Introduction to Symplectic Dirac Operators

Lecture Notes in Mathematics Editors: J.-M. Morel, Cachan F. Takens, Groningen B. Teissier, Paris 1887 K. Habermann ·...

Author: Katharina Habermann Lutz Habermann

37 downloads 571 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Lecture Notes in Mathematics Editors: J.-M. Morel, Cachan F. Takens, Groningen B. Teissier, Paris

1887

K. Habermann · L. Habermann

Introduction to Symplectic Dirac Operators

ABC

Authors Katharina Habermann State and University Library Göttingen Platz der Göttinger Sieben 1 37073 Göttingen Germany e-mail: [email protected] Lutz Habermann Department of Mathematics University of Hannover Welfengarten 1 30167 Hannover Germany e-mail: [email protected]

Library of Congress Control Number: 2006924i23 Mathematics Subject Classification (2000): 53-02, 53Dxx, 58-02, 58Jxx ISSN print edition: 0075-8434 ISSN electronic edition: 1617-9692 ISBN-10 3-540-33420-3 Springer Berlin Heidelberg New York ISBN-13 978-3-540-33420-0 Springer Berlin Heidelberg New York DOI 10.1007/b138212

This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable for prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springer.com c Springer-Verlag Berlin Heidelberg 2006 Printed in The Netherlands The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. A EX package Typesetting: by the authors and SPI Publisher Services using a Springer LT Cover design: design & production GmbH, Heidelberg

Printed on acid-free paper

SPIN: 11736684

V A 41/3100/ SPI

543210

To Karen

Preface

This book aims to give a systematic and self-contained introduction to the theory of symplectic Dirac operators and to reflect the current state of the subject. At the same time, it is intended to establish the idea that symplectic spin geometry and symplectic Dirac operators may give valuable tools in symplectic geometry and symplectic topology, which have become important fields and very active areas of mathematical research. The basic idea of symplectic spin geometry goes back to the early 1970s, when Bertram Kostant introduced symplectic spinors in order to give the construction of the half-form bundle and the half-form pairings in the context of geometric quantization [37]. During the next two decades, however, almost no attention has been given to a closer study of symplectic spin geometry itself. In 1995, the first author introduced symplectic Dirac operators [24] and started a systematical investigation [25, 26, 27]. These symplectic Dirac operators are called Dirac operators, since they are defined in an analogous way as the classical Riemannian Dirac operator known from Riemannian spin geometry (cf. e.g. [21]). They are called symplectic because they are constructed by use of the symplectic setting of the underlying symplectic manifold. All tools which were necessary for that construction have already been known and accepted, mainly in mathematical physics. These are the symplectic Clifford algebra (also known as Weyl algebra), the metaplectic group, the metaplectic representation (Segal–Shale–Weil representation) acting on L2 (Rn ), metaplectic structures, and symplectic connections. One of the basic ideas in differential geometry is that the study of analytic properties of certain differential operators acting on sections of vector bundles yields geometric and topological properties of the underlying base manifold. There are several classical results in that direction. An example is Hodge– de Rham theory. Here, one considers the Hodge–Laplace–Beltrami operator ∆

VIII

Preface

acting on differential forms. This operator is one of the most studied operators in global Riemannian geometry and his spectrum gives important topological invariants. In particular, the dimension of the kernel of ∆ on p-forms over a closed Riemannian manifold is the p-th Betti number. Other well known and well studied operators are the Kodaira–Hodge–Laplace operator on differential forms with values in a holomorphic vector bundle or the classical Dirac operator on Riemannian manifolds. Now, symplectic spinor fields are sections in an L2 (Rn )-Hilbert space bundle over a symplectic manifold and symplectic Dirac operators, acting on symplectic spinor fields, are associated to the symplectic manifold in a very natural way. Hence they may be expected to give interesting applications in symplectic geometry and symplectic topology. It is our opinion that, besides the already stated, there are further close relations to mathematical physics. Some steps towards this direction have been made by the first author and Andreas Klein in [28, 29, 30]. Another perspective could be the extension of Clifford analysis and spin geometry to super differential geometry. According to the most used version of super geometry developed by Bertram Kostant, geometrical structures over a supermanifold consist of Z2 -graded objects and thus have an even as well as an odd part. Then one can imagine that a metric has to satisfy some kind of graded symmetry which, roughly speaking, corresponds to a symmetric object on the even part as well as to a skew symmetric object on the odd part of the supermanifold. For the first one, we have the classical Riemannian spin geometry, whereas the second one is basically given by symplectic spin geometry. The main aspects of that idea are treated in a paper by Frank Sommen [45]. Although the construction of symplectic Dirac operators follows the same procedure as for the classical Riemannian Dirac operator, using the symplectic structure of the underlying manifold instead of the Riemannian metric, there are essential differences to the Riemannian case. These are caused by the fact that the algebraic structure of the symplectic Clifford algebra is completely different from that of Riemannian spin geometry. For the classical Clifford algebra, we have the relation v2 = −v2 , whereas the algebraic structure of the symplectic Clifford algebra is given by v · w − w · v = −ω0 (v, w). This implies essentially different properties for the Clifford multiplications, which enter into the definition of the Dirac operators. Moreover, the non-compactness of the symplectic group leads to analytic difficulties. Namely, since the typical fiber of the symplectic spinor bundle is the Hilbert space L2 (Rn ), we deal with operators acting on sections of a vector bundle of infinite rank. For elliptic formally self-adjoint pseudo-differential operators with positive definite leading symbol acting on sections in a vector bundle of finite rank, one has a completely developed theory. So, in order to be able to apply these techniques, we are interested in equivariance properties

Preface

IX

of our operators with respect to a certain decomposition of the symplectic spinor bundle into a series of subbundles of finite rank. It turns out that an associated second order operator respects this decomposition provided that a technical assumption, which always can be realized, holds true. Let us now briefly describe the content of each chapter of this book. The first chapter is introductory. It contains preliminaries and basic material needed for our considerations. Chapter 2 is devoted to symplectic connections. In particular, we introduce a further Ricci tensor, which we will call symplectic Ricci tensor. To our knowledge, no attention has been given to this tensor in previous studies. In fact, this symplectic Ricci tensor is a new object in the case of non-vanishing torsion. To date, mostly only torsion-free symplectic connections have been considered. It turns out that, in our context, it is convenient to work also with symplectic connections with torsion and that the symplectic Ricci tensor is more suitable for our purposes. The next chapter introduces the symplectic spinor bundle and the spinor derivative and analyzes a splitting property of the spinor bundle. In Chapter 4, we give the definition of symplectic Dirac operators and describe in detail how these operators depend on the objects from which they are built. Chapter 5 is concerned with an associated second order operator of Laplace type and addresses properties of this operator. The objective of Chapter 6 is the situation for a special class of symplectic manifolds, namely K¨ ahler manifolds. Here, we also investigate the example of CP 1 . The aim of Chapter 7 is to construct a Fourier transform for symplectic spinor fields and to derive consequences for the symplectic Dirac operators. The last chapter focuses on relations to mathematical physics, in particular to quantization. This closes the circle to the beginnings by Bertram Kostant. The present text is originated in research ideas of the first author and provides an extended version of her “Habilitationsschrift”, which was never published separately. Starting with helpful discussions from the beginning of the investigations and proposing many improvements in the selection and presentation of the material, the second author became more and more involved into the subject. We decided to write this book tree years ago and, for this book, he made a thorough revision of the material, in particular, to improve the strictness of the presentation. Then it took time to compose it in a natural and organic way. Furthermore, our working places are no longer as close to each other as before and it became difficult to keep our discussions at its intensive level. Now we consider the text ready for publication and hope to present the reader a mature work. During the work on the book, we received financial support from DFG, the German Research Foundation, contract HA 3056/1-1,2. Many thanks are due to our students Paul Rosenthal and Steffen Rudnick for proof-reading the LATEX-type-written manuscript.

X

Preface

We dedicate this book to our daughter Karen. She accepts our mathematical family circle and enriches it with her interest. We are grateful to being able to give her an understanding of the fascination of mathematics. Karen, you are a wonderful teenager which makes us enjoying the challenge of teaching mathematics. G¨ ottingen and Greifswald, November 2004

Katharina & Lutz Habermann

Contents

1

Background on Symplectic Spinors . . . . . . . . . . . . . . . . . . . . . . . .

1

1.1 Symplectic Group and Clifford Algebra . . . . . . . . . . . . . . . . . . . . .

1

1.2 The Stone–von Neumann Theorem . . . . . . . . . . . . . . . . . . . . . . . .

5

1.3 Metaplectic Representation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7

1.4 Symplectic Clifford Multiplication . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.5 Hermite Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 2

Symplectic Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.1 Symplectic Manifolds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 2.2 Constructions and Torsion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.3 Symplectic Curvature and Ricci Tensors . . . . . . . . . . . . . . . . . . . . 29

3

Symplectic Spinor Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.1 Metaplectic Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.2 Symplectic Spinor Bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 3.3 Splitting of the Spinor Bundle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43

4

Symplectic Dirac Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.1 Definition of the Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.2 Dependence on the Symplectic Connection . . . . . . . . . . . . . . . . . . 52 4.3 Dependence on the Metaplectic Structure . . . . . . . . . . . . . . . . . . . 57 4.4 Dependence on the Almost Complex Structure . . . . . . . . . . . . . . 62 4.5 Formal Self-Adjointness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

XII

5

Contents

An Associated Second Order Operator . . . . . . . . . . . . . . . . . . . . 67 5.1 Definition and Ellipticity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 5.2 A Weitzenb¨ ock Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68 5.3 Splitting of the Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

6

The K¨ ahler Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 6.1 The Operator P on K¨ ahler Manifolds . . . . . . . . . . . . . . . . . . . . . . 81 6.2 Lower Bound Estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 6.3 The Spectrum of P on CP 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87

7

Fourier Transform for Symplectic Spinors . . . . . . . . . . . . . . . . . . 97 7.1 Definition of the Transform . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 7.2 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 . . . . . . . . . . . . . . . . . . . . . . 99 7.3 Symmetry of the Spectra of D and D

8

Lie Derivative and Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 8.1 Lie Derivative of Symplectic Spinor Fields . . . . . . . . . . . . . . . . . . 101 8.2 Schr¨ odinger Equation for Quadratic Hamiltonians . . . . . . . . . . . 109 8.3 Lie Derivative as Quantization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

1 Background on Symplectic Spinors

This chapter serves two purposes. First it gives a survey on fundamental relations between the symplectic Clifford algebra, the metaplectic group, and its Lie algebra. The second is to provide several elementary facts used in the later computations. Most of it is well known, but we have summarized the material in a form that makes it applicable for our considerations.

1.1 Symplectic Group and Clifford Algebra We consider the 2n-dimensional real vector space R2n equipped with its standard symplectic form ω0 . We write any v ∈ R2n as v v= v with vectors v , v ∈ Rn and any 2n × 2n-matrix A as ab A= cd with n × n matrices a, b, c, d. Let I be the unit element of the linear group GL(n, R) and set 0 −I J0 = . I 0 Then

ω0 (v, w) = J0 v, w = v , w − v , w

for v, w ∈ R2n , where , denotes the Euclidean inner product on R2n as well as on Rn . Hence the standard basis {a1 , . . . , an , b1 , . . . , bn } of R2n forms a symplectic basis, which means that

2


ω0 (aj , ak ) = ω0 (bj , bk ) = 0

and ω0 (aj , bk ) = δjk

(1.1.1)

for j, k = 1, . . . , n. The symplectic group Sp(n, R) is the group of all automorphisms of R2n which preserve the symplectic form ω0 . That is, Sp(n, R) is the group of those A ∈ GL(2n, R) that satisfy ω0 (Av, Aw) = ω0 (v, w) for all v, w ∈ R2n , which is equivalent to AT J0 A = J0 as well as AJ0 AT = J0 . In particular, J0 ∈ Sp(n, R). We denote the space of all symmetric real n × n-matrices by S(n) and set a 0 D(n) = : a ∈ GL(n, R) −1 0 aT

and N (n) =

Ib 0I

: b ∈ S(n) .

Obviously, D(n) and N (n) are subgroups of Sp(n, R). Moreover, we have (cf. [16]) Proposition 1.1.1 Sp(n, R) is generated by D(n) ∪ N (n) ∪ {J0 }.

Identifying R2n with Cn via v → v + iv , the Hermitian inner product of two vectors v, w ∈ R2n is v, w − iω0 (v, w) . Therefore, the intersection of Sp(n, R) with the orthogonal group O(2n) is the unitary group U(n). One can prove (cf. [16]) Proposition 1.1.2 (1) U(n) is a maximal compact subgroup of Sp(n, R). 2

(2) Sp(n, R) is homeomorphic to the product U(n) × Rn

+n

.

Corollary 1.1.3 Sp(n, R) is connected and its fundamental group is Z.

1.1 Symplectic Group and Clifford Algebra

3

Proof. This follows from Proposition 1.1.2 and the corresponding properties of U(n). The symplectic Lie algebra, i.e. the Lie algebra sp(n, R) of the symplectic group Sp(n, R) is given by the space of all endomorphisms X of R2n satisfying ω0 (Xv, w) + ω0 (v, Xw) = 0 for all v, w ∈ R2n . We identify the space S 2 (R2n ) of symmetric 2-tensors of R2n with sp(n, R) by assigning to v1 v2 ∈ S 2 (R2n ) the endomorphism v ∈ R2n → ω0 (v, v1 )v2 + ω0 (v, v2 )v1 ∈ R2n . Then we have Lemma 1.1.4 For every X ∈ sp(n, R), 1 (Xaj bj − aj Xbj ) . 2 j=1 n

X=

Proof. The assertion follows from n

(ω0 (v, Xaj )bj + ω0 (v, bj )Xaj − ω0 (v, aj )Xbj − ω0 (v, Xbj )aj )

j=1

=

n

(ω0 (Xv, bj )aj − ω0 (Xv, aj )bj )

j=1

  n +X  (ω0 (v, bj )aj − ω0 (v, aj )bj ) j=1

= 2 Xv for any v ∈ R2n .

The Lie bracket of sp(n, R) now writes as [v1 v2 , v3 v4 ] = −ω0 (v2 , v4 )v1 v3 − ω0 (v2 , v3 )v1 v4 −ω0 (v1 , v4 )v2 v3 − ω0 (v1 , v3 )v2 v4 . In matrix notation, the above identification is given by

v1 ⊗ v2 + v2 ⊗ v1 −v1 ⊗ v2 − v2 ⊗ v1 v1 v2 = . v1 ⊗ v2 + v2 ⊗ v1 −v1 ⊗ v2 − v2 ⊗ v1 Here, x ⊗ y for x, y ∈ Rn means the endomorphism

(1.1.2)

4


z ∈ Rn → x, z y ∈ Rn . Clearly, the endomorphism adjoint to x ⊗ y is then (x ⊗ y)T = y ⊗ x .

(1.1.3)

Next we define the symplectic Clifford algebra, which is also referred to as Weyl algebra. In contrast to the Riemannian case, this algebra is infinite dimensional. Definition 1.1.5 The symplectic Clifford algebra Cl(n) is the associative unital algebra over R generated by R2n with the relations v · w − w · v = −ω0 (v, w) for v, w ∈ R2n . A basis of Cl(n) is formed by β1 β2 α2 αn βn 1 aα 1 · a2 · · · · · an · b1 · b2 · · · · · bn ,

where αj , βj are non-negative integers. According to Equation (1.1.1), aj · ak = ak · aj ,

bj · bk = bk · bj ,

and aj · bk − bk · aj = −δjk

for j, k = 1, . . . , n. As usual, we set [v, w] = v · w − w · v for v, w ∈ Cl(n), which gives Cl(n) the structure of a real Lie algebra. Let a(n) denote the subspace of Cl(n) which is spanned by v · w + w · v for v, w ∈ R2n . Lemma 1.1.6 The space a(n) is a Lie subalgebra of Cl(n) which is isomorphic to the symplectic Lie algebra sp(n, R). Proof. Let v1 , v2 , v ∈ R2n and v ∈ a(n). First we observe that [v1 · v2 , v] = v1 · v2 · v − v · v1 · v2 = v1 · v · v2 + ω0 (v, v2 )v1 − v · v1 · v2 = ω0 (v, v1 )v2 + ω0 (v, v2 )v1 = (v1 v2 )v . Thus [v1 · v2 + v2 · v1 , v] = 2(v1 v2 )v and

(1.1.4)

1.2 The Stone–von Neumann Theorem

[v, v] ∈ R2n .

5

(1.1.5)

Further we have [v1 · v2 , v] = v1 · [v2 , v] + [v1 , v] · v2 and hence [v1 · v2 + v2 · v1 , v] = v1 · [v2 , v] + [v2 , v] · v1 +v2 · [v1 , v] + [v1 , v] · v2 .

(1.1.6)

Equations (1.1.5) and (1.1.6) imply that a(n) is a Lie subalgebra of Cl(n). From Equation (1.1.4) and the Jacobi identity, we get the second part of the assertion.

1.2 The Stone–von Neumann Theorem In this section, we want to recall some facts of representation theory and formulate the Stone–von Neumann theorem. This fundamental theorem will be used in the next section to construct the metaplectic representation. For details and proofs, we refer to [1, 16, 49]. Let G be a connected Lie group and let H be a separable complex Hilbert space. We endow the group GL(H) of invertible bounded linear operators on H with the strong topology. That means that a sequence (Tk ) in GL(H) converges to T0 if and only if the sequence (Tk h) converges to T0 h for all h ∈ H. Definition 1.2.1 A representation of G on H is a continuous group homomorphism r : G → GL(H). It is called unitary if it maps into the unitary group U(H) of H. Definition 1.2.2 Let r be a representation of G on H. (1) A subspace W ⊂ H is said to be r-invariant if r(a)W ⊂ W for all a ∈ G. (2) r is called irreducible if the only r-invariant closed subspaces of H are {0} and H. The following proposition is a version of Schur’s lemma.

6


Proposition 1.2.3 A unitary representation r of G on H is irreducible if and only if the only bounded linear operators T : H → H such that T ◦ r(a) = r(a) ◦ T for all a ∈ G are scalar multiples of the identity.

Let r be a fixed representation of G on H and let H∞ denote the space of smooth vectors of r, i.e. the set of all h ∈ H such that the map a ∈ G → r(a)h ∈ H is smooth. Theorem 1.2.4 (G˚ arding) H∞ is a dense invariant subspace of H.

The differential of r is the homomorphism r∗ : g → End(H∞ ) from the Lie algebra g of G into the endomorphism algebra End(H∞ ) of H∞ given by d r∗ (X)h = r(exp(tX))h dt t=0 for X ∈ g and h ∈ H∞ . By Theorem 1.2.4, each operator r∗ (X) can be considered as an unbounded operator on H. Definition 1.2.5 Let r1 : G → U(H1 ) and r2 : G → U(H2 ) be representations of G on Hilbert spaces H1 and H2 . Then r1 is said to be equivalent to r2 if there exists a bounded linear bijection T : H1 → H2 such that T ◦ r1 (a) = r2 (a) ◦ T for all a ∈ G. If, in addition, T can be chosen unitary, r1 is called unitary equivalent to r2 . Now we consider the Heisenberg group H(n) of R2n , i.e. H(n) = R2n × R with group multiplication given by 1 (v, s) · (w, t) = v + w, s + t + ω0 (v, w) . 2 Setting

(rS (v, s)f )(x) = e−i(s+v ,x−v ,v

/2)

f (x − v )

for (v, s) ∈ H(n), f : Rn → C and x ∈ Rn , we obtain a unitary representation rS of H(n) on the Hilbert space L2 (Rn ) of square integrable functions on Rn . This representation, which is called Schr¨ odinger representation, is irreducible. Furthermore, rS (0, s)f = e−is f for any s ∈ R and f ∈ L2 (Rn ). These properties turn out to be characteristic for rS .

1.3 Metaplectic Representation

7

Theorem 1.2.6 (Stone–von Neumann) Let r be an irreducible unitary representation of the Heisenberg group H(n) on a separable complex Hilbert space H such that r(0, s)h = e−is h for all s ∈ R and h ∈ H. Then r is unitary equivalent to the Schr¨ odinger representation rS .

1.3 Metaplectic Representation In view of Corollary 1.1.3, the symplectic group Sp(n, R) has a unique connected double cover Mp(n, R). This covering group is called metaplectic group. Let ρ : Mp(n, R) → Sp(n, R) denote the covering homomorphism. By Lemma 1.1.6, we may identify the Lie algebra mp(n, R) of Mp(n, R) with the subalgebra a(n) of Cl(n) and think the differential ρ∗ : mp(n, R) → sp(n, R) of ρ to be realized by the homomorphism given in the proof of this lemma. So ρ∗ (v)v = [v, v]

(1.3.1)

ρ∗ (v · w + w · v) = 2 v w

(1.3.2)

and for v ∈ mp(n, R) and v, w ∈ R2n . The inverse of ρ∗ is explicitly described by Lemma 1.3.1 For every X ∈ sp(n, R), 1 (bj · Xaj − aj · Xbj ) . 2 j=1 n

ρ−1 ∗ (X) =

Proof. By bj · Xaj = Xaj · bj − ω0 (bj , Xaj ) , aj · Xbj = Xbj · aj − ω0 (aj , Xbj ) and ω0 (bj , Xaj ) = ω0 (aj , Xbj ) , one has n

(bj · Xaj − aj · Xbj )

j=1

1 (Xaj · bj + bj · Xaj − aj · Xbj − Xbj · aj ) . 2 j=1 n

=

8


Applying Lemma 1.1.4 and Equation (1.3.2), one gets the assertion.

In the following, we outline the construction of a unitary representation of Mp(n, R) on L2 (Rn ). Let A ∈ Sp(n, R). Then τA (v, s) = (Av, s) defines an automorphism τA of the Heisenberg group H(n). Thus, composing the Schr¨ odinger representation rS with τA , we obtain an irreducible unitary representation rA S = rS ◦ τA of H(n). Obviously, −is rA f S (0, s)f = e

for all s ∈ R and f ∈ L2 (Rn ). Therefore, by the Stone–von Neumann theorem, there exists an operator U (A) ∈ U(L2 (Rn )) such that U (A) ◦ rS (v, s) = rA S (v, s) ◦ U (A)

(1.3.3)

for all (v, s) ∈ H(n). Due to Proposition 1.2.3, U (A) is determined up to a scalar factor of modulus one. Since A ∈ Sp(n, R) → τA ∈ Aut(H(n)) is a group homomorphism, it follows that the operators U (A) give rise to a projective unitary representation of Sp(n, R), which means that U (AB) = c(A, B) U (A) ◦ U (B) for any A, B ∈ Sp(n, R) with some function c : Sp(n, R) × Sp(n, R) → S 1 . Let us compute U (A) in the cases that A is one of the generators of Sp(n, R) according to Proposition 1.1.1. Let F : L2 (Rn ) → L2 (Rn ) be the Fourier transform, i.e. that unitary operator on L2 (Rn ) which is given by −n/2 (Ff )(x) = (2π) e−ix,y f (y) dy Rn

for any f in the Schwartz space S(R ) of rapidly decreasing smooth functions on Rn . n

Lemma 1.3.2 Set U (A) = F −1 for A = J0 . Then Equation (1.3.3) is fulfilled. Proof. Let (v, s) ∈ H(n) and f ∈ S(Rn ). Substituting y = z − v , we conclude ((rS (v, s) ◦ F)f )(x) −n/2 −i(s+v ,x−v ,v /2)

= (2π)

e

e−ix−v

,y

f (y) dy

Rn

e−ix−v ,z−v f (z − v ) dz = (2π)−n/2 e−i(s+v ,x−v ,v /2) Rn −n/2 e−ix,z e−i(s−v ,z+v ,v /2) f (z − v ) dz = (2π) Rn

= ((F ◦ rS (J0 v, s))f )(x) .

1.3 Metaplectic Representation

9

Thus rS (v, s) ◦ F = F ◦ rS (J0 v, s) ,

which gives the assertion. Lemma 1.3.3 Let

A=

a 0 −1 0 aT

∈ D(n) .

Choose λ(a) ∈ C with λ(a)2 = det(a) and define U (A) : L2 (Rn ) → L2 (Rn ) by (U (A)f )(x) = λ(a)f aT x . Then U (A) is a unitary operator and satisfies Equation (1.3.3). Proof. The unitarity of U (A) follows from the transformation formula of Lebesgue integration. It remains to verify that (U (A) ◦ rS (v, s))f = (rS (Av, s) ◦ U (A))f for (v, s) ∈ H(n) and f ∈ L2 (Rn ). This can be directly calculated, using −1 T . v v , a x = av , x and v , v = av , aT

Lemma 1.3.4 Let A=

Ib 0I

∈ N (n)

and define U (A) ∈ U(L2 (Rn )) by (U (A)f )(x) = e−ibx,x/2 f (x) . Then U (A) satisfies Equation (1.3.3). Proof. One immediately sees that the statement is equivalent to 1 1 bx, x + v , x − v , v 2 2 1 1 = v + bv , x − v + bv , v + b(x − v ), x − v 2 2 for x, v , v ∈ Rn . The last equation reduces to bx, v = bv , x , which is true, since b ∈ S(n).

It was proved by Weil in [50] (see also [16, 32, 49]) that the projective unitary representation A ∈ Sp(n, R) → U (A) ∈ U(L2 (Rn )) described above lifts to a unitary representation of the metaplectic group Mp(n, R). More precisely, we have

10


Proposition 1.3.5 There exists a unique unitary representation m : Mp(n, R) → U(L2 (Rn )) which satisfies m(q) ◦ rS (v, s) = rS (ρ(q)v, s) ◦ m(q) for all q ∈ Mp(n, R) and (v, s) ∈ H(n).

The unitary representation m is known as Segal–Shale–Weil representation or metaplectic representation. As we will see, it will take the role of the spinor representation of the orthogonal group in the Riemannian case. So, when one speaks of symplectic spinors, one means the elements of the representation space L2 (Rn ). We point out some properties of m (cf. e.g. [42] and the references therein). Proposition 1.3.6 (1) The metaplectic representation m is faithful. It decomposes into the sum of two inequivalent irreducible unitary representations, which are the restrictions of m to the spaces of even and odd functions, respectively. (2) The space of smooth vectors of m is precisely the Schwartz space S(Rn ). In particular, S(Rn ) is m-invariant. It turns out that, for q ∈ Mp(n, R) with ρ(q) ∈ D(n) ∪ N (n) ∪ {J0 }, the operator m(q) coincides, up to sign, with the operator U (ρ(q)) given in the above lemmas. Thus we have the following case-by-case realization of the metaplectic representation (cf. e.g. [12, 32] and Lemma 1.5.4). There, we consider N (n) as a subgroup of Mp(n, R), which may be done, since N (n) is simply connected. Proposition 1.3.7 Let q ∈ Mp(n, R) and f ∈ L2 (Rn ). (1) If ρ(q) = J0 , then

m(q)f = ±einπ/4 F −1 f .

(2) If

ρ(q) =

then

a 0 T −1 0 a

∈ D(n) ,

(m(q)f )(x) = λ(a)f aT x ,

where λ(a) is a suitably chosen root of det(a). (3) If

q=

then

Ib 0I

∈ N (n) ,

(m(q)f )(x) = e−ibx,x/2 f (x) .

1.4 Symplectic Clifford Multiplication

11

1.4 Symplectic Clifford Multiplication In this section, we will define Clifford multiplication in the symplectic context and derive a formula expressing the differential of the metaplectic representation in terms of this multiplication. We consider the position and momentum operators Qj and Pj , j = 1, . . . , n, defined by ∂f (Qj f )(x) = ixj f (x) and Pj f = ∂xj for smooth functions f : Rn → C. Generally, we will regard them as continuous operators on the Schwartz space S(Rn ). Set σ(aj ) = Qj

and σ(bj ) = Pj

for the standard basis {a1 , . . . , an , b1 , . . . , bn } of R2n . By linear continuation, this defines, for any v ∈ R2n , an operator σ(v) : S(Rn ) → S(Rn ). As one easily sees, (σ(v)f )(x) = iv , x f (x) + v , grad(f )(x) for all f ∈ S(Rn ). Lemma 1.4.1 (1) For any v, w ∈ R2n , σ(v) ◦ σ(w) − σ(w) ◦ σ(v) = −iω0 (v, w) . (2) For any v ∈ R2n and any f1 , f2 ∈ S(Rn ), σ(v)f1 , f2 = −f1 , σ(v)f2 , where , denotes the inner product of L2 (Rn ). (3) For any v ∈ R2n ,

F ◦ σ(J0 v) = σ(v) ◦ F .

Proof. Assertion (1) follows from Qj ◦ Qk = Qk ◦ Qj , Pj ◦ Pk = Pk ◦ Pj , Pk ◦ Qj = iδjk + Qj ◦ Pk . Obviously, Qj f1 , f2 = −f1 , Qj f2 and, by partial integration, Pj f1 , f2 = −f1 , Pj f2 .

12


This implies (2). Assertion (3) is a reformulation of the well known relations Qj ◦ F = F ◦ Pj

and Pj ◦ F = −F ◦ Qj .

(1.4.1)

In view of Lemma 1.4.1(1), the map σ : R2n → End(S(Rn )) extends to a linear map defined on the whole symplectic Clifford algebra Cl(n) by setting σ(1)f = if for f ∈ S(Rn ) and σ(v1 · v2 · · · · · vm ) = σ(v1 ) ◦ σ(v2 ) ◦ · · · ◦ σ(vm ) for v1 , . . . , vm ∈ R2n . Lemma 1.4.2 (1) For any v ∈ mp(n, R) and any v ∈ R2n , σ(v) ◦ σ(v) − σ(v) ◦ σ(v) = iσ(ρ∗ (v)v) . (2) For any v ∈ mp(n, R) and any f1 , f2 ∈ S(Rn ), σ(v)f1 , f2 = f1 , σ(v)f2 . (3) The kernel of the homomorphism σ : Cl(n) → End(S(Rn )) is trivial. Proof. By Lemma 1.4.1(1), σ(v1 ) ◦ σ(v2 ) ◦ σ(v) = σ(v1 ) ◦ σ(v) ◦ σ(v2 ) − iω0 (v2 , v)σ(v1 ) = σ(v) ◦ σ(v1 ) ◦ σ(v2 ) − iω0 (v1 , v)σ(v2 ) − iω0 (v2 , v)σ(v1 ) for v1 , v2 , v ∈ R2n . It follows that σ(v1 · v2 + v2 · v1 ) ◦ σ(v) − σ(v) ◦ σ(v1 · v2 + v2 · v1 ) = 2iσ(ω0 (v, v1 )v2 + ω0 (v, v2 )v1 ) = iσ(ρ∗ (v1 · v2 + v2 · v1 )v) , where we have used Equation (1.3.2). This gives (1). Assertion (2) is obtained by applying Lemma 1.4.1(2) twice. To prove (3), let v ∈ Cl(n) and suppose that σ(v)f = 0 for all f ∈ S(Rn ). Expressing v in the basis of Cl(n) given after Definition 1.1.5 and applying σ(v) to functions f ∈ S(Rn ) which are polynomial near 0, one easily gets v = 0. We remark that, in contrast to Lemma 1.4.2(3), the Clifford multiplication in Riemannian spin geometry generally has a non-trivial kernel.


13

For convenience, in the following, we will usually write v · f instead of σ(v)f with the convention v · w · f = v · (w · f ) for v, w ∈ Cl(n). Definition 1.4.3 The symplectic Clifford multiplication is the map µ0 : R2n ⊗ S(Rn ) → S(Rn ) defined by µ0 (v ⊗ f ) = v · f . Lemma 1.4.4 The Clifford multiplication is Mp(n, R)-equivariant. That is, for any q ∈ Mp(n, R), v ∈ R2n , and f ∈ S(Rn ), µ0 (ρ(q)v ⊗ m(q)f ) = m(q)µ0 (v ⊗ f ) . Proof. We apply Proposition 1.3.7 and Lemma 1.4.1(3). First we assume that ρ(q) = J0 . Then µ0 (ρ(q)v ⊗ m(q)f ) = ±einπ/4 σ(J0 v) ◦ F −1 f = ±einπ/4 F −1 ◦ σ(v) f = m(q)µ0 (v ⊗ f ) . Now let

ρ(q) =

In this case,

a 0 −1 0 aT

.

grad(m(q)f )(x) = λ(a)a grad(f ) aT x .

It follows that µ0 (ρ(q)v ⊗ m(q)f )(x)

−1 = iav , x (m(q)f )(x) + aT v , grad(m(q)f )(x) = i v , aT x (m(q)f )(x) + v , a−1 grad(m(q)f )(x) = λ(a) i v , aT x f aT x + v , grad(f ) aT x = (m(q)µ0 (v ⊗ f ))(x) .

Finally, if

q=

then

Ib 0I

∈ N (n) ,

grad(m(q)f )(x) = e−ibx,x/2 (grad(f )(x) − if (x)bx)

14


and hence µ0 (ρ(q)v ⊗ m(q)f )(x) = iv + bv , x (m(q)f )(x) + v , grad(m(q)f )(x) = e−ibx,x/2 (iv , x (f )(x) + v , grad(f )(x) ) = (m(q)µ0 (v ⊗ f ))(x) . Since the considered elements of Mp(n, R) generate the whole group, the lemma is proved. We now turn to the calculation of the differential m∗ : mp(n, R) → End(S(Rn )) of the metaplectic representation m. Proposition 1.4.5 For any v ∈ mp(n, R) and f ∈ S(Rn ), m∗ (v)f = −iv · f . Proof. We have to show that m∗ (v) = −iσ(v)

(1.4.2)

for all v ∈ mp(n, R). For this, we may assume that v = v · w + w · v for v, w ∈ R2n such that (a) v = w = 0, (b) v = w = 0, or (c) v = w = 0. Suppose (a). Then, by Equations (1.1.2) and (1.3.2), ρ(exp(tv)) = exp(tρ∗ (v)) = exp(2t v w) exp(2t w ⊗ v ) 0 = 0 exp(−2t v ⊗ w ) for all t ∈ R. In particular, by Equation (1.1.3), ρ(exp(tv)) ∈ D(n). Hence, according to Proposition 1.3.7, (m(exp(tv))f )(x) = det(exp(2t w ⊗ v ))f (exp(2t v ⊗ w )x) for f ∈ S(Rn ). We conclude that


15

d (m(exp(tv))f )(x) (m∗ (v)f )(x) = dt t=0 d = Tr(w ⊗ v )f (x) + f (exp(2t v ⊗ w )x) dt t=0 = v , w f (x) + 2v , x w , grad(f )(x) .

On the other hand, using Lemma 1.4.1(1), we obtain (σ(v)f )(x) = ((σ(v) ◦ σ(w) + σ(w) ◦ σ(v))f )(x) = iω0 (v, w)f (x) + 2((σ(v) ◦ σ(w))f )(x) = iv , w f (x) + 2iv , x w , grad(f )(x) . This implies Equation (1.4.2). Suppose (b). As above, we see 1 −2t(v ⊗ w + w ⊗ v ) ρ(exp(tv)) = ∈ N (n) . 0 1 Hence

(m(exp(tv))f )(x) = e2itv ,xw ,x f (x) . It follows that (m∗ (v)f )(x) = 2iv , x w , x f (x) = −2i((σ(v) ◦ σ(w))f )(x) = −i(σ(v)f )(x) as claimed. Suppose (c). In this case ρ(exp(tv)) = Since

1 0 2t(v ⊗ w + w ⊗ v ) 1

1 0 2t(v ⊗ w + w ⊗ v ) 1

=

J0−1

.

1 −2t(v ⊗ w + w ⊗ v ) 0 1

J0

= J0−1 ρ(exp(t (J0 v · J0 w + J0 w · J0 v)))J0 , we get m(exp(tv)) = F ◦ m(exp(t (J0 v · J0 w + J0 w · J0 v))) ◦ F −1 . By means of case (b) and Lemma 1.4.1(3), this yields

16


m∗ (v) = F ◦ m∗ (J0 v · J0 w + J0 w · J0 v) ◦ F −1 = −iF ◦ (σ(J0 v) ◦ σ(J0 w) + σ(J0 w) ◦ σ(J0 v)) ◦ F −1 = −i(σ(v) ◦ σ(w) + σ(w) ◦ σ(v)) = −iσ(v) ,

so we are done.

We note that σ : Cl(n) → End(S(Rn )) is not an algebra homomorphism. However, the previous proposition implies Corollary 1.4.6 For any v, w ∈ mp(n, R), σ(v) ◦ σ(w) − σ(w) ◦ σ(v) = iσ([v, w]) . Proof. This is a consequence of Equation (1.4.2) and the fact that m∗ is a Lie algebra homomorphism. Alternatively, Corollary 1.4.6 can be obtained by a direct calculation, using Equations (1.1.6) and (1.3.1) and Lemma 1.4.2(1).

1.5 Hermite Functions Let U(n) ⊂ Mp(n, R) denote the double cover of U(n) realized as the inverse image of U(n) ⊂ Sp(n, R) under the covering map ρ : Mp(n, R) → Sp(n, R). In this section, we study the restriction u : U(n) → U(L2 (Rn )) of the metaplectic representation m to U(n). As we will see, u is strongly related to the harmonic oscillator and the Hermite functions. We start with describing the Lie algebra u(n) of U(n). First observe that the Lie algebra u(n) of U(n) ⊂ Sp(n, R) consists of those X ∈ sp(n, R) that satisfy Xv, w + v, Xw = 0 or, equivalently, J0 XJ0 = −X . Hence u(n) is spanned by the endomorphisms X − J0 XJ0 for X ∈ sp(n, R). Obviously, J0 ∈ u(n). Moreover, since J0 (X − J0 XJ0 ) = J0 X + XJ0 = (X − J0 XJ0 )J0 , J0 lies in the center of u(n). By means of

1.5 Hermite Functions

17

ω0 (v, J0 v1 )J0 v2 = −ω0 (J0 v, v1 )J0 v2 = −J0 (ω0 (J0 v, v1 )v2 ) , we see that, with the identification sp(n, R) ≡ S 2 (R2n ), the Lie algebra u(n) u(n) = ρ−1 is spanned by v1 v2 + J0 v1 J0 v2 for v1 , v2 ∈ R2n . Hence ∗ (u(n)) is spanned by the elements v1 · v2 + v2 · v1 + J0 v1 · J0 v2 + J0 v2 · J0 v1 . In particular, aj · ak + bj · bk for 1 ≤ j ≤ k ≤ n and aj · bk − ak · bj for 1 ≤ j < k ≤ n form a basis of u(n). Let 0 = ρ−1 ∗ (J0 ). Then, by Lemma 1.3.1, 1 (aj · aj + bj · bj ) . 2 j=1 n

0 =

(1.5.1)

u(n). So Furthermore, 0 is in the center of [0 , v] = 0

(1.5.2)

for all v ∈ u(n). Now recall that the Hamilton operator of the n-dimensional harmonic oscillator is the operator H0 : S(Rn ) → S(Rn ) defined by

n 1 ∂2f 2 (x) − xj f (x) . (H0 f )(x) = 2 j=1 ∂x2j Notice that, by the definition of σ and Equation (1.5.1), H0 = σ(0 ) . Equivalently,

1 (aj · aj + bj · bj ) · f . 2 j=1

(1.5.3)

n

H0 f =

(1.5.4)

Let N0 denote the set of non-negative integers and let hα ∈ S(Rn ) for multiindices α = (α1 , . . . , αn ) ∈ Nn0 be the Hermite functions on Rn . These functions are defined by hα (x) = hα1 (x1 ) · · · hαn (xn ) , where hl for l ∈ N0 are the classical Hermite functions on R, i.e. 2

hl (t) = et

/2

dl −t2 e . dtl

As is well known (cf. [16]), the Hermite functions form a complete orthogonal system in L2 (Rn ) of eigenfunctions of H0 . In particular,

18


n H0 hα = − |α| + hα , 2

(1.5.5)

where |α| = α1 + · · · + αn . Let Wl denote the eigenspace of H0 with eigenvalue −(l+n/2). An elementary combinatorial computation shows that the complex dimension of Wl is n+l−1 dimC Wl = . (1.5.6) l Proposition 1.5.1 For all q ∈ U(n), H0 ◦ u(q) = u(q) ◦ H0 on S(Rn ). Proof. Equations (1.4.2), (1.5.2), and (1.5.3) and Corollary 1.4.6 yield that, for all v ∈ u(n), H0 ◦ u∗ (v) − u∗ (v) ◦ H0 = −i(σ(0 ) ◦ σ(v) − σ(v) ◦ σ(0 )) = σ([0 , v]) =0. Since U(n) is connected, this implies the assertion.

Corollary 1.5.2 The spaces Wl , l ∈ N0 , form an orthogonal decomposition of L2 (Rn ) into u-invariant subspaces. Proof. The u-invariance of each Wl follows from Proposition 1.5.1. The rest is clear. We denote the restriction of the unitary representation u of U(n) to the subspace Wl by ul . For the sake of completeness, we note that each ul is irreducible (cf. [5]). We conclude this section with the following two lemmas. Lemma 1.5.3 The Fourier transform F commutes with H0 , F ◦ H 0 = H0 ◦ F . Moreover, if f ∈ Wl , then

Ff = i−l f .

1.5 Hermite Functions

19

Proof. The first part follows from 1 2 Pj + Q2j 2 j=1 n

H0 =

and Equation (1.4.1). For the second part, it suffices to show that Fhα = (−i)|α| hα

(1.5.7)

for all α ∈ Nn0 . First recall that this equation is valid in case α = 0. Then check that, for j = 1, . . . , n and any α ∈ Nn0 , (iQj + Pj )hα = hα+ε(j) , where ε(j) denotes that multi-index whose only non-vanishing entry is a 1 at the j-th position. Furthermore, again by Equation (1.4.1), F ◦ (iQj + Pj ) = −i(iQj + Pj ) ◦ F . By induction, this proves the desired relation. Lemma 1.5.4 It is

π u exp = einπ/4 F −1 . 0 2

Proof. By Equations (1.4.2) and (1.5.3), we have d u(exp(t0 ))f = u∗ (0 )f = −iH0 f dt t=0 for any f ∈ S(Rn ). Applying Equation (1.5.5), it follows d n u(exp(t0 ))hα hα . = i |α| + dt 2 t=0 Thus u(exp(t0 ))hα = ei(|α|+n/2)t hα for all t ∈ R. Setting t = π/2, we obtain π u exp 0 hα = einπ/4 i|α| hα . 2 Equations (1.5.7) and (1.5.8) imply the assertion.

(1.5.8)

2 Symplectic Connections

In order to define symplectic Dirac operators, we have to fix a symplectic connection on the underlying symplectic manifold. Every symplectic manifold admits symplectic connections. Unfortunately, there is no canonical condition to make a symplectic connection unique as one has in Riemannian geometry for defining the Levi–Civita connection. There are systematical investigations on (almost-)symplectic connections by Tondeur [46]. Further results, in particular on curvature properties, were established by Vaisman [47]. Gelfand, Retakh, and Shubin [22, 44] studied geometrical properties of symplectic connections. Moreover, symplectic connections play a key role in Fedosov’s construction of a canonical deformation quantization (cf. [13, 14, 15]). Recently, Bourgeois and Cahen [6, 7] have introduced a variational principle for symplectic connections to single out a preferred symplectic connection. Furthermore, Cahen, Gutt, and Rawnsley [8, 9, 10] have studied the corresponding field equation. However, these considerations were made for torsion-free connections only. Our work suggests to take into account also symplectic connections with torsion. It seems that such connections are more convenient for our purposes.

2.1 Symplectic Manifolds This section contains basic facts on symplectic manifolds. References for this are [11, 31, 39]. Let M be a manifold. Unless otherwise stated, all manifolds, all tensor fields on it, and all maps are assumed to be smooth, i.e. of class C ∞ . Moreover, we suppose that all considered manifolds are Hausdorff, second countable, connected, and without boundary.

22


Definition 2.1.1 A symplectic structure on M is a 2-form ω on M which is closed and non-degenerate. The pair (M, ω) is then called a symplectic manifold. Here, the non-degeneracy of ω means that, for every p ∈ M , the bilinear form ωp on the tangent space Tp M is non-degenerate. Hence M is necessarily of even dimension 2n. Moreover, since ω is non-degenerate if and only if the n-fold wedge product ωn = ω ∧ · · · ∧ ω is a volume form, M is orientable. Examples for symplectic structures are the constant form ω0 on R2n and the K¨ ahler form of a K¨ ahler manifold. Furthermore, for any manifold N , the cotangent bundle T ∗ N carries a canonical symplectic structure ωN , given as the exterior derivative of the canonical 1form on T ∗ N . Note that a symplectic structure ω on a compact manifold M cannot be an exact form. Indeed, if ω = dη for a 1-form η, then by Stokes’ theorem, ωn = dη ∧ ω n−1 = d η ∧ ω n−1 = 0 . M

M

M n

On the other hand, since ω is a volume form, ω n = 0 , M

which gives a contradiction. Therefore, a compact manifold which admits a symplectic structure ω has non-trivial second cohomology H 2 (M ; R). This implies that there are orientable even-dimensional manifolds without any symplectic structure. For example, there exist no symplectic structures on the sphere S 2n for n ≥ 2. Definition 2.1.2 A symplectomorphism between two symplectic manifolds (M1 , ω1 ) and (M2 , ω2 ) is a diffeomorphism F : M1 → M2 satisfying F ∗ ω2 = ω1 . (M1 , ω1 ) and (M2 , ω2 ) are then called symplectomorphic. Theorem 2.1.3 (Darboux) Every symplectic manifold (M, ω) is locally symplectomorphic to (R2n , ω0 ). In view of this theorem, the only local symplectic invariant is the dimension. Let (M, ω) be a fixed symplectic manifold. For some of our later considerations, we need an additional structure.

2.1 Symplectic Manifolds

23

Definition 2.1.4 An almost complex structure J on M , i.e. an endomorphism J : T M → T M such that J 2 = −id, is called ω-compatible if g(X, Y ) = ω(X, JY ) for X, Y ∈ Γ(T M ) defines a Riemannian metric g on M . Here, Γ(T M ) means the space of smooth sections of the tangent bundle T M , i.e. the space of all vector fields on M . Lemma 2.1.5 An ω-compatible almost complex structure J preserves the symplectic structure ω as well as the associated Riemannian metric g. That means ω(JX, JY ) = ω(X, Y ) and g(JX, JY ) = g(X, Y ) for all X, Y ∈ Γ(T M ). Proof. By the skew-symmetry of ω and the symmetry of g, ω(JX, JY ) = g(JX, Y ) = g(Y, JX) = −ω(Y, X) = ω(X, Y ) and g(JX, JY ) = −ω(JX, Y ) = ω(Y, JX) = g(Y, X) = g(X, Y ) . It is not hard to show that, for any symplectic structure ω, there exists an ω-compatible almost complex structure J. Such a structure is not unique. However, one has Proposition 2.1.6 The space of all ω-compatible almost complex structures is contractible. For p ∈ M , let Rp denote the set of all symplectic basises of Tp M , i.e. the set of all 2n-tuples (e1 , . . . , en , f 1 , . . . , f n ) of tangent vectors at p satisfying ωp (ej , ek ) = ωp (f j , f k ) = 0

and ωp (ej , f k ) = δjk

for j, k = 1, . . . , n. The fiber bundle πR : R → M whose fiber over p is Rp is called the symplectic frame bundle of (M, ω). On each Rp , the symplectic group Sp(n, R) acts simply transitively. The Darboux theorem guarantees the existence of local sections of R. Thus R is locally trivial and hence a principal Sp(n, R)-bundle. We call the local sections of R symplectic frames. We write these frames as (e1 , . . . , en , f1 , . . . , fn ) and their dual frames as (e∗1 , . . . , e∗n , f1∗ , . . . , fn∗ ). Then, of course, (2.1.1) e∗j (X) = ω(X, fj ) and fj∗ (X) = −ω(X, ej )

24


for j = 1, . . . , n and X ∈ Γ(T M ). It is clear that the ω-compatible almost complex structures are in 1 : 1 correspondence with the U(n)-reductions of the symplectic frame bundle. Given an ω-compatible almost complex structure J, we call the local sections of the U(n)-reduction unitary frames and write them as (e1 , . . . , e2n ). Obviously, such a frame is characterized by g(ej , ek ) = δjk ,

g(ej , en+k ) = 0 ,

and Jej = en+j

for j, k = 1, . . . , n, where g is again the Riemannian metric associated to J. Definition 2.1.7 Let u ∈ C ∞ (M ). The vector field Xu on M determined by du(Y ) = ω(Xu , Y ) for all Y ∈ Γ(T M ) is called the Hamiltonian vector field of the function u. The vector field Xu is also referred to as the symplectic gradient of u. Obviously, if J is an ω-compatible almost complex structure, then JXu is the gradient of u taken with respect to the associated Riemannian metric g. Definition 2.1.8 The Poisson bracket of two functions u, v ∈ C ∞ (M ) is the function {u, v} = ω(Xu , Xv ) . The Lie bracket of Hamiltonian vector fields and the Poisson bracket are related by Proposition 2.1.9 For all u, v ∈ C ∞ (M ), [Xu , Xv ] = −X{u,v} . Proof. As is known, dη(Y1 , Y2 , Y3 ) = Y1 (η(Y2 , Y3 )) − Y2 (η(Y1 , Y3 )) + Y3 (η(Y1 , Y2 )) −η([Y1 , Y2 ], Y3 ) + η([Y1 , Y3 ], Y2 ) − η([Y2 , Y3 ], Y1 ) for any 2-form η on M and all Y1 , Y2 , Y3 ∈ Γ(T M ). Thus, since ω is closed, we have 0 = Xu (ω(Xv , Y )) − Xv (ω(Xu , Y )) + Y (ω(Xu , Xv )) −ω([Xu , Xv ], Y ) + ω([Xu , Y ], Xv ) − ω([Xv , Y ], Xu ) = Xu Y (v) − Xv Y (u) + Y ({u, v}) −ω([Xu , Xv ], Y ) − [Xu , Y ](v) + [Xv , Y ](u) = Y ({u, v}) − ω([Xu , Xv ], Y ) + Y Xu (v) − Y Xv (u)

2.2 Constructions and Torsion

25

for all u, v ∈ C ∞ (M ) and Y ∈ Γ(T M ). Applying that Xu (v) = ω(Xu , Xv ) = −{u, v} , we obtain Y ({u, v}) = −ω([Xu , Xv ], Y ) , and the proposition is proved.

2.2 Constructions and Torsion Let (M, ω) be a symplectic manifold of dimension 2n. Definition 2.2.1 A connection ∇ on M is called symplectic if ∇ω = 0, i.e. if X(ω(Y, Z)) = ω(∇X Y, Z) + ω(Y, ∇X Z) for all X, Y, Z ∈ Γ(T M ). Symplectic connections on M are in 1 : 1 correspondence with connection 1forms on the symplectic frame bundle R. So any symplectic manifold admits symplectic connections. However, as already mentioned, there is no canonically distinguished symplectic connection. Lemma 2.2.2 Let ∇ be a symplectic connection and let A be a (2, 1)-tensor field on M . Then the connection ∇ given by ∇X Y = ∇X Y + A(X, Y ) is symplectic if and only if ω(A(X, Y ), Z) = ω(A(X, Z), Y ) for all X, Y, Z ∈ Γ(T M ). Proof. The symplecticity of ∇ is equivalent to ω(∇X Y − ∇X Y, Z) + ω(Y, ∇X Z − ∇X Z) = 0 , which gives the assertion.

Later on, we will need symplectic connections which are adapted to an ωcompatible almost complex structure J. The following proposition illustrates a method for constructing such connections.

26


Proposition 2.2.3 Let J be an ω-compatible almost complex structure and let ∇ be a connection on M such that (a) ∇ g = 0 or (b) ∇ is symplectic, where g is the Riemannian metric associated to J. Then the connection ∇ on M defined by 1 ∇X Y = ∇X Y + (∇X J)(JY ) 2 is symplectic and satisfies ∇J = 0

and

∇g = 0 .

Proof. We suppose (a). The proof for (b) proceeds analogously. Recall that

(∇X J)Y = ∇X (JY ) − J(∇X Y ) .

Using this and Lemma 2.1.5, we conclude 1 ω(∇X Y, Z) = ω(∇X Y, Z) + ω((∇X J)(JY ), Z) 2 1 = −g(∇X Y, JZ) − g((∇X J)(JY ), JZ) 2 1 1 = g(∇X (JY ), Z) − g(∇X Y, JZ) . 2 2 Hence ω(Y, ∇X Z) = −ω(∇X Z, Y ) 1 1 = g(JY, ∇X Z) − g(Y, ∇X (JZ)) . 2 2 Since ∇ g = 0 is equivalent to X(g(Y, Z)) = g(∇X Y, Z) + g(Y, ∇X Z) , it follows that 1 1 X(g(JY, Z)) − X(g(Y, JZ)) 2 2 = −X(g(Y, JZ))

ω(∇X Y, Z) + ω(Y, ∇X Z) =

= X(ω(Y, Z)) , which proves that ∇ is symplectic. From

(2.2.1)

2.2 Constructions and Torsion

27

1 ∇X (JY ) = ∇X (JY ) − (∇X J)Y 2 1 1 = ∇X (JY ) + J(∇X Y ) 2 2 and 1 J(∇X Y ) = J(∇X Y ) + J((∇X J)(JY )) 2 1 1 = J(∇X Y ) + ∇X (JY ) , 2 2 we get ∇X (JY ) − J(∇X Y ) = 0 , which shows ∇J = 0. The identity ∇g = 0 now follows from (∇X g)(Y, Z) = (∇X ω)(Y, JZ) + ω(Y, (∇X J)Z) . We recall that a connection ∇ which satisfies Equation (2.2.1) is called Hermitian. Given a connection ∇ on M , let T denote its torsion, which is the (2, 0)-tensor field on M defined by T(X, Y ) = ∇X Y − ∇Y X − [X, Y ] . ∇ is said to be torsion-free if T = 0. We assign to T a vector field as follows. Definition 2.2.4 The torsion vector field T of ∇ is the vector field on M defined by n T(ej , fj ) . T= j=1

Here and subsequently, (e1 , . . . , en , f1 , . . . , fn ) is an arbitrary symplectic frame. One easily checks that, up to the factor 1/2, the vector field T arises from a contraction of T with respect to the symplectic structure ω and hence is well defined. For a vector field X on M , let div(X) denote the divergence of X with respect to a given connection ∇ on M , i.e. the function on M defined by div(X) = Tr(Y → ∇Y X) . Lemma 2.2.5 For any X ∈ Γ(T M ), div(X) =

n ω ∇ej X, fj − ω ∇fj X, ej . j=1

28


Proof. One gets the assertion from div(X) =

n

e∗j ∇ej X + fj∗ ∇fj X

j=1

and Equation (2.1.1).

In the next proposition, we use the following notations. For X ∈ Γ(T M ), let ηX be the 1-form on M given by ηX (Y ) = ω(X, Y ) , and let dM denote the volume element of (M, ω), i.e. dM =

1 n ω . n!

Proposition 2.2.6 Let ∇ be a symplectic connection on (M, ω). Then, for any X ∈ Γ(T M ), d ηX ∧ ω n−1 = (n − 1)! (div(X) + ω(X, T)) dM . Proof. By dM (e1 , f1 , . . . , en , fn ) = 1 , we have to show that d ηX ∧ ω n−1 (e1 , f1 , . . . , en , fn ) = (n − 1)! (div(X) + ω(X, T)) . Since ω is closed,

d ηX ∧ ω n−1 = dηX ∧ ω n−1 .

Consequently,

d ηX ∧ ω

n−1

(e1 , f1 , . . . , en , fn ) = (n − 1)!

n

dηX (ej , fj ) .

j=1

On the other hand, from the symplecticity of ∇ and the well known formula dηX (Y, Z) = Y (ηX (Z)) − Z(ηX (Y )) − ηX ([Y, Z]) , we derive dηX (Y, Z) = Y (ω(X, Z)) − Z(ω(X, Y )) − ω(X, [Y, Z]) = ω(∇Y X, Z) − ω(∇Z X, Y ) +ω(X, ∇Y Z) − ω(X, ∇Z Y ) − ω(X, [Y, Z]) = ω(∇Y X, Z) − ω(∇Z X, Y ) + ω(X, T(Y, Z)) .

2.3 Symplectic Curvature and Ricci Tensors

29

Using Lemma 2.2.5, it follows that n j=1

  n n ω ∇ej X, fj − ω ∇fj X, ej + ω X, dηX (ej , fj ) = T(ej , fj ) j=1

j=1

= div(X) + ω(X, T) . This proves the proposition.

Corollary 2.2.7 Let ∇ be a symplectic connection on (M, ω). Then, for any compactly supported vector field X on M , (div(X) + ω(X, T)) dM = 0 . M

Proof. This follows from Proposition 2.2.6 by Stokes’ theorem.

2.3 Symplectic Curvature and Ricci Tensors Let again (M, ω) be a 2n-dimensional symplectic manifold and let ∇ be a connection on M. We denote the curvature of ∇ by R. Here, we use the sign convention that R(X, Y )Z = ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z . As already mentioned, we do not want restrict our considerations on symplectic Dirac operators to torsion-free connections. For that reason, we recall the following general relation between curvature and torsion (cf. [35]). Lemma 2.3.1 For any X, Y, Z ∈ Γ(T M ), R(X, Y )Z + R(Y, Z)X + R(Z, X)Y = (∇X T)(Y, Z) + (∇Y T)(Z, X) + (∇Z T)(X, Y ) +T(T(X, Y ), Z) + T(T(Y, Z), X) + T(T(Z, X), Y ) . Of course, if ∇ is torsion-free, then R satisfies the Bianchi identity R(X, Y )Z + R(Y, Z)X + R(Z, X)Y = 0 . Definition 2.3.2 The symplectic curvature tensor S of ∇ is the (4, 0)-tensor field on M defined by S(X1 , X2 , X3 , X4 ) = ω(R(X1 , X2 )X3 , X4 ) for X1 , X2 , X3 , X4 ∈ Γ(T M ).

30


The symplectic curvature tensor has the following symmetries. Proposition 2.3.3 (1) For any connection ∇ on M , S(X1 , X2 , X3 , X4 ) = −S(X2 , X1 , X3 , X4 ) . (2) If ∇ is symplectic, then S(X1 , X2 , X3 , X4 ) = S(X1 , X2 , X4 , X3 ) (3) If ∇ is symplectic and torsion-free, then S(X1 , X2 , X3 , X4 ) + S(X2 , X3 , X4 , X1 ) +S(X3 , X4 , X1 , X2 ) + S(X4 , X1 , X2 , X3 ) = 0 . Proof. The first assertion is obvious. Let ∇ be symplectic. Then X1 X2 (ω(X3 , X4 )) − X2 X1 (ω(X3 , X4 )) − [X1 , X2 ](ω(X3 , X4 )) = X1 (ω(∇X2 X3 , X4 ) + ω(X3 , ∇X2 X4 )) −X2 (ω(∇X1 X3 , X4 ) + ω(X3 , ∇X1 X4 )) −ω(∇[X1 ,X2 ] X3 , X4 ) − ω(X3 , ∇[X1 ,X2 ] X4 ) = ω(∇X1 ∇X2 X3 , X4 ) + ω(X3 , ∇X1 ∇X2 X4 ) −ω(∇X2 ∇X1 X3 , X4 ) − ω(X3 , ∇X2 ∇X1 X4 ) −ω(∇[X1 ,X2 ] X3 , X4 ) − ω(X3 , ∇[X1 ,X2 ] X4 ) = ω(R(X1 , X2 )X3 , X4 ) + ω(X3 , R(X1 , X2 )X4 ) = S(X1 , X2 , X3 , X4 ) − S(X1 , X2 , X4 , X3 ) . Since the first expression of the equation vanishes, this gives (2). Now we additionally assume that ∇ is torsion-free. Then, by the Bianchi identity, S(X1 , X2 , X3 , X4 ) + S(X2 , X3 , X1 , X4 ) + S(X3 , X1 , X2 , X4 ) = 0 , S(X1 , X3 , X4 , X2 ) + S(X3 , X4 , X1 , X2 ) + S(X4 , X1 , X3 , X2 ) = 0 . Thus, applying (2), we get S(X1 , X2 , X3 , X4 ) + S(X2 , X3 , X4 , X1 ) = −S(X3 , X1 , X2 , X4 ) , S(X3 , X4 , X1 , X2 ) + S(X4 , X1 , X2 , X3 ) = −S(X1 , X3 , X2 , X4 ) . Together with (1), this yields (3). Moreover, we have


31

Proposition 2.3.4 Let J be an ω-compatible almost complex structure and let ∇ be a symplectic connection on M that satisfies ∇J = 0. Then S(X1 , X2 , JX3 , JX4 ) = S(X1 , X2 , X3 , X4 ) and S(X1 , X2 , JX3 , X4 ) = −S(X1 , X2 , JX4 , X3 ) . Proof. Since ∇J = 0, R(X, Y ) ◦ J = J ◦ R(X, Y ) for any X, Y ∈ Γ(T M ). This implies S(X1 , X2 , JX3 , JX4 ) = ω(R(X1 , X2 )(JX3 ), JX4 ) = ω(J(R(X1 , X2 )X3 ), JX4 ) = ω(R(X1 , X2 )X3 , X4 ) = S(X1 , X2 , X3 , X4 ) and, further, by Proposition 2.3.3(2), S(X1 , X2 , JX3 , X4 ) = −S(X1 , X2 , X3 , JX4 ) = −S(X1 , X2 , JX4 , X3 ) . Now we are going to introduce the notion of a symplectic Ricci tensor. Let us first recall the definition of the usual Ricci tensor. Definition 2.3.5 Let ric be the (2, 0)-tensor field on M defined by ric(X, Y ) = Tr(Z → R(Z, X)Y ) for X, Y ∈ Γ(T M ). Expressing ric in a symplectic frame (e1 , . . . , en , f1 , . . . , fn ), one gets Lemma 2.3.6 For any X, Y ∈ Γ(T M ), ric(X, Y ) =

n

(S(ej , X, Y, fj ) − S(fj , X, Y, ej )) .

j=1

Proof. By means of Equation (2.1.1), one sees

32


ric(X, Y ) =

n ∗ ej (R(ej , X)Y ) + fj∗ (R(fj , X)Y ) j=1

=

=

n j=1 n

(ω(R(ej , X)Y, fj ) − ω(R(fj , X)Y, ej )) (S(ej , X, Y, fj ) − S(fj , X, Y, ej )) .

j=1

Thus the Ricci tensor ric is obtained by contracting the symplectic curvature tensor S with respect to the symplectic form ω. We use another contraction of S to define the symplectic Ricci tensor sric. As we will see later (cf. Proposition 5.3.6 and Definition 2.3.15), this notion of a Ricci tensor is more suitable for studying symplectic Dirac operators. Definition 2.3.7 We define the symplectic Ricci tensor sric to be the (2, 0)tensor field on M given by sric(X, Y ) =

n

S(ej , fj , X, Y ) .

j=1

For completeness, we note Lemma 2.3.8 If ∇ is symplectic, then for any X, Y ∈ Γ(T M ), the endomorphism Z ∈ Γ(T M ) → R(X, Y )Z ∈ Γ(T M ) is trace-free. Proof. As in the proof of Lemma 2.3.6, one sees Tr(Z → R(X, Y )Z) =

n

(S(X, Y, fj , ej ) − S(X, Y, ej , fj )) .

j=1

Hence the assertion follows from Proposition 2.3.3(2).

Proposition 2.3.9 For any symplectic connection ∇ on M , the symplectic Ricci tensor sric is symmetric. Proof. This also follows from Proposition 2.3.3(2).

We point out that, in general, the Ricci tensor ric and the symplectic Ricci tensor sric differ. In particular, also for symplectic connections, it can happen that ric is not symmetric. However, we have


33

Proposition 2.3.10 If ∇ is symplectic and torsion-free, then ric = sric . Proof. Using Proposition 2.3.3, Lemma 2.3.6, and the Bianchi identity, one concludes ric(X, Y ) = −

n

(S(X, ej , fj , Y ) + S(fj , X, ej , Y ))

j=1

=

n

S(ej , fj , X, Y )

j=1

= sric(X, Y ) . Combining the previous two propositions, we obtain (cf. also [47]) Corollary 2.3.11 If ∇ is symplectic and torsion-free, then the Ricci tensor ric is symmetric. The following definition (cf. [47]) generalizes the notion of constant holomorphic sectional curvature on a K¨ ahler manifold. Definition 2.3.12 A symplectic connection ∇ on (M, ω) is called reducible if there exists a symmetric (2, 0)-tensor field b such that S(X1 , X2 , X3 , X4 ) = ω(X1 , X3 )b(X2 , X4 ) + ω(X1 , X4 )b(X2 , X3 ) −ω(X2 , X3 )b(X1 , X4 ) − ω(X2 , X4 )b(X1 , X3 ) +2ω(X1 , X2 )b(X3 , X4 ) for all X1 , X2 , X3 , X4 ∈ Γ(T M ). Proposition 2.3.13 Let ∇ be a reducible symplectic connection on M . Then sric = 2(n + 1)b . Proof. For any X, Y ∈ Γ(T M ), one concludes sric(X, Y ) =

=

n j=1 n j=1

S(ej , fj , X, Y ) (ω(ej , X)b(fj , Y ) + ω(ej , Y )b(fj , X)

34


−ω(fj , X)b(ej , Y ) − ω(fj , Y )b(ej , X) + 2ω(ej , fj )b(X, Y ))   n = b  (ω(ej , X)fj − ω(fj , X)ej ), Y  j=1

 +b 

n

 (ω(ej , Y )fj − ω(fj , Y )ej ), X  + 2nb(X, Y )

j=1

= 2(n + 1)b(X, Y ) . Remark 2.3.14 In dimension 2, i.e. for n = 1, the symplectic curvature tensor S of a connection ∇ writes as S(X1 , X2 , X3 , X4 ) = ω(X1 , X2 )sric(X3 , X4 ) . Indeed, if (e, f) is a symplectic frame, then S(e, f, X, Y ) = sric(X, Y ) = ω(e, f)sric(X, Y ) . Moreover, it is readily verified that, in this dimension, every symplectic connection is reducible. Since ω is skew-symmetric and, for symplectic connections, sric is symmetric, there is no way to intrinsically define symplectic scalar curvature. Therefore, to introduce a symplectic counterpart of scalar curvature, we need an auxiliary structure, which in our case is provided by an ω-compatible almost complex structure J. Definition 2.3.15 Let ∇ be any symplectic connection on (M, ω). We define the symplectic scalar curvature of ∇ with respect to J to be the function R on M given by 2n sric(ej , ej ) , R= j=1

where (e1 , . . . , e2n ) is an arbitrary unitary frame. Keeping in mind the definition of sric, the symplectic scalar curvature writes as 2n 1 R= S(ej , Jej , ek , ek ) . (2.3.1) 2 j,k=1

Of course, if (M, ω, J) is K¨ ahler and ∇ is the Levi–Civita connection, then R equals the Riemannian scalar curvature. In general, however, R turns out to be something complete different.

3 Symplectic Spinor Fields

This chapter is devoted to the symplectic spinor bundle, which is the Hilbert space bundle associated to a metaplectic structure via the metaplectic representation. Since the fibers of this bundle are infinite-dimensional, several complications in handling differential operators acting on spinor fields occur. To circumvent some of these difficulties, we should like to have a suitable decomposition of the symplectic spinor bundle into a sequence of finite rank subbundles. Actually, such a decomposition is given by the irreducible components of the metaplectic representation restricted to the double cover of the unitary group.

3.1 Metaplectic Structures Metaplectic structures are the symplectic analogs of spin structures on orientable Riemannian manifolds. Definition 3.1.1 A metaplectic structure on a symplectic manifold (M, ω) is an equivariant lift of the symplectic frame bundle πR : R → M with respect to the double covering ρ : Mp(n, R) → Sp(n, R). That means, a metaplectic structure on (M, ω) is a principal Mp(n, R)-bundle πP : P → M together with a map FP : P → R such that πR ◦ FP = πP and FP (pq) = FP (p)ρ(q) for all p ∈ P and q ∈ Mp(n, R). Then, of course, P is connected and FP is a double covering, too.

36


The topological obstruction to the existence of metaplectic structures is precisely the same as in Riemannian spin geometry (cf. [3, 17] and the references therein). Proposition 3.1.2 A symplectic manifold (M, ω) admits a metaplectic structure if and only if the second Stiefel–Withney class w2 (M ) ∈ H 2 (M ; Z2 ) of M vanishes. If this is the case, the isomorphy classes of metaplectic structures on (M, ω) are classified by the first cohomology group H 1 (M ; Z2 ). In particular, because of H 1 (M ; Z2 ) ∼ = Hom(π1 (M ), Z2 ), a simply connected symplectic manifold possesses, up to isomorphism, at most one metaplectic structure. Definition 3.1.3 The first Chern class c1 (M, ω) of a symplectic manifold (M, ω) is defined to be the first Chern class of the complex vector bundle (T M, J) for any ω-compatible almost complex structure J. In view of Proposition 2.1.6, the class c1 (M, ω) ∈ H 2 (M ; Z) does not depend on the choice of the ω-compatible almost complex structure J. The obstruction to the existence of metaplectic structures can now be formulated as a condition on c1 (M, ω) (cf. [37]). Proposition 3.1.4 A symplectic manifold (M, ω) admits a metaplectic structure if and only if the first Chern class c1 (M, ω) is even, i.e. if there exists a class a ∈ H 2 (M ; Z) such that c1 (M, ω) = 2a. Example 3.1.5 The symplectic frame bundle R of (R2n , ω0 ) is trivial. Thus the unique isomorphy class of metaplectic structures on (R2n , ω0 ) is that of the trivial principal Mp(n, R)-bundle over R2n . Example 3.1.6 Let N be any manifold and let its cotangent bundle T ∗ N be endowed with the canonical symplectic structure ωN . If b ∈ H 1 (T ∗ N ; Z2 ) is the pull back of the first Stiefel–Whitney class w1 (N ) ∈ H 1 (N ; Z2 ) of N , then w2 (T ∗ N ) = b2 . This implies that (T ∗ N, ωN ) carries a metaplectic structure if w1 (N ) = 0, i.e. if N is orientable. Example 3.1.7 Consider the complex projective space CPn endowed with the K¨ ahler form ωFS of the Fubini–Study metric. Then, c1 (CPn , ωFS ) coincides with the first Chern class c1 (CPn ) of the complex tangent bundle of CPn . Since c1 (CPn ) = (n + 1)a , where a is the positive generator of H 2 (CPn ; Z), the symplectic manifold (CPn , ωFS ) has a metaplectic structure if and only if n is odd. Since CPn is simply connected, such a structure has to be unique.

3.2 Symplectic Spinor Bundle

37

3.2 Symplectic Spinor Bundle Let (M, ω) be a 2n-dimensional symplectic manifold and let P be a fixed metaplectic structure on (M, ω). Definition 3.2.1 The symplectic spinor bundle Q is defined to be the Hilbert space bundle Q = P ×m L2 (Rn ) associated to the metaplectic structure P via the metaplectic representation m : Mp(n, R) → U(L2 (Rn )). We recall that P ×m L2 (Rn ) is the orbit space of the Mp(n, R)-action on P × L2 (Rn ) defined by (p, f )q = pq, m q −1 f for p ∈ P, f ∈ L2 (Rn ), and q ∈ Mp(n, R). As usual, we will write the equivalence class of an element (p, f ) ∈ P × L2 (Rn ) as [p, f ]. Then, by definition, [pq, f ] = [p, m(q)f ] for every q ∈ Mp(n, R). Although the bundle Q has only the structure of a topological vector bundle, it is possible to define smooth sections of Q. For this, we apply that sections ϕ of Q correspond to mappings ϕˆ : P → L2 (Rn ) that are Mp(n, R)-equivariant, i.e. satisfy ˆ . ϕ(pq) ˆ = m q −1 ϕ(p) Explicitly, this correspondence is described by ϕ(πP (p)) = [p, ϕ(p)] ˆ . Definition 3.2.2 A section ϕ of the symplectic spinor bundle Q is said to be smooth if the corresponding mapping ϕˆ : P → L2 (Rn ) is smooth. A smooth section of Q will also be referred to as a symplectic spinor field. As for smooth bundles, we denote the space of smooth sections of Q by Γ(Q). Lemma 3.2.3 Let ϕ ∈ Γ(Q). Then ϕ(p) ˆ ∈ S(Rn ) for all p ∈ P.

38


Proof. Since ϕˆ is smooth, also the map ˆ ∈ L2 (Rn ) q ∈ Mp(n, R) → ϕˆ pq −1 = m(q)ϕ(p) for any p ∈ P is smooth. Thus ϕ(p) ˆ is a smooth vector of the metaplectic representation m, which is equivalent to ϕ(p) ˆ ∈ S(Rn ) by Proposition 1.3.6(2). Besides Q, one may consider the associated vector bundle S = P ×m S(Rn ) . By the above lemma, all smooth sections of Q are in fact sections of S. In analogy with Definition 3.2.2, we say that a section ϕ of S is smooth if the corresponding mapping ϕˆ : P → S(Rn ) is smooth, where S(Rn ) is endowed with its usual local convex topology. Since S(Rn ) is continuously embedded into L2 (Rn ), the space Γ(S) of smooth sections of S is contained in Γ(Q). The symplectic Clifford multiplication µ0 : R2n ⊗ S(Rn ) → S(Rn ) , defined in Definition 1.4.3, gives rise to a symplectic Clifford multiplication µ : TM ⊗ S → S on the bundle level as follows. First observe that T M = P ×ρ R2n .

(3.2.1)

Hence the tangent vectors to M can be written as equivalence classes [p, v] of pairs (p, v) ∈ P × R2n . We set µ([p, v] ⊗ [p, f ]) = [p, µ0 (v ⊗ f )] for [p, v] ∈ T M and [p, f ] ∈ S. Since µ0 is Mp(n, R)-equivariant by Lemma 1.4.4, this leads to a well defined endomorphism µ : T M ⊗ S → S. We will write µ(X ⊗ ϕ) = X · ϕ for X ∈ Γ(T M ) and ϕ ∈ Γ(Q). Again, we will use the convention X · Y · ϕ = X · (Y · ϕ) for iterative Clifford multiplication. Then, by Lemma 1.4.1(1) and the fact that ω([p, v], [p, w]) = ω0 (v, w) for [p, v], [p, w] ∈ T M , we have (X · Y − Y · X) · ϕ = −iω(X, Y )ϕ .

(3.2.2)


39

Remark 3.2.4 At this point, a comment is needed. By Lemma 3.2.3, for any X ∈ Γ(T M ) and ϕ ∈ Γ(Q), the Clifford product X · ϕ is well defined and again a section of the spinor bundle Q. This section, however, is not necessarily smooth, since the position and momentum operators Qj and Pj are not continuous on L2 (Rn ). On the other hand, if ϕ ∈ Γ(S), then, of course, also X ·ϕ ∈ Γ(S). Therefore, to avoid additional notations, we will understand ϕ ∈ Γ(Q) → X · ϕ ∈ Γ(Q) as an “unbounded” operator on Γ(Q), i.e. as an operator defined on a suitable subspace of Γ(Q) containing Γ(S). The same will be done for iterative Clifford multiplications and for operators involving the Clifford multiplication as the spinor derivative defined below and the Dirac operators introduced in the next chapter. For [p, f1 ], [p, f2 ] ∈ Q, we set [p, f1 ], [p, f2 ] = f1 , f2 , where f1 , f2 is the L2 -product of the functions f1 , f2 ∈ L2 (Rn ). Since Mp(n, R) acts by unitary operators on L2 (Rn ), this defines a Hermitian structure on Q. The symplectic Clifford multiplication and the Hermitian structure on Q are related by Proposition 3.2.5 For any X ∈ Γ(T M ) and any ϕ, ψ ∈ Γ(Q), X · ϕ, ψ = −ϕ, X · ψ . Proof. This follows from Lemma 1.4.1(2).

Any symplectic connection ∇ on M induces a covariant derivative, called spinor derivative, on the symplectic spinor bundle Q. This can be described in the following way. Consider the connection 1-form Z : T R → sp(n, R) on the symplectic frame bundle R of (M, ω) which corresponds to ∇ and take the lift Z of Z to the metaplectic structure P, i.e. that connection 1-form Z : T P → mp(n, R) on P that satisfies

FP∗ Z = ρ∗ ◦ Z .

Then the covariant derivative on Q associated to Z is the spinor derivative induced by ∇. We denote this covariant derivative on Q also by ∇.

40


Next we want to derive an expression for the spinor derivative in terms of a symplectic frame, which will be useful for our computations. For this, we fix the following notations. For a symplectic frame, i.e. a local section s : U → R of R on an open subset U of M , let s : U → P denote a lift of s to P. Thus FP ◦ s = s . Furthermore, for ϕ ∈ Γ(Q), we set ϕs = ϕˆ ◦ s . Then ϕ = [s, ϕs ] on U . Proposition 3.2.6 Let ∇ be a symplectic connection on M . Then i (fj · ∇X ej − ej · ∇X fj ) · ϕ 2 j=1

(3.2.3)

i (∇X fj · ej − ∇X ej · fj ) · ϕ = [s, X(ϕs )] + 2 j=1

(3.2.4)

n

∇X ϕ = [s, X(ϕs )] −

n

for X ∈ Γ(T M ) and ϕ ∈ Γ(Q), where s = (e1 , . . . , en , f1 , . . . , fn ) is any symplectic frame. Proof. Let Z and Z be as above. Then, by the definition of the spinor derivative and Proposition 1.4.5, ∇X ϕ = s, X(ϕs ) + m∗ s∗ Z(X) ϕs = s, X(ϕs ) − i s∗ Z(X) · ϕs . (3.2.5) Furthermore, from Lemma 1.3.1 and ρ∗ ◦ s∗ Z = s∗ ρ∗ ◦ Z = s∗ FP∗ Z = s∗ Z , we conclude ∗ i s∗ Z(X) · ϕs = i ρ−1 ∗ (s Z(X)) · ϕs n i = (bj · s∗ Z(X)aj − aj · s∗ Z(X)bj ) · ϕs . 2 j=1

On the other hand, since ej = [s, aj ]

and fj = [s, bj ]

(3.2.6)


41

and hence ∇X ej = [s, s∗ Z(X)aj ] and ∇X fj = [s, s∗ Z(X)bj ] , we have n

(fj · ∇X ej − ej · ∇X fj ) · ϕ

j=1



= s,

n

 (bj · s∗ Z(X)aj − aj · s∗ Z(X)bj ) · ϕs  .

j=1

This proves Equation (3.2.3). Further, using Equation (3.2.2) and the symplecticity of ∇, we see (∇X fj · ej − ∇X ej · fj ) · ϕ = (ej · ∇X fj − fj · ∇X ej ) · ϕ − i(ω(∇X fj , ej ) − ω(∇X ej , fj ))ϕ = (ej · ∇X fj − fj · ∇X ej ) · ϕ + iX(ω(ej , fj ))ϕ = (ej · ∇X fj − fj · ∇X ej ) · ϕ ,

which implies Equation (3.2.4).

We next want to prove some useful properties of the spinor derivative. For this, we fix a symplectic connection on M . We start with a product rule. Proposition 3.2.7 For any X, Y ∈ Γ(T M ) and ϕ ∈ Γ(Q), ∇X (Y · ϕ) = ∇X Y · ϕ + Y · ∇X ϕ . Proof. Let s : U → R be a symplectic frame. According to Equation (3.2.1), Y = [s, Ys ] for some mapping Ys : U → R2n . With this, ∇X Y = [s, X(Ys ) + s∗ Z(X)Ys ] and Y · ϕ = [s, Ys · ϕs ] . Hence, by Equation (3.2.5), ∇X Y · ϕ = [s, X(Ys ) · ϕs + (s∗ Z(X)Ys ) · ϕs ] , Y · ∇X ϕ = s, Ys · X(ϕs ) − i Ys · s∗ Z(X) · ϕs , ∇X (Y · ϕ) = s, X(Ys · ϕs ) − i s∗ Z(X) · Ys · ϕs .

42


Clearly, X(Ys · ϕs ) = X(Ys ) · ϕs + Ys · X(ϕs ) . Moreover, by Lemma 1.4.2(1) and Equation (3.2.6), s∗ Z(X) · Ys · ϕs = i (s∗ Z(X)Ys ) · ϕs + Ys · s∗ Z(X) · ϕs .

Altogether, this yields the assertion. The following proposition states that the spinor derivative is Hermitian. Proposition 3.2.8 For any X ∈ Γ(T M ) and ϕ, ψ ∈ Γ(Q), Xϕ, ψ = ∇X ϕ, ψ + ϕ, ∇X ψ . Proof. With respect to a symplectic frame s, we have ∇X ϕ, ψ + ϕ, ∇X ψ = X(ϕs ), ψs + ϕs , X(ψs ) −i s∗ Z(X) · ϕs , ψs + i ϕs , s∗ Z(X) · ψs

by the definition of the Hermitian structure on Q and Equation (3.2.5). Since ∗ s Z(X) · ϕs , ψs = ϕs , s∗ Z(X) · ψs by Lemma 1.4.2(2) and Xϕ, ψ = Xϕs , ψs = X(ϕs ), ψs + ϕs , X(ψs ) ,

the proposition is proved. The curvature RQ of the spinor derivative, defined by RQ (X, Y )ϕ = ∇X ∇Y ϕ − ∇Y ∇X ϕ − ∇[X,Y ] ϕ , is related to the curvature R of the underlying symplectic connection by

Proposition 3.2.9 Let s = (e1 , . . . , en , f1 , . . . , fn ) be any symplectic frame. Then i (ej · R(X, Y )fj − fj · R(X, Y )ej ) · ϕ 2 j=1

(3.2.7)

i (R(X, Y )fj · ej − R(X, Y )ej · fj ) · ϕ = 2 j=1

(3.2.8)

n

RQ (X, Y )ϕ =

n

for X, Y ∈ Γ(T M ) and ϕ ∈ Γ(Q).

3.3 Splitting of the Spinor Bundle

43

Proof. One easily checks that the right hand side of Equation (3.2.7) does not depend on the choice of the symplectic frame s. Thus, without loss of generality, we may assume that the form s∗ Z vanishes at an arbitrarily fixed p ∈ M . Then, at p, R(X, Y )ej = [s, (X(s∗ Z(Y )) − Y (s∗ Z(X))aj ] , R(X, Y )fj = [s, (X(s∗ Z(Y )) − Y (s∗ Z(X))bj ] , On the other hand, again applying Equation (3.2.6), one gets RQ (X, Y )ϕ = −i s, X s∗ Z(Y ) − Y s∗ Z(X) · ϕs ∗ ∗ = −i s, ρ−1 ∗ (X(s Z(Y )) − Y (s Z(X))) · ϕs at p. Now Equation (3.2.7) follows by Lemma 1.3.1 as in the proof of Proposition 3.2.6. Equation (3.2.8) is a consequence of Proposition 2.3.3(2) and Equation (3.2.2).

3.3 Splitting of the Spinor Bundle We consider again a symplectic manifold (M, ω) of dimension 2n with a fixed metaplectic structure P. Let J be an ω-compatible almost complex structure and let RJ denote the corresponding U(n)-reduction of the symplectic frame bundle R. So RJp for p ∈ M is the set of all unitary basises of Tp M . We set P J = FP−1 (RJ ) . Clearly, P J is a principal U(n)-bundle. Moreover, the symplectic spinor bunJ dle Q is associated to P via the restriction u of the metaplectic representation m to U(n), Q = P J ×u L2 (Rn ) . Therefore, the splitting of the unitary representation u into its components ul described in Section 1.5 gives rise to an orthogonal decomposition of Q into finite rank subbundles QJl . Here, QJl is the vector bundle associated to P J via ul , QJl = P J ×ul Wl . According to Equation (1.5.6), rankC QJl

=

n+l−1 l

.

The Hamilton operator H0 of the harmonic oscillator gives rise to an endomorphism of the bundle S as follows.

44


Definition 3.3.1 We define HJ : S → S by HJ ([p, f ]) = [p, H0 f ] for p ∈ P J and f ∈ S(Rn ). This definition is justified by Proposition 1.5.1. Clearly, n QJl = q ∈ S : HJ (q) = − l + q . 2

(3.3.1)

For some computations, it will be useful to understand the endomorphism HJ in the following way. Lemma 3.3.2 For any unitary frame s = (e1 , . . . , e2n ) and any ϕ ∈ Γ(Q), 1 ej · ej · ϕ . 2 j=1 2n

HJ ϕ =

Proof. Writing ϕ = [s, ϕs ], we get HJ ϕ = [s, H0 ϕs ] n n 1 1 = [s, aj · aj · ϕs ] + [s, bj · bj · ϕs ] 2 j=1 2 j=1 =

1 1 ej · ej · ϕ + en+j · en+j · ϕ 2 j=1 2 j=1

=

1 ej · ej · ϕ , 2 j=1

n

n

2n

where we have used Equation (1.5.4).

Lemma 3.3.3 For any ϕ, ψ ∈ Γ(Q), HJ ϕ, ψ = ϕ, HJ ψ . Proof. This follows from H0 f1 , f2 = f1 , H0 f2 for all f1 , f2 ∈ L2 (Rn ).

Next we prove that the spinor derivative commutes with HJ provided the underlying connection on M is Hermitian.


45

Proposition 3.3.4 Let ∇ be a symplectic connection on M and suppose that ∇J = 0. Then, for any ϕ ∈ Γ(Q) and X ∈ Γ(T M ), ∇X (HJ ϕ) = HJ (∇X ϕ) . Proof. Let (e1 , . . . , e2n ) be a unitary frame. From ∇X ej =

2n

ω(∇X ej , Jek )ek

k=1

and ω((∇X J)ej , ek ) = ω(∇X (Jej ) − J(∇X ej ), ek ) = ω(∇X ek , Jej ) + ω(∇X ej , Jek ) , we deduce that 2n 2n (∇X ej · ej + ej · ∇X ej ) · ϕ = ω(∇X ej , Jek )(ek · ej + ej · ek ) · ϕ j=1

j,k=1

=

2n

ω((∇X J)ej , ek )ek · ej · ϕ

j,k=1

=0. By means of Proposition 3.2.7 and Lemma 3.3.2, it follows that 1 ∇X (ej · ej · ϕ) 2 j=1 2n

∇X (HJ ϕ) =

1 (∇X ej · ej · ϕ + ej · ∇X ej · ϕ + ej · ej · ∇X ϕ) 2 j=1 2n

=

= HJ (∇X ϕ) , so we are done.

Of course, the above proposition is originated in Proposition 1.5.1 and the fact that the connection 1-form of a Hermitian connection on M reduces to the subbundle RJ . Equation (3.3.1) and Proposition 3.3.4 imply that, for a Hermitian connection, the spaces Γ(QJl ) of smooth sections of QJl are preserved by the spinor derivative. That means, we have Corollary 3.3.5 Let ∇ be a symplectic connection on M satisfying ∇J = 0. Then, for any l ∈ N0 , it holds that, if ϕ ∈ Γ(QJl ) and X ∈ Γ(T M ), then also ∇X ϕ ∈ Γ(QJl ).

46


In the remaining of this section, we will give alternative descriptions of the complex line bundle QJ0 . !0 of all functions f ∈ S(Rn ) such that Lemma 3.3.6 The space W J0 v · f = iv · f for all v ∈ R2n coincides with the one-dimensional subspace W0 . !0 if and only if Proof. It is clear that f ∈ S(Rn ) belongs to W ∂f (x) = −xj f (x) ∂xj

(3.3.2)

for j = 1, . . . , n and that the Hermite function h0 ∈ W0 solves this system. On the other hand, differentiating Equation (3.3.2), we obtain ∂2f (x) = −f (x) + x2j f (x) . ∂x2j This implies

n H0 f = − f 2

and hence f ∈ W0 .

Corollary 3.3.7 The space Γ(QJ0 ) of smooth sections of QJ0 consists exactly of those symplectic spinor fields ϕ that satisfy JX · ϕ = iX · ϕ for all X ∈ Γ(T M ). Proof. One applies Lemma 3.3.6 and J([p, v]) = [p, J0 v] for p ∈ P J and v ∈ R2n . Let

T C M = T 1,0 M ⊕ T 0,1 M

be the decomposition of the complexified tangent bundle T C M given by the ±i eigenspaces of J. Since, in view of Proposition 3.1.4, the first Chern class c1 (M, ω) is even, there exists a square root of the so-called anti-canonical line bundle KJ = Λn T 1,0 M . More explicitly, we have Proposition 3.3.8 The line bundle QJ0 is a square root of the anti-canonical line bundle KJ .


Proof. We first note that T 1,0 M ∼ = P J ×ρˆ Cn , where ρ : U(n) → GL(n, C) is the restriction of ρ to U(n) composed with a −b ι: ∈ U(n) → a + ib ∈ GL(n, C) . b a Indeed, if ΦJ : RJ × Cn → T 1,0 M is given by 1 zj (ej − iJej ) , 2 j=1 n

ΦJ ((e1 , . . . , e2n ), z) = then

[p, z] ∈ P J ×ρˆ Cn → ΦJ (FP (p), z) ∈ T 1,0 M is a well defined bundle isomorphism. Hence KJ ∼ = P J ×det ◦ρˆ C . Consequently, we have proved the proposition, if we have shown, that (det ◦ ρ)(q) = (u0 (q))2 for all q ∈ U(n). For this, it suffices to verify that (det ◦ ρ)∗ = 2(u0 )∗ . Using Proposition 1.4.5, we compute ((u0 )∗ (aj · ak + bj · bk )h0 )(x) = (m∗ (aj · ak + bj · bk )h0 )(x) ∂ 2 h0 = i xj xk h0 (x) − (x) ∂xj ∂xk = iδjk h0 (x) and ((u0 )∗ (aj · bk − ak · bj )h0 )(x) = (m∗ (aj · bk − ak · bj )h0 )(x) ∂h0 ∂h0 = xj (x) − xk (x) ∂xk ∂xj =0. On the other hand, by means of the easily checked fact that (Tr ◦ ι∗ )(v w + J0 v J0 w) n = 2i (ω0 (v, aj )ω0 (w, aj ) + ω0 (v, bj )ω0 (w, bj )) , j=1

47

48


we obtain (det ◦ ρ)∗ (aj · ak + bj · bk ) = (Tr ◦ ι∗ )(aj ak + bj bk ) = 2iδjk and (det ◦ ρ)∗ (aj · bk − ak · bj ) = (Tr ◦ ι∗ )(aj bk − ak bj ) = 0 . This concludes the proof.

4 Symplectic Dirac Operators

Having introduced all the necessary material, we now define symplectic Dirac These operators, acting on symplectic spinor fields, are operators D and D. defined in a similar way as the Dirac operator on Riemannian manifolds. Ac depend on a choice of a metaplectic structure cordingly, the operators D and D as well as on a choice of a symplectic connection on the underlying manifold. These dependences can be investigated by methods developed by Friedrich for the Riemannian situation (cf. [20, 18]). To describe the dependence of the on the ω-compatible almost complex structure, we will adapt the operator D strategy used by Baum for studying the transformation of the Riemannian Dirac operator under a conformal change of the Riemannian metric (cf. [2]). At this point, we wish to emphasize that, although the whole construction follows the same procedure as for the classical Riemannian Dirac operator, using the symplectic structure on M instead of the Riemannian metric, the symplectic Dirac operators and the Riemannian Dirac operator have essentially different properties. These are particularly caused by the completely different structures of the Clifford algebras and Clifford multiplications.

4.1 Definition of the Operators Let a 2n-dimensional symplectic manifold (M, ω) be given and suppose that (M, ω) admits a metaplectic structure. After fixing a metaplectic structure P and a symplectic connection ∇ on M , we may consider the symplectic spinor bundles Q and S as well as the spinor derivative ∇ : Γ(Q) → Γ(T ∗ M ⊗ Q) . In view of Lemma 3.2.3, the Clifford multiplication µ : T M ⊗ S → S induces a map

50


µ : Γ(T M ⊗ Q) → Γ(Q) . Further, let J be an ω-compatible almost complex structure and let g denote the associated Riemannian metric on M . Definition 4.1.1 The symplectic Dirac operator D of (M, ω) with respect to the metaplectic structure P and the symplectic connection ∇ is defined as the composition ω

D = µ ◦ ∇ : Γ(Q) → Γ(T ∗ M ⊗ Q) ∼ = Γ(T M ⊗ Q) → Γ(Q) , where we identify the bundles T ∗ M and T M by putting a tangent vector into the first argument of ω. Using the Riemannian metric g instead of ω for identifying T ∗ M and T M , we obtain a second symplectic Dirac operator g

= µ ◦ ∇ : Γ(Q) → Γ(T ∗ M ⊗ Q) ∼ D = Γ(T M ⊗ Q) → Γ(Q) . are As explained in Remark 3.2.4, the symplectic Dirac operators D and D understood as operators defined on a suitable subspace of Γ(Q). Locally, D can be expressed as follows. and D Lemma 4.1.2 Let (e1 , . . . , en , f1 , . . . , fn ) be any symplectic frame. Then Dϕ =

n ej · ∇fj ϕ − fj · ∇ej ϕ j=1

and = Dϕ

n Jej · ∇fj ϕ − Jfj · ∇ej ϕ . j=1

Proof. Writing ∇ϕ =

n

e∗j ⊗ ∇ej ϕ + fj∗ ⊗ ∇fj ϕ

j=1

and using Equation (2.1.1), one has   n −fj ⊗ ∇ej ϕ + ej ⊗ ∇fj ϕ  Dϕ = µ  j=1

and



 n = µ Dϕ −Jfj ⊗ ∇ej ϕ + Jej ⊗ ∇fj ϕ  , j=1

which yields the assertion. An immediate consequence is

4.1 Definition of the Operators

51

Lemma 4.1.3 Let (e1 , . . . , e2n ) be any unitary frame. Then Dϕ = −

2n

Jej · ∇ej ϕ

= Dϕ

and

j=1

2n

ej · ∇ej ϕ .

j=1

Example 4.1.4 We consider the symplectic manifold (R2 , ω0 ). Its unique, up to isomorphism, metaplectic structure is the trivial Mp(1, R)-bundle P = R2 × Mp(1, R) . Consequently, also the bundle S is trivial. Thus a section of it is a mapping ϕ : R2 → S(R), i.e. a mapping ϕ : R3 → C such that, for any (s, t) ∈ R2 , x ∈ R → ϕ(s, t, x) ∈ C lies in S(R). Now choose the canonical flat connection ∇ on R2 as symplectic connection and J0 as ω0 -compatible almost complex structure. Then, according to take the form Lemma 4.1.2, the symplectic Dirac operators D and D (Dϕ)(s, t, x) = ix

∂2ϕ ∂ϕ (s, t, x) − (s, t, x) ∂t ∂x∂s

(4.1.1)

and

∂ϕ ∂2ϕ (s, t, x) + (s, t, x) . (4.1.2) ∂s ∂x∂t Obviously, any symplectic spinor field ϕ : R3 → C independent of (s, t) is Moreover, using the complex an element of the kernel of D as well as of D. coordinate z = s + it, Equations 4.1.1 and 4.1.2 read as ∂ϕ ∂ϕ ∂ ∂ϕ ∂ϕ (z, x) − (z, x) − i (z, x) + (z, x) (Dϕ)(z, x) = ix ∂z ∂z ∂x ∂z ∂z (Dϕ)(s, t, x) = ix

and (Dϕ)(z, x) = ix

∂ϕ ∂ϕ ∂ ∂ϕ ∂ϕ (z, x) + (z, x) + i (z, x) − (z, x) . ∂z ∂z ∂x ∂z ∂z

This implies that all mappings ϕ : C × R → C given by ϕ(z, x) = e−x

2

/2

h(z)

= 0. The for any holomorphic function h on C satisfy Dϕ = 0 and Dϕ equivalence of the last two equations for such ϕ can also be derived from the following fact. Namely, by Equations (1.4.1), (4.1.1), and (4.1.2), for any = 0, symplectic spinor field ϕ on R2 , one has Dϕ = 0 if and only if DFϕ where Fϕ means the Fourier transformed of ϕ with respect to the variable x. In Chapter 7, we will come back to this relation in a more general setting.

52


4.2 Dependence on the Symplectic Connection Here, we carry over the concepts developed for Riemannian Dirac operators defined with respect to different metric connections from the orthogonal case to the symplectic one. Let a metaplectic structure P on (M, ω) as well as an ω-compatible almost complex structure J be fixed and let ∇ and ∇ be two symplectic connections on M . We define a (2, 1)-tensor field A on M by A(X, Y ) = ∇X Y − ∇X Y . Clearly,

A(X, Y ) − A(Y, X) = (T − T)(X, Y ) ,

(4.2.1)

where T and T are the torsions of ∇ and ∇ , respectively. Let the (3, 0)-tensor defined by fields B and B B(X, Y, Z) = ω(A(X, Y ), Z)

and B(X, Y, Z) = B(X, JY, JZ)

and set C(X, Y, Z) = B(X, Y, Z) + B(Y, Z, X) + B(Z, X, Y ) and

C(X, Y, Z) = B(X, Y, Z) + B(Y, Z, X) + B(Z, X, Y ) .

Lemma 4.2.1 (1) B(X, Y, Z) = B(X, Z, Y ) for any X, Y, Z ∈ Γ(T M ). are symmetric. (2) The (3, 0)-tensor fields C and C (3) B is symmetric if and only if T = T . if and only if ∇J = ∇ J. (4) B = B Proof. Assertion (1) follows from Lemma 2.2.2. Assertion (2) is a consequence of (1). Assertion (3) is obtained from Equation (4.2.1) and (1). The last assertion is seen from B(X, Y, Z) − B(X, Y, Z) = ω((∇X J)Y − (∇X J)Y, JZ) for all X, Y, Z ∈ Γ(T M ).

Let div : Γ(T M ) → C ∞ (M ) be the divergence operator with respect to ∇ ! : Γ(T M ) → C ∞ (M ) be defined by and let div ! div(X) = Tr(Y → J(∇Y X)) . Then ! div(X) =

n ω ∇fj X, Jej − ω ∇ej X, Jfj , j=1

! denote where (e1 , . . . , en , f1 , . . . , fn ) is any symplectic frame. Let div and div the corresponding operators with respect to ∇ . One easily checks

4.2 Dependence on the Symplectic Connection

53

Lemma 4.2.2 For any X ∈ Γ(T M ), div (X) − div(X) =

n

(B(ej , X, fj ) − B(fj , X, ej ))

j=1

and

n

! ! (X) − div(X) = div

j , JX, fj ) − B(f j , JX, ej )) . (B(e

j=1

be the symplectic Dirac operators with respect Theorem 4.2.3 Let D and D be the corresponding operators with respect to ∇ . to ∇ and let D and D Then D and D coincide if and only if the tensor field C identically vanishes =D if and only if and the operators div and div coincide. Analogously, D ! ! C = 0 and div = div . Proof. Let ϕ ∈ Γ(Q) and let (e1 , . . . , e2n ) be a unitary frame. Applying Proposition 3.2.6, we compute ∇X ϕ

i i − ∇X ϕ = − Jek · ∇X ek · ϕ + Jek · ∇X ek · ϕ 2 2 2n

2n

k=1

k=1

2n i =− Jek · A(X, ek ) · ϕ 2 k=1

=

i 2

2n

B(X, ek , el )Jek · Jel · ϕ

k,l=1

for the difference of the spinor derivatives induced by ∇ and ∇ . Together with Lemma 4.1.3, this implies D ϕ − Dϕ = −

2n i B(ej , ek , el )Jej · Jek · Jel · ϕ 2 j,k,l=1

=

2n i B(Jej , Jek , Jel )ej · ek · el · ϕ . 2 j,k,l=1

Next, observe that 2n

2n

B(Jek , Jel , Jej )ω(el , ej )ek = −

j,k,l=1

B(Jek , Jel , ej )ω(el , Jej )ek

j,k,l=1 2n

=−

j,k=1

=0,

B(Jek , Jej , ej )ek

54


since substituting Jej for ej and Lemma 4.2.1(1) yield 2n

2n

B(Jek , Jej , ej )ek = −

j,k=1

B(Jek , ej , Jej )ek

j,k=1 2n

=−

B(Jek , Jej , ej )ek .

j,k=1

Further, 2n

B(Jek , Jel , Jej )ω(ek , ej )el = −

j,k,l=1

2n

B(ek , Jel , Jej )ω(Jek , ej )el

j,k,l=1

=

2n

B(ej , Jel , Jej )el .

j,l=1

Consequently, 2n

B(Jek , Jel , Jej )ek · el · ej · ϕ

j,k,l=1 2n

=

B(Jek , Jel , Jej )(ej · ek · el − iω(el , ej )ek − iω(ek , ej )el ) · ϕ

j,k,l=1 2n

=

B(Jek , Jel , Jej )ej · ek · el · ϕ − i

j,k,l=1

2n

B(ej , Jel , Jej )el · ϕ .

j,l=1

Analogously, 2n

B(Jel , Jej , Jek )ω(el , ej )ek =

j,k,l=1

and

2n

B(Jel , Jej , Jek )ω(el , ek )ej =

j,k,l=1

2n

B(ej , Jek , Jej )ek

j,k=1 2n

B(ek , Jej , Jek )ej ,

j,k=1

which leads to 2n

B(Jel , Jej , Jek )el · ej · ek · ϕ

j,k,l=1

=

2n

B(Jel , Jej , Jek )(ej · ek · el − iω(el , ej )ek − iω(el , ek )ej ) · ϕ

j,k,l=1

=

2n j,k,l=1

B(Jel , Jej , Jek )ej · ek · el · ϕ − 2i

2n j,k=1

B(ej , Jek , Jej )ek · ϕ .

4.2 Dependence on the Symplectic Connection

55

It follows that 2n

3

B(Jej , Jek , Jel )ej · ek · el · ϕ

j,k,l=1 2n

=

B(Jej , Jek , Jel )ej · ek · el · ϕ +

j,k,l=1

+

2n

B(Jek , Jel , Jej )ek · el · ej · ϕ

j,k,l=1

2n

B(Jel , Jej , Jek )el · ej · ek · ϕ

j,k,l=1 2n

=

C(Jej , Jek , Jel )ej · ek · el · ϕ − 3i

j,k,l=1

2n

B(ej , Jek , Jej )ek · ϕ .

j,k=1

By Lemma 4.2.2, 2n

B(ej , Jek , Jej ) = div (Jek ) − div(Jek ) .

j=1

Thus we arrive at D ϕ − Dϕ =

2n i C(Jej , Jek , Jel )ej · ek · el · ϕ 6

+

j,k,l=1 n

1 2

(div (Jek ) − div(Jek ))ek · ϕ .

k=1

Now, arranging the factors of ej · ek · el in ascending order and applying Lemma 1.4.2(3), this proves the assertion for the Dirac operators D and D . Since

2n j , ek , el )ej · ek · el · ϕ ϕ − Dϕ = i D B(e 2 j,k,l=1

and, according to Lemma 4.2.2, 2n

! ! B(Je j , ek , ej ) = div (Jek ) − div(Jek ) ,

j=1

and D . the same reasoning also yields the assertion for the operators D

Corollary 4.2.4 If ∇ and ∇ are two symplectic connections on M such that T = T , then their Dirac operators D and D coincide if and only if ∇ = ∇ . That is, different symplectic connections having the same torsion induce different Dirac operators D.

56


Proof. By Lemma 4.2.1(3), in case T = T , the conditions C = 0 and ∇ = ∇ are equivalent. Hence the corollary is an easy consequence of the previous theorem. =D . Corollary 4.2.5 If ∇J = ∇ J, then D = D if and only if D by Lemma 4.2.1(4). This implies Proof. Suppose ∇J = ∇ J. Then B = B C = C and, by Lemma 4.2.2,

! (X) − div(X) ! div = div (JX) − div(JX) . The corollary now immediately follows from the above theorem.

To illustrate Theorem 4.2.3, we look at the following examples. Here, we restrict ourselves to the Dirac operator D. Examples explaining the relations may be found analogously. for the operator D Example 4.2.6 We consider the symplectic manifold (R2n , ω0 ) with n ≥ 2 and fix any symplectic connection ∇ on it. Let Ai : R2n × R2n → R2n , i = 1, 2, 3, be defined by A1 (v1 , v2 ) = ω0 (a1 , v1 )(ω0 (b1 , v2 )b2 + ω0 (b2 , v2 )b1 ) , A2 (v1 , v2 ) = ω0 (b1 , v1 )(ω0 (a1 , v2 )b2 + ω0 (b2 , v2 )a1 ) , A3 (v1 , v2 ) = ω0 (b2 , v1 )(ω0 (a1 , v2 )b1 + ω0 (b1 , v2 )a1 ) . Clearly, ω0 (Ai (v1 , v2 ), v3 ) = ω0 (Ai (v1 , v3 ), v2 )

(4.2.2)

for i = 1, 2, 3. We set A = t1 A1 + t2 A2 + t3 A3 for real numbers t1 , t2 , t3 . By Lemma 2.2.2 and Equation (4.2.2), understanding A as a constant (2, 1)-tensor field on R2n , ∇X Y = ∇X Y + A(X, Y )

(4.2.3)

defines a further symplectic connection ∇ on R2n . One easily computes that, in this situation, C(v1 , v2 , v3 ) = (t1 + t2 + t3 ) ω0 (a1 , vς(1) )ω0 (b1 , vς(2) )ω0 (b2 , vς(3) ) , ς∈S3

where S3 denotes the symmetric group. In particular, C(b1 , a1 , a2 ) = t1 + t2 + t3 . Moreover,

4.3 Dependence on the Metaplectic Structure n

57

(B(aj , v, bj ) − B(bj , v, aj )) = (t1 − t2 )ω0 (b2 , v) .

j=1

Hence C = 0 if and only if t1 +t2 +t3 = 0, whereas, by Lemma 4.2.2, div = div if and only if t1 = t2 . Specializing the parameters such that (t1 , t2 , t3 ) = (1, t, −1 − t) , we get C = 0 for all t, but div = div only for t = 1. On the other hand, if (t1 , t2 , t3 ) = (t, t, −2) , then C = 0 only if t = 1, whereas div = div for all t. Of course, if (t1 , t2 , t3 ) = (t, 1, 1) , then C = 0 or div = div for all t. Furthermore, in all cases, ∇ = ∇. Note that, multiplying A by a suitable cut-off function and applying the Darboux theorem, the example can be carried over to any symplectic manifold of dimension at least 4. Example 4.2.7 We again consider (R2n , ω0 ), where now the case n = 1 is allowed, and define A : R2n × R2n → R2n by A(v1 , v2 ) = ω0 (a1 , v1 )ω0 (a1 , v2 )a1 . As in the above example, the connection ∇ defined by Equation (4.2.3) for a given symplectic connection ∇ on R2n is symplectic, too. Moreover, we see that div = div . Thus, in contrast to the Riemannian situation, in dimension 2, there are examples of different symplectic connections having the same divergence.

4.3 Dependence on the Metaplectic Structure We again adapt the techniques used for the classical Riemannian Dirac operator to study symplectic Dirac operators with respect to different metaplectic structures. Let P1 and P2 be two metaplectic structures on the symplectic manifold (M, ω). We set = {(p1 , p2 ) ∈ P1 × P2 : FP (p1 ) = FP (p2 )} C 1 2 by and define an action of Z2 = {±1} on C

58


(p1 , p2 )(−1) = (p1 e− , p2 e− ) . Here, e− denotes that element of Mp(n, R) that lies over the unit element 2 and let of Sp(n, R) but is not the unit element of Mp(n, R). Let C = C/Z [p1 , p2 ] ∈ C be the equivalence class of (p1 , p2 ). The group Sp(n, R) × Z2 acts on C via [p1 , p2 ](ρ(q), 1) = [p1 q, p2 q] , [p1 , p2 ](ρ(q), −1) = [p1 q, p2 qe− ] for q ∈ Mp(n, R). Defining πC : C → M by πC [p1 , p2 ] = πP1 (p1 ) = πP2 (p2 ) , C gets the structure of a principal Sp(n, R)×Z2 -bundle. Moreover, the double covering FC : C → R given by FC ([p1 , p2 ]) = FP1 (p1 ) = FP2 (p2 ) satisfies πR ◦ FC = πC and FC ([p1 , p2 ](A, a)) = FC ([p1 , p2 ])A for any (A, a) ∈ Sp(n, R) × Z2 . Thus C together with FC is an equivariant lift of the symplectic frame bundle R with respect to the homomorphism (A, a) ∈ Sp(n, R) × Z2 → A ∈ Sp(n, R) . This lift is called the deformation of the metaplectic structures P1 and P2 . Proposition 4.3.1 The metaplectic structures P1 and P2 are isomorphic if and only if their deformation C is isomorphic to the trivial lift R × Z2 of R. Proof. Let the metaplectic structures P1 and P2 be isomorphic. That means, there is an isomorphism Φ : P1 → P2 of the principal Mp(n, R)-bundles such that (4.3.1) FP2 ◦ Φ = FP1 . In particular, Φ(p1 e− ) = Φ(p1 )e− for any p1 ∈ P1 . Therefore z(FP1 (p1 )) = [p1 , Φ(p1 )] ,

(4.3.2)

defines a section z : R → C of the covering FC . Conversely, if the deformation C is isomorphic to R × Z2 , then there exists a section z : R → C of FC such that z(rA) = z(r)(A, 1)

4.3 Dependence on the Metaplectic Structure

59

for any r ∈ R and A ∈ Sp(n, R). Let Φ : P1 → P2 be defined by Equation (4.3.2). Then, clearly, Φ satisfies Equation (4.3.1). Moreover, for any q ∈ Mp(n, R), we have [p1 q, Φ(p1 q)] = z(FP1 (p1 q)) = z(FP1 (p1 )ρ(q)) = z(FP1 (p1 ))(ρ(q), 1) = [p1 q, Φ(p1 )q] and hence Φ(p1 q) = Φ(p1 )q . Thus Φ is an isomorphism of the metaplectic structures P1 and P2 .

Corollary 4.3.2 The metaplectic structures P1 and P2 are isomorphic if and only if their deformation C is not connected. Let E = C ×τ C be the complex line bundle associated to the deformation C via the representation τ : Sp(n, R)×Z2 → GL(C) given by the projection onto the second factor and let Q1 and Q2 denote the symplectic spinor bundles associated to P1 and P2 , respectively. Lemma 4.3.3 Setting ΦE ([[p1 , p2 ], z] ⊗ [p1 , f ]) = [p2 , zf ] , one gets a bundle isomorphism ΦE : E ⊗ Q1 → Q2 . Proof. Let q ∈ Mp(n, R). On the one hand,

[[p1 , p2 ], z] ⊗ [p1 , f ] = [[p1 q, p2 q], z] ⊗ p1 q, m q −1 f = [[p1 q, p2 qe− ], −z] ⊗ p1 q, m q −1 f .

On the other hand, since m(e− )f = −f , [p2 , zf ] = p2 q, zm q −1 f = p2 qe− , −zm q −1 f . This shows that ΦE is correctly defined. Obviously, ΦE is surjective. The injectivity of ΦE can be easily checked. Consequently, if the complex line bundle E is trivial, then the symplectic spinor bundles Q1 and Q2 are isomorphic. In particular, E is trivial, if the metaplectic structures P1 and P2 are isomorphic. Since E is the complexification of a real line bundle, one has 2c1 (E) = 0. Thus E is also trivial, if the second cohomology group H 2 (M ; Z) has no torsion of order 2. Now we fix a symplectic connection ∇ on M and an ω-compatible almost complex structure J. Let Z C : T C → sp(n, R) be the lift of the connection 1-form Z : T R → sp(n, R) of ∇ to the deformation C and let ∇E denote the covariant derivative in E induced by Z C .

60


Lemma 4.3.4 ∇E is flat. Proof. Let C ∈ Γ(E). Since the differential τ∗ of the representation τ is trivial, writing C = [r, Cr ] for a local section r of C, we have " # ∗ C ∇E X C = r, X(Cr ) + τ∗ r Z (X) Cr = [r, X(Cr )] ,

which implies the assertion.

Let ∇1 and ∇2 denote the spinor derivatives in the symplectic spinor bundles 1 and D2 , D 2 be the corresponding Q1 and Q2 , respectively, and let D1 , D Dirac operators. Replacing the spinor derivative by the covariant derivative ∇1,E = id ⊗ ∇1 + ∇E ⊗ id : Γ(E ⊗ Q1 ) → Γ(T ∗ M ⊗ E ⊗ Q1 ) in Definition 4.1.1, we obtain the twisted Dirac operators ω

∼ Γ(T M ⊗ E ⊗ Q ) DE = µ ◦ ∇1,E : Γ(E ⊗ Q1 ) → Γ(T ∗ M ⊗ E ⊗ Q1 ) = 1 → Γ(E ⊗ Q1 ) and g

∼ Γ(T M ⊗ E ⊗ Q ) E = µ ◦ ∇1,E : Γ(E ⊗ Q1 ) → Γ(T ∗ M ⊗ E ⊗ Q1 ) = D 1 → Γ(E ⊗ Q1 ) . E are given as follows. Locally, the operators DE and D Lemma 4.3.5 Let (e1 , . . . , en , f1 , . . . , fn ) be any symplectic frame. Then DE (C ⊗ ϕ) = C ⊗ D1 ϕ +

n

E ∇E fj C ⊗ ej · ϕ − ∇ej C ⊗ fj · ϕ

j=1

and E (C ⊗ ϕ) = C ⊗ D 1 ϕ + D

n

E ∇E fj C ⊗ Jej · ϕ − ∇ej C ⊗ Jfj · ϕ

.

j=1

Proof. Proceed as for Lemma 4.1.2. A direct calculation yields E and D2 , D 2 satisfy Proposition 4.3.6 The operators DE , D ΦE ◦ DE = D2 ◦ ΦE

and

E = D 2 ◦ ΦE . ΦE ◦ D

4.3 Dependence on the Metaplectic Structure

61

To describe the relation between symplectic Dirac operators with respect to different metaplectic structures, we choose a local non-vanishing section C : U → E. Then, by E ΦE # (ϕ) = Φ (C ⊗ ϕ) for ϕ ∈ Γ(Q1 ), an isomorphism ΦE # between the restrictions of Q1 and Q2 to U is given. Let η# be the local connection form of ∇E with respect to C, i.e. the complex-valued 1-form on U defined by ∇E X C = η# (X)C . Since ∇E is flat, η# is closed. Let X# be the vector field dual to η# , i.e. the complex vector field on U given by η# (Y ) = ω(X# , Y ) , where the symplectic structure ω is extended complex linearly. Theorem 4.3.7 Let C be any local non-vanishing section in E. Then, for all ϕ ∈ Γ(Q1 ), E D2 ◦ ΦE # (ϕ) = Φ# (D1 ϕ + X# · ϕ) and

2 ◦ ΦE (ϕ) = ΦE (D 1 ϕ + JX# · ϕ) . D # #

Proof. Let (e1 , . . . , en , f1 , . . . , fn ) be a symplectic frame. By Lemma 4.3.5, DE (C ⊗ ϕ) = C ⊗ D1 ϕ + 

n

(η# (fj )C ⊗ ej · ϕ − η# (ej )C ⊗ fj · ϕ)

j=1

= C ⊗  D1 ϕ +

n

 (ω(X# , fj )ej + ω(ej , X# )fj ) · ϕ

j=1

= C ⊗ (D1 ϕ + X# · ϕ) and E (C ⊗ ϕ) = C ⊗ D 1 ϕ + D 

n j=1

(η# (fj )C ⊗ Jej · ϕ − η# (ej )C ⊗ Jfj · ϕ) 

1 ϕ + J  = C ⊗ D

n





(ω(X# , fj )ej + ω(ej , X# )fj ) · ϕ

j=1

1 ϕ + JX# · ϕ) . = C ⊗ (D Together with Proposition 4.3.6, this gives the assertion.

Remark 4.3.8 Pay attention to the fact that Theorem 4.3.7 is a local statement depending on the choice of the local section C in E. Of course, in the case that the complex line bundle E is trivial, one can choose a global nonvanishing section in E to make the relations global.

62


4.4 Dependence on the Almost Complex Structure The task of this section is to show that a change of the ω-compatible al most complex structure in the definition of the symplectic Dirac operator D corresponds to an isomorphism of the symplectic spinor bundle and a gauge transformation of the symplectic connection. Let J and K be two ω-compatible almost complex structures and set Θ = −J ◦ K . Since ω(ΘX, ΘY ) = ω(X, Y ) for all vector fields X and Y , ΦΘ (e1 , . . . , en , f 1 , . . . , f n ) = (Θe1 , . . . , Θen , Θf 1 , . . . , Θf n ) defines an isomorphism ΦΘ of the symplectic frame bundle R. Let ΨΘ : P → P be a lift of ΦΘ to the metaplectic structure P. Considering the tangent bundle T M as the associated vector bundle P ×ρ R2n , one has Θ[p, v] = [ΨΘ (p), v] . Let Ξ : Q → Q be the isomorphism of the symplectic spinor bundle Q given by Ξ([p, f ]) = [ΨΘ (p), f ] . Lemma 4.4.1 For any X ∈ Γ(T M ) and ϕ ∈ Γ(Q), Ξ(X · ϕ) = ΘX · Ξ(ϕ) . Proof. Let s : U → P be a local section of the metaplectic structure P. Writing X = [s, Xs ] and ϕ = [s, ϕs ], we have Ξ(X · ϕ) = Ξ([s, Xs ] · [s, ϕs ]) = Ξ([s, Xs · ϕs ]) = [ΨΘ ◦ s, Xs · ϕs ] and ΘX · Ξ(ϕ) = [ΨΘ ◦ s, Xs ] · [ΨΘ ◦ s, ϕs ] = [ΨΘ ◦ s, Xs · ϕs ] , which already proves the assertion.

Let again Z : T R → sp(n, R) be the connection form of the chosen symplectic connection ∇ on M and Z : T P → mp(n, R) the lift of Z to the metaplectic structure P. Then ∗ Z Z Θ = Φ−1 Θ is the connection form of a symplectic connection ∇Θ on M and the spinor derivative induced by ∇Θ , which is also denoted by ∇Θ , is associated to the connection form ∗ Θ Z = ΨΘ−1 Z on P.

4.4 Dependence on the Almost Complex Structure

63

Lemma 4.4.2 For any X ∈ Γ(T M ) and ϕ ∈ Γ(Q), Ξ(∇X ϕ) = ∇Θ X (Ξ(ϕ)) . Proof. Writing again ϕ = [s, ϕs ] with respect to a local section s of P, we have Ξ(∇X ϕ) = Ξ s, X(ϕs ) + m∗ s∗ Z(X) ϕs " # Θ = ΨΘ ◦ s, X(ϕs ) + m∗ (ΨΘ ◦ s)∗ Z (X) ϕs = ∇Θ X (Ξ(ϕ)) . Θ denote the symplectic Dirac operator which is defined with respect to Let D J Θ K be the corresponding operator with respect to ∇ and ∇ and J and let D K. Θ and D K are related by Theorem 4.4.3 The symplectic Dirac operators D J K ϕ) = D Θ (Ξ(ϕ)) Ξ(D J for all ϕ ∈ Γ(Q). Proof. Let (e1 , . . . , en , f1 , . . . , fn ) be any symplectic frame. By means of Lemmas 4.1.2, 4.4.1, and 4.4.2 and the relation Θ ◦ K = −J ◦ K 2 = J , we compute K ϕ) = Ξ(D

n Ξ Kej · ∇fj ϕ − Ξ Kfj · ∇ej ϕ j=1

n Θ ◦ K(ej ) · Ξ ∇fj ϕ − Θ ◦ K(fj ) · Ξ ∇ej ϕ =

=

j=1 n

Θ Jej · ∇Θ (Ξ(ϕ)) − Jf · ∇ (Ξ(ϕ)) j fj ej

j=1

JΘ (Ξ(ϕ)) . =D

64


4.5 Formal Self-Adjointness In the last section of this chapter, we want to show that, under certain con are formally self-adjoint with respect to ditions, the Dirac operators D and D 2 the L -product on Γ(Q) defined by ϕ, ψ dM (ϕ, ψ) = M

for ϕ ∈ Γ(Q) and ψ ∈ Γ0 (Q). Here, Γ0 (Q) denotes the space of compactly supported smooth sections of Q. Lemma 4.5.1 For any ϕ, ψ ∈ Γ(Q), Dϕ, ψ = ϕ, Dψ + div(Y1 ) , where the vector field Y1 on M is given by Y1 =

2n

ϕ, Jej · ψ ej

j=1

for any unitary frame (e1 , . . . , e2n ). Proof. By means of Propositions 3.2.5, 3.2.7, and 3.2.8 and Lemma 4.1.3, we see Dϕ, ψ = −

2n

Jej · ∇ej ϕ, ψ

j=1

=

2n

∇ej ϕ, Jej · ψ

j=1

=

2n

ej ϕ, Jej · ψ − ϕ, ∇ej (Jej · ψ)

j=1

= ϕ, Dψ +

2n ej ϕ, Jej · ψ − ϕ, ∇ej (Jej ) · ψ . j=1

On the other hand, we have div(Y1 ) =

2n ω ∇ej Y1 , Jej j=1

=

2n ej ϕ, Jek · ψ ω(ek , Jej ) + ϕ, Jek · ψ ω ∇ej ek , Jej j,k=1

4.5 Formal Self-Adjointness

=

65

2n ej ϕ, Jek · ψ δjk − ϕ, Jek · ψ ω ek , ∇ej (Jej ) j,k=1

=

2n

ej ϕ, Jej · ψ − ϕ, ∇ej (Jej ) · ψ ,

j=1

and the lemma is proved. Lemma 4.5.2 Suppose that ∇J = 0. Then, for any ϕ, ψ ∈ Γ(Q), ψ = ϕ, Dψ − div(Y2 ) , Dϕ, where Y2 ∈ Γ(T M ) is given by Y2 =

2n

ϕ, ej · ψ ej

j=1

for any unitary frame (e1 , . . . , e2n ). Proof. We proceed as before to derive that ψ = Dϕ,

2n

ej · ∇ej ϕ, ψ

j=1

=−

2n ∇ej ϕ, ej · ψ j=1

=

2n

ϕ, ∇ej (ej · ψ) − ej ϕ, ej · ψ

j=1

+ = ϕ, Dψ

2n

ϕ, ∇ej ej · ψ − ej ϕ, ej · ψ .

j=1

Furthermore, div(Y2 ) =

2n ej ϕ, ek · ψ ω(ek , Jej ) + ϕ, ek · ψ ω ∇ej ek , Jej j,k=1

=

2n ej ϕ, ek · ψ δjk − ϕ, ek · ψ ω ek , ∇ej (Jej ) . j,k=1

Since, by assumption, ∇J = 0, we obtain

66


div(Y2 ) =

2n ej ϕ, ek · ψ δjk − ϕ, ek · ψ ω ek , J ∇ej ej j,k=1

=

2n ej ϕ, ej · ψ − ϕ, ∇ej ej · ψ . j=1

This finishes the proof.

Theorem 4.5.3 Let the symplectic connection ∇ on M be such that its torsion vector field T vanishes. Then the Dirac operator D is formally selfadjoint, i.e., (Dϕ, ψ) = (ϕ, Dψ) for all ϕ ∈ Γ(Q) and ψ ∈ Γ0 (Q). If, in addition, ∇J = 0, then also the Dirac is formally self-adjoint. operator D Proof. If T = 0, then, by Corollary 2.2.7, div(X) dM = 0 M

for all compactly supported vector fields X on M . Hence the theorem follows from Lemmas 4.5.1 and 4.5.2.

5 An Associated Second Order Operator

In this chapter, we introduce an elliptic operator of second order which is of Laplace type and will play a central role in the further investigations. In a sense, this operator can be considered as the symplectic counterpart of the square of the Dirac operator in the Riemannian case.

5.1 Definition and Ellipticity Equation (3.2.2) suggests to study the following differential operator. Definition 5.1.1 Let P : Γ(Q) → Γ(Q) be the second order operator defined by D] = i(DD − DD) . P = i[D, Let us compute the principal symbol of P. We start with Lemma 5.1.2 For any u ∈ C ∞ (M ) and ϕ ∈ Γ(Q), D(uϕ) = uDϕ + Xu · ϕ

and

+ JXu · ϕ . D(uϕ) = uDϕ

Proof. This immediately follows from the definitions of D and D.

For a covector ξ ∈ Tp∗ M , let ξ & ∈ Tp M denote the dual tangent vector, i.e. ξ = ωp (ξ & , ) = gp (Jξ & , ) . of the symplectic Dirac Lemma 5.1.3 The principal symbols σ(D) and σ(D) operators D and D are given by σ(D)(p, ξ)s = ξ & · s where ξ ∈ Tp∗ M and s ∈ Sp .

and

σ(D)(p, ξ)s = Jξ & · s ,

68


Proof. Let u ∈ C ∞ (M ) and ϕ ∈ Γ(Q) be such that (du)p = ξ and ϕ(p) = s. From ξ & = Xu (p) and Lemma 5.1.2, we get σ(D)(p, ξ)s = D((u − u(p))ϕ)(p) = D(uϕ)(p) − u(p)Dϕ(p) = u(p)Dϕ(p) + Xu (p) · ϕ(p) − u(p)Dϕ(p) = ξ& · s as well as σ(D)(p, ξ)s = D((u − u(p))ϕ)(p) = D(uϕ)(p) − u(p)Dϕ(p) = u(p)Dϕ(p) + JXu (p) · ϕ(p) − u(p)Dϕ(p) & = Jξ · s . Corollary 5.1.4 The principal symbol σ(P) of the operator P is given by σ(P)(p, ξ)s = −gp (ξ & , ξ & )s for ξ ∈ Tp∗ M and s ∈ Sp . In particular, P is an elliptic operator of Laplace type. Proof. By Lemma 5.1.3 and Equation (3.2.2), σ(P)(p, ξ)s = i(σ(D)(p, ξ) ◦ σ(D)(p, ξ) − σ(D)(p, ξ) ◦ σ(D)(p, ξ))s & & & & = i(Jξ · ξ · s − ξ · Jξ · s) = ωp (Jξ & , ξ & )s = −gp (ξ & , ξ & )s .

5.2 A Weitzenb¨ ock Formula Due to Corollary 5.1.4, the highest order part of the operator P is a Laplace operator. The aim of this section is to prove the corresponding Weitzenb¨ ock formula. In Section 4.5, we have defined an L2 -product on Γ(Q). Similarly, we now equip Γ(T ∗ M ⊗ Q) with an L2 -product. Let Γ0 (T ∗ M ⊗ Q) denote the space of compactly supported smooth sections of T ∗ M ⊗ Q. For α ∈ Γ(T ∗ M ⊗ Q) and β ∈ Γ0 (T ∗ M ⊗ Q), we set

5.2 A Weitzenb¨ ock Formula

69

α, β dM ,

(α, β) = M

where the function α, β on M is given as α, β =

2n

α(ej ), β(ej ) .

j=1

Here and subsequently, (e1 , . . . , e2n ) is any unitary frame. Let

∇∗ : Γ(T ∗ M ⊗ Q) → Γ(Q)

be the formal L2 -adjoint operator of the spinor derivative. Using the torsion vector field T and the divergence operator div with respect to the chosen symplectic connection on M , the operator ∇∗ can be written as follows. Lemma 5.2.1 For all α ∈ Γ(T ∗ M ⊗ Q), ∇∗ α = −

2n

∇ej (α(ej )) + div(ej )α(ej ) + α(JT) .

j=1

Proof. By definition, (∇∗ α, ψ) =

α, ∇ψ dM M

for any ψ ∈ Γ0 (Q). By means of Proposition 3.2.8 and div(uX) = X(u) + u div(X) for u ∈ C ∞ (M ) and X ∈ Γ(T M ), we derive α, ∇ψ =

2n

α(ej ), ∇ej ψ

j=1

=

2n ej α(ej ), ψ − ∇ej (α(ej )), ψ j=1

=

2n div(α(ej ), ψ ej ) − α(ej ), ψ div(ej ) − ∇ej (α(ej )), ψ j=1

= div(Y ) −

2n ∇ej (α(ej )) + div(ej )α(ej ), ψ j=1

with

70


Y =

2n

α(ej ), ψ ej .

j=1

Since Y is a globally defined vector field with compact support, we can apply Corollary 2.2.7 to obtain ∇∗ α = −

2n ∇ej (α(ej )) + div(ej )α(ej ) + ω(ej , T)α(ej ) , j=1

which proves the lemma. Definition 5.2.2 We call the Bochner–Laplace operator ∆Q = ∇∗ ∇ : Γ(Q) → Γ(Q) the symplectic spinor Laplacian. Proposition 5.2.3 For all ϕ ∈ Γ(Q), ∆Q ϕ = −

2n ∇ej ∇ej ϕ + div(ej )∇ej ϕ + ∇JT ϕ . j=1

Proof. This is Lemma 5.2.1 applied to α = ∇ϕ.

To prove the announced Weitzenb¨ ock formula, we proceed similarly as in the Riemannian case. For X ∈ Γ(T M ), we set P(X) =

2n

ej · ∇Jej X ,

j=1

P(X) =

2n

ej · ∇ej X ,

j=1

P(J)(X) =

2n ∇Jej J (X) · ej . j=1

Then P(X), P(X), and P(J)(X) are globally defined expressions which act via symplectic Clifford multiplication on symplectic spinor fields. Lemma 5.2.4 For any X ∈ Γ(T M ) and ϕ ∈ Γ(Q), D(X · ϕ) = X · Dϕ + P(X) · ϕ − i∇X ϕ , + P(X) D(X · ϕ) = X · Dϕ · ϕ + i∇JX ϕ .


71

Proof. By means of Equation (3.2.2), Proposition 3.2.7, and Lemma 4.1.3, one computes D(X · ϕ) =

2n

ej · ∇Jej (X · ϕ)

j=1

= P(X) · ϕ +

2n

ej · X · ∇Jej ϕ

j=1

= P(X) · ϕ + X · Dϕ − i

2n

ω(ej , X)∇Jej ϕ

j=1

= P(X) · ϕ + X · Dϕ − i∇X ϕ and D(X · ϕ) =

2n

ej · ∇ej (X · ϕ)

j=1

= P(X) ·ϕ+

2n

ej · X · ∇ej ϕ

j=1

+i = P(X) · ϕ + X · Dϕ

2n

ω(X, ej )∇ej ϕ

j=1

+ i∇JX ϕ . = P(X) · ϕ + X · Dϕ Using the torsion T of the fixed symplectic connection ∇, we have Lemma 5.2.5 For any X ∈ Γ(T M ), P(X) + P(JX) = −P(J)(JX) − i div(X) +

2n

(ej (ω(ek , JX)) − ek (ω(ej , JX)))ej · Jek

j,k=1

−

2n

ω(T(ej , ek ) + [ej , ek ], JX)ej · Jek .

j,k=1

Proof. By Lemma 2.2.5 and Equation (3.2.2), we get P(X) = −

2n j=1

Jej · ∇ej X

72


=−

2n

∇ej X · Jej − i

j=1

=−

2n

2n ω ∇ej X, Jej j=1

∇ej X · Jej − i div(X) .

j=1

Further, by the symplecticity of ∇, ω ek , ∇ej (JX) = ej (ω(ek , JX)) − ω ∇ej ek , JX = ej (ω(ek , JX)) − ω(T(ej , ek ) + ∇ek ej + [ej , ek ], JX) = ej (ω(ek , JX)) − ω(T(ej , ek ) + [ej , ek ], JX) −ek (ω(ej , JX)) + ω(ej , ∇ek (JX)) , which implies P(JX) =

2n

ej · ∇ej (JX)

j=1

=

2n

ω(ek , ∇ej (JX))ej · Jek

j,k=1

=

2n

(ej (ω(ek , JX)) − ek (ω(ej , JX)))ej · Jek

j,k=1

−

2n

ω(T(ej , ek ) + [ej , ek ], JX)ej · Jek

j,k=1

−

2n

J(∇ek (JX)) · Jek .

k=1

Since 2n

2n ∇ej X + J ∇ej (JX) · Jej = − ∇ej J (JX) · Jej = P(J)(JX) ,

j=1

j=1

the assertion follows. After these preparations, we can prove Theorem 5.2.6 The operator P satisfies the Weitzenb¨ ock formula Pϕ = ∆Q ϕ + i

2n

Jej · ek · RQ (ej , ek )ϕ + i

j=1

j,k=1

−∇JT ϕ − i

2n j,k=1

2n

Jej · ek · ∇T(ej ,ek ) ϕ .

P(J)(Jej ) · ∇ej ϕ


Proof. By Lemmas 4.1.3 and 5.2.4, − DD)ϕ (DD =

2n

2n Jej · ∇e ϕ D D ej · ∇ej ϕ + j

j=1

j=1

2n = ej · D ∇ej ϕ + P(ej ) · ∇ej ϕ − i∇ej ∇ej ϕ j=1

+

2n ∇e ϕ + P(Je Jej · D j ) · ∇ej ϕ − i∇ej ∇ej ϕ j j=1

2n = Jej · ek · ∇ek ∇ej ϕ − ej · Jek · ∇ek ∇ej ϕ j,k=1

+

2n

(P(ej ) + P(Je j )) · ∇ej ϕ − 2i

j=1

2n

∇ej ∇ej ϕ .

j=1

Since 2n

Jej · ek · ∇ek ∇ej ϕ − ej · Jek · ∇ek ∇ej ϕ

j,k=1

=

2n

2n Jej · ek · ∇ek ∇ej ϕ − ∇ej ∇ek ϕ + i ω(ek , Jej )∇ej ∇ek ϕ

j,k=1

=−

j,k=1

2n

Jej · ek · RQ (ej , ek )ϕ −

j,k=1

2n

Jej · ek · ∇[ej ,ek ] ϕ + i

2n

∇ej ∇ej ϕ

j=1

j,k=1

and, by Lemma 5.2.5, 2n (P(ej ) + P(Je j )) · ∇ej ϕ j=1

=−

2n

P(J)(Jej ) · ∇ej ϕ − i

j=1

−

2n

div(ej )∇ej ϕ

j=1

2n

ω(T(ej , ek ) + [ej , ek ], Jel )ej · Jek · ∇el ϕ

j,k,l=1

=−

2n

P(J)(Jej ) · ∇ej ϕ − i

j=1

+

2n j,k=1

2n

div(ej )∇ej ϕ

j=1

Jej · ek · ∇T(ej ,ek ) ϕ +

2n j,k=1

Jej · ek · ∇[ej ,ek ] ϕ ,

73

74


we arrive at − DD)ϕ (DD =i

2n 2n ∇ej ∇ej ϕ + div(ej )∇ej ϕ + Jej · ek · RQ (ej , ek )ϕ j=1

+

2n

j,k=1 2n

P(J)(Jej ) · ∇ej ϕ −

j=1

Jej · ek · ∇T(ej ,ek ) ϕ .

j,k=1

Together with Proposition 5.2.3, this yields the claimed formula.

By Proposition 2.2.3, for any ω-compatible almost complex structure J, one can find a Hermitian connection. For such a connection, P(J)(X) vanishes. Thus we have Corollary 5.2.7 If the symplectic connection ∇ on M satisfies ∇J = 0, then Pϕ = ∆Q ϕ + i

2n

Jej · ek · RQ (ej , ek )ϕ

j,k=1 2n

−∇JT ϕ − i

Jej · ek · ∇T(ej ,ek ) ϕ

j,k=1

for all ϕ ∈ Γ(T M ).

In Section 6.1, we will study the Weitzenb¨ ock formula for the case of K¨ ahler manifolds.

5.3 Splitting of the Operator We will now show that the operator P leaves the subbundles QJl of the symplectic spinor bundle Q given in Section 3.3 invariant provided the chosen connection ∇ on M is Hermitian. Thus P splits into a family of Laplace type operators Pl : Γ(QJl ) → Γ(QJl ) obtained by restricting P to Γ(QJl ). Throughout this section, it is assumed that ∇J = 0. We first derive the and the following relations between the symplectic Dirac operators D and D endomorphism HJ . Proposition 5.3.1 For any ϕ ∈ Γ(Q), HJ (Dϕ) = D(HJ ϕ) + iDϕ

and

= D(H J ϕ) − iDϕ . HJ (Dϕ)

5.3 Splitting of the Operator

75

Proof. By means of Lemmas 3.3.2 and 4.1.3 and Proposition 3.3.4, we calculate HJ (Dϕ) = −

2n 1 ej · ej · Jek · ∇ek ϕ 2 j,k=1

=−

2n 1 (ej · Jek · ej · ∇ek ϕ − iω(ej , Jek )ej · ∇ek ϕ) 2 j,k=1

=−

2n 1 (Jek · ej · ej · ∇ek ϕ − 2iω(ej , Jek )ej · ∇ek ϕ) 2 j,k=1

=−

2n

Jek · HJ (∇ek ϕ) + i

2n

ej · ∇ej ϕ

j=1

k=1

=−

2n

Jek · ∇ek (HJ ϕ) + i

2n

ej · ∇ej ϕ

j=1

k=1

= D(HJ ϕ) + iDϕ and, analogously, 2n = 1 HJ (Dϕ) ej · ej · ek · ∇ek ϕ 2 j,k=1

=

2n 1 (ek · ej · ej · ∇ek ϕ − 2iω(ej , ek )ej · ∇ek ϕ) 2 j,k=1

=

2n

ek · HJ (∇ek ϕ) − i

k=1

=

2n

2n

ω(Jej , ek )Jej · ∇ek ϕ

j,k=1

ek · ∇ek (HJ ϕ) + i

2n

Jej · ∇ej ϕ

j=1

k=1

J ϕ) − iDϕ . = D(H Proposition 5.3.2 For all ϕ ∈ Γ(Q), HJ (Pϕ) = P(HJ ϕ) . In particular, for any l ∈ N0 , it holds that, if ϕ is a section of QJl , then so is Pϕ. Proof. By Proposition 5.3.1,

76


HJ (Pϕ) = iHJ (DDϕ) − iHJ (DDϕ) 2ϕ J (Dϕ)) + D2 ϕ − iD(HJ (Dϕ)) +D = iD(H J + D2 ϕ − iD(D(H J ϕ) − iDϕ) + D 2ϕ ϕ) + iDϕ) = iD(D(H J J ϕ) ϕ) − iDD(H = iDD(H

= P(HJ ϕ) . The second part of the assertion follows by Equation (3.3.1).

Remark 5.3.3 Proposition 5.3.1 and Equation (3.3.1) imply that the vector bundles QJl are not invariant under the Dirac operators D and D. In the remainder of this section, we want to describe the operator P0 in more detail. For this, we prove the following lemmas. The first one shows that the curvature part of the Weitzenb¨ ock formula for P is related to the symplectic scalar curvature R of ∇ with respect to J. Lemma 5.3.4 If ϕ ∈ Γ(QJ0 ), then 2n

1 Q i R (Jej , ej )ϕ = Rϕ . 2 j=1 4 2n

Jej · ek · RQ (ej , ek )ϕ =

j,k=1

Proof. Let ϕ ∈ Γ(QJ0 ). Then, by Corollary 3.3.5, RQ (ej , ek )ϕ is also a section of QJ0 . Applying Corollary 3.3.7, we get 2n

2n


j,k=1

ek · Jej · RQ (ej , ek )ϕ

j,k=1

=i

2n

ek · ej · RQ (ej , ek )ϕ

j,k=1

=

2n i (ej · ek − ek · ej ) · RQ (ek , ej )ϕ 2 j,k=1

=

2n 1 ω(ej , ek )RQ (ek , ej )ϕ 2 j,k=1

1 Q R (Jej , ej )ϕ . 2 j=1 2n

=

To see the second equation, first observe that, by Proposition 3.2.9 and Corollary 3.3.7,


77

i ej · R(X, Y )(Jej ) · ϕ 2 j=1 2n

RQ (X, Y )ϕ =

=−

2n i S(X, Y, Jej , ek )ej · Jek · ϕ 2 j,k=1

=

1 2

2n

S(X, Y, Jej , ek )ej · ek · ϕ .

j,k=1

We use Proposition 2.3.4 and continue with RQ (X, Y )ϕ =

2n 1 S(X, Y, Jej , ek )(ej · ek − ek · ej ) · ϕ 4 j,k=1

=−

2n i ω(ej , ek )S(X, Y, Jej , ek )ϕ 4 j,k=1

= Hence

2n

i 4

2n

S(Y, X, ej , ej )ϕ .

j=1

RQ (Jej , ej )ϕ =

j=1

2n i S(ej , Jej , ek , ek )ϕ , 4 j,k=1

which, by Equation (2.3.1), gives the desired equation.

Lemma 5.3.5 If ϕ ∈ Γ(QJ0 ), then 2n

Jej · ek · ∇T(ej ,ek ) ϕ = −∇T ϕ .

j,k=1

Proof. Proceeding as in the proof of Lemma 5.3.4, one sees 2n j,k=1

1 ∇T(Jej ,ej ) ϕ = −∇T ϕ 2 j=1 2n

Jej · ek · ∇T(ej ,ek ) ϕ =

for ϕ ∈ Γ(QJ0 ). Proposition 5.3.6 For any ϕ ∈ Γ(QJ0 ), 1 Pϕ = ∆Q ϕ − Rϕ − ∇JT ϕ + i∇T ϕ . 4

Proof. This is a consequence of Corollary 5.2.7 and Lemmas 5.3.4 and 5.3.5.

78


Proposition 5.3.7 Let ϕ ∈ Γ(QJ0 ) be such that ∇JX ϕ = i∇X ϕ for all X ∈ Γ(T M ). Then

(5.3.1)

Pϕ = 0 .

Proof. According to Lemma 5.3.4 and Proposition 5.3.6, we have to show that i Q R (Jej , ej )ϕ . 2 j=1 2n

∆Q ϕ = −

(5.3.2)

By Proposition 5.2.3 and 2n

RQ (Jej , ej )ϕ =

j=1

2n ∇Jej ∇ej ϕ − ∇ej ∇Jej ϕ − ∇[Jej ,ej ] ϕ j=1

=−

2n

i∇Jej ∇Jej ϕ + i∇ej ∇ej ϕ + ∇[Jej ,ej ] ϕ

j=1

= −2i

2n

∇ej ∇ej ϕ −

j=1

2n

∇[Jej ,ej ] ϕ ,

j=1

Equation (5.3.2) is equivalent to 2n

i ∇[Jej ,ej ] ϕ . 2 j=1 2n

div(ej )∇ej ϕ = i∇T ϕ −

j=1

Now observe that 1 1 [Jej , ej ] = (T(ej , Jej ) − [Jej , ej ]) 2 j=1 2 j=1 2n

T−

2n

1 ∇ej (Jej ) − ∇Jej ej 2 j=1 2n

=

=

2n

∇ej (Jej )

j=1

and 2n

div(ej )ej =

j=1

2n

ω(∇ek ej , Jek )ej

j,k=1

=

2n

ω(∇ek (Jek ), ej )ej

j,k=1

=

2n k=1

J(∇ek (Jek )) .

(5.3.3)


79

This, together with Equation (5.3.1), yields Equation (5.3.3), and we are done. Remark 5.3.8 Equation (5.3.1) is a generalized Cauchy–Riemann equation and hence gives a notion of a “holomorphic” symplectic spinor field. It is, however, questionable whether, for n > 1 and generic almost complex structure J, non-trivial local holomorphic spinor fields exist.

6 The K¨ ahler Case

In this chapter, we direct our considerations to K¨ ahler manifolds seen as symplectic manifolds, where the symplectic structure is the K¨ ahler form. In particular, we discuss the operator P on K¨ ahler manifolds of constant holomorphic sectional curvature and derive lower bound estimates for its eigenvalues in this situation. The chapter concludes with the computation of the spectrum of P on CP 1 . Focusing to the kernel of P, this leads to an interesting number theoretical equation.

6.1 The Operator P on K¨ ahler Manifolds As before, let (M, ω) be a symplectic manifold of dimension 2n and let J be an ω-compatible almost complex structure. Throughout this chapter, we let ∇ be the Levi–Civita connection of the Riemannian metric g associated to J and assume that ∇J = 0, which means that (M, ω, J) is K¨ ahler. Since ∇ is then also symplectic, after fixating a metaplectic structure P on (M, ω), and therewith the we can construct the symplectic Dirac operators D and D operator P. Proposition 6.1.1 If (M, ω, J) is a K¨ ahler manifold and ∇ is its Levi–Civita connection, then P is formally self-adjoint. are Proof. We know from Theorem 4.5.3 that, on a K¨ ahler manifold, D and D formally self-adjoint. Hence also P has this property. As given in the next proposition, the difference between P and the spinor Laplacian ∆Q expresses in curvature terms. Proposition 6.1.2 On a K¨ ahler manifold,

82


Pϕ = ∆Q ϕ + i

2n


j,k=1

holds true for all ϕ ∈ Γ(Q). Proof. This follows directly from Corollary 5.2.7.

The following equation looks very similar to the famous Weitzenb¨ ock formula for the square of the Riemannian Dirac operator (cf. [38, 43]). As mentioned in Section 2.3, in the present situation, the symplectic scalar curvature R coincides with the scalar curvature of the Riemannian metric g. Proposition 6.1.3 On a K¨ ahler manifold, 1 Pϕ = ∆Q ϕ − Rϕ 4 for any ϕ ∈ Γ(QJ0 ). Proof. This is an immediate consequence of Proposition 5.3.6.

The last proposition can also be obtained by using the following two lemmas. These lemmas are of their own interest, since they link the Ricci tensor with the symplectic spinor bundle. In Riemannian spin geometry, this is a well known fact (cf. [19]). However, in our context, it only occurs in a very special situation. Let Ric denote the Ricci tensor of g understood as a (1, 1)-tensor field on M . Lemma 6.1.4 For any symplectic spinor field ϕ on a K¨ ahler manifold, 2n

R (ej , Jej )ϕ = i Q

j=1

2n

Ric(ej ) · ej · ϕ .

j=1

Proof. By Lemma 2.3.6, we have ω(Ric(X), JY ) = ric(X, Y ) =

2n

S(ej , X, Y, Jej ) .

j=1

Using Propositions 2.3.3 and 2.3.4, it follows that 1 (S(ej , X, Y, Jej ) − S(Jej , X, Y, ej )) 2 j=1 2n

ω(Ric(X), JY ) =

1 (S(X, ej , Jej , Y ) + S(Jej , X, ej , Y )) 2 j=1 2n

=−

6.1 The Operator P on K¨ ahler Manifolds

=

1 S(ej , Jej , X, Y ) 2 j=1

=

1 S(ej , Jej , JX, JY ) , 2 j=1

83

2n

2n

which gives the known formula (see [36]) 1 R(ej , Jej )(JX) . 2 j=1 2n

Ric(X) =

Together with Proposition 3.2.9, this yields 2n

RQ (ej , Jej )ϕ =

j=1

2n i ek · R(ej , Jej )(Jek ) · ϕ 2 j,k=1

=

2n i R(ej , Jej )(Jek ) · ek · ϕ 2 j,k=1

=i

2n

Ric(ek ) · ek · ϕ .

k=1

Lemma 6.1.5 For any symplectic spinor field ϕ on a K¨ ahler manifold, 2n

i Ric(ej ) · Jej · ϕ = − Rϕ . 2 j=1

Proof. By Propositions 2.3.4 and 2.3.9, ω(Ric(X), Y ) = ω(X, Ric(Y )) for all X, Y ∈ Γ(T M ). This yields 2n

Jej · Ric(ej ) · ϕ = −

j=1

2n

ω(Ric(ej ), ek )Jej · Jek · ϕ

j,k=1

=−

2n

ω(ej , Ric(ek ))Jej · Jek · ϕ

j,k=1

=−

2n k=1

Consequently,

Ric(ek ) · Jek · ϕ .

84


iRϕ = i

2n

ω(Ric(ej ), Jej )ϕ

j=1

=

2n

Jej · Ric(ej ) · ϕ −

j=1

= −2

2n

Ric(ej ) · Jej · ϕ

j=1 2n

Ric(ej ) · Jej · ϕ ,

j=1

which concludes the proof.

We now want to specialize Proposition 6.1.2 to the case of K¨ ahler manifolds of constant holomorphic sectional curvature. Below we describe those manifolds by means of the symplectic curvature tensor S. Definition 6.1.6 Let h ∈ C ∞ (M ). A K¨ ahler manifold (M, ω, J) is said to be of constant holomorphic sectional curvature h if S(X, JX, X, X) = h (ω(X, JX))2 for all X ∈ Γ(T M ). For a proof of the next proposition, see e.g. [36]. Proposition 6.1.7 A K¨ ahler manifold (M, ω, J) is of constant holomorphic sectional curvature h if and only if S(X1 , X2 , X3 , X4 ) =

h (ω(X1 , X3 )ω(X2 , JX4 ) + ω(X1 , X4 )ω(X2 , JX3 ) 4 −ω(X2 , X3 )ω(X1 , JX4 ) − ω(X2 , X4 )ω(X1 , JX3 ) +2ω(X1 , X2 )ω(X3 , JX4 ))

for all X1 , X2 , X3 , X4 ∈ Γ(T M ).

In other words, (M, ω, J) is of constant holomorphic sectional curvature h if and only if its Levi–Civita connection ∇ is reducible in the sense of Definition 2.3.12 with h b= g. 4 If this is the case, in view of Proposition 2.3.13, (M, ω, J) is an Einstein manifold with scalar curvature R = n(n + 1)h and hence, for n ≥ 2, as for the Riemannian sectional curvature, the function h is forced to be constant.

6.1 The Operator P on K¨ ahler Manifolds

85

Proposition 6.1.8 On a K¨ ahler manifold of constant holomorphic sectional curvature h, it holds Pϕ = ∆Q ϕ +

h n(n − 1)ϕ − 2h(HJ )2 ϕ 4

(6.1.1)

for all ϕ ∈ Γ(Q). Proof. According to Propositions 3.2.9 and 6.1.7, we have i

2n


j,k=1

=

2n

1 2

S(ej , ek , el , Jem )Jej · ek · Jel · em · ϕ

j,k,l,m=1

=−

2n h (Jej · Jek · ej · ek + Jej · Jek · ek · ej 8 j,k=1

+ej · ek · ek · ej + ej · ek · ej · ek + 2ej · ej · ek · ek ) · ϕ =−

2n h (Jej · ej · Jek · ek + Jej · Jek · ek · ej + 4ej · ej · ek · ek 8 j,k=1

−iω(Jek , ej )Jej · ek − 3iω(ek , ej )ej · ek ) · ϕ =−

2n h (Jej · ej · Jek · ek + Jej · Jek · ek · ej + 4ej · ej · ek · ek ) · ϕ 8 j,k=1

h +i Jej · ej · ϕ . 4 j=1 2n

Using 2n

1 i (Jej · ej − ej · Jej ) · ϕ = ω(ej , Jej )ϕ = inϕ , 2 j=1 2 j=1 2n

Jej · ej · ϕ =

j=1

2n

we arrive at i

2n


j,k=1

=

h h 2 2n ϕ − 16(HJ )2 ϕ − nϕ 8 4 h n(n − 1)ϕ − 2h(HJ )2 ϕ . 4

This together with Proposition 6.1.2 implies the desired equation.

Example 6.1.9 The flat complex torus TCn = Cn /Z2n has vanishing holomorphic sectional curvature. Thus, by Proposition 6.1.8,

86


P = ∆Q for any metaplectic structure on TCn , which gives that the kernel of P consists of all parallel symplectic spinor fields. Hence, with respect to the canonical metaplectic structure, this kernel is nothing else but the space S(Rn ).

6.2 Lower Bound Estimates We apply Proposition 6.1.8 to obtain estimates for the eigenvalues of P. Keep in mind that, by Proposition 6.1.1, all eigenvalues of P are real. Proposition 6.2.1 Let (M, ω, J) be a 2n-dimensional closed K¨ ahler manifold of constant holomorphic sectional curvature h and suppose that the function h is constant and non-positive. Then any eigenvalue λ of P satisfies λ≥

h n(n − 1) . 4

Proof. Let λ ∈ R be an eigenvalue of P and let ϕ be a non-vanishing symplectic spinor field such that Pϕ = λϕ . Taking the L2 -product of Equation (6.1.1) with ϕ and using Lemma 3.3.3, one gets λϕ2 = (∆Q ϕ, ϕ) + = ∇ϕ2 + Thus

λ−

h n(n − 1)ϕ2 − 2h (HJ )2 ϕ, ϕ 4

h n(n − 1)ϕ2 − 2hHJ ϕ2 . 4

h n(n − 1) ϕ2 = ∇ϕ2 − 2hHJ ϕ2 . 4

Since, by assumption, h ≤ 0, this yields the asserted inequality.

Taking advantage of the splitting of Q described in Section 3.3 and the corresponding splitting of P derived in Section 5.3, the above estimate can be refined in the following way. Proposition 6.2.2 Let (M, ω, J) be a 2n-dimensional closed K¨ ahler manifold of constant holomorphic sectional curvature h and suppose that h is constant. Then, for all l ∈ N0 and any eigenvalue λ of Pl , one has h λ ≥ − (n(n + 1) + 8l(l + n)) . 4

6.3 The Spectrum of P on CP 1

87

Proof. Let ϕ ∈ Γ(QJl ) be an eigensection of Pl with eigenvalue λ ∈ R. By Equation (3.3.1), n ϕ. HJ ϕ = − l + 2 Therefore, as in the proof of Proposition 6.2.1, one obtains h n 2 λϕ2 = ∇ϕ2 + n(n − 1)ϕ2 − 2h l + ϕ2 4 2 h = ∇ϕ2 − (n(n + 1) + 8l(l + n))ϕ2 , 4

which gives the assertion.

As noted in the preceding section, for n ≥ 2, the assumption in the above propositions that the function h is constant is automatically satisfied. The above estimates indicate that, for h > 0, in general, one can not found a common lower bound for the eigenvalues of P. That this is actually true is clarified by the results in the next section.

6.3 The Spectrum of P on CP 1 For the computation of the spectrum of the operator P on the complex projective space CP 1 , we use tools of harmonic analysis. For details, we refer to [48]. First we realize CP 1 as follows. We consider the canonical action of the compact Lie group w1 −w2 2 2 SU(2) = : w1 , w2 ∈ C and |w1 | + |w2 | = 1 w2 w1 on CP 1 . This action is transitive and has U(1) as isotropy group of the point [1:0] ∈ CP 1 , where U(1) is understood as z0 : z ∈ C and |z| = 1 . U(1) = 0z Hence CP 1 is the homogeneous space SU(2)/U(1). We endow the Lie algebra su(2) of SU(2) with the inner product b defined by 1 b(X, Y ) = − Tr(XY ) , 2 which is a negative multiple of the Killing form of su(2). Then the matrices i 0 0 1 0 i E0 = , E1 = , E2 = 0 −i −1 0 i 0

88


form an orthonormal basis of su(2) with E0 ∈ u(1). Furthermore, we have the commutation relations [E0 , E1 ] = 2E2 ,

[E1 , E2 ] = 2E0 ,

[E2 , E0 ] = 2E1 .

(6.3.1)

In particular, the orthogonal complement w of u(1) ⊂ su(2) satisfies [w, w] = u(1) . That way we have described CP 1 as symmetric space. Let J denote the complex structure on w determined by JE1 = E2 and define a symplectic structure ω on w by ω(X, Y ) = b(JX, Y ) . Clearly, ω(E1 , E2 ) = 1 . Let J be the left-invariant almost complex structure on CP 1 induced by J and let ω be the left-invariant symplectic structure on CP 1 induced by ω. More concretely, let κ : U(1) → GL(w) be the isotropy representation of SU(2)/U(1), i.e. the restriction of the adjoint representation of SU(2) to U(1) acting on w, and identify the tangent bundle T CP 1 of CP 1 with the vector bundle SU(2) ×κ w associated to SU(2) via κ by d a exp(tX)[0 : 1] ∈ T CP 1 . [a, X] ∈ SU(2) ×κ w → dt t=0 Then J and ω are given by J[a, X] = [a, JX]

and ω([a, X], [a, Y ]) = ω(X, Y ) .

One readily sees that (ω, J) is the usual K¨ ahler structure on CP 1 . Hence the associated Riemannian metric g, which is given by g([a, X], [a, Y ]) = b(X, Y ) , is exactly the Fubini–Study metric. According to Example 3.1.7, the projective space CP 1 admits a unique metaplectic structure. We now construct this bundle. For this, we identify the real vector space w with C, seen as R2 , via the isomorphism 0 w ∈w. w ∈ C → −w 0


89

Then {E1 , E2 } is the standard basis and the complex structure J on w is the multiplication with i. Equation (6.3.1) implies that 0 −2 . κ∗ (E0 ) = 2 0 In particular, the image of κ is U(1) = Sp(1, R) ∩ O(2). From Section 1.5, we know that the Lie algebra u(1) of the inverse image U(1) of U(1) under the double covering ρ : Mp(1, R) → Sp(1, R) is spanned by E1 · E1 + E2 · E2 . Moreover, by Equation (1.3.2), 0 −2 . ρ∗ (E1 · E1 + E2 · E2 ) = 2 0 Thus a lift κ ˆ : U(1) → U(1) of the isotropy representation κ is determined by requiring that (6.3.2) κ ˆ ∗ (E0 ) = E1 · E1 + E2 · E2 . Since, with the above identifications, the symplectic frame bundle of CP 1 is R = SU(2) ×κ Sp(1, R) , we get that the metaplectic structure of CP 1 is P = SU(2) ×κˆ Mp(1, R) , where the covering map FP : P → R is given by FP ([a, q]) = [a, ρ(q)] . Consequently, the associated symplectic spinor bundle is Q = SU(2) ×m◦ˆκ L2 (R) and the Clifford multiplication writes as [a, X] · [a, f ] = [a, X · f ] . Furthermore, the splitting of Q into the subbundles Ql , l ∈ N0 , corresponds to the decomposition of the unitary representation m ◦ κ ˆ = u◦κ ˆ into the representations ˆ : U(1) → GL(Wl ) ul ◦ κ and hence Ql = SU(2) ×ul ◦ˆκ L2 (R) . Here, u and ul are as given in Section 1.5. In particular, Wl is the onedimensional complex vector space spanned by the Hermite function hl . In view of Proposition 1.4.5 and Equations (1.5.4) and (6.3.2),

90


(m ◦ κ ˆ )∗ (E0 )f = m∗ (E1 · E1 + E2 · E2 )f = −i(E1 · E1 + E2 · E2 ) · f

(6.3.3)

= −2iH0 f for any f ∈ S(R). Together with Equation (1.5.5), this implies (ul ◦ κ ˆ )∗ (E0 )hl = (2l + 1)ihl , which says that the weight of the representation ul ◦ κ ˆ is 2l + 1. Next observe that the Levi–Civita connection of CP 1 is induced by the canonical connection of the symmetric space SU(2)/U(1), and hence so is the spinor derivative. Thus, identifying Γ(Q) with the spaces of U(1)-equivariant maps ϕ : SU(2) → L2 (R), the spinor derivative is given by ∇[a,X] ϕ = X(ϕ)(a) for [a, X] ∈ T CP 1 . Here, X(ϕ) means the map X(ϕ) : SU(2) → L2 (R) defined by d ϕ(a exp(tX)) X(ϕ)(a) = . dt t=0 on CP 1 , which are obviously For the symplectic Dirac operators D and D SU(2)-invariant, we then have Lemma 6.3.1 For any symplectic spinor field ϕ on CP 1 , Dϕ = E1 · E2 (ϕ) − E2 · E1 (ϕ) and

= E1 · E1 (ϕ) + E2 · E2 (ϕ) . Dϕ

Proof. Since {E1 , E2 } is a unitary basis of w, this follows again from Lemma 4.1.3. We now relate our operator P to the Casimir operator Ω = E02 + E12 + E22 of su(2). Proposition 6.3.2 For any symplectic spinor field ϕ on CP 1 , Pϕ = −Ω(ϕ) − 12 H02 ◦ ϕ .


91

Proof. Let ϕ : SU(2) → L2 (R) be a symplectic spinor field on CP 1 . Then ϕ is U(1)-equivariant and it follows by Equation (6.3.3) that d ϕ(a exp(tE0 )) E0 (ϕ)(a) = dt t=0 d (m ◦ κ ˆ )(exp(−tE0 ))ϕ(a) = dt t=0 = −(m ◦ κ ˆ )∗ (E0 )ϕ(a) = 2i H0 (ϕ(a)) . Using this, Lemma 6.3.1, and X(Y · ϕ) = Y · X(ϕ) for X, Y ∈ w, we conclude that − DD)ϕ = E1 · E1 (E1 · E2 (ϕ)) + E2 · E2 (E1 · E2 (ϕ)) (DD −E1 · E1 (E2 · E1 (ϕ)) − E2 · E2 (E2 · E1 (ϕ)) −E1 · E2 (E1 · E1 (ϕ)) − E1 · E2 (E2 · E2 (ϕ)) +E2 · E1 (E1 · E1 (ϕ)) + E2 · E1 (E2 · E2 (ϕ)) = (E1 · E1 + E2 · E2 ) · (E1 (E2 (ϕ)) − E2 (E1 (ϕ))) +(E2 · E1 − E1 · E2 ) · E12 (ϕ) + E22 (ϕ) = i E12 (ϕ) + E22 (ϕ) + (E1 · E1 + E2 · E2 ) · [E1 , E2 ](ϕ) = iΩ(ϕ) − iE02 (ϕ) + 4H0 ◦ E0 (ϕ) = iΩ(ϕ) + 12i H02 ◦ ϕ , which proves the proposition.

Taking into account that the sections of Ql are identified with U(1)equivariant maps ϕ : SU(2) → Wl , we obtain Corollary 6.3.3 On CP 1 , for all l ∈ N0 and any ϕ ∈ Γ(Ql ), one has Pl ϕ = −Ω(ϕ) − 3(2l + 1)2 ϕ . Proof. This is a consequence of Proposition 6.3.2 and Equation (1.5.5).

Therefore, in order to compute the spectrum of each operator Pl , it suffices to determine the spectrum of the Casimir operator acting on Γ(Ql ), and this can be done by means of the Frobenius reciprocity. For convenience, we now formulate this result specialized to our situation. Let Vk for k ∈ N0 denote the complex vector space of homogeneous polynomials of order k in two complex variables z1 and z2 and let

92


τk : SU(2) → GL(Vk ) be the representation of SU(2) on Vk induced by the canonical action of SU(2) on C2 . One checks that, for the basis {pk,0 , . . . , pk,k } of Vk defined by pk,j (z1 , z2 ) = z1k−j z2j , it holds (τk )∗ (E0 )pk,j = i(2j − k)pk,j . Thus the representation τk has the weights −k, −k + 2, . . . , k − 2, k. As is well known (cf.[53]), the representations τk are irreducible and pairwise inequivalent and represent all possible irreducible representations of SU(2). Let ν : U(1) → GL(W) be a one-dimensional complex representation of U(1) and let HomU(1) (Vk , W) denote the space of all U(1)-equivariant homomorphisms φ : Vk → W. We embed Vk ⊗ HomU(1) (Vk , W) into the space Γ(SU(2) ×ν W) of sections of the line bundle SU(2) ×ν W by assigning to p ⊗ φ ∈ Vk ⊗ HomU(1) (Vk , W) the U(1)-equivarient map a ∈ SU(2) → φ τk a−1 p ∈ W . Then we have Theorem 6.3.4 (Frobenius reciprocity) (1) The space Γ(SU(2) ×ν W) decomposes into the direct sum ∞

Vk ⊗ HomU(1) (Vk , W) .

k=0

(2) If HomU(1) (Vk , W) is non-trivial, then Vk ⊗ HomU(1) (Vk , W) is an eigenspace of the Casimir operator Ω with eigenvalue 1 − (k + 1)2 . Now we can prove the following theorem. For a generalization of this result to all complex projective spaces CP n with odd n, see [52]. Theorem 6.3.5 On the complex projective space CP 1 , the operator Pl has the eigenvalues λl,j = 4(l + j + 1)2 − 3(2l + 1)2 − 1 for j = 0, 1, 2, . . ., where the multiplicity of λl,j is ml,j = 2(l + j + 1) .


93

Proof. We apply Theorem 6.3.4 for ν = ul ◦ κ ˆ . Thus we have to check for which k ∈ N0 the space HomU(1) (Vk , Wl ) is non-trivial. By Schur’s lemma, this is ˆ is a weight of τk , i.e., if the case if and only if the weight of ul ◦ κ 2l + 1 ∈ {−k, −k + 2, . . . , k − 2, k} . Hence HomU(1) (Vk , Wl ) is non-trivial if and only if k = 2l + 2j + 1 for j ∈ N0 . For such k, by Corollary 6.3.3 and Theorem 6.3.4, Vk ⊗ HomU(1) (Vk , Wl ) is an eigenspace of Pl with eigenvalue 4(l + j + 1)2 − 3(2l + 1)2 − 1 and dimC Vk ⊗ HomU(1) (Vk , Wl ) = dimC Vk = 2(l + j + 1) .

This yields the assertion.

The spectrum spec(P) of P is now the set of all λl,j , i.e. the set of the values of the function F (x, y) = 4(x + y + 1)2 − 3(2x + 1)2 − 1 for non-negative integer arguments. Note that different pairs (l, j) may give the same eigenvalue λl,j .

3000 2000 1000 0 -1000 -2000 -3000 20 10 10

0 -10 -20

-15

-10

-5

0

15

20

5

As the picture of the graph of F illustrates, spec(P) is neither bounded from below nor from above. However, we see that λl,j ≥ λl,0 = −8l2 − 4l for all l, j ∈ N0 . This estimate is better than the estimate given in Proposition 6.2.2. Indeed, since the symplectic curvature tensor S of CP 1 writes as

94


S([a, X1 ], [a, X2 ], [a, X3 ], [a, X4 ]) = b([X1 , X2 ], [X3 , JX4 ]) for [a, X1 ], [a, X2 ], [a, X3 ], [a, X4 ] ∈ T CP 1 , we have S([a, E1 ], J[a, E1 ], [a, E1 ], [a, E1 ]) = b([E1 , E2 ], [E1 , E2 ]) = 4 , which shows that CP 1 is of constant holomorphic sectional curvature 4. Consequently, Proposition 6.2.2 would only give λl,j ≥ −8l2 − 8l − 2 . In the picture below, it is visualized for which indices l and j the eigenvalues λl,j are positive and for which they are negative. l 5

*

*

4

*

*

3

*

*

*

*

2

*

*

*

*

*

*

* * positive eigenvalues

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

*

negative eigenvalues 1

*

*

*

*

*

*

*

*

0

*

*

*

*

*

*

*

*

0

1

2

3

4

5

6

7

k

-1

-2 -2

-1

8

Clearly, the two regions are separated by the hyperbola 4(x + y + 1)2 − 3(2x + 1)2 = 1 . The following proposition results immediately from Theorem 6.3.5. Proposition 6.3.6 On CP 1 , the kernel KerPl of Pl is non-trivial if and only if there is an non-negative integer j such that 4(l + j + 1)2 − 3(2l + 1)2 = 1 .

(6.3.4)

If this is the case, then dimC KerPl = 2(l + j + 1) .


95

In the remainder of this section, we solve Equation (6.3.4) within the nonnegative integers. Obviously, (l, j) = (0, 0) is a solution of this equation. Furthermore, any solution (l, j) ∈ N20 of Equation (6.3.4) yields a solution (k, m) ∈ N20 of the equation k 2 − 3m2 = 1 .

(6.3.5)

Conversely, as one readily checks, any solution (k, m) ∈ N20 of Equation (6.3.5) such that m is odd gives a solution (l, j) ∈ N20 of Equation (6.3.4). Let us consider first Equation (6.3.5). This equation is a certain Diophantine equation called Pell’s equation. From the general theory of Pell’s equation (cf. e.g. [40]), one knows that, since the number 3 occuring in our equation is not a square, this equation has infinitely many solutions. Moreover, one has Proposition 6.3.7 The solutions (ki , mi ) ∈ N20 , i = 1, 2, 3, . . ., of Equation (6.3.5) are determined by √ i √ ki + mi 3 = 2 + 3 . Remark 6.3.8 To obtain the pairs (ki , mi ) from the last equation, one has to compute the power and then to compare the rational and purely irrational terms. For example, from √ 2 √ √ k2 + m2 3 = 2 + 3 = 7 + 4 3 ,

one gets (k2 , m2 ) = (7, 4).

We now have to detect those solutions (ki , mi ) for which mi is odd. Obviously, mi is odd if and only if ki is even. From √ √ √ ki + mi 3 = ki−1 + mi−1 3 2 + 3 √ = (2ki−1 + 3mi−1 ) + (ki−1 + 2mi−1 ) 3 , we see that ki = 2ki−1 + 3mi−1 and mi = ki−1 + 2mi−1 . Therefore, mi is odd if and only if mi−1 even. Thus, since (k1 , m1 ) = (2, 1), the solutions of Equation (6.3.5) which yield a solution of Equation (6.3.4) are (k2i+1 , m2i+1 ), i = 0, 1, 2, . . ., and we obtain

96


Proposition 6.3.9 The solutions (li , ji ) ∈ N20 , i = 0, 1, 2, . . ., of Equation (6.3.4) are determined by √ √ 2i+1 2(li + ji + 1) + (2li + 1) 3 = 2 + 3 . Proof. We relate (li , ji ) to (k2i+1 , m2i+1 ) by k2i+1 = 2(li + ji + 1)

and m2i+1 = 2li + 1 .

The assertion then follows from Proposition 6.3.7.

Corollary 6.3.10 The solutions (li , ji ) ∈ N20 , i = 0, 1, 2, . . ., of Equation (6.3.4) are given by (l0 , j0 ) = (0, 0) and (li , ji ) = (11li−1 + 4ji−1 + 7, 8li−1 + 3ji−1 + 5) for i = 1, 2, 3, . . .. Proof. From Proposition 6.3.9, we know that, for i = 1, 2, 3, . . ., √ 2(li + ji + 1) + (2li + 1) 3 √ 2 √ 2i−1 2+ 3 = 2+ 3 √ √ = 2(li−1 + ji−1 + 1) + (2li−1 + 1) 3 7 + 4 3 √ = 38li−1 + 14ji−1 + 26 + (22li−1 + 8ji−1 + 15) 3 . It follows that li + ji + 1 = 19li−1 + 7ji−1 + 13 , 2li + 1 = 22li−1 + 8ji−1 + 15 , which implies li = 11li−1 + 4ji−1 + 7 , ji = 8li−1 + 3ji−1 + 5 . Since, also by Proposition 6.3.9, (l0 , j0 ) = (0, 0), this proves the corollary.

7 Fourier Transform for Symplectic Spinors

This chapter presents a globally defined Fourier transform F J for symplectic spinor fields, which was worked out by Klein in [34]. We show how this transform behaves with respect to Clifford multiplication and spinor derivative. This has consequences for the symplectic Dirac operators. Applications in mathematical physics can be found in [29] and [34]. In particular, there it is explained how the quantization procedure of [41] can be expressed by means of the Fourier transform F J .

7.1 Definition of the Transform The following proposition states that the Fourier transform F on L2 (Rn ) is an intertwining operator of the restriction u of the metaplectic representation m to U(n). Proposition 7.1.1 For all q ∈ U(n), F ◦ u(q) = u(q) ◦ F . Proof. Since U(n) is connected and 0 = ρ−1 ∗ (J0 ) lies in the center of the lie algebra u(n) of U(n), exp(t0 )q = q exp(t0 ) and hence u (exp(t0 )) ◦ u(q) = u(q) ◦ u (exp(t0 )) hold true for all q ∈ U(n) and t ∈ R. Together with Lemma 1.5.4, this yields the assertion. Now, fixing an ω-compatible almost complex structure J on M and writing the symplectic spinor bundle as Q = P J ×u L2 (Rn ), Proposition 7.1.1 allows to introduce a Fourier transform F J for symplectic spinors as follows.

98


Definition 7.1.2 Let F J : Q → Q be the endomorphism on Q defined by F J ([p, f ]) = [p, Ff ] for p ∈ P J and f ∈ L2 (Rn ).

7.2 Basic Properties Proposition 7.2.1 For any X ∈ Γ(T M ) and any ϕ ∈ Γ(Q), F J (JX · ϕ) = X · F J ϕ .

Proof. This follows from Lemma 1.4.1(3).

Proposition 7.2.2 Let ∇ be a symplectic connection on M which satisfies ∇J = 0. Then the Fourier transform F J commutes with the symplectic spinor derivative, i.e. F J (∇X ϕ) = ∇X (F J ϕ) for all X ∈ Γ(T M ) and ϕ ∈ Γ(Q). Proof. Let s = (e1 , . . . , e2n ) be a unitary frame. Writing ϕ = [s, ϕs ], we have i Jej · ∇X ej · ϕ 2 j=1 2n

∇X ϕ = [s, X(ϕs )] − by Proposition 3.2.6. Obviously,

F(X(ϕs )) = X(F(ϕs )) . Using Proposition 7.2.1, we conclude i J F (Jej · ∇X ej · ϕ) 2 j=1 2n

F J (∇X ϕ) = [s, F(X(ϕs ))] − = [s, X(F(ϕs ))] −

i ej · F J (∇X ej · ϕ) 2 j=1

= [s, X(F(ϕs ))] +

i ej · J(∇X ej ) · F J ϕ . 2 j=1

2n

2n

Since, by assumption, J(∇X Y ) = ∇X (JY ), we obtain

7.3 Symmetry of the Spectra of D and D

99

i ej · ∇X (Jej ) · F J ϕ 2 j=1 2n

F J (∇X ϕ) = [s, F(X(ϕs ))] + = ∇X (F J ϕ) .

Finally, we have Proposition 7.2.3 The Fourier transform F J leaves the subbundles QJl of Q invariant, F J (QJl ) = QJl . Moreover, for any l ∈ N0 and ϕ ∈ Γ(QJl ), F J ϕ = i−l ϕ .

Proof. This is a consequence of Lemma 1.5.3.

! 7.3 Symmetry of the Spectra of D and D The next proposition shows that the Fourier transform F J is the link between the two symplectic Dirac operators D and D. are defined with respect to a symplectic conProposition 7.3.1 If D and D nection ∇ satisfying ∇J = 0, then the relations , D ◦ FJ = FJ ◦ D

◦ F J = −F J ◦ D D

and

P ◦ FJ = FJ ◦ P

hold true. Proof. Applying Lemma 4.1.3 and Propositions 7.2.1 and 7.2.2, we see DF J ϕ = −

2n

Jej · ∇ej (F J ϕ)

j=1

=−

2n

Jej · F J ∇ej ϕ

j=1

=

2n

F J ej · ∇ej ϕ

j=1

= F J Dϕ and

100


Jϕ = DF

2n

ej · ∇ej (F J ϕ)

j=1

=

2n

ej · F J ∇ej ϕ

j=1

=

2n

F J Jej · ∇ej ϕ

j=1

= −F J Dϕ for any unitary frame (e1 , . . . , e2n ). The third identity follows from the first two. Indeed, J J ϕ) PF J ϕ = i(DDF ϕ − DDF + DF J Dϕ) J Dϕ = i(DF

+ F J DDϕ) = i(−F J DDϕ = F J Pϕ . Proposition 7.3.1 immediately implies Corollary 7.3.2 If the chosen symplectic connection ∇ satisfies ∇J = 0, then D ◦ (F J )2 = −(F J )2 ◦ D

and

◦ (F J )2 = −(F J )2 ◦ D . D

For the spectra of the symplectic Dirac operators, we obtain have the same specCorollary 7.3.3 Suppose that ∇J = 0. Then D and D trum. Moreover, the spectrum is symmetric with respect to the origin. Proof. Apply Proposition 7.3.1.

8 Lie Derivative and Quantization

The aim of this supplementary chapter is to present new aspects of quantization theory related to symplectic spinors. For background information, we refer to the extensive literature on quantization, for instance the books [15, 51] or survey articles such as [4, 23, 33]. First we describe the construction of a Lie derivative for symplectic spinor fields in the direction of a symplectic vector field. This clarifies Kostant’s remark in [37], §5.5 that Hamiltonian vector fields clearly operate as Lie differentiations on smooth symplectic spinor fields. Second we explain how a one parameter subgroup of Mp(n, R) can be assigned to any quadratic Hamiltonian on R2n such that the family of operators associated via the metaplectic representation satisfies the corresponding Schr¨ odinger equation. As is shown in the last section, this equation gives exactly the Lie derivative of a constant symplectic spinor field over R2n in the direction of the Hamiltonian vector field. That means, the Lie derivative of symplectic spinor fields gives rise to a quantization procedure.

8.1 Lie Derivative of Symplectic Spinor Fields We consider a 2n-dimensional symplectic manifold (M, ω) and suppose that a metaplectic structure P on M is fixed. Let θ be a local symplectomorphism of M . Such a mapping gives rise to a local isomorphism θ˜ of the symplectic frame bundle R by ˜ 1 , . . . , en , f 1 , . . . , f n ) = (dθ(e1 ), . . . , dθ(en ), dθ(f 1 ), . . . , dθ(f n )) . θ(e Let θ¯ be a lift of θ˜ to the metaplectic structure P. Then, for any ϕ ∈ Γ(Q), the composition ϕˆ ◦ θ¯ of the Mp(n, R)-equivariant mapping ϕˆ : P → L2 (Rn ) corresponding to ϕ with the local isomorphism θ¯ is also Mp(n, R)-equivariant

102


and thus defines a local symplectic spinor field. We denote this spinor field by (θ¯−1 )∗ ϕ. Let ∇ be a symplectic connection on M . Then it is known that, for any X, Y ∈ Γ(T M ), (8.1.1) ∇(θ−1 )∗ X (θ−1 )∗ Y = (θ−1 )∗ (∇X Y ) , where, as usual, (θ−1 )∗ X denotes the local vector field on M given by −1 (θ )∗ X (p) = dθ−1 (X(θ(p))) . The spinor derivative and the Clifford multiplication have analogous properties. Lemma 8.1.1 For any X ∈ Γ(T M ) and any ϕ ∈ Γ(Q),

and

∇(θ−1 )∗ X (θ¯−1 )∗ ϕ = (θ¯−1 )∗ (∇X ϕ)

(8.1.2)

(θ−1 )∗ X · (θ¯−1 )∗ ϕ = (θ¯−1 )∗ (X · ϕ) .

(8.1.3)

ˆ : P → R2n be the Mp(n, R)-equivariant Proof. Let X ∈ Γ(T M ) and let X mapping corresponding to X according to Equation (3.2.1). Then one easily ˆ ◦ θ¯ corresponds to the local sees that the Mp(n, R)-equivariant mapping X −1 vector field (θ )∗ X. Using this, Equation (8.1.3) follows directly from the definition of the Clifford multiplication. Equation (8.1.2) can be derived from Proposition 3.2.6 and Equations (8.1.1) and (8.1.3). For completeness, we now recall the following. Definition 8.1.2 A vector field X on M is called symplectic if LX ω = 0 . Here, LX denotes the Lie derivative in the direction of X. Lemma 8.1.3 A vector field X is symplectic if and only if its flow (θt ) consists of local symplectomorphisms. Proof. This follows from d d ∗ θt ω = θs∗ θt∗ ω = θs∗ LX ω . dt dt t=s t=0 Lemma 8.1.4 Any Hamiltonian vector field is symplectic.

8.1 Lie Derivative of Symplectic Spinor Fields

103

Proof. Let X be the Hamiltonian vector field of a function u ∈ C ∞ (M ) and let iX denote the interior product of X. Then, by Definition 2.1.7, du = iX ω . Since the symplectic structure ω is a closed form and since the Lie derivative LX applied to forms satisfies the well known relation LX = d ◦ iX + iX ◦ d , we get LX ω = diX ω + iX dω = diX ω = ddu = 0 . Lemma 8.1.5 If X is a symplectic vector field and ∇ is a torsion-free symplectic connection on M , then X(ω(Y, Z)) = ω([X, Y ], Z) + ω(Y, [X, Z])

(8.1.4)

ω(∇Y X, Z) = ω(∇Z X, Y )

(8.1.5)

and for all Y, Z ∈ Γ(T M ). Proof. As is known, (LX ω)(Y, Z) = X(ω(Y, Z)) − ω([X, Y ], Z) − ω(Y, [X, Z]) for arbitrary X, Y, Z ∈ Γ(T M ). This immediately implies Equation (8.1.4). The second equation is an easy consequence of the first one. Let X be a symplectic vector field on M . According to Lemma 8.1.3, the flow (θt ) of X induces a local one parameter group (θ˜t ) of local isomorphisms of R. Let (θ¯t ) denote the lift of (θ˜t ) to P. By means of this local group, the Lie derivative of symplectic spinor fields can now be defined in the usual way. Definition 8.1.6 The Lie derivative of a symplectic spinor field ϕ in the direction of a symplectic vector field X is the symplectic spinor field LX ϕ given by d ¯−1 (θt )∗ ϕ LX ϕ = , dt t=0 where (θ¯t ) is as above. In the following, for brevity, we write ∇X · ϕ =

n

∇fj X · ej − ∇ej X · fj · ϕ ,

j=1

where ∇ is a connection on M and (e1 , . . . , en , f1 , . . . , fn ) is again any symplectic frame. With this, the Lie derivative LX ϕ can be expressed as follows.

104


Proposition 8.1.7 Let ∇ be a torsion-free symplectic connection on M . Then, for any symplectic vector field X on M and any ϕ ∈ Γ(Q), i LX ϕ = ∇X ϕ − ∇X · ϕ . 2 Proof. Let X be a symplectic vector field on M and let (θt ), (θ˜t ), and (θ¯t ) again be the associated local groups. Further, let s = (e1 , . . . , en , f1 , . . . , fn ) : U → R be a symplectic frame and let s : U → P denote a lift of s to the metaplectic structure P. We set st = θ˜t−1 ◦ s ◦ θt : θt−1 (U ) → R and Clearly,

st = θ¯t−1 ◦ s ◦ θt : θt−1 (U ) → P . st = (θt−1 )∗ e1 , . . . , (θt−1 )∗ en , (θt−1 )∗ f1 , . . . , (θt−1 )∗ fn

and st is a lift of st . Let qt : U ∩ θt−1 (U ) → Mp(n, R) be defined by ¯st = ¯sqt . Then At = ρ ◦ qt : U ∩ θt−1 (U ) → Sp(n, R) satisfies st = sAt . Thus, in the standard basis {a1 , . . . , an , b1 , . . . , bn } of R2n , At is given by At ai =

n n ω (θt−1 )∗ ei , fj aj − ω (θt−1 )∗ ei , ej bj j=1

and At bi =

n n ω (θt−1 )∗ fi , fj aj − ω (θt−1 )∗ fi , ej bj . j=1

Since

j=1

j=1

d −1 (θt )∗ Y = LX Y = [X, Y ] dt t=0

for all Y ∈ Γ(T M ), this implies n n d At ai = ω([X, ei ], fj )aj − ω([X, ei ], ej )bj dt t=0 j=1 j=1 and


d At bi dt

= t=0

n

ω([X, fi ], fj )aj −

j=1

n

105

ω([X, fi ], ej )bj

j=1

Hence, by Lemma 1.3.1 and d d At qt = ρ∗ , dt dt t=0 t=0 we obtain n d 1 qt = (ω([X, ei ], fj )bi · aj − ω([X, ei ], ej )bi · bj dt t=0 2 i,j=1 −ω([X, fi ], fj )ai · aj + ω([X, fi ], ej )ai · bj ) . Using Equation (8.1.4), we arrive at n d 1 qt = (ω([X, ej ], fi )ai · bj − ω([X, ej ], ei )bi · bj dt t=0 2 i,j=1 −ω([X, fj ], fi )ai · aj + ω([X, fj ], ei )bi · aj ) .

(8.1.6)

Now let ϕ ∈ Γ(Q). We have ϕ = [s, ϕs ] on U , where ϕs = ϕˆ ◦ s, and, accordingly, (θ¯t−1 )∗ ϕ = [st , ϕˆ ◦ θ¯t ◦ st ] = [st , ϕs ◦ θt ] = [s, m(qt )(ϕs ◦ θt )] on U ∩ θt−1 (U ). Thus, on U , d [s, m(qt )(ϕs ◦ θt )] LX ϕ = dt t=0 % $ d d = s, ϕs ◦ θt + m(qt )ϕs dt dt t=0 t=0 % $ d qt = [s, X(ϕs )] + s, m∗ ϕs . dt t=0 Moreover, by Proposition 1.4.5 and Equation (8.1.6), we compute % $ n d i qt s, m∗ (ω([X, fj ], fi )ei · ej − ω([X, fj ], ei )fi · ej ϕs = dt t=0 2 i,j=1 −ω([X, ej ], fi )ei · fj + ω([X, ej ], ei )fi · fj ) · ϕ

106


=

i ([X, fj ] · ej − [X, ej ] · fj ) · ϕ 2 j=1

=

i (∇X fj · ej − ∇X ej · fj ) · ϕ 2 j=1

n

n

i ∇fj X · ej − ∇ej X · fj · ϕ . 2 j=1 n

− Altogether, we obtain

i i (∇X fj · ej − ∇X ej · fj ) · ϕ − ∇X · ϕ . 2 j=1 2 n

LX ϕ = [s, X(ϕs )] +

Together with Proposition 3.2.6, this yields the assertion.

Recall that the Lie derivative applied to tensor fields satisfies L[X,Y ] = [LX , LY ] for all X, Y ∈ Γ(T M ). Since the commutator of two symplectic vector fields is again a symplectic vector field, one may ask if this relation remains true for the Lie derivative of symplectic spinor fields. In fact, one has Proposition 8.1.8 Let X and Y be symplectic vector fields on M and let ϕ ∈ Γ(Q). Then L[X,Y ] ϕ = [LX , LY ]ϕ . In order to prove this, we first show the following lemmas, where J is an ω-compatible almost complex structure and (e1 , . . . , e2n ) is a unitary frame. Lemma 8.1.9 Let ∇ be a symplectic connection on M . Then 2n

∇Jej X · ∇Y ej · ϕ = −

j=1

2n

∇∇Y (Jej ) X · ej · ϕ

j=1

for any X, Y ∈ Γ(T M ) and ϕ ∈ Γ(Q). Proof. By the symplecticity of ∇, we conclude 2n

∇Jej X · ∇Y ej · ϕ =

j=1

2n

ω(∇Y ej , Jei )∇Jej X · ei · ϕ

i,j=1

=

2n

ω(∇Y (Jei ), ej )∇Jej X · ei · ϕ

i,j=1

=−

2n

∇∇Y (Jei ) X · ei · ϕ .

i=1


107

Lemma 8.1.10 Let ∇ be a torsion-free symplectic connection on M . Then, for all symplectic vector fields X and Y and any ϕ ∈ Γ(Q), ∇X · ∇Y · ϕ − ∇Y · ∇X · ϕ = 2i

2n

∇∇Jej X Y · ej · ϕ

j=1

−2i

2n

∇∇Jej Y X · ej · ϕ .

j=1

Proof. By Equation (3.2.2), ∇X · ∇Y · ϕ =

2n

∇Jei X · ei · ∇Jej Y · ej · ϕ

i,j=1

=

2n

∇Jei X · ∇Jej Y · ei · ej · ϕ

i,j=1

−i

2n

ω ei , ∇Jej Y ∇Jei X · ej · ϕ

i,j=1

=

2n

∇Jei X · ∇Jej Y · ei · ej · ϕ − i

i,j=1

2n

∇∇Jej Y X · ej · ϕ

j=1

and ∇Y · ∇X · ϕ =

2n

∇Jej Y · ∇Jei X · ej · ei · ϕ − i

i,j=1

=

2n


j=1

∇Jej Y · ∇Jei X · ei · ej · ϕ − i

i,j=1

−i

2n

2n

2n


j=1

∇ei Y · ∇Jei X · ϕ .

i=1

Moreover, by Equation (8.1.5), 2n

2n ω ∇Jei X, ∇Jej Y ei · ej · ϕ = ω ∇∇Jej Y X, Jei ei · ej · ϕ

i,j=1

i,j=1

=

2n j=1

and

∇∇Jej Y X · ej · ϕ

108

8 Lie Derivative and Quantization 2n

∇ei Y · ∇Jei X · ϕ =

i=1

2n

ω (∇Jei X, Jej ) ∇ei Y · ej · ϕ

i,j=1

=

2n

ω ∇Jej X, Jei ∇ei Y · ej · ϕ

i,j=1

=

2n

∇∇Jej X Y · ej · ϕ .

j=1

Hence, applying once again Equation (3.2.2), one gets the assertion.

Lemma 8.1.11 Let ∇ be a torsion-free connection on M . Then ∇X ∇Z Y − ∇Y ∇Z X − ∇Z [X, Y ] = R(X, Y )Z + ∇[X,Z] Y − ∇[Y,Z] X for all X, Y, Z ∈ Γ(T M ). Proof. It is ∇X ∇Z Y − ∇Y ∇Z X − ∇Z [X, Y ] = ∇X ∇Z Y − ∇Y ∇Z X − ∇Z ∇X Y + ∇Z ∇Y X = R(X, Z)Y + R(Z, Y )X + ∇[X,Z] Y + ∇[Z,Y ] X . By the Bianchi identity, this implies the lemma.

Proof of Proposition 8.1.8. Let ∇ be a torsion-free symplectic connection on M . Then, by Proposition 8.1.7, i i 1 LX LY ϕ = ∇X ∇Y ϕ − ∇X (∇Y · ϕ) − ∇X · ∇Y ϕ − ∇X · ∇Y · ϕ . 2 2 4 Using Proposition 3.2.7, we see that ∇X (∇Y · ϕ) =

2n ∇X ∇Jej Y · ej + ∇Jej Y · ∇X ej · ϕ + ∇Y · ∇X ϕ . j=1

Thus [LX , LY ]ϕ − L[X,Y ] ϕ i ∇X ∇Jej Y · ej + ∇Jej Y · ∇X ej 2 j=1 −∇Y ∇Jej X · ej − ∇Jej X · ∇Y ej − ∇Jej [X, Y ] · ej · ϕ 1 − (∇X · ∇Y · ϕ − ∇Y · ∇X · ϕ) . 4 2n

= RQ (X, Y )ϕ −

8.2 Schr¨ odinger Equation for Quadratic Hamiltonians

109

According to Proposition 3.2.9 and Lemma 8.1.11, i ∇X ∇Jej Y · ej − ∇Y ∇Jej X · ej − ∇Jej [X, Y ] · ej · ϕ 2 j=1 2n

i R(X, Y )(Jej ) · ej + ∇[X,Jej ] Y · ej − ∇[Y,Jej ] X · ej · ϕ 2 j=1 2n

=

i ∇[X,Jej ] Y · ej − ∇[Y,Jej ] X · ej · ϕ . 2 j=1 2n

= RQ (X, Y )ϕ +

By Lemmas 8.1.9 and 8.1.10 and the torsion-freeness of ∇, it follows that i ∇Jej Y · ∇X ej − ∇Jej X · ∇Y ej 2 j=1 +∇[X,Jej ] Y · ej − ∇[Y,Jej ] X · ej · ϕ 1 − (∇X · ∇Y · ϕ − ∇Y · ∇X · ϕ) 4 2n i ∇∇X (Jej ) Y · ej − ∇∇Y (Jej ) X · ej = 2 j=1 2n

[LX , LY ]ϕ − L[X,Y ] ϕ = −

−∇[X,Jej ] Y · ej + ∇[Y,Jej ] X · ej −∇∇Jej X Y · ej + ∇∇Jej Y X · ej · ϕ =0,

and the proposition is proved.

8.2 Schr¨ odinger Equation for Quadratic Hamiltonians We consider R2n with its standard symplectic structure ω0 and a quadratic Hamiltonian u on R2n , i.e. a function u : R2n → Rn defined by u(v) = Υ v, v , where Υ is any symmetric real 2n × 2n-matrix. First we note that the Hamiltonian vector field Xu of u is given by an element of the Lie algebra sp(n, R). Lemma 8.2.1 (1) One has Xu (v) = −2J0 Υ v for all v ∈ R2n .

110


(2) J0 Υ and Υ J0 lie in sp(n, R). Proof. Assertion (1) follows from (du)v (w) = 2Υ v, w = −2ω0 (J0 Υ v, w) for any v, w ∈ R2n . To verify (2), use the fact that a real 2n × 2n-matrix A is an element of sp(n, R) if and only if AT J0 + J0 A = 0 . Let Hu denote the Hamilton operator associated to u via normal ordering quantization. That means that Hu =

n

(Υ ai , aj Qi ◦ Qj + Υ ai , bj (Qi ◦ Pj + Pj ◦ Qi )

i,j=1

+Υ bi , bj Pi ◦ Pj ) , where Qi and Pi are again the position and momentum operators. Lemma 8.2.2 One has the relation i m∗ ◦ ρ−1 ∗ (Υ J0 ) = − Hu . 2 Proof. By Lemma 1.3.1, 1 (ai · Υ ai + bi · Υ bi ) 2 i=1 n

ρ−1 ∗ (Υ J0 ) = =

n 1 (Υ ai , aj ai · aj + Υ ai , bj ai · bj 2 i,j=1

+Υ bi , aj bi · aj + Υ bi , bj bi · bj ) n 1 = (Υ ai , aj ai · aj + Υ ai , bj (ai · bj + bj · ai ) 2 i,j=1 +Υ bi , bj bi · bj ) .

Now apply Equation (1.4.2).

Let (AΥt ) be the one parameter subgroup of Sp(n, R) generated by 2Υ J0 , i.e. AΥt = exp(2tΥ J0 ) for t ∈ R, and let (qtΥ ) denote the lift of (AΥt ) to Mp(n, R).

8.3 Lie Derivative as Quantization

111

Proposition 8.2.3 For any f ∈ S(Rn ), the curve (m(qtΥ )f ) in S(Rn ) satisfies the Schr¨ odinger equation d m(qtΥ )f = −iHu m(qtΥ )f . dt Proof. We have d Υ d Υ d Υ ρ∗ q ρ(q ) A = = 2Υ J0 . = ds s s=0 ds s s=0 ds s s=0 Hence

d Υ q = 2 ρ−1 ∗ (Υ J0 ) . ds s s=0

By means of Lemma 8.2.2, we get

d d m(qtΥ )f = m(qsΥ )m(qtΥ )f dt ds s=0 d Υ q = m∗ m(qtΥ )f ds s s=0 Υ = 2 m∗ ρ−1 ∗ (Υ J0 ) m(qt )f = −iHu m(qtΥ )f

for any f ∈ S(Rn ).

8.3 Lie Derivative as Quantization We use the symplectic standard basis (a1 , . . . , an , b1 , . . . , bn ) understood as a global section of the symplectic frame bundle R of (R2n , ω0 ) to identify R2n × Sp(n, R) with R. That means, we assign to (v, A) ∈ R2n × Sp(n, R) the symplectic basis (Aa1 , . . . , Aan , Ab1 , . . . , Abn ) of Tv R2n = R2n . The metaplectic structure of (R2n , ω0 ) is then simply the product bundle P = R2n × Mp(n, R) , where the map FP : P → R is given by FP (v, q) = (v, ρ(q)) . We identify R2n × L2 (Rn ) with the symplectic spinor bundle Q via (v, f ) ∈ R2n × L2 (Rn ) → v, e+ , f ∈ Q , where e+ denotes the unit element of Mp(n, R). Then sections of Q are maps ϕ : R2n → L2 (Rn ). In the next lemma, we use that elements of Sp(n, R) can also be considered as symplectomorphisms of R2n .

112


Lemma 8.3.1 Let B ∈ Sp(n, R) and let qB ∈ Mp(n, R) such that ρ(qB ) = B. Then one has ˜ : R → R induced by B is given by (1) The isomorphism B ˜ B(v, A) = (Bv, BA) for (v, A) ∈ R2n × Sp(n, R). ¯ : P → P defined by (2) The isomorphism B ¯ B(v, q) = (Bv, qB q) ˜ for (v, q) ∈ R2n × Mp(n, R) is a lift of B. (3) It is

−1 ¯ )∗ ϕ (v) = m q −1 ϕ(Bv) (B B

for ϕ ∈ Γ(Q) and v ∈ R2n . Proof. Assertion (1) follows from (dB)v (w) = Bw for v, w ∈ R2n . Assertion (2) is easily checked. To see (3), first observe that the Mp(n, R)-equivariant mapping ϕˆ : P → L2 (Rn ) corresponding to ϕ ∈ Γ(Q) is given by ϕ(v, ˆ q) = m q −1 ϕ(v) for v ∈ R2n and q ∈ Mp(n, R). Hence, by (2), −1 ¯ ϕ(Bv) . ϕˆ ◦ B(v, q) = m q −1 m qB This implies the stated relation.

We now give an interpretation of the Lie derivative in the direction of an Hamiltonian vector field in the language of normal ordering quantization. For this, let f & for f ∈ S(Rn ) denote the symplectic spinor field over R2n that is constantly equal to f . Proposition 8.3.2 Let u be a quadratic Hamiltonian on R2n . Then & LXu f & = −i F −1 ◦ Hu ◦ F(f ) for any f ∈ S(Rn ). Proof. Let Υ , (AΥt ) and (qtΥ ) be as in Section 8.2. By Lemma 8.2.1, the flow (θt ) of the Hamiltonian vector field Xu is θt = exp(−2tJ0 Υ ) = J0 AΥ−t J0−1 .

8.3 Lie Derivative as Quantization

113

By Proposition 1.3.7(1) and Lemma 8.3.1, it follows that & (θ¯t−1 )∗ f & = F −1 ◦ m(qtΥ ) ◦ F(f ) for f ∈ S(Rn ). Now, using Proposition 8.2.3, one gets the assertion.

Although the above relation cannot be carried over to arbitrary manifolds, we can prove a Heisenberg relation in the general situation. Proposition 8.3.3 Let (M, ω) be a symplectic manifold that admits a metaplectic structure. Setting q(u)ϕ = iLXu ϕ , one has [q(u), q(v)]ϕ = −iq({u, v})ϕ for all u, v ∈ C ∞ (M ) and all symplectic spinor fields ϕ over M with respect to any metaplectic structure. Proof. By Propositions 2.1.9 and 8.1.8, one concludes that [q(u), q(v)]ϕ = − [LXu , LXv ] ϕ = −L[Xu ,Xv ] ϕ = LX{u,v} ϕ = −iq({u, v})ϕ .

References

1. A.O. Barut, R. Raczka: Theory of group representations and applications. 2nd rev. ed. Singapore: World Scientific (1986) 2. H. Baum: Spin-Strukturen und Dirac-Operatoren u ¨ber pseudoriemannschen Mannigfaltigkeiten. Teubner-Texte zur Mathematik, Bd. 41. Leipzig: BSB B. G. Teubner Verlagsgesellschaft (1981) 3. N. Berline, E. Getzler, M. Vergne: Heat Kernels and Dirac Operators. Grundlehren der Mathematischen Wissenschaften. 298. Berlin etc.: SpringerVerlag (1992) 4. R.J. Blattner: Some remarks on quantization. Symplectic geometry and mathematical physics, Proc. Colloq., Aix-en-Provence/Fr. 1990, Prog. Math. 99, 37-47 (1991) 5. A. Borel, N. Wallach: Continuous cohomology, discrete subgroups, and representations of reductive groups. 2nd ed. Mathematical Surveys and Monographs. 67. Providence, RI: American Mathematical Society (AMS) (2000) 6. F. Bourgeois, M. Cahen: Can one define a preferred symplectic connection? Rep. Math. Phys. 43, No.1-2, 35-42 (1999) 7. F. Bourgeois, M. Cahen: A variational principle for symplectic connections. J. Geom. Phys. 30, No.3, 233-265 (1999) 8. M. Cahen, S. Gutt, J. Rawnsley: Symplectic connections with parallel Ricci tensor. Grabowski, Janusz (ed.) et al., Poisson geometry. Stanislaw Zakrzewski in memoriam. Warszawa: Polish Academy of Sciences, Institute of Mathematics, Banach Cent. Publ. 51, 31-41 (2000) 9. M. Cahen, S. Gutt, J. Rawnsley: Symmetric symplectic spaces with Ricci-type curvature. Dito, Giuseppe (ed.) et al., Conférence Moshé Flato 1999: Quantization, deformations, and symmetries, Dijon, France, September 5-8, 1999. Volume II. Dordrecht: Kluwer Academic Publishers. Math. Phys. Stud. 22, 81-91 (2000) 10. M. Cahen, S. Gutt, J. Horowitz, J. Rawnsley: Homogeneous symplectic manifolds with Ricci-type curvature. J. Geom. Phys. 38, No.2, 140-151 (2001) 11. A. Cannas da Silva: Lectures on symplectic geometry. Lecture Notes in Mathematics. 1764. Berlin: Springer (2001). 12. M. de Gosson: Maslov classes, metaplectic representation and Lagrangian quantization. Mathematical Research. 95. Berlin: Akademie Verlag (1997) 13. C. Emmrich, A. Weinstein: The differential geometry of Fedosov’s quantization. Brylinski, Jean-Luc (ed.) et al., Lie theory and geometry: in honor of Bertram

116

14. 15. 16. 17. 18.

19.

20. 21. 22. 23.

24. 25. 26. 27. 28.

29. 30. 31. 32.

References Kostant on the occasion of his 65th birthday. Invited papers, some originated at a symposium held at MIT, Cambridge, MA, USA in May 1993. Boston, MA: Birkh¨ auser. Prog. Math. 123, 217-239 (1994). B.V. Fedosov: A simple geometrical construction of deformation quantization. J. Differ. Geom. 40, No.2, 213-238 (1994) B.V. Fedosov: Deformation quantization and index theory. Mathematical Topics. 9. Berlin: Akademie Verlag (1996) G.B. Folland: Harmonic analysis in phase space. Annals of Mathematics Studies, 122. Princeton, NJ: Princeton University Press (1989). M. Forger, H. Hess: Universal metaplectic structures and geometric quantization. Commun. Math. Phys. 64, 269-278 (1979) Th. Friedrich: Einige differentialgeometrische Untersuchungen des DiracOperators einer Riemannschen Mannigfaltigkeit. Habilitation thesis, Humboldt University Berlin (1979) Th. Friedrich: Der erste Eigenwert des Dirac-Operators einer kompakten Riemannschen Mannigfaltigkeit nichtnegativer Skalarkr¨ ummung. Math. Nachr. 97, 117-146 (1980) Th. Friedrich: Zur Abh¨ angigkeit des Dirac Operators von der Spin-Struktur. Colloq. Math. 48, 57-62 (1984) Th. Friedrich: Dirac operators in Riemannian geometry. Graduate Studies in Mathematics 25. Providence, RI: American Mathematical Society (2000) I. Gelfand, V. Retakh, M. Shubin: Fedosov Manifolds. Adv. Math. 136, No.1, 104-140 (1998) S. Gutt: Variations on deformation quantization. Dito, Giuseppe (ed.) et al., Conférence Moshé Flato 1999: Quantization, deformation, and symmetries, Dijon, France, September 5-8, 1999. Volume I. Dordrecht: Kluwer Academic Publishers. Math. Phys. Stud. 21, 217-254 (2000) K. Habermann: The Dirac Operator on Symplectic Spinors. Ann. Global Anal. Geom. 13, 155-168 (1995) K. Habermann: Basic Properties of Symplectic Dirac Operators. Commun. Math. Phys. 184, 629-652 (1997) K. Habermann: Harmonic Symplectic Spinors on Riemann Surfaces. Manuscr. Math. 94, 465-484 (1997) K. Habermann: Symplectic Dirac Operators on K¨ ahler Manifolds. Math. Nachr. 211, 37-62 (2000) K. Habermann: An M pC -spinorial approach to a geometric-type quantization for symplectic manifolds. Gesztesy, Fritz (ed.) et al., Stochastic processes, physics and geometry: New interplays. II. A volume in honor of Sergio Albeverio. Proceedings of the conference on infinite dimensional (stochastic) analysis and quantum physics, Leipzig, Germany, January 18-22, 1999. Providence, RI: American Mathematical Society (AMS). CMS Conf. Proc. 29, 259-268 (2000) K. Habermann, A. Klein: A Fourier transform for symplectic spinors and applications. Preprint, Greifswald University (2002) K. Habermann, A. Klein: Lie derivative of symplectic spinor fields, metaplectic representation, and quantization. Rostocker Math. Kolloq. 57, 71-91 (2003) H. Hofer, E. Zehnder: Symplectic invariants and Hamiltonian dynamics. Birkh¨ auser Advanced Texts. Basel: Birkh¨ auser (1994) M. Kashiwara, M. Vergne: On the Segal-Shale-Weil representations and harmonic polynomials. Invent. Math. 44, 1-47 (1978)

References

117

33. A.A. Kirillov: Geometric quantization. Dynamical systems. IV. Symplectic geometry and its applications, Encycl. Math. Sci. 4, 137-172 (1990); translation from Itogi Nauki Tekh., Ser. Sovrem. Probl. Mat., Fundam. Napravleniya 4, 141-178 (1985) 34. A. Klein: Eine Fouriertransformation f¨ ur symplektische Spinoren und Anwendungen in der Quantisierung. Diploma Thesis, Technische Universit¨ at Berlin (2000) 35. S. Kobayashi, K. Nomizu: Foundations of differential geometry. I. New YorkLondon: Interscience Publishers, a division of John Wiley & Sons. XI (1963) 36. S. Kobayashi, K. Nomizu: Foundations of Differential Geometry. Vol. II. New York-London-Sydney: Interscience Publishers a division of John Wiley and Sons (1969) 37. B. Kostant: Symplectic Spinors. Symp. math. 14, Geom. simplett., Fis. mat., Teor. geom. Integr. Var. minim., Convegni 1973, 139-152 (1974) 38. A. Lichnerowicz: Spineurs harmoniques. C. R. Acad. Sci., Paris 257, 7-9 (1963) 39. D. McDuff, D. Salamon: Introduction to symplectic topology. Oxford Mathematical Monographs. Oxford: Clarendon Press (1995) 40. I. Niven, H.S. Zuckerman: An introduction to the theory of numbers. 3rd ed. New York etc.: John Wiley & Sons, Inc. XII (1972) 41. M.J. Pflaum: The normal symbol on Riemannian manifolds. New York J. Math. 4, 97-125, electronic only (1998) 42. P.L. Robinson, J.H. Rawnsley: The metaplectic representation, M pc structures and geometric quantization. Mem. Am. Math. Soc. 410 (1989) 43. E. Schr¨ odinger: Diracsches Elektron im Schwerefeld. I. Sitzungsber. Preuß. Akad. Wiss., Phys.-Math. Kl. 1932, No. 11/12, 105-128 (1932) 44. M. Shubin: A sequence of connections and a characterization of K¨ ahler manifolds. Farber, Michael (ed.) et al., Tel Aviv topology conference: Rothenberg Festschrift. Proceedings of the international conference on topology, Tel Aviv, Israel, June 1-5, 1998 dedicated to Mel Rothenberg on the occasion of his 65th birthday. Providence, RI: American Mathematical Society. Contemp. Math. 231, 265-270 (1999). 45. F. Sommen: An extension of Clifford analysis towards super-symmetry. Ryan, John (ed.) et al., Clifford algebras and their applications in mathematical physics. Papers of the 5th international conference, Ixtapa-Zihuatanejo, Mexico, June 27-July 4, 1999. Volume 2: Clifford analysis. Boston, MA: Birkh¨ auser. Prog. Phys. 19, 199-224 (2000) 46. Ph. Tondeur: Affine Zusammenh¨ ange auf Mannigfaltigkeiten mit fast-symplektischer Struktur. Comment. Math. Helv. 36, 234-244 (1961) 47. I. Vaisman: Symplectic curvature tensors. Monatsh. Math. 100, 299-327 (1985) 48. N.R. Wallach: Harmonic Analysis on Homogeneous Spaces. Pure and Applied Mathematics. 19. New York: Marcel Dekker, Inc. (1973) 49. N.R. Wallach: Symplectic Geometry and Fourier Analysis. With an appendix on quantum mechanics by Robert Hermann. Lie Groups: History, Frontiers and Applications. Vol. V. Brookline, Mass.: Math. Sci. Press. (1977) 50. A. Weil: Sur certains groupes d’op´ erateurs unitaires. Acta Math. 111, 143-211 (1964) 51. N.M.J. Woodhouse: Geometric quantization. 2nd ed. Oxford Mathematical Monographs. Oxford: Clarendon Press (1997) 52. Ch. Wyss: Symplektische Diracoperatoren auf dem komplexen projektiven Raum. Diploma Thesis, Universit¨ at Bremen (2003)

118

References

ˇ 53. D.P. Zelobenko: Compact Lie Groups and their representations. Translated from the Russian by Israel Program for Scientific Translations. Translations of Mathematical Monographs. Vol. 40. Providence, R.I.: American Mathematical Society (AMS). VIII (1973)

Index

almost complex structure, 23 ω-compatible, 23 anti-canonical line bundle, 46 Baum, 49 Bianchi identity, 29 Bourgeois, 21 Cahen, 21 Casimir operator, 90 Cauchy–Riemann equation generalized, 79 Chern class first, 36 complex projective space, 36, 87, 92 complex torus, 85 connection canonical, 90 Hermitian, 27 Levi–Civita, 34, 81 symplectic, 25 reducible, 33 torsion-free, 27 cotangent bundle, 22, 36 curvature, 29 of the spinor derivative, 42 Darboux theorem, 22 deformation of metaplectic structures, 58 divergence, 27 Einstein manifold, 84

Fedosov, 21 flow, 102, 112 formally self-adjoint, 64, 66, 81 Fourier transform, 8 for symplectic spinors, 97 Friedrich, 49 Frobenius reciprocity, 91, 92 Fubini–Study metric, 36, 88 G˚ arding, 6 Gelfand, 21 Gutt, 21 Hamiltonian quadratic, 109 vector field, 24 harmonic oscillator, 17 Heisenberg group, 6 relation, 113 Hermite function, 17, 89 Hermitian connection, 27 structure, 39 holomorphic sectional curvature constant, 33, 84, 94 invariant subspace, 5 K¨ ahler form, 22, 81 manifold, 22, 81 Lie derivative, 102

120

Index

of symplectic spinor fields, 103 metaplectic group, 7 representation, 10 differential of, 14 structure, 35 momentum operator, 11 normal ordering quantization, 110, 112 orthogonal group, 2 Pell’s equation, 95 Poisson bracket, 24 position operator, 11 quadratic Hamiltonian, 109 quantization, 97, 101 normal ordering, 110, 112 Rawnsley, 21 representation, 5 adjoint, 88 differential of, 6 equivalent, 6 unitary, 6 irreducible, 5 isotropy, 88 metaplectic, 10 differential of, 14 projective unitary, 8 unitary, 5 Retakh, 21 Ricci tensor, 31 symplectic, 31 scalar curvature, 34 symplectic, 34 Schr¨ odinger equation, 111 representation, 6 Schur’s lemma, 5, 93 Schwartz space, 8 sectional curvature, 84 holomorphic, 33, 84 Segal–Shale–Weil representation, 10 Shubin, 21 smooth section, 37 vector, 6 spectrum, 93, 100

spinor derivative, 39 curvature of, 42 Stiefel–Whitney class first, 36 second, 36 Stone–von Neumann theorem, 7 symmetric space, 88 symplectic basis, 1, 23 Clifford algebra, 4 multiplication, 13, 38 connection, 25 curvature tensor, 29 Dirac operator, 50 form, 1 frame, 23 frame bundle, 23 gradient, 24 group, 2 Lie algebra, 3 manifold, 22 Ricci tensor, 31 scalar curvature, 34 spinor, 10 bundle, 37 field, 37 Laplacian, 70 structure, 22 vector field, 102 symplectomorphic, 22 symplectomorphism, 22 Tondeur, 21 torsion, 27 vector field, 27 unitary frame, 24 group, 2 Vaisman, 21 vector field Hamiltonian, 24 symplectic, 102 volume element, 28 weight, 90, 92 Weil, 9 Weitzenb¨ ock formula, 72, 82 Weyl algebra, 4

Lecture Notes in Mathematics For information about earlier volumes please contact your bookseller or Springer LNM Online archive: springerlink.com

Vol. 1681: G. J. Wirsching, The Dynamical System Generated by the 3n+1 Function (1998) Vol. 1682: H.-D. Alber, Materials with Memory (1998) Vol. 1683: A. Pomp, The Boundary-Domain Integral Method for Elliptic Systems (1998) Vol. 1684: C. A. Berenstein, P. F. Ebenfelt, S. G. Gindikin, S. Helgason, A. E. Tumanov, Integral Geometry, Radon Transforms and Complex Analysis. Firenze, 1996. Editors: E. Casadio Tarabusi, M. A. Picardello, G. Zampieri (1998) Vol. 1685: S. König, A. Zimmermann, Derived Equivalences for Group Rings (1998) Vol. 1686: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXII (1998) Vol. 1687: F. Bornemann, Homogenization in Time of Singularly Perturbed Mechanical Systems (1998) Vol. 1688: S. Assing, W. Schmidt, Continuous Strong Markov Processes in Dimension One (1998) Vol. 1689: W. Fulton, P. Pragacz, Schubert Varieties and Degeneracy Loci (1998) Vol. 1690: M. T. Barlow, D. Nualart, Lectures on Probability Theory and Statistics. Editor: P. Bernard (1998) Vol. 1691: R. Bezrukavnikov, M. Finkelberg, V. Schechtman, Factorizable Sheaves and Quantum Groups (1998) Vol. 1692: T. M. W. Eyre, Quantum Stochastic Calculus and Representations of Lie Superalgebras (1998) Vol. 1694: A. Braides, Approximation of Free-Discontinuity Problems (1998) Vol. 1695: D. J. Hart el, Markov Set-Chains ( 1998) Vol. 1696: E. Bouscaren (Ed.): Model Theory and Algebraic Geometry (1998) Vol. 1697: B. Cockburn, C. Johnson, C.-W. Shu, E. Tadmor, Advanced Numerical Approximation of Nonlinear Hyperbolic Equations. Cetraro, Italy, 1997. Editor: A. Quarteroni (1998) Vol. 1698: M. Bhattacharjee, D. Macpherson, R. G. Möller, P. Neumann, Notes on In nite Permutation Groups (1998) Vol. 1699: A. Inoue,Tomita-Takesaki Theory in Algebras of Unbounded Operators (1998) Vol. 1700: W. A. Woyczyński, Burgers-KPZ Turbulence (1998) Vol. 1701: Ti-Jun Xiao, J. Liang, The Cauchy Problem of Higher Order Abstract Differential Equations (1998) Vol. 1702: J. Ma, J. Yong, Forward-Backward Stochastic Differential Equations and Their Applications (1999) Vol. 1703: R. M. Dudley, R. Norvaiša, Differentiability of Six Operators on Nonsmooth Functions and pVariation (1999) Vol. 1704: H. Tamanoi, Elliptic Genera and Vertex Operator Super-Algebras (1999) Vol. 1705: I. Nikolaev, E. Zhuzhoma, Flows in 2-dimensional Manifolds (1999) Vol. 1706: S. Yu. Pilyugin, Shadowing in Dynamical Systems (1999)

Vol. 1707: R. Pytlak, Numerical Methods for Optimal Control Problems with State Constraints (1999) Vol. 1708: K. Zuo, Representations of Fundamental Groups of Algebraic Varieties (1999) Vol. 1709: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXIII (1999) Vol. 1710: M. Koecher, The Minnesota Notes on Jordan Algebras and Their Applications (1999) Vol. 1711: W. Ricker, Operator Algebras Generated by Commuting Projećtions: A Vector Measure Approach (1999) Vol. 1712: N. Schwartz, J. J. Madden, Semi-algebraic Function Rings and Re ectors of Partially Ordered Rings (1999) Vol. 1713: F. Bethuel, G. Huisken, S. Müller, K. Steffen, Calculus of Variations and Geometric Evolution Problems. Cetraro, 1996. Editors: S. Hildebrandt, M. Struwe (1999) Vol. 1714: O. Diekmann, R. Durrett, K. P. Hadeler, P. K. Maini, H. L. Smith, Mathematics Inspired by Biology. Martina Franca, 1997. Editors: V. Capasso, O. Diekmann (1999) Vol. 1715: N. V. Krylov, M. Röckner, J. Zabczyk, Stochastic PDE’s and Kolmogorov Equations in In nite Dimensions. Cetraro, 1998. Editor: G. Da Prato (1999) Vol. 1716: J. Coates, R. Greenberg, K. A. Ribet, K. Rubin, Arithmetic Theory of Elliptic Curves. Cetraro, 1997. Editor: C. Viola (1999) Vol. 1717: J. Bertoin, F. Martinelli, Y. Peres, Lectures on Probability Theory and Statistics. Saint-Flour, 1997. Editor: P. Bernard (1999) Vol. 1718: A. Eberle, Uniqueness and Non-Uniqueness of Semigroups Generated by Singular Diffusion Operators (1999) Vol. 1719: K. R. Meyer, Periodic Solutions of the N-Body Problem (1999) Vol. 1720: D. Elworthy, Y. Le Jan, X-M. Li, On the Geometry of Diffusion Operators and Stochastic Flows (1999) Vol. 1721: A. Iarrobino, V. Kanev, Power Sums, Gorenstein Algebras, and Determinantal Loci (1999) Vol. 1722: R. McCutcheon, Elemental Methods in Ergodic Ramsey Theory (1999) Vol. 1723: J. P. Croisille, C. Lebeau, Diffraction by an Immersed Elastic Wedge (1999) Vol. 1724: V. N. Kolokoltsov, Semiclassical Analysis for Diffusions and Stochastic Processes (2000) Vol. 1725: D. A. Wolf-Gladrow, Lattice-Gas Cellular Automata and Lattice Boltzmann Models (2000) Vol. 1726: V. Marić, Regular Variation and Differential Equations (2000) Vol. 1727: P. Kravanja M. Van Barel, Computing the Zeros of Analytic Functions (2000) Vol. 1728: K. Gatermann Computer Algebra Methods for Equivariant Dynamical Systems (2000) Vol. 1729: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.)

Séminaire de Probabilités XXXIV (2000) Vol. 1730: S. Graf, H. Luschgy, Foundations of Quantization for Probability Distributions (2000) Vol. 1731: T. Hsu, Quilts: Central Extensions, Braid Actions, and Finite Groups (2000) Vol. 1732: K. Keller, Invariant Factors, Julia Equivalences and the (Abstract) Mandelbrot Set (2000) Vol. 1733: K. Ritter, Average-Case Analysis of Numerical Problems (2000) Vol. 1734: M. Espedal, A. Fasano, A. Mikelić, Filtration in Porous Media and Industrial Applications. Cetraro 1998. Editor: A. Fasano. 2000. Vol. 1735: D. Yafaev, Scattering Theory: Some Old and New Problems (2000) Vol. 1736: B. O. Turesson, Nonlinear Potential Theory and Weighted Sobolev Spaces (2000) Vol. 1737: S. Wakabayashi, Classical Microlocal Analysis in the Space of Hyperfunctions (2000) Vol. 1738: M. Émery, A. Nemirovski, D. Voiculescu, Lectures on Probability Theory and Statistics (2000) Vol. 1739: R. Burkard, P. Deu hard, A. Jameson, J.-L. Lions, G. Strang, Computational Mathematics Driven by Industrial Problems. Martina Franca, 1999. Editors: V. Capasso, H. Engl, J. Periaux (2000) Vol. 1740: B. Kawohl, O. Pironneau, L. Tartar, J.-P. Zolesio, Optimal Shape Design. Tróia, Portugal 1999. Editors: A. Cellina, A. Ornelas (2000) Vol. 1741: E. Lombardi, Oscillatory Integrals and Phenomena Beyond all Algebraic Orders (2000) Vol. 1742: A. Unterberger, Quantization and Nonholomorphic Modular Forms (2000) Vol. 1743: L. Habermann, Riemannian Metrics of Constant Mass and Moduli Spaces of Conformal Structures (2000) Vol. 1744: M. Kunze, Non-Smooth Dynamical Systems (2000) Vol. 1745: V. D. Milman, G. Schechtman (Eds.), Geometric Aspects of Functional Analysis. Israel Seminar 19992000 (2000) Vol. 1746: A. Degtyarev, I. Itenberg, V. Kharlamov, Real Enriques Surfaces (2000) Vol. 1747: L. W. Christensen, Gorenstein Dimensions (2000) Vol. 1748: M. Ruzicka, Electrorheological Fluids: Modeling and Mathematical Theory (2001) Vol. 1749: M. Fuchs, G. Seregin, Variational Methods for Problems from Plasticity Theory and for Generalized Newtonian Fluids (2001) Vol. 1750: B. Conrad, Grothendieck Duality and Base Change (2001) Vol. 1751: N. J. Cutland, Loeb Measures in Practice: Recent Advances (2001) Vol. 1752: Y. V. Nesterenko, P. Philippon, Introduction to Algebraic Independence Theory (2001) Vol. 1753: A. I. Bobenko, U. Eitner, Painlevé Equations in the Differential Geometry of Surfaces (2001) Vol. 1754: W. Bertram, The Geometry of Jordan and Lie Structures (2001) Vol. 1755: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXV (2001) Vol. 1756: P. E. Zhidkov, Korteweg de Vries and Nonlinear Schrödinger Equations: Qualitative Theory (2001) Vol. 1757: R. R. Phelps, Lectures on Choquet’s Theorem (2001) Vol. 1758: N. Monod, Continuous Bounded Cohomology of Locally Compact Groups (2001) Vol. 1759: Y. Abe, K. Kopfermann, Toroidal Groups

(2001) Vol. 1760: D. Filipović, Consistency Problems for HeathJarrow-Morton Interest Rate Models (2001) Vol. 1761: C. Adelmann, The Decomposition of Primes in Torsion Point Fields (2001) Vol. 1762: S. Cerrai, Second Order PDE’s in Finite and In nite Dimension ( 2001) Vol. 1763: J.-L. Loday, A. Frabetti, F. Chapoton, F. Goichot, Dialgebras and Related Operads (2001) Vol. 1764: A. Cannas da Silva, Lectures on Symplectic Geometry (2001) Vol. 1765: T. Kerler, V. V. Lyubashenko, Non-Semisimple Topological Quantum Field Theories for 3-Manifolds with Corners (2001) Vol. 1766: H. Hennion, L. Hervé, Limit Theorems for Markov Chains and Stochastic Properties of Dynamical Systems by Quasi-Compactness (2001) Vol. 1767: J. Xiao, Holomorphic Q Classes (2001) Vol. 1768: M.J. P aum, Analytic and Geometric Study of Strati ed Spaces (2001) Vol. 1769: M. Alberich-Carramiñana, Geometry of the Plane Cremona Maps (2002) Vol. 1770: H. Gluesing-Luerssen, Linear DelayDifferential Systems with Commensurate Delays: An Algebraic Approach (2002) Vol. 1771: M. Émery, M. Yor (Eds.), Séminaire de Probabilités 1967-1980. A Selection in Martingale Theory (2002) Vol. 1772: F. Burstall, D. Ferus, K. Leschke, F. Pedit, U. Pinkall, Conformal Geometry of Surfaces in S4 (2002) Vol. 1773: Z. Arad, M. Muzychuk, Standard Integral Table Algebras Generated by a Non-real Element of Small Degree (2002) Vol. 1774: V. Runde, Lectures on Amenability (2002) Vol. 1775: W. H. Meeks, A. Ros, H. Rosenberg, The Global Theory of Minimal Surfaces in Flat Spaces. Martina Franca 1999. Editor: G. P. Pirola (2002) Vol. 1776: K. Behrend, C. Gomez, V. Tarasov, G. Tian, Quantum Comohology. Cetraro 1997. Editors: P. de Bartolomeis, B. Dubrovin, C. Reina (2002) Vol. 1777: E. García-Río, D. N. Kupeli, R. VázquezLorenzo, Osserman Manifolds in Semi-Riemannian Geometry (2002) Vol. 1778: H. Kiechle, Theory of K-Loops (2002) Vol. 1779: I. Chueshov, Monotone Random Systems (2002) Vol. 1780: J. H. Bruinier, Borcherds Products on O(2,1) and Chern Classes of Heegner Divisors (2002) Vol. 1781: E. Bolthausen, E. Perkins, A. van der Vaart, Lectures on Probability Theory and Statistics. Ecole d’ Eté de Probabilités de Saint-Flour XXIX-1999. Editor: P. Bernard (2002) Vol. 1782: C.-H. Chu, A. T.-M. Lau, Harmonic Functions on Groups and Fourier Algebras (2002) Vol. 1783: L. Grüne, Asymptotic Behavior of Dynamical and Control Systems under Perturbation and Discretization (2002) Vol. 1784: L.H. Eliasson, S. B. Kuksin, S. Marmi, J.-C. Yoccoz, Dynamical Systems and Small Divisors. Cetraro, Italy 1998. Editors: S. Marmi, J.-C. Yoccoz (2002) Vol. 1785: J. Arias de Reyna, Pointwise Convergence of Fourier Series (2002) Vol. 1786: S. D. Cutkosky, Monomialization of Morphisms from 3-Folds to Surfaces (2002) Vol. 1787: S. Caenepeel, G. Militaru, S. Zhu, Frobenius and Separable Functors for Generalized Module Categories and Nonlinear Equations (2002)

Vol. 1788: A. Vasil’ev, Moduli of Families of Curves for Conformal and Quasiconformal Mappings (2002) Vol. 1789: Y. Sommerhäuser, Yetter-Drinfel’d Hopf algebras over groups of prime order (2002) Vol. 1790: X. Zhan, Matrix Inequalities (2002) Vol. 1791: M. Knebusch, D. Zhang, Manis Valuations and Prüfer Extensions I: A new Chapter in Commutative Algebra (2002) Vol. 1792: D. D. Ang, R. Goren o, V. K. Le, D. D. Trong, Moment Theory and Some Inverse Problems in Potential Theory and Heat Conduction (2002) Vol. 1793: J. Cortés Monforte, Geometric, Control and Numerical Aspects of Nonholonomic Systems (2002) Vol. 1794: N. Pytheas Fogg, Substitution in Dynamics, Arithmetics and Combinatorics. Editors: V. Berthé, S. Ferenczi, C. Mauduit, A. Siegel (2002) Vol. 1795: H. Li, Filtered-Graded Transfer in Using Noncommutative Gröbner Bases (2002) Vol. 1796: J.M. Melenk, hp-Finite Element Methods for Singular Perturbations (2002) Vol. 1797: B. Schmidt, Characters and Cyclotomic Fields in Finite Geometry (2002) Vol. 1798: W.M. Oliva, Geometric Mechanics (2002) Vol. 1799: H. Pajot, Analytic Capacity, Recti ability, Menger Curvature and the Cauchy Integral (2002) Vol. 1800: O. Gabber, L. Ramero, Almost Ring Theory (2003) Vol. 1801: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXVI (2003) Vol. 1802: V. Capasso, E. Merzbach, B.G. Ivanoff, M. Dozzi, R. Dalang, T. Mountford, Topics in Spatial Stochastic Processes. Martina Franca, Italy 2001. Editor: E. Merzbach (2003) Vol. 1803: G. Dolzmann, Variational Methods for Crystalline Microstructure – Analysis and Computation (2003) Vol. 1804: I. Cherednik, Ya. Markov, R. Howe, G. Lusztig, Iwahori-Hecke Algebras and their Representation Theory. Martina Franca, Italy 1999. Editors: V. Baldoni, D. Barbasch (2003) Vol. 1805: F. Cao, Geometric Curve Evolution and Image Processing (2003) Vol. 1806: H. Broer, I. Hoveijn. G. Lunther, G. Vegter, Bifurcations in Hamiltonian Systems. Computing Singularities by Gröbner Bases (2003) Vol. 1807: V. D. Milman, G. Schechtman (Eds.), Geometric Aspects of Functional Analysis. Israel Seminar 20002002 (2003) Vol. 1808: W. Schindler, Measures with Symmetry Properties (2003) Vol. 1809: O. Steinbach, Stability Estimates for Hybrid Coupled Domain Decomposition Methods (2003) Vol. 1810: J. Wengenroth, Derived Functors in Functional Analysis (2003) Vol. 1811: J. Stevens, Deformations of Singularities (2003) Vol. 1812: L. Ambrosio, K. Deckelnick, G. Dziuk, M. Mimura, V. A. Solonnikov, H. M. Soner, Mathematical Aspects of Evolving Interfaces. Madeira, Funchal, Portugal 2000. Editors: P. Colli, J. F. Rodrigues (2003) Vol. 1813: L. Ambrosio, L. A. Caffarelli, Y. Brenier, G. Buttazzo, C. Villani, Optimal Transportation and its Applications. Martina Franca, Italy 2001. Editors: L. A. Caffarelli, S. Salsa (2003) Vol. 1814: P. Bank, F. Baudoin, H. Föllmer, L.C.G. Rogers, M. Soner, N. Touzi, Paris-Princeton Lectures on Mathematical Finance 2002 (2003)

Vol. 1815: A. M. Vershik (Ed.), Asymptotic Combinatorics with Applications to Mathematical Physics. St. Petersburg, Russia 2001 (2003) Vol. 1816: S. Albeverio, W. Schachermayer, M. Talagrand, Lectures on Probability Theory and Statistics. Ecole d’Eté de Probabilités de Saint-Flour XXX-2000. Editor: P. Bernard (2003) Vol. 1817: E. Koelink, W. Van Assche(Eds.), Orthogonal Polynomials and Special Functions. Leuven 2002 (2003) Vol. 1818: M. Bildhauer, Convex Variational Problems with Linear, nearly Linear and/or Anisotropic Growth Conditions (2003) Vol. 1819: D. Masser, Yu. V. Nesterenko, H. P. Schlickewei, W. M. Schmidt, M. Waldschmidt, Diophantine Approximation. Cetraro, Italy 2000. Editors: F. Amoroso, U. Zannier (2003) Vol. 1820: F. Hiai, H. Kosaki, Means of Hilbert Space Operators (2003) Vol. 1821: S. Teufel, Adiabatic Perturbation Theory in Quantum Dynamics (2003) Vol. 1822: S.-N. Chow, R. Conti, R. Johnson, J. MalletParet, R. Nussbaum, Dynamical Systems. Cetraro, Italy 2000. Editors: J. W. Macki, P. Zecca (2003) Vol. 1823: A. M. Anile, W. Allegretto, C. Ringhofer, Mathematical Problems in Semiconductor Physics. Cetraro, Italy 1998. Editor: A. M. Anile (2003) Vol. 1824: J. A. Navarro González, J. B. Sancho de Salas, C ∞ – Differentiable Spaces (2003) Vol. 1825: J. H. Bramble, A. Cohen, W. Dahmen, Multiscale Problems and Methods in Numerical Simulations, Martina Franca, Italy 2001. Editor: C. Canuto (2003) Vol. 1826: K. Dohmen, Improved Bonferroni Inequalities via Abstract Tubes. Inequalities and Identities of Inclusion-Exclusion Type. VIII, 113 p, 2003. Vol. 1827: K. M. Pilgrim, Combinations of Complex Dynamical Systems. IX, 118 p, 2003. Vol. 1828: D. J. Green, Gröbner Bases and the Computation of Group Cohomology. XII, 138 p, 2003. Vol. 1829: E. Altman, B. Gaujal, A. Hordijk, DiscreteEvent Control of Stochastic Networks: Multimodularity and Regularity. XIV, 313 p, 2003. Vol. 1830: M. I. Gil’, Operator Functions and Localization of Spectra. XIV, 256 p, 2003. Vol. 1831: A. Connes, J. Cuntz, E. Guentner, N. Higson, J. E. Kaminker, Noncommutative Geometry, Martina Franca, Italy 2002. Editors: S. Doplicher, L. Longo (2004) Vol. 1832: J. Azéma, M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXVII (2003) Vol. 1833: D.-Q. Jiang, M. Qian, M.-P. Qian, Mathematical Theory of Nonequilibrium Steady States. On the Frontier of Probability and Dynamical Systems. IX, 280 p, 2004. Vol. 1834: Yo. Yomdin, G. Comte, Tame Geometry with Application in Smooth Analysis. VIII, 186 p, 2004. Vol. 1835: O.T. Izhboldin, B. Kahn, N.A. Karpenko, A. Vishik, Geometric Methods in the Algebraic Theory of Quadratic Forms. Summer School, Lens, 2000. Editor: J.P. Tignol (2004) Vol. 1836: C. Nastasescu, F. Van Oystaeyen, Methods of Graded Rings. XIII, 304 p, 2004. Vol. 1837: S. Tavaré, O. Zeitouni, Lectures on Probability Theory and Statistics. Ecole d’Eté de Probabilités de Saint-Flour XXXI-2001. Editor: J. Picard (2004) Vol. 1838: A.J. Ganesh, N.W. O’Connell, D.J. Wischik, Big Queues. XII, 254 p, 2004. Vol. 1839: R. Gohm, Noncommutative Stationary Processes. VIII, 170 p, 2004.

Vol. 1840: B. Tsirelson, W. Werner, Lectures on Probability Theory and Statistics. Ecole d’Eté de Probabilités de Saint-Flour XXXII-2002. Editor: J. Picard (2004) Vol. 1841: W. Reichel, Uniqueness Theorems for Variational Problems by the Method of Transformation Groups (2004) Vol. 1842: T. Johnsen, A.L. Knutsen, K3 Projective Models in Scrolls (2004) Vol. 1843: B. Jefferies, Spectral Properties of Noncommuting Operators (2004) Vol. 1844: K.F. Siburg, The Principle of Least Action in Geometry and Dynamics (2004) Vol. 1845: Min Ho Lee, Mixed Automorphic Forms, Torus Bundles, and Jacobi Forms (2004) Vol. 1846: H. Ammari, H. Kang, Reconstruction of Small Inhomogeneities from Boundary Measurements (2004) Vol. 1847: T.R. Bielecki, T. Björk, M. Jeanblanc, M. Rutkowski, J.A. Scheinkman, W. Xiong, Paris-Princeton Lectures on Mathematical Finance 2003 (2004) Vol. 1848: M. Abate, J. E. Fornaess, X. Huang, J. P. Rosay, A. Tumanov, Real Methods in Complex and CR Geometry, Martina Franca, Italy 2002. Editors: D. Zaitsev, G. Zampieri (2004) Vol. 1849: Martin L. Brown, Heegner Modules and Elliptic Curves (2004) Vol. 1850: V. D. Milman, G. Schechtman (Eds.), Geometric Aspects of Functional Analysis. Israel Seminar 20022003 (2004) Vol. 1851: O. Catoni, Statistical Learning Theory and Stochastic Optimization (2004) Vol. 1852: A.S. Kechris, B.D. Miller, Topics in Orbit Equivalence (2004) Vol. 1853: Ch. Favre, M. Jonsson, The Valuative Tree (2004) Vol. 1854: O. Saeki, Topology of Singular Fibers of Differential Maps (2004) Vol. 1855: G. Da Prato, P.C. Kunstmann, I. Lasiecka, A. Lunardi, R. Schnaubelt, L. Weis, Functional Analytic Methods for Evolution Equations. Editors: M. Iannelli, R. Nagel, S. Piazzera (2004) Vol. 1856: K. Back, T.R. Bielecki, C. Hipp, S. Peng, W. Schachermayer, Stochastic Methods in Finance, Bressanone/Brixen, Italy, 2003. Editors: M. Fritelli, W. Runggaldier (2004) Vol. 1857: M. Émery, M. Ledoux, M. Yor (Eds.), Séminaire de Probabilités XXXVIII (2005) Vol. 1858: A.S. Cherny, H.-J. Engelbert, Singular Stochastic Differential Equations (2005) Vol. 1859: E. Letellier, Fourier Transforms of Invariant Functions on Finite Reductive Lie Algebras (2005) Vol. 1860: A. Borisyuk, G.B. Ermentrout, A. Friedman, D. Terman, Tutorials in Mathematical Biosciences I. Mathematical Neurosciences (2005) Vol. 1861: G. Benettin, J. Henrard, S. Kuksin, Hamiltonian Dynamics – Theory and Applications, Cetraro, Italy, 1999. Editor: A. Giorgilli (2005) Vol. 1862: B. Helffer, F. Nier, Hypoelliptic Estimates and Spectral Theory for Fokker-Planck Operators and Witten Laplacians (2005) Vol. 1863: H. Fürh, Abstract Harmonic Analysis of Continuous Wavelet Transforms (2005) Vol. 1864: K. Efstathiou, Metamorphoses of Hamiltonian Systems with Symmetries (2005) Vol. 1865: D. Applebaum, B.V. R. Bhat, J. Kustermans, J. M. Lindsay, Quantum Independent Increment Processes I. From Classical Probability to Quantum Stochastic Calcu-

lus. Editors: M. Schürmann, U. Franz (2005) Vol. 1866: O.E. Barndorff-Nielsen, U. Franz, R. Gohm, B. Kümmerer, S. Thorbjønsen, Quantum Independent Increment Processes II. Structure of Quantum Lévy Processes, Classical Probability, and Physics. Editors: M. Schürmann, U. Franz, (2005) Vol. 1867: J. Sneyd (Ed.), Tutorials in Mathematical Biosciences II. Mathematical Modeling of Calcium Dynamics and Signal Transduction. (2005) Vol. 1868: J. Jorgenson, S. Lang, Posn (R) and Eisenstein Sereies. (2005) Vol. 1869: A. Dembo, T. Funaki, Lectures on Probability Theory and Statistics. Ecole d’Eté de Probabilités de Saint-Flour XXXIII-2003. Editor: J. Picard (2005) Vol. 1870: V.I. Gurariy, W. Lusky, Geometry of Müntz Spaces and Related Questions. (2005) Vol. 1871: P. Constantin, G. Gallavotti, A.V. Kazhikhov, Y. Meyer, S. Ukai, Mathematical Foundation of Turbulent Viscous Flows, Martina Franca, Italy, 2003. Editors: M. Cannone, T. Miyakawa (2006) Vol. 1872: A. Friedman (Ed.), Tutorials in Mathematical Biosciences III. Cell Cycle, Proliferation, and Cancer (2006) Vol. 1873: R. Mansuy, M. Yor, Random Times and Enlargements of Filtrations in a Brownian Setting (2006) Vol. 1874: M. Yor, M. Émery (Eds.), In Memoriam PaulAndré Meyer - Séminaire de Probabilités XXXIX (2006) Vol. 1875: J. Pitman, Combinatorial Stochastic Processes. Ecole d’Eté de Probabilités de Saint-Flour XXXII-2002. Editor: J. Picard (2006) Vol. 1876: H. Herrlich, Axiom of Choice (2006) Vol. 1877: J. Steuding, Value Distributions of L-Functions (2006) Vol. 1878: R. Cerf, The Wulff Crystal in Ising and Percolation Models, Ecole d'Eté de Probabilit és de Saint-Flour XXXIV-2004. Editor: Jean Picard (2006) Vol. 1879: G. Slade, The Lace Expansion and its Applications, Ecole d'Eté de Probabilités de Saint-Flour XXXIV-2004. Editor: Jean Picard (2006) Vol. 1880: S. Attal, A. Joye, C.-A. Pillet, Open Quantum Systems I, The Hamiltonian Approach (2006) Vol. 1881: S. Attal, A. Joye, C.-A. Pillet, Open Quantum Systems II, The Markovian Approach (2006) Vol. 1882: S. Attal, A. Joye, C.-A. Pillet, Open Quantum Systems III, Recent Developments (2006) Vol. 1883: W. Van Assche, F. Marcellàn (Eds.), Orthogonal Polynomials and Special Functions, Computation and Application (2006) Vol. 1884: N. Hayashi, E.I. Kaikina, P.I. Naumkin, I.A. Shishmarev, Asymptotics for Dissipative Nonlinear Equations (2006) Vol. 1885: A. Telcs, The Art of Random Walks (2006) Vol. 1886: S. Takamura, Splitting Deformations of Degenerations of Complex Curves (2006) Vol. 1887: K. Habermann, L. Habermann, Introduction to Symplectic Dirac Operators (2006)

Recent Reprints and New Editions Vol. 1618: G. Pisier, Similarity Problems and Completely Bounded Maps. 1995 – Second, Expanded Edition (2001) Vol. 1629: J.D. Moore, Lectures on Seiberg-Witten Invariants. 1997 – Second Edition (2001) Vol. 1638: P. Vanhaecke, Integrable Systems in the realm of Algebraic Geometry. 1996 – Second Edition (2001) Vol. 1702: J. Ma, J. Yong, Forward-Backward Stochastic Differential Equations and their Applications. 1999. – Corrected 3rd printing (2005)

Introduction to Symplectic Dirac Operators

Introduction to Symplectic Dirac Operators

Introduction to Symplectic Dirac operators