expect a subclass of the actions of (B.4) to
describe a fundamentally {\underline {new}} class of 2D, N = 2
superstrings. As long ago as 1989, we reported that at the level
of superfields\footnote{See S.J.Gates, Jr., R. Oerter and L. Lu,
Phys. Lett. 218B (1989) 33.} there were at least three different
N = 2 superstring actions. One of these, which actually has an N = 4
rigid supersymmetry (one chiral plus one twisted chiral
multiplet) is known to permit a non-trivial axion background
unlike the other two version. However, the axion occurs as
the second derivative of a potential. Our new theories are not
subject to this constraint. So we believe with (B.4) we have yet
again increased the number of known 2D, N = 2 superstrings. These
new N = 2 superstrings are associated with different choices of
auxiliary fields. So even for string theory we have evidence that
auxiliary fields matter...a point totally absent in superconformal
field theory.
\newpage
|
\section{Introduction}
The success of chiral perturbation theory ($\chi$PT)\cite{Wphilo}
for understanding properties of the pseudoscalar mesons is
now well established\cite{GassLeut}. The approach is based on the
existence of a systematic expansion in derivatives of the
pion's field and the pion's mass,
whereby $m_\pi$ divided by some large scale,
generated by the theory itself and typically $\sim 4 \pi f_\pi$, becomes
the perturbative expansion parameter. For the purely mesonic sector this
expansion is in fact quadratic in the pion's mass, so that even for the
$SU_f(3)$ generalization, $(m_K/4 \pi f_\pi)^2$ is still a
reasonablely small parameter.
The application to the baryon sector has however from the outset been
confronted with a variety of difficulties~\cite{GassNuc}.
For example, how to handle the nucleon's mass was
a problem only relatively recently solved~\cite{Wcount,JenkMan1}.
One unavoidable complication when including baryons is that
the chiral expansion is itself more complicated~\cite{GassNuc,Wcount}
than in the purely mesonic case,
and the expansion parameter is now only linear in $m_\pi$ ($m_K$).
A second pertinent complication involves the issue of resonances.
\footnote{Because of the large mass difference between the
$\rho$ and the pion, as
well as the quadratic nature of the chiral expansion,
this issue does not arise when
considering the sector of purely pseudoscalar, goldstone bosons.
The effects of the $\rho$ and other
such states are incorporated as part of the needed
input coupling constants of the theory.}
Originally it was conjectured~\cite{Wcount} that all such resonances
(and most notably the $\Delta$) need not be included as an explicit
degree of freedom,
{\it i.e.} that they could be ``integrated out''.
While at least one notably active group~\cite{BerKaisM}
have maintained this viewpoint,
\footnote{Arguing that for a very limited region near
threshold in the $SU_f(2)$ defined theory,
the $\Delta$-nucleon mass difference can be considered ``large''}
most researchers have subsequently found it untenable~\cite{Kolck}.
The $\Delta$ degree of freedom was first introduced into $\chi$PT
by Jenkins and Manohar in Ref.~\cite{JenkMan2}.
Recently, the importance of the $\Lambda^*(1405)$ for
understanding threshold kaon--nucleon
scattering lengths has also been realized~\cite{MRho1,Savage,MRho2}.
In this paper we discuss the role of higher--multiplets
for the properties of the
$\Delta$ decuplet at the one--loop level in $\chi$PT.
We consider the ${\cal O}(p^3)$ correction to the decuplet masses
and the ${\cal O}(p^2)$, one--loop correction to the magnetic moments of the
decuplet (the electromagnetic vertex has chiral power $-1$ excluding
whatever power maybe assigned to electric charge).
Our criterion for which multiplets to consider
is that the average mass--splitting, $\delta_h$, between the multiplet and the
$\Delta$ decuplet be less than the mass of the kaon,
\footnote{$m_\eta$ is obtained at this order in $\chi$PT using
GMOR~\cite{GMOR}.}
\begin{equation}
\delta_h = \mid M_H - M_{10} \mid < m_K.\label{criteria}
\end{equation}
This criterion is based on the fact that when it is met, an expansion
in $m_K/\delta_h$ is not justified so that loop effects involving these
higher--mulitplet members as intermediate states {\it cannot} be absorbed into
higher order terms in the chiral lagrangian.
Such loop effects place a fundamental limitation on any formulation of
$\chi$PT that omits the higher--multiplet as an explicit degree of freedom,
{\it even if the chiral expansion was then executed to all orders.}
By the convention of Ref. \cite{ourselves},
note that the chiral power of all $\delta$ is $1$.
For the case of the nucleon octet Eq. (\ref{criteria}) is clearly met,
\begin{equation}
\delta_N = M_{10} - M_8 = 226{\rm MeV},\label{deltaN}
\end{equation}
($M_{10} = 1377$MeV and $M_8 = 1151$MeV)
and is indeed the driving phenomenological reason for expecting that
the $\Delta$ cannot
be ignored for descriptions of the nucleon.
We focus here on the properties of the
$\Delta$ decuplet as opposed to those of the nucleon octet because of the
simple
reason that the mass splitting $\delta_h$ satisfies the criterion specified by
Eq.~(\ref{criteria}) for at least two resonances.
We will show that the Roper has a nontrivial effect on both decuplet
mass splittings
and magnetic moments. We should also note that these results lead to a
more general statement. When adding a loop one must also add resonances
whose masses are within a kaon mass of the resonances already included.
The rest of this paper is organized thusly. In Section (II) we
enumerate the various multiplets considered and their interactions
(and experimentally obtained couplings) with the
$\Delta$ decuplet utilizing the heavy baryon formalism
of Jenkins and Manohar~\cite{JenkMan1}.
In Section (III) we present the one--loop, $O(p^3)$ contributions to
the decuplet masses focussing on the violation to the
Decuplet Equal Spacing rule~\cite{nobel},
the sole quantity for which $\chi$PT makes a prediction at this order
in the chiral expansion~\cite{Jenkins,ourselves}.
In Section(IV) we consider the
one--loop, ${\cal O}(p^2)$ results for the magnetic moments of the decuplet,
a subject first discussed in $\chi$PT in
Ref.~\cite{bss} where though adherence
to the order of the chiral expansion was not strictly maintained.
We demonstrate later that strict adherence
is crucial for renormalizability. The focus in both
Sections (III) and (IV) is the quantitative importance of the
higher--multiplets.
In addition new relations for the magnetic moments at this order in
$\chi$PT are derived that are independent of the intermediate multiplets
considered. Their violation would be a clear measure of the
importance of higher chiral power terms in the expansion.
In Section (V) we conclude with a
discussion of the consequence of these results to the
loop--expansion in general in $\chi$PT.
\section{Higher Multiplets: Definitions and Couplings}
A number of multiplets
\footnote{See {\it e.g.} Table 30.4 in the Particle Data Group~\cite{PDG}.
As we are at present only concerned with the average coupling of these
multiplets with the $\Delta$ decuplet, we ignore potentially
interesting questions as to the exact
$SU_f(3)$ composition of any particular excited state.~\cite{yasuo}
We also omit from consideration possible exotics.}
satisfy the criteria Eq.~(\ref{criteria}). Fortunately most of
these can be eliminated due to symmetry constraints. For example,
flavor singlets such as the
$\Lambda^*(1405)$ do not couple to a decuplet via an $SU_f(3)$ octet
(the goldstone bosons).
Only slightly less straightfoward, a $1/2^-$ octet ({\it e.g.} the
N(1535) multiplet)
couples only through the lower components of the baryon spinors,
\begin{equation}
{\cal L}^i \sim \overline{T}^\mu \gamma_5 A_\mu B^*,
\end{equation}
which vanishes to lowest order in the
heavy baryon expansion~\cite{JenkMan1}.
(Such states would, in principle, need to be considered in
higher--order calculations.)
Coupling to the $5/2^-$ octet likewise vanishes at lowest order.
With these eliminations,
we obtain that only octets or decuplets of baryons with quantum numbers
$1/2^+, \, 3/2^+, \, 3/2^-$ and $5/2^+$ need be considered.
The most important $1/2^+$ multiplet (beyond of course the nucleon's) is the
octet containing the Roper, N(1440). A slight difficulty
arises in determining
\begin{equation}
\delta_R = M_8(1440) - M_{10}
\end{equation}
because one member of the Roper multiplet, the excited Cascade,
has not yet been identified. To get a reasonable approximate
value for its mass, we use the
corresponding GMO relation~\cite{GMO}
\begin{equation}
M_{\Xi^*}=\frac32\,M_{\Lambda^*}(1600) +
\frac12\,M_{\Sigma^*}(1660) - M_{N^*}(1440),
\end{equation}
by which one estimates that $M_{\Xi^*} = 1790$.
The average value of the Roper multiplet is thereby
\begin{eqnarray}
M_{8^*} &=& \frac18 (2 M_{\Xi^*} + 3 M_{\Sigma^*} +
M_{\Lambda^*} +2 M_{N^*})\nonumber\\
&=& 1630 \, {\rm Mev},
\end{eqnarray}
from which one gets that
\begin{equation}
\delta_R = 253{\rm MeV}.
\end{equation}
For the $\Delta N^* \pi$ interaction one has
\begin{equation}
{\cal L}^i = {\tilde C} (\overline{T}^\mu A_\mu B^* + {\rm h. c.})
\end{equation}
in complete analogy to the leading $\Delta N\pi$ interaction in the
heavy fermion limit.~\cite{JenkMan2} The coupling ${\tilde C}^2$ can
be obtained from
the observed decay of the $N^*(1440) \rightarrow \Delta \pi$ which has
a partial width
$\Gamma_{N^* \rightarrow \Delta \pi} \approx 90$MeV.
Comparing this with the decay of the $\Delta$,
$\Gamma_{\Delta \rightarrow N \pi} \approx 120$MeV one obtains that
\begin{equation}
{\tilde C}^2 \approx \frac12 {\cal C}^2.\label{couplingt}
\end{equation}
In principle, one other $1/2^+$ multiplet meets our criteria, the octet
containing the
N(1710), with a $\delta_h \approx 470$MeV. However, best
estimates~\cite{PDG} for the partial width for
N(1710)$\rightarrow \Delta \pi$ is $\approx 30$MeV, which implies
that the relevant coupling constant is
significantly suppressed
compared to that in Eq.~(\ref{couplingt}). We therefore
ignore this multiplet in our subsequent calculations as
it amounts to only a small correction to those of the Roper octet.
For the lowest lying $3/2^-$ octet, we obtain directly from the
experimentally measured
masses~\cite{PDG} that
\footnote{As a self--consistent test of the assignment of
hadrons to this multiplet, note that the violation to the
corresponding GMO relation for this octet is only $15$MeV.}
\begin{equation}
\delta_{8^*} = 296{\rm MeV}.
\end{equation}
The interaction with the $\Delta$ decuplet is, to leading order in the
chiral lagrangian,
\begin{equation}
{\cal L}^i = C^*\left(\overline{T}^\mu \gamma_\nu A^\nu T^*_\mu +
{\rm h. c.}\right).\label{3halfminus}
\end{equation}
The coupling $C^*$ can be determined from the observed decay of the
$N^*(1520)$, $\Gamma(N^*(1520) \rightarrow \Delta \pi) \approx 25$MeV,
whence one finds that
\begin{equation}
\left(\frac{C^*}{\tilde C}\right)^2 \approx \frac{1}{25}.\label{suppC}
\end{equation}
This relative suppression results both from the smaller branching width
as well as overall kinematic factors that otherwise tend to enhance
$N^*(1520) \rightarrow \Delta \pi$ w.r.t. $N^*(1440) \rightarrow \Delta \pi$.
There are two other $3/2^-$ multiplets, one octet and one decuplet,
listed in the Particle Data Group that could potentially satisfy our
criteria, Eq.~(\ref{criteria}). Each are very poorly determined, containing
merely one member each, the N(1700) and the $\Delta(1700)$, respectively.
In the case of the N(1700), its coupling to the $\Delta \pi$ is
experimentally negligible and hence can be safely ignored.
On the other hand, the coupling with the $\Delta(1700)$ is
not so readily ignored, having a decay width
$\Gamma(\Delta(1700) \rightarrow \Delta \pi) \approx 120$MeV.
We therefore explicitly keep this decuplet,
assigning for its intermultiplet spacing with the
$\Delta$ decuplet a value
\begin{equation}
\delta_{10^*} = 1700 - 1232 = 468 {\rm MeV}.
\end{equation}
{}From the aforementioned decay width, and an interaction of the form
Eq.~(\ref{3halfminus}),
we obtain that for the $\Delta^* \Delta \pi$ coupling
\begin{equation}
{\cal H}^{*2} \approx .15.\label{suppH}
\end{equation}
Here we have used the convention of Ref.~\cite{Jenkins} for the
$SU_f(3)$ algebra factors
(whereby, for the $\Delta \Delta \pi$ coupling, ${\cal H}^2 \approx 4$
is typical).
As in the case of the $N(1520)$ overall kinematic factors in addition to the
available phase space, yields a rather suppressed value of the coupling.
Indeed, as we will soon see, due to
Eqs.~(\ref{suppC}) and (\ref{suppH}) little
would have been lost had we ignored the $3/2^-$ multiplets altogether.
Nevertheless, they have been included for completeness.
We come finally to the $5/2^+$ states. The lowest such multiplet is
the N(1680) octet. It has an intermultiplet mass splitting with the
$\Delta$ decuplet of $\delta_h = 496$MeV. We conclude that there is
no $5/2^+$ multiplet that meets our criteria Eq.~(\ref{criteria}).
This then concludes our examination of the relevant multiplets. By far
the most important, as we will presently see, is the Roper octet.
\newpage
\section{Decuplet Equal Spacing Rule}
The one--loop, $O(p^3)$ results for the masses of the decuplet
involving intermediate
$\Delta$ decuplet and nucleon octet states have been published
previously~\cite{ourselves}. The contribution, $\delta M_{10}$,
from the $3/2^-$ multiplet, for the case $m > \delta_{8^*}$, is
\begin{eqnarray}
&\delta M_{10} &= \frac{- 3 \beta \delta_{8^*}}{16 \pi^2 f^2_\pi}
\times\nonumber\\
&&\left[
\left( \delta_{8^*}^2 - \frac12 m^2 \right)
\left( \frac{1}{\epsilon} - \gamma_E +
{\rm ln}(4\pi) + 2- {\rm ln}\frac{m^2 }{\mu^2} \right)
\right. \nonumber\\
&&+\left.
2 \delta_{8^*} \sqrt{m^2 - \delta_{8^*}^2}
\left(\frac{\pi}{2} - {\rm tan}^{-1}\frac{\delta_{8^*}}
{ \sqrt{m^2-\delta_{8^*}^2} }\right)
\right]
\label{uvterms}
\end{eqnarray}
where $\beta$ represents SU(3) algebra factors~\cite{Jenkins,ourselves}.
As discussed in Ref.~\cite{ourselves}, the counterterms neccessary
to renormalize these terms are either of the
form $\delta_{8^*}^2 {\cal L}_0^{\pi N}$ (for the $\delta_{8^*}^3$ divergences)
or $\delta_{8^*} {\cal L}_1^{\pi N}$ (for the $\delta_{8^*} m^2$ divergences).
As the DES rule is exact for all terms through $m^2$,
all counterterms (divergences) cancel at this order of the chiral expansion
in the violation to the DES rule.
Including all multiplets, the one--loop, ${\cal O}(p^3)$,
violation to the DES rule is
\begin{eqnarray}
(M_{\Sigma^*} &-& M_{\Delta}) - (M_{\Xi^*}-M_{\Sigma^*})
=\nonumber\\
(M_{\Xi^*} &-& M_{\Sigma^*}) - (M_{\Omega^-}-M_{\Xi^*})
=\nonumber\\
\frac12\{(M_{\Sigma^*} &-& M_{\Delta}) -(M_{\Omega^-}-M_{\Xi^*})\}
=\nonumber \\
&&\frac29\, \left( {\cal C}^2 V(-\delta_N) + {\tilde C}^2 V(\delta_R) +
(C^*)^2 V^*( \delta_{8^*}) \right)\nonumber\\
&&- \frac{20}{81}\, \left( {\cal H}^2 V(0) +
{\cal H^*}^2 V^*( \delta_{10^*}) \right).
\label{DESf}
\end{eqnarray}
\footnote{Note that unlike our convention in ~\cite{ourselves}, all
$\delta_h$ are now strictly positive and hence the explicit
sign in the function $V$ above.}
$V$ and $V^*$ are given by
\begin{eqnarray}
V(\delta) &=& W(m_K,\delta,\mu)\nonumber\\
&&-\frac34\,W(m_\eta,\delta,\mu)
-\frac14\,W(m_\pi,\delta, \mu),\nonumber\\
V^*(\delta) &=& W^*(m_K,\delta,\mu)\nonumber\\
&&-\frac34\,W^*(m_\eta,\delta,\mu)
-\frac14\,W^*(m_\pi,\delta, \mu),
\label{Vequation}
\end{eqnarray}
wherein the function $W(m,\delta,\mu)$ is~\cite{JenkManconf,BerMeis1}
\footnote{Note the correction from ~\cite{ourselves} regarding the
arctangent term in the case $m>\mid\delta\mid$.}
\begin{eqnarray}
\delta=0,\,\,\,\, && W(m,\delta, \mu)
=\frac{1}{16 \pi f^2_\pi}m^3, \label{del0}\\
m>\mid\delta\mid,\,\,\,\, && W(m,\delta,\mu)
=\nonumber\\ &&\frac{1}{8 \pi^2 f^2_\pi}(m^2 - \delta^2)^{3/2}
\left(\frac{\pi}{2} - {\rm tan}^{-1}\frac{\delta}{ \sqrt{m^2-\delta^2}}\right)
\nonumber\\
&&- \frac{3\delta}{32 \pi^2
f^2_\pi}\left(m^2-\frac{2}{3}\delta^2\right)
{\rm ln}\frac{m^2}{\mu^2}\nonumber\\
\mid\delta\mid>m,\,\,\,\, &&W(m,\delta,\mu) =\nonumber\\
&&\frac{-1}{16 \pi^2f^2_\pi}(\delta^2-m^2)^{3/2}
{\rm ln}\frac{\delta-\sqrt{\delta^2-m^2}}{
\delta+\sqrt{\delta^2-m^2}}\nonumber\\
&&- \frac{3\delta}{32 \pi^2
f^2_\pi}\left(m^2-\frac{2}{3}\delta^2\right)
{\rm ln}\frac{m^2}{\mu^2}\nonumber\\
\end{eqnarray}
and $W^*(m,\delta,\mu)$ is
\begin{eqnarray}
m>\mid\delta\mid,\,\,\,\, && W^*(m,\delta,\mu)
=\nonumber\\ &&\frac{3 \delta^2}{8 \pi^2 f^2_\pi}\sqrt{m^2 - \delta^2}
\left(\frac{\pi}{2} - {\rm tan}^{-1}\frac{\delta}{ \sqrt{m^2-\delta^2}}\right)
\nonumber\\
&&-\frac{3\delta}{16 \pi^2f^2_\pi}( \delta^2-\frac12m^2 )
{\rm ln}\frac{m^2}{\mu^2}\nonumber\\
\mid\delta\mid>m,\,\,\,\, && W^*(m,\delta, \mu)=\nonumber\\
&&\frac{3\delta^2}{16 \pi^2f^2_\pi}\sqrt{\delta^2-m^2}
{\rm ln}\frac{\delta-\sqrt{\delta^2-m^2}}{
\delta+\sqrt{\delta^2-m^2}}\nonumber\\
&&-\frac{3\delta}{16 \pi^2f^2_\pi}( \delta^2-\frac12m^2 )
{\rm ln}\frac{m^2}{\mu^2}.
\end{eqnarray}
As was already mentioned in Section (II),
the contribution from the $3/2^-$ multiplets is
essentially negligible due to their suppressed coupling constants,
Eqs.~(\ref{suppC}) and (\ref{suppH}). Explicitly we find that,
\begin{eqnarray}
\frac29 C^{*\,2} V^*(+\delta_{8^*}) &\approx& -.06{\rm MeV},
\nonumber\\
-\frac{20}{81}{\cal H}^{*2} V^*(+\delta_{10^*})
&\approx& +.3{\rm MeV} \label{m3/2}
\end{eqnarray}
which are indeed negligible. Hence we omit further considerations of
the $3/2^-$ multiplets.
This is not true of the Roper octet. Evaluating,
one finds that the ratio of the Roper to Nucleon multiplet contribution
\begin{equation}
\frac {{\tilde C}^2 V(+\delta_R)} {{\cal C}^2 V(-\delta_N)} = .21
\end{equation}
where result (\ref{couplingt}) has been used.
Taking ${\cal C}^2 = 2$ one obtains that the Roper contribution
\begin{equation}
\frac29{\tilde C}^2 V(+\delta_R) = -5.7{\rm MeV}, \label{mR}
\end{equation}
which is, in absolute magnitude, as large as the average, experimental
value of $6.8$MeV. It clearly cannot be ignored.
\newpage
\section{Decuplet Magnetic Moments}
The topic of the magnetic moments of the decuplet in the context of chiral
perturbation theory was first discussed in the work of Ref.~\cite{bss}.
Apart from the inclusion of the Roper as an intermediate state, our work
differs from Ref.~\cite{bss} in the treatment of $SU_f(3)$ symmetry.
In our calculation, the symmetry of the decuplet states
is broken through the meson masses appearing in the one loop calculation.
The meson masses are taken to be proportional to the current quark masses
with the up and down quark masses being equal. The strangeness~\cite{Jenkins}
and charge dependences~\cite{LL} of the baryon masses are regarded as effects
of chiral power 1 or more. The quantity $f_K-f_\pi$ has chiral singularity
$\sim m_\pi^2log m_\pi^2$~\cite{DGH} and has chiral power $2$.
Our calculations of the decuplet mass splittings and magnetic moments are
limited to chiral power ${\cal O}(p^3)$ and ${\cal O}(p^2)$, respectively.
Hence, we set $f_K=f_\pi$ and do not include in one loop calculation the sigma
terms from ${\cal L}_1$ which produce strangeness dependence of baryon
masses at the tree level. We ignore charge-dependence of the masses altogether.
The advantage of this strategy in the calculation of baryon masses is
well-known~\cite{Jenkins,ourselves}. The counter-terms which appear at one
loop level simply renormalizes the sigma terms. We find a similar result in the
magnetic moment calculation at one loop level, namely, that the counter terms
are strictly proportional to the baryon charge and hence renormalize
the tree level decuplet magnetic moment term, Eq.~(\ref{treemag}) below.
These advantages are lost if $f_K$ is not set equal to $f_\pi$\cite{bss}.
The lowest order term in the chiral lagrangian for the
magnetic moment of the $i$th member of the $\Delta$
decuplet is given by~\cite{gangof4}
\begin{equation}
{\cal L}_M = - \imath \frac{e}{M_N}
\mu_c q_i \overline{T}^\mu_i T^\nu_i F_{\mu \nu}
\label{treemag}
\end{equation}
where $q_i$ is the charge of the $i$th member.
The one--loop, ${\cal O}(p^2)$ corrections to
the magnetic moments result from vertex corrections in which the
external photon attaches
to the meson propagator~\cite{bss} and receives
contributions from intermediate states with
either a $3/2^+$ or $1/2^+$ baryon.
Photon attachments to the intermediate baryon are $m_\pi/M_N$
further suppressed as are also the contributions from $3/2^-$ baryons.
These latter are hence ignored as they
form part of the higher--order contribution in the chiral expansion.
Note that the $\eta$ meson, being electrically neutral,
also does not contribute
at the order being considered.
Following the notation of ~\cite{gangof4}, the magnetic
moment of the decuplet members, $\mu^{10}_i$, at the ${\cal O}(p^2)$,
one--loop contribution in the chiral expansion
is, in nuclear magneton units ($e/2 M_N$):
\begin{eqnarray}
&&\mu^{10}_i = q_i \mu_c + \sum_{ j=\pi,K}\frac{M_N}{32 \pi^2 f_\pi^2}
\left(\alpha^i_j \, \frac49 {\cal H}^2 {\cal F}(0,m_j,\mu) \right.
\nonumber\\
&& \left. -\beta^i_j\left[ {\cal C}^2
{\cal F}(-\delta_N,m_j,\mu)
+ {\tilde C}^2 {\cal F}(\delta_R,m_j,\mu) \, \right] \right).
\label{mags}
\end{eqnarray}
The function ${\cal F}(\delta,m,\mu)$ is ultraviolet divergent and given by
\begin{eqnarray}
m>\mid\delta\mid,\,\,\,\, &&{\cal F}(\delta,m,\mu) = \nonumber\\
&&\delta \left( -\frac{1}{\epsilon} + \gamma_E - {\rm ln}(4\pi) - \frac43 +
{\rm ln}(\frac{m^2}{\mu^2})\right)
\nonumber\\
&& - 2 \sqrt{m^2-\delta^2}
\left(\frac{\pi}{2} - {\rm tan}^{-1}\frac{\delta}{ \sqrt{m^2-\delta^2}}\right)
\nonumber\\
\mid\delta\mid>m,\,\,\,\, &&{\cal F}(\delta,m,\mu) = \nonumber\\
&& \delta \left( -\frac{1}{\epsilon} + \gamma_E - {\rm ln}(4\pi) +
{\rm ln}(\frac{m^2}{\mu^2})\right)
\nonumber\\
&&+ \sqrt{\delta^2 - m^2}
{\rm ln}\frac{\delta+\sqrt{\delta^2-m^2}}{\delta-\sqrt{\delta^2-m^2}}.
\label{Fdef}
\end{eqnarray}
The expression for the nonanalytic terms in ${\cal F}(\delta,m,\mu)$
appeared in Ref.~\cite{bss}.
The coefficients $\alpha^i_j$ and $\beta^i_j$ are simply related to the
coefficients $\alpha_{ij}$ and $\beta_{ij}$ of Ref.~\cite{bss}.
We multiply the coefficients $\alpha_{ij}$ by $3$ so that
they add up to the charge of decuplet $i$.
Unlike Ref.~\cite{bss}, we use the same mass for all members of a
baryon multiplet.
Accordingly we add the contributions of $\pi^\pm$ and of $K^\pm$.
The sum over $j$ in Eq.~(\ref{mags}) runs over two terms -- $\pi$ and K.
The resulting coefficients have surprizing simplicity. First,
$\alpha^i_j=\beta^i_j$.
Second, they may be expressed in terms of any two of the following
three -- charge,
$q^i$, isospin, $I^i_3$, and hypercharge $Y^i$ of decuplet $i$.
All three are traceless in any $SU(3)$ multiplet space.
We choose to use charge and isospin.
\begin{eqnarray}
\alpha^i_\pi=\beta^i_\pi&=&\frac23 I^i_3, \label {ab1} \\
\alpha^i_K=\beta^i_K&=&-\frac23 I^i_3+q^i, \label {ab2} \\
\alpha^i_\pi+\alpha^i_K&=& \beta^i_\pi+\beta^i_K=q^i. \label{ab3}
\end{eqnarray}
Eq.~(\ref{ab3}) is the key result for renormalizability. As a
consequence of these relations, the counterterm for the
ultraviolet divergences in ${\cal F}(\delta,m,\mu)$
(which are $m$ independent) is simply proportional to $\delta{\cal L}_M$,
Eq.~(\ref{treemag}), and therefore absorbed into a redefinition of $\mu_c$.
Note that this is precisely the same procedure as for the one--loop,
$\delta$--dependent contributions to the masses. We emphasize that this
procedure, and hence renormalizability, is tightly wedded to the systematics of
the chiral expansion (whereby $\delta$ and $f_\pi$ have
fixed values in Eq. (\ref{mags})).
The simplicities of the coefficients $\alpha^i_j$ and $\beta^i_j$
allows great simplification of Eq.~(\ref{mags}).
We introduce the combination:
\begin{eqnarray}
&&{\cal G}_j=\frac{M_N}{32 \pi^2 f_\pi^2}\left[
\frac49 {\cal H}^2 {\cal F}(0,m_j,\mu) \right. \nonumber \\
&& \left. - {\cal C}^2 {\cal F}(-\delta_N,m_j,\mu)
- {\tilde C}^2{\cal F}(\delta_R,m_j,\mu) \, \right], \label{mags2}
\end{eqnarray}
and rewrite the decuplet magnetic moments in the form:
\begin{equation}
\mu^{10}_i=q_i(\mu_c+{\cal G}_K)+\frac23\,I^i_3({\cal G}_\pi-{\cal G}_K).
\label{mags3} \end{equation}
Note that the form of Eq.~(\ref{mags3}), in particular the modification of
the coefficient of charge, reflects the choice of charge and isospin as the
two traceless quantities. The form would be different if we chose charge and
hypercharge or some other combination of isospin and hypercharge.
The second term is present only because the $SU(3)$ symmetry of the states is
broken through the difference in $\pi$ and $K$ masses. If we had used the same
masses the states would be pure decuplets and the Wigner-Eckart theorem
for $SU_f(3)$ would ensure that the magnetic moments are simply proportional
to the charge.
At tree level the decuplet magnetic moments are proportional to the
charges only.
The main one--loop results is the appearance of the second term of
Eq.~(\ref{mags3}). Now we need two magnetic moments to fix the two
coefficients in Eq.~(\ref{mags3}), viz,
$\mu_c+{\cal G}_K$ and ${\cal G}_\pi - {\cal G}_K$.
The only decuplet magnetic moment which has been measured is
$\mu_{\Omega^-} = -1.94 \pm .17 \pm .14$ nbm ~\cite{PDG}.
It fixes the coefficient of charge in Eq.~(\ref{mags3}). For the other
magnetic
moment we choose $\mu_{\Delta^0}$.\footnote{The $\Omega^-$, decaying only
weakly, is sufficiently long lived to allow such measurements.
Since all other members of the decuplet decay through the
strong interaction, it is a challenge to extract their magnetic moments
from experiment.} While not measured yet, it is given entirely by the loop
effect.
\begin{equation}
\mu_{\Omega^-}=-\mu_c-{\cal G}_K,\,\,\,\,\,\mu_{\Delta^0} =
\frac13\,({\cal G}_K - {\cal G}_\pi). \label{mags4}
\end{equation}
We can express the magnetic moments of all other decuplets in terms of these
two magnetic moments. Specifically, we derive the new relation that
\begin{equation}
\mu^{10}_i=-q_i\,\mu_{\Omega^-} - 2\,I^i_3\,\mu_{\Delta^0}.
\label{mags5}
\end{equation}
Explicit predictions for the eight other decuplet magnetic moments at the
one-loop ${\cal O}(p^2)$ level are listed below.
\begin{eqnarray}
\mu_{\Delta^{++}} &=& - 2 \mu_{\Omega^-} - 3 \mu_{\Delta^{0}}\nonumber\\
\mu_{\Delta^{+}} &=& - \mu_{\Omega^-} - \mu_{\Delta^{0}}\nonumber\\
\mu_{\Delta^{-}} &=& \mu_{\Omega^-} + 3 \mu_{\Delta^{0}}\nonumber\\
\mu_{\Sigma^{*+}} &=&- \mu_{\Omega^-} - 2 \mu_{\Delta^{0}}\nonumber\\
\mu_{\Sigma^{*-}} &=& \mu_{\Omega^-} + 2 \mu_{\Delta^{0}}\nonumber\\
\mu_{\Xi^{*-}} &=& \mu_{\Omega^-} + \mu_{\Delta^{0}}\nonumber\\
\mu_{\Xi^0} &=&- \mu_{\Delta^{0}}\nonumber\\
\mu_{\Sigma^{*0}} &=& 0.\label{murelate}
\end{eqnarray}
Independent of the explicit mulitplets included as intermediate states,
violations to these relations are {\it strictly} due to higher--order terms
in the chiral expansion.
Note that the magnetic moment of the $\Sigma^{*0}$
continues to be zero at this order in the expansion. Analogous relations
follow for the quadropole moments\cite{comment}. We note that
these relations are not obeyed by quenched lattice QCD\cite{Derek}.
This last result is perhaps not surprising as the quenched calculations
do not contain disconnected, quark loop diagrams\cite{Derek2}.
The explicit expression for $\mu_{\Delta^0}$ in terms of the functions
${\cal F}(\delta, m, \mu)$ is given below.
\begin{eqnarray}
\mu_{\Delta^0} =&& \frac{M_N}{96\pi^2 f_\pi^2}
\left( {\cal H}^2 \frac{4}{9}
\left( {\cal F}(0,m_K,\mu) - {\cal F}(0,m_\pi,\mu) \right) \right.\nonumber\\
&&-{\cal C}^2 \left( {\cal F}(-\delta_N,m_K,\mu) -
{\cal F}(-\delta_N,m_\pi,\mu)
\right) \nonumber\\
&&\left.-{\tilde C}^2 \left( {\cal F}(\delta_R,m_K,\mu) -
{\cal F}(\delta_R,m_\pi,\mu) \right) \right).\label{deltalong}
\end{eqnarray}
With the help of Eqs.~(\ref{Fdef}) it is easy to verify that
$\mu_{\Delta^{0}}$ is renormalization scale independent.
Since the magnetic moment of the ${\Delta^0}$ is given strictly by
loop--effects, it is an appropriate measure of the relative importance of
the Roper at the one--loop level. Explicitly one has from
Eq.~(\ref{deltalong}) that
\begin{equation}
\mu_{\Delta^{0}} = -.055 {\cal H}^2 + .229 {\cal C}^2 + .086 {\tilde C}^2,
\label{deltanumber}
\end{equation}
As in the case of the masses\cite{Jenkins} there is a strong
cancellation between the $\Delta$ decuplet and nucleon
octet intermediate states. This implies that $\mu_{\Delta^0}$
is a very delicate function of ${\cal H}^2$ and $ {\cal C}^2$
and therefore potentially very sensitive to the Roper contribution
(${\tilde C}^2$). Ambiguity in this regard resides in the fact
that ${\cal H}^2$ and $ {\cal C}^2$ are not sufficiently well--known
that a reliable prediction of $\mu_{\Delta^{0}}$ can be made.
To illustrate this point, we quote the results using
two ``representative'' values for the couplings, both with and without
the Roper included. For the coupling values used in Ref. \cite{ourselves},
one obtains only a mild dependence on the Roper,
\begin{eqnarray}
\mu_{\Delta^{0}} ( {\cal H}^2 = 4.4, {\cal C}^2 = 2,
{\tilde C}^2 = 0) &=& .21,\nonumber\\
\mu_{\Delta^{0}} ( {\cal H}^2 = 4.4, {\cal C}^2 = 2,
{\tilde C}^2 = 1) &=& .29,
\end{eqnarray}
while for the couplings used by Ref. {\cite{bss}}, we find a
much more dramatic dependence on the Roper,
\begin{eqnarray}
\mu_{\Delta^{0}} ( {\cal H}^2 = 4.84, {\cal C}^2 = 1.44,
{\tilde C}^2 = 0.0) &=& .05,\nonumber\\
\mu_{\Delta^{0}} ( {\cal H}^2 = 4.84, {\cal C}^2 = 1.44,
{\tilde C}^2 = .72) &=& .12.
\end{eqnarray}
The difference in $\mu_{\Delta^{0}}$ from these two parameter sets
is clearly sizable, as is the relative importance of the Roper.
If instead we chose to use as input the recent, model dependent
extraction\cite{experiment2} of the magnetic moment of the $\Delta^{++}, \,
\mu_{\Delta^{++}} = 4.5 \pm .5$ to infer $\mu_{\Delta^{0}}$ using the
relations Eq.~(\ref{murelate}), one then obtains that
$\mu_{\Delta^{0}} = -.2 \pm .2$. If this is indeed the data, then by
Eq.~(\ref{deltanumber}) the contribution of the Roper is, in absolute
magnitude, significant. As in the case of the mass--splittings,
a formulation of $\chi$PT without the Roper as an explicit degree of
freedom is intrinsically incapable of predicting such ``data''.
\newpage
\section{Conclusions}
{}From the results of the last two sections a few points are worth discussing.
First, that while the magnetic moment of the $\Delta^0$ sensitively depends
on the cancellation between terms depending on relatively ill--determined
coupling constants, the relations between the magnetic moments of the $\Delta$
decuplet given in Eq. (\ref{murelate}) are rigorous predictions of $\chi$PT
at $O(p^2)$.
We urge experimental activity to confront these predictions with
data. A new measurement of $\mu_{\Omega^-}$ with higher precision
will be most useful. At least two other decuplet magnetic moments need
to be measured, hopefully with a precision of $\sim 0.1$ n.m..
Second, that on the level of one--loop corrections in $\chi$PT, the
contribution of
the Roper octet to properties of the $\Delta$ decuplet is as important
as any other multiplet's contribution. We have seen this result
quantitatively in the case of the DES rule and the magnetic moment of
the $\Delta^0$. Both of these quantities are good measures of the
one--loop effects as they are each zero at lower--order in the chiral
expansion.
We expect that these results are illustrative and that they generalize to all
one--loop calculations for the $\Delta$ decuplet.
Since the mass--splitting, $\delta_R$, between the Roper octet and
$\Delta$ decuplet is less than that of the kaon's mass, a Taylor expansion
in $m_K/\delta_R$ is not permissible. Hence, these loop
effects cannot be be absorbed within higher order
terms of the chiral expansion of a theory not containing
the Roper as an explicit degree of freedom.
In such a theory it is indeed difficult to justify going to one--loop
or higher without inclusion of the Roper. Phenomenologically successful
results would have to be considered merely fortuitous unless shown
to be a result of more general considerations
(as in the relations of Eq. (\ref{murelate}) for the magnetic moments).
Third, that the above argument can be repeated in kind for the one--loop
corrections to the Roper resonance. That is, even higher--multiplets,
separated by $\delta_h$ in mass from the Roper octet by an
amount less than the kaon mass, will {\it a priori} be as important
quantitatively as the $\Delta$ decuplet for
properties of the Roper at the one--loop level. Since such corrections
are neccessary for a two loop calculation
\footnote{See though ~\cite{ourselves} for a discussion as to the
likely feasibility of such a program.} of the baryon masses,
we are led to conclude that the loop expansion in
general in the baryon sector of $\chi$PT is inevitably wedded to the
neccessity of including more and more multiplets in the theory
as fundamental fields. While such a result may not
be true in a particular limit of QCD
({\it e.g.} $m_{u, d, s} \rightarrow 0$ or $N_c \rightarrow \infty$),
it is a consequence of the experimental fact
that on the average, $m_\pi \approx \delta_h$.
\bigskip
{\centerline{ACKNOWLEDGEMENTS}} We thank
Y. Umino and T. D. Cohen for useful discussions.
This work was supported in part by DOE Grant DOE-FG02-93ER-40762.
|
\section{Introduction}
Let $G$ be a connected semisimple Lie group with finite center.
Let ${\bf g}$ be the Lie algebra of $G$ and $K$ be its maximal
compact subgroup. By ${{\cal HC}(\gaaa,K)}$ we denote the category
of Harish-Chandra modules of $G$.
The Harish-Chandra module $(\pi,V_{\pi,K})\in {{\cal HC}(\gaaa,K)}$ has natural globalizations
(analytic, smooth, distribution, and hyperfunction vectors)
$$V_{\pi,\omega}\subset V_{\pi,\infty }\subset V_{\pi,-\infty}\subset
V_{\pi,-\omega}\ .$$
Here $V_{\pi,*}$ is a complete locally convex vector space admitting
a continuous $G$-action such that the underlying $({\bf g},K)$-module
is $V_{\pi,K}$.
Let $\Gamma\subset G$ be a torsion-free
discrete subgroup of finite covolume.
Then a natural problem is to study the cohomology of
$\Gamma$ with coefficients in the different globalizations $V_{\pi,*}$.
The cohomology groups again become locally convex topological vector spaces.
Our work in \cite{bunkeolbrich947} led us to the following conjecture.
\begin{con}\label{grt}
If $\Gamma$ is cocompact, then for all $p\ge 0$ there are isomorphisms
\begin{eqnarray}H^p(\Gamma,V_{\pi,\omega})&=&
H^p(\Gamma,V_{\pi,\infty})\label{yu1}\\
H^p(\Gamma,V_{\pi,-\omega})&=& H^p(\Gamma,V_{\pi,-\infty})\label{entz}\\
H^p(\Gamma,V_{\pi,*})^*&=& H^{\dim(G/K)-p}(\Gamma,V_{\tilde{\pi},-*}),
\:\:*=\omega,\infty\label{yu2}
\end{eqnarray}
and all vector spaces are Hausdorff and finite-dimensional.
If $\Gamma$ has finite covolume, then
$H^\ast(\Gamma,V_{\pi,-\infty})$ is finite-dimensional and Hausdorff.
\end{con}
In \cite{bunkeolbrich947} we settled this conjecture in the rank-one case
for $*=\omega$ and cocompact $\Gamma$.
In the present paper we prove the conjecture in the case ${\mbox{\rm rank}}_{\bf R}(G)=1$
for $*=\infty$ admitting non-cocompact $\Gamma$ of finite covolume
(Section \ref{ggkk}).
As indicated in \cite{bunkeolbrich947}, the main result needed
to extend the method of \cite{bunkeolbrich947} to the case $*=\infty$
was the surjectivity of the $B:=\Omega-\mu$ ($\mu\in{\bf C}$, $\Omega$ - Casimir of
$G$)
on the space of sections of moderate growth of homogeneous bundles over
$X:=G/K$.
This is proved in Theorem \ref{sur} of the present paper.
We also obtain a more concrete description of the cohomology groups
in terms of automorphic and cusp forms.
Let now $G$ be of general rank and $\Gamma$ be cocompact.
Employing the recent result of Kashiwara-Schmid \cite{kashiwaraschmid94}, Thm.
6.13,
one can show (\ref{yu2}) for $*=\omega$
and $\dim\:H^*(\Gamma,V_{\pi,\pm\omega})<\infty$.
Using Theorem
\ref{casss} below for $*=\infty$ it is possible to prove
(\ref{yu1}) and (\ref{entz}).
We will explain this in a forthcoming paper.
The case of non-cocompact $\Gamma$ remains open for
${\mbox{\rm rank}}_R(G)>1$.
Our main motivation for considering the $\Gamma$-cohomology of
globalizations of Harish-Chandra modules was to prove a conjecture of Patterson
about the singularities of the Selberg zeta functions associated to $\Gamma$.
Let $G$ be a semisimple Lie group of real rank one and
and $P=MAN$ be a minimal parabolic subgroup of $G$.
Let ${\bf a},{\bf n}$ be the Lie-algebras of $A,N $.
Let $(\sigma,V_\sigma)\in\hat{M}$ be
an irreducible representation of $M$ and $\Gamma\subset G$ be a discrete
cocompact
torsion-free subgroup.
Then the Selberg zeta function $Z_\Gamma(s,\sigma)$, $s\in{\aaaa_\C^\ast}$,
is defined as the analytic continuation of the infinite product
$$Z_\Gamma(s,\sigma)=\prod_{[g]\in C\Gamma,n_\Gamma(g)=1}\prod_{k=0}^\infty
\:\det\left(1-a_g^{-\rho-s}S^k(Ad(m_ga_g)_{\bf n}^{-1})\otimes
\sigma(m_g)\right).$$
Here $C\Gamma$ is the set of hyperbolic conjugacy classes in $\Gamma$,
$n_\Gamma(g)$
is the maximal number $n\in{\bf N}$ such that $g=h^n$ for some $h\in\Gamma$
and $m_g\in M$, $a_g\in A^+$ are such that $g$ is conjugated in $G$ to
$m_ga_g$.
$S^k({\mbox{\rm Ad}}(m_ga_g)_{\bf n}^{-1})$ stands for the $k$'th symmetric power
of ${\mbox{\rm Ad}}(m_ga_g)^{-1}$ restricted to ${\bf n}$ and $\rho\in{\bf a}^\ast$ is defined
by
$\rho(H):=\frac{1}{2}{\mbox{\rm tr}}({\mbox{\rm ad}}(H)_{\bf n})$.
The infinite product converges for ${\rm Re }(s) > \rho$. In this generality
the Selberg zeta function was introduced by Fried \cite{fried86}.
He applied Ruelle's techniques of hyperbolic dynamics and gave a
meromorphic continuation of $Z_\Gamma(s,\sigma)$.
The parameters $(\sigma,\lambda)\in \hat{M}\times{\aaaa_\C^\ast}$ also define a principal
series
representation of $G$
on the Hilbert space
$$H^{\sigma,\lambda}=\{f:G\rightarrow
V_\sigma\:|\:f(gman)=a^{\lambda-\rho}\sigma(m)^{-1}f(g),\:\forall man\in MAN,
f_{|K}\in L^2\},$$
where $G$ acts by the left regular representation.
By $H^{\sigma,\lambda}_{-\infty}$ we denote the space of its distribution
vectors.
S. Patterson \cite{patterson93} conjectured the relationship of the
singularities of Selberg zeta functions
with the cohomology of $\Gamma$ with coefficients in the distribution vectors
of principal series representations. In \cite{bunkeolbrich947} we have shown an
analog of Pattersons conjecture for the hyperfunction vectors of principal
series representations.
Pattersons original conjecture now follows from (\ref{entz}).
\begin{theorem}\label{umth1}
The cohomology $H^p(\Gamma,H^{\sigma,\lambda}_{-\infty})$ is finite dimensional
for all $p\ge 0$,
\begin{eqnarray}
\chi(\Gamma,H^{\sigma,\lambda}_{-\infty})=\sum_{p=0}^\infty (-1)^p \dim
H^p(\Gamma,H^{\sigma,\lambda}_{-\infty})&=&0,\label{uass1}\\
\chi(\Gamma,\hat{H}^{\sigma,\lambda}_{-\infty})=\sum_{p=0}^\infty (-1)^p \dim
H^p(\Gamma,\hat{H}^{\sigma,\lambda}_{-\infty})&=&0,\label{uass2}
\end{eqnarray}
and the order of $Z_\Gamma(s,\sigma)$ at $s\in{\aaaa_\C^\ast}$
can be expressed in terms of the group cohomology of $\Gamma$ with coefficients
in $H^{\sigma,\lambda}_{-\infty}$ as follows :
\begin{eqnarray}{\mbox{\rm ord}}_{s=\lambda\not=0} Z_\Gamma(s,\sigma)&=&-\sum_{p=0}^\infty
(-1)^p p \dim H^p(\Gamma,H^{\sigma,\lambda}_{-\infty}),\label{pat}\\
\mbox{\em ord}_{s=0} Z_\Gamma(s,\sigma)&=&-\sum_{p=0}^\infty (-1)^p p \dim
H^p(\Gamma,\hat{H}^{\sigma,0}_{-\infty}),
\label{pat0}
\end{eqnarray}
where $\hat{H}^{\sigma,\lambda}_{-\infty}$ is a certain non-trivial extension
of $H^{\sigma,\lambda}_{-\infty}$ with itself.
\end{theorem}
Our description of the cohomology for non-cocompact $\Gamma$ of finite covolume
implies that the analog
of Pattersons conjecture is false in that case.
A very detailed analysis of the cohomology of Fuchsian groups of the first kind
with coefficients in principal series
representations is given in Section \ref{fufu}. We express the cohomology
in terms of automorphic and cusp forms. For the non-irreducible
principal series representations we find explicit formulas
for the dimensions of the cohomology groups in terms of the topology of
$Y=\Gamma\backslash X$.
The methods of this paper can also be used to study the ${\bf n}$-cohomology
of globalizations of Harish-Chandra modules.
Let $G$ be of general rank and
$P\subset G$ be a parabolic subgroup
with Langlands decomposition $P=M_PA_PN_P$.
Let ${\bf n}_P$ be the Lie algebra of $N_P$.
If $(\pi,V_{\pi,K})\in{{\cal HC}(\gaaa,K)}$, then by a theorem of Hecht-Schmid
\cite{hechtschmid83}
the Lie algebra cohomology groups
$H^*({\bf n}_P,V_{\pi,K})$ have natural structures of
modules in ${\cal HC}({\bf m}_P,K_P)$,
where $K_P=K\cap M_P$ and ${\bf m}_P$ is the Lie algebra of $M_P$.
Moreover, all elements of $H^*({\bf n}_P,V_{\pi,K})$ are finite
under ${\bf a}_P$, where ${\bf a}_P$ is the Lie algebra of $A_P$.
It is natural to ask whether globalization is compatible with
${\bf n}_P$-cohomology.
The standard ${\bf n}_P$-cohomology complex induces a natural topology on
$H^p({\bf n}_P,V_{\pi,*})$ such that $H^p({\bf n}_P,V_{\pi,*})$ becomes a
continuous
representation of $M_PA_P$.
The natural theorem is
\begin{theorem}\label{casss}
There are topological isomorphisms of $M_PA_P$ modules
\begin{eqnarray*}H^p({\bf n}_P,V_{\pi,*})&\cong& H^p({\bf n}_P,V_{\pi,K})_*\ , \\
H^p({\bf n}_P,V_{\pi,-*})^\ast&\cong&
H^{\dim({\bf n}_P)-p}({\bf n}_P,V_{\tilde{\pi},K})_{ *}\otimes
\Lambda^{\dim({\bf n}_P)}{\bf n}_P\ ,\quad \forall p\ge 0\end{eqnarray*}
for $*= \infty, \omega$, where $(\tilde{\pi},V_{\tilde{\pi},K})$ is the dual
Harish-Chandra module.
\end{theorem}
For $*=\infty$ this was proved by Casselman (unpublished).
In the rank-one case there were previous partial results by Osborne
\cite{osborne72}.
In the present paper we give a proof in the rank-one case using different
methods (Theorem \ref{haha}).
In \cite{bunkeolbrich947} we were able to prove this theorem in the case
${\mbox{\rm rank}}_{\bf R}(G)=1$ for
$*=\omega$.
For $G$ of general rank and $*=\omega$ the theorem
was announced by Bratten and Hecht-Taylor (compare \cite{hechttaylor93}).
It is now an easy consequence of the above-mentioned result
\cite{kashiwaraschmid94}, Thm. 6.13.
\section{Surjectivity on weighted spaces}
{}From now on let $G$ be a connected semisimple Lie
group with finite center of {\em real rank one} and $K$ be its maximal compact
subgroup.
Let $G\times_KV_\gamma=E$ be a $G$-homogeneous bundle over
the rank-one symmetric space $X=G/K$. Let ${\cal E}=C^\infty(X,E)$
be the space of smooth sections of $E$.
We define the increasing sequence of subspaces $S_R{\cal E}$, $R\in{\bf R}$,
by
$$S_R{\cal E}:=\{f\in{\cal E}\:|\: p_{-R,L }(f)<\infty\: \forall L\in{\cal U}({\bf g}) \},$$
where the seminorms $p_{R,L}$ are defined by
$$p_{R,L}(f) :=\sup_G {\rm e}^{R{\rm dist}(gK,{\cal O})} |f(Lg)|\ , $$
${\cal O}\in X$ is the class $[K]\in G/K$ considered as the base point of $X$,
and we view $f$ as a function on $G$
with values in $V_\gamma$ satisfying $f(gk)=\gamma(k)^{-1}f(g)$, $\forall k\in
K$.
$S_R{\cal E}$ is a Fr\'echet space and
for $R^\prime\ge R$ we have a continuous
inclusion $S_R{\cal E}\hookrightarrow S_{R^\prime}{\cal E}$.
Let $S_R{\cal E}^\prime:=(S_{-R}{\cal E})^\ast$ be the topological conjugate dual of
$S_{-R}{\cal E}$.
Then for $R \ge R^\prime$ there is a continuous projection
$S_{R^\prime}{\cal E}^\prime\rightarrow S_R{\cal E}^\prime$.
We also consider
\begin{eqnarray*}
S_\infty{\cal E}^\prime&:=&\lim_{\stackrel{\longrightarrow}{R}} S_R{\cal E}^\prime\ ,\\
S_\infty{\cal E}&:=&\{f\in{\cal E}\:|\:\forall L\in {\cal U}({\bf g})\: \exists R\in{\bf R} \:s.t.
\:p_{R,L}(f)<\infty\} \ .
\end{eqnarray*}
The space $S_\infty{\cal E}$ is a limit of Fr\'echet spaces in the following way.
Let ${\cal I}$ be the partially ordered set of all monotone functions
$u:{\bf N}_0\rightarrow {\bf R}$,
where $u\ge u^\prime$, iff $u(n)\ge u^\prime(n)$, $\forall n\in{\bf N}$.
For any $u\in {\cal I}$ we define the Fr\'echet space
$$S_u{\cal E}:=\{f\in{\cal E}\:|\: p_{-u(\deg(L)),L}(f)<\infty\:\:\: \forall
L\in{\cal U}({\bf g}) \}\ ,$$
where $\deg(L)$ is the order of $L$ as a differential operator on $G$.
Then
$$S_\infty {\cal E}=\lim_{\stackrel{\longrightarrow}{u\in {\cal I}}} S_u{\cal E}\ .$$
We equip $S_\infty {\cal E}$ with the topology of the direct limit.
The symbol $S_\infty{\cal E}$ is an abuse
of notation since $$S_\infty{\cal E}\not= \lim_{\stackrel{\longrightarrow}{R\in
{\bf R}}}S_R{\cal E}\ .$$
$S_\infty{\cal E}$ is called the space of sections of $E$ of moderate growth.
The space $S_\infty{\cal E}^\prime$ is the topological conjugate dual of $\cap_{R}
S_R{\cal E}$.
Since $C_c^\infty(X,E)$ is dense in $\cap_{R} S_R{\cal E}$,
we have an inclusion $S_\infty{{\cal E}^\prime}\hookrightarrow C^{-\infty}(X,E)$.
Let $\Omega$ be the Casimir operator of $G$ and for $\mu\in{\bf C}$ let
$B:=\Omega-\mu$.
Let ${\bf C}[B]$ be the ring of all polynomials in $B$.
We consider the functor ${\mbox{\rm Fin}}_\mu$ on the category of ${\bf C}[B]$-modules
defined by
$${\mbox{\rm Fin}}_\mu(V)=\{v\in V\:|\: \exists l\ge 0 \quad B^lv=0 \}\ .$$
This functor is left exact and its higher derived functors are denoted by
${\mbox{\rm Fin}}_\mu^i(V)$.
Since the inductive limit is an exact functor on the category of
${\bf C}[B]$-modules
and $${\mbox{\rm Fin}}_\mu(V)\cong\lim_{\stackrel{\longrightarrow}{j}} {\mbox{\rm Hom}}_{{\bf C}[B]}(
{\bf C}[B]/(B )^j,V)$$
we have
\begin{equation}\label{iop}{\mbox{\rm Fin}}_\mu^i(V)\cong
\lim_{\stackrel{\longrightarrow}{j}} {\mbox{\rm Ext}}^i_{{\bf C}[B]}({\bf C}[B]/(B )^j,V)\
.\end{equation}
Since ${\bf C}[B]$ is a regular ring of dimension one
we have ${\mbox{\rm Fin}}_\mu^i(V)=0$ for $i\ge 2$.
A ${\bf C}[B]$-module $V$ is called ${\mbox{\rm Fin}}_\mu$-acyclic, iff ${\mbox{\rm Fin}}_\mu^1(V)=0$.
\begin{theorem}\label{fin}
$S_\infty{{\cal E}^\prime}$ is ${\mbox{\rm Fin}}_\mu$-acyclic.
\end{theorem}
{\it Proof.$\:\:$}
We consider the Schwartz space
$$S_{-log}{\cal E}:=\{f\in{\cal E}\:|\: q_{L,k}(f)<\infty \quad \forall k\in{\bf N} , L\in
{\cal U}({\bf g})\}\ ,$$
where $$q_{L,k}(f)^2=\int_G {\rm dist}(gK,{\cal O})^k |f(Lg )|^2 dg$$
and its topological conjugate dual space
$$S_{log}{{\cal E}^\prime}=(S_{-log}{\cal E})^\ast\ .$$
\begin{lem}\label{ssur}
The inclusion $S_{log}{{\cal E}^\prime} \hookrightarrow S_\infty{{\cal E}^\prime}$
induces a surjection
$${\mbox{\rm Fin}}^1_\mu(S_{log}{{\cal E}^\prime})\rightarrow {\mbox{\rm Fin}}_\mu^1(S_\infty{{\cal E}^\prime})\ .$$
\end{lem}
{\it Proof.$\:\:$}
Let $V$ be a ${\bf C}[B]$-module.
We can compute
${\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B )^k,V)$ using the Koszul complex
$$0\rightarrow {\mbox{\rm Ext}}^0_{{\bf C}[B]}({\bf C}[B]/(B )^k,V)\rightarrow
V\stackrel{B^k}{\rightarrow}V\rightarrow
{\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B )^k,V)\rightarrow 0\ .$$
In order to perform the direct limit in (\ref{iop})
note that the map
$${\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B^k),V)\rightarrow
{\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B^{k+1}),V)$$
is induced by $B:V\rightarrow V$.
Hence Lemma \ref{ssur} follows from the next result.
\begin{lem}\label{iter}
Let $k\ge 1$.
For any $f\in S_\infty{{\cal E}^\prime}$ there exists a $g\in S_\infty{{\cal E}^\prime}$ with
$f-B^k g\in S_{log}{{\cal E}^\prime}$.
\end{lem}
In fact, Lemma \ref{iter} implies
that any element $f\in S_\infty{{\cal E}^\prime}$ representing some class in\linebreak[4]
${\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B )^k,S_\infty{{\cal E}^\prime})$ can be replaced by $f-B^kg\in
S_{log}{{\cal E}^\prime}$
representing the same class in ${\mbox{\rm Ext}}^1_{{\bf C}[B]}({\bf C}[B]/(B )^k,S_\infty{{\cal E}^\prime})$.\\
{\it Proof.$\:\:$}
It is enough to show that for $f\in S_R{{\cal E}^\prime}$ there exists
$g\in S_R{{\cal E}^\prime}$ such that $f-B^kg\in S_{R-1}{{\cal E}^\prime}$.
In fact, it is sufficient that this holds for all weights $R$
in a dense subset of ${\bf R}$.
Then the assertion of the lemma follows by a finite iteration.
We consider polar coordinates $ {\bf a}^+\times K/M $ of $X\setminus \{{\cal O}\}$.
We identify the fibre of $E$ over $(a,kM)$ with the fibre over $(a^\prime,kM)$
using the radial parallel transport.
There is a constant coefficient
operator $B^k_{rad}$ on ${\bf a}^+$ such that $B^k-B^k_{rad}={\rm e}^{-a}Q$ (we
identify
${\bf a}^+$ with ${\bf R}_+$ such that the short root has length one) and
$Q$ has bounded coefficients up to infinity.
Note that $Q$ also contains differentiations along ${\bf a}^+$ but with
exponentially decreasing
prefactors.
There are an endomorphism $F(kM)\in{\mbox{\rm End}}(E_{(a,kM)})$ and $c\in {\bf C}$ such that
$$B_{rad} = - \frac{d^2}{da^2}+c\frac{d}{da}+F(kM)\ .$$
We decompose the bundle $E_{|X\setminus {\cal O}}=\sum_{\sigma}E(\sigma)$
according to the
eigenvalues of $F(kM)$ (which are independent of $kM$).
There are $x_\sigma,y_\sigma\in {\bf C}$ such that
$B_{rad}=\sum_{\sigma}B_{rad}(\sigma)$ and
$B_{rad}(\sigma)=-(\frac{d}{da}-\bar{x_\sigma})(\frac{d}{da}-\bar{y_\sigma}) $.
We exclude the weights $R= {\rm Re }(x_\sigma), {\rm Re }(y_\sigma)$ for all $\sigma$
from the following consideration.
Let $S_R{\cal E}_0\subset S_R{\cal E}$ be the subspace of all sections vanishing in a
neighbourhood
of the origin ${\cal O}\in X$. Then $S_R{\cal E}_0$ is a limit of Fr\'echet spaces. Let
$S_R{{\cal E}^\prime}_0$
be the topological conjugate dual of $S_R{\cal E}_0$. Then $S_R{{\cal E}^\prime}\subset
S_R{{\cal E}^\prime}_0$.
Choose a cut-off function $\chi\in C^\infty({\bf a}^+)$ such that $\chi(a)=0$ for
$a<1$ and $\chi(a)=1$ for $a>2$. Then the multiplication by $\chi$ defines a
continuous
operator $S_R{{\cal E}^\prime}_0 \rightarrow S_R{{\cal E}^\prime}$.
Below we will construct a continuous solution operator
$H:S_R{\cal E}_0\rightarrow S_R{\cal E}_0$ such that
$ H(B^*_{rad})^k -{\mbox{\rm id}} : S_R{\cal E}_0\rightarrow C_c^\infty(X\setminus{\cal O},E)$
is continuous and ${\mbox{\rm supp}} (H(B^*_{rad})^k -{\mbox{\rm id}})f \subset [1,2]\times K/M$,
$\forall f\in S_R{\cal E}_0$.
Here $(B_{rad}^*)^k $ is the formal adjoint of $B^k_{rad}$.
The adjoint $H^*:S_R{{\cal E}^\prime}_0\rightarrow S_R{{\cal E}^\prime}_0$ then has the property that
$B^k_{rad}H^*\phi-\phi\in {{\cal E}^\prime}$ for $\phi\in S_R{{\cal E}^\prime}_0$.
Here ${{\cal E}^\prime}$ is the space of distributions with compact support.
The composition $\chi\circ H^*_{|S_R{{\cal E}^\prime}}:S_R{{\cal E}^\prime}\rightarrow S_R{{\cal E}^\prime}$
satisfies $B^k_{rad}\chi\circ H^*\phi-\phi\in {{\cal E}^\prime}$, $\phi\in S_R{{\cal E}^\prime}$. Hence
$B^k H^*\phi-\phi=e^{-a}QH^*\phi\:({\rm mod}\: {\cal E}^\prime)$.
But $e^{-a}QH^*\phi\in S_{R-1}{{\cal E}^\prime}$.
This proves the lemma assuming that we have already constructed $H$.
Note that $ B^*_{rad}(\sigma)^k$
is a product of operators of first order
$\frac{d}{da}+x_\sigma$, $\frac{d}{da}+y_\sigma$.
It is enough to construct solution operators
$H_{x_\sigma},H_{y_\sigma}:S_R{\cal E}_0(\sigma)\rightarrow S_R{\cal E}_0(\sigma)$
such that ${\mbox{\rm supp}} (H_{x_\sigma}(\frac{d}{da}+x_\sigma)f-f)\subset [1,2]\times
K/M$, $\forall f\in S_R{\cal E}_0(\sigma)$,
and similarly for $H_{y_\sigma}$.
Then $H:=(-1)^k\sum_{\sigma}H_{x_\sigma}^kH_{y_\sigma}^k$ has the required
properties.
For $R-{\rm Re }(x_\sigma)>0$ we set (simplifying the notation by omitting the
angular variable)
$$(H_{x_\sigma}f)(a):=-\chi(a){\rm e}^{-x_\sigma a}\int_a^\infty {\rm e}^{x_\sigma b}
f(b) db\ .$$
Since $f\in S_R{\cal E}_0(\sigma)$ we have for some $C<\infty$ and all
$b\in{\bf a}^+$ that
$|f(b)|\le C {\rm e}^{-Rb}$. Hence the integral converges.
If $R-{\rm Re }(x_\sigma)< 0$, then we set
$$(H_{x_\sigma}f)(a):=\chi(a){\rm e}^{-x_\sigma a}\int_0^a {\rm e}^{x_\sigma b} f(b)
db\ .$$
It is easy to check that ${\mbox{\rm supp}} (H_{x_\sigma} (\frac{d}{da}+x_\sigma)
f-f)\subset [1,2]\times K/M$.
We must show that $H_{x_\sigma}:S_R{\cal E}_0(\sigma)\rightarrow S_R{\cal E}_0(\sigma)$
and that $H_{x_\sigma}$ is continuous.
To define the topology of $S_R{\cal E}_0(\sigma)$ it is sufficient to consider the
set of seminorms
$\tilde{p}_{R,L}$ with $L=(L_1,L_2)\in{\cal U}({\bf a})\times {\cal U}({\bf k})$.
Set $(Lf)(a,kM)=f(L_1a,L_2kM)$.
Then $$\tilde{p}_{R,L}: =\sup_{a\in{\bf a}^+, k\in K} {\rm e}^{Ra} |(Lf)(a,kM)| \ .$$
It is clear that $LH_xf=H_x Lf + W_{L}f$, where
$W_{L}:S_R{\cal E}_0(\sigma)\rightarrow C_c^\infty(X\setminus {\cal O},E(\sigma))$ is
continuous.
Thus in order to show that $H_{x_\sigma}$ is continuous it is enough
to verify the estimate
$$\tilde{p}_{R,1}(H_{x_\sigma}f) \le C \tilde{p}_{R,1}(f)\ .$$
We employ that $|f(a,kM)|\le \tilde{p}_{R,1}(f){\rm e}^{-Ra}$.
If $R-{\rm Re }(x_\sigma)>0$, then
\begin{eqnarray*} \tilde{p}_{R,1}(H_{x_\sigma}f) &=&
\sup_{a\in{\bf a}^+,k\in K} {\rm e}^{Ra}|(H_{x_\sigma}f)(a,kM)|\\ &\le &
\sup_{a\in{\bf a}^+,k\in K} {\rm e}^{Ra}\tilde{p}_{R,1}(f) {\rm e}^{-{\rm Re }(x_\sigma)a}
\int_{a}^\infty {\rm e}^{({\rm Re }(x_\sigma)-R)b}db\\
&\le & C \tilde{p}_{R,1}(f)\ .
\end{eqnarray*}
If $R-{\rm Re }(x_\sigma)<0$, then
\begin{eqnarray*}\tilde{p}_{R,1}(H_{x_\sigma}f) &=&
\sup_{a\in{\bf a}^+,k\in K} {\rm e}^{Ra}|(H_{x_\sigma}f)(a,kM)|\\ &\le &
\sup_{a\in{\bf a}^+,k\in K} {\rm e}^{Ra}\tilde{p}_{R,1}(f){\rm e}^{-{\rm Re }(x_\sigma)a}
\int_{1}^a {\rm e}^{ ({\rm Re }(x_\sigma)-R)b}db\\
&\le & C \tilde{p}_{R,1}(f)\ .
\end{eqnarray*}
This finishes the construction of $H$ and hence the proof of Lemma
\ref{ssur}.\hspace*{\fill}$\Box$ \newline\noindent
\begin{lem}\label{logfin}
$S_{log}{{\cal E}^\prime}$ is ${\mbox{\rm Fin}}_\mu$-acyclic.
\end{lem}
{\it Proof.$\:\:$}
Using the generalization of the Helgason-Fourier transform to bundles
(see \cite{bunkeolbrich955}, 1.4.3 , Branson-Olafsson-Schlichtkrull
\cite{bransonolafssonschlichtkrull92})
and a result of Arthur \cite{arthur75}
we can identify the Schwartz space
$S_{-log}{\cal E}$ with
$$[{\cal S}(\imath {\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M V_\gamma)]^W\oplus
\oplus_{i=1}^r V_{i,\infty } \ .$$
Here $V_{i,\infty}$ are the spaces of smooth vectors of certain irreducible
discrete series representations, ${\cal S}(\imath{\bf a}^\ast)$
is the Schwartz space on $\imath{\bf a}^\ast$ and $W={\bf Z}_2$ is the Weyl group.
$\hat{\otimes}$ denotes the nuclear tensor product.
The operation of $W$ on ${\cal S}(\imath {\bf a}^*)\hat{\otimes} C^\infty(K\times_M
V_\gamma)$
is implemented by a family of unitary operators
$$A_\lambda:C^\infty(K\times_M V_\gamma)\rightarrow C^\infty(K\times_M
V_\gamma),\quad \lambda\in\imath {\bf a}^*\ ,$$
which are closely related to
Knapp-Stein intertwining operators of principal series representations.
The non-trivial element $w\in W$ acts by
$(wf)(-\lambda,k)=(A_\lambda f(\lambda,.))(k)$.
Dually, we can identify
$S_{log}{{\cal E}^\prime}$ with
$$[{\cal S}^\prime(\imath {\bf a}^*)\hat{\otimes} C^{-\infty}(K\times_M
V_\gamma)]^W\oplus \oplus_{i=1}^r V_{i,-\infty}^\prime \ .$$
On the Fourier image the operator $B$ acts as multiplication
by the polynomial $P(\lambda):=\lambda^2+\rho^2-\gamma(\Omega_M)-\mu$,
$\lambda\in\imath{\bf a}^*$, on the first
summand and as a scalar on $V_{i,-\infty}^\prime$.
It follows that ${\mbox{\rm Fin}}_\mu^1(V_{i,-\infty}^\prime)=0$.
In order to see that
$${\mbox{\rm Fin}}_\mu^1([{\cal S}^\prime(\imath {\bf a}^*)\hat{\otimes} C^{-\infty}(K\times_M
V_\gamma)]^W)=0$$
we show that the multiplication by $P$
is surjective on $[{\cal S}^\prime(\imath {\bf a}^*)\hat{\otimes}
C^{-\infty}(K\times_M V_\gamma)]^W$.
We consider the dual situation and show that the multiplication by
$P$ on $[{\cal S}(\imath {\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M V_\gamma)]^W$
is injective and has a closed range.
Injectivity is easy since $P(\lambda)$ is non-invertible at most on a set of
codimension $1$
in $\imath {\bf a}^*\times K/M$.
Note that $C^\infty(K\times_M V_\gamma)=\oplus_{\sigma\in \hat{M}}
\oplus_{h=1}^{[\gamma:\sigma]}
C^\infty(K\times_M V_{\sigma})$
and $P(\lambda)$ acts by the scalar polynomial
$P_\sigma(\lambda)=\lambda^2+\rho^2-\sigma(\Omega_M)-\mu$
on each summand ${\cal S}^\prime(\imath {\bf a}^*)\hat{\otimes} C^\infty(K\times_M
V_{\sigma})$.
Let $Z_\sigma$ be the set of zeros of $P_\sigma(\lambda)$.
Let $n_\sigma(\lambda)$ be the order of the zero of $P_\sigma$ at $\lambda \in
Z_\sigma$.
Then the range of the multiplication by $P_\sigma$ on ${\cal S}(\imath
{\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M V_{\sigma})$ consists of all
sections
$f\in {\cal S}(\imath {\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M V_{\sigma}) $
such that $f(\lambda,kM)$ has a zero of order $n_\sigma$ for all $\lambda \in
Z_\sigma$ and for all $k\in K$.
This vanishing condition is a closed condition.
The range of $P$ on $[{\cal S}(\imath {\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M
V_\gamma)]^W$
consists of all $f\in [{\cal S}(\imath {\bf a}^\ast)\hat{\otimes} C^\infty(K\times_M
V_\gamma)]^W$
such that $f_\sigma(\lambda,kM)$ has a zero of order $n_\sigma(\lambda)$ for
all $\lambda\in Z_\sigma$, $k\in K$ and $\sigma$.
This condition is again closed.
This finishes the proof of Lemma \ref{logfin} and hence of Theorem
\ref{fin}.\hspace*{\fill}$\Box$ \newline\noindent
\begin{theorem}\label{sur}
The operator $B:S_\infty{\cal E}\rightarrow S_\infty{\cal E}$
is surjective.
\end{theorem}
{\it Proof.$\:\:$}
Let $S_\infty{{\cal E}^\prime}(B)$ and $C$ be defined by the following exact sequence:
\begin{equation}\label{mmnnj}0\rightarrow S_\infty{{\cal E}^\prime}(B)\rightarrow
S_\infty{{\cal E}^\prime}\stackrel{B}{\rightarrow} S_\infty {{\cal E}^\prime}\rightarrow C\rightarrow 0\
.\end{equation}
\begin{lem}\label{dissur}
$C=0$.
\end{lem}
{\it Proof.$\:\:$}
We have $BC=0$ and $BS_\infty{{\cal E}^\prime}(B)=0$. Hence
${\mbox{\rm Fin}}_\mu^1(S_\infty{{\cal E}^\prime}(B) )=0$, ${\mbox{\rm Fin}}_\mu(S_\infty{{\cal E}^\prime}(B) )=S_\infty{{\cal E}^\prime}(B)
$, ${\mbox{\rm Fin}}_\mu^1(C)=0$, and
${\mbox{\rm Fin}}_\mu(C)=C$.
Thus applying ${\mbox{\rm Fin}}_\mu$ to the exact sequence of ${\mbox{\rm Fin}}_\mu$-acyclic spaces
(\ref{mmnnj}) we obtain the exact sequence
$$ 0\rightarrow S_\infty{{\cal E}^\prime}(B) \rightarrow {\mbox{\rm Fin}}_\mu
(S_\infty{{\cal E}^\prime})\stackrel{B}{\rightarrow} {\mbox{\rm Fin}}_\mu(S_\infty {{\cal E}^\prime})\rightarrow
C\rightarrow 0\ .$$
Let $S_\infty{{\cal E}^\prime}(B^k) :=\ker(B^k:S_\infty{{\cal E}^\prime}\rightarrow S_\infty{{\cal E}^\prime})$. Then
$${\mbox{\rm Fin}}_\mu(S_\infty{{\cal E}^\prime})=\lim_{\stackrel{\longrightarrow}{k}}S_\infty{{\cal E}^\prime}(B^k)
$$
and in order to show that $C=0$ it is enough to show that
\begin{equation}\label{loo}
\lim_{\stackrel{\longrightarrow}{k}}{\rm coker}(B:S_\infty{{\cal E}^\prime}(B^k)\rightarrow
S_\infty{{\cal E}^\prime}(B^k))=0\ .
\end{equation}
The space of $K$-finite vectors $S_\infty{{\cal E}^\prime}(B^k)_K$
is an admissible $({\bf g},K)$-module and $S_\infty{{\cal E}^\prime}(B^k)$
is its canonical distribution vector globalization (\cite{wallach92}, Ch. 11).
Since the globalization functor is exact (Wallach \cite{wallach92}, Ch. 11,
Casselman \cite{casselman89})
to show (\ref{loo}) is suffices to verify that
\begin{equation}\label{huhu}
\lim_{\stackrel{\longrightarrow}{k}}{\rm coker}(B:S_\infty{{\cal E}^\prime}(B^k)_K\rightarrow
S_\infty{{\cal E}^\prime}(B^k)_K)=0 \ .
\end{equation}
Since the multiplication by $B$ on
$({\cal U}({\bf g})\otimes_{{\cal U}({\bf k})}V_{\tilde{\gamma}})_K$ is injective we obtain
by $K$-type wise algebraic dualization
$${\rm coker}(B:{\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K\rightarrow
{\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K)=0\ .$$
Thus ${\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K$ is ${\mbox{\rm Fin}}_\mu$-acyclic.
Applying ${\mbox{\rm Fin}}_\mu$ to the exact sequence of ${\mbox{\rm Fin}}_\mu$-acyclic spaces
$$
0\rightarrow {\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K(B)\rightarrow
{\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K
\stackrel{B}{\rightarrow}{\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K \rightarrow 0
$$
we obtain the exact sequence
\begin{eqnarray*}
\lefteqn{0\rightarrow {\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K(B)\rightarrow
{\mbox{\rm Fin}}_\mu( {\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K)
\stackrel{B}{\rightarrow}}\\
&&\hspace{5cm}{\mbox{\rm Fin}}_\mu({\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K) \rightarrow 0\
{}.
\end{eqnarray*}
The exactness at the last place is equivalent to
$$
\lim_{\stackrel{\longrightarrow}{k}}{\rm coker}(B:{\mbox{\rm Hom}}_{{\cal U}({\bf k})}
({\cal U}({\bf g}),V_\gamma)_K(B^k)\rightarrow
{\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K(B^k))=0 \ .
$$
The claim (\ref{huhu}) now follows from
$S_\infty{{\cal E}^\prime}(B^k)_K\cong {\mbox{\rm Hom}}_{{\cal U}({\bf k})}({\cal U}({\bf g}),V_\gamma)_K(B^k)$
(compare \cite{bunkeolbrich947}, Lemma 3.5).
\hspace*{\fill}$\Box$ \newline\noindent
We have shown that $B:S_\infty {{\cal E}^\prime}\rightarrow S_\infty {{\cal E}^\prime}$
is surjective.
The $G$-invariant Hermitian scalar product on $E$ induces an inclusion
$S_\infty{\cal E}\subset S_\infty{{\cal E}^\prime}$.
Theorem \ref{sur} is now a consequence
of the following regularity result.
\begin{lem}\label{reg}
If $f\in S_\infty{{\cal E}^\prime}$, $Bf\in S_\infty{\cal E}$, then $f\in S_\infty{\cal E}$.
\end{lem}
{\it Proof.$\:\:$}
Let $\tilde{W}$ be a $G$-invariant fundamental solution of $B$, i.e.,
$B\tilde{W}=\tilde{W}B={\mbox{\rm id}}$ on $C_c^\infty(X,E)$.
Consider $X\times X$ with the diagonal $G$-action.
Let $\chi\in {}^GC^\infty(X\times X)$ be a $G$-invariant cut-off
function defined by $\chi(x,y)=\rho({\rm dist}(x,y))$, where $\rho\in C^\infty({\bf R})$,
$\rho(r)\in [0,1]$,
$\rho(r)=1$ for $r<1$ and $\rho=0$ for $r>2$.
Multiplying the distributional kernel of $\tilde{W}$ by $\chi$
we obtain a $G$-invariant parametrix $W$ of $B$.
It has finite propagation
(the support of $W\phi$ is contained in a $1$-neighbourhood
of the support of $\phi$ for all $\phi\in C_c^\infty(X,E)$)
and is applicable to arbitrary distributions in $C^{-\infty}(X,E)$.
Let $f \in S_\infty{{\cal E}^\prime}$ and $Bf=F\in S_\infty{\cal E}$.
Then $f=WF+Sf$, where $S$ is a $G$-invariant smoothing
operator on $E$.
We must show that $WF,Sf\in S_\infty{\cal E}$.
Let $L\in {\cal U}({\bf g})$. Then
$$(Sf)(Lg) = \langle f ,s(LgK,. ) \rangle \ ,$$
where $s$ is the integral kernel of $S$ and
$s(LgK,.)\in C_c^\infty(X, E)$.
For any $L_1\in {\cal U}({\bf g})$ and $R\in {\bf R}$ there is a $P\in {\bf R}$ depending on
$L_1,L$ such that
$$p_{R,L_1}(s(LgK, .)) = \sup_{h\in G} {\rm e}^{R{\rm dist}({\cal O},hK)} |s(LgK,L_1hK)|
\le C {\rm e}^{(P+R){\rm dist}({\cal O},gK)}\ ,$$
where $C$ may depend on $L ,L_1,R,P$ but not on $g$.
In fact, for all $A,B\in {\cal U}({\bf g})$ we have $\sup_{g,h\in G}|s(gA,hB)|<
\infty$ since
$S$ is $G$-invariant.
If $f\in S_R{{\cal E}^\prime}$ for some $R$, then there is a finite set
$\{L_1,\dots,L_r\}\subset {\cal U}({\bf g})$
such that for all $L\in {\cal U}({\bf g})$ there exists
$P\in {\bf R}$ with
\begin{eqnarray*}
p_{-(R+P),L}(Sf) &=& \sup_{g\in G} {\rm e}^{-(R+P){\rm dist}({\cal O},gK)} \langle f,
s(LgK,. ) \rangle \\
&\le& C(f) \sup_{g\in G, i=1,\dots, r}
{\rm e}^{-(R+P){\rm dist}({\cal O},gK)}p_{R,L_i}(s(LgK, .))\\&<&\infty \ .\end{eqnarray*}
This proves $Sf \in S_\infty{\cal E}$.
We now show that $WF\in S_\infty{\cal E}$.
Let $L\in {\cal U}({\bf g})$. We must show that for some $R\in{\bf R}$
$$p_{R,L}(WF)=\sup_{g\in G} {\rm e}^{R{\rm dist}({\cal O},gK)} |(WF)(LgK)|<\infty\ .$$
Let $w(gK,hK)$ be the distributional kernel of $W$.
The family $gK\mapsto w(LgK,.)$ is a smooth family of distributions
such that ${\mbox{\rm supp}}(w(LgK,.) )$ is contained in the unit ball in $X$ with center
at $gK$.
Thus we can write
$$(WF)(LgK)=\langle w(LgK,.) , F\rangle\ .$$
Since $W$ is $G$-invariant there is a finite set
$\{L_1,\dots,L_r\}\subset {\cal U}({\bf g})$ such that we have
$$|(WF)(LgK)|\le {\rm e}^{P{\rm dist}({\cal O},gK)} \sup_{h\in G, {\rm dist}(hK,gK)\le 1,
i=1,\dots, r}| F(hL_iK)| $$
for some $P\in R$ depending on $L$.
Hence there is another finite set
$\{L^\prime_1,\dots,L^\prime_{r^\prime}\}\subset {\cal U}({\bf g})$ and an exponent
$Q\in {\bf R}$ such that
\begin{eqnarray*}|(WF)(LgK)|&\le& {\rm e}^{Q{\rm dist}({\cal O},gK)} \sup_{h\in G,
{\rm dist}(hK,gK)\le 1, i=1,\dots, r^\prime}|F(L_i^\prime hK)|\\
&\le&
{\rm e}^{(Q-R){\rm dist}({\cal O},gK)}\sup_{i=1,\dots,
r^\prime}p_{R,L_i^\prime}(F)\end{eqnarray*}
Hence $p_{-(Q-R),L}(WF)<\infty$ if $R$ is small enough.
We conclude that $WF\in S_\infty{\cal E}$.
This finishes the proof of the Lemma and of Theorem \ref{sur}.
\hspace*{\fill}$\Box$ \newline\noindent
\section{De Rham complexes}\label{sec15}
In this section we prove the local acyclicity
of the weighted de Rham complex on $X$ twisted with
the functions of moderate growth on $G$,
the global acyclicity of the weighted de Rham complex on $X$ twisted with the
functions of moderate growth on $\Gamma\backslash G$, and
the ${\bf n}$-acyclicity of $S_\infty{\cal E}$ for a homogeneous vector bundle
$E\rightarrow X$.
A complex is called acyclic if it is exact in all non-zero degrees.
\begin{lem}\label{amdt}
For any vector bundle $E\rightarrow X$
the space $S_\infty{\cal E}$ is ${\bf n}$-acyclic.
\end{lem}
{\it Proof.$\:\:$}
Fix a basis $\{X_i\}_{i=1}^{\dim({\bf n})}$ of ${\bf n}$. Let $\{X^i\}$ be the
basis of ${\bf n}^*$ dual to $\{X_i\}$.
Let $Xn\in T_nN$, $X\in{\bf n}$, $n\in N$, be the fundamental vector at $n$
corresponding to
$-X$.
Let $I_p$ be the set of all multi-indices $\{1\le i_1<i_2<\dots < i_p\le
\dim({\bf n})\}$.
For $I\in I_p$ we define $X^In\in \Lambda^pT^*_nN$ by
$X^In(X_Jn)=\delta_{IJ}$, $\forall J\in I_p$. Here $X_Jn=X_{j_1}n\wedge\dots
\wedge X_{j_p}n$.
Now identify $N\times A\stackrel{\sim}{\rightarrow}X$, $(n,a)\mapsto naK$.
We also employ the ${\bf n}$-equivariant trivialization $G\times_K
V_\gamma=E\stackrel{\sim}{\rightarrow}N\times A\times V_\gamma$,
$[nak,v]\mapsto (n,a)\times \gamma^{-1}(k)v$.
We identify $C^\infty(N,\Lambda^*T^\ast N\otimes C^\infty(A)\otimes V_\gamma)$
with $\Lambda^*{\bf n}^*\otimes {\cal E}$ such that
$\omega\in C^\infty(N,\Lambda^pT^\ast N\otimes C^\infty(A)\otimes V_\gamma)$
corresponds
to $(X_I,na)\to \omega(na)(X_In)$, $I\in I_p$, $n\in N$, $a\in A$,
$\omega(na)(X_In)\in V_\gamma$.
Under this identification the ${\bf n}$-cohomology
complex of ${\cal E}$ becomes the de Rham complex of $N$
twisted with $C^\infty(A)\otimes V_\gamma$.
Using the trivialization $\{X^In\}_{I \in I_p}$ of the $p$-form
bundle we write \linebreak[4]$\omega(na)=\sum_{I\in I_p} \omega_I(na) X^In$.
The subspace $\Lambda^p{\bf n}\otimes S_\infty {\cal E}$ is identified with
\begin{eqnarray*}
\Lambda^p{\bf n}\otimes S_\infty {\cal E}&=&\{\omega\in C^\infty(N,\Lambda^pT^\ast
N\otimes C^\infty(A)\otimes V_\gamma)\\&&|\:\: \forall L_1\in{\cal U}({\bf n}),
L_2\in {\cal U}({\bf a})\:\:\: \exists R\in {\bf R}\\
&& s.t. \sup_{na\in NA, I\in I_p} {\rm e}^{-R{\rm dist}({\cal O},naK)}
|\omega_I(L_1n,L_2a)| < \infty\}\ .
\end{eqnarray*}
We define the contraction $\Psi_t$, $t\in[0,1]$, of $N$ by
$\Psi_t(n):=\exp(t\log(n))$.
Let $T_t:=\frac{d}{ds}_{|s=0}\Psi_{t+s}(n)\in T_{\Psi_t(n)}N$.
We set $H_t\omega:=\Psi_t^*(i_{T_t}\omega)$
and $H=\int_0^1 H_t dt$.
Then $$H:C^\infty(N,\Lambda^pT^\ast N\otimes C^\infty(A)\otimes V_\gamma)
\rightarrow C^\infty(N,\Lambda^{p-1}T^\ast N\otimes C^\infty(A)\otimes
V_\gamma)$$
is a zero homotopy of the de Rahm complex
$(C^\infty(N,\Lambda^*T^\ast N\otimes C^\infty(A)\otimes V_\gamma),d)$.
In order to prove the Lemma we have to show
that this zero homotopy is compatible with the subspaces
$ \Lambda^*\otimes S_\infty {\cal E}$. It is enough to show that
$$t\to H_t\in {\mbox{\rm Hom}}^{cont}_{\bf C}(\Lambda^p\otimes S_\infty {\cal E},\Lambda^{p-1}\otimes
S_\infty {\cal E})$$ is continuous.
Here we equip ${\mbox{\rm Hom}}^{cont}$ with the strong topology
(pointwise convergence).
We call a function $f$ on $N$ a polynomial, if $f(\exp(n))$ is a polynomial
on ${\bf n}$. Using the fact that $N$ is nilpotent, one can easily show that
$$(H_t\omega)(n)=\sum_{J\in I_{p-1},I\in
I_p}\Phi_{I,J}(t,n)\omega_I(\Psi_t(n))X^Jn\ , $$
where $n\to \Phi_{I,J}(t,n)$ are polynomial functions on $N$ (and in fact
also polynomials in $t$).
Let $L_r$ be the set of
tuples $\{i_1,\dots,i_{\dim({\bf n})}\}$, $i_k\in {\bf N}_0$, with
$\sum_{k=1}^{\dim({\bf n})}i_k\le r$.
Let $X(l):=X_1^{l_1}\dots X_{\dim({\bf n})}^{l_{\dim({\bf n})}}\in {\cal U}({\bf n})$,
$l\in L_r$.
By $X(l)n$ we denote the corresponding differential operator $(X_1n)^{l_1}
\dots (X_{\dim({\bf n})}n)^{l_{\dim({\bf n})}}$.
If $L_1\in {\cal U}({\bf n})$, $\deg(L_1)=r$, $L_2\in {\cal U}({\bf a})$, then also
$$(H_t\omega)_J(L_1n,L_2a)=\sum_{l\in L_r,I\in I_p} \Phi_{l,I,J}(t,n)
\omega_I(X(l)\Psi_t(n),L_2a)$$
with polynomial functions $\Phi_{l,I,J}(t,n)$.
We obtain the estimate
$$\sup_{J\in I_{p-1}}|(H_t\omega)_J(L_1n,L_2a)|\le C (1+|\log(n)|)^P \sup_{l\in
L_r,I\in I_p} |\omega_I(X(l)\Psi_t(n),L_2a)|\ ,$$
where $P$ is sufficiently large.
Note that ${\rm e}^{Q{\rm dist}({\cal O},naK)} \ge c (1+|\log(n)|)^P$ for some $Q,c>0$
and all $na\in NA$.
We also have ${\rm dist}({\cal O},naK)\ge {\rm dist}({\cal O},\Psi_t(n)aK)$ for all $t\in
[0,1]$.
Fix $R\in {\bf R}$.
Then for all $\omega$ with
$$\sup_{l\in L_r,na\in NA, I\in I_p} {\rm e}^{-R{\rm dist}({\cal O},naK)}
|\omega_I(X(l)n,L_2a)|=:S(\omega) < \infty$$
we have $$ \sup_{na\in NA, J\in I_{p-1}} {\rm e}^{-(R+Q){\rm dist}({\cal O},naK)}
|(H_t\omega)_J(L_1n,L_2a)| < CS(\omega)$$
with $C<\infty$ independent of $\omega$.
These estimates imply that if $\omega\in \Lambda^p{\bf n}\otimes S_u{\cal E}$, $u\in
{\cal I}$,
then $H_t\omega \in \Lambda^{p-1}{\bf n}\otimes S_{u+Q}{\cal E}$ and depends
continuously on $t\in[0,1]$.
This proves the required continuity of $H_t$.
Thus $H$ provides a zero homotopy of the ${\bf n}$-cohomology
complex for $S_\infty {\cal E}$, and the lemma is proved.
\hspace*{\fill}$\Box$ \newline\noindent
Fix an Iwasawa decomposition $G=NAK$ and identify $X\cong NA$. Using these
coordinates
we attach a boundary $\partial X=N\times\{\infty\}$
to $X$ obtaining $\bar{X}$. Note that this boundary is different from
the geodesic boundary.
The motivation for this definition is that small neighbourhoods
of $x\in\partial X$ look like small neighbourhoods
of points in the boundary of the Borel-Serre compactification
of quotients $Y=\Gamma\backslash X$ of finite volume.
We want to twist the de Rham complex of $X$ with the space of functions
of moderate growth on $G$. Let $\pi:X\times G\rightarrow X$ denote the
projection onto
the first factor.
Let $L^p:=\Lambda^pT^*X$.
We define
\begin{eqnarray*}
S_\infty{\cal L}^p[G]&:=&\{\omega\in C^\infty(X\times G,\pi^*L^*)\\
&&|\:\forall L_1,L_2\in {\cal U}({\bf g})\:\:\: \exists R,Q\in{\bf R}\\
&&s.t.\:\: \sup_{g,h\in G}{\rm e}^{-R{\rm dist}({\cal O},gK)}{\rm e}^{-Q{\rm dist}({\cal O},hK)}
|\omega(L_1g,L_2h)|<\infty
\} \ ,
\end{eqnarray*}
where we consider $\omega$ as a function from $G\times G$ with values in
$\Lambda^pT_{\cal O}^*X $.
There is a natural sheafification $\underline{S_\infty{\cal L}^\ast[G]}$ on
$\bar{X}$
such that for an open set $U\subset \bar{X}$ the vector space
$\underline{S_\infty{\cal L}^\ast[G]}(U)$ consists of those forms $\omega\in
C^\infty(U\times G,\pi^* L^\ast)$
which satisfy
$$\sup_{g\in U,h\in G}
|\omega(L_1g,L_2h)|{\rm e}^{-R{\rm dist}({\cal O},gK)}{\rm e}^{-Q{\rm dist}({\cal O},hK)} <\infty$$
for any $L_1,L_2\in {\cal U}({\bf g})$ and appropriate $Q,R\in {\bf R}$ which may depend on
$L_1,L_2$.
Let $d$ be the differential of the de Rham complex acting trivially with
respect to the second variable $g\in G$.
\begin{lem}\label{mmarrt}
The complex of sheaves $(\underline{S_\infty{\cal L}^\ast[G]},d)$
is acyclic.
\end{lem}
{\it Proof.$\:\:$}
Let $x\in \bar{X}$. Then we have to show that the complex of germs
$(\underline{S_\infty{\cal L}^\ast[G]}_x,d)$ is acyclic.
If $x\in X$, then we employ the standard homotopy formula
associated to the radial contraction of small balls in $X$ with center at $x$.
We leave to the reader to verify that the zero homotopy is compatible with the
growth conditions with respect to second variable.
We discuss a similar problem in the proof of Lemma \ref{mglobal}.
It remains to consider $x\in \partial X$. Without loss of generality
we can assume $x=e\times\{\infty\}$, where $e\in N$ is the identity.
Let $U_i\subset {\bf n}$, $i\in {\bf N}$, be a fundamental
system of balls around $0$. Then $W_i:=\exp(U_i) (i,\infty]\subset NA$
is a fundamental system of neighbourhoods of $x$.
Here we identified ${\bf a}\cong{\bf R}$ such that
$(i,\infty]=\exp((i,\infty))\cup\{\infty\}$.
Let $da$ be the one-form dual to the fundamental vector field $H^*$ on $A$
corresponding to
$H\in {\bf a}^+$ with $|H|=1$.
We decompose forms $\omega$ in ${\cal L}^p$ as $\omega=\omega_1\oplus da\wedge
\omega_2$,
where $i_{H^*}\omega_j=0$, $j=1,2$, and $i_{H^*}$ is the insertion of $H^*$.
One can check that
$$\underline{S_\infty{\cal L}^p[G]}(W_i)=V^p(W_i)\oplus da\wedge V^{p-1}(W_i)\ ,$$
where
\begin{eqnarray*}
V^p(W_i)&:=&\{\omega\in C^\infty(U_i\times (i,\infty)\times
G,\pi_1^*\Lambda^pT^*U_i)|\\
&&\forall L_1\in{\cal U}({\bf n}),L_2\in{\cal U}({\bf a}),L_3\in{\cal U}({\bf g})\:\:\:\exists
R,Q\in{\bf R}\\
&&s.t.\:\:\:\sup_{a\in(i,\infty),h\in G} a^{-R} {\rm e}^{-Q{\rm dist}({\cal O},hK)}
|\omega(L_1n,L_2a,L_3h)|<\infty\}\ ,
\end{eqnarray*}
and $\pi_1:U_i\times (i,\infty)\times G\rightarrow U_i$ is the projection onto
the first factor.
The complex $(\underline{S_\infty{\cal L}^*[G]}(W_i),d)$
is the total complex
of a double complex. The latter consists of two rows equal
$(V^*(W_i),d)$, where $d$ is the differential of the de Rham complex
of $U_i$. The vertical differential of the double complex
is the differentiation along the $A$-direction given by $\pm H$.
The balls $U_i$ can be contracted radially.
The associated zero homotopy of the de Rham complex of $U_i$
extends to $V^*(W_i)$.
Again we leave the verification to the reader (see also the proof of Lemma
\ref{amdt}).
Thus the rows $(V^*(W_i),d)$ of the double complex are acyclic.
The zeroth horizontal cohomology is equal to ${}^{\bf n} V^0(W_i)$,
the functions in $V^0(W_i)$ that do not depend on $n\in U_i$.
To finish the proof of the Lemma we must show that
$H:{}^{\bf n} V^0(W_i)\rightarrow {}^{\bf n} V^0(W_i)$
is surjective.
In fact the equation $Hf=g$, $g\in {}^{\bf n} V^0(W_i)$, can explicitly be
solved by integration
such that $f\in {}^{\bf n} V^0(W_i)$. Set $$f(n,a,h)=\int_i^a g(n,b,h) db\ .$$
\hspace*{\fill}$\Box$ \newline\noindent
Let $\pi:X\times \Gamma\backslash G$ be the projection onto the first factor.
We define
\begin{eqnarray*}
S_\infty{\cal L}^p[\Gamma\backslash G]&:=&\{\omega\in C^\infty(X\times
\Gamma\backslash G ,\pi^*L^*)\\
&&|\:\forall L_1,L_2\in {\cal U}({\bf g})\:\:\: \exists R,Q\in{\bf R}\\
&&s.t.\:\: \sup_{g,h \in G}{\rm e}^{-R{\rm dist}({\cal O},gK)}
{\rm e}^{-Q{\rm dist}_Y(\Gamma{\cal O},\Gamma hK)} |\omega(L_1g,L_2h)|<\infty
\} \ ,
\end{eqnarray*}
where ${\rm dist}_Y(\Gamma{\cal O},\Gamma hK)$ is the distance of $\Gamma hK$ from
$\Gamma {\cal O}$ in $Y:=\Gamma\backslash X$.
\begin{lem}\label{mglobal}
The de Rham complex
$(S_\infty{\cal L}^*[\Gamma\backslash G ],d)$ is acyclic.
\end{lem}
{\it Proof.$\:\:$}
As in the proof of Lemma \ref{mmarrt} the de Rham complex
$(S_\infty{\cal L}^*[\Gamma\backslash G ],d)$ is the total complex of a double
complex
consisting of two rows each equal to the ${\bf n}$-cohomology complex
$(\Lambda^*{\bf n}\otimes S_\infty{\cal L}^0[\Gamma\backslash G ],d)$.
Let $\pi_1:N\times A\times \Gamma\backslash G \rightarrow N$ be the projection
onto the first
factor.
We identify
$\Lambda^p{\bf n}\otimes S_\infty{\cal L}^0[\Gamma\backslash G ]$ with
\begin{eqnarray*}
&&\{\omega\in C^\infty(N\times A\times \Gamma\backslash G
,\pi_1^*\Lambda^pT^\ast N)
\\&&|\:\: \forall L_1\in{\cal U}({\bf n}), L_2\in {\cal U}({\bf a}),L_3\in{\cal U}({\bf g})\:\:\:
\exists R,Q\in {\bf R}\\
&& s.t. \sup_{na\in NA,g\in G , I\in I_p}
{\rm e}^{-R{\rm dist}({\cal O},naK)}{\rm e}^{-Q{\rm dist}_Y(\Gamma {\cal O},\Gamma gK)}
|\omega_I(L_1n,L_2a,L_3g)| < \infty\}\ .
\end{eqnarray*}
We show that the complex $(\Lambda^*{\bf n}\otimes S_\infty{\cal L}^0[\Gamma\backslash
G ],d)$ is acyclic.
In the proof of Lemma \ref{amdt} we constructed a zero homotopy $H=\int_0^1
H_t$, where
$(H_t\omega)(n,a,g)=\Psi_t^\ast(i_{T_t}\omega(\Psi_t(n),a,g))$.
We claim that $H:\Lambda^p{\bf n}\otimes S_\infty{\cal L}^0[\Gamma\backslash G
]\rightarrow \Lambda^{p-1}{\bf n}\otimes S_\infty{\cal L}^0[\Gamma\backslash G ]$.
The same discussion as in the proof of Lemma \ref{amdt} leads to the following
estimate.
Fix $L_1\in{\cal U}({\bf n})$, $L_2\in{\cal U}({\bf a})$, $L_3\in{\cal U}({\bf g})$.
Let $r:=\deg(L_1)$.
For all
$\omega$ with
$$\sup_{l\in L_r,na\in NA,g\in G , I\in I_p} {\rm e}^{-P{\rm dist}_Y(\Gamma {\cal O},\Gamma
gK)}{\rm e}^{-R{\rm dist}({\cal O},naK)} |\omega_I(X(l)n,L_2a,L_3g)|=:S(\omega) < \infty$$
we have
$$ \sup_{na\in NA,g\in G ,J\in I_{p-1}}
{\rm e}^{-P{\rm dist}_Y(\Gamma {\cal O},\Gamma gK)}{\rm e}^{-(R+Q){\rm dist}({\cal O},naK)}
|(H_t\omega)_J(L_1n,L_2a,L_3g)| < CS(\omega)$$
with $C<\infty$ independent of $\omega$.
This easily implies the claim.
Hence the rows of the double complex are acyclic
and their zeroth cohomology is equal to
\begin{eqnarray*}
V&:=&\{f\in C^\infty(A\times \Gamma\backslash G )\:\: |\:\: \forall L_2\in
{\cal U}({\bf a}), L_3\in{\cal U}({\bf g})\:\:\: \exists R,Q\in {\bf R}\\
&&\hspace{3cm} s.t. \sup_{a\in A,g\in G } {\rm e}^{-R|\log(a)|}
{\rm e}^{-Q{\rm dist}_Y(\Gamma {\cal O},\Gamma gK)} |f(L_2a,L_3g)| < \infty\}\ .
\end{eqnarray*}
The vertical differential given by $H:V\rightarrow V$ is surjective.
In fact, let $f\in V$ and set
$$F(a,g):=\int_1^a f(b,g)db\ .$$
Then $HF=f$ and $F\in V$.
This proves the lemma. \hspace*{\fill}$\Box$ \newline\noindent
\newcommand{{{\cal HC}(\gaaa,K)}}{{{\cal HC}({\bf g},K)}}
\section{The standard resolution}
For the convenience of the reader we repeat here the construction
of the standard resolution given in \cite{bunkeolbrich947}.
Let $(\pi,V_{\pi,K})\in{{\cal HC}(\gaaa,K)}$ be a Harish-Chandra module.
Then $V_{\pi,K}$ decomposes into a direct sum of joint
generalized eigenspaces of ${\cal Z}({\bf g})$.
Since the summands can be treated separately, without loss of
generality we can assume
that there exist $\mu\in {\bf C}$ and $k\in{\bf N}$ such
that $B:=(\Omega-\mu)^k\in {\mbox{\rm Ann}}(V_{\pi,K})$, i.e., $BV_{\pi,K}=0$.
Let $W$ be a finite-dimensional $K$-stable subspace of
the dual $V_{\tilde\pi,K}$ of $V_{\pi,K}$ in the category
${{\cal HC}(\gaaa,K)}$, which generates $V_{\tilde\pi,K}$ as a
${\cal U}({\bf g})$-module. Let $E_0\rightarrow X$ be
the homogeneous vector bundle $G\times_K \tilde W$
and ${\cal E}_0$ be the space of its smooth sections.
Using any globalization $ V_\pi $ of $V_{\pi,K}$ (i.e. a
representation of $G$ such that $ V_\pi=V_{\pi,K}$)
we can define an embedding
$$
i: V_{\pi,K}\hookrightarrow {\cal E}_0\cong [C^\infty(G)\otimes \tilde W]^K
$$
that is characterized by
$$\langle i(v)(g),w\rangle:= \langle w,\pi(g^{-1})v\rangle,
\qquad v\in V_{\pi,K},w\in W,g\in G.$$
In fact $i$ maps into $S_\infty{\cal E}_0$
and the closure of $i(V_{\pi,K})$ in $S_\infty{\cal E}_0$ is contained in
$S_\infty{\cal E}_0(B)$ and constitutes the distribution vector
globalization $V_{\pi,-\infty}$
of $V_{\pi,K}$ (Wallach \cite{wallach92}, Ch. 11, Casselman
\cite{casselman89}).
We will also consider the space $V_{\pi,for}:=V_{\tilde\pi, K}^*$
of formal power series vectors
of $V_{\pi,K}$. There is an exact functor from ${{\cal HC}(\gaaa,K)}$ to the
category of (not necessarily $K$-finite) $({\bf g},K)$-modules
which sends $V_{\pi,K}$ to $V_{\pi,for}$. Note that
$V_{\pi,for}=\prod_{\gamma\in\hat K} V_{\pi,K}(\gamma)$.
For homogeneous vector bundles $E$ and $F$ on $X$ we denote by $D(E,F)$ the set
of $G$-invariant differential operators $D:{\cal E}\rightarrow {\cal F}$.
\begin{prop}\label{1.zeile}
There exist homogeneous vector bundles $E_1,E_2,\dots$ on $X$ and
$G$-\linebreak[4] invariant differential operators $D_i\in D(E_i,E_{i+1})$,
$i=0,1,\dots$, such that the embedding $i:V_{\pi,-\omega}\hookrightarrow
{\cal E}_0(B)$ can be
extended to a (possibly infinite) exact sequence
\begin{equation}\label{*}
0\rightarrow V_{\pi,-\infty}\stackrel{i}{\rightarrow}S_\infty{\cal E}_0(B)
\stackrel{D_0}{\rightarrow}S_\infty{\cal E}_1(B)
\stackrel{D_1}{\rightarrow}S_\infty{\cal E}_2(B)
\stackrel{D_2}
{\rightarrow}\dots\ .
\end{equation}
This sequence remains to be exact on the level of formal power
series :
\begin{equation}\label{*+}
0\rightarrow V_{\pi,for}\stackrel{i}{\rightarrow}{\cal E}_0^{for}(B)
\stackrel{D_0}{\rightarrow}{\cal E}_1^{for}(B)
\stackrel{D_1}{\rightarrow}{\cal E}_2^{for}(B)
\stackrel{D_2}{\rightarrow}\dots\ .
\end{equation}
\end{prop}
{\it Proof.$\:\:$}
Let ${\cal Z}(E)$ be the image of ${\cal Z}({\bf g})$ in $D(E,E)$.
For any vector bundle $E\rightarrow X$ the ${\bf C}[B]$-module ${\cal Z}(E)$
is finitely generated (\cite{bunkeolbrich947}, Lemma 2.3).
\begin{lem}\label{isntit}
For any vector bundle $E\rightarrow X$ we have ${\cal E}(B)_K\in{{\cal HC}(\gaaa,K)}$.
\end{lem}
{\it Proof.$\:\:$}
Let $(\gamma,V_\gamma)$ be the finite dimensional representation
of $K$ corresponding to $E$ and $(\tilde{\gamma},V_{\tilde{\gamma}})$
be its dual.
We consider the $K$-equivariant embedding
$$i:V_{\tilde{\gamma}}\hookrightarrow \widetilde{{\cal E}(B)_K}$$
defined by
$$i(\tilde{v})(f):=\langle\tilde{v},f(e)\rangle\ ,$$
where we identify the fibre of $E$ at $e=[K]$
with $V_\gamma$. Let $T:={\cal U}({\bf g})(i(V_{\tilde{\gamma}}))$.
For any $t\in T$ the dimension of ${\cal Z}({\bf g})t$ can be estimated
by the dimension of a generating subspace of the
${\bf C}[B]$-module ${\cal Z}(E)$. Thus $T$ is a locally
$Z({\bf g})$-finite and finitely generated ${\cal U}({\bf g})$-module. By a theorem of
Harish-Chandra (\cite{wallach88}, 3.4.7), $T\in{{\cal HC}(\gaaa,K)}$. The canonical map
${\cal E}(B)_K\rightarrow \tilde{T}$
is injective by the analyticity of solutions of the equation $Bf=0$. In fact,
an element in the kernel of this map
would have a vanishing Taylor series at $e$.
We obtain that $T\hookrightarrow \widetilde{{\cal E}(B)_K}$ is surjective.
Thus $T=\widetilde{{\cal E}(B)_K}$ and ${\cal E}(B)_K\in{{\cal HC}(\gaaa,K)}$ since
the dual of a Harish-Chandra module is a Harish-Chandra module, too
(\cite{wallach88}, 4.3.2).
\hspace*{\fill}$\Box$ \newline\noindent
\begin{lem}\label{hno}
Let $V_{\pi,K}$ be a Harish-Chandra submodule of ${\cal E}(B)_K$.
Then there exist a homogeneous vector bundle $F$ and an
operator $D\in D(E,F)$ such that $\ker D \cap {\cal E}(B)_K=
V_{\pi,K}$. We also have $\ker D \cap S_\infty{\cal E}(B)=V_{\pi,-\infty}$.
\end{lem}
{\it Proof.$\:\:$}
According to the proof of Lemma \ref{isntit}
there is a surjection
$$ {\cal U}({\bf g})\otimes_{{\cal U}({\bf k})} V_{\tilde\gamma}
\rightarrow \widetilde{{\cal E}(B)_K}\ .$$ Let $W$ be a finite-dimensional
$K$-stable generating subspace of the Harish-Chandra module
$V_{\pi,K}^\perp\subset\widetilde{{\cal E}(B)_K}$. Then we choose a
$K$-equivariant map $\alpha$ such that the following diagram
$$\begin{array}{ccc}
W&\stackrel{\alpha}{\longrightarrow}&{\cal U}({\bf g}) \otimes_{{\cal U}({\bf k})}
V_{\tilde\gamma}\\
\downarrow&&\downarrow\\
V_{\pi,K}^\perp&\longrightarrow &\widetilde{{\cal E}(B)_K}
\end{array}$$
commutes. This is possible since
${\cal U}({\bf g})\otimes_{{\cal U}({\bf k})}V_{\tilde\gamma}$ is
$K$-semisimple.
We set $F:=G\times_K \tilde W$. The map $\alpha$ can be considered
as an element of
$$[{\cal U}({\bf g})\otimes_{{\cal U}({\bf k})} V_{\tilde\gamma}\otimes
\tilde W]^K\cong [{\cal U}({\bf g})\otimes_{{\cal U}({\bf k})}
{\mbox{\rm Hom}} (V_{\gamma},\tilde W)]^K\ .$$
The latter space is canonically isomorphic to $D(E,F)$ via the
right regular representation $R$ of ${\cal U}({\bf g})$ on $C^\infty(G)\otimes
V_\gamma$. Thus $\alpha$ defines an element
$D\in D(E,F)$. If $\alpha(w)=\sum X_i\otimes v_i$, then
$$ \langle w,Df\rangle_F = \sum \langle v_i, R_{X_i}f\rangle_E \in
C^\infty(G),\ w\in W,v_i\in V_{\tilde\gamma}, X_i\in {\cal U}({\bf g})\ .$$
Let $f\in {\cal E}(B)$, $X\in{\cal U}({\bf g})$ and $w\in W$.
Then we have
\begin{eqnarray}\label{mufti}
\langle w, L_XDf(1)\rangle_F&=&\langle w, DL_Xf\rangle_F\nonumber\\
&=&\sum\langle v_i,R_{X_i}L_Xf(1)\rangle_E\\
&=&\langle w ,L_Xf \rangle_{{\cal E}(B)}\nonumber\\
&=& \langle L_{X^{op}}w ,f \rangle_{{\cal E}(B)}\ ,\nonumber
\end{eqnarray}
where $X\rightarrow X^{op}$ is the anti-automorphism of
${\cal U}({\bf g})$ induced by the multiplication with $-1$ on
${\bf g}$. By construction $Df=0$ iff the left hand side of (\ref{mufti})
vanishes for all $X\in{\cal U}({\bf g})$ and $w\in W$,
while $f\in V_{\pi,-\infty}$ iff the right hand side
does. The lemma follows.\hspace*{\fill}$\Box$ \newline\noindent
In order to construct the bundles
$E_i$ and operators $D_i$ of Proposition \ref{1.zeile} we
iterate Lemma \ref{hno}. $D_i({\cal E}_i(B)_K)$ is a Harish-Chandra submodule of
${\cal E}_{i+1}(B)_K$.
Therefore we find a bundle $E_{i+2}$ and an operator $D_{i+1}\in
D(E_{i+1},E_{i+2})$ such that $\ker D_{i+1}\cap{\cal E}_{i+1}(B)_K =
D_i({\cal E}_i(B)_K)$. We obtain an exact sequence of Harish-Chandra modules
$$ 0\rightarrow V_{\pi,K}\stackrel{i}{\rightarrow}{\cal E}_0(B)_K
\stackrel{D_0}{\rightarrow}{\cal E}_1(B)_K \stackrel{D_1}{\rightarrow}{\cal E}_2(B)_K
\stackrel{D_2}{\rightarrow}\dots\ .$$
Applying the distribution vector globalization functor (which
is exact) we end up with (\ref{*}). Analogously, we want to obtain (\ref{*+})
by taking
formal power series vectors. This is possible since for any homogeneous vector
bundle $E$ we have
$$ ({\cal E}(B)_K)_{for}={\cal E}^{for}(B)$$
(\cite{bunkeolbrich947}, Lemma 3.5).
The Proposition \ref{1.zeile} provides a resolution
of $V_{\pi,-\infty}$ by spaces $S_\infty{\cal E}_i(B)$.
We now employ a Koszul complex construction in order to get rid of
the eigenspaces. We recall the following fact from \cite{bunkeolbrich947}.
\begin{lem}\label{mlem}
Let $E$, $F$ be homogeneous vector bundles on $X$ and
$A\in D(E,F)$
such that $A {\cal E}(B) =0$.
Then $A = H B$ for some $H\in D(E,F)$.
\end{lem}
Let $V_{\pi,K},E_i,D_i$ be constructed as in Proposition \ref{1.zeile}.
\begin{prop}\label{stare}
There exist $H_i\in D(E_i,E_{i+2})$, $i\ge 0$, making the
following into an exact complex:
\begin{equation}\label{**}
0\rightarrow V_{\pi,-\infty}\rightarrow S_\infty{\cal E}_0
\stackrel{{\scriptsize \left(\begin{array}{c}D_0\\B\end{array}\right)}}
{\longrightarrow}
\begin{array}{c}S_\infty{\cal E}_1\\ \oplus\\ S_\infty{\cal E}_0\end{array}
\stackrel{ \left({\scriptsize\begin{array}{cc}D_1&H_0\\-B&D_0\end{array}}
\right)}{\longrightarrow}
\begin{array}{c}S_\infty{\cal E}_2\\ \oplus\\S_\infty{\cal E}_1\end{array}
\stackrel{ \left({\scriptsize\begin{array}{cc}
D_2&H_1\\B&D_1\end{array}}\right)}{\longrightarrow}\dots\ .
\end{equation}
\end{prop}
We shall call (\ref{**}) a standard resolution of $V_{\pi,-\infty}$.\\
{\it Proof.$\:\:$}
In order to construct the operators $H_i$ we apply
Lemma \ref{mlem} for $A=D_{i+1}D_i$.
The exactness of (\ref{**}) is easily reduced to the exactness of
(\ref{*}) and the surjectivity
of $B:S_\infty{\cal E}_i\rightarrow S_\infty{\cal E}_i$ proved in Theorem \ref{sur}.
\hspace*{\fill}$\Box$ \newline\noindent
\section{${\bf n}$-cohomology
}
Let $B=(\Omega_G-\lambda)^l$ for some $\lambda\in{\bf C}$, $l\in{\bf N}$, where
$\Omega_G$
is the Casimir operator of $G$,
and $S_\infty{\cal E}(B)=\{f\in S_\infty{\cal E}\:|\: Bf=0\}$.
\begin{lem}\label{lez}
We have
$$H^p({\bf n},S_\infty{\cal E}(B))=0,\quad \forall p\ge 1.$$
\end{lem}
{\it Proof.$\:\:$}
By Theorem \ref{sur} and Lemma \ref{amdt}
$$0 \rightarrow S_\infty{\cal E}(B)\rightarrow S_\infty
{\cal E}\stackrel{B}{\rightarrow}S_\infty {\cal E}\rightarrow 0$$
is an ${\bf n}$-acyclic resolution of $S_\infty{\cal E}(B)$.
Taking ${\bf n}$-invariants we obtain the complex
\begin{equation}\label{ueeqq1}0\rightarrow {}^{\bf n} S_\infty {\cal E}(B)\rightarrow
S_\infty C^\infty(A)\otimes V_\gamma\stackrel{{}^{\bf n} B}{\rightarrow}
S_\infty C^\infty(A)\otimes V_\gamma\rightarrow 0\ .\end{equation}
Here ${}^{\bf n} B$ is the restriction of $B$
to the subspace of ${\bf n}$-invariant vectors.
It is a second order translation invariant differential operator on $A$.
The complex (\ref{ueeqq1}) is again exact and the Lemma follows.
\hspace*{\fill}$\Box$ \newline\noindent
Let $(\pi,V_{\pi,K}) \in{{\cal HC}(\gaaa,K)}$.
Recall that $H^*({\bf n},V_{\pi,-\infty})$ carries a natural $MA$-module
structure.
\begin{theorem}\label{haha}
The inclusion $V_{\pi,-\infty}\hookrightarrow V_{\pi,for}$
induces an isomorphism
$$H^p({\bf n},V_{\pi,-\infty})\stackrel{\sim}{\longrightarrow}
H^p({\bf n},V_{\pi,for})\ .$$
Moreover
$H^p({\bf n},V_{\pi,-\infty})=H^p({\bf n},V_{\pi,-\omega})= H^p({\bf n},V_{\pi,for})$
and all spaces are finite dimensional.
The ${\bf n}$-cohomology of $V_{\pi,-\infty}$ satisfies Poincar\'e duality
\begin{equation}\label{ukk1}
H^p({\bf n},V_{\pi,-\infty})^*\cong H^{\dim({\bf n})-p}({\bf n},
V_{\tilde{\pi},\infty})\otimes\Lambda^{\dim{{\bf n}}}{\bf n} .
\end{equation}
We also have $$ H^p({\bf n},V_{\pi,\infty})\cong H^p({\bf n},V_{\pi,K}).
$$
\end{theorem}
{\it Proof.$\:\:$}
By (\cite{bunkeolbrich947}, Lemma 2.3 and Proposition 4.1), Lemma \ref{amdt}
and Proposition \ref{1.zeile} the cohomology $H^p({\bf n},V_{\pi,*})$ for
$*=-\infty$, $for$ is isomorphic to the cohomology of the subcomplex of
${\bf n}$-invariants of (\ref{*}), (\ref{*+}), respectively.
Hence the following
lemma implies the first assertion of the theorem.
\begin{lem}\label{hihi}
For any homogeneous vector bundle $E\rightarrow X$ associated to $V_\gamma$ we
have
$$ {}^{\bf n} S_\infty {\cal E}(B)={}^{\bf n}{\cal E}^{for}(B) \ .$$
\end{lem}
{\it Proof.$\:\:$}
The ${\cal U}({\bf a})$-module
$${}^{\bf n} {\cal E}^{for}(B)\cong (\widetilde {{\cal E}(B)_K}/{\bf n}(\widetilde
{{\cal E}(B)_K}))^*$$
is finite dimensional (see \cite{wallach88}, Ch.4). Therefore it splits into
generalized weight spaces ${}^{\bf n} {\cal E}^{for}(B)_\mu, \mu\in{\bf a}^*_{\bf C}$. $f\in
{}^{\bf n} {\cal E}^{for}(B)_\mu$,
considered as a formal power series on ${\bf a}$,
satisfies the differential equation
\begin{equation}\label{muuh}(H+\mu(H))^kf=0\quad \forall
H\in{\bf a}\end{equation}
for a certain $k\in {\bf N}$.
The solutions of (\ref{muuh}) have the form
$$P(H){\rm e}^{-\mu(H)},\qquad P\in S({\bf a}^*)\otimes V_\gamma\ .$$
They extend to smooth ${\bf n}$-invariant sections in ${}^{\bf n} S_\infty {\cal E}(B)$.
\hspace*{\fill}$\Box$ \newline\noindent
The proof of the remaining assertions of the theorem is parallel
to the proofs of the corresponding facts in \cite{bunkeolbrich947}.
\hspace*{\fill}$\Box$ \newline\noindent
\section{$S_\infty{\cal E}$ is $\Gamma$-acyclic}
Let $\Gamma\subset G$ be a discrete, torsion-free subgroup of finite covolume.
Let $E\rightarrow X$ be a $G$-homogeneous vector bundle
and $S_\infty{\cal E}$ the space of its sections of moderate growth.
\begin{theorem}\label{gaz}
$S_\infty{\cal E}$ is $\Gamma$-acyclic, i.e.,
$$H^p(\Gamma,S_\infty{\cal E})=0\quad\forall p\ge 1\ .$$
\end{theorem}
{\it Proof.$\:\:$}
We first consider the space of functions of moderate growth
$S_\infty C^\infty(G)$ on $G$ defined by
$$S_\infty C^\infty(G):=\{f\in{\bf C}^\infty(G)\:|\: \forall L\in{\cal U}({\bf g})\:\exists
R\in{\bf R}\:s.t. \:
\sup_{g\in G} {\rm e}^{-R{\rm dist}(gK,{\cal O})} |f(Lg)| <\infty\}\ .$$
As a topological vector space $S_\infty C^\infty(G)$ is a limit of Fr\'echet
spaces.
\begin{prop}\label{acc}
$S_\infty C^\infty(G)$ is $\Gamma$-acyclic.
\end{prop}
{\it Proof.$\:\:$}
$H^\ast(\Gamma,S_\infty C^\infty(G))$ is the cohomology of the
de Rham complex
of $Y=\Gamma\backslash X$ twisted with
the flat bundle associated to the $\Gamma$-module
$S_\infty C^\infty(G)$.
In greater detail let $L^\ast:=\Lambda^\ast T^\ast X$ and
${\cal C}^\ast:=C^\infty(X,L^\ast\otimes S_\infty C^\infty(G))$.
Moreover, let $d:{\cal C}^\ast\rightarrow {\cal C}^{\ast+1}$ denote the differential of
the de Rham complex.
The complex $({\cal C}^\ast,d)$ is a complex of $\Gamma$-modules.
If we view $\omega\in{\cal C}^p$ as a function on $G$ with values in ${\cal L}^p$,
then the action of $\gamma\in\Gamma$ on $\omega$ is given by
$(\gamma\omega)(g)=(L_\gamma^*\omega)(\gamma^{-1}g)$,
where $L_\gamma^\ast$ is the pull back of forms associated to the
diffeomorphism $L_\gamma:X\rightarrow X$
given by $L_\gamma(x):=\gamma^{-1}x$.
The complex $({\cal C}^\ast,d)$ is exact
(the contraction of $X$ along radial rays induces a zero-homotopy
of the de Rham complex)
and it consists of $\Gamma$-acyclic modules (\cite{bunkeolbrich947}, Lemma
2.4).
Hence $H^\ast(\Gamma,S_\infty C^\infty(G))$ is the cohomology of the complex
$({}^\Gamma{\cal C}^\ast,d)$.
Let ${\cal S}^\ast:=S_\infty {\cal L}^\ast[S_\infty(G)]$
(see Section \ref{sec15} for notation).
Then $({\cal S}^\ast,d)\hookrightarrow ({\cal C}^\ast,d)$
is a subcomplex.
\begin{lem}\label{incl}
The inclusion $({}^\Gamma{\cal S}^\ast,d)\hookrightarrow ({}^\Gamma{\cal C}^\ast,d)$
induces an isomorphism in cohomology.
\end{lem}
{\it Proof.$\:\:$}
The manifold $Y$ has finitely many cusps each diffeomorphic to
$B_i\times [0,\infty)$, where $B_i$ is some compact nil-manifold.
The Borel-Serre compactification of $Y$ is obtained
by attaching copies of the cusp bases $B_i$
as a boundary at infinity obtaining a manifold with boundary $\bar{Y}$.
There is a natural sheafification
$(\underline{{}^\Gamma{\cal S}^\ast},d)\hookrightarrow
(\underline{{}^\Gamma{\cal C}^\ast},d)$
of the inclusion $({}^\Gamma{\cal S}^\ast,d)\hookrightarrow ({}^\Gamma{\cal C}^\ast,d)$
on $\bar{Y}$.
To any finite open covering of $\bar{Y}$ there is an associated
partition of unity
which is compatible with the sheafs $\underline{{}^\Gamma{\cal S}^\ast}$.
Thus $\underline{{}^\Gamma{\cal S}^\ast}$ and $\underline{{}^\Gamma{\cal C}^\ast}$
are acyclic with respect to the global section functor.
The complex of sheaves corresponding to $({}^\Gamma{\cal C}^\ast,d)$
is locally acyclic by the standard Poincar\'e Lemma.
Moreover the inclusion of sheaves
$(\underline{{}^\Gamma{\cal S}^\ast},d)\hookrightarrow
(\underline{{}^\Gamma{\cal C}^\ast},d)$
induces an isomorphism of the zeroth
cohomology sheaves.
By Lemma \ref{mmarrt} $(\underline{{}^\Gamma{\cal S}^\ast},d)$
is locally acyclic.
Thus the inclusion
$(\underline{{}^\Gamma{\cal S}^\ast},d)\hookrightarrow
(\underline{{}^\Gamma{\cal C}^\ast},d)$
is a quasi-isomorphism.
Since the sheaves $\underline{{}^\Gamma{\cal S}^\ast},\underline{{}^\Gamma{\cal C}^\ast}$
are acyclic with respect
to the global section functor, the induced map of the complexes
of global sections $({}^\Gamma {\cal S}^*,d)\hookrightarrow ({}^\Gamma {\cal C}^*,d)$ is
a quasi-isomorphism, too.
\hspace*{\fill}$\Box$ \newline\noindent
\begin{lem} \label{moac}
$H^p({}^\Gamma{\cal S}^.,d)=0$, $\forall p\ge 1$.
\end{lem}
{\it Proof.$\:\:$}
The map $T:X\times G\rightarrow X\times G$ given by
$(x,g)\rightarrow (gx,g)$ intertwines the $\Gamma$-action on the second factor
$G$ with the diagonal $\Gamma$-action
on the product $X\times G$.
We claim that $T$ induces an isomorphism
$T^\ast:S_\infty{\cal L}^\ast[S_\infty(G)]\rightarrow S_\infty{\cal L}^\ast[S_\infty(G)]$
intertwining the $\Gamma$-module structure given above with the
$\Gamma$-module structure
given by $(\gamma\omega)(g)=\omega(\gamma^{-1}g)$
(again viewing $\omega$ as a function from $G$ to ${\cal L}^*$).
The inverse of $T^{\ast}$ is induced by $T^{-1}:X\times G\rightarrow X\times
G$,
$T^{-1}(x,g)=(g^{-1}x,g)$.
In fact, $(T^\ast\omega)(g)=L_{g^{-1}}^\ast\omega(g)$ and
$$(T^\ast\gamma\omega)(g)=L_{g^{-1}}^\ast L_\gamma^\ast
\omega(\gamma^{-1}g))=(T^\ast\omega)(\gamma^{-1}g)\ . $$
In order to prove the claim we must show that $T^\ast,(T^*)^{-1}$
are compatible with the weighted spaces.
It is at this point that we have to consider the weighted
de Rham complex ${\cal S}^*$.
Since $T$ mixes the $G$- and the $X$ directions
it does not act on ${\cal C}^\ast$.
Let $\Delta:{\cal U}({\bf g})\rightarrow {\cal U}({\bf g})\otimes {\cal U}({\bf g})$
be the co-product induced by $X\mapsto X\otimes 1 + 1 \otimes X$, $X\in {\bf g}$.
Fix $L_1,L\in {\cal U}({\bf g})$ and let $(\Delta\otimes {\mbox{\rm id}} )\Delta(L) =\sum_\alpha
A_\alpha\otimes B_\alpha\otimes C_\alpha$.
Let $\omega\in {\cal S}^p$.
Then $$(T^\ast\omega)(L_1h,Lg)=\sum_{\alpha} \omega(B_\alpha gL_1h,C_\alpha
g)\circ DL_{g^{-1}A^{op}_\alpha}\ ,$$
where $DL_{g}$ is the differential of $L_g$ acting on $TX$.
Let $r=\deg(L_1)+\deg(L)$. Assume that
$$\sup_{g,h\in G, l,l_1\in L_r,}
{\rm e}^{-R{\rm dist}({\cal O},gK)}{\rm e}^{-Q{\rm dist}({\cal O},hK)} |\omega(X(l)h,X(l_1)g)|<\infty
$$
for appropriate $Q,R\in {\bf R}$.
Here $\{X(l)\}_{l\in L_r}$ is a basis of the differential operators on $G$ of
order $\le r$
(similarly to the notation in the proof of Lemma \ref{amdt}).
Note that ${\rm dist}({\cal O},ghK)\le {\rm dist}({\cal O},hK)+ {\rm dist}({\cal O},gK)$.
Moreover we have
$$|DL_{g^{-1}A^{op}_\alpha}|\le C {\rm e}^{-P{\rm dist}({\cal O},gK)} \quad \forall \alpha,
\:\:\forall g\in G $$
and
$B_\alpha gL_1h=\sum_{l\in L_r} e_{\alpha,l}(g) X(l)gh$ with
$$|e_{\alpha,l}(g)|\le C {\rm e}^{-P{\rm dist}({\cal O},gK)}
\quad \forall l\in L(r),\:\:\:\forall \alpha, \:\:\forall g\in G$$
for $P\in {\bf R}$ large enough.
We conclude that
$$\sup_{g,h\in G} |(T^\ast\omega)(L_1h,Lg)|
{\rm e}^{-(R+Q+2P){\rm dist}({\cal O},gK)}{\rm e}^{- Q {\rm dist}({\cal O},hK)}
<\infty\ .$$
Hence $T^*\omega\in {\cal S}^p$.
In a similar manner one can handle $(T^*)^{-1}$ thus proving the claim.
We see that $H^\ast({}^\Gamma {\cal S}^.,d)$ is isomorphic to the cohomology of
$(S_\infty{\cal L}^\ast[ \Gamma \backslash G],d)$.
Lemma \ref{moac} now follows from Lemma \ref{mglobal}. \hspace*{\fill}$\Box$ \newline\noindent
We now finish the proof of Theorem \ref{gaz}.
Note that $S_\infty C^\infty(G)$ carries a right $K$-module structure which
commutes with the left $\Gamma$-module structure.
This induces a $K$-action on $({}^\Gamma {\cal C}^*,d)$.
$H^*(\Gamma,S_\infty{\cal E})$ is the cohomology of the complex
$([{}^\Gamma {\cal C}^\ast\otimes V_\gamma]^K,d)$.
Let $[z]\in H^p(\Gamma, S_\infty {\cal E})$, $p\ge 1$,
be represented by the $K$-invariant cycle $z\in [{}^\Gamma {\cal C}^p\otimes
V_\gamma]^K$.
Then by Lemma \ref{acc} we have
$z=d b$ for some possibly non-invariant $p-1$ cochain
$b\in {}^\Gamma {\cal C}^{p-1}\otimes V_\gamma$. Let $\bar{b}$ the
average of $b$ with respect to $K$. Then also $d\bar{b}=z$ and hence
$0=[z]\in H^p(\Gamma, S_\infty{\cal E})$.
This proves Theorem \ref{gaz} .\hspace*{\fill}$\Box$ \newline\noindent
\section{The cokernel of $B:S_\infty{\cal E}_Y\rightarrow
S_\infty{\cal E}_Y$}\label{klop}
In this section we relate the dimension of the cokernel of $B:=\Omega-\mu$ on
$S_\infty{\cal E}_Y:={}^\Gamma S_\infty{\cal E}$ with the
dimension of corresponding spaces of cusp forms.
Since it is not a priori clear that the range of $B$ is closed note that
we consider the algebraic cokernel of $B$.
Let $E_Y:=\Gamma\backslash E$ be the locally homogeneous
bundle over $Y$. A section $f\in C^\infty(Y,E_Y)=:{\cal E}_Y$
can be viewed as a function on $f:G\rightarrow V_\gamma$
satisfying $f(h g k)=\gamma^{-1}(k)f(g)$ for all $h\in \Gamma$, $k\in K$.
The space $S_\infty{\cal E}_Y$ has the alternative description
$$
S_\infty{\cal E}_Y=\{f\in {\cal E}_Y\:|\: \forall L\in{\cal U}({\bf g})\: \exists R\in {\bf R} \:\:
s.t. \:\:\sup_{g\in G}{\rm e}^{R{\rm dist}_Y(\Gamma{\cal O},\Gamma gK)}|f(Lg)| <\infty\}\ .
$$
A section $f\in {\cal E}_Y$ which is an eigenfunction of
$\Omega$ and satisfies
$$\sup_{g\in G} {\rm e}^{R{\rm dist}_Y(\Gamma{\cal O},\Gamma gK)}
|f(g)|<\infty\:\:\:\forall R\in{\bf R}$$
is called a cusp form.
Let $V_\mu$ be the space of cusp forms in $\ker(B)$.
If $V_\mu\not=\{0\}$, then necessarily $\mu\in\ R$, since $\Omega$
is symmetric.
The main result of the present section is the following theorem.
\begin{theorem}\label{koko}
$$\dim {\rm coker}(B:S_\infty{\cal E}_Y\rightarrow S_\infty{\cal E}_Y)=\dim V_\mu\ .$$
\end{theorem}
The proof of the theorem occupies the remainder of the section.
\begin{lem}\label{larger}
$$\dim {\rm coker}(B:S_\infty{\cal E}_Y\rightarrow S_\infty{\cal E}_Y)\ge \dim V_\mu\ .$$
\end{lem}
{\it Proof.$\:\:$}
We show that the projection of $V_\mu$ to ${\rm coker}(B:S_\infty{\cal E}_Y\rightarrow
S_\infty{\cal E}_Y)$
is an inclusion.
We can assume that $V_\mu\not=0$ and hence $\mu\in{\bf R}$.
Thus $B$ is symmetric.
If $f\in V_\mu$, then it vanishes rapidly and can be integrated against
elements
of $S_\infty{\cal E}_Y$.
Thus let $f\in V_\mu$ and assume that $[f]=0$ in
${\rm coker}(B:S_\infty{\cal E}_Y\rightarrow S_\infty{\cal E}_Y)$.
Then $f=Bh$ for some $h\in S_\infty{\cal E}_Y$.
We have
\begin{eqnarray*}
0&=&\int_Y \langle (Bf)(y)\:,\:h(y)\rangle_{E_{Y,y}} dy \\
&=&\int_Y \langle f(y)\:,\:(Bh)(y)\rangle_{E_{Y,y}} dy\\
&=& \|f\|^2_{L^2(Y,E_Y)}\ .
\end{eqnarray*}
It follows $f=0$.
This proves Lemma \ref{larger}. \hspace*{\fill}$\Box$ \newline\noindent
We define an increasing sequence of Fr\'echet spaces $S_R{\cal E}_Y$, $R\in{\bf R}$, by
$$S_R{\cal E}_Y:=\{f\in{\cal E}_Y\:|\: p_{Y;-R,L}(f)<\infty\: \:\forall L\in{\cal U}({\bf g})\}\
,$$
where the seminorms are defined by
$$p_{Y;R,L}(f):=\sup_{g\in G} {\rm e}^{R{\rm dist}(\Gamma{\cal O},\Gamma gK)} |f(gL)|\ .$$
Note that (in contrast to the definition of $S_\infty{\cal E}$)
we employ the right action of ${\cal U}({\bf g})$ to define $p_{Y;R,L}(f)$.
In fact, to define $S_R{\cal E}_Y$ it is sufficient to consider the seminorms
$p_{Y;R,\Omega^k}(f)$,
$k\in{\bf N}_0$.
Let
$$
S_{-\infty}{\cal E}_Y = \bigcap_{R\in{\bf R}}S_R{\cal E}_Y$$
with the natural topology
of the intersection and $S_\infty{{\cal E}^\prime}_Y$ be the topological conjugate dual of
$S_{-\infty}{\cal E}_Y$.
The Hermitian scalar product of $E_Y$ induces an inclusion
$S_\infty{\cal E}_Y\hookrightarrow S_\infty{{\cal E}^\prime}_Y$.
\begin{lem}\label{smmo}
The inclusion $S_\infty{\cal E}_Y\hookrightarrow S_\infty{{\cal E}^\prime}_Y$
induces an injection
$${\rm coker}(B:S_\infty{\cal E}_Y\rightarrow S_\infty{\cal E}_Y)\hookrightarrow
{\rm coker}(B:S_\infty{{\cal E}^\prime}_Y\rightarrow S_\infty{{\cal E}^\prime}_Y)\ .$$
\end{lem}
{\it Proof.$\:\:$}
We reduce the proof to Lemma \ref{reg}.
Let $\pi:X\rightarrow Y$ be the projection.
We define a continuous map $\pi_*:S_{-\infty}{\cal E}\rightarrow S_{-\infty}{\cal E}_Y$.
For $f\in S_{-\infty}{\cal E}$ let $(\pi_\ast f)(\Gamma g):=\sum_{\gamma\in \Gamma}
f(\gamma g)$.
We show that the sum converges and $\pi_\ast$ is continuous.
There is a $Q\in R$ such that
$\sup_{g\in G }\sum_{\gamma\in\Gamma} {\rm e}^{Q{\rm dist}(\gamma gK,{\cal O})}=:D<\infty
$.
For $k\in {\bf N}_0$ we have
\begin{eqnarray*}
p_{Y;R,\Omega^k}(\pi_\ast f)&= &\sup_{g\in G}{\rm e}^{R {\rm dist}_Y(\Gamma {\cal O},\Gamma
gK)} |\sum_{\gamma\in \Gamma} f(\gamma g\Omega^kK)|\\
&\le & \sup_{g\in G}{\rm e}^{R {\rm dist}_Y(\Gamma {\cal O},\Gamma gK)}\sum_{\gamma\in
\Gamma}
|f(\gamma \Omega^kgK)|\\
&\le& \sup_{g\in G}{\rm e}^{R {\rm dist}_Y(\Gamma {\cal O},\Gamma gK)}\sum_{\gamma\in
\Gamma}
{\rm e}^{ (Q-R){\rm dist}(\gamma gK,{\cal O})} p_{(R-Q),\Omega^k}(f)\\
&\le & \sup_{g\in G} p_{(R-Q),\Omega^k}(f)
\sum_{\gamma\in \Gamma}
{\rm e}^{ Q {\rm dist}(\gamma gK,{\cal O})}\\
&\le& D p_{(R-Q),\Omega^k}(f)\ .
\end{eqnarray*}
This estimate shows the continuity of $\pi_\ast$.
Let $\pi^*:S_\infty{{\cal E}^\prime}_Y\rightarrow S_\infty{{\cal E}^\prime}$ be the adjoint of $\pi_*$.
To prove the Lemma it suffices to show that
if $f\in S_\infty{{\cal E}^\prime}_Y$ with $F=Bf\in S_\infty{\cal E}_Y$, then $f\in
S_\infty{\cal E}_Y$.
Let $W$ be the parametrix of $B$ constructed in the proof of Lemma \ref{reg}.
Then we have $\pi^\ast f=WF + S\pi^\ast f$ (viewing $F\in {}^\Gamma
S_\infty{\cal E}$).
We have already shown that $WF,S\pi^*f\in S_\infty{\cal E}$.
Since $W,S$ are $G$-equivariant we obtain $\pi^*f\in {}^\Gamma S_\infty{\cal E} $.
This finishes the proof of the Lemma. \hspace*{\fill}$\Box$ \newline\noindent
\begin{lem}\label{clos}
The operator $B^*:S_{-\infty}{\cal E}_Y\rightarrow S_{-\infty}{\cal E}_Y$
has closed range.
\end{lem}
{\it Proof.$\:\:$}
Let $\{f_i\}$ be a sequence in $S_{-\infty}{\cal E}_Y$ such that
$B^*f_i=:h_i\rightarrow h\in S_{-\infty}{\cal E}_Y$.
We are to find $f\in S_{-\infty}{\cal E}_Y$ with $B^*f=h$.
If $V_\mu\not=0$, then
$V_\mu=\ker(B^*:S_{-\infty}{\cal E}_Y\rightarrow S_{-\infty}{\cal E}_Y) $, and
we can project $f_i$ to the $L^2$-orthogonal complement of $V_\mu$.
Thus we can assume that $f_i\perp V_\mu$.
We can assume that for some $R\in {\bf R}$ the sequence $p_{Y;R,1}(f_i)$ is bounded.
If not, we divide $f_i$ by $p_{Y;R,1}(f_i)$ obtaining a sequence $\tilde{f}_i$
with $p_{Y;R,1}(\tilde{f}_i)=1$ and $ B^*\tilde{f}_i\to 0$.
We show below that $\tilde{f}_i$ has a subsequence converging to $F\in
S_{-\infty}{\cal E}_Y$.
Now $ B^*F=0$ and $F\perp V_\mu$.
Hence $F=0$ contradicting $p_{Y;R,1}(F)=1$.
Since for all $k\in{\bf N}_0$ the sequence $p_{Y;R,\Omega^k}(h_i)$ is bounded we
conclude
that $p_{Y;R,\Omega^k}(f_i)$ is bounded, too.
Consider a cusp $c_l$ of $Y$ associated to the minimal parabolic subgroup
$P=MAN\subset G$.
Then a neighbourhood $U$ of infinity of this cusp can be identified with
$\Gamma\cap N\backslash N \times [d_l,\infty)$, $d_l\in A$ large.
If $F\in {\cal E}_Y$, then we define the constant term $F_P\in C^\infty(U,E_{Y|U})$
by
$$F_P(na)=\frac{1}{{\rm vol}(\Gamma\cap N\backslash N)}\int_{\Gamma\cap N\backslash
N} F(n^\prime na) dn^\prime \ . $$
Let $\chi\in C^\infty(Y)$ be a cut-off function being one on $\Gamma\cap
N\backslash N\times [d_l+1,\infty)$ and
zero outside of $U$. Then $F_{c_l}=\chi F_P \in {\cal E}_Y$.
If $Y$ has the cusps $c_l$, $l=1,\dots,r$, then we set $F_c:=\sum_{l=1}^r
F_{c_l}$.
We apply this construction to our sequence $f_i$ obtaining a decomposition
$f_i=f_{i,c}+f_{i,r}$ with $f_{i,r}:=f_i-f_{i,c}$.
Since $f_i$ is bounded in $S_R{\cal E}_Y$, by a Lemma
of Gelfand \cite{franke95}, Thm. 5, the sequence $f_{i,r}$ is
bounded in $S_{-\infty}{\cal E}_Y$.
Since for $R<R^\prime$ the embedding
$S_R{\cal E}_Y\hookrightarrow S_{R^\prime}{\cal E}_Y$ is compact, any bounded
sequence in $S_{-\infty}{\cal E}_Y$ has a converging subsequence.
Thus by taking a subsequence we can assume that $f_{i,r}$ converges in
$S_{-\infty}{\cal E}_Y$.
For
${\rm dist}_Y(\Gamma{\cal O},y)\ge {\rm max}_l d_l+1$ we have $(B^*f)_c(y)= B^*f_c(y)$.
Consider again the cusp $c_l$ and the coordinates $ \Gamma\cap N\backslash
N\times [d_l,\infty)$.
There are commuting $x,y\in {\mbox{\rm End}}(V_\gamma)$ such that
$$( B^*F_{c_l})(n,a)=-((\frac{d}{da}+x)(\frac{d}{da}+y)F_{c_l})(n,a)\ ,$$
$a>d_l+1$.
Assume that $R< - {\rm max}( \|x\|) , \|y\| ) $. We set for $F\in S_R{\cal E}_Y$
$$(H_{c_l}F)(n,a):=-\chi(a) {\rm e}^{-ya} \int_{a}^\infty
{\rm e}^{(y-x)a_1}\int_{a_1}^\infty {\rm e}^{ xb} F_{c_l}(n,b) db da_1\ .$$
Again, if $Y$ has cusps $c_l$, $l=1,\dots, r$, then we set $H_c:=\sum_{l=1}^r
H_{c_l}$.
Then $H_c:S_R{\cal E}_Y\rightarrow S_R{\cal E}_Y$ is continuous,
and ${\mbox{\rm supp}} (H_c B^* F-F_c)\subset V$, where $V\subset Y$ is compact and
independent of $F$.
The proof of continuity is similar to the corresponding argument
in the proof of Lemma \ref{iter}.
Notice that
$\Omega^k H_c-H_c \Omega^k =W$
is a continuous operator $W : S_R{\cal E}_Y\rightarrow C_c^\infty(Y,E_Y)$
and ${\mbox{\rm supp}} Wf \subset V^\prime$, where $V^\prime\subset Y$ is compact and
independent of $f$.
Thus $F_i:= H_c B^* h_i-f_{i,c}$ is a bounded sequence in $S_{-\infty}{\cal E}_Y$.
Hence $F_i$ has a subsequence converging in $S_{-\infty}{\cal E}_Y$.
Since $H_c B^* h_i$ converges in $S_{-\infty}{\cal E}_Y$
by taking a subsequence we can assume that $f_{i,c}$ converges
in $S_{-\infty}{\cal E}_Y$, too.
Let $f$ be the limit of $f_i=f_{i,c}+f_{i,r}$ for this subsequence.
Then $ B^*f=h$. This finishes the proof of the lemma.
\hspace*{\fill}$\Box$ \newline\noindent
We now finish the proof of Theorem \ref{koko}.
By Lemma \ref{clos} we have
$$\dim \:V_\mu=\dim \ker ( B^*: S_{-\infty}{\cal E}_Y\rightarrow
S_{-\infty}{\cal E}_Y)=\dim{\rm coker}(
B:S_\infty{{\cal E}^\prime}_Y\rightarrow S_\infty{{\cal E}^\prime}_Y)\ .$$
By Lemma \ref{smmo}
we have
$$\dim{\rm coker}(B:S_\infty{\cal E}_Y\rightarrow S_\infty{\cal E}_Y)\le \dim V_\mu\ .$$
Combining this with Lemma \ref{larger} we obtain the theorem.
\hspace*{\fill}$\Box$ \newline\noindent
\section{$\Gamma$-cohomology}\label{ggkk}
In this section we discuss properties of the $\Gamma$-cohomology
of distribution vector globalizations of admissible representations
of $G$.
Let $(\pi,V_{\pi,K})\in {{\cal HC}(\gaaa,K)}$ and $\Gamma\subset G$ be a discrete torsion-free
subgroup
of finite covolume.
Let $B=(\Omega-\mu)^k$ such that $BV_{\pi,K}=0$.
\begin{prop}\label{ug1}
We have
$$\dim H^p(\Gamma,V_{\pi,-\infty})<\infty,\quad \forall p\ge 0.$$
\end{prop}
{\it Proof.$\:\:$}
Let
\begin{equation}\label{ss.ss}
0\rightarrow V_{\pi,-\infty}\rightarrow S_\infty{\cal E}_0\stackrel{\scriptsize
\left(\begin{array}{c}D_0\\B\end{array}\right)}
{\longrightarrow}\begin{array}{c}S_\infty{\cal E}_1\\ \oplus\\
S_\infty{\cal E}_0\end{array}
\stackrel{\scriptsize
\left(\begin{array}{cc}D_1&H_0\\-B&D_0\end{array}\right)}
{\longrightarrow}\begin{array}{c}S_\infty{\cal E}_2\\ \oplus\\
S_\infty{\cal E}_1\end{array}
\stackrel{
\scriptsize\left(\begin{array}{cc}D_2&H_1\\B&D_1\end{array}\right)}
{\longrightarrow}\dots
\end{equation}
be a standard resolution (see Proposition \ref{stare}) of $V_{\pi,-\infty}$.
It is a $\Gamma$-acyclic resolution of $V_{\pi,-\infty}$ by Theorem \ref{gaz}.
The cohomology of the subcomplex of $\Gamma$-invariant vectors is isomorphic to
$H^\ast(\Gamma,V_{\pi,-\infty})$.
For any locally homogeneous vector bundle $E_Y\rightarrow Y$
we denote by ${\cal E}_Y(B)_{cusp}$ the space of cusp-forms in ${\cal E}_Y(B)$
and by $S_\infty{\cal E}_Y(B)$ the kernel of $B$ in $S_\infty{\cal E}_Y$.
We consider the subcomplex of the complex of $\Gamma$-invariants of
(\ref{ss.ss})
\begin{equation}\label{rr.rr}
0 \rightarrow S_\infty{\cal E}_{0,Y}(B) \stackrel{\scriptsize
\left(\begin{array}{c}D_0\\0\end{array}\right)}
{\longrightarrow}\begin{array}{c}S_\infty{\cal E}_{1,Y}(B) \\ \oplus\\
{\cal E}_{0,Y}(B)_{cusp}\end{array}
\stackrel{\scriptsize
\left(\begin{array}{cc}D_1&H_0\\0&D_0\end{array}\right)}
{\longrightarrow}\begin{array}{c}S_\infty{\cal E}_{2,Y}(B) \\ \oplus\\
{\cal E}_{1,Y}(B)_{cusp}\end{array}
\stackrel{
\scriptsize\left(\begin{array}{cc}D_2&H_1\\0&D_1\end{array}\right)}
{\longrightarrow}\dots
\end{equation}
and claim that its cohomology is $H^\ast(\Gamma,V_{\pi,-\infty})$.
In fact, let $(f_i,f_{i-1})\in S_\infty{\cal E}_{i,Y}\oplus S_\infty{\cal E}_{i-1,Y}$
be a cochain in the complex of $\Gamma$-invariants of (\ref{ss.ss}).
By the results of Section \ref{klop} there is a unique decomposition
$f_{i-1}=f_{i-1}^{im}+f_{i-1}^{cusp}$, where $f_{i-1}^{im}\in B
S_\infty{\cal E}_{i,Y}$
and $f_{i-1}^{cusp}\in {\cal E}_{i-1,Y}(B)_{cusp}$.
Notice that ${\rm coker} B = {\rm coker} (\Omega-\mu)$.
Thus modulo a coboundary the cochain $(f_i,f_{i-1})$
is equivalent to $(\tilde{f_i},f_{i-1}^{cusp})$. If in addition $(f_i,f_{i-1})$
and hence $(\tilde{f_i},f_{i-1}^{cusp})$ is a cocycle, then
$B\tilde{f_i}=0$, since $D_{i-1}:{\cal E}_{i-1,Y}(B)_{cusp}\rightarrow
{\cal E}_{i,Y}(B)_{cusp}$
and the range of $B:S_\infty{\cal E}_{i,Y}\rightarrow S_\infty{\cal E}_{i,Y}$ is
transverse
to ${\cal E}_{i,Y}(B)_{cusp}$. We conclude that the cohomology of (\ref{rr.rr})
surjects
onto $H^*(\Gamma,V_{\pi,-\infty})$.
Assume now that $f_i=H_{i-2}g_{i-2}+D_{i-1}g_{i-1}$,
$f_{i-1}=(-1)^{i-1} Bg_{i-1}+D_{i-2}g_{i-2}$, and
$f_{i-1}\in {\cal E}_{i-1,Y}(B)_{cusp}$, $g_{i-2}\in {\cal E}_{i-2,Y}(B)_{cusp}$.
It follows that $Bg_{i-1}=0$ and thus $(g_{i-1},g_{i-2})$ is
a $i-1$-cochain of (\ref{rr.rr}). This show that the cohomology of
(\ref{rr.rr})
maps injectively to $H^*(\Gamma,V_{\pi,-\infty})$ proving the claim.
The lemma now follows since (\ref{rr.rr}) is a complex of finite-dimensional
vector spaces (see e.g. \cite{harishchandra68}, Thm.1).
\hspace*{\fill}$\Box$ \newline\noindent
If $\Gamma$ is cocompact, then we can prove a Poincar\'e duality theorem as in
\cite{bunkeolbrich947}, Proposition 5.2.
It relates the $\Gamma$-cohomology of the distribution vector globalization
of an admissible representation of $G$ with the $\Gamma$-cohomology
of the smooth globalization of its dual.
\begin{prop} \label{ug2}
Let $\Gamma$ be cocompact. The $\Gamma$-cohomology of $V_{\pi,\pm\infty}$
satisfies Poincar\'e duality
$$H^p(\Gamma,V_{\pi,-\infty})^\ast\cong H^{n-p}(\Gamma,V_{\tilde\pi,\infty}),$$
where $n=\dim(X)$.
\end{prop}
Let $V_{\pi,\pm\omega}$ be the minimal and maximal globalizations of
$V_{\pi,K}$,
respectively. If $\Gamma$ is cocompact, then
${\cal E}_Y(B)_{cusp}={\cal E}_Y(B)=S_\infty{\cal E}_Y(B)$.
Combining \cite{bunkeolbrich947}, Proposition 5.1 with Proposition \ref{ug1} we
obtain
\begin{kor}
Let $\Gamma$ be cocompact. Then for $(\pi,V_{\pi,K})\in{{\cal HC}(\gaaa,K)}$ we have
\begin{eqnarray*}
H^\ast(\Gamma,V_{\pi,\omega})&=&H^\ast(\Gamma,V_{\pi,\infty})\\
H^\ast(\Gamma,V_{\pi,-\infty})&=&H^\ast(\Gamma,V_{\pi,-\omega})\ ,
\end{eqnarray*}
where the identifications are induced by the natural inclusions
of the globalizations.
\end{kor}
If $\Gamma$ has torsion, then there exists a cofinite torsion free normal
subgroup
$\Gamma^\prime\subset \Gamma$. Employing the spectral sequence
for the group cohomology associated to the extension
$$0\rightarrow \Gamma^\prime\rightarrow \Gamma\rightarrow F\rightarrow 0\ ,$$
where $F$ is some finite group, one easily obtains a generalization of the
results
of the present section to $\Gamma$.
The spectral sequence degenerates at the second term since
the higher cohomology of a finite group with
coefficients in a vector spaces over ${\bf C}$ vanishes.
It follows $H^*(\Gamma,V)=H^*(\Gamma^\prime,V)^F$ for any $\Gamma$-module
$V$ over ${\bf C}$.
\section{Fuchsian groups of the first kind}\label{fufu}
In this section we give a detailed discussion of the $\Gamma$-cohomology
of a Fuchsian group of the first kind with coefficients
in the distribution vector globalization of principal series
representations. In this case we know explicit standard resolutions.
Let $\Gamma\subset PSL(2,{\bf R})=:G$ be a discrete torsion-free subgroup
of finite covolume. Such a $\Gamma$ is called a Fuchsian group of the first
kind
and it acts freely on the hyperbolic plane $X=H^2$.
The quotient $Y=\Gamma\backslash X$ is a complete Riemann surface of finite
volume.
The group $G$ acts on the circle $S^1$
which can be identified with the
boundary $\partial X$ of $X$ using the Poincar\'e disc model.
Let $T\rightarrow S^1$ be the complexified tangent bundle of $S^1$.
It is $G$-homogeneous and we can form complex powers $T^\lambda\rightarrow
S^1$, $\lambda\in{\bf C}$.
The number $\lambda\in{\bf C}$ parametrizes a principal series representation
$(\pi^\lambda,H^\lambda)$ of $G$ on the Hilbert space
$L^2(S^1,T^{\lambda-1/2})$.
By $H^\lambda_{-\infty}$ we denote the space of its distribution vectors.
For $\lambda\not=-1/2,-3/2,-5/2,\dots$
combining a theorem of Helgason
(\cite{helgason70}, \cite{helgason84} Introduction Thm. 4.3)
with the characterization of the distribution vector globalization
(Wallach \cite{wallach92} Ch. 11, Casselman \cite{casselman89}) we see that
the Poisson transform $P_\lambda$ is an $G$-equivariant isomorphism
$$P_\lambda:H^\lambda_{-\infty}\stackrel{\sim}{\rightarrow} S_\infty{\cal E}(B)\ ,$$
where $B=\Omega-1/4+\lambda^2$ and $E=X\times{\bf C}$ is the trivial bundle
(in this special situation this fact was first obtained by Lewis
\cite{lewis78}).
Thus a standard resolution of the principal series representation
$H^\lambda_{-\infty}$
for $\lambda\not=-1/2,-3/2,-5/2,\dots$
is simply
$$0\rightarrow H^\lambda_{-\infty}\stackrel{P_\lambda}{\rightarrow}S_\infty{\cal E}
\stackrel{B}{\rightarrow}S_\infty{\cal E}\rightarrow 0\ .$$
The complex (\ref{rr.rr}) reduces to
\begin{equation}\label{red}0\rightarrow
S_\infty{\cal E}_Y(B)\stackrel{0}{\rightarrow}{\cal E}_Y(B)_{cusp}\rightarrow 0\
.\end{equation}
\begin{prop}\label{gener}
For $\lambda\not=-1/2,-3/2,-5/2,\dots$ we have
\begin{eqnarray*}
H^0(\Gamma,H^\lambda_{-\infty})&=&S_\infty{\cal E}_Y(B)\\
H^1(\Gamma,H^\lambda_{-\infty })&=&{\cal E}_Y(B)_{cusp}\\
H^2(\Gamma,H^\lambda_{-\infty })&=& 0.
\end{eqnarray*}
Moreover, $\chi(\Gamma,H^\lambda_{-\infty})=\dim
H^0(\Gamma,H^\lambda_{-\infty})-\dim H^1(\Gamma,H^\lambda_{-\infty })=r$,
where $r$ is the number of cusps of $Y$.
If $H^1(\Gamma,H^\lambda_{-\infty})\not=0$, then $\lambda\in
\imath{\bf R}\cup(-1/2,1/2)$.
\end{prop}
{\it Proof.$\:\:$}
The first part of the Proposition follows immediately from (\ref{red}).
${\cal E}_Y(B)_{cusp}\subset \ker_{L^2}(B:{\cal E}_Y\rightarrow {\cal E}_Y)$ and
${\mbox{\rm spec}}_{L^2}\Delta_Y\subset [0,\infty)$
implies the last assertion.
Let $p:S_\infty{\cal E}_Y(B)\rightarrow {\bf C}^{2r}$ be the linear map taking the
constant
term. For any cusp the constant term has two components
(the incoming and the outcoming).
It is known that $\dim{\mbox{\rm im}}(p)=r$.
In fact, the range of $p$ is generated by the constant terms of regular
Eisenstein
series and their residues. The scattering matrix fixes the relation
between the two components of the constant term.
Since $\ker(p)={\cal E}_Y(B)_{cusp}$, the assertion about the Euler characteristic
follows. \hspace*{\fill}$\Box$ \newline\noindent
Now we discuss the case $\lambda=-k/2$, $k=1,3,5\dots$
taking the structure of $H^{\pm k/2}_{-\infty}$ as a $G$-module into account.
We first exploit the exact sequence
\begin{equation}\label{wert}0\rightarrow F_k \rightarrow
H^{k/2}_{-\infty}\rightarrow D^+_k\oplus D^-_k\rightarrow 0\ ,\end{equation}
where $F_k$ is the finite-dimensional representation of $G$ of dimension $k$
and $D^\pm_k$ are the distribution vectors of
holomorphic and anti-holomorphic discrete series representations.
Let $r$ denote the number of cusps of $Y$ and $g$ denote the genus.
We have $H^i(\Gamma,H^{k/2}_{-\infty})=0$, $i\ge 1$, and
$\dim H^0(\Gamma,H^{k/2}_{-\infty})=\dim S_\infty{\cal E}_Y(B) = r$ by Proposition
\ref{gener}.
The long exact cohomology sequence associated to (\ref{wert}) gives
\begin{eqnarray*}
&&0\rightarrow H^0(\Gamma,F_k)\rightarrow
H^0(\Gamma,H^{k/2}_{-\infty})\rightarrow H^0(\Gamma,D_k^+\oplus D_k^-)
\stackrel{\delta}{\rightarrow} H^1(\Gamma,F_k)\rightarrow 0\\
&&\hspace{2cm}H^1(\Gamma,D_k^+\oplus D_k^-) = 0 \ .
\end{eqnarray*}
An investigation of the long exact sequence associated to
$$0\rightarrow D_k^+\oplus D_k^-\rightarrow H^{-k/2}_{-\infty}\rightarrow
F_k\rightarrow 0$$
leads to
\begin{eqnarray*}&&0\rightarrow H^0(\Gamma,D_k^+\oplus D_k^-)\rightarrow
H^0(\Gamma,H^{-k/2}_{-\infty}) \rightarrow H^0(\Gamma,F_k)\rightarrow 0\\
&&\hspace{2cm}H^1(\Gamma,H^{-k/2}_{-\infty})=H^1(\Gamma,F_k)\ .\end{eqnarray*}
\begin{prop}\label{wwqqq}
We have
\begin{eqnarray*}
\dim H^0(\Gamma,D_k^+\oplus D_k^-)&=&r+\dim H^1(\Gamma,F_k)-\dim
H^0(\Gamma,F_k)\\
\dim H^0(\Gamma,H^{-k/2}_{-\infty})&=&r+ \dim H^1(\Gamma,F_k)\\
\dim H^1(\Gamma,H^{-k/2}_{-\infty})&=&\dim H^1(\Gamma,F_k)\\
\chi(\Gamma,H^{-k/2}_{-\infty})&=&r
\end{eqnarray*}
\end{prop}
Since $r>1$ or $g>0$ we have for $k>1$
\begin{eqnarray*}
H^0(\Gamma,F_1)&=&1\\
H^1(\Gamma,F_1)&=&2g+r-1\\
H^0(\Gamma,F_k)&=&0\\
H^1(\Gamma,F_k)&=&k(2g+r-2)\ .
\end{eqnarray*}
It follows that
\begin{eqnarray*}
\dim H^0(\Gamma,D_1^+\oplus D_1^-)&=&2g+2r-2 \\
\dim H^0(\Gamma,D_k^+\oplus D_k^-)&=&k(2g -2)+(k+1)r\\
\dim H^0(\Gamma,H^{-1/2}_{-\infty})&=&2g+2r-1\\
\dim H^1(\Gamma,H^{-1/2}_{-\infty})&=&2g+r-1\\
\dim H^0(\Gamma,H^{-k/2}_{-\infty})&=&k(2g-2)+(k+1)r\\
\dim H^1(\Gamma,H^{-k/2}_{-\infty})&=&k(2g+r-2) \ .
\end{eqnarray*}
The aim of the following discussion is to understand Proposition \ref{wwqqq}
in the framework of standard resolutions.
Let ${\tt K}$ be the canonical bundle of $X$ (viewing $X$ as a complex manifold)
and ${\tt K}^i$ be its $i$'th power.
\begin{lem}\label{reso}
A standard resolution of $H^{-k/2}_{-\infty}$ is given by
\begin{eqnarray*}
\lefteqn{0\longrightarrow H^{-k/2}_{-\infty}\stackrel{P}{\longrightarrow}
S_\infty{\cal K}^{(k+1)/2}\oplus S_\infty{\cal K}^{-(k+1)/2}}\hspace{3cm}\\
&&\stackrel{\scriptsize\left(\begin{array}{c}
\bar{\partial}^{(k+1)/2}\oplus\partial^{(k+1)/2}\\\Omega+
\frac{k^2-1}{4}\end{array}\right)}{
\longrightarrow}\begin{array}{c}S_\infty{\cal K}^0\\\oplus\\
S_\infty{\cal K}^{(k+1)/2}\oplus
S_\infty{\cal K}^{-(k+1)/2} \end{array}
\longrightarrow\\
&&\stackrel{\scriptsize
(-\Omega-\frac{k^2-1}{4},\bar{\partial}^{(k+1)/2}
\oplus\partial^{(k+1)/2})}{\longrightarrow}
S_\infty{\cal K}^0\longrightarrow 0\ .
\end{eqnarray*}
Here $\bar{\partial}:{\cal K}^{(l+1)/2}\rightarrow {\cal K}^{(l-1)/2}$, $l>0$ odd,
is the contraction of the anti-holomorphic part of the canonical connection
with the K\"ahler form and similarly
$\partial:{\cal K}^{-(l+1)/2}\rightarrow{\cal K}^{-(l-1)/2}$
is the contraction of the holomorphic part of the canonical connection with the
K\"ahler form.
In abuse of notation we write $\bar{\partial}^{(k+1)/2}, \partial^{(k+1)/2}$
for the composition of the corresponding number of $\bar{\partial}$'s,
$\partial$'s,
respectively.
\end{lem}
{\it Proof.$\:\:$}
Define $B:=\Omega+\frac{k^2-1}{4}$.
As $G$-modules the eigenspaces
$S_\infty{\cal K}^{\pm(k+1)/2}(B)$ have three-step composition series'
with composition factors $D^\pm_k,F_k,D^\mp_k$ (see \cite{olbrichdiss}).
The operators $\bar{\partial}^{(k+1)/2}$, $\partial^{(k+1)/2}$, respectively,
annihilate exactly the first composition factor.
Above we have seen that
$S_\infty{\cal K}^0(B) = H^{k/2}_\infty$ with composition factors $D_k^+\oplus
D_k^-,F_k$.
This discussion shows that the complex above resolves
$H^{-k/2}_\infty$ on a ${\bf K}$-theoretic level.
In order to show that the complex is in fact a resolution of the specific
extension
$H^{-k/2}_\infty$ of the composition factors note that there is an injective
Poisson transform $P$.
\hspace*{\fill}$\Box$ \newline\noindent
Now we take the $\Gamma$-invariant vectors in the standard resolution.
The complex (\ref{rr.rr}) reduces to
\begin{eqnarray*}
\lefteqn{0\longrightarrow S_\infty{\cal K}^{(k+1)/2}_Y(B)\oplus
S_\infty{\cal K}^{-(k+1)/2}_Y(B)}\hspace{3cm}\\
&&\stackrel{\scriptsize\left(\begin{array}{c}
\bar{\partial}^{(k+1)/2}\oplus\partial^{(k+1)/2}\\0
\end{array}\right)}{\longrightarrow}\begin{array}{c}
S_\infty{\cal K}_Y^0(B)\\\oplus\\ {\cal K}^{(k+1)/2}_Y(B)_{cusp}
\oplus {\cal K}^{-(k+1)/2}_Y(B)_{cusp} \end{array}
\longrightarrow\\
&&\stackrel{\scriptsize(0,\bar{\partial}^{(k+1)/2}
\oplus\partial^{(k+1)/2})}{\longrightarrow} {\cal K}^0_Y(B)_{cusp}
\longrightarrow 0\ .
\end{eqnarray*}
In order to make further reductions we assume that $Y$ is not compact
(the case of compact $Y$ is an easy exercise and left to the reader).
Since for $k>1$ the operator $B$ on $L^2(Y)$ is strictly positive and
$${\cal K}^0_Y(B)_{cusp}\subset \ker_{L^2}(B:{\cal K}^0_Y\rightarrow {\cal K}^0_Y)$$ for $k>1$
we have ${\cal K}^0_Y(B)_{cusp}=0$. This is also true for $k=1$ since then
$\ker_{L^2}(B:{\cal K}^0_Y\rightarrow {\cal K}^0_Y)$ is generated by the constant
function which
is not a cusp form.
We claim that for $k>1$
\begin{eqnarray}0&=&\bar{\partial}^{(k+1)/2}:
S_\infty{\cal K}^{(k+1)/2}_Y(B)\rightarrow S_\infty
{\cal K}^{0}_Y(B)\label{firstl}\\
0&=&\partial^{(k+1)/2}:S_\infty{\cal K}^{-(k+1)/2}_Y(B)
\rightarrow S_\infty{\cal K}^0_Y(B)\ .\label{secondl}
\end{eqnarray}
To prove (\ref{firstl}) we argue that already
$$ 0=\bar{\partial}:S_\infty{\cal K}^{(k+1)/2}_Y(B)\rightarrow
S_\infty{\cal K}^{(k-1)/2}_Y(B)\ .$$
In upper half-plane coordinates $(x,y)$, $y>0$, $z=x+\imath y$,
we trivialize ${\tt K}^{(k+1)/2}$ using the section $(\frac{dz}{y})^{(k+1)/2}$.
Then the operator $B$ has the form $y^2\Delta + \imath (k+1)
\frac{\partial}{\partial x}+\frac{k^2-1}{4}$.
If $\phi\in S_\infty{\cal K}^{(k+1)/2}(B)$, then its constant term is a linear
combination of
$y^{-k}dz^{(k+1)/2}$ and $dz^{(k+1)/2}$. It follows that $\bar{\partial}\phi\in
L^2$.
But $\ker_{L^2}(B)=0$ on ${\tt K}^{ (k-1)/2}$, $k >1$.
Equation (\ref{secondl}) follows
by complex conjugation.
The spaces
\begin{eqnarray*}{\cal A}_{k+1}&:=&\ker(\bar{\partial}:
S_\infty{\cal K}^{(k+1)/2}_Y(B)\rightarrow S_\infty{\cal K}^{(k-1)/2}_Y(B))\\
\bar{{\cal A}}_{k+1}&:=&\ker( \partial: S_\infty{\cal K}^{-(k+1)/2}_Y(B)\rightarrow
S_\infty{\cal K}^{-(k-1)/2}_Y(B))
\end{eqnarray*}
are called the holomorphic and anti-holomorphic automorphic forms of weight
$k+1$.
The subspaces
\begin{eqnarray*}
{\cal S}_{k+1}&:=&{\cal K}^{(k+1)/2}_Y(B)_{cusp}\subset {\cal A}_{k+1}\\
\bar{{\cal S}}_{k+1}&:=&{\cal K}^{-(k+1)/2}_Y(B)_{cusp}\subset \bar{{\cal A}}_{k+1}
\end{eqnarray*}
are the holomorphic and anti-holomorphic cusp-forms of weight $k+1$.
Since for $k>1$ the space $S_\infty{\cal K}^0_Y(B)$ is generated by Eisenstein
series
its dimension is equal to the number of cusps $r$.
For $k=3,5,\dots$ we have
\begin{eqnarray*}
\dim H^0(\Gamma,H^{-k/2}_{-\infty})&=&\dim({\cal A}_{k+1} \oplus \bar{{\cal A}}_{k+1})\\
\dim H^1(\Gamma,H^{-k/2}_{-\infty})&=&\dim({\cal S}_{k+1} \oplus \bar{{\cal S}}_{k+1})+r\
{}.
\end{eqnarray*}
Employing $\dim {\cal A}_{k+1} = \dim \bar{{\cal A}}_{k+1} = k(g-1) +r (k+1)/2$
and $\dim {\cal S}_{k+1} = \dim \bar{{\cal S}}_{k+1} = k(g-1) + r(k-1)/2$ (see
\cite{shimura94}, Thm. 2.23) we recover the result of Proposition \ref{wwqqq}.
We now consider the case $k=1$. In this case both
\begin{eqnarray}&&\bar{\partial} :S_\infty{\cal K}^{1}_Y(B)\rightarrow
S_\infty{\cal K}^{0}_Y(B)\ , \\
&&\partial :S_\infty{\cal K}^{-1}_Y(B)\rightarrow S_\infty{\cal K}^0_Y(B)
\end{eqnarray}
have a one-dimensional range spanned by the constant function.
Note the $r$-dimensional space $ S_\infty{\cal K}^0_Y(B)$ is generated by the
regular Eisenstein series and the constant function.
We obtain for $k=1$
\begin{eqnarray*}
\dim H^0(\Gamma,H^{-1/2}_{-\infty})&=&\dim({\cal A}_2 \oplus \bar{{\cal A}}_2)+1\\
\dim H^1(\Gamma,H^{-1/2}_{-\infty})&=&\dim({\cal S}_2 \oplus \bar{{\cal S}}_2)+r-1\ .
\end{eqnarray*}
Using $\dim {\cal A}_2=\dim \bar{{\cal A}}_2=g-1+r$, $ \dim {\cal S}_1 = \dim \bar{{\cal S}}_2 =
g$ (see \cite{shimura94}, Thm. 2.23)
we again recover the results of Proposition \ref{wwqqq}.
We finish this section with a discussion of the relation of the
sequence (\ref{wert}) with the Eichler homomorphism
$E:{\cal S}_{k+1} \rightarrow H^1(\Gamma,F_k)$
(see e.g. \cite{shimura94}).
We first recall its definition.
If we restrict $F_k$ to the maximal compact subgroup $K=S^1\subset G$,
then it decomposes as $(F_k)_{|K}=\sum_{m=1}^k {\bf C}_{ m-(k+1)/2} $, where
${\bf C}_l$ is the representation of $S^1$ on ${\bf C}$ given by $z\mapsto z^l$.
Note that ${\tt K}^l=G\times_K {\bf C}_{-l}$. The map
$(gK,f)\mapsto [g,g^{-1}f]$, $g\in G$, $f\in F_k$ defines an isomorphism
$$G/K\times F_k\stackrel{\sim}{\rightarrow} G\times_K (F_k)_{|K}=\oplus_{m=1}^k
{\tt K}^{ m-(k+1)/2}\ .$$
We denote the canonical $G$-equivariant embedding
${\tt K}^{\pm(k-1)/2}\hookrightarrow X\times F_k\cong L^2\otimes F_k$
by $j_\pm$, where the second identification is given by the Hodge-$*$-operator
and $L^l$ is the bundle of $l$-forms on $X$. Taking the tensor product with
${\tt K}^{\pm 1}$ we obtain embeddings
$i_\pm:{\tt K}^{\pm(k+1)/2}\hookrightarrow L^1\otimes F_k$.
If $f\in {\cal S}_{k+1}={\cal K}^{(k+1)/2}(B)_{cusp}$, then $i_+(f)$ is a closed
form and represents $E(f)$.
The long exact cohomology sequence associated to (\ref{wert}) induces a
boundary map
$$\delta:H^0(\Gamma,D^+_k\oplus D^-_k)\rightarrow H^1(\Gamma,F_k)\ .$$
\begin{lem}\label{eichler}
The map $\delta$ restricted to the cusp forms coincides with $E$.
\end{lem}
{\it Proof.$\:\:$}
In order to compute $\delta$ we need a $\Gamma$-acyclic resolution of
(\ref{wert})
as a sequence.
One possibility is the following:
\begin{equation}{\scriptsize \label{saur}
\begin{array}{ccccccccc}
&&0&&0&&&&\\
&&\uparrow&&\uparrow&&&&\\
0&\rightarrow&S_\infty{\cal L}^2\otimes
F_k&\stackrel{=}{\rightarrow}&S_\infty{\cal L}^2\otimes F_K&\rightarrow&0&&\\
&&d\uparrow&&a\uparrow&&\uparrow&&\\
0&\rightarrow&S_\infty{\cal L}^1\otimes F_k&\rightarrow&S_\infty{\cal L}^1\otimes
F_K\oplus S_\infty{\cal K}^{(k-1)/2}\oplus S_\infty{\cal K}^{-(k-1)/2}&\rightarrow
&S_\infty{\cal K}^{(k-1)/2}\oplus S_\infty{\cal K}^{-(k-1)/2}&\rightarrow&0\\
&&d\uparrow&&b\uparrow& &c\uparrow&&\\
0&\rightarrow&S_\infty{\cal L}^0\otimes F_k&\rightarrow&S_\infty{\cal L}^0\otimes
F_K\oplus S_\infty{\cal K}^{(k+1)/2}\oplus S_\infty{\cal K}^{-(k+1)/2}&\rightarrow
&S_\infty{\cal K}^{(k+1)/2}\oplus S_\infty{\cal K}^{-(k+1)/2}&\rightarrow&0\\
&&\uparrow&&\uparrow& &\uparrow&&\\
0&\rightarrow& F_k&\rightarrow &H^{k/2}&\rightarrow &D^+_k\oplus D^-_k
&\rightarrow&0\\
&&\uparrow&&\uparrow&&\uparrow&&\\
&&0&&0&&0&&\ .
\end{array}}
\end{equation}
Here the maps are given by
\begin{eqnarray*}
a&:=&d-j_+-j_-\\
b&:=&\left(\begin{array}{ccc}d&i_+&i_-
\\0&\bar{\partial}&0\\0&0&\partial\end{array}\right)\\
c&:=&\bar{\partial}\oplus\partial\ .
\end{eqnarray*}
The horizontal maps are the obvious embeddings and projections, respectively.
We leave the verification of the commutativity of the diagram and the exactness
of the middle column to the reader.
The boundary map $\delta$ is now obtained by the usual diagram chasing. Let
$[\alpha]\in H^0(\Gamma,D^+_k)$ be represented by a holomorphic $\Gamma$-
invariant form $\alpha\in{}^\Gamma S_\infty {\cal K}^{(k+1)/2}$. Then $\delta
[\alpha] \in H^1(\Gamma,F_k)$
is represented by the closed form $ i_+(\alpha)\in S_\infty {\cal L}^1\otimes F_k$.
Comparing this with the definition of the Eichler map $E$ given above
we finish the proof of the lemma.
\hspace*{\fill}$\Box$ \newline\noindent
\bibliographystyle{plain}
|
\section{Introduction}
When we construct a Grand Unified Theory(GUT) based on SO(10) \cite{so10},
in general, we have singlet fermions under the Standard Model(SM)
-what we call right-handed neutrino. Under the SM right-handed
neutrinos can have Majorana masses because they are singlet.
Then the scale of the right-handed neutrinos($\equiv M_{\nu_R}$)
is expected
to be a scale below which the SM is realized.
It is well known that in the Minimal Supersymmetric
Standard Model (MSSM) the present experimental values
of gauge couplings are successfully unified at a unification scale
$M_U \simeq 10^{16}$GeV \cite{Amal}. This fact implies
that if we would like to consider the gauge unification,
it is favorable that the symmetry of the GUT breaks down
to that of the SM at the unification scale.
In this case the scale of the right-handed neutrinos $M_{\nu_R}$ is
expected to be the unification scale $M_U$. This means also
there is no intermediate scale between
the Supersymmetry(SUSY) breaking scale and the unification scale.
On the other hand
it is said that $M_{\nu_R} \sim 10^{10-12} {\rm GeV}$\cite{yana}.
The experimental data on the deficit
of the solar neutrino can be explained by
the Mikheyev-Smirnov-Wolfenstein(MSW) solution \cite{MSW}.
According to one of the MSW solutions,
the mass of the muon neutrino seems to be
$m_{ \nu_\mu}\simeq10^{-3}$ eV.
Such a small mass can be led by the seesaw
mechanism \cite{seesaw}: A muon neutrino can
acquire a mass of O($10^{-3}$) eV
if the Majorana mass of the right-handed muon neutrino
is about $10^{12}$ GeV.
Then how can the right-handed neutrinos
acquire mass of about $10^{12}$ GeV?
It was our question in our previous paper \cite{BST},
because if we take the prediction of the MSSM serious,
$M_{\nu_R}$ is expected to be $M_U \simeq 10^{16}$ GeV.
Our point of view was that it is more natural to consider
that one energy scale corresponds to a dynamical phenomenon,
for instance a symmetry breaking.
Mass is given by a renormalizable coupling
is also the crucial point of our view.
This idea is consistent with the survival hypothesis.
Thus we were led to a possibility that a certain group breaks
down to the SM group at the intermediate scale at which right-handed
neutrinos gain mass through a {\it renormalizable coupling}.
In the previous work
we have searched possibilities to construct such a SUSY
SO(10) GUT with an intermediate symmetry.
We have a possibility to construct a SUSY SO(10) GUT with an intermediate
symmetry ${\rm SU(2)}_L\otimes{\rm SU(2)}_R\otimes
{\rm U(1)}_{B-L}\otimes {\rm SU(3)}_C$ ($\equiv G_{2231}$)
\footnote{We use a notation $G_{lmn...}$ to represent
${\rm SU(l)}\otimes{\rm SU(m)}\otimes{\rm SU(n)}...$.If $l=1$,
it means U(1).}
which breaks down to the SM group at an intermediate
scale $M_{\nu_R} \sim 10^{10-12} {\rm GeV}$ where
a right-handed neutrino gains mass.
In such a scenario, as we showed in the previous work,
to make the model consistent with the gauge unification
we have to introduce several multiplets at the intermediate region
between the GUT scale and the intermediate scale,
in addition to ordinary matters, three generations of quarks and
leptons and a pair of so-called Higgs doublets.
Although we showed a possibility to construct
a SUSY SO(10) GUT with an intermediate symmetry $G_{2231}$
it is not trivial whether
it is actually possible to construct such a GUT
since there are many extra fields in the intermediate region.
We did not show the superpotential for the theory explicitly
which can realize such a scenario that we have suggested
in ref.\cite{BST}.
The purpose of this paper is to show an explicit form of
a superpotential for a SUSY SO(10) GUT to construct a
SUSY SO(10) GUT whose symmetry breaks
down to $G_{2231}$ at a GUT scale $M_U$ and $G_{2231}$
breaks down to the SM symmetry at the intermediate scale
$M_{\nu_R}$.
We give the scenario and the model briefly in sect. 2
where we give a candidate for the matter content in the intermediate
region (the spectrum (\ref{eq:spectrum})).
Then in sect. 3
we show the most general form of the superpotential
and a symmetry breaking condition
as preparation for our analysis.
In sect. 4 first we calculate parameters of the theory,
namely parameters appearing in the superpotential, which produce
the spectrum (\ref{eq:spectrum})
at the intermediate region. Then
we show the exact parameters which realize the MSSM
below $M_{\nu_R}$. Finally (in sect. 5)
we give a summary and discussion.
\section{Scenario and Model}
\subsection{Scenario}
We construct a SUSY SO(10) GUT whose symmetry breaks
down to $G_{2231}$ at a GUT scale $M_U$ and $G_{2231}$
breaks down to the SM symmetry at the intermediate scale
$M_{\nu_R}$. When $G_{2231}$ breaks down to the SM symmetry
the right-handed neutrinos gain mass through
a {\it renormalizable Yukawa coupling}.
Let us first recapitulate the content of the previous work\cite{BST}.
To achieve the gauge unification in the scenario
we have to introduce a certain combination of multiplets.
Because in our model right-handed neutrinos
acquire mass of O($M_{\nu_R}$) via a renormalizable
Yukawa coupling by the symmetry breaking,
we have to introduce at least a pair of
(1,3,1,6) + h.c multiplet under $G_{2231}$.
We adopt the
normalization for $U(1)_{B-L}$, $T^{15}_4 = {\rm diag}(-1,-1,-1,3)$.
When we introduce only (1,3,1,6) + h.c multiplet in addition to the
ordinary matter,
gauge couplings do not unify. Then we have to introduce
certain matter content under $G_{2231}$.
We found very many
candidates for matter content in the intermediate region
between the GUT scale and the intermediate scale
which lead the gauge unification.
Among them we showed two candidates for the matter content as
the simplest example.
In this article we use another candidate
which was not showed in the previous
paper. In the examples appearing in it
a (1,3,1,0) multiplet under $G_{2231}$ was not included.
In constructing a GUT following the idea,
however, we have to introduce a
(1,3,1,0) multiplet in the intermediate region.
The reason why we have to introduce a (1,3,1,0) multiplet
is stated in the appendix A.
Thus we have to use another candidate for matter content.
The matter content
other than quarks and leptons (including right-handed neutrinos),
which we assume
survive until $G_{2231}$ breaks down to the SM group
at the intermediate scale,
are given below.
\begin{eqnarray}
\begin{array}{ccccl}
(1,3,1,-6) & 1 & (1,3,1,6) & 1
& {\rm responsible\,for }\nu_R {\rm mass}\\
(2,2,1,0) & 2 && & {\rm ordinary\, Higgs\, doublets}\\
(2,1,3,-1) & 1 & (2,1,\overline{3},1) & 1& \\
(2,1,1,3) & 1 & (2,1,1,-3) & 1 &\\
(1,3,1,0) & 1 &&&\\
(1,1,8,0) & 1 &&&
\end{array}
\label{eq:spectrum}
\end{eqnarray}
In this list,
for example, (1,3,1,-6) 1 stands
for that the representation
of the matter under $G_{2231}$ is (1,3,1,-6) and its number is one.
When we have the particle content
listed here in the intermediate region
the unified coupling $\alpha_U (M_U)$ is about 1/18
if we take the intermediate scale to be $10^{12}$ GeV.
As a candidate which contains (1,3,1,0), this candidate
leads the smallest unified coupling.
In our scenario, at the GUT scale $M_U$ where SO(10) breaks
down to $G_{2231}$ almost of all particles
have mass of O($M_U$) while the particles listed in
(\ref{eq:spectrum}) as well as quarks and leptons
are left massless. Then at the intermediate scale
where $G_{2231}$ breaks down to $G_{231}$ the SM group
all extra multiplets in (\ref{eq:spectrum})
besides a pair of Higgs doublets
and right-handed neutrinos have mass of O($M_{\nu_R}$),
that is, they decouple from the spectrum. Thus below
$M_{\nu_R}$ the MSSM is realized.
\subsection{Model}
\subsubsection{Matter content}
To have multiplets (\ref{eq:spectrum}) and quarks/leptons
at the intermediate region we introduce
following multiplets of SO(10).
\begin{eqnarray}
\begin{array}{rcrl}
&&{\rm SO(10)}&G_{2231}\\
H&:&10&(2,2,1,0),...\\
A&:&45&(1,3,1,0),(1,1,8,0),...\\
\Phi&:&126&(1,3,1,-6),(2,2,1,0),...\\
\overline{\Phi}&:&\overline{126}&(1,3,1,6),(2,2,1,0),...\\
\Delta&:&210&(1,3,1,0),(1,1,8,0),...\cr
\Psi_{i=1\sim 4}&:&16&(2,1,3,-1),(2,1,1,3),{\rm quarks/leptons}\\
\overline{\Psi}&:&\overline{16}&(2,1,\overline{3},1),(2,1,1,-3),...\\
\end{array}
\label{eq:so10multiplets}
\end{eqnarray}
In this list numbers in the column of SO(10) means SO(10)
representation. In the last column we show what representation
in (\ref{eq:spectrum}) is contained in the corresponding SO(10)
multiplet.
By the requirement that the right-handed neutrinos get mass through
a renormalizable coupling, we introduce 126 and $\overline{126}$.
As a candidate of ordinary Higgs doublets 10 is introduced.
There are other candidates for ordinary Higgs doublets
in 126 and $\overline{126}$. Then the ordinary Higgs doublets
will be a mixture of these three.
To break SO(10) to the SM group
via $G_{2231}$, namely to have the intermediate symmetry
$G_{2231}$, we use 45 and 210\footnote{
Using only 210 it is impossible to break SO(10) to $G_{231}$
through $G_{2231}$ \cite{Lee}.
We can break SO(10) to the SM group via $G_{2231}$
using 45 + 54. In this case
if there is no multiplet which
contains (1,3,1,0) other than 45 (3,1,1,0)
is also massless. The reason is that
mass terms for (1,3,1,0) and (3,1,1,0) come from
the mass term of 45 and the vacuum expectation value of 54
through the coupling $45^2 54$
and because of D parity\cite{DP} they are same as each other's.
Thus we can get rid of the possibility of using 45+54.}.
These also contain (1,3,1,0) and (1,1,8,0).
4 16's and 1 $\overline{16}$ represent 4 generation +
1 anti-generation. The reason why we introduce a pair of
16 and $\overline{16}$ is that they contain
(2,1,3,-1) + h.c and (2,1,1,3) + h.c .
At this stage
the matter content (\ref{eq:so10multiplets}) is just a
candidate which may realize our scenario.
As we will see, we can write down the superpotential
with these matter which realize our idea.
\subsubsection{Singlets under the SM group}
In the SO(10) multiplets (\ref{eq:so10multiplets})
there are many singlets under the SM symmetry
(see appendix B for the meaning of subscripts 1,...,0):
\begin{eqnarray}
\begin{tabular}{lllr}
Field&:&Component&Little Group\cr
A&:&$a_{12+34+56}\equiv \alpha $&$G_{2231}$\cr
&:&$a_{78+90}\equiv \beta $&$G_{241}$\cr
$\Phi$&:&$\phi_{1-2i,3-4i,5-6i,7-8i,9-0i}
\equiv \phi$&SU(5)\cr
$\overline{\Phi}$&:&$\overline{\phi}_{1+2i,3+4i,5+6i,7+8i,9+0i}
\equiv \overline{\phi}$&SU(5)\cr
$\Delta$&:&$\delta_{7890}\equiv a$&$G_{224}$\cr
&&$\delta_{1234+3456+5612}\equiv b$&$G_{2231}$\cr
&&$\delta_{(12+34+56)(78+90)}\equiv c$&$G_{2311}$\cr
$\Psi_{i=1\sim 4}$&:&$\psi_{i=1\sim 4}$&SU(5)\cr
$\overline{\Psi}$&:&$\overline{\psi}$&SU(5)\cr
\end{tabular}
\label{eq:vevtable}
\end{eqnarray}
where $a, b,$... stand for vacuum expectation values (VEV)
of the corresponding fields. Little group means
a remaining symmetry
when only a corresponding component has a VEV.
For example, when only $a$ gets a VEV SO(10) breaks down to
$G_{224}$.
Among them $a, b$ and $\alpha$ are $G_{2231}$ singlets
and hence their order of magnitudes is expected to
be the GUT scale $M_U \sim 10^{16}$
GeV. By assumption that SO(10) breaks down to $G_{2231}$ at the GUT scale,
$b$ or $\alpha$ must be of order $M_U$.
Others must be of order at most $M_{\nu_R} \equiv M_U \epsilon$
by assumption
because they are not $G_{2231}$ singlets.
Also $\overline\phi$ is required to be of order $M_{\nu_R}$,
\begin{equation}
\overline\phi
\sim M_{\nu_R} ( = M_U \epsilon)
\label{eq:Condphi}
\end{equation}
because it gives masses to the right-handed neutrinos.
Of course, as we will see, there are constraints among
VEVs in addition to the well known constraints - F-flat
and D-flat condition
because we require certain multiplets must have mass
of O($M_{\nu_R}$).
\section{Preparation}
\subsection{Superpotential}
With the multiplets (\ref{eq:so10multiplets})
the most general form
of the superpotential $W$ is written as
\begin{eqnarray}
W = W_{mass} + W_{int} + W_{\Psi} .
\label{eq:superpotential}
\end{eqnarray}
$W_{mass}$ consists of the most general bilinear terms:
\begin{eqnarray}
W_{mass} = {1\over 2}M_H H^2 + M_\Phi \overline{\Phi} \Phi
+{1\over 2}M_\Delta \Delta^2+{1\over 2} M_A A^2
+ M_\Psi \overline{\Psi} \Psi_4 .
\label{eq:supermass}
\end{eqnarray}
We define only $\Psi_4$ has a mass term with $\overline{\Psi}$,
because
by a redefinition of $\Psi_4$ , namely by a rotation among
$\Psi_{i=1-4}$,
it is possible that only the new $\Psi_4$
has a mass term with $\overline\Psi$.
We require all mass parameters are O($M_U$) because
$M_U$ is the natural order for them.
$W_{int}$ has the most general interaction terms without
16 and $\overline{16}$:
\begin{eqnarray}
W_{int} &=& Y_{H\Phi\Delta}H\Phi\Delta
+ Y_{H\overline{\Phi}\Delta}H\overline{\Phi}\Delta
+ {1\over {3!}} Y_\Delta \Delta^3
+ Y_{\Phi\Delta}\overline{\Phi} \Delta \Phi
+Y_{\Phi A}\overline{\Phi} A \Phi \nonumber\\
& + &{1\over 2} Y_{\Delta A^2} A^2 \Delta
+{1\over 2} Y_{\Delta^2A} A \Delta^2 .
\label{eq:superyukawa}
\end{eqnarray}
We require all Yukawa couplings are at most O(1).
More exactly, as an expansion parameter for the perturbation
we require they are at most O(1). As a expansion parameter for the
perturbation they appear with multiplied by a certain overall
factor.
The overall factors for each couplings are given in appendix B.3.
Finally, $W_\Psi$ represents the most general interaction terms
with 16 and $\overline{16}$.
\begin{equation}
\begin{array}{ccl}
\displaystyle
W_\Psi& = &
\displaystyle
\sum^4_{i=3} Y_{\Psi\Delta i} \overline{\Psi}\Delta\Psi_i
+ \sum^4_{i=2} Y_{\Psi Ai} \overline{\Psi} A \Psi_i
+ \sum_{ij} y_{ij} \Psi_i \Psi_j \overline\Phi
+ y' \overline{\Psi} \,\overline{\Psi} \Phi \\
&
\displaystyle
+ &
\displaystyle
\sum_{ij} \tilde{y}_{ij} \Psi_i \Psi_j H
+ \tilde{y}' \overline{\Psi}\, \overline{\Psi} H .\hfill
\end{array}
\label{eq:superpsi}
\end{equation}
By the same reason that only $\Psi_4$ has a mass term with
$\overline\Psi$, only $\Psi_{3,4}$ have couplings with $\Delta$
and only $\Psi_{2,3,4}$ have couplings with $A$.
To see in which direction the gauge group SO(10) can break down
we examine the D-term and the F-term conditions.
\subsection{D-flat condition}
To keep the SUSY all D-terms must be zero up to SUSY braking scale:
$$
\Phi^\dagger T^a_\Phi \Phi +
\overline\Phi^\dagger T^a_{\overline\Phi} \overline\Phi +
\sum_i \Psi^\dagger_i T^a_\Psi \Psi_i
+\overline\Psi^\dagger T^a_{\overline\Psi} \overline\Psi
+\Delta^\dagger T^a_\Delta \Delta + A^\dagger T^a_A A = 0 .
$$
Since the D-term for real representations automatically
vanishes \cite{HeMe,BuDeSa},
\begin{eqnarray}
2 (|\phi|^2 - |\overline\phi|^2) +
(\sum_{i=1}^4 |\psi_i|^2 - |\overline\psi|^2) = 0
\label{eq:Dphipsi}
\end{eqnarray}
must be satisfied. The factor 2 reflects the difference of
U(1) charge which corresponds to a broken generator.
Later we put $\psi_i$'s and $\psi$ zero. In this case
\begin{equation}
|\phi|^2 - |\overline\phi|^2 = 0 .
\label{eq:Dphi}
\end{equation}
\subsection{F-flat condition}
First we examine the F-flat condition for 16 and $\overline{16}$
with a mass term for $(1,2,1,-3) + {\rm h.c}$ component because
the singlet components of 16 and $\overline{16}$
are contained in it and therefore
there is a relation between the mass term and the F-flat condition.
By such an examination we see
that $\psi_i$ and $\overline\psi$ should be zero though it is not a strict
reason for it.
The F-flat condition for 16 and $\overline{16}$ are as follows:
(See appendix B to know how to calculate Clebsch-Gordan (CG)
coefficient)
\begin{eqnarray}
{\partial W \over \partial \psi_1}=
2 \sum_{j=1}^4 y_{1j} \psi_j \overline{\phi}=0,
\label{eq:Fpsi1}
\end{eqnarray}
\begin{eqnarray}
{\partial W \over \partial \psi_2}=
2 \sum_{j=1}^4 y_{2j} \psi_j \overline{\phi}
-Y_{\Psi A2}(\sqrt{6} i \alpha + 2 i \beta) \overline\psi=0,
\label{eq:Fpsi2}
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial \psi_3} &
=&
\displaystyle
2 \sum_{j=1}^4 y_{3j} \psi_j \overline{\phi}
-Y_{\Psi A3}(\sqrt{6} i \alpha + 2 i \beta) \overline\psi
\nonumber \\
&-&
\displaystyle
Y_{\Psi\Delta 3}(2 \sqrt{6} a + 6 \sqrt{2} b + 12 c) \overline\psi
\label{eq:Fpsi3}
\\&
\displaystyle
=&
\displaystyle
0,\nonumber
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial \psi_4}&=&
\displaystyle
2 \sum_{j=1}^4 y_{4j} \psi_j \overline{\phi}
-Y_{\Psi A4}(\sqrt{6} i \alpha + 2 i \beta) \overline\psi
\nonumber \\
&-&
\displaystyle
Y_{\Psi\Delta 4}(2 \sqrt{6} a + 6 \sqrt{2} b + 12 c) \overline\psi
+M_\Psi \overline\psi
\label{eq:Fpsi4} \\
&=&0,\nonumber
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial \overline\psi}&=&
\displaystyle
2\, y'\overline{\psi} \phi+
\sum_{i=2}^4
- Y_{\Psi Ai}(\sqrt{6} i \alpha + 2 i \beta) \psi_i\nonumber \\
&-&
\displaystyle
\sum_{j=3}^4
Y_{\Psi\Delta i}(2 \sqrt{6} a + 6 \sqrt{2} b + 12 c) \psi_i
+M_\Psi \psi_4
\label{eq:Fbarpsi} \\
& =& 0. \nonumber
\end{eqnarray}
By the way in the intermediate region where $G_{2231}$ is realized,
$\beta = c = 0$ and the mass term for (1,2,1,-3)+h.c is given by
\begin{equation}
{\partial^2 W \over {\partial \psi_i \partial \overline\psi}}
=\pmatrix{0\cr
- \sqrt{6} i Y_{\Psi A2} \alpha\cr
-\sqrt{6} i Y_{\Psi A3} \alpha
- 2 \sqrt{6} Y_{\Psi\Delta 3} (a + \sqrt{3} b)\cr
-\sqrt{6} i Y_{\Psi A4} \alpha
- 2 \sqrt{6} Y_{\Psi\Delta 4} (a + \sqrt{3} b)
+M_\Psi
}.
\label{eq:m1213}
\end{equation}
If $\phi, \overline\phi, \psi_i, \overline\psi \sim O(\epsilon)$,
using
F-flat conditions (\ref{eq:Fpsi2})\, -\, (\ref{eq:Fpsi4}),
all of elements of the mass term for (1,2,1,-3)+h.c,
(\ref{eq:m1213}), are calculated to be of order $M_{\nu_R}$.
This, however, contradicts with
the mass spectrum (\ref{eq:spectrum}).
Though we may be able to make the some elements of the mass term
O($M_U$), for example, by making
$\overline\psi$ O($\epsilon^2$)
(with an appropriate value of $\psi_i,\overline\phi \sim
O(\epsilon)$), we put $\psi_i$ and
$\overline\psi$ zero since what we try to do is to show a possibility
of SUSY SO(10) GUT with an intermediate scale
and to take $\psi_i = \overline\psi = 0$ as the solution of
the F-flat conditions for 16
and $\overline{16}$ is the easiest way of it.
Then other F-term conditions are as follows:
\begin{eqnarray}
{\partial W \over \partial a}=
24\,{\sqrt{2}}\,i Y_{\Delta^2A}\,\alpha b -
{{ Y_{\Delta A^2}\,\beta^2 }\over {2\,\sqrt{6}} } +
{{{ Y_\Delta}\,{c^2}}\over {12\,{\sqrt{6}}}} + { M_\Delta} \,a +
{{{ Y_{\Phi\Delta}}\,{ \overline{\phi}}\phi}\over {10\,{\sqrt{6}}}}=0,
\label{eq:Fa}
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial b}&=&
24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,a { \alpha} -
{{{ Y_{\Delta A^2}}\,{{{ \alpha}}^2}}\over {3\,{\sqrt{2}}}} +
{{{ Y_\Delta}\,{b^2}}\over {18\,{\sqrt{2}}}} \nonumber \\
&+&
\displaystyle
24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,{ \beta}\,c +
{{{ Y_\Delta}\,{c^2}}\over {18\,{\sqrt{2}}}} + { M_\Delta}\,b +
{{{ Y_{\Phi\Delta}}\,\phi\,{ \overline{\phi}}}\over
{10\,{\sqrt{2}}}}
\label{eq:Fb}\\
&=&0,\nonumber
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial c}&=&
-{{{ Y_{\Delta A^2}}\,{ \alpha}\,{ \beta}}\over {{\sqrt{6}}}} +
24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,b\,{ \beta} +
{{{ Y_\Delta}\,a c}\over {6\,{\sqrt{6}}}} \nonumber \\
&+&
\displaystyle
16\,{\sqrt{6}}\,i\,{ Y_{\Delta^2A}}\,{ \alpha}\,c +
{{{ Y_\Delta}\,b\,c}\over {9\,{\sqrt{2}}}} + { M_\Delta} c+
{{{ Y_{\Phi\Delta}}\,\phi\,{ \overline{\phi}}}\over {10}}
\label{eq:Fc}\\
&=&0,\nonumber
\end{eqnarray}
\begin{eqnarray}
\displaystyle
{\partial W \over \partial \alpha}&=&
24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,a\,b -
{{{\sqrt{2}}\,{ Y_{\Delta A^2}}\,{ \alpha}\,b}\over 3} -
{{{ Y_{\Delta A^2}}\,{ \beta}\,c}\over {{\sqrt{6}}}}\nonumber\\ &+&
\displaystyle
8\,{\sqrt{6}}\,i\,{ Y_{\Delta^2A}}\,{c^2} + { M_A}\,\alpha +
{{{\sqrt{6}}\,{ Y_{\Phi A}}\,\phi\,{ \overline{\phi}}}\over 10}
\label{eq:Falpha}\\
& =&0,\nonumber
\end{eqnarray}
\begin{eqnarray}
{\partial W \over \partial \beta}=
-{{{ Y_{\Delta A^2}}\,a\,{ \beta}}\over {{\sqrt{6}}}} -
{{{ Y_{\Delta A^2}}\,{ \alpha}\,c}\over {{\sqrt{6}}}} +
24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,b\,c + M_A \,\beta +
{{{ Y_{\Phi A}}\,\phi\,{ \overline{\phi}}}\over 5}=0,
\label{eq:Fbeta}
\end{eqnarray}
\begin{eqnarray}
{\partial W \over \partial \phi}=
\left( { Y_{\Phi A}}\,\left( {{{\sqrt{6}}\,{ \alpha}}\over 10} +
{{{ \beta}}\over 5} \right) +
{ Y_{\Phi\Delta}}\,
\left( {a\over {10\,{\sqrt{6}}}} + {b\over {10\,{\sqrt{2}}}} +
{c\over {10}} \right) + { M_\phi} \right) \,{ \overline{\phi}}
=0.
\label{eq:Fphi}
\end{eqnarray}
\section{Analysis}
The purpose of this paper is to give the input parameters appearing
in the superpotential (\ref{eq:superpotential}). Though
VEVs listed in (\ref{eq:vevtable})
are functions of the input parameters
we will express them in the term of the VEVs since
we know the desirable values of the VEVs.
\subsection{First Step}
First we check whether it is possible to break $SO(10)$ down to
$G_{2231}$ consistently with the requirement that
the spectrum (\ref{eq:spectrum})
remains massless up to O($\epsilon$) $\sim$ O($M_{\nu_R}/M_U$).
\subsubsection{Multiplets under $G_{2231}$}
First we show what multiplets exist in the SO(10) multiplets
(\ref{eq:so10multiplets}).
\begin{eqnarray}
\begin{array}{lrrrr}
{\rm Multiplet\, under}\, G_{2231}&{\rm under SO(10),\, contained in}&
{\rm NG1}&{\rm NG2}\cr
\hline
(2,2,1,0)&10,126,\overline{126}&&\cr
(1,1,3,2)+{\rm h.c}&10,126,\overline{126}&&\cr
(3,1,1,0)&45,210&&\cr
(1,3,1,0)&45,210&&\tilde z\cr
(1,1,3,-4)+{\rm h.c}&45,210&x&\tilde x\cr
(1,1,8,0)&45,210&&\cr
(2,2,3,2)+{\rm h.c}&45,210&y&\tilde y\cr
(3,1,1,6)+{\rm h.c}&126+\overline{126}&&\cr
(3,1,3,2)+{\rm h.c}&126+\overline{126}&&\cr
(3,1,6,-2)+{\rm h.c}&126+\overline{126}&&\cr
(1,3,1,-6)+{\rm h.c}&126+\overline{126}&&\tilde z\cr
(1,3,\overline{3},-2)+{\rm h.c}&126+\overline{126}&&\tilde x\cr
(1,3,\overline 6,2)+{\rm h.c}&126+\overline{126}&&\cr
(2,2,3,-4)+{\rm h.c}&126,\overline{126}&&\tilde y\cr
(2,2,8,0)+{\rm h.c}&126,\overline{126}&&\cr
(3,1,3,-4)+{\rm h.c}&210&&\cr
(1,3,3,-4)+{\rm h.c}&210&&\tilde x\cr
(3,1,8,0)+{\rm h.c}&210&&\cr
(1,3,8,0)+{\rm h.c}&210&&\cr
(2,2,1,6)+{\rm h.c}&210&&\cr
(2,1,3,-1)+{\rm h.c}&16+\overline{16}&&\tilde y\cr
(1,2,\overline 3,1)+{\rm h.c}&16+\overline{16}&&\tilde x\cr
(2,1,1,3)+{\rm h.c}&16+\overline{16}&&\cr
(1,2,1,-3)+{\rm h.c}&16+\overline{16}&&\tilde z\cr
\end{array}
\label{eq:multiplets}
\end{eqnarray}
In this table NG1 means a Nambu-Goldstone (NG) mode associated with the
breakdown of SO(10) to $G_{2231}$.
An NG mode associated with the SO(10) breaking down to
the SM group $G_{231}$ is contained in a multiplet with $\tilde x,
\tilde y$ and $\tilde z$ in the column NG2. Under $G_{231}$,
certain components of the multiplets with
$\tilde x$ ($\tilde y, \tilde z$)
have same quantum number and mix with each other.
One of combinations of $\tilde x$ ($\tilde y, \tilde z$)
is massless which is swallowed by a gauge boson.
There are also singlets of $G_{2231}$ which we denote $a, b$
and $\alpha$.
\subsubsection{F-flat condition}
In the intermediate region $c ,\beta , \phi =0$.
And hence the F-term conditions (\ref{eq:Fa}) - (\ref{eq:Fphi})
are reduced to
\begin{eqnarray}
{\partial W \over \partial a}=
24\,i\,{\sqrt{2}}\,{ Y_{\Delta^2A}}\,{ \alpha}\,b
+ { M_\Delta} a = 0 ,
\label{eq:Fa2231}
\end{eqnarray}
\begin{eqnarray}
{\partial W \over \partial b}=
24\,i\,{\sqrt{2}}\,a\,{ Y_{\Delta^2A}}\,{ \alpha} -
{{{ Y_{\Delta A^2}}\,{{{ \alpha}}^2}}\over {3\,{\sqrt{2}}}} +
{{{ Y_\Delta}\,{b^2}}\over {18\,{\sqrt{2}}}} +
{ M_\Delta} b = 0,
\label{eq:Fb2231}
\end{eqnarray}
\begin{eqnarray}
{\partial W \over \partial \alpha}=
24\,i\,{\sqrt{2}}\,{ Y_{\Delta^2A}} a\,b -
{{{\sqrt{2}}\,{ Y_{\Delta A^2}}\,{ \alpha}\,b}\over 3}
+ { M_A} { \alpha} = 0.
\label{eq:Falpha2231}
\end{eqnarray}
\subsubsection{Tuning of parameters}
{}From now on
as we stated at the top of this section,
we express the input parameters in the term of the VEVs.
Using the F-flat conditions
(\ref{eq:Fa2231}) and (\ref{eq:Falpha2231}),
$M_\Delta$ and $M_A$ are expressed as follows:
\begin{eqnarray}
{ M_{\Delta}} = M_\Delta (Y_{\Delta^2A} , a , b , \alpha) =
{{-24\,{\sqrt{2}}\,i\,{ Y_{\Delta^2A}}\,{ \alpha}\,b}\over a},
\label{eq:Md1}
\end{eqnarray}
\begin{eqnarray}
M_A = M_A(Y_{\Delta^2A} , Y_{\Delta A^2} , a , b ,
\alpha) =
{-72\,\sqrt{2}\,i\,Y_{\Delta^2A} a\,b +
\sqrt{2}\,Y_{\Delta A^2}\,\alpha\,b
\over {3 \alpha}}.
\label{eq:Ma1}
\end{eqnarray}
There is an additional constraint which is obtained by
eliminating $M_\Delta$ from equations (\ref{eq:Fa2231}) and
(\ref{eq:Fb2231}):
\begin{eqnarray}
-24\,{\sqrt{2}}\,i\,Y_{\Delta^2A} a^2\,{ \alpha} +
{ {Y_{\Delta A^2}\,a\,\alpha^2}\over {3\,\sqrt{2}} } -
{{Y_{\Delta}\,a\,b^2}\over {18\,\sqrt{2}}} +
24\,\sqrt{2}\,i\,Y_{\Delta^2A}\,\alpha\,b^2
=0.
\end{eqnarray}
We can interpret that
this constraint with (\ref{eq:Md1}) and (\ref{eq:Ma1})
is equivalent with that determinant of the mass matrix for (1,1,3,-4)
($\equiv M(1,1,3,-4)$
an explicit form is given at appendix C)
vanishes because
(1,1,3,-4) is an NG mode and hence
when we substitute VEVs into the mass matrix for it
there must be one massless mode which mean the determinant
vanishes.
\begin{eqnarray}
\displaystyle
\left.
\begin{array}{cl}
&\displaystyle
\det M(1,1,3,-4)\\
\displaystyle
=&
\displaystyle
M_A\,M_\Delta +
{{Y_\Delta \,M_A \,b}\over {18\,\sqrt{2}}} -
{{Y_{\Delta A^2}\,M_\Delta\,b}\over {3\,\sqrt{2}}}\\
\displaystyle
+&
\displaystyle
1152\,Y^2_{\Delta^2A}\,a^2
+ 16\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,a \alpha -
{{Y^2_{\Delta A^2}\,\alpha^2}\over 18} -
{{Y_\Delta\, Y_{\Delta A^2}\,b^2}\over 108} \\
\displaystyle
=&
\displaystyle
0.
\end{array}
\right.
\label{eq:detMass1134}
\end{eqnarray}
Now we required that one (1,1,8,0) mode be massless and therefore
determinant of the mass matrix for it ($\equiv M(1,1,8,0)$)
should vanish.
\begin{eqnarray}
\left.
\begin{array}{cl}
&
\displaystyle
\det M(1,1,8,0)\\
=&
\displaystyle
M_A\,M_\Delta -
{{Y_\Delta \,M_A \,b}\over {18\,\sqrt{2}}} +
{{Y_{\Delta A^2}\,M_\Delta\,b}\over {3\,\sqrt{2}}}\\
\displaystyle
+&
\displaystyle
1152\,Y^2_{\Delta^2A}\,a^2
+ 16\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,a \alpha -
{{Y^2_{\Delta A^2}\,\alpha^2}\over 18} -
{{Y_\Delta\, Y_{\Delta A^2}\,b^2}\over 108} \\
\displaystyle
=&
\displaystyle
0.
\end{array}
\right.
\label{eq:detMass1180}
\end{eqnarray}
Combining (\ref{eq:detMass1134}) and (\ref{eq:detMass1180})
with substituting (\ref{eq:Md1}) and (\ref{eq:Ma1}),
we find
\begin{eqnarray}
{{-8\,i}\over 3}\,Y_\Delta\,Y_{\Delta^2A}\,a^2 +
{{Y_\Delta \,Y_{\Delta A^2}\,a\,\alpha}\over 27} +
16\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,\alpha^2 = 0
\label{eq:M113411801}\\
\left( \left(
(\ref{eq:detMass1134}) - (\ref{eq:detMass1180}) \right) a \alpha /b^2
\right)\nonumber
\end{eqnarray}
and
\begin{eqnarray}
2304\,Y^2_{\Delta^2A} a^3+
32\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,a\,\alpha -
{{Y^2_{\Delta A^2}\,a\,\alpha^2}\over 9} -
{{Y_{\Delta}\,Y_{\Delta A^2}\,a\,b^2}\over 54}\nonumber\\ -
2304\,Y^2_{\Delta^2A}\,a\,b^2 -
32\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,\alpha\,b^2 = 0
\label{eq:M113411802}\\
( ( (\ref{eq:detMass1134}) + (\ref{eq:detMass1180}) ) * a ).
\nonumber
\end{eqnarray}
Solving a simultaneous equation (\ref{eq:M113411801})
and (\ref{eq:M113411802}) we get forms of
$Y_\Delta$ and $ Y_{\Delta A^2}$
as a function of $ Y_{\Delta^2A},a, b, \alpha $. Then
by substituting these expressions into
(\ref{eq:Md1}) and (\ref{eq:Ma1}) we find the following
three sets of solutions for
$M_\Delta$ , $M_A$ ,$Y_\Delta$ and $ Y_{\Delta A^2}$
as a function of $ Y_{\Delta^2A},a, b, \alpha $:
\vspace*{0.5cm}
$
\left.
\begin{array}{l}
M_{\Delta} \\
M_A \\
Y_\Delta \\
Y_{\Delta A^2}
\end{array}
\right\} =
$
solution 1::
\begin{eqnarray}
\left\{
\begin{array}{@{\,}l}
\displaystyle
{{-24\,\sqrt{2}\,i\,Y_{\Delta^2A}\,\alpha\,b}\over a}\\
\displaystyle
{{24\,\sqrt{2}\,i\,Y_{\Delta^2A}\,a\,b}\over \alpha}\\
\displaystyle
{{-864\,i\,Y_{\Delta^2A}\,\alpha}\over a} \\
\displaystyle
{{144\,i\,Y_{\Delta^2A} a}\over \alpha}
\end{array}
\right.
&\qquad\qquad\qquad\,&\qquad\qquad\qquad\qquad
\label{eq:solution1}
\end{eqnarray}
solution 2::
\begin{eqnarray}
\left\{
\begin{array}{@{\,}l}
\displaystyle
{{-24\,\sqrt{2}\,i\,Y_{\Delta^2A}\,\alpha\,b} \over a}\\
\displaystyle
{{- 24 i Y_{\Delta^2A}\,b}\over {{\sqrt{2}}\,a\,\alpha}}
\left( - a^2 + 3\,{b^2} -
\sqrt{a^4-10 a^2 b^2 + 9 b^4}
\right)\\
\displaystyle
{{-432\,i\,{ Y_{\Delta^2A}}\,{ \alpha}\,
\left( -3\,{a^2} + 3\,{b^2} -
{\sqrt{{a^4} - 10\,{a^2}\,{b^2} + 9\,{b^4}}} \right) }\over
{\left( -{a^3} + 3\,a\,{b^2} -
a\,{\sqrt{a^4 - 10\,a^2\,b^2 + 9\,b^4}}\right)}} ,\\
\displaystyle
{- 36 i\, Y_{\Delta^2A} \over {a \alpha}}\,
\left( - 3\,{a^2} + 3\,{b^2} -
\sqrt{a^4-10 a^2 b^2 + 9 b^4}
\right)\\
\end{array}
\right.
\label{eq:solution2}
\end{eqnarray}
solution 3::
\begin{eqnarray}
\left\{
\begin{array}{@{\,}l}
\displaystyle
{{-24\,\sqrt{2}\,i\,Y_{\Delta^2A}\,\alpha\,b}\over a}\\
\displaystyle
\displaystyle
{{- 24 i Y_{\Delta^2A}\,b}\over {{\sqrt{2}}\,a\,\alpha}}
\left( - a^2 + 3\,{b^2} +
\sqrt{a^4-10 a^2 b^2 + 9 b^4}
\right)\\
\displaystyle
{{-432\,i\,{ Y_{\Delta^2A}}\,{ \alpha}\,
\left( -3\,{a^2} + 3\,{b^2} +
{\sqrt{{a^4} - 10\,{a^2}\,{b^2} + 9\,{b^4}}} \right) }\over
{-{a^3} + 3\,a\,{b^2} +
a\,{\sqrt{{a^4} - 10\,{a^2}\,{b^2} + 9\,{b^4}}}}}\\
\displaystyle
{- 36 i\, Y_{\Delta^2A} \over {a \alpha}}\,
\left( - 3\,{a^2} + 3\,{b^2} +
\sqrt{a^4-10 a^2 b^2 + 9 b^4}
\right)\\
\end{array}
\right.
\label{eq:solution3}
\end{eqnarray}
In other words, once $M_{\Delta}, M_A, Y_\Delta$ and $Y_{\Delta A^2}$
are set to be one of these solutions,
the VEVs of $a, b$ and $\alpha$ can be chosen at our will
and one of (1,1,8,0) mode becomes massless.
Because we require also that one (1,3,1,0) mode be massless,
determinant of the mass matrix for it ($\equiv M(1,3,1,0)$)
must be zero.
\begin{eqnarray}
\begin{array}{cl}
&
\displaystyle
\det M(1,3,1,0)\\
\displaystyle
= &
\displaystyle
-{{Y_\Delta\,Y_{\Delta A^2}\,a^2 }\over {36}} -
16\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,a\,\alpha -
{{Y^2_{\Delta A^2}\,\alpha^2}\over 6} -
{{Y_\Delta\,Y_{\Delta A^2}\,a\,b}\over {18\,\sqrt{3}}} \\
\displaystyle
+&
\displaystyle
16\,{\sqrt{3}}\,i\,Y_{\Delta A^2}\,Y_{\Delta^2A}\,\alpha\,b
+ 1152\,Y^2_{\Delta^2A}\,b^2 +
{{Y_\Delta\,M_A\,a}\over {6\,\sqrt{6}}} \\
\displaystyle
+&
\displaystyle
16\,\sqrt{6}\,i\,Y_{\Delta^2A}\,M_A\,\alpha+
{{Y_\Delta\,M_A\,b}\over {9\,\sqrt{2}}} -
{{Y_{\Delta A^2}\,M_\Delta\,a}\over {\sqrt{6}}} + M_A\,M_\Delta \\
\displaystyle
= &
\displaystyle
0.
\end{array}
\label{eq:detMass1310}
\end{eqnarray}
Using (\ref{eq:detMass1310}) and
(\ref{eq:solution1})-(\ref{eq:solution3}),
we obtain a following equation
which determine a relation between $a$ and $b$
corresponding to a set of above solutions respectively:
solution 1::
\begin{eqnarray*}
{a^2}\,
\left( -3\,{a^2} + 7\,{\sqrt{3}}\,a\,b - 6\,{b^2} \right) = 0.
\end{eqnarray*}
solution 2::
\begin{eqnarray*}
-15\,{a^6} +
62\,{\sqrt{3}}\,{a^5}\,b +
237\,{a^4}\,{b^2} - 280\,{\sqrt{3}}\,{a^3}\,{b^3} -
249\,{a^2}\,{b^4} + 234\,{\sqrt{3}}\,a\,{b^5} + 27\,{b^6} \\
=\left(
33\,{a^4}\, -
50\,{\sqrt{3}}\,{a^3}\,b\, -
78\,{a^2}\,{b^2}\, +
78\,{\sqrt{3}}\,a\,{b^3}\, +
9\,{b^4}\right)
{\sqrt{{a^4} - 10\,{a^2}\,{b^2} + 9\,{b^4}}}.
\end{eqnarray*}
solution 3::
\begin{eqnarray*}
15\,{a^6} - 62\,{\sqrt{3}}\,{a^5}\,b -
237\,{a^4}\,{b^2} + 280\,{\sqrt{3}}\,{a^3}\,{b^3} +
249\,{a^2}\,{b^4} - 234\,{\sqrt{3}}\,a\,{b^5} - 27\,{b^6} \\
=\left(
33\,{a^4}\, -
50\,{\sqrt{3}}\,{a^3}\,b\, -
78\,{a^2}\,{b^2}\, +
78\,{\sqrt{3}}\,a\,{b^3}\, +
9\,{b^4}\right)
{\sqrt{{a^4} - 10\,{a^2}\,{b^2} + 9\,{b^4}}}.
\end{eqnarray*}
Numerically $a$ and $b$ must satisfy the following relation
respectively:
solution 1::
\begin{eqnarray}
a = \left\{
\begin{array}{@{\,}l}
{b}/{\sqrt{3}} ,\\
2\,{\sqrt{3}}\,b
\label{eq:ab1}
\end{array}
\right.
\end{eqnarray}
solution 2::
\begin{eqnarray}
a = \left\{
\begin{array}{@{\,}l}
-0.987293\,b\\
\left( -0.120361 -
0.724007\,i \right) \,b\\
\left( -0.120361 +
0.724007\,i \right) \,b\\
5.11238\,b
\label{eq:ab2}
\end{array}
\right.
\end{eqnarray}
solution 3::
\begin{eqnarray}
a = \left\{
\begin{array}{@{\,}l}
-3.13416\,b\\
-0.0643986\,b \\
\left( 1.10047 -
0.0616122\,i \right) \,b \\
\left( 1.10047 +
0.0616122\,i \right) \,b
\label{eq:ab3}
\end{array}
\right.
\end{eqnarray}
The solution 1 is the exact solution and the others are
exact up to O($\epsilon$).
In other words, if $a$ and $b$ satisfy these relations,
one (1,3,1,0) mode becomes massless.
Other requirements that two (2,2,1,0) modes,
one (1,3,1,-6) + h.c mode,
one (2,1,3,1) + h.c mode and one (2,1,1,-3) + h.c mode
be massless are easily satisfied
by tuning parameters such as $M_\Phi, M_H, Y_{H\Phi\Delta}$,
$Y_{H\overline{\Phi}\Delta}$ and so on.
To make (1,3,1,-6) + h.c mode massless,
from the mass term for it (see appendix C)
\begin{eqnarray}
{ M_\Phi}=
-\left({{{\sqrt{6}}\, Y_{\Phi A}\alpha}\over 10} +
{{Y_{\Phi\Delta}\,a}\over {10\,{\sqrt{6}}}} +
{{{ Y_{\Phi\Delta}}\,b}\over {10\,{\sqrt{2}}}}
\right).
\label{eq:Mphi1}
\end{eqnarray}
To make two (2,2,1,0) mode massless
we tune parameters $M_H, M_\Phi, Y_{H\Phi\Delta}$
and $Y_{H\overline{\Phi}\Delta}$ so that the eigenvalue equation
for the mass matrix of (2,2,1,0)
\begin{eqnarray}
\displaystyle
{\lambda^3}
- { M_H}\,{\lambda^2}
+\left( -{{ Y^2_{H\overline{\Phi}\Delta} \,b^2
}\over 10} - {{ Y^2_{H\Phi\Delta} \,b^2 }\over {10}} -
( {{{ Y_{\Phi\Delta}}\,b}\over {15\,{\sqrt{2}}}}
+ { M_{\Phi}} )^2 \,
\right) \,\lambda & \cr
-\left( {{{ Y_{\Phi\Delta}}\,b}\over {15\,{\sqrt{2}}}}
+ { M_{\Phi}} \right) \,
\left(
M_H\,\left( {{{ Y_{\Phi\Delta}}\,b}\over {15\,{\sqrt{2}}}}
+ { M_{\Phi}} \right)
+ {{ Y_{H\overline{\Phi}\Delta A^2}
\,Y_{H\Phi\Delta}\,{b^2}} \over 5} \right) =0
\label{eq:eigeneq2210}
\end{eqnarray}
has two 0 solutions (exactly these two solutions may
have at most O($\epsilon$) solution)\footnote{Implicitly
it is assumed that the mass matrix for (2,2,1,0) is hermite, that is,
all parameters appearing in the mass matrix are real}.
The way of getting two zero eigenvalues is to tune
the zeroth and first terms of $\lambda$
zero. More exactly the zeroth term must be at most O($\epsilon^2$)
and the first term must be at most O($\epsilon$).
To satisfy these constraint
\begin{eqnarray}
\begin{array}{c}
M_\Phi+{ {Y_{\Phi\Delta} b} \over {15 \sqrt{2}}} \sim O(\epsilon),
\\
Y_{H\Phi\Delta} \sim Y_{H\overline\Phi\Delta} \sim O(\sqrt\epsilon).
\end{array}
\label{eq:M221001}
\end{eqnarray}
(\ref{eq:Mphi1}) and the first equation of (\ref{eq:M221001})
lead
\begin{eqnarray}
Y_{\Phi A}=-{ {\sqrt{3} a + b}\over {6 \sqrt{3} \alpha}}
Y_{\Phi\Delta}
\label{eq:Yda1}
\end{eqnarray}
up to O($\epsilon$).
Finally to make one (2,1,3,-1) + h.c mode and one (2,1,1,3) + h.c mode
massless, foe example, we can switch only couplings with subscript 4 on
and tune
\begin{eqnarray}
Y_{\psi\Delta} = {7 \over {16 \sqrt{3}}} i Y_{\psi A} \alpha/b,
\label{eq:ypd0}
\end{eqnarray}
\begin{eqnarray}
M_\Psi = -{3 \over {4 \sqrt{6}}} i Y_{\psi A} \alpha
-{7 \over {4 \sqrt{2}}} i Y_{\psi A} {a\over b}\alpha.
\label{eq:mp0}
\end{eqnarray}
\subsubsection{check mass matrices}
Now we know the necessary condition for the parameters
realizing the spectrum (\ref{eq:spectrum}).
Then we check all the mass matrices to examine
whether these parameters really
produce the spectrum (\ref{eq:spectrum}).
solution 1::
The solution 1 does not produce the spectrum (\ref{eq:spectrum}),
because by substituting the solution 1 (\ref{eq:solution1})
into the mass matrix of (2,2,6,2) multiplet,
this multiplet is calculated to be massless.
solution 2::
First to see whether the solution 2 ,
(\ref{eq:solution2}) with a relation between $a$ and $b$
(\ref{eq:ab2}), is usable, we substitute (\ref{eq:ab2})
into (\ref{eq:solution2}).
\begin{eqnarray}
&\left.
\begin{array}{@{\,}l}
M_{\Delta} \\
M_A
\end{array}
\right\}=
\left\{
\begin{array}{@{\,}l}
\left\{
\begin{array}{@{\,}l}
24.3089\,i\,{\sqrt{2}}\,Y_{\Delta^2A}\,\alpha\\
19.1441\,i\,{\sqrt{2}}\,Y_{\Delta^2A}
b^2/ \alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
\left( 32.2574 +
5.36258\,i \right) \,{\sqrt{2}}\,Y_{\Delta^2A}\,
\alpha\\
\left( -4.78842 +
0.510831\,i \right)
\,{\sqrt{2}}\,Y_{\Delta^2A} b^2 / \alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
\left( -32.2574 +
5.36258\,i \right) \,{\sqrt{2}}\,Y_{\Delta^2A}\,
\alpha \\
\left( 4.78842 +
0.510831\,i \right)
\,{\sqrt{2}}\,Y_{\Delta^2A} b^2 / \alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
-4.69449\,i\,{\sqrt{2}}\,Y_{\Delta^2A}\,\alpha\\
103.023\,i\,{\sqrt{2}}\,Y_{\Delta^2A} b^2
/ \alpha
\end{array}
\right.
\end{array}
\right.
\label{eq:sol2mdma}\\
&Y_{\Delta A^2} =
\left\{
\begin{array}{@{\,}l}
{-13.6527\,i\, Y_{\Delta^2A}\,b} / \alpha \\
{\left( 37.7632 -
7.13352\,i \right) \,Y_{\Delta^2A}\,b} /
\alpha \\
{\left( -37.7632 -
7.13352\,i \right) \,{ Y_{\Delta^2A}}\,b}
/ \alpha \\
677.159\,i\,Y_{\Delta^2A}\,b/ \alpha
\end{array}
\right.
\label{eq:sol2ydaa}\\
& Y_\Delta =
\left\{
\begin{array}{@{\,}l}
-104.016\,i\,Y_{\Delta^2A}\,\alpha / b\\
\left( -1560.23 -
131.862\,i \right) \,Y_{\Delta^2A}\,\alpha
/ b\\
\left( 1560.23 -
131.862\,i \right) \,Y_{\Delta^2A}\,\alpha
/ b \\
-185.139\,i\,Y_{\Delta^2A}\,\alpha / b
\end{array}
\right.
\label{eq:sol2yd}
\end{eqnarray}
In each of these equations, four expressions correspond to
the four relations
between $a$ and $b$ in (\ref{eq:ab2}) respectively.
As we required that Yukawa couplings are not too big
(see the statement below (\ref{eq:superyukawa}))
only the first expression of the solution 2
is meaningful. This means that only the first relation
between $a$ and $b$ in (\ref{eq:ab2}) is meaningful.
By substituting (\ref{eq:solution2}) with the first equation of
(\ref{eq:ab2}) it is easy to check that all multiplets
other than those in (\ref{eq:spectrum}) have their mass of
O($M_U$) which spread around $M_U$ up to one order of magnitude
and multiplets in (\ref{eq:spectrum}) are massless.
Therefore this solution can be a solution of
our scenario.
solution3::
First we substitute (\ref{eq:ab3}) (relation between $a$ and $b$)
into (\ref{eq:solution3}) to see an explicit form of solution 3.
\begin{eqnarray}
&\left.
\begin{array}{@{\,}l}
M_{\Delta} \\
M_A
\end{array}
\right\}=
\left\{
\begin{array}{@{\,}l}
\left\{
\begin{array}{@{\,}l}
7.65756\,i\,{\sqrt{2}}\,Y_{\Delta^2A}\,\alpha\\
-15.8066\,i\,{\sqrt{2}}\,Y_{\Delta^2A} b^2 /
\alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
372.679\,i\,{\sqrt{2}}\,Y_{\Delta^2A}\,\alpha\\
1115.98\,i\,{\sqrt{2}}\,Y_{\Delta^2A} b^2
/ \alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
\left( 1.21719 -
21.7407\,i \right) \,{\sqrt{2}}\,Y_{\Delta^2A}\,
\alpha\\
\left( 17.3100 -
22.7812\,i \right)
\,{\sqrt{2}}\,Y_{\Delta^2A} b^2 /\alpha
\end{array}
\right.\\
\left\{
\begin{array}{@{\,}l}
\left( -1.21719 -
21.7407\,i \right) \,{\sqrt{2}}\,
Y_{\Delta^2A}\,\alpha\\
\left( -17.3100 -
22.7812\,i \right) \,
{\sqrt{2}}\,{ Y_{\Delta^2A}} b^2 / \alpha
\end{array}
\right.
\end{array}
\right.\\
&Y_{\Delta A^2} =
\left\{
\begin{array}{@{\,}l}
-273.079\,i\,Y_{\Delta^2A}\,b / \alpha\\
3343.29\,i\,Y_{\Delta^2A}\,b / \alpha\\
\left( 56.3660 +
10.8904\,i \right) \,Y_{\Delta^2A}\,b
/ \alpha \\
\left( -56.3660 +
10.8904\,i \right) \,Y_{\Delta^2A}\,b
/ \alpha
\end{array}
\right.\\
& Y_\Delta =
\left\{
\begin{array}{@{\,}l}
793.766\,i\,Y_{\Delta^2A}\,\alpha / b \\
6698.93\,i\, Y_{\Delta^2A}\,\alpha / b \\
\left( 241.144 -
102.803\,i \right) \,Y_{\Delta^2A}\,\alpha
/ b \\
\left( -241.144 -
102.803\,i \right) Y_{\Delta^2A}\, \alpha
/ b
\end{array}
\right.
\end{eqnarray}
In each of these equations, four expressions correspond to
the four relations
between $a$ and $b$ in (\ref{eq:ab3}) respectively.
By the same way as we picked
only the first expression up from four cases in solution 2,
the last two relations
between $a$ and $b$ in (\ref{eq:ab3}) are meaningful.
By substituting (\ref{eq:solution3}) with the third or fourth
equation of (\ref{eq:ab3}) it is easy to check that all multiplets
other than those in (\ref{eq:spectrum}) have their mass of
O($M_U$) which spread around $M_U$ up to one order of magnitude
and multiplets in (\ref{eq:spectrum}) are massless.
Therefore this solution can be a solution of
our scenario too.
\subsection{Second step}
In this section we find a parameter region
which produces our scenario exactly.
\subsubsection{Deviation from the previous solutions}
Because the accuracy of the previous calculation
is O($\epsilon$),
all parameters besides
$b, \alpha$ and $Y_{\Delta^2A}$
can deviate from the value which is obtained at the previous
section and therefore we can expand the deviation
in the power of $\epsilon$ as follows.
\begin{eqnarray}
\displaystyle
a = a_0 + \sum_{i=1} a_i \epsilon^i,
\label{eq:deva}
\end{eqnarray}
\begin{eqnarray}
\displaystyle
M_{\Delta} = M_{\Delta 0} + \sum_{i=1} M_{\Delta i} \epsilon^i ,
\label{eq:devMd}
\end{eqnarray}
\begin{eqnarray}
M_A = M_{A 0} + \sum_{i=1} M_{A i} \epsilon^i ,
\label{eq:devMa}
\end{eqnarray}
\begin{eqnarray}
Y_\Delta = Y_{\Delta 0} + \sum_{i=1} Y_{\Delta i} \epsilon^i ,
\label{eq:devAd}
\end{eqnarray}
\begin{eqnarray}
Y_{\Delta A^2} = Y_{\Delta A^20} + \sum_{i=1}
Y_{\Delta A^2 i} \epsilon^i ,
\label{eq:devAdaa}
\end{eqnarray}
\begin{eqnarray}
\beta = \sum_{i=1} \beta_i \epsilon^i ,
\label{eq:devbe}
\end{eqnarray}
\begin{eqnarray}
c = \sum_{i=1} c_i \epsilon^i .
\label{eq:devc}
\end{eqnarray}
In these expressions, variables with subscript 0 stand for
those which are obtained in the previous section.
Substituting (\ref{eq:deva}) - (\ref{eq:devc}) into
the F-flat condition (\ref{eq:Fa}) - (\ref{eq:Fphi}),
we get following relations.
{}From (\ref{eq:Fa}), (\ref{eq:Fb}) and (\ref{eq:Falpha}) we get
\begin{eqnarray}
M_{\Delta 1} &=& -
{M_{\Delta 0}\over a_0} a_1,\nonumber\\
M_{A1} &=& {b^3 \over {9 \sqrt{2} \alpha^2}} Y_{\Delta 1}
+{{24 \sqrt{2} i Y_{\Delta^2A} b }\over \alpha}
(1+2 {b^2 \over a^2_0}) a_1,
\label{eq:deva1Md1Ma1Adaa1Ad1}\\
Y_{\Delta A^2 1}&=& (b^2/6 \alpha^2) Y_{\Delta 1}
+{{144 i Y_{\Delta^2A}}\over \alpha}(1+{b^2 \over a^2_0}) a_1.
\nonumber
\end{eqnarray}
We obtain the relation between $\beta_1$ and $c_1$ by
substituting (\ref{eq:deva}) - (\ref{eq:devc}) with
(\ref{eq:deva1Md1Ma1Adaa1Ad1}) into
(\ref{eq:Fc}) and (\ref{eq:Fbeta}) as follows:
First we note (\ref{eq:Fc}) and (\ref{eq:Fbeta})
can be rewritten
\begin{eqnarray}
M(1,3,1,0) \pmatrix{\beta\cr c}
=-{1 \over 10} \pmatrix{2 Y_{\Phi A} \cr Y_{\Phi\Delta} }
\phi\overline\phi
\end{eqnarray}
and therefore
\begin{eqnarray}
\pmatrix{\beta\cr c}=
-{1 \over 10}
{\rm M}(1,3,1,0)^{-1}
\pmatrix{2 Y_{\Phi A} \cr Y_{\Phi\Delta} }
\phi\overline\phi.
\label{eq:betaandc}
\end{eqnarray}
where M(1,3,1,0) is a mass matrix for (1,3,1,0) and by assumption
$\phi, \overline\phi \sim $ O($\epsilon$).
Let us decompose the inverse of M(1,3,1,0)
\begin{eqnarray}
{\rm M}(1,3,1,0)^{-1}=
\det\left({\rm M}(1,3,1,0)\right)^{-1}
\left({\rm A}+{\rm O}({\epsilon})\right).
\end{eqnarray}
Since by assumption there is one massless mode in (1,3,1,0)
up to O($\epsilon$), det(M(1,3,1,0)) $ \sim $ O($\epsilon$)
and the first row in A is parallel to the second row in A, that is
\begin{eqnarray}
{a_{11} \over a_{21}}= {a_{12} \over a_{22}}
\label{eq:aa}
\end{eqnarray}
where A $\equiv (a_{ij})$.
Then up to the leading order of $\epsilon$
\begin{equation}
\beta={a_{21} \over a_{11}} c
\label{eq:betac0}
\end{equation}
namely, as a exact relation
\begin{equation}
\beta_1={a_{21} \over a_{11}} c_1
\label{eq:betac1}
\end{equation}
is obtained.
To see this explicitly, we follow the above calculation
in the case of the first relation of solution 2.
\begin{eqnarray*}
\det({\rm M}(1,3,1,0))=
(-26423.4 Y^2_{\Delta^2A} b\, a_1 + {{16.1727\,i
Y_{\Delta^2A} Y_{\Delta 1} {b^3}}
\over \alpha})\,{ \epsilon}
+{\rm O} (\epsilon^2)
\end{eqnarray*}
as we expected the determinant is O($\epsilon$).
A is calculated to be
\begin{eqnarray*}
A=
\pmatrix{
72.3850\,i\,{ Y_{\Delta^2A}}\,{ \alpha}
,&
-39.5148\,i\,{ Y_{\Delta^2A}}\, b
\cr
-39.5148\,i\,{ Y_{\Delta^2A}}\,b
,&
{{21.5710\,i\,
{ Y_{\Delta^2A}}\,{b^2}} / {{ \alpha}}}
}
\end{eqnarray*}
Apparently A satisfies (\ref{eq:aa}).
Then
\begin{eqnarray}
\beta_1=-1.83185 {\alpha \over b} c_1
\label{eq:betac}
\end{eqnarray}
is obtained.
\subsubsection{Determination of input parameters of the theory}
Though we can determine the parameters in the power of
$\epsilon$ order by order,
instead of doing so we will give the parameters of the theory
in a term of the VEVs
because the purpose of the paper is to find a parameter region
for the theory, $M$'s and $Y$'s,
which leads to the spectrum ({\ref{eq:spectrum}).
As we will see,
by the VEVs $a, b, c, \alpha$ and $\beta$
we can express the input parameters of the theory.
To do this, first we see the F-flat conditions (\ref{eq:Fa})
- (\ref{eq:Fbeta}). These equation can be rewritten
\begin{eqnarray}
C
\pmatrix{M_\Delta\cr
M_A\cr
Y_\Delta\cr
Y_{\Delta A^2}\cr
Y_{\Delta^2A}}=
-\pmatrix{1/(10 \sqrt{6}) Y_{\Phi\Delta}\cr
1/(10 \sqrt{2}) Y_{\Phi\Delta}\cr
1/10 Y_{\Phi\Delta}\cr
\sqrt{6} /10 Y_{\Phi A}\cr
1/5 Y_{\Phi A}}
\phi\overline\phi
\end{eqnarray}
where
\begin{eqnarray}
C=\pmatrix{a,& 0,&{1\over{12 \sqrt{6}}} c^2,& - {1\over{2 \sqrt{6}}} \beta^2,&
24 \sqrt{2} i \alpha b \cr
b,&0,& {1\over{18 \sqrt{2}}} b^2 + {1\over{18 \sqrt{2}}} c^2
,&-{1\over{3 \sqrt{2}}}\alpha^2,&
24 \sqrt{2} i a \alpha + 24 \sqrt{2} i \beta c \cr
c ,&0,& {1\over {6 \sqrt{6}}} a c + {1\over {9 \sqrt{2}}} b c,
&- {1\over{ \sqrt{6}}} \alpha \beta,
&16 \sqrt{6} i \alpha c + 24 \sqrt{2} i b \beta \cr 0,&
\alpha,&0,& - {\sqrt{2}\over 3} \alpha b - {1\over\sqrt{6}} \beta c,&
24 \sqrt{2} i a b + 8 \sqrt{6} i c^2 \cr
0,&\beta,&0,& - {1\over\sqrt{6}} \alpha c - {1\over\sqrt{6}} a \beta,&
24 \sqrt{2} i b c}
\end{eqnarray}
As we know from the previous argument that
$b,c$ and $\alpha$ can be chosen freely and
$a$ and $\beta$ are given by
\begin{eqnarray}
a = a_0 + a_1 \epsilon,\cr
\beta = \beta_1 \epsilon + \beta_2 \epsilon^2
\label{eq:constraintabeta}
\end{eqnarray}
where $a_0$ is given by the first equation of (\ref{eq:ab2})
or one of the last two equation of (\ref{eq:ab3})
and $\beta_1$ is given by (\ref{eq:betac1}). Note that
higher orders in (\ref{eq:deva}) and (\ref{eq:devbe})
can be absorbed into $a_1$ and $\beta_2$ respectively.
Then the input parameters are reduced to
\begin{eqnarray}
\pmatrix{M_\Delta\cr
M_A\cr
Y_\Delta\cr
Y_{\Delta A^2}\cr
Y_{\Delta^2A}}=
-C^{-1}
\pmatrix{1/(10 \sqrt{6}) Y_{\Phi\Delta}\cr
1/(10 \sqrt{2}) Y_{\Phi\Delta}\cr
1/10 Y_{\Phi\Delta}\cr
\sqrt{6} /10 Y_{\Phi A}\cr
1/5 Y_{\Phi A}}
\phi\overline\phi .
\label{eq:parameters2}
\end{eqnarray}
For example, in the case of solution 2,
\begin{eqnarray*}
C^{-1}&=&(\det C)^{-1} C'\epsilon\\
\det C&=&(-3.76350\,i\,\alpha^2\,b^4\,\beta_2\,
c_1-2.25347 i \alpha^3 b^2 a_1 c^2_1)
\,\epsilon^3 + {{{\rm O}({ \epsilon})}^4}
\end{eqnarray*}
\begin{eqnarray*}
C'&=&
\pmatrix{ 0 ,& 0,&
-2.68018\,i\,\alpha^3\,b^3\,c_1 ,& 0,&
-1.08826\,i\,\alpha^4\,{b^2}\,c_1\cr
0 ,& 0,&
-2.11074\,i\,\alpha\,{b^5}\, c_1 ,& 0,&
-0.857040\,i\,\alpha^2\,{b^4}\,c_1\cr
0 ,& 0,&
8.10927\,i\,\alpha^3\,{b^2}\,c_1 ,& 0,&
3.29268\,i\,\alpha^4\,b\,c_1 \cr
0 ,& 0,&
1.06439\,i\,\alpha\,{b^4}\,c_1 ,&0 ,&
0.432184\,i\,\alpha^2\,{b^3}\,c_1 \cr
0 ,& 0,&
-0.0779620\,\alpha^2\,{b^3}\,c_1 ,&0 ,&
-0.0316556\,\alpha^3\,{b^2}\, c_1} \\
&+&{\rm O}(\epsilon)
\end{eqnarray*}
{}From this equation it is easy to see that all parameters
are of order $\epsilon^0$ and they satisfy the first solution of
the solution 2.
Finally from (\ref{eq:Fphi}) $M_\Phi$ is determined:
\begin{equation}
{ M_\phi}=-
{ Y_{\Phi A}}\,\left( {{\sqrt{6}\,{ \alpha}}\over 10} +
{{{ \beta}}\over 5} \right) -
{ Y_{\Phi\Delta}}\,
\left( {a\over {10\,{\sqrt{6}}}} + {b\over {10\,{\sqrt{2}}}} +
{c\over {10}} \right) .
\label{eq:Mphi2}
\end{equation}
\subsubsection{check mass matrices}
The multiplets in (\ref{eq:spectrum}) besides one (2,2,1,0)
must decouple at $M_{\nu_R}$,that is, they must acquire
mass of O$(M_{\nu_R})$.
{}From now on we check whether they have mass of O$(M_{\nu_R})$.
First we note one (2,1,3,-1) + h.c and (2,1,1,3) + h.c
can have mass of O$(M_{\nu_R})$ by the following two reasons:
(1) Parameters $Y_{\psi\Delta}$ and $M_\Psi$ may deviate from the
value given by (\ref{eq:ypd0}) and (\ref{eq:mp0})
respectively
\footnote{Though (2,1,3,-1) + h.c has a same
quantum number under the SM group as an NG mode
associated with the breakdown of SO(10)
the SM group (see table (\ref{eq:multiplets})),
it does not mix with others
because the VEV of $\psi=0$
and therefore
this NG mode does not consist of it.
(2,1,1,3) + h.c has a same quantum number as that of (2,2,1,0)
under the SM group but by the same reason they do not mix with
(2,2,1,0).
See the superpotential
(\ref{eq:superpotential}) - (\ref{eq:superpsi}). }.
(2) There exist couplings with $c$ and $\beta$.
Then we see the mass matrix for (2,2,1,0).
Under SM it has a quantum number (2,1,$\pm 1/2$).
(2,2,1,6) + h.c also includes the same component.
Then the mass matrix is
\begin{eqnarray}
M(2,1,\pm 1/2)&=&\pmatrix{
\tilde{M}_\Delta,&x,&y,&0\cr
x',&M_H,& u,& v\cr
0,&u,& 0,&, w-z\cr
y',&v,& w+z,&0}
\label{eq:massdoublets}
\end{eqnarray}
where
\begin{eqnarray}
\begin{array}{ccl}
\displaystyle
\tilde{M}_\Delta&=&{\rm M}(2,2,1,6)+{1\over12} Y_\Delta c + 24 i
Y_{A\Delta^2} \beta\\
x&=&-{1\over \sqrt{5}} Y_{H\overline{\Phi}\Delta} \overline\phi
\sim {\rm O} (\epsilon^{3/2})\\
x'&=&-{1\over \sqrt{5}} Y_{H\Phi\Delta} \phi
\sim {\rm O} (\epsilon^{3/2})\\
y&=&-{1\over 40} Y_{\Phi\Delta} \overline\phi
\sim {\rm O} (\epsilon)\\
y'&=&-{1\over 40} Y_{\Phi\Delta} \phi
\sim {\rm O}(\epsilon)\\
u&=&-{1\over \sqrt{10}} Y_{H\Phi\Delta}\,b +
{1\over {2 \sqrt{5}}} Y_{H\Phi\Delta}\,c \sim {\rm O}
(\sqrt{\epsilon}), \\
v&=& {1 \over \sqrt{10}}Y_{H\overline{\Phi}\Delta}\,b+
{1\over {2 \sqrt{5}}} Y_{H\overline{\Phi}\Delta}\,c
\sim {\rm O} (\sqrt{\epsilon}), \\
w&=& { M_\Phi}+
{{{ Y_{\Phi\Delta}}\,b}\over {15\,{\sqrt{2}}}} \sim {\rm O}
(\epsilon) , \\
z&=&{{{ Y_{\Phi\Delta}}\,c}\over 30} +
{{{ Y_{\Phi A}}\,\beta}\over 10} \sim {\rm O} (\epsilon) .
\end{array}
\label{eq:ahpd}
\end{eqnarray}
M(2,2,1,6) is given in the appendix C.
Orders of $x,y,..$ are followed from (\ref{eq:M221001})
Because one (2,1,$\pm 1/2$) multiplet remains massless
after $G_{2231}$ breaks down to the SM group
\begin{eqnarray}
\det\left(M(2,1,\pm 1/2)\right) = \{ \tilde{M}_\Delta (z^2-w^2)+
y\, y'(w-z)\} M_H + 2 \tilde{M}_\Delta u v w ... =0,
\label{eq:detdoublets}
\end{eqnarray}
and hence $M_H$ is determined as follows:
\begin{eqnarray}
M_H= { {2 u v w} \over {w^2 - z^2}} + {\rm O} (\epsilon) .
\label{eq:mh}
\end{eqnarray}
In this case the higher order terms must be included
to have a pair of light Higgs doublets.
Next let us consider (1,1,8,0). This multiplet
becomes (1,8,0) under the SM group and therefore
it mixes with $T_{3R} = 0$ component of
(1,3,8,0) under the SM.
Then the mass matrix for (1,8,0) is represented as $3 \times 3$
matrix.
\begin{equation}
M(1,8,0)=\left(
\begin{array}{c|c}
M(1,1,8,0)&mixing\cr\hline
mixing&M(1,3,8,0)
\end{array}
\right).
\label{eq:Mass180}
\end{equation}
After $G_{2231}$ breaks down to the SM group,
there is a correction of O($M_U \epsilon \sim M_{\nu_R}$)
to the mass
matrices M(1,1,8,0) and M(1,3,8,0)
because parameters appearing in them
are different by O($\epsilon$)
from those calculated at the previous
section. It is directly calculated using (\ref{eq:parameters2})
(or equivalently (\ref{eq:deva}) - (\ref{eq:devAdaa})
and (\ref{eq:deva1Md1Ma1Adaa1Ad1}) ) that
one of the eigenvalues of M(1,1,8,0) is of O($M_U$)
which has already suggested at the previous section and
the other is O($M_{\nu_R}$). As M(1,3,8,0) is O($M_U$),
even though there is a correction of O($M_{\nu_R}$),
M(1,3,8,0) is still O($M_U$). Contributions of $c$ and $\beta$
to the mass matrix (\ref{eq:Mass180}) appear at mixing terms
between (1,1,8,0) and (1,3,8,0)\footnote{There is no contribution
of $c$ and $\beta$ to M(1,1,8,0) and M(1,3,8,0).
The reason is as follows. Under $G_{2231}$
$c$ and $\beta$ are contained in (1,3,1,0). Because
(1,3,1,0)$(1,1,8,0)^2$ contains no singlet, $c$ and $\beta$
do not couple to $(1,1,8,0)^2$. Though (1,3,1,0)$(1,3,8,0)^2$
can appear, as there is no three point coupling of $T_{3R}=0$
component of SU(2) triplet, $c$ and $\beta$ do not couple to
$T_{3R}=0$ component of (1,3,8,0)}
and they are of O($M_{\nu_R}$). Then M(1,8,0) takes the following
form
\begin{equation}
\left(
\begin{array}{ccc}
\rm O (M_U) &0&\rm O (M_U \epsilon) \\
0& O (M_U \epsilon)&\rm O (M_U \epsilon)\\
\rm O (M_U \epsilon)&\rm O (M_U \epsilon)&\rm O (M_U)
\end{array}
\right).
\end{equation}
Apparently two eigenvalues are of O($M_U$) and
the other is of O($M_{\nu_R}$). This fact suggests that
the lightest element of (1,1,8,0) under $G_{2231}$ decouples
at the scale $M_{\nu_R}$.
Finally we check the mass of (1,3,1,0) and (1,3,1,-6) + h.c.
Under the SM (1,3,1,0) is decomposed into one
neutral singlet
and a pair of charged singlet with
hypercharge $Y = \pm 1$. (1,3,1,,-6) + h.c becomes
two neutral singlets, a pair of $Y = \pm 1$ and a pair of
$Y = \pm 2$ singlets. Then $Y = \pm 1$
component of them will
mix with each other.
Mass for $Y = \pm 2$ component takes the following form
\begin{equation}
{ Y_{\Phi A}}\,\left( {{\sqrt{6}\,\alpha}\over 10} -
{{{ \beta}}\over 5} \right) +
{ Y_{\Phi\Delta}}\,
\left( {a\over {10\,{\sqrt{6}}}} + {b\over {10\,{\sqrt{2}}}} -
{c\over {10}} \right) + { M_\phi}
= -{2 \over 5 }Y_{\Phi A}\beta - {1\over 5} Y_{\Phi\Delta} c
\end{equation}
where (\ref{eq:Mphi2}) is used.
{}From this equation obviously the $Y = \pm 2$ component
has a mass of O($M_{\nu_R}$).
Mass matrix of $Y = \pm 1$ component is
\begin{equation}
\pmatrix{ -{{ Y_{\Delta A^2}\,a}\over {\sqrt{6}}} + M_A,&
-{{{ Y_{\Delta A^2}}\,{ \alpha}}\over {{\sqrt{6}}}} +
24\,i\,{\sqrt{2}}\,{ Y_{\Delta^2A}}\,b,&
-{{Y_{\Phi A}\phi }\over 5} \cr
-{{{ Y_{\Delta A^2}}\,{ \alpha}}\over {{\sqrt{6}}}} +
24\,i\,{\sqrt{2}}\,{ Y_{\Delta^2A}}\,b,&
{{Y_{\Delta}\,a}\over {6\,{\sqrt{6}}}} +
16\,i\,{\sqrt{6}}\,{ Y_{\Delta^2A}}\,{ \alpha} +
{{{ Y_{\Delta}}\,b}\over {9\,{\sqrt{2}}}} + { M_{\Delta}},&
-{{Y_{\Phi\Delta}\phi }\over 10} \cr
-{{Y_{\Phi A}\,\overline\phi}\over 5},&
-{{Y_{\Phi\Delta}\,\overline\phi }\over 10},&
-{{Y_{\Phi A}\,\beta}\over 5} -
{{{ Y_{\Phi\Delta}}\,c}\over {10}} }.
\label{eq;MassY1}
\end{equation}
Since it is an NG mode associated with the breakdown of
$G_{2231}$ to $G_{231}$ there is one massless mode.
It is easy to see that this matrix has 0 eigenvalue because
${\rm 1st\ row} \times {\beta/\phi} + {\rm 2nd\ row}
\times c/\phi + {\rm 3rd\ row} = 0$ using the F-flat conditions
(\ref{eq:Fc}) and (\ref{eq:Fbeta}). It is also explicitly
calculated that one eigenvalue is of O($M_U$) and
the other is of O($M_{\nu_R}$).
\section{summary}
As we saw, by constructing the input parameters for the theory
using (\ref{eq:parameters2}),
(\ref{eq:Mphi2}), (\ref{eq:ahpd}) and (\ref{eq:mh})
from the desired values of VEVs $a, b, c, \alpha, \beta, \phi$
and $\overline\phi$ which satisfy (\ref{eq:Dphi}) and
(\ref{eq:constraintabeta}),
we can have particles (\ref{eq:spectrum}) in the intermediate region.
They decouple from the spectrum at $M_{\nu_R}$ except
a pair of what we call Higgs doublets.
It means that it is possible to construct a SUSY SO(10) GUT
with an intermediate scale
consistent with the gauge unification. It suggests also
that the right-handed neutrinos acquire mass through a
renormalizable coupling.
and it can be understood as a reflection of the breakdown of $G_{2231}$
to $G_{231}$
There are many variations for a SUSY SO(10) GUT with an
intermediate scale because there are many candidates
for the particle content which exist in the intermediate region
and we have many variations for content of SO(10) multiplets
which contain one of the candidates.
For example, we can replace (2,2,1,0) to (2,1,1,3) + h.c
in the spectrum (\ref{eq:spectrum}) and vise versa,
because their contribution to the running of the gauge coupling
relevant to $G_{231}$ is as same as that of each other.
When we remove one (2,2,1,0) from the spectrum (\ref{eq:spectrum})
and add one (2,1,1,3) + h.c to it,
by adding a pair of SO(10) multiplets $16 + \overline{16}$
which contains (2,1,1,3) + h.c under $G_{2231}$
we can have such a spectrum at the intermediate region.
At that time while we have to tune couplings relevant to
SO(10) multiplets $16 + \overline{16}$,
we can release the constraint (\ref{eq:M221001})
(or equivallently (\ref{eq:ahpd})).
Of course, there is a quite different type of content
for the candidates. Using them we can construct
quite a different SO(10) GUT with an intermediate scale.
Though the gauge unification by the MSSM is a very attractive idea,to
take into account a right handed-neutrino mass
we should consider a possibility of a GUT with an intermediate
symmetry.
\begin{center}
ACKNOWLEDGEMENTS
\end{center}
The author wish to acknowledge T.~ Kugo, M.~Bando and
T.~Takahashi for valuable comments and discussion.
|
\section{Shor's Algorithm}
We outline the idea of Shor's algorithm below (See Refs.~\cite{Shor,RMP} for
details). To factorize a composite number $M$ (which is assumed not to be a
prime power), we prepare our system in the state
\begin{equation}
|\Psi\rangle = \frac{1}{2^{L/2}} \sum_{a=1}^{2^L} |a\rangle \mbox{~,}
\label{E:A1}
\end{equation}
with $2^L \approx M^2$. This can be achieved by, say, setting $L$ quantum
spin-1/2 particles with their spins pointing towards the positive
$x$-direction (and all our measurements are performed in the $z$-direction).
\par
Then we evolve our wavefunction to
\begin{equation}
|\Psi\rangle = \frac{1}{2^{L/2}} \sum_{a=1}^{2^L} |a,m^a \mbox{mod~} M\rangle
\mbox{~,} \label{E:A2}
\end{equation}
for some randomly chosen integer $1<m<M$ with $\gcd (m,M) = 1$. (If $\gcd
(m,M) > 1$, then we are so lucky that we have found a non-trivial factor of
$M$ by chance. The probability for this to happen scales exponentially with
$\log M$, and is therefore negligible.) The above evolution can be done using
the power algorithm \cite{Knuth}, which takes $\mbox{O}(L)$ multiplications in
${\Bbb Z}_M$. As mentioned in the text, multiplying two $L$-bit numbers using
the Sch\"{o}nhagen and Strassen method, which is asymptotically the fastest
known algorithm, requires $\mbox{O}(L \log L \log\log L)$ elementary bit
operations \cite{Knuth,Fast_Mult}. Consequently, evolving the wavefunction
from state~(\ref{E:A1}) to state~(\ref{E:A2}) takes $\mbox{O}(L^2 \log L
\log\log L)$ elementary q-bit operations. Besides, it requires $\mbox{O}(L)$
extra q-bits as working space during the computation.
\par
Now we make a measurement on the second set of q-bits in our system (which
should take at most $\mbox{O}(L)$ time). Thus, the wavefunction of our system
for the first set of q-bits collapses to
\begin{equation}
|\Psi\rangle = \frac{1}{\sqrt{k}} \sum_{a=0}^{k} |a_0 + p a\rangle \mbox{~,}
\end{equation}
where $p$ is the order of the number $m$ under multiplication modulo $M$,
$0\leq a_0 < k$ is some constant, and $k = \left[ (2^L - a_0) / p \right]$.
\par
To extract the order $p$, we perform a discrete Fourier transform, which
evolves our system to
\begin{equation}
|\Psi\rangle = \frac{1}{\sqrt{k s^L}} \sum_{c=0}^{2^L -1} \sum_{a=0}^{k} \exp
\left[ \frac{2\pi i (a_0 + p a) c}{2^L} \right] |c\rangle \mbox{~.}
\end{equation}
This can be done in $\mbox{O}(L^2)$ elementary q-bit operations \cite{RMP}.
Now the amplitude of our wavefunction is sharply peaked at $|p\rangle$. It can
be shown that by making a measurement on the first set of q-bits, we have a
probability of at least $\mbox{O}(1 / \log L)$ of getting the correct order
$p$ \cite{Shor,RMP}. So, by repeatedly running our machine $\mbox{O}( \log L)$
times, we are almost sure to get the order $p$ of the multiplicative group
modulo $M$ generated by the integer $m$. Therefore, it requires $\mbox{O}(
(\log M)^2 (\log\log M)^2 \log\log\log M)$ elementary q-bit operations to find
the order $p$ of the group $\langle m\rangle$.
\par
Now we hope that $p$ is an even number and that $\gcd ((m^{p/2}-1) \mbox{mod~}
M, M)$ is a non-trivial factor of $M$. It can be shown that for a randomly
chosen $m$, the probability that the above algorithm does give a non-trivial
factor of $M$ is at least 1/2 provided that $M$ is not of the form $p^k$ or
$2p^k$ for some odd prime $p$ \cite{RMP}.
\par
The remaining case is to recognize and factorize an odd number $M$ in the form
of a prime power. This can be done by classical probabilistic algorithms
whose run time scales like $\mbox{O}((\log M)^2 (\log\log M)^2 \log\log\log
M)$ \cite{Cohen}, which is negligible in comparison with Shor's algorithm.
(See Appendix~B for an equally efficient quantum prime power factorization
algorithm.) Thus, we have an efficient way to factorize a composite number
$M$.
\par
Combining Shor's algorithm with the classical factorization of prime powers,
we are almost sure to find a non-trivial factor of $M$ after $\mbox{O}((\log M
)^2 (\log\log M)^2 \log\log\log M)$ elementary q-bit operations. Moreover, the
power algorithm is one of the major bottle necks in this method.
\section{Quantum Prime Power Factorization Algorithm}
Here we discuss a variant of Shor's algorithm that is useful for factoring a
number $M$ that is of the form $p^n$ for some odd prime $p$. Following Shor,
we can find the order of a number $m$ which is relatively prime to $M \equiv
p^n$ in $\mbox{O}((\log M)^2 (\log\log M)^2 \log\log\log M)$ time. We denote
the set of all integers in ${\Bbb Z}_M$ which are relatively prime to $M$ by
$U({\Bbb Z}_M)$. It can be shown that $U({\Bbb Z}_M)$ is a cyclic group of
order $p^{n-1} (p-1)$ under multiplication modulo $M$ \cite{MNT}. The group
generated by $m$ under multiplication modulo $M$, $\langle m\rangle$, is a
sub-group of $U({\Bbb Z}_M)$. The probability that the order of $\langle m
\rangle$ is divisible by $p^{n-1}$ equals the probability that a randomly
chosen element of ${\Bbb Z}_{p^{n-1} (p-1)}$ is relatively prime to $p^{n-1}$,
which is in turn equal to $1 - 1/p \geq 2/3$. So, the greatest common divisor
of the order of $m$ and $M$ has at least 2/3 chance of being $p^{n-1}$. Thus,
we have a probability of at least 2/3 of finding $p$ by calculating $M / \gcd
(M,r)$ where $r$ is the order of $m$. Once $p$ is found, $M$ can be factorized
easily. The total time required scales as $\mbox{O}((\log M)^2 (\log\log M)^2
\log\log\log M)$.
|
\section{Introduction}
The lattice Boltzmann equation (LBE) method has achieved great success
for simulation of transport phenomena in recent years.
Among different LBE methods, the lattice Boltzmann BGK model is considered
more robust \cite{sdqo}.
Besides, theoretical discussion is easier for the LBGK due to its simple
form.
Some recent theoretical discussions on LBGK \cite{he2,zou,he1}
have enhanced our understanding of the method and the effect of boundary
conditions.
In \cite{zou}, analytical solutions of distribution functions
for plane Poiseuille flow with forcing
and plane Couette flow
have been obtained for the 2-D triangular and square
lattice Boltzmann BGK models. It is found that the bounce-back boundary
condition produces
distribution functions with a first-order error compared with the
analytical distribution functions.
In \cite{he1},
a new technique was developed to seek the analytic solution of
LBGK model for some simple flow.
For example, the velocity profile from the
2-D square and triangular LBGK models are shown to satisfy a second-order
difference equation of the Navier-Stokes equation in the case of plane
Poiseuille flow with forcing and Couette flow.
The technique is generalized in \cite{he2}
to include steady-state flows with both $x$ and $y$ velocities,
which are assumed to be independent of $x$.
The analysis provides a framework to analyze any velocity
boundary condition. For example, the analysis explains why the
velocity boundary condition for the 2-D triangular LBGK
model proposed in \cite{noble}
generates results of machine accuracy
for plane Poiseuille flow with forcing.
In practice, however, a flow is often driven by pressure difference.
In general, the pressure gradient through the
flow field is not a constant and the local pressure gradient is unknown
before solving the flow. Hence the pressure gradient in many cases cannot
be replaced by an external force in LBGK computations.
In this situation, boundary conditions usually need be implemented by
giving prescribed pressure or velocity on some ``flow boundaries'',
which are not solid walls or interfaces of two distinct fluids.
Instead, they are imaginary boundaries inside a flow domain
(e.g. inlet and outlet in a pipe flow). Their existence is purely for the
convenience of study.
The implementation of these boundary conditions in LBGK
is very important but it has not yet been well studied.
Since in lattice Boltzmann method, the pressure is related
to the density by the isothermal
equation of state as $p = c_s^2 \rho$ ($c_s$ is the sound speed of the
model), a specification of pressure difference amounts to a specification
of density difference.
Early works (see, for example, \cite{daryl})
to implement pressure (density) flow boundary condition is simply
to assign the equilibrium distribution computed with the specified
density and some velocity (maybe zero)
to the distribution function.
This method introduces significant errors: the real pressure gradient obtained
in
the simulation for the Poiseuille flow is not a constant. It is approximately
a constant only some distance away from the inlet and outlet of the channel.
Besides, even away from the inlet and outlet region,
the pressure gradient is different from the intended value.
Maier {\it et al.} \cite{bob} proposed an alternative
pressure or velocity flow boundary condition
for the 3-D 15-velocity direction LBGK model,
and their results are greatly improved over the equilibrium distribution
approach.
The pressure or velocity flow boundary condition in \cite{bob}
is obtained through a post-streaming rule to the distribution functions
based on an extrapolation.
However,
this pressure or velocity boundary condition is still to be improved
due to some
inconsistency (see discussion in Section 2).
Its inaccuracy is most noticeable in the following case:
when this pressure or velocity boundary condition is applied to
the modified LBGK \cite{zoui},
which corresponds to a macroscopic momentum equation
having the analytical solution of Poiseuille flow with pressure (density)
gradient, the simulation results are far from the analytical results.
In this paper, we propose a general way to specify
pressure or velocity on flow boundaries.
The implementation is a natural extension of the wall boundary condition
described in our previous paper \cite{he2}.
The result shows a clear improvement to the flow boundary
conditions in \cite{bob} for ordinary LBGK models.
Besides, for the modified LBGK model,
These flow boundary conditions produce results of machine accuracy for
Poiseuille flow.
\section{Pressure or Velocity Flow Boundary Condition of the 2-D Square
Lattice LBGK Model}
\subsection{Governing Equation}
The square lattice LBGK model (d2q9) is expressed as
(\cite{ccmm},\cite{qian2},\cite{hchen}):
\begin{equation}
f_i({\bf x}+\delta {\bf e}_i, t+\delta)-f_i({\bf x},t)=
-\frac 1{\tau}[f_i({\bf x},t)-f_i^{(eq)}({\bf x},t)], \ \ i=0,1,...,8,
\label{eq:lbgk}
\end{equation}
where the equation is written in physical units.
Both the time step and the lattice spacing have the value of $\delta$
in physical units.
$f_{i}({\bf x},t) $ is the density distribution
function along the direction ${\bf e}_i$ at (${\bf x}, t$). The
particle speed ${\bf e}_i$'s are given by
${\bf e}_{i} = (\cos(\pi(i-1)/2), \sin(\pi(i-1)/2), i = 1,2,3,4$, and
${\bf e}_{i} = \sqrt{2} (\cos(\pi(i-4-\frac{1}{2})/2),
\sin(\pi(i-4-\frac{1}{2})/2), i = 5,6,7,8$.
Rest particles of type 0 with ${\bf e}_{0} = 0$ is also allowed
(see Fig.~1).
The right hand side represents the collision term
and $\tau $ is the single relaxation time which controls the rate
of approach to equilibrium.
The density per node, $ \rho $, and the macroscopic flow velocity,
${\bf u} = (u_x, u_y)$,
are defined in terms of the particle distribution function by
\begin{equation}
\sum_{i=0}^8 f_i=\rho, \ \ \ \
\sum_{i=1}^8 f_i {\bf e}_i =\rho {\bf u}.
\label{eq:dens}
\end{equation}
The equilibrium distribution functions
$f_i^{(eq)}({\bf x},t) $
depend only on local density and
velocity and they
can be chosen in the following form (the model d2q9 \cite{qian2}):
\begin{eqnarray}
f_{0}^{(eq)}&=&\frac{4}{9}\rho[1-\frac{3}{2}{\bf u}\cdot{\bf u}],
\mbox{\hspace{1.65 in}}
\nonumber\\
f_{i}^{(eq)}&=&\frac{1}{9}\rho[1+3({\bf e}_{i}\cdot
{\bf u})+\frac{9}{2}({\bf e}_{i}\cdot {\bf u})^2-\frac{3}{2}{\bf u}\cdot
{\bf u}],\;\; i = 1,2,3,4 \\
f_{i}^{(eq)}&=&\frac{1}{36}\rho[1+3({\bf e}_{i}\cdot
{\bf u})+\frac{9}{2}({\bf e}_{i}\cdot {\bf u})^2-\frac{3}{2}{\bf u}\cdot
{\bf u}], \;\; i = 5,6,7,8 . \nonumber
\label{eq:equil}
\end{eqnarray}
A Chapman-Enskog procedure can be applied to Eq.~(\ref{eq:lbgk}) to derive
the macroscopic equations of the model. They are given by:
the continuity equation (with an error term $O(\delta^2)$ being omitted):
\begin{equation}
\frac{\partial \rho}{\partial t} + \nabla \cdot (\rho {\bf u})=0,
\label{eq:cont}
\end{equation}
and the momentum equation (with terms of $O(\delta^2)$ and $O(\delta u^3)$
being omitted):
\begin{equation}
\partial_{t}(\rho u_{\alpha})+\partial_{\beta}(\rho u_{\alpha} u_{\beta})
=
-\partial_{\alpha}(c_s^2 \rho)+\partial_{\beta}(2 \nu \rho S_{\alpha \beta}),
\label{eq:momen}
\end{equation}
where the Einstein summation convention is used.
$S_{\alpha \beta}=\frac{1}{2}(\partial_{\alpha}u_{\beta}+\partial_{\beta}
u_{\alpha})$ is the strain-rate tensor. The pressure is given by
$p = c_s^2 \rho$, where $c_s$ is the speed of sound with
${\displaystyle c_s^2= \frac{1}{3}, }$
and
${\displaystyle \nu=\frac{2 \tau -1}{6} \delta}$,
with $\nu $ being the the kinematic viscosity.
The form of the error terms and the derivation of these equations can be found
in \cite{qian3,hou1}.
In this paper, we will take the Poiseuille flow
as an example to study the pressure (density) or velocity inlet/outlet
condition.
The analytical solution of Poiseuille flow
in a channel with width $2 L $ for the Navier-Stokes equation is given by:
\begin{equation}
u_x = u_0 (1 - \frac{y^2}{L^2}) ,
\;\;\;\; u_y = 0, \;\;\;\;
\frac{\partial p}{\partial x} = - G, \;\;\;\;
\frac{\partial p}{\partial y} = 0, \;\;\;\;
\label{eq:pois}
\end{equation}
where the pressure gradient
$G$ is a constant related to the centerline velocity $u_0$ by
\begin{equation}
G = 2 \rho \nu u_0/L^2 ,
\label{eq:g}
\end{equation}
and the flow density $\rho$ is a constant. The Reynolds number is defined
as Re = $u_0 (2L)/\nu$.
The Poiseuille flow is an exact solution of the steady-state
incompressible Navier-Stokes equations:
\begin{equation}
\nabla \cdot {\bf u}=0.
\label{eq:a64}
\end{equation}
\begin{equation}
\partial_{\beta}( u_{\alpha} u_{\beta})
=- \partial_{\alpha} (\frac{p}{\rho_0}) + \nu \partial_{\beta \beta}
u_{\alpha},
\label{eq:a65}
\end{equation}
On the other hand, the steady-state macroscopic equations of LBGK model,
Eq.~(\ref{eq:lbgk}), is given by:
\begin{equation}
\nabla \cdot (\rho {\bf u})=0 ,
\label{eq:conts}
\end{equation}
\begin{equation}
\partial_{\beta}(\rho u_{\alpha} u_{\beta})
=- \partial_{\alpha}(c_s^2 \rho)+\frac{2\tau-1}{6} \delta
\partial_{\beta} \partial_{\beta} (\rho u_{\alpha}) .
\ \
\label{eq:momens}
\end{equation}
These equations are different from the incompressible Navier-Stokes equations
Eqs.~(\ref{eq:a64},\ref{eq:a65})
by terms containing the spatial derivative of $\rho$. These discrepancies
are called compressibility error in LBE model.
Thus, the Poiseuille flow given by
Eq.~(\ref{eq:pois}) is not the exact solution of Eqs.~(\ref{eq:conts},
\ref{eq:momens}) when pressure (density) gradient drives the flow.
That is, due to change of pressure (density) in the $x-$direction,
$u_x$ is not constant in the $x-$direction, and
the velocity profile of the solution of
Eqs.~(\ref{eq:conts}, \ref{eq:momens})
is no loger a parabolic profile. For a fixed Mach number
($u_0$ fixed),
as $\delta \rightarrow 0$, the velocity
of the LBGK simulation will not converge to the velocity in
Eq.~(\ref{eq:pois}) because the compressibility error becomes
dominating. This phenomenon is seen in the result in \cite{bob},
where the error of velocity increases as the number of lattice grid
increases.
Besides, from $\partial_x (\rho u_x) = 0$ (suppose
$u_y =0$ in the simulation), one can see that
$u_x$ should be increasing linearly along the flow direction since
$\rho$ is decreasing linearly.
This makes the comparison of $u_x$ with the analytical velocity of
Poiseuille flow somehow ambiguous.
To make a more sensible study for Poiseuille flow with pressure (density)
or velocity flow boundary condition, it is better to use the improved
incompressible LBGK model proposed in \cite{zoui}.
The model (called d2q9i)
is given by Eq.~(\ref{eq:lbgk}) with the same ${\bf e}_i$ and
the following equilibrium distributions:
\begin{eqnarray}
f_{0}^{(eq)}&=&\frac{4}{9} [ \rho - \frac{3}{2} {\bf v}\cdot {\bf v} ] ,
\nonumber \\
f_{i}^{(eq)}&=&\frac{1}{9} [ \rho + 3 {\bf e}_{i} \cdot {\bf v} +
\frac{9}{2}({\bf e}_{i} \cdot {\bf v})^2 - \frac{3}{2} {\bf v}\cdot{\bf v} ],
\;
i=1,2,3,4 , \nonumber \\
f_{i}^{(eq)}&=&\frac{1}{36} [ \rho + 3 {\bf e}_{i} \cdot {\bf v} +
\frac{9}{2}({\bf e}_{i} \cdot {\bf v})^2 - \frac{3}{2}{\bf v}\cdot {\bf v} ]
,\; i = 5,6,7,8 ,
\label{eq:eqi}
\end{eqnarray}
and
\begin{equation}
\sum_{i=0}^8 f_{i} = \sum_{i=0}^8 f_{i}^{(eq)} =\rho, \ \ \ \
\sum_{i=1}^8 f_i {\bf e}_{i} =
\sum_{i=1}^8 f_i^{(eq)} {\bf e}_{i} ={\bf v},
\label{eq:densi}
\end{equation}
where ${\bf v}=(v_x, v_y)$ (like the momentum in the ordinary LBGK model)
is used to represent the flow velocity.
The macroscopic equations of d2q9i in the steady-state case (apart from error
terms of $O(\delta^2)$:
\begin{equation}
\nabla \cdot {\bf v}=0 ,
\label{eq:conti}
\end{equation}
\begin{equation}
\partial_{\beta}( v_{\alpha} v_{\beta})
=- \partial_{\alpha}(c_s^2 \rho)+ \nu
\partial_{\beta \beta} v_{\alpha} ,
\label{eq:momeni}
\end{equation}
are exactly the steady-state incompressible
Navier-Stokes equation. In the model,
pressure is related to density by
$c_s^2 \rho = p/\rho_0$ ($c_s^2 = 1/3$) for a flow with constant density like
Poiseuille flow, and $\nu = \frac{2\tau-1}{6} \delta$.
If the flow is steady, d2q9i is superior to d2q9 to simulating incompressible
flows. If the flow is unsteady,
one may consider the continuity equation derived from d2q9 given by
$\partial_t \rho + (\nabla \cdot \rho) {\bf u} +
\rho \nabla \cdot {\bf u} = 0$. In a situation where the first two terms
likely
cancel each other, d2q9 may be of advantage (to approximate the
continuity equation $\nabla \cdot {\bf u}$). If the first two terms have
the same sign, then d2q9i is better.
\subsection{Review of The Velocity Wall Boundary Condition}
It is proved in \cite{he2,he1} that if the flow is steady and independent of
$x$, then the solution $f_i$ of
Eq.~(\ref{eq:lbgk}) produces a velocity
profile that satisfies a difference equation which is a second-order
approximation of the Navier-Stokes equation. If the boundary condition
is chosen correctly,
then the difference equation near the boundary is
consistent with the difference equation inside.
A velocity wall boundary condition is proposed in \cite{he2}
as follows:
take the case of a bottom node in Fig.~1, the boundary is aligned with
$x-$direction with
$f_4, f_7, f_8$ pointing into the wall.
After streaming, $f_0, f_1, f_3, f_4, f_7, f_8$ are known.
Suppose that $u_x, u_y$ are specified on the wall, we need to
determine $f_2, f_5, f_6$ and $\rho$ from Eqs.~(\ref{eq:dens}),
which can be put into the form:
\begin{equation}
f_2+f_5+f_6 =
\rho - (f_0+f_1+f_3+f_4+f_7+f_8) ,
\label{eq:rho}
\end{equation}
\begin{equation}
f_5-f_6 = \rho u_x - (f_1-f_3-f_7+f_8) , \mbox{\hspace{.9in}}
\label{eq:u0}
\end{equation}
\begin{equation}
f_2+f_5+f_6 = \rho u_y + (f_4+f_7+f_8) .\mbox{\hspace{.9in}}
\label{eq:v0}
\end{equation}
Consistency of Eqs.~(\ref{eq:rho},\ref{eq:v0}) gives
\begin{equation}
\rho = \frac{1}{1-u_y} [f_0+f_1+f_3+2(f_4+f_7+f_8)].
\label{eq:rhod}
\end{equation}
\par
We assume the bounce-back rule is still correct for the
non-equilibrium part of the particle distribution normal to
the boundary (in this case, $f_2 - f_2^{(eq)}=f_4 - f_4^{(eq)}$).
With $f_2$ known, $f_5, f_6$ can be found as
\begin{eqnarray}
f_2&=&f_4+\frac 23 \rho u_y , \nonumber \\
f_5&=&f_7-\frac 12 (f_1-f_3)+\frac 12 \rho u_x +\frac 16 \rho u_y , \nonumber
\\
f_6&=&f_8+\frac 12 (f_1-f_3)-\frac 12\rho u_x +\frac 16 \rho u_y.
\label{eq:f256}
\end{eqnarray}
The collision step is applied to the boundary nodes also.
For non-slip boundaries,
this boundary condition is reduced to that in \cite{bob}.
\subsection{Specification of Pressure on a Flow Boundary }
Now let us turn to pressure (density) flow boundary condition.
Its derivation is based on Eq.~(\ref{eq:dens}) as for velocity wall
boundary condition.
Suppose a flow boundary (take the inlet in Fig.~1 as example)
is along the $y-$direction,
and the pressure is to be specified on it. Suppose that $u_y$ is also
specified (e.g. $u_y=0$ at the inlet in a channel
flow).
After streaming, $f_2,f_3,f_4,f_6,f_7$ are known, $\rho
= \rho_{in}, u_y=0$ are specified
at inlet. We need to determine
$u_x$ and $ f_1, f_5, f_8$ from Eq.~(\ref{eq:dens}) as following:
\begin{equation}
f_1+f_5+f_8 =
\rho_{in} - (f_0+f_2+f_3+f_4+f_6+f_7) ,
\label{eq:rhoi}
\end{equation}
\begin{equation}
f_1+f_5+f_8 =
\rho_{in} u_x + (f_3+f_6+f_7) , \mbox{\hspace{.9in}}
\label{eq:u0i}
\end{equation}
\begin{equation}
f_5-f_8 = f_2-f_4+f_6-f_7 .\mbox{\hspace{.9in}}
\label{eq:v0i}
\end{equation}
Consistency of Eqs.~(\ref{eq:rhoi},\ref{eq:u0i}) gives
\begin{equation}
u_x = 1 - \frac{ [f_0+f_2+f_4+2(f_3+f_6+f_7)]}{\rho_{in}} .
\label{eq:u0d}
\end{equation}
\par
We use bounce-back rule for the
non-equilibrium part of the particle distribution normal to
the inlet to find $f_1 - f_1^{(eq)}=f_3 - f_3^{(eq)}$.
With $f_1$ known, $f_5, f_8$ are obtained by the remaining two
equations:
\begin{eqnarray}
f_1&=&f_3+\frac 23 \rho_{in} u_x , \nonumber \\
f_5&=&f_7-\frac 12 (f_2-f_4)+\frac 16 \rho_{in} u_x , \nonumber \\
f_8&=&f_6+\frac 12 (f_2-f_4)+\frac 16 \rho_{in} u_x .
\label{eq:f158}
\end{eqnarray}
The corner node at inlet needs some special treatment. Take the
bottom node at inlet as an example,
after streaming, $ f_3,f_4,f_7$ are known; $\rho$ is specified,
and $u_x=u_y = 0$. We need to determine
$ f_1,f_2,f_5,f_6,f_8$. We use bounce-back rule for the
non-equilibrium part of the particle distribution normal to
the inlet and the boundary to find:
\begin{equation}
f_1 = f_3 + (f_1^{(eq)}- f_3^{(eq)}) = f_3 , \;\;
f_2 = f_4 + (f_1^{(eq)}- f_3^{(eq)}) = f_4,
\label{eq:f13}
\end{equation}
Using these $f_1, f_2$ in Eqs.~(\ref{eq:dens}),
we find:
\begin{equation}
f_5 = f_7, \;\; f_6=f_8=\frac 12 [\rho_{in}-(f_0+f_1+f_2+f_3+f_4+f_5+f_7)] .
\label{eq:f5678}
\end{equation}
Similar procedure can be applied to top inlet node and outlet nodes
including outlet corner nodes.
The case that $\rho$ and
non-zero $u_y$ is specified at a flow boundary along $y-$direction
can be handled in the same way.
Here, it is useful to compare our pressure boundary condition
with that proposed in \cite{bob}, which is given by the following
post-streaming rule (an extrapolation) at inlet:
after streaming, $f_1, f_5, f_8$ are calculated as
\begin{equation}
f_i({\bf x}, t)= f_i^+({\bf x}, t-\delta) -
(f_j({\bf x}, t)-f_j^+({\bf x}, t-\delta) ) , \;\; i = 1, 5,8
\label{eq:posts}
\end{equation}
where $f_i^+({\bf x}, t-\delta)
\equiv f_i({\bf x}, t-\delta) - \frac{1}{\tau}(f_i({\bf x}, t-\delta)
-f_i^{(eq)}({\bf x}, t-\delta))$
is the distribution functions at previous
time step after collision and before streaming, $f_j$ is along ${\bf e}_j$
with ${\bf e}_j = {\bf e}_i - 2 {\bf e}_n$ (the inner normal
${\bf e}_n = {\bf e}_1$ in the case).
Thus, for $i=1,5,8$, $j=3,6,7$ respectively.
The density $\rho $ is set to the specified inlet value and $u_y$ is set to
zero
to compute $f_i^{(eq)}({\bf x}, t)$.
At the bottom, $f_1, f_8$ are computed using
Eq.~(\ref{eq:posts}) and then $f_2, f_5, f_6$ are obtained in the
treatment of wall boundary condition.
Notice, however,
that at the inlet,
$ \sum_{i = 0}^{8} f_i$ may not be equal to the specified density and
$ \sum_{i = 1}^{8} e_{iy} f_i$ may not be equal to zero with this
post-streaming
operation. This inconsistency
causes some inaccuracy in simulations and leaves room
for improvement.
\subsection{Specification of Velocity on a Flow Boundary }
In some calculations, velocities $u_x, u_y$ are specified at a flow boundary
(take the inlet in Fig.~1 as example).
In the case of channel flow, after streaming, $f_2,f_4,f_3,f_6,f_7$ are known
at inlet.
$u_x, u_y $ are specified at inlet
(for the special case of Poiseuille flow, $u_y=0$),
we need to determine
$\rho$ and $ f_1, f_5, f_8$.
This is actually equivalent to a velocity wall boundary
condition. Using our velocity wall boundary condition in \cite{he2}
previously described, we find:
\begin{eqnarray}
\rho&=& \frac{1}{1-u_{x}} [f_0+f_2+f_4+2(f_3+f_6+f_7)], \nonumber \\
f_1&=&f_3+\frac 23 \rho u_{x} , \nonumber \\
f_5&=&f_7-\frac 12 (f_2-f_4)+\frac 12 \rho u_y +
\frac 16 \rho u_{x} , \nonumber \\
f_8&=&f_6+\frac 12 (f_2-f_4)-\frac 12 \rho u_y +
\frac 16 \rho u_{x} .
\label{eq:bc}
\end{eqnarray}
The effect of specifying velocity at inlet is similar to specifying
pressure (density) at inlet. Density difference in the flow
can be generated by the velocity inlet condition.
At the inlet bottom (non-slip boundary), special treatment is needed.
After streaming,
$f_1, f_2, f_5, f_6, f_8$ need to be determined. Using bounce-back
on normal distributions gives:
\[ f_1 = f_3, \;\; f_2 = f_4 , \]
expressions of $x, y$ momenta give:
\begin{eqnarray}
f_5-f_6+f_8 = -(f_1-f_3-f_7)= f_7 , \nonumber \\
f_5+f_6-f_8 = -(f_2-f_4-f_7)= f_7 ,
\label{eq:veli}
\end{eqnarray}
which gives
\begin{eqnarray}
f_5&=&f_7 , \nonumber \\
f_6&=&f_8 = \frac 12 [\rho - (f_0+f_1+f_2+f_3+f_4+f_5+f_7) ] ,
\label{eq:veli1}
\end{eqnarray}
but there is no more equation available to determine $\rho$. The situation
is similar to a corner wall node (the intersection of two perpendicular
walls).
In the situation, $\rho$ at the inlet bottom node can be taken as the
$\rho$ of its neighboring flow node, thus the velocity inlet condition
is specified.
{}From the discussion given above, we can unify boundary conditions (on a wall
boundary or in a flow boundary) in 2-D
simulation on a straight boundary as:
\begin{itemize}
\item Given $u_x, u_y$, find $\rho$ and unknown $f_i$'s.
\item Given $\rho$ and the velocity along the boundary, find the velocity
normal to the boundary and unknown $f_i$'s.
\end{itemize}
Eq.~(\ref{eq:dens}) is used to determine $\rho$ or the normal velocity and
the unknown $f_i$'s. The formula are given in sections 2.2, 2.3, 2.4.
Again, it is useful to compare our
velocity flow boundary condition with that proposed in \cite{bob},
which is given by the following
post-streaming rule (a
zeroth-order extrapolation) at inlet:
after streaming, $f_1, f_5, f_8$ are calculated using
\begin{equation}
f_i({\bf x}, t)= f_i^+({\bf x}, t-\delta) , \;\; i = 1,5,8
\label{eq:postsv}
\end{equation}
where $f_i^+({\bf x}, t-\delta)$ is the distribution function at previous
time step after collision.
After $\rho $ is computed using $f_i$'s,
$f_i^{(eq)}({\bf x}, t)$ can be computed using this $\rho$ and the specified
velocities $u_x, u_y$.
In this approach, the determination of unknown $f_i$'s does not use the
information of known $f_i$'s at present time. This is inconsistent with
the present distribution in the flow.
Suppose that initially, $f_i^{(eq)},\ i = 0, \cdots, 8$ are computed by using
some density $\rho_0$ and the specified $u_x, u_y$, and one assigns
$f_i = f_i^{(eq)}, \ i = 0, \cdots, 8$, then collision does not change
$f_i$, and the post-streaming rule Eq.~(\ref{eq:postsv}) does not
change $f_i$ and $\rho$. Hence $f_i=f_i^{(eq)}$
for all time
in the simulation. This velocity inlet condition amounts to assign
the equilibrium distribution to $f_i$ and it makes a significant error.
The result is worse than that of pressure inlet condition \cite{bob}.
\subsection{Boundary Conditions for the Modified Incompressible Model d2q9i}
The velocity wall boundary condition and flow boundary conditions for d2q9i
are similar to
that of d2q9. It is from equations $\sum_{i=0}^8 f_i = \rho$ and
$\sum_{i=1}^8 {\bf e}_i f_i = {\bf v}$ and hence some modifications are
needed as follows:
\begin{itemize}
\item In wall boundary condition, Eq.~(\ref{eq:rhod}) is replaced by
\begin{equation}
\rho = v_y - [f_0+f_1+f_3+2(f_4+f_7+f_8)].
\label{eq:rhodi}
\end{equation}
\par
and in Eq.~(\ref{eq:f256}), $\rho u_x, \rho u_y$ are replaced by
$v_x, v_y$ respectively.
\item In pressure flow boundary condition, Eq.~(\ref{eq:u0d}) is replaced by
\begin{equation}
v_x = \rho - [f_0+f_2+f_4+2(f_3+f_6+f_7)] ,
\label{eq:u0di}
\end{equation}
and in Eq.~(\ref{eq:f158}), $\rho_{in} u_x $ is replaced by
$v_x $.
\item Similar replacement in velocity flow boundary condition,
Eq.~(\ref{eq:bc}).
\end{itemize}
\section{Numerical Results and Discussion }
We report and discuss the numerical results for Poiseuille flow with
pressure (density) or velocity flow boundary condition. The simulation
is performed on both models d2q9 and d2q9i.
The main result in the simulation of d2q9i is the achievement of machine
accuracy.
The main result in the simulation of d2q9 is the achievement of
second-order accuracy of the boundary conditions.
The width of the channel is assumed to be $2 L = 2$.
We use $nx, ny$ lattice nodes on the $x-$ and $y-$directions, thus,
$\delta = 2/(ny-1)$.
The initial condition is to assign $f_i = f_i^{(eq)}$ computed using
a constant density $\rho_0$, and zero velocities.
The steady-state is reached if
\[ \frac{ \sum_{i} \sum_{j} | u_x(i,j, t+\delta) - u_x(i,j,t)|
+ | u_y(i,j, t+\delta) - u_y(i,j,t)|}
{ \sum_{i} \sum_{j} | u_x(i,j,t)|
+ | u_y(i,j,t)|} \leq \delta \cdot Tol . \]
For model d2q9i, $u_x,u_y$ are replaced by $v_x, v_y$. $Tol$ is a tolerance
usually set to $10^{-10}$.
On the wall, boundary condition discussed in section 2.2
is used to make non-slip boundaries.
We also define a L1 error as:
\begin{equation}
err_1 \equiv \frac{ \sum_{i} \sum_{j} | u_x^t(i,j) - u_x(i,j)|
+ | u_y^t(i,j) - u_y(i,j)|}
{ \sum_{i} \sum_{j} | u_x^t(i,j)|
+ | u_y^t(i,j)|} ,
\label{eq:err1}
\end{equation}
where $u_x^t, u_y^t$ is the analytical velocity.
\subsection{Results of Model d2q9i}
For model d2q9i, we carried out
simulations with a variety of Re, $nx, ny, u_0$ ($u_0$ is the peak
velocity in the channel) using the pressure or velocity flow
boundary condition.
The range of Re is from 0.0001 to 30.0;
the range of $\tau $ is from 0.56 to 20.0 and
the range of $u_0 $ is from 0.001 to 0.4;
the largest density difference simulated (not the limit)
is $\rho_{in} = 5.6, \ \rho_{out}=4.4$ with $nx=5, ny=3$ corresponding to
a pressure gradient of $G=0.1$.
The magnitude of average density
$\rho_0$ is irrelevant for the simulation \cite{zoui}.
For all cases where the simulation
is stable, the steady-state velocity and density show:
\begin{itemize}
\item The velocity field $v_x$ is accurate up to machine accuracy
compared to the analytical solution in Eq.~(\ref{eq:pois}),
$v_y$ is very small with maximum of $|v_y|$ in the whole region being
in the order
of $10^{-12}$. For example, for
$nx = 5, ny = 3, u_0= 0.1, \tau=0.56, $ Re =10, the relative L1 error
of ${\bf v}$ in the whole flow region is
$0.485 \cdot 10^{-10}$.
\item The density is uniform in the cross channel direction, and linear in
the flow direction. The density difference $\rho(i+1,j) - \rho(i,j)$
is a constant through the flow region,
its value is equal (up to machine accuracy) to the analytical
value set by the constant pressure gradient.
\item Velocity $v_x$ is uniform in the $x-$direction, the results are
the same for different $nx$.
\end{itemize}
If the computed velocity were plotted with the analytical velocity,
there would be no difference to naked-eyes.
It is also noticed that with pressure (density) gradient to drive
Poiseuille flow, the maximum Reynolds
number which makes the simulation stable is far less than that with
external forcing. For $ny = 5$, the maximum Re is about 30.
Refinement of mesh can increase the maximum Re.
When the pressure flow boundary condition is replaced by the method in
\cite{bob},
machine accuracy can no longer be obtained,
for a simple case
$nx=17, ny= 9, u_0= 0.03542, \tau=0.67, $ Re =5, with a moderate
pressure gradient of $G=0.001004$, $\rho_{in} = 5.006, \rho_{out} = 4.994$,
the relative L1 error of ${\bf v}$ in the whole flow region is
$0.1824 \cdot 10^{-2}$, with maximum of $|v_y|$ being
$0.1364 \cdot 10^{-3}$, not very small
compared to $u_0$.
The density difference $(\rho(i+1,j) - \rho(i,j))/\delta$
is no longer a constant, its range is from -0.002094 to -0.003872
(the analytical value is $-0.003012 = -G /c_s^2= -3G$).
The result indicates that the pressure or velocity flow boundary condition
proposed in this paper is a clear improvement of that in \cite{bob}.
Similar results of machine accuracy
are obtained by specifying the analytical velocity profile
given in Eq.~(\ref{eq:pois})
at inlet and pressure (density) at outlet by using the flow boundary
conditions in this paper. In the case, there is a uniform
pressure (density) difference in the region. The value of the density
difference
depends on $u_0$ and the outlet density.
\subsection{Results of Model d2q9}
Since the ordinary LBGK model d2q9 is still widely used
for simulations. It is worthwhile to do some
simulations with d2q9 with our flow boundary conditions and show that
they give a second-order accuracy.
We use d2q9 to Poiseuille flow with pressure or velocity flow boundary
condition. Since as $\delta \rightarrow 0$ the computed solution
does not approach the analytical solution, we will use the result
of the finest mesh as the exact solution to compute the error.
Simulations with successively doubled lattice steps are carried out to
observe the convergence.
The example uses fixed Re = 10, $u_0=0.1, \rho_0 = 5$.
The pressure gradient $G$ in Eq.~(\ref{eq:pois})
and then pressure (density) at inlet/outlet can be
obtained as
$G=0.02$ and $\rho_{in}=5.12, \rho_{out}=4.88$ respectively
to be used in the pressure (density) flow boundary condition.
For the velocity flow boundary condition, the analytical velocities in
Eq.~(\ref{eq:pois}) are used at inlet, and $\rho_0$ is specified at outlet.
Thus, the results of the pressure flow boundary condition and the
velocity flow
boundary condition are similar but not identical.
Define $lx=nx-1, ly=ny-1$ to represent the number of lattice steps in
$x-$ and $y-$directions, we use $lx = 4,8,16,32,64,128,256,
ly = 2,4,8,16,32,64,128$ respectively to do the
simulation. The L1 error is defined in Eq.~(\ref{eq:err1}) with
$u_x^t, u_y^t$ being replaced by the computed velocities of $lx=256, ly=128$.
$\tau$ has to be changed as $lx, ly$ are changed to keep the same Re, $u_0$
(values of $\tau$ are included in Table~I).
The convergence result is summerized in Table~I.
The case of $lx=4, ly=2$
was not shown in Table I, because the simulation is unstable
for the velocity inlet condition.
The ratio of two consecutive L1 errors is also shown.
The ratio is approximately equal to 4, indicating a second-order accuracy.
For all runs, it is also observed that
\begin{itemize}
\item Velocity $u_x$ is monotonically increasing in the $x-$direction,
the result is
not sensitive to the value of $lx$, we usually take $lx=2 \ ly$.
If the
pressure boundary condition in \cite{bob} is used, $u_x$ on the centerline
of the channel decreases first, then increases, and decreases again
near the outlet. This behavior deviates from the macroscopic continuity
equation of LBGK $\partial_x (\rho u_x) = 0$ in the case, indicating that
errors are introduced by the pressure boundary condition in \cite{bob}.
An example of centerline velocity $u_x$ as a function of $x$ is presented
in Fig.~4.
\item $u_y$ is very small compared to $u_0$,
with maximum of $|u_y|$ being approximately
of $10^{-6} \cdot u_0$ in typical cases.
If the pressure boundary condition in \cite{bob} is used, maximum value of
$|u_y|$ is like $10^{-3} \cdot u_0$ typically.
\item The density is uniform in the cross channel direction, and linear in
the flow direction. The density difference $\rho(i+1,j) - \rho(i,j)$
is almost a constant through the flow region, its value is equal to the
analytical
value with a fluctuation of
less than 0.03~\% in a worst case observed with Re= 5.
If the pressure boundary condition in \cite{bob} is used,
The fluctuation of density difference $\rho(i+1,j) - \rho(i,j)$
from the analytical value reaches 20~\% for the same case mentioned above with
Re = 5.
\end{itemize}
In the case where the density gradient is small, the computed velocity profile
are close to the analytical velocity profile of Poiseuille flow.
We present an example here with
$nx =9, ny = 5$, Re=0.8333, $u_0=0.008333$ , $G = 0.001667$,
$\rho_{in} = 5.01, \rho_{out} = 4.99$ with both pressure (density) flow
boundary
conditions in this paper and in \cite{bob}.
Fig.~2 shows the velocity $u_x$ as a function of $y$ at
$i=5$ (the middle section of the channel),
Fig.~3 shows the centerline density profile along the $x$ direction, and
Fig.~4 shows the centerline $u_x$ along the $x$ direction.
the solid line represents the corresponding analytical
solution and the symbols $\diamond, +$ represent the computed solutions
with the pressure boundary conditions in this paper and in \cite{bob}
respectively.
Both computed velocities $u_x$ at the mid-channel
has no difference to the analytical solution
to naked eyes (the relative error are of order $10^{-3}$ for both
pressure boundary conditions).
The computed centerline density with our method looks identical to the
analytical solution
(given as
a linear function crossing $\rho_{in}$ and $\rho_{out}$ at inlet and outlet
respectively),
while the centerline density with the method in \cite{bob}
has a discernible difference with the linear function especially near the
inlet and outlet.
The centerline $u_x$ with our method is monotonically increasing,
the behavior is
consistent with the continuity equation of LBGK $\partial_x (\rho u_x) = 0$
in the case,
while the centerline $u_x$ with the method in \cite{bob} has a behavior
inconsistent with the continuity equation near the inlet and outlet.
This again shows an clear
improvement of our method to that in \cite{bob}.
\section{Flow Boundary Conditions and
Preliminary Results for the 3-D 15-velocity
LBGK Model}
Since 3-D model is needed in practical problems, we give a discussion
of the pressure or velocity flow boundary condition for the 3-D
15-velocity LBGK model (d3q15)
and present a brief statement about its simulation results.
The model is based on the LBGK equation Eq.~(\ref{eq:lbgk})
with $i = 0, 1, \cdots, 14$, where ${\bf e}_i, i = 0, 1, \cdots, 14$ are the
column vectors of the following matrix:
\[ E = \left[ \begin{array}{rrrrrrrrrrrrrrr}
0 &1 &-1 &0 &0 &0 &0 &1 &-1 &1 &-1 &1 &-1 &1 &-1 \\
0 &0 &0 &1 &-1 &0 &0 &1 &-1 &1 &-1 &-1 &1 &-1 &1 \\
0 &0 &0 &0 & 0 &1 &-1 &1 &-1 &-1 &1 &1 &-1 &-1 &1
\end{array} \right] \]
and ${\bf e}_i, i = 1, \cdots, 6$ are clasified as type I,
${\bf e}_i, i = 7, \cdots, 14$ are clasified as type II.
The density per node, $ \rho $, and the macroscopic flow velocity,
${\bf u} = (u_x, u_y, u_z)$,
are defined in terms of the particle distribution function by
\begin{equation}
\sum_{i=0}^{14} f_i=\rho, \ \ \ \
\sum_{i=1}^{14} f_i {\bf e}_i =\rho {\bf u}.
\label{eq:3ddens}
\end{equation}
The equilibrium can be chosen as:
\begin{eqnarray}
f_{0}^{(eq)}&=&\frac{1}{8}\rho- \frac 13 \rho {\bf u}\cdot{\bf u} ,
\mbox{\hspace{1.65 in}}
\nonumber\\
f_{i}^{(eq)}&=&\frac{1}{8}\rho +\frac 13 \rho{\bf e}_{i}\cdot
{\bf u}+ \frac{1}{2}\rho({\bf e}_{i}\cdot {\bf u})^2-\frac{1}{6}\rho
{\bf u}\cdot {\bf u} ,\;\; i = I \nonumber \\
f_{i}^{(eq)}&=&\frac{1}{64}\rho +\frac{1}{24} \rho{\bf e}_{i}\cdot
{\bf u}+ \frac{1}{16}\rho({\bf e}_{i}\cdot {\bf u})^2-\frac{1}{48}\rho
{\bf u}\cdot {\bf u} .\;\; i = II
\label{eq:3dequil}
\end{eqnarray}
Since we will use this model for 2-D simulation in this paper, it is
clear to give a projection of the velocities in the $xz$ plane
as shown in Fig.~1. The $y-$axis is pointing into the paper, so are
velocity directions 3,7,9,12,14, while the
velocity directions 4,8,10,11,13 are pointing out (velocity
directions 3,4 have
a projection at the center and are not shown in the figure).
The flow direction is still $x$, and the cross channel direction is $z$.
The macroscopic equations of the model is the same as
Eqs.~(\ref{eq:cont},\ref{eq:momen}) with $c_s^2 = 3/8$, and
$\nu = (2 \tau -1)\delta /6$.
The velocity wall
boundary condition proposed in \cite{he2} has the following version
for the model d3q15:
take the case of a bottom node (wall node) as shown in Fig.~1, the wall
is on $xy$ plane.
After streaming, $f_i, (i = 0,1,2,3,4,6,8,9,12,13)$ are known,
Suppose that $u_x, u_y, u_z$ are specified on the wall, we need to
determine $f_i, i = 5,7,10,11,14$
and $\rho$ from Eqs.~(\ref{eq:3ddens}).
Similar to the derivation in d2q9, $\rho$ is determined by a consistency
condition as:
\begin{equation}
\rho = \frac{1}{1-u_z} [f_0+f_1+f_2+f_3+f_4+2(f_6+f_8+f_9+f_{12}+f_{13})].
\label{eq:3drhod}
\end{equation}
The expression of $z-$ momentum gives:
\begin{equation}
f_5+f_7+f_{10}+f_{11}+f_{14} =
\rho u_z + (f_6+f_8+f_9+f_{12}+f_{13}) ,
\label{eq:3duz}
\end{equation}
If we use bounce-back rule for the
non-equilibrium part of the particle distribution
$ f_i, (i=5,7,10,11,14)$ to set
\begin{eqnarray}
f_i = f_{i+1} + (f_i^{(eq)} - f_{i+1}^{(eq)}) , \; i = 5,7,11 \nonumber \\
f_i = f_{i-1} + (f_i^{(eq)} - f_{i-1}^{(eq)}) , \; i = 10,14
\label{eq:bbn}
\end{eqnarray}
then Eq.~(\ref{eq:3duz}) is satisfied, and all $f_i$ are defined.
In order to get the correct
$x-, y-$momenta, we further fix this $f_5$
(bounce-back of non-equilibrium $f_i$ in the normal direction)
and modify
$f_7, f_{10}, f_{11}, f_{14}$ as in \cite{bob}:
\begin{equation}
f_i \leftarrow f_i + \frac 14 e_{ix}\delta_x +\frac 14 e_{iy}
\delta_y . \;\; i = 7, 10, 11, 14
\label{eq:bbm}
\end{equation}
This modification leaves $z-$ momentum unchanged but adds
$\delta_x, \delta_y$ to the $x-, y-$momenta respectively. A suitable
choice of $\delta_x$ and $\delta_y$ then gives the correct
$x-, y-$momenta. Finally, we find:
\begin{eqnarray}
f_5&=&f_6 + \frac 23 \rho u_z , \nonumber \\
f_i&=&f_{j} + \frac{1}{12} \rho u_z + \frac 14 [e_{ix}(\rho u_x-f_1+f_2) +
e_{iy}(\rho u_y-f_3+f_4) ] ,
\label{eq:3dbc}
\end{eqnarray}
where $j$ is the index corresponding to
${\bf e}_j = -{\bf e}_i$ (e.g., $j=8$ for $i=7$ and $j=9$ for $i=10$).
In the case of non-slip boundary, this boundary condition is reduced to
that in \cite{bob}.
The derivation of pressure (density) flow boundary condition uses a similar
way as for velocity wall boundary condition.
Suppose the boundary (take the case of inlet in Fig.~1)
is on $yz$ plane with specified $\rho_{in}$ and
$u_y=u_z =0$. After streaming, we need to determine
$u_x$ and $f_i, (i = 1,7,9,11,13)$ from Eq.~(\ref{eq:3ddens}).
The consistency condition gives:
\begin{equation}
\rho_{in} u_x = \rho_{in} - [f_0+f_3+f_4+f_5+f_6+2(f_2+f_8+f_{10}+f_{12}+
f_{14}) ] ,
\label{eq:3din1}
\end{equation}
which determines $u_x$ at inlet, using a similar procedure as in deriving
the boundary condition, we find:
\begin{eqnarray}
f_1&=&f_2 + \frac 23 \rho_{in} u_x , \nonumber \\
f_i&=&f_j + \frac{1}{12} \rho_{in} u_x - \frac 14 [e_{iy}(f_3-f_4) +
e_{iz}(f_5-f_6) ] , \;\; i = 7,9,11,13
\label{eq:3din2}
\end{eqnarray}
where $j$ direction is opposite to $i$ direction.
Same procedure as in d2q9 is used at the inlet bottom node (non-slip) to
derive:
\begin{eqnarray}
f_1&=&f_2 , \;\; f_5 = f_6 , \nonumber \\
f_7&=&f_8 - \frac 12 (f_3-f_4) ,\;\;
f_{11} = f_{12} + \frac 12 (f_3-f_4) , \nonumber \\
f_9&=&f_{10}=f_{13}=f_{14} \nonumber \\
&=& \frac 14 [\rho_{in} -
(f_0+f_1+f_2+f_3+f_4+f_5+f_6+f_7+f_8+f_{11}+f_{12})] ,
\label{eq:3din3}
\end{eqnarray}
and similar results for inlet top and outlet condition.
The pressure (density) boundary condition in \cite{bob} is specified
as post-streaming rule (take the inlet as an example):
\begin{eqnarray}
f_i({\bf x}, t)= f_i^+({\bf x}, t-\delta) -
(f_j({\bf x}, t)-f_j^+({\bf x}, t-\delta) ) , \;\; i = 1,7,9,11,13
\label{eq:3din4}
\end{eqnarray}
where $f_i^+({\bf x}, t-\delta)$ is the distribution functions at previous
time step after collision, $f_j$ is along ${\bf e}_j$ with
${\bf e}_j = {\bf e}_i - 2 {\bf e}_n$ (the inner normal
${\bf e}_n = {\bf e}_1$ in the case).
Thus, $j=2,14,12,10,8$ for $i=1,7,9,11,13$ respectively.
Then $\rho$ is set to $\rho_{in}$ and $u_y, u_z$ are set to zero to compute
$f_i^{(eq)}$. Again, the density and $z-$ momentum from $f_i$ may not be
correct, indicating some inconsistency.
The velocity flow boundary condition, which can be viewed
simply a velocity wall boundary condition, can be derived similarly.
But the
corner node with non-slip condition needs some treatment as in the d2q9 case.
The details are easy to work out and omitted here.
Simple modifications are used to derive wall boundary condition, pressure
(density) or velocity flow boundary condition for the improved incompressible
model d3q15i, which is the counterpart of d2q9i in 3-D case.
Simulations on plane Poiseuille flow are performed on d3q15 and d3q15i
using the pressure or velocity flow boundary condition.
The only difference with
2-D simulations is periodic condition is used on $y-$direction, and
initial condition is uniform in $y$ direction with zero velocity.
It is found that the
result is uniform in the $y$-direction and is independent of the number
of nodes in $y-$direction. $u_y$ is identically zero (in the order of
$10^{-16}$ because of round-off error).
The results are very similar, although not
identical, to the results of d2q9, d2q9i.
Again, d3q15i with our boundary condition and pressure (density) flow boundary
condition gives results with machine accuracy, showing a clear improvement
over the pressure boundary condition in \cite{bob}.
\section{Acknowledgments}
Discussions with R. Maier, R. Bernard are appreciated.
Q.Z. would like to thank the Associated Western Universities Inc.
for providing a fellowship and to thank G. Doolen and S. Chen
for helping to arrange his visit to
the Los Alamos National Lab.
\vfill\eject
|
\section{Introduction}
Campi and Krivine~\cite{Campi1988,Campi1992}
introduced a method for distinguishing fragmentation models from one
another. By comparing various expectation
values in which the largest cluster is excluded but the particle
number and fragment multiplicity are held fixed, they showed that
the percolation model has a distinctly different behavior than many
competing nuclear fragmentation models.
In this paper, we analyze another statistical weight we have been
using recently and show
that it shares many of the same properties as percolation theory, a
point already apparent from a consideration of critical
exponents~\cite{Chase1995}.
The method used in this paper is unusual in that it is an exact
computational method. Monte Carlo sampling is avoided by exploiting
some properties of the partition functions, and as such there are no
statistical errors, an immense improvement over earlier
methods.
We begin by assuming that each fragmentation outcome happens with a
probability proportional to the Gibbs
weight~\cite{Mekjian1990,Mekjian1991,Mekjian1992,Cole1989}
\begin{equation}
W(\vec{n})
= \prod_{k \ge 1} {x_{k}^{n_{k}} \over n_{k}!} \;,
\label{eq:weight}
\end{equation}
where $n_{k}$ is the number of fragments of size (or charge) $k$,
and $x_{k}$ is a parameter associated with $k$ sized fragments.
We then define the microcanonical partition function as
\begin{equation}
Z_{A}^{(m)}(\vec{x})
= \sum_{\pi_{m}(A)} W(\vec{n}) \;,
\end{equation}
where $\pi_{m}(A)$ is the set of partitions of $A$
nucleons into $m$ fragments,
i.e. $\sum_{k} k n_{k} = A$, $\sum_{k} n_{k} = m$.
These partition functions satisfy the identity
\begin{equation}
{\partial Z_{A}^{(m)} \over \partial x_{k}}
= Z_{A-k}^{(m-1)}(\vec{x}) \;,
\end{equation}
which allows for the computation of the partition functions
recursively from $\sum_{k} \langle n_{k} \rangle = m$, since
\begin{equation}
\langle n_{k} \rangle
= x_{k} {\partial \over \partial x_{k}} \ln Z_{A}^{(m)}
= x_{k} {Z_{A-k}^{(m-1)} \over Z_{A}^{(m)}} \;.
\end{equation}
Campi defined reduced moments as moments in which
the largest cluster is excluded from the measure, i.e.
\begin{equation}
M_{s}(\vec{n}) \equiv \sum_{k} k^{s} n_{k} - k_{\rm max}^{s}
\end{equation}
where $k_{\rm max}$ is the size of the largest cluster.
This suggested the definition of the reduced variance
$\gamma_{2}$~\cite{Campi1988,Campi1992} for a fragmentation event
should be
\begin{equation}
\gamma_{2}(\vec{n}) = {M_{2} M_{0} \over M_{1}^{2}} = (m-1)
{\sum_{k} k^2 n_{k} - k_{\rm max}^{2} \over
(A - k_{\rm max})^{2}} \;,
\end{equation}
Its expectation value can be computed by breaking events into classes
specified by $k_{\rm max}$, and summing the expectation values
over those classes with the appropriate weight, i.e.
\begin{equation}
\langle \gamma_{2} \rangle = (m-1)
\sum_{k_{\rm max}} {\rm Pr}(k_{\rm max})
{\sum_{k} k^{2} \langle n_{k} \rangle(k_{\rm max}) -
k_{\rm max}^{2} \over (A-k_{\rm max})^{2}} \;,
\end{equation}
where ${\rm Pr}(k_{\rm max})$ is the probability of
$k_{\rm max}$ being the largest cluster size and
$\langle n_{k} \rangle(k_{\rm max})$ is the expectation
value of $n_{k}$ when $k_{\rm max}$ is fixed.
To compute expectation values in which the largest cluster size is
fixed we need to compute the partition function for such ensembles.
Clearly this partition function is given
by all the terms in the microcanonical partition function which have
$x_{k_{\rm max}}$ as the highest $x_{k}$ in the term.
Consider
\begin{eqnarray}
\Delta Z_{A}^{(m)}(k_{\rm max}) & \equiv &
Z_{A}^{(m)}(x_{1}, \ldots, x_{k_{\rm max}}, 0, \ldots, 0)
\nonumber \\
&& -Z_{A}^{(m)}(x_{1}, \ldots, x_{k_{\rm max}-1}, 0, \ldots, 0) \;.
\label{eq:deltaZ}
\end{eqnarray}
We see that $\Delta Z_{A}^{(m)}(k_{\rm max})$ is the partition
function for ensembles with fixed maximum cluster size
$k_{\rm max}$, since the first term collects all terms with
$x_{k_{\rm max}}$ or lower, and the second term eliminates those
terms which don't have an $x_{k_{\rm max}}$. From this result
we can determine ${\rm Pr}(k_{\rm max})$ and
$\langle n_{k} \rangle(k_{\rm max})$, which are given by
\begin{eqnarray}
{\rm Pr}(k_{\rm max})
& = & {\Delta Z_{A}^{(m)}(k_{\rm max}) \over
Z_{A}^{(m)}(x_{1}, \ldots, x_{A})} \\
\langle n_{k} \rangle(k_{\rm max})
& = & \left\{
\begin{array}{ll}
0 & k > k_{\rm max} \\
{Z_{A-k}^{(m-1)}(x_{1}, \ldots, x_{k_{\rm max}}, 0, \ldots, 0) \over
\Delta Z_{A}^{(m)}(k_{\rm max})} & k = k_{\rm max} \\
{\Delta Z_{A-k}^{(m-1)}(k_{\rm max}) \over
\Delta Z_{A}^{(m)}(k_{\rm max})} & k < k_{\rm max}
\end{array} \right.
\end{eqnarray}
This method is quite general and can be applied to other models. For
example, equipartitioning models, which have weights given by
\begin{equation}
W(\vec{n}) = \prod_{k \ge 1} x_{k}^{n_{k}}
\end{equation}
can also be analyzed by this method with some minor modifications.
For example, $x_{k} = 1$ is the model used by Sobotka and
Moretto~\cite{Sobotka1985}.
With these identities there is sufficient information to compute
$\langle \gamma_{2} \rangle$ and other reduced moments for any
$x_{k}$. We use $x_{k} = x/k^{\tau}$ for a variety of reasons
discussed elsewhere~\cite{Chase1994}.
Campi and Krivine~\cite{Campi1992} following Mekjian~\cite{Mekjian1990}
considered this model with $\tau = 1.0$ and showed
that its reduced variance and other related expectation values
had a distinctly different behavior than the percolation
model. Plotting the expected reduced variance vs.
$(m-1)/A$, $\langle \gamma_{2} \rangle(m)$ has a single peak.
The location, height and width of this peak for the two models (and
other models they considered) are completely different,
suggesting the usefulness of this plot in distinguishing fragmentation
models. The choice $\tau = 0$ was considered by Gross,
et.~al.~\cite{Gross1992a,Gross1992b,Jaqaman1992}, and a different
model was analyzed by Pan and Das Gupta~\cite{Pan1995}.
Since that time our interest has turned to the choice
$\tau = 2.5$ because of similarities with percolation theory and Bose
condensation. Namely, the sudden appearance of an infinite cluster in
the infinite $A$ limit and the presence of condensation phenomena. As
such, we have recomputed the Campi plots for
this model and have discovered that they
duplicate the percolation model results in many respects.
Figure~\ref{fig:gamma2kmax}(a) shows the results.
The height and location of the peak
of the reduced variance $\langle \gamma_{2} \rangle$
are the same in both models. The only significant difference is the
width of the peak which is larger in the Gibbs model than
in the percolation model.
Plots of $\langle k_{\rm max} \rangle(m)$ vs. $(m-1)/A$ given
in Fig.~\ref{fig:gamma2kmax}(b) are also
very similar for both models, and the scaling behavior of the
position, width and height with changing $A$ also agree.
Another plot suggested by Campi~\cite{Campi1986} is to divide the
event space by
the maximum cluster size of the event
$k_{\rm max}$ and the reduced second moment $M_{2}$ and
plot the probability of the canonical model being at any particular
point on the graph. This can be also be done exactly for Gibbs
models in a way completely analogous to the way given above. Define
the partition function $Z_{A}(m_{2}; \vec{x})$ as the sum of the Gibbs
weight Eq.~(\ref{eq:weight}) over all partition vectors $\vec{n}$
which satisfy $\sum_{k} k n_{k} = A$, $\sum_{k} k^{2} n_{k} = m_{2}$.
This can be computed by the following recursion,
\begin{equation}
Z_{A}(m_{2}; \vec{x})
= {1 \over A} \sum_{k} k x_{k} Z_{A-k}(m_{2}-k^{2}; \vec{x})
\end{equation}
with $Z_{0}(m_{2}; \vec{x}) = \delta_{m_{2},0}$. If we define
$\Delta Z$ as in Eq.~(\ref{eq:deltaZ}), we find again the partition
function conditioned on $k_{\rm max}$ being fixed, which is
proportional to the probability of having an event with both $m_{2}$
and $k_{\rm max}$. Figure~\ref{fig:campiplots} plots this
probability profile as a contour plot, which reveals the events are
centered on a particular region in this phase space. The slopes of the
edges of this region are related to the critical exponents
according to Campi~\cite{Campi1986}.
Clearly there are differences between percolation theory and a
Gibbs model, but the differences are not as large as
originally suggested by early computations. Indeed the reduced
variance might not reliably distingush percolation from a simple
Gibbs model. A different method is needed to distingush these
models. However, the idea of excluding the largest cluster from the
ensemble averages is a standard procedure in percolation
theory~\cite{Stauffer1992}, and this new technique for doing that
analytically in the Gibbs models shows a particular advantage of
these models over percolation models, which we hope will encourage
further interest in them.
This work supported in part by the National Science Foundation
Grant No. NSFPHY 92-12016.
|
\section{\label{intro}Introduction}
A Newtonian two-body system cannot be completely
isolated from all other masses
in the universe as a consequence of the universality of the gravitational
interaction. In fact, the attraction of the other masses would cause the
binary system to move through approximately inertial spacetime.
This center-of-mass motion should be distinguished from the
relative motion, which is affected by the gradient of the disturbing
forces. Consider, for instance,
the equations of motion for an ``isolated'' two-body system in Newtonian
mechanics
\begin{align}
m_1 \frac{d^2 {\bold{X}}_1}{dt^2} + \frac{G_0 m_1m_2}{|{\bold{X}}_1 - {\bold{X}}_2|^3}
({\bold{X}}_1 - {\bold{X}}_2) = & - m_1 \nabla \Phi({\bold{X}}_1),\notag \\
m_2 \frac{d^2 {\bold{X}}_2}{dt^2} + \frac{G_0 m_1m_2}{|{\bold{X}}_1 - {\bold{X}}_2|^3}
({\bold{X}}_2 - {\bold{X}}_1) = & - m_2 \nabla \Phi({\bold{X}}_2),\label{BasicEq1}
\end{align}
where $G_0$ is Newton's constant of gravitation and
\[\Phi({\bold{X}}):=-\sum_p\frac{G_0m_p}{|{\bold{X}}-{\bold{X}}_p|}\]
represents the combined gravitational potential of all
other masses $m_p$ at ${\bold{X}}_p$ in the universe. Here and throughout this work,
the finite size of astronomical bodies is neglected.
If, in the inertial space coordinates $(X^1,X^2,X^3)$,
the binary system is so far away from the other masses that the
relative distance between the masses comprising the binary
is very small compared to the distance of
the center of mass of the binary to the external masses,
then, to first order in this small ratio, the equation of relative motion
has the form
\begin{equation}\label{beqn}
\frac{d^2 r^i}{dt^2} + \frac{k r^i}{r^3} = - K_{ij}(t) r^j,
\end{equation}
where ${\bold{r}}=(r^1,r^2,r^3): = {\bold{X}}_1 - {\bold{X}}_2$,
$r$ is the length of ${\bold{r}}$, and
$k$ = $G_0(m_1 + m_2)$. Here,
$K_{ij}$, the tidal matrix, is given by
\[
K_{ij} = \frac{\partial^2 \Phi}{\partial X^i \partial X^j}
\]
evaluated at the center of mass of the binary system.
In Newtonian mechanics the gravitational potential
$\Phi$ is a harmonic function; therefore, the symmetric
tidal matrix is trace-free.
It turns out that \eqref{beqn} holds approximately in general
relativity as well, except that $K_{ij}$ would be represented by the
``electric" components of
the Riemannian curvature of the underlying spacetime projected
onto a Fermi frame along the center-of-mass worldline
\cite{mashoon1,mashoon2}. That is,
the equation of relative motion can be considered to be the
Newton-Jacobi equation in the sense that once the
internal Newtonian attraction is
neglected, equation~\eqref{beqn} reduces to the Jacobi equation in Fermi
normal coordinates for the
relative motion of two neighboring geodesics in the underlying
spacetime manifold.
Thus the spacetime coordinates in \eqref{beqn} refer to a local
Fermi system established along
the path of the center of mass of the system.
In our approximate treatment,
we neglect relativistic effects in the binary system.
On the other hand, the
external influences may now include gravitational radiation.
It should be noted in this connection that classical celestial mechanics has
been mainly concerned with the $n$-body problem; however, the
``vacuum'' between these bodies is expected to abound with gravitational
radiation as well as with other radiation fields. It is therefore interesting
to consider the interaction of gravitational waves with $n$-body systems, since
it is estimated that half of all stars are members of binary or multiple
systems.
In this paper, attention is focused on a Newtonian binary system that
undergoes perturbation due to an incident gravitational wave. Let the
spacetime metric due to the gravitational wave be given by
\[
g_{\mu\nu} = \eta_{\mu\nu} + \epsilon\chi_{\mu\nu},
\]
where $\eta_{\mu\nu}$ is the Minkowski
metric, $\epsilon$ is the strength of the perturbation, $0<\epsilon\ll 1$,
and $\chi_{\mu\nu}$ represents the gravitational radiation field.
In the transverse traceless gauge, $\chi_{0\mu}$ = 0 and $\chi_{ij}$
is a symmetric traceless matrix that satisfies the wave equation
$\Box^2 \chi_{ij} = 0$ and the transversality condition
$\partial_j \chi_{ij} = 0$.
It turns out that in this gauge,
\begin{equation}\label{bbeqn}
K_{ij}(t) =
- \frac{1}{2}\epsilon \frac{\partial^2 \chi_{ij}}{\partial t^2}
(t,{\bold{X}}_{\operatorname{cm}}),
\end{equation}
where $(m_1 + m_2){\bold{X}}_{\operatorname{cm}} = m_1{\bold{X}}_1 + m_2{\bold{X}}_2$.
It is possible to fix the position of the center of mass (e.g.,
$\bold{X}_{\operatorname{cm}}={\bold 0}$) in the approximation under consideration here,
since $K_{ij}$ is considered only to first order in $\epsilon$. The
perturbing field $\chi_{ij}$ may be expressed as a Fourier sum
of plane monochromatic waves with wavelengths much larger than
the semimajor axis of the binary system. Such waves could be generated
by the motion of masses during the Hubble expansion, or could be
primordial waves left over from the big bang era. It is therefore
important to note that in our analysis there is no need to specify
the {\em initial conditions} for the interaction of the
waves with the binary; instead, we concentrate on the
``steady-state'' situation involving a dominant plane wave
of frequency $\Omega$ incident on the binary. Hence, the symmetric
and traceless tidal matrix in equation~\eqref{beqn} is given
in Cartesian coordinates by
\begin{align*}
{K}_{11} = & \epsilon\alpha \Omega^2 {\cos^2{\Theta}} \cos{(\Omega t)}, \\
{K}_{12} = & \epsilon\beta \Omega^2 \cos{\Theta} \cos{(\Omega t + \rho)}, \\
{K}_{13} = &-\epsilon \alpha \Omega^2 \cos{\Theta}
\sin{\Theta} \cos{(\Omega t)}, \\
{K}_{22} = & -\epsilon\alpha \Omega^2 \cos{(\Omega t)}, \\
{K}_{23} = & -\epsilon\beta \Omega^2\sin{\Theta} \cos{(\Omega t + \rho)},
\end{align*}
where $\alpha$, $\beta$, and $\rho$ are constants,
and $\Theta$ is the polar angle from the normal to the plane of the
unperturbed orbit to the propagation vector of the incident radiation.
However, for the sake of simplicity we will
explicitly consider only the case of normal incidence, i.e. $\Theta$ = 0.
In this case, the orbital plane will remain fixed; that is,
when the waves are normally incident, the problem reduces to planar
motion under the approximations considered here.
Let the unperturbed Keplerian motion be confined to the $(x,y)$-plane.
The transverse nature of the radiation
field implies that the orbital plane will be unchanged
under the perturbation of normally incident waves.
Thus in \eqref{beqn},
we have $r^1 = x$, $r^2 = y$, and $r^3 = 0$.
The nonzero elements of the tidal matrix for our single-frequency
radiation are
\begin{alignat*}{2}
K_{11} & =& - K_{22} =& \epsilon\alpha \Omega^2 \cos{(\Omega t)},\\
K_{12}& = &K_{21} =& \epsilon\beta \Omega^2 \cos{(\Omega t + \rho)},
\end{alignat*}
where $\epsilon\alpha$ and $\epsilon\beta$ represent the
amplitudes of the two independent linear polarization states of
the low-frequency gravitational wave,
and $\rho$ represents the constant phase difference between them.
The justification for replacing the actual problem with this
rather simplified nonlinear model is that it becomes amenable to
mathematical analysis. It should also be remarked that---within the
limitations discussed in this section---equation~\eqref{beqn}
for the relative motion holds
generally in the Fermi coordinate system established along the
center-of-mass worldline. Thus, for this system,
$K_{ij} = R_{0i0j}$, where $R_{\mu\nu\rho\sigma}$ denotes the Riemannian
curvature due to external sources projected onto the frame
of the center of mass.
In the Newtonian limit of general relativity, each $K_{ij}$
reduces to a second partial derivative of the external
Newtonian potential $\Phi({\bold{X}})$ evaluated along the
path of the center of mass.
On the other hand, for a weak gravitational wave,
equation~\eqref{beqn} holds to first order in the amplitude of the
gravitational potential.
Thus, in general,
the matrix $(K_{ij})$ is a function of the proper time along the path of
the center of mass.
It is always possible to diagonalize this symmetric matrix;
however, its dependence upon time implies that \eqref{beqn}
must then be written in a rotating system of coordinates.
In electrodynamics, the interaction of electromagnetic radiation with a
two-body system constitutes a basic problem (e.g., the scattering and
absorption of light by the Rutherford-Bohr atom).
The gravitational analog of
this problem in the classical regime would involve the scattering and
absorption of gravitational radiation by a Keplerian two-body system.
The wavelength of light is much larger than the Bohr radius of the atom;
therefore, the dominant interaction takes place via the
electric dipole moment
of the atom since electromagnetism is a spin - 1 field.
We expect by analogy
that for gravitational radiation with a (reduced) wavelength
that is much larger
than the semimajor axis of the Keplerian orbit, the dominant
interaction would involve the mass quadrupole moment of the binary since
gravitation is a spin - 2 field.
This approximation corresponds precisely to
dropping higher-order terms in the tidal equation~\eqref{beqn}.
The reciprocity between emission and absorption of
radiation should be noted.
In the quadrupole approximation for the emission of
gravitational radiation, the
waves carry away energy and angular momentum but not linear momentum.
The same holds in the inverse process as well,
except that in general the system can gain or lose energy.
Moreover, a Newtonian binary system emits gravitational radiation
of frequency $\Omega=m\omega$, $m=1,2,3\ldots$, where $\omega$
is the Keplerian frequency of the elliptical orbit. Similarly,
resonant absorption of gravitational waves by an elliptical
binary occurs at $\Omega=m\omega$, $m=1,2,3\ldots$, according
to the linear perturbation analysis~\cite{mashoon2}.
A two-body system continuously emits gravitational radiation according to
general relativity. Gravitational energy in the form of radiation is thus
carried away from the system. Hence, the relative orbit evolves in such a way
that the semimajor axis of the osculating ellipse monotonically shrinks.
This phenomenon of inward spiraling of the members of the binary is
consistent with the timing observations of
the Hulse-Taylor binary pulsar~\cite{ht}~\cite{ht1}.
Direct observational evidence for gravitational radiation does not
exist at present; however, efforts are under way to detect in
the laboratory gravitational waves emitted by astrophysical sources with
$\epsilon\approx 10^{-20}$.
In this work, we will ignore the emission of gravitational
radiation by the binary
system, and concentrate our attention instead on the absorption
process. The flow of energy between the incident radiation and the binary is
not unidirectional, however. The self-gravitating system can absorb energy
from the radiation field or deposit energy into the wave so as to induce an
amplification of the radiation. These issues were first discussed
in connection with the problem of ionization~\cite{mashoon2}
in the context of linear perturbation analysis that in general
breaks down over time. Here we employ the concepts introduced
by Poincar\'e~\cite{hp} for the treatment of nonlinear problems.
These enable us to
prove that periodic orbits exist in the perturbed system for which
energy must steadily flow back and forth between the wave and the binary.
It is
important to emphasize that such periodic orbits occur near resonance
conditions when certain definite phase relationships are satisfied.
If the binary system monotonically absorbs energy from the wave, then the
semimajor axis of the osculating ellipse will grow with time and the system
eventually ionizes. We provide a qualitative picture for such a process in
\S~\ref{scn}.
The gravitational quadrupole interaction may be
illustrated by considering the
Hamiltonian for the relative motion
\begin{equation}\label{bhamiltonian}
{\cal H} = \frac{1}{2} p^2 - \frac{k}{r} + \frac{1}{6} K_{ij}(t) Q_{ij},
\end{equation}
where the quadrupole moment per unit mass is defined by
$Q_{ij} = 3 r^i r^j - r^2 \delta_{ij}$ and in the most general case
of a normally incident gravitational wave packet considered
in this paper the matrix $K_{ij}$
is a traceless symmetric matrix of periodic functions.
Thus $K_{11} = - K_{22} = h_1(t)$,
$K_{12} = K_{21} = h_2(t)$, and there exist $\tau_1$ and $\tau_2$
such that $h_1(t) = h_1(t + \tau_1)$ and $h_2(t) = h_2(t + \tau_2)$.
Here $h_1$ and $h_2$ represent the amplitudes of the two
linearly independent polarization
states of the perpendicularly incident gravitational wave.
Now let
\begin{equation*}
{\cal E} = \frac{1}{2} p^2 - \frac{k}{r}, \quad
{\cal L}^i = \epsilon_{ijk} r^j p^k, \quad
\eta^i = \epsilon_{ijk} p^j {\cal L}^k - \frac{k}{r} r^i
\end{equation*}
denote the Newtonian energy, orbital angular momentum and the
Runge-Lenz vector associated with the relative motion of the
system (per unit reduced mass); then, these otherwise conserved
quantities vary as a consequence
of the coupling of the quadrupole moment of the system to the
curvature of the background spacetime.
Thus, in this quadrupole approximation, the Keplerian
orbit exchanges energy and angular momentum with the radiation field.
At each instant, the relative motion can be described by the
orbital elements of the
osculating ellipse. This osculating ellipse continuously makes transitions
to other osculating Keplerian orbits with different energies and
angular momenta as a consequence of interaction with the external wave.
It is interesting to describe the path of
the system in the six-dimensional manifold of orbital elements;
in fact, this paper is
devoted to the description of periodic paths in this manifold.
The interaction of gravitational radiation with
matter may have played a
significant role in the evolution of the universe.
The treatment presented here
is confined, however, to the interaction of an incident wave
with a {\em Newtonian} binary system. In particular, we neglect all
relativistic effects
in the relative motion of the binary system.
Let the system have a Keplerian frequency $\omega$ and semimajor
axis $a$; then, by Kepler's third law, $\omega^2=k/a^3$.
Relativistic two-body effects may be neglected
provided $k\ll c^2 a$, where $c$ is the speed of light in vacuum.
Moreover, the quadrupole approximation for the interaction of the system
with the gravitational wave is valid if $\Omega a\ll c$.
More generally, our approach is sound provided
\[
\big(\frac{k}{c^2 a}\big)^{1/2}\ll\frac{\Omega}{\omega}
\ll\big(\frac{k}{c^2a}\big)^{-1/2}.
\]
Furthermore, the requirement that the external wave be a small
perturbation of the system implies that
$\Omega/\omega\ll 1/\sqrt{\epsilon }$,
since the strength of the interaction is given by
$\epsilon\Omega^2/\omega^2$, a quantity that must be much smaller than
unity.
Our objective is to determine the
conditions under which periodic Keplerian motions of the binary are
continued to periodic motions under the interaction.
It is an important consequence of our methods, which are originally due
to Poincar\'e~\cite{hp}, that the existence of higher-order perturbing
influences on the orbit, i.e., terms of order at least $\epsilon^2$,
can only affect the shape of the resulting periodic orbit but not its
existence.
In this first treatment of the nonlinear case,
we consider only special cases of the other
interesting questions---such as gravitational
ionization---that are suggested by the electrodynamic
analogy discussed in \cite{mashoon2}.
A treatment of the general problem on the basis of
linear perturbation theory is
contained in previous work \cite{mashoon2}.
Superposition of linear
perturbations due to the Fourier components of a
pulse of gravitational
radiation permits a general analysis in that case;
however, the validity of the
treatment is restricted in time as temporal evolution leads
to a breakdown of
the linear perturbation theory.
Therefore, the intrinsic nonlinearity of the
system must in general be taken into account for applications
in celestial mechanics. In this regard, we mention that the equations
of motion of a binary influenced by the gravitational attraction
of a massive distant third body
can also be treated using the methods developed here; this constitutes
a special limiting case of the three-body problem and is discussed
in Appendix B.
We will be mainly concerned with
the continuation of Keplerian orbits under perturbation
by a resonant gravitational wave. In general, we show that
all but a finite
set of such resonant orbits are not continued to periodic orbits under
the influence of the incident wave and that in general all elements of
the finite exceptional set are, in fact, continued.
The plan of this paper is the following. In \S~\ref{de}, we transform
the perturbation problem to Delaunay elements and obtain explicit
expressions for the transformed perturbation in terms of Fourier
series with coefficients that involve the Bessel functions.
In \S~\ref{ct}, we outline a continuation
method based on the Lyapunov-Schmidt reduction that is
adapted from \cite{ccc}.
In \S~\ref{cko}, the results of the continuation theory are applied
to the perturbation
of a binary influenced by a normally
incident wave. In particular, we find bifurcation equations
for the problem, that is, a system of equations whose simple zeros correspond
to the continuable periodic Keplerian orbits and we show that
these equations indeed have simple zeros.
In \S~\ref{cpw}, we consider the special case of circularly polarized
gravitational waves, a case that is not covered by the
results of \S~\ref{cko}. In this case we show that there are periodic
solutions.
We also show for sufficiently weak perturbations of Keplerian ellipses
that the semimajor axis of the osculating ellipse
remains bounded for all time.
The final section, \S~\ref{scn}, contains a brief discussion
of some additional speculative results on the ionization problem.
Some standard formulas are relegated to Appendix~A, and
in Appendix~B we consider a special case of the three body problem:
A binary influenced by a distant massive third body.
\section{\label{de}Delaunay Elements and Fourier Series Expansion}
In terms of the canonical variables $(p_r,p_\theta,r,\theta)$, that is
\[
x=r\cos\theta,\quad y=r\sin\theta,\quad p_r=\dot r,\quad p_\theta=
r^2\dot\theta,
\]
the Hamiltonian for our perturbation problem, with perturbation
parameter $\epsilon$, may be expressed in the
following general form
\begin{equation}\label{perHam}
{\cal{H}}(p_r,p_\theta,r,\theta)=
\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}
+\epsilon r^2 (\phi(t)\cos{2\theta}+\psi(t)\sin{2\theta}),
\end{equation}
where $\phi$ and $\psi$ are periodic functions with a common period.
We will continue to
use this general form in order to show how our theory can be applied.
However, for the sinusoidal monochromatic gravitational wave
model we will consider only the case where
\begin{equation}\label{phipsi}
\phi(t)=\frac{1}{2}\alpha\Omega^2\cos(\Omega t),\qquad
\psi(t)=\frac{1}{2}\beta\Omega^2\cos(\Omega t+\rho).
\end{equation}
The unperturbed Hamiltonian
\begin{equation}\label{KepHam}
H(p_r,p_\theta,r,\theta)=\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}
\end{equation}
is called the Kepler Hamiltonian. We will consider only those
motions corresponding to negative energy $E=H(p_r,p_\theta,r,\theta)$.
Following S.~Sternberg~\cite[Vol. 2, pp. 234-247]{Stern},
we define
\[ L:=\big(\frac{-k^2}{2E}\big)^{1/2},\qquad G:=p_\theta,\]
and we let $a(1\pm e)$ denote the roots of the quadratic
polynomial
\[
r^2-\frac{2L^2}{k}r+\frac{G^2L^2}{k^2}=0
\]
so that
\begin{equation}\label{ea}
a=\frac{L^2}{k},\qquad e=\frac{1}{L}(L^2-G^2)^{1/2}.
\end{equation}
Here,
$a$ is the semimajor axis and $e$ is the eccentricity of the Keplerian ellipse
with $0\le e<1$.
However, we will only consider the case
$e>0$, that is noncircular orbits.
With this restriction in force,
we define $\widehat{u}$, the eccentric anomaly and $v$, the true anomaly,
implicitly by the formulas
\begin{equation}\label{anomaly}
r=a(1-e\cos \widehat{u}),\qquad r=\frac{a(1-e^2)}{1+e\cos v},
\end{equation}
and new variables $\ell$ and $g$ by
\[
\ell=\widehat{u}-e\sin \widehat{u},\qquad g=\theta-v.
\]
As proved in \cite{Stern}, the change of coordinates
\[
(p_r,p_\theta,r,\theta)\to (L,G,\ell,g)
\]
is canonical. Here $\ell$ and $g$ are ``angle variables'', defined
modulo $2\pi$, while $L$ and $G$ are ``action variables''. The
new coordinates $(L,G,\ell,g)$ are called the Delaunay elements.
In Delaunay variables, our Hamiltonian~\eqref{perHam}
is transformed to
\[
{\cal{H}}=
-\frac{k^2}{2L^2}+\epsilon({\cal{C}}(L,G,\ell,g)\phi(t)+{\cal{S}}(L,G,\ell,g)\psi(t)),
\]
where ${\cal{C}}$ (resp.\ ${\cal{S}}$) is the function obtained by
expressing $r^2\cos{2\theta}$
(resp. $r^2\sin{2\theta}$) in terms of the
Delaunay elements. Using the fact that the change of coordinates
is canonical, the differential equations of motion are given
by the Hamiltonian system
\begin{alignat}{2}\label{eqmotion}
\notag \dot L=&&-&\epsilon(
\frac{\partial {\cal{C}}}{\partial\ell}(L,G,\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial\ell}(L,G,\ell,g)\psi(t)),\\
\notag \dot G=&&-&\epsilon(
\frac{\partial {\cal{C}}}{\partial g}(L,G,\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial g}(L,G,\ell,g)\psi(t)),\\
\notag \dot \ell=&&\omega+&\epsilon(
\frac{\partial {\cal{C}}}{\partial L}(L,G,\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial L}(L,G,\ell,g)\psi(t)),\\
\dot g=&&&\epsilon(
\frac{\partial {\cal{C}}}{\partial G}(L,G,\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial G}(L,G,\ell,g)\psi(t)),
\end{alignat}
where $\omega:=k^2/L^3$ is the frequency of the elliptical Keplerian
orbit.
In order to analyze system~\eqref{eqmotion}, we must find computable
expressions for the partial derivatives of ${\cal{C}}$ and ${\cal{S}}$.
This can be done in several
ways; however, for our purposes, the most useful expressions are obtained
from Fourier series expanded as functions of the angle
variable $\ell$.
The determination of these series is outlined in Appendix A, and the result
can be expressed as
\begin{align}
\notag{\cal{C}}(L,G,\ell,g)=&\frac{5}{2}a^2e^2\cos{2g}
+a^2\sum_{\nu =1}^\infty (A_\nu\cos{2g}\cos {\nu\ell}
-B_\nu\sin{2g}\sin{\nu \ell}),\\
\label{CSfs}{\cal{S}}(L,G,\ell,g)=&\frac{5}{2}a^2e^2\sin{2g}
+a^2\sum_{\nu =1}^\infty (A_\nu\sin{2g}\cos {\nu\ell}
+B_\nu\cos{2g}\sin{\nu \ell}),
\end{align}
where
\begin{align}\label{fcoeff}
\notag A_\nu=&\frac{4}{\nu^2e^2}
(2\nu e(1-e^2)J_\nu'(\nu e)-(2-e^2)J_\nu(\nu e)),\\
B_\nu=&-\frac{8}{\nu^2e^2}\sqrt{1-e^2}\,
(e J_\nu'(\nu e)-\nu (1-e^2)J_\nu(\nu e)).
\end{align}
\section{\label{ct}Continuation Theory}
In order to analyze the continuation (persistence) of periodic solutions
of the Kepler system to system~\eqref{eqmotion}, we use a method
proposed in~\cite{ccc}. Here, we outline the main ideas; the reader
is referred to~\cite{ccc} for the details.
Consider a system of the form
\begin{equation}
\label{bifeq}
\dot u=F(u)+\epsilon h(u,t),
\end{equation}
where $u$ is a coordinate on a manifold $M$ consisting of a
cross product of Euclidean spaces and
tori, $h$ is $2\pi/\Omega$ periodic in its second variable, and
$\epsilon$ is a small parameter.
Let $t\mapsto u(t,\xi,\epsilon)$ denote the solution
of~\eqref{bifeq} with initial condition $u(0,\xi,\epsilon)=\xi$,
$\xi\in M$. Also, we define the $m$th order Poincar\'e map by
${\cal{P}}^m(\xi,\epsilon)=u(2m\pi/\Omega,\xi,\epsilon)$;
it corresponds to a strobe that illuminates the orbit after $m$ cycles
of the perturbation. Of course, a fixed point of
$\xi\mapsto {\cal{P}}^m(\xi,\epsilon)$
corresponds to a periodic orbit of~\eqref{bifeq}.
If $m$ is the smallest such integer for which $\xi$ is a fixed point,
then $\xi$ is the initial point of a subharmonic of order $m$.
Suppose that there is a submanifold ${\cal{Z}}\subset M$ consisting entirely
of fixed points of the unperturbed order $m$ Poincar\'e map, defined by
$p^m(\xi):={\cal{P}}^m(\xi,0)$.
Our continuation theory is a method, one among many, to decide if
any of these fixed points survive after perturbation. More precisely,
we say a point $z\in {\cal{Z}}$, and therefore the unperturbed periodic
orbit of~\eqref{bifeq} with initial point $z$,
is continuable (or that it persists) if
there is a continuous curve $\epsilon\mapsto \gamma(\epsilon)$ in $M$
such that
$\gamma(0)=z$ and
${\cal{P}}^m(\gamma(\epsilon),\epsilon)\equiv \gamma(\epsilon)$.
Here, $\gamma(\epsilon)\in M$ is the initial point of a periodic
solution of~\eqref{bifeq}.
In order to apply the method of~\cite{ccc}, namely Lyapunov-Schmidt
reduction to the Implicit Function Theorem, the fixed-point manifold
(resonance manifold) ${\cal{Z}}$ must satisfy a nondegeneracy condition
relative to the unperturbed Poincar\'e map. To specify this condition,
consider $z\in {\cal{Z}}$ and the derivative $Dp^m(z)$ viewed as a linear
transformation of the tangent space $T_zM$. The base point stays fixed
because $p^m$ is the identity on ${\cal{Z}}$. Moreover, every vector
in $T_zM$ that is tangent to the submanifold ${\cal{Z}}$ is fixed by
$Dp^m(z)$, or, as we will say, every such vector is in the kernel
of the infinitesimal displacement ${\cal{D}}(z)= Dp^m(z)-I$. The manifold
${\cal{Z}}$ is called normally nondegenerate if the kernel of the infinitesimal
displacement is exactly the tangent space $T_z{\cal{Z}}\subset M$.
Equivalently, ${\cal{Z}}$ is normally nondegenerate, if for each $z\in {\cal{Z}}$,
the dimension of the kernel of the infinitesimal displacement at $z$
is equal to the dimension of the manifold ${\cal{Z}}$.
Suppose ${\cal{Z}}$ has dimension $\Delta$ and that it is a
normally nondegenerate submanifold of $M$. In this case the range of
the infinitesimal displacement at each point in ${\cal{Z}}$ has codimension
$\Delta$.
Thus, for $z\in {\cal{Z}}$, there is a vector space complement $\tilde{\cal S}(z)$,
to the range of ${\cal{D}}(z)$. We let $\tilde{s}(z)$ denote the projection of
$T_zM$ to $\tilde{\cal S}(z)$. By choosing local coordinates, we note that
both ${\cal{Z}}$ and $\tilde{\cal S}(z)$ may be identified with ${\Bbb{R}}^\Delta$.
Let $z\in {\cal{Z}}$ and consider the curve in $M$ given by
$\epsilon\mapsto {\cal{P}}^m(z,\epsilon)$. This curve passes through $z$ at
$\epsilon=0$. Its tangent vector at $\epsilon=0$, which
may be identified with the partial derivative ${\cal{P}}^m_\epsilon(z,0)$,
is in $T_zM$.
We define the bifurcation function ${\cal{B}}$ to be the map,
from ${\cal{Z}}$ to the complement $\tilde{\cal S}$ of the
range of the infinitesimal displacement, given by
\[
{\cal{B}}(z)=\tilde{s}(z){\cal{P}}^m_\epsilon(z,0).
\]
In local coordinates ${\cal{B}}:{\Bbb{R}}^\Delta\to{\Bbb{R}}^\Delta$.
We will say $z\in{\cal{Z}}$ is a simple zero of the bifurcation function provided
${\cal{B}}(z)=0$ and the derivative $D{\cal{B}}(z)$ is invertible.
A result in~\cite{ccc} is the following continuation theorem:
\begin{thm}\label{conthm}
If ${\cal{Z}}$ is a normally nondegenerate fixed-point
submanifold of $M$ for system~\eqref{bifeq} and if
$z\in {\cal{Z}}$ is a simple zero of the associated bifurcation function,
then the unperturbed periodic orbit
of~\eqref{bifeq} with initial point $z$ is continuable.
\end{thm}
To use Theorem~\ref{conthm} as a practical tool, we must be able to
compute ${\cal{P}}^m_\epsilon(z,0)$. Fortunately,
this partial derivative can usually be computed. In fact, if
$\Omega=\Omega(\epsilon)$, then
\[
{\cal{P}}^m_\epsilon(z,0)=
-\frac{2m\pi}{\Omega(0)^2}\Omega'(0)\dot u(2m\pi/\Omega(0),z,0)
+u_\epsilon(2m\pi/\Omega(0),z,0).
\]
Thus,
if $t\mapsto W(t)$ is
the solution of the second variational initial value problem
\[
\dot W=DF(u(t,z,0))W+h(u(t,z,0),t),\quad W(0)=0,
\]
then
\[{\cal{P}}^m_\epsilon(z,0)=
-\frac{2m\pi}{\Omega(0)^2}\Omega'(0)F(u(2m\pi/\Omega(0),z,0))+
W(2m\pi/\Omega(0)).\]
In effect,
$W(t)=u_\epsilon(t,z,0)$ with
$W(0)=0$ because $u(0,z,\epsilon)=z$.
In our gravitational
radiation model, it seems appropriate that the frequency
of the gravitational wave is independent of the amplitude
of the wave. Thus, we will assume below that $\Omega$ does not
depend on $\epsilon$. This simplifies the expression for
${\cal{P}}^m_\epsilon(z,0)$ by removing the
``detuning'' .
\section{\label{cko}Continuation of Kepler Orbits}
To apply Theorem~\ref{conthm} to our perturbation problem~\eqref{eqmotion},
we must define a normally nondegenerate fixed-point manifold. For this,
we consider the Kepler orbits that are
in resonance with the periodic perturbation.
If there are fixed relatively prime positive integers $m$ and $n$
and a fixed value of $\omega$, the frequency of the Keplerian
orbit, such that
\[
m\frac{2\pi}{\Omega}=n\frac{2\pi }{\omega},
\]
then the unperturbed solution of~\eqref{eqmotion} starting at
$(L,G,\ell,g)$ is given by
\[t\mapsto (L,G,\omega \widehat{t}+\ell,g),\]
where $\widehat{t}=t-t_0$ and $t_0$ is an integration constant that
denotes the starting instant of time.
A detailed analysis shows that $t_0$ can be set equal to zero here
without loss of generality.
Since $\ell$ is defined modulo $2\pi$, this solution is periodic
of period $2\pi /\omega$. Moreover, the $m$th order unperturbed
Poincar\'e map is defined by
\[
p^m(L,G,\ell,g)=(L,G,2\pi m\frac{\omega}{\Omega}+\ell,g).
\]
If we define the three-dimensional manifold
\[
{\cal{Z}}^L:=\{(L,G,\ell,g): m\omega=n\Omega\},
\]
and recall that $\ell$ and $g$ are defined modulo $2\pi$, then
it follows immediately that ${\cal{Z}}^L$ is fixed by $p$.
To check that ${\cal{Z}}^L$ is normally nondegenerate, we compute
\[{\cal{D}}(L,G,\ell,g)=Dp^m(L,G,\ell,g)-I=
\left(
\begin{array}{cccc}
0 & 0 & 0 & 0 \\
0 & 0 & 0 & 0 \\
-6\pi n/L & 0 & 0 & 0 \\
0 & 0 & 0 & 0
\end{array}
\right)
\]
and we note that the infinitesimal displacement
has a three-dimensional kernel that is spanned by the usual basis vectors
\[
\left(
\begin{array}{c}
0 \\
1 \\
0 \\
0
\end{array}
\right),\quad
\left(
\begin{array}{c}
0 \\
0 \\
1 \\
0
\end{array}
\right),\quad
\left(
\begin{array}{c}
0 \\
0 \\
0 \\
1
\end{array}
\right).
\]
Moreover, the range of the infinitesimal displacement is complemented
by the span of the vectors
\[
\left(
\begin{array}{c}
1 \\
0 \\
0 \\
0
\end{array}
\right),\quad
\left(
\begin{array}{c}
0 \\
1 \\
0 \\
0
\end{array}
\right),\quad
\left(
\begin{array}{c}
0 \\
0 \\
0 \\
1
\end{array}
\right).
\]
To compute the bifurcation function associated with~\eqref{eqmotion},
we must compute the partial derivative
${\cal{P}}^m_\epsilon(G,L,g,\ell,0)$ on the manifold
${\cal{Z}}^L$ and then project the result into the complement
of the range of the infinitesimal displacement.
To do this, we simply solve the variational initial value problem
\begin{alignat*}{2}
\dot L_\epsilon=&&
&-\frac{\partial {\cal{C}}}{\partial\ell}(L,G,\omega t+\ell,g)\phi(t)-
\frac{\partial {\cal{S}}}{\partial\ell}(L,G,\omega
t+\ell,g)\psi(t),\\
\dot G_\epsilon=&&
&-\frac{\partial {\cal{C}}}{\partial g}(L,G,\omega t+\ell,g)\phi(t)-
\frac{\partial {\cal{S}}}{\partial g}(L,G,\omega
t+\ell,g)\psi(t),\\
\dot \ell_\epsilon=&&
-\frac{3k^2}{L^4}L_\epsilon&+
\frac{\partial {\cal{C}}}{\partial L}(L,G,\omega t+\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial L}(L,G,\omega
t+\ell,g)\psi(t),\\
\dot g_\epsilon=&&
&\quad\frac{\partial {\cal{C}}}{\partial G}(L,G,\omega t+\ell,g)\phi(t)+
\frac{\partial {\cal{S}}}{\partial G}(L,G,\omega t+\ell,g)\psi(t)
\end{alignat*}
with zero initial values and then project the solution computed
at $t=m2\pi/\Omega$ into the complement of the range of the infinitesimal
displacement.
{}From this procedure, we obtain the following bifurcation function
\[
{\cal{B}}(G,\ell,g)=
(B^L(G,\ell,g),B^G(G,\ell,g),B^{g}(G,\ell,g)),
\]
where
\begin{equation}\label{biffunc}
B^L(G,\ell,g):=-\frac{\partial {\cal{I}}}{\partial\ell},
\quad
B^G(G,\ell,g):=-\frac{\partial {\cal{I}}}{\partial g},
\quad
B^g(G,\ell,g):=\frac{\partial {\cal{I}}}{\partial G},
\end{equation}
and
\[
{\cal{I}}:=\int_0^{m 2\pi/\Omega}\big[
{\cal{C}}(L,G,\omega t+\ell,g)\phi(t)
+{\cal{S}}(L,G,\omega t+\ell,g)\psi(t)\big] \,dt.
\]
Using the resonance relation, we have
\[
{\cal{I}}=\int_0^{m 2\pi/\Omega}\big[
{\cal{C}}(L,G,\frac{n}{m}\Omega t+\ell,g)\phi(t)
+{\cal{S}}(L,G,\frac{n}{m}\Omega t+\ell,g)\psi(t)\big]\,dt
\]
and, after changing the variable to $\widehat{\sigma}=\Omega t/m+\ell/n$,
we obtain
\begin{align*}
{\cal{I}}=&\frac{m}{\Omega}\int_{\ell/n}^{2\pi+\ell/n}\big[
{\cal{C}}(L,G,n\widehat{\sigma},g)\phi(m(\widehat{\sigma}-\ell/n)/\Omega) \\
&\hspace*{.4in}+{\cal{S}}(L,G,n\widehat{\sigma},g)\psi(m(\widehat{\sigma}-\ell/n)/\Omega)\big]\,d\widehat{\sigma}.
\end{align*}
Using the fact that the last integrand is periodic with period $2\pi$
as a function of $\widehat{\sigma}$ and substituting
$\phi$ and $\psi$ given in~\eqref{phipsi},
we find
\begin{align*}
{\cal{I}}=&\frac{m\Omega}{2}\int_{0}^{2\pi}\big[
\alpha{\cal{C}}(L,G,n\widehat{\sigma},g)\cos(m\widehat{\sigma}-\frac{m\ell}{n}) \\
&\hspace*{.4in}+\beta{\cal{S}}(L,G,n\widehat{\sigma},g)\cos( m\widehat{\sigma}-\frac{m\ell}{n}+\rho)\big]
\,d\widehat{\sigma}.
\end{align*}
To compute ${\cal{I}}$, we substitute the Fourier series~\eqref{CSfs}
for ${\cal{C}}$ and ${\cal{S}}$ into the last expression for ${\cal{I}}$ and use
trigonometric relations together with the fact that
$m$ and $n$ are relatively prime, to conclude that
${\cal{I}}=0$ unless $n=1$. Of course, there may be continuable
orbits for $n>1$, but they are not detected by our first
order method. In case $n=1$, that is for the $(m:n)=(m:1)$ resonance,
we find that
\begin{equation*}
\begin{split}
{\cal{I}}=\frac{1}{2}\pi ma^2\Omega\big(
\alpha(&A_m\cos{m\ell}\cos{2g}-B_m\sin{m\ell}\sin{2g})\\
+&\beta (A_m\cos(m\ell-\rho)\sin{2g}
+B_m\sin(m\ell-\rho)\cos{2g})\big).
\end{split}
\end{equation*}
This result can be rewritten as
\begin{align}
\notag {\cal{I}}=&\frac{1}{4}\pi m a^2\Omega\big[
(A_m+B_m)(\alpha-\beta\sin\rho)\cos(2g+m\ell) \\
\notag &+(A_m-B_m)(\alpha+\beta\sin\rho)\cos(2g-m\ell)
+\beta(A_m+B_m)\cos\rho\sin(2g+m\ell) \\
&+\beta(A_m-B_m)\cos\rho\sin(2g-m\ell) \big].
\label{preB}
\end{align}
It is possible to express equation~\eqref{preB} in a more
compact form, in the usual manner, by defining
\begin{align}
\notag{\cal{E}}\cos\sigma:=
\beta\cos\rho,\quad &{\cal{E}}\sin\sigma:=\alpha+\beta\sin\rho,\\
\label{EF}{\cal{F}}\cos\tau:=
\beta\cos\rho,\quad &{\cal{F}}\sin\tau:=\alpha-\beta\sin\rho,
\end{align}
so that
\[
{\cal{E}}=(\alpha^2+\beta^2+2\alpha\beta\sin\rho)^{1/2},\quad
{\cal{F}}=(\alpha^2+\beta^2-2\alpha\beta\sin\rho)^{1/2},
\]
and
\[
{\cal{I}}=\frac{1}{4}\pi m a^2\Omega\big[
(A_m+B_m){\cal{F}}\sin(2g+m\ell+\tau)
+(A_m-B_m){\cal{E}}\sin(2g-m\ell+\sigma)\big].
\]
The simple zeros of the bifurcation function are then the same as the
simple zeros of
\begin{align}
\notag {\cal{F}}(A_m+B_m)\cos(2g+m\ell+\tau)=&0,\\
\notag {\cal{E}}(A_m-B_m)\cos(2g-m\ell+\sigma)=&0,\\
\label{B}
\Big(\frac{\partial A_m}{\partial G}+\frac{\partial B_m}{\partial G}\Big){\cal{F}}
\sin(2g+m\ell+\tau)+
\Big(\frac{\partial A_m}{\partial G}-\frac{\partial B_m}{\partial G}\Big){\cal{E}}
\sin(2g-m\ell+\sigma)=&0.
\end{align}
To obtain explicit formulas for the partial derivatives of the
functions $A_m$ and $B_m$ with respect to $G$, we assume that
$G>0$ so that
\[\frac{\partial e}{\partial G}=-\sqrt{1-e^2}\,\frac{1}{Le}.\]
In case $G<0$, the partial derivative has the opposite
sign and the subsequent analysis is similar. We use
\eqref{A2},\eqref{A3},\eqref{A4} (from Appendix A), and \eqref{fcoeff}
to obtain
\begin{align}
\notag\frac{\partial A_m}{\partial G}=&-\sqrt{1-e^2}\,\frac{4}{L m^2 e^4}
\big((2m^2(1-e^2)^2+4)J_m(me)-me(6-e^2)J_m'(me)\big),\\
\label{parAB}\frac{\partial B_m}{\partial G}=& -\frac{8}{Lm^2e^4}
\big(-3m(1-e^2)J_m(me)+e((2-e^2)(1-m^2e^2)+m^2)J_m'(me)\big).
\end{align}
The simple zeros of the function ${\cal{B}}$ given by~\eqref{biffunc}
correspond to the continuable periodic orbits. Equivalently,
the continuable periodic orbits correspond to the simple solutions of the
system of equations given by~\eqref{B}. In order to find the
simple zeros of~\eqref{B}, we will use
the following proposition:
\begin{prop}\label{gbaseprop}
If ${\cal{E}}{\cal{F}}\ne 0$ (that is, $|\alpha|\ne |\beta|$, or
$|\alpha|=|\beta|$ but $|\sin\rho| <1$) and if the system of
equations~\eqref{B} has a solution,
then \begin{equation}\label{gbase}
(A_m^2-B_m^2)\Big\{
\Big(\frac{\partial A_m}{\partial G}+\frac{\partial B_m}{\partial G}\Big)^2
{\cal{F}}^2-
\Big(\frac{\partial A_m}{\partial G}-\frac{\partial B_m}{\partial G}\Big)^2
{\cal{E}}^2\Big\}
\end{equation}
is zero.
\end{prop}
\begin{pf}
If~\eqref{B} has a solution and
equation~\eqref{gbase} does not vanish, then, since the first factor
of equation~\eqref{gbase}
is not zero, we have
\[\cos(2g+m\ell+\tau)=0,\qquad \cos(2g-m\ell+\sigma)=0.\]
This implies that
\[\sin(2g+m\ell+\tau)=\pm 1,\qquad \sin(2g-m\ell+\sigma)=\pm 1,\]
and, since the third equation of~\eqref{B} is zero, that
\[
\Big(\frac{\partial A_m}{\partial G}
+\frac{\partial B_m}{\partial G}\Big){\cal{F}}\pm
\Big(\frac{\partial A_m}{\partial G}
-\frac{\partial B_m}{\partial G}\Big){\cal{E}}=0,
\]
in contradiction to the fact that~\eqref{gbase} is not zero.
\end{pf}
Proposition~\ref{gbaseprop}
reduces the search for solutions to several cases.
Just note that as soon as one of the factors of~\eqref{gbase}
vanishes, the value of $e$ and hence the
values of
\[
A_m-B_m,\quad A_m+B_m,\quad \frac{\partial A_m}{\partial G},\quad
\frac{\partial B_m}{\partial G}
\]
are fixed. Thus, equations~\eqref{B} reduce to solvable
trigonometric equations. It is important to note that
Proposition~\ref{gbaseprop} does not cover the interesting case
of circular polarization, which is therefore deferred to~\S~\ref{cpw}.
To study the zeros of~\eqref{gbase}
we will use the following proposition.
\begin{prop}\label{zeroprop}
For the $(1:1)$ resonance, the functions $A_1+B_1$ and $A_1-B_1$
appearing in equations~\eqref{B} and viewed as functions of $e$
are both negative on the interval $0<e<1$. The functions
$\partial A_1/\partial G$ and $\partial B_1/\partial G$, viewed
as functions of $e$, each
have exactly one simple zero on the interval and their zeros
are distinct.
For the $(m:1)$ resonance with $m>1$,
the range of the function
$(\partial A_m/\partial G)/(\partial B_m/\partial G)$, viewed
as a function of $e$ on the interval $0<e<1$, contains the interval
$[-1,1]$.
\end{prop}
\begin{pf}
We will outline the proof of the proposition. Some of the computations
were checked using a computer algebra system.
Consider the case $m=1$.
We will use the following elementary lemma~\cite[Lemma 3.5]{cj}
to show that the function $f$ defined by
\begin{align}
\notag f(e):=&\frac{e^2}{4}(A_1-B_1)\\
\label{fe} =&2e((1-e^2)+\sqrt{1-e^2})J'_1(e)
-((2-e^2)+2(1-e^2)^{3/2})J_1(e)
\end{align}
is negative on the interval $I_0:=\{e:0<e<1\}$.
\begin{lem}\label{cjlem}
Suppose $f$ is a smooth function defined on an interval $[a,b)$
with the additional property that there is a number
$\epsilon>0$ such that $f(x)f'(x)>0$ for $a<x<a+\epsilon$.
If there are smooth functions $p$, $q$, and $r$
defined on $(a,b)$ such that $p(x)r(x)>0$ and
\begin{equation}\label{zde}
p(x)f''(x)=q(x)f'(x)+r(x)f(x)
\end{equation}
on the interval $(a,b)$, then $f$ is strictly monotone on $[a,b)$.
In particular, $f(x)$ has the same sign on $(a,b)$ that
it does on $(a,a+\epsilon)$.
\end{lem}
The function $f$ defined by~\eqref{fe} satisfies a differential
equation of the form~\eqref{zde}
with $x=e$, $w:=\sqrt{1-e^2}$, and
\begin{align*}
p(e):=&e^4w^2(5w+2),\\
q(e):=&e^3(15w^3 +4w^2+2),\\
r(e):=&e^2(5w^5-8w^4-16w^3+26w^2+24w+4).
\end{align*}
To test the sign of $p(e)r(e)$,
we change the variable to $w$ and note that
$0<w<1$. Let
\[
p^*(w):=p(\sqrt{1-w^2}),\quad r^*(w):=r(\sqrt{1-w^2}).
\]
Clearly, $p^*$ is positive on $0<w<1$. The second factor of
$r^*$ is easily shown to be positive on the same interval.
For example, the second factor is positive at $w=0$ and has no
roots in the interval. The fact that there are no roots can
be checked by computing a Sturm sequence (cf.~\cite{hen}).
This proves $p(e)r(e)>0$ for $e\in I_0$.
To complete the proof of this case, it remains only to
show that there is some $\epsilon>0$ such that
$f(e)<0$ and $f'(e)<0$ on the interval
$0<e<\epsilon$. To do this, we simply note that the Taylor series of
$f$ and $f'$ at $e=0$ are given by
\[
f(e)=-\frac{7}{48}e^5+O(e^7),\quad f'(e)=-\frac{35}{48}e^4+O(e^6).
\]
The fact that $A_1+B_1$ is negative on the interval $I_0$ can
be proved in a similar manner.
We must show that $\partial A_1/\partial G$ has exactly one simple
root on the interval $I_0$.
To do this it suffices to
prove that the function given by
\[
e\mapsto (6-e^2)J_2(e)+e(2e^2-3)J_1(e)
\]
has exactly one simple zero on $I_0$.
Equivalently, using the fact that $J_2(e)$ does not vanish on
$I_0$, it suffices to show that the function
\[f_0(e):=\frac{6-e^2}{2e^2-3}+e\frac{J_1(e)}{J_2(e)}\]
has the same property.
This fact follows from the expression~\eqref{parAB} for
$\partial A_1/\partial G$,
the recurrence formula
\begin{equation}\label{newrec}
\nu J_\nu(x)-xJ_\nu'(x)=xJ_{\nu+1}(x),
\end{equation}
that is a simple consequence of equation~\eqref{A2} of
Appendix A,
and the connection between $e$ and
$G$.
The fact that $f_0$ has exactly one simple zero is
a consequence of the following three propositions:
$(a)$ The function $f_1(e):=(6-e^2)/(2e^2-3)$ is monotone decreasing on
$I_0$.
$(b)$ The function $f_2(e):=eJ_1(e)/J_2(e)$ is monotone decreasing on
$I_0$.
$(c)$ The function $f_0$ has a zero in $I_0$.
Statement $(a)$ is immediate: $f_1'$ is negative on the interval.
Statement $(b)$ follows from the product representation of the
Bessel function $J_\nu$ given by~\cite{as}
\[
J_\nu(x)=\frac{x^\nu}{2^\nu\Gamma(\nu+1)}\prod_{s=1}^\infty
\big(1-\frac{x^2}{j_{\nu,s}^2}\big),
\]
where $j_{\nu,s}$ denotes the $s$th zero of $J_\nu$ and the fact that
the zeros are interlaced as follows:
\[j_{\nu,1}<j_{\nu+1,1}<j_{\nu,2}<j_{\nu+1,2}<j_{\nu,3}<\cdots . \]
Statement $(c)$ follows from two facts:
$\lim_{e\to 0^+}f_0(e)=2$ and $f_0(1)<0$.
The first fact is immediate from the above product representation or
from the Taylor series for the Bessel
functions at $e=0$. To obtain the inequality, we use the
product representation of the Bessel functions to deduce
\[
f_0(1)=-5+4 \prod_{s=1}^\infty\big(1-\frac{1}{j_{1,s}^2}\big)/
\prod_{s=1}^\infty\big(1-\frac{1}{j_{2,s}^2}\big).
\]
Since $j_{2,s}>j_{1,s}>1$ for all $s$, the
quotient of the two products is less than unity, as required.
The fact that $\partial B_1/\partial G$ has a unique simple
zero is proved using a similar analysis.
Just as before, it suffices to show that the following function
has exactly one simple zero on $I_0$:
\[
\tilde{f}(e):=-e^3\frac{J_1(e)}{J_2(e)}+e^4-3 e^2+3.
\]
Both terms are monotone decreasing on $I_0$ with $\tilde{f}(0)=3$ and
$\tilde{f}(1)<0$.
We claim the zeros of $\partial A_1/\partial G$
and $\partial B_1/\partial G$ do not occur at the same point.
After division by nonzero factors and the substitution $m=1$
in~\eqref{parAB}, it suffices to show that the following functions
do not have a common zero on $I_0$:
\begin{align*}
\widehat{\alpha}(e):=&(2(1-e^2)^2+4)J_1(e)-e(6-e^2)J_1'(e),\\
\widehat{\beta}(e):=&-3(1-e^2)J_1(e)+e((2-e^2)(1-e^2)+1)J_1'(e).
\end{align*}
If the functions do have a common zero, then the function
\[
\widehat{f}(e):=\widehat{\alpha}(e)-e\widehat{\beta}(e)
\]
has at least one zero on $I_0$. To prove the claim, we
show that $\widehat{f}$ is negative on $I_0$.
Using the recurrence formulas for Bessel's functions, we find that
\[
\widehat{f}(e)=(-e^6+3e^4+e^3-3e^2-6e)J_0(e)+(e^5+2e^4-6e^3-5e^2+6e+12)J_1(e).
\]
To prove this function is negative on $I_0$ we will apply
Lemma~\ref{cjlem}. We find that $\widehat{f}$ satisfies a differential
equation of the specified form with
\begin{align*}
\wp(e):=& e^2(e^{10}-7e^8-4e^7+31e^6+14e^5-55e^4-45e^3+84e^2+9e-36),\\
\widehat{q}(e):=& e(11e^{10}-63e^8-32e^7+217e^6+84e^5-275e^4 -180e^3+252e^2+18e-36),\\
\wr(e):=&-e^{12}-27e^{10}+12e^9+95e^8+11e^7-235e^6\\
&+273e^5-19e^4-468e^3+351e^2+90e-108.
\end{align*}
Using Sturm sequences, it can be proved that
$\wp(e)\wr(e)>0$ for $e\in I_0$. Moreover, we find that
\[
\widehat{f}(e)=-\frac{3}{4}e^3+O(e^4),\quad\widehat{f}(e)\widehat{f}'(e)=\frac{27}{16}e^5+O(e^6).
\]
This completes the proof of the claim.
In case $m>1$, it suffices to consider the range of the function
$F_m$ given by
$e\mapsto (\partial A_m/\partial e)/(\partial B_m/\partial e)$.
A computation shows that the Taylor series of both the numerator and
the denominator of $F_m$ is given by
$-5 e+O(e^2)$ in case $m=2$ and, in case $m>2$, by
\[
\frac{8m^m(m-1)(m-2)}{2^mm!m^2}e^{m-3}+O(e^{m-2}).
\]
It follows that $\lim_{e\to 0^+}F_m(e)=1$.
We claim that
$e\mapsto \partial B_m/\partial e$ has at least one zero on $I_0$.
If not, then $B_m$ is a monotone function of $e$. A computation
shows that $B_m$ has a removable singularity at $e=0$ and that
\[
B_m(e)=\frac{8m^m(m-1)}{2^mm!m^2}e^{m-2}+O(e^{m-1}).
\]
If $m>2$, then $B_m$ is increasing for $0<e<<1$. But,
from the definition of $B_m$, we have
$\lim_{e\to 1^-}B_m(e)=0$, in contradiction.
For $m=2$, we find that
\[B_2(e)=1-\frac{5}{2}e^2+O(e^4) \]
and $B_2$ decreases for $0<e<<1$. But, $B_2$ is negative
for $0<<e<1$. To see this just note that near $e=1$, the
sign of $B_2$ is determined by $-J_2'(2e)$. By standard properties
of the Bessel functions (cf.~\cite{as}), $J_2(x)$ is positive
on the interval $(0,j_{2,1}')$, where $j_{\nu,s}'$
denotes the $s$th zero of $J_\nu'$.
Since $\nu\le j_{\nu,s}'$, we have that $-J_2'(2e)<0$ for $0<< e<1$.
Again, since $\lim_{e\to 1^-}B_2(e)=0$, we have
a contradiction.
This proves the claim.
Suppose for the moment that $\partial B_m/\partial e>\partial A_m/\partial e$
on the interval $0<e<1$ and consider the first zero $e_*$ of
$e\mapsto \partial B_m/\partial e$. It follows that
$\partial A_m/\partial e(e_*)<0$
while $\partial B_m/\partial e>0$ on the interval $0<e<e_*$. Thus,
we have
$\lim_{e\to e_*^-}F_m(e)=-\infty$ and the range
of $F_m$ contains the interval
$(-\infty,1]$, as required.
To complete the proof, it suffices to show that the function $G_m$
given by
\[
e\mapsto m^2 e^3\big( \partial B_m/\partial e-\partial A_m/\partial e\big)
\]
for $m>1$ is positive on the interval $0<e<1$.
This fact follows from Lemma~\ref{cjlem}.
The Taylor series of $G_m$ at $e=0$ is given by
\[
G_m(e)=\frac{(5m+2)m^m}{2^mm!(m+1)}e^{m+4}+O(e^{m+5}).
\]
Thus, it follows that $G_m(e)G_m'(e)>0$ for $0<e<<1$. We also find that
there are functions $p_m(e)$, $q_m(e)$ and $r_m(e)$
such that $p_m(e)r_m(e)>0$ and
\[p_m(e)G_m''(e)=q_m(e) G_m'(e)+r_m(e) G_m(e)\]
on the interval $0<e<1$.
In fact,
\begin{align*}
p_m:=&
e^3{w}^{6}
\big(5 {w} ^{5}{m}^{3}+(-8w^4+30 {w}^{2}){m}^{2}
+(4 {w}^{3 }+24w)m+8\big), \\
q_m:=&
e^2{w}^{4}\big ((25 {w}^{7}-10 {w}^{5 } ){m}^{3}
+ (-32 {w}^{6}+68 {w}^{4}+30 {w}^{2}){m}^{2}\\
&+ (12 {w}^{5}+24 {w}^{3}+48 w )m+24\big ),\\
r_m:=&
e{w}^{3}m\big (5 {w}^{10}{m}^{4}
+(-18 {w}^{9}+60 {w}^{7}){m}^{3}+ (-12 {w}^{8}
-52 {w}^{6} +180 {w}^{4} ){m}^{2}\\
&+ (30 {w}^{7}-76 {w}^{5}
+166 {w}^{3})m-12 {w}^{6}+32 {w}^{4}-12 {w}^{2}+24\big ).
\end{align*}
(To verify that $G_m$ satisfies the second
order differential equation with these coefficients, we compute
the derivatives of $G_m$ and then convert all the expressions to
the variable $w$.) Finally, to show $p_m(e)r_m(e)>0$, it suffices to
show that the inequality holds for $0<w<1$. To do this,
view $p_m$ and $r_m$ as
polynomials in $m$ and note that all their
coefficients are positive functions
of $w$.
\end{pf}
\subsection{The $(1:1)$ Resonance}
The fundamental physical result of this section
is the following proposition:
Among the periodic Keplerian orbits in $(1:1)$
resonance with an incident
gravitational wave, there are generally a (nonzero) finite number of
continuable periodic motions. In fact,
the frequency of the gravitational wave fixes the
semimajor axis, $a$, while the amplitudes and the
phase shift, $\alpha$, $\beta$, $\rho$, of the wave
fix the eccentricity
of the unperturbed Keplerian orbits that are excited
by the perturbation. The inclination of the major axis and
the angular position on the ellipse that complete the set
of initial conditions for a continuable orbit on the excited
ellipse are given by formulas presented below.
However, two facts complicate the
mathematical analysis: there are exceptional choices of the wave
amplitudes $\alpha$ and $\beta$ such that none of the periodic
orbits in $(1:1)$ resonance with the incident
gravitational wave are continuable and there are zeros of the
bifurcation function that are not simple.
The precise mathematical result that we will prove
requires a genericity
assumption. For this we will say that a property of the zero
set of~\eqref{B} is generic relative to the parameter vector
$(\alpha,\beta,\rho)\in {\Bbb{R}}^3$, if it holds for an open and dense subset
of ${\Bbb{R}}^3$.
\begin{prop}\label{11res}
If $m=1$, then, generically relative to the parameters
$(\alpha,\beta,\rho)$, the zero set
of system~\eqref{B}
is a nonempty finite set consisting entirely of simple zeros.
If $m=1$, $\alpha^2+\beta^2\ne 0$, and $\alpha\beta\sin\rho=0$,
then system~\eqref{B}
has a nonzero finite number of zeros which are all simple.
\end{prop}
\begin{pf}
The first generic assumption is ${\cal{E}}{\cal{F}}\ne 0$, the second
generic assumption is that $\alpha^2+\beta^2\ne 0$. (Of course,
if $\alpha^2+\beta^2=0$, then there is no perturbation of the Keplerian
orbits.)
According to Proposition~\ref{zeroprop}, we have $A_1^2-B_1^2\ne 0$ for
$0<e<1$. Thus, if system~\eqref{B} has a solution, then, according
to Proposition~\ref{gbaseprop}, we must have
\[
\big(\frac{\partial A_1}{\partial G}+\frac{\partial B_1}{\partial G}\big)^2
{\cal{F}}^2-
\big(\frac{\partial A_1}{\partial G}-\frac{\partial B_1}{\partial G}\big)^2
{\cal{E}}^2=0.
\]
Define
\[\kappa:=\frac{{\cal{E}}-{\cal{F}}}{{\cal{E}}+{\cal{F}}}\]
and note that, after a simple algebraic manipulation and after
taking into account the
obvious fact that the partial derivatives with respect to $G$
can be replaced with no
loss of generality by partial derivatives with respect to the eccentricity $e$,
the last condition is equivalent to the requirement that
\[
\big(\frac{\partial A_1}{\partial e}-\kappa
\frac{\partial B_1}{\partial e}\big)
\big(\frac{\partial B_1}{\partial e}-\kappa
\frac{\partial A_1}{\partial e}\big) =0.
\]
Also, taking into account the fact that ${\cal{E}}\ge 0$ and ${\cal{F}}\ge 0$,
our assumption that
$\alpha^2+\beta^2\ne 0$ implies $0<|\kappa|\le 1$.
To find the solutions of system~\eqref{B}, suppose for the moment
that the equation
\begin{equation}\label{eqe1}
\frac{\partial A_1}{\partial e}-\kappa
\frac{\partial B_1}{\partial e} =0
\end{equation}
has a solution. For this value of $e$ the third equation of system~\eqref{B}
vanishes provided $\sin(2g+\ell+\tau)$ and $\sin(2g-\ell+\sigma)$ are both
equal to one or both equal to minus one. In either case,
$\cos(2g+\ell+\tau)$ and $\cos(2g-\ell+\sigma)$ both vanish. Thus, for
all integers ${\cal{M}}$ and ${\cal{N}}$ such that
\[
2g+\ell+\tau=\frac{\pi}{2}+2\pi{\cal{M}},\quad 2g-\ell+\sigma=
\frac{\pi}{2}+2\pi {\cal{N}},
\]
or such that
\[
2g+\ell+\tau=-\frac{\pi}{2}+2\pi {\cal{M}},\quad
2g-\ell+\sigma=-\frac{\pi}{2}+2\pi {\cal{N}},
\]
the fixed value of $e$ together with the (nonzero)
finite number of simultaneous solutions
of these last equations with the property that $0\le g,\ell<2\pi$
give a set of solutions of system~\eqref{B}. A similar result
is valid in case
\begin{equation}\label{eqe2}
\frac{\partial B_1}{\partial e}-\kappa
\frac{\partial A_1}{\partial e} =0.
\end{equation}
To determine the simplicity of these solutions, we must compute the
Jacobian of system~\eqref{B} with respect to the variables $(G,\ell,g)$
at the given solution.
This Jacobian is easily computed by expanding along
the first column of the Jacobian matrix. Up to a nonzero constant
multiple, we find the value of the Jacobian to be
\[
(A_1^2-B_1^2)\Big(\big(\frac{\partial^2 A_1}{\partial G^2}
+\frac{\partial^2 B_1}{\partial G^2}\big)
{\cal{F}}\sin(2g+\ell+\tau)+
\big(\frac{\partial^2 A_1}{\partial G^2}
-\frac{\partial^2 B_1}{\partial G^2}\big)
{\cal{E}}\sin(2g-\ell+\sigma)\Big).
\]
In particular, using the fact that
$e$ is a monotone function of $G$,
it follows that the solution $(G,\ell,g)$ of system~\eqref{B}
is simple provided the corresponding solution $e$
of equation~\eqref{eqe1}, respectively~\eqref{eqe2}, is simple.
To finish the proof, we must determine the existence and simplicity
of the solutions of equations~\eqref{eqe1} and~\eqref{eqe2}.
If either $\alpha=0$, $\beta=0$, or $\sin\rho=0$,
then $\kappa=0$ and both
equations~\eqref{eqe1} and~\eqref{eqe2} have unique simple solutions
by Proposition~\ref{zeroprop}. This proves the second statement
of the proposition.
Since the left-hand sides of equations~\eqref{eqe1} and~\eqref{eqe2} are
both analytic functions of $e$, there are at most a finite number
of solutions on the interval $0<e<1$. Moreover, since the map
$(\alpha,\beta,\rho)\mapsto \kappa(\alpha,\beta,\rho)$
is regular on an open and dense subset of ${\Bbb{R}}^3$, if some of the
solutions of one of the equations are not simple, then there is
an arbitrarily small perturbation of the triplet $(\alpha,\beta,\rho)$
such that the perturbed equations have a finite number of simple zeros.
The existence part of the first statement of the proposition
is a consequence of the following facts. If $\kappa\ne 1$ (a generic
assumption), then
equation~\eqref{eqe2} has at least one zero. To see this, we note
that the function
$e\mapsto {\partial B_1}/{\partial e}$ has value $-3$ at $e=0$ and
has limit $\infty$ as $e\to 1^-$ while the function
$e\mapsto {\partial A_1}/{\partial e}$ has value $-3$ at $e=0$ and
has a finite value at $e=1$. As long as $\kappa\ne 1$, then
the function
\[
e\mapsto \frac{\partial B_1}{\partial e}-\kappa
\frac{\partial A_1}{\partial e}
\]
has a negative value at $e=0$ and has limit $\infty$ as $e\to 1^-$.
\end{pf}
The proposition does not give a complete description of the
zero set of system~\eqref{B}. However, since this
description is reduced to an investigation of equations~\eqref{eqe1}
and~\eqref{eqe2} that are algebraic combinations of
Bessel's functions, numerical approximations suggest
the following scenario.
If $\kappa\ne 1$, then the function
\begin{equation}\label{fe1}
e\mapsto \frac{\partial B_1}{\partial e}-\kappa
\frac{\partial A_1}{\partial e}
\end{equation}
has exactly one simple zero on the interval $0<e<1$.
If $\kappa=1$, then~\eqref{fe1}
vanishes at $e=0$ and increases monotonically
to $\infty$ as $e\to 1^-$.
If $\kappa\le 0$,
then the function
\begin{equation}\label{fe2}
e\mapsto \frac{\partial A_1}{\partial e}-\kappa
\frac{\partial B_1}{\partial e}
\end{equation}
has exactly one simple zero on the interval $0<e<1$.
There is a number $\kappa_*\approx 0.036$ such that
if $0<\kappa<\kappa_*$, then~\eqref{fe2}
has exactly two simple zeros, while if $\kappa>\kappa_*$, then~\eqref{fe2}
has no zeros. If $\kappa=\kappa_*$, then~\eqref{fe2} has exactly
one zero which is not simple.
\begin{remark}
In case $\beta=0$, that is the wave is plane polarized in a very
special direction,
it appears that
the zeros of~\eqref{gbase} are all near $e=1$. The smallest
occurs for the $(m:1)=(1:1)$ resonance in case
$\partial A_1/\partial G=0$. Its root is larger than
$e=0.68$. The root is larger for the higher order resonances. This
suggests that only some ``comets'' could remain periodic after perturbation
by a gravitational wave with this particular polarization.
\end{remark}
\subsection{The $(m:1)$ Resonance, $m>1$}
For the $(m:1)$ resonance we will prove that there are
perturbed periodic solutions for the generic $\alpha$, $\beta$ and $\rho$.
This is the content of the next proposition.
\begin{prop}
If $m>1$, then generically, relative to the parameters
$\alpha$, $\beta$ and $\rho$, system~\eqref{B}
has at least one simple zero.
\end{prop}
\begin{pf}
It suffices to consider $\alpha$, $\beta$ and $\rho$ such that
${\cal{E}}{\cal{F}}\ne 0$. Let $g$ and $\ell$
denote a solution of the equations
\[2g+m\ell+\tau=\pi/2,\qquad 2g-m\ell+\sigma=\pi/2,\]
and note that
with this choice of $g$ and $\ell$, system~\eqref{B} has a solution
provided $G$ , equivalently $e$, is chosen such that
\[
\big(\frac{\partial A_m}{\partial G}
+\frac{\partial B_m}{\partial G}\big){\cal{F}}
+\big(\frac{\partial A_m}{\partial G}
-\frac{\partial B_m}{\partial G}\big){\cal{E}}=0.
\]
Equivalently, as in Proposition~\eqref{11res}, there is a solution provided
\begin{equation}\label{m1eq}
\frac{\partial A_m}{\partial G}-\kappa
\frac{\partial B_m}{\partial G}=0.
\end{equation}
Since $|\kappa|<1$, an application of
Proposition~\eqref{zeroprop} shows that~\eqref{m1eq}
has at least one solution.
Moreover, since $A_m^2-B_m^2$ is not the zero function, it has only
a finite number of zeros for $0<e<1$. Also, its zeros do not depend
on the choice of the parameters $\alpha$, $\beta$ and $\rho$.
Thus, if necessary, after a perturbation of the
parameters we can be sure that our solution of~\eqref{m1eq} is not a zero of
$A_m^2-B_m^2$ and that it is a simple zero of the
left hand side of~\eqref{m1eq}. As in Proposition~\eqref{11res}, it follows
that the corresponding choice of $(G,\ell,g)$ is a simple zero of
system~\eqref{B}.
\end{pf}
\section{\label{cpw}Circularly Polarized Waves}
In this section we consider the equations of motion~\eqref{beqn} for the
case of a circularly polarized incident wave. This corresponds to
the special case where, in the components of
the tidal matrix $K$ in~\eqref{bbeqn}, we take
$\alpha=\beta$ and $\rho=\pm \pi/2$. The minus sign
corresponds to a right circularly
polarized wave, while the plus sign corresponds to
a left circularly polarized wave. We note that this is the
main case excluded from the analysis of the previous section. There,
the bifurcation function does not have simple zeros for the bifurcation
problem corresponding to circular polarization.
The equations of motion for the right circularly polarized wave have
the form
\begin{align*}
\frac{d^2 x}{dt^2}+\frac{kx}{r^3}=&-\epsilon\alpha\Omega^2
(x\cos\Omega t+y\sin\Omega t),\\
\frac{d^2 y}{dt^2}+\frac{ky}{r^3}=&\epsilon\alpha\Omega^2
(-x\sin\Omega t+y\cos\Omega t).
\end{align*}
This system can be treated in a similar manner as the analogous system
that arises in
Hill's lunar theory (cf.~\cite{Kov}~\cite{hp}~\cite{Stern}).
The key idea is to view the system in a new Cartesian coordinate
system that rotates relative to the inertial system
with half the frequency of the gravitational wave.
This factor of $1/2$ is due to the fact that the wave has helicity $+2$.
In these rotating coordinates,
that we again call $x$ and $y$,
we obtain the following system
\begin{align}
\notag\frac{d^2 x}{dt^2}-\Omega\frac{dy}{dt}
-(\frac{1}{4}-\epsilon \alpha) \Omega^2 x +\frac{kx}{r^3}&=0,\\
\label{rsys}\frac{d^2 y}{dt^2}+\Omega\frac{dx}{dt}
-(\frac{1}{4}+\epsilon \alpha) \Omega^2 y +\frac{ky}{r^3}&=0.
\end{align}
We note that the replacements $t\to -t$ and $\Omega\to-\Omega$
leave the system invariant. Thus, it suffices to consider the
equations of motion for either state of circular polarization.
\begin{figure}[t]
\caption[]{
Orbits of a Poincar\'e map for the Hamiltonian system with
Hamiltonian~\eqref{hamrp}. The parameters are
$\alpha=1$, $\Omega=1$, $k=1$, $\epsilon=0$ (left hand panel),
$\epsilon=.026$ (right hand panel), and
the energy is $H(p_r,p_\theta,r,\theta)=H(0,1,1,0)$.
For post script figures contact
carmen\@chicone.cs.missouri.edu
\label{fig1}}
\end{figure}
A remarkable fact, also utilized by Hill,
is that system~\eqref{rsys} is equivalent to a
Hamiltonian system with Hamiltonian
\[
H=\frac{1}{2}(X^2+Y^2)+\frac{\Omega}{2}(yX-xY)-\frac{k}{r}+
\frac{\epsilon}{2}\alpha\Omega^2(x^2-y^2),
\]
where $X:=\dot x-\Omega y/2$, and $Y:=\dot y+\Omega x/2$.
This Hamiltonian is given in polar coordinates by
\begin{equation}\label{hamrp}
H=\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}
-\frac{\Omega}{2}p_\theta
+\frac{\epsilon}{2}\alpha\Omega^2r^2 \cos 2\theta,
\end{equation}
where
$p_r=(xX+yY)/r$ and $p_\theta=xY-yX$,
and in Delaunay elements by
\[
H=-\frac{k^2}{2L^2}-\frac{\Omega}{2}G+\frac{\epsilon}{2}\alpha\Omega^2
{\cal{C}}(L,G,\ell,g).
\]
The Delaunay form of the Hamiltonian is expressed in action-angle
variables and is in the correct form to apply the
Kolmogorov-Arnold-Moser~(KAM) theory
(see for example~\cite{ArnoldCM}~\cite{ds3}).
Here, the Hamiltonian is degenerate. But, as pointed out in
Sternberg~\cite[Vol. 2, p. 257]{Stern}, the Hamiltonian system
with Hamiltonian $H^2$ has the same trajectories as the original
system, only the speed along trajectories is changed.
Moreover, the unperturbed part of $H^2$ satisfies the nondegeneracy
assumption for the KAM theorem---its Hessian, with
respect to the actions, has a nonzero determinant. Thus, the
perturbed trajectory remains bounded in time, being
trapped between two-dimensional invariant tori in the three-dimensional
energy surfaces of our {\em two-degree} of freedom
Hamiltonian. Thus, sufficiently weak
circularly polarized gravitational waves do
not ``ionize'' the Keplerian ellipses; that is, the
osculating semimajor axes do not become unbounded. This is illustrated
in Fig.~\ref{fig1}, where ``phase portraits'' for a
typical Poincar\'e map for the Hamiltonian
system corresponding to~\eqref{hamrp} is depicted.
After an energy $H_0$ is fixed, each orbit
on the graph is produced by first choosing an initial
point $(p_r,r)$ in the depicted plane and then by marking the
position of the $(p_r,r)$ coordinates of the Hamiltonian orbit,
with initial condition $(p_r,r)$, $\theta=0$, and with $p_\theta$
the implicit solution of
$H(p_r,p_\theta,r,0)=H_0$, at each time when $\theta(t)$ is
a multiple of $2\pi$ and $\dot\theta(t)\dot\theta(0)>0$. In the actual
computation, $\theta$ is reset to zero each time a crossing is marked.
The figure contrasts the foliation by invariant tori for the
unperturbed system, where there appears to be an
incidental resonance of order two
and one of order three,
with the existence of several invariant tori, as
well as a strongly stochastic layer, for the perturbed Poincar\'e map.
We mention that the existence of periodic solutions of the
equations of motion in
the rotating coordinate system (these correspond to periodic
or quasiperiodic motions of the original system) can be
proved along the lines of Poincar\'e's periodic solutions
of the first and the second kind for the restricted three-body problem.
The continuation theory for the periodic solutions of
the first kind does not depend on the perturbation terms, only
on the Floquet multipliers of the ``circular'' periodic orbits
of the Kepler problem in the rotating coordinate system
(see~\cite{MH}~\cite{Stern}). The unperturbed orbits that continue
are given by $p_\theta=C$, $p_r=0$, $r=C^2/k$ for a fixed constant
$C$.
The continuation theory for the elliptical orbits,
periodic orbits of the second kind, can be completed along the
lines attributed to Poincar\'e and subsequent authors as outlined
in~\cite[Vol 2, p. 274]{Stern}. However, these can also be
found using the continuation theory of \S~\ref{ct}. In the following
brief outline of the procedure, we will use
the Delaunay formulation of the equations of motion.
After isoenergetic reduction,
by implicitly solving for the angular momentum $G$
in the perturbed Hamiltonian as a power series
in $\epsilon$,
and reformulation of the reduced system as a system
of differential
equations with the timelike independent variable $g$, one obtains
a system of two differential equations that are $\pi$-periodic
in $g$. The continuation theory of \S~\ref{ct} can be applied to
this reduced system.
Here, the Poincar\'e section is given by the submanifold
defined by $g=0$, and the return map is an iterate of the strobe after
each $g$ interval of length $\pi$.
An $(m:n)=(m:1)$ resonant unperturbed orbit corresponds to
an invariant one dimensional torus in the Poincar\'e section.
All such tori are normally
nondegenerate due to the fact that the periods of the unperturbed
orbits, in the reduced unperturbed
system, change monotonically with $L$ (this is equivalent to the
twist condition for the Poincar\'e map).
The corresponding bifurcation function maps
the angular variable $\ell$ along the unperturbed orbit
to the average of the first order part of the reduced
differential equation for the action $L$ over the unperturbed
resonant orbit with initial value $\ell$.
In fact, the function is given (up to a nonzero constant
multiple) by $\ell\mapsto (A_m(e)-B_m(e))\sin m\ell$
where $A_m$ and $B_m$ are defined in~\eqref{fcoeff}.
This function has simple zeros (for almost all values of
the eccentricity). Hence, the unperturbed resonant
orbits have continuations.
In particular, our method produces a periodic
orbit of the form $g\mapsto (L(g,\epsilon),\ell(g,\epsilon))$
for the reduced system with independent variable $g$.
Using the fact that $G$ is implicitly given as a function
of the form
\[G:=G(L(g,\epsilon),\ell(g,\epsilon),g,\epsilon),\]
we see that $G$ is also periodic in $g$. Finally, to obtain
$g$ as a function of time, we use the first order differential
equation
\[
\frac{dg}{dt}=-\frac{\Omega}{2}+\epsilon\frac{1}{2}\alpha\Omega^2
\frac{\partial {\cal{C}}}{\partial G}
(L(g,\epsilon),G(g,\epsilon),\ell(g,\epsilon),g).
\]
This last equation, at least for sufficiently small $\epsilon$,
has solutions $g(t)$ such that, for some period $T(\epsilon)>0$,
its solution satisfies
$g(t+T(\epsilon))=g(t)-2\pi$. Since $g$ is an angular variable,
the corresponding function
\[t\mapsto (L(g(t),\epsilon),G(g(t),\epsilon),\ell(g(t),\epsilon),g(t))\]
produces a periodic solution of the original perturbation problem
in the rotating coordinate system. These solutions are analogous
to Poincar\'e's periodic solutions of the second kind.
\section{\label{scn}Speculations, Conjectures and Numerics}
Will a Keplerian binary perturbed by an incident gravitational wave ionize?
To make this question precise, we consider the unperturbed system to be
a Keplerian ellipse, that is, the eccentricity $e$ of the unperturbed
orbit satisfies $0\le e<1$; equivalently, the energy of the
unperturbed system defined
by Hamiltonian~\eqref{KepHam} is negative.
The corresponding perturbed orbit
(the Hamiltonian trajectory given by~\eqref{perHam})
generally does not lie on an ellipse.
However, we define its osculating conic section
at the instant the perturbed motion reaches the state
$(p_r,p_\theta,r,\theta)$
to be the conic obtained as the Keplerian orbit with this initial data, that
is the Keplerian motion that would be obtained if the perturbation were
``turned off'' at this instant of time.
To ionize, the flow of energy between the binary and the wave
must turn unidirectional in a time averaged sense
in the course of the perturbation such that the binary would steadily
absorb energy; in time, the binary system would be permanently disrupted
and the two bodies would eventually be infinitely far apart from each
other.
On the other hand, the basic equation of motion~\eqref{beqn} breaks
down once the semimajor axis of the osculating ellipse becomes
comparable to the (reduced) wavelength of the incident gravitational
wave. To study the route to ionization, we therefore
introduce the notion of dissociation.
We say the Keplerian ellipse determined by the initial data
$(p_r,p_\theta,r,\theta)$ at time $t_0$ {\em dissociates}
under the influence of the perturbation if at some later time the
osculating conic along the perturbed orbit is a hyperbola.
Equivalently, if one wishes to
remove the geometric language of this definition, the requirement for
dissociation may be recast as follows:
the Keplerian energy $H(p_r(t),p_\theta(t),r(t),\theta(t))$, where
$H$ is given by~\eqref{KepHam}, defined along the perturbed orbit
becomes positive in the course of time.
The ionization question probably does not have a simple answer.
However,
two facts are clear. If the strength of the perturbation is sufficiently
small, there are Keplerian binaries that do not ionize. Independent of
the strength of the perturbation, there are Keplerian binaries that do
dissociate. The first fact is proved in this paper: some of the resonant
Keplerian orbits continue to periodic orbits of the perturbed system.
We also recall that in the case of sufficiently weak
circularly polarized incident
gravitational waves, as discussed in~\S~\ref{cpw}, none of the
Keplerian orbits ionize.
On the other hand, to see that dissociation is possible and
to speculate on the fate of
all orbits, we must review the geometry of our problem.
Recall that Hamiltonian~\eqref{perHam} defines a
$2\frac{1}{2}$-degree of freedom Hamiltonian system.
This system is equivalent to the three-degree of freedom
Hamiltonian system given by
\begin{equation}\label{3deg}
{\cal{H}}^*(J,p_r,p_\theta,s,r,\theta)=J+
\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}
+\epsilon r^2 (\phi(s)\cos{2\theta}+\psi(s)\sin{2\theta}),
\end{equation}
where $J$ is a ``fictitious'' action variable conjugate to the ``time''
$s$.
Note here that the phase space of~\eqref{3deg}
is six dimensional and that the
five dimensional submanifold ${\cal{P}}$ given by
\[
H(p_r,p_\theta,\theta,r)=\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}=0
\]
separates the phase space.
This submanifold corresponds to the parabolic
Kepler orbits while a Keplerian binary corresponds to an initial point
in the region of the six dimensional phase space given by
$H(p_r,p_\theta,\theta,r)<0$. Dissociation occurs provided the perturbed
orbit of the Keplerian binary
eventually crosses the manifold ${\cal{P}}$.
To determine that some
perturbed orbits do in fact cross the manifold ${\cal{P}}$,
in both directions, we simply compute the derivative of $H$ along the
perturbed orbit to obtain
\[
\dot H=-2\epsilon\big(
rp_r(\phi(s)\cos 2\theta+\psi(s)\sin 2\theta)+
p_\theta(\psi(s)\cos 2\theta -\phi(s)\sin 2\theta)
\big).
\]
This derivative may be interpreted as a measure of the cosine of the
angle between the
perturbed Hamiltonian velocity field and the normal to the submanifold
${\cal{P}}$. It is apparent that there are open sets on ${\cal{P}}$ where
$\dot H>0$. Thus, there are open subsets (obtained by reversing the
flow on the boundary set)
of the region
$H<0$ such that every point of the subset corresponds to a Keplerian
ellipse that dissociates. However, we emphasize the fact that time, represented
by $s$ in our three-degree of freedom system,
is one of the variables under
consideration when these open sets are determined.
Thus, the initial data
for a Keplerian ellipse that dissociates include an initial time $t=t_0$.
We do not know, from this analysis, how far back in time the reversed orbits
remain in the region where $H<0$.
A more sophisticated analysis might be based on the diffusion
properties of the
orbits of ${\cal{H}}^*$. The geometric picture that is believed to
hold for the dynamics of a nearly integrable Hamiltonian system
with at least
three degrees of freedom is easy to describe informally: there
is a dense set of invariant tori coexisting with
a dense set of orbits some of which are
dense in their respective energy
surfaces (see~\cite{ArnoldCM}~\cite{ds3}).
For the
Hamiltonian~\eqref{3deg}, it is easy to see that each five dimensional
energy surface intersects ${\cal{P}}$. In fact, each energy surface
intersects the subsets of ${\cal{P}}$ defined by $\dot H>0$ and $\dot H<0$.
Thus, in the situation of the
conjectured dynamics, for a sufficiently small perturbation strength,
some of the orbits not on invariant tori ionize while a large subset
of the orbits on invariant tori do not. Of course, some of the invariant
tori of the perturbed system might intersect ${\cal{P}}$; under our definition,
the corresponding orbits will dissociate even though these
same orbits will repeatedly
return to the region where their osculating conics are ellipses.
We note that the usual theory that is used to
prove the existence of invariant
tori for nearly integrable Hamiltonian systems, namely the KAM theory,
is not directly applicable to the Hamiltonian given by~\eqref{3deg} because
the unperturbed system does not meet the required nondegeneracy
conditions. In fact, this Hamiltonian is degenerate
and isoenergetically degenerate (cf.~\cite[p. 408]{ArnoldCM}).
In particular, these facts are evident from the Delaunay action-angle
coordinates for~\eqref{3deg}, where
we see that the three-dimensional unperturbed
invariant tori given by fixing
$J$, $L$, and $G$ do not even have dense trajectories because $\dot g=0$.
To obtain
``nondegenerate'' tori, one must consider
the two dimensional tori given by fixing $J$, $L$, $G$,
and $g$ while leaving $\ell$ and $s$ free. Some of these tori
may survive perturbation and, given their dimensions, it is possible
that some of the perturbed tori are ``whiskered'': they have
stable and unstable manifolds (each with dimension at most three).
The existence of these invariant manifolds---together with
the stable
and unstable manifolds associated with periodic solutions
(one dimensional invariant
tori) and the
intersections among them---is likely responsible
for the diffusion of some of the orbits not in the union of
these invariant sets and the invariant tori.
\begin{figure}[th]
\caption[]{
Projections into $(p_r,r)$ plane of part of one orbit, approximately
5000 iterates in each panel,
of the Poincar\'e map
for the Hamiltonian system with
Hamiltonian~\eqref{perHam}. The parameters are
$\alpha=\beta=2$, $k=1$, $\rho=0$,
and $\epsilon$ in the panels from left
to right is $0.0$, $0.001$, $0.002$ and $0.0025$.
The initial values are
$(p_r,p_\theta,r,\theta)=(.5, 1, 1, 0)$. In this case,
$\Omega$ is chosen ($\Omega\approx 3.897$) so that
the unperturbed Keplerian ellipse has frequency approximately
$1/6$th the frequency of the incident gravitational wave.
The region bounded by the branches of the curve given by $rp_r^2=2k$ shown
in the panels corresponds to elliptical motion.
To obtain the .ps files for this figure or hard copies of the figure,\\ contact
carmen\@chicone.cs.missouri.edu
\label{fig2}}
\end{figure}
We end this section with a short description of some of the numerical
experiments performed on the Hamiltonian system~\eqref{perHam}
given by
\begin{equation*}
{\cal{H}}(p_r,p_\theta,r,\theta)=
\frac{1}{2}(p_r^2+\frac{p_\theta^2}{r^2})-\frac{k}{r}
+\epsilon r^2 (\phi(t)\cos{2\theta}+\psi(t)\sin{2\theta}),
\end{equation*}
where $\phi$ and $\psi$ are given by equation~\eqref{phipsi}.
The results of a typical experiment that suggests the possibility of
dissociation for an elliptical Keplerian orbit
with eccentricity $e=0.5$ are depicted in
Fig.~\ref{fig2}. To obtain the figure,
the above $2\frac{1}{2}$-degree of freedom
Hamiltonian system is integrated numerically, and the values of
the solution are projected into the $(p_r,r)$-plane
after each time interval of length $2\pi/\Omega$. The figure
suggests that dissociation will occur for values of $\epsilon$ that
exceed $\epsilon\approx 0.002$.
To gain some insight into the absorption of gravitational waves by
a Newtonian binary, we have also tested the ``rate of dissociation'', defined
to be inversely proportional to
the number of iterates of the Poincar\'e map required before
the osculating conic of the perturbed ellipse becomes a hyperbola,
by numerical integration of the Hamiltonian system for various
values of the frequency $\Omega$ and
phase shift $\rho$ of the incident wave. We assume here that
$\alpha=\beta$; moreover, the initial elliptical motion is
counterclockwise. Although these experiments are somewhat
difficult to interpret, one fact seems to emerge
regarding the sensitivity of the rate of dissociation to
the polarization of the wave. For fixed $\Omega$,
the maximal dissociation rate is in the vicinity of $\rho=-\pi/2$ while
the minimal dissociation rate is in the vicinity of $\rho=\pi/2$. This rate
also depends on $\Omega$, but in a seemingly unpredictable manner.
\section{Appendix A: ${\cal{C}}$ and ${\cal{S}}$ in terms of Delaunay elements}
The purpose of this appendix is to express ${\cal{C}}=r^2\cos{2\theta}$ and
${\cal{S}}=r^2\sin{2\theta}$ in terms of Delaunay elements. It follows from
the relation $\theta=v+g$ that
\begin{align*}
{\cal{C}}=&r^2\cos(2g+2v)
=r^2\cos{2v}\cos{2g}-r^2\sin{2v}\sin{2g},\\
{\cal{S}}=&r^2\sin(2g+2v)
=r^2\sin{2v}\cos{2g}+r^2\cos{2v}\sin{2g}.
\end{align*}
Moreover,
\[
\cos v =\frac{\cos \widehat{u}-e}{1-e\cos \widehat{u}},\quad
\sin v=\sqrt{1-e^2}\,\frac{\sin \widehat{u}}{1-e\cos \widehat{u}};
\]
these relations follow from~\eqref{anomaly},
and the fact that by definition $v\to \widehat{u}$ as
$e\to 0$ (cf.~\cite[Vol.~1, p.~100]{Stern}).
Therefore,
\begin{align*}
r^2\cos{2v}=&a^2(\frac{3}{2}e^2-2e\cos \widehat{u}+\frac{1}{2}(2-e^2)\cos{2\widehat{u}}),\\
r^2\sin{2v}=&a^2\sqrt{1-e^2}\,(\sin{2\widehat{u}}-2e\sin \widehat{u}).
\end{align*}
There are classical expansions
for $\cos{j\widehat{u}}$ and $\sin{j\widehat{u}}$ in Fourier series whose coefficients are
expressible in terms of the Bessel function $J_\nu$ of order $\nu$. Here,
this Bessel function
is most conveniently defined by
\[
J_\nu(x):=\frac{1}{2\pi}\int_0^{2\pi}\cos(\nu t-x\sin t)\,dt.
\]
Following, for example, J.~Kovalevsky~\cite[p.~49]{Kov},
one finds that
\begin{align*}
\cos \widehat{u}=&-\frac{e}{2}
+\sum_{\nu=1}^\infty\frac{1}{\nu}
[J_{\nu-1}(\nu e)-J_{\nu+1}(\nu e)]\cos{\nu\ell},\\
e\sin \widehat{u}=&2\sum_{\nu=1}^\infty\frac{1}{\nu}J_\nu(\nu e)\sin{ \nu\ell},
\end{align*}
and, for $j>1$,
\begin{align*}
\cos j\widehat{u}=&\sum_{\nu=1}^\infty\frac{j}{\nu}
[J_{\nu-j}(\nu e)-J_{\nu+j}(\nu e)]\cos{\nu \ell},\\
\sin j\widehat{u}=&\sum_{\nu=1}^\infty\frac{j}{\nu}
[J_{\nu-j}(\nu e)+J_{\nu +j}(\nu e)]\sin{\nu \ell}.
\end{align*}
Using these expansions, we obtain the Fourier series
given in~\eqref{CSfs}
where
\begin{align}
\notag A_\nu=&\frac{1}{\nu}
\big((2-e^2)[J_{\nu-2}(\nu e)-J_{\nu+2}(\nu e)]
-2e[J_{\nu-1}(\nu e)-J_{\nu+1}(\nu e)]\big),\\
\label{A1}B_\nu=& \frac{2}{\nu}\sqrt{1-e^2}\,
\big([J_{\nu-2}(\nu e)+J_{\nu+2}(\nu e)]-2J_{\nu}(\nu e)\big).
\end{align}
We note that the functions
${\cal{C}}$ and ${\cal{S}}$ are analytic and
$2\pi$ periodic in the angle variables $\ell$ and $g$. Moreover,
partial derivatives with respect to the Delaunay elements
can be obtained by differentiation
of their Fourier series term by term.
To simplify the expressions for the Fourier coefficients computed
above, we will use
the following elementary identities for
the Bessel functions~\cite[p.~48]{Kov}:
\begin{align}
\notag J_\nu(x)=&\frac{x}{2\nu}[J_{\nu-1}(x)+J_{\nu+1}(x)],\\
\label{A2} J_\nu'(x)=&\frac{1}{2}[J_{\nu-1}(x)-J_{\nu+1}(x)],\\
\label{A3}J_\nu''(x)=&\frac{1}{4}[J_{\nu-2}(x)-2J_\nu(x)+J_{\nu+2}(x)],
\end{align}
and Bessel's equation
\begin{equation}\label{A4}
x^2J_\nu''(x)+xJ_\nu'(x)+(x^2-\nu^2)J_\nu(x)=0.
\end{equation}
The final expressions for $A_\nu$ and $B_\nu$ given in~\eqref{fcoeff}
are obtained from~\eqref{A1} using~\eqref{A2}-~\eqref{A4}; in
fact, the formula for $A_\nu$ is obtained by standard methods using~\eqref{A2},
and the formula for $B_\nu$ is derived from the original expressions~\eqref{A1}
and~\eqref{A3} after noticing that $B_\nu$
is proportional to $J_\nu''(\nu e)$
and using Bessel's equation~\eqref{A4}.
\section{Appendix B: A binary influenced by a distant massive third body}
The purpose of this appendix is to explore the possibility
of applying our results to the three body problem. For a binary system
influenced by a distant massive third body, our ``tidal'' approach
results in a limiting case of the celebrated problem of three bodies and
the question is whether the continuation
theory of \S~\ref{ct} would be applicable
in this case. The existence of periodic orbits in the three-body
problem has been established in the classical work of Poincar\'e~\cite{hp}.
We study the effect of a massive body,
metaphorically the Sun,
on a binary, metaphorically the Earth-Moon system, where the
Sun is viewed as giving rise to a periodic perturbation of the Earth-Moon
orbit by tidal forces.
To derive the equations of motion that will be considered in this appendix,
we
will model the Earth-Moon-Sun system according to the following scenario.
The motion of the Sun is neglected due to its great mass,
its gravitational attraction brings about the motion of the Earth-Moon
system as a whole on an almost Keplerian orbit about the Sun and its tidal
influence perturbs the orbit of the Moon about the Earth. It is this latter
motion that constitutes the lunar problem under investigation here.
To obtain the mathematical model, let us consider the equations of motion of
$m_1$ and $m_2$---the Earth and Moon in our approximation, respectively---as
given by~\eqref{BasicEq1}, with a single perturbing body, namely the Sun, with
potential
\begin{equation}\label{B1}
\Phi(\bold{X}) = - \frac{G_0 M_{\odot}}{|\bold{X} - \bold{X}_{\odot}|}.
\end{equation}
It is useful to transform the equations of motion of $m_1$ and $m_2$ from
${\bold{X}}_1(t)$ and ${\bold{X}}_2(t)$ to the relative coordinates
${\bold{r}} = {\bold{X}}_1-{\bold{X}}_2$ and the center-of-mass coordinates
${\bold{X}}_{\operatorname{cm}} = (m_1{\bold{X}}_1 + m_2{\bold{X}}_2)/(m_1+m_2)$.
In terms of these new coordinates, the
solar gravitational attraction involves the relative coordinates ${\bold{r}}$ and
${\bold{R}} = {\bold{X}}_{\operatorname{cm}}-\bold{X}_\odot$.
Let us further assume that $\mu = r/R \ll 1$, where $r$ and $R$ denote
the magnitudes of the corresponding vectors, and
consider the expansion of the solar influence in these equations in terms of
$\mu$. Using the fact that
\[
\frac{{\bold{R}} + \eta {\bold{r}}}{|{\bold{R}} + \eta {\bold{r}}|^3}
= \frac{{\bold{R}}}{R^3} + \frac{\eta}{R^3}\big({\bold{r}}
-3\frac{{\bold{R}}\cdot{\bold{r}}}{R^2} {\bold{R}}\big) + O(\mu^2),
\]
where $\eta$ is a constant parameter with $|\eta|<1$,
the equations of motion reduce to
\begin{align}\label{B2}
\frac{d^2 \bold{X}_{\operatorname{cm}}}{dt^2} = &
- G_0 M_{\odot} \frac{{\bold{R}}}{R^3} + O(\mu^2), \\
\label{B3}
\frac{d^2 r^i}{dt^2} + \frac{k r^i}{r^3} =& - K_{ij} r^j + O(\mu^2),
\end{align}
where we have introduced the ``tidal matrix" $K_{ij}$ such that
\begin{equation}\label{B4}
K_{ij} = \frac{G_0 M_{\odot}}{R^3}
\big(\delta_{ij} - 3\frac{R^iR^j}{R^2}\big).
\end{equation}
We consider the external body (the Sun) to be so massive
($M_{\odot} >> m_1, m_2$) that it is essentially
unaffected by the presence of $m_1$ and $m_2$
(the Earth and Moon, respectively).
Thus, we can take ${\bold{X}}_{\odot} = {\bold 0}$
and therefore the Sun remains fixed at the
origin of the inertial coordinate system
under consideration.
Neglecting terms of order $r^2/R^2$ in~\eqref{B2}, this equation
reduces to the Newtonian two-body equation for the relative
motion of the center-of-mass about the Sun. We take this orbit to be slightly
elliptic (for example, for the Earth-Moon orbit about the Sun, the eccentricity
$e_1$
is approximately $0.017$).
The resulting expression for
${\bold{R}}$ can be substituted
into \eqref{B3} to give the equations describing
the dynamics of the Earth-Moon system in the presence of the Sun. We further
assume that the relative orbit as well as the center-of-mass motion occurs in
the equatorial plane of the Sun.
This should be a reasonable approximation
as the
Earth-Moon orbital plane makes an angle of approximately $5^\circ$
with the ecliptic
(the ecliptic is essentially the plane of the Earth's orbit around the Sun)
while the
ecliptic makes an angle of approximately $ 7^\circ$ with the
equatorial plane of the
Sun. It is clear that
this ``tidal" approach to the three-body problem is somewhat
different from the standard ``restricted'' approach; in the latter
case, the mass of the Moon is effectively set equal to zero.
The Earth-Moon orbit about the Sun has a small eccentricity; therefore,
the tidal
matrix in \eqref{B3} will be determined to first order in the
eccentricity. To this end, let
$\Omega^2 = G_0 M_{\odot}/a_{\odot}^3$
(with $a_{\odot}$ being the semimajor axis of the Earth-Moon orbit around the
Sun)
and note that the eccentric anomaly is
$\widehat{u} \approx \Omega t + e_1 \sin{\Omega t}$,
the true anomaly is
$v \approx \Omega t + 2 e_1 \sin{\Omega t}$, and
$R \approx a_{\odot}(1 - e_1\cos{\Omega t})$.
Using $R^1=R\cos v$ and $R^2=R\sin v$, the
Cartesian components of the tidal matrix are given by
\begin{align}\label{B5}
K_{11} = & - \Omega^2(\frac{1}{2} + \frac{3}{2} \cos{2\Omega t} -
\frac{3}{2} e_1 \cos{\Omega t} (3 - 7 \cos{2\Omega t})), \notag \\
K_{12} = & - \frac{3}{2}\Omega^2(\sin{2\Omega t}
+ e_1 \sin{\Omega t}(3+7\cos{2\Omega t})), \notag \\
K_{22} = &- \Omega^2(\frac{1}{2} - \frac{3}{2} \cos{2\Omega t}
+ \frac{3}{2} e_1 \cos{\Omega t}(5 - 7\cos{2\Omega t})), \\
K_{13} = & K_{23} = 0.\notag
\end{align}
As the tidal matrix is traceless and symmetric, the above equations determine
all of its elements.
Using \eqref{B3} and \eqref{B5} we can write the
associated Hamiltonian for this system. Because the motion is taken to be
in the equatorial plane, polar coordinates are convenient. In
these coordinates the Hamiltonian is
\begin{align}
{\cal H} = &\frac{1}{2} \left(p_r^2 + \frac{p_{\theta}^2}{r^2}\right) -
\frac{k}{r} - \frac{\Omega^2 r^2}{4} \{ 1 + 3e_1 \cos{\Omega t}
\notag \\
&\quad + 3 \cos{2\theta}[\cos{2\Omega t}
- e_1\cos{\Omega t}(4-7\cos{2\Omega t})] \notag\\
&\quad + 3 \sin{2\theta}[\sin{2\Omega t}
+ e_1 \sin{\Omega t}(3 +7\cos{2\Omega t})]\}. \label{B6}
\end{align}
Upon expressing the Hamiltonian equations in terms of
intrinsic dimensionless quantities, it becomes clear that the
strength of the interaction between the binary and the third body
is $\Omega^2/\omega^2\ll 1$, but the square root of this perturbation
parameter also occurs in the harmonic terms that render
the Hamiltonian~\eqref{B6}
explicitly time-dependent. In particular, the period of the harmonic
terms becomes unbounded as the perturbation parameter goes to zero.
Therefore, the continuation method of \S~\ref{ct} is
not directly applicable in this case;
in fact, the resolution of this problem is
due to Hill (cf.~\cite{Kov}~\cite{hp}). In Hill's approach, the equation
of relative motion~\eqref{B3} is referred to a Cartesian system of
coordinates $\bold{r}'$ that rotates with frequency $\Omega$ with respect
to the inertial system. Let $r^i=S_{ij}{r'}^j$, where the nonzero
elements of the orthogonal matrix $S$ are given by
\[
S_{11}=S_{22}=\cos\Omega t, \quad -S_{12}=S_{21}=\sin\Omega t,\quad S_{33}=1;
\]
then, the equations of motion in the new system are $(r'=r)$
\begin{align}
\frac{d^2x'}{dt^2}-2\Omega\frac{dy'}{dt}-\Omega^2 x'+\frac{kx'}{{r'}^3}=&
-(K_{11}'x'+ K_{12}'y'),\notag\\
\frac{d^2y'}{dt^2}+2\Omega\frac{dx'}{dt}-\Omega^2 y'+\frac{ky'}{{r'}^3}=&
-(K_{12}'x'+ K_{22}'y'),\label{B7}
\end{align}
where $K'=S^TKS$, i.e.,
\begin{equation}
\label{B8}
K_{11}'=-2\Omega^2(1+3 e_1\cos\Omega t),\quad
K_{12}'=-6\Omega^2 e_1\sin\Omega t,\quad
K_{22}'=-\frac{1}{2}K_{11}',
\end{equation}
and $K_{13}'=K_{23}'=0$. The system~\eqref{B7} is
autonomous for $e_1=0$. In this case, periodic solutions exist as
originally demonstrated by Hill and Poincar\'e (cf.~\cite{hp}).
The continuation of such solutions using $e_1$, $0\le e_1\ll 1$, as
the expansion parameter can be proved, using the
Implicit Function Theorem, as originally conceived by
Poincar\'e (cf.~\cite{MH}~\cite{Kov}~\cite{hp}) for the restricted
three-body problem.
Of course, the method of \S~\ref{ct} is also applicable by following
the ideas for isoenergetic reduction as discussed in~\S~\ref{cpw}.
|
\section{Introduction}
In a recent work~\cite{jamal} we computed the Green's function in light cone
gauge $A^+=0$ for the small fluctuations about a background
Weizs\"acker--Williams gluon field. This background field is generated by the
valence quarks in a large $A$ nucleus. For small x partons, these valence
quarks constitute a static and well localized source of color
fields~\cite{larry1}. The average color charge squared per unit area of the
valence quarks is denoted by $\mu^2$ and it is of the order of $A^{1/3}$
fm$^{-2}$. The quantity
$\mu$ is the only dimensionful parameter in the theory and as a
result, the coupling constant will run as a function of it.
Previously, two of us~(McL.--V.) have argued that we can compute the gluon
distribution function from the light cone gauge Green's function~\cite{larry3}.
In the present paper we use that Green's function to compute the corrections
induced by quantum fluctuations on the Weizs\"acker--Williams distribution
function.
There are several reasons for this computation to be of interest. From the
practical point of view, we hope that an analysis which provides us with a
better understanding of the initial conditions for the evolution of partons
in the collision of heavy nuclei to form a quark--gluon plasma will help
establish a firm foundation~\cite{heribert} for partonic cascade models
simulating such collisions~\cite{klaus}. For an alternative approach to the
problem of initial conditions in heavy ion collisions, see
reference~\cite{sasha}.
{}From the theoretical point of view, we hope to understand the small x
behavior
of a nucleus starting from a QCD based approach. Let us recall that it is
believed that in the small x region, the gluon distribution function computed
perturbatively including leading logarithmic contributions in $\ln(1/x)$ for a
single nucleon behaves like~\cite{smallx}
\begin{eqnarray}
\frac{dN}{dx} \sim \frac{1}{x^{1+C\alpha_{s}}} \nonumber.
\end{eqnarray}
The steep rise of the gluon distribution function for small x is sometimes
referred to as the Lipatov enhancement. It is obtained by solving the
Balitsky--Fadin--Kuraev--Lipatov (BFKL) kernel for the t--channel exchange of a
perturbative pomeron~\cite{delduca}. This behavior is also exhibited in a
hadron where the large x part of the hadron wavefunction is taken to be a heavy
quark--antiquark state. By applying Hamiltonian perturbation theory to this
state, it is possible to reproduce the kernel of the BFKL
equation~\cite{mueller} for the emission of a large number of soft gluons.
However, the rapid rise of the gluon distribution function with smaller values
of x is in conflict with unitarity when considering the hadron scattering total
cross section at asymptotically high energies~\cite{smallx}. Physically, this
violation of unitarity can be understood to result from ignoring effects that
arise due to the large density of partons at very small values of x~\cite{glr}.
When the density of partons is so large that neighboring partons overlap, the
t--channel picture of an independent parton cascade in x breaks down. The
former signals that at very small values of x, the picture in which the partons
do not interact with each other has to be modified in order to comply with the
Froissart bound~\cite{froissart} on the growth of cross sections at
asymptotically high energies. Although some work has been done in recent years
to include these ``higher twist'' effects in describing parton evolution at
high densities~\cite{glr,muellqiu,laenena}, more still remains to be done in
devising a quantitative mechanism to limit this growth.
The regime of high parton density is the screening regime. This screening is
presumably responsible for the shadowing phenomena observed in deep inelastic
scattering experiments off nuclei at small x. It can be addressed as a
collective or many body effect. It is precisely this many body problem of
parton interactions that we seek to address in our work. As outlined in
references~\cite{larry1}~\cite{larry3}~\cite{larry2}, our formalism provides
us, by means of a novel weak coupling approach, with a technique to solve the
many body problem of wee parton distributions in a large nucleus.
In this work, we will focus on one of the theoretical aspects of the problem
--the nature of the small x terms in the gluon distribution function of the
Weizs\"{a}cker--Williams fields generated by the valence quarks in the infinite
momentum frame. We do this by computing a formula for the gluon distribution
function which includes the effects of all orders in the parameter
$\alpha_s\mu$. Working in the weak coupling regime $\alpha_s\mu \ll k_t$, we
extract from this formula an expression for the distribution function in
perturbation theory to second order in $\alpha_s$. Although we approach the
problem in weak coupling, we will show that after the corrections are
considered, the series expansion parameter becomes $\alpha_s \ln(1/x)$ which
may be large for small x's. We will argue that this forces us to devise a
method to isolate and sum up the leading contributions to all orders in that
effective expansion parameter in order to compute the modification to the
zeroth order $1/x$ distribution function.
The outline of this work is as follows: In section 2, we briefly review the
basic aspects of the model in which we treat the nuclear valence quarks as
static sources of color charge as seen by small x partons. We also go briefly
through the formalism that allows us to compute the Green's function for the
small fluctuations equation and finish the section by writing the formula for
this Green's function. In section 3, we use this Green's function and exploit
its relation to the gluon density in order to compute the gluon distribution of
small x gluons. We also derive from this result a formula for the leading small
x terms of the distribution function in perturbation theory valid to second
order in $\alpha_s$. In section 4, we compute the corrections to the background
field induced by the fluctuation field and show that the only effect that this
introduces can be absorbed into the renormalization of the background field
(which is related to the renormalization of the coupling constant to one loop).
Details of the calculations in section 3 and section 4 are discussed
in appendix I and appendix II respectively. Finally we summarize our results in
section 5.
\section{The model}
In QCD, a hadron is a cloud of virtual particles with a rather complicated
structure. The picture gets simpler when we consider the hadron as a quantum
system composed of quasi real particles (partons) with lifetimes much larger
than the characteristic interaction times. This can be done in a reference
frame where the hadron has a large momentum~\cite{glr}. Partons with large
lifetimes can produce new partons carrying smaller fractions x of the initial
hadron's momentum. The small x partons will therefore densely populate the
hadron and see the rest of it with its longitudinal dimension Lorentz
contracted to a thin disk. In our model, we look at the small x partons in a
large $A$ nucleus ($x\ll A^{-1/3}$) where the high parton density allows us to
use weak coupling techniques. The rest of the nucleus consists of the valence
quarks which carry most of the nuclear momentum. They are described as a static
(recoilless) source of color charges, in a reference frame in which they move
with the speed of light (infinite momentum frame)~\cite{larry1}.
The problem is well suited to be described using light cone
variables~\cite{lightcone}
\begin{eqnarray}
y^{0}, y^{3} \longrightarrow y^{\pm} = (y^{0} \pm y^{3})/\sqrt{2}
\label{eq:lightconevar}
\end{eqnarray}
In order to compute ground state properties of the wee partons, we define a
partition function for the system. This partition function includes the sum
over a large number of color configurations. To simplify the problem, we
resolve the transverse space direction into cells which contain a large number
of valence quarks, or equivalently, a large number of color charges. This
allows us to treat the sum over color configurations classically~\cite{larry1}.
To write the average over the color charges, we introduce a Gaussian weight by
inserting into the path integral representation of the partition function the
term
\begin{eqnarray}
\exp \left\{ - \frac{1}{2\mu^2} \int d^{2}x_{t}~ \rho^{2}(x) \right\} \, ,
\label{eq:coloraverage}
\end{eqnarray}
where $\rho$ is the color charge density (per unit area) and the parameter
$\mu^2$ is the average color charge density squared (per unit area) in units of
the coupling constant $g$. The introduction of the partition function, where we
average over the sources of external charge, allows us to formulate the theory
as a many body problem with modified propagators and vertices.
We treat the system perturbatively and the first step is to solve the classical
equations of motion
\begin{eqnarray}
D_{\mu}F^{\mu\nu} = g J^{\nu} , \; \;
J^{\nu} = \delta^{+\nu} \rho(x^{+},x_{t}) \delta (x^-) \, ,
\end{eqnarray}
for which (working in the light cone gauge $A_{-}=-A^{+}=0$) there exists a
solution with $A_{+}=-A^{-}=0$. We require
\begin{eqnarray}
A_{i}(x)=\theta(x^{-})\alpha_{i}(x_{t}) \, ,
\label{eq:trans}
\end{eqnarray}
(hereafter, latin indices refer to transverse variables) and furthermore
$F^{ij}=0$. The latter condition implies that $\alpha(x_{t})$ is a pure gauge
transform of the vacuum~\cite{larry2},
\begin{eqnarray}
\tau \cdot \alpha_{i} =\frac{i}{g} U(x_t)
\nabla_{i} U^{\dagger}(x_t), \label{eq:alphau}
\end{eqnarray}
where $U(x_t)$ is a $SU(3)$ local gauge transformation whose
spatial dependence is only on the two--dimensional transverse space.
It is subject to the physical gauge condition
\begin{eqnarray}
\nabla \cdot \left(U(x_t) \nabla U^{\dagger}(x_t)\right) =
-ig^2 \rho \label{eq:ufield} \, .
\end{eqnarray}
The $x^+$ (light cone time) dependence of the charge density is a consequence
of the extended current conservation law
\begin{eqnarray}
D_{\mu}J^{\mu}=0.
\label{eq:gaugeinv}
\end{eqnarray}
The integration over the sources $\rho$ in equation~(\ref{eq:coloraverage})
may be written as
\begin{eqnarray}
\int [dU] \exp \left( -\frac{1}{\mu^{2}g^{4}}{\bf Tr}
\left( \nabla_i \cdot U \frac{1}{i} \nabla_i U^{\dagger}
\right)^{2} \right),\label{eq:CA}
\end{eqnarray}
where we have ignored the Faddeev--Popov determinant. Note that the effective
coupling constant for this theory is $g^{2}\mu$ so that the expansion parameter
becomes $\alpha_s \mu / p_{t}$.
The Green's function can be computed from the relation
\begin{eqnarray}
G^{\alpha\beta}_{ij}(x,x') = \int\frac{d\lambda}{\lambda - i \epsilon}
\int\frac{d^{4}p}{(2\pi)^{4}}\delta(\lambda-p^{2})
(A^{\alpha}_{i})_{\lambda}(x)(A^{\beta}_{j})^{\dagger}_{\lambda}(x')
\label{eq:green}
\end{eqnarray}
and the gluon distribution function can be computed from the Green's function
by the relation
\begin{eqnarray}
\frac{dN}{d^{3}k} = i \frac{2k^{+}}{(2\pi)^{3}}
\sum_{\alpha , i} G^{\alpha \alpha}_{i i}
(x^{+}, \vec{k} ;x^{+}, \vec{k})\label{eq:distfunc} \, .
\end{eqnarray}
To relate the Green's functions to the distribution function by the above
relation, the former must be averaged over the external sources of color
charge.
Indeed, the distribution function, to all orders, is related by the above
expression to the fully connected two point Green's function. This Green's
function is given by the relation
\begin{eqnarray}
\langle \langle AA\rangle\rangle_{\rho}=\langle \langle A_{cl}\rangle
\langle A_{cl}\rangle + \langle A_{q}A_{q}\rangle \rangle_{\rho} \, .
\label{eq:schdy1}
\end{eqnarray}
In the above, $\langle A_{cl}\rangle$ is the expectation value of the classical
field to all orders in $\hbar$. It can be expanded as
\begin{eqnarray}
\langle A_{cl}\rangle = A_{cl}^{(0)}+A_{cl}^{(1)}+...... \, ,
\label{eq:schdy2}
\end{eqnarray}
where $A_{cl}^{(0)}$ is the solution discussed in Eqs.~(4)--(5).
The one loop correction to the classical field, $A_{cl}^{(1)}$, is computed in
section 4 of this paper. The term $\langle A_{q}A_{q}\rangle$ above is the
small fluctuations Green's function computed to each order in the classical
field. The symbol $<\cdots>_{\rho}$ indicates that we have to average over the
external sources of color charge with the Gaussian weight described
above.
{}From the above, it is clear that the zeroth order contribution is $\langle
A_{cl}^{(0)} A_{cl}^{(0)}\rangle_{\rho}$. This contribution is the QCD analog
of the well known Weizs\"acker--Williams distribution in classical
electrodynamics. The general form of the solution is given in
reference~\cite{larry2}. In the range of momenta $\alpha_{s}\mu\ll k_{t} \ll
\mu$, the zeroth order solution $A^{\mu}$ yields a distribution function that,
written in terms of $x\equiv k^{+}/P^{+}$, with $P^{+}$ the nuclear
longitudinal light cone momentum, looks like
\begin{eqnarray}
\frac{1}{\pi R^{2}} \frac{dN}{dxd^{2}k_{t}} =
\frac{\alpha_{s} \mu^{2} (N_{c}^{2} - 1)}{\pi^{2}}
\frac{1}{xk^{2}_{t}} \label{eq:WWdist} \, .
\end{eqnarray}
The above is the well known Weizs\"acker--Williams distribution scaled by
$\mu^2\approx A^{1/3}$ fm$^{-2}$.
With this formalism at hand, we can proceed to compute the next order
contribution to the gluon distribution function. Our strategy is to compute the
small fluctuations correction to the classical equation of motion. Writing the
field in terms of its background and fluctuation parts
\begin{eqnarray}
A^{\mu}(x)=B^{\mu}(x) + b^{\mu}(x),
\end{eqnarray}
we are able to express the equation obeyed by $b^{\mu}$ as~\cite{jamal}
\begin{eqnarray}
\left( D(B)^2 g^{\mu \nu} - D^\mu (B) D^\nu (B) \right)
b_{\nu} - 2 F^{\mu \nu} b_{\nu} = 0.
\label{eq:SM}
\end{eqnarray}
where $B$ is the background field which according to equation~(\ref{eq:trans})
is non--vanishing only for its transverse components. $D^{\mu}(B)$ is the
covariant derivative with $B$ as the gauge field (notice that $D^{\pm} =
\partial^{\pm}$). As discussed in reference~\cite{jamal}, the set of
equations~(\ref{eq:SM}) can be unambiguously solved in the gauge $A^-=0$, and
by means of equation~(\ref{eq:green}) we can compute the Green's function for
the fluctuation fields in this gauge. To obtain the Green's function in the
gauge $A^+=0$ (light cone gauge), we perform a gauge transformation on the
Green's function in the $A^-=0$ gauge and obtain finally (in the matrix
representation)
\begin{eqnarray}
G_{ij}^{\alpha\beta ;\alpha^\prime \beta^\prime}(x,y)
\!\!\!\!\!& = &\!\!\!\! - \! \int \!\!{{d^4p}\over {(2\pi)^4}}
{{e^{ip(x-y)}}\over {p^2-i\epsilon}}
\Bigg\{\! \bigg[\delta_{ij}\!+\!{p_ip_j\over {p^-p^+}}(2e^{ip^+(x^-\! - y^-)}
\!-\! e^{-ip^+y^-}\!-\! e^{ip^+x^-})\bigg]\nonumber \\
&\times & \bigg[\theta(-x^{-})\theta(-y^{-})\tau_{a}^{\alpha\beta}
\tau_{a}^{\alpha^\prime \beta^\prime} + \theta(x^-)\theta(y^-)\,
F_{a}^{\dagger\alpha\beta}(x_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}(y_{t})
\bigg]\nonumber \\
\!\!\!\!\!&+ &\!\!\!\!
\theta(-x^-) \theta(y^-)
\int \!{{d^2 q_t} \over {(2\pi)^2}} d^2 z_t \,\, e^{i(q^{+}-p^{+})y^{-}}
e^{i(p_t-q_t)(y_t-z_t)} \nonumber \\
\!\!\!\!\!&\times &\!\!\!\!
F_{a}^{\dagger\alpha \beta}(z_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}\!(y_{t})\bigg[\delta_{ij} +
\frac{p_{i}p_{j}}{p^-p^{+}}
(e^{ip^{+}x^{-}}\!-1)\nonumber \\
\!\!\!\!\!& + &\!\!\!\!\!\! \frac{q_{i}q_{j}}{p^-q^{+}}
(e^{-iq^{+}y^{-}}\!-1)
+ \frac{p_{i}q_{j}p_{t}\cdot q_{t}}
{(p^{-}p^{+})(p^{-}q^{+})}(e^{ip^{+}x^{-}}\!-1)
(e^{-iq^{+}y^{-}}\!-1)\bigg]\nonumber \\
\!\!\! & + &\!\!\!\!\!
\theta(x^-) \theta(-y^-)
\int\! {{d^2 q_t} \over {(2\pi)^2}} d^2 z_t \,\,
e^{i(p_t-q_t)(z_t - x_t)} e^{-i(q^{+}-p^{+})x^{-}}\nonumber \\
\!\!\!\!\!&\times &\!\!\!\! F_{a}^{\dagger\alpha \beta}(x_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}\!(z_{t})\bigg[\delta_{ij} +
\frac{p_{i}p_{j}}{p^{-}p^{+}}(e^{-ip^{+}y^{-}}\!\!-1)
+ \frac{q_{i}q_{j}}{p^{-}q^{+}}(e^{iq^{+}x^{-}}\!-1)\nonumber \\
\!\!\!\!\!& + &\!\!\!\!\!\!
\frac{q_{i}p_{j}p_{t}\cdot q_{t}}{(p^{-}p^{+})(p^{-}q^{+})}
(e^{-ip^{+}y^{-}}\!-1)
(e^{iq^{+}x^{-}}\!-1)\bigg]\Bigg\} \, ,
\label{eq:GF}
\end{eqnarray}
where $q^{+}=p^{+}+ \frac{q_{t}^{2}-p_{t}^{2}}{2p^{-}}$ and
\begin{eqnarray}
F_{a}^{\dagger\alpha \beta}(x_{t}) =\bigg( U(x_{t})\tau_{a}U^{\dagger}
(x_{t})\bigg)^{\alpha\beta}.
\end{eqnarray}
The above result was derived in reference~\cite{jamal}. The regularization of
the poles in $p^-$ and $p^+$ is a rather subtle issue and was discussed at some
length in the above mentioned paper.
The Green's function $G_{ij}(x,y)$ contains the physical degrees of freedom
that we need to relate to the gluon density. In the following section, the
above small fluctuations propagator will be used to compute corrections to the
Weizs\"acker--Williams distribution function. We will show that logarithmic
corrections, both in x and $k_t$, arise from here. In section~4, we discuss the
corrections that the small fluctuations induce on the classical field to one
loop. These corrections, in contrast, provide no such logarithmic terms to the
distribution function and their only effect is to renormalize the coupling
constant and the background field.
In appendix I, we compute the Fourier transform of the Green's function.
This result is useful for the computation of the distribution functions
performed in the next two sections.
\section{The gluon distribution function}
In this section, we will show how one can obtain the gluon distribution
function from the Green's function~(equation~\ref{eq:GF}), corresponding to the
small fluctuations about the Weizs\"{a}cker--Williams background color field.
We will be concerned with the structure of the leading terms for small x and at
the end of the section compute a formula for the gluon distribution function
valid to $\alpha_{s}^{2}$. To compute the distribution function we need to sum
over all possible color configurations. This color average involves the two
dimensional gauge fields $U$. We start by recalling the properties of these
gauge fields under our color averaging.
\subsection{Correlation functions involving the gauge transformations $U$}
According to equation~(\ref{eq:alphau}) the gauge transformations $U(x_t)$
carry the information on the background field which enters the Green's
function~(\ref{eq:GF}). These gauge transformations have the interesting
property that the color average with the Gaussian weight (defined by
equation~(\ref{eq:CA})), of the combination
\begin{eqnarray}
U^{\dagger}(x_t)\tau^{a}U(x_t)U^{\dagger}(y_t)\tau^{a}
U(y_t)
\end{eqnarray}
can be written as~\cite{larry3}
\begin{eqnarray}
<\! Tr U^{\dagger}(x_t)\tau^{a}U(x_t)U^{\dagger}(y_t)\tau^{a}U(y_t)\!>=
\!\frac{(N_c^2 -1)}{2}\!\Gamma (x_{t}\!-\!y_{t}) \label{eq:correl}
\end{eqnarray}
where $N_c$ is the number of colors and summation over repeated indices is
implied.
As pointed out in reference~\cite{larry1}, the average over the color sources
yields the information about the ground state properties of the system. The
average with the Gaussian weight is an artifact that simplifies the computation
and we expect that as long as we resolve the nucleus on a transverse size
much larger than the typical transverse quark separation, such an artifact
is justified.
The function $\Gamma$ factorizes the dependence on the transverse
coordinates $x_t$,~$y_t$ and is a function of their
difference. Moreover, from equation~(\ref{eq:correl}) we see that
$\Gamma$ is real and also
\begin{eqnarray}
\Gamma(0)=1.
\label{eq:gammaofzero}
\end{eqnarray}
Defining the Fourier transform of $\Gamma(x_t)$
\begin{eqnarray}
\gamma(p_t) = \int d^{2}x_t e^{-ip_t x_t}\Gamma(x_t)
\label{eq:Foutransgamma}
\end{eqnarray}
we have, together with~(\ref{eq:gammaofzero}), the sum rule
\begin{eqnarray}
\int \frac{d^2 p_t}{(2\pi)^2} \gamma(p_t) = 1.
\label{eq:sumrule}
\end{eqnarray}
The color charge at a given transverse location will be zero on average and the
only way to generate a non--zero color charge will be by fluctuations.
Equation~(\ref{eq:CA}) can be thought of as the generator of those
fluctuations and thus the function $\Gamma(x_t, y_t)$ represents the
correlator of fluctuating fields at the transverse locations $x_t$ and
$y_t$.
In momentum space, the function $\gamma(p_t)$ can be formally computed by
expanding the exponential in~(\ref{eq:CA}) in powers of the coupling parameter
$\alpha_{s}\mu / p_t$ (weak coupling regime). This was done in
reference~\cite{larry3} for scalars. For gluons, the result for $\alpha_s\mu
\ll p_t$ is
\begin{eqnarray}
\gamma(p_t) = (4\pi)^2\frac{\alpha_{s}^{2}\mu^2}{p_{t}^{4}}N_c\,.
\label{eq:gammaweak}
\end{eqnarray}
We notice that the expansion is only necessary in order to analytically compute
expressions of the form
\begin{eqnarray}
\int d^{2}p_t f(p_t)\gamma(p_t) \, ,
\label{eq:example}
\end{eqnarray}
with $f(p_t)$ a non--trivial function of $p_t$. However, in principle, we can
perform a numerical analysis to take into account the many possible different
configurations of the external field contributing to expressions such
as~(\ref{eq:example}). This is equivalent to considering the effect to all
orders in $\alpha_{s}\mu / p_t$ of the different configurations of the
background field. For a quantitative discussion about the properties of the
distribution function, we will restrict ourselves to the weak coupling regime
for which $\gamma(p_t)$ is given by equation~(\ref{eq:gammaweak}).
\subsection{The distribution function}
With the above remarks in mind, we proceed to the computation of the gluon
distribution function. We use the formula for the distribution function
\begin{eqnarray}
\frac{dN}{d^{3}k} = i \frac{2k^{+}}{(2\pi)^{3}}
\lim_{k^+\rightarrow {k'}^+}\int \frac{dk^-}{2\pi}
\int \frac{d{k'}^-}{2\pi} <D^{aa}_{ii}(k,k')\!\!> \, ,
\end{eqnarray}
where $<D^{aa}_{ii}(k,k')\!\!>$ is the small fluctuations propagator in
momentum space, traced over the color and Lorentz indices and averaged over the
external sources of color charge. This formula follows from computing
$<a^\dagger(p) a(p)>$ as an expectation value for the gluon Fock space
distribution function in the ground state generated by the external valence
charges~\cite{larry3}.
In appendix I, we derive explicitly an expression for
$<D^{aa}_{ii}(k,k')\!\!>$. Using this result, we obtain the following integral
expression for the distribution function
\begin{eqnarray}
{1 \over {\pi R^2}} {{dN} \over {d^3k}}
& = &
{{2ik^+} \over {(2\pi)^3}}(N_c^2-1)
\lim_{k^+ \rightarrow k^{\prime +}}
\int
{{dp^+d^2p_tdk^-} \over {(2\pi)^4}}
\nonumber \\
& &
\Bigg\{
{1 \over {p_t^2-2p^+k^--i\epsilon}}
\Bigg(
2 +
{p_t^2 \over {k^-k^+k^{\prime +}}}
(2p^+-k^+-k^{\prime +})
\Bigg)
\nonumber \\
&\times&
\Bigg(
-(2\pi)^2 \delta^{(2)} (p_t-k_t)
{1 \over {p^+-k^++i\epsilon}}
{1 \over {p^+-k^{\prime +} - i\epsilon}}
\nonumber \\
&-&
\gamma(p_t-k_t)
{1 \over {p^+-k^+-i\epsilon}}
{1 \over {p^+-k^{\prime +} + i\epsilon}}
\Bigg)
\nonumber \\
&+&
\gamma(p_t-k_t)
{1 \over {k_t^2-2p^+k^- - i\epsilon}}
\Bigg(
2 -
{{k_t^2+p_t^2} \over {k^-k^+}} +
{{(p_t \cdot k_t)^2} \over {(k^{-2}k^{+2})}}
\Bigg)
\nonumber \\
&\times&
\Bigg(
{1 \over {p^+-k^+ - i\epsilon}}
{1 \over {q^+-k^+ - i\epsilon}}+
{1 \over {p^+-k^+ + i\epsilon}}
{1 \over {q^+-k^+ + i\epsilon}}
\Bigg)
\Bigg\}
\end{eqnarray}
In the last term of this equation, we have taken the limit that $k^+
\rightarrow k^{\prime +}$ since this term has no singularity in that limit.
We now do the integral over $k^-$. We assume that
$k^+ > 0$. When we do the integral, two classes
of terms result. The first set of terms arise from the explicit
$k^-$ dependence in the above equation and are non-zero. The second
set of terms arise from the $q^+$ in the last terms of the above equation.
These terms result in an unrestricted integral over $p^+$.
One can show that all the singularities of the resulting integrand
are on the same side of the $p^+$ integration contour in the complex
$p^+$ plane. They therefore integrate to zero. (There is a possible
ambiguity in the closing of contours associated with the contour at
infinity, but this term does not have any contribution proportional to
$\ln(1/x)$.) Therefore, we only get the contribution from the first term,
which is only nonzero for $p^+ < 0$,
\begin{eqnarray}
{1 \over {\pi R^2}} {{dN} \over {d^3k}} & = &
(N_c^2-1){{4k^+} \over {(2\pi)^3}}
\int^0_{-\infty} {{dp^+d^2p_t} \over {(2\pi)^3}}
\nonumber \\
&\times&
\Bigg\{ {1 \over {p^+(p^+-k^+)(k^++p^+p_t^2/k_t^2)}}
\left[\gamma(p_t-k_t) - (2\pi)^2\delta^{(2)} (p_t-k_t)\right]
\nonumber \\
&\times&
\Bigg[ 1 - {p^+ \over k^+} \left( 1 + {p_t^2 \over k_t^2} \right)
+2 \left( {p^+ \over k^+} \right)^2 { (p_t \cdot k_t)^2 \over k_t^4}
\Bigg] \Bigg\}\, .
\end{eqnarray}
Now in this expression, we shall only be concerned with those terms which are
proportional to $\ln(1/x)$. The terms not proportional to $\ln(1/x)$ are
non-leading for small x. Moreover, we have found that within our approach,
these terms are inherently ambiguous. This is due to the fact that the
$\ln(1/x)$ terms can only arise by regulating the singularity in the above
integral as $p^+ \rightarrow \infty$. We do this by making the upper limit of
integration, to be of the same
order than the total momentum of a typical nucleon in the nucleus. Of course
different regularization schemes will affect the non--leading terms in
different
ways. Presumably, the detailed longitudinal structure of the valence quark
charge distribution must be known before these terms may be evaluated.
After some straightforward algebra, we find
\begin{eqnarray}
\Bigg(\frac{1}{\pi R^{2}}\frac{dN}{dxd^{2}k_{t}}\Bigg)_q\!\!\!
&=&\!\!\! \frac{8(N^{2}_{c}-1)}{(2\pi )^{4}}
\frac{1}{x}\int {{d^2 p_t }\over {(2\pi )^2}}
\gamma (p_t -k_t ) \bigg[ 1 -
\frac{(p_t \cdot k_t )^2}{p_t^2 k_t^2}\bigg]\ln \left(\frac{1}{x}\right)
\label{eq:MAINRES}
\end{eqnarray}
where the subindex $q$ in the left hand side of the above equation refers to
the correction to the distribution function from the small quantum fluctuation
field. Equation~(\ref{eq:MAINRES}) is our main result. It is normalized so that
the vacuum density is zero. This can be checked by setting $U=1$ in the above
equation.
The terms above can be written in the form
\begin{eqnarray}
\Bigg(\frac{1}{\pi R^{2}}\frac{dN}{dxd^{2}k_{t}}\Bigg)_{q}\!\!=
\left(\frac{N_c^2-1}{2\pi^4}\right)\frac{1}{x}\int \frac{d^2p_t}{(2\pi)^2}
\gamma(p_t-k_t)
\Bigg\{ \frac{p_{t}^{2}k_{t}^{2} - (p_t \cdot k_t)^2}
{p_{t}^{2}k_{t}^{2}}
\Bigg\} \ln \left(\frac{1}{x}\right) .
\label{eq:secandthird}
\end{eqnarray}
Now, shift the variable of integration $p_t \rightarrow p_t+k_t$ and expand
$\gamma(p_t)$ in weak coupling. This restricts the lower limit of integration
for the radial component of $\vec{p_t}$ to be $\alpha_s\mu$, which comes from
the weak coupling expansion of $\gamma$. Thus the above expression becomes
\begin{eqnarray}
\Bigg(\frac{1}{\pi R^{2}}\frac{dN}{dxd^{2}k_{t}}\Bigg)_{q}&=&
8N_c(N_c^2-1)\frac{\alpha_{s}^{2}\mu^2}{(\pi)^2}\frac{1}{x}\nonumber \\
&\times&\int_{0}^{2\pi} d\theta \int_{\alpha_s\mu}^{\infty}
dp_t \frac{(1-\cos^2\theta)}
{p_t[p_{t}^{2}+k_{t}^{2}+2p_tk_t\cos\theta]}
\ln \left(\frac{1}{x}\right)
\label{eq:interval}
\end{eqnarray}
where $\theta$ is the angle between $\vec{p_t}$ and $\vec{k_t}$.
The integral above can be performed exactly and the contribution
to~(\ref{eq:distfunc}) from the $\ln (1/x)$ terms becomes
\begin{eqnarray}
\Bigg(\frac{1}{\pi R^{2}}\frac{dN}{dxd^{2}k_{t}}\Bigg)_{q1}=
N_c(N_{c}^{2}-1)
\frac{\alpha_{s}^{2}\mu^{2}}{xk_{t}^{2}}\frac{C(k_t)}{\pi^3}
\ln \left(\frac{1}{x}\right).
\label{eq:contsecandthird}
\end{eqnarray}
with $C(k_t)$ given by
\begin{eqnarray}
C(k_t)=
2\left( \ln (\frac{k_t}{\alpha_s\mu}) + \frac{1}{2}\right).
\label{eq:coeffitientCk}
\end{eqnarray}
Notice that the above expression means that the second term in the
perturbative expansion of $dN/dxd^2k_t$ in $\alpha_s$,
develops the large factor $\ln(1/x)$ and that in the kinematical region of
interest, the product $\alpha_s\ln(1/x)$ is not small.
Furthermore, let us impose ordering in transverse momentum. This is the
statement that the main contribution to the distribution function comes from
the momentum region for which the emitted gluon's transverse momentum is larger
than that of the original one~\cite{glr}. The
effect is to restrict the integration interval for the radial component of
$\vec{p_t}$ in equation~(\ref{eq:interval}) which now runs between
$\alpha_s\mu$ and $k_t$. The reader can check that the above results in the
modification of~(\ref{eq:coeffitientCk}) which now reads like
\begin{eqnarray}
C(k_t)=
2\ln \left(\frac{k_t}{\alpha_s\mu}\right).
\label{eq:coeffitientC}
\end{eqnarray}
We can now include the contribution from the background
Weizs\"{a}cker--Williams field as given by equation~(\ref{eq:WWdist}). Thus
finally, the perturbative expression for the gluon distribution function to
second order in $\alpha_s$ becomes
\begin{eqnarray}
\frac{1}{\pi R^{2}} \frac{dN}{dxd^{2}k_{t}}& = &
\frac{\alpha_{s} \mu^{2} (N_{c}^{2} - 1)}{\pi^{2}}
\frac{1}{xk^{2}_{t}} \Bigg\{ 1 + \frac{\alpha _s N_c}{\pi} C(k_t)
\ln \left(\frac{1}{x}\right) \Bigg\}.
\label{eq:gluondistfuncsec}
\end{eqnarray}
Equation~(\ref{eq:gluondistfuncsec}) contains both $\ln(1/x)$ and $\ln(k_t)$
corrections to the $1/(xk_t^2)$ distribution and they represent the first order
contributions to the perturbative expansion for the distribution function. In
the kinematical region of validity, these corrections are large. This signals
that in order to properly account for the perturbative corrections one has to
devise a mechanism to isolate and sum up these leading contributions. Also
notice that equation~(\ref{eq:MAINRES}) is more general. In principle, it
contains the information about the non--perturbative corrections as well. That
information is in the function $\gamma(p_t)$ and it can be extracted by means
of a Monte Carlo analysis for the whole $k_t$ domain. These issues will be
treated in a future work.
Diagrammatically, we can represent the background gluon field coupled
to the external source (valence quarks), in momentum space, by means of
figure 1a. The background field (wavy line) is by itself of order $1/g$,
according to equation~(\ref{eq:alphau}) and the coupling to the external source
(cross) can be considered to n-th order in the parameter $g^2\mu/k_t$ by means
of the weak coupling expansion of equation~(\ref{eq:CA}). As an example, the
Weizs\"{a}cker--Williams distribution is obtained through the correlation of
the
background field taking the average over the source to first order (n=1) in
$g^2\mu/k_t$. This can be represented as in figure 1b where the broken wavy
line means that the momentum $k$ is not integrated over.
The gluon propagator in the presence of the background field can be computed
perturbatively in the coupling constant $g$ and the m-th order gluon propagator
can be represented as in figure~2a. This is because the perturbative expansion
of the gluon field involves its coupling to the background field through the
covariant derivative and the background field acts as the gauge field. Notice
that m has to be even since the gluon propagator is the correlator of two gluon
fields and each time we couple the background field to the gluon field we
introduce one power of $g$. In particular, the gluon propagator to second order
in $g$ can be represented as in figure 2b. As suggested in this figure, the
explicit dependence on $\alpha_s$ of a quantity such as the gluon distribution
function (which involves the gluon propagator) comes about only after
performing our color average through the expansion in $g^2\mu/k_t$. This is
because the coupling constant dependence of the background field and the order
of the perturbative expansion offset each other. As shown in section 3,
when computing the leading small x terms for the gluon distribution function,
any term for which we can use the sum rule~(\ref{eq:sumrule}) will not exhibit
an explicit coupling constant dependence. This becomes the criterion to decide
that such terms are vacuum contributions.
The gluon distribution function computed in section 3 can be represented by the
diagram in figure 3, where we expanded the coupling with the external source
to first order in $g^2\mu/k_t$.
\section{Loop corrections to the classical field}
Thus far, we have been concerned exclusively with the contribution of the small
fluctuations propagator to the gluon distribution function. We have shown that
this propagator induces large corrections proportional to $\alpha _{s}\ln(1/x)
\ln(k_t^2)$
and $\alpha _{s}\ln(1/x)$ to the distribution function and have argued
that the presence of these large logarithms signals the need to devise a
method to sum them up to all orders in the perturbative regime. Before we do
that we need to consider another contribution, to the same order, which comes
from the corrections to the lowest order classical field induced by quantum
fluctuations (see figure 4). This is apparant from equation~(\ref{eq:schdy1})
and equation~(\ref{eq:schdy2}) where one sees that there is a contribution
$\langle
A_{cl}^{(1)} A_{cl}^{(0)}\rangle_{\rho}$ of the same order as $\langle \langle
A_{q} A_{q} \rangle \rangle_{\rho}$.
In this section, we will compute the correction to the lowest order classical
field induced by the quantum fluctuations. We will start by writing the total
field $A^{\mu}$ in terms of background (classical) and fluctuation
(quantum) pieces allowing for the possibility that the background field may now
be different from our lowest order classical (Weizs\"acker--Williams) solution.
We will then write the equations of motion in terms of these new background
and fluctuation fields keeping terms up to and including second order in the
fluctuation fields.
Our strategy will be to consider the expectation value of the equations of
motion (in the path integral sense) and to relate the correlator of two quantum
fields to the gluon propagator in equation~(\ref{eq:GF}). We will show that
only the $+$ component of the equations of motion is modified and that the
change could be thought of as the appearance of an induced current generated by
the loop of fluctuation fields. We then proceed to explicitly compute this
induced current and show that its effect is to renormalize the coupling
constant $g$ and the original background field. In other words, the
modification induced by quantum fluctuations on the classical equations of
motion can be cast into the standard expression for the renormalization of the
coupling constant and the original background field to one loop in the light
cone gauge. This result in itself is not surprising to QCD practitioners (
see for instance reference~\cite{doksh}). What is surprising is that
this result
persists to all orders in the effective coupling $\alpha_S\mu$.
We start with the classical equations of motion
\begin{eqnarray}
D_{\mu}F^{\mu\nu}_a = g J^{\nu}_a
\label{eq:classical}
\end{eqnarray}
and expand the full gluon field as
\begin{eqnarray}
A^{\mu} = B^{\mu} + b^{\mu}
\end{eqnarray}
where $B^{\mu}$ is the background (classical) field, that is
$<A^{\mu}>\, =B^{\mu}$ while $b^{\mu}$ is the fluctuation (quantum) field
with $<b^{\mu}>\, =0$. Keeping up to quadratic terms in $b^{\mu}$,
the $+$ component of the equations of motion can be written as
\begin{eqnarray}
\partial_{-}\partial_{-}B_{a}^{-}
+(D_i\partial_- B^i)_{a} = gj_a^+ + g<J^{+}_{a}>
\label{eq:plus}
\end{eqnarray}
where $j_a^+(x)= f_{abc}<b_{b}^{i}(x)\partial^+b_{c}^{i}(x)>$. Also, $D^{\mu}$
is the covariant derivative with $B^{\mu}$ as the gauge potential.
The corresponding expressions for the minus and transverse components
of the equations of motion look like
\begin{eqnarray}
(D_iD^iB^-)_{a} - (D_i\partial^-B^i)_{a} +
(D_+\partial^+ B^-)_a + \nonumber \\
gf_{abc}\Bigg\{ <b_+^b\partial^+b_c^->+(D_i<b^ib^->)_{bc}+\nonumber \\
<b_b^i(D^ib^-)_c>-<b_b^i(D^-b^i)_c>
\Bigg\} = 0
\label{eq:minus}
\end{eqnarray}
\begin{eqnarray}
(D_jD^jB^i)_a - (D_j\partial^iB^j)_a+(D_+\partial^+ B^i)_a -
(\partial_-D^iB^-)_a + \nonumber \\
(\partial_{-}\partial^- B^i)_a + gf_{abc} \Bigg\{
<b_+^b\partial^+b_i^c>+\partial_-<b_b^-b_c^i> +
\nonumber \\
(D_j<b^jb^i>)_{bc} - <b_b^j(D^ib^j)_c> + <b_b^j(D^jb^i)_c> \Bigg\} = 0.
\label{eq:transverse}
\end{eqnarray}
The expectation values of bilinear products
of fields are related to the gluon propagator by the relation
\begin{eqnarray}
<b_a^{\mu}(x)b_b^{\nu}(y)> = -iG_{ab}^{\mu\nu}(x,y)\, .
\label{eq:relgreenexp}
\end{eqnarray}
In the above, $a,b,c$\ldots are color indices and $\mu$, $\nu$ are Lorentz
indices with $i,j$ representing the transverse components.
The reader may verify that all the terms involving bilinear products of $b^\mu$
in the minus and transverse components of the equations of motion either vanish
by explicit computation, or, because they are symmetric in the color indices
$b$ and $c$, obviously do not contribute since they are always contracted with
the totally antisymmetric structure constants $f_{abc}$. In other words, the
minus and transverse components of the classical equations of motion are not
modified by the quantum fluctuations and the set of equations reduces to
\begin{eqnarray}
-\partial_{-}\partial^{+}B_{a}^{-}
-(D_i\partial^+B^i)_{a}& = &
gj^+_a + g<J^{+}_{a}>,\nonumber \\
(D_iD^iB^-)_{a} - (D_i\partial^-B^i)_{a} +
(D_{+}\partial^+B^{-})a & = & 0, \nonumber \\
(D_jD^jB^i)_a - (D_j\partial^iB^j)_a + (D_+\partial^+B^i)_a -
(\partial_-D^iB^-)_a & = & 0.
\label{eq:set}
\end{eqnarray}
{}From now on we will concentrate only on the plus component of the equation
of motion given by equation~(\ref{eq:plus}). It is clear that
this equation is modified by the quantum fluctuations due to the presence
of the induced current. In order to understand this effect, we need to
evaluate this term explicitly. For this purpose, we write it in the following
way
\begin{eqnarray}
j^+_a(x)=f_{abc}<b_{b}^{i}(x)\partial^+b_{c}^{i}(x)>=
i f_{abc}\lim_{y\rightarrow x} \frac{\partial}{\partial y^-}
G_{bc}^{ii}(x,y).
\label{eq:term}
\end{eqnarray}
Diagramatically, this term can be represented as in figure~4 where the wavy
line is the background field and the spiral represents the loop of the
fluctuation field. The loop is the vacuum polarization tensor and the component
which contributes to the induced current and modifies the background field
(which is purely transverse) is $\Pi^{+i}$. This allows us to represent the
term $\partial^+G^{ii} \equiv D^+G^{ii}$ as $\Pi^{+i}B^i$.
We now proceed to compute the induced current explicitly. We will use
our expression for the gluon propagator as given by equation~(\ref{eq:GF}).
The first observation is that the terms in the Green's function with both
$x^-$ and $y^-$ negative will be symmetric in the color indices
and will not contribute. Also, it can be shown that the terms with both $x^-$
and $y^-$ positive yield (after we implement the limit $y \rightarrow x$ in a
Lorentz covariant way) an infinite constant (independent of the transverse
loop momentum) which vanishes upon dimensional regularization~\cite{kaku}.
However, the terms in the Green's function with opposite signs of
$x^-$ and $y^-$ are a bit tricky. For these terms taking the partial
derivative with respect to $y^-$ followed by the limit $y \rightarrow x $
is a very delicate operation and must be performed carefully. We find it more
convenient to rewrite the terms with opposite signs of $x^-$ and $y^-$ in the
Green's function in such a way as to avoid acting with $\partial/\partial y^-$
on the terms $\theta(\pm y^-)$. To do so, we will change the two dimensional
integral over $q_t$ to a four dimensional integral over $q$. We can show
that as a result, the product of theta functions of $x^-$ and $y^-$ will be
replaced by theta functions of the light cone energy $p^-$. After some
long but straightforward algebra we can rewrite the terms with opposite signs
of $x^-$ and $y^-$ in the Green's function (which we call $D^{bc}_{ij}$) as
\begin{eqnarray}
D_{ij}^{bc}(x,y)\equiv
(2i)\int\frac{d^4p}{(2\pi)^4}\frac{d^4q}{(2\pi)^4}(2\pi)
\delta(p^- - q^-)(p^- + q^-)
\frac{e^{ip(x-y)}}{(p^2-i\epsilon)(q^2-i\epsilon)}\nonumber \\
\times\Bigg\{\theta(p^-)e^{-i(p-q)x}Tr[U^{\dagger}(x_t)\tau^bU(x_t)
\tilde{F}^c(p_t-q_t)]\times
\Bigg[\delta_{ij} +\frac{p_ip_j}{p^-p^+}(e^{-ip^+y^-}-1)\nonumber \\
+\frac{q_iq_j}{q^-q^+}(e^{iq^+x^-}-1)
+\frac{q_ip_j(q_t\cdot p_t)}{p^-p^+q^-q^+}
(e^{-ip^+y^-}-1)(e^{iq^+x^-}-1)\Bigg]\nonumber \\
-\theta(-p^-)e^{i(p-q)y}Tr[\tilde{F}^b(q_t-p_t)
U^{\dagger}(y_t)\tau^cU(y_t)]\times
\Bigg[\delta_{ij} +\frac{q_iq_j}{q^-q^+}(e^{-iq^+y^-}-1)\nonumber \\
+\frac{p_ip_j}{p^-p^+}(e^{ip^+x^-}-1)
+\frac{p_iq_j(q_t\cdot p_t)}{p^-p^+q^-q^+}
(e^{-iq^+y^-}-1)(e^{ip^+x^-}-1)\Bigg]\Bigg\}\label{eq:crossterms}
\end{eqnarray}
where $\tilde{F}^b(p_t)$ is the Fourier transform of
$F^b(x_t)\equiv U^{\dagger}(x_t)\tau^bU(x_t)$. In the above expression, we are
working with the color components of expression~(\ref{eq:GF}) (as opposed to
the matrix notation), which is more suitable for the computation at hand.
According to equation~(\ref{eq:term}), we now have to compute
\begin{eqnarray}
j^+_a(x)=if_{abc}\lim_{y\rightarrow x}\frac{\partial}{\partial y^-}
D_{ii}^{bc}(x,y).
\label{eq:renorm}
\end{eqnarray}
There are three distinct pieces to the computation corresponding to the
different number of factors of $p^-$ ($q^-$) in the denominator of each
term in expression~(\ref{eq:crossterms}). For the rest of the section, we will
outline the procedure for the computation of one of them, namely, the term
proportional to $\delta_{ij}$ and quote the result for the other two terms.
In appendix II, we will show the detailed computation of some of the
integrals necessary to fill in the intermediate steps.
There are two important details to keep in mind while computing explicitly
expression~(\ref{eq:renorm}): first, we have to implement the limiting
procedure in a Lorentz covariant way. Second, when carrying out the integration
over the relative and total (light cone) energies, we have to allow for a
non--zero energy flow into the loop by not setting the total energy to zero, in
spite of the structure of the integral which seems to require it to be so. The
procedure is nothing but the well known {\it point splitting method}.
In order to implement the limit $y\rightarrow x$ in a Lorentz covariant
way, we first transform the induced current to momentum space and then
integrate over the relative (loop) momentum. We make use of the following
identity. Let $f(x,y)$ be a function of the four--dimensional variables $x$ and
$y$ for which we want to compute the limit when $x \rightarrow y$.
Fourier transform $f$ to momentum space with respect to both $x$ and $y$
\begin{eqnarray}
\tilde{f}(k,k')= \int d^4xd^4y~f(x,y)
e^{ikx}e^{ik'y}
\end{eqnarray}
and then make the change of variables
\begin{eqnarray}
s=\frac{k-k'}{2}\nonumber \\
S=\frac{k+k'}{2}\label{eq:transformation}
\end{eqnarray}
where $s$ and $S$ can be thought of as the relative and total momenta
respectively. Then
\begin{eqnarray}
\tilde{f}(s,S)= \int d^4xd^4yf(x,y)
e^{is(x-y)}e^{iS(x+y)}.
\end{eqnarray}
Integrating over $d^4s/(2\pi)^4$ will give $\delta^4(x-y)$
which will set $y=x$ upon integrating over $y$. Hence,
\begin{eqnarray}
\int d^4s \tilde{f}(s,S)= \int d^4x~e^{i2Sx}f(x,x)\, ,
\end{eqnarray}
which is the expression for the Fourier transform of $f(x,y)$ in the limit when
$y \rightarrow x$ as a function of $2S$. With the above remarks in mind, we
proceed to take the partial derivative with respect to $y^-$ in
equation~(\ref{eq:crossterms}) and then to Fourier transform with respect to
$x$ and $y$ to the momentum variables $k$ and $k'$. Let us look at the piece
proportional to $\delta_{ij}$. Set $i=j$ and then perform the $p^{\pm}$,
$q^{\pm}$ and $p_t$ integrations to get
\begin{eqnarray}
\int d^4xd^4ye^{ikx}e^{ik'y}\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}&\!\!\!=\!\!\!&
(-2\pi)\delta({k'}^- + k^-)
\int \frac{d^2q_t}{(2\pi)^2}\nonumber \\
&\!\!\!\times\!\!\!&
\frac{({k'}^- - k^-){k'}^+}{({k'}^-)(k^-)}
Tr[\tilde{F}^a(k_t+q_t)\tilde{F}^b({k'}_t-q_t)]\nonumber \\
&\!\!\!\times\!\!\!&
\left\{
\frac{\theta({k'}^-)}{({k'}^+ - \frac{({k'}_{t}^{2} -i\epsilon )}{2{k'}^-})
(k^+ - \frac{(q_{t}^{2} -i\epsilon )}{2k^-})}\right.\nonumber \\
&\!\!\!-\!\!\!& \left.
\frac{\theta({-k'}^-)}{({k}^+ - \frac{({k}_{t}^{2} -i\epsilon )}{2{k}^-})
({k'}^+ - \frac{(q_{t}^{2} -i\epsilon )}{2{k'}^-})}\right\},
\end{eqnarray}
where the index $1$ refers to our considering the first of the terms in
equation~(\ref{eq:crossterms}), namely, the term proportional to $\delta_{ij}$.
Now, perform the change of variables~(\ref{eq:transformation}) to write the
above expression in terms of relative and total momentum and integrate over
the relative momentum. The expression to evaluate becomes
\begin{eqnarray}
\int \frac{d^4s}{(2\pi)^4}
d^4xd^4ye^{is(x-y)}e^{iS(x+y)}
\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}=
(-2\pi)\delta(2S^-)\nonumber \\
\times \int \frac{d^4s}{(2\pi)^4}
\frac{d^2q_t}{(2\pi)^2}
\frac{(-2s^-)(S^+-s^+)}{(S^--s^-)(S^-+s^-)}
Tr[\tilde{F}_b(s_t+S_t+q_t)\tilde{F}_c(S_t-s_t-q_t)]\nonumber \\
\times\left\{\frac{\theta(S^- - s^-)}
{(S^+ - s^+ - \frac{(S_t -s_t)^{2} -i\epsilon }{2(S^- - s^-)})
(S^+ + s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^- + s^-)})}\right.
\nonumber \\
-\left.\frac{\theta(-S^- + s^-)}
{(S^+ + s^+ - \frac{(S_t + s_t)^{2} -i\epsilon }{2(S^- + s^-)})
(S^+ - s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^- - s^-)})}\right\}.
\label{eq:midstep}
\end{eqnarray}
We will perform the transverse momentum integrations last. For the moment,
let us concentrate on the $s^+$ and $s^-$ integrals. It is more convenient to
do the $s^+$ integral first since this can be done by contour integration. The
$s^+$ dependent integral of the above expression, denoted by $I$, is
\begin{eqnarray}
I&\!\!\!=\!\!\!&\int \frac{ds^+}{(2\pi)}\left\{
\frac{\theta(S^--s^-)}
{(S^+ - s^+ - \frac{(S_t -s_t)^{2} -i\epsilon }{2(S^- - s^-)})
(S^+ + s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^- + s^-)})}\right.
\nonumber \\
&\!\!\!-\!\!\!&\left.
\frac{\theta(-S^-+s^-)}
{(S^+ + s^+ - \frac{(S_t+s_t)^{2} -i\epsilon }{2(S^-+s^-)})
(S^+-s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^--s^-)})}\right\}
(S^+ - s^+).
\end{eqnarray}
This integral has a logarithmically divergent piece which can be isolated
by adding and subtracting the term $(S_t -s_t)^{2}/2(S^- - s^-)$ to
the numerator. This divergent piece can be shown to give a constant
independent of the transverse loop momentum and therefore it vanishes upon
dimensional regularization in the transverse direction. The remaining piece
reads as
\begin{eqnarray}
I&\!\!\!=\!\!\!& - \int \frac{ds^+}{(2\pi)}
\left\{
\frac{\theta(S^--s^-)}
{(S^+ - s^+ - \frac{(S_t -s_t)^{2} -i\epsilon }{2(S^- - s^-)})
(S^+ + s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^- + s^-)})}\right.
\nonumber \\
&\!\!\!-\!\!\!&\left.
\frac{\theta(-S^-+s^-)}
{(S^+ + s^+ - \frac{(S_t+s_t)^{2} -i\epsilon }{2(S^-+s^-)})
(S^+-s^+ - \frac{(q_{t}^{2} -i\epsilon )}{2(S^--s^-)})}\right\}
\frac{(S_t-s_t)^2}{2(S^--s^-)}.
\label{eq:tworplusint}
\end{eqnarray}
Let us investigate the above expression in some detail. The
integrand has two poles in the complex $s^+$ plane. Their
location depends on the signs of $(S^- - s^-)$ and $(S^- + s^-)$. If the two
poles are on the same side of the real axis, then we can close the contour
of integration on the other side of the real axis and the integral vanishes.
So in order to get a non--vanishing result, the two poles must be on opposite
sides of the real axis. Recall that $2S$ is the total or external momentum
flowing into the loop and thus it has to be kept fixed (and finite for a
non--trivially zero loop integral) while working in momentum space. Thus for a
given sign of $S^-$ only one of the two terms in the
integral~(\ref{eq:tworplusint}) above contributes. This can be seen as follows:
the first term in~(\ref{eq:tworplusint}) is
non--zero only if the two conditions
\begin{eqnarray}
S^--s^->0,\;\;\;\; S^-+s^->0
\label{eq:condition1}
\end{eqnarray}
are satisfied, whereas the second term is non--vanishing only if
\begin{eqnarray}
S^--s^-<0,\;\;\;\; S^-+s^-<0.
\label{eq:condition2}
\end{eqnarray}
First, take $S^-$ positive. Then only the first condition~(\ref{eq:condition1})
gives overlapping intervals for $s^-$ namely $S^->s^-$ and $s^->-S^-$ or
$-S^-<s^-<S^-$ whereas the second condition~(\ref{eq:condition2}) does not.
Therefore the second term in equation~(\ref{eq:tworplusint}) can be disregarded
and only the first one is non--vanishing. The opposite is true for $S^-$
negative in which case only the second of the terms in
equation~(\ref{eq:tworplusint}) contributes.
The integral~(\ref{eq:midstep}) however turns out to be independent of the sign
of $S^-$ (as we shall describe below). The reason is the {\it scaling} property
of the integral over $s^-$. This can be understood by recalling that after all,
the overall expression, equation~(\ref{eq:midstep}), is explicitly proportional
to $\delta(2S^-)$ and any term proportional to $S^-$ can be thrown away after
scaling the integration variable $s^-$ by $S^-$. With these remarks in mind,
let us continue working with a definite sign of $S^-$, say, $S^->0$. Performing
the integral~(\ref{eq:tworplusint}) we get
\begin{eqnarray}
I=\frac{-i\theta(S^--s^-)\theta(S^-+s^-)(S^-+s^-)(S_t-s_t)^2}
{S^-[4S^+S^-(1-\xi)(1+\xi) - q_{t}^{2}(1-\xi) - (S_t - s_t)^2(1+\xi)]}
\end{eqnarray}
in terms of which, equation~(\ref{eq:midstep}) is
\begin{eqnarray}
\int \frac{d^4s}{(2\pi)^4}
d^4xd^4ye^{is(x-y)}e^{iS(x+y)}
\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}=\nonumber \\
(-2i)\delta(2S^-)\int \frac{d^2s_t}{(2\pi)^2}\frac{d^2q_t}{(2\pi)^2}
Tr[\tilde{F}^a(s_t+S_t+q_t)\tilde{F}^b(S_t-s_t-q_t)](S_t -s_t)^2\nonumber \\
\int_{-S^-}^{S^-} \frac{s^-ds^-}{(S^-\!\! -\! s^-)
[4S^+(S^-\! +\! s^-)(S^-\!\! -\! s^-)\! -\! q_{t}^{2}(S^-\!\! -\! s^-)
\!-\!(S_t\! -\! s_t)^2(S^-\! +\! s^-)]}.
\end{eqnarray}
Let us now look at the integration over $s^-$. We scale $s^-$ by $S^-$, that is
we define the variable $\xi = s^-/S^-$.
We notice that after scaling we can make use of the explicit factor
$\delta (2S^-)$ in equation~(\ref{eq:midstep}) and
we can safely throw away any term which is still explicitly proportional
to $S^-$. Thus, the term $[4S^+S^-(1-\xi)(1+\xi)]$ in the denominator of the
above drops. As a result, the overall expression~(\ref{eq:midstep}) becomes
\begin{eqnarray}
\int \frac{d^4s}{(2\pi)^4}
d^4xd^4ye^{is(x-y)}e^{iS(x+y)}
\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}=\nonumber \\
(2i)\delta(2S^-)\int \frac{d^2s_t}{(2\pi)^2}\frac{d^2q_t}{(2\pi)^2}
Tr[\tilde{F}^a(s_t+S_t+q_t)\tilde{F}^b(S_t-s_t-q_t)](S_t -s_t)^2\nonumber \\
\int_{-1}^{1} \frac{\xi d\xi}{(1-\xi)
[q_{t}^{2}(1 - \xi)+(S_t - s_t)^2(1 + \xi)]}.
\label{eq:overallscale}
\end{eqnarray}
To proceed further, it is convenient to shift $q_t \rightarrow q_t - s_t$.
Then the arguments of ${F'}^s$ in the trace of the above expression become
independent of $s_t$ and can be taken outside the $s_t$ integral which will
be evaluated next. The $s_t$ integral is a formally divergent integral
and must be regulated. This is done in appendix II using the dimensional
regularization method. We find that
\begin{eqnarray}
\int \frac{d^{2\omega} s_t}{(2\pi)^2}
\frac{(S_t -s_t)^2}{[(q_t - s_t)^2(1-\xi) + (S_t -s_t)^2(1+\xi)]}
=\frac{\Gamma(-\omega )}{16\pi}\xi(1-\xi)(S_t-q_t)^2,
\label{eq:divergent}
\end{eqnarray}
which brings expression~(\ref{eq:overallscale}) to read like
\begin{eqnarray}
\int \frac{d^4s}{(2\pi)^4}
d^4xd^4ye^{is(x-y)}e^{iS(x+y)}
\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}=
(2i)\frac{\Gamma(-\omega)}{16\pi}
\delta(2S^-)\nonumber \\
\times \int \frac{d^2q_t}{(2\pi)^2}
Tr[\tilde{F}^a(S_t+q_t)\tilde{F}^b(S_t-q_t)](S_t -q_t)^2
\int_{-1}^{1} \xi^{2} d\xi .
\end{eqnarray}
The integration over $\xi$ can now be done easily. It just gives a factor of
$2/3$ and the remaining transverse momentum integral can be computed by using
the explicit form of $\tilde{F}$ in terms of the gauge transforms $U$. We also
show this in appendix II where we find that
\begin{eqnarray}
\int \frac{d^2q_t}{(2\pi)^2}
Tr[\tilde{F}_b(S_t+q_t)\tilde{F}_c(S_t-q_t)](S_t -q_t)^2 =
\frac{g^2}{2}f^{bcd}\tilde{\rho}_d(2S_t)
\end{eqnarray}
with
\begin{eqnarray}
\tilde{\rho}_d(2S_t)=\int d^2x_t e^{2iS_tx_t}\rho_d(x_t)
\end{eqnarray}
being the Fourier transform of the charge density with respect to $2S_t$.
Putting everything together, we get finally the result that the expression for
the Fourier transform of the induced current coming from the term proportional
to $\delta_{ij}$ as a function of $2S$ is
\begin{eqnarray}
\tilde{j}^+_a(2S)_{(1)}&\!\!\!\equiv \!\!\! & i f_{abc}
\int \frac{d^4s}{(2\pi)^4}
d^4xd^4ye^{is(x-y)}e^{iS(x+y)}
\frac{\partial}{\partial y^-}
D^{ii}_{bc}(x,y)_{(1)}\nonumber \\
&\!\!\!=\!\!\!&
-\frac{\Gamma(-\omega)}{8\pi}g^2
\tilde{\rho}_d(2S_t)\delta(2S^-) \, .
\label{eq:contfirstterm}
\end{eqnarray}
Above, we have written only the divergent part of the gamma function when
$\omega \rightarrow 1$ and have used that for $SU(3)$
the product $f_{abc}f_{bcd}\equiv C_A\delta_{ad}$ with $C_A=3$. The remaining
three terms in equation~(\ref{eq:crossterms}) can be evaluated in a similar
fashion. Here we just quote the result
\begin{eqnarray}
\tilde{j}^+_a(2S)_{(2)}&=&
\frac{\Gamma(-\omega)}{8\pi}g^2
\tilde{\rho}_d(2S_t)\delta(2S^-)\nonumber \\
\tilde{j}^+_a(2S)_{(3)}&=&
\frac{\Gamma(-\omega)}{8\pi}g^2
\tilde{\rho}_d(2S_t)\delta(2S^-)\nonumber \\
\tilde{j}^+_a(2S)_{(4)}&=&
4\frac{\Gamma(-\omega)}{8\pi}g^2
\tilde{\rho}_d(2S_t)\delta(2S^-)\, .
\label{eq:contrestterms}
\end{eqnarray}
The final result for the Fourier transform of the induced current is obtained
by adding the four terms given by equations~(\ref{eq:contfirstterm})
and~(\ref{eq:contrestterms}) and it becomes
\begin{eqnarray}
\tilde{j}^+_a(p) = g^2\Gamma(-\omega)\frac{5}{8\pi}
\tilde{\rho}_a(p_t)\delta(p^-) \, ,
\label{eq:totalrenorm}
\end{eqnarray}
where we have renamed $2S \rightarrow p$. Thus the term that modifies the plus
component of the equations of motion in momentum space just becomes
$g\tilde{j}^+_a(p)$.
We proceed to argue that by rewriting this result in terms of an expression
involving the components of the polarization operator, we can absorb the effect
of the loop corrections on the equations of motion, into the renormalization of
the coupling constant. Let us first recall that according to
equations~(\ref{eq:trans}) and~(\ref{eq:alphau}), which are the classical
solutions to the equations of motion, the expression for the zeroth
order background field in momentum space, can be written as
\begin{eqnarray}
A^{i(0)}_b(p)&\!\!\!\equiv\!\!\!&\int d^4xe^{ipx}A^{i(0)}_b(x)\nonumber \\
&\!\!\!=\!\!\!&(2\pi)g\left(\frac{p^i}{p_t^2p^+}\right)
\tilde{\rho}_b(p_t)\delta(p^-)\, .
\end{eqnarray}
Therefore, notice that $g\tilde{j}^+_a(p)$ can be written as
\begin{eqnarray}
g\tilde{j}^+_a(p)=\Pi_{ab}^{+i}(p)A_b^{i(0)}(p)
\end{eqnarray}
with $\Pi_{ab}^{+i}(p)$ given by
\begin{eqnarray}
\Pi_{ab}^{+i}(p)= g^2p^+p^i\left(\frac{5\Gamma(-\omega)}{16\pi^2}\right)
\delta_{ab}
\end{eqnarray}
which is the standard expression for the $\;+i\;$ components of the
polarization
operator in light cone gauge~\cite{leibbrandtreview}. The fact that we recover
this well known result is truly remarkable and is one indicator of the
success of our formalism.
We can now take an {\it ansatz} for the formal solution of the system of
equations~(\ref{eq:set}) to be
\begin{eqnarray}
B^-(x)&\!\!\!=\!\!\!&0,\nonumber \\
B^i(x)&\!\!\!=\!\!\!&\theta(x^-)\alpha^i_R(x_t),\nonumber \\
\tau\cdot\alpha^i_R(x)&\!\!\!=\!\!\!&\frac{i}{g_R}U(x_t)
\nabla^iU^{\dagger}(x_t).
\label{eq:anzatsforB}
\end{eqnarray}
where $\;g_R\;$ is the renormalized coupling constant whose expression is
obtained
through the computation of $\Pi^{\mu\nu}$ and will be given explicitly by
\begin{eqnarray}
g_{R}^{2}=Z_3g^2,\;\;\;\;\;
Z_3=\left(1+\frac{11g^2}{16\pi^2(1-\omega)}\right)\, .
\end{eqnarray}
This in turn means, according to~(\ref{eq:anzatsforB}), that the field $B$ gets
renormalized by the inverse of the constant that renormalizes $g$:
\begin{eqnarray}
B^{i}(x)=Z^{-1/2}A^{i(0)}(x).
\end{eqnarray}
The above exercise has taught us the important lesson that the modifications to
the background field introduced by the quantum fluctuations do not induce extra
terms in the expression for the distribution function~(\ref{eq:MAINRES}).
Further, their effect can be included by replacing the coupling constant $g$ by
the renormalized coupling constant $g_R$.
\section{Summary}
We have presented in this paper an expression for the quantum corrections to
the Weizs\"acker--Williams gluon distribution at small x valid to all orders in
the parameter $\alpha_s\mu$. We used this expression to compute explicitly the
leading $\ln(1/x)$ and $\ln(k_t)$ terms in the momentum regime
$\alpha_{s}\mu\ll k_{t}$. We have shown that the perturbative approach
introduces a series expansion parameter $\alpha_s \ln(1/x)$ which is large and
thus, forces us to devise a method to sum up the leading contributions to all
orders in that expansion parameter. Nevertheless, the present result already
signals that at small x values the gluon distribution function will be modified
significantly from the $1/(xk_t^2)$ behavior.
We wish to emphasize that our central result, equation~(\ref{eq:MAINRES}),
contains in principle the information about the quantum correction to the
classical distribution to all orders in the parameter $\alpha_s\mu$.
We have found that the only effect the quantum corrections have on the
classical background field can be absorbed into the renormalization of the
field and the running of the coupling constant.
We have not addressed the issue of summing up the perturbative series in this
paper. In the weak coupling limit, this is equivalent to solving an integral
equation for virtual corrections to the gluon propagator. Another issue we
would like to address is whether we can relax some of the constraints in the
model. In particular, we need to pay attention to the restriction set by the
necessity of having a large (perhaps too large) $A$ nucleus in order to compare
our predictions to experimental data. These issues will be addressed in a
future work.
\section*{Appendix I: The propagator in momentum space}
In this appendix, we shall Fourier transform the small fluctuations propagator
to momentum space. This result will be useful for computing
the distribution function and may also be useful for other computations.
We will first consider the propagator without averaging over all
possible color orientations of the source fields.
We shall derive two representations of the propagator. One representation
will include an integration over the variable $p^+$. This representation
will be useful for computing distribution functions. We will also present
a second representation where we have performed this integration over $p^+$.
Finally, we will present a result for the propagator after averaging over
the colors of the external field.
We recall that in coordinate space the propagator has the form
\begin{eqnarray}
G_{ij}^{\alpha\beta ;\alpha^\prime \beta^\prime}(x,y)
\!\!\!\!\!& = &\!\!\!\! - \! \int \!\!{{d^4p}\over {(2\pi)^4}}
{{e^{ip(x-y)}}\over {p^2-i\epsilon}}
\Bigg\{\! \bigg[\delta_{ij}\!+\!{p_ip_j\over {p^-p^+}}(2e^{ip^+(x^-\! - y^-)}
\!-\! e^{-ip^+y^-}\!-\! e^{ip^+x^-})\bigg]\nonumber \\
&\times & \bigg[\theta(-x^{-})\theta(-y^{-})\tau_{a}^{\alpha\beta}
\tau_{a}^{\alpha^\prime \beta^\prime} + \theta(x^-)\theta(y^-)\,
F_{a}^{\dagger\alpha\beta}(x_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}(y_{t})
\bigg]\nonumber \\
\!\!\!\!\!&+ &\!\!\!\!
\theta(-x^-) \theta(y^-)
\int \!{{d^2 q_t} \over {(2\pi)^2}} d^2 z_t \,\, e^{i(q^{+}-p^{+})y^{-}}
e^{i(p_t-q_t)(y_t-z_t)} \nonumber \\
\!\!\!\!\!&\times &\!\!\!\!
F_{a}^{\dagger\alpha \beta}(z_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}\!(y_{t})\bigg[\delta_{ij} +
\frac{p_{i}p_{j}}{p^-p^{+}}
(e^{ip^{+}x^{-}}\!-1)\nonumber \\
\!\!\!\!\!& + &\!\!\!\!\!\! \frac{q_{i}q_{j}}{p^-q^{+}}
(e^{-iq^{+}y^{-}}\!-1)
+ \frac{p_{i}q_{j}p_{t}\cdot q_{t}}
{(p^{-}p^{+})(p^{-}q^{+})}(e^{ip^{+}x^{-}}\!-1)
(e^{-iq^{+}y^{-}}\!-1)\bigg]\nonumber \\
\!\!\! & + &\!\!\!\!\!
\theta(x^-) \theta(-y^-)
\int\! {{d^2 q_t} \over {(2\pi)^2}} d^2 z_t \,\,
e^{i(p_t-q_t)(z_t - x_t)} e^{-i(q^{+}-p^{+})x^{-}}\nonumber \\
\!\!\!\!\!&\times &\!\!\!\! F_{a}^{\dagger\alpha \beta}(x_{t})
F_{a}^{\dagger\alpha^\prime \beta^\prime}\!(z_{t})\bigg[\delta_{ij} +
\frac{p_{i}p_{j}}{p^{-}p^{+}}(e^{-ip^{+}y^{-}}\!\!-1)
+ \frac{q_{i}q_{j}}{p^{-}q^{+}}(e^{iq^{+}x^{-}}\!-1)\nonumber \\
\!\!\!\!\!& + &\!\!\!\!\!\!
\frac{q_{i}p_{j}p_{t}\cdot q_{t}}{(p^{-}p^{+})(p^{-}q^{+})}
(e^{-ip^{+}y^{-}}\!-1)
(e^{iq^{+}x^{-}}\!-1)\bigg]\Bigg\} \, ,
\label{eq:GF1}
\end{eqnarray}
where $q^{+}=p^{+}+ \frac{q_{t}^{2}-p_{t}^{2}}{2p^{-}}$ and
\begin{eqnarray}
F_{a}^{\dagger\alpha \beta}(x_{t}) =\bigg( U(x_{t})\tau_{a}U^{\dagger}
(x_{t})\bigg)^{\alpha\beta}.
\end{eqnarray}
We now wish to define the Fourier transformed Green's function
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{ij} (k,k^\prime)
= \int d^4xd^4y e^{-ikx+ik^\prime y}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{ij} (x,y) \, ,
\end{eqnarray}
which we can divide into four pieces as
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{ij} (k,k^\prime) & = &
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{--ij} (k,k^\prime) +
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{++ij} (k,k^\prime)\nonumber \\
& + &
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{+-ij} (k,k^\prime) +
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{-+ij} (k,k^\prime)
\end{eqnarray}
In this equation, the first index $\pm$ refers to the index of
$\theta (\pm x^-)$ and the second to $\theta (\pm y^-)$ in the definition
of the coordinate space Green's function.
Let us first consider the $--$ component
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{--ij} (k,k^\prime)
&\!\!\! = \!\!\! &
- \int {{d^4p} \over {(2\pi)^4}} d^4xd^4y e^{i(p-k)x-i(p-k^\prime)y}
\theta(-x^-) \theta (-y^-) \tau_a^{\alpha \beta} \tau_a^{\alpha^\prime
\beta^\prime} {1 \over {p^2-i\epsilon}} \nonumber \\
&\!\!\!\times\!\!\!& \left\{ \delta_{ij}
+{{p_ip_j} \over {p^-p^+}} (2e^{ip^+(x^--y^-)} - e^{-ip^+y^-}
-e^{ip^+x^-}) \right\} \, .
\end{eqnarray}
We can perform the integrations over $p^-$, $p_t$, $\vec{x}$
and $\vec{y}$ to obtain,
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{--ij} (k,k^\prime)
= - (2\pi) \delta(k^- - k^{\prime -}) (2\pi)^2 \delta^{(2)}(k_t - k^\prime_t)
\tau_a^{\alpha \beta} \tau_a^{\alpha^\prime \beta^\prime}
\Delta_{--}(k,k^\prime)
\end{eqnarray}
where
\begin{eqnarray}
\Delta_{--}(k,k^\prime) & = & \int dx^-dy^- {{dp^+} \over {2\pi}}
\frac{\theta (-x^-) \theta (-y^-) e^{-i(p^+-k^+)x^- + i (p^+-k^{\prime +})y^-}}
{k_t^2 - 2p^+k^- - i \epsilon}\nonumber \\
&\times&\Bigg\{ \delta_{ij}
+{{k_ik_j} \over {k^-p^+}} (2e^{ip^+(x^--y^-)} - e^{-ip^+y^-}
-e^{ip^+x^-}) \Bigg\}
\end{eqnarray}
Now let us do the $x^-$ and $y^-$ integrations. In performing these
integrations, $i\epsilon$ factors will appear which will guarantee
the convergence of the integrals at infinity. We find
\begin{eqnarray}
\Delta_{--}(k,k^\prime) & = & \int {{dp^+} \over {2\pi}}
\,{1 \over {k_t^2 - 2 p^+k^- -i\epsilon}}\, {1 \over {p^+-k^++i\epsilon}}
\,{1 \over {p^+-k^{\prime+}-i\epsilon}} \nonumber \\
&\times& \left\{ \delta_{ij} + {{k_ik_j} \over {k^-(k^+-i\epsilon)
(k^{\prime +} + i\epsilon)}} (2p^+ - k^+ -k^{\prime +}) \right\}
\end{eqnarray}
Finally, we can perform the integration over $p^+$ to find
\begin{eqnarray}
\Delta_{--}(k,k^\prime)& =& {i \over {k^{\prime +} - k^+ +i\epsilon}}
\,{1\over {k^{\prime 2} -i\epsilon}}\,\Bigg\{ \theta (k^-) \Big( \delta_{ij}
\nonumber \\
&+& {{k_ik_j} \over {k^-(k^+-i\epsilon)
(k^{\prime +} + i\epsilon)}} (k^{\prime +} -k^+) \Big) +
k,k^\prime \rightarrow -k^\prime,-k \Bigg\}
\end{eqnarray}
Now, to fully define this Green's function, we must specify the nature
of the singularity at $k^- = 0$. In the last equation, we would like
to use the Leibbrandt--Mandelstam prescription on $1/(k^- +i\epsilon /k^+)$
whenever we have the combination $1/k^-k^+$ and
$1/(k^-+i\epsilon /k^{\prime +})$ whenever we have the combination
$1/k^-k^{\prime +}$. We would like to go further however
and define $1/k^+k^-$ as $1/(k^-k^+ + i\epsilon)$. This can be done
as follows: we use $1/(k^+-i\epsilon) = 2\pi i \delta (k^+) +
1/(k^++i\epsilon)$
whenever we have the constraint that $k^- > 0$, and a similar modification
when $k^- < 0$ for ${k^\prime}^+$.
On the other hand, it is more difficult to implement the Leibbrandt-Mandelstam
prescription in the expression which involves the integral over $p^+$.
However, we will only use this result when both $k^+$ and $k^{\prime +}$ have
the same sign and are non-zero. In this case the Leibbrandt-Mandelstam
prescription is unambiguous.
Our results for the two representations are therefore
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{--ij} (k,k^\prime)
& = & -2\pi i \delta(k^- - k^{-\prime}) (2\pi)^2 \delta^{(2)}
(k_t - k_t^\prime) \tau_a^{\alpha \beta} \tau_a^{\alpha^\prime \beta^\prime}
{1\over {k^{\prime 2} -i\epsilon}}
\nonumber \\
&\times&\Bigg\{ {1 \over {k^{\prime +} - k^+ +i\epsilon}}
\Big[ \theta (k^-) \left( \delta_{ij} + {{k_ik_j} \over {k^-k^+
k^{\prime+}}} (k^{\prime +} -k^+) \right) \nonumber \\
&+& k,k^\prime \rightarrow -k^\prime,-k \Big]
+2\pi i \delta (k^+) \theta (k^-) {{k_ik_j} \over {
k^-k^{\prime+}}}\nonumber\\
&+& k,k^\prime \rightarrow -k^\prime,-k \Bigg\}
\end{eqnarray}
Here the $k^+$, $k^{\prime +}$ and $k^-$ singularities are treated using
the Leibbrandt-Mandelstam prescription.
For $k^+$ and $k^{\prime +}$ both non-zero and of the same sign, we have the
representation
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{--ij} (k,k^\prime)
& = & -2\pi \delta(k^- - k^{-\prime}) (2\pi)^2 \delta^{(2)}
(k_t - k_t^\prime) \tau_a^{\alpha \beta} \tau_a^{\alpha^\prime \beta^\prime}
\nonumber \\
&\times& \int {{dp^+} \over {2\pi}}
{1 \over {k_t^2 - 2 p^+k^- -i\epsilon}}\, {1 \over {p^+-k^++i\epsilon}}\,
{1 \over {p^+-k^{\prime+}-i\epsilon}} \nonumber \\
&\times& \left\{ \delta_{ij} + {{k_ik_j} \over {k^-k^+
k^{\prime +}}} (2p^+ - k^+ -k^{\prime +}) \right\}\, ,
\end{eqnarray}
where the $k^-$ singularity is treated using the Leibbrandt-Mandelstam
prescription.
The evaluation of the remaining contributions to the Green's functions can be
done by the same methods as above. There is nothing really new in the analysis
except that it is longer and more involved. The subtlety is in the treatment
of the singularities in the $k^\pm$ and $k^{\prime +}$ variables. This has been
discussed above and is treated as such. The results are
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{++ij} (k,k^\prime)
&\!\!\! = \!\!\!& -2\pi i \delta (k^- - k^{-\prime})
\int {{d^2p_t} \over {(2\pi)^2}}
F_a^{\dagger \alpha \beta}(p_t-k_t)
F_a^{\dagger \alpha^\prime \beta^\prime}(k_t^\prime -p_t) \nonumber \\
&\!\!\!\times\!\!\!& \Bigg\{ {1 \over {k^+ - k^{\prime +} +i\epsilon}}
\Bigg[ \left( \delta_{ij} +
{{p_ip_j} \over {k^-k^+k^{\prime +}}}(k^+-k^{\prime +}) \right)
{ \theta (k^-)\over {p_t^2 - 2k^-k^+ - i\epsilon}} \nonumber \\
&\!\!\!+\!\!\!& k, k^\prime \rightarrow -k^\prime k \Bigg]
+2\pi i \delta (k^{\prime +}) \theta (k^-) {{p_ip_j} \over {k^-k^+}}
{1 \over {p_t^2-2k^-k^+-i\epsilon}} \nonumber \\
&\!\!\!+\!\!\!& k,k^\prime \rightarrow -k^\prime, -k
\Bigg\}\, .
\end{eqnarray}
For $k^+$ and $k^{\prime +}$ both of the same sign and non-zero, the above is
equivalent to
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{++ij} (k,k^\prime)
& = & -2\pi \delta (k^- - k^{-\prime}) \int {{dp^+d^2p_t} \over {(2\pi)^3}}
F_a^{\dagger \alpha \beta}(p_t-k_t)
F_a^{\dagger \alpha^\prime \beta^\prime}(k_t^\prime - p_t)
\nonumber \\
&\times&{ 1 \over {p_t^2-2p^+k^--i\epsilon} }\,
{1 \over {p^+-k^+-i\epsilon}}\,{1 \over {p^+-k^{\prime +} + i\epsilon}}
\nonumber \\
&\times& \left( \delta_{ij} + {{p_ip_j} \over {k^-k^+k^{\prime +}}}
(2p^+-k^+-k^{\prime +}) \right) \, .
\end{eqnarray}
We finally also obtain an expression for $G_{+-}$. It turns out that in this
expression, no restrictions on the values of $k^+$ and $k^{\prime +}$ are
needed to get the singularities in $1/k^\pm$ or $1/k^{\prime +}$ into the
Leibbrandt--Mandelstam form. The results are
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{+-ij} (k,k^\prime)
& = & 2\pi i \delta (k^- - k^{-\prime}) 2k^- \theta (k^-)
{1 \over {k^{\prime 2}-i\epsilon}} \int {{d^2p_t} \over {(2\pi)^2}}
F_a^{\dagger \alpha \beta}(p_t-k_t)
\nonumber \\
&\times&
F_a^{\dagger \alpha^\prime \beta^\prime}(k_t^\prime -p_t)
{1 \over {p_t^2-2k^-k^+ -i\epsilon}} \Bigg\{ \delta_{ij}
- {{k^\prime_ik^\prime_j} \over {k^-k^{\prime +}}}
-{{p_ip_j} \over {k^-k^+}} \nonumber \\
&+& {{p_ik^\prime_j p_t \cdot k^\prime} \over {(k^-k^+)(k^-k^{\prime +})}}
\Bigg\} \, .
\end{eqnarray}
We also have the equivalent integral representation where
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{+-ij} (k,k^\prime)
& = & 2\pi \delta (k^- - k^{-\prime}) \int {{dp^+d^2p_t} \over {(2\pi)^3}}
F_a^{\dagger \alpha \beta}(p_t-k_t)
F_a^{\dagger \alpha^\prime \beta^\prime}(k_t^\prime -p_t) \nonumber \\
&\times& {1 \over {k_t^2 -2k^-p^+ -i\epsilon}}\, {1 \over {q^+-k^+-i\epsilon}}
\,{1 \over {p^+-k^{\prime +} -i\epsilon}} \nonumber \\
&\times& \Bigg\{ \delta_{ij} -{{k^\prime_i k^\prime_j} \over
{k^-k^{\prime +}}}
-{{p_ip_j} \over {k^-k^+}} + {{p_ik^\prime_j p_t \cdot k^\prime_t}
\over {(k^-k^+) (k^-k^{\prime +})}}\Bigg\} \, .
\end{eqnarray}
In this equation,
$q^+ = p^+ + {{p_t^2-k^{\prime 2}} \over {2k^-}}$.
The expression for $G_{-+}$ is
\begin{eqnarray}
G^{\alpha \beta; \alpha^\prime \beta^\prime}_{+-ij} (k,k^\prime)
= G^{\alpha \beta; \alpha^\prime \beta^\prime}_{-+ij} (-k^\prime,-k)
\end{eqnarray}
We now want to convert these expressions for the propagator from the
matrix basis to the component basis. To do this, we make the
transformation
\begin{eqnarray}
(U(x)\tau^c U^\dagger (x)) (U(y) \tau^c U^\dagger (y))
\rightarrow 4 (Tr\, \tau^a U(x) \tau^c U^\dagger (x)) (Tr\,
\tau^b U(y) \tau^c U^\dagger (y) )
\end{eqnarray}
We then use the identity
\begin{eqnarray}
\tau^c_{\alpha \beta} \tau^c_{\alpha^\prime \beta^\prime}
= {1 \over 2} \left( \delta_{\alpha \beta^\prime} \delta_{\alpha^\prime \beta}
- {1 \over N_c} \delta_{\alpha \beta} \delta_{\alpha^\prime \beta^\prime}
\right)\, ,
\end{eqnarray}
which results in following transformation for the definition of the propagator
\begin{eqnarray}
F^{\dagger a}_{\alpha \beta} (x)
F^{\dagger a}_{\alpha^\prime \beta^\prime} (y)
\rightarrow 2 \,Tr \, F^a(x) F^b(y)
\end{eqnarray}
We now want to proceed to derive formulae for the propagator
which has been averaged over all values of the color charges of
the valence quarks. To do this we define
\begin{eqnarray}
< Tr \,U^\dagger (x) \tau^a U(x) U^\dagger (y) \tau^b U(y) >
= {1 \over 2} \Gamma^{ab}(x-y) \, .
\end{eqnarray}
Notice that
\begin{eqnarray}
\Gamma^{ab} (0) = \delta^{ab}
\end{eqnarray}
We now define
\begin{eqnarray}
< G^{ab}_{ij} (k,k^\prime ) > = (2\pi ) \delta (k^- - k^{\prime -} )
(2\pi )^2 \delta (k_t - k^\prime_t) D_{ij}^{ab} (k,k^\prime) \, .
\end{eqnarray}
As before, we can write $D_{ij}^{ab}$ as
\begin{eqnarray}
D_{ij}^{ab} = D_{-- ij}^{ab}(k,k^\prime) + D_{++ ij}^{ab}(k,k^\prime)
+ D_{+- ij}^{ab}(k,k^\prime) + D_{-+ ij}^{ab}(k,k^\prime)\, .
\end{eqnarray}
Using the above substitutions we find that
\begin{eqnarray}
D_{--ij}^{ab} & = & -\delta^{ab} \int {{dp^+} \over {2\pi}}
\, {1 \over {k_t^2 - 2 p^+ k^- -i\epsilon}}\,
{1 \over {p^+-k^+ +i\epsilon}}\, { 1 \over {p^+-k^{\prime +} -i\epsilon}}\,
\nonumber \\
&\times&\left[ \delta_{ij} + {{k_ik_j} \over {k^-k^+k^{\prime +}}}
(2p^+-k^+-k^{\prime +}) \right]\, ,
\end{eqnarray}
which is valid for $k^+,k^{\prime +}$ nonzero and $sign(k^+) = sign(k^{\prime
+} )$ . The expression with the $p^+$ integral completed is
\begin{eqnarray}
D_{--ij}^{ab} & = & -i \delta^{ab} \Bigg\{ \theta(k^-)
{1 \over k^{\prime 2}}
\Bigg[ {1 \over {k^{\prime +} -k^+ +i\epsilon }}
\left( \delta_{ij} + {{k_ik_j} \over {k^-k^+k^{\prime +}}}
(k^{\prime +} - k^+) \right)
\nonumber \\
&+&2\pi i \delta (k^+) {{k_ik_j} \over {k^-k^{\prime +}}} \Bigg]
+ k,k^\prime \rightarrow -k^\prime, -k \Bigg\} \, .
\end{eqnarray}
For $D_{++}$ we obtain
\begin{eqnarray}
D_{++ij}^{ab} & = & -\int {{dp^+d^2p_t} \over {(2\pi )^3}}
\Gamma^{ab}(p_t-k_t)
\, {1 \over {p_t^2 - 2p^+k^- -i\epsilon}} \nonumber \\
&\times&{1 \over {p^+-k^+-i\epsilon}}\,{1 \over {p^+-k^{\prime +} + i\epsilon
}} \left[ \delta_{ij} + {{p_ip_j} \over {k^-k^+k^{\prime +}}}
(2p^+-k^+-k^{\prime +}) \right] \, ,
\end{eqnarray}
which is valid for $k^+,k^{\prime +}$ nonzero and $sign(k^+) = sign(k^{\prime
+} )$ . After completing the $p^+$ integral,
\begin{eqnarray}
D_{++ij}^{ab} & = & \int {{d^2p_t} \over {(2\pi )^2}}
\Gamma^{ab}(p_t-k_t)
\Bigg\{ \theta (k^-) {1 \over {p_t^2 - 2k^-k^+ -i\epsilon}}
\nonumber \\ &\times&\Bigg[{1 \over {k^+-k^{\prime +}}} \left( \delta_{ij} +
{{p_ip_j} \over {k^-k^+k^{\prime +}}} (k^+-k^{\prime +} ) \right) + 2\pi i
\delta (k^{\prime +}) {{p_ip_j} \over {k^-k^+}} \Bigg]\nonumber \\
&+& k,k^\prime \rightarrow -k^\prime, -k \Bigg\} \, .
\end{eqnarray}
We also have
\begin{eqnarray}
D_{+-ij}^{ab} & = & \int {{dp^+d^2p_t} \over {(2\pi)^3}}
\Gamma^{ab}(p_t-k_t)
\,{1 \over {k_t^2-2k^-p^+-i\epsilon}}\, {1 \over {q^+-k^+-i\epsilon}}\,
{1 \over {p^+-k^{\prime +} +i\epsilon}}\,
\nonumber \\
&\times&\Bigg[\delta_{ij} - {{k_ik_j} \over {k^-k^{\prime +}}}
-{{p_ip_j} \over {k^-k^+}} + {{p_i k_j p_t \cdot k_t} \over {(k^-k^+)
(k^-k^{\prime +})}} \Bigg] \, ,
\end{eqnarray}
where, again,
$q^+ = p^+ + {{p_t^2-k_t^2} \over {2k^-}}$.
Doing the integral over $p^+$ gives
\begin{eqnarray}
D_{+-ij}^{ab} & = & i 2k^- \theta (k^-) {1 \over {k^{\prime 2}}}
\int {{d^2p_t} \over {(2\pi )^2}}\, \Gamma^{ab} (p_t-k_t)
{1 \over {p_t^2 - 2k^-k^+ -i\epsilon}}
\nonumber \\ &\times&
\Bigg[\delta_{ij} - {{k_ik_j} \over {k^-k^{\prime +}}}
-{{p_ip_j} \over {k^-k^+}} + {{p_i k_j p_t \cdot k_t} \over {(k^-k^+)
(k^-k^{\prime +})}} \Bigg] \, .
\end{eqnarray}
Finally, the expression for $D_{-+}$ is given by
\begin{eqnarray}
D_{-+ij}^{ab} (k,k^\prime) = D^{ba}_{+-ji} (-k^\prime,-k) \, .
\end{eqnarray}
\section*{Appendix II}
In this appendix, we will explicitly evaluate two of the integrals we
use in the computation of the one loop corrections to the classical background
field. The first integral we will evaluate is
\begin{eqnarray}
I_1= \int \frac{d^2s_t}{(2\pi)^2}
\frac{(S_t -s_t)^2}{[(q_t - s_t)^2(1-\xi) + (S_t -s_t)^2(1+\xi)]}.
\end{eqnarray}
Let us first shift $s_t \rightarrow s_t+S_t$ and write the above as
\begin{eqnarray}
I_1= \int \frac{d^2s_t}{(2\pi)^2}
\frac{s_{t}^{2}}{[(s_t+S_t-q_t)^2(1-\xi) + s_{t}^{2}(1+\xi)]}\, .
\label{eq:shift}
\end{eqnarray}
The denominator in the integral can be written as
\begin{eqnarray}
2s_{t}^{2} + (2s_t\cdot(S_t-q_t) + (S_t-q_t)^2)(1-\xi).
\label{eq:denominator}
\end{eqnarray}
Performing the change of variables
\begin{eqnarray}
v_t&=&s_t + \frac{(1-\xi)}{2}(S_t-q_t)\nonumber \\
d^2v_t&=&d^2s_t\nonumber \\
v_t^2&=&s_{t}^{2} + s_t\cdot(S_t-q_t)(1-\xi)
+ \frac{(1-\xi)^2}{4}(S_t-q_t)^2
\label{eq:changevar}
\end{eqnarray}
Eq.~(\ref{eq:denominator}) can be re--written as
\begin{eqnarray}
2[v_t^2 + \frac{(1-\xi^2)}{4}(S_t-q_t)^2]
\label{eq:newdenominator}
\end{eqnarray}
and the integral~(\ref{eq:shift})
\begin{eqnarray}
I_1= \int \frac{d^2v_t}{(2\pi)^2}
\frac{[v_t - \frac{(1-\xi)}{2}(S_t-q_t)]^2}
{2[v_t^2 + \frac{(1-\xi^2)}{4}(S_t-q_t)^2]}.
\label{eq:newshift}
\end{eqnarray}
Written in this form, we can drop the linear term in $v_t$ and the contributing
terms to~(\ref{eq:newshift}) will be
\begin{eqnarray}
I_1 &\rightarrow&\int \frac{d^2v_t}{(2\pi)^2}
\frac{v_t^2}{2[v_t^2 + \frac{(1-\xi^2)}{4}(S_t-q_t)^2]}\nonumber \\
& +&\int \frac{d^2v_t}{(2\pi)^2}
\frac{(1-\xi)^2/4(S_t-q_t)^2}{2[v_t^2 +
\frac{(1-\xi^2)}{4}(S_t-q_t)^2]}.
\label{eq:twoterms}
\end{eqnarray}
The first and second terms in expression~(\ref{eq:twoterms}) are
quadratically and logarithmically divergent. To regulate the divergences we
compute them by dimensional regularization. We thus write~(\ref{eq:twoterms})
as
\begin{eqnarray}
I_1&=& \frac{1}{2(2\pi)^2} \Bigg( \frac{2\pi^2}{\Gamma(d/2)} \Bigg)
\Bigg\{
\int_{0}^{\infty}
dv_t \frac{v_t^{d+1}}{[v_t^2 + \frac{(1-\xi^2)}{4}(S_t-q_t)^2]}\nonumber \\
&+&
\int_{0}^{\infty}
dv_t \frac{v_t^{d-1}(1-\xi)^2/4(S_t-q_t)^2}
{[v_t^2 + \frac{(1-\xi^2)}{4}(S_t-q_t)^2]}
\Bigg\}.
\label{eq:dimreg}
\end{eqnarray}
To evaluate the integrals in the above expression we use the well known
formula
\begin{eqnarray}
\int_{0}^{\infty}
\frac{u^{\beta}du}{(u^2+C^2)^\alpha} =
\frac{\Gamma(\frac{1}{2}(1+\beta))\Gamma(\alpha-\frac{1}{2}(1+\beta))}
{2(C^2)^{\alpha-(1+\beta)/2}\Gamma(\alpha)}
\label{eq:identity}
\end{eqnarray}
by means of which~(\ref{eq:dimreg}) becomes
\begin{eqnarray}
I_1= (\frac{d}{4})(\frac{1}{2\pi})^2 2\pi^{d/2}\Gamma(-d/2)
(\frac{x}{1+x})[(1-\xi^2)(S_t-q_t)^2/4]^{d/2}.
\label{eq:almostresult}
\end{eqnarray}
We are interested in the divergent part of this expression when
$d\rightarrow 2$ thus for that we take $d=2$ everywhere except in the argument
of $\Gamma$. We obtain finally
\begin{eqnarray}
I_1= \frac{\Gamma(-\frac{d}{2})}{16\pi}2\xi(1-\xi)(S_t-q_t)^2,
\end{eqnarray}
Next, we want to compute the following integral
\begin{eqnarray}
I_2=\int \frac{d^2q_t}{(2\pi)^2}
Tr[F^a(S_t+q_t)F^b(S_t-q_t)][(S_t -q_t)^2 - (S_t +q_t)^2].
\label{eq:chargeprop}
\end{eqnarray}
For this purpose let us write $\tilde{F}$ in its explicit form in terms of $U$.
\begin{eqnarray}
\tilde{F}^a(p_t)=\int d^2x_t e^{ip_tx_t} U^{\dagger}(x_t)\tau^aU(x_t),
\label{eq:tildeF}
\end{eqnarray}
and thus the integrand in~({\ref{eq:chargeprop}) can be written as
\begin{eqnarray}
\int d^2x_td^2y_t e^{i(S_t+q_t)x_t}e^{i(S_t-q_t)y_t}
[(S_t-q_t)^2 - (S_t+q_t)^2]\nonumber \\
Tr[U^{\dagger}(x_t)\tau^aU(x_t)U^{\dagger}(y_t)\tau^bU(y_t)].
\label{eq:FTint}
\end{eqnarray}
The factors $(S_t\pm q_t)^2$ are to be interpreted as derivatives acting on the
corresponding exponential. Integrating by parts, ignoring the surface terms and
with the help of Eq.~(\ref{eq:FTint}) we can write Eq.~(\ref{eq:chargeprop}) as
\begin{eqnarray}
I_2&=& -\int\frac{d^2q_t}{(2\pi)^2}d^2x_td^2y_t
e^{i(S_t+q_t)x_t}e^{i(S_t-q_t)y_t} \nonumber \\
&\times&\{Tr[U^{\dagger}(x_t)\tau^aU(x_t)
\nabla^2(U^{\dagger}(y_t)\tau^bU(y_t))]\nonumber \\
&-&Tr[\nabla^2(U^{\dagger}(x_t)\tau^aU(x_t))
U^{\dagger}(y_t)\tau^bU(y_t)]\}.
\end{eqnarray}
Performing the $q_t$
and $y_t$ integrals, Eq.~(\ref{eq:chargeprop}) becomes
\begin{eqnarray}
I_2&=& -\int d^2x_t e^{2iS_tx_t}
\Bigg\{ Tr[U^{\dagger}(x_t)\tau^aU(x_t)
\nabla^2(U^{\dagger}(x_t)\tau^bU(x_t))]\nonumber \\
&-& Tr[\nabla^2(U^{\dagger}(x_t)\tau^aU(x_t))
U^{\dagger}(x_t)\tau^bU(x_t)]
\Bigg\}\, .
\label{eq:about}
\end{eqnarray}
Since the gauge transformations $U$ and the charge density are related by
Eq.~(\ref{eq:ufield}), one can prove the following identity
\begin{eqnarray}
Tr[U^{\dagger}\tau^aU
\nabla^2(U^{\dagger}\tau^bU) -
\nabla^2(U^{\dagger}\tau^aU)
U^{\dagger}\tau^bU]= -\frac{g^2}{2}\Bigg(f^{abc}\rho_c(x_t) -
f^{bac}\rho_c(x_t)\Bigg)\!.
\label{eq:tracediff}
\end{eqnarray}
Using the antisymmety of $f^{abc}$ and plugging the above back into
Eq.~(\ref{eq:about}) we finally obtain the result
\begin{eqnarray}
I_2= g^2f^{abc}\int d^2x_t e^{2iS_tx_t} \rho_c(x_t) \, ,
\end{eqnarray}
which is the Fourier transform of the charge density with respect to $2S_t$.
\section*{Acknowledgments}
Two of the authors, A.A and J.J.M., would like to thank R. Rodriguez and
R. Madden for useful discussions.
Research supported by the U.S. Department of Energy under grants No.
DOE High Energy DE--AC02--83ER40105, No. DOE Nuclear DE--FG02--87ER--40328,
No. DOE Nuclear DE--FG06--90ER--40561, and by the DGAPA/UNAM/M\'exico.
|
\chapter{Hamiltonian Formulation of Gravity}
\section{Classical Theory}
To proceed with the Hamiltonian approach to gravity we perform a $3+1$
decomposition of the spacetime manifold, singling out a time coordinate.
A Hamiltonian is then identified that propagates the three geometry along
the time direction. The subtlety involved in this procedure is caused by
the reparameterization invariance of the Einstein-Hilbert action: due to
the existence of redundant variables, time evolution is not uniquely
defined unless a gauge is specified. However, there is a well developed
formalism for dealing with such systems \cite{dirac},
as will now be discussed.
The key to separating out the physical and redundant degrees of freedom
is to write the metric in $3+1$ form as \cite{adm}
\begin{equation}
ds^2 = -(N^t dt)^2 + h_{ij}(dx^i+N^i dt)(dx^j+N^j dt)
\end{equation}
With this labelling, the Einstein-Hilbert action becomes:
\begin{equation}
{\cal L}_{G} = \frac{1}{16\pi}\sqrt{-g}\,{\cal R} = \frac{1}{16\pi}
\sqrt{h}\,N^t\,[^{3}{\cal R} +K_{ij}K^{ij}-K^2],
\end{equation}
where $^{3} {\cal R}$ is the Ricci scalar associated with $h_{ij}$,
$K_{ab}$ is the extrinsic curvature of a constant $t$ hypersurface:
\begin{equation}
K_{ab}=\frac{1}{2N^t}[\dot{h}_{ij}-N_{i|j}-N_{j|i}],
\end{equation}
and $K\equiv K_{a}^{a}$. Throughout, Latin indices are raised and lowered by
$h_{ij}$, and $|$ denotes covariant differentiation with respect to $h_{ij}$.
Now, it is seen that no time derivatives of $N^t$ or $N^i$ appear in the
gravitational action, nor will they when matter is included, so that the
canonical momenta conjugate to these variables vanish identically:
\begin{equation}
\pi_{N^t}\equiv \frac{\partial {\cal L}}{\partial \dot{N}^t}=0
{\mbox \ \ \ \ ; \ \ \ \ } \pi_{N^i}\equiv \frac{\partial {\cal L}}{\partial
\dot{N}^i}=0.
\end{equation}
These are constraints on the phase space. $h_{ij}$, on the other hand, have
nonvanishing canonical momenta $\pi_{ij}$. These can be computed and used
to put ${\cal L}_{G}$ in canonical form:
\begin{equation}
{\cal L}_G = \pi_{ij}\dot{h}^{ij} - N^{t}{\cal H}_{t}-N_{i}{\cal H}^{i}
\end{equation}
where
\begin{equation}
{\cal H}_{t} = 8\pi h^{-\frac{1}{2}}(h_{ik}h_{jl}+h_{il}h_{jk}-h_{ij}h_{kl})
\pi^{ij}\pi^{kl} - \frac{1}{16\pi}h^{\frac{1}{2}} {\,^{3} {\cal R}},
\end{equation}
\begin{equation}
{\cal H}^{i}=-2\pi^{ij}_{|j}.
\end{equation}
The constraints $\pi_{N^{t}}=\pi_{N^{i}}=0$ must hold at all times, which
means that their Poisson brackets with the Hamiltonian must vanish. This
condition yields the secondary constraints
\begin{equation}
{\cal H}_{t}={\cal H}^{i}=0.
\end{equation}
All the dynamics of gravity is contained in these constraints. It should be
stressed that these constraints hold ``weakly'' --- they are to be imposed
only after all Poisson brackets have been computed.
\subsection{Surface Terms}
In deriving the canonical form of the action we ignored all surface terms
which arose. However, for gravity in asymptotically flat space surface terms
play an important role and cannot be neglected, so we write
\begin{equation}
H_{G}=\int\! d^{3}\!x [N^t{\cal H}_{t}+N_i {\cal H}^i]+\mbox {surface terms}.
\label{Hnew}
\end{equation}
Regge and Teitelboim \cite{regge}
showed that Einstein's equations are equivalent to
Hamilton's equations applied to $H_G$ only for a particular choice of
surface terms. The point is that if Einstein's equations are written as
\begin{equation}
\dot{h}_{ij}=A_{ij}(h,\pi) \mbox{ \ \ \ \ ; \ \ \ \ } \dot{\pi}_{ij}
=-B_{ij}(h,\pi)
\end{equation}
then we need the variation of $H_G$ to be
\begin{equation}
\delta H_G = A_{ij}\delta \pi^{ij} + B_{ij}\delta h^{ij}
\end{equation}
to ensure consistency with Einstein's equations. To put $\delta H_G$ in this
form we must integrate by parts, because space derivatives of $h_{ij}$ and
$\pi_{ij}$ appear in $H_G$, and then demand that the resulting surface term
cancels with the one we have included in (\ref{Hnew}). In \cite{regge} the
surface terms were worked out assuming a particular rate of fall-off of
$h_{ij}$, $\pi_{ij}$, $N^t$, and $N^i$ at infinity. This condition
can be relaxed \cite{ferraris}, and in general the full Hamiltonian is:
\begin{equation}
H_G=\int_{\Sigma}\!d^3\!x[N^t {\cal H}_{t}+N_{i}{\cal H}^i]+\int_{\partial
\Sigma} ds_{l}[{\cal H}_{1s}^{l}+{\cal H}_{2s}^{l}]
\end{equation}
with
$$
H_{1s}^{l}=\frac{1}{16\pi}(N^t\sqrt{h}h^{ij}U_{ij}^{l}+2\pi^{li}N_{i})
$$
\begin{equation}
H_{2s}^{l}=\frac{1}{16\pi}\left[\frac{1}{N^t}\sqrt{h}(N^{i}\partial_{i}
N^l-N^l\partial_{i}N^i)\right]
\label{surface}
\end{equation}
where
\begin{equation}
U_{jk}^{i} = \Gamma_{jk}^{i}-\delta^{i}_{(j}\Gamma_{k)}^l.
\end{equation}
The result of Regge and Teitelboim corresponds to choosing fall-off
conditions such that $H_{2s}^{l}$ vanishes at infinity.
Since ${\cal H}_{t}$ and ${\cal H}^i$ vanish when the constraints are
satisfied, the numerical value of the Hamiltonian is given solely by the
surface terms. This numerical value is the ADM mass of the system.
\subsection{Spherical Symmetry}
We now specialize to the case of spherically symmetric geometries
\cite{bcmn,polch,thiemann}. It is
possible to proceed substantially further in this case because there are no
propagating degrees of freedom -- after solving the constraints there will
only be one free parameter remaining: the ADM mass. The metric is written
\begin{equation}
ds^2= - (N^t dt)^2 +L^2(dr+N^r dt)^2+R^2(d\theta^2+{\sin{\theta}}^2
d\phi^2)
\end{equation}
and the action is
\begin{equation}
S=\frac{1}{16\pi}\int\! d^4\!x\sqrt{-g}\,{\cal R}=\int\!dt\,dr\,[\pi_{R}\dot{R}
+\pi_{L}\dot{L}-N^t{\cal H}_{t}-N^r{\cal H}_r]
\end{equation}
with
\begin{equation}
{\cal H}_t=\frac{L\pi_{L}^{2}}{2R^2}-\frac{\pi_{L}\pi_{R}}{R}+\left(
\frac{RR'}{L}\right)'-\frac{R'^{2}}{2L}-\frac{L}{2} \mbox{ \ \ \ \ ;
\ \ \ \ } {\cal H}_r=R'\pi_{R}-L\pi_{L}'
\label{theconstraints}
\end{equation}
where $'$ represents $d/dr$.
As before, the Hamiltonian is
\begin{equation}
H=\int\! dr[N^t{\cal H}_t+N^r{\cal H}_r] + M
\end{equation}
and the constraints are
\begin{equation}
{\cal H}_t={\cal H}_r=0.
\end{equation}
The constraints can be solved as follows. $\pi_R$ is eliminated by forming
the linear combination of constraints
$$
\frac{R'}{L}{\cal H}_t+\frac{\pi_{L}}{RL}{\cal H}_r=0.
$$
Defining
\begin{equation}
{\cal M}(r)\equiv \frac{\pi_{L}^{2}}{2R}+\frac{R}{2}-\frac{RR'^2}{2L^2},
\end{equation}
this constraint is equivalent to ${\cal M}'=0$. By comparing with
(\ref{surface}) it can be shown that ${\cal M}(\infty)$ is the ADM mass.
The constraints can now be solved for the momenta:
\begin{equation}
\pi_{L}=\eta R \sqrt{(R'/L)^2-1+2M/R} \mbox{ \ \ \ \ ; \ \ \ \ }
\pi_{R}=\frac{L}{R'}\pi_{L}'
\label{momsolve}
\end{equation}
where $\eta=\pm 1$.
Since we have defined the new variable, ${\cal M}(r)$, it is natural to ask
what the momentum conjugate to ${\cal M}(r)$ is. From the fundamental
Poisson brackets,
\begin{equation}
\{L(r),\pi_{L}(r')\}=\{R(r),\pi_{R}(r')\}=\delta(r-r')
\end{equation}
it is straightforward to check that
\begin{equation}
\pi_{{\cal M}}\equiv \frac{L\sqrt{(R'/L)^2-1+2{\cal M}/R}}{1-2{\cal M}/R}
\end{equation}
satisfies
\begin{equation}
\{{\cal M}(r),\pi_{{\cal M}}(r')\}=\delta(r-r').
\end{equation}
$\pi_{{\cal M}}$ has a simple physical interpretation. Consider the geometry
in Schwarzschild coordinates:
\begin{equation}
ds^2=-(1-2M/r_s)dt_{s}^2+\frac{dr_{s}^{2}}{1-2M/r_s}+r_{s}^{2}(d\theta^2
+\sin^{2}\theta d\phi^2)
\end{equation}
and then transform to new coordinates $t(t_s,r_s),\: r(t_s,r_s)$. One
finds that in the new coordinates,
\begin{equation}
R=r_s(t,r) \mbox{ \ \ \ \ ; \ \ \ \ }
L^2=\frac{1}{1-2M/r_s}\left(\frac{\partial r_s}{\partial r}\right)^2
-(1-2M/r_s)\left(\frac{\partial t_s}{\partial r}\right)^2
\end{equation}
which then yields
\begin{equation}
\pi_{{\cal M}}(r)=\frac{\partial t_s}{\partial r}.
\end{equation}
In other words, $\pi_{{\cal M}}$ measures the rate at which Schwarzschild time
changes along a hypersurface. $P_M \equiv \int\pi_{{\cal M}}(r)dr$ is conjugate
to the ADM mass: $\{M,P_M\}=1$. This gives clear expression to the often heard
statement that the energy generates time translations at infinity.
Since
$P_M$ is related to the behavior at infinity, it is not surprising that it is
invariant only under small diffeomorphisms:
$$
\delta P_M =\left\{\int\! dr[f(r){\cal H}_t+g(r){\cal H}_r]\,,P_M\right\}
$$
vanishes provided
$$
\left.
f(\infty), g(\infty) \le \frac{\pi_L}{RL}\right|_{r=\infty}
$$
as can be checked explicitly. For most coordinate choices, the inequality
reduces to $f(\infty)=g(\infty)=0$.
\section{Quantization}
Once the Hamiltonian formulation of gravity has been developed, the
transition to the quantum theory is quite straightforward. This is true
only in a formal sense, though,
since operator ordering ambiguities will be left
unresolved, as will the problem of ultraviolet divergences. Furthermore,
the interpretation
of the resulting theory is not at all clear, as we will briefly discuss.
More detailed discussion of interpretational issues can be found in
\cite{unwald}.
To quantize, we realize the phase space variables $h_{ij}$, $\pi_{ij}$ as
operators satisfying the commutation relations
\begin{equation}
[h_{ij}(x^i),\pi^{mn}(x'^i)]=i\delta^{m}_{i}\delta^{n}_{j}\delta^{3}(x^i-x'^i)
\end{equation}
We therefore make the replacements,
$$
\pi^{ij} \rightarrow -i\frac{\delta}{\delta h_{ij}}.
$$
The states of the system are then described by functions of the three-geometry,
{\em i.e.} by the wave functionals $\Psi[h_{ij}(x^i)]$. Physical states are
required to be annihilated by the constraints:
\begin{equation}
{\cal H}^i \Psi[h_{ij}]=2i\left(\frac{\delta}{\delta h_{ij}}\right)_{|j}
\Psi[h_{ij}]=0
\end{equation}
\begin{equation}
{\cal H}_{t} \Psi[h_{ij}]=-\left[\frac{8\pi}{m_{p}^2}h^{\frac{1}{2}}G_{ijkl}
\frac{\delta}{\delta h_{ij}}\frac{\delta}{\delta h_{kl}}+\frac{m_{p}^{2}}
{16\pi}h^{\frac{1}{2}} {\,^{3}{\cal R}}\right]\Psi[h_{ij}]=0
\label{WD}
\end{equation}
where $G_{ijkl}=h_{ik}h_{jl}+h_{il}h_{jk}-h_{ij}h_{kl}$, and we have restored
the Planck mass. The first set of constraints simply says that physical states
should be invariant under reparameterizing the spacelike hypersurfaces; they
are analogous to Gauss' law as they are linear in derivatives. The final
constraint is quadratic in derivatives and is known as the Wheeler-Dewitt
equation. In asymptotically flat space, the wavefunction also satisfies a
non-trivial Schrodinger equation:
\begin{equation}
H\Psi[h_{ij},t]=-i\frac{\partial \Psi[h_{ij},t]}{\partial t}.
\end{equation}
It is then assumed that $\Psi^{*}[h_{ij}]\Psi[h_{ij}]$ can somehow be
interpreted as a probability density over three-geometries.
Even if we knew the correct operator ordering prescription, the full
Wheeler-Dewitt equation would still be much too difficult to solve. However,
it can be simplified greatly by employing the WKB approximation, which amounts
to taking $m_{p}\rightarrow \infty$. In the WKB approximation the
wavefunction is written in the form
\begin{equation}
\Psi[h_{ij}]=e^{im_{p}^{2}S[h_{ij}]}.
\end{equation}
Inserting this expression into (\ref{WD}) and keeping only terms of order
$m_{p}^{2}$, we find that $S$ satisfies:
\begin{equation}
16\pi G_{ijkl}\frac{\delta S}{\delta h_{ij}}\frac{\delta S}{\delta h_{kl}}
-\frac{1}{16\pi}h^{\frac{1}{2}} {\,^{3}{\cal R}}=0.
\end{equation}
This is the Einstein-Hamilton-Jacobi equation, which means that $S[h_{ij}]$
is the classical action associated with a solution to Einsteins's equations.
Specifically, given some solution $g_{\mu \nu}(x^{\mu})$, $S[h_{ij}]$ is
found by integrating $S_{G}=\frac{1}{16\pi}\int\!d^4\!x\sqrt{-g}\,{\cal R}$ in
the region between some reference hypersurface and the hypersurface
$h_{ij}(x^i)$. Thus the WKB approximation reduces the task of solving a
functional differential equation to solving a set of partial differential
equations.
As we remarked earlier, it is difficult to interpret the wavefunction
$\Psi[h_{ij}]$. In particular, what does it mean to say that $|\Psi|^2$ is the
probability to find some three-geometry? It may well be that the wavefunction
only has meaning once matter is included. Then, writing $\Psi[h_{ij},\phi]$,
we would interpret
$$
\frac{|\Psi[h_{ij},\phi]|^2}{\int\! D\!\phi |\Psi[h_{ij},\phi]|^2}
$$
as the relative probability of finding various matter configurations on the
hypersurface $h_{ij}$. We will return to this intepretation in later sections
when we couple a scalar field to gravity.
\subsection{Spherical Symmetry}
In the spherically symmetric case we have the operators $L$, $R$ and
$\pi_{L}=-i\frac{\delta}{\delta L}$, $\pi_R=-i\frac{\delta}{\delta R}$, and
the wavefunction, $\Psi[L,R]$, which satisfies the constraints
\begin{equation}
{\cal H}_{t}\Psi[L,R]={\cal H}_{r}\Psi[L,R]=0.
\end{equation}
In this simplified theory the quadratic constraint is still too difficult to
solve so we again employ the WKB approximation and write
\begin{equation}
\Psi[L,R]=e^{iS[L,R]}.
\end{equation}
Since the WKB approximation amounts to the replacements $\pi_{L}\rightarrow
\frac{\delta S}{\delta L}$, $\pi_{R}\rightarrow \frac{\delta S}{\delta R}$,
and using (\ref{momsolve}), $S$ is found to satisfy
$$
\frac{\delta S}{\delta L} =\eta R\sqrt{(R'/L)^2-1+2M/R}
$$
\begin{equation}
\frac{\delta S}{\delta R}=\frac{L}{R'}\frac{d}{dr}\left(\eta R
\sqrt{(R'/L)^2-1+2M/R}\right).
\end{equation}
These expressions are most easily integrated as follows. Start from some
arbitrary geometry $L$, $R$, and vary $L$ while holding $R$ fixed until
$(R'/L)^2-1+2M/R=0$. The contribution to $S$ from this variation is
$$
S[L,R]=\eta R\int\! dr\, dL \sqrt{(R'/L)^2-1+2M/R}
$$
$$
=\eta\int\! dr \left[RL\sqrt{(R'/L)^2-1+2M/R}\right.
$$
\begin{equation}
\left. \mbox{\ \ \ \ } +RR' \log{\left|\frac{R'/L
-\sqrt{(R'/L)^2-1+2M/R}}{\sqrt{|1-2M/R|}}\right|}\,\right].
\label{Ssolve}
\end{equation}
Now vary $L$, $R$, while keeping $(R'/L)^2-1+2M/R=0$, to some set geometry.
There is no contribution to $S$ from this variation, so (\ref{Ssolve}) is the
complete solution. The wavefunction, $\Psi$, either oscillates or decays
exponentially depending on the sign of $(R'/L)^2-1+2M/R$. This is also the
condition which determines whether the hypersurface $L$, $R$, can be embedded
in the classical Schwarzschild geometry.
The description of the state in terms of the wavefunction $\Psi[L,R]$ is, of
course, highly redundant since, according to the constraints, the wavefunction
must take the same value for many different $L$'s and $R$'s. If we wish to
eliminate this redundancy we should describe the state in terms of physical
observables, {\em i.e.} those which commute with the constraints. In the
present case there are only two such observables \cite{thiemann,kuchar}
: the ADM mass, M, and its
canonical momentum $P_M$. To satisfy the commutation relation $[M,P_M]=i$
we make the replacement $M=i\partial/\partial P_M$, so that a state of
definite mass is written
\begin{equation}
\Psi_M(P_M)=e^{-iMP_M}.
\end{equation}
If the time dependence is also included we have
\begin{equation}
\Psi_M(P_M,t)=e^{-iM(P_M+t)}.
\end{equation}
It is seen that $P_M$ functions as an intrinsic time variable.
The emergence of an intrinsic time in this system is quite interesting in
light of the so-called ``problem of time'' in quantum gravity. In the
asymptotically flat case this problem is relatively benign, since the
wavefunction depends on the time $t$, which can be measured by clocks at
infinity. However, for a closed universe there is no such asymptotic region
and no obvious variable to play the role of time. This makes the wavefunction
of a closed universe especially difficult to interpret. In quantum mechanics,
we interpret the wavefunction $\psi(x,t)$ by saying that at fixed time $t$ the
probability to find $x$ is $|\psi(x,t)|^2$. To carry this procedure over to
the quantum theory of a closed universe we need to specify a variable which
can be fixed in order to determine probabilities. In the spherically
symmetric, asymptotically flat case, $P_M$, like $t$, can play this role, but
no such variable is known in general.
\pagebreak
This completes our discussion of the
quantization of the pure gravity theory.
In what follows
we turn to the effects of incorporating matter into the
system.
\chapter{Quantum Field Theory Near Black Holes}
\section{A Useful Coordinate System}
Calculations of physical effects in the presence of black holes are greatly
simplified by employing appropriate coordinates. In this section, we will
describe the properties of a little known set of coordinates
\cite{KW2} which will be
used repeatedly in the material that follows.
Schwarzschild found his remarkable exact solution for
the geometry outside a star in general relativity
quite soon after Einstein
derived the field equations. Further
study of this geometry over the course of several decades
revealed a series of surprises: the existence and physical
relevance of pure vacuum ``black hole'' solutions; the
incompleteness of the space-time covered by the
original Schwarzschild coordinates, and the highly
non-trivial global structure of its completion; and the
dynamic nature of the physics in this geometry despite its
static mathematical form, revealed perhaps most dramatically
by
the Hawking radiance \cite{hawking}. Discussions
of this material can now be found
in advanced
textbooks \cite{birrel}, but they are hardly limpid.
In the course of investigating an improvement to the standard
calculation of this radiance to take into account its
self-gravity, as we detail in later sections,
we came upon a remarkably simple form for the line element of
Schwarzschild (and Reissner-Nordstrom) geometry.
This line element has an interesting history~
\cite{israel}, but
as far as we know it has never been discussed from a modern
point of view.
We have found
that several of the more subtle features of the geometry become
especially easy
to see when this line element is used.
To motivate the form of the coordinates we reconsider the constraint
equations of spherically symmetric gravity (\ref{theconstraints}):
$$
{\cal H}_t =\frac{\pi_{L}^{2}}{2R^2}-\frac{\pi_L \pi_R}{R} +\left(\frac{
RR'}{L}\right)'-\frac{R'^2}{2L}=0,
$$
\begin{equation}
{\cal H}_r=R'\pi_R -L\pi'=0.
\end{equation}
The canonical momenta are given by
\begin{equation}
\pi_L=\frac{N^rR'}{N^t}-\frac{R\dot{R}}{N^t} \mbox{ \ \ \ \ ; \ \ \ \ }
\pi_R=\frac{(N^rLR)'}{N^t}-\frac{\dot{(LR)}}{N^t}.
\end{equation}
We can arrive at a particular set of coordinates by choosing a gauge and
solving the constraints.
Our choice is simply $L=1$, $R=r$. With this choice the
equations simplify drastically, and
one easily solves to find
$\pi_L = \sqrt {2Mr }~,~ \pi_R = \sqrt {M\over 2r}$
and then $N^t = \pm 1,~ N^r = \pm \sqrt {2M\over r}$.
Thus for the line element we have
\begin{equation}
ds^2~=~ -dt^2 + (dr \pm \sqrt {2M \over r} dt )^2 +
r^2 (d\theta^2 + \sin^2 \theta d\phi^2 )~.
\label{nicemetric}
\end{equation}
$M$, which appears as an integration constant, of course
is to be interpreted as the mass of the black hole described by
this line element.
For the Reissner-Nordstrom geometry, the same gauge choice leads
to a metric of the same form, with the only change that
$2M \rightarrow 2M - Q^2/r$.
These line elements are stationary -- that is, invariant under
translation of $t$,
but not static -- that is, invariant under reversal
of the sign of
$t$. Indeed reversal of this sign interchanges the $\pm$
in (\ref{nicemetric}), a feature we will interpret further below.
Another peculiar feature is that each
constant time slice
$dt=0$ is simply flat Euclidean space!
We can obtain a physical interpretation of these coordinates by
comparing them to those of Lemaitre ~\cite{lemaitre}, in terms of
which the line element reads
\begin{equation}
ds^2~=~-d\tau^2 + {(2M)^{2/3} \over [{3 \over 2} (r_{L} +
\tau)]^{2/3}}~dr_{L}^2~-~(2M)^{2/3} [{3 \over 2}(r_{L} -\tau)]^{4/3}
(d\theta^2 + \sin^2{\theta} d\phi^2).
\label{lemaitre}
\end{equation}
As the Lemaitre coordinates are synchronous ($g_{\tau \tau}=-1$ ,
$g_{\tau \rm i}=0$),~ a class of timelike geodesics is given by
motion along the time lines ($r_{L},\theta, \phi =$constant), and
the proper time along the geodesics is given by the coordinate
$\tau$. To arrive at (\ref{lemaitre}) we retain the Lemaitre time
coordinate, but now demand that the radial coordinate squared be
the coefficient multiplying $d\theta^2 + \sin^{2} \theta d\phi^2$.
In other words, we write
\begin{equation}
t=\tau\mbox{ \ \ \ \ ; \ \ \ \ } r=(2M)^{1/3}[{3\over 2}(r_{L} -
\tau)]^{2/3}.
\label{coords}
\end{equation}
A simple calculation then leads from (\ref{lemaitre}) to (\ref{nicemetric})
with the upper choice of sign; the lower choice is obtained by repeating
the same steps starting from Lemaitre coordinates with the sign of
$\rm \tau$ reversed.
Finally, let us note that these coordinates are related to
Schwarzschild coordinates,
\begin{equation}
ds^{2}=-(1-\frac{2M}{r})dt_{s}^{2} + \frac{dr^{2}}{1-\frac{2M}{r}} +
r^{2}(d\theta^{2}
+\sin{\theta}^{2} d\phi^{2})
\end{equation}
by a change of time slicing,
\begin{equation}
t_{s}=t-2\sqrt{2Mr} - 2M \log{\left[\frac{\sqrt{r}-\sqrt{2M}}{\sqrt{r}+
\sqrt{2M}}\right]}.
\label{tdef}
\end{equation}
In contrast to the surfaces of constant $t_{s}$, the constant $t$ surfaces pass
smoothly through the horizon and extend to the future singularity
free of coordinate singularities.
In terms of $r$ and $t$, the radially ingoing and outgoing null
geodesics are given by
$$
\hspace{-10mm}\mbox{ingoing: \ \ \ } t+r-2\sqrt{2Mr}+4M\log[{\sqrt{r}+\sqrt{2M}]}
=v = \mbox{ constant}
$$
\begin{equation}
\mbox{outgoing: \ \ \ } t-r-2\sqrt{2Mr} -4M\log{[\sqrt{r}-\sqrt{2M}]}=u
=\mbox{ constant }.
\label{geo}
\end{equation}
\subsection{Global Structure}
Now let us discuss the global properties of our coordinate system.
Perhaps the clearest approach to such questions is via
consideration of the properties
of light rays. Taking for definiteness the upper sign in (\ref{nicemetric}),
and without any essential
loss of generality restricting to the case
$d\theta = d\phi =0$ appropriate to the $\theta = \pi/2$ sections,
we find that $ds^2=0$ when
\begin{equation}
{dr\over dt } ~=~ - \sqrt {2M\over r } \mp 1~.
\label{lightrays}
\end{equation}
For the class of light rays governed by
the upper sign, we can cover the
entire range $0 < r < \infty $ as $t$ varies. In particular one
meets
no obstruction, nor any special structure,
at the horizon $r=2M$. For the class of light rays
governed by the lower sign there is structure at $r=2M$.
When $r > 2M$ one has a positive slope for ${dr\over dt}$, and
$r$ ranges over $2M < r < \infty$. When $r < 2M$ one has a negative
slope for ${dr\over dt}$, and $r$ ranges over $0 < r < 2M$.
When $r = 2M$ it does not vary with $t$. From these properties, one
infers that our light rays cover regions I and II in the Penrose
diagram, as displayed in Fig. ~\ref{metfig}. Let us emphasize that the
properties of the Penrose diagram can be {\em inferred} from the
properties of the light rays, although we will not belabor that
point here.
\renewcommand\floatpagefraction{.9}
\renewcommand\topfraction{.9}
\renewcommand\bottomfraction{.9}
\renewcommand\textfraction{.1}
\begin{figure}[htb]
\centerline{\psfig{file=metfig.epsi,width=\hsize,angle=-90}}
\caption{Penrose diagram for the Schwarzschild geometry. As described in the
text, $r$ and $t$ in one coordinate patch, for the upper sign of the line
element, cover regions I and II. As is clear from the diagram, the ingoing
light rays are captured in their entirety (into the singularity), whereas the
outgoing light rays cannot be traced back past the horizon.}
\label{metfig}
\end{figure}
If one chooses instead the lower sign in (\ref{nicemetric}) , and performs
a similar analysis, one finds that regions I and II$^\prime$
are covered.
Patching these together with the sectors found previously,
one still does not have a complete space-time.
However our line element is not yet exhausted. For in drawing
Figure 1 we have implicitly assumed that $t$ increases along light
rays which point up (``towards the future''). Logically, and to
maintain symmetry, one should
consider also the opposite case, that the coordinate
$t$ increases towards the past. By doing this, one generates
coordinate systems covering regions
I$^\prime$ and II$^\prime$
respectively I$^\prime$ and II$^\prime$,
for the upper and lower signs
in (\ref{nicemetric}) . Thus the complete Penrose diagram is covered
with patches each governed by a stationary -- but not static --
metric, and with non-trivial regions of overlap.
In the Reissner-Nordstrom case the generalization of (\ref{nicemetric})
has
a coordinate singularity at $r=Q^2/2M$. However this singularity
is inside both horizons, and does not pose a serious obstruction to
a global analysis. One obtains the complete Penrose diagram also in
this case by
iterating
constructions similar to those just sketched.
The usual
Schwarzschild line element appears to be time reversal symmetric,
but when the global structure of the space-time it defines is taken into
account one sees that this appearance is misleading. The fully extended
light-rays in Figure 1 go from empty space to a singularity as $t$
advances (they pass from region I into region II),
which is definitely distinguishable from the
reverse process. There is a
symmetry which relates these to the corresponding rays going from
region II$^\prime$ to region I$^\prime$, however it involves not merely
changing the sign of
$t$ in the Schwarzschild metric,
but rather going to a completely disjoint
region of the space-time. This actual symmetry of the space-time is if
anything more obvious in our construction than in the standard one.
Thus by taking the line-element
in region I stationary rather than static
we have lost some false symmetry while making the true symmetry --
and its
necessary
connection with the existence of
region I$^\prime$ (constructed, as we have seen, by simultaneously
reversing the
sign of $t$ {\it and\/} interchanging the future with the past) --
more obvious.
\section{Quantum Field Theory in Curved Space}
In this section we briefly present some of the important results in the
theory of quantum fields in curved space. Our goal is primarily to establish
notation and to write down some formulas which will be referred to in later
sections. Comprehensive reviews of the subject can be found in
\cite{birrel,fulling}.
Quantum field theory in curved space is a hybrid of quantum and classical
field theory, which one hopes reliably describes the behavior of matter in
regions of relatively low space-time curvature. The approach is to couple
a quantum field to a metric tensor, which is treated as a classical variable.
For simplicity, we will consider a massless scalar field with action
\begin{equation}
S=-\frac{1}{2}\int\! d^4\! x\sqrt{-g}\,g^{\mu \nu}
\partial_{\mu}\phi\, \partial_{\nu}\phi.
\end{equation}
The field satisfies the wave equation,
\begin{equation}
(\sqrt{-g}\,g^{\mu \nu}\phi_{,\mu})_{,\nu}=0
\end{equation}
Associated with this wave equation is the conserved inner product
\begin{equation}
(\phi_1,\phi_2)\equiv -i\int_{\Sigma}d^3\! x\sqrt{h}\,n^{\mu}\phi_1(x^i,t)
\stackrel{\leftrightarrow}{\partial_{\mu}}\phi_{2}^{*}(x^i,t)
\end{equation}
where $\phi_1$ and $\phi_2$ are two solutions of the wave equation. Here
$\Sigma$ is a Cauchy surface, and $n^{\mu}$ is a future pointing unit vector
normal to $\Sigma$. In order to second quantize the field we must write down
a complete set of ``positive frequency'' solutions, $u_{i}(x^{\mu})$, which
satisfy $(u_i,u_j)=\delta_{ij}$. This is where the ambiguity in the
quantization process occurs, since there are many different sets of $u_i$'s,
which will be shown to lead to inequivalent quantizations. Once we have chosen
a set, the field operator is written in terms of creation and
annihilation operators:
\begin{equation}
\phi=\sum_{i}\,[a_i u_i+a_{i}^{\dagger} u_{i}^{*}]
\label{phiu}
\end{equation}
where $a$, $a^{\dagger}$ obey
\begin{equation}
[a_i,a_{j}^{\dagger}]=\delta_{ij}.
\end{equation}
The vacuum state is defined by $a_i |0_u \rangle=0$, and particle states
are created by applying $a^{\dagger}$'s to $|0_u \rangle$.
On the other hand, suppose that instead of $u_i$ we had chosen the modes
$v_i$ in which to expand the field:
\begin{equation}
\phi=\sum_{i}\,[b_i v_i +b_{i}^{\dagger}v_{i}^{*}].
\label{phiv}
\end{equation}
Then the vacuum would be defined by $b_i |0_v \rangle =0$, and particle
states would be created by applying $b^{\dagger}$'s to $|0_v \rangle$.
The obvious question which arises is: what is the relation between the states
defined by $u_i$ and those defined by $v_i$. To answer this, we
note that by equating (\ref{phiu}) and (\ref{phiv}), and taking inner products,
we obtain the ``Bogoliubov transformation'':
\begin{equation}
b_j = \sum_{j}(\alpha_{ji}^{*} a_{i}-\beta_{ji}^{*} a_{i}^{\dagger})
\end{equation}
where
\begin{equation}
\alpha_{ij}=(v_i,u_j) \mbox{ \ \ \ \ ; \ \ \ \ } \beta_{ij}=-(v_i,u_{j}^{*})
\end{equation}
The Bogoliubov coefficients obey the completeness relations,
$$
\sum_{k}(\alpha_{ik}\alpha_{jk}^{*}-\beta_{ik}\beta_{jk}^{*})=\delta_{ij}
$$
\begin{equation}
\sum_{k}(\alpha_{ik}\beta_{jk}-\beta_{ik}\alpha_{jk})=0.
\label{complete}
\end{equation}
It can then be shown that the states are related by
\begin{equation}
|\psi_u \rangle =C :\!\exp{[\frac{1}{2}a(\alpha^{-1}\beta)a+a(\alpha^{-1}
-1)a^{\dagger}+
\frac{1}{2}a^{\dagger}(-\beta^{*}\alpha^{-1})a^{\dagger}]}\!:|\psi_v
\rangle
\end{equation}
where $C$ is a constant. The average number of $v$ particles in the $u$
vacuum is
\begin{equation}
\langle 0_u |b_{i}^{\dagger}b_{i}|0_{u}\rangle=\sum_{j}|\beta_{ji}|^2.
\end{equation}
\section{Black Hole Radiance}
Now we turn to the quantization of a massless scalar field in the presence of
collapsing matter. The goal is to compute the flux of particles on
${\cal J}^+$ given some initial state defined on ${\cal J}^-$. To completely
determine this flux on all of ${\cal J}^+$, one would have to solve
the wave equation
mode by mode in the region bounded by ${\cal J}^-$ and $\Sigma$,
where $\Sigma$ is a constant $t$ hypersurface which
crosses the horizon to the future of the collapsing matter.
This is clearly intractable, since the geometry
inside the collapsing matter may be very complicated, and the scalar field
might interact with the matter in an arbitrarily complex fashion. Fortunately,
all of the dependence of the particle flux on ${\cal J}^+$ on these factors
dies out at sufficiently late times, and the radiation becomes, at least to
leading order, completely independent of the details of the collapse process.
Thus we shift our goal to calculating this late-time radiation and to showing
that it is indeed universal.
The first important observation is that the geometry is entirely smooth as long
as one stays sufficiently far from the singularity at $r=0$. In particular,
for a black hole much larger than $m_p$, the geometry is smooth near the
horizon --- in fact, the curvature can be made arbitrarily small by making the
black hole arbitrarily massive. This strongly suggests the conlusion that
however complicated the state of the field is after propagating through the
matter, it should certainly appear nonsingular to inertial observers near the
horizon. Of course, this presupposes that the initial state on ${\cal J}^-$
is nonsingular.
Next, we examine the outgoing null geodesics in the region exterior to the
collapsing matter. These obey:
\begin{equation}
u \equiv t-2\sqrt{2Mr}-r-4M\log{(\sqrt{r/2M}-1)}=\mbox{constant}.
\end{equation}
Let us consider two geodesics, labelled by $u_1$ and $u_2$, which are
separated in time by $\Delta u= u_2 - u_1$. On a constant $t$
surface, $\Sigma$, the geodesics have a radial separation given by
\begin{equation}
\Delta u=r_1 -r_2 +2(\sqrt{2Mr_1}-\sqrt{2Mr_2})+4M\log{\left( \frac{\sqrt{r_1}
-\sqrt{2M}}{\sqrt{r_2}-\sqrt{2m}}\right)}.
\end{equation}
For large $u_1$, $u_2$, which corresponds to geodesics which reach ${\cal J}^+$
at late times, the radial separation of the geodesics near the horizon is
determined by
\begin{equation}
\Delta u \approx 4M\log{\left(\frac{\sqrt{r_1}-\sqrt{2M}}{\sqrt{r_2}-\sqrt{2M}
}\right)}
\end{equation}
or
\begin{equation}
r_2 - 2M \approx (r_1 -2M) e^{-\Delta u/4M}.
\end{equation}
In other words, the outoging geodesics pile up along the horizon.
Alternatively, we can note that all the geodesics which reach ${\cal J}^+$
after $u=\hat{u}$, where $\hat{u}\gg M$, were contained on the surface $t=0$ in
the region between $r=2M$ and $r=2M+4Me^{-\hat{u}/4M}$.
This has two important consequences. First, we see that the late time
radiation is entirely determined by the state of the field at distances
arbitrarily close to the horizon. Second, an outgoing wave suffers an
arbitrarily large redshift in escaping from near the horizon to ${\cal J}^+$.
So, putting it all together, we see that to compute the late-time radiation
we need only consider nonsingular states in a region near the horizon.
It is easy to construct a nonsingular state if we first define a new time
coordinate. The trouble with the coordinate $t$ used in (\ref{nicemetric})
is that its
flow is spacelike inside the horizon. However, this can be rectified if we
define a new time, $\tau$, as the value of $t$ along the curve
$dr + \sqrt{2M/r}dt=0$. $\tau$ has the virtue of being nonsingular and
timelike. We can choose a set of modes which are positive frequency with
respect to $\tau$, and expand the field in terms of them. Then, by the
equivalence principle and standard quantum field theory in flat space, the
corresponding vacuum will have a nonsingular stress-energy as seen by inertial
observers. Of course, this choice of state is not unique since other
nonsingular states can be obtained by applying particle creation operators to
this state. This is irrelevant if one is only interested in late-time
radiation, since the excitations above the vacuum will eventually redshift
away.
Now at spatial infinity (more accurately: conformal infinity
${\cal I}_+$) the vacuum state is defined
locally by the requirement that
modes having positive frequency with respect to the variable
$u = t_{\rm s} -r_*$ are unoccupied, where $t_{\rm s}$
is Schwarzschild
time and $r_* = r + 2M\ln (r -2M)$ is the tortoise coordinate.
We wish to find the relationship between this
requirement and the preceding one.
The relationship between $t$ and $t_{\rm s}$ is
\begin{equation}
t = t_{\rm s} + 2 \sqrt{2Mr} +
2M \ln {{\sqrt r - \sqrt{2M} } \over {\sqrt r + \sqrt{2M} }}
\label{times}
\end{equation}
so that
\begin{equation}
u = t_{\rm s} - r_* =
t - 2 \sqrt{2Mr} -r -4M \ln (\sqrt r - \sqrt{2M} )~,
\label{uexpression}
\end{equation}
and thus one finds that along a curve with $dr+ \sqrt {2M\over r} dt =0$,
\begin{equation}
{du\over dt}~=~ 2 + \sqrt {2M\over r} + {2M \over r - \sqrt{2Mr} }~.
\label{frequencyfactor}
\end{equation}
Because the last term on the right hand side
is singular, the two definitions of positive
frequency -- with respect to $u$ or to $t$ -- do not coincide.
To remove the singularity, note that along any of
the curves of interest
$e^{-u/4M}$ has a simple zero at $r=2M$, but is otherwise positive.
Clearly then demanding positive frequency with respect to
$t$ along such curves requires positive frequency not with respect
to $u$ but rather with respect to
\begin{equation}
U = - e^{-u/4M}~.
\label{kruskalU}
\end{equation}
In this way we have arrived at the famous Unruh
boundary conditions \cite{unruh}.
To summarize, the appropriate construction near the horizon is to expand the
field as
\begin{equation}
\phi= \int\! \frac{d\omega}{\sqrt{2\pi 2\omega}}\,[b_{\omega}e^{i\omega U}
+b_{\omega}^{\dagger}e^{-i\omega U}]
\end{equation}
and take the state resulting from collapse to be $|0_U\rangle$, where
$b_{\omega}|0_{U}\rangle =0$. For simplicity, we have suppressed the
field's dependence on the ingoing modes. It remains to describe this state in
terms of particles defined by the modes $e^{-i\omega u}$. Since $u \rightarrow
\infty$ at the horizon, we need another set of modes in which to expand the
field
inside the horizon. These will be referred to as $u_{\omega}^{in}(u)$; their
explicit form will not be needed. We then have
\begin{equation}
\phi = \int\! \frac{d\omega}{\sqrt{2\pi 2\omega}}\,[(a_{\omega}e^{-i\omega u}
+a_{\omega}^{\dagger} e^{i\omega u})\,\Theta\!\left(-U\right)
+\left(a_{\omega}^{in}u_{\omega}^{in}+a_{\omega}^{in \dagger}
u_{\omega}^{in *}\right)\Theta\! \left(U\right)].
\end{equation}
The relation between the modes for $U<0$ is given by
\begin{equation}
e^{i\omega' U(u)}=\int\! \frac{d \omega}{2\pi}\,[\alpha_{\omega' \omega}
e^{-i\omega u}+\beta_{\omega' \omega}e^{i\omega u}] \mbox{ \ \ \ ; \ \ \ }
U<0
\end{equation}
so
\begin{equation}
\alpha_{\omega' \omega}=\int_{-\infty}^{\infty}\!du\,
e^{i(\omega' U(u)+\omega u)}
\mbox{ \ \ \ ; \ \ \ } \beta_{\omega' \omega}=\int_{-\infty}^{ \infty}\!du\,
e^{i(\omega'U(u)-\omega u)},
\end{equation}
or, since $U(u)=-e^{-u/4M}$,
\begin{equation}
\alpha_{\omega' \omega}=4M\, \frac{\omega^{4\pi i \omega}}{\omega'}\,
e^{2\pi M\omega}\, \Gamma(1-4iM \omega) \mbox{ \ \ ; \ \ }
\beta_{\omega' \omega}=4M\, \frac{\omega^{-4\pi i \omega}}{\omega'}\,
e^{-2\pi M \omega}\, \Gamma(1+4iM \omega).
\end{equation}
So
\begin{equation}
|\alpha_{\omega' \omega}|^2= \frac{(4M)^3 \pi \omega}{\omega'^2}
\frac{1}{1-e^{-8 \pi M\omega}} \mbox{ \ \ \ ; \ \ \ }
|\beta_{\omega' \omega}|^2 = \frac{(4M)^3 \pi \omega}{\omega'^2}
\frac{e^{-8\pi M \omega}}{1-e^{-8\pi M \omega}}.
\end{equation}
If we are only interested in the radiation which flows to infinity it is
appropriate to form a density matrix by tracing over the states inside the
horizon. To find the average number of particles radiated we should integrate
$|\beta_{\omega' \omega}|^2$. But because the black hole radiates for an
infinite amount of time at a constant rate (in the current approximation),
this yields infinity. To find the {\em rate} of emission,
we can place the hole in a
large box and use the density of states $d\omega/2\pi$ for outgoing particles.
For normalized modes the completeness relation, (\ref{complete}),
and the relation
\begin{equation}
|\alpha_{\omega' \omega}|^2=e^{8\pi M \omega}|\beta_{\omega' \omega}|^2
\end{equation}
imply
\begin{equation}
\int\! d\omega' |\beta_{\omega' \omega}|^2 = \frac{1}{e^{8\pi M \omega}-1}.
\end{equation}
The rate of emission of particles in the range $\omega$ to $\omega + d\omega$
is then:
\begin{equation}
F(\omega)=\frac{d\omega}{2\pi}\frac{1}{e^{8\pi M \omega}-1}.
\end{equation}
This is precisely the rate of emission from a black body, in one dimension,
at temperature $T=1/8\pi M$. $F$ does not quite give the flux seen at infinity
since some fraction, $1-\Gamma(\omega)$, of the particles will be reflected by
the spacetime curvature back into the hole. Thus, for the flux at infinity
we write,
\begin{equation}
F_{\infty}(\omega)=\frac{d\omega}{2\pi}\frac{\Gamma(\omega)}{e^{8\pi M \omega}
-1}.
\end{equation}
The preceding analysis shows that the average flux is that of a thermal body,
not that the full density matrix is thermal. That the density matrix is
exactly thermal is, in fact, easy to show by utilizing a clever trick due to
Unruh \cite{unruh}.
We have not chosen to use this method here, and instead have gone
through the rather laborious procedure of computing the Bogoliubov coefficients
directly, because only the latter method can be used when one wants to find
corrections to the spectrum.
\subsection{Reissner-Nordstrom}
A similar analysis can be carried out for the Reissner-Nordstrom geometry, the
metric for which is, in the $L=1$, $R=r$ gauge:
\begin{equation}
ds^2=-dt^2+(dr+\sqrt{2M/r-Q^2/r^2\,}\,dt)^2 +
r^2(d\theta^2+{\sin}^2\theta \,d\phi^2).
\label{rnmetric}
\end{equation}
The geometry has two horizons which are located at the two values of $r$ for
which $t$ goes from being timelike to spacelike, and vice versa. The outer
horizon radius is
\begin{equation}
R_+=M+\sqrt{M^2-Q^2}
\end{equation}
and the inner horizon radius is
\begin{equation}
R_- =M-\sqrt{M^2-Q^2}.
\end{equation}
It should be noted that the coordinates in (\ref{rnmetric}) fail to cover the
region inside the inner horizon. This will not pose any obstacle to determining
the radiance, though, as the calculation only depends in the geometry in the
vicinity of the outer horizon.
In the present case, the state resulting from collapse is specified by
demanding the absence of particles positive in frequency with respect to $t$
along the curve $dr+\sqrt{2M/r-Q^2/r^2}$. Now the calculation can proceed in
exact analogy to the uncharged case. The resulting temperature is:
\begin{equation}
T(M,Q)=\frac{1}{2\pi}\frac{\sqrt{M^2-Q^2}}{(M+\sqrt{M^2-Q^2})^2}.
\end{equation}
The temperature vanishes for the extremal black hole $M=Q$. If $Q>M$,
consideration of the global geometry reveals that there is no black hole at
all, but rather a naked singularity, which we will refer to as the
meta-extremal case. Finally, if the radiated particles are themselves charged,
the factor governing the emission probability is not the Boltzmann factor,
$e^{-\omega/T}$, but rather:
\begin{equation}
\exp{\left(-\frac{\omega-Qq/R_+(M,Q)}{T(M,Q)}\right)}
\end{equation}
where the particles carry charge q. The factor $Q/R_+$ is the electrostatic
potential at the outer horizon, and appears as a chemical potential causing
particles with the same sign of charge as the hole to be preferentially
emitted.
\subsection{Breakdown of the Semiclassical Approximation}
In the preceding treatment we have followed Hawking and considered the
propagation of free fields on a classical background geometry, ignoring the
back reaction altogether. It is usually assumed that the dominant effects of
the back reaction can be incorporated by allowing the black hole to lose mass
quasistatically, so that the radiation at infinity is thermal but with a
slowly varying temperature. Indeed, this is how a normal thermal body is
expected to behave during cooling. While this scenario seems quite plausible
for a wide range of configurations, both for black holes and normal objects,
there are notable instances where it is certain to be invalid, even
qualitatively \cite{psstw}.
In particular, if the emission of a single typical quantum
induces a relatively large change in temperature, a quasistatic description is
clearly inappropriate. Since a typical quantum has energy equal to $T$, the
temperature of the hole, we expect non-trivial back reaction effects to become
important when
\begin{equation}
T\frac{\partial T}{\partial M}\approx T.
\end{equation}
For a Schwarzschild black hole this happens when $M\approx m_p$, whereas for
a Reissner-Nordstrom hole it happens when $M\approx Q$. In the Schwarzschild
case, the breakdown occurs in a regime of high curvatures and large quantum
fluctuations, presumably inaccessible to semiclassical methods. Not so,
however, for a large near-extremal Reissner-Nordstrom hole, $M\approx Q\gg
m_p$, for which the curvature near the horizon remains small, suggesting that
meaningful calculations can be performed without the need for a full theory of
quantum gravity. In the next chapter
we perform such a calculation and see explicitly
that the radiance from the near extremal hole is markedly different from what
free field calculations indicate.
\section{Effect of Fluctuations}
The semiclassical approach to back reaction effects in quantum gravity is to
compute $<T_{\mu \nu}>$ for a field quantized on a classical background
geometry, and then to use this quantity on the right hand side of Einstein's
equations,
\begin{equation}
{\cal R}_{\mu \nu}-\frac{1}{2}g_{\mu \nu}{\cal R}=8\pi <T_{\mu \nu}>.
\label{semiclass}
\end{equation}
The (generally intractable) problem then becomes to compute $<T_{\mu \nu}>$
for an arbitrary metric. One expects this approach to provide a reasonable
description of the gross effects of back reaction provided that the
fluctuations in $T_{\mu \nu}$ are not too large. For example, replacing
$<T_{\mu \nu}>$ by $\sqrt{<T_{\mu \nu}\,^2>}$
should not lead to any substantial
change in the geometry if (\ref{semiclass}) is to be physically relevant.
In order to be quantitative, we shall consider the computation of stress
energy fluctuations in the moving mirror model, which is discussed in
\cite{Dav77,chung,hotta}. We consider this model here
because it provides a vivid demonstration of just how misleading the
semiclassical approximation can be, and because we will referring to it later
in another context. By ``moving mirror model'', we simply mean quantizing a
scalar field in two dimensions subject to the condition that the field vanish
along the timelike worldline $z_m(t)$. We will use the coordinates
\begin{equation}
u= t-z \mbox{ \ \ \ \ ; \ \ \ \ } v=t+z.
\end{equation}
Without the mirror present the propagator is
\begin{equation}
G(1,2)\equiv \langle 0|T[\phi(1)\phi(2)]|0\rangle
=\frac{1}{4\pi} \log{[(u_2-u_1)(v_2-v_1)]}
\label{minkprop}
\end{equation}
and for a mirror at rest, $z_m(t)=0$:
\begin{equation}
G(1,2)=\frac{1}{4\pi}\log{\left[\frac{(u_2-u_1)(v_2-v_1)}{(u_2-v_1)
(v_2-u_1)}\right]}.
\end{equation}
Now consider a general mirror trajectory, which we shall write in terms of
$u,v$ as $v_m(u)$. It is easiest to proceed by defining new coordinates in
terms of which the mirror is at rest. Working in the region to the right of
the mirror and defining
\begin{equation}
U(u)\equiv v_m(u) \mbox{ \ \ \ \ ; \ \ \ \ } V(v)\equiv v,
\end{equation}
we see that the trajectory $v=v_m(u)$ corresponds to $U=V$, which is the
desired result. In the new coordinates the metric is
\begin{equation}
ds^2=\left(U'(u)\right)^{-1}dU dV,
\end{equation}
and because of conformal invariance, the propagator has the same form as
before:
\begin{equation}
G(1,2)=\frac{1}{4\pi}\log{\left[\frac{\left(U(u_2)-U(u_1)\right)
\left(v_2-v_1\right)}{\left(U(u_2)-v_1\right)\left(v_2-v_1\right)}\right]}.
\label{mmprop}
\end{equation}
The basic object we will be considering is the renormalized propagator defined
by subtracting (\ref{minkprop}) from (\ref{mmprop}):
\begin{equation}
G_R(1,2)=\frac{1}{4\pi}\log{\left[\frac{U(u_2)-U(u_1)}{(u_2-u_1)(U(u_2)-v_1)
(v_2-U(u_1))}\right]}.
\end{equation}
This prescription simply corresponds to normal ordering with respect to
standard Minkowski space creation and annihilation operators. $<T_{\mu \nu}>$
can now be found by operating on $G_R(1,2)$ according to
\begin{equation}
T_{\mu \nu}=\partial_{\mu} \phi\, \partial_{\nu}\phi-\frac{1}{2}g_{\mu \nu}
\partial_{\rho}\phi\,\partial^{\rho}\phi,
\end{equation}
so
\begin{equation}
<T_{\mu \nu}>=\lim_{1\leftrightarrow 2}\,[\,
\frac{\partial}{\partial x_{1}^{\mu}}
\frac{\partial}{\partial x_{2}^{\nu}}-\frac{1}{2}g_{\mu \nu} g^{\alpha \rho}
\frac{\partial}{\partial x_{1}^{\rho}}\frac{\partial}{\partial x_{2}^{\alpha}}
\,]G_R(1,2).
\end{equation}
In the present case the only nonzero element is
\begin{equation}
<T_{uu}>=\lim_{1 \leftrightarrow 2}
\frac{\partial}{\partial u_1}\frac{\partial}
{\partial u_2} G_R(1,2)
\end{equation}
which, after some algebra, is found to be
\begin{equation}
<T_{uu}>=\frac{1}{12\pi}\sqrt{v_m(u)}\frac{d^2}{du^2}\sqrt{\frac{1}{v_m(u)}}.
\end{equation}
The result can be written in terms of the mirror trajectory $z_m(t)$ using
\begin{equation}
\frac{dv_m(u)}{du}=\frac{1+\dot{z_m}}{1-\dot{z_m}},
\end{equation}
yielding
\begin{equation}
<T_{uu}>=-\frac{1}{12\pi}\frac{(1-\dot{z_m}^2)^{\frac{1}{2}}}{(1-\dot{z_m})^2}
\frac{d}{dt}\left[\frac{\ddot{z_m}}{(1-\dot{z_m}^2)^{\frac{3}{2}}}\right].
\end{equation}
These expressions hold in the region to the right of the mirror. The
expressions to the left of the mirror are obtained by interchanging $u$ and
$v$.
These results can be used to find the radiation reaction force on the mirror.
The force four-vector is, by energy conservation:
\begin{equation}
F_{\mu}=-\Delta T_{\mu \nu}v^{\nu}
\end{equation}
where $\Delta T_{\mu \nu}=T_{\mu \nu}(+)-T_{\mu \nu}(-)$ is the difference in
$T_{\mu \nu}$ evaluated on the right and left hand sides of the mirror, and
$v^{\nu}$ is the mirror's velocity four-vector. We obtain:
\begin{equation}
F^{\mu}=\frac{1}{12\pi}\left(\frac{d^2v^{\mu}}{d\tau^2}-v^{\mu}v^{\nu}
\frac{d^2 v_{\nu}}{d\tau^2}\right)
\end{equation}
where $\tau$ is proper time. It is amusing to compare this to the
radiation reaction force on a point charge in classical electrodynamics,
\begin{equation}
F^{\mu}=\frac{2e^2}{3}\left(\frac{d^2v^{\mu}}{d\tau^2}-v^{\mu}v^{\nu}
\frac{d^2v_{\nu}}{d\tau^2}\right).
\end{equation}
This expression, containing as it does third derivatives of position of with
respect to time, has some well known peculiar features such as runaway
solutions and pre-acceleration. These carry over to the present case, although
we hasten to add that a realistic mirror has a high frequency cutoff
above which it becomes transparent; the worried reader need not be concerned
with the hazards of runaway mirrors.
$<T_{\mu \nu}T_{\alpha \beta}>$ is found by operating on the renormalized four
point function, which by Wick's theorem is:
\begin{equation}
G_R(1,2,3,4)=G_R(1,2)G_R(3.4)+G_R(1,3)G_R(2,4)+G_R(1,4)G_R(2,3).
\end{equation}
We find, in the region to the right of the mirror:
$$
<(T_{uu})^2>=3<T_{uu}>^2 \mbox{ \ \ \ \ ; \ \ \ \ } <(T_{vv})^2>=0
$$
\begin{equation}
<T_{uu}T_{vv}>=-\frac{1}{8\pi^2}\frac{(dv_m/du)^2}{[v-v_m(u)]^4}
\end{equation}
The fluctuations become infinite at the position of the mirror, $v=v_m(u)$.
To see how this might affect the motion of the mirror we can consider the
fluctuations in the radiation reaction force. Specifically, let us look at
the force normal to the mirror,
\begin{equation}
F_N=-F_{\mu}n^{\mu}
\end{equation}
where $n_{\mu}n^{\mu}=-1$ and $n_{\mu}v^{\mu}=0$. The mean squared value
is found to be:
$$
<F_{N}^{2}>=n^{\mu}v^{\nu}n^{\alpha}v^{\beta}[<T_{\mu \nu}(-)T_{\alpha \beta}
(-)>+<T_{\mu \nu}(+)T_{\alpha \beta}(+)>
$$
\begin{equation}
-2<T_{\alpha \beta}(-)><T_{\mu \nu}(+)>]\mbox{ \ \ \ \ }.
\end{equation}
This is clearly infinite; for instance, the second term contributes
\begin{equation}
\left.
n^u n^v v^u v^v <T_{uu}(+)T_{vv}(+)>=\frac{1}{8\pi^2}\frac{(dv_m/du)^2}
{[v-v_m(u)]^4}\right|_{v=v_m(u)}.
\end{equation}
The fluctuations diverge even for a stationary mirror.
In the spirit of the semiclassical approach we might have attempted to model
the back reaction by solving
\begin{equation}
ma_{\mu}=F_{\mu}.
\end{equation}
However, in the case of the moving mirror this can not be interpreted as the
leading term of a fully quantum mechanical treatment, since the fluctuations
diverge. Instead, it seems more likely that a fully quantum mechanical
treatment simply does not exist.
While these effects are most important for the moving mirror, there is no
reason to believe that fluctuations are so violent in the case of a black
hole much larger then the Planck mass. The point is simply that spacetime
is locally flat near the horizon on a scale much larger than the Planck
length. The divergences in the mirror model arose from the sharp boundary
condition at the mirror, a feature which has no analog in the black hole case.
\chapter{Self-Interaction Corrections}
Black hole radiance \cite{hawking} was originally derived in an approximation
where the background geometry was given, by calculating the
response of quantum fields to this (collapse) geometry. As we have seen,
in this
approximation the radiation is thermal, and much has been made both
of the supposed depth of this result and of the paradoxes that
ensue if it is taken literally. For if the radiation is accurately
thermal there is no connection between what went into the hole and
what comes out, a possibility which is difficult to reconcile with
unitary evolution in quantum theory -- or, more simply, with the
idea that there are equations uniquely connecting the past with the
future. To address such questions convincingly, one must
go beyond the approximation of treating the geometry as given, and
treat it too as a quantum variable. This is not easy, and as
far as we know no concrete correction to the original result
has previously
been derived in spite of much effort over more than twenty years.
Here we shall calculate what is plausibly the leading correction
to the emission rate of single particles in the limit of large
Schwarzschild holes, by a method that can be generalized in several
directions, as we shall outline.
There is a semi-trivial fact about the classic results
for black hole radiation,
that clearly prevents the radiation from being accurately
thermal. This is the effect that
the temperature
of the hole depends upon its mass, so that in calculating the
``thermal'' emission rate one must know what mass of the black hole
to use -- but the mass is different, before and after the radiation!
(Note that a rigorous identification of the
temperature of a hot body from its radiation,
can only be made for sufficiently high frequencies,
such that the
gray-body factors approach unity. But it is just in this limit
that the ambiguity mentioned above is most serious.) As we
have emphasized, this problem is particularly
quantitatively
acute for near-extremal holes --- it is a general problem for
bodies with finite heat capacity, and in the near-extremal
limit the heat capacity of the black hole vanishes.
To resolve the above-mentioned
ambiguity, one clearly must allow the geometry to fluctuate,
namely to support black holes of different mass. Another point
of view is that one must take into account the {\it self-gravitational
interaction\/} of the radiation.
\section{The Thin Shell Model}
To obtain a
complete description of a self-gravitating particle it would be necessary to
compute the action for an arbitrary motion of the particle and gravitational
field. While writing down a formal expression for such an object is
straightforward, it is of little use in solving a concrete problem due to the
large number of degrees of freedom present. To arrive at a more workable
description of the particle-hole system, we will keep only those degrees of
freedom which are most relevant to the problem of particle emission from
regions of low curvature. The first important restriction is made by
considering only spherically symmetric field configurations, and treating the
particle as a spherical shell. This is an interesting case
since black hole
radiation into a scalar field occurs primarily in the s-wave,
and virtual transitions to higher
partial wave configurations are formally suppressed by powers of
$\hbar$\begin{footnote}{Since
we do not address the ultraviolet problems of quantum gravity
these corrections are actually infinite, but one might anticipate that
in gravity theory with satisfactory ultraviolet behavior the virtual
transitions will supply additive corrections of order
$\frac{\omega^2 \Lambda^2}{M^4}$,
where $\Lambda$ is the effective cutoff, and $M$ is the mass of the black hole,
but will not alter the
exponential factors we compute.}\end{footnote}.
Before launching into the detailed calculation, which becomes rather
intricate, it seems
appropriate briefly to describe its underlying logic.
After the truncation to s-wave, the remaining dynamics
describes a
shell of matter interacting with a black hole of fixed mass and
with itself. (The mass as seen from infinity is the
total mass, including that from the shell variable, and is allowed
to vary. One could equally well have chosen the
total mass constant, and allowed the hole mass to vary.)
There is effectively
one degree of freedom, corresponding to the position of the
shell, but to isolate it one must choose appropriate
variables and solve
constraints, since the original action
superficially appears to contain much more than
this. Having done that, one obtains an effective action for the
true degree of freedom.
This effective
action is nonlocal, and its full quantization would require one to
resolve factor-ordering ambiguities, which appears very difficult.
Hence we quantize it semi-classically, essentially by
using the WKB approximation. After doing this one arrives at a
non-linear first order partial differential equation for the phase of
the wave function. This differential equation may be solved by the
method of characteristics. According to this method, one solves
for the characteristics, specifies the
function to be determined along a generic initial surface
(intersecting the
characteristics transversally), and evolves the function away from
the initial surface, by integrating the action along the characteristics.
(For a nice brief account of this,
see \cite{whitham}.)
When the background geometry is regarded as
fixed the characteristics for particle motion are simply the
geodesics in that geometry, and they are essentially
independent of
the particle's
mass or energy --- principle of equivalence --- except that
null geodesics are used for massless particles, and
timelike geodesics for massive particles. Here we find that the
characteristics depend on the mass and energy in a highly non-trivial
way. Also the action along the
characteristics, which would be zero for a massless particle
and proportional to the length for a massive particle, is
now a much more
complicated expression.
Nevertheless we can solve the equations, to obtain the
proper modes for our
problem.
Having obtained the modes, the final step is to identify the state
of the quantum field --- that is, the occupation of the modes ---
appropriate to the physical conditions we
wish to describe. We do this
by demanding that a freely falling observer passing through the
horizon see no singular behavior, and that
positive frequency modes are unoccupied in the distant past. This,
it has been argued,
is plausibly the appropriate prescription for the state of the
quantum field excited by collapse of matter into a black hole, at
least in so far as it leads to late-time radiation. Using it, we
obtain a mixture of positive- and negative- frequency modes at
late times, which can be interpreted as a state of radiation from
the hole. For massless scalar particles, we carry the explicit
calculation
far enough to identify the leading correction to the exponential
dependence of the radiation intensity on frequency.
\subsection{Effective Action}
We now derive the Hamiltonian effective action
for a self-gravitating particle
in the s-wave.
First, we would like to explain why the Hamiltonian form of the
action is particularly well suited to our problem.
As explained above, our physical problem really contains just
one degree of freedom, but the original action appears to contain
several. The reason of course is that
Einstein gravity is a theory with constraints and one
should only include a
subset of the spherically symmetric configurations in the
physical description, namely those satisfying the constraints.
In general, in eliminating constraints
Hamiltonian methods are more flexible than Lagrangian
methods.
This appears to be very much the case for our problem,
as we now discuss.
In terms of the variables appearing in the Lagrangian description,
the constraints have the form
$$
{\cal C}_{L} \left[ \hat{r}, \dot{\hat{r}}; g_{\mu \nu}, \dot{g_{\mu \nu}} \right] =0,
$$
where $\hat{r}$ is the shell radius, and $ \,\dot{}\,$ represents $\frac{d}{dt}$.
When applied to the spherically symmetric, source free, solutions, one obtains
the content of Birkhoff's theorem -- the unique solution is the Schwarzschild
geometry with some mass, $M$. Since this must hold for the regions interior
and exterior to the shell (with a different mass $M$ for each), and since $M$
must be time independent, we see that only those shell trajectories which are
``energy conserving'' are compatible with the constraints. This feature makes
the transition to the quantum theory rather difficult, as one desires an
expression for the action valid for an arbitrary shell trajectory. This defect
is remedied in the Hamiltonian formulation, where the constraints are
expressed in terms of momenta rather than time derivatives,
$$
{\cal C}_{H} \left[ \hat{r}, p; g_{ij}, \pi_{ij} \right] =0.
$$
At each time, the unique solution is again some slice of the Schwarzschild
geometry, but the constraints no longer prevent $M$ from being time dependent.
Thus, an arbitrary shell trajectory $\hat{r} (t)$, $p(t)$, is perfectly consistent
with the Hamiltonian form of the constraints, making quantization much more
convenient.
As before, we begin by writing the metric in ADM form:
\begin{equation}
ds^{2}=-N^{t}(t,r)^{2} dt^{2} + L(t,r)^{2}[dr +N^{r}(t,r)dt]^{2}
+R(t,r)^{2}[d\theta ^{2} + {\sin{\theta}}^{2} d\phi ^{2}]
\end{equation}
In considering the above form, we have restricted ourselves to spherically
symmetric geometries at the outset. With this choice of variables, the action
for the shell is written
\begin{equation}
S^{s}=-m\int\! \sqrt{-\hat{g}_{\mu \nu} d\hat{x}^{\mu} d\hat{x}^{\nu}}
=-m\int\! dt\, \sqrt{\hat{N}^{t^{2}} -\hat{L}^{2}\left(\dot{\hat{r}}+
\hat{N}^{r}\right)^{2}},
\end{equation}
$m$ representing the rest mass of the shell,
and the carets instructing one to evaluate quantities at the
shell $\left(\hat{g}_{\mu \nu}=g_{\mu \nu}(\hat{t},\hat{r})\right)$.
The action for the gravity-shell system is then
\begin{equation}
S= \frac{1}{16\pi} \int\!d^{4} x\, \sqrt{-g}\, {\cal R} - m \int\! dt\,
\sqrt{({\hat{N}}^{t})^{2} - {\hat{L}}^{2}(\dot{\hat{r}} + {\hat{N}}^{r})^{2}}
+ \mbox{ boundary terms}
\end{equation}
and can be written in canonical form as
\begin{equation}
S=\int\! dt\, p\, \dot{\hat{r}}\, + \int\! dt\, dr\, [\pi_{R}\dot{R} + \pi_{L}\dot{L}
-N^{t}({\cal H}_{t}^{s}+{\cal H}_{t}^{G}) - N^{r}({\cal H}_{r}^{s}+{\cal H}_{r}
^{G})] - \int\! dt\, M_{\mbox{{\scriptsize ADM}}}
\label{act}
\end{equation}
with
\begin{equation}
{\cal H}_{t}^{s}=\sqrt{(p/\hat{L})^{2}+m^{2}\,}\: \delta(r-\hat{r})
\ \ \ \ \mbox{;} \ \ \ \
{\cal H}_{r}^{s}=-p\, \delta(r-\hat{r})
\end{equation}
\begin{equation}
{\cal H}_{t}^{G}=\frac{L {\pi_{L}}^{2}}{2R^2} - \frac{\pi_{L} \pi_{R}}{R}
+\left(\frac{RR'}{L}\right)' - \frac{{R'}^{2}}{2L} - \frac{L}{2}
\ \ \ \ \mbox{;} \ \ \ \
{\cal H}_{r}^{G}=R'\pi_{R} - L \pi_{L}'
\label{gravcons}
\end{equation}
where $'$ represents $\frac{d}{dr}$.
$M_{\mbox{{\scriptsize ADM}}}$ is the ADM mass of the system,
and is numerically equal to the total mass of the
combined gravity-shell system.
We now wish to eliminate the gravitational degrees of freedom in order to
obtain an effective action which depends only on the shell variables. To
accomplish this, we first identify the constraints which are
obtained by varying with respect to $N^{t}$ and $N^{r}$:
\begin{equation}
{\cal H}_{t}={\cal H}_{t}^{s}+{\cal H}_{t}^{G}=0 \mbox{ \ \ \ ; \ \ \ }
{\cal H}_{r}={\cal H}_{r}^{s}+{\cal H}_{r}^{G}=0.
\label{con}
\end{equation}
By solving these constraints, and inserting the solutions back into (\ref{act})
we can eliminate the dependence on $\pi_{R}$ and $\pi_{L}$. We first consider
the linear combination of constraints
\begin{equation}
0=\frac{R'}{L}{\cal H}_{t} + \frac{\pi_{L}}{RL} {\cal H}_{r}= -{\cal M}'
+\frac{\hat{R}'}{\hat{L}}{\cal H}_{t}^{s}+\frac{\hat{\pi}_{L}}{\hat{R}\hat{L}}
{\cal H}_{r}^{s}
\label{mcon}
\end{equation}
where
\begin{equation}
{\cal M} = \frac{{\pi_{L}}^{2}}{2R} + \frac{R}{2} - \frac{R{R'}^{2}}{2L^{2}}.
\label{mdef}
\end{equation}
Away from the shell the solution of this constraint is simply ${\cal M}=$
constant. By considering a static slice ($\pi_{L}=\pi_{R} = 0$), we see that
the solution is a static slice of the Schwarzschild geometry with $\cal{M}$ the
corresponding mass parameter. The presence of the shell causes ${\cal M}$ to
be discontinuous at $\hat{r}$, so we write
$$
{\cal M}=M \ \ \ \ r<\hat{r}
$$
\begin{equation}
{\cal M}=M_{+} \ \ \ \ r>\hat{r}.
\end{equation}
As there is no matter outside the shell we also have
$M_{\mbox{{\scriptsize ADM}}}=M_{+}$.
Then, using (\ref{mcon}) and (\ref{mdef}) we can solve the constraints to find
$\pi_{L}$ and $\pi_{R}$:
$$
\pi_{L}=R \mo{R'} \mbox{ \ \ \ ; \ \ \ } \pi_{R} = \frac{L}{R'} \pi_{L}'
\ \ \ \ \ \ \ \ \ \ \ r<\hat{r}
$$
\begin{equation}
\pi_{L}=R\mop{R'} \mbox{ \ \ \ ; \ \ \ } \pi_{R}=\frac{L}{R'} \pi_{L}';
\ \ \ \ \ \ \ \ \ \ \ r>\hat{r}.
\label{mom}
\end{equation}
The relation between $M_{+}$ and $M$ is found by solving the constraints at the
position of the shell. This is done most easily by choosing coordinates such
that $L$ and $R$ are continuous as one crosses the shell, and $\pi_{R,L}$
are free of singularities there. Then, integration of the constraints
across the shell yields
$$
\pi_{L}(\hat{r}+\epsilon) - \pi_{L}(\hat{r}-\epsilon) = -p/ \hat{L}
$$
\begin{equation}
R'(\hat{r}+\epsilon) - R'(\hat{r}-\epsilon) = - \frac{1}{\hat{R}} \sqrt{p^{2} + m^{2} \hat{L}^{2}}
\label{cons}
\end{equation}
Now, when the constraints are satisfied a variation of the action takes the
form
\begin{equation}
dS= p\, d\hat{r} + \int dr (\pi_{R} \delta\! R + \pi_{L} \delta\!L) - M_{+}\, dt
\label{var}
\end{equation}
where $\pi_{R,L}$ are now understood to be given by (\ref{mom}),
and $M_{+}$ is determined by solving (\ref{cons}).
We wish to integrate the expression (\ref{var}) to find the action for an
arbitrary shell trajectory. As discussed above, the
geometry inside the shell is taken to be fixed (namely, $M$ is held constant)
while the geometry outside the shell will vary in order to satisfy the
constraints. It is easiest to integrate the action by initially varying the
geometry away from the shell. We first consider starting from an arbitrary
geometry and varying $L$ until $\pi_{R}=\pi_{L}=0$, while holding
$\hat{r}, p, R, \hat{L}$ fixed:
\begin{equation}
\begin{array}{l}
\int\! dS = \int_{r_{\mbox{{\tiny min}}}}^{\infty}\!dr \int_{\pi = 0}^{L}\! \delta\!L\, \pi_{L}
\vspace{4mm}
\\
=\int_{r_{\mbox{{\tiny min}}}}^{\hat{r}-\epsilon}\! dr \int_{\pi=0}^{L}\! \delta\!L\, R \mo{R'}
+\int_{\hat{r}+\epsilon}^{\infty}\! dr \int_{\pi=0}^{L}\! \delta\!L\, R \mop{R'}
\vspace{4mm}
\\
= \int_{r_{\mbox{{\tiny min}}}}^{\hat{r}-\epsilon}\! dr \left[ RL \mo{R'} + RR'
\log{\left|\frac{ \smo{R'}}{\sq{2M/R}}\right|}\right]
\vspace{1mm}
\\
\ \ \ + \int_{\hat{r}+\epsilon}^{\infty}\! dr \left[ RL \mop{R'} +
RR' \log{\left|\frac{\smop{R'}}{\sq{2M_{+}/R}}\right|}\right]
\label{varyL}
\end{array}
\end{equation}
where the lower limit of integration, $r_{{\scriptsize min}}$,
properly extends to the
collapsing matter forming the black hole; its precise value will not be
important. We have discarded the constant arising from the lower limit
of the $L$ integration. In the next stage we can vary $L$ and $R$,while
keeping
$\pi_{R,L}=0$, to some set geometry. Since the momenta vanish, there
is no contribution to the action from this variation.
It remains to consider nonzero variations at the shell. If an arbitrary
variation of $L$ and $R$ is inserted into the final expression of (\ref{varyL})
one finds
\begin{equation}
dS=\int_{r_{\mbox{{\tiny min}}}}^{\infty}\! dr\, [\pi_{R} \delta\!R + \pi_{L} \delta\!L]
- \left [\frac{\partial S}{\partial \hat{R}'}(\hat{r}+\epsilon) -
\frac{\partial S}{\partial \hat{R}'}(\hat{r}-\epsilon)\right]d\!\hat{R}
+ \frac{\partial S}{\partial M_{+}} d\!M_{+}.
\label{try}
\end{equation}
Since $R'$ is discontinuous at the shell,
$$
\frac{\partial S}{\partial \hat{R}'}
(\hat{r}+\epsilon) - \frac{\partial S}{\partial \hat{R}'}(\hat{r}-\epsilon)
$$
is nonvanishing and needs to be subtracted in order that the relations
$$
\frac{\delta S}{\delta R} = \pi_{R} \mbox{ \ \ ; \ \ }
\frac{\delta S}{\delta L} = \pi_{L}
$$
will hold. From (\ref{varyL}), the term to be subtracted is
\pagebreak
$$
-\left[\frac{\partial S}{\partial \hat{R}'}(\hat{r}+\epsilon) -
\frac{\partial S}{\partial \hat{R}'}(\hat{r}-\epsilon)\right] d\!\hat{R}
\vspace{4mm}
$$
$$
=-d\!\hat{R} \hat{R} \log{\left|\frac{\smoh{R'(\hat{r}-\epsilon)}}{\sq{2M/\hat{R}}}\right|}
$$
\begin{equation}
\ \ \ + d\!\hat{R} \hat{R} \log{\left|\frac{\smohp{R'(\hat{r}+\epsilon)}}{\sq{2M_{+}/\hat{R}}}\right|}.
\end{equation}
Similarly, arbitrary variations of $L$ and $R$ induce a variation of $M_{+}$
causing the appearance of the final term in (\ref{try}). Thus we need to
subtract
\begin{equation}
\frac{\partial S}{\partial M_{+}} d\!M_{+} =
-\int_{\hat{r}+\epsilon}^{\infty}\! dr\, L \frac{\mop{R'}}{1-2M_{+}/R} d\!M_{+}.
\end{equation}
Finally, we consider variations in $p, \hat{r}$, and $t$. $t$ variations
simply give $dS = - M_{+} dt$. We do not need to separately consider variations
of $p$ and $\hat{r}$, since when the constraints are satisfied their variations
are already accounted for in our expression for $S$, as will be shown.
Collecting all of these terms, our final expression for the action reads
$$
S= \int_{r_{\mbox{{\tiny min}}}}^{\hat{r}-\epsilon}\!dr\,\left[ RL \mo{R'}
+RR'\log{\left|\frac{\smo{R'}}{\sq{2M/R}}\right|}\right]
\vspace{2mm}
$$
$$
+\int_{\hat{r}+\epsilon}^{\infty}\! dr\, \left[ RL \mop{R'}
+RR' \log{\left|\frac{\smop{R'}}{\sq{2M_{+}/R}}\right|}\right]
\vspace{2mm}
$$
$$
-\int\! dt\, \dot{\hat{R}} \hat{R}\left [\log{\left|
\frac{\smoh{R'(\hat{r}-\epsilon)}}{\sq{2M/\hat{R}}}\right|}\right.
\vspace{2mm}
$$
$$
+\left. \log{\left|\frac{\smohp{R'(\hat{r}+\epsilon)}}{\sq{2M_{+}/\hat{R}}}\right|}\right]
\vspace{2mm}
$$
\begin{equation}
+ \int\! dt \int_{\hat{r}+\epsilon}^{\infty}\! dr\, \frac{L \mop{R'}}{1-2M_{+}/R}
\dot{M_{+}} - \int\! dt\, M_{+}.
\label{mess}
\end{equation}
To show that this is the correct expression we can differentiate it; then
it can
be seen explicitly that when the constraints are satisfied (\ref{var}) holds.
We now wish to write the action in a more conventional form as the time
integral of a Lagrangian. As it stands, the action in (\ref{mess})
is given for
an arbitrary choice of $L$ and $R$ consistent with the constraints. There is,
of course, an enormous amount of redundant information contained in this
description, since many $L$'s and $R$'s are equivalent to each other through a
change of coordinates. To obtain an action which only depends on the truly
physical variables $p, \hat{r}$ we make a specific choice for $L$ and
$R$, {\it ie.} choose a gauge. In so doing, we must respect the condition
$$
R'(\hat{r}+\epsilon) - R'(\hat{r}-\epsilon) = -\frac{1}{\hat{R}} \sqrt{p^2 + m^2 \hat{L}^2}
$$
which constrains the form of $R'$ arbitrarily near the shell. Suppose we
choose $R$ for all $r>\hat{r}$; then $R'(\hat{r}-\epsilon)$ is fixed by the constraint, but we
can still choose $R$ for $r<\hat{r}-\epsilon$, in other words, away from the
shell. We will let $R_{<}'$ denote the value of $R'$ close to the shell but
far enough away such that $R$ is still freely specifiable. We employ the
analogous definition for $R_{>}'$, except in this case we are free to choose
$R_{>}' = R'(\hat{r}+\epsilon)$.
In terms of this notation the time derivative of $S$ is
$$
L=\frac{dS}{dt}=\dot{\hat{r}}\hat{R}\hat{L}\left[ \moh{R_{<}'} - \moph{R_{>}'}
\right]
$$
$$
\ \ \ \ \ - \dot{\hat{R}} \hat{R} \log{\left| \frac{\smoh{R'(\hat{r}-\epsilon)}}{\smoh{R_{<}'}}
\right|}
$$
\begin{equation}
\ \ \ \ +\int_{r_{\mbox{{\tiny min}}}}^{\hat{r}-\epsilon}\! dr [\pi_{R} \dot{R} +\pi_{L}\dot{L}]
+ \int_{\hat{r}+\epsilon}^{\infty}\! dr [\pi_{R} \dot{R} + \pi_{L} \dot{L}] -M_{+}.
\label{lag}
\end{equation}
At this point we will, for simplicity, specialize to a massless particle
($m=0$) and define $\eta = \pm = \mbox{sgn($p$)}$. Then the constraints
(\ref{cons}) read
$$
R'(\hat{r}-\epsilon) = R'(\hat{r}+\epsilon) + \frac{\eta p }{\hat{R}}
$$
\begin{equation}
\moh{R'(\hat{r}-\epsilon)} = \moph{R'(\hat{r}+\epsilon)} + \frac{p}{\hat{L} \hat{R}}.
\label{newcon}
\end{equation}
These relations can be inserted into (\ref{lag}) to yield
$$
L= \dot{\hat{r}} \hat{R} \hat{L} \left[ \moh{R_{<}'} - \moph{R_{>}'}\right]
$$
$$
-\eta \dot{\hat{R}} \hat{R}\log{\left|\frac{R_{>}'/\hat{L}
- \eta \sqrt{(R_{>}'/\hat{L})^{2} -1 +2M_{+}/\hat{R}}}
{R_{<}'/\hat{L} - \eta \sqrt{(R_{<}'/\hat{L})^{2}-1+2M/\hat{R}}}\right|}
$$
\begin{equation}
\ \ \ + \int_{r_{\mbox{{\tiny min}}}}^{\hat{r}-\epsilon}\! dr [\pi_{R}\dot{R} + \pi_{L}\dot{L}]
+\int_{\hat{r}+\epsilon}^{\infty}\! dr [\pi_{R} \dot{R} +\pi_{L} \dot{L}] - M_{+}.
\label{nlag}
\end{equation}
Now we can use the freedom to choose a gauge to make (\ref{nlag}) appear as
simple as possible. It is clearly advantageous to choose $L$ and $R$ to be
time independent, so $\pi_{R} \dot{R} + \pi_{L} \dot{L}=0$. Also, having
$R'=L$ simplifies the expressions further. Finally, it is crucial that the
metric be free of coordinate singularities. A gauge which conveniently
accommodates these features is
$$
L=1 \ \ ; \ \ \ R=r
$$
The $L=1, R=r$ gauge reduces the Lagrangian to
\begin{equation}
L=\dot{\hat{r}} [\sqrt{2M \hat{r}} - \sqrt{2M_{+} \hat{r}}] -\eta \dot{\hat{r}} \hat{r}
\log{\left|\frac{\sqrt{\hat{r}}-\eta \sqrt{M_{+}}}{\sqrt{\hat{r}}- \eta \sqrt{2M}}
\right|} - M_{+}
\label{laga}
\end{equation}
where $M_{+}$ is now found from the constraints (\ref{newcon}) to be related to
$p$ by
\begin{equation}
p= \frac{M_{+} - M}{\eta - \sqrt{2M_{+}/r}}.
\end{equation}
The canonical momentum conjugate to $\hat{r}$ obtained from \ref{laga} is
\begin{equation}
p_{c}= \frac{\partial L}{\partial \dot{\hat{r}}} = \sqrt{2M\hat{r}} - \sqrt{2M_{+}\hat{r}}
- \eta \hat{r} \log{\left|\frac{\sqrt{\hat{r}}-\eta \sqrt{2M_{+}}}{\sqrt{\hat{r}}
-\eta \sqrt{2M}}\right|}
\label{pcan}
\end{equation}
in terms of which we write the action in canonical form as
\begin{equation}
S=\int\! dt [p_{c} \dot{\hat{r}} - M_{+}]
\label{Scan}
\end{equation}
which identifies $M_{+}$ as the Hamiltonian. We should point out that $M_{+}$
is the Hamiltonian only for a restricted set of gauges. If we look back at
(\ref{nlag}) we see that the terms $\pi_{R} \dot{R} + \pi_{L} \dot{L}$ will in
general contribute to the Hamiltonian.
\subsection{Quantization}
In this section we discuss the quantization of the effective action
(\ref{Scan}).
First, it is convenient to rewrite the action in a form which explicitly
separates out the contribution from the particle. We write
$$
M_{+}=M-p_{t}
$$
so
\begin{equation}
S=\int\! dt [p_{c} \dot{\hat{r}} +p_{t}]
\label{Act}
\end{equation}
and the same substitution is understood to be made in (\ref{pcan}). We have
omitted a
term, $\int\! dt M$, which simply contributes an overall constant to our
formulas. In order to place our results in perspective, it
is useful to step back and consider the analogous expressions in flat space.
Our results are an extension of
\begin{equation}
p=\pm\sqrt{{p_{t}}^{2} -m^{2}}
\label{pflat}
\end{equation}
\begin{equation}
S=\int\! dt [p\dot{r}+p_{t}].
\end{equation}
Indeed, the $G \rightarrow 0$ limit of (\ref{pcan}), (\ref{Scan})
yields precisely these expressions (with \linebreak $m=0$).
To quantize, one is tempted to
insert the substitutions $p \rightarrow -i\frac{\partial}{\partial r}$,
$p_{t} \rightarrow -i\frac{\partial}{\partial t}$ into (\ref{pflat}),
so as to satisfy the canonical commutation relations.
This
results in a rather unwieldy, nonlocal differential equation. In this
trivial case we know, of course, that the correct description of the particle
is obtained by demanding locality and squaring both sides of (\ref{pflat})
before substituting $p$ and $p_{t}$. So for this example it is
straightforward to move from the point particle description
to the field theory
description, {\it i.e}. the Klein-Gordon equation.
Now, returning to (\ref{pcan})
we are again met with the question of how to implement the substitutions
$p \rightarrow -i \partial$. In this case the difficulty is more severe;
we no longer have locality as a guiding criterion instructing us how to
manipulate (\ref{pcan}) before turning the $p$'s into differential operators.
This is because we expect the effective action (\ref{Act}) to be nonlocal on
physical grounds, as it was obtained by including the gravitational field of
the shell.
There is, however, a class of solutions to the field equations for which this
ambiguity is irrelevant to leading order, and which is sufficient to determine
the late-time radiation from a black hole. These are the short-wavelength
solutions, which are accurately described by the geometrical optics, or WKB,
approximation. Writing these solutions as
$$
\phi(t,r)=e^{i S(t,r)},
$$
the condition determining the validity of the WKB approximation is that
$$
|\partial S| \gg |\partial^{2}S|^{1/2}, \ |\partial^{3} S|^{1/3} \ \ldots
$$
and that the geometry is slowly varying compared to $S$. In this regime,
derivatives acting on $\phi(t,r)$ simply bring down powers of $\partial S$,
so we can make the replacements
$$
p_{c} \rightarrow \frac{\partial S}{\partial r} \mbox{ \ \ \ ; \ \ \ }
p_{t} \rightarrow \frac{\partial S}{\partial t}
$$
and obtain a Hamilton-Jacobi equation for $S$. Furthermore, it is well known
that the solution of the Hamilton-Jacobi equation is just the classical action.
So, if $\hat{r} (t)$ is a solution of the equations of motion found by
extremizing (\ref{Act}), then
\begin{equation}
S\left(t,\hat{r} (t)\right)= S\left(0,\hat{r} (0)\right) +
\int_{0}^{t}\! dt \left[p_{c}\left(\hat{r} (t) \right) \dot{\hat{r}} (t) + p_{t} \right]
\label{phase}
\end{equation}
where
\begin{equation}
p_{c}\left(0,\hat{r} \right) = \frac{\partial S}{\partial r}\left(0,\hat{r} \right).
\label{pinit}
\end{equation}
Since the Lagrangian in (\ref{Act}) has no explicit time dependence, the
Hamiltonian $p_{t}$ is conserved. Using this fact, it is easy to verify that
the trajectories, $\hat{r} (t)$, which extremize (\ref{Act}) are simply the null
geodesics of the metric
\begin{equation}
ds^{2}=-dt^{2}+\left(dr+\sqrt{\frac{2M_{+}}{r}} dt\right)^{2}.
\end{equation}
From (\ref{geo}) the geodesics are:
$$
\mbox{ingoing: \ \ \ } t+\hat{r} (t)+2\sqrt{2M_{+} \hat{r} (t)} + 4M_{+}
\log{[\sqrt{\hat{r} (t)}+\sqrt{2M_{+}}]}
$$
$$
\hspace{20mm} =\hat{r} (0)+2\sqrt{2M_{+}\hat{r} (0)} +4M_{+}\log{[\sqrt{\hat{r} (0)}+\sqrt{2M_{+}}]}
\vspace{4mm}
$$
$$
\mbox{outgoing: \ \ \ } t-\hat{r} (t) -2\sqrt{2M_{+}\hat{r} (t)}-4M_{+}
\log{[\sqrt{\hat{r} (t)}-\sqrt{2M_{+}}]}
$$
\begin{equation}
\hspace{22mm}=-\hat{r} (0)-2\sqrt{2M_{+} \hat{r} (0)}-4M_{+}\log{[\sqrt{\hat{r} (0)}-\sqrt{2M_{+}}]}.
\label{geodef}
\end{equation}
$M_{+}$, in turn, is determined by the initial condition $S(0,r)$ according to
(\ref{pcan}) and (\ref{pinit}):
$$
\hspace{-7mm}\mbox{ingoing: \ \ \ } \frac{\partial S}{\partial r}(0,\hat{r} (0))=
\sqrt{2M\hat{r} (0)}-\sqrt{2M_{+}\hat{r} (0)}+\hat{r} (0)\log{\left|\frac{\sqrt{\hat{r} (0)}
+\sqrt{2M_{+}}}{\sqrt{\hat{r} (0)}+\sqrt{2M}}\right|}
$$
\begin{equation}
\mbox{outgoing: \ \ \ } \frac{\partial S}{\partial r}(0,\hat{r} (0))
=\sqrt{2M\hat{r} (0)}-\sqrt{2M_{+}\hat{r} (0)}-\hat{r} (0)\log{\left|\frac{\sqrt{\hat{r} (0)}
-\sqrt{2M_{+}}}{\sqrt{\hat{r} (0)}-\sqrt{2M}}\right|}.
\label{mpdef}
\end{equation}
Finally, we can use this value of $M_{+}$ to determine $p_{c}(t)$:
$$
\hspace{-7mm}\mbox{ingoing: \ \ \ } p_{c}(t)=\sqrt{2M\hat{r} (t)}-\sqrt{2M_{+}\hat{r} (t)}
+\hat{r} (t)\log{\left|\frac{\sqrt{\hat{r} (t)}+\sqrt{2M_{+}}}{\sqrt{\hat{r} (t)}
+\sqrt{2M}}\right|}
$$
\begin{equation}
\mbox{outgoing: \ \ \ } p_{c}(t)=\sqrt{2M\hat{r} (t)}-\sqrt{2M_{+}\hat{r} (t)} -\hat{r} (t)
\log{\left|\frac{\sqrt{\hat{r} (t)}-\sqrt{2M_{+}}}{\sqrt{\hat{r} (t)}-\sqrt{2M}}
\right|}.
\label{pcdef}
\end{equation}
These formulas are sufficient to compute $S(t,r)$ given $S(0,r)$.
As will be discussed in the next section, the relevant solutions needed to
describe the state of the field following black hole formation are those with
the initial condition
\begin{equation}
S(0,r)=kr \ \ \ \ \ \ k>0
\label{sinit}
\end{equation}
near the horizon. Here, $k$ must be large ($\gg1/M$) if the solution is to be
accurately described by the WKB approximation. In fact, the relevant $k$'s
needed to calculate the radiation from the hole at late times become
arbitrarily large, due to the ever increasing redshift experienced by the
emitted quanta as they escape to infinity. We also show in the next section
that to compute the emission probability of a quantum of frequency $\omega$,
we are required to find the solution for all times in the region between
$r=2M$ and $r=2(M+\omega)$. That said, we turn to the calculation of $S(t,r)$
in this region, and with the initial condition (\ref{sinit}). The solutions
are determined from (\ref{phase}), (\ref{geodef})-(\ref{pcdef}). Because of
the large redshift, we only need to keep those terms in these relations which
become singular near the horizon. We then have for the outgoing solutions:
\begin{equation}
S(t,r)=k\,\hat{r} (0)-\int_{\hat{r} (0)}^{r}\! d\hat{r}\,\hat{r} \log{\left[\frac{\sqrt{\hat{r}}
-\sqrt{2M_{+}}}{\sqrt{\hat{r}}-\sqrt{2M}}\right]} - (M_{+}-M)t
\label{phasecal}
\end{equation}
\begin{equation}
t-4M_{+}\log{[\sqrt{r}-\sqrt{2M_{+}}]}=
-4M_{+}\log{[\sqrt{\hat{r}(0)}-\sqrt{2M_{+}}]}
\label{tcal}
\end{equation}
\begin{equation}
k=-\hat{r} (0) \log{\left[\frac{\sqrt{\hat{r} (0)}-\sqrt{2M_{+}}}
{\sqrt{\hat{r} (0)} -\sqrt{2M}}\right]}.
\label{kcal}
\end{equation}
To complete the calculation, we need to invert (\ref{tcal}) and (\ref{kcal})
to find $M_{+}$ and $\hat{r} (0)$ in terms of $t$ and $r$, and then insert these
expressions into (\ref{phasecal}). One finds that to next to leading order,
$$
\sqrt{2M_{+}}=\sqrt{2M}+(\sqrt{r}-\sqrt{2M})\frac{(e^{k/2M'}-1)e^{-t/4M'}}
{1+(e^{k/2M'}-1)e^{-t/4M'}}
$$
\begin{equation}
\sqrt{\hat{r} (0)}=\sqrt{2M}+(\sqrt{r}-\sqrt{2M})\frac{e^{(k/2M' -t/4M')}}
{1+(e^{k/2M'}-1)e^{-t/4M'}}
\label{Mr}
\end{equation}
where
\begin{equation}
M'=M+\sqrt{2M}(\sqrt{r}-\sqrt{2M})\frac{e^{(k/2M-t/4M)}}{1+e^{(k/2M-t/4M)}}.
\label{Mprime}
\end{equation}
Plugging these relations into (\ref{phasecal}) and keeping only those terms
which contribute to the late-time radiation, one finds after some tedious
algebra,
\begin{equation}
S(t,r)=-(2M^{2}-r^{2}/2)\log{\left[1+e^{(k/2M'-t/4M')}\right]}.
\label{sol}
\end{equation}
\subsection{Results}
We will now discuss the application of these results to the problem of black
hole radiance. The procedure is a slight modification of the one we discussed
in the free field theory case. As before, the point is that there are
two inequivalent sets of modes which need to be considered:
those which are natural from the standpoint of an observer making measurements
far from the black hole, and those which are natural from the standpoint of an
observer freely falling through the horizon subsequent to the collapse of the
infalling matter. The appropriate modes for the observer at infinity are those
which are positive frequency with respect to the Killing time, $t$.
Writing these modes as
$$
u_{k}(r) e^{-i\omega_{k} t},
$$
$\hat{\phi} (t,r)$ reads
\begin{equation}
\hat{\phi}(t,r)=\int\! dk \left[ \hat{a}_{k}u_{k}(r) e^{-i\omega_{k} t}
+ \hat{a}_{k}^{\dagger}u_{k}^{*}(r) e^{i \omega_{k} t}\right].
\label{phik}
\end{equation}
These modes are singular at the horizon,
$$
\frac{du_{k}}{dr} \ \rightarrow \infty \mbox{ \ \ as \ \ } r \rightarrow 2M.
$$
Symptomatic of this is that the freely falling observer would measure an
infinite energy-momentum density in the corresponding vacuum state,
$$
\langle 0_{t} | T_{\mu \nu} | {0}_{t} \rangle \rightarrow \infty
\mbox{ \ \ as \ \ } r \rightarrow 2M
$$
where $\hat{a}_{k} | 0_{t}\rangle =0$. However, we do not expect this to be
the state resulting from collapse, since the freely falling observer is not
expected to encounter any pathologies in crossing the horizon, where the local
geometry is entirely nonsingular for a large black hole.
To describe the state resulting from collapse, it is more appropriate to use
modes which extend smoothly through the horizon, and which are positive
frequency with respect to the freely falling observer.
Denoting a complete set of
such modes by $v_{k}(t,r)$, we write
\begin{equation}
\hat{\phi}(t,r) \int\! dk \left[\hat{b}_{k}v_{k}(t,r)+
\hat{b}_{k}^{\dagger} v_{k}^{*}(t,r)\right].
\label{phih}
\end{equation}
Then, the state determined by
$$
\hat{b}_{k}|0_{v}\rangle=0
$$
results in a non-singular energy-momentum density at the horizon, and so is a
viable candidate.
The operators $\hat{a}_{k}$ and $\hat{b}_{k}$ are related by the Bogoliubov
coefficients,
\begin{equation}
\hat{a}_{k}=\int\! dk' \left[\alpha_{kk'}\hat{b}_{k'}+\beta_{kk'}\hat{b}_{k'}
^{\dagger}
\right]
\label{bog}
\end{equation}
where
$$
\alpha_{kk'}=\frac{1}{2\pi u_{k}(r)}\int_{-\infty}^{\infty}\! dt\,
e^{i\omega_{k} t} v_{k'}(t,r)
$$
\begin{equation}
\beta_{kk'}=\frac{1}{2\pi u_{k}(r)}\int_{-\infty}^{\infty}\! dt\,
e^{i\omega_{k} t} v_{k'}^{*}(t,r).
\label{bcoef}
\end{equation}
Note that here we compute the Bogoliubov coefficients by performing a $t$
integration, rather than an integration over a spatial coordinate, as is
conventional. We are forced to do this in the present case since we do not
know the spatial dependence of the definite energy modes.
The flux seen at infinity is
\begin{equation}
F_{\infty}(\omega_{k})=\frac{d\omega_{k}}{2\pi}\frac{\Gamma(\omega_{k})}
{|\alpha_{kk'}/\beta_{kk'}|^{2}-1}.
\label{flux}
\end{equation}
Next, we consider the issue of determining the modes $v_{k}(t,r)$. As stated
above, we require these modes to be nonsingular at the horizon. Since the
metric near the horizon is a smooth function of $t$ and $r$, a set of such
modes can be defined by taking their behaviour on a constant time surface,
say $t=0$, to be
$$
v_{k}(0,r)\approx e^{ikr} \mbox{ \ \ as \ \ } r\rightarrow 2M.
$$
This is, of course, the initial condition given in (\ref{sinit}). Now, the
integrals in (\ref{bcoef}) determining the Bogoliubov coefficients depend on
the values of $v_{k}(t,r)$ at constant $r$. Since $v_{k}$ is evaluated in the
WKB approximation, the highest accuracy will be obtained when $r$ is as close
to the horizon as possible, since that is where $v_{k}$'s wavelength is short.
On the other hand, in calculating the emission of a particle of energy
$\omega_{k}$, we cannot take $r$ to be less than $2(M+\omega_{k})$, since the
solution $u_{k}(r)e^{-i\omega_{k} t}$ cannot be extended past that point.
Therefore, we calculate the integrals with $r=2(M+\omega_{k})$.
The results of the previous section give us an explicit expression
for $v_{k}$. From (\ref{sol}),
\begin{equation}
v_{k}(t,2(M+\omega_{k}))=
e^{iS(t,2(M+\omega_{k}))}=
e^{i(4M\omega_{k}+2\omega^{2})\log{[1+e^{(k/2M'-t/4M')}] }}
\end{equation}
where $M'$ is
\begin{equation}
M'=M+\sqrt{2M}(\sqrt{2(M+\omega_{k})}-\sqrt{2M})\frac{e^{(k/2M-t/4M)}}
{1+e^{(k/2M-t/4M)}}\approx M+\omega_{k}
\frac{e^{(k/2M-t/4M)}}{1+e^{(k/2M-t/4M)}}.
\end{equation}
Then, the integrals are,
\begin{equation}
\int_{-\infty}^{\infty} dt e^{i\omega_{k} t}e^{\pm i(4M\omega_{k'}+2\omega_{k'}
^{2 })\log{[1+e^{(k'/2M'-t/4M')}]}},
\label{integral}
\end{equation}
the upper sign corresponding to $\alpha_{kk'}$, and the lower to $\beta_{kk'}$.
We
can compute the integrals using the saddle point approximation.
It is readily seen
that for the upper sign, the saddle point is reached when
$$
e^{(k'/2M'-t/4M')}
\rightarrow \infty,
$$
so $t$ is on the real axis. For the lower sign, the saddle
point is
$$
e^{(k'/2M'-t/4M')} \approx -1/2,
$$
which, to zeroth order in $\omega_{k}$, gives
$$
t=4i\pi M + \mbox{real}
$$
and to first order in $\omega_{k}$, gives
$$
t=4i\pi(M-\omega_{k})+ \mbox{real}.
$$
Inserting these values of the saddle point into the integrands gives for the
Bogoliubov coefficients,
\begin{equation}
\left|\frac{\alpha_{kk'}}{\beta_{kk'}}\right|
= e^{4\pi(M-\omega_{k})\omega
_{k}}.
\end{equation}
The flux of radiation from the black hole is given by (\ref{flux}),
\begin{equation}
F_{\infty}(\omega_{k})=\frac{d\omega_{k}}{2\pi}\frac{\Gamma(\omega_{k})}
{e^{8\pi(M-\omega_{k})\omega_{k}}-1}.
\end{equation}
There is an alternative way of viewing the saddle point calculation, which
provides additional insight into the physical origin of the radiation. Let us
rewrite the integral (\ref{integral}) as
\begin{equation}
\int_{-\infty}^{\infty}\! dt e^{i\omega_{k} t \pm i S(t,2(M+\omega_{k'}))}.
\label{newint}
\end{equation}
The saddle point is given by that value of $t$ for which the derivative of the
expression in the exponent vanishes:
$$
\omega_{k} \pm \frac{\partial S}{\partial t} (t,2(M+\omega_{k}))=0.
$$
But $\partial S/\partial t$ is just the negative of the Hamiltonian,
$$
\frac{\partial S}{\partial t}=p_{t}=M-M_{+}
$$
so the saddle point equation becomes
$$
M_{+}=M\pm \omega_{k}.
$$
To find the corresponding values of $t$, we insert this relation into
(\ref{tcal}) and (\ref{kcal}):
\begin{equation}
t=4(M\pm \omega_{k})\log{\left[\frac{\sqrt{2(M+\omega_{k})+\epsilon\,}
-\sqrt{2(M+\omega_{k})\,}}
{\sqrt{\hat{r} (0)}
-\sqrt{2(M\pm\omega_{k})}}\right]}
\label{tsad}
\end{equation}
\vspace{2mm}
\begin{equation}
k=-\hat{r} (0)\log{\left[\frac{\sqrt{\hat{r} (0)}-\sqrt{2(M\pm\omega_{k})}}
{\sqrt{\hat{r} (0)} -\sqrt{2M}}\right]},
\label{ksad}
\end{equation}
where we have written $\hat{r}=2(M+\omega_{k})+\epsilon$ to make explicit that
$\hat{r}$ must lie outside the point where the solutions $u_{k}(r)$ break down.
We desire to solve for $t$ as $k\rightarrow \infty$. For the upper choice of
sign, we find from (\ref{ksad}) that
$$
\sqrt{\hat{r} (0)}=\sqrt{2(M+\omega_{k})}+\mbox{O}\!\left(e^{-k/2M}\right),
$$
which, from (\ref{tsad}), then shows that the corresponding value of $t$ is
purely real.
For the lower choice of sign we have,
$$
\sqrt{\hat{r} (0)}=\sqrt{2(M-\omega_{k})}-\mbox{O}\!\left(e^{-k/2M}\right).
$$
Continuing $t$ into the upper half plane, we find from (\ref{tsad}) that
$$
t=4i\pi (M-\omega_{k}) + \mbox{ real.}
$$
These results of course agree with our previous findings.
The preceding derivation invites us to interpret
the radiation as being due to
negative energy particles propagating in imaginary time. The particles
originate from just inside the horizon, and cross to the outside in an
imaginary time interval $4\pi (M-\omega_{k})$. This, perhaps, helps clarify
the analogy between black hole radiance and pair production in an electric
field, which, in an instanton approach \cite{affleck},
is also calculated by considering
particle trajectories in imaginary time.
Finally, let us return to the question of thermality. One might have
guessed that the correct exponential suppression factor could be
the Boltzmann factor for nominal temperature corresponding
to the mass of the hole before the radiation, after the radiation,
or somewhere in between. Thus one might have guessed that the
exponential suppression of the radiance could take the form
$e^{-\omega/T_{\rm before}}$, $e^{-\omega/T_{\rm after}}$, or
something in between. Our result, to lowest order, corresponds
to the nominal temperature for emission being equal
to $T_{\rm after}$.
\section{Corrections to Charged Black Hole Radiance}
In this section, two additional
things are done. First, we extend the calculations
to include a charged black hole, and charged matter. Although this step
does not present any significant formal difficulties,
the physical results we obtain are considerably richer
than what we found in our previous calculations involving neutral holes
and shells. In the neutral case the final result could be
summarized as a simple replacement of the nominal temperature governing
the radiation by the Hawking temperature for the mass after
radiation, so that the ``Boltzmann factor'' governing emission of energy
$\omega$ from a hole of mass $M$ became
\begin{equation}
e^{-\omega / T_{\rm eff.} } ~=~ e^{-\omega 8\pi (M -\omega )}~.
\label{teff}
\end{equation}
Note that the argument of the exponential is {\it not\/}
simply proportional to the energy $\omega$, so that the spectrum is
not, strictly speaking, thermal. While
the deviation from thermality is important in principle its structure,
in this case, is
rather trivial,
and one is left wondering whether that is
a general result. Fortunately we find that for charged holes the
final results are much more complex. We say ``fortunately'', not only
because this relieves us of the nagging fear that we have done a
simple calculation in a complicated way, but also for more physical
reasons.
For one knows on general grounds
that the thermal description of black
hole radiance breaks down completely for near-extremal holes.
One might anticipate, therefore, that something more drastic than a simple
modification of the nominal temperature will occur -- as indeed we
find. A particularly gratifying consequence of the accurate formula is
a form of ``quantum cosmic censorship''. Whereas a literal application
of the conventional thermal formulas for radiation yields a non-zero
amplitude for radiation past extremality -- that is, radiation leaving
behind a hole with larger charge than mass -- we find (within our
approximations) {\it vanishing\/} amplitude for such processes.
Second, we discuss in a more detailed fashion the relationship between
our method of calculation, which proceeds by reduction to an effective
particle theory, and more familiar approximations. We show that it amounts
to saturation of the functional integral of the
underlying s-wave field theory with one-particle intermediate
states, or alternatively to neglect of vacuum polarization. It is
therefore
closely related to conventional eikonal approximations. We demonstrate
the reduction of the field theory to a particle theory
explicitly in the related problem of particle creation
by a strong spherically symmetric charge source,
which is a problem of independent
interest.
\subsection{Self-Interaction Correction}
Our system consists of a matter shell of rest mass m and charge q interacting
with the electromagnetic and gravitational fields. The corresponding action is
\begin{equation}
S=\int [ -m \sqrt{-\hat{g}_{\mu \nu} d\hat{x}^{\mu} d\hat{x}^{\nu}}
+ q \hat{A}_{\mu} d\hat{x}^{\mu}]
+\frac{1}{16\pi} \int\! d^4\! x \sqrt{-g}\, [{\cal R}-F_{\mu \nu}F^{\mu \nu}]
\label{acta}
\end{equation}
The gravitational contribution to the Hamiltonian is the same as in
(\ref{gravcons}), and the shell and electromagnetic contributions are
\begin{equation}
{\cal H}_{t}^{s}=\left( \sqrt{(p/\hat{L})^{2}+m^{2}}-q\hat{A}_{t}
\right)\delta(r-\hat{r})
\mbox{ \ \ \ ; \ \ \ } {\cal H}_{r}^{s}=-p\,\delta(r-\hat{r})
\label{hshell}
\end{equation}
\begin{equation}
{\cal H}_{t}^{EM} = \frac{N^t L {\pi_{\!A_{r}}}^2}{2R^2} - A_t\,
\pi_{\!A_{r}}'
\label{hem}
\end{equation}
To arrive at this
form we have chosen a gauge such that $A_t$ is the only nonvanishing component
of $A_\mu$. Of course, we set $A_{r}=0$ only {\em after}
computing the canonical momentum $\pi_{\!A_{r}}$.
Constraints are found by varying the action with respect to $N^t$, $N^r$,
and $A_t$,
$$
{\cal H}_{t} \equiv {\cal H}_{t}^{s}+{\cal H}_{t}^{G}+{\cal H}_{t}^{EM}=0
\mbox{ \ \ \ ; \ \ \ } {\cal H}_{r}\equiv {\cal H}_{r}^{s} +{\cal H}_{r}^{G}=0
$$
\begin{equation}
\pi_{A_r}'+q\, \delta(r-\hat{r})=0.
\end{equation}
$\pi_R$ can be eliminated by forming the linear combination of constraints
\begin{equation}
0=\frac{R'}{L}{\cal H}_{t} +\frac{\pi_L}{RL}{\cal H}_{r}
=-{\cal M}' + \frac{R'}{L}({\cal H}_{t}^{s}+{\cal H}_{t}^{EM})
+\frac{\pi_L}{RL}{\cal H}_{r}^{s}
\label{mcona}
\end{equation}
where
\begin{equation}
{\cal M}=\frac{{\pi_{L}}^2}{2R^2}+\frac{R}{2}-\frac{RR'^2}{2L^2}.
\label{mdefa}
\end{equation}
We see from the Gauss' law constraint that $-\pi_{\!A_r}(r)$ is the charge
contained within a sphere of size $r$, so we define:
$Q(r) \equiv -\pi_{\!A_r}(r)$
Now, if the shell was absent $ {\cal M}(r) $ and $Q(r)$ would be given by
\begin{equation}
{\cal M}(r) = M-\int_{r}^{\infty}\! dr\, {R'(r){\cal H}_{t}^{EM}(r)\over L(r)}
\mbox{ \ \ \ ; \ \ \ } Q(r)=Q
\end{equation}
with $M$ and $Q$ being the mass and charge of the black hole as seen from
infinity. In the gauge $L=1, R=r$ these become
\begin{equation}
{\cal M}(r)=M-Q^{2}/2r \mbox{ \ \ \ ; \ \ \ } Q(r)=Q.
\label{msola}
\end{equation}
With the shell present we retain the expression (\ref{msola}) for the region
inside the shell, $r<\hat{r}$, whereas outside the shell we write
(with $L=1,\ R=r$),
\begin{equation}
{\cal M}(r) = M_{+} - (Q+q)^{2}/2r \mbox{ \ \ \ ; \ \ \ } Q(r)=Q+q
\label{out}
\end{equation}
where $M_+$ and $Q+q$ are the mass and charge of the hole-shell system as
measured at infinity.
By using the constraints we can determine $\pi_{R}$, $\pi_{L}$, and an
expression for $M_+$, in terms of the shell variables. These relations can
then be inserted in the action to give an effective action
depending only on the shell variables. The calculation for the present
case runs precisely parallel to the uncharged case, resulting in
$$
S=\int\! dt\left[\dot{\hat{r}}\left(\sqrt{2M\hat{r}-Q^{2}} -
\sqrt{2M_{+}\hat{r}
- (Q+q)^{2}\,}\right)\right.
$$
\begin{equation}
\left.
- \eta \dot{\hat{r}}\hat{r} \log{\left|\frac{
\sqrt{\hat{r}}-\eta\sqrt{M_{+}-(Q+q)^{2}/2\hat{r}}}{\sqrt{\hat{r}}-\eta
\sqrt{M-Q^{2}/2\hat{r}}}\right|} - M_{+}\right]
\label{seff}
\end{equation}
where $\eta \equiv$ sgn$\,(p)$, and we have now specialized to a massless shell
($m=0$). The canonical momentum is then
\begin{equation}
p_{c} = \sqrt{2M\hat{r}-Q^{2}}-\sqrt{2M_{+} \hat{r}- (Q+q)^{2}}
-\eta \hat{r} \log{\left|\frac{\sqrt{\hat{r}}-\eta \sqrt{M_{+}-(Q+q)^{2}/
2\hat{r}}}{\sqrt{\hat{r}}-\eta\sqrt{M-Q^{2}/2\hat{r}}}\right|}.
\label{pcana}
\end{equation}
We need wish to find the short wavelength solutions which are accurately
described by the WKB approximation. Writing these solutions as
$v(t,r)=e^{iS(t,r)}$ with $S$ rapidly varying, we can make the replacements
$$
p_{c} \rightarrow \frac{\partial S}{\partial r}
\mbox{ \ \ \ ; \ \ \ } M_{+}-M \rightarrow \frac{\partial S}{\partial t}.
$$
$S(t,r)$ satisfies the Hamilton-Jacobi equation, and so is found by computing
classical action along classical trajectories. We first choose the initial
conditions for $S(t,r)$ at $t=0$:
\begin{equation}
S_{k}^{q}(0,r)=kr.
\label{init}
\end{equation}
We have a appended a subscript and a superscript to denote the initial
condition and charge of the solution. The corresponding classical
trajectory has the initial condition $p_{c}=k$ at $t=0$. $S_{k}^{q}(t,r)$ is
then given by
\begin{equation}
S_{k}^{q}(t,r)=k\hat{r}(0) + \int_{\hat{r}(0)}^{r}d\hat{r}\,
p_{c}(\hat{r}) -(M_{+}-M)t.
\label{sola}
\end{equation}
To determine the radiance from the hole we will will only need to consider
the behaviour of the solutions near the horizon. Furthermore, only the most
rapidly varying part of the solutions will contribute to the late-time
radiation. With this in mind, we can write the momentum as (choosing
$\eta =1$ for an outgoing solution)
\begin{equation}
p_{c}(\hat{r}) \approx -\hat{r} \log{\left| \frac{\hat{r}-R_{+}(M_{+},Q+q)}
{(\hat{r}-R_{+}(M,Q))(\hat{r}-R_{-}(M,Q))}\right|}
\label{pnew}
\end{equation}
so that the initial condition becomes
\begin{equation}
k=-\hat{r}(0) \log{ \left| \frac{(\hat{r}(0)-R_{+}(M_{+},Q+q))(\hat{r}(0)
-R_{-}(M_{+},Q+q))}{(\hat{r}(0)-R_{+}(M,Q))(\hat{r}(0)-R_{-}(M,Q))}\right|}.
\label{kinit}
\end{equation}
Similarly, the classical trajectory emanating from $\hat{r}(0)$ is given by
approximately,
$$
t \approx \frac{2}{R_{+}(M_{+},Q+q)-R_{-}(M_{+},Q+q)} \left[
R_{+}(M_{+},Q+q)^{2} \log{\left|\frac{\hat{r}-R_{+}(M_{+},Q+q)}{\hat{r}(0)
-R_{+}(M_{+},Q+q)}\right|}\right.
$$
\begin{equation}
\left. - R_{-}(M_{+},Q+q)^{2} \log{ \left| \frac{
\hat{r}-R_{-}(M_{+},Q+q)}{\hat{r}(0)-R_{-}(M_{+},Q+q)}\right|}\,\right].
\label{geoa}
\end{equation}
These trajectories are in fact null geodesics of the metric
\begin{equation}
ds^2 = -dt^2 + (dr+\sqrt{2M_{+}/r - Q^2}\, dt)^2.
\label{met}
\end{equation}
The relations (\ref{kinit}) and (\ref{geoa}) allow us to determine $M_{+}$ and
$\hat{r}(0)$ in terms of the other variables, so that after integrating
(\ref{sola}) we can obtain an expression for $S_{k}^{q}(t,r)$ as a function
of $k$, $t$, and $r$.
We can now write down an expression for the field operator:
\begin{equation}
\hat{\phi}(t,r)=\int\! dk\, [\hat{a}_{k}v_{k}^{q}(t,r)+\hat{b}_{k}^{\dagger}
v_{k}^{-q}(t,r)^{*}].
\label{phia}
\end{equation}
The modes $v_{k}^{q}(t,r)$ are nonsingular at the horizon, and so the state
of the field is taken to be the vacuum with respect to these modes:
$$
\hat{a}_{k}\left|0_{v}\rangle = \hat{b}_{k} \left|0_{v}\rangle = 0.
\right.\right.
$$
Alternatively, we can consider modes which are positive frequency with
respect to the Killing time $t$. We write these modes as
$u_{k}^{q}(r) e^{-i\omega_{k}t}$ where the $u_{k}^{q}(r)$ are singular at
the horizon, $r=R_{+}(M+\omega_{k},Q+q)$. Then
\begin{equation}
\hat{\phi}(t,r)=\int\! dk\, [\hat{c}_{k}u_{k}^{q}(r) e^{-i\omega_{k}t}
+\hat{d}_{k}^{\dagger}u_{k}^{-q}(r)^{*}e^{i\omega_{k}t}].
\label{phising}
\end{equation}
The two sets of operators are related by Bogoliubov coefficients,
\begin{equation}
\hat{c}_{k} = \int\! dk\, [\alpha_{kk'} \hat{a}_{k'} +\beta_{kk'}\hat{b}_{k'}
^{\dagger}].
\end{equation}
From
(\ref{phia},\ \ref{phising}) $\alpha_{kk'}$ and $\beta_{kk'}$ are found to be
$$
\alpha_{kk'} = \frac{1}{2\pi u_{k}^{q}(r)}\int_{-\infty}^{\infty}\! dt\,
e^{i\omega_{k}t}v_{k'}^{q}(t,r)
$$
\begin{equation}
\beta_{kk'}=\frac{1}{2\pi u_{k}^{q}(r)}\int_{-\infty}^{\infty}\! dt\,
e^{i\omega_{k}t}v_{k'}^{-q}(t,r)^{*}.
\label{boga}
\end{equation}
Here, $r$ is taken to be slightly outside the horizon,
$r=R_{+}(M+\omega_{k},Q+q) + \epsilon$. These coefficients can be evaluated
in the saddle point approximation. Recalling that $v_{k}^{q}(t,r)
=e^{iS_{k}^{q}(t,r)}$, the saddle point equation for $\alpha_{kk'}$ becomes
\begin{equation}
\omega_{k}=-\frac{\partial S_{k'}^{q}}{\partial t} = M_{+}^{q} - M.
\label{alpha}
\end{equation}
This leads to a purely real value of $t$ for the saddle point.
For $\beta_{kk'}$ we have
\begin{equation}
\omega_{k}=\frac{\partial S_{k'}^{-q}}{\partial t}=M-M_{+}^{-q}.
\label{beta}
\end{equation}
\mbox{}From (\ref{kinit},\ \ref{geoa}) we find that the saddle point
value for t has an
imaginary part given by
\begin{equation}
\mbox{Im}(t_{s})=\frac{2\,R_{+}(M-\omega_{k},Q-q)^{2}}{R_{+}(M-\omega_{k},Q-q)
-R_{-}(M-\omega_{k},Q-q)}\,\pi=\frac{1}{2\,T(M-\omega_{k},Q-q)}.
\end{equation}
Therefore,
\begin{equation}
|{\beta_{kk'}/\alpha_{kk'}}|
= \frac{1}{|2\pi u_{k}(r)|}\exp{\left(\omega_{k}/T(M-
\omega_{k},Q-q) + \mbox{Im}[S_{k'}^{-q}(t_{s})^{*}]\right)}.
\end{equation}
The terms in $S_{k'}^{-q}$ which contribute to the second term in the
exponent are
$$
\int_{\hat{r}(0)}^{r} d\hat{r}\, p_{c}(\hat{r}) +\omega_{k}\,\mbox{Im}(t_{s}).
$$
Using (\ref{pnew}-\ref{geoa}) this can be evaluated to give
\begin{equation}
\mbox{Im}[S_{k'}^{-q}(t_{s})^{*}]=\frac{M\omega + \sqrt{M^{2}-Q^{2}\,}
\left(\sqrt{(M-\omega)^{2}-(Q-q)^{2}}-\sqrt{M^{2}-Q^{2}}\right)}
{2\,T(M-\omega,Q-q)\, R_{+}(M,Q)}
\end{equation}
resulting in
\begin{equation}
\left|\frac{\beta_{kk'}}{\alpha_{kk'}}\right|^{2}
=\exp{\left(-\frac{\sqrt{M^{2}-Q^{2}}\, [\omega - \sqrt{(M-\omega)^{2}
-(Q-q)^{2}}+\sqrt{M^{2}-Q^{2}\,}\,]}{T(M-\omega,Q-q)\,R_{+}(M,Q)}\right)}.
\label{bol}
\end{equation}
This is the effective Boltzmann factor governing emission.
Sufficiently far from extremality, when $\omega$, $q \ll
\sqrt{M^{2}-Q^{2}}$, we can expand (\ref{bol}) to give
\begin{equation}
\left|\frac{\beta_{kk'}}{\alpha_{kk'}}\right|^{2} \approx \exp{\left(-\frac{
\omega - \frac{Qq}{R_{+}(M,Q)} + \frac{M^{2}q^{2}+Q^{2}\omega^{2}
-2MQ\omega q}
{2(M^{2}-Q^{2})R_{+}(M,Q)}}{T(M-\omega,Q-q)}\right)}
\end{equation}
as compared to the free field theory result \cite{hawking},
\begin{equation}
\left|\frac{\beta_{kk'}}{\alpha_{kk'}}\right|^{2} =
\exp{\left(-\frac{\omega-\frac{Qq}{R_{+}(M,Q)}}{T(M,Q)}\right)}.
\label{free}
\end{equation}
Near extremality, the self-interaction corrections cause the emission to
differ substantially from (\ref{free}).
We might ask whether it is possible to reach extremality after a finite
number of emissions. Since $T(M-\omega,Q-q)$ appears in the denominator
of the exponent of (\ref{bol}), the transition probability to the extremal
state is in fact zero. We can also ask whether there are transitions to a
meta-extremal ($Q>M$) hole. This would have rather dramatic implications
as the meta-extremal hole is a naked singularity. To address this question
we return to the saddle point equation (\ref{beta}). When $Q>M$, $R_{+}$
and $R_{-}$ become complex. From (\ref{kinit}) we see that a saddle point
solution would require that $k$ be complex, but we do not allow this since a
complete family of initial conditions $S_{k}(0,r)=kr$ was defined with $k$
real. Therefore, in the saddle point approximation the extremal hole is
stable.
Modes with $|\beta /\alpha | > 1$ formally require larger amplitudes
for higher occupation numbers, and thus require special interpretation.
Considering for simplicity the free field form of these
coefficients, (\ref{free}), we see that such modes occur when
$\omega < q Q/ R_+$, that is when the incremental energy gain from
discharging the Coulomb field overbalances the cost of creating
the charged particle. Under these conditions one has dielectric
breakdown of the vacuum, just as for a uniform electric field in
empty space. Since this physics is not our primary concern in the
present work, we shall restrict ourselves to a few remarks.
The occupation factor appearing in the formula for
radiation in these ``superradiant''
modes is negative, but the reflection probability
exceeds unity, so the radiation flux is positive as it should be.
And in general the formulas for physical quantities will appear sensible,
although Fock space occupation
numbers are not. We can avoid superradiance altogether by considering
a model with only
massive charged fundamental particles, and holes with a charge/mass ratio
small compared to the minimal value for fundamental quanta.
Another interesting variant is to consider a {\it magnetically\/} charged
hole interacting with neutral matter. In that case, one simply
puts $q=0$ in the fomulae above (but $Q \ne 0$). One could also consider
the interaction of dyonic holes with charged matter, and other variants
({\it e.g}. dilaton black holes) but we shall not do that here.
\section{Discussion}
We have arrived at our results by what may have appeared to be a
somewhat circuitous route. Inspired by a field theory question, we
calculated the solutions of a single self-gravitating particle at the
horizon, and then passed back to field theory by interpreting the solutions
as the modes of a second quantized field operator. In this section we hope
to clarify the logic of this procedure, and show that it is both correct and
efficient, by demonstrating how a single particle action emerges from the
truncation of a complete field theory.
We can illustrate this explicitly if we consider the simpler model of
spherically symmetric
electromagnetic and charged scalar fields interacting in flat space. Our
goal is to show that the propagator for the scalar field can be expressed as a
Hamiltonian path integral for a single charged shell.
To achieve this, two important
approximations will be made. The first is that the effects of vacuum
polarization will be assumed to be small, so we can ignore scalar loop
diagrams. The second is to assume that the dominant interactions involve
soft photons, so that the difference in the scalar particle's energy before
and after emission or absorption of a photon is small compared to the energy
itself. Thus we expect that our expression will be valid for cases where
the scalar particle has a large energy, so that the energy transfer per
photon is relatively small, and is far from the origin, so that
the classical electromagnetic self energy of the particle is a slowly varying
function of the radial coordinate. Field theory in this domain is in fact
well described by the eikonal approximation, which implements the same
approximations we have just outlined. What follows is then essentially a
Hamiltonian version of the eikonal method.
We start from the action
$$
S=-\frac{1}{4\pi}\int \!d^{4}x\,\left[(\partial_{\mu}-iqA_{\mu})\phi^{*}
\, (\partial^{\mu}+iqA^{\mu})\phi +m^{2}\phi^{*}\phi
+\frac{1}{4}F_{\mu \nu}F^{\mu \nu}\right]
$$
$$
=\int\!dt\,dr\left[\pi_{\phi^{*}}\dot{\phi}^{*}+\pi_{\phi}\dot{\phi}
-\left(\frac{\pi_{\phi^{*}}\pi_{\phi}}{r^2}+r^{2}{\phi^{*}}'
\phi'+m^{2}r^{2}\phi^{*}\phi+\frac{{\pi_{\!A_{r}}}^{2}}{2r^2}
\right)\right.
$$
\begin{equation}
\left.
-A_{t}\left(iq[\pi_{\phi^{*}}\phi^{*}-\pi_{\phi}\phi]-\pi_{\!A_{r}}'
\right)\right].\mbox{ \ \ \ \ }
\end{equation}
Defining the charge density
\begin{equation}
\rho(r)\equiv iq[\pi_{\phi^{*}}(r)\phi^{*}(r)-\pi_{\phi}(r)\phi(r)]
\end{equation}
the solution of the Gauss' law constraint is
\begin{equation}
Q(r)\equiv -\pi_{\!A_{r}}=\int_{0}^{r}\!dr\,\rho(r)
\end{equation}
and so the scalar field Hamiltonian becomes
\begin{equation}
H=\int_{0}^{\infty}\!dr\left[
\frac{\pi_{\phi^{*}}\pi_{\phi}}{r^2}+r^{2}{\phi^{*}}'\phi'+m^{2}\phi^{*}\phi
+\frac{Q(r)^2}{2r^2}\right].
\end{equation}
The fields are now written as second quantized operators:
$$
\hat{\phi}=\int\!\frac{dk}{\sqrt{2\pi\,2\omega_k}}\,
\frac{[\hat{a}_{k} e^{ikr}+\hat{b}^{\dagger}_{k} e^{-ikr}]}{r}
$$
\begin{equation}
\hat{\pi}_{\phi}=i\int\!\frac{dk}{\sqrt{2\pi}}\sqrt{\frac{\omega_k}{2}}\,r\,
[\hat{a}^{\dagger}_{k} e^{-ikr}-\hat{b}_{k} e^{ikr}]
\end{equation}
where $\omega_{k}=\sqrt{k^{2}+m^{2}}$, and we also have $\hat{\phi}^{*}
=\hat{\phi}^{\dagger}$ , $\hat{\pi}_{\phi^{*}}=\hat{\pi_{\phi}}^{\dagger}$. To
ensure that the field is nonsingular at the origin we impose the conditions
$\hat{a}_{-k}=-\hat{a}_{k}$ , $\hat{b}_{-k}=-\hat{b}_{k}$, and take the limits of all $k$
integrals to be from $-\infty$ to $\infty$.
We now write the Hamiltonian in terms of the creation and annihilation
operators. In doing so we shall normal order the operators, which corresponds
to omitting vacuum polarization since we do not allow particle-antiparticle
pairs to be created out of the vacuum.
Also when evaluating $\phi'$ we shall use the geometrical optics
approximation, $(e^{ikr}/r)'\approx ike^{ikr}/r$, valid for $k\gg 1/r$.
Then the quadratic part of the Hamiltonian becomes,
\begin{equation}
\int_{0}^{\infty}\!dr\left[\frac{\hat{\pi}_{\phi^{*}}\hat{\pi}_{\phi}}{r^2}
+r^{2} \hat{\phi}^{*}{'} \hat{\phi}'+m^{2}\hat{\phi}^{*}\hat{\phi}\right]
=\frac{1}{2}\int\!dk\,\omega_{k}[\hat{a}^{\dagger}_{k}\hat{a}_{k} +\hat{b}^{\dagger}_{k}\hat{b}_{k}].
\end{equation}
Next we consider the interaction term. When evaluating this there will arise
factors of $\sqrt{\omega_{k'}/\omega_{k}}$. The essence of the soft photon
approximation is that we replace these factors by $1$, since we are assuming
that $\Delta \omega/\omega \ll 1$ for the emission or absorption of a single
photon. Then, after normal ordering, we can evaluate the charge density to
be:
\begin{equation}
\hat{\rho}(r)=q\int\!\frac{dk\,dk'}{2\pi}[\hat{a}^{\dagger}_{k}\hat{a}_{k'}-\hat{b}^{\dagger}_{k}\hat{b}_{k'}]e^{i(k-k')r}.
\end{equation}
We now wish to calculate matrix elements of the Hamiltonian between one
particle states. A basis of one particle states labelled by position is given
by
\begin{equation}
|r\rangle=\int\!\!\frac{dk}{\sqrt{2\pi}}\,e^{-ikr}\,\hat{a}^{\dagger}_{k}|0\rangle.
\end{equation}
The free part of the Hamiltonian then has matrix elements
\begin{equation}
\langle r_{2}|\,\frac{1}{2}\int\!dk\,\omega_{k}[\hat{a}^{\dagger}_{k}\hat{a}_{k}+\hat{b}^{\dagger}_{k}\hat{b}_{k}]|r_{1}\rangle
=\int\!\frac{dk}{2\pi}\,\omega_{k}[e^{ik(r_{2}-r_{1})}-e^{ik(r_{2}+r_{1})}].
\end{equation}
The second term in the brackets corresponds to the path from $r_1$ to $r_2$
which passes through the origin. These paths will not contribute to local
processes far from the origin, so we drop this term. The matrix elements
of the interaction term for closely spaced points $r_1$ and $r_2$ are:
\begin{equation}
\langle r_{2}|\int_{0}^{\infty}\!\!dr\,\frac{\hat{Q}(r)^{2}}{2r^2}|r_{1}\rangle
=\frac{q^{2}}{2r_{1}}\int\!\frac{dk}{2\pi}e^{ik(r_{2}-r_{1})} .
\end{equation}
Putting these together, we find the matrix elements of the Hamiltonian,
\begin{equation}
\langle r_{2}|\hat{H}|r_{1}\rangle=
\int\! \frac{dk}{2\pi}\, e^{ik(r_{2}-r_{1})}
(\sqrt{k^{2}+m^{2}}+q^{2}/2 r_{1}).
\end{equation}
Now we can follow the standard route which leads from matrix elements of
the Hamiltonian to a path integral expression for the time evolution
operator, with the result
\begin{equation}
\langle r_{f}|e^{-i\hat{H}t}|r_{i}\rangle
=\int_{r(0)=r_{i}}^{r(t)=r_{f}}{\cal D}p\,{\cal D}r\,
e^{i\int_{0}^{t}\!dt'\,(p\dot{r}-\sqrt{p^{2}+m^{2}}-q^{2}/2r)}.
\end{equation}
The action in the exponent is precisely that of a charged shell, with
$q^{2}/2r$ being the electromagnetic self energy.
We now discuss how this analysis can be applied to the case where
we include gravitational interactions. The resulting field Hamiltonian
is much more complex, and so we will not be able to explicitly calculate the
effective shell action. However, the preceding derivation allows us to argue
that were we to do so, we would simply derive the effective action obtained in
section 2. The nature of the black hole radiance calculation makes us believe
that the approximations used to arrive at a shell action are justified. This
is so because for a large ($M\gg m_{p}$) hole the relevant field
configurations are short wavelength solutions moving in a region of
relatively low curvature, and these are the conditions which we argued make
the eikonal approximation valid.
For simplicity, we will consider an uncharged self-gravitating scalar field.
If we truncate to the s-wave we arrive at what is known as the BCMN model,
originally considered in \cite{bcmn} and corrected in \cite{unruh}. The
action is
$$
S=\frac{1}{4\pi}\int\! d^4\!x\, \sqrt{-g}\left[\frac{1}{4}{\cal R}
- \frac{1}{2}g^{\mu \nu}\partial_{\mu}\phi\partial_{\nu}\phi\right]
$$
\begin{equation}
=\int\! dt\, dr\,\left[\pi_{\phi}\dot{\phi}+\pi_{R}\dot{R}+\pi_{L}\dot{L}
-N^{t}({\cal H}_{t}^{\phi}+{\cal H}_{t}^{G})-N^{r}({\cal H}_{r}^{\phi}
+{\cal H}_{r}^{G})\right] -\int\! dt\, M_{ADM}
\label{sphi}
\end{equation}
with
\begin{equation}
{\cal H}_{t}^{\phi}=\frac{1}{2}\left(\frac{{\pi_{\phi}}^{2}}{LR^2}
+\frac{R^2}{L}{\phi '}^{2}\right) \mbox{ \ \ \ ; \ \ \ }
{\cal H}_{r}^{\phi}=\pi_{\phi}\phi'.
\end{equation}
The analog of (\ref{mcona}) is now
\begin{equation}
{\cal M}'=\frac{R'}{L}{\cal H}_{t}^{\phi}+\frac{\pi_{L}}{RL}{\cal H}_{r}^{s}
=\frac{R'}{2L^2}\left(\frac{{\pi_{\phi}}^{2}}{R^2}+R^{2}{\phi'}^{2}\right)
+\frac{\pi_{L}\pi_{\phi}\phi'}{RL}
\label{mphi}
\end{equation}
The Hamiltonian is
\begin{equation}
H=M_{ADM}={\cal M}(\infty)=M+\int_{0}^{\infty}\!dr\left[\frac{R'}{L}
{\cal H}_{t}^{\phi}+\frac{\pi_{L}}{RL}{\cal H}_{r}^{s}\right].
\end{equation}
To obtain an expression for $H$ which depends only on $\phi$ and $\pi_{\phi}$
we must choose a gauge and solve the constraints. We can obtain an explicit
result if we choose the gauge $R=r$, $\pi_{L}=0$. Then, defining
\begin{equation}
h(r)\equiv
\frac{1}{2}\left(\frac{{\pi_{\phi}}^{2}}{r^2}+r^{2}\phi'^{2}\right),
\end{equation}
$L$ is determined from (\ref{mphi}),
\begin{equation}
{\cal M}'(r)=\left(\frac{r}{2}-\frac{r}{2L^2}\right)'=\frac{h(r)}{L^2}
\end{equation}
so
\begin{equation}
\frac{1}{L^2}=-\frac{2M}{r}e^{-2\int_{0}^{r}\!dr'\,h(r')/r'}
+\frac{1}{r}e^{-2\int_{0}^{r}\!dr'\,h(r')/r'}
\int_{0}^{r}\!dr'\,e^{2\int_{0}^{r'}\!dr''\,h(r'')/r''}
\end{equation}
which then leads to
\begin{equation}
H=Me^{-2\int_{0}^{\infty}\!dr\, h(r)/r}
+\int_{0}^{\infty}\!dr\,h(r)e^{-2\int_{r}^{\infty}\!dr'\,
h(r')/r'}.
\label{phiham}
\end{equation}
This generalizes the result of \cite{unruh} to include a nonzero mass $M$
for the pure gravity solution. To make a direct comparison with our work in
the previous section, it would be preferable to obtain the Hamiltonian in
$L=1, R=r$ gauge. This is more difficult and we do not know the explicit
expression. For the moment, though, we are mainly interested in the
qualitative structure of the Hamiltonian, and (\ref{phiham}) will be
sufficient for our purposes. The various nonlocal
terms contained in the Hamiltonian (\ref{phiham}) correspond to gravitons
attaching onto the particle's worldline.
If we expand the exponentials in (\ref{phiham}), we see that
there arise an infinite series of bi-local, tri-local, \ldots, terms resulting
from the non-linearity of gravity. Now we could, in principle, repeat the
analysis which led to an effective shell action for the charged field in
flat space. In that case the calculation could be done with only modest
effort because there was only a single quartic interaction term. In the
present case we would have to sum the infinite series of terms that arise;
our point is that handling all of these terms is cumbersome,
to say the least, and that it is much simpler to proceed as in
section 2.
\section{Multi-particle Correlations}
We have seen how to obtain the self-interaction correction to the probability
of single particle emission. The natural next step would be to obtain similar
results for multi-particle processes, involving some combination of ingoing
and outgoing particles. The ultimate goal, of course, it to construct a
complete S-matrix relating arbitrary in and out states. This brings to the
fore what is usually regarded as the central conceptual puzzle of black hole
physics: does such an S-matrix exist which unitarily relates states described
on ${\cal J}^+$ and ${\cal J}^-$, or is there ``information loss'' is the sense
that pure states on ${\cal J}^-$ can evolve into mixed states on ${\cal J}^+$?
While the latter possibility clearly violates a tenet of quantum physicists,
there is at present no satisfactory proposal as to how the former possibility
can be realized.
If information is preserved in gravitational collapse, and an S-matrix exists,
it will require the existence of intricate correlations between particles on
${\cal J}^+$ which encode the details of the state on ${\cal J}^-$. An
understanding of how this situation might arise is currently precluded by the
fact that essentially nothing is known about how to compute {\em any}
correlations on ${\cal J}^+$, much less those which would preserve all
information.
In free field theory, the state on ${\cal J}^+$ is described by an exactly
thermal density matrix, and no one knows how to go beyond the free field
approximation, except in the case of the single particle processes we have
been discussing. However, we can envision calculating the correlations between
two emitted particles by an extension of our previous methods, simply including
two shells instead of one. Presumably, the correlations would be quite
complicated in the case of short time separation between the particles, but
upon going to to large time separations one would see the later particle being
emitted with a probability corresponding to a hole of mass $M-\omega$, where
$\omega$ is the energy of the earlier particle. In addition, there is the
possibility for the phases of the particles to be correlated even for large
time separation, owing to the fact that the two shells were closely spaced near
the horizon at early times.
There is no obstacle to writing fown the action for two shells,
\begin{equation}
S=\int\! dt\, [p_1 \dot{r_1} +p_2 \dot{r_2} - H(r_1,r_2,p_1,p_2)].
\end{equation}
$H$ is again given by the ADM mass,
which can be expressed in terms of the shell
variables by solving the constraints. The action $S_{k_1 k_2}(t,r_1,r_2)$ can
then be computed for the initial condition
\begin{equation}
S_{k_1 k_2}(0,r_1,r_2)=k_1 r_1 +k_2 r_2
\end{equation}
by integrating the Hamilton-Jacobi equation in precisely the same manner as for
the one particle case. Presumably, all correlations between the particles are
encoded in this action; unfortunately, we do not know the code, for reasons
that we now discuss.
In the single particle case it was possible to make progress because we know
how to pass from the first quantized description in terms of $S_k(t,r)$, to a
field description; namely by writing
\begin{equation}
\phi(t,r)=\sum_{k}\,[a_k e^{iS_k(t,r)}+a_{k}^{\dagger} e^{-iS_k(t,r)}].
\label{field}
\end{equation}
This then put the full power of Bogoliubov transformations at our disposal,
which enabled us to determine the emission probabilities. By constrast, in
the multi-particle case the relation between the first quantized description
and the field description is unclear --- we are not aware of any extension of
(\ref{field}) which takes into account the two particle solutions. This
represents a major obstacle, since a proper interpretation of the theory
requires that we have a field description. On the other hand, it seems most
likely that this problem is merely a technical one, as there is no reason to
believe that the theory is ill defined or inconsistent.
\chapter{Black Hole Entropy}
Almost all researchers agree that a black hole has an entropy equal to one
quarter the area of its event horizon, even though there is no consensus as
to what the entropy represents physically. The most appealing possibility
is that the entropy counts the number of black hole microstates, {\em i.e.}
the black
hole Hamiltonian has $e^{S(M)}$ eigenvalues between $0$ and $M$. Since it
may seem rather remarkable that this could be deduced from the study of free
field theory on a classical background geometry, we will attempt put the
argument in perspective by recalling the corresponding derivation in ordinary
thermodynamics. Suppose we wish to determine the entropy of some body of
matter. To proceed, we would first perform the experiment of placing the
body in contact with a heat bath of temperature $T$, waiting for equilibrium
to occur, and then measuring $E(T)$, the energy of the body. Then, assuming
that equilibrium occurs when the total number of states of the system is
maximized, we have
\begin{equation}
\frac{\partial S_{\mbox{\scriptsize body}}(E)}{\partial E}=\frac{\partial
S_{\mbox{\scriptsize bath}}}{\partial E}.
\end{equation}
Finally, since $\partial S_{\mbox{\scriptsize bath}}/\partial E =
1/T$, we obtain
\begin{equation}
S_{\mbox{\scriptsize body}}(E)=\int_{0}^{E}\frac{dE}{T(E)}.
\end{equation}
Clearly, the derivation relies on two features: a) our ability to measure
$E(T)$, or $T(E)$, b) the assumption that in equilibrium the number of states
is maximized. Now let is return to the black hole case. Obviously, no one
has performed an experiment to see whether a black hole of mass $M$ is in
equilibrium with a heat bath of temperature $1/8\pi M$. The point is that
it appears that any nonsingular state of the black hole satisfies this
property. We saw that by requiring regularity of the stress energy
tensor at the horizon we were inevitably led to consider states which radiate
at the Hawking temperature. The status of point (b) is much less clear. Just
because the hole is in equilibrium with the heat bath we cannot conclude that
the black hole is making frequent transitions among its microstates, and that
the probability distribution of these states is Boltzmann. A rather trivial
example of this behaviour is provided by a perfectly reflecting mirror;
it can certainly be in equilibrium with thermal radiation without satisfying
these other conditions. In fact, we shall se shortly that it does not seem to
be possible to assign a Boltzmann probability distribution to the microstates
of the hole. Nevertheless, let us assume for now that assumption (b) is
justified. Then,
\begin{equation}
S_{\mbox{\scriptsize hole}}(M)=\int_{0}^{M}\!\frac{dM}{1/8\pi M}
=4\pi M^2=\frac{A}{4}.
\end{equation}
The belief that the entropy does count microstates is bolstered by the
existence of a completely unrelated derivation due to Gibbons and Hawking
\cite{gibbons}.
We will not go through the details but simply sketch the idea. Their approach
relies on the fact that the partition function for a system can be expressed
as a path integral over configurations periodic in imaginary time:
\begin{equation}
Z=\mbox{Tr}\,e^{-\beta H}=\int_{\mbox{\scriptsize periodic}}
{\cal D}g_{\mu \nu}e^{-S}.
\end{equation}
Here, the path integral is over Euclidean metrics with periodicity $\Delta
\tau_{E}=i\beta$. One then calculates the path integral in saddle point
approximation and notes that only the analytic continuation of the black
hole geometry with $\beta=8\pi M$ contributes, because other geometries have
a conical singularity at the horizon. After calculating the action and
subtracting the flat space contribution, they obtain:
\begin{equation}
Z(\beta)=e^{A/4}e^{-\beta M}
\end{equation}
leading to the identification $S=A/4$.
This derivation also has serious problems, not least of which is the fact
that the Euclidean path integral for gravity may not even exist due to the
well known conformal instability. Nevertheless, it has the virtue that the
answer agrees with the previous result. In any event, it seems overwhelmingly
likely that the entropy has some sort of deep significance, and the most
natural interpretation is that it counts microstates.
\section{State Counting}
If we are inclined to believe that a black hole has the enormous number of
states $e^{S}$, it behooves us to explain how these states can be understood
in terms of the underlying Hamiltonian. It would be most satisfying if
quantization of this Hamiltonian revealed a discrete set of states whose number
could be counted to yield the entropy. In this section we will describe some
efforts along these lines, and the divergence problems
\cite{thooft} which ensue. In keeping
with the spirit of this thesis, the discussion will be confined to spherically
symmetric configurations. While we have no particular reason to believe that
all of the states can be accounted for by considering only the s-wave, the
sorts of problems that arise seem to afflict the higher partial waves in the
same way. Pure gravity in the s-wave has no dynamical degrees of freedom, and
so to have a nontrivial theory to quantize we include a massless scalar field.
It is a little unsettling that we are forced to include to matter in order to
calculate, since the expression for black hole entropy has no explicit
dependence on the matter content of the world. However, it is possible that
there {\em is} matter dependence, but it can be absorbed into the value of
Newton's constant which appears in the entropy formula
\cite{susskind,larsen}.
As a first approach, we can try to proceed in the same way as if we were
quantizing a soliton in flat space --- by quantizing the quadratic fluctuations
of the field about the background solution. Considering only the quadratic
fluctuations amounts to doing free field theory. For simplicity, we will
work in Schwarzschild coordinates and use modes of definite energy,
$e^{-i\omega(t-r_*)}$, where $r_*=r+2m\log{(r/2M-1)}$, and write:
\begin{equation}
\phi=\sum_{\omega}\,[a_{\omega}e^{-i\omega(t-r_*)}+a_{\omega}^{\dagger}
e^{i\omega(t-r_*)}].
\end{equation}
We consider only the outgoing modes since they alone lead to the divergence
problems. In order to count modes we must make the frequencies discrete in
some way. The easiest thing to do is to impose periodic boundary conditions
at $r=2M+\epsilon$ and $r=L$. The $\epsilon$ is included because the modes
become singular at the horizon, and $L$ is some arbitrarily chosen radius
outside the horizon. We take $L\gg2M\gg\epsilon$.
This leads to the allowed frequencies:
\begin{equation}
\omega_n \approx\frac{\pi n}{M \log{(L/\epsilon)}}.
\end{equation}
This means that there are
\begin{equation}
n=\frac{M\omega \log{(L/\epsilon)}}{\pi}
\end{equation}
single particle states for the black hole with energies between $M$ and
$M+\omega$. The problem is that there is no physical reason to keep $\epsilon$
finite, but the number of states diverges as $\epsilon \rightarrow 0$. The
reason for this behavior is due to the fact that $r_* \rightarrow -\infty$
as $r\rightarrow 2M$, so the modes oscillate an infinite number of times before
they reach the horizon.
It seems quite possible that the infinite number of oscillations is simply a
result of using free field theory, and that once the proper self-interaction
corrections are included a finite result will be obtained. To illustrate this,
let us recall the expression for the canonical momentum of a self-interacting
particle,
\begin{equation}
p_c=\sqrt{2Mr}-\sqrt{2M_+ r}-r\log{\left|\frac{\sqrt{r}-\sqrt{2M_+}}
{\sqrt{r}-\sqrt{2M}}\right|}.
\end{equation}
For an energy eigenstate $M_+=M+\omega$, and the corresponding mode is
\begin{equation}
\psi(r,t)=e^{i\int^{r}\!p_c(r')dr'-i\omega t}.
\label{modes}
\end{equation}
Thus the number of oscillations is finite or infinite depending on whether
$\int^{2(M+\omega)}p_c(r')dr'$ is finite or infinite. The singular part of
$p_c$ as $r\rightarrow 2M$ is:
\begin{equation}
p_c \approx -2(M+\omega) \log{[r-2(M+\omega)]}
\end{equation}
which leads to,
\begin{equation}
\int^{2(M+\omega)} p_c(r') dr' = \mbox{finite}.
\end{equation}
The (incorrect) free field theory result is recovered by expanding in $\omega$.
Then $p_c$ has the singular part
\begin{equation}
p_c=\frac{4M\omega}{r-2M}
\end{equation}
and $\int^{2M}p_c dr$ is again infinite.
Thus by including self-interaction we seem to have solved the divergence
problem. However, there is a major caveat. The mode solutions (\ref{modes})
were obtained using the WKB approximation. As we discussed earlier,
the WKB approximation for a
solution of the form $e^{iS}$ is only valid provided
\begin{equation}
\left|\frac{\partial S}{\partial r}\right|\gg \left|\frac{\partial^2 S}
{\partial r^2}\right|^{\frac{1}{2}},\,
\left|\frac{\partial^3 S}{\partial r^3}\right|
^{\frac{1}{3}} \ldots
\end{equation}
In other words, $S$ should be rapidly oscillating but the rate of change of
oscillation must not be too large. Applying this condition to (\ref{modes})
gives
\begin{equation}
|p_c(r)|\gg \left|\frac{dp_c(r)}{dr}\right|^{\frac{1}{2}}
\end{equation}
or:
\begin{equation}
\left|2(M+\omega)\log{[r-2(M+\omega)]}\right|\gg
\left|\frac{2(M+\omega)}{r-2(M+\omega)}
\right|^{\frac{1}{2}}.
\end{equation}
This condition is not satisfied as $r\rightarrow 2(M+\omega)$. Therefore, we
cannot trust the behavior of (\ref{modes}) near the horizon, including the
conclusion that it oscillates a finite number of times. The most convincing
resolution would be to go beyond the WKB approximation and obtain the correct
result for $\psi$. But, as we have discussed previously, this requires the
resolution of factor ordering problems which we do not know how to solve at
the present time. The most we can say is that the preceding analysis suggests
quite strongly that the divergences are connected with an improper treatment
of the self-interaction of the modes near the horizon.
The breakdown of the WKB approximation occurs when we insist on using modes
of definite energy. Earlier, we saw how to derive a complete set of modes
which were nonsingular at the horizon and are accurately described by the
WKB approximation. These were:
\begin{equation}
u_k(t,r)=e^{iS_k(t,r)}
\label{smoothies}
\end{equation}
where:
\begin{equation}
S_{k}(t,r)=-(2M^2-r^2/2)\log{[1+e^{(k/2M' - t/4M')}]}
\end{equation}
with $M'$ given by (\ref{Mprime}).
The coordinates are those of (\ref{nicemetric}).
Since these modes do
not have definite energy, there is no straightforward way to count states
using the microcanonical ensemble. However, we can imagine putting the black
hole in contact with a distant heat bath at the Hawking temperature. If the
black hole behaved as an ordinary thermodynamic body we could proceed by
computing the partition function, and from that extract the entropy by
standard thermodynamic formulas. Since the partition function is a trace,
\begin{equation}
Z=\mbox{Tr}\,(e^{-\beta H}),
\end{equation}
it is independent of which basis we choose to describe the states, and so the
modes (\ref{smoothies}) are as good as any other.
Our working assumption is that
we can describe all the states as fluctuations about a fixed black hole of mass
$M$. Since in the s-wave there are no purely gravitational fluctuations, the
Fock space built on the modes (\ref{smoothies}) should provide a complete
description of these states. Therefore we can write
\begin{equation}
Z=<0|e^{-\beta H}|0>+\int\! dk_1 <k_1|e^{-\beta H}|k_1>
+\int\! dk_1\, dk_2 <k_1 k_2|e^{-\beta H}|k_1 k_2> + \ldots.
\end{equation}
$H$ is the total energy as measured at infinity, so $e^{-\beta H}$ generates
translations in imaginary time:
\begin{equation}
e^{\beta H} \phi(t) e^{-\beta H}=\phi(t-i\beta).
\end{equation}
This feature makes $Z$ easy to calculate. The first term gives simply
\begin{equation}
<0|e^{-\beta H}|0>=1
\end{equation}
by definition of the vacuum. The next term is more interesting,
\begin{equation}
<k|e^{-\beta H}|k>=\frac{1}{2\pi}\int_{-\infty}^{\infty}\! dr\,
u_{k}^{*}(0,r) u_{k}(-i\beta,r)=\frac{1}{2\pi}\int_{-\infty}^{\infty}\!dr\,
e^{-iS_k(0,r)}e^{iS_k(-i\beta,r)}.
\end{equation}
Let us concentrate on the behavior for large $k$. For $k\gg M$ we have
\begin{equation}
S_k(-i\beta,r) \approx 2M(r-2M)(k/2M' + i\beta/4M') \mbox{ \ \ \ \ ; \ \ \ \ }
M' \approx M+ \frac{1}{2}(r-2M).
\end{equation}
This is strictly valid only near the horizon, but that is the only region in
which we need the solutions since we can always construct wavepackets
localized near the horizon. The point we now wish to stress is that
$<k|e^{-\beta H}|k>$ goes to a (nonzero) constant independent of $k$ as
$k \rightarrow \infty$. The precise value of the constant depends on the form
of the wavepackets, but at any rate
\begin{equation}
\int\! dk <k|e^{-\beta H}|k>=\infty
\end{equation}
so that $Z$ cannot be defined.
There is a simple physical explanation for this behavior. A nonsingular mode
has positive energy density for points outside the horizon, and negative
energy density for points inside. Now imagine being at fixed radius, $r$,
initially outside the horizon, and letting $k$ increase. As $k$ increases, the
total mass inside radius $r$ increases. However, the mass can never increase
past $M=2/r$, since if it did one would be inside the horizon, but the modes
have negative energy here and so cannot lead to an increase in mass. So as
$k$ goes to infinity, the effect on the geometry simply goes to a constant,
which explains why the matrix element of $e^{-\beta H}$ also goes to a
constant.
The lesson to be learned from all of this is presumably that a black hole
cannot be treated as an ordinary thermodynamic body in the sense that its
states are distributed according to a Boltzmann distribution when in
equilibrium with a heat bath. There are simply too many low energy states
localized near the horizon for this distribution to make sense. Let us point
out, though, that this conclusion is based on the assumption that the states are
correctly described by a local quantum field theory. This assumption may be
incorrect, and the divergences may disappear when the correct theory at short
distances, such as string theory, is taken into account. That string theory
plays a crucial role in black hole physics is argued in, for example,
Ref. ~\cite{susskind}.
\chapter{Black Holes and Quantum Tunneling}
The radiation of particles from matter evolving along a classical trajectory
has been heavily studied in recent years. Less well studied is the radiation
accompanying quantum tunneling from one classically allowed trajectory to
another. The following question is of interest: if a matter system impinges
upon a potential barrier with a radiation field in a certain state, what is
the state of the field given that the matter is subsequently observed to be
on the other side of the barrier? A method to answer this question in the
context of false vacuum decay in flat space was developed by Rubakov
\cite{Rub84} and has been generalized to include gravity as well as topology
changing processes \cite{Rub87,Kan89}. The spectrum of radiation is
found by solving an imaginary time Schr\"{o}dinger equation, the occurrence
of which leads to novel features. Instead of solving field equations in real
time, one is naturally led to consider propagation on the Euclidean solution
interpolating between the two classical trajectories. As phase factors in real
time are converted into exponential damping factors in imaginary time, the
resulting particle creation can be distinctly different and is accompanied by
the systematic supression of excitations present before tunneling.
Given this situation, it is natural to ask how the radiation from black holes
might be affected by the presence of tunneling. If we consider a distribution
of matter, initally outside its Schwarzschild radius, which tunnels through a
potential barrier to form a black hole, the conventional calculation
\cite{hawking} of the radiation does not apply. On the other hand, it would be
shocking if the same answer was not obtained for the radiation at late times,
as this is thought to depend only on the hole's late time geometry and not on
its history at early times. Here we compute the radiation for this process
and show that while the Euclidean time evolution has an effect at early times,
it has none at late times so that the standard result is in fact obtained.
In order illustrate the technique of Ref.~\cite{Rub84} in a simpler setting, we
first study the effect of tunneling on another well known radiating system ---
the moving mirror \cite{Dav77}. We show in Sect.~(2)
how an imaginary
time Schr\"{o}dinger equation emerges from a Born-Oppenheimer
approximation, and use this result to calculate the shift in the spectrum of
radiated particles as a result of the tunneling. It is shown that the
initial spectrum is shifted to favor low energy excitations,
as is understood by realizing that the probability to tunnel is increased if
energy is transferred from the radiation to the mirror.
In Sect.~(3) this approach is extended to include gravity in
asymptotically flat space. A WKB approximation to the Wheeler-DeWitt equation,
as considered in \cite{Lap79,Banks85}, is used to obtain an imaginary time
Schr\"{o}dinger equation which can then be solved as before. In
Sect.~(4) we use this result to examine the radiation from a
black hole which is formed by tunneling. In particular, we consider the
tunneling of a false vacuum bubble, a system extensively studied in
Refs.~\cite{Sato81} ---~\cite{Guth87}. This example involves a
complication due to the peculiar structure that arises;
Refs.~\cite{Guth90,polch} show that the sequence of
three-geometries encountered during tunneling can not be stacked together to
form a manifold. Employing a slight modification of the
standard approach, we show how the behaviour of fields on the Euclidean
Schwarzschild manifold protects the late time radiation from being affected
by tunneling. An intuitive reason for this is that the bubble's tunneling
probability
is unchanged by the presence of Hawking radiation, which involves the creation
of pairs of particles with zero total energy.
\section{Tunneling Mirror}
\label{secmir}
Consider a mirror moving in a one dimensional potential in the presence of a
massless scalar field. The Schr\"{o}dinger equation for this system is
\begin{equation}
[\hat{H}_{m}+\hat{H}_{\phi}]\Psi[\phi,x_{m};t]=i\frac{\partial}{\partial
t}\Psi[\phi,x_{m};t]
\end{equation}
where
\begin{equation}
\hat{H}_{m}=-\frac{1}{2m}\frac{\partial^{2}}{\partial x_{m}^{\;2}}+V(x_{m})
\end{equation}
and
\begin{equation}
\hat{H}_{\phi}=\frac{1}{2}\int_{x_{m}}^{\infty}dx\left[-\frac{\delta^{2}}
{\delta\phi(x)^{2}}+\left(\frac{d\phi}{dx}\right)^{2}\right].
\end{equation}
Note that $\Psi$ is a function of the mirror coordinate $x_{m}$, and a
functional of the field configuration $\phi(x)$. The mirror boundary
condition is imposed by demanding that the field vanish at $x_{m}$,
\begin{equation}
\Psi[\phi,x_{m};t]=0\; \mbox{ \ if \ } \phi(x_{m})\neq0.
\end{equation}
The system is solved by assuming that the backreaction of the field on the
mirror is a small perturbation of the mirror's motion, and that the mass and
momenta of the mirror are large enough that it can be described by a well
localized wave packet. In this domain the system admits a Born-Oppenheimer
approximation, which amounts to an expansion in $1/m$. In particular, we seek
a solution to the time independent Schr\"{o}dinger equation
\begin{equation}
[\hat{H}_{m}+\hat{H}_{\phi}]\Psi[\phi,x_{m}]=E\Psi[\phi,x_{m}]
\label{e:ti}
\end{equation}
valid to zeroth order in $1/m$. Following Refs.~\cite{Rub84,Banks85} the
Born-Oppenheimer approximation is implemented by writing $\Psi$ in the form
\begin{equation}
\Psi[\phi,x_{m}]=\psi_{VV}(x_{m})\,e^{iS(x_{m})}\,\chi[\phi,x_{m}]
\end{equation}
where $\psi_{VV}$ is a slowly varying function to be identified with the
Van Vleck determinant. To lowest order in $1/m$, (\ref{e:ti}) reduces to
the Hamilton-Jacobi equation.
\begin{equation}
\frac{1}{2m}\left(\frac{dS}{dx_{m}}\right)^{2}+V(x_{m})=E
\label{e:hj}
\end{equation}
since $dS/dx_{m}$, $V(x_{m})$ and $E$ are all of order $m$.
To zeroth order:
\begin{equation}
-\frac{i}{2m}\,\frac{d^{2}S}{dx_{m}^{\;2}}\,\psi_{VV}\,\chi[\phi,x_{m}]
-\frac{i}{m}\frac{dS}{dx_{m}}\frac{d\psi_{VV}}{dx_{m}}\chi[\phi,x_{m}]
\end{equation}
$$
-\frac{i}{m}\psi_{VV}\frac{dS}{dx_{m}}\frac{\partial}{\partial
x_{m}}\chi[\phi,x_{m}]
+ \psi_{VV}\,\hat{H}_{\phi}\,\chi[\phi,x_{m}]=0.
$$
$\psi_{VV}$ is chosen so that the first two terms cancel, leaving
\begin{equation}
\hat{H}_{\phi}\, \chi[\phi,x_{m}]=\frac{i}{m}\frac{dS}{dx_{m}}
\frac{\partial}{\partial x_{m}}\chi[\phi,x_{m}].
\end{equation}
This can be put in a familiar form by defining the time variable $\tau(x_{m})$.
In a classically allowed region, where $E-V(x_{m})>0$ and $dS/dx_{m}$ is real,
$ \tau $ is defined by
\begin{equation}
\frac{d\tau}{d x_{m}} = \frac{m}{dS/dx_{m}}
\mbox{ \ \ \ allowed regions}
\end{equation}
whereas in a classically forbidden region with $dS/dx_{m}$ imaginary,
\begin{equation}
\frac{d\tau_{E}}{dx_{m}} = i \frac{m}{dS/dx_{m}}
\mbox{ \ \ \ forbidden regions.}
\end{equation}
The resulting zeroth order equations for $\phi$ are:
\begin{equation}
\hat{H}_{\phi} \,\chi[\phi,\tau] = i \frac{\partial}{\partial \tau}\chi[\phi,
\tau] \mbox{ \ \ \ allowed regions}
\end{equation}
\begin{equation}
-\hat{H}_{\phi} \,\chi[\phi,\tau_{E}] = \frac{\partial}{\partial \tau_{E}}
\chi[\phi,\tau_{E}] \mbox{ \ \ \ forbidden regions.}
\end{equation}
These are the fundamental equations governing the evolution of the scalar
field in the presence of the mirror. In the allowed regions we have
recovered the time-dependent Schr\"{o}dinger equation with the postion of the
mirror playing the role of a clock, whereas in the forbidden regions we have
obtained a diffusion equation, which we shall refer to as the Euclidean
Schr\"{o}dinger equation, with the Euclidean time $\tau_{E}$ measuring the
position of the mirror in the potential barrier.
\renewcommand\floatpagefraction{.9}
\renewcommand\topfraction{.9}
\renewcommand\bottomfraction{.9}
\renewcommand\textfraction{.1}
\begin{figure}[htb]
\centerline{\psfig{file=bubfig1.epsi,width=\hsize,angle=-90}}
\caption{A generic mirror potential. The turning points for energy E are
indicated.}
\label{bubfig1}
\end{figure}
Now, choose the potential to be of the form illustrated in Fig. ~\ref{bubfig1}
and let the
mirror come from right to left. In the allowed region to the right of
$x_{m}^{i}$
the state $\chi[\phi,\tau]$ obeys the normal Schr\"{o}dinger equation, and so
standard methods can be used to find $\chi[\phi,\tau^{i}]$. Between
$x_{m}^{i}$ and $x_{m}^{f}$ the mirror is in a forbidden region, so the
state evolves according to
\begin{equation}
-\frac{1}{2}\int_{x_{m}(\tau_{E})}
^{\infty} dx \left[-\frac{\delta^{2}}{\delta \phi(x)^{2}}
+ \left(\frac{d\phi}{dx}\right)^{2}\right] \chi[\phi,\tau_{E}]
=\frac{\partial}{\partial \tau_{E}} \chi[\phi,\tau_{E}]
\end{equation}
with $\chi[\phi,\tau_{E}^{i}]=\chi[\phi,\tau^{i}]$.
We wish to solve this equation in order to find the state at the final turning
point $x_{m}^{f}$. It is useful to transform the mirror to rest by defining the
coordinate
\begin{equation}
y(x,\tau_{E})=x-x_{m}(\tau_{E})
\end{equation}
in terms of which the Euclidean Schr\"{o}dinger equation is
\begin{equation}
-\frac{1}{2} \int_{0}^{\infty} dy \left[ -\frac{\delta^{2}}{\delta \phi(y)
^{2}} + 2\,\frac{dx_{m}}{d\tau_{E}} \frac{d\phi}{dy}\frac{\delta}{\delta
\phi(y)} + \left(\frac{d\phi}{dy}\right)^{2}\right] \chi[\phi,\tau_{E}]
=\frac{\partial}{\partial \tau_{E}} \chi[\phi,\tau_{E}]
\end{equation}
or
\begin{equation}
-\hat{H}_{\phi}^{E}(\tau_{E}) \chi[\phi,\tau_{E}] =
\frac{\partial}{\partial \tau_{E}} \chi[\phi,\tau_{E}].
\end{equation}
The solution is
\begin{equation}
\chi[\phi,\tau_{E}]=T\exp\left[-\int_{\tau_{E}^{i}}^{\tau_{E}}
\hat{H}_{\phi}^{E}(\tau_{E}^{'}) d\tau_{E}^{'}\right] \chi[\phi,\tau_{E}^{i}]
=\hat{U}_{E}(\tau_{E},\tau_{E}^{i})\, \chi[\phi,\tau_{E}^{i}].
\end{equation}
Here T represents time ordering with respect to $\tau_{E}^{'}$.
The crucial point
is that the Euclidean time evolution operator, $\hat{U}_{E}$, is non-unitary.
This is natural since we know that wavefunctions decay exponentially during
tunneling. If $\hat{U}_{E}$ was unitary, the easiest way to calculate it
would be to transform to the Heisenberg picture, solve the field equations
mode by mode, and compute Bogolubov coefficients. However, as emphasized in
Ref.~\cite{Rub84} the non-unitarity of $\hat{U}_{E}$ implies that the
Schr\"{o}dinger and Heisenberg pictures are inequivalent, making the standard
method inapplicable. Instead, one can use the method developed in
Ref.~\cite{Rub84} which closely resembles the standard one but is more general.
We first describe the state right before tunneling. For convenience, set
$x_{m}^{i}=\tau^{i}=\tau_{E}^{i}=0$. Let $\xi_{\omega}(x,\tau)$ be a complete
set of positive norm solutions to the Klein-Gordon equation which vanish
vanish at the mirror:
\begin{equation}
\left[-\frac{\partial^{2}}{\partial \tau^{2}} + \frac{\partial^{2}}
{\partial x^{2}}\right] \xi_{\omega}(x,\tau)=0
\end{equation}
\begin{equation}
i \int dx \left[\xi_{\omega}^{*}(x,\tau)\,\frac{\partial}{\partial \tau}
\xi_{\omega^{'}}(x,\tau)\,
-\frac{\partial}{\partial \tau}\xi_{\omega}^{*}(x,\tau)\xi_{\omega^{'}}
(x,\tau)\right]
=\delta_{\omega \omega^{'}}
\end{equation}
\begin{equation}
\xi_{\omega}(x_{m}(\tau),\tau)=0.
\end{equation}
The set of allowed frequencies $\omega$ is taken to be discrete, and
$\sum_{\omega}$ represents summation over this set. The field operators can
then be expanded in terms of these modes:
\begin{equation}
\hat{\phi}(x,\tau)=\sum_{\omega}\left[\hat{a}_{\omega} \xi_{\omega}(x,\tau)
+\hat{a}_{\omega}^{\dagger}\xi_{\omega}^{*}(x,\tau)\right]
\end{equation}
\begin{equation}
\hat{\pi}_{\phi}(x,\tau)=\frac{\partial}{\partial \tau}\hat{\phi}(x,\tau)
=\sum_{\omega}\left[
\hat{a}_{\omega}\frac{\partial}{\partial \tau}
\xi_{\omega}(x,\tau)+\hat{a}_{\omega}^{\dagger}
\frac{\partial}{\partial \tau}\xi_{\omega}(x,\tau)\right]
\end{equation}
with $\left[\hat{a}_{\omega},\hat{a}_{\omega'}^{\dagger}\right]=\delta_{
\omega \omega '}$.
Now define Euclidean fields $\hat{\phi}^{E}(y,\tau_{E})$,
$\hat{\pi}_{\phi}^{E}(y,\tau_{E})$ which agree with $\hat{\phi}(x,\tau)$,
$\hat{\pi}_{\phi}(y,\tau)$ at $\tau=\tau^{E}=0$, but evolve according to
\begin{equation}
\hat{\phi}^{E}(y,\tau_{E})=\hat{U}_{E}^{-1}(\tau_{E},0)\,\hat{\phi}^{E}(y,0)\,
\hat{U}_{E}(\tau_{E},0)
\label{e:phi}
\end{equation}
\begin{equation}
\hat{\pi}_{\phi}^{E}(y,\tau_{E})=\hat{U}_{E}^{-1}(\tau_{E},0)\,
\hat{\pi}_{\phi}^{E}(y,0)\, \hat{U}_{E}(\tau_{E},0).
\label{e:pi}
\end{equation}
We will calculate $\hat{U}_{E}(\tau_{E},0)$ by first finding $\hat{\phi}
^{E}(y,\tau_{E})$, $\hat{\pi}^{E}_{\phi}(y,\tau_{E})$.
The field equations for these operators are
\begin{equation}
\frac{\partial \hat{\phi}^{E}}{\partial \tau_{E}} = -\left[\hat{\phi}^{E},
\hat{H}_{\phi}^{E}\right]=-i\hat{\pi}_{\phi}^{E}+\frac{dx_{m}}{d\tau_{E}}
\frac{\partial \hat{\phi}^{E}}{\partial y}
\end{equation}
\begin{equation}
\frac{\partial \hat{\pi}_{\phi}^{E}}{\partial \tau_{E}}=
-\left[\hat{\pi}_{\phi}^{E},\hat{H}_{\phi}^{E}\right]=
-i\frac{\partial^{2} \hat{\phi}^{E}}{\partial y^{2}}
+\frac{dx_{m}}{d\tau_{E}}\frac{\partial \hat{\pi}_{\phi}^{E}}{\partial y}.
\end{equation}
So
\begin{equation}
\hat{\pi}_{\phi}^{E}=i\left(\frac{\partial \hat{\phi}^{E}}{\partial \tau_{E}}
-\frac{dx_{m}}{d\tau_{E}}\frac{\partial \hat{\phi}^{E}}{\partial y}\right)
\end{equation}
and
\begin{equation}
\frac{\partial^{2} \hat{\phi}^{E}}{\partial \tau_{E}^{\;2}}
+\left[1+\left(\frac{dx_{m}}{d\tau_{E}}\right)^{2}\right]\frac{\partial^{2}
\hat{\phi}^{E}}{\partial y^{2}} - 2\frac{dx_{m}}{d\tau_{E}}
\frac{\partial^{2} \hat{\phi}^{E}}{\partial y \partial \tau_{E}}-\frac
{d^{2}x_{m}}{d\tau_{E}^{\;2}}\frac{\partial \hat{\phi}^{E}}{\partial y}=0.
\label{e;phi}
\end{equation}
Equation (\ref{e;phi}) can be obtained by varying the action
\begin{equation}
S=\frac{1}{2}\int dy\, d\tau_{E}\, \sqrt{g_{E}\,}\, g_{E}^{\mu \nu}
\partial_{\mu}\phi\partial_{\nu}\phi
\end{equation}
with the Euclidean metric
\begin{equation}
ds_{E}^{2}=g^{E}_{\mu \nu}dx^{\mu}dx^{\nu}=d\tau_{E}^{\;2}+2\,\frac{dx_{m}
}{d\tau_{E}}\,dx\,d\tau_{E}+dx^{\;2}.
\end{equation}
$\hat{\phi}^{E}$, $\hat{\pi}_{\phi}^{E}$ can be expanded in terms of modes
$f_{\omega}$ which satisfy the Euclidean Klein-Gordon equation (\ref{e;phi})
and which vanish at $y=0$,
\begin{equation}
\hat{\phi}^{E}(y,\tau_{E})=\sum_{\omega}\hat{b}_{\omega}f_{\omega}(y,
\tau_{E})
\label{e:sta}
\end{equation}
\begin{equation}
\hat{\pi}_{\phi}^{E}(y,\tau_{E})=i\sum_{\omega}\hat{b}_{\omega}
\left(\frac{\partial}{\partial \tau_{E}}f_{\omega}(y,\tau_{E})
-\frac{dx_{m}}{d\tau_{E}}\frac{\partial}{\partial y}f_{\omega}(y,\tau_{E})
\right).
\end{equation}
As the Euclidean Klein-Gordon equation is elliptic, one cannot in general
impose Cauchy boundary conditions at $\tau_{E}=0$ on $f_{\omega}$. The
resulting solutions would not satisfy the mirror boundary condition. With the
appropriate boundary conditions, either Dirichlet or Neumann, imposed at
$\tau_{E}=0$ and $\tau_{E}=\tau_{E}^{f}$, a detailed calculation is, of
course, required to find $f_{\omega}$ for a generic mirror trajectory. We shall
take the solutions as given and only use their specific forms in a region far
from the mirror, where they are simple.
Now, using the condition that the two sets of operators $\hat{\phi}$,
$\hat{\pi}_{\phi}$ and $\hat{\phi}^{E}$, $\hat{\pi}_{\phi}^{E}$ are equal at
$\tau=\tau_{E}=0$, and taking inner products, the operators $\hat{b}_{\omega}$
can be expressed as a linear combination of $\hat{a}_{\omega}$,
$\hat{a}_{\omega}^{\dagger}$:
\begin{equation}
\hat{b}_{\omega}=\sum_{\omega'}\left[\alpha_{\omega \omega'}\hat{a}_{\omega'}
+\beta_{\omega \omega'}\hat{a}_{\omega'}^{\dagger}\right].
\end{equation}
Then using
\begin{equation}
\hat{\phi}^{E}(y,\tau_{E}^{f})=\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,
\hat{\phi}_{E}(y,0)\,\hat{U}_{E}(\tau_{E}^{f},0)=\hat{U}_{E}^{-1}
(\tau_{E}^{f},0)\,
\hat{\phi}(y,0)\,\hat{U}_{E}(\tau_{E}^{f},0)
\end{equation}
and the analogous expression for $\hat{\pi}_{\phi}^{E}$, the following
equations for $\hat{U}^{E}$ are obtained:\pagebreak
$$
\sum_{\omega}\sum_{\omega'}\left[\alpha_{\omega \omega'} \hat{a}_{\omega'}
+\beta_{\omega \omega'} \hat{a}_{\omega'}^{\dagger}\right]
f_{\omega}(y,\tau_{E}^{f})
$$
\begin{equation}
=\sum_{\omega}\left[\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}\,
\hat{U}_{E}(\tau_{E}^{f},0)\,\xi_{\omega}(y,0)
+\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}^{\dagger}\,
\hat{U}_{E}(\tau_{E}^{f},0)\,\xi_{\omega}^{*}(y,0)\right]
\end{equation}
and
$$
i\sum_{\omega}\sum_{\omega'}\left[\alpha_{\omega \omega'}\hat{a}_{\omega'}
+\beta_{\omega \omega'}\hat{a}_{\omega'}^{\dagger}\right]
\frac{\partial}{\partial \tau_{E}}f_{\omega}(y,\tau_{E}^{f})
$$
\begin{equation}
=\sum_{\omega}\left[\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}\,
\hat{U}_{E}(\tau_{E}^{f},0)\,\frac{\partial}{\partial \tau}\xi_{\omega}(y,0)
+\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}^{\dagger}\,
\hat{U}_{E}(\tau_{E}^{f},0)\,\frac{\partial}{\partial \tau}
\xi_{\omega}^{*}(y,0)
\right].
\end{equation}
Again taking inner products, this leads to relations of the form
\begin{equation}
\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}\,\hat{U}_{E}(\tau_{E}^{f},0)
=\sum_{\omega'}\left[u_{\omega \omega'}\hat{a}_{\omega'}
+v_{\omega \omega'}\hat{a}_{\omega'}^{\dagger}\right]
\label{e:tev}
\end{equation}
\begin{equation}
\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}^{\dagger}\,
\hat{U}_{E}(\tau_{E}^{f},0)=\sum_{\omega'}\left[w_{\omega \omega'}
\hat{a}_{\omega'}+z_{\omega \omega'}\hat{a}_{\omega'}\right].
\end{equation}
Then it can be shown that \cite{Rub84}
\begin{equation}
\hat{U}_{E}(\tau_{E}^{f},0)=\mbox{ const. }\times:\exp\sum_{\omega}
\sum_{\omega'}\left[\frac{1}{2}D_{\omega \omega'}\hat{a}_{\omega}^{\dagger}
\hat{a}_{\omega'}^{\dagger}+F_{\omega \omega'}\hat{a}_{\omega}\hat{a}
_{\omega'}+\frac{1}{2}G_{\omega \omega'}\hat{a}_{\omega}\hat{a}_{\omega'}
\right]:
\end{equation}
where the matrices $D$, $F$, and $G$ are defined by
\begin{equation}
D=vz^{-1}\mbox{ \ ; \ } F=\left(z^{T}\right)^{-1}-1\mbox{ \ ; \ }G=-z^{-1}w.
\end{equation}
The state after tunneling is then determined,
\begin{equation}
\left|\chi(\tau_{E}^{f})\right\rangle = \hat{U}_{E}(\tau_{E}^{f})
\left|\chi(0)\right\rangle
\label{end}
\end{equation}
and is expressed in terms of occupation numbers with respect to the modes
$\xi_{\omega}(y,0)$, where now $y=x-x_{m}^{f}$. All of the information about
the final state is contained in the matrices $D$, $F$, and $G$, which are in
turn given in terms of inner products between the modes $f_{\omega}$ and
$\xi_{\omega}$.
As a simple application of these formul{\ae} we will calculate the shift in the
spectrum of outgoing particles which are far from the mirror at the time of
tunneling. It is assumed that the mirror was initially at rest and the field
in its ground state. The mirror subsequently accelerates in the potential
$V(x_{m})$ until it reaches the classical turning point $x_{m}^{i}$. It is well
known that as a result of the mirror's acceleration, a flux of outgoing
particles is created whose spectrum is calculable by standard methods
\cite{Dav77}. Outgoing particles far from the mirror are wavepackets composed
of superpositions of plane waves,
\begin{equation}
\xi_{\omega}(x,\tau)=\frac{1}{2\sqrt{\omega}}\,e^{-i\omega(\tau-x)}
\end{equation}
The spectrum of outgoing particles located at $x=\bar{x}\gg\omega^{-1}$
at $\tau=0$ is written as
\begin{equation}
\sum_{\{n_{\omega}\}}S_{\bar{x}}\left(\{n_{\omega}\}\right)\left|\{n_{\omega}\}
\right\rangle
\end{equation}
where $\{n_{\omega}\}$ is a set of occupation numbers and $S_{\bar{x}}\left(
\{n_{\omega}\}\right)$ is the amplitude for the set to occur.
Far from the mirror, the modes $f_{\omega}$ are easy to calculate since the
mirror boundary condition is irrelevant. They are of two types,
$$
f_{\omega}^{\mbox{--}}=\frac{1}{2\sqrt{\omega}}\,e^{-\omega \tau_{E}+i\omega x}
=\frac{1}{2\sqrt{\omega}}\,e^{-\omega \tau_{E}+i\omega \left(y+x_{m}(\tau_{E}
)\right)}
$$
\begin{equation}
f_{\omega}^{\mbox{+}}=\frac{1}{2\sqrt{\omega}}\,e^{\omega \tau_{E}+i\omega x}
=\frac{1}{2\sqrt{\omega}}\,e^{\omega \tau_{E}+i\omega \left(
y+x_{m}(\tau_{E})\right)}
\end{equation}
Then $\hat{\phi}$, $\hat{\pi}$ and $\hat{\phi}^{E}$, $\hat{\pi}_{\phi}^{E}$
are equal at $\tau=\tau_{E}=0$ if
\begin{equation}
\hat{b}_{\omega}^{\mbox{--}}=\hat{a}_{\omega} \mbox{ \ ; \ }
\hat{b}_{\omega}^{\mbox{+}}=\hat{a}_{\omega}^{\dagger}.
\end{equation}
Equation (\ref{e:tev}) gives:
$$
\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}\,\hat{U}_{E}(\tau_{E}
^{f}),0)
=e^{-\omega \tau_{E}^{f}+i\omega x_{m}^{f}}\,\hat{a}_{\omega}
$$
\begin{equation}
\hat{U}_{E}^{-1}(\tau_{E}^{f},0)\,\hat{a}_{\omega}^{\dagger}\,
\hat{U}_{E}(\tau_{E}^{f},0)=e^{\omega \tau_{E}^{f}+i\omega x_{m}^{f}}
\,\hat{a}_{\omega}^{\dagger}
\end{equation}
leading to
\begin{equation}
D=G=0 \mbox{ \ ; \ } F_{\omega \omega'}=\left(e^{-\omega \tau_{E}^{f}
-i\omega x_{m}^{f}}-1\right)\delta_{\omega \omega'}
\end{equation}
and
$$
\hat{U}_{E}(\tau_{E}^{f},0)=\mbox{ const. }\times :\exp\sum_{\omega}
\left[e^{-\omega \tau_{E}^{f}-i\omega x_{m}^{f}}-1\right]\hat{a}_{\omega}
^{\dagger}\hat{a}_{\omega}:
$$
\begin{equation}
=\mbox{ const. }\times:\exp\sum_{\omega}\left[e^{-i\omega x_{m}^{f}}
-1\right]\hat{a}_{\omega}^{\dagger}\hat{a}_{\omega}::\exp
\sum_{\omega}\left[e^{-\omega \tau_{E}^{f}}-1\right]\hat{a}_{\omega}^{\dagger}
\hat{a}_{\omega}:
\end{equation}
The first factor is a translation operator which expresses the state in terms
of the modes $\xi_{\omega}(x,0)$ instead of $\xi_{\omega}(x+x_{m}^{f},0)$, and
the second factor acts on a state $\left|\{n_{\omega}\}\right\rangle$ to give
$e^{-E\left(\{n_{\omega}\}\right)\tau_{E}^{f}}\left|\{n_{\omega}\}
\right\rangle$, where $E\left(\{n_{\omega}\}\right)=\sum n_{\omega}\omega$
is the energy of the state. Therefore, the state after tunneling is
\begin{equation}
\mbox{const. }\times\sum_{\{n_{\omega}\}}e^{-E\left(\{n_{\omega}\}\right)
\tau_{E}^{f}}\,S_{\bar{x}}\left(\{n_{\omega}\}\right)\left|\{n_{\omega}\}
\right\rangle.
\label{e:spec}
\end{equation}
The result of the tunneling is simply to shift the spectrum from $S_{\bar{x}}$
to $e^{-E({n_{\omega}})\tau_{E}^{f}}S_{\bar{x}}$.
It is not difficult to understand this result. Since the total energy is fixed
, the state before tunneling is given by a superpostion of the various ways
of distributing the energy between the mirror and the radiation. As the
mirror's probability to tunnel depends exponentially on its energy, we expect
an inverse exponential correlation between tunneling and energy in radiation.
Thus an observer measuring the spectrum of radiation, conditional on the mirror
tunneling, finds the result (\ref{e:spec}).
Far from the mirror the shift in the
spectrum depends only on $\tau_{E}^{f}$, the amount of Euclidean time spent
during tunneling. This is because the tunneling amplitude in the WKB
approximation is $e^{-S}$, and the derivative of $S$ with respect to energy is
just the Euclidean time.
If we were to identify the Euclidean time with an inverse temperature, the
shift would become a Boltzmann factor. This makes it easy to generate thermal
distributions of radiation. Specifically, if the distribution before tunneling
was a constant, then after tunneling tracing over the states of the mirror
would yield a thermal density matrix for the radiation. A number of authors
have been led by this fact to seek a connection between the thermal radiation
that arises in the contexts of cosmology and black holes and an occurrence of
tunneling \cite{Kan89,Bro91,Cas92}.
Such a connection relies upon assumptions about what is on the
other side of the barrier and what the spectrum of radiation is there. In this
work we only consider situations
where there is a well defined classical trajectory
on either side of the barrier; we are interested in the case in which there is
collapsing matter on side of the barrier and a black hole on the other. The
treatment of this process requires an extension of the previous method to
include gravity.
\section{Application to Gravity}
\label{secgrav}
In this section we make a WKB approximation to gravity in a manner which
directly parallels that for the moving mirror. The action for gravity plus
matter takes the form
$$
S=\frac{m_{p}^{\,2}}{16\pi}\int d^{4}x\sqrt{-g}\left({\cal R}-2\Lambda\right)
+S_{M}+\mbox{ boundary terms}
$$
\begin{equation}
=\int d^{4}x\left(\pi_{\phi_{i}}\dot{\phi}^{i}+\pi_{ij}\dot{h}^{ij}
-N^{t}{\cal H}_{t}-N_{i}{\cal H}^{i}\right)+\mbox{ boundary terms}.
\end{equation}
The Wheeler-DeWitt equation resulting from this action is
\begin{equation}
\hat{{\cal H}}_{t}\Psi=\left[-\frac{16\pi}{m_{p}^{\;2}}G_{ijkl}
\frac{\delta}{\delta h_{ij}}\frac{\delta}{\delta h_{kl}}-\frac{m_{p}^{\;2}}
{16\pi}h^{\frac{1}{2}}\left(^{3}{\cal R}-2\Lambda\right)+\hat{{\cal H}}_{t_{M}}
\right]\Psi=0
\end{equation}
Proceeding as before, we seek a semiclassical
solution of the form
\begin{equation}
\Psi\left[h_{ij},\phi_{i}\right]=\psi_{VV}[h_{ij}]\,e^{im_{p}^{\;2}S[h_{ij}]}
\,\chi\left[\phi_{i},h_{ij}\right].
\end{equation}
At first order the Einstein-Hamilton-Jacobi equation is obtained:
\begin{equation}
\frac{16\pi}{m_{p}^{\;2}}G_{ijkl}\frac{\delta S}{\delta h_{ij}}\frac{\delta S}
{\delta h_{kl}}-\frac{m_{p}^{2\;}}{16\pi}h^{\frac{1}{2}}\left(^{3}{\cal R}-
2\Lambda \right)=0.
\label{e:ehj}
\end{equation}
Zeroth order yields
\begin{equation}
-\frac{16\pi}{m_{p}^{\;2}}i\,G_{ijkl}\frac{\delta S}{\delta h_{ij}}\frac{\delta
\chi}{\delta h_{kl}}+\hat{{\cal H}}_{t_{M}}\chi=0
\label{e:sc}
\end{equation}
provided $\psi_{VV}$ satisfies
\begin{equation}
G_{ijkl}\frac{\delta^{2}S}{\delta h_{ij} \delta h_{ij}}\psi_{VV}
+G_{ijkl}\frac{\delta S}{\delta h_{ij}}\frac{\delta \psi_{VV}}{\delta h_{kl}}
=0.
\end{equation}
The momentum constraints at first order are
\begin{equation}
\left(\frac{\delta S}{\delta h_{ij}}\right)_{|j}=0
\end{equation}
and at zeroth order are
\begin{equation}
2i\left(\frac{\delta \chi}{\delta h_{ij}}\right)_{|j}+\hat{{\cal H}}_
{M}^{i}\,\chi=0.
\label{e:mo}
\end{equation}
Equations (\ref{e:sc}) and (\ref{e:mo}) describe how the matter wave function
evolves as the spatial geometry changes. Quantum field theory in curved space
can be recovered by writing $\chi$'s dependence on $h_{ij}$ in terms of a
time functional $\tau[x;h_{ij}]$, and by reintroducing a lapse $N^{\tau}$ and
shift $N_{i}$, demanding that they obey
\begin{equation}
G_{ijkl}\frac{\delta S}{\delta h_{ij}}=\frac{m_{p}^{\,2}}{16\pi N^{\tau}}
\left(\int dy \frac{\delta h_{kl}}{\delta \tau[y;h_{ab}]}
-N_{i|j}-N_{j|i}\right).
\label{e:con}
\end{equation}
Then
\begin{equation}
-i\frac{16\pi}{m_{p}^{\;2}}\int\left(N^{\tau}G_{ijkl}\frac{\delta S}
{\delta h_{ij}} \frac{\delta \chi}{\delta h_{kl}}+2iN_{i}\left(\frac{\delta
\chi}{\delta h_{ij}}\right)_{|j}\right)=-i\int\frac{\delta h_{ij}}{\delta
\tau}\frac{\delta \chi}{\delta h_{ij}}
\end{equation}
so that the equation for $\chi$ becomes
\begin{equation}
\int d^{3}x
\left[N^{\tau}\hat{{\cal H}}_{t_{M}}+N_{i}\hat{{\cal H}}_{M}^{i}\right]
\chi[\phi_{i};\tau]= i\frac{\partial}{\partial \tau}\chi[\phi_{i};\tau].
\label{e:scg}
\end{equation}
The condition (\ref{e:con}) agrees with the classical relation between
$\pi_{ij}$ and $h_{ij}$, demonstrating that $\tau[x;h_{ij}]$ is the classical
time and that (\ref{e:scg}) is the Schr\"{o}dinger picture version of quantum
field theory in curved space.
As with the mirror example, $\tau$ becomes imaginary during tunneling so we
define a Euclidean time $\tau_{E}$ along with a Euclidean lapse $N^{\tau_{E}}
=iN^{\tau}$, in terms of which $\chi$ obeys
\begin{equation}
-\int dx [N^{\tau_{E}}\hat{{\cal H}}_{t_{M}}+iN_{i}\hat{{\cal H}}_{M}^{i}]\chi
[\phi_{i},\tau_{E}]=\frac{\partial}{\partial \tau_{E}}\chi[\phi_{i},\tau_{E}].
\end{equation}
For a massless scalar field with action
\begin{equation}
S=-\frac{1}{2} \int d^{4}x \sqrt{-g\,}g^{\mu \nu}\,\partial_{\mu}\phi
\partial_{\nu}\phi,
\end{equation}
we have
\begin{equation}
\hat{{\cal H}}_{t_{M}}=\frac{1}{2}\left(h^{-\frac{1}{2}}\hat{\pi}_{\phi}^{2}
+h^{\frac{1}{2}} h^{ij} \partial_{i}\hat{\phi}\, \partial_{j} \hat{\phi}\right)
\end{equation}
\begin{equation}
\hat{{\cal H}}_{i_{M}}=\partial_{i}\hat{\phi}\, \hat{\pi}_{\phi}.
\end{equation}
To evolve $\chi$ through the tunneling region one is required to calculate the
Euclidean time evolution operator
\begin{equation}
\hat{U}_{E}(\tau_{E}^{f},\tau_{E}^{i})=T \exp\left[-\int_{\tau_{E}^{i}}
^{\tau_{E}^{f}} \hat{{\cal H}}^{E}_{\phi} d\tau_{E}\right]
\end{equation}
with
\begin{equation}
\hat{H}^{E}_{\phi}=\int d^{3}x\, [N^{\tau_{E}}(-\frac{1}{2}h^{-\frac{1}{2}}
\frac{\delta^{2}}{\delta \phi^{2}} +\frac{1}{2}h^{\frac{1}{2}}h^{ij}
\partial_{i}\phi\, \partial_{j} \phi)+N_{i} \partial_{i} \phi \frac{\delta}
{\delta \phi}].
\end{equation}
As before, one proceeds by defining Euclidean fields obeying (\ref{e:phi},\ref{e:pi}). In the present case the resulting field equations are:
$$
\left(\sqrt{g_{E}}\,g_{E}^{\mu \nu}\partial_{\mu}\hat{\phi}^{E}\right)_{,\nu}=0
$$
\begin{equation}
\hat{\pi}_{\phi}^{E}=i\frac{h^{\frac{1}{2}}}{N^{\tau_{E}}}\left(
\frac{\partial \hat{\phi}_{E}}{\partial \tau_{E}}-N^{i}\partial_{i}\phi\right)
\end{equation}
with
\begin{equation}
ds_{E}^{\;2}=g_{\mu \nu}^{E}\,dx^{\mu} dx^{\nu}=
\left(N^{\tau_{E}}d\tau_{E}\right)^{2}
+h_{ij}\left(dx^{i}+N_{i}d\tau_{E}\right)\!\left(dx^{j}+N^{j}d\tau_{E}\right).
\end{equation}
The evolution operator, and therefore the state after tunneling, is determined
by solving the field equations mode by mode, and repeating the steps leading
from (\ref{e:sta}) to (\ref{end}).
\label{s:grav}
\section{Black Hole Radiance in the Presence of Tunneling}
\label{sechole}
We can now apply this method to determine how the radiation from a black hole
is affected by tunneling. It is well known that a black hole formed
classically from collapsing matter radiates in a complicated manner at early
times due to the time dependent geometry, but at late times will inevitably
radiate as a black body at the Hawking temperature. Is this scenario altered
if the black hole is formed while tunneling? We shall show that it is not.
The form of the late time radiation is insensitive to the hole's
unconventional history in a way that is consistent with the intuitive picture
of Hawking radiation being caused by pair production near the horizon.
We consider the behaviour of a scalar field on the background of a false
vacuum bubble which tunnels leading to the formation of a black hole. The
action for a false vacuum bubble in the thin wall approximation is
\begin{equation}
S=\frac{1}{16\pi} \int d^{4}x \sqrt{-g\,}\,{\cal R} -\frac{\Lambda_{I}}
{8\pi}\int_{bubble} \! d^{4}x \sqrt{-g\,}\, -\frac{\mu}{4\pi}\int_{wall}d^{3}A
\end{equation}
where $\Lambda_{I}$ is the cosmological constant of the false vacuum, and
$\mu$ is the energy density of the bubble wall. The classical solutions for
this action have been derived in Refs.~\cite{Sato81}-\cite{Guth87}.
In what follows
we refer to the treatment of Ref.~\cite{Guth87}. The spherically symmetric
solutions are characterized by three parameters: $\Lambda_{I}$, $\mu$, and
the total mass $M$. In addition, for given $\Lambda_{I}$ and $\mu$ there is
a critical mass $M_{cr}$ below which there are two solutions: type (a),
where the bubble emerges from a singularity with zero radius, subsequently
expands to a maximum radius, and then recollapses; type (b), where the bubble
initially collapses from infinite radius, reaches a minimum radius, and then
reexpands. Using the results of Refs.~\cite{Guth90,polch}, we focus on an
expanding solution of type (a) which tunnels to an expanding solution of
type (b).
We confine our interest to the region outside the bubble where
the metric, written in terms of Schwarzschild time $t$ and $r_{*}=r+2M
\ln({r}/{2M}-1)$, is
\begin{equation}
ds^{2}=\left(1-\frac{2M}{r}\right)\left(-dt^{2}+dr_{*}^{2}\right)+r^{2}
d\Omega^{2}.
\end{equation}
As $t$ and $r_{*}$ cover only part of the complete manifold, we introduce
Kruskal-Szekeres coordinates,
\begin{equation}
ds^{2}=\frac{32M^{3}e^{-r/2M}}{r}(-dT^{2}+dX^{2}) +r^{2}d\Omega^{2}.
\end{equation}
The two sets of coordinates are related by
$$
\left(\frac{r}{2M}-1\right)e^{r/2M}=X^{2}-T^{2}
$$
\begin{equation}
t=\left\{\begin{array}{ll}
4M\tanh^{-1}(T/X) & \mbox{if \, $|T/X|<1$}\\
4M\tanh^{-1}(X/T) & \mbox{if \, $|T/X|>1$}
\end{array}\right.
\end{equation}
Using these cordinates the type (a) and (b) solutions of interest are depicted
in Fig. ~\ref{bubfig2}.
\renewcommand\floatpagefraction{.9}
\renewcommand\topfraction{.9}
\renewcommand\bottomfraction{.9}
\renewcommand\textfraction{.1}
\begin{figure}[htb]
\centerline{\psfig{file=bubfig2.epsi,width=\hsize,angle=-90}}
\caption{The type (a) and (b) solutions. The heavy lines represent the bubble
trajectory, and the dashed lines are the initial and final surfaces of the
tunneling solution. In these figures, only the regions to the right of the
trajectory are of interest, as they are outside of the bubble.}
\label{bubfig2}
\end{figure}
The tunneling amplitude for this process has been computed by two different
methods. In Ref.~\cite{polch} the solution to the Wheeler-DeWitt equation
is found in the
WKB approximation by solving the Einstein-Hamilton-Jacobi equation
(\ref{e:ehj}). Since the solution behaves as $e^{-S}$, and the tunneling
amplitude is given by the ratio of the wavefunction evaluated at the initial
and final geometries, the tunneling amplitude is
\begin{equation}
\exp\left(S[h_{ij}^{\mbox{\scriptsize{initial}}}]-S[h_{ij}^
{\mbox{\scriptsize{final}}}]\right)
\end{equation}
No difficulties arise in this approach; the calculation of tunneling
amplitude proceeds in a straightforward fashion.
In Ref.~\cite{Guth90} the calculation is performed using the functional
integral.
In this formalism one looks for a manifold which interpolates between the
initial and final surfaces and which is a solution to the Euclidean Einstein
equations. The tunneling amplitude is $e^{-S}$, where S is the action of the
solution. It is found, however, that solving the field equations leads to a
sequence of three geometries which do not form a manifold. To see this, first
note that the geometry outside the bubble is Euclidean Schwarzschild space,
obtained by $t\rightarrow it_{E}$, $T\rightarrow iT_{E}$,
\begin{equation}
ds_{E}^{2}=\left(1-\frac{2M}{r}\right)\left(dt_{E}^{2}+dr_{*}^{2}\right)
+r^{2}d\Omega^{2}
=\frac{32M^{3}e^{-r/2M}}{r}\left(dT^{2}+dX^{2}\right)+r^{2}d\Omega^{2}
\end{equation}
with
\begin{equation}
\left(\frac{r}{2M}-1\right)e^{r/2m}=X^{2}+T_{E}^{2} \mbox{\ ;\ }
t_{E}=4M\tan^{-1}(T_{E}/{X}).
\end{equation}
It remains to describe the motion of the bubble wall. Solving the equations
of motion leads to the trajectory in Fig. \ref{bubfig3}.
\renewcommand\floatpagefraction{.9}
\renewcommand\topfraction{.9}
\renewcommand\bottomfraction{.9}
\renewcommand\textfraction{.1}
\begin{figure}[htb]
\centerline{\psfig{file=bubfig3.epsi,width=4truein,angle=-90}}
\caption{Bubble trajectory in Euclidean Schwarzschild space.}
\label{bubfig3}
\end{figure}
It is seen that the bubble wall
crosses the initial surface during the course of its motion, creating a
situation in which it is impossible to identify a region which is swept out
by the evolving hypersurface. Some regions of the manifold are crossed
twice by the hypersurface, some once, and some not at all. The authors of
Ref.~\cite{Guth90} call
this object a pseudomanifold and give a prescription to calculate its action
by assigning covering numbers to the various regions, but this is not needed
for what follows.
With these results in hand, the technique of Sect.~(3) can be
used to calculate the state of the scalar field after tunneling. It was seen
that once the solution of the Einstein-Hamilton-Jacobi equation is given, the
field wave functional $\chi$ is fully determined by equations (\ref{e:sc}) and
(\ref{e:mo}). Since $S[h_{ij}]$ is calculated in Ref.~\cite{polch},
we have all
that we need to find $\chi$. This would, however, require finding the
solution
to an unfamiliar functional differential equation. To cast it in in the form
of the Schr\"{o}dinger equation a lapse $N^{\tau_{E}}$, shift $N_{i}$
and time $\tau_{E}$ were reintroduced leading to the appearance of the
Euclidean metric $g_{\mu \nu}^{E}$. In the present case there is no true
interpolating Euclidean manifold, so that any choice of $N^{\tau_{E}}$ and
$N_{i}$ which define a well behaved $g_{\mu \nu}^{E}$ will lead to a bubble
trajectory that is a multivalued function of time. Alternatively, a choice
of time functional which gives a single valued bubble trajectory will
necessarily lead to a Euclidean metric with vanishing determinant at some
point. In either case, it is not clear that the resulting Schr\"{o}dinger
equation is well defined. This is apparent from Fig. \ref{bubfig3},
where it can be seen
that boundary conditions imposed on the initial surface and on the bubble wall
may contradict each other. These difficulties arise as a result of trying to
compute the final state of the field in one step, which requires a Euclidean
manifold interpolating all the way from the initial surface to the final
surface, and can be avoided by calculating the state on a series of
intermediate hypersurfaces. In this approach, it does not matter that the
bubble wall eventually crosses the initial surface since once the state is
calculated at some intermediate point we can forget about what preceded it.
For simplicity, we will consider only the s-wave component of the scalar field
and frequencies high enough such that the geometrical optics approximation is
valid. This means that the field equation is taken to be
\begin{equation}
g_{E}^{\mu \nu}\partial_{\mu} \partial_{\nu} \phi =0.
\end{equation}
The state of the field on the initial surface, $t=T=0$, is most conveniently
expressed in terms of the coordinates $r_{*}$ and $t$. We divide the modes
into ingoing and outgoing,
$$
\xi_{\omega}^{\mbox{\scriptsize{in}}}(r_{*},t)=C_{\omega}\,
e^{-i\omega(t+r_{*})}
$$
\begin{equation}
\xi_{\omega}^{\mbox{\scriptsize{out}}}(r_{*},t)=
C_{\omega}\,e^{-i\omega(t-r_{*})}
\end{equation}
and write the field operator as
\begin{equation}
\hat{\phi}(r_{*},t)=\sum_{\omega}\left[\hat{a}_{\omega}^
{\mbox{\scriptsize{in}}}
\xi_{\omega}^{\mbox{\scriptsize{in}}}+\hat{a}_{\omega}^
{\mbox{\scriptsize{in}}\dagger}\xi_{\omega}^{\mbox{\scriptsize{in}*}}
(r_{*},t) + \mbox{ in} \rightarrow \mbox{out}\right].
\end{equation}
$C_{\omega}$ are normalization constants whose values will not be important.
We shall only consider the {\em in} modes as the treatment of the {\em out}
modes is exactly the same. We also suppress the {\em in} superscript.
In the first stage of the evolution the hypersurface is pivoted around
$r_{*}=r_{*}^{b}$ by $180^{\circ}$, where $r_{*}^{b}$ is the position
of the bubble wall on the
initial surface. The solutions to the Euclidean field
equations are most conveniently obtained by choosing Cauchy boundary conditions
on the initial surface, (clearly a valid procedure in this case)
$$
f_{\omega}^{+}(r_{*},0)=\xi_{\omega}(r_{*},0) \mbox{ \ ; \ }
\frac{\partial}{\partial t_{E}}f_{\omega}^{+}(r_{*},0)
=-i\frac{\partial}{\partial t}\xi_{\omega}^{*}(r_{*},0)
$$
\begin{equation}
f_{\omega}^{-}(r_{*},0)=\xi_{\omega}^{*}(r_{*},0) \mbox{ \ ; \ }
\frac{\partial}{\partial t_{E}}f_{\omega}^{-}(r_{*},0)
=-i\frac{\partial}{\partial t}\xi_{\omega}^{*}(r_{*},0)
\end{equation}
It is also easiest to use the $X$, $T$ coordinates as they are well behaved
everywhere. Since the evolution of the hypersurface is simply a reflection
about the point $X=X^{b}$, a mode which has the form $f(X,T_{E})$ on the
initial
surface has the form $f(-X+2X^{b},T)$ on the new surface. Using the relations
\begin{equation}
r_{*}=4M\ln\sqrt{X^{2}+T_{E}^{2}\,} \mbox{ \ ; \ } t_{E}=4M\tan^{-1}(T_{E}/X)
\end{equation}
and that
\begin{equation}
f_{\omega}^{\pm}(r_{*},t_{E})=C_{\omega}\,e^{\pm \omega t_{E} -i\omega r_{*}}
\end{equation}
near the initial surface, one sees that near the new surface,
\begin{equation}
f_{\omega}^{\pm}(X,T_{E})=C_{\omega}\exp\left(\frac{\mp 4M\omega
T_{E}}{-X+2X^{b}}
-4iM\omega\ln(-X+2X^{b})\right).
\end{equation}
Since on the new surface,
$f_{\omega}^{+}=(f_{\omega}^{-})^{*}$ and $\partial f_
{\omega}^{+}/\partial t_{E}=-(\partial f_{\omega}^{-}/\partial t_{E})^{*}$,
the
evolution operator $\hat{U}_{E}$ is unitary. This means that the state on the
new surface has the same form as it did on the initial surface, but is now
expressed in terms of the modes
\begin{equation}
\xi_{\omega}(X,T)=C_{\omega}\exp\left(\frac{-4iM\omega T}{-X+2X^{b}}
+4iM\omega \ln(-X+2X^{b})\right).
\end{equation}
These modes can be approximated near $T=0$ as
\begin{equation}
\xi_{\omega}= \left\{\begin{array}{ll}
C_{\omega}\, e^{i\omega (t-r_{*})} & \mbox{ if $ \; |X|\gg X^{b}$} \\
C_{\omega}\,e^{-(2iM\omega/X^{b})(T-X)} & \mbox{ if $ \; |X|\ll X^{b}$}
\end{array} \right.
\end{equation}
Now it is useful to express the state in terms of modes which are nonzero only
inside or outside the horizon,
$$
\eta_{\omega}^{<}=\left\{\begin{array}{ll}
D_{\omega}\,e^{i\omega(t-r_{*})} & \mbox{ if $\; X<0$}\\
0 & \mbox{ if $ \; X>0$}
\end{array} \right.
$$
\begin{equation}
\eta_{\omega}^{>}=\left\{\begin{array}{ll}
0 & \mbox{ if $\; X<0$} \\
D_{\omega}\,e^{-i\omega(t+r_{*})} & \mbox{ if $\; X>0$.}
\end{array} \right.
\end{equation}
A fundamental result \cite{hawking,unruh} in the derivation of black hole
radiance is that the vacuum state with respect to modes which have a time
dependence $e^{-i\omega T}$ is the state
\begin{equation}
\mbox{const.}\times \sum_{\{n_{\omega}\}}e^{-E(\{n_{\omega}\})/2T_{H}}
\left| \{n_{\omega}\}\right\rangle_{<}\left| \{n_{\omega}\}\right\rangle_{>}
\label{e:dm}
\end{equation}
with respect to the modes $\eta_{\omega}^{<}$ and $\eta_{\omega}^{>}$. The
sum runs over all sets of occupation numbers, $E=\sum n_{\omega}\omega$, and
$T_{H}=1/8\pi M$ is the Hawking temperature. Further, near the horizon, any
deviation of $\left|\chi \right \rangle$ from the
vacuum state can be ignored because
of the arbitrarily large redshift as $r_{*}\rightarrow -\infty$. Far from the
horizon $\xi_{\omega}$ and $\eta_{\omega}^{<}$ agree so the form of the state
is unchanged there.
Now the hypersurface can be evolved the remainder of the way. If we restrict
our attention to the region $X<X^{b}$, then the motion of the hypersurface
is simply a translation, $t_{E}\rightarrow t_{E}-\Delta t_{E}$.
This causes states with time dependence $e^{i\omega t}$ to be damped by a
factor $e^{-\omega \Delta t_{E}}$, and states with time dependence
$e^{-i\omega t}$ to be amplified by a factor $e^{\omega \Delta t_{E}}$.
Near the
horizon, the state $\left|\chi \right\rangle$ consists of pairs of positive and
negative frequency states according to (\ref{e:dm}). One member of the pair
is damped but the other is amplified by a compensating amount so as to leave
the state $\left|\chi \right\rangle$ unchanged.
The final state of the field can then be summarized as follows.
Far from the hole, where there is no pairing, the initial state is damped:
\begin{equation}
\sum_{\{n_{\omega}\}} S(\{n_{\omega}\})\left|\{n_{\omega}\}\right\rangle
\longrightarrow \mbox{const.}\times \sum_{\{n_{\omega}\}}e^{-E(\{n_{\omega}\})
\Delta t_{E}}\,S(\{n_{\omega}\})\left|\{n_{\omega}\}\right\rangle.
\end{equation}
Near the horizon the final state is given by (\ref{e:dm}). This is true for
both the in and out modes, so an observer stationed on either side of the
horizon would observe a thermal distribution of both ingoing and outgoing
particles. As time passes, all of the ingoing particles will eventually
cross the horizon and be swallowed by the hole, whereas the outgoing particles
will propagate out to infinity where they can be detected at arbitarily late
times as a flux of thermal radiation at the Hawking temperature.
\section{Comments}
It was shown that the standard picture of black hole radiance is unchanged by
tunneling. At late times, the hole radiates just as it would have had it been
formed from a classical collapse. This makes sense if one thinks of Hawking
radiation as pair production. The probability of tunneling is not affected
by the creation of a pair, since the pair has zero total energy. From this
point of view it is also clear that what happens at early times cannot
possibly affect the late time radiation, since the produced pairs only see the
late time geometry. The conventional derivation of radiance obscures this
point somewhat and it seems desirable to find an approach which makes
this feature manifest from the outset.
For the two systems considered here, and presumably this is true in general,
the effect of the tunneling was to shift the distribution of any particles
that were
present before tunneling. In the present case initial excitations were damped
because the final surface is rotated clockwise relative to the initial
surface. A counterclockwise rotation would have led to amplification.
In \cite{Guth90} numerical investigations are quoted which show that the
rotation is always clockwise for the false vacuum bubble. One is led to
speculate whether this is a general phenomenon --- whether all tunneling
transitions lead to damping.
\chapter{Introduction}
Perhaps the most conspicuous deficiency in a physicist's current understanding
of nature is the absence of a usable quantum mechanical theory of gravity.
Standing in the way has been an utter lack of experimental data to serve as a
guide toward such a theory. On one hand, the Standard Model of particle
physics, while not without unresolved issues of its own, provides a coherent
description of all phenomena which have been observed at existing accelerators;
on the other hand, string theory, while remaining a viable candidate for a
theory of everything, is presently incapable of making further predictions
which can be tested by these same accelerators. In the face of such a
situation there is still reason to believe that progress can be made by
extracting clues from the structure of the theories that we know today. The
remarkable quantum mechanical properties of black holes are an important
example -- the phenomenon of black hole thermodynamics is tantalizingly
suggestive of a more comprehensive theory.
Before discussing the relevance of
black holes to quantum gravity, which is the focus of the present work, it will
be useful to recall some of the development of the classical theory of black
holes.
Black holes are hard to see. For the experimentalist, this is true for obvious
reasons: the defining property of black holes as regions of spacetime
from which nothing can emerge means that their existence can only be inferred,
either through their gravitational effects on neighboring bodies, or by
viewing the characteristic glow of matter as it is sucked into a hole.
Nevertheless, recent observations have essentially removed any doubt as to the
existence of black holes in our universe. For many years, theorists were
similarly unable to see black holes in the equations of general relativity,
although in this case the difficulty was due to faulty vision rather than a
lack of evidence. Although Schwarzschild wrote down the metric describing the
geometry outside a spherically symmetric body,
$$
ds^2=-(1-2M/r)dt^2+\frac{dr^2}{1-2M/r}+r^2(d\theta^2 +\sin^2
\theta d\phi^2),
$$
immediately after the discovery of the field equations, an incomplete
understanding of the geometry's global structure led to the erroneous
conclusion that a singularity occurs at the horizon, $r=2M$, rendering the
solution for $r\le 2M$ unphysical. Only much later was it realized that the
singularity was merely a coordinate artifact, and that physical quantities are
entirely well behaved at the horizon. With this came the understanding that
there exists a {\em true} singularity at $r=0$, but it was suspected that
deviations from spherical symmetry would cause the singularity to be smoothed
out in a realistic collapse process. However, once Penrose \cite{penrose}
proved that singularities {\em do} occur in classical relativity under very
generic conditions, the modern age of black hole research was begun.
Two theorems in classical relativity served to presage the subsequent
development of black hole thermodynamics. First, it was found that the mass
and angular momentum of black holes obey the so-called ``first law of black
hole thermodynamics'':
$$
d\!M=\frac{1}{8\pi}\kappa d\!A + \Omega d\!J,
$$
$J$ being the hole's angular momentum, $A$ and $\Omega$ the horizon's area and
angular velocity. $\kappa$ is the surface gravity, defined as the force which
a person at infinity would need to exert in order to keep an object suspended
at the horizon. For the Schwarzschild hole, $\kappa =1/4M$. Second, it was
found that in any process, the total area of black hole horizons necessarily
increases --- the so-called ``second law of black hole thermodynamics''.
These results led Bekenstein \cite{bekenstein} to suggest that a black hole
possessed an entropy proportional to $A$, and that the conventional second
law of thermodynamics would be upheld when this entropy was taken into account.
While the formulae suggest that something proportional to $\kappa$ plays the
role of temperature, this point was entirely obscure from the standpoint of
classical relativity, since in this context a black hole can't emit anything,
much less thermal radiation.
At this point, quantum mechanics enters the game: Hawking \cite{hawking}
showed that quantum effects cause a black hole to emit exactly thermal
radiation with a temperature $T=\kappa/2\pi$. This result immediately leads
to the identification $S=A/4$. As we shall consider the derivation of this
effect in detail later, let us simply remark here that it is intimately
connected to the existence of the horizon as a surface of infinite redshift.
This allows particles of negative energy to exist inside the horizon, leading
to the possibility of particle-antiparticle creation, one particle flowing
into the hole, one flowing out, without violating conservation of energy.
Perhaps the most startling implication of this result is that, although it only
relies on fairly conservative assumptions about the interaction of quantum
fields with gravity, it leads to a fundamental conflict with the usual
formulation of quantum mechanics. For if the black hole continually emits
radiation it will eventually disappear altogether, and since thermal radiation
is uncorrelated, there will not be any trace left of the matter which
originally formed the hole. More concisely, the process is not described by
by unitary evolution since any initial state forming the black hole leads to
the same thermal density matrix after the hole has evaporated. If this is
true, then quantum mechanics will have to be revised in order to be
compatible with this phenomenon of information loss. On the other hand, this
conclusion may be premature, as the approximation employed in computing the
radiance breaks down once the hole becomes sufficiently small, so that there
is in fact no compelling reason why the black hole has to disappear. Various
alternatives to the information loss scenario have been the subject of much
discussion recently (two reviews, written from quite different viewpoints,
are \cite{strominger,banks}) but no consensus has been established. Much of
the recent work has been done in the context of two dimensional models
\cite{CGHS} which, owing to conformal invariance, allow more analytical
progress than was previously possible. Unfortunately, the key questions
remain unanswered.
The other closely related problem is to obtain a proper understanding of the
formula $S=A/4$. It seems quite likely that a resolution of the information
loss puzzle is contingent upon understanding in what sense, if any, a black
hole has the number of states $e^{A/4}$. One possibility is that a black hole
has this number of weakly coupled states localized near the horizon, proper
accounting of which will lead to correlations in the outgoing radiation.
Although the radiation may look approximately thermal, the correlations could
be sufficient to encode all details of the initial state. Any computation
of such correlations must go beyond the free field approximation originally
employed by Hawking; with this goal in mind, a large part of the present work
is devoted to obtaining the leading corrections to the free field results.
In the course of this work we shall be discussing these problems mainly in the
context of the semiclassical approximation. However, we should point out that
the validity of this approximation is by no means assured, due to the peculiar
properties of the horizon. Indeed, it has been argued that strong coupling
effects \cite{schoutens} and quantum fluctuations \cite{ortiz} make this
description unreliable.
The remainder of this thesis is organized as follows. In Chapter 2 we develop
the Hamiltonian formulation of gravity, which will play a key role in the
material that follows. We consider both classical and quantum aspects, paying
special attention to the spherically symmetric case. Chapter 3 provides an
overview of quantum field theory in curved space, and a detailed discussion of
black hole radiance. The discussion is phrased in terms of a nonstandard
coordinate system for the black hole geometry \cite{KW2},
which is particularly convenient
for doing computations near the horizon. This coordinate system also provides
insights into the global structure of the geometry, as we discuss. Finally,
the role of fluctuations in the stress-energy is considered, with emphasis on
the impact these effects have on the moving mirror model. In Chapter 2 we
turn to the main focus of this thesis: the calculation of self-interaction
corrections to the black hole emission spectrum \cite{KW1,KW3}.
Drawing upon material
developed in the previous chapters, we show how these corrections can be
deduced by calculating the effective action for a gravitating thin shell. We
then discuss the relation between the first and second quantized approaches to
this problem, and finally, the difficulty in obtaining multi-particle
correlations by a straightforward extension of the single particle calculation.
In Chapter 5 we describe some attempts at gaining a microscopic understanding
of the black hole entropy by counting fluctuations of the geometry. As we will
see, the problem is not how to obtain a sufficiently large number of states,
but rather how to control the divergences which inevitably occur. The topic
of Chapter 6 is the effect of quantum tunneling upon radiance phenomena
\cite{bubble}. The
main example concerns black hole formation via the tunneling of a false vacuum
bubble, a process which nicely illustrates the interplay between the
Hamiltonian and Lagrangian approaches to quantum gravity. We will show how
thermal radiation emerges from the black hole at late times,
even though the
standard calculation of the radiance is inapplicable.
|
\section{Introduction}
Quantum chromodynamics (QCD), the fundamental theory of strong
interactions, can be perturbatively solved only in the region of
asymptotic freedom, i.e., for large momenta of quarks and gluons
\cite{Pol}. At low momenta, quarks and gluons interact strongly and are
confined inside hadrons. In this case, the expansion parameter of
perturbation theory, the running coupling constant $\alpha _s,$ is of the
order of unity, so that perturbative methods are not applicable and one
has to use non-perturbative methods. One such method is the construction
of effective theories to describe the behaviour of hadronic matter, such
as the $\sigma -\omega $ model \cite{Val}. However, one of the most
successful non-perturbative methods is the use of lattice gauge
calculations \cite{Cre}, which are especially suitable for studying
perturbative as well as non-perturbative effects in QCD. The presently
available lattice data mostly concern simulations of $SU(N)$ pure gauge
theory \cite{B,D}, since the treatment of dynamical fermions on a lattice
is difficult \cite{Cre}. Moreover, lattice artefacts are believed to be
well under control only in the pure gauge case \cite{EE}. A striking
feature of lattice simulations of $SU(2)$ and $SU(3)$ pure gauge theory
is a phase transition (of apparently first order for $SU(3)$
\cite{B,D,Ben} and second order for $SU(2)$ theory \cite{Fig}) from a
phase of confined gluons (``glueballs'')
to one of deconfined gluons (``gluon plasma''), leading to a sharp
rise in the energy density of a gluon gas as a function of temperature
at a phase transition temperature $T_c$ \cite{B,D} (Figs. 1,2). In the
$\sigma-\omega $ model, a similar phase transition (or, at least, a rapid
increase of the energy density in a small temperature interval) appears
at small net baryon densities \cite{The} (Fig. 3). This is due to a
strong enhancement of the scalar meson interaction at $T_c\simeq 200$
MeV, leading to a transition from a phase of massive baryons to a phase
of massless baryons. Apparently, a phase transition to a weakly
interacting phase seems to be a fundamental feature of strongly
interacting matter\footnote{It should be, however, mentioned that a phase
transition in the $\sigma -\omega $ model may be an artefact of the
approximations made in the calculation of the thermodynamic quantities in
Fig. 3 (in the mean-field approximation) \cite{HE}.} \cite{Ri}.
To understand the lattice data from a simple physical point of view,
Rischke {\it et al.} \cite{Ri,Ri1} have constructed a phenomenological
model for the gluon plasma, in which gluons with large momenta are
considered as an ideal gas with perturbative corrections of order
$O(\alpha _s),$ while gluons with low momenta are subject to confining
interactions and do not contribute to the energy spectrum of free gluons.
The equation of state for this model, although it quantitatively
reproduces the lattice data for the thermodynamic functions of $SU(3)$
pure gauge theory above the deconfinement transition temperature
(Fig. 2), it has a rather complicated form and is therefore not suitable
for practical use, e.g., in astrophysics (for description of stellar
structure), or in cosmolody (for treatment of a hadron-plasma phase
transition in the early universe). In this paper, we suggest another
equation of state, which reflects the main properties of strongly
interacting matter (i.e., a phase transition to a weakly interacting
phase) and agrees qualitatively with the lattice data for $SU(2)$ pure
gauge theory $both$ above and below the transition temperature, and has a
much simpler form, as compared to that of ref. \cite{Ri}.
First, let us note that in strongly interacting matter, particles
undergoing continual mutual interaction are necessarily off-shell.
Therefore, the effect of strong interaction in such a system may be
represented by the off-shellness of its particles. The equilibrium state
of such a system should be characterized by a well-defined relativistic
mass distribution around the on-shell value. Thus, instead of dealing
with interaction explicitly, we reduce the problem to the description of
the relativistic off-shell ensemble. The role of interaction then
consists in determining the effective thermodynamic parameters
governing the mass distribution in a strongly interacting system.
The physical framework for the description of a relativistic off-shell
ensemble has been established by Horwitz and Piron \cite{HP} as a
manifestly covariant relativistic dynamics whose consistent formulation
is based on the ideas of Fock \cite{Fock} and Stueckelberg \cite{Stu},
in which the four components of energy-momentum are considered as
independent degrees of freedom, permitting fluctuations from the mass
shell. In this framework, the dynamical evolution of a system of $N$
particles, for the classical case, is governed by equations of motion
that are of the form of Hamilton equations for the motion of $N$ $events$
which generate the space-time trajectories (particle world lines) as
functions of a continuous Poincar\'{e}-invariant parameter $\tau $
\cite{HP,Stu}. These events are characterized by their positions $q^\mu =
(t,{\bf q})$ and energy-momenta $p^\mu =(E,{\bf p})$ in an
$8N$-dimensional phase-space. For the quantum case, the system is
characterized by the wave function $\psi _\tau (q_1,q_2,\ldots ,q_N)\in
L^2(R^{4N}),$ with the measure $d^4q_1d^4q_2\cdots d^4q_N\equiv d^{4N}q,$
$(q_i\equiv q_i^\mu ;\;\;\mu =0,1,2,3;\;\;i=1,2,\ldots ,N),$ describing
the distribution of events, which evolves with a generalized
Schr\"{o}dinger equation \cite{HP}. The collection of events (called
``concatenation'' \cite{AHL}) along each world line corresponds to a
{\it particle,} and hence, the evolution of the state of the $N$-event
system describes, {\it a posteriori,} the history in space and time of
an $N$-particle system.
For a system of $N$ interacting events (and hence, particles) one takes
\cite{HP} (we use the system of units in which $\hbar =c=k_B=1;$ we also
use the metric $g^{\mu \nu }=(-,+,+,+))$
\beq
K=\sum _i\frac{p_i^\mu p_{i\mu }}{2M}+V(q_1,q_2,\ldots ,q_N),
\eeq
where $M$ is a given fixed parameter (an intrinsic property of the
particles), with the dimension of mass, taken to be the same for all the
particles of the system. The Hamilton equations are
$$\frac{dq_i^\mu }{d\tau }=\frac{\partial K}{\partial p_{i\mu }}=\frac{p_
i^\mu }{M},$$
\beq
\frac{dp_i^\mu }{d\tau }=-\frac{\partial K}{\partial q_{i\mu }}=-\frac{
\partial V}{\partial q_{i\mu }}.
\eeq
In the quantum theory, the generalized Schr\"{o}dinger equation
\beq
i\frac{\partial }{\partial \tau }\psi _\tau (q_1,q_2,\ldots ,q_N)=K
\psi _\tau (q_1,q_2,\ldots ,q_N)
\eeq
describes the evolution of the $N$-body wave function
$\psi _\tau (q_1,q_2,\ldots ,q_N).$
In the present paper we restrict ourselves to a relativistic Bose gas, in
order to compare the results with experimental data of pure gauge theory
lattice simulations. We show that our results agree with those for
$SU(2)$ pure gauge theory. It should be, however, noted that since the
underlying theory is basically different from QCD, a comparison with the
$SU(2)$ lattice data can only be qualitative. From this point of view,
the similarity is remarkable.
\section{Ideal relativistic Bose gas}
Gibbs ensembles in a manifestly covariant relativistic classical and
quantum mechanics were derived by Horwitz, Schieve and Piron \cite{HSP}.
To describe an ideal gas of events obeying Bose-Einstein statistics in
the grand canonical ensemble, we use the expression for the number of
events found in \cite{HSP} (for our present purposes we assume no
degeneracy),
\beq
N=V^{(4)}\sum _{k^\mu }n_{k^\mu }=
V^{(4)}\sum _{k^\mu }\frac{1}{e^{(E-\mu -\mu _K\frac{m^2}{2M})/T}-1},
\eeq
where $V^{(4)}$ is the system's four-volume and $m^2\equiv -k^2=-k^\mu
k_\mu $ is the variable dynamical mass. Here, in addition to the usual
chemical potential $\mu ,$ there is the mass potential $\mu _K$
corresponding to the Lorentz scalar function $K(p,q)$ (Eq. (1.1)), here
taken in the ideal gas limit, on the $N$-event relativistic phase space;
in order to simplify subsequent considerations, we shall take it to be a
fixed parameter (which determines
an upper bound of the mass distribution in the ensemble we are studying,
as we shall see below). To ensure a positive-definite value for $n_{k^\mu
},$ the number of bosons with four-momentum $k^\mu ,$ we require that
\beq
m-\mu -\mu _K\frac{m^2}{2M}\geq 0.
\eeq
The discriminant of Eq. (2.2) must be nonnegative, which gives
\beq
\mu \leq \frac{M}{2\mu _K}.
\eeq
For such $\mu ,$ (2.2) has the solution
\beq
\frac{M}{\mu _K}\left( 1-\sqrt{1-\frac{2\mu \mu _K}{M}}\right) \leq m\leq
\frac{M}{\mu _K}\left( 1+\sqrt{1-\frac{2\mu \mu _K}{M}}\right) .
\eeq
For small $\mu \mu _K/M,$ the region (2.4) may be approximated by
\beq
\mu \leq m\leq \frac{2M}{\mu _K}.
\eeq
One sees that $\mu _K$ plays a fundamental role in determining an upper
bound of the mass spectrum, in addition to the usual lower bound $m\geq
\mu .$
Replacing the sum over $k^\mu $ (2.1) by an integral, one obtains for
the density of events per unit space-time volume $n\equiv N/V^{(4)}$
\cite{ind},
\beq
n=\frac{1}{4\pi ^3}\int _{m_1}^{m_2}\frac{m^3\;dm\;\sinh ^2\beta \;d
\beta }{e^{(m\cosh \beta -\mu -\mu _K\frac{m^2}{2M})/T}-1},
\eeq
where $m_1$ and $m_2$ are defined in Eq. (2.4), and we have used the
parametrization \cite{HSS} $$\begin{array}{lcl}
p^0 & = & m\cosh \beta , \\
p^1 & = & m\sinh \beta \sin \theta \cos \phi , \\
p^2 & = & m\sinh \beta \sin \theta \sin \phi , \\
p^3 & = & m\sinh \beta \cos \theta ,
\end{array} $$ $$0\leq \theta <\pi ,\;\;\;0\leq \phi <2\pi ,\;\;\;-\infty
<\beta <\infty .$$
In what follows we shall take $\mu \simeq 0$ (as for the case of the
ensemble of gauge bosons). The integral (2.6) is calculated in ref.
\cite{BHS} (in the high-temperature Boltzmann approximation for the
integrand):
\beq
n=\frac{T^4}{4\pi ^3}\left[ 2-x^2K_2(x)\right] ,\;\;\;\;\;x=\frac{
2M}{T\mu _K}\equiv \frac{2m_c}{T},
\eeq
where $K_\nu (z)$ is the Bessel function of the third kind (imaginary
argument), $m_c=M/\mu _K$ is the central value around which the mass of
the particles are distributed, in view of (2.4), and $2m_c$ is the upper
limit of the mass distribution in our case of small $\mu ,$ in view of
(2.5). One calculates the characteristic averages \cite{BHS} to be
\bqry
\langle E\rangle & = & T\;\frac{8-x^3K_3(x)}{2-x^2K_2(x)}, \\
\langle E^2\rangle & = & T^2\;\frac{40-x^4K_4(x)+x^3K_3(x)}{
2-x^2K_2(x)}, \\
\langle m^2\rangle & = & T^2\;\frac{16-x^4K_4(x)+4x^3K_3(x)}{
2-x^2K_2(x)}, \\
\langle {\bf p}^2\rangle & = & 3T^2\;\frac{8-x^3K_3(x)}{2-x^2K_2(x)}\;=
\;3T\langle E\rangle ,
\eqry
and obtains the following thermodynamic functions (the particle number
density, pressure and energy density) \footnote{If we had not used the
Boltzmann limit for the integrand in (2.6), one would obtain the factors
$\zeta (3)\approx 1.202$ in Eq. (2.12) and $\zeta (4)=\pi ^4/90\approx
1.082$ in Eqs. (2.13),(2.14).}:
\bqry
N_0 & = & \langle J^0\rangle \;=\;\frac{T^3}{\pi ^2}\;\frac{8-x^3K_3(
x)}{x^2}, \\
p & = & \frac{1}{3}\langle T^{ii}\rangle g_{ii}=\;\frac{T^4}{\pi ^2}\;
\frac{8-x^3K_3(x)}{x^2}\;=\;N_0T, \\
\rho & = & \langle T^{00}\rangle \;=\;\frac{T^4}{\pi ^2}\;\frac{
40-x^4K_4(x)+x^3K_3(x)}{x^2},
\eqry
where $\langle J^\mu \rangle $ and $\langle T^{\mu \nu }\rangle $ are
the average $particle$ four-current and energy-momentum tensor,
respectively, given by \cite{HSS}:
\beq
\langle J^\mu \rangle =\frac{T_{\triangle V}}{M}n\langle p^\mu \rangle ,
\;\;\;\langle T^{\mu \nu }\rangle =
\frac{T_{\triangle V}}{M}n\langle p^\mu p^\nu \rangle .
\eeq
In (2.15), $T_{\triangle V}$ is the average passage interval in $\tau $
for the events which pass through the small (minimal typical) four-volume
$\triangle V$ in the neighborhood of the $R^4$-point; it is related to
a width of the mass distribution around the central value, $\triangle m,$
as follows \cite{BH}:
\beq
T_{\triangle V}\triangle m=2\pi .
\eeq
The expressions (2.13),(2.14) for $p$ and $\rho $ are obtained from Eq.
(2.15) for $\langle T^{\mu \nu }\rangle .$ They are, moreover,
thermodynamically consistent. One may varify easily that the relation
\beq
\rho =T\frac{dp}{dT}-p
\eeq
is satisfied \cite{BHS}.
For low $T,$ Eqs. (2.13),(2.14) reduce, through the asymptotic formula
\cite{AS1}
\beq
K_\nu (z)\sim \sqrt{\frac{\pi }{2z}}e^{-z}\left( 1+\frac{4\nu ^2-1}{8z}+
\ldots \right) ,\;\;\;z>>1,
\eeq
to
\beq
p=\frac{2T^6}{\pi ^2m_c^2},\;\;\;\rho =\frac{10T^6}{\pi ^2m_c^2}=5p,
\eeq
consistent with the ``realistic'' equation of state suggested by
Shuryak for strongly interacting hadronic matter \cite{Shu}.
For high $T,$ we use another asymptotic formula \cite{AS2},
\beq
K_\nu (z)\sim \frac{1}{2}\Gamma (\nu )\left( \frac{z}{2}\right) ^{-\nu }
\left[ 1-\frac{z^2}{4(\nu -1)}+\ldots \right] ,\;\;\;z<<1,
\eeq
and obtain
\bqry
p & = & \frac{T^4}{\pi ^2}\left( 1-\frac{m_c^2}{2T^2}\right) \;=\;
p_{{\rm SB}}\left( 1-\frac{m_c^2}{2T^2}\right) , \\
\rho & = & \frac{T^4}{\pi ^2}\left( 3-\frac{m_c^2}{2T^2}\right) \;=\;
\rho _{{\rm SB}}\left( 1-\frac{m_c^2}{6T^2}\right) ,
\eqry
where $p_{{\rm SB}}=T^4/\pi ^2$ and $\rho _{{\rm SB}}=3p_{{\rm SB}}$ are
the pressure and energy density of an ideal ultrarelativistic
Stefan-Boltzmann gas. Therefore, as $T\rightarrow \infty ,$ the
thermodynamic functions of the relativistic Bose gas we are considering
become asymptotically those of a Stefan-Boltzmann gas, implying a phase
transition to an effectively weakly interacting phase, from the phase of
strong interactions described by Eq. (2.19).
It follows from Eq. (2.10), via the asymptotic formulas (2.18),(2.20),
that
\beq
\frac{\langle m^2\rangle }{T^2}=\left\{ \begin{array}{ll}
8, & T<<2m_c, \\
2m_c^2/T^2, & T>>2m_c.
\end{array} \right.
\eeq
The dependence of $p/p_{{\rm SB}},$ $\rho /\rho _{{\rm SB}}$ and $\langle
m^2\rangle /T^2$ on temperature are shown in Figs. 4,5. At $T\simeq 0.2\;
m_c$ (corresponding to $z\simeq 0.1$ in Figs. 4,5), there is a smooth
phase transition to a weakly
interacting phase described by Eqs. (2.21),(2.22).
\section{Concluding remarks}
The manifestly covariant framework discussed in the present paper can be
an effective tool in dealing with realistic physical systems. The
equation of state (2.12)-(2.14) obtained in our work reflects the main
properties of strongly interacting matter (i.e., a phase transition to a
weakly interacting phase), and agrees qualitatively
with the lattice data for $SU(2)$ pure gauge theory.
The question naturally arises of why the $SU(2)$ lattice data \cite{Fig}
appear to contain a second order phase transition but the $SU(3)$ data
\cite{B,D} appear to contain first order one. There is no {\it a priori}
difference between $SU(2)$ and $SU(3)$ pure gauge theories from a
statistical mechanical point of view. As shown in ref. \cite{B}, $SU(3)$
pure gauge theory simulations on $24^3\times N_T$ lattices indicate a
rapid rise in $\rho +p$ as a function of temperature which takes place
in the case of $N_T=4,$ reflecting a sharp first order phase transition.
This rise is broadened considerably in the case of $N_T=6,$ with the
slope of the curve diminished by almost a factor 3, indicating a smoother
transition in this case. As remarked by the authors, this softening of
the structure of the transition as $N_T$ is increased may well continue
as the continuum limit is aproached. Thus, in ref. \cite{B} the apparent
first order nature of the transition in the case of $SU(3)$ pure gauge
theory has been called in question. Moreover, there are indications from
lattice QCD calculations that when fermions are included, the phase
transition may be of second or higher order \cite{Ben1}. Altogether,
these observations suggest that a realistic equation of state of strongly
interacting hadronic matter should be expected to contain a second or
higher order phase transition, as reflected by the equation of state
obtained in our work.
As remarked by Ornik and Weiner \cite{OW}, a $single$ equation of state
which, at high temperature, describes a quark-gluon phase and, at low
temperature, a hadronic phase and which contains a phase transition of
either first or higher order provides a more satisfactory theoretical
description than one in which each phase is described by a different
equation of state. In this way, the equation of state that we obtained
can be considered as a candidate for such a realistic equation of state
of strongly interacting matter, in contrast to the equations of state of,
e.g., Rischke {\it et al.} \cite{Ri} and Shuryak \cite{Shu} each of which
describes just one of the phases (above and below the transition,
respectively). The introduction of the quark degrees of freedom in this
equation of state, as well as taking into account an effective
interaction potential in an explicit form in the strongly interacting
phase, and perturbative corrections in the weakly interacting phase,
should enable one to derive an equation of state which we expect to be
more accurate for the description of the phenomena taking place in
strongly interacting hadronic matter. The derivation of such an equation
of state and its possible implications in astrophysics and cosmology
are now being worked out by the authors.
\section*{Acknowledgements}
We wish to thank J. Eisenberg for very valuable discussions, and E.
Eisenberg for his help in drawing the pictures for the present paper.
\newpage
|
\section*{References} \begin{list}{}{
\setlength{\parsep}{0pt} \setlength{\itemsep}{0pt}
\setlength{\leftmargin}{0pt} } }{ \end{list} }
\title{A Search for Ultra-High Energy Counterparts to Gamma-Ray Bursts}
\author{S.P. Plunkett, M. Delaney, B. McBreen, K.J. Hurley and C.T. O'Sullivan}
\date{astro-ph/9508083: presented at 29 ESLAB Symposium, April 1995}
\begin{document}
\maketitle
\begin{abstract}
A small air shower array operating over many years has been used to
search for ultra-high energy (UHE) gamma radiation ($\geq 50$ TeV)
associated with gamma-ray bursts (GRBs) detected by the BATSE
instrument on the Compton Gamma-Ray Observatory (CGRO). Upper
limits for a one minute interval after each burst are presented for
seven GRBs located with zenith angles $\theta < 20^{\circ}$. A
$4.3\sigma$ excess over background was observed between 10 and 20
minutes following the onset of a GRB on 11 May 1991. The confidence
level that this is due to a real effect and not a background
fluctuation is 99.8\%. If this effect is real then cosmological
models are excluded for this burst because of absorption of UHE
gamma rays by the intergalactic radiation fields.
\end{abstract}
\section{Introduction}
Despite the large number of GRBs detected, the sources responsible for
this extraordinary phenomenon remain unidentified (Harding, 1994;
Hurley, 1994). To date, the majority of GRB counterpart searches have
been carried out at low energies (e.g.\ Schaefer, 1994), with few
attempts to search for a UHE component (e.g.\ Alexandreas {\it et al.}, 1994;
Borione {\it et al.}, 1993; Connaughton {\it et al.}, 1993; Vallania {\it et al.}, 1993).
Kazanas and Ellison (1986) have predicted that proton acceleration due
to diffuse shock acceleration in the atmosphere of neutron stars could
produce UHE emission, while the lightning model of McBreen {\it et al.}\
(1994) and the fireball model of M\'{e}sz\'{a}ros and Rees (1993)
suggest a similar result. Delayed high energy gamma ray emission,
including an 18 GeV photon detected 1.5 hours after the onset of
GRB940217, has been detected by the EGRET instrument onboard CGRO
(Dingus {\it et al.}, 1994; Hurley {\it et al.}, 1994). Models predicting such
emission were subsequently proposed by M\'{e}sz\'{a}ros and Rees (1994)
and Katz (1994).
A positive detection of UHE emission would indicate an upper limit to
the distance to GRB sources, since photons of such energy undergo pair
production with soft photons of the intergalactic radiation fields.
Absorption by the cosmic microwave background (CMB) would impose a
distance constraint of $\sim\! 100$ Mpc for 100 TeV photons (Gould and
Schreder 1966; Jelley 1966), though greater attenuation by the
intergalactic infrared radiation field (IIRF) may reduce this to
$\sim\! 10$ Mpc (Stecker {\it et al.}, 1994).
A small air shower array (Plunkett {\it et al.}, 1991a; 1991b) has been in
operation near sea level at University College, Dublin, since early
1989. Briefly, the array consists of eight plastic scintillation
counters of area 1 m$^{2}$, which sample air showers over a collection
area approximately $1.2 \times 10^{3}$ m$^{2}$ above a threshold
primary particle energy of about 50 TeV for showers close to the
zenith and a median primary particle energy of about 100 TeV. The
angular resolution of the array is 3.1$^{\circ}$ when all eight
detectors are triggered, and about 4.5$^{\circ}$ when three are
triggered. The latter resolution was used because showers triggering
at least three detectors were accepted for analysis.
The long term stability of the array was checked by the analysis of
$10^{4}$ hours of accumulated data. Over this period, the number of
air showers recorded over intervals of one minute were compared to a
measure of the mean rate of arrival of showers, determined over a
period of $\sim\! 6$ hours to account for variations imposed by
changes in atmospheric pressure. The mean shower arrival rate was
typically 35 min$^{-1}$, allowing the normal approximation to the
Poisson distribution to be applied to this analysis. The statistical
significance of any positive excesses was computed by the maximum
likelihood method of Li and Ma (1983). Both the observed and expected
number of one minute time intervals which registered an excess of air
showers with significances in the ranges $3.5\!-\!4\sigma$,
$4\!-\!4.5\sigma$, $4.5\!-\!5\sigma$ and $>\!5\sigma$, were shown to
give good agreement, thus confirming the stability of the array.
\section{Data Analysis}
The BATSE GRB catalogues (Fishman {\it et al.}, 1994) were searched for
bursts with a reported location within the field of view of the array
and a Poisson positional uncertainty of $<\! 5^{\circ}$, not including
systematic errors. This search yielded a sample of seven GRBs with a
zenith angle $\theta\! <\! 20^{\circ}$.
The search for UHE emission was conducted for a period of up to three
hours before and after each burst, while the reported burst location
was at $\theta\! <\! 20^{\circ}$. This period was divided into
non-overlapping time intervals of ten minutes duration, to search for
weak sustained emission, and an additional single interval of one
minute duration after the GRB onset time. The number of showers from
the burst direction was computed using a optimally sized circular
source bin, which is 1.6 times the angular resolution of the array
(Alexandreas {\it et al.}, 1993), of radius $7^{\circ}$ centred on the
reported location of the burst. The expected background was computed
using up to eight non-overlapping bins, identical in size to the
source bin, separated by $14^{\circ}/\cos\delta$ in right ascension
$\alpha$, but having the same declination $\delta$ as the GRB
location. These bins followed the same path on the sky as the source
bin, but at different times, and ensured a proper account of the
variation in shower arrival rate with zenith angle for all bins.
Upper limits to the photon flux were computed by the method of Gatto
{\it et al.}\ (1988) over a period of one minute after the GRB onset. In the
case of only one burst, GRB910511, a significant delayed excess of
$>\! 3 \sigma$ was found at the reported location, prompting a more
thorough search for the location of an optimum excess. The search
process was repeated 28 times with the source and background bins
shifted first in increments of $3^{\circ}$ in $\alpha$ and $\delta$ to
determine within which quadrant the excess is optimised, and
subsequently in smaller steps down to $1^{\circ}$ to further localise
the optimum source bin.
\section{Results}
The analysis of the seven GRBs in the sample showed no evidence for
coincident UHE emission. Upper limits at the 95\% confidence level,
calculated over an interval of one minute after the GRB onset, are
listed in Table~\ref{limits}.
\begin{table}[t]
\caption{\label{limits}Upper limits to the number of
air showers detected by the
array and the UHE photon flux at the 95\% confidence level for the GRBs
listed.}
\begin{center}
\begin{tabular}{lllll}\hline
GRB Date & Time & Altitude & Air Showers & Photon Flux \\
(yymmdd) & (UT) & ($90^{\circ} - \theta$) & & (cm$^{-2}$ s$^{-1}$)\\
\hline
910511 & 02:11:47.720 & 81.9 & $<\!6$ & $<\!1.7 \times 10^{-8}$ \\
910512 & 15:15:09.449 & 71.6 & $<\!8$ & $<\!2.2 \times 10^{-8}$ \\
910718 & 02:24:57.230 & 75.9 & $<\!3$ & $<\!8.3 \times 10^{-9}$ \\
911004 & 01:27:09.760 & 70.2 & $<\!5$ & $<\!1.4 \times 10^{-8}$ \\
920320 & 12:18:58.879 & 73.6 & $<\!5$ & $<\!1.4 \times 10^{-8}$ \\
921101 & 18:03:55.738 & 72.9 & $<\!5$ & $<\!1.4 \times 10^{-8}$ \\
931220 & 15:50:01.000 & 75.3 & $<\!7$ & $<\!1.9 \times 10^{-8}$ \\
\hline
\end{tabular}
\end{center}
\end{table}
The analysis of GRB910511 revealed an excess of events from the
reported location, between 10 and 20 minutes after the burst onset.
This excess was noted long before the report on delayed emission by
Hurley {\it et al.}\ (1994). As determined by BATSE this was a burst of
duration 7.2 s with a peak photon flux of 4.7 cm$^{-2}$ s$^{-1}$,
located at $\alpha\! =\! 266.5^{\circ}$, $\delta\! =\! 57.2^{\circ}$,
and with a Poisson positional uncertainty of $3.1^{\circ}$. This
location remained at $\theta\! <\! 20^{\circ}$ for 1.6 hours before
the burst and 3.1 hours afterwards. A detailed search, described in
section two, revealed an optimum excess in a bin centred on the
location $\alpha\! =\! 269.5^{\circ}$, $\delta\! =\! 63.2^{\circ}$
with a positional error of about $3^{\circ}$. The number of counts in
ten minute intervals from this source bin is shown in Fig.\ 1. During
a ten minute interval beginning at 02:21:47 UT, a total of 36 air
showers were detected compared to an expected background of 14.6
showers. This corresponds to an excess of $21.4 \pm 6.2$ events (the
errors quoted are $\pm 1 \sigma$), giving a statistical significance
of $4.3 \sigma$ (Li and Ma 1983). Upon examination, the arrival times
of these showers appear consistent with a random distribution.
\begin{figure}[t]
\epsfxsize=\textwidth
\epsffile{mtlb.ps}
\caption{The solid line gives the number of air showers
detected as a function of time from the optimum location
within the BATSE uncertainty
in the position of GRB910511. The expected background rate in each time
interval is shown as a dashed line. Time is measured in minutes
relative to the GRB onset at 02:11:47 UT. The $4.3 \sigma$ and an
earlier $2.9 \sigma$ excess are shown.}
\end{figure}
The initial number of trials made in the analysis of all seven GRBs
was 150. A further 62 trials were conducted during the overlapping
bin search for the optimum location of the delayed excess of air
showers in the interval 10 to 20 minutes after the onset of GRB910511,
and in examining the behaviour of the source over a six hour period at
this location (see Fig. 1). This yields a total of 212 trials and
results in a confidence level of 99.8\% that this excess was caused by
an UHE burst from the direction of the GRB, which occurred more than
ten minutes earlier, and not by a background fluctuation. The flux of
photons which is required to give the observed excess is $(5.9 \pm
1.7) \times 10^{-9}$ cm$^{-2}$ s$^{-1}$, assuming an efficiency of 0.5
for detection of air showers. Taking a typical primary particle energy
of 100 TeV for the showers detected by the array, this corresponds to
a fluence of $(9.5 \pm 2.8) \times 10^{-7}$ erg cm$^{-2}$ s$^{-1}$
averaged over the ten minute duration of the burst. An earlier excess
of showers appears in Fig. 1, though with a low significance of $2.9
\sigma$ this was rejected as a possible UHE detection.
\section{Conclusions}
For all seven of the GRBs in the sample, upper limits were obtained
for UHE emission in a one minute time interval after the GRB onset.
Evidence for delayed UHE emission accompanying GRB910511, in the
interval 10--20 minutes after the initial detection, has been
presented. If this effect is real an upper limit of 100 Mpc due to
absorption by the CMB is imposed on the distance to the source, which
might be further reduced to the order of 10 Mpc by the IIRF. It may
therefore be concluded that a non-cosmological origin is expected for
the source of GRB910511. Given that a large number of number of
ingenious mechanisms have been proposed to account for the generation
of GRBs (Nemiroff, 1994), it may be considered possible that there are
contributions to the observed number of bursts from sources at both
cosmological and non-cosmological distances.
\section*{Acknowledgements}
{The authors are indebted to Catherine Handley for her able
assistance with the data analysis and to FORBAIRT for their support.}
\begin{refs}
\item Alexandreas, D.E. {\it et al.}, 1993, Nucl. Instrum. Meth. Phys. Res.,
A328, 570.
\item Alexandreas, D.E. {\it et al.}, 1994, Astrophys. J., 426, L1.
\item Borione, A. {\it et al.}: 1993,
{\it Proc. 23rd ICRC}, Calgary, {\bf 1}, 57.
\item Connaughton, V. {\it et al.}: 1993,
{\it Proc. 23rd ICRC}, Calgary, {\bf 1}, 69.
\item Dingus, B.L. {\it et al.}: 1994,
{\it AIP Conf. Proc. 307}, AIP:New York, 22.
\item Fishman, G.J. {\it et al.}, 1994, Astrophys. J. Suppl., 92, 229.
\item Gatto, R. {\it et al.}, 1988, Phys. Lett., B204, 81.
\item Gould, R.J. and Schreder, G. {\it et al.}, 1966, Phys. Rev.
Lett., 16, 252.
\item Harding, A.: 1994, in C. Fichtel {\it et al.}\ (eds.)
{\it AIP Conf. Proc. 304}, AIP:New York, 30.
\item Hurley, K.C., 1994, Astrophys. J. Suppl., 90, 857.
\item Hurley, K.C. {\it et al.}, 1994, Nature, 372, 652.
\item Jelley, J.V. {\it et al.}, 1966, Phys. Rev. Lett., 16, 179.
\item Katz, J., 1994, Astrophys. J., 432, L27.
\item Kazanas, D. and Ellison, D.C., 1986, Adv. Space Res., 6, 81.
\item Li, T.P. and Ma, Y.Q., 1983, Astrophys. J., 272, 317.
\item M\'{e}sz\'{a}ros, P. and Rees, M.J., 1993, Astrophys. J., 418, L59.
\item M\'{e}sz\'{a}ros, P. and Rees, M.J., 1994, Mon. Not. Roy. Astr. Soc.,
269, L41.
\item McBreen, B. {\it et al.}, 1994, Mon. Not. Roy. Astr. Soc., 271, 662.
\item Nemiroff, R.J.: 1994,
{\it AIP Conf. Proc. 307}, AIP:New York, 730.
\item Plunkett, S. {\it et al.}, 1991a, Nucl. Instrum. Meth. Phys. Res., A300, 197.
\item Plunkett, S. {\it et al.}, 1991b, Proc. 22nd ICRC (Dublin), 1, 89.
\item Schaefer, B.E.: 1994,
{\it AIP Conf. Proc. 307}, AIP:New York, 382.
\item Stecker, F.W. {\it et al.}, 1994, Nature, 369, 294.
\item Vallania, P. {\it et al.}: 1993,
{\it Proc. 23rd ICRC}, Calgary, {\bf 1}, 61.
\end{refs}
\end{document}
|
\section*{\centering
References\\}\bigskip\list
{\arabic{enumi}.}{\settowidth\labelwidth{#1}\leftmargin\labelwidth
\advance\leftmargin\labelsep\usecounter{enumi}}
\def\hskip .11em plus .33em minus .07em{\hskip .11em plus .33em minus .07em}
\sloppy\clubpenalty4000\widowpenalty4000 \sfcode`\.=1000\relax}
\newcommand{\under}[2]{\mathrel{\mathop{#1}\limits_{\scriptstyle #2}}}
\def\op#1{\mathop{\fam0 #1}\limits}
\newcommand{{\rm Id\,}}{{\rm Id\,}}
\def{\rm Ker\,}{{\rm Ker\,}}
\newcommand{\begin{eqnarray}}{\begin{eqnarray}}
\newcommand{\end{eqnarray}}{\end{eqnarray}}
\newcommand{\begin{eqnarray*}}{\begin{eqnarray*}}
\newcommand{\end{eqnarray*}}{\end{eqnarray*}}
\newcommand{\begin{eqalph}}{\begin{eqalph}}
\newcommand{\end{eqalph}}{\end{eqalph}}
\newcommand{{\cal L}}{{\cal L}}
\newcommand{{\cal E}}{{\cal E}}
\newcommand{{\cal H}}{{\cal H}}
\newcommand{{\cal F}}{{\cal F}}
\newcommand{{\cal D}}{{\cal D}}
\newcommand{\alpha}{\alpha}
\newcommand{\beta}{\beta}
\newcommand{\theta}{\theta}
\newcommand{\Lambda}{\Lambda}
\newcommand{\Omega}{\Omega}
\newcommand{\pi}{\pi}
\newcommand{\delta}{\delta}
\newcommand{\lambda}{\lambda}
\newcommand{\omega}{\omega}
\newcommand{\mu}{\mu}
\newcommand{\nu}{\nu}
\newcommand{\otimes}{\otimes}
\newcommand{\gamma}{\gamma}
\newcommand{\Gamma}{\Gamma}
\newcommand{\varepsilon}{\varepsilon}
\newcommand{\epsilon}{\epsilon}
\newcommand{\sigma}{\sigma}
\newcommand{\Sigma}{\Sigma}
\newcommand{\wedge}{\wedge}
\newcommand{\widetilde}{\widetilde}
\newcommand{\widehat}{\widehat}
\newcommand{\overline}{\overline}
\newcommand{\partial}{\partial}
\newcounter{eqalph}
\newcounter{equationa}
\def\arabic{equationa}\alph{equation}{\arabic{equation}}
\newenvironment{eqalph}{\stepcounter{equation}
\setcounter{equationa}{\value{equation}}
\setcounter{equation}{0}
\def\arabic{equationa}\alph{equation}{\arabic{equationa}\alph{equation}}
\begin{eqnarray}}{\end{eqnarray}
\setcounter{equation}{\value{equationa}}}
\newenvironment{remark}{{\bf Remark.}}{$\Box$\medskip}
\hyphenation{ma-ni-fold La-gran-gi-ans di-men-si-o-nal
-di-men-si-o-nal La-gran-gi-an Ha-mil-to-ni-an}
\begin{document}
\hbox{}
\centerline{\large\bf FERMIONS IN GRAVITATION THEORY}
\bigskip
\centerline{\bf Gennadi A Sardanashvily}
\bigskip
\centerline{Department of Theoretical Physics}
\centerline{Moscow State University, 117234 Moscow, Russia}
\centerline{<EMAIL>}
\bigskip
\bigskip
\centerline{\sc ABSTRACT}
\bigskip
\noindent
In gravitation theory, a fermion field must be regarded only in a pair with a
certain tetrad gravitational field. These pairs can be represented by
sections of the composite spinor bundle
$$S\to\Sigma\to X^4$$
where values of gravitational fields play the role of
parameter coordinates, besides the familiar world coordinates.
\bigskip
\section{Problem}
There are different spinor bundles
$$
S\longrightarrow X^4
$$
over a world manifol $X^4$.
To describe realistic fermion fields $s$,
one must construct the representation
$$
dx^\mu \longrightarrow \gamma^a
$$
of cotangent vectors to $X^4$ by the Dirac's $\gamma$-matrices.
The problem lies in the fact that the tranformation group of cotangent
vectors is $$GL(4,{\bf R}),$$ whereas that of the $\gamma$-matrices is the
Lorentz group $$SO(3,1).$$
Given a gravitational field $h$, we have such a representation
$$
\gamma_h: dx^\mu \longrightarrow \widehat dx^\mu =h^\mu_a\gamma^a.
$$
However, different gravitational fields $h$ and $h'$ yield the nonequivalent
representations
$$
\gamma_{h'\neq h} \neq \gamma_h.
$$
It
follows that a fermion field must be regarded only in a pair with a
certain gravitational field. These pairs $(s_h,h)$ can not be represented by
sections of any one spinor bundle over $X^4$, but the composite spinor bundle
$$
S\longrightarrow\Sigma\longrightarrow X^4
$$
over the manifold $\Sigma$ which is coordinatized by values $ \sigma^\mu_a$
of gravitational fields, besides the familiar world coordinates $x^\mu$.
\section{Technical preliminary}
We follow the generally accepted geometric description of classical fields
as sections of bundles $$\pi: Y\to X.$$
Their dynamics is phrased in terms of jet manifolds. In field theory, we can
restrict our consideration to the first order Lagrangian formalism. In
this case, the jet manifold $J^1Y$ plays the role of a finite-dimensional
configuration space of fields.
Recall that the 1-order jet manifold $J^1Y$ of a bundle
$Y\to X$ comprises the equivalence classes
$j^1_xs$, $x\in X$, of sections $s$ of $Y$ identified by their values
and the values of their first derivatives
at $x$. Given bundle coordinates $(x^\lambda,y^i)$
of $Y\to X$, it is endowed with the adapted
coordinates $(x^\lambda,y^i,y^i_\lambda)$ where coordinates $y^i_\lambda$ make the
sense of values of first order partial derivatives $\partial_\lambda y^i(x)$ of field
functions $y^i(x)$.
Jet manifolds have been widely used in the
theory of differential operators. Their application to differential
geometry is based on the 1:1 correspondence
between the connections
on a bundle $Y\to X$ and the global sections
\[
\Gamma=dx^\lambda\otimes(\partial_\lambda+\Gamma^i_\lambda(y)\partial_i)
\]
of the jet bundle
$J^1Y\to Y$. The jet bundle $J^1Y\to Y$
is an affine bundle modelled on the
vector bundle
\[
T^*X\op\otimes_Y VY
\]
where $VY$ denotes the vertical tangent bundle of $Y\to X$.
In jet terms, a first order Lagrangian density of fields is represented
by an exterior horizontal density
\[
L={\cal L}(x^\mu, y^i, y^i_\mu)\omega, \qquad \omega=dx^1\wedge...\wedge dx^n,
\]
on $J^1Y\to X$.
In field theory, all Lagrangian densities are polynomial forms on the
affine bundle $J^1Y\to Y$
with respect to velocities $y^i_\lambda$.
They are factorized as
$$
L: J^1Y\op\to^D T^*X\op\otimes_Y VY\to\op\wedge^n T^*X
$$
where $D$ is the covariant differential
\[
D=(y^i_\lambda-\Gamma^i_\lambda)dx^\lambda\otimes\partial_i
\]
with respect to some connection $\Gamma$
on $Y\to X$.
\section{World manifold}
In gravitation theory, $X$ is
a 4-dimensional oriented world manifold $X^4$.
The tangent bundle $TX$ and the cotangent bundle
$T^*X$ of $X^4$ are provided with atlases of the induced coordinates
$(x^\lambda, \dot x^\lambda)$ and $(x^\lambda,\dot x_\lambda)$ relative to the holonomic
fibre bases $\{\partial_\lambda\}$ and $\{dx^\lambda\}$ respectively.
The structure group of these bundle is the general linear group
\[
GL_4=GL^+(4,{\bf R}).
\]
The associated principal bundle is the bundle $LX\to X^4$ of linear frames
in tangent spaces to $X^4$.
A family $\{z_\xi\}$ of local sections of the
principal bundle $LX$ sets up an atlas of $LX$ and the associated atlases
of $TX$ and $T^*X$ which we treat conventionally the distribution of
local reference frames $z_\xi(x)$ at points $x\in X^4$.
\section{Dirac spinors}
Different spinor models of the fermion matter have been suggested.
But all
observable fermion particles are the Dirac fermions. There are several
ways to introduce Dirac fermion fields. We follo the algebraic approach.
Given
a Minkowski space $M$ with the Minkowski metric $\eta$, let
${\bf C}_{1,3}$ be the complex Clifford algebra generated by elements
of $M$.
A spinor space $V$ is defined to be a minimal left ideal of ${\bf C}_{1,3}$ on
which this algebra acts on the left. We have the {\bf representation}
\begin{eqnarray}
&&\gamma: M\otimes V \to V, \nonumber\\
&& \widehat e^a=\gamma(e^a)=\gamma^a,\label{521}
\end{eqnarray}
of elements of the Minkowski space $M$ by
Dirac's matrices $\gamma$ on $V$.
Let us consider the transformations preserving the representation (\ref{521}).
These are pairs $(l,l_s)$ of Lorentz transformations $l$ of the Minkowski
space $M$ and invertible elements $l_s$ of ${\bf C}_{1,3}$ such that
\[\gamma (lM\otimes l_sV)=l_s\gamma (M\otimes V).\]
Elements $l_s$ constitute the Clifford group whose action on $M$
however fails to be effective. Therefore we restrict ourselves to its spinor
subgroup $ SL(2,{\bf C})$. For the sake of simplicity, let us identify
$ SL(2,{\bf C})$ with the Lorentz group $L=SO(3,1)$ whose generators on
the spinor space $V$ read
\[
I_{ab}=\frac14[\gamma_a,\gamma_b].
\]
\section{Dirac spinor fields}
To describe spinor fields,
let us consider a bundle of complex Clifford algebras ${\bf C}_{3,1}$
over $X^4$. It contains both a spinor bundle $S_M\to X^4$ and the
bundle $Y_M\to X^4$ of Minkowski spaces of generating elements of
${\bf C}_{3,1}$.
{\bf To describe Dirac fermion fields on a world manifold, one must
require that $Y_M$ is isomorphic to the cotangent bundle $T^*X$
of a world manifold $X^4$.} It takes place if only the structure group of
$LX$ is reducible to the Lorentz group
$L$ and $LX$
contains a reduced $L$ subbundle $L^hX$ such that the bundles
are associated with
\begin{eqnarray*}
&& Y_M=Y_h=(L^hX\times M)/L,\\
&& S_M=S_h=(L_hX\times V)/L
\end{eqnarray*}
are associated with $L^hX$.
In accordance with the well-known theorem, there is the 1:1 correspondence
between the reduced subbubdles $L^hX$ of $LX$ and
the {\bf tetrad gravitational fields} $h$ identified with global sections
of the quotient bundle
\[
\Sigma:= LX/L\to X^4.
\]
This bundle is the 2-fold cover of the bundle $\Sigma_g$ of pseudo-Riemannian
bilinear forms in cotangent spaces to $X^4$. Global sections of $\Sigma_g$
are pseudo-Riemannian metrics $g$ on $X^4$.
{\bf It follows that existence of a
gravitational field is necessary condition in order that Dirac fermion
fields live on a world manifold.}
Given a tetrad field $h$, let $\Psi^h$ be an atlas of
$LX$ such that the corresponding local sections $z_\xi^h$ of $LX$
take their values in its reduced subbundle $L^hX$. This atlas has
$L$-valued transition functions. It fails to be a holonomic atlas in general.
With respect to an atlas $\Psi^h$ and a
holonomic atlas $\Psi^T=\{\psi_\xi^T\}$ of $LX$, the tetrad field $h$
can be represented by a family of $GL_4$-valued tetrad functions
\begin{eqnarray*}
&& h_\xi=\psi^T_\xi\circ z^h_\xi,\\
&&dx^\lambda= h^\lambda_a(x)h^a,
\end{eqnarray*}
which carry gauge transformations between the holonomic fibre bases
$\{dx^\lambda\}$ of $T^*X$ and the fibre bases $\{h^a\}$ associated with
$\Psi^h$. The well-known relation
\[
g^{\mu\nu}=h^\mu_ah^\nu_b\eta^{ab}
\]
takes place.
Given a tetrad field $h$, one can define the {\bf representation}
\[
\gamma_h: T^*X\otimes S_h=(L_hX\times (M\otimes V))/L\to (L_hX\times
\gamma(M\otimes V))/L=S_h,
\]
\[
\widehat dx^\lambda=\gamma_h(dx^\lambda)=h^\lambda_a(x)\gamma^a,
\]
of cotangent vectors to a world manifold $X^4$ by Dirac's $\gamma$-matrices
on elements of the spinor bundle $S_h$.
Let $A_h$ be a connection on $S_h$ associated with a principal
connection on $L^hX$ and $D$
the corresponding covariant differential. Given the
representation $\gamma_h$, one can construct the first order differential
{\bf Dirac operator}
\[
{\cal D}_h=\gamma_h\circ D: J^1S_h\to T^*X\op\otimes_{S_h}VS_h\to VS_h
\]
on $S_h$.
Thus, one can say that sections of
the spinor bundle $S_h$ describe Dirac fermion fields in the presence of
the tetrad gravitational field $h$.
The crucial point lies in the fact that,
for different tetrad fields $h$ and $h'$,
the representations $\gamma_h$ and $\gamma_{h'}$
are {\bf not equivalent}.
It follows that
Dirac fermion field must be regarded only in a pair with a certain
tetrad gravitational field $h$. These pairs constitute the so-called
{\bf fermion-gravitation complex}. They can not be represented by
sections of any product $$S\op\times_{X^4}\Sigma$$
where $S\to X^4$ is some standard
spinor bundle.
At the same time, there is the 1:1 correspondence between
these pairs and the sections of the {\bf composite spinor bundle}
\[
S\to\Sigma\to X^4
\]
where $S\to\Sigma$ is a spinor bundle associated with the $L$ principal
bundle $LX\to\Sigma$. In particular, every spinor bundle
$S_h$ is {\bf isomorphic to the restriction of} $S$ to $h(X^4)\subset\Sigma$.
\section{Composite bundles}
By a composite bundle is meant the composition
\begin{equation}
Y\to \Sigma\to X \label{I1}
\end{equation}
of bundles $Y_\Sigma :=Y\to\Sigma$ and $\Sigma\to X$. It is
provided with the bundle coordinates $( x^\lambda ,\sigma^m, y^i) $ where
$( x^\lambda ,\sigma^m)$ are fibred coordinates of $\Sigma$.
Application of composite bundles to field theory is
founded on the following speculations. Given
a global section $h$ of $\Sigma$, the restriction $ Y_h$
of $Y_\Sigma$ to $h(X)$ is a subbundle
of the composite bundle $Y\to X$. There is the 1:1 correspondence between
the global sections $s_h$ of $Y_h$ and the global sections of
the composite bundle (\ref{I1}) which cover $h$.
{\bf Therefore, one can say that sections $s_h$ of $Y_h$
describe fields in the presence of a background parameter
field $h$, whereas sections
of the composite bundle $Y$ describe all pairs $(s_h,h)$.}
The configuration space of these pairs is the
first order jet manifold $J^1Y$ of the composite bundle (\ref{I1}).
The feature of the dynamics of field systems on composite bundles
consists in the following.
Let $Y$ be a composite bundle (\ref{I1}). Every connection
\[
A=dx^\lambda\otimes(\partial_\lambda+ A^i_\lambda\partial_i)
+ d\sigma^m\otimes(\partial_m+A^i_m\partial_i)
\]
on $Y\to\Sigma$ yields splitting
\[
VY=VY_\Sigma\op\oplus_Y (Y\op\times_\Sigma V\Sigma)
\]
and, as a consequence,
the first order differential operator
\begin{eqnarray*}
&&\widetilde D: J^1Y\to T^*X\op\otimes_Y VY_\Sigma, \\
&&\widetilde D= dx^\lambda\otimes(y^i_\lambda- A^i_\lambda -A^i_m\sigma^m_\lambda)\partial_i,
\end{eqnarray*}
on $Y$.
Let $h$ be a global section
of $\Sigma$ and $Y_h$ the restriction of the bundle $Y_\Sigma$ to $h(X)$. The
restriction of $\widetilde D$ to $J^1Y_h\subset J^1Y$
{\bf comes to the familiar covariant differential} relative to a certain
connection on $Y_h$.
One can use the differential operator $\widetilde D$ in order to construct a
Lagrangian density of sections of the composite bundle $Y\to\Sigma\to X$.
It is never regular, and we have the {\bf constraint conditions}
\[
\partial^\mu_m{\cal L} +A^i_m\partial^\mu_i{\cal L} =0.
\]
\section{Composite spinor bundles}
In gravitation theory, we have the composite bundle
\[
LX\to\Sigma\to X^4
\]
where $\Sigma=LX/L$ and
\[ LX_\Sigma:=LX\to\Sigma\]
is the trivial principal bundle with the structure Lorentz group $L$.
Let us consider the composite spinor bundle $$S\to\Sigma\to X^4$$ where the
spinor bundle
$S_\Sigma:= S\to\Sigma$
is associated with the $L$-principal bundle $LX_\Sigma$, that is,
\[
S_\Sigma= (LX_\Sigma\times V)/L.
\]
It is readily observed
that, given a global section $h$ of $\Sigma$, the restriction $S_\Sigma$ to
$h(X^4)$ is the spinor bundle $S_h$ whose sections describe Dirac
fermion fields in the presence of the tetrad field $h$.
Let us provide the principal bundle $LX$ with a holonomic atlas
$\{\psi^T_\xi\}$ and the principal bundle $LX_\Sigma$
with an atlas $\Psi^h=\{z^h_\xi\}$.
With respect to these atlases, the composite spinor bundle $S\to\Sigma\to
X^4$ is coordinatized by $(x^\lambda,\sigma_a^\mu, y^A)$ where $(x^\lambda,
\sigma_a^\mu)$ are coordinates of the bundle $\Sigma$ such that,
given a section $h$ of $\Sigma$, we have the tetrad functions
\[
(\sigma^\lambda_a\circ h)(x)= h^\lambda_a(x).
\]
The corresponding jet manifold $J^1S$ is coordinatized by
\[
(x^\lambda,\sigma^\mu_a, y^A,\sigma^\mu_{a\lambda}, y^A_\lambda).
\]
Note that, whenever $h$, the jet manifold $J^1S_h$ is a
subbundle of $J^1S\to X^4$ given by the coordinate relations
\[\sigma^\mu_a=h^\mu_a(x), \qquad \sigma^\mu_{a\lambda}=\partial_\lambda h^\mu_a(x).\]
Let us consider the bundle of Minkowski spaces
\[(LX\times M)/L\to\Sigma\]
associated with the $L$-principal bundle $LX_\Sigma$. Since $LX_\Sigma$ is
trivial, it is isomorphic to the pullback $\Sigma\op\times_X T^*X$
which we denote by the same symbol $T^*X$.
Then one can define the {\bf representation}
\[
\gamma_\Sigma: T^*X\op\otimes_\Sigma S_\Sigma= (LX_\Sigma\times (M\otimes V))/L
\to (LX_\Sigma\times\gamma(M\otimes V))/L=S_\Sigma,
\]
\[\widehat dx^\lambda=\gamma_\Sigma (dx^\lambda) =\sigma^\lambda_a\gamma^a,\]
over $\Sigma$. When restricted to $h(X^4)\subset \Sigma$,
the morphism $\gamma_\Sigma$ {\bf comes to} the morphism $\gamma_h$.
We use this morphism in order to construct the total Dirac
operator on sections of the composite spinor bundle $S\to\Sigma\to X^4$.
Let
\begin{equation}
A=dx^\lambda\otimes (\partial_\lambda + A^B_\lambda\partial_B) + d\sigma^\mu_a\otimes
(\partial^a_\mu+A^B{}^a_\mu\partial_B) \label{200}
\end{equation}
be a connection on the bundle $S_\Sigma$. It yields the
splitting of the vertical tangent bundle $VS$ and determines the modified
differential
\[
\widetilde D= dx^\lambda\otimes(y^B_\lambda- A^B_\lambda -A^B{}_\mu^a\sigma^\mu_{a\lambda})\partial_B
\]
which one can use in order to construct a Lagrangian density of the
total fermion-gravitation complex.
The composition of
the morphisms $\gamma_\Sigma$ and $\widetilde D$ is the first order differential
operator
\[{\cal D}=\gamma_\Sigma\circ\widetilde D:J^1S\to T^*X\op\otimes_SVS_\Sigma\to VS_\Sigma,\]
\[\dot y^A\circ{\cal D}=\sigma^\lambda_a\gamma^{aA}{}_B(y^B_\lambda- A^B_\lambda -
A^B{}^c_\mu\sigma^\mu_{c\lambda}),\] on $S$.
One can think of it as being the {\bf total Dirac operator} since, whenever a
tetrad field $h$, the restriction of ${\cal D}$ to $J^1S_h\subset J^1S$ {\bf
comes to} the Dirac operator ${\cal D}_h$
relative to the connection
\begin{equation}
A_h=dx^\lambda\otimes[\partial_\lambda+(\widetilde A^B_\lambda+A^B{}^a_\mu\partial_\lambda h^\mu_a)\partial_B] \label{L10}
\end{equation}
on $S_h$.
One can provide the explicit form of the total Dirac operator where
\begin{eqnarray}
&&A^B_\lambda=A^{ab}{}_\lambda I_{ab}{}^B{}_Cy^C=\frac12 K^\nu{}_{\mu\lambda}
\sigma^\lambda_c (\eta^{cb}\sigma^a_\nu -\eta^{ca}\sigma^b_\nu )I_{ab}{}^B{}_Cy^C,\nonumber\\
&&A^B{}_\mu^c=A^{ab}{}^c_\mu I_{ab}{}^B{}_Cy^C=
\frac12(\eta^{cb}\sigma^a_\mu -\eta^{ca}\sigma^b_\mu)I_{ab}{}^B{}_Cy^C,\label{M4}
\end{eqnarray}
where $K$ is some symmetric connection on $TX$ and (\ref{M4})
corresponds to the {\bf canonical left-invariant free-curvature connection} on
the bundle
$GL_4\to GL_4/L.$
Given a tetrad field $h$, the connection (\ref{L10}) {\bf is reduced to} the
Levi-Civita connection on $L^hX$.
In particular, we get the constraints of the total fermion-gravitation complex
\begin{equation}
P^{c\lambda}_\mu+\frac18\eta^{cb}\sigma^a_\mu(y^B[\gamma_a,\gamma_b]^A{}_B
P^\lambda_A+P^{A\lambda}_+[\gamma_a,\gamma_b]^{+B}{}_Ay^+_B)=0 \label{M2}
\end{equation}
where $P^{c\lambda}_\mu$ and
$P^\lambda_A$ are the momenta corresponding the tetrad and spinor coordinates
$(\sigma^\mu_c,y^A)$ of the
composite spinor bundle $S\to\Sigma\to X^4$. The condition (\ref{M2}) modifies
the standard gravitational constraints
\[
P^{c\lambda}_\mu=0.
\]
\bigskip
\bigskip
\centerline{\bf References}
\bigskip
\begin{itemize}
\item
G.Sardanashvily and O. Zakharov, {\it Gauge Gravitation Theory}, 1992,
World Scientific, Singapore
\item
G.Sardanashvily, On the geometry of spontaneous symmetry breaking,
{\it Journal of Mathematical Physics}, 1992, {\bf 33}, 1546
\item
G.Sardanashvily, {\it Gauge Theory in Jet Manifolds}, 1993, Hadronic
Press Inc., Palm Harbor
\item
G.Sardanashvily, Differential geometry of composite fibred manifolds,
E-print: dg-ga/9412002
\item
G.Sardanashvily, Composite spinor bundles in gravitation theory,
E-print: gr-qc/9502022
\end{itemize}
\end{document}
|
\section{Introduction}
In this letter, we consider the pure electroweak three loop contribution
to the neutron electric
dipole moment in the Kobayashi-Maskawa (KM)
model with four generations. Besides its
interesting predictions for B-meson physics \cite{HSS}, the
enlarged variant of the KM model leads to the enhancement of the neutron
electric dipole moment (EDM) in comparison with Standard Model
case \cite{HamPos}.
This enhancement is linked with the short distance contribution to
the electric and chromoelectric dipole moments (CEDM) of quarks.
The origin of CP-violation in the Kobayashi-Maskawa (KM) model resides in
the complexity of the quark mixing matrix. This requires four
semi-weak vertices to generate a flavour-diagonal CP-odd
amplitude. It was shown that the EDMs and CEDMs of quarks cannot be generated
at the lowest possible two-loop order
\cite{Shab}. Thus, the QCD corrections are brought in to
prevent these quantities from the identical cancellation.
The detailed study of these operators at three loop order (two
electroweak plus one gluonic) was done in
SM by Khriplovich \cite{Kh} and by us for its four generation
modification \cite{HamPos}.
Here, we replace the hard gluon loop by another one of electroweak
in the EDM inducing graphs.
Specifically, we concentrate on the
large renormalization factor of the axial coupling of Z-boson
with fermion proportional to $m_t^2$ in SM
and its relevance for the induced EDM in the presence
of an extra heavy generation of quarks \cite{NOV}.
The purpose of this work is to compute EDMs and CEDMs at three
loop electroweak order and then compare our results with
corresponding QCD induced values \cite{HamPos,Kh}.
\section{Electroweak corrections to EDM}
At first glance the problem of this calculation appears
to be very complicated.
However, taking into account the explicit mass hierarchy in this problem,
we reduce the three-loop calculation
to one-loop integrals.
We assume that:
\begin{equation}
m_h^2;\;m_g^2\gg m_t^2\gg m_w^2\gg m_i^2,
\label{eq:scale}
\end{equation}
where we have denoted the heaviest quarks as h and g; i
represents the standard set of "light" flavours: u, d, s, c and b.
The first inequality is assumed in order to single
out parametrically the contributions proportional to $m_h^2$ or
$m_g^2$.
The typical representatives of the diagrams
to be calculated are depicted in Fig. 1. The solid line represents
the fermions; wavy lines are charged electroweak bosons
and the dashed line are the neutral ones. The position of an
external photon or gluon is not indicated here.
In the following calculation, we consider the EDM and CEDM
operators of the strange quark. Then the arrangement of flavours
along the fermion
line is determined in SM uniquely by \cite{Kh}:
\begin{equation}
i\mbox{Im}(V^*_{ts}V_{td}V^*_{cd}V_{cs})
s[t(b-d)u-u(b-d)t+u(b-d)c-c(b-d)u+c(b-d)t-t(b-d)c]s.
\label{eq:FS1}
\end{equation}
The enlarged KM matrix possesses in general three independent
CP-odd invariants, one of which being
distinguished by the dynamical
enhancement \cite{HamPos}:
\begin{equation}
i\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})
s[t(b-g)h-h(b-g)t+U(b-g)t-t(b-g)U+h(b-g)U-U(b-g)h]s.
\label{eq:FSfin}
\end{equation}
The capital U here denotes the propagation of u and c quarks which we
are free to regard massless and degenerate inside the loops.
This degeneracy is the factor which leads to the identical cancellation of
the amplitude (\ref{eq:FS1}) in the SM. Therefore (\ref{eq:FS1})
is forced to be
proportional to $m_c^2$ whereas (\ref{eq:FSfin}) is determined
by heavy mass scale like $m_t^2$ \cite{HamPos}. This is the main
source of the EDM enhancement in the KM model with four
generations. The additional source of the enhancement probably lies in
the numerical significance of the CP-odd combination
$\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})$ which could naturally reach
the level of $\lambda^5$, where $\lambda$ is the Wolfenstein
parameter \cite{HSS,HamPos}.
We begin from the shortest distances and calculate first the
effective one-loop induced flavour changing neutral currents
"electroweak penguin", which is well known from the
kaon physics:
\begin{equation}
{\cal L}^{(1)}=\sum_{i,j,f}a_jV_{ij}V^*_{fj}\bar{q}_i\mbox{$\gamma_{\mu}$}(1-\mbox{$\gamma_{5}$})
q_fZ_\mu
\end{equation}
To a good accuracy, it is sufficient to take only leading
contribution to the coefficients $a_j$ which are proportional
to the square of the heaviest masses:
\begin{equation}
{\cal L}^{(1)}=\fr{g_w}{\cos\theta_W}\fr{\mbox{$\alpha$}_w}{16\pi}Z_\mu\left[
\fr{m_h^2}{m_w^2}\sum_{i,f}V_{hi}V^*_{hf}\bar{q}_f\mbox{$\gamma_{\mu}$}\fr{(1-\mbox{$\gamma_{5}$})}{2}
q_i-
\fr{m_g^2}{m_w^2}\sum_{i',f'}V_{i'g}V^*_{f'g}\bar{q}_{f'}\mbox{$\gamma_{\mu}$}\fr{(1-\mbox{$\gamma_{5}$})}{2}
q_{i'}\right],
\label{eq:eff1}
\end{equation}
in analogy to the SM expression with dependence on $m_t^2$.
The heavy mass dependence here originates from the longitudinal
part of W-boson propagator or from the equivalent graph with charged
non-physical higgs boson \cite{NOV}.
Taking the new effective vertex
generated by (\ref{eq:eff1}), we reduce the remaining computation to
the one
presented in Fig. 2.
The second step is to integrate out the neutral
boson, which could be done along the same line. At
this point, however, it is useful to treat the three and four
generation models separately.
In the four generation case, the characteristic
loop momenta are large, which allow us again to take only the longitudinal
part of the Z-boson propagator. This leads us
to obtain effective charged right handed
currents in an easy way. Taking into account the flavour structure
in (\ref{eq:FSfin}), we obtain the effective lagrangian for the s -
t transition:
\begin{equation}
{\cal L}^{(2)}\simeq -\fr{g_w}{\sqrt{2}}V_{tb}V^*_{hb}V_{hs}
(\fr{\mbox{$\alpha$}_w}{16\pi})^2
\fr{m_tm_s}{m_w^2}\fr{(m_h^2+m_g^2)}{m_w^2}\log(\fr{m_{h(g)}^2}{m_t^2})
\bar{q}_t\mbox{$\gamma_{\mu}$}\fr{(1+\mbox{$\gamma_{5}$})}{2}q_sW^+_\mu +(h.c.),
\label{eq:eff2}
\end{equation}
where we have omitted all constants in comparison with "large"
logarithmic factors. This logarithmic accuracy is motivated by
theoretical considerations. However, in the final numerical result,
we set all logarithms to
unity. This allows us to avoid the problem of a true
two-loop calculation which is not reducible to two independent
integrations. It is important to remember that the vertices
(\ref{eq:eff1}) and (\ref{eq:eff2}) are indeed effective and do
not survive if the incoming momenta are larger than all masses
of particles flowing inside the loops. For this reason, the upper limit
for logarithmic integral over the loop momentum coincides with
$m_h$ if $m_h<m_g$ and with $m_g$ if $m_g<m_h$. The analysis of
the precision electroweak data suggests that h and g quarks
must be sufficiently degenerate in masses. From here to below,
we put $m_g \simeq m_h$. It is worth to note also the constructive
interference between the two terms in (\ref{eq:eff2}) proportional
to $m_g^2$ and $m_h^2$, in contrary to the mass dependence of
the electroweak parameter $\rho$ \cite{NOV}. The contribution of
the fourth generation to $\rho$ vanishes at $m_h=m_g$.
The situation is quite different in the SM. Due to the presence of
right-handed currents in the s-c or d-c transition, those transition
amplitudes are suppressed not
only by $m_{s(d)}$ and $m_c$ but also by the factor
$m_b^2/m_Z^2$ reflecting the GIM property of
(\ref{eq:FS1}). Taking into account that the result for EDM of
s or d quark must be
proportional to $m_c^2$, although the interchange of a gluon loop by
an electroweak one replaces
$\mbox{$\alpha$}_s(q^2\simeq m_b^2)$ by
$\mbox{$\alpha$}_w\fr{m_t^2m_b^2}{m_w^4}$, we deduce that the total effect is
negligibly small.
The presence of right-handed currents itself does not
necessarily imply CP-violation. The latter arises through the
complex phase of $V_LV^{\dagger}_R$, the product of right- and left-handed
KM matrices,
\begin{equation}
{\cal L}=\fr{g_w}{2\sqrt{2}}W^+_\mu\sum_{f,i}
\left[\bar{q}_f(V_L)_{fi}\mbox{$\gamma_{\mu}$} (1-\mbox{$\gamma_{5}$})q_i+
\bar{q}_f(V_R)_{fi}\mbox{$\gamma_{\mu}$} (1+\mbox{$\gamma_{5}$})q_i\right]+(h.c.),
\end{equation}
and is known to exist even in two generations. This
is exactly what happens at the last stage of our calculation.
EDM and CEDM of s-quark results from the mixing between second
and third generations in the presence of right-handed currents
given by (\ref{eq:eff2}) depicted in Fig. 3. The results are:
\begin{eqnarray}
\label{eq:EDM}
d_s=-\fr{5e}{3}\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})
(\frac{\alpha_w}{4\pi})^2\frac{1}{16\pi^2}\fr{G_F}{\sqrt{2}}m_s
\frac{m_t^2m_h^2}{4m_w^4}\log(\fr{m_{h}^2}{m_t^2})\\
\tilde{d}_s=-g_s\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})
(\frac{\alpha_w}{4\pi})^2\frac{1}{16\pi^2}\fr{G_F}{\sqrt{2}}m_s
\frac{m_t^2m_h^2}{4m_w^4}\log(\fr{m_{h}^2}{m_t^2}),
\label{eq:CEDM}
\end{eqnarray}
where $d_s$ and $\tilde{d_s}$ are determined as the coefficients
in front of $\fr{1}{2}\bar{q}(F\mbox{$\sigma$})\mbox{$\gamma_{5}$} q$ and
$\fr{1}{2}\bar{q}t^a(G^a\mbox{$\sigma$})\mbox{$\gamma_{5}$} q$ respectively. Due to
the inequality (\ref{eq:scale}), we have taken into account only the
longitudinal part of W-propagator. The use of longitudinal
parts of gauge boson propagators throughout the calculation of EDM in
four generation case maximizes the size of CP-violation. This
also means that the obtained result arises entirely from the Higgs sector
phenomenon. Then, the whole calculation could be performed in
the t'Hooft-Feynman gauge and therefore only scalar bosons should be taken
into account. Thus, we can rewrite the results given in
(\ref{eq:CEDM}) and (\ref{eq:EDM}) in terms of Yukawa couplings
$f_i=m_i/v$ of heavy quarks:
\begin{equation}
\fr{\tilde{d}_s}{g_s}=\fr{3d_s}{5e}=-\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})
\frac{1}{1024\pi^6}\fr{G_F}{\sqrt{2}}m_s
f_t^2f_h^2
\log(\fr{f_{h}^2}{f_t^2}),
\label{eq:ans}
\end{equation}
where $v=246 $GeV is the vacuum expectation value of the scalar
field.
The extraction of the EDM of neutron resulting from the effective
interaction (\ref{eq:CEDM}) depends on our understanding of
low-energy hadronic physics. Based on the chiral perturbation
estimation proposed in \cite{KKZ}, we find the EDM of neutron
at the level:
\begin{equation}
\label{eq:NEDM}
d_N\sim
e\mbox{Im}(V^*_{ts}V_{tb}V^*_{hb}V_{hs})f_t^2f_h^2\cdot
2\cdot10^{-26}\;cm
\end{equation}
The comparison of (\ref{eq:NEDM}) with the corresponding
QCD-induced contribution to EDM obtained earlier in
Ref.\cite{HamPos} shows, in principle, the same order of magnitude
for both results. In the most optimistic scenario about the
values of CP-odd phase invariant, combined with the masses of heavy
quarks around $500$GeV, we obtain the neutron EDM to arise at the
level:
\begin{equation}
d_N\sim10^{-29}\;e\cdot cm.
\label{eq:num}
\end{equation}
The main source of the numerical smallness is connected with
the tiny numerical coefficient in expression (\ref{eq:ans}).
\section{ Discussions and Conclusions}
We would like to point out
that there exists another contribution to the EDM of neutron
from the physical Higgs
boson loop which should be taken into account as well. One may
undertake that calculation along the same line and obtain the
effective right-handed currents as well. This means that the
value of EDM depends not only on unknown $m_g$ and $m_h$ but also on the
mass of real Higgs boson. To our logarithmic
accuracy, these corrections are unimportant if we believe that
$m_{Higgs}$ is also very large, somewhere around $m_{h(g)}$.
The last diagram which could contribute to EDM at this order is
the "rainbow" graph, Fig. 4, where all wavy lines are W-bosons.
It could be checked, however, that even if this diagram provides
any nonvanishing contribution, it cannot generate
an $m_{h(g)}^2$-dependence in the result.
Our estimate shows that the electroweak contributions to the EDM of
neutron are comparable with QCD ones in the case of four
generations and negligibly small in SM. The QCD effects dominate
over pure electroweak contribution at $m_{h(g)}\sim m_t$, whereas
at $m_{h(g)}\sim 500-600$GeV the latter dominates. The numerical result
(\ref{eq:num}) then could be regarded as the maximal value of
EDM which can be obtained through the Kobayashi-Maskawa type of
CP-violation in the presence of an additional heavy generation.
Despite the significant enhancement in comparison with SM case,
it is too low to be detected even at the future generation
of experiments aiming at the search of EDM. This implies also
that the reliable information and limits on the parameter of the
model with four generations may come only from the analysis of
the electroweak precision data and K, B-meson physics
\cite{HSS}. The latter case
requires a further analysis on the role of large electroweak
radiative corrections.
As a concluding remark, we have emphasized the main mechanism
for the induced EDM of neutron in the hypothetical case of two or
more additional heavy generations, (h, g); (h', g');..., with
large mixing between them. Then the maximum of CP-violation at
low energies comes from Weinberg operator \cite{Wein} which is
known to exist in KM model at the lowest possible three-loop
order \cite{Posp}. In the four generation case, this operator is
suppressed by the factor $m_b^2/m_w^2$ \cite{HamPos}, which
disappears at five or more generations. At the same time, there
is no severe limits on the mixing between heavy generations from
the low energy phenomenological data and one may expect the
corresponding CP-odd invariant, $\mbox{Im}(V^*_{tg'}V_{tb}V^*_{hb}V_{hg'})$
to be large.
\eject \hfill
\begin{flushleft}{\Large\bf Acknowledgments}\end{flushleft}
We thank I.B. Khriplovich, V.N. Novikov and M.I.
Vysotsky for attracting our attention to the significance of
electroweak radiative corrections to EDM and for helpful
discussions. We thank also F. Boudjema for his kind advise
and comments. The work of C.H. was partially funded by N.S.E.R.C.
of Canada, and the research of M.P. was supported in part by the
I.C.F.P.M., INTAS grant \#\,93-2492.
|
\section{Introduction}
A cellular automaton is a system of discrete variables on a lattice
which is updated according to some simple and local
rule~\cite{wolfram86a,toffoli87}. Though applications of such models
range from physics to biology to social science~\cite{weisbuch91}, one
of the most exciting areas of interest in recent years has been
hydrodynamics. This interest derives from the discovery by Frisch,
Hasslacher, and Pomeau in 1986 that discrete cellular-automaton models
of fluids, known specifically as {\em lattice-gas automata}, may be
constructed for the numerical solution of the Navier-Stokes
equations~\cite{frisch86,frisch87}. Since their introduction, lattice
gases have been used to study a variety of problems in hydrodynamics.
Perhaps the greatest interest in the method is for the simulation of
problems that either involve complex boundaries, such as porous
media~\cite{rothman88a}, or complex fluids, such as
suspensions~\cite{ladd88,vanderhoef90} or immiscible
mixtures~\cite{rothman88b,appert90,rothman94}.
As in most other endeavors in computational physics, there is a great
need for fast, low-cost simulations. Indeed, it has long been
recognized that the simplicity of lattice gases allows them to be
simulated on special-purpose hardware, called ``cellular-automata
machines''~\cite{margolus86,clouqueur87,toffoli87}. Such machines
have been constructed for the simulation of general cellular automata,
and have been applied to a variety of systems. They combine the
performance of supercomputers with the hardware simplicity of a
personal computer or workstation for these particular applications.
In this paper we describe an implementation of a three-dimensional
lattice-gas model~\cite{dhumieres86a} on a new cellular-automata
machine called CAM-8~\cite{margolus90,margolus95}. CAM-8 retains the
high performance-to-cost ratio of previous cellular-automata machines,
but with considerably increased flexibility. It is precisely this
flexibility which makes it attractive for simulating lattice gases.
Lattice gases (as well as other cellular automata) may be simulated by
the application of a lookup table, which gives an output state from the
input state at every site of the lattice. The size of this table
increases with the number of degrees of freedom per site. For this
reason lattice gases for three-dimensional
hydrodynamics~\cite{dhumieres86a}, which typically require 24 bits of
state (and hence $2^{24}$ possible states) per site, have proved
challenging to implement in situations where memory is limited, such as
on a distributed-memory multiprocessor. Following the pioneering work
of H\'{e}non~\cite{henon87a,henon87b,henon89}, a considerable reduction
of table size was achieved by Somers and Rem~\cite{somers92}. Precisely
what algorithm and collision table to use, however, remains a
machine-dependent question. Because the CAM-8 allows one to work most
efficiently with units of 16 bits of state per site at any instant of
time, we have chosen to design a new collision strategy built upon
successive 16-bit table lookups. Our results, in terms of minimization
of viscosity (and thus maximization of efficiency for the types of
simulations that interest us) are roughly comparable to those obtained
by previous workers.
The outline of the paper is as follows. We first briefly summarize the
principal architectural features of the CAM-8. We then review the key
features of the face-centred hypercubic (FCHC) lattice-gas model, upon
which our implementation is based. Next,
we show how the collision step of the FCHC lattice gas can be broken
down into operations that involve no more than 16 bits at a time. We
also show how the collision rules can be optimized in a simple way to
give a small viscosity. We compare our theoretical predictions of the
viscosity to results obtained by numerical simulation on CAM-8. Finally,
we report a number of test simulations of flow through channels, pipes,
and a periodic array of spheres, and compare our results to analytic
predictions. Our goal in this latter exercise is to provide some
practical guidelines for the application of our methods to the
measurement of the permeability of disordered media.
\section{CAM-8}
CAM-8 is an indefinitely scalable parallel computer optimized for the
large-scale simulation of three-dimensional cellular automata (CA)
systems. It emphasizes simulation size, flexibility, and cost
efficiency, rather than ultimate performance---this emphasis leads to
a {\em virtual processor} design\cite{margolus95}.
\subsection{Hardware overview}
In CAM-8, each processor simulates the operation of up to millions of
spatial sites, with the site data stored in conventional DRAM-memory
chips. This DRAM is scanned sequentially in an addressing pattern
that maximizes the speed of access, and data is read, updated (by
SRAM-memory table lookup) and put back. This update cycle, which is
illustrated in Figure~\ref{fig1}a, resembles the operation of a video
framebuffer, and indeed CAM-8 is genetically more closely related to
framebuffer hardware than to conventional microprocessors.
Figure~\ref{fig1}b schematically illustrates an array of CAM-8
processing nodes (the array is actually three-dimensional in the
machine). A uniform spatial calculation is handled in parallel by
these nodes, with each node containing an equal ``chunk'' of the
space. In the diagram the solid lines between nodes indicate a local
mesh interconnection, used for spatial data movement between
processors. Since each processor handles millions of spatial sites,
but only one at a time, these mesh wires are time-shared among
millions of sites. The consequent reduction in interconnect, compared
to what would be needed in a fully parallel machine, makes large
three-dimensional simulations practical. There is also a
communications network (not shown) connecting all nodes to a front-end
workstation that controls the CAM-8 machine. The workstation uses
this network to broadcast simulation parameters to all or some of the
processing nodes.
Although the initial CAM-8 chip design accomodates machines with many
thousands of processing nodes, the first prototypes had only 8 nodes.
All of the simulations discussed in this paper were performed on such
an 8-node prototype, which has about the same amount and quality of
hardware as one might find in a low-end workstation: 2 Mbytes of SRAM,
64 Mbytes of DRAM, about two million gates of 1.2 micron CMOS logic,
with the whole thing running at 25 MHz.
\subsection{Programmable resources}
CAM-8 is a virtual-processor simulator of a fully parallel CA space.
Its virtual nature makes it possible to reconfigure and redirect its
computing resources, allowing the programmer to directly control
parameters such as
\begin{itemize}
\item
Number of dimensions
\item
Size and shape of the space
\item
Number of bits at a site
\item
Directions and distances of data movement
\item
Rules of data interaction (lookup tables)
\end{itemize}
The most novel facet of CAM-8's operation is the data movement, which
is lattice-gas-like. Corresponding bits, one from each spatial site,
constitute a {\em bit-field}: each bit-field can be shifted in any
direction by a chosen amount. The direction is chosen independently
for each bit-field, and the amount can be quite large---up to a few
thousand positions.\footnote{Such large spatial shifts are useful in
generating the high-quality random variables that are needed by
statistical mechanical simulations.} Data movements are uniform
across the entire space---the hardware hides the fact that the space
is divided up among separate processor nodes.
The hardware only allows 16 separate bit-fields to be manipulated at a
time. The 16 bits that ``collide'' at a given spatial site (i.e.,
land there due to their bit-field shifts) are replaced with a new
value given by a lookup table. In fact, all data interaction is
performed by 16-input/16-output lookup tables. Interactions that
involve more than 16 bits at each site must be synthesized as a
sequence of 16-bit interactions.
The lookup tables in the hardware are double-buffered: this means that
while one table is being actively used to scan all the sites and
update each in turn, the front-end workstation can broadcast the table
that will be used next. Since the data movement is controlled by a
few pointers within each processor, these movements can be changed
quickly from one scan to the next. Thus there is very little overhead
involved in applying a different lookup table {\em and} a different
set of bit-field movements at every scan of the space.
Finally, we note that most of our data analysis and data collection is
performed directly by dedicated statistics-gathering hardware. The CA
space is split up into bins of a chosen size and lookup tables are
used to evaluate a function on each of the sites in each bin. These
function data are collected by event counters and are continuously
reported back to the front-end as the simulation runs.
\section{The FCHC lattice gas}
Two fluids with quite different microscopic interactions may still have
the same macroscopic behavior. The reason for this is that the form of
the macroscopic equations of motion that describe this behavior depend
only on the conservation laws obeyed by these interactions and not on
their detailed form. This was one of the main motivations for the
introduction of lattice-gas automata as a method to simulate fluid flow.
A lattice gas models a fluid as a large number of particles undergoing
simple interactions that conserve mass and momentum. While the aim of
molecular dynamics is to simulate the real physics at the molecular, or
microscopic, scale, the microworld of lattice gases is fictitious.
Nevertheless, realistic macroscopic behavior is recovered when space and
time averages are performed.
One of the first lattice-gas models was introduced by Hardy, de Pazzis,
and Pomeau in 1976~\cite{hardy76}. This model, constructed on a square
lattice, succeeded in describing isotropically propagating sound waves,
and, if two particle species were introduced, diffusion. The
hydrodynamics of this model, however, were anisotropic, as was the
damping of the sound waves.
The simplest and first lattice-gas model that produced isotropic
two-dimensional hydrodynamic behavior was introduced on a triangular
lattice by Frisch, Hasslacher, and Pomeau in
1986~\cite{frisch86,frisch87,doolen91,rothman94}. The triangular
lattice is crucial for the isotropy of the flow dynamics, which,
technically, relies on the isotropy of the second and fourth order
tensors constructed from the basis vectors of the lattice. In three
dimensions no regular lattice with this property exists. However, the
four-dimensional face-centered hyper-cubic (FCHC) lattice has the
required symmetry~\cite{dhumieres86a}, as do its projections to three
dimensions.
\subsection{Description of the FCHC lattice}
The FCHC lattice is formed from the 4 basis vectors, $(\pm 1, \pm 1 , 0
,0)$, and the 20 additional vectors obtained by permuting the components
of these vectors. The FCHC lattice thus has a total of 24 basis vectors,
and we list these as follows:
{\footnotesize
\begin {equation}
\begin{array}{ccc}
{A}& {B}& {C} \\
\begin{array}{rrrrrr}
(&1& 0& 0& 1&) \\
(&1& 0& 0& -1& )\\
(& -1& 0& 0& 1&) \\
(& -1& 0& 0& -1&) \\
(& 0,& 1,& 1,& 0 &)\\
(& 0,& 1,& -1,& 0 &)\\
(& 0,& -1,& 1,& 0 &)\\
(&0,& -1,& -1,& 0 &)
\end{array}
&
\begin{array}{rrrrrr}
(& 0,& 1,& 0,& 1&) \\
(& 0,& 1,& 0,& -1& )\\
(& 0,& -1,& 0,& 1&) \\
(& 0,& -1,& 0,& -1&) \\
(& 1,& 0,& 1,& 0 &)\\
(& -1,& 0,& 1,& 0 &)\\
(& 1,& 0,& -1,& 0 &)\\
(& -1,& 0,& -1,& 0&)
\end{array}
&
\begin{array}{rrrrrr}
(& 0,& 0,& 1,& 1&) \\
(& 0,& 0,& 1,& -1&) \\
(&0,& 0,& -1,& 1&) \\
(& 0,& 0,& -1,& -1 &)\\
(& 1,& 1,& 0,& 0&) \\
(& 1,& -1,& 0,& 0 &)\\
(& -1,& 1,& 0,& 0& )\\
(& -1,& -1,& 0,& 0 &)
\end{array}
\end{array}.
\label{vectors}
\end{equation}
}
This grouping of the lattice vectors into three subgroups of eight will
be discussed further below. For now we note that, if $\theta (n_A,m_B)$
is the angle between the $n$th vector of group $A$ and the $m$th vector
of group $B$, then $\theta (n_A,m_B) = \theta (n_B,m_C) = \theta
(n_C,m_A)$, and $\theta (n_A,m_A) = \theta (n_B,m_B) = \theta
(n_C,m_C)$, and each group contains 4 oppositely oriented pairs of
vectors. In other words, the ordering of the velocity vectors is such
that group $B$ is related to group $A$ in exactly the same way as group
$C$ is related to $B$ and $A$ is related to $C$.
It is possible to visualize the FCHC lattice by the geometrical
construction shown in Figure~\ref{fchcviz}. Consider four unit circles
packed into a square of side 4 in two dimensions. It is straightforward
to see that the inscribed circle, tangent to each of these four unit
circles, has radius $\sqrt{2}-1$. In three dimensions, when eight unit
spheres are packed into a cube of side 4, the inscribed sphere has
radius $\sqrt{3}-1$. Likewise, in $D$ dimensions, when $2^D$ unit
hyperspheres are packed into a hypercube of side 4, the inscribed
hypersphere has radius $\sqrt{D}-1$. In particular, in four dimensions
($D=4$), the inscribed hypersphere is also a unit hypersphere. This
nontrivial packing of unit hyperspheres provides an
alternative description, within a factor of $\sqrt{2}$,
of the FCHC lattice: the centers of the
hyperspheres are the lattice vertices, and tangent hyperspheres
correspond to linked vertices.
This geometrical description makes it clear that sixteen of a given
lattice site's neighbors lie on the corners of a perfect
four-dimensional hypercube, while the other eight neighbors correspond
to the eight faces of that hypercube\footnote{Recall that a
$D$-dimensional hypercube has $2D$ faces.}. Presumably, this is how the
face-centered hypercubic lattice derives its name. Note also that there
are three distinct ways in which this partition of the lattice vectors
into a group of sixteen and a group of eight may be carried out. These
three ways correspond to letting each of the three subgroups in the list
(\ref{vectors}) be the group of eight.
To see this explicitly, note that the proper orthogonal matrix
\begin{equation}
\Lambda = \frac{1}{\sqrt{2}} \left(
\begin{array}{cccc}
+1&+1&0&0 \\ +1&-1&0&0 \\ 0&0&+1&+1 \\ 0&0&+1&-1
\end{array} \right),
\label{omat}
\end{equation}
when applied to each of the twenty-four lattice vectors in
equation~(\ref{vectors}), yields~\footnote{For brevity, we have omitted the
normalization factor $1/\sqrt{2}$ in front of each of these vectors.}
{\footnotesize
\begin{equation}
\begin{array}{ccc}
{A}& {B}& {C} \\
\begin{array}{rrrrrr}
(&+1,&+1,&+1,&-1&)\\
(&+1,&+1,&-1,&+1&)\\
(&-1,&-1,&+1,&-1&)\\
(&-1,&-1,&-1,&+1&)\\
(&+1,&-1,&+1,&+1&)\\
(&+1,&-1,&-1,&-1&)\\
(&-1,&+1,&+1,&+1&)\\
(&-1,&+1,&-1,&-1&)
\end{array}
&
\begin{array}{rrrrrr}
(&+1,&-1,&+1,&-1&)\\
(&+1,&-1,&-1,&+1&)\\
(&-1,&+1,&+1,&-1&)\\
(&-1,&+1,&-1,&+1&)\\
(&+1,&+1,&+1,&+1&)\\
(&-1,&-1,&+1,&+1&)\\
(&+1,&+1,&-1,&-1&)\\
(&-1,&-1,&-1,&-1&)
\end{array}
&
\begin{array}{rrrrrr}
(& 0,& 0,& +2,& 0&)\\
(& 0,& 0,& 0,& +2&)\\
(& 0,& 0,& 0,&-2&)\\
(& 0,& 0,&-2,& 0&)\\
(& +2,& 0,& 0,& 0&)\\
(& 0,& +2,& 0,& 0&)\\
(& 0,&-2,& 0,& 0&)\\
(&-2,& 0,& 0,& 0&)
\end{array}
\end{array}.
\label{vectors2}
\end{equation}
}
In this representation, it is manifest that the lattice vectors of
subgroups $A$ and $B$ lie at the corners of a perfect four-dimensional
hypercube, and that those of subgroup $C$ lie on its faces. In what
follows, we therefore refer to this representation as the {\it explicit
hypercubic frame}. This frame will be useful for describing the
isometries of the FCHC lattice in Section~\ref{ssiso}.
Finally, we note that isotropic three-dimensional hydrodynamics can be
obtained from three-dimensional projections of this four-dimensional
lattice. Due to the staggered nature of the FCHC lattice, this requires
only two lattice spacings in the fourth dimension. Alternatively (and
equivalently), we can simply project out the fourth coordinate of the
lattice vectors in equation~(\ref{vectors}). The resulting set of
three-vectors are the lattice vectors of an irregular lattice in three
dimensions. The neighbors of a point $(0,0,0)$ on this lattice can be
described in Cartesian coordinates as the six on-axis neighbors at
distance one ($(\pm 1, 0, 0)$, $(0, \pm 1, 0)$, and $(0, 0, \pm 1)$),
each counted with multiplicity two (double bonds), and the twelve
neighbors at distance $\sqrt{2}$ ($(\pm 1, \pm 1, 0)$ and permutations).
The four-tensor constructed by these lattice vectors is perfectly
isotropic.
\subsection{Microdynamics}
The two basic steps of the LGA are 1) propagation of the particles, and 2)
collisions. The particles reside on the lattice sites only, and there
can be at most one particle per direction at any given site and time.
The hydrodynamic behavior of the model depends on the fact that the
collision step conserves mass and momentum (conservation of these
quantitites holds trivially in the propagation step).
In order to introduce flow with solid walls present one must introduce
new collisions which prevent particles from moving across the
boundaries. These might be either of the bounce-back type which send
particles back into the direction from which they came, or of the mirror
reflection type where only one component of the particle's momentum is
changed. When particle velocities are averaged the effect of the
bounce-back collision is a hydrodynamic no-slip boundary
condition~\cite{cornubert91,ginzbourg94}.
In order to introduce a body-force, like gravity, an additional
collision step is needed that puts momentum into the system. This can
be done in several ways, one of which is to flip particle velocities at
a few randomly chosen sites into the direction of the forcing.
The state at a single site (at position $\mbox{{\bf x}}$ at time $t$) is
given by the 24 occupation numbers $n_i (\mbox{{\bf x}} , t)\in\{
0,1\}$, which are simply the particle number in direction $i$.
The time development of the $n_i$'s is given by the microdynamical
equation
\begin{equation}
n_i(\mbox{{\bf x}} + \mbox{{\bf c}}_i , t+1) = n_i(\mbox{{\bf x}}, t)
+\Omega_i (\{ {\bf n}(\mbox{{\bf x}} ,t)\} ).
\label{Boltzmann}
\end{equation}
The term $\Omega_i (\{ {\bf n}(\mbox{{\bf x}} ,t)\} )$, where ${\bf n} =
(n_1, n_2, \ldots , n_{24})$, is the change in $n_i$ due to collisions,
$\mbox{\bf c}_i$ are the velocity vectors connecting neighboring lattice
sites, $t$ is the time, and the time increment corresponding to a
combined propagation and collision step is unity.
The full detailed state of the lattice gas is given by the set of all
$n_i$'s on all sites. In simulations on a computer, all this information
is packed into 24 bits of information at each site, and is stored and
updated. However, the quantities of physical interest are the (space
and/or time and/or ensemble) averaged mass and momentum densities. On
the CAM-8 this averaging can be performed by the application of look-up
tables and read to the front end without loss of speed.
\subsection{The 16-bit collision operator and random isometries}
\label{ssiso}
In a three-dimensional lattice gas the number of possible collision
rules, represented by the collision operator $ \Omega$, is very large.
If one 24-bit output configuration for each of the $2^{24} = 16$M input
configurations were stored in a table, the table would have a size of at
least 48 Mbytes. This is bigger than what the local SRAM memory in each
node of the CAM-8 permits, and it is also bigger than the local memory of some
massively parallel machines, like the Connection Machine CM2. The
restriction for the CAM-8 is the 16-bit limit of the lookup tables. In
order to adapt to this restriction each 24-bit collision is split into
two 16-bit collisions. The structure of this factorization is shown in
Figure~\ref{fig2}. The particle velocities are split into the three
groups of eight introduced in equation~(\ref{vectors}). The collisions,
which are given by a single table, involve 16 particles at a time and
act first on the first 16 particles (subgroups $A$ and $B$) and then on
the last 16 (subgroups $B$ and $C$). The overall mass and momentum of
the 16 particles is conserved at each step.
Such a division of the collision rule
into three fixed sets of spatial directions can introduce
anisotropies. In order to prevent this effect, the particles are first
subjected to a random precursor transformation---an isometry $p$---before
the collision $C$; after the collision they are then subjected to
the inverse isometry $p^{-1}$. The isometries $p$ form a group $G$
defined as the set of transformations that map the set of lattice
vectors $\{\mbox{\boldmath c}_i \}$ to itself, preserving the vectors'
lengths and the angles between them. The total collision step can thus
be represented by the operator $p^{-1} C p$, where $C$ is deterministic
and is given by the 16-bit table, and $p$ is randomly chosen from the
isometry group $G$ at every timestep. Note that $p$ can be the same for
all sites on the lattice.
It can be shown that there are altogether 1152 elements in
$G$~\cite{henon87b}, and each element is a map acting on all the 24
bits. It is therefore crucial that the isometries can be factorized
into factors that, like $C$, decompose into operations on subsets of the
24 bits.
This factorization takes a convenient form when specified by the action
of the isometries on the coordinates of the velocity vectors in the
explicit hypercubic frame defined by equation (\ref{vectors}).
In this frame the
isometries take the form of a product of simple coordinate inversions and
permutations, known as $R$ and $P$ isometries, respectively,
combined with the forward or reverse cyclic interchange of
the subgroups $A$, $B$ and $C$~\cite{henon87a}. Specifically, the
isometries corresponding to the forward ($(A,B,C)\rightarrow (C,A,B)$)
and reverse ($(A,B,C)\rightarrow (B,C,A)$) cyclic interchange of the
three subgroups are denoted by $S_1$ and $S_2$, respectively. No tables
are needed for these operations. The
$R$ and $P$ isometries are also particularly simple because they do
not map vectors between $A + B$ and $C$. Thus they can be encoded as
16-bit and 8-bit tables acting on $A + B$ and $C$, respectively.
As a concrete example, consider the axis-inversion isometry $R_4$,
implemented by negating the fourth coordinate in the explicit hypercubic
frame. As before, we enumerate the lattice vectors in this frame by
$1_A,\ldots, 8_A, 1_B,\ldots, 8_B, 1_C,\ldots, 8_C$, where, for example,
$n_A$ denotes the $n$th entry of column $A$ in equation~(\ref{vectors2}). It
is evident that the application of $R_4$ amounts to swapping the lattice
vectors $1_A$ and $5_B$, $2_A$ and $7_B$, $3_A$ and $6_B$, $4_A$ and
$8_B$, $5_A$ and $1_B$, $6_A$ and $2_B$, $7_A$ and $3_B$, $8_A$ and
$4_B$, and $2_C$ and $3_C$. The orthogonal matrix, equation~(\ref{omat}),
then maps these swaps back to the corresponding swaps in the original
frame, equation~(\ref{vectors}), even though it is evident that $R_4$ is no
longer a simple axis inversion in that frame. So, for example, ordering
the lattice vectors and their corresponding bits according to equation
(\ref{vectors}), we see that the state
\begin{equation}
101010101010101010101010
\end{equation}
maps to the state
\begin{equation}
110010101010110011001010
\end{equation}
under the isometry $R_4$.
To apply a general isometry $p$ to the lattice,
we exploit the fact that $p$ can be written
in the product form
\begin{equation}
\left( \begin{array}{c} I \\ S_1 \\ S_2 \end{array} \right)
\left( \begin{array}{c}I\\ R_1 \end{array} \right)
\left( \begin{array}{c}I \\ R_2\end{array} \right)
\left( \begin{array}{c} I \\ R_3\end{array} \right)
\left( \begin{array}{c}I \\ R_4 \end{array} \right)
\left( \begin{array}{c}I\\P_{12}\end{array} \right)
\left( \begin{array}{c}I\\P_{13}\\P_{23} \end{array} \right)
\left( \begin{array}{c} I\\P_{14}\\P_{24} \\P_{34}\end{array} \right),
\end{equation}
where the indices of the $R$ and $P$ isometries refer to the axis to be
inverted or the axes to be permuted, respectively. The elementary $R$
and $P$ isometries are their own inverses whereas $S_1$ is the inverse
of $S_2$.
To choose a random isometry at each time step,
eight random numbers are
computed and used to choose one isometry from each column above.
The resulting combined isometry is then performed by the successive
application of tables corresponding to each of the elementary
isometries. The inverse isometry is obtained by the application of the
tables in the reverse order.
\subsection{Memory usage and implementation of the algorithm}
The memory in the CAM-8 is arranged so that there are four 16-bit
subcells at each lattice site. The initial organization of the memory
is to occupy the first eight bits of the first three subcells with the
three groups, $A$, $B$ and $C$, of lattice vectors, respectively. The
subcells are given the same labels, $A$, $B$ and $C$, as the velocity
groups they contain. The extra bits are used as logical flags and swap
space which can minimize the number of steps required to implement the
collision algorithm, trading abundant memory for computational speed.
In addition, there is a subcell $D$ which contains flags that indicate
whether or not a lattice site is solid, and whether or not forcing is to
be performed (if possible) at that site. Thus bits 0 through 7 of the
subcell $A$ indicate the presence or absence of particles moving in
lattice directions $A$ , while bits 0-7 of subcell $B$ indicate
similarly for lattice directions $B$, and bits 0-7 of subcell $C$
indicate similarly for lattice directions $C$.
The collision algorithm proceeds in five steps:
\begin{enumerate}
\item
Apply forcing and reorganize the memory to prepare subsequent steps.
\item
Apply randomly chosen $R$, $P$, and $S$ isometries.
\item
Apply collision table.
\item
Apply inverses of isometries applied in step 2 in reverse order.
\item
Reflect lattice vectors at solid sites and restore initial organization
of memory.
\end{enumerate}
Forcing is applied by adding momentum to a particle if possible and if
the flag at the site permits. The bit plane of flags (or planes, if
forcing in more than one direction is done simultaneously) is randomly
shifted at each time step so that forcing is on average uniform over the
space. Because the same lookup tables are necessarily applied at every
site, the amount of forcing is set in the initial conditions by setting
the number of positively set flags. The amount of forcing applied
cannot be precisely tuned, since not all particle configurations can be
changed to increase the momentum in a given direction. For this reason the
applied force must be measured along with the other observables.
At the solid sites all momenta are reversed. This is an operation that
can be performed on each of the three groups of bits separately, since
each group contains four pairs of oppositely oriented momenta. The
momentum reversals are handled in the following manner. The flag that
indicates whether or not a site is solid is copied into each of the
first three subcells. If the site is solid, the momentum information in
each subcell (in bits 0-7) is moved to the higher eight bits (8-15) and
the lower bits are set to zero. The subsequent steps in the collision
algorithm are thus applied to the eight bits of zero instead of the
momentum information, and the net result is no change. At the end of
the collision step, the presence of zeros in the first eight bits of
each subcell indicate that the site is solid, and that the momentum
information in the high eight bits should be reflected through the
origin and repositioned in the lower eight bits of each subcell. This
algorithm is robust in the case where the site is not solid but no
particles are present, since, in this case, the reflection of the bits
will cause no change.
When forcing is being applied in only a single direction, the collision
algorithm requires 26 applications of lookup tables when an $S$ isometry
is applied, and 14 applications when the $S$ isometry chosen is the
identity, which happens 33\% of the time. Thus the average number of
lookup table applications required per time step is 22. The performance
of the present algorithm on the 8 node machine (1 node handles 4M
16-bit sites ) is 7M site updates per second. It benefits greatly from
the fact that the CAM-8 can download one table while being busy using
another, without loss of speed.
\section{Optimization of the viscosity}
Typically, lattice-gas collision rules are chosen to maximize
a ``Reynolds coefficient'' and thus the Reynolds number of a simulation.
For the application we envision---namely, slow flow through disordered
porous media---accuracy and efficiency of the simulation improve
as the kinematic viscosity is reduced \cite{ferreol94}.
In this section we show how we choose our 16-bit collision tables to
obtain a low viscosity.
By employing the Boltzmann approximation, which assumes that the
particles entering a collision are
uncorrelated, it is possible to derive a closed expression for the
kinematic viscosity $\nu$ in terms of the collision rules and the
average density $d \leq 1$ per direction $i$.
The formula, due to H\'{e}non~\cite{henon87b}, has
the form
\begin{equation}
\nu = \frac{1+Q}{6(1-Q)} \; ,
\label{viscosity}
\end{equation}
where
\begin{equation}
Q = \frac{1}{18} \sum_{ss'} \; A(s \rightarrow s') \; d^{m-1}
(1-d)^{23 - m} \; Y_{\alpha \beta} (s) Y_{\alpha \beta} (s')
\label{Q}
\end{equation}
and $A(s \rightarrow s')$ is the probability that the state $s$
goes into $s'$.
The sum runs over all possible input and output states $s$ and $s'$,
where $s = \{ s_i \}$, $m = \sum_i s_i$ is the mass at a site,
${\alpha }$ and ${ \beta}$ are cartesian indices with summation
over repeated indices implied, and
\begin{equation}
Y_{\alpha \beta} = \sum_{i=1}^{24} s_i \left(
c_{i\alpha}c_{i\beta} - \frac{1}{2} \delta_{\alpha \beta} \right)
\label{Y} \end{equation}
where $\delta_{\alpha \beta}$ is the Kronecker delta.
The term $Y_{\alpha \beta} (s) Y_{\alpha \beta} (s')$
is easily evaluated as
\begin{eqnarray}
Y_{\alpha \beta} (s) Y_{\alpha \beta} (s') &=&
\sum_{ij} s_i s'_j \left(
c_{i\alpha}c_{i\beta} -
\frac{1}{2} \delta_{\alpha \beta} \right)
\left(
c_{j\alpha}c_{j\beta} -
\frac{1}{2} \delta_{\alpha \beta} \right) \nonumber \\
&=& \sum_{ij} s_i s'_j \left( ( \mbox{\bf c}_i
\cdot \mbox{\bf c}_j )^2 -1 \right) \nonumber \\
&=& \sum_{ij} s_i s'_j {\cal A}_{ij}
\end{eqnarray}
where
\begin{equation}
{\cal A}_{ij}=( \mbox{\bf c}_i \cdot \mbox{\bf c}_j )^2 -1.
\end{equation}
The angle between lattice vectors $i$ and $j$ different subgroups is
either $\pi /3$ or $2 \pi /3$, and $|{\bf c}_i|=\sqrt{2}$, so $\mbox{\bf
c}_i\cdot \mbox{\bf c}_j = \pm 1$, so ${\cal A}_{ij}=0$ for $i$ and $j$
in different subgroups. It follows that the $24\times 24$ matrix ${\cal
A}$ is block diagonal,
\begin{equation}
{\cal A} =
\left(
\begin{array}{ccc} A & 0 & 0 \\ 0 & A & 0 \\
0 & 0 & A \end{array} \right),
\end{equation}
where the $0$'s denote $8\times 8$ null matrices, and the $8\times 8$
matrix of components $A$ for lattice vectors in the same subgroup is
given by
\begin{equation}
A = \left(\begin{array}{cccccccc}
+3 & -1 & -1 & +3 & -1 & -1 & -1 & -1 \\
-1 & +3 & +3 & -1 & -1 & -1 & -1 & -1 \\
-1 & +3 & +3 & -1 & -1 & -1 & -1 & -1 \\
+3 & -1 & -1 & +3 & -1 & -1 & -1 & -1 \\
-1 & -1 & -1 & -1 & +3 & -1 & -1 & +3 \\
-1 & -1 & -1 & -1 & -1 & +3 & +3 & -1 \\
-1 & -1 & -1 & -1 & -1 & +3 & +3 & -1 \\
-1 & -1 & -1 & -1 & +3 & -1 & -1 & +3
\end{array} \right) \; .
\end{equation}
Thus, heuristically speaking, collision events that take a particle from
one lattice vector to another---neither the original one nor its
negation---in the same subgroup of eight are the most preferred in
terms of minimizing $Q$ (score $=-1$). Collisions that take a particle
to a different subgroup are next (score $=0$). Finally, collisions that
take particles to themselves or their negation are least preferred
(score $=+3$). The collision set that minimizes this scoring
is optimal.
Due to the symmetric ordering of the velocity vectors, $Q$ splits into a
sum of three terms
\begin{equation}
Q = Q_A + Q_B + Q_C \; ,
\end{equation}
corresponding to the group A, B and C respectively. We consider only
the first two terms, which corresponds to a 16-bit table, and choose the
collisions that minimize $(Q_A + Q_B)$. Note that $\nu = \nu (Q)$ is an
increasing function of $Q$, so that minimizing $Q$ is equivalent to
minimizing $\nu$. Minimizing $(Q_A + Q_B)$, however, is not equivalent
to minimizing $Q_A + Q_B + Q_C$. The former quantity results from a
single application of the 16-bit table, whereas the latter quantity
results from applying the 16-bit table twice. Only $Q_A$ remains the
same after the second application.
The reason for doing only the restricted optimization is twofold.
First, by restricting the optimization to the 16-bit space with $2^{16}$
velocities, an exhaustive search through all possible output states
corresponding to every input state is feasible and requires only
about one hour of computation time on a Sun Sparcstation 2.
By comparison,
an exhaustive search through all output states corresponding to the
$2^{24}$ possible input states of the complete collision table would
require increase computation time by a factor of $(2^8)^2 \approx 65000$.
Second, we argue that the process of minimizing $Q_A + Q_B$ approximates
the process of minimizing $Q$. From the above equations it is seen that
minimizing $Q$ means finding output states that select the matrix
elements $-1$ in $A$. This corresponds to maximizing the number of
perpendicular directions between the velocities of the input and output
states within each subgroup of eight. Physically it corresponds to
minimizing the momentum flux and thereby the viscosity. For this
purpose it is advantageous to let a particle scatter into one of the
directions of its own subgroup which give a contribution $-1$ to the
matrix product, rather than into another subgroup, which would give no
contribution to the matrix product. The 16-bit collisions certainly
allow for the particles to go into their own subgroups.
However, the selection of output states is restricted, not only by mass
and momentum conservation, but by the limited number of velocities in
the 16-bit states as well. Also, since the 16-bit table is applied
twice to give the full collision, the eight particles that are involved
in both collisions may scatter back into a disadvantageous direction.
The probability that such backscattering occurs, however,
appears to be small enough to give a
relatively low viscosity. However, a refined scheme that takes into
account that the 16-bit collisions are indeed applied twice would be expected
to decrease the viscosity further.
Our optimized viscosity, determined from
the Boltzmann approximation of equations~(\ref{viscosity}-\ref{Y}),
is given as a function of the reduced
density $d$ in Figure~\ref{fig3}. Here it is also compared to measurements
of the viscosity by simulation of a relaxing shear wave on the CAM-8.
Specifically, simulations were initialized with the velocity profile
\begin{equation}
u_z (x,t=0) = U \sin (k x) ,
\end{equation}
where $u_z$ is the $z$-component of the velocity, $k = 2\pi /L$ , $L =
64$ is the linear size of the system, $U=0.07$ is the maximum flow
velocity, $x$ is a Cartesian coordinate, and $t$ is the time. According
to the Stokes equation, the velocity should evolve as
\begin{equation}
u_z (x,t) = U \sin ( k x) \exp (-k^2 \nu t) \; .
\end{equation}
The viscosity can thus be deduced from the integral $\int_0^L \; dx
|u_z| $, which decays exponentially with time.
The minimum viscosity is obtained at a density $d=0.5$ and has the value
$\nu = 0.095$. For comparison, the value obtained from a purely random
table is $\nu = 0.1667$. When Q is minimized by choosing the collisions
by an exhaustive search in the (severely) restricted space of 24-bit
{\em isometries}~\cite{henon87b}, the resulting viscosity is within 20\%
of the value resulting from the random table \footnote{In this case the
problem is slightly different because the minimization is of a Reynolds
coefficient rather than of the viscosity.}.
\section{Test simulations}
To check the behavior of the model when the fluid is forced in the
presence of solid walls we performed simulations of flow through a pipe
and channel as well as flow through a simple cubic array of solid
spheres. The latter geometry creates the only truly three-dimensional flow
because the flow passes through constrictions.
The simulations of pipe flow were carried
out to obtain an estimate of the lower limit for the spatial size of
obstructions and constrictions.
\begin{table}
\begin{center}
\begin{tabular}{|| l |l| l | l |l ||} \hline
Geometry & $\phi $ & $R$ & $(\kappa -
\kappa_{theory})/\kappa_{theory}$
& $R_{eff}$\\
\hline
\hline
Channel & 1.0 & 1.5 & +0.04 $\pm $ 0.03 & 1.6 \\
& 1.0 & 2.5 & $-$0.04 $\pm $ 0.02 & 2.4\\
& 1.0 & 3.5 & $-$0.03 $\pm $ 0.05 & 3.4 \\
& 1.0 & 7.5 & +0.03 $\pm $ 0.05 & 7.7\\
& 1.0 & 15.5 & +0.04 $\pm $ 0.05 * & 16.1\\
\hline
Pipe & 1.0 & 1.5 & +0.09 $\pm $ 0.07 & 1.6 \\
& 1.0 & 2.5 & +0.01 $\pm $ 0.04 & 2.5 \\
& 1.0 & 8.5 & $-$0.03 $\pm $ 0.02 & 8.4 \\
& 1.0 & 15.5 & $-$0.07 $\pm $ 0.05 *& 14.9 \\
\hline
Array of spheres
& 0.713 & 4.5 & $-$0.07 $\pm $ 0.08 & 4.6 \\
& 0.907 & 4.5 & $-$0.11 $\pm $ 0.02 & 4.7 \\
& 0.719 & 6.5 & $+$0.05 $\pm $ 0.06 & 6.4 \\
& 0.882 & 19.5 & $-$0.12 $\pm $ 0.04 * & 20.4\\
& 0.882 & 19.5 & $-$0.04 $\pm $ 0.02 & 19.8\\
& 0.735 & 25.5 & $-$0.04 $\pm $ 0.07 * & 25.8\\
\hline
\end{tabular}
\end{center}
\caption{
Comparison of calculations of the permeability $\kappa$ to theoretical
predictions $\kappa_{theory}$ for channels, pipes, and a periodic cubic
array of spheres.
$R$ denotes the radius of the pipes and spheres and
the half-width of the channels, $R_{eff}$ denotes
the effective radius (half-width) corresponding to the theoretical
expressions for $\kappa_{theory}$; $\phi$ is the void fraction.
The asterisk (*) shows the simulations that were performed on CAM-8.
(The remaining small-scale simulations were carried out on a workstation.)
}
\end{table}
Table 1 summarizes the results of the simulations of flow through
channels of half-width $R$, pipes with radius $R$, and periodic cubic arrays
of spheres with radius $R$.
The pipes were constructed with periodic boundary conditions in the flow
direction, whereas the channels had periodic boundaries in two directions
and a flat-wall boundary perpendicular to the third direction.
In the table, the permeability $\kappa$ of each particular geometry
is compared to the theoretical prediction of the permeability,
$\kappa_{theory}$.
The permeability is defined as
\begin{equation}
\kappa = \frac{J \nu \phi}{F}
\end{equation}
where $J$ is the total massflux, $F$ is the total force applied to the fluid,
and $\phi $ the fraction of void space, i.e., the porosity. Note that
$\phi = 1$ for the pipe and the channel.
In the simulations the permeabilities were deduced from the measurement
of the average flow rate and the average forcing of the fluid, and the
uncertainties are estimated from the noise in these data.
The theoretical expression for the permeability of the pipe,
\begin{equation}
\kappa_{theory} = \frac{R^2}{ 8} \; ,
\end{equation}
is easily obtained from the Stokes equation, as is
the permeability of a channel
\begin{equation}
\kappa_{theory} = \frac{R^2}{3} \; .
\end{equation}
The theoretical expression for the permeability of the simple cubic
array of spheres is obtained (in a slightly different form)
by Hasimoto~\cite{hasimoto59} and Sangani et al.~\cite{sangani82}, and
has the form
\begin{equation}
\kappa_{theory} = \frac{2}{9 K(\phi )}
\left( \frac{\phi}{1-\phi} \right) R^2
\end{equation}
where $K(\phi )$ is given in~\cite{sangani82}. In this case the radii
$R=19.5$ and $26.5$ correspond to the porosities $\phi = 0.882$ and
$\phi = 0.735$ respectively.
The drag $F$ on a single sphere in the
array is
\begin{equation}
F = K(\phi ) \; 6 \pi \rho \nu R \; ,
\label{drag}
\end{equation}
where $\rho $ is the mass density and $\nu$ is the viscosity of the
fluid. It is seen that $K(\phi ) \geq 1$ is the correction factor to
Stokes law. When the spheres are infinitely far apart $\phi =1 $ and $K=
1$.
The non-integer values of $R$ are discussed below.
They are related to the
definitions of the (discretized) geometries in the following way:
For the pipe a site will be a void site if its distance
$r$ from the symmetry axis satisfies the inequality
\begin{equation}
r < R+0.5 \; .
\end{equation}
The remaining sites on the lattice will be solid sites.
The solid sites of the sphere are given by the same inequality
if $r$ is taken to be the distance from the center of the sphere.
The effective radius $R_{eff}$ given in the above table is the radius
(half-width) that gives the observed permeability when inserted in the
theoretical expressions for $\kappa_{theory}$. Especially in the case
of the spheres the permeability varies strongly with
$R$, and the difference between $R$ and $R_{eff}$ corresponding
to the discrepancies between $\kappa$ and $\kappa_{theory}$
is small.
The discrepancies between the theoretical and simulated
results have three principal causes. These are
\begin{itemize}
\item Uncertainty in the effective position of the wall;
\item Discretization errors; and
\item Effects of compressibility.
\end{itemize}
In the case of the channel and pipe simulations, only the first and
second problems come into play, whereas all three cause errors in the
three-dimensional case of the spheres. The discretization errors are
due both to the approximations made when smoothly curved surfaces are
represented on a lattice, and to non-hydrodynamic effects at sufficiently
small scales.
For the hydrodynamic behavior of the lattice gas to be described
by the equations of incompressible hydrodynamics, several requirements
must be fullfilled.
First, the geometry must be such that
there is a scale separation between
the lattice constant and the scale over which the flow varies.
The ratio between these scales is denoted by the small number
$\epsilon$.
Second, the flow must be forced sufficiently weakly so that
the flow velocity $u$ does not exceed $\epsilon$.
Finally, density
variations must remain small. More precisely, the variations in $\rho$
must be of order $\epsilon^2$ or smaller.
In geometries where the overall permeability is small, a relatively
large force will be necessary to drive the flow
and spatial variations in the permeability (caused, for instance, by
constrictions) may result in large
density variations, even when $u$ is small.
This is the case in particular for the CAM8-simulation with the
$R=19.5$ sphere
where the forcing applied produced a maximum flow velocity $u=0.10$.
This simulation was repeated on a workstation with approximately half the
forcing, and the resulting discrepancy between theory and simulations
was correspondingly decreased.
When the requirements on the scale separation
are obeyed and there is no
forcing, the lattice
gas is described by the following equation~\cite{doolen91}
\begin{equation}
\frac{\partial {\bf u}}{\partial t} +
g(\rho ){\bf u}\cdot\nabla {\bf u} =
-\frac{\nabla P}{\rho} +
\nu \nabla^2 {\bf u}
\label{NS}
\end{equation}
where ${\bf u}$ and $\rho$ are the average flow velocity and the
density, respectively,
\begin{equation}
P = (1/2) \rho \; (1- g(\rho ) u^2 )
\label{P}
\end{equation} is the pressure ~\cite{doolen91}
and $\nu$ is the kinematic viscosity.
The $g$-factor is given as
\begin{equation}
g(\rho ) = \frac{4}{3} \left(\frac{12- \rho}{24 - \rho}\right)
\label{g} \; .
\end{equation}
The velocity dependence in the pressure results from the discreteness of
the velocities and the exlusion principle. For the purpose of finite
Reynolds number simulations, the $g$-factor may be removed by rescaling
the velocity as $u' = g(\rho ) \; u$ \cite{doolen91}. In the
simulations the average density $\overline{\rho } = 12$ was chosen so
that $g = 0$ and the flow is described by the Stokes equation in the
limit where there are no density variations.
As an example of how to quantify the density variations in steady state
flow, we give an estimate of the density variations that will result in
the case of the array of spheres. By the above equation of state
(\ref{P}) we have that the density difference across the sphere is given
as
\begin{equation}
\Delta \rho \approx 2 \Delta P \; .
\end{equation}
But this pressure difference
can be estimated as the
drag divided by some area which is characteristic
for the sphere, say $\pi R^2$.
Hence, by Eq.~(\ref{drag})
\begin{equation}
\Delta \rho \approx K(\phi) \; 12 \nu \; U/R \; .
\end{equation}
Since the drag coefficient $K( \phi )$ increases sharply
with decreasing porosity $\phi $, $\Delta \rho $ may become
significant even when $U/R \sim \epsilon^2 $ is small.
Care must be taken in order for this effect of compressibility
in the lattice gas to be negligible. In the present case when $\rho$
is chosen so that $g(\rho ) =0$, and the Stokes equations
are simulated, deviations in $\rho$ from its average value
will cause a finite value of $g(\rho )$.
This means that the $g$-dependent term of Eq.~(\ref{NS})
must be taken into account and we are no longer in the regime of
linear hydrodynamics.
The effective wall positions have been studied theoretically by several
authors~\cite{cornubert91,ginzbourg94}. It can be shown (theoretically)
that in the case of Couette flow along a flat wall oriented in a
direction parallel to the lattice directions, the position of vanishing
velocity is not on the wall sites, but rather half a lattice unit
measured from the wall sites into the fluid region. In the case when
the velocity field has a non-vanishing second derivative, there is a
correction to the effective wall position of relative order
$\epsilon$. This correction is
significant only when the constrictions in the flow are very small, and
it depends on the viscosity.
In the channel flow simulations referred above, we fitted the velocity
profiles with a parabola, as predicted by the Stokes equation. Within
the noise of the measurements we found agreement between simulations and
a parabolic profile going through the position halfway between the first
wall site and the first fluid site. We also found good agreement when
the wall-normal was in an angle of $\pi /4$ to the closest lattice
directions. This result slightly generalizes the predicted result for
the wall positions~\cite{ginzbourg94}. In the table the values of
$R$ are obtained by assuming that the effective wall position is allways
halfway between the first wall site and the first fluid site. In the
case of curved geometries this approximation creates discrepancies
in addition to the discretization approximation.
The relatively good agreement between theory and simulations in the
case of the two smallest pipes is coincidential, since the
crossection of these pipes are squares rather than circles.
In simulations of the flow around the periodic
array of spheres (in the simulations, a single sphere and periodic
boundary conditions were used) the result is highly sensitive to the
exact position of the boundary, as can be seen from the
values of $R_{eff}$ given in the table, and the small discrepancy between
simulation and theory can be accounted for by a correspondingly small
shift of the effective boundary.
In general, the lower limit on the size of obstructions
and constrictions depends on
the particular application as well as the required precision. In the
case of flow in porous media, the coarse grained characteristics depend
mainly on the flow in the widest channels, and the averaged behavior
depends only weakly on the flow in the narrower passages through the
medium. In this case the small scale hydrodynamics may not matter much.
But the permeability typically depends strongly on the effective
positions of the walls, and care must be taken to ensure that these are
correctly represented or that the channels are sufficiently wide.
\section{Conclusion}
We have designed an algorithm for the implementation of a
three-dimensional (FCHC) lattice gas model on the CAM-8. It has been
shown how a proper geometric grouping of the velocity vectors on the
lattice makes it possible to decompose both the collisions and the
random isometries acting on the full 24-bit states into operations
involving only 16 or 8 bits at a time. The corresponding 16-bit
collision table has been optimized to obtain a minimum viscosity, and
tested against analytic results from the Boltzmann theory. The
architecture of the CAM-8 has been described, and the performance of an
eight node machine has been shown to be comparable to that of existing
supercomputers. Some practical limitations of the model have been
established and discussed, and it has been shown that the model behaves
according to the hydrodynamic predictions for various permeable media.
We have thus demonstrated that this implementation of the FCHC lattice
gas represents a working tool for large-scale simulation of flows in
simple and complex geometries.
\section{Acknowledgements}
We thank John Olson for helpful discussions. One of us (BMB) would like
to acknowledge helpful conversations with David Levermore and Washington
Taylor. This work was supported in part by the sponsors of the
MIT Porous Flow Project
and NSF Grant 9218819-EAR. E. G. Flekk\o y acknowledges support by
NFR, the Norwegian Research Council for Science and the Humanities.
The CAM-8 project is supported by ARPA under grant N00014-94-1-0662.
|
\section{Introduction}
\label{sec:intro}
Atoms of a hard-sphere lattice gas occupy sites of a lattice and
interact with infinite repulsion of nearest neighbour pairs.
The grand partition function is
\begin{equation}
{\mit \Xi_{V}(z)}
=
\sum_{N} z^{N}\,Z_{V}(N),
\label{eqn:gpf}
\end{equation}
where $z$ is an activity and
$Z_{V}(N)$ is the number of configurations in which
there are $N$ atoms in the lattice of $V$ sites.
At $z=+\infty$ a ground state configuration is that the atoms occupy
all the sites of one sublattice and the other is vacant.
There is no atom at $z=0$.
A continuous phase transition occurs at a critical activity.
There are many studies of the system by various methods:
series expansions
\cite{Domb58,Temperley59,Burley60,GauntFisher65,Gaunt67},
finite-size scaling and transfer matrix
\cite{KamieniarzBlote93},
Bethe and ring approximations
\cite{Burley61b},
transfer matrix
\cite{Runnels65,RunnelsCombs66,ReeChesnut66},
corner transfer matrix and series expansions
\cite{Baxteretal80},
exact calculations
\cite{Baxter80},
and Monte Carlo simulations
\cite{Meirovitch83,Yamagata95}.
The critical activity is obtained on various lattice.
The square lattice:
3.80(2)
\cite{GauntFisher65},
3.80(4)
\cite{RunnelsCombs66},
3.7966(3)
\cite{ReeChesnut66},
3.7962(1)
\cite{Baxteretal80},
and
3.796255174(3)~\cite{BloteWu90}.
The triangular lattice:
11.12(11)
\cite{RunnelsCombs66},
11.05(15)
\cite{Gaunt67},
and
$\frac{1}{2}(11+5\sqrt{5}) = 11.090 \ldots$
\cite{Baxter80}.
The simple cubic lattice:
1.09(7)
\cite{Gaunt67}
and
1.0588(3)
\cite{Yamagata95}.
The body-centred cubic lattice:
0.77(5)
\cite{Gaunt67}
and
0.73(2)
\cite{Landau77,Racz80}.
The critical exponents on the square lattice are as follows.
The thermal exponent, $y_t$, is 1.000000(1),
the magnetic exponent, $y_h$, is 1.875000(1)
\cite{KamieniarzBlote93},
$\beta/\nu$ = 0.125(5),
and
$\gamma/\nu$ = 1.74(2)
\cite{Meirovitch83}
where $\beta$, $\gamma$, and $\nu$ are the critical exponents
for the order parameter, the staggered compressibility, and
the correlation length, respectively.
There are exact values on the triangular lattice
\cite{Baxter80,BaxterPearce82}:
$\alpha=1/3$ for the compressibility,
$\beta=1/9$,
$\mu=5/6$ for the interfacial tension, and
$\nu=5/6$.
The value, 0.61071(2)
\cite{KamieniarzBloteIsing93},
of the fourth-order cumulant
\cite{Binder81}
of the order parameter on the square lattice agrees with those
of the Ising model,
0.611(1)
\cite{Bruce85},
0.6103(7)
\cite{NicolaidesBruce88},
0.611(1),
and
0.610690(1)
\cite{KamieniarzBloteIsing93}.
These results support that the hard-sphere lattice gas on the square
lattice belongs to the universality class of the two-dimensional
Ising model and that on the triangular lattice does to one of the
three-state Potts model in two dimensions~\cite{Wu82}.
There is no result of critical exponents in three-dimensional
hard-sphere lattice gas.
We carry out Monte Carlo simulations of the hard-sphere lattice gas
on the simple cubic lattice.
Using the finite-size scaling,
we estimate the critical exponents.
In the next section we define physical quantities measured.
In section~\ref{sec:mcr} we describe procedure of analyses and
present Monte Carlo results.
A summary is given in section~\ref{sec:sum}.
\section{Monte Carlo simulations}
We use the Metropolis Monte Carlo technique
\cite{Binder79,BinderStauffer87}
to simulate the hard-sphere lattice gas~(\ref{eqn:gpf})
on the simple cubic lattice
of $V$ sites, where $V$ = $L \times L \times L$
($L$ = $2 \times n$, $n$ = 2, 3, $\ldots$ , 30),
under fully periodic boundary conditions.
According to Meirovitch
\cite{Meirovitch83},
we adopt the grand canonical ensemble.
The algorithm is described in the references
\cite{Meirovitch83,Yamagata95}.
We start each simulation with a ground state configuration
at the critical activity,
$z_{\rm c} = 1.0588$
\cite{Yamagata95}.
The pseudorandom numbers are generated by the Tausworthe method
\cite{ItoKanada88,ItoKanada90}.
We measure physical quantities
over $1.2 \times 10^{6}$ Monte Carlo steps per site (MCS/site)
for $L \ge 44$ and $10^{6}$ MCS/site for $L \le 42$
after discarding $5 \times 10^{4}$ MCS/site to attain equilibrium.
We have checked that simulations from the ground state configuration
and no atom one gave consistent results.
Each run is divided into $M$ blocks.
Let us the average of a physical quantity, $O$, in each block
$\langle O \rangle_{i}$; $i$ = $1, 2, \ldots , M$.
The expectation value is
\[
\overline{\langle O \rangle}
=
\frac{1}{M}\,\sum_{i=1}^{M} \langle O \rangle_{i}.
\]
The standard deviation is
\[
{\mit \Delta} \langle O \rangle
=
\left(
\overline{\langle O \rangle^{2}} - \overline{\langle O \rangle}^{2}
\right)^{1/2}/\sqrt{M-1}.
\]
We have taken $M=12$ for $L \ge 44$ and $M=10$ for $L \le 42$.
Let us define an order parameter by
\[
R
=
2\,(N_{\rm A}-N_{\rm B})/V
\]
where $N_{\rm A}$ ($N_{\rm B}$) is the number of the atoms
in the A (B)-sublattice and $N = N_{\rm A} + N_{\rm B}$.
We measure the staggered densities,
\begin{equation}
m^{\dagger}
=
\overline{
\langle |R| \rangle
}
\label{eqn:stagdens}
\end{equation}
and
\begin{equation}
m^{\dagger \prime}
=
\overline{
\langle R^{2} \rangle^{1/2}
},
\label{eqn:rmsdensi}
\end{equation}
the staggered compressibilities,
\begin{equation}
\chi^{\dagger}
=
V
\overline{
\left(
\langle R^{2} \rangle - \langle |R| \rangle^{2}
\right)
}
/ 4
\label{eqn:stagcomp}
\end{equation}
and
\begin{equation}
\chi^{\dagger \prime}
=
V
\overline{
\langle R^{2} \rangle
}
/ 4,
\label{eqn:rmscompr}
\end{equation}
and the fourth-order cumulant of $R$ \cite{Binder81},
\begin{equation}
U
=
1-\frac{1}{3}\,
\overline{
\langle R^{4} \rangle / \langle R^{2} \rangle^{2}
}.
\label{eqn:stagcumu}
\end{equation}
\section{Monte Carlo results}
\label{sec:mcr}
We estimate a critical exponent and an amplitude
by using the finite-size scaling
\cite{Fisher70,Barber83,Privman90}.
For a physical quantitiy, $O$,
we use the nonlinear chi-square fitting
\cite{Pressetal92}
with a function of $L$,
\[
O(L)=A\,L^{p},
\]
where $p$ and $A$ are fitting parameters.
$p=-\beta/\nu$ when $O$ = $m^{\dagger}$ or $m^{\dagger \prime}$.
$p=\gamma/\nu$ when $O$ = $\chi^{\dagger}$ or $\chi^{\dagger \prime}$.
We calculate the chi-square per degrees of freedom,
$\chi^{2}/{\rm DOF}$,
and the goodness of fit, $Q$,
\cite{Pressetal92}
for the data set of the sizes
$L = L_{\rm min}, L_{\rm min}+2, \ldots , L_{\rm max}-2$, and
$L_{\rm max}$.
The values of $L_{\rm min}$ and $L_{\rm max}$ are selected so that
the difference between $\chi^{2}/{\rm DOF}$ and 1 is the smallest.
We get the results as follows.
For $m^{\dagger}$ defined by (\ref{eqn:stagdens}),
$\beta/\nu$ = 0.28(1),
$A$ = 0.39(2),
$\chi^{2}/{\rm DOF} = 1.21$, and
$Q = 0.27$.
For $m^{\dagger \prime}$ defined by (\ref{eqn:rmsdensi}),
$\beta/\nu$ = 0.313(9),
$A$ = 0.48(2),
$\chi^{2}/{\rm DOF} = 1.22$, and
$Q = 0.26$.
For $\chi^{\dagger}$ defined by (\ref{eqn:stagcomp}),
$\gamma/\nu$ = 1.96(2),
$A$ = 0.040(3),
$\chi^{2}/{\rm DOF} = 0.98$, and
$Q = 0.47$.
For $\chi^{\dagger \prime}$ defined by (\ref{eqn:rmscompr}),
$\gamma/\nu$ = 2.37(2),
$A$ = 0.058(4),
$\chi^{2}/{\rm DOF} = 1.21$, and
$Q = 0.28$.
$L_{\rm min} = 36 $ and $L_{\rm max} = 60$
for $m^{\dagger}$, $m^{\dagger \prime}$, and $\chi^{\dagger \prime}$.
$L_{\rm min} = 30 $ and $L_{\rm max} = 54$
for $\chi^{\dagger}$.
The values of the critical exponents of $m^{\dagger \prime}$ and
$\chi^{\dagger \prime}$ satisfy the scaling relation,
$2 \beta + \gamma = d\,\nu$
where $d$ is the dimensions, within errors.
Thus we adopt them as the reliable estimates.
Figure~1 shows the size dependence of $m^{\dagger \prime}$
at $z = z_{\rm c}$.
The solid line indicates $0.48 L^{-0.313}$.
Figure~2 shows the size dependence of $\chi^{\dagger \prime}$
at $z = z_{\rm c}$.
The solid line indicates $0.058 L^{2.37}$.
Figure~3 shows the size dependence of $U$
defined by (\ref{eqn:stagcumu}) at $z = z_{\rm c}$.
We cannot extract a reliable estimate in $L = +\infty$
since we do not know the correction to scaling of this system.
It will have a value between 0.54 and 2/3.
\section{Summary}
\label{sec:sum}
We carry out the Monte Carlo simulations of the hard-sphere lattice
gas on the simple cubic lattice with nearest neighbour exclusion
under fully periodic boundary conditions.
Using the finite-size scaling,
we obtain the critical exponents,
$\beta/\nu$ = 0.313(9) and $\gamma/\nu$ = 2.37(2).
They satisfy the scaling relation, $2 \beta + \gamma = d \nu$.
The critical exponents of the three-dimensional Ising model
have been estimated,
$\beta/\nu$ = 0.518(7) and $\gamma/\nu$ = 1.9828(57)
\cite{FerrenbergLandau91}.
These values do not agree with
our results of the hard-sphere lattice gas.
The value of the fourth-order cumulant of the Ising model
on the simple cubic lattice under fully periodic boudary conditions
is as follows.
0.44(2)
\cite{Binder81},
0.4677(23)
\cite{LaiMon89},
and
0.4637(13)
\cite{BloteKamieniarz93}.
We see that these values are smaller than those of the hard-sphere
lattice gas as seen in figure~3.
While there is evidence that the hard-sphere lattice gas
on the square lattice
belongs to the Ising universality class in two dimensions
as described in section~\ref{sec:intro},
it does not seem that the hard-sphere lattice gas
on the simple cubic lattice falls into
the universality class of the three-dimensional Ising model.
\section*{Acknowledgements}
We have carried out the simulations on the HITAC S-3600 computer
under the Institute of Statistical Mathematics Cooperative Research
Program (94-ISM$\cdot$CRP-43 and 95-ISM$\cdot$CRP-37) and
on two personal computers with the 486DX2/66MHz CPU and
the Linux operating system
(SLS 1.03 + JE 0.9.3, Slackware 2.0.0 + JE 0.9.5).
This study was supported by a Grant-in-Aid for Scientific Research
from the Ministry of Education, Science and Culture, Japan.
\clearpage
|
\section{Introduction}
The {\v C}erenkov Correlated Timing (CCT) Detector is a new particle identification concept based on
precision timing measurements to determine the {\v C}erenkov angle of photons emitted
by particles passing through a transparent radiator~\cite{mats-cct}.
In order to investigate the detector concept, several prototypes have been tested
at the KEK-PS test beam line, as tabulated in Table 1.
Material tests have been done in parallel with measurements of the optical
characteristics~\cite{mats-rich}.
The beam test data are well reproduced by Monte Carlo simulation,
so that the feasibility of the CCT detector concept has been
demonstrated.
We describe the results mainly from a quartz bar prototype
and summarize the results from other prototypes for comparison.
\section{Experiment}
\subsection{Beam test setup}\label{subsec:setup}
The beam test setup at the KEK-PS T1 beam line is shown in Figure~\ref{fig:setup}.
CCT prototypes are placed on a rotating table which is mounted on a moving stage, so that
the beam position and the crossing angle can be adjusted independently.
The accuracies in positioning the prototype counter were 0.5 cm along the counter
axis and 1$^o$ in tilt angle, respectively.
The beam was defined by two sets of beam definition counters D1 and D2$\times $D3, which were
separated by about 5 m. Their areas were 2$\times $2 cm$^2$ and 1$\times $1 cm$^2$,
respectively, thus the divergence of the beam was expected to be about 2 mrad.
{\v C}erenkov light yields and timing characteristics were investigated as function of
beam position (Z) and tilt angle ($\theta$) of the counter, using a 1.5 GeV/c $\pi^-$ beam.
\subsection{CCT prototypes}\label{subsec:Zygo CCT pmt}
The tested CCT prototypes are tabulated in Table~\ref{tab:CCTpro}.
Details of the material characteristics are described in Ref.~\cite{mats-rich}.
The Acrylic bars had been annealed for 10 hours at 80$^o$C to release internal stress
after machining and polishing.
A support for the Zygo quartz bar CCT was carefully designed
to avoid a sizable loss of photons during total internal reflection.
As the edges were machined very sharply and fragile,
the mechanical contacts were made of Teflon and designed to
touch only the surfaces and not the edges with the contact areas minimized.
This point is very important to realize an attenuation length of several 10's of m,
whose reflectivity per bounce is expected to be on the order of 0.9995.
One end of the counter was viewed by a phototube and the opposite end
was blackened to avoid reflection.
\begin{table}\begin{center}\caption{CCT prototypes}\label{tab:CCTpro}
\vspace{0.4cm}
\begin{tabular}{|c|c|c|c|c|} \hline
Prot. & (1) & (2) & (3) & (4) \\ \hline
Test Run & Mar.95 & \multicolumn{3}{|c|}{ Nov.94 } \\ \hline
Prep. & KEK & U.I. & \multicolumn{2}{|c|}{ KEK } \\ \hline
Mater.& Quartz & Quartz & Acrylic & Acrylic \\
Prod. & Zygo & S. Finish & Mitsubishi & Kuraray \\ \hline
Index & 1.47 &1.47 &1.51 &1.51 \\ \hline
Thickness & 2 cm & 4 cm & 3 cm & 5 cm \\
Width & 4 cm & 4 cm & 4 cm & 6 cm \\
Length & 120 cm & 100 cm & 150 cm & 100 cm \\ \hline
Couplant & Viscasil & \multicolumn{3}{|c|}{ OKEN 6262A } \\ \hline
PMT & \multicolumn{4}{|c|}{ H2431 } \\ \hline
$N_o$(pe/cm) & 136 & 120 & 66 & 82 \\
$\lambda_{att}$(m) & 13.6 & 1.8 & 1.43 & 1.45 \\
$\sigma_{CCT}$(ps) & 80 & 80 & 105 & 95 \\
$\delta_{\Theta}$(mrad) & 15 & 15 & 20 & 20 \\ \hline
\end{tabular}
\end{center}
\end{table}
\begin{figure}
\centerline{ \epsfxsize=7.5cm \epsfbox{beam_setup.eps} }
\caption{Beam test setup }
\label{fig:setup}
\end{figure}
The same photomultiplier tube (Hamamatsu H2431) was used to measure the light yield
and timing characteristics of all of the prototypes.
The rise time and the transit time spread ($\sigma_{CCT}$)
are 0.7 ns and 160 ps.
The gain of the tube was calibrated by using a laser pulser (Hamamatsu PLP-02) in single photon mode.
A gain of 3.65 ADC counts per photoelectron was obtained at an operating voltage of 3100V,
with an error of 10$\%$.
The H2431 is 2" tube, with an effective photocathode diameter of 46 mm,
and a borosilicate window with a bi-alkali photocathode.
It is sensitive in the wavelength range between 300 nm to 500 nm
and has a peak quantum efficiency of 25\% at about 350 nm.
The photocathode sensitivity curve $QE(\lambda)$ is taken into
account in MC simulation.
\subsection{ {\v C}erenkov light yield of Zygo bar CCT }
\label{subsec:CCT yield}
The signals from the CCT counters were read out by CAMAC TDC and ADC modules,
with a least count of 25 ps and 0.25 pC, respectively.
Light yields were evaluated from the ADC data and the gain described above.
Figure~\ref{fig:Zygo N p.e.} shows the light yield measured for a Zygo quartz bar CCT,
as a function of tilt angle $\theta$ at Z= 60 cm.
The light yield shows a rapid decrease from $0^o$ to
a cutoff around $\theta$= -7$^o$ corresponding to a critical angle. However,
there is seen a non zero light yield below $\theta$= -10$^o$,
where none of {\v C}erenkov photons could reach the readout phototube.
These are due to {\v C}erenkov photons from $\delta$-rays produced by the beam passage.
The details of $\delta$-ray {\v C}erenkov contribution will be discussed in Section~\ref{sec:delta-ray}.
The dotted line shows a Monte Carlo prediction with a
{\v C}erenkov quality factor $N_o$ of 155 pe/cm, obtained by a fit to the data,
as described below and in Section~\ref{section:MC}.
Our measurement of $N_o$ is 20\% larger than the results measured
at SLAC~\cite{BaBar} (124 pe/cm) and Princeton Univ~\cite{PU} (121 pe/cm) with
cosmic rays. The solid line shows the prediction with $N_o$= 124 pe/cm.
This excess is attributable to the photocathode sensivity
and/or some contribution from $\delta$-ray {\v C}erenkov photons from the 1.5GeV $\pi ^- $ beam.
$N_0$ was obtained from the data
using Equation~\ref{eq:it1},
\begin{equation}
N_{pe}^{meas} = \frac{N_0 sin^2{\Theta_c}d} { cos\theta} \! ~G
\label{eq:it1}
\end{equation}
where $\Theta_c$ is {\v C}erenkov angle, d is thickness of CCT
counter, and $\theta$ is a tilt angle of the counter to the beam.
G is a collection efficiency of photons
in a given CCT counter geometry, and
it is evaluated by a Monte Carlo simulation using Equation~\ref{eq:it2},
\begin{equation}
G = \frac{N_{pe}(MC)} {\int\!\frac{dN}{d\lambda}\!~QE(\lambda)~d\lambda}\!
\frac{cos\theta}{d}\!
\label{eq:it2}
\end{equation}
$N_{pe}(MC)$ is number of photoelectrons expected by MC simulation,
$\frac{dN}{d\lambda}\!$ is {\v C}erenkov photon density and $QE(\lambda)$ is photocathode
sensitivity.
Figure~\ref{fig:Zygo att} shows {\v C}erenkov light yield as a function of distance Z
at three tilt angles: $\theta$= 0$^o$, 20$^o$ and 40$^o$.
Reflectivity per bounce $\epsilon_{ref}$ is found to be 0.9996$\pm$0.0003 by a fit.
The dotted lines show the predictions with $N_0$=155 pe/cm and $\epsilon_{ref}$= 0.9996.
The peaks at shortest distances are not predicted by this MC simulation.
The attenuation lengths in the region of Z$\ge$20 cm were found to be 10.7$\pm$0.4 m,
13.6$\pm$2.2 m and 43.8$\pm$26.7 m, respectively.
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{npe.eps} }
\caption{ {\v C}erenkov light yield vs. $\theta$ }
\label{fig:Zygo N p.e.}
\end{figure}
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig4.eps} }
\caption{ {\v C}erenkov Attenuation in Zygo bar }
\label{fig:Zygo att}
\end{figure}
\subsection{ Timing characteristics of the Zygo bar CCT }\label{subsec:Zygo CCT}
The discriminator level was set at 50 mV, which corresponded to
a few photoelectron signal.
The start time jitter for each run was evaluated to be 40$\sim$50 ps, and it was subtracted
in quadrature to obtain the intrinsic time resolution for each of the CCT prototypes.
The start time for each event was evaluated using a pair of start scintillation counters S1 and S2.
Each TDC measurement was corrected from time walk by using a formula $T_C \ =\ (TDC - TDC_{start}) + a - b/\sqrt{ADC} $.
Figure~\ref{fig:CCT twc} shows the fitted slope parameter $b$ as function of
Z and $\theta$.
The parameter b shows very little dependence on Z at $\theta$= 40$^o$,
while it shows a strong correlation with Z at $\theta$= 20$^o$ and 0$^o$.
This is due to contamination from $\delta$-ray {\v C}erenkov photons.
The resultant time resolution was investigated as a function of $b$,
and a single value was chosen to correct all of the TDC data.
It corresponded to a value of about 400 found at $\theta$= 40$^o$ over all Z, to which
all of the values converge at Z$\le$15 cm, as shown in Figure~\ref{fig:CCT twc}.
Figure~\ref{fig:CCT timing} shows corrected timing $T_C$ as a function of $\theta$ at Z= 60 cm.
Prediction by Monte Carlo simulation shows good agreement with the data, with an {\it rms} of 20 ps.
In the region of $\theta$ $\le$ -10$^o$, $T_C$ is almost independent of $\theta$,
suggesting a uniform cosine distribution due to $\delta$-ray {\v C}erenkov photons.
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig7.eps} }
\caption{ Slope parameters of time walk as a function of Z }
\label{fig:CCT twc}
\end{figure}
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig2.eps} }
\caption{ CCT timing vs. $\theta$ at z=60cm }
\label{fig:CCT timing}
\end{figure}
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig6.eps} }
\caption{ $\sigma_{CCT}$ vs. Z at (a) $\theta$= 40$^o$, (b) 20$^o$
and (c) 0$^o$, compared with Monte Carlo prediction }
\label{fig:CCT sigma}
\end{figure}
Figure~\ref{fig:CCT sigma} shows time resolution $\sigma_{CCT}$ as a function of Z
at the three tilt angles (a) 40$^o$, (b) 20$^o$ and (c) 0$^o$, respectively.
The $\sigma_{CCT}$'s being 30 ps at Z= 5$\sim$10 cm,
increasing with Z up to 50 ps, 80 ps and 105 ps at Z=100 cm, respectively.
The MC predictions ( dotted lines ) show good agreement with the data.
A systematic discrepancy is observed at $\theta$= 0$^o$, which increases
up to $\sim$20\% at Z= 100 cm.
\section{ Monte Carlo Simulation of CCT prototype }\label{section:MC}
According to the given functions~\cite{mats-cct} for the {\v C}erenkov photon density
and for the dispersion,
{\v C}erenkov photons are emitted along the path of particle passage through the CCT counter.
Each photon propagates by total internal reflection, at a critical angle
given by its wavelength.
After many bounces on the counter surface and going through the couplant,
some fraction of the {\v C}erenkov photons arrive at read out phototube
and emit photoelectrons.
The data of M.Griot in Ref.~\cite{BaBar} was used to simulate photon absorption in the quartz bar.
In this simulation, only two parameters were assumed to be free: reflectivity per bounce
($\epsilon_{ref}$), and a normalization factor ($Q$).
This $Q$ corresponds to a peak quantum efficiency (nominally $\sim$25\%).
These parameters were found to be 0.9996$\pm$0.0003 for $\epsilon_{ref}$,
and 25\% ($N_o$=155 pe/cm) for $Q$ by fitting to the data.
We take into account the timing characteristics of
the phototube as described in section~\ref{subsec:Zygo CCT pmt}.
In addition, we introduce a pulse height fluctuation for single photoelectron signals~\cite{mats-cct}~\cite{pmt}.
For this, a Gaussian distribution was assumed, having an average
of 3.65 ADC counts and a standard deviation of a half of the average.
To simulate the time walk effect, a discriminator was also included.
An assumption for $\sigma_{TTS}$ of 160 ps is well reproduced for single photoelectron signals,
by applying the time walk correction to Monte Carlo data.
\section{ Effects from $\delta$-rays }\label{sec:delta-ray}
A Monte Carlo simulation using GEANT code has been carried out to study the effects of
$\delta$-rays on quartz bar CCT counters, with 1.5GeV $\pi ^- $ beam.
The fraction of the photons from $\delta$-rays is expected to be 15\% at normal beam incidence.
The number of beam {\v C}erenkov photons shows a Gaussian (Poisson) distribution,
while that from $\delta$-ray {\v C}erenkov photons shows a rapidly decaying tail peaked at zero.
The observed ADC distribution is a sum of those two components.
The angular divergence of the photons from $\delta$-rays is about 100$^o$ in FWHM,
showing an almost uniform cosine distribution with respect to the beam direction.
The spread in arrival times of photons is expected to be proportional to their propagation length.
In contrast, those from a beam particle are well collimated
with a divergence of less than 1$^o$ in FWHM.
The two components behave very differently as function of Z and $\theta$, so that
they are expected to separate at longer Z and at smaller $\theta$.
Figure~\ref{fig:delta-ray}(a) shows ADC data at Z=60 cm and $\theta$=$0^o$.
The tail in the ADC and the TDC distributions were separated by a fit to a Gaussian plus a tail.
The results are shown in Figure~\ref{fig:delta-ray}(c).
10$\sim$12\% of the photons are attributable to $\delta$-rays.
Subtracting this contribution, a quality factor $N_o$ is calculated to be 136 pe/cm.
Figure~\ref{fig:delta-ray}(b) shows a TDC distribution after time walk correction.
The tail at earlier times corresponds to $\delta$-ray {\v C}erenkov photons.
Figure~\ref{fig:delta-ray}(d) shows the results
as a function of Z at $\theta$=$0^o$.
This indicates that 16\% of events are triggered by $\delta$-ray
{\v C}erenkov photons at Z= 100 cm.
As seen in Figure~\ref{fig:delta-ray}(b), about 10\% of events would be triggered at a time
earlier by 2$\sigma$ than expected from beam {\v C}erenkov photons.
Such a wrong trigger probability due to $\delta$-ray {\v C}erenkov photons may depend
on discrimination level, as well as Z and $\theta$.
\begin{figure}
\centerline{ \epsfxsize=8.5cm \epsfbox{delta.eps} }
\caption{ (a) ADC and (b) TDC distributions at z=60cm and at $\theta$=0$^o$,
and fraction of $\delta$-ray events
as function of (c) $\theta$ at Z= 60 cm
and (d) Z at $theta$= 0$^o$, estimated by a Gaussiun
fit to ADC and TDC distributions. }
\label{fig:delta-ray}
\end{figure}
\section{ Results from other prototypes }
In Table~\ref{tab:CCTpro} are listed
the measured {\v C}herenkov quality factors, attenuation length and time
resolution at $\theta$= 20$^o$, and {\v C}erenkov angle resolution ($\delta_{\Theta_C}$)
at $\theta$= 0$^o$.
The attenuation length of Surface Finishes (SF) quartz bar is measured to be 1.8 m, which is mainly
attributable to its beveled edges. The acrylic bars show a short attenuation length
of about 1.4 m and quality factors of 66$\sim$82 pe/cm.
Figure~\ref{fig:CCT_sigma} shows the measured time resolution of a Zygo bar at $\theta$=20$^o$,
as a function of Z, together with other prototype counters.
These are separated into groups of quartz and acrylic bars. The curves are fits to a function
$\sigma_{CCT}=\sqrt{a+b\times L}$, assuming that the resolution is a function of
the number of bounces of the photons on the surface.
The measured resolutions of the quartz bars are worse than the MC prediction by about 25\%.
As described in section~\ref{subsec:CCT yield},~\ref{subsec:Zygo CCT} and in Ref~\cite{mats-rich},
the optical characteristics of the Zygo bar are extremely good,
while the obtained time resolutions are almost the same as for the SF bar.
The acrylic bars show fairly good time resolutions,
in spite of a poorer surface quality and a cutoff at lower wavelength.
These are worse only by about 20\% as shown in Figure~\ref{fig:CCT_sigma}.
The fraction of $\delta$-ray {\v C}erenkov photons
are also found to be about 16\% for the acrylic bars.
Figure~\ref{fig:Angle separation} shows {\v C}erenkov angle resolution as a function of
Z at $\theta$= $0^o$ and $20^o$, using the fitted curves to the Zygo quartz bar data.
A {\v C}erenkov angle resolution of 15 mrad is expected at a propagation distance of 100 cm and at
normal incidence, which corresponds to 3$\sigma$-$\pi/K$ separation up to a momentum of 1.4 GeV/c.
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig9.eps} }
\caption{ $\sigma_{CCT}$ vs. Z for quartz and acrylic CCT
at $\theta$=20$^o$}
\label{fig:CCT_sigma}
\end{figure}
\section{Summary}
The {\v C}erenkov light yield and timing characteristics have been investigated
for quartz and acrylic bar prototypes with a 1.5 GeV/c $\pi^-$ beam,
and the feasibility of the CCT detector concept
has been demonstrated.
The results for a Zygo quartz bar CCT are in good agreement with the previous measurements with cosmic rays.
A {\v C}erenkov quality factor $N_o$ of 136 pe/cm and a reflectivity per bounce of 0.9996$\pm$0.0003
have been obtained.
The measured {\v C}erenkov light yield and CCT timing performance are well reproduced by Monte Carlo
simulation, with the exception of $\sigma_{CCT}$ at smaller tilt angles and at larger distances.
A Surface Finishes quartz bar showed almost the same $\sigma_{CCT}$ as the Zygo bar,
in spite of its shorter attenuation length.
Acrylic bars have been investigated and found to show a better than expected $\sigma_{CCT}$,
worse only by about 20\% than the quartz bars,
in spite of a smaller $N_o$ of 66$\sim$82 pe/cm, a shorter attenuation
length of 1.4 m and a marginal surface quality.
The number of bounces of photons on the surface seems to be a key factor in determining
the time resolution. However, requirements for the surface quality
seem not to be strict for CCT performance.
We gratefully acknowledge the support of the Department of Energy, the
National Science Foundation, and the A. P. Sloan Foundation.
We would also like to express our thanks to Prof. S.Iwata and
Prof. F.Takasaki at KEK for their support of this work.
\begin{figure}
\centerline{ \epsfxsize=6.5cm \epsfbox{fig12.eps} }
\caption{ {\v C}erenkov angle resolution vs. Z }
\label{fig:Angle separation}
\end{figure}
|
\section{Introduction}
This article is meant to describe some preliminary applications of duality to
unravelling the implications of string theory for black hole physics in four
dimensions \cite{bhsoln,bhdual}. Before summarizing the points to be outlined,
some motivation is in order as to why duality transformations, and their
applications to black hole physics, are worth thinking about.
Duality transformations \cite[gives an extensive review]{dualityreview} are a
particular type of change of variable which have come to play increasingly
important roles in extracting the physical content of string theory, and of some
ordinary quantum field theories. Field theories which are related by duality
transformations are thought to be completely equivalent, even though they may
appear to be very different. Since it is sometimes true that intractible
strongly-coupled theories turn out to be dual to calculable weakly-coupled ones,
the physical equivalence of the dual theories can be exploited to infer the
behaviour of otherwise unsolvable systems.
The development of this new tool may open new approaches to studying old
problems of compelling physical interest which have hitherto been too difficult
to crack. In particular, a current hope is that they may be useful for
extracting the predictions of string theory for the dynamics of spacetimes
having very strong, or even singular, curvatures, such as are believed to arise
in the final state of runaway gravitational collapse, or in the earliest moments
of the universe. The predictions of string theory are of particular interest
for this long-standing puzzle, since string theory is the only known theory
which gives sensible results for the simpler, but related, problem of the
scattering of particles in flat space at energies at and above the Planck-scale.
The good news (so far) is that there are preliminary indications that string
theory has qualitatively new features which may figure importantly in our final
understanding of gravitational collapse. The main new feature to emerge so far
is the realization that string theories appear to be quite forgiving in their
notion of what constitutes a physically unacceptable singularity. What might
be a malignantly singular field configuration from the point of view of ordinary
point-particle field theories, can be completely benign as a background for
string propagation. This understanding has emerged as more solutions have been
constructed \cite[for a recent review]{exactsolns} to the full string equations,
including some singular ones. It is also indicated by the existence of duality
transformations which relate spacetimes with singularities to duals which are
equivalent, and yet are absolutely nonsingular (including even Minkowski space).
Some of the dual spacetimes which have been related in this way even include
black hole spacetimes \cite{twodbhone,twodbhtwo,twodbhthree}, although
admittedly only in two spacetime dimensions.
All of this motivates further study of black-hole-type field configurations
within the context of string theory. This article reports on the results of two
lines of inquiry in this direction. These two lines may be summarized as
follows:
\begin{enumerate}
\item
The first investigation \cite{bhsoln} uses duality transformations, and some of
their extensions, to generate new classical solutions to the string equations of
motion, starting from simple initial solutions. In particular, they are used to
systematically construct all possible spherically-symmetric, time-independent
and asymptotically-flat solutions in four dimensions. There are two new features
to these results. ($i$) First, because using duality to generate new solutions
is a purely algebraic proceedure, it is much simpler than a direct attempt to
solve the relevant set of nonlinear, coupled, partial-differential equations.
This simplicity has permitted the inclusion of more nontrivial background
fields than previously had been tractable. ($ii$) Second, keeping in mind the
broad-mindedness of string theory towards singularities, {\em all} solutions are
presented, including some that are singular. Singular solutions were often
omitted in previous constructions.
\item
The second application \cite{bhdual} is to critically analyze the situation
under which dual configurations can be expected to be physically equivalent. In
particular, an important class of duality transformations --- including most of
those applied to black holes --- are argued to {\em not} relate
physically-equivalent spacetimes in the usual sense. The same considerations
also lead to restrictions on the boundary conditions which must be satisfied by
some fields in order for the duality transformations to take their usual form.
\end{enumerate}
These two developments are presented in the remainder of this article in the
following way. \S 2 sets the stage by briefly presenting a reminder
of the connection between string field configurations, and two-dimensional
conformal field theories. This is followed by the two applications listed
above. First, \S 3 through \S 5 present the use of duality-related
transformations for constructing solutions to the low-energy field equations.
\S 3 summarizes the transformations themselves, while \S 4 gives
the simplest --- {\it i.e.}\/\ dilaton-metric --- solutions. \S 5 then applies the
transformations of \S 3 to the field configurations of \S 4, thereby
generating solutions also incorporating nonzero axion and gauge fields.
The second line of argument is the topic of \S 6 through \S 8. \S 6 states
the form of the duality transformations in a manner which is sufficiently
general for the desired applications. Their implications for the general
spherically-symmetric, time-independent and asymptotically-flat solutions,
including in particular the black hole solutions, are the topic of \S 7. The
result, as applied to black hole configurations, makes it difficult to see how
the dual solutions can be physically equivalent. \S 8 is dedicated to the
resolution of this puzzle.
The conclusions which follow from these sections are finally summarized in \S 9.
\section{Strings and 2D Field Theories}
This section is meant to review the connection between string field
configurations, and two-dimensional conformal field theories, since this
underlies all of what follows. This connection allows the application of
results for two-dimensional field theories to draw conclusions about
solutions to the string field equations.
\subsection{String Field Configurations}
A classical solution to the string theory equations of motion is equivalent to
a conformally-invariant, two-dimensional field theory. The connection is
simplest to see in the case that the two-dimensional field theory is a sigma
model, whose two-dimensional scalar fields, $x^\mu(\sigma)$, can be interpreted
as describing the worldsheet of a string propagating within a `target'
spacetime. For instance, suppose the sigma model action has the form
\begin{eqnarray}
\label{sigmamodelaction}
S[x^\mu] &=& - \, {1 \over 2 \pi \alpha'} \int d^2\sigma \; \sqrt{\gamma} \,
\Bigl[ G_{\mu\nu}(x) \, \gamma^{\alpha\beta} + B_{\mu\nu}(x) \,
\epsilon^{\alpha\beta} \Bigr] \; \partial_\alpha x^\mu \partial_\beta x^\nu
\nonumber\\
&& \qquad \qquad \qquad - \, {1 \over 8 \pi} \int d^2\sigma \; \sqrt{\gamma}
\; \phi(x) \Sc{R} ,
\end{eqnarray}
where $\gamma_{\alpha\beta}$ and $\epsilon^{\alpha\beta}$ are the
two-dimensional metric and antisymmetric Levi-Civita symbol and $\Sc{R}$ is the
curvature scalar for the metric $\gamma_{\alpha\beta}$.
The coupling functions, $G_{\mu\nu}(x)$, $B_{\mu\nu}(x)$ and $\phi(x)$, are
interpreted in string theory as being background values for three fields ---
the metric, the `axion', and the `dilaton' --- which represent three of the
modes of the string. These particular modes would have been massless if the
string were propagating through Minkowski space. There are typically other such
nearly `massless' modes as well in a string theory, and we shall consider later a
spin-one gauge fields, $A_\mu(x)$, in addition to the above three.
Now comes the main point. There are two ways in which one might imagine
defining the equations of motion for these fields, given our presently limited
understanding of string theory itself. The direct way is to infer the
interactions of the various string modes by computing their tree-level
scattering on simple spacetimes, such as for Minkowski space. One finds an
action which reproduces these scattering amplitudes, and then computes the
equation of motion using this action.
The alternative way to determine the equations of motion for $G_{\mu\nu}$,
$B_{\mu\nu}$, $\phi$ and $A_\mu$ is to compute the conditions which these
quantities must satisfy in order for the corresponding two-dimensional theory to
be conformally invariant at the quantum level. It is an amazing fact that these
two methods produce results which agree with one another --- up to the
ubiquitous freedom to perform field redefinitions --- in all cases for which
they have been compared.
The resulting field equations can be explicitly written as an expansion in
powers of derivatives of the background fields times the dimensionful constant
$\alpha'$. In four dimensions the action which reproduces the leading terms in
these equations for the fields of interest is \cite{GSW}:
\begin{equation}
\label{leaction}
\Sc{L} = {e^\phi \over 8 \pi \alpha'} \; \sqrt{- G} \, \left[ R(G) + (\nabla
\phi)^2 - {1 \over 12} \, H^{\mu\nu\lambda} H_{\mu\nu\lambda} - {1\over 8} \,
F^{\mu\nu}F_{\mu\nu} \right] + \cdots \, ,
\end{equation}
where $H_{\mu\nu\lambda} = \partial_\mu B_{\nu\lambda} - \nth{4} \, A_\mu
F_{\nu\lambda} + \hbox{(cyclic permutations)}$\footnote{By including the
gauge-field-dependent Chern-Simons term, $\nth{4} \, AF$, in the definition of
$H$ we restrict our discussion of gauge fields to the heterotic string. By
contrast, the discussion for $A_\mu = 0$ applies equally well to bosonic
and superstrings.} and $F_{\mu\nu} = \partial_\mu A_\nu - \partial_\nu A_\mu$
are, respectively, the field strengths for the axion and electromagnetic fields.
$R(G)$ is the Ricci scalar for the metric, $G_{\mu\nu}$, and the square root
involves the quantity: $G = \det G_{\mu\nu}$. This metric is often called the
`sigma-model' metric to distinguish it from the one for which a field
redefinition has been performed to put the Einstein term into standard form. The
ellipses in eq.~\pref{leaction} represent terms which involve other massless
fields and others involving more derivatives that arise at higher orders in
$\alpha'$.
Some of the higher-order corrections to these equations in the derivative
expansion have also been computed. For later purposes we quote the quantity
which is responsible for the higher-order corrections to configurations
involving only the metric and the dilaton, which turns out to be \cite{GSW}:
\begin{equation}
\label{corrtoleaction}
\delta \Sc{L} = {\lambda e^\phi \over 2} \; \sqrt{-G} \;
R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma} \ .
\end{equation}
Here $\lambda$ is $\frac{1}{2}$ for bosonic strings, $\nth{4}$ for heterotic strings,
and $0$ for supersymmetric strings.
This establishes a direct connection between background string configurations
and two-dimensional conformal field theories. The spectrum of string states
corresponds to a class of operators in these field theories which represent the
conformal group in a particular way. The scattering of these string states can
be computed by evaluating appropriate correlation functions involving these
operators in the conformal field theory.
Given this connection it is immediate how to interpret the equivalence between
any two conformal field theories in terms of string states. If two conformal
theories can be shown to have precisely the same content, then they must
describe identical string scattering about identical background string
configurations. Duality transformations are interesting because they imply
precisely this kind of equivalence between two-dimensional field theories.
\section{Classical Transformations}
There is a broad class of transformations which are guaranteed to map conformal
field theories into other conformal field theories, even though the theories
which are related in this way need not be physically equivalent. That is to
say, this broader class of transformations are guaranteed to take classical
string vacua into other classical vacua, but the full Hilbert space of string
modes constructed about these vacua can be different. Although these
transformations cannot therefore be considered to be {\it bona-fide} string
symmetries, they are nevertheless very useful for generating new classical
solutions from known ones.
There are two transformations of this type which are used in what follows. We
outline both of these in the following two subsections.
\subsection{$SL(2,\relax{\rm I\kern-.18em R})$ Transformations}
The first class of such transformations is a group of $SL(2,\relax{\rm I\kern-.18em R})$
transformations \cite{STW,MDR,Sen} which includes the classical string
$S$-duality transformations \cite{Wit,BFQ,LPS} of the low-energy effective
theory. To formulate these transformations it is useful to use the scalar
variable, $a(x)$, which is dual to the three-index axion field,
$H_{\mu\nu\lambda}$:
\begin{equation} \label{axion}
H_{\mu\nu\rho} = -e^{-2\phi}\epsilon_{\mu\nu\rho\kappa}\nabla^\kappa a .
\end{equation}
Here all indices are raised and lowered with the Einstein metric,
$g_{\mu\nu}$, which (in four spacetime dimensions) is related to the sigma-model
metric, $G_{\mu\nu}$ by: $g_{\mu\nu} = e^\phi \; G_{\mu\nu}$, and the
Levi-Civita tensor is also constructed using $g_{\mu\nu}$. Many properties of
the theory take a simple form if the field, $a$, is combined with the dilaton,
$\phi$, into the complex combination, $S = a + i\, e^{\phi}$. In terms of this
variable the $SL(2,\relax{\rm I\kern-.18em R})$ transformation becomes
\begin{eqnarray}
\label{sltrt}
S &\longrightarrow& { {\rm a}\, S + {\rm b} \over {\rm c} \,
S +{\rm d}} \nonumber\\
(F_+)_{\mu\nu} &\longrightarrow& \left( {\rm c}\, S +{\rm d} \right)\,
\left( F_+ \right)_{\mu\nu} \\
(F_-)_{\mu\nu} &\longrightarrow& \left( {\rm c}\, S^* +{\rm d} \right)\,
\left( F_- \right)_{\mu\nu} . \nonumber
\end{eqnarray}
where $S^*$ is the complex conjugate of $S$, and $(F_\pm)_{\mu\nu} \equiv
F_{\mu\nu} \pm {i\over2} \; \epsilon_{\mu\nu\rho\kappa} \, F^{\rho\kappa}$.
Once again it is the Einstein metric which is involved in these definitions. The
quantities a, b, c and d are real numbers which must satisfy ${\rm ad} - {\rm
bc} = 1$. If $S$, $g_{\mu\nu}$, and $F_{\mu\nu}$ all satisfy the string
equations of motion, then so must the transformed variables as defined by
eq.~\pref{sltrt}.
\subsection{The $O(1,1)$ Transformations}
There is a second transformation which can be used to generate new classical
string solutions from old ones \cite{CFG,MV,Sen92,GMV,HS}. In the simplest case
of a field configuration that is independent of a single coordinate, $s$, there
is an $O(1,1)$ group of such transformations. The action of these
transformations is most easily written when the background fields are written as
the following $9 \times 9$ matrix
\begin{equation}
\label{Mmatrix}
{\cal M} =
\pmatrix{{\cal K}_-^TG^{-1} {\cal K}_- & {\cal K}_-^TG^{-1}{\cal K}_+ & -{\cal K}_-^TG^{-1}A \cr
{\cal K}_+^TG^{-1}{\cal K}_- & {\cal K}_+^TG^{-1}{\cal K}_+ & -{\cal K}_+^TG^{-1}A \cr
-A^TG^{-1}{\cal K}_- & -A^TG^{-1}{\cal K}_+ & A^TG^{-1}A \cr}
\end{equation}
where
\begin{equation}
\label{Kmatrix}
({\cal K}_\pm)_{\mu\nu}=-B_{\mu\nu}-\Sc{G}_{\mu\nu} \pm \eta_{\mu\nu} ,
\end{equation}
and $\eta_{\mu\nu}$ is the flat Minkowski metric in four dimensions. In these
expressions the quantity $\Sc{G}_{\mu\nu}$ is defined by: $\Sc{G}_{\mu\nu} =
G_{\mu\nu} + \nth{4} \, A_\mu A_\nu$. For a detailed statement of the
conventions used, see ref. \cite{bhsoln}.
With these variables the $O(1,1)$ transformations can be expressed in matrix
form, ${\cal M}\to \Omega {\cal M}\Omega^T$, where the transformation matrix is
given by
\begin{equation}
\label{Omatrix}
\Omega=\pmatrix{I_7&0&0\cr0&x&
\sqrt{x^2-1}\cr0&\sqrt{x^2-1}&x} .
\end{equation}
Here $I_7$ is the $7\times7$ unit matrix, and $x$ is a parameter satisfying
$x^2 \geq 1$. The action of the $O(1,1)$ transformations on the dilaton is
given by
\begin{equation}
\label{dilatonp}
e^{\phi} \to \left({{\rm det}G\over{\rm
det}G^\prime}\right)^{1\over2} e^\phi\ \ ,
\end{equation}
where $G'_{\mu\nu}$ denotes the transformed sigma-model metric.
\section{Dilaton-Metric Configurations}
The goal now is to use these transformations to generate the most general
spherically-symmetric, asymptotically-flat and static solutions to the string
equations. This will be done by applying the transformations to a particularly
simple class of solutions. The first step --- identifying this initial simple
class of solutions --- is the topic of the present section.
\subsection{The Lowest-Order Solution}
The starting point is the most general four-dimensional, asymptotically flat,
static and spherically-symmetric configuration which solves the string
equations to leading order in $\alpha'$, and which involves only the metric and
dilaton fields. We therefore directly solve the equations:
\begin{eqnarray}
\label{eqofmotion}
R_{\mu\nu}(g) &=& \frac{1}{2} \nabla_\mu \phi \nabla_\nu \phi \nonumber\\
\nabla^2 \phi &=& 0 ,
\end{eqnarray}
using the {\em ansatz}
\begin{eqnarray}
\label{sssconfig}
g_{\mu\nu} dx^\mu dx^\nu &=& - f(r) dt^2 + g(r) dr^2 + h^2(r) (
d\theta^2 + \sin^2 \theta \, d\varphi^2 ) , \nonumber\\
\phi &=& \phi(r), \qquad
\end{eqnarray}
The result is a family of field configurations whose study actually predates
the earliest advent of string theory itself \cite{Buchdahl,JRW}:
\begin{eqnarray}
\label{answer}
f = {1 \over g} &=& \Lambda^\delta(r) \nonumber\\
h^2 &=& r^2 \Lambda^{1-\delta}(r) \\
e^\phi &=& e^{\phi_0} \; \Lambda^\gamma(r) , \nonumber
\end{eqnarray}
where $\Lambda(r)$ is a shorthand for the function $1 - (\ell / r)$. $\ell$ is
the only dimensionful parameter in the solutions, and so it simply sets their
overall scale. We assume both $\ell$ and $r$ to be positive in what follows.
The two dimensionless quantities, $\delta$ and $\gamma$, are the arbitrary
parameters which label the solutions, subject to the condition $\delta^2
+ \gamma^2 = 1$.
$\phi_0$ is the asymptotic value which is obtained by the dilaton field as $r
\to \infty$. It is convenient in what follows to shift $\phi$ by a constant to
ensure that it tends to zero for large $r$, and so to completely remove $\phi_0$
from the solutions. This is always possible in string theory, with the general
result --- for classical string solutions --- that $\phi_0$ only appears as an
overall factor in the low-energy lagrangian of eqs.~\pref{leaction} and
\pref{corrtoleaction}. It can therefore be absorbed into the definition of
Newton's constant, which we take (in four dimensions) to be $G_{\scrs N} = \frac{1}{2} \,
e^{-\phi_0} \, \alpha'$.
The choice $(\delta,\gamma)=(1,0)$ yields the Schwarzschild metric with constant
dilaton field, and $\ell$ is in this case related to Newton's constant, $G_{\scrs N}$,
and the black hole mass, $M$, by $\ell = 2 G_{\scrs N} M$. So long as $\gamma \ell \neq
0$, however, the metric of eq.~\pref{answer} has a real curvature singularity at
$r=\ell$.
\subsection{The Conserved Charges}
The two independent parameters which label this family of solutions have a
useful interpretation in terms of the asymptotic behaviour, as $r \to \infty$,
of the fields involved. In general, for large $r$, the solutions take the form:
\begin{eqnarray}
\label{gasymform}
f(r) &=& 1 - {\Sc{A} \over r} + \cdots; \nonumber\\
g(r) &=& 1 + {\Sc{B} \over r} + \cdots ; \\
h(r) &=& r^2 \left[ 1 + O\!\left( {1 \over r} \right) \right] \, , \nonumber\\
e^{\phi(r)} &=& 1 - {Q_{\ss{D}} \over r} + \cdots \, , \nonumber
\end{eqnarray}
with constants, $\Sc{A} = \Sc{B} = \delta \ell$ and $Q_{\ss{D}} = \gamma
\ell$.
These constants have a physical interpretation as specifying the corresponding
conserved charges which are carried by the solutions. That is, the conserved
(ADM \cite{ADM}) mass, $M$, of the solution is related to the constant $\Sc{A}$
by $2 G_{\scrs N} \, M = \Sc{A}$. The constant, $Q_{\ss{D}}$, similarly defines a {\em
dilaton charge} for the solution.
The utility of labelling the solutions by their values for $M$ and $Q_{\ss{D}}$
is that these quantities are equally well defined for the complete solution to
the string equations. Given these asymptotic expressions, the complete equations
may be solved order by order in $\alpha'$, giving a unique solution for {\em
any} choice of $M$ and $Q_{\ss{D}}$. The same is not true for the parameters
$\delta$ and $\gamma$, which are defined in terms of the specific form of the
solution to these equations only at lowest order in $\alpha'$. For instance, if we
focus on the black hole solution --- which we take to be the solution for which
the potential singularity at $r = \ell$ is only a coordinate artifact --- and
work to next order in $\alpha'$ by including eq.~\pref{corrtoleaction} into the
low-energy string action, then the result is still characterized by the
quantities $M$ and $Q_{\ss{D}}$, but with \cite{CMP}:
\begin{eqnarray}
\label{bhasymform}
2 G_{\scrs N} M = \Sc{A} = \Sc{B} &=& \ell + {11 \lambda \alpha' \over 6 \ell}
+ O(\alpha'^2) ,\nonumber\\
Q_{\ss{D}} &=& - \; {2 \lambda \alpha' \over \ell} + O(\alpha'^2) .
\end{eqnarray}
Recall that the constant $\lambda$ appearing here is $\frac{1}{2}$ in the bosonic
string, $\nth{4}$ in the heterotic string, and 0 in the superstring.
\section{More General Classical Solutions}
It is now possible to compute more general classical solutions, simply by
repeatedly applying the $SL(2,\relax{\rm I\kern-.18em R})$ and $O(1,1)$ transformations of the
previous sections to the above solutions. Before doing so explicitly, it is
first worth characterizing the possible parameters which can describe these
solutions by first analyzing the asymptotic behaviour which is permitted for
the fields at large $r$.
\subsection{The Conserved Charges}
Once the metric and dilaton fields are supplemented by nonzero axion and gauge
fields, more complicated solutions become possible. We take the following
spherically-symmetric and time-independent {\em ansatz} for the axion
and gauge fields:
\begin{equation}
\label{sssdilem}
a = a(r), \qquad F_{tr} = E(r), \qquad F_{\theta \varphi} = B(r) \sin \theta.
\end{equation}
and we generalize, for later convenience, our metric {\em ansatz} to include
stationary, but not static, metrics:
\begin{equation}
\label{taubnutansatz}
g_{\mu\nu} dx^\mu dx^\nu = - f(r) (dt + 2 N \cos\theta \; d\phi)^2 + g(r) dr^2 +
h^2(r) ( d\theta^2 + \sin^2 \theta \, d\varphi^2 ).
\end{equation}
The parameter $N$ here is called the {\em NUT} parameter, due to the
similarity of eq.~\pref{taubnutansatz} with the Taub-NUT metric
\cite{Taub,NUT,CEtoE}.
The corresponding new conserved charges can be inferred by examining the
large-$r$ behaviour that is implied by the field equations for these fields.
This is given, for the dilaton and metric, by eqs.~\pref{gasymform}, and for the
axion and gauge fields by:
\begin{eqnarray}
\label{emaform}
a(r) &=& - \; {Q_{\ss{A}} \over r} + \cdots; \nonumber\\
A_t &=& {Q_{\ss{E}} \over r} + \cdots, \\
A_{\varphi} &=& -\; Q_{\ss{M}}\, \cos\theta + \cdots .\nonumber
\end{eqnarray}
The constants $Q_{\ss{A}}$, $Q_{\ss{E}}$ and $Q_{\ss{M}}$ are the solution's
axion, electric and magnetic charges, respectively.
\subsection{Dilaton-Axion-Metric Solutions}
The simplest generalization is the inclusion of a nonzero axion field in
addition to the original dilaton and metric configurations. These may be
generated by applying the $SL(2,\relax{\rm I\kern-.18em R})$ transformations, eq.~\pref{sltrt}, to
the solution of eq.~\pref{answer}, being careful to preserve the boundary
conditions for $\phi$ and to ensure that $a \to 0$ at large $r$. (There is no
loss of generality in choosing this boundary condition for the axion field,
since the definition, eq.~\pref{axion}, only defines $a(r)$ up to an additive
constant.)
Performing the $SL(2,\relax{\rm I\kern-.18em R})$ transformation, we obtain the same metric
configuration as before, but the new dilaton and axion fields, $\hat\phi$ and
$\hat a$:
\begin{equation}
\label{firstt}
e^{\hat\phi} = \left(1+\omega^2\right){\Lambda^\gamma(r)
\over \omega^2 \Lambda^{2\gamma}(r) +1} , \qquad
\hat a ={\omega \left[ \Lambda^{2\gamma}(r) -1 \right] \over
\omega^2 \Lambda^{2\gamma}(r) +1} ,
\end{equation}
$\omega$ is the new real parameter of the solution, while $\delta$ and $\gamma$
are the labels of the original dilaton-metric configuration. The starting
dilaton-metric solution is re-obtained as the special case $\omega = 0$. These
parameters are related to the three conserved charges of the
dilaton-axion-metric system by:
\begin{equation}
\label{charges}
2 G_{\scrs N} \, M = \delta \ell , \qquad
Q_{\ss{D}}= {1-\omega^2 \over 1+\omega^2}\; \gamma \, \ell \qquad
\hbox{ and} \qquad Q_{\ss{A}} ={2\omega \, \gamma \, \ell\over 1+\omega^2} .
\end{equation}
All of these solutions have real singularities at $r=\ell$, except for the
special case $\gamma \ell =0$. An interesting special limit of these solutions
is the case $\omega=\pm1$, for which the dilaton charge, $Q_{\ss{D}}$, vanishes.
Notice that even for this choice, however, the dilaton field is nonvanishing
due to the nontrivial axion configuration.
\subsection{Including a Gauge Potential}
It is conceptually no more difficult, although algebraically more tedious, to
generate the general dilaton--axion--metric--gauge-potential configuration.
This is obtained by repeatedly performing $SL(2,\relax{\rm I\kern-.18em R})$ and $O(1,1)$
transformations to the previous solutions. Three new parameters, $x$, $\epsilon$
and $\rho$, enter the solution in this way before these transformations start
just regenerating previously-obtained configurations. These three parameters
are related to the electric and magnetic charges, $Q_{\ss{E}}$ and $Q_{\ss{M}}$,
of the background electromagnetic fields, and to the NUT parameter, $N$,
of the metric (which is defined by eq.~\pref{taubnutansatz}).
The result of this process is a fairly complicated field configuration, whose
explicit form is not particularly illuminating and so is not repeated here.
Detailed expressions are given in \cite{bhsoln}. Instead, we turn to the action
of duality on these solutions, and on their extensions to higher order in the
$\alpha'$ expansion.
\section{Duality Transformations}
The remainder of this article is devoted to exploring the implications of
duality transformations for the black-hole, and singular, solutions just
constructed. Some of the results we obtain also apply to the exact solutions
to the string equations.
\subsection{The Transformation Rules}
The first step is to define what is meant by duality transformations.
A reasonably general algorithm has emerged with which it is always possible to
systematically generate dual field theories from a given one
\cite{algorithm,AAL,AABL}, generalizing an earlier construction which had
been developed for the earliest known string duality \cite{Buscher}. The
algorithm applies to any field theory which admits a continuous symmetry.
It is simple to state the result for the case of the sigma model,
eq.~\pref{sigmamodelaction}, when the symmetry corresponds to the independence
of one of the coordinate directions, of the fields $G_{\mu\nu}$, $B_{\mu\nu}$,
$\phi$ and $A_\mu$. Denoting this direction by $s$, then the sigma model which
is dual to the original one involves the fields $\twi{G}_{\mu\nu}$,
$\twi{B}_{\mu\nu}$, $\tilde\phi$ and $\twi{A}_\mu$, where
\cite{Buscher,GRV,SW,AAB}:
\begin{eqnarray}
\label{dualitytransformation}
\twi{\Sc{G}}_{ss} &=& 1/\Sc{G}_{ss},\qquad
\twi{\Sc{G}}_{s\mu}=-B_{s\mu}/\Sc{G}_{ss},\nonumber\\
\twi{\Sc{G}}_{\mu\nu} &=& \Sc{G}_{\mu\nu} - \left[
{\Sc{G}_{s\mu}\Sc{G}_{s\nu} + B_{s\mu} B_{s\nu}\over \Sc{G}_{ss}} \right] \\
\twi{B}_{s\mu} &=& -{\Sc{G}_{s\mu}}/{\Sc{G}_{ss}},\qquad
\twi{B}_{\mu\nu}=B_{\mu\nu}+{\Sc{G}_{s\mu} B_{s\nu}
-\Sc{G}_{s\nu}B_{s\mu}\over \Sc{G}_{ss}}, \nonumber \\
\twi{A}_s &=& - \; {A_s \over \Sc{G}_{ss}} \nonumber\\
\twi{A}_\mu &=& A_\mu -A_s \; {\Sc{G}_{s\mu} -B_{s\mu} \over \Sc{G}_{ss}} \nonumber\\
e^{\tilde\phi} &=& e^\phi\left({\det G\over \det\twi G}\right)^{1/2} .\nonumber
\end{eqnarray}
As before, the quantity $\Sc{G}_{\mu\nu}$ is defined by:
$\Sc{G}_{\mu\nu} = G_{\mu\nu} + \nth{4} \, A_\mu A_\nu$, and the index $\mu$
runs over all values except for $\mu = s$. (Ref. \cite{AABL} presents the
duality transformations in a more manifestly covariant way.) These
transformations in general can acquire higher-derivative corrections in powers
of $\alpha'$ as well.
\subsection{An Example: The Torus}
Perhaps the simplest example brings out many of the main features of these
duality transformations.\footnote{See ref.~\cite{dualityreview} for references
to the extensive literature on toroidal duality.} The simplest example consists
of a toroidal spacetime, for which the metric is flat and all other background
fields are taken to be trivial. If we base the duality on the symmetry of
translations along one of the compact coordinate directions of the torus ---
call it $s$, say --- then the general expression of
eq.~\pref{dualitytransformation} simply reduces to the replacement:
\begin{equation}
\label{toruscase}
G_{ss} \to {1 \over G_{ss}}, \qquad \phi_0 \to \phi_0 + \log \Bigl| G_{ss}
\Bigr|.
\end{equation}
Here both $G_{ss}$ and $\phi_0$ are constants. The dual configuration is once
more a torus, but eq.~\pref{toruscase} states that if the circumference in the
$s$ direction is initially $R$, then for the dual theory it becomes $\twi{R}
\propto \alpha'^2/R$.
The beauty of the toroidal example is that for this background the complete
spectrum of string fluctuations is known, and the correspondence of states in
the dual theories can be explicitly followed. States turn out to be labelled
by two integers, $m$ and $n$, in addition to other quantum numbers, where $m$
labels the quantized momentum in the $s$ direction and $n$ gives the number of
times which the string winds around this direction. It turns out that duality
acts to interchange $m$ and $n$, while leaving all of the other quantum numbers
fixed. The important role which is played by the winding modes, for which $n
\neq 0$, emphasizes the intrinsically `stringy' nature of the duality symmetry.
When the direction associated with the symmetry coordinate is compact, such as
for the translations of a torus just considered, then the sigma models having
the dual field configurations can be shown to be completely equivalent, and so
describe {\em exactly} the same string physics \cite{algorithm}.
Eqs.~\pref{dualitytransformation} can therefore be considered as full quantum
symmetries of string theory.
The relation of the dual theories is less clear when the relevant symmetry
coordinate is not compact. We argue in what follows that the equivalence need
{\em not} be true in the noncompact case.
\section{Applications of Duality}
We now explore the implications of duality for general spherically-symmetric,
time-independent and asymptotically-flat field configurations.
\subsection{The Dilaton-Metric Configuration}
To get an idea for what is implied by a duality transformation, we first apply
one to the dilaton-metric configuration considered in \S 4.1, above. We
base the duality transformation on the symmetry of time translation of the
original solution. This solution, given by eq.~\pref{answer}, is characterized
by the two constants, $\delta$ and $\gamma$, together with the dimensional
quantity $\ell$. Applying the duality transformation,
eq.~\pref{dualitytransformation}, to this configuration leads to another
solution of the form as eq.~\pref{answer}, but with the constants $\delta$ and
$\gamma$ interchanged:
\begin{equation}
\label{dualityondilmet}
\delta \longleftrightarrow \gamma.
\end{equation}
Given the relation between these parameters and the conserved charges we see
that eq.~\pref{dualityondilmet} implies:
\begin{equation}
\label{dualityondilmetchg}
2 G_{\scrs N} \, M \longleftrightarrow Q_{\ss{D}}.
\end{equation}
As will now be discussed, this last way of writing the duality transformation
law applies equally well to the solutions at higher-orders in $\alpha'$.
\subsection{Application to General Configurations}
Since part of the promise of studying duality transformations lies in their
potential for leading to exact information concerning the theory involved, it
is of enormous interest to understand how duality acts on the exact string
solutions, rather than simply having its action on the approximate solutions
to low orders in the derivative expansion. The obvious difficulty lies in
determining this action when explicit expressions for the solutions, and the
transformation laws, are themselves not known beyond the leading order in
$\alpha'$.
It is nonetheless possible to draw general conclusions for the case of
time-independent, spherically-symmetric, asymptotically flat field
configurations, since these are in principle completely characterized by their
values for the conserved charges $M$, $N$, $Q_{\ss{D}}$, $Q_{\ss{A}}$,
$Q_{\ss{E}}$, and $Q_{\ss{M}}$. Furthermore, these charges are themselves
completely determined by the behaviour of the fields in the asymptotic,
large-$r$, regime for which all fields are very slowly-varying. As a result,
it is possible to determine the action of duality on the conserved charges of a
solution using just the leading expressions in the derivative expansion.
By applying the transformation law, eq.~\pref{dualitytransformation}, to the
asymptotic forms for the background fields, eqs.~\pref{gasymform} and
\pref{emaform}, the action of duality on the space of four-dimensional,
asymptotically-flat, time-independent and spherically-symmetric string solutions
may be determined in general. We find that the solution which is labelled by
the charges $M$, $N$, $Q_{\ss{D}}$, $Q_{\ss{A}}$, $Q_{\ss{E}}$, and $Q_{\ss{M}}$
becomes mapped to the solution whose charges are $\twi M$, $\twi N$,
$\twi Q_{\ss{D}}$, $\twi Q_{\ss{A}}$, $\twi Q_{\ss{E}}$, and $\twi Q_{\ss{M}}$,
where:
\begin{equation}
\label{dualsum}
2 G_{\scrs N} \, \twi{M} = Q_{\ss{D}}, \qquad \twi{Q}_{\ss{D}} = 2 G_{\scrs N} \, M, \qquad
\twi{Q}_{\ss{A}} = 2 N, \qquad 2 \twi{N} = Q_{\ss{A}},
\end{equation}
and
\begin{equation}
\label{dualnochange}
\twi{Q}_{\ss{E}} = Q_{\ss{E}}, \qquad \twi{Q}_{\ss{M}} = Q_{\ss{M}}.
\end{equation}
That is, duality simply interchanges the mass with the dilaton charge, as well
as interchanging the axion charge with the NUT parameter, while leaving
the electric and magnetic charges untouched.
\subsection{The Black Hole Special Case}
This transformation law has some odd consequences, as may be seen by focussing
on the black-hole solutions. At lowest order in $\alpha'$ these are characterized by
the parameters $\delta = 1$ and $\gamma = 0$, and so their image under duality
must have $\tilde\delta = 0$ and $\tilde\gamma = 1$. If we use the expressions
for $M$ and $Q_{\ss{D}}$, as given to $O(\alpha')$ in eqs.~\pref{bhasymform}, then
we find for the dual configuration:
\begin{eqnarray}
\label{smdasymform}
2 G_{\scrs N} \twi M &=& - \; {2 \lambda \alpha' \over \ell} + O(\alpha'^2) , \nonumber\\
\twi Q_{\ss{D}} &=& \ell + {11 \lambda \alpha' \over 6 \ell} + O(\alpha'^2) .
\end{eqnarray}
This transformation rule contains the seeds of a puzzle. Notice, in this regard,
the mapping from positive to negative mass:
\begin{equation}
\label{mdual}
\twi{M} = - \, {k \over \alpha' M} + O\!\left({1\over\alpha'^2M^3}\right),
\end{equation}
where $k = \lambda \alpha'^2 / 2 G_{\scrs N}^2 = 2\lambda e^{2\phi_0}$ is a dimensionless
constant. (In the special case of the superstring, for which $\lambda = 0$, it
is necessary to work to still higher order in $\alpha'$ to infer the sign of
$\twi{M}$, again giving the result \cite{bhdual} that $M$ and $\twi{M}$ have
opposite signs.) The puzzle is to understand how the physical equivalence of
dual solutions can be reconciled with a change of sign of the solution's mass.
The next section is devoted to the resolution of this puzzle.
\section{Equivalence Under Duality}
Intuitively, it seems absurd that a background field configuration having
negative mass could be physically equivalent to one for which the mass is
positive. The goal of this section is to pin down this intuition, in order to
better understand the circumstances under which dual spacetimes can be
equivalent.
A more precise reason for doubting the equivalence of spacetimes having opposite
masses would be the expectation that such spacetimes should gravitationally
scatter incident test particles differently. After all, based on experience
with the Newtonian force law for universal gravitation, a positive mass should
exert an attractive gravitational force on a distant, slowly-moving particle,
while the force due to a negative mass should be repulsive in this limit. One
might worry, however, whether the presence of other massless background fields
--- such as the dilaton or axion \etc\ --- may invalidate any intuition which
is based on the Newtonian limit of pure gravity.
\subsection{Test-Particle Scattering}
It is straightforward to test these ideas, by computing the scattering of test
particles which remain at large $r$, and so which experience only slowly varying
fields. This calculation may be performed explicitly in string theory by
choosing massless string states as test particles. The result is particularly
simple in the limit for which the wavelength of the incident state is both
longer than the Planck length, and shorter than the typical distances over which
the background fields vary. This is the regime of geometric optics, for which
simple arguments show that massless string states simply follow the null
geodesics of the metric, regardless of the presence of axion and dilaton fields
\cite{bhdual}.
For example, in this limit the scattering angle, $\Delta \varphi$, of a photon
which remains at large $r$ throughout its interaction with the background
field is given (for vanishing NUT parameter) by \cite{bhdual}
\begin{equation}
\label{simpang}
\Delta \varphi \roughlyup- {4 G_{\scrs N} \, M \over r_0} ,
\end{equation}
where $r_0$ is the smallest radial coordinate that is attained by the photon.
As advertised, this result depends only on the mass $M$ of the source, and not
on its other charges. The result for the dual spacetime is given by the same
expression, with $M \to \twi{M}$.\footnote{This assumes only that the dual
NUT parameter also vanishes, which implies the vanishing of the axion
charge, $Q_{\ss{A}}$, of the original solution. This class of configurations
is amply big for the present purposes.}
\subsection{A Better Argument}
Convincing as the previous calculation may seem, it leaves room for doubt
concerning the inequivalence of the dual field configurations we have
considered. This is because it leaves open a loophole, whose existence becomes
clear once the well-understood example of toroidal duality is reconsidered. As
was stated in \S 6.2, duality acts to interchange the labels $m$ and $n$
which respectively characterize the momentum and winding number along the
symmetry direction. It is the scattering of `winding modes' ($n \neq 0$) in the
dual theory which is equal to the scattering of `momentum modes' ($m \neq 0$)
in the original theory. The problem with the scattering calculation of \S 8.1
arises because it is not clear whether the photon in the original theory is
mapped to the same photon in the dual theory. If not there is no reason for a
result like eq.~\pref{simpang} to be the same for both the original spacetime
and its dual.
In fact, since the symmetry on which duality is based for the configurations
considered here is time translation, it is not completely clear precisely how
the winding and momentum modes should be defined. As a result, an alternative
argument in favour of the inequivalence of the dual solutions is now presented
which does not rely on being able to trace the detailed action of duality on
the various string states. This is done by identifying a one-parameter family
of equivalent background field configurations, and showing that these dualize
to a one-parameter family of inequivalent configurations. We are led to the
conclusion that eqs.~\pref{dualitytransformation} need not be a string
symmetry, at least when the coordinate `$s$' labels a noncompact direction.
The key point on which this argument relies is the observation that the
transformations of eqs.~\pref{dualitytransformation} depend in detail upon the
boundary condition which is satisfied at large $r$ by the gauge potential in the
time direction, $A_t$. In particular, the duality transformations obtained in
\S 7.2 for the conserved charges of the spherically-symmetric solutions are
only correct if it is assumed that $A_t$ falls off like $1 / r$. This is
required if the quantity $\Sc{G}_{tt} = G_{tt} + \nth{4} \, A_t A_t$ is to have
the same asymptotic form as does $G_{tt}$ --- a result which is required in
deriving eqs.~\pref{dualsum} and \pref{dualnochange}.
Suppose, then, we instead assume the following asymptotic behaviour for $A_t$:
\begin{equation}
\label{newgauge}
A_t=2v+{Q_{\ss{E}} \over r}+\cdots,
\end{equation}
where $v$ is a constant. This form does not affect the asymptotic behavior of
the electric and magnetic fields, and so does not change the values of the
electric and magnetic charges. In order to not change the asymptotic form for
the field strength, $H=dB-{1\over4}A\,dA$, it is also necessary to alter the
asymptotic behavior of the Kalb-Ramond field, $B$, which we assume to have done
in what follows.
When performing the duality transformation, eq.~\pref{dualitytransformation},
using the asymptotic form of eq.~\pref{newgauge} it is necessary to rescale the
time coordinate to preserve the limit $G_{tt} \to -1$ as $r \to \infty$, and to
shift the dilaton by a constant in order to recover its previous limiting value
at infinity. With these choices the action of duality on the asymptotic charges
is no longer given by eqs.~\pref{dualsum} and \pref{dualnochange}, but rather by
\cite{bhdual}:
\begin{eqnarray}
\label{longeq}
2{\twi{G_{\scrs N}}} \twi{M} &=& {1\over1-v^2} \left[ Q_{\ss{D}} -2v^2 G_{\scrs N} M -v
Q_{\ss{E}} \right] \nonumber\\
\twi{Q}_{\ss{D}} &=& {1\over1-v^2} \left[ 2G_{\scrs N} M -v^2 Q_{\ss{D}} +v
Q_{\ss{E}} \right] \nonumber\\
2\twi{N} &=& {1\over1-v^2} \left[ Q_{\ss{A}} - 2v^2 N \right] \\
\twi{Q}_{\ss{A}} &=& {1\over1-v^2} \left[ (1-2v^2)2N + v^2 Q_{\ss{A}} \right]
\nonumber\\
\twi{Q}_{\ss{E}} &=& {1\over1-v^2} \left[ (1+v^2) Q_{\ss{E}} +2v (2 G_{\scrs N}
M - Q_{\ss{D}} ) \right] \nonumber\\
\twi{Q}_{\ss{M}} &=& Q_{\ss{M}} + {v\over1-v^2} \left[ 2N - Q_{\ss{A}} \right]
\ \ . \nonumber
\end{eqnarray}
It is assumed in these equations that $v^2<1$. Notice that the previous results
are obtained as $v \to 0$. The `tilde' is written over Newton's constant in the
first line to emphasize that this constant is modified because of the shift
which was required of the dilation field in order to preserve its boundary
conditions at large $r$.
We are now in a position to argue against the necessity of physical equivalence
between dual solutions when the duality which relates the solutions is based on
a noncompact symmetry. The first step is the observation that the asymptotic
value, $v$, which is taken by $A_t$ drops out of all physical quantities
because it can be changed by performing a gauge transformation. The
one-parameter family of field configurations which differ only in their
asymptotic value for $A_t$ are therefore all physically equivalent to one
another. In particular, all would agree on their prediction for the scattering
of an incident massless string mode.
Now, if duality based on eqs.~\pref{dualitytransformation} were always to
generate physically equivalent string configurations, it would follow that the
one-parameter family of dual configurations which are obtained from the original
gauge-equivalent family must also all be physically equivalent. But
eqs.~\pref{longeq} show that the dual configurations have conserved charges
which depend explicitly on the value of the parameter, $v$. They therefore
differ in their physical predictions, such as for the scattering of massless
string modes, and so cannot be physically equivalent. We conclude that for the
class of configurations considered here, eqs.~\pref{dualitytransformation} are
{\em not} symmetries of string theory.
Notice that the compactness of the symmetry plays a role in this argument.
If the symmetry direction were compact, the above argument would fail since it
would no longer be possible to shift $A_t$ by an arbitrary constant value simply
by performing a gauge transformation. Different constant values for $A_t$ would
be physically distinct since they could be characterized by different values for
a gauge-invariant quantity: the Wilson lines around the compact direction.
\subsection{Flat Space Revisited}
The above considerations can be further focussed by reconsidering the case of
flat space and toroidal compactifications. We can avoid the complication of
understanding the meaning of winding and momentum modes in the time direction
by considering instead duality based on translations in one of the noncompact
spatial directions. It is instructive, in this case, to consider this direction
as the limit of a compact, circular, direction as its circumference tends to
infinity.
For any finite circumference, duality indeed represents an exact string
symmetry, with all momentum modes dualizing to winding modes in the symmetry
direction, and {\em vice versa}. For infinite circumference the momentum states
degenerate into a continuum of permitted eigenvalues, while the masses of all of
the winding modes tend to infinity. As required by duality, precisely the
opposite happens as the circumference tends to zero: winding states form a
continuum and momentum states become infinitely massive. So long as both
winding- and momentum- modes are both kept in the infinite (and zero-)
circumference limit, duality continues to relate equivalent theories.
But string theory, as it is normally defined on (noncompact) flat Minkowski
space, is the path integral over all possible string embeddings in this space.
This corresponds to what is obtained from a compact space in the
infinite-circumference limit {\em provided} that the winding modes are dropped
in this limit. Once the winding modes are omitted, the equivalence of dual
configurations is no longer guaranteed.
This picture is similar to that obtained by Ref.~\cite{AABL}, who analyze more
systematically the equivalence of solutions related by duality based on a
noncompact symmetry. They find equivalence, but only if one of the dual pair of
solutions is quantized as a vortex gas. In this unconventional string theory,
there are no local vertex operators carrying momentum in the symmetry direction,
but instead one constructs nonlocal `vortex' operators carrying
`winding-number' for this direction.
This line of argument indicates that a string theory is not uniquely specified
by listing its configuration of background fields, since additional choices ---
such as whether or not to include nonlocal vortex operators --- must also be
made. Although duality can always produce an equivalent string theory from any
given one, the two string theories generically will {\it not} both be
interpretable as the sum over string embeddings in the given background field
configurations. In comparing the dual theories, one must keep in mind this
distinction. For example, for two-dimensional black holes, the propagation and
interactions of tachyons near the singularity are equivalent to the propagation
and interactions near the horizon of some dual vortex states, rather than of
tachyons.
\section{Conclusions}
The purpose here has been to present some results of a preliminary investigation
\cite{bhsoln,bhdual} into the consequences of duality, and related,
transformations on our understanding of string propagation through
black-hole-like spacetimes. This has been accomplished by outlining two
qualitatively different types of results.
The first class of result simply uses these transformations to
generate new solutions to the string equations. This technique has been used
to generate the most general static, spherically-symmetric and
asymptotically-flat string configuration which solves the string equations to
lowest order in $\alpha'$. Solutions of this form are now known which involve all of
the metric, dilaton, axion and gauge fields --- a broader class than had
hitherto been constructed. They are labelled by a collection of five conserved
quantities: the mass $M$, and the dilaton, axion, electric and magnetic
charges: $Q_{\ss{D}}$, $Q_{\ss{A}}$, $Q_{\ss{E}}$ and $Q_{\ss{M}}$. If the
static condition is relaxed to include some stationary metrics, then a sixth
charge, the NUT parameter, $N$, is also required.
Using duality, and related, transformations to generate these solutions has the
practical advantage of requiring the use of only algebraic techniques, instead
of having to solve several coupled, nonlinear, partial-differential equations.
This makes its application relatively straightforward to apply to more
complicated, less symmetric, field configurations.
Most of the solutions which were generated in this way have real curvature
singularities at fixed, nonzero $r$, and these singularities are in some cases
naked. We regard it to be premature to discard these solutions until it is
better understood precisely what configurations string theory considers to be
singular.
The second type of investigation reported here concerns the action and
implications of duality on these low-energy solutions in particular, and on
static, spherically-symmetric and asymptotically-flat string configurations in
general.
The result can be presented in a very general form. By investigating the action
of duality on the asymptotic form of a general field configuration for large
$r$, it becomes clear that duality simply interchanges the configuration's mass
and dilaton charge, as well as interchanging the axion charge with the NUT
parameter. The electric and magnetic charges do not change. These results are
very robust --- applying equally well once higher orders in $\alpha'$ are included
--- depending as they do only on the large-$r$ behaviour of the solution.
Somewhat surprisingly we find that the dual field configuration depends in an
important way on the assumed asymptotic form for the electrostatic potential,
$A_t$. The conclusions of the previous paragraph assumed $A_t$ to be chosen
to vanish asymptotically.
For the special case of the four-dimensional `Schwarzschild-like'
solutions --- {\it i.e.}\/\ those that are nonsingular at the Schwarzschild radius ---
a black hole of mass $M$ is mapped by duality onto a singular configuration
whose mass, $\twi M$, is given by $\twi{M} = - k/(\alpha' M)$, where $k = \lambda
\alpha'^2 / 2 G_{\scrs N}^2 = 2\lambda e^{2\phi_0}$. $\lambda$ here is $\frac{1}{2}$ in the bosonic,
and $\nth{4}$ in the heterotic string. For the superstring, it happens that
$\lambda = k = 0$, and so $O(\alpha'^3)$ corrections are required in order to
determine the dual mass. The result in this case turns out to be: $\twi{M} = -
\, k' / ( \alpha'^3 M^5)$, where $k' = 3 \zeta(3) \, e^{6 \phi_0}$.
We have also addressed the potential inequivalence of the two dual solutions in
the case where the symmetry on which duality is based is not compact. An argument
was presented which led to one of the following two options: ($i$) dual
solutions, defined by eqs.~\pref{dualitytransformation}, can be physically
inequivalent; or ($ii$) constant shifts of the electrostatic potential, $A_t$,
can change the physical content of the theory. Arguments in favour of the first
option were presented, using the well-understood example of flat space.
The potential implications of duality are just beginning to be explored, both
for black holes, and for other physical systems. May the rewards be both rich
and varied!
\section{Acknowledgements}
C.B. would like to thank the organizers for their kind invitation to speak at
the conference. Our funds have been provided by N.S.E.R.C.\ of Canada, les Fonds
F.C.A.R.\ du Qu\'ebec and the Swiss National Foundation.
\def\pr#1#2#3{{\it Phys.~Rev.} {\bf #1} (19#2) #3}
\def\np#1#2#3{{\it Nucl.~Phys.} {\bf #1} (19#2) #3}
\def\pl#1#2#3{{\it Phys.~Lett.} {\bf #1} (19#2) #3}
\def\prc#1#2#3{{\it Phys.~Rev.} {\bf C#1} (19#2) #3}
\def\prd#1#2#3{{\it Phys.~Rev.} {\bf D#1} (19#2) #3}
\def\prl#1#2#3{{\it Phys. Rev. Lett.} {\bf #1} (19#2) #3}
\def\plb#1#2#3{{\it Phys. Lett.} {\bf B#1} (19#2) #3}
\def\npb#1#2#3{{\it Nucl. Phys.} {\bf B#1} (19#2) #3}
\def\mpla#1#2#3{{\it Mod.\ Phys.\ Lett.} {\bf A#1}, (19#2) #3}
\def{\it et.al. \/}{{\it et.al. \/}}
\bibliographystyle{unsrt}
|
\section{SYSTEM CHARACTERISTICS}
During the earliest discussions of the feasibility of the MACHO
project the team recognized that the software would be one of the two
most challenging system components (the other being the ccd camera).
Before the software design process began, we set ourselves a number of
goals for the system. We discuss them first, as a prelude to
the following sections, which cover the architecture and implementation
of the system.
\subsection{Functionality}
The scientific goals of the MACHO project \markcite{(Alcock 1992, 1993)} require that
the system operate every clear nighttime hour with high reliability.
The duration of the data collection phase was initially set as 4
years. We recognized that the data must be processed, as
well as collected, in near real-time. Otherwise, the project was
likely to die from data indigestion. Additionally, there is significant
scientific gain from being able to detect microlensing events while they
are still in progress. We decided at the outset that
these requirements could be met only by a system which encompassed in
an integrated framework all the separate tasks required for operation
of the experiment, and one which operated in a highly automated way
with minimal human intervention. The main tasks required of the system
are:
1. Scheduling of observations
2. Control of the telescope
3. Operation of the camera system
4. Flatfielding of the ccd images
5. Archiving the images to tape
6. Preparing the images for photometry
7. Photometry
8. Storing all information associated with each observation in a
data base
9. Interacting with the human operator
Since the Macho telescope is dedicated to a single experiment, this
software can run continously, and can be highly specialized.
Typical exposure times are 150 to 600 sec, with times as short as a few sec
sometimes being employed. This time is short enough that the time
required to move the telescope and to read out the ccd camera
system is significant. To maintain the maximum possible rate of
observations, it is therefore crucial that the various system tasks
be overlapped as much as possible (e.g move the telescope at the
same time as reading out the camera; flat field the preceding image
while exposing the next).
Additionally, given the very large number of pixels in the camera system
and the high star densities that are typical of our fields, the photometry
process is very cpu intensive. The use of multiple cpu's is absolutely
essential to process data at the rate it is taken.
\subsection{Hardware Configuration}
The hardware configuration, and the underlying operating system
software, has evolved significantly since the beginning of the
project. At the outset, the main computer was a Solbourne 5/804, a 4
processor Unix system running a version of SunOS 4.1. Today, the main
computer is a Sun SparcServer 1000 with 8 cpu's and 0.5GB of memory,
running Solaris 2.4, supported by a secondary SparcServer 1000 with 4
cpu's. Online disk storage has grown from an initial 2GB to
approximately 150GB today, and is still growing rapidly. The current
hardware configuration is shown in Fig. 1.
\subsection{Need for Robustness}
Given the complexity of the system, a high priority was given
to achieving robustness in an environment where a wide variety of
error conditions can occur, and occasional software ``crashes'' are
inevitable. Several sub-goals were identified that together would
contribute to the desired robustness:
\subsubsection{Persistent State}
The system inevitably has a large amount of ``state'' associated with
it. Examples of state variables include the current observation
number, the status of an observation in the observe cycle, and the
information needed to identify and communicate with the processes
that make up the system. We recognized early that making the state
persistent would be a big contributor to robustness. Our meaning of
persistence is simply that information is preserved when a process
dies, rather than disappearing with the process. Persistent state
allows a replacement process to be started and take its place in the
system without special action to tell it what it needs to know.
\subsubsection{Central Heartbeat Monitoring}
A failure mode which is particularly troublesome in systems of this
nature occurs when a process enters a ``hung'' state in which it
continues to exist but is no longer doing its job. When this occurs,
gridlock often ensues as other processes wait for the hung process to
do something useful. A human operator, faced with such gridlock,
seldom has the information necessary to determine the cause and
generally has no option but to restart the entire system.
We decided that many such failures could be recognized if every
member process was endowed with an externally visible ``heartbeat''.
One process could then be assigned the task of monitoring the
heartbeats of all component processes. Disappearance of the
heartbeat would occur not only when the process itself disappears
(i.e. ``crashes''), but also in many cases where the process ``hangs''.
Heartbeat existence is useful both for automated recovery
procedures and for the human operator, who can now see which
process is hung and take action less drastic than a global restart.
\subsubsection{Capability to post and respond to alarms}
When an error condition is detected it is important to have a variety
of options for responding to it. The response can range from an
automatic ``reflex'' on the part of the process which recognizes the
error, such as ``retry the operation'', to one which alerts the human
operator and gives him the opportunity to respond in an informed
way. We decided that it was crucial to have a universal mechanism
for posting alarm conditions and allowing a variety of responses.
\section{SYSTEM ARCHITECTURE}
\subsection{Observation Cycle}
The architecture we chose has as its central feature the observation
cycle, and the mapping of that cycle to communicating Unix processes.
The observation cycle, shown schematically in Fig. 2, is a fixed set
of operations that are performed in the same order each time a new
observation is taken. Additionally, we make use of the same
observation cycle for ``reobserving'', which is our term for performing
the data reduction steps on a previously taken observation. This
situation comes up frequently, usually because photometry was not
performed at the time of exposure, or because some problem requires it
to be repeated. During reobserving, all stages of the observing
pipeline are active, but some of them (e.g. Telescope and Camera) have
little to do.
A key concept in the observation cycle is the observation descriptor.
Observation descriptors are used by all phases of the observation cycle
and completely describe the status of an observation and the parameters
associated with it as it moves through the pipeline. When an
observation is completed, its descriptor is saved permanently in a
data base. Table 1 shows the contents of a representative observation descriptor,
as printed out by a viewer utility.
The observation descriptor has proved to be an important unifying data
structure, and it has expanded several times in the course of the
project when it became clear that it would be valuable to include more
data (for example, the inclusion of telescope temperature data
was a recent addition). Its uniform use throughout all phases of the
observation cycle has considerably simplified both the initial
implementation and subsequent changes to the observation cycle. It has
also made it easy for the progress of an observation to be monitored
from the user interface (UI).
Here is a brief description of each operation in the observation cycle:
\subsubsection{Scheduler}
The scheduler gathers the information needed to specify the next observation,
builds its observation descriptor, and a creates a number of associated
directories. The required information includes the field number (if it is
a prespecified Macho field, otherwise the pointing coordinates), the exposure
time, and whether or not photometry is to be performed. The information
can come either directly from the observer through the user interface (UI),
from an ascii file, or from an external source that places observation
requests on the scheduler's work queue. Which source is active is determined
by the observer through the UI.
\subsubsection{Slotter}
Every new observation requires a number of records in the data base structure
to be created and initialized for it. This is the main role of the slotter.
\subsubsection{Telescope}
The telescope drive is controlled by a dedicated VAX/VMS computer that is not
itself part of the Macho system \markcite{(Hart 1995)}. The main function of the telescope
process is to issue the proper commands to the telescope computer so that it
points and tracks at the coordinates required for the observation.
The telescope coordinates are continuously monitored to ensure that the
correct pointing is in fact achieved. The telescope process maintains on
the bulletin board (see Section 3.1.3) a current descriptor of the telescope state
for use by other processes. Additionally, the telescope process has the
capability to automatically set the telescope focus. This feature is not
currently in use, because it has proved difficult to create a satisfactory
focus model - but we think we are now close.
\subsubsection{Camera}
The camera process issues commands to the camera controller (a micro
computer running Forth) to properly initialize the camera, take the
exposure, and read out the image \markcite{(Stubbs 1993)}. It then
makes use of a separate server process which flatfields the image,
and corrects for crosstalk between ccd channels \markcite{(Stubbs 1995)},
and finally stores the resulting images on disk as FITS files.
\subsubsection{Image Archiver}
The archiver writes the flatfielded images to Exabyte tape for archival
storage. It is a simple procedure, but must nonetheless contend with a
variety of error conditions (I/O errors, end of tape, etc) in a robust
way.
\subsubsection{Slicer}
All data reductions are performed not on whole ccd images, which at 2K x 2K are
too large for convenience, but on ``chunks'', which are roughly 500 pixels square.
These chunks are overlapping, so that unnecesssary edge effects are not
introduced. The slicer is responsible for creating the chunk images, creating
for each chunk a data structure that binds together all data necessary for processing it,
and mailing it off to the photometry work queue.
\subsubsection{Photometry}
The photometry process is responsible for running the SoDoPhoT \markcite{(Bennett 1995)} photometry
package on each of the image chunks. A data structure is created for each chunk
that binds together all resulting photometry data for the chunk. This data structure
is then mailed off to the collector work queue. A typical chunk contains 5000
stars and takes 20 sec to process.
\subsubsection{Collector}
The collector is the final stage of the observation cycle. The photometry data for
each chunk is loaded into the photometry data base, where it is then available for
later analysis. Database loading rates are typically 5000 stars/sec, so that
a chunk requires roughly one sec to process.
\subsubsection{User Interface}
The principal functions of the UI are to :
1. Allow the observer to initiate, pause and stop the observation cycle,
and specify parameters that affect it, such as exposure times.
2. Present process status (heartbeats) in a graphical form.
3. Present alarms for action by the observer.
4. Allow the observer to assess the progress of any observation through
the pipeline and intervene if necessary.
\subsubsection{Controller}
The controller process coordinates the overall functioning of the system.
It handles the startup of all component processes, monitors their health,
and restarts failed processes
when required. It also contains the central sequencing logic for the observation
cycle and the maintenance cycle (discussed below).
\subsection{Parallel Architecture}
The component operations of the observation cycle form a data
pipeline, with the output of one operation being the input to the next.
As with any pipeline, under conditions of continuous operation
higher throughput may be achieved by assigning each pipeline stage
to its own hardware, so that it operates in parallel with the others.
Because of the unchanging structure of the observation cycle, the ``macro''
nature of each of the pipeline stages, and the desirability of having
the stages handled by different processors, we decided to map each
stage to a Unix process. In the case of the photometry stage, which is
by far the most compute intensive phase of the observation cycle, we
assign a group of processes (typically 4 or 5) to the function, further increasing
parallelism. This could readily be done for other compute intensive
stages of the pipeline if their throughput ever becomes an issue.
The MACHO system makes explicit two levels of parallelism. The first
level, as mentioned above, is that between the Unix processes that
cooperate to form the observation pipeline. Within each of these
processes there is a second level of parallelism, which is present to
deal with the demands of handling asynchronous events such as the
arrival of messages and the expiration of timers.
The motivation for the process level, or coarse grain, parallelism is
principally the need to exploit multiple CPU's for increased
performance. Secondarily, the multiple process structure is desirable
from a software engineering point of view, because the process
boundaries enforce a useful modularity on the system, and eliminate
many opportunities for unforseen interaction between components.
The principal motivation for the second level of parallelism, which we
will call the fine grain level, is quite different. The problems of
creating reliable real-time control systems that handle multiple
classes of asynchronous events are well known \markcite{(Everett
1995)}. Avoiding deadlocks and other forms of ``hangs'', and ensuring
that all incoming events are processed in a fair way are all difficult
in a sequential program. On the other hand, a programming method that
uses parallel threads of control can solve many of these problems,
especially if it is practical to spawn a new thread to process each
incoming event.
Secondarily, multiple threads have allowed major performance improvements
to be made in I/O intensive processes, such as the collector.
\subsection{Petri Nets: Unifying Coarse- and Fine-grained parallelism}
One innovation of the Macho system is that it ties together the coarse-
and fine-grain levels of parallelism into a single unified view of the
system. This is valuable both for system understanding, and for
efficient implementation of the system. This unified view is based on
Petri nets, which we now briefly discuss.
Petri nets are a well known methodology for designing and, less
commonly, for implementing parallel systems \markcite{(Murata 1989)}.
In its most basic form a Petri net consists of a set of places; a set
of transitions; and a set of tokens (Fig. 3). The execution rules for
the net are:
1. A transition ``fires'' when all of its input places are marked
with tokens.
2. When a transition ``fires'', it marks each of its output
places with a token.
This structure allows a surprising variety of parallel, sequential, and hybrid
systems to be modeled. However, in this simple form, a Petri net lacks a number
of features that are required in a real system implementation:
1. There is no way of performing ``useful work'', e.g. performing a
computation or writing a file.
2. There is no conditional behavior
3. There is no way of expressing timeout conditions.
Petri nets have been extended in many different ways to overcome these,
and other, difficulties. Our extensions are inspired by \markcite{(DiStefano
1991)}, one of the few examples of using Petri nets for the actual
implementation of a real time system, and consist of the following:
1. A transition can optionally have an associated function that
is executed whenever the transition fires. Tokens are allowed
to carry optional pointers to arbitrary data which the
transition function can utilize as input and output arguments.
2. Conditional transitions are defined, which mark either a
``true'' or a ``false'' output place, based on the return value of
an associated function.
3. Timeout transitions are defined, which always have two
input, and two output places. Marking one of the input places
starts the timer, while marking the other stops it. If the
timer runs for longer than a preset timeout interval, the first
output place is marked, while if the timer is stopped before
the interval expires, the second output place is marked. This
definition is identical to that used in \markcite{(DiStefano 1991)}.
\subsection{Petri Nets that cross process boundaries}
Now we must take a step closer to implementation, and consider Petri
nets in the context of Unix processes executing in parallel. It is
here that we are able to unify the coarse- and fine-grained levels of
parallelism.
It is natural to express each stage of the observation cycle as an
extended Petri net. These nets are connected because the
pipeline stages must communicate. If we have a mechanism to allow
tokens to cross process boundaries, we can then use this to connect the
nets, and thereby end up with a Petri net description of the entire
system that spans process boundaries in a transparent way. A sketch of
this concept is shown in Fig. 4.
\subsection{Database}
The operation of the data pipeline requires a large amount of input
data in addition to the images taken by the camera system. This data
includes flat and dark images for the camera system, the definitions of
field centers, photometric reduction parameters, locations of fiducial stars, and
a great deal more. Additionally, the pipeline generates a large amount of
output data for every observation. The largest volume of output data
consists of images and star photometry, but there is much else,
including log files, temperature data, and so on.
The data base system is designed to organize all this input and output
data in a manner that supports a variety of analysis tasks. It is
essential that the data base preserve the history behind each observation:
full details of the environmental conditions when the exposure was taken,
the operations that were performed on the data, and the identity of the
software that performed them.
The ramifications of this requirement can be quite extensive. As an
example, several of the focalplane ccd's were recently replaced. The
replacements have somewhat different orientations than their predecessors,
as well as different darks, flats, readout noise and crosstalk parameters.
The data base must keep track of these changes, and make sure that the proper ccd
data is associated with each observation in the data base, and that this data
is transparently routed to any analysis programs that require it.
The data base defines a few fundamental geometrical data structures:
1. A Field is the nominal portion of the sky captured on the ccd
during an exposure. Fields may overlap. The geometry of a field is
considered inviolate once it is ingested into the Field data base.
Fields are currently 43 arc minutes on a side, and frequently contain
500,000 stars or more.
2. A Chunk is a portion of ccd image, typically of order 500 pixels on
a side. Its corresponding position in the sky is known only to an
approximation. Chunks overlap in order to avoid needless photometry
errors near chunk boundaries. The geometry of a chunk may change over
the life of the project.
3. A Domain is a bounded region in the sky which totally encompasses a
pre-defined set of fields. A Domain is divided into an evenly spaced
rectangular grid. Each grid region is known as a Tile. A Tile is
typically 4 arc minutes on a side, and contains several thousand
stars. The Domain geometry and its grid definition are considered
inviolate once they are ingested into the Domain data base.
4. A Template is a set of stars from a given Field which is fabricated
to completely cover a ccd Chunk. A Template may contain stars from
multiple Tiles. The geometry of a Template is tied to the geometry of
a Chunk and may change over the life of the project. Also, the star set
may be modified without changing the geometry.
5. An Image is an aggregate output of 4 red- and 4 blue-band ccd's.
The ordering and color assignments of ccd's may change over the life of
the project.
The relationship of these structures is shown in Fig. 5.
The data base system has two levels of organization. The primary data base files contain
some particular record type spanning some range of time, for example,
field boundary definitions or ccd characteristics. The second tier collects
the timestamped primary data bases into a set containing the full span of time varying
characteristics of observations and their photometry.
The most important data base files are as follows:
1. The Domain data base defines Domain boundaries and tiling size. The tile size
is selected so each Tile contains an average of 5 thousand stars.
2. The Field data base defines the Telescope pointing coordinates.
3. The FieldChunk data base defines the template geometry. Since its
data may change over the course of the project, it is tagged so that the
version corresponding to any observation is automatically retrieved.
4. A Star data base encompasses the coordinates of all fiducial and observed
stars within a single Tile's boundary.
5. The CCD data base defines the focal plane characteristics. Since its
data may change over the course of the project, it is tagged.
6. The CDF data base contains crosstalk, dark, and flatfield information for
the focalplane. Like the CCD data base, it is tagged.
7. The ObsState data base defines the environmental state of an observation and
the status of its subsequent photometry reduction.
8. The SodObs data base contains the photometry reduced by SoDoPhot for a single
Tile's stars.
9. The Tile data base defines the Tile boundaries. Although the information is
derivable from data within the Domain data base, it is expanded
in order to speed photometry reduction and analysis.
10. The TileHit data base defines every observation that overlaps a single Tile's
boundary. Although the information is available in the ObsState data base,
it is expanded in order to speed photometry reduction.
\section{SYSTEM IMPLEMENTATION}
We now turn to some of the details of the actual system implementation.
The process-spanning Petri nets that are used to implement the Macho
system are based on a lower level set of interprocess communication
mechanisms. These are discussed in Section 3.1. The Petri nets
themselves are implemented as C++ classes that utilize the Sun LWP
library (in the SunOS implementation) or the Sun Threads library (in
the Solaris implementation). This is discussed in Section 3.2. The
maintenance cycle, which implements the heartbeat monitoring function
discussed in Section 1, is described in Section 3.3. Section 3.4
describes the functioning of alarms, while 3.5 covers the data base
implementation, and 3.6 the user interface.
\subsection{Interprocess Communication Mechanisms}
Interprocess communication is handled by three mechanisms (Fig. 6):
mailboxes, work queues, and the bulletin board. Each of these is a
persistent object in the sense described in Section 1.3.1, and is implemented
as a memory mapped file.
\subsubsection{Mailboxes}
Each process has a single mailbox to which any other process in the
system can send messages. A message consists of a header used for
routing purposes, followed by an arbitrarily long string of bytes. The
byte string can be interpreted by a process in whatever way it
wishes, for example as a C structure. Messages are used to request a
service from a process, to notify a process that some condition has
occurred, or to request status information from a process. A process
is notified by the mail delivery system whenever a message arrives.
\subsubsection{Work queues}
Work queues are similar to mailboxes but differ in that they are
shared by a group of recipient processes instead of being exclusively
owned by a single process. They are used, as the name implies, to
deliver work to a group of server processes. The first to become free
will remove the item from the work queue and perform the
requested task.
\subsubsection{Bulletin board}
In addition to the pipeline style of information flow supported by
mailboxes and work queues, it is very useful to have a globally
shared data structure that can be used to contain information that is
of general use. The bulletin board serves this function. An item can
be posted to the bulletin board by any process, and read by any
process. Items are identified by name and have an associated time
tag so that customer processes know how current the information is
that they are getting. Items can be of arbitrary length, and as with
messages, are usually interpreted as a C structure by customer
processes.
The bulletin board is used to post items such as the number of the
next observation, the current state of the telescope, temperatures
from a variety of sensors, and observation descriptors for all
observations that currently inhabit the pipeline.
\subsection{Petri net}
The Petri nets are built on two C++ base classes, Place and
Transition. Specialized types of Transitions, the TimeoutTransition
and ConditionalTransition, are derived from the Transition class by
inheritance. Additional container classes, PlaceSet, TransitionSet,
Net, and SubNet, are used to bundle together related groups of Places
and Transitions, to define complete Petri nets, and subsets of Petri nets,
respectively. The Tokens that mark the Places of the Net are in
fact pointers to Messages, those objects that can be mailed from
process to process via the mail system. If a Token does not need to
have any associated information, as is frequently the case, a null
pointer is used.
Places come in three flavors, Normal, Input, and Output. Input Places
are marked by the arrival of a Message from another process addressed
to them. The marking of an Output Place causes a Message to be
delivered to some Input Place in another process. Normal Places do not
involve interprocess communication. Each Place has an associated FIFO queue
for Tokens. This modification to the basic Petri net behavior provides
for some buffering between Transitions, and improves performance.
The parallelism of the Petri nets is achieved through the use of the
Solaris Threads library \markcite{(SunSoft 1994)}, which allows applications to make use of
large numbers of parallel threads of control with low resource expenditure.
Each Transition has an associated thread, which implements the Petri
net firing rules and invokes any associated function when the firing
rules are satisfied. At any moment the majority of these threads
are suspended, waiting for the input Places associated with their
Transition to be marked.
Associated with each Net is a single thread which processes Messages
arriving in the Net's Mailbox, and then distributes each one to the
Place for which it is bound. The arrival of such a Message marks the
associated Input Place, and therefore may cause the associated Transition
to fire. Similarly, marking of an Output Place spawns a thread
which is responsible for sending the mail that will the deliver the
associated Message to its destination Input Place.
\subsection{Maintenance cycle}
In addition to the observation cycle, all processes participate in the
maintenance cycle. This cycle, which executes in parallel with the
observation cycle, implements the heartbeat monitoring function.
Any process can monitor the health of another by periodically
sending a ``status request'' message to its mailbox. If a ``status reply''
message is not received within a fixed timeout period, the
heartbeat of that process is flagged as stopped. The ``status reply''
message also can contain detailed information about the internal
state of the process. Currently, both the UI and the Controller
processes perform heartbeat monitoring functions.
\subsection{Alarms}
Any process that encounters an error condition, or any condition that
may require human intervantion, is free to post an alarm. Alarms have
two types: a ``Show'' alarm merely informs the operator, without requiring
a response, while a ``Resolve'' alarm requires an operator response. As
examples of the two types, a Show alarm is sent on software startup to
inform the observer that the telescope is disconnected and must be
reconnected if he wishes to take a new field, while a Resolve alarm is sent
if disk space runs out and a file cannot be created. The latter requires
an operator response because an operation attempted by the software (file
create) has failed, and further progress cannot be made until the
cause is corrected.
The response to a Resolve alarm is made by picking one of three choices
presented by the UI. These are ``Retry'', ``Continue'', or ``Exit''.
Retry indicates the process should retry the failed operation, usually
after the operator has intervened in some way (e.g. freed some disk
space). Continue indicates the process should simply continue
processing without attempt to handle the error, while Exit indicates
the process should terminate itself. This latter option, which is
very rarely used, will then lead to another alarm when the Controller
realizes that the process has died, and thereby an opportunity to
start a new copy of the process.
Alarms are implemented using the mail system. When a process posts an
alarm, mail is sent to the UI containing the text of an alarm message
and the identity of the poster. If the type of the alarm is Resolve
mail containing the choice made by the operator is returned to the
poster, who then proceeds accordingly.
\subsection{Database}
In order to implement the Macho data base subsystem in minimal time,
we initially sought a commercial product. We
quickly restricted our search to object oriented data base products, which
more readily accomodate time history data ingest, manipulation, and retrieval
rather than do relational data bases.
We found the few existing (circa 1991) object oriented data base products
to be too ``leading edge'' for production use. We then reluctantly embarked
on the development of a handcrafted data base system, the architecture of
which was sketched in Section 2.5.
The data base is not a monolithic object, but rather a large collection
of data base files organized by a multi-level directory tree. The
majority of the data is maintained in specially constructed files which
are tuned for the data content and its ingest and retrieval patterns.
The data bases are implemented using a C++ library based on GNU C++
Associative Maps \markcite{(Lea 1992)}. The base data base class
provides functions common to all data base types (open, close, create,
etc). It also defines alternate access methods - either random disk
access or direct memory mapping.
Each specific data base class additionally defines the ingest, retrieval and
status functions particular to the data. For example, the segmentation and
chaining of the time-sequenced photometry data base files is derived
from the base class and is tuned for efficiency of photometry installation
and extraction. The data base classes, which provide a uniform
interface, greatly simplify implementation of new data base types.
The Database system now manages more than 90,000 files, a number expected
to grow to more than 200,000 as data collection continues.
\subsection{User Interface}
The UI receives and processes information from the pipeline in two ways.
Firstly, it polls the bulletin board at regular intervals (normally every 5
seconds) using XView notifier timer functions. This allows such items as the
status of all processes and the contents of the pipeline (number and state of
observation descriptors) to be updated on displays. Secondly, the UI sends
and receives mail messages. These messages contain incoming alarms which are
posted for action by the observer or outgoing requests for action by the
controller or other component in the pipeline. Outgoing mail could be either
observer responses to alarm conditions or intervention action like the
removal of an invalid observation descriptor from the pipeline.
On startup, the UI is presented as a root window with icons for each pipeline
component. The icons are panel buttons that reveal customised sub-windows.
Graphics, colour and animation are used widely in all sub-windows to
enhance the observers ability to quickly assess the state of a process or to
be alerted if action is required.
The UI is written in C and uses the XView 3.0 toolkit \markcite{(Heller 1991)}.
\section{EXPERIENCE WITH THE SYSTEM}
We have now had over two years of experience with the system described
here, during which over 30000 observations have been taken and nearly 3 terabytes of
data collected. For the most part, it has fully met our expectations and continues
to fulfill the tasks originally set for it. The best features of the
system have been:
1. Rapid implementation
Design of the system began in July of 1991. Stable operation of the
first version of the system was achieved in September of 1992 with
an expenditure of about 2.5 man-years. This version lacked many
capabilities that were added later, such as flatfielding of images
and online photometry, but it was sufficient to collect data for
analysis.
2. Ease of modification
The system has been continually modified since it first became
operational, both in incremental, and occasionally in more major
ways. The incremental modifications have generally been to add
new features or correct bugs, while the major changes have been
associated with changes in the underlying hardware and OS support.
For the most part, these changes have been surprisingly easy to
make. Most importantly, none of them have required significant
change to the underlying Petri net and data base concepts that the system is
based on. This stability, combined with the extensibility of
C++, has made the system easy to live with in the real world.
3. Robustness
The system has turned out to be quite robust to a variety of
failures, as we originally intended. The original
multiprocessor it was implemented on suffered many system
crashes, and the use of persistent state in the observing
software allowed it to be restarted after a system crash with
minimal loss of time or data. Similarly, the system recovers
well from crashes caused by its own bugs, with the ability to
easily restart a component process generally making recovery
fairly painless. The use of alarms has also turned out well,
providing a simple but flexible way for the operator to
usefully respond to the exceptional conditions that always
arise in the real world.
4. Performance
The performance of the system has increased continuously, as
more powerful hardware has been added and software has been
improved. Today the pipeline meets all of the performance
goals that we set for it. First, the interval between
successive images is within a few seconds of the minimum set by
exposure time + ccd readout time. This maximizes the rate at
which raw data is taken.
Secondly, photometric reductions and data base loading happen
nearly as quickly as images are taken. This has been critical
to the project's ability to ``alert'' on microlensing events
very shortly after they begin. Currently, data from a full
winter night's observing that ends at 6 am are typically fully
processed by 8 am when using 4 cpu's for photometry. Templates
have been made for only about half of our fields, however.
Although our current templates include all of the densest
fields, and therefore represent more than half of the total
stars, a full set of templates will significantly increase the
processing load. When we have the full set, we will need to
use 7 or 8 cpu's for photometry (out of the 12 available) to
maintain the same performance.
Finally, data base retrieve performance is good, and adequately
supports analysis of lightcurves.
The problem areas have been:
1. Bifurcation of the system
As stated above, the initial version of the system supported
neither flatfielding nor photometry. With data pouring in,
there was a powerful motivation to complete the analysis path
as quickly as possible. There was a division of opinion as to
whether this should be done by completing the data pipeline, or
by creating a separate ad-hoc ``offline'' system fed from the
existing ``online'' data pipe to perform the analysis. In the
event, the development of the system bifurcated, with an
offline system being created while data pipeline development
continued at a slowed rate.
Today, both systems are complete and are in active use. The
offline system performed the photometry and analysis used in
all of the project's publications to date. The online system
is currently used for all new photometric reductions, is the
only repository for recent photometry, and provides the only
capability for alerts. Older photometric data is partially
duplicated between the two systems, with the online system
expected to have a complete set of the project's photometry
within the next six months.
Although our plan is that the online system will ultimately
be the basis for all the project's analysis, this is not an
easy transition to make. Each is the result of a large effort,
and has a group of committed users. It is likely that we
will be dealing with the effects of the bifurcation for
the life of the project.
2. Difficulty of text-based Petri net specification
Petri nets are most naturally specified and thought about
graphically. Our software, however, requires a textual
specification of the net which is directly input to the C++
compiler. This has caused a continual, and costly, mismatch
between the human specification and the machine specification.
It has been the most fertile source of bugs, which have
frequently arisen because the text specification was an
improper translation of the graphical object that the designer
had in mind.
3. Difficulty of debugging
Debugging parallel, real-time applications is difficult at the
best of times. In addition to this generic problem, some
specific difficulties made our lives difficult. One of these,
of course, is the problem of rapidly and reliably translating
back and forth from the text-based Petri net specification
required by the compilers to the graphical notation required by
human understanding, as mentioned above. Additionally, most of
the life of the system has been under SunOS 4.1, where the
Petri nets were built on top of the LWP library. This library
had many bugs and little support, and therefore required much
investment of time to find workarounds. This problem has gone
away with the switch to Solaris and the Threads library.
Similarly, under LWP there was no debugger support of any kind,
which made it necessary to debug largely with print statements!
The debugger under Threads has sophisticated built-in support
for threads, and is making life much easier.
4. Database issues
Although the hand-crafted data base system is efficient, robust,
and reliable, its greatest strength - being a no frills data
entry and retrieval engine - is also its greatest liability.
Database record definitions are difficult to alter. Definition
changes require either retrofitting the extant data base records
or using versioning to determine, on the fly, the appropriate
record format to use. To date, we have opted for the former.
An early decision was taken to physically load photometry
records in time-ordered sequence in the anticipation that this
would greatly enhance data retrieval speed. This subverted the
classic data base management premise that a record is extracted
by its selection key, and has caused a number of difficulties.
In the event, all our current analysis tools access records via
key.
5. User interface and XView
The high-level nature of the XView calls (as compared to raw
Xlib or Motif) allowed the UI to be developed quickly.
However, XView has proven to be a poor choice for
a UI that relies on timers and interrupts. Incoming mail in the
SunOS implementation of the pipeline was signalled to the UI
using the standard UNIX signal SIGUSR1. This required an
asynchronous interrupt to the notifier loop that often failed
and caused the UI to crash.
Furthermore, this situation is exacerbated by the presence of
multiple notifier timer functions that periodically read the
bulletin board and run animation graphics. XView under Solaris
seems to have similar notifier interaction problems. A robust
UI in an interrupt driven and polling application like the
pipeline will probably require a move to a toolkit better
suited to realtime applications.
6. Stress on system software
We rapidly discovered that this system works many features
of the operating system software quite hard. In addition to
the persistent LWP problems mentioned above, there were
many early problems due to our intensive use of memory
mapping and network locking of memory mapped files. Unreliable
signal delivery has also been a major problem. These
problems have generally gotten better, but finding workarounds
has been time consuming.
\section{FUTURE DEVELOPMENT}
Looking to the future, we expect to operate the system in a fully
distributed mode, where member processes can exist on fully separate
machines (today they can run on multiple processors, but all on the
same machine). Although the underlying process-spanning Petri net
model fits perfectly into a fully distributed environment, the
low level machinery will not make the transition readily. Largely,
this is because the Solaris Threads library itself is not designed
to cross machine boundaries. We have not yet decided how best to
take this next step.
Additionally, we expect to implement a graphical method for
specifying the Petri nets that form the core of the system. This
software is straightforward (although time consuming) to create, and
will be a major aid to system understanding and debugging. Also, it will
help ensure that documentation and software are kept synchronized.
As part of this effort, we will reexamine the features of
our extended Petri nets, and possibly modify them to reduce the need
to propagate state information outside of the nets.
There are a number of ways in which we plan to improve and extend
the pipeline. The two most important are to include a process which
analyzes images in the pipeline to automatically focus the telescope;
and to include the ``alert'' software, which detects new microlensing events,
within the pipeline.
Finally, the volume of our reduced data is rapidly outstripping our
ability to purchase online disk storage for it. We plan to move to a
system that holds older data on ``nearline'' storage at the ANU
Supercomputer Facility (ANUSF), transparently migrating it as it is
required. Our data base organization, based as it is on large numbers
of individual files, will adapt well to this form of distribution.
\section{FIGURES}
\begin{figure}
\plotone{fig1.eps}
\caption{Major components of the MACHO computer hardware}
\end{figure}
\begin{figure}
\plotone{fig2.eps}
\caption{The processes that make up the observation cycle are shown. The arrows connecting them indicate the sequence in which the operations are performed for each observation.}
\end{figure}
\begin{figure}
\plotone{fig3.eps}
\caption{A simple example of a Petri net. The circles denote places and the horizontal lines transitions. The filled circle in place P1 indicates it is marked with a token. This will allow transition T1 to fire, thereby marking places P3 and P4. T3 will then fire, marking P5. Unless place P2 is marked, no further activity can take place in this net. If it is marked by some external agency, however, Transition T2 will then fire, and the whole cycle will begin again.}
\end{figure}
\begin{figure}
\plotone{fig4.eps}
\caption{The same net as in Fig. 3 is shown, but now the dotted line shows the boundaries between two separate processes, A and B. When transition T1 fires, output place OP is marked, causing an interprocess message to be sent that marks the input place IP in process B. The functionality of the net is exactly the same as before.}
\end{figure}
\begin{figure}
\plotone{fig5.eps}
\caption{Geometry of fundamental data base structures}
\end{figure}
\begin{figure}
\plotone{fig6.eps}
\caption{The functionality of the various interprocess communication entities are indicated. MB denotes a mailbox, and WQ a work queue. Note that processes 2 and 3 share the same work queue. A work queue item will be processed by whichever of them is first free.}
\end{figure}
\newpage
\begin{verbatim}
Table 1. Display of the information contained in a typical observation descriptor.
observation state:
obs id: 33415
ObsNumber: 33415 time: Wed Jul 26 18:25:26 1995
field_id: 113 RA: 18:0:20.9363 DEC: -28:56:45.12
telescope pointing: RA: 18:00:20.8514 DEC: -28:56:45.8466 West of Pier
exposure time: 150
seeing: 0
airmass: 1.37137
parallactic angle: 4.32645
refraction: 58.2
focus: 3.92
Locators:
Active: /meta_tmp/PActive
Archive: /meta_tmp/PArchive
Camera Configuration: test.camconfig
archive image CCD tag: 950723
Camera Status:
readout[0]: 1100
readout[1]: 2200
readout[2]: 0
readout[3]: 0
readout[4]: 1124
readout[5]: 2148
readout[6]: 1
readout[7]: 1
readout[8]: 0
readout[9]: 0
readout[10]: 0
readout[11]: 0
readout[12]: 10
nclears1: 0 nclears2: 5 nwait: -1
Controller flags:
check_list: 0x101ff
obs_type: 0
reobs_cam_tag: 0
reobs_date: invalid reobs_check_list: 0x0
photometry type: Sod
Flatfielding status:
flags: XTALK BYROW DARK VOODOO1 FLAT RESCALE STATS WARN
rescaled_above = 62200
config_file_tag = 3
ctm_file_tag = 4
dark_image_tag = 5
flat_image_tag = 7
debias_region = [1071,1120:1,2148]
voodoo1_rows = 701 to 1041
amp_info:
gain itop bias noise sky xtalk dark stuck
_0_0: 0.827 78650 0 4.14 0 2 0 FF80
_0_1: 1.45 57475 0 4.40 2687 3 -6 0000
_1_0: 1.59 57475 0 3.97 2626 4 -2 0000
_1_1: 1.48 64735 0 4.24 2670 0 0 0000
_2_0: 1.47 57475 0 4.36 2523 4 -1 0000
_2_1: 1.52 58685 0 4.09 2508 -2 0 0000
_3_0: 1.54 59895 0 4.03 2176 0 -1 0000
_3_1: 1.52 58685 0 3.81 2592 3 -2 0000
_4_0: 1.66 61105 0 4.36 1450 -1 1 C000
_4_1: 1.53 58685 0 3.89 1600 2 1 0000
_5_0: 1.52 58685 0 4.84 1803 3 0 8000
_5_1: 1.67 53845 0 5.38 1600 0 0 8000
_6_0: 1.50 50215 0 5.23 1660 0 0 0000
_6_1: 1.42 50215 0 5.19 1701 0 0 8000
_7_0: 1.40 51425 0 4.51 1721 -1 0 0000
_7_1: 1.33 50215 0 4.55 1794 1 0 8000
Sensor log:
time_stamp: Wed Jul 26 18:24:56 1995
Tout = 5.084 C
Tinsd = 5.815 C
Tmtlw = 5.645 C
Ttlwr = 5.385 C
Tmirr = 6.122 C
Tmtmd = 5.690 C
Ttmid = 5.449 C
Ttupr = 6.261 C
Dap template tag: -99999999
PR source version: -99999999
PR CCD tag: -99999999
PR date: invalid
Sod group template tag: 940510
PR source version: 87
PR CCD tag: 950723
PR date: Wed Jul 26 18:34:33 1995
chunk count: 121
chunk: 0
chunk:
chunk id: 0
sod template observation: 7793
star count: 9582
psf parameter(0): 9.273
psf parameter(1): 3.021
psf parameter(2): 2.708
psf parameter(3): -0.03729
average sky value(0): 9337
average sky value(1): 3986
transformation parameter(0): 1.003
transformation parameter(1): 0.000962
transformation parameter(2): -0.0009312
transformation parameter(3): 1.003
transformation parameter(4): -16.3
transformation parameter(5): 6.021
transformation parameter(6): 0.0109
transformation parameter(7): -0.01907
transformation parameter(8): 0
transformation parameter(9): 0
transformation parameter(10): 0
transformation parameter(11): 0
dap template observation: 0
ddap template sky per pixel: 0
sky per pixel: 0
aperture size used: 0
sky correction: 0
flux calculation: 0
transformation parameter(0): 0
transformation parameter(1): 0
transformation parameter(2): 0
# normalization stars: 25
normalization stars
star id:: tile: 18942 seqn: 42
star id:: tile: 19072 seqn: 47
star id:: tile: 18942 seqn: 44
star id:: tile: 18942 seqn: 46
star id:: tile: 18941 seqn: 50
star id:: tile: 18941 seqn: 53
star id:: tile: 19071 seqn: 57
star id:: tile: 18941 seqn: 54
star id:: tile: 18941 seqn: 59
star id:: tile: 18941 seqn: 60
star id:: tile: 18942 seqn: 53
star id:: tile: 18942 seqn: 54
star id:: tile: 18942 seqn: 55
star id:: tile: 18942 seqn: 56
star id:: tile: 18943 seqn: 20
star id:: tile: 18941 seqn: 69
star id:: tile: 18942 seqn: 62
star id:: tile: 19072 seqn: 66
star id:: tile: 18942 seqn: 64
star id:: tile: 18942 seqn: 65
star id:: tile: 18943 seqn: 21
star id:: tile: 18942 seqn: 67
star id:: tile: 18941 seqn: 74
star id:: tile: 19071 seqn: 76
star id:: tile: 18941 seqn: 75
.....
Information from additional 120 chunks deleted
.....
byte count of Sod input parameters: 1878
Sod Input Parameters:
N0STARMN = 6
SEEMIN = 2.0
HWHM_TOL = 2.0
NPIXCELL = 10
AXIS_RATIO = 1.0
TILT = 0.0
ISKYROWMN = 25
ISKYROWMX = 500
ISKYROWINT = 25
SKYPERCENT = 0.3
DELSKYHIST = 30.
SKYMIN = 0.
SKYMAX = 40000.
TOLMATCH = 3.
NFITBOX_X = 11
NFITBOX_Y = 11
MASKBOX_X = 5
MASKBOX_Y = 5
APBOX_X = 17.0
APBOX_Y = 17.0
THRESHMIN = 40.0
THRESHMAX = 20000.0
THRESHDEC = 1.0
THRESHDEC2 = 1.5
THRESHDEC3 = 1.5
NLOOPMN = 2
NLOOPMN2 = 2
NLOOPMN3 = 2
IDTYPE = 1
LEVPSFFIT = 2
ISAMEC = 1
VMINUSR = 0.
DFMAGSIG = 5.0
DIFMAGMN = 0.05
FWHMPSF = 5.0
ILOOKUP = 1
NPERFWHM = 80
STPAR2CRIT = 20000.
NSTRFIND = 100
NTRI_IMG = 5000
NSTAR_FID = 100
NTRI_FID = 5000
NTRI_MATCH = 100
MIN_SET_SIZE = 35
BOX_RADIUS = 1
NTCELL = 40
DMIN = 30.
DMAX = 700.
TOL = 2.0
SLOPE = 0.005
FAT = 20.
HIST_LEVEL = 0.003
MN_MATCH = 7
DEL_X_MX = 275.
DEL_Y_MX = 275.
THETA_MX = 0.03
DEL_X_TOL = 7.0
DEL_Y_TOL = 7.0
THETA_TOL = 0.015
R_TOL = 2.0
RATIN_MX = 1.10
COLORTERM_V = 0.108
COLORTERM_R = 0.025
MAGFIDNORM = 0
SCALE_RAT0 = 1.
NPOSTFIT = 1
NSKYBOXS2 = 8
ROLDMX = 2.5
PRETHRESHFAC = 3.
CHI2CUT = 10000.
DMAGCUT = 0.1
DMSECUT = 7.0
RXFITMX = 5.0
NSKYAVX = 10
NSKYAVY = 10
NSKYGRIDX = 40
NSKYGRIDY = 40
AUTOTHRESH = 'YES'
OBJ_IN_NORM = 1.
LOGVERBOSITY = 1
FOOTPRINT_NOISE = 1.3
NPHSUB = 1
NPHOB = 1
ICRIT = 40
CENTINTMAX = 2.5e5
CTPERSAT = 8000.
STARGALKNOB = 8.e6
STARCOSKNOB = 1.0
KCOS = 2
SNLIM7 = 7.0
SNLIM = 0.5
SNLIMMASK = 4.0
SNLIMCOS = 3.0
NBADLEFT = 0
NBADRIGHT = 0
NBADTOP = 0
NBADBOT = 0
SKYTYPE = 'PLANE'
OBJTYPE_IN = 'SMALL'
OBJTYPE_OUT = 'UNFORM'
BADTYPE_IN= = Input
IALERT = 1
NFITBOXFIRST_X = 31
NFITBOXFIRST_Y = 31
CHI2MINBIG = 16
XTRA = 25
SIGMA1 = 0.10
SIGMA2 = 0.10
SIGMA3 = 0.10
ENUFF4 = 0.50
ENUFF7 = 0.80
COSOBLSIZE = 0.9
APMAG_MAXERR = 0.1
PIXTHRESH = 1.0
BETA4 = 1.0
BETA6 = 0.5
FITBOXMIN = 5.0
SCALEAPBOX = 6.0
APBOXMIN = 7.0
SCALEMASKBOX = 1.5
AMASKBOXMIN = 5.0
SIGMAIBOTTOM = 10.0
SIGMATHRESHMIN = 2.0
END = =
\end{verbatim}
|
\section{Introduction}
The role of phonons in high-temperature superconductors is complex. On the one
hand there is ample evidence that the cuprates are not conventional
electron-phonon BCS superconductors. This evidence includes, for the optimally
doped materials a linear resistivity, in the normal state, with a
zero-tem\-pera\-ture intercept at zero resistance \cite{Gurvitch87,Martin90}, a
collapse of transport scattering at the superconducting transition
\cite{Romero92,Bonn92b,Yu92}, and a lack of an isotope effect
\cite{Batlogg87,Bourne87,Franck90,Crawford90b}.
On the other hand, a large number of studies show that certain phonons at least
are coupled to the carriers, and that this coupling is unconventional
\cite{Litvinchuk94,Pintschovius94}. The frequencies and widths of certain
phonon lines, again in the optimally-doped materials (precisely those where the
isotope effect is small), undergo abrupt changes at $T_c$. This effect was
first observed by infrared spectroscopy
\cite{Bonn87b,Wittlin87a,Litvinchuk92,Macfarlane87}, then by Raman spectroscopy
\cite{Macfarlane87, Friedel90b,Altendorf93}, resonant neutron-absorption
spectroscopy \cite{Mook90} and neutron scattering \cite{Pyka93}.
Broad minima in the {\it ab}-plane optical conductivity are seen at frequencies
that are in the phonon region. These have been shown to be associated with {\it
c}-axis longitudinal optic (LO) phonons \cite{Timusk91,Reedyk92e} that couple
to the {\it ab}-plane conductivity by a symmetry-breaking mechanism.
In this paper we focus on the changes in the {\it c}-axis phonon spectra of
YBa$_2$Cu$_3$O$_{6+x}$ as we change the oxygen doping. These spectra can be
separated from the weak and smooth electronic background conductivity whose
intensity rises with doping from a few $\Omega^{-1}{\rm cm}^{-1}$ in the
$x=0.5$ material\footnote{C.C. Homes, T. Timusk, D.A. Bonn, R. Liang and W.N.
Hardy. Unpublished results.} to 450~$\Omega^{-1}{\rm cm}^{-1}$ in overdoped
samples \cite{schutzmann94}. The background is due to inter-cell hopping of
free carriers \cite{Cooper94}. Superimposed on this electronic background are
five strong phonon lines. These five lines are also seen the in the spectra of
ceramic superconductors \cite{bonn88a,genzel89}, and have been studied by a
large number of groups. The extensive literature on ceramic spectra has been
summarized by Feile in ref.~\citen{feile89}. However, in ceramics, effects due
to the background {\it ab}-plane conductivity distort the spectra, particularly
the line strengths, and the {\it c}-axis polarized phonons are best studied in
single crystals. Single crystal studies are difficult due to the small area of
the available flux-grown samples. Nevertheless, a number of studies have
appeared, with YBa$_2$Cu$_3$O$_{6+x}$ being the most studied material [see
footnote 3 and
refs.~\citen{schutzmann94,bozovic87,collins89c,koch90,koch90b,cooper92,homes93a}].
In the present work single crystals with cleaved surfaces are used and the
region of doping ranges from the optimally doped, that is the material with the
highest $T_c=93.5$~K with $x=0.95$, to an underdoped sample with $x=0.5$ which
is still superconducting with a $T_c=53$ K. Preliminary reports of this work
have been published \cite{homes93a, timusk92b,timusk95a}, and the electronic
background is the subject of a separate publication [see footnote 3].
\section{Experimental Details}
\subsection{Experimental}
The spectra were obtained by a Kramers-Kronig transformation of {\it c}-axis
polarized reflectance of millimeter-size, high-quality single crystals. The
growth of the single crystals and experimental techniques used have been
reported on previously \cite{homes93a,liang92}. To maximize the signal from the
{\it ac} face of the crystal, an infrared beam larger than the crystal was
used. The sample was coated with gold {\it in situ} and then remeasured at each
temperature to obtain a reference spectrum \cite{homes93b}.
The reflectance of YBa$_2$Cu$_3$O$_{6+x}$, for two oxygen concentrations,
$x=0.6$ and 0.95, for light polarized parallel to the {\it c} axis is shown in
the phonon frequency range in Fig.~\ref{reflec} at several temperatures. The
reflectance has a continuous background, rising towards unity at low frequency,
characteristic of a poor metal, with sharp resonances due to optical phonons.
In the superconducting state, the development of the superconducting condensate
has the effect of producing a reflectance edge which shifts down to lower
frequency as the doping is reduced. Also, as the temperature is lowered, the
shapes of the phonon features change due to the changing background
conductivity.
Changes in oxygen doping have two principal effects on the phonon spectra:
first, as the doping level is lowered, a new phonon feature appears at $\approx
610$ cm$^{-1}$. Secondly, a very broad feature, weak at first, but stronger at
low oxygen dopings, appears in the normal state at $\approx 400$ cm$^{-1}$, and
becomes stronger at low temperature. These effects are seen more clearly in the
optical conductivity, discussed in the following section.
\subsection{Optical conductivity}
A Kramers-Kronig transformation of the reflectance was us\-ed to calculate the
optical conductivity. The reflectance was extended to regions outside the
actual measurements as follows: below the lowest measured frequency
Hagen-Rubens frequency dependence: [$(1-R)\propto \omega^{-1/2}$] was assumed
to hold in the normal state. In the superconducting state, where a plasma edge
develops, it was assumed that the reflectance was given by
$(1-R)\propto\omega^{-2}$. This assumes that the conductivity is finite and
constant to zero frequency. There is good evidence from comparisons with
transport measurements that both assumptions are valid [see footnote 3].
Temperature-dependent reflectance measurements were carried out to
5000~cm$^{-1}$. The room-temperature measurements of Koch et al. \cite{koch90}
were used from 5000 cm$^{-1}$ to $3.5\times 10^5$ cm$^{-1}$ and above this
frequency free-electron behavior was assumed: $R\propto \omega^{-4}$.
\begin{figure}[t]
\caption{The reflectance of YBa$_2$Cu$_3$O$_{6+x}$ for radiation polarized
along the {\it c}-axis from $\approx 50$ to 700 cm$^{-1}$ at several
temperatures above and below $T_c$, for two oxygen dopings $x$: ({\it a})
fully-doped material with $x=0.95$ ($T_c=93$~K), and ({\it b}) an underdoped
crystal with $x=0.6$ ($T_c=58$~K). The main change in the phonon spectrum in
the underdoped material is a new line at 610~cm$^{-1}$ that we associate with
the bridging oxygen at the two-fold coordinated copper sites in the chain
layer. Another new feature in the $x=0.6$ material is a broad band at $\approx
400$ cm$^{-1}$ that appears at low temperature.}%
\vspace*{0.4cm}%
\centerline{\includegraphics[width=11.0cm]{figure1.eps}}%
\vspace*{1.2cm}%
\label{reflec}%
\end{figure}
The 295~K conductivity along the {\it c}-axis is shown in Fig.~\ref{sigma295}
for a wide range of dopings. The 10~K conductivity is shown in
Fig.~\ref{sigma10}. Superimposed on a continuous background that increases in
strength with doping are a series of sharp phonon lines. In general, the
strength of the phonon bands does not depend strongly on doping, but as the
doping is reduced a new phonon appears at $\approx 610$~cm$^{-1}$ in the
nearly fully-doped material and moves to higher frequency as the oxygen
concentration is reduced. We will call this the 610 cm$^{-1}$ line in what
follows, but its actual frequency changes with doping and is as high as
635~cm$^{-1}$ in the $x=0.5$ material. The other major effect of doping occurs
at low temperature where a new broad band appears at 400~cm$^{-1}$ as the
doping level is reduced to $x=0.7$. This band seems to increase its oscillator
strength at the expense of the two high-frequency modes and the band at
310~cm$^{-1}$. Also, its center frequency appears to decrease as the doping is
reduced, reaching a value of $\approx 375$~cm$^{-1}$ in the $x=0.5$ crystal.
\begin{figure}
\caption{The optical conductivity ($\sigma_1$) at 295~K for
YBa$_2$Cu$_3$O$_{6+x}$ for radiation polarized along the {\it c}-axis from
$\approx 50$ to 700 cm$^{-1}$ for five oxygen dopings, $x=0.5 \rightarrow
0.95$. As the doping is reduced, the background conductivity is reduced
uniformly at all frequencies. At high frequency, in addition to the apical
oxygen [O(4)] peak at 570~cm$^{-1}$ a new phonon peak grows at $\approx
610$~cm$^{-1}$; this peak is associated with the O(4) mode and the growing
density of two-fold coordinated copper sites in the underdoped materials. As
the doping is reduced, the O(4) modes acquire an increasingly dispersive line
shape.}
\vspace*{-0.4cm}%
\centerline{\includegraphics[width=10.5cm]{figure2.eps}}%
\vspace*{-1.5cm}%
\label{sigma295}
\end{figure}
\section{Results}
\subsection{Calculation of the phonon parameters}
In what follows we refer to the six phonons that we observe by their room
temperature frequencies in the fully-doped material, bearing in mind that the
actual center frequencies change with both temperature and doping. There are
five strong phonons, and at room temperature their frequencies are: 155, 194,
279, 312 and 570 cm$^{-1}$, plus a weaker one at 610 cm$^{-1}$. All the lines
have symmetric line shapes except for the two highest ones. The 570 cm$^{-1}$
and the 610 cm$^{-1}$ lines have Lorentzian shapes at room temperature, but at
low temperature they become asymmetric \cite{burlakov}. Such lines are usually
described in terms of a Fano line shape
\begin{displaymath}
\sigma_1(\omega) = A \left[ {{(x+q)^2}\over{(1+x^2)}} \right]
\end{displaymath}
where $\sigma_1(\omega)$ is the conductivity, $A$ is a constant, $x
=(\omega - \omega_i)/\gamma_i$, $\omega_i$ is the phonon frequency, $\gamma_i$ is
the phonon line width and $q_i$ is a parameter that describes the asymmetry of
the $i$th phonon \cite{fano61}.
\begin{figure}
\caption{The {\it c}-axis optical conductivity ($\sigma_1$) at 10~K for five
oxygen dopings, $x = 0.5 \rightarrow 0.95$. The curves have been offset
vertically in 200 $\Omega^{-1}$cm$^{-1}$ increments for clarity. There is a
major redistribution of spectral weight from the O(4) peaks at $\approx 570$
cm$^{-1}$ and $\approx 610$~cm$^{-1}$, and the in-plane buckling mode at 310
cm$^{-1}$, to an new broad band in the $\approx 400$ cm$^{-1}$ range. The peak
shifts to lower frequency as the doping is reduced, starting as a shoulder on
the apical-oxygen line at 550~cm$^{-1}$ and ending up centered at $\approx 375$
cm$^{-1}$ in the $x=0.5$ sample.}
\vspace*{0.4cm}%
\centerline{\includegraphics[width=10.5cm]{figure3.eps}}%
\vspace*{1.2cm}%
\label{sigma10}
\end{figure}
\begin{table*}[t]
\topcaption{The parameters used to fit Lorentzian line shapes to the
peaks in the optical conductivity of YBa$_2$Cu$_3$O$_{6+x}$, for
$x=0.95, 0.85$, $0.7$, and $0.5$ at 10~K. The two high-frequency
copper-oxygen vibrations have been fitted using non-zero rotations.$^a$ }
\begin{tabular}{ccccccccccccccc}
\hline*
\multicolumn{3}{c}{$x=0.95$} & & \multicolumn{3}{c}{$x=0.85$} & &
\multicolumn{3}{c}{$x=0.7$} & & \multicolumn{3}{c}{$x=0.5$} \\
%
\cline{1-3} \cline{5-7} \cline{9-11} \cline{13-15}
$\omega_{{\rm TO},i}$ & $\gamma_i$ & $\omega_{pi}$ & & $\omega_{{\rm TO},i}$ &
$\gamma_i$ & $\omega_{pi}$ & & $\omega_{{\rm TO},i}$ & $\gamma_i$ &
$\omega_{pi}$ & & $\omega_{{\rm TO},i}$ & $\gamma_i$ & $\omega_{pi}$ \\
\hline
155 & 2.0 & 398 &
$\left\{ \begin{array}{c} \\ \end{array} \right.$ \hspace*{-0.8cm}&
\begin{tabular}{c} 147 \\ 153 \end{tabular} &
\begin{tabular}{c} 8.4 \\ 3.8 \end{tabular} &
\begin{tabular}{c} 248 \\ 351 \end{tabular} &
$\left\{ \begin{array}{c} \\ \\ \end{array} \right.$ \hspace*{-0.8cm} &
\begin{tabular}{c} 136 \\ 147 \\ 153 \end{tabular} &
\begin{tabular}{c} 15 \\ 9.8 \\ 4.2 \end{tabular} &
\begin{tabular}{c} 167 \\ 322 \\ 267 \end{tabular} &
$\left\{ \begin{array}{c} \\ \\ \\ \\ \end{array} \right.$ \hspace*{-0.8cm} &
\begin{tabular}{c} 93 \\ 113 \\ 125 \\ 145 \\ 161 \end{tabular} &
\begin{tabular}{c} 11 \\ 5.0 \\ 5.2 \\ 9.0 \\ 5.9 \end{tabular} &
\begin{tabular}{c} 116 \\ 80 \\ 117 \\ 363 \\ 120 \end{tabular} \\
194 & 2.2 & 128 & $\rightarrow$ \hspace*{-0.8cm} &
192 & 2.8 & 133 & $\rightarrow$ \hspace*{-0.8cm} &
189 & 4.2 & 133 & $\rightarrow$ \hspace*{-0.8cm} &
184 & 5.9 & 110 \\
279 & 19 & 357 & $\rightarrow$ \hspace*{-0.8cm} &
281 & 19 & 329 & $\rightarrow$ \hspace*{-0.8cm} &
280 & 22 & 323 & $\rightarrow$ \hspace*{-0.8cm} &
272 & 33 & 299 \\
312 & 5.2 & 487 & $\rightarrow$ \hspace*{-0.8cm} &
312 & 7.0 & 417 & $\rightarrow$ \hspace*{-0.8cm} &
309 & 18 & 333 & $\rightarrow$ \hspace*{-0.8cm} &
301 & 28 & 184 \\
570 & 14 & 483 &
$\left\{ \begin{array}{c} \\ \end{array} \right.$ \hspace*{-0.8cm} &
\begin{tabular}{c} 566 \\ 619 \end{tabular} &
\begin{tabular}{c} 23 \\ 23 \end{tabular} &
\begin{tabular}{c} 321 \\ 205 \end{tabular} &
$\left\{ \begin{array}{c} \\ \end{array} \right.$ \hspace*{-0.8cm} &
\begin{tabular}{c} 564 \\ 630 \end{tabular} &
\begin{tabular}{c} 30 \\ 24 \end{tabular} &
\begin{tabular}{c} 188 \\ 250 \end{tabular} &
$\left\{ \begin{array}{c} \\ \end{array} \right.$ \hspace*{-0.8cm} &
\begin{tabular}{c} 551 \\ 637 \end{tabular} &
\begin{tabular}{c} 30 \\ 24 \end{tabular} &
\begin{tabular}{c} 127 \\ 251 \end{tabular} \\
\hline*
\end{tabular} \\
{\small
\indent NOTE: All parameters are in cm$^{-1}$. \\
\ $^a$ The two high-frequency apical-oxygen modes have
been fit using a Lorentzian with a phase. The angles (in radians)
used are for $x=0.95$, $\theta=-0.40, 0.0$, for $x=0.85$, $\theta=-0.50,
-0.15$, for $x=0.7$, $\theta=-0.62, -0.45$, for $x=0.5$, $\theta=-0.75,
-0.80$, for the low- and high-frequency components, respectively.
}
\vspace*{0.3cm}%
\label{doping}%
\end{table*}
While the Fano line shape describes the 570~cm$^{-1}$ phonon at low
temperatures, it fails as the line shape becomes more Lorentzian
($q\rightarrow\infty$). A better fit to the data is in terms of a classical
Drude-Lorentz model for the dielectric function, for each phonon band, but with
an empirical line shape made up of a mixture of the real and imaginary parts of
the Lorentzian oscillator:
\begin{displaymath}
\tilde\epsilon(\omega) = { {\omega_p^2\,e^{i\theta}} \over {\omega_{\rm TO}^2
- \omega^2 - i\omega\gamma}} \hspace*{4.0cm} (1)
\end{displaymath}
where $\omega_{p}$ is the effective plasma frequency, $\omega_{\rm TO}$ the
center frequency and $\gamma$ the width of the phonon. The parameter $\theta$
is the phase associated with the asymmetry of the phonon. The line shape is
that of a classical oscillator for $\theta=0$ but becomes asymmetric for
nonzero $\theta$.
An advantage of this rotated-Lorentzian line shape over the Fano profile is
that one can associate with it an oscillator strength. For a pure Lorentzian,
the integrated oscillator strength is defined to be $\omega_{p}^2/8$; for the
rotated Lorentzian, to leading order the oscillator strength is
$\cos(\theta)\omega_p^2/8$, so that for small rotations, the introduction of an
asymmetric line shape does not alter the oscillator sum rule too much. However,
for rotations of greater than $\approx 0.4$ radians, the error introduced is
greater than 10\%, and the other contributions to the sum rule begin to play an
increasingly large role, further compounding this error. The asymmetry of the
rotated Lorentzian is related to the Fano parameter $q$ by $q^{-1} \propto\tan
(\theta/2)$.
All the phonon lines have been fitted to (1) with a non-linear least-squares
method. In addition to the parameters of the modified-Lorentzian oscillator, a
linear background was used in the fits. We find that within the accuracy of
our experiments, the four low-frequency lines are symmetric and only the
570~cm$^{-1}$ line, and the new feature associated with it at 610~cm$^{-1}$,
have clearly asymmetric shapes which become more pronounced at low
temperatures.
It should be noted that if the Kramers-Kronig analysis is done on data that
deviates from the true reflectance, for example by using room-temperature data
for the high-frequency extrapolation for a low-temperature data set, there is a
tendency for sharp lines to acquire an artificial asymmetry. As the
experimental data become more accurate, the lines become more symmetric. The
lines that have the highest reflectance are most susceptible to this error,
which in our case is the 155~cm$^{-1}$ line, which we find accurately
symmetric, thus confirming the position of our 100\% line. We find that an
error of greater than 2\% in the unity reflectance line would show an
observable asymmetry.
The phonon parameters obtained by the least-squares fitting procedure applied
to the rotated-Lorentzian profile with of all the measured spectra are
assembled in Table~\ref{doping}. We estimate the error in the parameters to be
$\approx 1$\% for the center frequencies, $\approx 5$\% for the line widths,
and $\approx 5$\% for the effective plasma frequencies. Where there is a
finite rotation angle $\theta$, the error could be larger.
\subsection{Assignment of phonons}
Group theory predicts seven $B_{1u}$ infrared-active modes polarized in the
c-direction for the fully-oxygenated ortho\-rhombic (Ortho~I)
YBa$_2$Cu$_3$O$_7$ shown in Fig.~\ref{struct} \cite{bates87,kress88}. There is
crystallographic evidence that the oxygen-reduced material, with a plateau in
the $T_c$ vs $x$ graph (in the $T_c\approx 60$~K region), has a doubled unit
cell normal to the chain direction, resulting from an alternation of
fully-occupied O(1) sites on chains with empty sites. This structure (Ortho~II)
is expected to have 13 infrared-active modes, but many of the new modes result
from extremely small splittings of modes involving atoms far away from the
chains. These small splittings may be partially responsible for the systematic
broadening of most of the phonon lines at reduced dopings. We expect the
largest changes for modes that involve atoms on and near the chains.
\begin{figure}[t]
\caption{The orthorhombic unit cell of YBa$_2$Cu$_3$O$_7$. The chain oxygen is
denoted as O(1), the apical copper as Cu(1), and the O(4) as the apical oxygen
atom. When all the O(1) sites are occupied, the Cu(1) atom is four-fold
coordinated. Deoxygenation results in the oxygens at the O(1) sites being
removed, lowering the coordination of the Cu(1) atom.}
\vspace*{0.4cm}%
\centerline{\includegraphics[width=8.0cm]{figure4.eps}}%
\vspace*{0.2cm}%
\label{struct}
\end{figure}
\begin{table}
\caption{The fitted phonon parameters for the {\it c}-axis $B_{1u}$ modes in
YBa$_2$Cu$_3$O$_{6.95}$ at 295~K determined from optical and inelastic
neutron-scattering measurements. The effective-plasma frequency based on the
shell model with formal charges in column 5 [Y = +3.0, Ba = +2.0, Cu1 = Cu2 =
2, O(1)
= O(2) = O(3) = O(4) = -2] and fitted charges in column 6, [Y = +2.8, Ba = +2.5,
Cu1 = Cu2 = 2, O(1)= -1.8, O(2) = -1.8, O(3) = -.17, O(4) = -2.6].}
\begin{tabular}{ccccccc}%
\hline*
\multicolumn{3}{c}{Optical$^{\rm a}$} & &
\multicolumn{3}{c}{Neutron$^{\rm b}$} \\
\cline{1-3} \cline{5-7}
$\omega_{{\rm TO}i}$ & $\gamma_i$ & $\omega_{pi}$ & & $\omega_i$ &
$\omega_{pi}^{\rm c}$ & $\omega_{pi}^{\rm d}$ \\
\hline
--- & --- & --- & & 111 & 100 & 115 \\
153 & 3.6 & 389 & & 149 & 283 & 330 \\
--- & --- & --- & & 186 & 22 & 14 \\
195 & 3.7 & 95 & & 198 & 79 & 94 \\
287 & 18.3 & 316 & & 291 & 455 & 400 \\
315 & 15.6 & 498 & & 322 & 493 & 457 \\
\ \ \ 568$^{\rm e}$\ \ \ & \ \ \ 18.4\ \ \ & \ \ \ 408\ \ \ & \ \ &
\ \ \ 556\ \ \ & \ \ \ 256\ \ \ & \ \ \ 365\ \ \ \\
\hline*
\end{tabular} \\
\small{
NOTE: All units are in cm$^{-1}$. \\
$^a$Fitted to optical conductivity, this work. \\
$^b$W.~Reichardt et al. Private communication. \\
$^c$Effective charges, formal valence charges. \\
$^d$Effective charges, fitted to optical data. \\
$^e$The high-frequency apical-oxygen mode has been fitted using an asymmetric
line shape described by a Lorentzian with a rotation angle of $\theta=-0.27$
rad.
}
\vspace*{0.3cm}
\label{neutrons}
\end{table}
The assignment of the infrared-active spectral lines to normal modes is most
reliable when the results from inelastic-neutron scattering are combined with a
realistic lattice-dynamic model \cite{cowley63}. Such a procedure has been
carried out recently for fully-oxygenated YBa$_2$Cu$_3$O$_7$,\footnote{W.
Reichardt. Private communication.} and we make extensive use of the
eigenvectors of these models in our discussion of {\it c}-axis the phonon
spectra. A list of calculated contributions the various modes to the
polarizability has been published by Timusk et al. \cite{timusk95a}. It is
clear from the shell models that infrared modes, unlike Raman modes, which can
often be assigned to the motion of a particular ion, usually involve the motion
of all the ions in the unit cell. The shell-model eigenvectors suggest that
there are some exceptions. For example the 570 cm$^{-1}$ mode is almost a pure
mode of the apical O(4) oxygen (a description of the different copper and
oxygen environments is shown in Fig.~\ref{struct}). Similarly, the 279
cm$^{-1}$ mode is an O(1) chain-oxygen vibration, polarized in the {\it
c}-direction and the 315 cm$^{-1}$ mode an O(2) and O(3) plane-bending
vibration. The low-frequency modes involve many more atoms, and while the 194
cm$^{-1}$ mode has been identified as the yttrium mode, it also involves
substantial motion of the O(2) planar-oxygen atoms. The 155 cm$^{-1}$ mode,
often called the barium mode, according to the shell model gets more spectral
weight from the O(4) oxygen motion than from the barium, and also involves the
motions of the Cu(1) and O(1) chain-copper and -oxygen atoms. This prediction
is in agreement with our observation that this modes splits in oxygen-reduced
materials.
\begin{figure*}[t]
\caption{The variation of the central frequency ($\omega_{{\rm TO},i}$), width
($\gamma_i$) and the effective plasma frequency ($\omega_{pi}$) as a function
of doping ($x=0.5\rightarrow 0.95$) of the five strong phonons observed in the
highly-doped materials, at 10~K ($\bullet$). The room-temperature frequencies
have been included for the three central panels ($\ast$). Also included in the
panels are two phonons that appear only at lower oxygen dopings ($\circ$): in
the first panel, at new mode appears at $\approx 148$ cm$^{-1}$ and gains
oscillator strength as the mode observed in the highly-doped materials at
$\approx 155$ cm$^{-1}$ becomes weaker. The second phonon appears at $\approx
610$ cm$^{-1}$ and, similarly to the 148 cm$^{-1}$ phonons, gains oscillator
strength with decreasing doping. Inset: rotation angles for the $\approx 570$
cm$^{-1}$ mode ($\bullet$), and the $\approx 610$ cm$^{-1}$ mode ($\circ$).}
\vspace*{-6.8cm}%
\centerline{\includegraphics[width=16.3cm]{figure5.eps}}%
\vspace*{-0.5cm}%
\label{parms}
\end{figure*}
\begin{figure}
\caption{The evolution of the shape of the 155 cm$^{-1}$ line, at 295~K and
10~K, for five oxygen dopings ($x=0.5 \rightarrow 0.95$). As the oxygen doping
is reduced, the line broadens, first developing shoulders, and then finally
splitting into several clearly resolved components.}
\vspace*{0.3cm}%
\centerline{\includegraphics[width=10.5cm]{figure6.eps}}%
\vspace*{1.3cm}%
\label{barium}
\end{figure}
In the highly-doped YBa$_2$Cu$_3$O$_{6.95}$ material five strong pho\-nons are
observed at 155, 194, 279, 312 and 570~cm$^{-1}$ at 10~K. These frequencies are
in excellent agreement with previous measurements on ceramics
\cite{bonn88a,Crawford88}. Table~\ref{neutrons} compares these frequencies
with those of the neutron-based shell model. As expected, since the \boldmath
$k=0$ \unboldmath frequencies have been fitted to the model, the agreement is
satisfactory. What is surprising is the excellent agreement between the
measured line strengths, and those predicted by the shell-model eigenvectors.
In calculating the line strengths we have used two separate approaches: the
simplest is to use the formal charges of the ions in all cases. The only case
where this seems to be inappropriate is for the 570~cm$^{-1}$ mode where the
observed line strength is larger, and a larger effective charge on the O(4)
would give better agreement between the model and the infrared observations.
This line is also anomalous in that it develops a strong asymmetry in the
oxygen-reduced materials, a sign of strong electron-phonon interaction which
may enhance the polarizability of the mode by dynamic charge transfer from the
O(4) to the planes.
The second approach is to use a least-squares fit to the effective charges to
the observed infrared oscillator strengths. The results are shown in column
six of Table~\ref{neutrons}. The fitted results, as expected, agree better with
the optical data. However, it is interesting to note that in order to get this
better fit, the main change was the removal of charge from the O(4) ion, and
the addition of that charge on the planar oxygen ions.
The two missing lines are predicted by the shell model to be very weak. Of
these there is some evidence for the lower one predicted to be at 111 cm$^{-1}$
in some of our spectra, but they are near the limit of the accuracy of the
experiments.
Two modes not predicted by the shell model are seen: the broad band at $\approx
400$ cm$^{-1}$, and weak peak at $\approx 610$~cm$^{-1}$. In what follows we
will argue that the 610~cm$^{-1}$ line is associated with the two-fold
coordinated copper neighbor of the O(4) oxygen. Such sites are found in
oxygen-reduced material where the 610 cm$^{-1}$ line is much stronger, and its
presence in our $x=0.95$ sample is an indication that this sample contains a
small concentration of broken chain fragments with two-fold coordinated copper.
Overall, at room temperature at least, there is good agreement with the shell
model and the observed phonon spectrum and there is no need to invoke special
electronic enhancements to explain the intensities of the phonon lines, nor is
there any evidence of strong screening by the electronic background.
\subsection{Doping dependence of the phonon parameters}
As the doping level is reduced, a number of changes in the phonon parameters
take place; these are summarized for the five fundamental phonons observed in
the highly-doped material, as well as the two strong splittings associated with
the 155 and 570 cm$^{-1}$ modes at 10~K observed at lower dopings, in
Fig.~\ref{parms}. To a first approximation one expects the phonon frequencies
to soften with the expansion of the {\it c}-axis lattice parameter. One
generally expects the frequency to vary approximately as the third power of the
lattice parameter \cite{Maroni89}. Since the {\it c}-axis lattice parameter
increases by 0.5\% between $x=0.93$ to $x=0.50$, one would expect an overall
softening of the order of 1.5\% as the result of the lattice expansion
In very rough agreement with this simple picture, many mod\-es soften, but by a
larger amount of 3.5\% in this doping range. There are also several notable
exceptions. First there is a tendency, at room temperature, for several modes
to harden as the lattice expands on removal of oxygen, for example the 312
cm$^{-1}$ mode. The 155 cm$^{-1}$ mode as well as the high-frequency O(4) mode
behave in a similar fashion: they split into two components, the higher
frequency component hardens, the lower frequency one softens. As we will show
below for the 570 cm$^{-1}$ mode, and its high-frequency partner the 610
cm$^{-1}$ mode the higher frequency component corresponds to the O(4) at the
two-fold coordinated copper sites whereas the lower frequency corresponds to
the O(4) at the four-fold coordinated copper sites. While the shell model
eigenvectors suggest that the 155~cm$^{-1}$ mode also involves a large amount
of O(4) motion, many other atoms in the unit cell participate in the motion,
and the net result is a behavior that is opposite to that for the 570 cm$^{-1}$
mode: it is the high-frequency, hard component that loses oscillator strength.
The mode at $\approx 155$ cm$^{-1}$ in the highly-doped material has split into
a doublet in the $x=0.85$ material, and consists of at least three lines in the
$x=0.7$ material; the phonon lines are significantly broader than the narrow
line in the $x=0.95$ material. Fig.~\ref{barium} shows the evolution of the
shape of this line with doping. The multi-component nature of the line and the
gradual shift, with reduced doping to lower frequency components is reminiscent
of the 500~cm$^{-1}$ Raman line which has been shown to have many components,
associated with chain disorder in underdoped samples \cite{Hadjiev93}. The
splitting of the $\approx 155$~cm$^{-1}$ mode is also associated with the loss
of O(1) chain oxygens. Unlike the 570~cm$^{-1}$ mode, where the O(4)--Cu(1)
chain-copper bond is stretched, the 155~cm$^{-1}$ mode involves the O(4)--Cu(2)
bond which is also expected to change its length as the Cu(1) changes from
four- to two-fold coordination.
\begin{figure*}[t]
\caption{The temperature dependence of the frequency ($\omega_{{\rm TO},i}$),
line width ($\gamma_i$) and effective plasma frequency ($\omega_{pi})$ for the
three oxygen-related {\it c}-axis modes of YBa$_2$Cu$_3$O$_{6.95}$ ($\bullet$).
The $\approx 279$ cm$^{-1}$ mode fails to show any discontinuity at $T_c$,
while the 312 and 570 cm$^{-1}$ modes both soften below $T_c$; the 570
cm$^{-1}$ feature also appears to gain oscillator strength below $T_c$. The
rotation angle used to fit the asymmetric line shape is shown in the O(4) panel
($\circ$), and shows a slightly increasing asymmetry with decreasing
temperature.}
\vspace*{-0.5cm}%
\centerline{\includegraphics[width=16.3cm]{figure7.eps}}%
\vspace*{-1.6cm}%
\label{phona}
\end{figure*}
As the doping decreases to $x\approx 0.7$, the 155~cm$^{-1}$ O(4) mode softens
and decreases in strength, while the mode at 148~cm$^{-1}$ hardens slightly,
and gains strength. Below $x\approx 0.7$, it is clear that the fundamental
nature of this mode has changed; the 155 cm$^{-1}$ feature hardens
dramatically, and loses almost all of its oscillator strength; in contrast, the
148 cm$^{-1}$ mode softens, and its oscillator strength approaches the value of
the 155 cm$^{-1}$ mode in the highly-doped material.
The 312~cm$^{-1}$ plane-oxygen mode displays drastically different behavior at
room temperature and at 10~K. At room temperature, the mode hardens from
315~cm$^{-1}$ to 320 cm$^{-1}$ and its oscillator strength remains unchanged as
the doping is reduced. However, at 10~K, this mode softens dramatically with
decreasing doping, furthermore, at $x\approx 0.5$ this mode, at low
temperature, has lost most of its considerable oscillator strength. The loss
of oscillator strength at low temperature in the underdoped materials results
from the transfer of their spectral weight to the 400 cm$^{-1}$ mode, which
will be discussed in more detail below.
In the fully-oxygenated material the 610~cm$^{-1}$ mode can just be seen as a
weak side band of the 570~cm$^{-1}$ mode, but as the doping is reduced the
610~cm$^{-1}$ grows substantially in spectral weight. The mode hardens
substantially to 645~cm$^{-1}$ in the $x=0.5$ material, while the 570~cm$^{-1}$
mode that it is associated with is steadily losing oscillator strength with
decreasing doping. This effect has been attributed by Burns et al. to the
shortening of the O(4)-Cu(1) bond by 6\% in the same concentration range
\cite{Hazen91}. In rough agreement with this model the 610 cm$^{-1}$ line
changes its frequency by 6\% as well in the same concentration range. In Burns'
picture the 570 cm$^{-1}$ line arising from the O(4) at the four-fold
coordinated copper, {\it softens} by 3\%, a value typical of the overall {\it
c}-axis expansion, suggesting a {\it lengthening} of that bond. If the
interpretation of Burns is correct we would expect a modulation of the O(4)
position in the {\it a}-direction by approximately 0.03~\AA\ (1~\AA\ $=
10^{-10}$~m) in the Ortho~II structure where there is an alternation between
two- and four-fold coordinated coppers. There have been reports, by several
authors \cite{Conradson90,Mustre90,Stern93}, of a double peak in the
Cu(1)--O(4) bond length distribution measured by XAFS by as much as 0.1~\AA .
\begin{figure*}[t]
\caption{The temperature dependence of the frequency ($\omega_{{\rm TO},i}$),
line width ($\gamma_i$), and effective plasma frequency ($\omega_{pi}$) for the
three oxygen-related {\it c}-axis modes of YBa$_2$Cu$_3$O$_{6.70}$, and a new
fourth feature associated with the change in coordination of the Cu(1) atom,
discussed in the text ($\bullet$). The feature at $\approx 285$~cm$^{-1}$
softens with decreasing temperature, while the two modes at $\approx 557$
cm$^{-1}$ and $\approx 630$ cm$^{-1}$ harden smoothly; the feature at 318
cm$^{-1}$ softens well above $T_c$. The three high-frequency modes begin to
loose oscillator strength at $\approx 150$~K, while the low-frequency mode
gains in strength primarily below $T_c$. The rotation angles ($\circ$) used to
fit the two high-frequency modes are shown in the O(4)$^\ast$ and O(4) panels:
the $\approx 570$ cm$^{-1}$ mode becomes increasing asymmetric at low
temperature, while the mode at $\approx 610$~cm$^{-1}$ becomes more symmetric.}
\vspace*{-0.4cm}%
\centerline{\includegraphics[width=16.3cm]{figure8.eps}}%
\vspace*{-1.4cm}%
\label{phonb}
\end{figure*}
The widths of the phonons increase steadily with decreasing doping, with the
exception of the 155~cm$^{-1}$ mode, which shows some signs of narrowing at low
dopings. The steadily increasing widths may be a reflection of the O(1)-site
disorder at low dopings. However, if there is some long-range ordering of the
O(1) sites at low dopings, as is expected in the Ortho~II phase, then there
should be some narrowing observed in the lines. While there is some evidence
of this in the 155~cm$^{-1}$ mode, the 279 cm$^{-1}$ mode shows no signs of
narrowing at low-oxygen dopings.
The oscillator strengths are a convenient guide by which to follow the effects
of the removal of oxygen from the O(1) sites. In particular, this is true for
the 279~cm$^{-1}$ mode, which is almost exclusively an O(1)-oxygen vibration.
In the $x=0.5$ material, we expect that the O(1) sites should be half empty. At
295~K, the effective-plasma frequency of this mode decreases from $\approx 310$
cm$^{-1}$ to $\approx 200$ cm$^{-1}$ in the $x=0.5$ material; taking the
na\"{\i}ve view that the ratio of the oscillator strengths ($\propto
\omega_p^2$) of 0.44 should be a reflection of the half-empty occupancy of the
O(1) sites, we find good agreement.
The asymmetric line shapes of the 570 and 610~cm$^{-1}$ modes require that they
be fitted using a rotated Lorentzian. While this line shape generally satisfies
oscillator sum rules for small rotations, it has been pointed out that for
large rotations ($\theta > 0.4$ rad) this is no longer true. The values for
$\omega_p$ may generally have increasingly large errors associated with them
for large rotations. The results show a decreasing oscillator strength for the
570 cm$^{-1}$ with decreasing doping, and the 610~cm$^{-1}$ mode show a
steadily increasing oscillator strength with decreasing doping with the two
oscillator strengths becoming equal at $x=0.7$. Below $x\approx 0.6$, the 610
cm$^{-1}$ line appears to be decreasing in intensity. This latter result is
consistent with the small value for $\omega_p$ observed in the tetragonal
YBa$_2$Cu$_3$O$_6$ material.
\subsection{Temperature dependence of the phonon parameters}
The temperature dependence of the phonon parameters for the three
high-frequency modes in YBa$_2$Cu$_3$O$_{6.95}$ is shown in Fig.~\ref{phona},
the results are in general agreement with previous work
\cite{Bonn87b,Wittlin87a,Litvinchuk92,Macfarlane87}. The temperature
dependence for the four high-frequency modes in YBa$_2$Cu$_3$O$_{6.7}$ is shown
in Fig.~\ref{phonb}.
In the highly-doped material, as the temperature is lowered, the phonon
frequencies generally increase due to anharmonic effects, but there are
exceptions, most notably the 279 cm$^{-1}$ chain-oxygen mode, which softens by
$\approx 6$~cm$^{-1}$ between 295 and 10~K. The two low-frequency modes
display little temperature dependence. The mode at $\approx 155$ cm$^{-1}$
hardens slightly from 153 cm$^{-1}$ at room temperature to 155 cm$^{-1}$ at
10~K. The feature at $\approx 194$ cm$^{-1}$ softens to 193 cm$^{-1}$ at 10~K;
this feature shows a slight anomaly at $T_c$.
The three remaining modes at 279, 312 and 570 cm$^{-1}$ are all associated with
oxygen vibrations, and show a more dramatic behavior. The 312 and
570~cm$^{-1}$ modes both harden with decreasing temperature, but show a
discontinuity and soften below $T_c$. The peak at $\approx 279$ cm$^{-1}$,
originally associated with the oxygen atoms in the CuO$_2$ planes, has been
reassigned as an O(1) chain-oxygen vibration \cite{thomsen91}; this mode
softens continuously from room temperature to below $T_c$, and shows only a
weak anomaly at $T_c$. However, like the 570 cm$^{-1}$ mode, the 279 cm$^{-1}$
mode shows some signs of hardening below $\approx 40$~K. The line widths of
the 279 cm$^{-1}$ mode shows only a weak temperature dependence, while the line
widths of the 312 cm$^{-1}$ and 570~cm$^{-1}$ modes both narrow with decreasing
temperature; interestingly, none show any sign of anomalous behavior at $T_c$.
There is substantial variation in the oscillator strengths with temperature.
The 279 cm$^{-1}$ feature weakens considerably with decreasing temperature, but
then becomes much stronger below $T_c$, while the 312 cm$^{-1}$ remains
relatively constant. The O(4) mode at 570 cm$^{-1}$ shows a rapid increase in
oscillator strength at $T_c$, and initially becomes more asymmetric below
$T_c$, suggesting the possibility of electron-phonon coupling (all the other
modes have been fit using simple Lorentzian oscillators).
The sharpness of the phonons in the highly-doped phase, especially at the low
frequency modes at 155 and 279 cm$^{-1}$, where anharmonic contribution to the
width is small, is a sign of good crystal quality. In previous single crystal
work with polished samples, these features were broad and weak, suggesting
either poor oxygen homogeneity, or that polishing the sample has induced
surface damage or altered the oxygen content of the material in the surface
layers, particularly in the chains. Furthermore, an examination of the
reflectance of polished {\it c}-axis crystals \cite{koch90,koch90b} shows that
they bear a strong similarity to the reflectance of freshly-annealed ceramics.
The 194~cm$^{-1}$ mode, according to the shell model, has its eigenvectors
confined to plane region of the unit cell, and therefore does not split when
alternate chains become deoxygenated, as expected.
The vibrations at 279 and 312 cm$^{-1}$ in the highly-doped material harden and
soften slightly with decreasing oxygen content, respectively, while also
broadening considerably.
The temperature dependence of the phonon parameters in the oxygen-reduced
system is dramatically different from the fully-doped case. As
Fig.~\ref{phonb} shows for YBa$_2$Cu$_3$O$_{6.70}$ there is now an additional
mode at $\approx 610$ cm$^{-1}$. The two apical-oxygen modes at 557 and 630
cm$^{-1}$ harden continuously with decreasing temperature, and there is no sign
of a phonon ano\-ma\-ly at the superconducting transition temperature. As noted
by Lit\-vin\-chuck \cite{Litvinchuk92} the plane feature at $\approx 318$
cm$^{-1}$ displays very interesting behavior in that it begins to soften
rapidly at $\approx 100$~K, well above the $T_c$ of 63~K. The chain feature at
$\approx 285$ cm$^{-1}$ at room temperature softens to $\approx 279$ cm$^{-1}$
at low temperature, none of these features shows any discontinuity at $T_c$.
The behavior of the line widths is similar to that of the oscillator strengths.
The strengths of the two high-frequency modes both begin to decrease at
$\approx 140$~K, well above $T_c$, as does the strength of the $\approx 312$
cm$^{-1}$ mode. This is in sharp contrast to the $\approx 279$~cm$^{-1}$ mode,
which loses strength with decreasing temperature until $T_c$, at which point it
begins to increase very rapidly, this is the only feature which shows any
sensitivity to $T_c$.
The rapid loss of oscillator strength, particularly in the 312 cm$^{-1}$ mode
in the $x=0.7$ material is even more noticeable in the $x\lesssim 0.6$
materials: in the $x=0.5$ material, this mode appears to have completely
collapsed at 10~K. The mode at $\approx 410$ cm$^{-1}$ (or combination of
modes), which is very weak in the $x=0.95$ material, is gradually becoming
stronger as the oxygen content is decreased, so that in the oxygen-reduced
$x=0.6$ material, it is one of the dominant features. Furthermore, as the
other oxygen related vibrations are losing oscillator strength, this feature is
becoming correspondingly stronger, as Fig.~\ref{lump} illustrates.
Interestingly, optical sum-rule calculations show no net increase or decrease
in the spectral weight in the $250-700$ cm$^{-1}$ region, suggesting that the
$\approx 400$~cm$^{-1}$ feature(s) is growing as a result of the transfer of
oscillator strength for the other modes, but in particular from the planar
copper-oxygen vibrations. The $\approx 400$ cm$^{-1}$ feature is identifiable
well above $T_c$, and appears to grow smoothly, and without any sudden
discontinuity at $T_c$. Furthermore, the 312 cm$^{-1}$ mode loses a great deal
of oscillator strength, and begins to do so well above $T_c$. This suggests
that while there is a major reorganization of the vibrational energy in the
CuO$_2$ planes (in the oxygen-reduced materials) as the temperature is lowered;
this is a separate mechanism which may affect superconductivity in these
materials, but does not appear to be affected by the superconducting
transition. Interestingly, similar behavior is observed in the high-frequency
copper-oxygen vibrations in cupric oxide (CuO) near the N\'eel transition
\cite{homes95}.
\begin{figure}[t]
\caption{The optical conductivity ($\sigma_1$) of YBa$_2$Cu$_3$O$_{6.60}$ along
the {\it c}-axis as a function of temperature, from 200 to 700~cm$^{-1}$, for
four temperatures each above and below $T_c$ ($\approx 58$~K). Note the
dramatic loss in the strength of the feature at $\approx 318$ cm$^{-1}$, and
the corresponding increase in the feature (or group of features) centered at
$\approx 400$ cm$^{-1}$, there is no anomalous behavior near $T_c$. }
\vspace*{-0.5cm}%
\hspace*{-0.4cm}%
\centerline{\includegraphics[width=10.5cm]{figure9.eps}}%
\vspace*{-1.4cm}%
\label{lump}
\end{figure}
\section{Discussion}
\subsection{The assignment of the 610 cm$^{-1}$ phonon}
The most striking change at room temperature with doping, is the growth of the
610~cm$^{-1}$ mode at the expense of the 570~cm$^{-1}$ mode. This
610~cm$^{-1}$ feature has been observed in other work and has been the subject
of some controversy, since there is lack of agreement in the Raman community on
the presence of the corresponding Raman mode at 600 cm$^{-1}$ \cite{burns91}.
In the tetrahedral YBa$_2$Cu$_3$O$_6$ material, which lacks the O(1) oxygen
chains, there is an infrared active $A_{2u}$ mode that is observed at $645-
650$~cm$^{-1}$ which, like the 573 cm$^{-1}$ mode in the orthorhombic phase,
has been identified with the O(4) oxygen \cite{burns88}. In removing the chain
oxygens, the four-fold coordinated Cu(1) ``squares'' become two-fold
coordinated ``sticks'' \cite{burns91}. The change in coordination number is
accompanied by a decrease in the Cu(1)--O(4) distance, and a resulting
hardening of the mode causing it to move from 570 to 610 cm$^{-1}$
\cite{burns91}. The feature at $\approx 610$ cm$^{-1}$ therefore appears to be
an $A_{2u}$ mode native to the tetrahedral phase involving the
O(4)--Cu(1)--O(4) sticks.
In accord with this picture we find that the oscillator strength of the two
modes, at 570 and 610~cm$^{-1}$, are nearly identical in the $x=0.70$ material.
One expects the ideal Ortho~II structure to occur at $x=0.50$ where there are
an equal number of two and four-fold coordinated coppers with every other chain
complete and every other chain empty. Also, the temperature dependence of the
various mode parameters seen in Fig.~\ref{phonb}, is nearly identical for the
two modes confirming that they are located on nearly identical structural
units.
Some recent Raman work does show intensity variation with doping that parallels
our work with weakening of the 500~cm$^{-1}$ line as the 600 cm$^{-1}$ line
grows. The two lines reach lines of equal strength at 500 and 600 cm$^{-1}$ in
the oxygen reduced, $x=0.75$ material.\footnote{M.N. Iliev. Private
communication.} Surprisingly, this is observed for the \boldmath $E\parallel b$
\unboldmath polarization.
Neutron measurements, too, find a mode in oxygen-reduced material $x=0.25$ at
\boldmath $q=0$ \unboldmath at the frequency that is consistent with Burns'
model of a high frequency mode for two-fold coordinated Cu(1).\footnote{N.
Pyka. Unpublished results quoted in ref. \citen{Pintschovius94}.} It should
also be noted that the neutron density of states determined by incoherent
scattering shows a high frequency band of modes in the $x=0$ phase centered at
650 cm$^{-1}$, whereas the fully-doped material has its highest branch at
$\approx 575$ cm$^{-1}$ \cite{Renker88}, in accord with the notion of a set of
high-frequency modes associated with the oxygen-reduced material.
\subsection{The 400~cm$^{-1}$ mode}
Another trend with decreasing doping is the appearance and growth of a broad
band at $\approx 400$ cm$^{-1}$ at low temperature. The band also shifts to
lower frequencies as the doping is reduced. As shown in Fig.~\ref{sigma10} and
in Fig.~\ref{lump}, this unusual feature increases in strength in the normal
state as the temperature is lowered, but does not show anomalous behavior at
$T_c$. At very low dopings, ($x \lesssim 0.6$) where this band is quite
strong, it appears to gain oscillator strength at the expense of the 312, 570,
and 610 cm$^{-1}$ phonons. At room temperature where the phonons are sharp and
well defined, the total spectral weight of the in the $250 - 700$~cm$^{-1}$
range, is equal to the total spectral weight, including the 400 cm$^{-1}$ mode,
at low temperature. This conservation of spectral weight is consistent with
the idea that the broad peak at $\approx 400$ cm$^{-1}$ is indeed a phonon.
The broad, weak mode at $\approx 408$ cm$^{-1}$ has been observed in ceramic
spectra of samples that appear to have been oxygen reduced \cite{Bonn87b}, but
this mode does not correspond to any of the calculated $B_{1u}$ normal modes
\cite{bates87,kress88}. A similar mode is seen in YBa$_2$Cu$_4$O$_8$
\cite{basov94c} as well as in Pb$_2$Sr$_2$RCu$_3$O$_8$ \cite{reedyk94}. The
common element in the three compounds is the third copper layer with two-fold
coordinated oxygen bonds. In all three systems the mode appears at low
temperature, growing in strength at the expense of both in-plane mode at 312
cm$^{-1}$ as well as the two apical-oxygen modes. This behavior suggests that
an unusual transition is taking place that involves several atoms in the unit
cell. The transition may be related to the zone-boundary mode reported on by
Reichardt et al. \cite{reichardt89}; located at approximately 400 cm$^{-1}$, it
was found to be anomalously broad and involved the ``breathing'' motions of the
ions around the planar copper. This mode is not normally infrared active, but
could become so as a result of some symmetry-breaking process. The mode
involves the {\it c}-axis motion of the O(4), but not the in-plane motion of
the planar oxygens. Its invocation would explain only half of our
observations, namely the loss of O(4) spectral weight to the new mode through
an electron-phonon process of the type discussed in connection with organic
molecules by Rice \cite{Rice80}. In such a process a symmetry-breaking
transformation allows totally-symmetric vibrations (here the zone-boundary
breathing mode) to become optically active. A necessary ingredient of the
process is the presence of low-lying charge-transfer bands. The excess spectral
weight and the asymmetric shape of the O(4) bands suggests that charge is being
pumped at the O(4) frequency between chains and planes. This process is already
active at high temperature above the formation temperature of the pseudogap. It
would then not be unreasonable to anticipate that when the symmetry-breaking
transition sets in at 150 K in the oxygen-reduced materials, that some of this
phonon-driven charge is transferred from the O(4) and plane modes, to the newly
activated breathing mode. The difficulty with this argument is that it is does
not apply to the other two compounds where this mode is observed
\cite{basov94c,reedyk94}.
Possibly related to this unusual phonon, polarized in the c-direction, is the
phenomenon in the {\it ab}-plane conductivity reported on by Reedyk and Timusk
where, in YBa$_2$Cu$_3$O$_{6.95}$ the {\it c}-axis longitudinal modes
interacted strongly with the {\it ab}-plane conductivity
\cite{timusk91,reedyk92e}. The strongest of these was a broad phonon with the
longitudinal frequency of 440 cm$^{-1}$.
The structural transition giving rise to the phonon at $\approx 400$ cm$^{-1}$
band appears to be related to the pseudogap seen at $\approx 280$ cm$^{-1}$ in
underdoped materials. It is strong in underdoped materials where the pseudogap
is also well developed. In Zn-doped YBa$_2$Cu$_4$O$_8$, both the pseudogap and
the 400~cm$^{-1}$ mode disappear at a 1.7\% doping level.\footnote{D.N.~Basov,
T.~Timusk and B.~Dabrowski (unpublished)} However, there are some differences.
Whereas the pseudogap appears gradually, in case of the $x=0.6$ material at a
temperature well above 300~K, the 400~cm$^{-1}$ band forms quite suddenly at
$\approx 150$~K, as shown by the sudden loss of O(4) spectral weight below this
temperature.
In summary, at this stage we have only the vaguest understanding of the nature
of the phonons giving rise to the band at 400~cm$^{-1}$, it is a feature that
appears at low temperature in oxygen-reduced materials only, and gets its
oscillator strength from plane-buckling and the O(4) modes. Its frequency
coincides with the anomalous broad mode seen in neutron scattering, and a broad
minimum in the {\it ab}-plane conductivity. Zn doping of YBa$_2$Cu$_4$O$_{8}$
seems to interfere with the formation of this mode. It should be noted that it
is not the result of a simple structural transition, since the {\it
frequencies} of the modes involved change little; there is no soft-mode
behavior. It appears that the transition is electronic in nature, the main
result being the shift of effective charges of the phonons, not their
frequencies.
\section{Conclusions}
The phonons in the oxygen-reduced systems provide a richly detailed structure.
At high dopings, there are five strong $B_{1u}$ modes. At low dopings
($x\lesssim 0.7$), many of the lines split into a number of different
components. In particular, the high-frequency apical-oxygen mode splits into
two components; the low-frequency component is associated with the four-fold
coordinated copper atom, and the high-frequency component with the two-fold
coordinated copper. What is most unusual is the development of a broad
feature, centered at $\approx 400$~cm$^{-1}$. This feature, attributed to a
(unknown) phonon, becomes more pronounced as oxygen is removed from the system;
it grows in strength as the temperature is lowered, appearing to draw
oscillator strength from the vibrations at $\approx 320$, 570 and 610~cm$^{-1}$
with decreasing temperature. This behavior develops well above $T_c$, and other
phonon anomalies are seen in the $x=0.7$ material at $\approx 150$~K.
Since these massive shifts of spectral weight to the 400~cm$^{-1}$ mode are not
seen in other high-temperature superconductors, such as La$_{2-x}$Sr$_x$CuO$_4$
and Bi$_2$Sr$_2$CaCu$_2$O$_{8+ z}$ \cite{Tamasaku94,Tajima94}, they may be
related to some special structural property of materials with a third copper
layer in the charge reservoir area of the unit cell, and not a fundamental
property of the cuprates.
\section*{Acknowledgements}
We would like to thank A.~Bianconi, J.C.~Irwin, W.~Reichardt, and J.~R\"ohler
for stimulating discussions. This work was supported by the Natural Sciences
and Engineering Research Council of Canada and the Canadian Institute for
Advanced Research.
|
\section{Introduction}
Thanks to the work done in expressing vector bundles, forms, integration,
etc., on locally compact topological spaces $X$, entirely in terms of the
algebra $C(X)$ of complex continuous functions on $X$ vanishing at infinity
which forms a commutative $C^{\ast}$ algebra; a generalization of ordinary
geometry can be introduced. Namely, when expressed in terms of a $C^{\ast}$
algebra the above cited notions make sense even when the $C^{\ast}$ algebra
is not commutative, therefore not of the form $C(X)$ \cite{Co}. The simplest
non-commutative geometries that have been studied are non-commutative and
non-cocommutative Hopf algebras, corresponding to both quantization and
curvature.\
Meanwhile in classical mechanics states are points of a manifold $M$ and
observables are functions on $M$; in the quantum case, states are
one-dimensional subspaces of a Hilbert space $H$ and observables are operators
in
$H$. Observables, in both classical and quantum mechanics, form an
associative algebra, which is commutative in the classical case and
non-commutative in the quantum case. So, we can think of quantization as a
procedure that replaces the classical algebra of observables by a
non-commutative quantum algebra of observables. The non-commutative Heisenberg
algebra, i.e. the algebra that comes up because momentum and space are not
simultaneously measurable (so called Heisenberg's uncertainty principle), is
the best example to illustrate this idea. Generally speaking, it is expected
that even using non commutative geometry, one might nevertheless extend our
regular notions of symmetry to the quantum world. If we consider the space of
states endowed by a group structure, the functions on this are observables. To
quantize such a system one has to construct a non-commutative associative
algebra of functions on a locally compact topological group space; i.e. a
quantum group \cite{Dr}.\
Thinking about quantization of the space-time metric itself, where we cannot
use path integration techniques to express quantization in terms of classical
fields; we claim the assumption of a smooth manifold structure for space-time
to be meaningless in extremely small scales from the experimental viewpoint.
The problem is that the finer the accuracy in the observation we ask for, the
heavier the test particle we need; eventually the space-time curvature due to
both the test particle and the space-time itself can be of the same magnitud.
In this context, by relaxing the assumption of smoothness of the space-time
manifold and introducing non-commutative algebraic geometry, we propose
a scheme called $q$-regularization, so we can regulate relevant quantities
in field theory before renormalising. $q$ (being $q^2\neq -1$) parametrizes the
deformation to the non-commutative and
non-cocommutative framework in which relevant quantities in quantum field
theories
are finite for $q\neq 1$, and reduce to the unregulated, divergent, physical
theory as $q\rightarrow 1$. Namely, as well as in dimensional regularization
we interpolate consistently to dimension $4-\epsilon $ where the relevant
quantities are finite (these would be infinite at dimension four); in
$q$-regularization we extend relevant quantities in quantum field theory to a
non-commutative and non-cocommutaive Hopf algebra or quantum group (by
introducing the parameter $q$ ) where the relevant quantities are finite
(these would be infinite at $q=1$; i.e. in $C(X)$, the commutative limit).\
We present two examples, the first one is constructed in a four dimensional
representation of a particular non-commutative space previously
reported \cite{Mj2}. The second example is proposed having in mind
$q$-spinors (two dimensional objects with the generators of $A^{2/0}_{q}$,
Manin's quantum plane \cite{Ma}, as entries) constructed by the projective
representation of the Heisenberg algebra, they are braided in a very specific
way to obtain a $q$-deformed space.\
Second example is intended as a first step to approach $q$-regularization in
$q$-Minkowski space-time. We work out this example in a $q$-deformed space
which can be related with both, first example's $q$-mutator algebra and
previously reported \cite{Ca1}\cite{Ca2} braided two copies of Manin's
quantum planes. Since we do not impose reality conditions, among others, we
are not working in anyway in $q$-Minkowski space-time.\
For the second example we want to learn more about the symmetries of our
measure, we study a projection in the $q$-deformed space used and its
relation to the $SU_{q}(2)$ measure. Moreover, we analyze the null directions
of the corresponding Hopf algebra that lead to a $q$-deformed Galilei group.\
This paper is organized as follows; in Section 2 we construct the Manin
quantum plane out of the non-commutative Heisenberg algebra and introduce
the $q$-spinors as a way to link $q$-regularization scheme with physically
meaningful concepts. In section 3, we present two examples of
$q$-regularization on $q$-deformed Euclidean spaces for $\lambda\phi^4$
theory. Our scheme can only be carried out in a very particular basis for
functions defined on the $q$-deformed spaces chosen such that we end with a
Haar weight that reduces to an ordinary integration. Further work should be
done to generalize this. Finally, in order to learn about desired properties
of symmetry in this $q$-regularization we study the zero time projection of
the measure we have just introduced in second example in terms of the
$SU_{q}(2)$ measure and the null directions of the Hopf algebra that lead to
a $q$-deformed Galilei group. The quantum Galilei group has been found as
symmetry in condensed matter \cite{Bo}.\
\section{From Heisenberg algebra to $q$-spinors.}
The goal of this section is to link non-commutative Heisenberg algebra with
two co-cycles and $q$-spinors as defined by Manin \cite{Ma}. Let us start by
the fundamental Heisenberg commutator algebra generated by translations on
phase space $({\bf r},{\bf p})$;
\begin{eqnarray}
&&\left[r^{i},p^{j}\right] = i \hbar \delta^{ij} \label{line1} \\
&&\left[r^{i},r^{j}\right] = \left[p^{i},p^{j}\right]=0 .\nonumber
\end{eqnarray}
We propose the following translation operator on phase space;
\begin{equation}
U({\bf a} ,{\bf b} )=e^{i ({\bf a}\cdot {\bf p} -
{\bf b} \cdot {\bf r} )/ \hbar }
\mbox{ where ${\bf a }$ and ${\bf b }\in R^n$ .}
\end{equation}
In a ray or projective representation, eq(2) obeys the composition law
\cite{Dj}.\
{\large\begin{equation}
U({\bf a}_{2},{\bf b}_{2})\cdot U({\bf a}_{1},{\bf b}_{1})= e^{ \left[2\pi
i \alpha_{2}({\bf r} ;
({\bf a}_1 ,{\bf b}_1),({\bf a}_2 ,{\bf b}_2))\right]}\cdot U({\bf a}_1
+{\bf a}_2 ,{\bf b}_1 +{\bf b}_2 ),\\
\end{equation}}
where ${\bf a}_{1},{\bf b}_{1},{\bf a}_{2},{\bf b}_{2}\in R^n$ and, for a
free particle in quantum
mechanics, the two co-cycle
$\alpha_2$ for translations in the phase space is given by
\begin{equation}
2 \pi \alpha_{2}\left({\bf r} ;
({\bf a}_1 ,{\bf b}_1),({\bf a}_2 ,{\bf b}_2)\right) =
\frac{1}{2\hbar}({\bf a}_1 \cdot {\bf b}_2 -{\bf a}_2\cdot {\bf b}_1 ) .\\
\end{equation}
Let us now consider the following infinitesimal Galilei transformation
\cite{Dj}
\begin{eqnarray}
{\bf r'}={\bf r}+{\bf a}_1={\bf r} +\hbar {\bf u} & & {\bf r''} ={\bf r} +
{\bf a}_2 ={\bf r} \\
{\bf p'}={\bf p} + {\bf b}_1 ={\bf p} & & {\bf p''}={\bf p}+{\bf b}_2 =
{\bf p} +\hbar {\bf u}, \nonumber
\end{eqnarray}
\nonumber
where {\bf u }is a unit vector in $R^{n}$.\
If we define
{\Large\begin{equation}
q=e^{-i\hbar} ,
\end{equation}}
impose eq(5) as symmetry in eq(4) and substitute the result in eq(3),
it is straightforward to prove that
\begin{equation}
U(\hbar\bf u ,\rm0)U(0,\hbar{\bf u})=qU(0,\hbar {\bf u})U(\hbar \bf u,\rm 0)
\end{equation}
is a realization of $A_{q}^{(2/0)}$; i.e. this fulfills the non-commutative
algebra of the Manin's quantum plane \cite{Ma}.\
Like other authors \cite{Ca1}\cite{Ca2}, we call the following two dimensional
object a $q$-spinor (more properly Weyl $q$-spinor).
\begin{equation}
Z^{\rho}=
\left[
\begin{array}{c}
Z^1 \\
Z^2
\end{array}
\right]=
\left[
\begin{array}{c}
U(\hbar {\bf u} ,\rm 0)\\
U(0,\hbar {\bf u})
\end{array}
\right]
\mbox{, i.e. $\rho =1,2$ . }
\end{equation}
In example 2 we use an approach \cite{Ca1} \cite{Ca2} in which the
$q$-deformed space can be related with the tensor product representation of
two $q$-spinor spaces called $(Z^{i} ,\tilde{Z}^{i})$. A pair of $q$-spinors
$(i=1,2)$ is introduced in each space.
Hereafter greek indices are for spinor suffix and roman ones for different
spinors. Besides it is required the following braiding
\begin{equation}
Z^{i}\tilde{Z}^{j}= \hat{R}^{ij}_{j'i'}\tilde{Z}^{j'}Z^{i'}
\end{equation}
where $\hat{R}^{ij}_{j'i'}$ is the Yang-Baxter matrix for $ SL_{q}(2,C)$.
\section{Examples on $q$-regularization}
In this section we present two examples of $q$-regularization for
$\lambda\phi^4$
theory on two apparently different $q$-deformed spaces, both Euclidean. The
first case involves a four dimensional version of a Hopf algebra previously
reported \cite{Mj2}; we propose to extend momenta internal to Feynman loops
to a non-commutative structure. Second example involves a braided 4-dimensional
representation of Manin's quantum plane
(so called $q$-spinors) where some particular transformations on the
generators of this Hopf algebra relates to the one used in example 1.
Actually, example 1 is posed in order to better explain example 2 which is
considered as a preliminary step for formulating $q$-regularization on
$q$-Minkowski space-time.\
{\bf EXAMPLE 1}. From reference 3, let us consider the Hopf algebra ${\it L}$
generated by $\left(l_1,l_2,l_3,l_4\right)$ and
\begin{equation}
\left[l_k,l_j\right]=il_jQ', \mbox{ for k=2,4 and j=1,3}
\end{equation}
where $Q'=\sqrt{1-q}$. Define on this, the antipode map as
\begin {equation}
S\left(l_k\right) = -l_k \;\;\;\;\;
S\left(l_j\right) = -q^{-i}l_jq^{-l_k/Q'},
\end{equation}
the coproduct map is given by
\begin{equation}
\bigtriangleup l_k = l_k\otimes 1+1\otimes l_k \;\;\;\;
\bigtriangleup l_j = l_j\otimes 1+q^{\frac{l_k}{Q'}}\otimes l_j
\mbox{ and}
\end{equation}
the counit
\begin{equation}
\epsilon\left(l_k\right)=\epsilon\left(l_j\right)=0.
\end{equation}
Additionally, ${\it L}$ can become a $C^{\ast}$ algebra if we define
\begin{equation}
l^{\ast}_k=l_k\;\;\;\;\;l^{\ast}_j=l^{\ast}_jq^{i/2}
\end{equation}
For every finite-dimensional Hopf algebra there is an invariant integration,
the Haar weight $\int$, unique up to normalization.\
A basis $B^{a_1,...a_4}=e^{ia_1l_1}...e^{ia_4l_4}$ where $a_n\in C$, being $C$
the complex is chosen, then the dual basis $D_{a'_1,...a'_4}$ is given via
\begin{equation}
B^{a_1...a_4}D_{a'_1...a'_4}=\delta(a'_1-a_1)...\delta(a'_4-a_4),\;\;\;
a'_n\in C
\end{equation}
where the Dirac delta functions $\delta$ have been defined with respect to the
usual Lebesgue integration, then it is straightforward, by analogy with the
case of finite dimensional Hopf algebras \cite{La}, to prove \cite{Su}
\begin{equation}
\int\int f=\left[2\pi\delta(0)\right]^{k}\int
\prod_{j}da'_jf'(0,a'_j(1-q^{-i}))
\end{equation}
for all $j$ and $f$ suitable of being written on the basis $B^{a_1...a_4}$.
In case $q\neq 1$ and assuming proper analycity and decay of $f'$
(the Fourier transform of the Wick ordered function $f$), eq(16) might be
finite for suitable $f$. If $q=1$ eq(16) certainly diverges.\
We propose, from $\lambda\phi^4$ theory, to $q$-regularize the vertex
corrections with
contributions given by
\begin{equation}
\Gamma(s)=\frac{\left(-i\lambda\right)^2}{2}\int\int\frac{d^4l}{(2\pi)^4}
\frac{i}{(l-p)^2-\mu^2_0-i\epsilon}\frac{i}{l^2-\mu^2_0+i\epsilon}
\end{equation}
where s is any of the Mandelstam variables. These corrections diverge
logarithmically.\
Let us extend the internal momentum in the Feynman loop in $\Gamma(s)$ to the
non-commutative algebraic framework by considering instead of the standard
Lebesgue integration, the Haar weight above defined on the basis
$B^{a_1...a_4}$, thus;
\begin{equation}
\Gamma_q(s)=
\end{equation}
$$
\frac{\lambda^2_0\delta(0)}{2(2\pi)^3}
\int\frac{d^jl'_j}
{\left[p^2_k+\left(l'_j(1-q^{-i})-p_j\right)^2-
\mu^2_0+i\epsilon\right]\left[\left(l'_j(1-q^{-i})\right)^2-\mu^2_0+
i\epsilon\right]}
$$
where $l'_j$ are the odd components of the dual internal momentum that was
extended to non-commutative geometry and $p_k(p_l)$ are the even (odd)
components of the external momentum in standard Euclidean commutative four
dimensional space-time. Unless $q=1$, eq(18) is finite, thus we have a
regularization scheme. An additional attempt of $q$-renormalization has
recently been presented \cite{Su}. Since we extend to the non-commutative
framework only the internal momenta degrees of freedom, the lack of locality,
consequence of this extention has not experimental consequences in this case.
\cite{Fr}.\
{\bf EXAMPLE 2}. Let us consider the Hopf algebra ${\it H}$ generated by 1 and
$(a,{\bar a},b,{\bar b})$ such that;
\begin{equation}
\left[b,{\bar b}\right]=0,\;\; \;\;\;
\left[a,{\bar a}\right]=2(q^{-1}-q)q^{\frac{1}{2Q'}(b+3{\bar b})}
\end{equation}
$$
\left[{\bar b},a\right]=\left[b,a\right]=2Q'{\bar a},\;\;\;\;\;
\left[{\bar b},{\bar a}\right]=\left[b,{\bar a}\right]=2Q'a .
$$
The coproduct map $\bigtriangleup $ in this Hopf algebra is
\begin{equation}
\bigtriangleup a=a\otimes 1+q^{\frac{b}{Q'}}\otimes a \;\;\;\;\;
\bigtriangleup b=b\otimes 1+1\otimes b
\end{equation}
$$
\bigtriangleup {\bar a}={\bar a}\otimes 1+q^{\frac{{\bar b}}{Q'}}\otimes
{\bar a} \;\;\;\;\;\bigtriangleup {\bar b}={\bar b}\otimes 1+1\otimes
{\bar b},
$$
the antipode map $S$ is
\begin{eqnarray}
S(a) &=& \frac{1}{2}
\left\{-(q^{-2}+q^2)aq^{-\frac{b}{Q'}}+(q^2-q^{-2}){\bar a}
q^{-\frac{b}{Q'}}\right\} \\
S({\bar a}) &=& \frac{1}{2}
\left\{(q^2-q^{-2})aq^{-\frac{{\bar b}}{Q'}}-(q^{-2}+q^2){\bar a}
q^{-\frac{{\bar b}}{Q'}}\right\}\nonumber \\
S(b)=-b & & S({\bar b})=-{\bar b} \nonumber,
\end{eqnarray}
and finally the counit map $\epsilon $ is
\begin{equation}
\epsilon (a)=\epsilon({\bar a})=\epsilon(b)=\epsilon({\bar b})=0
\end{equation}
Furthermore, we can make this into a $\ast$-algebra via
$$
b^{\ast}=b,\;\;\;{\bar b}^{\ast}={\bar b}\;\;\;
a^{\ast}=aq^{i/2},\;\;\;{\bar a}^{\ast}={\bar a}q^{i/2}
$$
iff q is a primitive root of unity such that $q^4=1$.\
We would like to relate ${\it H}$ with
\begin{equation}
X^{ij}=\tilde{Z}^{i}Z^{j} \; \in A^{2/0}_{q}\otimes A^{2/0}_{q}\;\; i,j=1,2;
\end{equation}
where $\tilde{Z}^i$ and $Z^j$ where introduced in Section 2 (eq(8) and
eq(9)).
It is straightforward to prove that $A^{2/0}_q\otimes A^{2/0}_q$ is isomorphic
to the real algebra generated by 1 and (A,$\bar{A}$,B,$\bar{B}$), where
\begin{equation}
A=X+Y ,\;\;\;\bar{A}=X-Y ,\;\;\;B=Z+T ,\;\;\;\bar{B}=Z-T
\end{equation}
and
\begin{equation}
X=q^{-1/2}X_{11},\;\;\;Y=q^{-1/2}X_{12},\;\;\;Z=\frac{q^{-1}X_{21}-qX_{22}}
{\sqrt{q+q^{-1}}},\;\;\;T=\frac{X_{21}+X_{22}}{q\sqrt{q+q^{-1}}}.
\end{equation}
To relate $H$ with $A^{2/0}_q\otimes A^{2/0}_q$ let us rewrite the
$(A,\bar{A},B,\bar{B})$ generators, for $q\neq 1$, as follows\
{\large\begin{equation}
A=a ,\;\;\;\bar{A}=\bar{a} ,
\end{equation}
$$
B=q^{\frac{b}{Q'}},\;\;\;
\bar{B}=q^{\frac{\bar{b}}{Q'}}.
$$}
On the other hand, it is straightforward to prove that in ${\it H}$,
$((a+\bar{a})$, $(a-\bar{a})$, $b$, $\bar{b})$ corresponds to the algebra
${\it L}$ with generators $(l_1,l_2,l_3,l_4)$ defined in example 1. If
$q\rightarrow 1$, the algebra becomes the commutative algebra of functions on
the space generated by ($a,\bar{a},b,\bar{b}$) and the unit.\
Like in example 1, we proceed defining the Haar measure $\int \int$ as a map
${\it H}\rightarrow C$, such that
\begin{equation}
\int \int f=f_{(1)}\int \int f_{(2)}\;\;\;\forall f\in {\it H}
\end{equation}
Here we have expressed the action of $\bigtriangleup$ on f as
$\bigtriangleup f=f_{(1)}\otimes f_{(2)}$. We remark that it is
well known in the theory of Hopf algebras \cite{Mj1} that
eq(27) is the dual formulation of the usual left invariance.\
By analogy with the case of finite dimensional Hopf algebras \cite{La},
we use the following formal expression for eq(27)
\begin{equation}
\int \int f=Tr_{\it H}L_{f} S^2
\end{equation}
where $L_{f}$ stands for f acting by left multiplication on ${\it H}$.\
{}From eq(21) follows
\begin{equation}
S^2(a-{\bar a})=w^{-1}(a-{\bar a})\;\;\;;\;\;
S^2(a+{\bar a})=w^{-1}(a+{\bar a})\;\;\;;\;\;S^2b=b
\;\;\;\;;\;\;S^2\bar{b}=\bar{b}
\end{equation}
where $w^{-1}=f(q)$ and $lim_{q\rightarrow 1}w^{-1}=1$. This shall be used
below.\
To compute $\int \int$ we propose the following basis in ${\it H}$:\
{\large\begin{equation}
F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}=
(F^{\lambda_1 \lambda_2},F^{\lambda_3 \lambda_4},
F^{\lambda_5 \lambda_6})=
\end{equation}}
$$
\left(e^{i\lambda_1 \bar{b}}e^{i\lambda_2 \frac{(a-\bar{a})}{2}},
e^{i\lambda_3 \bar{b}}e^{i\lambda_4 \frac{(a+\bar{a})}{2}},
e^{i\lambda_5 \bar{b}}e^{i\lambda_6 b}\right)
$$
where
$$
F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}
\in {\it H}
\mbox{ and } (\lambda_1 ,\lambda_2 ,\lambda_3 ,\lambda_4 ,\lambda_5 ,\lambda_6)
\in R .
$$
We associate to
$F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
a dual basis
$F_{\lambda'_1 \lambda'_2 , \lambda'_3 \lambda'_4 , \lambda'_5 \lambda'_6}
\in (A^{2/0}_{q}\otimes A^{2/0}_{q})^{!}$
where $({A}^{2/0}_{q}\otimes A^{2/0}_{q})^{!}$ is the dual Hopf
algebra of ${A}^{2/0}_{q}\otimes A^{2/0}_{q}$, such that
\begin{equation}
F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}
F_{\lambda'_1 \lambda'_2 , \lambda'_3 \lambda'_4 , \lambda'_5 \lambda'_6}=
\left(
\delta (\lambda'_1-\lambda_1)\delta (\lambda'_2-\lambda_2),
\right.
\end{equation}
$$
\left.
\delta (\lambda'_3-\lambda_3)\delta (\lambda'_4-\lambda_4),
\delta (\lambda'_5-\lambda_5)\delta (\lambda'_6-\lambda_6)
\right)
$$
In basis eq(30) we have introduced six parameters $\lambda_i$, one for each
generator involved. They are dual variables to the non-commutative parameter.\
{\bf Theorem 1}. The Haar weight
$\int\int F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
defined in eq(28), for a basis
$F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
chosen as in eq(30), reduces to an ordinary integration.\
{\bf Proof}. From eq(19) we know that
$$
\left[\bar{b},(a-\bar{a})\right]=-2Q'(a-\bar{a})
$$
$$
\left[\bar{b},(a+\bar{a})\right]=2Q'(a+\bar{a})
$$
$$
\left[\bar{b},b\right]=0
$$
Note that $\frac{(a+\bar{a})}{2}=X$, $\frac{(a-\bar{a})}{2}=Y$ in eq(25).
Substituting the basis given by eq(30) in eq(28) and using the Glaube
formula for operators we obtain the ordinary integral
\begin{equation}
\int \int F^{\lambda_1 \lambda_2 ,\lambda_3 \lambda_4 ,\lambda_5 \lambda_6}=
\left(\int^{\infty}_{-\infty}d\lambda'_1d\lambda'_2\delta
(\lambda'_1-(\lambda_1+\lambda'_1))
\delta(\lambda'_2-(\lambda_2e^{-2i\lambda'_1Q'}+w^{-1}\lambda'_2)),
\right.
\end{equation}
$$
\int^{\infty}_{-\infty}d\lambda'_3d\lambda'_4\delta
(\lambda'_3-(\lambda_3+\lambda'_3))
\delta(\lambda'_4-(\lambda_4e^{2i\lambda'_3Q'}+w^{-1}\lambda'_4)),
$$
$$
\left.
\int^{\infty}_{-\infty}d\lambda'_5d\lambda'_6\delta
(\lambda'_5-(\lambda_5+\lambda'_5))
\delta (\lambda'_6-(\lambda_6+\lambda'_6))\right)
$$
Q.E.D.\
The basis $
F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
in eq(30) admits an expression in terms of the $q$-spinor defined in eq(8) out
of the projective representation for the Heisenberg algebra. Furthermore,
this basis can be rewritten in terms of $q$-Majorana spinors built using
$q$-Weyl spinors in analogy with the commutative algebraic formulation. As a
result of this we can show how does not matter if we think in terms of
integrating out non-commutative light cone coordinates, Weyl $q$-spinors or
Majorana $q$-spinors degrees of freedom; the result is exactly the same.
Furthermore, the Haar measure $\int \int$ defined on ${\it H}$ can be written
in terms of ordinary integration.\
{\bf Theorem 2}. For a suitable $f\in {\it H}$
that can be expressed on the basis
$F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
given in eq(30) (or any of their different $q$-spinor representations),
$\int\int f$ as defined in eq(28) contains a component that can be
$q$-regularized, i.e. is finite provided $q\neq 1$, but infinite in the
limit $q=1$.\
{\bf Proof.}\
It is straightforward to show that for any function f
defined on ${\it H}$ with basis
$F^{\lambda_1 \lambda_2 , \lambda_3 \lambda_4 , \lambda_5 \lambda_6}$
the following transformation holds
\begin{equation}
f=:f':=\int^\infty_{-\infty}d\lambda_1d\lambda_2\tilde{f}
(\lambda_1,\lambda_2)F^{\lambda_1 \lambda_2}+
\end{equation}
$$
\int^\infty_{-\infty}d\lambda_3d\lambda_4\tilde{f}
(\lambda_3,\lambda_4)F^{\lambda_3 \lambda_4}+
\int^\infty_{-\infty}d\lambda_5d\lambda_6\tilde{f}
(\lambda_5,\lambda_6)F^{\lambda_5 \lambda_6}
$$\
where we express f as a normal ordered form of f',
in terms of the generators. Namely putting $\bar{b}$ to the left of $a$,
$\bar{a}$ and b in the light cone coordinate approach;
$\{\sigma^3, \sigma^0\}$ to the left of $\{\sigma^+, \sigma^-\}$ in the Weyl
$q$-spinor formulation and finally $\{\gamma^3,\gamma^0\}$ to the left of
$\{\gamma^1,\gamma^2\}$ in the Majorana $q$-spinor basis. Here $(\sigma^0,
\sigma^3, \sigma^+, \sigma^-)$ and $(\gamma^0, \gamma^1, \gamma^2, \gamma^3)$
are $q$-deformed Pauli and Dirac matrices \cite{Ca2}. Additionally $\tilde{f}$
is the Fourier transform of f', i.e.\
{\large\begin{equation}
\tilde{f}(\lambda_i,\lambda_j)=(2\pi)^{-2}\int^\infty_{-\infty}
d\mu_id\mu_jf'(\mu_i \mu_j)e^{-i\mu_i\lambda_i}e^{-i\mu_j\lambda_j}
\end{equation}}
$$
i,j=(1,2),(3,4),(5,6).
$$
Then carrying on integration on $\lambda_1$, $\lambda_3$,
$\lambda_5$, $\lambda_6$ we obtain
{\large\begin{equation}
\int \int f=\int^\infty_{-\infty}d\lambda'_1d\lambda'_2d\lambda_2
\tilde{f}(0,\lambda_2)\delta(\lambda'_2(1-w^{-1})-\lambda_2e^
{-2i\lambda'_1 Q'})+
\end{equation}
$$
\int^\infty_{-\infty}d\lambda'_3d\lambda'_4d\lambda_4
\tilde{f}(0,\lambda_4)\delta(\lambda'_4(1-w^{-1})-\lambda_4e^
{-2i\lambda'_3 Q'})+
$$
$$
\int^\infty_{-\infty}
d\lambda'_5d\lambda'_6\tilde{f}(0,0),
$$ }
that after changing the order of integration and integrating on
$\lambda_2$ and $\lambda_4$ becomes
{\large\begin{equation}
\int \int f=\int^\infty_{-\infty}d\lambda'_1d\lambda'_2e^{\lambda'_1 Q'}
\tilde{f}(0,\lambda'_2(1-w^{-1})e^{2i\lambda'_1 Q'})+
\end{equation}
$$
\int^\infty_{-\infty}d\lambda'_3d\lambda'_4e^{-\lambda'_3 Q'}
\tilde{f}(0,\lambda'_4(1-w^{-1})e^{-2i\lambda'_3 Q'})+
$$
$$
\int^\infty_{-\infty}d\lambda'_5d\lambda'_6\tilde{f}(0,0)
$$}
The last term in eq(36) corresponds to the ordinary divergent term
that appears in the standard commutative algebraic formulation of quantum
field theory; there is no way we can recover a finite term out of this
in the limit $q\rightarrow 1$. Checking the non-commutative Hopf algebra
generated by $X^{ij}$ we find the reason why this happens to be so; T is
central with respect to $(X,Y,Z)$, so this part of the Haar measure is not
really defined on a non-commutative algebraic variety. Therefore we can
extract out of $\int\int f$ a $q$-regularizable part
\begin{equation}
\int \int f-\int^\infty_{-\infty}d\lambda'_5d\lambda'_6\tilde{f}(0,0)=
\end{equation}
$$
(2\pi \delta (0)) \left\{\int^\infty_{-\infty}d\lambda'_2
\tilde{f}(0,\lambda'_2(1-w^{-1}))+
\int^\infty_{-\infty}d\lambda'_4
\tilde{f}(0,\lambda'_4(1-w^{-1}) \right\}.
$$
But $lim_{q\rightarrow 1}w^{-1}=1$ thus, as $q\rightarrow 1$,
$\int \int f-\int^\infty_{-\infty}d\lambda'_5d\lambda'_6\tilde{f}(0,0)$
diverges, by contrast at $q\neq 1$ and assuming suitable analicity and decay
of $\tilde{f}$ to allow contour integration, eq(37) can be made finite
for suitable f; moreover, this is proportional to $(1-w^{-1})^{-1}$.\
In the limit $q\rightarrow 1$, the transformation described in eq(26) is non
sense, because in this limit the map ${\it H}\rightarrow {\it L}$ is
singular. We remark that this does not mean that the $q$-regularization
scheme performed on the algebra ${\it H}$ and described up to here is lacking
sense in case $q=1$ but only the map that relates this with
$A^{2/0}_q\otimes A^{2/0}_q$. We are interested in this map because might be
of some help in the future construction of $q$-regularization on $q$-Minkowski
space-time. Further work should be done in this direction.\
Q.E.D.\
For obvious reasons, the vertex correction for $\lambda\phi^4$ theory described
in example 1 is suitable of being $q$-regularized on ${\it H}$ as well as in
example 2 was on ${\it L}$. Further work should be done to generalize these
examples to more interesting cases. Since this scheme is strongly basis
dependant,
a complete analysis of the class of functions suitable of being $q$-regularized
on physically interesting basis is needed. Note that $q$-regularization may
be considered equivalent to dimensional regularization in a similar sense to
the McKane and Parisi-Sourlas case \cite{MPS}.\
\section{Comments and Remarks.}
In the paper wherein Woronowicz \cite{Wo} proves the existence and uniqueness
of the Haar measure, i.e. the unique state invariant under left
(and simultaneusly right) shifts, for any compact quantum group, he proposes
the following $q$-integration on $SU_q(2)$
\begin{equation}
\int^1_{q^0}{\bf f}=(1-q)\sum^\infty_{k=0} q^k{\bf f}(q^k)
\;\;\;\mbox{for any } {\bf f}\in SU_q(2),
\end{equation}
On the other hand, let us set $T=0$ in eq(25) and from eq(7), we get
\begin{equation}
U_2(\hbar {\bf u},0)=q^{\frac{1}{2}}\tilde{u}_1(0,\hbar {\bf u})
\;\mbox{ and }\;U_2(0,\hbar {\bf u})=-q^{-\frac{1}{2}}
\tilde{u}_1(\hbar {\bf u},0),
\end{equation}
this is equivalent to set ${\tilde Z}^{\rho}$$=\epsilon^{\rho\sigma}
{\bar Z}_{\sigma}$ in eq(9).
Thus
$$
A=X+Y,\;\;\;\bar{A}=X-Y\;\;\;B=\bar{B}=Z.
$$
We call $X^1=\frac{1}{2}(A+\bar{A})$,$X^2=\frac{1}{2}(A-\bar{A})$ and
$X^3=B$; then in terms of this $(X^1, X^2,X^3)$ 3-dimensional vector
representation we propose the following basis to be used;
\begin{equation}
F^{\lambda_1 \lambda_2 ,\lambda_3 \lambda_4}=
\left(
e^{i\lambda_1x^3}e^{i\lambda_2x^1},e^{i\lambda_3x^3}e^{i\lambda_4x^2}
\right)
\end{equation}
where $X^1=x^1$, $X^2=x^2$, and $X^3=q^{\frac{x^3}{Q'}}$, as was done in
eq(26).\\
{}From the work done on the category of representations of a Hopf algebra we
can write the action of any function f of the Hopf algebra $SU_q(2)$ on its
vector representation V, through the corresponding basis;
\begin{equation}
{\bf f}.e^{j}_{m}=\sum_{i=+,-,0}f^{i}{\bf e_i}.e^{j}_{m}\in V
\;\;\;\;\;\;\forall {\bf f}\in SU_q(2)
\end{equation}
where ${\bf e_{+}}$=$X^{+}$,${\bf e_{-}}$=$X^{-}$, ${\bf e_{0}}$=$H$ is the
$SU_q(2)$ basis and $f^{i}\in C$.\
{}From eq(41) is clear that the Woronowicz's map $\int {\bf f} \rightarrow C$
for the $SU_q(2)$ Haar measure induces a $\int \int f\rightarrow C$ map for
the vector representation of the Hopf algebra $SU_{q}(2)$, inducing another
for $\Lambda $ which is written in terms of $SU_q(2)$. We define the
$\Lambda$ matrix as
\begin{equation}
\Lambda^{(ij)}_{(i'j')} \equiv \tilde{M}^{i}_{i'}M^{j}_{j'}
\;\;\;\;;\;\;M\in SL_q(2,C),\; \tilde{M}\in \tilde{SL}_q(2,C)\;.
\end{equation}
Thus, the similarity of eq(34) and eq(38) can be understood in these terms.\
{}From these two facts and the obvious similarity of the $q$-integration
carried out in eq(34) and the one depicted in eq(38), we think that the
$q$-Time zero-projection in the $q$-deformed space defined for the second
example corresponding to the $\tilde{M}^{\dagger}=M^{-1}$ identification,
reduces $\int\int f-\int^{\infty}_{-\infty}d\lambda'_{5}d\lambda'_{6}
\tilde{f}(0,0)$ in Theorem 2 to the Haar weight on the vector representation
of $\Lambda$ written in terms of $SU_q(2)$.\
Finally, we show how null directions in $\Lambda$ can lead us to obtain the
quantum mechanical Galilei group. By imposing the following null bi-ideals
$$
u^1_2=0,\;\;\;u^2_1=0,
$$
$$
u^1_1u^2_2=u^2_2u^1_1=1,
$$
being $u^i_j\in M$ ( the same is asked for ${\tilde u}^i_j$ being
${\tilde u}^i_j\in {\tilde M}$), we obtain a direct product representation of
the quantum Galilei group.\
We can see this from the viewpoint of cohomological formalism. We construct
the quantum mechanical Galilei group, choosing the following Galilei
transformations
\begin{equation}
{\bf r'}={\bf r}+{\bf v}t\;\;\;\;\;{\bf p'}={\bf p}+m{\bf v}
\end{equation}
where m is the particle mass.\\
Then eq(2) transforms to
{\large\begin{equation}
U({\bf v})=e^{i\frac{{\bf v}\cdot ({\bf p}t-m{\bf r})}{\hbar}},
\end{equation}}
and its action on a wave function $\Psi({\bf r})$ introduces a phase
(one-cocycle) $\alpha_1$, i.e.
\begin{equation}
U({\bf v})\cdot \Psi({\bf r})=e^{2i\pi \alpha_1({\bf r};{\bf v})}
\cdot\Psi({\bf r}+{\bf v}t).
\end{equation}
We shall consider this one-cocycle as trivial, so
\begin{equation}
\alpha_1(\bf {r;v})=\delta \alpha_0=\alpha_0({\bf r'})-\alpha_0({\bf r})
\end{equation}
where $\alpha_0$ is a function, called 0-cocycle, which depends only on
{\bf r}.\
Therefore, the group law of the quantum mechanical Galilei group
for translations on phase space (or $U(1)$ extended Galilei group) is
expressed such that
\begin{equation}
e^{2i\pi \alpha_0({\bf r'})}=e^{2i\pi (\alpha_0({\bf r})+\phi+\alpha_1({\bf
r;v}))}
\end{equation}
where
\begin{equation}
2\pi \alpha_1=\frac{1}{\hbar}(m{\bf v}\cdot {\bf r}+\frac{1}{2}mv^2t)
\end{equation}
and $\phi$ is the central parameter of the quantum mechanical Galilei group.\
On the other hand, we require
\newline
$M=\left(\begin{array}{cc}
u^1_1 & u^1_2 \\
u^2_1 & u^2_2
\end{array}
\right)\in SL_q(2) \mbox{ and } $
$M=\left(\begin{array}{cc}
{\tilde u}^1_1 & {\tilde u}^1_2 \\
{\tilde u}^2_1 & {\tilde u}^2_2
\end{array}
\right)\in {\tilde SL}_q(2)$ to belong to the quantum mechanical Galilei
group; i.e. $M$ ( equivalently ${\tilde M}$) must fulfill eq(5). It is
straightforward to prove that, in this case, the following null bi-ideals
have to be imposed on $M$ (equivalently on ${\tilde M}$)
\begin{eqnarray}
u^1_2=0 & & u^2_1 = 0 \\
u^1_1u^2_2 & = & u^2_2u^1_1=1 ,\nonumber
\end{eqnarray}
so to end up with a group that has only one generator as should be.\
Besides, we can prove that the null bi-ideals once imposed on $M\in SL_q(2)$
(thereby defining the quantum mechanical Galilei group) produce the following
pairing
\begin{eqnarray}
<u^1_2,t^{\dagger k}_m>=R^{1k}_{2m}=0 \;\;\;k,m,s=1,2 \nonumber \\
<u^2_1,t^{\dagger k}_m>=R^{2k}_{1m}=0 \\
<u^1_1u^2_2,t^{\dagger k}_s>=R^{1k}_{1m}R^{2m}_{2s}=1 \nonumber \\
<u^2_2u^1_1,t^{\dagger k}_s>=R^{2k}_{2m}R^{1m}_{1s}=1 \nonumber
\end{eqnarray}
where $t^{\dagger k}_m \;\; (k,m=1,2)$ are generators of the dual Hopf
algebra for $SL_q(2)$ and $R^{ij}_{kl} \;\;(i,j,k,l=1,2)$ entries of
the Yang-Baxter matrix $R_{G}$ associated with the quantum mechanical
Galilei group. This does not determine $R_{G}$ but restricts the solution
to block diagonal matrices.\
Finally, if we impose the null-directions given by eq(50) in $\Lambda$ we
obtain the following representation of the quantum Galilei group \newline
{\large\begin {equation}
\Lambda=
\left(
\begin{array}{cccc}
(\tilde{u}^1_1)^{-1}u^1_1 & 0 & 0 & 0 \\
0 & \frac{\tilde{u}^1_1u^1_1+q^2(\tilde{u}^1_1u^1_1)^{-1}}{1+q^2} & 0 &
\frac {q^2(\tilde{u}^1_1u^1_1-(\tilde{u}^1_1u^1_1)^{-1})}{1+q^2} \\
0 & 0 & \tilde{u}^1_1(u^1_1)^{-1} & 0 \\
0 & \frac{\tilde{u}^1_1u^1_1-(\tilde{u}^1_1u^1_1)^{-1}}{1+q^2} & 0 &
\frac {q^2(\tilde{u}^1_1u^1_1+(\tilde{u}^1_1u^1_1)^{-1})}{1+q^2}
\end{array}
\right)
\end{equation}}
Summarizing, in this paper we have introduced the concept of $q$-regularization
and used the projective representation of the non-commutative Heisenberg
algebra to construct the Manin quantum plane, thereby defining $q$-spinors.
Using this as a building block we present a $q$-regularization in terms of a
four dimensional representation of a particular two dimensional non-commutative
space
at first. Besides, regularization on a $q$-deformed space that can be maped
into a particular braided product of Manin's quantum plane and it is related
to the first $q$-space is studied too. We show how to extract, from relevant
quantities, finite components (provided $q\neq 1$) that can become infinite
at $q=1$. To compute the Haar weight, we propose particular basis projected
from $q$-deformed spaces, so the functions to be $q$-regularized are to be
considered on this frame of reference. An example for $\lambda\phi^4$ field
theory is presented. Additional work must be done to generalize our scheme
to any arbitrary function on $q$-Minkowski space-time basis.\
Finally, in order to learn about the general scheme and its symmetries, we
study the T=0 Haar measure in terms of the $SU_q(2)$ measure and the null
directions in the Hopf algebra that lead to a quantum mechanical Galilei
group.\
Although in this paper we can $q$-regularize only a class of suitable
functions (restricted by the particular basis chosen), we think that the full
prescription, derived from physical considerations, might apply to make
relevant quantities in field theory finite at $q\neq 1$.
\section{Acknowledgments}
This work was partially supported by CONACYT, M\'exico. I wish to thank ITD
in UC Davis for its hospitality
|
\section{Introduction}
An affine connection is one of the basic objects of interest in
differential geometry. It provides a simple and invariant way of
transferring information from one point of a connected manifold $M$ to
another and, not surprisingly, enjoys lots of applications in many
branches of mathematics, physics and mechanics. Among the most
informative characteristics of an affine connection is its (restricted)
holonomy group which is defined, up to conjugacy, as the subgroup of
$Gl(T_tM)$ consisting of all automorphisms of the tangent space $T_tM$
at $t \in M$ induced by parallel translations along $t$-based loops in
$M$.
Which reductive Lie groups $G$ can be irreducibly acting holonomies of
affine connections?
By a result of Hano and Ozeki \cite{HO}, {\em any} (closed) Lie group
representation $G \subseteq Gl(V)$ can be realized in this way. The
same question, if posed in the subclass of {\em torsion-free} affine
connections, has a very different answer. Long ago, Berger \cite{Ber}
presented a very restricted list of possible holonomies of a
torsion-free affine connection which, as he claimed, is complete up to
a finite number of missing terms. His list is separated into two
parts. The first part corresponds to the holonomies of {\em metric}
connections, the second part to the {\em non-metric} ones. While Berger
gave detailed arguments for the proof of the metric part, the proof of
the second part was omitted.
The list of metric connections has been studied extensively in the
intervening years. In fact, it is by now well-known which entries of
this list actually {\em do} occur as holonomies of torsion-free
connections, and how the holonomy relates to both the geometry and the
topology of the underlying manifold. See \cite{Al}, \cite{Br1},
\cite{Br2}, \cite{J}, \cite{Si}, as well as the surveys in \cite{Sa}
and \cite{Bes}.
Despite the lack of proof, the second part of Berger's holonomy list
seems to have been generally accepted as correct. Even when Bryant
found examples of holonomies which are not on Berger's list \cite{Br3},
he called them {\em exotic holonomies}, suggesting that such holonomies
should exist in very special dimensions only, and therefore should be
thought of as analogous to the {\em exceptional holonomies} in the
metric case. Further examples of exotic holonomies were found in
\cite{CS}.
However, it is the subject of the present article to prove that, even
up to finitely many missing terms, Berger's list is still incomplete.
This is done by proving the existence of an infinite family of
irreducible representations which are not on this list, yet do occur as
holonomy of torsion-free connections. These representations are:
\begin{equation} \label{replist} \begin{array}{llll}
Sl(2,{\Bbb C}) SO(n,{\Bbb C}), & {\mbox{acting on}} & {{\Bbb R}^{8n}} \cong {\Bbb C}^2
\otimes {\Bbb C}^n, & \mbox{where $n \geq 3$},\\
Sl(2,{\Bbb R}) SO(p,q), & {\mbox{acting on}} & {{\Bbb R}^{2(p+q)}} \cong {\Bbb R}^2
\otimes {\Bbb R}^{p+q}, & \mbox{where $p + q \geq 3$},\\
Sl(2,{\Bbb R}) SO(2,{\Bbb R}), & {\mbox{acting on}} & {{\Bbb R}^4} \cong {\Bbb R}^2
\otimes {\Bbb R}^2.\\
\end{array} \end{equation}
\begin{thm} \label{thm:ConnectionsExist} All representations in {\em
(\ref{replist})} occur as holonomies of torsion-free connections which
are not locally symmetric.
\end{thm}
These candidates were first discovered by twistor theoretical methods
which shall be described in section \ref{sec:twistor}. In fact, the
formal existence of holomorphic torsion-free connections whose holonomy
is listed in the first family of (\ref{replist}) can be shown, in
principle, by deformation theory only.
We shall, however, present a different, technically simple method to
prove their existence in section \ref{sec:construct}. This new approach
will allow us to treat all representations in (\ref{replist})
simultaneously, and to assert some global geometric properties and
rigidity of connections with these holonomies which seem hard to
achieve otherwise. In fact, we shall classify {\em all} connections
with the above holonomies by showing that they all arise from this
construction. This implies, for example, the following:
\begin{thm} \label{thm:summary} Let $G \subseteq Gl(V)$ be one of the
representations in {\em (\ref{replist})} other than \linebreak $Sl(2,{\Bbb R})
SO(2,{\Bbb R})$. Then the following hold.
\begin{itemize} \item[{\em (1)}] Any torsion-free connection whose holonomy is
contained in $G$ is analytic.
\item[{\em (2)}] The moduli space of torsion-free connections whose
holonomy is contained in $G$ is finite dimensional. In particular, the
2nd derivative of the curvature at a single point completely determines
the connection.
\item[{\em (3)}] If $n \equiv 0,1 \bmod 4$ then any torsion-free
connection whose holonomy is contained in $G$ admits a non-trivial
infinitesimal symmetry, i.e. a vector field whose flow preserves the
connection.
\item[{\em (4)}] If $G \neq Sl(2,{\Bbb R}) SO(n,{\Bbb R})$, then a torsion-free
connection whose holonomy is contained in $G$ is geodesically
incomplete, unless the connection is locally symmetric. If the latter
is the case, then the holonomy is a proper subgroup of $G$.
\end{itemize}
\end{thm}
The final step towards the proof Theorem \ref{thm:ConnectionsExist} was
accomplished when the second and third author met at the conference
{\em Geometry and Physics} in Aarhus, Denmark. We are grateful to the
organizers of this conference for inviting us; the third author wishes
to thank Professor Friedrich Hirzebruch for providing financial support
for his participation in that conference.
\section{Connections and Legendre moduli spaces} \label{sec:twistor}
Let $G \subseteq Gl(V)$ be an effective irreducible representation of a
complex connected reductive Lie group $G$ on the finite dimensional
complex vector space $V$. Without further comment, we shall always
regard $G$ as a Lie group with a fixed representation on $V$. Clearly,
$G$ also acts on $V^*$ via the dual representation. Let $\tilde{X}$ be
the $G$-orbit of a highest weight vector in $V^*$. Then the quotient
$X:= \tilde{X}/{{\Bbb C}}^*$ is a generalized flag variety which is
canonically embedded into ${\Bbb P}(V^*)$. In fact, $X=G_s/ P $, where
$G_s$ is the semi-simple part of $G$, and where $P$ is the parabolic
subgroup of $G_s$ leaving the highest weight vector invariant up to a
scalar. Let $L_X$ be the restriction of the hyperplane bundle ${\cal
O}(1)$ on ${\Bbb P}(V^*)$ to the submanifold $X\hookrightarrow {\Bbb
P}(V^*)$. Then $V \cong H^0(X,L_X)$, and the Lie algebra ${\frak g}$ of $G$ is
contained in the Lie algebra
\[
H^0\left(X,L_X\otimes \left(J^1L_X\right)^*\right) \cong H^0(X,TX) \oplus {{\Bbb C}}.
\]
Therefore, to the representation $G \subseteq Gl(V)$, we canonically
associate a pair $(X,L_X)$, consisting of a generalized flag variety
$X$ and an ample line bundle $L_X$ on $X$. Most of the relevant
information about $G$ can be restored from $(X,L_X)$. The very few
cases when important information gets lost by the transition from
$(X,L_X)$ back to $G \subseteq Gl(V)$ are listed in \cite{Stei}.
A crucial step in deciding whether $G$ can occur as the holonomy of a
torsion-free affine connection is the computation of the formal
curvature space $K({\frak g})$, which is defined as the kernel of the
composition
\[
\Lambda^2 V^*\otimes {{\frak g}} \rightarrow \Lambda^2 V^* \otimes V^*\otimes V \rightarrow
\Lambda^3 V^* \otimes V.
\]
Namely, it follows from the {\em Ambrose-Singer Holonomy Theorem}
\cite{AS} that if there is a {\em proper} subalgebra ${\frak g}' \subset {\frak g}$
such that $K({\frak g}) \subseteq \Lambda^2 V^* \otimes {\frak g}' \subset \Lambda^2
V^* \otimes {\frak g}$, then $G$ {\em cannot} be holonomy of a torsion-free
connection. This condition is called {\em Berger's first criterion},
and was used for the original classification in \cite{Ber}.
It is desirable to interpret $K({\frak g})$ in terms of the associated pair
$(X,L_X)$, because one will then be able to make use of the powerful
tools of complex analysis (such as, say, the vanishing theorems) to
attack the Berger classification problem. The twistor theory developped
in \cite{Me} does indeed provide us with such an interpretation. We
shall briefly describe this twistor construction, and use it to
calculate $K({\frak g})$ for the representations in (\ref{replist})
explicitly. For details and more general statements, we refer to
\cite{Me}.
Let $Y$ be a complex $(2n + 1)$-dimensional manifold. A {\em complex
contact structure}\, on $Y$ is a rank $2n$ holomorphic subbundle
$D\subseteq TY$ of the holomorphic tangent bundle to $Y$ such that the
Frobenius form
\begin{eqnarray*}
\Phi: D \times D & \longrightarrow & TY/D\\
(v,w) & \longmapsto & [v,w]\bmod D
\end{eqnarray*}
is non-degenerate. Define the contact line bundle $L$ by the exact
sequence
\[
0 \longrightarrow D \longrightarrow TY
\stackrel{\theta}{\longrightarrow} L \longrightarrow 0.
\]
One can easily verify that the maximal non-degeneracy of the
distribution $D$ is equivalent to the fact that the above defined
``twisted'' 1-form $\theta \in H^0(Y, L\otimes \Omega^1 M)$ satisfies
the condition
\[
\theta\wedge (d\theta)^n \neq 0.
\]
A complex $n$-dimensional submanifold $X$ of the complex contact
manifold $Y$ is called a {\em Legendre submanifold}\, if $TX \subseteq
D$.
Suppose $X \hookrightarrow Y$ is a compact complex Legendre submanifold
of a complex contact manifold $(Y, L)$, and denote the restriction
$L|_X$ by $L_X$. If $H^1(X, L_X) = 0$, then there exists a complete
family $\{X_t \hookrightarrow Y\ |\ t \in M \}$ of compact complex
Legendre submanifolds which is obtained from $X$ by all possible
holomorphic Legendre deformations of $X$ inside $Y$. The parameter
space $M$ is an $h^0(X, L_X)$-dimensional complex manifold, called the
{\em Legendre moduli space}. Moreover, if two more cohomology groups
vanish, namely if
\begin{equation} \label{eq:cohomsvanish}
H^0\left(X, L_X \otimes S^2\left(J^1 L_X\right)^*\right) = H^1\left(X, L_X \otimes
S^2\left(J^1 L_X\right)^*\right) = 0,
\end{equation}
then the Legendre moduli space $M$ comes equipped not only with an
induced complex manifold structure, but also with a uniquely induced
holomorphic torsion-free affine connection whose curvature tensor is a
field on $M$ with values in the cohomology group $H^1\left(X, L_X \otimes
S^3\left(J^1 L_X\right)^*\right)$. In fact, {\em any} holomorphic torsion-free
affine connection with reductive irreducibly acting holonomy group is
an induced connection on an appropriate Legendre moduli space
\cite{Me}.
This suggests the following strategy to look for new exotic
holonomies: find a pair $(X, L_X)$ consisting of a generalized flag
variety $X = G/P$ and an ample line bundle $L_X \rightarrow X$ such that the
cohomology groups in (\ref{eq:cohomsvanish}) vanish, while $H^1\left(X,
L_X \otimes S^3\left(J^1 L_X\right)^*\right)$ is non-zero. Then the twistor theory
guarantees that there is a natural injection
\begin{equation} \label{inj}
\imath : H^1\left(X, L_X \otimes S^3\left(J^1 L_X\right)^*\right) \longrightarrow
\Lambda^2 V^* \otimes {\frak g},
\end{equation}
where ${{\frak g}}:= H^0\left(X,L_X\otimes \left(J^1 L_X\right)^*\right)$ and $V:=
H^0(X,L_X)$, whose image equals $K({\frak g})$. In particular, $K({\frak g}) \cong
H^1\left(X, L_X \otimes S^3\left(J^1 L_X\right)^*\right)$ as a ${\frak g}$-vector space. We
let $K_0({\frak g}) \subseteq K({\frak g})$ be the set of elements with {\em full
curvature}, i.e.
\begin{equation} \label{eq:fullcurvature}
K_0({\frak g}) := \left\{ \rho \in K({\frak g})\ \left|\ \left< \left\{ \rho(x,y)\ |\ x,y \in V
\right\} \right> = {\frak g} \right. \right\}.
\end{equation}
One can show that either $K_0({\frak g}) = \emptyset$ or $K_0({\frak g})$ is {\em
dense in} $K({\frak g})$.
As a particular example, let us consider the reducible homogeneous
manifold $X =$ \linebreak ${\Bbb C} {\Bbb P}_1 \times {\Bbb Q}_m$, where ${\Bbb Q}_m
\hookrightarrow {\Bbb C} {\Bbb P}_{m+1}$ is the non-degenerate quadric. We define the
line bundle
\[
L_X := \pi_1^*({\cal O}_{{\Bbb C} {\Bbb P}_1}(1)) \otimes \pi_2^*( {\cal O}_{
{\Bbb C} {\Bbb P}_{m+1}}(1)|_{{\Bbb Q}_m}),
\]
where $\pi_1: X \rightarrow {\Bbb C} {\Bbb P}_1$ and $\pi_2: X \rightarrow {\Bbb Q}_m$ are the canonical
projections. Then
\[
V:= H^0(X,L_X) = W_2\otimes W_n,
\]
where $W_2 := H^0({\Bbb C} {\Bbb P}_1, {\cal O}_{{\Bbb C} {\Bbb P}_1}(1))$ and $W_n := H^0({\Bbb Q}_m,
{\cal O}_{{\Bbb C} {\Bbb P}_{m+1}}(1)|_{{\Bbb Q}_m})$ are vector spaces of dimensions $2$ and $n
:= m+2$, respectively, and
\[
{\frak g}:= H^0\left(X,L_X\otimes \left(J^1 L_X\right)^*\right) = {{\frak g}}_0 \oplus {{\Bbb C}},
\]
where ${{\frak g}}_0 = {\frak sl}(W_2) \oplus {\frak so}(W_n)$. Note that $W_2$ and $W_n$
carry a ${\frak g}_0$-invariant area form $<\!\ ,\ \!>$ and a
${\frak g}_0$-invariant inner product $(\ ,\ )$, respectively. It is easy to
check that for $m \geq 2$, i.e. $n \geq 4$, the cohomology groups in
(\ref{eq:cohomsvanish}) vanish, while
\begin{equation} \label{eq:Kg-is-H1}
K({\frak g}) \cong H^1\left(X, L_X \otimes S^3\left(J^1L_X\right)^*\right) \cong S^2(W_2)
\oplus \Lambda^2 W_n.
\end{equation}
A calculation shows that $\Lambda^2 V^* \otimes {\frak g}$ contains only two
summands isomorphic to $S^2(W_2)$ and three isomorphic to $\Lambda^2
W_n$. After that, it is easy to obtain explicit expressions for the
elements of $K({\frak g})$. As it turns out, all elements of $K({\frak g})$ are
contained in $\Lambda^2 \otimes {\frak g}_0 \subset \Lambda^2 \otimes {\frak g}$. It
follows that {\em every torsion-free connection whose holonomy is
contained in $Gl(2,{\Bbb C}) SO(n,{\Bbb C})$ for $n \geq 3$ must be contained in
$Sl(2,{\Bbb C}) SO(n,{\Bbb C})$}.
In order to present the description of $K({\frak g})$, we use the
identifications ${\frak sl}(W_2) \cong S^2(W_2)$ and ${\frak so}(W_n) \cong \Lambda^2
W_n$, given by the identities
\begin{equation} \label{eq:LieIsos} \begin{array}{cccccc} (e_1 e_2)
\cdot e_3 & := & <\! e_1, e_3 \!> e_2 & + & <\! e_2, e_3 \!> e_1, &
\mbox{and}\\
(x_1 \wedge x_2) \cdot x_3 & := & (x_1, x_3) x_2 & - & (x_2, x_3) x_1
\end{array} \end{equation}
for all $e_i \in W_2$, $x_i \in W_n$.
\begin{prop} \label{prop:compute-Kg} Let $G \subseteq Gl(V)$ be one of
the representations in (\ref{replist}) other than \linebreak $Sl(2,{\Bbb R})
SO(2,{\Bbb R})$, and let ${\frak g} \subseteq {\frak gl}(V)$ be its Lie algebra. For $A \in
{\frak sl}(2,{\Bbb C})$ and $M \in {\frak so}(n,{\Bbb C})$ ($A \in {\frak sl}(2,{\Bbb R})$ and $M \in
{\frak so}(p,q)$, respectively), define $\rho_{A+M}: \Lambda^2 V \rightarrow {\frak g}$ by
\begin{equation} \label{eq:curvature} \begin{array}{ll}
\rho_{A + M} (e_1 \otimes x_1, e_2 \otimes x_2) := & <\! e_1, e_2 \!>
((x_1,x_2) (A + M)\ +\ (x_1 \wedge M x_2 + x_2 \wedge M x_1))\\
& +\ (x_1, M x_2) e_1 e_2\ -\ <\! Ae_1, e_2 \!> x_1 \wedge x_2.
\end{array} \end{equation}
Then $\rho_{A+M} \in K({\frak g})$, and the map
\[
\begin{array}{lcll} \rho: & {\frak g} & \longrightarrow & K({\frak g})\\
& A + M & \longmapsto & \rho_{A+M} \end{array}
\]
is a $G$-equivariant isomorphism.
\end{prop}
\begin{proof} It is straightforward to verify that $\rho_{A+M} \in
K({\frak g})$ and that $\rho$ is a $G$-equivariant injection. If $G = Sl(2,{\Bbb C})
SO(n,{\Bbb C})$ with $n \geq 4$, then the surjectivity of $\rho$ follows from
(\ref{eq:Kg-is-H1}). This implies the surjectivity of $\rho$ for $G =
Sl(2,{\Bbb R}) SO(p,q)$ with $p+q \geq 4$ as well.
If $n=3$ or $p+q=3$, then the surjectivity follows from a direct
calculation.
\end{proof}
\begin{rem} \label{remark} {\em \begin{itemize}
\item[(1)] If $G = Sl(2,{\Bbb R}) SO(2,{\Bbb R})$ then $\rho$ is injective but not
surjective. In fact, $dim(K({\frak g})) = 9 > dim({\frak g})$ in this case.
\item[(2)] The case $G = Sl(2,{\Bbb F}) SO(3,{\Bbb F})$ with ${\Bbb F} = {\Bbb R}$ or ${\Bbb C}$ was
treated in \cite{CS}, where this representation was called $H_{12}$.
Note that the twistor approach from above does not work for $n=3$;
indeed, the cohomology $H^1\left(X, L_X \otimes S^2\left(J^1 L_X\right)^*\right) \neq
0$, thus (\ref{eq:cohomsvanish}) is {\em not} satisfied. Nevertheless,
it follows from the results in \cite{CS} that every Legendre moduli
space with \linebreak $X \cong {\Bbb C} {\Bbb P}_1 \times {\Bbb Q}_1 \cong {\Bbb C} {\Bbb P}_1 \times {\Bbb C} {\Bbb P}_1$
and $L_X$ as above, i.e. $L_X \cong {\cal O}(1,2)$, {\em does} carry an
induced torsion-free connection. This shows, in particular, that
(\ref{eq:cohomsvanish}) is not a necessary condition for the existence
of torsion-free induced connections on a Legendre moduli space (cf.
\cite{Me}).
\item[(3)] If $A,M \neq 0$ then $\rho_{A+M}$ is surjective. Thus, for
the representations in (\ref{replist}), $K_0({\frak g}) \subseteq K({\frak g})$ is
open dense.
\end{itemize}} \end{rem}
By (3), it follows that if there exists a holomorphic torsion-free
affine connection $\nabla$ with ``generic'' curvature tensor, then its
holonomy will be the full group $Sl(2,{\Bbb C}) SO(n,{\Bbb C})$.
If $(Y,L)$ is a complex contact manifold containing $X = {\Bbb C} {\Bbb P}_1 \times
{\Bbb Q}_m$ as a Legendre submanifold, then the latter is stable under
arbitrary holomorphic deformations of the contact structure on $Y$.
Therefore, one possible way to prove the existence of the required
holomorphic affine connections is to use the Kodaira deformation theory
\cite{Ko} to construct a sufficiently ``generic'' contact manifold
$Y$.
We shall, however, use a different method for the existence proof,
which will be presented in the following section.
\section{Construction of torsion-free connections}
\label{sec:construct}
Let us briefly recall the definition and basic properties of a Poisson
manifold. For a more detailed exposition, see e.g. \cite{L} or
\cite{V}.
\begin{df} \label{def:Poisson} A Poisson structure on a differentiable
manifold $P$ is a bilinear map, called the Poisson bracket
\[
\{\ ,\ \}: \otimes^2 C^\infty(P, {\Bbb R}) \longrightarrow C^\infty(P, {\Bbb R}),
\]
satisfying the following identities:
\begin{itemize}
\item[{\em (i)}] the bracket is skew-symmetric:
\[ \{f,g\} = -\{g,f\}, \]
\item[{\em (ii)}] the bracket satisfies the Jacobi identity:
\[ \{f,\{g,h\}\} + \{g,\{h,f\}\} + \{h,\{f,g\}\} = 0, \]
\item[{\em (iii)}] the bracket is a derivation in each of its
arguments:
\[ \{f,gh\} = \{f,g\} h + g \{f,h\}. \]
\end{itemize}
\end{df}
It is well-known that on every Poisson manifold $(P, \{\ ,\ \})$, there
exists a unique smooth bivector field $\Lambda \in \Gamma(P, \Lambda^2
TP)$ such that the Poisson bracket is given by
\begin{equation}
\label{eq:bivector} \{f,g\} = \Lambda(df,dg).
\end{equation}
We define the homomorphism $\Lambda^\#: T^*P \rightarrow TP$ by the equation
\begin{equation} \label{eq:defLambda} \begin{array}{ll}
(\Lambda^\# df)(g) = \{f, g\} & \mbox{for all $f,g \in C^\infty(P,
{\Bbb R})$.}
\end{array} \end{equation}
The {\em half-rank at $p \in P$} of the Poisson structure is the
smallest integer $r$ such that
\[
\Lambda^{r+1}(p) = 0,
\]
and the {\em rank at $p \in P$} is twice the half-rank. It follows that
the rank at $P$ equals the rank of $\Lambda^\#_p: T_p^*P \rightarrow T_pP$. The
Poisson structure is called {\em non-degenerate at $p$} if
$\Lambda^\#_p$ is an isomorphism, i.e. if the rank at $p$ equals the
dimension of $P$. In particular, if $P$ is non-degenerate at a point
then $P$ must be even dimensional, and the set of non-degenerate points
is open in $P$. If $P$ is non-degenerate {\em everywhere}, then there
is a natural symplectic 2-form $\sigma$ on $P$ such that $\Lambda^\#$
is precisely the index-raising map associated to $\sigma$. In fact, it
is well known that symplectic structures are in a natural one-to-one
correspondence with non-degenerate Poisson structures.
The {\em characteristic field} of the Poisson structure is the subset
of $TP$ given by
\[
{\cal C} = \Lambda^\#(T^*P).
\]
Thus, the dimension of ${\cal C}_p$ equals the rank at $p$. A {\em
characteristic leaf} $\ \Sigma \subseteq P$ is a submanifold for which
$T_p\Sigma = {\cal C}_p$ for all $p \in \Sigma$. From
(\ref{eq:defLambda}), it follows that the set of functions which vanish
on $\Sigma$ form a {\em Poisson ideal;} hence there is a naturally
induced Poisson structure on $\Sigma$. Clearly, this Poisson structure
on $\Sigma$ is non-degenerate. Thus it follows that {\em every
characteristic leaf of a Poisson manifold carries a natural symplectic
structure}.
\begin{df} Let $(P, \{\ ,\ \})$ be a Poisson manifold. A symplectic
realization of $P$ is a symplectic manifold $(S, \sigma)$ and a
submersion
\[
\pi: S \longrightarrow P
\]
which is compatible with the Poisson structures, i.e.
\begin{equation} \label{eq:PoissonMorph} \begin{array}{ll}
\{\pi^*(f), \pi^*(g)\}_S = \pi^*(\{f,g\}) & \mbox{for all $f,g \in
C^\infty(P,{\Bbb R})$,}
\end{array} \end{equation}
where the Poisson bracket $\{\ ,\ \}_S$ on $S$ is induced by the
symplectic structure.
\end{df}
\begin{prop} \label{prop:SymplReal} If $p_0 \in P$ has locally constant
rank, i.e. the rank is constant on an open neighborhood $U$ of $p_0$,
then there is a local symplectic realization at $p_0$, i.e. a
symplectic realization $\pi: S \longrightarrow U'$ with $p_0 \in U'
\subseteq U$.
\end{prop}
\begin{proof} By Darboux's Theorem, there exists a local coordinate
system $(p_i, q_i, t_\alpha)$, \linebreak $i = 1, \ldots, r$, $\alpha = 1,
\ldots, s$, in a neighborhood $U$ of $p_0$ such that the Poisson
bracket is given by
\[
\{f,g\} = \frac {\partial f}{\partial p_i} \frac {\partial g}{\partial
q_i} - \frac {\partial f}{\partial q_i} \frac {\partial g}{\partial
p_i}.
\]
Let $S := U \times {\Bbb R}^s$ with coordinates $u_\alpha$ on ${\Bbb R}^s$. Define
the symplectic 2-form \linebreak $\sigma := dp_i \wedge dq_i + dt_\alpha \wedge
du_\alpha$ on $S$. Then it is easily verified that the natural
projection $\pi: S \rightarrow U$ is a symplectic realization of $U$.
\end{proof}
We now turn to the construction of torsion-free connections via Poisson
structures. First of all, let us set up some notation.
Let $V$ be a finite dimensional vector space, ${\frak g} \subseteq {\frak gl}(V)$ a
Lie sub-algebra, and let $G \subseteq Gl(V)$ be the corresponding
connected Lie group. $G$ acts canonically on $V$, and on ${\frak g}$ via the
adjoint representation. This induces $G$-actions on all tensor powers
and direct sums of ${\frak g}$ and $V$ which we will call the {\em canonical
action} of $G$ on these spaces.
Recall that the {\em curvature space} $K({\frak g})$ is defined by the exact
sequence
\[
0 \longrightarrow K({\frak g}) \longrightarrow \Lambda^2 V^* \otimes {\frak g}
\longrightarrow \Lambda^3 V^* \otimes V,
\]
where the latter map is the composition of the inclusion and the
skew-symmetrization map $\Lambda^2 V^* \otimes {\frak g} \rightarrow \Lambda^2 V^*
\otimes V^* \otimes V \rightarrow \Lambda^3 V^* \otimes V$. Likewise, we
define the {\em 2nd curvature space} $K^1({\frak g})$ by the exact sequence
\[
0 \longrightarrow K^1({\frak g}) \longrightarrow V^* \otimes K({\frak g})
\longrightarrow \Lambda^3 V^* \otimes {\frak g},
\]
where again, the latter map is given by the composition of an inclusion
and skew-symmetrization, namely $V^* \otimes K({\frak g}) \rightarrow V^* \otimes
\Lambda^2 V^* \otimes {\frak g} \rightarrow \Lambda^3 V^* \otimes {\frak g}$. In other
words, $K({\frak g})$ and $K^1({\frak g})$ consist of those linear maps $\Lambda^2 V
\rightarrow {\frak g}$ and $V \rightarrow K({\frak g})$ which satisfy the 1st and 2nd Bianchi
identity, respectively.
Let $W := {\frak g} \oplus V$. Denote elements of ${\frak g}$ and $V$ by $A, B,
\ldots$ and $x, y, \ldots$, respectively, and elements of $W$ by $w,
w', \ldots$. We may regard $W$ as the semi-direct product of Lie
algebras, i.e. we define a Lie algebra structure on $W$ by the equation
\[
[A+x, B+y] := [A,B] + A \cdot y - B \cdot x.
\]
It is well-known \cite{L} that this induces a natural Poisson structure
on the dual space $W^*$. Now, we wish to perturb this Poisson
structure. For this, we need the
\mbox{}
\begin{df} A $C^\infty$-map $\phi: {\frak g}^* \rightarrow \Lambda^2 V^*$ is called
admissible if
\begin{itemize} \item[{\em (i)}] $\phi$ is $G$-equivariant,
\item[{\em (ii)}] $d\phi(p) \in K({\frak g}) \subseteq \Lambda^2 V^* \otimes
{\frak g}$ for all $p \in {\frak g}^*$.
\end{itemize}
\end{df}
In order for condition (ii) to make sense, we use the natural
identification $T^*_p{\frak g}^* \cong {\frak g}$. Now, the following important
observation is easily proven.
\begin{prop} \label{prop:bracket}
Let $V$, ${\frak g} \subseteq {\frak gl}(V)$, $W$ and $K({\frak g})$ as above, and let
$\phi: {\frak g}^* \rightarrow \Lambda^2 V^*$ be an admissible map. Let $\Phi := \phi
\circ pr$, where $pr: W^* \rightarrow {\frak g}^*$ is the natural projection. Then the
following bracket on $W^*$ is Poisson:
\begin{equation} \label{eq:bracket}
\{f,g\}(p) := p([A+x, B+y]) + \Phi(p)(x,y).
\end{equation}
Here, $df_p = A + x$ and $dg_p = B + y$ are the decompositions of
$df_p, dg_p \in T^*_pW^* \cong W$.
\end{prop}
Note that for $\phi = 0$, we simply obtain the Poisson structure
induced by the Lie algebra structure on $W$.
Let us now consider a Poisson structure on $W^*$ induced by an
admissible map \linebreak $\phi: {\frak g}^* \rightarrow \Lambda^2 V^*$. Let $\pi: S \rightarrow
U$ be a symplectic realization of the open subset $U \subseteq W^*$.
For each $w \in W$, we define the vector fields
\[
\eta_w := \Lambda^\#(w) \in {\frak X}(W^*),
\]
and
\[
\xi_w := \#(\pi^*(w)) \in {\frak X}(S),
\]
where $w \in W \cong T^*W^*$ is regarded as a 1-form on $W^*$. Since
$\pi$ preserves the Poisson structure (\ref{eq:PoissonMorph}), we have
\begin{equation} \label{eq:pixi=eta} \begin{array}{ll}
\pi_*(\xi_w) = \eta_w & \mbox{for all $w \in W$}.
\end{array} \end{equation}
In contrast to the map $w \mapsto \eta_w$, the map $w \mapsto \xi_w$ is
{\em pointwise injective}. Thus, \linebreak $\xi := \{\xi_w\ |\ w \in W \}
\subseteq TS$ is a distribution on $S$ whose rank equals the dimension
of $W$. For the bracket relations, we compute
\begin{equation} \label{eq:xibracket} \begin{array}{llll}
[\xi_A, \xi_B] & = & \xi_{[A,B]}\\
{[\xi_A, \xi_x]} & = & \xi_{A \cdot x}\\
{[\xi_x, \xi_y]}(s) & = & \xi_{d\Phi(p)(x,y)} & \mbox{ where $p =
\pi(s)$}.
\end{array} \end{equation}
This implies, of course, that the distribution $\xi$ on $S$ is {\em
integrable}. Moreover, the first equation in (\ref{eq:xibracket})
implies that the flow along the vector fields $\xi_A$ induces a local
$G$-action on $S$. Let $F \subseteq S$ be a maximal integral leaf of
$\xi$. Clearly, $F$ is $G$-invariant, and we can define a $W$-valued
coframe $\theta + \omega$ on $F$, where $\theta$ and $\omega$ take values in $V$
and ${\frak g}$, respectively, by the equation
\[
v_s = \xi_{(\omega + \theta) (v_s)} (s), \mbox{ all $v_s \in T_sF$}.
\]
The equations dual to (\ref{eq:xibracket}) then read
\begin{equation} \label{eq:structureeqn} \begin{array}{lll}
d\theta & = & - \omega \wedge \theta\\
d\omega & = & - \omega \wedge \omega - \pi^*(d\Phi) \circ (\theta \wedge \theta).
\end{array} \end{equation}
Here, $d\Phi$ is regarded as a map with values in $K({\frak g}) \subseteq
\Lambda^2 V^* \otimes {\frak g}$.
After shrinking $S$ and $U$ if necessary, we may assume that $M := F/G$
is a {\em manifold}. From (\ref{eq:structureeqn}) it follows that there
is a unique torsion-free connection on $M$ and a unique immersion
$\imath: F \hookrightarrow {\frak F}_V$ into the $V$-valued coframe bundle
${\frak F}_V$ of $M$ such that $\theta = \imath^*(\underline\theta)$ and $\omega =
\imath^*(\underline\omega)$, where $\underline\theta$ and $\underline\omega$ are the tautological
and the connection 1-form on ${\frak F}_V$, respectively. Clearly, the holonomy
of this connection is contained in $G$.
\begin{df} Let $\phi: {\frak g}^* \rightarrow \Lambda^2 V^*$ be an admissible map.
Then a torsion-free connection which is obtained from the above
construction is called a Poisson connection induced by $\phi$.
\end{df}
We then get the following result.
\begin{thm} \label{th:ConnExist} Let $V$, ${\frak g} \subseteq {\frak gl}(V)$ and
$K({\frak g})$ be as before, and let $K_0({\frak g}) \subseteq K({\frak g})$ be as in {\em
(\ref{eq:fullcurvature})}. Consider an admissible map $\phi: {\frak g}^* \rightarrow
\Lambda^2 V^*$. Furthermore, suppose that
\[
{\frak g}^* \supseteq U_0 := (d\phi)^{-1} (K_0({\frak g})) \neq \emptyset.
\]
Then there exist Poisson connections induced by $\phi$ whose holonomy
representations are equivalent to ${\frak g}$. Moreover, if $\phi|_{U_0}$ is
not affine, then not all of these connections are locally symmetric.
\end{thm}
\begin{proof} Let $U^{reg} \subseteq U_0 \oplus V^* \subseteq W^*$ be
the subset of points for which the rank is locally constant. By upper
semi-continuity of the rank, $U^{reg}$ is open dense in $U_0 \oplus
V^*$.
Now Proposition \ref{prop:SymplReal} implies that there are symplectic
realizations $\pi: S \to U$, with open $U \subseteq U^{reg} \subseteq
W^*$. Then the above construction produces Poisson connections induced
by $\phi$ on some manifold $M = F/G$. By (\ref{eq:fullcurvature}),
(\ref{eq:structureeqn}) and the {\em Ambrose-Singer Holonomy Theorem}
\cite{AS}, the holonomy of this connection equals ${\frak g}$.
To show the last part, let us assume that {\em all} connections which
arise in this way are locally symmetric. Let $w := (p, q) \in U^{reg}$
with $p \in U_0$, $q \in V^*$. Then we may choose the symplectic
realization $\pi: S \rightarrow U$ and $F \subseteq U$ such that $w \in
\pi(F)$. It is then easy to show by (\ref{eq:structureeqn}) that the
corresponding connection on $M := F/G$ is locally symmetric iff
${\frak L}_{\xi_x}(\pi^*(d\Phi)) = 0$ for all $x \in V$. By
(\ref{eq:pixi=eta}) and because $\pi$ is a submersion, this is
equivalent to ${\frak L}_{\eta_x}(d\Phi) = 0 \mbox{ for all $x \in V$},$ or
${\frak L}_{pr_*(\eta_x)}(d\phi) = 0 \mbox{ for all $x \in V$.}$ But now an
easy calculation shows that for all $A \in {\frak g}$,
\[
(pr_*(\eta_x)_w)(A) = -q(A \cdot x) = -\jmath(q \otimes x)(A),
\]
where $\jmath: V^* \otimes V \rightarrow {\frak g}^*$ is the natural projection. Thus,
by our assumption, it follows that
\[
{\frak L}_{\jmath(q \otimes x)} (d\phi)_p = 0 \mbox{ for all $x \in V$, $(p,q)
\in U^{reg}$.}
\]
By density of $U^{reg}$, this equation holds for {\em all} $p \in U_0$,
$q \in V^*$, and since $\jmath$ is surjective, it follows that
\[
{\frak L}_\alpha(d\phi)_p = 0 \mbox{ for all $\alpha \in {\frak g}^*$, $p \in U_0$},
\]
i.e. $d\phi|_{U_0}$ is constant, hence $\phi|_{U_0}$ is affine.
\end{proof}
By Theorem \ref{th:ConnExist} it will suffice to address the question
of {\em existence} of admissible maps $\phi$ in order to construct
connections with prescribed holonomy.
Define the $k$-th jet space of ${\frak g}$ by
\[
J_k({\frak g}) := \left(S^k({\frak g}) \otimes \Lambda^2 V^*\right) \cap \left( S^{k-1}({\frak g})
\otimes K({\frak g}) \right),
\]
where both are regarded as subspaces of $S^{k-1}({\frak g}) \otimes {\frak g} \otimes
\Lambda^2 V^*$. Suppose there is a $G$-invariant element $\phi_k \in
J_k({\frak g})$. If we regard $\phi_k$ as a polynomial map of degree $k$,
$\phi_k: g^* \rightarrow \Lambda^2 V^*$, then it follows that $\phi_k$ is
admissible. Conversely, given an {\em analytic} map $\phi: {\frak g}^* \rightarrow
\Lambda^2 V^*$ with analytic expansion at $0 \in {\frak g}^*$
\[
\phi = \phi_0 + \phi_1 + \cdots,
\]
then it is straightforward to show that $\phi$ is admissible iff all
$\phi_k$ are, iff $\phi_k \in J_k({\frak g})^G$.
Consider an element $\phi_2 \in J_2({\frak g})^G$. On the one hand, we may
regard $\phi_2$ as an element of ${\frak g} \otimes K({\frak g})$, on the other hand,
it is easy to verify that also $\phi_2 \in V \otimes K^1({\frak g}) \subseteq
V \otimes V^* \otimes K({\frak g})$. Thus, by the natural contractions,
$\phi_2$ induces $G$-equivariant linear maps
\begin{equation} \label{eq:inducedmaps} \begin{array}{llll}
\phi_2': & {\frak g}^* & \longrightarrow & K({\frak g})\\
\phi_2'': & V^* & \longrightarrow & K^1({\frak g}).
\end{array} \end{equation}
We shall now demonstrate the existence of torsion-free connections with
prescribed holonomy.
\begin{thm} \label{thm:PoissonExist} Let $G \subseteq Gl(V)$ be one of
the following representations:
\begin{itemize} \item[{\em (i)}] For ${\Bbb F} = {\Bbb R}$ or ${\Bbb C}$, $G = Sl(2,{\Bbb F})$ acting on
the space $V_3 \subseteq {\Bbb F}[x,y]$ of homogeneous polynomials in $x$ and
$y$ of degree $3$,
\item[{\em (ii)}] any of the representations in {\em (\ref{replist})}.
\end{itemize}
Then there is a $G$-invariant 2-form $\sigma \in \Lambda^2 V^*$ which
is unique up to a scalar. Also, $J_2({\frak g})$ is one-dimensional and acted
on trivially by $G$. The generic Poisson connection induced by the
admissible map \begin{equation} \label{eq:admissible} \phi = \phi_2 + c \sigma, \end{equation}
with $0 \neq \phi_2 \in J_2({\frak g})$ and some constant $c$, has full
holonomy $G$ and is not locally symmetric.
\end{thm}
This shows, in particular, that the representations in (\ref{replist})
do occur as holonomy representations, and hence proves Theorem
\ref{thm:ConnectionsExist}. Also, the representation in (i) is
precisely the representation $H_3$ which has already been shown to
occur as an exotic holonomy in \cite{Br3}.
\begin{proof} We shall present the proof for the representations in
(ii) only. In this case, the $G$-invariant symplectic form is
\[
\sigma(e_1 \otimes x_1, e_2 \otimes x_2) = <\! e_1, e_2 \!>
(x_1,x_2),
\]
where, as before, $<\! \ ,\ \!>$ and $(\ ,\ )$ denote the area form
on ${\Bbb F}^2$ and the inner product on ${\Bbb F}^n$, respectively, with ${\Bbb F} = {\Bbb R}$
or ${\Bbb C}$. We shall identify ${\frak g}$ and ${\frak g}^*$ via the Killing form $B$ on
${\frak g}$, given by
\[
B(A + M, e_1 e_2 + x_1 \wedge x_2) = <\! Ae_1, e_2 \!> + (M x_1, x_2)
\mbox{ with $A \in {\frak sl}({\Bbb F}^2)$ and $M \in {\frak so}({\Bbb F}^n)$.}
\]
Here, once again we used the identifications ${\frak sl}({\Bbb F}^n) \cong
S^2({\Bbb F}^n)$ and ${\frak so}({\Bbb F}^n) \cong \Lambda^2 {\Bbb F}^n$ from
(\ref{eq:LieIsos}). Now from the explicit description of $K({\frak g})$ in
Proposition \ref{prop:compute-Kg} it is straightforward to show that
$J_2({\frak g}) = \left({\frak g} \otimes K({\frak g})\right) \cap \left(S^2({\frak g}) \otimes \Lambda^2
V\right)$ is one-dimensional and spanned by the element $\phi_2$, given by
\begin{equation} \label{eq:phi2} \begin{array}{ll}
\phi_2(A_1 + M_1, A_2 + M_2, & e_1 \otimes x_1, e_2 \otimes x_2) := \\
& \sigma(e_1 \otimes x_1, e_2 \otimes x_2) (B(A_1 + M_1, A_2 + M_2)) \\
& - (B(A_1, e_1 e_2) B(M_2, x_1 \wedge x_2) + B(A_2, e_1 e_2) B(M_1, x_1 \wedge
x_2))\\
& + <\! e_1, e_2 \!> ((M_1 x_1, M_2 x_2) + (M_1 x_2, M_2 x_1)).
\end{array} \end{equation}
The isomorphism $\rho$ from (\ref{eq:curvature}) then coincides with
the map $\phi_2': {\frak g}^* \rightarrow K({\frak g})$ from (\ref{eq:inducedmaps}), again
after identifying ${\frak g}$ and ${\frak g}^*$ via the Killing form $B$. Moreover,
by (3) in Remark \ref{remark}, $K_0({\frak g}) \subseteq K({\frak g})$ is {\em open
dense}. Then Theorem \ref{th:ConnExist} completes the proof.
\end{proof}
We can show even more. Namely, surprisingly enough, the converse of
Theorem \ref{thm:PoissonExist} is true:
\begin{thm} \label{th:allarePoisson} Let $G \subseteq Gl(V)$ be one of
the representations in Theorem \ref{thm:PoissonExist} other than
$Sl(2,{\Bbb R}) SO(2,{\Bbb R})$. Then all torsion-free affine connections whose
holonomy is contained in $G$ are Poisson connections induced by the
admissible maps {\em (\ref{eq:admissible})}.
\end{thm}
This will follow immediately from the next Theorem, together with the
explicit description of the spaces $J_2({\frak g})$ in each case.
\begin{thm} \label{thm:allareinduced} Let $G \subseteq Gl(V)$ and ${\frak g}
\subseteq {\frak gl}(V)$ be as before. Suppose the following conditions are
satisfied:
\begin{itemize} \item[{\em (i)}] $G$ is connected, reductive and acts irreducibly
on $V$,
\item[{\em (ii)}] there is a $\phi_2 \in J_2({\frak g})^G$ such that the
corresponding $G$-equivariant maps $\phi_2'$ and $\phi_2''$ from {\em
(\ref{eq:inducedmaps})} are isomorphisms.
\end{itemize}
Then every torsion-free affine connection whose holonomy is contained
in $G$ is a Poisson connection induced by a polynomial map
\[
\phi = \phi_2 + \tau,
\]
with $\phi_2 \in J_2({\frak g})$ from above and a (possibly vanishing)
$G$-invariant 2-form $\tau \in \Lambda^2 V^*$.
\end{thm}
For the proof, we shall need the following Lemma whose proof will be
given in the appendix.
\begin{lem} \label{lem:reps} Let $G \subseteq Gl(V)$ be an irreducible
representation of a connected, reductive Lie group $G$, and let ${\frak g}
\subseteq {\frak gl}(V)$ be the corresponding Lie algebra. If $\tau \in V^*
\otimes V^*$ satisfies the condition
\[
\tau(x, A \cdot y) = \tau(y, A \cdot x) \mbox{\ \ for all $x,y \in V$
and $A \in {\frak g}$,}
\]
then $\tau$ is skew-symmetric and hence a $G$-invariant 2-form.
\end{lem}
\medskip \noindent {\bf Proof of Theorem.} Let $F \subseteq {\frak F}_V$ be a
$G$-structure on the manifold $M$ where ${\frak F}_V \rightarrow M$ is the $V$-valued
coframe bundle of $M$, and denote the tautological $V$-valued 1-form on
$F$ by $\theta$. Suppose that $F$ is equipped with a torsion-free
connection, i.e. a ${\frak g}$-valued 1-form $\omega$ on $F$. Since $\phi_2'$ is
an isomorphism, the {\em first and second structure equations} read
\begin{equation} \label{eq:struct} \begin{array}{ll}
d\theta & = - \omega \wedge \theta\\
d\omega & = - \omega \wedge \omega - 2 (\phi_2'({\bf a})) \circ (\theta \wedge \theta),
\end{array} \end{equation}
where ${\bf a}: F \rightarrow {\frak g}^*$ is a $G$-equivariant map. Differentiating
(\ref{eq:struct}) and using that $\phi_2''$ is an isomorphism yields
the {\em third structure equation} for the differential of ${\bf a}$:
\begin{equation} \label{eq:struct3}
d{\bf a} = -\omega \cdot {\bf a} + \jmath({\bf b} \otimes \theta),
\end{equation}
for some $G$-equivariant map ${\bf b}: F \rightarrow V^*$, where $\jmath: V^*
\otimes V \rightarrow {\frak g}^*$ is the natural projection. The multiplication in
the first term refers to the coadjoint action of ${\frak g}$ on ${\frak g}^*$. In
other words, (\ref{eq:struct3}) should be read as
\[ \begin{array}{lll}
(\xi_A {\bf a})(B) & = & {\bf a}([A,B])\\
(\xi_x {\bf a})(B) & = & {\bf b} (B \cdot x).
\end{array} \]
Let us define the map ${\bf c}: F \rightarrow V^* \otimes V^*$ by
\begin{equation} \label{eq:struct4}
{\bf c}_p (x,y) := d{\bf b}(\xi_x) (y) - \phi_2({\bf a}_p, {\bf a}_p, x, y).
\end{equation}
Differentiation of (\ref{eq:struct3}) yields
\begin{equation} \label{eq:cinvariance} \begin{array}{ll}
{\bf c}_p(x, Ay) = {\bf c}_p(y, Ax) & \mbox{for all $x,y \in V$ and all $A \in
{\frak g}$.}
\end{array} \end{equation}
Then by (i), (\ref{eq:cinvariance}) and Lemma \ref{lem:reps} we
conclude that ${\bf c}_p \in \Lambda^2 V^*$ is $G$-invariant. Moreover,
differentiation of (\ref{eq:struct4}) implies that $\xi_A({\bf c}) = 0$ and
$(\xi_x {\bf c})(y,z) = (\xi_y {\bf c})(x,z)$ for all $A \in {\frak g}$ and $x,y,z \in
V$. Since ${\bf c}$ is skew-symmetric, it follows that
\[
d{\bf c} = 0,
\]
i.e. ${\bf c}_p \equiv \tau \in \Lambda^2 V^*$ is {\em constant}. Thus, the
$G$-equivariance of ${\bf b}$ and (\ref{eq:struct4}) yield
\begin{equation} \label{eq:newstruct4}
d{\bf b} = -\omega \cdot {\bf b} + \left( {\bf a}_p^2 \mbox{}\begin{picture}(10,10)\put(1,0){\line(1,0){7} \phi_2 + \tau \right) \circ \theta,
\end{equation}
where $\mbox{}\begin{picture}(10,10)\put(1,0){\line(1,0){7}$ refers to the contraction of ${\bf a}_p^2 \in S^2({\frak g}^*)$ with
$\phi_2 \in S^2({\frak g}) \otimes \Lambda^2 V^*$. In other words,
(\ref{eq:newstruct4}) should be read as
\[ \begin{array}{lll}
(\xi_A {\bf b})(y) & = & {\bf b}(A \cdot y)\\
(\xi_x {\bf b})_p(y) & = & \phi_2({\bf a}_p, {\bf a}_p, x, y) + \tau (x, y).
\end{array} \]
Let us now define the Poisson structure on $W^* = {\frak g}^* \oplus V^*$
induced by $\phi := \phi_2 + \tau$, and let $\pi := {\bf a} + {\bf b}: F \rightarrow
W^*$. From (\ref{eq:struct3}) and (\ref{eq:newstruct4}) it follows that
$\pi_*(\xi_w) = \eta_w$ for all $w \in W$, and from there one can show
that, at least locally, the connection is indeed a Poisson connection
induced by $\phi$.
{\hfill \rule{.5em}{1em}\mbox{}\bigskip}
From the complete characterization in Theorem \ref{th:allarePoisson},
we can deduce the following properties which summarize our discussion
so far:
\begin{cor} \label{cor:summary} Let $M$ be a manifold which carries a
torsion-free connection whose holonomy is contained in one of the
groups $G \subseteq Gl(V)$ from Theorem \ref{thm:PoissonExist} other
than $Sl(2,{\Bbb R}) SO(2,{\Bbb R})$, and let $\phi = \phi_2 + c \sigma$ be the
admissible map which induces this connection. Then we have the
following.
\begin{itemize} \item[{\em (1)}] The connection is analytic.
\item[{\em (2)}] The map $\pi:= {\bf a} + {\bf b}: F \rightarrow W^*$ has constant even
rank $2k$ which we shall call the rank of the connection. $k=0$ iff the
connection is flat.
\item[{\em (3)}] $\pi(F)$ is contained in a $2k$-dimensional
characteristic leaf $\ \Sigma$ of the Poisson structure on $W^*$
induced by $\phi$. In particular, $\pi: F \rightarrow \Sigma$ is a submersion
onto its image.
\item[{\em (4)}] Conversely, given a characteristic leaf $\Sigma
\subseteq W^*$ of maximal rank, then $\Sigma$ can be covered by open
neighborhoods $\{ U_\alpha \}$ such that there are Poisson connections
with $\pi(F_\alpha) = U_\alpha$.
\item[{\em (5)}] Let ${\frak s} \subseteq {\frak X}(F)$ be the Lie algebra of
infinitesimal symmetries of the connection, i.e. those vector fields
whose flows preserve the connection. Then $dim({\frak s}) = dim (W^*) -
2k$. In particular, if $n \equiv 0,1 \bmod 4$ then $dim({\frak s}) >
0$.
\item[{\em (6)}] The moduli space of torsion-free connections with any
of the above holonomies is finite dimensional. In particular, the 2nd
derivative of the curvature at a single point in $M$ completely
determines the connection on all of $M$.
\end{itemize}
\end{cor}
\begin{proof} (2) and (3) follow from the identity $\pi_*(\xi_w) =
\eta_w$. (1) and (4) are clear from the construction of the Poisson
connections and the analyticity of $\phi$, whereas (6) follows from the
structure equations in the proof of Theorem \ref{thm:allareinduced}.
To show (5), let $f: W^* \rightarrow {\Bbb F}$ be a function which vanishes on
$\Sigma = \pi(F)$. Then it is easy to see that $\# \pi^*(df)$ is an
infinitesimal symmetry. It follows that $dim({\frak s}) \geq dim (W^*)
- 2k$. On the other hand, if $X \in {\frak s}$ then $\pi_*(X) = 0$,
hence $dim({\frak s}) \leq dim (W^*) - 2k$.
\end{proof}
Of course, (4) is not an optimal statement. One would like to show that
there are connections such that $\pi(F)$ is an {\em entire
characteristic leaf}. Also, our method does not prove the existence of
connections whose rank is not maximal. The difficulty is that, in
general, one cannot expect to have a {\em global} symplectic
realization $\pi: S \rightarrow W^*$. In fact, even if we restrict to the
subset $W^{reg} \subseteq W^*$ where the Poisson structure has maximal
rank, then the obstruction for the existence of a global symplectic
realization is given by a class in $H^3_{rel}(W^{reg},{\cal F})$, where
$\cal F$ is the foliation by symplectic leafs \cite{V}.
Also, (5) is not an optimal statement either. Namely, as was shown in
\cite{CS}, if $n = 3$ then the dimension of the Lie algebra ${\frak s}$
of infinitesimal symmetries is at least 2. In fact, it seems likely
that $dim({\frak s}) > 0$ in {\em any} dimension, even though we do not
have a proof of this assertion.
For $H_3$-connections, all statements in Corollary \ref{cor:summary}
have been shown in \cite{Br3} by {\sc MAPLE} calculations. The same
kind of calculations was used in \cite{CS} to prove
Corollary~\ref{cor:summary} for $n = 3$. Our current approach, however,
seems more conceptual. In fact, it was a deeper understanding of the
structure of $H_3$-connections which led us to the construction of
Poisson connections presented in this paper.
There is another global result whose analogue for $H_3$-connections has
been demonstrated in \cite{Sch}:
\begin{thm} \label{thm:incomplete} Let $M$ be a manifold which carries
a torsion-free connection whose holonomy is contained in one of the
groups $G \subseteq Gl(V)$ from {\em (\ref{replist})} other than
$Sl(2,{\Bbb R}) SO(n,{\Bbb R})$. Then $M$ is geodesically incomplete, unless the
connection is locally symmetric. If the latter is the case, then the
holonomy is a proper subgroup of $G$.
\end{thm}
\begin{proof} The holonomy of a locally symmetric connection must leave
the curvature tensor invariant. Since $K({\frak g})$ does not contain any
non-zero $G$-invariant element, the last statement follows.
Throughout this proof, we shall use the same notations and
identifications as in the proof of Theorem \ref{thm:PoissonExist}. In
particular, whenever it is convenient we shall regard ${\bf a}$ as a
${\frak g}$-valued form, and write ${\bf a}_p = A_p + M_p$ with $A_p \in {\frak sl}(2,{\Bbb F})$
and $M_p \in {\frak so}(n,{\Bbb F})$.
$M$ is geodesically complete iff the vector fields $\xi_x \in {\frak X}(F)$
are complete for all $x \in V$, and we shall assume this from now on.
To prove the Theorem, we have to show that the connection is locally
symmetric, which is the case iff ${\bf b} \equiv 0$. If $M_p \equiv 0$ then
(\ref{eq:struct3}) implies that ${\bf b} \equiv 0$. Thus, we assume that at
some $p_0 \in F$, $M_{p_0} \neq 0$ and ${\bf b}_{p_0} \neq 0$ and shall
deduce a contradiction.
Let ${\cal N} := \{ y \in {\Bbb F}^n\ |\ (y,y) = 0 \}$. By the indefiniteness
of $(\ ,\ )$, ${\cal N}$ spans all of ${\Bbb F}^n$. Therefore, ${\bf b}_{p_0}(e_1
\otimes y) \neq 0$ for some $e_1 \in {\Bbb F}^2, y \in {\cal N}$. Let $\nu :=
(M_{p_0} \cdot y, M_{p_0} \cdot y)$. If ${\Bbb F} = {\Bbb R}$ and $(\ ,\ )$ has
signature $(1, n-1)$, then it follows that $\nu \leq 0$. Therefore,
after possibly replacing $(\ ,\ )$ by its negative, we may assume that
$\nu \geq 0$ and, if ${\Bbb F} = {\Bbb R}$, the signature is $(p,q)$ with $p>1$. It
follows that there is a basis $\{x_1, \ldots, x_{n-2}, y, z\}$ of
${\Bbb F}^n$ and $e_1, e_2 \in {\Bbb F}^2$ such that the following hold:
\begin{equation} \label{eq:conditions} \begin{array}{c}
(x_i, x_j) = \delta_i^j \epsilon_i \mbox{\ \ with\ \ }
\epsilon_i = \pm 1, \mbox{\ \ \ \ } \epsilon_1 = 1,\\
(x_i, y) = (x_i, z) = (y, y) = (z, z) = 0, \mbox{\ \ \ \ }
(y, z) = 1,\\
<\! e_1, e_2 \!> = 1, \mbox{\ \ \ \ }
(M_{p_0} \cdot y, M_{p_0} \cdot y) \geq 0, \mbox{\ \ and\ \ }
{\bf b}_{p_0}(e_1 \otimes y) \neq 0.
\end{array} \end{equation}
Now, let us define
\[ \begin{array}{lllll}
\xi_1 := \xi_{e_1 \otimes x_1}, &
\xi_2 := \xi_{e_2 \otimes y}, &
A_0 := -e_1^2 & \mbox{and} &
M_0 := 2 y \wedge x_1,
\end{array} \]
and the functions
\[ \begin{array}{llll}
f_1(p) := {\bf a}_p(A_0), &
f_2(p) := {\bf a}_p(M_0) & \mbox{and} &
g(p) := - 2 {\bf b}_p(e_1 \otimes y).
\end{array} \]
From (\ref{eq:phi2}), (\ref{eq:struct3}),
(\ref{eq:newstruct4}) and (\ref{eq:conditions}) we calculate
\begin{equation} \label{eq:derivatives} \begin{array}{c}
\xi_1(f_1) = \xi_2(f_2) = 0, \mbox{\ \ \ \ }
\xi_1(f_2) = \xi_2(f_1) = g,\\
\xi_1(g) = 2 f_1 f_2, \mbox{\ \ and\ \ }
\xi_2(g) = 4 (M_p \cdot y, M_p \cdot y).
\end{array} \end{equation}
For $v \in {\Bbb F}^n$, we have $(v,v) = \sum_i \epsilon_i (v,x_i)^2 + (v,y)
(v,z)$. Thus,
\[
(M_p \cdot y, M_p \cdot y) = \sum_i \epsilon_i (M_p \cdot y, x_i)^2 =
\sum_i \epsilon_i ({\bf a}_p (y \wedge x_i))^2.
\]
We have ${\bf a}_p (y \wedge x_1) = B(M_p, y \wedge x_1) = \frac 12 f_2(p)$, and
from (\ref{eq:struct3}) and (\ref{eq:conditions}) we get that
$\xi_k({\bf a}_p (y \wedge x_i)) = {\bf b}_p((y \wedge x_i) \cdot \xi_k) = 0$ for $i>1$.
Therefore,
\begin{equation} \label{eq:squareplusconst} \begin{array}{ll}
\xi_k \left( \xi_2(g) - f_2^2 \right) = 0 & \mbox{for $k = 1,2$.}
\end{array} \end{equation}
For arbitrary constants $0 \neq c_1, c_2 \in {\Bbb F}$, define the vector
field $\xi$ and the function $f$ by
\[
\xi := c_1 \xi_1 + c_2 \xi_2 \mbox{\ \ and\ \ }
f := c_1^2 f_1 + 2 c_1 c_2 f_2.
\]
From (\ref{eq:derivatives}) we compute that
\[ \begin{array}{ll}
\xi^2(f) & = 3 c_1^2 c_2 \xi(g)\\
& = 3 c_1^2 c_2 (2 c_1 f_1 f_2 + c_2 \xi_2(g))\\
& = f^2 - c_1^2 \left( (c_1 f_1 - c_2 f_2)^2 - 3 c_2^2 (\xi_2(g) - f_2^2)
\right).
\end{array} \]
By (\ref{eq:squareplusconst}), $\xi \left( (c_1 f_1 - c_2 f_2)^2 - 3 c_2^2
(\xi_2(g) - f_2^2) \right) = 0$. It follows that along the flow line of
$\xi$, $f$ satisfies the differential equation
\begin{equation} \label{eq:ODE}
y'' = y^2 + C,
\end{equation}
where $C$ is a constant. By the assumption of completeness, $f$ must be
a {\em global} solution of (\ref{eq:ODE}). The following Lemma will be
shown in the appendix.
\mbox{}
\begin{lem} \label{lem:ODE} \begin{itemize}
\item[{\em (i)}] The only holomorphic solutions of {\em (\ref{eq:ODE})}
which are defined on all of ${\Bbb C}$ are constants.
\item[{\em (ii)}] Suppose there is a real solution of {\em
(\ref{eq:ODE})} which is defined on all of ${\Bbb R}$. If for $x_0 \in {\Bbb R}$,
$y(x_0) > 0$ and $y'(x_0) \neq 0$, then $y''(x_0) < 0$.
\end{itemize} \end{lem}
We shall use this Lemma to get the desired contradiction. If ${\Bbb F} = {\Bbb C}$,
the Lemma implies that $3 c_1^2 c_2 g = \xi(f) \equiv 0$, contradicting
$g(p_0) = -2 {\bf b}_{p_0}(e_1 \otimes y) \neq 0$.
Consider now the case ${\Bbb F} = {\Bbb R}$. From (\ref{eq:derivatives}) and
(\ref{eq:squareplusconst}) we have $(\xi_2)^3(f_1) = 0$. Moreover,
$(\xi_2)^2(f_1)_{p_0} = 4 (M_{p_0} \cdot y, M_{p_0} \cdot y) \geq 0$
and $\xi_2(f_1)_{p_0} = g(p_0) \neq 0$. Therefore, since $\xi_2$ is
complete, there is a point $q_0$ on the flow line of $\xi_2$ through
$p_0$ which, in addition to (\ref{eq:conditions}), also satisfies
$f_1(q_0) > 0$. W.l.o.g. we assume that $q_0 = p_0$.
Now $y(x_0) = f(p_0) = (c_1^2 f_1 + 2 c_1 c_2 f_2)(p_0)$ and $y''(x_0)
= \xi^2(f)_{p_0} = 3 c_1^2 c_2 (2 c_1 f_1 f_2 + c_2 \xi_2(g))(p_0)$.
Since $f_1(p_0) > 0$ and $\xi_2(g)_{p_0} \geq 0$ it follows that for a
suitable choice of $c_1, c_2$, both $y(x_0) > 0$ and $y''(x_0) \geq 0$
can be achieved. But $y'(x_0) = g(p_0) \neq 0$, hence again, we get a
contradiction from the Lemma.
\end{proof}
It is now clear that Theorem \ref{thm:summary} is simply a restatement
of parts of Corollary \ref{cor:summary} and Theorem
\ref{thm:incomplete}.
\pagebreak
\section{Appendix}
We shall now give the proof of two technical Lemmas.
\noindent {\bf Lemma \ref{lem:reps}} {\em Let $G \subseteq Gl(V)$ be an
irreducible representation of a connected, reductive Lie group $G$, and
let ${\frak g} \subseteq {\frak gl}(V)$ be the corresponding Lie algebra. If $\tau
\in V^* \otimes V^*$ satisfies the condition
\[
\tau(x, A \cdot y) = \tau(y, A \cdot x) \mbox{\ \ for all $x,y \in V$
and $A \in {\frak g}$,}
\]
then $\tau$ is skew-symmetric and hence a $G$-invariant 2-form.}
\begin{proof} Clearly, the Lemma is invariant under complexification,
hence we may assume that $G, {\frak g}$ and $V$ are complex. Also, if ${\frak g}_s
\subseteq {\frak g}$ is the semi-simple part of ${\frak g}$, then ${\frak g}_s$ acts
irreducibly on $V$ as well. Thus, we may assume w.l.o.g. that ${\frak g}$ is
semi-simple. Let
\[ \begin{array}{lll}
{\frak g} = {\frak t} \oplus \bigoplus_{\alpha \in \Delta} {\frak g}_{\alpha} & \mbox{and} &
V = \bigoplus_{\lambda \in \Lambda} V_\lambda
\end{array} \]
be a Cartan and weight space decomposition of ${\frak g}$ and $V$,
respectively. We shall denote elements of ${\frak t}$, ${\frak g}_\alpha$ and
$V_\lambda$ by $A_0$, $A_\alpha$ and $x_\lambda$, respectively. Also,
we always write $\alpha, \beta, \ldots$ for roots, whereas $\lambda,
\mu, \ldots$ stand for weights. Let $\lambda_{max} \in {\frak t}^*$ be the
maximal weight, and $x_{max} \in V_{\lambda_{max}}$ be a highest weight
vector.
The proof now proceeds as follows. Given $\tau$ as above, we shall
prove:
\[ \begin{array}{ll}
\mbox{Step 1:} & \mbox{if ${\frak g} \cong {\frak sl}(2,{\Bbb C})$ and $\lambda + \mu \neq
0$, then $\tau(x_\lambda, x_\mu) = 0$.}\\
\mbox{Step 2:} & \mbox{$\tau(x_\lambda, x_\mu) \mu = \tau(x_\mu,
x_\lambda) \lambda$ for all $\lambda, \mu \in \Lambda$. In particular,
$\tau(x_\lambda, x_\mu) = 0$ if $\lambda, \mu$}\\ & \mbox{are linearly
independent.}\\
\mbox{Step 3:} & \mbox{if $\lambda, \mu$ are scalar multiples of
$\lambda_{max}$ and $\lambda, \mu, \lambda + \mu \neq 0$ then}\\ &
\tau(x_\lambda, x_\mu) = \tau(x_\mu, x_\lambda) = 0.\\
\mbox{Step 4:} & \mbox{$\tau(x_{max}, x_0) = \tau(x_0, x_{max}) =
0$.}\\
\mbox{Step 5:} & \mbox{$\tau$ is skew-symmetric.}
\end{array} \]
{\em Proof of step 1:} Suppose ${\frak g} \cong {\frak sl}(2, {\Bbb C})$, and $V \cong V_n$
is the (unique) irreducible $n + 1$-dimensional representation of ${\frak g}$.
Then it is an easy exercise to show that the only elements $\tau \in
V^* \otimes V^*$ which satisfy the above identity are $\tau = 0$ if $n$
is even, and a multiple of the symplectic ${\frak g}$-invariant 2-form on $V$
if $n$ is odd. From this, step 1 follows. The proof is omitted.
{\em Proof of step 2:} $\tau(x_\lambda, A_0 \cdot x_\mu) = \tau(x_\mu,
A_0 \cdot x_\lambda)$, hence $\mu(A_0) \tau(x_\lambda, x_\mu) =
\lambda(A_0) \tau(x_\mu, x_\lambda)$ for all $A_0 \in {\frak t}$. This proves
step 2.
{\em Proof of step 3:} Let $\lambda, \mu$ be as above, and let $\alpha
\in \Delta$ be linearly independent of $\lambda_{max}$. We compute
$\tau(x_\lambda, A_\alpha A_{-\alpha} \cdot x_\mu) = \tau(A_{-\alpha}
\cdot x_\mu, A_\alpha \cdot x_\lambda) = 0$ by step 2; indeed,
$A_{-\alpha} \cdot x_\mu \in V_{\mu - \alpha}$, $A_\alpha \cdot
x_\lambda \in V_{\lambda + \alpha}$, and $\mu - \alpha$, $\lambda +
\alpha$ are linearly independent if $\lambda + \mu \neq 0$. Likewise,
$\tau(x_\lambda, A_{-\alpha} A_\alpha \cdot x_\mu) = 0$, and hence $0 =
\tau(x_\lambda, [A_\alpha, A_{-\alpha}] \cdot x_\mu) = \mu ([A_\alpha,
A_{-\alpha}]) \tau(x_\lambda, x_\mu)$. Thus, we have $\lambda_{max}
([A_\alpha, A_{-\alpha}]) \tau(x_\lambda, x_\mu) = 0$ and similarly,
$\lambda_{max} ([A_\alpha, A_{-\alpha}]) \tau(x_\mu, x_\lambda) = 0$.
But if $\lambda_{max} ([A_\alpha, A_{-\alpha}]) = 0$ for all roots
$\alpha$ independent of $\lambda_{max}$, then it follows that ${\frak g}$ must
contain ${\frak sl}(2,{\Bbb C})$ as a summand, and $\lambda_{max}$ is a multiple of
the root corresponding to this summand.
In this case, however, since the representation is faithful, it must be
equivalent to the action of ${\frak sl}(2,{\Bbb C})$ on $V_n$. This observation
together with step 1 implies step 3.
\pagebreak
{\em Proof of step 4:} The equation $\tau(x_0, x_{max}) = 0$ follows
immediately from step 2.
Since the representation is irreducible, we have $V_0 = span
\{A_{-\alpha} \cdot V_\alpha\ |\ \alpha \in \Delta\}$. Thus, in order
to show step 4, we need to show
\begin{equation} \label{eq:lemproof}
\tau(x_{max}, A_{-\alpha} \cdot x_\alpha) = 0 \mbox{ for all } \alpha
\in \Delta, x_\alpha \in V_\alpha.
\end{equation}
If $\alpha \neq \lambda_{max}$ we have $\tau(x_{max}, A_{-\alpha} \cdot
x_\alpha) = \tau(x_\alpha, A_{-\alpha} \cdot x_{max}) = 0$ by steps 2
and 3; indeed, $A_{-\alpha} \cdot x_{max} \in V_{\lambda_{max} -
\alpha}$ and $\lambda_{max} - \alpha \notin \{ 0, -\alpha \}$. Thus,
(\ref{eq:lemproof}) follows in this case.
Next, suppose that $\alpha = \lambda_{max}$, and let $\beta$ and
$x_{\alpha + \beta}$ be such that $x_{max} = A_{-\beta} \cdot x_{\alpha
+ \beta}$. By irreducibility, this is always possible. Note that $\beta
\neq \alpha$, since by maximality of $\alpha = \lambda_{max}$, $2
\alpha$ is not a root. Then
\[ \begin{array}{ll}
\tau(x_{max}, A_{-\alpha} \cdot x_\alpha) & =
\tau(x_\alpha, A_{-\alpha} A_{-\beta} \cdot x_{\alpha + \beta})\\ & =
\tau(x_\alpha, ([A_{-\alpha}, A_{-\beta}] +
A_{-\beta} A_{-\alpha}) \cdot x_{\alpha + \beta})\\ & =
\tau(x_{\alpha + \beta}, [A_{-\alpha}, A_{-\beta}] \cdot x_\alpha) +
\tau(A_{-\alpha} \cdot x_{\alpha + \beta}, A_{-\beta} \cdot x_\alpha)\\
& := I + II.
\end{array} \]
Now $x_{\alpha + \beta} \in V_{\alpha + \beta}$ and $[A_{-\alpha},
A_{-\beta}] \cdot x_\alpha \in V_{-\beta}$. Thus, by step 2 and the
fact that $\alpha \neq \beta$ it follows that $I = 0$. For $II$, we
have $A_{-\alpha} \cdot x_{\alpha + \beta} \in V_\beta$ and $A_{-\beta}
\cdot x_\alpha \in V_{\alpha - \beta}$. Now if $\beta$ is independent
of $\alpha$, step 2 implies that $II = 0$. If $\beta = -\alpha$ then
$A_{-\beta} \cdot x_\alpha \in V_{\alpha - \beta} = V_{2 \alpha} = 0$
by maximality of $\lambda_{max} = \alpha$, and again, $II = 0$
follows. Thus, (\ref{eq:lemproof}) is shown in this case as well, and
step 4 follows.
{\em Proof of step 5:} Let $\tau^{sym}(x,y) := \tau(x,y) + \tau(y,x)$.
From steps 2 through 4, it follows that $\lambda_{max} + \lambda \neq
0$ implies that $\tau(x_{max}, x_\lambda) = \tau(x_\lambda, x_{max}) =
0$. Also, if $x_{-max} \in V_{-\lambda_{max}}$, then by step 2,
$\tau^{sym}(x_{max}, x_{-max}) = 0$.
Thus, we conclude that $\tau^{sym}(x_{max},\underline{\ \ }) = 0$. But this
is true for {\em any} choice of Cartan decomposition. Exploiting all
possible decompositions, we conclude
\[
\tau^{sym}(g \cdot x_{max}, \underline{\ \ }) = 0 \mbox{ for all } g \in G.
\]
By irreducibility, the $G$-orbit of $x_{max}$ spans all of $V$, and
from there it follows that \linebreak $\tau^{sym} = 0$, hence $\tau$ is
skew-symmetric.
\end{proof}
\noindent {\bf Lemma \ref{lem:ODE}} {\em Consider the differential
equation $y'' = y^2 + C$ {\em (\ref{eq:ODE})} for a constant $C$.
\begin{itemize} \item[{\em (i)}] The only holomorphic solutions of {\em
(\ref{eq:ODE})} which are defined on all of ${\Bbb C}$ are constants.
\item[{\em (ii)}] Suppose there is a real solution of {\em
(\ref{eq:ODE})} which is defined on all of ${\Bbb R}$. If for $x_0 \in {\Bbb R}$,
$y(x_0) > 0$ and $y'(x_0) \neq 0$, then $y''(x_0) < 0$.
\end{itemize}}
\begin{proof} First of all, we integrate (\ref{eq:ODE}) to
\begin{equation} \label{eq:ODE2}
(y')^2 = \frac 23 y^3 + 2 C y + C_1
\end{equation}
for some constant $C_1$.
(i) Let $y$ be a holomorphic solution which is an entire function.
Define
\[ \begin{array}{llll}
\phi: & {\Bbb C} & \longrightarrow & {\Bbb C} {\Bbb P}_2\\ & x & \longmapsto &
[y(x) : y'(x) : 1].
\end{array} \]
By (\ref{eq:ODE2}), the image of $\phi$ is contained in the cubic $\cal
C$ given by $3 q^2 r - 2 p^3 - 6 C p r^2 - 3 C_1 r^3 = 0$, where
$p,q,r$ are the coordinates in ${\Bbb C} {\Bbb P}_2$.
We calculate that $\cal C$ is a regular curve of genus 1 if $16 C^3 + 9
C_1^2 \neq 0$; it is a rational curve with a node if $16 C^3 + 9 C_1^2
= 0$ and $C \neq 0$, and it is a rational curve with a cusp if $C = C_1
= 0$.
Since $im(\phi)$ does not contain the point $[0:1:0] \in \cal C$, it
follows that in the first case $\phi$ maps ${\Bbb C}$ non-surjectively to a
genus 1 curve. By Picard's Theorem, this implies that $\phi$, and hence
$y$, is constant.
In the second case, we find a parametrization of $\cal C$ and compute
that there must be a function $t = t(x)$ such that
\[ \begin{array}{lll}
y = \frac 3 {2C} \left( C_1 - 4 C^2 t^2 \right) & \mbox{and} &
y' = \frac{2 C} {C_1} \left( 3 C_1 - 8 C^2 t^2 \right) t.
\end{array} \]
But this implies that either $t \equiv 0$ or $t' = \frac{4 C^2}{3 C_1}
t^2 - \frac 12$. It is straightforward to verify that the only global
solutions of the latter equation are constants.
In the third case $C = C_1 = 0$, a parametrization is given by $y = 6
t^2, y' = 12 t^3$ for some entire function $t$. But then $12 t t' = 12
t^3$, which again has no global non-constant solution.
(ii) Consider a global real solution $y$ and $x_0 \in {\Bbb R}$ as above.
After possibly replacing $y(x)$ by the solution $y(-x)$ and $x_0$ by
$-x_0$, we may assume that $y'(x_0) > 0$. If $y^2(x_0) + C = y''(x_0)
\geq 0$, then elementary calculus shows that $\lim_{x \rightarrow \infty} y =
\infty$. Thus, from (\ref{eq:ODE2}), we have for sufficiently large
$x$ that $(y')^2 > \frac 13 y^3$, and hence
\[
\left( y^{-\frac 12} \right)' = - \frac 12 y^{-\frac 32} y' <
- \frac 1 {2 \sqrt 3},
\]
contradicting $y^{-\frac 12} > 0$ for large $x$. Thus, $y''(x_0) \geq
0$ is impossible.
\end{proof}
|
\section{ Introduction }
The goal of this paper to study the Landau - Pomeranchuk - Migdal (LPM) effect
\cite{LPM} for the emission of photons and gluons in nuclear matter.
We show that the typical distances which characterize
the succesive rescaterring of quarks and gluons in nuclear matter are
small enough to justify the QCD approach to this case.
We shall
generalized the formalism suggested in refs. \cite{GW}\cite{BDPS} to the case
of nuclear matter using the QCD approach for the interaction
of quarks and gluons.
We consider the emission of soft gluons ( photons ) with
energies $ \omega \,\ll
\, E$, where $E$ is the energy of the projectile. The assumption that the mean
free path $\lambda$ of the projectile is much larger than the screening radius
in the nuclear matter, $ \lambda\,\gg\,\mu^{-1}$, allows one to treat successive
rescatterings as independent ( see ref. \cite{LPM} ) and simplifies
all formulae reducing the problem to an eikonal picture of classical
propagation of a relativistic particle with $E\,\gg\, \mu$ through a medium.
As is well known, the QED emission amplitude for a single scattering
can be written in terms of a transverse velocity ${\vec{u}}_{\perp}\,=\,
\,\frac{{\vec{k}}_{\perp}}{\omega}$ where ${\vec{k}}_{\perp}$ is the transverse
momentum of the photon with respect to the direction of the fast (massless)
charged
particle. For the emission of a photon from the scattering center
$i$ it reads:
\begin{equation}
\vec{J}_{i}\,\,=\,\,\frac{{\vec{u_i}}_{\perp}}{u^2_{i \perp}}\,\,-\,\,
\frac{\vec{u}_{i - 1 \perp}}{u^2_{i - 1, \perp}}\,\,;\,\,\,\,
\vec{u}_{i \perp}\,\,=\,\,\vec{u}_{(i - 1) \perp}\,\,-\,\,\frac{\vec{
q}_{i \perp}}{E}\,\,;
\end{equation}
where
$\vec{q_{i\perp}} $ is the momentum transfer to the scattering center $i$.
The LPM effect comes from the intenference between the amplitudes (1) in
the calculation of the inclusive cross section of photon production.
The relative
eikonal phase of the two amplitudes due to centers $i $ and $j$ in
relativistic kinematics is equal to:
\begin{equation}
\phi_{ij}\,\,=\,\,k^{\mu}(x_i - x_j)_{\mu}
\,\,=\,\,\sum^{j - 1}_{m = i}\,\,\frac{z_m - z_{m +
1}}{\tau_m(\omega)}\,\,;\,\,\,\,
\tau_m\,=\,\frac{2\omega}{(\vec{k}_{m \perp})^2}\,=\,
\frac{2}{\omega u^2_{m \perp}}\,\,;
\end{equation}
where $ \tau_m$ is the radiation formation time
and $z_m$ is the longitudinal coordinate of the
$m$-th centre.
One can easily estimate the average phase \cite{LPM}, assuming that all
$z_m - z_{m + 1}$ are equal to the mean free path $\lambda$, and obtain:
\begin{equation} \label{PHI}
\phi( n = j - i )\,\,\approx\,\,\frac{\lambda \omega}{2} \,\sum^{j - i - 1}_{m = 0} \,\,
u^2_{i + m}\,\,\approx\,\,\frac{\lambda \omega}{2}\,\,\{\, n u^2_{i \perp}\,\,+\,\,
\frac{n ( n - 1 )}{2} \,\frac{\mu^2}{E^2}\,\}
\end{equation}
Here we have
used the assumption that the transverse momentum of the projectile,
after $m$ collisions with a typical momentum transfer $\VEV{q^2_{\perp}}
\,=\,\mu^2$, can be treated as a random walk in the transverse plane which
gives :
$ u^2_{i + m,\perp}\,\,=\,\,u^2_{i \perp} \,\,+\,\,m \,\frac{\mu^2}{E^2}$.
Since the amplitude of Eq.(1) is small at
$u^2_{i \perp}\,\,>\,\,\frac{\mu^2}{E^2}$ the second term in \eq{PHI} dominates
at large $ N\,\,\gg\,\,1$.
The cross section vanishes if $\phi \,>\,1$. It means that the number
( $n$ ) of scattering centres
that act coherently and can be treated as a single radiator,
is restricted from above by the value $N_{coh}( \omega )$:
\begin{equation} \label{NCOH}
n\,\,\leq\,\,N_{coh}( \omega )\,\,=\,\,2 \,\sqrt{\frac{ E^2}{\lambda \mu^2 \omega}}
\end{equation}
Considering a group of centeres with $ i - j \,<\,N_{coh}$ as a single
radiator, we can estimate the radiation density ( see refs. \cite{LPM}
\cite{BDPS} \cite{LPMREV} for details):
\begin{equation} \label{RADAPPR}
\omega\,\frac{d^2 I}{ d \omega \,d z}\,\,=\,\,\frac{1}{\lambda} \,
\{\omega \frac{d I}{ d \omega}\}_{one\,\,centre}\,\cdot\,\frac{1}{N_{coh}( \omega )}
\,\,\propto\,\,\frac{\alpha_{em}}{\lambda N_{coh}}
\end{equation}
One can see three kinematic regions in which \eq{RADAPPR} gives different
answers for the radiation density:
1. $N_{coh}( \omega )\,\,>\,\,1 $ or $\omega \,<\,\omega_{BH}\,=\,\frac{2 E^2}{\lambda
\mu^2}\,=\,
E \cdot\frac{E}{E_{LPM}}$ where $ E_{LPM} \,=\,\frac{1}{2} \mu^2 \lambda$.
Here
\begin{equation}
\omega\,\frac{d^2 I}{ d \omega \,d z}\,\,\propto\,\,\alpha_{em} \,\sqrt{\frac{\mu^2}
{\lambda E^2}\,\omega}\,=\,\frac{\alpha_{em}}{\lambda} \,\sqrt{\frac{\omega}{\omega_{BH}}}
\end{equation}
This equation gives the Migdal result \cite{LPM}.
2. $N_{coh}(\omega)\,\,\approx\,\,1$ or $\omega \,\geq \omega_{BH}$.
In this kinematic region each scattering centre radiates separately and
we derive the well known Bethe - Heitler limit:
\begin{equation} \label{BH}
\omega\,\frac{d^2 I}{ d \omega \,d z}\,\,=\,\,\frac{1}{\lambda} \,
\{\,\omega \frac{d I}{ d \omega}\}_{one\,\,centre}\,\propto \,\frac{\alpha_{em}}{\lambda}
\end{equation}
3. In a medium with a final longitudinal size $L$, we have to assume
that $L$ is large enough so as to have $N_{coh}$ successive interactions.
It means:
\begin{equation} \label{FINSIZE}
\lambda N_{coh}\,\,<\,\,L \,\,\,\,or\,\,\,\,
\omega\,\,>\,\,\omega_{fact}\,=\,\frac{\lambda E^2}{L^2\mu^2}
\end{equation}
For $\omega \,\
<\,\omega_{fact}$ one recovers the factorization limit in which the whole
medium radiates as one centre with an amplitude
\begin{equation}
\sum^n_{i=1}\,\vec{J_i}\,\,=\,\,\frac{\vec{u}_{n\perp}}{u^2_{n \perp}}\,\,
-\,\,
\frac{\vec{u}_{1 \perp}}{u^2_{1 \perp}}\,\,;
\end{equation}
At first sight, the factorization region is irrelevant for the emission in a
nuclear matter. However, it is not so, since each produced quark or gluon
lives only a finite time ( $ \tau_{max} \,\propto\,
\frac{E}{\VEV{p^2_{\perp}}}$
where $p_{\perp}$ is the transverse momentum of the
produced charged particle which
embodies successive rescattering)
inside a nuclear matter. Substituting $ L = \tau_{max}$ in \eq{FINSIZE}
we obtain $\omega_{fact}\,=\,\frac{\lambda \VEV{p^2_{\perp}}}{\mu^2}$.
Therefore we expect the LPM effect for the photon energy range
\begin{equation}
\frac{\lambda \VEV{p^2_{\perp}}}{\mu^2}\,=\,\omega_{fact}\,<\,\omega\,<\,\omega_{BH}\,
=\,\frac{E^2}{E_{LPM}}
\end{equation}
The situation becomes even more interesting in the case of gluon emission.
As was discussed in ref.\cite{BDPS}, the produced gluon itself embodies the
successive rescatterings in a medium. The life time of the
gluon is $\tau_{max}\,
\propto\,\frac{\omega}{k^2_{\perp}}\,\,\approx\,\,\frac{1}{\omega}$. To discuss
the contribution of gluon rescattering to the LPM effect we have to demand
that
\begin{equation}
\lambda N_{coh} ( \omega ) \,\,<\,\,\tau_{max} ( gluon ) \,\,\approx \,\,\frac{1}{\omega}\,
\,\,\,or\,\,\,\,
\omega \,\,<\,\,\frac{1}{\lambda^2 \omega_{BH}}
\end{equation}
Therefore~~ we~~ anticipate~~ for~ gluons~~ the classic~~ LPM
suppression~~ only~~ for
$\omega \,\,<\,\,min\{\,\omega_{BH},\frac{1}{\lambda^2 \omega_{BH}}\,\}$
and\,~~ some ~~\,\,new behaviour of\,\,~~ the radiation \,\, density\,\,~~ for
$
min\{\,\omega_{BH},\frac{1}{\lambda^2 \omega_{BH}}\,\}$
$ \,\,\,<\,\,\,\omega\,\,<\,\,\,$
$ max\{\,\omega_{BH},\frac{1}{\lambda^2 \omega_{BH}}\,\}
$.
{}From the above sketch of the physics of the LPM effect it is clear that
there are a number of questions that we have to answer: (i)
what is the value of
the mean free path $\lambda$ in QCD; (ii)
what is the correct procedure of averaging
over transverse momentum and what is the value of $\mu$; and, (iii) what kind of
LPM supression do we expect for gluon emission. We will answer these
questions in this paper, starting with the photon emission
from a fast quark in a
nuclear matter in the second section.
In the third section we shall consider the interconnection
between the LPM effect
and the Mueller technique \cite{MUT}, which was widely used to calculate
the inclusive spectra of produced particles.
In the fourth section we discuss the LPM like supression for gluon emission
in a nuclear matter.
A summary and final discussion are presented in the conclusion.
In general, we attempt to adjust our notation and normalization to agree with
those of Baier et al \cite{BDPS}, using $E$ and $p$ for
the energy and momentum of a projectile which propagates through
a nuclear matter;
$\omega$ and $k$ for the energy and momentum of a radiated photon (gluon) and
$q_{i \perp}$ for the momentum transfer to the $i$-th scattering centre.
\section{ The LPM effect for photon radiation}
The inclusive soft photon spectrum from
a fast quark with energy $E$ can be written in the form ( see for example
\cite{BDPS}):
\newpage
\begin{equation} \label{INCSP}
\omega\frac{dI}{d\omega\,d^2 u} =
\frac{\alpha_{em}}{\pi^2}
\VEV{ \sum^{N}_{i=1}\,\vert \vec{J}_i e^{i\,k_{\mu} x^{\mu}}\vert^2}\,\,=
\end{equation}
$$\frac{\alpha_{em}}{\pi^2}
\VEV{ 2 \mathop{\rm Re} \,\sum^{N}_{i=1} \sum^{N}_{j=i+1} \, \vec{J}_i \vec{J}_j
\left[\, e^{ik_{\mu} (x_i-x_j)^{\mu}}\,\, -1\,\right]\,\,+\,\,
\vert \sum^N_{i=1} \vec{J}_i\vert^2 }
$$
The brackets $\VEV{\,\, }$ indicate the averaging procedure discussed in
\cite{GW} \cite{BDPS}, and we shall discuss it below in detail.
The differential energy distribution of photons radiated
{\em per unit length}\/ is given by \cite{BDPS}
\begin{equation} \label {ENSP}
\omega\frac{dI}{d\omega\, dz} =
\, \frac{\alpha_{em}}{\pi}\,
\int\frac{d^2 U_0}{\pi} \VEV{ 2\mathop{\rm Re}\, \sum^{\infty}_{n=0}
\vec{J}_1 \vec{J}_{n+2}
\left[\,
\exp\left\{ i \frac{1}{\tau} \sum^{n + 1}_{l=1}\ U^2_l\,(
\,z_{l +1}-z_l\,)\,\right\}\,\,\, -1 \,\right] }.
\end{equation}
Here we have expressed the relative phases in terms of the
rescaled transverse velocity,
\begin{equation}\label{UDEF}
\vec{U}_{l} \equiv \vec{u}_{l} \cdot E\,\,=\,\,\frac{\vec{k}_{\perp}}{x_F}
\,\,-\,\,\sum^{l}_{i = 1} \,\vec{q}_{i\,\perp}\,\,;\,\,\,\,\,
\vec{U}_{0}\,\,=\,\,\frac{\vec{k}_{\perp}}{x_F}
\end{equation}
where $x_F \,\,=\,\,\frac{\omega}{E}$ is the fraction of quark energy
carried by the photon.
We also introduced the characteristic parameter
\begin{equation} \label{TAUDEF}
\frac{1}{\tau}\,\, = \,\,\frac{\omega}{2\,E^2}
\end{equation}
Now we shall
define the averaging procedure that has to be implemented in \eq{ENSP},
starting with the simplest problem, namely, with the propagation of the
quark through a nuclear matter. This problem was solved a long time ago
\cite{LR1}.
Indeed, let us consider the inclusive spectrum of a fast quark
produced at a point $z$, with respect to
its transverse momentum ( $p_{\perp}$ )
after $n$ rescatterings with nucleons at points
$z_1,z_2,
...z_n$. This spectrum can be written as \cite{LR1} ( see Fig.1)
$$
\frac{d \sigma_n}{d^2 p_{\perp}}\,\,=\,\,
e^{- \sigma_{tot}\,(\, z_1 - z\,)} \,d z_1 \,\rho\,\frac{d \sigma}{ d^2 q^2_{1 \perp}}
\,\cdot\,e^{- \sigma_{tot}\,(\, z_2 - z_1\,)} \,d z_1
\,\rho\,\frac{d \sigma}{ d^2 q^2_{2 \perp}}\,...\,
e^{- \sigma_{tot}\,(\,z_1 + L - z_n\,)} \,d z_n \,\rho\,\frac{d \sigma}
{ d^2 q^2_{n \perp}}
$$
\begin{equation} \label{QUARKSP}
\,\,\delta^{( 2 )} \(\,\sum^{n}_{m = 1} \,{\bf \vec{q}_{m \perp}\,\,-\,\,
\vec{p}_{\perp}\,}\) \,\prod_{m} \,d^2 q_{m \perp}\,\,\,\,\,\,(\,z_1\,<\,z_2
\,<...<\,z_n \,\,<\,z_1 + ( L\,=\,\tau_{max})\,)
\end{equation}
Here, $\rho$ is the nucleon density, $\frac{d \sigma}{ d^2 q_{m \perp}}$ is the
spectrum of the quark due to rescattering with momentum transfer $q_{m \perp}$
which can be written through the unintegrated parton density $\phi$
\cite{LR1} ( see also relevant papers \cite{LR2} \cite{MU90}):
\begin{equation} \label{QCRSEC}
\frac{d \sigma}{ d^2 q_{m \perp}}\,\,=\,\,\frac{8\pi^2 \alpha_{\rm S}^2 ( q^2_{ m \perp} )}{ 9
\,
q^4_{m \perp}}\,\cdot\,( x \phi_N (x, q^2_{n \perp}))
\end{equation}
where $ x \phi_N( x, q^2_{\perp} )$ is the nucleon parton density
with \,\,\,$x =
\frac{1}{\tau_{max}}$. The relation between the unintegrated parton
density $\phi$ and the nucleon's gluon structure function
$x G_N (x, q^2_{\perp})$ can be calculated using the following
equation:
$$
\alpha_{\rm S}(Q^2)\,\cdot\,x G(x, Q^2)\,\,=\,\,\int^{Q^2}\,\,d\,q^2_{\perp}\,
\alpha_{\rm S}(q^2_{\perp})\,\,\phi (x, q^2_{\perp})
$$
The factor $ exp ( - \sigma_{tot}$\,\,
$ \,(\, z_m\, -\,z_{m - 1}\,)\,)$ in \eq{QUARKSP}
describes the fact that the quark has no inelastic
collisions between the centres $z_{m - 1}$
and $ z_m$. The total cross section is equal to $\sigma_{tot} \,=\,\int
\,d^2 q_{m \perp} \,\frac{d \sigma}{ d^2 q_{m \perp}}$ and it is formally
divergent at $q_{m \perp} \,\rightarrow\,0$. However, this divergency cancels against
the divergency of the inelastic cross section. To see this fact, it is
convenient to describe the process in the transverse coordinate space
$r_{\perp}$ using the well know representation of the $\delta$ function:
$$
\int \,\frac{d^2 r_{\perp}}{ ( 2 \pi )^2} \,\exp[ i \, \sum_m
{\bf\( \vec{q}_{m \perp} \,\,-\,\,\vec{p}_{\perp}\,\)\,\cdot
\,\vec{r}_{\perp}}]
\,\,=\,\,\delta^{( 2 )} \(\,\sum^{n}_{m = 1} \,{\bf \vec{q}_{m \perp}\,\,-\,\,
\vec{p}_{\perp}\, }\)
$$
Summing over $n$ we obtain
\begin{equation} \label{INCQUARK}
\frac{d \sigma}{d^2 p_{\perp}}\,\,=\,\,\int \,\frac{d r^2_{\perp}}{ 4\, \pi^2 }
\,\,J_0 (\, p_{\perp}\,r_{\perp}\,)\,\,\exp\( -\,\sigma(\, r^2_{\perp}\, )\,
\rho\,L\,\)
\end{equation}
with
\begin{equation} \label{QUARKCRSEC}
\sigma ( r^2_{\perp})\,=\,\int d q^2_{\perp}
\,\{\, 1\, - \,J_0 (\, q_{\perp}\,
r_{\perp}\, \}\,\frac{d \sigma}{ d q^2_{\perp}}
\,\,=\,\,\frac{\alpha_{\rm S}(\frac{4}{r^2_{\perp}})}{3}
\,\,\pi^2\,\,r^2_{\perp}\,\,
\(\,\, x G_N( \frac{4}{r^2_{\perp}},x )\,\,\)
\end{equation}
Let us investigate $ \frac{d \sigma}{d^2 p_{\perp}}$, adopting the following method
of integration over $r_{\perp}$ in \eq{INCQUARK}:(i) we consider
$xG_N(\frac{4}{r^2_{\perp}}, x)$ as a smooth function of $r^2_{\perp}$ since
it depends on $r^2_{\perp}$ only logarithmically; (ii) we integrate over
$r_{\perp}$ taking only into account $\sigma \,\propto \,r^2_{\perp}$; and,
(iii) we put in $xG_N(\frac{4}{r^2_{\perp}}, x)$
$r^2_{\perp} \,=\,r^2_{0 \perp}$, where $r^2_{0 \perp}$ is the typical value
of $r^2_{\perp}$ in the integral.
The result of the integration is obvious, namely
\begin{equation} \label{RDEP}
\frac{d \sigma}{d^2 p_{\perp}}\,\,=\,\,\frac{1}{ 4 \,\rho\,L\,\,
\pi^2\,[\, \frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)\,]}\,\,e^{ -\,\,
\frac{p^2_{\perp}}{4 \,\rho\,L\,\, \frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)}}
\end{equation}
One can find the value of $r^2_{0 \perp}$ by calculating the saddle point in
$r_{\perp}$ integration, which gives:
\begin{equation} \label{TYPR}
r^2_{0 \perp}\,\,=\,\,\frac{p^2_{\perp}}{ 4 \,( \,\rho\,L\,\,
\frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)\,)^2} \,\,\,\,for\,\,\,\,
p^2_{\perp}\,r^2_{0 \perp}\,\,\gg\,\,1\,\,;
\end{equation}
$$
r^2_{0 \perp}\,\,=\,\,\frac{1}{ 2 \,( \,\rho\,L\,\,
\frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)\,)} \,\,\,\,for \,\,\,\,
p^2_{\perp}\,r^2_{0 \perp}\,\,\ll\,\,1\,\,;
$$
We can trust our calculation in pQCD only if $r_{\perp} \,\,\ll\,\,r_{soft}$,
where $r_{soft} $ is the scale at which the nonperturbative QCD corrections
become essential. It means that the value of $p_{\perp}$ should be
smaller than $ 2 \,( \,\rho\,L\,\,
\frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)\,)$. For bigger $p_{\perp}$, in
\eq{INCQUARK}, we take $r_{\perp} \,\propto \,\frac{1}{p_{\perp}}$ which gives
$\frac{d \sigma}{d^2 p_{\perp}}\,\,\propto\,\,\frac{1}{p^2_{\perp}}\,
\sigma(\frac{1}{p^2_{\perp}})\,\, \rho\,L$.
The main message from this calculation is that
the typical distances that work in the rescattering inside the nuclear matter
turns out to be small. This justifies the use of pQCD in
this case. Therefore, we can make a guess that the parameter $\mu$ that has been
discussed in the introduction is equal to $\mu^2 \,=\,\frac{1}{r^2_{0 \perp}}$.
However, we have to be very carefull because the $p_{\perp}$ - distribution
has a tail at large values of $p_{\perp}$. We will show that we can safely use
\eq{RDEP} to study the LPM effect.
The value of the mean free path $\lambda$ can also be estimated from
\eq{INCQUARK} and it is equal $\lambda \,=\,\frac{1}{\sigma(r^2_{0 \perp}) \,\rho}$.
This is clear if we rewrite \eq{INCQUARK} as
\begin{equation} \label{AVER}
\frac{d \sigma ( \Delta z)}{ d^2 p_{\perp} \,d z} \,\,=\,\,
\int \,\frac{d r^2_{\perp}}{ 2 }\,\sigma( r^2_{\perp})\,\rho\,
J_0 (\, p_{\perp}\,r_{\perp}\,)\,\exp\( -\,\sigma(\, r^2_{\perp}\, )\,
\rho\,\Delta z\, \)
\end{equation}
where\,\,\, $ \Delta z$\,\,
{}~~ is ~~the longitudinal~~ distance ~~that ~~ the quark~~ passes ~in ~a ~
nuclear
{}~~ matter (\,$ \sigma(\, r^2_{\perp}\, )\,
\rho\,\Delta z\,$~~~$ \,\,>\,\,1$\,).
Actually, \eq{AVER} sets the averaging procedure for \eq{ENSP}.
Indeed
\begin{equation} \label{AVER1}
\VEV{\,\,}\,\,=\,\,
\prod^{n + 2}_{ l = 1} \,\frac{d^2 U_{l \perp}}{\pi}\,\prod^{n+2}_{l = 1}
\, d z_l\,\,\frac{d \sigma (z_l - z_1)}{ d^2 p_{l\,\perp} \,d z}
\,\,;\,\,\,
z_1\,<\,z_2\,...\,<\,z_{l - 1}\,<\,z_{l}\,<\,...<\,z_{n+2}\,
<\,z_1 \,+\,L
\end{equation}
Using the explicit expression for $\vec{J}_{i}$ of eq.(1), one can see, after
integration over the azimuthal angle,
that only big transfer momenta $q_{i \perp}
\,\,\gg\,\,\frac{k_{\perp}}{x_F}$ contribute to \eq{ENSP}. It means that
$ p_{l \perp} \,\,\rightarrow\,\,U_l$. This fact allows us to make the integration over
$U_l$ in \eq{AVER1} explicitly which leads to
\begin{equation} \label{UINT}
\VEV{ \vert \frac{d I_l}{d z_l}\vert}\,\,\equiv\,\,
\int^{\infty}_{U^2_0}\,\,d\, U^2_{l \perp}\,\,
\frac{d \sigma (z_l - z_1)}{ d^2 U_{l}\,\,d \,z_l}
\,\,\exp\left\{ i \frac{1}{\tau} U^2_l\,(
\,z_{l +1}-z_l\,)\,\right\}
\end{equation}
Introducing a new variable $\tilde z_l \,\,=\,\, \,\rho\,z_l\,\,{\bf D}\,\,=\,\,
\rho\,\,z_l\,\,
\frac{\alpha_{\rm S} \pi^2}{3}
x \,G_N ( \frac{4}{r^2_{0 \perp}}, x)$ one can take the integral over $U_l$
and obtain:
\begin{equation} \label{AVERU}
\VEV{ \vert \frac{d I_l}{d z_l}\vert}\,\,=\,\,{\bf D}\,\rho\,\frac{d}{d \tilde
z_l}
\,\{\,\frac{1}{i \,\kappa \,\tilde z_l\,\Delta \tilde z_l\,-\,\frac{1}{4}}\,\cdot\,
\exp[\,-\,\frac{U^2_0}{4 \tilde z_l}\,\,+\,\,i\,\kappa\,U^2_0\,\Delta \tilde z_l\,]
\,\}\,=
\end{equation}
$$
\,=\,\,\{\,16\,i\,\kappa\,\Delta \,\tilde z_l\,\,-\,\,\frac{U^2_0}{{\tilde
z_l}^2}\,\}
\,\cdot\,\exp[\,-\,\frac{U^2_0}{4 \tilde z_l}\,\,+\,\,
i\,\kappa\,U^2_0\,\Delta \tilde z_l\,]
$$
where $\kappa \,=\,\frac{1}{\tau\,\cdot\,{\bf D}\,\rho}$ and $\Delta \tilde z_l
\,=\, \tilde z_{l + 1}\,-\,\tilde z_l$. We anticipate that $\kappa \tilde z_l
\Delta \tilde z_l\,\,\ll\,\,1$ and have expanded the answer with respect to this
parameter.
The above equation allows us to obtain the functional equation for
\begin{equation} \label{DEFP}
\Phi_n (U^2_0, z )\,\,=\,\,\prod^{n}\,\,\VEV{ \vert \frac{d I_l}{d z_l}\vert}
\end{equation}
Namely,
\begin{equation} \label{FUNCEQ}
\Phi_{n - 1} (U^2_0, z - \Delta z)\,\{ \,16 i \kappa \Delta z \,\,-\,\,
\frac{U^2_0}{ z^2_l}\,\}\,\cdot\,\exp\,
[\,-\,\frac{U^2_0}{4 z}\,\,+\,\,i \,\kappa
\,\Delta z\,]\,\,=\,\,\Phi_{n} (U^2_0, z)
\end{equation}
Considering $\Delta z \,\ll\,z$, we can solve \eq{FUNCEQ} and obtain
\begin{equation} \label{SOLPHI}
\Phi_{n}(\kappa,U^2_0, z)\,\,=\,\,
\left( -\frac{U^2_0}{ {\tilde z}^2}\right)^{n + 1}
\,\exp\,[\,-\,\frac{U^2_0}{4 \tilde z}\,\,(\,n\,+\,1\,)
\,\,+\,\,i\,\kappa U^2_0\,\tilde z\,\,+
\,\,\frac{16 i \kappa\,{\tilde z}^3}{3 \,U^2_0}\,]
\end{equation}
Substituting \eq{SOLPHI} in \eq{ENSP}, summing over $n$ and substructing the
value of the integral at $\kappa = 0$, which corresponds to 1 in \eq{ENSP},
we obtain the answer:
\begin{equation} \label{ANSWER1}
\omega\,\frac{d I}{ d \omega d z}\,\,=\,\,
\frac{\alpha_{em}}{\pi}\,4\,{\bf D}\,\rho\,
\int \,\frac{d U^2_0}{U^2_0}\,\int d \tilde z
\,\,\frac{sin^2\{\frac{\kappa}{2}\,[\,\frac{16 {\tilde z}^3}{3\,U^2_0}\,\,+\,\,
U^2_0\,\tilde z\,]\,\}\,( \,- \frac{U^2_0}{{\tilde z}^2}\,)^2\,
e^{-\,\frac{U^2_0}{2 \tilde z}}}
{1\,\,+\,\,\frac{U^2_0}{{\tilde z}^2}\,e^{-\,\frac{U^2_0}{4 \tilde z}}}
\end{equation}
Introducing a new variable $\xi \,=\,\frac{U^2_0}{\tilde z}$ one can rewrite
\eq{ANSWER1} in the form:
\begin{equation} \label{ANSWER2}
\omega\,\frac{d I}{ d \omega d z}\,\,=\,\,
\frac{\alpha_{em}}{\pi}\,4\,{\bf D}\,\rho\,
\int \,\,\frac{d U^2_0}{U^2_0} \,\,\int^{\infty}_{\xi_0} \,\,\xi^2\,\,d \,\xi
\,\,\frac{sin^2(\frac{\kappa\,U^2_0}{2}\,[\frac{16}{3 \xi^3}\,+\,\frac{1}{\xi}]
\,e^{\,-\,\frac{\xi}{2}}}{U^2_0\,\,+\,\,\xi^2\,e^{\,-\,\frac{\xi}{4}}}
\end{equation}
The lower limit in the
$\xi$ integral is $\xi_0\,\,=\,\,\frac{U^2_0}{\rho\,
{\bf D}\,L}$. For $\kappa \,U^4_0\,\,\gg\,\,1$ $\xi \,\approx\,1$ contributes
to the integral and we recover the BH limit. For $\kappa\,U^4_0\,\,\ll\,\,1$
the main contribution comes from the region of small $\xi$, but $\xi\, >\,
\xi_{min}\,=\,( \frac{8 \kappa U^4_0}{3} )^{\frac{1}{3}}$. In this region
the argument of the $sin$ in \eq{ANSWER2} turns out to be small. Expanding
$ sin$ and doing all integrations we get
\begin{equation} \label{FINANS}
\omega\,\frac{d I}{ d \omega d z}\,\,=\,\,
\frac{\alpha_{em}}{\pi}\,4\,{\bf D}\,\rho\,\frac{8}{9}\,\,\sqrt{\kappa}\,\,
=\,\,\frac{32\alpha_{em}}{9\,\pi}\,\,\sqrt{\frac{\omega \,{\bf D}\,\rho}{2\,\,E^2}}
\end{equation}
To obtain the final answer we need to specify the argument of the gluon
structure function in ${\bf D}$. It turns out that the value of the typical
$r^2_{0 \perp}\,\,\propto\,\sqrt{\kappa}$ and it is small for small $\kappa$
for which we have
derived \eq{FINANS}. Therefore we are justified in our approach and the
final answer is
\begin{equation} \label{FINANS1}
\omega\,\frac{d I}{ d \omega d z}\,\,=\,\,
\frac{32\alpha_{em}}{9\,\pi}\,\,\sqrt{\frac{\omega \,\frac{\alpha_{\rm S} \pi^2}{3}
\,\,( x\,G(x, r^2_{soft} \frac{1}{\sqrt{\kappa}})\,\rho}{2\,\,E^2}}
\end{equation}
The region of applicability of \eq{FINANS1} can be found from the condition
that the typical distances in our calculations are small so as
to use perturbative
calculation. In other words, it can be found
from the integral of \eq{ANSWER2} in the region
$\xi\,\approx \,1, \kappa\,U^4_0 \,\gg\,1$. It gives
\begin{equation} \label{BHQCD}
\omega_{BH}\,\,=\,\,\sigma(r^2_{soft})\,\rho\,E^2
\end{equation}
where $r^2_{soft}$ is the scale from where we can use perturbative QCD
($r^2_{\perp} \,<\,r^2_{soft}$), and which
should be calculated in a nonperturbative QCD approach.
The factorization limit comes from the contstraint that $\xi_{min}
\,\leq\,\xi_0$
and gives
\begin{equation} \label{OFACT}
\omega_{fact}\,=\,\frac{\VEV{p^2_{\perp}}}{\rho \sigma(\frac{1}{\VEV{p^2_{\perp}}})}
\end{equation}
Summarizing this section we want to mention that we did not obtain the
$\\sqrt{kappa}\,
ln \kappa$ behaviour of the radiation density in a nuclear matter
as it was done for ``hot"
deconfined plasma in ref.\cite{BDPS}. The difference comes from the averaging
over momentum transfers $\vec{q}_{l \perp}$ which turns out to be quite
different from the screened Coulomb potential applied in ref.\cite{BDPS} for
a ``hot" plasma.
\section{The LPM effect and Mueller technique.}
In this section we are going to discuss the interrelation between our approach
and the Mueller's technique \cite{MUT},
which, together with the AGK cutting rules \cite{AGK},
remarkably simplify the calculation of the inclusive spectra of produced
particles. It is widely used in all
Reggeon inspired calculations. The Mueller diagrams for
the inclusive spectra of emitted photons are shown in Fig.2.
There are two main assumptions that make the Mueller technique simple and
attractive: (i) the spectrum of a produced particle does not depend on
the kinematics of the incoming one; and, (ii) the longitudital part of the
momentum transfer turns out to be so small at high energy that it
can be neglected.
The first assumption holds for the emission of sufficiently soft particles
and it corresponds to the factorization properties of the production of
particles from sufficiently small distances in QCD \cite{FACTOR}. At first
sight
one can prove that the diagrams of Fig.2b do not contribute to the inclusive
cross section using the AGK cutting rules and factorization. However, we are
dealing with the emission from the fastest particle and factorization does not
work in this case. As a result the sum of the diagrams of Fig.2 gives the
factorization limit, or in other words the second term in \eq{INCSP}. We plan
to discuss this term in a separate paper where we shall consider the
inclusive cross section for the emitted photon or gluon. However, even
the sum of all Mueller diagrams cannot reproduce the LPM effect.
The LPM effect comes from a more carefull consideration of the dependance of
the
production amplitude on the longitudinal part of the momentum transfer
$q_l$. From Fig.3 we can see that the
interaction with the centre $`l'$ in the first term of \eq{INCSP} has the
longitudinal component of the momentum transfer $q_z$ ($z$ is the direction
of the incoming quark) which is equal:
\begin{equation} \label{LONGQ}
q_z\,\,=\,\,\frac{1}{ 2 E}\,\cdot\,\{\,(\,p\,+\,k\,)^2\,\,-
\,\,(\,p'\,+\,k\,)^2\,\}\,\,=\,\,\frac{1}{2 E}\,\cdot\,\{\,2\,k_{\mu}
\,q_{l \mu}\,\}\,\,=\,\,\frac{\omega}{2 E^2}\,\{\,U^2_{l + 1} \,-\,U^2_l\,\}
\end{equation}
where we have
neglected the change in the energy of the fast quark due to
the photon emission, as well as due to rescattering.
On the other hand, the longitudinal
part of the momentum $q'_l$ is small. At high energy we are doing all
calculations for quark - nucleon interactions in the leading log(1/x)
approximation ( see for example ref.\cite{GLR} ), in which we neglect
the longitudinal momentum dependance for all gluon and quark propagators.
The only dependance on the longitudinal momentum comes from
the fact that the total
momentum transfer for the scattering off the
$l$-th centre ($Q_l \,= q_l + q'_l$)
differs from zero and it is equal $Q_{z\,l}\,\,=\,\,q_{z\,l} $. Considering
nucleons, in a nuclear matter, as nonrelativistic particles, we have neglected
the change of the energy component of $Q_l$. Finally, we have for each
rescattering
\begin{equation} \label{PSIFUNC}
\Psi^{*}_{initial}( l ) \,\cdot\,\Psi_{final} (l )
\,\,=\,\,\rho\,\,e^{i\,Q_{zl}\,z_l}\,\,=\,\,\rho\,e^{i \frac{1}{\tau}\,
(\, U^2_{l + 1}\,-\,U^2_l\,)}\,\,,
\end{equation}
where $\Psi(l)$ is the wave function of the $l$-th nucleon in a nuclear matter.
One recognizes the phase that we have taken into account in \eq{ENSP}.
Therefore, the LPM effect can be derived from Mueller diagrams if one takes
into account the dependance on the longitudinal part of the momentum transfer.
We will show that the Mueller diagrams
will be very useful to reach simple and transparent understanding of the
LPM - like effect in the case of gluon emission.
\section{The LPM effect for gluon emission.}
In this section we consider the gluon radiation in a nuclear matter using the
Mueller diagrams of Fig.3. To write down the diagrams of Fig.3 we have to
specify the expression for $\vec{J}_{i}$ for gluon emission and calculate
the longitudinal part of the momentum transfer (see \eq{LONGQ} ).
The first
ingredient in the soft limit (\, $\omega\,\,\ll\,\,E$ \,)
was calculated many years ago by Lipatov and collaborators
\cite{BFKL} and has been confirmed using quite different technique
(see ref. \cite{GW} in which such a calculation was done just for the case of
induced gluon radiation in a medium. It is very relelevant to our
approach). The answer is (see Fig.4):
\begin{equation} \label{JQCDGLUON}
\vec{J}_{i}\,\,=\,\,\frac{\vec{k}_{\perp}}{k^2_{\perp}}\,\,-\,\,
\frac{ \vec{k}_{\perp}\,\,-\,\,\vec{q}_{l \,\perp}}{ (\,\vec{k}_{\perp}\,\,-
\,\,\vec{q}_{l \,\perp}\,)^2}
\end{equation}
with the colour factor which is equal to the colour factor of the Feynman
diagram of Fig.4(2). To simplify the calculation of the colour factor we
perform them for QCD with a large
number of colours $N_c$ neglecting all terms
of the order of $1/N_c$.
For a gluon we have two rescatterings which are shown in Fig.4b and Fig.4c.
The value of $q_{l z}$, for the quark rescattering, has been calculated (
see \eq{LONGQ} ) and it is small in the soft region where $E\,\gg\,\omega$.
For the gluon rescattering $q_{lz}$ is equal to:
\begin{equation} \label{LONGG}
q_z\,\,=\,\,\frac{1}{ 2 E}\,\cdot\,\{\,(\,p\,+\,k\,)^2\,\,-
\,\,(\,p\,+\,k'\,)^2\,\}\,\,=\,\,\frac{1}{2 E}\,\cdot\,\{\,2\,p_{\mu}
\,q_{l \mu}\,\}\,\,=\,\,\frac{\omega}{2 \omega}\,\{\,U^2_{G,l + 1} \,-\,U^2_{G,l}\,\}
\end{equation}
where
\begin{equation} \label{UGDEF}
\vec{U}_{G,l}\,\,=\,\,\vec{k}_{\perp}
\,\,-\,\,\sum^{l}_{i = 1} \,\vec{q}_{i\,\perp}\,\,;\,\,\,\,\,\vec{U}_{G,0}\,
\,=\,\,\vec{k}_{\perp}\,\,.
\end{equation}
Comparing \eq{LONGQ} and \eq{LONGG} one can see that $q_{lx}$ due to quark
rescattering is much smaller than $q_{l z}$ for a gluon collision. Therefore,
we can neglect the quark rescattering in the first approximation to the
problem.
Therefore, we can write an equivalent
expression to the one in \eq{ENSP} for gluon radiation
density, using the following substitutes:
\begin{equation} \label{SUBS}
\vec{U}_{l}\,\,\,\rightarrow\,\,\vec{U}_{G,l}\,\,\,;\,\,\,\kappa\,\,\,\rightarrow\,\,\,
\kappa_G\,\,\,=\,\,\,\frac{1}{2 \,\omega\,{\bf D}\,\rho}\,\,.
\end{equation}
We have to make two comments:
1 . Inspite of the fact that we are dealing with
gluon rescattering in the averanging procedure over the transverse
momentum of the produced gluon $k_{\perp}$ given by \eq{AVER1}, we
should use the same quark cross section given by
\eq{QUARKCRSEC}. Indeed, the gluon -
quark pair scatters in the medium and the transverse separation between them
due to many rescatterings off nucleons in the limit
$q_{lz}\,\,\rightarrow\,\,0$, as we
will show below, is of the order $r^2_{\perp}\,\,\propto \,\,\sqrt{\kappa_G}$.
The average momentum transfer in a single collision $\VEV{q_{l \perp}}$ is
of the order of $\VEV{q^2_{l \perp}}\,\,\approx\,\,
\frac{1}{r^2_{\perp}\,N_{col}}\,\,\ll\,\,\frac{1}{r^2_{\perp}}$, where
$N_{col}$
is the number of collisions which is big enough
$\propto \,\,\frac{1}{\sqrt{\kappa_G}}$. Therefore, each gluon in the nucleon
cross section carries a transverse momentum $q_{l \perp}$ which is much
smaller than $\frac{1}{ r_{\perp}}$. It means that such a gluon interacts with
the total colour charge of the gluon - quark pair, which is equal to the
charge
of the quark at $N_c\,\gg\,1$. The typical time that a gluon lives is
$\tau_G \,=\,\frac{\omega}{U^2_0}$ which is much smaller than the life-time of
the quark $\tau_Q\,=\,\frac{E}{\VEV{p^2_{\perp}}}$. It means that for each
emitted gluon the quark plays the role of a spectator that neutralizes
half of the gluon colour charge.
2. It was shown in ref.\cite{GW}
( see also ref. \cite{BDPS}) that quark rescattering contributes to the
radiation density of gluons, if one takes into account the
$1/N_c$ corrections.
The origin of this contribution is the dynamic gluon correlations that have
been studied in ref. \cite{HT}. Such correlations change the Glauber-type
formula of \eq{INCQUARK} that was used for the
averaging in the nuclear matter.
The final answer is \eq{ANSWER2} with the substitutions
defined in \eq{SUBS}. The value of
$\xi_0$ is equal to $\xi_0 \,=\,\frac{U^2_0}{\rho \,{\bf D} \,L}\,=\,
2 \kappa_G\,U^2_0$. For $\kappa_G U^4_0\,\,\ll\,\,1$, we derive the answer:
\begin{equation} \label{ANSWERG}
\omega\,\frac{d I_G}{ d \omega d z}\,\,=\,\,
\frac{ N_C \alpha_{S}}{2 \pi}\,4\,{\bf D}\,\rho\,\frac{8}{9}\,\,\sqrt{\kappa}
\,\,
=\,\,\frac{16 N_c \alpha_{S}}{9\,\pi}\,\,\sqrt{\frac{ \,\frac{\alpha_{\rm S} \pi^2}{3}
\,\,( x\,G(x, \frac{1}{r^2_{soft}\,\,\sqrt{\kappa}})\,\rho}{2\,\,\omega}}
\end{equation}
$r^2_{soft}$ in \eq{ANSWERG} is a scale of the ``soft" interaction.
We can trust a
perturbative calculation for the nucleon gluon distribution only for
$r_{\perp}
\,<\,r_{soft}$.
The value of $1/x$ in \eq{ANSWERG} is equal to
$\frac{\tau_G}{\tau_{soft}}$ where $\tau_G\,=\,\frac{\omega}{U^2_0}\,\,=\,\,
\sqrt{\kappa_G} \,\omega$ and $\tau_{soft}$ is a typical time for the ``soft"
processes which we cannot specify in the leading log (1/x) approximation
(LL(1/x)A)
which we have used to obtain the answer. To trust the
LL(1/x)A we have to assume
that thr emitted gluon energy
is so big that $\alpha_{\rm S} ln \frac{\tau_{soft}}{\tau_G}
\,\,\geq\,\,1 $. It means that $\sqrt{\kappa_G}\,\omega \,\,\gg\,\,\tau_{soft}$.
The value of $\omega_{BH}$ can be obtained from the equation $\sqrt{\kappa }
\,\,=\,\,
r^2_{soft} $ which gives
\begin{equation} \label{OBHG}
\omega^{G}_{BH}\,\,= \,\,\frac{1}{r^2_{soft}\,\,\sigma (r^2_{soft})\, \,\rho}
\end{equation}
It should be stressed that unlike the QED case we cannot find the
factorization limit for induced gluon emission. The physical reason for this
is
obvious since the phase of the propagating gluon cannot be small due to
its rescattering in a nuclear matter, at least in the kinematic region where
$\kappa_G$ is small. We would like to recall that we cannot trust our formulae
for big valus of $\kappa_G$ as has been discussed above.
The radiation density of \eq{ANSWERG} is very close to the result of ref.
\cite{BDPS} and quite different from other attemps to estimate the LPM effect
for gluon radiation \cite{GW} \cite{RY}. The main difference between
\eq{ANSWERG} and the result of ref. \cite{BDPS} is the fact that the gluon
nucleon density
enters our answer while the radiation density of ref. \cite{BDPS}
is proportional to $\frac{1}{\sqrt{\omega}} \,\ln \omega$. Since HERA data \cite{HERA}
shows sufficiently steep energy behaviour of the gluon structure function
\cite{HERA} which could be parametrized as $\frac{1}{x^{\omega_0}}$ with the
value of $\omega_0 \,\,\sim \,0.3 - 0.4$, the $\omega$ behaviour of
the gluon radiation density can be evaluated as
$\omega\,\frac{d I_G}{ d \omega d z}\,\,\propto\,\,\omega^{-\,
\frac{1}{2}\,+\,\frac{\omega_0}{4}}
$. The energy losses of the fast quark due to gluon emission can be estimated
integrating \eq{ANSWERG} over $\omega$ up to $E$ :
\begin{equation} \label{ELOSS}
-\,\frac{d E}{ d z}\,\,\propto\,\frac{N_c \alpha_{\rm S}}{ 2 \pi}\,\cdot\,\frac{1}{
\frac{1}{2} \,+\,\frac{\omega_0}{4}} \,\cdot\,E^{\frac{1}{2} \,+\,\frac{\omega_0}{4}}\,
\,.
\end{equation}
However, we shall be very careful with such kind of estimates since the
value of $\omega_0$ depends crussially on the value of the gluon virtuality which
depends on $\omega$ in its turn (see \eq{ANSWERG}). The integral over $\omega$
in $\frac{d E}{d z}$ concentrates at $\omega \,\approx\,\,E$ where the soft energy
approximation is not valid. Because of this we can consider \eq{ELOSS}
only as a rough estimate which we are going to improve later.
\section{Conclusions}
In this letter we have considered the LPM effect for photon and gluon
radiation off a fast quark propagating in a nuclear matter. The close
analogy betwen photon and gluon emission suggested in ref. \cite{BDPS}
has been confirmed and the relation between the Mueller approach \cite{MUT}
and traditional calculations has been established.
The main result reads:
\begin{equation} \label{FINANS}
\omega\,\frac{d I}{ d \omega d z}\,\,=\,\,
\,\,\frac{32 \alpha}{9\,\pi}\,\,\sqrt{\frac{ \,\frac{\alpha_{\rm S} \pi^2}{3}
\,\,( x\,G(x, \frac{1}{r^2_{soft}\,\,\sqrt{\kappa}})\,\rho}{\tau}}
\end{equation}
where $\kappa$ is defined in \eq{AVERU} and
for QED:
$\alpha = \alpha_{em}\,; \frac{1}{\tau}\,=\,\frac{\omega}{2 E^2}\,;
x\,\,=\,\,\frac{\VEV{p^2_{\perp}}}{m E}$.
For QCD:
$\alpha = \frac{N_c \alpha_{S}}{2}\,; \frac{1}{\tau}\,=\,\frac{1}{2\omega }
\,;
x\,\,=\,\,\frac{r^2_{soft}}{\omega \sqrt{\kappa}}$.
The answer does not depend on the nonperturbative QCD
scale. It depends on the ratio
$\frac{\mu^2}{\lambda} \,=\,\rho\,\,\cdot\,\,\{ \,\frac{\alpha_{\rm S} \pi^2}{3}
\,\,( x\,G(x, \frac{1}{r^2_{soft}\,\,\sqrt{\kappa}}\,\}$ which
depends on the QCD
scale only in the argument of the gluon structure function. However, the value
of $\omega$ for which we can apply the above formula comes from $\kappa\,\ll\,
r^2_{soft}$ and crucially depends on the value of $r^2_{soft}$ ( see \eq{OBHG})
{}.
$r_{soft}$ establishes the scale of distances
( $r_{\perp} \,\,<\,\,r_{soft}$ )
where we can trust the perturbative QCD approach for nucleon interactions.
This scale has clear nonperturbative origin, however, it becames large and
grows with $x$ in the small $x$ region where the saturation of gluon density
in a nucleon should be reached \cite{GLR}. Taking $r^2_{soft}\,\,
=\,\, 0.5 GeV^{-2}$
and $\rho \,=\,0.17 Fm^{-3}$ we derive that our formula
for gluon induced radiation can be justified for $ \omega\,>\,\omega_{BH} = 100 GeV$.
We are going to present more reliable estimates, as well as the application
to the gluon inclusive cross section, in further publications.
|
\section{Introduction}
\label{sec:intro}
Recently, much interest is focused on systems whose classical
(mean-field) ground states exhibit an infinite number of local continuous
degeneracies due to frustration(``floppy" systems).
In such a system, construction of the linear spin-wave theory based on
one of the classical ground states leads to at least one spin-wave
mode whose frequency vanishes for all wave vectors(zero-energy mode).
The zero-energy mode corresponds to local deformation of a spin configuration
that does not raise the energy.
The set of the ground state configurations is a manifold with dimensions
proportional to the system size.
The classical ground state thus may be considered as disordered. %
We can find examples of floppy spin systems in all dimensions.\cite{ramirez94}
It is an interesting and a challenging problem to investigate how the quantum
effects manifest themselves in such spin systems.
The central issue is whether quantum effects lift
the ground state degeneracy and select some long-range order.
Also the low-energy excitation spectrum is of interest,
since it may lead to peculiar thermodynamic properties.
A typical example of such systems is the antiferromagnetic Heisenberg
(AFH) model on the {\it kagom\'e} lattice.
Intensive studies
\cite{elser89,zeng-e90,chalker-ws92,keren94,reimers-b93,harris-kb92,%
chubukov92,sachdev92,leung-e93,elstner-y94,zeng-e95,nakamura-m95}
on this system have been inspired by the experiments on the $^3$He layer
adsorbed on graphite \cite{greywall89,franco86}
and also on the compound SrCr$_8$Ga$_4$O$_{19}$.\cite{fioriani-dts85}
What mainly have been concerned with are
the existence of a double peak in the specific heat at low temperatures
and
whether a magnetic order is realized in the ground state or not.
Several approximate analyses \cite{harris-kb92,chubukov92,sachdev92}
have been done but they are still far from giving a common understanding
on the ground state property.
Numerical studies on finite systems support the existence of the double
peak.\cite{elser89,zeng-e90,zeng-e95,nakamura-m95}
In this paper we study a simple one-dimensional floppy
system called the $\Delta$-chain.\cite{hamada-knn88,doucot-k89}
We consider that this model shares general features of quantum floppy
systems, and an understanding of its properties might give some insight into
that of the {\it kagom\'e} antiferromagnet.
The ground state of this system is exactly known to be a dimer
state.\cite{monti-s91,monti-s92}
One of the authors(K. K.) examined the low-lying
excitations and the specific heat in a previous paper.\cite{kubok93}
The numerical diagonalization study of small clusters (up to 20 spins)
exhibited that the low-lying excitation modes in the periodic
chains are almost dispersionless.
The first and the second lowest mode were revealed to converge
to dispersionless modes with the same energy in the
thermodynamic limit.
The specific heat was shown to have a double peak in common with
the {\it kagom\'e} antiferromagnet.
This double peak structure was also observed
recently by a Monte Carlo method \cite{nakamura-s95}
and by a recursion method.\cite{otsuka95}
The dispersionless aspect of excitations may be considered to imply immediately
their localization and hence very weak size-dependence of various quantities.
The energy gap obtained by numerical diagonalizations,\cite{kubok93}
however, does exhibit fairly large size-dependence.
Also a broad bump of the spin correlation was observed between the most distant
spin pairs in the lowest triplet excitations.\cite{kubokunp}
These results suggest that the excited states are not localized.
We solve this puzzle and clarify the character of
low-lying excited states in the following sections by mainly employing
numerical diagonalization of finite size systems.
Elementary excitations are revealed to be
`kink'-`antikink' type domain walls created in the singlet dimer ground state.
An antikink is shown to move freely in a region bounded by kinks at both ends.
The kinetic energy of an antikink leads to the size dependence
of the energy gap.
On the other hand, a kink is localized and gives the dispersionless
property to a kink-antikink pair excitation mode.
Properties of an isolated kink and an antikink, and also interactions between
them will be discussed in Sec. \ref{sec:elem}.
In Sec. \ref{sec:num}, we show that the size dependence of the
energy gap in the periodic chain agrees with that of the
kinetic energy of an antikink.
The specific heat due to kink-antikink excitations is discussed in
\ref{sec:spec},
and the position of the low-temperature peak in the thermodynamic limit is
estimated.
Conclusion is given in Sec. \ref{sec:conc}.
\section{Elementary excitations}
\label{sec:elem}
\subsection{Introduction of a kink and an antikink}
The $\Delta$-chain is described by the following Hamiltonian.
\begin{equation}
H=\sum_{i=1}^N h_i,
\label{eq:hamil}
\end {equation}
where
\begin{equation}
h_i=\mbox{\boldmath $S$}_{2i-1}\cdot\mbox{\boldmath $S$}_{2i }
+\mbox{\boldmath $S$}_{2i }\cdot\mbox{\boldmath $S$}_{2i+1}
+\mbox{\boldmath $S$}_{2i-1}\cdot\mbox{\boldmath $S$}_{2i+1}.
\label{eq:hi}
\end {equation}
$\mbox{\boldmath $S$}_i$ is the spin with size $1/2$ at the site $i$
and $N$ denotes the number of triangles in the chain.
We first explain the ground state of this system and then forward to the
explanation of elementary excitations.
Under periodic boundary conditions, the ground state is the perfect singlet
dimer state, since the ground state of the local Hamiltonian $h_i$ is
realized by pairing any two of three spins into the singlet state.
The two-fold degenerate ground states
are written as
\begin{eqnarray}
\psi_{\rm g}^{\rm L}&=&[1, 2]\otimes[3, 4]\otimes[5, 6]
\otimes\cdots\otimes[2N-1, 2N]\nonumber \\
\psi_{\rm g}^{\rm R}&=&[2, 3]\otimes[4, 5]\otimes[6, 7]
\otimes\cdots\otimes[2N, 1].
\label{eq:psig}
\end {eqnarray}
Here $[i, j]$ denotes the singlet dimer of the spins at the site $i$ and $j$,
i.e., $[i, j]=(\alpha_{i}\beta_{j}-\beta_{i}\alpha_{j})/\sqrt{2}$,
where $\alpha_{i}$($\beta_{i}$) is the state with $S_{i}^{z}=1/2(-1/2)$.
They are schematically depicted in Fig. \ref{fig:scheme} (a) and (b).
We call a dimer located on the left(right) side of a triangle an L-dimer
(R-dimer) and
hence the state $\psi_{\rm g}^{\rm L}$($\psi_{\rm g}^{\rm R}$) an L-dimer
(R-dimer) state.
These two states are linearly independent but not
orthogonal to each other for finite $N$.
They become orthogonal in the limit where $N\rightarrow \infty$.
The existence of the excitation energy gap above the ground state was
rigorously shown.\cite{monti-s91,monti-s92}
Under open boundary conditions, the ground state of a system with $N$
triangles is highly degenerate
since a ground state consists of $N$ singlet dimers and one free spin.
One of the ground states is shown in Fig. \ref{fig:scheme}(c).
A dimer configuration is uniquely determined by fixing the position of
the free spin.
It is seen in the figure that the free spin plays a role of a domain wall
between an L-dimer state and an R-dimer state.
The number of possible positions for a free spin in the ground state
is equivalent to the number of sites, and therefore
$2N+1$ different dimer configurations can be considered.
However, those $2N+1$ states are linearly dependent, since only two out of
three dimer configurations in a triangle are linearly independent.
The dimensions of the ground state reduce to
$2\times (N+1)$;\cite{monti-s91,monti-s92}
only one positional freedom per triangle is allowed.
Let us consider a simple trial excited state in a periodic chain
constructed by changing one dimer singlet
in the ground state into a triplet state.
Then the expectation value of the excitation energy $\Delta E$ is identical
to the singlet-triplet gap; $\Delta E=1$.
Since the energy gap is known as $\Delta E\sim 0.22$
in the thermodynamic limit,\cite{kubok93}
the lowest excited state is not such a simple state.
In fact, the above state is not an eigenstate and the two spins coupled to a
triplet are separated when the Hamiltonian is operated.
The average excitation energy may decrease if the two spins are
apart
as depicted in Fig. \ref{fig:scheme} (d).
Each of two spins is regarded as a domain wall
between an L-dimer and an R-dimer state.
We show in the following that isolated domain walls are the elementary
excitations in this system.
A domain wall which has L-dimers on the left side and R-dimers on the right
has a singlet dimer in its own triangle, and thus
is the ground state of the local Hamiltonian $h_i$.
We call this type of a domain wall a `kink'.
A free spin appearing in the ground state of the open $\Delta$-chain is
a domain wall of this type.
Therefore the excitation energy of a kink is null.
There is ambiguity in the definition of the position of a kink,
since ground states with different positions of the free spin
are not orthogonal to each other.
Let us consider a kink as a state where a free spin is located at the top of a
triangle as depicted in Fig. \ref{fig:scheme} (c) for a convention.
Then the overlap between the states with an up(or a down)-spin at the $i$th
and the $j$th($i\ne j$) triangle is $-(-2)^{-|i-j|}$.
We treat a kink as a localized object since the state with a kink alone
is an eigenfunction of the system,
although its position is not a good quantum
number due to the non-orthogonality.
We call another type of a domain wall an `antikink',
which has R-dimers on the left side and L-dimers on the right.
The state with an antikink at the $i$th triangle is not a ground state of
$h_i$ as the triangle can not accommodate a singlet dimer.
A finite energy is necessary to excite an antikink.
It spreads out to other triangles and propagates among them.
The energy of an antikink is lowered by this motion,
which will be discussed in the next subsection.
Both a kink and an antikink have a spin $1/2$ as a whole, and
must appear alternatively.
Elementary excitations in the Majumdar-Ghosh model
is known as propagating domain walls.\cite{majumdar-g69,shastry-s81}
They are also kink-antikink type domain walls.
The physical properties of a kink and an antikink are same in this
model, since they are transformed to each other by symmetry operations.
On the other hand, the properties of a kink and those of an antikink are
quite different in the $\Delta$-chain:
kinks are localized and antikinks move about
in a region bounded by two localized kinks.
In the following,
we study first an isolated antikink and then the interaction between a kink and
an antikink.
\subsection{An isolated antikink}
An antikink is necessarily accompanied by a kink in a periodic chain and it is
not easy to extract the properties of an isolated antikink
because the translational symmetry mixes the states with different positions
of a kink and an antikink.
We can solve this difficulty by considering open systems as shown in
Fig. \ref{fig:openscheme} (a), which we call `Open-A' systems.
In this subsection we treat only Open-A systems.
In this system,
two additional edge bonds have strong tendency to form dimer singlet states
and they force the left(right)-hand side of the system
to be in the R(L)-dimer state.
Then inconsistency of the dimer configurations results in an antikink
to appear in the middle of the chain.
Thus we expect only an antikink to exist
in the ground state of an Open-A system.
We confirm this speculation and clarify the
properties of an isolated antikink through numerical and variational analyses.
The energy of an antikink $\delta E$ is obtained from the total energy $E$
of the system with $N$ triangles as $\delta E = E +3(N+2)/4$.
Figure \ref{fig:open-acorsz} shows the nearest-neighbor spin correlation
$\langle \mbox{\boldmath $S$}_i\cdot \mbox{\boldmath $S$}_{i+1} \rangle$
and the local magnetization $\langle S_{i}^{z} \rangle$
in the lowest four states of an `Open-A' system with $S^{z}=1/2$.
We consider the ground state first.
It is seen that the spin correlations at the both edge bonds are very close to
the value of a dimer singlet state $-3/4$, while those of
the bonds next to the edges almost vanish.
These features show that the both edges are almost in the perfect dimer states.
We notice that the state close to the left(right) edge
is approximately the R(L)-dimer state.
As the position of a bond approaches the center,
the singlet coupling becomes weak and finally the phase changes at the center.
Since the L-dimers are located in the right-hand side of the domain wall,
we may conclude that there is an antikink.
The local magnetization oscillates extended to the whole system,
and its amplitude has a broad maximum in the middle of the system.
The system-size dependence of the domain wall profile in the ground state is
shown in Fig. \ref{fig:open-adom}.
We plotted as the profile the deviation of the nearest-neighbor correlation
from that in the R(L)-dimer state on the left(right) half of the system.
That is,
$0.75+\langle \mbox{\boldmath $S$}_i\cdot \mbox{\boldmath $S$}_{i+1} \rangle$
for odd positive or even negative $i$ and
$0-\langle \mbox{\boldmath $S$}_i\cdot \mbox{\boldmath $S$}_{i+1} \rangle$
for even positive or odd negative $i$.
We notice in this figure
that the profile of the domain wall is extended to almost whole region
of the chain.
Results of the profile and the magnetization in the ground state of Open-A
systems imply that the antikink is in a state
extended to the entire region of the system with large amplitude in the
middle.
In the first excited state, the profile of the magnetization has two bellies
at about a quarter of the system size from both edges and a node
at the center.
If we assume that the local magnetization reflects
the probability of existence of an antikink, the above result suggests that
the wave function of an antikink is something like a sine-function with a node
at the center.
The spin correlation profile is very flat with its value $\sim -0.35$ in the
middle region.
The correlation at a bond is averaged to be $-3/8$ if the probability for
an antikink to be at the left of the bond is equal to that at the right.
So the result is consistent with a wave function of an antikink
which has a node in the center and two bellies.
We can also speculate that the wave functions of
the second and the third excited states have two and three nodes, respectively.
The flat region in the spin correlation
in the third excited state suggests existence of a node at the center.
We can not, however, extract detailed properties of the wave functions
for these excited states from the present numerical data alone.
We see in the following that our speculation is correct by
comparing the diagonalization results with those by a simple trial functions.
Numerical results of Open-A systems suggest that an antikink behaves like a
free particle propagating in the whole system,
since the data are consistent with that the wave function
of an antikink is a sine function in terms of the position.
In order to check the validity of this picture we have made a variational
analysis.
We take the state where an antikink consists of one free spin
as the first trial function.
Then the variational basis is the set of $ \psi_i^{(1)}$
which consists of $(N+1)$ dimer singlet
pairs and the one free spin located at the top of the $i$th triangle
as depicted in Fig. \ref{fig:varbase}(a).
The dimension of the basis is $N+2$ since the free spin may occupy the sites at
the edges.
These basis functions are not orthogonal to each other and satisfy the
following relations.
\begin{eqnarray}
\langle \psi_i^{(1)}|\psi_j^{(1)}\rangle&=&\left (-\frac{1}{2}\right )^{|i-j|}
\\
\langle \psi_i^{(1)}|H|\psi_j^{(1)}\rangle&=&-\frac{3}{4}
[(N+2)\langle \psi_i^{(1)}|\psi_j^{(1)}\rangle-\delta_{ij}]
\end {eqnarray}
Here, $\delta_{ij}$ is the Kronecker delta.
The trial function $\Psi_{\rm var}^{(1)}$, which is assumed as
\begin{equation}
\Psi_{\rm var}^{(1)}\equiv \sum_{i=0}^{N+1} C_i\psi_i^{(1)},
\label{eq:varf}
\end {equation}
is determined as an eigenfunction of the matrix
$\langle \psi_i^{(1)}|\psi_j^{(1)}\rangle$ ($0\leq i,j \leq N+1$).
In the limit of infinite $N$ we can neglect the boundary effect, and
the eigenfunction is easily obtained as
$C_i\propto (-1)^i \sin {qi}$ with the excitation energy
$\delta E_{\rm 1-spin}=5/4-\cos q$.
For a finite $N$, we have solved the eigenvalue problem numerically
and have found that the $k$th state($k= 1, 2, \cdots$) and
its antikink energy is well approximated by
\begin{eqnarray}
C_i&\propto &(-1)^i \sin \frac{\pi k (i+3/2)}{N+4}
\end{eqnarray}
and
\begin{eqnarray}
\label{eq:1spinvar}
\delta E_{\rm 1-spin}&=&\frac{5}{4}-\cos \frac{\pi k}{N+4}.
\end {eqnarray}
The variational energy gap converges to
$\delta E_{\rm 1-spin}= 0.25$ as $N\to\infty$.
The fact that the true energy gap converges to about 0.22 implies
that the above choice
of the variational basis is too simple to describe an antikink.
We have to take into account of the fact that
a free spin spreads out to neighboring triangles
destroying singlet dimers nearby.
As an improved trial function, we have employed $\Psi_{\rm var}^{(5)}$
which is given by Eq. (\ref{eq:varf}) with $\psi_i^{(1)}$ replaced with
$\psi_i^{(5)}$, which consists of $(N-1)$ dimer singlet pairs and
the ground state of the cluster
with five spins as depicted in Fig. \ref{fig:varbase} (b).
We numerically diagonalized the effective Hamiltonian
for $\Psi_{\rm var}^{(5)}$.
The wave functions for the lowest three levels
obtained numerically for the system with 98 triangles
are shown in Fig. \ref{fig:varwave}.
The wave functions ($C_{i}\times (-1)^{i}$) look like sine functions
and is consistent with the numerical
results of the local magnetization shown in Fig. \ref{fig:open-acorsz}.
There is no qualitative difference between the 1-spin variational
wave functions and the 5-spin ones.
The variational energy by $\Psi_{\rm var}^{(5)}$ is lower
than that by $\Psi_{\rm var}^{(1)}$ for low-lying states($k\ll N$) and
leads to the energy gap $\Delta E_{\rm 5-spin}\sim 0.23$
in the limit $N\to\infty$.
The 5-spin variational energy, however, rapidly increases
as $k$ approaches $N$ and exceeds the 1-spin variational energy,
which implies that the 5-spin wave function is not a good approximation
in the high momentum region.
This is because we only took into account the
{\it ground state} of the 5-spin Hamiltonian.
In the high momentum region the 1-spin variational function seems to be a
better approximation than the 5-spin one.
The expectation value of the local magnetization in the 1-spin variational
wave function $\Psi_{\rm var}^{(1)}$ is compared with the exact results
for the lowest three states in Fig. \ref{fig:varsiz}.
Overall behaviors of both results agree very well not only for the lowest
state but for the excited states.
This fact supports that the low energy states of the
system are those with a freely propagating antikink.
The amplitude of the oscillations of the variational results is larger
than the exact ones.
This reflects that the real antikink is not a single spin
as supposed in the 1-spin variational function but
extended to several triangles and thus the magnetization is smeared.
The size dependence of the variational energy is shown
in Fig. \ref{fig:altogether}
and is compared with the diagonalization results.
All the data show almost linear dependence on $N^{-2}$.
It is seen that the slope of the results of the 5-spin trial function
agrees very well with the diagonalization results.
Another important fact shown in this figure is that
the behavior of the eigen energy of the Open-A system
agrees with that of the excitation energy of the periodic system.
A low energy excitation of the periodic system is considered to be
a kink-antikink pair excitation but
the above result implies that the excitation energy
in the periodic chain is mainly determined by a freely moving antikink.
\subsection{kink-antikink interaction}
In this subsection, we investigate the wave function of a kink and
the interaction between a kink and an antikink.
For this purpose we study the excited states of finite clusters as depicted in
Fig. \ref{fig:openscheme} (b), which we call `Open-B' systems.
The L-dimer state is the unique ground state of this Open-B system.
Hence low energy excitations in this system necessarily involve
a kink and an antikink and
the open boundary conditions make each of them visible as shown below.
We first consider the first excited state.
This is a triplet state and we show the nearest-neighbor spin correlation
$\langle \mbox{\boldmath $S$}_i\cdot \mbox{\boldmath $S$}_{i+1}\rangle$ and
the local magnetization $\langle S^z_i\rangle$ in the state with $S^z=1$
in Fig. \ref{fig:open-bedge}(a).
The numerical results show that the leftmost bond forms a complete
triplet state for all system sizes, i.e.,
$\langle \mbox{\boldmath $S$}_1\cdot \mbox{\boldmath $S$}_{2}\rangle =1/4$
(note that $\mbox{\boldmath $S$}_1\cdot \mbox{\boldmath $S$}_{2}$
commutes with the Hamiltonian).
The data for the three leftmost spins approach the limiting values
when $N\to \infty$ as:
$\langle \mbox{\boldmath $S$}_2\cdot \mbox{\boldmath $S$}_{3}\rangle \to -1/2$,
$\langle S_1^z\rangle=\langle S_2^z\rangle\to 1/3$ and
$\langle S_3^z\rangle\to -1/6$.
In fact the difference from the limiting values is almost negligible
in the data of the system with $N=11$ (not shown in the figure).
This result implies that the total wave function reduces to
a direct product of the wave function of these three spins given by the
following equation
and that of the remaining $(2N-1)$ spins in the limit $N\to \infty$.
\begin{equation}
\psi_{\rm kink}=[\alpha_1(\alpha_2\beta_3-\beta_2\alpha_3)
+\alpha_2(\alpha_1\beta_3-\beta_1\alpha_3)
]/\sqrt{6}
\label{eq:kink}
\end {equation}
Therefore a kink is localized in the leftmost triangle
though it is a linear combination of the wave function
with two different positions of a free spin.
An antikink is distributed in the rest part of the system
which corresponds to an Open-A system with $N-2$ triangles.
It also feels a magnetic field at the 4th and 5th site
caused by a magnetization $\langle S_3\rangle\sim -1/6$
of the kink at the third site.
Otherwise the antikink is freely propagating in this region.
The domain wall profile shown in Fig. \ref{fig:open-bedge}(a) is
fairly well reproduced by that of the Open-A system with 6 triangles
and with a magnetic field $-1/6$ at the site 1and 2.
Almost complete reproduction of the profile is made by
adjusting the magnetic field to be $-0.225$,
which implies that the antikink feels a larger effective field
due to the overlap of the wavefunction with that of the kink for finite $N$.
We have examined higher excited states up to the fourth
in the subspace of $S^{z}=1$. (Not shown in the figure.)
We found that a kink is located in the $n$th triangle
in the $n$th excited state and
the left side of that triangle is in the L-dimer state.
Therefore the wave function of the $n$th state is
nothing but a direct product of the wave function of the L-dimer
state of $n-1$ triangles and that of the first excited state of
a smaller Open-B system with $N-(n-1)$ triangles.
Excitations corresponding to excited states of the Open-A systems
can be obtained by restricting the phase space
to that where the edge bond forms a triplet state
(shown in the Fig. \ref{fig:open-bedge}(b),(c)).
This restriction ensures the existence of a kink at the leftmost triangle
and then the domain for an antikink is same with the first excited state.
It is seen that the results at the leftmost triangle
scarcely vary among four excitations in Fig. \ref{fig:open-bedge}.
This implies that the state of a kink is almost same in these three states
and is localized within the leftmost triangle.
The results on the right of the leftmost triangle agree very well with
those in the `Open-A' system shown in Fig. \ref{fig:open-acorsz}.
Only a small difference is seen at the center of the chain where the
domain wall profile is not flat for the Open-B system.
This difference shows that the antikink is pulled to the left
due to the effective exchange interaction with the kink.
Since the leftmost triangle is occupied by a kink,
the region for the motion of an antikink in the
system with $N$ triangles is same with that in a Open-A system
with $N-2$ triangles.
The interaction with a kink attracts the wave function of an antikink,
and as a result, an antikink feels a larger region for its motion.
The wave function in the $S=1$ subspace roughly corresponds
to that in the Open-A system with $N-1$ triangles.
We discuss this problem in detail in the next section.
We have also obtained the singlet excited states in the subspace
where the edge bond forms a triplet state.
The result of the lowest excited state is shown in Fig. \ref{fig:open-bs01}(a)
together with that in the triplet state.
We find that the results at the leftmost triangle is almost same with
those in the corresponding triplet states.
On the other hand, the results on the right side of the system are different
from those in the the triplet states.
That is, the domain wall profile is shifted to the right
by nearly one triangle compared to that in the triplet states.
This is the effect of the exchange
interaction with the kink, which acts as a repulsion in the singlet state.
The profile of the singlet state shown in Fig.\ref{fig:open-bs01}(a)
is fairly well reproduced by that in the ground state
of the Open-A system with 6 triangles under
the magnetic field $1/6$ at the site 1 and 2.
More precise reproduction of the profile can be made by
adjusting the magnetic field to be $0.320$.
The fact that the antikink feels stronger effective field
in the singlet state than in the triplet one seems to imply that
the kink is extended to outside of the leftmost triangle
for this size of the system.
The antikink in the singlet state is repelled stronger
by the kink than that in the triplet state
because the state must be orthogonal to the ground state.
This effect might be the cause of the stronger effective field,
and as a result, it is nearly prohibited for the antikink
to occupy the site 4 in the singlet state.
The wave function of an antikink in the $S=0$ subspace
roughly corresponds to that in the Open-A system with $N-3$ triangles.
In Fig. \ref{fig:open-bs01}(a),
we compare the domain wall profiles of the lowest
triplet and the singlet excitations.
We see that the center of the profile is different nearly by
one triangle between two cases,
which implies that the effective length
of the region where the antikink can move is different by about two triangles.
The profiles of the the ground state of the Open-A system with $N$ triangles
under the magnetic field $-0.225$,
that of the $S=0$ excited state of the Open-B system with $N+3$ triangles,
and the mirror image of the profile of the $S=1$ excited state
of the Open-B system with $N+1$ triangles agree surprisingly well
in the central region of the antikink(Fig \ref{fig:open-bs01}(b)).
We may conclude by this evidence that the wave functions of an antikink
in these three states are approximately equal.
We have seen above that a kink is localized
and an antikink is moving in the region
bounded by the kink and by the open boundary in the low energy excited states
of the Open-B systems.
The interaction between a kink and an
antikink is approximated by a hard-core repulsion on the same triangle and a
ferromagnetic exchange interaction on the nearest neighbor triangles.
The exchange constant is approximated by $-1/3$ in the limit of infinite $N$.
Otherwise an antikink can be considered as a free particle.
A kink is pushed to the left edge by an antikink as to lower
its kinetic energy in Open-B systems.
We have studied the interaction between a kink and an antikink
when only one antikink is on the right side of the kink.
In periodic systems and/or higher excitations,
an antikinks exists also on the left side of a kink.
We expect that a kink will be localized even
when antikinks exist at both sides.
Its wave function, however, would be different from that studied above
which is asymmetric with respect to the inversion.
So we have obtained the lowest excitation with $S^{z}=3/2$
of the Open-A system with 17 spins(7 triangles) to
investigate excitations with a kink and two antikinks.
The system size, however, seems to be too small to study a
detailed nature of the kink-antikink interaction
from the obtained domain wall profile and the local magnetization.
We could not obtain the corresponding state in the system
with 21 spins by the Lanczos method because of a poor convergence of the
iteration.
We conjecture that
a ferromagnetic exchange interaction acts between a kink and an antikink
when they are at the nearest neighbor triangles.
The ferromagnetic nature of the interaction is confirmed
from the result of the periodic systems
where the lowest excited state is triplet.
More study is necessary to determine the magnitude of the exchange coupling.
\section{Size dependence of the energy gap}
\label{sec:num}
In the previous section, we focused on a kink and an antikink
as elementary excitations and clarified their individual characteristics
and the interaction between them.
We found that the excitation energy of the system
is determined mainly by the excited antikink.
The antikink is described as a particle propagating freely
in a region bounded by kinks and/or boundaries.
If the region of its motion is bounded by a kink,
it feels an exchange field which modifies its wave function.
The expectation value of the exchange interaction is usually very small,
since the probability for a kink and an antikink to be
at the nearest neighbor triangles is very small.
The effect of the interaction appears rather as a change
in the effective length of the region of motion for the antikink.
We therefore postulate that the excitation energy of the system is
given by the energy of a freely-propagating antikink as a first approximation.
Then the excitation energy may be written as
\begin{equation}
\delta E=\epsilon_{\rm a}
=\epsilon_{0} + \frac{1}{m}(1-\cos \frac{\pi k}{N'+1}).
\label{eq:deltaE}
\end {equation}
Here $\epsilon_{\rm a}$ denotes the energy of an antikink, $m$ its mass,
$k-1$ stands for the number of nodes in the wave function and
$N'$ the effective length of the region where the antikink propagates.
The creation energy $\epsilon_{0}$ of an antikink is equal to the excitation
energy gap of the $\Delta$-chain in the thermodynamic limit.
Of course, the cosine form in Eq. (\ref{eq:deltaE}) is an approximate
form only valid in the region where $\pi k/(N'+1)$ is small compared to unity.
We treat only such cases.
In this section, we show that
the excitation energies in the periodic systems as well as
those in the open systems considered in the previous section
are well described by Eq. (\ref{eq:deltaE}).
Before comparing Eq. (\ref{eq:deltaE}) with numerical results in detail,
we examine the size dependence of the excitation energy in the periodic chain.
In Fig. \ref{fig:periodicgap} we show the size dependence of
the excitation energies $\delta E$ in the subspace with zero total momentum
classified by the total spin and the parity
in the periodic systems with up to 13 triangles.
The energy levels up to the 7th excitation
in each subspace have been obtained by the Lanczos method.
In the figures for the $S=0$ and the $S=1$ subspace, we find that $\delta E$
for the lowest levels converge to the same value $\sim 0.22$
in the thermodynamic limit.
In the subspace with even parity, $\delta E$
of the second level also seems to converge to the same value.
These levels are considered to be states with a single pair of
a kink and an antikink.
The effect of the exchange interaction between a kink and an antikink
vanishes as $N\to \infty$ and so $\delta E$ for $S=1$ and
$S=0$ converge to the same value corresponding to $\epsilon_0$.
On the other hand, the lowest $\delta E$ in the $S=2$ and $S=3$ subspaces
seem to converge to $2\epsilon_0$ and $3\epsilon_0$, respectively.
The results are consistent with the picture that
the lowest excitation in the $S=2$($S=3$) subspace is a state
with two(three) pairs of a kink and an antikink.
It is still not conclusive whether they are precisely twice or three times,
since the systems we have treated are not large enough
to extrapolate such high excitations.
We also observe that $\delta E$ of higher excitations in $S=0$ and $S=1$
subspaces seems to converge to $2\epsilon_0$.
These states are also regarded as states
with two pairs of a kink and an antikink.
The results mentioned above indicate that
kinks and antikinks behave as independent $S=1/2$ particles
when they are far apart.
We further confirm the above picture by comparing the size dependence of
all the systems we have considered in this paper.
We examine here $\delta E$ of the ground state of the Open-A systems,
the first excited state of the Open-B systems with $S=0$ and that with $S=1$
and
the lowest three excited states of the periodic systems with $S=0$ and those
with $S=1$ in the subspace with zero total momentum.
We assume that
Eq. (\ref{eq:deltaE}) with $k=1$ expresses
$\delta E$ of the ground state of the Open-A systems,
the first excited state of the Open-B system with $S=0$ and that with $S=1$.
The first, the second, and the third excited states of the periodic systems
with $S=0$ and $S=1$ are also assumed to be expressed by
Eq. (\ref{eq:deltaE}) with $k = 1, 2,$ and $3$, respectively.
The parity of these states are consistent with this assignment
(parity = even for $k =1$ and 3; parity = odd for $k=2$).
In Fig. \ref{fig:cosineplot}, we show the plot of $\delta E$ versus
$1-\cos(\pi k/(N'+1))$,
where $N'=N+\Delta N$ is chosen so that all the data fall onto one curve.
$\Delta N$ is a correction to $N$ induced by the boundary effect and/or
the interaction with kinks.
It is remarkable that all data lie on a curve.
This is a clear evidence for that the excitation energy is mainly
contributed by an antikink.
For the Open-A systems we take $\Delta N=3$ for a convention.
Then the best choice of $\delta N$ for other cases
are shown in Table I.
Let us try to understand the above results in an intuitive way.
In Open-B systems, a kink occupies the leftmost triangle.
Then the region allowed for an antikink to move is assumed to be equivalent to
an Open-A system with less triangles by two.
The attractive interaction elongates the effective region for $S=1$ state,
and repulsive interaction and the requirement of the orthogonality to the
ground state shortens it for $S=0$.
We see from the numerical results (Table I)
that the effect of the interaction amounts
to $1.3 (-0.2)$ triangles for $S=1(0)$.
It amounts to $-0.2 (-3)$ triangles for $k=1$ and $S=1(0)$ states
in the periodic systems,
since the allowed region for an antikink is equivalent to
an Open-A system with less triangles by three.
The difference of the effect of the interaction and/or the orthogonality
between the Open-B systems and the periodic systems
is not clarified yet at present.
The least-square extrapolation by using all the data in the figure
gives the creation energy $\epsilon_0$ in the limit $N \to \infty$
and the mass of an antikink $m$.
They are
\begin{equation}
\epsilon_{0}=0.215 \label{eq:ep}\\
\end{equation}
and
\begin{equation}
m =1.21. \label{eq:mass}
\end {equation}
\section{The Low-temperature specific heat}
\label{sec:spec}
In this section, we show that the elementary excitations discussed
in the previous sections generate the low-temperature peak in the specific
heat.
First we calculate the specific heat of a finite system
by numerical diagonalization, and examine whether the result can be
reproduced by the elementary excitation with the spectrum given by
Eq. (\ref{eq:deltaE}) with the parameters (\ref{eq:ep}) and (\ref{eq:mass}).
We consider a periodic system with 8 triangles ($N=8$).
All the $2^{16}$ states can be diagonalized numerically, and
thus the exact specific heat is obtained as shown in Fig. \ref{fig:spec}.
Now we calculate the specific heat due to kink-antikink excitations.
We take only low-lying states into account, i.e.,
the ground state (doubly degenerate), all the
states with one pair of a kink and an antikink,
and some states with two pairs.
Among the states with two kink-antikink pairs,
we take those where the distance between two kinks are largest.
We consider that these are more important than others since
the region for the motion of two antikinks are comparatively large.
We checked that these states
contribute to the low-temperature behavior of the specific heat
by comparing the results that we took all the configurations into account.
The expectation value of energy are written approximately as
\begin{equation}
\langle E\rangle _{\beta} = E_{\rm g}+
\frac{\displaystyle{
8\sum_{S_1=0}^1 \sum_{k_1=1}^7 A_{S_1}
E_{S_1}^{(8)}\exp[-\beta E_{S_1}^{(8)}]
+4\sum_{S_1, S_2=0}^1 \sum_{k_1, k_2=1}^3 A_{S_1}A_{S_2}
(E_{S_1}^{(4)}+E_{S_2}^{(4)})\exp[-\beta (E_{S_1}^{(4)}+E_{S_2}^{(4)})]
}}
{\displaystyle{
2
+8\sum_{S_1=0}^1 \sum_{k_1=1}^7 A_{S_1}
\exp[-\beta E_{S_1}^{(8)}]
+4\sum_{S_1, S_2=0}^1 \sum_{k_1, k_2=1}^3 A_{S_1}A_{S_2}
\exp[-\beta (E_{S_1}^{(4)}+E_{S_2}^{(4)})]
}}, \label{eq:Eav}
\end {equation}
with
\begin{eqnarray}
E_{\rm g} &\equiv& -\frac{3}{4}N \\
E_{S_i}^{(N)}&\equiv& \epsilon_0+\frac{1}{m}\cos \frac{\pi k_i}{N+1}
~~~\mbox{\rm for $S_i=1$} \label{eq:ET} \\
E_{S_i}^{(N)}&\equiv& \epsilon_0+\frac{1}{m}\cos \frac{\pi k_i}{N-2}
~~~\mbox{\rm for $S_i=0$}. \label{eq:ES}
\end {eqnarray}
$\epsilon_0$ and $m$ are given by Eqs. (\ref{eq:ep}) and (\ref{eq:mass}).
$S_i$ denotes the total spin value of the $i$th pair of a kink and an antikink.
Here we have treated the states with two kink-antikink pairs
as two independent pairs for simplicity.
Also we have used integer values for $\Delta N$ in Eqs. (\ref{eq:ET})
and (\ref{eq:ES}).
The spin multiplicity is denoted by $A_S$,
namely $A_S=3$ for $S=1$ and $A_S=1$ for $S=0$.
The specific heat is calculated by differentiating
this equation with respect to the temperature.
The result coincides with the exact specific heat at low temperatures
as shown in Fig. \ref{fig:spec}.
The peak position and the low-temperature side
of the peak are correctly reproduced.
The peak height is a little smaller than the exact value.
The approximate result gives much smaller value than
the exact one at higher temperatures
because of the negligence of higher excited states.
We may conclude that the low-temperature peak is caused
by the low-lying kink-antikink excitations whose spectrum
is described by Eq. (\ref{eq:deltaE}).
Next we calculate the specific heat in the thermodynamic limit.
We consider the system at low temperatures
where the thermal excitations are described by kinks and antikinks.
Let us first consider the case where the mass $m$ is infinite.
In this case, antikinks are localized with the excitation
energy $\epsilon_{0}$.
This system is thermodynamically equivalent to that
where both kinks and antikinks have the excitation energy $\epsilon_{0}/2$,
and it shows a Schottky-type specific heat with an excitation energy
$\epsilon=\epsilon_{0}/2$.
If $m$ is finite, the excitation energy of an antikink
$\epsilon_{\rm a}$ depends on the length $L$
of the region of the motion and on the level $k$ as
$\epsilon_0+(1/m)\cos(\pi k/L)$.
In order to obtain thermodynamic quantities, we must
average them over all states with possible values of $L$ and $k$.
Instead of doing this summation, we calculate the specific heat
in an approximate way.
We assume that the thermal excitations of kink-antikink pairs
at a temperature $T\equiv \beta ^{-1}$ are regulated by the
averaged antikink energy defined by
\begin{equation}
\langle \epsilon_{\rm a}\rangle _{\beta} = \frac{\displaystyle{
\sum_{k=1}^{l} (\epsilon_0+\frac{1}{m}\cos\frac{\pi k}{l})
\exp[-\beta (\epsilon_0+\frac{1}{m}\cos\frac{\pi k}{l})]
}}
{\displaystyle{
\sum_{k=1}^{l}
\exp[-\beta (\epsilon_0+\frac{1}{m}\cos\frac{\pi k}{l})]
\label{eq:epsav} }}.
\end {equation}
Here the integer $l$ is the average distance between two excited kinks, i.e.,
the length of the region where an antikink can move.
We have neglected the effects of interactions between kinks and antikinks.
This may be justified if $l \gg 1$, which will be shown to be the case below.
We assume that the specific heat is written as a Schottky-type one as
\begin{equation}
C(T) = \frac{(\beta\epsilon)^2 \exp[\beta\epsilon]}
{(1+\exp[\beta\epsilon])^2},
~~~ \epsilon=\langle \epsilon_{\rm a}\rangle _{\beta}/2.
\label{eq:c}
\end {equation}
The average length $l$, which is equal to the distance between two
antikinks, is given by
\begin{equation}
l = {\rm int}(\exp[\beta \langle \epsilon_{\rm a}\rangle _{\beta}]),
\label{eq:l}
\end{equation}
since an antikink is excited with the probability
$\exp[-\beta \langle \epsilon_{\rm a}\rangle _{\beta}]$.
We can solve self-consistent equations (\ref{eq:epsav}) and (\ref{eq:l})
by iteration for a given $T$.
Then with obtained $\langle \epsilon_{\rm a}\rangle _{\beta}$,
we can calculate the specific heat at the temperature through Eq. (\ref{eq:c}).
We have employed the parameters given by Eqs. (\ref{eq:ep}) and (\ref{eq:mass})
and plotted the obtained specific heat in Fig. \ref{fig:spec}.
The peak position shifts from the finite-size result toward the low
temperature and locates at $T\sim 0.05$.
The profile of the peak is very steep on its low-temperature side
but it has a tail on the other side.
These features are common with the numerical results of finite-size systems.
This agreement implies that above approximation is
correctly taking into account of the temperature dependence of
$\langle \epsilon_a\rangle _{\beta}$.
The approximation is based on the idea that
an antikink can be treated as a localized object
after averaging the kinetic energy in
$\langle \epsilon_{\rm a}\rangle _{\beta}$.
We consider that this picture is a good approximation
if the discreteness of the antikink kinetic energy is negligible
and the motion of antikinks is well thermalized.
In fact, the average antikink energy $\langle\epsilon_{\rm a}\rangle_{\beta}$
obtained in the above calculation turns out to be $0.244$
at $T\sim0.05$, the peak position of the specific heat.
Therefore the averaged kinetic energy
$\langle\epsilon_{\rm a}\rangle_{\beta} - \epsilon_{0} = 0.029$
is larger than $T/2=0.025$,
the classical thermal average of the kinetic energy.
The result implies that the phase
space of the $l$ states is well thermalized at this temperature.
The average distance between antikinks $l$ is estimated as $\sim 130$,
which is much greater than unity.
We find that the average kinetic energy is much smaller than the creation
energy at this temperature range .
Above results indicate that our approximate scheme
is a consistent one and is correctly
describing the thermodynamics of the system
in the temperature region of the peak in the specific heat.
\section{Conclusion}
\label{sec:conc}
In the previous sections, we have clarified the elementary excitations
in the $\Delta$-chain to a great extent.
They are two types of domain walls called a kink and an antikink.
A kink is essentially localized in a finite range (about one triangle).
An antikink propagates with kinetic energy within a region bounded by kinks.
The dispersionless mode found previously \cite{kubok93} originates
in the localized character of a kink.
The low-temperature peak in the specific heat
is caused by thermal excitations of kinks and antikinks.
The peak position is mainly determined
by the creation energy of the antikink,
while its kinetic energy causes the size-dependence
of the excitation energies as well as
the broadening of the peak of the specific heat.
Above understanding of the $\Delta$-chain might give some insight into the
properties of the {\it kagom\'e} antiferromagnet.
Though the dimensionality of two systems are different,
they have some common features.
Both systems have macroscopically degenerate classical
ground states and show the second peak in the specific heat at low
temperatures.
The lower peak in the specific heat is observed in finite {\it kagom\'e}
antiferromagnets.\cite{elser89,elstner-y94,zeng-e95,nakamura-m95}
The uniform susceptibility seems to decrease suddenly
at the same temperature.\cite{elstner-y94,nakamura-m95}
In the $\Delta$-chain, the spin gap corresponds to the first excitation gap.
Above similarities suggest that the elementary excitations with a spin gap
in both systems may have common characters that create the second peak of
the specific heat.
Recently, Zeng and Elser \cite{zeng-e95}
investigated the {\it kagom\'e} antiferromagnet by dimer calculations.
Their results support this speculation, but
further investigation on the {\it kagom\'e} antiferromagnet
is necessary for a concrete understanding of the system.
\acknowledgments
Numerical diagonalization was done with TITPACK Ver. 2. programed by
Professor H. Nishimori.
Use of the random number generator RNDTIK programmed by N. Ito and
Y. Kanada is gratefully acknowledged.
Computations were done partly on FACOM VPP500
at the Science Information Processing Center, University of Tsukuba,
and partly on FACOM VPP500 at the Institute for Solid State Physics,
University of Tokyo.
|
\section{Introduction}
An increasing number of actual problems in physics find their natural
expression not in the the static (equilibrium) aspects of quantum
systems, but in the kinematical (non-equilibrium) development of their
average properties. Examples include studies of early universe
expansion\cite{calzetta1,pi1,lawrie1}, heavy ion collisions and the
postulated quark-gluon plasma\cite{heinz1}, lasers and other driven
systems\cite{korenman1,senitzky1,senitzky2}, and particle creation in changing
fields of
force\cite{brown1,mottola1}.
The migration from equilibrium to non-equilibrium involves a shift of
paradigm. In common with zero-temperature field theory, particle
systems at equilibrium are often treated by a scattering formalism,
with an initial (in) state and a final (out) state; this is only
sensible if both are known and are at equilibrium with the same
thermodynamic reservoir. The physics of a non-equilibrium system
demands different boundary conditions. The initial and final states
are (by definition) not characterized by the same ensemble and it is
more appropriate to define the state (spectral profile or density
matrix) of the system at some initial time $\overline t_{i}$ and
compute the final state and its consequence at a later time $\overline
t$. This describes to an initial value problem which is deftly
handled by Schwinger's closed time path (CTP) action principle. The
new picture also implies a concern with probabilities, or expectation
values rather than amplitudes.
In equilibrium, one is used to the notion of translational invariance
in space and time, implying that physical
quantities only depend on the differences of coordinates $x-x'$. When
the field is driven into disequilibrium, it acquires an additional
dependence on the average position and time $\overline x =
\frac{1}{2}(x+x')$. This is measured relative to an initial point of
reference $\overline x_{i}$. In practical applications it is usually
necessary to assume that the dependence on the average coordinate is
{\em quasi-static} or of low adiabatic order in order to make
computations tractable. The dependence on the average coordinate has important
features: the preservation of unitarity demands that the statistical
state of the field only depend on $\overline x$ and not $x-x'$. Since
the state of the field can only be altered by the intervention of
sources or sinks (hereafter referred to collectively as sources), the
sources must also develop with respect to the average coordinate. Since
one is interested both in fluctuations and the average kinematics, it
is convenient to work with variables and sources which are
bi-local objects rather than working with
the field itself. This is in contrast to the pure field approach used
by Schwinger\cite{schwinger2}. Self-interacting theories are a special case in
which
the field is its own source; they pose mainly calculational
problems---conceptually no new issues are introduced other than
self-consistency.
Since the external sources affect the average state, they can be
regarded as thermodynamical reservoirs, with the caveat that they must
suffer a `recoil' or back-reaction as a result of their effect on the
system. This is not negligeable off equilibrium. Many quantum systems
(the laser, for example) can be treated as two-reservoir systems in
which the `external' reservoir is of comparable magnitude to the local
one.
The nomenclature `open system' is used to describe a system coupled
to independent sources. The name `closed system' is given to
a system without sources or one in which the field is its
own source; in the latter case, the source must become
effectively impotent with regard to the further development
of the average state, as equilibrium is approached.
Equilibrium is only achieved when, either the whole system
approaches some driven limit cycle, or the contact with
external sources is effectively terminated.
In a closed system, the final equilibrium is a thermal state,
or a state of maximum entropy. Some authors define equilibrium
to mean a thermal state rather than merely a static one---and
non-equilibrium to mean anything else. This is somewhat
misleading since a non-thermal but static state is still final
in the absence of new perturbations and must therefore be
considered a point of equilibrium for the system.
The purpose of the present paper is to extend Schwinger's method of
analysis to treat non-equilibrium ensembles of bosons. This
touches on and extends a number of apparently different approaches to
non-equilibrium\cite{calzetta1,lawrie1,eboli1}. Since Schwinger's
original work \cite{schwinger2} on the initial-value problem, most
authors have been seduced by the functional integral and have
therefore missed the often subtle advantages of Schwinger's
methodology. It is intended that the present work should convey a
pedagogical flavour of the suggested approach, which overlaps with the
existing literature in strategic places without actually following any
of them. In particular, conventions and definitions (which differ
from most other accounts) have been chosen rather carefully for
practical purposes. Some well known results are rederived in order to
make the present work as self-contained as possible. The paper begins
with a summarial discussion of the formalism, paying special attention
to the action principle and unitarity; later the most general
quadratic theory which maintains unitarity is presented and the Green
functions are calculated for prescribed sources. Particular attention
is paid to the effects of non-locality in the sources---an issue which
has been largely neglected in previous work, and turns out to place
strong requirements on the behaviour of stable systems. Finally a
brief comparison is made between the present work and other approaches.
\section{Formalism and conventions}
The conceptual framework for the decription of non-equilibrium processes
will include operator field theory, the method of sources
and the local momentum space Green functions. In addition it
proves convenient to use Schwinger's quantum action principle.
This is a statement about the unitary
development of the field with respect to the variation of certain
variables. Since it embodies the equations of motion and the
fundamental commutation relations
for the field, it is both compact and elegant. One begins
with the action operator, which is defined to be the classical
action with the classical field replaced by the field operator, together
with a suitable ordering prescription for the fields.
Here the ordering will be the usual time-ordering and the action
that for a real scalar field without self-interactions.
The Minkowski metric-signature is $(- + + +\cdots)$ which allows
straightforward comparison with the Euclidean theory.
\begin{equation}
S = \int dV_x \lbrace \frac{1}{2}(\partial^\mu\phi)(\partial_\mu\phi) +
\frac{1}{2}m^2\phi^2 - J\phi\rbrace\label{eq:1}
\end{equation}
where $dV_x$ is the Minkowski volume element. The operator
equations of motion now follow from the quantum action
principle\cite{schwinger1}
\begin{equation}
\delta \langle \phi| \phi'\rangle = i \langle \phi|\delta S |\phi'\rangle
\label{eq:2}
\end{equation}
giving
\begin{equation}
\left( -{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}} + m^2 \right) \phi(x) =
J(x)\label{eq:3}
\end{equation}
Given that $\phi(x)=\phi(x_i)$ at initial time $t_i$ (or, more
generally, on the the space-like hypersurface $\sigma_i$), the solution
to (\ref{eq:3}) may be written
\begin{equation}
\phi(x) = \phi(x_i) + \int_{\sigma_i}^{\sigma} dV_{x'} G_c(x,x')J(x')
\label{eq:4}
\end{equation}
where $\sigma_i$ and $\sigma$ are the initial and final hypersurfaces
and $G_c(x,x')$ is a Green function which satisfies
\begin{equation}
(-\stackrel{x}{{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}} +m^2) G_c(x,x') =
\delta(x,x')
\label{eq:5}.
\end{equation}
Both the Feynman propagator and the retarded Green function have this
property.
The surface integral under the variation of the action vanishes
independently implying that the generator of infinitesimal
unitary transformations on the field is\cite{schwinger1}
\begin{equation}
\chi_\phi = \int d\sigma^\mu \phi\partial_\mu\phi.\label{eq:6}
\end{equation}
Since it is easily established\cite{schwinger1} that
the unitary variation of any operator $A$ is
\begin{equation}
\delta A = -i \lbrack A,\chi_A \rbrack\label{eq:7}
\end{equation}
it follows that, on any spacelike hypersurface with
orthogonal vector $\hat{n}^\mu$, one has
\begin{equation}
\lbrack \phi, \Pi_\mu \rbrack \hat{n}^\mu = i\delta({\bf x},{\bf
x'})\label{eq:8}
\end{equation}
with $\Pi_\mu = \partial_\mu \phi$.
This is the covariant statement of the canonical commutation
relations for the field and its conjugate momentum.
To avoid unnecessary notation it is convenient to write this simply as
\begin{equation}
\lbrack\phi({\bf x},t), \partial_t\phi({\bf x'},t) \rbrack = i\delta({\bf
x},{\bf x'})\label{eq:9}
\end{equation}
with the understanding that general covariance is easily restored
by introducing a suitable time-like vector.
{}From the action principle (\ref{eq:2}) it can be shown by
repeated functional differentiation with respect to the
source that
\begin{equation}
\frac{\delta^n\langle\phi_2|\phi_1\rangle_J}{\delta J(x_1)\ldots \delta J(x_n)}
= i^n \langle \phi_2|T(\phi(x_1)\ldots\phi(x_n)) |\phi_1 \rangle\label{eq:10},
\end{equation}
thus the Taylor expansion of the amplitude may be written in the shorthand
form
\begin{equation}
\langle\phi_2|\phi_1\rangle_J = \langle \phi_2|T
e^{iJ\phi}|\phi_1\rangle\label{eq:11}
\end{equation}
where $T$ denotes time-ordering (latest time to the left). This formula
may be regarded as a generating functional for the $n$-point functions
of the theory.
The complex conjugate of this quantity is
\begin{equation}
\langle\phi_2|\phi_1\rangle_J^\dagger= \langle\phi_1|\phi_2\rangle_J = \langle
\phi_1|T^\dagger e^{-iJ\phi}|\phi_2\rangle\label{eq:12}
\end{equation}
where $T^\dagger$ stands for anti-time-ordering (latest field to the right).
This reverse-ordering is necessary to ensure the cancellation of intermediate
fields in the identity:
\begin{equation}
\langle\phi_2|\phi_1\rangle_J \times\langle\phi_1|\phi_2\rangle_J = \langle
\phi_2|T^\dagger e^{-iJ\phi}T e^{iJ\phi}|\phi_2\rangle = 1\label{eq:13}.
\end{equation}
This is the key observation for the construction of the expectation values.
Notice how the operator ordering in (\ref{eq:13}) starts from an early
time, increases to a final time (at the centre of the operator product) and
then reverses back to the initial time. Each field, at each instant along the
closed time path has a mirror counterpart required for the cancellation of the
intermediate operators in (\ref{eq:13}). This property can now be used to
advantage to construct a `closed time path' action
principle\cite{schwinger2,bakshi1}.
Consider an expectation value of the form
\begin{equation}
\langle t| X(t')|t\rangle = \sum_{i,i'} \langle t|i\rangle\langle i|X(t')|i'
\rangle\langle i'|t \rangle\label{eq:14}
\end{equation}
where the sum over intermediate states $i,i'$ is a sum over all states
and $\langle t|$ is a shorthand which refers to either a pure state of the
system, or a
mixed state, specified at time $t$. The expectation value specifies
the average value of the operator $X$ at the time $t'$ given the state
of the system at time $t$. It involves conjugate amplitudes and hence the
conjugate forms of the
action principle:
\begin{eqnarray}
\delta \langle t|i\rangle &=& i\langle t|\delta S_{ti}|i \rangle\label{eq:15}\\
\delta \langle i|t \rangle &=& -i\langle i|\delta S_{it}^\dagger|t
\rangle\label{eq:16}\\
S_{ab} &=& \int_a^b dt L\label{eq:17}.
\end{eqnarray}
To obtain (\ref{eq:14}) from an action principle one would therefore like to
introduce the operator $X$ by functional differentiation with respect to an
appropriate source (or combination of sources) between an amplitude and its
conjugate. This is achieved in the following way. First one observes that
\begin{equation}
\delta \langle t|t \rangle = \delta(\sum_i \langle t|i \rangle \times \langle
i|t \rangle)
= \delta(1) = 0\label{eq:18},
\end{equation}
so differentiation of this object is to no avail.
However, if we make an artificial distinction between the amplitude and its
conjugate by labelling all objects in the former with a $+$ symbol and all
objects in the latter with a $-$ symbol,
\begin{equation}
\delta \langle t|t\rangle = \lim_{+\rightarrow -}i \langle t|\delta S_+-\delta
S_-^\dagger|t\rangle
\label{eq:19}
\end{equation}
then we can use the solution of this quantity as a generating functional
for (\ref{eq:14}) since $X$ can be expanded in terms of either
$\frac{\delta}{\delta J_+}$
or$\frac{\delta}{\delta J_-}$. This breaks the symmetry of symbols in
(\ref{eq:18}).
At the end of a variational calculation one removes the $+$ and $-$ symbols
restoring the conjugate relationship between the two amplitudes,
having inserted the appropriate operators by differentiation with respect
to the source of only one of them. Note in (\ref{eq:19})
that, for any unitary field theory, the action is self-adjoint, thus we may
drop the dagger symbol in future. Also, in treating the $+$ and $-$ parts of
the
field as being artificially independent, the condition
\begin{equation}
\phi_+(t_{\infty}) = \phi_-(t_{\infty})\label{eq:20}
\end{equation}
is required to ensure that the limit $+\rightarrow -$ restores the single
identity of the field operators, and additionally one must have that all
$\phi_-$ fields (at any time) must stand to the left of all $\phi_+$ fields
(at any time). Since $-$ fields are anti-time-ordered and $+$ fields
are time ordered, this condition arises naturally and ensures the triviality
of (\ref{eq:13}).
The meaning of the above procedure can be illustrated by noting that the
solution
to (\ref{eq:19}) may be written
\begin{equation}
\langle t_i|t_i\rangle_{J_{\pm}} = \langle t_i|T^\dagger e^{-iJ_i\phi_-}T
e^{iJ_+\phi_+} |t_i\rangle
\label{eq:21}.
\end{equation}
The expectation value of the field is found using the ordered expression
\begin{eqnarray}
-i\frac{\delta}{\delta J_+(x)} \langle t_i|t_i\rangle \Bigg |_{+=-} &=& \langle
\phi(x) \rangle\\
&=& \lim_{+\rightarrow -}\langle t_i|
\exp\left(-i\int_{t_i}^{\infty}J_-\phi_-
+i\int_{t}^{\infty}J_+\phi_+ \right)
\phi_+(x)
\exp\left(i\int_{t_i}^{t}\right)
|t_i\rangle\label{eq:22}.
\end{eqnarray}
Taking the limit $+\rightarrow -$, one has
\begin{equation}
\langle \phi(x)\rangle = \langle t_i|\exp\left(-i\int_{t_i}^{t}J\phi\right)
\phi(x)
\exp\left(i\int_{t_i}^{t}J\phi\right)
|t_i\rangle.
\label{eq:23}
\end{equation}
This shows that the average value of the operator depends only on the
past (retarded) history of the system beginning from the initial time $t_i$.
It can be shown (see appendix A) that the closed time path generating
functional is closely related to the generator for the retarded $n$-point
functions.
The acausal (advanced) pieces cancel in the limit $+\rightarrow -$.
So far, the discussion has used the slightly trivial example of pure states
$\langle t|$.
As noted implicitly by Schwinger\cite{schwinger2}, the same action principle
holds when $\langle t| \ldots |t\rangle$ is replaced by $\langle t|\rho(t)
\ldots |t\rangle$
(a mixture of states) since this does not affect
the conjugate relationship between amplitudes.
The nature of the expectation value can therefore be left out
of the discussion for the most part. Indeed, in practice, the effect of a
non-trivial
density matrix in the expectation value can be mimicked by the introduction of
suitable sources\cite{schwinger2,calzetta2}---a procedure which will be adopted
in the next section. To present the formalism in a way conducive to
generalization, the next step is to present the Green functions for the
case of pure-state vacuum expectation values and then introduce the finite
temperature (mixed state) modifications which will be the starting point for
writing down
an ansatz for non-equilibrium.
The above use of generating functionals is closely related to the path integral
approaches of Calzetta and Hu\cite{calzetta1}, and Lawrie\cite{lawrie1}.
It proves useful not to pass directly to the path integral however, but to
follow Schwinger's approach.
For the remainder of the paper, equation (\ref{eq:19}) will be considered the
starting point for the discussion of non-equilibrium field theory.
{}From equations (\ref{eq:1}) and (\ref{eq:19}) one obtains the operator
equations
of motion for the field. Taking the initial time to be
$t_i$, the furthermost future time to be $t_\infty$ and
the final time at which expectation values are to be computed
as $t_f$, then using the boundary condition in equation (\ref{eq:20}),
\begin{eqnarray}
\phi_+(x) &=& \phi(x_i) + \int_{t_i}^{t} G_c(x,x')J_+(x') dV_{x'}\nonumber\\
\phi_-(x) &=& \phi_+(t_\infty) + \int_{t_\infty}^{t_f} G_c(x,x') J_-(x')
dV_{x`}\nonumber\\
&=& \phi(x_i) + \int_{t_i}^{t_\infty} G_c(x,x')J_+(x') dV_{x'}+
\int_{t_\infty}^{t_f} G_c(x,x') J_-(x') dV_{x`}\label{eq:24}
\end{eqnarray}
where $G_c(x,x')$ is a retarded (causal) Green function.
Notice that, as the distinction between $+$ and $-$ is removed, these equations
reduce to (\ref{eq:4}). Substituting these into the exponential solution to
(\ref{eq:19}) and defining a vector and its transverse by $J^T = (J_+ J_-)$,
one
may write
\begin{equation}
\ln \langle 0,t_i|0,t_i\rangle = -i\int_{t_f}^{t_\infty}
J_-(x')dV_{x'}\phi(x_i)
+ i \int_{t_i}^{t_\infty} J^T(x) G(x,x') J(x') dV_{x}dV_{x'}
\label{eq:25}
\end{equation}
where
\begin{equation}
G(x,x') = \left(
\begin{array}{cc}
\theta(x-x')G_c(x,x') & 0 \\
-G_c(x,x') & \theta(x'-x)G_c(x,x')
\end{array}
\right),
\label{eq:26}
\end{equation}
and $\theta(x-x')$ is the step function which satisfies
\begin{equation}
\theta(x) + \theta(-x) = 1\label{eq:27}.
\end{equation}
As a result of this property, the sum of rows and columns in (\ref{eq:26}) is
zero. This is a reflection of the triviality of equation (\ref{eq:13}).
It further implies the causality of expectation values derived from this
generating functional. While (\ref{eq:26}) has a simple physical
derivation in terms of the equations of motion, a more symmetrical
form can be obtained by attaching a variational interpretation to
$G_c(x,x')$ directly. Again, following Schwinger, and varying with respect
to the sources
\begin{equation}
\delta_2\delta_1 \langle t_i|t_i \rangle = (i)^2 \langle
t|(\delta_2S_+-\delta_2S_-)(\delta_1S_+-\delta_1S_-) |\rangle t>
\label{eq:28}
\end{equation}
where, according to the ordering rule, this equals
\begin{equation}
\delta_2\delta_1 \langle t_i,t_i \rangle = (i)^2 \langle
t|\phi_+(x_2)\phi_+(x_1)
+\phi_-(x_2)\phi_-(x_1) - \phi_-(x_2)\phi_+(x_1) - \phi_+(x_2)\phi_-(x_1)
|\rangle t>.
\label{eq:29}
\end{equation}
Comparing the solution of this to
\begin{equation}
\exp\left( \frac{i}{2}\int dV_x dV_{x'} J^T G(x,x') J(x')\right) \label{eq:30}
\end{equation}
one has
\begin{equation}
G(x,x') = \left(
\begin{array}{cc}
G_{++} & G_{+-} \\
G_{-+} & G_{--}
\end{array}
\right),
\label{eq:31}
\end{equation}
where
\begin{eqnarray}
\langle \phi_+(x)\phi_+(x')\rangle &=& -iG_{++}(x,x')\label{eq:32}\\
\langle\phi_+(x)\phi_-(x') \rangle &=& iG_{+-}(x,x')\label{eq:33}\\
\langle\phi_-(x)\phi_+(x') \rangle &=& iG_{-+}(x,x')\label{eq:34}\\
\langle \phi_-(x)\phi_-(x')\rangle &=& -iG_{--}(x,x')\label{eq:35}.
\end{eqnarray}
As the distinction between $+$ and $-$ is lifted, the assumed ordering
implies that
\begin{eqnarray}
G_{++}(x,x') &=& i \langle T \phi(x)\phi(x')\rangle = G_F(x,x')\label{eq:36}\\
G_{+-}(x,x') &=& -i \langle \phi(x)\phi(x')\rangle =
-G^{(-)}(x,x')\label{eq:37}\\
G_{-+}(x,x') &=& -i \langle\phi(x')\phi(x)\rangle =
G^{(+)}(x,x')\label{eq:38}\\
G_{--}(x,x') &=& i \langle T^\dagger\phi(x)\phi(x') \rangle =
G_{AF}(x,x')\label{eq:39}\\
\end{eqnarray}
where $G_F$ is the Feynman propagator, $G^{(\pm)}$ are the positive and
negative
frequency Wightman functions and $G_{AF}$ is the anti-time ordered propagator.
In the limit of zero source, these quantities satisfy the equations
\begin{eqnarray}
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}} + m^2) G_F(x,x') &=&
\delta(x,x')\label{eq:40}\\
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}} + m^2) G^{(\pm)}(x,x')
&=& 0\label{eq:41}\\
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}} +m^2) G_{AF}(x,x') &=& -
\delta(x,x')\label{eq:42}\\
\end{eqnarray}
by virtue of equation (\ref{eq:5}) for the field operator. The
non-zero right hand side of (\ref{eq:40}) and (\ref{eq:41}) are
due to the time ordering. From the time-ordering (\ref{eq:36}) it
follows that
\begin{equation}
-iG_F(x,x') = i\theta(t-t')G^{(+)}(x,x') -i\theta(t'-t)G^{(-)}(x,x').
\label{eq:43}
\end{equation}
Substituting this relation into (\ref{eq:36}) and using the commutation
relations for the field (\ref{eq:9}) proves (\ref{eq:36}).
Similarly (\ref{eq:42}) follows from the relation
\begin{equation}
G_{AF}(x,x') = - G_F^*(x,x')\label{eq:44}.
\end{equation}
Using (\ref{eq:27}), it is now straightforward to see that the sum
of the rows and columns in (\ref{eq:31}) is vanishing, as required
for causality. A number of additional relations between the
Green functions can be proven. The retarded and advanced Green
functions satisfy:
\begin{eqnarray}
G_{ret}(x,x') &=& -\theta(t-t')\lbrack \phi(x),\phi(x')\rbrack\nonumber\\
G_{adv}(x,x') &=& \theta(t'-t)\lbrack \phi(x),\phi(x')\rbrack\label{eq:45}.
\end{eqnarray}
Also, in virtue of (\ref{eq:27}) it is easy to see that
\begin{eqnarray}
G_F &=& G_{ret} + G^{(-)} = G_{adv} - G^{(+)}\label{eq:46}\\
G_{AF} &=& = G_{ret}+G^{(+)} = G_{adv} - G^((-))\label{eq:47}.
\end{eqnarray}
The unequal-time commutator and anti-commutator Green-functions are defined
by
\begin{eqnarray}
\tilde{G}(x,x') &=& \lbrack \phi(x),\phi(x')\rbrack =
G^{(+)}+G^{(-)}\nonumber\\
\overline{G}(x,x') &=& \lbrace \phi(x),\phi(x')\rbrace = G^{(+)}-G^{(-)}
\label{eq:48}
\end{eqnarray}
These will be useful later and serve to pin-point the conventions used
in this work. Before considering the momentum-space
representation of these functions it is useful to note
that $G(x,x')$ can be written entirely in terms of the formal quantity
\begin{equation}
H(x,x') \equiv i \langle t|\phi(x)\phi(x')|t\rangle\label{eq:49}
\end{equation}
as
\begin{equation}
G(x,x') = \theta(t-t')\left(
\begin{array}{cc}
H(x,x') & -H(x',x) \\
-H(x,x') & H(x',x)
\end{array}
\right)
+
\theta(t'-t)
\left(
\begin{array}{cc}
H(x',x) & -H(x',x) \\
-H(x,x') & H(x,x')
\end{array}
\right)
,
\label{eq:50}
\end{equation}
where
\begin{equation}
H(x',x) = H(x,x')^*
\label{eq:51}
\end{equation}
and the sum of rows and columns is manifestly zero.
Since the spectrum of the operator $-{\vcenter{\vbox{\hrule
height.4pt\hbox{\vrule width.4pt height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}+m^2$ on the
complex wave $e^{ikx}$ is solved for any $k$ satisfying
a dispersion relation, the solution to (\ref{eq:41}) is
the most general linear combination of plane waves satisfying
the dispersion relation $k^2+m^2=0$. This implies that the vacuum
positive and negative frequency Wightman functions can be
written, in $n$ spacetime dimensions,
\begin{eqnarray}
G^{(+)} &=& - 2\pi i \int \frac{d^nk}{(2\pi)^n}e^{ik_\mu(x-x')^\mu}
\theta(k_0)\delta(k^2+m^2)\label{eq:52}\\
G^{(-)} &=& 2\pi i \int \frac{d^nk}{(2\pi)^n}e^{ik_\mu(x-x')^\mu}
\theta(-k_0)\delta(k^2+m^2)\label{eq:53}.
\end{eqnarray}
Defining the fourier transform of $G(x,x')$ by
\begin{equation}
\int \frac{d^nk}{(2\pi)^n} e^{ik_\mu(x-x')^\mu} G(k),\label{eq:54}
\end{equation}
and using the integral representation
\begin{equation}
\theta(t-t') = i \int_{-\infty}^{+\infty} \frac{d\omega}{2\pi}
\frac{e^{-i\omega(t-t')}}{\omega+i\epsilon}\label{eq:55}
\end{equation}
it is straightforward to show from (\ref{eq:43}) that
\begin{equation}
G_F(k) = \frac{1}{k^2+m^2-i\epsilon},\label{eq:56}
\end{equation}
which is fully consistent with (\ref{eq:40}). $G_{AF}(k)$ is
easily obtained from (\ref{eq:44}).
Note that, had the symmetrical form of $G(x,x')$ not been used,
similar results could still have been obtained. It is possible, in
the manner of a symmetry transformation to redefine the
Wightman functions so that positive and negative frequencies are
mixed. This simply mixes up the Feynman and anti-Feynman propagator
also. For instance, if one defines
\begin{equation}
G^{(+)}(k) = 2\pi i\delta(k^2+m^2) \lbrack\theta(k_0) +\alpha\theta(k_0)
+\beta\theta(-k_0)\rbrack
\label{eq:57},
\end{equation}
then the corresponding Feynman propagator becomes
\begin{equation}
G_F(k) = \frac{1+\alpha}{k^2+m^2-i\epsilon} -
\frac{\beta}{k^2+m^2+i\epsilon}\label{eq:58}
\end{equation}
where the last term is evidently a piece from $G_{AF}$.
Since this only complicates the matter, such redefinitions will not
be pursued further.
So far, this summary of the action principle has only
explicitly encompassed pure state
expectation values, which are comparatively trivial. A statistical system
with real particle densities, as well as perhaps a temperature and entropy,
is described by a mixture of such vacuum expectation values (since the
character of the actual pure state is not usually knowable), with the
the statistical weight given by the density matrix $\rho$. The simplest
example of such is a system in thermal equilibrium ($\rho = \exp(-\beta H)$).
Although a thermal system is quite extraordinary as many particle systems
go, it serves as a useful reference point, both from the viewpoint of
formalism and from a physical perspective, since very many physical systems
can be characterized by a temperature of sorts. A statistical
expectation value for some operator $X$ may be written
\begin{equation}
\langle t| X(t')|t \rangle_s \equiv \frac{{\rm Tr} \langle t| \rho(t) X(t') |t
\rangle}
{{\rm Tr} \langle t| \rho(t) |t \rangle}
\label{eq:59}
\end{equation}
and characterizes the average value of $X$ at the time $t'$ given the
state of the system at time $t$. Notice that the trace is over
{\em probabilities} of the form $\langle t|t \rangle$ rather than
amplitudes $\langle t|t'\rangle$. The latter would be meaningless.
The structure of the expectation value is therefore simply that in
equation (\ref{eq:14}) and the closed time path action principle
applies. Indeed, it is noteworthy that the density matrix itself
is merely an operator which can effectively by introduced into
the pure state generating functional by functional differentiation
with respect to an appropriate source. There is therefore no loss
of generality in taking the closed time path action principle
at face value and making no special reference to $\rho$..
The cyclic property of the trace in (\ref{eq:59}) has noteworthy
implications for the Green functions and sources in the CTP formalism.
Consider the expectation value in (\ref{eq:59}). This may be
rewritten as
\begin{eqnarray}
\langle X(t')\rangle_s &=&
\frac{{\rm Tr}\langle \rho(t) e^{iH(t'-t)}X(t)e^{-iH(t'-t)}\rangle}
{{\rm Tr} \langle t| \rho(t) |t \rangle}\nonumber\\
&=& \frac{{\rm Tr}\langle e^{-iH(t'-t)} \rho(t) e^{iH(t'-t)}X(t) \rangle}
{{\rm Tr} \langle t| \rho(t) |t \rangle}
\label{eq:60}
\end{eqnarray}
where $H$ is the Hamiltonian of the system.
Using this `relativity' between the time-dependence of $\rho$ and $X$, it
is possible to place all of the dynamical development of the system
in either one or the other. An example of the use of density matrix
time-development is given in ref. \cite{eboli1}. In the CTP formalism,
the distinction between forward moving times and backward moving
times makes equation (\ref{eq:60}) effectively
\begin{equation}
\langle X(t')\rangle_s=
\frac{\langle e^{-iH_-(t'-t)} \rho(t) e^{iH_+(t'-t)}X(t) \rangle}
{ {\rm Tr} \langle t | \rho(t) |t \rangle}.
\label{eq:61}
\end{equation}
The cyclic property of the trace therefore implies that the density
matrix $\rho$ always sits between the $+$ and $-$ branches of the
operator product and hence it must always be reflected by the
{\em off-diagonal} terms in $\pm$-space. In the special case of
a thermal density matrix, the same observation leads to the well-known
KMS condition\cite{martin1}, by identifying the inverse-temperature $\beta$
with imaginary time. This is seen by considering the thermal Wightman function
\begin{eqnarray}
G^{(+)}_\beta(x,x') &=& \frac{{\rm Tr} \langle t| e^{-\beta H}
\phi(x)\phi(x')|t \rangle}
{{\rm Tr} \langle t| e^{-\beta H} |t \rangle}\nonumber\\
&=& \frac{{\rm Tr} \langle t| e^{-\beta H} \phi(x)e^{\beta H}e^{-\beta
H}\phi(x')|t \rangle}
{{\rm Tr} \langle t| e^{-\beta H} |t \rangle}\nonumber\\
&=&\frac{{\rm Tr} \langle t| e^{-\beta H} \phi(x')\phi({\bf x},t+i\beta)|t
\rangle}
{{\rm Tr} \langle t| e^{-\beta H} |t \rangle}\nonumber\\
&=& - G^{(-)}({\bf x},t+i\beta,x')
\label{eq:62}
\end{eqnarray}
where the cyclic property of the trace has been used.
The left and right hand sides are precisely the elements
of the off-diagonal $G_{+-}$ and $G_{-+}$.
This property is, in fact, sufficient to determine the
thermal Green functions.
To determine these, in a form which manifestly reduced to
the vacuum case, one writes
\begin{eqnarray}
G^{(+)}(k) = -2\pi i \lbrack\theta(k_0)+X \rbrack\delta(k^2+m^2)\nonumber\\
G^{(-)}(k) = 2\pi i \lbrack\theta(-k_0)+Y \rbrack\delta(k^2+m^2)
\label{eq:63}
\end{eqnarray}
with $X$ and $Y$ to be determined. Since the commutator
function $\tilde{G}(x,x')$ must be independent of the state
of the system (in order to preserve the canonical commutation
relations), it follows immediately that $X=Y$. If one then employs
the KMS condition which, in momentum space, becomes the definite relation
\begin{equation}
G^{(+)}(k) = - e^{\beta k_0}G^{(-)}(k)
\label{eq:64}
\end{equation}
it follows that
\begin{equation}
\theta(k_0) +X = e^{|k_0|\beta} \lbrack \theta(-k_0)+X\rbrack
\label{eq:65}
\end{equation}
and hence
\begin{equation}
X = \theta(k_0)f(|k_0|) = f_{>0}(k_0)
\label{eq:66}
\end{equation}
where
\begin{equation}
f(|k_0|) = \frac{1}{e^{\beta|k_0|}-1}.
\label{eq:67}
\end{equation}
By considering (amongst other things) $\overline{G}(x,x')$, it
follows that $f(|k_0|)$ is an even function of $|k_0|$ thus
$f(k_0)\theta(k_0)=f(-k_0)\theta(-k_0)$, whereupon it is
trivial to show the unitarity relation
\begin{equation}
G^{(+)}(x,x') = G^{(-)}(x,x')^*.
\label{eq:68}
\end{equation}
Note that the fact that $G^{(+)}$ consists only of positive
frequencies $k_0$ is pivotal in this derivation. The Feynman
propagator can now be obtained from equation (\ref{eq:43}) by using
the integral representation of the step-function (\ref{eq:55}). The
thermal Green functions are therefore summarized by
\begin{eqnarray}
G^{(+)}(k) &=& -2\pi i \theta(k_0) \lbrack 1+f(|k_0|) \rbrack
\delta(k^2+m^2)\label{eq:69}\\
G_F(k) &=& \frac{1}{k^2+m^2-i\epsilon} +2\pi i
f(|k_0|)\delta(k^2+m^2)\theta(k_0)\label{eq:70}.
\end{eqnarray}
Another important form of $G^{(\pm)}$ is obtained by
performing the integral over $k_0$, thereby enforcing the
role of the delta-function in (\ref{eq:63}). This gives
a result which will prove more useful for calculations later and is more
closely
related to the ansatz used by Lawrie in ref. \cite{lawrie1}:
\begin{eqnarray}
G^{(+)}(x,x')&=& -2\pi i \int \frac{d^{n-1}k}{(2\pi)^{n-1}}
e^{i({\bf k\cdot} ({\bf x}-{\bf x'})-\omega (t-t'))}
\frac{(1+2f_{>0}(|\omega|))}{2|\omega|}\nonumber\\
&=& -2\pi i \int \frac{d^{n-1}k}{(2\pi)^{n-1}}
e^{i({\bf k\cdot} ({\bf x}-{\bf x'})-\omega (t-t'))}
\frac{(1+f(|\omega|))}{2|\omega|}
\label{eq:71}
\end{eqnarray}
where $f_{>0}(k_0)$ is the function composed of only positive
frequencies.
It is now straightforward to verify that the canonical commutation
relations are satisfied, by differentiating $\tilde{G}(x,x')$
with respect to $t'$ (see equation (\ref{eq:48})).
This completes the presentation of conventions
to be used in the remainder of the paper. It is convenient to
add here that a bar (e.g. $\overline a$) represents
an object which is even, while an object with a tilde (e.g. $\tilde a$)
represents one which is odd with respect to its arguments.
\section{Interaction with sources}
The formalism demonstrated so far has been for free fields. Free
fields are always in a state of equilibrium and therefore the discussion needs
to be
widened to incorporate collisions or interactions. The present work
will deal with interactions which can be mediated by sources of the
type $\phi(x)A(x,x')\phi(x')$. This includes a variety of self interactions,
contact with external forces and noise or impurity scattering,
depending on the nature of $A(x,x')$.
The self-energy of an interacting field theory has this form,
for instance, thus sources of the quadratic type can also be a representation
of the
lowest order, self-consistent `particle dressings'. In Feynman diagram
language,
these represent the re-summed one-loop approximation, Hartree approximation
and so on. Lawrie\cite{lawrie1} uses the notion of such sources to effect
a renormalization (resummation) of perturbation theory in a real scalar
field theory. The same idea is expressed in a different language in the
work of Calzetta and Hu\cite{calzetta1}.
Since it is not the aim of this paper to discuss specific
models, the specific nature of the source terms will not be specified here.
Rather
the discussion will centre around what general properties such a system
might have and a discussion of possible applications will follow.
In an interacting theory, one normally perturbs about the free
field theory. Unfortunately, the dispersion relation (or `mass shell'
constraint) for
free particles is no longer appropriate, since it reflects none of the
interactions which `dress' the particles. A more satisfactory starting
point would be a `quasi-particle' mass shell, including some of the
interaction effects as the basis for a perturbation theory.
This is the essence of a renormalization and can be effected by the
use of sources\cite{burgess99}.
The starting point for the investigation of non-equilibrium fields will
therefore be the closed time path (CTP) action principle, taking the action $S$
for a neutral scalar boson and supplementing it by quadratic sources.
Observing the CTP operator ordering, one has
\begin{eqnarray}
S_+-S_- \rightarrow S_+-S_- &+&\frac{1}{2}\int dV_x dV_{x'} \lbrack
T(\phi_+(x)A_{++}(x,x')\phi_+(x')) \nonumber\\
&+& \phi_-(x)A_{-+}(x,x')\phi_+(x')\nonumber\\
&+& \phi_-(x')A_{+-}(x,x')\phi_+(x)\nonumber\\
&+& T^\dagger(\phi_-(x)A_{--}(x,x')\phi_-(x'))
\rbrack
\label{eq:72}
\end{eqnarray}
It should be clear that no fundamental field theory may contain
off-diagonal terms in $\pm$-space. The CTP action principle is, by
construction,
diagonal, being the difference between $S_+$ and $S_-$ (see equation
(\ref{eq:19})).
However, it was remarked earlier that the effect of a density matrix must
be reflected in off-diagonal terms so, while such terms are certainly
not fundamental, they can exist as off-diagonal self-energies representing
the dynamics of a density matrix. Moreover, since off-diagonal terms represent
a point of contact between fields moving forward in time and fields moving
backward in time, one might anticipate that off-diagonal sources would be
at least partly responsible for choosing an arrow of time (the generation
of entropy). The explicit coupling will therefore play an important role
in both non-equilibrium kinematics and dynamics.
The essential unitarity of the CTP formalism is seen, from equations
(\ref{eq:15})
and (\ref{eq:16}), to be summarized by the
following property of the transformation function:
\begin{equation}
\langle t_i | t_i\rangle^*_{\pm} = \langle t_i | t_i\rangle_{\mp}\label{eq:73}
\end{equation}
namely that complex conjugation merely exchanges $+$ labels with $-$ labels
and vice versa.
If one defines indices $a,b = +,-$, then the operator defined by the
second variation of (\ref{eq:72}) with respect to the field $\phi_a$,
$S_{ab}=\delta_a\delta_b(S_+-S_-)$, with
$S_{ab} = S_{++}, S_{+-}\ldots$, satisfies the relations
\begin{eqnarray}
S_{++}^*(x,x') &=& -S_{--}(x,x')\nonumber\\
S_{+-}^*(x,x') &=& - S{-+}(x,x').
\label{eq:74}
\end{eqnarray}
This, in turn, implies that $S_{ab}$ may be written
in terms of real constants $A,B,C$ and $\gamma_\mu$.
\begin{eqnarray}
S_{ab}(x,x') =
\left(
\begin{array}{cc}
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}+m^2)\delta(x,x')+A(x,x')+iC(x,x') & B(x,x') + \gamma^\mu(x,x')
\stackrel{\leftrightarrow}{D^\gamma_{\mu}}' -iC(x,x')\\
-B(x,x') - \gamma^\mu(x,x') \stackrel{\leftrightarrow}{D^\gamma_{\mu}}'
-iC(x,x') & ({\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}-m^2)\delta(x,x')-A(x,x')+iC(x,x')
\end{array}
\right)
\label{eq:75}
\end{eqnarray}
where a new derivative has been defined to commute with the function
$\gamma^\mu(x,x')$:
\begin{equation}
\stackrel{x}{D_\mu^\gamma}~ \equiv ~\stackrel{x}{\partial}_\mu +
\frac{1}{2}\gamma_\nu^{-1}(x,x')\stackrel{x}{\partial}_\mu\gamma_\nu(x,x').
\label{eq:76}
\end{equation}
Note, first of all, that the sum of rows and columns in this operator
is zero, as required for unitarity and subsequent causality.
Derivatives higher than first order in the sources could be rewritten
using the field equations (still to be found) and absorbed into other
terms, thus such terms are redundant. There can be no non-vanishing
terms of the form $\phi_+\partial\phi_+$ without violating time
reversal invariance or merely adding total derivatives to the
action. Finally $C\not=0$ is clearly disallowed in a fundamental
theory on the grounds of unitarity. It turns out, by considering the
field equations, that the only fully consistent choice is $C=0$, even
though such a term does not violate equation (\ref{eq:73}). Equation
(\ref{eq:75}) agrees with the form given by Lawrie, up to differences
in conventions and the inclusion here of $B(x,x')$.
The significance of the off-diagonal terms involving $\gamma^\mu$ can be seen
by writing out the coupling fully:
\begin{equation}
\gamma^\mu(x,x') \cdot \left( \phi_1 D^\gamma_\mu \phi_2 - \phi_2 D^\gamma_\mu
\phi_1\right).
\label{eq:gamma}
\end{equation}
The term in parentheses has the form of a current between components $\phi_1$
(the the forward moving field) and $\phi_2$ (the backward moving field).
When these two are in equilibrium there will be no dissipation to the
external reservoir and these
off-diagonal terms will vanish. This indicates that these off-diagonal
components
(which are related to off-diagonal density matrix elements, as noted earlier)
can be
understood as the mediators of a detailed balance condition for the field.
A similar conclusion was reached in reference \cite{calzetta1} by a different
argument for the quantity referred to here as $\tilde B$. When the term is
non-vanishing, it represents a current flowing in one particular direction,
pointing out the arrow of time for either positive or negative frequencies.
The current is a `canonical current' and is clearly related to the fundamental
commutator for the scalar field in the limit $+\rightarrow -$.
Using equation (\ref{eq:75}) it is now possible to express (\ref{eq:72})
in the form
\begin{equation}
S_{CTP} = \int dV_x dV_{x'} \frac{1}{2}\phi^a S_{ab} \phi^b
\label{eq:77}
\end{equation}
and thus the closed time path field equations may be found by varying
this action with respect to the $+$ and $-$ fields:
\begin{eqnarray}
\frac{\delta S_{CTP}}{\delta\phi_+(x)} &=& (-{\vcenter{\vbox{\hrule
height.4pt\hbox{\vrule width.4pt height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}+m^2)\phi_+(x)
+\frac{1}{2}\int dV_{x'}
\Bigg\lbrace
(\overline A+i\overline C)\phi_+(x') + (\tilde{B} -i\overline C)\phi_-(x')
+\overline \gamma^\mu(\partial_\mu\phi_-(x'))
+\overline{\partial^\mu\gamma_\mu}\phi_-(x')
\Bigg\rbrace\label{eq:78}\\
\frac{\delta S_{CTP}}{\delta\phi_-(x)} &=& ({\vcenter{\vbox{\hrule
height.4pt\hbox{\vrule width.4pt height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}-m^2)\phi_-(x)
-\frac{1}{2}\int dV_{x'}
\Bigg\lbrace
(\overline A-i\overline C)\phi_+(x') + (\tilde{B} +i\overline C)\phi_-(x')
+\overline \gamma^\mu(\partial_\mu\phi_+(x'))
+\overline{\partial^\mu\gamma_\mu}\phi_+(x')
\Bigg\rbrace\label{eq:79}\\
\end{eqnarray}
and setting the right hand side to zero,
which introduces the notation
\begin{eqnarray}
\overline A(x,x') &=& \frac{1}{2}\left(A(x,x')+A(x',x) \right)\nonumber\\
\tilde{A}(x,x') &=& \frac{1}{2}\left(A(x,x')-A(x',x) \right)\nonumber\\
\overline{\partial^\mu\gamma_\mu}(x,x') &=& \frac{1}{2}\left(
\stackrel{x}{\partial^\mu}
\gamma_\mu(x,x')+ \stackrel{x'}{\partial^\mu}
\gamma_\mu(x',x)
\right)
\label{eq:80}.
\end{eqnarray}
The canonical commutation relations which derive from the action
principle (see equation (\ref{eq:9})) are unchanged by these
modifications, since they depend only on $\overline x$
and therefore cancel in the commutator. This is key feature in any
consistent description of non-equilibrium phenomena.
To solve this system of non-local equations, the best strategy will
be to look for the Green functions, or the inverse of the operator
$S_{ab}$. This is the method adopted by Lawrie\cite{lawrie1}.
Although a common strategy will be used here, the method will
be somewhat different in spirit. The variational approach
used in ref. \cite{calzetta1} will not be used here.
Owing to the non-locality, it is clear that the inverse
operator cannot be a translationally invariant function in the
general case. It must be formally dependent on both Cartesian
coordinate differences and the average coordinate:
\begin{eqnarray}
\tilde{x} \equiv \frac{1}{2}(x-x')\label{eq:81}\\
\overline{x} \equiv \frac{1}{2}(x+x')\label{eq:82}.
\end{eqnarray}
Moreover, since the operator contains off-diagonal terms,
which typically signify a non-trivial density matrix, it
is natural to look for a solution based on the form of
equation (\ref{eq:69}), generalized to include a dependence
on $\overline x$. Although this sounds like an
innocent step, it is far from a trivial undertaking since it
introduces non-linearities in the spectrum of excitations which
must be handled in a self-consistent way. It is useful to
work with the quantity $H(x,x')$, from which all the Green
functions can be obtained (actually the Wightman
function in disguise). Using either the field equations or
the matrix equation
\begin{equation}
S_{ab}G^{bc} = \delta_a^{~c}\delta(x,x')\label{eq:83}
\end{equation}
one obtains equations of motion for the quantity $H(x,x')$ (see
equation (\ref{eq:50})). Not all of these equations are independent,
owing to the symmetry in equation (\ref{eq:73}). In particular, their
consistency requires that $C=0$ which is now chosen explicitly. It is
sufficient to consider
\begin{equation}
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}+m^2)H(x,x') + \int
dV_{x''} \left(\overline A(x,x'')
-\tilde{B}(x,x')-\overline\gamma^\mu(x,x'')\stackrel{x''}{\partial}_\mu
-\overline{\partial_\mu\gamma^\mu}(x,x'')\right)H(x'',x) = 0
\label{eq:84}
\end{equation}
on the understanding that $H(x,x') = H(x',x)^*$.
This relation is to be supplemented by the canonical
commutation relations for the field, which appear in
equation (\ref{eq:83}) in the form
\begin{equation}
\partial_{t'} \left( H(x,x')-H(x',x) \right)\Bigg|_{t=t'}=i\delta({\bf x},{\bf
x'})
\label{eq:85}
\end{equation}
and complete the consistency of the equations of motion.
The next step in obtaining an intuitive formalism is to introduce
a (local) momentum space technique by Fourier transforming
$\tilde x$ and retaining a dependence on the average coordinate
$\overline x$:
\begin{equation}
H(x,x') = \int \frac{d^nk}{(2\pi)^n} e^{ik(x-x')}H(k,\overline x)\label{eq:86}.
\end{equation}
A suitable ansatz for this function, which generalizes the
dispersion relation and the one-particle distribution function $f(k_0)$
is
\begin{equation}
H(k,\overline x) = 2\pi \theta(k_0)\lbrack 1+f(k_0,\overline x)\rbrack
\delta(-k_0^2+\omega^2(k,\overline x))
\label{eq:87}.
\end{equation}
The spacetime dependent function $f(k,\overline x)$ is often referred to
as the Wigner function and signifies the inhomogeneity in particle
occupation numbers.
The generalized dispersion relation takes generic form
$-k_0^2+\omega^2 =0$. In the free particle limit $\omega^2={\bf k}^2+m^2$.
It is the determination of this dispersion relation which is
of specific importance, since this determines the spectrum of
excitations for particles in the plasma-field, and forms the
basis of all perturbation theory when the sources represent
self-interactions.
It can be verified that, since $H(k,\overline x)$ depends only on
the average coordinate, the commutation relations are preserved (see equation
(\ref{eq:85})).
A more useful form of (\ref{eq:86}) is obtained on performing the integration
over $k_0$. This eliminates the dubious derivative of the delta-function
from subsequent relations and leads to a number of helpful insights.
\begin{equation}
H(x,x') = 2\pi \int \frac{d^{n-1}k}{(2\pi)^{n-1}}e^{ik_\mu(x-x')^\mu}
\frac{(1+f(k_0,\overline x))}{2|\omega|}
\label{eq:88}
\end{equation}
where it is understood that $k_0=|\omega|$.
Finally, it is useful to define the derivative with respect
to the average coordinate $\overline \partial =
\frac{1}{2}(\partial_x+\partial_{x'})$
and the quantities
\begin{eqnarray}
F_\mu &=& \frac{\partial_\mu f}{1+f} = \frac{1}{2}\overline F_\mu =
\frac{1}{2}\overline \partial_\mu \ln(1+f)\label{eq:89}\\
\Omega_\mu &=& \frac{\partial_\mu\omega}{\omega} =
\frac{1}{2}\overline\Omega_\mu =
\frac{1}{2}\overline\partial_\mu\ln|\omega|.\label{eq:90}
\end{eqnarray}
\section{Dispersion relations}
To determine the dispersion relations for given sources it is useful
to distinguish three cases which will be referred to as the local,
translationally invariant and inhomogeneous cases respectively. In the
local case, the sources are proportional to a delta function. In the
translationally invariant case $A(x,x')=A(x-x')$ and in the inhomogeneous
case $A(x,x')=A(\tilde x,\overline x)$.
There are two ways in which one can proceed with the determination of
the dispersion relations. One is to separate real and imaginary parts
and the other is to used complexified momenta. The latter has several
advantages and makes straightforward contact with the classical theory
of normal modes. It will be used exclusively for determining the
spectral relations. Separating real and imaginary parts on the other
hand allows one to identify imaginary contributions as a
Boltzmann/Vlasov equation, illustrating nicely the intimate
relationship between transport and dissipation\cite{calzetta2}.
In order to extract information from the equations it is necessary
to undertake an approximation scheme in which only low order
derivatives are kept in $x$. This is equivalent to an adiabatic
(or quasi-static) scheme in which the development of the system
if slow in comparison to fluctuations, so that fast and slow moving
variables separate in an assumed way. In fact this is already
built into the assumed form of the solution for the Green function,
since without such an assumption, there are no grounds for assuming
that $\tilde x$ and $\overline x$ would separate in the prescribed
manner. For most purposes this approximation is quite sound. For
the present, there seems to be no way of eliminating the approximation.
\subsection{Local sources}
In the local case, the equation satisfied by $H(x,x')$ is
\begin{equation}
\lbrack-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}+m^2+\overline A( x)
-\overline{\partial^\mu\gamma_\mu}
-\overline\gamma^\mu\partial_\mu \rbrack=0.
\label{eq:91}
\end{equation}
Note that, since $\tilde B$ is an odd quantity, it does not
appear in the local limit.
Since one is interested in the variables $x-x'$ and $\overline x$, it
is convenient to Taylor expand $x$ around $\overline x$.
Under the Fourier transform this takes the appearance
\begin{equation}
\overline A(x)H(x,x') \rightarrow \lbrack \overline A(\overline x)
+\frac{i}{2}(\overline \partial_\mu \overline A)\frac{\partial}{\partial k_\mu}
+\ldots\rbrack H(k,\overline x)
\label{eq:92}.
\end{equation}
It is useful to define a new quantity by
\begin{equation}
T_\mu = \frac{\frac{\partial f}{\partial k}}{(1+f)}
\label{eq:93}
\end{equation}
(the steepness of the spectral envelope for the Wigner function)
so that
\begin{equation}
\frac{1}{H}\frac{\partial H}{\partial k_\mu} = T^\mu - v_g^\mu/\omega
\label{eq:94}
\end{equation}
where $v_g=\frac{\partial \omega}{\partial k}$ is the group velocity of the
dispersing wave-packets.
In terms of the quantities (\ref{eq:89}) and (\ref{eq:90})
the action of the spacetime derivative operator on $H(x,x')$ gives
\begin{equation}
\stackrel{x}{\partial}_\mu H(x,x')= 2\pi \int \frac{d^{n-1}k}{(2\pi)^{n-1}}
e^{ik(x-x')}\frac{(1+f)}{2|\omega|}\lbrack ik_\mu+F_\mu-\Omega_\mu\rbrack
\label{eq:95}
\end{equation}
and subsequent derivatives are obtained in a straightforward way.
Substituting $H(x,x')$ into the equation of motion (\ref{eq:91}) now
leads to a second order differential equation for the
frequency $\omega(\overline x)$:
\begin{eqnarray}
\omega^2 &-&2(ik^\mu +F^\mu+\frac{1}{2}\overline\gamma^\mu)\Omega_\mu +
\Omega^2
= {\bf k}^2 +m^2 + \overline A(\overline
x)+\frac{i}{2}(\overline\partial_\mu\overline A)(T^\mu-v_g^\mu/\omega)
\nonumber\\
&-&\overline{\partial^\mu\gamma_\mu}-F^2 -2ik^\mu F_\mu
-\partial^\mu F_\mu -ik^\mu\overline\gamma_\mu - \gamma^\mu F_\mu
\label{insert}
\end{eqnarray}
Clearly this equation presents an insurmountable problem for the
purposes of analytic calculation, thus an approximation
must be made, based on the adiabatic evolution of the average
properties of the system. The lowest order case (which will
be sufficient to reveal the features of interest in this paper) is when
$\Omega_\mu$ and derivatives of $F_\mu$ may be neglected.
This corresponds to a near classical transport of particles, with relatively
few of the fluctuations added by the quantum nature of the field. With
this approximation the dispersion relation may be written:
\begin{equation}
k^2+m^2+\overline A(\overline x)
+\frac{i}{2}(\overline\partial_\mu\overline A)(T^\mu-v_g^\mu/\omega)
-\overline{\partial^\mu\gamma_\mu} -F^2 -2ik^\mu F_\mu -i\gamma^\mu k_\mu
-\gamma^\mu F_\mu=0.
\label{eq:96}
\end{equation}
This can be separated into a more appealing form as the
dispersion relation for a damped oscillator array
\begin{equation}
-\omega^2 -i\Gamma\omega +\omega_0^2=R
\label{eq:97}
\end{equation}
where one identifies the natural frequency,
\begin{equation}
\omega_0^2 = {\bf k}^2 +m^2 -F^2
\label{eq:98}
\end{equation}
the decay constant,
\begin{equation}
\Gamma = -\frac{1}{2\omega}(\overline\partial_\mu\overline
A)(T^\mu-v_g^\mu/\omega)
+\frac{2}{\omega}k^\mu(F_\mu-\frac{1}{2}\gamma_\mu)
\label{eq:99}
\end{equation}
and force term
\begin{equation}
R = \gamma^\mu F_\mu+\overline{\partial^\mu\gamma_\mu}-\overline A(\overline
x).
\label{eq:100}
\end{equation}
One notices how the effective mass of the theory is reduced by the
gradient of the Wigner function $F_\mu$, indicating that rapid transport
could lead to a second order phase transition. This might also
lead to anomalous dispersion.
In a true linear oscillator array $R$, $\Gamma$ and $\omega_0$
would all be independent of the frequency $\omega$. In
equation (\ref{eq:99}) only the zeroth component of the last term is
independent
of $\omega$. This indicates that the decay/amplification of certain modes in
time
is oscillator-like, but that the spatial modes are multiplied by a factor of
${\bf k}/\omega$, the inverse phase velocity, which has a critical
value when $m/{\bf k}$ is a maximum. This signifies the effect which
a gap in the frequency spectrum can have in leading to anomalous dispersion
in the `plasma'. At high frequencies $\Gamma\rightarrow
k^0(F_0-\frac{1}{2}\gamma_0)$
and the system is oscillator-like. At low frequencies, damping is dominated
by the external potential $\overline A$ and by transport as one might
expect.
\subsection{Translationally-invariant sources}
In the translationally invariant case, all variables are a function of
$x-x'$. One may therefore fully Fourier transform the sources:
\begin{eqnarray}
\overline A(x-x') = \int \frac{d^{n}k}{(2\pi)^{n}}e^{ik_\mu(x-x')^\mu}
\overline A(k)\nonumber\\
\overline \gamma^\mu(x-x') = \int \frac{d^{n}k}{(2\pi)^{n}}e^{ik_\mu(x-x')^\mu}
\overline \gamma^\mu(k)\nonumber\\
\tilde B(x-x') = \int \frac{d^{n}k}{(2\pi)^{n}}e^{ik_\mu(x-x')^\mu} i\tilde
B(k)
\label{eq:101}.
\end{eqnarray}
Note that, since $\tilde B$ is an antisymmetric function a factor of $i$
is introduced to keep $\tilde B(k)$ real.
The equation of motion for $H(x,x')$ is now
\begin{equation}
\lbrack-\stackrel{x}{{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}}+m^2 \rbrack H(x,x') +
\int dV_{x''}
\lbrack\overline A(x-x'') -\tilde B(x-x')
-\overline\gamma^\mu(x-x')\stackrel{x''}{\partial}_\mu
- (\overline{\partial^\mu\gamma_\mu})\rbrack H(x'',x')=0.
\label{eq:102}
\end{equation}
The translational invariance enables the latter spacetime integral to be
performed immediately, yielding the dispersion relation
\begin{equation}
k^2+m^2+\overline A(k) -2ik^\mu\gamma_\mu -i\tilde B(k) = 0.
\label{eq:103}
\end{equation}
An apparent consequence of the translational invariance is that $F_\mu=0$
owing to the steady state nature of the system. Comparing the dispersion
relation
to equation (\ref{eq:97}), one identifies
\begin{eqnarray}
\Gamma &=& \frac{2}{\omega}k^\mu\gamma_\mu + \tilde B(k)\nonumber\\
\omega_0^2 &=& {\bf k}^2+m^2\nonumber\\
R &=& - \overline A(k)
\label{eq:104}
\end{eqnarray}
Although the translationally invariant theory describes only
steady state disequilibria, it is nevertheless seen that the
field oscillations are concentrated around the usual mass-shell
$\omega_0^2$ with an amplitude driven by the external force
\begin{equation}
\frac{\overline A(k)}{\lbrack(\omega^2-\omega_0^2)^2
+(\Gamma\omega)^2\rbrack^{\frac{1}{2}}}
\label{eq:105}
\end{equation}
and a quality factor $Q=\omega_0/\Gamma$. Such a steady-state description
would be appropriate for an `infinite laser' i.e. a device which is not
affected by any finite size considerations.
\subsection{Inhomogeneous sources}
The main case of interest is when the sources and Green functions have a
residual dependence on the average position and time. This includes
the local limit as a special case:
\begin{eqnarray}
\overline A(x,x') &=& \alpha(x-x')\beta(x+x')\nonumber\\
\alpha &\rightarrow& \delta(x-x')\nonumber\\
\beta &\rightarrow& \overline A(x) = \overline A(x').
\label{eq:106}
\end{eqnarray}
As usual, one is looking for the eigenspectrum of the quadratic
operator acting on $H(x,x')$. The equation satisfied
by $H(x,x')$ is now:
\begin{equation}
\lbrack-\stackrel{x}{{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}}+m^2 \rbrack H(x,x')
+\int dV_{x''}\left[ \overline A(x,x'')-\tilde B(x,x'')-\overline
\gamma^\mu(x,x'')
\stackrel{x''}{\partial_\mu} - \overline{\partial^\mu\gamma_\mu}\right]
H(x'',x')=0
\label{eq:new}
\end{equation}
In the inhomogeneous case there is no dispersion relation
consisting of continuous frequencies in general so the dispersion
relation will only exist for a discrete set. It is convenient to
divide the discussion into two parts: the determination of the
dispersion relation and the nature of the restricted set of values
which satisfy the dispersion relation.
The problem to be addressed is contained in
following form in momentum space:
\begin{equation}
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}+m^2)H(x,x')+\int
dV_{x''} \frac{d^nk}{(2\pi)^n}\frac{d^np}{(2\pi)^n}
e^{ik(x-x'')+ip(x''-x')}
S(p,x''+x')H(k,x+x') = \lambda H(x,x').
\label{eq:107}
\end{equation}
The integral over $x''$ is no longer a known quantity in general, but
it is possible to extract an overall Fourier transform
by shifting the momentum $p\rightarrow p+k$ and defining
the average variable of interest $\overline x=\frac{1}{2}(x+x')$:
\begin{eqnarray}
(k^2 &+&
ik^\mu\overline\partial_\mu-\frac{1}{4}\overline{{\vcenter{\vbox{\hrule
height.4pt\hbox{\vrule width.4pt height8pt\kern8pt\vrule width.4pt}\hrule
height.4pt}}}} +m^2)H(k,\overline x)+\nonumber\\
&+&\int dV_{x''}\frac{d^np}{(2\pi)^n} S(k,x+x'')H(k+p,x''+x') = \lambda
H(k,\overline x)
\label{eq:108}.
\end{eqnarray}
In order to find eigenvalues, it is necessary to extract the
factor of $H(k,\overline x)$ from this expression.
This is not possible for arbitrary values of $k$. It is possible,
however,
if the momenta are restricted to a denumerable set expressed by
the property
\begin{equation}
H(k+p,x''+x') = H(k,x''+x')
\label{eq:109}
\end{equation}
which implies that $H(k,\overline x)$ is a periodic function of the
momenta.
The absence of eigenvalues or the failure of this property
leads to the consideration of an infinite iterative mapping
of states, which---in the absence of a stable limit---suggests chaotic
excitations of
the field. This can also be argued geometrically (see the final
section).
Given this mitigating condition, one has
\begin{equation}
\int dV_{x''}\frac{d^np}{(2\pi)^n} S(k,x+x'')H(k+p,x''+x')
= S(k,\overline x)H(k,\overline x).
\label{eq:110}
\end{equation}
The dispersion relation is now obtained in a straightforward
fashion, adopting the same adiabatic approximation
as before, and is given by the implicit relation
\begin{eqnarray}
k^2&+&m^2+\overline A(k,\overline x) +\frac{i}{2}(\overline \partial_\mu
\overline A)(T^\mu-v_g^\mu/\omega)
-i\tilde{B}(k)\nonumber\\
&-& \overline{\partial^\mu\gamma_\mu}(k,\overline x)
-(F-N)^2 -2ik^\mu(F-N)_\mu -2i\overline \gamma^\mu k_\mu -\overline \gamma^\mu
(F-N)_\mu=0.
\label{eq:111}
\end{eqnarray}
where it is noted that $\lbrace k\rbrace$ is now discontinuous.
Note that the antisymmetry of $\tilde B(k)$ makes it independent of $\overline
x$.
Comparing to the oscillator equation, one has
\begin{eqnarray}
\Gamma &=& -\frac{1}{2\omega}(\overline \partial_\mu\overline
A)(T^\mu-v_g^\mu/\omega)
+\frac{2}{\omega}k^\mu(F_\mu-N_\mu+\overline \gamma_\mu)+\tilde B(k)\nonumber\\
\omega_0^2 &=& {\bf k}^2 +m^2 -(F-N)^2\nonumber\\
R &=& \overline \gamma^\mu F_\mu +
\overline{\partial^\mu\gamma_\mu}(k,\overline x) -\overline A(k,\overline x)
\label{eq:112}
\end{eqnarray}
where $N_\mu$ will be defined presently.
We now turn to the consequences of the condition in equation (\ref{eq:109}).
There are various precedents for such a relation: one is Green functions
defined on a torus (finite temperature, Matsubara formalism,
electron band structure); another is the case of Landau levels on a torus.
The periodicity is clearly the important factor here. In most
of these cases the periodicity is one in real space and the result
is a discrete spectrum of eigenvalues. Here the periodicity lies in
the momentum itself. In fact the two notions are closely related and
a periodic system in real space has Green functions which are periodic
in momentum space owing to an infinite summation over discrete frequencies
(which is therefore invariant under shifts by a whole number of
periods). The relation (\ref{eq:109}) must be satisfied for all legal
values of the momentum, thus the implication is that the system is
degenerate---i.e. there exist bands of energy which leave the
Green function invariant under certain shifts. These need not all
refer to the same band.
It is therefore possible to write
\begin{equation}
H(k) = H\left(\sum_l \frac{2\pi l^\mu k_\mu}{P_\mu}\right)
\label{eq:113}
\end{equation}
where $P_\mu$ is the {\em momentum} periodicity length (which has dimensions
of inverse space-length).
This finite length must diverge to infinity when the inhomogeneities
vanish. There is only one natural momentum/length scale which has these
properties, namely
\begin{equation}
L_\mu = P_\mu^{-1} = \overline\partial_\mu H(k,\overline x).
\label{eq:114}
\end{equation}
In deriving (\ref{eq:110}) we have used the fact that
\begin{equation}
\int \frac{d^nk}{(2\pi)^n} \exp(ikx)=\delta(x)
\label{eq:115}
\end{equation}
Since $k$ is now restricted to a discrete set, the correctness
of this relation could now be an issue. It can easily be
verified using the formulae
\begin{eqnarray}
\sum_{k=1}^{n} \sin(kx) &=& \sin\frac{(n+1)}{2}\sin\frac{(nx)}{2}{\rm
cosec}\frac{x}{2}\nonumber\\
\sum_{k=0}^{n} \cos(kx) &=& \sin\frac{(n+1)}{2}\cos\frac{(nx)}{2}{\rm
cosec}\frac{x}{2}
\label{eq:116}
\end{eqnarray}
that an extra finite {\em imaginary} contribution can arise from the
discrete nature of the spectrum, which vanishes in the continuous
limit. It will be assumed that such a contribution can be absorbed by
a redefinition of the sources.
Although one is looking at periodic functions, the solution for $H(x,x')$
need not be sinusoidal. In the case of Landau levels on the torus
\cite{laughlin1,burgess6} periodicity is only achieved at the expense of
a flux-quantization condition which, again, involves a degeneracy of
solutions. There is, in fact, an analogy to this situation here.
The extraordinary properties of Landau levels on a torus can be attributed
to the non-translational invariance of the electro-magnetic vector potential.
The similarity here is the non-translational invariance of the many-particle
state as expressed by the dependence on $\overline x$. This point will be
discussed at greater length in the final section, to avoid its meaning
being lost in the present analysis.
The extra terms containing $N_\mu$ can now be explained. They arise from the
$\overline x$ dependence of the momentum space measure:
\begin{equation}
\int \frac{d^nk}{(2\pi)^n} \rightarrow \prod_\mu
\left(\frac{1}{L_\mu}\sum_{l_\mu} \right)
\label{eq:117}
\end{equation}
giving a contribution
\begin{equation}
N_\mu = \frac{\partial_\mu(L_0\ldots L_{n-1})}{(L_0\ldots L_{n-1})}
\label{eq:118}
\end{equation}
which compounds the non-linearity.
The above restrictions have no special consequences for the Feynman
propagator, since the nature of the momentum is not used to obtain it.
This is gratifying since the Feynman propagator must always be the
literal inverse of the quadratic part of the $+$ time-ordered action. Only the
nature of the singularity is altered in accord with the modified
dispersion:
\begin{equation}
G_F(k) = \frac{1}{-k_0^2+\omega^2-i\epsilon} + 2\pi i f(k,\overline
x)\delta(-k_0^2+\omega^2)
\theta(k_0).
\label{eq:119}
\end{equation}
The appearance of a natural length scale, connected to the inhomogeneities
of a non-equilibrium system, is an important feature for two reasons.
Firstly, the spontaneous generation of a length scale implies the
possibility of domain formation, or a cellular localization in the field.
Secondly, the dependence of the Green functions on themselves implies that
the stable solutions of the system can be regarded as fixed points of
an iterative map. Such maps have been studied in connection with
classical chaotic systems\cite{lauwerier1}.
In the present case, the function $H(k,\overline x)$ depends not
merely on itself but on its derivative. For exponential-like solutions
one could expect that this would amount to the same thing, up to
a constant multiplier. The situation would then be something
akin to $H=H(\lambda H)$, for some constant $\lambda$. This bears a
noteworthy similarity to Feigenbaum's functional equation which
can be written
\begin{equation}
g(x) = \alpha g(g(x/\alpha))
\label{eq:120}
\end{equation}
subject to a boundary value, or rewriting:
\begin{equation}
g(g(\lambda x)) = \lambda g(x).
\label{eq:121}
\end{equation}
This equation has an analytic solution as a power series
\begin{equation}
g(z) = 1 + c_1z^2 + c_2 z^4 + \ldots
\label{eq:122}
\end{equation}
where a limiting value is approached through a geometric
progression with Feigenbaum ration ${\cal F}=4.66$ and
universal scaling factor $\alpha=-2.5$.
Solutions to this equation which fall outside the
fixed point behaviour can be expected to lead to chaotic
behaviour. This strongly suggests that the non-equilibrium
Green functions must exhibit universal behaviour or chaos
in their approach to stable behaviour. In other words, the
approach to equilibrium need not be of the simple damped
or over-damped form of a linear oscillator array---it could
easily entail a chaotic attractor.
\section{Entropy, temperature and the KMS condition}
For systems close enough to a thermal state, it is possible to
define an approximate temperature and entropy. The entropy of the system may
be defined in various ways, often based only on combinatorial
considerations of the micro-canonical picture. Here it is
convenient to define an `oscillator effective entropy' which
is easily related to quantities which arise in the
analysis.
Suppose the
Wigner function is given by the approximate equilibrium form
\begin{equation}
f(k_0,\overline x) = \left( \exp(\beta(\overline x)\omega(\overline
x))-1\right)^{-1}
\label{eq:123}
\end{equation}
then one has
\begin{equation}
\overline F_\mu = - f^2 \left[ \left(\frac{\overline\partial_\mu\beta}{\beta}
+\overline\Omega_\mu\right)\beta\omega\right]
\label{eq:124}
\end{equation}
and, classically, the statistical entropy $S$ is
\begin{equation}
S = k \left( \ln Z + \beta\langle\omega \rangle\right).
\label{eq:125}
\end{equation}
For a harmonic oscillator, one has (see for example ref. \cite{reif1})
\begin{equation}
\ln Z = - {\rm Tr} (1-e^{-\beta\omega}) - \frac{1}{2}\beta\langle\omega\rangle
\label{eq:126},
\end{equation}
thus the oscillator entropy may be defined as
\begin{equation}
S = \frac{1}{2}\beta \langle\omega\rangle + {\rm Tr}\ln(1+f).
\label{eq:127}
\end{equation}
This motivates the definition of a simple measure of entropy for
the oscillator array, given by
\begin{equation}
S_{E}(\overline x) = \int \frac{d^nk}{(2\pi)^n}
\theta(k_0)\ln(1+f)\delta(-k_0^2+\omega^2).
\label{eq:128}
\end{equation}
The rate of change of this entropy is then
\begin{eqnarray}
\overline\partial_\mu S_{E} &=&\overline\partial_\mu
\int \frac{d^{n-1}k}{(2\pi)^{n-1}}
\frac{\ln(1+f)}{2|\omega|}\Bigg|_{k>0}\nonumber\\
&=& \int \frac{d^{n-1}k}{(2\pi)^{n-1}} \frac{(1+f)}{2|\omega|}\lbrack\overline
F_\mu-\overline\Omega_\mu \rbrack
\label{eq:129}
\end{eqnarray}
This quantity can be compared to (\ref{eq:95}). It shows that the entropy
gradient
can be thought of as a `connection' for the field modes. The generation of
entropy is therefore fundamentally connected with the flow of particle
occupation numbers and the `downgrading' of the frequency spectrum---i.e.
the rate at which energy becomes unavailable to do work.
As mentioned earlier, the effect of a non-trivial density matrix, either at
the initial time or later, is reflected in the off-diagonal sources and Green
functions. If one imagines that the sources $A_{\pm\pm}$ arise from a
coupling to another oscillator system\cite{schwinger2} or that they represent
the self-interaction of the field to order $\phi^4$, then $A_{\pm\pm}$ is
essentially the Green function for the field concerned and one would therefore
expect the KMS condition to hold for the sources at equilibrium---now in the
form
\begin{equation}
\theta(|\omega|)A_{+-}(\omega) = e^{\beta\omega}
\theta(-|\omega|)A_{-+}(\omega).
\label{eq:130}
\end{equation}
This condition does not hold in general, but for an
isoentropic process, in terms of the defined quantities at $\gamma^\mu=0$, one
therefore has
\begin{equation}
\tilde B(\omega) = - e^{\beta\omega} \tilde B(\omega)
\label{eq:131}
\end{equation}
It is verified that
\begin{equation}
\frac{\theta(\omega)A_{+-}(\omega)}{\theta(-\omega)A_{-+}(\omega)} =
e^{\beta|\omega|}
\label{eq:132}
\end{equation}
giving $A_{+-}=\sinh(\frac{1}{2}\beta|\omega|)a(\omega)$ for some $a(\omega)$
or
\begin{equation}
B(\omega) = \frac{1-e^{-\beta|\omega|}}{1+e^{\beta|\omega|}} =
\tanh(\frac{1}{2}\beta|\omega|)
\label{eq:133}
\end{equation}
which agrees with Schwinger's result\cite{schwinger2}.
Note that the initial state $f(\overline x_i)$ and its subsequent
development enter only as boundary conditions to the Green functions
and the Wigner function. The changing form of $f(\overline x)$ is
determined solely by the sources $A_{\pm\pm}$. Thus, if the sources
do not evolve, neither does $f(\overline x)$ and neither does
the implicit density matrix. In the perturbation around free
field theory \cite{calzetta1}, $f(\overline x)$ always represents
the state of the system at the initial time.
In the approach to equilibrium one normally expects that dependence on the
average coordinate $\overline x$ to disappear. This is an expression of what is
often called `loss of memory' of the initial state, since $\overline x$ is
measured
relative to the initial time. An equilibrium state (thermal or otherwise)
is, by its nature, either static or periodic, thus the resulting Wigner
function $f(k_0,\overline x)$ must either be independent of $\overline x$ or
a periodic function of this variable. One of the advantages of the present
formulation is that one sees how the sources are responsible for this
loss of memory. Since the sources drive the system, $f(k_0,\overline x)$
can never become $\overline x$ independent as long as the sources are
$\overline x$ dependent. Thus equilibrium will only be secured by
accounting for the back-reaction of the sources to the behaviour of the field.
Explicit equations of motion for the sources have not been considered here.
An example of a periodic `equilibrium' is the case of Rabi oscillations in
the laser (see ref. \cite{firth1} for a review), in which the source and
the field enter into a pendulum-like flip-flop behaviour. An example
of this will be given in the final section.
The decay of field modes is exponential, per mode and is mediated by the
source $\gamma^\mu(x,x')$ and the gradient of the potential $A(x,x')$.
This does not preclude other behaviour for the Wigner function. For example,
in the simplest case close to equilibrium in which the system is quasi-static
and $\overline A = \tilde B = 0$, with almost no external force (see equation
(\ref{eq:112})),
one has $\gamma_\mu\sim F_\mu$ and thus $\partial^\mu F_\mu +F^2\sim0$ giving
$F_\mu \sim x_\mu^{-1}$--- a `long tail' power law decay which parallels
the decay of harmonic waves in curved spacetime\cite{ching1}.
\section{Calculation of expectation values}
The closed time path formalism codifies the causal relationship
between source and response, for the computation of expectation values
in a general mixed state. Since it is redundant except as a
calculational aid, it's introduction should be justified by an
example. The causality of the method is not affected by the
introduction of the sources $A_{\pm\pm}$, but the dissipative dynamics
are. Normally a fundamental Gaussian theory can never show
dissipation, but in the present situation one has sources which can
drive the field modes and redistribute energy.
There are two cases of interest.
In a self-interacting theory one might identify $A_{\pm\pm}$ with
the correlation function for the field itself $\lambda G_{\pm\pm}$,
giving rise to dispersion relation of the approximate form
\begin{equation}
k^2+m^2 + \lambda {\rm Tr} (k^2+m^2)^{-1} = 0.
\label{eq:134}
\end{equation}
This is like the variational method used in ref. \cite{calzetta1}.
Lawrie\cite{lawrie1} takes the view that the sources can effect a
renormalization of a self-interacting theory by choosing them in such
a way as to `minimize' the effect of higher order perturbative
contributions. In either case, the effective `resummation' induced by
the sources makes it possible to see damping of field modes at the
one-loop (Gaussian) level.
Consider the response of the field to the source $J(x)$, in the
presence of $A_{\pm\pm}$. One is interested in the causal expectation
value of the field at time $\overline t$, given the state of the
system at the initial time. The time dependence, in the present
formalism is now contained entirely within the sources---or
equivalently the dispersion relation. That the CTP generator leads to
a causal result is easily verified by realizing that the expectation
value of the field is always coupled to the sources by the retarded
$n$ point functions. For an arbitrary action $S[\phi]$,
\begin{eqnarray}
\langle\phi(x)\rangle &=& -i\frac{\delta}{\delta J_+(x)}\Bigg|_{+=-} \langle
0|0 \rangle_{\pm}\nonumber\\
&=& \frac{1}{2}\int dV_{x'} \lbrack
2G_{++}(x,x')+G_{+-}(x,x')+G_{-+}(x,x')\rbrack J(x') +\ldots\nonumber\\
&=& \frac{1}{2}\int dV_{x'} \lbrack 2G_{++}(x,x')+G^{(+)}-G^{(-)}\rbrack J
+\ldots\nonumber\\
&=& \int dV_{x'} \lbrack G_{++}(x,x')-G^{(-)}\rbrack J(x') +\ldots\nonumber\\
&=& \int dV_{x'} G_{ret}(x,x') J(x')+\ldots
\label{eq:135}
\end{eqnarray}
thus the expectation value depends only on retarded times.
Furthermore, the result is real (being a probability) since the retarded Green
function
is explicitly the real part of the Wightman functions, restricted
to retarded times by a step function:
\begin{equation}
G_{ret}(x,x') = - \theta(t-t') \lbrack G^{(+)}(x,x') +G^{(+)*}(x,x') \rbrack.
\label{eq:136}
\end{equation}
Making use of the integral representation (\ref{eq:55}), one has
\begin{equation}
G_{ret}(x,x')= -i \int \frac{d\omega}{(2\pi)}\frac{d^nk}{(2\pi)^n}
\frac{\exp(-i\omega(t-t')+i{\bf k}({\bf x}-{\bf x'}))}{(\omega+i\epsilon)}
\left( G^{(+)}(k)+G^{(-)}(k)\right).
\label{eq:137}
\end{equation}
Relabelling and inserting the momentum-space forms for the
Wightman functions from (\ref{eq:69}), one has
\begin{equation}
G_{ret}(k)= - \int d\omega \frac{1}{k_0-\omega+i\epsilon}
\left( \frac{1}{\omega_+}\delta(\omega-\omega_+)-
\frac{1}{\omega_-}\delta(\omega+\omega_-)\right)
\label{eq:138}
\end{equation}
where $\omega_+$ and $-\omega_-$ are the positive and negative frequency
solutions
to the appropriate dispersion relation. These are complex numbers in general,
owing
to the non-vanishing imaginary part labelled as $\Gamma$. Now, since
unitarity demands that $G^{(+)}(x,x')$ be the complex conjugate of
$G^{(-)}(x,x')$,
it is clear that
\begin{equation}
\omega_+^* = \omega_-.
\label{eq:139}
\end{equation}
It is assumed here that the dispersion relation has two complex roots.
The quantity appearing in the delta function in equation (\ref{eq:87})
is then $-k_0^2+\omega_+\omega_-$ which may also be written
$-k_0^2+\omega^*\omega$.
To avoid confusion with previous notation for the absolute value, the complex
modulus will not be denoted $|\omega|$. This indicates that, in spite of the
complex momenta in the dispersion relation, whose role it is to capture
dissipation and transport/kinetic effects, the
`mass shell' constraint is real. The simplest expression
for the retarded Green function is therefore
\begin{equation}
G_{ret}(k)= - \left( \frac{1}{2\omega_+(k_0-\omega_++i\epsilon)}
- \frac{1}{2\omega_-(k_0-\omega_-+i\epsilon)}\right).
\label{eq:140}
\end{equation}
This expression is not manifestly real, since it is a momentum space
result. However, if one defines $2i\tilde\omega = \omega_--\omega_+$ and
$2\overline\omega=\omega_++\omega_-$, where $\tilde \omega$ and
$\overline\omega$ are real, then it is possible to write
\begin{equation}
G_{ret}(k)= \frac{1}{\omega^*\omega}\left(
\frac{(i\tilde\omega k_0 -\omega^*\omega
+2\overline\omega^2)(-k_0^2+\omega^*\omega - 4ik_0\tilde\omega)}
{(-k_0^2+\omega^*\omega)^2+16k_0^2\tilde\omega^2}
\right).
\label{eq:141}
\end{equation}
This may be compared to equation (\ref{eq:105}) and reduces to
\begin{equation}
\frac{1}{-k_0^2+\omega^2}
\end{equation}
when $\omega^* = \omega$ and $\epsilon\rightarrow 0$. Since the
imaginary part of (\ref{eq:141}) is odd with respect to the momentum
variable $k$, the Fourier transform back to configuration space is
real, as expected. The desired expectation value is therefore
manifestly real and causal, and the time dependence since the initial
time is contained entirely in the $\overline x$ dependence of the
frequency $\omega$.
\section{Reformulation}
In the preceding sections, it has been shown how dissipation and
amplification of spectral modes can be incorporated into the dispersion
of a quadratic theory, for suitably adiabatic processes. It is
now practical interest to show that the same results can be presented in
another significant form by introducing a `covariant derivative' $D_\mu$
which commutes with the average development of the field state.
This description parallels the structure of a gauge theory
(in momentum space) with an imaginary charge. Alternatively one may
speak of a generalized chemical potential for the `gauge' field.
Consider the derivative
\begin{equation}
D_\mu = \partial_\mu - a_\mu
\label{eq:200}
\end{equation}
and its square
\begin{equation}
D^2 = {\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}} - \partial^\mu a_\mu -
2a^\mu\partial_\mu + a^\mu a_\mu.
\label{eq:201}
\end{equation}
Without any approximation, it is straightforward to show that, in the
general inhomogeneous case,
\begin{eqnarray}
(-{\vcenter{\vbox{\hrule height.4pt\hbox{\vrule width.4pt
height8pt\kern8pt\vrule width.4pt}\hrule height.4pt}}}+m^2)H(x,x') = 2&\pi&
\int \frac{d^{n-1}k}{(2\pi)^{n-1}}\frac{(1+f)}{2|\omega|}
e^{ik(x-x')}\nonumber\\
\lbrace &-& (ik_\mu+F_\mu-\Omega_\mu-N_\mu)^2
-\partial^\mu(ik_\mu + F_\mu -\Omega_\mu-N_\mu)\rbrace.
\label{eq:202}
\end{eqnarray}
It is then natural to make the identification
\begin{eqnarray}
a_\mu &=& F_\mu - \Omega_\mu - N_\mu +\overline \gamma_\mu \nonumber\\
&=& \partial_\mu S_E(k) - N_\mu +\overline \gamma_\mu
\label{eq:203}
\end{eqnarray}
where the meaning of this notation is such that the expression only defined
when all
objects are under the momentum integration---this is to be understood
in all future expressions.
The field $a_\mu$ is clearly related to the rate of increase of the entropy
$S_E$, the damping factor $\gamma^\mu$ and the rarefaction of the localized
cells $N_\mu$.
One now has:
\begin{eqnarray}
(-D^2+m^2)H(x,x') &=& 2\pi \int
\frac{d^{n-1}k}{(2\pi)^{n-1}}\frac{(1+f)}{2|\omega|}
\lbrace - (ik_\mu -\overline\gamma_\mu)^2) -
\partial^\mu(ik_\mu-\overline\gamma_\mu)\rbrace
\nonumber\\
&=& 2\pi \int \frac{d^{n-1}k}{(2\pi)^{n-1}}\frac{(1+f)}{2|\omega|}
\lbrace k^2 + 2ik^\mu\overline\gamma_\mu -\overline\gamma^2 -i(\partial^\mu
k_\mu)
+(\partial^\mu \overline \gamma_\mu)\rbrace.
\label{eq:204}
\end{eqnarray}
Adding the appropriate source combinations for the inhomogeneous case one has,
without approximation, the differential equation satisfied by $H(x,x')$:
\begin{equation}
\left[ -D^2+m^2 +\overline\gamma^2(k,\overline x) +\overline A(k,\overline x)
- \tilde B(k)+\frac{i}{2}(\overline\partial_\mu\overline
A)(T^\mu-v_g^\mu/\omega) \right]_k H(x,x') = 0
\label{eq:205}
\end{equation}
where the appearence of the subscript $k$ to the
bracket serves to remind that the equation only exists under the
momentum integral. The local limit is simply
\begin{equation}
\lbrack -D^2+m^2-\overline\gamma^2(x) + \overline A(x) \rbrack_k H(x,x')=0.
\label{eq:206}
\end{equation}
The 'gauge' field $a_\mu$ couples via an imaginary unit-charge plays the
role of a generalized chemical potential on the manifold of positive
energy solutions for the real scalar field (the chemical potential has no
meaning
for the full field, since particle numbers are not conserved).
Suppose now that one defines the analogue of the field strength tensor
\begin{equation}
f_{\mu\nu} = \partial_\mu a_\nu - \partial_\nu a_\mu.
\label{eq:207}
\end{equation}
In many cases one will have $f_{\mu\nu}=0$, thus one can 'gauge transform' the
field, which maps
\begin{eqnarray}
\phi(k) &\rightarrow & \phi(k) e^{\int a_\mu dx^\mu}\nonumber\\
&=& \phi(k) e^{-\int
(F_\mu-\Omega_\mu-N_\mu-\overline\gamma_\mu)dx^\mu}\nonumber\\
&=& \phi(k) e^{-S_E-\int (N_\mu+\overline\gamma_\mu)dx^\mu}.
\label{eq:208}
\end{eqnarray}
This shows the explicit decay (amplification) of the $k$-th field mode.
The latter relation shows that this process involves an increase
in the effective oscillator entropy of the system.
In terms of the above formulation, the spectral content of the
bosonic theory reduces to the problem of finding the eigenvalues
of the operator $D^2$. In particular one can use the body of experience
gained in the study of gauge theories to attack the problem.
With an adiabatic approximation for $f(\overline x,k)$, $a_\mu$
has a series expansion in powers of $\overline x$. Thus for
quasi static systems
\begin{equation}
a_\mu = (c_0 + c_1 \overline x+\ldots)_\mu.
\label{eq:209}
\end{equation}
The effective field strength
$f_{\mu\nu}$ need not always be zero. Two situations might arise:
(i) the Wigner function might contain a logarithmic singularity, as in
the case where vortices are present, and (ii) the source $\overline\gamma_\mu$
could contain components which specifically drive the macroscopic field
in a given way. A simple example of the latter is the analogue of
Rabi oscillations in the laser, in which the field oscillates between
two states in a regular way. Here, this oscillation is driven by
the source $\overline \gamma_\mu$ or perhaps by a pulsation of the
inhomogeneity scale, and occurs from the linear terms in equation
(\ref{eq:209}).
The current $J=\phi_2 \stackrel{\leftrightarrow}{D}^\gamma \phi_1 $
behaves like a magnetic influence on the system (doing no net work).
Simplifying to the case of a ($1+1$) dimensional system, one may write
\begin{equation}
a_\mu \sim \overline\gamma_\mu = |\gamma| \epsilon_{\mu\nu}\overline x^\nu
\label{eq:210}
\end{equation}
for constant $|\gamma|$ and $\mu,\nu=0,1$.
This corresponds to a harmonic `flip-flop' motion between field
and source. It is also directly analogous to the well known problem
of Landau levels in an effective magnetic field $|\gamma|$.
The localization in spacetime resulting from the inhomogeneity
scale suggests that such oscillations may take place locally
in cellular regions. A simplified model for this is to impose
periodic boundary conditions on the cells, generating a
kind of global field coherence (this is admittedly
motivated by technical simplicity rather than physical reasoning).
One is therefore led to the study of Landau levels on the
torus---a system which has been studied at some
length\cite{burgess6,laughlin1},
and will not be re-analyzed here.
A significant feature of the Landau problem on the torus is that the
periodicity enforces a flux quantization condition on the field. Here
this translates into the following relation:
\begin{equation}
\overline \partial_0 H(k,\overline x) \overline \partial_1 H(k,\overline x)
| \gamma| = 2\pi n
\end{equation}
for integer $n$. This relation indicates that nearest neighbour
cells might engage in cooperative oscillations (i.e. the size
of cells is quantized in units of the local inhomogeneity scale).
This is clearly a far less stringent condition here than in the
case of a true periodic torus, since the inhomogeneity scale
varies in space and time thus the meaning of strict quantization
is lost. However, it indicates that one can expect a tiling
of spacetime by oscillation cells. Since the size of the cells
might be highly irregular, the tiling behaviour is most likely to be
chaotic unless special geometrical boundary conditions
can enforce a regularity on the field. This is an alternative
expression of the behaviour deduced from the Green function
in equation (\ref{eq:109}).
\section{Conclusion}
Schwinger's closed time path action principle has been applied to the
neutral scalar meson, off-equilibrium, in the presence of long-range,
inhomogeneous sources. The method of dispersion relations is used to
find formal expressions for Green functions which reflect the
absorbative and amplifying processes in the normal modes of the field.
In the case of self-interacting theories, the sources can be thought
of as representing $\phi^{2n}$ interactions to one-loop order, effecting
a resummation of the theory. The effect of rapid transport
(large $F_\mu$) is to induce a change in the sign of the mass
squared, indicating a second order phase transition and anomalous
dispersion.
If significant inhomogeneities or long range interactions exist, the
field naturally forms localized cells with (to lowest order) a
periodic relationship to the natural inhomogeneity length/time scale.
This is shown from the viewpoint of the Green functions and by
recourse to an analogy with Landau levels on the torus.
Since the length scale is determined by non-linear considerations
one can expect chaotic behaviour with islands of order (stable
fixed points) along the approach to equilibrium.
A simple analogue of Rabi oscillations in the laser is shown to
arise as a leading order behaviour in $\overline x$.
The method used in the this work has the advantage of combining the
fundamental aspects of an operator field theory with the usefulness of
the action principle. The use of generating functional ultimately
leads to functional integral forms, as used almost exclusively in the
literature. However, the introduction of the functional integral is
scarcely necessary using the present method and often has the
undesirable effect of turning the discussion of causality into one of
complicated paths of integration in the complex plane.
Comparing to other works reveals both differences and
similarities. Lawrie\cite{lawrie1}, for example, treats the quantity
$\gamma^\mu$ as an explicitly written imaginary part of the spectrum
of excitations. He ignores $F_\mu$, but does not ignore
$\Omega_\mu$. This is a somewhat different approximation which has a
more distant relationship to classical transport theory. In fact,
since the appearance of $F_\mu$ and $\Omega_\mu$ in $a_\mu$ is
identical, up to a sign, the form of dynamics might well be
independent of the approximation used in this work---understandable as
a reparameterization of an equivalent problem. Lawrie further
considers $\phi^4$ theory and uses a renormalization-like philosophy
to determine the sources self-consistently thereby effecting a
resummation as noted in equation (\ref{eq:134}). Calzetta and
Hu\cite{calzetta1} use a variational principle to determine the
effective action for a self-interacting boson theory. This is the same
idea as in ref. \cite{lawrie1}, expressed in extremely aesthetic
formalism and containing important insights into the subject; the
solution to their method is, in practice, more difficult to attain
however and thus results are mainly formal. Neither of these works
consider the implications of non-local effects. Another interesting
approach is the Schr\" odinger quantization approach in
ref. \cite{eboli1}. This makes a contact with the Schwinger action
principle at a more subtle level and, focusing on somewhat different
issues, uncovers features absent in other formulations of
non-equilibrium physics.
It is important to extend the present analysis to include
both fermions and spin-1 bosons (true gauge fields). The
latter is probably a difficult task in view of the
problems which can arise in gauge fixing. Again, the action principle
approach, starting from the operator field theory is likely
to be the most informative approach.
The appearance of discrete spectra and magnetic like effects
makes the present work very interesting to the study of the
fractional quantum Hall effect. In particular, the pseudo-gauge
field formulation might have interesting connections with the
statistical gauge field employed in the Chern-Simons gauge
theory picture. These and other outstanding issues will be discussed in
future work.
I am grateful to I.D. Lawrie for helpful discussions.
|
\section{Introduction}
The effects of the exchange interaction on the appearance
of macroscopic spin structures have been
studied in semiconductor microstructures in reduced dimensions
by several researchers both theoretically and in experiments.
The enhancement of the effective $g$-factor, $g^*$, of a
two-dimensional electron gas (2DEG) in the quantum Hall regime has been
reviewed by Ando, Fowler, and Stern.\cite{Ando82:437}
For the unbounded 2DEG
Ando and Uemura \cite{Ando74:1044} presented a model
where the broadening of the Landau levels due to
impurity scattering is treated in the self-consistent
Born approximation (SCBA). The dielectric function is calculated
with the inclusion of the lowest order exchange energy of the
screened Coulomb interaction in the self-energy of the electrons.
For a strong magnetic field the overlapping of Landau levels with
different indices is neglected.
The enhancement of $g^*$
can lead to a spin polarization of the 2DEG at certain values of the filling
factor $\nu$, and, in addition, the exchange interaction
can lead to the spontaneous formation of spin-density \cite{Gruener:xx} or
charge-density waves \cite{Gerhardts81:1339,Yoshioka:4986}.
The onset of a spin-density wave state in a parabolic quantum well
has been studied by Brey and Halperin using a modified Hartree-Fock
approximation (HFA)
with a point-contact exchange interaction. They find a divergence
of the electric susceptibility in the presence of a
magnetic field of intermediate strength parallel to the quantum well
and an infinitesimal fictitous magnetic field perpendicular to
the quantum well.\cite{Brey89:11634} This spin-density wave state
has a wavevector along the quantum well parallel to the intermediate
magnetic field and occurs only when the quantum
well is wide enough and the exchange interaction has a strength
larger than a critical value. The calculated optical properties of a
$\delta$-doped quantum well in the HFA
due to spin- and charge-density excitations have been found to be
in good agreement with experiments,\cite{Luo93:11086} as well as those
of donor states in 2DEG in strong magnetic fields.\cite{Hawrylak94:2943}
In the quantum well Hembree {\em et al.}\cite{Hembree93:9162}
discovered abrupt spin polarization of the system
at high magnetic fields and a spin-inversion regime where the net spin
alignment strongly varies across the well. They studied the effect
in different approximation schemes and in the presence of
impurity scattering. Recently the effects of the $g$-factor enhancement
on various transport coefficients has been reported by the
same group.\cite{Hembree94:15197}
As to microstructures of further reduced dimensionality
the spontaneous polarization of
of an array of quantum dots into a ferroelectric or antiferroelectric state
has been investigated by Kempa, Broido, and Bakshi.\cite{Kempa:9343}
The spin degree of freedom together with the exchange interaction
and correlation effects have also been found to be essential
to model few electrons in a single quantum dot in magnetic
field.\cite{Pfannkuche93:2244,Pfannkuche93:6,Hawrylak93:3347,Palacios94:5760,Ferconi94:14722}
In this paper we are concerned with the spin-related phenomena associated
with the exchange interaction that can occur in quantum dots with a large
number of electrons. We study the spin splitting of Landau bands
(LB's) due to the enhancement of $g^*$,
and the formation of a spin-inversion state (SIS)
in a strictly two-dimensional {\em finite size} electron system in a
perpendicular magnetic field of intermediate strength.
The system size is chosen to be of the order of
several magnetic
lengths, $l=\sqrt{\hbar c/(eB)}$. The LB's in the center of the
system do approach flat Landau levels indicating that an
electron in the center does not feel the boundary. We are thus
able to study the crossing from the quantum regime in which the
electronic confinement dominates over the electron-electron interaction to
the regime in which electrostatics plays a dominant role.
Finally we show how the formation of
a SIS can be detected in the far-infrared (FIR)
absorption spectrum of the system.
\section{Model}
We consider $N_s$ strictly two-dimensional electrons to model qualitatively
a real heterostructure where the 2DEG is confined to the lowest
electrical subband. The 2DEG is confined to a disk of radius $R$
in the 2D-plane by a potential step
\begin{equation}
V_{\mbox{conf}}(r)=U_0
\left[ \exp\left(\frac{R-r}{4\Delta r}\right) +1\right] ^{-1},
\label{vconf}
\end{equation}
where $\Delta r =22\: ${\AA}. To ensure charge neutrality of the system
a positive background charge $+en_b$ resides on the disk
\begin{equation}
n_b(r)={\bar n_s }
\left[ \exp\left(\frac{r-R}{\Delta r}\right) +1\right] ^{-1},
\label{nb}
\end{equation}
with the average electron density of the system given by
$\bar n_s=N_s/(\pi R^2)=\langle n_s(r)\rangle$.
In the HFA the state of each electron is
described by a single-electron Schr{\"o}dinger equation
\begin{eqnarray}
\lefteqn{\{ H^0+V_H(r)+V_{\mbox{conf}}(r)\}\psi_{\alpha}(\vec r)} \nonumber\\
& & -\int d^2 r^{\prime}\:
\Delta(\vec r,\vec r\: ')\psi_{\alpha}(\vec r\: ')
=\epsilon_{\alpha}\psi_{\alpha}(\vec r) \label{schr}
\end{eqnarray}
for an electron moving in a Hartree potential
\begin{equation}
V_H(r)={e^2\over\kappa}\int d^2 r^{\prime}
{n_s(r^{\prime})-n_b(r^{\prime})\over |\vec r-\vec r\: '|}
\label{vh}
\end{equation}
caused by the charge density $-e\{ n_s(r)-n_b(r)\}$,
and a nonlocal exchange potential with
\begin{equation}
\Delta(\vec r,\vec r\: ')={e^2\over\kappa}\sum_{\beta}
f(\epsilon_{\beta}-\mu ){\psi_{\beta}^*(\vec r\: ')
\psi_{\beta}(\vec r)\over |\vec r-\vec r\: '|}.\label{del}
\end{equation}
The equilibrium occupation of the electronic states
is according to the Fermi distribution
$f(\epsilon_{\beta}-\mu )$ at finite temperature $T$.
The density of the electrons $n_s(r)$ is constructed from
the energy spectrum $\{ \epsilon_{\alpha}\}$ and the wave
functions $\{ \psi_{\alpha}\}$
\begin{equation}
n_s(r)=\sum_{\alpha}|\psi_{\alpha}(\vec r)|^2
f(\epsilon_{\alpha}-\mu), \label{ns}
\end{equation}
together with the chemical potential $\mu$.
The label $\alpha$ represents the
radial quantum number $n_r$, the angular quantum number $M$, and the
spin quantum number $s=\pm\frac{1}{2}$. $H^0$ is the single particle
Hamiltonian for one electron with spin in a constant perpendicular
external magnetic field.\cite{Landau58,Gudmundsson90:63}
A Landau band index $n$ can be constructed from the quantum numbers
$n_r$ and $M$ as $n=(|M|-M)/2+n_r$.
The Landau levels of $H^0$ with energy $E_{n,M,s}
=\hbar\omega_c(n+\frac{1}{2})+sg^*(\mu_B/\hbar )B$
are degenerate with respect to $M$
with the degeneracy $n_0=(2\pi l^2)^{-1}$ per spin orientation.
$\mu_B$ is the Bohr magneton $(e\hbar /2mc)$.
The cyclotron frequency is given by $\omega_c=eB/(mc)$.
The Hartree-Fock energy spectrum $\{ \epsilon_{\alpha}\}$ and the
corresponding wave functions $\{ \psi_{\alpha}\}$ are now found by solving
(\ref{schr})-(\ref{ns}) iteratively
in the basis
of $H^0$.\cite{Gudmundsson90:63,Gudmundsson89:517,Pfannkuche93:2244}
The chemical potential $\mu$ is recalculated in each iteration
in order to preserve the total number of electrons $N_s$.
The number of basis functions used in the diagonalization is chosen
such that a further increase of the subset results in an unchanged
density $n_s(r)$.
The calculations have been repeated for several initial
conditions with different spin configuration
in order to search for the ground state of the
Hartree-Fock equations (\ref{schr})-(\ref{ns}).
The total energy of the system $E_{\mbox{tot}}$ can be found
by summing up the single electron contributions and carefully
counting the interaction energy of each electron pair
only once.\cite{Pfannkuche93:2244}
The FIR-absorption of the system is calculated as a self-consistent
linear response to an external potential\cite{Gudmundsson94:xx}
\begin{equation}
\phi^{\mbox{ext}}(\vec r,t)={\cal E}^{\mbox{ext}}r\exp\left\{
-iN_p\varphi
-i(\omega +i\eta )t\right\} ,
\label{phi_ext}
\end{equation}
where $\eta\rightarrow 0^+$.
$N_p=\pm 1$ corresponds to left or right circular polarization.
The small size of the system compared to the wavelength of the
external radiation makes possible to use
a electrostatic potential representing a time dependent
but spatially constant external electrical field
$\vec E^{\mbox{ext}}=-\vec\nabla\phi^{\mbox{ext}}$.
In this so-called time-dependent HFA the change
of the density matrix due to an adiabatically switched-on
total electrostatic potential
$\phi^{\mbox{sc}}$ is calculated within a linear
approximation.
The total potential consists of the external
potential and the induced potential $\phi^{\mbox{ind}}=\phi^H+\phi^F$
due to the direct and the exchange
interaction of the electrons. The induced potential in turn depends on
the density matrix, thus closing the circle and allowing for a
self-consistent evaluation of the total potential together with an
expression for the frequency dependent dielectric tensor
$\varepsilon_{\alpha\beta ,\delta\gamma}(\omega)$. The power
absorption is then calculated from the Joule heating of the system
due to $\phi^{\mbox{ext}}$
\begin{equation}
P(\omega )=e{\cal
E}^{\mbox{ext}}\sum_{\alpha\beta}\frac{(E_{\beta}-E_{\alpha})}
{\hbar}\langle\beta |r|\alpha\rangle
2\pi\delta_{M_{\beta},M_{\alpha}
\pm N_p}\Im \left\{ f^{\alpha\beta}(\omega )
\langle\alpha|(-e\phi^{\mbox{sc}})|\beta\rangle\right\} \, ,
\label{Pomega}
\end{equation}
where ${\cal E}^{\mbox{ext}}$ is the strength of the external field
and
\begin{equation}
f^{\alpha \beta}(\omega ) = \frac{1}{\hbar} \, \left\{
\frac{f_{\beta}-f_{\alpha}}
{\omega + (\omega_{\beta}-\omega_{\alpha}) + i \eta}
\right\} \,
\end{equation}
with the Fermi distribution $f_{\alpha}=f(\epsilon_{\alpha}-\mu)$.
\section{Results}
The calculations for the box-like confinement (\ref{vconf})
are carried out with
GaAs parameters: $m^*=0.067m_e$, $\kappa =12.4$, and $g^*=-0.44$.
The occupation of the LB's is varied by changing
the number of electrons $N_s$ at a
constant strength of the magnetic field $B=3.0\: $T (we could equally
have changed the magnetic field keeping constant the number of particles).
Since the radius of the system $R\geq 1000\: ${\AA} is much larger than the
magnetic length $l\approx 148\: ${\AA} and the effective Bohr
radius $a^*_0\approx 97.9\: ${\AA} we can use as a convenient label
an effective filling factor $\nu $ describing the
occupation of the LB's in the
interior of the system. The cyclotron energy $\hbar\omega_c\approx 5.2\: $meV,
so a sufficient height of the confining potential is $U_0=60\: $meV in order
to include several LB's in the calculation.
For $B=3.0\: $T the bare spin splitting of the LB's
$E_{\mbox{Zeeman}}=(g^*\mu_B/\hbar )B\approx 0.076\: $meV
is much smaller than their separation
$\hbar\omega_c$ and corresponds to the thermal energy $k_BT$
at $T\approx 0.9\: $K.
Figure \ref{fig1} shows the HFA quasiparticle energy spectrum for four
values of $N_s$ such that the filling factor $\nu$ (defined as the
number of occupied bands in the central
region of the box) ranges from 4 to 2.
Figure \ref{fig1}a corresponds to the case $\nu=4$. Electrons in the
first LB ($n=0$) form a large paramagnetic
compact droplet while electrons in the second LB ($n=1$) form a smaller one.
Figures \ref{fig1}b and \ref{fig1}c
show clearly a large spin splitting of the LB's
due to the enhancement of $g^*$ when the 2nd LB is half filled ($\nu=3$)
and the electrons in it form a ferromagnetic compact droplet. Finally,
Fig.\ \ref{fig1}d shows the case corresponding to the droplet at $\nu=2$ when
no electrons are left in the 2nd LB. In our case
the 2nd LB behaves in all respects like an independent, smaller quantum dot,
and its properties are identical to those studied previously.
\cite{Palacios94:5760}
However, the first LB present a more complicated behavior:
The energy spectra for six values of $N_s$ such that the chemical
potential $\mu$ lies in the neighborhood of the first LB is seen
in Fig.\ \ref{E_U}. Now the filling factor lies within the range
$1\leq \nu \leq 2$. For $N_s=42$ (Fig.\ \ref{E_U}a) both spin
bands of the first LB in the bulk region are still filled
($\mu$ is still lying between the first and second LB's but closer to the
former one). The small Zeeman energy makes
the LB's look degenerate with respect to the spin degree of
freedom. As the number of electrons is reduced to 38 (Fig.\ \ref{E_U}b)
the spin bands split up near the edge and the number of spin-down
electrons becomes smaller than that of spin-up electrons. In addition to
this splitting (which was also present in the 2nd LB in Fig.\ \ref{fig1})
we can observe two instability points (one for each
spin band) rising near the center of the system. By instability points
we mean bumps in the spin bands approaching the chemical
potential. Thus, one should expect some transition as the number of electrons
keeps changing (or the magnetic field)
and these bumps touch the chemical potential. Before
that can happen one can even see signatures of such transitions in the density
plotted in Fig.\ \ref{d3}. The finite temperature "reveals" a budding
spin-inversion state due to the difference in distance to
the chemical potential
between the spin-up and spin-down bands for a given position ($M$).
The onset of such a SIS takes place
when the number of particles is reduced further and those bumps cross the
chemical potential (Figs.\ \ref{E_U}c to \ref{E_U}e). Finally, the
compact droplet at $\nu=1$ is formed (Fig.\ \ref{E_U}f).
In order to understand better the formation of the SIS let us
consider a simpler but equivalent situation.
The number of LB's cannot be reduced considerably in the calculations
with the box-like confinement (\ref{vconf}) since we are using a
basis constructed of the eigenfunctions of noninteracting
electrons in an infinite system. However, by considering parabolically
confined interacting electrons and using the one-electron basis
set of such a system we are left only with band mixing due to the
electron-electron interaction.
We have thus calculated the energy spectra of a 2DEG in a parabolic
quantum dot with confinement frequency $\hbar\omega_0$
for an increasing number of LB's (from one to three) at $T=1.0\: $K.
First, we analyze the case of one LB at zero temperature.
Figure \ref{jjp} shows the
evolution with $B$ of the band structure for $N_s=30$. The spin splitting
opens up from the edge to the center of the
LB in the parabolic confinement
as we go from $\nu=2$ to $\nu=1$. Surprisingly, one can see how both spin bands
near the center of the system bend upward, and, eventually, one of them
crosses the chemical
potential. This cannot happen for a smaller number of electrons since in that
case the $s=-\frac{1}{2}$ electrons
leave their band before the unstable point (in
the center) crosses the chemical potential.
Such central instability requires a certain size of the electronic droplet and
constitutes the initial stage of the SIS.
If we include higher LB's the spectra becomes more complicated and
the instability points of each spin band shift from each other
due to the mixing with higher LB's.
This result is presented in
Fig.\ \ref{E_para}. The total energy of the system $E_{\mbox{tot}}$
is $1789.07\: $meV for one LB, $1728.3\: $meV for two LB's, and
$1718.5\: $meV for three.
Obviously the calculation with one LB does
not represent well the ground state for the given values of
$\hbar\omega_c$ and $\hbar\omega_0$, but helps us to get an insight on the spin
instability.
{\em No twisting of the spin bands is ever seen for the calculation
with one LB.} This can be verified by checking the analytical
expressions for the matrix elements of the exchange interactions.
The additional degree of freedom introduced to the system by allowing
coupling of states of higher LB's into the lowest LB for interacting
electrons is essential in order to obtain the full richness of the
spin band structure.
The formation of the SIS invokes clear signs in the
FIR spectrum $P(\omega )$ of the 2DEG detailed in Fig.\ \ref{FIR-absorption}.
The first two subfigures show the spectrum in the Hartree approximation (HA)
and the HFA, respectively.
In the HA the exchange interaction is neglected both in the ground
state and the excited states.
A common feature is the occurrence of two strong absorption
lines, the lower one in energy corresponding to $N_p=+1$ and the higher
one corresponding to $N_p=-1$. These two lines can either be identified
as the ones corresponding to the center of mass motion predicted by the
generalized Kohn theorem for quantum dots with parabolic
confinement,\cite{Kohn61:1242,Maksym90:108,Brey89:10647,Gudmundsson91:12098,Pfannkuche94:1221}
or more appropriately here as the low energy excitation of an edge plasmon and
the 2D bulk plasmon at energy slightly higher than the cyclotron resonance
$E_c=\hbar\omega_c$.\cite{Gudmundsson94:xx}
Both approximation then show small absorption peaks above the
bulk magnetoplasmon that have been identified as absorption
due to single electron transitions.\cite{Malshukov94:2,Gudmundsson94:xx}
The spin splitting itself does not have large effects on the absorption
due to the bulk magnetoplasmon but the finer details of the corresponding
absorption peak in a parabolic quantum well have been studied by
Hembree et.\ al.,\cite{Hembree94:15197} here we shall concentrate on the
effects of the SIS. By comparing the spectra for the
two approximations at energy below the energy of the edge plasmon
we find small peaks for $N_p=-1$ that
are enlargened in the last subfigure of Fig.\ \ref{FIR-absorption}.
No such peaks are found in the HA. They are only present when the
SIS occurs and the ones with the lowest energy are
caused by single electron
transitions in the lowest LB, intra-Landau-band transitions with
$M\rightarrow M-1$ that are only possible because of the twisting
of the LB's. Corresponding absorption peaks of the opposite
polarization $N_p=+1$ can also be found in the center subfigure
at similar energy, but the peaks with $N_p=-1$ are much more
characteristic of the SIS since otherwise peaks of that
polarization are never found for low energy.
As soon as the spin Landau bands of the lowest LB cross twice a second
absorption peak appears with energy above the edge magnetoplasmon but
below the bulk plasmon. The occurrence of this second row of peaks
has to be correlated with the fact that the twisting of the lowest
and the next LB, that did mirror one another for lower $N_s$, are now
out of phase for the higher values of $N_s$ corresponding to
$\nu$ just below 2.
\section{Discussion and summary}
In a system of a confined 2DEG we have been able to
demonstrate both bulk effects and phenomena caused by the finite
size of the system, in the absence of any impurity scattering of the
electrons.
The 2D system is large enough so that the LB's approach flat
Landau levels for low values of the angular quantum number $M$.
This can be interpreted as the formation of 2D bulk states inside
the system. The ensuing singular density of states together with
the exchange interaction causes the well known oscillations of the
energy separation of the LB's with the same Landau level index $n$
but opposite spin orientations as a function of the filling factor
$\nu$. Here we have seen that the enhancement of $g^*$ occurs not
only in the LB where $\mu$ is located but in all the LB's included
in the model. Similar behavior has been established in optical
measurements of a 2DEG by Kukushkin.\cite{Kukushkin:511}
We have observed the spontaneous formation of concentric
circular regions of different spin phases when the spin
splitting of the first LB's is opening up with a decreasing $\nu$
at a low temperature. The shape of this SIS depends on the
size, shape of the system, and filling factor $\nu$,
such that the wavelength decreases as $\nu$ approaches an even
integer. The coupling of the states of higher Landau bands
into the lowest band by the Coulomb interaction of the 2DEG
is {\em essential} for the fine structure of the SIS.
Even though we have been using a restricted HFA here (total angular momentum
and spin are good quantum numbers)
different results can be attained by choosing different initial spin
configurations. In Fig.\ \ref{E_Ns30_Varg} we show three stable
states with higher energy than the ground state seen in Fig.\ \ref{E_U}c.
It is interesting to note that the state with no crossing of
spin bands is not the ground state.
The exact shape of the SIS
does strongly depend on the confining potential and, thus, also the size
of the system. As was noted earlier the LB's do not twist when $\mu$
is crossing higher LB's and the spin splitting is opening up, but the
uneven opening up produces strong modulation of the spin densities.
To exclude the possibility that numerical deficiencies are
causing the twisting of the Landau bands we have
tested the stability of the spin-density structures by increasing the
number of basis states included in the numerical calculation
and tested different schemes in attaining the convergence
of the self-consistent problem.
No visible changes in the ground state properties were observed.
On the other hand, the exact shape and formation
of the SIS does depend on the
size of the system emphasizing that we are observing a confined
spin-density wave (SDW) here.\cite{Gudmundsson94:xx,Gudmundsson94:ps}
Two possible problems associated with the HFA come to mind.
First, the HFA may lead to a ground state that is quite different
from the physical one due to the strong exchange force that may be
reduced in better approximations where higher order correlation effects
or impurity broadening to a high order are included.\cite{Uenoyama89:11044}
It is thus, very reassuring that this type of spin inversion and
formation of a SDW has been observed in models employing
the local density approximation (LDA) where the SDW has
been observed for different approximations of the correlation
effects.\cite{Hembree93:9162} The on-set of the SDW is
also found to depend on the amount of collision broadening of the LB's,
but neither the broadening nor the correlation effects
prevent it.\cite{Hembree93:9162} The spatial correlation of the
2DEG in two approaching finite-size layers for the common filling factor
of unity is quite similar to the formation of the SIS here.
The layer index can be treated as isospin for vanishing separation
and the numerical diagonalization of the
many-electron Hamiltonian in a large subspace of noninteracting
many-electron states includes, in principle, all correlation effects
in the model to a high degree of accuracy.\cite{Palacios95:1769}
An important difference of the present SIS
in the two-dimensional plane to the SDW parallel to $\vec B$ investigated
by Brey and Halperin \cite{Brey89:11634} is the fact that the
wavelength of the present modulation varies strongly with
$\nu$. This is caused by the strong dependence of the effective
interaction, or the screening, in the 2D plane on
$\nu$.\cite{Gudmundsson90:63,Wulf88:4218,Labbe88:1373,McEuen92:11419}
The SDW found by Brey and Halperin has strong
reassemblance with the more ``traditional one'' known
in 1D electroninc systems.\cite{Gruener:xx} The notation SIS
is, therefore, used here to emphazise this difference.
The region of filling factors when the electrons
are not fully spin polarized yet ($1\leq\nu\leq 2$)
but the system has not entered the
regime of the integer quantum Hall effect with the lowest LB
filled ($\nu=1$) has attracted much interest lately. It has been shown that
in absence of Zeeman energy the lowest
energy charged excitations at $\nu =1$ are skyrmions, spin textures
with a unit winding number in two
dimensions.\cite{Sondhi93:16419,Fertig94:11018}
At large $g$ the quasi-particles, analogous to the single particles, have
unit charge $\pm e$ and spin half, $s=\pm 1/2$, but as $g$ is reduced
to zero the excitation gap survives and the size of the quasi-particles
diverges with the spin becoming macroscopic - skyrmions.\cite{Fertig94:11018}
This effect has also been studied in double-layered electron systems
when the distance between the layers, each having no spin degree of freedom,
is reduced since these models can be mapped directly onto the spin system
identifying the layer index as an
isospin.\cite{Palacios95:1769,Moon95:5138}
It has also been found that
these spin textures might eventually dominate the ground state properties
at filling factors
$1\leq\nu\leq 2$.\cite{Palacios94:5760,Palacios95:1769,Schmidt95:5570}
The SIS's found in the present work are
{\em not} related to the skyrmions observed in such regime of filling factors,
but the skyrmions and the SIS's
may coexist, which emphasizes the
very complex and interesting structure of the 2DEG in such a regime.
The spin-density modulation was found to cause clear signs
in the FIR-absoption of the confined 2DEG. The signs may be
weak since they partly reflect single-electron transitions rather than
collective oscillations and they may be in the low frequency part
of the spectrum most difficult to measure, but the final
word about the appropriatness of the HFA or the LDA for the
current model will come from experiments.
\acknowledgements
The authors are indebted to R.\ R.\ Gerhardts, D.\ Pfannkuche,
A.\ H.\ MacDonald, M.\ Ferconi, G.\ Vignale,
and G.\ P{\'a}lsson for fruitful discussion.
This research was supported in part by the Icelandic Natural Science
Foundation, the University of Iceland Research Fund,
NATO collaborative research Grant No. CRG 921204, NFS contract
NSF-DMR9416906, and, for
one of the authors (J.J.P.), by a NATO postdoctoral fellowship.
\bibliographystyle{prsty}
|
\section{Introduction}
The Kondo effect\cite{kondo}
has been of great interest in condensed matter physics
since its observation. The proposed model Hamiltonian, a magnetic
$S_I=1/2$ local moment interacting with the conduction electron gas,
looked very
simple but was non-trivial due to many body nature of the problem.
Ever since, many generalized models have been studied to extend our
understanding and to relate to real materials.
The simplest $S_I=1/2$ orbitally nondegenerate
Anderson model\cite{anderson} and s-d exchange model are now
well understood for a single impurity case using several techniques.
The numerical renormalization group (NRG)\cite{nrg}
method was able to provide complete information about crossover
from the high temperature fixed point to the low temperature
fermi liquid fixed point for these models.
Subsequently, the exact diagonalization of these
models was realized by the Bethe Ansatz (BA)\cite{bethe},
which also gives the exact
solution for thermodynamics of these models.
However, it has not proven possible to compute dynamical properties with
the BA.
Through non-crossing approximation (NCA)\cite{nca},
dynamics as well as thermodynamics\cite{bcw}
have been extensively studied for the infinite on-site Coulomb interaction
models.
The Quantum monte Carlo method (QMC)\cite{qmc}
has also been applied
to study statics and dynamics for the simple $S_I=1/2$ models.
Recently conformal field theory (CFT)\cite{cft,ludwig}
has been used to study all properties near
the low temperature fixed points.
However we are still far from a complete understanding for realistic
models which, for example, include the strong spin-orbit coupling, the
crystal electric field effects, and multiple (more than two) configurations.
In this paper, we study a realistic extension of the conventional simple
approach to the Kondo effect for Ce$^{3+}$ ions,
including the strong spin-orbit coupling,
the crystal electric field effect, and multiple configurations
in their simplest form.
Ever since Nozi\`{e}res and Blandin\cite{nozbland} introduced the
multichannel Kondo model, its realization in real materials
has been controversial.
On the theoretical side, the multichannel Kondo models
are well understood\cite{cft,twoch1,twoch2,sacramento}
irrespective of the experimental situations.
When the degenerate channel number ($N_{\rm ch}$), which is the number of the
conduction electron ``bands" coupled to the impurity site, is greater than
$2S_I$ (the impurity spin), the impurity magnetic moment is
overscreened and the strong coupling fixed point (Fermi liquid) becomes
unstable leading to a non-trivial fixed point (non-Fermi liquid behavior).
In the underscreened and the completely screened cases,
$N_{\rm ch} \leq 2S_I$, the strong coupling fixed point is stable
resulting in a Fermi liquid ground state.
The spin susceptibility $\chi (T)$ and
specific heat coefficient $ C(T) / T$ for the
two-channel $S_I=1/2$ magnetic Kondo model
are proportional to $\log (T_K/T)$ at low temperatures
\cite{cft,ludwig,twoch1,twoch2,sacramento}, where
$T_K$ is the Kondo energy scale.
The dynamic susceptibility shows Marginal Fermi liquid behavior
\cite{coxruck}. The resistivity increases logarithmically and saturates to
a constant with $\rho (T) = \rho (0) ~[~ 1 - a \sqrt{T/T_{\rm K}} ~]$ with
decreasing temperature\cite{cft}.
On the other hand, the one-channel $S_I=1/2$
Kondo model leads to the Fermi liquid ground state.
In that case, magnetic susceptibility $\chi (T)$ and
specific heat coefficient $ C(T) / T$ saturate to constants
of order $1/T_K$\cite{nrg,bethe}.
Resistivity increases logarithmically and saturates to
a constant with $\rho (T) = \rho (0) ~[~ 1 - a (T/T_{\rm K})^2 ~]$ with
decreasing temperature\cite{noz}.
In this paper we study
a model Hamiltonian for Ce$^{3+}$ ions in cubic
metals with three configurations ($f^0$, $f^1$, $f^2$).
The nominal ground configuration $f^1$
can fluctuate to $f^0$ and $f^2$ configurations by hybridizing with
the conduction electrons. A one-channel Anderson hybridization interaction
is present between $f^0$ and $f^1$ configurations. A two-channel Anderson
hybridization interaction
is present between $f^1$ and $f^2$ configurations.
We report detailed studies of our simplified Hamiltonian
of one-channel \& two-channel Anderson model using the NCA.
This simple model is quite intriguing in that we can study
the competition between the two different Kondo effects, that is,
the Fermi liquid and the non-Fermi liquid fixed points.
The distinct ground state physics for the different numbers of channels
is classified using the zero temperature analysis of NCA integral equations
and the third order scaling theory. We calculate the thermodynamics and the
dynamics of this simple model and find that all the calculated physical
quantities show the behavior appropriate for the accessible different
channel numbers. The static magnetic
susceptibility displays a scaling behavior agreeing with the exact Bethe
ansatz results in the two- and three-channel cases. NCA calculation of
entropy and specific heat is also compared with the Bethe ansatz results.
The resistivity shows the correct temperature dependence near zero temperature
agreeing with the conformal field theory results in the two- and
three-channel cases. The sign and the magnitude of the thermopower are
dependent sensitively on the relevant channel numbers. The peak position
in the dynamic magnetic susceptibility is almost linear in temperature
in the overscreened cases. A short paper which presents some of these
results will appear elsewhere\cite{kimcox}.
Our study is motivated by a recent discovery of non-Fermi liquid system,
Ce$_x$La$_{1-x}$Cu$_{2.2}$Si$_2$ ($x=0.1$)\cite{andraka}.
We summarize the experimental findings of this alloy system.
The logarithmic divergence in both the magnetic
susceptibility $\chi(T)$ and the specific heat linear coefficient
$\gamma (T)$ have been observed for
Ce$_x$La$_{1-x}$Cu$_{2.2}$Si$_2$ ($x=0.1$).
The two-channel $S_I=1/2$ magnetic Kondo physics
\cite{cft,ludwig,twoch1,twoch2,sacramento} provides a theoretical framework
to explain the thermodynamics of this system at low temperatures.
$\gamma (T)$ initially increases in the presence of the magnetic field,
which qualitatively agrees with the two-channel Kondo effect coming from
the lifting of residual entropy\cite{sacramento}.
In the one-channel Kondo effect, the
Sommerfeld coefficient decreases in the magnetic field due to
the destruction of the Kondo effect.
The Wilson ratio is estimated to be $R \approx 2.7$ from the slopes of two
curves ($\chi (T)$ and $\gamma (T)$), which compares well with
theoretical value $8/3$ for the two-channel magnetic $S_I=1/2$ Kondo
model\cite{ludwig}.
The good agreement between the theoretical and the experimental Wilson
ratios supports our crystalline electric field energy scheme described
below. This system is pseudo-cubic (i.e., the crystal field scheme on the
Ce$^{3+}$ site appears cubic). The best superconducting
system with excess Cu shows almost isotropic magnetic
susceptibility\cite{chitep}.
That the pseudo-cubic $\Gamma_7$ magnetic doublet in $f^1$ lies lowest
is inferred from
the neutron scattering experiments\cite{neutron}.
The thermopower for CeCu$_2$Si$_2$
changes sign near $70$ K and stays negative below with a large
extremum ($-20~\mbox{to}~-30 \mu$V/K)\cite{chitep},
suggesting the presence of
strong hole resonance scattering.
As we will show below, these thermopower
results also support our interpretation of the two channel magnetic
Kondo physics. Though other experiments (e.g., specific heat and magnetic
susceptibility) support the interpretation of them in terms of the
two-channel Kondo effect, a linear temperature dependence in the resistivity
remains as a puzzle.
The $\sqrt{T}$ behavior in the resistivity
is predicted from conformal field theory treatment of the two-channel
Kondo models\cite{cft,ludwig}. In Fig.~\ref{resfit}, we present our
numerical calculation of resistivity and the experimental results
measured in the alloy system Ce$_x$La$_{1-x}$Cu$_{2.2}$Si$_2$ ($x=0.1$)
\cite{andraka}. It can be seen that the data curves downwards at lower
temperatures which may indicate a crossover to a new fixed point.
{}From all these experimental findings, we believe
that the alloy system
Ce$_x$La$_{1-x}$Cu$_{2.2}$Si$_2$ ($x=0.1$) is a strong candidate
for two-channel $S_I=1/2$ magnetic Kondo system.
In addition, the two-channel, $S_I=1/2$ Kondo effect may be realized
in other materials: two-channel quadrupolar Kondo effect\cite{quad}
in some U alloys and two-level systems\cite{tls} in metallic point contacts.
U alloy systems include U$_{0.2}$Y$_{0.8}$Pd$_3$
\cite{uypd}, U$_x$Th$_{1-x}$Ru$_2$Si$_2$ \cite{urusi},
UCu$_{3.5}$Pd$_{1.5}$ \cite{ucupd},
U$_{0.1}$Th$_{0.9}$Ni$_2$Al$_3$ \cite{unial},
U$_{0.1}$Pr$_{0.9}$Ni$_2$Al$_3$ \cite{unial}, U$_x$Sc$_{1-x}$Pd$_3$
\cite{uscpd}, U$_{0.9}$Th$_{0.1}$Be$_{13}$\cite{ube13},
and U$_{x}$Th$_{1-x}$Pd$_2$Al$_3$\cite{updal}.
All the above systems show a logarithmic divergence at low temperature
in the linear specific heat coefficient and a square root saturation in the
static magnetic susceptibility. U$_x$Th$_{1-x}$Ru$_2$Si$_2$ \cite{urusi}
is an exception showing a logarithmically divergent magnetic susceptibility.
The two channel quadrupolar Kondo physics has been invoked to
explain non-Fermi liquid behavior in thermodynamic and
transport properties of U$_x$Y$_{1-x}$Pd$_3$ for $x=0.2$ \cite{uypd} and
other U alloy systems. Recently, the resistivity in a metallic constriction
was observed to obey $\sqrt{T}$ behavior and was interpreted as due to
two-channel Kondo scattering from atomic two-level tunneling systems
\cite{cornel,ufl}.
Our paper is organized as follows.
In section II, we introduce our simple model Hamiltonian and analyze
this model using the third order scaling equation.
We briefly introduce the NCA in section III.
Zero temperature analysis of NCA integral equations follows in section IV.
In section V, we present the detailed numerical analysis of our simple model
Hamiltonian. We conclude in the section VI.
\section{Model Hamiltonian}
The single impurity Anderson model\cite{anderson} has been very successful
\cite{bcw}
in describing Kondo systems (meaning magnetic transition metal elements
embedded in normal metals, and dilute rare earth or actinide alloys).
The thermodynamics is rather well explained by the single impurity
properties for even highly concentrated Ce alloys\cite{bcw,singleion}.
Coherence effects, arising from the lattice of Anderson or
Kondo ions at low temperatures, do not play such an important role
in thermodynamics. Transport properties also are well explained by
the single impurity model except for the low temperature regime
where a coherence effect prevails due to the coherent Bloch state
formation leading to vanishing resistivity at zero temperature
(the residual resistivity is larger than the room temperature one
in the dilute impurity limit).
This one-channel Anderson model can explain the complete screening of the
magnetic moment at the local moment sites leading to
the local Fermi liquid ground state discussed in the introduction.
The two-channel orbital Kondo model (quadrupolar or two-level system)
keeps the channel symmetry guaranteed by time reversal symmetry, but
can suffer from the ground state degeneracy lifting due to the Jahn-Teller
effect.
In general, the orbital Kondo model has the exchange anisotropy.
It has been shown that the exchange anisotropy is irrelevant for
the two-channel $S_I=1/2$ Kondo models\cite{anisoexchange}.
As will be shown in other publication\cite{kimcoxtobe},
the two-channel magnetic Kondo model
for Ce$^{3+}$ is vulnerable to the channel asymmetry due to its orbital
nature in the channel degrees of freedom\cite{nozbland}. However,
the channel-mixing Kondo interaction can save the two-channel physics
\cite{kimcoxtobe}.
The conventional theory of Ce$^{3+}$ impurities assumes the infinite
on-site Coulomb interaction, which removes the $f^2$ configuration
from consideration\cite{bcw}, and
as a result has no chance to get the two-channel Kondo effect
which we shall explain below. When we relax our assumption about the
strong on-site Coulomb interaction and we include the detailed
atomic energy structure, we can develop a variety of
model Hamiltonians\cite{kimcoxtobe}.
In our simple model, we assume the magnetic $f^1~J=5/2~\Gamma_7$ CEF
doublet lies lowest in the $f^1$ configuration, and we keep the two
excited states -- a singlet $f^0$ and the nonmagnetic $f^2 ~ \Gamma_3$
CEF doublet.
We find the one-channel Anderson model in mixing between $f^0$ and $f^1$
configurations and the two-channel Anderson model in mixing
between $f^1$ and $f^2$ configurations\cite{coxham}.
Other interesting Kondo interactions\cite{kimcoxtobe} arise
when the excited triplets in the $f^2$
configuration are included.
According to the group theory analysis, the hybridization is mediated
only by the cubic $\Gamma_8$ conduction electrons between $f^1$ and $f^2$
($\Gamma_3 \otimes \Gamma_7 = \Gamma_8$),
and by $\Gamma_7$ between $f^0$ and $f^1$ for the mixing potential
allowed in the cubic crystal. CEF states are schematically drawn in
the Fig.~\ref{cef} for this simple model.
To see the essential physics, we restrict our attention to
the simple case of isotropic hybridization and a free conduction band
with Lorentzian/Gaussian density of states (DOS). In this simple case,
two components of conduction partial wave $j_c =5/2, 7/2$ can mix with
the atomic orbitals.
For the moment, we will consider only one angular momentum component,
say, $j_c =5/2$ of the conduction band for our model study.
We shall examine the effects of relaxing this in the other paper
\cite{kimcoxtobe}.
Our model Hamiltonian is
\begin{eqnarray}
H &=& H_{\rm cb} + H_{\rm at} + H_1, \\
H_{\rm cb}
&=& \sum_{\epsilon \alpha} \epsilon
c_{\epsilon \Gamma_7\alpha}^{\dagger} c_{\epsilon \Gamma_7\alpha}
+ \sum_{\epsilon n \alpha} \epsilon
c_{\epsilon \Gamma_8n \alpha}^{\dagger}
c_{\epsilon \Gamma_8n \alpha}, \\
H_{\rm at}
&=& \epsilon_0 |f^0><f^0| + \epsilon_1 \sum_{\alpha} |f^1 \Gamma_7\alpha >
< f^1 \Gamma_7\alpha |
+ \epsilon_2 \sum_{n}|f^2 \Gamma_3n >< f^2 \Gamma_3n |,
\\
H_1 &=& V_{01} \sum_{\epsilon \alpha} c_{\epsilon \Gamma_7\alpha}^{\dagger}
| f^0 >< f^1 \Gamma_7\alpha | + h.c. \nonumber\\
&& + V_{12} \sum_{\epsilon n \alpha} (-1)^{\alpha +1/2}
c_{\epsilon \Gamma_8n \alpha}^{\dagger}
|f^1 \Gamma_7\bar{\alpha} > < f^2 \Gamma_3 n | + h.c.
\end{eqnarray}
$c_{\epsilon\Gamma_7\alpha} (c_{\epsilon\Gamma_7\alpha}^{\dagger})$
and $c_{\epsilon \Gamma_8n \alpha} (c_{\epsilon \Gamma_8n \alpha}^{\dagger})$
are the annihilation (creation)
operators for conduction electrons of $\Gamma_7$ and $\Gamma_8$,
respectively. $\alpha = \uparrow,\downarrow $ denotes the time reversal
pair for the irreducible $\Gamma_7$ representation, and
$n = \pm$ is the quadrupolar index for $\Gamma_3$ and here acting
as the channel indices. The conduction electrons
are assumed to be described by the uncorrelated Lorentzian/Gaussian
density of states with bandwidth $D=3$ eV.
$\epsilon_{0,1,2}$ are the configuration energies for the
empty ($\epsilon_0 = 0$), single, and double occupancies, respectively.
We lumped all Clebsch-Gordon
coefficients into the hybridization constants, $V_{01}$ and $V_{12}$,
except for the phase dependence on the cubic degeneracy indices.
The phase dependence on the Kramers doublet index comes from time
reversal symmetry.
We will consider these two hybridization constants to be independent of
each other to probe the competition between the one-channel
and the two-channel Kondo physics.
When the real charge fluctuations are removed from the model system
in the Kondo limit, we have to construct the tensor operators representing
each CEF states for the $f^1$ configuration and the projected conduction
electron CEF states. In this paper, the relevant tensor operators are
for $\Gamma_{7,8}$ CEF states. We can show using the character table
\cite{group} that
\begin{eqnarray}
\Gamma_7 \otimes \Gamma_7
&=& \Gamma_1 \oplus \Gamma_4, \\
\Gamma_8 \otimes \Gamma_8
&=& \Gamma_1 \oplus \Gamma_2 \oplus \Gamma_3 \oplus
2 \Gamma_4 \oplus 2\Gamma_5.
\end{eqnarray}
In the direct product, the first CEF states are written as ket, and the
second as bra. The $\Gamma_7$ tensor operator ($2\times 2$ tensor)
is a direct sum of charge operator ($\Gamma_1$) and
spin operator ($\Gamma_4$). Indeed,
the Schrieffer-Wolff transformation leads to two interaction
terms: the spin exchange
interaction and the pure potential scattering term.
The relevant terms of $\Gamma_8$ tensor operators are $2\Gamma_4$ in our
model. In the conduction electron $\Gamma_8$ tensor space, one of the two
$\Gamma_4$'s gives rise to the ordinary $S_c=1/2$ spin operators with two
degenerate orbital channels
and the other to the $S_c=3/2$ spin operator with one channel\cite{kimcoxtobe}.
Hence, there are three distinct channel labels for conduction electron partial
wave states about the Ce impurity in this simple model.
One channel is just the $\Gamma_7$
doublet. The other two are the $\Gamma_{3}\pm$ ``orbital" states of the
$\Gamma_8$ quartet. Each $\Gamma_{3}\pm$ orbital has a $\Gamma_7$
``spin" doublet (recall $\Gamma_8 = \Gamma_7 \otimes \Gamma_3$.)
As shown in Fig.~\ref{channel} for $J=5/2$ conduction partial waves,
the $+$ orbital is ``stretched" along
the quantized axis (one of the three principal cubic axes, taken to be
$\hat{z}$ here for definiteness.)
The ``$-$" orbital is ``squashed" in the $xy$ plane.
We note that the
simplest example of $\Gamma_8$ partial wave quartet is, for zero spin-orbit
coupling, $d$-wave states with $+ \to 3z^2 - r^2$ and $- \to x^2-y^2$.
The ``spin" index is then real spin of the electrons.
In the Kondo limit with the stable $f^1$ configuration, we may
remove the $f^0-f^1, f^1-f^2$ charge fluctuations from the Hamiltonian of
Eq. (4) using the Schrieffer-Wolff
transformation \cite{sw} to find the effective Hamiltonian.
\begin{eqnarray}
\tilde{H}_1
&=& J_1 \vec{S}_{c\Gamma_7} (0) \cdot \vec{S}_{\Gamma_7}
+ J_2 \sum_{n=\pm} \vec{S}_{c\Gamma_8 n} (0) \cdot
\vec{S}_{\Gamma_7} \\
J_1 &=& { 2|V_{01}|^2 \over - \epsilon_1 }, ~~
J_2 = { 2|V_{12}|^2 \over \epsilon_2 - \epsilon_1 }, \\
\vec{S}_{\Gamma_7}
&=& { 1\over 2} \sum_{\alpha\beta}
|f^1;\Gamma_7\alpha > \vec{\sigma}_{\alpha\beta} <f^1;\Gamma_7\beta |.
\end{eqnarray}
$\vec{S}_{\Gamma_7}$ is the $f^1$ pseudo spin.
$\vec{S}_{c\Gamma_7} (0)$ and $\vec{S}_{c\Gamma_8 \pm} (0)$
are the conduction electron pseudo spin densities at the impurity
site of symmetry $\Gamma_7$ and $\Gamma_8$, respectively.
When the $f^1$ configuration is stable, its pseudo spin is coupled to
the conduction band in a one-channel via $f^0$ configuration,
and in a two-channel via $f^2$ configuration.
Their couplings are both antiferromagnetic.
The unique feature of our Hamiltonian is that it can generate
the one-, two-, and three-channel ground states depending on
the model parameters. The competition between the Fermi liquid fixed point
and the non-Fermi liquid fixed point can thus be investigated
using this model Hamiltonian.
We first analyze our simple model Hamiltonian using the third
order scaling argument, i.e., the perturbative renormalization group (RG)
\cite{nozbland,scale}.
At temperature $T$, only the conduction electrons (thermally excited)
inside the band of order $T$ with respect to the Fermi level play an
important role in determining physical properties.
Thus we can integrate out the band edge states (virtually excited states)
to find the effective Hamiltonian. Though the following analysis is
restricted to $\omega \ll D$(the conduction band width)
and the perturbative regime (weak coupling limit),
we can derive qualitative results out of this. For quantitative results,
a full numerical renormalization group (NRG) study is required.
It can be deduced from the scaling theory that the low
temperature and the low energy physics is dominated by the one-channel
or the two-channel Kondo effects depending on their relative
magnitude of the antiferromagnetic couplings.
We introduce an exchange coupling matrix in the channel space which
is convenient for the derivation of the scaling equations.
We can thus rewrite the one-channel and two-channel Kondo model
in the following form\cite{kimcoxtobe}.
\begin{eqnarray}
\tilde{H}_1
&=& {\bf J} \otimes \vec{S}_c (0) \cdot \vec{S}_I, \\
{\bf J} &=& \pmatrix{J_1 & 0 & 0 \cr
0 & J_2 & 0 \cr 0 & 0 & J_2}.
\end{eqnarray}
Here $\vec{S}_c$ and $\vec{S}_I$ are $S=1/2$ operators.
The scaling equations
of our simple model Hamiltonian up to the third order diagrams of
Fig.~\ref{scale} are
\begin{eqnarray}
{\partial {\bf g} \over \partial x}
&=& {\bf g}^2 - {1\over 2} ~ {\bf g} ~ \mbox{Tr} [{\bf g}^2],
\\
{\bf g} &=& N(0) {\bf J}.
\end{eqnarray}
The scaling equations in components are
\begin{eqnarray}
{\partial g_1 \over \partial x}
&=& g_1^2 - {1\over 2} ~g_1~[~g_1^2 + 2 g_2^2 ~], ~~ g_1 = N(0)J_1 > 0, \\
{\partial g_2 \over \partial x}
&=& g_2^2 - {1\over 2} ~g_2~[~g_1^2 + 2 g_2^2 ~], ~~ g_2 = N(0)J_2 > 0.
\end{eqnarray}
Here $x = \log(D/T)$. We can identify three fixed points related
to one-, two-, and three-channel Kondo physics.
The one-channel, strong coupling
fixed point $(g_1^*, g_2^*) = (\infty, 0)$ is stable leading
to the Fermi liquid ground state\cite{scale}.
The three-channel fixed point $(2/3, 2/3)$ is
stable along the line $g_1 = g_2$ in the $g_1-g_2$ plane, but unstable for
any small perturbation from $g_1 = g_2$.
Finally, the two-channel fixed point
$(0, 1)$ is stable leading to the logarithmically divergent thermodynamic
properties at zero temperature.
{}From the scaling analysis, we can infer the ground state physics:
one-channel for $J_1 > J_2$; two-channel for $J_1 < J_2$;
and three-channel for $J_1 = J_2$. As will be shown in
section IV, the zero temperature analysis of the NCA equations leads to
the same conclusion.
We now discuss the neglected other 8 $\Gamma_3$ irreps in the
$f^2$ configuration\cite{kimcoxtobe}.
The 9 $\Gamma_3$ CEF states all contribute to the enhancement of the
two-channel exchange coupling between $f^1 \Gamma_7$ spin and the $\Gamma_8$
conduction electron spins.
\begin{eqnarray}
H_1 &=& \sum_{\epsilon} \sum_{i n\alpha} (-1)^{\alpha +1/2} V_{12}^i ~
c_{\epsilon \Gamma_8 n \alpha}^{\dagger} ~
|f^1 \Gamma_7\bar{\alpha} > <f^2 \Gamma_3^i n|
+ h.c.
\end{eqnarray}
The NCA can treat this problem with
the extension that now the $f^2 ~ \Gamma_3$ Green's function becomes
an $9\times 9$ matrix. See the Appendix for more details.
The Schrieffer-Wolff transformation leads to
\begin{eqnarray}
\tilde{H}_1
&=& J ~ \sum_{n=\pm} \vec{S}_{c\Gamma_8 n} \cdot \vec{S}_{\Gamma_7}, \\
J &=& \sum_{i=1}^{11} ~ {2 |V_{12}^{i}|^2 \over \epsilon_2^{i} - \epsilon_1}.
\end{eqnarray}
Here $\epsilon_2^{i}$ is the energy level for the $i$-th $f^2 \Gamma_3$
state. Hence multiple $\Gamma_3$ states in the $f^2$ configuration
lead to the enhancement of the two-channel exchange coupling.
Particle-hole asymmetry in the conduction band density of states (DOS)
is also important in determining the ground state weights of the
one-channel ($f^0-f^1$) and the two-channel ($f^1-f^2$)
contributions. In the scaling approach, the particle-hole
asymmetry is completely neglected. However the NCA can
take into account this particle-hole asymmetry.
The occupied conduction electron states (hole) contribute to the
$f^0$ self energy (see the section III), while the unoccupied
states (particle) to the $f^2$ self energy. Hence the more
weight in the particle side can enhance the effective
hybridization strength between $f^1$ and $f^2$ configurations.
The particle-dominant conduction band DOS will lead to the
enhancement of the two-channel exchange coupling.
\section{Non-crossing approximation (NCA).}
We now apply the non-crossing approximation (NCA) \cite{nca}
to study our simple model system -- one-channel \& two-channel Anderson
model. Though this model is highly simplified compared with the full
model\cite{kimcoxtobe}, we can simulate the full model using this
simple model Hamiltonian to study the different physics for varying
channel number.
In the NCA, our starting basis is the conduction band plus
the atomic Hamiltonian
projected to the atomic electron Fock space and treat the hybridization
between the conduction band and the atomic orbital as a perturbation.
The strength of this approach is that the strong on-site Coulomb interaction
for atomic electrons is treated accurately at the outset.
Pseudo particle Green's functions are introduced for each atomic
electron occupation state which is neither fermionic nor bosonic.
The price we pay is that we cannot apply the conventional Feynmann
diagram techniques to this strongly correlated problems. Thus special
Green's function technique was developed by many investigators
\cite{nca,bcw,grewe,kuramoto,coleman,zhanglee}.
This approach may be justified by expansion
in $1/N$ which reorders the diagrams by treating $NV^2$ as ${\cal O}(1)$.
Here $N$ is the nominal atomic ground state degeneracy and $V$ is the
hybridization strength between the conduction band and the atomic orbitals.
In the NCA, pseudo particle self energy diagrams
include only the leading order skeleton (non-crossing) diagrams
and they are solved self-consistently.
For the one-channel models, this theory includes all the diagrams
up to ${\cal O} (1/N)$ order and a subset of higher order diagrams up to
the infinite order. Though vertex corrections, which is ${\cal O} (1/N^2)$,
are not included, this approach is a conserving approximation\cite{kuramoto}.
For the overscreened multi-channel Anderson models, it has been shown
\cite{coxruck} that the $1/N$ approach becomes exact in the limit
$M,N\to \infty$ with fixed $M/N$ ratio ($M$ is the number of channels).
When we study the most general three-configuration model,
the same symmetry conduction electron can be involved in the two mixing
processes, e.g., $f^0-f^1$ and $f^1-f^2$ for Ce$^{3+}$ atoms.
Generally, a specific vertex correction is required to get the right Kondo
energy scales in this case.
Recently, vertex corrections were included in the study of
the finite $U$ spin $1/2$ Anderson model\cite{pruschke}.
A simplifying feature of our model Hamiltonian
is that the leading vertex correction vanishes, since two different
symmetry conduction electrons are involved in the hybridizations
$f^0 - f^1$ ($\Gamma_7$) and $f^1 - f^2$ ($\Gamma_8$).
This feature greatly simplifies the numerical work and formalism.
From the leading order skeleton diagrams of Fig.~\ref{nca},
we find the self-consistent integral equations:
\begin{eqnarray}
\Sigma_0 (z)
&=& {\Gamma_{01} \over \pi} \sum_{\alpha} \int d\epsilon ~
\tilde{N} (\epsilon) f(\epsilon) G_1 (z+\epsilon), \\
\Sigma_1 (z)
&=& {\Gamma_{01} \over \pi} \int d\epsilon ~\tilde{N} (\epsilon)
f(-\epsilon) G_0 (z-\epsilon) \nonumber\\
&& + {\Gamma_{12} \over \pi} \sum_{n} \int d\epsilon ~\tilde{N} (\epsilon)
f(\epsilon) G_2 (z+\epsilon), \\
\Sigma_2 (z)
&=& {\Gamma_{12} \over \pi} \sum_{\alpha} \int d\epsilon ~
\tilde{N} (\epsilon) f(-\epsilon) G_1 (z-\epsilon), \\
G_N (z) &=& {1 \over z - \epsilon_N -\Sigma_N (z) }, ~~
\Gamma_{ij} \equiv \pi N(0) |V_{ij}|^2.
\end{eqnarray}
Here $\tilde{N} (\epsilon)$ is the normalized conduction band DOS at the
fermi level such that $\tilde{N} (0) =1$. $\Sigma_{0,1,2} (z)$ and
$G_{0,1,2} (z)$ are the self energy equations and Green's functions
for $f^0$, $f^1 \Gamma_7$, and $f^2 \Gamma_3$ atomic states, respectively.
$f(\epsilon)$ is the Fermi-Dirac distribution function. $\Gamma_{ij}$
is the hybridization strength characterizing the width of the renormalized
atomic electron spectral function peak. One of the strong points of the NCA
approach is that we can easily study any form of the conduction band DOS,
as opposed to other approaches, e.g., Bethe ansatz or conformal field theory,
to Anderson/Kondo models.
As a concrete example,
we will use the structureless Lorentzian/Gaussian DOS to see
the many body physics.
We solved these coupled integral equations numerically
to study the thermodynamics and the dynamics of the model Hamiltonian.
Pseudo particle Green's functions are not directly measurable. All the
physically measurable quantities are given by the convolution of the
pseudo particle Green's functions. Now it is convenient to introduce
the spectral function ($A_N (\omega)$) for each pseudo particle
Green's function and its negative frequency spectral function
($a_N (\omega)$).
\begin{eqnarray}
A_N (\omega)
&\equiv & - { 1 \over \pi} \hbox{Im}
{ 1 \over \omega - \epsilon_N - \Sigma_N (\omega) }, \\
a_N (\omega)
&\equiv& e^{-\beta \omega} ~ A_N (\omega).
\end{eqnarray}
Since $a_N (\omega)$ always appears in combination
with the impurity partition function $Z_{\rm f}$, there is arbitraryness
in overall factor in its definition. The above NCA integral equations
do not have $a_N (\omega)$ in them. Whenever
we calculate any measurable physical quantities, the spectral functions
appear with the Boltzman thermal factor divided by the impurity partition
function.
In the numerical work, we calculate $a_N (\omega)$'s self-consistently.
Roughly speaking, $a_N (\omega)$'s are occupancy
distribution function weighted by the thermal factor.
The impurity partition function for our simple model is defined by
\begin{eqnarray}
Z_{\rm f}
&\equiv& \int d\omega ~
\left[ a_0 (\omega) + 2 a_1 (\omega) + 2 a_2 (\omega) \right].
\end{eqnarray}
This partition function includes the many body effect of the
interaction between impurity and the conduction band and is exact
in its form. The approximation comes in when we choose
the approximate Green's functions for the atomic states.
In our simple model, only two symmetry electrons are present: $\Gamma_7$ and
$\Gamma_8$. Their spectral functions (measurable) are defined
as a convolution of two neighboring configurations' spectral functions.
It can be shown that these two interconfiguration atomic spectral
functions are given by
\begin{eqnarray}
\rho_{\Gamma_7} (\omega)
&=& { 1+ e^{-\beta\omega} \over Z_{\rm f} } \int d\zeta ~
a_0 (\zeta) ~ A_1 (\zeta + \omega) = \rho_{01} (\omega), \\
\rho_{\Gamma_8} (\omega)
&=& { 1+ e^{-\beta\omega}\over Z_{\rm f} } \int d\zeta ~
a_1 (\zeta) ~ A_2 (\zeta + \omega) = \rho_{12} (\omega)
\end{eqnarray}
in this conserving approximation\cite{kuramoto}.
{}From now on, we will use the notations
$\rho_{01}, \rho_{12}$ in favor of $\rho_{\Gamma_7}, \rho_{\Gamma_8}$.
This approximation does not include any
vertex correction.
From the leading bubble diagram\cite{bcw},
the static magnetic susceptibility per Ce$^{3+}$ impurity is
\begin{eqnarray}
\chi (T)
&=& {1\over 3} ~\mu_{\rm eff}^2 ~ \tilde{\chi}_{\rm f} (T),
~~ \mu_{\rm eff}^2 = {75 \over 49} ~ \mu_{\rm B}^2, \\
\tilde{\chi} (T)
&=& - {4 \over Z_{\rm f} } \int d\zeta ~ a_1 (\zeta,T)~
\hbox{Re} G_1 (\zeta,T).
\end{eqnarray}
Here $\mu_{\rm B}$ is Bohr magneton. The reduced dynamic magnetic
susceptibility is
\begin{eqnarray}
\tilde{\chi}^{''} (\omega, T)
&=& { 1 - e^{-\beta\omega} \over Z_{\rm f} }
\int d\zeta ~a_1 (\zeta,T) ~ A_1 (\zeta + \omega,T).
\end{eqnarray}
We are going to calculate these physical quantities and transport
coefficients.
\section{Zero temperature analysis.}
Zero temperature analysis\cite{coxruck,zero} of the NCA integral equations
leads to a qualitative understanding of our model system.
Recently, conformal field theory approach\cite{cft}
calculated the scaling dimensions
exactly for the multi-channel Kondo exchange models with any size of
impurity spin.
In Ref. \cite{coxruck}, the evaluation of the scaling dimensions is extended
to the multichannel Anderson/Coqblin-Schrieffer models using the functional
integral formulation. Both models become congruent when the impurity spin is
$S_I=1/2$.
The main results of the zero temperature analysis of the
NCA integral equations are:
(1) We can find the criterion which predicts whether the ground state is
that of the 1, 2, or 3 channel model.
(2) The Kondo energy scale ($T_0$) can be estimated analytically for the case
$\Gamma_{01} = \Gamma_{12}$. $T_0$ in the one- and two-channel model
parameter regimes is shown to vanish as the $f^2 ~ \Gamma_3$ energy level
approaches that of the $f^0$ configuration.
(3) We obtain the correct scaling dimensions for the overcompensated
cases which agree with the conformal field theory results\cite{ludwig}.
(4) The crossover physics between the parameter regimes for different
numbers of channels can be identified.
The self consistent NCA integral equations can be transformed into
the differential equations for the flat conduction band
in the wide band limit: $D \gg |\epsilon_{1,2} |$.
We analyze the zero temperature NCA equations in the asymptotic
limit $|\omega - E_0| \ll T_0$. $E_0$ is the threshold energy below
which the pseudo particle Green's functions become purely real.
We introduce the inverse Green's functions and transform the self
energy equations at zero temperature into the coupled non-linear
differential equations.\cite{zero}
\begin{eqnarray}
g_0 (\omega)
&\equiv& - 1/G_0 (\omega); ~~ g_1 (\omega)
\equiv - 1/G_1 (\omega); ~~ g_2 (\omega)
\equiv - 1/G_2 (\omega), \\
{d \over d\omega} g_0 (\omega)
&=& -1 - { 2 \Gamma_{01} \over \pi} { 1 \over g_1 (\omega) };
~~ g_0 (-D) = D, \\
{d \over d\omega} g_1 (\omega)
&=& -1 - { \Gamma_{01} \over \pi} { 1 \over g_0 (\omega) }
- { 2 \Gamma_{12} \over \pi} { 1 \over g_2 (\omega) };
~~ g_1 (-D) = D + \epsilon_1, \\
{d \over d\omega} g_2 (\omega)
&=& -1 - { 2 \Gamma_{12} \over \pi} { 1 \over g_1 (\omega) };
~~ g_2 (-D) = D + \epsilon_2.
\end{eqnarray}
Then $g_0$ and $g_2$ can be shown to be related by
\begin{eqnarray}
{g_2 \over \Gamma_{12}}
&=& { g_0 \over \Gamma_{01} }
+ \left[ {1\over \Gamma_{01}} - {1 \over \Gamma_{12} } \right] ~
( \omega - E_0 ) + \gamma_c, \\
\gamma_c
&=& {\epsilon_2 - E_0 \over \Gamma_{12}}
+ { E_0 \over \Gamma_{01}}.
\end{eqnarray}
In the zero temperature analysis, it is more convenient to define the
``negative frequency" spectral functions by $a_i (\omega) \equiv
e^{-\beta \omega} ~ A_i (\omega) / Z_{\rm f}$.
These spectral functions vanish above the threshold energy $E_0$ and satisfy
\begin{eqnarray}
{d \over d\omega} [~a_0 (\omega) |g_0 (\omega)|^2 ~]
&=& - { 2 \Gamma_{01} \over \pi} ~ a_1 (\omega), \\
{d \over d\omega} [~a_1 (\omega) |g_1 (\omega)|^2 ~]
&=& - { \Gamma_{01} \over \pi} ~a_0 (\omega)
- { 2 \Gamma_{12} \over \pi} ~ a_2 (\omega), \\
{d \over d\omega} [~a_2 (\omega) |g_2 (\omega)|^2~]
&=& - { 2 \Gamma_{12} \over \pi}~ a_1 (\omega).
\end{eqnarray}
It can be shown from the above relations that
\begin{eqnarray}
{d \over d\omega}
\left[~a_0 (\omega) g_0 (\omega)
+ 2 a_1 (\omega) g_1 (\omega) + 2 a_2 (\omega) g_2 (\omega) ~\right]
&=& a_0 (\omega) + 2 a_1 (\omega) + 2 a_2 (\omega).
\end{eqnarray}
By integrating this equation from $\omega = -\infty$ to $\omega = E_0$,
we find the additional relation,
\begin{eqnarray}
\left[~a_0 (\omega) g_0 (\omega)
+ 2 a_1 (\omega) g_1 (\omega) + 2 a_2 (\omega) g_2 (\omega)
\right]_{\omega = E_0} &=& 1.
\end{eqnarray}
This identity will be useful in finding the asymptotic behavior of the
``negative frequency" spectral functions. As a corollary, we have another
identity in the asymptotic limit,
\begin{eqnarray}
{a_0 (\omega) |g_0 (\omega)|^2 \over \Gamma_{01} }
&=& {a_2 (\omega) |g_2 (\omega)|^2 \over \Gamma_{12} }.
\end{eqnarray}
This relation can be proved by using the boundary condition
at $\omega = E_0$.
We calculate the physical atomic spectral functions and the
dynamic magnetic susceptibility defined in section IV.
\begin{eqnarray}
\rho_{01} (\omega)
&=& \int d\epsilon ~ \left[ a_0 (\epsilon) A_1 (\epsilon + \omega)
+ A_0 (\epsilon) a_1 (\epsilon + \omega) \right], \\
\rho_{12} (\omega)
&=& \int d\epsilon ~ \left[ a_1 (\epsilon) A_2 (\epsilon + \omega)
+ A_1 (\epsilon) a_2 (\epsilon + \omega) \right], \\
\tilde{\chi}^{''} (\omega)
&=& \int d\epsilon ~ \left[ a_1 (\epsilon) A_1 (\epsilon + \omega)
- A_1 (\epsilon) a_1 (\epsilon + \omega) \right].
\end{eqnarray}
{}From the $\omega$ dependence of the spectral functions near $\omega = 0$,
we can infer the finite temperature dependence of transport coefficients
as will be discussed below.
We now discuss the phase diagram in the model parameter space using the
zero temperature analysis. $\gamma_c$ decides the low energy and the low
temperature behaviors of our model Hamiltonian.
$\gamma_c$ measures the relative magnitude of the antiferromagnetic
coupling strengths when the charge fluctuation is removed in the model
Hamiltonian. Noting that $E_0 \approx \epsilon_1 + {\cal O} (V_{01}^2,
V_{12}^2)$, we find
\begin{eqnarray}
\gamma_c &\approx& {2\over \pi N(0)} ~
\left[ {1\over J_2} - {1\over J_1} \right]
\end{eqnarray}
which illustrates the correspondence to the scaling analysis.
If $\gamma_c$ is greater than zero, the singular behavior
shows up in the $f^0$ Green's function, and not in the $f^2$ Green's
function.
Hence the system will be dominated by the $f^0$ and $f^1$ sector leading to
the one-channel Kondo effect.
When $\gamma_c$ is less than zero, the singular behavior
shows up in the $f^2$ Green's function, and not
in the $f^0$ Green's function.
In this case $f^1$ and $f^2$ sector (the two-channel Kondo physics)
determines the low temperature behavior of
the system.
When $\gamma_c =0$, $f^0$ and $f^2$ become equivalent
asymptotically ($|\omega - E_0| \ll T_0$).
Both Green's functions develop the singular behaviors
at the ground state energy.
In this model parameter space, the three-channel Kondo model
is realized.
The characteristic Kondo energy scale $T_0$ is found from
an integration constant
which connects the low and high energy states. We can obtain the
integration constant for the case $\Gamma_{12} = \Gamma_{01} (=\Gamma)$.
We will analyze this case in detail and indicate subsequently
how to extend the zero temperature analysis
to the general hybridization strength.
\subsection{The symmetric hybridization limit:
$\Gamma_{12} = \Gamma_{01} (=\Gamma)$. }
When we take a symmetric hybridization limit
$\Gamma_{12} = \Gamma_{01} (=\Gamma)$, the equations are simplified.
\begin{eqnarray}
\gamma_c &=& {\epsilon_2 \over \Gamma}, \\
g_2 &=& g_0 + \epsilon_2, \\
a_0 g_0^2 &=& a_2 g_2^2.
\end{eqnarray}
The last two relations hold true in the asymptotic limit,
$|\omega - E_0| \ll T_0$.
The different ground states then are determined solely by the sign of
the $f^2$ configuration energy relative to the $f^0$ configuration
energy. Removing the variable $\omega$, we can find the
differential equations between the inverse Green's functions.
Integrating from their values at $\omega = -D$ to those at $\omega$, we
find the integration constant
\begin{eqnarray}
\exp\left[ {\pi \over 2 \Gamma} ~
[~ g_0 - g_1 + \epsilon_1 ~] \right]
&=& \left[ {g_1 \over D} \right] ~
\left[{g_0 \over D} \right]^{-1/2} ~
\left[{g_0 + \epsilon_2 \over D} \right]^{-1}
\end{eqnarray}
in the wide conduction band limit, $D >> |\epsilon_{1,2}|$.
We identify three cases for evaluating $T_0$. \\
(1) {\it One-channel case:} $\gamma_c > 0$ or $\epsilon_2 > 0$.
$f^0$ Green's function develops
a singular behavior at the threshold energy, while $f^2$ Green's function
does not. Thus the ground state physics is dominated by the sector
$f^0 - f^1$ leading to the one-channel Kondo effect.
The integration constant and the Kondo temperature are
\begin{eqnarray}
{g_1 \over T_0}
&=& \left[{g_0 \over \Delta} \right]^{1/2} ~
\left[ 1 + {g_0 \over \epsilon_2} \right] ~
\exp \left[ {\pi (g_0 - g_1) \over 2 \Gamma} \right], \\
T_0
&=& D ~ \left[{\epsilon_2 \over D} \right] ~
\left[{\Gamma \over \pi D}\right]^{1/2} ~
\exp \left[ {\pi \epsilon_1 \over 2 \Gamma} \right].
\end{eqnarray}
Here $\Delta \equiv \Gamma / \pi$.
We can find the asymptotic behavior of the Green's functions for each
atomic state.
\begin{eqnarray}
{g_0 (\omega) \over \Delta}
&\approx& \tilde{\Omega}^{\alpha_0}, \\
{g_1 (\omega) \over T_0}
&\approx& \tilde{\Omega}^{\alpha_1}, \\
g_2 (\omega)
&=& g_0 (\omega) + \epsilon_2, \\
\tilde{\Omega}
&=& 3{ E_0 - \omega \over T_0 }, \\
\alpha_0 &=& {2 \over 3}; ~~ \alpha_1 = {1 \over 3}.
\end{eqnarray}
Though the scaling dimension is not correct, the estimated Kondo
temperature is correct within order unity and $\chi (0) \sim 1/T_0$.
An interesting observation
is that Kondo temperature vanishes as $\epsilon_2$ tends to zero, i.e.,
the three-channel parameter regime (see the discussion below).
The detailed derivation of the asymptotic behavior is not included since
NCA does not produce the Fermi liquid ground state in the one-channel
model. We just give a brief summary of the zero temperature analysis
which is relevant to our study. $A_2(\omega)$ vanishes as
$\tilde{\Omega}^{2/3}$ at the threshold energy, while $A_0$ ($A_1$) is
singular as $\tilde{\Omega}^{-2/3}$ ($\tilde{\Omega}^{-1/3}$)
at $\omega = E_0$.
Thus, the physical spectral function $\rho_{12}$ vanishes
as $|\omega|^{4/3}$ at the Fermi energy, while $\rho_{01}$ is finite.
(2) {\it Two-channel case:} $\gamma_c < 0$ or $\epsilon_2 < 0$.
In contrast to the one-channel case,
the $f^2$ spectral function has a singular structure leading to two-channel
ground state. The integration constant and the Kondo temperature are
\begin{eqnarray}
{g_1 \over T_0}
&=& \exp\left[{\pi (g_2 - g_1) \over 2 \Gamma}\right] ~
\left[ 1 + {g_2 \over |\epsilon_2|} \right]^{1/2} ~
{g_2 \over \Delta}, \\
T_0 &=& D ~ \left[{|\epsilon_2| \over D }\right]^{1/2} ~
\left[ {\Gamma \over \pi D } \right] ~
\exp \left[{\pi (\epsilon_1 - \epsilon_2)
\over 2 \Gamma } \right].
\end{eqnarray}
Note that the Kondo temperature vanishes with $\epsilon_2 \to 0$.
We can find the asymptotic behavior following
the standard zero temperature analysis.
\begin{eqnarray}
g_0 (\omega) &=& g_2 (\omega) + |\epsilon_2|, \\
{g_1 (\omega) \over T_0}
&=& \tilde{\Omega}^{1/2} ~[~ 1 + c_1 \tilde{\Omega}^{1/2}
+ {\cal O} (\tilde{\Omega})~], \\
{g_2 (\omega) \over \Delta}
&=& \tilde{\Omega}^{1/2} ~[~ 1 + c_2 \tilde{\Omega}^{1/2}
+ {\cal O} (\tilde{\Omega})~], \\
\tilde{\Omega} &=& 4 {E_0 -\omega \over T_0}, \\
c_1 &=& {1\over 6} ~ \left[ 2\left(1 + {\Delta \over |\epsilon_2|}\right)
- {T_0 \over \Delta} \right]; ~~
c_2 = -{1\over 6} ~ \left[ \left(1 + {\Delta \over |\epsilon_2|}\right)
- 2 {T_0 \over \Delta} \right].
\end{eqnarray}
The asymptotic behavior above the ground state energy $E_0$ can be obtained
from the expressions below $E_0$ by the analytic continuation.
Furthermore, we find for the ``negative frequency" spectra that
\begin{eqnarray}
{d \over d\tilde{\Omega} } a_0 g_0^2
&=& {T_0 \Delta \over 2} ~ a_1, \\
{d \over d\tilde{\Omega} } a_1 g_1^2
&=& {T_0 \Delta \over 2} ~ [~ {1\over 2} a_0 + a_2~], \\
{d \over d\tilde{\Omega} } a_2 g_2^2
&=& {T_0 \Delta \over 2} ~ a_1,
\end{eqnarray}
which implies for $|\omega - E_0| \ll T_0$
\begin{eqnarray}
a_0 &=& {\Delta \over 4 |\epsilon_2|^2} ~\tilde{\Omega}^{1/2}
~[~ 1 - x_0 \tilde{\Omega}^{1/2} + {\cal O} (\tilde{\Omega})~], \\
a_1 &=& {1\over 4 T_0} ~\tilde{\Omega}^{-1/2}
~[~ 1 - x_1 \tilde{\Omega}^{1/2} + {\cal O} (\tilde{\Omega})~], \\
a_2 &=& {1\over 4 \Delta} ~\tilde{\Omega}^{-1/2}
~[~ 1 - x_2 \tilde{\Omega}^{1/2} + {\cal O} (\tilde{\Omega})~], \\
x_0 &=& {2\over 3} (2c_1 + c_2) + 2 {\Delta \over |\epsilon_2|}
= {1\over 3} ~\left(1 + {7\Delta \over |\epsilon_2|} \right), \\
x_1 &=& {4\over 3} (2c_1 + c_2) =
{2\over 3} ~\left(1 + {\Delta \over |\epsilon_2|} \right), \\
x_2 &=& {4\over 3} (c_1 + 2c_2) = {2\over 3} ~ {T_0 \over \Delta}.
\end{eqnarray}
Note that $x_{0,1,2} > 0$.
Now we can find the asymptotic behavior of the pseudo particle
spectral functions.
\begin{eqnarray}
A_0 (\omega)
&=& {\Delta \over \pi |\epsilon_2|^2} ~ \theta (\omega - E_0) ~
\left[ |\tilde{\Omega}|^{1/2} + {\cal O} (|\tilde{\Omega}|^{3/2} )
\right], \\
A_1 (\omega)
&=& {1\over \pi T_0} ~ \theta (\omega - E_0) ~
\left[ |\tilde{\Omega}|^{-1/2} + {\cal O} (|\tilde{\Omega}|^{1/2} )
\right], \\
A_2 (\omega)
&=& {1\over \pi \Delta} ~ \theta (\omega - E_0) ~
\left[ |\tilde{\Omega}|^{-1/2} + {\cal O} (|\tilde{\Omega}|^{1/2} )
\right].
\end{eqnarray}
As expected, $A_0 (\omega)$ vanishes at the threshold energy and does not
develop any singular behavior. On the other hand, $A_1 (\omega)$ and
$A_2 (\omega)$ are singular at the threshold energy.
Finally the physical spectral functions are given by in the asymptotic
limit.
\begin{eqnarray}
\rho_{01} (\omega > 0)
&=& {\Delta \over 32 |\epsilon_2|^2} ~
\left[\tilde{\omega}
- {8x_0 \over 3\pi} ~ \tilde{\omega}^{3/2} + \cdots \right], \\
\rho_{01} (\omega < 0)
&=& {\Delta \over 32 |\epsilon_2|^2} ~
\left[|\tilde{\omega}|
- {4x_1 \over 3\pi} ~ |\tilde{\omega}|^{3/2} + \cdots \right], \\
\rho_{12} (\omega > 0)
&=& {1\over 16 \Delta} ~
\left[ 1 - {2x_1 \over \pi} ~ \sqrt{\tilde{\omega}} + \cdots \right], \\
\rho_{12} (\omega < 0)
&=& {1\over 16 \Delta} ~
\left[ 1 - {2x_2 \over \pi} ~ \sqrt{|\tilde{\omega}|} + \cdots \right], \\
\chi^{''} (\omega)
&=& {\mbox{sign} (\omega) \over 16 \Delta} ~
\left[ 1 - {2x_1 \over \pi} ~ \sqrt{|\tilde{\omega}|} + \cdots \right], \\
\tilde{\omega} &=& 4 {\omega \over T_0}.
\end{eqnarray}
$\rho_{01} (\omega)$ vanishes at $\omega = 0$ and increases as $|\omega|$
near the fermi level. This spectral depletion at the fermi level
is also confirmed in the finite temperature NCA calculation.
$\rho_{12}$ is peaked right at the fermi level and has more spectral weight
below than above the fermi level since $x_2 < x_1$ for $T_0 < \Delta$.
The finite temperature NCA results also confirm this observation.
The spectral functions become non-analytic at the Fermi level at zero
temperature.
Note that the dynamic magnetic susceptibility is step function like at
$\omega = 0$, which is none other than marginal fermi liquid behavior
\cite{coxruck}.
{}From the asymptotic form of the zero temperature spectral functions, we
can infer the finite temperature dependence of the resistivity using the
Kubo formula
\begin{eqnarray}
{1\over \rho (T)}
&=& {ne^2 \over m} ~ \int d\omega \tau (\omega) ~
\left(-{\partial f \over \partial \omega} \right).
\end{eqnarray}
For $\omega, T \ll T_{0}$, the relaxation rate for the conduction electron
is approximately
\begin{eqnarray}
{1\over \tau (\omega)}
&\propto& \rho_{12} (\omega, T).
\end{eqnarray}
Near the zero temperature, we may replace $\rho_{12} (\omega, T)$ by our zero
temperature one and find the $\sqrt{T}$ temperature dependence.
In fact one complication arises due to the angular averaging.
Since the angular harmonics conjugate to $\rho_{12}$ is positive definite,
still the above simple argument applies. In the one-channel case, we cannot
use this kind of simple argument to find the low temperature behavior of
resistivity.
The scaling dimensions agree with those from the conformal field theory
approach.
We summarize the low temperature properties:
(i) The atomic spectral function $\rho_{12}$ is peaked right at the Fermi
level independent of the occupancy, while $\rho_{01}$ vanishes at
$\omega =0$ and is depleted near the fermi level;
(ii) The dynamic magnetic susceptibility is step function-like
at zero frequency;
(iii) The resistivity obeys the scaling behavior
$\rho (T) = \rho (0)~[~ 1 - a \sqrt{T/T_0} ~]$
near zero temperature.
All these results will be shown consistent with numerical NCA calculations
at finite temperatures.
(3) {\it Three-channel case:} $\gamma_c = 0$ or $\epsilon_2 = 0$.
In this case, the $f^0$ and
$f^2$ configurations are equivalent asymptotically leading to three-channel
ground state. The integration constant and the Kondo temperature are
\begin{eqnarray}
{g_1 \over T_0}
&=& \left[{g_0 \over \Delta} \right]^{3/2} ~
\exp \left[ {\pi \over 2 \Gamma} ~ (g_0 - g_1) \right], \\
T_0 &\approx& D ~
\left[{\Gamma \over \pi D }\right]^{3/2} ~
~ \exp \left[{\pi \epsilon_1 \over 2 \Gamma } \right].
\end{eqnarray}
See the Appendix for the detailed derivation of asymptotic behavior.
This model case is not different from the analysis of the general
overcompensated models. We simply list the asymptotic behavior here.
\begin{eqnarray}
g_0 (\omega) &=& g_2 (\omega), \\
{g_1 (\omega) \over T_0}
&=& \tilde{\Omega}^{3/5} ~ \left[
1 + c_1 ~ \tilde{\Omega}^{2/5} + d_1 ~ \tilde{\Omega}^{3/5}
+ \cdots ~\right], \\
{g_2 (\omega) \over \Delta}
&=& \tilde{\Omega}^{2/5} ~ \left[
1 + c_2 \tilde{\Omega}^{2/5} + d_2 ~ \tilde{\Omega}^{3/5}
+ \cdots ~\right], \\
\tilde{\Omega}
&=& 5 { E_0 - \omega \over T_0 }, \\
c_1 &=& {2\over 7}, ~~ d_1 = - {T_0 \over 8\Delta }; ~~
c_2 = - {1\over 7 }, ~~ d_2 = {T_0 \over 4\Delta }.
\end{eqnarray}
The nagative spectral functions are
\begin{eqnarray}
a_1 (\omega)
&=& {1 \over 5T_0}~ \tilde{\Omega}^{-3/5} ~
\left[1 - x_1 ~ \tilde{\Omega}^{2/5} - y_1 ~ \tilde{\Omega}^{3/5}
+ \cdots \right] \theta (\tilde{\Omega}), \\
a_2 (\omega)
&=& {1 \over 5\Delta}~ \tilde{\Omega}^{-2/5} ~
\left[1 - x_2 ~ \tilde{\Omega}^{2/5} - y_2 ~ \tilde{\Omega}^{3/5}
+ \cdots \right]
\theta (\tilde{\Omega}), \\
x_1 &=& 2c_1 = {4\over 7}, ~~y_1 = 0; ~~x_2 = 0, ~~
y_2 = 2 d_2 = {T_0 \over 2\Delta}.
\end{eqnarray}
And the pseudo particle spectral functions are
\begin{eqnarray}
A_1 (\omega)
&=& {\sin (3\pi/5) \over \pi T_0}~
|\tilde{\Omega}|^{-3/5} ~
\left[ 1 - x_1 ~\cos (2 \pi/5) ~
|\tilde{\Omega}|^{2/5} + \cdots \right]
\theta (-\tilde{\Omega}), \\
A_2 (\omega)
&=& {\sin (2 \pi/5) \over \pi \Delta}~
|\tilde{\Omega}|^{-2/5} ~
\left[ 1 - y_2 ~\cos (2\pi/5) ~
|\tilde{\Omega}|^{3/5} + \cdots \right]
\theta (-\tilde{\Omega}).
\end{eqnarray}
{}From the pseudo particle spectral functions, we can find the
physical spectral functions in the asymptotic limit.
\begin{eqnarray}
\rho_{01} (\omega)
&=& \rho_{12} (-\omega), \\
\rho_{12} (\omega > 0)
&=& {\sin(2\pi/5) \over 25 \pi \Delta}
~[~ B(2/5,3/5)
- x_1 ~ B(3/5,4/5)~ |\tilde{\omega}|^{2/5} \nonumber\\
&& \hspace{2.0cm} - y_2~ \cos(2\pi/5)~ B(2/5,6/5) ~|\tilde{\omega}|^{3/5}
+ \cdots ~], \\
\rho_{12} (\omega < 0)
&=& {\sin(3\pi/5) \over 25 \pi \Delta}
~[~ B(2/5,3/5)
- x_1~ \cos(2\pi/5)~ B(3/5,4/5) ~|\tilde{\omega}|^{2/5} \nonumber\\
&& \hspace{2.0cm}- y_2 ~ B(2/5,6/5) ~ |\tilde{\omega}|^{3/5} + \cdots ~], \\
\tilde{\chi}^{''} (\omega)
&=& \mbox{sign} (\omega) ~ {\sin(3\pi/5) \over 25 \pi T_0}
~[~ B(2/5,2/5) ~ |\tilde{\omega}|^{-1/5} \nonumber\\
&& \hspace{2.0cm}- x_1 ~[~1 + \cos(2\pi/5)~]~ B(2/5,4/5)
~ |\tilde{\omega}|^{1/5} + \cdots ~], \\
\tilde{\omega}
&=& 5 ~ {\omega \over T_0}.
\end{eqnarray}
Here $B(p,q)$ is the Beta function.
The spectral functions become non-analytic at the Fermi level at zero
temperature.
One important observation is that the power laws agree with those
obtained from the conformal field theory treatments for the overcompensated
multichannel $S=1/2$ models. Since $\rho_{01}$ and $\rho_{12}$ are
equivalent in the asymptotic limit, the angular functions are factored
out in the conduction electron scattering time. Thus we can read off the
power law of the resistivity, $\alpha = 2/5$.
Again, the low temperature properties are summarized below.
(i) The atomic spectral function is peaked right at the Fermi level
independent of the occupancy;
(ii) The dynamic magnetic susceptibility diverges as $\omega^{-1/5}$
at zero frequency;
(iii) The resistivity obeys the scaling behavior
$\rho (T) = \rho (0)~[~ 1 - a (T/T_0)^{0.4} ~]$
near zero temperature. All these results will be shown consistent
with numerical NCA calculations at finite $T$.
\subsection{Crossover physics.}
We expect the smooth crossover from the high temperature
regime to the low temperature regime where the relevant channel number
physics shows up. We can estimate the crossover temperature in the one- and
two-channel model parameter regime using the zero temperature
analysis. Noting that the asymptotic analysis assumes $g_0 << \epsilon_2$
for the one-channel regime and $g_2 << |\epsilon_2|$ for the two-channel
regime (see the integration constant), we can estimate the crossover energy
scale below which one- or
two-channel physics is realized. In the one-channel case, the integration
constant can be rewritten as
\begin{eqnarray}
{g_1 \over T_0^{(3)} }
&=& \left[ 1 + {\epsilon_2 \over g_0} \right] ~
\left[{g_0 \over \Delta} \right]^{3/2} ~
\exp \left[ {\pi (g_0 - g_1) \over 2 \Gamma} \right].
\end{eqnarray}
We can see that the relative magnitude of $g_0$ and $\epsilon_2$
determines the differing channel behavior. The crossover
energy scale can be defined by the relation $g_0 = \epsilon_2$.
The crossover temperature is found in the one-channel case
\begin{eqnarray}
T_{\rm x}^{(1)}
&=& {T_0 \over 3} ~ \left[{\epsilon_2 \over \Delta} \right]^{3/2}.
\end{eqnarray}
And in the two-channel case, we find the crossover temperature
\begin{eqnarray}
T_{\rm x}^{(2)}
&=& {T_0 \over 4} ~ \left[{|\epsilon_2| \over \Delta} \right]^2.
\end{eqnarray}
Substituting the Kondo energy scale, we can see that the crossover
energy scales are
\begin{eqnarray}
T_{\rm x}^{(1)}
&=& {1\over 3} ~ \left[ \epsilon_2 \over \Delta \right]^{5/2} ~
D ~ \left[{\Gamma \over \pi D} \right]^{3/2} ~
\exp \left[{\pi \epsilon_1 \over 2\Gamma} \right], \\
T_{\rm x}^{(2)}
&=& {1\over 4} ~ \left[ |\epsilon_2| \over \Delta \right]^{5/2} ~
D ~ \left[{\Gamma \over \pi D} \right]^{3/2} ~
\exp \left[{\pi (\epsilon_1 - \epsilon_2) \over 2\Gamma} \right].
\end{eqnarray}
Note that the crossover temperature vanishes as $|\epsilon_2|^{5/2}$
with $\epsilon_2 \to 0$. We intentionally write the crossover temperature
as the prefactor times the corresponding three-channel Kondo temperature.
If this prefactor is greater than 1, the system will not display the
three-channel behavior with decreasing temperatures but will flow directly
to the one- or two-channel fixed point.
When this prefactor is less than 1, the system
will display the three-channel behavior before finally flowing to each
channel fixed point.
In the one-channel regime, the system flows directly from the
high temperature regime (local moment regime) to the one-channel fixed
point when $\epsilon_2 >> \Delta$ with decreasing temperatures. In the
opposite limit, $\epsilon_2 << \Delta$, the system will flow from the
local moment regime to the three-channel regime, and finally to
the one-channel fixed point after passing through the crossover temperature.
In the two-channel regime the same argument can be applied as in
the one-channel case.
\subsection{General asymmetric hybridization case.}
Though the Kondo energy scale can not be estimated analytically
for the general case $\Gamma_{01} \neq \Gamma_{12}$,
still the asymptotic behavior ($\omega \leq E_0$) can be sorted out. Since
the physics is the same as the above analysis,
we shall just write the asymptotic
properties of Green's functions. \\
(1) $\gamma_c > 0$:
\begin{eqnarray}
g_0 (\omega)
&\approx& {\Gamma_{01} \over \pi}
\left[ 3{ E_0 - \omega \over T_0 } \right]^{\alpha_0},
\\
g_1 (\omega)
&\approx& T_0 \left[ 3
{ E_0 - \omega \over T_0 } \right]^{\alpha_1}, \\
g_2 (\omega)
&=& {\Gamma_{12} \over \Gamma_{01}} ~ g_0 (\omega)
+ \left[ {\Gamma_{12} \over \Gamma_{01}} - 1 \right] ~
( \omega - E_0 ) + \gamma_c ~ \Gamma_{12} , \\
\alpha_0 &=& {2 \over 3}; ~~
\alpha_1 = {1 \over 3}.
\end{eqnarray}
\\
(2) $\gamma_c < 0$:
\begin{eqnarray}
g_0 (\omega)
&=& {\Gamma_{01} \over \Gamma_{12}} ~ g_2 (\omega)
+ \left[ {\Gamma_{01} \over \Gamma_{12}} - 1 \right] ~
( \omega - E_0 ) + |\gamma_c|~\Gamma_{01}, \\
g_1 (\omega)
&\approx& T_0 \left[ 4
{ E_0 - \omega \over T_0 } \right]^{\alpha_1}, \\
g_2 (\omega)
&\approx& {\Gamma_{12} \over \pi}
\left[ 4{ E_0 - \omega \over T_0 } \right]^{\alpha_2},
\\
\alpha_1 &=& {1 \over 2}; ~~
\alpha_2 = {1 \over 2}.
\end{eqnarray}
\\
(3) $\gamma_c = 0$: In this case, $f^0$ and $f^2$ are equivalent
asymptotically.
\begin{eqnarray}
g_2 (\omega)
&=& {\Gamma_{12} \over \Gamma_{01}} ~ g_0 (\omega)
+ \left[ {\Gamma_{12} \over \Gamma_{01}} - 1 \right] ~ (\omega - E_0 )
\\
{d \over d\omega} g_0 (\omega)
&=& - 1 - { 2 \Gamma_{01} \over \pi} { 1 \over g_1 (\omega) };
~~ g_0 (-D) = D, \\
{d \over d\omega} g_1 (\omega)
&=& -1 - {\Gamma_{01} \over \pi}
~\left[ { 1 \over g_0 (\omega) }
+ { 2 \over g_0 (\omega) + (1- \Gamma_{01}/ \Gamma_{12} )
(\omega - E_0) }
\right]; ~~ g_1 (-D) = D + \epsilon_1.
\end{eqnarray}
Then the asymptotic form follows
\begin{eqnarray}
g_0 (\omega)
&\approx& {\Gamma_{01} \over \pi}
\left[ 5{ E_0 - \omega \over T_0 } \right]^{\alpha_0},
\\
g_1 (\omega)
&\approx& T_0 ~ \left[ 5{ E_0 - \omega \over T_0 } \right]^{\alpha_1}, \\
\alpha_0 &=& {2 \over 5}; ~~
\alpha_1 = {3 \over 5}.
\end{eqnarray}
Though the Kondo energy scale cannot be evaluated analytically, it can
be estimated from the numerical
NCA calculation of, e.g., magnetic susceptibility
and fitting it onto the universal results.
\section{Numerical Analysis.}
In this section we present results from our full numerical study at
finite temperatures.
We studied the model for the parameters listed in Table \ref{modelpar}.
This covers the
one-, two-, and three-channel Kondo regimes. For simplicity, we have chosen
the same hybridization strength in the $f^0 - f^1$ and the $f^1 - f^2$
sectors. The relevant channel-number-dependent physics can then be studied by
varying the
relative position of the $f^0$ and the $f^2$ configuration energies.
This simple choice of the hybridization further makes it possible to find
the characteristic Kondo energy scales analytically, which is described
in the previous section in detail.
Our main results are:
(1) The magnetic susceptibility shows a scaling behavior and
agrees well with the exact Bethe ansatz results in the two-
and three-channel parameter regimes;
(2) The NCA results of residual entropy in the two-, and three-channel
models agree with the exact ones within order ${\cal O} (1/N)$. The Kondo
anomaly peak in the specific heat also agrees with the exact one
in its magnitude;
(3) The thermopower is a diagnostic to display the ground states
for different numbers of relevant channels for our model;
(4) The dynamic magnetic susceptibility is distinct between the one-channel
model and overcompensated (two- and three-channel) models;
(5) Due to the simplifying features of our model, the resistivity shows
a bendover at low temperatures in the one-channel parameter regime.
The resistivity in the two- and
three-channel parameter regimes shows temperature dependences
near $T = 0K$ agreeing with the conformal field theory results;
(6) We confirmed in detail that NCA is a valid
numerical self-consistent non-perturbative method
in studying the overcompensated
multi-channel $S_I=1/2$ Anderson model at $T>0$.
\subsection{Entropy and specific heat.}
The entropy and the specific heat due to the magnetic impurity can be
calculated from the free energy obtained in NCA through numerical
differentiation.
These thermodynamic quantities include very important information
about the nature of the ground state. We can estimate the characteristic
energy scales in the Kondo models from the temperature variation of
the entropy. In general, the entropy will increase with increasing
temperatures, until the frozen impurity degrees of freedom are released.
Our model Hamiltonian is expected to have the entropy $S = k_B \ln 5$
at high enough temperatures, while the residual entropy at zero temperature
will show behavior expected for different numbers of relevant channels.
The entropy and the specific heat are displayed in Fig.~\ref{entropy} and
\ref{spheat} for the model
parameter sets studied here. In the one-channel case, the Kondo anomaly
peak is well separated from the Schottky anomaly peak coming from the
interconfiguration excitations. The Kondo anomaly peak has a magnitude
comparing well with the exact results\cite{sacramento}
for the Kondo exchange model.
No residual entropy remains with $T\to 0$.
In the two-channel case, the Kondo anomaly peak is not clearly
separated from the charge fluctuation peak for most of our
model parameters.
The residual entropy agrees within 5 \% with the exact
one\cite{sacramento}
and the discrepancy can be explained by the ${\cal O}(1/N^2)$ and
higher order corrections. To see this, we note that the entropy
for spin $1/2$ conduction electrons has an explicit dependence
on the impurity spin degeneracy\cite{cft,ludwig}.
Strictly speaking, the NCA results are valid for $N$-fold impurity
spin and $M$-fold channel degeneracies as $N\to \infty$ with
$N/M$ fixed. Hence it is natural to expect ${\cal O} (1/N^2)$
corrections to the entropy through the neglected vertex
corrections.
In the three-channel case, the Kondo anomaly peak is reduced
further due to the increased residual entropy and is almost merged with
the charge fluctuation contribution for model set 4. For model set 5,
we can see a weak indication of separation. The residual entropy is increased
compared to that of two-channel case and its magnitude is again a little bit
smaller than the exact one\cite{sacramento}
due to the neglect of the higher
order contribution in $1/N$ expansion.
\subsection{Static magnetic susceptibility.}
The static magnetic susceptibility is a direct indicator of
the nature of the ground state for the magnetic Kondo model.
As is well documented,
the magnetic susceptibility diverges logarithmically
$\chi(T) \propto \log (T_0 /T)$ for the two-channel $S_I=1/2$
magnetic Kondo model at zero temperature,
and diverges algebraically for the three
or the higher channel $S_I=1/2$ Kondo model
($\chi (T) \propto \left[ T_0 /T \right]^{1-2\Delta_n};
{}~~\Delta_n = 2/(n+2)$: $n$ is the channel number ($\geq 3$))
\cite{cft,ludwig,sacramento}.
The one-channel Kondo model has
a constant magnetic susceptibility at zero temperature ($\chi(0) \sim
1/T_0$). From the NCA, using the leading order bubble diagram\cite{bcw},
the magnetic susceptibility for one Ce$^{3+}$ impurity is given by
\begin{eqnarray}
\chi (T)
&=& {1\over 3} \mu_{\rm eff}^2 ~ \tilde{\chi} (T),
~~ \mu_{\rm eff}^2 = {75 \over 49} ~ \mu_{\rm B}^2, \\
\tilde{\chi} (T)
&=& - {4 \over Z_{\rm f} } \int d\zeta ~ a_1 (\zeta,T)~
\hbox{Re} G_1 (\zeta,T).
\end{eqnarray}
Here $\mu_{\rm B}$ is the Bohr magneton.
Our numerical results for
$\tilde{\chi} (T)$ clearly show the right tendency
for each possible low $T$ channel number. The magnetic susceptibility in the
one-channel
regime, for the model parameter sets 6, 7, 8 shows
approximate scaling behavior and clearly has the negative
curvature at low temperatures indicating approach to the Fermi liquid ground
state. The deviation of the scaling behavior
at low temperatures seems to come from the pathological behavior of NCA
in this one-channel case\cite{zero}.
The $\tilde{\chi} (T)$ curves in the two-channel regime, for
the model parameter sets 1, 2, 3, also show scaling
behavior and diverge logarithmically at low temperatures (Fig.~\ref{chi}).
Our results
are compared with those of Ref.\cite{sacramento}. The fitting to
the exact Bethe ansatz numerical results\cite{sacramento} is quite good.
We believe that the high temperature deviation comes
from the $M=3$ to $M=2$ crossover physics described in the previous section.
Note that Bethe ansatz results are for the pure two-channel s-d
exchange model. To get the fitting to the Bethe ansatz, we slide
the temperature axis to find $T_K = 0.3 \times T_0$
($T_K$ from the Bethe Ansatz). Here $T_0$ is the
Kondo energy scale estimated from the zero temperature analysis.
At high temperatures we cannot distinguish
the different numbers of relevant channels clearly.
Note that distinct physics for different
numbers of relevant channels
shows up at low temperatures below the crossover temperature which
was estimated in the previous section. This observation is supported
by results from the three-channel model parameter sets. The high temperature
deviation from scaling is very weak in this case which will not show
the crossover physics.
The magnetic susceptibility in the three-channel model also shows
a scaling behavior.
Since the three-channel case lies exactly on the boundary between
the one-channel and the two-channel regime, we probed the three-channel
case by varying the position of the bare $f^1$ configuration energy while
the two excited configurations $f^0, f^2$ energies are kept equal.
Fitting to the exact results is quite good. The Kondo energy scale $T_0$
estimated from the zero temperature analysis agrees with that in the
exact Bethe ansatz ($T_K$ of Ref\cite{sacramento}).
\subsection{Spectral functions.}
The interconfiguration spectral functions show
a distinct behavior for different numbers of relevant channels
near the the Fermi level ($\omega=0$) depending on the model
parameters. In our simple model, we have two symmetry
atomic spectral functions of $\Gamma_7$ and $\Gamma_8$:
$\rho_{\Gamma_7} (\omega) = \rho_{01} (\omega)$
and $\rho_{\Gamma_8} (\omega) = \rho_{12} (\omega)$.
\begin{eqnarray}
\rho_{01} (\omega)
&=& { 1+ e^{-\beta\omega} \over Z_{\rm f} } \int d\zeta ~
e^{-\beta \zeta} ~ A_0 (\zeta) ~ A_1 (\zeta + \omega) \\
\rho_{12} (\omega)
&=& { 1+ e^{-\beta\omega}\over Z_{\rm f} } \int d\zeta ~
e^{-\beta \zeta} ~ A_1 (\zeta) ~ A_2 (\zeta + \omega)
\end{eqnarray}
In the one-channel model parameter regime (see Fig.~\ref{rho1ch}),
$\rho_{01} (\omega)$ develops
the Kondo resonance peak just above the Fermi level and
$\rho_{12} (\omega)$ is depleted near $\omega =0$ and tends to zero
with $T\to 0$ at $\omega = 0$. This confirms our zero temperature
analysis in section IV. For comparison, note that the Kondo
resonance amplitude is big compared to the two-channel or three-channel
cases.
In the two-channel regime (see Fig.~\ref{rho2ch}),
$\rho_{12} (\omega)$ is peaked below $\omega =0$ and its peak
position tends to $\omega = 0$ with decreasing temperature. Note that
the Kondo resonance amplitude is reduced compared to the one-channel
Kondo resonance amplitude. On the other hand,
$\rho_{01}(\omega)$ is depleted near $\omega = 0$.
In the three-channel parameter regime (see Fig.~\ref{rho3ch}),
the two spectral functions are equivalent asymptotically when
$\omega \to -\omega$ transformation is accounted for. Peak positions
tend to zero with decreasing temperature. Both develop
the Kondo resonance peak with further reduced amplitude.
As mentioned above, the positions of the Kondo resonance peak
(see Fig. \ref{kondopeak})
show a distinct behavior for different numbers of relevant channels.
The peak position saturates to a constant value in the one-channel
cases while it vanishes in the two- and three-channel cases with
decreasing temperature. In addition, the detailed functional
forms of the Kondo resonance peak
are different depending on the relevant channel numbers.
In the overscreened cases ($M=2,3$), the peak structure becomes
non-analytic with decreasing temperature as shown in the
zero temperature analysis of the NCA integral equations.
Finite temperature washes out this non-analytic behavior
at the Kondo resonance peak. In the one-channel case,
the atomic spectral functions remain analytic down to
zero temperature.
Since we are not considering all the atomic energy levels, the
full atomic spectral functions (measured from the photoemission
experiments) cannot be defined in our simple model. Though
the high energy physics of real systems can not be properly
treated within our simple model Hamiltonian, the low energy or
low temperature properties can be studied
using the restricted spectral functions.
Note that low temperature/energy physics is governed by the
Kondo resonance peak development.
Though the spectral depletion is found in the interconfiguration
spectral functions, we do not expect that the photoemission spectroscopy
can observe this feature unless it can distinguish the atomic electron
symmetry. Measurable atomic spectral functions
are shown in Fig.~\ref{rhof} for each relevant channel case.
\subsection{Dynamic magnetic susceptibility.}
The dynamic magnetic susceptibility
measures the magnetic excitation structure.
Since the properties of the magnetic excitations are related to the
interaction of the local magnetic moment with the conduction electrons,
the relevant channel number will determine the nature of the magnetic
excitations.
The dynamic magnetic susceptibility is expected to be distinct for
each number of relevant channels.
We have already seen this channel dependence
in the static magnetic susceptibility.
Dynamic magnetic susceptibility is defined as the spin-spin correlation
function and can be measured directly from the neutron scattering
experiments. From the leading bubble diagram, the reduced dynamic magnetic
susceptibility is
\begin{eqnarray}
\tilde{\chi}^{''} (\omega, T)
&=& { 1 - e^{-\beta\omega} \over Z_{\rm f} }
\int d\zeta ~a_1 (\zeta,T) ~ A_1 (\zeta + \omega,T).
\end{eqnarray}
Our reduced static magnetic susceptibility is related to the above magnetic
response function by the Kramer-Kronig relation:
\begin{eqnarray}
\tilde{\chi} (T)
&=& 2 \int {d\omega \over \pi} ~
{ \tilde{\chi}^{''} (\omega, T) \over \omega }.
\end{eqnarray}
The neutron scattering experiments measure the structure function
$S(\omega, T) \propto ~[~b(\omega) + 1~]~\chi^{''} (\omega, T)$.
The dynamic susceptibility can be quantitatively characterized
by its `linewidth' dependence on temperature in addition to its
overall functional shape.
We may define the linewidth $\Gamma (T)$
by the peak position of $\chi^{''} (\omega, T)$, which
can be measured directly in the inelastic neutron
scattering experiments.
The variation of $\chi^{''} (\omega, T)$ and $\Gamma (T)$
with temperature are displayed in Fig.~\ref{dchi} and \ref{gamma}.
For the dynamic magnetic susceptibility, we display the reduced
form $\tilde{\chi}^{''} (\omega,T) / \tilde{\chi}^{''} (\Gamma(T),T)$
as a function of $\omega/\Gamma (T)$.
The distinct behavior for different number of relevant channels
is clearly evidenced: in the one-channel
regime, the $\tilde{\chi}^{''} (\omega, T)$ curves converge with decreasing
temperatures (not shown). This is clearly supported by the saturation of
$\Gamma (T)$ at low temperatures. In the two- and three-channel regimes,
$\Gamma (T)$ vanishes algebraically (close to linear)
and the dynamic magnetic susceptibility does not
converge (not shown) in contrast to the one-channel case,
but instead develops non-analytic behavior at $\omega =0$.
The reduced dynamic magnetic susceptibilities (defined above,
see Fig.~\ref{dchi})
roughly show a scaling behavior between two extrema.
The physics of the linewidth is quite important in understanding the
nature of the magnetic spin screening. The distinct behavior for different
numbers of channels
shows up at low temperatures below $T_0$. The impurity
spin flipping time ($\tau_{\rm f}$) due to the hybridization will be given
by the inverse of the linewidth, $\tau_{\rm f} \sim 1/\Gamma (T)$.
On the other hand,
thermally excited conduction electrons close to the impurity site
will pass through it in a time $\tau_c$ of
order of $1/T$ from the uncertainty principle.
At high temperatures above the Kondo temperature, $\tau_{\rm f} \gg \tau_c$.
Thus the impurity spin rarely flips while the conduction electrons
pass by the impurity site. Hence the Curie behavior is expected
in the magnetic susceptibility. At low temperatures below the Kondo
temperature, differing channel number physics shows up. In the one-channel
regime, $\tau_{\rm f} \ll \tau_c$ and the impurity spin flips
vigorously averaging out its spin to zero
and leading to the Pauli behavior. In the two- and three-channel regimes,
$\tau_{\rm f} \approx \tau_c$ and spin screening is
not complete
leading to the non-Fermi liquid ground state. This interpretation
agrees with the low temperature behaviors of magnetic susceptibility.
\subsection{Transport coefficients.}
Using the Kubo formula\cite{bcw} in a dilute impurity limit, where
the inter-impurity correlation can be neglected, we have calculated
the transport coefficients, resistivity and thermopower.
The anisotropic conduction electron scattering rate is
\begin{eqnarray}
\tau^{-1} (\hat{k}, \omega)
&=& {8\pi n_{\rm imp} \over N(0) }
~[~\Gamma_{01}~\Theta_7^{(5/2)} (\hat{k}) ~ \rho_{01} (\omega)
+ \Gamma_{12}~\Theta_8^{(5/2)} (\hat{k}) ~ \rho_{12} (\omega) ~],
\\
\Theta_7^{(5/2)} (\hat{k})
&=& {1\over 16\pi} ~[~ 6 - \Phi (\hat{k}) ~]; ~~
\Theta_8^{(5/2)} (\hat{k}) = {1\over 16\pi}
~[~ 6 + \Phi (\hat{k}) ~], \\
\Phi (\hat{k})
&=& 15 \cos^4 \theta - 10 \cos^2 \theta + 1
+ 5\sin^4 \theta \cos^2 2\varphi.
\end{eqnarray}
Here $n_{\rm imp}$ is the impurity concentration.
Note that the conduction electron scattering rate contains the
hybridization since the interconfiguration spectral functions are
involved. The crystal harmonics are normalized such that the integration of
$\Theta_7^{(5/2)} (\hat{k})$ and $\Theta_8^{(5/2)} (\hat{k})$
over the solid angle leads to 1 and 2, respectively.
Note that the the crystal harmonic $\Theta_7^{(5/2)} (\hat{k})$
vanishes at ``hot spot" angles of
$(\theta, \varphi) = (0,-)$, $(\pi, -)$, $(\pi/2, 0)$,
$(\pi/2, \pi/2)$,$(\pi/2, \pi)$,$(\pi/2, 3\pi/2)$, while $\Theta_8^{(5/2)}
(\hat{k})$ is positive definite.
This feature, combined with the dip structure of
the interconfiguration spectral function
($f^1 - f^2$ or $\rho_{\Gamma_8} (\omega)$),
leads to a bendover in resistivity at low temperatures in the one-channel
parameter regime within our simplified model. We believe this feature
will go away in more realistic models.
For example, another contribution to the
$\Gamma_8$ atomic spectral function comes from the convolution between
$f^0$ and $f^1~J=5/2~\Gamma_8$. This spectral function will not be
depleted at the fermi level, but instead will build up its spectral
weight due to the weak Kondo resonance structure just above $\omega =0$.
Thus the low temperature bendover in the resistivity
will disappear.
The channel-dependence of the spectral functions, combined with the
angular dependence, leads to a distinct behavior in the angle-averaged
conduction electron lifetime for different numbers of channels.
The spectral depletion becomes pronounced
in the one-channel regime due to the angular average over the $\Gamma_7$
``hot spots." The development
of the dip structure just below the Fermi level is correlated to
the resistivity bendover at low temperatures. This feature does not occur
in the two- and three-channel regimes.
A Kondo resonance related peak develops near
the Fermi energy, whose position with respect to the Fermi level
depends on the
channel character of the model parameters.
Details are displayed in Fig.~\ref{tau}.
Note that the anisotropic scattering
rate includes explicitly the hybridization as opposed to the other
spectral functions like the atomic electron spectral function and the
dynamic susceptibility.
The transport coefficients are calculated using the Kubo formula under
the assumption of dominant scattering in the $l=3$ channel. Our results
are defined in terms of transport integrals $I_n$ given by
equations.
\begin{eqnarray}
I_n (T) &=& \int d\omega ~ \omega^n ~ \tau (\omega, T) ~
\left[ -{\partial f(\omega) \over \partial \omega} \right],
\\
\tau (\omega, T)
&\equiv& \int {d\hat{k} \over 4\pi} ~ \tau (\hat{k}, \omega, T).
\end{eqnarray}
Here $f(\omega)$ is Fermi function.
The resistivity is calculated using the equation:
\begin{eqnarray}
{1\over \rho (T) } &=& {ne^2 \over m} ~ I_0 (T).
\end{eqnarray}
As we stated before, the resistivity shows a bendover
at low temperatures in the one-channel regime within our simple model
(not shown). This derives from the depleted
spectral weight in the $f^1 - f^2$ sector and zeros of
$\Theta_7^{(5/2)} (\hat{k})$.
In the two-channel and three-channel model parameter regimes,
the resistivity initially increases logarithmically through $T_0$
and then saturates with a power law to a constant
with further decreasing temperatures.
The resistivity near zero temperature obeys a scaling behavior as shown in
Fig.~\ref{res23ch} confirming our zero temperature analysis with the scaling
dimensions, $\Delta_2 = 1/2, \Delta_3 = 2/5$. These results
agree with the conformal field theory results\cite{cft,ludwig},
$\rho (T) = \rho(0) ~[~1- a [T/T_0]^{\Delta_n}~]$ for $T\le 0.06T_0$,
$~~\Delta_n = 2/(n+2)$
for the overcompensated multi-channel Kondo models (note that the power
law exponent is independent of the impurity spin size).
We note that the region where strict $T^{\Delta_n}$ behavior holds
is below about $0.05~T_0$. A fit to the resistivity of
Ce$_x$La$_{1-x}$Cu$_{2.2}$Si$_2$ (see Fig. \ref{resfit}) is good until
low temperatures where the data breaks below theory.
This suggests a possible crossover to a new fixed point which could be
set by intersite interactions (producing a spin molecular field)
or a weak noncubic symmetry for the Ce$^{3+}$ ions.
The thermopower is a sensitive measure of the asymmetry in
the scattering rate and the density of states (DOS)
near the Fermi level. Since we are assuming the symmetrical
Lorentzian/Gaussian
DOS for the conduction band, the sign of the thermopower
is determined by the asymmetrical scattering rate.
\begin{eqnarray}
Q(T) &=& - {1\over eT} ~ {I_1 (T) \over I_0 (T) }.
\end{eqnarray}
As shown in Fig.~\ref{tep}, the thermopower shows explicitly
the distinct behavior for different numbers of channels
varying from the two-channel regime,
through the three-channel , finally to the one-channel regime.
The hybridization between $f^0$ and $f^1$ leads to the Kondo resonance
peak just above the Fermi level
and more spectral weight above the Fermi level.
Since particle scattering dominates, holes are the main carriers in the
one-channel cases, thus leading to a positive thermopower.
On the other hand, negative thermopower arises in
the $f^1 - f^2$ sector. In this case, hole scatterings
dominate and particles are the main carriers.
In the three-channel regime, energy structures are symmetric.
However, the double degeneracy in the $f^2$ configuration leads to
a weak hole scattering dominance resulting in a negative thermopower at low
temperatures. The overall magnitude of the thermopower is
slightly reduced when anisotropy at the cubic sites is included.
\section{Discussion and Conclusion}
We have introduced and studied a realistic model Hamiltonian for Ce$^{3+}$
impurities with three configurations ($f^0$, $f^1$, $f^2$), which
are embedded in cubic normal metals. This simple model shows competition
between the Fermi liquid fixed point of the one-channel $S=1/2$ Kondo
model and the non-Fermi liquid fixed point of the two-channel $S=1/2$
Kondo model.
We studied a simplified Anderson model using the NCA.
This simple model covers one-, two-, and three-channel Kondo
physics depending on the model parameters. All the calculated
physical quantities show the signatures of the Kondo effect
appropriate to the different numbers of relevant channels.
The magnetic susceptibility agrees with
the exact Bethe Ansatz results for the two- and three-channel model
parameter regime and has the correct scaling dimension agreeing
with the conformal field theory results.
Entropy and specific heat calculations also
agree with the Bethe ansatz results within ${\cal O} (1/N^2)$ approximation.
Though the resistivity in the one-channel regime bends over
with decreasing temperatures, we do not believe that this feature
will survive when all the energy level structures, especially
$f^1 J=5/2 \Gamma_8$, are included.
In the two- and three-channel parameter regimes,
the resistivity increases logarithmically
and saturates with decreasing temperatures. The low temperature
behavior leads to power laws in agreement with the conformal
field theory results.
Since the thermopower is very sensitive to the density of states
structure and the
scattering mechanism near the Fermi level, its sign and its magnitude
depend on which fixed point is stable at zero temperature.
In the one-channel regime, electrons are strongly scattered due to the
Kondo resonance above the Fermi level and the thermopower remains positive
definite and large due to the resonant scattering. In the two-channel case,
holes are scattered off the impurity sites stronger than electrons. Thus
electrons are the main carriers leading to the negative thermopower. In the
three-channel regime, though hole scattering is reduced compared to
the two-channel case, electrons are still the main carriers due to the
degeneracy imbalance between the singlet $f^0$ and the doublet
$f^2 J=4 \Gamma_3$.
The thermopower remains negative with reduced value.
We calculated the dynamic magnetic susceptibility and characterized it with
its peak position ($\Gamma (T)$) as a function of temperature.
We can see the clear difference between one-channel and the overcompensated
cases. While $\Gamma (T)$ decreases and saturates to a constant value of order
$T_K$ with decreasing temperatures in the one-channel case,
$\Gamma (T)$ goes to zero almost linearly with decreasing temperatures
in the two- and three-channel case.
We now discuss the experimental relevance of our model study to the
Ce$_{1-x}$La$_x$Cu$_{2.2}$Si$_2$ alloy\cite{andraka}.
We already stressed the experimental evidences supporting
the two-channel Kondo effect in this alloy system in the section I.
The thermopower for CeCu$_{2}$Si$_2$ changes sign around 70 K and is
negative and large below\cite{chitep,ceybtep}.
As our numerical calculation shows, the thermopower is negative and
large in the two-channel regime. This result compares well with
the experimental findings for the stoichiometric system with $x=1$.
We believe the sign change comes from the Kondo resonance of
$f^0$ and $f^1 \Gamma_8$, which lies above the Fermi level.
Further experiments are required for the alloy system with excess Cu.
For comparison, we note that
CeAl$_2$ or CeAl$_3$\cite{ceybtep} have a positive thermopower
large compared to transition
metals at high temperatures and have a sign change at low temperature
which is still bigger than the Kondo temperature.
Our thermopower calculation and the thermopower dependence on the
unit cell volume\cite{jaccard} suggests that the alloy system
Ce$_{1-x}$La$_x$Cu$_{2.2}$Si$_2$ can go through the three-channel
model parameter regime with external pressure.
Renormalized atom calculations further suggest a destabilization of
$f^2$ relative to $f^0$ with initial increasing pressure
\cite{herbst}.
In addition to the thermopower,
the future neutron scattering experiments for
Ce$_{1-x}$La$_x$Cu$_{2.2}$Si$_2$ alloy
can search for the dependence of $\Gamma (T)$ (the peak position of the
dynamic magnetic susceptibility) consistent with our proposal for
the channel number.
\acknowledgments
This research was supported by a grant
from the U.S. Department of Energy, Office of Basic Energy Sciences,
Division of Materials
Research. We thank Eunsik Kim for her careful reading of this paper
and thank L. N. Oliveira and J.W. Wilkins for
stimulating interactions.
|
\section{Introduction}
\indent\indent Now that the existence of the top quark is firmly
established,\cite{TOP} we should begin to explore the opportunities for
top-quark physics. This is an enormous topic, as evidenced by the two-day
workshop devoted to top-quark physics following this Symposium. In this
talk, I restrict myself to top-quark physics at Fermilab, both in the
immediate and the more-distant future. My emphasis is on top-quark physics
in the next ten years or so, during which Fermilab will have a monopoly
on the top quark. In a final section I speculate on the role of Fermilab
during the LHC era.
The talk is divided into several subsections:
\begin{itemize}
\item Top-quark yields
\item Mass
\item Decay
\item Production
\item Not-so-rare decays
\item Speculations
\end{itemize}
The top-quark mass from combining the measured CDF and D0 values is
$179 \pm 12$ GeV. For definiteness, I use $m_t=175$ GeV throughout this talk.
\section{Top-quark yields}
\indent\indent The machine parameters and running schedule of the
Fermilab Tevatron are given in Table~1. Run I is now coming to a close,
and each experiment will have accumulated an integrated luminosity in
excess of 100 $pb^{-1}$ by the end of the run. The peak luminosity
achieved thus far is about ${\cal L}=2 \times 10^{31}/cm^2/s$, impressive for
a machine that was designed for ${\cal L}=10^{30}/cm^2/s$.
\begin{table}[ht]
\begin{center}
\caption[fake]{Schedule and machine parameters for Run I and II at Fermilab.}
\bigskip
\begin{tabular}{cc}
\underline{Run I} & \underline{Run II} \\
\\
1992-1995 & 1999- \\
$\sqrt s = 1.8$ TeV & $\sqrt s = 2$ TeV \\
${\cal L}=2 \times 10^{31}/cm^2/s$ & ${\cal L}=2 \times 10^{32}/cm^2/s$ \\
$\int {\cal L} dt >$ 100 $pb^{-1}$ & $\int {\cal L} dt >$ 1 $fb^{-1}$ \\
\end{tabular}
\end{center}
\end{table}
Run II will begin in late 1998/early 1999, with a machine energy of 2 TeV.
The increase in energy is made possible by cooling the magnets to a lower
temperature, thereby allowing a higher field strength. This increases
the top-quark production cross section by about $35\%$.
The most important change that will occur in Run II is a ten-fold
increase in luminosity, to ${\cal L}=2 \times 10^{32}/cm^2/s$. This
will be achieved by two additions to the existing accelerator complex:
\begin{itemize}
\item Main Injector: The original Main Ring in the Tevatron collider
tunnel is a bottleneck to higher luminosity. It will be replaced by the
Main Injector, a 120 GeV synchrotron housed in a separate tunnel, now
under construction. The Main Injector will enable the production of many
more antiprotons, yielding a five-fold increase in luminosity.
\item Recycler: \cite{FOSTER,J} A new development within the past year is the
addition
of another element to the Main Injector project, the Recycler ring. It is
an 8 GeV, low-field, permanent-magnet ring which will be installed in the
Main Injector tunnel. The primary function of the Recycler is to allow
more efficient accumulation of antiprotons. Its secondary role, from
which it takes its name, is to allow the reuse of antiprotons left over
from the previous store. The Recycler will yield a two-fold increase in
luminosity.
\end {itemize}
There will also be a variety of detector upgrades for Run II. One of the
most significant is an improved silicon vertex detector (SVX),
used to detect secondary vertices from $b$ quarks. This is of obvious
importance for top physics, since the top quark decays via $t\to W b$.
Both CDF and D0 will have an SVX in Run II, and they will be longer than
the existing CDF SVX, allowing for nearly $100\%$ acceptance of $b$ quarks
from top decays. The silicon detector will also be more sophisticated,
providing a stereo view of the events. This will increase the SVX tagging
efficiency of fiducial $b$ jets ($p_T>20$ GeV and within the SVX) from the
present value of about $40\%$ up to nearly $60\%$. As a result of these
improvements, the fraction of top events with at least one $b$ tag will
increase from $50\%$ to $85\%$. The fraction with two $b$ tags will
increase dramatically, from $13\%$ to $40\%$.\cite{DAN}
Taken together, the improvements in the accelerator and the detectors
will result in a dramatic increase in the potential for top-quark physics
in Run II. In $t\bar t$ events, the final state with the most kinematic
information is
$W+4j$, where the $W$ is detected via its leptonic decay. These events
are fully reconstructable. To reduce backgrounds, it is best to demand at
least one $b$ tag. The number of such events is about 500/$fb^{-1}$. The
number of events with two $b$ tags, which have very small background, is
about 250/$fb^{-1}$. Depending on the length of Run II, the integrated
luminosity delivered to each detector will be between 1 and a few
$fb^{-1}$. Thus there will be on the order of 1000 tagged,
fully reconstructed top-quark events in
Run II, to be compared with the approximately 20 $W+4j$ single-tagged top
events
in Run I.
\section{Mass}
\indent\indent Due in part to their SVX, CDF has the best measurement of the
top-quark mass,
$m_t = 176 \pm 8 \pm 10$ GeV. It is anticipated that the errors will be
reduced to $\pm 6 \pm 8$ GeV at the end of Run I. If one assumes that
the error scales like the inverse of the square root of the number of
events, the error will be reduced to $\pm 3$ GeV in Run II. It may be
optimistic to assume that all the systematic errors scale like the
statistical error, so $\Delta m_t \sim 3-5$ GeV is a more conservative
prognosis.\cite{DAN}
There will also be an improved measurement of the $W$ mass in Run II, as
well as a measurement at LEP II. An error of $\pm 50$ MeV is anticipated
from each experiment, to be compared with the current error of $\pm 180$
MeV.\cite{WMASS} Figure~1 shows the well-known plot of $M_W$ vs. $m_t$,
with bands of constant Higgs mass. The contours show the one- and
two-sigma fits to data from LEP and SLC. The large cross indicates the
present direct measurement of $M_W$ from CDF and UA2, and $m_t$ from CDF
and D0. The small cross indicates the errors expected in Run II; $\Delta
M_W = 50$ MeV, $\Delta m_t = 5$ GeV, placed arbitrarily on the plot.
Note that the length of the Run II $\Delta M_W$ and $\Delta m_t$
error bars are similar. Since the lines of
constant $m_H$ are sloped towards the horizontal, a reduction of
the uncertainty in $M_W$ yields more sensitivity to the Higgs mass than a
reduction of the uncertainty in $m_t$, a point I will return to in the
last section.
\begin{figure}[htb]
\begin{center}
\epsfxsize= 0.8\textwidth
\leavevmode
\epsfbox{fig1.ps}
\end{center}
\caption{$W$ mass vs. top-quark mass, with bands of constant Higgs mass.
The contours are the one- and two-sigma regions from precision LEP and SLC
data.
The large cross is the direct measurement of $M_W$ and $m_t$. The small cross,
placed arbitrarily on the figure, is the anticipated uncertainty in $M_W$ and
$m_t$ in Run II. Adapted from Ref.~5.}
\end{figure}
There is another perspective on the top-quark mass that is interesting to
consider. Ultimately we want to find a theory of fermion masses, and we
can ask how well we want to know the top-quark mass to help pin down
this theory. It is reasonable to strive for a measurement of $m_t$ which
is as good (fractionally) as the best-known quark mass. This is the $b$ mass,
which is
$\overline{m_b}(\overline{m_b}) =4.0 \pm 0.1$ GeV,\footnote{This is the running
$\overline{MS}$ mass evaluated at the quark mass.} extracted from the
Upsilon spectrum
calculated with lattice QCD.\cite{NRQCD} The top-quark mass is
already the second best-known quark mass. Since the uncertainty in
$\overline{m_b}$ is entirely theoretical, one can anticipate that it will be
reduced
by perhaps a factor of two, corresponding to an uncertainty of $\pm 1.3\%$.
This is comparable to a 3 GeV uncertainty in the top-quark mass.
So $\Delta m_t \sim 3$ GeV is a good benchmark.
\section{Decay}
\indent\indent This section and the next are devoted to studying the
decay and production of the top quark. To gain some perspective, I will
first ask:
\medskip
Is the Top Quark Exotic?
\medskip
\noindent There are two extreme viewpoints on this question:
\begin{itemize}
\item Yes. The top quark is much heavier than the other known fermions,
and its mass is close to the electroweak scale (e.g., the Higgs-field
vacuum-expectation value). It seems likely that the top quark is related
to electroweak symmetry breaking. This point of view is embodied, for
example, by top-quark-condensate models.\cite{TOPCON}
\item No. The top-quark Yukawa coupling to the Higgs field is close to
unity, a natural value. The other known fermions have Yukawa couplings
$<<1$, which must be explained. This point of view is embodied, for
example, by grand-unified Yukawa-matrix models.\cite{DHR}
\end{itemize}
The way to decide this issue is to study the properties of the top quark.
If the top quark is exotic, a study of its properties may reveal that fact.
One can imagine that non-standard interactions of fermions are
proportional to the fermion mass, in which case the top quark is the best
hope to discover such effects.\cite{PZ,M}
With 1000 fully-reconstructed $t\bar t$ events, the statistical accuracy
on the measurement of top-quark properties should be around $3\%$.
Including a comparable systematic error, we anticipate a knowledge of
the properties of the top quark at about the $5\%$ level at the end of
Run II.
\begin{figure}[htb]
\begin{center}
\epsfxsize= 0.25\textwidth
\leavevmode
\epsfbox{fig2.eps}
\end{center}
\caption{The matrix element of the top-quark charged current, probed via
top-quark decay.}
\end{figure}
The weak decay of the top quark is pictured in Fig.~2. The decay
involves the matrix element of the top-quark charged current. Using only
Lorentz invariance, we can write down the structure of this current in
terms of four \footnote{There are two additional form factors,
\begin{displaymath}
\bar u(b)[\frac{i}{2m_t}F^3_L\sigma^{\mu\nu}P_\nu (1-\gamma_5) +
\frac{i}{2m_t}F^3_R\sigma^{\mu\nu}P_\nu (1+\gamma_5)]u(t)
\end{displaymath}
where $P=p_t-p_b$. However, these terms do not contribute to the top-quark
decay amplitude if the $W$ boson decays to massless fermions.}
form factors \cite{KLY}
\begin{eqnarray}
\bar u(b)\Gamma^\mu u(t)& = &\frac{g}{2\sqrt 2} V_{tb}
\bar u(b)[F^1_L\gamma^\mu(1-\gamma_5) +
F^1_R\gamma^\mu(1+\gamma_5) \nonumber \\
& & \;\;\;\;\;\;\;\;\;\; - \frac{i}{2m_t}F^2_L\sigma^{\mu\nu}q_\nu (1-\gamma_5)
-
\frac{i}{2m_t}F^2_R\sigma^{\mu\nu}q_\nu (1+\gamma_5)]u(t) \nonumber
\end{eqnarray}
where the form factors $F^{1,2}_{L,R}(q^2)$ are evaluated at $q^2=M_W^2$.
These form factors are calculable in the standard model, and are given
by \footnote{The form factors $F_1^R$ and $F_2^L$ are
zero to all orders in the standard model in the limit $m_b =0$.}
\begin{eqnarray}
F^1_L & = & 1 + {\cal O}(\alpha) + {\cal O}(\alpha_s) \nonumber \\
F^1_R & = & F^2_L\;\; =\;\; F^2_R\;\; =\;\; 0 + {\cal O}(\alpha)
+ {\cal O}(\alpha_s) \nonumber
\end{eqnarray}
An example of a measurement which provides information on the form
factors is the fraction of top decays in which the $W$ boson is
longitudinal (helicity zero) in the top-quark rest frame.\cite{KLY}
Consider the case where we set $F^1_R=F^2_L=0$, as is true in the standard
model in the limit $m_b=0$. The ratio of
the longitudinal and transverse partial widths is given by
\begin{displaymath}
\frac{\Gamma_L}{\Gamma_T}=\frac{m_t^2}{2M_W^2}
\frac{\left|1+\frac{M_W^2}{2m_t^2}\frac{F^2_R}{F^1_L}\right|^2}
{\left|1+\frac{1}{2}\frac{F^2_R}{F^1_L}\right|^2} \nonumber
\end{displaymath}
which grows quadratically with
the top-quark mass. However, the quantity that is measured is the
branching ratio of the top quark to longitudinal $W$ bosons,
\begin{displaymath}
B(t\to W_Lb)=\Gamma_L/(\Gamma_L+\Gamma_T) \nonumber
\end{displaymath}
which is much less sensitive to the top-quark mass. At leading order in the
standard model, $B(t\to W_Lb)= 0.70$. A measurement of this quantity to
$5\%$ corresponds to an uncertainty in the top-quark mass of 15 GeV, much
greater than the uncertainty in a direct measurement. Our ability to
predict this branching ratio is therefore not limited by the uncertainty
in the top-quark mass.
The form factor $F^2_R$ is non-zero in the standard model, arising dominantly
from gluon loops.\cite{L}
QCD decreases the ratio
$\Gamma_L/\Gamma_T$ by about $6\%$, which decreases the longitudinal
branching ratio by about $2\%$, to $B(t \to W_Lb) =
0.69$.\footnote{This also includes the effect of real gluon emission.\cite{L}}
A measurement of $B(t \to W_Lb)$ to $5\%$ is sensitive to a non-standard value
of $|F^2_R/F^1_L|>0.2$.
\section{Production}
\indent\indent The QCD production of top-quark pairs occurs via the
quark-antiquark annihilation and gluon-fusion processes.
The quark-antiquark annihilation process accounts for about
$80\%$ of the cross section at the Tevatron. This process is sensitive
to the gluon coupling to top quarks,\cite{AKR,HNW} and to resonances which
might occur in this channel.\cite{HP} The gluon-fusion process,
although suppressed, could be
greatly enhanced if there is a resonance, such as a techni-eta.\cite{AT,EL}
The measured cross section is within one sigma of the band of theoretical
predictions,\cite{LSV,BC} so there is no indication of new physics in the
production of top-quark pairs at this time.
There are two processes which produce a single top quark, rather than a
$t\bar t$ pair: the $W$-gluon-fusion process,\cite{DW,Y,EP} depicted in
Fig.~3(a),
and $q\bar q \to t\bar b$,\cite{CP,SW} shown in Fig.~3(b). Both involve
the weak interaction, so they are suppressed relative to the QCD
production of $t\bar t$; however, this suppression is partially compensated by
the presence of only one heavy particle in the final state. Both processes
probe the charged-current weak interaction of the top quark.
The single-top-quark production cross
sections are proportional to the square of the Cabbibo-Kobayashi-Maskawa
matrix element $V_{tb}$, which cannot be measured in top quark decays since
the top quark is so short-lived.
\begin{figure}[htb]
\begin{center}
\epsfxsize= 0.6\textwidth
\leavevmode
\epsfbox{fig3.eps}
\end{center}
\caption{Single-top-quark production at hadron colliders: (a) $W$-gluon fusion;
(b) quark-antiquark annihilation.}
\end{figure}
The single-top-quark processes lead to a final state of $Wb\bar b$ (plus
an additional jet, for $W$-gluon fusion).
The backgrounds are more serious for the single-top-quark processes than for
$t\bar t$, but they are manageable. The dominant background is $Wb\bar
b$ from ordinary QCD/weak interactions. Fig.~4 shows a recent study of the
signal and backgrounds for $W$-gluon fusion; \cite{BH} the distribution
of the reconstructed top-quark mass is plotted for $Wjj$ events with
a single $b$ tag.\footnote{Most of the events in the signal contain one
$b$ jet plus the spectator jet from the radiation of the virtual $W$
boson.} Fig.~5 shows a similar plot for the process $q\bar q \to t\bar
b$, but with two $b$ tags.\cite{SW} Both processes should be observed
in Run II. The process $q\bar q \to t\bar b$ will yield a measurement of
$|V_{tb}|$ which is limited mostly by statistics; a $10\%$ measurement
may be possible in Run II.
\begin{figure}[htb]
\begin{center}
\epsfxsize= 0.5\textwidth
\leavevmode
\epsfbox{fig4.ps}
\end{center}
\caption{Signal and backgrounds for single-top-quark production via
$W$-gluon
fusion at the Tevatron, via $Wjj$ with a single $b$ tag. From Ref.~25.}
\end{figure}
\begin{figure}[htb]
\begin{center}
\epsfxsize= 0.5\textwidth
\leavevmode
\epsfbox{fig5.ps}
\end{center}
\caption{Signal and backgrounds for single-top-quark production via
$q\bar q \to t\bar b$ at the Tevatron, via $Wjj$ with a double $b$ tag.
{}From Ref.~24.}
\end{figure}
\section{Not-so-rare decays}
\indent\indent The top quark can also provide a window into new physics via
its decay to new particles. In this section I discuss several decays
which could occur at the few to tens of percent level. These decays
could be accessible even in Run I.
The decay of the top quark to a charged Higgs boson is a well-known decay
mode, and has been sought for several years.\cite{CHARGED}
The charged Higgs is sought via its decay to
$\bar\tau\nu$ or $c\bar s$. These are the dominant decay modes in a
multi-Higgs doublet model with natural flavor conservation.\cite{GW} However,
it is conceivable that the Higgs coupling to fermions are
generation-changing, in which case the decay $H^+\to c\bar b$ could
dominate.\cite{HW} If the other top quark decays conventionally,
this would give rise to events with three $b$ quarks, which could be
distinguished by tagging all three $b$ jets.
Continuing along this line of thought, one can also imagine tree-level
flavor-changing neutral-current decays of the top quark, such as $t \to
ch^0$.\cite{HW,H} Flavor-changing neutral currents are severly
restricted in the first two generations of fermions, but could be large
in the third generation. The dominant decay of the neutral Higgs would
likely be $h^0 \to b\bar b$. If the other top quark decays
conventionally, the events would again have three $b$ quarks in the final
state.
If the top squark is light, it could potentially be discovered in
top-quark decays.\cite{BDGGT,MKKW} The decay mode
$t \to \tilde t_1 \tilde\chi^0_1$
could have a significant branching ratio ($\tilde\chi^0_1$ is the lightest
neutralino). If it is kinematically
allowed, the top squark will decay via $\tilde t_1 \to \tilde\chi^+_1 b$;
otherwise, the loop-induced decay $\tilde t_1 \to \tilde\chi^0_1 c$ would
dominate. The extraction of these signals from backgrounds is challenging.
\section{Speculations}
\indent\indent As discussed in the section on top-quark yields,
the Main Injector and
the Recycler will allow the Tevatron to achieve a luminosity of $2\times
10^{32}/cm^2/s$ in Run II. One can ask if even higher luminosity can be
achieved. The answer seems to be yes.\cite{FOSTER} A design exists
which would achieve a luminosity of at least $10^{33}/cm^2/s$. The
main requirement to achieve this luminosity is to increase the rate of
antiproton production. This can be attained by directing more bunches
from the Main Injector onto the antiproton-production target, a technique
called ``multibatch targeting''.\cite{FOSTER2} The cost of this scheme is
modest in
comparison with the Main Injector, and could be in place for Run II.
A luminosity of $10^{33}/cm^2/s$ would produce about $5,000$ tagged and
fully-reconstructed top-quark pairs per year. One can imagine pushing
the uncertainty on the top-quark mass down to $2$ GeV, and the accuracy
on the measurement of the top-quark properties down to $2-3\%$. There
are other physics opportunities which become available as well. The
error on the $W$ mass could potentially be pushed down to 20 MeV. This
may be even more interesting than improving the accuracy on the top-quark
mass, as remarked in section 3. The production of the Higgs boson in
association with a $W$ boson, followed by $H\to b\bar b$, may also become
accessible, in the mass range $m_H = 80-120$ GeV.\cite{SMW,BBD,MK}
Given the physics opportunities afforded by ${\cal L} = 10^{33}/cm^2/s$,
why don't we do it? The stumbling block is not the accelerator, but the
detectors, which cannot operate at such a high luminosity. Significant
detector upgrades, or perhaps a new detector, are needed to take
advantage of this luminosity. One can even imagine this occuring during the
LHC era, especially if some of the physics objectives of such a machine are
complementary to the LHC.
Can one contemplate a luminosity for a $p\bar p$ collider as high as
${\cal L} = 10^{34}/cm^2/s$, the LHC design luminosity? There doesn't
seem to be any reason why not. If such a luminosity can be attained, it
might remove the advantage of $pp$ colliders over that of $p\bar p$. A
$p\bar p$ collider requires only a single ring of magnets, so it can
potentially be built more economically than a $pp$ collider, which
requires either two rings, or a 2-in-1 magnet such as for the LHC.
Magnets are a significant fraction of the cost of an accelerator; they
account for roughly two thirds of the cost of the LHC, for example.
The next hadron collider after the LHC might be a return
to $p\bar p$.\cite{FOSTER}
\section*{Acknowledgements}
\indent\indent I am grateful for conversations with D.~Amidei, W.~Foster,
G.~Jackson, and T.~Liss, and for assistance from P.~Baringer, A.~Heinson,
R.~Keup, and J.~Womersley.
This work was supported in part by Department of Energy grant
DE-FG02-91ER40677.
|
\section{Introduction}
Recently, {\em Parasupersymmetry} \cite{r-s,b-d,p8} and {\em Fractional
Supersymmetry} \cite{frac} have been attracting much attention.
This may be best explained by noting the achievements of workers
in {\em Supersymmetry} \cite{susy}. Afterall parasupersymmetry and fractional
supersymmetry may be viewed as generalizations of the ordinary
supersymmetry. This is most easily seen in the structure of the
defining algebraic expressions.
For the case of ($p=2$)-parasupersymmetry, it is shown, in the
most general setting, that the degeneracy structure is almost fully
determined by the defining parasuperalgebra \cite{p8}. In fact, for
a large class of ($p=2$)-parasupersymmetric quantum systems one
can even define the analog of the Witten index \cite{witten}
of supersymmetry \cite{p8}. Like the Witten index, this integer
is a topological invariant linked to the indices of Fredholm (resp.~
elliptic) operators for the known cases \cite{p9}. Physically,
it signifies the exactness or breaking of parasupersymmetry \cite{p8}.
These indications of the similarity between supersymmetry
and parasupersymmetry urges one to seek for a better understanding of both
the classical and quantum versions of parasupersymmtry.
Parasupersymmetric quantum mechanics (PSQM) has been studied to some extent
in the framework of specific quantum mechanical examples
\cite{r-s,b-d,khare}. Its classical counterpart, however, has not
been studied properly, to the best of author's knowledge. A discouraging
factor in such a study would be the complicated algebraic structure of
the associated {\em para-Grassmann variables}. The latter were introduced
in the study of {\em parastatistics} \cite{o-k} which is directly
related with parasupersymmetry.
In 1953, Green \cite{green} proposed a generalization of quantum field
theory that allowed for dynamical fields with generalized statistics
or parastatistics. Such theories were studied in a series of articles
in 60's and 70's before the advent of supersymmetry \cite{g-m,o-k}.
Parastatistics of Green has found some application in string theory
\cite{a-m} and provided an alternative point of view for theories with
internal symmetries \cite{gro}. A thorough reveiw of the subject is
provided in Ref.~\cite{o-k}.
To relate the new ($p=2$)-PSQM with the old parastatistics of Green, one may
begin with a study of its classical analog. The corresponding classical
parasupersymmetric systems will involve para-Grassmann variables $\psi$ of
order 2, i.e., $\psi^3=0$. As is the case for ordinary fermionic variables
($\psi^2=0$), the Lagrangian formulation is most convenient to study such
systems. This observation stems from the fact that fermionic coordinates,
due to the form of their kinetic term in the Lagrangian, are proportional
with their corresponding conjugate momenta. Thus these systems are indeed
constrained and the proper treatment of the constraints is necessary in
their (Hamiltonian) canonical quantization \cite{goodman}.
The Lagrangian formulation lacks the apparent difficulties with these first
class constraints. A quantization scheme applicable in the framework of
Lagrangian mechanics was proposed by Peierls \cite{peierls} for bosonic
systems, and generalized for fermionic and superclassical systems by
De~Witt \cite{bd}. For a demonstration of the application of this method
to supersymmetric systems see Refs.~\cite{bd,p5}.
The aim of the present article is to provide a simple formalism
which would allow for a concise and unified treatment of both parafermionic
and parabosonic dynamical variables of order 2. This allows for a
development of the Lagrangian formulation of para-classical mechanics and
a generalization of Peierls quantization scheme.
In section~2 a brief reveiw of the algebra of creation and annihilation
operators for parafermionic and parabosonic degrees of freedom
and their classical counterparts is provided.
Section~3 specializes to the case $p=2$. Here
a parabracket is introduced which summarizes the algebra of Green's
components and unifies the treatment of both types of degrees of freedom.
Section~4 first discusses the Peierls bracket quantization scheme for
classical systems involving ordinary fermionic and bosonic variables.
A generaliztion of this approach for ($p=2$) Green's components is then
proposed. In section~5, the Peierls bracket quantization is applied
to a simple one-dimensional parafermi system. The requirement
of the consistency of the canonical and Peierls quantization methods leads
to the determination of the kinetic term in the Lagrangian. Section~6
includes author's final remarks.
\section{Green's Parastatistics}
As defined in Ref.~\cite{gro}, parafermionic (parabosonic) statistics
of order $p$, is a type of statistics -- called {\em parastatistics} -- which
restricts the number of identical particles in a totally symmetric
(resp.~ antisymmetric) state to be at most $p$. Clearly, for $p=1$,
one recovers the ordinary fermionic or Fermi-Dirac (bosonic or Bose-Einstein)
statistics.
Parastatistics is generally signified with the following set of algebraic
relations \cite{green,g-m,o-k}:
\begin{eqnarray}
[ a_k,[ a_l^\dagger,a_m]_\mp ] &=&2\delta_{kl}a_m
\;,\nonumber\\
{[} a_k , {[} a_l^\dagger,a_m^\dagger {]_\mp} {]} &=&2\delta_{kl}
a_m^\dagger\mp 2\delta_{km}a_l^\dagger\;,
\label{q1}\\
{[} a_k,{[} a_l,a_m{]_\mp}{]} &=&0\;,\nonumber
\end{eqnarray}
where $a_k^\dagger$ and $a_k$ denote the creation and annihilation
operators, $\ll x,y\right] _\mp :=xy\mp yx\;,~~\forall x,y$, and the signs
$-$ and $+$ correspond to parafermions and parabosons, respectively.
In general, the order $p$ of parastatistics appears as the label
of a representation of the algebra {\large\bf$a$}, generated by
$a_k$ and $a_k^\dagger$ with the rules (\ref{q1}). An irreducible
representation is provided by choosing a unique vacuum state vector $|0\rangle$:
\[ a_k|0\rangle=0\;,~~~~\forall k\;,\]
and constructing a Hilbert-Fock space ${\cal A}$ using the basic vectors:
\[ |k_1,\cdots,k_l\rangle:=a_{k_1}^\dagger\cdots a_{k_l}^\dagger
|0\rangle\;.\]
One can show \cite{g-m,o-k} that in this representation
\[ a_ka_l^\dagger=p\delta_{kl}|0\rangle\;,\]
for some non-negative integer $p$.
In his original article, Green \cite{green} proposed another
(reducible \cite{g-m}) representation of the algebra {\large\bf$a$},
which involved bilinear algebraic relations rather than the complicated
trilinear relations (\ref{q1}). Green defined the algebra
{\large\bf$b$}, generated by the generators $\zeta^\alpha_k$,
and $\zeta^{\alpha\dagger}_k$, $\alpha=0,\cdots,p-1$, and rules:
\begin{eqnarray}
\ll \zeta_k^\alpha,\zeta^{\alpha\dagger}_j\right] _\mp
&=&\delta_{kj}\;,\nonumber\\
\ll \zeta_k^\alpha,\zeta^{\alpha}_j\right] _\mp
&=&0\;,
\label{q2}\\
\ll \zeta_k^\alpha,\zeta^{\beta\dagger}_j\right] _\pm
&=& \ll \zeta_k^\alpha,\zeta^{\beta}_j\right] _\pm
\:=\: 0\;,~~~~(\alpha\neq\beta)\,.\nonumber
\end{eqnarray}
These relations together with the identification:
\begin{equation}
a_k=\sum_{\alpha=0}^{p-1}\zeta_k^\alpha\;,
\label{q3}
\end{equation}
lead to the defining relation of {\large\bf$a$}, i.e., Eqs.~(\ref{q1}).
Choosing the same vacuum state vector, $|0\rangle$, requiring:
\[
\zeta_k^\alpha|0\rangle=0\;~~~~\forall \alpha,k\;,\]
and defining a Hilbert-Fock space ${\cal B}$ using the basic vectors:
\[
|k_1,\alpha_1;\cdots ;k_m,\alpha_m\rangle:=\zeta_{k_1}^{\alpha\dagger}
\cdots \zeta_{k_m}^{\alpha_m\dagger}|0\rangle\;,\]
one obtains a representation of {\large\bf$b$}. In view of Eq.~(\ref{q3}),
this also provides a representation for {\large\bf$a$}. This representation
is known as the {\em Green representation} and $\zeta^\alpha_k$ are
called the {\em Green components} of $a_k$. Note that by construction
the spaces ${\cal A}$ and ${\cal B}$ are in one-to-one correspondence
with the polynomial rings generated by $a_k^\dagger$ and
$\zeta^{\alpha\dagger}_k$, respectively. Thus according to Eq.~(\ref{q3})
${\cal A}$ may be viewed as a subring of ${\cal B}$. In fact, the physical
space to be considered is ${\cal A}$ and not ${\cal B}$. The latter is
introduced for practical convenience.
Another important point is the possibility of the existence of particles
with different types of parastatistics. This is especially the case
for parasupersymmetric systems. The treatment of this case leads to
the introduction of relative parastatistics \cite{g-m}.
Consider two species of particles $a$ and $b$, with creation and
annihilation operators $a_i^\dagger,~a_i$ and $b_j^\dagger,~b_j$,
and orders of parastatistics $p_a$ and $p_b$, respectively. Then,
it can be shown \cite{g-m} that if $p_a\neq p_b$, the
operators $a_i^\dagger$ and $a_i$ either commute or anticommute with
$b_j^\dagger$ and $b_j$. If $p_a=p_b=:p$ then there is a set of
trilinear relations between these operators. The latter can be more
easily expressed in terms of the corresponding Green's components:
\begin{equation}
a_i=\sum_{\alpha=0}^{p-1}\zeta_i^\alpha~,~~~
b_j=\sum_{\alpha=0}^{p-1}\xi_j^\alpha \;,
\label{q4}
\end{equation}
with $a_i|0\rangle=b_j|0\rangle=\zeta^\alpha_i|0\rangle=\xi_j^\alpha|0\rangle=0$ and
$i=1,\cdots,n_a$ and $j=1,\cdots,n_b$, for some positive integers
$n_a$ and $n_b$. The following relations express the relative
parastatistics of species of particles $a$ to $b$:
\begin{eqnarray}
[\zeta_i^\alpha,\xi_j^\alpha]_{-\eta}&=&
[\zeta_i^\alpha,\xi_j^{\alpha\dagger}]_{-\eta}\:=\:0\;,
\label{q5.1}\\
\ll \zeta_{i}^{\alpha},\xi_{j}^{\beta} \right] _\eta&=&
\ll \zeta_i^\alpha,\xi_j^{\beta\dagger}\right] _{\eta}\:=\:0\;,~~~(\alpha
\neq\beta)\;,
\label{q5.2}
\end{eqnarray}
where $\eta=\pm$ determines the relative statistics. $\eta=+$
(resp.~$\eta=-$) corresponds to the relative parabosonic
(resp.~ parafermionic) statistics.
The relative parastatistics is not determined by physical
reasoning, however there is a so called {\em normal relative
parastatistics} \cite{g-m} that generalizes the known case of $p=1$.
It is described as follows:
\begin{itemize}
\item[I)] If $p_a\neq p_b$, then $a_i^\dagger,~a_i$
and $b_j^\dagger,~b_j$ anticommute if both particles $a$ and $b$ are
parafermions. Otherwise, they commute.
\item[II)] If $p_a=p_b=p$, then in Eqs.~(\ref{q5.1}) and (\ref{q5.2})
$\eta=-$, if both $a$ and $b$ are parafermions. Otherwise, $\eta=+$.
\end{itemize}
The latter case says that two species of parafermionic particles
of the same order have parafermionic relative statistics. Whereas
a parabosonic particle has relative parabosonic statistics with respect
to both parabosonic and parafermionic particles of the same order.
This is a direct generalization of the $p=1$ case.
We conclude this section with a comment on the classical counterparts
of the quantum operators encountered above.
In the spirit of the work of Berezin \cite{berezin}, one defines
the classical analogs of $a_i^\dagger,~a_i$ and $\zeta_i^{\alpha\dagger},
{}~\zeta_i^\alpha$ as generators of algebras
defined by the rules given by (\ref{q1}) and (\ref{q2})
with the right hand side set to zero. Again the formula (\ref{q3})
establishes the relation between these algebras. The generators of
the former algebra, i.e., the one defined by setting the right hand side
of (\ref{q1}) to zero, with the sign ($-$) chosen in
Eqs.~(\ref{q1}), are called
{\em para-Grassmann variables of order $p$}. There is an alternative
definition of para-Grassmann variables advocated by Fillipov et al.
\cite{fillip} which is relevant to fractional supersymmetry. The latter
will not be employed in this article.
\section{The ($p=2$) Case and the Parabracket}
For $p=2$, the defining relations (\ref{q1}) simplify considerably
\cite{green,g-m}. One has:
\begin{eqnarray}
a_ka_l^\dagger a_m\pm a_ma_l^\dagger a_k &=&
2\delta_{kl}a_m\pm 2\delta_{lm}a_k\;,\nonumber\\
a_ka_l a_m^\dagger\pm a_m^\dagger a_l a_k &=&
2\delta_{lm}a_k\;,
\label{q10}\\
a_ka_la_m\pm a_ma_la_k&=&0\;,\nonumber
\end{eqnarray}
which can be easily checked using the Green representation (\ref{q3}).
In these equations, the signs ($+$) and ($-$) correspond to ($p=2$)
parafermionic and parabosonic operators, respectively. Since we would like to
treat both of these operators simultaneously, the introduction of a
grading index $\mu=0,1$ is convenient, i.e., we attach $\mu$ to
operators $a_k$ and their Green components $\zeta_k^{\alpha}$ as a
superindex, and interpret $a_k^\mu$ and $\zeta_k^{\alpha\mu}$ as
parabosonic if $\mu=0$ and parafermionic if $\mu=1$. Now, we can
use $\mu$ to express the ($\pm$) signs in Eqs.~(\ref{q10}). In terms
of the Green components:
\begin{equation}
a_i^\mu=\sum_{\alpha=0}^1 \zeta_i^{\alpha\mu}=
\zeta_i^{0\mu}+\zeta_i^{1\mu}\;,
\label{q11}
\end{equation}
one has:
\begin{eqnarray}
[\zeta_i^{\alpha\mu},\zeta_j^{\alpha\mu\dagger}]_{(-1)^{\mu+1}}&=&
\delta_{ij}\;,\nonumber\\
{[}\zeta_i^{\alpha\mu},\zeta_j^{\alpha\mu}{]_{(-1)^{\mu+1}}}&=& 0\;,
\label{q12}\\
{[}\zeta_i^{\alpha\mu},\zeta_j^{\beta\mu\dagger}{]_{(-1)^\mu}}&=&
{[}\zeta_i^{\alpha\mu},\zeta_j^{\beta\mu}{]_{(-1)^\mu}}\:=\: 0\,,~~~
(\alpha\neq\beta)\;.\nonumber
\end{eqnarray}
Note that throughtout the rest of this article the Green indices, $\alpha,
\beta,\cdots$, take values 0 and 1, for $p=2$.
It turns out that it is easier to work with self-adjoint (``real") operators
(variables). Thus we introduce yet another index $m=1,2$ and consider the
self-adjoint operators:
\begin{equation}
\theta^{\alpha\mu}_{i1}:=\sqrt{\frac{\hbar}{2}}(\zeta_i^{\alpha\mu}
+\zeta_i^{\alpha\mu\dagger})\;,~~
\theta^{\alpha\mu}_{i2}:=-i\sqrt{\frac{\hbar}{2}}(\zeta_i^{\alpha\mu}
-\zeta_i^{\alpha\mu\dagger}).
\label{q15}
\end{equation}
Now, if one defines the {\em parabracket} by:
\begin{equation}
[\hspace{-.3mm}[
\theta^{\alpha\mu}_{im},\theta_{jn}^{\beta\nu}
]\hspace{-.3mm}]:=
\theta^{\alpha\mu}_{im} \theta^{\beta\nu}_{jn}-(-1)^{
\mu\nu+\alpha+\beta}
\theta^{\beta\nu}_{jn}\theta^{\alpha\mu}_{im}\;,
\label{q16}
\end{equation}
then the relation:
\begin{equation}
[\hspace{-.3mm}[
\theta^{\alpha\mu}_{im},\theta_{jn}^{\beta\nu}
]\hspace{-.3mm}]
=\hbar\delta_{ij}\delta^{\alpha\beta}\ll
i(1-\mu)(1-\nu)\epsilon_{mn}+\mu\nu\delta_{mn}\right] \;,
\label{q16a}
\end{equation}
not only summarizes the defining relations (\ref{q2}) and hence (\ref{q1}),
but it also includes the statement of the normal relative parastatistics.
In Eq.~(\ref{q16a}), $\delta$ and $\epsilon$ are the Kronecker delta function
and the Levi Civita symbol, respectively.
One might view Eq.~(\ref{q16a}) as the statement of canonical quantization
for the ($p=2$) para-classical systems. In fact, the factor $\hbar$ has
been introduced so that (\ref{q16a}) yields the definition of the
classical counterparts of the quantum operators, i.e., ($p=2$)
parafermionic and parabosonic variables, in the limit $\hbar\to 0$.
The definition of the parabracket (\ref{q16}) may be extended to
polynomials in $\theta_{im}^{\alpha\mu}$. This is done by
defining it for the monomials, e.g.
\begin{equation}
M:=\theta_{i_1m_1}^{\alpha_1\mu_1}\cdots
\theta_{i_rm_r}^{\alpha_r\mu_r}\;,~~
N:=\theta_{j_1n_1}^{\beta_1\nu_1}\cdots
\theta_{j_sn_s}^{\beta_s\nu_s}\;,
\label{mono}
\end{equation}
by
\begin{eqnarray}
[\hspace{-.3mm}[
M,N
]\hspace{-.3mm}]
:=MN-(-1)^{\eta(M,N)}NM\;,
\label{q20}\\
\eta(M,N):=(\sum_{k=1}^r \mu_k)(\sum_{l=1}^s\nu_l)+
r\sum_{l=1}^s\beta_l+s\sum_{k=1}^r\alpha_k\;,
\label{q20b}
\end{eqnarray}
and requiring bilinearity. In the classical limit, for any two polynomials
$P$ and $Q$ in $\theta_{im}^{\alpha\mu}$, one has
\begin{equation}
[\hspace{-.3mm}[
P, Q
]\hspace{-.3mm}]
=0\;.
\label{q20a}
\end{equation}
A more substantial result is a generalization of the Jacobi identity.
The following lemma can be easily proved by the application of
Eqs.~(\ref{q20}) and (\ref{q20b}).
\begin{itemize}
\item[ ]{\bf Lemma 1:} {\em
Let $M,~N$, and $O$ be monomials in $\theta_{im}^{\alpha\mu}$
and the parabracket $[\hspace{-.3mm}[~,~]\hspace{-.3mm}]$, is defined by Eq.~(\ref{q20}),
then the relation:
\begin{eqnarray}
(-1)^{\eta(M,O)}[\hspace{-.3mm}[ M,[\hspace{-.3mm}[ N,O]\hspace{-.3mm}]\,]\hspace{-.3mm}]&+&
(-1)^{\eta(O,N)}[\hspace{-.3mm}[ O,[\hspace{-.3mm}[ M,N]\hspace{-.3mm}]\,]\hspace{-.3mm}]\: +~~~
\label{q31}\\
&&
(-1)^{\eta(N,M)}[\hspace{-.3mm}[ N,[\hspace{-.3mm}[ O,M]\hspace{-.3mm}]\,]\hspace{-.3mm}]\:=\: 0\;,\nonumber
\end{eqnarray}
holds as an identity.}\footnote{Eq.~(\ref{q31}) is a generalization
of the the super-Jacobi identity \cite{bd} encountered in the
study of supersymmetry. Thus it might be called the
{\em para-Jacobi identity}.}
\end{itemize}
Before proceeding further, we would like to make a further remark about
Eq.~(\ref{q16a}). This equation also provides a description of
the ($p=1$) case. This is done by making the Green indices vanish,
i.e., $\alpha,\beta,\cdots=0$.
This reveals the well-known fact that for the bosonic case ($\mu=\nu=0$),
the variables $\theta^{00}_{i2}$ correspond to the momenta
conjugate to the coordinates $\theta_{i1}^{00}$. This suggests
a similar pattern for the ($p=2$) case. That is, the
parabosonic coordinate variables in the Lagrangian formulation are
$\theta_{i1}^{\alpha 0}$. Whereas there is no such restriction on the
parafermionic variables. To demonstrate this in a unified notation, one
may introduce a collective index $I=(i;m)$, i.e., consider
$\theta_I^{\alpha\mu}$ , and require that for $\mu=0$,
$I=(i=1,\cdots,n_{\pi b};m=1)$ and for $\mu=1$, $I=(i=1,\cdots,n_{\pi f};
m=1,2)$, where $n_{\pi b}$ and $2n_{\pi f}$ are the number of
parabosonic and parafermionic degrees of freedom, respectively. In view of
this notation, one rewrites (\ref{q16a}) in the classical limit, as follows:
\begin{equation}
[\hspace{-.3mm}[
\theta^{\alpha\mu}_{I},\theta_{J}^{\beta\nu}
]\hspace{-.3mm}]
=0\;.
\label{q16b}
\end{equation}
One also must emphasize that the physical dynamical coordinate varaibles
are:
\begin{equation}
\psi^\mu_I:=\sum_{\alpha=0}^1 \theta_I^{\alpha\mu}\;,
\label{q17}
\end{equation}
and not the Green components $\theta_I^{\alpha\mu}$ themselves. In other
words, it is the algebra (ring) of polynomials ${\cal P}$ in $\psi^\mu_I$
that serves as the space of physical quantities. In view of Eq.~(\ref{q17}),
${\cal P}$ is a subalgebra (subring) of the algebra (ring) of
polynomials ${\cal T}$ in $\theta_I^{\alpha\mu}$. ${\cal T}$
is used as a larger space in which the calculations are performed.
To extract the physical results, one is bound to
project to the subspace ${\cal P}$. ${\cal P}$ and ${\cal T}$ have some
important subspaces. These are the even subalgebras ${\cal P}_2$ and
${\cal T}_2$, and the subalgebras generated by only the parabosonic
(parafermionic) variables $\psi^1_I$ (resp.~$\psi^0_I$) of ${\cal P}$,
and $\theta_I^{\alpha 1}$ (resp.~$\theta_I^{\alpha 0}$) of ${\cal T}$.
These are denoted by ${\cal P}^\mu$ and ${\cal T}^\mu$, respectively.
In view of Eq.~(\ref{q20a}), one finds that for example
the monomials in ${\cal T}_2$ either commute or anticommute.
In fact, if there is an even number of parafermionic factors
in an even monomial it commutes with all the even monomials and
two even monomials with odd numbers of parafermionic factors anticommute.
Furthermore, the even subalgebras of both ${\cal T}^\mu$
and hence ${\cal P}^\mu$, ($\mu=0,1$), are commutative. This is
important, because one would ordinarily like to choose ``physical"
quantities such as a Lagrangian to be a commutative object.
This cannot be fully achieved with ($p=2$) variables in general.
However, one might suffice to require that the Lagrangian be
chosen as a linear sum of even monomials each consisiting of an
even number of parafermionic factors. We shall offer a justification
for the latter requirement in Sec.~4.
In order to carry out the program of Lagrangian mechanics, one
also needs a differential calculus for the variables $\psi$'s or
alternatively for $\theta$'s. The latter also is addressed in the
earlier work in parastatistics \cite{o-k}. The results can be best
demonstrated using an extension of the definition of parabracket which also
applies
to ``partial derivatives":
\begin{eqnarray}
[\hspace{-.3mm}[
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta_I^{\alpha\mu}},
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta_J^{\beta\nu}}
]\hspace{-.3mm}]
&=&
[\hspace{-.3mm}[
\frac{\stackrel{\leftarrow}{\partial}}{\partial\theta_I^{\alpha\mu}},
\frac{\stackrel{\leftarrow}{\partial}}{\partial\theta_J^{\beta\nu}}
]\hspace{-.3mm}]
\:=\:0,
\label{q21} \\
[\hspace{-.3mm}[
\theta_I^{\alpha\mu},
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta_J^{\beta\nu}}
]\hspace{-.3mm}]
&=&
[\hspace{-.3mm}[
\frac{\stackrel{\leftarrow}{\partial}}{\partial\theta_I^{\alpha\mu}},
\theta_J^{\beta\nu}
]\hspace{-.3mm}]
\:=\:
\delta_{\mu\nu}\delta_{\alpha\beta}\delta_{IJ}\,,
\label{q21a}
\end{eqnarray}
where one defines the left hand sides of the latter equations
by replacing $\theta$'s in Eq.~(\ref{q16}) by either of
$\stackrel{\leftarrow}{\partial}\!\!/\partial\theta$ or $\stackrel{\rightarrow}{\partial}\!\!/\partial\theta$, with the same
indices.
Eqs.~(\ref{q21a}) may be used to obtain a generalized Leibniz rule.
One has:
\begin{itemize}
\item[ ] {\bf Lemma 2}: {\em
Let $M$ and $N$ be monomials in ${\cal T}$ as given by
Eq.~(\ref{mono}), then
\begin{equation}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta}(MN)=
(\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta}M)N-(-1)^{\eta(M,N)}
(\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta}N)M\;,
\label{q30}
\end{equation}
where $\eta(M,N)$ is defined by Eq.~(\ref{q20b}), and
the indices of $\theta$'s are suppressed for simplicity.}
\end{itemize}
A proof of Lemma 2 involves a lengthy two step induction on the
orders $r$ and $s$ of the monomials. Here, one makes extensive use
of Eqs.~(\ref{q16}), (\ref{q16b}) and (\ref{q21a}). Eq.~(\ref{q30}) is
of great practical use in performing computations with $\theta$'s.
A similar result may be proven for $\stackrel{\leftarrow}{\partial}\!\!/\partial\theta$.
We conclude this section with a discussion of the {\em reality}
condition. As is the case in the analysis of supernumbers \cite{bd},
we define a real element of the (complex) algebra ${\cal T}$, and
similarly ${\cal P}$, by introducing a $*$-operation. This is already
implicit in the quantum level in the definition of the Hermitian
conjugation. Following the ($p=1$) case \cite{bd}, we require
\begin{equation}
\left( \lambda~
\theta^{\alpha_1\mu_1}_{I_1}\cdots\theta^{\alpha_r\mu_r}_{I_r}
\right)^* :=
\lambda^*
\theta^{\alpha_r\mu_r}_{I_r}\cdots\theta^{\alpha_1\mu_1}_{I_1}
\;,\nonumber
\end{equation}
and (additive) linearity of $*$-operation. Here $\lambda$ is a complex number
and $\lambda^*$ stands for its complex conjugate. A real element of ${\cal T}$
(resp.\ ${\cal P}$) is one whose $*$-conjugate equals itself.
\section{Peierls Bracket Quantization}
A generalization of Peierls bracket quantization to systems involving
bosonic (commuting) and fermionic (anticommuting) dynamical variables is
carried out in Ref.~\cite{bd}. Here a brief review is presented.
Consider a (non-relativistic) classical system whose dynamics is described by
the action functional:
\begin{equation}
S[\Phi]:=\int_{\Phi} L(\Phi^i(t),\dot{\Phi}^i(t),t) dt \;,
\label{q41}
\end{equation}
where $\Phi^i$ are the coordinate variables, $\dot{\Phi}^i$ are their
corresponding velocities, $t\in [0,T]$ is the time variable,
and $\Phi=(\Phi(t))$ is a path in the configuration space. Following
Ref.~\cite{bd}, let us denote the right and left functional derivatives
by
\begin{equation}
S_{,i'}\equiv
S[\Phi]\frac{\stackrel{\leftarrow}{\delta}}{\delta\Phi^i(t')}\;,
~~~_{i',}\!S\equiv
\frac{\stackrel{\rightarrow}{\delta}}{\delta\Phi^i(t')} S[\Phi]\;,
\label{q42}
\end{equation}
respectively. In this (condensed) notation, the indices represent both
the discrete and continuous (time) labels and repeated indices imply
summation over the discrete and integration over the continuous labels.
In particular, note that the prime on the index $i$ in (\ref{q42})
is associated with the continuous index, $t'$.
The dynamical equations are given by
\begin{equation}
S_{,i}=0\;.
\label{q43}
\end{equation}
The second functional derivatives of the action functional yield the Jacobi
operator: $(_{i,}\!S_{,j'})$. The Green's functions of the latter
are defined according to their boundary conditions and the familiar relation:
\begin{equation}
_{i,}S_{,j'}\,G^{j'k''}=-\,_{i}\delta^{k''}\;,
\label{q44}
\end{equation}
where the repeated index $j'$ is summed and integrated over, and
\[
_{i}\delta^{k''} \equiv \delta_{i}^{k}\delta(t-t'')\;.
\]
Denoting the advanced and retarded Green's functions by $G^+$ and $G^-$,
one defines the Peierls bracket of the fields $A=A[\Phi^i]$ and
$B=B[\Phi^i]$ according to:
\begin{equation}
(A,B) := A_{,i}\, \tilde{G}^{ij'}\, _{j',}\! B\;,
\label{q45}
\end{equation}
where the Green's function $\tilde{G}$ is defined by
\begin{equation}
\tilde{G}:=G^+-G^- \;,
\label{q46}
\end{equation}
It is called the {\em supercommutator function} by De~Witt \cite{bd}. One also
has the useful relation:
\begin{equation}
(\Phi^i,\Phi^{j'}):=\tilde{G}^{ij'}\;.
\label{q45a}
\end{equation}
The Peierls quantization scheme invloves the promotion of the classical
fields to linear operators acting on a Hilbert space and satisfying the
following (not necessarily equal time) supercommutation relations:
\begin{equation}
[\hat{A},\hat{B}]_{\rm super}=i\hbar\widehat{(A,B)}\;.
\label{q47}
\end{equation}
Here the hats are placed to emphasize that the corresponding quantities are
operators. They will be dropped where possible. If $A$ and $B$ have definite
parity, then the {\em superbracket} $[~,~]_{\rm super}$ becomes the ordinary
commutator if either of $A$ or $B$ is bosonic. Otherwise it becomes the
anticommutator. In practice, one usually uses Eq.~(\ref{q47}) written for
the coordinate variables, i.e.,
\begin{equation}
[\hat{\Phi}^i,\hat{\Phi}^{j'}]_{\rm super}=
i\hbar \widehat{\tilde{G}^{ij'}}\;,
\end{equation}
and properties of the Peierls bracket \cite{bd} (more conveniently those of
the superbracket) to compute the superbracket of other fields.
Employing the parity indices $\mu,~\nu,~\cdots$ of Sec.~2, i.e., considering
$\Phi^{i\mu}$, with $\mu=0$ corresponding to bosonic coordinates and
$\mu=1$ to the fermionic coordinates, one has
\begin{equation}
[\hat{\Phi}^{i,\mu},\hat{\Phi}^{j'\nu}]_{\rm super}:=
\hat{\Phi}^{i\mu}\hat{\Phi}^{j'\nu}-(-1)^{\mu\nu}
\hat{\Phi}^{j'\nu}\hat{\Phi}^{i\mu} \;.
\label{q48}
\end{equation}
Ref.~\cite{bd} uses the same indices to label the coordinates and their
parity.
For this procedure to make sense, the Peierls bracket must possess a
series of properties. These are essentially the properties of
the supercommutator, namely the {\em supersymmetry} property:
\begin{equation}
(A^\mu,B^\nu)=-(-1)^{\mu\nu}(B^\nu A^\mu) \;,
\label{q49}
\end{equation}
and {\em super-Jacobi identity}:
\begin{equation}
(-1)^{\mu\pi}(A^\mu,(B^\nu,C^\pi))
+(-1)^{\pi\nu}(C^\pi,(A^\mu,B^\nu))
+(-1)^{\nu\mu}(B^\nu,(C^\pi,A^\mu)) =0 \;,
\label{q50}
\end{equation}
where $A^\mu,~B^\nu$ and $C^\pi$ are functions of $\Phi^i$ and have definite
parities $\mu,~\nu$ and $\pi$, respectively. Note that relations (\ref{q49})
and (\ref{q50}) are quite nontrivial. A proof of Eqs.~(\ref{q49}) and
(\ref{q50}) uses the symmetries of the Jacobi operator $_{i,}S_{,j'}$,
the supercommutator function $\tilde{G}^{ij'}$, and their functional
derivatives under the exchange of their indices \cite{bd}.
In view of the developments presented in the last section, we proceed
to generalize the Peierls scheme to systems involving ($p=2$)
parabosonic and parafermionic variables.\footnote{Inclusion of ordinary
fermionic and bosonic variables to such systems can also be carried out
within the framework presented in the present article.} In order to pursue
in this direction, we consider a Lagrangian $L$ built up of parabosonic
and parafermionic variables $\psi_I^\mu$ ($\mu=0,1$) and the corresponding
velocities, $\dot{\psi}_I^\mu$, i.e.,
\begin{equation}
L=L(\psi^\mu_I,\dot{\psi}^\mu_I,t)\;.
\label{q51}
\end{equation}
Note that the velocities are considered as independent
variables with the same parastatistical properties. In general, we shall
consider real Lagrangians which are even polynomials in both parafermionic
and parabosonic variables. The latter condition will prove essential
in having a consistent quantization scheme. For practical purposes, we
then switch to the Green's components $\theta_I^{\alpha\mu}$ and
$\dot{\theta}_I^{\alpha\mu}$. Using the calculus developed for
Green's components, one can define the notion of functional differentiation,
e.g., according to
\begin{eqnarray}
F_{,i'} &\equiv& F[\theta(t)]\frac{\stackrel{\leftarrow}{\delta}}{
\delta\theta^i(t')}\nonumber\\
&:=& \left(
F[\theta^1(t),\cdots,\theta^i
(t+\epsilon\delta(t-t')),\cdots, \theta^d(t)]
- F[\theta(t)] \right)\left.
\frac{\stackrel{\leftarrow}{\partial}}{\partial\epsilon}\right|_{\epsilon=0}\;,\nonumber
\end{eqnarray}
where the index $i$ is a collective index representing $(I,~\alpha,
{}~\mu$) and $\epsilon$ is a variable with the same parastatistical
properties as $\theta^i$. The left functional derivative is defined
similarly.
Identifying the coordinate variables $\Phi^i$ of the beginning of
this section by $\theta^i$, with $i\equiv (I,\alpha,\mu)$,
the action functional, the dynamical equations, the Jacobi operator,
and its Green's functions are given according to Eqs.~(\ref{q41}),
(\ref{q43}) and (\ref{q44}), respectively. The {\em para-generalization}
of the Peierls bracket is obtained by Eqs.~(\ref{q45}) and (\ref{q46}). The
following analog of Eq.~(\ref{q47}) then yields the Peierls quantization
condition:
\begin{equation}
[\hspace{-.3mm}[ \hat{A}, \hat{B}]\hspace{-.3mm}]= i\hbar \widehat{(A,B)}
\label{q52}
\end{equation}
where $[\hspace{-.3mm}[~,~]\hspace{-.3mm}]$ is the parabracket defined by Eq.~(\ref{q20}),
and $A$ and $B$ are polynomials in $\theta_I^{\alpha\mu}$. In particular,
one has:
\begin{equation}
[\hspace{-.3mm}[ \hat{\theta}^i(t),\hat{\theta}^j(t')]\hspace{-.3mm}] =
i\hbar \widehat{\tilde{G}^{ij'}}\;.
\label{q53}
\end{equation}
The above procedure would be consistent provided that the para-generalized
Peierls bracket satisfies the symmetry properties of the parabracket,
namely the {\em parasupersymmetry} properties:
\begin{equation}
(M,N)=-(-1)^{\eta(M,N)}(N,M)\;,
\label{q54}
\end{equation}
and the {\em para-generalized Jacobi identity}:
\begin{equation}
(-1)^{\eta(M,O)}(M,(N,O))+
(-1)^{\eta(O,N)}(O,(M,N))+
(-1)^{\eta(N,M)}(N,(O,M))=0\;,
\label{q55}
\end{equation}
Here the function $\eta$ is the one defined by Eq.~(\ref{q20b}) and
$M,~N$, and $O$ are monomials in $\theta$'s.
A proof of Eq.~(\ref{q54}) follows from the following symmetry property
of {\em paracommutator function} $\tilde{G}^{ij'}$:
\begin{itemize}
\item[]{\bf Proposition~1}: {\em
Let $\tilde{G}^{ij'}$ be defined by Eq.~(\ref{q46}),
$i=(I,\alpha,\mu)$ and $j=(J,\beta,\nu)$, then:
\begin{equation}
\tilde{G}^{ij'}=-(-1)^{\mu\nu+\alpha+\beta}
\tilde{G}^{j'i}\;.
\label{q56}
\end{equation}}
\end{itemize}
To arrive at a proof of Prop.~1, we first state a couple of related results
which are labeled as Lemmas~3 and 4:
\begin{itemize}
\item[]{\bf Lemma~3}: {\em
Let $M=\theta_{J_1}^{\gamma_{1}\rho_{1}}\cdots
\theta_{J_D}^{\gamma_D\rho_D}$ be a monomial of order $D$,
then:
\begin{equation}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta^{\alpha\mu}_I}M
=(-1)^{\mu+\sum_{a=1}^{D}
\eta(\theta^{\alpha\mu}_I,\theta^{\gamma_a\rho_a}_{J_a})}
(M\frac{\stackrel{\leftarrow}{\partial}}{\partial\theta^{\alpha\mu}_I})\;.
\label{q57}
\end{equation}}
\end{itemize}
A proof of this result is obtained by a direct computation of both
sides of Eq.~(\ref{q57}) using the result of Lemma~2, i.e., Eq.~(\ref{q30}).
Next, we have:
\begin{itemize}
\item[]{\bf Lemma~4}: {\em
Let M be as in Lemma~3, and let $i$ and $j$ label
$(I,\alpha,\mu)$ and $(J,\beta,\mu)$ respectively,
then
\begin{equation}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta^i}M\frac{\stackrel{\leftarrow}{\partial}}{\partial
\theta^j}=
(-1)^{(1+\sum_{a=1}^D\rho_a)(\mu+\nu)+\mu\nu+
(D+1)(\alpha+\beta)}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta^j}M\frac{\stackrel{\leftarrow}{\partial}}{\partial
\theta^i}\;.
\label{q58}
\end{equation}
In particular, if $M$ is an even monomial in both parabosonic
and para\ -fermionic variables, then
\begin{equation}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta^i}M\frac{\stackrel{\leftarrow}{\partial}}{\partial
\theta^j}=
(-1)^{\mu\nu+\mu+\nu+\alpha+\beta}
\frac{\stackrel{\rightarrow}{\partial}}{\partial\theta^j}M\frac{\stackrel{\leftarrow}{\partial}}{\partial
\theta^i}\;.
\label{q59}
\end{equation}}
\end{itemize}
Lemma~4 is a straightforward consequence of Lemma~3. The statement of
Lemma~4 generalizes to the case of functional derivatives as well. Namely:
\begin{itemize}
\item[] {\bf Corollary}: {\em
If the action functional $S$ consists of terms which are
even monomials in both parabosonic and parafermionic variables,
then one has:
\begin{equation}
_{i,}\!S_{,j'}=(-1)^{\mu\nu+\mu+\nu+\alpha+
\beta}~ _{j',}\!S_{,i}\;.
\label{q60}
\end{equation}}
\end{itemize}
Eqs.~(\ref{q44}) and (\ref{q60}) together with the observation that
both $_{i,}\!S_{,j'}$ and $G^\pm$ are even polynomials, lead to the
desired reciprocity relation:
\begin{equation}
G^{\pm ij'}=(-1)^{\mu\nu+\alpha+\beta}G^{\mp j'i}
\label{q61}
\end{equation}
This equation and the definition (\ref{q46}) yield a proof of Prop.~1.
Proof of the para-Jacobi identity (\ref{q55}) follows essentially the
same procedure as in the ($p=1$) case \cite{bd}, but the computations are
more involved.
In the next section, we consider a simple example of application
of Peierls quantization program for a ($p=2$)--parafermionic system.
\section{One-dimensional Parafermi System}
Let us denote by $\psi$ a classical ($p=2$)--parafermionic (para-Grassmann)
variable with the Green components $\tau^\alpha:=\theta^{\alpha,\mu=1}_{I=1}$.
Then the defining relations (\ref{q10}), in the classical limit, imply
\begin{equation}
\begin{array}{cc}
\psi^3=\dot{\psi}^3=0\;,&\psi^2\dot{\psi}^2=\dot{\psi}^2\psi^2\;,\\
\psi\dot{\psi}^2=-\dot{\psi}^2\psi\;,&
\psi^2\dot{\psi}=-\dot{\psi}\psi^2\,
\end{array}
\label{q62}
\end{equation}
where $\psi$ and $\dot{\psi}$ are treated as independent ($p=2$)--parafermi
variables. Furthermore, one has the realtions:
\begin{equation}
(\psi\dot{\psi})^2=(\dot{\psi}\psi)^2=0\;,
\label{q63}
\end{equation}
which are most easily verified using the Green representation.
In view of Eqs.~(\ref{q62}) and (\ref{q63}), the most general real even
polynomial in dynamical variables -- upto an unimportant multiplicative
constant and additive total time derivatives -- has the form:
\begin{equation}
L=\frac{A}{2}\psi^2+\frac{B}{2}\dot{\psi}^2+
\frac{C}{4}\psi^2\dot{\psi}^2+ \frac{i}{4}(\psi\dot{\psi}-
\dot{\psi}\psi)\;.
\label{q64}
\end{equation}
Here $A,~B$ and $C$ are real numerical parameters. Eq.~(\ref{q64}) serves
as the most general possible form for the Lagrangian. In the following we
shall make a further demand, namely that the Peierls bracket quantization
and the canonical quantization of this system be consistent.
To carry out Peierls' program we first rewrite the Lagrangian (\ref{q64})
in terms of the Green components $\tau^\alpha$ and $\dot{\tau}^\alpha$
and compute the Jacobi operator. Here we suffice to state the results:
\begin{eqnarray}
L&=&\sigma_{\alpha\beta}(
\frac{A}{2}\tau^\alpha\tau^\beta+\frac{B}{2}\dot{\tau}^\alpha
\dot{\tau}^\beta + \frac{C}{4}\sigma_{\gamma\delta}
\tau^\alpha\tau^\beta\dot{\tau}^\gamma\dot{\tau}^\delta)
+\frac{i}{4}\delta_{\alpha\beta}\tau^\alpha\dot{\tau}^\beta\;,
\label{q65}\\
_{\beta',}\!S_{,\alpha}&=&\left\{\left[\sigma_{\alpha\beta}
(-B+\frac{C}{2}\sigma_{\gamma\delta}\tau^\gamma\tau^\delta)\right]
\frac{\partial^2}{\partial t^2}+ \right.
\label{q66}\\
&& \left[ -i\delta_{\alpha\beta}
-C([(-1)^{\alpha+\beta}+1]\sigma_{\alpha
\gamma}\sigma_{\delta\beta}-
\sigma_{\alpha\beta}\sigma_{\gamma\delta})
\tau^\gamma\dot{\tau}^\delta\right]
\frac{\partial}{\partial t}+\nonumber\\
&& \left. \left[ A\sigma_{\alpha\beta}-C( [
\sigma_{\alpha\delta}\sigma_{\gamma\beta}
-\frac{1}{2}\sigma_{\alpha\beta}\sigma_{\gamma\delta}]
\dot{\tau}^\gamma\dot{\tau}^\delta+
\sigma_{\alpha\delta}\sigma_{\beta\gamma}
\tau^\gamma\ddot{\tau}^\delta)\right]
\right\}\delta(t-t')\;,\nonumber
\end{eqnarray}
where $\sigma$ denotes the Pauli matrix $\sigma_1$, i.e.,
\begin{equation}
\sigma_{\alpha\beta}=\left\{\begin{array}{cc}
1 & {\rm if}~~~ \alpha=\beta\\
0 & {\rm if}~~~ \alpha\neq \beta.
\end{array}\right.
\label{sigma}
\end{equation}
The Green's functions can be computed as power series in $(t-t')$,
similarly to the ($p=1$) case \cite{bd,p5}. A simple analysis of the
Green's functions, shows that if $B\neq 0$ or $C\neq 0$, then
$[\hspace{-.3mm}[\tau^\alpha(t),\tau^\beta(t)]\hspace{-.3mm}]=0$, which is inconsistent with
the result of canonical quantization (\ref{q16a}), namely:
\begin{equation}
[\hspace{-.3mm}[\tau^\alpha(t),\tau^\beta(t)]\hspace{-.3mm}]= \hbar\delta^{\alpha\beta}\;.
\label{q67}
\end{equation}
Setting $B=C=0$, and carrying out the computation of the Green's functions,
one finds:
\begin{equation}
G^{\pm\alpha\beta'}=\left[
\mp i\delta^{\alpha\beta}+ O(t-t')\right]\Theta[\pm(t-t')]
\;,
\end{equation}
where $\Theta$ is the step function: $\Theta(t)=1$ if $t>0$,
$\Theta(t)=0$ if $t<0$, $\Theta(0)=1/2$.
The latter relation directly leads to Eq.~(\ref{q67}) and confirms the
consistency of the canonical and Peierls quantization programs.
Enforcing $B=C=0$ in the expression for the Lagriangian (\ref{q64}), one
has:
\begin{equation}
L=\frac{i}{4}(\psi\dot{\psi}-\dot{\psi}\psi)+\frac{A}{2}\psi^2\;.
\label{q68}
\end{equation}
The first couple of terms in the right hand side of (\ref{q68}) has the
same form as the kinetic term for ordinary fermionic systems. We shall
refer to these also as the kinetic term of the parafermionic system. The
last term serves as a potential term which does not have a counterpart
in fermionic systems.
A similar analysis shows that for ($p=2$)--parabosonic systems, choosing
the kinetic term to be of the same form as the bosonic kinetic term,
one ensures the consistency of the canonical and Peierls quantization
schemes.
\section{Conclusion}
The Lagrangian formulation of classical mechanics is shown to be applicable
to systems involving ($p=2$) parafermionic and parabosonic variables.
The introduction of the parabracket for the Green's components of the ($p=2$)
dynamical variables facilitates computations considerably. It also allows
for a generalization of the Peierls quantization program to such systems.
The internal consistency of the Peierls program requires the Lagrangian to
be an even polynomial in both parafermi and parabose variables.
The consistency of the results of the canonical and Peierls quantization
programs leads to the specification of the form of the parafermionic and
parabosonic kinetic terms in the Lagrangian.
The material developed in this article has direct application in the
study of systems involving both the parafermi and parabose variables of order
($p=2$). Some examples of such systems have been encountered in the
context of parasupersymmetric quantum mechanics \cite{b-d}. There are still
quite a few unsettled issues regarding the true meaning of parafermi-parabose
(super)symmetry. Some of these issues are addressed in a companion paper
\cite{p11} using the formalism developed above.
\newpage
|
\section{Introduction}
The first step of the transcription of deoxyribonucleic acid (DNA) is a local
opening of the double helix which extends over about 20 base pairs. Such
local unwindings of the helix can be obtained by heating DNA to about
$70^{\circ}~C$. But in the life of an organism they must occur at
physiological temperature. This is achieved by the action of an
enzyme\cite{CD}. However one may wonder how this can be possible since,
whatever its origin, the local opening requires the breaking of the same
number of hydrogen bonds, hence the same amount of energy, and the enzyme
does not bring in energy. However, under normal physiological conditions
there are thermal fluctuations along the DNA chain. They can be weakly
localized by nonlinear effects to generate what biologists call the
``breathing of DNA''. But their intensity is not high enough to open the
double helix over many base pairs. A possible pathway to the opening would
be to collect the thermal energy that is present along the molecule. This
could be the role of the enzyme. From a physicist point of view, the effect
of an enzyme can be considered as a perturbation to the DNA lattice.
Recently Forinash {\it et al} considered the interaction between a mass
impurity on a DNA chain, and thermal nonlinear waves described as
breathers traveling along the chain.\cite{FPM} They found that the
impurity is selective toward the breather it can trap. Although this is
a first indication that a defect can contribute to localize energy in a
nonlinear chain, it does not appear to be a good model for the action of
an enzyme because, with such a localized defect, only some predefined
frequencies of the thermal fluctuations would contribute to bring in the
energy. Therefore one may ask whether there exist any other mechanism
more efficient to trap energy.
One learns from biological studies that some proteins, make contact with DNA
at multiple sites\cite{A,EYSB}. Moreover the transcription enzyme actually
bends DNA toward itself. It has the effect not only to modify the mass at
some sites but also to modify the coupling constants along the strands. The
bases which are inside the bend are brought closer to each other while the
ones which are outside are moved farther apart. Although the variation of
the distances between neighboring bases may be rather small, it can have a
large effect because the interaction between bases is due to the overlap of
$\pi$ electrons over the whole surface of the planar bases. We examine in
this paper whether the interaction of the enzyme with more than one site
might be more efficient for trapping breathers than isolated impurities by
studying the effect of an extended modification of the coupling along the
DNA chain.
The effect of bending and twisting to modify the elasticity of DNA has been
considered previously by Barkley and Zimm\cite{BZ},
and by Marko and Siggia\cite{MS} but they did not study the
consequences of base pair opening. Salerno\cite{S} considered the dynamical
properties of a DNA promoter which has some similarities with our problem
because we treat here the enzyme as an inhomogeneity due to an external
effect while he considers inhomogeneities from the DNA composition itself.
However he was interested by kinks while we study breathing modes. In a more
abstract level we are investigating here a nonlinear model, with an
``extended defect'', and we try to understand the interplay between
nonlinearity and disorder. In the harmonic case, a one dimensional chain with
isolated defects has been considered before by Montroll and Potts\cite{MP}.
However, besides the introduction of nonlinearity, one should also notice
that for the type of extended defect that we consider, there is no
evanescent local mode which would couple to a breather as in the case
considered by Forinash {\it et al}, so that the mechanism for energy
localization must be different.
\section{DNA lattice model}
If one neglects the small longitudinal motion and concentrates on the
stretching of the base pairs, DNA can be described by a simple
one-dimensional model\cite{PB} which consists of an array of harmonically
coupled particles subjected to a Morse potential. Such a model is
sufficient to provide a good qualitative description of the thermal
denaturation of the molecule\cite{Dauxois}. If one treats a chain with
inhomogeneous coupling, the equations of motion read
\begin{eqnarray}
\label{discrete}
m {{\partial^2 Y_n} \over {\partial T^2}} -
K_{n+1} ( Y_{n+1} - Y_n ) + K_n ( Y_n - Y_{n-1} )
\nonumber \\
- 2 D \alpha e^{- \alpha Y_n}
( e^{- \alpha Y_n} - 1) = 0 \; ,
\end{eqnarray}
in which $\alpha$ and $D$ are parameters for the Morse potential\cite{M},
which have dimensions of inverse length and energy respectively, and $n$ is
the site index.
Fig.~\ref{line1} shows the geometry and the coordinate used. It is convenient
for the analytical calculations to transform these equations into a
dimensionless form by defining the following dimensionless variables:
\begin{eqnarray}
y_n &=& \alpha Y_n, \\
t_{~} &=& \sqrt {{D \alpha^2 } \over m} T , \\
k_n &=& {K_n \over {D \alpha^2 }}.
\end{eqnarray}
The equations become
\begin{eqnarray}
\label{ND}
{{\partial^2 y_n} \over {\partial t^2}} -
k_{n+1} ( y_{n+1} - y_n ) +
k_n ( y_n - y_{n-1} )
\nonumber \\
- 2 e^{- y_n}
( e^{- y_n} - 1) = 0 \; .
\end{eqnarray}
One notices that the last set of equations contain only one parameter, the
coupling constant. In order to represent the perturbation due to the enzyme,
one could imagine to modify locally any of the parameters of
Eq.~(\ref{discrete}), but it is likely that the presence of an enzyme will
affect the coupling constant through the bending of the molecule. Moreover
previous studies of the role of disorder on the dynamics of the DNA
model\cite{Tashkent} have shown that the formation of open regions in the
model are much more sensitive to modulations of the coupling constant than
to changes in other parameters. Therefore we only consider here an extended
perturbation of the coupling constant. An additional possibility to model
the enzyme specificity is however examined in the discussion.
Since we do not know how to solve in the discrete case, we transform the set
of equations, Eqs.~(\ref{ND}), into the corresponding continuous PDE. In the
continuum limit, with a Taylor expansion in the potential term which assumes
small amplitude oscillation, Eq.~(\ref{ND}) becomes,
\begin{equation}
\label{continue}
{{\partial^2 y} \over {\partial t^2}}
- {\partial \over {\partial x} } (k_1
{{\partial y} \over {\partial x}} ) d^2
+ 2 ( y -
{3 \over 2} y^2 +
{7 \over 6} y^3 ) = 0 \; ,
\end{equation}
in which $d$ is the lattice spacing and $k_1$ is a space dependent coupling
constant. We set $d$ equal to unity in the following calculations.
\section{The Perturbed Nonlinear Schr\"odinger Equation}
Equation (\ref{continue}) can be transformed into a perturbed Nonlinear
Schr\"odinger equation by a multiple-scale expansion\cite{H,R}. Assuming
that the amplitude of the thermal oscillation is small, $y \approx
\epsilon \phi $, we perform the expansion
\begin{eqnarray}
\label{eq:approx}
\phi & = & F_0 + \epsilon F_1 + \epsilon^2 F_2 + O (\epsilon^3) \; , \\
{\partial \over {\partial t}} & = &
{\partial \over {\partial t_0}} + {\partial \over {\partial t_1}} \epsilon +
{\partial \over {\partial t_2}} \epsilon^2 + O ( \epsilon ^3 ) \; , \\
{\partial \over {\partial x}} & = &
{\partial \over {\partial x_0}} + {\partial \over {\partial x_1}} \epsilon +
{\partial \over {\partial x_2}} \epsilon^2 + O ( \epsilon ^3 ) \; .
\end{eqnarray}
Moreover we assume a modulation of the coupling constant of the order of
$\epsilon$, {\it i.e.}
\begin{equation}
\label{assume}
{{\partial k_1} \over {\partial x}} \approx
{{\partial k_1} \over {\partial x_1}} \; \epsilon \; .
\end{equation}
Equating like powers of $\epsilon$ yields a sequence of equations, in
ascending powers of $\epsilon$:
\begin{eqnarray}
\label{eq:order}
& & {\partial^2 F_0 \over \partial t_0^2} -
k_0 {\partial^2 F_0 \over \partial x_0^2} + 2 F_0 = 0 \; ,\\
& & ({\partial^2 F_1 \over \partial t_0^2} +
2 {{\partial^2 F_0} \over {\partial t_0 \partial t_1}}) -
{{\partial k_1} \over \partial x_1}{\partial F_0 \over \partial x_0} -
k_0 ({\partial^2 F_1 \over \partial x_0^2} +
2 {{\partial^2 F_0} \over {\partial x_0 \partial x_1}}) \nonumber \\
& &+ 2 (F_1 - {3 \over 2} F_0^2 ) = 0 \; ,\\
& &{\text{ and higher order equations}} \; , \nonumber
\end{eqnarray}
in which $k_0$ is the unperturbed coupling constant. Solving for equations
in each order of $\epsilon$ sequentially one obtains,
\begin{eqnarray}
F_0 &=& u ( x_1, x_2, t_1, t_2) e^{ i (q x_0 - \omega t_0 )} + c.c. \; ,\\
F_1 &=& {3 \over 2} | u |^2 +
{{ 3 u^2 }\over { -4 \omega^2 + 4 k_1 q^2 + 2 }}
e^{ 2 i (q x_0 - \omega t_0 )} \nonumber \\
& &+ c.c. \; ,
\end{eqnarray}
and the dispersion relation: $\omega^2 = \omega_0^2 + {k_0 } q^2 \;$, with
$\omega_0^2 = 2 $. From the vanishing of the secular equation at $ q = 0 $
one obtains the perturbed Nonlinear Schr\"odinger equation (NLS) at order
$\epsilon^2$:
\begin{equation}
2 i \omega { \partial u \over \partial t_2} +
{{\partial k_1} \over {\partial x_1}}
{{\partial u} \over {\partial x_1}} +
k_1 {{\partial^2 u} \over {\partial x_1^2}}
+ 8 u | u |^2 = 0 \;.
\end{equation}
We can further rescale the equation into a standard form:
Defining the following new dimensionless variables,
\begin{eqnarray}
\hat k ( \hat x ) &=& {k_1 \over k_0} - 1 \, ,\\
\hat u &=& \sqrt{ { 8 } \over { \omega^2}} u \, , \label{u}\\
\hat x &=& \sqrt{ { \omega^2} \over { 2 k_0 } } x_1\, , \label{x}\\
\hat t &=& {\omega \over 2} t_2 \, ,
\end{eqnarray}
with $\hat k$ being the normalized deviation coupling in the vicinity of the
enzyme, one obtains the following perturbed dimensionless NLS,
\begin{equation}
\label{PNSE}
i {\hat u}_{\hat t} + {1 \over 2} \hat u_{\hat x \hat x} +
\hat u | \hat u |^2 + {1 \over 2}
{{\partial } \over {\partial {\hat x}}} ( {\hat k}
{\hat u}_{\hat x} ) = 0 \;,
\end{equation}
and the corresponding Lagrangian density,
\begin{equation}
\label{Lagrangian}
\Lambda = {i \over 2} ( {\hat u}^* {\hat u}_{\hat t} -
{\hat u} {\hat u}_{\hat t}^*)
- {1 \over 2} ( { 1 } + {\hat k} ) | {\hat u}_{\hat x} |^2
+ {1 \over 2} | {\hat u} | ^ 4 \; .
\end{equation}
In the following section we drop the $\;\hat{}\;$ for
nomenclature simplicity.
\section{One soliton collective coordinate analysis.}
The collective coordinate method, which is a particle description of the
soliton in contrast to the field description given by the Lagrangian,
provides a good way to study the influence of a perturbation on a soliton.
The spirit is the same as using the center of mass to analyze the behavior
of a system of particles.
Without the perturbating term in Eq.~(\ref{PNSE}), one has a breather
solution
\begin{equation}
\label{breather}
u ( x , t ) = \eta
{\rm sech}
[ \eta ( x - u_e t ) ]
e^{i {u_e} ( x - u_c t )} + c.c.
\;,
\end{equation}
in which $\eta = \sqrt {(u_e^2 -2 u_e u_c)/(2PQ)}$,where $u_e$ is the
envelope velocity, $u_c$ the carrier velocity, and $P = 1 / 2$, $Q = 1$
are coefficients of the second space derivative and the nonlinear terms in
Eq.~(\ref{PNSE}) respectively. In view of this solution, we use an {\em
ansatz} for the collective coordinate analysis
\begin{equation}
\label{ansatz}
u ( x , t ) = \eta\, {\rm sech} ( {{\eta x} } - \zeta )\,
e^{i (\phi + \xi x )} \;,
\end{equation}
where the parameters $\eta$, $\zeta$, $\phi$, $\xi$
are functions of $t$.
For an unperturbed system this implies the following relations between
the parameters:
\begin{eqnarray}
\eta &=& \sqrt {{u_e^2 - 2 u_e u_c} } \, , \label{eta}\\
\zeta &=& {u_e \eta t }\, , \label{zeta}\\
\xi &=& {u_e }\, , \label{xi}\\
\phi &=& - {{u_e u_c t} } \,.
\end{eqnarray}
At $t = 0, \zeta = 0$ and $\phi = 0$ and there are only two parameters left,
which is consistent with Eq.~(\ref{breather}), because the NLS breather is a
two-parameter solution. Even when the breather is far away from the defect,
because the ansatz extends to infinity and always feels the defect, we do
not expect these relations to hold for a perturbed system. Hence in what
follows we examine the whole four-parameter space for the equations of
motion.
Introducing this ansatz into the Lagrangian density,
Eq.~(\ref{Lagrangian}), and integrating over space, one obtains an
effective Lagrangian,
\begin{eqnarray}
\label{eqlagrange}
L &=& - 2 \eta \phi_t - 2 \zeta \xi_t +
{{ \eta^3 } \over 3} - \xi^2 \eta
- {1 \over 2} \int_{- \infty}^{+ \infty} k | u_x |^2 d x
\; .
\end{eqnarray}
and the corresponding Hamiltonian
\begin{eqnarray}
\label{Hamiltonian}
H &=& - {\eta^3 \over 3} + \xi^2 \eta
+ {1 \over 2} \int_{- \infty}^{+ \infty} k | u_x |^2 d x
\; ,
\end{eqnarray}
which contains no momentum term.
At this point, we must specify an expression for $k(x)$ to proceed. For
algebraic convenience let us choose
\begin{equation}
\label{step}
k = \kappa [ \Theta ( x + l ) - \Theta ( x - l ) ] \; ,
\end{equation}
in which $\Theta$ is the Heaviside step function and $l$ is the half-length
of the defect. This form of $k$ violates Eq.~(\ref{assume}), however
previous works showed that the collective coordinate results are generally
robust for the treatment of dynamics in the presence of
perturbation\cite{SB3}, therefore we can expect to get results which are at
least qualitatively correct in spite of this rather crude approximation.
Moreover we shall check them against full numerical simulations in the next
section.
Introducing the following abbreviated notation:
\begin{eqnarray}
T_+ &=& \tanh (\eta l + \zeta ) \, ,\\
T_- &=& \tanh (\eta l - \zeta ) \, ,\\
S_+ &=& {\rm sech} (\eta l + \zeta ) \, ,\\
S_- &=& {\rm sech} (\eta l - \zeta ) \, ,
\end{eqnarray}
one obtains
\begin{equation}
\int_{- \infty}^{+ \infty} k | u_x |^2 d x = {\kappa \over 3}
( T_+^3 + T_-^3 ) \eta^3 + \kappa ( T_+ + T_- ) \xi^2 \eta \; ,
\end{equation}
which characterizes the effect of the defect and decays fast towards zero
soon outside of the impurity region, and the equations of motion:
\begin{eqnarray}
\phi_t &=& {{ {\eta^2} \over 2} } - { {\xi^2 \over 2} }
- { \kappa \over 4} ( T_+ + T_- ) \xi^2
- {{\kappa l } \over 4} ( S_+^2 + S_-^2 ) \xi^2 \eta
\nonumber \\
& &- {{\kappa l } \over 4} ( S_+^2T_+^2 + S_-^2T_-^2 ) \eta^3
- {{ \kappa} \over 4} ( T_+^3 + T_-^3 ) \eta^2 \, ,
\\
\xi_t &=&
- {{\kappa } \over 4} ( S_+^2T_+^2 - S_-^2T_-^2 ) \eta^3
- {{\kappa } \over 4} ( S_+^2 - S_-^2 ) \xi^2 \eta \, , \\
\zeta_t &=& \xi \eta + {{\kappa} \over 2} ( T_+ + T_- ) \xi \eta \, ,
\label{CZ}\\
\eta_t &=& 0 \label{CA} \, .
\end{eqnarray}
As expected, far away from the defect, {\it i.e.} when $S$ and $T$ vanish,
one recovers the usual relations for the NLS equation because in this case
$\xi_t = 0$ so that $\xi$ is a constant that we can denote by $u_e$. Then
$\zeta_t = \xi \eta$ gives $\zeta = u_e \eta t$ as expected, and $\phi_t
= ( \eta^2 - \xi^2) / 2$ gives
$\phi = - u_e u_c t $ if $u_c$ is defined by Eq.~(\ref{eta}) for $\eta$.
In the presence of the defect, the set of nonlinear differential equations
for the collective variables cannot be integrated analytically. It is
however much simpler than the full set of discrete equations since it
contains only 3 equations. It can be integrated by a fourth-order
Runge-Kutta method. One can however make general remarks on the properties
of the solution before resorting to numerical calculations. The soliton
described by the ansatz is an unbreakable entity and moreover the energy
given by Eq.~(\ref{Hamiltonian}) is conserved even when a potential well is
encountered.
Therefore when the soliton reaches a defect, it may speed up to compensate
for the extra energy requirement due to a decrease in coupling energy, as
shown in Fig.~\ref{CC}(a), in which a large carrier velocity was chosen to
exaggerate the effect. However, the behavior is richer than the one
generally found for topological solitons because, in addition to the time
dependent position $\zeta(t)$, the ansatz contains an internal degree of
freedom, $\xi$, so that the energy can also be transferred between different
collective coordinates. With $\kappa < 0$ the last term of
Eq.~(\ref{Hamiltonian}) decreases in the region of the defect, therefore
$\xi^2 $ has to increase accordingly. If $\kappa > 0$ and the breather is
initially inside the defect, it simply slips away as in Fig.~\ref{CC}(b).
If it is initially outside, for some suitable range of amplitude it is
first slowed down and eventually reflected, as shown in Fig.~\ref{CC}(c);
reflection occurs beyond $\eta \approx 0.25$. For smaller $\eta$, breathers
pass through the defect, indicating that the broader breathers are less
influenced by the presence of defects, just as a large-wheel bike will not
be stopped by a pebble or a ditch. In Fig.\ref{CC}(c) the breather has
actually penetrated into the defect before being reflected. When the
breather is trapped it oscillates between two positions, which may not be
the defect boundary, as shown in Fig.~\ref{CC}(d). For values of $\eta$ close
to the threshold between trapping and non-trapping, the breather slowly
turns around at the boundaries as in Fig.~\ref{CC}(e).
A {\it necessary} condition for a moving breather to be trapped in the above
defect is $\kappa < 0$. This statement can be proved through the following
argument: a necessary condition for trapping is $\zeta_t = 0$ more than
twice, which, according to Eq.~(\ref{CZ}), is equivalent to
\begin{equation}
\cosh^2 \zeta =
1 - \cosh^2 ( \eta l ) - {\kappa} \sinh ( \eta l ) \cosh ( \eta l ).
\label{trap}
\end{equation}
Since $\cosh^2 > 1$, $k_0 > 0$ and $\eta > 0$,
$\kappa $ has to be less than zero.
We have therefore showed that trapping occurs only if the perturbed
coupling constant is less than the unperturbed one, which is consistent
with our simulations although it has been proven only in the collective
coordinate approach.
Since Eq.~(\ref{trap}) contains only $\zeta$, $\eta$, $\kappa$, and $l$, if
the characters of the defect, {\it i.e.} the length, $l$, and the strength,
$\kappa$, have been fixed for a given system, and the initial position of
the breather is chosen, the only factor which characterizes trapping is the
breather amplitude. The initial value of $\phi $ seems to have no
consequence on the results. In general, if one finds that a breather passes
through a defect for
$\kappa < 0$ as in Fig.\ref{CC}(a), one can obtain trapping by increasing
its amplitude.
Because of the helicoidal structure of DNA, a given strand is alternatively
inside and outside the bend so that it experiences a periodical modulation
of its elasticity by an attached enzyme. We examined the consequence of such
a modification by considering the following coupling constant modulation:
\begin{equation}
k = \kappa [ \Theta ( x + l ) - 2 \Theta ( x ) + \Theta ( x - l ) ]
\label{SS}
\end{equation}
for which the coupling is first increased by $\kappa$ and then decreased by
the same amount if $\kappa > 0$. It can be viewed as a step approximation
of one period of a sinusoidal modulation. In general one finds nothing
essentially new with this perturbation because it is only a superposition
of two step defects. However, this perturbation is asymmetric in space. By
changing the sign of $\kappa$ we can reverse the orientation of the
perturbation with respect to an incoming breather. If $\kappa < 0$, a breather
starting from the left side of the defect encounters first the region where
the coupling constant is decreased.
Table I summarizes the behavior of breathers with various amplitudes and
the two possible signs of $\kappa$ and $\xi$. For negative $\kappa$ and
positive $\xi$, the breather is reflected for intermediate $\eta$ while for
large enough $\eta$ it is trapped. If one switches to positive $\kappa$
there is still a range of $\eta$ values that produce reflection, but for
large $\eta$ the breather passes through. If the breather starts from the
side where coupling constant is decreased the trapping can still exist even
if the initial position of the breather is far away from the defect, but the
pass-through region disappears as expected. These results show that it is
the first encounter which determines the trapping. However, for this case of
a composite defect the collective coordinate calculation can, in some cases,
lead to qualitatively wrong results. The full numerical calculation shown in
Fig.\ref{all}(a) indicate that the breather can be trapped even if it were
coming from the higher side of the defect. This points out the limit of the
collective coordinate method for successive perturbations of the breather.
The first interaction of the breather with a perturbation appears to be
qualitatively well described. But then the perturbed breather is not
accurately described by the ansatz. Thus when it encounters a second
perturbation (here the second step in coupling constant), the collective
coordinate description fails to describe the interaction.
\section{Direct numerical simulations}
Since the last example has shown that the collective coordinates cannot
provide a full description of the breather dynamics, it is necessary to
check them against full numerical simulations of Eqs.~({\ref{ND}). Using the
breather solution given by Eq.~(\ref{breather}) as an initial condition,
and periodic boundary conditions, we integrate Eqs.~({\ref{ND}) with a
4$^{\text{th}}$ order Runge-Kutta scheme and a time step chosen to provide
a conservation of energy to an accuracy better than $10^{-6}$ over a full
simulation. The calculations have been tested on different system sizes to
make sure that the results are not modified by boundary effects. The ansatz
(\ref{breather}) is not an exact solution of the full set of equations
because the transformation to the NLS form involved several approximations,
however, except for very discrete cases or large amplitude breathers, it
provides a rather good solution far away from the defect. As long as the
breather is far away from the defect, one generally notices only a small
decay of the initial energy peak due to radiation.
Full numerical calculations have the advantage of allowing radiation and
breaking of a breather. Furthermore, although the collective coordinate
method starts from the perturbed NLS which requires small $u_e$ and $u_c$
and hence small amplitude, the full numerical calculations do not have this
restriction. In what follows we show both energy distribution and the
breather amplitude. The energy distribution is more relevant to the opening
of DNA chain while the breather amplitude allows a comparison with the
results of the collective coordinate calculations.
Fig.~\ref{DS}(a) is a typical case for trapping at an equivalent amplitude
$\eta = 0.19$. The correspondence is made from Eq.~(\ref{eta}) and
Eq.~(\ref{xi}). The threshold for trapping predicted by collective
coordinate is higher ($\eta = 1.01$ for a breather initially at
$x = 12$ when both systems have the same dimensionless coupling constant).
As noticed earlier, it is not surprising to find such a discrepancy because
we have used a sharp perturbation that violates the condition (\ref{assume}).
However the full simulations confirm the qualitative predictions of the
collective coordinate calculations: small amplitude breathers are
transmitted, while larger ones are trapped. Fig.~\ref{DS}(b) shows that the
energy distribution around the breather gets sharper in the region of
the defect. Therefore a negative perturbation, which tends to trap
breathers, is also favorable for base-pair opening since it concentrates the
energy of the incoming breathers in a narrow domain. This sharpening of the
breather shape occurs when the breather is inside the perturbation domain,
whether it will stay trapped or not. One can also find on the contrary
that, if a breather meets a positive perturbation, its energy distribution
broadens. This behavior is similar to that of a vortex in shallow water: the
vortex becomes wider when it is in shallower water and thinner in deeper
water.\cite{T} In the amplitude plot of Fig.~\ref{DS}(b), when the breather
reaches the boundary of the defect, one can see two small reflected waves.
They were not included in the collective coordinate analysis, and their
presence explains part of the quantitative discrepancy between the
analytical approach and the full simulations. Sometimes one can also notice
that the breather changes its oscillation frequency after the collision
with the defect.
The results of the full numerical simulations show that, although the
collective coordinate analysis is able to predict qualitatively the main
features, in particular the existence of a threshold for trapping when the
breather amplitudes increases, it is quantitatively wrong. The same
conclusion had been found for an isolated impurity\cite{FPM}. There are
several reasons for that. Firstly we do not know an appropriate ansatz for
the original equations of motion (\ref{discrete}) and we start from a
perturbed NLS Lagrangian which is already approximate. Then we use an ansatz
which is localized in space and does not allow for the breaking of the
breather or the emission of reflected waves. And finally the calculation
assumes a smooth evolution of the coupling constant while we later use a
sharp variation to make the analytical calculation possible. In spite of
all their weaknesses, the collective coordinate calculations are however
useful to get an insight of the behavior of the breather in the presence of
the defect or even draw general conclusions on the kind of defects that can
trap energy as explained above.
Another point of interest is the trapping of {\it several} breathers in the
defect region which could really enhance the energy density locally and
cause local openings in DNA. We show in Fig.~\ref{all} examples for trapping
for two kinds of coupling constant shapes: Fig.~\ref{all}(a) is an example
when the breather comes from the higher side of a two-step defect but is
trapped. However, unless in favorable conditions we seldom find that two
breathers can be trapped inside the same perturbation.
When the second breather gets trapped it often kicks out the first breather
that was trapped before, as shown in Fig.~\ref{all}(b). In other cases we
noticed that, when a first breather is trapped in the defect, a second
breather that would have been trapped if it were alone, is on the contrary
reflected. Therefore, if one studies only the positions of the breathers
during their first interactions with the extended defect, it seems that the
defect will never collect more than the energy of one breather. This is in
fact not true, but the complete phenomena require a more detailed analysis.
It is interesting to study the evolution of the energy in the
region of the defect versus time.
An example is shown on Fig.~\ref{figu5}. In this
case the first breather that interacts with the defect has an amplitude
$\eta = 0.2$ which is above the trapping threshold and the second one has an
amplitude $\eta = 0.1$ below the threshold. As expected the first breather
is trapped and oscillates around the defect. The second one passes
through the defect region that contains the first breather. However, if one
looks at the energy density on the
three-dimensional plot of Fig.~\ref{figu5}(a), one
can notice a significant increase of energy density after the interaction of
the second breather with the defect. The reason is that the second breather
is only {\it partly} transmitted. A large part of its energy is given to the
trapped breather, i.e. it stays in the defect region. The same phenomenon
occurs again when the second breather collides a second time with the
trapped breather. Due to this complex process, the time evolution of the
energy inside the defect region
(Fig.~\ref{figu5}-c) is a complicated curve, but it is
important to notice that it tends to grow, and never falls again to a small
value, indicating that the multiple collision process does cause a
concentration of energy in the defect region. The origin of this
localization of energy does not lie in breather trapping but in breather
interactions in the presence of a perturbation, and therefore it is not
included the collective coordinate description of Sect.~IV. The result is
very reminiscent of a mechanism described recently for energy localization
due to discreteness effects in nonlinear lattices\cite{Nlenloc}. In both
cases the collisions of breathers, perturbed either by a defect or by
discreteness, cause energy transfers that, on average, favor the big
excitation at the expense of the small one. We have checked that the
mechanism is not restricted to a particular case.
Fig.~\ref{figu6} shows another
example in which 3 breathers with the same initial amplitude
$\eta=0.2$ were sent to the defect. Although the details of the process are
different, they lead to the same final result: breather interactions in the
presence of the defect tend to favor the formation of a large amplitude
breather that concentrates a large part of the energy of the three incoming
breathers and is finally trapped at the defect site. Hence the energy in the
defect region settles to a high value. Tests have been performed with
various breather amplitudes, leading to the same general result.
\section{Conclusion}
Using a simple DNA model we have modeled the effect of a transcription enzyme
by an extended modification of the coupling constant along the strands. The
results show that such a perturbation is more efficient than an isolated
impurity to trap breathers, in particular because trapping can occur provided
that the amplitude of the incoming breather exceeds a threshold instead of
requiring breathers with a well defined frequency. This conclusion can be
derived from collective coordinate calculations as well as from numerical
integration of the full set of equations of motion, although the collective
coordinate method overestimates the trapping threshold. One cannot expect
quantitative results from the collective coordinate analysis because we have
violated at least one basic assumption, Eq.~(\ref{assume}), to allow the
analytical calculations, but it gives insight into the physics, and in
particular a necessary condition for breather trapping which is confirmed by
the full simulations.
We have also showed that energy exchanges between a first breather, already
trapped, and other incoming breathers can lead to a concentration of energy
in the region of the defect.
One may wonder whether the results obtained above for specific perturbations
are extendable to more realistic cases. Although it is difficult to give
general answers to this question, one can get insights through numerical
simulations of the full system. In real DNA one has
$D = 0.03 eV$, $\alpha = 4.45 \AA^{-1}$, $k_1 = 0.08 eV/\AA^2$,
$m=300$ a.m.u. for AT base pairs, while for GC pairs we have $D=0.035 eV$,
$k_1 =0.104$. \cite{Dauxois} This is equivalent to $k_n \approx 0.13$ to
$0.15$. In this range of coupling large amplitude breathers are trapped by
discreteness\cite{BangP}. The collective coordinate calculations suggest
that the low amplitude breathers, which can move, will not be trapped by the
20 base-pair defect. Simulations show that it is not necessarily so. For
instance a 20 base-pair defect with $K_n=0.12$ in a chain with $K_n=0.15$
can trap breathers of various amplitudes. The energy exchange mechanism in
the presence of the defect, discussed above, interferes with discreteness
effects that can have similar effects to localize energy\cite{Nlenloc}.
Therefore, although we have exhibited a mechanism which is active in a wider
frequency range than an isolated defect, the calculations performed on a
simple model are not be sufficient to draw a conclusion about its validity
to describe the effect of an enzyme on DNA transcription. It may however
deserve attention because of its greater efficiency compared to the case
of a point defect that was considered previously.
In this work we have modeled the role of the enzyme by modulating only the
coupling constant along the strands. As mentioned in the introduction, other
possibilities could be considered, particularly if one attempts to take into
account the enzyme specificity which suggests that the enzyme could have
another role than merely bending locally the molecule. As a first step in
this direction, we have considered a local change of the Morse potential in
addition to the effect of the bending.
Figure \ref{figu7} shows the result of a
numerical simulation where all the conditions are the same as for
Fig.~\ref{figu6},
except that, in addition to changing the coupling constant inside the defect
to model the bending, we have also multiplied the denaturation energy of the
base pairs (parameter $D$ of Eq.~(\ref{discrete})~) by a factor 0.8. This
means that we also assume that the enzyme can have some chemical effect to
reduce the base pairing interaction. The comparison of Fig.~7 and Fig.~6c
shows that this modification has a rather drastic effect on the results. It
is easy to understand qualitatively why because locally the vibrating
frequency of base pairs has been reduced. As we consider low energy
breathers which are the most likely to be excited at physiological
temperatures, their frequency, situated below the base-pair linear frequency
of the unperturbed region because of the soft nonlinearity of the Morse
potential, is however very close to the bottom of the phonon band of the
unperturbed part of the molecule. As the enzyme lowers the frequencies of
phonon band in the defect region, the breather frequency is now {\it in
resonance} with some modes of the phonon band of the defect. Therefore when
the breather is trapped at the defect site by the bending, it is trapped in
a region where it resonates with phonons. As a result it loses energy by
radiation, but, as the emitted modes have a frequency below the lowest
frequency of the unperturbed lattice, the vibrations are trapped in the
defect region. One observes that the trapped breather spreads out its energy
inside the defect region. When a second breather comes to this excited
defect it is no longer repelled by a highly
localized breather as in Fig.~\ref{figu6}.
Thus it is more likely to penetrate in the defect region too. This makes the
energy localization effect more efficient and instead of the large
oscillations that were observed in Fig.~\ref{figu6}-c,
Fig.~\ref{figu7} shows that the energy in
the defect region now grows steadily, each new breather having a high
probability to add its contribution. Although it is still preliminary, this
example shows that, if one combines the bending effect of the enzyme with
some model for its specific action on the promoter site, one can perhaps
provide a mechanism to achieve the local opening of the double helix which
is required by DNA transcription.
\section{Acknowledgments}
We would like to thank T. Dauxois for helps on computer programming.
J.J.-L.T. acknowledges the hospitality of the Laboratoire de
Physique de l'Ecole Normale Sup\'erieure de Lyon where part of this work
was done. J.J.-L.T. also acknowledges
partial support of National Science Council, Taiwan,
grant No. 84-2911-I-007-030-B21. Part of this work has been supported by the
EU Science Program through grant SC1*CT91-0705.
|
\section{Introduction}
Recently there has been a lot of interest in the difference in the
measured values of $\alpha_s$ in high and low energy experiments. A
very useful summary of the issues have been given in a recent paper by
Shifman\cite{shifman} suggesting that the observed discrepancy between
the higher values of $\approx 0.125$ for $\alpha_{s}(M_Z)$ derived
from the global fit of LEP/SLC data assuming only the SM particle
content and interactions on the one hand and lower values near
$\approx 0.11$ derived from the low energy data such
as deep inelastic electron scattering \cite{altarelli,vir}, lattice
calculations \cite{el-kadra}
involving the upsilon and the $J/\Psi$ system etc on the other should be
considered to be an indication of the presence of new
physics\cite{langa}. If this new physics is identified with the
supersymmetric version of the standard model at low energy, then one
can attempt to do a global fit to all LEP/SLC data including the
supersymmetric particles and interactions and see if indeed the high
value of $\alpha_{s}(M_Z)$ indicated there is lowered. Such an
analysis has been carried out recently by several groups\cite{kane},
who have shown that if the stop and the chargino masses
are kept below a 100 GeV, then there are new contributions to the
$Z\rightarrow b\overline{b}$ decay which increase its decay width. In
the presence of these contributions, the global fit to LEP/SLC data indeed
leads to a value for $\alpha_{s}(M_Z)
\simeq .112$ which is what the low energy data give for this parameter.
Of course it should be noted that the values of parameters used in the
above discussion may not be obtainable in simple SUSY GUT theories.
It should be noted that there are a number of other suggestions that
could also lead to a higher value for the $Z\rightarrow b\overline{b}$
width. It could very well be that one of these scenarios rather than
the SUSY contribution is at the real heart of the problem. But for our
discussion of unification, it important that either the experimental
value of the $Z\rightarrow b\overline{b}$ come down or that the
supersymmetric scenario provide an explanation for its enhancement
over the standard model value so that the value of $\alpha_{s}(M_Z)$
gravitates towards its lower value. As we discuss below, this low value
will have profound implications for the nature of SUSY GUT theories.
The LEP measurements of the gauge couplings
$\alpha_1,\alpha_2$ and $\alpha_{s}$ combined with the
coupling constant evolution dictated by the minimal supersymmetric
standard model, have in the past three years led to speculations that
the present data while providing overwhelming support for the standard
model may in fact be indicating that the next level of physics
consists of a single scale supersymmetric grand unified theory, with
new physics beyond supersymmetry appearing only at the scale of
$10^{16}$ GeV\cite{amaldi}. This has generated a great deal of
excitement and activity in the area of supersymmetric grand unified
theories(SUSY GUT). Consequently, the renormalization group (RG) analysis
have become more
refined over the time including the various low and high scale threshold
corrections. The effect of low energy threshold corrections has been
included in the simple step function approximation as well as by including
a detailed mass dependent renormalization scheme \cite{bagger, mar, chan}.
These analysis conclude that the effect of the low energy threshold effects
is to increase the predicted value of $\alpha_s $. The GUT scale threshold
corrections to $\alpha_s$ come mainly from two sources in the minimal scheme-
the doublet-triplet splitting and the mass splitting among the heavy scalars
present at the GUT scale. The threshold effects from the well known
doublet-triplet splitting always increases the
prediction of $\alpha_s$. The threshold effects from
the splitting among the heavy scalars can be of either sign depending on
the mass spectrum of the heavy scalars. In the minimal SU(5) GUT the heavy
scalars reside in the 24 dimensional adjoint representation of SU(5) and the
spectrum the masses of these heavy scalars are quite constrained. A
combined analysis including the heavy and light threshold corrections in
the minimal SU(5) GUT predicts a value of $\alpha_s$ around 0.126\cite{bagger}
when the
superpartner masses are at the TeV scale. If the s-particle masses are
less than 1 TeV the prediction of $\alpha_s$ increases further. For example,
for s-particle masses around 500 GeV, the predicted value of $\alpha_s$ is
0.130, even beyond the value measured at LEP. Clearly, this leads to a
conflict. On the one hand to reduce the prediction of $\alpha_s$ in a SUSY
GUT one needs a high value of the s-particle masses; on the other hand, to fit
the LEP/SLC data with a low $\alpha_s$ in a supersymmetric model, one needs the
stop and the chargino below 100 GeV.
In this brief review, we will summarize the arguments leading one to
conclude that if the value of
the QCD fine structure constant $\alpha_{s}(M_Z)$ at the weak scale turns out
to be around $0.112$ as is indicated by several low energy experiments then
to reconcile this value with the predictions for supersymmetric GUTs
one necessarily needs new physics beyond the usual minimal SUSY GUT
scenarios, either at $10^{11}-10^{12}$ GeV or at
the GUT scale. Several new physics possibilities
have been identified in recent
literature such as, (a) introducing non-universality of gaugino masses at
the GUT scale; (b) introducing heavy threshold effects coming from the heavy
scalar multiplets at the GUT scale, when the scalar sector of the minimal
SUSY SU(5) model is altered to include $50$, $\overline{50}$, and $75$
representations of SU(5); (c) introducing a superstring-inspired scalar
spectrum in a supersymmetric SO(10) GUT and making room to incorporate an
intermediate B-L symmetry breaking scale; (d) introducing an extra mini
doublet-triplet splitting near the GUT scale. This paper is organized as
follows. In section 2, we state the basic notions; in section 3 we
outline how the non-universality of
gaugino masses can lead to a lowering of the prediction of $\alpha_s$; in
section 4, we describe the heavy threshold corrections in the SU(5) model
leading to the change in the prediction of $\alpha_s$; in section 5, we
describe the introduction of the string inspired scalar spectrum in the
supersymmetric SO(10) GUT - and the consequent lowering of the prediction
of $\alpha_s$ and in section 6, we describe the extra mini doublet triplet
splitting in the GUT scale. In section 7, we present our conclusions.
\section{Basic notions}
To illustrate our basic procedures, let us consider a toy example in which
all the superpartners are degenerate at the scale $M_Z$, excepting the
gluinos and the winos, which are somewhat heavier than the scale
$M_Z$. In such a scenario, the three gauge couplings at the scale $M_Z$ can
be related to the unification coupling by the relations,
\begin{eqnarray}
2 \pi \alpha^{-1}_s~(M_Z)&=& 2 \pi \alpha^{-1}_U + b^{susy}_s \ln {M_U
\over M_{\tilde{g}} } + [ b^{susy}_s - \Delta_{\tilde{g}}] \ln
{M_{\tilde{g}} \over M_Z}, \nonumber\\
2 \pi \alpha^{-1}_2~(M_Z)&=& 2 \pi \alpha^{-1}_U + b^{susy}_2 \ln
{M_U \over M_{\tilde{w}} } + [ b^{susy}_2 - \Delta_{\tilde{w}}]
\ln{M_{\tilde{w}} \over M_Z},
\nonumber\\
2 \pi \alpha^{-1}_1~(M_Z)&=& 2 \pi \alpha^{-1}_U + b^{susy}_1 \ln{M_U
\over M_Z}, \label{rge}
\end{eqnarray}
where, $b^{susy}_i$ are the usual one-loop supersymmetric beta function
coefficients and $\Delta_{\tilde{g}}$ and $\Delta_{\tilde{w}}$ are the
contributions from the gluino and the wino loops to $b^{susy}_s$ and
$b^{susy}_2$ respectively and $M_U$ is the grand
unification scale. We notice that the quadratic Casimirs of
the SU(2), U(1) and SU(3) groups have the values 2,0 and 3 respectively.
Using a vector orthogonal to (2,0,3) and (1,1,1) we construct the
combination,
\begin{equation}
c= 2 \pi [ 3 \alpha^{-1}_2(M_Z)-\alpha^{-1}_{1Y}(M_Z)-2
\alpha^{-1}_s(M_Z)]. \label{c}
\end{equation}
Combining Eqn.(\ref{rge}) and Eqn.(\ref{c}) we have,
\begin{eqnarray}
c=[3 b^{susy}_2-b^{susy}_1-2b^{susy}_s] \ln{M_U \over M_Z}
- 3 \Delta_{M_{\tilde{w}} } \ln {M_{\tilde{w}} \over M_Z} +2
\Delta_{\tilde{g}} \ln {M_{\tilde{g}} \over M_Z}. \label{comb}
\end{eqnarray}
In the right hand side of Eqn.(\ref{comb}) the gauge and the fermionic
contributions to the beta function coefficients cancel out due to the
orthogonality that has been mentioned earlier, where as the scalar
contribution does not. At this stage, using $\Delta_{\tilde{w}}=4/3$
and $\Delta_{\tilde{g}}=2$, we have the prediction of $\alpha_s$, as,
\begin{equation}
\alpha^{-1}_s(M_Z)={ 1\over 2}~[3 \alpha^{-1}_2-\alpha^{-1}_1] -{3 \over
5 \pi} \ln{M_U \over M_Z} + { 1\over \pi} \ln {M_{\tilde{w}} \over M_Z} -
{ 1\over \pi} \ln {M_{\tilde{g}} \over M_Z}, \label{predic}
\end{equation}
where, the second term is the threshold correction due to the doublet
triplet splitting, in which $M_U$ and $M_Z$ are the masses of the
superheahy triplet and the light doublet higgs scalars. It is interesting to
note that the `gluino'-term
and the `wino'-term in the right hand side of Eqn.(\ref{predic}) have a
relative sign among them.
\section{Non-universality of gaugino masses}
In the minimal scenario the spontaneous breaking of supergravity yields a
global supersymmetric theory supplemented by a set of soft
supersymmetry-breaking parameters. In particular in SUSY GUT
theories, gauge invariance implies that at the GUT scale, one must have
a universal gaugino mass ($m_{1/2}$). Using the RG evolution, if one runs
the gaugino masses from the GUT scale to the lower scales, a relation
between the mass of gluionos and the mass of the winos is gotten
\cite{nilles}. If
the masses of the s-particles are near the electroweak scale one gets,
\begin{equation}
x \equiv {m_{\tilde{g}}(M_Z) \over m_{\tilde{w}}(M_Z)} = {\alpha_{2}(M_Z)
\over \alpha_s (M_Z)} \sim 3.3. \label{ratio}
\end{equation}
This relation is of considerable significance regarding the prediction of
$\alpha_s$ in a supersymmetric GUT as we will explain now. To predict $\alpha_s$
we use the complete relation,
\begin{eqnarray}
\alpha^{-1}_s(M_Z)&=&{1 \over 2}~[3 \alpha^{-1}_2(M_Z) -\alpha^{-1}_{1Y}(M_Z)]
- {1 \over \pi} \ln{M_{\tilde{g}} \over M_Z}
{}~\theta(M_{\tilde{g}}-M_Z) \nonumber\\
&+& {1 \over \pi} \ln{M_{\tilde{w}} \over M_Z} ~\theta(M_{\tilde{w}}-M_Z)
+ T_{heavy} + T_{others}. \label{alstrong}
\end{eqnarray}
The theta functions in Eqn.(\ref{alstrong}) have the value 1
whenever the argument is positive and is zero otherwise. $T_{heavy}$
parametrizes the heavy threshold
corrections arising from the doublet-triplet splitting as well as from
the splitting among the heavy scalars. $T_{others}$ parametrizes the
effect of the light degrees of freedom apart from the winos and the
gluinos. We notice that, when the mass of the gluino is less than the mass
of Z,
neither wino nor the gluino contributes to Eqn(\ref{alstrong}). When the
mass of the gluino crosses the $M_Z$ range, the `gluino'-term, having a
negative
sign, starts contributing to the $\alpha^{-1}_s$, and consequently $\alpha_s$
starts to
increase. Only when the mass of the wino becomes comparable to the
electroweak scale, the `wino'-term starts contributing, which compensates
the increasing effect. In the Figure 1 \cite{mar} this phenomenon
is depicted \footnote{ There is a
10 \% decrease of $\alpha^{-1}_s$ at the 2 loop level.}. Notice that the
mass of the gluino is always more than the
mass of the wino due to the constraining relation given in Eqn.(\ref{ratio}).
\begin{figure}[htb]
\begin{center}
\epsfxsize=8.5cm
\epsfysize=8.5cm
\mbox{\hskip -1.0in}\epsfbox{mar.eps}
\caption{ The one and two loop predictions of $\alpha_s(M_Z)$ in
MSSM with universal scalar mass term at the GUT scale.\label{}}
\end{center}
\end{figure}
Clearly, to reverse the effect, one needs to relax the mass relation
between the gluino and the wino \cite{nonuni}. This approach has been
taken by Roszkowski
and Shifman \cite{rs}. It has been noted by them that out of all soft
masses the soft masses of the wino and the gluino have the dominant
influence on the prediction of $\alpha_s$. The reason for this is two fold,
\noindent (a) The soft mass scale of the wino influences only the SU(2)
beta function coefficient, whereas the soft mass scale of the gluino
influences only the SU(3) beta function coeffcient only.
\noindent (b) The contribution of the wino and the gluino in the beta
function coefficients of SU(2) and SU(3) groups respectively are the
largest among all superparticles.
Along with this observations we may also observe that if the constraint in
Eqn.(\ref{ratio}) is
relaxed [the x parameter is varied], one can alter the prediction of
$\alpha_s$ considerably. The two loop predictions of $\alpha_s$ by choosing
different values of x have been summarized in Figure 2 \cite{rs}.
\begin{figure}[htb]
\begin{center}
\epsfxsize=8.5cm
\epsfysize=8.5cm
\mbox{\hskip -1.0in}\epsfbox{shifman.eps}
\caption{ The contours of constant $\alpha_s (M_Z)$ for various
choices of the parameter `x'. \label{}} \end{center}
\end{figure}
One needs the gluino mass in the ball park of 100 GeV to predict
$\alpha_s=0.11$ by the above mechanism. Such a light gluino will have further
phenomenological consequences. For example, the gluino correction enhances
the hadronic width of $Z$. However, the $Z \rightarrow b
\overline{b}$ width increases too much for $M_{\tilde{g}} \sim 100$ GeV,
and one has to descend to unacceptably low sqark masses to reconcile with
the experimental $Z \rightarrow b \overline{b}$ width.
\section{Heavy thresholds and missing doublet SU(5)}
In the presence of heavy threshold corrections, the unification scale is
no longer well-defined. Let us, therefore, define a scale $\Lambda$,
which is larger than any GUT scale mass. Above the scale $\Lambda$ all the
couplings remain unified. Again we will consider a toy SU(5) example. Now,
as we are interested in the heavy thresholds only, let all the
superpartners of the standard model fermions and gauge bosons be
degenerate at the scale $M_Z$. The heavy spectrum of the minimal SU(5)
model is simple. The $(3, 2, 5/6) + (\overline{3}, 2, 5/6)$ components of
the $24$-scalar get absorbed by the heavy gauge bosons with a common mass
$M_V$. The SU(3)-octet and the SU(2)-triplet have a common mass
$M_\Sigma$ whereas the singlet has a mass $0.2M_\Sigma$. We can relate
the three gauge couplings at the scale $M_Z$ to the unified coupling at
the scale $\Lambda$ as \cite{hisano},
\begin{eqnarray}
\alpha_s^{-1}(M_Z)&=&[\alpha_U^{-1}(\Lambda) - { 2 \over \pi} \ln{
\Lambda \over M_V} + { 3 \over 2 \pi} \ln{ \Lambda \over M_\Sigma} +
{1 \over 2 \pi} \ln { \Lambda \over M_{H_3}}]
+ { 1 \over 2 \pi} b_s^{susy} \ln{\Lambda \over M_Z}, \nonumber\\
\alpha_2^{-1}(M_Z)&=&[\alpha_U^{-1}(\Lambda) - { 3 \over \pi}
\ln{\Lambda \over M_V} + {1 \over \pi} \ln{ \Lambda \over M_\Sigma}]
+ { 1 \over 2 \pi} b_2^{susy} \ln{\Lambda \over M_Z},
\nonumber\\
\alpha_1^{-1}(M_Z)&=&[\alpha_U^{-1}(\Lambda) - { 5 \over \pi}
\ln{\Lambda \over M_V} + {1 \over 5 \pi} \ln { \Lambda \over
M_{H_3}}]+{1 \over 2 \pi} b_1^{susy} \ln{\Lambda \over M_Z}.
\end{eqnarray}
Taking the combination c as in the previous section,
we find that it is independent of any field that
is in the adjoint of SU(5) and we recover the result,
\begin{equation}
\alpha_s^{-1}(M_Z)={1 \over 2} [3 \alpha_2^{-1}-\alpha_1^{-1}] - { 3 \over
5 \pi} \ln{M_{H_3} \over M_Z}.
\end{equation}
Now, it is clear why we interpreted the second term in Eqn. (\ref{predic}) as
the effect of doublet triplet splitting.
We notice a special property of the $24$-dimensional scalar in relation
to the combination c. While the components $(3,2,5/6) + (\overline{3},2,
-5/6)$ are absorbed as the logitudinal components of the heavy SU(5)
gauge bosons, the rest are in the adjoint representation of the low
energy groups. We have already noticed that fields which are in the
adjoint representation of the low energy groups and having a common mass
cannot contribute to the combination c. Now, we also note that even the
would be goldstone bosons do not contribute to c; and no other component
of the adjoint is left; except the singlet which does not contribute to
the beta function coefficients. So the splitting in the adjoint $24$,
does not affect the prediction of $\alpha_s$ at all.
The situation changes if we introduce 75-dimensional scalar instead of
the adjoint to break the unification symmetry along with the $50$ and
$\overline{50}$ dimensional representations needed for the missing
doublet mechanism \cite{masiero}. Twelve components of $75$ will be eaten
up by the
the heavy gauge bosons. Remaining 63 components will be split in mass
\cite{yamada}.
Moreover, the color triplets from $50$ and $\overline{50}$ Higgs scalars
get mixed with the color triplets of $5$ and the $\overline{5}$. The rest
of the components of $50$ and the $\overline{50}$ components are not
split among themselves. They acquire a common mass $M_{50}$. The
typical masses of the various GUT scale heavy scalars \cite{yamada} have
been summerized in Table\ref{split}.
\begin{table}[htb]
\begin{center}
\[
\begin{array}{|c||c||c|}
\hline
component&mass &comments\\
\hline
(8,3,0) & M_\Sigma & \\
(3,1, \pm 5/3) & 0.8 M_\Sigma & \\
(6,2, \pm 5/6) & 0.4 M_\Sigma &in~75 \\
(1,1,0) & 0.4 M_\Sigma &of~SU(5) \\
(8,1,0) & 0.2 M_\Sigma & \\
\hline
(3,1, \pm 1/3) & M_{D_1}& in~5+\overline{5} \\
(3,1, \pm 1/3) & M_{D_2}& and~50+\overline{50} \\
\hline
rest ~of~ 50+\overline{50}&M_{50} & \\
\hline
\end{array}
\]
\end{center}
\caption{The representations and masses of the heavy scalars in the
missing doublet SU(5) model. Bars have been suppressed in certain
representation for the compactness of presentation}
\label{split}
\end{table}
Given the masses and the transformation properties in Table\ref{split},
it is straightforward to calculate the change in the prediction of
$\alpha_s(M_Z)$. Indeed such a calculation have been performed in
Ref.\cite{bagger} in which they have carefully taken into account both
the heavy and the light threshold effects. They conclude that a large
negative correction to $\alpha_s$ comes in the missing doublet model due
to the splitting in the $75$ Higgs. The value of $\alpha_s=0.112$ is
achievable (taking into account a lower bound on a combination of
$M_{D_1}$ and $M_{D_2}$ from proton decay) with the s-particle masses
around the electroweak scale. For further details the reader is referred
to their papers.
\section{Intermediate scale in supersymmetric SO(10)}
There is a general understanding in the literature that the LEP measurements
of the gauge couplings at the scale $M_Z$ is a hint to a one step unified
theory. There are however several physical arguments based on neutrino
physics \cite{hdm,numass} as well as strong CP problem \cite{kim} suggesting
that there may
be an intermediate scale corresponding to a gauged (local) B-L symmetry
breaking somewhere around $10^{11}$ to $10^{12}$ GeV. However, such a scale
must not affect the gauge unification constraints. To examine this, let
us write down the RGE of the three couplings
relating them to the gauge coupling at
the unification scale $M_U$ to the ones at
the weak scale $M_Z$ and introducing a general intermediate scale $M_I$, as,
\begin{eqnarray}
2 \pi \alpha_1^{-1}(M_Z)&=& 2 \pi \alpha_U^{-1}(M_U) + b_1 \ln{M_I \over M_Z} + b_1
^\prime {M_U \over M_I}, \nonumber\\
2 \pi \alpha_2^{-1}(M_Z)&=& 2 \pi \alpha_U^{-1}(M_U) + b_2 \ln{M_I \over M_Z} +
b_2^\prime {M_U \over M_I}, \nonumber\\
2 \pi \alpha_s^{-1}(M_Z)&=& 2 \pi \alpha_U^{-1}(M_U) + b_s \ln{M_I \over M_Z}
+b_s^\prime {M_U \over M_I}. \label{rgint}
\end{eqnarray}
At first, we notice that the one-loop beta function coefficients $b_i$ in
MSSM
for the groups SU(3), SU(2) and U(1) are -3, 1 and 33/5 respectively.
Presently we are interested in the beta function coefficients
$b_i^\prime$ governing the slopes in the region between $M_I$ and $M_U$.
So, by taking a vector orthogonal to \\(-3,1,33/5) and {(1,1,1)} we
construct a second combination $c_1$ in which $b_i$'s get eliminated but
$b_i^\prime$ s survive, namely,
\begin{equation}
c_1=2 \pi (7 \alpha_s^{-1} - 12 \alpha_2^{-1} + 5 \alpha_1^{-1}). \label{c2}
\end{equation}
We combine, Eqn.(\ref{rgint}) and Eqn.(\ref{c2}) to get,
\begin{equation}
c_1=(7 b_s -12 b_2 + 5 b_1) \ln {M_I \over M_Z} +
(7 b_s^{\prime} -12 b_2^\prime +5 b_1^\prime) \ln {M_U \over M_I}
\label{c2cond}.
\end{equation}
Due to the orthogonality in the construction of
$c_1$ the coefficient of $\ln (M_I / M_Z)$
vanishes and we are left with,
\begin{equation}
c_1=(7 b_s^\prime -12 b_2^\prime +5 b_1^\prime)
\ln {M_U \over M_I} \label{c2b}.
\end{equation}
Moreover, if we assume that the values of $\alpha_1$, $\alpha_2$ and
$\alpha_s$ are such that an one-step unification is possible with
MSSM and a big desert - we get a condition from Eqn.(\ref{c2cond}) in the
limit $M_U=M_I$, which is,
\begin{equation}
c_1=0. \label{c20}
\end{equation}
Combining, Eqn.(\ref{c20}) and Eqn.(\ref{c2b}), we get,
\begin{equation}
(7 b_s^\prime -12 b_2^\prime +5 b_1^\prime)
\ln {M_U \over M_I} =0
\label{c2c}.
\end{equation}
We note the two solutions of Eqn.(\ref{c2c}).
\noindent (a) $M_U=M_I$ which leads us back to the one-step unification.
\noindent (b) Due to the presence of new fields above the scale $M_I$, the
beta function coefficients $b_i^\prime$ conspire among themselves to
produce\cite{sato},
\begin{equation}
7 b_s^\prime -12 b_2^\prime +5 b_1^\prime=0.
\label{cond}
\end{equation}
One can restrict the type of Higgs representations above the intermediate
scale $M_I$ by requiring that the supersymmetric SO(10) GUT emerges from an
underlying
superstring theory. If we restrict ourselves to only those Higgs scalars
which can arise from simple superstring models with Kac-Moody levels one
or two, we can have a restricted number of solutions of Eqn.(\ref{cond}).
We will consider an intermediate symmetry group ${G_I=U_{B-L}\times SU(2)_L
\times SU(2)_R \times SU(3)_C}$. The solutions can be characterized by three
integers ($n_C,n_H,n_X$), where $n_C$ refers to the number of
(0,1,1,8), $n_H$ means the number of (0,2,2,1)
fields and $n_X$ means the number of {(1,1,2,1)+(-1,1,2,1)} fields under
the intermediate symmetry gauge group $G_I$. The various
scenarios\cite{leemoh,we} are tabulated in Table\ref{table2}.
\begin{table}[htb]
\begin{center}
\[
\begin{array}{|c||c||c||c||c||c||c||c|}
\hline
model&I &II & III& IV&V&VI&VII\\
\hline
(n_C,n_H,n_X)&(0,1,2)&(0,1,3) &(0,1,4) &(0,2,3) &(0,2,4)&(0,2,5)&(1,2,4) \\
\hline
\end{array}
\]
\end{center}
\caption{The various models which satisfy the condition in Eqn.~15
approximately.} \label{table2}
\end{table}
\begin{figure}[htb]
\begin{center}
\epsfxsize=8.5cm
\epsfysize=8.5cm
\mbox{\hskip -1.0in}\epsfbox{lee.eps}
\caption{ The predictions of $\alpha_s$ are displayed as constant $\alpha_s$
contours for the various models given in Table 2.\label{fig3}}
\end{center}
\end{figure}
The predictions of $\alpha_s$ for different models \cite{leemoh,we} have been
plotted in Figure \ref{fig3}. We notice from Figure \ref{fig3} that the
models VI and VII
are particularly interesting for the value of $\alpha_s$ in the range of
0.11. A comment is in order. The
value of the intermediate scale ($M_I$) is a new parameter. But we notice
that even this new parameter is quite constrained by the the LEP measurements
of the couplings at the scale $M_Z$. Moreover, only three out of the seven
intermediate scale models in Table\ref{table2} will survive if
the value of $\alpha_s(M_Z)$ turns to be 0.112.
There is another interesting aspect of such intermediate scale SO(10)
GUTs. We see that the requirement of unification forces us to introduce
new scalar fields above the scale of $B-L$ symmetry breaking. These new
scalars will also affect the evolution of Yukawa couplings above the
intermediate symmetry breaking scale. The evolution of various Yukawa
couplings have been calculated in Ref \cite{we}. The results are
tabulated in Table\ref{table3} and Table\ref{table4}.
\begin{table}[htb]
\begin{center}
\[
\begin{array}{|c||c||c||c||c||c||c||c||c|}
\hline
Y_1&Y_2&h_t(M_t)&h_b(M_t)&h_\tau(M_t)&
tan\beta&m_b(M_t)&m_t(M_t)&{m_b(M_t) \over m_\tau(M_t)}\\
\hline
1&1 &1.010&0.96&0.62&60.43&2.77&176.83& 1.56 \\
1&10^{-1} &1.060&0.84&0.52&51.26&2.85&184.46&1.60 \\
1&10^{-2} &1.094&0.460&0.270&26.7&3.01&190.34&1.69 \\
1&10^{-3} &1.103&0.160&0.095&9.25&3.06&190.90&1.72 \\
1&10^{-4} &1.104&0.054&0.030&2.80&3.07&181.03&1.73 \\
1&{1 \over 2}10^{-4} &1.104&0.037&0.021&1.85&3.08&169.16&1.73 \\
\hline
\end{array}
\]
\end{center}
\caption{The values of $h_t(M_t)$, $h_b(M_t)$, $h_\tau(M_t)$ and
calculated by RGE for
$\alpha_s=0.11$ in model VI. The prediction of the
masses $m_b$ and $m_t$ at the scale
$M_t$ has been quoted in GeV. $M_t$ is defined as 170 GeV. $tan \beta$
has been calculated assuming $m_\tau(M_Z)=1.777$ GeV.} \label{table3}
\end{table}
\begin{table}[htb]
\begin{center}
\[
\begin{array}{|c||c||c||c||c||c||c||c||c|}
\hline
Y_1&Y_2&h_t(M_t)&h_b(M_t)&h_\tau(M_t)&
tan\beta&m_b(M_t)&m_t(M_t)&{m_b(M_t) \over m_\tau(M_t)}\\
\hline
1&1 &0.99&0.960&0.59 &57.81 &2.87&173.90&1.61 \\
1&10^{-1} &1.05&0.830&0.50 &48.78 &2.97&182.65&1.67 \\
1&10^{-2} &1.09&0.460&0.26 &25.41 &3.17&189.11&1.78 \\
1&10^{-3} &1.10&0.160&0.09 &8.81 &3.23&189.69&1.82 \\
1&10^{-4} &1.10&0.052&0.023 &2.65 &3.24&178.80&1.82 \\
1&{1 \over 2}10^{-4} &1.104&0.037&0.020&1.74&3.24&165.70&1.82 \\
\hline
\end{array}
\]
\end{center}
\caption{The values of $h_t(M_t)$, $h_b(M_t)$, $h_\tau(M_t)$ and
calculated by RGE for $\alpha_s=0.11$ in model VII. The
prediction of the masses $m_b$ and $m_t$
at the scale $M_t$ has been quoted in GeV. $M_t$ is
defined as 170 GeV. $tan \beta$ has been calculated assuming
$m_\tau(M_Z)=1.777$ GeV.} \label{table4}
\end{table}
It is widely known that in SUSY GUTs with one step breaking predict a
large value of $m_b$ for the major part of the parameter space \cite{barg}.
The study of b-$\tau$ unification including a right handed neutrino has also
been performed \cite{fv}. However, in this study no new
gauge interactions beyond the intermediate scale was considered
and due to renormalization effects of the new Yukawa coupling,
a 10-15\% increase in the mass of the b-quark was obtained. However one
sees that the inclusion of the new left-right symmetric gauge and Higgs
interactions at $M_{B-L}$ surviving from a string inspired SO(10) GUT, and
constrained by Eqn.(\ref{cond}), the running of the b-quark Yukawa coupling
is altered and as a result an attractive reconciliation with the
experimental measurements \footnote{ Taking $m_b(m_b)$ in the range
4.1 to 4.5 GeV and $m_\tau(m_\tau)$ to be 1.777 GeV \cite{pdg}, one
typically gets ${m_b(M_t) \over m_\tau(M_t)}$ in the range 1.7 to 1.9 \cite{
barg, fv}.} can be achieved.
\section{Mini doublet-triplet splitting}
In this section we will explore a possible reverse doublet-triplet
splitting \cite{marbr}
which will have an effect opposite to the conventional doublet-triplet
splitting on the prediction of $\alpha_s$. Such a strange reverse
doublet-triplet splitting is indeed possible in a SO(10) model when
there is a mechanism to strongly suppress the Higgsino mediated proton
decay \cite{babr} as will be displayed below.
We consider the prediction of $\alpha_s$
including the threshold effects in SUSY SU(5), which is well-studied
in the literature \cite{barhall,yamada,hisano,mar,bagger}. Throughout
this section we will assume that including the threshold corrections,
the minimal SUSY SU(5) GUT predicts $\alpha_s=0.126$ \cite{bagger}; we
will also assume that the mass of the color triplet Higgs scalars in a
minimal SU(5) GUT is $10^{16.6}$ GeV \cite{mar}. In particular the
prediction of $ \alpha_s$ in the minimal SUSY SU(5) can be written as,
\begin{equation}
\alpha^{-1}_s(M_Z)={1 \over 2} ~[3 \alpha^{-1}_2(M_Z)-\alpha^{-1}_1(M_Z)]
- { 3 \over 5 \pi} \ln[{M_3 \over M_2}] + T_L \label{a3su5},
\end{equation}
Where, $M_3$ and $M_2$ are the masses of the triplet and the doublet
Higgs scalars present in the $5$ and $\overline{5}$ representations of
SU(5), and $T_L$ parametrizes the contribution from all other light
degrees of freedom (excluding the light Higgs doublets) \cite{hisano},
and in a simple step function approximation\footnote{ $M_{SUSY}$ can
be considered in the simplest approach as a common susy breaking
scale, or as an effective susy mass parameter \cite{lang} resuming the
effect of the detailed susy spectrum, and in this sense it can be
either more or less than $M_Z$ depending on the super-partner masses.}
$T_L= {1 \over 2 \pi}
\ln {M_{SUSY} \over M_Z}$. In the minimal model the triplet-mass,
which is bounded from below from the non-observation of proton decay,
remains at the GUT scale. On the contrary the mass of the doublet is
of the order of the electroweak scale. In such a generic situation,
that is, whenever $M_3>M_2$ the doublet-triplet splitting increases
the prediction of $\alpha_s$. However, notice the hypothetical
possibility, that if the mass of the doublet were more than the mass
of the triplet, we would have had a reverse effect on $\alpha_s$.
Keeping this in mind we add one more $5+\overline{5}$ Higgs scalars
with doublet and triplet masses as $M^\prime_2$ and $M^\prime_3$ GeV
respectively. In that case the Eqn.(\ref{a3su5}) gets modified to,
\begin{equation}
{\alpha^\prime}^{-1}_s(M_Z)={1 \over 2} ~[3
\alpha^{-1}_2(M_Z)-\alpha^{-1}_1(M_Z)] - { 3 \over 5
\pi}
\ln[{M_3 M^\prime_3
\over M_2 M^\prime_2}] + T_L \label{a3}.
\end{equation}
Taking the difference of Eqn.(\ref{a3su5}) and Eqn.(\ref{a3}) and
assuming,
\begin{equation}
M_3=10^{16.6}~;~M_2=10^2~;~M^\prime_3=10^x~;~M^\prime_2=10^y,
\end{equation}
we get,
\begin{equation}
\Delta \alpha^{-1}_s={\alpha^\prime}^{-1}_s(M_Z)-\alpha^{-1}_s(M_Z)
={3 \over 5 \pi} (y-x)
\ln 10. \label{diff}
\end{equation}
It is easy to check from Eqn.(\ref{diff}) that taking $y-x=2.26$ we
can get $\Delta \alpha^{-1}_s =0.99$ and consequently $\alpha_s$
decreases by 11\%, from 0.126 to 0.112. Instead if we add n extra
pairs of $5+\overline{5}$ the required splitting in each SU(5)
multiplet is only 2.26/n orders of magnitude.
We can give an example of incorporating this mechanism in a realistic
SUSY SO(10) GUT. The main problem to lower the mass of the color
triplet Higgs Scalars comes out of the stringent experimental upper
bounds imposed on the amplitude of the Higgsino mediated proton decay
diagrams. Babu and Barr \cite{babr} have shown that it is possible to
suppress the Higgsino mediated proton decay strongly in an SO(10)
model by a judicious choice of the fields, couplings and VEVs at the
GUT scale. Consider the SO(10) invariant superpotential
\cite{babr}
\begin{equation}
W= \lambda 10_{1H} 45_H 10_{2H}+\lambda^\prime 10_{2H} 45^\prime_H
10_{3H}+M 10_{3H} 10_{3H}+ \sum^{3}_{i,j=1} f_{ij} 16_i 16_j 10_{1H}.
\end{equation}
If $45$ and $45^\prime$ get VEVs in the directions \cite{dw,babr}
\begin{equation}
\langle 45\rangle = \eta \otimes diag(a,a,a,0,0)~~;~~
\langle 45^\prime \rangle = \eta \otimes diag(0,0,0,b,b)~~;~~
\eta \equiv \pmatrix{0&1 \cr -1&0},
\end{equation}
the super-heavy mass matrices of the doublets and the triplets are of
the form,
\begin{equation}
\pmatrix{\overline{2}_1& \overline{2}_2 & \overline{2}_3}
\pmatrix{0&0&0\cr0&0&\lambda^\prime b\cr0&-\lambda^\prime b &M}
\pmatrix{2_1 \cr 2_2 \cr 2_3}
{}~~{\rm and}~~
\pmatrix{\overline{3}_1& \overline{3}_2 & \overline{3}_3}
\pmatrix{0& \lambda a&0\cr -\lambda a &0& 0 \cr 0& 0 &M }
\pmatrix{3_1 \cr 3_2 \cr 3_3}.
\end{equation}
The absence of any direct coupling between $\overline{3}_1$ and $3_1$
suppresses the Higgsino mediated proton decay. The absolute values of
the masses for the doublets ($M_D$) and the triplets ($M_T$) are given
by\footnote{ When supersymmetry is broken the masses receive
correction of the order of $m_{3/2}$.}, \begin{equation} M_D=(0,{M \pm
\sqrt{M^2-{\lambda^\prime}^2b^2} \over 2})~~;~~ M_T=(\lambda a,\lambda
a,M).
\end{equation}
There can be various choices of parameter space leading to the
required lowering in the prediction of $\alpha_s$ [see
Eqn.(\ref{diff})]. The simplest choice is, $ M^2={\lambda^\prime}^2
b^2. $ In this case the masses are,
\begin{equation}
M_D=(0,M/2,M/2) ~~{\rm and}~~ M_T=(\lambda a, \lambda a, M).
\end{equation}
We notice that, in this model one pair of doublet-triplet is almost
degenerate. Now, using the fact that in minimal SUSY SU(5) the mass
of the color triplet is $10^{16.6}$ GeV and using,
\begin{equation}
\lambda a = 10^x~~,~~M=10^y,
\end{equation}
we get,
\begin{equation}
\Delta \alpha^{-1}_s={3 \over 5 \pi}[(16.6-x)+(y-x)-{\ln 4 \over \ln 10}]
\ln 10 . \label{dsso10}
\end{equation}
We can achieve the desired suppression of $\alpha_s$ if for example
$x=15.2$ and $y=16.6$. We expect that such a splitting is not
difficult to achieve given the number of parameters in the SO(10)
invariant superpotential. The threshold effects due to heavy SO(10) gauge
bosons have been discussed in Ref \cite{marbr}.
\section{Conclusion}
To summarize, we have discussed various ways to reconcile a possible
low value of $\alpha_s$ with the idea that all gauge couplings eventually
unify . They all indicate new physics beyond the canonical minimal
single SUSY GUT scenarios and point towards models with intermediate
scales\cite{we,martin} or new Higgs fields at the GUT scale\cite{marbr}
or perhaps even non-GUT type physics\cite{rs}. Any realistic model
building in the SUSY GUT framework must therefore respect either one
of these scenarios. There are of course other suggestions to change
the predictions of $\alpha_s$ by including gravitational effects at the
GUT scale\cite{utpal,nath} provided these effects are assumed to occur
with an enhanced strength, a possibility which may not be so natural
in many theories (although one cannot be completely sure about
the strength of the gravitational effects at this stage of our knowledge).
\section{Acknowledgements}
B.B would like to thank Mar Bastero-Gil for a number of discussions and
computational help.
\newpage
|
\section*{Acknowledgements}
\noindent We are grateful to Gilberto Colangelo and J\"urg
Gasser for helpful discussions and to Veronique Bernard and Ulf Mei\ss ner
for a useful correspondence. G.E. thanks Pedro Pascual and the Benasque
Center of Physics (Spain) for the kind hospitality during the
final stage of this work. M.M. thanks the members of the
Institut f\"ur Theoretische Physik, Universit\"at Wien for their
warm hospitality.
\newpage
\newcommand{\PL}[3]{{Phys. Lett.} {#1} {(19#2)} {#3}}
\newcommand{\PRL}[3]{{Phys. Rev. Lett.} {#1} {(19#2)} {#3}}
\newcommand{\PR}[3]{{Phys. Rev.} {#1} {(19#2)} {#3}}
\newcommand{\NP}[3]{{Nucl. Phys.} {#1} {(19#2)} {#3}}
|
\section{Introduction}
\section{BACKGROUND}
In 1931
Dirac\cite{PAM}
suggested that the existence of magnetic monopoles can explain the quantization
of electric charge. Dirac showed that a monopole must carry a magnetic
charge which is an integer multiple of $g_D = 68.5$ e in CGS units. After
many unsuccessful searches for monopoles, interest gradually waned until the
Grand Unified Theory (GUT) magnetic monopole was introduced by t'
Hooft\cite{hooft} and independently by Polyakov\cite{poly} in 1974. They
found that there is a solution with magnetic monopoles in any unified gauge
theory which has an unbroken U(1) factor embedded in it. Since the U(1)
factor is essential for describing electromagnetism, virtually all unified
theories predict the existence of magnetic monopoles. The most notable
property of the GUT monopoles is their extreme mass. In all GUTs the
monopole mass ($m_M$) is of the order $\sim 4\pi m_X/e^2$ where {\it X} is
the mass of the superheavy vector boson, which determines the unification
energy scale\cite{four}. For the SU(5) GUT, $m_X$ is $\sim 10^{14}$ GeV
which implies $m_M \approx 10^{16}$ GeV ($\sim0.01~\mu$g). In SO(10),
monopoles with masses $\sim 10^4$ GeV could exist. On the other hand
magnetic monopoles from GUTs with still larger groups, such as Kaluza-Klein
theory or supersymmetry generally have higher mass, $m_M
\mathrel{\rlap {\raise.6ex\hbox{$>$} 10^{19}$ GeV.
The magnetic charge of the GUT monopoles is predicted to be either 1 or 2
Dirac charges ($g_D$).
Because of the extreme mass of the GUT monopoles, a total rethinking of
search strategies was necessary. It is generally agreed that GUT monopoles
can only have been created in the early ($\sim 10^{-35}$ sec) hot Big Bang
universe, possibly abundantly\cite{four}. Thus the only GUT monopoles we can
possibly find are primordial.
Most of the experiments searching for them have been cosmic ray (or
in-flight) experiments\cite{five}. An alternate, possibly more efficient way
is to search for magnetic monopoles trapped in matter. In the Big-Bang
theory the GUT monopoles would be produced when the universe was $\sim
10^{-35}$ seconds old. Hydrogen and helium synthesis began when the universe
was $\sim 1$ second old and ended when $\sim 10^2$ seconds old. Stars and
galaxies are believed to form much later than that, when the universe is
$\mathrel{\rlap {\raise.6ex\hbox{$>$} 10^{12}$ seconds old. Heavy nuclei such as Fe and Ni are produced in
stars by nuclear fusion. Magnetic monopoles, which might have been accreted
in the stars, should mostly have survived the evolution process because of
the exact conservation of magnetic charge. Annihilations are inhibited by
even a small magnetic field and would not be expected to decrease the
monopole density by a large factor. Eventually the monopoles may be recycled
in supernova, along with the heavy nuclei, to form new stars and planets.
Thus most of the magnetic monopoles may remain trapped in stars and other
astronomical bodies, probably near the core. Presently the only macroscopic
samples we have of material from inside such bodies are meteorites and
meteoritic dust.
Monopoles can be bound to bulk ferromagnetic material by an image force. For
iron (magnetite) the binding force near the boundary is $\sim 100$ eV/nm (60
eV/nm)\cite{six}, compared to the gravitational force $\sim 1$ eV/nm on a
monopole with mass $10^{17}$ GeV/c$^2$ near the surface of the Earth. Thus a
monopole with mass $\mathrel{\rlap {\raise.6ex\hbox{$<$} 10^{19}$ GeV/c$^2$ would be bound. However if the
sample containing the monopole is subjected to an acceleration, the monopole
could be dislodged. Meteorites entering the Earth's atmosphere undergo
accelerations of 100--1000 g.\cite{seven} The interiors of meteorites remain
cold during their passage through the atmosphere.\cite{eight} Any monopoles
they contain are likely to remain trapped in ferromagnetic grains in the
interior. Thus the effective upper limit on the
monopole masses that meteorites might contain is $\sim 10^{16}-10^{17}$
GeV/c$^2$. Monopoles also lose $\sim 0.02$ eV/nm in remagnetizing hard
magnetic material such as magnetite or meteoritic iron, so that monopoles
with mass $\mathrel{\rlap {\raise.6ex\hbox{$<$} 10^{14}$ GeV/c$^2$ would only be extracted
by large sustained accelerations.
In general, primordial massive monopoles are unlikely to remain in material
near the Earth's surface, because most of the Earth's material was at one
time at a temperature well above the Curie point. This means that any
primordial massive monopoles are likely to have fallen to the center. If
there are magnetic monopoles with masses $\mathrel{\rlap {\raise.6ex\hbox{$>$} 10^{16}$ GeV/c$^2$ in rocks
near the Earth's surface, they would
have to be captured after the rocks formed and cooled down below the Curie
temperature. However it is easy to show that the chance of a massive
monopole being stopped near the Earth's surface is extremely low because the
range of a monopole is typically much greater than the diameter of the
Earth.\cite{nine}
Most meteors are fragments of small planets or asteroids.\cite{eight} This
means that the meteorites falling on the Earth are samples from cores of
small planets which might contain monopoles. Meteor ages are comparable to
that of the Earth. Most have never been re-heated above the Curie point and
have a high content of iron. This will insure that monopoles will reman
trapped in them via the monopole-image force if they have not been subjected
to overly large accelerations as explained above.
In addition to meteorites themselves, sedimentary rocks and ocean floor
sediments contain non-negligible amounts of meteoritic dust; this consists of
very small grains of meteoritic material constantly falling down on the Earth.
\cite{ten} Typically $\sim 20~\mu$g of meteoritic dust is found in 100 g
of ordinary dust.\cite{eleven} Also ferromanganese nodules found on the deep
sea bed have especially large numbers of meteoritic spherules rich in Fe and
Ni.\cite{twelve} Another possibility is that monopoles could pass through
part of the Sun, losing enough kinetic energy to be captured in solar orbits
and eventually be stopped in the Earth. While it is not easy to estimate the
possible monopole concentration due to these mechanisms, the rocks in the
Earth's crust are also of some interest in trapped monopole searches
considering their availability.
More than 20 years ago, Alvarez, Eberhard and collaborators\cite{thirteen}
tested $\sim 20$ kg of lunar rocks and other materials including $\sim 2$ kg
of meteorites. Another somewhat similar experiment was done by Ebisu and
Watanabe.\cite{fourteen} They tested iron ore from one of the deep mines in
Japan. The iron ore was heated above the Curie temperature\cite{fifthteen}
and they looked for monopoles which fell from the ore.
\section{THE MONOPOLE DETECTOR}
Physically the monopole detector is a cylindrically-shaped vacuum chamber as
shown in Fig.~1. Its length is 1295 mm and its diameter is 457 mm. The
outer wall was made of mild steel to reduce the Earth's magnetic field. A
mu-metal shield was placed just inside the outer wall for the same purpose.
The center hole (warm bore) is 152.4 mm in diameter and served as the test
path. To reduce thermal noise (so-called Johnson noise), this was made of
G-10, an electrically non-conductive glass fiber/epoxy laminate. A
liquid helium dewar, made of stainless steel, was placed concentrically.
Welded stainless steel is an ideal material for constructing cryostats;
however it is electrically conductive and, thus, a source of thermal noise.
In this case the thermal noise was acceptable, since the dewar was at
liquid helium temperature ($\sim 4.2^\circ$ K) and samples with ambiguous
monopole content could easily be passed through repeatedly.
The passage of a magnetic monopole
(trapped in a sample) will cause a jump in the persistent current in the
superconducting coil. This minute change of current in the coil is measured
reliably with a SQUID. For a superconducting coil with {\em N} turns and
inductance {\em L,} the change in current is $\Delta i = 4\pi Ng/L =
2\Delta i_o,$
where $\Delta i_o$ is the current change for a change of one flux quantum of
superconductivity, $\Phi_o = \hbar c/2e = 2.07 \times 10^{-7}$ G-cm$^2$.
Here we assume that the magnetic charge of a monopole is equal to 1 minimal
Dirac charge ($g_D = 2\Phi_o$). A magnetic dipole will cause no net change
of magnetic flux and thus no persistent current in the coil.
To reduce vibration the detector was suspended by 3 nylon lines attached to the
3 ends of a T-shaped aluminum bar. The bar was bolted to the ceiling via a
vibration damper. Natural frequencies of the system were $\sim 1$ Hz
vertically and $\sim 0.5$ Hz horizontally.
Most of the samples we tested had a fair amount of ferromagnetic material,
and thus were strong magnetic dipoles in general. Thus, with some samples
the SQUID output signal was changing even when the samples moved to the
outermost positions. Sometimes these changes were indistinguishable from
that expected from a magnetic monopole. This was mainly due the permanent
dipole moment of the sample, but sometimes the dipole moment induced by the
magnetic field outside the detector was as big as or bigger than the permanent
dipole moment. The dipole signal was reduced significantly by placing the
sample in a tight fitting mu-metal container when necessary.
The expected response of the detector to a magnetic monopole was simulated
with a ``pseudopole", which is a very long solenoid with a diameter of 9.5 mm.
The pseudopole was also used to verify that the SQUID measured the flux
charge properly when it was subjected to a huge transient dipole field.
The signal from the SQUID controller was digitized by an analog-to-digital
converter. Internal feedback was not used in the SQUID controller, so the
electronics did not need to follow the large signals from dipole fields. The
low pass filter in the SQUID controller was set to 1 Hz, and we digitized the
signal at a rate of $\sim 10$ per second. The digitized signal was recorded
for 4 complete (up/down) cycles for each sample. One cycle took $\sim 40$
seconds, or $\sim 160$ seconds to collect one set of data. When the detector
had settled down, about 4 hours after filling with liquid helium, the noise
level was roughly one-third of the smallest Dirac charge (=2$\Phi_0$), as
simulated by the pseudopole.
\section{DISCUSSION AND CONCLUSIONS}
We tested meteorites and other samples, amounting to more than 743 in number
and 331 kg in total mass (Table I). This is about 10 times the total
material tested by Alvarez, Eberhard et al.\cite{thirteen} We did not find a
magnetic monopole.\cite{sixteen} With these results, we can
conclude that the average overall ratio of the monopoles to nucleons in
our sample is $n_M/n_N < 1.2 \times 10^{-29}.$
If $m_M \mathrel{\rlap {\raise.6ex\hbox{$>$} 10^{17}$ GeV, this experiment may not necessarily be a very good
test of the monopole/nucleon ratio of the universe because the trapped
monopoles could have been dislodged by accelerations which occurred during
the meteorites entry into the atmosphere, their impact, or subsequent
handling.
We wish to acknowledge the efforts of James Stone in early phases of the
experiment. H.R. Gustafson provided considerable help and advice. We
especially wish to thank J. Budai, E. Essene, D. Peacor, and Y. Zhang of the
University of Michigan Department of Geological Sciences for loaning us
many samples. We are grateful to the NASA Meteorite Working Group for the
loan of 100 Antarctic meteorite samples. This work was supported by the
National Science Foundation and grants from the Michigan Memorial Phoenix
Project and the Office of the Vice President for Research.
|
\section{Introduction}
Over the recent years, there has been a
growing interest in the origin of
the observed large scale structure of the universe,
due to the data coming from COBE \cite{COBE}
and from the extensive IRAS survey
\cite{largescale}.
One of the most important
conclusions of both measurements
is that the standard cold or hot
dark matter (CDM or HDM) scenarios
with a
Harrison-Zeldovich spectrum of
primeval fluctuations,
fail to account on their own
for the complete spectrum of
energy density fluctuations.
In the case of cold dark matter,
which has been considered as the
most promising solution,
the scale invariance of the spectrum
results in a discrepancy:
either the data fits well with the theory
at large scales and then the
predicted structure at smaller scales is
unacceptably large, or
the data is normalised to agree with the
observations at small scales and
then there is not enough power at
the long wavelength components of
the theoretical
spectrum.
In several particle physics models,
there has been a lot of effort
to provide additional sources
of fluctuations at large scales
\cite{hil}
In one of these attempts,
we have
applied percolation theory in order to
perform a detailed analysis of the
density perturbations that are to be expected
from the domain walls forming
during the phase transition of very light fields
\cite{LLOR}.
This statistical method, which has first been
introduced in the study of domain wall distributions
in the universe by Lalak, Ovrut and Thomas \cite{p2},
allowed us to formulate
a picture of the spatial
distribution of the
overdensities
in a post-inflationary
universe.
We found that
the domain walls
may act as seeds of structure
formation and enhance the
standard cold dark matter spectrum, in
such a way as to account for
the whole range of observations of IRAS and
COBE and still be consistent with the
measurements
of the cosmic microwave background radiation.
This occurs, provided that one of the minima
of the potential of the scalar field is favoured,
and in \cite{LLOR,CLB} it has been demonstrated
why this is true after inflation has taken
place\footnote{
In the case that both minima appear with the same probability,
horizon size domain walls which would result to
unacceptably large
fluctuations in the cosmic microwave background
radiation arise \cite{wal}.}.
Here,
we will extend the analysis
in the small scale distribution of matter.
In these lines, it is of particular interest
to find at what redshift the
first structures are expected to form
and how large is the
amount of mass
that has become non-linear
at that time.
The objects with the higher observed
redshift, $z_{q}$, are quasars.
In the recent years the limits
on $z_{q}$ have increased
and for the more distant quasar that has been
seen up to now,
$z_{q} \sim 5 $ \cite{quas}.
The standard
cold dark matter picture can
account for
early quasar formation
with difficulty and the
situation will become worse if
new, more distant objects are seen at even higher
redshifts. For this reason,
one would like to see whether
domain walls may trigger
sufficiently large density
fluctuations which
lead to quasar
production at early times.
This was found to be the case in \cite{p1},
where models with unstable
domain walls have been considered.
We showed that, due to a small
degeneracy in the minima of the potential
of a pseudo-Goldstone boson, which may arise
in string models, there exists a critical horizon
scale at which the true vacuum
dominates and all walls disappear.
However, when the wall bubbles that
surround a region of true vacuum
expand, rapid collisions
and domain wall annihilation occurs,
resulting in large local overdensities,
which can host quasars. Redshifts as large
as $10$ were predicted, which are much larger
than those expected in CDM models.
In schemes with stable walls, where
the field may roll to the minima of its potential
with different probability,
overdensities at the local level are
also to be expected.
Using percolation theory, it is possible
to examine not only at what
red-shifts the mass in the non-linear
regime becomes sufficient for the
formation of stellar objects, but
also what is their spatial distribution.
The data indicates
that at high and low redshifts
there exists
a decrease in the distribution of
quasars, the peak being at
$z \approx 2$.
In this work, we will attempt to
gain some insight, as to why this
occurs.
\section{Domain walls in the percolation theory picture}
Domain walls are associated
with discrete symmetries,
which arise commonly in many particle
physics models, after the
explicit breaking of a continuous
symmetry \cite{hilross}.
The resulting potential
of the pseudo-Goldstone bosons is of the form
\begin{equation}
V = V_{0} \left[ \cos \left(\frac{\phi}{\upsilon}\right) +1
\right]
\label{e1}
\end{equation}
and obeys the discrete symmetry
$\phi \rightarrow
\phi + 2 \pi n \upsilon$, $n = 1,2 ...$
The equation of motion corresponding to
the above potential, admits domain wall
solutions that interpolate between
two adjacent vacua \cite{walls}.
The width of the walls,
$\Delta$, is given by
\begin{equation}
\Delta = \frac{\upsilon}{\sqrt{V_{0}}} =
m^{-1}
\label{eq:e16}
\end{equation}
$m$ being the mass of
$\phi$ evaluated at any minimum.
The surface energy density of
the wall is
\begin{equation}
\sigma = \int_{-\infty}^{+\infty}
2 V_{0} \left[ \cos \left(\frac{\phi}
{\upsilon}\right) +1 \right] \, dz = 8
\, \upsilon^2 m
\label{eq:e17}
\end{equation}
The space distribution of
the domain walls is found by
partitioning the three-dimensional
space into cubic lattice sites with
lattice spacing $\Lambda$ \cite{p2}.
In this letter, as in \cite{LLOR}, we
work with a system that has two minima.
It is assumed that at each lattice site
the physical system can be in one of the two vacua,
denoted by $(+)$ and $(-)$ respectively.
The probability that
a lattice site is in the $(+)$ vacuum
is denoted by $p$, $0 \leq p \leq 1$,
while the probability that a lattice site
is in the $(-)$ vacuum is $q=1-p$.
Provided there is no correlation between the
vacuum structures at any two different lattice
sites, it is possible to calculate
the spatial distribution of the two vacua
and, hence, the spatial distribution of
domain walls, by applying three-dimensional
percolation theory \cite{3dp}.
It has been shown that for
$p < p_{c}$, where $p_{c}=0.311$ is the
critical probability for a cubic
lattice in three dimensions,
the $(-)$ vacua lie predominantly
in a large percolating cluster (since necessarily $q>p_c$),
while the $(+)$
vacua are in finite
s-clusters.
Here, $s$ denotes the
number of nearest neighbour lattice sites that
are occupied by $(+)$.
Moreover,
it was recently found that the scaling
behaviour of the percolating cluster is
not persistent until $p$ reaches the value
$p=1/2$ \cite{CLB}.
This indicates that for a large range of
probabilities only finite wall bubbles are present.
On a given lattice, the number of s-clusters
falls rather quickly with growing s.
Indeed,
the probability per lattice site that a given lattice
site is an element of an s-cluster,
$n_{s}(p)$, (which is
given by the ratio of the
total number of s-clusters, $N_{s}$, over the total number
of lattice sites, $N$) is
\begin{equation}
n_{s}(p) = 0.0501 s^{-\tau}\exp \left \{-0.6299
\left(\frac{p-p_{c}}{p_{c}}\right) s^{\sigma} \left[(\frac{p-p_{c}}{p})
s^{\sigma}+ 1.6679\right]\right\}
\label{ns}
\end{equation}
where $\tau = 2.17$ and $\sigma = 0.48$ \cite{p2}.
Since for a given lattice
there exists an upper
statistical cut-off on the size of observable clusters,
no unacceptable fluctuations of
the cosmic microwave background radiation
due to domain wall bubbles are generated
\cite{LLOR}.
The mean radius of a wall bubble
at a specific redshift
is well characterized by the average radius
of gyration, $R_{s}(p)$, of an s-cluster.
This quantity
(for $p < p_{c}$ and $s > s_{\xi}$)
is found to be
\begin{equation}
R_{s}(p) = f_{s}(p) \Lambda \equiv
0.702 (p_{c}-p)^{0.322} s^{0.55} \Lambda
\label{Rs}
\end{equation}
where
\begin{equation}
s_{\xi} = \left(\frac{0.311}{|p-0.311|}\right)^{2.08}
\label{sz}
\end{equation}
Initially, $R_{s}$ is larger than the
horizon.
However, the horizon radius grows
faster than that of the
bubble (whose radius just
grows linearly with the expansion),
and thus at some
redshift $z_{a}(s)$ the bubble comes
within the horizon.
At this stage the bubble
shrinks under its surface tension,
undergoes a few cycles of oscillations,
and finally
looses its energy in the form of scalar waves
\cite{WWW}.
The energy stored in the domain wall is
\begin{equation}
E_{s} = f t_{s} \sigma \Lambda^{2}, \; \; \; t_{s} =
s \left ( \frac{1-p}{p} \right )
\label{col
\end{equation}
where $t_{s}$ is the average surface area of an
s-cluster.
The parameter $f$ is $1 \leq f \leq 6$ and
here we use the moderate value $f=3$.
Due to the expansion of the
universe, the lattice spacing $\Lambda$ is a function of
the redshift $z$.
The redshift
when the wall bubbles shrink, $z_{a}(s)$,
is taken to be the redshift
when an s-cluster enters
the horizon. The overdensities that are
expected to arise after the bubbles shrink are calculated in a
subsequent section. However, even at this stage, it is
possible to predict the qualitative picture that arises
and to see its relevance for quasar formation.
\section{Quasar production and spectrum}
Quasars are high-redshift active galaxies,
with a very energetic central source of energy,
which may not be coming from nuclear fusion.
The most popular scheme is that quasars are powered
by the accretion of matter in a supermassive
black hole ($M_{h} \approx 10^{8} M_{\odot}$)
in the center of a host galaxy
\cite{bookqua}. Then, before any quasar activity can
begin, some galaxies must have formed and
virialised at redshifts higher than that of the
actual quasar and subsequently evolved
to the stage of developing a
massive black hole.
This indicates that the very existence of
quasars implies that non-linear structures
must have appeared at redshifts higher than
$5$.
Studies of the space distribution of quasars show
that their number density
exhibits a peak at a redshift
$z \approx 2$; for smaller redshifts the observed
abundance of
quasars decreases.
The space density of quasars with redshift $z < 2$
has been measured to be roughly
$10^{-5} \; {\rm Mpc}^{-3}$.
For redshifts $z > 2$ there is
also a decrease of the number of quasars as the redshift
goes up,
except for the very bright ones. This decrease however
is gradual, rather than a steep cut-off.
It is also possible to estimate the mass
that ought to collapse at a redshift $z_{q}$, in order
to lead to quasar production at a later time.
For example, for a quasar with a
redshift $4$
it is found that the minimal collapsing mass should be
at least $O(10^{12} M_{\odot})$,
and this value can be even larger.
However, such a picture for the local
fluctuations is exactly
what we would expect from percolation theory.
The basic points to note are that:
$\bullet$
the larger domain wall bubbles
enter the horizon and shrink under
surface tension later than the smaller ones.
Since
$\bullet$
wall-driven fluctuations redshift
slower than radiation or matter,
$\bullet$
the local fluctuations that involve larger
bubbles may become non-linear at higher redshifts.
On the other hand,
percolation theory predicts that
$\bullet$
the number of lattice sites
$n_{s}$ decreases exponentially with $s$,
thus the number density of bubbles decreases
with their size.
The combined effect is that
{\em non-linear fluctuations
at very high redshifts
become rare}.
As for the decrease in the
number of quasars as the redshift
drops below $2$ it can also be explained.
This is because
$\bullet$
For larger wall bubbles, the amount of
mass in the non-linear regime is
larger. It is possible therefore that
at $z = 2$ we approach a critical mass scale which is
slightly higher from the minimum value that we need in order for
quasar formation to proceed.
Then, at that redshift the quasar
density will be expected to be maximal, while
at lower redshifts the number of objects
will decrease.
$\bullet$
Finally, for the brighter quasars
more mass is required
in the non-linear regime,
therefore they will appear only at high redshifts.
Note that these features are generic and lead
to the same qualitative picture for
different regions of the parameter space and in particular for
different choices of the mass of the field, $m$, which
sets the time-scale of the transition\footnote{
Only when the length scale $m^{-1}$ comes within the horizon,
can the field roll towards its minima \cite{LLOR}
.}.
\section{Local density fluctuations}
While the qualitative behaviour that has been discussed is generic,
in order to gain a better understanding, we will calculate the local
density fluctuations in a specific scheme. Here, we will chose to
work with the same parameter space which in \cite{LLOR} was found
to lead to the best agreement with the data at large scales.
In \cite{LLOR}, in order to compare the
wall driven fluctuations with the ones observed at the
large scales, the energy density
$E_{s}$ had been averaged over the
mean distance between s-clusters
at $z_{a}$.
The Fourier analysis of a quasi-periodic matter distribution
shows that the amplitude of the
Fourier coefficients is peaked in the
momentum space around a set of discrete
points corresponding to the wavelengths
$\lambda =\infty,d,d/2,...,R$,
where $d$ is the mean distance between seeds
and $R$ is the typical radius of the overdensity
produced by the accretion of matter onto the seed
in question. In the scenario described in \cite{LLOR}
it has been assumed that $R \approx d$
(that is a big amount of the
overdensity detached by
each cluster is dispersed over a
region not much smaller than $d$).
This is a reasonable assumption as the seeds
considered in \cite{LLOR} are produced
in the radiation dominated epoch,
and one can argue that the accretion is not very effective at that time.
However, we do not know how large $R$ really is.
One may expect that the larger clusters, which {\em enter the
horizon at a much later stage}, may
drive collapse more efficiently and leave behind more compact
overdensities whose radii are smaller than $d$ and as small as $R_s$.
We think that this is a reasonable assumption,
for the following reasons:
$\bullet$
Although the phase transition occurs deep in the radiation
dominance, the larger clusters enter
the horizon at the very end of
the radiation era. At that time matter has already started
to slow down significantly, therefore part of it may
accrete around the wall just before the latter collapses.
In this case we can define an ``efficiency parameter'',
$\gamma_{\ell}$,
which scales with the redshift. This parameter should become
unity in matter dominance (where subsequently
the fluctuations grow proportionally to the redshift),
while the earlier the wall bubble enters the horizon,
as compared to the beginning of the matter dominance
era, the smaller the efficiency parameter becomes.
We come back to this point in the quantitative examples
that will be presented.
$\bullet$
The larger walls store more energy, therefore they
drive more efficient collapse.
$\bullet$
In addition, although in \cite{LLOR} we have
prefered to work with non-interacting
light scalar fields, as they are out of equilibrium and
their mass is naturally in the correct range for
structure formation, it is also
possible that some {\em interaction} between the
light fields and ordinary matter is present.
In \cite{mh} it has been shown that very light pseudo-Goldstone
bosons may give rise to long range forces.
In the case that an interaction exists, localized density
fluctuations due to domain walls may appear
\cite{masarotti}. If this interaction is very weak,
it still will not be sufficient
to change the out-of-equilibrium property of the system,
which has the general behaviour that we have described
in \cite{LLOR}.
Nevertheless, even a very soft interaction can result in
amplifying the local energy dissipation
mechanism in comparison to the situation where only gravity is present
in the theory. As we are going to see below, an
efficiency for matter accretion as low as $ \approx
10\%$, around collapsing
domain walls which enter the horizon at the end of the radiation era,
is sufficient to support the picture we propose.
$\bullet$
Finally, while here we chose to work with the parameter space that
was found in \cite{LLOR} to be optimal for the creation of
large scale structure due to a single phase transition, it
would be possible to relax this condition. In such a case, we
can assume that the mass of the pseudo-Goldstone boson, $m$,
can be smaller, such that the larger walls which
give rise to the local density fluctuations that subsequently
will host quasars, enter the horizon and dissipate their energy
in the matter dominance era.
In \cite{LLOR} $m$ was fixed
by demanding that the peak of the density fluctuations as
a function of the scale occurs at $\approx 30$ Mpc, as
observations indicate. What happens when $m$ is smaller?
The first point to make is
that the density fluctuations at large scales will
decrease, since the fluctuations now have less time to grow
(here we should note that for large scale structure, the relevant
fluctuations are of super-horizon size and grow
as the square of the red-shift during the radiation era
\cite{KT}, while the local fluctuations practically grow only
in matter dominance).
The larger domain walls may now enter
the horizon deep inside
matter dominance and the resulting local
density fluctuations
are amplified. As we have pointed out,
the basic features which determine the quasar
distribution {\em are generic} and give rise to a similar
qualitative behaviour if a smaller $m$ is chosen.
The picture we have therefore
is that overdensities
at scales larger than the respective $R_s$,
as well as the local overdensities which do not become nonlinear
early enough, form a kind of a diffusive
background on top of which some
overdensities at the local level
may form gravitationally
bound structures, for instance galaxies
hosting quasars.
In what follows we will try
to estimate the expected spectrum of such structures.
For this, we need to calculate the overdensity at a local
level and the scale over which the averaging is done is
the one that may provide mass greater than
this of a quasar,
in the non-linear region $\delta \rho/\rho \geq 1$.
The smallest possible distance
over which we may average
(leading to the larger local
fluctuations) is the horizon
at a redshift $z_{a}$,
$R_{H_{a}} \equiv
R_{H}(z_{a})$.
Let us now pass to specific formulas:
The redshift $z_{a}$ is obtained
by equating the mean radius of gyration
for an s-cluster,
to the horizon at that redshift.
We find that
\begin{equation}
1+z_{a} = \frac{1+z_t}{\alpha f_{s}(p)}
\ee
with
\begin{equation}
(1+z_t)^2 = \frac{R_{{H}_0}} {R_{H}(z_t)}
(1+z_{d})^{1/2}
\ee
where $z_d$ is the redshift when
matter domination begins
and $R_{H_{0}}= 6000$ {\rm Mpc}
is the horizon size today\footnote
{Throughout the calculation we are going to
assume that the reduced
Hubble constant $h$ is unity (that is the
Hubble constant today is
$H_{0} = 100 \,{\rm km} \,
{\rm s}^{-1} \,
{\rm Mpc}^{-1}$), for
simplicity of presentation. A different value of
$h$ does not alter the picture we have. In this case,
the input model parameters that are needed to
fit the large scale data, which are
the same that we use here for small scale
predictions, are shifted to
$\sigma h^2$ and $\upsilon h$,
as explained in \cite{LLOR}
.}. The factor $\alpha \equiv
H(z_t)/H(z_f)$, where
$z_t$ is the redshift where the field
starts rolling down the potential towards
one of its minima, and $z_f < z_t$ denotes the time
at which the system actually settles in one of the vacua,
after a period of oscillations \cite{LLOR}.
The above formulas hold for
$z_{a} \geq z_{d}$.
Assuming
that the mean total energy density
is equal to the critical
density, we may express the
critical energy density at $z_{a}$ in
terms of the present day critical density $\rho_0$
as
\begin{equation}
\rho_c (z_{a}) = \rho_0 \, \frac{(1+z_{a})^4}{1+z_{d}}
\ee
The local energy density perturbation due to an
s-cluster with diameter $R_{H_a}$ at $z_{a}(s)$ is
\begin{equation}
\left. \frac{\delta \rho}{\rho}
\right|_{a} \equiv
\left. \frac{\delta \rho}{\rho}
\right|_{\ell oca\ell} (z_{a})
= \frac{ 6 \, f \, \sigma \,
(1-p) \, s \, \Lambda_{a}^{2}}
{p \, \rho_{c}(z_{a}) \,
\pi \, R_{H_{a}}^{3}}
\label{loc}
\end{equation}
where
\begin{equation}
\Lambda_{a} \equiv
\Lambda(z_{a}) = \frac{\alpha}{m} \,
\frac{1+z_{t}}{1+z_{a}}
\label{loc1}
\end{equation}
is the lattice spacing at $z_{a}$\footnote{ $\Lambda(z_t)
\equiv a/m$ \cite{p2,LLOR}.}
and
\begin{equation}
R_{H_{a}} = \frac{1}{m} \, \left
(\frac{1+z_{t}}{1+z_{a}} \right)^{2}
\label{loc2}
\end{equation}
The local
fluctuations are
(in contrast to the fluctuations that give rise
to the large scale structure)
always sub-horizon and therefore grow
logarithmically with the redshift during radiation
dominance and linearly during matter dominance.
Then the redshift $z_{q}$ at which the
fluctuations become
non-linear is given by
\begin{equation}
1+z_{q} = \gamma_{\ell} \;
\left.
\frac{\delta \rho}{\rho}
\right|_{a}
(1+z_{d}) \left(
1 + 2 \log \frac{1+z_{a}}{1+z_{d}}
\right)
\label{loc3}
\end{equation}
The factor $\gamma_{\ell} < 1$ which
appears in (\ref{loc}), has been added
in order to take into account that for the parameter
space where we work, at
the end of the radiation dominance
only a part of the overdensities
that are produced by the wall bags
will remain localized.
The total amount of mass
in the non-linear region at $z_{q}$,
$M_{q}$, in terms of solar masses $M_{\odot}$,
is given by
\begin{equation}
M_{q} = \frac{ \pi L_{q}^{3}}{6 M_{\odot}}
\rho_{c}(1+z_q)^{3}
\label{loc4}
\end{equation}
where the scale of the perturbation
at $z_{q}$ is
\begin{equation}
L_{q} = R_{H_{a}} \, \frac{1+z_{a}}{1+z_{q}}
\label{loc5}
\end{equation}
and $M_{\odot}$
is the solar mass.
Finally, we can identify the space distribution of
the local fluctuations.
The average distance between s-clusters at a
redshift $z$ is
\begin{equation}
d(z) = \left(\frac{V(z)}{V(z)n_{s}}\right)^{1/3}
\Lambda(z)
\label{loc6}
\end{equation}
thus today
\begin{equation}
d(z=0) = \frac{1}{n_{s}^{1/3}} \frac{a}{m}
(1+z_{t})
\label{loc7}
\end{equation}
\section{Numerical analysis}
In \cite{LLOR} we have found that
large scale structure may form
as a result of the global density fluctuations
(fluctuations averaged on scales $d$),
and some indicative
combinations for $\alpha =10$
appear on Table 1.
Here we want to use the same set of parameters
to examine the local overdensities of the model.
However, in \cite{LLOR}
we had also introduced a parameter
$\gamma_{s}$ (to account for the
fact that before a wall bag disappears,
it may stay around sufficiently long
to cause the collapse of amounts of matter).
This coefficient may not be determined
precisely without a more detailed analysis
in the framework of the spherical collapse model.
In \cite{LLOR} we took the value $\gamma_{s} =10$, however
if this parameter is of order unity,
the only modification in our
results would be that we need a higher value of
$\upsilon$ to fit the IRAS and COBE data. The
model parameters that lead to solutions,
for $\gamma_{s} = 1$,
are given in Table 2.
In the present work
we take $\gamma_{s}$ at its minimum value and
on top of that we have introduced
$\gamma_{\ell}$, to account for the fact that
the wall bubbles that we consider appear
in the end of the radiation dominance (if $\gamma_{s} > 1$, even larger
efficiency in
the accretion of the local overdensities would
be expected).
To see whether it is possible to
get the correct qualitative behaviour
for the distribution of local overdensities,
we consider the following possibilities:
(i) $\gamma_{\ell}$ $O(0.2)$.
Such a constant factor (especially
shifted towards larger values)
may be expected if some
soft interactions are present.
(ii) $\gamma_{\ell}$ $O(0.1)$.
(iii) In the absence of interactions, the
most realistic approach is
to take into account
that larger bubbles
lead to a more efficient energy dissipation,
since they enter the horizon nearer the matter dominance era,
where the overdensities grow linearly with the
redshift.
For this reason we set
$\gamma_{\ell} = (1+z_d)/(1+z_a)$. For
$z_{a} \approx z_{d}$ the efficiency parameter
is unity, since we are in matter dominance,
while the higher $z_{a}$ is, the smaller the
parameter becomes.
Using the model parameters of Table 2, we
have calculated the local density fluctuations
as well as their space distribution,
the redshift $z_{q}$ where the
fluctuations become non-linear and
the amount of mass
that is involved in the non-linear regime.
The results appear on
Tables 3,4 and 5
for the three choices of
input parameters respectively.
These tables indicate that
for all three choices,
the amount of mass in the non-linear region
can be sufficiently large to allow for
early quasar formation.
Moreover, we reproduce
qualitatively the observed
space distribution
of quasars, at redshifts $z \geq 2$,
that is quasars at
larger redshifts appear with
larger space separation.
We also observe that the mass in the non-linear
regime reduces with the red-shift, indicating that
after a certain redshift the total available amount of
mass will be near the lower limit that we
need for quasar production. We find that for this
to occur at $z =2$, the mass should be
$O(10^{13} \; M_{\odot})$.
Concerning the scale of the perturbation,
in all cases is found to be $O({\rm Mpc})$.
We also see that the number of quasars at a specific redshift
is sensitive to the parameter $p$.
Indeed, for p = 0.11 we find that $s = 50$
leads to one quasar every 1614 Mpc, while
for p =0.15 the distance is 461 Mpc.
The number of quasars and the total mass
in the non-linear regime, as functions of the red-shift,
for the three cases of Table 2,
are given in figures 1-6.
We see that the qualitative behaviour is in agreement
with observations and that even with the
rough approximations that we have made,
the quantitative agreement is also good.
\section{Conclusions}
To summarise,
we have looked at the local density fluctuations
generated by domain walls
after the phase transition of light pseudo-Goldstone
bosons. In particular, we have analyzed the
expected density perturbations and
their spatial distribution, as well as
the redshifts at which they become non-linear.
We have found that, complementary to the generation of the
observed large scale structure,
the same overdensities
may lead at the local level to an early
appearance of non-linear
fluctuations which may result to
early quasar production.
The scale of the overdensities
is naturally of the correct order of magnitude.
Concerning the spectrum of these objects, we
show that
quasars are expected to
appear with larger space separation
as the redshift increases, in consistency with
observations. A decrease to the number of objects as
the redshift falls below a critical value is also
predicted.
The total amount of mass that is involved in
this non-linear process is
from $10^{12}-10^{14} M_{\odot}$, which is interesting, given
that $10^{12} M_{\odot}$
is the minimal possible value for
an overdensity to evolve to a galaxy that may
host a quasar.
\vspace{0.2 cm}
\noindent
{\bf Acknowledgment }
\noindent
I am grateful to Z. Lalak,
B. Ovrut and G. G. Ross for many
enlightening discussions
on the subject.
|
\section*{}
In the early days of hadron physics, the contours of a satisfying dynamic
theory were far from obvious.
The analysis of resonance scattering had to be performed
with a minimum of theoretical assumptions.
Basics of the S-matrix theory which was developed in this context
and its application to the description
of resonances may be found e.g. in \cite{eden,sakurai,landau}.
The present understanding of the gauge theory of electroweak
interactions \cite{gws} allows for
detailed and precise theoretical predictions of electroweak scattering
including the precise measurements of $Z\:$ boson interactions at LEP100.
Nevertheless, it is of some interest
to perform also model-independent fits to the $Z\:$ line shape
\cite{usual,borrelli}.
In this respect, we consider the S-matrix theory to be the most consequent
approach. An introduction to the necessary formalism
and its application is the subject of the present article.
\vspace{.4cm}
The annihilation of electrons and positrons into
lepton pairs or hadrons at LEP100,
\begin{equation}
e^+e^- \longrightarrow (\gamma, Z) \longrightarrow f^+f^-(\gamma),
\label{ee}
\end{equation}
is used to determine mass $M_Z$ and width $\Gamma_Z$ of the $Z\:$ boson. These
observable
quantities correspond uniquely to the location of a pole of the S-matrix
describing
(\ref{ee}) in the complex energy plane:
\begin{equation}
{\bf \cal M}(s) = \frac {R_{\gamma}}{s} + \frac {R_{Z}}{s-s_{Z}} + F(s).
\label{smatrix}
\end{equation}
The poles of ${\bf \cal M}$ have complex residua
$R_{Z}$ and $R_{\gamma}$, the latter corresponding to the photon,
and $F(s)$ is an analytic function without poles. Further,
\begin{equation}
s_Z = M_Z^2 - i M_Z \Gamma_Z.
\label{s0}
\end{equation}
The analysis of the $Z\:$ line shape
will be based here on the cross section
\begin{equation}
\sigma(s) = \sum_{i=1}^4 \sigma^i(s) = \frac{1}{4} \sum_{i=1}^4
s |{\bf \cal M}^i (s)|^2,
\label{sigma}
\end{equation}
where the sum must be performed over four helicity amplitudes with different
residua
$R_Z^i$ and functions $F^i(s)$
\footnote{An application of the
S-matrix formalism to $e^+e^-$-annihilation has been proposed also in
\cite{stuart}. It is not pointed out there that one has to rely on helicity
amplitudes and a simple-minded application of the formulae
discussed there would fail.}.
Although we will not perform a field theoretic interpretation here,
for the reader's convenience the Born predictions of $R_{\gamma}$ and
$R_Z$ in terms of
vector and axial vector couplings are shown:
\begin{equation}
R_{\gamma}^B = \sqrt{\frac{4\pi}{3} c_f (1+\frac{\alpha_s}{\pi}) }
Q_e Q_f \alpha(s),
\label{rgamma}
\end{equation}
\begin{eqnarray}
R_Z^{0,B}=R_Z(e_L^{-}e_R^{+}\rightarrow f_L^{-}f_R^{+})=c(v_e+a_e)(v_f+a_f),
\nonumber \\
R_Z^{1,B}=R_Z(e_L^{-}e_R^{+}\rightarrow f_R^{-}f_L^{+})=c(v_e+a_e)(v_f-a_f),
\nonumber \\
R_Z^{2,B}=R_Z(e_R^{-}e_L^{+}\rightarrow f_R^{-}f_L^{+})=c(v_e-a_e)(v_f-a_f),
\nonumber \\
R_Z^{3,B}=R_Z(e_R^{-}e_L^{+}\rightarrow f_L^{-}f_R^{+})=c(v_e-a_e)(v_f+a_f).
\label{helirz}
\end{eqnarray}
In the standard theory,
\begin{equation}
c = \sqrt{\frac{4\pi}{3}c_f(1+\frac{\alpha_s}{\pi})}
\frac{G_{\mu}}{\sqrt2} \frac{M_Z^2}{2\pi},
\hspace{1cm} a_f=\pm\frac{1}{2}, \hspace{1cm}
v_f=a_f(1-4|Q_f|\sin^2\vartheta_W),
\label{const}
\end{equation}
where $c_f$ is a possible color factor in case of hadron production.
The corresponding functions $F^{i}(s)$ vanish in Born approx\-imation,
$F^{i,B}(s)=0$.
In general, the $F^{i}(s)$ contain non-resonating radiative corrections.
More details on the realization of ansatz (\ref{smatrix}) in the standard
theory may be found in \cite{stuart}.
Instead referring to field theory, we parametrize the
cross section (\ref{sigma}) as follows:
\begin{equation}
\sigma(s) = \sum_{A} \sigma_A(s), \hspace{1cm} A =
Z, \gamma, F, \gamma Z,ZF, F\gamma,
\label{sign}
\end{equation}
with the contributions:
\begin{equation}
\begin{array}{rlrl}
\sigma_Z(s) = &
\displaystyle{\frac{s r_Z}{|s-s_Z|^2} },
&r_Z = & \frac{1}{4} \sum |R_Z^i|^2,
\vspace{.4cm} \nonumber \\
\sigma_{\gamma}(s) = &
\displaystyle{\frac{r_\gamma}{s}} ,
&r_{\gamma} = & |R_{\gamma}|^2,
\vspace{.4cm} \nonumber \\
\sigma_F(s) = & s r_F(s), &
r_F(s) = & \frac{1}{4} \sum |F^i(s)|^2,
\vspace{.4cm} \nonumber \\
\sigma_{\gamma Z}(s)=& 2 {\rm Re}
\displaystyle{\frac {C_{\gamma}^{\ast} C_Z} {s-s_Z}}, &
C_{\gamma} = & R_{\gamma}, \hspace{.5cm} C_Z = \frac{1}{4} \sum R_Z^i,
\vspace{.4cm} \nonumber \\
\sigma_{ZF}(s) = & 2 {\rm Re}
\displaystyle{\frac {s C_{ZF}(s)} {s-s_Z}}, &
C_{ZF}(s) = &\frac{1}{4} \sum R_Z^i F^{i\ast}(s),
\vspace{.4cm} \nonumber \\
\sigma_{F\gamma}(s) =& 2 {\rm Re} \left[ C_{\gamma}^{\ast} C_F(s)\right],&
C_F(s) = &\frac{1}{4} \sum F^i(s).
\end{array}
\label{sigA}
\end{equation}
After making denominators real one remains with the following formula
for the line shape:
\begin{equation}
\displaystyle{\sigma (s) = \frac{R + (s - M_Z^2) I }{|s-s_Z|^2}
+ \frac{r_{\gamma}}{s} + r_0 + (s-M_Z^2) r_1 + \ldots}
\label{sigfin}
\end{equation}
Besides $M_Z, \Gamma_Z$, the real constants $R, I, r_0$ and $r_1$ are
introduced:
\begin{eqnarray}
R =& M_Z^2 \left[ r_Z
+ 2 (\Gamma_Z / M_Z ) \left( \Im m C_R + M_Z \Gamma_Z \Re e (C'_R) \right)
\right],
\nonumber\\
I =& r_Z + 2 \Re e C_R,
\nonumber\\
C_R(s) =& C_\gamma^{\ast} C_Z + s_Z C_{ZF}(s), \nonumber\\
r_0 =& M_Z^2 \left[ r_F -M_Z \Gamma_Z \Im m (r'_F) \right] +
\Re e C_r - M_Z \Gamma_Z \Im m C'_r, \nonumber \\
r_1 =& r_F + M_Z^2 \left[ \Re e (r'_F) - (\Gamma_Z / M_Z ) \Im m (r'_F) \right]
+ \Re e C'_r, \nonumber \\
C_r(s) =& C_{\gamma}^{\ast} C_F(s) + C_{ZF}(s).
\label{fitconst}
\end{eqnarray}
The energy-dependent functions $C_{ZF}, C_F, r_F$,
and their (primed) derivatives with
respect to s have to be taken at $s=s_Z$. As may be seen, the cross section
may be described by only six real parameters as long as one takes into account
only the first two terms in the expansion of the functions $F^i(s)$ around
$s=s_Z$ and at most terms of the order $(s-M_Z^2)^n, n=0,1$ in the cross
section parametrization.
Next we have to discuss a conceptual problem due to QED bremsstrahlung.
Initial state radiation of photons leads to a deviation of the
{\it effective}
energy variable $s'$ in (\ref{sigma})
from $s$ with $s'<s$. At least events with soft photon
emission unavoidably become part of the measured cross section. Taking them
into account means summing up an infinitely dense chain of single poles
leading to a new singularity structure (including a cut in the
complex plane) compared to the original
ansatz (\ref{sigma}). Indeed, the well-known formulae for initial state
radiation in reaction (\ref{ee}) (see e.g. \cite{usual,h0t} and refs. therein)
contain in one way or the other the complex logarithm
\begin{equation}
L(s_Z) = \ln \frac{1-\Delta-s_Z/s}{1-s_Z/s},
\label{log}
\end{equation}
where $0 < \Delta < 1$ stands for some cut on the allowed photon energies.
A function like (\ref{log}) with its highly singular behaviour at $s=s_Z$
cannot be absorbed into the function $F(s)$ as introduced in (\ref{smatrix}).
The QED bremsstrahlung must be treated as follows. Initial state radiation
has
to be taken into account as exactly as possible,
e.g. using a convolution formula \cite{usual},
\begin{equation}
\sigma_T(s) = \int ds' \sigma (s') \rho_{\rm ini}(1-s'/s).
\label{convolution}
\end{equation}
Final state radiation can be
either calculated similarly or
formally simply
neglected. It doesn't influence the singularity structure of the
cross section and leads to some modifications of parameters other
than
$s_Z$ in (\ref{smatrix})
\footnote{If one wants to interpret the residua $R_{\gamma}$ and $R_Z$,
and $F(s)$ in terms of a field
theory, one should of course make an explicit calculation of final state
radiation.}.
The radiation connected with initial-final state
interferences can be taken into account by an analogue formula to
(\ref{convolution}) with
a slightly more complicated structure \cite{h0t,intconv}:
\begin{equation}
\sigma_{\rm int}(s) = \int ds' \sigma(s,s') \rho_{\rm int}(1-s'/s).
\label{convigif}
\end{equation}
The correct ansatz for the S-matrix based cross section is:
\begin{equation}
\sigma(s,s') = \frac{1}{8} s' \sum_i
\left[ {\bf \cal M}_i(s) {\bf \cal M}_i^{\ast}(s')
+ {\bf \cal M}_i^{\ast}(s){\bf \cal M}_i(s') \right].
\label{sigif}
\end{equation}
We only mention that a representation like (\ref{sigA})
may be obtained easily also for $\sigma(s,s^{\prime})$.
If necessary, cross section (\ref{convigif}) may be added to
(\ref{convolution}).
Its numerical contribution is very small at LEP100 energies under usually
applied cut conditions.
\\
\vspace{.5cm}
Using (\ref{smatrix}) - (\ref{sigma}) for a fit to data, one is free of any
model-dependent
assumption, or
some choice of gauge, or a truncation of perturbation theory as must be usually
taken into account (see e.g. \cite{usual} and the recent discussion in
\cite{stuart,sirlin,spanier}).
The actual
configuration of cuts applied to the data is as unimportant as are details
of the final state. In case of a differential cross section, the S-matrix
would depend on additional variables.
\\
\vspace{.5cm}
In order to demonstrate that the S-matrix approach may have some practical
relevance, we use a simple code \cite{zpole} for the calculation of the QED
corrections
(\ref{convolution})
including soft photon exponentiation.
We perform four fits with a rising number of degrees of freedom to published
data at seven different beam energies $E$, $s=4E^2$, taken from
an analysis of the hadronic line
shape (Table 2 of \cite{l3}).
Without loss of generality, one can assume that the behaviour of the
running coupling constant of the photon $\alpha(s)$ is known
at LEP100 energies \cite{alfrun}:
\begin{equation}
\alpha \left( s=(91.2 {\rm GeV})^2 \right) = \alpha_0(1.0660-i0.0189).
\label{alfalep}
\end{equation}
For comparison, we give also the field theoretic Born estimates
for hadron production:
\begin{eqnarray}
R_{\gamma}^{B,h} &=& r^{1/2}
\sqrt{ \frac{4\pi}{3} c_f (1+\frac{\alpha_s}{\pi})}
\alpha(s),
\nonumber \\
r_Z^{B,h} &=& c^2 (v_e^2+a_e^2)
[3(v_d^2+a_d^2)+2(v_u^2+a_u^2)] \sim 6.32\ 10^{-4},
\nonumber \\
C_Z^{B,h} &=& \frac{c}{r^{1/2}} Q_e v_e [3 Q_d v_d + 2 Q_u v_u ]
\sim 7.77 \ 10^{-4},
\nonumber \\
r &=& 3 Q_d^2 + 2 Q_u^2,
\nonumber \\
C_{ZF}^{B,h} &=& C_F^{B,h} = r_F^{B,h} = 0.
\label{prediction}
\end{eqnarray}
Further,
\begin{eqnarray}
R^{B,h} =& M_Z^2 r_Z^{B,h} &= 5.45 \: {\rm GeV^2},
\nonumber \\
I^{B,h} =& r_Z^{B,h} + 2 (\Re e C_{\gamma}^{\ast}) C_Z &= 7.05 \times 10^{-3},
\nonumber \\
r_0^{B,h} =& 0.0 \: {\rm GeV^{-2}},
\nonumber \\
r_1^{B,h} =& 0.0 \: {\rm GeV^{-4}}.
\label{predic}
\end{eqnarray}
These numerical estimates are obtained with the weak parameters quoted in
\cite{l3}.
In our first fit with four free parameters we fix all quantities which are zero
in the Born approximation.
Then we allow for additional parameters ($r_0, r_1$) to be fitted.
The numerical results are shown in Table \ref{table}.
The small number of available data points is
certainly disadvantageous for the fit results.
Nevertheless, the table gives some impression on the potential
value of the approach. The gain of accuracy for a smaller number of
floating parameters
is a measure of the degree of biasing the fits with certain assumptions usually
done in a specific ansatz. From our starting point it is evident what would
be a completely unbiased fit - taking into account all higher powers of
$(s-M_Z^2)$ in the cross section ansatz and of $(s-s_Z)$ in the Taylor
expansions of the $F^i(s)$.
\begin{table}\centering
\begin{tabular}[h]{|c|*{4}{r@{$\pm$}l|}}
\hline
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
nb. of
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
parameters
&\multicolumn{2}{c|}{4} &\multicolumn{2}{c|}{5}
&\multicolumn{2}{c|}{5} &\multicolumn{2}{c|}{6} \\
\multicolumn{1}{|c|}{ }&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
\hline
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
$M_{Z}$
&91.134&.020 &91.130 & .020 &91.120& .032 &91.128 & .046
\\
$\Gamma_{Z}$
& 2.506 & .018 & 2.484 & .040 & 2.490 & .034 & 2.484 & .041
\\
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
\hline
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
$R, {\rm GeV^{2}} $
&5.49 & 0.08 & 5.38 & 0.21 & 5.41 & 0.17 & 5.38 & 0.21
\\
$I \times 10^{3}$
&8.9 & 2.4 & 9.5 & 2.5 & 12.1 & 6.3 & 10.1 & 11.3
\\
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
\hline
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
$r_0 \times 10^7, {\rm GeV^{-2}}$
&\multicolumn{2}{c|}{--} & 3.5 & 5.9 &\multicolumn{2}{c|}{--} & 3.0 & 13.1
\\
$r_1 \times 10^{10}, {\rm GeV^{-4}} $
&\multicolumn{2}{c|}{--} &\multicolumn{2}{c|}{--}
& --5.6 & 10.5 & --1.3 & 23.
\\
\multicolumn{1}{|c|}{ } &\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ }
&\multicolumn{2}{c|}{ } &\multicolumn{2}{c|}{ } \\
\hline
\end{tabular}
\caption{Results of S-matrix based fits to the hadronic line shape
as measured at LEP100. An uncertainty of $20$ MeV in the energy scale of
LEP must yet be added to the error of $M_Z$.}
\label{table}
\end{table}
As is known from earlier fits, the determination of the residuum
and of the non-resonating terms is not too stringent if one analyzes only the
line shape. While our accuracy for the $Z$ width is comparable to other
determinations, we have a larger error for the mass. This is due to a strong
correlation between $M_Z$ and the parameters $I, r_1$ in (\ref{sigfin}),
which are not fixed here from the beginning. If one would assume them to be
known from other sources, the mass determination would be better. Similarly,
the $Z$ width is correlated with $R, r_0$. The smaller error of $R$
compared to that of $I$ leads to the relatively small error of $\Gamma_Z$
compared to that of $M_Z$.
\vspace{.4cm}
The measured $Z\:$ mass value differs from earlier determinations
by a non-negligible shift. This is an immediate
consequence
of the S-matrix approach.
The parametrization of the Breit-Wigner resonance formula
for the $Z\:$ peak as being inspired by perturbation theory
assumes usually (but {\it not necessarily} \cite{consoli,willenbrock})
an s-dependent width function $\bar{\Gamma}_Z(s)$
\cite{usual,l3,bbhvn,opal,aleph,delphi}.
In our notations, this would correspond to the following ansatz:
\begin{eqnarray}
{\bf \cal M}^i(s) = \frac {R_{\gamma}}{s} + \frac {\bar{R}_{Z}^i}
{s-\bar{s}_{Z}(s)} + \bar{F}^i(s),
\nonumber \\
\bar{s}_Z(s) = \bar{M}_Z^2 - i \bar{M}_Z \bar{\Gamma}_Z(s).
\label{sbarmatrix}
\end{eqnarray}
The difference between $M_Z, \Gamma_Z, R_{Z}^i$ and $\bar{M}_Z,
\bar{\Gamma}_Z, \bar{R}_{Z}^i$ is described by a transformation
proposed earlier in another context \cite{ztrafo}:
\begin{eqnarray}
\bar{M}_Z = M_Z \sqrt{1+\Gamma_Z^2/M_Z^2}
\approx M_Z + \frac{1}{2} \Gamma_Z^2/M_Z
= M_Z+34 {\rm MeV}, \nonumber
\label{mbar}
\end{eqnarray}
\begin{eqnarray}
\bar{\Gamma}_Z = \Gamma_Z \sqrt{1+\Gamma_Z^2/M_Z^2}
\approx \Gamma_Z + \frac{1}{2} \Gamma_Z^3/M_Z^2
= \Gamma_Z + 1 {\rm MeV}, \nonumber
\label{gbar}
\end{eqnarray}
\begin{eqnarray}
\bar{R}_Z = R_Z (1 + i \Gamma_Z/ M_Z).
\label{zz}
\end{eqnarray}
This transformation
is exact as long as there are no thresholds (opening new
decay channels) or rapidly changing radiative corrections in the vicinity of
the $Z\:$ peak position. Then,
\begin{equation}
\bar{\Gamma}_Z(s) = \frac {s} {\bar{M}_Z^2} \bar{\Gamma}_Z.
\label{gammas}
\end{equation}
If (\ref{gammas}) would be exact, it would follow $\bar{F}(s)=F(s)$.
A dependence of mass and width determinations on
the
theoretical ansatz for a line shape description has been observed earlier, see
e.g. \cite{sakurai}. There, formulae (\ref{zz}) may also be applied in order
to
relate different approaches to the hadron resonances under discussion.
We further mention that the ratio of width and mass is invariant:
\begin{equation}
\frac{ \bar{\Gamma}_Z} {\bar{M}_Z} = \frac{ \Gamma_Z} {M_Z}.
\label{wm}
\end{equation}
\vspace{.5cm}
\\
Although we think that the natural application of the S-matrix approach
to $Z\:$ boson physics is the line shape analysis, we indicate also how
other observables than $\sigma_T(s)$ may be treated.
For a calculation of e.g. the initial state bremsstrahlung contribution to the
left-right asymmetry $A_{LR}$,
the convolution for the numerator $\sigma_{LR}(s)$ must be performed with
a modified ansatz:
\begin{equation}
A_{LR} = \frac{\sigma_{LR}(s)}{\sigma_T(s)}, \hspace{1cm}
\sigma_{LR}(s) =
\int ds' \left[ \sigma_0 + \sigma_1 - \sigma_2 - \sigma_3 \right] (s')
\rho_{\rm ini}(1-s'/s).
\label{alr}
\end{equation}
The helicity cross sections $\sigma_i$ are introduced in (\ref{sigma}).
Similarly, the final state polarisation
$A_{pol}$ may be obtained:
\begin{equation}
A_{pol} = \frac{\sigma_{pol}(s)}{\sigma_T(s)}, \hspace{1cm}
\sigma_{pol}(s) =
\int ds' \left[ \sigma_0 - \sigma_1 - \sigma_2 + \sigma_3 \right] (s')
\rho_{\rm ini}(1-s'/s).
\label{apol}
\end{equation}
The forward-backward asymmetry $A_{FB}$
deserves additional comments.
Due to the different angular
integrations, the weight functions (flux factors) $R_{\rm ini}(z)$ etc.
of the forward-backward difference cross section $\sigma_{FB}(s)$
differ from those for $\sigma_T(s)$. Nevertheless, a convolution
may be derived \cite{h0t,intconv,ringberg}:
\begin{equation}
A_{FB} = \frac{\sigma_{FB}(s)}{\sigma_T(s)}, \hspace{1cm}
\sigma_{FB}(s) =
\int ds' \left[ \sigma_0 - \sigma_1 + \sigma_2 - \sigma_3 \right] (s')
R_{\rm ini}(1-s'/s).
\label{afb}
\end{equation}
Strictly speaking, the $\sigma_i$ for the forward-backward asymmetry are
different from those for the total cross section.
They agree in Born approximation. Further, we know from one-loop calculations
in the standard theory
which changes are to be expected after the introduction of e.g. weak
form factors as proposed in \cite{realistic}.
A study of the usefulness of these formulae should be performed with
a more sophisticated code than ZPOLE in order to describe more realistic
cut situations. A modified version of ZFITTER \cite{alfrun} for this purpose
is in preparation. A larger number of experimental
data points would also be highly desirable
due to the large number of degrees of freedom of the line shape.
\vspace{.8cm}\\
To summarize, we formulate an alternative approach to the $Z\:$ line shape
assuming the existence of an analytic, unitary S-matrix and the validity
of QED.
We demonstrate that
the approach allows reasonable fit results for mass and width of the
$Z\:$ boson. The $Z\:$ mass obtained this way
differs by a well-understood shift of -34 MeV from earlier measurements.
The other line shape parameters determined by the fit can also be
interpreted e.g. in the standard theory.
A dedicated application of the S-matrix approach
to polarized scattering and to $Z'$ physics would also be
interesting.
\vspace{1.cm}
\nopagebreak
\begin{center}
{\bf Acknowledgements}
\end{center}
\vspace{.3cm}
We would like to thank D. Bardin,
A. B\"ohm, W. Hollik, W. Lohmann, R. Stuart, B. Ward
for interesting and stimulating
discussions
and W. Lohmann additionally for support.
|
\section{Introduction}
It has been recognized that turbulence, with its manifold experimental
realizations, is one of the challenging problems
to which very different
approaches, ranging from pure mathematics to engineering applications,
have been developed in an interesting complementary way.
One of the most important methods to study turbulence is, in fact,
the formulation via field theory, based on its relationship with stochastic
partial differential equations \cite{domi,yak}.
However, such technique is far from being
well established and complete, so that new ideas and important improvements
are constantly appearing in the subject.
Recently, Polyakov suggested that
non-unitary minimal models of conformal field theory could be used to describe
the physics of two-dimensional turbulence \cite{poly}.
The advantage of this proposal is that one can deal in a controllable way
with a set of anomalous dimensions and short distance products.
An infinite number of inertial range exponents follows from this approach
\cite{lowe,falko,matsuo}
and one of the still open problems is how to find
``selection rules," which would define the experimentally relevant
minimal models or the connection between them and
statistical characterizations of the random forces acting on the system.
These ideas have attracted the attention of many
authors, and generalizations have
been investigated, as, for instance, possible boundary effects \cite{chung},
alternative physical pictures for the enstrophy and energy cascades
\cite{cateau},
and magnetohydrodynamic turbulence \cite{ferri}.
We will consider, in this paper, the problem of conformal turbulence
including in its formalism the influence of three-dimensional effects.
Our motivation comes from a number of experimental studies, in
which approximately two-dimensional fluids were observed, together with the
unavoidable presence of three-dimensional perturbations
\cite{hopfi,mory1,mory2,max}.
It was verified that a quasi two-dimensional fluid is perturbed by small scale
forces originated from the the degrees of freedom related to the direction
perpendicular to the plane of motion.
We will take this fact into account, noticing that there are
also compressibility effects which cannot be neglected in an effective
two-dimensional theory of the perturbed system.
A generalization of the conformal approach will be devised and new
inertial range exponents will be obtained here, in reasonable agreement
with the experimental data.
This paper is organized as follows. In the next section we briefly review
the most important and practical aspects of conformal turbulence, in order
to make the paper as self contained as possible. In section III, we discuss
some of the experiments carried out to investigate two-dimensional
turbulence. This will motivate us to define an effective (and infinite)
set of stochastic partial differential equations which represents a quasi
two-dimensional fluid under the influence of three-dimensional perturbations.
The conformal approach is then introduced in order to solve the Hopf equations
for the turbulence problem. Furthermore, the constant enstrophy and energy
flux conditions are also studied. Explicit solutions are found and described
in section IV and, in section V, the problem of boundary effects is discussed.
Finally, in section VI we comment on our results and on possible
directions for future investigations.
\section{Conformal Turbulence}
The minimal models of conformal field theory \cite{bpz} are
generically defined by a
pair of relatively prime numbers, $(p,q)$, with $p<q$. These models contain
a subset of
$(p-1)(q-1)/2$ scalar primary operators, $\psi_{(m,n)}$,
labelled by $1\leq m <p$
and $1 \leq n \leq (q-1)/2$, if $p$ is even, or
$1\leq m \leq (p-1)/2$ and $1 \leq n <q$, otherwise, having
dimensions $\Delta_{(m,n)}=\left( (pn-qm)^2-
(p-q)^2 \right)/4pq$. The reason for the choice of scalar operators is that
we will be dealing with isotropic correlation functions in the
turbulence problem. The operator product expansion (OPE) of two primary
operators $\psi_{(r_1,s_1)}(z)$ and $\psi_{(r_2,s_2)}(z')$,
with $|z-z'| \rightarrow 0$ is written as
\begin{eqnarray}
&&\psi_{(r_1,s_1)}(z)\psi_{(r_2,s_2)}(z')=
\sum_{(r_3,s_3)} (a \bar a)^{ \left( \Delta_{(r_3,s_3)}
-\Delta_{(r_1,s_1)}-\Delta_{(r_2,s_2)} \right)}
\sum_{(n,m)} C^{(r_3,s_3)}_{ \{ (n_1,...,n_k);
(m_1,...,m_l) \} } \nonumber \\
&&\times
L_{-n_1}...L_{-n_k} \bar L_{-m_1}...\bar L_{-m_l}
a^{\sum n} \bar a^{\sum m} \psi_{(r_3,s_3)}(z) \ , \ \label{ct1}
\end{eqnarray}
where $|r_1-r_2|+1 \leq r_3 \leq \hbox{min}(r_1+r_2-1,2p-r_1-r_2-1)$,
$|s_1-s_2|+1 \leq s_3 \leq \hbox{min}(s_1+s_2-1,2q-s_1-s_2-1)$ and we have
introduced, in (\ref{ct1}), the Virasoro generators of conformal
transformations, $L_{-n}$ and $\bar L_{-n}$.
The interest in these models is related not only to their finite number
of primary operators, but also to the fact that their dimensions and the form of
short distance products are completely known.
Let us look now at the problem of turbulence in two dimensions and show how it
may be matched \cite{poly} with the above operator structures.
The motion of an
incompressible fluid is assumed, even in the turbulent regime, to be described
by the Navier-Stokes equations for the velocity field,
\begin{equation}
\partial_t v_\alpha + \left( \delta_{\alpha \gamma} - {{\partial_\alpha
\partial_\gamma} \over {\partial^2}} \right) v_\beta \partial_\beta v_\gamma
= f_\alpha + \nu \partial^2 v_\alpha \ , \ \label{ct2}
\end{equation}
where $f_\alpha$ represents a random force acting at large scales, determined
by a characteristic length $L$, and $\nu
\rightarrow 0$ is
the viscosity, associated to the small scale where dissipation effects
come into play, yielding a natural UV cutoff for the system.
In terms of the stream function, $\psi$, related to the velocity
field by $v_\alpha = \epsilon_{\beta \alpha} \partial_\beta \psi$, we may write
the following equation for the vorticity field, $\omega = \partial^2 \psi$,
\begin{equation}
\partial_t \omega + \epsilon_{\alpha \beta} \partial_\alpha \psi
\partial^2 \partial_\beta \psi = \epsilon_{\alpha \beta} \partial_\alpha
f_\beta + \nu \partial^2 \omega \ . \ \label{ct3}
\end{equation}
One of the fundamental problems of turbulence theory is to find solutions of
the Hopf equations, for statistical averages over realizations of the velocity
field,
\begin{equation}
\partial_t [ <\omega(x_1,t) \omega(x_2,t)...\omega(x_n,t)> ] = 0 \ , \
\label{ct4}
\end{equation}
where the time derivative is expressed through the use of equations (\ref{ct3}).
In the inertial range, the standard view of the problem is that both forcing
and viscosity terms may be neglected in order to formulate an effective set
of Hopf equations. Considering, furthermore, the convection term in (\ref{ct3})
as a vanishing point-splitted product of fields, that is, $\oint_{|z-z'|=|a|}
(dz' / a) \epsilon_{\alpha \beta} \partial_\alpha \psi(z)
\partial^2 \partial_\beta \psi(z') \rightarrow 0$, when $|z-z'| \rightarrow 0$,
we would have, then, an exact solution of (\ref{ct4}). A concrete
realization of this possibility
may be achieved if we regard the stream function $\psi$ as
a primary operator of some conformal minimal model. In this case we may use
all the available information on operator dimensions and OPE's to
extract physical results from the analysis of the problem.
According to this assumption, let $\phi$ be the primary operator which has
the lowest dimension, $\Delta \phi$, appearing in the OPE $\psi \psi$,
between fields with the same dimension $\Delta \psi$. Taking $a \equiv
|a| \exp(i \theta)$, we will have, thus,
\begin{eqnarray}
&& \lim_{|a| \rightarrow 0} \oint_{|z-z'|=|a|} {{dz'} \over a}
\epsilon_{\alpha \beta}
\partial_\alpha \psi(z) \partial^2 \partial_\beta \psi(z') \nonumber \\
&&\sim \int d \theta \left [ \partial^2_{\bar a} \partial_a \partial_z
-\partial^2_a \partial_{\bar a} \partial_{\bar z} \right ]
(a \bar a)^{\left( \Delta \phi - 2 \Delta \psi \right)}
\sum C_{\{n;m\}}L_{-n_1}...L_{-n_k} \bar L_{-m_1}...\bar L_{-m_l}
a^{\sum n} \bar a^{\sum m} \phi(z, \bar z) \nonumber \\
&&\sim (a \bar a)^{\left( \Delta \phi - 2 \Delta \psi \right)}
\left [ L_{-2} \bar L_{-1}^2 - \bar L_{-2} L_{-1}^2 \right ] \phi
\ , \ \label{ct5}
\end{eqnarray}
as the dominant contribution in this short distance product. It is important to
note that in order to get (\ref{ct5}) it was necessary to set $C_{\{1;2\}}=
C_{\{2;1\}}$ and $C_{\{1;(1,1)\}}=C_{\{(1,1);1\}}$,
as it follows from the pseudoscalar nature
of the $\epsilon$ factor above. We see, then, that (\ref{ct5}) vanishes with
$|a| \rightarrow 0$ if
\begin{equation}
\Delta \phi > 2 \Delta \psi \ , \ \label{ct6}
\end{equation}
which is one of the constraints that the chosen minimal model has to satisfy.
An additional constraint comes from the condition of a constant
enstrophy or energy flux through the inertial range, meaning that $<\dot \omega
(x) \omega(0)> \sim x^0$ or $<\dot v_\alpha(x) v_\alpha(0)> \sim x^0$,
respectively. In the case of a constant enstrophy flux, we have
\begin{equation}
<\dot \omega(x) \omega(0)> \sim
(a \bar a)^{\left( \Delta \phi - 2 \Delta \psi \right)}
<\left [ \left( L_{-2} \bar L_{-1}^2 - \bar L_{-2} L_{-1}^2 \right)
\phi(x) \right ] \partial^2 \psi(0)> \ . \ \label{ct7}
\end{equation}
The correlation function at the RHS of (\ref{ct7}) is now evaluated by means
of a purely dimensional argument, as $L^{-2 \left( \Delta \phi + \Delta \psi
+3 \right)}$, which makes sense if one thinks that there is an effective
IR cutoff in the theory at the scales where the forcing terms act. Imposing
(\ref{ct7}) to be independent of $L$, we get
\begin{equation}
\Delta \phi + \Delta \psi + 3 = 0 \ . \ \label{ct8}
\end{equation}
In the case of an energy cascade, the argument is the same and the constraint
turns out to be $\Delta \phi + \Delta \psi + 2 = 0$. It is known that there is
an infinite number of minimal models compatible with (\ref{ct6}) and
(\ref{ct8}) \cite{lowe}. The general belief, and still
an open problem, is that there
may be universality classes, associated to the statistical properties of the
forcing terms, which would single out one or another of the possible solutions.
An alternative analysis of conformal turbulence regards the existence of
boundary effects on the vacuum expecation values (VEV's) of single operators
in non-unitary theories \cite{zamo}. In this case, one has to consider the OPE
between $\phi(x)$ and $\psi(0)$ in (\ref{ct7}), picking up the most relevant
operator, let us say, $\chi$. Now, (\ref{ct8}) is modified to
$\Delta \phi + \Delta \psi -\Delta \chi +3 = 0$, with an analogous change
for the constant energy flux condition. Some of these further solutions
(in the enstrophy cascade picture)
were obtained in ref. \cite{chung}.
The connection of the conformal approach with real experiments or numerical
simulations is made
through the computation of inertial range exponents,
which describe the decrease of energy in
the region of higher Fourier modes. In the situation where VEV's of single
operators vanish, the inertial range exponents are given by $4 \Delta \psi
+1$ and, in the opposite case, by $4 \Delta \psi -2\Delta \phi +1$.
A good agreement has been reached between the former possibility, for the
the direct enstrophy cascade case, and numerical simulations \cite{legras,babiano} of the two-dimensional Navier-Stokes equations.
\section{Three-dimensional effects}
In a series of interesting experiments, Hopfinger et al.
\cite{hopfi,mory1,mory2} studied
the turbulence phenomenon as it happens in a rotating tank, where at its
bottom there was an oscilating grid responsible for perturbations of the
fluid motion. According to the Taylor-Proudman theorem
\cite{proudman,taylor,inglis} a rotating
fluid tends to behave as if it were two-dimensional and in fact this was
observed in the form of coherent structures
(vortices) organized in the direction parallel to the rotation axis of
the tank. However, ``defects" in the vortices were seen to propagate from
the very turbulent region at the bottom of the tank up to the effectively
two-dimensional system. The essential picture extracted from these observations
is that the fluid should be best described in terms of two-dimensional equations
containing not only large scale forcing terms but also small scale random
perturbations, originated from either vortex-breakdown or soliton pulses
propagating along
vorticity filaments.
The experimental data suggested then the existence of an inertial range,
likely to be related to a direct enstrophy cascade
and well approximated by $E(k) \sim k^{-2.5}$,
which represents a less steep energy spectrum
than the one obtained by Kraichnan \cite{kraich}, $E(k) \sim
k^{-3}$, or even other proposals \cite{saff,moff}, not excluding
conformal turbulence
\cite{lowe}. This puzzingly result is presently understood to be due only to the
measurement techniques used in the experiments, based on the analysis of the
dispersion of suspended particles in the fluid \cite{mory2}.
More recently, similar
experiments were conduced by Narimousa et al. \cite{max} and direct measurements
of the turbulent velocity field were recorded. The results pointed out the
existence of a possible energy spectra $E(k) \sim k^{-5/3}$ at lower
wavenumbers, in agreement with the conjecture of an inverse energy cascade
\cite{kraich},
and a range at higher wavenumbers, where $E(k) \sim k^{-5.5 \pm 0.5}$. In
this region, the spectral slope was seen to depend on the controlling
external conditions, with results varying from $E(k) \sim k^{-5.0}$ up to
$E(k) \sim k^{-6.0}$. It is worth to note that a spectral law
$E(k) \sim k^{-5}$ follows from Rhines theory of $\beta$-plane turbulence
\cite{rhines} and, alternatively, is closely approximated by some solutions
of the constant enstrophy flux condition in the conformal approach, like
the minimal models $(9,71)$ or $(11,87)$.
The variation of exponents obtained in the experiments may have a theoretical
counterpart in the existence of a set of operator anomalous dimensions,
making it interesting to analyze the problem from the conformal field theory
point of view. It is clear, however, that the inertial range exponents, found in
ref. \cite{lowe}, cannot reproduce the experimental situation. We believe that the important ingredient, missing in the previous conformal approach, is precisely the existence of three-dimensional perturbations, which must be taken into account in any realistic model of a quasi two-dimensional fluid.
In view of the above considerations, let us write the two-dimensional
Navier-Stokes equations as
\begin{equation}
\partial_t v_\alpha + v_\beta \partial_\beta v_\alpha =
\nu \partial^2 v_\alpha + f^{(1)}_\alpha + gf^{(2)}_\alpha -\partial_\alpha P
\ , \ \label{tde1}
\end{equation}
where $f^{(1)}_\alpha$ and $f^{(2)}_\alpha$ are stirring forces defined at
large $(L)$ and small $( \mu << L)$ scales, respectively. The dimensionless
constant $g$ represents,
roughly, a coupling with the three-dimensional modes of the fluid.
We assume that the dissipation scale, $a$, is related , in principle, to the
other scales of the problem as $a<< \mu << L$.
This means that even though the perturbations act at very small scales, when
compared to the macroscopic size of the system, they are still much larger
than the scale where dissipation occurs.
An important point here is that the condition of incompressibility, when
formulated in three dimensions, reads $\partial_1 v_1 +
\partial_2 v_2 +\partial_3 v_3 = 0$,
suggesting that the ``projection" of this
constraint
to the two-dimensional
world has to be given by $\partial_\alpha v_\alpha = {\cal O} (g)$,
in the framework of
equations (\ref{tde1}). The velocity field may be described, then,
by means of a stream function, $\psi$, and a velocity potential, $\phi$, as
\begin{equation}
v_\alpha = \epsilon_{\beta \alpha} \partial_\beta \psi + g \partial_\alpha
\phi \ . \ \label{tde2}
\end{equation}
It is of further interest to study, besides the vorticity $\omega$,
the divergence of $v_\alpha$,
given by $\rho = g \partial^2 \phi$. An exact, although infinite, chain of
equations may be generated if we expand $\psi$ and $\phi$ in powers
of $g$, substituing them into (\ref{tde1}) and collecting the coefficients
of the obtained series. Defining, in this way,
\begin{eqnarray}
&&\psi = \sum^\infty_{n=0} g^n \psi^{(n)} \ , \
\omega= \sum^\infty_{n=0} g^n \omega^{(n)} \ , \ \nonumber \\
&&\phi = \sum^\infty_{n=0} g^n \phi^{(n)} \ , \
\rho= \sum^\infty_{n=0} g^{n+1} \rho^{(n)} \ , \
\label{tde3}
\end{eqnarray}
we get the following set of coupled equations,
\begin{eqnarray}
\hbox{{\it i) }} \partial_t \omega^{(n)} &&+ \sum^n_{p=0} \epsilon_{\alpha \beta}
\partial_\alpha \psi^{(p)}\partial_\beta \partial^2 \psi^{(n-p)} +
\sum^{n-1}_{p=0} \left[ \partial_\beta \phi^{(p)} \partial_\beta
\partial^2 \psi^{(n-p-1)} +
\partial^2 \phi^{(p)} \partial^2 \psi^{(n-p-1)} \right] \nonumber \\
&&= \nu \partial^2 \omega^{(n)} + \epsilon_{\alpha \beta} \partial_{\alpha}
f^{(2)}_\beta \delta_{n,1} \ , \
\nonumber \\
\hbox{{\it ii) }} \partial_t \omega^{(0)} &&+ \epsilon_{\alpha \beta}
\partial_\alpha \psi^{(0)}\partial_\beta \partial^2 \psi^{(0)} =
\nu \partial^2 \omega^{(0)}+\epsilon_{\alpha \beta} \partial_{\alpha}
f^{(1)}_\beta \ , \ \nonumber \\
\hbox{{\it iii) }} \partial_t \rho^{(n)} &&+ \sum^{n-1}_{p=0}
\left[ \partial_\alpha \partial_\beta \phi^{(p)}
\partial_\alpha \partial_\beta \phi^{(n-p-1)} +
\partial_\alpha \phi^{(p)} \partial_\alpha \partial^2 \phi^{(n-p-1)}
\right] \nonumber \\
&&+ \sum^n_{p=0} \left[2 \epsilon_{\alpha \beta}
\partial_\beta \partial_\sigma \phi^{(p)}
\partial_\alpha \partial_\sigma \psi^{(n-p)}
+\epsilon_{\alpha \beta} \partial_\alpha \psi^{(n-p)} \partial_\beta
\partial^2\phi^{(p)} \right] \nonumber \\
&&+\sum^{n+1}_{p=0} \left[ \partial_\alpha \partial_\beta \psi^{(p)}
\partial_\alpha \partial_\beta \psi^{(n-p+1)}
- \partial^2 \psi^{(p)} \partial^2 \psi^{(n-p+1)} \right]
= \nu \partial^2 \rho^{(n)} \ , \
\nonumber \\
\hbox{{\it iv) }} \partial_t \rho^{(0)} &&+2\partial_\alpha \partial_\beta
\psi^{(0)}\partial_\alpha \partial_\beta \psi^{(1)}
+2\epsilon_{\alpha \beta}\partial_\beta \partial_\sigma \phi^{(0)}
\partial_\alpha \partial_\sigma \psi^{(0)}+
\epsilon_{\alpha \beta} \partial_\alpha \psi^{(0)} \partial_\beta
\partial^2\phi^{(0)} \nonumber \\
&&-2 \partial^2 \psi^{(0)} \partial^2 \psi^{(1)}
=\nu \partial^2 \rho^{(0)}+\partial_\alpha f^{(2)}_\alpha
\ , \ \label{tde4}
\end{eqnarray}
and, finally, the constraint of incompressibility for the
$g$-independent part of the velocity field, which defines the pressure term,
\begin{equation}
\left( \partial_\alpha \partial_\beta \psi^{(0)} \right)
\left( \partial_\alpha \partial_\beta \psi^{(0)} \right) - \partial^2
\psi^{(0)} \partial^2 \psi^{(0)} = \partial_\alpha f^{(1)}_\alpha -
\partial^2 P \ . \ \label{tde4.5}
\end{equation}
In the above expressions, $n \geq 1$. We have obtained, therefore, a set
of stochastic partial differential equations.
In a statistical
description, reflecting a stable asymptotic limit for the correlation
functions of $\omega$ and $\rho$, Hopf equations
may be straightforwardly written as
\begin{equation}
\partial_t < \prod_{i=1}^{N} \omega^{(n_i)}(x_i,t) \prod_{j=N+1}^{M}
\rho^{(n_j)}(x_j,t) > = 0 \ . \ \label{tde5}
\end{equation}
We observe now that in (\ref{tde4}), equation $ii)$ is identical
to the one which corresponds to an unperturbed ($g=0$) two-dimensional fluid.
This means that the field $\psi^{(0)}$ will be related to an enstrophy
or energy cascade, even in the presence of three-dimensional effects. This
field plays the role of an external random variable
in the other equations, since its dynamics is independent of the other
components $\psi^{(n)}$ or to the field $\phi$
(in general, the subset $\{ \psi^{(0)}, \psi^{(1)},...,\psi^{(n)},
\phi^{(0)},\phi^{(1)},...,\phi^{(n-1)} \}$ contains fields which act like
external random perturbations in the equations for $\psi^{(p)}$ and
$\phi^{(p-1)}$, with $p \geq n+1$).
Considering that (\ref{tde4}) gives relatively complex equations, the
analysis of the problem might seem hopeless, being perhaps adressed only to
a numerical treatment. However, we can extend the conformal approach, applied
previously to the unperturbed case, to find here solutions of the Hopf
equations. Our basic assumption is that not only $\psi^{(0)}$ but
also the other components in the power expansions of $\psi$ and $\phi$
are primary operators which belong to some minimal model in a conformal field
theory. It is necessary, then, to define a scale $\ell$, possibly associated to
intermittency effects, which allows us to write the following dimensionally
correct expansion,
\begin{eqnarray}
&&\psi = \sum_{n=0}^\infty f_n \ell^{2 (\Delta \psi^{(n)} - \Delta \psi^{(0)})}
g^n \psi^{(n)} \ , \ \nonumber \\
&&\phi = \sum_{n=0}^\infty f_n' \ell^{2 (\Delta \phi^{(n)} - \Delta \psi^{(0)})}
g^n \phi^{(n)} \ , \ \label{tde6}
\end{eqnarray}
where $\psi^{(n)}$ and $\phi^{(n)}$ have dimensions $\Delta \psi^{(n)}$ and
$\Delta \phi^{(n)}$, respectively.
The introduction of a scale $\ell$ in
(\ref{tde6}) means that the perturbed system exhibits a breaking of scale invariance in the inertial range. We will see later that this phenomenon is
signaled by the existence of constant enstrophy or energy fluxes which
depend on the small scales of three-dimensional perturbations. It is conceptually important to understand the physical origin of $\ell$. A clue for this comes from the structure of couplings between $\psi^{(0)}$ and the other fields, as expressed in equation (\ref{tde4}). As we have already observed,
$\psi^{(0)}$ is effectively an external field in the equations for $\psi^{(n)}$
(with $n \geq 1$) and $\phi^{(n)}$ (for any $n$). In this way, it is plausible to have a relation between $\ell$ and the scales involved in the dynamics of
$\psi^{(0)}$. Now, if we consider the turbulent limit of the equations for
$\psi^{(0)}$, corresponding to $\nu \rightarrow 0$ (or, alternatively,
$a \rightarrow 0$), we are left essentially with the correlation length $L$ of
large scale random forces. A simple choice, thus, is to consider $\ell = L$.
In this respect, one may observe that the small scale $\mu$ could also be used in the definition of $\ell$. We have, however, physical reasons to believe that this does not happen: $\mu$ is related to the forcing terms in the equations for $\psi^{(1)}$ and $\phi^{(1)}$, which we expect to be irrelevant when compared to the nonlinear convection terms in the range of wavenumbers given by $|\vec k| << 1/ \mu$.
It is interesting to note that there is an analogy between our problem and the statistical mechanics of second order phase transitions for a system close to its critical point. In this case, one can study deviations of the critical temperature $T_c$ by means of an expansion in $(T-T_c)$ and through the use of the operator structure of the critical theory \cite{amit}. Here, in the turbulence context, the ``critical theory" is just what we get when $g \rightarrow 0$.
In order to simplify the notation we will keep using (\ref{tde3}), with the above observations in mind. We are interested to obtain possible
combinations of primary operators, in eq. (\ref{tde6}),
that would not affect, in the limit $\mu \rightarrow 0$, the constant enstrophy or energy fluxes, obtained from the dynamics of the field $\psi^{(0)}$.
Within this point of view,
it is important, therefore, to consider short distance products of a certain
number operators, as it follows from (\ref{tde4}). Taking two
generic primary operators, $O^{(p)}_1$ and $O^{(p')}_2$
(for example, $\phi^{(p)}$ and $\psi^{(p')}$),
with dimensions $\Delta O^{(p)}_1$ and $\Delta O^{(p')}_2$,
respectively, we may write
\begin{eqnarray}
&&O^{(p)}_1(z,\bar z) O^{(p')}_2(z', \bar z')
= (a \bar a)^{\left(
\Delta A^{(p,p')}_{O_1,O_2} - \Delta O^{(p)}_1 - \Delta O^{(p')}_2 \right)}
\nonumber \\
&&\times \sum C^{O_1^{(p)},O_2^{(p')}}_{\{(n_1,...,n_k);(m_1,...,m_\ell) \}}
L_{-n_1}...L_{-n_k} \bar L_{-m_j}...\bar L_{-m_\ell}
a^{\sum n} \bar a^{\sum m} A^{(p,p')}_{O_1,O_2}(z,\bar z) \ , \
\label{tde7}
\end{eqnarray}
where $A^{(p,p')}_{O_1,O_2}$ is the primary operator with the
lowest dimension in the above OPE. The short distance products
appearing in (\ref{tde4}) are listed below
together with the conformal field
theory representation, obtained after straightforward computations:
\begin{eqnarray}
&&\bullet \epsilon_{\alpha \beta} \partial_\alpha
\psi^{(p)}\partial_\beta \partial^2 \psi^{(p')}
\sim (a \bar a)^{\left (\Delta A^{(p,p')}_{\psi \psi}- \Delta \psi^{(p)}
-\Delta \psi^{(p')} \right)}
\left[ L_{-2} \bar L_{-1}^2 - \bar L_{-2} L_{-1}^2 \right]
A^{(p,p')}_{\psi \psi}
\nonumber \\
&&\bullet \partial_\beta \phi^{(p)} \partial_\beta \partial^2 \psi^{(p')}
\sim \partial^2 \phi^{(p)} \partial^2 \psi^{(p')} \sim
(a \bar a)^{\left (\Delta A^{(p,p')}_{\phi \psi}- \Delta \phi^{(p)}
-\Delta \psi^{(p')} -1\right)}
L_{-1} \bar L_{-1} A^{(p,p')}_{\phi \psi} \nonumber \\
&&\bullet \partial_\alpha \partial_\beta \phi^{(p)}
\partial_\alpha \partial_\beta \phi^{(p')}
\sim \partial_\alpha \phi^{(p)} \partial_\alpha \partial^2 \phi^{(p')} \sim
(a \bar a)^{\left (\Delta A^{(p,p')}_{\phi \phi}- \Delta \phi^{(p)}
-\Delta \phi^{(p')} -1 \right)}
L_{-1} \bar L_{-1} A^{(p,p')}_{\phi \phi} \nonumber \\
&&\bullet \epsilon_{\alpha \beta} \partial_\beta \partial_\sigma
\phi^{(p)} \partial_\alpha \partial_\sigma \psi^{(p')}
\sim \epsilon_{\alpha \beta} \partial_\alpha
\psi^{(p')}\partial_\beta \partial^2 \phi^{(p)} \sim
(a \bar a)^{\left (\Delta A^{(p,p')}_{\phi \psi}- \Delta \phi^{(p)}
-\Delta \psi^{(p')} \right)} \nonumber \\
&&\times \left[ L_{-2} \bar L_{-1}^2 - \bar L_{-2} L_{-1}^2 \right]
A^{(p,p')}_{\phi \psi}
\nonumber \\
&&\bullet \partial_\alpha \partial_\beta \psi^{(p)}
\partial_\alpha \partial_\beta \psi^{(p')} \sim
\partial^2 \psi^{(p)} \partial^2 \psi^{(p')} \sim
(a \bar a)^{\left (\Delta A^{(p,p')}_{\psi \psi}- \Delta \psi^{(p)}
-\Delta \psi^{(p')} -1 \right)} A^{(p,p')}_{\psi \psi}
\ , \ \label{tde8}
\end{eqnarray}
with $p,p' \geq 0$, except in the last relation, for the product
of the type $\psi \psi$, where $p+p'\geq 1$.
From (\ref{tde8}) we see clearly that Hopf equations are satisfied if
\begin{eqnarray}
&&\hbox{{\it i)} } \Delta A^{(0,0)}_{\psi \psi}- 2 \Delta \psi^{(0)}
> 0 \ , \ \nonumber \\
&&\hbox{{\it ii)} } \Delta A^{(p,p')}_{\psi \psi}- \Delta \psi^{(p)}
-\Delta \psi^{(p')}-1> 0 \ , \ \nonumber \\
&&\hbox{{\it iii)} } \Delta A^{(p,p')}_{\psi \phi}- \Delta \psi^{(p)}
-\Delta \phi^{(p')}-1> 0 \ , \ \nonumber \\
&&\hbox{{\it iv)} } \Delta A^{(p,p')}_{\phi \phi}- \Delta \phi^{(p)}
-\Delta \phi^{(p')}-1> 0 \ , \ \label{tde9}
\end{eqnarray}
with $p+p' \geq 1$ in {\it ii)} and $p,p' \geq 0$ in {\it iii)}
and {\it iv)}. These equations
are the first step in the generalization
of unperturbed conformal turbulence, in order to deal with a larger set
of primary operators. We will find more inequalities, restricting, then, up
to some extent the number of different operators allowed in the theory.
Let us now write the conditions for constant enstrophy or energy fluxes
through the inertial range. Here we will assume that vacuum expectation
values of primary operators are zero.
The alternative possibility is discussed in the next section.
The case of a constant enstrophy flux requires,
as commented before, that we compute $<\dot \omega(x) \omega(0)>$.
From relations (\ref{tde4}) we get
\begin{eqnarray}
&&<\dot \omega (x) \omega (0)> =
\nonumber \\
&&\sum_{n,m}^\infty \sum_{p=0}^n g^{n+m} < \left \{
\epsilon_{\alpha \beta}
\partial_\alpha \psi^{(p)} \partial_\beta \partial^2 \psi^{(n-p)}
+g \left[ \partial_\beta
\phi^{(p)} \partial_\beta
\left. \partial^2 \psi^{(n-p)} + \partial^2 \phi^{(p)} \partial^2 \psi^{(n-p)}
\right] \right \} \right \vert_x \nonumber \\
&&\times \partial^2 \psi^{(m)}(0)> \ . \ \label{tde10}
\end{eqnarray}
In the above expression we may define a ``large scale" part as the one
which depends solely on the field $ \psi^{(0)}$ and a ``small scale" part,
involving the fields $\phi^{(p)}$ and the other components of $\psi$.
Regarding the large scale part, we already know, from the study of the
unperturbed case, that the constant enstrophy flux condition is
\begin{equation}
\Delta A^{(0,0)}_{\psi \psi} + \Delta \psi^{(0)}
+ 3 = 0 \ . \ \label{tde11.1}
\end{equation}
It is natural to assume, like in the case of unperturbed
conformal turbulence, that the correlation functions in the small scale
part may also be evaluated by means of a dimensional argument, where,
instead of using the typical large scale
parameter $L$, the correct choice turns out to be the small length scale $\mu$.
Assuming, furthermore, that $\mu \rightarrow 0$ leads to a well defined
limit, we just require the powers of $\mu$ in the most relevant terms
belonging to the small scale part of (\ref{tde10}) (contributions which have
the lowest power of $a \bar a$) to be non-negative numbers.
This discussion may be restated by saying that we will need to select one
or more of the following conditions,
\begin{eqnarray}
&&\hbox{{\it i)} } \Delta A^{(p,p')}_{\phi \psi} + \Delta \psi^{(p'')}
+ 2 \leq 0 \ , \ \nonumber \\
&&\hbox{{\it ii)} } \Delta A^{(p,p')}_{\psi \psi} + \Delta \psi^{(p'')}
+ 3 \leq 0 \ , \ \label{tde11}
\end{eqnarray}
according to the analysis of the dominant terms in the small scale part of
$<\dot \omega(x) \omega(0)>$.
In the derivation of (\ref{tde11}) we have used the OPE's computed from
the Hopf equations, given by (\ref{tde8}).
An additional care must be taken if it happens that
$A^{(p,p')}_{\psi \psi} = \psi^{(p'')}$ or
$A^{(p,p')}_{\phi \psi}=\psi^{(p'')}$ for some values of $p,p'$ and $p''$.
In this circumstance it is necessary to have
$\Delta \psi^{(p'')} = -3/2$ or $\Delta \psi^{(p'')} = -1$, respectively,
to assure spatially independent correlation functions and hence a constant
enstrophy flux.
Let us turn now to the case of a constant energy flux. We have
\begin{eqnarray}
&&< \dot v_\alpha(x) v_\alpha(0)> = \nonumber \\
&&- \sum_{n,m}^\infty \sum_{p=0}^n
g^{n+m} < \left \{ g^2 \partial_\beta \phi^{(p)} \partial_\beta
\partial_\alpha \phi^{(n-p)} + g \left[ \epsilon_{\gamma
\alpha} \partial_\beta \phi^{(p)} \partial_\beta \partial_\gamma \psi^{(n-p)}
\right.
+\epsilon_{\sigma \beta}
\partial_\sigma \psi^{(p)} \partial_\beta
\partial_\alpha \phi^{(n-p)} \right] \nonumber \\
&&\left. \left. +\partial_\alpha \partial^{-2} \left[
\partial^2 \psi^{(0)} \partial^2 \psi^{(0)}
-\partial_\sigma \partial_\beta \psi^{(0)}
\partial_\sigma \partial_\beta \psi^{(0)} \right] \delta_{n,0}
+ \epsilon_{\sigma \beta} \epsilon_{\gamma \alpha}
\partial_\sigma \psi^{(p)} \partial_\beta
\partial_\gamma \psi^{(n-p)}
\right \} \right \vert_x \nonumber \\
&&\times \left(
\epsilon_{\eta \alpha} \partial_\eta \psi^{(m)}(0)+g \partial_\alpha \phi^{(m)}
(0) \right)> \ . \
\label{tde12}
\end{eqnarray}
Here we cannot refer immediately to the Hopf equations and formulate
a set of conditions, as we did in the constant enstrophy flux case.
There are, in (\ref{tde12}), OPE's which do not appear in (\ref{tde4}), viz.
\begin{eqnarray}
&&\bullet \epsilon_{\sigma \beta} \epsilon_{\gamma z}
\partial_\sigma \psi^{(p)} \partial_\beta \partial_\gamma \psi^{(p')}
\sim (a \bar a)^{\left (\Delta A^{(p,p')}_{\psi \psi}- \Delta \psi^{(p)}
-\Delta \psi^{(p')} -1 \right)} L_{-1} A^{(p,p')}_{\psi \psi} \nonumber \\
&&\bullet \epsilon_{\sigma \beta} \epsilon_{\gamma z}
\partial_\sigma \psi^{(0)} \partial_\beta \partial_\gamma \psi^{(0)}
+\partial_z \partial^{-2} \left[
\partial^2 \psi^{(0)} \partial^2 \psi^{(0)}
-\partial_\sigma \partial_\beta \psi^{(0)}
\partial_\sigma \partial_\beta \psi^{(0)} \right] \nonumber \\
&&\sim (a \bar a)^{ \left( \Delta A^{(0,0)}_{\psi \psi} -2 \Delta \psi^{(0)}
\right) }
\left[ \bar L_{-1} L_{-2} -4 \partial^{-2} L_{-1}^3 \bar L_{-2} \right]
A^{(0,0)}_{\psi \psi} \nonumber \\
&&\bullet \epsilon_{\gamma z} \partial_\beta \phi^{(p)} \partial_\beta
\partial_\gamma \psi^{(p')} \sim
\epsilon_{\sigma \beta} \partial_\sigma \psi^{(p')} \partial_\beta
\partial_z \phi^{(p)} \sim
(a \bar a)^{\left (\Delta A^{(p,p')}_{\phi \psi}- \Delta \phi^{(p)}
-\Delta \psi^{(p')} -1\right)} L_{-1} A^{(p,p')}_{\phi \psi}
\nonumber \\
&&\bullet \partial_\beta \phi^{(p)} \partial_\beta \partial_z \phi^{(p')}
\sim (a \bar a)^{\left (\Delta A^{(p,p')}_{\phi \phi}- \Delta \phi^{(p)}
-\Delta \phi^{(p')} -1\right)} L_{-1} A^{(p,p')}_{\phi \phi}
\ . \ \label{tde13}
\end{eqnarray}
In (\ref{tde13}) we have $\epsilon_{\gamma z} \equiv (
\epsilon_{\gamma 1} - i \epsilon_{\gamma 2} )/2$.
Using the above point-splitted products, we can get the constant energy
flux conditions.
First, the equation following from the large scale part of $< \dot v(x) v(0)>$,
\begin{equation}
\Delta A^{(0,0)}_{\psi \psi} + \Delta \psi^{(0)} +2 =0
\label{tde14.1}
\end{equation}
and then the inequalities which come from the small scale terms,
\begin{eqnarray}
&&\hbox{{\it i) }} \Delta A^{(p,p')}_{\psi \psi} + \Delta \psi^{(p'')} +1
+ \delta_{p+p',0} \leq 0
\ , \ \nonumber \\
&&\hbox{{\it ii) }} \Delta A^{(p,p')}_{\psi \psi} + \Delta \phi^{(p'')} +1
+ \delta_{p+p',0} \leq 0
\ , \ \nonumber \\
&&\hbox{{\it iii) }} \Delta A^{(p,p')}_{\phi \phi}+ \Delta \psi^{(p'')}+1
\leq 0
\ , \ \nonumber \\
&&\hbox{{\it iv) }} \Delta A^{(p,p')}_{\phi \phi}+ \Delta \phi^{(p'')}+1
\leq 0
\ , \ \nonumber \\
&&\hbox{{\it v) }} \Delta A^{(p,p')}_{\phi \psi}+ \Delta \phi^{(p'')}+1
\leq 0
\ , \ \nonumber \\
&&\hbox{{\it vi) }} \Delta A^{(p,p')}_{\phi \psi}+ \Delta \psi^{(p'')}+1
\leq 0
\ , \ \label{tde14}
\end{eqnarray}
where, analogously to the enstrophy cascade case, only a subset
of (\ref{tde14}) has to be considered.
Supplementary relations, like the ones obtained after equation (\ref{tde11}),
are in order, to avoid possible $x$-dependent correlation functions.
We are led, here, to
\begin{eqnarray}
&&\hbox{{\it i) }} \Delta \psi^{(p)} = -1 \hbox{, if }
A_{\psi \psi}^{(0,0)} = \psi^{(p)} \ , \ \nonumber \\
&&\hbox{{\it ii) }} \Delta \phi^{(p)} = -1 \hbox{, if }
A_{\psi \psi}^{(0,0)} = \phi^{(p)} \ , \ \nonumber \\
&&\hbox{{\it iii) }} \Delta \psi^{(p'')} = -1/2 \hbox{, if }
A_{\psi \psi}^{(p,p')} = \psi^{(p'')} \hbox{, for $p+p'>0$, or }
A_{\phi \psi}^{(p,p')} = \psi^{(p'')} \nonumber \\
&&\hbox{ or }
A_{\phi \phi}^{(p,p')} = \psi^{(p'')} \ , \ \nonumber \\
&&\hbox{{\it iv) }} \Delta \phi^{(p'')} = -1/2 \hbox{, if }
A_{\psi \psi}^{(p,p')} = \phi^{(p'')} \hbox{, for $p+p'>0$, or }
A_{\phi \psi}^{(p,p')} = \phi^{(p'')} \nonumber \\
&&\hbox{ or }
A_{\phi \phi}^{(p,p')} = \phi^{(p'')} \ . \ \label{tde15}
\end{eqnarray}
Once we have some solution at hand, derived from the conditions
obtained here, we may associate inertial range exponents to each one
of the fields $\psi^{(p)}$ and $\phi^{(p)}$, expressed by $4 \Delta
\psi^{(p)}+1$ and $4 \Delta \phi^{(p)}+1$, respectively.
From these values, we have to select the one which will appear effectively
in experimental situations. This problem is investigated in the
following section.
\section{analysis of the constant flux conditions}
We have obtained so far all the conditions necessary to find
minimal models related to an enstrophy or energy cascade in
a quasi two-dimensional fluid. In order to explore them, the first
observation we can make is that these models
must belong to the infinite set of
solutions found in the former study of unperturbed conformal turbulence.
This follows directly from the conditions which depend only on
$\psi^{(0)}$.
A strategy of computation could be, thus, just a numerical analysis of all
possible combinations of fields for these previously known minimal models.
As straight it may sound, this approach is hardly useful when the number of
primary operators becomes large, a fact that happens already for the first few
minimal models.
A more interesting computational scheme is provided if we look for solutions of
the form
\begin{eqnarray}
&& \psi = \psi_0 + f_a(g)\psi_1 \ , \ \nonumber \\
&& \phi = f_b(g) \phi_0 \ , \ \label{afc1}
\end{eqnarray}
where $f_a(0)=f_b(0)=0$, that is, we are considering solutions with
$\psi^{(p)} = \psi_1$, for $p \geq 1$, and $\phi^{(p)} = \phi_0$, for any $p$.
This approach is valuable since a little reflection shows that if
it is impossible to satisfy the constant flux conditions through any pair
of fields $\psi_1$ and $\phi_0$, then there are no further solutions
for the model under consideration.
All our task is, therefore, to consider the set of minimal models representing
conformal turbulence without perturbations, from which the fields $\psi_0$
may be immediately obtained, and add, according to the new constraints
associated to three-dimensional effects, the fields $\psi_1$ and $\phi_0$.
In the study of the inertial range exponents, we may think of, at least, three
limits for $f_{a,b}(g)$: a) $ g \rightarrow 0$, that is, $f_{a,b}(g)
\rightarrow 0$, b) $f_{a,b}(g) \simeq 1$, and c) $g >>1$, which may be defined
as a ``strong coupling" regime. In the first case, the perturbations play a
negligible role and everything is described by unperturbed conformal turbulence.
A competition between exponents shows up in the second case, where the less
steep spectral slope will be the most relevant in the limit of higher
wavenumbers. We see, in this way, that cases a) and b) cannot give any of
the steeper spectral slopes observed in real experiments. The third case is,
in fact, where we have some hope to find a relation with experimental results.
It would be unphysical to have $f_{a,b}(g) \rightarrow 0$, for large values of
$g$, since in this limit we would recover the unperturbed system.
Also, it is unlikely to have $f_{a,b}(g) \rightarrow$ {\it const.}: taking, for instance, gaussian random forces $f^{(1)}_\alpha$ and $f^{(2)}_\alpha$, with $<f^{(1)}_\alpha (\vec x,t) f^{(2)}_\beta (\vec x',t')>=0$ and $<f^{(2)}_\alpha (\vec x,t) f^{(2)}_\beta (\vec x',t')>= D_{\alpha \beta} (|\vec x - \vec x'|) \delta (t-t')$, it may be proved, from the retarded nature of the diffusion propagator, that
$<f^{(2)}_\alpha (\vec x,t) v_\beta (\vec x',t)> = g D_{\alpha \beta} (|\vec x - \vec x'|)$, yielding
\begin{equation}
{\partial \over {\partial g} } \left [
{ {<f^{(2)}_\alpha (\vec x,t) v_\beta (\vec x',t)>} \over
{D_{\alpha \beta} (|\vec x - \vec x'|)}} \right ] = 1 \ . \ \label{afc2}
\end{equation}
Let us assume, thus, that $f_{a,b}(g)$ diverges as $g \rightarrow \infty$.
This means that the inertial range exponent derived from $\psi_0$ may be
discarded and we have to analyze only the competition between the exponents
obtained from $\psi_1$ and $\phi_0$.
We performed an investigation of the first six minimal models for both the
enstrophy and energy cascade cases. In the enstrophy case we found solutions
for all the models studied. They are represented in table 1 and in
figure 1. In table 1, we show the fields $\psi_1$ and $\phi_0$
for the minimal models (2,21), (3,25) and (3,26), together with their
associated inertial range exponents. As the number of solutions became larger,
we had to represent the other three models, (6,55), (7,62) and (8,67) in
figure 1, where we plotted the most relevant exponents, found from the
competition between $\psi_1$ and $\phi_0$, in the strong coupling regime.
We observe, from the results, that there is a good agreement with experimental
verifications, with the only considerable deviation occuring for the very small
set of two solutions for the model (2,21).
The solutions, excluding the model (2,21), were organized in table 3, where
values of mean exponents and standard deviations are described. It is clearly
seen that the perturbed exponents are in general lesser than the exponents
of the unperturbed fluid.
In the energy case, an interesting fact happened: most of the models studied
did not yield any solution for the fields $\psi_1$ and $\phi_0$. Only the
model (10,59), represented in table 2, gave solutions, all of them
with inertial range exponents close to $-3.0$, which do not support
the conjecture of a Kolmogorov exponent, $-5/3$, for the range of lower
wavenumbers. However, more theoretical and experimental work is necessary in
order to arrive at a conclusive answer on this point.
\section{boundary perturbations}
It is worth to understand what happens when boundary effects are supposed to
have some influence in the problem of conformal turbulence. Below, we obtain
the set of conditions needed to account for it, leaving a numerical
analysis for future investigations.
The basic modification here is that we have to study further OPE's in the
conditions of constant enstrophy or energy fluxes, found in section III,
since now VEV's of single operators do not necessarily vanish. In this
way, let us define the primary operator
$A_{\left( O_1O_2 \right) O_3}^{(p,p')p''}$ as the one with the
lowest dimension appearing in the OPE
$\left( O_1^{(p)} O_2^{(p')} \right) O_3^{(p'')}$, where the product of
$O_1^{(p)}$ and $O_2^{(p)}$ was computed first. The conditions we are looking
for must be obtained from the analysis of the $x$-dependence of the dominant
terms in (\ref{tde10}) and (\ref{tde12}). In the situation of a constant
enstrophy flux, the large scale part of (\ref{tde10}) gives
\begin{equation}
\Delta A^{(0,0)}_{\psi \psi} + \Delta \psi^{(0)}
-\Delta A_{\left( \psi \psi \right) \psi }^{(0,0)0}+ 3 = 0
\ , \ \label{bp1}
\end{equation}
which is nothing else than the condition established in section II, in a
diferent notation. On the other side, the small scale part of (\ref{tde10})
gives one or more of the following conditions:
\begin{eqnarray}
&&\hbox{{\it i)} } \Delta A^{(p,p')}_{\phi \psi} + \Delta \psi^{(p'')}
-\Delta A_{\left( \phi \psi \right) \psi}^{(p,p')p''}
+ 2 = 0 \ , \ \nonumber \\
&&\hbox{{\it ii)} } \Delta A^{(p,p')}_{\psi \psi} + \Delta \psi^{(p'')}
- \Delta A_{\left( \psi \psi \right) \psi}^{(p,p')p''}
+ 3 = 0 \ . \ \label{bp2}
\end{eqnarray}
A similar analysis for the case of an energy cascade yields, for the large and
small scale parts of (\ref{tde12}), respectively,
\begin{equation}
\Delta A^{(0,0)}_{\psi \psi}
+ \Delta \psi^{(0)}
- \Delta A_{\left( \psi \psi \right) \psi}^{(0,0)0}+2 =0
\label{bp3}
\end{equation}
and
\begin{eqnarray}
&&\hbox{{\it i) }} \Delta A^{(p,p')}_{\psi \psi} + \Delta \psi^{(p'')}
- \Delta A_{\left( \psi \psi \right) \psi}^{(p,p')p''}
+1 + \delta_{p+p',0} = 0
\ , \ \nonumber \\
&&\hbox{{\it ii) }} \Delta A^{(p,p')}_{\psi \psi} + \Delta \phi^{(p'')}
- \Delta A_{\left( \psi \psi \right) \phi}^{(p,p')p''}
+1 + \delta_{p+p',0} = 0
\ , \ \nonumber \\
&&\hbox{{\it iii) }} \Delta A^{(p,p')}_{\phi \phi}+ \Delta \psi^{(p'')}
- \Delta A_{\left( \phi \phi \right) \psi}^{(p,p')p''} + 1 = 0
\ , \ \nonumber \\
&&\hbox{{\it iv) }} \Delta A^{(p,p')}_{\phi \phi}+ \Delta \phi^{(p'')}
- \Delta A_{\left( \phi \phi \right) \phi}^{(p,p')p''} + 1 = 0
\ , \ \nonumber \\
&&\hbox{{\it v) }} \Delta A^{(p,p')}_{\phi \psi}+ \Delta \phi^{(p'')}
- \Delta A_{\left( \phi \psi \right) \phi}^{(p,p')p''} + 1 = 0
\ , \ \nonumber \\
&&\hbox{{\it vi) }} \Delta A^{(p,p')}_{\phi \psi}+ \Delta \psi^{(p'')}
- \Delta A_{\left( \phi \psi \right) \psi}^{(p,p')p''} + 1 = 0
\ . \ \label{bp4}
\end{eqnarray}
The computation of inertial range exponents is also modified. We now have to
consider all possible combinations like $\psi^{(p)} \psi^{(p')}$ and
$\phi^{(p)} \phi^{(p')}$ in the evaluation of the velocity-velocity
correlation function. The observed inertial range exponent
must be obtained from
$2 \Delta \psi^{(p)} + 2 \Delta \psi^{(p')} - 2 \Delta A_{\psi \psi}^{(p,p')}
+1$ or $2 \Delta \phi^{(p)} + 2 \Delta \phi^{(p')}
- 2 \Delta A_{\phi \phi}^{(p,p')}+1$.
\section{conclusion}
The problem of two-dimensional turbulence was investigated, taking into account
the presence of three-dimensional perturbations. They were introduced in
an effective way, represented by random forcing terms which act at small
scales in the two-dimensional Navier-Stokes equations, as suggested by
experimental observations. A coupling constant, $g$, related to
the strength of these
additional forces, allowed us to write a power expansion for the velocity
field, containing also a compressible part. An infinite set
of equations was found by just collecting terms with the same powers of $g$.
The components $\psi^{(p)}$ and $\phi^{(p)}$, appearing in the power expansion
of the velocity field were assumed to be primary operators of
some conformal minimal model. We obtained, then, from
point-splitted products of operators, a group of conditions in order to have a
solution of the Hopf equations and to reproduce the situation of a constant
enstrophy or energy flux through the inertial range.
In the constant flux conditions, large and small scale terms were defined and
evaluated by means of an extension of the dimensional argument employed
formely in the study of analogous correlation functions.
An analysis of the first six minimal models of unperturbed conformal turbulence
was performed, showing that the picture of a constant enstrophy cascade is in
good agreement with experimental data, yielding inertial range exponents, for
the strong coupling regime, $g>>1$, very close to the ones observed in the
laboratory. Regarding the energy cascade case, we noticed that most of the
minimal models considered in our study were unable to give solutions for the
perturbed system. Only one solution was obtained, with inertial range exponents
around $-3.0$. It would be interesting to investigate further minimal models
for the energy case, in order to see if a closer connection with the results
indicated in experiments could be reached.
From tables 1 and 2, and figure 1, we see clearly that there are many solutions,
differing by just one of the fields $\psi_1$ or $\phi_0$, which give exactly
the same inertial range exponents, in the strong coupling regime. It is
tempting, then, to conjecture that one could find ``plateaux" for the
spectral slopes, while varying some set of external parameters. This question
is contained, of course, in the deeper problem of how to match large scale
properties of the fluid with the minimal models describing the inertial range.
A point which deserves attention is the crossover between unperturbed conformal turbulence and the results obtained in the strong coupling regime. A bridge between these two situations may be investigated not only by varying $g$, as we did in section IV, but also through $\mu \rightarrow 0$, when the effects of small scale three-dimensional perturbations on the constant enstrophy or energy fluxes become negligible. Finally, it is important to stress that a standard direct numerical simulation of equations (\ref{tde4})-(\ref{tde4.5}), up to some level $n$ in their hierarchy, would be an interesting way to study the above questions and the physical assumptions addressed in the present work.
\acknowledgments
I would like to thank C. Callan and D. Gross for the kind hospitality at the
Physics Department of Princeton University, where this paper was completed.
Also, I would like to thank A. Migdal for sharing his insights on the general
turbulence problem, and M. Moriconi for interesting comments. This work was
supported by CNPq (Brazil).
|
\section{Introduction}
The usual kinematic variables used
for discussing deep-inelastic electron-proton scattering
are derived from the four-momenta $p$ of the incoming proton
and $q$ of the exchanged virtual photon: $Q^2\equiv -q^2$ and
the Bj{\o}rken variable $x\equiv Q^2/2\, p\cdot q$.
In the region where higher twist effects are likely to be negligible, i.\ e.,
for $W^2\equiv Q^2\,(1/x-1)\gtrsim 10$ $\mbox{GeV}^2$ , the
(Dokshitzer-Gribov-Lipatov-)Altarelli-Parisi [(DGL)AP] set of equations
\cite{ap,dgl} describes the evolution of the structure function $F_2$ with
$Q^2$ very well. In leading order (LO) all powers of
$\alpha_s\,\ln (Q^2/\mu^2)$ are summed by the AP evolution equations,
which take into account just strongly ordered parton-$k_T$ ladders.
Nowadays usually the next-to-leading order (NLO) set is used, where terms
of the form $\alpha_s^n\,\ln^{n-1}(Q^2/\mu^2)$ are also summed,
taking {\em non-\/}ordered $k_T$ contributions (covariantly) into account
as well. Even in NLO, the AP equations do obviously not contain all
leading logarithms in $x$. Thus one might na\"\i vely expect
the AP framework to break down at some small value of $x$, where a
resummation of all powers of $\alpha_s\,\ln(1/x)$ should be necessary,
although no perturbative instability between LO and NLO has been
observed thus far in the presently relevant kinematic regime
\cite{grv92,grs,grv94}.
Such a resummation in LO is provided by the Balitskij-Fadin-Kuraev-Lipatov
(BFKL) equation \cite{bfkl}. The equation treats the gluons only,
which are expected to be the dominant partons at small $x$.
This might be deduced from the AP splitting functions \cite{ap},
which become $\sim 1/x$ for splitting to gluons and constant for splitting to
quarks in the small $x$ limit. It should be remarked though,
that this argument may be too simple, as it neglects the influence of the
particular shape of the parton distributions used \cite{grs,grv94} and
furthermore does not respect the fundamental energy-momentum conservation
constraint.
The BFKL resummation is formally correct for all $Q^2$, but
a fixed coupling $\overline{\alpha}_s$ was used in its derivation.
Furthermore it is based on LO perturbative QCD, using (non-covariant) $k_T$
cut-off regularizations, thus we are limited to
sufficiently large $Q^2$. Since the resummation assumes that subleading
terms in $x$ are small, including those involving logarithms of $Q^2$,
it will also not be valid at high $Q^2$.
A conservative range for $Q^2$ would be 4 to 50 $\mbox{GeV}^2$ , and the range
0.8 to 120 $\mbox{GeV}^2$\ explored in this work should be regarded as the
extreme limit.
To estimate the effects of the subleading terms, a NLO resummation
in $x$ would be necessary. But even though there has been some progress
in that direction \cite{xnlo}, a final result has not yet been obtained.
Ultimatively it should even be possible to find a unified evolution equation
covering the whole perturbative region \cite{unify,ColEl,KMS}.
But a calculation
that can be confronted with experiment is still missing,
thus we stay with the usual BFKL formalism for the time being.
Since a fixed coupling constant seems unreasonable in view of the
running coupling of the AP framework we wish to connect to, the replacement
$\overline{\alpha}_s\rightarrow\alpha_s(Q^2)$ is done by hand.
There is really no rigorous motivation for this step. Some trust
in this procedure can be gained by considering the representation
of the evolution equations as ladder diagrams. It is well known that
the LO AP equations can be represented in a physical gauge by a sum of
ladder graphs which are strongly ordered in the transverse momentum
$k^2$: $Q^2\gg k^2\gg k_{n-1}^2\gg\ldots\gg k_1^2\gg k_0^2$.
In the small $x$ limit of the AP equation we consider only the dominant
gluon ladder (see Fig. \ref{leiter}) and keep only the terms
with double leading logarithms (DLL)
$\alpha_s\,\ln (1/x)\,\ln(Q^2/\mu^2)$. This corresponds to introducing
an additional strong ordering in $x$: $x\ll x_{n-1}\ll\ldots\ll x_0$.
The BFKL evolution can also be described in terms of a (reggeized) gluon
ladder, using the strong ordering in $x$ only, to get the
leading logarithms in $x$. If we let the coupling run, then
BFKL will reduce to LO DLL upon imposing the ordering in transverse momentum.
The main feature of the BFKL evolution with fixed coupling constant
$\overline{\alpha}_s$ is the growth of the unintegrated gluon
$\sim x^{-\lambda}$ with $\lambda=\frac{3\,\overline{\alpha}_s}{\pi}\,4\,\ln 2
\simeq 0.5$.
Due to the dominance of the gluons at small $x$ we expect a corresponding
rise at small $x$ in the structure functions too, once the off-shell
gluons have been appropriately coupled to the quark sector. This expectation
has been confirmed \cite{AKMS_mod,AKMS_rev} for a small range
of medium $Q^2$ even
in the case of running coupling BFKL, with $F_2\sim x^{-\lambda}$ and
$\lambda\gtrsim 0.5$ for $x<10^{-3}$.
Experimentally the situation has improved drastically since the
advent of the HERA $ep$ collider. Before HERA only the Fermilab
experiment E665 \cite{E665} was able to reach the small $x$ region,
but at rather
low $Q^2$. Now HERA takes data \cite{HERA_93,HERA_pre}
with $x/Q^2\gtrsim 3\cdot 10^{-5}$
over a wide range of $Q^2$. The observed strong rise of $F_2$
at small $x$ has boosted the interest in the BFKL formalism. On the
other hand the dynamically generated AP partons \cite{grv94} create a steep
gluon differently, via a long evolution length in $Q^2$, and successful
parameter-free predictions \cite{grv92,grv94,grv90} have been given long
before HERA started to operate. Alternatively, the present data can be
fitted using the NLO AP evolution equations: Then a term $\sim x^{-\lambda}$
has to be {\em assumed\/} for the gluon distribution,
e.\ g.\ in MRSG \cite{MRSG}. Possibly this
term mimicks the BFKL behaviour. But this is not clear, since both methods
describe the data equally well.
We conclude that a detailed comparison of the standard BFKL
formalism with the new data is necessary. Only this can tell us
if BFKL can rival conventional field theoretic renormalization group
(AP) evolution equations in describing the measured structure function
$F_2$.
Our calculations are based on the methods employed by Askew, Kwieci\'nski,
Martin and Sutton (AKMS) \cite{AKMS_mod,AKMS_rev,AKMS_box},
which will be described briefly in the following.
\subsection{\label{introsub} BFKL equation and $k_T$-factorization}
The unintegrated gluon distribution $f(x,k^2)$, which is related
to the familiar integrated gluon distribution used in the AP equations
by
\newcounter{temp}
\setcounter{temp}{\value{equation}}
\stepcounter{temp}
\defA.\arabic{equation}{\arabic{temp}\alph{equation}}
\setcounter{equation}{0}
\begin{eqnarray}
\label{fg_int}
x\, g(x,Q^2)&=&\int_0^{Q^2}\,\frac{dk^2}{k^2}\, f(x,k^2),\\
\label{fg_diff}
f(x,k^2)&=&\left.\frac{\partial x\, g(x,Q^2)}{\partial\ln(Q^2)}
\right|_{Q^2=k^2},
\end{eqnarray}
\setcounter{equation}{\value{temp}}
\defA.\arabic{equation}{\arabic{equation}}
depends on the transverse momentum $k$. Using it, one can write the BFKL
equation as \cite{nBFKL}
\begin{eqnarray*}
f(x,k^2)&=&f_0(x,k^2)+\int_x^1\,\frac{dy}{y}\,\int\,dk'^2\,
K(k^2,k'^2)\, f(y,k'),\\
K(k,k')&=&\frac{3\,\alpha_s(k^2)}{\pi}\, k^2\,\left\{\frac{1}{k'^2\,
|k^2-k'^2|}-\beta (k^2)\, \delta(k^2-k'^2)\right\},\\
\beta (k^2)&=&\int\,\frac{dk'^2}{k'^2}\, \left\{\frac{1}{|k^2-k'^2|}-\frac{1}{
(4\, k'^4+k^4)^\frac{1}{2}}\right\}.
\end{eqnarray*}
One could use a suitable input $f_0$, the so called ``driving term'',
and solve the equation iteratively \cite{KMS}, but this procedure
allows no simple connection to the known AP region.
Instead we can obtain an evolution equation in $x$ from the integral
equation by differentiating with respect to $\ln (1/x)$
\begin{equation}
\label{master}
-x\,\frac{\partial\, f(x,k^2)}{\partial x}=\frac{3\,\alpha_s (k^2)}{\pi}
\, k^2\int\,\frac{dk'^2}{k'^2}\,\left[\frac{f(x,k'^2)-f(x,k^2)}{|k'^2-k^2|}+
\frac{f(x,k^2)}{(4\, k'^4+k^4)^\frac{1}{2}}\right],
\end{equation}
assuming that the derivative of the driving term $f_0$
can be neglected. Since the driving term describes the gluon content
without any BFKL evolution, it is reasonable to assume that it is
connected to the non-perturbative ``soft'' pomeron \cite{SoftP}.
Because of the soft pomeron's weak $x$ dependence $\sim x^{-0.08}$, we
expect $\partial f_0/\partial\ln (1/x)$ to be small.
Eq.\ (\ref{master}) can then be used to evolve the unintegrated gluon
to smaller $x$,
using a suitably modified AP input as boundary condition at $x_0=10^{-2}$
by applying Eq.\ (\ref{fg_int}).
An obvious problem in Eq.\ (\ref{master}) is posed by the integration
over $k'^2$, which starts at zero. Even for the AP gluon distributions
we use in this work the limit of validity is about 1 $\mbox{GeV}^2$ . We employ a simple
ansatz to continue the gluon into the infrared (IR) region,
which will be described later in this paper. A similar comment applies
to the upper limit of the $k'^2$ integration. The upper limit introduced
by energy conservation is close to infinity \cite{AKMS_mod,AKMS_rev},
so that for practical calculations an artificial ultraviolet cutoff
has to be introduced. The dependence on the IR and UV treatment will be
thoroughly discussed later on.
To obtain predictions for $F_2$, we have to convolute the (off-shell) BFKL
gluon,
i.\ e.\ the gluon ladder in Fig. \ref{leiter}, with the
photon-gluon fusion quark box $F^{(0)}$ using the $k_T$-factorization theorem
\cite{ColEl,Factor}
\begin{equation}
\label{kt}
F_i(x,Q^2)=\int\,\frac{dk'^2}{k'^4}\,\int_x^1\,\frac{dy}{y}\,
f\left(\frac{x}{y},k'^2\right)\, F_i^{(0)}(y,k'^2,Q^2),
\end{equation}
with $i=T,L$ denoting the transverse and longitudinal parts, respectively.
The expressions for $F^{(0)}$ can be found in \cite{AKMS_box}
and references therein. The BFKL prediction for $F_2$ is then simply
the sum of the calculated $F_T$ and $F_L$. Obviously here the same
problems with the $k'^2$ integration occur, which are circumvented by the
methods mentioned above.
It is important to notice that the BFKL gluons are {\em off-shell\/}
$(k^2\not=0)$. The term $F^{(0)}/k^2$ in the above equation then corresponds
to the structure function of a virtual gluon. In contrast the AP
formalism is based on {\em on-shell\/} gluons, which is a good approximation
due to the strong ordering in transverse momentum encountered in AP evolutions
in LO.
This strong ordering allows us to perform the $k'^2$ integration in
Eq.\ (\ref{kt}).
Thus, ignoring complications due to the collinear
singularities for simplicity, one arrives at the usual mass factorization
equation
\[ F_i(x,Q^2)=\int_x^1\,\frac{dy}{y}\, g\left(\frac{x}{y},Q^2\right)\,
\hat{F}_i(y,Q^2).\]
Here $\hat{F}_i$ plays the r\^{o}le of the on-shell gluon structure
function, whose $x$ dependence stems from the AP splitting function
$P_{qg}$, and $g$ is the integrated gluon.
In NLO AP evolutions the strong ordering in $k^2$ does not hold anymore due to
the emission of a second gluon, but the interacting gluon is still
considered to be on-shell in comparison with the hard scattering
scale $Q^2$.
This shows that it is inconsistent to simply feed the
evolved BFKL gluon via Eq.\ (\ref{fg_int}) back into the AP equations
below the limit set by $x_0$. Calculations attempting to use
BFKL gluons below and AP gluons above $x_0$ within the AP
formalism \cite{StroKot} ignore the essential {\em off-shellness\/}
of the BFKL gluons. This casts first doubts on a recent BFKL
analysis of the HERA H1 data using this method \cite{h1glue}.
No such problem persists when we use just this gluon mixture
to drive the general $k_T$-factorization Eq.\ (\ref{kt}), which
reduces to the mass factorization in the AP region.
Even though the dominant contribution at small $x$ should come from
the BFKL gluons, a certain amount of ``background'' in $F_2$ due to
quarks and non-perturbative effects should be taken into account.
We expect the background to be comparatively small and also to vary
much less with $x$. We shall use an ansatz motivated by the soft
pomeron $C_{I\! P}\, x^{-0.08}$, where the constant $C_{I\! P}$ is fitted to
the data.
After this general outline of our method, we will now proceed to a detailed
discussion of the underlying formalism.
\section{Suitable input for the BFKL evolution}
We focus our analysis on the gluon
distribution used in \cite{AKMS_rev}, i.e.\ a gluon
based on the MRS $D_0$-set of parton distributions \cite{d0d-},
but evolved with the leading order Altarelli-Parisi
equations \cite{Sutton}.
Since the BFKL evolution deals with an {\em unintegrated\/}
gluon distribution,
we calculate its derivative using the well-known singlet
Altarelli-Parisi equation given by
\[
f^{\mathrm{AP}}(x,Q^2)=\frac{\alpha_s(Q^2)}{2\pi}
\int_x^1 \frac{dy}{y}\left(P^{(0)}_{gg}\left(\frac{x}{y}\right)\,
g(y,Q^2)+
P^{(0)}_{gq}\left(\frac{x}{y}\right)\,\Sigma(y,Q^2)\right),
\]
where $\Sigma$ denotes the quark singlet part, and $P^{(0)}_{gg}$,
$P^{(0)}_{gq}$
the usual LO splitting functions. In the same way we produce a leading
order MRS $D_-$-type gluon, based on an input given in \cite{d-type}.
We also use a dynamically generated gluon distribution, for definiteness
the GRV '92 LO parametrization \cite{grv92}, that has the
advantage of (a) being based on an explicit LO calculation and (b) being
positive definite down to a low value of $Q^2$.
Furthermore, we take a look at the MRSA-Low $Q^2$ gluon \cite{mrsalow}, which
extends the valid $Q^2$ range down to $0.625$~$\mbox{GeV}^2$\ using an ad
hoc form-factor-like
ansatz similar to the one we employ for $f$. However, as MRSA is a
NLO analysis, there is no consistent way to implement it in our BFKL evolution,
which is neither an $\overline{\mathrm{MS}}$- nor DIS-renormalization scheme
calculation.
\subsection{Treatment of the infrared region}
As mentioned in the Introduction, we need a suitable description of the
infrared region; for our calculations we use the ansatz explained in
detail in \cite{AKMS_mod,AKMS_rev}. This ansatz introduces three
parameters:
\begin{itemize}
\item An IR-``cutoff'' $k_c^2$, i.e.\ a parameter which
separates the infrared region, where an assumption on the $k^2$-behaviour
of $f$ has to be made, from the region where the Lipatov equation
is solved numerically.
\item A parameter $k_a^2$ that controls the infrared behaviour of $f$.
For $k^2 < k_c^2$ we set:
\begin{equation}
f(x,k^2)= C' \frac{k^2}{k^2+k_a^2} f(x,k_c^2). \label{eqiransatz}
\end{equation}
The proportionality constant is given by
\[
C'=\frac{k_c^2+k_a^2}{k_c^2},
\]
to guarantee continuity at $k^2=k_c^2$.
This ansatz ensures that for $k^2\rightarrow 0$,
$f(x,k^2)\sim k^2$, as required by gauge invariance.
\item A scale $k_b^2$ where we ``freeze'' the running coupling constant,
i.e.\
\[
\alpha_s(k^2) \longrightarrow \alpha_s(k^2+k_b^2).
\]
\end{itemize}
This procedure applies also to our boundary
condition $f(x_0, k^2)$, although there is a slight modification to
the pure AP gluon in order to soften its low $k^2$ behaviour:
\begin{equation}
f^{\mathrm{AP}}(x_0,k^2) \longrightarrow f^{\mathrm{AP}}
(x_0,k^2+k_s^2) \label{eqshift}.
\end{equation}
The only purpose of the additional parameter $k_s^2$
is to ensure that we do not approach too closely a region where the gluon is
unreliable; e.g.\ the $D_0$-type gluon already approaches zero at
$Q^2=1$~$\mbox{GeV}^2$, and the $D_-$-type gluon is not defined at all for such a low
scale.
Thus,
\begin{equation}
f(x_0,k^2) = \left\{ \begin{array}{ll}
\left(\frac{k^2}{k^2+k_a^2}\right)f^{\mathrm{AP}}
(x_0,k_c^2+k_s^2) & \quad\mbox{for~} k^2 < k_c^2 \\
\left(\frac{k^2}{k^2+k_a^2}\right)f^{\mathrm{AP}}
(x_0,k^2+k_s^2) & \quad\mbox{for~} k^2 \ge k_c^2. \\
\end{array} \right. \label{eqbound}
\end{equation}
This provides an infrared behaviour according to (\ref{eqiransatz}), and for
$k^2$ large enough $f(x_0,k^2)$ approaches $f^{\mathrm{AP}}(x_0,k^2)$.
In ref.\ \cite{AKMS_mod}, the parameter $k_s^2$ is set equal to
$k_a^2$, whereas in \cite{AKMS_rev} $k_s^2$ equals $k_b^2$. We prefer setting
$k_s^2=k_a^2$ for reasons given below.
\subsection{Fixing $k_a^2$}
It seems to be clear that the parameters introduced above should be small;
so the simplest choice for these would be \cite{AKMS_rev}:
\begin{equation}
k_a^2 = k_b^2 = k_c^2 = k_s^2 = 1\mbox{~GeV}^2. \label{eqsimple}
\end{equation}
However, there is a self-consistency constraint on the choice of $k_a^2$,
depending on
$k_c^2$ and $x_0$. The boundary condition (\ref{eqbound}) should inversely be
related
to the (integrated) Altarelli-Parisi gluon $g$ in Eq.\ (\ref{fg_int}), that is
\begin{equation}
x_0\,g(x_0,Q^2)-\int_0^{Q^2} \frac{d k^2}{k^2} f(x_0,k^2)=0. \label{eqconst}
\end{equation}
In a stricter approach, we apply the shift introduced in (\ref{eqshift}) also
to the integrated gluon $g$ in this equation. Then it is not very difficult to
show that
\begin{equation}
\int_{k_c^2+k_a^2}^{Q^2+k_a^2}dl^2 \left(1+\frac{k_s^2-k_a^2}{l^2}\right)
\left.\frac{\partial x_0\,g(x_0,Q^2)}{\partial Q^2}\right|_{Q^2=
l^2+k_s^2-k_a^2}=x_0\,g(x_0,Q^2+k_s^2)-x_0\,g(x_0,k_c^2+k_s^2).
\label{eqdetks}
\end{equation}
It is obvious that this equation is fulfilled by setting $k_s^2=k_a^2$, hence
motivating our previous assumption. Using this and the asymptotic behaviour
of $g$,
\begin{equation}
x\,g(x,Q^2+k_s^2) \stackrel{Q^2\gg k_s^2}{\simeq} x\,g(x,Q^2),
\end{equation}
we may return to (\ref{eqconst}) for simplicity, which gives us upon
inserting our boundary condition (\ref{eqbound})
\begin{eqnarray*}
x_0\,g(x_0,Q^2) & = &
\int_0^{k_c^2} \frac{d k^2}{k^2+k_a^2} f^{\mathrm{AP}}(x_0,k_c^2+k_a^2)
+ \int_{k_c^2}^{Q^2} \frac{d k^2}{k^2+k_a^2} f^{\mathrm{AP}}(x_0,k^2+k_a^2) \\
& \simeq & f^{\mathrm{AP}}(x_0,k_c^2+k_a^2)\ln\frac{k_c^2+k_a^2}{k_a^2}+
x_0\,g(x_0,Q^2)-x_0\,g(x_0,k_c^2+k_a^2).
\end{eqnarray*}
Thus we obtain an implicit equation for $k_a^2$:
\begin{equation}
\ln\frac{k_c^2+k_a^2}{k_a^2}\simeq\frac{x_0\,g(x_0,k_c^2+k_a^2)}
{f^{\mathrm{AP}}(x_0,k_c^2+k_a^2)}. \label{eqestka}
\end{equation}
One can see that the value of $k_a^2$ mainly depends
on the ratio $g/f$ in the region of low $Q^2$, and as the variation of $f$ for
different parton distributions is relatively small in that region compared to
the variation of $g$, we conclude that it is basically the absolute value of
$g$ that determines the size of $k_a^2$.
Hence, as a rule of thumb, the higher
the value of $x_0\,g(x_0,Q^2)$ for small $Q^2$, the smaller the resulting
$k_a^2$. It should be emphasized that, although the shift in Eq.\
(\ref{eqshift})
is taken into account in Eq.\ (\ref{eqdetks}) for both $f$ and $g$, but in
Eq.\ (\ref{eqestka}) only for $f$, the
solution of Eq. (\ref{eqestka}) provides a good estimate on $k_a^2$,
very close to the value we get by minimizing the l.h.s.\ of Eq.\
(\ref{eqconst}) for $Q^2>100$~$\mbox{GeV}^2$.
The result for all the parton distributions under consideration is given in
Table \ref{tbparam}, together with the value of
$\Lambda_{\mathrm{QCD}}$\ for four flavors used with each distribution.
Figure \ref{bound} shows the unmodified gluons and the modified boundary
conditions according to Table \ref{tbparam}.
Regarding the $D_0$-type gluon, we see that the optimized value of
$k_a^2=0.95$~$\mbox{GeV}^2$\ is indeed very close to the na\"\i ve estimate of 1~$\mbox{GeV}^2$, as
was already noticed in \cite{AKMS_rev}.
It was also mentioned there that a reasonable choice of
$k_a^2$ should lie in the
range of $0.5 - 2$~$\mbox{GeV}^2$, and we see that the $D_-$-type value lies well
within this range,
while the MRSA set is already close to its lower edge. The most extreme
boundary condition in this respect is derived from the GRV parametrization,
with a value of only $0.19$~$\mbox{GeV}^2$.
\begin{table}
\centering
\begin{tabular}{|c|c|c|c|c|}
\hline
\ostrut\ustrut} \def\uostrut{\oustrut Set & $\Lambda_{\mathrm{QCD}}^{(4)}$~[MeV] & $x_0$ &
$k_c^2$~[$\mbox{GeV}^2$] & $k_a^2$~[$\mbox{GeV}^2$] \\
\hline
\hline
\ostrut\ustrut} \def\uostrut{\oustrut $D_0$-type & 173.2 & 0.01 & 1.0 & 0.95 \\
\ostrut\ustrut} \def\uostrut{\oustrut $D_-$-type & 230.4 & 0.01 & 1.0 & 1.51 \\
\ostrut\ustrut} \def\uostrut{\oustrut GRV '92-LO & 200.0 & 0.01 & 1.0 & 0.19 \\
\ostrut\ustrut} \def\uostrut{\oustrut MRSA-Low $Q^2$ & 230.0 ($\overline{\mathrm{MS}})$ & 0.01 &
1.0 & 0.44 \\
\hline
\end{tabular}
\caption{Parameters for various sets of parton distributions}
\label{tbparam}
\end{table}
As the more recent parton distributions favor a larger gluon $g$, the
assumption of $k_a^2=1$~$\mbox{GeV}^2$\ (a good choice for the relatively small
$D_0$-type gluon) does not appear to be the best choice for these gluons.
Let us now emphasize the importance of the constraint (\ref{eqconst}) on
$k_a^2$: Starting with the simple assumption (\ref{eqsimple}) that all the
parameters introduced should be equal to 1~$\mbox{GeV}^2$, one gets a slope
\[
\lambda = \left.\frac{1}{f}\frac{\partial f}{\partial\ln(1/x)}
\right|_{x=10^{-4}} = 0.5 - 0.6
\]
for the unintegrated, BFKL evolved gluon distribution,
as expected, no matter if $D_0$-type, $D_-$-type, or GRV is chosen
as input for the BFKL evolution. If we deviate each of the infrared parameters
from (\ref{eqsimple}), we see that it is $k_a^2$ which has the biggest impact
on $\lambda$. With $k_a^2$ decreasing, the slope rises, resulting in a slope
$\lambda\simeq 0.9$ for GRV partons ($k_a^2=0.19$~$\mbox{GeV}^2$) as the extreme limit.
Keeping these considerations in mind, we will now concentrate our analysis on
the structure functions. We will demonstrate that the slopes of $f$ and
$F_2$ are related in such a way that for large $\lambda$ our calculated
$F_2$ is too steep to match it to the recent HERA data.
\section{Varying the parameters}
We construct the boundary condition at $x_0=10^{-2}$.
The consistency constraint (\ref{eqconst}) determines $k_a^2$ and
as a standard value for the other two IR parameters we choose $k_b^2=k_c^2=
1\,\mbox{GeV}^2$. The UV cutoff is set to $10^4\, \mbox{GeV}^2$
and we stay with the $\Lambda_{\mathrm{QCD}}$\ of the AP partons. We have written an evolution
program to solve (\ref{master}) iteratively.
Below $k_c^2$ the necessary
integration can easily be done analytically, above we use
the Gau{\ss}-Legendre quadrature. This transforms the integro-differential
equation (\ref{master}) into a set of coupled differential equations,
allowing us to use the standard Runge-Kutta method to calculate the
evolution. The evolved gluon below $x_0$ is combined with the unintegrated
AP gluon above $x_0$ to obtain predictions
for $F_T$ and $F_L$ by performing the integration in Eq.\ (\ref{kt}) with
Monte-Carlo methods. For
convenience a fit of $F_2=F_T+F_L$ is given
in the Appendix. Finally we fit the background to the data, as discussed at
the end of Section \ref{introsub}, and obtain our BFKL prediction.
It is vital to check the dependence of the results on the parameters
used for the IR and UV treatment. In Fig. \ref{vary} we varied all relevant
parameters, using the $D_0$-type gluon. All curves have been calculated
at $Q^2=15\,\mbox{GeV}^2$ and are already fitted to the shown HERA
data with the soft pomeron background $C_{I\! P}\, x^{-0.08}$.
The strong dependence on $k_a^2$, which we expect from the
corresponding variation in the gluon slope, is obvious. The curves for
low $k_a^2$ are much too steep. At $k_a^2=0.2$ $\mbox{GeV}^2$\
for example, the pure BFKL prediction is too high, even without any background.
A background fit would then give a {\em negative\/} contribution, i.\ e.\
$C_{I\! P}<0$,
which is unphysical. In all such cases we set the background to zero.
It is crucial, that $k_a^2$ can be precisely determined from the
consistency constraint. Without the constraint, we could vary the slope
of $F_2$ from 0.5 to 0.8 by choosing $k_a^2$ within the shown range
of 0.2 to 2.0 $\mbox{GeV}^2$ ,
rendering any serious prediction impossible. In the already mentioned
comparison of a BFKL calculation with HERA data \cite{h1glue}, $k_a^2$
is treated as a {\em free\/} parameter. If the fitted $k_a^2$ should not
be close to the consistent value by chance, it is doubtful that
any strong conclusions can be drawn from a successful description of the data.
The influence of the IR cutoff $k_c^2$ is comparably small. It should
be kept in mind that $k_a^2$ has to be fitted separately for each $k_c^2$.
Actually the induced variations in $k_a^2$ dampen the dependence
of the predictions on $k_c^2$ slightly. It is no surprise, that
variations of the UV cutoff do not introduce much uncertainty into the
predictions, as the running coupling already serves as an effective
UV cutoff \cite{AKMS_mod,AKMS_rev}. The remaining free IR parameter,
$k_b^2$, has a sizeable effect on the curves. The slope of $F_2$ varies from
0.53 to 0.58 in the shown $Q^2$ range. The standard value of 1 $\mbox{GeV}^2$\
gives a slope of about 0.55, and the variations of $k_b^2$ represent an
effective error-band of our calculations.
A considerable dependence on $\Lambda_{\mathrm{QCD}}$\ is expected and can be
seen in Fig. \ref{vary}. Fortunately this parameter is fixed, since we are
using AP gluons with a given $\Lambda_{\mathrm{QCD}}$\ as boundary condition. It is
interesting to note, that higher values of $\Lambda_{\mathrm{QCD}}$\ than the rather
low 173.2 MeV used in the $D_0$-type partons give {\em un\/}favourably steep
slopes! Finally we take a look at variations of $x_0$, noting that
we have to determine $k_a^2$ separately again.
We expect the steep curve for $x_0=5\cdot 10^{-2}$ due to the long
evolution length in $x$. But it is daring to use the BFKL equations at
such high values of $x$ and the data do not support such a choice.
We can also see that the comparably flat curve created
by a short evolution in $x$ starting from $x_0=10^{-3}$ does fit the data
well. But this success can obviously not be claimed by the BFKL evolution,
since the data do not extend far below that $x_0$. In order to test if
the BFKL equations can describe the {\em current\/} data, we have to choose
a larger $x_0\sim 10^{-2}$. The typical steep BFKL slope simply needs enough
evolution length to develop. Small variations
to lower $x_0$ from the usually used $10^{-2}$ do not affect the
calculations strongly. The standard values \cite{AKMS_mod,AKMS_rev} used for
$k_b^2$ and $x_0$ give slopes slightly below and above the
average expected from the variations, respectively.
Thus this choice of parameters
will give a sensible prediction while still allowing comparisons with
earlier calculations \cite{AKMS_mod,AKMS_rev}.
The results presented in this sections show that the dependence
of the calculations on the parameters is under control. This
statement would be impossible, if $k_a^2$ did not obey the
consistency constraint (\ref{eqconst}).
\subsection{Using other input distributions}
Besides our detailed analysis of the $D_0$-type gluon, we also studied
the effect of feeding different parton distributions into our evolution,
namely GRV'92 LO \cite{grv92} and MRSA-Low~$Q^2$ \cite{mrsalow}. The results
for $F_2$ can be seen in Fig.\ \ref{mrsa}.
Obviously by using GRV and MRSA gluons as inputs for the BFKL evolution, a
structure function is generated which is far too steep
for the data. Even though we have got, in principle, some freedom in fitting
an appropriate background, this is useless here, since already the pure
BFKL part of $F_2$ is too high for the HERA data, so that we must set
the background to zero.
The steepness of $F_2$ is closely related to the fact that the proper
$k_a^2$ is much smaller than 1~$\mbox{GeV}^2$, resulting in a larger
slope $\lambda$ of the BFKL gluon $f$. As we explained above, this is mainly
an effect of the size of the integrated gluon input $g$. This suggests the
conclusion that the
whole BFKL procedure has only a chance to work with the older (smaller) gluons.
Modern parton distributions imply small values of
the crucial infrared parameter $k_a^2$,
and since this has a strong effect on the calculated structure functions,
it is likely that one does not succeed in matching these to present
experimental data.
\section{Comparison with data}
In Fig. \ref{e6pre} we compare our calculations with very low $Q^2$ data from
the Fermilab E665 experiment \cite{E665}.
Preliminary data of the HERA $ep$ collider \cite{HERA_pre} at low $Q^2$
are also shown in this figure. The published 1993 HERA data \cite{HERA_93}
for low to medium $Q^2$ are presented in Fig. \ref{hera} together with our BFKL
predictions. All graphs show the BFKL calculations based on the $D_0$-type
and the $D_-$-type partons, as well as the latest dynamical NLO renormalization
group predictions (GRV), as
presented in \cite{grv94}. Also shown is the soft pomeron background,
which is included in the $D_0$-type curve. A general feature of all
figures is that the difference between the $D_0$-type and the $D_-$-type
BFKL predictions is very small after fitting the background.
For this reason we do not discuss them separately. We have
checked, that the $R=F_L/F_T$ values used to extract $F_2$ from
the experimental data are close enough to those predicted by BFKL.
Thus it is not necessary to
reanalyze the data in terms of $R_{\mathrm{BFKL}}$.
We first turn our attention to the E665 data \cite{E665}.
The data do not extend very far into the small $x$ region, making
a check of the BFKL behaviour difficult. On the other hand the difference
between the GRV (AP) and BFKL predictions is potentially large at small $x$.
We also notice that the added background is comparable in size to the
BFKL part at higher $x$. This flattens the steep BFKL behaviour, giving
a good description of the data. But it is just the large contribution
of the background which makes the very procedure used doubtful.
A further hint, that the successful description of the E665 data should
not be taken too seriously, is provided by comparing the curves for
the E665 data at 2.8 $\mbox{GeV}^2$\ and those for preliminary ZEUS
data at 3.0 $\mbox{GeV}^2$ . While the E665 background is strong $(C_{I\! P}=0.189)$
for $D_0$-type, the optimal ZEUS background would be negative
$(C_{I\! P}=-0.172)$
and therefore is set to zero. Thus the natural requirement, that the BFKL
predictions for similar $Q^2$ values should be approximately equal,
is only fulfilled at very small $x$. For $x\gtrsim 10^{-4}$ the
influence of the background quickly becomes stronger and the curves deviate.
We conclude that BFKL can {\em not\/} describe both data sets consistently.
It is also obvious from the figures that just this is possible using the
usual NLO renormalization group equations, see the curves labeled GRV '94
in Fig.\ \ref{e6pre}.
Turning to the preliminary HERA data in Fig. \ref{e6pre},
we find that the BFKL slope of $F_2$ is evidently too steep. The same
tendency can also be found in the 1993 HERA data up to approximately
15 $\mbox{GeV}^2$\ as shown in Fig. \ref{hera}. Everywhere
in this region the background is very small or would even be negative,
if it were allowed. Due to the larger spread in $x$, the preliminary
data show the discrepancy rather clearly. The problem is rooted in
the almost $Q^2$ independent slope of BFKL, which is always $\gtrsim 0.5$
and grows weakly with $Q^2$. The dynamical GRV (AP) calculations give on
the contrary
flat slopes at low $Q^2$, which rise significantly with larger $Q^2$.
This is obviously in much better agreement with the data.
For example, using the preliminary ZEUS data at 4.5 $\mbox{GeV}^2$ ,
we get a total $\chi^2$ of 15.3 ($D_0$-type), 15.0 ($D_-$-type)
and of 2.81 (GRV '94 NLO) for the four data points.
At $Q^2$ higher than 15 $\mbox{GeV}^2$ , we see that
the slope of the data becomes compatible with the one predicted
by BFKL. The rising slope of the dynamical GRV prediction fits the data
at least as well. It is interesting to note that even at higher
$Q^2$ a slight extension in the $x$ range could provide an indication
which evolution should be used. If future data should conform to the
already visible tendency that a Lipatov-like slope is only obtainable at
high $Q^2$, say $\gtrsim 50$ $\mbox{GeV}^2$ , then the simple LO BFKL
formalism would become implausible, since its validity at such high $Q^2$ is
questionable.
\section{Conclusions}
We have shown that the AKMS method for calculating BFKL predictions
of $F_2$ remains stable under variations of the introduced parameters,
if the consistency constraint Eq.\ (\ref{eqconst}) on $k_a^2$ is applied.
As boundary
conditions one has to choose older, i.\ e.\ smaller, gluon densities, since
the large recent AP gluons lead to small $k_a^2$, which in turn produces
overly steep slopes of $F_2$. The BFKL boost at small $x$ is simply
too large for gluons constructed to produce the measured large $F_2$
slope via the conventional renormalization group (AP) evolution equations.
Thus we use the old $D_0$-type and $D_-$-type gluons with
consistently fixed $k_a^2$ as input for the BFKL evolution.
A further complication is introduced by
the necessity to add a background contribution to the BFKL prediction for
$F_2$. Then it seems
reasonable to require that the main growth of $F_2$ is not driven
by the chosen background, and that this background is comparably small. The
last condition is not fulfilled in the region of the E665 experiment,
casting serious doubts on the good agreement with the data.
In the HERA region we find in contrast small background contributions,
allowing for
reliable comparisons with experiment. Especially the preliminary HERA
data in Fig. \ref{e6pre} show that
the almost $Q^2$ independent BFKL slope is too steep for $Q^2\lesssim 15$
$\mbox{GeV}^2$ . But even up to the expected limit of applicability in $Q^2$ of
the BFKL evolution we find that the predicted slope is somewhat too steep.
With improved statistics and maybe a slightly extended coverage
in $x$, HERA should be able to assess the LO BFKL predictions for
$F_2$. If the tendency visible in the current data is an
indication for future developments, we expect LO BFKL to fail the test.
It remains to be seen whether extensions of LO BFKL --- inclusion of the
quark sector, NLO BFKL or theoretically consistent (energy-momentum
conservation, etc.\ ) unified evolution equations --- can
improve the agreement with the data.
Our results also indicate that conventional (dynamical) renormalization group
evolutions are still the best method for calculating and analyzing $F_2$.
\section{Acknowledgments}
We are grateful to E.\ Reya and M.\ Gl\"uck for helpful guidance
and fruitful discussions, and to E.\ Reya, M.\ Stratmann and S.\ Kretzer
for carefully reading the manuscript. We thank P.\ J.\ Sutton for making
the $D_0$-type partons available to us. I.\ B.\ wishes
to thank D.\ Wegener for unbureaucratically granting additional
computing time on his workstations.
This work has been supported in part by the 'Bundesministerium f\"ur Bildung,
Wissenschaft, Forschung und Technologie', Bonn.
\section*{Appendix: Parametrization}
\setcounter{equation}{0}
\defA.\arabic{equation}{A.\arabic{equation}}
It should be convenient to have a simple parametrization of our theoretical
results for $F_2$.
The following is a parametrization of the BFKL part only,
to which an appropriate background still has to be added.
We use an ansatz of the form:
\begin{equation}
\label{ansatz}
F_2^{BFKL}=\alpha\, x^{-\lambda}+\beta\, x+\gamma,
\end{equation}
which describes all curves shown in Figs. \ref{vary} -- \ref{hera} well.
It is even
possible to parametrize {\em all\/} pure BFKL results for the $D_0$-type
and $D_-$-type gluons shown in Figs. \ref{e6pre} and \ref{hera} by
choosing the following $Q^2$ dependence
of the coefficients [$t=\ln (Q^2/\Lambda_{\mathrm{QCD}}^2)$]
\begin{eqnarray}
\label{anskoeff}
\alpha (Q^2)&=&A+B\, t+C\, t^2+D\, t^3,\nonumber\\
\beta (Q^2)&=&E\, t^2+F\, t^3,\\
\gamma (Q^2)&=&G+H\, (Q^2/\mathrm{GeV}^2) +I\, t,\nonumber\\
\lambda (Q^2)&=&J+K\, t.\nonumber
\end{eqnarray}
The corresponding coefficients are given in Table \ref{koeff}.
It is interesting to note that a small growth of $\lambda$ with $Q^2$
has to be taken into account.
\begin{table}
\begin{center}
\begin{tabular}{|c||r|r|}
\hline
\ostrut\ustrut} \def\uostrut{\oustrut&\multicolumn{1}{c|}{$D_0$-type}&\multicolumn{1}{c|}{$D_-$-type}\\
\hline\hline
$\ostrut\ustrut} \def\uostrut{\oustrut A$&$4.818\cdot 10^{-3}$&$8.07\cdot 10^{-4}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut B$&$-2.460\cdot 10^{-3}$&$4.1\cdot 10^{-5}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut C$&$7.386\cdot 10^{-4}$&$1.579\cdot 10^{-4}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut D$&$3.32\cdot 10^{-6}$&$4.370\cdot 10^{-5}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut E$&$-2.23\cdot 10^{-1}$&$-1.89\cdot 10^{-1}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut F$&$1.15\cdot 10^{-2}$&$1.86\cdot 10^{-2}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut G$&$-5.3\cdot 10^{-3}$&$-1.46\cdot 10^{-2}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut H$&$-6.66\cdot 10^{-4}$&$-8.72\cdot 10^{-4}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut I$&$6.54\cdot 10^{-3}$&$1.026\cdot 10^{-2}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut J$&$-5.113\cdot 10^{-1}$&$-5.700\cdot 10^{-1}$\\ \hline
$\ostrut\ustrut} \def\uostrut{\oustrut K$&$-7.10\cdot 10^{-3}$&$-2.89\cdot 10^{-3}$\\ \hline
\ostrut\ustrut} \def\uostrut{\oustrut $\Lambda_{\mathrm{QCD}}$ & 173.2 MeV & 230.4 MeV \\ \hline
\end{tabular}
\caption{\label{koeff} Coefficients defined in (\ref{anskoeff}) of the
parametrization (\ref{ansatz})}
\end{center}
\end{table}
It has to be stressed, that this parametrization is {\em only\/}
valid within the range $10^{-5}\le x\le 10^{-2}$ and 0.8 $\mbox{GeV}^2$\
$\le Q^2\le$ 120 $\mbox{GeV}^2$ . On the one hand BFKL is not expected to be applicable
even at the edges of this region. On the other hand we note, that the
form of the ansatz has been tailored for this region only. The term
$\beta\, x$, for example, will lead to wrong results for $x>10^{-2}$.
\clearpage
|
\section{Three-dimensional $\pi N\!N$ equations}
Quantum field theory requires the use of four dimensions to assure manifestly
covariant descriptions. However, solving four-dimensional equations presents
an
enormous numerical task, and consequently three-dimensional formulations of the
$\pi N\!N$\ system are most desirable and indeed have been the most numerous.
Until recently, perhaps the most sophisticated few-body description of the
$\pi N\!N$\ system has been the ``unitary $N\!N\!-\pi N\!N$\ model'' \ctt{Avishai}{AB_80}.
This
is a field-theoretic model based on time-ordered perturbation theory, it takes
into account pion absorption, and has the desirable properties of two- and
three-body unitarity. The essential feature of the model is that it describes
all the processes $\pi d\rightarrow \pi d$, $\pi d\rightarrow \pi N\!N$, $N\!N
\rightarrow \pi N\!N$, $N\!N \rightarrow \pi d$, and $N\!N \rightarrow N\!N$
all
within the one set of coupled equations.
The derivation of the unitary $N\!N\!-\pi N\!N$\ equations is based upon truncating
Hilbert space to states of at most one pion. In practise, this means retaining
all diagrams contributing to subsystem $\pi N$\ and $N\!N$\ potentials, but
neglecting
all other diagrams having two or more pions in an intermediate state. Many
calculations have been performed with the $N\!N\!-\pi N\!N$\ model \ctt{AB_81}{Mizutani}.
In general, one can say that the model can account for an extensive amount of
data, albeit only in a qualitative way.
\bigskip
\noindent{\em 2.1\hspace{2mm} The renormalisation problem}
\bigskip
\begin{figure}[b]
\hspace{1cm} \epsffile{figz.ps}
\caption{\fign{flaw} Allowed dressing in the unitary $N\!N\!-\pi N\!N$\ model, with
associated $Z$ renormalisation factors. (a) $\pi N$\ nucleon pole graph, (b) $N\!N$\
OPE graph.} \end{figure}
Despite the modest successes of the unitary $N\!N\!-\pi N\!N$\ model, it has become clear
that the model itself has a serious theoretical
inconsistency \cite{Sauer,FewBody_Adelaide}. The origin of the problem lies in
the truncation of Hilbert space used to derive the $N\!N\!-\pi N\!N$\ equations. This
truncation has serious consequences for the renormalisation of both the
two-nucleon propagator and the $\pi N\!N$\ vertex. In \fig{flaw}(a) we show the
$\pi N$\ nucleon pole diagram where the intermediate state nucleon is dressed by
one-pion loops; however, the initial and final state nucleons do not include
dressing since two-pion states are neglected in the truncation. Since close to
the nucleon pole the dressed one-nucleon propagator is of the form $g(E) \sim
Z/(E-m)$, where $Z$ is the residue at the pole, \fig{flaw}(a) illustrates how
each $\pi N\!N$\ vertex $f(E)$ gets effectively renormalised by a factor of
$Z^{1/2}$. Thus $f_{\pi N\!N} = Z^{1/2}f(m)$ is essentially the $\pi N\!N$\ coupling
constant, and this fact is used to fix the strength parameter in the form
factor
$f(E)$. With all other parameters of $f(E)$ fixed to reproduce experimental
$\pi N$\ phase shifts, this form factor then enters the unitary $N\!N\!-\pi N\!N$\ equations
as an input. As illustrated in \fig{flaw}(b), when the $N\!N$\ one pion exchange
(OPE) amplitude is calculated in the unitary $N\!N\!-\pi N\!N$\ model, the initial and
final nucleons are dressed by pions and consequently each external nucleon
obtains a renormalisation factor of $\tilde{Z}^{1/2}$. The first
renormalisation
problem is the fact that $\tilde{Z}\ne Z$. This arises because two nucleons
cannot be dressed at the same time in the truncated Hilbert space; thus, each
nucleon in a two-nucleon state cannot obtain its full dressing. This, however,
may not be such a serious problem since, in practice, the difference between
$Z$
and $\tilde{Z}$ turns out to be quite small. The serious problem, instead, is
the size of the effective $\pi N\!N$\ coupling constant in the $N\!N\!-\pi N\!N$\
equations. Taking $Z\approx \tilde{Z}$, \fig{flaw}(b) illustrates that each
vertex gets renormalised by a factor of $Z$, so that the effective $\pi N\!N$\
coupling constant here becomes $Zf(m)$; this is a factor $Z^{1/2}$ times the
physical coupling constant. With $Z$ being typically between $0.6$ and $0.8$,
we
come to the disturbing conclusion that the effective $\pi N\!N$\ coupling constant
in
the $N\!N\!-\pi N\!N$\ equations is smaller than the one used in constructing the $\pi N$\
input. This observation helps explain why one typically obtains much too small
$pp\rightarrow\!\pi^+ d$\ cross sections using this model \ctt{Rinat2}{Fayard}.
\newpage
\noindent{\em 2.2\hspace{2mm} The $\pi N\!N$\ convolution equations}
\bigskip
Here we describe a new formulation of the $\pi N\!N$\ problem where unitary
equations
are obtained without having to truncate the Hilbert space to some maximum
number
of pions. Consequently, all possible dressings of one-particle propagators and
vertices are retained in our model. The essential technique that enables the
calculation of all such dressings is a novel use of convolution integrals.
In this way we overcome the renormalisation problems discussed above.
As an explicit derivation of the new $\pi N\!N$\ equations can be found in
Ref.\cite{KB_PL}, here we prefer to simply state the final equations and to
describe their essential features. The new $\pi N\!N$\ equations can be expressed in
many different forms, all of which are equivalent. The form we shall choose
here
is the one that most closely resembles the unitary $N\!N\!-\pi N\!N$\ equations as given
by Afnan and Blankleider\cite{AB_80} (AB). Choosing this AB form has two
essential advantages: firstly, we are able to directly compare the differences
between our $\pi N\!N$\ convolution equations and the unitary $N\!N\!-\pi N\!N$\ equations,
and
secondly, this form is ideal for numerical solution, especially since advantage
may be taken of existing codes for the unitary $N\!N\!-\pi N\!N$\ equations where
essentially the AB form has been used. We note, however, that for the sake of
easy comparison, we give the equations here only for the case of
distinguishable
nucleons. Including proper antisymmetry of the nucleons is a technical
formality
which, for the $AB$ form, will be presented elsewhere (in this regard, note
that
the derivation of the four-dimensional $\pi N\!N$\ equations presented in Ref.\
\cite{KB_NP}, assumes identical nucleons from the beginning).
The $\pi N\!N$\ convolution equations may be expressed as a set of coupled equations
for the reactions $N\!N\!\rightarrow\!N\!N$, $N\!N\!\rightarrow\!\pi d$, $\pi d\!\rightarrow\!N\!N$, and $\pi d\!\rightarrow\!\pi d$\ using the following
($4\times 4$) matrix form:
\begin{equation} \left(
\begin{array}{cc} T_{N\!N} & \bar{\underline{T}}_{N} \\
\underline{T}_{N} & \underline{T} \end{array} \right)= \left( \begin{array}{cc}
V_{N\!N} & \bar{\underline{F}} \\
\underline{F} & G_0^{-1}\underline{\cal{I}} \end{array} \right)\left\{ I+
\left(
\begin{array}{cc} G_{N\!N} & \underline{0} \\
\underline{0} & G_0 \underline{w}^0 G_0 \end{array} \right) \left(
\begin{array}{cc} T_{N\!N} & \bar{\underline{T}}_{N} \\
\underline{T}_{N} & \underline{T} \end{array} \right) \right\}. \eqn{AB}
\end{equation}
Before explaining the symbols in this equation, let us define what we mean by a
product of two symbols. For any two quantities $B$ and $A$, describing
processes
$m\rightarrow k$ and $k\rightarrow n$, respectively, we define the product
symbol $AB$ to mean the the integral
\begin{equation}
AB \equiv \int d{\bf p}''_1\ldots d{\bf p}''_k \,
A({\bf p}'_1\ldots{\bf p}'_n,{\bf p}''_1\ldots{\bf p}''_k;E)
\, B({\bf p}''_1\ldots{\bf p}''_k,{\bf p}_1\ldots{\bf p}_m;E) \eqn{product}
\end{equation}
where ${\bf p}_i$ is the three-momentum of particle $i$ and $E$ is the total energy.
Although, momentum conserving $\delta$-functions are assumed to be contained in
both $A$ and $B$, it is easy to see that such $\delta$-functions can be
factored
out without affecting the symbolic equations. In \eq{AB} the unknown quantities
are $T_{N\!N}$, together with $T_{\alpha N}$, $T_{N \beta}$, and $T_{\alpha\beta}$ which are
elements of the matrices $\underline{T}_N$, $\bar{\underline{T}}_{N}$, and
$\underline{T}$, respectively (here indices $\alpha$ and $\beta$ take on values $1$,
$2$, and $3$). The physical amplitudes for $N\!N\!\rightarrow\!N\!N$, $N\!N\!\rightarrow\!\pi d$, $\pi d\!\rightarrow\!N\!N$, and $\pi d\!\rightarrow\!\pi d$,
are then given by
\begin{eqnarray}
\begin{array}{ccccccc}
X_{N\!N} = T_{N\!N} & ; &
X_{dN} = \bar{\Psi}_dT_{3N} & ; &
X_{Nd} = T_{N3}\Psi_d & ; &
X_{dd} = \bar{\Psi}_d T_{33}\Psi_d ,
\end{array}
\end{eqnarray}
respectively, where $\Psi_d$ is the deuteron wave function in the presence of a
spectator pion. On the r.h.s.\ of \eq{AB} $G_0$ is the fully dressed $\pi N\!N$\
propagator, $G_{N\!N}$ is the fully dressed $N\!N$\ propagator,
$\underline{\cal{I}}$ is a $3\times 3$ matrix whose elements are
$\bar{\delta}_{\alpha\beta}=1-\delta_{\alpha\beta}$, and $V_{N\!N}$ is the dressed one-pion
exchange potential given by
$
V_{NN}=\sum_{i,j=1}^{2}\bar{F}_i\bar{\delta}_{ij}G_0 F_j
$
where $F_i$ and $\bar{F}_i$ are fully dressed $\pi N\!N$\ vertices in the
two-nucleon
sector as illustrated in \fig{F}. Finally we have the matrices
\vspace{-5mm}
\begin{eqnarray}
\begin{array}{ccccc}
\underline{F}=\left( \sum_{j=1}^2 \bar{\delta}_{\alpha j}F_j \right) & ; &
\bar{\underline{F}}=\left( \sum_{i=1}^2\bar{F}_i \bar{\delta}_{i\beta} \right)& ;
&
\underline{w}^0 =\left( \begin{array} {ccc}
w^0_1 & w^0_4 & 0 \\
w^0_5 & w^0_2 & 0 \\
0 & 0 & w^0_3 \end{array} \right)
\end{array}
\end{eqnarray}
where the $w^0_\alpha$ ($\alpha=1\ldots 5$) are the disconnected $N\!N$-irreducible
amplitudes for $\pi N\! N\!\rightarrow\!\pi N\!N$, to be discussed shortly.
\begin{figure}[t]
\hspace{3cm} \epsffile{figf.ps}
\caption{\fign{F} The vertices $F_1$ and $\bar{F}_1$. The dark circles
represent all possible intermediate states consistent with the requirement that
$F_1$ and $\bar{F}_1$ be amplitudes with chopped external legs. Vertices
$F_2$ and $\bar{F}_2$ are obtained by interchanging 1 and 2.} \end{figure}
\begin{figure}[b]
\hspace{1cm} \epsffile{figw.ps}
\caption{\fign{w} The amplitudes $w_{\alpha}$. The dark circles represent all
possible intermediate states consistent with the requirement that the
$w_{\alpha}$ be amplitudes with chopped external legs. Amplitudes $w_2$ and
$w_5$ are obtained by interchanging the two nucleons in $w_1$ and $w_4$
respectively.}
\end{figure}
By form, \eq{AB} is very similar to the unitary $N\!N\!-\pi N\!N$\ equations as given in
Eq.\ (59) of AB. However, the essential feature of \eq{AB} that distinguishes
it
from the unitary $N\!N\!-\pi N\!N$\ equations, is that all input quantities in \eq{AB}
are
fully dressed. In this way the renormalisation problems of the $N\!N\!-\pi N\!N$\
equations have been overcome. However this would only be a formal solution to
the renormalisation problem if it were not for the fact that all the necessary
dressings can be calculated exactly using convolution integrals. That this is
so
follows from Ref.\cite{KB_PRC} where we showed that any disconnected Green
function is equal to the convolution of all its disconnected parts; thus for
example, the dressed two-nucleon propagator $G_{N\!N}$ is expressed in terms of
the dressed one-nucleon propagators $g_1$ and $g_2$ as
\begin{equation}
G_{N\!N}(E) = -\frac{1}{2\pi i}\int_{-\infty}^{\infty}dz\,g_1(E-z)g_2(z)
\eqn{G}
\end{equation}
where, for the sake of simplicity, we have set the momenta of the nucleons to
zero. To further save on notation, we introduce the shorthand
$G_{N\!N}=g_1\otimes g_2$ to mean the convolution integral of \eq{G}. Giving
labels 1 and 2 to the two nucleons, and label 3 to the pion, in the same way we
have that the fully dressed $\pi N\!N$\ propagator $G_0$ is given by the double
convolution
$
G_0 = g_1 \otimes g_2 \otimes g_3 .
$
To see how the amplitudes $w_\alpha^0$ are calculated, we first define the
amplitudes $w_\alpha$ to be the disconnected $\pi N\! N\!\rightarrow\!\pi N\!N$\ amplitudes, illustrated in
\fig{w}, each corresponding to a different type of disconnectedness, and
containing {\em all} possible contributing diagrams. It is just because the
$w_\alpha$ contain all possible contributions that one can express them through
convolution integrals as
\begin{eqnarray}
\begin{array}{ccccccccc}
\tilde{w}_1 = \tilde{t}_1 \otimes g_2 & ; &
\tilde{w}_2 = \tilde{t}_2 \otimes g_1 & ; &
\tilde{w}_3 = \tilde{t}_3 \otimes g_3 & ; &
\tilde{w}_4 = \tilde{f}_1 \otimes \tb{f}_2 & ; &
\tilde{w}_5 = \tilde{f}_2 \otimes \tb{f}_1 \eqn{w_convolutions}
\end{array}
\end{eqnarray}
where the ``tilde'' denotes a Green function quantity consisting of the
corresponding amplitude with additional initial and final-state propagators;
thus, for example, $\tilde{w}_\alpha=G_0w_\alpha G_0$, and $\tilde{t}_1 = g_{\pi N_1}
t_1 g_{\pi N_1}$ where $t_1$ is the t-matrix and $g_{\pi N_1}$ the dressed
propagator for scattering of a pion off nucleon 1. As we have shown in
Ref.\cite{KB_PRC}, the convolution integrals effectively sum over all the
relative time orderings of one subamplitude of a disconnected diagram with
respect to another. Similarly, the vertices $F_1$ and $F_2$ are also
expressed via convolutions as
\begin{equation}
\tilde{F}_1 = \tilde{f}_1\otimes g_2\hspace{1cm};\hspace{1cm}
\tilde{F}_2 = \tilde{f}_2\otimes g_1.
\end{equation}
Once the $w_\alpha$ are calculated, we may then write them as
$
w_\alpha = w_\alpha^0 + w_\alpha^P \eqn{w_A}
$
where $w_\alpha^P$ is the part of $w_\alpha$ that is two-nucleon reducible, while
$w_i^0$ is two-nucleon irreducible. Since we
consider all possible contributions, it is clear that ($i=1,2$)
$w_i^P = F_i G_{N\!N} \bar{F}_i$, $w_3^P = 0$, $w_4^P = F_1 G_{N\!N}
\bar{F}_2$,
and $w_5^P = F_2 G_{N\!N} \bar{F}_1.$
In this way, all the essential input to \eq{AB} has been specified.
We may finally note a second major difference between \eq{AB} and the unitary
$N\!N\!-\pi N\!N$\ equations. The input matrix $\underline{w}^0$ in \eq{AB} has
off-diagonal elements, while the corresponding matrix for the $N\!N\!-\pi N\!N$\
equations
is diagonal. Recalling that the amplitudes of $\underline{w}^0$ are two-nucleon
irreducible, we can see from \fig{w}, that the off-diagonal elements
$\underline{w}_4^0$ and $\underline{w}_5^0$ correspond to what has been called
the Jennings terms. As pointed out by Jennings\cite{Jennings}, these terms may
be important for the understanding of $\pi d$\ scattering. In our case, the
Jennings terms are also fully dressed, and form an essential part of the
convolution equations. Indeed, since our $N\!N$\ propagator $G_{N\!N}$ is fully
dressed, it also contains two-pion states coming from intermediate
Jennings-like
terms. It is then necessary to retain $\underline{w}_4^0$ and
$\underline{w}_5^0$ in the convolution equations because they combine with
$G_{N\!N}$ in just the right way to guarantee three-body unitarity.
We recall, that the only approximation made in deriving the convolution
equations of \eq{AB} is the neglect of all connected $\pi N\!N$-irreducible diagrams
for the $\pi N\! N\!\rightarrow\!\pi N\!N$\ process\cite{KB_PL}. Yet it is very easy to include some
types of connected contributions. One such contribution would involve
intermediate state potentials $V_{N\!N}^{(1)}$ that are $\pi N\!N$-irreducible. Then
\eq{AB} would be modified simply by replacing $V_{N\!N}$ with
$V_{N\!N}+V_{N\!N}^{(1)}$. This observation suggests that a way to include
heavy
meson exchange into our $N\!N$\ potential would be as a phenomenological model for
$V_{N\!N}^{(1)}$.
\bigskip
\noindent{\bf 3.\hspace{3mm} Four-dimensional formulation}
\bigskip
Although three-dimensional equations may be easier to solve than those in
four-dimensions, there are important reasons why the formulation of
four-dimensional equations is necessary. Firstly, such equations are based on
relativistic quantum field theory, and retain the fundamental property of
off-shell covariance. Secondly, having the correct four-dimensional equations,
one can then do a three-dimensional reduction using one of the well-known
reduction schemes. We may also add, that with the ever increasing power of
computers, the numerical solution of four-dimensional equations becomes ever
more feasible.
The first attempts to formulate few-body equations using relativistic quantum
field theory were made already in the early 1960's \ctt{Taylor}{Tucciarone}.
Both
such general formulations and ones more specific to the $\pi N\!N$\ system have been
pursued until the present time \ctt{Avishai4d}{Haberzettl}. Yet as in the
three-dimensional case, all these attempts have had theoretical
inconsistencies. In particular, all previous attempts have contained either
overcounting or undercounting of Feynman diagrams.
\bigskip
\begin{figure}[t]
\hspace*{5mm} \epsffile{figt.ps}
\caption{\fign{t} Example of overcounting in $N\!N\!\rightarrow\!\pi d$. (a) The $N\!N\!\rightarrow\!\pi d$\
Feynman diagram where dark circles represent all possible contributions.
(b) One of the contributions included in (a). (c) Another way of drawing
diagram (b) showing how overcounting arises. } \end{figure}
\noindent{\em 3.1\hspace{2mm} Overcounting and undercounting problems}
\bigskip
\begin{figure}[b]
\hspace*{3mm} \epsfxsize=15.5cm\epsffile{figu.ps}
\caption{\fign{u} Example of undercounting in $N\!N\!\rightarrow\!\pi N\!N$. (a) A $\pi N\! N\!\rightarrow\!\pi N\!N$\ graph
that has usually been neglected since it corresponds to a three-body force.
(b)
The coupling of the graph in (a) to the $N\!N$\ channel. (c) Another way of
drawing
diagram (b) reveals a two-body process. }
\end{figure}
Perhaps the easiest way to illustrate the overcounting problem in the $\pi N\!N$\
system is with an example. Consider the ``triangle'' diagram of \fig{t}(a) for
the process $N\!N\!\rightarrow\!\pi d$, where the dark circles represent the full $\pi N\!\rightarrow\!\pi N$\
amplitude, the dressed $\pi N\!N$\ vertex, and the dressed deuteron vertex.
If one were to calculate this diagram in four dimensions, as is, using
covariant
forms for the off-shell $\pi N$\ t-matrix, $\pi N\!N$\ vertex, and the deuteron vertex,
then one would have the mistake of overcounting of diagrams. This is
illustrated
in \fig{t}(b) where we consider just the crossed-pion graph contribution to the
input $\pi N$\ t-matrix. As these are Feynman graphs, there is no meaning
associated with the slope of the lines, and one could just as well have drawn
\fig{t}(c). However \fig{t}(c) clearly illustrates that this contribution
corresponds to the dressing of the already fully dressed deuteron vertex.
This type of overcounting arises in four-dimensional approaches whenever one
tries to formulate multiple-scattering graphs in terms of fully dressed
vertices
and full amplitudes for all subprocesses. In once-off cases, like that of
\fig{t}(a), one can easily fix the overcounting problem by making a necessary
subtraction (here one would subtract the graph of \fig{t}(b) from the
calculation of \fig{t}(a)). However, the way to solve the overcounting problem
for the case of coupled integral equations is highly non-trivial as an infinite
number of overcounted contributions are involved.
In a similar way, let us illustrate how undercounting arises in the covariant
$\pi N\!N$\ problem. As in the three-dimensional formulation, one neglects
three-body
forces also in the four-dimensional case. Only in this way can one obtain
few-body equations where (in the c.m.) no more than two independent momenta are
involved. However, one does need to be very careful about neglecting three-body
forces in the four-dimensional theory. Consider, for example the Feynman
diagram
of \fig{u}(a). This is a graph for the process $\pi N\! N\!\rightarrow\!\pi N\!N$\ that is both
connected
and $\pi N\!N$-irreducible. It therefore corresponds precisely to what is meant by
a
three-body force. However, neglecting this contribution from a few-body theory
of the $\pi N\!N$\ system would be a bad mistake. This is illustrated in \fig{u}(b)
where we allow the graph of \fig{u}(a) to couple to the $N\!N$\ channel. Again no
meaning can be attached to the slope of the propagator lines, and we can
equally
well draw this diagram as in \fig{u}(c). This, however, reveals that the
three-body force of the $\pi N\! N\!\rightarrow\!\pi N\!N$\ process has now become a two-body
rescattering contribution in the $N\!N\!\rightarrow\!\pi N\!N$\ process. Thus neglecting the
three-body force of \fig{u}(a) would lead to an undercounting of important
two-body contributions.
\bigskip
\noindent{\em 3.2\hspace{2mm} Four-dimensional $\pi N\!N$ equations}
\bigskip
\begin{figure}[b]
\hspace*{3mm} \epsfxsize=15.5cm\epsffile{figx.ps}
\caption{\fign{x} The subtraction terms in the four-dimensional $\pi N\!N$\
equation:
(a) $W_{\pi\pi}$, (b) $W_{\pi N}$, (c) $W_{N\!N}$, (d) $X$, (e) $Y$, and (f)
$B$. The dark circles represent the following two-body amplitudes: (a) full
$\pi\pi$ t-matrix, (b) one-nucleon irreducible $\pi N$ t-matrix, and (c) full
$N\!N$\ t-matrix minus the $N\!N$\ one-pion-exchange potential.}
\end{figure}
In a recent paper, we have solved both the overcounting and undercounting
problems in the formulation of few-body equations in field theory\cite{KB_NP}.
The few-body equations for the $\pi N\!N$\ system then follow as a particular case.
The method used to derive the equations involves the classification of Feynman
diagrams according to their irreducibility. The overcounting problem is handled
by a procedure where, in formally identical cases like that of \figs{t}(b) and
(c), one of the two right-most vertices is ``pulled out'' further to the
right. The undercounting of diagrams is handled simply by retaining all
three-body forces until the end of the derivation where the ones that did not
lead to two-body interactions are safely neglected. It is gratifying that
Phillips and Afnan\cite{PA} have recently confirmed our equations using a
modified version of Taylor's original classification of diagram
scheme \cite{Taylor,PA2}.
With three-body forces neglected as described, one might still find it useful
to
retain, as in the three-dimensional case, the $N\!N\!\rightarrow\!N\!N$\ $\pi N\!N$-irreducible
potential
$V_{N\!N}^{(1)}$, as well as the simultaneously $N\!N$- and $\pi N\!N$-irreducible
connected $N\!N\!\rightarrow\!\pi N\!N$\ amplitude $F_{\pi}$. However, let us at first consider
the simplest case where these contributions are neglected (they are in fact
completely absent in the usual case of a $\phi\bar{\psi}\psi$ interaction).
In this case the four-dimensional $\pi N\!N$\ equations can be written as \eq{AB},
but with the following modifications: (1) The product of two quantities $A$ and
$B$ is defined as in \eq{product}, but now with all momenta and integrations
being four-dimensional (in particular the replacement $d{\bf p}''_i\rightarrow
d^4p_i/(2\pi)^4$ needs to be made), (2) all convolutions of Green functions are
replaced by usual products, (3) the matrix $\bar{w}^0$ is now diagonal
(i.e. with $w^0_4=w^0_5=0$), and (4) the following replacements are made,
\begin{eqnarray}
\begin{array}{ccccc}
\underline{F}\rightarrow \underline{F}-\underline{B'} & ; &
\underline{\bar{F}}\rightarrow \underline{\bar{F}}-\underline{\bar{B}'} & ; &
V_{N\!N}\rightarrow V_{N\!N}-\Delta
\end{array}
\end{eqnarray}
where the terms $\underline{B'}$ and $\Delta$ are subtraction terms that
exactly compensate all the overcounting due to the use of full off-shell
amplitudes and fully dressed vertices in the coupled scattering equations. They
are defined with the help of \fig{x} as follows:
\begin{equation}
\Delta = W_{\pi\pi} + W'_{\pi N} + W_{N\!N} + X + Y' - \bar{B}'G_0B'
\eqn{sub}
\end{equation}
where the dashed quantities are the sums $W'_{\pi N}=W_{\pi N}+PW_{\pi N}P$,
$Y'=Y+PYP$, and $B'=B+PBP$, $P$ being the nucleon label exchange operator.
$\underline{B'}$ is a column matrix with each element given by $B'$.
In the more general case where $V_{N\!N}^{(1)}$ and $F_{\pi}$ are retained, it
turns out that only the subtraction terms need be modified. In
particular, we need to do the replacements $\Delta \rightarrow \Delta-
V_{N\!N}^{(1)}-\bar{F}_\pi(F_1+F_2)+
\bar{F}_{2\pi}F_1F_2-(\bar{F}_1+\bar{F}_2)F_\pi-\bar{F}_1\bar{F}_2F_{2\pi}$ and
$B'\rightarrow B'-F_\pi$, where $F_{2\pi}$ is the connected $N\!N$- and $\pi N\!N$-
irreducible amplitude for $N\!N\rightarrow \pi\pi N\!N$.
\vspace{1cm}
\noindent
$^\dagger$ Permanent address:\,\, Mathematical Institute of Georgian Academy of
Sciences, \\ Z. Rukhadze 1, 380093 Tbilisi, Georgia.
|
\section{INTRODUCTION}
Due to the nature of its microscopic origins, attention has focused
on both magnetic\cite{mag} and structural
properties\cite{adams,nishi,lorenzo,arai,pouget,kamimura,hirota,roessli%
,harris,stpaul,nishi2,chen,poirier1,poirier2,poirier3,wink,yama}
of the recently discovered spin-Peierls system $CuGeO_3$.
It is now well established that this material is the first example
of an inorganic compound
which exhibits the simultaineous opening of a gap from a singlet
ground state and a lattice dimerization in S=1/2 spin chains.
In contrast with previously studied organic spin-Peierls systems,
large single crystals of $CuGeO_3$ can be made which facilitate
neutron and X-ray diffraction studies in addition to the measurments of
crystal-direction-dependent thermodynamic properties.
Such measurements have revealed large anisotropic
strain effects due to the dimerization, in contrast with simple
quasi-one-dimensional models of $Cu$ displacements.
Such effects are examined in the
present work based on a phenomenological Landau-type model which describes
the atomic displacements below the transition temperature $T_c$
as proposed by Hirota {\it et al.}\cite{hirota}
In contrast with the principal $Cu$ displacements which give rise to anomalies
{\it at} $T_c$, secondary $O$ distortions are
found to be important in determining the temperature dependence of elastic
properties below $T_c$. A linear coupling between the two
types of ionic displacements is proposed which is electronic in origin.
Good agreement with available experimental results is demonstrated.
This study also serves to compliment and extend the brief Landau-type
descriptions of some elastic properties given in Refs.\onlinecite{nishi,harris%
,poirier2}.
Of the ten ions per high-temperature (above $T_c$) orthorhombic
unit cell ($a$ = 4.81, $b$ = 8.47, and $c$ = 2.941 $\AA$),
at sites ${\bf R} = {\bf R}_l + {\bf v}_j$ (where
${\bf R}_l$ are Bravais lattice vectors and $j=1-10$), there are
two $Cu$ and four $O$ ions which displace at $T_c$ in the model of
Ref\onlinecite{hirota}. These are shown schematically in Fig. 1, with the
high-T positional parameters ${\bf v}_j$ given by
\begin{eqnarray}
Cu_1:~v_1 &=&(0,0,0);~~~~~~~~~~Cu_2:~v_2 = (0,{\textstyle \frac12},0)
\nonumber \\
O_3: ~v_3 &=&(x,-y,{\textstyle \frac12});
~~~~~~~~O_4: ~v_4 = (-x,y,{\textstyle \frac12}) \\ \nonumber
O_5: ~v_5 &=&(-x,{\textstyle \frac12}-y,{\textstyle \frac12});
~~~O_6:~v_6 = (x,y+{\textstyle \frac12},{\textstyle \frac12}).
\end{eqnarray}
These displacements thus double the unit cell in the $a$ and $c$ directions
below $T_c$. A simple order-parameter description of the ionic movements
below $T_c$ may be written as
\begin{equation}
\mbox{\boldmath $\rho(R)$} = \mbox{\boldmath $\rho$}_j \cos ({\bf Q \cdot R}_l)
\end{equation}
where ${\bf Q} = (\frac12,1,{\textstyle \frac12})$ and here $j=1-6$.
The displacements shown in Fig. 1 are then described with
\begin{eqnarray}
\mbox{\boldmath $\rho$}_1 = - \rho_z {\bf \hat z};
~~\mbox{\boldmath $\rho$}_2 = - \mbox{\boldmath $\rho$}_1 \nonumber \\
\mbox{\boldmath $\rho$}_3 = - \rho_x {\bf \hat x} - \rho_y {\bf \hat y};
~~\mbox{\boldmath $\rho$}_4 = - \mbox{\boldmath $\rho$}_3 \\ \nonumber
\mbox{\boldmath $\rho$}_5 = - \rho_x {\bf \hat x} + \rho_y {\bf \hat y};
~~\mbox{\boldmath $\rho$}_6 = - \mbox{\boldmath $\rho$}_5
\end{eqnarray}
where $\rho_x$ and $\rho_y$ are associated with the $O$ distortions
and $\rho_z$ represents $Cu$-ion displacements. It is argued below that
simple ion-ion interactions derived from symmetry arguments can
reproduce the configuration of displacements depicted in Fig. 1. A
desirable feature of any such model is to account for the simultaneous
displacement of both $Cu$ and $O$ ions (which is also relevant to possible
critical behavior). It is for this reason that
that ionic density-density coupling must also be considered. In addition,
simple coupling between $\mbox{\boldmath $\rho$}$
and the strain tensor reproduces the principal
features of measured elastic properties without evoking higher-order
interactions as in Refs.\onlinecite{nishi,harris,poirier2}.
\section{LANDAU FREE ENERGY}
Landau-type free energies may be constructed using symmetry
arguments\cite{landau}. It is assumed here that magnetic degrees
of freedom are integrated out so that the free energy
depends explicitly on \mbox{\boldmath $\rho$} only.
In the case of orthorhombic symmetry, there
are three independent invariants which contribute to free energy density
$f$ at second order in the vector \mbox{\boldmath $\rho$},
which may be chosen as
\begin{equation}
f_2 = D'_x \rho_x^2 + D'_y \rho_y^2 +D'_z \rho_z^2.
\end{equation}
The independence of these three coefficients requires that
$\rho_x \not= \rho_y \not= \rho_z$, as observed experimentally.
Alternatively, one can choose the three terms as
$f_2 = A \rho^2 + D_x \rho_x^2 +D_y \rho_y^2$,
where the $A$-term represents an isotropic interaction.
Minimization of the anisotropy at the $Cu$ sites, for
example, gives
$\mbox{\boldmath $\rho$}_1, \mbox{\boldmath $\rho$}_2
\parallel \pm {\bf \hat z}$
if $D_x(Cu) > 0$ and $D_y(Cu) > 0$.
In the following
analysis, only isotropic terms are considered in detail, which is
sufficient for the purposes of the present study.
\subsection{Second-order terms}
A general expression for isotropic two-ion interactions is
\begin{equation}
F_2 = \frac12 \sum_{{\bf R,R'}} A{\bf(R - R')}
\mbox{\boldmath $\rho$}({\bf R}) \cdot \mbox{\boldmath $\rho$}({\bf R'}).
\end{equation}
Using the form (2) for $\mbox{\boldmath $\rho$({\bf R})}$
in this relation yields
\begin{equation}
F_2 = \frac12 \sum_{j,k} {\tilde A}_{jk}({\bf Q})
\mbox{\boldmath $\rho$}_j \cdot \mbox{\boldmath $\rho$}_k,
\end{equation}
where
\begin{equation}
{\tilde A}_{jk}({\bf Q}) = \sum_{\mbox{\boldmath $\tau$}}
A(\mbox{\boldmath $\tau$} + {\bf v}_j - {\bf v}_k)
\cos ({\bf Q} \cdot \mbox{\boldmath $\tau$})
\end{equation}
with $\mbox{\boldmath $\tau$} = {\bf R}_l - {\bf R'}_l$.
Considered below are the near-neighbor interactions within and between
the $CuO_2$ chains which are likely to be important. We recall
that $CuGeO_3$ is somewhat quasi-one-dimensional.
\subsubsection{Cu-Cu within chains}
Interactions between near-neighbor $Cu$ ions along the $c$ axis
involve $j=k=(1,2)$ with ${\tilde A}_{11} = A({\bf c}) \cos \pi
+ A({-\bf c}) \cos \pi$,
and with a similar expression for ${\tilde A}_{22}$, so that
\begin{eqnarray}
F_{2A} &=& -A({\bf c})[\mbox{\boldmath $\rho$}_1 \cdot \mbox{\boldmath
$\rho$}_1
+ \mbox{\boldmath $\rho$}_2 \cdot \mbox{\boldmath $\rho$}_2 ] \nonumber \\
&=& -2A_{11} \rho_C^2
\end{eqnarray}
where $A_{11} = A({\bf c})$.
\subsubsection{Cu-Cu between chains}
Here $j \not= k=(1,2)$ with, e.g.,
${\tilde A}_{12} = A(-{\bf v}_2) + A({-\bf v}_2 + b{\bf \hat y}) \cos2 \pi$,
giving
\begin{equation}
F_{2B} = 2A({\bf v}_2)
\mbox{\boldmath $\rho$}_1 \cdot \mbox{\boldmath $\rho$}_2.
\end{equation}
If $A({\bf v}_2) > 0$, then this term is minimized by the configuration
$\mbox{\boldmath $\rho$}_1 \parallel -\mbox{\boldmath $\rho$}_2$, as in
Fig. 1. It can be expected, however, that this term is smaller
than $F_{2A}$ due to the larger interaction distance involved.
Its contribution is given by $F_{2B} = -2A_{12} \rho_C^2$,
where $A_{12} = A({\bf v}_2)$.
\subsubsection{O-O within chains}
Interactions between the two $O$ ions which surround the $Cu$ of Fig. 1
correspond with $j \not= k=(3,4)$ and $(5,6)$. Using
${\tilde A}_{34} = A({\bf v}_3 - {\bf v}_4) = A_{43} = A_{56} = A_{65}$,
the energy is
\begin{equation}
F_{2C} = A({\bf v}_3 - {\bf v}_4)
[\mbox{\boldmath $\rho$}_3 \cdot \mbox{\boldmath $\rho$}_4
+ \mbox{\boldmath $\rho$}_5 \cdot \mbox{\boldmath $\rho$}_6].
\end{equation}
If $A({\bf v}_3 - {\bf v}_4) > 0$, then the configuration
$\mbox{\boldmath $\rho$}_3 \parallel -\mbox{\boldmath $\rho$}_4$ and
$\mbox{\boldmath $\rho$}_5 \parallel -\mbox{\boldmath $\rho$}_6$,
as in Fig. 1, is stabilized by this term, giving
$F_{2C} = -2A_{34} \rho_O^2$, where $A_{34} = A({\bf v}_3 - {\bf v}_4)$.
\subsubsection{O-O between chains}
With $j \not= k=(4,5)$ and $(3,6)$, the resulting contribution to the
free energy is
\begin{equation}
F_{2D} = A({\bf v}_4 - {\bf v}_5)
[\mbox{\boldmath $\rho$}_4 \cdot \mbox{\boldmath $\rho$}_5
+ \mbox{\boldmath $\rho$}_3 \cdot \mbox{\boldmath $\rho$}_6].
\end{equation}
Although parallel or anti-parallel configurations are minimized
by this term, it is expected to be rather small. Fourth-order
terms considered below prefer a configuration
$\mbox{\boldmath $\rho$}_4 \perp \mbox{\boldmath $\rho$}_5$ and
$\mbox{\boldmath $\rho$}_3 \perp \mbox{\boldmath $\rho$}_6$
giving $F_{2D} = 0$. As emphasized above, however, anisotropy effects
will alter somewhat this idealized state.
\subsubsection{Cu-O within chains}
Considered here are the near-neighbor $Cu-O$ interactions, so that
$j \not= k=(1,3)$, (1,4), (2,5), as well as (2,6). Due to the fact
that neighboring $Cu$ ions along the $c$ axis displace in opposite
directions, and that the $O$ ions (3,4) and (5,6) are
located mid-way between these $Cu$ ions, this two-ion interaction term
is zero. This is illustrated by considering
${\tilde A}_{13} = A(-{\bf v}_3) + A({\bf c} - {\bf v}_3) \cos \pi = 0$,
where $A(-{\bf v}_3) = A({\bf v}_3) = A({\bf c} - {\bf v}_3)$ has
been used. In the absence of $Cu-O$ coupling at second order, there
is not a simultaneous displacement of both species. This result is
not in agreement with the observed behavior.
\subsubsection{Cu-O within chains: density-density interactions}
In addition to the two-ion intereactions which involve the vector
displacements $\mbox{\boldmath $\rho(R)$}$, there are also ionic
charge density couplings which involve the perturbed density
$\rho({\bf R})~=~|\mbox{\boldmath $\rho(R)$}|$.
Such interactions are of the type
which have been considered within the framework of density functional
theory\cite{plum} and are of the form
\begin{eqnarray}
F_G &=& \frac12 \sum_{{\bf R,R'}} G{\bf( R - R')}
\rho({\bf R}) \rho({\bf R'}) \nonumber \\
&=& \frac12 \sum_{j,k} {\tilde G}_{jk} \rho_j \rho_k
\end{eqnarray}
where
\begin{equation}
{\tilde G}_{jk} = \sum_{\mbox{\boldmath $\tau$}}
G(\mbox{\boldmath $\tau$} + {\bf v}_j - {\bf v}_k).
\end{equation}
For simplicity, only the near-neighbor $Cu-O$ coupling is explicitly
considered here (the other interactions discussed above are simply
renormalized by charge density coupling). This gives a contribution
\begin{equation}
F_G = 4 G_{13} \rho_C \rho_O,
\end{equation}
where $G_{13} = G({\bf v}_3)$.
Linear coupling of this type yields (see below)
$\rho_O \propto \rho_C$, as expected.
Temperature dependence appears through the usual entropy
term of mean-field theory\cite{landau}
\begin{equation}
F_e = {\textstyle \frac12} aT \sum_{\bf R} \rho^2({\bf R}).
\end{equation}
With the simplifying assumption that $\rho_x \simeq \rho_y$ (see below),
all of the second-order terms considered above
can be written as
\begin{equation}
F_2 = A_C \rho^2_C + A_O \rho^2_O + A_{CO} \rho_C \rho_O,
\end{equation}
where
\begin{equation}
A_C = aT - 2A_{11} - 2A_{12}, ~~~ A_O = aT - 2A_{34}, ~~~ A_{CO} = 4G_{13}.
\end{equation}
\subsection{Fourth-order terms}
We omit here consideration of fourth-order anisotropy terms and note that,
in general, isotropic interactions involving the vector
\mbox{\boldmath $\rho(R)$} are nonlocal and can be expressed as\cite{plum2}
\begin{equation}
F_4 = \frac14 \sum_{\lbrace{\bf R}_i \rbrace} B{\bf(R_1,R_2;R_3,R_4)}
[\mbox{\boldmath $\rho$({\bf $R_1$})}
\cdot \mbox{\boldmath $\rho$({\bf $R_2$})}]
[\mbox{\boldmath $\rho$({\bf $R_3$})}
\cdot \mbox{\boldmath $\rho$({\bf $R_4$})}],
\end{equation}
where $B$ depends only on the relative distances $\mid {\bf R_i - R_j} \mid$.
For simplicity, and for the puposes of demonstrating that the present model
is plausible, only two types of terms will be considered explicitly here:
\begin{eqnarray}
F_{4a} &=& \frac14 B_0 \sum_{\bf R}
[\mbox{\boldmath $\rho$({\bf R})}
\cdot \mbox{\boldmath $\rho$({\bf R})}]^2 \nonumber \\
&=& B_0[ \rho_C^2 + 2 \rho_O^2 ]^2,
\end{eqnarray}
which comes from entropy considerations,\cite{landau}
as well as the simple two-ion interactions of the form
\begin{equation}
F_{4b} = \frac14 \sum_{{\bf R,R'}} B{\bf(R - R')}
[\mbox{\boldmath $\rho$({\bf R})} \cdot \mbox{\boldmath $\rho$}({\bf R'})]^2.
\end{equation}
The five near-neighbor couplings as considered above for $F_2$ yield a result
\begin{eqnarray}
F_{4b} &=& B_{11} [\rho_1^4 + \rho_2^4]
+ B_{12}(\mbox{\boldmath $\rho$}_1
\cdot \mbox{\boldmath $\rho$}_2)^2 \nonumber \\
&+& {\textstyle \frac12} B_{34}[(\mbox{\boldmath $\rho$}_3
\cdot \mbox{\boldmath $\rho$}_4)^2
+ (\mbox{\boldmath $\rho$}_5
\cdot \mbox{\boldmath $\rho$}_6)^2]
+ {\textstyle \frac12} B_{45}[(\mbox{\boldmath $\rho$}_4
\cdot \mbox{\boldmath $\rho$}_5)^2
+ (\mbox{\boldmath $\rho$}_3
\cdot \mbox{\boldmath $\rho$}_6)^2] \nonumber \\
&+& B_{13}[(\mbox{\boldmath $\rho$}_1 \cdot \mbox{\boldmath $\rho$}_3)^2
+ (\mbox{\boldmath $\rho$}_1 \cdot \mbox{\boldmath $\rho$}_4)^2
+ (\mbox{\boldmath $\rho$}_2 \cdot \mbox{\boldmath $\rho$}_5)^2
+ (\mbox{\boldmath $\rho$}_2 \cdot \mbox{\boldmath $\rho$}_6)^2].
\end{eqnarray}
Configurations minimized by these terms are as follows:
\begin{eqnarray}
For ~ B_{12} & < & 0, ~~~
\mbox{\boldmath $\rho$}_1 \parallel \pm \mbox{\boldmath $\rho$}_2; \nonumber
\\
for ~ B_{34} & < & 0, ~~~
\mbox{\boldmath $\rho$}_3 \parallel \pm \mbox{\boldmath $\rho$}_4; ~~
\mbox{\boldmath $\rho$}_5 \parallel
\pm \mbox{\boldmath $\rho$}_6;\nonumber \\
for ~ B_{45} & > & 0, ~~~
\mbox{\boldmath $\rho$}_4 \perp \mbox{\boldmath $\rho$}_5; ~~
\mbox{\boldmath $\rho$}_3 \perp \mbox{\boldmath $\rho$}_6;\nonumber \\
for ~ B_{13} & > & 0, ~~~
\mbox{\boldmath $\rho$}_C \perp \mbox{\boldmath $\rho$}_O.
\end{eqnarray}
Note that the complete minimization of the term $B_{45}$ occurs only if
$\rho_x = \rho_y$ when (3) is used, but that
this equality is in competition with anisotropy terms. It
appears, however, that this condition is nearly satisfied in $CuGeO_3$.
These results, together with the minimization conditions from consideration
of $F_2$, fully explain the displacements of Fig. 1, as described by Eqs.
(1)-(3).
\subsection{Equilibrium Properties}
The full free energy is then given by
\begin{equation}
F = A_C \rho^2_C + A_O \rho^2_O + A_{CO} \rho_C \rho_O
+ B_C \rho^4_C + B_O \rho^4_O + B_{CO} \rho_C^2 \rho_O^2,
\end{equation}
where $B_C = B_0 + B_{11} + B_{12}$, $B_O = 4B_0 + B_{34}$, and $B_{CO} =
4B_0$.
Due to the nature of the spin-Peierls interaction, it is assumed here that
$Cu$ and $O$ displacements represent primary and secondary order parameters,
respectively.
Minimization of $F$ with respect to $\rho_O$ yields a result which can
be expressed as an expansion in odd powers of $\rho_C$, at least near the
transition temperature,
\begin{equation}
\rho_O = \alpha \rho_C + \beta \rho^3_C + \cdot \cdot \cdot,
\end{equation}
where
\begin{equation}
\alpha = - {\textstyle \frac12} A_{CO} / A_O; ~~~
\beta = - \alpha / A_{O} (B_{CO} + {\textstyle \frac12} B_O A_{CO}^2 / A_O^2).
\end{equation}
This establishes a linear relationship between $Cu$ and $O$ displacements
near $T_c$, but that a more complicated behavior can be expected at lower
temperatures. Using this result in the above expression for $F$ then
gives simply
\begin{equation}
F = A \rho^2_C + {\textstyle \frac12} B \rho^4_C,
\end{equation}
where $A = A_C - \frac12 \alpha A_{CO}$ and
$\frac12 B = B_C + \alpha^2 B_{CO} + \alpha^4 B_O$.
The transition temperature is defined by $A \equiv a(T - T_c)$, so that
\begin{equation}
aT_c = A_+ + \left \lbrace A_-^2 + 4G_{13}^2 \right \rbrace ^\frac12,
\end{equation}
where $A_\pm = A_{11} + A_{12} \pm A_{34}$ and $A_{CO} = 4 G_{13}$ has been
used.
For later purposes, we note here that minimization of $F$ above yields
$\rho_C^2 = -A/B = (a/B)(T_c-T)$ for the primary order parameter,
and that the specific-heat anomaly at $T_c$ is given by
\begin{equation}
\Delta C = C(T_c^+) - C(T_c^-) = -T_c \frac{a^2}{B}.
\end{equation}
\section{ELASTIC COUPLING}
The orthorhombic symmetry of $CuGeO_3$ allows for nine independent
elastic constants: $C_{11}, C_{12}, C_{13}, C_{22}, C_{23}, C_{33},
C_{44}, C_{55}, C_{66}$. The elastic energy is conveniently
written as
\begin{equation}
F_{el} = \frac12 \sum_{\mu \nu} C_{\mu \nu} e_\mu e_\nu
\end{equation}
where Voigt notation $(1 \equiv xx, 2 \equiv yy, 3 \equiv zz,
4 \equiv yz, 5 \equiv zx, 6 \equiv xy)$ has been used.
Lowest order coupling between the strain tensor and displacement
vector \mbox{\boldmath $\rho$} is of the form $ \sim e \rho^2$
and also has nine independent terms
\begin{eqnarray}
F_{\rho e} = \frac12 \sum_{{\bf R,R'}} \lbrack &D_1& e_{xx} \rho_x \rho_x
+ D_2 e_{yy} \rho_y \rho_y + D_3 e_{zz} \rho_z \rho_z
+ D_4 (e_{xx} \rho_y \rho_y + e_{yy} \rho_x \rho_x) \nonumber \\
+ &D_5& (e_{xx} \rho_z \rho_z + e_{zz} \rho_x \rho_x)
+ D_6 (e_{yy} \rho_z \rho_z + e_{zz} \rho_y \rho_y) \\
+ &D_7& e_{yz} \rho_y \rho_z + D_8 e_{xz} \rho_x \rho_z
+ D_9 e_{xy} \rho_x \rho_y \rbrack, \nonumber
\end{eqnarray}
where the short-hand notation $D_i = D_i({\mbox{\boldmath $\tau$}})$ and
$\rho_\alpha \rho_\beta = \rho_\alpha({\bf R}) \rho_\beta({\bf R'})$
has been used. The same near-neighbor interactions as considered
above for $F_2$ and $F_{4b}$ yield
\begin{eqnarray}
F_{\rho e} = 2 \lbrack &D_1& e_{xx} \rho_x^2
+ D_2 e_{yy} \rho_y^2 + D_3 e_{zz} \rho_z^2
+ D_4 (e_{xx} \rho_y^2 + e_{yy} \rho_x^2)\nonumber \\
+ &D_{5C}& e_{xx} \rho_z^2 + D_{5O} e_{zz} \rho_x^2
+ D_{6C} e_{yy} \rho_z^2 + D_{6O} e_{zz} \rho_y^2
+ D_9 e_{xy} \rho_x \rho_y \rbrack.
\end{eqnarray}
Recall that $Cu$ displacements are along the $z$ axis and that
$O$ displacements are in the $xy$ plane. The interaction constants
$D_1, D_2, D_4, D_{5O}, D_{6O}$, and $D_9$ thus involve $O-O$ couplings
whereas $D_3, D_{5C}$, and $D_{6C}$ involve $Cu-Cu$ couplings.
$Cu-O$ interactions are zero for the same reasons as described in section
2A-5, although those inlvolving density-density couplings could be
added if desired. With the assumption $\rho_x =\rho_y = \rho_O/ \sqrt{2}$,
as well as the result from Eq. (24)
\begin{equation}
\rho_O^2 = \chi \rho_C^2, ~~~ \chi \simeq \alpha^2 + 2 \alpha \beta \rho_C^2,
\end{equation}
the elastic coupling term reduces to
\begin{eqnarray}
F_{\rho e} &=& \lbrack K_1 e_{xx} +K_2 e_{yy} + K_3 e_{zz}
+ K_6 e_{xy} \rbrack \rho_C^2 \nonumber \\
&=& \sum_\mu K_\mu e_\mu \rho_C^2,
\end{eqnarray}
where
\begin{eqnarray}
K_1 &=& (D_1 + D_4) \chi + 2 D_{5C} \nonumber \\
K_2 &=& (D_2 + D_4) \chi + 2 D_{6C} \nonumber \\
K_3 &=& 2 D_3 (D_{5O} + D_{6O}) \chi \nonumber \\
K_6 &=& D_9 \chi.
\end{eqnarray}
Note that $\chi$ is temperature dependent through $\rho_C^2$ and that it
is proportional to the square of the density-density coupling constant,
$\chi \sim G_{13}^2$. It is also convenient to write
\begin{equation}
K_\mu = K_{\mu o} + k_\mu \rho_C^2,
\end{equation}
where $K_{\mu o}$ is independent of temperature and depends
only on $Cu-Cu$ interactions, whereas the coefficient of the
temperature-dependent contribution $k_\mu$ depends only on $O-O$ interactions.
\subsection{Thermal expansion and elastic constants}
The full free energy now takes the simple form\cite{plum2}
\begin{equation}
F = A \rho^2_C + {\textstyle \frac12} B \rho^4_C
+ {\textstyle \frac12} \sum_{\mu \nu} C_{\mu \nu} e_\mu e_\nu
+ \sum_\mu K_\mu e_\mu \rho_C^2.
\end{equation}
Minimization with respect to the strain tensor yields the following
results for the effects of $Cu$ and $O$ displacements on
thermal expansion, the expansion coefficient, and the elastic constants
\begin{eqnarray}
e_\mu &=& -\sum_\nu s_{\mu \nu} \lbrack K_{\nu o}
+ k_\nu \rho_C^2 \rbrack \rho_C^2 \nonumber \\
\Delta \alpha_\mu &=& - (a/B) \sum_\nu s_{\mu \nu} \lbrack K_{\nu o}
+ 2 k_\nu \rho_C^2 \rbrack \nonumber \\
\Delta C_{\mu \nu} &=& (K_{\nu o} + k_\nu \rho_C^2)
(K_{\mu o} + k_\mu \rho_C^2) /B
\end{eqnarray}
where $s_{\mu \nu}$ is the compliance matrix
$(s_{\mu \nu} = C^{-1}_{\mu \nu})$,
$\Delta \alpha_\mu = \alpha_\mu (T_c^+) - \alpha_\mu (T_c^-)$,
and with a similar definition for $\Delta C_{\mu \nu}$.
Due to the temperature dependence of $\rho_C^2$, thermal
expansion is not simply proportional to the square of the
order parameter, and the expansion coefficient does not exhibit a simple
step discontinuity, as is usually the case with low order
($ \sim e \rho^2$) interactions. In addition, the elastic constants exhibit
temperature dependence below $T_c$, also due to the displacement
of $O$ ions (since $k_\mu \rho_C^2 \sim \rho_O^2$).
Such effects can also be mimicked by including higher-order
couplings.\cite{nishi,harris,poirier2}
\subsection{Numerical fits}
With available experimental data, it is possible to fit a number
of the Landau model parameters.
{}From the specific heat data of Ref.\onlinecite{wink},
the estimate $\Delta C \simeq -4.5 \times 10^4 J/(K m^3)$ can be made.
Comparing this result with Eq. (28), and using $T_c = 14.2K$, gives
$a^2/B \simeq 3.2 \times 10^3 J/m^3$ (where the free energy $F$ is in units
of energy/volume). The order parameter at low temperature takes the form
$\rho_C(0) = (a T_c/B)^\frac12$. This expression can then be used with
the above result, and the estimate from Hirota {\it et al.} of
$\rho_C(0) \simeq 0.0014 c$, to give
$a \simeq 2.6 \times 10^{29} J/(K m^5)$ and
$B \simeq 2.2 \times 10^{55} J/(K m^7)$ in (cumbersome) $SI$ units.
There are also available limited data on the elastic constants near
$T_c$. In particular, from the results of Ref.\onlinecite{poirier1}
on $C_{11}, C_{22}$, and $C_{33}$, it is clear that only $C_{33}$
exhibits a detectable discontinuity so that $K_{10} \simeq 0$,
$K_{20} \simeq 0$. From these data, the estimate
$\Delta C_{33}/C_{33} \simeq 2.5 \times 10^{-3}$ can be made. Using
this result, along with\cite{private} $C_{33} \simeq 3.7 \times 10^{11} N/m^2$,
as well as Eq. (37) and the above value for $B$, the estimate
$K_{30} \simeq 1.5 \times 10^{32}J/m^5$
can be made. These results, together with the relations (37),
imply that the only other elastic constants which could display a
discontinuity at $T_c$ are $C_{36}$ and $C_{66}$, provided that $K_{60}$ has
appreciable magnitude.
The temperature dependence of $\Delta C_{\mu \nu}$ below $T_c$ from Eq. (37)
is of the form $\sim \Delta C_0 + c_1 \rho^2_C + c_2 \rho^4_C$,
where in general $\rho_C \sim (T_c - T)^\beta$. From the data of
Ref.\onlinecite{poirier1}, there appears to be a region near $T_c$ where
$\Delta C_{11}$, $\Delta C_{22}$, and $\Delta C_{33}$ increase
{\it nearly} linearly with decreasing temperature,
followed by a less rapid increase at lower temperatures. A more detailed
analysis of these results\cite{poirier2}
suggests a value $\beta = 0.42(2)$. This estimate may be
compared with $\beta = 0.25(5)$ found in an earlier study
of elastic properties.\cite{stpaul} A discussion of
possible critical behaviors is deferred to the next section.
It is also of interest to examine the expression for the effective
fourth-order coefficient $B' = B - \sum_{\mu \nu} K_{\mu 0} K_{\nu 0}$
which results from elastic coupling.\cite{plum2} With the reasonable
assumptions that the dominant contribution is from $K_{30}$, and
that $s_{33} \sim 1/C_{33}$, one finds $(B' - B)/B \sim 10^{-3}$. The
argument given in Ref.\onlinecite{harris} that elastic coupling
could be responsible for reducing the magnitude of the fourth-order
coefficient to near zero appears to be inconsistent with the above
analysis. Thus, the proposal in Ref.\onlinecite{harris}
of tricriticality is not supported by the present work.
Experimental results on the three diagonal components of the
strain tensor may be found in Ref.\onlinecite{wink}. From Eq. (37)
and the approximations $K_{10} = K_{20} = 0$,
temperature dependence of the form
\begin{equation}
e_\mu = -\lbrack s_{\mu 3} K_{30} + (s_{\mu 1} k_1 + s_{\mu 2} k_2)
+ s_{\mu 3} k_3) \rho^2_C \rbrack \rho^2_C ,
\end{equation}
is expected.
It is clear that such predicted behavior is consistent with the
data. Detailed fitting to extract estimates of the three parameters
$k_\mu$ must await determination of the absolute values of
the other elastic constants (only $C_{11}, C_{22}$, and $C_{33}$ are
presently available) and hence knowledge of the compliance matrix.
With the assumption of quadratic order-parameter dependence, the authors
of Ref.\onlinecite{wink} extracted estimates $2 \beta = 0.61(2)-0.69(8)$.
Discontinuities in $\Delta \alpha_\mu$ at $T_c$ along the three principle
axes $a, b$, and $c$ were found to be $1.3(1)$, $-2.3(1)$,
and $-0.4(1) \times 10^{-5}/K$, respectively.\cite{wink}
{}From (37), as well as the approximation $s_{33} \sim 1/C_{33}$, the
c-axis expansivity is given by $\Delta \alpha_c \simeq -(a/B)K_{30}/C_{33}$.
Using the parameter estimates given above, this yields the model
prediction $\Delta \alpha_c \simeq -5 \times 10^{-6}$, in very good
agreement with the data. Comparison between the model predictions
$\Delta \alpha_a = -(a/B)K_{30}s_{13}$ and
$\Delta \alpha_b = -(a/B)K_{30}s_{23}$ at $T_c$
with corresponding experimental results
requires knowledge of the compliance matrix.
It is also of interest to note that the temperature dependences observed
below $T_c$ in the expansivity are accounted for by Eq. (37).
\section{CONCLUSIONS}
In spite of the great interest in the novel properties of the new
inorganic spin-Peierls system $CuGeO_3$ as reflected by the large number
of recently published experimental studies, there has been relatively
few theoretical investigations. This is partly due to the complex
nature of the relevant interactions caused by the non-negligible
interchain interactions in this not-so quasi-one-dimensional system.
Phenomenological models are thus particularly useful due to their
simplicity and serve to guide further study of a more microscopic nature.
The Landau-type free energy studied in the present work was constructed
from symmetry arguments with dominant close-neighbor interactions
included. The analysis demonstrates that the simplest of interactions can
account for the model of rather complicated ionic displacements proposed in
Ref.\onlinecite{hirota}. $O$-ion distortions are coupled linearly to
the principal $Cu$-ion displacements by charge-density interactions.
Elastic distortions as a function of temperature below $T_c$, not
present in simple Landau models, are seen here to be a consequence of
this incipient $O$-ion movement. This contrasts with the elastic anomalies
{\it at} $T_c$, which are due to the $Cu$ displacements.
A simple argument based on the results of this work suggest that the
transition should belong to the standard $XY$ universality class
(where, e.g., $\alpha = -0.012(10), \beta=0.349(4), \nu = 0.671(5)$).
One component is simply the $Cu$ displacement along the $c$ axis,
$\rho_z = \rho_C$, and the other is the concomitant $O$ distortion
in the $ab$ plane $\rho_\bot = \rho_O$. Note that there is not rotational
degeneracy in the $ab$ plane due to the second-order anisotropy terms:
A specific direction is favored.
In addition to the wide ranging estimates for the critical exponent
$\beta= 0.25 - 0.42$ determined from the elastic properties discussed above,
there have been disparate results from other techniques. Analysis of neutron
diffraction data for $\beta$ and $\nu$ indicated a simple mean-field transition
at $T_c$,\cite{nishi,pouget}
in agreement with one recent study of the specific heat.\cite{kuo}
In contrast, an earlier specific-heat measurement yielded an estimate
$\alpha \simeq 0.4$,\cite{sahling} whereas the most recently published result
claims $-0.15 \leq \alpha \leq 0.15$.\cite{liu} The above results for
$\beta$ may also be compared with those from X-ray
scattering data:\cite{harris,lum} $\beta=0.26(3)$ and $\beta=0.37(3)$.
One may conclude from all of these results
that the most recent of studies of the critical behavior appear to favor a
scenario where the spin-Peierls transition in $CuGeO_3$ belongs to
one of the standard universality classes (Ising, XY, or Heisenberg).
It is of interest to note that some of the above results suggest
tricitical behavior ($\alpha = \frac12, \beta=\frac14$). Simulations
of simple microscopic models of structural phase transitions indeed indicate
that the fourth-order coefficient of a corresponding Landau-type expansion
can be rather small.\cite{giddy} Analysis of recent X-ray data on
$CuGeO_3$ demonstrates
that the critical region is rather small and that tricritical behavior is
evident if exponents are estimated using larger values of the reduced
temperature.\cite{lum} Similar conclusions were recently made based on
Monte Carlo simulations of a magnetic system that was previously thought to be
tricritical.\cite{plum3}
\acknowledgements
I thank A Caill\'e for suggesting this problem, as well as M. Poirier,
M. Ain, B. Gaulin and M. Azzouz for useful discussions.
This work was supported by NSERC of Canada and FCAR du Qu\'ebec.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.