Symmetry Results for Finite-Temperature, Relativistic Thomas-Fermi Equations
Extracted source text ยท arXiv math-ph/0012021
\documentstyle[amsfonts,amssymb,epsfig,12pt]{article}
%\parindent 0pt
%\parskip 0.5\baselineskip
%\sloppy
\setlength{\textwidth}{460pt}
\setlength{\voffset}{-12pt}
\setlength{\hoffset}{-12pt}
\setlength{\textheight}{530pt}
\makeatletter
\renewcommand \thesection {\@arabic\c@section}
\renewcommand\thesubsection {\thesection.\@arabic\c@subsection}
\renewcommand\thesubsubsection{\thesubsection .\@arabic\c@subsubsection}
\renewcommand\theparagraph {\thesubsubsection.\@arabic\c@paragraph}
\renewcommand\section{\@startsection {section}{1}{\z@}%
{-3.5ex \@plus -1ex \@minus -.2ex}%
{1.9ex \@plus.2ex}%
{\normalfont\large\bfseries\centering}}
\renewcommand\subsection{\@startsection{subsection}{2}{\z@}%
{-2ex\@plus -1ex \@minus -.2ex}%
{1.2ex \@plus .2ex}%
{\normalfont\normalsize\bfseries\centering}
}
\renewcommand\subsubsection{\@startsection{subsubsection}{3}{\z@}%
{-2ex\@plus -1ex \@minus -.2ex}%
{.5ex \@plus .2ex}%
{\normalfont\normalsize\em}}
\renewcommand\paragraph{\@startsection{paragraph}{4}{\z@}%
{3.25ex \@plus1ex \@minus.2ex}%
{-1em}%
{\normalfont\normalsize\em}}
\renewcommand\subparagraph{\@startsection{subparagraph}{5}{\parindent}%
{3.25ex \@plus1ex \@minus .2ex}%
{-1em}%
{\normalfont\normalsize\em}}
\makeatother
\newcounter{subequation}
\newenvironment{subequation}%
{\addtocounter{equation}{-1}%
\stepcounter{subequation}%
\def\theequation{\arabic{equation}\alph{subequation}}%
\begin{equation}}%
{\end{equation}%
\def\theequation{\arabic{equation}}}
\newcommand{\resetseq}{\setcounter{subequation}{0}}
\newcommand{\stepeq}{\stepcounter{equation}}
%%%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\beq}{\begin{equation}}
\newcommand{\eeq}{\end{equation}}
\newcommand{\bseq}{\begin{subequation}}
\newcommand{\eseq}{\end{subequation}}
\newcommand{\bea}{\begin{eqnarray}}
\newcommand{\eea}{\end{eqnarray}}
\newcommand{\refeq}[1]{(\ref{#1})}
\newcommand{\op}[1]{{\bf #1}}
\newcommand{\dd}{{\mathrm d}}
\newcommand{\pd}{\partial}
\newcommand{\rhs}{{\rm right\ hand\ side\ }}
\newcommand{\lhs}{{\rm left\ hand\ side\ }}
\newcommand{\rank}{{\rm rank\ }}
\newcommand{\sign}{{\rm sign\ }}
\newcommand{\kb}{k_{\rm B}}
\newcommand{\eps}{\epsilon}
\newcommand{\lf}{\left}
\newcommand{\rg}{\right}
\newcommand{\lfa}{\lf\langle}
\newcommand{\rga}{\rg\rangle}
\newcommand{\noin}{\noindent}
\newcommand{\cG}{{\cal G}}
\newcommand{\cN}{{\cal N}}
\newcommand{\cM}{{\cal M}}
\newcommand{\RR}{{\Bbb R}}
\newcommand{\SS}{{\Bbb S}}
\newcommand{\NN}{{\Bbb N}}
\newcommand{\ZZ}{{\Bbb Z}}
\newcommand{\pr}{\prime}
\newcommand{\ppr}{{\prime\prime}}
\newcommand{\QED}{{\rm Q.E.D.}}
\begin{document}
\title{SYMMETRY RESULTS FOR FINITE-\\
TEMPERATURE, RELATIVISTIC\\
THOMAS-FERMI EQUATIONS$^*$}
\author{Michael K.-H. Kiessling\\
\textit{Department of Mathematics, Rutgers University}\\
\textit{110 Frelinghuysen Rd., Piscataway, N.J. 08854}}
\bigskip
\date{Original version: Dec. 06, 2000; final: Dec. 14, 2001.
To appear in: \textbf{Communications in Mathematical Physics}}
\maketitle
\begin{abstract}
\noindent
In the semi-classical limit the relativistic quantum mechanics
of a stationary beam of counter-streaming (negatively charged) electrons and
one species of positively charged ions is described by a nonlinear system of
finite-temperature Thomas--Fermi equations.
In the high temperature / low density limit these
Thomas--Fermi equations reduce to the (semi-)conformal system of
Bennett equations discussed earlier by Lebowitz and the author.
With the help of a sharp isoperimetric inequality it is
shown that any hypothetical particle density function which is not
radially symmetric about and decreasing away from the beam's axis
would violate the virial theorem.
Hence, all beams have the symmetry of the circular cylinder.
\end{abstract}
\vskip 1truecm
\smallskip
\hrule
\smallskip
\noindent
$^*$\textit{In celebration of the $70^{th}$ birthday of Joel L. Lebowitz.}
\smallskip
\noindent
\copyright{2001} The author. Reproduction of this article, in its
entirety, for non-commercial purposes is permitted.
\newpage
\section{INTRODUCTION}
Modern books on charged-particle beams,
e.g.~\cite{humphries}, usually contain a chapter about the Bennett
model~\cite{bennettPAPa}, but back in the early 50's when regular
research on charged-particle beams came into sharper focus, W.H. Bennett's
pioneering pre-WWII paper~\cite{bennettPAPa} on the statistical mechanics
of a relativistic, stationary particle beam had been forgotten, apparently,
and so in 1953 Bennett sent out a reminder note~\cite{bennettPAPb}.
For some reason or other, Bennett's note did not appear until
1955~\cite{bennettPAPb}, the very year when Joel L. Lebowitz was launching
his stellar career~\cite{joelA} with center of gravity in stationary
non-equilibrium statistical mechanics~\cite{joelB, joelC, joelD}.
At that time, a single issue of {\it The Physical Review} was still
of a decent size and could be consumed from first to last page by an
individual with huge scientific appetite such as Joel, and Bennett's
note~\cite{bennettPAPb} did not pass unnoticed before Joel's hungry eyes.
All this happened a few years before I was born, but when I came
to spend some postdoctoral time with Joel nearly 40 years later,
several interesting questions raised by Bennett's work were still unanswered,
and so we began to answer some of these~\cite{kiejllPhPl}.
One of the problems we had to leave open was that of the symmetry
of a beam.
Following Bennett we only inquired into circular-cylindrically
symmetric solutions.
While it is a natural conjecture that in the absence of
external fields an unbounded straight particle beam with finite electrical
current through its cross-section necessarily posesses the symmetry of the
circular cylinder, how to prove it is not quite so obvious.
It is with great pleasure when in this paper I present a
rigorous proof to Joel.
Fitting for the occasion, the proof of the cylindrical
symmetry of the beam involves statistical mechanics in an essential way.
Namely, it is shown that any hypothetical
stationary beam with finite electrical current whose particle density
functions are not radially symmetric about and decreasing away from
the beam's axis would violate the virial theorem for this many-particle
system.
This symmetry proof covers Bennett's strictly classical model
as well as its semi-classical upgrade, i.e. a system of relativistic,
finite-temperature Thomas--Fermi equations which in the high-temperature
/ low-density limit reduce to the (semi-)conformal Bennett equations.
The proof is, however, restricted to a system of merely two
equations because the coefficient matrix for the beam equations has rank 2.
Our symmetry theorem therefore does not apply to beams that consist
of the negatively charged electrons and more than one, differently positively
charged ion species; but then again, our method of proof not only yields the cylindrical
symmetry of the beam, it also yields monotonic radial decrease of
the particle density functions.
Hence, it is conceivable that monotonicity of the density
functions may be violated in an electron / multi-ion species beam
while cylindrical symmetry might still hold --- yet to prove that
would seem to require an entirely new argument.
Incidentally, our result also sheds some new light on the
theory of white dwarfs~\cite{chandra}.
These Earth-sized, expired stellar objects shine in bright white light
because they are still incredibly hot compared to our Sun, yet they are
relatively cold compared to their Fermi temperature and therefore
essentially in their quantum ground state.
This justifies using zero-temperature Thomas--Fermi theory for
the description of their overall structure~\cite{chandra} --- a fortunate
happening, for finite-temperature Thomas--Fermi theory could not be
used in three dimensions since it does not have solutions with finite mass.
Interestingly, the finite-temperature Thomas--Fermi equations of the
two-dimensional caricature of such a white dwarf star should have solutions
with finite mass, because the gravitational potential in two dimensions is
sufficiently strongly confining for this purpose.
In any event, relatively little is known
rigorously\footnote{More is known rigorously~\cite{chandra, liebyau}
for the locally neutral approximation of the model,
where the positive and negative charges are distributed
identically and Coulomb's law is discarded.
In particular, radial symmetry of solutions for
this locally neutral model has been proven by energy minimization
through radial rearrangement~\cite{liebyau}.
We remark that due to the enormous ratio of the electrical and
gravitational coupling constants the locally neutral approximation
is expected to be an excellent approximation for a white dwarf;
however, this is not generally the case for a
particle beam, where the ratio of electric and magnetic
coupling constants may be arbitrarily close to $1$.}
for such a gravitating plasma of negative electrons and positive nuclei
(all species treated as fermions) in either two or three dimensions; see
the discussion of this model by W.E. Thirring in the preface to the E.H. Lieb
jubilee volume~\cite{liebJUBILEE}, where Thirring gives an amusing account of
the pitfalls associated with the fact
that the Thomas--Fermi equations are the Euler--Lagrange
equations for the saddle points of a variational functional.
When dealing with saddle points, existence and symmetry of
solutions via minimization by radial decreasing
rearrangement~\cite{almgrenlieb, brothersziemer, carlenloss, kawohl}
is not an option, and neither is symmetry via uniqueness by
convexity~\cite{kawohl} of the functional.
Now recall that by the Biot-Savart law the magnetic interactions
of straight, parallel electrical current filaments are attractive, with
a distance law that is identical to the Newtonian gravity law in two
dimensions.
From this it follows that the finite-temperature Thomas--Fermi
beam equations are identical to the finite-temperature Thomas--Fermi
equations of the two-dimensional caricature of a white dwarf model,
with the magnetic flux function re-interpreted as the gravitational
Newton potential in two dimensions, and the mean electric current of
each species (positive after at most a joint space rotation) re-interpreted
as the mass of that species.
Our symmetry result can be rephrased thus: two-dimensional
finite-temperature white dwarfs are radially symmetric.
Our proof of symmetry, which is based on the
Rellich~\cite{rellich}--Pokhozaev~\cite{pokhozaev} identity
(which expresses the virial theorem) and the classical
isoperimetric inequality~\cite{bandleBOOK, brothersziemer},
does involve radial rearrangements in a
strategy that goes back at least as far as~\cite{bandleBOOK},
where it is applied to Liouville's
equation\footnote{The elliptic Liouville equation, known from
two-dimensional differential geometry, is meant
and not the evolution equation on phase space
known from statistical mechanics.}
in a disk $\subset\RR^2$~\cite{bandleBOOK}.
In~\cite{chakieGAFA} this strategy was generalized to systems of
PDEs of Liouville type in all $\RR^2$ which are unrestricted in size but
which have a symmetric, fully stochastic coefficient matrix of full rank.
The Bennett equations also constitute a Liouville system, but
are not covered by the theorem of~\cite{chakieGAFA} because
their coefficient matrix is generally not symmetric, has some
negative elements, and is always rank 2.
The present paper develops the necessary generalizations
of~\cite{chakieGAFA} to overcome the first two peculiarities of the
Bennett equations, but the rank 2 restricts the proof to a system of
two equations.
By adapting the treatment of single PDEs with more general
nonlinearities developed in~\cite{lionsPAPa} (cf. also~\cite{kesavanpacella})
and~\cite{chakieCR, chakieNA} to the system case we are able to extend our
proof of symmetry for the Bennett equations to the relativistic,
finite-temperature Thomas--Fermi beam equations.
Our proof simplifies considerably when the systems of
Thomas--Fermi and Bennett equations are restricted to a disk
with $0$-Dirichlet boundary conditions for the electric and
magnetic potentials.
In this compact case, an alternate proof of the radial
symmetry and decrease of the solutions to systems of PDE which
includes the finite-temperature Thomas--Fermi and Bennett equations,
was given by Troy~\cite{troy}, who exploited Alexandroff's method of
moving planes.
For more on the moving-planes method,
see~\cite{serrin, gidasninirbg, li, chenli, chakieCMP, chakieDUKE}.
Troy's proof has been extended to Liouville systems in
unbounded domains, the Bennett equations not included though,
in~\cite{chipotshafrirwolansky}.
Presumably, the moving planes method can be made to work
also for the system of Thomas--Fermi equations studied here; however,
this is not done in this paper.
While the present paper addresses only the question whether
invariance of the PDEs under rotations implies radial symmetry of
their solutions, these PDEs feature other symmetries which deserve
mentioning.
The system of Thomas--Fermi equations is invariant under
the isometries of Euclidean space, simple gauges, and Lorentz boosts
along the beam.
The Bennett equations are in addition to that invariant under
isotropic scaling in $\RR^2$, and for a special family of parameter
values also under Kelvin transformations, in which case
they are invariant under the Euclidean conformal group of $\RR^2$.
In this fully conformal case the conformal orbit of the finite current
solutions is connected and itself invariant~\cite{kiejllPhPl}.
Invariance under the Euclidean conformal group holds also for
the Liouville systems studied in~\cite{chakieGAFA}, but their conformal
orbit of finite mass solutions is generally not connected, and each
component not invariant under inversions~\cite{chipotshafrirwolansky}.
Toda systems in $\RR^2$, which are Liouville systems with
symmetric coefficient matrix given by the $SU(N)$ Cartan
matrix, are studied in~\cite{jostwangA, jostwangB}.
The distribution of negative and positive signs in the $SU(2)$
Cartan matrix is opposite to that in our Bennett equations, and sure
enough, our radial symmetry proof fails in this case.
Interestingly, in this case one can show that radial symmetry is
in fact broken by some solutions, see the bifurcation argument
with $n=2$ in (1.7) of~\cite{chakieGAFA}, and see~\cite{jostwangB} for
the construction of the complete solution family with finite masses.
Another interesting topic not discussed further here is
whether the translation invariance along the beam can be broken, as
is suggested by various dynamical beam instabilities~\cite{weinberg}.
The remainder of this paper is structured as follows.
In the next section we formulate the basic equations of
the semi-classical beam model and its classical limit.
Existence of solutions is briefly touched upon.
In section 3 we state our two main theorems, and in section 4
we present their proofs.
%\newpage
\section{RELATIVISTIC BEAM EQUATIONS}
We let $a\in\SS^2$ denote the fixed axis of the beam,
$x\in\RR^2$ a point in the cross-section of the beam containing
the coordinate origin, and $p\in \RR^3$ the kinematical particle
momentum.
The self-consistent electric field of the beam is given by
$E(x)=-\nabla\phi(x)$, where $\phi$ is the electric potential,
and the magnetic field by $B(x)= \nabla \psi(x)\wedge a$, where
$\psi$ is the magnetic flux function.
The beam consists of spin $1/2$ electrons (negatively charged,
thus indexed by $s=-$) and one species of positively charged spin $1/2$
fermions (indexed by $s=+$),
characterized by the following parameters: the particle charges $q_s$ and rest
masses $m_s$; the rest frame temperatures $T_s$; the external
chemical potentials $\mu_s$; and lab frame drift speeds $c\nu_s$,
where $c$ is the speed of light and $\nu_s\in (-1,1)$.
We demand $\nu_+\neq \nu_-$, as appropriate
for counter-streaming particle species.
The temperatures and drift speeds combine into the thermal
lab frame parameters $\beta_s^{-1} = \kb T_s \sqrt{1 -\nu_s^2}$.
\subsection{The semi-classical model (Thomas--Fermi theory)}
The finite-temperature Thomas--Fermi model of a straight,
relativistic beam is set up as follows.
In the lab frame the density of $s$-particles at $x$ is given by
$\rho_s(x)={\cG}_s^{\rm TF}\!\left(\phi,^{\phantom{b}}\!\!\psi\right)(x)$,
where
\beq
{\cG}_s^{\rm TF}(\phi,\psi)
=
\frac{2}{h^3}\int_{\RR^3}
\frac{\dd p}{ 1+e^{-\beta_s\!
\left(\mu_s-c\sqrt{m_s^2c^2+|p|^2}+\nu_scp\cdot a-q_s[\phi-\nu_s\psi]\right)}}
\label{TFdensity}
\eeq
is the finite-temperature Thomas--Fermi density function for the
relativistic $s$-species, which is subjected to the integrability condition
\beq
\int_{\RR^2}
{\cG}_s^{\rm TF}\!\left(\phi,^{\phantom{b}}\!\!\psi\right)(x)\dd x
=
N_s,
\label{Ns}
\eeq
where $N_s$ is the number of $s$-particles per unit length of beam.
The phase-space density function under the integral in
\refeq{TFdensity} is the drifting Fermi--Dirac--J\"uttner
function~\cite{juettnerQM} with local chemical self-potential
$-q_s(\phi(x)-\nu_s \psi(x))$.
The electric charge and current densities in the Poisson equations
for the electric potential $\phi$ and the magnetic flux function $\psi$
are computed with the density functions \refeq{TFdensity}, which
leads to the system of nonlinear PDEs
\bea
&&
-\Delta \phi \,
=
4\pi{\textstyle\sum_s}\ q_s\, {\cG}_s^{\rm TF}\!\left(\phi, \psi\right)
\label{TFeqPHI}
\\
&&
-\Delta \psi
=
4\pi{\textstyle\sum_s} \nu_s q_s {\cG}_s^{\rm TF}\!\left(\phi, \psi\right)
. \label{TFeqPSI}
\eea
Here and in the following, $\sum_s$ or $\sum_t$ always stands for
summation over the particle species, i.e. $s=\mp$ and $t=\mp$.
The Thomas--Fermi equations \refeq{TFeqPHI}, \refeq{TFeqPSI}, are
invariant under the isometries of three-dimensional Euclidean space,
Lorentz boosts along the beam's axis $a$, and the gauge transformation
\beq
\phi(x)\to\phi(x)+\phi_0;
\quad \psi(x)\to\psi(x)+\psi_0;
\quad \mu_s\to \mu_s +q_s(\phi_0-\nu_s\psi_0)
,\eeq
where $\phi_0$ and $\psi_0$ are arbitrary constants.
Since we are interested in the beam's natural symmetries,
we will not allow ``sources at infinity'' which would deform the
beam; hence, we supplement \refeq{TFeqPHI} and \refeq{TFeqPSI}
with the asymptotic conditions that, uniformly as $|x|\to\infty$,
\beq
\lim_{|x|\to \infty}\ \frac{\phi(x)}{ Q \ln \frac{1}{|x|}}
\ = 2 =\
\lim_{|x|\to \infty}\ \frac{c \psi(x)}{I \ln \frac{1}{|x|}}
, \label{AC}
\eeq
with $I\neq 0$ and $Q\neq 0$, where $I=\sum_s N_sq_s\nu_sc$ is the total
electrical current through the beam's cross-section and $Q=\sum_s N_sq_s$
the total charge per unit length of beam in the lab system;
if $Q=0$, the left equation in \refeq{AC} is to be replaced
by the condition that $\phi(x)\to\, const$ uniformly as $|x|\to\infty$.
The situation $I=0$ is not considered here, for then
of course there is no stationary beam.
\smallskip\noin
{\it Remark:} There are good reasons to conjecture that the asymptotic
conditions \refeq{AC} {are in fact implied by}
\refeq{TFdensity}-\refeq{TFeqPSI}.
Analogous results have been proven for Liouville's
equation~\cite{chenli} and for some Liouville
systems~\cite{chakieGAFA, chipotshafrirwolansky}.
No attempt will be made here to generalize
these results to \refeq{TFdensity}-\refeq{TFeqPSI}.
However, we note that such a generalization would have the
interesting {\it physical} implication (within the limits of applicability
of the model) that one cannot maintain a stationary straight beam of finite
current, whatever the geometry of its cross section, when there are
magnetic or electric multipole sources ``at infinity.''
\smallskip
To the best of the author's knowledge, the existence of beam
solutions in the Thomas--Fermi model \refeq{TFdensity}-\refeq{TFeqPSI}
with asymptotics \refeq{AC} has not yet been studied rigorously.
However, this semi-classical model is surely more regular than
the classical one, addressed next.
\subsection{The classical limit (Bennett theory)}
In the high-temperature / low-density limit, i.e. formally
$0<\beta_s\ll 1$ and $\beta_s\mu_s \ll -1$,
the Fermi--Dirac--J\"uttner functions \cite{juettnerQM} reduce to
the Maxwell--Boltzmann--J\"uttner functions \cite{juettnerCLASSa}
(see also \cite{degrootBOOK}, p.46, eq. (24)),
so that the Thomas--Fermi densities \refeq{TFdensity} simplify to
Boltzmann densities,
\beq
{\cG}_s^{\rm B}(\phi,\psi)
=
\frac{2}{h^3}\int_{\RR^3}
e^{-\beta_s\! \left(c\sqrt{m_s^2c^2+|p|^2} - \nu_s cp\cdot a \right)}
\dd p \, e^{\beta_s (\mu_s - q_s[\phi -\nu_s\psi])}
,\label{Bdensity}
\eeq
and \refeq{Ns} becomes
\beq
\int_{\RR^2}
{\cG}_s^{\rm B}\!\left(\phi,^{\phantom{b}}\!\!\psi\right)(x)\dd x
=
N_s.
\label{NsB}
\eeq
The system of equations \refeq{TFeqPHI} and \refeq{TFeqPSI}
then reduces to the Bennett equations
\bea
&&
-\Delta \phi
=
4\pi {\textstyle\sum_s} N_s q_s
\frac{\, e^{-\beta_s q_s(\phi - \nu_s \psi)} }
{\int_{\RR^2}e^{-\beta_s q_s(\phi-\nu_s \psi)}\dd x}
\label{BeqPHI}
\\
&&
-\Delta \psi
=
4\pi {\textstyle\sum_s} N_{s} q_s \nu_{s}
\frac{\,e^{-\beta_s q_{s} (\phi - \nu_{s} \psi)}}
{\int_{\RR^2}e^{-\beta_s q_{s} (\phi - \nu_{s} \psi)} \dd x}
, \label{BeqPSI}
\eea
see \cite{bennettPAPa} eq.'s(8),(9), and \cite{bennettPAPb}
eq.(7),\footnote{
In his papers~\cite{bennettPAPa, bennettPAPb}, Bennett
employed a classical, semi-relativistic setup, assuming
drifting Maxwell-Boltzmann distributions with relativistic
drift speeds, yet with non-relativistic velocity
dispersion in the cross-section of the beam; the
relativistic model with drifting J\"uttner functions
was used in \cite{benford}. It should be noticed, though,
that after integration over momentum space the very system
of equations \refeq{BeqPHI}, \refeq{BeqPSI} results in either
case, and it does so also in the strictly non-relativistic
limit~\cite{kiejllPhPl} --- except for minor re-interpretations
of the parameters in each case.}
where we have eliminated the external chemical potentials $\mu_s$ via
\refeq{NsB}.
The Bennett system is invariant under the isometries of
three-dimensional Euclidean space and under Lorentz boosts along
the beam's axis, $a$.
Restricted to the beam's cross-section, it is also invariant
under isotropic scaling, and in the special case when the parameters satisfy
\beq
\beta_s q_s (\nu_s c^{-1}I - Q) = 2,\qquad s = \mp
, \label{conform}
\eeq
also invariant under translated inversions.
Thus, \refeq{conform} implies invariance of the Bennett system
under the conformal group of two-dimensional Euclidean space, acting
in the beam's cross-section.
In addition, the Bennett equations are invariant under
a gauge transformation $\phi(x) \to \phi (x)+\phi_0$,
$\psi(x) \to \psi(x)+ \psi_0$.
Recall that we already eliminated the external chemical potentials
via the constraint equations \refeq{Ns} in the Bennett limit.
In the conformally invariant case \refeq{conform},
Bennett's Ansatz\footnote{Bennett actually made the Ansatz that
$\rho_+(x)/\rho_-(x) =\ const$, which up to gauge
freedom for the potentials is equivalent to
\refeq{AgleichPHI}.}
\beq
I^{-1}c\psi(x) = v(x) = Q^{-1}\phi(x)
\label{AgleichPHI}
\eeq
maps \refeq{BeqPHI} and \refeq{BeqPSI} separately into
Liouville's equation~\cite{liouville}
\beq
-\Delta {v}
=
4\pi \frac{\;e^{2v}}
{\int_{\RR^2}e^{2v}\,\dd x}
. \label{liouEQ}
\eeq
As remarked above, it has been proven in \cite{chenli} that any
regular solution of \refeq{liouEQ}, with the understanding that
$\int\exp(2v)\dd x <\infty$, satisfies
\beq
\lim_{|x|\to\infty}\ \frac{v(x)}{\ln\frac{1}{|x|}}
=
2
,\label{PSIasympCONF}
\eeq
uniformly as $|x|\to\infty$, which implies
that the asymptotic conditions \refeq{AC} are
automatically satisfied if $\phi$ and $\psi$ are given by \refeq{AgleichPHI}.
It has also been proven in \cite{chenli}, and subsequently
in \cite{chakieGAFA, chouwan} by using alternate techniques,
that \refeq{liouEQ} has only one regular family of solutions, given by
\beq
v(x|x_0;k) = v_0 + \ln \frac{1}{1+k^2 |x-x_0|^2}
, \label{BpinchPOT}
\eeq
where $k^{-1}>0$ is an arbitrary scale length, $x_0$ the
arbitrary center of rotational symmetry of the solution, and
$v_0$ an arbitrary gauge constant.
The corresponding current density $j(x)$ and charge density
$q(x)$ are given by
\beq
I^{-1} j(x)
=
\frac{1}{\pi}\frac{k^2}{\left(1+ k^2 |x-x_0|^2\right)^2}
=
Q^{-1}q(x)
. \label{BpinchDENS}
\eeq
The density profile \refeq{BpinchDENS} is the celebrated
Bennett beam profile.
%~\cite{benford, kiejllPhPl}.
Bennett speculated about the existence of other solutions to
\refeq{BeqPHI} and \refeq{BeqPSI} with asymptotics \refeq{AC};
see \cite{bennettPAPa} p.893, and \cite{bennettPAPb} p.1587.
(In the punctured plane additional solutions are readily found,
see e.g. \cite{benford}; however, they all lack regularity, due
to a point source, at the origin.)
In~\cite{kiejllPhPl} we proved that in the conformal case
\refeq{conform}, Bennett's system of equations
\refeq{BeqPHI} and \refeq{BeqPSI}, supplemented by the asymptotic conditions
\refeq{AC}, are in fact equivalent to \refeq{liouEQ} (with
asymptotic condition \refeq{PSIasympCONF} automatically satisfied,
see above) so that \refeq{BpinchPOT} then exhausts all possibilities.
Moreover, for the semi-conformal case where \refeq{conform} does
not hold we proved the existence of a continuous parameter family of
smooth radial solutions to \refeq{BeqPHI} and \refeq{BeqPSI} with asymptotics
\refeq{AC} which are not invariant under inversions.
All the solutions of our beam equations are
automatically also stationary solutions of the equations of Vlasov's
relativistic kinetic theory~\cite{vlasovBOOK}.
In \cite{kiejllPhPl} we showed that the Bennett equations
can also be realized as transversal part of stationary
{\it dissipative} kinetic equations in which the dissipation,
modeled by a thermostat, compensates the action of an applied
longitudinal electromotive force that drives the current.
In \cite{kiejllPhPl} we also gave a rigorous proof that
all radial solutions of \refeq{BeqPHI} and \refeq{BeqPSI} satisfying
\refeq{AC} also satisfy the Bennett identity
\beq
c^{-2} I^2 - Q^2
=
2 {\textstyle\sum_s} N_s\kb T_s \sqrt{1 -\nu_s^2}
. \label{bennettID}
\eeq
The identity \refeq{bennettID} was originally obtained by
Bennett~\cite{bennettPAPb} in a formal (and not entirely compelling)
manner by studying the radial time-dependent virial.
In this paper we will show that the Bennett identity
\refeq{bennettID}, respectively it's counterpart for the Thomas--Fermi
model, holds {\it a priori} without assuming symmetry, and this fact will
be one major ingredient in our proof of the cylindrical symmetry of
the beams.
\section{MAIN RESULTS}
To state our virial theorem, we introduce the thermodynamic
potentials (per unit length of beam), given by
\beq
J^{\rm TF}\!
=
\!\sum_s\beta_s^{-1}\frac{2}{h^3}\int_{\RR^2}\!\int_{\RR^3}\!
\ln\left[
1+e^{\beta_s\! \left(
\mu_s - c\sqrt{m_s^2c^2+|p|^2} +\nu_s cp\cdot a -q_s[\phi -\nu_s\psi]
\right)}\!
\right]\!\dd p\;\dd x
\label{THERMpotTF}
\eeq
for the semi-classical model, respectively by
\beq
J^{\rm B}
=
\sum_s\beta_s^{-1}\frac{2}{h^3}\int_{\RR^2}\int_{\RR^3}
e^{\beta_s\! \left(
\mu_s - c\sqrt{m_s^2c^2+|p|^2} +\nu_s cp\cdot a -q_s[\phi -\nu_s\psi]
\right)}
\dd p\;\dd x
\label{THERMpotB}
\eeq
for the classical model.
\smallskip
{\bf Theorem 3.1: } {(Virial identity.) \it
Let $\phi \in C^{2,\alpha}(\RR^2)$ and $\psi\in C^{2,\alpha}(\RR^2)$
solve \refeq{TFeqPHI} and \refeq{TFeqPSI} under the constraints
\refeq{Ns}, respectively solve \refeq{BeqPHI} and \refeq{BeqPSI}
under the constraints \refeq{NsB}, $s=\mp$, in either case
subjected to the asymptotic conditions \refeq{AC}.
Then
\beq
c^{-2} I^2 - Q^2
=
2 J
, \label{virialID}
\eeq
where $J$ stands for either $J^{\rm TF}$ or $J^{\rm B}$. }
\newpage
%\smallskip
We also show that deviations from cylindrical symmetry
violate \refeq{virialID}, which is expressed in the next theorem.
\smallskip
{\bf Theorem 3.2: } {(Cylindrical symmetry.) \it
Let $\phi \in C^{2,\alpha}(\RR^2)$ and $\psi\in C^{2,\alpha}(\RR^2)$
solve \refeq{TFeqPHI} and \refeq{TFeqPSI} under the constraints
\refeq{Ns}, respectively solve \refeq{BeqPHI} and \refeq{BeqPSI}
under the constraints \refeq{NsB}, $s=\mp$,
subjected to the asymptotic conditions \refeq{AC}.
Then there exists a point $x_0\in\RR^2$ such that both
$\phi$ and $\psi$ are radially symmetric about $x_0$, and the density
functions ${\cG}_s\!\left(\phi,^{\phantom{b}}\!\! \psi\right)(x)$
are decreasing away from $x_0$, where ${\cG}_s$ here stands for
either the Thomas--Fermi or the Boltzmann density function.}
%\newpage
\section{PROOFS}
We rewrite the Thomas--Fermi, respectively
Bennett system in two equivalent versions, which may be called the
``density potential representation'' and
the ``chemical self-potential representation.''
We will switch between these representations at our
convenience to obtain the asymptotic estimates, as $|x|\to\infty$, and the
isoperimetric estimates needed for our proofs of Theorems 3.1 and 3.2.
\subsection{The alternate PDE representations}
The chemical self-potentials $U_s(x),\ x \in \RR^2$, are given by
\beq
U_s = - q_s( \phi - \nu_s \psi)
. \label{chemUs}
\eeq
We also introduce density potentials $u_s(x),\ x \in \RR^2$,
defined by the invertible linear system
\bea
&& \phi = {\textstyle\sum_s} q_s u_s ,
\label{PHIus}
\\
&& \psi = {\textstyle\sum_s} \nu_s q_s u_s
.\label{PSIus}
\eea
Clearly,
\beq
U_s = {\textstyle\sum_t}\gamma_{s,t} u_{t}
,\label{usUs}
\eeq
where
\beq
\gamma_{s,t} = - q_s q_t (1 - \nu_s \nu_t)
\label{gMATRIX}
\eeq
denotes the entries of the matrix of coupling constants.
Notice that
\beq
\det (\gamma) = - (q_+ q_-)^2 (\nu_+ - \nu_-)^2
, \label{detMATRIX}
\eeq
so that for $\nu_+\neq \nu_-$, we have
\beq
\rank (\gamma) = 2
,\label{rankMATRIX}
\eeq
hence
\beq
u_s = {\textstyle\sum_t}\gamma_{s,t}^{-1} U_{t}
\label{Usus}
\eeq
where $\gamma_{s,t}^{-1}$ denotes the entries of the inverse
matrix $\gamma^{-1}$ to $\gamma$.
Now let ${\cG}_s$ stand for either ${\cG}_s^{\rm TF}$
or ${\cG}_s^{\rm B}$.
We note that ${\cG}_s (\phi,\psi)$ depends on $\phi$ and
$\psi$ only through the combination $-q_s(\phi -\nu_s \psi)=U_s$; thus
we can write ${\cG}_s(\phi,\psi)= G_s(U_s)=G_s(\sum_t\gamma_{s,t}u_t)$,
where of course $G_s$ stands for either $G_s^{\rm TF}$ or $G_s^{\rm B}$.
In either case, the map $w\mapsto G_s(w)$ is monotonic
increasing.
It then follows at once that the chemical self-potentials $U_s$
solve
the system of nonlinear PDEs
\beq
-\Delta U_s = 4\pi {\textstyle\sum_t}\gamma_{s,t}
G_{t}( U_{t})
,\label{UsPDE}
\eeq
supplemented by the integrability conditions
\beq
\int_{\RR^2} G_s(U_s)\dd x
=
N_s
\label{UsNs}
\eeq
and by the asymptotic conditions that, uniformly as $|x|\to\infty$,
\beq
\lim_{|x|\to\infty}\ \frac{U_s(x)}{\ln\frac{1}{|x|}}
=
2 {\textstyle\sum_t}\gamma_{s,t}N_t
.\label{UsAC}
\eeq
Alternately, in terms of the $u_s$
we get the following representation for our Thomas--Fermi / Bennett models,
\beq
-\Delta u_s = 4\pi G_s\big( {\textstyle\sum_t}\gamma_{s,t}u_t\big) ,
\label{usPDE}
\eeq
supplemented by the integrability conditions
\beq
\int_{\RR^2} G_s\big( {\textstyle\sum_t}\gamma_{s,t}u_t\big)\;\dd x
=
N_s
\label{usNs}
\eeq
and by the asymptotic conditions that, uniformly as $|x|\to\infty$,
\beq
\lim_{|x|\to\infty}\ \frac{u_s(x)}{\ln\frac{1}{|x|}}
=
2N_s,
\label{usAC}
\eeq
for $s=\mp$.
This constitutes the density potential representation of our
Thomas--Fermi / Bennett models.
\smallskip
\noin
{\it Remark:}
For the sake of completeness, we also state the PDEs of
the Bennett model explicitly as a Liouville system.
We readily eliminate the $\mu_s$ in terms of the $N_s$, using
\refeq{usNs}.
Setting now $u_s = N_s v_s$ and $\beta_s\gamma_{s,t}N_t =\kappa_{s,t}$,
and furthermore ${\textstyle\sum_t}\kappa_{s,t}v_t = 2V_s$
(equivalently, $\beta_s U_s= 2V_s$), with $s$ and $t$ taking the
``values'' $\pm$, we rewrite \refeq{usPDE} into the form
\beq
-\Delta v_s = 4\pi
\frac{\, \exp\left( {\textstyle\sum_t}\kappa_{s,t}v_t \right)}
{\int_{\RR^2}\exp\left({\textstyle\sum_t}\kappa_{s,t}v_t\right)\dd x},
\label{BLsyst}
\eeq
and \refeq{UsPDE} into
\beq
-\Delta V_s = 2\pi {\textstyle\sum_t}\kappa_{s,t}
\frac{\, \exp\left( 2V_t \right)}
{\int_{\RR^2}\exp\left(2V_t\right)\dd x},
\label{LsystemB}
\eeq
Equations \refeq{BLsyst} and \refeq{LsystemB} are
explicit alternate representations of the Liouville system associated to
the Bennett model.
The coefficient matrix~$\kappa$ is manifestly non-symmetric
in general, having negative diagonal and positive off-diagonal elements.
Note that in the conformal case \refeq{conform}, viz.
${\textstyle\sum_t}\kappa_{s,t} =2$ for $s=\pm$, the Ansatz $v_+=v_- = v$ in
\refeq{BLsyst}, respectively $V_+=V_- = v$ in \refeq{LsystemB}, reduces both
\refeq{BLsyst} and \refeq{LsystemB} to Liouville's equation \refeq{liouEQ}.
\smallskip
%\newpage
\subsection{Isoperimetric estimates}
Let $G_s$ continue to stand for either
$G_s^{\rm TF}$ or $G_s^{\rm B}$.
We introduce $g_s$, the primitive of $G_s$, i.e.,
$g_s^\pr(w) = G_s(w)$ for $w\in\RR$, such that the integrals
\beq
\int_{\RR^2} g_s(U_s)\;\dd x
=
M_s
\label{Ms}
\eeq
exist (notice that $M_s$ is defined by \refeq{Ms}).
In each case this primitive $g_s$ is unique and given by
\beq
g_s^{\rm TF}(U_s)
=
\beta_s^{-1}\frac{2}{h^3}\int_{\RR^3}\!
\ln\left[
1+e^{\beta_s\! \left(
\mu_s - c\sqrt{m_s^2c^2+|p|^2} +\nu_s cp\cdot a +U_s\right)}\!
\right]\!\dd p
\label{TFdensPRIM}
\eeq
for the semi-classical model, and by
\beq
g_s^{\rm B}(U_s)
=
\beta_s^{-1}\frac{2}{h^3}\int_{\RR^3}
e^{\beta_s\! \left(
\mu_s - c\sqrt{m_s^2c^2+|p|^2} +\nu_s cp\cdot a + U_s \right) }
\dd p
\label{BdensPRIM}
\eeq
for the classical model.
Notice that in the classical model we have $M_s =\beta_s^{-1}N_s$,
while in the semi-classical model we have $M_s > \beta_s^{-1}N_s$ by the
simple convexity inequality $\ln x \leq -1 +x$, with ``$=$'' only for $x=1$.
Notice furthermore that, in either case, the map
$w\mapsto g_s(w)$ is monotonic increasing.
\smallskip
\noin
{\bf Lemma 4.1:}
{\it Let the pair $(u_+,u_-)$ solve equations \refeq{usPDE} and
\refeq{usNs}, $s=\mp$, under the asymptotic conditions
\refeq{usAC}, with $\gamma$ given in \refeq{gMATRIX} satisfying
\refeq{detMATRIX}.
Then
\beq
\frac{1}{2} {\textstyle\sum_{s,t}}
\gamma_{s,t} N_sN_t - {\textstyle\sum_s} M_s \geq 0
,\label{isoperEST}
\eeq
and equality in \refeq{isoperEST} holds if and only if both $u_+$ and
$u_-$ are radially symmetric and decreasing about the same point.}
\smallskip
\noin
{\it Proof:}
We follow the general reasoning of
\cite{chakieGAFA, chakieCR, chakieNA}.
Since, by hypothesis, the pair $(u_+,u_-)$ solves the equations
\refeq{usPDE} and \refeq{usNs}, $s=\mp$, under the asymptotic
conditions \refeq{usAC}, then $(U_+,U_-)$ satisfies \refeq{UsPDE}
and \refeq{usNs}, $s=\mp$, under the asymptotic conditions \refeq{UsAC}.
Therefore, as $|x|\to\infty$,
\beq
G_s(U_s)(x)
=
\frac{2}{h^3}\int_{\RR^3}
e^{\beta_s\! \left(\mu_s - c\sqrt{m_s^2c^2+|p|^2}+\nu_s cp\cdot a\right)}
\dd p\,
{\displaystyle |x|^{-2\beta_s\sum_t\gamma_{s,t}N_t(1+\theta(x))}}
,\label{GsAC}
\eeq
with $\theta(x) = o(1)$.
Also by hypothesis, \refeq{UsNs} is satisfied, so that from
\refeq{GsAC} we conclude that $\beta_s{\sum_t}\gamma_{s,t}N_t \not\!\!{<}\,1$.
Then, by \refeq{UsAC} again, and since $U_s\in C^{2,\alpha}$ (hence,
$U_s\in C^\infty$ by bootstrapping), the level sets
$\Lambda^s_\xi = \{x| U_s \geq \xi_s\}$ are compact, hence
$|{\Lambda_\xi^s}| <\infty$.
Let $x\mapsto U_s^\ast(|x|)$
denote the equi-measurable, radially symmetric, non-increasing
rearrangement of $x\mapsto U_s(x)$, centered at the origin, and
denote by ${\Lambda^s_\xi}^\ast = \{x|\; U_s^\ast \geq \xi_s\}$ the
ball of radius $r_\xi^s$, centered at the origin.
By Sard's theorem the $C^\infty$ regularity of the
$U_s$ implies that the outward normal $\hat{\lambda}$ to $\pd\Lambda_\xi^s$
exists except at most for $\xi$-values in a set of measure zero, so that
the ensuing manipulations involving $\hat{\lambda}$ to $\pd\Lambda_\xi^s$
are well defined $\xi$-a.e.
First, recalling that $G_s>0$, we note that on $\pd\Lambda_\xi^s$
we have $-\langle\hat{\lambda}, \nabla U_s\rangle = |\nabla U_s|$, by the Hopf lemma.
Integration of this identity over $\pd\Lambda_\xi^s$, a trivial
rewriting, and
an application of the Cauchy--Schwarz inequality now gives the estimate
\beq
-\int_{\partial\Lambda^s_\xi}
\langle\hat{\lambda}, \nabla U_s\rangle\, \dd\sigma
=
\int_{\partial\Lambda^s_\xi}
| \nabla U_s|^2 \frac{1}{|\nabla U_s|}\, \dd\sigma
\geq
\left(\int_{\partial\Lambda^s_\xi}\, \dd\sigma\right)^2
\left(\int_{\partial\Lambda^s_\xi}\frac{1}{|\nabla U_s|}\,\dd\sigma\right)^{-1},
\label{BANDLEa}
\eeq
with equality holding if and only if $|\nabla U_s|$ is constant on
$\partial\Lambda^s_\xi$.
Noting that
\beq
\int_{\partial\Lambda^s_\xi}\dd\sigma = |\Lambda^s_\xi|,
\label{BANDLEb}
\eeq
and applying the classical isoperimetric inequality~\cite{bandleBOOK}, we have
\beq
|\Lambda^s_\xi| \geq |\partial{\Lambda_\xi^s}^\ast|,
\label{BANDLEc}
\eeq
with equality holding if and only if, up to translation,
$\partial\Lambda^s_\xi=\partial{\Lambda_\xi^s}^\ast$.
By the co-area formula~\cite{federer},
\beq
\int_{\partial\Lambda^s_\xi}\frac{1}{|\nabla U_s|}\, \dd\sigma
=
\int_{\partial{\Lambda^s_\xi}^*}\frac{1}{|\nabla U_s^*|}\, \dd\sigma.
\label{BANDLEd}
\eeq
Pulling these estimates together we have
\beq
-\int_{\partial\Lambda^s_\xi}
\langle\hat{\lambda}, \nabla U_s\rangle\, \dd\sigma
\geq
|\partial{\Lambda_\xi^s}^\ast|^2
\left(\int_{\partial{\Lambda^s_\xi}^*}\frac{1}{|\nabla U_s^*|}\, \dd\sigma
\right)^{-1},
\label{BANDLEe}
\eeq
with equality holding if and only if,
(i), $|\nabla U_s|$ is constant on $\partial\Lambda^s_\xi$, and
(ii), $\partial\Lambda^s_\xi=\partial{\Lambda_\xi^s}^\ast$, up to translation.
This last remark implies in particular that we can restate
\refeq{BANDLEe} as~\cite{bandleBOOK},
\beq
-\int_{\partial\Lambda^s_\xi}
\langle\hat{\lambda}, \nabla U_s\rangle\, \dd\sigma
\geq
-\int_{\partial{\Lambda_\xi^s}^\ast}\partial_r U_s^\ast\, \dd\sigma .
\label{isoperINEQcoarea}
\eeq
Next, using Green's theorem and \refeq{UsPDE}, then rearrangement identity for $s=t$,
then rearrangement inequality for $s\neq t$ (in which case $t= -s$), noting that $\gamma_{s,-s} >0$
and recalling that $w\mapsto G_s(w)$ is increasing, we have
\bea
-\int_{\partial\Lambda^s_\xi}
\langle\hat{\lambda}, \nabla U_s\rangle\, \dd\sigma
\!\!\!\!\!\!
&&
=
-\int_{\Lambda^s_\xi}\Delta U_s\,\dd x
\\
&&
=
4\pi {\textstyle\sum_t}\gamma_{s,t}
\int_{\Lambda^s_\xi}G_t({U_{t}})\,\dd x
\\
&& =
4\pi\Biggl(\gamma_{s,s}\int_{{\Lambda^s_\xi}^\ast}\!\!
G_s({U^\ast_{s}})\,\dd x
+ \gamma_{s,-s}\int_{\Lambda^s_\xi}\!\!G_{-s}({U_{-s}})\,\dd x\Biggr)
\\
&& \leq
4\pi {\textstyle\sum_t}\gamma_{s,t}
\int_{{\Lambda^s_\xi}^\ast}G_t({U^\ast_{t}})\,\dd x
\label{isoperINEQprodfunc}
\eea
where equality in \refeq{isoperINEQprodfunc}
can hold only if $U_{t}$ and $U_s$ share their level lines (up to the labelling)
in $\Lambda_\xi^{t}$, for our $\gamma$ is irreducible.
Combining inequalities \refeq{isoperINEQcoarea} and \refeq{isoperINEQprodfunc}, we arrive
at the inequality
\beq
-\int_{\partial{\Lambda_\xi^s}^\ast}\partial_r U_s^\ast\, \dd\sigma
\leq
4\pi {\textstyle\sum_t}\gamma_{s,t}
\int_{{\Lambda^s_\xi}^\ast}G_t({U^\ast_{t}})\,\dd x ,
\label{MAINineq}
\eeq
where equality can hold if and only if each $\Lambda^s_\xi$ is a disk,
with $|\nabla U_s|$ constant on $\partial\Lambda_\xi^s$, and all the $U_s$
share their level lines (up to the labelling).
Thus, in case of equality in \refeq{MAINineq}, from the first two conditions for equality
it follows that the family of disks $\Lambda^+_\xi$ and the family of disks $\Lambda^-_\xi$
are separately concentric, while from the third condition for equality it then follows that
the families of disks must be jointly concentric.
On the other hand, if at least one of the $U_s$ is not radially
symmetric decreasing about any point, let $\Xi^s$ be the image under $U_s$ of the
(generally non-radial) set $\subset\RR^2$ which supports the non-radial parts of $U_s$.
Then $\Xi^s$ has finite measure.
Since equality in \refeq{MAINineq} cannot hold for $\xi\in \Xi^s$,
for $\xi\in \Xi^s$ we now conclude that we have strict inequality in
\refeq{MAINineq},
\beq
-2\pi r^s_\xi U_s^{\ast\prime}(r^s_\xi)
<
4\pi{\textstyle\sum_t}\gamma_{s,t}
\int_{{\Lambda^s_\xi}^\ast}G_{t}({U^\ast_{t}})\,\dd x
\label{isostrictINEQ}
\eeq
for {\it both} $s=\mp$.
We now set
\beq
\cN_s(r)=\int_{B_r(0)}G_s({U_s^\ast})\,\dd x,
\eeq
and
\beq
\cM_s(r)=\int_{B_r(0)}g_s({U_s^\ast})\,\dd x
. \eeq
We have $\lim_{r\to\infty}\cN_s(r) = N_s$
and $\lim_{r\to\infty}\cM_s(r) = M_s$, for
\beq
\int_{\RR^2}f({U_s^\ast})\,\dd x = \int_{\RR^2}f({U_s})\,\dd x
, \eeq
where $f$ stands for either $g_s$ or $G_s$.
By \refeq{isostrictINEQ},
\beq
2\pi r U_s^{\ast\prime}(r) \geq
- 4\pi{\textstyle\sum_t}\gamma_{s,t}\; \cN_{t}(r)
, \eeq
from which we conclude that
\beq
r \cM_s^{\ppr}(r) \geq \cM_s^\pr(r)-
2 \cN^\pr_s(r){\textstyle\sum_t}\gamma_{s,t}\; \cN_{t}(r)
,\label{isoANW}
\eeq
with ``$>$'' valid for all $r>0$ for which $U_s^\ast(r) \in \Xi^s$,
while ``$=$'' holds for $U_s^\ast(r) \not\in \Xi^s$.
We now sum \refeq{isoANW} w.r.t. $s=\mp$, obtaining
\beq
{\textstyle\sum_{s}} r \cM_s^{\ppr}(r)
\geq
{\textstyle\sum_{s}} \cM_s^\pr(r)
-
{\textstyle\sum_{s,t}}\gamma_{s,t}\Bigl(\cN_s(r)\cN_{t}(r)\Bigr)^\pr
, \label{totDIFF}
\eeq
where we made use of the fact that $\gamma$ is real symmetric.
Next we integrate \refeq{totDIFF} from $r=0$ to $r=\infty$,
using integration by parts on the left-hand side.
Since $g_s({U_s^\ast})\in L^1(\RR^2)$ is radially
decreasing, we have
\beq
\lim_{r\to\infty} r \cM_s^\pr(r) =0
, \eeq
thus we get the result
\beq
\frac{1}{2}{\textstyle\sum_{s,t}}\gamma_{s,t}N_sN_t-{\textstyle\sum_s}M_s
\geq
0
. \label{isoINEQfinal}
\eeq
Now, if ``$=$'' holds in \refeq{isoINEQfinal},
then all the level curves $\pd\Lambda_\xi^s$ are circles
with $|\nabla U_s|$ constant on $\pd\Lambda_\xi^s$;
hence~\cite{brothersziemer} the circular level curves of each $U_s$ are
concentric, and then $U_s(x) = U_s^*(|x-x_0^s|)$ for some $x_0^s$.
Moreover, since in case of ``$=$'' in \refeq{isoINEQfinal}
\refeq{isoperINEQprodfunc} tells us that the two $U_s$ must share their
level curves (with generally different level values, of course),
we conclude that $x_0^+ = x_0^-$, i.e. $U_+$ and $U_-$ are then
radially symmetric and decreasing about the same center of
symmetry, $x_0$.
On the other hand, if at least one of the $U_s$ is not
radially symmetric and decreasing about any point, then
the integration picks up all the strict inequalities
from $\xi\in \Xi^s$, and ``$>$'' holds in \refeq{isoINEQfinal}.
Finally, since $\rank (\gamma) =2$, it follows that at
least one $u_s$ is not radially symmetric and decreasing
about any point if at least one $U_s$ is not.
In the same vein, $u_+$ and $u_-$ are radially symmetric
and decreasing about the same center of symmetry, $x_0$, whenever both
$U_+$ and $U_-$ are.
This proves Lemma 4.1. $\QED$
\subsection{Asymptotic control near infinity}
Standard harmonic analysis gives us:
\smallskip
\noindent
{\bf Proposition 4.2:}
{\it Under the hypothesis stated in Lemma 4.1, each solution pair
$(u_+,u_-)$ of \refeq{usPDE}, \refeq{usNs}, \refeq{usAC}
satisfies the integral representation
\beq
u_s(x) - u_s(0)
=
\int_{\RR^2}\left(\ln \frac{1}{|x - y|^2}-\ln\frac{1}{|y|^2}\right)
G_s \Big({\textstyle\sum_t} \gamma_{s,t}u_t\Big)(y)\dd y
. \label{usINTrep}
\eeq
}
\smallskip
\noindent
{\bf Corollary 4.3:}
{\it By \refeq{usINTrep} we have
\beq
\nabla u_s(x)
=
- 2\int_{\RR^2}\frac{x - y\,}{|x - y|^2}\,
G_s \Big({\textstyle\sum_t} \gamma_{s,t}u_t\Big)(y)\dd y
. \label{GRADusINTrep}
\eeq
}
\smallskip
With the help of \refeq{usINTrep} and \refeq{GRADusINTrep}
we obtain asymptotic control over the r.h.s. of \refeq{usPDE},
expressed in terms of the $U_s$.
\smallskip
\noin
{\bf Lemma 4.4:}
{\it Under the hypotheses of Lemma 4.1, there exists an $r_0(U_s) >0$,
a constant $C_s>0$, and a monotonic decreasing $h_s(|x|) > 0$ satisfying
\beq
\lim_{|x|\to\infty} |x|^{-h_s(|x|)} =0
,\label{hDECAY}
\eeq
such that for $s=\mp$ we have, for $|x| > r_0$,
\beq
G_s(U_s)(x)
\leq
C_s |x|^{-2 -h_s(|x|)}
.\label{GsUsASYMP}
\eeq
Furthermore, for at least one $s$, we have
$h_s(|x|) \geq \eps_s>0 $ for $|x| > r_0$.}
\smallskip
\noin
{\it Proof:}
The bound \refeq{GsUsASYMP}, with $h_s(|x|) = O(1)$
monotonic decreasing, follows directly from \refeq{GsAC} and \refeq{UsNs}.
Furthermore, by \refeq{GsAC} and \refeq{UsNs} we can find $h$
such that $\int_{1}^\infty |x|^{-1-h_s(|x|)} \dd |x| <\infty$,
but this is impossible if $ |x|^{-h_s(|x|)} \not\to 0$; hence,
\refeq{hDECAY} follows.
This still allows $h_s(|x|) = o(1)$ for both $s$, but by
Lemma 4.1 and the fact that $M_s\geq \beta_s^{-1}N_s$ (see the definition
of the $g_s$ above), we find, after multiplying \refeq{isoINEQfinal}
by $-2$ and re-grouping terms, that
\beq
{\textstyle\sum_s} \beta_s^{-1} N_s
\left(2 - \beta_s {\textstyle\sum_{t}}\gamma_{s,t} N_t\right)
\leq
0
. \label{isoINEQestimate}
\eeq
Thus, if for one of the $s$ we have $h_s(|x|) = o(1)$, say for $s=+$,
then by \refeq{GsAC} we have $\beta_+ \sum_t\gamma_{+,t} N_t = 1$, so that
\refeq{isoINEQestimate} gives us right away
\beq
\beta_{-}{\textstyle\sum_{t}}\gamma_{-,t} N_t
\geq
2+ \frac{\beta_-}{\beta_+}\frac{N_+}{N_-}
. \label{isoINEQestimateB}
\eeq
By symmetry, the analog conclusion holds for
$\beta_{+}{\textstyle\sum_{t}}\gamma_{+,t} N_t$ if $h_-(|x|)=o(1)$.
Hence, $h_s(|x|) = o(1)$ for at most one of the $s$.
This proves Lemma 4.4. $\QED$
\smallskip
\noin
{\bf Corollary 4.5:}
{\it For at least one $s$, we have
$\left|\int_{\RR^2}\ln |y|\; G_s \big(U_s\big)(y)\dd y\right|<\infty$,
so that for this $s$, we have
\beq
\lim_{|x|\to\infty}\Big(u_s(x) + 2N_s\ln|x|\Big)
=
u_s(0) + 2 \int_{\RR^2}\ln |y| G_s \big(U_s\big)(y)\dd y
.\eeq
}
\smallskip
We proceed with gradient estimates.
\smallskip
\noin
{\bf Lemma 4.6:}
{\it Under the hypothesis stated in Lemma 4.1, each solution pair
$(u_+,u_-)$ of \refeq{usPDE}, \refeq{usNs}, \refeq{usAC}
satisfies the gradient estimates
\beq
\limsup_{|x|\to\infty}|x||\nabla u_s|
\leq
2N_s
.\label{modGRADuEST}
\eeq
}
\smallskip
\noin
{\it Proof:}
By Corollary 4.3, we have
\beq
\big|\nabla u_s(x)\big|
\leq
2\int_{\RR^2}{\frac{G_s\!\left(U_s\right)(y)}{|x-y|}}\, \dd y
.\label{modGRADuESTa}
\eeq
After multiplying \refeq{modGRADuESTa} by $|x|$, a simple rewriting
of the r.h.s. gives
\beq
|x|\big|\nabla u_s(x)\big|
\leq
2 \int_{\RR^2}G_s(U_s)(y)\dd y
+ 2 \int_{\RR^2}\left({\frac{|x|}{|x-y|}}-1\right)G_s(U_s)(y)\dd y
.\label{modGRADuESTb}
\eeq
By \refeq{UsNs} the first integral on the r.h.s. of
\refeq{modGRADuESTb} equals $N_s$.
By the triangle inequality, the second integral on the
r.h.s. of \refeq{modGRADuESTb} is bounded in absolute value by
\beq
2\int_{\RR^2}{\frac{|y|}{|x-y|}}G_s(U_s)(y) \dd y
.\label{modGRADuESTc}
\eeq
We now show that
\beq
\lim_{|x|\to\infty} \int_{\RR^2}{\frac{|y|}{|x-y|}}G_s(U_s)(y)\dd y=0
,\label{modGRADuESTd}
\eeq
from which the lemma follows.
We split the domain of integration in \refeq{modGRADuESTc} as follows:
\hbox{$\RR^2 = \Omega_1{\cup}\Omega_2{\cup}\Omega_3$,} with
$\Omega_1 = \{ y\mid|y|<|x|/2 \}$,
$\Omega_2 =\{y\mid|x|/2\leq|y|\leq 2|x|\}$, and
$\Omega_3 = \{y\mid|y|>2|x|\}$.
If $G_s(U_s)(y) \leq C |y|^{-2-\eps}$,
with $0<\epsilon <1$, then the estimates are precisely the same as
in \cite{chakieGAFA}, section 2, with $\exp$ replaced by $G_s$; this
is the case for at least one of the $s$.
It remains to provide estimates when $h_s(|x|) = o(1)$
for one of the $s$.
To estimate the contribution from $\Omega_1$ when
$G_s(U_s)(y) \leq C |y|^{-2-h_s(|y|)}$ with $h_s(|y|) = o(1)$,
we note that
\bea
\int_{\Omega_1}\!{\frac{|y|}{|x{-}y|}} G_s(U_s)(y) \dd y
%\!\!\!\!\!\!
%&&
\leq
\frac{C}{|x|}\int_{\Omega_1}\!|y|G_s(U_s)(y) \dd y
%\\
%&&
\leq
\frac{C'}{|x|}\int_{0}^{|x|} \zeta^{-h_s(\zeta)} \dd \zeta
.\label{modGRADuESTe}
\eea
As for the \rhs of \refeq{modGRADuESTe}, L'Hopital's Rule gives us
\beq
\lim_{|x|\to\infty}\frac{C'}{|x|}\int_{0}^{|x|}\zeta^{-h_s(\zeta)} \dd \zeta
=
C' \lim_{|x|\to\infty} |x|^{-h_s(|x|)} =0
,\label{modGRADuESTeB}
\eeq
the last step by Lemma 4.4.
Hence, the l.h.s.\refeq{modGRADuESTe} vanishes as $|x|\to\infty$.
Similarly, the contribution from $\Omega_2$ is
estimated by using again that $G_s(U_s)(y) \leq C |y|^{-2-h_s(|y|)}$,
so that for $|x|$ large enough we have the bound
\beq
\int_{\Omega_2}{\frac{|y|}{|x-y|}}G_s(U_s)(y)\dd y
\leq
\frac{C}{|x|^{1+h_s(|y|)}}\int_{|y|<4|x|}{\frac{\dd y}{|y|}}
\leq
C|x|^{-h_s(|x|)}
.\label{modGRADuESTg}
\eeq
Clearly r.h.s.\refeq{modGRADuESTg}$\to 0$ as $|x|\to\infty$,
by the same reasoning as for $\Omega _1$.
Finally, the contribution from $\Omega_3$ is dominated by
\beq
\int_{\Omega_3}{\frac{|y|}{|x-y|}}G_s(U_s)(y) \dd y
\leq
C\int_{|y|>2|x|}G_s(U_s)(y) \dd y
,\label{modGRADuESTf}
\eeq
which vanishes as $|x|\to\infty$, by hypothesis \refeq{UsNs}.
This concludes the proof of Lemma 4.6. $\QED$
\smallskip
\noin
{\bf Lemma 4.7:}
{\it Under the hypotheses of Lemma 4.1, we have, uniformly in $x$,
\beq
\lim_{|x|\to\infty}\langle x,\nabla u_s\rangle = - 2 N_s.
\label{LEMMAfourSEVEN}\eeq
}
\smallskip
\noin
{\it Proof:}
Let $\hat{x}={x/{|x|}}$ and $\hat{y}={y/{|y|}}$, with $|x|=|y|$.
Now fix $\hat{x}\in \SS^1$.
By \refeq{usAC}, we have
\beq
\lim_{\tau\to\infty}\frac{u_s(\tau\hat{x})}{\ln \tau}
=
- 2N_s
\eeq
Thus, by L'Hopital's Rule,
\beq
\lim_{\tau\to\infty}\tau\frac{\dd}{\dd \tau}u_s(\tau\hat{x})
=
\lim_{|x|\to\infty} \langle x,\nabla u_s\rangle
=
-2N_s
\label{lhospital}
\eeq
for $x =|x|\hat{x}$.
It remains to establish uniformity of \refeq{lhospital}.
To this extent, we show that there exist $R$ and $\delta$,
such that, if $|x|>R$ and $|\hat{x}-\hat{y}|<\delta$, then
\beq
\big|\langle x,\nabla u_s(x)\rangle - \langle y,\nabla u_s(y)\rangle\big|
<
\eps
.\label{uniformGRAD}
\eeq
We first show that for $|x|>R$ and $|x-y|<{|x|/ {10}}$, we have,
\beq
|x|\big|\nabla u_s(x)-\nabla u_s(y)\big|
\leq
C|\hat{x}-\hat{y}|+ C' |x|^{-h_s(|x|)} .
\eeq
By Corollary 4.3,
\beq
\big|\nabla u_s(x)-\nabla u_s(y)\big|
\leq
2\int_{\RR^2}G_s(U_s)(z)\left|{\frac{x-z}{|x-z|^2}}
-{\frac{y-z}{|y-z|^2}}\right|\dd z
.\eeq
We break up the domain of integration in the above integral
exactly as in the proof of Lemma 4.6.
(Notice the integration variable is now $z$.)
The integration over $\Omega_1$ is estimated exactly as in
section 2 of \cite{chakieGAFA} to get
\beq
\int_{\RR^2}G_s(U_s)(z){\frac{|x-y|}{|x|^2}}\dd z
\leq
C{\frac{|x-y|}{|x|^2}}
.\eeq
The integral over $\Omega_2$ is dominated by
\beq
\int_{|z|\sim |x|}G_s(U_s)(z)\left(\frac{1}{|x-z|}
+\frac{1}{|y-z|}\right)\dd z
\leq
C |x|^{-1-h_s(|x|)}
.\eeq
The final estimate above was identical to that made in the
proof of Lemma~4.6.
Use was made of $G_s(U_s)(z)\leq C|x|^{-2- h_s(|x|)}$ on $\Omega_2$,
which holds by Lemma 4.4.
The contribution from $\Omega_3$ is estimated once again exactly as
in section 2 of \cite{chakieGAFA} to be dominated by
\beq
C|x-y|\int_{|z|>2|x|}{\frac{G_s(U_s)(z)}{|z|^2}}\,\dd z
\leq
C' {\frac{|x-y|} {|x|^2}}\int_{\RR^2} G_s(U_s)\dd x
\leq
C^\ppr {\frac{|x-y|}{|x|^2}}
,\eeq
where the last step follows by \refeq{UsNs}.
By these estimates,
\bea
\big|\langle x,\nabla u_s(x)\rangle-\langle y,\nabla u_s(y)\rangle\big|
\!\!\!\!\!\!\!\!\!
&&
\leq
|x||\hat{x}-\hat{y}|\big|\nabla u_s(x)\big|
+ |y|\big|\nabla u_s(x)-\nabla u_s(y)\big|\qquad
\\
&&
\leq
|x|\big|\nabla u_s(x)\big||\hat{x}-\hat{y}|+|\hat{x}-\hat{y}|+
C|x|^{-h_s(|x|)}
.\quad
\eea
By Lemma 4.6, the last expression above is at most
$C' \delta + C|x|^{-h_s(|x|)}$.
Thus our claim \refeq{uniformGRAD} follows now from Lemma 4.4
for suitably large $R$ and small $\delta$.
Since $\SS^1$ is compact, uniformity of the limit in Lemma 4.7
now follows. $\QED$
\smallskip
Lemmata 4.6 and 4.7 imply
\smallskip
\noin
{\bf Corollary 4.8:}
{\it Under the hypotheses expressed in Lemma 4.1, we have,
uniformly in $x$,
\beq
\lim_{|x|\to\infty}|x||\nabla u_s|
=
2 N_s
.\eeq
}
\smallskip
\noin
{\it Proof:}
Follows essentially verbatim \cite{chakieGAFA}, proof of
Corollary 2.2, with $\exp$ replaced by $G_s$. $\QED$
\smallskip
Let $\Omega_\xi=\{x|\; u_s(x)\geq \xi\}$, where $\xi\ll -1.$
By \refeq{usAC} it follows that if $x\in\partial \Omega_\xi$,
then $|x|\geq R(c)$ with $R(c)$ large.
For such $x$, it follows from Corollary 4.8 that $\nabla u_s\neq 0$.
Since $u\in C^{2,\alpha}_{loc}$ we easily see that therefore
$\partial\Omega_\xi\in C^{2,\alpha}$.
Thus the unit outward normal $\hat{\omega} (x)$
to $\partial \Omega_\xi$ exists at all $x\in\partial\Omega_\xi$
for $\xi$ sufficiently negative.
\noin
{\bf Lemma 4.9.}
{\it Let $\hat{\omega} (x)$ be the unit outward normal to
$\partial\Omega_\xi$ at $x$, and let $\hat{x} = x/|x|$.
We have, uniformly in $x$,
\beq
\lim_{c\to -\infty}\langle \hat{x},\hat{\omega}\rangle = 1
.\eeq
}
\smallskip
\noin
{\it Proof:}
Identical to \cite{chakieGAFA}, proof of Lemma 2.8. $\QED$
\smallskip
\noin
{\it Remark:}
Lemma 4.9 implies that asymptotically for large $x$
the $\partial\Omega_\xi$ become concentric circles.
\smallskip
We are now in a position to prove our Theorem 3.1.
\subsection{Proof of the Virial Theorem}
\noin
{\bf Proposition 4.10:}
(Rellich--Pokhozaev identity.)
{\it Under the hypotheses expressed in Theorem 3.1, any solution pair
$(u_+,u_-)$, of \refeq{usPDE}, \refeq{usNs}, \refeq{usAC} satisfies
the Rellich--Pokhozaev identity
\beq
\frac{1}{2}{\textstyle{\sum_{s,t}}}\gamma_{s,t} N_s N_t
- {\textstyle\sum_s} M_s
=
0
. \label{relpohIDliouEQsyst}
\eeq}
\smallskip
\noin
{\it Remark:}
The Rellich--Pokhozaev identity \refeq{relpohIDliouEQsyst}
is identical to the identity expressed in the Virial Theorem 3.1.
\smallskip
\noin
{\it Proof of Proposition 4.10:}
For $(u_+,u_-)$ a solution pair of \refeq{usPDE}, \refeq{usNs}, \refeq{usAC},
we have the partial differential identity
\beq
{\rm div} ( \langle x,\nabla u_{t}\rangle \nabla u_s )
=
\langle \nabla u_s , (1 + \langle x,\nabla \rangle )
\nabla u_{t}\rangle
-4\pi \langle x,\nabla u_{t}\rangle G_s ({\textstyle\sum_{t}}\gamma_{s,t}u_t)
.\label{relpohANSliouEQsyst}
\eeq
We will multiply \refeq{relpohANSliouEQsyst} by $\gamma_{s,t}$,
sum over $s$ and $t$, integrate over $B_R$, use some partial integrations,
then take the limit $R\to\infty$.
We evaluate first the left-hand side of \refeq{relpohANSliouEQsyst}.
Green's theorem gives us
\beq
\int_{B_R}
{\rm div} ( \langle x,\nabla u_{t}\rangle \nabla u_s )\dd x
=
\int_{\partial B_R}
|x|^{-1} \langle x,\nabla u_{t}\rangle
\langle x, \nabla u_s\rangle\, \dd\sigma
.\label{partINTrelpohSYST}
\eeq
Taking the limit $R\to \infty$, using \refeq{LEMMAfourSEVEN},
we get
\beq
\lim_{R\to\infty}
\int_{\partial B_R}
|x|^{-1} \langle x,\nabla u_{t}\rangle
\langle x, \nabla u_s\rangle\, \dd\sigma
= 8 \pi N_sN_t
,\label{firstLIMIT}
\eeq
hence
\beq
\lim_{R\to\infty}
{\textstyle{\sum_{s,t}}}\gamma_{s,t}
\int_{B_R}
{\rm div} ( \langle x,\nabla u_{t}\rangle \nabla u_s )\dd x
=8\pi {\textstyle\sum_{s,t}} \gamma_{s,t} N_s N_t
.\eeq
On the other hand, the last term in the right-hand side of
\refeq{relpohANSliouEQsyst} gives us
\bea
&&
\!\!\!\!\!\! \!\!\!
{\textstyle{\sum_{s,t}}}\gamma_{s,t}\!
\int_{B_R}\!\! \langle x,\nabla u_{t}\rangle
G_s({\textstyle\sum_{t}}\gamma_{s,t}u_t)\dd x
=
{\textstyle{\sum_s}}
\int_{B_R}\!\!
\langle x,\nabla g_s({\textstyle\sum_{t}}
\gamma_{s,t}u_{t})\rangle\dd x
\nonumber \\ &&
=
{\textstyle{\sum_s}}
\int_{\partial B_R}\! |x|\;g_s({\textstyle\sum_{t}}\gamma_{s,t}u_{t})\dd\sigma
-2{\textstyle{\sum_s}}\int_{B_R}\!\!g_s({\textstyle\sum_t}\gamma_{s,t}u_{t})
\dd x\,
.\label{lastterm}
\eea
We now take the limit $R\to \infty$ in \refeq{lastterm}.
As for the surface integrals, we note that by Lemma 4.4 we
have $g_s(U_s)(x)\sim C G_s(U_s)(x)$ as $|x|\to\infty$, so that
once again by Lemma 4.4, we have
\beq
\lim_{R\to\infty}
\int_{\partial B_R}
| x |\; g_s({\textstyle\sum_t} \gamma_{s,t}u_{t}) \,\dd\sigma = 0
.\eeq
As for the volume integrals, we get
\beq
\lim_{R\to\infty}
\int_{B_R}g_s({\textstyle\sum_t} \gamma_{s,t} u_{t})\,\dd x = M_s
.\eeq
Turning now to the first term in the right-hand side of
\refeq{relpohANSliouEQsyst}, we use the symmetry of $\gamma$,
an integration by parts and \refeq{usPDE}, to get
\bea
&&
\!\!\!\!\!\!\!\!\!\!\!
{\textstyle\sum_{s,t}}\gamma_{s,t}
\int_{B_R}
\langle \nabla u_s , (1\! +\! \langle x,\nabla \rangle )
\nabla u_{t}\rangle \,\dd x
=
{1\over 2}{\textstyle\sum_{s,t}}\gamma_{s,t}
\int_{\partial B_R}
| x | \langle \nabla u_s, \nabla u_{t}
\rangle \,\dd\sigma\qquad
\\
&&
=
{1\over 2}
{\textstyle\sum_{s,t}}\gamma_{s,t}
\int_{\partial B_R} \Big(|x|^{-1}
\langle x, \nabla u_s\rangle
\langle x, \nabla u_{t} \rangle
+ | x | \langle \nabla_{\rm T} u_s, \nabla_{\rm T} u_{t}\rangle\Big) \dd\sigma
\label{alltogether}
\eea
where $\nabla_{\rm T}$ denotes tangential derivative.
By Lemma 4.7 and Corollary 4.8, we have
\beq
|x|^{2}|\nabla_{\rm T}u_s|^2 =
|x|^{2}|\nabla u_s|^2 - \langle x, \nabla u_s\rangle^2 \to 0
,\eeq
uniformly as $|x|\to \infty$.
Thus as $R\to \infty$,
\beq
\int_{\partial B_R}
| x | \langle \nabla_{\rm T} u_s, \nabla_{\rm T} u_{t}\rangle \,\dd\sigma \to 0
\eeq
and therefore
\beq
\lim_{R\to\infty}
{\textstyle\sum_{s,t}}\gamma_{s,t}
\int_{B_R}
\langle \nabla u_s , (1 + \langle x,\nabla \rangle )
\nabla u_{t}\rangle \,\dd x
=
4\pi{\textstyle\sum_{s,t}}\gamma_{s,t}N_s N_t
.\label{therefore}
\eeq
Pulling all limit results together, we obtain Proposition
4.10. $\QED$
\smallskip
\noin
{\it Remark.} The proof of the virial theorem extends to situations
when $\gamma$ does not have full rank, hence to more-than-two species beams.
\smallskip
\subsection{Concluding the proof of the Symmetry Theorem}
By Lemma 4.1, and by Proposition 4.10, the solutions
$u_s$ of \refeq{usPDE}, \refeq{usNs}, \refeq{usAC},
have to be radially symmetric and decreasing about a common center $x_0$.
Since the coupling matrix $\gamma$ is invertible, the
same conclusion holds for the solutions $U_s$ of
\refeq{UsPDE}, \refeq{UsNs}, \refeq{UsAC}.
The proof is complete. $\QED$
\medskip
\noin
{\bf Acknowledgement} The results reported here spun out of
my collaboration with Sagun Chanillo. Work supported by
NSF Grant DMS 9623220.
%\newpage
\begin{thebibliography}{[99]}
\bibitem{almgrenlieb}
Almgren, F. J., and Lieb, E. H.,
{\it Symmetric decreasing rearrangement is sometimes
continuous},
J. Amer. Math. Soc. {\bf 2}, 683--773 [1989].
\bibitem{bandleBOOK}
Bandle, C.,
{\it Isoperimetric inequalities and applications,}
Pitman, Boston [1980].
%
\bibitem{benford}
Benford, G., Book, D. L., and Sudan, R. N.,
{\it Relativistic beam equilibria with back currents},
Phys. Fluids {\bf 13}, 2621--2623 [1970].
%
\bibitem{bennettPAPa}
Bennett, W. H.,
{\it Magnetically self-focussing streams,}
Phys. Rev. {\bf 45}, 890--897 [1934].
%
\bibitem{bennettPAPb}
Bennett, W. H.,
{\it Self-focussing streams,}
Phys. Rev. {\bf 98}, 1584--1593 [1955].
%
\bibitem{brothersziemer}
Brothers, J. E., and Ziemer, W. P.,
{\it Minimal rearrangement of Sobolev functions,}
J. reine angew. Math. {\bf 384}, 153--179 [1988].
%
\bibitem{carlenloss}
Carlen, E.A, and Loss, M.,
{\it Competing symmetries, the logarithmic HLS
inequality, and Onofri's inequality on $\SS^n$},
Geom.\ Funct.\ Anal. {\bf 2}, 90--104 [1992].
%
\bibitem{chandra}
Chandrasekhar, S.,
{\it The highly collapsed configurations of a stellar mass},
Month. Not. R. astr. Soc. {\bf 91}, 456--466 [1931];
{\it Part II}, ibid. {\bf 95}, 207--225 [1935].
%
\bibitem{chakieCMP}
Chanillo, S., and Kiessling, M. K.-H.,
{\it Rotational symmetry of solutions to some nonlinear
problems in statistical mechanics and geometry,}
Commun. Math. Phys. {\bf 160}, 217--238 [1994].
%
\bibitem{chakieGAFA}
Chanillo, S., and Kiessling, M. K.-H.,
{\it Conformally Invariant Systems of Nonlinear
PDE of Liouville type,}
Geom. Funct. Anal. {\bf 5}, 924--947 [1995].
%
\bibitem{chakieCR}
Chanillo, S., and Kiessling, M. K.-H.,
{\it Symmetry of solutions of Ginzburg--Landau equations},
Compt. Rendus Acad. Sci. Paris, {\bf t. 321}, 1023--1026 [1995].
%
\bibitem{chakieNA}
Chanillo, S., and Kiessling, M. K.-H.,
{\it Curl-free Ginzburg--Landau vortices},
Nonlinear Anal. {\bf 38}, 933--949 [1999].
%
\bibitem{chakieDUKE}
Chanillo, S., and Kiessling, M. K.-H.,
{\it Surfaces with Prescribed Gauss Curvature,}
Duke Math. J., {\bf 105}, 309--353 [2000].
%
\bibitem{chenli}
Chen, W., and Li, C.,
{\it Classification of solutions of some
nonlinear elliptic equations},
Duke Math. J., {\bf 63 }, 615--622 [1991].
%
\bibitem{chipotshafrirwolansky}
Chipot, M., Shafrir, I., Wolansky, G.,
{\it On the solutions of Liouville systems},
J. Diff. Eq. {\bf 140}, 59--105 [1997];
and {\it errata}, ibid. (in press).
%
\bibitem{chouwan}
Chou, K. S., and Wan, T.Y.,
{\it Asymptotic radial symmetry for
solutions of $\Delta u+e^u=0$ in a punctured disc,}
Pac. J. Math. {\bf 163}, 269--276 [1994].
%
\bibitem{degrootBOOK}
de Groot, S. R., van Leeuwen, W. A., and van Wert, Ch. G.,
{\it Relativistic Kinetic Theory},
North Holland, Amsterdam [1980].
%
\bibitem{federer}
Federer, H.,
{\it Geometric measure theory},
Springer Verlag, Berlin [1996].
%
\bibitem{gidasninirbg}
Gidas, B., Ni., W.M., and Nirenberg, L.,
{\it Symmetry and related properties via the maximum principle},
Commun.\ Math.\ Phys. {\bf 68}, 209--243 [1979].
%
\bibitem{gilbargtrudinger}
Gilbarg, D., and Trudinger, N. S.,
{\it Elliptic partial differential equations of second order},
Springer Verlag, New York [1983].
%
\bibitem{humphries}
Humphries, S.
{\it Charged particle beams}, pp. 582ff.,
Wiley, New York [1990].
%
\bibitem{jostwangA}
Jost, J., and Wang, G.,
{\it Analytic aspects of a Toda system I:
a Moser-Trudinger inequality},
e-print, arXiv:math-ph0011039 [2000],
Commun. Pure Appl. Math. {\bf 54}, pp. 1289-1319 [2001].
%
\bibitem{jostwangB}
Jost, J., and Wang, G.,
{\it Classification of solutions of a Toda system in $\RR^2$},
e-print, arXiv:math-ph0105045 [2001],
Int. Math. Res. Notices 2002 (no.6), pp. 277-290 [2002].
%
\bibitem{juettnerCLASSa}
J\"uttner, F.,
{\it Das Maxwellsche Gesetz der Geschwindigkeitsverteilung in der
Relativtheorie},
Ann. Physik u. Chemie {\bf 34}, 856--882 [1911].
%
\bibitem{juettnerCLASSb}
J\"uttner, F.,
{\it Die Dynamik eines bewegten Gases in der
Relativtheorie},
Ann. Physik u. Chemie {\bf 35}, 145--161 [1911].
%
\bibitem{juettnerQM}
J\"uttner, F.,
{\it Die relativistische Quantentheorie des idealen Gases},
Z. Physik {\bf 47}, 542--566 [1928].
%
\bibitem{kawohl}
Kawohl, B.,
{\it Rearrangements and convexity of level sets in PDE},
Springer-Verlag, New York [1985].
%
\bibitem{kesavanpacella}
Kesavan, C., and Pacella, F.,
{\it Symmetry of positive solutions of a quasilinear
elliptic equation via isoperimetric inequalities},
Appl. Anal. {\bf 54}, 27--37 [1994].
%
\bibitem{kiejllPhPl}
Kiessling, M. K.-H., and Lebowitz, J. L.,
{\it Dissipative Stationary Plasmas: Kinetic Modeling,
Bennett's Pinch and Generalizations,}
Phys. Plasmas {\bf 1}, 1841--1849 [1994].
%
\bibitem{joelA}
Lebowitz, J. L., and Bergmann, P.G.,
{\it New Approach to Nonequilibrium Process,}
Phys. Rev. {\bf 99}, 578--587 [1955].
%
\bibitem{joelB}
Lebowitz, J. L., and Bergmann, P.G.,
{\it Irreversible Gibbsian Ensembles,}
Annals of Physics (N.Y.), {\bf 1}, 1--23 [1957].
%
\bibitem{joelC}
Lebowitz, J. L., and Frisch, H.,
{\it Model of a Nonequilibrium Ensemble: The Knudsen Gas,}
Phys. Rev. {\bf 107}, 917--923 [1957].
%
\bibitem{joelD}
Lebowitz, J. L.,
{\it Stationary Nonequilibrium Gibbsian Ensembles},
Phys. Rev. {\bf 114}, 1192--1202 [1959].
%
\bibitem{li}
Li, C.-M.,
{\it Monotonicity and symmetry of solutions of fully nonlinear
elliptic equations on unbounded domains},
Commun. PDE {\bf 16}, 585--615 [1991].
%
\bibitem{liebyau}
Lieb, E.H., and Yau, H.-T.,
{\it The Chandrasekhar theory of stellar collapse as the limit
of quantum mechanics},
Commun. Math. Phys. {\bf 112}, 147--174 [1987].
%
\bibitem{lionsPAPa}
Lions,~P.-L.,
{\it Two geometrical properties of solutions of semilinear problems,}
Appl.\ Anal. {\bf 12}, 267--272 [1981].
%
%\bibitem{lionsPAPb}
%Lions,~P.-L.,
%{\it The concentration-compactness principle in the calculus
%of variations. The locally compact case,} part 1,
%Ann.\ Inst.\ H.\ Poincar\'{e}, Anal. Nonlin., {\bf 1}, 109--145;
%part 2, ibid. 223-283 [1984].
%
%\bibitem{lionsPAPc}
%Lions,~P.-L.,
%{\it The concentration-compactness principle in the calculus
%of variations. The limit case,} part 1,
%Rev. Math. Iberoamer., {\bf 1}, 145--201; part 2, ibid. 45-121 [1985].
%
\bibitem{liouville}
Liouville, J.,
{\it Sur l'\'equation aux diff\'erences partielles
$\partial^2 \log \lambda/\partial u\partial v \pm \lambda/ 2 a^2 =0$},
J. de Math. Pures Appl. {\bf 18}, 71--72 [1853].
%
\bibitem{pokhozaev}
Pokhozaev, S. I.,
{\it Eigenfunctions of the equation $\Delta u + \lambda f(u) = 0$,}
Sov. Math. Dokl. {\bf 5}, 1408--1411 [1965].
%
\bibitem{rellich}
Rellich, F.,
{\it Darstellung der Eigenwerte von $\Delta u + \lambda u = 0$
durch ein Randwertintegral,}
Math. Z. {\bf 46}, 635--636 [1940].
%
\bibitem{serrin}
Serrin, J.,
{\it A symmetry problem in potential theory},
Arch. Rat. Mech. Anal. {\bf 43}, 304--318 [1971].
%
\bibitem{liebJUBILEE}
Thirring, W.E.,
Preface to: {\it The State of Matter} (M. Aizenman and H. Araki, Ed.),
Adv. Ser. Math. Phys. {\bf 20}, World Scientific, Singapore [1994].
%
\bibitem{troy}
Troy, W.C.,
{\it Symmetry properties in systems of semilinear
elliptic equations},
J. Diff. Eq. {\bf 42}, 400--413 [1981].
%
\bibitem{vlasovBOOK}
Vlasov, A. A.,
{\it Many-Particle Theory and its Application to Plasma},
Gordon and Breach, New York [1961].
%
\bibitem{weinberg}
Weinberg, S.,
{\it General theory of resistive beam instabilities},
J. Math. Phys. {\bf 8}, 614--641 [1967].
%
\end{thebibliography}
%
\end{document}