Induced $C^*$-complexes in metaplectic geometry
Extracted source text ยท arXiv 1711.09937
\documentclass{article}
%dat mezi docuclass a article [draft]
\usepackage{amssymb,amsmath, tikz}
\usepackage{graphics}
\input xy
\xyoption{all}
\usepackage{accents}
\usepackage{braket}
\usepackage{hyperref}
\newlength{\dhatheight}
\newcommand{\doublehat}[1]{%
\settoheight{\dhatheight}{\ensuremath{\hat{#1}}}%
\addtolength{\dhatheight}{-0.35ex}%
\hat{\vphantom{\rule{1pt}{\dhatheight}}%
\smash{\hat{#1}}}}
\newcommand\phantomarrow[2]{%
\setbox0=\hbox{$\displaystyle #1\to$}%
\hbox to \wd0{%
$#2\mapstochar
\cleaders\hbox{$\mkern-1mu\relbar\mkern-3mu$}\hfill
\mkern-7mu\rightarrow$}%
\,}
\begin{document}
\title{Induced $C^*$-complexes in metaplectic geometry}
\author{Svatopluk Kr\'ysl\footnote{{\it E-mail address}:
Svatopluk.Krysl@mff.cuni.cz}\\ {\it \small Charles University, Faculty of
Mathematics and Physics, Czechia}
\thanks{The author thanks for financial supports from the founding
No. 17-01171S granted by the Czech Science Foundation. We thank to the
anonymous reviewer for his
comments and suggestions.}}
\maketitle \noindent
\centerline{\large\bf Abstract}
For a symplectic manifold admitting a metaplectic structure and for a Kuiper map, we
construct a complex of differential operators acting on exterior differential
forms with values in the dual of Kostant's symplectic spinor bundle.
Defining a Hilbert $C^*$-structure on this bundle for a suitable $C^*$-algebra,
we obtain an elliptic $C^*$-complex in the sense of Mishchenko--Fomenko. Its cohomology groups
appear to be finitely generated projective Hilbert $C^*$-modules. The paper can
serve as a guide for handling of differential complexes and
PDEs on Hilbert bundles.
\section{Introduction}
An important assumption is made in
classical works on the Hodge theory of elliptic complexes beside the compactness of the base manifold.
Namely, the bundles in the complexes are supposed to be of finite rank. (See Hodge
\cite{Hodge} and Palais \cite{Palais}.) Under these two assumptions,
elements of the complexes can be decomposed into closed, co-closed
and harmonic parts uniquely. It follows that cohomology groups of the
complexes are bijectively representable
by harmonic elements. Moreover, they are finite
dimensional vector spaces.
The study of elliptic complexes has a long history which goes back to
W. Hodge and G. de Rham. See Maurin \cite{Maurin} and also Klein
\cite{Klein} for references and for the (pre-)history of this subject.
Elliptic complexes are studied also in areas `extraterritorial' to
differential geometry and global analysis or in cases
violating the compactness. See Tsai at
al. \cite{Tsai}, Schmid, Vilonen \cite{SchmidVilonen}, Li \cite{Li},
Nekov\'a\v{r}, Scholl
\cite{Nekovar}, Hain \cite{Hain} and Albin et al. \cite{ALMP} for recent
contributions.
However, elliptic complexes can be studied in the case of
infinite rank bundles as well.
Troitsky in \cite{Troitsky} and Schick, subsequently, in \cite{Schick} elaborate a theory for
elliptic complexes on finitely generated
projective Hilbert $C^*$-bundles, which is based on the index theory of
Mishchenko and Fomenko described in \cite{Mishchenko} and \cite{FM}.
In \cite{Troitsky} and \cite{Schick}, $C^*$-compact perturbations of the
complexes' differentials are allowed. For the infinite rank bundles, one cannot
expect that, in general,
the space of harmonic elements represents the cohomology groups
homeomorphically (quotient topology) and linearly (quotient projections).
Therefore, it makes sense to find conditions when this happens. This was
investigated by the author in the past years (see \cite{KryslAGAG2}, \cite{KryslJGP2} or
an overview in \cite{KryslHabil}).
In the connection to sheaf cohomology, Banach and Fr\'echet bundles are
studied in the papers of Illusie
\cite{Illusie} and R\"ohrl \cite{Rohrl}. To present a sample of the broad
context
(foremost connected to $C^*$-algebras, $K$-theory and homological algebra) in which
such bundles are considered, let us mention the papers of Maeda, Rosenberg
\cite{MaedaRosenberg}, Freed, Lott \cite{Freed}, Larra\'in-Hubach
\cite{Larrain}, the author \cite{KryslJGP2} and Fathizadeh, Gabriel
\cite{FathiGabr}. There are also works
in which holomorphic Banach bundles and their sheaf cohomology groups are
treated. See Lempert \cite{Lempert} and Kim \cite{Kim}. A further reason for
an investigation of these complexes might originate in BRST-theories
(\cite{HenTeitel}), which we do not touch here explicitly, although we give a
modest interpretation of the introduced structures in the realm of quantum
theory.
The aim of our paper is to study elliptic complexes for the case of infinite
rank Hilbert $C^*$-bundles which are
induced by Lie group representations, and to show as well
how the theory of connections and appropriately generalized
elliptic complexes apply in this situation. In order to make the considered
situation not too general,
we choose a specific Lie group representation (the Segal--Shale--Weil representation of the metaplectic group).
We hope that the reader may follow the text easier. Our further aim
is to find examples for the theory of Mishchenko and Fomenko
(\cite{FM}) that would not be trivial or covered by the theory developed around
the classical Atiyah--Singer index theorem, and to show how the Hodge theory for certain $C^*$-bundles
(derived in \cite{KryslAGAG2} and \cite{KryslJGP2}) make us able to describe
the cohomology of elliptic complexes
on Hilbert bundles easily.
Let $(V,\omega_0)$ be a symplectic vector space and $L$ be a Lagrangian
subspace of it. The Segal--Shale--Weil representation is a non-trivial representation of
the connected double cover $\widetilde{G}$ of the symplectic group on a
complex Hilbert space by unitary operators. The Hilbert space is formed by the
$L^2$-functions defined on $L.$
For references to this representation see, e.g., Weil
\cite{Weil}, Borel, Wallach \cite{BorelWallach}, Wallach \cite{Wallach} and
Folland \cite{Folland}. Let us denote the dual of the Segal--Shale--Weil
representation by $\sigma$ and the continuous dual of the Hilbert space by $H.$ We call the resulting representation
the {\it oscillator representation}.
The double cover, the metaplectic group, acts by
$\sigma_k: \widetilde{G} \to \mbox{Aut}(H^k)$ on the tensor products
$H^k=\bigwedge^k V^*\otimes H$ in a natural way.
Let $CH$ denote the algebra of compact operators on
the vector space carrier of the Segal--Shale--Weil representation, i.e., on $L^2(L).$
On the vector space space $CH,$ we consider the representation, denoted by
$\rho,$ that is given by the conjugation by the
Segal--Shale--Weil representation.
We introduce a right $CH$-module structure $H \times CH \to H$ on $H$
and a $CH$-product $H\times H \to CH$ obtaining a Hilbert $CH$-module -- a
mixture of a module and of a normed vector space as
described by Paschke \cite{Paschke} and Rieffel \cite{Rieffel}.
The
spaces $H^k,$ $k\geq 1,$
are made Hilbert $CH$-modules as well.
Inducing $(\sigma_k,H^k)$ to a principal $\widetilde{G}$-bundle over a
symplectic manifold $M$ and in particular, to the so-called metaplectic
bundle of a metaplectic structure (Kostant
\cite{Kostant}), one
gets Banach bundles $\mathcal{H}^k$ (Sect. 2).
Further, we associate $(\rho, CH)$
to the metaplectic bundle, getting a
continuous Banach bundle. We denote it by $\mathcal{CH}$ and call it the bundle of measurements.
Then we define certain intermediate maps $\mathcal{H} \times_{M}
\mathcal{CH} \to \mathcal{H}$ and $\mathcal{H} \times_{M} \mathcal{H} \to
\mathcal{CH},$ called pointwise structures, which are families of
$C^*$-algebra actions and of $C^*$-algebra valued products.
We are not aware that any substantial modification was
done regarding the assumptions in the theory of Mishchenko--Fomenko from \cite{FM}.
Thus, we adapt the pointwise structures to get the honest required ones.
Recall that we have to define maps $\mathcal{H} \times CH \to \mathcal{H}$ and
$\mathcal{H} \times_M \mathcal{H} \to CH$ instead of $\mathcal{H} \times_M
\mathcal{CH} \to \mathcal{H}$ and
$\mathcal{H}\times_M \mathcal{H} \to \mathcal{CH},$ respectively.
This can be assured by choosing a trivialization $J$ of $\mathcal{H},$ called a
Kuiper map, which is possible in the case of a unitary representation.
Nevertheless, let us note
that we can choose the maps
$\mathcal{H} \times CH \to \mathcal{H}$ and $\mathcal{H} \times_M \mathcal{H}
\to CH$ by using a Kuiper map without defining the
pointwise structures. We do not do so and
work with the pointwise maps as far as possible instead. Both the $CH$-action
and the
$CH$-valued fiber product on $\mathcal{H}^{\bullet}$ are defined in Sect. 4.2.
We focus our attention on the continuity and differentiability of
bundles induced by unitary or isometric representations of Lie groups
on infinite dimensional Hilbert or Banach spaces, respectively.
We show that one may speak about $C^{\infty}$-differentiable structures
when a particular smooth principal bundle atlas is chosen, but only on $C^0$-differentiable structures for general representations.
We treat connections, vertical lifts, connectors and induced covariant
derivatives in the context of Banach bundles, showing that the general
approach to connections as given, e.g., in Kol\'a\v{r}, Michor,
Slov\'ak \cite{KMS} is possible to
apply in the case of infinite bundles as well.
Since the oscillator representation
is known to be not differentiable, we do not get
a $C^{\infty}$-structure that is induced canonically. As already mentioned, it is possible to
choose a Kuiper map. The Kuiper structure on $\mathcal{H}$ is defined, in a non-canonical way, as the maximal
smooth atlas containing the Kuiper map.
Some facts from Lie groups' representations based on Bruhat \cite{Bruhat} ($l=\infty$) and
Goodman \cite{Goodman} ($l \in \mathbb{N}_0$) provide
us with further $C^l$-structures as well. We define them in
Sect. 3.2. However, only the Kuiper structure seems to have
some purpose in the theory of $C^*$-indices and consequently, for
$C^*$-elliptic complexes.
Let us mention that we don't use other algebras than the $C^*$-algebras
for several
reasons. The first one is that we would like to stay close to the approach of
quantum physics, for which $C^*$-algebras are fundamental objects until now.
The second one is more pragmatic. As far as we know, the Hodge
theory has not yet been developed for bundles over other algebras in a
satisfactory form. Let us mention that a different approach might be represented
possibly
by considering bundles over Fr\'echet $*$-algebras (see Keyl et al.
\cite{Keyl} for these algebras and their connection to quantum physics). This
approach eliminates problems with the smoothness of the representations.
Further, we give a straightforward proof of the triviality of
the bundle of measurements (Sect. 3.3), which is not easy to find in the
literature, but which is colloquial if one is acquainted with certain constructions in
non-abelian sheaf cohomology adapted for Banach bundles.
In the last section of the paper, we use the Kuiper map of the oscillator
bundle $\mathcal{H}$ to define a
product type connection $\Phi$
which is flat by construction and induces a cochain
complex $d_{\bullet}^{\Phi}=(\Gamma(\mathcal{H}^k), d_k^{\Phi})_{k\in
\mathbb{Z}}$ of exterior covariant derivatives acting between sections of
$\mathcal{H}^k.$
When the symplectic manifold is compact, we prove the central result of the
paper (Theorem 18) that the cohomology groups of $d_{\bullet}^{\Phi}$ are
isomorphic to
the tensor product of the de Rham groups and $H.$
We show also how to use the Hodge
theory
for projective and finitely generated Hilbert $CH$-bundles in the case of
$d_{\bullet}^{\Phi},$ that is elliptic, for to produce a substantially shorter proof of Thm. 18.
\bigskip
In order to specify our terminology, we start with a preamble to
which the reader may come back during the reading.
\bigskip
{\bf \large Preamble}
\begin{itemize}
\item[a1)] Let $X$ be a topological space. The symbol $\mbox{Aut}(X)$ denotes
the set of all homeomorphisms of $X.$
\item[a2)] All vector spaces are considered to be Fr\'echet. Hilbert spaces
are considered with the norm topologies induced by
scalar products and the scalar products are anti-linear in the left variable
(physicists' convention).
\item[a3)] If $X$ is a vector space, maps in $\mbox{Aut}(X)$ are additionally
supposed to be linear.
By $\mbox{End}(X),$ we understand continuous linear maps $X \to X.$ In
particular, $\mbox{End}(X)=B(X)$ where $B(X)$ are bounded operators on $X.$
By a functional on a Banach space, we mean a continuous linear
map of this space into the field of complex numbers
\item[b)] By a $C^l$-differentiable structure on a manifold, we understand a
maximal $C^l$-atlas with respect to the inclusion. A
$C^{\infty}$-differentiable structure is called a smooth structure.
\item[c)] All manifolds are $C^0$-differentiable (topological) Banach
manifolds and all base spaces
of fiber bundles are finite dimensional smooth manifolds. All
bundles are locally trivial.
\item[d)] Representations of Lie groups are considered
continuous in the classical representation theory sense. For a homomorphism
$\sigma: G \to
\mbox{Aut}(X),$ of a Lie group $G$ into the linear homeomorphisms of a normed
vector space $X$ to be a representation means
that $\widetilde{\sigma}: G \times X \to X$ given by $\widetilde{\sigma}: G
\times
H \ni (g,f) \mapsto \sigma(g)f \in X$ is continuous.\footnote{If more details
needed, see, e.g., Knapp
\cite{KnappOverview}.}
\item[e)] When an index of a labeled object or a labeled morphism exceeds its
allowed range, we consider it as zero.
\end{itemize}
\section{$\widetilde{G}$-module and $CH$-module structures}
In this section, we introduce the higher oscillator representation of the metaplectic
group and the Hilbert $CH$-module structure on the underlying space of this representation,
and we prove some properties of these structures.
Let $(V,\omega_0)$ be a real symplectic vector space of dimension
$2n \geq 1$ and $G$ denote the symplectic group $Sp(V,\omega_0).$
Let us choose a compatible positive complex structure $J_0$ on $V.$ By a
compatible positive complex structure,
we mean an $\mathbb{R}$-linear map $J_0:V \to V$ such that $J_0^2 = -\mbox{Id}_V$ and
such that $g_0$ defined by
$g_0(u,v) = \omega_0(J_0u,v),$ $u,v \in V,$ is a positive definite bilinear
form.
Denoting the orthogonal group of $(V,g_0)$ by $O(V,g_0),$ we set $K =
O(V,g_0)\cap Sp(V,\omega_0)$ for the
unitary group of the hermitian vector space $(V,J_0,g_0).$
The symplectic group is known to retract smoothly onto $K$
(see, e.g., \cite{Folland}) the first fundamental group of which is
isomorphic to $\mathbb{Z}.$
By the theory of covering spaces, there exists a unique smooth connected
$2$-fold covering
$\lambda:\tilde{G} \to G$ of $G$ up to a smooth covering isomorphism.
Choosing an element in the $\lambda$-preimage of the neutral element of $G$
and declaring it as the neutral element of $\widetilde{G},$
define a unique Lie group
structure on $\widetilde{G}.$
This is the so called {\it metaplectic group}, denoted
by $Mp(V,\omega_0).$ Further, we set $\widetilde{K} = \lambda^{-1}(K)$ for
the corresponding double cover of the unitary group.
\bigskip
{\bf Basics on oscillator representation:} After Shale \cite{Shale}
and Weil \cite{Weil}, it is known that there exists a specific
non-trivial unitary
representation
of $\widetilde{G},$ called metaplectic, Shale--Weil, Segal--Shale--Weil,
oscillator or symplectic spinor representation. It is called a
spinor representation probably because it
does not descend to a true representation of the symplectic group.
Since
$\widetilde{G}$ is non-compact,
the space carrying this representation is an infinite dimensional Hilbert
space.
The representation intertwines so-called twisted Schr\"{o}dinger representations
and
can be made a representation (of the metaplectic group) on the Hilbert space
$L^2(L)$ of
complex valued square Lebesgue integrable functions defined on a Lagrangian
subspace
$L \subseteq (V,\omega_0).$ The Lebesgue measure is determined by the norm on
$L$ induced
by the inner product $g_0$ restricted to $L \times L.$ For the uniqueness of the
representation,
we refer to Wallach \cite{Wallach}, p. 224. We denote the Hilbert
space
$L^2(L)$ by $\check{H}$ and fix a Segal--Shale--Weil representation, denoting
it by
$\check{\sigma}: \widetilde{G} \to U(L^2(L))$ where $U(X)$
denotes the space of unitary operators on a Hilbert space $X.$
\bigskip
{\bf Notation for dualities $\&$ musical isomorphisms:} Let $(\check{H},
(,)_{\check{H}})$ be a complex Hilbert space, and let us denote its
continuous dual by $H.$
For each $f \in H,$ we denote its dual by $f^{\sharp},$ which is the unique
element in $\check{H}$ such that
$f(v)=(f^{\sharp},v)_{\check{H}}$ for any $v \in \check{H}.$ By the Riesz's
representation theorem for Hilbert spaces, the element $f^{\sharp}$
exists also for non-separable Hilbert spaces.
For any $v \in \check{H},$ we denote by $v^{\flat}$ the element of $H$ defined
by
$v^{\flat}(w)=(v,w)_{\check{H}},$ $v,w \in \check{H}.$ We set $(f,g)_H =
(g^{\sharp},f^{\sharp})_{\check{H}}$ in order the product be anti-linear in the
left variable.
With this notation, $(H,(,)_H)$ is a complex Hilbert space and $\sharp:
\check{H} \to H$ and $\flat:H \to \check{H}$ are
anti-unitary maps, i.e., they are complex anti-linear and intertwine the
corresponding scalar products, i.e.,
$(v^{\flat}, w^{\flat})_H = (w,v)_{\check{H}},$ $(f^{\sharp}, \,
g^{\sharp})_{\check{H}} = (g,f)_H$.
For any pair $u \in \check{H}$ and $v \in H,$ we denote by $u \otimes v$ the
endomorphism $u \otimes v: w \ni \check{H} \mapsto v(w)u \in \check{H}.$
For any linear continuous map $A:X \to Y$ between Hilbert spaces $X$ and $Y,$
$A^*:Y \to X$ denotes its adjoint.
\bigskip
Because we would like to consider right $C^*$-modules, it is more convenient to
use the dual of the Segal--Shale--Weil representation. We denote it
by $\sigma$ and call it the {\it oscillator representation}. Using the
notation defined, we have a homomorphism
$$\sigma: \widetilde{G} \to U(H)$$ The homomorphism is a
representation, i.e., it is continuous in the sense described in
Preamble, Par. d.
Let us notice that the oscillator representation decomposes into two
irreducible subrepresentations, the space of functionals on even and that
of odd functions in $\check{H}.$
See Wallach \cite{Wallach} and Folland \cite{Folland}
for more information on $Mp(V,\omega_0)$ and the Segal--Shale--Weil
representation.
Since $V$ is equipped with $g_0$ and $\omega_0$ (the latter gives an
orientation
to $V$ which is induced by $\omega_0^{\wedge n}$),
we get a canonical scalar product (Hodge type product) on $\bigwedge^k V^*,$
$k=0,\ldots, 2n.$ We denote it by $g_0$ as well.
For $k=0,\ldots, 2n,$ the tensor product
$H^k = \bigwedge^k V^* \otimes H$ is equipped with the norm topology induced by
the canonical Hilbert space inner
product on the tensor product of Hilbert spaces. Let
${\lambda^*}^{\wedge k}$ denote the $k$-th wedge product of
the dual of the representation (and covering) $\lambda:\widetilde{G}\to
Sp(V,\omega_0) \subseteq \textrm{Aut}(V).$
The tensor product
representations
$\sigma^k:\widetilde{G} \to \mbox{Aut}(H^k)$ are given by
$$\sigma^k(g)(\alpha \otimes f) = {\lambda^*}^{\wedge k}(g)\alpha\otimes
\sigma(g)f$$ where
$g\in \widetilde{G},$ $\alpha \in \bigwedge^k V^*$ and $f\in H.$
The above prescription for $\sigma^k$ is extended linearly to
non-homogeneous
elements.
Let us remark that $H^0 = H$ is the representation space of the oscillator
representation $\sigma =\sigma^0.$
We set
$H^{\bullet} = \bigoplus_{k=0}^{2n}H^k
=\bigoplus_{k=0}^{2n} \bigwedge^k V^* \otimes H.$
In a parallel to ordinary spinors, we call
$H^{\bullet}=\bigoplus_{k=0}^{2n} H^k$ the {\it higher oscillator module} and
the representation
$\sigma^{\bullet} = \bigoplus_{k=0}^{2n}\sigma^k$ the {\it higher oscillator
representation}.
\bigskip
Let us denote the space of compact operators on $\check{H}$ equipped with the
operator norm by $CH,$ and consider the representation $$\rho:
\widetilde{G} \to
\mbox{Aut}(CH)$$ of the metaplectic group $\widetilde{G}$ on $CH$
given by $\rho(g)a = \check{\sigma}(g) \circ a \circ \check{\sigma}(g)^{-1},$
where $g \in
\widetilde{G}$ and $a \in CH.$
The representation is well established since $CH$ is a closed two-sided
ideal in the space $\textrm{End}(\check{H})$ of bounded operators on
$\check{H}.$
\bigskip
With the adjoint of a Hilbert space map as the $*$-operation,
$(CH,||\,||_{CH},*)$ becomes a $C^*$-algebra, where $||\,||_{CH}$ is the
restriction
of the operator norm on $B(\check{H})$ to $CH.$
\bigskip
{\bf Definition:}
The {\it right action} $H^{\bullet} \times CH \to H^{\bullet}$
of $CH$ on $H^{\bullet}$ is defined on homogeneous elements by $$(\alpha
\otimes u) \cdot a = \alpha
\otimes (u \circ a)$$ where $\alpha \otimes u \in
\bigwedge^{\bullet} V^* \otimes H$ and $a\in CH.$
It is extended linearly to all elements of $H^{\bullet}.$
The {\it $CH$-product} $(,): H^{\bullet}\times H^{\bullet} \to CH$ is given by
$$(\alpha \otimes u, \beta \otimes v) =
g_0(\alpha,\beta) u^{\sharp} \otimes v$$ where
$\alpha, \beta \in \bigwedge^{\bullet} V^*$ and $u,v \in H.$
The product is extended by linearity. It maps $H^{\bullet} \times H^{\bullet}$
into finite rank operators, especially, into $CH.$
We denote the induced ($C^*$-module) norm on $H^{\bullet}$ by $||\,||.$ It is
given by $||f|| =
\sqrt{||(f, f)||_{CH}},$ $f\in H^{\bullet}.$
\bigskip
{\bf Lemma 1:} For $k=0,\ldots, 2n$ and $\alpha \otimes f \in H^k$
$$||\alpha \otimes f|| = \sqrt{g_0(\alpha,\alpha)}||f||_H$$ In
particular, the norm $||\,||$ coincides on homogeneous elements with the norm induced by the inner
product on $H^k.$
{\it Proof.} The appearance of the first factor, $\sqrt{g_0(\alpha, \alpha)},$
is easy to show. For the second one, we compute
the operator norm $||\,||_{CH}$ of the map $f^{\sharp} \otimes f: \check{H} \ni
k
\mapsto f(k)f^{\sharp},$ where $f \in H.$ On one hand,
we have $||(f,f)(k)||_{\check{H}}^2=||f^{\sharp}f(k)||^2_{\check{H}}$ $=|f(k)|^2
(f^{\sharp},f^{\sharp})_{\check{H}}=
|f(k)|^2(f,f)_H \leq $ $||f||^2_H$
$||k||_{\check{H}}^2$
$||f||_H^2.$
Consequently,
$||(f,f)||_{CH} \leq ||f||_H^2.$
On the other hand,
for $k=f^{\sharp}\neq 0,$ we get
$||f^{\sharp}f(f^{\sharp})||_{\check{H}}^2/$ $/||f^{\sharp}||_{\check{H}}^2
= ||f||^2_H(f,f)_H/||f||^2_H=||f||^2_H.$ Thus,
$||(f,f)||_{CH}=||f||_H^2=(f,f)_H.$
For the norm induced by the $CH$-product on $H$, we obtain
$||f||=\sqrt{||(f,f)||_{CH}}=||f||_H,$
and consequently, $||f||=\sqrt{(f,f)_H}.$
\hfill\(\Box\)
\bigskip
{\bf Lemma 2:} The pair $(H^{\bullet}, (,))$ is a finitely generated
projective Hilbert $CH$-module.
{\it Proof.}
It is obvious that $H^{\bullet}$ is a right $CH$-module.
Proving that $(,)$ is hermitian symmetric and $CH$-sesquilinear, is
a routine
check. Indeed, for $a\in CH,$ $\alpha \otimes f, \beta \otimes h \in H^k,$ and
$u\in
\check{H}$
\begin{align*}
((\alpha \otimes f) \cdot a, \beta \otimes h)(u)&= g_0(\alpha,
\beta)\left((f\cdot a)^{\sharp} \otimes h \right)(u)
=g_0(\alpha, \beta) \left((f \circ a)^{\sharp} \otimes h \right)(u)\\
&=g_0(\alpha,\beta) \left(a^*(f^{\sharp}) \otimes h \right)(u)=g_0(\alpha,
\beta) h(u) a^*(f^{\sharp})\\
&=a^*\left(g_0(\alpha, \beta)h(u) f^{\sharp}\right)
=a^*\left(g_0(\alpha,\beta) \left(f^{\sharp} \otimes h\right) (u) \right)\\
&=\left(a^*\circ \left(g_0(\alpha, \beta) \left(f^{\sharp} \otimes h
\right)\right)\right) (u)\\
&= \left(a^* \circ (\alpha \otimes f, \beta \otimes
h)\right) (u)
\end{align*}
Thus, $(,)$ is anti-linear over $CH$ in the left variable. The hermitian
symmetry
is proved similarly. The right $CH$-linearity follows from the hermitian
symmetry and
left $CH$-linearity.
The projectivity follows from the Magajna theorem (Theorem 1,
Magajna \cite{Magajna}).
For the existence of a finite set of generators, let us choose a basis
$\mathfrak{C}$ of $\bigwedge^{\bullet}V^*$ and take a non-zero vector
$w \in H.$ For any $v \in H^{\bullet},$ there exist scalars
$\lambda_{\mathfrak{a}} \in \mathbb{C}$ and vectors $v_{\mathfrak{a}} \in H,$
$\mathfrak{a} \in \mathfrak{C},$ such that
$v= \sum_{\mathfrak{a} \in \mathfrak{C}} \lambda_{\mathfrak{a}} \mathfrak{a}
\otimes v_{\mathfrak{a}}.$
For each $v' \in H,$ let us set
$p_{v',w}(k)= \frac{v'(k)}{||w||_H^2} w^{\sharp},$ $k \in \check{H}.$
Then $v=\sum_{\mathfrak{a} \in \mathfrak{C}} (\mathfrak{a} \otimes w) \cdot
(\lambda_{\mathfrak{a}} p_{v_{\mathfrak{a}},w}).$
Thus, $\{\mathfrak{a} \otimes w|\, \mathfrak{a} \in \mathfrak{C}\}$ represents a
finite set of generators of
$H^{\bullet}$.
Since the $C^*$-module norm $||\,||$ restricted to $H^0$ is the dual of the
Hilbert
norm on $\check{H}$ (Lemma 1), we see that $H^{\bullet}$ is complete.
\hfill\(\Box\)
\bigskip
{\bf Remark:} A similar proof can be found in \cite{KryslDGA}
where, though, a determination of the inducing scalar product for the
$C^*$-norm is
omitted. The above construction of a Hilbert $CH$-module
structure can be done for an arbitrary Hilbert space.
Note that in the bra--ket convention, $p_{v',w} =
\frac{\ket{w}\bra{v'}}{\braket{w|w}} = \frac{\ket{w}\bra{v'}}{||w||^2}.$
\bigskip
{\bf Definition:} For $a,b \in CH,$ $u \in \check{H}$ and $h \in H,$ let us
set
\begin{tabular}{lll}
$\circ: CH \times CH \to CH$ & $\circ
(a,b)= a \circ b$ & (composition)\\
$\otimes: \check{H} \times H \to CH$ &
$ \otimes(u,h) = u \otimes h$ & (tensor product) \\
$ev: H \times \check{H} \to
\mathbb{C}$ & $ev(h,u) = h(u)$ & (evaluation) \\
\end{tabular}
\bigskip
{\bf Lemma 3:} The $CH$-action, the composition, the tensor
product and the evaluation
are $\widetilde{G}$-equivariant. The $CH$-product $(,): H \times H \to CH$ and
the induced norm $||\,||:H^{\bullet} \to \mathbb{R}$ are
$\widetilde{K}$-equivariant.
For any $g \in \widetilde{G},$ $v \in \check{H}$ and $h \in H$
\begin{eqnarray} \label{formule}
\rho(g) (v \otimes h) = \check{\sigma}(g)v \otimes \sigma(g)h
\end{eqnarray}
{\it Proof.} For $g \in \widetilde{G},$ $v, w \in \check{H}$ and $h \in H$
\begin{align*}
\rho(g)(v \otimes h)(w) &= \left(\check{\sigma}(g)\circ (v \otimes h)\circ
\check{\sigma}(g)^{-1}\right)(w)\\
&=\check{\sigma}(g)(v)h(\check{\sigma}(g)^{-1}w) \\
&= (\check{\sigma}(g)v \otimes \sigma(g)h)(w)
\end{align*}
obtaining formula (\ref{formule}).
(\ref{formule}) is used to prove the $\widetilde{G}$-equivariance
of the tensor product readily.
We check the $\widetilde{G}$-equivariance of the action $H^{\bullet}\times CH
\to H^{\bullet}$ and the $\widetilde{K}$-equivariance of the
$CH$-product.
For $g \in \widetilde{G},$ $f =\alpha \otimes u \in H^k,$ $k=0,\ldots, 2n,$ and
$a \in CH$
\begin{align*}
(\sigma^k(g)f) \cdot (\rho(g)a) &= \left(\lambda^{* \wedge k}(g)\alpha \otimes
\sigma(g) u\right) \cdot (\check{\sigma}(g) \circ a \circ
\check{\sigma}^{-1}(g))\\
& = \lambda^{* \wedge k}(g)\alpha \otimes \left(\sigma(g) u \circ
(\check{\sigma}(g) \circ a \circ \check{\sigma}^{-1}(g))\right)\\
& = \lambda^{* \wedge k}(g)\alpha \otimes \left((u \circ
\check{\sigma}^{-1}(g))\circ (\check{\sigma}(g) \circ a \circ
\check{\sigma}^{-1}(g))\right)\\
& =\lambda^{* \wedge k}(g)\alpha \otimes (u \circ a \circ
\check{\sigma}^{-1}(g))\\
& =\lambda^{* \wedge k}(g)\alpha \otimes \sigma(g)(u \circ a) =
\sigma^k(g)(\alpha \otimes u \cdot a) = \sigma^k(g)(f \cdot a)
\end{align*}
The equivariance on non-homogeneous elements follows from linearity.
Let $g \in \widetilde{K}$ and $h = \beta \otimes v \in H^k.$ Since $g \in
\widetilde{K},$ we have \\
$g_0(\lambda^{* \wedge k}(g)\alpha, \lambda^{* \wedge k}(g)\beta)= g_0(\alpha,
\beta).$ Further
\begin{align*}
(\sigma^k(g)f, \sigma^k(g)h) &= g_0(\lambda^{* \wedge k}(g)\alpha,\lambda^{*
\wedge k}(g)\beta)\left((\sigma(g)u)^{\sharp} \otimes (\sigma(g)v)\right)\\
& = g_0(\alpha, \beta) \check{\sigma}(g)u^{\sharp} \otimes \sigma(g)v \\
& = g_0(\alpha, \beta) \rho(g)(u^{\sharp} \otimes v) = \rho(g) g_0(\alpha,
\beta) (u^{\sharp} \otimes v) \\
& = \rho(g)(\alpha \otimes u, \beta \otimes v) = \rho(g)(f,h)
\end{align*}
where formula (\ref{formule}) is used in the second row.
For the norm, consider $||\sigma^k(g)f||^2=||(\sigma^k(g)f,\sigma^k(g)f)||_{CH}
=||\rho(g)(f,f)||_{CH}=||(f,f)||_{CH}=||f||^2,$ $g\in \widetilde{K},$
since $\rho(g)$ is isometric.
$\widetilde{G}$-equivariances of $\circ$ and $ev$ can be proved in a similar
way.
\hfill\(\Box\)
\section{Geometric lifts of $\widetilde{G}$- and $CH$-module structures}
In this chapter, we introduce the higher oscillator bundle and the bundle of
measurements. We investigate continuity and smoothness of canonical
atlases of these bundles. For this, we examine the
Segal--Shale--Weil representation from the analytic point of view.
\bigskip
For a symplectic manifold $(M^{2n},\omega),$ we consider
$$\mathcal{Q} = \{ A: V \to T_m^*M| \, \omega_m(Au, Av) = \omega_0(u,v) \mbox{ for any } u,v \in V, m \in M \}$$
together with the bundle projection $\pi_Q(A)=m$ if and only if $A:V \to T_m^*M.$
For the topology on $\mathcal{Q},$ we take the final one with respect to
inverses of all bundle maps for the projection $\pi_Q.$
The right action of $G=Sp(V,\omega_0)$ on $\mathcal{Q}$ is given by the composition from the right, i.e.,
$(A, g) \mapsto A \circ g$ for $A\in \mathcal{Q}$ and $g \in G.$ This make $\pi_Q: \mathcal{Q} \to M$ a principal
$G$-bundle.
\bigskip
Let us consider
pairs $(\pi_P,\Lambda),$ where $\pi_P:\mathcal{P}\to M$
is a principal $\widetilde{G}$-bundle over $M$
and $\Lambda: \mathcal{P} \to \mathcal{Q}$
is a smooth map. Such a pair is
called a {\it metaplectic structure}
if $\Lambda$ intertwines the actions of $G$ and $\widetilde{G}$ on
$\mathcal{Q}$ and $\mathcal{P},$ respectively, with respect to the homomorphism
$\lambda: \widetilde{G} \to G$.
In other words, the following diagram has to commute.
$$\begin{xy}\xymatrix{
\mathcal{P} \times \widetilde{G} \ar[dd]^{\Lambda\times \lambda} \ar[r]&
\mathcal{P} \ar[dd]^{\Lambda} \ar[dr]^{\pi_P} &\\
& &M\\
\mathcal{Q} \times G \ar[r] & \mathcal{Q} \ar[ur]_{\pi_Q} }\end{xy}$$
(The horizontal arrows represent the right actions of the groups
$\widetilde{G}$ and
$G$ on the total spaces $\mathcal{P}$ and $\mathcal{Q},$
respectively.) The bundle $\mathcal{P}$ is called a {\it metaplectic bundle}.
The bundle $\mathcal{Q}$ is isomorphic to the quotient of $\mathcal{P}$ by the
action of the $2$-element group $\mbox{Ker} \, \lambda.$
Further, we may consider a version of a metaplectic structure over the complex numbers
$\mathbb{C}$ as well,
getting a so-called $Mp^c$-structure. Notice that a metaplectic structure is formally similar to that of a spin structure.
\bigskip
{\bf Remark:}
Let us mention that a metaplectic
structure exists on a
symplectic manifold $(M,\omega)$ if and only if
the second Stiefel--Whitney class of $TM$ vanishes. Their set modulo an
appropriate equivalence relation is isomorphic to the singular
cohomology group $H^1(M,\mathbb{Z})$ as a set. See Kostant
\cite{Kostant}.
Note that the $Mp^c$-structures are known to exist
without any restriction on the topology of the underlying symplectic manifold
(Forger, Hess \cite{FH} or Robinson, Rawnsley
\cite{RR}). Compare approaches of
Habermann, Habermann in \cite{HH} and Cahen, Gutt, La Fuente-Gravy,
Rawnsley in \cite{CGLR}.
\bigskip
Similarly as in the orthogonal spinor case,
we obtain by induction from representations $\sigma^k:\widetilde{G} \to
\mbox{Aut}(H^k)$ and the principal bundle
$\mathcal{P} \to M$ bundles
$$p_k: \mathcal{H}^k = \mathcal{P} \times_{\sigma^k} H^k \to M$$ where
$k=0,\ldots, 2n.$ We call the bundle $\mathcal{H}^{\bullet} = \bigoplus_{k=0}^{2n}
\mathcal{H}^k$ the {\it higher oscillator bundle}.
The basic oscillator bundle $\mathcal{H} = \mathcal{H}^0 = \mathcal{P}
\times_{\sigma} H$ coincides with the dual of the Kostant's or symplectic
spinor bundle as it is called by Habermann, Habermann \cite{HH}.
Their continuity and
differentiability are discussed in the next sections.
\bigskip
{\bf Definition:} The Banach bundle $\mathcal{CH}$ associated to $\mathcal{P}$
via $\rho:
\widetilde{G} \to \mbox{Aut}(CH),$ i.e., $$\mathcal{CH} =
\mathcal{P} \times_{\rho} CH$$ is called the {\it bundle of measurements}.
\bigskip
The bundle $\mathcal{CH}$ exists regardless of the existence of the
metaplectic structure because $\mbox{Ker}\, \lambda =\mbox{Ker}\, \rho.$
\bigskip
{\bf Remark} (Terminology related to measurements in Quantum theory):
The names `bundle of measuring devices' and `bundle of filters' for the bundle of
measurements
$\mathcal{CH}$ seem appropriate as well. The word
{\it filter} is used in quantum theory for a theoretical
measuring device. See Ludwig \cite{Ludwig} for the role of
filters, and still topical texts of
Birkhoff, von Neumann \cite{BirkhoffNeumann} and von Neumann \cite{NeumannKM}
for a mathematical framework for quantum (and also classical)
physical theories.
In each
point $m$ of a symplectic manifold $(M,\omega)$ (a `curved' phase space), we
imagine
a measuring
device (filter) constituting of filtering channels, which are represented by
elements of $\mathcal{CH}_m,$ the fiber in $m.$
Assuming that for any quantum measurement's result,
there is a corresponding measuring device, $\mathcal{CH}_m$ presents a stack
for all possible results of quantum measurements that can be
obtained on a system found classically in the point $m.$
An ideal filter in point $m$ is a measuring device corresponding to a rank $1$
operator in $\mathcal{CH}_m$. Let $k: \mathcal{CH} \to M$ be the bundle
projection of the bundle of measurements onto $M.$
Suppose that we have measuring devices filling-up a submanifold
$\iota: M' \hookrightarrow M$ (possibly only a discrete set of points).
Then we have the pull-back bundle $k_{M'}: \iota^*(\mathcal{CH}) \to
M'.$ In real measurements, each measuring device corresponds to a
finite linear combination of rank 1
operators in the fibers $(\iota^*(\mathcal{CH}))_m, m\in M'.$
\subsection{Continuity of $G$-structures}
We make a comment on a canonical continuity of atlases of induced Banach
bundles.
Let us recall that for a principal
$Q$-bundle $\pi: \mathcal{R} \to M$
and a representation $\kappa: Q \to \mbox{Aut}(X)$ of
a Lie group $Q$ on a topological vector space $X$, the induced
(associated) bundle is the quotient $\mathcal{F}=(\mathcal{R} \times X)
/\simeq$ together with the appropriate bundle projection onto $M$.
The equivalence
relation $\simeq$ is defined
for any $r, s \in \mathcal{R}$ and $v,w \in X$ by
$(r,v) \simeq (s,w)$ if and only if there exists an element $g\in Q$ such that
$s = r g^{-1}$ and $w=\kappa(g)v.$ The topology induced on the quotient
is the final topology, i.e.,
the finest one for which the canonical projection $q: \mathcal{R} \times X \to
(\mathcal{R} \times X)/ \simeq$ is continuous.
The bundle projection $p:\mathcal{F} \to M$ is given by $p([(r,v)])=\pi(r),$
where $[(r,v)] \in \mathcal{F}.$
For $U\subseteq M,$ let $\mathcal{F}_U$ denote the set $p^{-1}(U).$
Especially,
$\mathcal{F}_{\{m\}},$ denoted by $\mathcal{F}_m,$ is the fiber of
$\mathcal{F}$
in $m\in M.$
Let us equip the higher oscillator bundle and the bundle of measurements
with the final topologies and prove that they form Hilbert and Banach
bundles, respectively. Actually, it is convenient to prove a more general
theorem.
\bigskip
{\bf Theorem 4:} Let $\kappa: Q \to \mbox{Aut}(X)$ be an isometric or a unitary
representation of a Lie group $Q$ on Banach or Hilbert space $X,$
respectively. Let $p:\mathcal{F} \to M$ be the associated bundle to a
principal $Q$-bundle $\pi: \mathcal{R} \to M.$ Then $p$ is canonically a
continuous Banach or a Hilbert bundle, respectively.
{\it Proof.} Let us take a principal bundle atlas
$(U_{\alpha},\phi_{\alpha})_{\alpha \in I}$ in the differentiable structure of
the
principal bundle $\pi.$ In particular, $\phi_{\alpha}: \pi^{-1}(U_{\alpha})
\to U_{\alpha} \times Q.$
For to get a bundle atlas on $p:\mathcal{F} \to M,$ we define maps
$\psi_{\alpha}': U_{\alpha} \times X \to p^{-1}(U_{\alpha})$
by $\psi_{\alpha}'(m,v)= q(\phi^{-1}_{\alpha}(m,e),v),$ where $e$ is the unit
element in $Q,$ $m \in U_{\alpha}$ and $v \in X.$ The maps $\psi'_{\alpha}$
are
continuous because
$\phi^{-1}_{\alpha}$ and $q$ are continuous. They are fiber preserving in the
sense that $p \circ \psi_{\alpha}' = \mbox{pr}_1,$ where
$\mbox{pr}_1: U_{\alpha} \times X \to U_{\alpha}$ is the projection onto the
first factor in the product. Their inverse maps, $\psi_{\alpha}:
p^{-1}(U_{\alpha}) \to
U_{\alpha} \times X$ are given by $\psi_{\alpha}([(r,v)])=(\pi(r),
\kappa(g^{-1})v)$ where $g$ is the unique element of $Q$ such that
$rg=\phi_{\alpha}^{-1}(m,e)$
and $m=\pi(r).$ The independence on the chosen representative of an
equivalence
class
follows from the definition of the associated bundle. Let $U \subseteq
U_{\alpha}$ and $Y \subseteq X$ be open sets.
The $\psi_{\alpha}$-preimage of $U \times Y$ is
$S=\{[(\phi^{-1}_{\alpha}(m,e),w)]|\, m \in U, w \in Y\}.$
Consequently, the $q$-preimage $q^{-1}(S) = \{(\phi_{\alpha}^{-1}(m,e)g^{-1},
\kappa(g)w)|\, m \in U, g \in Q, w \in Y\}.$
The preimage can be written as $\bigcup_{g\in Q}(\phi_{\alpha}^{-1}(U,e)g^{-1},
\kappa(g)Y).$\footnote{More properly, the image equals to
$\bigcup_{g \in Q}\{(\phi_{\alpha}^{-1}(m,e)g^{-1},
\kappa(g)y)|\, m \in U, y \in Y\}.$}
For each $g\in Q,$ both of the sets $\{\phi_{\alpha}^{-1}(m,e)g^{-1}|\, m \in
U \}$ and $\{ \kappa(g)y|\, y \in Y \}$ are open since $\phi_{\alpha}^{-1}$
is a homeomorphism and the map $\kappa(g)$ is a homeomorphism as
well. Due to the definition of the topology on Cartesian products, $q^{-1}(S)$
is
open and thus $\psi_{\alpha}$ is continuous.
Summing-up, $\psi_{\alpha}$ is a homeomorphism and
$\mathfrak{A}=(U_{\alpha},\psi_{\alpha})_{\alpha \in I}$ is
a bundle atlas.
Choosing a different atlas in the differentiable structure of
$\mathcal{R},$
leads to an induced atlas $\widetilde{\mathfrak{A}}=(\widetilde{U}_{\alpha},
\widetilde{\psi}_{\alpha})_{\alpha \in \widetilde{I}}$ that is different from
$\mathfrak{A}$ in general. However, the transition functions $\psi_{\alpha
\beta}$ for $\mathfrak{A}
\cup \widetilde{\mathfrak{A}}$ are given by
composition of the representation with the
transition functions $(\phi_{\alpha \beta})_{\alpha, \beta}$ of the principal
bundle, i.e., $\psi_{\alpha \beta} = \kappa \circ \phi_{\alpha \beta}.$
Since $\kappa$ is continuous (as a map $Q \times X \to X)$ and unitary, it is
continuous
as a map $Q \to U(X)$ where $U(X)$ is considered with the strong topology (see
Schottenloher \cite{Schotten} or
Knapp \cite{KnappOverview}, pp. 10 and 11). Thus we have defined an independent continuous
bundle structure.
Suppose that $\kappa$ is a unitary representation on a Hilbert space $X$. The
fiber-wise scalar product is defined by $([(r,v)],[(r,w)])_S = (v,w)_X$
where at the right-hand side, the scalar product on $X$ is meant. The
independence of the scalar product on the chosen representatives follows from
the fact that the representation is unitary.
Let $|| \,||_S:\mathcal{F} \to \mathbb{R},$ given by $||a||_S=\sqrt{(a,a)_S},$
$a\in \mathcal{F},$ be the induced norm, and let $(a_n)_n$ converge to an
element $a$ in the
quotient topology of $\mathcal{F}.$ By definition of the quotient topology,
there exists a
selector $(b_n)_n$ on $\{q^{-1}(\{a_n\})|\, n \in \mathbb{N}\}$ which converges
in
$\mathcal{R} \times X.$ Denoting its limit by $b,$ we have that
1) $c_n=\mbox{pr}_2(b_n)$ converges to $c=\mbox{pr}_2(b)$ in $X$, and 2)
$a=[b],$ where $\mbox{pr}_2$ denotes the projection $\mathcal{R}
\times X \to X$ onto the second factor. Summing-up,
$||a_n||_S=\sqrt{(a_n,a_n)_S}=\sqrt{(c_n,c_n)_X}
\rightarrow ||c||_X=||[b]||_S =||a||_S.$ Thus, $||\,||_S$ is continuous.
The continuity of $(,)_S$ in each of the arguments follows from the
polarization
identity.
Rewriting the definition of $(,)_S$, we have
$(q(r,v),q(r,w))_S=(v,w)_X.$
Further $(\psi_{\alpha}'(m,v),\psi_{\alpha}'(m,w))_S=
(q(\phi_{\alpha}^{-1}(m,e),v), q(\phi_{\alpha}^{-1}(m,e),w))_S=$
$(v,w)_X$, \\ proving the unitarity of $\psi_{\alpha}'.$
Consequently, $\mathcal{F}$ is a Hilbert bundle.
Similarly one proceeds in the case of the isometric
Banach representation.
\hfill\(\Box\)
\bigskip
Let us call any atlas $(U_{\alpha}, \psi_{\alpha})_{\alpha \in I}$ on an
induced
bundle $\mathcal{F}$ constructed as in the preceding proof
the {\it canonically induced atlas} or, in more detail, the
atlas canonically induced
by the principal bundle atlas $(U_{\alpha}, \phi_{\alpha})_{\alpha \in I}.$ As
follows from the above proof,
canonical atlases induced by different but compatible
atlases
are $C^0$-compatible. In particular, one can speak about the induced
$C^0$-differentiable structure.
\bigskip
{\bf Corollary 5:} For a symplectic manifold $(M,\omega)$ and a metaplectic
structure
$(\pi_P, \Lambda)$ on it, the higher oscillator bundle
and the bundle
of measurements, both equipped with the canonically induced atlases by a
principal bundle atlas, are continuous Banach bundles. The basic oscillator
bundle is a
continuous Hilbert bundle.
\subsection{Differentiable structures on oscillator bundle}
A substantial obstacle for analysis on induced Banach or Hilbert bundles of
infinite rank is that they need not be smooth or not even $l$-times
differentiable for some $l \in \mathbb{N}.$
When speaking about $C^l$-sections, one has to take care to which atlas the
degree of differentiability refers.
\bigskip
Let us recall that for a representation $\kappa: Q \to \mbox{Aut}(X)$ of a Lie
group $Q$ on a
Banach space $X,$ a vector $v \in X$ is called smooth if
$Q\ni g \mapsto \kappa(g)v \in X$ is smooth, i.e., it has each Fr\'echet
differential in each point $g.$
\bigskip
Let $\phi_{\alpha \beta}: U_{\alpha} \cap U_{\beta} \to Q$ be a
cocycle of transition functions of a principal $Q$-bundle $\mathcal{R}$
corresponding to a
smooth atlas $(U_{\alpha}, \phi_{\alpha})_{\alpha \in I}$ of $\mathcal{R}.$
Using the definition of the canonically induced atlas, we obtain that its
transition maps $\psi_{\alpha \beta}: (U_{\alpha} \cap U_{\beta}) \times X \to
(U_{\alpha} \cap U_{\beta}) \times X$ are given by
$\psi_{\alpha \beta}(m,v) = (m, \kappa(\phi_{\alpha \beta}(m))v).$
We agree that a Banach bundle atlas is called smooth if the transition
functions $\psi_{\alpha \beta}$ considered as maps of $U_{\alpha} \cap
U_{\beta} \to \mbox{Aut}(X) \subseteq \mbox{End}(X)$ are
smooth. On $\mbox{Aut}(X),$ we consider the topology
induced from the strong operator topology on $\mbox{End}(X).$
This happens if and only if $m \mapsto
\kappa(\phi_{\alpha \beta}(m))$ is
smooth. For that it is sufficient the representation to be smooth as a map into
$\mbox{End}\,(X)$ with the strong topology.
Thus we get
\bigskip
{\bf Lemma 6:} The bundle atlas canonically induced by a principal bundle
atlas
is smooth if the inducing
representation is
smooth.
\bigskip
The opposite implication does not hold. For it, take an arbitrary
manifold $M$, and the metaplectic group $Mp(V,\omega)$ for the Lie group $Q.$
Let us consider the product bundle $M \times Q \to
M$ as the principal $Q$-bundle and the oscillator representation $\sigma$
as the representation $\kappa.$
For the smooth atlas on the product bundle take the one containing the
single chart $U = M$ and the only bundle map $\phi(m,g)=(m,g),$ where $m\in M$
and $g\in Q.$ The set of transition functions of this atlas consists just of
the constant map $m \mapsto e,$ where $e$ is the unit element in the metaplectic group.
The set of transition functions of the canonically induced atlas on
$\mathcal{F} =
(M \times Q) \times_{\sigma} \check{H}$ contains only $\psi(m,v)=(m,\kappa(e)v)
= (m,v)$, which is clearly smooth, although the
representation is known not to be. (See Borel, Wallach \cite{BorelWallach}, p.
153, or below.)
\bigskip
{\bf Smooth Hilbert bundle \`a la Kuiper}
\bigskip
Because the unitary group of $H$ is contractible in the strong operator
topology (Dixmier, Douday \cite{DixmierDouady})\footnote{It is contractible
also in the operator norm topology (Kuiper \cite{Kuiper}).},
there exists an isomorphism $J: \mathcal{H} \to
M \times H$ of topological vector bundles. We declare the smooth structure
of $\mathcal{H}$ as the
maximal set of smoothly compatible atlases containing the chart
$(M,J)$ and
call it the {\it Kuiper structure} induced by $J$.
\bigskip
{\bf Smooth Fr\'echet bundle \`a la Bruhat}
\bigskip
Let $S(L)$ denote the Schwartz space of rapidly decreasing smooth functions on
$L.$ Vectors in $\check{H} \setminus S(L)$ are exactly the ones which are
not smooth for $\check{\sigma}$ (Borel, Wallach \cite{BorelWallach}).
Consequently, smooth vectors of $H^k$ acted on by $\sigma^k$ form
spaces $S^k = \bigwedge^{k}V^* \otimes
S,$ where $S \subseteq H$ denotes the $\flat$-image of $S(L) \subseteq
L^2(L)$ in $L^2(L)'.$
Note that
$S$ is not complete in the norm topology inherited from $H.$ To get a complete space, we
retopologize it as it is usual in representation theory (see below).
We recall and combine some known facts from Lie groups' representations (based on
the Bruhat's dissertation thesis \cite{Bruhat})
and some known facts specific for the Segal--Shale--Weil representation.
\begin{itemize}
\item[1)] Representations $\sigma^k$ preserve the space of smooth vectors
$S^k.$
\item[2)] Let us denote the restrictions $\sigma^k$ to $S^k$ by $\sigma^k_0$
and consider $S^k$ to be equipped with the topology inherited by the
injection $S^k \to C^{\infty}(\widetilde{G}, H^k),$ $S^k \ni v \mapsto (\pi^v:\widetilde{G} \ni g \mapsto \sigma^k(g)(v) \in H^k).$
Here, $C^{\infty}(\widetilde{G}, H^k)$ is considered with the compact $C^{\infty}$-topology.
(See, e.g., Warner \cite{Warner}, p. 484 for a definition of
compact $C^{\infty}$-topology if necessary.) The restricted representation
$\sigma_0^k$ is continuous in the new topology (\cite{Warner}, Sect. 4.4.).
\item[3)] The compact $C^{\infty}$-topology on $S$ coincides with the classical
semi-norm Fr\'echet topology on this space (see Borel, Wallach \cite{BorelWallach}).
\item[4)] Since (a) $\sigma^k_0$ is a representation on $S^k$ with
the compact $C^{\infty}$-topology, which is Fr\'echet, and since (b) the space
$(E_{\infty})_{\infty}$ of smooth vectors of the space $E_{\infty}$ of smooth
vectors of a continuous representation
on a Fr\'echet space $E$\footnote{Both $E_{\infty}$ and $(E_{\infty})_{\infty}$
are considered with the compact $C^{\infty}$-topology.} is topologically
isomorphic to $E_{\infty},$ we conclude that $\sigma_0^k: \widetilde{G} \to
\mbox{Aut}\,(S^k)$ are smooth maps
when $\mbox{Aut}(S^k) \subseteq \mbox{End}(S^k)$ is considered with the strong
topology.
Technically speaking, this is because the assumption of Thm. 4.4.1.7
(see Warner \cite{Warner}) is satisfied, and thus Remark (3) at pages 258 and 259
of \cite{Warner} applies.
\end{itemize}
Taking representations $\sigma_0^k$ instead of $\sigma^k,$ we get
Fr\'echet bundles $\mathcal{S}^k= \mathcal{P} \times_{\sigma^k_0} S^k.$
By what was mentioned in item 4 above and by Lemma 6, each of the
bundles $\mathcal{S}^k$ are smooth with respect to the atlas induced
canonically by any smooth principal bundle atlas (on $\mathcal{P}).$
We call the bundles $\mathcal{S}^k \to M$ with the smooth atlases {\it Bruhat structures}. The atlases
are constructed as in the proof of Theorem 4.
\bigskip
{\bf $C^0$-Hilbert bundle \`a la Goodman}
\bigskip
We can consider spaces $W^l$ of $C^l$-vectors in $H$ as well, $l \geq
1,$ obtaining representations $\sigma^l: \widetilde{G} \to
\mbox{Aut}(W^l)$. See Goodman \cite{Goodman} or Neeb
\cite{Neeb}. Consider the tensor product representations
$\sigma^{k,l}:\widetilde{G} \to \mbox{Aut}(\bigwedge^k V^* \otimes W^l).$
\bigskip
From the known form of the differential of the Segal--Shale--Weil
representation (see Habermann \cite{HabermannAn}, or Kirillov \cite{Kirillov},
p. 184 for the case $n=1$) and from the definition of the so-called
Hermite--Sobolev spaces (as defined, e.g., in Bongioanni, Torrea \cite{Bong}),
spaces $W^l$ are isomorphic to the Hermite--Sobolev spaces denoted
by $W^{2l}_2$ or sometimes $W^{2l,2}$ in \cite{Bong}.
Moreover, the compact $C^l$-topology on $W^l,$ as described in
Goodman \cite{Goodman}, coincides with the norm topology, generated by the oscillator ladder operators, considered in Bongioanni, Torrea \cite{Bong}.
(Notice that it is easy to see that spaces $W^l,$ with this topology, are
Hilbert spaces as well as that $\sigma^{0,l}$ is not $C^1.$)
\bigskip
The associated bundles $\mathcal{W}^{k,l} =\mathcal{P}
\times_{\sigma^{k,l}}(\bigwedge^k V^* \otimes W^l)$ when equipped with the
canonically induced atlases are Hilbert
bundles with in general, only continuous structures. We call
these bundles $\mathcal{W}^{k,l} \to M$ with the continuous atlases the
{\it Goodman structures}.
\bigskip
{\bf Remark:} In the case of the Bruhat or the Goodman structures, we
either loose the completeness with
respect to a norm or the smoothness of the representation, respectively.
\subsection{Differentiable structure on bundle of measurements}
Operations $\sharp: H \to \check{H}$ and $\flat: \check{H} \to H$ lift to
bundles
$\mathcal{H}\to M$ and $\check{\mathcal{H}} \to M,$ respectively. We denote
them by the same symbols, i.e.,
$\sharp: \mathcal{H} \to \check{\mathcal{H}}$ and $\flat: \check{\mathcal{H}}
\to \mathcal{H}.$
We shall denote the operator $\mbox{Id}_M \times \sharp: M \times H \to M
\times
\check{H}$ by $\sharp_M$
and the operator $\mbox{Id}_M \times \flat: M \times \check{H} \to M \times H$
by $\flat_M.$
Let us consider the map
$$\otimes_{\mathcal{CH}}: \check{\mathcal{H}} \times_{M} \mathcal{H} \to
\mathcal{CH} \mbox{ defined by } \, [(r,v)]\otimes_{\mathcal{CH}} [(r,f)] =
[(r, v \otimes f)]$$ where $v\in \check{H}, f\in H$ and $r\in \mathcal{P}.$
Since the operator norm topology on $CH$ is coarser than
the Hilbert tensor product topology on $\check{H} \otimes H,$ described
equivalently by the Hilbert--Schmidt norm,
$\otimes_{\mathcal{CH}}$ is
continuous. For a moment, let us denote by $\underline{\mathbb{C}}$ the trivial
complex line
bundle $M \times \mathbb{C}$ over $M$. This bundle is isomorphic to
the
bundle associated to $\mathcal{P}$ by the
trivial representation of $\widetilde{G}$ on $\mathbb{C}.$ Further, we set
\begin{align*}
ev_{\mathcal{H}}: \mathcal{H} \times_M \check{\mathcal{H}} \to
\underline{\mathbb{C}}& \qquad
ev_{\mathcal{H}}([(r,f)],[(r,v)])= (\pi_P(r),f(v))\\
\end{align*}
Instead of
$\mbox{pr}_2\left(ev_{\mathcal{H}}([(r,f)],[(r,v)])\right),$ we write
$[(r,f)]([(r,v)]),$ where $\mbox{pr}_2: M\times
\mathbb{C} \to \mathbb{C}$ denotes the projection onto the second factor.
The operations $\sharp: \mathcal{H} \to \check{\mathcal{H}}, \flat:\mathcal{H}
\to \check{\mathcal{H}}, \otimes_{\mathcal{CH}}:\check{\mathcal{H}}\times_M
\mathcal{H}
\to \mathcal{CH}$ and $ev_{\mathcal{H}}:\mathcal{H}\times_M\check{\mathcal{H}}
\to\underline{\mathbb{C}}$ are
well defined. The proof of this is based on the $\widetilde{G}$-equivariance
(Lemma 3) and goes in the same lines as the one of the
correctness of definitions of the $\mathcal{CH}$-product and the
$\mathcal{CH}$-action, explained below (Sect. 4. 1).
\bigskip
{\bf Notation for bundle $\underline{H}:$} For the trivial bundle
$\underline{H} = M \times H \to M,$ we set $(m,f')+_M (m,f'') = (m,f'+f''),$
$t\odot_M (m,f) = (m,tf),$ and $((m,f'),(m,f'')) =
(f',f'')_H \in \mathbb{C}$
where $t\in \mathbb{R},$ $f,f',f'' \in H,$ and $m\in M.$
This notation may look like superfluous. However, we use
it later (sect. 4.3) when it appears to be helpful.
\bigskip
{\bf Normalized Kuiper maps:}
Since $\sigma$ is a unitary representation of $\widetilde{G}$ on $H,$ the
associated bundle $\mathcal{H}$ is a vector bundle with fiber $H$ and structure
group
$U(H)$ with the strong operator topology (Thm. 4).
The set of isomorphism classes of vector bundles with fiber $H$ and structure group $U(H)$ is in a set bijection
with the first \v{C}ech cohomology group $H^1(M, U(H))$ (see, e.g., Raeburn, Williams \cite{Raeburn}, pp. 96--97).
Because the unitary group of an infinite dimensional Hilbert space is contractible,
the \v{C}ech cohomology group consists only of one element. Consequently, each bundle must be isomorphic to
the trivial one. This means that by definition (p. 93, \cite{Raeburn}), there
is a fiber preserving homeomorphism $J_N$ of $\mathcal{H}$ and $\underline{H}$
which is a map into $U(H)$ when considered composed with bundle maps. In particular, this implies that $J_N$ is a unitary map in each fiber.
We call such a map a normalized Kuiper map. {\bf By a Kuiper map, we shall only mean a normalized Kuiper map}, i.e.,
a unitary isomorphism.
\bigskip
Since the bundle of measurements $\mathcal{CH}$ is a specific associated
bundle to the principal bundle $\mathcal{P},$
the transition functions of $\mathcal{CH}$ are compositions of
the transition functions of $\check{\mathcal{H}},$ the representation
$\sigma,$ and the $\mbox{Ad}$-representation. Consequently, $\mathcal{CH}$ is a
trivial bundle by the Dixmier--Douady theory. (See Raeburn, Williams \cite{Raeburn}, Thm. 4.85
(a), pp. 109--110.)
Because the cited result is often scattered through texts,
we give an elementary proof based only on the triviality of
$\mathcal{H}$ and $\check{\mathcal{H}}$ and
functional analysis' methods.
\bigskip
{\bf Theorem 7:} Let $(M,\omega)$ be a symplectic manifold admitting a
metaplectic structure $(\pi_P, \Lambda).$
The bundle of measurements $\mathcal{CH} \to M$ and the
product bundle $M \times CH \to M$ are isomorphic as Banach bundles.
Consequently, the bundle of measurements is trivial.
{\it Proof.}
Let us consider the Banach bundle $\check{\mathcal{H}}=
\mathcal{P} \times_{\check{\sigma}} \check{H}$ and let $J$ be a Kuiper map of
the bundle $\mathcal{H} \to
M$, i.e., $J: \mathcal{H} \to M \times H$ is an isomorphism of continuous
Hilbert bundles which is isometric on each fiber.
It induces a map $\check{J}: \check{\mathcal{H}} \to M \times \check{H}$
(trivializing $\check{\mathcal{H}}$)
given by the formula
$\check{J}(\tilde{v}) = (J(\tilde{v}^{\flat}))^{\sharp_M},$ where $\tilde{v}
\in
\check{\mathcal{H}}.$
Because $J$, $\sharp_M$ and $\flat$ are continuous, so is $\check{J}.$
The inverse to $\check{J}$ is given by $\check{J}^{-1}(v) =
(J^{-1}(v^{\flat_M}))^{\sharp},$ $v \in M \times \check{H}.$
Since $\flat_M, J$ and $\sharp$ are continuous, we get that $\check{J}$
is a homeomorphism and thus $(M,\check{J})$ is a bundle chart for
$\check{\mathcal{H}}.$
For bundle rank 1 operators, we set
$$J_{CH}^0([(r, v \otimes f)])=
\left(\pi_P(r),\check{\mbox{pr}}_2(\check{J}([(r,v)])) \otimes
\mbox{pr}_2(J([(r,f)]))\right)$$
where $f \in H,$ $v \in \check{H},$
$r \in \mathcal{P},$ $\mbox{pr}_2: M \times H \to H$ and $\check{\mbox{pr}}_2:
M
\times \check{H} \to \check{H}$ denote the
projections onto the second factors of the respective products.
To prove that $J_{CH}^0$ is well defined, we use formula (\ref{formule}) from
Lemma 3. Indeed,
\begin{align*}
J_{CH}^0([(rg ^{-1}, \rho(g)&(v \otimes f))])= J_{CH}\left([(rg^{-1},
\check{\sigma}(g)v \otimes \sigma(g)f)]\right)\\
& = \left(\pi_P(rg^{-1}),
\check{\mbox{pr}}_2\check{J}([(rg^{-1},\check{\sigma}(g)v)]) \otimes
\mbox{pr}_2J([(rg^{-1},\sigma(g)f)])\right)\\
& =\left(\pi_P(r), \check{\mbox{pr}}_2 \check{J}([(r,v)]) \otimes
\mbox{pr}_2J([(r,f)])\right) = J_{CH}^0([(r, v \otimes f)])
\end{align*}
Since defined by composition of continuous maps, $J_{CH}^0$ is continuous.
The inverse of $J_{CH}^0$ on rank 1 operators is given by
$(J_{CH}^0)^{-1}(m,\tilde{v} \otimes \tilde{f}) = \check{J}^{-1}(m,\tilde{v})
\otimes_{\mathcal{CH}} J^{-1}(m,\tilde{f}),$ $m \in M,$ $\tilde{v} \in
\check{H}$ and
$\tilde{f} \in H,$ which is a composition of continuous maps.
On finite rank operators, $J^0_{CH}$ is extended linearly. Obviously,
$(J_{CH}^0)^{-1}$ on finite rank operators is
continuous as well.
Finally for $r\in \mathcal{P}$ and $a \in CH,$ we set
$$J_{CH}([(r,a)])=\left(\pi_P(r),\lim_i \mbox{pr}_2^{CH}J_{CH}^0([(r,
a_i)])\right)$$ where
$(a_i)_{i\in \mathbb{N}}$ is any
sequence of finite rank operators converging
to $a$ in the operator norm topology, and $\mbox{pr}_2^{CH}$ denotes the
projection of
$M
\times CH$ onto the second factor. The map $J_{CH}$ is independent of
the chosen sequence. Indeed, let $(a_i)_{i\in \mathbb{N}}$ and $(b_i)_{i\in
\mathbb{N}}$
be sequences of finite rank operators which converge to $a$ in the operator
norm.
Their difference $c_i = a_i - b_i$ converges to the zero operator.
Moreover, each $c_i$ is a finite rank operator. Therefore, there are
$\lambda^{jk}_i \in \mathbb{C}$ such that
$c_i=\sum_{j,k=1}^{n_i}\lambda^{jk}_i h_j \otimes \eta_k,$ where $(h_j)_j$ and
$(\eta_k)_k$ are complete orthonormal systems of $\check{H}$ and $H,$
respectively. We have $||c_i||_{CH}^2 = \sum_{j,k}|\lambda^{jk}_i|^2.$ Since
$c_i
\to 0,$ we get $\lambda^{jk}_i \to 0$ for all $j,k.$
This fact and the linearity of $J$ and $\check{J}$ make us able to conclude
that $J_{CH}^0([(r,c_i)])=\lim_i
\sum_{j,k}\lambda^{jk}_i\left(\check{\mbox{pr}}_2\check{J}([(r,h_j)]) \otimes
\mbox{pr}_2 J([(r,\eta_k)])\right)=
0.$ This proves that the definition of $J_{CH}$ is correct.
Map $J_{CH}$ is continuous since it is defined as the continuous extension in
the
second argument and
since $\pi$ is continuous.
The inverse of $J_{CH}$ is given by $J_{CH}^{-1}(m,a)=\lim_i
(J^{CH}_0)^{-1}(m,a_i),$
where $(a_i)_{i\in \mathbb{N}}$ is a sequences of finite rank operators
converging
to $a$ in the operator norm.
Summing-up, $J_{CH}: \mathcal{CH} \to M \times CH$ is a homeomorphism of
$\mathcal{CH}$ and $M\times CH.$ Since $J_{CH}$ is fiber preserving, $J_{CH}$
is an
isomorphism of continuous bundles. Since the Kuiper maps are taken to be
normalized, $J_{CH}$
is a fiber-wise isometry and thus, it is an isomorphism of Banach bundles and
consequently, $\mathcal{CH}$ is trivial as a Banach bundle.
\hfill\(\Box\)
\bigskip
{\bf Remark:} It is known that the set of equivalence classes of
bundles of Banach spaces, with fibers the algebra of
compact operators $CH$ on a Hilbert space $H$ and
whose structure group is the projective unitary group of $H,$ is isomorphic to
the third
singular cohomology
group $H^3(M,\mathbb{Z})$
of $M.$ The equivalence is the isomorphism of continuous fiber bundles.
See Dixmier, Douady \cite{DixmierDouady} or Dixmier \cite{Dixmier}.
We refer also to Mathai, Melrose and Singer \cite{MMS} and Schottenloher
\cite{Schotten} for a context and for a nice topological study,
respectively. (Moreover, the set of such Banach fiber bundles
forms a group under the spatial tensor product (the Brauer group). This group
is isomorphic to the additive structure on $H^3(M,\mathbb{Z}).$ See
Rosenberg \cite{Rosenberg}.)
\section{Hilbert $C^*$-bundle and Kuiper complex}
In this section, we first introduce pointwise structures which we use for
defining a Hilbert $C^*$-structure on the higher oscillator bundle.
Then we use a product type connection to establish the
elliptic complex, mentioned in the Introduction.
\subsection{Pointwise analytical structures}
Let us define a map $\cdot_{\mathcal{H}}:
\mathcal{H}^{\bullet} \times_M \mathcal{CH} \to
\mathcal{H}^{\bullet}$ by
$$[(q,f)] \cdot_{\mathcal{H}} [(q,a)] = [(q, f \cdot a)] $$
where $q \in \mathcal{P},$ $f\in
H^{\bullet},$ and $a\in CH.$ We call it the {\it $\mathcal{CH}$-action}.
The independence on the representative of an equivalence class is proved below.
\bigskip
For to define a pointwise $\mathcal{CH}$-valued product on
$\mathcal{H}^{\bullet},$ we
reduce the structure group of $\mathcal{P}$
from $\widetilde{G}$ to $\widetilde{K}$ which is possible since this group
is a deformation retract of the structure group $Mp(V,\omega_0)$ by the
lifting
property for coverings
and the known fact that $Sp(V,\omega_0)$ deformation retracts onto
$K$.
The reduction is non-unique. Therefore, we choose a compatible positive almost
complex structure on $(M,\omega)$ and consider the $\Lambda$-preimage of the bundle of
unitary
frames. We denote the resulting principal
$\widetilde{K}$-bundle by $\mathcal{P}_R.$
The corresponding
associated bundles are denoted by $\mathcal{H}^{\bullet}_{R}$ and
$\mathcal{CH}_{R}.$ We call the former one the reduced higher oscillator
bundle. The total spaces of these vector bundles
coincide, as topological spaces, with $\mathcal{H}^{\bullet}$ and
$\mathcal{CH},$ respectively.
Using this reduction, we define the pointwise {\it $\mathcal{CH}$-product}
$(,)_{\mathcal{CH}}: \mathcal{H}^{\bullet}_R \times_M \mathcal{H}^{\bullet}_R
\to
\mathcal{CH}_R \mbox{ by the formula }$
$$([(q,v)],[(q,w)])_{\mathcal{CH}}=[(q,(v,w))] \in \mathcal{CH}_R$$ where $q
\in
\mathcal{P}_R,$ and $v,w \in H^{\bullet}.$
Note that whereas at the right-hand side, $(v,w) \in CH,$ at the left-hand
side, a map
$(,)_{\mathcal{CH}}:\mathcal{H}^{\bullet}_R \times_M \mathcal{H}^{\bullet}_R
\to
\mathcal{CH}_R$ is prescribed as we prove yet.
\bigskip
{\bf Lemma 8}: For a symplectic manifold $(M,\omega)$ admitting a metaplectic
structure $(\pi_P,\Lambda)$, the maps
\begin{eqnarray*}
\cdot_{\mathcal{H}}&:& \mathcal{H}^{\bullet} \times_M \mathcal{CH} \to
\mathcal{H}^{\bullet}\\
(,)_{\mathcal{CH}}&:& \mathcal{H}^{\bullet}_R \times_M \mathcal{H}^{\bullet}_R
\to
\mathcal{CH}_R
\end{eqnarray*}
are correctly defined.
{\it Proof.}
We have to check that these maps do not depend on the choice of
representatives.
\begin{itemize}
\item[1)] Let us verify that the definition of the action
$\cdot_{\mathcal{H}}$
is
correct. For $q\in \mathcal{P},$ $g \in
\widetilde{G},$ $k=0,\ldots, 2n,$ $a\in CH,$ and
$v = \alpha \otimes h \in H^{\bullet}$
\begin{align*}
[(q,v)]\cdot_{\mathcal{H}} [(q g, \rho(g)^{-1}a)] &= [\left(q g,
\sigma^k(g^{-1})v\right)]\cdot_{\mathcal{H}} [\left(q
g,(\rho(g)^{-1}a)\right)]\\
&= [\left(q g, \left(\sigma^k(g)^{-1}(v)\right) \cdot
\left(\rho(g)^{-1}a\right)\right]\\
&= [\left(q g, \sigma^k(g)^{-1} (v \cdot a)\right)]\\
&= [(q,v)]\cdot_{\mathcal{H}}[(q,a)]
\end{align*}
where the equivariance of the $CH$-action (Lemma 3) is used in the second row.
In a similar way, one proves that the definition of $\cdot_{\mathcal{H}}$ does
not depend on the choice of a
representative of the element acted upon.
\item[2)] We prove the correctness of the definition of $(,)_{\mathcal{CH}}.$
For a frame $q \in \mathcal{P}_R,$ $\alpha, \beta \in \bigwedge^{k}V^*,$ $v,w
\in H$ and $g\in \widetilde{K}$
\begin{align*}
\left([(q g,\sigma^k(g^{-1})v)],[(q ,
w)]\right)_{\mathcal{CH}}&=\left([\left(qg,\sigma^k(g^{-1})v\right)], [\left(q
g,\sigma^k(g^{-1})w)\right]\right)_{\mathcal{CH}}\\
&=\left[\left(qg, \left(\sigma^k(g^{-1})v,
\sigma^k(g^{-1})w\right)\right)\right]\\
&=\left[\left(q g, \rho(g)^{-1}(v,w)\right)\right] = [\left(q,(v,w)\right)]
\end{align*}
In the second row above, the $\widetilde{K}$-equivariance of the
$CH$-product from Lemma 3 is used.
The independence on representatives of the second argument of
$(,)_{\mathcal{CH}}$ follows from
the hermitian symmetry of $(,): H^{\bullet} \times H^{\bullet} \to CH$.
\end{itemize}
\hfill\(\Box\)
\bigskip
{\bf Remark:}
Consider the composition
$$\xymatrix{
h \in \mathcal{H} \ar@{|->}[r]^-{(,)_{\mathcal{CH}}} &
(h,h)_{\mathcal{CH}} \in \mathcal{CH} \ar@{|->}[r]^-{||\,||_{\mathcal{CH}}} &
||(h,h)_{\mathcal{CH}}||_{\mathcal{CH}} \in \mathbb{R}_{\geq 0}
}$$
where $||[(q,a)]||_{\mathcal{CH}} = ||a||_{CH}$ for $q\in \mathcal{P}_R$ and
$a\in
CH.$
This map is well defined since $\rho$ is an isometry. Let us recall that there
is also a fiber-wise scalar product
$(,)_S$ on $\mathcal{H}\to M$ as constructed in the proof of Thm. 4
that makes $\mathcal{H}$ a Hilbert bundle. One can prove easily
that
$||h||_S^2=||(h,h)_{\mathcal{CH}}||_{\mathcal{CH}}$ where $||\,||_S$ is the
fiber-wise norm induced by $(,)_S.$
\subsection{Hilbert $C^*$-bundles}
According to Fomenko, Mishchenko \cite{FM}, we shall define a
smooth action of $CH$ on the total space $\mathcal{H}^{\bullet}_R,$ which
restricts to a Hilbert
$CH$-module action on each fiber; and a smooth
$CH$-valued map $(,):\mathcal{H}^{\bullet}_R\times_M \mathcal{H}^{\bullet}_R
\to
CH,$ which restricts to a Hilbert $CH$-product on each fiber.
Further, on these bundles, Hilbert $C^*$-atlases have to be fixed
such that their transition functions are maps into
$D^{\bullet}=\mbox{Aut}_{CH}(H^{\bullet}),$ i.e., the set of all homeomorphisms
commuting with the action of $CH$ on $H^{\bullet}.$
\bigskip
We define a Hilbert $C^*$-bundle atlas on $\mathcal{H}^{0}_R$ by requiring that it is
the
maximal smooth atlas satisfying the following conditions
\begin{itemize}
\item[1)] it contains $(M,J)$ and
\item[2)] its transition functions are maps into $D^0$.
\end{itemize}
Note that $D^0=\mbox{Aut}_{CH}(H^0) \simeq \mathbb{C}^{\times}$, which is quite
restrictive but it gives a clear picture of the situation.
Namely, the atlas equals $$\mathcal{A}= \{(U,\mu J_{|(p_0)^{-1}(U)})| \,
\mu \in \mathbb{C}^{\times},\, U \mbox{ open in } M \}$$
This atlas is a subset of the Kuiper structure on $\mathcal{H}
^0$ induced by $J.$
On $\mathcal{H}^{k}$ we consider the tensor product atlases $\mathcal{A}_k,$
which are the tensor products of the $C^{\infty}$-structure on
$\bigwedge^kT^*M$ and of $\mathcal{A}.$ We call them the {\it
Hilbert $C^*$-bundle structure induced by $J$}.
\bigskip
{\bf Definition:} For a symplectic manifold $(M,\omega)$ possessing a
metaplectic structure $(\pi_P,\Lambda)$, a compatible positive almost complex
structure, a Kuiper map $J: \mathcal{H} \to M \times H,$ $m\in M,$
$f, h \in (\mathcal{H}^{\bullet}_R)_m,$ and $a \in CH,$ we set
\begin{align*}
&h \cdot a = h \cdot_{\mathcal{H}} J_{CH}^{-1}(m,a) \hspace{.5cm} \mbox{and}
\\
&(f,h) =\mbox{pr}_2^{CH} \left( J_{CH}(f,h)_{\mathcal{CH}}\right)
\end{align*}
where $\mbox{pr}_2^{CH}: M
\times CH \to
CH$ denotes the projection onto the second factor.
\bigskip
{\bf Remark:} The symbols $\cdot$ and $(,)$ should not be confused with the
ones
referring to the action $H^{\bullet} \times CH \to H^{\bullet}$ and the
$CH$-product $H^{\bullet} \times H^{\bullet} \to CH.$
\bigskip
Let us introduce the following operations. For $r \in \mathcal{P},$ $a,b \in
CH,$ $v \in \check{H}$
and $f \in H,$ we set
\begin{align*}
{\circ}_{\mathcal{CH}}: \mathcal{CH} \times_M \mathcal{CH} \to \mathcal{CH}&
\quad [(r,a)] {\circ}_{\mathcal{CH}} [(r,b)] = [(r,a b)]\\
{\circ}_{CH}: (M \times CH) \times_M (M \times CH) \to M \times CH & \quad
(m,a) {\circ}_{CH} (m,b) = (m,ab) \\
\otimes_{CH}: (M \times \check{H}) \times_M (M \times H) \to M \times
CH &\quad
(m,v) \otimes_{CH} (m,f) = (m, v \otimes f)
\end{align*}
These maps are well defined by Lemma 3.
\bigskip
{\bf Lemma 9:} Let $a,b \in CH,$ $f\in H$ and $m \in M.$ Then
\begin{align*}
&J^{-1}_{CH}(m,a\circ b)=J^{-1}_{CH}(m,a) \circ_{\mathcal{CH}} J_{CH}^{-1}(m,b)
\mbox{
and }\\
&J^{-1}(m,f) \cdot_{\mathcal{H}} J^{-1}_{CH}(m,a) = J^{-1}(m,f \circ a)
\end{align*}
{\it Proof.} For any $q\in \mathcal{P},$ $v_a, v_b \in \check{H},$ and $f_a,
f_b
\in
H,$
$([(q,v_a)]\otimes_{\mathcal{CH}}[(q,f_a)])\circ_{\mathcal{CH}}([(q,v_b)]
\otimes_{
\mathcal { CH } } [ (q , f_b) ]) = [(q, (v_a\otimes f_a) \circ (v_b\otimes
f_b))] = [(q,f_a(v_b)v_a \otimes f_b)].$
Thus for $v,w \in \check{\mathcal{H}}_m$ and $f,h \in \mathcal{H}_m$
\begin{eqnarray}
(v \otimes_{\mathcal{CH}} f) \circ_{\mathcal{CH}}
(w\otimes_{\mathcal{CH}}
h)
= f(w)(v \otimes_{\mathcal{CH}} h) \label{blbost}
\end{eqnarray}
For $a=v_a \otimes f_a \in CH$ and $b= v_b \otimes f_b \in CH$
\begin{align*}
J_{CH}^{-1}(m,a \circ b) &= J_{CH}^{-1}\left(m, (v_a\otimes f_a) \circ (v_b
\otimes f_b)\right)=
f_a(v_b) J_{CH}^{-1}(m, v_a \otimes f_b)
\end{align*}
Since $J$ is a fiber-wise isometry
\begin{align*}
f_a(v_b) J_{CH}^{-1}(m, v_a &\otimes f_b)=f_a(v_b) \check{J}^{-1}(m,v_a)
\otimes_{\mathcal{CH}} J^{-1}(m,f_b)\\
&= (J^{-1}(m,f_a),J^{-1}((m,v_b)^{\flat_M}))_S
\check{J}^{-1}(m,v_a)\otimes_{\mathcal{CH}} J^{-1}(m,f_b)\\
&=(J^{-1}(m,f_a)[\check{J}^{-1}
(m,v_b)])\check{J}^{-1}(m,v_a)\otimes_{\mathcal{CH}}J^{-1}(m,f_b)
\end{align*}
Using (\ref{blbost}), we have
$\left(J^{-1}(m,f_a)[\check{J}^{-1}(m,v_b)]\right)\check{J}^{-1}(m,v_a)
\otimes_{\mathcal{CH}} J(m,f_b)$
\begin{align*}
&=(\check{J}^{-1}(m,v_a) \otimes_{\mathcal{CH}} J^{-1}(m,f_a))
\circ_{\mathcal{CH}} (\check{J}^{-1}(m,v_b) \otimes_{\mathcal{CH}}
J^{-1}(m,f_b))\\
&=J_{CH}^{-1}(m, v_a \otimes f_a) \circ_{\mathcal{CH}} J_{CH}^{-1}(m, v_b
\otimes
f_b)=J_{CH}^{-1}(m,a) \circ_{\mathcal{CH}} J_{CH}^{-1}(m,b)
\end{align*}
Summing-up, $J_{CH}^{-1}(m,a\circ b)=J_{CH}^{-1}(m,a) \circ_{\mathcal{CH}}
J_{CH}^{-1}(m,b)$
for rank 1 operators $a$ and $b.$
For a linear combination of finite rank maps, the result follows by linearity.
For a product of compact operators, one proceeds by taking limits which is
justified by noting that $J_{CH}$ is a homeomorphism (Sect. 3.3, proof of Thm.
7).
The second equation is proved in a similar way.
\hfill\(\Box\)
\bigskip
{\bf Theorem 10:} Let $(M,\omega)$ be a symplectic manifold admitting a
metaplectic structure $(\pi_P, \Lambda)$ and $J$ be a Kuiper map.
Then the reduced higher oscillator bundle $\mathcal{H}^{\bullet}_R \to M$ with
the Hilbert $C^*$-bundle structure induced by a Kuiper map
$J$ is a smooth Hilbert $CH$-bundle for any compatible positive almost complex
structure.
{\it Proof.}
First, we have to verify that the map $\cdot: \mathcal{H} \times CH \to
\mathcal{H}$ is
an action. Recall that according to the definitions of the
$\mathcal{CH}$-action and the composition $\circ_{\mathcal{CH}}$, we have
$h \cdot_{\mathcal{H}} (a \circ_{\mathcal{CH}} b) = (h \cdot_{\mathcal{H}} a)
\cdot_{\mathcal{H}} b,$ whenever these operations make sense.
For $m \in M,$ $h\in (\mathcal{H}^{\bullet}_R)_m$ and $a,b \in CH,$ we get
according to Lemma 9
\begin{align*}
h \cdot (ab) &= h \cdot_{\mathcal{H}} J^{-1}_{CH}(m,ab) = h
\cdot_{\mathcal{H}}
\left(J^{-1}_{CH}(m,a) \circ_{\mathcal{CH}} J^{-1}_{CH}(m,b)\right) \\
&= \left(h \cdot_{\mathcal{H}} J^{-1}_{CH}(m,a)\right) \cdot_{\mathcal{H}}
J^{-1}_{CH}(m,b) = \left(h
\cdot a\right) \cdot b
\end{align*}
Second, using Lemma 3, it is easy to verify that
$$(f,h \cdot_{\mathcal{H}} b)_{\mathcal{CH}} = (f,h)_{\mathcal{CH}}
\circ_{\mathcal{CH}} b$$
for suitable $h,f \in \mathcal{H}_R^{\bullet}$ and $b \in \mathcal{CH}_R.$ For
such elements and $a\in CH$ we get, using Lemma 9
\begin{align*}
(f, h \cdot a) &= \mbox{pr}_2^{CH}\left( J_{CH}\left( f, h \cdot_{\mathcal{H}}
J_{CH}^{-1}(m,a)\right) \right)\\
\quad &=\mbox{pr}_2^{CH}\left( J_{CH}\left(
(f,h)_{\mathcal{CH}}
\circ_{\mathcal{CH}} J_{CH}^{-1}(m,a)\right) \right)\\
\quad
&=\mbox{pr}_2^{CH}\left(J_{CH}\left[J_{CH}^{-1}\left(J_{CH}(f,h)_{\mathcal{CH}}
\right)
\circ_{\mathcal{CH}} J_{CH}^{-1}(m,a)\right]\right)\\
\quad &=\mbox{pr}_2^{CH}\left(J_{CH} \left[J_{CH}^{-1}\left(m,
\mbox{pr}_2^{CH}\left(J_{CH}\left(f,h\right)_\mathcal{CH}\right)\circ
a\right)\right]
\right)\\
\quad &= \mbox{pr}_2^{CH}\left(m,\mbox{pr}_2^{CH}(J_{CH}(f,h)_{\mathcal{CH}})\circ
a\right)\\
\quad &=\mbox{pr}_2^{CH}\left(J_{CH}\left(f,h\right)_{\mathcal{CH}}\right)
\circ
a =
(f,h) \circ a
\end{align*}
The Hilbert $C^*$-bundle structure induced by $J$ on $\mathcal{H}^k,$
$k=0,\ldots,
2n,$ is a maximal set of smooth Hilbert $C^*$-bundle atlases smoothly compatible with
$(M,J)$. Let us verify
that the $CH$-action
is
smooth. The $CH$-action $\mathcal{H} \times CH \to \mathcal{H}$ in the map
$(M,J)$ is given by
$(m,f) \mapsto (m, f\circ a),$ where $m\in M,$ $f\in
(H_R^{\bullet})_m$ and $a\in CH.$ This map is infinitely many times
Fr\'echet differentiable. Similarly, the $CH$-product is given by
$((m,f),(m,h)) \mapsto (f,h) \in CH$ in $(M,J),$ $m\in M$
and $f,h \in (H_R^{\bullet})_m,$ which is thus Fr\'echet
smooth.
\hfill\(\Box\)
\bigskip
{\bf Hilbert $C^*$-bundle structure on $\underline{H}$:} Let $g$ be the
(Hodge type) Riemannian metric on the fibers of $\bigwedge^k T^*M$ induced
by $\omega$ and a chosen compatible positive almost complex structure.
The tensor product bundle $\bigwedge^{\bullet}T^*M \otimes \underline{H} \to M$
of $\bigwedge^{\bullet}T^*M \to M$ and of the trivial
bundle $\underline{H} = M \times H \to M$ is a Hilbert
$C^*$-bundle by setting $(\alpha \otimes (m,f))\cdot_M a = \alpha \otimes
(m,f\cdot a) \in \bigwedge^{\bullet}T^*M \otimes
\underline{H}$ and $(\alpha \otimes (m,f), \beta \otimes (m,h))_M =
g(\alpha,\beta)(f,h) \in CH,$ where
$m\in M,$ $\alpha,\beta \in \bigwedge^{\bullet} T^*_mM,$ $f,h \in
H$ and $a\in CH.$ For the smooth Hilbert $C^*$-bundle structure on
$\bigwedge^{\bullet}T^*M \otimes \underline{H}$, we take the product of
the maximal atlas of $\bigwedge^{\bullet}T^*M$ with the atlas
$\{\left(U,\mu \mbox{Id}_{U \times H}\right)| \, \mu
\in \mathbb{C}^{\times},\, U \mbox{ open in } M\}$ of $\underline{H}.$
\bigskip
{\bf Lemma 11:} Let $(M,\omega)$ be a symplectic manifold admitting a
metaplectic structure $(\pi_P, \Lambda).$ For any
compatible positive almost complex structure and a Kuiper structure $J,$ the
Kuiper map extends to a unitary isomorphism
$\widetilde{J}: \mathcal{H}^{\bullet}_R \to \bigwedge^{\bullet}T^*M \otimes
\underline{H}$ of Hilbert
$CH$-bundles.
{\it Proof.} Obviously, $\mathcal{H}_R^k \simeq \bigwedge^kT^*M \otimes
\mathcal{H}_R^0.$
We set $\widetilde{J}(\alpha \otimes h)= \alpha \otimes J(h),$ $\alpha \in
\bigwedge^kT^*M$ and $h\in \mathcal{H}^0_R,$ and extend it linearly.
Since $CH$ does not act on the form part, it is sufficient to verify that $J(h
\cdot a) =J(h) \cdot_M a$ for any $h \in \mathcal{H}^0_R$ and $a\in CH.$
Representing $h=J^{-1}(m,f)$ and using Lemma 9
\begin{align*}
J(h\cdot a) &= J(J^{-1}(m,f) \cdot_{\mathcal{H}} J^{-1}(m,a)) = J(J^{-1}(m,h
\circ a))\\
&= (m,h \circ a) = (m,h) \cdot_M a.
\end{align*}
It is easy to see
\begin{eqnarray}
(f,h)_{\mathcal{CH}} = f^{\sharp} \otimes_{\mathcal{CH}} h \label{vzor}
\end{eqnarray}
for $f, h \in \mathcal{H}^0_R.$
Indeed, representing $f, h$ by equivalence classes, we obtain \\
$([(q,v)],[(q,w)])_{\mathcal{CH}} = [(q, v^{\sharp} \otimes w)] = f^{\sharp}
\otimes_{\mathcal{CH}} h.$
When proving the unitarity of $\widetilde{J}$, we choose $J$-preimages of
$f,h
\in \mathcal{H}^0_R$ and use (\ref{vzor}), proceeding in a similar way
as in the case of $CH$-equivariance. The unitarity of $\widetilde{J}$ follows
then from the fact that $CH$ does not act on the wedge form part of
$\bigwedge^{\bullet}T^*M \otimes \underline{H}.$
\hfill\(\Box\)
\subsection{Associated elliptic complex to a Kuiper connection}
We introduce the complex and prove that it is a complex of
$C^*$-operators. Further we investigate its cohomology groups. For that we
focus to the topology of section spaces of the higher oscillator bundles
and the topology of their tensor products and quotients.
Let $(M,\omega)$ be a symplectic manifold admitting a metaplectic structure
$(\pi_P, \Lambda)$ and $J: \mathcal{H} \to M \times H$ be a
normalized Kuiper map.
The $T(M\times H)$-valued $1$-form $\Phi^H$ defining the canonical
product connection on $M \times H
\to M$ is
given by $$\Phi^H(t_m, v_h)=(0_m,v_h)$$
where $m\in M,$ $t_m \in T_mM,$ $h\in H,$ $v_h \in T_hH,$ and $0_m$ is the
zero tangent vector in $T_mM.$ Note that we identify $T(M\times H) \simeq TM
\times
TH$.
The Kuiper map gives rise to a connection $\Phi$ in $\mathcal{H}$ given by
the prescription $\Phi=TJ^{-1} \circ \Phi^H \circ TJ$ which we call the
{\it Kuiper connection induced by} $J,$ expressing its roots.
\bigskip
Let us notice that a
different description of a connection on symplectic spinors is chosen by
Herczeg, Waldron \cite{HerczegWaldron} in the
setting of quantum physics for contact manifolds. This connection is used
already in Kr\'ysl \cite{KryslDGA0}.
\bigskip
{\bf Remark:}
The distribution $$\mathcal{H} \ni f=[(r,h)] \mapsto \mathcal{L}_f =
\left(T_{(\pi_P(r),h)}J^{-1}\right)(T_mM,0)\subseteq
T_f\mathcal{H}$$ where
$r \in \mathcal{P}, h \in H, m=\pi_P(r)$ and $0 \in T_hH$ (the zero vector
of $T_hH$) is the horizontal distribution defined by $\Phi.$ In this way, we
get
a splitting of the vector bundle $T\mathcal{H} =
\mathcal{L} \oplus \textrm{Ker}\,(Tp_0).$ The vertical part of the bundle
$\mathcal{H}$ is denoted by $\mathcal{V}.$
\bigskip
{\bf Gateaux and Fr\'echet differentials in geometric setting:} We
introduce a notation and recall some facts from analysis on Banach spaces.
Let $(Df)(h,v)$ denote the Gateaux derivative of $f$ at $h$ in direction $v.$
For each $h \in H,$ $T_hH = \{v_h, v \in H\},$ where $v_hf=(Df)(h,v)$ for any
smooth function $f$ defined in a neighborhood of $h \in H.$
For any point $h \in H,$ let us define $i_h:T_hH \to H$ by $i_h(v_h)=v,$ $v \in
H.$ It is easy to see that
$i_h$ is an isomorphism of vector spaces.
\bigskip
{\bf Vertical lifts, connectors and induced connections:} We recall
basic principles for handling of associated
connections. It is based on the approach described in Michor
\cite{Michor} and Kol\'a\v{r}, Michor, Slov\'ak \cite{KMS}.
For a vector field $X\in \mathfrak{X}(M)(=\Gamma(TM)),$ the connection on
$\mathcal{H}$ induces a
covariant derivative $\nabla_X: \Gamma(\mathcal{H}) \to \Gamma(\mathcal{H})$
defined by $\nabla_X s = \kappa \circ Ts \circ X,$ where
$X: M \to TM$ is a vector field on $M,$ $Ts: TM \to T\mathcal{H}$ is the tangent
map of a bundle section $s: M \to \mathcal{H},$ and
$\kappa: T\mathcal{H} \to \mathcal{H}$ is the connector of $\Phi.$
We have also a covariant derivative $\nabla^H_X$ acting on sections of $M\times
H \to M$ which is given by $\nabla^H_X s = \kappa^H \circ Ts \circ X,$ where
$s: M \to M \times H$ and $\kappa^H$ is the connector of $\Phi^H.$
Let us denote the vertical lifts of $(H\times M \to M,\Phi^H)$ and
$(\mathcal{H}
\to M, \Phi)$ by $vl^H$
and $vl,$ respectively. Recall that $vl: \mathcal{H} \times_M \mathcal{H} \to
\mathcal{V}$ and similarly
$vl^H: (M \times H) \times_M (M \times H) \to TM \times H.$
We denote the projections $\mbox{pr}_2: \mathcal{H} \times_M
\mathcal{H} \to \mathcal{H}$ and $\mbox{pr}_2^H:(M\times H)\times_M (M\times H)
\to M\times H$
onto the respective second factors. With this notation, the connector $\kappa$
is the map $\mbox{pr}_2 \circ vl^{-1} \circ \Phi$ and similarly for $\kappa^H.$
\bigskip
{\bf Lemma 12:} In the situation described in the previous paragraph, we have
for any $m\in M,$ and $h,v \in H$
\begin{itemize}
\item[1)] $(vl^H)^{-1}(0_m,v_h)=\left((m,h),(m,v)\right) \in (M\times H)
\times_M
(M\times H)$
\item[2)] $vl = TJ^{-1} \circ vl^H \circ (J \times_M J)$
\item[3)] $\mbox{pr}_2^H \circ (J \times_M J) = J \circ \mbox{pr}_2 $
\end{itemize}
{\it Proof.} For $m \in M$ and
$h, v \in (M \times H)_m,$
$vl^H\left((m,h),(m,v)\right)=$\\
$=\frac{d}{dt}_{|t=0}\left( (m,h)
+_M t\odot_M (m,v)\right),$ where
$+_M$ and $\odot_M$ denote the addition and the multiplication by
scalars in
$M \times H \to M,$ respectively, as introduced in 3.3.
We have $\frac{d}{dt}_{|t=0} ((m,h)+_Mt\odot_M(m,v))=\frac{d}{dt}_{|t=0}(m,
h+tv) =
(0_m,v_{h}).$
For $x,y \in \mathcal{H}_m,$ $vl(x,y) = \frac{d}{dt}_{|t=0}(x+ty).$ Because
$J$ is fiber-wise linear, $vl(x,y)=
\frac{d}{dt}_{|t=0}J^{-1}J(x+ty)=
TJ^{-1}\frac{d}{dt}_{|t=0}(h +_M t\odot_M v) = TJ^{-1}vl^H(h,v),$ where
$h = J(x)$ and $v=J(y).$
The third equality is obvious.
\hfill\(\Box\)
\bigskip
{\bf Lemma 13:} For $X \in \mathfrak{X}(M)$ and $s\in \Gamma(\mathcal{H})$
$$J \circ \nabla_X s = \nabla_X^H (J \circ s)$$
{\it Proof.} We use Lemma 12 to prove the identity. For $X\in \mathfrak{X}(M)$
and $s\in \Gamma(\mathcal{H})$
\begin{eqnarray*}
\nabla_X^H(J \circ s) &=& \kappa^H \circ TJ \circ Ts \circ X\\
&=& \mbox{pr}_2^H \circ {vl^H}^{-1} \circ \Phi^H \circ TJ \circ Ts \circ X\\
&=& \mbox{pr}_2^H \circ (J \times_M J) \circ vl^{-1} \circ TJ^{-1} \circ
\Phi^H
\circ
TJ \circ Ts \circ X\\
&=& \mbox{pr}_2^H \circ (J \times_M J) \circ vl^{-1} \circ \Phi \circ Ts \circ
X\\
&=& J \circ \mbox{pr}_2 \circ vl^{-1} \circ \Phi \circ Ts \circ X = J \circ
\nabla_Xs
\end{eqnarray*}
\hfill\(\Box\)
\bigskip
{\bf Remark:} Since we shall compute derivations of the $CH$-action on
$\underline{H}$ (by a fixed element),
we use a notation which is more usual for maps.
For any $a \in CH$ and $(m,v) \in M \times H,$ we set $G_a(m,v)=(m, v \circ a)$
and $g_a(v) = v \circ a,$
defining maps $G_a: M \times H \to M \times H$ and $g_a: H \to H.$ Note that
$G_a =
\mbox{Id}_M \times g_a.$
\bigskip
{\bf Lemma 14:} For any $X \in \mathfrak{X}(M),$ $s\in \Gamma(\underline{H})$
and
$a \in CH$
$$\nabla^H_X(G_a \circ s) = G_a \circ \nabla^H_X s$$
{\it Proof.}
For $v,h\in H$ and $a \in CH,$ $(Dg_a)(h,v)=\lim_{t\to 0} [t^{-1}(h \circ a
+ t v \circ a - h \circ a)] = v \circ a$.
For $f\in C^{\infty}(H),$ $(Tg_a)(v_h)f = v_h(f \circ g_a) =
D(f \circ g_a)(h,v)= $ $(Df)(h \circ a,$ $(Dg_a)(h,v)) = (Df)(h \circ a, v
\circ
a)$ by the chain rule.
The last expression equals to $(v \circ a)_{h \circ a} f.$
Thus $(Tg_a)v_h = (v \circ a)_{h \circ a},$ which implies
\begin{align}
TG_a(t_m,v_h) = \left(t_m, (Tg_a)(v_h)\right) = (t_m, (v \circ a)_{h \circ a})
\label{pomoc}
\end{align}
where $m \in M$ and $t_m \in T_mM.$
By Lemma 12
\begin{eqnarray*}
(G_a \circ \kappa^H)(t_m,v_h)&=&(G_a \circ \mbox{pr}_2^H \circ{vl^H}^{-1}
\circ \Phi^H)(t_m,v_h) \\
&=& (G_a \circ \mbox{pr}_2^H \circ {vl^H}^{-1})(0_m,v_h)\\
&=&(G_a \circ \mbox{pr}_2^H)((m,h),(m,v))=G_a(m,v) = (m,v \circ a)
\end{eqnarray*}
Further, we get by Lemma 12 and (\ref{pomoc}) that
\begin{eqnarray*}
(\kappa^H \circ TG_a)(t_m,v_h)&=& (\mbox{pr}_2^H \circ{vl^H}^{-1}\circ \Phi^H
\circ TG_a)(t_m,v_h)\\
&=&(\mbox{pr}_2^H \circ{vl^H}^{-1} \circ\Phi^H)(t_m,i^{-1}_{h \circ a}(v \circ
a))\\
&=& (\mbox{pr}_2^H \circ{vl^H}^{-1})(0_m, i^{-1}_{h \circ a}(v \circ a)) = (m,v
\circ a)
\end{eqnarray*}
Summing-up, $G_a \circ \kappa^H = \kappa^H \circ TG_a.$
The theorem follows by
$\nabla_X(G_a \circ s)= \kappa^H \circ TG_a \circ Ts \circ X = G_a \circ
\kappa^H \circ Ts \circ X = G_a \circ \nabla^H_X s$ for any
$s\in \Gamma(\underline{H})$ and $X \in \mathfrak{X}(M).$
\hfill\(\Box\)
\bigskip
{\bf Hilbert $C^*$-module structure on section spaces:}
Notice that $CH$ acts not only on the higher oscillator bundle but
also on its smooth
sections space $\Gamma(\mathcal{H}^{\bullet}).$
Namely, $(s \cdot a)_m = s_m\cdot a$
for any $a\in CH,$ $s \in \Gamma(\mathcal{H}^{\bullet})$ and $m \in
M.$\footnote{At the left-hand side a new action is defined.} In this
way,
$\Gamma(\mathcal{H}^{\bullet})$ is a right $CH$-module. The space of smooth
sections of
the reduced higher oscillator bundle $\mathcal{H}_R^{\bullet}$ is also a
pre-Hilbert
$CH$-module if $(M, \omega)$ is compact.
The $CH$-valued product is given by $$(s,s') = \int_{m \in M} (s_m,s'_m)
\mbox{vol}_m$$
where $s,s' \in \Gamma(\mathcal{H}^{\bullet})$ and $\mbox{vol}_m$ denotes the
volume form in $m$
induced by $\omega_m.$ The integral is the Pettis integral (for the measure on
$M$
induced by the volume form $\mbox{vol}_m$) of
Pettis $CH$-valued integrable function on $M.$ (See Ryan
\cite{Ryan} or Grothendieck \cite{Groth}.)
Similarly, we consider the
space
$\Gamma(\underline{H})$ of smooth sections of the product bundle $\underline{H}
\to M$ as a
pre-Hilbert $CH$-module as well. Completions of this pre-Hilbert module
(see Fomenko, Mishchenko \cite{FM}) with respect to the induced norm make
it a Hilbert $C^*$-module. (Let us notice that for reasons of analysis, several
further $C^*$-valued products of `Sobolev type' are introduced in \cite{FM}
besides
the $CH$-product given above. However, the integral seems to be rather not
specified there.)
\bigskip
{\bf Corollary 15:} For a symplectic manifold $(M,\omega)$ admitting a
metaplectic structure
$(\mathcal{P}, \pi_P),$ a Kuiper map $J$ and a vector field $X$ on $M,$
the operator
$$\nabla_X:\Gamma(\mathcal{H}) \to \Gamma(\mathcal{H})$$ is a first order
$CH$-equivariant differential operator.
{\it Proof.}
The map $\nabla_X$ is a first order differential operator for any $X\in
\mathfrak{X}(M)$ (all operations involved in its definition are of zero
order, except of $s \mapsto Ts$).
Using Lemmas 9, 13 and 14 for $a \in CH,$ $X \in \mathfrak{X}(M)$ and
$s \in \Gamma(\mathcal{H})$
\begin{eqnarray*}
\nabla_X (s \cdot a) &=& J^{-1}\circ \nabla_X^H (J \circ (s \cdot a))=
J^{-1}\circ \nabla^H_X(J \circ J^{-1} \circ G_a \circ J \circ s)\\
&=& J^{-1}\circ \nabla_X^H (G_a \circ J \circ s) =(J^{-1} \circ G_a) \circ
\nabla^H_X (J \circ s)\\
&=& (J^{-1} \circ G_a \circ J) \circ \nabla_X s = (\nabla_X s) \cdot a
\end{eqnarray*}
\hfill\(\Box\)
\bigskip
{\bf Remark} (Connection to non-commutative geometry): It is easy to see that
$\nabla_X$ is $CH$-hermitian
in the sense that $X(s,s') = (\nabla_X s, s')+ (s,\nabla_X s')$ where $s,s' \in
\Gamma(\mathcal{H}_R^{\bullet}).$
In Dubois-Violette, Michor \cite{DM}, central $A$-bimodules and connections
for them are investigated. To some extent, our structures fit into the specific
{\it non-commutative geometry concept} described there.
Namely, for the not necessarily normed algebra $A$, considered in \cite{DM}, we
take the smooth
sections of the bundle of measurements, $A=\Gamma(\mathcal{CH}),$ and for the
module, we take the smooth sections $\Gamma(\mathcal{H}^{\bullet}).$ The
$A$-valued metric
$h$ (p. 229 in \cite{DM}) is then the `sectionised' $\mathcal{CH}$-product,
i.e., $h: \Gamma(\mathcal{H}^{\bullet}_R) \times
\Gamma(\mathcal{H}_R^{\bullet}) \to
\Gamma(\mathcal{CH}),$ $h(s,s')(m)=(s_m,s'_m)_{\mathcal{CH}}$ where
$s,s' \in \Gamma(\mathcal{H}^{\bullet})$ and $m\in M.$
However, the
vector fields $X$ in our paper are generically not all derivations of
$A=\Gamma(\mathcal{CH}).$ (They coincide iff the manifold is discrete as a
topological space.) The module $\Gamma(\mathcal{H}_R^{\bullet})$ with the
$\Gamma(\mathcal{CH})$-hermitian connection may be considered as a modification of
a non-commutative hermitian manifold as considered in the program of Connes \cite{Connes}.
They have a global realization by topological and differential structures.
\bigskip
Let $(e_i)_{i=1}^{2n}$ and $(\epsilon^i)_{i=1}^{2n}$ be dual local frames of
$TM \to M$ and $T^*M \to M$ respectively.
Recall that the exterior covariant derivatives
$d_k^{\Psi}:\Gamma(\bigwedge^kT^*M \otimes \mathcal{F}) \to
\Gamma(\bigwedge^{k+1}T^*M \otimes \mathcal{F})$ induced by a vector bundle
connection $\Psi,$ defined on a vector bundle $\mathcal{F} \to M,$ are given by
$$d^{\Psi}_k: \alpha \otimes s \mapsto
d_k\alpha \otimes s + \sum_{i=1}^{2n}\epsilon^i \wedge \alpha \otimes
\nabla_{e_i}^{\Psi}s$$
where $\nabla^{\Psi}$ denotes the covariant derivative associated to $\Psi$ and
$d_k$ denotes the de Rham differential. They are extended linearly to
non-homogeneous elements.
\bigskip
{\bf Theorem 16:} The exterior covariant derivatives induced by the
Kuiper connection $\Phi$ form
a cochain
complex in the category of right-modules over the ring $CH$ of compact
operators, i.e.,
\begin{align*}
&d^{\Phi}_k(s \cdot a) = (d^{\Phi}_k s)\cdot a \, \mbox{ and}\\
&d_{k+1}^{\Phi} \circ d_k^{\Phi} = 0
\end{align*}
for any $a\in CH$ and $s\in \Gamma(\mathcal{H}^{\bullet}_R).$
{\it Proof.}
The second formula follows by applying the definition of the exterior
covariant derivatives and using the fact that $\Phi$ is flat.
Using Corollary 15, for any $\phi = \alpha \otimes s \in
\Gamma(\mathcal{H}^k)$ and $a\in CH$
\begin{eqnarray*}
d^{\Phi}_k \left((\alpha \otimes s) \cdot a\right) &=& d^{\Phi}_k\left(\alpha
\otimes (s
\cdot a)\right)\\
&=& d_k\alpha \otimes (s \cdot a) + \sum_{i=1}^{2n}
\epsilon^i \wedge
\alpha \otimes \nabla_{e^i}^{\Phi} (s\cdot a)\\
&=& (d_k\alpha \otimes s)\cdot a + \sum_{i=1}^{2n}\epsilon^i \wedge \alpha
\otimes
(\nabla_{e_i}^{\Phi} s) \cdot a\\
&=& \left( d_k\alpha \otimes s + \sum_{i=1}^{2n} \epsilon^i \wedge \alpha
\otimes \nabla_{e_i}^{\Phi} s \right) \cdot a\\
&=& \left(d^{\Phi}_k(\alpha \otimes s)\right)\cdot a
\end{eqnarray*}
\hfill\(\Box\)
\bigskip
{\bf Definition:} For a symplectic manifold admitting a metaplectic structure
$(\pi_{P}, \Lambda),$ a compatible positive almost complex structure and a
normalized Kuiper map $J,$ we call the cochain complex $d^{\Phi}_{\bullet}=
(\Gamma(\mathcal{H}_R^k), d^{\Phi}_k)_{k\in \mathbb{Z}}$ in the category of
pre-Hilbert $CH$-modules
the {\it Kuiper complex induced by $J$}.
\bigskip
{\bf Remark:} Let $A$ be a $C^*$-algebra. In categories
of pre-Hilbert and Hilbert $A$-modules and adjointable maps between them, the
images of morphisms need not have closed range. (See Solovyev, Troitsky
\cite{ST}.)
Consequently, the quotient topology on the cohomology groups of complexes in
these categories may be non-Hausdorff and thus, the quotients may not be
Hilbert $A$-modules in the quotient topology. (See \cite{KryslJGP2} for simple
examples.)
\bigskip
{\bf Convention on the cohomology groups and tensor products:} For a compact
manifold $M,$ we denote the space of $k$-cocycles and the $k$-th cohomology
group of a cochain complex $D_{\bullet}$ on $\mathcal{H}^{\bullet}_R$ by
$Z^k(D_{\bullet})$ and
$H^k(D_{\bullet}),$ respectively. The topology on
$\Gamma(\mathcal{H}^{\bullet}_R)$
is determined by the
norm induced from the pre-Hilbert $C^*$-module structure of
$\Gamma(\mathcal{H}^{\bullet}_R).$ For the topology on the cohomology groups,
we take
the quotient topology. In the case of the de Rham complex on $M$, we denote the
space of $k$-cocycles by $Z^k_{dR}(M),$ the space of $k$-coboundaries by
$B^k_{dR}(M)$ and the cohomology group by
$H^k_{dR}(M),$ and
consider the quotient topology on it as well. On the space of exterior
differential $k$-forms $\Omega^k(M)$ on the manifold $M$, we take the norm
topology given by the Hodge scalar product.
When we form a tensor product of pre-Hilbert spaces, we mean
the algebraic tensor product equipped with the scalar product induced by the
scalar products on the individual factors (especially, no
completions involved).
\bigskip
{\bf Lemma 17:} Let $(h_i)_{i\in \mathbb{N}}$ be a complete orthonormal system
on $H$ and, for $i \in \mathbb{N},$ let $\underline{h}_i$ be the constant
extension of $h_i$ to a section of $\Gamma(\underline{H}).$ If $h=
\sum_{i=1}^{\infty} c_i\underline{h}_i$ is a $C^1$-section of $\underline{H}$,
then
for any $i \in \mathbb{N},$ $c_i$ are $C^1$-functions and for any $v\in TM$
$$\nabla_v^H h =
\sum_{i=1}^{\infty} v(c_i) \underline{h}_i$$
{\it Proof.} Since the pointwise scalar product $(,)$ and $h$ are
continuous, we see
that $c_j$ is continuous for any $j\in \mathbb{N}$ by taking the scalar product
of $h$ and $\underline{h}_j.$
Let us choose a map $(U,\phi),$ $x\in U,$ $v\in T_xM$ and compute `in coordinates'
\begin{eqnarray*}
\nabla_v h&=&\lim_{t\to 0} \frac{1}{t} \sum_{i=1}^{\infty}
(c_i(x+tv)\underline{h}_i(x+tv) - c_i(x)\underline{h}_i(x)) \\
&=&\lim_{t \to 0} \sum_{i=1}^{\infty} \frac{1}{t}(c_i(x+tv)-c_i(x))
\underline{h}_i(x).
\end{eqnarray*}
Taking the inner product with $\underline{h}_j(x),$ we get
\begin{eqnarray*}
(\nabla_v h, \underline{h}_j(x)) &=& (\lim_{t \to 0} \sum_{i=1}^{\infty}
\frac{1}{t}(c_i(x+tv)-c_i(x)) \underline{h}_i(x), \underline{h}_j(x))
\end{eqnarray*}
Since the pointwise inner product $(,),$ $c_i$'s and $\underline{h}_i$'s are continuous, we can take the limit out of the bracket getting
\begin{eqnarray*}
&&\lim_{t \to 0} \frac{1}{t}(\sum_{i=1}^{\infty} (c_i(x+tv)-c_i(x))
\underline{h}_i(x), \underline{h}_j(x)) = \lim_{t\to 0}
\frac{1}{t}(c_j(x+tv)-c_j(x))\\
&&= v(c_j)
\end{eqnarray*}
Especially, the derivation of $c_j$ in direction $v$ exists and it is
continuous since the scalar product and $\underline{h}_j$ are continuous
and since $\nabla_v h$ is continuous by the assumption.
From the uniqueness of coordinates in a Hilbert space, we get that
$\nabla_v h = \sum_{i=1}^{\infty} v(c_i)\underline{h}_i.$
\hfill\(\Box\)
\bigskip
{\bf Theorem 18:} Let $(M,\omega)$ be a compact symplectic manifold of
dimension $2n$
admitting a metaplectic structure $(\pi_P,\Lambda)$ and $J$ be a
normalized Kuiper map.
Then for any $k \in \mathbb{Z}$ and any compatible positive almost complex
structure, the cohomology groups
$H^k(d_{\bullet}^{\Phi})$ of the Kuiper complex induced by $J$
have a structure of finitely generated projective Hilbert $CH$-modules whose
norm topology coincides with the quotient topology.
{\it Proof.}
1) By Lemma 11, the bundles $\mathcal{H}^k_R$ and $\bigwedge^kT^*M \otimes
\underline{H}$ are isomorphic
as smooth Hilbert $C^*$-bundles.
The map $$j: \Gamma(\mathcal{H}^{\bullet}_R) \to
\Gamma(\bigwedge^{\bullet}T^*M \otimes \underline{H}) \mbox{ given by }
j(s) = \widetilde{J} \circ s$$ where $s \in
\Gamma(\mathcal{H}_R^{\bullet}),$
preserves the $CH$-products and it is a $CH$-homomorphism as follows from
Lemma 11 and the linearity of the Pettis integral with respect to continuous
maps (see, e.g., \cite{Ryan}), respectively.
The same is true for the inverse
$j^{-1}(s) = \widetilde{J}^{-1} \circ s$. In particular,
$\Gamma(\bigwedge^k T^*M \otimes \underline{H})$ and
$\Gamma(\mathcal{H}_R^{\bullet})$ are isomorphic as pre-Hilbert $CH$-modules.
2) For $\alpha \otimes s \in \Gamma(\mathcal{H}_R^k),$ we have using Lemma 13
\begin{align*}
&d_k^{\Phi^H} \left( j (\alpha \otimes s) \right) =d_k^{\Phi^H} \left(\alpha
\otimes (J \circ s)\right) = d_k\alpha \otimes J \circ s + \sum_{i=1}^{2n}
\epsilon^i \wedge \alpha \otimes \nabla_{e_i}^H
(J \circ s) \\
&= d_k\alpha \otimes J \circ s + \sum_{i=1}^{2n} \epsilon^i \wedge \alpha \otimes J
\circ \nabla_{e_i} s = \widetilde{J} \circ d_k^{\Phi}(\alpha \otimes
s)=j\left(d_k^{\Phi}(\alpha \otimes s)\right)
\end{align*}
where $(e_i)_{i=1}^{2n}$ and $(\epsilon)_{i=1}^{2n}$ are dual local frames of
$M.$ Thus $j$ is a cochain map.
Because $j$ is a pre-Hilbert $CH$-isomorphism (item 1),
the corresponding cohomology
groups $H^k(d_{\bullet}^{\Phi})$ and $H^k(d_{\bullet}^{\Phi^H})$ of
$(\Gamma(\mathcal{H}_R^k), d_k^{\Phi})_k $ and $(\Gamma(\bigwedge^kT^*M
\otimes \underline{H}), d_k^{\Phi^H})_k$ are isomorphic as
$CH$-modules and topological vector spaces when considered with the
quotient topologies.
3) Note that $H^k_{dR}(M) \otimes H$ is a $CH$-module by defining
$([\alpha]\otimes h)\cdot a = [\alpha] \otimes (h \cdot a)$ for $\alpha \in
\Omega^k(M),$ $h\in H$ and $a\in CH.$ We show that
the cohomology groups of
$(\Gamma(\bigwedge^kT^*M \otimes
\underline{H}), d_{k}^{\Phi^H})_k$ are isomorphic to $H^k_{dR}(M) \otimes H$ as
$CH$-modules. Let us consider the linear
extension, denoted by $[\gamma],$ of the map
$H^k_{dR}(M) \otimes H \ni [\alpha]\otimes h \mapsto [\alpha \otimes
\underline{h}]
\in H^k(d_{\bullet}^{\Phi^H})$
where $\underline{h}$ is the constant extension of an element $h \in H.$
It is easy to check that the resulting map is well defined (maps cocycles into
cocycles and vanishes on coboundaries).
Let us analyze the continuity of $[\gamma].$ For it, we denote the
extension map $h \mapsto \underline{h}$ by $\mbox{ext}.$ Then
$[\gamma]$ fits
into the following commutative diagram
$$\begin{xy}\xymatrix{
Z^k_{dR}(M)\otimes H \ar[d]^{q_1 \otimes \mbox{Id}_H} \ar@{->}[r]^-{\mbox{Id}
\otimes
ext}
&Z^k(d_{\bullet}^{\Phi^H})\ar[d]^{q_2}\\
H^k_{dR}(M)\otimes H \ar@{->}[r]^-{[\gamma]}
&H^k(d_{\bullet}^{\Phi^H})}\end{xy}$$
where $q_i,$ $i=1,2,$ denote the appropriate quotient projections and
$\mbox{Id}$
denotes $\mbox{Id}_{Z^k_{dR}(M)}$.
Since $q_2,$ $\mbox{Id}$ and $\mbox{ext}$ are continuous and
$q_1\otimes \mbox{Id}_H$ is open, $[\gamma]$ is continuous.
For to prove that $[\gamma]$ is bijective,
let us choose a complete orthonormal system $(h_i)_{i \in \mathbb{N}}$ of $H$
and let us consider an element $f=[\sum_{j=1}^r \alpha_j \otimes
(\sum_{i=1}^{\infty}c_{ij}\underline{h_i})]$ of $H^k(d_{\bullet}^{\Phi^H}),$
where
$c_{ij}$ are functions on $M.$ The element $f'=\sum_{i=1}^{\infty}[\sum_{j=1}^r
c_{ij}\alpha_j]\otimes h_i \in H^k_{dR}(M)\otimes H$ is a well defined
element of the tensor product, as can
be seen by the following consideration. Since
$c_{ij}$ are smooth (by an iterative use of Lemma 17),
$\sum_{j=1}^rc_{ij}\alpha_j$ is smooth for each $i \in \mathbb{N}.$ Further, the
sum $\sum_{i=1}^{\infty}\sum_{j=1}^r
c_{ij}\alpha_j\otimes h_i,$ denoted by $f'',$ converges pointwise on the
manifold to
$\sum_{j=1}^r \alpha_j\otimes \sum_{i=1}^{\infty}c_{ij}\underline{h_i},$ that converges
pointwise in each point on the manifold by the
choice of the representation for $f$.
Let us consider the operator
$T: Z^k_{dR}(M) \otimes H \to H^k_{dR}(M) \otimes H$ given as the
linear extension of $T(\alpha \otimes
h) = [\alpha] \otimes h.$ Let us see that this operator is continuous.
Since $M$ is compact, $H^k_{dR}(M)$ is finite dimensional and $B^k_{dR}(M)$ is
orthogonally complemented by the usual Hodge theory. Therefore, we can define the
orthogonal projection $\mbox{pr}$ of $Z^k_{dR}(M)$
onto the complement, which is the space of harmonic forms.
The quotient norm of an element $[\alpha]$ in the de Rham cohomology group
equals to the norm of $\mbox{pr}(\alpha)$
in the cocycle space. Consequently, the norm of $T$ is bounded by the norm of
$\mbox{pr}$ which is one. Now, $f'$ is the $T$-image of $f''$ by the
continuity of $T.$ Especially, $f'$ is well defined (from the point of view
of convergence).
The facts that $\sum_{j=1}^r \alpha_j \otimes
\sum_{i=1}^{\infty}c_{ij}\underline{h_i}$ is closed and that $\underline{h}_i$'s are
constant imply
\begin{eqnarray*}
0 &=& d_{k}^{\Phi^H}(\sum_{j=1}^r \alpha_j \otimes
\sum_{i=1}^{\infty}c_{ij}\underline{h_i}) \\
&=&\sum_{j=1}^r \left(d_{k}\alpha_j \otimes
\sum_{i=1}^{\infty} c_{ij}\underline{h}_i + \sum_{p=1}^{2n}\epsilon^p \wedge
\alpha_j \otimes \sum_{i=1}^{\infty} e_p(c_{ij})\underline{h}_i\right)\\
&=&\sum_{j=1}^r\left(d_{k}\alpha_j \otimes\sum_{i=1}^{\infty}
c_{ij}\underline{h}_i + \sum_{i=1}^{\infty}(d_0c_{ij})\wedge
\alpha_j \otimes \underline{h}_i\right)\\
&=& \sum_{i=1}^{\infty}\sum_{j=1}^rd_k(c_{ij}\alpha_j) \otimes \underline{h}_i
\end{eqnarray*}
Since $(h_i)_{i\in \mathbb{N}}$ is linearly independent,
$\sum_{j=1}^rc_{ij}\alpha_j$ is closed for each $i\in \mathbb{N}.$ Especially,
$f'\in H^k_{dR}(M)\otimes H.$
Since $[\gamma]$ is continuous,
$f'=\sum_{i=1}^{\infty}[\sum_{j=1}^r
c_{ij}\alpha_j]\otimes h_i$ is a preimage of $f.$ Summing-up, $[\gamma]$ is surjective.
For the injectivity, let us
suppose that there is an element $\sum_{l=1}^s\sum_{i=1}^{\infty}
[\alpha_l] \otimes c_{li}h_i$ mapped to $[\sum_{l=1}^s
\sum_{i=1}^{\infty}\alpha_l \otimes c_{li}\underline{h}_i]=0$ by $[\gamma],$
where $c_{li}$ are complex numbers.
Thus, there exist smooth exterior differential $(k-1)$-forms $\widetilde{\alpha}_j,$
smooth functions $\widetilde{c}_j$ and smooth sections
$v_j \in \Gamma(\underline{H}),$ $j=1, \ldots, r,$ such that
\begin{eqnarray*}
\sum_{l=1}^s\sum_{i=1}^{\infty}c_{li}\alpha_l \otimes \underline{h}_i =
d^{\Phi^H}_{k-1} \left(\sum_{j=1}^r\widetilde{c}_j\widetilde{\alpha}_j \otimes
v_j\right)
\end{eqnarray*} For any $v_j,$ there are
smooth functions $\widetilde{c}_{ij}$ on $M$ (Lemma 17) for which
$v_j=\sum_{i=1}^{\infty}\widetilde{c}_{ij}\underline{h}_i.$
Using the Leibniz rule, we get
\begin{eqnarray*}
&&\sum_{l=1}^s\sum_{i=1}^{\infty}c_{li} \alpha_l \otimes
\underline{h}_i=d_{k-1}^{\Phi^H}\left(\sum_{j=1}^r \widetilde{c}_j\widetilde{\alpha}_j
\otimes \sum_{i=1}^{\infty}\widetilde{c}_{ij}\underline{h}_i\right)=\\
&&d_{k-1} \left(\sum_{j=1}^r \widetilde{c}_j\widetilde{\alpha}_j\right) \otimes \sum_{i=1}^{\infty}\widetilde{c}_{ij}\underline{h}_i
+ \sum_{p=1}^{2n}\sum_{j=1}^r \widetilde{c}_j\epsilon^p \wedge
\widetilde{\alpha}_j \otimes \nabla^H_{e_p}(\sum_{i=1}^{\infty}
\widetilde{c}_{ij} \underline{h}_i) =:(*)
\end{eqnarray*}
Using Lemma 17, we can take the derivative behind the infinite sum, getting
\begin{eqnarray*}
(*) &=& \sum_{j=1}^r \left( d_{k-1}(c_j \widetilde{\alpha}_j) \otimes \sum_{i=1}^{\infty} \widetilde{c}_{ij}
\underline{h}_i\right) + \sum_{j=1}^r \left(\sum_{p=1}^{2n} \widetilde{c}_j
\epsilon^p \wedge \widetilde{\alpha}_j \otimes
\sum_{i=1}^{\infty}e_p(\widetilde{c}_{ij}) \underline{h}_i\right) \\
&=& \sum_{i=1}^{\infty} \sum_{j=1}^{r} \left(\widetilde{c}_{ij}
d_{k-1}(\widetilde{c}_j\widetilde{\alpha}_j) + \sum_{p=1}^{2n}
e_p(\widetilde{c}_{ij})\epsilon^p \wedge \widetilde{c}_j \widetilde{\alpha}_j
\right) \otimes \underline{h}_i\\
&=& \sum_{i=1}^{\infty} \sum_{j=1}^{r} \left(d_{k-1}(\widetilde{c}_{ij} \widetilde{c}_j \widetilde{\alpha}_j)\right) \otimes \underline{h}_i
\end{eqnarray*}
since $d_0\widetilde{c}_{ij}=\sum_{p=1}^{2n}e_p(\widetilde{c}_{ij})\epsilon^p.$
Summing-up, $\sum_{l=1}^sc_{li} \alpha_l = \sum_{j=1}^r
d_{k-1}(\widetilde{c}_{ij}\widetilde{c}_j\alpha_j)$
for each $i \in \mathbb{N}$.
In particular, $\sum_{l=1}^s c_{li}\alpha_l$ is exact for each $i\in
\mathbb{N}.$
Thus, $\sum_{i=1}^{\infty} \sum_{l=1}^s[\alpha_l]\otimes c_{li}h_i = \sum_{i=1}^{\infty} [\sum_{l=1}^sc_{li}\alpha_l]\otimes h_i = 0$ and
$[\gamma]$ is injective.
The verification of the $CH$-equivariance of $[\gamma]$
is straightforward.
4) The inverse to $[\gamma]$ which we denote by $[\delta],$ is defined by
$[\delta]([\sum_{j=1}^r
\alpha_j \otimes \sum_{i=1}^{\infty}c_{ij}\underline{h}_i]) =
\sum_{i=1}^{\infty}[\sum_{j=1}^r
c_{ij}\alpha_j] \otimes h_i$ where $c_{ij}$ are smooth
functions on $M$ and $\alpha_j \in \Omega^k(M),$ $j=1,\ldots, r$.
The sum at the right-hand side can be shown to converge by undertaking
a similar procedure as in the proof of the surjectivity of $[\gamma].$
Further, it is immediate to check that $[\delta]$ maps cocycles into
cocycles and vanishes on coboundaries. Summing-up, $[\delta]$ is well
defined.
Let us define a map $\Xi(\sum_{j=1}^r\alpha_j \otimes \sum_{i=1}^{\infty}c_{ij}
\underline{h}_i) = \sum_{i=1}^{\infty}\sum_{j=1}^r c_{ij} \alpha_j \otimes
h_i.$ Since $M$ is compact, the norm of $\Xi$ is easily seen to be finite and
thus, $\Xi$ is continuous as a linear map between normed spaces.
The continuity of $[\delta]$ is established
by the use of the commutative diagram ($q_1 \otimes \mbox{Id}_H$ and $\Xi$
are continuous, and $q_2$ is open)
$$\begin{xy}\xymatrix{
Z^k_{dR}(M)\otimes H \ar[d]^{ q_1 \otimes \mbox{Id}_H } &
Z^k(d_{\bullet}^{\Phi^H}) \ar@{->}[l]_-{\Xi} \ar[d]^{q_2} \\
H^k_{dR}(M)\otimes H & H^k(d_{\bullet}^{\Phi^H})
\ar@{->}[l]_-{[\delta]}}
\end{xy}$$
Summing-up, we have
$H^k(d^{\Phi^H}_{\bullet}) \simeq H^k_{dR}(M)\otimes H$ as $CH$-modules (by item 3) and
as topological spaces by this item.
5) Notice that we already proved that $H^k(d^{\Phi}_{\bullet})$ and
$H^k(d^{\Phi^H}_{\bullet})$ are isomorphic as $CH$-modules and
topological spaces (item 2).
Since $H^k_{dR}(M)$ is finite dimensional, we may use the projection
$\mbox{pr}$ defined in item 3. The quotient topology on the cohomology
group is generated by the scalar product
$([\alpha],[\beta])_{H^k_{dR}}=(\mbox{pr}(\alpha),\mbox{pr}(\beta))_{\Omega^k}$
where at the right-hand side, the Hodge scalar product on differential forms
is meant.
Consequently, $H^k_{dR}(M)\otimes H$ possesses a pre-Hilbert
$C^*$-product given by $([\alpha]\otimes h, [\beta]\otimes
h')_{\otimes}=([\alpha],[\beta])_{H^k_{dR}} (h,h')$ where at the
right-hand side, $(h,h')$ denotes the Hilbert $CH$-product on
$H$. The norm induced by this
pre-Hilbert
$CH$-product coincides with the norm
induced by the scalar product considered on the tensor product (Lemma 1).
By the finite dimensionality of the de Rham cohomology groups,
$H^k_{dR}(M)\otimes H$ is a complete inner product space.
In particular, the pre-Hilbert $CH$-product on the tensor product of
the de Rham cohomology group and $H$ is a Hilbert $CH$-product.
Since $H^k(d_{\bullet}^{\Phi^H})$ is homeomorphic
and $CH$-isomorphic to $H^k_{dR}(M) \otimes H$ (item 4) we can equip it with
the inner product
$([s_1],[s_2])=\left([\delta]([s_1]), [\delta]([s_2])\right)_{\otimes}$ where
$s_1,s_2$ are cocycles for $d^{\Phi^H}_{\bullet}$.
Because $[\delta]$ is an isomorphism of $CH$-modules, the resulting
map is a well defined
Hilbert $C^*$-product. Since $[\delta]$ is a homeomorphism,
the topology on
$H^k(d_{\bullet}^{\Phi^H})$ generated by the norm induced by the scalar product
coincides with the
quotient topology on $H^k(d_{\bullet}^{\Phi^H}).$ Similarly, by
the use of the map $j$ (item 1 and item 2),
we induce the Hilbert $CH$-product from $H^k(d_{\bullet}^{\Phi^H})$ to
$H^k(d_{\bullet}^{\Phi})$ making it a Hilbert $CH$-module isomorphic to
$H^k_{dR}(M) \otimes H$ as well.
Because $H^k_{dR}(M)$ is finite dimensional and
$H$ is finitely generated (Lemma 2), $H^k_{dR}(M)\otimes H$
is also finitely generated.
Consequently, it is projective by the Magajna theorem.
Summing-up, $H^k(d^{\Phi}_{\bullet})$ are finitely generated projective Hilbert
$CH$-modules for each $k.$
\hfill\(\Box\)
\bigskip
We give a much shorter proof of Thm. 18 based on the Hodge theory
for $CH$-bundles.
\bigskip
{\it Second proof of Theorem 18.}
We use the Mishchenko--Fomenko theory elaborated for complexes in
\cite{KryslAGAG2} and \cite{KryslJGP2}.
It is immediate to compute that the symbols $s_k$ of $d_k^{\Phi}$ are given
by
$s_k(\alpha \otimes s, \tau) = (\tau \wedge \alpha) \otimes s,$
$\tau \in T_m^*M,$ $\alpha \otimes s \in (\mathcal{H}_R^k)_m$ (see
\cite{KryslDGA}),
and a matter of multilinear algebra to see that they form an exact sequence in
places $k=0,\ldots, 2n$ for any
$\tau \neq 0.$ In the places $k=0,\ldots, 2n,$ the Kuiper complex is thus
elliptic. Due to Lemma 2, it is a complex on finitely generated projective
Hilbert $CH$-bundles.
Consequently (compactness and ellipticity), its cohomology groups are finitely
generated and projective
Hilbert $CH$-modules by Thm. 9 in Kr\'ysl \cite{KryslJGP2}. In places
$\mathbb{Z}\setminus \{0,\ldots, 2n\},$ the cohomology groups
are zero.
\hfill\(\Box\)
\bigskip
{\bf Remark:} 1) Since the cohomology groups are Hilbert $C^*$-modules,
the images of $d_{\bullet}^{\Phi}$ and $d_{\bullet}^{\Phi^H}$ are necessarily
closed.
2) The Hodge theory holds for the Kuiper complex by Thm. 9 in \cite{KryslJGP2} as
well. From the proof of Theorem 18 (and the proof of Lemma 2), we see that
the $CH$-rank of $H^k(d_{\bullet}^{\Phi})$ equals to the $k$-th Betti number
of $M.$ (See Baki\'c, Gulja\v s \cite{Bakic} for a definition of the $CH$-rank
if needed.)
\bibliographystyle{amsplain}
\begin{thebibliography}{99}
\bibitem{ALMP} Albin, P., Leichtnam, E., Mazzeo, R., Piazza, P., Hodge theory
on Cheeger spaces, \url{https://doi.org/10.1515/crelle-2015-0095}, on-line
published in 2016. ArXiv--preprint \url{https://arxiv.org/abs/1307.5473}.
\bibitem{Bakic} Baki\'c, D., Gulja\v s, B., Operators on Hilbert $H^*$-modules, J. Oper. Theory 46 (2001), No. 1,
123--137.
\bibitem{BirkhoffNeumann} Birkhoff, G., von Neumann, J., The logic of
quantum mechanics, Ann. Math. (2) 37, Princeton (1936), 823--843.
\bibitem{Bong} Bongioanni, B., Torrea, J.,
Sobolev spaces associated to the harmonic oscillator. In: Proc. Indian Acad. Sci. Math. Sci. 116 (2006), No. 3, 337--360.
\bibitem{BorelWallach} Borel, A., Wallach, N., Continuous cohomology, discrete
subgroups, and representations of reductive groups. Second edition. Mathematical
Surveys and Monographs, 67. American Mathematical Society, Providence, RI, 2000.
\bibitem{Bruhat} Bruhat, F., Sur les repr\'esentations induites des groupes de Lie,
Bull. Soc. Math. France 84, 1956, 97--205.
\bibitem{CGLR} Cahen, M., Gutt, S., La Fuente Gravy, L., Rawnsley, J., On
$Mp^c$-structures and symplectic Dirac operators. J. Geom. Phys. 86 (2014),
434--466.
\bibitem{Connes} Connes, A., Noncommutative geometry. Academic Press, Inc., San
Diego, CA, 1994.
\bibitem{DixmierDouady} Dixmier, J., Douady, A.,
Champs continus d'espaces hilbertiens et de $C^*$-alg\'ebres,
Bull. Soc. Math. France 91 (1963), 227--284.
\bibitem{Dixmier} Dixmier, J., Les $C^*$-alg\'ebres et leurs
repr\'esentations. Gauthier-Villars, Paris, 1969.
\bibitem{DM} Dubois-Violette, M., Michor, P.,
Connections on central bimodules in noncommutative differential geometry, J.
Geom. Phys. 20 (1996), Issue 2--3, 218--232.
\bibitem{FathiGabr} Fathizadeh, F., Gabriel, O., On the
Chern--Gauss--Bonnet theorem and conformally twisted spectral triples for
$C^*$-dynamical systems, SIGMA Symmetry Integrability Geom. Methods Appl. 12
(2016), paper No. 016.
\bibitem{Folland} Folland, G., Harmonic analysis in phase space. Annals of
Mathematics Studies, 122. Princeton Univ. Press, Princeton, NJ, 1989.
\bibitem{FM} Fomenko, A., Mishchenko, A., The index of elliptic operators over
$C^*$-algebras, Math. USSR-Izv. 15, (1980), 87--112.
\bibitem{FH} Forger, M., Hess, H., Universal metaplectic structures and
geometric quantization, Comm. Math. Phys. 64 (1979), No. 3, 269--278.
\bibitem{Freed} Freed, D., Lott, J., An index theorem in differential
$K$-theory, Geom. Topol. 14 (2010), No. 2,
903--966.
\bibitem{Goodman} Goodman, R., Analytic and entire vectors for representations
of Lie groups. Trans. Amer. Math. Soc. 143 (1969), 55--76.
\bibitem{Groth} Grothendieck, A., Produits tensoriels topologiques et Espaces
nucl\'eaires. Amer. Mat. Soc., Providence, RI, 1966.
\bibitem{HabermannAn} Habermann, K., The Dirac operator on symplectic spinors.
Annals Glob. Anal. Geom. 13 (1995), No. 2, 155--168.
\bibitem{HH} Habermann, K., Habermann, L., Introduction to Symplectic Dirac Operators, Lecture Notes in Mathematics 1887.
Springer-Verlag, Berlin, 2006.
\bibitem{Hain} Hain, R., The Hodge--de Rham theory of modular groups, In: Recent
advances in Hodge theory, London Math. Soc. Lecture Note Ser., 427.
Cambridge Univ. Press, Cambridge, 2016, 422--514.
\bibitem{HenTeitel} Henneaux, M., Teitelboim, C., Quantization of Gauge
Systems, Princeton University Press, Princeton, NJ, 1992.
\bibitem{HerczegWaldron} Herczek, G., Waldron, A., Contact geometry and
Quantum mechanics, Physics Letters, Section B: Nuclear, Elementary Particle and
High-Energy Physics, 781, 2018, 312--315,
\url{https://doi.org/10.1016/j.physletb.2018.04.008}.
\bibitem{Hodge} Hodge, W., The theory and applications of harmonic integrals.
2nd ed. Cambridge, at the Univ. Press, 1952.
\bibitem{Illusie} Illusie, L., Complexes quasi-acycliques directs de fibr\'es
banachiques, C. R. Acad. Sci. Paris 260, 1965, 6499--6502.
\bibitem{Keyl} Keyl, M., Kiukas, J., Werner, R.,
Schwartz operators, Rev. Math. Phys. 28 (2016), No. 3, 1630001, 60 pp.
\bibitem{Kim} Kim, J., A splitting theorem for holomorphic Banach bundles,
Math. Z. 263, 2009, No. 1, 89--102.
\bibitem{Kirillov} Kirillov, A., Lectures on the orbit method. Graduate
Studies in Mathematics, 64. American Mathematical Society, Providence, RI, 2004.
ISBN 0-8218-3530-0.
\bibitem{Klein} Klein, F., \"Uber die Entwicklung der Mathematik im 19.
Jahrhundert. Band 1, Springer, Berlin, 1926. Engl. transl. by M. Ackermann:
Development of Mathematics in the 19th Century with Appendices
"Kleinian Mathematics from an Advanced Standpoint" by R. Hermann, 1st Edition,
Math Sci Press, Brookline, MA, 1979.
\bibitem{KnappOverview} Knapp, A., Representation Theory of Semisimple Groups. An Overview Based on Examples,
Reprint of the 1986 original. Princeton Landmarks in Mathematics, Princeton Univ. Press, Princeton, 2001.
\bibitem{KMS} Kol\'a\v{r}, I., Michor, P., Slov\'ak, J., Natural operations in
differential geometry. Springer-Verlag, Berlin, 1993.
\bibitem{Kostant} Kostant, B., Symplectic spinors, In: Symposia Mathematica, Vol. XIV
(Convegno di Geometria Simplettica e Fisica Matematica, INDAM, Rome, 1973). Academic Press, London, 1974, 139--152.
\bibitem{KryslDGA0} Kr\'ysl, S., Classification of 1st order symplectic spinor operators over contact projective geometries,
Vol. 26 (2008) Issue 5, 553--565.
\bibitem{KryslDGA} Kr\'ysl, S., Cohomology of the de Rham complex twisted by the
oscillatory representation, Diff. Geom. Appl. 33 (2014), No. 5,
290--297.
\bibitem{KryslAGAG2} Kr\'ysl, S., Hodge theory for self-adjoint parametrix
possessing complexes over $C^*$-algebras, Annals Glob. Anal. Geom. 47
(2015), No. 4, 359--372.
\bibitem{KryslJGP2} Kr\'ysl, S., Elliptic complexes over $C^*$-algebras of
compact operators, Journal of Geometry and Physics 101 (2016), 27--37.
\bibitem{KryslHabil} Kr\'ysl, S., Hodge Theory and Symplectic Spinors.
Habilitation thesis. Faculty of Mathematics and Physics, Charles University,
Prague, 2016, electronically available at
\url{http://msekce.mff.cuni.cz/~krysl/habil.pdf},
short version at ArXiv \url{https://arxiv.org/abs/1708.02026}.
\bibitem{Kuiper} Kuiper, N., The homotopy type of the unitary group of Hilbert space, Topology 3 (1965), 19--30.
\bibitem{Larrain} Larra\'in-Hubach, A., $K$-theories for classes of infinite
rank bundles. In: Analysis, geometry and quantum field theory,
Contemp. Math., 584. Amer. Math. Soc., Providence, RI, 2012, 79--97.
\bibitem{Lempert} Lempert, L., On the cohomology groups of holomorphic Banach
bundles, Trans. Amer. Math. Soc. 361 (2009), No. 8, 4013--4025.
\bibitem{Li} Li, P., Cauchy--Schwarz-type inequalities on K\"{a}hler
manifolds--II, J. Geom. Phys. 99 (2016), 256--262.
\bibitem{Ludwig} Ludwig, G., Versuch einer axiomatischen Grundlegung der
Quantenmechanik und allgemeinerer physikalischer Theorien, Zeitschrifft f\"ur
Physik 181 (1964), Issue 3, 233--260.
\bibitem{MaedaRosenberg} Maeda, Y., Rosenberg, S., Traces and characteristic
classes in infinite dimensions, Geometry and analysis on manifolds, 413--435,
Progr. Math., 308, Birkh\"{a}user/Springer, Cham (CH), 2015.
\bibitem{Magajna} Magajna, B., Hilbert $C^*$-modules in which all closed submodules are complemented,
Proc. Amer. Math. Soc. 125 (1997), No. 3, 849--852.
\bibitem{MMS} Mathai, V., Melrose, R., Singer, I., The index of projective
families of elliptic operators: the decomposable case. In:
Dai, X. (ed.) et al., From probability to geometry II (Volume in honor of
the 60th birthday of Jean-Michel Bismut). Paris SMF,
Ast\'erisque 328, 255--296 (2009).
\bibitem{Maurin} Maurin, K., The Riemann legacy. Riemannian ideas in
mathematics and physics. Mathematics and its Applications, 417. Kluwer Academic
Publishers Group, Dordrecht, 1997.
\bibitem{Michor} Michor, P., Topics in differential geometry.
Graduate Studies in Mathematics 93. Providence, RI, American Mathematical Society, 2008.
\bibitem{Mishchenko} Mishchenko, A., The theory of elliptic operators over
$C^*$-algebras, Dokl. Akad. Nauk SSSR 239 (1978), No. 6, 1289--1291.
\bibitem{Neeb} Neeb, K., On differentiable vectors for representations of
infinite dimensional Lie groups. J. Funct. Anal. 259 (2010), No. 11, 2814--2855.
\bibitem{Nekovar} Nekov\'a\v{r}, J., Scholl, A., Introduction to plectic
cohomology. In: Jiang, D., Shahidi, F., Soudry, D. (eds.), Advances in the theory of automorphic forms and their
$L$-functions, 321--337, Contemp. Math., 664, Amer. Math. Soc., Providence, RI, 2016.
\bibitem{NeumannKM} von Neumann, J., Zur Operathorenmethode in der klassischen
Mechanik, Ann. Math. (2) 33 (1932), 257--642.
\bibitem{Palais} Palais, R., Seminar on the Atiyah--Singer index
theorem (with contributions by M. F. Atiyah, A. Borel, E. E. Floyd, R. T.
Seeley, W. Shih and R. Solovay), Annals of Mathematics Studies, No. 57. Princeton
Univ. Press, Princeton, NJ, 1965.
\bibitem{Paschke} Paschke, W., Inner product modules over $B^*$-algebras,
Trans. Amer. Math. Soc. 182 (1973), 443--468.
\bibitem{Raeburn} Raeburn, I., Williams, D., Morita equivalence and continuous-trace $C^*$-algebras.
Mathematical Surveys and Monographs, 60. American Mathematical Society, Providence, RI, 1998.
ISBN 0-8218-0860-5.
\bibitem{Rieffel} Rieffel, M., Induced representations of
$C^*$-algebras, Advances in Mathematics 13 (2) (1973),
176--257.
\bibitem{RR} Robinson, P., Rawnsley, J., The metaplectic representation, $Mp^c$-structures and geometric quantization,
Mem. Am. Math. Soc. 410, 1989.
\bibitem{Rohrl} R\"{o}hrl, H., \"{U}ber die Kohomologie berechenbarer Fr\'echet
Garben, Comment. Math. Univ. Carolinae 10 (1969), 625--640.
\bibitem{Rordam} R\o rdam, M., Larsen, F., Laustsen, N. J., An introduction to $K$-theory for $C^*$-algebras, London Mathematical Society, Cambridge University Press,
2000.
\bibitem{Rosenberg} Rosenberg, J., Topology, $C^*$-algebras, and string duality.
CBMS Regional Conference Series in Mathematics, 111. Published for the Conference Board of the Mathematical Sciences, Washington, DC,
American Mathematical Society, Providence, RI, 2009.
\bibitem{Ryan} Ryan, R., Introduction to tensor products of Banach Spaces,
Springer Monographs in Mathematics. London, 2002.
\bibitem{Schick} Schick, T., $L^2$-index theorems, $KK$-theory,
and connections, New York J. Math. 11 (2005), 387--443.
\bibitem{SchmidVilonen} Schmid, W., Vilonen, K., Hodge theory and
unitary representations. Representations of reductive groups, 443--453, Progr.
Math., 312, Birkh\"{a}user/Springer, Cham (CH), 2015.
\bibitem{Schotten} Schottenloher, M., The unitary group and its strong
topology. ArXiv--preprint, \url{https://arxiv.org/abs/1309.5891}.
\bibitem{Shale} Shale, D., Linear symmetries of free boson fields, Trans. Amer. Math. Soc. 103 (1962), 149--167.
\bibitem{ST} Solovyev (Solov'\"{e}v), Y., Troitsky (Troickij), E., $C^*$-algebras and
elliptic operators in differential topology. Transl. of Mathem. Monographs
192, Amer. Math. Soc., Providence, RI, 2001.
\bibitem{Troitsky} Troitsky, E., The index of equivariant elliptic operators
over $C^*$-algebras, Annals Glob. Anal. Geom., 5 (1987), No. 1, 3--22.
\bibitem{Tsai} Tsai, C., Tseng, L., Yau, S., Cohomology and
Hodge theory on symplectic manifolds: III. J. Differential Geom. 103 (2016), No.
1, 83--143.
\bibitem{Wallach} Wallach, N., Symplectic geometry and Fourier analysis. With
an appendix on quantum mechanics by Robert Hermann. Lie Groups: History,
Frontiers and Applications, Vol. V. Math Sci Press, Brookline, MA, 1977.
\bibitem{Warner} Warner, G., Harmonic analysis on semi-simple Lie groups I.
Die Grundlehren der mathematischen Wissenschaften, Band 188. Springer-Verlag,
New York--Heidelberg, 1972.
\bibitem{Weil} Weil, A., Sur certains groupes d'op\'erateurs unitaires, Acta Math. No. 111 (1964), 143--211.
\end{thebibliography}
\end{document}