Fedosov *-products and quantum momentum maps
Extracted source text ยท arXiv q-alg/9608006
\documentclass[11pt]{article}
\usepackage{amssymb}
\usepackage{latexsym}
\textwidth 6.5in
\textheight 8.5in
\oddsidemargin .1truein
\newtheorem{thm}{Theorem}[section]
\newtheorem{lem}[thm]{Lemma}
\newtheorem{cor}[thm]{Corollary}
\newtheorem{pro}[thm]{Proposition}
\newtheorem{ex}[thm]{Example}
\newtheorem{rmk}[thm]{Remark}
\newtheorem{defi}[thm]{Definition}
%\newremark{ex}[thm]{Example}
\newcommand{\alp }{\alpha }
\newcommand{\bet }{\beta }
\newcommand{\gm }{\Gamma }
\newcommand{\lon }{\longrightarrow }
\newcommand{\be }{\begin{eqnarray*}}
\newcommand{\ee }{\end{eqnarray*}}
%\newcommand{\pp }{\begin{flushright} {\bf Q.E.D. }\ \ \
% \end{flushright} \bigskip }
\newcommand{\gt }{ {\frakg}\otimes {\frakg} }
\newcommand{\per }{\backl }
\newcommand{\te }{\otimes }
\newcommand{\st }[1]{\stackrel{#1}{\lon}}
\newcommand{\poidd }[2]{#1\gpd #2}
\newcommand{\poiddd }[3]{ (#1\gpd #2, \alpha_{#3}, \beta_{#3})}
\newcommand{\gr}{ {\frakg } }
\newcommand{\tu}{\tilde{u}}
\newcommand{\tg}{\tilde{g}}
\newcommand{\dress}{\lambda_{\psi (v^{-1})}u}
\newcommand{\dre}[1]{\lambda_{#1}}
\newcommand{\gmstar}{\gm_{G^{*}}}
\newcommand{\del}[1]{\delta_{x_{#1}}}
\setlength{\baselineskip}{1.8\baselineskip}
%Alan's macros:
%\input mssymb
\input amssym.def
\input amssym
\def\nin{\noindent}
\def\eg{e.g.\ }
\def\ie{i.e.\ }
%\def\pt#1#2{{\partial #1\over \partial #2}}
\newcommand{\pf}{\noindent{\bf Proof.}\ }
%\newcommand{\del}{\partial}
\newcommand{\qed}{\begin{flushright} $\Box$\ \ \ \ \ \
\end{flushright}}
\newcommand{\complex}{{\Bbb C}}
\newcommand{\reals}{{\Bbb R}}
%\newcommand{\r}[1]{{\Bbb R}^{#1}}
\newcommand{\ctn}{\complex^{2n}}
\newcommand{\rtn}{\reals^{2n}}
\newcommand{\rn}{\reals^{n}}
\newcommand{\rtm}{\reals^{2m}}
\newcommand{\frakb}{{\frak b}}
\newcommand{\frakd}{{\frak d}}
\newcommand{\frakf}{{\cal F}}
\newcommand{\frakg}{{\frak g}}
\newcommand{\frakh}{{\frak h}}
\newcommand{\frakl}{{\frak l}}
\newcommand{\fraks}{{\frak s}}
\newcommand{\bstar}{{\frak b}^{*}}
\newcommand{\gstar}{{\frak g}^{*}}
\newcommand{\hstar}{{\frak h}^{*}}
\newcommand{\lstar}{{\frak l}^{*}}
\newcommand{\sstar}{{\frak s}^{*}}
\newcommand{\starh}{*_{\hbar}}
\newcommand{\zbar}{\overline{z}}
\newcommand{\torus}{{\Bbb T}}
\newcommand{\integers}{{\Bbb Z}}
\newcommand{\boldphi}{\mbox{\boldmath $\phi$}}
\newcommand{\boldpi}{\mbox{\boldmath $\pi$}}
\newcommand{\inner}{\backl}
\newcommand{\grad}{\nabla}
\newcommand{\half}{\frac{1}{2}}
\newcommand{\quarter}{\frac{1}{4}}
\newcommand{\cinf}{C^{\infty}}
\newcommand{\canoncoords}{q_{1},\dots,q_{n},p_{1},\dots ,p_{n}}
\newcommand{\canonbracket}{\sum _{i=1}^{n}
[\frac{\del f}{\del q_{i}}\frac{\del g}{\del p_{i}}-
\frac{\del g}{\del q_{i}}\frac{\del f}{\del p_{i}}]}
\newcommand{\symplecticform}{\sum_{i=1}^{n}dq_{i}\wedge dp_{i}}
\newcommand{\poissend}[1]{#1\times\overline{#1}}
\newcommand{\oppositepoissend}[1]{\overline{#1}\times #1}
\newcommand{\threetimes}[1]{#1 \times \overline{#1}\times\overline{#1}}
\newcommand{\fourtimes}[1]{#1\times\overline{#1}\times\overline{#1}\times
#1}
\newcommand{\inverse}{^{-1}}
\newcommand{\boldc}{{\bf c}}
\newcommand{\boldi}{{\bf i}}
\newcommand{\boldj}{{\bf j}}
\newcommand{\boldk}{{\bf k}}
\newcommand{\boldR}{{\bf R}}
\newcommand{\boldS}{{\bf S}}
\newcommand{\boldx}{{\bf x}}
\newcommand{\boldy}{{\bf y}}
\newcommand{\D}{{\bf D}}
% \newcommand{\F}{{\bf F}}
\newcommand{\pr}{\mbox{pr}}
\newcommand{\tr}{\trace}
\newcommand{\trace}{\mbox{\rm{Tr}}}
\newcommand{\area}{\mbox{\rm{Area}}}
\newcommand{\id}{\mbox{\rm{id}}}
\newcommand{\backl}{\mathbin{\vrule width1.5ex height.4pt\vrule height1.5ex}}
% \newcommand{\comment}{\tt}
\newcommand{\cala}{{\cal A}}
\newcommand{\calb}{{\cal B}}
\newcommand{\calc}{{\cal C}}\newcommand{\cald}{{\cal D}}
\newcommand{\cale}{{\cal E}}
\newcommand{\calf}{{\cal F}}
\newcommand{\calg}{{\cal G}}
\newcommand{\calh}{{\cal H}}
\newcommand{\cali}{{\cal I}}
\newcommand{\calj}{{\cal J}}
\newcommand{\calk}{{\cal K}}
\newcommand{\call}{{\cal L}}
\newcommand{\calm}{{\cal M}}
\newcommand{\caln}{{\cal N}}
\newcommand{\calo}{{\cal O}}
\newcommand{\calp}{{\cal P}}
\newcommand{\calq}{{\cal Q}}
\newcommand{\calr}{{\cal R}}
\newcommand{\calx}{{\cal X}}
\newcommand{\caly}{{\cal Y}}
\newcommand{\defequal}{\stackrel{\mbox {\tiny {def}}}{=}}
\newcommand{\smalcirc}{\mbox{\tiny{$\circ $}}}
\newcommand{\pp}{\qed}
% Dawn's newcommands
\newcounter{bean}
\newcounter{bacon}
\newcounter{in}
\def\description label#1{\hfil\bf[#1]\hfil}
\parskip 5pt plus 1pt
\topmargin 4pt
% \newcommand{\Z}{\bf{Z}\rm}
% \newcommand{\C}{\bf{C}\rm}
% \newcommand{\R}{$\cal R$ }
% \newcommand{\N}{\cal N}
% \newcommand{\K}{\cal K}
\newcommand{\M}{\cal M}
\newcommand{\X}{\cal X}
\newcommand{\s}{\cal S}
\newcommand{\g}{\frak g}
\newcommand{\h}{\frak h}
\newcommand{\slm}{\boldS_{\lambda}\otimes \boldS_{\mu} }
\newcommand{\rlm}{\calr_{\lambda \mu}}
\newcommand{\rml}{\calr_{\mu \lambda }}
\newcommand{\sla}{S_{\lambda}}
\newcommand{\sml}{\boldS_{\mu}\otimes \boldS_{\lambda} }
\newcommand{\smu}{S_{\mu}}
\newcommand{\gmdouble}{\gmstar \times \gmstar}
\newcommand{\jml}{J_{\mu}\otimes J_{\lambda}}
\newcommand{\jlm}{J_{\lambda}\otimes J_{\mu}}
\newcommand{\jm}{J_{\mu}}
\newcommand{\jl}{J_{\lambda }}
\newcommand{\ssss}{S\times S\lon S\times S}
\newcommand{\hcalr}{\hat{\calr}}
\newcommand{\oo}{o}
\newcommand{\gdouble}{\gr \oplus (\gr^{*})^{\oo}}
\newcommand{\hdouble}{\h \oplus (\h^{*})^{\oo}}
\newcommand{\su}[1]{SU(#1)}
\newcommand{\sbb}[1]{SB(#1, \complex )}
\newcommand{\sll}[1]{SL(#1, \complex )}
\newcommand{\sustar}[1]{\frak{su}^{*}(#1)}
\newcommand{\liesl}[1]{\frak{sl}(#1, \complex )}
\newcommand{\ddouble}{H\times (H^{*})^{\oo}}
\newcommand{\hhdouble}{H\bowtie H^{*}}
\newcommand{\ggdouble}{G\bowtie G^{*}}
\newcommand{\hr}{\hat{R}}
\newcommand{\Hom}{\mbox{\bf Hom}}
\newcommand{\sect}{\gm(A)}
\newcommand{\secdual}{\gm(A^{*})}
\newcommand{\secc}[1]{\gm(\wedge^{#1}A)}
\newcommand{\secd}[1]{\gm(\wedge^{#1}A^{*})}
% \newcommand{\sect}{\gm^{\infty}(A)}
% \newcommand{\secdual}{\gm^{\infty}(A^{*})}
% \newcommand{\secc}[1]{\gm^{\infty}(\wedge^{#1}A)}
% \newcommand{\secd}[1]{\gm^{\infty}(\wedge^{#1}A^{*})}
\newcommand{\parr}[1]{\frac{\partial}{\partial x^{#1}}}
\newcommand{\parrr}[2]{\frac{\partial #1}{\partial x^{#2}}}
\newcommand{\du}{\dot{\mu}}
\newcommand{\dx}{\dot{x}}
\newcommand{\anc}{a}
\newcommand{\ancd}{a_{*} }
\newcommand{\ot}{\omega \wed \theta}
\newcommand{\xia}{\xi_{a}}
\newcommand{\Ga}{\Gamma}
\newcommand{\pib}{\pi^{\#}}
\newcommand{\talp}{\tilde{\alpha}}
\newcommand{\tbet}{\tilde{\beta}}
\newcommand{\difft}{\frac{d}{dt}}
\newcommand{\rt}{\Lambda}
\newcommand{\rr}{\Lambda^{\#}}
\newcommand{\rrr}{[\Lambda , \Lambda ] ^{\#}}
\newcommand{\diff}[1]{\frac{d}{d#1}}
\newcommand{\pttt}[2]{\frac{\partial #1}{ \partial #2}}
\newcommand{\ttt}{t_{0}}
\newcommand{\gmex}{U_{ex}(\gm )}
\newcommand{\gmm}{U_{0}(\gm )}
\newcommand{\eex}{U_{0}(E)}
%\newcommand{\goth}{\frak}
\newcommand{\ih}{\frac{i}{\hbar}}
\newcommand{\poll}{pol (\frakg^{*})}
\newcommand{\pol}{U\frakg }
\newcommand{\ad}{ad}
\newcommand{\muh}{\mu_{\hbar}}
\newcommand{\commutant}{commutant }
\newcommand{\equivariant}{equivariant }
\newcommand{\preferred}{regular }
\newcommand{\nat}{natural }
% Kirill:
\renewcommand{\theenumi}{{\rm{(\roman{enumi})}}}
\def\Clin{C^\infty_{\ell in}}
\def\mvf{multiplicative vector field}
\def\mmvf{multiplicative multivector field}
\def\cdo{covariant differential operator}
\def\LAgpd{${\cal LA}$-groupoid}
\def\VBLalgd{${\cal VB}$-Lie algebroid}
\def\VBgpd{${\cal VB}$-groupoid}
\def\CDO{\mathop{\rm CDO}}
\def\End{\mathop{\rm End}}
\def\ST{\ \vert\ }
\def\sdp{\mathbin{\hbox{$\mapstochar\kern-.3333em\times$}}}
\def\pds{\mathbin{\hbox{$\times\kern-.55em\mapstochar\,$}}}
\newcommand{\wed}{\mathbin{\lower1.5pt\hbox{$\scriptstyle{\wedge}$}}}
\let\Tilde=\widetilde
\let\Bar=\overline
\let\Vec=\overrightarrow
\let\ceV=\overleftarrow
\let\Hat=\widehat
\let\isom=\cong
\let\sol=\bullet
\def\chigh{{\raise1.5pt\hbox{$\chi$}}}
\let\phi=\varphi
\def\til0{\Tilde{0}}
\def\dpl{\mathbin{+\hskip-6pt +\hskip4pt}}
\def\dminus{\raise2pt\hbox{\vrule height1pt width 2ex}\hskip3pt}
\def\dtimes{\mathbin{\hbox{\huge.}}}
\def\llangle{\langle\!\langle}
\def\rrangle{\rangle\!\rangle}
\def\pback#1{\mathbin{{{\lower1.2ex\hbox{$\times$}}\atop #1}}}
\def\ddt#1{\left.\frac{d}{dt}#1\right|_0}
\def\brev{\:\breve{\rule{0pt}{8pt}}}
\def\hatt{\Hat{\phantom{X}}}
\def\vlra{\hbox{$\,-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!
-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!\longrightarrow\,$}}
\def\vleq{\hbox{$\,=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\!\!\!
=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\!\!\!=\,$}}
\def\lrah{\hbox{$\,-\!\!\!-\!\!\!
-\!\!\!-\!\!\!-\!\!\!-\!\!\!-\!\!\!\longrightarrow\,$}}
\def\surj{-\!\!\!-\!\!\!-\!\!\!\gg}
\def\inj{>\!\!\!-\!\!\!-\!\!\!-\!\!\!>}
%The following \gpd is for use in text to indicate groupoid structure.
\def\gpd{\,\lower1pt\hbox{$\longrightarrow$}\hskip-.24in\raise2pt
\hbox{$\longrightarrow$}\,}
\def\lgpd{\,\lower1pt\hbox{$\vlra$}\hskip-1.02in\raise2pt\hbox{$\vlra$}\,}
\def\llgpd{\,\lower1pt\hbox{$\vvlra$}\hskip-1.3in\raise2pt\hbox{$\vvlra$}\,}
\def\vgpd{\Bigg\downarrow\!\!\Bigg\downarrow}
\def\vequals{\Bigg\|}
% -----
\hyphenation{group-oid group-oids}
\begin{document}
\title{{\bf Fedosov $*$-products and quantum momentum maps}}
%\thanks{1991 {\em Mathematics
%Subject Classification.} Primary 58F05. Secondary 22E65, 22A22, 58B25, 58H05.}}
\author{ PING XU \thanks{ Research partially supported by NSF
grant DMS95-04913}\\
Department of Mathematics\\The Pennsylvania State University \\
University Park, PA 16802, USA\\
{\sf email: ping@math.psu.edu }}
\date{August, 1996}
\maketitle
\begin{abstract}
The purpose of the paper is to study various aspects of star products on
a symplectic manifold related to the Fedosov method.
By introducing the notion of ``quantum exponential maps",
we give a criterion characterizing Fedosov connections.
As a consequence, a geometric realization is
obtained for the equivalence between an arbitrary $*$-product and
a Fedosov one.
%which was a result due to Nest-Tsygan and Deligne.
Every Fedosov $*$-product is shown to be a Vey $*$-product.
Consequently, one obtains that
every $*$-product is equivalent to a Vey $*$-product, a classical result
of Lichnerowicz.
Quantization of a hamiltonian $G$-space, and in particular, quantum
momentum maps are studied. Lagrangian submanifolds are also
studied under a deformation quantization.
\end{abstract}
\section{Introduction}
In classical mechanics, observables are smooth functions on a phase
space, which consist of a Poisson algebra, while in quantum mechanics
observables become a noncommutative associative algebra.
Deformation quantization, as laid out by
Bayen, Flato, Fronsdal, Lichnerowicz and Sternheimer in 1970's \cite{BFFLS}, is
one of the important attempts
aiming to establish a correspondence principle between these two mechanics.
A classical phase space is usually a symplectic manifold $M$.
A deformation quantization, or more precisely a star-product is a
family of associative multiplication $*_{\hbar}$ (depending on the Planck
constant $\hbar$) on $C^{\infty}(M)[[\hbar]]$, the
space of formal power series with coefficients in $C^{\infty}(M)$:
\begin{equation}
f*_{\hbar}g=fg -\frac{i\hbar }{2}\{f , g\} + \cdots +\hbar^{k}C_{k}(f, g)+\cdots , \ \ \forall
f, g\in C^{\infty}(M)[[\hbar]]
\end{equation}
such that
\begin{enumerate}
\item $C_{k}(f, g)=(-1)^{k}C_{k}(g, f)$;
\item $C_{k}(1, f)=C_{k}(f, 1)=0$, for $k\geq 1$;
\item each $C_{k}(\cdot , \cdot )$ is a bidifferential operator.
\end{enumerate}
Here $\{f , g\}$ is the Poisson bracket on the symplectic manifold $M$.
Such a $*$-product always exists on a symplectic vector space $V$, which is
known as Moyal-Weyl formula:
\begin{equation}
\label{eq:mw}
f *_{\hbar} g=\sum_{k=0}^{\infty} (-\frac{i\hbar}{2})^{k}\frac{1}{k!}\pi^{i_{1}j_{1}}
\cdots \pi^{i_{k}j_{k}}
\frac{\partial^{k} f}{\partial y^{i_{1}}\cdots \partial y^{i_{k}}}
\frac{\partial^{k} g}{\partial y^{j_{1}}\cdots \partial y^{j_{k}}}, \ \ \forall f , g\in C^{\infty}(V)[[\hbar ]],
\end{equation}
where $y^{1}, \cdots , y^{2n}$ are linear coordinates on $V$, and $\pi^{ij}=\{y^{i} , y^{j}\}$.
It is simple to see that this definition is independent of
the choice of linear coordinates.
Moyal-Weyl formula has a straightforward generalization
to the case of a symplectic manifold admitting a flat torsion free
symplectic connection $\nabla$, as shown in \cite{BFFLS}.
In this case, for any $f, \ g\in C^{\infty}(M)[[\hbar ]]$, set
\begin{equation}
\label{eq:generalized-mw}
(f*_{\hbar }g)(x)=[(exp^{*}_{x}f)(y)*_{\hbar }(exp^{*}_{x}f)(y)]|_{y=0},
\end{equation}
where $exp_{x}: T_{x}M\lon M$ is the exponential map, defined
in a neighborhood of the origin, corresponding to the connection
$\nabla$, and the $*_{\hbar }$-product on the RHS refers to the standard
Moyal-Weyl $*$-product on the symplectic vector space $T_{x}M$.
More explicitly, one has
\begin{equation}
(f*_{\hbar }g)(x) =
\sum_{k=0}^{\infty} (-\frac{i\hbar}{2})^{k}\frac{1}{k!}
\pi^{i_{1} j_{1}}\cdots \pi^{ i_{k} j_{k}}(\partial_{i_{1}}\cdots \partial_{i_{k}}f)
(\partial_{j_{1}}\cdots \partial_{j_{k}}g).
\end{equation}
However, the multiplication defined by Equation (\ref{eq:generalized-mw})
fails to be associative when $\nabla$ has a curvature.
The existence proof of $*$-products on a general
symplectic manifold was first obtained by de Wilde and Lecomte \cite{DeL}
using a homological argument. Later, an alternate proof using Weyl
manifolds was found by Omori et al.\cite{OMY}.
Guillemin showed that their results in \cite{GV} also
implies the existence of $*$-products.
Recently, Fedosov has given a nice geometrical existence proof \cite{F1}
\cite{F2}, which
provides a useful tool for understanding $*$-products geometrically.
This paper grew out from an attempt to understand the Fedosov method, as well as
an attempt to solve some other related problems
using such a method.
Roughly speaking, Fedosov's method is to make a ``quantum correction"
to Equation (\ref{eq:generalized-mw}) when the curvature exists.
As discussed early in this introduction, each tangent space $T_{x}M$ is
a symplectic vector space, hence can be quantized using the standard
Moyal-Weyl product. This enables us to obtain a bundle of algebras $W\lon M$,
called Weyl bundle by Fedosov, which can be thought of as
a kind of ``quantum tangent bundle". Fedosov found a nice iteration method
of constructing a flat connection on the Weyl bundle, whose
parallel sections can be naturally identified with
$C^{\infty}(M)[[\hbar ]]$. Thus the product on the bundle induces
a $*$-product on $C^{\infty}(M)[[\hbar ]]$. Intuitively, Fedosov
connection can be thought of as a ``quantum connection" on the ``quantum
tangent bundle" $W$,
which is indeed obtained by adding some ``quantum correction" to
the usual affine connection on the tangent bundle (see Section 2).
Then, the correspondence between $C^{\infty}(M)[[\hbar ]]$ and
its parallel sections can be considered as taking the exponential map for such
a quantum connection. This viewpoint has already been pointed out by Weinstein
(see \cite{Weinstein:1995} \cite{EW}), and is essential to our introduction
of the notion of ``quantum exponential maps".
By a ``quantum exponential map", we mean a map from $C^{\infty}(M)[[\hbar ]]$
to the space of sections $\gm W$ of the Weyl bundle, which
satisfies certain axioms (see Section 3 for details). In fact, we
will show that any quantum exponential map is equivalent
to a Fedosov connection, which is what is expected: connections
and exponential maps are equivalent. This result will enable us to
characterize those subalgebras of $\gm W$ arising from
Fedosov connections. As an application, we will
realize geometrically
the equivalence between any given $*$-product and
a Fedosov one, a result proved by several authors including
Nest-Tsygan \cite{NT1}, Deligne \cite{Deligne} and recently Bertelson-Cahen-Gutt \cite{BCG}.
% Such an equivalence is essential for the
%classification of $*$-products on a symplectic manifold (see \cite{NT1}).
Since a ``quantum exponential map" is a ``quantum correction" to the usual
exponential map, the $*$-products obtained by Fedosov method, called
Fedosov $*$-products in the paper,
should be closely related to
Equation (\ref{eq:generalized-mw}). A Vey $*$-product,
by definition, is a $*$-product
where the principal terms coincide with those as in Equation (\ref{eq:generalized-mw})
(see Section 4 for the precise definition).
Vey $*$-products have played an important role since the beginning of
deformation quantization theory (see \cite{BFFLS} \cite{L} \cite{V}).
In this paper, we will show that every Fedosov $*$-product
is a Vey $*$-product. As a consequence, we recover
the following well known result of Lichnerowicz \cite{L}:
any $*$-product is equivalent to a Vey $*$-product.
Because of the basic simplicity of its construction,
Fedosov method provides us a useful tool for studying some other problems
in deformation quantization theory. In this paper,
in particular, we will study the following question:
\begin{quote}
What do lagrangian submanifolds correspond under a deformation
quantization?
\end{quote}
Given a lagrangian submanifold $L$, the space $C^{\infty}_{L}(M)$ of smooth functions
vanishing on $L$ forms a Poisson subalgebra. For any
$*$-product, we will show that under some ``quantum
correction" $C^{\infty}_{L}(M)[[\hbar ]]$ becomes a subalgebra.
In symplectic geometry, hamiltonian $G$-spaces, and in particular, momentum
maps play a very special role. It is natural to ask: what
is the quantum analogue of momentum maps? The second part of the paper,
as another application of Fedosov method,
is devoted to the study of quantum momentum maps.
In particular, we will derive a sufficient condition
for their existence.
When a quantum momentum map exists, we obtain a pair of
mutual commutants which can be considered as a
quantum analogue of the well known Poisson dual pair $\frakg^{*}\stackrel{J}{\longleftarrow}M
\stackrel{pr}{\lon}M/G$ of Weisntein \cite{Weinstein:1983}.
Derivations are always important for associative algebras. In Appendix A,
we will collect some basic results regarding
derivations on a $*$-algebra $C^{\infty}(M)[[\hbar ]]$,
which are needed for the study of quantum momentum maps.
Most of them are proved using Fedosov method. Even for
some well known results, we will see that Fedosov method provides
a simpler way in understanding them.
While the paper is in writing,
it came to the author's attention that quantum momentum maps are also being
under study by some other authors including Astashkevich \cite{Ast}, Kostant
and Tsygan. In particular, Theorem \ref{eq:mom-exist} has also been proved by Kostant.
%Recently, we noticed that a result similar to Theorem \ref{thm:vey}
%has also been obtained independently by Bordemann et al. (NEED TO
%BE VERIFIED!).
{\bf Acknowledgements.} The author would like to thank Pierre Bieliavsky, Jean-Luc Brylinski,
Ranee Brylinski, Moshe Flato,
Victor Guillemin, Yuri Manin, Marc Rieffel, Daniel Sternheimer,
Boris Tsygan and Alan Weinstein for useful discussions. In
addition to the funding sources mentioned
in the first footnote, he would also like to thank IHES and Max-Planck-Institut
for their hospitality and financial support while part of this project was being done.
\section{Fedosov quantization}
In this section, we will recall some basic
ingredients of
Fedosov construction of
$*$-products on a symplectic manifold, as well as some
useful notations, which will
be needed in the future. For details, readers should
consult \cite{F1} \cite{F2}.
Let $(M, \omega )$ be a symplectic manifold of dimension $2n$.
Then, each tangent space $T_{x}M$ is equipped with a linear symplectic
structure, which can be quantized by the standard Moyal-Weyl product.
The resulting space is denoted by $W_{x}$.
More precisely,
\begin{defi}
A formal Weyl algebra $W_{x}$ associated to $T_{x}M$
is an associative algebra with a unit over $\complex $,
whose elements consist of formal power series in the formal parameter
$\hbar $ with coefficients being formal polynomials
in $T_{x}M$. In other words,
each element has the form
\begin{equation}
\label{eq:general}
a(y, \hbar )=\sum \hbar^{k}a_{k, \alpha }y^{\alpha }
\end{equation}
where $y=(y^{1}, \cdots , y^{2n})$ is a linear coordinate
on $T_{x}M$, $\alpha =(\alpha_{1}, \cdots , \alpha_{2n})$
is a multi-index and $y^{\alpha}=(y^{1})^{\alpha_{1}}\cdots
(y^{2n})^{\alpha_{2n}}$.
The product is defined by the Moyal-Weyl rule:
\begin{equation}
\label{eq:moyal}
a *b=\sum_{k=0}^{\infty} (-\frac{i\hbar}{2})^{k}\frac{1}{k!}\pi^{i_{1}j_{1}}
\cdots \pi^{i_{k}j_{k}}
\frac{\partial^{k} a}{\partial y^{i_{1}}\cdots \partial y^{i_{k}}}
\frac{\partial^{k} b}{\partial y^{j_{1}}\cdots \partial y^{j_{k}}} .
\end{equation}
\end{defi}
%Note that although Equation (\ref{eq:moyal}) is given using local coordinates
%it is in fact coordinate free.
Let $W=\cup_{x\in M}W_{x}$. Then $W$ is a bundle of algebras over $M$, called
the Weyl bundle over $M$.
Its space of sections $\gm W$ forms an associative algebra
with unit under
the fiberwise multiplication.
One may think of $W$ as a ``quantum tangent bundle"
of $M$, whose space of sections $\gm W$ gives
rise to a deformation quantization for the
tangent bundle $TM$ considered as a Poisson manifold,
with fiberwise linear symplectic structure on $T_{x}M, \ \ x\in M$.
The center $Z (W)$ of $\gm W$ consists of sections
not containing $y's$, thus can be naturally
identified with $C^{\infty}(M)[[\hbar ]]$.
By assigning degrees to $y's$ and $\hbar$
with $deg y^{i}=1$ and $deg \hbar =2$, there
is a natural filtration
$$\gm (W)\subset \gm (W_{1})\subset \cdots \gm (W_{i})\subset \gm (W_{i+1})\cdots $$ with respect
to the total degree $2k+l$ of the series terms in Equation (\ref{eq:general}).
A differential form with values in $W$ is a section of the
bundle $W\otimes \wedge^{q}T^{*}M$, which
can be expressed locally as
\begin{equation}
a(x, y, \hbar , dx)=\sum \hbar^{k} a_{k, i_{1}\cdots i_{p}, j_{1}\cdots j_{q}}
y^{i_{1}}\cdots y^{i_{p}}dx^{j_{1}}\wedge \cdots \wedge dx^{j_{q}}.
\end{equation}
Here the coefficient $a_{k, i_{1}\cdots i_{p}, j_{1}\cdots j_{q}} $
is a covariant tensor symmetric with respect to
$i_{1} \cdots i_{p}$ and antisymmetric in $j_{1}\cdots j_{q}$.
For short, we denote the space of
sections of the bundle $W\otimes \wedge^{q}T^{*}M$ by
$\gm W\otimes \Lambda^{q}$.
The usual exterior derivative on differential forms
extends, in a straightforward way,
to an operator $\delta$ on $W$-valued
differential forms:
\begin{equation}
\delta a=dx^i\wedge \frac{\partial a}{\partial y^i}, \ \ \ \forall a\in \gm W\otimes \Lambda^* .
\end{equation}
By $\delta ^{-1}$, we denote its ``inverse" operator as defined by:
\begin{equation}
\delta ^{-1}a=\frac{1}{p+q}y^{i} ( \frac{\partial }{\partial x^i}\per a)
\end{equation}
when $p+q>0$, and $\delta ^{-1}a=0$ when $p+q=0$.
Here $a\in \gm W\otimes \Lambda^{q}$ is
homogeneous of degree $p$ in $y$.
There is a ``Hodge"- decomposition:
\begin{equation}
a=\delta \delta ^{-1}a+\delta ^{-1}\delta a+a_{0 0}, \ \ \ \forall a\in \gm W\otimes \Lambda^{*},
\end{equation}
where $a_{00}(x)$ is the constant term of $a$, i.e, the $0$-form
term of $a|_{y=0}$ or $a_{00}(x)=a(x, 0, 0, 0)$.
The operator $\delta$ resembles most
basic properties of the usual exterior
derivatives. For example,
$$\delta^{2}=0 \ \ \mbox{and } (\delta^{-1})^{2}=0 .$$
Let $\nabla$ be a torsion-free symplectic connection on $M$, and $\partial
: \gm W \lon \gm W \otimes \Lambda^{1}$ its induced covariant
derivative.
Consider a connection on $W$ of the form:
\begin{equation}
\label{eq:connection}
D=-\delta +\partial + \ih [\gamma ,\cdot \ ],
\end{equation}
with $\gamma \in \gm W \otimes \Lambda^{1}$.
Clearly, $D$ is a derivation with respect to the Moyal-Weyl product, i.e.,
\begin{equation}
D(a* b)=a * Db +Da* b.
\end{equation}
A simple calculation yields that
\begin{equation}
D^{2}a=-[\ih \Omega , a] , \ \forall a\in \gm W,
\end{equation}
where
\begin{equation}
\label{eq:curvature}
\Omega =\omega -R+\delta \gamma -\partial \gamma -\ih \gamma^{2}.
\end{equation}
Here $R=\frac{1}{4}R_{ijkl}y^{i}y^{j}dx^{k}\wedge dx^{l}$ and
$R_{ijkl}=\omega_{im}R^{m}_{jkl}$ is the curvature tensor
of the symplectic connection.
A connection of the form (\ref{eq:connection})
is called {\em Abelian} if
$\Omega $ is a scalar 2-form, i.e., $\Omega \in \Omega^{2}(M)[[\hbar ]]$.
It is called a {\em Fedosov
connection } if it is Abelian and in addition $\gamma \in \gm W_{3}\otimes \Lambda^{1}$.
For an Abelian connection, the Bianchi identity implies that
$d\Omega =D\Omega =0$, i.e., $\Omega \in Z^{2}(M)[[\hbar ]]$.
In this case, $\Omega $ is called the Weyl curvature.
\begin{thm} (Fedosov)
\label{label:f1}
Let $\partial $ be any torsion free symplectic connection,
and
$\Omega =\omega +\hbar \omega_{1} +\cdots \in
Z^{2}(M)[[\hbar ]]$ a perturbation of the symplectic form
in the space $Z^{2}(M)[[\hbar ]]$.
There exists
a unique $\gamma \in \gm W_{3}\otimes \Lambda^{1}$
such that $D$, given by Equation (\ref{eq:connection}),
is a Fedosov connection, which has Weyl curvature $\Omega$ and
satisfies
$$\delta^{-1}\gamma =0.$$
\end{thm}
\pf It suffices to solve the equation:
\begin{equation}
\omega -R+\delta \gamma -\partial \gamma -\ih \gamma^{2}=\Omega,
\end{equation}
which is equivalent to
\begin{equation}
\label{eq:r}
\delta \gamma =\tilde{\Omega}+\partial \gamma +\ih \gamma^{2} ,
\end{equation}
where $\tilde{\Omega}=\Omega -\omega+R$.
Applying the operator $\delta^{-1}$ to Equation (\ref{eq:r})
and using the Hodge decomposition, we obtain
\begin{equation}
\label{eq:r-iteration}
\gamma =\delta^{-1}\tilde{\Omega}+\delta^{-1}(\partial \gamma +\ih \gamma^{2} ).
\end{equation}
Here we note that $\gamma_{00}=0$ since $\gamma$ is a $1$-form.
Since the operator
$\partial$ preserves the filtration and $\delta^{-1}$ raises it by
$1$, the iteration formula (\ref{eq:r-iteration}) has a
unique solution. Moreover
since $\delta^{-1}\tilde{\Omega}$ is at least of degree 3,
the solution $\gamma$ is indeed in $\gm W_{3}\otimes \Lambda^{1}$. \qed
{\bf Remark} The theorem above indicates that a Fedosov
connection is uniquely determined by a torsion free symplectic
connection $\nabla$ and a Weyl curvature $\Omega=\sum_{i}\hbar^{i}\omega_{i}\in
Z^{2}(M)[[\hbar ]]$. For this reason, we will say that the connection $D$
defined above is a Fedosov connection corresponding to the pair
$(\nabla , \Omega )$.
If $D$ is a Fedosov connection, the space of all parallel
sections $W_{D}$ automatically becomes an associative
algebra. Fedosov
proved that $W_{D}$ can be naturally
identified with $C^{\infty}(M)[[\hbar ]]$, and therefore
induces a $*$-product on $C^{\infty}(M)[[\hbar ]]$, which we will
call a {\em Fedosov $*$-product}.
Let $\sigma$ denote the projection from $W_{D}$ to
its center $C^{\infty}(M)[[\hbar ]]$ defined as
$\sigma (a)=a|_{y=0}$.
\begin{thm} (Fedosov)
\label{thm:f2}
$\sigma$ establishes an isomorphism between $W_{D}$ and
$C^{\infty}(M)[[\hbar ]]$ as vector spaces. Therefore,
it induces an associative algebra
structure on $C^{\infty}(M)[[\hbar ]]$, which is a $*$-product.
\end{thm}
\pf To prove that the two vector spaces are
isomorphic,
it suffices to show that for any $a_{0}(x, \hbar)\in C^{\infty}(M)[[\hbar ]]$ there is a unique
section $a\in W_{D}$ such that $\sigma (a)=a_{0}$.
The equation $Da=0$ can be written as
$$\delta a=\partial a +[\ih \gamma , a].$$
Applying the operator $\delta^{-1}$, it follows from the Hodge
decomposition that
\begin{equation}
\label{eq:parallel}
a=a_{0}+\delta^{-1}(\partial a+[\ih \gamma , a]).
\end{equation}
By the iteration method, we see that the equation above
has a unique solution since $\delta^{-1}$ increases the filtration.
The rest of the claim can be easily verified. \qed
\section{Quantum exponential maps}
If $\nabla$ is flat and $\Omega=\omega$, the Fedosov connection
is simply given by $D=-\delta +\partial $. In this case, the
solution to Equation (\ref{eq:parallel}) can be expressed explicitly
as
$$a=\sum_{k=0}^{\infty}\frac{1}{k!} (\partial_{i_{1}}\cdots
\partial_{i_{k}}a_{0})y^{i_{1}}\cdots y^{i_{k}}, $$
which is just the Taylor expansion of $exp_{x}^{*}a_{0}$ at
the origin. So the correspondence from $C^{\infty}(M)[[\hbar ]]$
to $W_{D}$ is indeed the pullback of the ($C^{\infty}$-jet at
the origin of the) usual exponential map.
Thus for a general Fedosov connection, one may consider
the correspondence
$C^{\infty}(M)[[\hbar ]]\lon W_{D}$ as a
``quantum correction" to the exponential map.
In this section, we will make this idea more
precise by introducing the notion of quantum exponential
maps, which gives a simple characterization for Fedosov connections.
As an application, we will realize geometrically
the equivalence between an arbitrary $*$-product and
one from Fedosov method, namely a Fedosov $*$-product.
\begin{defi}
A quantum exponential map is an $\hbar$-linear map
$\rho : \ \ C^{\infty}(M)[[\hbar ]] \lon \gm W $ such that
\begin{enumerate}
\item $\rho (C^{\infty}(M)[[\hbar ]] )$ is a subalgebra of $\gm W$;
\item $\rho (a)|_{y=0}=a, \ \ \forall a\in C^{\infty}(M)[[\hbar ]]$;
\item $\rho (a)=a +\delta^{-1}da , \ \ \forall a\in C^{\infty}(M), \ \ \mbox{mod } W_{2}$;
\item $\rho (a)$ can be expressed as a formal power series
in $y$ and $\hbar$, with coefficients being derivatives of $a$.
\end{enumerate}
\end{defi}
Given a quantum exponential map $\rho$, the Condition (ii)
implies that $\rho$ establishes an isomorphism
between $C^{\infty}(M)[[\hbar ]]$
and its image as vector spaces. Therefore,
$C^{\infty}(M)[[\hbar ]]$
becomes an associative algebra because of the first condition.
It is simple to see that the third condition implies that
this is indeed a deformation of
the symplectic structure, and the last one implies that it is
a $*$-product.
Clearly, for any Fedosov connection $D$, the
map from $C^{\infty}(M)[[\hbar ]]$ to $W_{D}$,
as constructed by Fedosov (see Theorem \ref{thm:f2}),
satisfies all the properties
of a quantum exponential map. So quantum
exponential map always exists, and one may
consider Fedosov construction as a
way of constructing a quantum exponential map.
In what follows, we will show that the converse
is also true. That is,
\begin{thm}
\label{thm:exponential}
Quantum exponential maps are equivalent to Fedosov connections.
\end{thm}
Before proving this theorem, we
start with investigating the following closely related question:
what kind of subalgebras of $\gm W$
arises from a Fedosov connection?
\begin{pro}
\label{pro:abelian}
Suppose that $\cala\subseteq \gm W$ is a subalgebra satisfying
the following conditions:
\begin{enumerate}
\item ``completeness"- for any $x_{0}\in M$, and any $a(y, \hbar )\in W_{x_{0}}$,
there is an element $\tilde{a}(x, y, \hbar )\in \cala $ such that
$\tilde{a}(x_{0}, y, \hbar )=a(y, \hbar )$.
\item ``uniqueness"- if $\tilde{a}$ and $\tilde{b}
\in \cala$ such that
$\tilde{a}|_{x_{0}}=\tilde{b}|_{x_{o}}$, then $\tilde{a}_{*}|_{x_{0}}
=\tilde{b}_{*}|_{x_{0}}$,
where $\tilde{a}$ and $\tilde{b}$ are considered as maps: $M\lon W$,
and $\tilde{a}_{*}$ and $\tilde{b}_{*}$ refer to their derivatives.
\end{enumerate}
Then, there exists a unique Abelian connection $D$
such that $\cala \subseteq W_{D}$.
\end{pro}
\pf Take a torsion free symplectic connection $\nabla$,
and let $\partial :\gm W\lon \gm W \otimes \Lambda^{1}$
be the corresponding covariant derivative.
For any $x_{0}\in M$, introduce an operator $\rho_{x_{o}}:
W_{x_{0}}\lon W_{x_{0}}\otimes \Lambda^{1}$ by
$$\rho_{x_{0}}(a(y, \hbar ))=(\delta -\partial ) \tilde{a}(x, y, \hbar )|_{x=x_{0}},$$
where $\tilde{a}(x, y, \hbar )\in \cala$ such that
$\tilde{a}(x_{0}, y, \hbar )=a(y, \hbar ) $.
By assumption, the map $\rho_{x_{0}}$ is well-defined and is in
fact a derivation of
the algebra $W_{x_{0}}$.
Therefore, there is a unique element $\gamma_{x_{0}}\in W_{x_{0}}\otimes
\Lambda^{1}$ with $\gamma_{x_{0}}|_{y=0}=0$ such that
$\rho_{x_{0}}=ad\ih \gamma_{x_{0}}=[\ih \gamma_{x_{0}}, \cdot ]$.
Applying this process pointwisely, we obtain a global
section $\gamma \in W_{1}\otimes \Lambda^{1}$
with $\gamma_{0}\stackrel{def}{=}\gamma|_{y=0}=0$.
Let $D=-\delta +\partial +[\ih \gamma , \cdot ]$.
Then, $D$ is a connection on the Weyl bundle $W$ and
satisfies the condition: $D \tilde{a}(x, y, \hbar )=0$
for all $\tilde{a}(x, y, \hbar )\in \cala$.
As in Section 2, let $\Omega =\omega -R+\delta \gamma -\partial \gamma -\ih \gamma^{2}$
denote the Weyl
curvature of the connection $D$. It thus follows that
$$D^{2} \tilde{a}=[\ih \Omega , \tilde{a} ]=0,
\ \ \ \forall \tilde{a}\in \cala. $$
Since $\cala$ is complete, $\Omega$
belongs to the center. Therefore, $\Omega \in Z^{2}(M)[[\hbar ]]$.
In other words, $D$ is an Abelian connection. \qed
%{\bf QUESTION} Is there any affine connection on the Weyl bundle
%$W$ of the form $D=-\delta +\partial +[\ih \gamma , \cdot ]$
%with $\gamma \in W_{1}\otimes \Lambda^{1}$. (CHECK WITH ALAN)
The following lemma gives a simple sufficient condition for a subalgebra
$\cala \subseteq \gm W$ being
complete.
\begin{lem}
\label{lem:complete}
Let $\cala \subseteq \gm W$ be a subalgebra with unit.
Suppose that for any $a_{0}(x)\in C^{\infty}(M)$, there is
$\tilde{a}\in \cala $ such that
$$\tilde{a}=a_{0}+\delta^{-1}da_{0}, \ \ \ \ \mbox{mod } W_{2}. $$
Then, $\cala$ is complete.
\end{lem}
\pf We will prove, by induction,
the following statement: $\forall x_{0}\in M \ \mbox{ and }
a(y, \hbar )\in W_{k}|_{x_{0}}$, there is an
element $\tilde{a}\in \cala$
such that $\tilde{a}(x_{0}, y, \hbar )-a(y, \hbar ) \in W_{k+1}|_{x_{0}} $.
The conclusion will then follows immediately from an
iteration argument.
By assumption, the statement holds for both $k=0$ and $k=1$.
Assume that $a(y, \hbar )= \hbar^{j}y^{i_{1}}\cdots y^{i_{p}} $
with $2j+p=k$.
Now
\be
&&a(y, \hbar )-\hbar^{j}y^{i_{1}}*\cdots *y^{i_{p}}\\
&=&\hbar^{j} (y^{i_{1}}\cdots y^{i_{p}}-y^{i_{1}}*\cdots *y^{i_{p}}) \\
&=& \sum_{2i +s=p} \hbar^{i+j}C_{i+j, j_{1} \cdots j_{s}}y^{j_{1}}\cdots
y^{j_{s}},
\ee
where $i\geq 1$, and therefore $s=p-2i=k-2j-2i<k$.
Applying the induction assumption for each term $y^{j_{1}}\cdots
y^{j_{s}}$
in the above summation,
we conclude that there is an element $\tilde{b}\in \cala$ such that
$$a(y, \hbar )-\hbar^{j}y^{i_{1}}*\cdots *y^{i_{p}}=\tilde{b}(x_{0}, y, \hbar ), \ \ \ \mbox{mod } W_{k+1} .$$
On the other hand, for each $y^{i_{l}}$, there
is $\tilde{a}^{i_{l}} \in \cala$ such that $\tilde{a}^{i_{l}}|_{x_{0}}=y^{i_{l}}$,
$\mbox{mod } W_{2}$, for $1\leq l\leq p$.
It is then clear that
$$ \hbar^{j}y^{i_{1}}*\cdots *y^{i_{p}}=\hbar^{j}\tilde{a}^{i_{1}}*\cdots
*\tilde{a}^{i_{p}} |_{x_{0}}, \ \ \mbox{mod } W_{k+1}.$$
Thus, we have
$$a(y, \hbar )=\hbar^{j}\tilde{a}^{i_{1}}*\cdots
*\tilde{a}^{i_{p}} |_{x_{0}} +\tilde{b}(x_{0}, y, \hbar ), \ \ \mbox{mod } W_{k+1}.$$
This concludes the proof.
\qed
Now we are ready to formulate the following
result, which gives a criterion characterizing
a Fedosov algebra.
\begin{thm}
\label{thm:fedosov-algebra}
Let $\cala$ be a subalgebra of $\gm W$ satisfying:
\begin{enumerate}
\item $\forall a_{0}\in C^{\infty}(M)$, there is an element
$\tilde{a}\in \cala $ such that
$$\tilde{a}=a_{0}+\delta^{-1}da_{0}, \ \ \ \ \mbox{mod } W_{2} ;$$
\item if $a$ and $b \in \cala$ such that
$a|_{x_{0}}=b|_{x_{o}}$, then $a_{*}|_{x_{0}}=b_{*}|_{x_{0}}$,
where $a$ and $b$ are considered as maps: $M\lon W$, and
$a_{*}$ and $b_{*}$ refer to their derivatives.
\end{enumerate}
Then, $\cala$ is a Fedosov algebra, i.e. $\cala$ arises
from a Fedosov connection.
\end{thm}
\pf According to Lemma \ref{lem:complete},
$\cala$ is complete. Thus by Proposition
\ref{pro:abelian}, there is an Abelian connection
$D$ such that $\cala \subseteq W_{D}$.
\begin{lem}
This Abelian connection $D$ is in fact a Fedosov connection.
\end{lem}
\pf By assumption, for any $a_{0}\in C^{\infty}(M)$,
there is $\tilde{a}\in \cala $ such that
$$\tilde{a}=a_{0}+\delta^{-1}da_{0}, \ \ \ \ \mbox{mod } W_{2}. $$
It follows from $D\tilde{a}=0$ that
$$(\delta -\partial )\tilde{a}=[\ih \gamma , \tilde{a} ]. $$
The degree zero term of the LHS is easily seen to be zero,
while the degree zero term of the RHS is $\{\gamma_{1}, \delta^{-1}da_{0}\}$,
where $\gamma_{1}$ is the degree one term of $\gamma$.
It thus follows that $\{\gamma_{1}, \delta^{-1}da_{0}\}=0$.
Since $a_{0}$ is arbitrary, $\gamma_{1}$ must be
constant with respect to $y$. However it is linear
in $y$, it has to be identically zero.
This implies that $\gamma \in W_{2}\otimes \Lambda^{1}$.
Denote by $\gamma_{2}$ the degree 2 term of $\gamma$, and assume
that $\gamma_{2}=\sum r_{ij, k}(x)y^{i}y^{j}dx^{k}$.
Note that this is the most general form since
$\gamma_{0}=0$.
Also, we may always assume that $r_{ij, k}=r_{ji, k}$.
Since $D$ is Abelian, its curvature $\Omega \in Z^{2} (M)[[\hbar ]] $.
Assume that
\begin{equation}
\label{eq:curvature2}
\Omega =\sum_{i=0}^{\infty}\hbar^{i}\omega_{i}
=\omega_{0}+\hbar \omega_{1}+\hbar^{2} \omega_{2}
+\cdots .
\end{equation}
On the other hand, according to Equation (\ref{eq:curvature}), we have
\begin{equation}
\label{eq:curvature3}
\Omega = \omega -R+\delta \gamma -\partial \gamma -\ih \gamma^{2}.
\end{equation}
Comparing the degree zero terms of Equations
(\ref{eq:curvature2}) and
(\ref{eq:curvature3}), it follows immediately
that $\omega_{0}=\omega $.
Now the terms $R$, $\partial \gamma $ and
$\ih \gamma^{2}$ are all of degree not less than $2$, so the
only degree $1$ term in Equation (\ref{eq:curvature3})
would be $\delta \gamma_{2}$.
Hence $\delta \gamma_{2}=0$. On the other hand,
a simple calculation yields that
\be
\delta \gamma_{2}&=&\sum ( r_{ij, k}(x)y^{j}dx^{i}\wedge dx^{k}
+r_{ij, k}(x)y^{i}dx^{j}\wedge dx^{k} )\\
&=&\sum r_{ij, k}(x)y^{j}dx^{i}\wedge dx^{k}
+\sum r_{ji, k}(x)y^{j}dx^{i}\wedge dx^{k}\\
&=&2\sum r_{ij, k}(x)y^{j}dx^{i}\wedge dx^{k} .
\ee
It thus follows that for any $i\neq k$,
$$ \sum_{j} r_{ij, k} (x) y^{j}=\sum_{j}r_{kj, i} (x) y^{j} . $$
That is, $ r_{ij, k}(x) = r_{kj, i} (x) $. Hence, $ r_{ij, k}$ is
completely symmetric with respect to $i, \ j, \ k$.
Let $\gm_{ijk}'=\gm_{ijk}+2r_{ij, k}$. Then, $\gm_{ijk}'$
defines a torsion free symplectic connection with induced
differential $\partial'=d+[\ih \gm ', \cdot ]$,
where $\gm '=\half \gm_{ijk}'y^{i}y^{j}dx^{k}$.
It is easy to see that $\partial '=\partial +[\ih \gamma_{2}, \cdot ]$.
Let ${\gamma} '=\gamma -\gamma_{2}$.
Then ${\gamma} '\in W_{3}\otimes \Lambda^{1}$, and
$D=-\delta +\partial ' +[\ih \gamma ' , \cdot ]$. This
shows that $D$ is indeed a Fedosov connection.
\begin{lem}
For any $\tilde{a}\in W_{D}$, there is $\tilde{a}_{0}\in \cala$ and
$\tilde{b}\in W_{D}$ such that
$$\tilde{a}=\tilde{a}_{0}+\hbar \tilde{b}. $$
\end{lem}
\pf Let $a=\tilde{a}|_{y=0} =a_{0}(x)+\hbar a_{1}(x)+\cdots \in
C^{\infty}(M)[[\hbar ]]$.
Take $\tilde{a}_{0}\in \cala $ such that
$\tilde{a}_{0}=a_{0}+\delta^{-1} da_{0} \ \ \ \mbox{mod } W_{2}$,
which is always possible by assumption. Thus,
$\tilde{a}_{0}|_{y=0}=a_{0}(x)+O (\hbar )$ and then
$(\tilde{a}-\tilde{a}_{0})|_{y=0} =O(\hbar )$.
However, we know that $\tilde{a}-\tilde{a}_{0} \in W_{D}$ since
$\tilde{a}_{0} \in \cala
\subseteq W_{D}$. This implies that $\tilde{a}-\tilde{a}_{0} =\hbar \tilde{b}$
for some $\tilde{b}\in W_{D}$.
This concludes the proof of the lemma. \qed
By using the lemma above repeatedly, one immediately
obtains the other inclusion, i.e.,
$W_{D}\subseteq \cala$. This concludes our proof of theorem
\ref{thm:fedosov-algebra}. \qed
Clearly, Theorem \ref{thm:exponential} is an immediate consequence of
Theorem \ref{thm:fedosov-algebra}. As another application, we will give a
geometric constructive proof for the following (see \cite{BCG} \cite{Deligne}
\cite{NT1}):
\begin{thm}
\label{thm:equivalent}
Any $*$-product is equivalent to a Fedosov $*$-product.
\end{thm}
\pf As in Section 2, let $P=TM$ be the regular Poisson manifold
equipped with the fiberwise linear symplectic structures.
Let $\calp=\gm (\cup_{m\in M}
\gm_{0}^{\infty}(T_{m}M, \reals ))$, where $\gm_{0}^{\infty}(T_{m}M, \reals )$
denotes the set of $\infty$-jets at $0$ of real valued functions
on $T_{m}M$.
Thus, $\calp$ is a Poisson algebra with a naturally induced
Poisson structure.
\begin{lem}
Any $*$-product on $C^{\infty}(M)[[\hbar ]]$ induces a
$*$-product on the Poisson manifold $P=TM$ so that
there is an algebra embedding $\rho :
C^{\infty}(M)[[\hbar ]]\lon \calp [[\hbar ]]$.
\end{lem}
\pf Take any torsion-free symplectic connection $\nabla$, and for
any fixed $m\in M$,
let $Exp_{m}$ be the formal symplectic exponential
map introduced by Emmrich and Weinstein \cite{EW}.
Then, $(Exp_{m})^{*}: \ C^{\infty}(M)\lon \gm_{0}^{\infty}(T_{m}M, \reals )$
is a Poisson algebra morphism, which in fact maps
$jet_{m}^{\infty}C^{\infty}(M)$ to $\gm_{0}^{\infty}(T_{m}M, \reals )$
isomorphically. Therefore, any $*$-product on $C^{\infty}(M)[[\hbar ]]$
induces a $*$-product
on $\gm_{0}^{\infty}(T_{m}M, \reals ) [[\hbar ]]$, hence on $C^{\infty}(T_{m }M) [[\hbar ]]$.
Thus, we obtain a \preferred $*$-product (see Appendix B for the definition),
denoted as $\tilde{*}_{\hbar}$,
on the Poisson manifold $P$,
and $Exp^{*}$ is clearly an embedding of the algebra. \qed
According to Proposition \ref{prop:main}, a \preferred $*$-product
on $P$ is essentially unique.
Hence, there exists an equivalence
operator: $T_{\hbar}=1+\hbar T_{1}+\hbar^{2}T_{2}+\cdots $, with
$T_{i}$ being leafwise differential operators, between
$(\calp [[\hbar ]], \tilde{*}_{\hbar})$ and the
standard Weyl quantization $\gm W$.
Let $\cala =T_{\hbar} Exp^{*}(C^{\infty}(M)[[\hbar ]])\subset \gm W$.
Then $\cala $ is a subalgebra of $\gm W$. It is simple
to see that $\cala$ satisfies all the conditions in Theorem
\ref{thm:fedosov-algebra}, so it is a Fedosov algebra. This concludes
the proof of the theorem. \qed
%For any $a_{0}(x)\in C^{\infty}(M)$, $Exp^{*}a_{0}=a_{0}+\delta^{-1}da_{0}
%\ \ \\mbox{Mod} W_{2}$. It is easy to see
%that $\tilde{a}=(T_{\hbar} Exp^{*})a_{0}
%=a_{0}+\delta^{-1}da_{0} \ \ \\mbox{Mod} W_{2}$.
%Assume that $\tilde{a} =(T_{\hbar} Exp^{*})a$ such that
%$\tilde{a}(x_{0}, y, \hbar )=0$. Then $Exp^{*}_{x_{0}}a=0$,
%which implies that $jet^{\infty}_{x_{0}}a=0$.
%It is thus clear that $\tilde{a}_{*}|_{x_{0}}=0$.
%Therefore, $\cala =W_{D}$ for some Fedosov connection
%and the map $T_{\hbar}\smalcirc Exp^{*}$ establishes
%the equivalence between the given $*$-product and
%the Fedosov $*$-product. \qed
Since every $*$-product is equivalent to a Fedosov $*$-product,
its {\em characteristic class}, as defined by Nest-Tsygan \cite{NT1},
is the class in $H^{2}(M)[[\hbar ]]$
of the Weyl curvature of its equivalent Fedosov $*$-product.
This is well-defined since the Weyl curvatures of equivalent Fedosov $*$-products
are cohomologous (see \cite{F2}). In fact, Fedosov showed that two
Fedosov $*$-products are equivalent iff their Weyl curvatures are cohomologous \cite{F2}.
Thus (see \cite{NT1}),
\begin{thm}
\label{thm:equivalent1}
Two $*$-products are equivalent iff they have the same characteristic class.
\end{thm}
\section{Vey-$*$ products}
Let $(M, \omega )$ be a symplectic manifold, and $\nabla $
a torsion free symplectic connection with covariant derivative
$\partial$. Define
$\partial u=du$, $\partial^{2}u=\partial (\partial u)$,
$\partial^{k}u= \partial (\partial^{k-1} u)$ and so on.
It is simple to see that $\partial^{k}u$ is a
symmetric contravariant $k$-tensor.
Let $\pi$ be the Poisson bivector field on $M$.
By $\pi^{k}$, we denote the $2k$-tensor:
$$(\pi^{k})^{i_{1}\cdots i_{k}, j_{1}\cdots j_{k}}=
\pi^{i_{1} j_{1}}\pi^{ i_{2} j_{2}}\cdots \pi^{ i_{k} j_{k}} . $$
Set
$$P^{k}_{\nabla }(u, v)=<\pi^{k}, \partial^{k}u\otimes \partial^{k}v >
=\pi^{i_{1} j_{1}}\cdots \pi^{ i_{k} j_{k}}(\partial_{i_{1}}\cdots \partial_{i_{k}}u)
(\partial_{j_{1}}\cdots \partial_{j_{k}}v). $$
In particular, $P^{0}_{\nabla }(u, v)= uv$ and $P^{1}_{\nabla }(u, v)
=\{u, v\}$.
\begin{defi} \cite{BFFLS}
A Vey $*$-product is a star product on $C^{\infty}(M)[[\hbar ]]$
such that
\begin{equation}
u*_{\hbar}v=\sum \frac{1}{k!}(-\frac{i\hbar}{2})^{k} Q_{k}(u, v),
\end{equation}
where $Q_{k}$ is a bidifferential operator
of maximum order $k$ in each argument
and its principal symbol coincides with that
of $P_{\nabla }^{k}(u, v)$.
\end{defi}
The main result of the section is
\begin{thm}
\label{thm:vey}
Any Fedosov $*$-product is a Vey $*$-product.
%The principal terms are determined by the symplectic
%connection $\nabla $ corresponding to $\partial$ in the
Moreover, if $D=-\delta +\partial +[\ih \gamma , \cdot ]$
as in Theorem \ref{label:f1}, then
%where $\gamma \in W_{3}\otimes \Lambda^{1}$.
%Moreover,
\begin{equation}
Q_{2}(u, v)=P^{2}_{\nabla}(u, v)+C_{2}(u, v),
\end{equation}
where $C_{2}$ is a bidifferential operator of maximum order $1$ in
each argument.
\end{thm}
Given a Fedosov connection $D=-\delta +\partial +[\ih \gamma , \cdot ]$,
any element
\begin{equation}
\label{eq:a}
a(x, y, \hbar )=\sum \hbar^{i}a_{i, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}}
\end{equation}
in $W_{D}$
is determined by the iteration
formula (\ref{eq:parallel}). Assume that $a_{0}(x)=\sigma (a)=a(x, 0, \hbar )\in C^{\infty}(M)$.
It is clear that each coefficient in Equation (\ref{eq:a}) can be expressed
as $a_{i, j_{1} \cdots j_{k}}(x)=\cald_{i, j_{1} \cdots j_{k}}a_{0} $
for certain differential operator $\cald_{i, j_{1} \cdots j_{k}}$.
We will say that the term $\hbar^{i}a_{i, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}}$ is of degree $(i+k, s)$ if
$\cald_{i, j_{1} \cdots j_{k}}$ is a differential operator of
degree $s$.
\begin{pro}
\label{pro:order}
Under the same hypothesis as above, assume that
$$a(x, y, \hbar )=\sum \hbar^{i}a_{i, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}} \in W_{D}. $$
Then
\begin{enumerate}
\item $a_{i, j_{1} \cdots j_{k}}(x) =\cald_{i, j_{1} \cdots j_{k}}a_{0}$,
where $\cald_{i, j_{1} \cdots j_{k}} $ is a differential operator
of degree not greater than $i+k$.
\item $$a(x, y, 0)=\sum_{k=0}^{\infty}\frac{1}{k!} (\partial_{i_{1}}\cdots
\partial_{i_{k}}a_{0})y^{i_{1}}\cdots y^{i_{k}}+
H, $$
where all terms in the remainder $H$ are of degree $(l, k)$ with $l<k$.
\item For any $j$, the order of the differential operator
$D_{1, j} $ is not greater than $1$.
\end{enumerate}
\end{pro}
\pf Since $a(x, y, \hbar ) $ is
generated by iteration formula (\ref{eq:parallel}), we prove (i) by induction.
% Assume that the claim holds up to the $p$-th step of iteration.
For this purpose, we need to analyses
the effect of the operators $\delta^{-1}\partial $ and
$\delta^{-1}[\ih \gamma , \cdot ]$ on any term of degree
$(i+k, s)$. It is obvious that
$\delta^{-1}\partial $ maps terms of degree $(i+k, s)$
to those of degree $(i+k+1, s+1)$.
On the other hand,
it is not difficult to check that
$\delta^{-1}[\ih \gamma , \cdot ]$ maps terms of degree $(i+k, s)$
to those of degree $(m, s)$ with $m\geq i+k$.
Claim (i) thus follows immediately.
To prove (ii), we need to analyses the $\hbar^{0}$-terms
produced from $\delta^{-1}[\ih \gamma , \cdot ]$
in iteration formula (\ref{eq:parallel}).
The only possibility would be $ \delta^{-1}(-i \hbar \{\ih \gamma_{0}, \tilde{a}\})$, where
$\gamma_{0}=\gamma (x, y, 0, dx )$ and
$\tilde{a}= a_{0, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}}$ is assumed of degree $(l, k)$.
By Part (i), we know that $l\leq k$.
Now $ \delta^{-1}(-i \hbar \{\ih \gamma_{0}, \tilde{a}\} )=
\delta^{-1}\{ \gamma_{0}, \tilde{a}\}$.
Since $\gamma_{0}$ is at least cubic in $y$ (deg $\gamma \geq 3$),
the latter is of degree $(l, k')$ with $k'$ being at least
$k+2$.
Therefore, all terms of degree $(k, k)$ in $a(x, y, 0)$
come solely from the iteration of
the operator $\delta^{-1}\partial$ on elements of the same
form.
The conclusion thus follows immediately.
To prove (iii), we will concentrate on those terms of
degree $(l, 2)$:
\begin{equation} \label{eq:form1}
\sum \hbar a_{1, j}(x)y^{j}.
\end{equation}
Since the operator $\delta^{-1}\partial $ maps any $(t, s)$-term
into that of degree $(t+1, s+1)$, and
$\delta^{-1}\partial $ does not produce any $\hbar$, the
only possible way to obtain such a term is when
$\delta^{-1}\partial $ acts on those of the form
$\hbar a_{1}(x)$. This however will never happen since
we already assume that $a(x, 0, \hbar )=a_{0}(x)$ is independent of $\hbar$.
As for the operator
$\delta^{-1}[\ih \gamma , \cdot ]$, the only possible way of generating
a term of the form (\ref{eq:form1}) is
$\delta^{-1}(-i\hbar \{\ih \gamma' , a'\})=
\delta^{-1}\{\gamma ' , a'\}$,
where $\gamma'$ is the term in $\gamma$ having the form:
$\hbar \gamma_{i, j}(x)y^{i}dx^{j}$ and $a'$ is the term in $a(x, y, \hbar )$
of the form:
$a'=a_{0, m}(x)y^{m} $.
Thus, $\delta^{-1}\{\gamma ' , a'\}=\hbar \gamma_{ij}(x) a_{0, m}(x)
\omega^{im}(x)y^{j} $ is of degree $(1, 2)$.
This concludes the proof. \qed
An immediate consequence of Proposition \ref{pro:order}
is the following:
\begin{cor}
Let $a(x, y, \hbar )\in W_{D}$, and let $x_{0}\in M$ be any point. Then,
$a(x_{0}, y, \hbar )=0$, $\forall y\in T_{x_{0}}M$ iff
$jet_{x_{0}}^{\infty} a(x, \hbar )=0$, where
$a(x, \hbar )=a(x, 0, \hbar )=\sigma (a( x, y, \hbar ) )$.
\end{cor}
\pf One direction follows directly from Part (1) of Proposition \ref{pro:order}.
To prove the other direction,
let $a_{0}(x)=a(x, 0 ,0)$ and assume that $a(x, \hbar )=
a_{0}(x)+\hbar a_{1}(x)+\hbar^{2}a_{2}(x) +\cdots $.
We shall prove,
as the first step, that $jet_{x_{0}}^{\infty} a_{0}(x)=0$.
Clearly, $a_{0}(x_{0})=0$. Assume that all derivatives
of $a_{0}$ up to the $k$-th order vanish at $x_{0}$.
According to Part (ii) of Proposition \ref{pro:order},
$$a(x_{0}, y, 0)=
\sum_{k=0}^{\infty}\frac{1}{k!} (\partial_{i_{1}}\cdots
\partial_{i_{k}}a_{0})y^{i_{1}}\cdots y^{i_{k}}+
H,$$
where all terms in $H$ are of degree $(l, k)$ with $l<k$.
Since the coefficient of $y^{i_{1}}\cdots y^{i_{k}}$
is zero, it follows that
$\frac{1}{k!} (\partial_{i_{1}}\cdots
\partial_{i_{k}}a_{0}) +(\cald a_{0} )(x_{0}) =0$, for
some differential operator $\cald$ of degree less than $k$.
By using the induction assumption, we deduce that
$(\partial_{i_{1}}\cdots \partial_{i_{k}}) a_{0}(x_{0})=0$.
This proves that $jet_{x_{0}}^{\infty} a_{0}(x)=0$.
Let $a_{0}(x, y, \hbar)\in W_{D}$ be the parallel section
corresponding to
$a_{0}(x)$. Then $a_{0}(x_{0}, y, \hbar)=0, \ \forall y\in
T_{x_{0}}M$.
By considering the element $\frac{1}{\hbar}[a(x, y, \hbar )- a_{0}(x_{0}, y, \hbar)]\in
W_{D}$, one yields that $jet^{\infty}_{x_{0}}a_{1}(x)=0$.
The conclusion thus follows by using this argument repeatedly. \qed
%\begin{thm}
%Any Fedosov $*$-product is a Vey $*$-product.
%The principal terms are determined by the symplectic
%connection $\nabla$ corresponding to $\partial$ in the
%Fedosov connection $D=-\delta +\partial +[\ih \gamma , \cdot ]$
%with $\gamma \in W_{3}\otimes \Lambda^{1}$.
%Moreover,
%
%\begin{equation}
%Q_{2}(u, v)=P_{2}(u, v)+C_{2}(u, v)
%\end{equation}
%where $C_{2}$ is a bidifferential operator of maximum order $1$ in
%each argument.
%\end{thm}
{\bf Proof of Theorem \ref{thm:vey}} Let $a_{0}(x) $ and $b_{0}(x)$ be
any functions on $M$, and $a(x , y, \hbar )$ and $b(x, y, \hbar )$
their corresponding parallel sections in $W_{D}$.
Assume that
$$a(x, y, \hbar )=\sum \hbar^{i}a_{i, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}}, \ \ \ \mbox{and} $$
$$b(x, y, \hbar )=\sum \hbar^{i}b_{i, j_{1} \cdots j_{k}}(x)
y^{j_{1}}\cdots y^{j_{k}}. $$
By the definition of Moyal-Weyl product,
\be
&&a(x, y, \hbar )*b(x, y, \hbar )|_{y=0}\\
&=&\sum (-\frac{i}{2})^{p} \frac{1}{p!} \hbar^{k+l+p}
a_{k, i_{1}\cdots i_{p}}(x)b_{l, j_{1}\cdots j_{p}} (x)\pi^{t_{1}s_{1}}\cdots
\pi^{t_{p}s_{p}}\frac{\partial (y^{i_{1}}\cdots y^{i_{p}} )}{\partial
y^{t_{1}} \cdots \partial y^{t_{p}} }\frac{\partial (y^{j_{1}}\cdots y^{j_{p}} )}{\partial
y^{s_{1}} \cdots \partial y^{s_{p}} } \\
&=&\sum_{n}\sum_{k+l+p=n} (-\frac{i}{2})^{p} \frac{1}{p!} \hbar^{k+l+p}
a_{k, i_{1}\cdots i_{p}}(x)b_{l, j_{1}\cdots j_{p}}(x)\pi^{t_{1}s_{1}}\cdots
\pi^{t_{p}s_{p}}\frac{\partial (y^{i_{1}}\cdots y^{i_{p}} )}{\partial
y^{t_{1}} \cdots \partial y^{t_{p}} }\frac{\partial (y^{j_{1}}\cdots y^{j_{p}} )}{\partial
y^{s_{1}} \cdots \partial y^{s_{p}} } \\
&=&\sum_{n}\sum_{k+l+p=n} (-\frac{i}{2})^{p} \frac{1}{p!} \hbar^{k+l+p}
(\cald_{k, i_{1}\cdots i_{p}} a_{0})( \cald_{l, j_{1}\cdots j_{p}} b_{0})
\pi^{t_{1}s_{1}}\cdots
\pi^{t_{p}s_{p}}\frac{\partial (y^{i_{1}}\cdots y^{i_{p}} )}{\partial
y^{t_{1}} \cdots \partial y^{t_{p}} }\frac{\partial (y^{j_{1}}\cdots y^{j_{p}} )}{\partial
y^{s_{1}} \cdots \partial y^{s_{p}} }
\ee
According to Proposition \ref{pro:order},
the order of the differential operator
$\cald_{k, i_{1}\cdots i_{p}}$ is not greater than $k+p$, which
is less than or equal to $n$, similarly for the order of
$\cald_{l, j_{1}\cdots j_{p}} $.
If $l\neq 0$, then $k+p=n-l<n$. The order of $\cald_{k, i_{1}\cdots i_{p}}$
is less than $n$. Similarly if $k\neq 0$, the order of
$\cald_{l, j_{1}\cdots j_{p}} $ is less than $n$.
So in order to have the maximum order in both arguments,
it is necessary that $k=l=0$, and in this case $p=n$.
Using Part (ii) of Proposition \ref{pro:order}, it
is simple to see that
the principal part is
$$ \sum (-\frac{i\hbar }{2})^{n} \frac{1}{n!} (\partial_{i_{1}}\cdots
\partial_{i_{n}}a_{0})(\partial_{j_{1}}\cdots \partial_{j_{n}}b_{0})
\pi^{i_{1}j_{1}}\cdots \pi^{i_{n}j_{n}} .$$
For the $\hbar^{2}$-term, we need $k+l+p=2$.
If $k=l=0$ and $p=2$, we obtain the principal term.
When $k=p=1$ and $l=0$, then
$a_{1, i_{1}}(x)=\cald_{1, i_{1}} a_{0}$
and $b_{0, j_{1}}(x)=\cald_{0, j_{1}}b_{0}$.
According to Proposition \ref{pro:order}, both $\cald_{1, i_{1}}$ and $\cald_{0, j_{1}}$
have degrees not greater than $1$;
similarly for the case that $l=p=1$ and $k=0$.
This concludes the proof. \qed
As an immediate consequence, we obtain the following result, which was first
proved by Lichnerowicz using homological methods \cite{L}.
\begin{cor} (Lichnerowicz)
Any $*$-product on a symplectic manifold is equivalent to a
Vey-$*$-product.
\end{cor}
{\bf Remark} As we know, the equivalence class of a
Fedosov $*$-product is determined by the class
of its Weyl curvature in $H^{2}(M)[[\hbar ]]$,
which is independent of the symplectic connection in the
construction. On the other hand, as we see in Theorem
\ref{thm:vey}, the symplectic connection is indeed completely
reflected in the Fedosov $*$-product itself. In fact
if two Fedosov connections
$D_{i}=-\delta +\partial_{i}+[\ih \gamma_{i}, \cdot ]$
induce an identical $*$-product on $C^{\infty}(M)[[\hbar ]])$,
it is necessary that their symplectic connections
coincide, i.e., $\partial_{1}=\partial_{2}$.
We would like to conjecture that:
{\em Two Fedosov connections
$D_{i}=-\delta +\partial_{i}+[\ih \gamma_{i}, \cdot ]$ induce an
identical $*$-product on $C^{\infty}(M)[[\hbar ]])$,
iff $D_{1}=D_{2}$.}
To prove this, one needs to prove that the Weyl curvatures
of $D_{1}$ and $D_{2}$ are not only cohomologous, but indeed coincide,
or equivalently $W_{D_{1}}=W_{D_{2}}$. In other words. one
needs to decode the Weyl curvature (not only its class in $H^{2}(M)[\hbar ]$)
from a Fedosov $*$-product directly.
We end this section with the following inverse question:
\begin{quote}
{\bf Question} Is every Vey $*$-product necessary a Fedosov
$*$-product?
\end{quote}
\section{Quantization of lagrangian submanifolds}
Lagrangian submanifolds play a fundamental role in the study
of symplectic manifolds. In fact, according to the ``symplectic creed",
everything can be thought of as a lagrangian submanifold \cite{Weinstein:1977}.
It is very natural to ask: what should lagrangian submanifolds
correspond to under a deformation quantization?
This is what we aim to answer in this section.
\begin{lem}
\label{lem:lag}
Let $*_{\hbar}$ be the Moyal-Weyl product for the symplectic
vector space $V\cong \reals^{2n}$ with the standard
symplectic structure. Suppose that $L\subset V$ is a lagrangian
subspace. If both $f$ and $g\in C^{\infty}(M) $ vanish on $L$,
so does $f*_{\hbar}g$.
\end{lem}
\pf Note that the Moyal-Weyl formula is independent of
the choice of linear coordinates.
Let us choose a lagrangian subspace $K$ supplementary to
$L$, and linear coordinates $(q_{1}, \cdots , q_{n})\in L$,
$(p_{1}, \cdots , p_{n})\in K$ such that
$\omega (q_{i}, p_{j})=\delta_{ij}$.
Then the claim follows directly from Equation (\ref{eq:mw}). \qed
Let $L$ be a lagrangian submanifold. For any $x\in L$,
by $W_{x}^{L}$ we denote the subspace of the Weyl algebra
$W_{x}$ consisting of all elements which vanish
when being restricted to $T_{x}L$. According to Lemma \ref{lem:lag},
$W_{x}^{L}$ is a subalgebra.
Let $W^{L}=\cup_{x\in L}W_{x}^{L}$. Then $W^{L}$ is
a subbundle of the Weyl bundle. Similarly, we can
define $W^{L}_{1}$, $W^{L}_{2}$, etc. according to the
natural filtration in $W$.
We say an element $a\in W_{p}\otimes \Lambda^{q}$ is in the
subspace $(W_{p}\otimes \Lambda^{q})_{L}$ if for any
$v_{1}, \cdots , v_{q}\in T_{x}L$, $a(v_{1}, \cdots , v_{q})\in
W_{p}^{L}$. By $(W\otimes \Lambda )_{L}$, we denote
the direct sum $\oplus_{p, q}(W_{p}\otimes \Lambda^{q})_{L}$.
It is clear that both $\delta$ and $\delta^{-1}$
preserve the space $(W\otimes \Lambda )_{L}$.
More precisely, $\delta $ maps $(W_{p}\otimes \Lambda^{q})_{L}$
into $(W_{p-1}\otimes \Lambda^{q+1})_{L}$, while $\delta^{-1}$
maps $(W_{p}\otimes \Lambda^{q})_{L}$
into $(W_{p+1 }\otimes \Lambda^{q-1})_{L}$.
Let $\nabla $ be a torsion-free symplectic connection on $M$ such that $L$
is totally geodesic. Then its induced differential
$\partial $ maps $(W\otimes \Lambda^{q})_{L}$ into
$(W\otimes \Lambda^{q+1})_{L}$.
\begin{pro}
\label{pro:lag}
As in Theorem \ref{label:f1}, if in addition, $L$ is totally
geodesic with respect to the symplectic connection $\nabla$,
and is lagrangian with respect to the Weyl curvature $\Omega$, i.e.,
$i^{*}\Omega =0$, where $i: L\lon M$ is the embedding,
then $\gamma $ belongs to $(W_{3}\otimes \Lambda^{1})_{L}$.
Therefore, the corresponding Fedosov connection
$D$ preserves $(W\otimes \Lambda )_{L}$.
\end{pro}
%Suppose that $\nabla $
% is a symplectic connection on $M$ such that $L$
%is totally geodesic. Let $D=-\delta +\partial +[\ih \gamma , \cdot ]$
%be its corresponding Fedosov connection with
%Weyl curvature $\Omega =\omega +\hbar \omega_{1}+\cdots \hbar^{i}\omega_{i} +\cdots $.
%Assume that $L$ is lagrangian with respect to $\Omega$, i.e.,
%$i^{*}_{L}\Omega =0$, where $i_{L}: L\lon M$ be
%the embedding. Then, $\gamma $ belongs
%$(W_{3}\otimes \Lambda^{1})_{L}$, and therefore $D$ preserves
%$(W\otimes \Lambda )_{L}$.
%\end{pro}
\pf According to Equation (\ref{eq:r-iteration}), $\gamma $ is determined
by the iteration formula
\begin{equation}
\label{eq:r-iteration1}
\gamma =\delta^{-1} \tilde{\Omega }+\delta^{-1}(\partial \gamma +\ih \gamma^{2} ),
\end{equation}
where $\tilde{\Omega }=\Omega -\omega +R$.
It is easy to see that $R\in (W_{2}\otimes \Lambda^{2})_{L}$
since $L$ is a totally geodesic lagrangian
submanifold. Therefore, it follows that $\tilde{\Omega }\in
(W_{2}\otimes \Lambda^{2})_{L}$.
It is then clear, from the above iteration formula, that $\gamma \in
(W_{3}\otimes \Lambda^{1})_{L}$ since $(W\otimes \Lambda^{*})_{L}$
is preserved under all the operations involved in Equation (\ref{eq:r-iteration1}). \qed
Let $W_{D}^{L}$ denote the subspace of $W_{D}$ consisting of
sections whose restriction to $L$ belong to $W^{L}$.
As an immediate consequence, we have
\begin{cor}
Under the same hypothesis as in Proposition \ref{pro:lag},
$W_{D}^{L}$ is a subalgebra of $W_{D}$.
\end{cor}
By $C_{L}^{\infty}(M)$, we denote the space of smooth functions on $M$ which
vanish on $L$.
\begin{pro}
Under the same hypothesis as in Proposition \ref{pro:lag},
$$ \sigma ( W^{L}_{D})=C^{\infty}_{L}(M)[[\hbar ]] ,$$
where $\sigma $
is the isomorphism between $W_{D}$ and $C^{\infty}(M)[[\hbar]]$ as introduced in
Section 2.
Therefore, $C^{\infty}_{L}(M)[[\hbar ]] $ is a subalgebra
of the Fedosov $*$-algebra $(C^{\infty}(M)[[\hbar ]], *_{\hbar})$.
\end{pro}
\pf According to Theorem \ref{thm:f2}, if $a\in W_{D}$ with
$\sigma (a)=a_{0}$,
$a$ is determined by the iteration formula
$$a=a_{0}+\delta^{-1}(\partial a+[\ih \gamma , a]). $$
If $a_{0}\in C^{\infty}_{L}(M)[[\hbar]]$, which means that
$a_{0}\in W^{L}$, it follows immediately that
$a\in W^{L}$. This is because $\gamma \in (W_{3}\otimes \Lambda^{1})_{L}$,
and all the operators involved preserve the space $(W\otimes \Lambda )_{L}$.
Conversely, if $a\in W_{D}^{L}$, it is obvious
that $a_{0}=\sigma (a)\in W^{L}$. \qed
\begin{ex}
If $f: M\lon N$ is a symplectic diffeomorphism, its graph
$G_{f}=\{(x, f(x) )|x\in M\}$ is a lagrangian
submanifold of $S=M\times \bar{N}$. Moreover if $f^{*}\Omega_{N} =\Omega_{M}$,
where $\Omega_{M}$ and $\Omega_{N}$ are Weyl curvatures on $M$ and $N$,
respectively,
$G_{f}$ is lagrangian with respect to
$(\Omega_{M}, \overline{\Omega_{N}} )$.
Let $S$ be equipped with a product symplectic
connection $\nabla \times \tilde{\nabla }$. It is easy
to see that $G_{f}$ is totally geodesic iff $\tilde{\nabla }=f_{*}\nabla $.
In this case, $C^{\infty}_{L}(S)$ is a subalgebra of
the corresponding Fedosov $*$-product $(C^{\infty}(S)[[\hbar ]], *_{\hbar})$.
This implies that $f_{*}$ is an algebra morphism
between Fedosov $*$-algebras $(C^{\infty}(M)[[\hbar]], *_{\hbar})$
and $(C^{\infty}(N)[[\hbar]], *_{\hbar})$.
\end{ex}
The following well known lemma indicates that a totally
geodesic symplectic connection always exists for any given lagrangian
submanifold.
\begin{lem}
Given any lagrangian submanifold $L\subset M$, there always exists
a torsion-free symplectic connection on $M$ such that $L$ is totally
geodesic.
\end{lem}
\pf First, take any torsion-free connection $\nabla $ on $M$
such that $L$ is totally geodesic.
Any other connection can be written
as
\begin{equation}
\tilde{\nabla}_{X}Y=\nabla_{X}Y +S(X, Y) , \ \ \ \forall X, Y\in \calx (M),
\end{equation}
where $S$ is a $(2, 1)$-tensor.
Clearly, $\tilde{\nabla}$ is torsion-free iff
$S$ is symmetric, i.e., $S(X, Y)=S(Y, X)$ for any $X, Y\in \calx (M)$.
$\tilde{\nabla}$ is symplectic iff $\tilde{\nabla}_{X}\omega =0$.
The latter is equivalent to
\begin{equation}
\omega (S(X, Y), Z)-\omega (S(X, Z), Y)=(\nabla_{X}\omega )(Y, Z).
\end{equation}
Let $S$ be the $(2, 1)$-tensor defined by
the equation:
\begin{equation}
\label{eq:s}
\omega (S(X, Y), Z)=\frac{1}{3}[(\nabla_{X}\omega )(Y, Z)+
(\nabla_{Y}\omega )(X, Z)] .
\end{equation}
Clearly, $S(X, Y)$, defined in this way, is symmetric
with respect to $X$ and $Y$. Now
\be
&&\omega (S(X, Y), Z)-\omega (S(X, Z), Y)\\
&=&\frac{1}{3}[(\nabla_{X}\omega )(Y, Z)+ (\nabla_{Y}\omega )(X, Z)]
-\frac{1}{3}[(\nabla_{X}\omega )(Z, Y)+(\nabla_{Z}\omega )(X, Y)]\\
&=&\frac{1}{3}[(\nabla_{X}\omega )(Y, Z)+(\nabla_{Y}\omega )(X, Z)
+(\nabla_{X}\omega )( Y, Z)+(\nabla_{Z}\omega )(Y, X)]\\
&=&(\nabla_{X}\omega ) (Y, Z),
\ee
where the last step follows from the identity:
$$(\nabla_{X}\omega )(Y, Z)+( \nabla_{Y}\omega )(Z, X)+
(\nabla_{Z}\omega )(X, Y)=0.$$
This means that $\tilde{\nabla}$ is a torsion-free symplectic connection.
>From Equation (\ref{eq:s}), it follows that
if $X$, $Y$ and $Z$ are all tangent
to $L$, $\omega (S(X, Y), Z)=0$ since $L$ is totally geodesic with
respect to $\nabla$.
Hence, $S(X, Y)$ is tangent to $L$ whenever $X$, $Y$ are tangent to $L$.
In other words, $L$ is totally geodesic with respect to $\tilde{\nabla}$. \qed
Since a
$*$-product is always equivalent to a
Fedosov $*$-product, as a consequence,
any lagrangian submanifold, under a deformation quantization,
becomes a subalgebra after
some ``quantum correction". More precisely, we have
\begin{thm}
Let $*_{\hbar}$ be a $*$-product on a symplectic manifold
$(M, \omega )$ with characteristic class $[\Omega ]\in H^{2}(M)[[\hbar ]] $.
Suppose that $L$ is a lagrangian submanifold
such that $i^{*}[\Omega] \in H^{2}(L)[[\hbar ]]$ vanishes, where
$i: L\lon M$ is the embedding.
Then there exists an operator: $T_{\hbar}=1+\hbar T_{1}+
\hbar^{2} T_{2}+\cdots $, with $T_{i}$ being
differential operators on $M$, such that
$T_{\hbar}(C^{\infty}_{L}(M)[[\hbar ]]) $ is a
subalgebra of $(C^{\infty}(M)[[\hbar ]], *_{\hbar})$.
\end{thm}
\pf By assumption, $i^* \Omega $ is an exact two-form on $L$, i.e.,
$i^* \Omega =d\theta_{L}$ for some $\theta_{L}\in \Omega^{1}(L)$.
Extending $\theta_{L}$ to a one-form on $M$, we may
assume that $i^* \Omega =d i^{*} \theta$ for some $\theta\in
\Omega^{1}(M)$. Let $\tilde{\Omega}=\Omega -d\theta$, and
take a torsion-free symplectic connection $\nabla $ such that
$L$ is totally geodesic.
Let $\bar{*}_{\hbar}$ be the corresponding
Fedosov $*$-product with Weyl curvature $\tilde{ \Omega }$.
Then $ C^{\infty}_{L}(M)[[\hbar ]] $ is a $\bar{*}_{\hbar}$-subalgebra.
According to Theorem \ref{thm:equivalent1}, $*_{\hbar}$ and $\bar{*}_{\hbar}$ are equivalent
$*$-products. The conclusion thus follows immediately. \qed
{\bf Remark}\ \ (1). The quantum counterparts of lagrangian submanifolds,
according to Lu \cite{Lu}, are left ideals. However, it is not
clear how this can be realized for $*$-algebras in our case.
It seems that a possible candidate would be the space
$ C^{\infty}_{L}(M)[[\hbar ]] $
modified in a certain way.
(2). For a symplectic manifold $M$, a coisotropic submanifold
is a submanifold $C$
such that the space of
functions vanishing on $C$ becomes a Poisson
subalgebra. It is natural to expect that the above result
can be generalized to any coisotropic submanifolds. But we cannot
prove this at the moment because it is not clear if there
always exists a symplectic connection such that $C$ is
totally geodesic for a general coisotropic
submanifold $C$.
\section{Quantum momentum maps}
This section is devoted to the study of deformation quantization of
a symplectic $G$-space. In particular, we will introduce the notion
of quantum momentum maps, which plays the role of a quantum
analogue of the usual momentum maps.
Let $(M, \omega )$ be a symplectic $G$-space with action
$\Phi_{g}: M\lon M$, $\forall g\in G$. A $*$-product on $M$ is called
{\em $G$-\equivariant} if for any $u$ and $v\in C^{\infty}(M)[[\hbar]]$,
\begin{equation}
\Phi^{*}_{g}(u*_{\hbar}v)=(\Phi^{*}_{g}u)*_{\hbar}(\Phi^{*}_{g}v).
\end{equation}
In general, $M$ does not necessary admit a $G$-\equivariant $*$-product.
It is known \cite{BFFLS} \cite{L} that the
existence of such a $*$-product is closely related to the
existence of a $G$-invariant connection on the
manifold. More precisely,
\begin{pro}
\label{pro:natural}
Let $M$ be a symplectic $G$-space. $M$ admits a $G$-\equivariant
\nat $*$-product
iff there exists a $G$-invariant connection on $M$,
where by a \nat $*$-product, we mean a $*$-product:
$$u*_{\hbar }v=\sum_{k} \hbar^{k} C_{k}(u, v),$$
where the $\hbar^2$-term $C_{2}$ is a bidifferential operator of
order $2$ in each argument.
\end{pro}
\pf Assume that there exists a $G$-\equivariant \nat $*$-product:
$$ u*_{\hbar }v=uv-\frac{i\hbar }{2}\{u , v\} + \half (\frac{i\hbar }{2})^{2}Q_{2}(u, v)+\cdots . $$
According to Proposition \ref{pro:nat-split}, there is a unique symplectic connection $\nabla$
such that
$$Q_{2}(u, v)=P_{\nabla}^{2}(u, v)+H(u, v), $$
where $H(u, v)$ is a bidifferential operator of maximum order $1$ in each argument.
It is then clear that $\nabla $ is $G$-invariant.
Conversely, suppose that there exists a $G$-invariant connection
on $M$. Using the standard method, we can always make it into
a $G$-invariant torsion free symplectic connection $\nabla $.
Then, the corresponding Fedosov $*$-product (with symplectic connection
$\nabla $, and Weyl curvature $\omega $) will be $G$-equivariant. \qed
A $G$-invariant connection always exists if $G$ is compact. However, when
$G$ is non-compact, there exist some cases where $G$-invariant connections do not
exist. Various attempts have been made in order to deal with such a situation.
For details, readers can consult \cite{AC1} \cite{AC2} \cite{FS}.
In what follows, nevertheless, we will always assume that $ *_{\hbar}$
is a $G$-\equivariant $*$-product.
Then, the corresponding infinitesimal action
$\xi \lon \hat{\xi}$ defines a
Lie algebra homomorphism from $\frakg$ to the derivation
space $Der C^{\infty}(M) [[\hbar ]]$ of the $*$-algebra.
By $\pol [[\hbar ]]$, we denote the space of formal power series of
$\hbar$ with coefficients in the universal
enveloping algebra $\pol$. Let $\frakg_{\hbar}$ be the deformed
Lie algebra: $\frakg_{\hbar}= \frakg [[\hbar]]$ with the bracket
being defined as
\begin{equation}
[X, Y]_{\hbar}=-i\hbar [X, Y], \ \ \ \forall X, Y\in \frakg [[\hbar]].
\end{equation}
Then $\pol [[\hbar ]]$ can be identified with the universal enveloping algebra
of $\frakg_{\hbar}$, and therefore inherits an associative algebra structure.
% we denote the space of polynomials on $\frakg^*$
%equipped with the $*_{\hbar}$-product induced from
%the universal enveloping algebra $U \frakg $.
\begin{defi}
A quantum momentum map is a homomorphism of associative
algebras:
$$\mu_{\hbar}: \pol [[\hbar ]] \lon C^{\infty}(M) [[\hbar ]], $$
such that for any $\xi \in \frakg$,
\begin{equation}
\label{eq:moment}
\hat{\xi} =\ad {\ih \mu_{\hbar}(\xi )},
\end{equation}
where both sides are considered as derivations on $C^{\infty}(M) [[\hbar ]]$.
\end{defi}
It is obvious, from definition, that
a necessary condition for the existence of a momentum map
is that the derivation $\rho_{\xi}f=\hat{\xi}f$,
$\forall \xi \in \frakg$ be inner.
Let us first assume that this is true.
Thus there is a linear map from $\frakg$ to
$C^{\infty}(M)[[\hbar ]]$, denoted by
$\xi \lon a_{\xi}$, such that
for any $f\in C^{\infty}(M)$ and $\xi \in \frakg$,
\begin{equation}
\label{eq:moment1}
\hat{\xi}f=[\ih a_{\xi }, f].
\end{equation}
Therefore,
\be
\widehat{[\xi , \eta ]}f &=&[\hat{\xi }, \hat{\eta }]f\\
&=& [\ih a_{\xi }, [\ih a_{\eta }, f]]-[\ih a_{\eta }, [\ih a_{\xi }, f]]\\
&=&[[\ih a_{\xi }, \ih a_{\eta }], f].
\ee
On the other hand, by definition,
$$\widehat{[\xi , \eta ]} f =[\ih a_{[\xi ,\eta ]}, f]. $$
Therefore,
$$[a_{[\xi , \eta ]}-\ih [a_{\xi }, a_{\eta } ], f]= 0, \ \ \ \forall f\in
C^{\infty}(M). $$
So $a_{[\xi , \eta ]}-\ih [a_{\xi }, a_{\eta } ]$, as a function in
$C^{\infty}( M)[[\hbar ]]$,
is constant.
Define $\lambda : \wedge^{2}\frakg \lon \complex [[h]]$ by
\begin{equation}
\label{eq:2cocycle}
\lambda (\xi ,\eta )=a_{[\xi , \eta ]}-\ih [a_{\xi }, a_{\eta } ], \ \ \ \forall \xi , \eta \in \frakg .
\end{equation}
\begin{pro}
\begin{enumerate}
\item $\lambda$ is a Lie algebra 2-cocycle;
\item its cohomology class $[\lambda ]\in H^{2}(\frakg , \complex [[\hbar]])
\cong H^{2}(\frakg )\otimes \complex [[\hbar ]] $ is independent
of the choice of the linear map $a_{\xi}$;
\item quantum momentum map exists iff $[\lambda ]=0$.
\end{enumerate}
\end{pro}
\pf Assertions (i)-(ii) are quite obvious, and left for the reader
to check.
For (iii), suppose that a quantum momentum map $\mu_{\hbar}$
exists. Then we may take $a_{\xi}=\mu_{\hbar}({\xi})$
as our linear map. In this case,
\be
\lambda (\xi, \eta )&=&a_{[\xi , \eta ]}-\ih [a_{\xi}, a_{\eta }]\\
&=&\mu_{\ih}{[\xi , \eta ]} -\ih [ \muh \xi , \muh {\eta}]\\
&=&\mu_{\ih}([\xi , \eta ]-\ih [{\xi} ,{\eta}]_{\hbar} ) \\
&=&0.
\ee
Conversely, if $[\lambda ]=0$, by adding a suitable coboundary we can always choose a
linear map $a:\frakg \lon C^{\infty}(M)[[\hbar ]]$
such that Equation (\ref{eq:moment1}) holds and
$a_{[\xi , \eta ]}-\ih [a_{\xi }, a_{\eta } ]=0$.
In other words, $a$ is a Lie algebra
homomorphism from $\frakg_{\hbar}$
to the commutator Lie algebra of $C^{\infty}(M)[[\hbar ]]$.
%where $\frakg_{\hbar}$ is the deformed Lie algebra defined
%on the same space $\frakg$ with Lie bracket
%$[\xi ,\eta ]_{\hbar}=-i \hbar [\xi ,\eta ], \forall \xi , \eta \in \frakg$.
Therefore, it extends to an associative algebra
morphism:
$$\muh : \ U\frakg_{\hbar}\lon (C^{\infty}(M)[[\hbar ]] ,
*_{\hbar}). $$
This concludes the proof. \qed
According to Theorem \ref{thm:derivation}, derivations are automatically inner if $H^{1}(M)=0$.
Thus we have
\begin{thm}
\label{eq:mom-exist}
There exists a quantum momentum
map if $H^{1}(M)=0$ and $H^{2}(\frakg )=0$. In particular,
a quantum momentum map exists if $M$ is simply connected
and $\frakg$ is semi-simple.
\end{thm}
We note that the above condition is exactly the same sufficient
condition for the existence of a classical momentum map \cite{AM}.
However, there are many cases where classical momentum maps still exist
even if this condition is no longer satisfied.
It is reasonable to expect that this phenomenon would happen
for quantum momentum maps as well. However,
we do not know too many examples except for the following:
\begin{ex}
Suppose that $Q$ is a $G$-manifold with action $\phi_{g}$,
which admits a $G$-invariant torsion free connection
$\nabla$. Let $ M=T^{*}Q$ be equipped with the standard cotangent
bundle symplectic structure.
The $G$-action naturally lifts to a symplectic action
$\Phi_{g}$ on $ M=T^{*}Q$
with an equivariant momentum map $J : T^{*}Q \lon \frakg^{*}$\cite{AM}:
$$ <J (\xi_{q}), X> =<\hat{X} , \xi_{q}>, \ \ \ \xi_{q}\in T_{q}Q , \forall X\in \frakg ,$$
where $\hat{X}$ is the vector field on $Q$ generated by $X\in \frakg $.
To any differential operator $D$, we assign its (complete) symbol
as the polynomial $S_{D}$
on $T^{*}Q$ given by
\begin{equation}
\label{eq:symbol}
S_{D} (\xi_{q} )=D e^{<exp_{q}^{-1}x , \xi_{q}>}|_{x=q}, \ \ \ \ \forall \xi_{q}\in T_{q}Q,
\end{equation}
where $exp_{q}: T_{q}Q\lon Q$ is the usual exponential map, defined in a
neighborhood of $0$, corresponding to the connection $\nabla$.
This assignment in fact establishes an isomorphism
between the space $\cald$ of differential operators and that of polynomials on $T^{*}Q$.
Deform the Lie bracket structure on $\calx (Q)$ according to:
\be
&& Z_{\hbar}f= -i \hbar Zf, \ \forall Z\in \calx (Q) \mbox{ and } f\in C^{\infty}(Q); \mbox{ and } \\
&&[Y, Z ]_{\hbar} =-i \hbar [Y, Z], \forall Y, Z\in \calx (Q).
\ee
Since differential operators are generated by $C^{\infty}(Q)$ and
$\calx (Q)$
over the module $C^{\infty}(Q)$, this deformed bracket induces
an $\hbar$-depending multiplication on $\cald $\footnote{A
more intrinsic viewpoint
is to think the tangent bundle $TQ$ as a Lie algebroid, and the above
construction as
deforming the algebroid structure by multiplying the factor $-i \hbar$. Then
such a construction admits an immediate generalization to algebroids, which
should give rise to a $*$-product for the Lie-Poisson structure associated
with a Lie algebroid.}, which in turn induces a $\hbar$-depending
multiplication on the space of polynomials on $T^{*}Q$, hence a $*$-product
on $T^{*}Q$.
It is simple to see that for any $D\in \cald$,
$$ \Phi_{g}^{*} S_{D} =S_{g^{-1} \cdot D}, $$
where $g\cdot D\stackrel{def}{=}\phi_{g^{-1}}^{*}\smalcirc D \smalcirc \phi_{g}^{*}$.
In the equation above, by letting $g=\exp{tX}$, $ \forall X\in \frakg$, and
taking derivative at $t=0$, one obtains immediately that
$$\hat{X}(S_{D})=S_{[\hat{X}, D]}, $$
where $\hat{X}$ on the LHS refers to the vector field on $Q$, while $\hat{X}$ on
the RHS stands for the one on $T^{*}Q$ corresponding to $X\in \frakg$.
This implies that $$\hat{X} f=[\ih J^{*}l_{X}, f], \ \ \ \forall f\in C^{\infty}(T^{*}Q).$$
In other words,
$\muh X=J^{*} l_{X}$ defines a quantum momentum map. However, it is not clear
how to express $\muh f$ explicitly for a general $f\in \pol [[\hbar]]$.
\end{ex}
When quantum momentum maps exist, they are, in general, not unique, as
in the classical case.
Assume that both $\mu_{\hbar}$ and $\nu_{\hbar}: \pol [[\hbar ]] \lon C^{\infty}(M) [[\hbar ]]$
are quantum momentum maps. Let
$\tau_{\hbar}: \frakg \lon C^{\infty}(M) [[\hbar ]] $ be the map defined as:
$$\tau_{\hbar}(\xi )= \mu_{\hbar} (\xi )-\nu_{\hbar} (\xi ), \ \ \ \ \forall \xi\in \frakg .$$
Then, for any $f \in C^{\infty}(M)$, $\ih [ \tau_{\hbar}(\xi ) , f]=0$. Thus it follows that
$\tau_{\hbar}(\xi )$ is a constant formal polynomial of
$\hbar$. Also, it is easy to see
that
$$\tau_{\hbar} ([\xi , \eta ])=0. $$
That is, $\tau_{\hbar}: \frakg \lon \complex [[\hbar ]]$
is a 1-cocycle.
Since all 1-coboundaries are trivial, it follows that the
quantum momentum map is unique if $H^{1}(\frakg )=0$.
\begin{pro}
If $H^{1}(\frakg )=0$, then the quantum momentum map is unique.
\end{pro}
In general, for $\xi\in \frakg$, we have $\mu_{\hbar}\xi = \nu_{\hbar} \xi +\tau_{\hbar} \xi$.
However, it is not clear how to express $\mu_{\hbar}f$,
for a general $f\in \pol [[\hbar ]]$,
in terms of $\nu_{\hbar}$ and $\tau_{\hbar}$.
To see the relation between a quantum momentum map and
a classical one, we start with
the following simple
\begin{lem}
\label{lem:qcm} Suppose that $$f*_{\hbar}g=\sum_{k}\hbar^{k} C_{k}(f, g)$$ is a $*$-product on a
symplectic manifold $M$. Let $X\in \calx (M)$ be
a vector field on $M$, which is an inner derivation
when being considered as an operator on $C^{\infty}(M)[[\hbar ]]$.
Assume that $X f=[\ih a , f]$, $\forall f\in C^{\infty}(M)[[\hbar ]]$.
Then, modulo a constant formal polynomial of $\hbar$, $a$ is of the form $a=\sum_{i}\hbar^{2i}a_{2i}$,
where $a_{0}$ is a hamiltonian function generating the
vector field $X$, and $a_{2k}$, for $k\geq 1$, is determined by
the equation:
$$ \{a_{2k}, f\}= \sum_{i+j=k-1, i\geq 0, j\geq 0} -2 \ih C_{2j+3}(a_{2i}, f).
$$
\end{lem}
\pf It is a direct verification, and is left for the reader. \qed
%is a $G$-equivariant $*_{\hbar}$, and $\hat{\xi}$, for $\xi \in
%\frakg$ is an inner derivation: $\hat{\xi}= ad\ih a_{\xi }$.
%%Then, $a_{\xi}$ is of the form $a_{\xi}=\sum_{i}\hbar^{2i}a_{2i}$,
%where $a_{0}$ is a hamiltonian function generating the
%vector field $\hat{\xi}$ and $a_{2k}$, for $k\geq 1$, is determined by
%the equation:
%$$ \{a_{2k}, f\}=-\sum_{i+j=k-1, i\geq 0, j\geq 0}C_{2j+3}(a_{2i}, f)
%$$
%%CHECK COIEFFICIENTS
%\end{lem}
As a vector space, $\pol [[\hbar ]]$ is canonically isomorphic to
$\poll [[\hbar ]]$, the space of formal power series of $\hbar$ with
coefficients being polynomials on $\frakg^*$.
The isomorphism is established by the symmetrization (see \cite{B} for
details). Therefore, the algebra structure on $\pol [[\hbar ]]$
induces a $*$-product on $\poll [[\hbar ]]$, which gives rise
to a deformation quantization for the Lie-Poisson structure $\frakg^*$. Below, we will identify
these two spaces and use them interchangeably if there is no confusion.
\begin{pro}
\label{pro:equ-moment}
Suppose that $*_{\hbar} $ is
a $G$-equivariant $*$-product on a symplectic manifold $M$. Assume that $\muh : \pol [[\hbar]]
\lon C^{\infty}(M)[[\hbar ]]$ is a quantum momentum
map.
Then, $M$ is a hamiltonian $G$-space, i.e., the symplectic
$G$-action admits
an equivariant (classical) momentum map $J$.
Moreover,
$$\muh f=J^{*}f +O(\hbar ), \ \ \ \ \forall f\in \poll .$$
\end{pro}
\pf Since $\muh \xi, \ \forall \xi\in \frakg$, depends on $\frakg$
linearly, it defines a map
$J:X \lon \frakg^*$ uniquely by
the relation:
$$\mu_{\hbar}(l_{\xi })=J^{*}l_{\xi }+ O(\hbar) , \ \ \forall \xi\in \frakg .$$
Then clearly $J$ is a (classical) momentum map according to
Lemma \ref{lem:qcm}.
Write the $2$-cocycle $\lambda$ defined by Equation (\ref{eq:2cocycle}) as
$$\lambda =\lambda_{0}+\hbar \lambda_{1}+\cdots . $$
Then each $\lambda_{i}: \wedge^{2}\frakg \lon \complex$ is a
Lie algebra $2$-cocycle. Moreover,
it is simple to see that
$$\lambda_{0}(\xi, \eta )= J^{*}l_{[\xi ,\eta ]}-\{J^{*}l_{\xi} , \ J^{*}l_{\eta }\} .$$
The vanishing of $\sigma$ implies that $\sigma_{0}=0$, which
means that $J$ is equivariant. The rest of the proposition
thus follows trivially. \qed
It is, however, not clear whether the converse
of Proposition \ref{pro:equ-moment} is true or not.
We end this section by posing the following
\begin{quote}
{\bf QUESTION} Does the existence of a classical moment
map imply the existence of a quantum moment map?
\end{quote}
\section{Quantum dual pair}
This is a continuation of the last section. We will assume the same
hypothesis as in the previous section, and in particular, assume that
a quantum momentum map $\mu_{\hbar}: \pol [[\hbar ]] \lon C^{\infty}(M) [[\hbar ]]$
exists.
Let $C^{\infty}(M)^{G}$ denote the space of all $G$-invariant
functions on $M$.
\begin{pro}
\label{prop:comm1}
$$(\muh \pol [[\hbar ]] )' \cong C^{\infty}(M)^{G}[[\hbar ]].$$
\end{pro}
\pf Let $f=\sum_{i}\hbar^{i} f_{i} \in C^{\infty}(M)[[\hbar ]]$.
Then $f$ commutes with $\muh \pol [[\hbar ]]$ iff it commutes with
its generators. That is,
$[ f , \muh \xi ]=0, \ \ \forall \xi \in \frakg$.
The latter is equivalent to that $\hat{\xi}f=0$, or in
other words, $f$ is $G$-invariant. \qed
According to this proposition, we have $\muh (\pol [[\hbar ]])\subseteq
(C^{\infty}(M)^{G}[[\hbar ]])'$. In order to describe the \commutant
$(C^{\infty}(M)^{G}[[\hbar ]])'$ completely, we need to extend the
quantum momentum map $\muh$.
It is well-known that the star-product on $\poll [[\hbar ]]$ naturally extends
to a star-product on smooth
functions $C^{\infty}(\frakg^{*})[[\hbar ]]$.
Below we will indicate that a quantum momentum map, if it exists,
extends to $C^{\infty}(\frakg^{*})[[\hbar ]]$ as well. More
precisely, we have
\begin{pro}
\label{prop:extension}
Let $\mu_{\hbar}: \poll [[\hbar ]] \lon C^{\infty}(M) [[\hbar ]]$ be a quantum
momentum map. Then it naturally extends to an algebra morphism,
denoted by the same notation $\mu_{\hbar}$, from $C^{\infty}(\frakg^{*})[[\hbar ]]$
to $ C^{\infty}(M) [[\hbar ]]$ such that
\begin{enumerate}
\item for any $f\in C^{\infty}(\frakg^{*})$, $\mu_{\hbar}f$
commutes with $C^{\infty}(M)^{G}[[\hbar ]]$, and
\item for any $f\in C^{\infty}(\frakg^{*})$,
$\muh f=J^{*}f +O(\hbar )$.
\end{enumerate}
\end{pro}
\pf Given a smooth function $f\in C^{\infty}(\frakg^{*})$,
for any $x_{0}\in M$, let $\tilde{f}_{x_{0}} (u), \ u\in \frakg^{*} $
denote its Talyor expansion
at the point $u_{0}=J(x_{0} )$. Define\footnote{We are grateful to
A. Weinstein for suggesting this method of extension.}
\begin{equation}
(\mu_{\hbar} f )(x_{0} )=(\mu_{\hbar} \tilde{f}_{x_{0}} )|_{x=x_{0}}.
\end{equation}
It is clear that this definition coincides with the original
$\mu_{\hbar}$ when $f$ is a polynomial, so it is indeed
an extension of the given quantum momentum map. It is
also obvious, from definition, that $\mu_{\hbar}f$
commutes with $C^{\infty}(M)^{G}[[\hbar ]]$.
It is simple to see that each term in the
expansion of $\tilde{f}_{x_{0}}-f(J(x_{0}))$,
as a function in $u$,
is a homogeneous polynomial in $u-u_{0}$.
Hence, we have $\muh f=J^{*}f +O(\hbar )$.
It remains to check that $\mu_{\hbar}$ is an algebra
homomorphism. This follows from the fact that
$\widetilde{(f*_{\hbar }g)}_{x_{0}}=
\tilde{f}_{x_{0}} *_{\hbar }\tilde{g}_{x_{0}}$,
for all $f$ and $g\in C^{\infty}(\frakg^* )$.
To see this, we write $f*_{\hbar }g=\sum \hbar^{k}C_{k}(f, g)$.
Then, this essentially follows from
the fact that $\widetilde{C_{k}(f, g)}_{x_{0}}=C_{k}(\tilde{f}_{x_{0}} , \tilde{g}_{x_{0}})$, which is in turn a consequence of the fact that
$\poll [[\hbar ]]$ is a subalgebra under the $*$-product on $C^{\infty}(\frakg^{*})[[\hbar ]]$. \qed
\begin{pro}
\label{prop:comm3}
Under the same hypothesis as in Proposition \ref{prop:extension}, if,
in addition, the $G$-action is free and proper,
$$(C^{\infty}(M)^{G}[[\hbar ]])'\cong \muh
(C^{\infty}(\frakg^{*})[[\hbar ]] ). $$
\end{pro}
\pf According to Proposition \ref{prop:extension}, $ \muh (C^{\infty}(\frakg^{*})[[\hbar ]] )
\subseteq (C^{\infty}(M)^{G}[[\hbar ]])'$.
To prove the other direction, let us assume that
$f=\sum_{i}\hbar^{i} f_{i} \in (C^{\infty}(M)^{G}[[\hbar ]])'$.
It is clear that for any $g\in C^{\infty}(M)$,
$$\ih [f , g] =\{f_{0}, \ g\}+O(\hbar ). $$
Then it follows that
$\{f_{0}, \ g\}=0$,
$ \forall \ g\in C^{\infty}(M)^{G}$.
This implies that
$f_{0}= J^{*}f'_{0} $ for some smooth function $f'_{0}\in C^{\infty}(\frakg^{*})$. Now
according to Proposition \ref{prop:extension},
$\muh f'_{0}=J^{*}f'_{0} +O(\hbar )
=f_{0}+O(\hbar )$. Therefore, $f-\muh f'_{0} =\hbar \tilde{f}$,
where $\tilde{f}\in (C^{\infty}(M)^{G}[[\hbar ]])' $ since
both $f$ and $\muh f'_{0}$ belong to $ (C^{\infty}(M)^{G}[[\hbar ]])'$.
By repeating the same argument on $\tilde{f}$ and so on, we deduce that
$f\in \muh (C^{\infty}(\frakg^{*})[[\hbar ]] )$. \qed
Combining Propositions \ref{prop:comm1}-\ref{prop:comm3}, we have
\begin{thm}
\label{thm:dual-pair}
Suppose that $*_{\hbar}$ is a $G$ equivariant $*$-product on $C^{\infty}(M)[[\hbar ]]$
with a quantum momentum map $\muh$.
Assume that the action is free and proper, then
\be
(C^{\infty}(M)^{G}[[\hbar ]])' &\cong & \muh
(C^{\infty}(\frakg^{*})[[\hbar ]] )\\
( \muh C^{\infty}(\frakg^{*})[[\hbar ]] )'&\cong
&C^{\infty}(M)^{G}[[\hbar ]] .
\ee
\end{thm}
Recall that if $A$ is an associative algebra and $B\subseteq A$ is a
subset, then $B'$, the commutatant of $B$,
is an associative subalgebra of $A$. If $B$ is
also the commutatnt of $B'$, then $B$ and $B'$ are called
{\em mutual commutants}.
% In other words,
%$C^{\infty}(M)^{G}[[\hbar ]]'$ and $ \muh (C^{\infty}(\frakg^{*})[[\hbar ]] )$
%are mutual commutants.
The notion of mutual commutants is an important concept in the
theory of associative algebras, especially in operator algebras.
It has been generalized to the context of groups by Roger Howe \cite{Howe},
called dual pairs of groups, in his study of representation
theory and mathematical physics.
On the classical level, or more
precisely on the level of Poisson manifolds, an analogue was introduced by
Weinstein \cite{Weinstein:1983}, which is called dual
pair of Poisson manifolds.
In fact, Poisson manifolds $\frakg^*$ and $M/G$,
together with the Poisson maps $J: M\lon \frakg^*$ and $p: M\lon M/G$,
consist a dual pair in terms of Weinstein
\cite{Weinstein:1983}.
%We can consider the above results as a quantum counterpart
%of a dual pair.
Now $(C^{\infty}(\frakg^{*})[[\hbar ]] , *_{\hbar}) $
provides a deformation quantization for the Lie-Poisson
structure $\frakg^*$ while $(C^{\infty}(M)^{G}[[\hbar ]], *_{\hbar})$
quantizes the reduced Poisson space $M/G$. So Theorem
\ref{thm:dual-pair} equivalently says that under
deformation quantization,
the classical dual pair $\frakg^*$ and $M/G$
becomes mutual commutants. For this reason, we shall
call the pair of algebras
$(C^{\infty}(\frakg^{*})[[\hbar ]] , *_{\hbar}) $
and $(C^{\infty}(M)^{G}[[\hbar ]], *_{\hbar})$
a {\em quantum dual pair}.
In general, given two Poisson manifolds
$P_{1}$ and $P_{2}$ and their deformation
quantization $(C^{\infty}(P_{1})[[\hbar ]], *_{\hbar})$ and
$(C^{\infty}(P_{2})[[\hbar ]], *_{\hbar})$, we say that
they consist a {\em quantum dual pair} if there is a
symplectic manifold $M$ and a star-product $(C^{\infty}(M)[[\hbar ]],
*_{\hbar})$ on $M$, and algebra morphisms $\rho_{1}: C^{\infty}(P_{1})[[\hbar ]]\lon C^{\infty}(M)[[\hbar ]]$ and
$\rho_{2}: C^{\infty}(P_{2})[[\hbar ]]\lon C^{\infty}(M)[[\hbar ]]$
such that $\rho_{1} (C^{\infty}(P_{1})[[\hbar ]])$
and $ \rho_{2} (C^{\infty}(P_{2})[[\hbar ]])$ are mutual commutants.
%Under this definition, Theorem \ref{thm:dual-pair} is equivalent to saying that
%$C^{\infty}(M)^{G}[[\hbar ]]$ and $C^{\infty}(\frakg^{*})[[\hbar ]]$
%consist a quantum dual pair.
Unfortunately, at the moment we only know a few
examples of deformation quantizable Poisson manifolds.
In fact, we do not know any other examples of quantum dual pairs
besides this and the trivial ones.
In particular, it is not clear in general whether
a classical dual pair (or Morita equivalent Poisson
manifolds \cite{Xu:1991}) can be quantized to a quantum
dual pair or not. The answer to all these
questions relies upon how successful it is the
deformation quantization theory of Poisson manifolds.
\section{Appendix A (Derivations of $*$-algebras)}
In this Appendix, we collect some basic
facts concerning derivations of a $*$-algebra $C^{\infty}(M)[[\hbar ]]$ on
a symplectic manifold $M$.
Some of them are well known. However, as we
shall see, Fedosov method even sheds new light on understanding
these results.
\begin{defi}
A derivation of
a $*$-algebra $(C^{\infty}(M)[[\hbar ]] , *_{\hbar}) $ is a
formal power series of $\hbar$ with coefficients
being linear operators on $C^{\infty}(M)$: $\delta=D_{0}+\hbar D_{1}
+\cdots +\hbar^{i}D_{i}+\cdots $ such that
\begin{equation}
\label{eq:der}
\delta (f*_{\hbar}g)=\delta f*_{\hbar}g+f *_{\hbar} \delta g , \forall f, g\in C^{\infty}(M)[[\hbar ]].
\end{equation}
A derivation is said to be inner if $\delta =ad \ih H=[\ih H , \cdot ]$ for some $H\in C^{\infty}(M)[[\hbar ]]$.
\end{defi}
Suppose that $\delta=\sum_{i} \hbar^{i}D_{i}$
is a derivation.
By expanding both sides of Equation (\ref{eq:der}),
the $\hbar^{0}$ terms yield that
$D_{0}(fg)=D_{0}(f)g+fD_{0}(g)$. That is, $D_{0}$ is
a vector field.
By considering the $\hbar^{1}$ terms, one obtains that
$$D_{0}\{f, g\}-\{D_{0}f, g\}-\{f, D_{0}g\}=
(D_{1}f)g+f(D_{1}g)-D_{1}(fg). $$
Since the LHS is skew-symmetric with respect to $f$ and $g$
while the RHS is symmetric, both terms have to vanish identically.
Thus, $D_{1}$ is a vector field and $D_{0}$ is a
symplectic vector field.
Moreover, we have the following (see \cite{BFFLS} for an equivalent
result, which is however concerning derivations of the corresponding
deformed Lie algebra; see also \cite{BCG}).
\begin{thm}
\label{thm:derivation}
Suppose that $D=D_{0}+ \hbar
D_{1} +\cdots \hbar^{i}D_{i}+\cdots $ is a derivation of a $*$-algebra
$(C^{\infty}(M)[[\hbar ]] , *_{\hbar})$ on a
symplectic manifold $M$. Then,
\begin{enumerate}
\item the operator $D_{i}$, for each $i$,
is a differential operator, and in particular
$D_{0}$ is a symplectic vector field;
\item there is a canonical one-one correspondence
between derivations and $Z^{1}(M)[[\hbar ]]$;
\item under such a correspondence, inner
derivations correspond to exact 1-forms in $B^{1}(M)[[\hbar ]]$.
\end{enumerate}
\end{thm}
To begin with, we will consider Fedosov algebras, and assume that $A=W_{D}$ for
some Fedosov connection $D$. We need a couple of lemmas first.
\begin{lem}
\label{lem:dk}
Let $K\in \gm W$ be a section. Then,
$\rho =ad\ih K=[\ih K, \cdot ]$ defines a derivation of $W_{D}$
iff $DK$ is a scalar closed one-form on $M$.
\end{lem}
\pf It is clear that $\rho $, defined in this way,
satisfies
the derivation property.
For any $a\in W_{D}$,
$$D\rho (a)=[\ih DK , a]+[\ih K, Da ].$$
If $\rho W_{D}\subset W_{D}$, it follows that $[DK, a]=0, \ \forall a\in
W_{D}$. Thus $DK=\theta $ is a scalar one-form on $M$,
i.e., $\theta \in \Omega^{1}(M)[[\hbar ]]$. Then $\theta $ must
be closed since $d\theta =D\theta =D^{2}K=0$.
The converse follows essentially from the same argument backwards. \qed
\begin{lem}
\label{lem:der-extension}
For a $*$-product on a symplectic manifold $M$,
any symplectic vector field $X$ may extend
to a derivation $\delta =\sum_{i} \hbar^{2 i}D_{2 i}$,
with $D_{2i}$ being differential operators, such that $D_{0}=X$.
If $X$ is a hamiltonian
vector field, $\delta$ may be chosen as an inner derivation.
\end{lem}
\pf Since every $*$-product is equivalent to a Fedosov $*$-product,
we may confine ourselves to a Fedosov algebra $W_{D}$.
Let $\theta =X\per \omega $. Then $\theta $ is a
closed one-form on $M$. Let $K\in \gm (W)$ be
the section satisfying
\begin{equation}
\label{eq:k0}
DK=\theta \ \ \mbox{and } K|_{y=0}=0.
\end{equation}
Note that such a section always
exists according to the Fedosov iteration method.
In fact, $K$ is uniquely determined by
the following iteration formula:
\begin{equation}
K=-\delta^{-1}\theta +\delta^{-1}(\partial K+[\ih \gamma , K]) .
\end{equation}
Take $\delta =ad\ih K $. Then $\delta$ is easily seen to be a required
derivation. In fact, locally $\theta $ is
exact, so $\delta$ can be expressed as an inner derivation generated by some function
on $M$, which is clearly a formal power series
of $\hbar^2$ with coefficients being differential operators.
If $X$ is a hamiltonian vector field with the hamiltonian function $H$,
we may take $\delta =[\ih H, \cdot ]$, which is an inner derivation having the
desired property. \qed
{\bf Proof of Theorem \ref{thm:derivation}}\ \ According to
the observation preceding Theorem \ref{thm:derivation}, $D_{0}$ is a
symplectic vector field. Thus,
it extends to a derivation $\delta_{0}=D_{0}+O(\hbar )$, whose
coefficients are differential operators, according to the lemma
above.
Let $ \tilde{\delta }=\frac{1}{\hbar} (\delta -\delta_{0})$.
Then $ \tilde{\delta } $ is a derivation and
$\delta =\delta_{0} +\hbar \tilde{\delta } $.
Applying this process repeatedly, one obtains
that $\delta =\sum \hbar^{i}\delta_{i}$, where every
$\delta_{i}$ is a derivation whose coefficients
are differential operators. So $\delta$ itself is such a
derivation as well.
To continue, without loss of generality,
we shall confine ourselves to the case of a Fedosov
algebra $W_{D}$. Assume that $\delta : W_{D}\lon W_{D}$
is a derivation. For any $x_{0}\in M$, we define a derivation
$\rho_{x_{0}}$ on $W_{x_{0}}$ by
\begin{equation}
\rho_{x_{0}} a(y ,\hbar )=(\delta \tilde{a})|_{x=x_{0}}, \ \ \ \ \mbox{where }
a(y ,\hbar ) \in W_{x_{0}} \ \mbox{and }
\tilde{a} \in W_{D} \ \mbox{ such that }
\tilde{a}(x_{0}, y, \hbar )=a(y, \hbar ).
\end{equation}
Clearly,
$\rho_{x_{0}} $ is well defined since $\delta $ is a local operator according
to Part (i).
Since all derivations on $W_{x_{0}}$ are inner, there is an
element $K(x_{0})\in W_{x_{0}}$ such that
$\rho_{x_{0}}=ad\ih K(x_{0})$. By requiring
that $K(x_{0})|_{y=0}=0$, $K(x_{0})$ will be unique.
Repeating this process pointwisely, we obtain a
global section $K\in \gm (W)$ with $K|_{y=0}= 0$ such that
$$\delta \tilde{a}=[\ih K , \tilde{a} ], \ \ \ \ \forall \tilde{a}\in W_{D}. $$
According to Lemma \ref{lem:dk}, $\theta =DK$ belongs to
$Z^{1}(M)[[\hbar ]]$. In this way, we obtain a map $\phi $
from the space of derivations $Der C^{\infty}(M)[[\hbar ]]$ to
that of closed 1-forms $Z^{1}(M)[[\hbar ]]$.
Conversely, given any $\theta \in Z^{1}(M)[[\hbar ]]$, the equation:
\begin{equation}
\label{eq:k}
DK=\theta , \ \ \ \mbox{ and } K|_{y=0}=0
\end{equation}
always has a unique solution. Thus,
$\delta =ad \ih K$ defines a derivation of $W_{D}$ such that $\phi \delta =\theta $.
In other words, $\phi$ is onto.
It is also simple to see that $\phi$ is injective since the
solution to Equation (\ref{eq:k}) is unique.
Suppose that $\delta$ is an inner derivation:
$\delta =ad \ih H$ for some $H\in W_{D}$. Thus, $K=H-H_{0}$, where $H_{0}=H|_{y=0}\in
C^{\infty}(M)[[\hbar ]]$, and
$\theta =DK=D(H-H_{0})=-dH_{0}$, which is clearly exact.
Conversely, if $ \phi \delta =\theta $ is exact, i.e., $\theta =dH$ for
some $H\in C^{\infty}(M)[[\hbar ]]$, we have
$D(K-H)= \theta -dH=0$. That is, $K-H \in W_{D}$.
Thus, $\delta =ad \ih K=ad \ih (K-H)$ is clearly inner.
This concludes the proof of the theorem. \qed
It is well known that the bracket of any two symplectic vector
fields is hamiltonian.
As another immediate application of
Fedosov method, we obtain the following ``quantum" analogue of this fact.
\begin{pro}
\label{pro:exact}
Let $*_{\hbar}$ be any $*$-product on a symplectic
manifold $M$. Then the bracket of any derivations
is an inner derivation.
\end{pro}
\pf Without loss of generality, assume that $A=W_{D}$ for
some Fedosov connections, and
$\delta_{1}=ad{\ih K_{1}} $ and $\delta_{2}=ad{\ih K_{2}}$ are
derivations of $W_{D}$,
where $K_{1}$ and $K_{2}$ are sections of $W$. Then, $DK_{1}$ and $DK_{2}$
belong to $Z^{1}(M)[[\hbar ]]$ according to Lemma \ref{lem:dk}.
Now $[\delta_{1} , \delta_{2}]=ad \ih (\ih [K_{1} , K_{2}])$.
It is clear that $K=\ih [K_{1} , K_{2}]$ is a section of $W$ and
$DK=0$. Therefore, $[\delta_{1} , \delta_{2}]$ is an inner
derivation. \qed
{\bf Remark} (1). We note that our definition of inner derivations differs
from the usual one by a factor $\ih$. If we modify the
usual Hochschild coboundary operator by multiplying
the factor $\ih$, Theorem \ref{thm:derivation} implies that the first
Hochschild cohomology $H^{1}(C^{\infty}(M)[[\hbar ]], C^{\infty}(M)[[\hbar ]])$ is
isomorphic to $H^{1}(M)[[\hbar ]]$ (see \cite{BFFLS}).
Similarly, for higher order cohomology, we expect that
$H^{*} (C^{\infty}(M)[[\hbar ]], C^{\infty}(M)[[\hbar ]]) \cong H^{*} (M) [[\hbar ]]$.
Under such an isomorphism, the Gerstenhaber bracket on Hochschild cohomology
would go to zero on the right hand side. Moreover,
the isomorphism at the second order should
provide an intrinsic explanation
for the characteristic class of a $*$-algebra (\cite{WeinsteinX:1996}).
(2). Using the identification as in Theorem \ref{thm:derivation}, one obtains
a Lie bracket on the space $Z^{1}(M)[[\hbar ]]$ such that the bracket
of any closed one-forms is exact (see Proposition \ref{pro:exact}).
It is easy to see that for any $\theta_{1} ,\ \theta_{2}$,
$[\theta_{1} , \theta_{2}]=\{\theta_{1} , \theta_{2}\}+O(\hbar )$,
where $\{ \cdot , \cdot \}$ refers to the standard Lie bracket
on one-forms induced from the Poisson bracket (see \cite{AM}). However, it is
difficult to find an explicit expression for the entire bracket
$[\cdot , \cdot]$. The latter should be related to
the Weyl curvature of the deformation.
%It would be interesting to explore this idea further.
Also, for a symplectic manifold, it is well known that
the Poisson bracket defined on closed one-forms extends to a bracket
on all one-forms. It is not clear, however, whether one can extend
the bracket $[\cdot , \cdot]$ above to $\Omega^{1}(M)[[\hbar ]]$.
It seems that these problems are all
related to the question raised by Weinstein regarding
``quantum Lie algebroids" \cite{Weinstein:1996}.
Another interesting consequence of Theorem \ref{thm:derivation} is the following:
\begin{cor}
If $\delta =D_{0}+\hbar D_{1}+\hbar^{2} D_{2} +\cdots =\sum_{i}
\hbar^{i}D_{i} $ is
a derivation. Then both
$\delta_{even}=D_{0}+\hbar^{2} D_{2} +\cdots =\sum_{i}\hbar^{2i}D_{2i}$
and $\delta_{odd}=D_{1}+\hbar^{2}D_{3} +\cdots =\sum_{i}\hbar^{2i}D_{2i+1}$,
are derivations.
\end{cor}
\pf Without loss of generality, we assume that this is a Fedosov $*$-product, and
consider $W_{D}$ as our algebra.
Then each derivation $\delta $ can be written as
$\delta =[\ih K, \cdot ] $, for some $ K\in \gm W$. Let $\theta =DK \in
Z^{1}(M)[[\hbar ]]$. Write $\theta =\theta_{even}+\hbar \theta_{odd}$, where
$\theta_{even}$ is the sum of all even terms in $\hbar$ while
$\theta_{odd}$ is the sum of all odd terms in $\hbar$ divided by $\hbar$.
Let $K_{even }$ and $K_{odd}$ be the sections of $W$ corresponding to
$\theta_{even}$ and $\theta_{odd}$, respectively, given by Equation (\ref{eq:k}), and
let
\be
\tilde{ \delta}_{even}&=&ad \ih K_{even } \ \ \mbox{and}\\
\tilde{ \delta}_{odd}&=&ad \ih K_{odd} .
\ee
Then, clearly both $\tilde{ \delta}_{even}$ and $\tilde{ \delta}_{odd}$
are derivations, and consist only even powers of $\hbar$ according to
Lemma \ref{lem:der-extension}.
Also $\delta = \tilde{ \delta}_{even}+\hbar \tilde{ \delta}_{odd}$ by construction.
Therefore, $\tilde{ \delta}_{even}=\delta_{even}$ and
$\tilde{ \delta}_{odd}=\delta_{odd}$.
This concludes the proof. \qed
\section{Appendix B}
A $*$-product
$$f*_{\hbar}g=\sum_{k}\hbar^{k}C_{k}(f, g)$$
on a regular Poisson manifold is said to be
{\em \preferred } if $C_{k} (\cdot , \cdot )$ is a
leafwise bidifferential
operator for every $k$.
In this section, for completeness,
we will outline a proof for the following result,
whose proof
% (incomplete ?? CHECK WITH BORIS)
can also be found in \cite{NT1}.
\begin{pro}
\label{prop:main}
Suppose that $P$ is a regular Poisson manifold whose symplectic
foliation is a fibration $P\lon M$.
Assume that the second Betti number of the fibers is zero. Then there exists
essentially a unique \preferred $*$-product on $P$.
I.e., any two \preferred $*$-products
are equivalent, and the equivalence operator
can be chosen as a formal power series in $\hbar$ with
coefficients being leafwise differential operators.
\end{pro}
Recall that, given a fibration $\calf$ on a manifold $P$,
a leafwise Hochschild cochain is a $k$-linear
form on $C^{\infty}(P)$ with value in $C^{\infty}(P)$, which
requires to be a leafwise $k$-differential operator on $P$.
The Hochschild differential is given by
\be
(\delta c)(u_{0}, \cdots , u_{k})&=&u_{0}c(u_{1}, \cdots , u_{k})+
\sum_{i=0}^{k-1}(-1)^{i+1}c(u_{0}, \cdots , u_{i}u_{i+1}, \cdots , u_{k})\\
&&+(-1)^{k+1}c(u_{0}, \cdots , u_{k-1})u_{k} .
\ee
As usual, the space of $k$-cochains is denoted by
$C^{k}_{\calf}(C^{\infty}(P), C^{\infty}(P))$, while
its cohomology is denoted by
$H^{k}_{\calf}(C^{\infty}(P), C^{\infty}(P))$.
\begin{lem}
(Nest-Tsygan \cite{NT1})
Let $\calf$ be a fibration on a manifold $P$. Then,
\begin{equation}
H^{k}_{\calf}(C^{\infty}(P), C^{\infty}(P)) \cong \gm (\wedge^{k}T\calf).
\end{equation}
%$(f, g)=<D, df\wedge dg>, \ \forall
%f, g\in C^{\infty}(P)$ for some leafwise bivector field $D$.
\end{lem}
An immediate consequence is the following
\begin{lem}
\label{lem:V}
\begin{enumerate}
\item If $c \in C^{2}_{\calf}(C^{\infty}(P), C^{\infty}(P))$ is an antisymmetric two-cocycle, then $c$
is a leafwise bivector field, i.e., $c\in \gm (\wedge^{2}T\calf )$.
\item If $c\in C^{2}_{\calf}(C^{\infty}(P), C^{\infty}(P))$ is
a symmetric two-cocycle, then it is a coboundary.
\end{enumerate}
\end{lem}
{\bf Proof of Proposition \ref{prop:main}}
The proof is simply a modification of the standard one for symplectic
manifolds. The idea, roughly speaking, is as follows.
The classification of $*$-products is equivalent to that
of their commutator Lie algebras $[\cdot , \cdot ]_{*}$,
which are Lie algebra deformations of the Poisson bracket
$\{\cdot , \cdot \}$.
The latter is classified by the 2nd order leafwise
Chevalley cohomology of the Poisson Lie algebra.
However, the 2nd order leafwise Chevalley cohomology is isomorphic
to the second leafwise de-Rham cohomology, and is zero by assumption.
Recall that the Chevalley coboundary (see \cite{L}) operator is
given by:
\begin{equation}
(\partial c)(u_{0}, \cdots , u_{p} )=\epsilon_{0\cdots p}^{\lambda_{0}
\cdots \lambda_{p}}[\frac{1}{p!}\{u_{\lambda_{0}}, c(u_{\lambda_{1}},
\cdots , u_{\lambda_{p}}) \}-\frac{1}{2(p-1)!}c(\{u_{\lambda_{0}}, u_{\lambda_{1}}\},
u_{\lambda_{2}}, \cdots , u_{\lambda_{p}}) ],
\end{equation}
where $u_{\lambda} \in C^{\infty}(P)$ and $\epsilon$ is the Kronecker symbol.
If $c\in C^{2}_{\calf}(C^{\infty}(P), C^{\infty}(P))$ is antisymmetric,
$$(\partial c)(u_{0}, u_{1}, u_{2})=
(\{u_{0}, c(u_{1}, u_{2})\}-c(\{u_{0}, u_{1}\}, u_{2}))+c.p.,
$$
where $c.p.$ stands for the cyclic permutation.
Hence if $c\in \gm(\wedge^{2} T \calf )$,
$\partial c=[\pi , c]=d_{\pi }c$,
where the bracket is the Schouten bracket on multivector fields and
$d_{\pi }: \gm (\wedge^{*}T \calf ) \lon \gm (\wedge^{*+1}T \calf )$ is
the differential operator defining the (leafwise)-Poisson cohomology.
Suppose that
\be
f*g&=& \sum \hbar^{k} C_{k}(f, g), \ \ \mbox{and }\\
f*'g&=& \sum \hbar^{k} C_{k}'(f, g)
\ee
are two \preferred $*$-products on $P$.
We need to construct an equivalence operator between them.
This will proceed by induction.
Assume that they are equivalent up to $i=2k$, i.e., we can find
an equivalence operator under which, $C_{i}=C_{i}'$ for $0\leq i\leq 2k$.
Then, $\delta (C_{2k+1}-C'_{2k+1})=0$. Since
$C_{2k+1}-C'_{2k+1}$ is skew-symmetric,
it belongs to $ \gm (\wedge^{2}T\calf )$ by Lemma \ref{lem:V}.
On the other hand, both $[\cdot , \cdot ]_{*}$
and $[\cdot , \cdot ]_{*'}$ are deformations
of the Poisson Lie algbra $\{\cdot , \cdot \}$.
Thus, $\partial (C_{2k+1}-C'_{2k+1})=0$.
That is, $[\pi , C_{2k+1}-C'_{2k+1} ]=0$. In other words, $C_{2k+1}-C'_{2k+1}$
is a 2-cocycle of the (leafwise)-Poisson cohomology, which is
isomorphic to the leafwise de-Rham cohomology. Since the 2nd leafwise de-Rham cohomology
is zero by assumption,
there is a vector field $X\in \gm (T\calf )$ such that
$ C_{2k+1}-C'_{2k+1}=[\pi , X]$. In other words,
$$C_{2k+1}(f, g)-C'_{2k+1}(f, g)=\{Xf ,g\}+\{f, Xg\}-X\{f ,g\}, \ \ \forall
f , g\in C^{\infty}(P). $$
It is then easy to see that $T=1+\hbar^{2k}X$ establishes
an isomorphism between $*$ and $*'$ up to $\hbar^{2k+1}$.
That is,
$$T(f*g)-Tf*'Tg=O(\hbar^{2k+2}). $$
Finally, we assume that $*$ and $*'$ are equivalent up to $n=2k-1$.
By applying an equivalence operator, we may assume that
$C_{i}=C_{i}'$ for $0\leq i\leq 2k-1$.
Then, $\delta (C_{2k}-C'_{2k})=0$. Since $C_{2k}-C'_{2k}$
is a symmetric 2-cochain, $c_{2k}-c'_{2k}=\delta D$,
for some $D\in C^{1}_{\calf}(C^{\infty}(P), C^{\infty}(P))$ according
to Lemma \ref{lem:V}.
Thus $T=1+\hbar^{2k}D$ establishes an isomorphism between
$*$ and $*'$ up to $\hbar^{2k}$. This concludes the proof. \qed
\section{Appendix C}
Recall that a {\em \nat } $*$-product on a symplectic manifold $M$ is
a $*$-product:
$$ u*_{\hbar }v=uv-\frac{i\hbar }{2}\{u , v\} + \half (-\frac{i\hbar }{2})^{2}Q_{2}(u, v)+\cdots , $$
where $Q_{2}(u, v)$ is a bidifferential operator of order $2$ in each argument.
It is well known \cite{BFFLS} \cite{L} that associated to a \nat
$*$-product there
is a canonical torsion-free symplectic connection. More precisely, we
have
\begin{pro}
\label{pro:nat-split}
Let $ u*_{\hbar }v=uv-\frac{i\hbar }{2}\{u , v\} +\half (-\frac{i\hbar }{2})^{2}Q_{2}(u, v)+\cdots $
be a \nat $*$-product on $M$. Then there exists a unique torsion-free symplectic connection
$\nabla$ such that
$$Q_{2}(u, v)=P_{\nabla}^{2}(u, v)+H(u, v), $$
where $H(u, v)$ is a bidifferential operator of maximum order $1$ in each argument.
\end{pro}
For completeness, we outline a proof here in this appendix. \\\\\\
{ \bf Proof of Proposition \ref{pro:nat-split}} Take a torsion free symplectic
connection $\tilde{\nabla}$. Then,
$$ u*_{\hbar }v=uv-\frac{i\hbar }{2}\{u , v\} +\half (-\frac{i\hbar }{2})^{2} P_{\tilde{\nabla}}^{2}(u, v)$$
is a deformation of order up to $\hbar^2$. Therefore,
$\delta (Q_{2}- P_{\tilde{\nabla}}^{2})=0$, where $\delta$ denotes the usual
Hochschild differential. Since $Q_{2}- P_{\tilde{\nabla}}^{2}$ is symmetric, it
must be a 2-coboundary. Hence there is a differential operator $D$ of maximum
order $3$ such that
$$Q_{2}- P_{\tilde{\nabla}}^{2}=\delta D. $$
The principal term of $D$ corresponds to a $3$-covariant symmetric tensor $T^{ijk}$.
In local coordinates, write $\tilde{\nabla}_{\parr{i}}\parr{j}=\tilde{\gm}_{ij}^{k}
\parr{k}$, and let $\tilde{\gm}^{ijk}=\tilde{\gm}_{lm}^{i}\pi^{lj}\pi^{mk}$.
Set
\begin{equation}
{\gm}^{ijk}=\tilde{\gm}^{ijk}+3 T^{ijk} .
\end{equation}
Since $T^{ijk}$ is a completely symmetric tensor, the equation above
defines a torsion free symplectic connection $\nabla_{\parr{i}}\parr{j}={\gm}_{ij}^{k}\parr{k}$
with $\gm^{ijk}={\gm}_{lm}^{i}\pi^{lj}\pi^{mk}$.
A simple calculation yields that
\begin{eqnarray}
P_{\tilde{\nabla}}^{2}(u, v)-P_{\nabla}^{2}(u, v)&=&\pi^{i_{1}j_{1}}\pi^{i_{2}j_{2}}
(\tilde{\gm}_{i_{1}i_{2}}^{k}\tilde{\gm}_{j_{1}j_{2}}^{l} -
{\gm}_{i_{1}i_{2}}^{k}{\gm}_{j_{1}j_{2}}^{l} )\parrr{u}{k}\parrr{v}{l} \nonumber \\
&& + 3T^{i_{1}i_{2}i_{3}}\parrr{u}{i_{1}}\frac{\partial^{2}v}{\partial x^{i_{2}} \partial x^{i_{3}}}
+3T^{i_{1}i_{2}i_{3}} \frac{\partial^{2}u}{\partial x^{i_{1}} \partial x^{i_{2}}} \parrr{v}{i_{3}}.
\label{eq:difference}
\end{eqnarray}
On the other hand,
$$(\delta D)(u, v)=-3T^{i_{1}i_{2}i_{3}}\parrr{u}{i_{1}}\frac{\partial^{2}v}{\partial x^{i_{2}}
\partial x^{i_{3}}}-3T^{i_{1}i_{2}i_{3}} \frac{\partial^{2}u}{\partial x^{i_{1}} \partial x^{i_{2}}} \parrr{v}{i_{3}}+
\tilde{H}(u, v),$$
where $\tilde{H}(u, v)$ is a bidifferential operator of maximum order $1$ in each argument.
Therefore,
$$ Q_{2}(u, v)-P_{\nabla}^{2}(u, v)=P_{\tilde{\nabla}}^{2}(u, v)-P_{\nabla}^{2}(u, v)+(\delta D)(u, v)$$
is clearly a bidifferential operator of maximum order $1$ in each argument.
To see that such a connection is unique, it
suffices to note that $P_{\tilde{\nabla}}^{2}(u, v)-P_{\nabla}^{2}(u, v)$
is a bidifferential operator of maximum order $1$ iff
$\tilde{\nabla}=\nabla$. This can be easily seen from Equation (\ref{eq:difference}). \qed
\begin{thebibliography}{99}
\bibitem{AM}
Abraham, R. and Marsden, J., Foundations of Mechanics,
Addison-Wesley, 2nd edition, 1985.
\bibitem{AC1}
Arnal, D. and Cortet, J. C.,
{\em J. Geom. Phys.} {\bf 2} (1985), 83-116.
\bibitem{AC2}
Arnal, D. and Cortet, J. C., Repr\'esentations $*$ des groupes
exponentiels,
{\em J. Funct. Anal. } {\bf 92} (1990), 103-135.
\bibitem{Ast}
Astashkevich, A., On Fedosov's quantization of semisimple coadjoint
orbits, Ph.D thesis (MIT), (1996).
\bibitem{BFFLS}
Bayen, F., Flato, M., Fronsdal, C., Lichnerowicz, A., and
Sternheimer, D., Deformation theory and quantization, {I} and {II},
{\em Ann. Phys.} {\bf 111} (1977), 61--151.
\bibitem{B}
Berezin, F.A.,
{\em Funktzionalnyi analiz i ego prilozheniya}
{\bf 1} (1967), 1-14.
\bibitem{BCG}
Bertelson, M., Cahen, M., and Gutt, S., Equivalence of star products,
{\em Classical and Quantum Gravity} to appear.
\bibitem{DeL}
De Wilde, M. and Lecomte, P., Existence of star-products and of
formal deformations of Poisson Lie algebra of arbitrary
symplectic manifolds,
{\em Lett. Math. Phys.} {\bf 7}
(1983), 487--496.
\bibitem{EW}
Emmrich, C., and Weinstein, A., The differential geometry of Fedosov's
quantization, {\em
Lie theory and geometry, in honor of B. Kostant},
Progress in Math. {\bf 123}, Birkhauser , New York, 1994.
\bibitem{Deligne}
Deligne, P., D\'{e}formations de l'alg\`{e}bre des fonctions d'une
vari\'{e}t\'{e} symplectique: Comparaison entre Fedosov et De Wilde, Lecomte,
{\em Selecta Mathematica, New Series}
{\bf 1 No. 4} (1995).
\bibitem{F1}
Fedosov, B., A simple geometrical construction of deformation quantization,
{\em J. Diff. Geom.}
{\bf 40} (1994), 213-238.
\bibitem{F2}
Fedosov, B., Deformation quantization and index
theory,
{\em Mathematical Topics} {\bf 9},
Akademie Verlag, Berlin, 1996.
\bibitem{F3}
Fedosov, B., Reduction and eigenstates in
deformation quantization,
{\em Pseudo-differential calculus and
mathematical physics},
Math. Top., {\bf 5}, Akademie Verlag,
Berlin, (1994), 277--297.
\bibitem{FS} Flato, M. and Sternheimer, D.,
Closedness of star products and cohomologies, {\em
Lie theory and geometry, in honor of B. Kostant},
Progress in Math. {\bf 123}, Birkhauser , New York, (1994), 241--259.
\bibitem{GV} V. Guillemin, Star products on compact pre-quantizable
symplectic manifolds,
{\em lett. Math. Phys.} {\bf 35} (1995),
85-89.
\bibitem{Howe}
Howe, R., Dual pairs in physics: harmonic oscillators,
photons, electrons, and singletons,
{\em lectures in Appl. Math.} {\bf 21} (1985), 179-207.
\bibitem{Lich1} Lichnerowicz, A., Les vari{\'e}t{\'e}s de Poisson
et leurs alg{\`e}bres de Lie associ{\'e}es, {\em J. Diff. Geom.}
{\bf 12} (1977), 253-300.
\bibitem{L}
Lichnerowicz, A., D\'eformations d'alg\'ebres associ\'ees \`a une vari\'et\'e
symplectique (les $*_{\nu}$-produits),
{\em Ann. Inst. Fourier, Grenoble},
{\bf 32 N.1} (1982), 157-209.
\bibitem{Lu}
Lu, J.-H., Moment maps at the quantum level, preprint.
\bibitem{NT1} Nest, R. and Tsygan, B., Algebraic index theorem for families,
{\em Advances in Math. } {\bf 113} (1995), 151-205.
\bibitem{OMY} Omori, H. Maeda, Y. and Yoshioka, A., Weyl manifolds
and deformation quantization,
{\em Advances in Math.} {\bf 85} (1991), 224-255.
\bibitem{V} Vey, J., D\'eformation du crochet de Poisson sur une vari\'et\'e symplectique,
{\em Comment. Math. Helv.} {\bf 50} (1975), 421-454.
\bibitem{Weinstein:1977}
Weinstein, A., Lectures on symplectic manifolds,
{\em CBMS Reg. Conf. Series in Math.} {\bf Vol 29}
AMS, 1977
\bibitem{Weinstein:1983}
Weinstein, A. The local structure of {P}oisson manifolds,
{\em J. Diff. Geom.} {\bf 18} (1983), 523--557.
\bibitem{Weinstein:1995} A. Weinstein, Deformation quantization,
S\'eminaie Bourbaki, 46\`eme ann\'ee, N. 789 (1993-1994),
{\em Asterisque} {\bf 227} (1995), 389-409.
\bibitem{Weinstein:1996} Weinstein, A., Private communication.
\bibitem{WeinsteinX:1996} Weinstein, A. and Xu, P., Work in progress.
\bibitem{Xu:1991}
Xu, P., Morita equivalence of {P}oisson manifolds,
{\em Commun. Math. Phys.} {\bf 142} (1991), 493--509.
\end{thebibliography}
\end{document}