Content-Type: multipart/mixed; boundary="-------------0007141802712"
This is a multi-part message in MIME format.
---------------0007141802712
Content-Type: text/plain; name="00-293.comments"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename="00-293.comments"
Submitted to J. Math. Phys., dedicated to J.M. Luttinger.
---------------0007141802712
Content-Type: text/plain; name="00-293.keywords"
Content-Transfer-Encoding: 7bit
Content-Disposition: attachment; filename="00-293.keywords"
translation symmetry breaking, bounded charge fluctuations, Coulomb systems, jellium, theta-states, spin chains, dimerization, one dimensional particle systems, tight co-cycles
---------------0007141802712
Content-Type: application/x-tex; name="BnddFluct.tex"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline; filename="BnddFluct.tex"
% LATEX DOCUMENT.
% "Bounded Fluctuations and Translation Symmetry Breaking ..."
% M. Aizenman, S. Goldstein and J.L. Lebowitz.
%%%%%%%%%%%%%%%%%%%%%%%%%%
% The main change needed to fit the Journal's style,
% is at the section headers, as explained below.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\documentclass[12pt]{article}
\usepackage{amsfonts,amsmath} % amsmath,times,psfig}
\date{ \vspace*{-.1truein} {\small July 12, 2000 }}
% \today } }
\input epsf %preferred figures program
\bibliographystyle{mabib}
% for BibTeX - sorted numerical labels
% by order of first citation.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\setlength{\oddsidemargin}{.3in}
\setlength{\evensidemargin}{.3in}
\setlength{\textwidth}{6.2in}
\setlength{\textheight}{8.3in}
\setlength{\topmargin}{0in}
\setlength{\leftmargin}{-0.3in}
\parskip=7pt
\parindent 0.4in
%% --Macro producing the abstract -- invoked by "\abst{text}:
\def\abst#1{\begin{minipage}{5.5in}
{\noindent \normalsize
{\bf Abstract} #1} %\\
\end{minipage} }
%%% equation shortcuts %%%%%%%%%%%%
\def\be{\begin{equation}}
\def\ee{\end{equation}}
\def\bea{\begin{eqnarray}}
\def\eea{\end{eqnarray}}
\def\mp#1{*\marginpar{#1}}
\def\ma{*\marginpar{MA}}
\newcommand{\eq}[1]{eq.~(\ref{#1})}
\newcommand{\lamst}{{\lambda^*}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\def\var{{\rm Var}}
\def\abs#1{\left\vert #1 \right\vert}
\def\blackbox{{\vrule height 1.3ex width 1.0ex depth -.2ex}
\hskip 1.5truecm}
\def\Cell{{\mathcal C}} %%% a curve
\def\c{\mathrm{const.}}
\def\Chi {{\cal X}}
\def\d {{\rm d}}
\def\deg {{\rm deg}}
\def\G { {\mathcal G} }
\def\T{ {\mathcal T}} %% the symmetry group of translations
\def\dist{\operatorname{dist}} %%% distance
\def\dim{\operatorname{dim}} %%% dimension
\def\dist {{\rm dist}}
\def\E{{\mathbb E}} %%% for "expectation value" (in math mode)
% \def\EE{{\mathbf E}} %%% set of bonds
\def\eps {\varepsilon}
\def\I{\mathrm I} %%% \I-- for the indicator function (in math)
\def\lg{\stackrel{\scriptstyle <}{_{_{\scriptstyle >} }} }
%% \lg for "less and greater"
\def\liminf{\mathop{\underline{\rm lim}}}
\def\limsup{\mathop{\overline{\rm lim}}} % prettier limsup, liminf
\def\norm#1{\left\| #1 \right\|}
\newcommand{\ovl}{\overline}
\def\P{\operatorname{Prob}} %%% appears in many equations Prob
\def\R{{\mathbb R}} %%% for R (in math mode)
\def\C{{\mathbb C}} %%% Complex (in math mode)
\def\too#1{\parbox[t]{.4in} {$\longrightarrow\\[-9pt] {\scriptstyle #1}$}}
% converges to (?) as #1 to infinity
\def\V{{\mathcal V}} %%% set of vertices
\def\Z{{\mathbb Z}} %%% for Z (in math mode)
\def\N{{\mathbb N}} %%% for N (in math mode)
\def\ran{\operatorname{Ran}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Explanation:
% the following macros have the purpose of:
% 1) changing the equation numbers to the format: (3.4) %
% 2) Labeling the Appendix Sections as A,B,... with eq: (A.3)
% The appendix starts with the declaration: \startappendix
% which resets the new counter used here for the section numbers.
%
% Changes needed for conversion to the bare-bones Latex:
% \masection => \section
% \masubsection => \subsection
% \startappendix => \appendix
% remove the space corrections next to the \section declarations
% (their purpose was to avoid a big gap with the \subsection title)
% and possibly redo the "Acknowledgment" and "Reference" lines. %
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%$$---Modified Section titles %%%%%%%%%
% invoke by:
% \masect{Introduction} %\vspace{-.6cm} --before first masubsect.
% \masubsect{Introduction}
% \maappendix{Title}
% \masubappendix{subsection title}
% be sure to place, just above the first appendix the line:
% \startappendix -- to restart the section counter
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newcounter{masectionnumber}
\setcounter{masectionnumber}{0}
\newcommand{\masect}[1]{\setcounter{equation}{0}
\refstepcounter{masectionnumber} \vspace{1truecm plus 1cm} \noindent
{\large\bf \arabic{masectionnumber}. #1}\par \vspace{.2cm}
\addcontentsline{toc}{section}{\arabic{masectionnumber}. #1}
}
\renewcommand{\theequation}
{\mbox{\arabic{masectionnumber}.\arabic{equation}}}
\newcounter{masubsectionnumber}[masectionnumber]
\setcounter{masubsectionnumber}{0}
\newcommand{\masubsect}[1]{
\refstepcounter{masubsectionnumber} \vspace{.5cm} \noindent
{\large\em \arabic{masectionnumber}.\alph{masubsectionnumber} #1}
\par\vspace*{.2truecm}
\addcontentsline{toc}{subsection}
{\arabic{masectionnumber}.\alph{masubsectionnumber}\hspace{.1cm}
#1}
}
\renewcommand{\themasubsectionnumber}
{\mbox{\arabic{masectionnumber}.\alph{masubsectionnumber}}}
%%%%%%%%%%%%%%%%% Numbering %%%%%%%%%%
% Separate numbering of main theorems.
% Call with \begin{Thm}
%\newtheorem{Thm}{Theorem}
% Consecutive numbering of lemmas, props., coros.,
% in the form section.lemma
% call with \begin{thm}
\newtheorem{lem}{Lemma}[masectionnumber]
\newtheorem{thm}[lem]{Theorem}
\newtheorem{prop}[lem]{Proposition}
\newtheorem{cor}[lem]{Corollary}
\newtheorem{df}{Definition}[masectionnumber]
\newtheorem{rem}[lem]{Remark}
\newtheorem{claim}[lem]{Claim}
\renewcommand{\baselinestretch}{1.5}
\newenvironment{proof_claim}{
\addtolength{\leftmargini}{-1em}\begin{quotation} \noindent
{\bf Proof of Claim:}}{
\hspace*{\fill} $\diamond$
\end{quotation}\addtolength{\leftmargini}{1em}}
\newenvironment{proof_of}[1]{
\noindent{\bf Proof of #1:} \hspace*{1em} }{
\hfill \blackbox\bigskip}
\newenvironment{proof}{\noindent{\bf Proof:}
\hspace*{1em}}{\hfill \blackbox\bigskip}
\def\remark{ \noindent {\bf Remark}\hspace*{1em} }
%%%%%%%%%%%% appendix sections:
\newcommand{\startappendix}{ \setcounter{masectionnumber}{0}
%% Equations in the form (A.1)
\renewcommand{\theequation}
{\mbox{\Alph{masectionnumber}.\arabic{equation}}}
%% Theorems/Lemmas in the form (A.1)
\renewcommand{\thelem}
{\mbox{\Alph{masectionnumber}.\arabic{lem}}}
\addcontentsline{toc}{section}{Appendix }
}
%%resetsection counter
\newcommand{\maappendix}[1]{ % Appendix Section
\setcounter{equation}{0}
\refstepcounter{masectionnumber} \vspace{1truecm plus 1cm} \noindent
{\large\bf \Alph{masectionnumber}. #1}\par \vspace{.2cm}
\addcontentsline{toc}{section}{\Alph{masectionnumber}. #1 } }
%% - test:
\newcounter{masubapp}[masectionnumber]
\setcounter{masubapp}{0}
\newcommand{\masubappendix}[1]{ % Appendix Subsection
\refstepcounter{masubapp} \vspace{.5cm} \noindent
{\large\em \Alph{masectionnumber}.\alph{masubapp} #1}
\par\vspace*{.2truecm}
\addcontentsline{toc}{subsection}
{\Alph{masectionnumber}.\alph{masubapp}\hspace{.1cm} #1 }
}
% any subsections of Appendices -do by hand (or modify the code)
%%%%%%%%%%%
% \topmargin=-.5in
\begin{document}
\title { \vspace*{-1in}
{\Large Bounded Fluctuations and Translation Symmetry Breaking} \\
{\Large in One-Dimensional Particle Systems}{$^*$}
}
\author{M. Aizenman $^{a}$ \ \ \ \ S. Goldstein $^b$ \ \ \ \
J. L. Lebowitz $^{b}$ \vspace*{0.25truein} \\
\normalsize \it $^{(a)}$ Departments of Physics and Mathematics,
Princeton University, \vspace*{-.1truein} \\
\normalsize \it Jadwin Hall, P. O. Box 708, Princeton, NJ 08544. \\
% \vspace*{-.2truein} \\
\normalsize \it $^{(b)}$ Departments of Mathematics and
Physics, Rutgers University, \vspace*{-.1truein} \\
\normalsize \it 110 Frelinghuysen Road, Piscataway,
NJ 08854-8019. } % \vspace*{-0.2truein} }
\maketitle
\thispagestyle{empty} %removes # on p.1
\begin{abstract} \abst{ We present general results for one-dimensional
systems of point charges (signed point measures)
on the line with a translation invariant
distribution $\mu$ for which the variance of the total charge in an
interval is uniformly bounded (instead of increasing with the interval
length). When the charges are restricted to multiples of a common unit,
and their average charge density does not vanish, then the boundedness of
the variance implies translation-symmetry breaking --- in the sense that
there exists a function of the charge configuration that is nontrivially
periodic under translations --- and hence that $\mu$ is not ``mixing.''
Analogous results are formulated also for one dimensional lattice systems
under some constraints on the values of the charges at the lattice sites
and their averages. The general results apply to one-dimensional Coulomb
systems, and to certain spin chains, putting on common grounds different
instances of symmetry breaking encountered there.}
\end{abstract}
\noindent $^*$Dedicated to the memory of J. M. (Quin) Luttinger, a master
of one-dimensional systems and much more.
%\vfill
% \noindent {{\bf AMS subject Classification:} } \\
%\noindent {\bf Key words:}
%\vskip .25truecm
%\newpage
%
% \vskip .25truecm % Table of contents
% \begin{minipage}[t]{\textwidth}
% \tableofcontents
% \end{minipage}
% \vskip .5truecm
%\newpage
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\masect{Introduction}
\label{sect:intro}
% \vskip-0.6cm
%
% \masubsect{Overview}
To fluctuate is normal, and normally fluctuations grow like the square root
of the volume. There are however non-trivial exceptions to this rule;
among the notable examples are systems of charges with Coulomb interaction
and certain quantum spin chains in their ground states. Curiously, the
known one-dimensional systems with uniformly bounded fluctuations in the
total charge or component of spin in an interval exhibit also other
exceptional properties, such as translation symmetry breaking (in jellium
and quantum spin chains), or non-uniqueness of the Gibbs state (in the
two-component neutral Coulomb system). We prove here that this is not
accidental: under some general assumptions, bounded fluctuations in a
one-dimensional particle system imply the existence there of a periodic, or
in the lattice case a quasi-periodic, structure. In particular, in such
systems the correlations of certain local variables do not decay.
We briefly describe the examples mentioned above to which our results
apply.
\noindent{\bf a) Coulomb systems:} In general terms, a Coulomb system
consists of several ($m$) species of particles with charges $q_\alpha$,
$\alpha = 1,...,m$. For a classical system containing $N_\alpha$ particles
of species $\alpha$
in a finite domain $V\subset {\Bbb R}^d$, the
(configurational) Gibbs
canonical distribution is the probability measure on the space of
configurations $\{(x_j, \alpha_j)\}_{j=1,\ldots,N},$\break $ \ x_j \in V,\
\alpha_j=1,\ldots,m$, and $N=\sum_{\alpha=1}^m N_{\alpha}$,
given by the density $\exp[-\beta U(\{(x_j, \alpha_j)\}) ] / Z$, with
\begin{equation}
U(\{(x_j, \alpha_j)\}) \ =\
\sum_{i\ne j} q_j q_i \ V_{C}(x_i - x_j) + \sum_{j} q_j V_{bg}(x_j)
+ \sum_{i,j}V_{sr}^{(\alpha_i, \alpha_j)}(x_i - x_j) \; ,
\end{equation}
where $q_j \equiv q_{\alpha_j}$, $V_{C}(x)$ is the Coulomb
potential,
satisfying $\Delta V_{C}(x) = -\delta(x)$, $V_{bg}(x) $ is the
potential induced by a (uniform)
background charge, of {\em charge} density
$\rho_{bg}$, and $V_{sc}^{(\alpha,\alpha')}(x) $ is a short range
interaction between particles of
species $\alpha$ and $\alpha'$. For quantum systems
the measure on the space of configurations is given by a more complicated
formula; for details, we refer the reader to the recent review article by
Brydges and Martin~\cite{BM}.
Under suitable conditions, such measures,
with $\{ q_j \}$ and $\rho_{bg}$ not all of the same sign,
admit translation invariant infinite volume limits. In the
limiting states the total
charge density, including the background, is zero, i.e., the
{\em particle densities} $\nu_{\alpha}$
satisfy $\sum_{\alpha} q_{\alpha} \nu_{\alpha} + \rho_{bg} = 0$
[1-8].
It is further expected, and proven in some cases, including all classical
one-dimensional Coulomb systems \cite{Lenard,AM}, that with respect to
these limiting measures the variance of the net charge in a region $\Lambda
$ increases with $\Lambda $ only like its surface area~\cite{vB,L}. That is,
\begin{equation}\label{lambda}
\var(Q_\Lambda) :=\langle
Q_\Lambda^2 \rangle - \langle Q_\Lambda \rangle^2
\ \sim \ |\partial \Lambda| \qquad
\mbox{(as $\Lambda \nearrow {\Bbb R}^d $) },
\end{equation}
where $Q_\Lambda = \sum_{x_i \in \Lambda} q_{\alpha_i}$.
In one dimension this corresponds to the statement that the variance of the
charge in an interval $I$ remains bounded as $|I| \to \infty$.
Analogous statements apply to the case where the charges are
restricted to lattice sites~\cite{Gaudin}. (Eq. (\ref{lambda}) follows from
the ''zero sum rule,'' which holds whenever the charge correlations have
sufficiently rapid decay \cite{BM,vB}).
A simple example of a Coulomb system is the so-called {\em jellium} model,
or the {\em one-component plasma} (OCP), with particles of unit charge
dispersed in a negatively charged uniform background. For this system
surface growth of charge fluctuations, which now correspond to
particle-number fluctuations, has been proven in $d = 1$ \cite{Lenard,AM}
and in $d \geq 2$ at high temperatures $(\beta << 1)$ and in $d=2$ also
for $\beta=2$ \cite{BM,Janco,Imbrie}.
For the one-dimensional jellium (with the $1$D Coulomb potential
$V_{C}(x) = -\frac12|x|$) it is also known that the limits of the Gibbs
measures exhibit ``translation-symmetry breaking''~\cite{BL,AM}.
The periodic structure found there
has been regarded as an example of the `Wigner lattice.'
However, in this case the symmetry breaking
was also understood to be related
to the boundedness of the charge fluctuations,
the two being connected through the properties of the electric
field~\cite{AM} (for which the existence of the limit
was previously established in ref.~\cite{Lenard}).
Here we show that this relation is an example of a more general
phenomenon.
\noindent {\bf b) Spin chains:}
Another example of a one-dimensional system with reduced
fluctuations is provided by the ground state of certain
quantum spin systems. These consist of `chains'
(also of interest are arrangements into `ladders') of
quantum spins
$\{ \underline{\sigma}_{n} \}_{n \in \Z}$,
of a common spin $S$, such that $2 S$ is an integer,
with the Hamiltonian
\begin{equation}
H \ = \ - \sum_{n\in \Z} P^{(0)}_{n,n+1} \; ,
\end{equation}
where $ P^{(0)}_{n,n+1} $ is the projection onto the singlet state, i.e.,
on the subspace in which
$(\underline{\sigma}_{n} + \underline{\sigma}_{n+1})=0$.
This class of Hamiltonians was introduced by
Affleck~\cite{Affleck} as an interesting extension of the spin $1/2$
Heisenberg antiferromagnetic spin chain. (It is also related to
the classical Potts antiferromagnet with $Q= 2S+1$ \cite{AN}).
The `$z$-components' of the spins, $\{ \sigma^{(3)}_n \}$, form a family of
commuting observables. Thus in any quantum state their joint distribution
may be described by a `classical' probability measure on the space of
configurations of one-component spin variables with values ranging over
$\{-S,\ -S+1, \ldots,\ S\}$. To distinguish these from the full quantum spin
variables let us refer to their values as `charges.' For these
systems the ground state ($|0\rangle$) admits a
representation in which the charges are organized into
\underline{neutral clusters}~\cite{AN}, i.e., randomly organized
clusters of sites for which $\sum \sigma^{(3)}
=0$, with relative spin flip symmetry between different clusters.
The clusters can intermingle, so this perspective can be of value
only when the distribution of these clusters is such that for any
interval $I$, the number of sites in $I$ belonging to
clusters which are not entirely contained in the given interval
is seldom large -- in a sense uniform in $I$.
A condition implying that this scenario is realized
is rapid decay of correlations:
\begin{equation}
\sum_{n>0} n \,
| \langle0|\ \sigma^{(3)}_0 \sigma^{(3)}_{n} \ |0\rangle | \
<\ \infty \; .
\label{eq:dimerization}
\end{equation}
One can show, using the cluster representation, that when
\eq{eq:dimerization} holds
the two point function satisfies the neutrality condition
$\sum_{x} \langle0|\ \sigma^{(3)}_0 \sigma^{(3)}_{x} \ |0\rangle =
0$, which is analogous to the ``zero sum rule'' of Coulomb
systems~\cite{BM,vB,L},
and the block spins
$S_{I}= \sum_{n\in I} {\sigma}_{n}^{(3)}$ have uniformly bounded
variance, with
\begin{equation}
\langle0|\ |S_{I}|^2 \ |0\rangle| \ \le \ 2 \sum_{n>0} n \,
| \langle0|\ \sigma^{(3)}_0 \sigma^{(3)}_{n} \ |0\rangle| \ \; .
\end{equation}
The validity of the condition (\ref{eq:dimerization}) depends on $S$; exact
calculations indicate that (\ref{eq:dimerization}) is satisfied for all $S
\ge 1$, but not for $S=1/2$~\cite{Affleck,Bat-Bar,Kl}.
In \cite{AN} it was shown, using the aforementioned cluster
representation, that
for ``half-integer'' spins (with $2S$ an odd integer greater than one)
condition (\ref{eq:dimerization}) implies, in addition to the bound
on block spin fluctuations, also that the translation invariant
infinite volume ground state decomposes into a mixture
of two states of period $2$. The symmetry breaking which occurs in the
above spin systems is akin to {\em dimerization}, though the
neutral clusters need not consist just of pairs of neighboring spins.
(A notable fact is that the lack of mixing in the ground state is
expressed through other correlators than the two-point
correlation function, since the spin-spin correlation
does decay to zero.)
The general
results presented here place this translation symmetry breaking
within a broader context, as we show that it is
not coincidental that the same condition (\ref{eq:dimerization})
implies, for the quantum spin chains
with {\em odd} values of $2S$,
both bounded variance of
the block spins and translation symmetry breaking.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\masect{Formulation and results}
\label{sect:statement}
% \vskip-0.6cm
Our main results relate reduced fluctuations to
{\em translation symmetry breaking}. A more precise formulation
of the conclusion is that the state, given by a translation
invariant measure $\mu$ on the space of configurations, has a
{\em cyclic factor}. Let us first introduce the relevant terminology.
\masubsect{Cyclic factors and symmetry breaking}
\begin{df}\label{def}
A system described by a measure $\mu(d\omega)$,
on the space of configurations $\omega$,
which is invariant under either the continuous
group of translations ( shifts $T_x$ by $x\in \R$)
or the group of lattice shifts ( $x\in \Z$) is said to have
a \underline{cyclic factor}, of period
$0< \lambda < \infty$, if
there is a measurable function $\phi(\omega)$
with values in $[0,2 \pi)$ that evolves under the shifts by
\begin{equation}
\phi(T_x \omega) \ = \ \phi(\omega) - 2 \pi \, x / \lambda
\quad \mbox{({\rm mod} $2 \pi $)},
\label{eq:phi}
\end{equation}
\end{df}
with, in the lattice case, $\lambda^{-1}\notin\Z$.
An equivalent formulation
is that the measure $\mu$ can be decomposed into a mixture:
\begin{equation}\label{eq:cc}
\mu(\cdot) = \int_0^{ 2 \pi} \frac{d \theta}{2 \pi} \ \ \mu_{\theta}(\cdot)
\end{equation}
of mutually singular measures which are cycled under the shifts $T_x$:
\begin{equation}
T_x \ \mu_{\theta} \ =
\mu_{(\theta - 2\pi x / \lambda)_{({\rm mod}\ 2\pi)} } \; , \qquad
d\theta \ {\rm a.e.} .
\end{equation}
In the lattice case, if $\lambda$ is rational then $\frac{d\theta}{2 \pi}$
in (\ref{eq:cc}) should be replaced by, or understood as, a probability
measure $\nu(d\theta)$ invariant under $\theta\longmapsto \theta -
\frac{2\pi}{\lambda}\ ({\rm mod}\ 2\pi)$.
The two formulations are related by the observation that
in the decomposition (\ref{eq:cc}) the \underline{cyclic component}
$\mu_{\theta}(\cdot)$ is supported
by configurations $\omega$ with $\phi(\omega)=\theta$.
If the invariance is under the continuous group of translations, the
existence of a cyclic factor implies that the measure
$\mu$ is not ergodic under the smaller group of shifts by $\lambda$.
% That is not the case for lattice systems.
For lattice systems, which are invariant only under shifts by
multiples of the lattice period, if $\lambda$ is incommensurate with
the
period (here 1) then each of the lattice shifts may still act ergodically
on $\mu$ and the translation symmetry breaking
is expressed only through the existence of a quasi-periodic
structure; if the two lengths are commensurate then there is loss of
ergodicity under shifts by some {\em multiple} of $\lambda$.
Nevertheless, in either the continuum or the lattice cases, the existence
of a cyclic factor implies that $\mu$ does not have good clustering
properties. In particular, the {\em mixing} condition, which implies that
for any bounded function $g(\omega)$ with expectation $\int g\,d\mu =0$
\begin{equation}
\int g\,T_x g\, d\mu \to 0 \; , \quad \mbox{as $|x|\to \infty$}\;,
\end{equation}
is not satisfied.
Let us note also that it
follows from (\ref{eq:phi}) that $\phi(\omega)$ is a tail function
(under translations), in fact
measurable with respect to the $\sigma$-algebra at $+\infty$ (or $-\infty$).
Thus, if
$\mu$ is a Gibbs state for some potential, or weakly
Gibbs~\cite{MS}, then
its cyclic components should also be Gibbs states, in the appropriate
sense, for that potential.
\masubsect{Results}
Our first result concerns systems of particles in the continuum. The two
cases of primary interest to us are: i) point processes, described by
measures on the space of locally finite particle configurations on the
line $\R$, and ii) point-charge processes on $\R$. A process of either
type may be described in terms of a random atomic measure on $\R$ for which
the collection of atoms (the locations of the points or the charges) is
(a.s.) locally finite. In the former case the measure, which will be
denoted by $N_{\omega}(I)$, is the number of particles in the
interval $I$; in the
latter case the measure is the sum of the point charges in $I$,
$Q_{\omega}(I)=\sum_{x_i\in I}q_{\alpha_i}$, in the notation
of example {\bf 1.a}.
\begin{thm}
Let $\mu(d\omega)$ be a translation invariant probability measure
describing a point process on $\R$.
If the variance of the number of points in an interval
$I$ (i.e., of the random variable $N_I(\omega)\equiv N_{\omega}(I)$)
is bounded uniformly in the size of the interval
\begin{equation}
\E\left([N_I - \E(N_I)]^2 \right) \ \le \ C \ (< \infty)
\label{eq:variance}
\end{equation}
then $\mu$ has a cyclic factor, of period
\begin{equation}
\lambda \ = \ 1 /\, \E(N_{[0,1]}) \; .
\end{equation}
More generally, let $\mu$ describe a translation invariant point-charge
process on the line, with the charge values restricted to
multiples of a common unit $e$.
Assume that for some $\rho \ne 0$ the family of random
variables $\{ F_{\omega}(a,b) \}$, defined by
\begin{equation}
F_{\omega}(a,b) \ = \
Q_{\omega}((a,b]) - \rho (b-a)
\label{eq:F}
\end{equation}
where $-\infty < a < b < \infty$
and $Q_\omega(I)$ is the total charge in the interval $I$, is
\underline{tight}.
Then the measure $\mu$ has a cyclic factor, of period
\begin{equation}
\lambda = e / \rho \; .
\end{equation}
\label{thm:1}
\end{thm}
The tightness of the family of random variables $\{ F_{\omega}(a,b)
\}_{a,b}$ means that the probabilities of large fluctuations have
asymptotically vanishing bounds that are uniform in $a,b \in \R$:
\begin{equation}
\P(\, | F_{\omega}(a,b)| \ge t \, ) \ \le \ p(t)
\end{equation}
with $p(t) \searrow 0$ as $t \to \infty$.
A sufficient condition for this
is that the variances of $Q_I$ are uniformly bounded, as in
(\ref{eq:variance}).
The quantity $\rho$ introduced in Theorem~\ref{thm:1}
is the {\it asymptotic mean\/} of the charge density.
Under the tightness condition formulated above
it is well defined even if
$\E(\, |Q_{\omega}(I)|\, ) = \infty$. Note also that ergodicity is
not assumed in Theorem~\ref{thm:1}.
% Let us also note that a sufficient condition
% for the tightness of the fluctuation variables $\{ F(a,b)
% \}_{a,b}$
% is that the variances of $Q_I$ are uniformly bounded, as in
% (\ref{eq:variance}).
% In that case, $\rho = \E(\, Q_{\omega}(I)\, ) / |I|$.
Our second result deals with systems of charges $\{ q_k \}_{k \in \Z}$
on the one-dimensional lattice $\Z$.
Here, for any interval $I\subset \Z$, the charge in $I$ is given by
$ Q_{\omega}(I)\ = \ \sum_{k\in I} q_k $, with $q_k $ the charge
variable at the lattice site $k \in \Z$.
\begin{thm}
Let $\mu$ describe a translation invariant system of charges on the
lattice $\Z$ with the following properties:
\begin{enumerate}
\item[a.]
The charges ($q_k$) are restricted to values of the form
$ (\gamma + n \, e) $
with $\gamma$ and $ e$ fixed, and $n$ taking only integer values.
\item[b.]
For some $\rho \in \R $ the family of random
variables
$\{ F_{\omega}(a,b) \equiv Q_{\omega}((a,b]) - \rho (b-a) \}$,
with $ a, b\in \Z$ and $a**0} \P_{\mu}(\, |F_{\omega }(-u,0)| \ge t\, )\; .
\label{eq:tight}
\end{eqnarray}
Under the tightness assumption of Theorem \ref{thm:derivative},
the right hand side vanishes as $t\to \infty$.
The above considerations suggest that compactness and continuity arguments
can be invoked to prove that the sequence of probability measures
$\nu_{L}(\cdot)$ has a convergent subsequence, as $L \to \infty$, and that
the limiting measure is strictly stationary under the flow $\widetilde
T$. The existence of a suitable limiting measure can indeed be proven using
a topology which is natural for $\omega$ in the present set-up (it is at
this point that we leave the argument at the level of a sketch), for which
\eq{eq:tight} implies that the sequence of measures $\nu_{L}(\cdot)$ is
tight, and thus has a convergent subsequence, with limit $\tilde\mu$.
Clearly the $\Omega-$marginal of $\tilde\mu$ is $\mu$. That $\tilde\mu$ is
also stationary would be immediate were $\widetilde T_x$ continuous on
$\widetilde \Omega$. However, there is here a slight complication in that
the second component on the right-hand-side of \eq{eq:t-tilde} is not
continuous in $\omega$. Nevertheless, the set of configurations $\omega$
for which $\widetilde T_x \widetilde \omega$ is discontinuous---those
configurations having atoms at $0$ or at $x$---has $\mu$-measure $0$, (an
easy consequence of the stationarity of $\mu$). Using this fact, the
conclusion of stationarity follows.
The above implies that in the larger space
$\widetilde \Omega$, the process
$X_{\omega}(x)$ has a stationary antiderivative $Y_{\widetilde
\omega}(x)$. While the antiderivative
thus obtained need not be determined by
$\omega$ alone, its stationarity implies that it has the necessary
asymptotic statistical regularity to yield
an antiderivative of the field $X_\omega$ which is a covariant
function of $\omega$ alone. This can be done
by selecting from among the one-parameter family of possible
antiderivatives (differing only by an overall shift)
the one whose median value, averaged along the
positive $x$-axis, is set at $Y=0$.
The median level for $Y(\cdot)$ is defined by
\begin{equation}
M[Y] \ = \inf_{M} \left\{ M \, :
\lim_{L\to \infty} \frac1L \int_0^L I[Y(x) \ge M] \, dx \ \le 1/2
\,\right\} \; ,
\end{equation}
provided the limit exists.
For $Y_{\widetilde \omega}(\cdot)$, which is sampled with a stationary
distribution, the existence of the limit (simultaneously for a countable
collection of $M$) follows by the ergodic theorem. Using the median, we define
\begin{equation}
{\mathcal E}_{\widetilde \omega}(x) \ =
\ Y_{\widetilde \omega}(x) - M[Y_{\widetilde \omega}] \; .
\end{equation}
Because $M[Y+\mbox{Const.}]=M[Y]+\mbox{Const.}$, ${\mathcal E}_{\widetilde
\omega}$
depends only on $\omega$ (and not on the other coordinate of $ \widetilde
\omega $), and thus it defines a suitable covariant antiderivative of
$X_{\omega}(x)$.
\end{proof_of}
\noindent{\bf Remarks:}
\noindent 1. Under the additional assumption that \eq{eq:uniform} holds,
the stationary antiderivative $Y_{\widetilde\omega}$ constructed in the
first step of the argument above satisfies
$\E[Y_{\widetilde\omega}(0)]<\infty$, as may easily be seen from the
construction of $\tilde\mu$. One may then alternatively define an
antiderivative functional ${\mathcal E}$ using the average $\bar Y$ of $Y$,
$\bar Y=\lim_{L\to \infty} \frac1L \int_0^L Y(x) \, dx $, in place of the
median $M[Y]$. Moreover, for this antiderivative we may obtain an explicit
formula: Averaging the expression ${\mathcal E}_{\omega}(0) - {\mathcal
E}_{\omega}(u)= -F_\omega(0,u)$ with respect to $du/L$ over the interval
$(0,L]$ and taking the limit $L \to \infty$, using the fact that, by
construction, $\lim_{L\to \infty} \frac1L \int_0^L {\mathcal E}_\omega(u)
\, du =0$, one arrives, in fact, at the formula given by \eq{eq:explicit}
(see ref.~\cite{AM} for further discussion of \eq{eq:explicit}).
\smallskip
\noindent 2. Suppose $X_{\omega}(x)$ is ergodic. Then it is easy to
see---either by using the covariant antiderivative of Theorem
\ref{thm:derivative} or directly---that a stationary antiderivative
$Y_{\tilde\omega}(x)$ of $X_{\omega}(x)$ is a covariant functional of
$\omega$ alone if and only if $Y_{\tilde\omega}(x)$ is ergodic. Moreover,
two ergodic antiderivatives of $X_{\omega}(x)$ differ by an absolute
constant. Thus if $X_{\omega}(x)$ is ergodic, instead of proceeding to step
2 as described above, we could obtain a covariant antiderivative ${\mathcal
E}$ by simply decomposing $\tilde\mu$, corresponding to the antiderivative
$Y_{\tilde\omega}(x)$ from step 1, into its ergodic components and choosing
any one of these. Moreover, using the ergodic theorem, it follows from the
manner of construction of $\nu_L$ that the process $Y_{\tilde\omega}(x)$
from step 1 is a mixture of the antiderivatives ${\mathcal E}(x) - y$ with
weights $\mu_{{\mathcal E}(0)}(dy)$, i.e., with $y$ random, with distribution
that of ${\mathcal E}(0)$. In particular, the distribution of
$Y_{\tilde\omega}(0)$ is that of the difference of two independent copies
of ${\mathcal E}(0)$. Note also that it follows that all limits of $\nu_L,\
L\to\infty,$ must agree, independent of which convergent subsequence is
chosen, so that in fact $\lim_{L\to\infty}\mu_L$ itself exists,
without passage to a subsequence.
\bigskip
\masubsect{Proof of the main results}
\begin{proof_of}{Theorem~\ref{thm:1}}
It suffices to consider the point-charge processes, since any point process
may be regarded as a special case with unit charges. Under the assumption
of tight fluctuations, Theorem~\ref{thm:derivative} says that there is a
measurable function ${\mathcal E}(\omega)$ of the charge configuration such
that
\begin{equation}
{\mathcal E}(T_x\omega) \ = {\mathcal E}(\omega) + Q_{\omega}((0,x]) - \rho
x \; .
\end{equation}
Since the charge $Q_{\omega}((0,x])$ assumes only values which are integer
multiples of $e$, we may conclude that $\phi(\omega)\equiv
\frac{2\pi}{e}{\mathcal E}(\omega)\ ({\rm mod}\ 2\pi)$ defines a cyclic
factor as described in Theorem~\ref{thm:1} and Definition \ref{def}.
\end{proof_of}
\begin{proof_of}{Theorem~\ref{thm:2}}
The argument in the discrete case is similar to the above, with the only
notable difference being that the charge $Q_{\omega}(\, (0,x]\, )$ now includes
$\gamma x$ in addition to an integer multiple of $e$. Thus in this case the
``antiderivative functional'' ${\mathcal E}(\omega)=g(T\omega)$ provided by
Theorem~\ref{thm:cocycle} satisfies
\begin{equation}
{\mathcal E}(T_x\omega) \ = {\mathcal E}(\omega) + (\gamma- \rho) x \quad ({\rm
mod}\ e)
\end{equation}
and hence $\phi(\omega)\equiv
\frac{2\pi}{e}{\mathcal E}(\omega)\ ({\rm mod}\ 2\pi)$ defines a cyclic
factor as described in Theorem~\ref{thm:2} and Definition \ref{def}.
\end{proof_of}
\masect{Extensions}
\label{sect:newapp}
\masubsect{Jellium tubes}
Our discussion of one-dimensional systems applies also to elongated tubes
which are locally of higher dimension, but are of finite cross section and
infinite in one direction. An example of such a system is the three
dimensional OCP with the flux lines of the three dimensional field
restricted to stay in the tube. (this is an idealized situation, possibly
mimicking some high contrast dielectric materials, or a Kalutza-Klein model
with only one unconfined dimension.) If the variances of the total charge
in tubes of length $L$ stays bounded, as may be expected under the Coulomb
interactions, then the theory presented here implies translation symmetry
breaking, along the unconfined direction. In such a situation the
projections of the positions of the point charges along the free direction
yield a one dimensional system of the type discussed in this paper. For
$d=2$ this system is considered in \cite{Choquard} for the explicitly
solvable case $\beta=2$.
\masubsect{$\Theta$-states of one-dimensional Coulomb gas}
Unlike the OCP the one-dimensional Coulomb gas consisting of a continuum
system of point charges of value $\pm e$, with respective densities
$\rho_+=\rho_-$, does not break translation symmetry, see Remark 1) of
Section 2.c. This system does however exhibit another form of
non-uniqueness of the Gibbs state: it admits a one-parameter family of
infinite-volume translation invariant Gibbs states, indexed by the
fractional part $(\Theta)$ of the boundary charge ``imposed at infinity''
\cite{AM,af}. While at first glance this ``anomaly'' appears to be of
a different kind than what is discussed here, let us point out that it can
also be viewed as a remnant of the translation symmetry breaking in
asymmetric charge systems.
The charge-symmetric model may be arrived at as a limit of Coulomb systems
with a uniform background of small charge density $\rho_{bg}=- \epsilon$,
and $\rho_+ = \rho_- +\epsilon $. In this situation our results do apply, and
prove that the system exhibits translation symmetry breaking, the
corresponding infinite volume Gibbs states having a cyclic factor of period
$\lambda=1 / \epsilon$. For each $\epsilon$ the resulting cyclic
components $\mu_\theta^{(\epsilon)}$ (of equation (\ref{eq:cc}) ) are
periodic under translations, with period $1/ \epsilon$. Moreover, the
dependence of $\mu_\theta^{(\epsilon)}$ on $\theta$ can be so chosen that
when the asymmetry parameter $\epsilon$ is taken down to zero, these
converge (locally) to a one-parameter family of translation invariant
states (i.e., probability measures on the space of configurations) which
form the $\Theta$-states of the symmetric Coulomb gas. The proof can be
obtained using the methods of ref.~\cite{AM}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%\vskip 1truecm
\noindent {\large \bf Acknowledgments\/} \\
We thank Yuval Peres for calling our attention to the relation of
our work with the theory of tight cocycles [22-24].
% \cite{Sch}).
MA gratefully acknowledges the hospitality accorded him at
Rutgers University. The work was supported in part by the NSF grants
PHY-9971149, DMR-9813268 and AFOSR Grant F49620--98--1--0207.
\addcontentsline{toc}{section}{References}
% \bibliography{fluctuations} % for Bibtex
\begin{thebibliography}{10}
\bibitem{BM} D.~C. Brydges and P.~A. Martin, ``Coulomb systems at low
density: A review,'' {\em J. Stat. Phys.}, {\bf 96}, 1163 (1999). See also
Ph. Martin, ``Sum Rules in Charged Fluids'', {\em Rev. Mod. Phys.}, {\bf
60}, 1075--1127 (1988).
\bibitem{Lenard}
A.~Lenard, ``Exact statistical mechanics of a one-dimensional system with
{C}oulomb forces. {I}{I}{I}. {S}tatistics of the electric field,'' {\em J.
Math. Phys.}, {\bf 4}, 533 (1963).
\bibitem{LL1} J.~L. Lebowitz and E.~H. Lieb, ``The existence of
thermodynamics for real matter with {C}oulomb forces,'' {\em
Phys. Rev. Lett.}, {\bf 22}, 631 (1969). E.~H. Lieb and J.~L. Lebowitz,
``The constitution of matter: {E}xistence of thermodynamics for systems
composed of electrons and nuclei,'' {\em Advances in Math.}, {\bf 9}, 316,
(1972).
\bibitem{AM}
M.~Aizenman and P.~A. Martin, ``Structure of {G}ibbs states of one-dimensional
{C}oulomb systems,'' {\em Comm. Math. Phys.}, {\bf 78}, 99. (1980).
\bibitem{vB} H.~van Beijeren and B.U.~Felderhof, {\em Mol. Phys.}, {\bf 38},
1179 (1979); Ch.~Gruber, Ch.~Lugrin, and Ph.~A.~ Martin, {\em
J. Stat. Phys.}, {\bf 22}, 193 (1980); Ph.A.~Martin and T.~Yalcin, {\em
J. Stat. Phys.}, {\bf 22},435 (1980).
\bibitem{L} J.L.~Lebowitz, {\em Phys.\ Rev.\ A} {\bf 72}, 773 (1983);
J.L.~Lebowitz in {\em Strongly Coupled Coulomb Systems},
G.J.~Kalman, J.M.~ Rommel and K.~Blagoev, eds., Plenum Press, New
York and London, 1998; D.~Levesque, J.-J.~Weis and J. L.~Lebowitz, ``Charge
Fluctuations in the Two-Dimensional One-Component Plasma'', {\em
J. Stat. Phys.}, {\bf 100}, 209-222, 2000.
\bibitem{Gaudin} M. Gaudin, {\em J. Phys.}, {\bf 34}, 511 (1973).
\bibitem{BL} H.~J. Brascamp and E.~H. Lieb, ``Some inequalities for
Gaussian measures and the long-range order of the one-dimensional
plasma,'' in Functional Integration and its Applications, A.~M.
Arthurs (ed.), (Oxford Univ. Press, 1975).
\bibitem{Janco} B.~Jancovici, {\em Phys. Rev. Lett.}, {\bf 46}, 366, (1981).
\bibitem{Imbrie} J.~Imbrie, {\em Commun. Math. Phys.}, {\bf 87}, 515, (1983).
\bibitem{Affleck}
I.~Affleck, ``Exact results on the dimerization transition in {SU}(n)
antiferromagnetic chains,'' {\em J. Phys. C: Cond. Matt.}, 405, (1990).
\bibitem{AN}
M.~Aizenman and B.~Nachtergaele, ``Geometric aspects of quantum spin states,''
{\em Comm. Math. Phys.}, {\bf 164}, no.~1, 17, (1994).
\bibitem{Bat-Bar}
M.~T. Batchelor and M.~N. Barber, ``Spin-$s$ quantum chains and
{T}emperley-{L}ieb algebras,'' {\em J. Phys. A}, {\bf 23}, no.~1, L15--L21,
(1990).
\bibitem{Kl}
A.~Kl{\"u}mper, ``The spectra of $q$-state vertex models and related
antiferromagnetic quantum spin chains,'' {\em J. Phys. A}, {\bf 23}, no.~5,
809, (1990).
\bibitem{af} M.~Aizenman and J.~Fr{\"o}hlich, {\em J. Stat. Phys.},
``States of one-dimensional Coulomb systems as simple examples of
$\Theta$-vacua and confinement'', {\bf 26}, 247 (1981).
\bibitem{MS} Ch.~Maes, A.~van Moffaert, and F.~Redig, ``Almost Gibbsian
versus Weakly Gibbsian Measures'', {\em Stoch. Proc. and Appl.}, {\bf 79},
1--15, (1999).
\bibitem{Mehta}
M.~L. Mehta, {\em Random Matrices}.
\newblock Academic Press, New York, 1990, 2nd ed.
\bibitem{Beck} J.~Beck, {\em Acta Math}, 1, {\bf 159}, (1987).
\bibitem{Choquard} P.~Choquard, P.J.~Forrester and E.R.~ Smith, ``The Two
Dimensional One Component Plasma at $T=2$; The Semiperiodic Strip'', {\em
J. Stat. Phys.}, {\bf 33}, 13 (1983); B.~Jancovici, private communication.
\bibitem{EntFerSok}
A.~C.~D. van Enter, R.~Fern{\'a}ndez, and A.~D. Sokal, ``Regularity properties
and pathologies of position-space renormalization-group transformations:
scope and limitations of {G}ibbsian theory,'' {\em J. Stat. Phys.}, {\bf
72}, no.~5-6, 879--1167, (1993).
\bibitem{Gg} H.-O.~Georgii, ``Gibbs Measures and Phase Transitions'', W. De
Gruyter, Berlin (1988).
\bibitem{Sch}
K.~Schmidt, {\em Cocycles of Ergodic Transformation Groups}.
\newblock Lect. Notes in Math. Vol. 1, {MacMillan Co. of India}, 1977.
\bibitem{Leo}
V.~Leonov, ``On the dispersion of time averages of a stationary random
process,'' {\em Theory Probab. Appl.}, {\bf 6}, 93, (1961).
\bibitem{Aa-W}
J.~Aaronson and B.~Weiss, ``Remark on tightness of cocycles,''
\newblock 1999 preprint.
\end{thebibliography}
\end{document}
---------------0007141802712--
**