Formal Geometric Quantization III, Functoriality in the spin-c setting
Extracted source text · arXiv 1704.06034
\documentclass[leqno,11pt]{article}
\usepackage{amssymb,amsmath,mathabx}
\usepackage{pgf,ifpdf,graphics,xcolor}
%\usepackage{latexsym}
\usepackage{amsfonts,amssymb,euscript,epsfig,color}
\usepackage[all]{xy}
%\renewcommand{\baselinestretch}{1.4}
%%\usepackage{MnSymbol}
%%%%%%%%%%%%%%%%% THEO %%%%%%%%%%%%%%%%
%\newcommand{\numero}{\setcounter{equation}{0}}
\newtheorem{theo}{Theorem}[section]
\newtheorem{defi}[theo]{Definition}
\newtheorem{defiprop}[theo]{Definition-Proposition}
\newtheorem{assum}[theo]{Assumption}
\newtheorem{lem}[theo]{Lemma}
\newtheorem{prop}[theo]{Proposition}
\newtheorem{rem}[theo]{Remark}
\newtheorem{coro}[theo]{Corollary}
\newtheorem{exam}[theo]{Example}
\newtheorem{conj}[theo]{Conjecture}
\renewcommand{\theequation}{\thesection.\arabic{equation}}
\newenvironment{proof}{{\bf Proof.}}
%%%%%%%%%%%%%%% SYMBOLES %%%%%%%%%%%%%%%%
%%%%%% GOTHIQUE %%%%%%%%%%%%%%%%%%%%%
\newcommand{\agot}{\ensuremath{\mathfrak{a}}}
\newcommand{\ggot}{\ensuremath{\mathfrak{g}}}
\newcommand{\hgot}{\ensuremath{\mathfrak{h}}}
\newcommand{\kgot}{\ensuremath{\mathfrak{k}}}
\newcommand{\lgot}{\ensuremath{\mathfrak{l}}}
\newcommand{\ngot}{\ensuremath{\mathfrak{n}}}
\newcommand{\glgot}{\ensuremath{\mathfrak{gl}}}
\newcommand{\rgot}{\ensuremath{\mathfrak{r}}}
\newcommand{\pgot}{\ensuremath{\mathfrak{p}}}
\newcommand{\qgot}{\ensuremath{\mathfrak{q}}}
\newcommand{\slgot}{\ensuremath{\mathfrak{sl}}}
\newcommand{\sgot}{\ensuremath{\mathfrak{s}}}
\newcommand{\tgot}{\ensuremath{\mathfrak{t}}}
\newcommand{\ugot}{\ensuremath{\mathfrak{u}}}
\newcommand{\zgot}{\ensuremath{\mathfrak{z}}}
\newcommand{\Rgot}{\ensuremath{\mathfrak{R}}}
%%%%%%%%%%%%%% CALLIGRAPHIQUE %%%%%%%%%%%
\newcommand{\lcal}{\ensuremath{\mathcal{l}}}
\newcommand{\Acal}{\ensuremath{\mathcal{A}}}
\newcommand{\Bcal}{\ensuremath{\mathcal{B}}}
\newcommand{\Ccal}{\ensuremath{\mathcal{C}}}
\newcommand{\Dcal}{\ensuremath{\mathcal{D}}}
\newcommand{\Ecal}{\ensuremath{\mathcal{E}}}
\newcommand{\Fcal}{\ensuremath{\mathcal{F}}}
\newcommand{\Hcal}{\ensuremath{\mathcal{H}}}
\newcommand{\Kcal}{\ensuremath{\mathcal{K}}}
\newcommand{\Lcal}{\ensuremath{\mathcal{L}}}
\newcommand{\Mcal}{\ensuremath{\mathcal{M}}}
\newcommand{\Ncal}{\ensuremath{\mathcal{N}}}
\newcommand{\Ocal}{\ensuremath{\mathcal{O}}}
\newcommand{\Pcal}{\ensuremath{\mathcal{P}}}
\newcommand{\Qcal}{\ensuremath{\mathcal{Q}}}
\newcommand{\Scal}{\ensuremath{\mathcal{S}}}
\newcommand{\Ucal}{\ensuremath{\mathcal{U}}}
\newcommand{\Tcal}{\ensuremath{\mathcal{T}}}
\newcommand{\Vcal}{\ensuremath{\mathcal{V}}}
\newcommand{\Wcal}{\ensuremath{\mathcal{W}}}
\newcommand{\Xcal}{\ensuremath{\mathcal{X}}}
\newcommand{\Ycal}{\ensuremath{\mathcal{Y}}}
\newcommand{\Zcal}{\ensuremath{\mathcal{Z}}}
\newcommand{\ocal}{\ensuremath{\mathcal{o}}}
%%%%%%%%%%%%%% corps %%%%%%%%%%%%%%%%%%%%%%%
\newcommand{\Pbb}{\ensuremath{\mathbb{P}}}
\newcommand{\Z}{\ensuremath{\mathbb{Z}}}
\newcommand{\C}{\ensuremath{\mathbb{C}}}
\newcommand{\Q}{\ensuremath{\mathbb{Q}}}
\newcommand{\R}{\ensuremath{\mathbb{R}}}
\newcommand{\N}{\ensuremath{\mathbb{N}}}
\newcommand{\tore}{\ensuremath{\mathbb{T}}}
\newcommand{\Lbb}{\ensuremath{\mathbb{L}}}
%%%%%%%%%%%%%%%% DIVERS %%%%%%%%%%%%%%%%%
\newcommand{\mm}{\ensuremath{\hbox{\rm m}}}
\newcommand{\trc}{\ensuremath{{\bf Tr^{\C}}}}
\newcommand{\tr}{\ensuremath{\hbox{\bf Tr}}}
\newcommand{\ntr}{\ensuremath{\hbox{\bf nTr}}}
\newcommand{\quant}{\ensuremath{\bf q}}
\newcommand{\esp}{\ensuremath{\varepsilon}}
\newcommand{\f}{\ensuremath{\mathcal{C}^{\infty}}}
\newcommand{\fgene}{\ensuremath{\mathcal{C}^{-\infty}}}
\newcommand{\croc}{\ensuremath{\hookrightarrow}}
\newcommand{\im}{\ensuremath{\imath}}
\newcommand{\indice}{\ensuremath{\hbox{\rm Index}}}
\newcommand{\image}{\ensuremath{\hbox{\rm Image}}}
\newcommand{\loc}{\ensuremath{\hbox{\rm P}}}
\newcommand{\p}{\ensuremath{\Delta}}
\newcommand{\pf}{\ensuremath{\overrightarrow{\Delta}}}
\newcommand{\Cr}{\ensuremath{\hbox{\rm Cr}}}
\newcommand{\Vol}{\ensuremath{\hbox{\rm vol}}}
\newcommand{\Eul}{\ensuremath{\hbox{\rm Eul}}}
\newcommand{\Char}{\ensuremath{\hbox{\rm Char}}}
\newcommand{\End}{\ensuremath{\hbox{\rm End}}}
\newcommand{\Hom}{\ensuremath{\hbox{\rm Hom}}}
\newcommand{\Thom}{\ensuremath{\hbox{\rm Thom}}}
\newcommand{\sthom}{\ensuremath{\hbox{\rm S-Thom}}}
\newcommand{\qfor}{\ensuremath{\mathcal{Q}^{-\infty}}}
\newcommand{\Rfor}{\ensuremath{R^{-\infty}}}
\newcommand{\Rforc}{\ensuremath{R^{-\infty}_{tc}}}
\newcommand{\GL}{\ensuremath{GL}}
\newcommand{\SL}{\ensuremath{SL}}
\newcommand{\spinq}{\ensuremath{\Qcal_{\hbox{\tiny \rm spin}}}}
\newcommand{\Sym}{\ensuremath{{\rm Sym}}}
\newcommand{\QS}{\ensuremath{\mathrm{Q}^{\mathrm{spin}}}}
\newcommand{\spin}{\ensuremath{\rm Spin}}
\newcommand{\spinc}{\ensuremath{{\rm spin}^{c}}}
\newcommand{\Qrm}{\ensuremath{\mathrm{Q}}}
\newcommand{\CW}{{\mathfrak t^*_{\geq 0}}}
\def \K {{\rm\bf K}}
\def \wK {\widehat{K}}
\def \T {{\rm T}}
\def \h {{\rm h}}
\def \what {\widehat}
\def \wtde {\widetilde}
\def \indr {{\rm Ind}^{^K}_{_{K_{\rho^{K_\sigma}}}}}
\def \indB {{\rm Ind}^{^K}_{_{K_{\beta}}}}
\def \indHa {{\rm Ind}^{^H}_{_{H_a}}}
\def \inds {{\rm Ind}^{^G}_{_{G_{\sigma}}}}
\def \ind {{\rm Ind}^{^K}_{_{T}}}
\def \carre {{\rm L}^{2}}
\def \HolT {{\rm Hol}^{^K}_{_T}}
\def \HolB {{\rm Hol}^{^K}_{_{K_{\beta}}}}
\def \HolTB {{\rm Hol}^{^{K_{\beta}}}_{_T}}
\def \Hols {{\rm Hol}^{^K}_{_{K_{\sigma}}}}
\def \HolHs {{\rm Hol}^{^{K_{\sigma}}}_{_T}}
\def \HolS {{\rm Hol}^{^G}_{_{S^1}}}
\def \sind {\hbox{\rm S-Ind}^{^K}_{_T}}
\def \tsind {\hbox{\rm S-Ind}^{^{\tilde{K}}}_{_{\tilde{T}}}}
%%%%%%%%%%%%%%
\def \so {{\rm SO}}
\def \u {{\rm U}}
\def \ad {{\rm ad}}
\def \Det {{\rm Det}}
\def \Pspin {{\rm P}_{{\rm Spin}^{\rm c}}}
\def \Pso {{\rm P}_{{\rm SO}}}
\def \Pu {{\rm P}_{{\rm U}}}
\def \clif {{\bf c}}
\def \Clif {{\rm Cl}}
%%%%%%%%%%%%%% table %%%%%%%
%\setcounter{tocdepth}{1}
\begin{document}
\title{Formal Geometric Quantization III \\ Functoriality in the $\spinc$ setting}
\author{Paul-Emile PARADAN
\footnote{Institut Montpelli\'erain Alexander Grothendieck,
Universit\'e de Montpellier, CNRS \texttt{paul-emile.paradan@umontpellier.fr}}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\maketitle
{\small
\tableofcontents}
%%%%%%%%%%%%%%%%%%%
\section{Introduction}
%%%%%%%%%%%%%%%%%%%
Let $(M,\Scal)$ be a $K$-manifold of even dimension, oriented, and equipped with a $K$-equivariant $\spinc$ bundle $\Scal$.
The orientation induces a decomposition $\Scal=\Scal^+\oplus \Scal^-$, and the corresponding $\spinc$ Dirac operator
is a first order elliptic operator $\Dcal_\Scal: \Gamma(M,\Scal^+)\to \Gamma(M,\Scal^-)$ \cite{Atiyah-Singer-2,B-G-V,Duistermaat96}.
When $M$ is compact, an important invariant is the equivariant index $\Qcal_K(M,\Scal)\in R(K)$ of the operator $\Dcal_\Scal$, that can be understood as the quantization of the data $(M,\Scal,K)$ \cite{Atiyah-Hirzebruch70,Hattori78}.
The determinant line bundle of the $\spinc$-bundle $\Scal$ is the line bundle $\det(\Scal)\to M$ defined by the relation
$$
\det(\Scal):=\hom_{\mathrm{Cl}}(\overline{\mathcal{S}},\mathcal{S})
$$
where $\overline{\mathcal{S}}$ is the Clifford module with opposite complex structure (see \cite{pep-vergne-acta}).
The choice of an invariant Hermitian connection $\nabla$ on $\det(\Scal)$ determines an equivariant map $\Phi_{\Scal}: M\to \kgot^{*}$ and a 2-form
$\Omega_\Scal$ on $M$ by means of the Kostant relations
\begin{equation}\label{eq:kostant-L}
\Lcal(X)-\nabla_{X_M}=2i \langle\Phi_\Scal,X\rangle\quad \mathrm{and} \quad \nabla^2= -2i \Omega_\Scal
\end{equation}
for every $X\in\kgot$. Here $\Lcal(X)$ denotes the infinitesimal action on the sections of $\det(\Scal)$.
We say that $\Phi_\Scal$ is the {\em moment map} for $\Scal$ (it depends however of the choice of a connection).
Assume now that $M$ is non-compact but that the moment map $\Phi_\Scal$ is {\em proper}. In this case the formal geometric
quantization of $(M,\Scal,K)$ is well-defined:
$$
\qfor_K(M,\Scal)\in \hat{R}(K)
$$
as an index localized on the zeros of the Kirwan vector field \cite{pep-ENS,pep-pacific,Ma-Zhang14,Hochs-Song-duke}. We will explain the construction in section \ref{sec:QR-non-compact}.
Consider now a closed connected subgroup $H\subset K$. Let ${\rm p}:\kgot^*\to \hgot^*$ be the canonical projection. The map
${\rm p}\circ \Phi_\Scal$ corresponds to the moment map for $\Scal$ relative to the $H$-action.
The main result of our note is the following
\begin{theo}\label{theo-principal-introduction}
Suppose that ${\rm p}\circ \Phi_\Scal$ is proper. Then the following holds:
\begin{itemize}
\item The $K$-module $\qfor_K(M,\Scal)$ is $H$-admissible.
\item We have $\qfor_K(M,\Scal)\vert_H=\qfor_H(M,\Scal)$.
\end{itemize}
\end{theo}
We obtained a similar result in the symplectic setting in \cite{pep-IF}. Here our method uses the compactifications of reductive groups \`a la
de Concini-Procesi and the multiplicative property of the functor $\qfor_K$ that has been proved recently by
Hochs and Song \cite{Hochs-Song-duke}.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{center}
\bf Notations
\end{center}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
Throughout the paper :
\begin{itemize}
\item $K$ denotes a compact connected Lie group with Lie algebra $\kgot$.
\item We denote by $R(K)$ the representation ring of $K$ : an element $E\in R(K)$ can be represented
as a finite sum $E=\sum_{\mu\in\what{K}}\mm_\mu \pi_\mu$, with $\mm_\mu\in\Z$. The multiplicity
of the trivial representation is denoted $[E]^K$.
\item We denote by $\hat R(K)$ the space of $\Z$-valued functions on $\widehat{K}$. An element $E\in
\hat R(K)$ can be represented
as an infinite sum $E=\sum_{\mu\in\what{K}}\mm(\mu) \pi_\mu$, with $\mm(\mu)\in\Z$.
\item An element $\xi\in \kgot^*$ is called regular if the stabilizer subgroup $K_\xi:=\{k\in K, k\cdot\xi=\xi\}$ is a maximal torus of $K$.
\item When $K$ acts on a manifold $M$, we denote $X_M(m):=\frac{d}{dt}\vert_{t=0} e^{-tX}\cdot m$
the vector field generated by $-X\in \kgot$. Sometimes we will also use the notation $X_M(m)=-X\cdot m$.
The set of zeroes of the vector field $X_M$ is denoted $M^X$.
\end{itemize}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{The $[Q,R]=0$ theorem in the $\spinc$ setting}\label{sec:QR-compact}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
In this section we suppose that $M$ is compact and we recall the results of \cite{pep-vergne-acta} concerning the multiplicities of $\Qcal_K(M,\Scal)\in R(K)$.
For any $\xi\in\kgot^*$, we denote $\kgot_\xi$ the Lie algebra of the stabilizer subgroup $K_\xi$. A coadjoint
orbit $\Pcal=K\eta$ is of type $(\kgot_\xi)$ if the conjugacy classes $(\kgot_\eta)$ and $(\kgot_\xi)$ are equal.
\begin{defi}
Let $(\kgot_M)$ be the generic infinitesimal stabilizer for the $K$-action on $M$. We says that the $K$-action on $M$ is {\em nice} if there exists $\xi\in\kgot^*$ such that
\begin{equation}\label{eq:nice}
([\kgot_M,\kgot_M])=([\kgot_\xi,\kgot_\xi]).
\end{equation}
\end{defi}
The first result of \cite{pep-vergne-acta} is the following
\begin{theo}
If the $K$-action on $M$ is not nice, then $\Qcal_K(M,\Scal)=0$ for any $\spinc$ bundle $\Scal$.
\end{theo}
We suppose now that the $K$-action on $M$ is nice. The conjugacy class $(\kgot_\xi)$ satisfying (\ref{eq:nice}) is unique and is denoted $(\hgot_M)$ (see Lemma 3 in \cite{pep-vergne:magic}).
\begin{defi}A coadjoint orbit $\Pcal\subset \kgot^*$ is admissible if $\Pcal$ carries a $\spinc$-bundle $\Scal_\Pcal$ such that
the corresponding moment map is the inclusion $\Pcal\croc \kgot^*$. We denote simply by
$\QS_K(\Pcal)$ the element $\Qcal_K(\Pcal,\Scal_\Pcal)\in R(K)$.
\end{defi}
We can check easily \cite{pep-vergne:magic} that $\QS_K(\Pcal)$ is either $0$ or an irreducible representation
of $K$, and that the map
$$
\Ocal\mapsto \pi_\Ocal^K:=\QS_K(\Ocal)
$$
defines a bijection between the regular admissible orbits and the dual $\widehat{K}$.
When $\Ocal$ is a regular admissible orbit, an admissible coadjoint orbit $\Pcal$
is called an ancestor of $\Ocal$ (or a $K$-ancestor of $\pi^K_\Ocal$)
if $\QS_K(\Pcal)=\pi^K_\Ocal$.
\medskip
Denote by $\Acal((\hgot_M))$ the set of admissible orbits of type $(\hgot_M)$. The following important fact is proved in section 5
of \cite{pep-vergne-acta}.
\begin{prop}
Let $\Pcal\in \Acal((\hgot_M))$.
\begin{itemize}
\item If $\Pcal$ belongs to the set of regular values of $\Phi_\Scal$, the reduced space
$$
M_\Pcal=\Phi^{-1}_\Scal(\Pcal)/K
$$
is an oriented orbifold equipped with a $\spinc$ bundle. The index of the corresponding Dirac operator on the orbifold $M_\Pcal$ is denoted
$\QS(M_\Pcal)\in \Z$ \cite{Kawasaki81}.
\item In general, the $\spinc$ index $\QS(M_\Pcal)\in \Z$ associated to the (possibly singular) reduced space $M_\Pcal$ is defined
by a deformation procedure.
\end{itemize}
\end{prop}
The $[Q,R]=0$ Theorem in the $\spinc$ setting takes the following form.
\begin{theo}[\cite{pep-vergne-acta}]\label{theo:QR-compact}
Let $\Ocal$ be a regular admissible orbit.
The multiplicity of the representation $\pi^K_\Ocal$ in $\Qcal_K(M,\Scal)$ is equal to
$$
\sum_{\Pcal} \QS(M_{\Pcal})
$$
where the sum runs over the ancestors of $\Ocal$ of type $(\hgot_M)$. In other words
$$
\Qcal_K(M,\Scal)=\sum_{\Pcal\in \Acal((\hgot_M))}\QS(M_{\Pcal})\,\QS_K(\Pcal).
$$
\end{theo}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Formal geometric quantization in the $\spinc$ setting}\label{sec:QR-non-compact}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
In this section the manifold $M$ is not necessarily compact, but the moment map $\Phi_\Scal$ is supposed to be proper.
We choose an invariant scalar product in $\kgot^*$ that provides an identification
$\kgot\simeq\kgot^*$.
\begin{defi}\label{defi:kir}
$\bullet$ The {\em Kirwan vector field} associated to $\Phi_\Scal$ is defined by
\begin{equation}\label{eq-kappa}
\kappa_{\Scal}(m)= -\Phi_\Scal(m)\cdot m, \quad m\in M.
\end{equation}
$\bullet$ We denote by $Z_\Scal$ the set of zeroes of $\kappa_{\Scal}$. Thus $Z_\Scal$ is a $K$-invariant closed subset of $M$.
\end{defi}
The set $Z_\Scal$, which is not necessarily smooth, admits an easy description. Choose a Weyl chamber $\mathfrak{t}^*_+\subset \mathfrak{t}^*$ in the dual
of the Lie algebra of a maximal torus $T$ of $K$. We see that
\begin{equation}\label{eq=Z-Phi-beta}
Z_\Scal=\coprod_{\beta\in\Bcal_\Scal} Z_\beta
\end{equation}
where $Z_\beta$ corresponds to the compact set $K(M^\beta\cap\Phi_\Scal^{-1}(\beta))$, and $\Bcal_\Scal=\Phi_\Scal(Z_\Scal)\cap \tgot^*_+$.
The properness of $\Phi_\Scal$ insures that for any compact subset $C\subset \tgot^*$ the intersection $\Bcal_\Scal\cap C$ is finite.
The principal symbol of the Dirac operator $D_\Scal$ is the bundle map
$\sigma(M,\Scal)\in \Gamma(\T^* M, \hom(\Scal^+,\Scal^-))$ defined by the Clifford action
$$\sigma(M,\Scal)(m,\nu)=\clif_{m}(\tilde{\nu}): \Scal\vert_m^+\to \Scal\vert_m^-.$$
where $\nu\in \T^* M\simeq \tilde{\nu}\in \T M$ is an identification associated to an invariant Riemannian metric on $M$.
\begin{defi}\label{def:pushed-sigma}
The symbol $\sigma(M,\Scal,\Phi_\Scal)$ shifted by the vector field $\kappa_{\Scal}$ is the
symbol on $M$ defined by
$$
\sigma(M,\Scal,\Phi_\Scal)(m,\nu)=\sigma(M,\Scal)(m,\tilde{\nu}-\kappa_\Scal(m))
$$
for any $(m,\nu)\in\T^* M$.
\end{defi}
For any $K$-invariant open subset $\Ucal\subset M$ such that $\Ucal\cap Z_\Scal$ is compact in $M$, we see that the restriction
$\sigma(M,\Scal,\Phi_\Scal)\vert_\Ucal$ is a transversally elliptic symbol on $\Ucal$, and so its equivariant index is a well defined element in
$\hat{R}(K)$ (see \cite{Atiyah74,pep-vergne:witten}).
Thus we can define the following localized equivariant indices.
\begin{defi}\label{def:indice-localise}
\begin{itemize}
\item A closed invariant subset $Z\subset Z_\Scal$ is called a {\em component} of $Z_\Scal$ if it is a union of connected components of $Z_\Scal$.
\item If $Z$ is a {\em compact component} of $Z_\Scal$, we denote by
$$
\Qcal_K(M,\Scal,Z)\ \in\ \hat{R}(K)
$$
the equivariant index of $\sigma(M,\Scal,\Phi)\vert_\Ucal$ where $\Ucal$ is an invariant neighbourhood of $Z$
so that $\Ucal\cap Z_\Scal=Z$.
\end{itemize}
\end{defi}
By definition, $Z=\emptyset$ is a component of $Z_\Scal$ and $\Qcal_K(M,\Scal,\emptyset)=0$. For any $\beta\in\Bcal_\Scal$, $Z_\beta$ is a
compact component of $Z_\Scal$.
When the manifold $M$ is compact, the set $\Bcal_\Scal$ is finite and we have the decomposition
$$
\Qcal_K(M,\Scal)=\sum_{\beta\in\Bcal_\Scal}\Qcal_K(M,\Scal,Z_\beta).
$$
See \cite{pep-RR,pep-vergne:witten}.
When the manifold $M$ is not compact, but the moment map $\Phi_\Scal$ is proper, we can defined
$$
\qfor_K(M,\Scal):=\sum_{\beta\in\Bcal_\Scal}\Qcal_K(M,\Scal,Z_\beta)
$$
The sum of the right hand side is not necessarily finite but it converges in $\hat{R}(K)$ (see \cite{pep-pacific,Ma-Zhang14,Hochs-Song-duke}).
We call $\qfor_K(M,\Scal)\in \hat{R}(K)$ the formal geometric quantization of the data $(M,\Scal,\Phi_\Scal, K)$.
Hochs and Song prove the following important property concerning the functoriality of $\qfor_K$ relatively to the product of manifolds.
\medskip
\begin{theo}[\cite{Hochs-Song-duke}]\label{theo:hochs-song}
Let $(M,\Scal)$ be a $\spinc$ $K$-manifold with a proper moment map $\Phi_\Scal$.
Let $(P,\Scal_P)$ be a compact $\spinc$ $K$-manifold (even dimensional and oriented). Then the $\spinc$ manifold
$(M\times P,\Scal\boxtimes \Scal_P)$ admits a proper moment map and we have the following equality in $\hat{R}(K)$:
$$
\qfor_K(M\times P,\Scal\boxtimes \Scal_P) = \qfor_K(M,\Scal)\otimes \Qcal_K(P,\Scal_P).
$$
\end{theo}
\medskip
With Theorem \ref{theo:hochs-song} in hand we can compute the multiplicities of $\qfor_K(M,\Scal)$ like in the compact setting by using the shifting trick.
Let $\Ocal$ be an admissible regular orbit of $K$. We denote by $[\pi^K_\Ocal:\qfor_K(M,\Scal)]$ the multiplicity of $\pi^K_\Ocal$ in
$\qfor_K(M,\Scal)\in \hat{R}(K)$. Let $\Ocal^*$ be the admissible orbit $-\Ocal$: we have $\Qcal_K(\Ocal^*,\Scal_{\Ocal^*})=(\pi^K_\Ocal)^*$.
Thanks to Theorem \ref{theo:hochs-song} we get
\begin{eqnarray*}
[\pi^K_\Ocal:\qfor_K(M,\Scal)]&=&\left[\qfor_K(M,\Scal)\otimes \Qcal_K(\Ocal^*,\Scal_{\Ocal^*})\right]^K\\
&=& \left[\qfor_K(M\times \Ocal^*,\Scal\otimes\Scal_{\Ocal^*})\right]^K.
\end{eqnarray*}
We consider the product $M\times \Ocal^*$ equipped with the $\spinc$-bundle $\Scal\otimes\Scal_{\Ocal^*}$.
The corresponding moment map is $\Phi_{\Scal\otimes\Scal_{\Ocal^*}}(m,\xi)=\Phi_\Scal(m)+\xi$.
We use the simplified notation $\Phi_\Ocal$ for $\Phi_{\Scal\otimes\Scal_{\Ocal^*}}$, $\kappa_\Ocal$ for the
corresponding Kirwan vector field on $M\times \Ocal^*$, and $Z_\Ocal:=\{\kappa_\Ocal=0\}$.
In \cite{pep-vergne-acta}, we introduced a locally constant function $d_\Ocal: Z_\Ocal\to \R$, and we denote
$Z_\Ocal^{=0}=\{d_\Ocal=0\}$. Using the localization\footnote{In \cite{pep-vergne-acta} we work in the compact setting, but
exactly the same proof works in the non compact setting, as noticed by Hochs and Song \cite{Hochs-Song-duke}.} done in \cite{pep-vergne-acta}[section 4.5], we get that
\begin{equation}\label{eq:Z-Ocal-0}
[\pi^K_\Ocal:\qfor_K(M,\Scal)]=
\left[\qfor_K(M\times \Ocal^*,\Scal\otimes\Scal_{\Ocal^*},Z_\Ocal^{=0})\right]^K
\end{equation}
Finally we obtain the same result like in the compact setting:
\begin{itemize}
\item If the $K$ action on $M$ is not nice, $Z_\Ocal^{=0}=\emptyset$ and then the multiplicity
$[\pi^K_\Ocal:\qfor_K(M,\Scal)]^K$ vanishes for any
regular admissible orbit $\Ocal$.
\item If the $K$ action on $M$ is nice, we have
$[\pi^K_\Ocal:\qfor_K(M,\Scal)]^K=\sum_{\Pcal} \QS(M_{\Pcal})$ where the sum runs over the ancestors of $\Ocal$ of type $(\hgot_M)$.
\end{itemize}
In other words,
\begin{itemize}
\item if the $K$ action on $M$ is not nice, then $\qfor_K(M,\Scal)=0$,
\item if the $K$ action on $M$ is nice, we have
$$
\qfor_K(M,\Scal)=\sum_{\Pcal\in \Acal((\hgot_M))}\QS(M_{\Pcal})\,\QS_K(\Pcal).
$$
\end{itemize}
\begin{rem}\label{rem:abelian-stabilizer}
We will use a particular case of identity (\ref{eq:Z-Ocal-0}) when the generic infinitesimal stabilizer of the $K$-action on $M$ is {\em abelian}, i.e.
$([\kgot_M,\kgot_M])=0$. In this case $Z_\Ocal^{=0}=\{\Phi_\Ocal=0\}$ and then
\begin{eqnarray*}
[\pi^K_\Ocal:\qfor_K(M,\Scal)]^K&=&
\left[\qfor_K(M\times \Ocal^*,\Scal\otimes\Scal_{\Ocal^*},\{\Phi_\Ocal=0\})\right]^K\\
&=& \QS(M_{\Ocal}).
\end{eqnarray*}
\end{rem}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\section{Functoriality relatively to a subgroup}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We come back to the setting of $K$-manifold $M$, even dimensional and oriented, equipped with an equivariant $\spinc$ bundle $\Scal$. We suppose that for some choice of connection on $\det(\Scal)$ the corresponding moment map $\Phi_\Scal$ is proper. As explained earlier,
the formal geometric quantization of $(M,\Scal,K)$ is well-defined :
$\qfor_K(M,\Scal)\in \hat{R}(K)$.
Consider now a closed connected subgroup $H\subset K$. Let ${\rm p}:\kgot^*\to \hgot^*$ be the canonical projection. The map
${\rm p}\circ \Phi_\Scal$ correspond to the moment map for the $\spinc$ bundle $\Scal$ relative to the $H$-action.
This section is dedicated to the proof of our main result.
\begin{theo}\label{theo-principal}
Suppose that ${\rm p}\circ \Phi_\Scal$ is proper. Then the $K$-module $\qfor_K(M,\Scal)$ is $H$-admissible, and
\begin{equation}\label{eq:main}
\qfor_K(M,\Scal)\vert_H=\qfor_H(M,\Scal).
\end{equation}
\end{theo}
We start with the following
\begin{lem}\label{lem:simple}
$\bullet$ $\qfor_K(M,\Scal)$ is $H$-admissible when ${\rm p}\circ \Phi_\Scal$ is proper.
$\bullet$ It is sufficient to prove (\ref{eq:main}) for manifolds with abelian generic infinitesimal stabilizers.
\end{lem}
\begin{proof} We have $\qfor_K(M,\Scal)=\sum_\Pcal \QS(M_\Pcal) \QS_K(\Pcal)$ where the sum runs over the admissible orbits of type
$(\hgot_M)$.
Thanks to the $[Q,R]=0$ Theorem we know that $\QS_K(\Pcal)\vert_H= \QS_H(\Pcal)= \sum_{\Pcal'}\QS(\Pcal_{\Pcal'}) \QS_H(\Pcal')$, where
$\Pcal_{\Pcal'}= \Pcal\cap {\rm p}^{-1}(\Pcal')/H$ is the reduction of the $K$-coadjoint orbit $\Pcal$ relatively to $H$-coadjoint orbit $\Pcal'$.
Hence $\qfor_K(M,\Scal)$ is $H$-admissible if for any $H$-coadjoint orbit $\Pcal'$, the sum
$$
\sum_\Pcal \QS(M_\Pcal)\QS(\Pcal_{\Pcal'})
$$
admits only a finite number of non-zero terms. We see that $\QS(\Pcal_{\Pcal'})\neq 0$ only if $\Pcal'\subset{\rm p}(\Pcal)$ and
$\QS(M_\Pcal)\neq 0$ only if $\Phi_\Scal^{-1}(\Pcal)\neq\emptyset$. Finally $\QS(M_\Pcal)\QS(\Pcal_{\Pcal'})\neq 0$ only if
$$
\Pcal\in K\Phi_\Scal\left(({\rm p}\circ \Phi_\Scal)^{-1}(\Pcal')\right).
$$
Since ${\rm p}\circ \Phi_\Scal$ is proper, we have only a finite number of $K$-admissible orbits contained in the compact set
$K\Phi_\Scal\left(({\rm p}\circ \Phi_\Scal)^{-1}(\Pcal')\right)$. The first point is proved.
\medskip
Let us check the second point. Suppose that (\ref{eq:main}) holds for manifolds with abelian generic infinitesimal stabilizers. Let $K\rho$ be the regular
admissible orbit such that $\QS_K(K\rho)$ is the trivial representation.
To any $\spinc$ manifold $(M,\Scal,K)$ with proper moment map $\Phi_\Scal$, we associate the product
$M\times K\rho$ which is a $\spinc$ $K$-manifold with proper moment map $\Phi_{\Scal\boxtimes \Scal_{K\rho}}$.
The multiplicative property (see Theorem \ref{theo:hochs-song}) gives
$$
\qfor_K(M,\Scal)=\qfor_K(M\times K\rho,\Scal\boxtimes \Scal_{K\rho}).
$$
Now we remark that the $K$-manifold $M\times K\rho$ has abelian infinitesimal stabilizers, and that ${\rm p}\circ \Phi_{\Scal\boxtimes \Scal_{K\rho}}$ is proper if
${\rm p}\circ \Phi_\Scal$ is proper. Then, when the moment map ${\rm p}\circ \Phi_\Scal$ is proper, we have
\begin{eqnarray*}
\qfor_K(M,\Scal)\vert_H&=&\qfor_K(M\times K\rho,\Scal\boxtimes \Scal_{K\rho})\vert_H\\
&=& \qfor_H(M\times K\rho,\Scal\boxtimes \Scal_{K\rho}) \hspace{21mm} [1]\\
&=& \qfor_H(M,\Scal)\otimes \Qcal_H(K\rho,\Scal_{K\rho})\hspace{15mm} [2]\\
&=&\qfor_H(M,\Scal).\hspace{42mm} [3]
\end{eqnarray*}
Here we see that $[1]$ is the identity (\ref{eq:main}) applied to $M\times K\rho$, $[2]$ is the multiplicative property relatively to the $H$-action, and
$[3]$ is due to the fact that $\Qcal_H(K\rho,\Scal_{K\rho})$ is the trivial $H$-representation. $\Box$
\end{proof}
%
%\section{Deformation using $\T^* K$}
%
%
%
%
%Let $K$ be a compact Lie group and let $H$ be a closed subgroup.
%Let $i$ denote the inclusion of $H$ into $K$, $i : \hgot \to \kgot$ the induced embedding of Lie algebras,
%and $\pi : \kgot^* \to \hgot^*$ the dual projection.
%
%We consider in this subsection the paradigmatic example of the cotangent space
%$\T^*K$ with the following action of $K\times K$ : $(k,h)\cdot a= k a h^{-1}$.
%
%The tangent bundle $\T K$ is identified with $K\times\tilde{\kgot}$ through the left translations: to
%$(a,X)\in K\times\kgot$ we associate $\frac{d}{dt} e^{tX}a\vert_{t=0}\in \T_a K$.
%The action of $K\times H$ on the cotangent bundle $\T^* K\simeq K\times \kgot^*$ is then
%$$
%(k,h)\cdot( a,\xi)= (kah^{-1},k\xi).
%$$
%
%The Liouville form $\lambda$ of $\T^*K$ is thus given by
%$\lambda_{a,\xi}(X,\eta)=\langle\xi,X\rangle$ for $(a,\xi)\in \T^*K\simeq K\times\kgot^*$ and
%$(X,\eta)\in \T_{(a,\xi)}(\T^*K)\simeq\kgot\times \kgot^*$.
%The symplectic form on $\T^*\tilde{K}$ is $\Omega:=d\lambda$.
%
%
%The corresponding moment map relative to the $K\times H$-action is the
%map $\Phi=\Phi_K \oplus \Phi_H : \T^* K \to \kgot^*\oplus \hgot^*$ defined by
%\begin{equation}\label{eq:momentcotangent}
%\Phi_{K}(a,\xi)=-\xi,\quad \Phi_H(a,\xi)=\pi(a^{-1}\xi).
%\end{equation}
%
%We consider the manifold $N:=\T^*K\times M$ with the $K\times H$ action: $(k,h)\cdot(a,\xi,m)=(kah^{-1},k\xi,km)$. The corresponding moment map is
%$$
%\Phi^N(a,\xi,m)=\left(\Phi_K^M(m)-\xi, \pi(a^{-1}\xi)\right).
%$$
%
%
%
%
%\begin{prop}We have
%$$
%\left[\qfor_{K\times H}(N,\Scal_N)\right]^K= \qfor_H(M,\Scal_M).
%$$
%
%\end{prop}
%
%\begin{proof}
%Let $\Ocal$ be a regular admissible orbit of $H$. Using the multiplicativity property of $\qfor$ we now that that multiplicity of $\pi^H_{\Ocal}$ in $[\qfor_{K\times H}(N,\Scal_N)]^K$ is
%$$
%m_{\Ocal}=\left[\qfor(P_\Ocal,\Scal_\Ocal)\right]^{K\times H}
%$$
%where $P_\Ocal$ is the $K\times H$-manifold $\T^*K\times M\times \overline{\Ocal}$. The moment map
%$\Phi_\Ocal$ is defined by
%$$
%\Phi_\Ocal(p)=(\Phi_M^K(m)-\xi, \pi(a^{-1}\eta)-\eta),\quad {\rm for}\ p=(a,\xi,m,\eta).
%$$
%Let $Z_\Ocal:=\{p\in P\, \vert\, \Phi_P(p)\cdot p=0\}$ be the critical set of the Kirwan vector field, and
%$d_\Ocal: Z_\Ocal\to \R$ the function introduced in \cite{spin2}
%
%\begin{lem}
%We have $d_\Ocal\geq 0$ on $Z_\Pcal$ and $\{d_\Ocal=0\}=\Phi_\Ocal^{-1}(0)$.
%\end{lem}
%The previous Lemma tells us that
%$$
%m_{\Ocal}=\left[\qfor(P_\Ocal,\Scal_\Ocal,\Phi_\Ocal^{-1}(0))\right]^{K\times H}
%$$
%We use then that the reduction $\Phi_\Ocal^{-1}(0)/K\times H$ can be obtain by a reduction in stage. If we start by the $K$-reduction we see that the reduction of $\T^*K\times M$ at $0$ corresponds to $M$ with the $H$-action.
%
%\end{proof}
%
%
%\end{proof}
%
%\section{Second Part}
%
%
%Here $N=\T^*K\times M$, with the $K\times H$-action. We want to prove that
%$$
%\qfor_{K\times H}(N,\Scal_N)= L^2(K)\otimes\qfor_K(M,\Scal_M)
%$$
%where $L^2(K)$ is viewed as a $K\times H$-module.
%
%For $n=(a,\xi,m)$,
%$$
%\T_n N\simeq \kgot\times \kgot\times \T_m M
%$$
%So $(X,Y,v)\in \T_n N$. If we take $(A,B)\in \kgot\times \hgot$, we have
%$$
%(A,B)\cdot n= (A-aB, [A,\xi],A\cdot m)
%$$
%Hence
%$$
%\T^*_{K\times H} N=
%\left\{ (X,Y,v)\in \T_n N\ ,\ X+[\xi,Y]+\varphi_{m,v}=0\ {\rm and}\ a^{-1}X\in\hgot^\perp\right\}
%$$
%where $(\varphi_{m,v}, X)=(v,X\cdot m)$.
%
%The Kirwan vector fields on $N$ is
%$$
%\kappa_n:=\left(\Phi_K(m)-\xi-a\pi(a^{-1}\xi), [\Phi_K(m),\xi], (\Phi_K(m)-\xi)\cdot m\right)
%$$
%
%The index $\qfor_{K\times H}(N,\Scal_N)$ is associated to the following $K\times H$-transversally elliptic symbol
%$$
%\sigma_N\vert_n(X,Y,v)=
%\clif\left(X-\Phi_K(m)+\xi +a\pi_H(a^{-1}\xi)\oplus Y - [\Phi_K(m),\xi]\right)\odot \clif\left(v-(\Phi_K(m)-\xi)\cdot m\right)
%$$
%
%We see that ${\rm support}(\sigma_N)\cap \T^*_{K\times H} N$ is defined by the relations
%\begin{equation}
%\begin{cases}
%1.\ \ X=\Phi_K(m)-\xi -a\pi_H(a^{-1}\xi) \\
%2. \ \ Y = [\Phi_K(m),\xi],\\
%3. \ \ v=(\Phi_K(m)-\xi)\cdot m,\\
%4. \ \ X+[\xi,Y]+\varphi_{m,v}=0,\\
% 5.\ \ a^{-1}X\in \hgot^\perp.
%\end{cases}
%\end{equation}
%
%Condition 1. and 5. give that $\Phi_H(a^{-1}m)=2\pi_H(a^{-1}\xi)$, $a^{-1}X=
%\Phi_{K/H}(a^{-1}X)-\pi_{K/H}(a^{-1}\xi)$, and then
%$$
%\left(X,\Phi_K(m)-\xi\right)=\|\Phi_{K/H}(a^{-1}m)-\pi_{K/H}(a^{-1}\xi)\|^2.
%$$
%Condition 2. gives that
%$$
%\left([\xi,Y],\Phi_K(m)-\xi\right)=\|[\Phi_K(m),\xi]\|^2
%$$
%Condition 3. gives
%$$
%\left(\varphi_{m,v},\Phi_K(m)-\xi\right)=\|(\Phi_K(m)-\xi)\cdot m\|^2
%$$
%Finally condition 4. gives that
%$$
%\|\Phi_{K/H}(a^{-1}m)-\pi_{K/H}(a^{-1}\xi)\|^2+ \|[\Phi_K(m),\xi]\|^2+ \|(\Phi_K(m)-\xi)\cdot m\|^2=0
%$$
%
%We see that $(a,\xi,m,X,Y,v)\in {\rm support}(\sigma_N)\cap \T^*_{K\times H} N$ iff
%\begin{equation}
%\begin{cases}
%1.\ \ X=Y=v=0 \\
%2. \ \ \Phi_K(m)-\xi= \frac{1}{2} \, a\Phi_H(a^{-1}m)\\
%3. \ \ a^{-1}m\in Z_{\Phi_H}.
%\end{cases}
%\end{equation}
%
%Note that $a^{-1}m\in Z_{\Phi_H}\Longrightarrow [\Phi_K(a^{-1}m),\Phi_H(a^{-1}m)]=0$.
%
%Note also that $\sigma_N$ is homotopic to the following $K\times H$-transversally elliptic symbol
%$$
%\sigma'_N=
%\clif\left(X-\Phi_K(m)+\xi +a\pi_H(a^{-1}\xi)\oplus Y \right)\, \odot\, \clif\left(v-(\Phi_K(m)-\xi)\cdot m\right)
%$$
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{De Concini-Procesi compactifications}\label{sec:Concini-Procesi}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
We recall that $T$ is a maximal torus of the compact connected Lie group $K$, and $W$ is the corresponding Weyl group.
We define a {\em $K$-adapted} polytope in $\tgot^*$ to be a $W$-invariant Delzant polytope $P$ in $\tgot^*$
whose vertices are regular elements of the weight lattice $\Lambda$. If $\{\lambda_1,\ldots,\lambda_r\}$
are the dominant weights lying in the union of all the closed one-dimensional faces of
$P$, then there is a $G \times G$-equivariant embedding of $G= K_\C$ into
$$\Pbb( \bigoplus_{i=1}^r (V^K_{\lambda_i})^* \otimes V^K_{\lambda_i})$$
associating to $g \in G$ its representation on $\bigoplus_{i=1}^r V^K_{\lambda_i}$.
The closure $\Xcal_{P}$ of the image of $G$ in this
projective space is smooth and is equipped with a $K\times K$-action. The restriction of the canonical K\"{a}hler structure on
$\Xcal_P$ defines a symplectic $2$-form $\Omega_{\Xcal_P}$.
We recall briefly the different properties of $(\Xcal_P,\Omega_{\Xcal_P})$ : all the details can be found in \cite{pep-IF}.
\begin{enumerate}
\item[(1)] $\Xcal_P$ is equipped with an Hamiltonian action of $K\times K$. Let
$\Phi:=(\Phi_l,\Phi_r): \Xcal_P\to\kgot^*\times\kgot^*$ be the corresponding moment map.
\item[(2)] The image of $\Phi$ is equal to $\{(k\cdot\xi,-k'\cdot\xi)\ \vert\ \xi\in P\
{\rm and}\ k,k'\in K\}$.
\item[(3)] The Hamiltonian $K\times K$-manifold $(\Xcal_P,\Omega_{\Xcal_P})$ has no multiplicities: the pull-back by $\Phi$ of a
$K\times K$-orbit in the image is a $K\times K$-orbit in $\Xcal_P$.
\item[(4)] The symplectic manifold $ (\Xcal_P,\Omega_{\Xcal_P})$ is prequantized by the restriction
of the hyperplane line bundle $\Ocal(1)\to\Pbb( \oplus_{i=1}^N (V^K_{\lambda_i})^* \otimes V^K_{\lambda_i})$ to $\Xcal_P$:
let us denoted $L_P$ the corresponding $K\times K$-equivariant line bundle.
\end{enumerate}
Let $\Ucal_P:= K\cdot P^\circ$ where $P^\circ$ is the interior of $P$. We define
$$
\Xcal_P^\circ:=\Phi_l^{-1}(\Ucal_P)
$$
which is an invariant, open and dense subset of $\Xcal_P$. We have the following
important property concerning $\Xcal_P^\circ$.
\begin{enumerate}
\item[(5)] There exists an equivariant diffeomorphism $\Upsilon : K\times \Ucal_P\to \Xcal_P^\circ$
such that $\Upsilon^*(\Phi_l)(g,\nu)= g\cdot \nu$ and $\Upsilon^*(\Phi_r)(g,\nu)=-\nu$.
\end{enumerate}
The manifold $\Xcal_P$ is equipped with a family of $\spinc$ bundles
$$
\Scal_P^n:= \bigwedge (\T \Xcal_P)^{1,0}\otimes L_P ^{\otimes n}, \quad n\geq 1,
$$
and we consider the corresponding $K\times K$-modules $\Qcal_{K\times K}(\Xcal_P,\Scal_P^n)$.
Let $\tgot^*_+\subset\tgot^*$ be a Weyl chamber and let
$\Lambda\subset \tgot^*$ be the lattice of weights: we denote by $\rho\in \tgot^*_+$ the half sum of the positive roots.
\begin{prop}\label{prop:Q-X-P}
We have the following decomposition
$$
\Qcal_{K\times K}(\Xcal_P,\Scal_P^n)=\sum_{\Ocal\cap \{n\,P^o+\rho\}\neq \emptyset}\ \pi_\Ocal\otimes \pi_{\Ocal^*} +
\sum_{\Ocal\cap \{n\,\partial P+\rho\}\neq \emptyset}a_{n,\Ocal}\ \pi_\Ocal\otimes \pi_{\Ocal^*}.
$$
\end{prop}
\begin{proof}
The result is a consequence of the Meinrenken-Sjamaar $[Q,R]=0$ theorem \cite{Meinrenken98,Meinrenken-Sjamaar,Tian-Zhang98,pep-RR}.
To explain it, we parametrize the dual $\widehat{K}$ with the highest weights. For any dominant weight
$\alpha\in \Lambda\cap \tgot^*_+$, let $V^K_\alpha$ be the irreducible representation of $K$ with highest weight $\alpha$. In other terms,
$V^K_\alpha=\pi^K_{\alpha+\rho}$.
The symplectic $[Q,R]=0$ theorem tells us that the multiplicity of $V_\alpha^K\otimes V_\gamma^K$ in $\Qcal_{K\times K}(\Xcal_P,\Scal_P^n)$ is equal to the
Riemann-Roch number of the symplectic reduced space $\Phi^{-1}(K\frac{\alpha}{n}\times K\frac{\gamma}{n})/K\times K$.
Points $(2)$ above tell us that $\Phi^{-1}(Ka\times Kb)/K\times K$ is non empty only if $Kb=-Ka$ and $Ka\cap P\neq \emptyset$. With point $(4)$ , we see that
the reduced space $\Phi^{-1}(Ka\times -Ka)/K\times K$ is a (smooth) point if $a\in P^o$. The proof is completed.
\end{proof}
\subsection{Cutting}
Let $M$ be a $K$-manifold of even dimension, oriented, equipped with a $K$-equivariant $\spinc$ bundle $\Scal$. Let $\Phi_\Scal$ be the moment map associated to a hermitian connection on $\det(\Scal)$. We assume that $\Phi_\Scal$ is {\em proper}.
We consider the manifold $\Xcal_P$ equipped with the $\spinc$ bundle
$\Scal_P^n:= \bigwedge (\T \Xcal_P)^{1,0}\otimes L_P ^{\otimes n}$. The determinant line bundle
$\det(\Scal_P^n)$ is equal to $$\det((\T \Xcal_P)^{1,0})\otimes L_P ^{\otimes 2n}.$$
Let $\varphi_l,\varphi_r:\Xcal_P\to\kgot^*$ be the moment maps associated to the action of $K\times K$
on the line bundle $\det((\T \Xcal_P)^{1,0})$. So the moment map
relative to the action of $K\times K$ on $\det(\Scal_P^n)$ is the map $\Phi^n=(\Phi^n_l,\Phi^n_r): \Xcal_P\to\kgot^*\times \kgot^*$
defined by $\Phi^n_l= n\Phi_l +\varphi_l$ and $\Phi^n_r= n\Phi_r +\varphi_r$.
On the dense open subset $\Xcal_P^o\simeq K\times \Ucal_P$ the line bundle $\det((\T \Xcal_P)^{1,0})$ admits a trivialization.
Let $c>0$ such that the closed ball $\{\|\xi\|\leq c\}$ is contained in $\Ucal_P$. By partition of unity, we can choose a connection on
$\det((\T \Xcal_P)^{1,0})$ such that the corresponding moment maps satisfy
\begin{equation}\label{eq:varphi=0}
\varphi_l(x)=\varphi_r(x)=0
\end{equation}
if $x\in \Xcal_P$ satisfies $\|\Phi_r(x)\|\leq c$.
We consider now the product $M\times \Xcal_P$ equipped with the $\spinc$ bundle
$\Scal\boxtimes\Scal_P^n$ and with the following $K\times K$ action:
$$
(k_l,k_r)\cdot (m,x)= (k_r\cdot m, (k_l,k_r)\cdot x).
$$
The moment map relative to the action of $K\times K$ on the line bundle $\det(\Scal\boxtimes\Scal_P^n)$ is
$$
(m,x)\longmapsto (\Phi^n_l(x), \Phi_\Scal(m)+\Phi^n_r(x)).
$$
We restrict the action of $K\times K$ on $M\times \Xcal_P$ to the subgroup $H\times K$. We see that the corresponding moment map
$(p\circ \Phi^n_l, \Phi_\Scal+\Phi^n_r)$ is proper, so we can consider the formal geometric quantization of the $\spinc$ manifold
$(M\times \Xcal_P,\Scal\boxtimes\Scal_P^n)$ relative to the action of $H\times K$:
$\qfor_{K\times H}(M\times \Xcal_P,\Scal\boxtimes\Scal_P^n)\in \hat{R}(K\times H)$.
We are interested in the following $H$-module
$$
E(n):= \left[\qfor_{H\times K}(M\times \Xcal_P,\Scal\boxtimes\Scal_P^n)\right]^K.
$$
The proof of Theorem \ref{theo-principal} will follows from the computation of the asymptotic behaviour of $E(n)$ by two means.
\subsection{First computation}
Let us write $\qfor_{K}(M,\Scal)= \sum_{\Ocal} m_\Ocal\,\pi_\Ocal^K$.
%$$
%\quad
%E(n)=\sum_{\mu\in\hat{H}} m(\mu,n) \pi_\mu^H.
%$$
\begin{prop}\label{prop:first-computation}We have
$$
\lim_{n\to\infty} E(n)=\qfor_{K}(M,\Scal)\vert_H:=\sum_{\lambda\in\hat{K}} m_\Ocal\, \pi_\Ocal^K\vert_H \in \hat{R}(H)
$$
\end{prop}
\begin{proof}
We start by using the multiplicative property (Theorem \ref{theo:hochs-song}):
$$
\qfor_{H\times K}(M\times \Xcal_P,\Scal\boxtimes\Scal_P^n)= \qfor_{K}(M,\Scal_M)\otimes \Qcal_{H\times K}(\Xcal_P,\Scal_P^n).
$$
Thanks to Proposition \ref {prop:Q-X-P}, we know that $\Qcal_{H\times K}(\Xcal_P,\Scal_P^n)$ is equal to
$$
\sum_{\Ocal\cap \{n P^\circ +\rho\}\neq\emptyset } \pi^K_\Ocal\vert_H\otimes (\pi^K_\Ocal)^* + R(n)
$$
with $R(n)= \sum_{\Ocal} a_{n,\Ocal}\, \pi^K_\Ocal\vert_H\otimes (\pi^K_\Ocal)^*$ where $a_{n,\Ocal}\neq 0$ only if $\Ocal\cap \{n \partial P +\rho\}\neq\emptyset$.
So we see that
$$
E(n)= \sum_{\Ocal\cap \{n P^\circ +\rho\}\neq\emptyset } m_\Ocal\,\pi^K_\Ocal\vert_H + r(n)
$$
with $ r(n)=\left[\qfor_{K}(M,\Scal)\otimes R(n)\right]^K$. It remains to check that $\lim_{n\to\infty} r(n)= 0$ in $\hat{R}(H)$.
If $\Ocal$ is a $K$-coadjoint orbit we denote $\|\Ocal\|$ the norm of any of its element. Note that there exists $d>0$ such that if
$\Ocal\cap \{n \partial P +\rho\}\neq\emptyset$, then $\|\Ocal\|\geq n d$.
Let $\Ocal'$ be a regular admissible $H$-orbit. By definition the multiplicity of $\pi_{\Ocal'}^H$ in $r(n)$ decomposes as follows
$$
[\pi_{\Ocal'}^H: r(n)]= \sum_{\Ocal} a_{n,\Ocal}\, m_\Ocal\, [\pi_{\Ocal'}^H: \pi_\Ocal^K\vert_H].
$$
Suppose that $r(n)$ does not tends to $0$ in $\hat{R}(H)$: there exists a regular admissible $H$-orbit $\Ocal'$ such that the set
$\{n\geq 1, [\pi_{\Ocal'}^H: r(n)]\neq 0\}$ is infinite. Hence there exists a sequence
$(n_k, \Ocal_k)$ such that $\lim_{k\to \infty} n_k=\infty$ and $a_{n_k,\Ocal_k}\, m_{\Ocal_k}\, [\pi_{\Ocal'}^H: \pi_{\Ocal_k}^K\vert_H]\neq 0$.
Thanks to the $[Q,R]=0$ property, we have
\begin{equation}
\begin{cases}
1.\ \|\Ocal_k\|\geq d\, n_k, \\
2.\ \Ocal_k\in \Phi_\Scal(M),\\
3.\ \Ocal'\subset {\rm p}(\Ocal_k),
\end{cases}
\end{equation}
where ${\rm p}:\kgot^*\to \hgot^*$ is the projection. Points $2.$ and $3.$ give that
$$
\Ocal_k\in K\Phi_\Scal\left(({\rm p}\circ \Phi_\Scal)^{-1}(\Ocal')\right).
$$
Since ${\rm p}\circ \Phi_\Scal$ is proper, we have only a finite number of $K$-admissible orbits contained in the compact set
$K\Phi_\Scal\left(({\rm p}\circ \Phi_\Scal)^{-1}(\Ocal')\right)$. This is in contradiction with the first point. $\Box$
\end{proof}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\subsection{Reduction in stage}\label{sec:stage}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
In this section, we explain the case of reduction in stages. Suppose that we have an action of the compact Lie group
$G\times K$ on the $\spinc$ manifold $(N,\Scal_N)$. Let $\Phi_{\Scal_N}=\Phi_{\Scal_N}^G\oplus\Phi_{\Scal_N}^K : N\to \ggot^*\oplus\kgot^*$ be the
corresponding moment map associated to the choice of an invariant connection $\nabla$ on $\det(\Scal_N)$. We suppose that
\begin{itemize}
\item $0$ is a regular value of $\Phi_{\Scal_N}^K$,
\item $K$ acts freely on $Z:=(\Phi_{\Scal_N}^K)^{-1}(0)$,
\item the set $\Phi_{\Scal_N}^{-1}(0)$ is compact.
\end{itemize}
We denote by $\pi: Z\to N_0:=Z/K$ the corresponding $G$-equivariant principal fibration.
On $Z$, we obtain an exact sequence $0\longrightarrow \T Z\longrightarrow \T M\vert_Z \stackrel{\T\Phi_K}{\longrightarrow} [\kgot^*]\to 0$,
where $[\kgot^*]$ is the trivial bundle $Z\times\kgot^*$. We have also an orthogonal decomposition
$\T Z= \T_{K} Z \oplus [\kgot]$ where $[\kgot]$ is the sub-bundle identified to $Z\times\kgot$ through the map
$(p,X)\mapsto X\cdot p$. So $\T M\vert_Z$ admits the orthogonal decomposition
$\T N\vert_Z \simeq \T_{K} Z\oplus [\kgot] \oplus [\kgot^*]$. We rewrite this as
\begin{equation}\label{eq:tangent-P}
\T N\vert_Z \simeq \T_{K} Z\oplus [\kgot_\C]
\end{equation}
with the convention $[\kgot]=Z\times(\kgot\otimes\R) $ and $[\kgot^*] = Z\times (\kgot\otimes i\R)$.
Note that the bundle $\T_{K} Z$ is naturally identified with $\pi^*(\T N_{0})$.
We can divide the $\spinc$-bundle $\Scal_N\vert_Z$ by the $\spinc$-bundle
$\bigwedge \kgot_\C$ for the vector space $\kgot_\C$ (see Section 2.2 in \cite{pep-vergne:witten}).
\begin{defi}\label{def:E-red}
Let $\Scal_{N_0}$ be the $\spinc$-bundle on $N_0$ such that
$$
\Scal_N\vert_Z \simeq \pi^*(\Scal_{N_0})\otimes [\bigwedge\kgot_\C]
$$
is an isomorphism of graded Clifford bundles on $\T N\vert_Z$.
\end{defi}
We see then that the line bundle $\det(\Scal_{N_0})$ is equal to $\det(\Scal_{N})\vert_Z/K$. Hence the connection $\nabla$ on $\det(\Scal_N)$
induces a $G$-invariant connection $\nabla_0$ on $\det(\Scal_{N_0})$. The corresponding moment map $\Phi_{\Scal_{N_0}}:N_0\to \ggot^*$ is the equivariant map
induced by $\Phi_{\Scal_N}^G: N\to \ggot^*$.
\begin{prop}\label{prop:localisation-stage}
We have the following relation
$$
\left[\Qcal_{G\times K}(N,\Scal_N,\{\Phi_{\Scal_N}=0\})\right]^K=
\Qcal_G(N_{0},\Scal_{N_0},\{\Phi_{\Scal_{N_0}}=0\}) \qquad \mathrm{in}\quad \hat{R}(G).
$$
\end{prop}
\begin{proof}
The proof is done in Section 3.4.2 of \cite{pep-vergne:witten} in the Hamiltonian setting. The same proof works here.
\end{proof}
\subsection{Second computation}
We consider $\qfor_H(M,\Scal)\in \hat{R}(H)$.
\begin{prop}\label{prop:second-computation}
Suppose that the generic infinitesimal stabilizer of the $K$-action on $M$ is {\em abelian}. Let $\Ocal'$ be a regular admissible $H$-orbit. There exists $n_{\Ocal'}\geq 1$ such that
$$
\left[\pi_{\Ocal'}^H : E(n)\right]=\left[\pi_{\Ocal'}^H :\qfor_H(M,\Scal)\right]
$$
when $n\geq n_{\Ocal'}$.
\end{prop}
\begin{proof} First of all, since the $K$-action on $M$ has generic abelian infinitesimal stabilizers,
we see that the $H\times K$-action on $M\times \Xcal_P$ has also generic abelian infinitesimal stabilizers.
Let us denote $\tilde{\Ocal}$ the $H\times K$ regular admissible orbit $\Ocal'\times K\rho$. We work with the $H\times K$ manifold
$$
N:= M\times \Xcal_P \times \tilde{\Ocal}^*
$$
which is equipped with the $\spinc$ bundles
$\Scal_N^n:= \Scal \boxtimes\Scal_P^n\boxtimes \Scal_{\tilde{\Ocal}^*}$. The moment map associated to the action of
$H\times K$ on $\det(\Scal_N^n)$ is $\Phi_{\Scal_N^n}=(\Phi^n_H,\Phi^n_K)$ where
$$
\Phi^n_H(m,x,\eta,\xi)={\rm p}\left(n\Phi_l(x)+\varphi_l(x)\right)+\eta,
$$
and
$$
\Phi^n_K(m,x,\eta,\xi)=\Phi_\Scal(m)+n\Phi_r(x)+ \varphi_r(x)+\xi
$$
for $(m,x,\eta,\xi)\in M\times \Xcal_P\times (\Ocal')^*\times (K\rho)^*$.
Thanks to the multiplicative property we have
$$
\left[\pi_{\Ocal'}^H : E(n)\right]= \left[\qfor_{H\times K}(N,\Scal_N^n)\right]^{H\times K}.
$$
Using the fact that the $H\times K$-action on $M\times \Xcal_P$ has generic abelian infinitesimal stabilizers, we know that
\begin{equation}\label{eq:multiplicity-localized}
\left[\pi_{\Ocal'}^H : E(n)\right]= \left[\Qcal_{H\times K}(N,\Scal_N^n,\{\Phi_{\Scal_N^n}=0\})\right]^{H\times K}.
\end{equation}
See Remark \ref{rem:abelian-stabilizer}. Now we are going to compute the right hand side of (\ref{eq:multiplicity-localized})
by using the reduction in stage (see Section \ref{sec:stage}).
We start with the
\begin{lem}
There exists $R,R'>0$, independent of $n$, such that
if \break $(m,x,\eta,\xi)\in \{\Phi_{\Scal_N^n}=0\}$ then
$\|\Phi_\Scal(m)\|\leq R$ and $ \|\Phi_r(x)\|\leq R'/n$.
\end{lem}
\begin{proof}
Let $(m,x,\eta,\xi)\in \{\Phi_{\Scal_N^n}=0\}$. We have ${\rm p}\left(n\Phi_l(x)+\varphi_l(x)\right)+\eta=0$ and
$\Phi_\Scal(m)+n\Phi_r(x)+ \varphi_r(x)+\xi=0$. Let $k\in K$ such that $k\Phi_r(x)=-\Phi_l(x)$
(see Point $(2)$ in Section \ref{sec:Concini-Procesi}). We get then
${\rm p}(\Phi_\Scal(km))+ {\rm p}(k\varphi_r(x)+\varphi_l(x)+\xi)+\eta=0$. The term
${\rm p}(k\varphi_r(x)+\varphi_l(x)+\xi)+\eta$ is bounded, and since ${\rm p}\circ\Phi_\Scal$ is proper, the variable $m$ belongs to a compact
of $M$ (independent of $n$). Finally the identity $\Phi_\Scal(m)+n\Phi_r(x)+ \varphi_r(x)+\xi=0$ shows that $n\Phi_r(x)$ is bounded by a quantity independent of $n$. $\Box$
\end{proof}
\medskip
So, if $n$ is large enough, the set $\{\Phi_{\Scal_N^n}=0\}$ is contained in the open subset $M\times \Xcal_P^o\times\tilde{\Ocal}^*\subset N$ that we can identify with
$$
\tilde{N}= M \times K\times \Ucal_P \times \tilde{\Ocal}^*
$$
through the diffeomorphism $\Upsilon : K\times \Ucal_P\to \Xcal_P^o$ (see Point $(4)$ in Section \ref{sec:Concini-Procesi}).
Moreover, thanks to (\ref{eq:varphi=0}), for $n$ large enough an element
$(m,g,\nu,\eta,\xi)\in \tilde{N}$ belongs to $\{\Phi_{\Scal_N^n}=0\}$ if and only if
\begin{equation}
\begin{cases}
n{\rm p}(g\nu)+\eta=0,\\
\Phi_\Scal(m) -n\nu+\xi=0.
\end{cases}
\end{equation}
We use now the reduction in stage for $n$ large enough. The map
$$
(m,\eta,\xi,g)\mapsto (m,g,\frac{\Phi_\Scal(m)+\xi}{n},\eta)
$$
defines a diffeomorphism between $M\times \tilde{\Ocal}^*\times K$ and the sub-manifold \break
$Z:=\{\Phi_K^n=0\}\subset \tilde{N}$ and induces a diffeomorphism
$$
\Psi: M\times \tilde{\Ocal}^* \stackrel{\sim}{\longrightarrow} N_0= Z/K.
$$
Through the diffeomorphism $\Psi$, the $H$-action on $N_0= Z/K$ corresponds to the induced action of the subgroup
$H\simeq \{(h,h), h\in H\}\subset H\times K$ on $M\times \tilde{\Ocal}^*$.
Through the diffeomorphism $\Psi$, the moment map $\Phi_{\Scal^n_{N_0}}: N_0\to \hgot^*$ becomes
$$
\Phi_{\tilde{\Ocal}}(m,\xi,\eta)={\rm p}(\Phi_\Scal(m)+\xi)+\eta,
$$
for $(m,\xi,\eta)\in M\times \tilde{\Ocal}^*$.
\begin{lem}\label{lem:diffeo-psi}
Through the diffeomorphism $\Psi$, the induced $\spinc$ bundle $\Scal^n_{N_0}$ corresponds to $\Scal_M\boxtimes \Scal_{ \tilde{\Ocal}^*}$.
\end{lem}
\begin{proof}
We consider the restriction of the $\spinc$ bundle $\Scal_P^n$ to the open subset $\Xcal^\circ_P$. Let
$\Scal^n:=\Upsilon^{-1}(\Scal_P^n\vert_{\Xcal^\circ_P})$ be the corresponding $K\times K$-equivariant $\spinc$ bundle on
$K\times\Ucal_P$. It must be of the form $\Scal^n\simeq F\times \bigwedge\kgot_\C\times K\times\Ucal_P$ where $F$ is a character of
$K\times K$. If we look at the value of $\Scal_P^n$ at the point $\Upsilon(1,0)\in\Xcal_P$, we see that $F$ is trivial. The Lemma follows.
$\Box$
\end{proof}
\medskip
Finally, for $n$ large enough, we get
\begin{eqnarray*}
\left[\pi_{\Ocal'}^H : E(n)\right]&=& \left[\Qcal_{H\times K}(N,\Scal_N^n,\{\Phi_{\Scal_N^n}=0\})\right]^{H\times K}
\hspace{13mm} [1]\\
&=&\left[\Qcal_{H}(N_0,\Scal_{N_0}^n,\{\Phi_{\Scal^n_{N_0}}=0\})\right]^{H}\hspace{19mm} [2]\\
&=&\left[\Qcal_{H}(M\times\tilde{\Ocal}^*,\Scal\boxtimes\Scal_{\tilde{\Ocal}^*},\{\Phi_{\tilde{\Ocal}}=0\})\right]^{H} \hspace{5mm} [3]\\
&=& \left[\qfor_{H}(M\times\tilde{\Ocal}^*,\Scal\boxtimes\Scal_{\tilde{\Ocal}^*})\right]^{H}\hspace{21mm} [4]\\
&=& \left[\qfor_{H}(M,\Scal)\otimes\Qcal_H(\tilde{\Ocal}^*,\Scal_{\tilde{\Ocal}^*})\right]^H\hspace{15mm} [5]\\
&=& \left[\pi_{\Ocal'}^H : \qfor_{H}(M,\Scal)\right].\hspace{36mm} [6]
\end{eqnarray*}
First we see that $[1]$ corresponds to (\ref{eq:multiplicity-localized}).
Equality $[2]$ is the reduction in stage (see Proposition \ref{prop:localisation-stage}) and Equality $[3]$ is a
consequence of the diffeomorphism $\Psi$ (see Lemma \ref{lem:diffeo-psi}). Equality $[4]$ follows from the fact that $M$ has
abelian generic infinitesimal stabilizers (see Remark \ref{rem:abelian-stabilizer}).
Equality $[5]$ is a consequence of the multiplicative property. Equality $[6]$ follows from the identity
$\Qcal_H(\tilde{\Ocal}^*,\Scal_{\tilde{\Ocal}^*})=(\pi_{\Ocal'}^H)^*$.
The proof of Proposition \ref{prop:second-computation} is completed. $\Box$
\end{proof}
\bigskip
We can now conclude our exposition. Proposition \ref{prop:first-computation} tell us that \break
$\lim_{n\to\infty} E(n)=\qfor_{K}(M,\Scal)\vert_H$
while Proposition \ref{prop:second-computation} says that $\lim_{n\to\infty} E(n)$ $=\qfor_{H}(M,\Scal)$
when the manifold $M$ has abelian generic infinitesimal stabilizers.
So we have proved Theorem \ref{theo-principal} for manifolds with abelian generic infinitesimal stabilizers.
But we have checked in Lemma \ref{lem:simple} that
it is sufficient to get the proof of Theorem \ref{theo-principal} in the general case.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
{\small
\begin{thebibliography}{99}
\bibitem{Atiyah74} {\sc M.F. Atiyah}, {\em Elliptic operators and
compact groups}, Lecture Notes in Mathematics {\bf 401}, Springer-Verlag, Berlin, 1974.
%\bibitem{Atiyah-Bott-YM} {\sc M.F. Atiyah} and {\sc R. Bott}, {\em The Yang-Mills equations over riemann surfaces},
%Philosophical Transactions of the Royal Society of London A (1983), 523--615.
%\bibitem{Atiyah82} {\sc M.F. Atiyah}, {\em Convexity and commuting
% Hamiltonians}, Bull. London Math. Soc. {\bf 14} (1982), 1--15.
%
%\bibitem{Atiyah-Segal68} {\sc M.F. Atiyah} and {\sc G.B. Segal},
%{\em The index of elliptic operators II}, Ann. of Math. {\bf 87} (1968), 531--545.
%
%\bibitem{Atiyah-Singer-1} {\sc M.F. Atiyah} and {\sc I.M. Singer},
%{\em The index of elliptic operators I}, Ann. of Math. {\bf 87} (1968), 484--530.
%
\bibitem{Atiyah-Singer-2} {\sc M.F. Atiyah} and {\sc I.M. Singer},
{\em The index of elliptic operators III}, Ann. of Math. {\bf 87} (1968), 546--604.
%
%\bibitem{Atiyah-Singer-3} {\sc M.F. Atiyah} and {\sc I.M. Singer},
%{\em The index of elliptic operators IV}, Ann. of Math. {\bf 93} (1971), 139--141.
\bibitem{Atiyah-Hirzebruch70} {\sc M.F. Atiyah} and
{\sc F. Hirzebruch}, {\em Spin Manifold and group actions}, Essays
on Topology and Related Topics (Geneva 1969), Springer-Verlag, Berlin-Heidelberg-New York,
1970.
\bibitem{B-G-V} {\sc N. Berline}, {\sc E. Getzler} and {\sc M. Vergne}, {\em Heat kernels and
Dirac operators}, Grundlehren {\bf 298}, Springer, Berlin, 1991.
%\bibitem{B-V.inventiones.96.1} {\sc N. Berline} and {\sc M. Vergne},
%{\em The Chern character of a transversally elliptic symbol and the
%equivariant index}, Invent. Math. {\bf 124} (1996), 11--49.%
%\bibitem{B-V.inventiones.96.2} {\sc N. Berline} and {\sc M. Vergne},
%{\em L'indice \'equivariant des op\'erateurs transversalement
%elliptiques}, Invent. Math. {\bf 124} (1996), 51--101.
\bibitem{Braverman-02} {\sc M. Braverman}, {\em Index theorem for equivariant Dirac operators on noncompact manifolds},
K-Theory {\bf 27} (2002), 61--101.
%\bibitem{Braverman-14} {\sc M. Braverman}, {\em The index theory on non-compact manifolds with proper group action},
%Journal of Geometry and Physics {\bf 98} (2015), 275--284.
%\bibitem{CdS-K-T}{\sc A. Cannas da Silva}, {\sc Y. Karshon} and
%{\sc S. Tolman}, {\em Quantization of presymplectic manifolds and circle
%actions}, Trans. A. M. S. {\bf 352} (2000), 525--552.
%
%\bibitem{du} {\sc M. Duflo}, {\em Construction de repr\'esentations unitaires d'un groupe de Lie}, CIME, Cortona (1980).
%
%
\bibitem{Duistermaat96} {\sc J. J. Duistermaat},
{\em The heat equation and the Lefschetz fixed point formula for the
Spin\textsuperscript{c}-Dirac operator}, Progress in Nonlinear
Differential Equation and Their Applications, vol. 18, Birkhauser,
Boston, 1996.
%\bibitem{Grossberg-Karshon94} {\sc M. Grossberg} and {\sc Y. Karshon}, {\em Bott towers, complete integrability, and the extended character of representations}, Duke Math. Journal {\bf 76} (1994), 23--58.
%
%\bibitem{Grossberg-Karshon98} {\sc M. Grossberg} and {\sc Y. Karshon}, {\em Equivariant index and the moment map for completely integrable torus actions}, Advances in Math. {\bf 133} (1998), 185--223.
%
%\bibitem{Guillemin95} {\sc V. Guillemin}, {\em Reduced phase spaces and Riemann-Roch in Proceedings: Lie
%Groups and Geometry in honor of B. Kostant}, Progress in Mathematics {\bf 123}, Birkh\"{a}user Boston (1995), 305--334.
%
%
%
%\bibitem{Guillemin-Sternberg82} {\sc V. Guillemin} and {\sc S. Sternberg},
%{\em Geometric quantization and multiplicities of group representations}, Invent. Math. {\bf 67} (1982), 515--538.
%
%\bibitem{Guillemin-Sternberg84} {\sc V. Guillemin} and {\sc S. Sternberg}, {\em A normal form for the moment map},
%Differential geometric methods in mathematical physics, {\bf 6} (1984), 161--175.
\bibitem{Hattori78} {\sc A. Hattori}, {\em $\spinc$-structures and $S^1$-actions},
Invent. Math. {\bf 48} (1978), 7--31.
\bibitem{Hochs-Mathai2014} {\sc P. Hochs}, and {\sc V. Mathai}, {\em Quantising proper actions on $\spinc$-manifolds},
preprint arXiv:1408.0085 (2014).
\bibitem{Hochs-Song-duke} {\sc P. Hochs}, and {\sc Y. Song}, {\em Equivariant indices of $\spinc$-Dirac operators
for proper moment maps}, Duke Math. J. {\bf 166}, (2017), 1125-1178.
%\bibitem{Jeffrey-Kirwan1997} {\sc L.C. Jeffrey}, and {\sc F.C. Kirwan}, {\em Localization and quantization conjecture},
%Topology {\bf 36} (1997), 647--693.
%
%\bibitem{Karshon-Tolman93} {\sc Y. Karshon} and {\sc S. Tolman},
%{\em The moment map and line bundles over presymplectic toric manifolds},
%J. Differential Geom {\bf 38} (1993), 465-484.
\bibitem{Kawasaki81} {\sc T. Kawasaki}, The index of elliptic operators over V-manifolds, {\em Nagoya Math. J.} {\bf 84} (1981), 135--157.
%\bibitem{Kirwan84} {\sc F.C. Kirwan}, {\em Cohomology of quotients in symplectic and algebraic geometry},
%Princeton University Press, {\bf 31} (1984).
%\bibitem{L-M-T-W} {\sc E. Lerman}, {\sc E. Meinrenken}, {\sc S. Tolman} and {\sc C. Woodward},
%{\em Non-Abelian convexity by symplectic cuts}, Topology {\bf 37} (1998), 245--259.
\bibitem{Ma-Zhang14} {\sc X. Ma} and {\sc W. Zhang}, {\em Geometric quantization for proper moment maps: the Vergne conjecture}, Acta Mathematica {\bf 212} (2014), 11--57.
\bibitem{Meinrenken98} {\sc E. Meinrenken}, {\em Symplectic surgery and the Spin\textsuperscript{c}-Dirac operator},
Advances in Math. {\bf 134} (1998), 240--277.
\bibitem{Meinrenken-Sjamaar} {\sc E. Meinrenken} and {\sc R. Sjamaar},
{\em Singular reduction and quantization}, Topology {\bf 38} (1999), 699--762.
%\bibitem{pep-1999} {\sc P-E. Paradan}, {\em Formules de localisation en cohomologie \'equivariante},
%Compositio Mathematica {\bf 117} (1999), 243--293.
\bibitem{pep-RR} {\sc P-E. Paradan}, {\em Localization of the Riemann-Roch character}, J. Functional Analysis {\bf 187} (2001), 442--509.
\bibitem{pep-ENS} {\sc P-E. Paradan}, {\em Spinc-quantization and the K-multiplicities of the discrete series},
Annales scientifiques de l'E.N.S. {\bf 36} (2003), 805–845.
\bibitem{pep-IF} {\sc P-E. Paradan}, {\em Formal geometric quantization}, Ann. Inst. Fourier {\bf 59} (2009), 199--238.
\bibitem{pep-pacific} {\sc P-E. Paradan}, {\em Formal geometric quantization II}, Pacific journal of mathematics, {\bf 253}, 169-211.
%\bibitem{pep-JSG} {\sc P-E. Paradan}, {\em Spin quantization commutes with reduction}, J. of Symplectic Geometry {\bf 10} (2012), 389--422.
%\bibitem{pep-vergne:bismut}{\sc P-E. Paradan} and {\sc M. Vergne}, {\em Index of transversally elliptic operators},
%Ast\'erique {\bf 328} (2009), 297--338.
%\bibitem{pep-vergne:cras}{\sc P-E. Paradan} and {\sc M. Vergne}, {\em The multiplicities of the equivariant index of twisted Dirac operators}, C. R. %A. S., Volume 352, Issue 9, (2014), 673-677.
\bibitem{pep-vergne:witten}{\sc P-E. Paradan} and {\sc M. Vergne}, {\em Witten non abelian localization
for equivariant $K$-theory and the $[Q,R]=0$ Theorem}, preprint arXiv 1504.07502 (2015), accepted in Memoirs of the A.M.S..
\bibitem{pep-vergne:magic}{\sc P-E. Paradan} and {\sc M. Vergne}, {\em Admissible coadjoint orbits for compact Lie groups},
preprint arXiv:1512.02367 (2015), accepted in Transformations Groups.
\bibitem{pep-vergne-acta}{\sc P-E. Paradan} and {\sc M. Vergne}, {\em Equivariant Dirac operators and differential geometric invariant theory},
preprint arXiv:1512.02367 (2015), accepted in Acta Mathematica.
%\bibitem{Sjamaar95} {\sc R. Sjamaar}, {\em Holomorphic Slices, Symplectic Reduction and Multiplicities of Representations},
%Annals of Mathematics {\bf 141} (1995), 87--129.
%%\bibitem{Sjamaar96} {\sc R. Sjamaar}, {\em Symplectic reduction and Riemann-Roch formulas for multiplicities},
%%Bulletin of the A.M.S. {\bf 33} (1996), 327--338.
%\bibitem{Szenes-Vergne} {\sc A. Szenes} and {\sc M. Vergne},
%{\em $[Q,R]=0$ and Kostant partition functions}, ArXiv:1006-4149.
%\bibitem{Teleman-Annals00} {\sc C. Teleman}, The quantization conjecture revisited, {\em Annals of Math.}
%{\bf 152} (2000), 1--43.
\bibitem{Tian-Zhang98} {\sc Y. Tian} and {\sc W. Zhang}, {\em An analytic proof of the geometric
quantization conjecture of Guillemin-Sternberg}, Invent. Math. {\bf 132} (1998), 229--259.
%\bibitem{Vergne96} {\sc M. Vergne}, {\em Multiplicity formula for geometric
%quantization, Part I, Part II, and Part III}, Duke Math. Journal {\bf 82} (1996), 143--179, 181--194, 637--652.
%
%\bibitem{Vergne-Bourbaki} {\sc M. Vergne}, {\em Quantification g\'eom\'etrique et
%r\'eduction symplectique}, S\'eminaire Bourbaki, vol. 2000/2001. Ast\'erisque {\bf 282} (2002), Exp. No 888, 249--278.
%
%
%\bibitem{Witten} {\sc E. Witten}, Two dimensional gauge theories
%revisited, {\em J. Geom. Phys.} {\bf 9} (1992), 303--368.
\end{thebibliography}
}
\end{document}