QUANTUM ISOMETRY GROUPS

arXiv:0907.0618v1 [math.OA] 3 Jul 2009

JYOTISHMAN BHOWMICK

Indian Statistical Institute, Kolkata

June, 2009
QUANTUM ISOMETRY GROUPS

JYOTISHMAN BHOWMICK

Thesis submitted to the Indian Statistical Institute

in partial fulfillment of the requirements
for the award of the degree of
Doctor of Philosophy.
June,2009
Thesis Advisor: Debashish Goswami

Indian Statistical Institute

203, B.T. Road, Kolkata, India.
In the memory of my grandmother
 Acknowledgements

I would like to start by expressing my deepest gratitudes to my supervisor Debashish

Goswami who introduced me to the theories of noncommutative geometry and compact
quantum groups and whose mere presence acted as a psychological support to an extent
unknown even to him. I thank each and every faculty member of Stat Math unit, ISI
Kolkata from each of whom I learnt something or the other. I am grateful to Prof.
Shuzhou Wang, whose valuable comments and suggestions helped me to have a better
understanding about the contents of this thesis and also led to an improvement of the
work. I thank the National Board for Higher Mathematics, India for providing me with
partial financial support. I should also mention the names of Max Planck Institute
fur Mathematics of Bonn and Chern Institute of mathematics of Nankai University for
allowing me to attend two workshop and conference on Non commutative Geometry
and Quantum Groups hosted by them, from where I had valuable exposures about the
subjects. I thank my parents, sister, aunt and uncle for their continuous support during
the time of working on this thesis. I would have been unable to fix some critical latex
problem unless my friend Rajat Subhra Hazra had spent his valuable time on it. I am
grateful to Biswarup for the discussions I had with him on operator algebras. Many
of my other friends including Koushik Saha, Abhijit da, Pusti da, Ashis da, Subhra,
Subhajit and obviously Rajat were a continuous source of encouragement. I thank
Subhajit for allowing me to use his room and providing me with a steady supply of 13
Tzameti et al. Lastly, I would like to mention the names of Subhra and Rajat again for
bearing with my bhoyonkor and haayre during a critical part of my thesis work.
Contents

Notations 1

0 Introduction 3

1 Preliminaries 9
1.1 Operator algebras and Hilbert modules . . . . . . . . . . . . . . . . . . . 9
1.1.1 C ∗ algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.1.2 von Neumann algebras . . . . . . . . . . . . . . . . . . . . . . . . 12
1.1.3 Free product and tensor product . . . . . . . . . . . . . . . . . . 13
1.1.4 Hilbert modules . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.2 Quantum Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.2.1 Hopf algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.2.2 Compact Quantum Groups: basic definitions and examples . . . 18
1.2.3 The CQG Uµ (2) . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.2.4 The CQG SUµ (2) . . . . . . . . . . . . . . . . . . . . . . . . . . 26
1.2.5 The Hopf ∗-algebras O(SUµ (2)) and Uµ (su(2)) . . . . . . . . . . 29
1.2.6 The Wang algebras . . . . . . . . . . . . . . . . . . . . . . . . . . 31
1.2.7 Action of a compact quantum group on a C ∗ algebra . . . . . . . 31
1.3 Rieffel deformation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
1.3.1 Rieffel Deformation of compact quantum group . . . . . . . . . . 38
1.4 Classical Riemannian geometry . . . . . . . . . . . . . . . . . . . . . . . 42
1.4.1 Classical Hilbert space of forms . . . . . . . . . . . . . . . . . . . 42
1.4.2 Isometry groups of classical manifolds . . . . . . . . . . . . . . . 43
1.4.3 Spin Groups and Spin manifolds . . . . . . . . . . . . . . . . . . 44
1.4.4 Dirac operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.5 Noncommutative Geometry . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.1 Spectral triples . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
1.5.2 The space of forms in noncommutative geometry . . . . . . . . . 48
1.5.3 Laplacian in Noncommutative geometry . . . . . . . . . . . . . . 50

v
2 Quantum isometry groups: approach based on Laplacian 51
2.1 Formulation of the quantum isometry group . . . . . . . . . . . . . . . . 51
2.1.1 Characterization of isometry group for a compact Riemannian
manifold . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.1.2 The definition and existence of the quantum isometry group . . . 52
2.2 Computation of QISOL . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
2.2.1 The commutative spheres . . . . . . . . . . . . . . . . . . . . . . 55
2.2.2 The commutative one-torus . . . . . . . . . . . . . . . . . . . . . 57
2.2.3 The commutative n-tori . . . . . . . . . . . . . . . . . . . . . . . 59

3 Quantum group of orientation preserving Riemannian isometries 67

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.2 Definition and existence of the quantum group of orientation-preserving
isometries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.2.1 The classical case . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.2.2 Quantum group of orientation-preserving isometries of an R-
twisted spectral triple . . . . . . . . . . . . . . . . . . . . . . . . 74
3.2.3 Stability and topological action . . . . . . . . . . . . . . . . . . . 80
3.2.4 Universal object in the categories Q or Q0 . . . . . . . . . . . . 84
3.3 Comparison with the approach of [30] based on Laplacian . . . . . . . . 87
3.4 Examples and computations . . . . . . . . . . . . . . . . . . . . . . . . . 91
3.4.1 Equivariant spectral triple on SUµ (2) . . . . . . . . . . . . . . . 91
3.4.2 A commutative example : spectral triple on T2 . . . . . . . . . . 99
3.5 QISO+ for zero dimensional manifolds . . . . . . . . . . . . . . . . . . 102
3.5.1 Inductive limit construction for quantum isometry groups . . . . 102
3.5.2 Quantum isometry groups for spectral triples on AF algebras . . 103

4 Quantum isometry groups for Rieffel deformed manifolds 107

4.1 Deformation of spectral triple . . . . . . . . . . . . . . . . . . . . . . . . 107
4.2 Some preparatory results . . . . . . . . . . . . . . . . . . . . . . . . . . 110
g + of a Rieffel deformed noncommutative manifold . . .
4.3 QISO . . . . . . 119
R
4.3.1 Derivation of the result . . . . . . . . . . . . . . . . . . . . . . . 119
4.3.2 Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
4.4 QISOL of a Rieffel deformed noncommutative manifold . . . . . . . . . 127
4.4.1 Derivation of the main result . . . . . . . . . . . . . . . . . . . . 127
4.4.2 Computations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

5 Quantum isometry groups of the Podles spheres 139

5.1 The Podles Spheres . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
 5.1.1 The original definition by Podles . . . . . . . . . . . . . . . . . . 139
5.1.2 The Podles spheres as in [24] . . . . . . . . . . . . . . . . . . . . 140
5.1.3 The Podles spheres as in [37] . . . . . . . . . . . . . . . . . . . . 141
5.1.4 The description as in [50] . . . . . . . . . . . . . . . . . . . . . . 142
5.1.5 Haar functional on the Podles spheres . . . . . . . . . . . . . . . 143
5.2 Spectral triples on the Podles spheres . . . . . . . . . . . . . . . . . . . 145
5.2.1 The spectral triple as in [24] . . . . . . . . . . . . . . . . . . . . . 145
5.2.2 SUµ (2) equivariance of the spectral triple . . . . . . . . . . . . . 146
5.2.3 The CQG SOµ (3) and its action on the Podles sphere . . . . . 147
5.3 Quantum Isometry Groups of the Podles sphere . . . . . . . . . . . . . . 148
5.3.1 Linearity of the action . . . . . . . . . . . . . . . . . . . . . . . . 150
5.3.2 Homomorphism conditions . . . . . . . . . . . . . . . . . . . . . 154
5.3.3 Relations from the antipode . . . . . . . . . . . . . . . . . . . . . 156
5.3.4 Identification of SOµ (3) as the quantum isometry group . . . . . 158
5.3.5 The quantum isometry group in the case µ = 1 . . . . . . . . . . 165
5.3.6 Existence of QISO g + (D) . . . . . . . . . . . . . . . . . . . . . . 167
5.4 The spectral triple by Chakraborty and Pal on Sµ,c 2 ,c > 0 . . . . . . . . 169
5.4.1 The spectral triple . . . . . . . . . . . . . . . . . . . . . . . . . . 169
5.4.2 Computation of the quantum isometry group . . . . . . . . . . . 171

Bibliography 185
1

Notations

IN The set of natural numbers

IN 0 IN ∪ {0}
R The set of real numbers
C The set of complex numbers
Mn (C) The set of all n × n complex matrices
S1 The circle group
Tn The n-torus
ev Evaluation map
id The identity map
l2 (IN ) The Hilbert space of square summable sequences
C ∞ (M ) The space of smooth functions on a smooth manifold M
Cc∞ (M ) The space of compactly supported smooth functions on M
V1 ⊗alg V2 Algebraic tensor product of two vector spaces V1 and V2
A⊗B Minimal tensor product of two C ∗ algebras A and B
M(A) The multiplier algebra of a C ∗ algebra A
L(E, F ) The space of adjointable maps from Hilbert modules E to F
L(E) The space of adjointable maps from a Hilbert module E to itself
K(E, F ) The space of compact operators from Hilbert modules E to F
K(E) The space of compact operators from a Hilbert module E to itself
B(H) The set of all bounded linear operators on a Hilbert space H
A∗B Free product of two C ∗ algebras A and B
G∗H Free product of two groups G and H
G >
H Semi direct product of two groups G and H
2
Chapter 0

Introduction

The theme of this thesis lies on the interface of two areas of the so called “ noncommuta-
tive mathematics ”, namely noncommutative geometry (NCG) a la Connes, cf [17] and
the theory of ( C ∗ -algebraic ) compact quantum groups (CQG) a la Woronowicz, cf [67]
which are generalizations of classical Riemannian spin geometry and that of compact
topological groups respectively.
The root of NCG can be traced back to the Gelfand Naimark theorem which says
that there is an anti-equivalence between the category of (locally) compact Hausdorff
spaces and (proper, vanishing at infinity) continuous maps and the category of (not
necessarily) unital C ∗ algebras and ∗-homomorphisms. This means that the entire
topological information of a locally compact Hausdorff space is encoded in the com-
mutative C ∗ algebra of continuous functions vanishing at infinity. This motivates one
to view a possibly noncommutative C ∗ algebra as the algebra of “functions on some
noncommutative space”.
In classical Riemannian geometry on spin manifolds, the Dirac operator on the
Hilbert space L2 (S) of square integrable sections of the spinor bundle contains a lot of
geometric information. For example, the metric, the volume form and the dimension of
the manifold can be captured from the Dirac operator. This motivated Alain Connes to
define his noncommutative geometry with the central object as the spectral triple which
is a triplet (A, H, D) where H is a separable Hilbert space, A is a (not necessarily closed)
∗-algebra of B(H), D is a self adjoint (typically unbounded) operator ( sometimes called
the Dirac operator of the spectral triple ) such that [D, a] admits a bounded extension.
This generalizes the classical spectral triple (C ∞ (M ), L2 (S), D) on any Riemannian spin
manifold M , where D denotes the usual Dirac operator.
On the other hand, quantum groups have their origin in different problems in math-
ematical physics as well as the theory of classical locally compact groups. It was S.L.
Woronowicz, who in [66] and [67] was able to pinpoint a set of axioms for defining

3
Chapter 0: Introduction 4

compact quantum groups (CQG for short) as the correct generalization of compact
topological groups.
The idea of a group acting on a space was extended to the idea of a CQG co-
acting on a noncommutative space (that is, a possibly noncommutative C ∗ algebra ).
Following suggestions of Alain Connes, Shuzhou Wang in [60] defined and proved the
existence of quantum automorphism groups on finite dimensional C ∗ algebras. Since
then, many interesting examples of such quantum groups, particularly the quantum per-
mutation groups of finite sets and finite graphs, were extensively studied by a number of
mathematicians (see for example [1], [2], [3], [4], [5], [61] and references therein ). The
underlying basic principle of defining a quantum automorphism group corresponding
to some given mathematical structure (for example a finite set, a graph, a C ∗ or von
Neumann algebra) consists of two steps: first, to identify (if possible) the group of auto-
morphism of the structure as a universal object in a suitable category, and then, try to
look for the universal object in a similar but bigger category, replacing groups by quan-
tum groups of appropriate type. However, most of the work done by them concerned
some kind of quantum automorphism group of a ‘finite’ structure, for example, of finite
sets or finite dimensional matrix algebras. It was thus quite natural to try to extend
these ideas to the ‘infinite’ or ‘continuous’ mathematical structures, for example classi-
cal and noncommutative manifolds. With this motivation, Goswami ( [30] ) formulated
and studied the quantum analogues of the group of Riemannian isometries called the
quantum isometry group. Classically, an isometry is characterized by the fact that its
action commutes with the Laplacian. Therefore, to define the quantum isometry group,
it is reasonable to consider a category of compact quantum groups which act on the
manifold (or more generally on a noncommutative manifold given by a spectral triple)
with the action commuting with the Laplacian, say L, coming from the spectral triple.
It is proven in [30] that a universal object in the category ( denoted by Q0L ) of such
quantum groups does exist (denoted by QISOL ) if one makes some mild assumptions
on the spectral triple all of which are valid for a compact Riemannian spin manifold.
The work of this thesis starts with the computation of the quantum isometry group in
several commutative and noncommutative examples ( [6], [7]).
However, the formulation of quantum isometry groups in [30] had a major drawback
from the view point of noncommutative geometry, since it needed a ‘good’ Laplacian to
exist. In noncommutative geometry it is not always easy to verify such an assumption
about the Laplacian, and thus it would be more appropriate to have a formulation in
terms of the Dirac operator directly. This is what is done in [8] where the notion of
a quantum group analogue of the group of orientation preserving isometries was given
and its existence as the universal object in a suitable category was proved. Then, a
number of computations for this quantum group were done in [8], [9] and [10].
5

Now we try to give an idea of the contents of each of the chapters. In chapter 1, we
discuss the concepts and results needed in the later chapters of the thesis. For the sake
of completeness, we begin with a glimpse of operator algebras and Hilbert modules, free
product and tensor products of C ∗ algebras and some examples. The next section is
on quantum groups which we start with the basics of Hopf algebras and then define
compact quantum groups (CQG) and give relevant definitions and properties including
a brief review of Peter Weyl theory. After that, we introduce the quantum groups
Uµ (2), SUµ (2), Au (Q) and Uµ (su(2)). We end this section by introducing the notion of
a C ∗ action of a compact quantum group on a C ∗ algebra and giving an account of
Shuzhou Wang’s work in [60]. The next section is on Rieffel deformation where we
recall a part of the work done in [46] and [62] which are relevant to us. We describe
some important examples which are going to appear in chapter 4. The 4th section is
on classical Riemannian geometry which includes, among other things, the definition
and properties of Dirac operator which will serve as a motivation for the definition of ‘
spectral triple’ in the 5th section. This section also contains a subsection on isometry
groups of classical Riemannian manifold in which the characterizing property of an
isometry, in the form given in [30], is given. In the 5th section, we define spectral triples,
give examples of them, introduce the Hilbert space of forms (as in [29]), noncommutative
volume form and the notion of Laplacian in noncommutative geometry.
Chapter 2 is on the Laplacian-based approach to quantum isometry groups as pro-
posed in [30]. Here we recall the formulation of quantum isometry groups from [30] and
then compute them for the space of continuous functions on the classical 2-spheres, the
circle and the n-tori. In each of these cases, the quantum isometry group turns out to be
the same as the classical ones, that is, C(O(3)), C(S 1 >
Z2 ) and C(Tn >
(Zn2 >
Sn ))
respectively ( Sn being the permutation group on n symbols ).
Chapter 3 deals with the quantum group of orientation preserving isometries. The
classical situation is stated clearly, which will serve as a motivation for the quantum
formulation. Then, the quantum group of orientation-preserving isometries of an R-
twisted spectral triple is defined ( see [31] for the definition of an R-twisted spectral
triple ) and its existence is proven. Given an R-twisted spectral triple (A∞ , H, D) of
compact type, we consider a category Q0 of pairs (Q, U ) where Q is a compact quantum
group which has a unitary (co)-representation U on H commuting with D, and such
that for all state φ on Q, (id ⊗ φ)adU maps A∞ inside A00∞ . Moreover, let Q0R be a
subcategory of Q0 consisting of those (Q, U ) for which adU preserves the R-twisted
volume form. In section 3.2, we have proved that Q0 R has a universal object to be
denoted by QISO^+ (D). The Woronowicz C ∗ subalgebra of QISO ^+ (D) generated by
R R
elements of the form hadU (a)(η ⊗ 1), η 0 ⊗ 1i ^+ (D) where η, η 0 are in H, a is in A∞
QISOR
Chapter 0: Introduction 6

and h. , .i ^+ (D) valued inner product of H ⊗ QISO

denotes the QISO ^+ (D)
^+ (D)
QISO R R
R
is defined to be the quantum group of orientation and volume preserving isometries of
+
the spectral triple (A∞ , H, D) and is denoted by QISOR (D). The next section explores
the conditions under which the action of this compact quantum group keeps the C ∗
algebra invariant and is a C ∗ action. Moreover, we have given some sufficient conditions
under which the universal object in the bigger category Q0 exists which is denoted by
QISO^ + (D) and the corresponding Woronowicz C ∗ subalgebra as above is denoted by

QISO+ (D). After this, we compare this approach with the Laplacian-based approach
in [30]. We obtain the following results:
( 1 ) Under some reasonable conditions, QISOI+ (D) is a sub-object of QISOL in
the category Q0L .
( 2 ) QISOL is isomorphic to a quantum subgroup of QISOI+ corresponding to the
Hodge Dirac operator coming from D.
( 3 ) Moreover, under some conditions which are valid for compact spin manifolds,
QISOL and the QISOI+ of the Hodge Dirac operator are isomorphic.
The next section is on examples and computations. To begin with, the compact
quantum group Uµ (2) is identified as the QISO+ of SUµ (2) corresponding to the spectral
triple constructed by Chakraborty and Pal in [13]. Then we derive that QISO+ for the
classical spectral triple on C(T2 ) is C(T2 ) itself. We end the chapter by showing that
QISO+ of spectral triples associated with some approximately finite dimensional C ∗
algebras arise as the inductive limit of QISO+ of the constituent finite dimensional
algebras. The results of this chapter are taken from [8] and [9].
+
Chapter 4 is about the QISOL and QISOR of a Rieffel deformed noncommutative
manifold. We first discuss the isospectral deformation of a spectral triple, followed by
the proof of some preparatory technical results which will be needed later. Then in the
+
final section we prove that QISOR and QISOL of a Rieffel deformed ( noncommutative
+
) manifold is a Rieffel deformation of the QISOR and QISOL ( respectively ) of the
original ( undeformed ) manifold.
In chapter 5, we compute the quantum group of orientation preserving isometries
for two different families of spectral triples on the Podles spheres, one constructed by
Dabrowski et al in [24] and the other by Chakraborty and Pal in [14]. We start by
giving the different descriptions of the Podles spheres ( as in [43], [24], [37], and [50] )
and the formula for the Haar functional on it. Then we introduce the spectral triples
on the Podles spheres as in [24] and show that it is indeed SUµ (2) equivariant and R-
twisted ( for a suitable R ). After this, the compact quantum group SOµ (3) is defined
and its action on the Podles sphere is discussed. In the 3rd section, the computation
+
for identifying SOµ (3) as QISOR for this spectral triple is given. In the 4th section,
the spectral triple defined in [14] is introduced and then the corresponding QISO+
7

is computed. In particular, it follows that QISO+ in general may not be a matrix

quantum group and that it may not have a C ∗ action.
Chapter 0: Introduction 8
Chapter 1

Preliminaries

1.1 Operator algebras and Hilbert modules

We presume the reader’s familiarity with the theory of operator algebras and Hilbert
modules. However, for the sake of completeness, we give a sketchy review of some basic
definitions and facts and refer to [56], [39] for the details. Throughout this thesis, all
algebras will be over C unless otherwise mentioned.

1.1.1 C ∗ algebras
A C ∗ algebra A is a Banach ∗-algebra satisfying the C ∗ property : kx∗ xk = kxk2 for
all x in A. The algebra A is said to be unital or non-unital depending on whether it
has an identity or not. Every commutative C ∗ algebra A is isometrically isomorphic
to the C ∗ algebra C0 (X) consisting of complex valued functions on a locally compact
Hausdorff space X vanishing at infinity( Gelfand’s theorem ). An arbitrary ( possibly
noncommutative ) C ∗ algebra is isometrically isomorphic to a C ∗ -subalgebra of B(H),
the set of all bounded operators on a Hilbert space H.
For x in A, the spectrum of x, denoted by σ(x) is defined as the complement of
the set {z ∈ C : (z1 − x)−1 ∈ A}. An element x in A is called self adjoint if x = x∗ ,
normal if x∗ x = xx∗ , unitary if x∗ = x−1 , projection if x = x∗ = x2 and positive if
x = y ∗ y for some y in A. When x is normal, there is a continuous functional calculus
sending f in C(σ(x)) to f (x) in A. where f 7→ f (x) is a ∗ isometric isomorphism from
C(σ(x)) onto C ∗ (x).
A linear map between two C ∗ algebras is said to be positive if it maps positive
elements to positive elements. A positive linear functional φ such that φ(1) = 1 is
called a state on A. A state φ is called a trace if φ(ab) = φ(ba) for all a, b in A
and faithful if φ(x∗ x) = 0 implies x = 0. Given a state φ on a C ∗ algebra A, there
exists a triple (called the GNS triple) (Hφ , πφ , ξφ ) consisting of a Hilbert space Hφ , a ∗

9
Chapter 1: Preliminaries 10

representation πφ of A into B(Hφ ) and a vector ξφ in Hφ which is cyclic in the sense

that {πφ (x)ξφ : x ∈ A} is total in Hφ satisfying

φ(x) = hξφ , πφ (x)ξφ i .

For a two-sided norm closed ideal I of a C ∗ algebra A, the canonical quotient norm
on the Banach space A/I is in fact the unique C ∗ norm making A/I into a C ∗ algebra.
Here we prove two results which we are going to need later on.

Lemma 1.1.1. [30] Let C be a C ∗ algebra and F be a nonempty collection of C ∗ -ideals

( closed two-sided ideals ) of C. Then for any x in C, we have

supI∈F kx + Ik = kx + I0 k ,

where I0 denotes the intersection of all I in F and kx + Ik = inf{kx − yk : y ∈ I}

denotes the norm in C/I.

Proof : It is clear that supI∈F kx + Ik defines a norm on C/I0 , which is in fact a C ∗

norm since each of the quotient norms k. + Ik is so. Thus the lemma follows from the
uniqueness of C ∗ norm on the C ∗ algebra C/I0 . 2

Lemma 1.1.2. Let C be a unital C ∗ algebra and F be a nonempty collection of C ∗ -

ideals (closed two-sided ideals) of C. Let I0 denote the intersection of all I in F, and
let ρI denote the map C/I0 3 x + I0 7→ x + I ∈ C/I for I in F. Denote by Ω the set
{ω ◦ ρI , I ∈ F, ω state on C/I}, and let K be the weak-∗ closure of the convex hull
S
of Ω (−Ω). Then K coincides with the set of bounded linear functionals ω on C/I0
satisfying kωk = 1 and ω(x∗ + I0 ) = ω(x + I0 ).

Proof : We will use Lemma 1.1.1. Clearly, K is a weak-∗ compact, convex subset of
the unit ball (C/I0 )∗1 of the dual of C/I0 , satisfying −K = K. If K is strictly smaller
than the self-adjoint part of unit ball of the dual of C/I0 , we can find a state ω on
C/I0 which is not in K. Considering the real Banach space X = (C/I0 )∗s.a. and using
standard separation theorems for real Banach spaces (for example, Theorem 3.4 of [49],
page 58), we can find a self-adjoint element x of C such that kx + I0 k = 1, and

sup ω 0 (x + I0 ) < ω(x + I0 ).

ω 0 ∈K

Let γ belonging to R be such that supω0 ∈K ω 0 (x + I0 ) < γ < ω(x + I0 ). Fix 0 <  <
ω(x+I0 )−γ, and let I be an element of F be such that kx+I0 k− 2 ≤ kx+Ik ≤ kx+I0 k.
Let φ be a state on C/I such that kx + Ik = |φ(x + I)|. Since x is self-adjoint, either
φ(x + I) or −φ(x + I) equals kx + Ik, and φ0 := ±φ ◦ ρI , where the sign is chosen so
11 Operator algebras and Hilbert modules

that φ0 (x + I0 ) = kx + Ik. Thus, φ0 is in K, so kx + I0 k = φ0 (x + I) ≤ γ < ω(x + I0 ) − .

But this implies kx + I0 k ≤ kx + Ik + 2 < ω(x + I0 ) − 2 ≤ kx + I0 k −  (since ω is a
state), which is a contradiction completing the proof. 2

For a C ∗ algebra A ( possibly non unital ), its multiplier algebra, denoted by M(A),
is defined as the maximal C ∗ algebra which contains A as an essential two sided ideal,
that is, A is an ideal in M(A) and for y in M(A), ya = 0 for all a in A implies y = 0.
The norm of M(A) is given by kxk = supa∈A,kak≤1 {kxak , kaxk}. There is a locally
convex topology called the strict topology on M(A), which is given by the family of
seminorms {k.ka , a ∈ A}, where kxka = Max(kxak , kaxk), for x in M(A). M(A) is the
completion of A in the strict topology.

We now come to the inductive limit of C ∗ algebras . Let I be a directed set

and {Ai }i∈I be a family of C ∗ algebras equipped with a family of C ∗ homomorphisms
Φij : Aj → Ai ( when j < i ) such that Φij = Φik Φkj when j < k < i. Then there
exists a unique C ∗ algebra denoted by limi Ai and C ∗ homomorphisms φi : Ai → limi Ai
with the universal property that given any other C ∗ algebra A0 and C ∗ homomorphisms
ψi : Ai → A0 satisfying ψj = ψi Φij for j < i, then there exists unique C ∗ homomorphism
χ : limi Ai → A0 satisfying χφi = ψi . limi Ai is called the inductive limit C ∗ algebra
corresponding to the inductive system (Ai , Φij ). Inductive limit of a sequence of finite
dimensional C ∗ algebras are called approximately finite dimensional C ∗ algebras or AF
algebras.

A large class of C ∗ algebras are obtained by the following construction. Let A0

be an associative ∗-algebra without any a-priori norm such that the set F = {π :
A0 → B(Hπ ) ∗ −homomorphism, Hπ a Hilbert space} is non empty and k.ku defined
by kaku = sup{kπ(a)k : π ∈ F} is finite for all a. Then the completion of A0 in k.ku is
a C ∗ algebra known as the universal C ∗ algebra corresponding to A0 .

Example 1: Noncommutative two-torus

Let θ belongs to [0, 1]. Consider the ∗ algebra A0 generated by two unitary symbols
U and V satisfying the relation U V = e2πiθ V U. It has a representation π on the Hilbert
space L2 (S 1 ) defined by π(U )(f )(z) = f (e2πiθ z), π(V )(f )(z) = zf (z) where f is in
L2 (S 1 ), z is in S 1 . Then kaku is finite for all a in A0 . The resulting C ∗ algebra is called
noncommutative two-torus and denoted by Aθ .

Example 2: Group C ∗ algebra

Let G be a locally compact group with left Haar measure µ. One can make L1 (G)
Chapter 1: Preliminaries 12

into a Banach ∗-algebra by defining

Z
(f ∗ g)(t) = f (s)g(s−1 t)dµ(s),
G

f ∗ (t) = ∆(t)−1 f (t−1 ).

Here f, g are in L1 (G), ∆ is the modular homomorphism of G.

L1 (G) has a distinguished representation πreg on L2 (G) defined by πreg (f ) =
f (t)π(t)dµ(t) where π(t) is a unitary operator on L2 (G) defined by (π(t)f )(g) =
R

f (t−1 g) (f ∈ L2 (G), t, g ∈ G). The reduced group C ∗ algebra of G is defined to be

B(L2 (G))
Cr∗ (G) := πreg (L1 (G)) .

Remark 1.1.3. For G abelian, we have Cr∗ (G) ∼

= C0 (G)
b where G
b is the group of
characters on G.

One can also consider the universal C ∗ algebra described before corresponding to
the Banach ∗-algebra L1 (G). This is called the free or full group C ∗ algebra and denoted
by C ∗ (G).

Remark 1.1.4. For the so-called amenable groups ( which include compact and abelian
groups ) we have C ∗ (G) ∼
= Cr∗ (G).

1.1.2 von Neumann algebras

We recall that for a Hilbert space H, the strong operator topology ( SOT ) , the
weak operator topology ( WOT ) and the ultra weak topology are the locally
convex topologies on B(H) given by families of seminorms F1 , F2 , F3 respectively where
F1 = {pξ : ξ ∈ H}, F2 = {pξ,η : ξ, η ∈ H}, F3 = {pρ : ρ is a trace class operator on H}
and pξ (x) = kxξk , pξ,η (x) = |hxξ, ηi| , pρ (x) = |Tr(xρ)| ( where Tr denotes the usual
trace on B(H)).
Now we state a well known fact.

Lemma 1.1.5. If Tn is a sequence of bounded operators converging to zero in SOT,

then for any trace class operator W, T r(Tn W ) → 0 as n → ∞.

For any subset B of B(H), we denote by B 0 the commutant of B, that is, B 0 = {x ∈

B(H) : xb = bx for all b ∈ B}. A unital C ∗ subalgebra A ⊆ B(H) is called a von
Neumann algebra if A = A00 which is equivalent to being closed in any of the three
topologies mentioned above.
A state φ on a von Neumann algebra A is called normal if φ(xα ) increases to φ(x)
whenever xα increases to x. A state φ on A is normal if and only if there is a trace class
13 Operator algebras and Hilbert modules

operator ρ on H such that φ(x) = Tr(ρx) for all x in A. More generally, we call a linear
map Φ : A → B ( where B is a von Neumann algebra ) normal if whenever xα increases
to x for a net xα of positive elements from A, one has that Φ(xα ) increases to Φ(x) in
B. It is known that a positive linear map is normal if and only if it is continuous with
respect to the ultra-weak-topology. In view of this fact, we shall say that a bounded
linear map between two von Neumann algebras is normal if it is continuous with respect
to the respective ultra-weak topologies.

1.1.3 Free product and tensor product

If (Ai )i∈I is a family of unital C ∗ algebras, then their unital C ∗ algebra free product
∗i∈I Ai is the unique C ∗ algebra A together with unital ∗-homomorphism ψi : Ai → A
such that given any unital C ∗ algebra B and unital ∗-homomorphisms φi : Ai → B there
exists a unique unital ∗-homomorphism Φ : A → B such that φi = Φ ◦ ψi .

Remark 1.1.6. It is a direct consequence of the above definition that given a family of
C ∗ homomorphisms φi from Ai to B, there exists a C ∗ homomorphism ∗i φi such that
(∗i φi ) ◦ ψi = φi for all i.

Remark 1.1.7. We recall that for discrete groups {Gi }i∈I , C ∗ (∗i∈I Gi ) ∼
= ∗i∈I C ∗ (Gi ).

For A and B two algebras, we will denote the algebraic tensor product of A and
B by the symbol A ⊗alg B. When A and B are C ∗ algebras, there is more than one
norm on A ⊗alg B so that the completion with respect to that norm is a C ∗ algebra.
Throughout this thesis, we will work with the so called injective tensor product, that is,
the completion of A⊗alg B with respect to the norm given on A⊗alg B by k ni=1 ai ⊗bi k =
P

supk ni=1 π1 (ai ) ⊗ π2 (bi )kB(H1 ⊗H2 ) where ai is in A, bi is in B and the supremum
P

runs over all possible choices of (π1 , H1 ), (π2 , H2 ) where H1 , H2 are Hilbert spaces
and π1 : A → B(H1 ) and π2 : A2 → B(H2 ) are ∗-homomorphisms. When A ⊆
B(H1 ), B ⊆ B(H2 ) are von Neumann algebras, then by the notation A ⊗ B, we mean
the von Neumann algebra tensor product, that is, the WOT closure of A ⊗alg B in
B(H1 ⊗ H2 ). We refer to [56] for more details.
We now prove a useful general fact.

Lemma 1.1.8. Let A be a C ∗ algebra and ω, ωj (j = 1, 2, ...) be states on A such that

ωj → ω in the weak-∗ topology of A∗ . Then for any separable Hilbert space H and for
all Y in M(K(H) ⊗ A), we have (id ⊗ ωj )(Y ) → (id ⊗ ω)(Y ) in the strong operator
topology.

Proof: Clearly, (id ⊗ ωj )(Y ) → (id ⊗ ω)(Y ) (in the strong operator topology) for
all Y in Fin(H) ⊗alg A, where Fin(H) denotes the set of finite rank operators on H.
Using the strict density of Fin(H) ⊗alg A in M(K(H) ⊗ A), we choose, for a given Y in
Chapter 1: Preliminaries 14

M(K(H) ⊗ A), ξ in H with kξk = 1, and δ > 0, an element Y0 in Fin(H) ⊗alg A such
that k(Y − Y0 )(|ξ >< ξ| ⊗ 1)k < δ. Thus,

k(id ⊗ ωj )(Y )ξ − (id ⊗ ω)(Y )ξk

= k(id ⊗ ωj )(Y (|ξ >< ξ| ⊗ 1))ξ − (id ⊗ ω)(Y (|ξ >< ξ| ⊗ 1))ξk
≤ k(id ⊗ ωj )(Y0 (|ξ >< ξ| ⊗ 1))ξ − (id ⊗ ω)(Y0 (|ξ >< ξ| ⊗ 1))ξk
+ 2k(Y − Y0 )(|ξ >< ξ| ⊗ 1)k
≤ k(id ⊗ ωj )(Y0 (|ξ >< ξ| ⊗ 1))ξ − (id ⊗ ω)(Y0 (|ξ >< ξ| ⊗ 1))ξk + 2δ,

from which it follows that (id ⊗ ωj )(Y ) → (id ⊗ ω)(Y ) in the strong operator topology.
2

Let A and B be two unital ∗-algebras. Then a linear map T from A to B is called
n-positive if T ⊗ Idn : A ⊗ Mn (C) → B ⊗ Mn (C) is positive for all k ≤ n but not
necessarily for k = n + 1. T is said to be completely positive ( CP for short ) if it is
n-positive for all n. It is a well known result that for a CP map T : A → B(H), one has
the following operator inequality for all a in A:

T (a)∗ T (a) ≤ kT (1)k T (a∗ a). (1.1.1)

Tensor product and composition of two CP maps are CP. The following is an useful
result about CP maps.

Proposition 1.1.9. If A and B are C ∗ algebras with A commutative, φ is a positive

map from A to B, then φ is CP. The same holds if φ is from B to A.

1.1.4 Hilbert modules

Given a ∗-subalgebra A ⊆ B(H) ( where H is a Hilbert space ), a semi-Hilbert A module
E is a right A-module equipped with a sesquilinear map h. , .i : E × E → A satisfying
hx, yi∗ = hy, xi , hx, yai = hx, yi a and hx, xi ≥ 0 for x, y in E and a in A. A semi
Hilbert module is called a pre-Hilbert module if hx, xi = 0 if and only if x = 0. It is
called a Hilbert module if furthermore, A is a C ∗ algebra and E is complete in the norm
1
x 7→ khx, xik 2 where k.k is the C ∗ norm of A.
The simplest examples of Hilbert A modules are the so called trivial A modules of
the form H ⊗ A where H is a Hilbert space with an A valued sesquilinear form defined
on H ⊗alg A by : hξ ⊗ a, ξ 0 ⊗ a0 i = hξ, ξ 0 i a∗ a0 . The completion of H ⊗alg A with respect
to this pre Hilbert module structure is a Hilbert A module and is denoted by H ⊗ A.
We recall that for a pre Hilbert A module E ( A is a C ∗ algebra ), the Cauchy
Schwarz inequality holds in the following form: 0 ≤ hx, yi hy, xi ≤ hx, xi khy, yik .
15 Quantum Groups

Let E and F be two Hilbert A modules. We say that a C linear map L from E to F is
adjointable if there exists a C linear map L∗ from F to E such that hL(x), yi = hx, L∗ (y)i
for all x in E, y in F. We call L∗ the adjoint of L. The set of all adjointable maps
from E to F is denoted by L(E, F ). In case, E = F, we write L(E) for L(E, E). For
an adjointable map L, both L and L∗ are automatically A-linear and norm bounded
maps between Banach spaces. We say that an element L in L(E, F ) is an isometry if
hLx, Lyi = hx, yi for all x, y in E. L is said to be a unitary if L is isometry and its range
is the whole of F. One defines a norm on L(E, F ) by kLk = supx∈R, kxk≤1 kL(x)k . L(E)
is a C ∗ algebra with this norm.
There is a topology on L(E, F ) given by a family of seminorms {k.kx , k.ky : x ∈

1

1
E, y ∈ F } ( where ktkx = htx, txi 2 and ktky = ht∗ y, t∗ yi 2 ) known as the strict

topology. For x in E, y in F, we denote by θx,y the element of L(E, F ) defined by

θx,y (z) = y hx, zi ( where z is in E ). The norm closure ( in L(E, F ) ) of the A linear
span of {θx,y : x ∈ E, y ∈ F } is called the set of compact operators and denoted
by K(E, F ) and we denote K(E, E) by K(E). These are not necessarily compact in
the sense of compact operators between two Banach spaces. One has the following
important result:

Proposition 1.1.10. The multiplier algebra M(K(E)) of K(E) is isomorphic with

L(E) for any Hilbert module E.

Using this, for a possibly non-unital C ∗ algebra B, we often identify an element V of

M(K(H) ⊗ B) with the map from H to H ⊗ B which sends a vector ξ of H to V (ξ ⊗ 1B )
in H ⊗ B.
Given a Hilbert space H and a C ∗ algebra A, and a unitary element U of M(K(H)⊗
A), we shall denote by αU (≡ adU ) the ∗-homomorphism αU (X) = U e (X ⊗ 1)Ue ∗ for
X belonging to B(H). For a not necessarily bounded, densely defined (in the weak
operator topology) linear functional τ on B(H), we say that αU preserves τ if αU maps
a suitable (weakly) dense ∗-subalgebra (say D) in the domain of τ into D ⊗alg A and
(τ ⊗ id)(αU (x)) = τ (x).1A for all x in D. When τ is bounded and normal, this is
equivalent to (τ ⊗ id)(αU (x)) = τ (x)1A for all x belonging to B(H).
We say that a (possibly unbounded) operator T on H commutes with U if T ⊗ I
(with the natural domain) commutes with U e . Sometimes such an operator will be called
U -equivariant.

1.2 Quantum Groups

In this section, we will recall the basics of Hopf algebras and then define compact
quantum groups ( as in [67], [66] ). After that, we will discuss a few examples of
Chapter 1: Preliminaries 16

quantum groups and the concept of an action of a compact quantum group on a C ∗

algebra. For more detailed discussion, we refer to [37], [35], [15] [42] and references
therein. In this thesis, we will be concerned about compact quantum groups only. For
other types of quantum groups, we refer to [37], [38], [58] etc.

1.2.1 Hopf algebras

We recall that an associative algebra with an unit is a vector space A over C together
with two linear maps m : A ⊗ A → A called the multiplication or the product and η :
C → A called the unit such that m(m⊗id) = m(id⊗m) and m(η ⊗id) = id = m(id⊗η).
Dualizing this, we get the following definition.

Definition 1.2.1. A coalgebra A is a vector space over C equipped with two linear
maps ∆ : A → A ⊗ A called the comultiplication or coproduct and  : A → C such that

(∆ ⊗ id)∆ = (id ⊗ ∆)∆,

( ⊗ id)∆ = id = (id ⊗ )∆.

Definition 1.2.2. Let (A, ∆A , A ) and (B, ∆B , B ) be co algebras. A C linear mapping

φ : A → B is said to be a cohomomorphism if

∆B ◦ φ = (φ ⊗ φ)∆A

A = B ◦ φ

Let σ denote the flip map : A ⊗ A → A ⊗ A given by σ(a ⊗ b) = b ⊗ a.

Definition 1.2.3. A coalgebra is said to be cocommutative if σ ◦ ∆ = ∆.

Definition 1.2.4. A linear subspace B of A is a subcoalgebra of A if ∆(B) ⊆ B ⊗ B.

Definition 1.2.5. A C linear subspace I of A is called a coideal if ∆(I) ⊆ A⊗I +I ⊗A

and (I) = {0}.

If I is a coideal of A, the quotient vector space A/I becomes a coalgebra with

comultiplication and counit induced from A.

Sweedler notation
We introduce the so called Sweedler notation for comultiplication. If a is an element
P
of a coalgebra A, the element ∆(a) in A ⊗ A is a finite sum ∆(a) = i a1i ⊗ a2i
where a1i , a2i belongs to A. Moreover, the representation of ∆(a) is not unique. For
notational simplicity we shall suppress the index i and write the above sum symbolically
17 Quantum Groups

as ∆(a) = a(1) ⊗ a(2) . Here the subscripts (1) and (2) refer to the corresponding tensor
factors.

Definition 1.2.6. A vector space A is called a bialgebra if it is an algebra and a coalge-

bra along with the condition that ∆ : A → A ⊗ A and  : A → C are algebra homomor-
phisms ( equivalently, m : A ⊗ A → A and η : C → A are coalgebra co-homomorphisms
).

Definition 1.2.7. A bialgebra A is called a Hopf algebra if there exists a linear map
κ : A → A called the antipode or the coinverse of A, such that m ◦ (κ ⊗ id)∆ = η ◦  =
m ◦ (id ⊗ κ) ◦ ∆.

Dual Hopf algebra

Let us consider a finite dimensional Hopf algebra A. Then the dual vector space A0 is
an algebra with respect to the multiplication f g(a) = (f ⊗g)∆(a). Moreover, for f in A0 ,
one defines the functional ∆(f ) ∈ (A ⊗ A)0 by ∆(f )(a ⊗ b) = f (ab), a, b in A. Since A is
finite dimensional, (A ⊗ A)0 = A0 ⊗ A0 and so ∆(f ) belongs to A0 ⊗ A0 . Then the algebra
A0 equipped with the comultiplication ∆, antipode κ defined by (κf )(a) = f (κ(a)),
counit A0 defined by A0 (f ) = f (1) and 1A0 (a) = (a) gives a Hopf algebra. This is
called the dual Hopf algebra of A.

Definition 1.2.8. A Hopf ∗-algebra is a Hopf algebra ( A, ∆, κ,  ) endowed with an

involution ∗ which maps a to an element denoted by a∗ satisfying the following proper-
ties:
1. For all a, b in A, α, β in C, (αa + βb)∗ = αa∗ + βb∗ , (a∗ )∗ = a, (a.b)∗ = b∗ a∗ .
2. ∆ : A → A ⊗ A is a ∗-homomorphism which means that ∆(a∗ ) = ∆(a)∗ where
the involution on A ⊗ A is defined by (a ⊗ b)∗ = a∗ ⊗ b∗ .

Proposition 1.2.9. In any Hopf ∗-algebra ( A, ∆, κ,  ), we have

1. (a∗ ) = (a) for all a in A.
2. κ(κ(a∗ )∗ ) = a for all a in A.

We recall that the dual algebra A0 of a Hopf ∗-algebra A is a ∗-algebra with involution
defined by
f ∗ (a) = f (κ(a)∗ ), for f in A0 .

Dual Pairing
A left action of a Hopf ∗-algebra ( U, ∆U , κU , U ) on another Hopf ∗-algebra (
A, ∆A , κA , A ) is a bilinear form . : U × A → C if the following conditions hold:

(1)f . (a1 a2 ) = ∆U (f ) . (a1 ⊗ a2 ), (f1 f2 ) . a = (f1 ⊗ f2 ) . ∆A (a);

Chapter 1: Preliminaries 18

(2)f . 1A = U (f ), 1U . a = A (a);

(3)f ∗ . a = f . κA (a)∗ (equivalently f . a∗ = κU (f )∗ . a)

for all f, f1 , f2 in U and a, a1 , a2 in A.

Similarly, a right action of a Hopf ∗-algebra ( U, ∆U , κU , U ) on another Hopf ∗-
algebra ( A, ∆A , κA , A ) is a bilinear form / : A × U → C if the following conditions
hold: a1 a2 /f = (a1 /f(1) )(a2 /f(2) ), a/(f1 f2 ) = ∆A (a)/(f1 ⊗f2 ), 1A /f = U (f ), a/1U =
A (a), a / f ∗ = κA (a)∗ / f ( equivalently a∗ / f = a / κU (f )∗ ) for all f, f1 , f2 in U and
a, a1 , a2 in A.
U = A0 gives a particular case of this duality pairing.

1.2.2 Compact Quantum Groups: basic definitions and examples

Definition 1.2.10. A compact quantum group (to be abbreviated as CQG from
now on) is given by a pair (S, ∆), where S is a unital separable C ∗ algebra equipped
with a unital C ∗ -homomorphism ∆ : S → S ⊗ S (where ⊗ denotes the injective tensor
product) satisfying
(ai) (∆ ⊗ id) ◦ ∆ = (id ⊗ ∆) ◦ ∆ (co-associativity), and
(aii) each of the linear spans of ∆(S)(S ⊗1) and of ∆(S)(1⊗S) are norm-dense in S ⊗S.

It is well-known (see [67], [66]) that there is a canonical dense ∗-subalgebra S0 of S,

consisting of the matrix elements of the finite dimensional unitary (co)-representations
(to be defined shortly) of S, and maps  : S0 → C (co-unit) and κ : S0 → S0 (antipode)
defined on S0 which make S0 a Hopf ∗-algebra.

The following theorem is the analogue of Gelfand Naimark duality for commutative
CQG s.

Proposition 1.2.11. Let G be a compact group. Let C(G) be the algebra of continuous
functions on G. If we define ∆ by ∆(f )(g, h) = f (g.h) for f in C(G), g, h in G, then
this defines a CQG structure on C(G).
Conversely, let (S, ∆) be a commutative CQG. Let H(S) denote the Gelfand spec-
trum of S and endow it with the product structure given by χχ0 = (χ ⊗ χ0 )∆ where χ, χ0
are in H(S). Then H(S) is a compact group.

Remark 1.2.12. In [57], A Van Daele removed Woronowicz’s separability assumption

(in [67] ) for the C ∗ algebra of the underlying compact quantum group. We remark
19 Quantum Groups

that although we assume that CQG s are separable, most of the results in this thesis go
through in the non separable case also.

Definition 1.2.13. Let (S, ∆S ) be a compact quantum group. A vector space M is said
to be an algebraic S co-module (or S co-module) if there exists a linear map α : M →
M ⊗alg S0 such that
1. α ⊗ id)α = (id ⊗ ∆S )α,
2. (id ⊗ )α(m) = m for all m in M.

In the notations as above, let us define αe : M ⊗ S → M ⊗ S by α e = (id ⊗ m)(α ⊗ id).

Then we claim that α e is invertible with the inverse given by T (m⊗q) = (id⊗κ)α(m)(1⊗
q), where m is in M, q is in S. As T is defined to be S0 linear, it is enough to check
that αeT (m ⊗ 1) = m ⊗ 1.

eT (m ⊗ 1)
α
e(m(1) ⊗ κ(m(2) )1)
= α
= m(1)(1) ⊗ m(1)(2) κ(m(2) )
= (id ⊗ m(id ⊗ κ)∆)α(m)
= (id ⊗ ().1)α(m)
= m ⊗ 1.

Similarly, T α
e = id. Thus,
e−1 .
T =α (1.2.1)

Definition 1.2.14. A morphism from a CQG (S1 , ∆1 ) to another CQG (S2 , ∆2 ) is a

unital C ∗ homomorphism π : S1 → S2 such that

(π ⊗ π)∆1 = ∆2 π.

It follows that in such a case, π preserves the Hopf ∗-algebra structures, that is, we
have
π((S1 )0 ) ⊆ (S2 )0 , πκ1 = κ2 π, 2 π = 1 ,

where κ1 , 1 denotes the antipode and counit of S1 respectively while κ2 , 2 denotes

those of S2 .

Definition 1.2.15. A Woronowicz C ∗ -subalgebra of a CQG (S1 , ∆) is a C ∗ subal-

gebra S2 of S1 such that (S2 , ∆|S2 ) is a CQG such that the inclusion map from S2 → S1
is a morphism of CQG s.

Definition 1.2.16. A Woronowicz C ∗ -ideal of a CQG (S, ∆) is a C ∗ ideal J of S

such that ∆(J) ⊆ Ker(π ⊗ π), where π is the quotient map from S to S/J.
Chapter 1: Preliminaries 20

It can be easily seen that a kernel of a CQG morphism is a Woronowicz C ∗ -ideal.

We recall the following isomorphism theorem.

Proposition 1.2.17. The quotient of a CQG (S, ∆) by a Woronowicz C ∗ -ideal I has

a unique CQG structure such that the quotient map π is a morphism of CQG s. More
e on S/I is given by ∆(s
precisely, the coproduct ∆ e + I) = (π ⊗ π)∆(s).

Definition 1.2.18. A CQG (S 0 , ∆0 ) is called a quantum subgroup of another CQG

(S, ∆) if there is a Woronowicz C ∗ -ideal J of S such that (S 0 , ∆0 ) ∼
= (S, ∆)/J.

Let us mention a convention which we are going to follow. We shall use most of the
terminologies of [59], for example Woronowicz C ∗ -subalgebra, Woronowicz C ∗ -ideal etc,
however with the exception that we shall call the Woronowicz C ∗ algebras just compact
quantum groups, and not use the term compact quantum groups for the dual objects
as done in [59].
Let (S, ∆) be a compact quantum group. Then there exists a state h on S, to be
called a Haar state on S such that (h ⊗ id)∆(s) = (id ⊗ h)∆(s) = h(s).1. We recall
that unlike the group case, h may not be faithful. But on the dense Hopf ∗-algebra S0
mentioned above, it is faithful. We have the following result.

Proposition 1.2.19. Let i : S1 → S2 be an injective morphism of CQG s. Then the

Haar state on S1 is the restriction of that of S2 on S1 .

Remark 1.2.20. In general, the Haar state might not be tracial. In fact, there exists a
multiplicative linear functional denoted by f1 in [66] such that h(ab) = h(b(f1 / a . f1 )).
Moreover, from Theorem 1.5 of [67], it follows that the Haar state of a CQG is tracial
if and only if κ2 = id.

Co-Representations of a compact quantum group

Definition 1.2.21. A co-representation of a compact quantum group (S, ∆) on a Hilbert

space H is a map U from H to the Hilbert S module H ⊗ S such that the element U e
belonging to M(K(H) ⊗ S) given by U e (ξ ⊗ b) = U (ξ)(1 ⊗ b) ( ξ in H, b in S) satisfies

(id ⊗ ∆)U
e =U
e(12) U
e(13) ,

where for an operator X in B(H1 ⊗H2 ) we have denoted by X(12) and X(13) the operators
X ⊗ IH2 in B(H1 ⊗ H2 ⊗ H2 ), and Σ23 X(12) Σ23 respectively and Σ23 is the unitary on
H1 ⊗ H2 ⊗ H2 which flips the two copies of H2 .
e is an unitary element of M(K(H) ⊗ S), then U is called a unitary co-
If U
representation.
21 Quantum Groups

From now on, we will drop the term co in the word co-representation unless there
is any confusion.
Remark 1.2.22. Let π be a CQG morphism from a CQG (S1 , ∆1 ) to another CQG
(S2 , ∆2 ). Then for every unitary representation U of S1 , (id ⊗ π)U is a unitary repre-
sentation of S2 .
Following the definitions given in the last part of subsection 1.1.4 and a unitary
representation U of a CQG on a Hilbert space H, and a not necessarily bounded,
densely defined (in the weak operator topology) linear functional τ on B(H), we will
use the notation αU and the terms “αU preserves τ ” and “ U equivariant ” throughout
this thesis.
A CQG (S, ∆) has a distinguished representation which corresponds to the right
regular representation in the group case. Let H be the GNS space of S associated with
the Haar state h, ξ0 be the associated cyclic vector and K be a Hilbert space on which
S acts faithfully and non-degenerately. There is a unitary operator u on H ⊗ K defined
by u(aξ0 ⊗ η) = ∆(a)(ξ0 ⊗ η) when a is in S, η is in K. Then u can be shown to be an
element of multiplier of K(H) ⊗ S and called the right regular representation of S.
Let v be a representation of a CQG (S, ∆) on a Hilbert space H. A closed subspace
H1 of H is said to be invariant if (e ⊗ 1)v(e ⊗ 1) = v(e ⊗ 1), where e is the orthogonal
projection onto this subspace. The representation v is called irreducible if the only
invariant subspaces are {0} and H. It is clear that one can make sense of direct sum of
(co)-representations in this case also. Moreover, for two representations v and w of a
CQG (S, ∆) on Hilbert spaces H1 and H2 , the tensor product of v and w is given by the
element v(13) w(23) . The intertwiner between v and w is an element x in B(H1 , H2 ) such
that (x⊗1)v = w(x⊗1). The set of intertwiners between v and w is denoted by Mor(v, w).
Two representations are said to be equivalent if there is an invertible intertwiner. They
are unitarily equivalent if the intertwiner can be chosen to be unitary.
Just like the case of compact groups, CQG s have an analogous Peter Weyl theory
which corresponds to the usual Peter Weyl theory in the group case. We will give a
sketch of it by mentioning the main results and refer to [41], [66] and [67] for the details.
Let v be a unitary representation of (S, ∆) on H. If H1 is an invariant subspace,
then the orthogonal complement of H1 is also invariant. Any non degenerate finite
dimensional representation is equivalent with a unitary representation.
Every irreducible unitary representation of a CQG is contained in the regular rep-
resentation. Let v be a representation on a finite dimensional Hilbert space H. If we
P
denote the matrix units in B(H) by (epq ), we can write v = epq ⊗ vpq . vpq are called
∗ .
P
the matrix elements of the finite dimensional representation v. Define v = epq ⊗ vpq
Then v is a representation and is called the adjoint of v. It can be shown that if v is
a finite dimensional irreducible representation, then v is also irreducible. Moreover, for
Chapter 1: Preliminaries 22

an irreducible unitary representation, its adjoint is equivalent with a unitary represen-

tation.
The subspace spanned by the matrix elements of finite dimensional unitary repre-
sentations is denoted by S0 . Firstly, S0 is a subalgebra as the product of two matrix
elements of finite dimensional unitary representations is a matrix element of the tensor
product of these representations. Moreover, as the adjoint of a finite dimensional uni-
tary representation is equivalent with a unitary representation, S0 is ∗ invariant. We
note that 1 is in S0 as 1 is a representation. Now, we will recall some basic facts about
the subalgebra S0 . We will denote the Haar state of S by h.

Proposition 1.2.23. (1) S0 is a dense ∗-subalgebra of S.

(2)Let {uα : α ∈ I} be a complete set of mutually inequivalent, irreducible unitary

representations. We will denote the representation space and dimension of uα by Hα
and n(α) respectively. Then the Schur’s orthogonality relation takes the following form:
For any α in I, there is a positive invertible operator F α acting on Hα such that for
any α, β in I and 1 ≤ j, q ≤ n(α), 1 ≤ i, p ≤ n(β)
∗
h((uβip ) uαjq ) = δαβ δpq Fijα .

(3) {uαpq : α ∈ I, 1 ≤ p, q ≤ n(α)} form a basis for S0 .

(4) Moreover, ∆ maps S0 into S0 ⊗ S0 . In fact, ∆ is given by ∆(uαpq ) = nk=1

P α α
upk ⊗
α
ukq . A counit and an antipode are defined on S0 respectively by the formulae,

(uαpq ) = δpq , κ(uαpq ) = (uαqp )∗ .

It follows that S0 becomes a Hopf ∗-algebra.

A compact matrix quantum group is a CQG such that there exists a distin-
guished unitary irreducible representation called the fundamental representation such
that the ∗-algebra spanned by its matrix elements is a dense Hopf ∗-subalgebra of the
CQG.
We now discuss the free product of CQG s which were developed in [59]. Let
(S1 , ∆1 ) and (S2 , ∆2 ) be two CQG s. Let i1 and i2 denote the canonical injections of S1
and S2 into the C ∗ algebra S1 ∗ S2 . Put ρ1 = (i1 ⊗ i1 )∆1 and ρ2 = (i2 ⊗ i2 )∆2 . By the
universal property of S1 ∗ S2 , there exists a map ∆ : S1 ∗ S2 → (S1 ∗ S2 ) ⊗ (S1 ∗ S2 ) such
that ∆i1 = ρ1 and ∆i2 = ρ2 . It can be shown that ∆ indeed has the required properties
so that (S, ∆) is a CQG.
Let {Sn }n∈IN be an inductive sequence of CQG s, where the connecting morphisms
πmn from Sn to Sm (n < m) are injective morphisms of CQG s. Then from Proposition
23 Quantum Groups

3.1 of [59], we have that the inductive limit S0 of Sn s has a unique CQG structure with
the following property: for any CQG S 0 and any family of CQG morphisms φn : Sn → S 0
such that φm πmn = φn , the uniquely defined morphism limn φn in the category of unital
C ∗ algebras is a morphism in the category of CQG s.
Combining the above two results, it follows that the free product C ∗ algebra of an
arbitrary sequence of CQG s has a natural CQG structure.
Moreover, the following result was derived in [59].

Proposition 1.2.24. Let Γ1 , Γ2 be a discrete abelian groups. Then the natural isomor-
phisms C ∗ (Γ1 ) ∼ c1 ) and C ∗ (Γ1 ) ∗ C ∗ (Γ2 ) ∼
= C(Γ = C ∗ (Γ1 ∗ Γ2 ) are isomorphism of CQG
s.

Let i1 and i2 be the inclusion of CQG s S1 and S2 into S1 ∗ S2 . If U1 and U2

are unitary representations of CQG s S1 and S2 on Hilbert spaces H1 and H2 respec-
tively, then the free product representation of U1 and U2 is a representation of
the CQG S1 ∗ S2 on the Hilbert! space H1 ⊕ H2 given by the S1 ∗ S2 valued matrix
(id ⊗ i1 )U1 0
.
0 (id ⊗ i2 )U2
Similarly, the free product representation of an arbitrary sequence of CQG repre-
sentations are defined.

The inductive limit of an arbitrary sequence of CQG s has the structure of a CQG.
The following lemma is probably known, but we include the proof ( taken from [9] ) for
the sake of completeness.

Lemma 1.2.25. Suppose that (Sn )n∈IN is a sequence of CQG s and for each n, m in
IN , n ≤ m there is a CQG morphism πn,m : Sn → Sm with the compatibility property

πm,k ◦ πn,m = πn,k , n ≤ m ≤ k.

Then the inductive limit of C ∗ -algebras (Sn )n∈IN has a canonical structure of a CQG.
It will be denoted S∞ or limn∈IN Sn . It has the following universality property:
for any CQG (S, ∆) such that there are CQG morphisms πn : Sn → S satisfying for all
m, n ∈ IN , m ≥ n the equality πm ◦ πn,m = πn , there exists a unique CQG morphism
π∞ : S∞ → S such that πn = π∞ ◦ πn,∞ for all n ∈ IN , where we have denoted by πn,∞
the canonical unital C ∗ -homomorphism from Sn into S∞ .

Proof:
Let us denote the coproduct on Sn by ∆n . We consider the unital C ∗ -homomorphism
ρn : Sn → S∞ ⊗ S∞ given by ρn = (πn,∞ ⊗ πn,∞ ) ◦ ∆n , and observe that these maps do
Chapter 1: Preliminaries 24

satisfy the compatibility property:

ρm ◦ πn,m = ρn ∀n ≤ m.

Thus, by the general properties of the C ∗ -algebraic inductive limit, we have a unique
unital C ∗ -homomorphism ∆∞ : S∞ → S∞ ⊗ S∞ satisfying ∆∞ ◦ πn,∞ = ρn for all n.
We claim that (S∞ , ∆∞ ) is a CQG.
We first check that ∆∞ is coassociative. It is enough to verify the coassociativity
on the dense set ∪n πn,∞ (Sn ). Indeed, for s = πn,∞ (a) (a ∈ Sn ), by using ∆∞ ◦ πn,∞ =
(πn,∞ ⊗ πn,∞ ) ◦ ∆n , we have the following:

(∆∞ ⊗ id)∆∞ (πn,∞ (a))

= (∆∞ ⊗ id)(πn,∞ ⊗ πn,∞ )(∆n (a))
= (πn,∞ ⊗ πn,∞ ⊗ πn,∞ )(∆n ⊗ id)(∆n (a))
= (πn,∞ ⊗ πn,∞ ⊗ πn,∞ )(id ⊗ ∆n )(∆n (a))
= (πn,∞ ⊗ (πn,∞ ⊗ πn,∞ ) ◦ ∆n )(∆n (a))
= (πn,∞ ⊗ ∆∞ ◦ πn,∞ )(∆n (a))
= (id ⊗ ∆∞ )((πn,∞ ⊗ πn,∞ )(∆n (a)))
= (id ⊗ ∆∞ )(∆∞ (πn,∞ (a)))

which proves the coassociativity.

Finally, we need to verify the quantum cancellation properties. Note that to show
that ∆∞ (S∞ )(1⊗S∞ ) is dense in S∞ ⊗S∞ it is enough to show that the above assertion
S
is true with S∞ replaced by a dense subalgebra n πn,∞ (Sn ).
Using the density of ∆n (Sn )(1 ⊗ Sn ) in Sn ⊗ Sn and the contractivity of the map
πn,∞ we note that (πn,∞ ⊗ πn,∞ )(∆n (Sn )(1 ⊗ Sn )) is dense in (πn,∞ ⊗ πn,∞ )(Sn ⊗ Sn ).
This implies that (πn,∞ ⊗ πn,∞ )(∆n (Sn ))(1 ⊗ πn,∞ (Sn )) is dense in πn,∞ (Sn ) ⊗ πn,∞ (Sn )
and hence ∆∞ (πn,∞ (Sn ))(1 ⊗ πn,∞ (Sn )) is dense in πn,∞ (Sn ) ⊗ πn,∞ (Sn ). The proof of
the claim now follows by noting that πn,∞ (Sn ) = πm,∞ πn,m (Sn ) ⊆ πm,∞ (Sm ) for any
m ≥ n, along with the above observations. The right quantum cancellation property
can be shown in the same way.
The proof of the universality property is routine and hence omitted.
2

We note that the proof remains valid for any other indexing set for the net, not
necessarily IN .
25 Quantum Groups

1.2.3 The CQG Uµ (2)

We now introduce the compact quantum group Uµ (2). We refer to [37] for more details.
As a unital C ∗ algebra, Uµ (2) is generated by 4 elements u11 , u12 , u21 , u22 satisfying:

u11 u12 = µu12 u11 (1.2.2)

u11 u21 = µu21 u11 (1.2.3)

u12 u22 = µu22 u12 (1.2.4)

u21 u22 = µu22 u21 (1.2.5)

u12 u21 = u21 u12 (1.2.6)

u11 u22 − u22 u11 = (µ − µ−1 )u12 u21 (1.2.7)

!
u11 u12
and the condition that the matrix u = is a unitary. Thus, the above
u21 u22
matrix u is the fundamental unitary for Uµ (2).
The CQG structure is given by
X
∆(uij ) = uik ⊗ ukj , κ(uij ) = uji ∗ , (uij ) = δij . (1.2.8)
k=1,2

The quantum determinant Dµ is defined by

Dµ = u11 u22 − µu12 u21 = u22 u11 − µ−1 u12 u21 . (1.2.9)

Then, Dµ ∗ Dµ = Dµ Dµ ∗ = 1. Moreover, Dµ belongs to the centre of Uµ (2).

We mention the following result for future use.

Proposition 1.2.26.

κ(u11 ) = u22 Dµ−1 , κ(u12 ) = −µ−1 u12 Dµ−1 , κ(u21 ) = −µu21 Dµ−1 , κ(u22 ) = u11 Dµ−1 .

Proof : By [37] ( Proposition 10, Page 314 ), we have that κ(uij ) = uf −1

ij Dµ where
uf
ij is the (i, j)th entry of a matrix ue satisfying
! u
eu = ue u = Dµ I2 . One can easily
u22 −1
−µ u12
check that the matrix satisfies this equation and this proves the
−µu21 u11
Proposition. 2
Chapter 1: Preliminaries 26

1.2.4 The CQG SUµ (2)

Let µ belongs to [−1, 1]. The C ∗ algebra SUµ (2) is defined as the universal unital C ∗
algebra generated by α, γ satisfying:

α∗ α + γ ∗ γ = 1, (1.2.10)

αα∗ + µ2 γγ ∗ = 1, (1.2.11)

γγ ∗ = γ ∗ γ, (1.2.12)

µγα = αγ, (1.2.13)

µγ ∗ α = αγ ∗ . (1.2.14)
!
α −µγ ∗
The fundamental representation of SUµ (2) is given by : .
γ α∗
There is a coproduct ∆ of SUµ (2) given by :

∆(α) = α ⊗ α − µγ ∗ ⊗ γ, ∆(γ) = γ ⊗ α + α∗ ⊗ γ

which makes it into a CQG. Let h denote the Haar state and H = L2 (SUµ (2)) be the
corresponding G.N.S space.
Haar state on SUµ (2)
We restate the content of Theorem 14, Chapter 4 ( page 113 ) of [37] in a convenient
form below. For all m ≥ 1, n, l, k ≥ 0, k 0 6= k 00 ,

1 − µ2 0 00
h((γ ∗ γ)k ) = 2k+2
, h(αm γ ∗n γ l ) = 0, h(α∗m γ ∗n γ l ) = 0, h(γ ∗k γ ∗k ) = 0. (1.2.15)
1−µ

( Co )- representations of SUµ (2)

For each n in {0, 1/2, 1, .....}, there is a unique irreducible representation T n of
dimension 2n + 1. Denote by tnij the ij th entry of T n . They form an orthogonal basis
of H. Denote by enij the normalized tnij s so that {enij : n = 0, 1/2, 1, ......, i, j = −n, −n +
1, .....n} is an orthonormal basis.
We recall from [37] that

1/2 1/2 1/2 1/2

t−1/2,−1/2 = α, t−1/2,1/2 = −µγ ∗ , t1/2,−1/2 = γ, t1/2,1/2 = α∗ . (1.2.16)

Moreover, if we define

fn,i = a(n, i)αn−i γ ∗n+i (1.2.17)

27 Quantum Groups

where a(n, i) s are some constants as in [37], then {fn,i : n = 0, 12 , 1, 23 , ....., − n ≤

i ≤ n} is an orthonormal basis of SUµ (2) and ∆(fn,i ) = nk=−n fn,k ⊗ tnk,i .
P

The following recursive relations will be useful to us.

Proposition 1.2.27.

l+1/2
ti,l+1/2
= c11 (i, l)γ ∗ tli+1/2,l + c12 (i, l)α∗ tli−1/2,l − l + 1/2 ≤ i ≤ l − 1/2,
= c21 (i, l)γ ∗ tli+1/2,l i = −l − 1/2,
= c31 (i, l)α∗ tli−1/2,l i = l + 1/2,
(1.2.18)

and for j ≤ l,

l+1/2
ti,j
= c(l, i, j)αtli+1/2,j+1/2 + c0 (l, i, j)γtli−1/2,j+1/2 − l + 1/2 ≤ i ≤ l − 1/2,
1 1
= d(l, j)αtl−l,j+1/2 + d0 (l, j)γ ∗ tl−l,j− 1 i = −l − 1/2, − l + ≤ j ≤ l − ,
2 2 2
1
= d00 (l, j)αtli+1/2,j+1/2 i = −l − 1/2, j = −l − ,
2
= e(l, j)γtli−1/2,j+1/2 + e0 (l, j)α∗ tli− 1 ,j− 1 i = l + 1/2,
2 2
(1.2.19)

where Cpq (il), c(l, i, j), d(l, j), d0l,j , d00 (l, j), e(l, j), e0 (l, j) are all complex numbers.

Proof : It can be easily seen that

fl+ 1 ,i = c(l, i)αfl,i+ 1 (1.2.20)

2 2

for some constants c(l, i).

Moreover, from ( 1.2.17 ) we have γ ∗ fl,i = a0 (l, i)αl−i γ ∗ l+i+1 for some constant
a0 (l, i). This means that

a0 (l, i)
γ ∗ fl,i = f 1 1. (1.2.21)
a(l + 12 , i + 12 ) l+ 2 ,i+ 2

We have fl+ 1 ,l+ 1 = a(l + 21 , l + 12 )γ ∗ 2l+1 and fl,l = a(l, l)γ ∗ 2l which implies that
2 2

a(l + 21 , l + 21 ) ∗
fl+ 1 ,l+ 1 = γ fl,l . (1.2.22)
2 2 a(l, l)
Now, we proceed to prove ( 1.2.18 ). Applying coproduct on ( 1.2.22 ) and using (
Chapter 1: Preliminaries 28

1.2.17 ) and ( 1.2.21 ), we have

1
l+ 2
X l+ 1
fl+ 1 ,k ⊗ tk,l+
2
1
2 2
k=−(l+ 21 )
l
a(l + 12 , l + 12 ) ∗ X
= (γ ⊗ α∗ + α ⊗ γ ∗ )( fl,k ⊗ tlk,l )
a(l, l)
k=−l
" l l
#
a(l + 12 , l + 12 ) X ∗ X
= γ fl,k ⊗ α∗ tlk,l + αfl,k ⊗ γ ∗ tlk,l
a(l, l)
k=−l k=−l
" l 0 ∗ l l fl+ 1 ,k− 1 ⊗ γ ∗ tlk,l
#
a(l + 12 , l + 12 ) X a (l, k)fl+ 12 ,k+ 12 ⊗ α tk,l X
2 2
= 1 1 +
a(l, l)
k=−l
a(l + 2 , k + 2 ) k=−l
c(l, k − 12 )
 
1 1 l+ 12 a 0 (l, k − 1 )f
1 ⊗ α ∗ tl l− 12 f 1 ⊗ γ ∗ tl
a(l + 2 , l + 2 )  X 2 l+ 2 ,k 1
k− 2 ,l X l+ 2 ,k 1
k+ 2 ,l 
= 1 + .
a(l, l) c(l, k)

1
a(l + 2 , k) 1
k=−l+ 2 k=−l− 2

1 l+ 1
Let −l + 2 ≤ k ≤ l − 12 . Then comparing coefficient of fl+ 1 ,k we have tk,l+
2
1 =
2 2
a(l+ 12 ,l+ 12 ) a0 (l,k− 21 ) ∗ l 1 ∗ l
a(l,l) [ a(l+ 1 ,k) α tk− 1 ,l + c(l,k) γ tk+ 1 ,l ] which proves the first equation of ( 1.2.18 ).
2 2 2
l+ 1 al+ 1 ,l+ 1
Applying the same procedure for k = −l − 12 , we have tk,l+
2
1 =
2
al,l
2 1
[ c(l,k) γ ∗ tlk+ 1 ,l ]
2 2
which proves the second equation of ( 1.2.18 ).
l+ 1 a(l+ 12 ,l+ 12 ) a0 (l,k− 12 ) ∗ l
Similarly, for k = l + 12 , we have tk,l+
2
1 = a(l,l) [ a(l+ 1 ,k) α tk− 1 ,l ] which proves
2 2 2
the third equation of ( 1.2.18 ). This completes the proof of ( 1.2.18 ).
Next, to prove ( 1.2.19 ), we apply coproduct on ( 1.2.20 ) and use ( 1.2.17 ) and (
1.2.21 ) to have

1
l+ 2
X l+ 1
fl+ 1 ,k ⊗ tk,i 2
2
k=−(l+ 12 )
l
X
= c(l, i)(α ⊗ α − µγ ∗ ⊗ γ)( fl,k ⊗ tlk,i+ 1 )
2
k=−l
l
X n
X
= c(l, i)( αfl,k ⊗ αtlk,i+ 1 − γ ∗ fl,k ⊗ µγtlk,i+ 1 )
2 2
k=−l k=−n
l l
X 1 l
X a0 (l, k)
= c(l, i) f 1 1 ⊗ αt 1 − c(l, i) f 1 1⊗
k=−l
c(l, k − 12 ) l+ 2 ,k− 2 k,i+ 2
k=−l
a(l + 12 , k + 21 ) l+ 2 ,k+ 2
µγtlk,i+ 1
2
29 Quantum Groups

1 1
l− 2 l+ 2
X 1 l
X a(l, k − 21 )
= c(l, i) f 1 ⊗ αtk+ 1 ,i+ 1 − c(l, i) f 1 ⊗
c(l, k) l+ 2 ,k 2 2 a(l + 12 , k) l+ 2 ,k
k=−l− 12 k=−l+ 12

µγtlk− 1 ,i+ 1 .
2 2
1 l+ 1
For −l + 2 ≤ k ≤ l − 12 , by comparing coefficient of fl+ 1 ,k we have tk,i 2 =
2
c(l,i) l c(l,i)a(l,k− 21 )
c(l,k) αtk+ 1 ,i+ 1 − a(l+ 12 ,k)
µγtlk− 1 ,i+ 1 which proves the first equation of ( 1.2.19 ).
2 2 2 2
l+ 1 c(l,i)
Comparing coefficient of fl+ 1 ,−l− 1 , we have t−l−2 1 ,i = c(l,−l− 12 )
αtl−l,i+ 1 from which
2 2 2 2
we get the second and the third equation of ( 1.2.19 ).
l+ 1
c(l,i)a(l,l)
Comparing coefficient of fl+ 1 ,l+ 1 , we have tl+ 21 ,i = − a(l+ 1
,l+ 1 )
µγtll,i+ 1 from which
2 2 2 2 2 2
we get the last equation of ( 1.2.19 ). 2

We recall the following multiplication rule from Page 74, [37] which we are going to
need :

1/2 0 0
X
tli,j ti0 ,j 0 = ck (l, i, j, i , j )tki+i0 ,j+j 0 (1.2.23)
k=|l−1/2|,.....l+1/2

(ck (l, i, j, i0 , j 0 ) are scalars).

1.2.5 The Hopf ∗-algebras O(SUµ (2)) and Uµ (su(2))

We define the Hopf ∗-algebra O(SUµ (2)) following the notations of [37].
O(SLµ (2)) is the complex associative algebra with generators a, b, c, d such that

ab = µba, ac = µca, bd = µdb, cd = µdc, bc = cb, ad − µbc = da − µ−1 bc = 1. (1.2.24)

The coproduct is given by

∆(a) = a ⊗ a + b ⊗ c, ∆(b) = a ⊗ b + b ⊗ d,

∆(c) = c ⊗ a + d ⊗ c, ∆(d) = c ⊗ b + d ⊗ d.

The antipode is
κ(a) = d, κ(b) = −b, κ(c) = −c, κ(d) = a

Finally, the counit is

(a) = (d) = 1, (b) = (c) = 0.

For all µ in R, there is an involution of the algebra O(SLµ (2)) determined by

a∗ = d, b∗ = −µc. (1.2.25)
Chapter 1: Preliminaries 30

The corresponding Hopf ∗-algebra is denoted by O(SUµ (2)).

Proposition 1.2.28. O(SUµ (2)) can be identified with (SUµ (2))0 , i.e the Hopf ∗-algebra
generated by the matrix elements of irreducible unitary representations of SUµ (2), via
the isomorphism given on the generators by

α 7→ a, γ 7→ c, α∗ 7→ d, γ ∗ 7→ −µ−1 b. (1.2.26)

Proof : (SUµ (2))0 is generated by the matrix elements of the fundamental unitary
of SUµ (2), that is, the ∗-algebra generated by α and γ. On the other hand, inserting (
1.2.25 ) in ( 1.2.24 ), we have that O(SUµ (2)) is generated by 4 elements a, b, c, d such
that ac = µca, ac∗ = µc∗ a, cc∗ = c∗ c, a∗ a + c∗ c = 1, aa∗ + µ2 c∗ c = 1. Comparing with
the defining equations of SUµ (2), that is, ( 1.2.10 ) - ( 1.2.14 ), it is clear that the above
correspondence gives the required isomorphism. 2

Next, we recall from [51] the Hopf ∗ algebra Uµ (su(2)) which is the dual Hopf ∗-
algebra of O(SUµ (2)). It is generated by elements F, E, K, K −1 with defining relations:

KK −1 = K −1 K = 1, KE = µEK, F K = µKF, EF − F E = (µ − µ−1 )−1 (K 2 − K −2 )

with involution E ∗ = F, K ∗ = K and comultiplication :

∆(E) = E ⊗ K + K −1 ⊗ E, ∆(F ) = F ⊗ K + K −1 ⊗ F, ∆(K) = K ⊗ K.

The counit is given by (E) = (F ) = (K − 1) = 0 and antipode κ(K) = K −1 , κ(E) =
−µE, κ(F ) = −µ−1 F.
There is a dual pairing h., .i of Uµ (su(2)) and O(SUµ (2)) given on the generators by
:
1
K , α = K , α = µ± 2 , hE, γi = hF, −µγ ∗ i = 1

±1 ∗ 
∓1 

and zero otherwise.

The left action . and right action / of Uµ (su(2)) on SUµ (2) are given by:



f . x = f, x(2) x(1) , x / f = f, x(1) x(2) , x ∈ O(SUµ (2)), f ∈ Uµ (su(2)) where
we use the Sweedler notation ∆(x) = x(1) ⊗ x(2) .
The actions satisfy :
(f . x)∗ = κ(f )∗ . x∗ , (x / f )∗ = x∗ / κ(f )∗ , f . xy = (f(1) . x)(f(2) . y), xy / f =
(x / f(1) )(y / f(2) ).
The action on generators is given by :
31 Quantum Groups

 
∗ E . γ = α∗ , E . γ ∗ = 0, E . α∗ = 0,
 E . α = −µγ
 

F . (−µγ ∗ ) = α F . α∗ = γ, F . α = 0, F . γ = 0, (1.2.27)
− 21 1
− 1 1
 
K . α = µ α, K . (γ ∗ ) = µ 2 γ ∗ , K . γ = µ 2 γ, K . α∗ = µ 2 α∗ .
 

 

 γ / E = α, α∗ / E = −µγ ∗ , α / E = 0, γ∗ / E = 0 

α / F = γ, ∗
−µγ / F = α , ∗ γ / F = 0, ∗
α / F = 0, (1.2.28)
− 21 − 12 ∗ 1 1
 
α / K = µ α, γ / K = µ γ , γ / K = µ 2 γ, α / K = µ 2 α∗ .
∗ ∗
 

1.2.6 The Wang algebras

Let us now recall the universal quantum groups as in [61], [59] and references therein.
For an n × n positive invertible matrix Q = (Qij ). let Au,n (Q) be the compact quantum
group defined and studied in [60], [61], which is the universal C ∗ -algebra generated by
{uQ Q
kj , k, j = 1, ..., n} such that u := ((ukj )) satisfies

uu∗ = In = u∗ u, u0 QuQ−1 = In = QuQ−1 u0 . (1.2.29)

˜ is given by,
Here u0 = ((uji )) and u = ((u∗ij )). The coproduct, say ∆,

n
uQ Q
X
˜ ij ) =
∆(u ik ⊗ ukj .
k=1

It may be noted that Au,n (Q) is the universal object in the category of compact quan-
tum groups which admit an action on the finite dimensional C ∗ algebra Mn (C) which
preserves the functional Mn 3 x 7→ Tr(QT x),( see [63] ) where the notion of a CQG and
that of preservation of a functional by an action are as in subsection 1.2.7. We refer
the reader to [61] for a detailed discussion on the structure and classification of such
quantum groups.

Remark 1.2.29. It was proved in [59] that in the case where Q = I, κ(uIij ) = uI∗
ji and
2
hence κ = id holds for Au,n (I).

1.2.7 Action of a compact quantum group on a C ∗ algebra

We say that the compact quantum group (S, ∆) (co)-acts on a unital C ∗ algebra B, if
there is a unital C ∗ -homomorphism (called an action) α : B → B ⊗ S satisfying the
following :
(bi) (α ⊗ id) ◦ α = (id ⊗ ∆) ◦ α, and
Chapter 1: Preliminaries 32

(bii) the linear span of α(B)(1 ⊗ S) is norm-dense in B ⊗ S.

It is known ( see, for example, [60] , [44] ) that (bii) is equivalent to the existence
of a norm-dense, unital ∗-subalgebra B0 of B such that α(B0 ) ⊆ B0 ⊗alg S0 and on B0 ,
(id ⊗ ) ◦ α = id.
We shall sometimes say that α is a ‘topological’ or C ∗ action to distinguish it from
a normal action of von Neumann algebraic quantum group.

Definition 1.2.30. Let (S, α) has a C ∗ action α on the C ∗ algebra B. We say that the
action α is faithful if there is no proper Woronowicz C ∗ -subalgebra S1 of S such that
α is a C ∗ action of S1 on B.

Definition 1.2.31. Let (S, α) has a C ∗ action α on the C ∗ algebra B. A continuous

linear functional φ on B is said to be invariant under α if

(φ ⊗ id)α(b) = φ(b).1S .

Now, we recall the work of Shuzhou Wang done in [60]. One can also see [4], [5].
The quantum permutation group QUn is defined to be the C ∗ algebra generated
by aij ( i, j = 1, 2, ...n ) satisfying the following relations:

a2ij = aij = a∗ij , i, j = 1, 2, ...n,

n
X
aij = 1, i = 1, 2, ...n,
j=1
n
X
aij = 1, i = 1, 2, ...n.
i=1

The name comes from the fact that the universal commutative C ∗ algebra generated
by the above set of relations is isomorphic to C(Sn ) where Sn denotes the permutation
group on n symbols.
Let us consider the category with objects as compact groups acting on on a n-point
set Xn = {x1 , x2 , ..., xn }. If two groups G1 and G2 have actions α1 and α2 respectively,
then a morphism from G1 to G2 is a group homomorphism φ such that α2 (φ × id) = α1 .
Then C(Sn ) is the universal object in this category. It is proved in [60] that the quantum
permutation group enjoys a similar property.
Pn
We have that C(Xn ) = C ∗ {ei : e2i = ei = e∗i , r=1 er = 1, i = 1, 2, ..., n}. Then
∗
QUn has a C action on C(Xn ) via the formula:
n
X
α(ej ) = ei ⊗ aij , j = 1, 2, ...n.
i=1
33 Quantum Groups

Proposition 1.2.32. Consider the category with objects as CQG s having a C ∗ action
on C(Xn ) and morphisms as CQG morphisms intertwining the actions as above. Then
QUn is the universal object in this category.

Now we note down a simple fact for future use.

Lemma 1.2.33. Let α be an action of a CQG S on C(X) where X is a finite set. Then
α automatically preserves the functional τ corresponding to the counting measure:

(τ ⊗ id)(α(f )) = τ (f ).1S .

Proof:
Let X = {1, ..., n} for some n ∈ IN and denote by δi the characteristic function of
P
the point i. Let α(δi ) = j δj ⊗ qij where {qij : i, j = 1 . . . n} are the images of the
canonical generators of the quantum permutation group as above. Then τ -preservation
of α follows from the properties of the generators of the quantum permutation group,
2
P P
which in particular imply that j qij = 1 = i qij .

Wang also identified the universal object in the category of all CQG s having a C ∗
action α1 on Mn (C) ( with morphisms as before ) such that the functional n1 Tr is kept
invariant under α1 . However, no such universal object exists if the invariance of the
functional is not assumed. The precise statement is contained in the following theorem.
Pn
Before that, we recall that Mn (C) = C ∗ {eij : eij ekl = δjk eil , e∗ij = eji , r=1 err =
1, i, j, k, l = 1, 2, ...n}.

Proposition 1.2.34. Let QUMn (C), 1 Tr be the C ∗ algebra with generators akl
ij and the
n
following defining relations:
n
X
akv vl kl
ij ars = δjr ais , i, j, k, l, r, s = 1, 2, ..., n,
v=1

n
ji
X
asr si
lv avk = δjr alk , i, j, k, l, r, s = 1, 2, ..., n,
v=1
∗
akl lk
ij = aji , i, j, k, l = 1, 2, ..., n,
n
X
akl
rr = δkl , k, l = 1, 2, ..., n,
r=1
n
X
arr
kl = δkl , k, l = 1, ..., n.
r=1

Then,
Chapter 1: Preliminaries 34

( 1 ) QUMn (C), 1 Tr is a CQG with coproduct ∆ defined by ∆(akl ) = nr,s=1 akl

P
n
ij rs ⊗
rs
aij , i, j, k, l = 1, 2, ..., n.
( 2 ) QUMn (C), 1 Tr has a C ∗ action α1 on Mn (C) given by α1 (eij ) = nk,l=1 ekl ⊗
P
n
akl
ij , i, j = 1, 2, ..., n. Moreover, QUMn (C), n1 Tr is the universal object in the category of
all CQG s having C ∗ action on Mn (C) such that the functional n1 Tr is kept invariant
under the action.
( 3 ) There does not exist any universal object in the category of all CQG s having
∗
C action on Mn (C).

Proposition 1.2.35. Since, any faithful state on a finite dimensional C ∗ algebra A is

of the form Tr(Rx) for some operator R, it follows from Theorem 6.1, ( 2 ) of [60] that
the universal CQG acting on any finite dimensional C ∗ algebra preserving a faithful
state φ exists and is going to be denoted by QUA,φ .

Notations:
We conclude this section on quantum groups by fixing some notations which will be
used throughout this thesis. In particular, given a compact quantum group (S, ∆), the
dense unital Hopf ∗-subalgebra of S generated by the matrix elements of the irreducible
unitary representations will be denoted by S0 . Moreover, given an action γ : B → B ⊗ S
of the compact quantum group (S, ∆) on a unital C ∗ -algebra B, the dense, unital ∗-
subalgebra of B on which the action becomes an action by the Hopf ∗-algebra S0 will be
denoted by B0 . We shall use the Sweedler convention of abbreviating γ(b) ∈ B0 ⊗alg S0 by
b(1) ⊗ b(2) , for b in B0 . This applies in particular to the canonical action of the quantum
group S on itself, by taking γ = ∆.
Moreover, for a linear functional f on S and an element c in S0 we recall the
‘convolution’ maps f / c := (f ⊗ id)∆(c) and c . f := (id ⊗ f )∆(c). We also define
convolution of two functionals f and g by (f  g)(c) = (f ⊗ g)(∆(c)).

1.3 Rieffel deformation

In this section, we recall the notions of Rieffel’s formulation of deformation quantization
( [46] ) as well as Rieffel type deformation of CQG s due to Rieffel and Wang ( as in [47]
and [62] ).
We begin with Rieffel deformation ( as in [46] ) from action of Rn on a C ∗ algebra.
In the following discussion and henceforth, the symbol e(x) will stand for e2πix . Let V be
a real vector space of dimension n and α be its strongly continuous isometric action on
a complex Frechet space A. Let {kkj } denote the family of seminorms which determine
the topology of A. It is assumed that α is isometric for each of the given seminorms on
A.
35 Rieffel deformation

Let A∞ denote the space of smooth vectors for the action α, that is, A∞ = {a ∈
A : v → αv (a) is C∞ }.
Let {X1 , X2 , ........, Xn } be a basis of V and δk denotes the operator of partial dif-
ferentiation on A∞ in the direction of Xk .
For any multi index µ = (µ1 , µ2 , .....µn ), we will let δ µ = δ µ1 δ µ2 ...δ µn , µ! =
Pn ∞ with the semi norms: kak
µ1 !......µn !, |µ| = i=1 µi . We will equip A jk =
P kδ µ aki
supi≤j |µ|≤k µ! .
Let Cb (V, A) denote the Frechet space of continuous bounded functions from V to A,
equipped with the semi norms kf kk = supv∈V kf (v)kk . There is also a natural action of
V on Cb (V, A) by translation and let Cu (V, A) denote the largest subspace of Cb (V, A)
on which this action is strongly continuous and let B A (V ) denote the space of smooth
vectors with respect to this action.
Let W = V × V. Then B A (W ) makes sense and for F in B A (W ), one can define the
RR
oscillatory integral F (u, v)e(u.v)dudv ( where u.v denote the usual inner product )
in the following way:
We choose a basis of W and let L denote the lattice of points of W which have
integer co-ordinates w.r.t this basis. Moreover, choose a positive φ0 in Cc∞ (W ) such
P
that Φ = p∈L φp vanishes nowhere on W where φp denotes the translate of φ by p
belonging to L. Let φ = φΦ0 .
It can be shown ( [46] ) that for F in B A (W ),
P R
p∈L (F φp )(u, v)e(u.v)dudv con-
RR
verges absolutely in A and F (u, v)e(u.v)dudv is defined to be this sum. Moreover,
this sum is independent of the choice of lattice and of φ. Thus,
Z Z XZ
F (u, v)e(u.v)dudv = (F φp )(u, v)e(u.v)dudv. (1.3.1)
p∈L

For more details of oscillatory integral, we refer to [46] and references therein.
We will need the following results from [46].

Proposition 1.3.1. ( Corollary 1.12, [46] ) Let F be a function in B A (V × V ) which

depends only on the first variable, so that it is essentially an element of B A (V ). Then
RR
F (u)e(u.v)dudv = F (0). The same is true if instead F depends only on the second
variable.

Proposition 1.3.2. ( Proposition 1.14, [46] ) Let S be a continuous linear transfor-

mation from A into a Frechet space C. Let F belongs to B A (W ). Then S ◦ F belongs to
B C (W ) and S(
RR RR
F (u, v)e(u.v)dudv) = S(F (u, v))e(u.v)dudv.

Now, let A be a Frechet algebra. Fix a skew symmetric matrix J on V. Then for all
a, b in A∞ , αJu (a)αv (b) belongs to B A (W ) and a new product ×J is defined on A∞ by
R R
declaring a ×J b = V V αJu (a)αv (b)e(u.v)dudv.
Chapter 1: Preliminaries 36

If the Frechet algebra has an involution ∗ which is continuous and if α acts by ∗

automorphisms, then ∗ is also an involution for the deformed product ×J .

Proposition 1.3.3. ( Lemma 2.20, [46] ) Let f, g belongs to B A and let g have the
lattice L as a period lattice so that g can be viewed as a smooth function on the compact
RR P R
group H = V /L. Then f (u)g(v)e(u.v)dudv = L f (p)( H g(v)e(p.v)dv). A similar
statement holds if instead it is f which is periodic.

Corollary 1.3.4. Z Z
e(θz1 )e(z2 )e(z1 .z2 )dz1 dz2 = e(−θ).

Proof : We have,
Z Z
e(θz1 )e(z2 )e(z1 .z2 )dz1 dz2
X Z
= e(θp)( e(z2 )e(p.z2 )dz2 )
p∈Z S1
X
= e(θp)δp,−1
p∈Z
= e(−θ).

Now, we will define the C ∗ algebra constructed by Rieffel corresponding to the data
(A, V, α, J) where A is also assumed to be a C ∗ algebra and α a C ∗ automorphism.
Let S A be the space of A valued smooth functions on V such that the product of
their derivatives with any complex valued polynomials on V are bounded under the
supremum norm of S A . Then S A is a pre Hilbert right A module with A valued inner
product defined by Z
hf, giA = f (v)∗ g(v)dv,
V

for f, g belonging to S A.
Then, for a in A, one defines the operator Lea on S A by
Z Z
Lea (f )(x) = αx+Ju (a)f (x + v)e(u.v)dudv,
V V

where f belongs to S A . Then

Proposition 1.3.5. ( Theorem 4.6 of [46] ) Lea is a bounded operator having an adjoint
on the pre Hilbert module S A and a 7→ Lea is a ∗ representation of the algebra (A∞ , ×J )
into the C ∗ algebra of bounded operators on S A .
37 Rieffel deformation

Now, by defining
kakJ = kLea k ,

we have a pre-C ∗ norm kkJ on A∞ endowed with the new product ×J .

The completion of this pre C ∗ algebra is the deformed C ∗ algebra and is denoted by
AJ .
One has a natural Frechet topology on A∞ J , given by a family of seminorms {kkn,J }
P kαX µ (a)kJ
where kakn,J = |µ|≤n µ!
We recall the following Proposition from [46].

Proposition 1.3.6. ( Proposition 4.10, [46] ) Let J be fixed. Then for large enough k
there is a constant ck such that for all a in A∞ , we have kakJ ≤ ck kak2k .

Proposition 1.3.7. ( Proposition 7.1, [46] ) Let α be an action of V on the C ∗ algebra

A, with A∞ its subalgebra of smooth vectors. Let J be a skew-symmetric operator on
V, and let α also denote the corresponding action of V on AJ . Then the subalgebra of
smooth vectors in AJ for α is exactly A∞ . Moreover, (AJ )−J ∼= A.

Corollary 1.3.8. A∞ and A∞

J coincide as topological ( Frechet ) spaces.

Proof : The proof is essentially contained in the proof of Proposition 7.1 in [46] (
Proposition 1.3.7 above ). By Proposition 1.3.6, we know that there is a constant ck
such that for any a in A∞ and for any µ,

kX µ akJ ≤ ck kX µ ak2k ≤ c0k kakj

for j = |µ| + 2k and a new constant c0k . Thus, the inclusion of A∞ into A∞
J is continuous
for their Frechet topologies. Similarly, using (AJ )−J = A, we deduce that the inclusion
of A∞J into A
∞ is continuous. This proves the result. 2

Examples

The Noncommutative Torus

Let A = C(Tn ). For v = (v1 , v2 , ..., vn ) in Rn , x = (x1 , x2 , ..., xn ) in Tn , f in C(Tn ),
the action α of Rn on A is given by αv f (x) = f (x1 e(v1 ), x2 e(v2 ), ...xn e(vn )). Let θ
be a n × n skew symmetric matrix and J = 2θ . Then AJ can be seen to be equal to
the noncommutative n tori Tnθ , that is the universal C ∗ algebra generated by unitaries
Ui , i = 1, 2, ..., n satisfying Ui Uj = e(θij )Uj Ui where θij denotes the (i, j) th entry of
the matrix θ. We will denote T2θ by the notation Aθ .

The Rieffel deformed spheres

For a skew symmetric matrix θ, we recall from [19], the definition of Sθn .
Chapter 1: Preliminaries 38

Let µ, ν = 1, 2, ....n. Let λµν = eiθµν where θµν is the (µ, ν) th entry of the matrix
θ.
Sθ2n−1 is the universal C ∗ algebra generated by 2n elements z µ , z µ with relations:

z µ z ν = λµν z ν z µ , z µ z ν = λµν z ν z µ , (1.3.2)

z µ z ν = λνµ z ν z µ , (1.3.3)

(z µ )∗ = z µ , (1.3.4)
n
X
z µ z µ = 1. (1.3.5)
µ=1

It can be easily seen that Sθ2n−1 is obtained by the Rieffel deformation of C(S 2n−1 ) using
µ,ν
the 2n × 2n skew symmetric matrix J whose (µ, ν) th entry is λ 2 and the R2n action
on C(S 2n−1 ) given by αv f (x1 , ..., x2n ) = f (x1 e(v1 ), ..., x2n e(v2n )) ( v = (v1 , ..., v2n ) is in
R2n , f is in C ∞ (S 2n−1 ) ).
Sθ2n is the universal C ∗ algebra generated by 2n + 1 elements {z µ , z µ , x : µ =
1, 2, .., n} where z µ , z µ satisfy ( 1.3.2 ) - ( 1.3.4 ), x is a self adjoint element satisfying
the relations xz µ = z µ x for all µ = 1, 2, ...n and nµ=1 z µ z µ + x2 = 1.
P

Sθ2n is the Rieffel deformation of C(S 2n ) by the action of R2n+1 on C(S 2n ) similar
µ,ν
to above and a (2n + 1) × (2n + 1) matrix J 0 such that (J 0 )µ,ν = λ 2 if µ ≤ 2n, ν ≤ 2n
and 0 otherwise.

1.3.1 Rieffel Deformation of compact quantum group

Here we describe the Rieffel deformation of a CQG as in [62].
Let (A, ∆) be a CQG with C(Tn ) as a quantum subgroup. Let π be the correspond-
ing CQG morphism from A to C(Tn ).
Let η be the canonical homomorphism from Rn to Tn given by η(x1 , x2 , ......, xn ) =
(e(x1 ), e(x2 ), .....e(xn )) and evx be the state on C(Tn ) obtained by evaluation of a func-
tion at the point x in Tn .
Now, put
λη(s) = (evη(−s) π ⊗ id)∆, (1.3.6)

ρη(u) = (id ⊗ evη(u) π)∆. (1.3.7)

We will use the notation Ω(u) for evη(u) π.

Then there is a R2n action on A defined by

χ(s,u) = λη(s) ρη(u) . (1.3.8)

39 Rieffel deformation

Fix a skew symmetric matrix J on Rn and put

Je = J ⊕ (−J).

Then, by the prescription of Rieffel as described above, we have a C ∗ algebra AJe.

Shuzhou Wang showed in [62] that AJe can be made into a CQG.
The ∗-algebra generated by the matrix elements of unitary irreducible representa-
tions of A ( denoted by A0 ) is dense in the space A∞ of smooth vectors of the action
χ under the Frechet topology and hence is dense in the C ∗ algebra AJe under the C ∗
norm of AJe. On A0 , the Hopf ∗-algebra structure remains unchanged and this extends
to a CQG structure on AJe.
We quote the following result ( Remark 3.10 ( 2 ), [62] ) which will be used later.

Proposition 1.3.9. The Haar measure hJe of AJe is still the same as the Haar measure
on the common subspace A0 .

Lemma 1.3.10. The Haar state (say h) of A coincides with the Haar state on AJe (
say hJ ) on the common subspace A∞ , and moreover, h(a ×Je b) = h(ab) for a, b in A∞ .

Proof : We recall ( Proposition 1.3.9 ) that h = hJ on A0 . By using h(Ω(−s) ⊗ id) =

Ω(−s)(id ⊗ h) and h(id ⊗ Ω(u)) = Ω(u)(h ⊗ id), we have for a in Q0 ,

h(χs,u (a))
= Ω(−s)(id ⊗ h)∆(id ⊗ Ω(u))∆(a)
= Ω(−s)(h((id ⊗ Ω(u))∆(a))1)
= h((id ⊗ Ω(u))∆(a))
= Ω(u)(h(a).1)
= h(a).

Therefore,
hχs,u (b) = h(b) for all b in Q0 . (1.3.9)

Now,

h(a×Jeb)
Z Z
= e (a)χv (b))e(u.v)dudv
h(χJu
Z Z
= h(χv (χJu−v
e (a)b))e(u.v)dudv
Z Z
= h(χt (a)b)e(s.t)dsdt,
Chapter 1: Preliminaries 40

where s = −u, t = Ju
e − v, which by Proposition 1.3.1 equals h(χ0 (a)b) = h(ab). That
is, we have proved
ha, biJ = ha, bi ∀a, b ∈ Q0 , (1.3.10)

where h·, ·iJ and h·, ·i respectively denote the inner products of L2 (hJ ) and L2 (h). We
now complete the proof of the lemma by extending ( 1.3.10 ) from Q0 to Q∞ , by using
the fact that Q∞ is a common subspace of the Hilbert spaces L2 (h) and L2 (hJ ) and
moreover, Q0 is dense in both these Hilbert spaces. In particular, taking a = 1 in Q0 ,
we have h = hJ on Q∞ . 2

Remark 1.3.11. Lemma 1.3.10 implies in particular that for every fixed a1 , a2 in Q0 ,
the functional Q0 3 b 7→ h(a1 ×Je b ×Je a2 ) = h(b ×Je a2 ×Je (f1 / a1 . f1 )) ( where f1 is as
in Remark 1.2.20 ) = h(b(a2 ×Je (f1 / a1 . f1 ))) extends to a bounded linear functional
on Q.

Let e be the identity of T2n and Un be a sequence of neighbourhoods of e shrink-

ing to e, fn smooth, positive functions with support contained inside Un such that
R
T2n fn (z)dz = 1 for all n.
Let us denote the action of T2n action on A∞ induced by χ by χe. Define λfn (a) =
R
T2n χ
ez (a)fn (z)dz. Then, we have the following result:

Lemma 1.3.12. λfn (a) belongs to Q∞ and

Z
ez (a)fn (z)dz → a as n → ∞.
χ
T2n

Proof : We note that, by using the translation invariance of Haar measure, for
eg (λfn (a)) = T2n fn (g −1 h)e
all g in T2n , χ
R
χh (a)dh. Therefore, χ eg (λfn (a)) − λfn (a) =
−1
R
T2n (fn (g h) − fn (h))eχh (a)dh which proves the first part.
R
Now we prove the second part. As T2n fn (z) = 1 for all n and supp (fn ) ⊆ Un , we
have
Z

2n χ
ez (a)fn (z) − a
T
Z Z

= ez (a)fn (z) − a
χ fn (z)dz

2n T2n
ZT

= 2n (eχz (a) − a)fn (z)dz
ZT

= χz (a) − a)fn (z)dz
(e .
Un

Now, using the fact that the map z 7→ χez (a) is continuous for all a, we deduce that
for all  > 0, there exists n such that for all z in Un , keχz (a) − χ
e0 (a)k < , that is,
41 Rieffel deformation

ke
χz (a) − ak < .
2
R R
Hence, T2n λz (a)fn (z)dz − a ≤  T2n fn (z) =  which proves the lemma.

Lemma 1.3.13. If h is faithful on Q, then hJ is faithful on QJe.

Proof : Let a ≥ 0, ∈ QJe be such that hJ (a) = 0. Let λfn be as defined above.
Then,

hJ (λfn (a))
Z
= hJ (e
χz (a))fn (z)dz
2n
ZT
= hJ (a)fn (z)dz
T2n
( by (1.3.9))
= 0,

so we have h(λfn (a)) = 0, since h and hJ coincide on Q∞ by Lemma 1.3.10 and λfn (a)
belongs to Q∞ .
Now we fix some notation which we are going to use in the rest of the proof. Let
L (h) and L2 (hJ ) denote the G.N.S spaces of Q and QJe respectively with respect to
2

the Haar states. Let i and iJ be the canonical maps from Q and QJe to L2 (h) and
L2 (hJ ) respectively. Also, let ΠJ denote the G.N.S representation of QJ . Using the
facts h(b∗ ×Je b) = h(b∗ b) for all b in Q∞ and h = hJ on Q∞ ( by Lemma 1.3.10 ) ,
we get kiJ (b)k2L2 (hJ ) = ki(b)k2L2 (h) for all b in Q∞ . So the map sending i(b) to iJ (b) is
an isometry from a dense subspace of L2 (h) onto a dense subspace of L2 (hJ ), hence it
extends to a unitary, say Γ : L2 (h) → L2 (hJ ). We also note that the maps i and iJ
agree on Q∞ .
Now, a ≥ 0 means that λfn (a) is positive in QJe and therefore, λfn (a) = b∗ ×Je b for
some b in QJe. So h(λfn (a)) = 0 implies hJ (λfn (a)) = 0 and therefore kiJ (b)k2L2 (hJ ) = 0.
Therefore, one has ΠJ (b∗ )iJ (b) = 0, and hence iJ (b∗ b) = iJ (λfn (a)) = 0. It thus follows
that Γ(i(λfn (a))) = 0, which implies i(λfn (a)) = 0. But the faithfulness of h means
that i is one one, hence λfn (a) = 0 for all n. Thus, recalling Lemma 1.3.12, we have
a = limn→∞ λfn (a) = 0, which proves the faithfulness of hJ . 2
At this point, we note a useful implication of the Lemma 1.3.10. Let us make use of
the identification of Q0 as a common vector-subspace of all QJe. To be precise, we shall
sometimes denote this identification map from Q0 to QJe by ρJ .

Corollary 1.3.14. Let W be a finite-dimensional (say, n-dimensional) unitary repre-

sentation of Q, with W
f belonging to Mn (C)⊗Q0 be the corresponding unitary. Then, for
any J, we have that W
fJ := (id⊗ρJ )(Wf ) is unitary in Q e, giving a unitary n-dimensional
J
Chapter 1: Preliminaries 42

representation of QJe. In other words, any finite dimensional unitary representation of

Q is also a unitary representation of QJe.

Proof:
Since the coalgebra structures of Q and QJe are identical, and W fJ is identical with Wf
as a linear map, it is obvious that W fJ gives a nondegenerate representation of Q e.
J
Let y = (id ⊗ h)(W f∗W
J J ). It follows from the proof of Proposition 6.4 of [41] that y is
f
1
fJ (y − 12 ⊗1) gives a unitary representation
invertible positive element of Mn and (y 2 ⊗1)W
of QJe. We claim that y = 1, which will complete the proof of the corollary. For
convenience, let us write W in the Sweedler notation: W = w(1) ⊗ w(2) . We note that
by Lemma 1.3.10, we have

f∗W
(id ⊗ h)(W J J)
f
∗ ∗
= w(1) w(1) h(w(2) ×Je w(2) )
∗ ∗
= w(1) w(1) h(w(2) w(2) )
f∗W
= (id ⊗ h)(W f ) = (id ⊗ h)(1 ⊗ 1) = 1.

2
Example The Rieffel deformed orthogonal groups
Let θ be a n × n skew symmetric matrix. C(Tn ) sits inside C(O(n)) as a quantum
subgroup. It can be easily seen that Oθ (n) is obtained by Rieffel deformation from
C(O(n)) by using the induced R2n action as given in the equation ( 1.3.8 ) and consid-
ering the matrix Je = −J ⊕ J when n is even and −J 0 ⊕ J 0 when n is odd where J and
J 0 are the matrices introduced while giving the definition of the θ deformed spheres.

1.4 Classical Riemannian geometry

In this section we recall some classical facts regarding manifolds which will be useful to
us later on.

1.4.1 Classical Hilbert space of forms

Let M be an n dimensional Riemannian manifold and Ωk (M ) ( k = 0, 1, 2, ...n ) be
the space of smooth k-forms. Set Ωk (M ) = {0} for k > n. The de-Rham differential
d maps Ωk (M ) to Ωk+1 (M ). Let Ω ≡ Ω(M ) = ⊕k Ωk (M ). We will denote the Rieman-
nian volume element by dvol. We recall that the Hilbert space L2 (M ) is obtained by
completing the space {f ∈ Cc∞ (M )} with respect to the pre-inner product given by
R
hf1 , f2 i = M f1 f2 dvol.
43 Classical Riemannian geometry

In an analogous way, one can construct a canonical Hilbert space of forms. The
Riemannian metric h. , .im ( for m in M ) on Tm M induces an inner product on the
vector space Tm∗ M and hence also Λk T ∗ M, which will be again denoted by h. , .i .
m m
This gives a natural pre-inner product on the space of compactly supported k-forms by
integrating the compactly supported smooth function m 7→ hω(m), η(m)im over M. We
will denote the completion of this space by Hk (M ). Let H = ⊕k Hk (M ).
Then, one can view d : Ω → Ω as an unbounded, densely defined operator ( again
denoted by d ) on the Hilbert space H with the domain Ω. It can be verified that it is
closable.

1.4.2 Isometry groups of classical manifolds

Let M be a Riemannian manifold of dimension n. Then the collection of all isometries of
M has a natural group structure and is denoted by ISO(M ). Let C and U be respectively
a compact and open subset of M and let W (C, U ) = {h ∈ ISO(M ) : h.C ⊆ U }. The
compact open topology on ISO(M ) is the smallest topology on ISO(M ) for which the
sets W (C, U ) are open. It follows ( see [34] ) that under this topology, ISO(M ) is a
closed locally compact topological group. Moreover, if M is compact, ISO(M ) is also
compact.
We recall that the Laplacian L on M is an unbounded densely defined self adjoint
operator −d∗ d on the space of zero forms H0 (D) = L2 (M, dvol) which has the local
expression
n
1 X ∂ p ∂
L(f ) = − p (g ij det(g) f)
det(g) ∂xj ∂xi
i,j=1

for f in C ∞ (M ) and where g = ((gij )) is the Riemannian metric and g −1 = ((g ij )).
It is well known that on a compact manifold, the Laplacian has compact resolvents.
Thus, the set of eigenvalues of L is countable, each having finite multiplicities, and
accumulating only at infinity. Moreover, there exists an orthonormal basis of L2 (M )
consisting of eigenvectors of L which belong to C ∞ (M ). It can be shown ( Lemma 2.3
of [30] ) that for a compact manifold, the complex linear span of the eigenvectors of L
is dense in C ∞ (M ) in the sup norm.

The following result is in the form in which it has been stated and proved in [30] (
Proposition 2.1 ).
Proposition 1.4.1. Let M be a compact Riemannian manifold. Let L be the Laplacian
of M. A smooth map γ : M → M is a Riemannian isometry if and only if γ commutes
with L in the sense that L(f ◦ γ) = (L(f )) ◦ γ for all f in C ∞ (M ).
Using this fact, we give an operator theoretic proof of the fact that for a compact
manifold, ISO(M ) is compact. As the action of ISO(M ) commutes with the Laplacian,
Chapter 1: Preliminaries 44

it has a unitary representation on L2 (M ). As the action preserves the finite dimensional

eigenspaces of the Laplacian, ISO(M ) is a subgroup of U (d1 ) × U (d2 ) × ..........( where
{di : i ≥ 0} denote the dimension of the eigenspaces of the Laplacian and U (d) denotes
the group of unitary operators on a Hilbert space of dimension d ) which is a compact
group. As ISO(M ) is closed, it is a closed subgroup of a compact group, hence compact.
We will see that this technique can be generalized in the noncommutative set-up in the
chapters 2 and 3.
Proposition 1.4.1 has the generalization in a more general context.
Let us fix some notations. Let Y be a compact metrizable space and θ : M ×Y → M.
Let ξy : M → M defined by ξy (m) = θ(m, y). Let α : C(M ) → C(M ) ⊗ C(Y ) ∼ =
C(M × Y ) be defined by α(f )(m, y) = f (θ(m, y)) for all y in Y, m in M. For a state
φ on C(Y ), denote by αφ , the map: (id ⊗ φ)α : C(M ) → C(M ). Lastly, let A∞ 0 be the
span of eigenvectors of the Laplacian L of M.
Then, we have the following( Lemma 2.5 of [30] ):

Proposition 1.4.2. The following are equivalent:

a. For every y in Y, ξy is smooth isometric.
b. For every state φ on C(Y ), we have αφ (A∞ ∞ ∞
0 ) ⊆ A0 and αφ L = Lαφ on A0 .

Example 1.4.3. 1. The isometry group of the n-sphere S n is O(n+1) where the action
is given by the usual action of O(n + 1) on Rn+1 . The subgroup of O(n + 1) consisting
of all orientation preserving isometries on S n is SO(n + 1).
2. The isometry group of the circle S 1 is S 1 >
Z2 . Here the Z2 ( = {0, 1} ) action
on S 1 is given by 1.z = z where z is in S 1 while the action of S 1 is its action on itself.
3. ISO(Tn ) = ∼ Tn >
(Zn >
Sn ) where Sn is the permutation group on n sym-
2
bols. Here an element of Sn acts on an element (z1 , z2 , ..., zn ) ∈ Tn by permutation.
If the generator of i-th copy of Zn2 is denoted by 1i , then the action of 1i is given by
1i (z1 , z2 , ..., zn ) = (z1 , ..., zi−1 , zi , zi+1 , ..., zn ) where (z1 , z2 , ..., zn ) ∈ Tn . Lastly, the ac-
tion of Tn on itself is its usual action.

1.4.3 Spin Groups and Spin manifolds

We begin with Clifford algebras. Let Q be a quadratic form on an n dimensional vector
space V. Then Cl(V, Q) will denote the universal associative algebra C equipped with
a linear map i : V → C, such that i(V ) generates C as a unital algebra satisfying
i(V )2 = Q(V ).1
Let β : V → Cl(V, Q) be defined by β(x) = −i(x). Then, Cl(V, Q) =
Cl (V, Q) ⊕ Cl1 (V, Q) where Cl0 (V, Q) = {x ∈ Cl(V, Q) : β(x) = x}, Cl1 (V, Q) =
0

{x ∈ Cl(V, Q) : β(x) = −x}.

45 Classical Riemannian geometry

We will denote by Cn and CnC the Clifford algebras Cl(Rn , −x21 − ... − x2n ) and
Cl(Cn , z12 + ... + zn2 ) respectively.
[n]
We will denote the vector space C2 2 by the symbol ∆n . It follows that CnC =
End(∆n ) if n is even and equals End(∆n ) ⊕ End(∆n ) is n is odd. There is a represen-
tation CnC → End(∆n ) which is the isomorphism with End(∆n ) when n is even and in
the odd case, it is the isomorphism with End(∆n ) ⊕ End(∆n ) followed by the projection
onto the first component. This representation restricts to Cn , to be denoted by κn and
called the spin representation. This representation is irreducible when n is odd and for
n even, it decomposes into two irreducible representations which decomposes ∆n into a
direct sum of two vector spaces ∆+ −
n and ∆n .
Pin(n) is defined to be the subgroup of Cn generated by elements of the form {x :
kxk = 1, x ∈ Rn }. Spin ( n ) is the group given by Pin(n)∩ Cn0 . There exists a continuous
group homomorphism from Pin(n) to O(n) which restricts to a two covering map λ :
Spin(n) → SO(n).

Let M be an n-dimensional orientable Riemannian manifold. Then we have the

oriented orthonormal bundle of frames over M ( which is a principal SO(n) bundle )
which we will denote by F.
Such a manifold M is said to be a spin manifold if there exists a pair (P, Λ) (
called a spin structure ) where
( 1 ) P is a Spin( n ) principal bundle over M.
( 2 ) Λ is a map from P to F such that it is a 2-covering as well as a bundle map
over M.
( 3 ) Λ(p.bg ) = Λ(p).g where λ(b
g ) = g.
Given such a spin structure, we consider the associated bundle S = P ×Spin(n) ∆n
called the ‘ bundle of spinors ’.

1.4.4 Dirac operators

We follow the notations of the previous subsection. On the space of smooth sections of
the bundle of spinors S, one can define an inner product by
Z
hs1 , s2 iS = hs1 (x), s2 (x)i dvol(x)
M

The Hilbert space obtained by completing the space of smooth sections with respect
to this inner product is denoted by L2 (S) and its members are called square integrable
spinors. The Levi Civita connection on M induces a canonical connection on S which
we will denote by ∇S .

Definition 1.4.4. The Dirac operator on M is the self-adjoint extension of the fol-
Chapter 1: Preliminaries 46

lowing operator D defined on the space of smooth sections of S :

n
X
(Ds)(m) = κn (Xi (m))(∇SXi s)(m),
i=1

where (X1 , ...Xn ) are local orthonormal ( with respect to the Riemannian metric ) vector
fields defined in a neighborhood of m. In this definition, we have viewed Xi (m) belonging
to Tm (M ) as an element of the Clifford algebra ClC (Tm M ), hence κn (Xi (m)) is a map
on the fibre of S at m, which is isomorphic with ∆n . The self-adjoint extension of D is
again denoted by the same symbol.

We recall three important facts about the Dirac operator:

Proposition 1.4.5. ( 1 ) C ∞ (M ) acts on S by multiplication and this action extends

to a representation, say π, of the C ∗ algebra C(M ) on the Hilbert space L2 (S).
( 2 ) For f in C ∞ (M ), [D, π(f )] has a bounded extension.
( 3 ) Furthermore, the Dirac operator on a compact manifold has compact resolvents.

As the action of an element f in C ∞ (M ) on L2 (S) is by multiplication operator, we

will use the symbol Mf in place of π(f ).
The Dirac operator carries a lot of geometric and topological information. We give
two examples.
( a ) The Riemannian metric of the manifold is recovered by

d(p, q) = supφ∈C ∞ (M ), k[D,Mφ ]k≤1 |φ(P ) − φ(q)| . (1.4.1)

2
( b ) For a compact manifold, the operator e−tD is trace class for all t > 0. Then
the volume form of the manifold can be recovered by the formula
2
Tr(Mf e−tD )
Z
f dvol = c(n)limt→0
M Tr(e−tD2 )

where dimM = n, c(n) is a constant depending on the dimension.

1.5 Noncommutative Geometry

In this section, we recall those basic concepts of noncommutative geometry which we
are going to need. We refer to [17], [40], [22] for more details.

1.5.1 Spectral triples

Motivated by the facts in Proposition 1.4.5, Alain Connes defined his formulation of
noncommutative manifold based on the idea of a spectral triple:
47 Noncommutative Geometry

Definition 1.5.1. A spectral triple or spectral data is a triple (A∞ , H, D) where H

is a separable Hilbert space, A∞ is a ∗ subalgebra of B(H), ( not necessarily norm closed
) and D is a self adjoint ( typically unbounded ) operator such that for all a in A∞ ,
the operator [D, a] has a bounded extension. Such a spectral triple is also called an odd
spectral triple. If in addition, we have γ in B(H) satisfying γ = γ ∗ = γ −1 , Dγ = −γD
and [a, γ] = 0 for all a in A∞ , then we say that the quadruplet (A∞ , H, D, γ) is an even
spectral triple. The operator D is called the Dirac operator corresponding to the spectral
triple.

Furthermore, given an abstract ∗-algebra B, an odd ( even ) spectral triple on B is

an odd ( even ) spectral triple ( π(B), H, D ) ( respectively, ( π(B), H, D, γ ) ) where
π : B → B(H) is a ∗-homomorphism.
Since in the classical case, the Dirac operator has compact resolvent if the manifold
is compact, we say that the spectral triple is of compact type if A∞ is unital and D
has compact resolvent.

Definition 1.5.2. We say that two spectral triples ( π1 (A), H1 , D1 ) and ( π2 (A), H2 , D2
) are said to be unitarily equivalent if there is a unitary operator U : H1 → H2 such
that D2 = U D1 U ∗ and π2 (.) = U π1 (.)U ∗ where πj , j = 1, 2 are the representations of A
in Hj , respectively.

Next, we will give two examples of spectral triples in classical geometry and a non-
classical example. We will give more examples in chapters 3 and 5.

Example 1.5.3. Let M be a smooth spin manifold. Then from Proposition 1.4.5, we
see that ( C ∞ (M ), H, D ) is a spectral triple over C ∞ (M ) and is of compact type if M
is compact.
We recall that when the dimension of the manifold is even, ∆n = ∆+ − 2
n ⊕ ∆n . An L
section s has a decomposition s = s1 + s2 where s1 (m), s2 (m) belongs to ∆+ n (m) and
− ±
∆n (m) ( for all m ) respectively where ∆n (m) denotes the subspace of the fibre over m.
This decomposition of L2 (S) induces a grading operator γ on L2 (S). It can be seen that
D anticommutes with γ.

Example 1.5.4. This example comes from the classical Hilbert space of forms discussed
in subsection 1.5.2. One considers the self adjoint extension of the operator d + d∗ on
H = ⊕k Hk (M ) which is again denoted by d + d∗ . C ∞ (M ) has a representation on
each Hk (M ) which gives a representation, say π on H. Then it can be seen that (
C ∞ (M ), H, d + d∗ ) is a spectral triple and d + d∗ is called the Hodge Dirac operator.
When M is compact, this spectral triple is of compact type.
Chapter 1: Preliminaries 48

Remark 1.5.5. Let us make it clear that by a ‘classical spectral triple’ we always mean
the spectral triple obtained by the Dirac operator on the spinors (so, in particular, man-
ifolds are assumed to be Riemannian spin manifolds), and not just any spectral triple
on the commutative algebra C ∞ (M ).

Example 1.5.6. The Noncommutative torus

We recall from subsection 1.1.1 that the noncommutative 2-torus Aθ is the universal
∗
C algebra generated by two unitaries U and V satisfying U V = e2πiθ V U where θ is a
number in [0, 1].
There are two derivations d1 and d2 on Aθ obtained by extending linearly the rule:

d1 (U ) = U, d1 (V ) = 0,

d2 (U ) = 0, d2 (V ) = V.

Then d1 and d2 are well defined on the following dense ∗-subalgebra of Aθ :

X  
A∞
θ ={ amn U m V n : supm,n mk nl amn  < ∞ for all k, l in IN }.
 
m,n∈Z

There is a unique faithful trace on Aθ defined as follows:

X
τ( amn U m V n ) = a00 .

Let H = L2 (τ )⊕L2 (τ ) where L2 (τ ) denotes the GNS Hilbert space of Aθ with respect
!
a 0
to the state τ. We note that A∞θ is embedded as a subalgebra of B(H) by a 7→ .
0 a
!
0 d1 + id2
Now, we define D = .
d1 − id2 0
Then, (A∞ θ , H, D) is a spectral triple of compact type. In particular, for θ = 0, this
coincides with the classical spectral triple on C(T2 ).

1.5.2 The space of forms in noncommutative geometry

We start this subsection by recalling the universal space of one forms corresponding to
an algebra.

Proposition 1.5.7. Given an algebra B, there is a ( unique upto isomorphism ) B − B

bimodule Ω1 (B) and a derivation δ : B → Ω1 (B) ( that is, δ(ab) = δ(a)b + aδ(b) for all
a, b in B ), satisfying the following properties:
(i) Ω1 (B) is spanned as a vector space by elements of the form aδ(b) with a, b be-
longing to B; and
49 Noncommutative Geometry

(ii) for any B − B bimodule E and a derivation d : B → E, there is an unique B − B

linear map η : Ω1 (B) → E such that d = η ◦ δ.

The bimodule Ω1 (B) is called the space of universal 1-forms an B and δ is called the
universal derivation.
We can also introduce universal space of higher forms on B, Ωk (B), say, for k =
2, 3, ..., by defining them recursively as follows: Ωk+1 (B) = Ωk (B) ⊗B Ω1 (B) and also set
Ω0 (B) = B.

Now we briefly discuss the notion of the noncommutative Hilbert space of forms
which will need noncommutative volume form for a spectral triple of compact type. We
refer to [29] ( page 124 -127 ) and the references therein for more details.
Definition 1.5.8. A spectral triple (A∞ , H, D) of compact type is said to be Θ-
2
summable if e−tD is of trace class for all t > 0. A Θ-summable spectral triple is called
p 2
finitely summable when there is some p > 0 such that t 2 Tr(e−tD ) is bounded on (0, δ]
for some δ > 0. The infimum of all such p, say p0 is called the dimension of the spectral
triple and the spectral triple is called p0 -summable.
Remark 1.5.9. We remark that the definition of Θ-summability to be used in this thesis
is stronger than the one in [17] ( page 390, definition 1. ) in which a spectral triple is
2
called Θ-summable if Tr(e−D ) < ∞.
1 2
Tr(T e− λ D )
For a Θ-summable spectral triple, let σλ (T ) = 1 2 for λ > 0. We note that
Tr(e− λ D )
λ 7→ σλ (T ) is bounded.
Let Z λ
1 du
τλ (T ) = σu (T ) for λ ≥ a ≥ e.
log λ a u
Now consider the quotient C∗
algebra B∞ = Cb ([a, ∞))/C0 ([a, ∞)). Let for T in
B(H), τ (T ) in B∞ be the class of λ → τλ (T ).
For any state ω on the C ∗ algebra B∞ , T rω (T ) = ω(τ (T )) for all T in B(H) defines
a functional on B(H). As we are not going to need the choice of ω in this thesis, we will
−tD 2 )
suppress the suffix ω and simply write Limt→0+ Tr(T e−tD2 for T rω (T ). This is a kind
Tr(e )
−tD 2
of Banach limit because if limt→0+ Tr(T e−tD2 ) exists, then it agrees with the functional
Tr(e )
Limt→0+ . Moreover, Trω (T ) coincides ( upto a constant ) with the Dixmier trace ( see
chapter IV, [17] ) of the operator T |D|−p when the spectral triple has a finite dimension
p > 0, where |D|−p is to be interpreted as the inverse of the restriction of |D|p on the
closure of its range. In particular, this functional gives back the volume form for the
classical spectral triple on a compact Riemannian manifold.
Let Ωk (A∞ ) be the space of universal k-forms on the algebra A∞ which is spanned
by a0 δ(a1 ) · · · δ(ak ), ai belonging to A∞ , where δ is as in Proposition 1.5.7. There
Chapter 1: Preliminaries 50

k ∞
L
is a natural graded algebra structure on Ω ≡ k≥0 Ω (A ), which also has a natu-
ral involution given by (δ(a))∗ = −δ(a∗ ), and using the spectral triple, we get a ∗-
representation Π : Ω → B(H) which sends a0 δ(a1 ) · · · δ(ak ) to a0 dD (a1 ) · · · dD (ak ), where
−tD 2
dD (a) = [D, a]. Consider the state τ on B(H) given by, τ (X) = Limt→0+ Tr(Xe−tD2 ) ,
Tr(e )
where Lim is as above. Using τ , we define a positive semi definite sesquilinear form on
Ωk (A∞ ) by setting hw, ηi = τ (Π(w)∗ Π(η)). Let K k = {w ∈ Ωk (A∞ ) : hw, wi = 0},
for k ≥ 0, and K −1 := (0). Let ΩkD be the Hilbert space obtained by completing the
quotient Ωk (A∞ )/K k with respect to the inner product mentioned above, and we define
HDk := P ⊥ Ωk , where P denotes the projection onto the closed subspace generated by
k D k
k−1 ). The map D := d + d∗ ≡ dD + d∗D on Hd+d∗ := k≥0 HD
0 k has a self-adjoint
L
δ(K
extension (which is again denoted by d + d∗ ). Clearly, HD k has a total set consisting

of elements of the form [a0 δ(a1 ) · · · δ(ak )], with ai in A∞ and where [ω] denotes the
equivalence class Pk⊥ (w + K k ) for ω belonging to Ωk (A∞ ). There is a ∗-representation
πd+d∗ : A → B(Hd+d∗ ), given by πd+d∗ (a)([a0 δ(a1 ) · · · δ(ak )]) = [aa0 δ(a1 ) · · · δ(ak )].
Then it is easy to see that

Proposition 1.5.10. (A∞ , Hd+d∗ , d + d∗ ) is a spectral triple.

1.5.3 Laplacian in Noncommutative geometry

Now, we discuss the notion of Laplacian in noncommutative geometry as introduced
in [30]. Let (A∞ , H, D) be a spectral triple of compact type. To define the Laplacian
in the noncommutative case ( as in [30] ), we need the following assumptions on the
spectral triple.
Assumptions
1. (A∞ , H, D) is a compact type spectral triple.
2. It is QC ∞ , that is, A∞ and {[D, a], a ∈ A∞ } are contained in the domains of
all powers of the derivation [|D|, ·].
3. The unbounded densely defined map dD from HD 0 to H1 given by d (a) = [D, a]
D D
∞
for a in A , is closable.
4. L := −d∗D dD has A∞ in its domain, and it is left invariant by L.
−tD 2
Under assumption 2., τ defined by τ (X) = Limt→0 Tr(Xe−tD2 ) is a positive trace on
Tr(e )
the C ∗ -subalgebra generated by A∞ and {[D, a] : a ∈ A∞ }.
5. We assume that it is also faithful on this subalgebra.

Then, L = −d∗D dD is defined to be the Laplacian for the spectral triple (A∞ , H, D).
It coincides with the Hodge Laplacian −d∗ d (restricted on space of smooth functions)
in the classical case, where d denotes the de-Rham differential.
The linear span of eigenvectors of L, which is a subspace of A∞ , is denoted by A∞0 ,
∞ ∞
and the ∗-subalgebra of A generated by A0 is denoted by A0 .
Chapter 2

Quantum isometry groups:

approach based on Laplacian

The idea of quantum isometry group of a noncommutative manifold (given by a spec-

tral triple), which has been defined by Goswami, is motivated by the definition and
study of quantum permutation groups of finite sets and finite graphs by a number of
mathematicians (see, e.g. [1], [2], [60], [61] and references therein).
In this chapter, we first recall the definition of quantum isometry groups as proposed
in [30] and then compute it for some examples.

2.1 Formulation of the quantum isometry group

2.1.1 Characterization of isometry group for a compact Riemannian
manifold
Let M be a compact Riemannian manifold. Consider the category with objects be-
ing the pairs (G, α) where G is a compact metrizable group acting on M by the
smooth and isometric action α. If (G1 , α) and (G2 , β) are two objects in this cate-
gory, Mor((G1 , α), (G2 , β)) consists of group homomorphisms π from G1 to G2 such
that β ◦ π = α. Then the isometry group of M is the universal object in this category.
More generally, the isometry group of a classical compact Riemannian manifold,
viewed as a compact metrizable space ( forgetting the group structure ), can be seen to
be the universal object of a category whose object class consists of subsets ( not generally
subgroups ) of the set of smooth isometries of the manifold. Then it can be proved that
this universal compact set has a canonical group structure. Thus, motivated by the
ideas of Woronowicz and Soltan ( [53], [68] ), Goswami considered in [30] a bigger
category with objects as the pair (S, f ) where S is a compact metrizable space and
f : S × M → M such that the map from M to itself defined by m 7→ f (s, m) is a

51
Chapter 2: Quantum isometry groups: approach based on Laplacian 52

smooth isometry for all s in S. The morphism set is defined as above ( replacing group
homomorphisms by continuous set maps ).
Therefore, to define the quantum isometry group, it is reasonable to consider a
category of compact quantum groups which act on the manifold (or more generally,
on a noncommutative manifold given by spectral triple) in a ‘nice’ way, preserving the
Riemannian structure in some suitable sense, which is precisely formulated in [30], where
it is also proven that a universal object in the category of such quantum groups does
exist if one makes some natural regularity assumptions on the spectral triple.

2.1.2 The definition and existence of the quantum isometry group

Let (A∞ , H, D) be a Θ-summable spectral triple of compact type. We recall from section
1.5 the Hilbert spaces of k-forms HDk , k = 0, 1, 2, ... and also the Laplacian L = −d∗ d .
D D
To define the quantum isometry group, we need the following assumptions:
Assumptions
1. dD is closable and A∞ ⊆ Dom(L) where A∞ is viewed as a dense subspace of
HD0 .

2. L has compact resolvents.

3. L(A∞ ) ⊆ A∞ .
4. Each eigenvector of L ( which has a discrete spectrum, hence a complete set of
eigenvectors ) belongs to A∞ .
5. ( connectedness assumption ) The kernel of L is one dimensional, spanned by the
identity 1 of A∞ , viewed as a unit vector in HD 0 .

6. The complex linear span of the eigenvectors of L, denoted by A∞ 0 is norm dense

in A .∞

Definition 2.1.1. We say that a spectral triple satisfying the assumptions 1. - 6.

admissible.

The following result is contained in Remark 2.16 of [30].

Proposition 2.1.2. If an admissible spectral triple (A∞ , H, D) satisfies the condition

Dom(Ln ) = A∞ , and if α : Ā → Ā ⊗ S is a smooth isometric action on A∞ by a
T

CQG S, then for all state φ on S, αφ (= (id ⊗ φ)α) keeps A∞ invariant.

In view of the characterization of smooth isometric action on a classical compact

manifold ( Proposition 1.4.1 and Proposition 1.4.2 in Chapter 1 ), Goswami gave the
following definition in [30].

Definition 2.1.3. A quantum family of smooth isometries of the noncommutative man-

ifold A∞ ( or more precisely on the corresponding spectral triple ) is a pair (S, α) where
53 Formulation of the quantum isometry group

S is a separable unital C ∗ algebra, α : A → A ⊗ S ( where A denotes the C ∗ alge-

bra obtained by completing A∞ in the norm of B(HD 0 )) is a unital C ∗ homomorphism,

satisfying the following:

a. Sp(α(A)(1 ⊗ S) = A ⊗ S
b. αφ = (id ⊗ φ)α maps A∞ ∞
0 into itself and commutes with L on A0 , for every state
φ on S.
In case, the C ∗ algebra has a coproduct ∆ such that (S, ∆) is a compact quantum
group and α is an action of (S, ∆) on A, we say that (S, ∆) acts smoothly and isomet-
rically on the noncommutative manifold.

Notations
1. We will denote by QL the category with the object class consisting of all quan-
tum families of isometries (S, α) of the given noncommutative manifold, and the set of
morphisms Mor((S, α), (S 0 , α0 )) being the set of unital C ∗ homomorphisms φ : S → S 0
satisfying (id ⊗ φ)α = α0 .
2. We will denote by Q0L the category whose objects are triplets (S, ∆, α) where
(S, ∆) is a CQG acting smoothly and isometrically on the given noncommutative mani-
fold, with α being the corresponding action. The morphisms are the homomorphisms of
compact quantum groups which are also morphisms of the underlying quantum families.

Let {λ1 , λ2 , ...} be the set of eigenvalues of L, with Vi being the corresponding (
finite dimensional ) eigenspace. We will denote by Ui the Wang algebra Au,di (I) (
as introduced in the chapter 1 ) where di is the dimension of the subspace Vi . We
fix a representation βi : Vi → Vi ⊗ Ui on the Hilbert space Vi , given by βi (eij ) =
P (i) (i) ≡ u(i)
k eik ⊗ ukj for j = 1, 2, ...di , where {eij } is an orthonormal basis for Vi , and u kj
are the generators of Ui . Thus, both u(i) and u(i) are unitaries. The representations βi
canonically induce the free product representation β = ∗i βi of the free product CQG
U = ∗i Ui on the Hilbert space HD 0 such that the restriction of β on V coincides with β
i i
for all i.
The following Lemma ( Lemma 2.12 of [30] ) will be needed later and hence we
record it.

Lemma 2.1.4. Consider an admissible spectral triple (A∞ , H, D) and let (S, α) be a
quantum family of smooth isometries of the spectral triple. Moreover, assume that the
action is faithful in the sense that there is no proper C ∗ subalgebra S1 of S such that
α(A∞ ) ⊆ A∞ ⊗ S1 . Then α e : A∞ ⊗ S → A∞ ⊗ S defined by α e(a ⊗ b) = α(a)(1 ⊗ b)
0
extends to an S linear unitary on the Hilbert S module HD ⊗ S, denoted again by α e.
∗
Moreover, we can find a C isomorphism φ : U/I → S between S and a quotient of U
by a C ∗ ideal I of U, such that α = (id ⊗ φ) ◦ (id ⊗ ΠI ) ◦ β on A∞ ⊆ HD 0 , where Π
I
denotes the quotient map from U to U/I.
Chapter 2: Quantum isometry groups: approach based on Laplacian 54

If furthermore, there is a CQG structure on S given by a coproduct ∆ such that α

is a C ∗ action of a CQG on A, then the map α : A∞ → A∞ ⊗ S extends to a unitary
representation ( denoted again by α ) of the CQG (S, ∆) on HD0 . In this case, the ideal

I is a Woronowicz C ∗ ideal and the C ∗ isomorphism φ : U/I → S is a morphism of

CQG s.

Using this, the following result has been proved in [30], which defines and gives the
existence of QISOL .

Theorem 2.1.5. For any admissible spectral triple (A∞ , H, D), the category QL has a
universal object denoted by (QISOL , α0 ). Moreover, QISOL has a coproduct ∆0 such
that (QISOL , ∆0 ) is a CQG and (QISOL , ∆0 , α0 ) is a universal object in the category
Q0L . The action α0 is faithful.

We very briefly outline the main ideas of the proof. The universal object QISOL is
constructed as a suitable quotient of U. Let F be the collection of all those C ∗ -ideals I
of U such that the composition ΓI = (id ⊗ ΠI ) ◦ β : A∞ ∞
0 → A0 ⊗alg (U/I) extends to a
C ∗ -homomorphism from A to A ⊗ (U/I). Then it can be shown that I0 (= ∩I∈F I ) is
again a member of F and (U/I0 , ΓI0 ) is the required universal object. Thus,

Remark 2.1.6. QISOL is a quantum subgroup of the CQG U = ∗i Au,di (I). As Au,di (I)
satisfies κ2 = id, ( by Remark 1.2.29 ) the same is satisfied by QISOL so that by Remark
1.2.20, QISOL has tracial Haar state.

Remark 2.1.7. It is proved in [30] that to ensure the existence of QISOL , the as-
sumption (5) can be replaced by the condition that the action α is τ preserving, that
is, (τ ⊗ id)α(a) = τ (a).1. In [30] it was also shown ( Lemma 2.5, b ⇒ a ) that for
an isometric group action on a not necessarily connected classical manifold, the volume
functional is automatically preserved. It can be easily seen that the proof goes verbatim
for a quantum group action, and consequently we get the existence of QISOL for a (
not necessarily connected ) compact Riemannian manifold.

Unitary representation of QISOL on a spectral triple

We shall also need the following result proved in section 2.4 of [30].

Proposition 2.1.8. QISOL has a unitary representation U ≡ UL on HD such that U

commutes with d + d∗ . Let δ be as in subsection 1.5.2. On the Hilbert space of k-forms,
k , U is defined by:
that is. HD

(1) (1) (1) (2) (2) (2)

U ([a0 δ(a1 ) · · · δ(ak )] ⊗ q) = [a0 δ(a1 ) · · · δ(ak )] ⊗ (a0 a1 · · · ak )q,
55 Computation of QISOL

where q belongs to QISOL , ai belongs to A∞ 0 , and for x in A0 , ( the ∗-subalgebra

generated by the eigenvectors of L ) we write in Sweedler notation α(x) = x(1) ⊗ x(2) ∈
A0 ⊗ (QISOL )0 (α denotes the action of QISOL ).

2.2 Computation of QISOL

Here we compute QISOL for three commutative examples, viz: the sphere, the circle
and the n tori. In Chapter 4, we will be able to compute it for two noncommutative
examples, namely Aθ and Sθn by using Theorem 4.4.7.

2.2.1 The commutative spheres

Let QISOL be the quantum isometry group of S 2 and let α be the action of QISOL
on C(S 2 ). Let L be the Laplacian on S 2 defined as

∂2 ∂ 1 ∂2
L= 2
+ cot(θ) + 2 ,
∂θ ∂θ sin (θ) ∂ψ 2

where the cartesian coordinates x1 , x2 , x3 for S 2 are given by x1 = r cos ψ sin θ, x2 =

∂2
r sin ψ sin θ, x3 = r cos θ. In the cartesian coordinates, L = 3i=1 ∂x
P
2.
i
The eigenspaces of L on S 2 are of the form
X
Ek = Sp{(c1 X1 + c2 X2 + c3 X3 )k : ci ∈ C, i = 1, 2, 3, c2i = 0},

where k ≥ 1. Ek consists of harmonic homogeneous polynomials of degree k on R3

restricted to S 2 ( See [33], page 29-30 ).
We begin with the following lemma, which says that any smooth isometric action
by a quantum group must be ‘linear’.
P3
Lemma 2.2.1. The action α satisfies α(xi ) = j=1 xj ⊗ Qij where Qij belongs to
L
QISO , i = 1, 2, 3.
Proof : Since α is a smooth isometric action of QISOL on C(S 2 ), α has to preserve
the eigenspaces of the Laplacian L. In particular, it has to preserve E1 = Sp{c1 x1 +
P3 2
c2 x2 + c3 x3 : ci ∈ C, i = 1, 2, 3, i=1 ci = 0}.
Now note that x1 + ix2 , x1 − ix2 are in E1 , hence x1 , x2 are in E1 . Similarly x3
belongs to E1 too. Therefore E1 = Sp{x1 , x2 , x3 }, which completes the proof of the
lemma. 2

Now, we state and prove the main result of this section, which identifies QISOL
with the commutative C ∗ algebra of continuous functions on the isometry group of S 2 ,
that is O(3).
Chapter 2: Quantum isometry groups: approach based on Laplacian 56

Theorem 2.2.2. The quantum isometry group QISOL is commutative as a C ∗ algebra,

and hence QISOL ∼
= C(O(3)).

Proof : We begin with the expression

3
X
α(xi ) = xj ⊗ Qij , i = 1, 2, 3,
j=1

and also note that x1 , x2 , x3 form a basis of E1 and {x21 , x22 , x23 , x1 x2 , x1 x3 , x2 x3 } is a
basis of E2 . Since x∗i = xi for each i and α is a ∗-homomorphism, we must have Q∗ij = Qij
for all i, j = 1, 2, 3. Moreover, the condition x21 + x22 + x23 = 1 and the fact that α is a
homomorphism gives:

Q21j + Q22j + Q23j = 1, ∀j = 1, 2, 3.

Again, the condition that xi ,xj commute for all i, j gives

Qij Qkj = Qkj Qij ∀i, j, k, (2.2.1)

Qik Qjl + Qil Qjk = Qjk Qil + Qjl Qik . (2.2.2)

Now, it follows from the Lemma 2.1.4 that α̃ : C(S 2 ) ⊗ QISOL → C(S 2 ) ⊗ QISOL
defined by α̃(X ⊗ Y ) = α(X)(1 ⊗ Y ) extends to a unitary of the Hilbert QISOL -module
L2 (S 2 ) ⊗ QISOL (or in other words, α extends to a unitary representation of QISOL
on L2 (S 2 )). But α keeps V = Sp{x1 , x2 , x3 } invariant. So α is a unitary representation
of QISOL on V , that is Q = ((Qij )) belonging to M3 (QISOL ) is a unitary, hence
Q−1 = Q∗ = QT , since in this case entries of Q are self-adjoint elements.
Clearly, the matrix Q is a 3-dimensional unitary representation of QISOL . From (
4 ) of Proposition 1.2.23, the antipode κ on the matrix elements of a finite-dimensional
unitary representation U α ≡ (uαpq ) is given by κ(uαpq ) = (uαqp )∗ .
So we obtain
κ(Qij ) = Q−1 T
ij = Qij = Qji . (2.2.3)

Now from ( 2.2.1 ) , we have Qij Qkj = Qkj Qij . Applying κ on this equation and using
the fact that κ is an antihomomorphism along with ( 2.2.3 ) , we have Qjk Qji = Qji Qjk
Similarly , applying κ on ( 2.2.2 ), we get

Qlj Qki + Qkj Qli = Qli Qkj + Qki Qlj ∀i, j, k, l.

Interchanging between k and i and also between l, j gives

Qjl Qik + Qil Qjk = Qjk Qil + Qik Qjl ∀i, j, k, l. (2.2.4)
57 Computation of QISOL

Now, by (2.2.2 ) - ( 2.2.4 ), we have

[Qik , Qjl ] = [Qjl , Qik ],

hence
[Qik , Qjl ] = 0.

Therefore the entries of the matrix Q commute among themselves. However, by

faithfulness of the action of QISOL , it is clear that the C ∗ -subalgebra generated by
entries of Q (which forms a quantum subgroup of QISOL acting on C(S 2 ) isometrically)
must be the same as QISOL , so QISOL is commutative.
So QISOL = C(G) for some compact group G acting by isometry on C(S 2 ) and G is
clearly universal in the category of compact metrizable groups acting on S 2 isometrically,
so that G ∼= O(3). 2

Remark 2.2.3. Similarly, it can be shown that QISO(S n ) is commutative for all n ≥ 2.

2.2.2 The commutative one-torus

Let C = C(S 1 ) be the C ∗ -algebra of continuous functions on the one-torus S 1 . Let
us denote by z and z the identity function (which is the generator of C(S 1 )) and its
conjugate respectively. The Laplacian coming from the standard Riemannian metric
is given by L(z n ) = −n2 z n , for n in Z, hence the eigenspace corresponding to the
eigenvalue −1 is spanned by z and z. Thus, the action of a compact quantum group
acting smoothly and isometrically (and faithfully) on C(S 1 ) must be linear in the sense
that its action must map z into an element of the form z ⊗ A + z ⊗ B. However, we show
below that this forces the quantum group to be commutative as a C ∗ algebra, that is it
must be the function algebra of some compact group .
Theorem 2.2.4. Let α be a faithful, smooth and linear action of a compact quantum
group (Q, ∆) on C(S 1 ) defined by α(z) = z ⊗ A + z ⊗ B. Then Q is a commutative C ∗
algebra.
Proof : By the assumption of faithfulness, it is clear that Q is generated (as a unital
C∗ algebra) by A and B. Moreover, recall that smoothness in particular means that
A and B must belong to the algebra Q0 spanned by matrix elements of irreducible
representations of Q . Since zz = zz = 1 and α is a ∗-homomorphism, we have
α(z)α(z) = α(z)α(z) = 1 ⊗ 1.
Comparing coefficients of z 2 , z 2 and 1 in both hand sides of the relation α(z)α(z) =
1 ⊗ 1, we get

AB ∗ = BA∗ = 0, AA∗ + BB ∗ = 1. (2.2.5)

Chapter 2: Quantum isometry groups: approach based on Laplacian 58

Similarly, α(z)α(z) = 1 ⊗ 1 gives

B ∗ A = A∗ B = 0, A∗ A + B ∗ B = 1. (2.2.6)

Let U = A + B , P = A∗ A , Q = AA∗ . Then it follows from (2.2.5) and (2.2.6) that

U is a unitary and P is a projection since P is self adjoint and

P 2 = A∗ AA∗ A = A∗ A(1 − B ∗ B) = A∗ A − A∗ AB ∗ B = A∗ A = P.

Moreover ,

UP
= (A + B)A∗ A = AA∗ A + BA∗ A = AA∗ A
( since BA∗ = 0 from (2.2.5))
= A(1 − B ∗ B) = A − AB ∗ B = A.

Thus, A = U P , B = U − U P = U (1 − P ) ≡ U P ⊥ , so Q = C ∗ (A, B) = C ∗ (U, P ).

We can rewrite the action α as follows:

α(z) = z ⊗ U P + z ⊗ U P ⊥ .

The coproduct ∆ can easily be calculated from the requirement (id ⊗ ∆)α = (α ⊗ id)α
, and it is given by :

∆(U P ) = U P ⊗ U P + P ⊥ U −1 ⊗ U P ⊥ , (2.2.7)

∆(U P ⊥ ) = U P ⊥ ⊗ U P + P U −1 ⊗ U P ⊥ . (2.2.8)

From this, we get

∆(U ) = U ⊗ U P + U −1 ⊗ U P ⊥ , (2.2.9)

∆(P ) = ∆(U −1 )∆(U P ) = P ⊗ P + U P ⊥ U −1 ⊗ P ⊥ . (2.2.10)

It can be checked that ∆ given by the above expression is coassociative.

Let h denote the right-invariant Haar state on Q. By the general theory of compact
quantum groups, h must be faithful on Q0 . We have (by right-invariance of h):

(id ⊗ h)(P ⊗ P + U P ⊥ U −1 ⊗ P ⊥ ) = h(P )1.

That is, we have

h(P ⊥ )U P ⊥ U −1 = h(P )P ⊥ . (2.2.11)
59 Computation of QISOL

Since P is a positive element in Q0 and h is faithful on Q0 , h(P ) = 0 if and only if

P = 0. Similarly , h(P ⊥ ) = 0, that is h(P ) = 1, if and only if P = 1. However, if P
is either 0 or 1, clearly Q = C ∗ (U, P ) = C ∗ (U ), which is commutative. On the other
hand, if we assume that P is not a trivial projection, then h(P ) is strictly between 0
and 1, and we have from ( 2.2.11 )

h(P )
U P ⊥ U −1 = P ⊥.
1 − h(P )

Since both U P ⊥ U −1 and P ⊥ are nontrivial projections, they can be scalar multiples
of each other if and only if they are equal, so we conclude that U P ⊥ U −1 = P ⊥ , that is
U commutes with P ⊥ , hence with P , and Q is commutative. 2

2.2.3 The commutative n-tori

Consider C(Tn ) as the universal commutative C ∗ algebra generated by n commut-
ing unitaries U1 , U2 , .....Un . It is clear that the set {Uim Ujn : m, n ∈ Z} is an
orthonormal basis for L2 (C(Tn ), τ0 ), where τ0 denotes the unique faithful normal-
ized trace on C(Tn ) given by, τ0 ( amn Uim Ujn ) = a00 which is just the integration
P

against the Haar measure. We shall denote by hA, Bi = τ0 (A∗ B) the inner prod-
uct on H0 := L2 (C(Tn ), τ0 ). Let C(Tn )fin be the unital ∗-subalgebra generated by
finite complex linear combinations of U m V n , m, n ∈ Z. The Laplacian L is given by
L(U1m1 ......Unmn ) = −(m21 + ...m2n )U1m1 ......Unmn , and it is also easy to see that the al-
gebraic span of eigenvectors of L is nothing but the space C(Tn )fin , and moreover, all
the assumptions 1. - 6. in subsection 2.1.2 required for defining the quantum isometry
group are satisfied.
Let QISOL be the quantum isometry group coming from the above Laplacian, with
the smooth isometric action of QISOL on C(Tn ) given by α : C(Tn ) → C(Tn )⊗QISOL .
By definition, α must keep invariant the eigenspace of L corresponding to the eigenvalue
−1, spanned by U1 , .....Un , U1−1 , ......., Un−1 . Thus, the action α is given by:
n
X n
X
α(Ui ) = Uj ⊗ Aij + Uj−1 ⊗ Bij ,
j=1 j=1

where Aij , Bij are in QISOL , i, j = 1, 2....n. By faithfulness of the action of quantum
isometry group, the norm-closure of the unital ∗-algebra generated by {Aij , Bij : i, j =
1, 2, ....n} must be the whole of QISOL .
Next we derive a number of conditions on Aij , Bij , i, j = 1, 2, ...n using the fact that
α is a ∗ homomorphism.
Chapter 2: Quantum isometry groups: approach based on Laplacian 60

Lemma 2.2.5. The condition U ∗ U = 1 = U U ∗ gives:

X
(A∗ij Aij + Bij
∗
Bij ) = 1, (2.2.12)
j

∗ ∗
Bij Aik + Bik Aij = 0 ∀j 6= k, (2.2.13)

A∗ij Bik + A∗ik Bij = 0 ∀j 6= k, (2.2.14)

A∗ij Bij = Bij

∗
Aij = 0, (2.2.15)
X
(Aij A∗ij + Bij Bij
∗
) = 1, (2.2.16)
j
∗ ∗
Aik Bij + Aij Bik = 0 ∀j 6= k, (2.2.17)

Bik A∗ij + Bij A∗ik = 0 ∀j 6= k, (2.2.18)

∗
Aij Bij = Bij A∗ij = 0. (2.2.19)

Proof : We get ( 2.2.12 ) - ( 2.2.15 ) by using the condition Ui∗ Ui = 1 along with the
fact that α is a homomorphism and then comparing the coefficients of 1, Uj Uk , Uj−1 Uk−1 ,
( for j 6= k), Uj−2 , Uk−2 .
Similarly the condition Ui Ui∗ = 1 gives ( 2.2.16 ) - ( 2.2.19 ). 2

Now, for all i 6= j, Ui∗ Uj , Ui Uj∗ and Ui Uj belong to the eigenspace of the Laplacian
with eigenvalue −2, while Uk2 , Uk−2 belong to the eigenspace corresponding to the eigen-
value −4. As α preserves the eigenspaces of the Laplacian, the coefficients of Uk2 , Uk−2
are zero for all k in α(Ui∗ Uj ), α(Ui Uj∗ ), α(Ui Uj ) for all i 6= j.
We use this observation in the next lemma.

Lemma 2.2.6. For all k and for all i 6= j,

∗
Bik Ajk = A∗ik Bjk = 0, (2.2.20)

Aik Bjk = Bik A∗jk = 0, (2.2.21)

Aik Ajk = Bik Bjk = 0. (2.2.22)

Proof : The equation ( 2.2.20 ) is obtained from the coefficients of Uk2 and Uk−2
in α(Ui∗ Uj ) while ( 2.2.21 ) and ( 2.2.22 ) are obtained from the same coefficients in
α(Ui Uj∗ ) and α(Ui Uj ) respectively. 2

Now, by Lemma 2.1.4 it follows that α̃ : C(Tn ) ⊗ QISOL → C(Tn ) ⊗ QISOL

defined by α̃(X ⊗ Y ) = α(X)(1 ⊗ Y ) extends to a unitary of the Hilbert QISOL -module
L2 (C(Tn ), τ ) ⊗ QISOL ( or in other words, α extends to a unitary representation of
61 Computation of QISOL

QISOL on L2 (C(Tn ), τ )). But α keeps W = Sp{Ui , Ui∗ : 1 ≤ i ≤ n} invariant. So α is

a unitary representation of QISOL on W . Hence, the matrix ( say M ) corresponding
to the 2n dimensional representation of QISOL on W is a unitary in ! M2n (QISOL ).
Aij Bij∗
From the definition of the action it follows that M = .
Bij A∗ij
Since M is the matrix corresponding to a finite dimensional unitary
! representation,
∗ ∗
Aji Bji
using ( 4 ) of Proposition 1.2.23, we have (k(Mkl )) = ( κ is the antipode
Bji Aji
of QISOL ).

Now we apply the antipode κ to get some more relations.

Lemma 2.2.7. :
For all k and i 6= j,

∗
A∗kj A∗ki = Bkj Bki = A∗kj Bki = Bkj Aki = Bkj A∗ki = Akj Bki = 0. (2.2.23)

Proof : The result follows by applying κ on the equations Aik Ajk = Bik Bjk =
∗ A
Bik jk = A∗ik Bjk = Aik Bjk = Bik A∗jk = 0 obtained from Lemma 2.2.6. 2

Lemma 2.2.8. :
Ali is a normal partial isometry for all l, i and hence has same domain and range.

Proof : From the relation ( 2.2.12 ) in Lemma 2.2.5, we have by applying κ,

P ∗ ∗ ) = 1. Applying A on the right of this equation, we have
(Aji Aji + Bji Bji li
Ali Ali Ali + j6=l (Aji Aji Ali + Bli Bli∗ Ali ) + j6=l Bji Bji
∗ ∗ ∗A =A .
P P
li li
∗
From Lemma 2.2.6, we have Aji Ali = 0 and Bji Ali = 0 for all j 6= l. Moreover, from
Lemma 2.2.5, we have Bli∗ Ali = 0. Applying these to the above equation, we have

A∗li Ali Ali = Ali . (2.2.24)

Again, from the relation j (Aij A∗ij + Bij Bij ∗ ) = 1 for all i in Lemma 2.2.5, applying κ
P

and multiplying by A∗li on the right, we have Ali A∗li A∗li + j6=l Aji A∗ji A∗li + Bli∗ Bli A∗li +
P
∗ ∗ ∗
P
j6=l Bji Bji Ali = Ali . From Lemma 2.2.6, we have Ali Aji = 0 for all j 6= l ( hence
A∗ji A∗li = 0 ) and Bji A∗li = 0. Moreover, we have Bli A∗li = 0 from Lemma 2.2.5. Hence,
we have
Ali A∗li A∗li = A∗li . (2.2.25)

From (2.2.24), we have

(A∗li Ali )(Ali A∗li ) = Ali A∗li . (2.2.26)
Chapter 2: Quantum isometry groups: approach based on Laplacian 62

By taking ∗ on (2.2.25), we have

Ali Ali A∗li = Ali . (2.2.27)

Using this in (2.2.26), we have

Ali A∗li Ali = Ali A∗li , (2.2.28)

and hence Ali is normal.

So, Ali = A∗li Ali Ali ( from ( 2.2.24 ) ) = Ali A∗li Ali .
Therefore, Ali is a partial isometry which is normal and hence has same domain and
range. 2

Lemma 2.2.9. :
Bli is a normal partial isometry and hence has same domain and range.

Proof : First we note that Aji is a normal partial isometry and Aji Bli = 0 for
all j 6= l ( obtained from Lemma 2.2.6 ) implies that Ran(A∗ji ) ⊆ Ker(Bli∗ ) and hence
Ran(Aji ) ⊆ Ker(Bli∗ ) which means Bli∗ Aji = 0 for all j 6= l.
To obtain Bli∗ Bli Bli = Bli , we apply κ and multiply by Bli on the right of ( 2.2.16
) and then use A∗li Bli = 0 from Lemma 2.2.5, Aji Bli = 0 for all j 6= l ( from Lemma
2.2.6 which implies Bli∗ Aji = 0 for all j 6= l as above ) and Bji Bli = 0 for all j 6= l from
Lemma 2.2.6 .
Similarly, we have Bli Bli∗ Bli∗ = Bli∗ by applying κ and multiplying by Bli∗ on the right
of ( 2.2.12 ) obtained from Lemma 2.2.5 and then use Ali Bli∗ = 0 ( Lemma 2.2.5 ),
Bli A∗ji = 0 for all j 6= l and Bli Bji = 0 for all j 6= l ( Lemma 2.2.6 ).
Using Bli∗ Bli Bli = Bli and Bli Bli∗ Bli∗ = Bli∗ as in Lemma 2.2.8, we have Bli is a
normal partial isometry. 2

Now, we use the condition α(Ui )α(Uj ) = α(Uj )α(Ui ) for all i, j.

Lemma 2.2.10. :
For all k 6= l,
Aik Ajl + Ail Ajk = Ajl Aik + Ajk Ail , (2.2.29)

Aik Bjl + Bil Ajk = Bjl Aik + Ajk Bil , (2.2.30)

Bik Ajl + Ail Bjk = Ajl Bik + Bjk Ail , (2.2.31)

Bik Bjl + Bil Bjk = Bjl Bik + Bjk Bil . (2.2.32)

Proof : The result follows by equating the coefficients of Uk Ul , Uk Ul−1 , Uk−1 Ul and
Uk−1 Ul−1 ( where k 6= l ) in α(Ui )α(Uj ) = α(Uj )α(Ui ) for all i, j.
63 Computation of QISOL

Lemma 2.2.11. :
Aik Bjl = Bjl Aik for all i 6= j, k 6= l.

Proof : From Lemma 2.2.10, we have for all k 6= l, Aik Bjl −Bjl Aik = Ajk Bil −Bil Ajk .
We consider the case where i 6= j.
We have, Ran(Aik Bjl − Bjl Aik ) ⊆ Ran(Aik ) + Ran(Bjl ) ⊆ Ran(Bjl ∗ B + A∗ A ) (
jl ik ik
using the facts that Aik and Bjl are normal partial isometries by Lemma 2.2.8 and 2.2.9
and also that Bjl∗ B and A∗ A are projections ).
jl ik ik
Similarly, Ran(Ajk Bil − Bil Ajk ) ⊆ Ran(Bil∗ Bil + A∗jk Ajk ).
Let
T1 = Aik Bjl − Bjl Aik , (2.2.33)

T2 = Ajk Bil − Bil Ajk , (2.2.34)

∗
T3 = Bjl Bjl + A∗ik Aik , (2.2.35)

T4 = Bil∗ Bil + A∗jk Ajk . (2.2.36)

Hence, T1 = T2 , RanT1 ⊆ RanT3 , RanT2 ⊆ RanT4 .

We claim that T4 T3 = 0.
Then Ran(T3 ) ⊆ Ker(T4 ).
But RanT1 ⊆ RanT3 will imply that RanT1 ⊆ KerT4 . Hence, Ran(T2 ) ⊆ Ker(T4 ) =
⊥ ⊥
Ran(T4∗ ) = Ran(T4 ) But Ran(T2 ) ⊆ Ran(T4 ) which implies that Ran(T2 ) = 0 and
hence both T2 and T1 are zero. Thus, the proof of the lemma will be complete if we can
prove the claim.
T4 T3 = (Bil∗ Bil + A∗jk Ajk )(Bjl
∗
Bjl + A∗ik Aik ) (2.2.37)

= Bil∗ Bil Bjl

∗
Bjl + Bil∗ Bil A∗ik Aik + A∗jk Ajk Bjl
∗
Bjl + A∗jk Ajk A∗ik Aik . (2.2.38)

From Lemma 2.2.6, we have for all i 6= j, Bil Bjl = 0 implying Bil Bjl∗ = 0 as B is a
jl
normal partial isometry.
Again, from Lemma 2.2.7 for all k 6= l, Bil Aik = 0. Then Aik is a normal partial
isometry implies that Bil A∗ik = 0 for all k 6= l.
Similarly, by taking adjoint of the relation Bjl A∗jk = 0 for all k 6= l obtained from
Lemma 2.2.7, we have Ajk Bjl ∗ = 0.

From Lemma 2.2.6, we have Ajk Aik = 0 for all i 6= j. Aik is a normal partial isometry
implies that Ajk A∗ik = 0 for all i 6= j.
Using these, we note that T4 T3 = 0 which proves the claim and hence the lemma. 2

Lemma 2.2.12. :
Chapter 2: Quantum isometry groups: approach based on Laplacian 64

Aik Bjk = 0 = Bjk Aik , (2.2.39)

Aki Bkj = 0 = Bkj Aki (2.2.40)

for all i 6= j and for all k.

Proof : By Lemma 2.2.6, we have Aik Bjk = 0 and Bjk A∗ik = 0 for all i 6= j. The
second relation along with the fact that Aik is a normal partial isometry implies that
Bjk Aik = 0 for all i 6= j.
Thus, Aik Bjk = 0 = Bjk Aik for all i 6= j.
Applying κ on the above equation and using Bkj and Aki are normal partial isome-
tries, we have Aki Bkj = 0 = Bkj Aki .
2

Lemma 2.2.13. : Aik Bik = Bik Aik for all i, k.

Proof : We have A∗ij Bij = 0 = Bij

∗A
ij from Lemma 2.2.5. Using the fact that
Bij and Aij are normal partial isometry we have A∗ij Bij
∗ = 0 = B ∗ A∗ and hence
ij ij
Aij Bij = Bij Aij . 2

Lemma 2.2.14. :
Aik Ajl = Ajl Aik for all i 6= j, k 6= l.

Proof : Using ( 2.2.29 ) in Lemma 2.2.10, we proceed as in Lemma 2.2.11 to get

Ran(Aik Ajl −Ajl Aik ) ⊆ Ran(Ajl A∗jl +Aik A∗ik ) and Ran(Ajk Ail −Ail Ajk ) ⊆ Ran(Ail A∗il +
Ajk A∗jk ).
We claim that (Aik A∗ik + Ajl A∗jl )(Ajk A∗jk + Ail A∗il ) = 0.
Then by the same reasonings as given in Lemma 2.2.11 we will have : Ajk Ail =
Ail Ajk .
To prove the claim, we use Aik Ajk = 0 for all i 6= j from Lemma 2.2.6 ( which implies
Ajk Aik = 0 for all i 6= j as Aik is a normal partial isometry ), A∗il A∗ik = 0 for all k 6= l
∗

from Lemma 2.2.7 ( which implies A∗il Aik = 0 for all k 6= l as Aik is a normal partial
isometry ) and Ail Ajl = 0 for all i 6= j from Lemma 2.2.6 ( which implies A∗jl Ail = 0 for
all i 6= j as A∗il is a normal partial isometry ).
2

Lemma 2.2.15. :

Aik Ail = Ail Aik for all k 6= l, (2.2.41)

Aik Ajk = Ajk Aik for all i 6= j. (2.2.42)

65 Computation of QISOL

Proof : From Lemma 2.2.6, we have Aki Ali = 0 for all k 6= l.

Applying κ and taking adjoint, we have Aik Ail = 0 for all k 6= l. Interchanging k
and l, we get Ail Aik = 0 for all k 6= l. Hence, Aik Ail = Ail Aik for all k 6= l.
From Lemma 2.2.6, we have Aik Ajk = 0 for all i 6= j. Interchanging i and j, we have
Ajk Aik = 0 for all i 6= j. 2

Remark 2.2.16. Proceeding in an exactly similar way, we have that Bij ’s commute
among themselves.

Theorem 2.2.17. The Quantum isometry group of C(Tn ) is commutative as a C ∗

algebra and hence coincides with the classical isometry group C(Tn >
(Zn2 >
Sn )).

Proof : Follows from the results in lemma 2.2.11 - 2.2.15 and the remark following
them. 2
Chapter 2: Quantum isometry groups: approach based on Laplacian 66
Chapter 3

Quantum group of orientation

preserving Riemannian isometries

3.1 Introduction
The formulation of quantum isometry groups in [30] had a major drawback from the
viewpoint of noncommutative geometry since it needed a ‘good’ Laplacian to exist. In
noncommutative geometry it is not always easy to verify such an assumption about the
Laplacian, and thus it would be more appropriate to have a formulation in terms of the
Dirac operator directly. This is what we aim to achieve in the present chapter.
The group of Riemannian isometries of a compact Riemannian manifold M can
be viewed as the universal object in the category of all compact metrizable groups
acting on M , with smooth and isometric action. Moreover, let us assume that the
manifold has a spin structure (hence in particular orientable, so we can fix a choice of
orientation) and D denotes the conventional Dirac operator acting as an unbounded
self-adjoint operator on the Hilbert space H of square integrable spinors. Then, it can
be proved that a group action on the manifold lifts as a unitary representation on the
Hilbert space H which commutes with D if and only if the action on the manifold is an
orientation preserving isometric action. Therefore, to define the quantum analogue of
the group of orientation-preserving Riemannian isometries of a possibly noncommutative
manifold given by a spectral triple (A∞ , H, D), it is reasonable to consider a category
Q0 of compact quantum groups having unitary (co-) representation, say U , on H, which
commutes with D, and the action on B(H) obtained by conjugation maps A∞ into its
weak closure. A universal object in this category, if it exists, should define the ‘quantum
group of orientation preserving Riemannian isometries’ of the underlying spectral triple.
It is easy to see that any object (S, U ) of the category Q0 gives an equivariant spectral
triple (A∞ , H, D) with respect to the action of S implemented by U . It may be noted

67
Chapter 3: Quantum group of orientation preserving Riemannian isometries 68

that recently there has been a lot of interest and work (see, for example, [13], [18], [23])
towards construction of quantum group equivariant spectral triples. In all these works,
given a C ∗ -subalgebra A of B(H) and a CQG Q having a unitary representation U on H
such that αU (≡ adU ) gives an action of Q on A, the authors investigate the possibility
of constructing a (nontrivial) spectral triple (A∞ , H, D) on a suitable dense subalgebra
A∞ of A such that U commutes with D ⊗ I, that is, D is equivariant. Our interest here
is in the (sort of) converse direction: given a spectral triple, we want to consider all
possible CQG representations with respect to which the spectral triple is equivariant;
and if there exists a universal object in the corresponding category, that is, Q0 , we
should call it the quantum group of orientation preserving isometries.
Unfortunately, even in the finite-dimensional (but with noncommutative A) situation
this category may often fail to have a universal object, as will be discussed later. It
turns out, however, that if we fix a suitable densely defined ( in the WOT ) functional
on B(H) (to be interpreted as the choice of a ‘volume form’) then there exists a universal
object in the subcategory of Q0 obtained by restricting the object-class to the quantum
group actions which also preserve the given functional. The subtle point to note here is
that unlike the classical group actions on B(H) which always preserve the usual trace,
a quantum group action may not do so. In fact, it was proved by Goswami in [31] that
given an object (Q, U ) of Q0 (where Q is a compact quantum group and U denotes its
unitary co-representation on H), we can find a positive invertible operator R in H so
that the given spectral triple is R-twisted in the sense of [31] and the corresponding
functional τR (which typically differs from the usual trace of B(H) and can have a
nontrivial modularity) is preserved by the action of Q. This makes it quite natural to
work in the setting of twisted spectral data (as defined in [31]).
Motivated by the ideas of Woronowicz and Soltan ( [68], [53] ), we actually consider a
bigger category called the category of ‘quantum families of smooth orientation preserving
Riemannian isometries’. The underlying C ∗ -algebra of the quantum group of orientation
preserving isometries (whenever exists) has been identified with the universal object in
this bigger category and moreover, it is shown to be equipped with a canonical coproduct
making it into a compact quantum group.
In this chapter, we discuss a number of examples, covering the classical spectral triple
on C ∞ (T2 ) as well as the equivariant spectral triples constructed recently on SUµ (2). It
may be relevant to point out here that it was not clear whether one could accommodate
the spectral triples on SUµ (2) and the Podles’ spheres Sµ,c 2 in the framework of [30],

since it is very difficult to give a nice description of the space of ‘noncommutative’ forms
and the Laplacian for these examples. However, the present formulation in terms of the
Dirac operator makes it easy to accommodate them, and we have been able to identify
Uµ (2) and SOµ (3) as the universal quantum group of orientation preserving isometries
69 Definition and existence of the quantum group of orientation-preserving isometries

2 respectively (the computations for S 2 have

for the spectral triples on SUµ (2) and Sµ,c µc
been presented in Chapter 5).
We conclude this section with an important remark about the use of the phrase
‘orientation -preserving’ in our terminology. We recall from Remark 1.5.5 that by a
‘classical spectral triple’ we always mean the spectral triple obtained by the Dirac op-
erator on the spinors. This is absolutely crucial in view of the fact that the Hodge
Dirac operator d + d∗ on the L2 -space of differential forms also gives a spectral triple
of compact type on any compact Riemannian (not necessarily with a spin structure)
manifold M , but the action of the full isometry group ISO(M ) (and not just the sub-
group of orientation preserving isometries ISO+ (M ), even when M is orientable) lifts
to a canonical unitary representation on this space commuting with d + d∗ . In fact, the
category of groups acting on M such that the action comes from a unitary representa-
tion commuting with d + d∗ , has ISO(M ), and not ISO+ (M ), as its universal object.
So, one must stick to the Dirac operator on spinors to obtain the group of orientation
preserving isometries in the usual geometric sense. This also has a natural quantum
generalization, as we shall see in section 3.3.

3.2 Definition and existence of the quantum group of

orientation-preserving isometries
3.2.1 The classical case
We first discuss the classical situation clearly, which will serve as a motivation for our
quantum formulation.
We begin with a few basic facts about topologizing the space C ∞ (M, N ) where M, N
are smooth manifolds. Let Ω be an open set of Rn . We endow C ∞ (Ω) with the usual
Frechet topology coming from uniform convergence (over compact subsets) of partial
derivatives of all orders. The space C ∞ (Ω) is complete with respect to this topology,
so is a Polish space in particular. Moreover, by the Sobolev imbedding theorem (
Corollary 1.21, [48] ), ∩k≥0 Hk (Ω) = C ∞ (Ω) as a set, where Hk (Ω) denotes the k-
th Sobolev space. Thus, C ∞ (Ω) has also the Hilbertian seminorms coming from the
Sobolev spaces, hence the corresponding Frechet topology. We claim that these two
topologies on C ∞ (Ω) coincide. Indeed, the inclusion map from C ∞ (Ω) into ∩k Hk (Ω),
is continuous and surjective, so by the open mapping theorem for Frechet space, the
inverse is also continuous, proving our claim.
Given two second countable smooth manifolds M, N , we shall equip C ∞ (M, N ) with
the weakest locally convex topology making C ∞ (M, N ) 3 φ 7→ f ◦ φ ∈ C ∞ (M ) Frechet
continuous for every f in C ∞ (N ).
Chapter 3: Quantum group of orientation preserving Riemannian isometries 70

For topological or smooth fibre or principal bundles E, F over a second countable

smooth manifold M , we shall denote by Hom(E, F ) the set of bundle morphisms from
E to F . We remark that the total space of a locally trivial topological bundle such that
the base and the fibre spaces are locally compact Hausdorff second countable must itself
be so, hence in particular Polish ( that is, a complete separable metric space ).
In particular, if E, F are locally trivial principal G-bundles over a common base, such
that the (common) base as well as structure group G are locally compact Hausdorff and
second countable, then Hom(E, F ) is a Polish space.
We need a standard fact, stated below as Lemma 3.2.2, about the measurable lift of
Polish space valued functions.
Before that, we introduce some notions.
A multifunction G : X → Y is a map with domain X and whose values are nonempty
subsets of Y. For A ⊆ Y, we put G−1 (A) = {x ∈ X : G(x) ∩ A 6= φ}.
A selection of a multifunction G : X → Y is a point map s : X → Y such that s(x)
belongs to G(x) for all x in X. Now let Y be a Polish space and σX a σ-algebra on X.
A multifunction G : X → Y is called σX measurable if G−1 (U ) belongs to σX for every
open set U in Y.
The following well known selection theorem is Theorem 5.2.1 of [55] and was proved
by Kuratowski and Ryll-Nardzewski.

Proposition 3.2.1. Let σX be a σ algebra on X and Y a Polish space. Then, every σX

measurable, closed valued multifunction F : X → Y admits a σX measurable selection.

A trivial consequence of this result is the following:

Lemma 3.2.2. Let M be a compact metrizable space, B, B̃ Polish spaces such that
there is an n-covering map Λ : B̃ → B. Then any continuous map ξ : M → B admits
a lifting ξ˜ : M → B̃ which is Borel measurable and Λ ◦ ξ˜ = ξ. In particular, if B̃ and B
are topological bundles over M , with Λ being a bundle map, any continuous section of
B admits a lifting which is a measurable section of B̃.

We shall now give an operator-theoretic characterization of the classical group of

orientation-preserving Riemannian isometries, which will be the motivation of our def-
inition of its quantum counterpart. Let M be a compact Riemannian n dimensional
spin manifold, with a fixed choice of orientation. We recall the notations as in subsec-
tion 1.4.3. In particular, the spinor bundle S is the associated bundle of a principal
Spin(n)-bundle, say P , on M which has a canonical 2-covering bundle-map Λ from
P to the frame-bundle F (which is an SO(n)-principal bundle), such that locally Λ
is of the form (idM ⊗ λ) where λ is the two covering map from Spin(n) to SO(n).
Moreover, the spinor space will be denoted by ∆n . Let f be a smooth orientation pre-
serving Riemannian isometry of M, and consider the bundles E = Hom(F, f ∗ (F )) and
71 Definition and existence of the quantum group of orientation-preserving isometries

Ẽ = Hom(P, f ∗ (P )) (where Hom denotes the set of bundle maps). We view df as a

section of the bundle E in the natural way. By the Lemma 3.2.2 we obtain a measurable
˜ : M → Ẽ, which is a measurable section of Ẽ. Using this, we define a map on
lift df
the space of measurable section of S = P ×Spin(n) ∆n as follows: given a (measurable)
section ξ of S, say of the form ξ(m) = [p(m), v], with p(m) in Pm , v in ∆n , we define
U ξ by (U ξ)(m) = [df ˜ (f −1 (m))(p(f −1 (m))), v]. Note that sections of the above form
constitute a total subset in L2 (S), and the map ξ 7→ U ξ is clearly a densely defined
linear map on L2 (S), whose fibre-wise action is unitary since the Spin(n) action is so
on ∆n . Thus it extends to a unitary U on H = L2 (S). Any such U , induced by the
map f , will be denoted by Uf ( it is not unique since the choice of the lifting used in
its construction is not unique ).
Theorem 3.2.3. Let M be a compact Riemannian spin manifold (hence orientable ,
and fix a choice of orientation) with the usual Dirac operator D acting as an unbounded
self-adjoint operator on the Hilbert space H of the square integrable spinors, and let S
denote the spinor bundle, with Γ(S) being the C ∞ (M ) module of smooth sections of
S. Let f : M → M be a smooth one-to-one map which is a Riemannian orientation
preserving isometry. Then the unitary Uf on H commutes with D and Uf Mφ Uf∗ = Mφ◦f ,
for any φ in C(M ), where Mφ denotes the operator of multiplication by φ on L2 (S).
Moreover, when the dimension of M is even, Uf commutes with the canonical grading
γ on L2 (S).
Conversely, suppose that U is a unitary on H such that U D = DU and the map
αU (X) = U XU −1 for X in B(H) maps A = C(M ) into L∞ (M ) = A00 , then there
is a smooth one-to-one orientation-preserving Riemannian isometry f on M such that
U = Uf . We have the same result in the even case, if we assume furthermore that
U γ = γU.
Proof: From the construction of Uf , it is clear that Uf Mφ Uf−1 = Mφ◦f . Moreover,
since the Dirac operator D commutes with the Spin(n)-action on S, we have Uf D =
DUf on each fibre, hence on L2 (S). In the even dimensional case, it is easy to see that
the Spin(n) action commutes with γ ( the grading operator ), hence Uf does so.
For the converse, first note that αU is a unital ∗-homomorphism on L∞ (M, dvol)
and thus must be of the form ψ 7→ ψ ◦ f for some measurable f . We claim that
f must be smooth. Fix any smooth g on M and consider φ = g ◦ f . We have to
argue that φ is smooth. Let δD denote the generator of the strongly continuous one-
parameter group of automorphism βt (X) = eitD Xe−itD on B(H) (with respect to the
weak operator topology, say). From the assumption that D and U commute it is clear
n ) into itself and since C ∞ (M ) ⊂ D, we conclude that
T
that αU maps D := n≥1 Dom(δD
αU (Mφ ) = Mφ◦g belongs to D. We claim that this implies the smoothness of φ. Let m
belongs to M and choose a local chart (V, ψ) at m, with the coordinates (x1 , ..., xn ), such
Chapter 3: Quantum group of orientation preserving Riemannian isometries 72

that Ω = ψ(V ) ⊆ Rn has compact closure, S|V is trivial and D has the local expression
D = i nj=1 µ(ej )∇j , where ∇j = ∇ ∂ denotes the covariant derivative (with respect to
P
∂xj

the canonical Levi Civita connection) operator along the vector field ∂x∂ j on L2 (Ω) and
µ(v) denotes the Clifford multiplication by a vector v. Now, φ ◦ ψ −1 ∈ L∞ (Ω) ⊆ L2 (Ω)
and it is easy to observe from the above local structure of D that [D, Mφ ] has the local
n ) implies φ ◦ ψ −1 is
P T
expression j iM ∂ φ ⊗ µ(ej ). Thus, the fact Mφ ∈ n≥1 Dom(δD
∂xj

in Dom(dj1 ...djk ) for every integer tuples (j1 , ..., jk ), ji ∈ {1, ..., n}, where dj := ∂x∂ j . In
other words, φ ◦ ψ −1 is in H k (Ω) for all k ≥ 1, where H k (Ω) denotes the k-th Sobolev
space on Ω (see [48]). By Sobolev’s Theorem (see, for example. [48],Corollary 1.21, page
24) it follows that φ ◦ ψ −1 is in C ∞ (Ω).
We note that f is one-to-one as φ → φ ◦ f is an automorphism of L∞ . Now, we shall
show that f is an isometry of the metric space (M, d), where d is the metric coming
from the Riemannian structure, and we have the explicit formula ( 1.4.1 )

d(p, q) = supφ∈C ∞ (M ),k[D,Mφ ]k≤1 |φ(p) − φ(q)|.

Since U commutes with D, we have k[D, Mφ◦f ]k = k[D, U Mφ U ∗ ]k = kU [D, Mφ ]U ∗ k =

k[D, Mφ ]k for every φ, from which it follows that d(f (p), f (q)) = d(p, q). Finally, f is
orientation preserving if and only if the volume form (say ω), which defines the choice
of orientation, is preserved by the natural action of df , that is, (df ∧ .... ∧ df )(ω) = ω.
This will follow from the explicit description of ω in terms of D, given by (see [65] Page
26, also see [20] )

ω(φ0 dφ1 ...dφn ) = τ (Mφ0 [D, Mφ1 ]...[D, Mφn ]),

where φ0 , ..., φn belong to C ∞ (M ),  = 1 in the odd case and  = γ ( the

grading operator ) in the even case and τ denotes the volume integral. In fact,
−tD 2
τ (X) = Limt→0+ Tr(e −tD2X) ( where Lim is as in subsection 1.5.2 ), which implies
Tr(e )
τ (U XU ∗ ) = τ (X) for all X in B(H) (using the fact that D and U commute). Thus,

ω(φ0 ◦ f d(φ1 ◦ f ) . . . d(φn ◦ f ))

= τ (U Mφ0 U ∗ U [D, Mφ1 ]U ∗ ...U [D, Mφn ]U ∗ )

= τ (U Mφ0 [D, Mφ1 ]...[D, Mφn ]U ∗ )

= τ (Mφ0 [D, Mφ1 ]...[D, Mφn ])

= ω(φ0 dφ1 ...dφn ).

2
73 Definition and existence of the quantum group of orientation-preserving isometries

Now we turn to the case of a family of maps. We are ready to state and prove the
operator-theoretic characterization of ‘set of orientation-preserving isometries’.

Theorem 3.2.4. Let X be a compact metrizable space and ψ : X × M → M is a map

such that ψx defined by ψx (m) = ψ(x, m) is a smooth orientation preserving Riemannian
isometry and x 7→ ψx ∈ C ∞ (M, M ) is continuous with respect to the locally convex
topology of C ∞ (M, M ) mentioned before.
Then there exists a ( C(X)-linear ) unitary Uψ on the Hilbert C(X)-module H ⊗
C(X) (where H = L2 (S) as in Theorem 3.2.3) such that for all x belonging to X, Ux :=
(id ⊗ evx )Uψ is a unitary of the form Uψx on the Hilbert space H commuting with D
and Ux Mφ Ux−1 = Mφ◦ψx−1 . If in addition, the manifold is even dimensional, then Uψx
commutes with the grading operator γ.
Conversely, if there exists a C(X)-linear unitary U on H ⊗ C(X) such that Ux :=
(id ⊗ evx )(U ) is a unitary commuting with D for all x, ( and Ux commutes with the
grading operator γ if the manifold is even dimensional ) and (id ⊗ evx )αU (L∞ (M )) ⊆
L∞ (M ) for all x in X, then there exists a map ψ : X ×M → M satisfying the conditions
mentioned above such that U = Uψ .

Proof: Consider the bundles F̂ = X × F and P̂ = X × P over X × M , with fibres

at (x, m) isomorphic with Fm and Pm respectively, and where F and P are respectively
the bundles of orthonormal frames and the Spin( n ) bundle discussed before. Moreover,
denote by Ψ the map from X × M to itself given by (x, m) 7→ (x, ψ(x, m)). Let πX :
Hom(F̂ , Ψ∗ (F̂ )) → X be the obvious map obtained by composing the projection map
of the X × M bundle with the projection from X × M to X, and let us denote by B
the closed subset of the Polish space C(X, Hom(F̂ , Ψ∗ (F̂ ))) consisting of those f such
that for all x, πX (f (x)) = x. Define B̃ in a similar way replacing F̂ by P̂ . The covering
0
map from P to F induces a covering map from B̃ to B as well. Let dψ : M → B be
the map given by d0ψ (m)(x) ≡ d0ψ (x, m) = dψx |m . Then by Lemma 3.2.2 there exists a
0 0
measurable lift of d , say df from M into B̃. Since d0 (x, m) ∈ Hom(F , F
ψ ψ ψ m ), it is
ψ(x,m)
0 (x, m) will be an element of Hom(P , P
clear that the lift df
ψ m ψ(x,m) ).
We can identify H ⊗ C(X) with C(X → H), and since H has a total set F (say)
consisting of sections of the form [p(·), v], where p : M → P is a measurable section of P
and v belongs to ∆n , we have a total set F̃ of H⊗C(X) consisting of F valued continuous
functions from X. Any such function can be written as [Ξ, v] with Ξ : X × M → P ,
v ∈ ∆n , and Ξ(x, m) ∈ Pm , and we define U on F̃ by U [Ξ, v] = [Θ, v], where

0 −1 −1
Θ(x, m) = df
ψ (x, ψx (m))(Ξ(x, ψx (m))).

It is clear from the construction of the lift that U is indeed a C(X)-linear isometry
which maps the total set F̃ onto itself, so extends to a unitary on the whole of H⊗C(X)
Chapter 3: Quantum group of orientation preserving Riemannian isometries 74

with the desired properties.

Conversely, given U as in the statement of the converse part of the theorem, we
observe that for each x in X, by Theorem 3.2.3, (id ⊗ evx )U = Uψx for some ψx such
that ψx is a smooth orientation preserving Riemannian isometry. This defines the map
ψ by setting ψ(x, m) = ψx (m). The proof will be complete if we can show that x 7→
ψx ∈ C ∞ (M, M ) is continuous, which is equivalent to showing that whenever xn → x
in the topology of X, we must have φ ◦ ψxn → φ ◦ ψx in the Frechet topology of C ∞ (M ),
for any φ ∈ C ∞ (M ). However, by Lemma 1.1.8, we have (id ⊗ evxn )αU ([D, Mφ ]) →
(id ⊗ evx )αU ([D, Mφ ]) in the strong operator topology where αU (X) = U XU −1 . Since
U commutes with D, this implies

(id ⊗ evxn )[D ⊗ id, αU (Mφ )] → (id ⊗ evx )[D ⊗ id, αU (Mφ )],

that is, for all ξ in L2 (S),

L2
[D, Mφ◦ψxn ]ξ → [D, Mφ◦ψx ]ξ.

By choosing φ with support in a local trivializing coordinate neighborhood for S,

and then using the local expression of D used in the proof of Theorem 3.2.3, we conclude
L2
that dk (φ ◦ ψxn ) → dk (φ ◦ ψx ) (where dk is as in the proof of Theorem 3.2.3). Similarly,
by taking repeated commutators with D, we can show the L2 convergence with dk
replaced by dk1 ...dkm for any finite tuple (k1 , ..., km ). In other words, φ ◦ ψxn → φ ◦ ψx
in the topology of C ∞ (M ) described before. 2

3.2.2 Quantum group of orientation-preserving isometries of an R-

twisted spectral triple
In view of the characterization of orientation-preserving isometric action on a classical
manifold ( Theorem 3.2.4 ), we give the following definitions.

Definition 3.2.5. A quantum family of orientation preserving isometries for the (

odd, compact type ) spectral triple (A∞ , H, D) is given by a pair (S, U ) where S is a
separable unital C ∗ -algebra and U is a linear map from H to H ⊗ S such that U e given by
e (ξ ⊗ b) = U (ξ)(1 ⊗ b) ( ξ in H, b in S ) extends to a unitary element of M(K(H) ⊗ S)
U
satisfying the following:
(i) for every state φ on S we have Uφ D = DUφ , where Uφ := (id ⊗ φ)(U e );
(ii) (id ⊗ φ) ◦ αU (a) ∈ (A∞ )00 for all a in A∞ and state φ on S, where αU (x) :=
e (x ⊗ 1)U
U e ∗ for x belonging to B(H).
In case the C ∗ -algebra S has a coproduct ∆ such that (S, ∆) is a compact quantum
group and U is a unitary representation of (S, ∆) on H, we say that (S, ∆) acts by
75 Definition and existence of the quantum group of orientation-preserving isometries

orientation-preserving isometries on the spectral triple.

In case the spectral triple is even with the grading operator γ, a quantum family of
orientation preserving isometries (A∞ , H, D, γ) will be defined exactly as above, with
the only extra condition being that U commutes with γ.

From now on, we will mostly consider odd spectral triples. However let us remark
that in the even case, all the definitions and results obtained by us will go through with
some obvious modifications. We also remark that all our spectral triples are of compact
type.
Consider the category Q ≡ Q(A∞ , H, D) ≡ Q(D) with the object-class consist-
ing of all quantum families of orientation preserving isometries (S, U ) of the given
spectral triple, and the set of morphisms Mor((S, U ), (S 0 , U 0 )) being the set of unital
C ∗ -homomorphisms Φ : S → S 0 satisfying (id ⊗ Φ)(U ) = U 0 . We also consider an-
other category Q0 ≡ Q0 (A∞ , H, D) ≡ Q0 (D) whose objects are triplets (S, ∆, U ), where
(S, ∆) is a compact quantum group acting by orientation preserving isometries on the
given spectral triple, with U being the corresponding unitary representation. The mor-
phisms are the homomorphisms of compact quantum groups which are also morphisms
of the underlying quantum families of orientation preserving isometries. The forgetful
functor F : Q0 → Q is clearly faithful, and we can view F (Q0 ) as a subcategory of Q.
Unfortunately, in general Q0 or Q will not have a universal object. It is easily seen
by taking the standard example A∞ = Mn (C), H = Cn , D = I. Any CQG having
a unitary representation on Cn is an object of Q0 (Mn (C), Cn , I). But by Proposition
1.2.34, there is no universal object in this category. However, the fact that comes to our
rescue is that a universal object exists in each of the subcategories which correspond to
the CQG actions preserving a given faithful functional on Mn .
On the other hand, given any equivariant spectral triple, it has been shown in [31]
that there is a (not necessarily unique) canonical faithful functional which is preserved
by the CQG action. For readers’ convenience, we state this result (in a form suitable to
us) briefly here. Before that, let us recall the definition of an R-twisted spectral data
from [31].

Definition 3.2.6. An R-twisted spectral data ( of compact type ) is given by a quadruplet

(A∞ , H, D, R) where
1. ( A∞ , H, D ) is a spectral triple of compact type.
2. R a positive (possibly unbounded) invertible operator such that R commutes with
D.
3. For all s ∈ R, the map a 7→ σs (a) := R−s aRs gives an automorphism of A∞ (not
necessarily ∗-preserving) satisfying sups∈[−n,n] kσs (a)k < ∞ for all positive integer n.

We shall also sometimes refer to (A∞ , H, D) as an R-twisted spectral triple.

Chapter 3: Quantum group of orientation preserving Riemannian isometries 76

Proposition 3.2.7. Given a spectral triple (A∞ , H, D) (of compact type) which is Q-
equivariant with respect to a representation of a CQG Q on H, we can construct a
positive (possibly unbounded) invertible operator R on H such that (A∞ , H, D, R) is a
twisted spectral data, and moreover, we have
αU preserves the functional τR defined at least on a weakly dense ∗-subalgebra ED of
B(H) generated by the rank-one operators of the form |ξ >< η| where ξ, η are eigenvec-
tors of D, given by
τR (x) = T r(Rx), x ∈ ED .
Remark 3.2.8. When the Haar state of Q is tracial, then it follows from the definition
of R in Lemma 3.1 of [31] and Theorem 1.5 part 1. of [67] that R can be chosen to be
I.
Remark 3.2.9. If Vλ denotes the eigenspace of D corresponding to the eigenvalue, say
2 2
λ, it is clear that τR (X) = etλ Tr(Re−tD X) for all X = |ξ >< η| with ξ, η belonging to
Vλ and for any t > 0. Thus, the αU -invariance of the functional τR on ED is equivalent
2
to the αU -invariance of the functional X 7→ Tr(XRe−tD ) on ED for each t > 0. If,
2
furthermore, the R-twisted spectral triple is Θ-summable in the sense that Re−tD is
trace class for every t > 0, the above is also equivalent to the αU -invariance of the
2
bounded normal functional X 7→ Tr(XRe−tD ) on the whole of B(H). In particular,
−tD 2
this implies that αU preserves the functional B(H) 3 x 7→ Limt→0+ Tr(xRe−tD2 ) , where
Tr(Re )
Lim is as defined in subsection 1.5.2.
This motivates the following definition:
Definition 3.2.10. Given an R-twisted spectral data (A∞ , H, D, R) of compact type,
a quantum family of orientation preserving isometries (S, U ) of (A∞ , H, D) is said to
preserve the R-twisted volume, (simply said to be volume-preserving if R is understood)
if one has (τR ⊗ id)(αU (x)) = τR (x).1S for all x in ED , where ED and τR are as in
Proposition 3.2.7. We shall also call (S, U ) a quantum family of orientation-preserving
isometries of the R-twisted spectral triple.
If, furthermore, the C ∗ -algebra S has a coproduct ∆ such that (S, ∆) is a CQG and
U is a unitary representation of (S, ∆) on H, we say that (S, ∆) acts by volume and
orientation-preserving isometries on the R-twisted spectral triple.
We shall consider the categories QR ≡ QR (D) and Q0R ≡ Q0R (D) which are the full
subcategories of Q and Q0 respectively, obtained by restricting the object-classes to the
volume-preserving quantum families.
Remark 3.2.11. We shall not need the full strength of the definition of twisted spectral
data here; in particular the third condition in the definition 3.2.6. However, we shall
continue to work with the usual definition of R-twisted spectral data, keeping in mind
that all our results are valid without assuming the third condition.
77 Definition and existence of the quantum group of orientation-preserving isometries

Let us now fix a spectral triple (A∞ , H, D) which is of compact type. The C ∗ -
algebra generated by A∞ in B(H) will be denoted by A. Let λ0 = 0, λ1 , λ2 , · · · be the
eigenvalues of D with Vi denoting the ( di -dimensional, 0 ≤ di < ∞ ) eigenspace for λi .
Let {eij , j = 1, ..., di } be an orthonormal basis of Vi . We also assume that there is a
positive invertible R on H such that (A∞ , H, D, R) is an R-twisted spectral data. The
operator R must have the form R|Vi = Ri , say, with Ri positive invertible di ×di matrix.
Let us denote the CQG Au,di (RiT ) by Ui , with its canonical unitary representation βi on
RT
Vi ∼
= Cdi , given by βi (eij ) =
P
k eik ⊗ u i . Let U be the free product of Ui , i = 1, 2, ...
kj
and β = ∗i βi be the corresponding free product representation of U on H. We shall also
consider the corresponding unitary element β̃ in M(K(H) ⊗ U).
Lemma 3.2.12. Consider the R-twisted spectral triple (A∞ , H, D) and let (S, U ) be a
quantum family of volume and orientation preserving isometries of the given spectral
triple. Moreover, assume that the map U is faithful in the sense that there is no proper
C ∗ -subalgebra S1 of S such that U e belongs to M(K(H) ⊗ S1 ). Then we can find a C ∗ -
isomorphism φ : U/I → S between S and a quotient of U by a C ∗ -ideal I of U, such
that U = (id ⊗ φ) ◦ (id ⊗ ΠI ) ◦ β, where ΠI denotes the quotient map from U to U/I.
If, furthermore, there is a compact quantum group structure on S given by a coprod-
uct ∆ such that (S, ∆, U ) is an object in Q0R , the ideal I is a Woronowicz C ∗ -ideal and
the C ∗ -isomorphism φ : U/I → S is a morphism of compact quantum groups.
(i)
Proof : It is clear that U maps Vi into Vi ⊗ S for each i. Let vkj (j, k = 1, ..., di ) be
P (i) (i)
the elements of S such that U (eij ) = k eik ⊗ vkj . Note that vi := ((vkj )) is a unitary
(i)
in Mdi (C) ⊗ S. Moreover, the ∗-subalgebra generated by all {vkj , i ≥ 0, , j, k ≥ 1} must
be dense in S by the assumption of faithfulness.
Consider the ∗-homomorphism αi from the finite dimensional C ∗ algebra Ai ∼ =
Mdi (C) generated by the rank one operators {|eij >< eik |, j, k = 1, ..., di } to Ai ⊗ S
given by αi (y) = U e (y ⊗ 1)U e ∗ |V . Clearly, the restriction of the functional τR on Ai
i

is nothing but the functional given by Tr(Ri ·), where Tr denotes the usual trace of
matrices. Since αi preserves this functional by assumption, we get, by the universality
(i) RT (i)
of Ui , a C ∗ -homomorphism from Ui to S sending ukj ≡ ukji to vkj , and by definition
of the free product, this induces a C ∗ -homomorphism, say Π, from U onto S, so that
U/I ∼ = S, where I := Ker(Π).
In case S has a coproduct ∆ making it into a compact quantum group and U is a
quantum group representation, it is easy to see that the subalgebra of S generated by
(i) (i) P (i) (i)
{vkj , i ≥ 0, j, k = 1, ..., di } is a Hopf algebra, with ∆(vkj ) = l vkl ⊗ vlj . From this,
it follows that Π is Hopf-algebra morphism, hence I is a Woronowicz C ∗ -ideal. 2

Theorem 3.2.13. For any R-twisted spectral triple of compact type (A∞ , H, D), the
category QR of quantum families of volume and orientation preserving isometries has a
Chapter 3: Quantum group of orientation preserving Riemannian isometries 78

universal (initial) object, say (G,

e U0 ). Moreover, Ge has a coproduct ∆0 such that (G,e ∆0 )
is a compact quantum group and (G, e ∆0 , U0 ) is a universal object in the category Q0 .
R
The representation U0 is faithful.

Proof : Recall the C ∗ -algebra U considered before, along with the representation
β and the corresponding unitary βe belonging to M(K(H) ⊗ U). For any C ∗ -ideal
I of U, we shall denote by ΠI the canonical quotient map from U onto U/I, and let
ΓI = (id⊗ΠI )◦β. Clearly, Γ fI = (id⊗πI )◦βe is a unitary element of M(K(H)⊗U/I). Let
F be the collection of all those C ∗ -ideals I of U such that (id ⊗ ω) ◦ αΓI ≡ (id ⊗ ω) ◦ adΓf
I
maps A∞ into A00 for every state ω (equivalently, every bounded linear functional) on
U/I. This collection is nonempty, since the trivial one-dimensional C ∗ -algebra C gives
an object in QR and by Lemma 3.2.12 we do get a member of F. Now, let I0 be the
intersection of all ideals in F. We claim that I0 is again a member of F. Indeed, in the
notation of Lemma 1.1.2, it is clear that for a in A∞ , (id ⊗ φ) ◦ Γ e I (a) belongs to A00
0
S
for all φ in the convex hull of Ω (−Ω). Now, for any state ω on U/I0 , we can find,
by Lemma 1.1.2, a net ωj in the above convex hull (so in particular kωj k ≤ 1 for all j)
such that ω(x + I0 ) = limj ωj (x + I0 ) for all x in U/I0 .
It follows from Lemma 1.1.8 that (id⊗ωj )(X) → (id⊗ω)(X) (in the strong operator
topology) for all X belonging to M(K(H) ⊗ U/I0 ). Thus, for a in A∞ , (id ⊗ ω) ◦ αΓeI (a)
0
is the limit of (id ⊗ ωi ) ◦ αΓeI (a) in the strong operator topology, hence belongs to A00 .
0
We now show that (Ge := U/I0 , ΓI ) is a universal object in QR . To see this, consider
0

any object (S, U ) of QR . Without loss of generality we can assume U to be faithful,

(i)
since otherwise we can replace S by the C ∗ -subalgebra generated by the elements {vkj }
appearing in the proof of Lemma 3.2.12. But by Lemma 3.2.12 we can further assume
that S is isomorphic with U/I for some I belonging to F. Since I0 ⊆ I, we have a
C ∗ -homomorphism from U/I0 onto U/I, sending x + I0 to x + I, which is clearly a
morphism in the category QR . This is indeed the unique such morphism, since it is
(i)
uniquely determined on the dense subalgebra generated by {ukj + I0 , i ≥ 0, j, k ≥ 1}
of G.
e
To construct the coproduct on Ge = U/I0 , we first consider U (2) : H → H ⊗ Ge ⊗ Ge
given by
U (2) = (ΓI0 )(12) (ΓI0 )(13) ,
e U (2) ) is an
where Uij is the usual ‘leg-numbering notation’. It is easy to see that (Ge ⊗ G,
object in the category QR , so by the universality of (G,e ΓI ), we have a unique unital
0

C ∗ -homomorphism ∆0 : Ge → Ge ⊗ Ge satisfying

(id ⊗ ∆0 )(ΓI0 ) = U (2) .

79 Definition and existence of the quantum group of orientation-preserving isometries

Letting both sides act on eij , we get

!
(i) (i) (i)
X X X
eil ⊗ (πI0 ⊗ πI0 ) ulk ⊗ ukj = eil ⊗ ∆0 (πI0 (ulj )).
l k l

˜ (i) ) = P u(i) ⊗ u(i) (where ∆

Comparing coefficients of eil , and recalling that ∆(u ˜
lj k lk kj
denotes the coproduct on U), we have

˜ = ∆0 ◦ πI
(πI0 ⊗ πI0 ) ◦ ∆ (3.2.1)
0

(i)
on the linear span of {ujk , i ≥ 0, j, k ≥ 1}, and hence on the whole of U. This implies
e maps I0 = Ker(πI ) into Ker(πI ⊗ πI ) = (I0 ⊗ 1 + 1 ⊗ I0 ) ⊂ U ⊗ U. In other
that ∆ 0 0 0

words, I0 is a Hopf C ∗ -ideal, and hence Ge = U/I0 has the canonical compact quantum
group structure as a quantum subgroup of U. It is clear from the relation (3.2.1) that
∆0 coincides with the canonical coproduct of the quantum subgroup U/I0 inherited
from that of U. It is also easy to see that the object (G, e ∆0 , ΓI ) is universal in the
0
0
category QR , using the fact that (by Lemma 3.2.12) any compact quantum group (S, U )
acting by volume and orientation preserving isometries on the given spectral triple is
isomorphic with a quantum subgroup U/I, for some Hopf C ∗ -ideal I of U.
Finally, the faithfulness of U0 follows from the universality by standard arguments
which we briefly sketch. If G1 ⊂ Ge is a ∗-subalgebra of Ge such that U f0 is an element of
M(K(H) ⊗ G1 ), it is easy to see that (G1 , U0 ) is also an object in QR , and by definition
of universality of Ge it follows that there is a unique morphism, say j, from Ge to G1 . But
the map i ◦ j is a morphism from Ge to itself, where i : G1 → Ge is the inclusion. Again
e 2
by universality, we have that i ◦ j = idGe, so in particular, i is onto, that is,. G1 = G.

Consider the ∗-homomorphism α0 := αU0 , where (G, e U0 ) is the universal object

obtained in the previous theorem. For every state φ on G, e (id ⊗ φ) ◦ α0 maps A into
A00 . However, in general α0 may not be faithful even if U0 is so, and let G denote the
C ∗ -subalgebra of Ge generated by the elements {(f ⊗ id) ◦ α0 (a) : f ∈ A∗ , a ∈ A}.

Remark 3.2.14. If the spectral triple is even, then all the proofs above go through with
obvious modifications.

Definition 3.2.15. We shall call G the quantum group of orientation-preserving isome-

+
tries of R-twisted spectral triple (A∞ , H, D, R) and denote it by QISOR (A∞ , H, D, R)
+ ^+ (D).
or even simply as QISOR (D). The quantum group Ge is denoted by QISO R
If the spectral triple is even, then we will denote G and G by QISO+ (D, γ) and
e
R
^+ (D, γ) respectively.
QISO R
Chapter 3: Quantum group of orientation preserving Riemannian isometries 80

3.2.3 Stability and topological action

In this subsection, we are going to use the notations as in subsection 3.2.2, in particular
e G, U0 , α0 . It is not clear from the definition and construction of QISO+ (D) whether
G, R
the C ∗ algebra A generated by A∞ is stable under α0 in the sense that (id ⊗ φ) ◦ α0
maps A into A for every φ. Moreover, even if A is stable, the question remains whether
+
α0 is a C ∗ -action of the CQG QISOR (D). In chapter 5, subsection 5.4.2 we have given
an example of a spectral triple for which the ∗-homomorphism α0 is not a C ∗ action.
However, one can prove that α0 is a C ∗ action for a rather large class of spectral triples,
including the cases mentioned below.
(i) For any spectral triple for which there is a ‘reasonable’ Laplacian in the sense
of [30]. This includes all classical spectral triples as well as their Rieffel deformation
(with R = I).
(ii) Under the assumption that there is an eigenvalue of D with a one-dimensional
eigenspace spanned by a cyclic separating vector ξ such that any eigenvector of D
belongs to the span of A∞ ξ and {a ∈ A∞ : aξ is an eigenvector of D} is norm-dense
in A ( to be proved in subsection 3.2.4 ).

Now we prove the sufficiency of the condition (i).

We begin with a sufficient condition for stability of A∞ under α0 . Let (A∞ , H, D)
be a (compact type) spectral triple such that
(1) A∞ and {[D, a], a ∈ A∞ } are contained in the domains of all powers of the deriva-
tions [D, ·] and [|D|, ·].

We will denote by Tet the one parameter group of ∗-automorphisms on B(H) given
by Tet (S) = eitD Se−itD for all S in B(H) which is clearly continuous in SOT. We will
denote the generator of this group by δ. For X such that [D, X] is bounded, we have
δ(X) = i[D, X] and hence
Z t

≤ tk[D, X]k.
Tt (X) − X = Ts ([D, X])ds
e e
0

Let us say that the spectral triple satisfies the Sobolev condition if
\
A∞ = A00 Dom(δ n ).
n≥1

Then we have the following result, which is a natural generalization of the classical
situation, where a measurable isometric action automatically becomes topological (in
fact smooth).
81 Definition and existence of the quantum group of orientation-preserving isometries

Theorem 3.2.16. (i) For every state φ on G,

\
(id ⊗ φ) ◦ α0 (A∞ ) belongs to A00 Dom(δ n ).
n≥1

(ii) If the spectral triple satisfies the Sobolev condition then A∞ ( and hence A ) is
stable under α0 .

Proof: Since U0 commutes with D ⊗ I, it is clear that the automorphism group

Tet commutes with α0φ ≡ (id ⊗ φ) ◦ α0 , and thus by the continuity of α0 in the strong
operator topology it is easy to see that, for a in Dom(δ),

Tet (α0φ (a)) − α0φ (a)

lim
t→0+ t
Tet (a) − a
= lim α0φ ( )
t→0+ t
= α0φ (δ(a)).

Thus, α0φ leaves Dom(δ) invariant and commutes with δ. Proceeding similarly, we prove
(i). The assertion (ii) is a trivial consequence of (i) and the Sobolev condition. 2

Let us now assume

2
(2) The spectral triple is Θ-summable, that is, for every t > 0, e−tD is trace-class
−tD 2 )
and the functional τ (X) = Limt→0 Tr(Xe−tD2 ( where Lim is as in subsection 1.5.2 ), is a
Tr(e )
positive faithful trace on the ∗ algebra, say S ∞ , generated by {Tes (A∞ ), Tes (A∞ )([D, a]) :
a ∈ A∞ }.
The functional τ is to be interpreted as the volume form ( we refer to [29], [30] for
the details ). The completion of S ∞ in the norm of B(H) is denoted by S, and we
1
shall denote by kak2 and k · k∞ the L2 -norm τ (a∗ a) 2 and the operator norm of B(H)
respectively.
From the definition of τ, it is also clear that Tet preserves τ, so extends to a group of
unitaries on N := L2 (S ∞ , τ ). Moreover, for X such that [D, X] is in B(H), in particular
for X in S ∞ , we have
2
Ts (X) − X
e
2
∗
= τ (Tes (X) (Tes (X) − X)) + τ (X ∗ (X − Tes (X)))

≤ 2 X − Tes (X) kXk2

∞
≤ 2sk[D, X]k∞ kXk2 ,
Chapter 3: Quantum group of orientation preserving Riemannian isometries 82

which clearly shows that s 7→ Tes (X) is L2 -continuous for X belonging to S ∞ , hence (by
unitarity of Tes ) on the whole of N , that is, it is a strongly continuous one-parameter
group of unitaries. Let us denote its generator by δ̃, which is a skew adjoint map, that
is, iδ̃ is self adjoint, and Tet = exp(tδ̃). Clearly, δ̃ = δ = [D, ·] on S ∞ .
We will denote L2 (A∞ , τ ) ⊂ N by HD 0 and the restriction of δ̃ to H0 (which is a
D
closable map from HD 0 to N ) by d . Thus, d is closable too.
D D
We now recall the assumptions made in chapter 2, subsection 2.1.2 for defining
the ‘Laplacian’ and the corresponding quantum isometry group of a spectral triple
(A∞ , H, D).
The following conditions will also be assumed throughout the rest of this subsection:
(3) A∞ ⊆ Dom(L) where L ≡ LD := −d∗D dD .
(4) L has compact resolvent.
(5) Each eigenvector of L ( which has a discrete spectrum , hence a complete set of
eigenvectors ) belongs to A∞ .
(6) The complex linear span of the eigenvectors of L, say A∞
0 ( which is a subspace of
∞
A by assumption (5) ), is norm dense in A . ∞

It is clear that L maps (A∞

0 ) into itself. The ∗-subalgebra of A
∞ generated by A∞
0
is denoted by A0 . We also note that L = P0 L̃P0 , where L̃ := (iδ̃)2 (which is a self
adjoint operator on N ) and P0 denotes the orthogonal projection in N whose range is
the subspace HD 0 .

Theorem 3.2.17. Let (A∞ , H, D) be a spectral triple satisfying the assumptions

(1) − (6) made above. In addition, assume that at least one of conditions (a) and
(b) mentioned below is satisfied:

(a) A00 ⊆ HD0 .

(b) α0φ (A∞ ) ⊆ A∞ for every state φ on G = QISOI+ (D).

Then α0 is a C ∗ -action of QISOI+ (D) on A.

Proof : Under either of the conditions (a) and (b), for any fixed φ, the map α0φ maps
A∞ into the subset HD 0 of N . Since αφ also commutes with [D, ·] on A∞ , it is clear
0
that α0φ maps S ∞ into N . In fact, using the complete positivity of the map α0φ and the
α0 -invariance of τ , we see that
∗
τ (α0φ (a) α0φ (a)) ≤ τ (α0φ (a∗ a)) = (id ⊗ φ)((τ ⊗ id)α0 (a∗ a)) = τ (a∗ a).1

which implies that α0φ extends to a bounded operator from N to itself. Since U0 com-
mutes with D, it is clear that α0φ (viewed as a bounded operator on N ) will commute
83 Definition and existence of the quantum group of orientation-preserving isometries

with the group of unitaries Tet , hence with its generator δ̃ and also with the self adjoint
operator L̃ = (iδ̃)2 .
On the other hand, it follows from the definition of G = QISOI+ (D) that (τ ⊗
id)(α0 (X)) = τ (X)1G for all X in B(H), in particular for X belonging to S ∞ , and thus
the map S ∞ ⊗alg G 3 (a⊗q) 7→ α0 (a)(1⊗q) extends to a G-linear unitary, denoted by W
(say), on the Hilbert G-module N ⊗ G. Note that here we have used the fact( which that
for any φ, (id ⊗ φ)(W )(S ∞ ⊗alg G) ⊆ N , since α0φ (S ∞ ) ⊆ N . The commutativity of α0φ
with Tet for every φ clearly implies that W and Tet ⊗ idG commute on N ⊗ G. Moreover,
α0φ maps HD 0 into itself, so W maps H0 ⊗ G into itself, and hence (by unitarity of W ) it
D
commutes with the projection P0 ⊗ 1. It follows that α0φ commutes with P0 , and (since
it also commutes with L̃) commutes with L = P0 L̃P0 as well.
Thus, α0φ preserves each of the (finite dimensional) eigenspaces of the Laplacian
L, and so is a Hopf algebraic action on the subalgebra A0 spanned algebraically by
these eigenvectors. Moreover, the G-linear unitary W clearly restricts to a unitary
representation on each of the above eigenspaces. If we denote by ((qij ))(i,j) the G-valued
unitary matrix corresponding to one such particular eigenspace, then by Proposition
1.2.23, qij must belong to G0 and we must have (qij ) = δij (Kronecker delta). This
implies (id⊗)◦α0 = id on each of the eigenspaces, hence on the norm-dense subalgebra
A0 of A, completing the proof of the fact that α0 extends to a C ∗ action on A. 2

Combining the above theorem with Theorem 3.2.16, we get the following immediate
corollary.

Corollary 3.2.18. If the spectral triple satisfies the Sobolev condition mentioned before,
in addition to the assumptions 1 − 6, then QISOI+ (D) has a C ∗ -action. In particular,
for a classical spectral triple, QISOI+ (D) has C ∗ -action.

Remark 3.2.19. Let us remark here that in case the restriction of τ on A∞ is normal,
that is, continuous with respect to the weak operator topology inherited from B(H), then
HD0 will contain A00 , which is the closure of A∞ in the weak operator topology of B(H),

so the condition (a) of Theorem 3.2.17 (and hence its conclusion) holds.

Remark 3.2.20. In a private communication to us, Shuzhou Wang has kindly pointed
out that a possible alternative approach to the formulation of quantum group of isome-
tries may involve the category of CQG which has a C ∗ -action on the underlying C ∗
algebra and a unitary representation with respect to which the Dirac operator is equiv-
ariant. However, we see from Corollary 5.4.17 of chapter 5 that the category proposed
by Wang does not admit a universal object in general.
Chapter 3: Quantum group of orientation preserving Riemannian isometries 84

3.2.4 Universal object in the categories Q or Q0

We shall now investigate further conditions on the spectral triple which will ensure the
existence of a universal object in the category Q or Q0 . Whenever such a universal
^+ (D), and denote by QISO+ (D) its largest
object exists we shall denote it by QISO
Woronowicz subalgebra for which αU on A∞ (where U is the unitary representation of
^+ (D) on H) is faithful.
QISO

Remark 3.2.21. If QISO ^+ (D) exists, then by Proposition 3.2.7, there will exist some
^+ (D) is an object in the category Q0 (D). Since the universal object
R such that QISO R
^ + ^+ (D), we have
in this category, that is, QISO (D), is clearly a sub-object of QISO
R
^+ (D) ∼
QISO ^+ (D) for this choice of R.
= QISO R

Let us state and prove a result below, which gives some sufficient conditions for the
^+ (D).
existence of QISO

Theorem 3.2.22. Let (A∞ , H, D) be a spectral triple of compact type as before and
assume that D has an one-dimensional eigenspace spanned by a unit vector ξ, which is
cyclic and separating for the algebra A∞ . Moreover, assume that each eigenvector of
D belongs to the dense subspace A∞ ξ of H. Then there is a universal object, (G, e U0 ).
Moreover, Ge has a coproduct ∆0 such that (G, e ∆0 ) is a compact quantum group and
e ∆0 , U0 ) is a universal object in the category Q0 .
(G,
If we denote by G the Woronowicz C ∗ subalgebra of Ge generated by elements of the
form hαU0 (a)(η ⊗ 1), η 0 ⊗ 1iGe where η, η 0 are in H, a in A∞ and h. , .iGe denotes the Ge
valued inner product of H ⊗ G, e we have Ge ∼ = G ∗ C(T).

Proof: Let Vi , {eij } be as before, and by assumption we have eij = xij ξ for a unique
xij in A∞ . Clearly, since ξ is separating, the vectors {eij = x∗ij ξ, j = 1, ..., di } are
linearly independent, so the matrix Qi = ((heij , eik i))dj,k=1
i
is positive and invertible.
Now, given a quantum family of orientation-preserving isometries (S, U ), we must have
Ue (ξ ⊗ 1) = ξ ⊗ q, say, for some q in S, and from the unitarity of U e it follows that q is a
unitary element. Moreover, U leaves Vi invariant, so let U e (eij ⊗ 1) = P eik ⊗ v (i) . But
k kj
this can be rewritten as

(i)
X
αU (xij )(ξ ⊗ q) = xik ξ ⊗ vkj .
k

(i)
xik ⊗ vkj q ∗ , and thus
P
Since ξ is separating and q is unitary, this implies αU (xij ) = k
we have

(i) (i)
X X
e (eij ⊗ 1) = αU (xij )∗ (ξ ⊗ q) =
U x∗ik ξ ⊗ q(vkj )∗ q = eik ⊗ q(vkj )∗ q.
k k
85 Definition and existence of the quantum group of orientation-preserving isometries

Taking the S-valued inner product h·, ·iS on both sides of the above expression, and
using the fact that U preserves this S-valued inner product, we obtain


eij , eij 0 .1
D E
= e (eij ⊗ 1) , U
U e (eij 0 ⊗ 1)
* +
X (i) ∗ X (i) ∗
= eik ⊗ q(vkj ) q, eik0 ⊗ q(vk0 j 0 ) q
k k0
(i) (i) ∗
X
= heik , eik0 i q ∗ vkj q ∗ q(vk0 j 0 ) q.
k,k0

This implies,

q(Qi )jj 0 q ∗ .1
(i) ∗
X (i)
= vkj heik , eik0 i (vk0 j 0 ) .
k,k0

Thus,
X (i) (i) ∗
(Qi )jj 0 = vkj (Qi )kk0 (vk0 j 0 ) .
k,k0

(i)
Hence, we have Qi = vi0 Qi vi (where vi = ((vkj ))). Thus, Q−1 0
i vi Qi must be the (both-
sided) inverse of vi . Thus, we get a canonical surjective morphism from Au,di (Qi ) to the
(i)
C ∗ algebra generated by {vkj : j, k = 1, 2, ...di }. This induces a surjective morphism
from the free product of Au,di (Qi ), i = 1, 2, ... onto S. The rest of the arguments for
showing the existence of Ge will be quite similar to the arguments used in the proof of
Theorem 3.2.13, hence omitted.
Now we come to the proof of the last part of the theorem. For a in A∞ , Ue (aξ ⊗ 1) =
αU (a)U e (ξ ⊗ 1) = αU (a)(ξ ⊗ q). Now, recalling that Span{aξ : a ∈ A∞ } is dense in H,
it is clear that Ge = G ∗ C ∗ (q) ∼
= G ∗ C(T). 2

Remark 3.2.23. Some of the examples considered in section 3.4 will show that the con-
+
ditions of the above theorem are not actually necessary; QISO
^ (D) may exist without
the existence of a single cyclic separating eigenvector as above.
Let (A∞ , H, D) be a spectral triple of compact type satisfying the conditions of
the above theorem. Let the faithful vector state corresponding to the cyclic separating
vector ξ be denoted by τ. Let A00 = span{a ∈ A∞ : aξ is an eigenvector of D}.
Moreover, assume that A00 is norm dense in A∞ .
Let D̂ : A00 → A00 be defined by :
D̂(a)ξ = D(aξ).
This is well defined since ξ is cyclic and separating.
Chapter 3: Quantum group of orientation preserving Riemannian isometries 86

Definition 3.2.24. Let A be a C ∗ algebra and A∞ be a dense ∗-subalgebra. Let

(A∞ , H, D) be a spectral triple of compact type as above.
Let Cb be the category with objects (Q, α) such that Q is a compact quantum group
with a C ∗ action α on A such that:
1. α is τ preserving ( where τ is as above ), that is, (τ ⊗ id)(α(a)) = τ (a).1.
2. α maps A00 inside A00 ⊗alg Q.
3. αD b = (Db ⊗ I)α.

Corollary 3.2.25. There exists a universal object Q

b of the category Cb and it is iso-
morphic to the Woronowicz C ∗ subalgebra G = QISO+ (D) of Ge obtained in Theorem
3.2.22.

Proof : The proof of the existence of the universal object follows verbatim from the
proof of Theorem 2.1.5 replacing L by D b and noting that D has compact resolvent. We
denote by α b the action of Q
b on A.
Now, we prove that Q b is isomorphic to G.
Each eigenvector of D is in A∞ by assumption. It is easily observed from the proof
of Theorem 3.2.22 that αU0 maps the norm-dense ∗-subalgebra A00 into A00 ⊗alg G0 ,
and (id⊗)◦αU0 = id, so that αU0 is indeed a C ∗ action of the CQG G. Moreover, it can
be easily seen that τ preserves αU0 and that αU0 commutes with D. b Therefore, (G, αU )
0

is an element of Obj(C)b and hence G is a quantum subgroup of Q b by the universality

of Q.
b
For the converse, we start by showing that α b induces a unitary representation W of
Qb ∗ C(T) on H which commutes with D, and the corresponding conjugated action αW
coincides with αb.
Define W (aξ) = α b(a)(ξ)(1 ⊗ q ∗ ) for all a in A∞0 where q is a generator of C(T).
Since we have (τ ⊗ id)(α(a)) = τ (a).1, it follows that W f is a (Q b ∗ C(T)-linear)
isometry on the dense subspace A00 ξ ⊗alg Q b and thus extends to H ⊗ (Q b ∗ C(T)) as an
isometry. Moreover, since α b(A)(1 ⊗ Q) b is norm dense in A ⊗ Q b (by the definition of a
CQG action) it is clear that the range of W f is dense, so W
f is indeed a unitary. It is
quite obvious that it is a unitary representation of Q b ∗ C(T).
We also have,

W D(aξ)
= W (D(a)ξ)
b =α
b(D(a))(ξ)(1
b ⊗ q∗)
α(a)ξ)(1 ⊗ q ∗ ) = (D ⊗ I)W (aξ),
= (D ⊗ I)(b

that is, W commutes with D.

Moreover, it is easy to observe that αW = α
b. This gives a surjective CQG morphism
from G = G ∗ C(T) to Q ∗ C(T), sending G onto Q,
e b b which completes the proof. 2
87 Comparison with the approach of [30] based on Laplacian

3.3 Comparison with the approach of [30] based on Lapla-

cian
Throughout this section, we shall assume the set-up of subsection 3.2.3 for the existence
of a ‘Laplacian’, including assumptions 1 − 6. Let us also use the notation of that
subsection.
We recall from chapter 2 that a CQG (S, ∆) which has an action α on A is said
to act smoothly and isometrically on the noncommutative manifold ( of compact type
) (A∞ , H, D) if (id ⊗ φ) ◦ α(A∞ ∞
0 ) ⊆ A0 for every state φ on S, and also (id ⊗ φ) ◦ α
commutes with the Laplacian L ≡ LD on A∞ ∞
0 ( where A0 is the complex linear span
of the eigenvectors of L ). One can consider the category Q0LD of all compact quantum
groups acting smoothly and isometrically on A, where the morphisms are quantum
group morphisms which intertwin the actions on A. We make the following additional
assumption throughout the present section:

(7) There exists a universal object in Q0LD ( the quantum isometry group for the
Laplacian L ≡ LD in the sense of [30]), and it is denoted by QISOL ≡ QISOLD

The following result now follows immediately from Theorem 3.2.17 of subsection 2.3.

Corollary 3.3.1. If (A∞ , H, D) is a spectral triple ( of compact type ) satisfying any

of the two conditions ( a ) or ( b ) of Theorem 3.2.17, then QISOI+ (D) is a sub-object
of QISOLD in the category Q0LD .

Proof: The proof is a consequence of the fact that QISOI+ (D) has the C ∗ -action α0
on A, and the observation already made in the proof of the Theorem 3.2.17 that this
action commutes with the Laplacian LD . 2

Now, we will need the Hilbert space of forms Hd+d∗ corresponding to a Θ-summable
spectral triple (A∞ , H, D) as discussed in subsection 1.5.2. We recall that one obtains
an associated spectral triple (A∞ , Hd+d∗ , d + d∗ ). We assume that this spectral triple is
of compact type, that is, d + d∗ has compact resolvents.
We will denote the inner product on the space of k forms coming from the spectral
triples (A∞ , H, D) and (A∞ , Hd+d∗ , d + d∗ ) by h , iHk and h , iHk ∗ respectively,
D d+d
k = 0, 1.
We will denote by πD , πd+d∗ the representations of A∞ in H and Hd+d∗ respectively.
Let Ud+d∗ be the canonical unitary representation of QISOI+ (d + d∗ ) on Hd+d∗ .
Hd+d∗ breaks up into finite dimensional orthogonal subspaces corresponding to the
distinct eigenvalues of ∆ := (d + d∗ )2 = d∗ d + dd∗ . It is easy to see that ∆ leaves each of
the subspaces HD i invariant, and we will denote by V i
λ,i the subspace of Hd+d∗ spanned
Chapter 3: Quantum group of orientation preserving Riemannian isometries 88

by eigenvectors of ∆ corresponding to the eigenvalue λ. Let {ej,λ,i }j be an orthonormal

0 .
basis of Vλ,i . Note that LD is the restriction of ∆ to HD
Now we recall Proposition 2.1.8. It was shown there that QISOLD has a uni-
∗
tary representation U ≡ UL on Hd+d such that U commutes with d + d∗ . Thus,
(A∞ , Hd+d∗ , d + d∗ ) is a QISOLD equivariant spectral triple. Moreover, by Remark
2.1.6, QISOLD has tracial Haar state, which implies, by Proposition 3.2.7 and Remark
3.2.8 that αU keeps the functional τI invariant. Summarizing, we have the following
result:

Proposition 3.3.2. The quantum isometry group (QISOLD , UL ) is a sub-object of

(QISOI+ (d + d∗ ), Ud+d∗ ) in the category QI (d + d∗ ), so in particular, QISOLD is
isomorphic to a quotient of QISOI+ (d + d∗ ) by a Woronowicz C ∗ ideal.

We shall give (under mild conditions) a concrete description of the above Woronowicz
ideal.
Let I be the C ∗ ideal of QISOI+ (d + d∗ ) generated by
D E
∪λ∈σ(∆) { (P0⊥ ⊗ id)Ud+d∗ (ejλ0 ), ejλi0 ⊗ 1 : j, i0 ≥ 1},
0 , h. , .i denotes the QISO + (d + d∗ ) valued inner
where P0 is the projection onto HD I
product and σ(∆) denotes the spectrum of ∆.
Since Ud+d∗ keeps the eigenspaces of ∆ = (d + d∗ )2 invariant, we can write
X X
Ud+d∗ (ejλ0 ) = ekλ0 ⊗ qkjλ0 + ek0 λi0 ⊗ qk0 jλi0 ,
k i0 6=0,k0

for some qkjλ0 , qk0 jλi0 in QISOI+ (d + d∗ ).

We note that qk0 jλi0 is in I if i0 6= 0.

Lemma 3.3.3. I is a co-ideal of QISOI+ (d + d∗ ).

Proof : It is enough to prove the relation ∆(X) ∈ I ⊗ QISOI+ (d + d∗ ) + QISOI+ (d +

d∗ ) ⊗ I for the elements X in I of the form (P0⊥ ⊗ id)Ud+d∗ (ejλ0 ), ejλi0 ⊗ 1 . We have:



D E
∆( (P0⊥ ⊗ id)Ud+d∗ (emλ0 ), ejλi0 ⊗ 1 )

D E
= (P0⊥ ⊗ id)(id ⊗ ∆)Ud+d∗ (emλ0 ), ejλi0 ⊗ 1 ⊗ 1 )

D E
= (P0⊥ ⊗ id)U(12) U(13) (emλ0 ), ejλi0 ⊗ 1 ⊗ 1
89 Comparison with the approach of [30] based on Laplacian

* +
X
= (P0⊥ ⊗ id)U(12) ( ekλ0 ⊗ 1 ⊗ qkmλ0 ) , ejλi0 ⊗ 1 ⊗ 1
k

X D E
+ (P0⊥ ⊗ id)U(12) (elλi0 ⊗ 1 ⊗ qlmλi0 ) , ejλi0 ⊗ 1 ⊗ 1
i0 6=0,l

XD E
= (P0⊥ ⊗ id)(ek0 λ0 ⊗ qk0 kλ0 ⊗ qkmλ0 ) , ejλi0 ⊗ 1 ⊗ 1
k,k0

X D E
+ (P0⊥ ⊗ id)(ek00 kλi0 ⊗ qk00 ,k,λ,i0 ⊗ qkmλ0 ) , ejλi0 ⊗ 1 ⊗ 1
i0 6=0, k, k00

X D E
+ (P0⊥ ⊗ id)(el0 λi0 ⊗ ql0 lλi0 ⊗ qlmλi0 ) , ejλi0 ⊗ 1 ⊗ 1
i0 6=0, l, l0

X D E
+ (P0⊥ ⊗ id)(e l00 λi00 ⊗q l00 lλi00 ⊗q lmλi0 ) , ejλi0 ⊗ 1 ⊗ 1
i0 6=0, i00 6=i0 , l, l00

X
= hek00 λi0 ⊗ qk00 kλi0 ⊗ qkmλ0 , ejλi0 ⊗ 1 ⊗ 1i
i0 6=0, k0 , k00

X
+ hel0 λi0 ⊗ ql0 lλi0 ⊗ qlmλi0 , ejλi0 ⊗ 1 ⊗ 1i
i0 6=0, l, l0

X
+ hel00 λi00 ⊗ ql00 lλi00 ⊗ qlmλi0 , ejλi0 ⊗ 1 ⊗ 1i ,
i0 6=0, i00 6=i0 , i00 6=0, l, l00

which is clearly in I ⊗ QISOI+ (d + d∗ ) + QISOI+ (d + d∗ ) ⊗ I, as qkjλi0 is an element of

I for i0 6= 0. 2

Theorem 3.3.4. If αUd+d∗ is a C ∗ action on A, then we have QISOLD ∼

= QISOI+ (d +
d∗ )/I.

Proof : By Proposition 3.3.2, we conclude that there exists a surjective CQG mor-
phism π : QISOI+ (d + d∗ ) → QISOLD . By construction ( as in Proposition 2.1.8 ), the
unitary representation UL of QISOLD preserves each of the HD i , in particular H0 . It is
D
Chapter 3: Quantum group of orientation preserving Riemannian isometries 90

then clear from the definition of I that π induces a surjective CQG morphism (in fact,
a morphism in the category Q0I (d + d∗ )) π 0 : QISOI+ (d + d∗ )/I → QISOLD .
Conversely, if V = (id ⊗ ρI ) ◦ Ud+d∗ is the representation of QISOI+ (d + d∗ )/I on
Hd+d∗ induced by Ud+d∗ (where ρI : QISOI+ (d + d∗ ) → QISOI+ (d + d∗ )/I denotes the
quotient map), then V preserves HD 0 (by definition of I), so commutes with P . Since
0
V also commutes with (d + d∗ )2 , it follows that V must commute with (d + d∗ )2 P0 = L,
that is,
Ve (d∗ d P0 ⊗ 1) = (d∗ dP0 ⊗ 1)Ve .

It is easy to show from the above that αV (which is a C ∗ action on A since αUd+d∗ is so
by assumption) is a smooth isometric action of QISOI+ (d + d∗ )/I in the sense of [30],
with respect to the Laplacian L. This implies that QISOI+ (d + d∗ )/I is a sub-object of
QISOLD in the category Q0LD , and completes the proof. 2

Now we prove that under some further assumptions which are valid for classical
manifolds as well as their Rieffel deformation, one even has the isomorphism QISOLD ∼
=
QISOI+ (d + d∗ ).
We assume the following:

(A) Both the spectral triples (A∞ , H, D) and (A∞ , Hd+d∗ , d + d∗ ) satisfy the
assumptions (1) − (7), so in particular both QISOLD and QISOLD0 exist (here D0 =
d + d∗ ).
(B) For all a, b in A∞ , we have

ha, biH0 = ha, biH0 0 , hdD a, dD biH1 = hdD0 a, dD0 biH1 0 .

D D D D

Remark 3.3.5. For classical compact spin manifolds these assumptions can be verified
by comparing the local expressions of D2 and the ‘Hodge Laplacian’ (D0 )2 in suitable
coordinate charts. In fact, in this case, both these operators turn out to be essentially
same, upto a ‘first order term’, which is relatively compact with respect to D2 or (D0 )2 .
By assumption (B), we observe that the identity map on A∞ extends to a unitary,
0 to H0 . Moreover, we have
say Σ, from HD D0

LD = Σ∗ LD0 Σ,

from which we conclude the following:

Proposition 3.3.6. Under the above assumptions, QISOLD ∼
= QISOLD0 .
We conclude this section with the following result, which identifies the quantum
isometry group QISOLD of [30] as the QISOI+ of a spectral triple, and thus, in some
sense, accommodates the construction of [30] in the framework of the present article.
91 Examples and computations

Theorem 3.3.7. If in addition to the assumptions already made, the spectral triple (
of compact type ) (A∞ , HD0 , D0 ) also satisfies the conditions of Theorem 3.2.17, so that
QISOI+ (D0 ) has a C ∗ -action, then we have the following isomorphism of CQG s:

QISOLD ∼
= QISOI+ (D0 ) ∼
= QISOLD0 .

Proof : By Proposition 3.3.2 we have that QISOLD is a sub-object of QISOI+ (D0 ) in

the category Q0I (D0 ). On the other hand, by Theorem 3.2.17 we have QISOI+ (D0 ) as a
sub-object of QISOLD0 in the category Q0L 0 . Combining these facts with the conclusion
D
of Proposition 3.3.6, we get the required isomorphism. 2

Remark 3.3.8. The assumptions, and hence the conclusions, of this section are valid
also for spectral triples obtained by Rieffel deformation of a classical spectral triple, to
be discussed in details in chapter 4.

3.4 Examples and computations

In this section we compute the quantum group of orientation preserving isometries for
spectral triples on SUµ (2) and C(T2 ). The computations for the Podles’ spheres and
Rieffel deformed manifolds are given in chapter 5 and Chapter 4 respectively.

3.4.1 Equivariant spectral triple on SUµ (2)

We recall from subsection 1.2.4 that by tni,j s, we will denote the (i, j) th matrix element
of the (2n+1) dimensional representation of SUµ (2) and eni,j s will denote the normalized
( with respect to the Haar state h ) tni,j s. We consider the spectral triple on SUµ (2)
constructed by Chakraborty and Pal ( [13] ) and also discussed thoroughly in [18] which
is defined by (A∞ , H, D) where A∞ is the linear span of tnij s, H = L2 (SUµ (2)) and D
is defined by :

D(enij )
= (2n + 1)enij , n 6= i
= −(2n + 1)enij , n = i.

Here, we have a cyclic separating vector 1SUµ (2) , and the corresponding faithful
state is the Haar state h. Thus, we are in the set up of the subsection 3.2.4, and as
ξ = 1, A00 = A∞ in this case. Therefore, an operator commuting with D ( equivalently
with Db ) must keep V l := Span{tl : j = −l, ........l} invariant for all fixed l and i where
i ij
D is the operator as in subsection 3.2.4.
b
Chapter 3: Quantum group of orientation preserving Riemannian isometries 92

In the notation of Corollary 3.2.25 , we have A00 = Span{tli,j : l = 0, 1/2, .........}. =

A∞ in this case. All the conditions of Theorem 3.2.22 and Corollary 3.2.25 are satisfied.
Thus, the universal object of the category Cb exists ( notation as in Corollary 3.2.25 )
and we denote it by Q.b
Before proving the next result, we note !
the following fact. We recall the fundamental
α −µγ ∗
unitary of SUµ (2) given by which is the matrix corresponding to the
γ α∗
1
coproduct ∆ on span {α. − µγ ∗ } as given in subsection 1.2.4. This implies that V−2 1 =
2
1
span {α, γ ∗ } and V 1
2
= span {α∗ , γ}.
2

Lemma 3.4.1. Given a CQG Q with a C ∗ action Φ on A, the following are equivalent
:
1.(Q, Φ) is an element of Obj(C).
b
1/2 1/2
2. The action is linear, in the sense that V−1/2 ( equivalently , V1/2 ) is invariant
1/2
under Φ and the representation obtained by restricting Φ to V1/2 is a unitary represen-
tation.
3. Φ is linear and Haar state preserving.
4. Φ keeps Vil invariant for each fixed l and i.

Proof : 1. ⇒ 2. Since Φ commutes with D, b Φ keeps each of the eigenspaces of

1/2
Db invariant and so in particular preserves V
−1/2 , that is Φ is linear. The condition
(h ⊗ id)Φ = h(·).1 implies the unitarity of the corresponding representation.

2 ⇒ 3. By linearity, write Φ(α) = α ⊗ X + γ ∗ ⊗ Y and Φ(γ ∗ ) = α ⊗ Z + γ ∗ ⊗ W.

Firstly, Φ-invariance of Span{tki,j } for k = 0 and k = 12 follow from the linearity and
the fact that Φ(1) = 1.
Next, we show that Φ keeps Span{t1ij : i, j = −1, 0, 1} invariant.
We recall the explicit form of the matrix ((t1ij )) from [43]:
α∗ 2 −(µ2 + 1)α∗ γ −µγ 2
 
 ∗ ∗
 γ α 1 − (µ2 + 1)γ ∗ γ αγ  .


−µγ ∗ 2 −(µ2 + 1)γ ∗ α α2

By inspection, we see that Φ(Vi1 ) ⊆ Vi1 ⊗ Q for i = −1, 1.
Hence, it is enough to check the Φ-invariance for αγ and 1 − (µ2 + 1)γ ∗ γ.
We have

Φ(αγ)
= (α ⊗ X + γ ∗ ⊗ Y )(α∗ ⊗ Z ∗ + γ ⊗ W ∗ )
= αα∗ ⊗ XZ ∗ + γ ∗ γ ⊗ Y W ∗ + αγ ⊗ XW ∗ + γ ∗ α∗ ⊗ Y Z ∗
93 Examples and computations

µ2 XZ ∗ − Y W ∗
= αγ ⊗ XW ∗ + γ ∗ α∗ ⊗ Y Z ∗ + 1 ⊗ XZ ∗ + (1 − (1 + µ2 )γ ∗ γ) ⊗
1 + µ2
Y W ∗ − µ2 XZ ∗
+1 ⊗
1 + µ2
Y W ∗ − µ2 XZ ∗
= αγ ⊗ XW ∗ + γ ∗ α∗ ⊗ Y Z ∗ + 1 ⊗ (XZ ∗ + ) + (1 − (1 + µ2 )γ ∗ γ)
1 + µ2
µ2 XZ ∗ − Y W ∗
⊗ .
1 + µ2

Thus, comparing coefficient of 1 in Φ(αγ), we can see that it belongs to V01 if and
only if XZ ∗ + Y W ∗ = 0.
In the case of 1 − (1 + µ2 )γ ∗ γ,

Φ(1 − (1 + µ2 )γ ∗ γ)
= 1 ⊗ 1 − (1 + µ2 )(αα∗ ⊗ ZZ ∗ + αγ ⊗ ZW ∗ + γ ∗ α∗ ⊗ W Z ∗ + γ ∗ γ ⊗ W W ∗ )
= 1 ⊗ 1 − (1 + µ2 )(1 − µ2 γ ∗ γ) ⊗ ZZ ∗ − αγ ⊗ (1 + µ2 )ZW ∗ − µα∗ γ ∗ ⊗ (1 + µ2 )
W Z ∗ − (1 + µ2 )γ ∗ γ ⊗ W W ∗
= 1 ⊗ 1 − (1 + µ2 ).1 ⊗ ZZ ∗ + (−1 + 1 − (1 + µ2 )γ ∗ γ) ⊗ −µ2 ZZ ∗ − αγ ⊗ (1 + µ2 )
ZW ∗ − µα∗ γ ∗ ⊗ (1 + µ2 )W Z ∗ + (−1 + 1 − (1 + µ2 )γ ∗ γ) ⊗ W W ∗
= 1 ⊗ (1 − (1 + µ2 )ZZ ∗ + µ2 ZZ ∗ − W W ∗ ) + (1 − (1 + µ2 )γ ∗ γ) ⊗ (−µ2 ZZ ∗ +
W W ∗ ) − αγ ⊗ (1 + µ2 )ZW ∗ − µα∗ γ ∗ ⊗ (1 + µ2 )W Z ∗ .

Comparing the coefficient of 1 in this case, we have the condition 1 − (1 + µ2 )ZZ ∗ +

µ2 ZZ ∗ − W W ∗ = 0, that is, ZZ ∗ + W W ∗ = 1. !
X∗ Z∗
But these conditions follow from the unitarity of the matrix , which
Y ∗ W∗
1/2
is nothing but the matrix corresponding to the restriction of Φ to V1/2 . Thus, Φ keeps
Span{t1ij : i, j = −1, 0, 1} invariant.
Moreover, we claim that by using the recursive relations ( 1.2.18 ), ( 1.2.19 ) and the
l+1/2 l−1/2 l+1/2
multiplication rule ( 1.2.23 ), we obtain that for all l ≥ 3/2, Φ(Vi ) ⊆ Vi ⊕Vi .
l+ 1 1
We prove this for ti,j 2 , − l + 2 ≤ i ≤ l − 12 , j ≤ l only, as the proofs of the others are
exactly similar. We have

l+ 1
Φ(ti,j 2 )
= c(l, i, j)Φ(α)Φ(tli+ 1 ,j+ 1 ) + c0 (l, i, j)Φ(γ)Φ(tli− 1 ,j+ 1 )
2 2 2 2
1 1
0
= c(l, i, j)Φ(t− 1 ,− 1 )Φ(tli+ 1 ,j+ 1 )
2
+ c (l, i, j)Φ(t 2
1
,− 12
)Φ(tli− 1 ,j+ 1 )
2 2 2 2 2 2 2
1 1 1
∈ Span{t− 1 ,k .tli+ 1 ,m , t
2 2
1
,k
.tli− 1 ,m : k = ± , m = −l, ...l} ⊗ Q
2 2 2 2 2
Chapter 3: Quantum group of orientation preserving Riemannian isometries 94

l− 21 l+ 12
⊆ Vi ⊗ Q + Vi ⊗ Q.

Using these observations, we conclude that Φ maps Span{tlij : l ≥ 1/2} into itself.
So, in particular, Ker(h) = Span{Vil : i = −l, , , , l, l ≥ 1/2} is invariant under Φ
which ( along with Φ(1) = 1 ) implies that Φ preserves h.

3. ⇒ 4.
We proceed by induction. The induction hypothesis holds for l = 21 since linearity
means that span {α, γ ∗ } is invariant under Φ and hence Span {α∗ , γ} is also invariant.
The case for l = 1 can be checked by inspection as in the proof of 2 ⇒ 3. Consider the
induction hypothesis that Φ keeps Vik invariant for all k, i with k ≤ l. From the proof of
l+1/2 l−1/2 l+1/2
2 ⇒ 3 we also have for all l ≥ 23 , Φ(Vi ) ⊆ Vi ⊕ Vi , by using linearity only.
1
l− 2 l+ 21
e leaves invariant the Hilbert Q module (V
Thus, Φ ⊕V )⊗Q, and is a unitary there
i i
1
e leaves invariant V l− 2 ⊗ Q by the induction
since Φ is Haar-state preserving. Since Φ i
l+ 12
hypothesis, it must keep its orthocomplement, Vi invariant as well.

4. ⇒ 3.
The fact that Φ keeps Vil invariant for l = 1/2 will imply that Φ is linear. The proof
of Haar state preservation is exactly the same as in 2 ⇒ 3.

4 ⇒ 1.
That Φ preserves the Haar state follows from arguments used in the proof of the
implication 2 ⇒ 3.. Since A00 = Span{tlij : l ≥ 0, i, j = −l, .......l}, and Φ keeps each Vil
invariant, it is obvious that Φ(A00 ) ⊆ A00 ⊗alg Q0 and ΦD b = (D b ⊗ id)Φ.
2
By Lemma 3.4.1, we have identified the category Cb with the category of CQG having
C∗ actions on SUµ (2) satisfying condition 3. of Lemma 3.4.1. Let the universal object
of this category be denoted by (Q,b Γ).
Then by linearity we can write:

Γ(α) = α ⊗ A + γ ∗ ⊗ B,

Γ(γ ∗ ) = α ⊗ C + γ ∗ ⊗ D.

Now we shall exploit the fact that Γ is a ∗-homomorphism to get relations satisfied
b is generated as a C ∗ algebra by the elements A, B, C, D.
by A, B, C, D where Q

Lemma 3.4.2.
A∗ A + CC ∗ = 1, (3.4.1)
95 Examples and computations

A∗ A + µ2 CC ∗ = B ∗ B + DD∗ , (3.4.2)

A∗ B = −µDC ∗ , (3.4.3)

B ∗ A = −µCD∗ . (3.4.4)

Proof : The proof follows from the relation ( 1.2.10 ) by comparing coefficients of
1, γ ∗ γ, α∗ γ ∗
and αγ respectively. 2

Lemma 3.4.3.
AA∗ + µ2 CC ∗ = 1, (3.4.5)

BB ∗ + µ2 DD∗ = µ2 .1, (3.4.6)

BA∗ = −µ2 DC ∗ . (3.4.7)

Proof : From the equation ( 1.2.11 ) by equating coefficients of 1 and α∗ γ ∗ , we get

respectively ( 3.4.5 ) and ( 3.4.7 ) whereas ( 3.4.6 ) is obtained by equating coefficients
of γ ∗ γ and using ( 3.4.5 ).
2

Lemma 3.4.4.
C ∗ C = CC ∗ , (3.4.8)

(1 − µ2 )C ∗ C = D∗ D − DD∗ , (3.4.9)

C ∗ D = µDC ∗ . (3.4.10)

Proof : The proof follows from the equation ( 1.2.12 ) by comparing the coefficients
of 1, γ ∗ γ, α∗ γ ∗ , respectively. 2

Lemma 3.4.5.
− µ2 AC ∗ + BD∗ − µD∗ B + µC ∗ A = 0, (3.4.11)

AC ∗ = µC ∗ A, (3.4.12)

BC ∗ = C ∗ B, (3.4.13)

AD∗ = D∗ A. (3.4.14)

Proof : The proof follows from the equation ( 1.2.13 ) comparing the coefficients of
γ ∗ γ, 1, α∗ γ ∗
and αγ respectively. 2

Lemma 3.4.6.
AC = µCA, (3.4.15)

BD = µDB, (3.4.16)
Chapter 3: Quantum group of orientation preserving Riemannian isometries 96

AD − µCB = DA − µ−1 BC. (3.4.17)

Proof : The proof follows from ( 1.2.14 ) from the coefficients of α2 , γ ∗2 , γ ∗ α respec-
tively. 2

Now we consider the antipode, say κ.

From the condition (h ⊗ id)Γ(a) = h(a).1, we see that Γ induces a unitary represen-
tation of the compact quantum group via Γ(a e ⊗ q) = Γ(a)(1 ⊗ q).
Now, the restriction of this unitary representation to the orthonormal
! set
q 2 A µC
{ 1+µ
p
µ2
α, 1 + µ2 γ ∗ } is given by the matrix : −1
.
µ B D
p p
Similarly, with respect to the orthonormal!set { 1 + µ2 α∗ , 1 + µ2 γ}, this repre-
A∗ C ∗
sentation is given by the matrix: .
B ∗ D∗
Thus, we have:
κ(A) = A∗ , κ(D) = D∗ , κ(C) = µ−2 B ∗ , κ(B) = µ2 C ∗ , κ(A∗ ) = A, κ(C ∗ ) =
B, κ(B ∗ ) = C, κ(D∗ ) = D.

Lemma 3.4.7.
AB = µBA, (3.4.18)

CD = µDC, (3.4.19)

BC ∗ = C ∗ B. (3.4.20)

Proof : The relations ( 3.4.18 ), ( 3.4.19 ), ( 3.4.20 ) follow by applying κ to the

equations ( 3.4.15 ), ( 3.4.16 ) and ( 3.4.13 ) respectively.

Lemma 3.4.8. There exists a ∗-homomorphism φ : Uµ (2) → Q

b defined by φ(u11 ) =
A, φ(u12 ) = µC, φ(u21 ) = µ−1 B, φ(u22 ) = D.

Proof : It is enough to check that the defining relations of Uµ (2) are satisfied.
1. φ(u11 u12 ) = φ(µu12 u11 ) ⇔ φ(u11 )φ(u12 ) = µφ(u12 )φ(u11 ) ⇔ A(µC) =
µ(µC)A ⇔ AC = µCA which is the same as ( 3.4.15 ).
2. φ(u11 u21 ) = φ(µu21 u11 ) ⇔ A(µ−1 B) = µ(µ−1 B)A ⇔ AB = µBA which is the
same as equation ( 3.4.18 ).
3. φ(u12 u22 ) = φ(µu22 u12 ) ⇔ µCD = µD(µC) ⇔ CD = µDC which is the same as
equation ( 3.4.19 ).
4. φ(u21 u22 ) = φ(µu22 u21 ) ⇔ µ−1 BD = µDµ−1 B ⇔ BD = µDB which is the same
as equation ( 3.4.16 ).
5. φ(u12 u21 ) = φ(u21 u12 ) ⇔ µCµ−1 B = µ−1 BµC ⇔ CB = BC.
97 Examples and computations

Now, BC ∗ = C ∗ B follows from equation ( 3.4.20 ). But by ( 3.4.8 ), C is normal,

which implies BC = CB.
6. φ(u11 u22 − u22 u11 ) = (µ − µ−1 )φ(u12 u21 ) ⇔ AD − DA = (µ − µ−1 )µCµ−1 B.
From ( 3.4.17 ), we have AD − DA = µCB − µ−1 BC = (µ − µ−1 )CB, using
BC = CB.
2

Lemma 3.4.9. The equations ( 3.4.1 ) - ( 3.4.17 ) are true when A, B, C, D are replaced
by u11 , µu21 , µ−1 u12 and u22 respectively.

Proof : We check some of the relations ( 3.4.1 ) - ( 3.4.17 by using the facts that Dµ
is a central element of Uµ (2), κ(uij ) = u∗ji ( ( 1.2.8 ) ), Proposition 1.2.26, the equations
( 1.2.2 ) - ( 1.2.7 ) and ( 1.2.9 ). The proofs of the others are exactly similar.
∗
Proof for ( 3.4.1 ) that is, u11 ∗ u11 + (µ−1 u12 )(µ−1 u12 ) = 1.

u∗11 u11 + µ−2 u12 u∗12 = u22 Dµ−1 u11 + µ−2 u12 (−µu21 Dµ−1 )
= (u22 u11 − µ−1 u12 u21 )Dµ−1 = Dµ Dµ−1 = 1.

∗
Proof for ( 3.4.2 ) that is, u∗11 u11 +µ2 (µ−1 u12 )(µ−1 u12 ) −((µu21 )∗ µu21 +u22 u∗22 ) = 0.

∗
u∗11 u11 + µ2 (µ−1 u12 )(µ−1 u12 ) − ((µu21 )∗ µu21 + u22 u∗22 )
= κ(u11 )u11 + u12 κ(u21 ) − (µ2 κ(u12 )u21 + u22 κ(u22 ))
= (u22 u11 − µu12 u21 )Dµ−1 − (−µu12 u21 + u22 u11 )Dµ−1
= 0.

Proof for ( 3.4.6 ) that is, µ2 u21 u∗21 + µ2 u22 u∗22 − µ2 .1 = 0.

µ2 u21 u∗21 + µ2 u22 u∗22 − µ2 .1

= µ2 (u21 κ(u12 ) + u22 κ(u22 ) − 1)
= µ2 (u21 (−µ−1 u12 Dµ−1 ) + u22 u11 Dµ−1 − 1)
= µ2 ((u22 u11 − µ−1 u21 u12 )Dµ−1 − 1)
= µ2 (Dµ Dµ−1 − 1)
= 0.
Chapter 3: Quantum group of orientation preserving Riemannian isometries 98

Proof for ( 3.4.7 ) that is, u21 u∗11 + u22 u∗12 = 0.

u21 u∗11 + u22 u∗12

= u21 κ(u11 ) + u22 κ(u21 )
= u21 u22 Dµ−1 − µu22 u21 Dµ−1
= (u21 u22 − µu22 u21 )Dµ−1
= 0.

Proof for ( 3.4.9 ) that is, (1 − µ2 )u∗12 u12 − µ2 (u∗22 u22 − u22 u∗22 ) = 0.

(1 − µ2 )u∗12 u12 − µ2 (u∗22 u22 − u22 u∗22 )

= (1 − µ2 )κ(u21 )u12 − µ2 (κ(u22 )u22 − u22 κ(u22 ))
= −µ(1 − µ2 )(u21 u12 Dµ−1 ) − µ2 (u11 u22 Dµ−1 − u22 u11 Dµ−1 )
= −µ[(1 − µ2 )(u21 u12 Dµ−1 ) − µ(u22 u11 Dµ−1 − u11 u22 Dµ−1 )]
= −µ[(1 − µ2 )u21 u12 Dµ−1 − µ(µ−1 − µ)u12 u21 Dµ−1 ]
= −µ(1 − µ2 )(u12 u21 − u12 u21 )Dµ−1
= 0.

Proof for ( 3.4.10 ) that is, u∗12 u22 − µu22 u∗12 = 0.

u∗12 u22 − µu22 u∗12

= κ(u21 )u22 − µu22 κ(u21 )
= µ2 u22 u21 Dµ−1 − µu21 u22 Dµ−1
= µ(µu22 u21 − u21 u22 )Dµ−1
= 0.

Proof for ( 3.4.11 ) that is, −µu11 u∗12 + µu21 u∗22 − µ2 u∗22 u21 + u∗12 u11 = 0.

−µu11 u∗12 + µu21 u∗22 − µ2 u∗22 u21 + u∗12 u11

= −µu11 κ(u21 ) + µu21 κ(u22 ) − µ2 κ(u22 )u21 + κ(u21 )u11
= −µu11 (−µu21 Dµ−1 ) + µu21 (u11 Dµ−1 ) − µ2 u11 u21 Dµ−1 − µu21 u11 Dµ−1
= µ2 (u11 u21 − u11 u21 )Dµ−1 + µ(u21 u11 − u21 u11 )Dµ−1
= 0.
99 Examples and computations

Proof for ( 3.4.17 ) that is, u11 u22 − µ(µ−1 u12 )µu21 = u22 u11 − µ−1 (µu21 )(µ−1 u12 ).

u11 u22 − µ(µ−1 u12 )µu21 − u22 u11 + µ−1 (µu21 )(µ−1 u12 )
= u11 u22 − µu12 u21 − u22 u11 + µ−1 u21 u12
= 0.

Lemma 3.4.10. There is a C ∗ action Ψ of Uµ (2) on SUµ (2) such that (Uµ (2), Ψ) is
an object of Obj(C)
b and Ψ is given by :

Ψ(α) = α ⊗ u11 + γ ∗ ⊗ µu21 ,

Ψ(γ ∗ ) = α ⊗ µ−1 u12 + γ ∗ ⊗ u22 .

Proof : The homomorphism conditions are exactly the conditions ( 3.4.1 ) - ( 3.4.17
) with A, B, C, D replaced by u11 , µu21 , µ−1 u12 and u22 respectively which are true by
Lemma 3.4.9.
1/2
Clearly, Ψ keeps V−1/2 invariant and the corresponding representation is a unitary.
It follows from Lemma 3.4.1 that (Uµ (2), Ψ) is an object of C. b
2

Corollary 3.4.11. There exists a surjective CQG morphism from Q

b to Uµ (2) sending
A, µC, µ−1 B, and D to u11 , u12 , u21 and u22 respectively.

b∼
Theorem 3.4.12. We have Q ^+ (D) ∼
= Uµ (2) and hence QISO = Uµ (2) ∗ C(T).

Proof : The first part follows from Lemma 3.4.8 and Corollary 3.4.11 and the second
part follows from Theorem 3.2.22. 2

3.4.2 A commutative example : spectral triple on T2

We consider the spectral triple (A∞ , H, D) 2 ∞ ∞ 2 2 2
! on T given by A = C (T ), H = L (T )⊕
0 d1 + id2
L2 (T2 ) and D = ,
d1 − id2 0
where we view C(T2 ) as the universal C ∗ algebra generated by two commuting
unitaries U and V, and d1 and d2 are derivations on A∞ defined by :

d1 (U ) = U, d1 (V ) = 0, d2 (U ) = 0, d2 (V ) = V. (3.4.21)
Chapter 3: Quantum group of orientation preserving Riemannian isometries 100

The vectors e1 = (1, 0) and e2 = (0, 1) form an orthonormal basis of the eigenspace
corresponding to the eigenvalue zero.
The Laplacian in the sense of chapter 2 exists in this case, and is given by
L(U m V n ) = −(m2 + n2 )U m V n . We recall that we denote the quantum isometry group
from the Laplacian L in the sense of [30] by QISOLD .

Lemma 3.4.13. Let (Q, e W ) be an object of Q0 (D). Then the ∗-homomorphism α = αW

must be of the following form:

α(U ) = U ⊗ z1 , (3.4.22)

α(V ) = V ⊗ z2 , (3.4.23)

where z1 , z2 are two commuting unitaries.

Proof: We denote the the maximal Woronowicz C ∗ subalgebra of Q e which acts on

C(T2 ) faithfully by Q.
We observe that D2 (aei ) = L(a)ei for i = 1, 2. Now, W commutes with D implies
that W commutes with D2 as well. Using this, we can show that (L ⊗ id)α(a)ei =
αL(a)ei , i = 1, 2. As the pair {e1 , e2 } is together separating for C(T2 ), we conclude that
α commutes with the Laplacian L. Therefore, Q is a quantum subgroup of QISOLD .
From Theorem 2.2.17, we conclude that QISOLD = C(T2 >
(Z22 >
Z2 )). Thus Q must
be of the form C(G) for a classical subgroup G of the orientation preserving isometry
group of T2 , which is T2 itself and whose (co )action is given by U 7→ U ⊗ U and
V 7→ V ⊗ V. 2
+
Theorem 3.4.14. The universal CQG QISO ^ (C ∞ (T2 ), H, D) exists and is isomorphic
with C(T2 ) ∗ C(T) ∼
= C ∗ (Z2 ∗ Z) (as a CQG). Moreover, QISO+ of this spectral triple
is isomorphic with C(T2 ).

e W ) be an object in Q0 (D) as in Lemma 3.4.13. Since {e1 , e2 } is an

Proof: Let (Q,
orthonormal basis for an eigenspace of D, we must have

W (e1 ) = e1 ⊗ q11 + e2 ⊗ q12 , (3.4.24)

W (e2 ) = e1 ⊗ q21 + e2 ⊗ q22 , (3.4.25)

for some qij in Q.

e
We now make use of the equation (D ⊗ id)W
f (U e1 ⊗ 1) = W
f (D ⊗ id)(U e1 ⊗ 1). Let
101 Examples and computations

z1 , z2 are as in Lemma 3.4.13. We compute

(D ⊗ id)W
f (U e1 ⊗ 1)

= (D ⊗ id)(α(U )W
f (e1 ⊗ 1))

= (D ⊗ id)(U ⊗ z1 )(e1 ⊗ q11 + e2 ⊗ q12 )

= (D ⊗ id)(U e1 ⊗ z1 q11 + U e2 ⊗ z1 q12 )
= U e2 ⊗ z1 q11 + U e1 ⊗ z1 q12 .

On the other hand,

f (D ⊗ id)(U e1 ⊗ 1)
W
f (U e2 ⊗ 1)
= W
= W f∗W
f (U ⊗ id)W f (e2 ⊗ 1)
f (e2 ⊗ 1)
= α(U )W
= (U ⊗ z1 )(e1 ⊗ q21 + e2 ⊗ q22 )
= U e1 ⊗ z1 q21 + U e2 ⊗ z1 q22 .

By comparing coefficients of U e1 and U e2 in the both sides of the equality (D ⊗

id)W (U e1 ) = W DU e1 , we have,
z1 q12 = z1 q21 (3.4.26)

and
z1 q11 = z1 q22 . (3.4.27)

Since z1 is a unitary, we have q11 = q22 and q12 = q21 .

Similarly, from the relation (D ⊗ I)W (V e1 ) = W DV e1 , we have q12 = −q21 , q22 =
q11 .
By the above two sets of relations, we obtain :
q12 = q21 = 0, q11 = q22 = q!( say ).
q11 q12
But the matrix is a unitary in M2 (Q),
e so q is a unitary.
q21 q22
Moreover, we note that W (aei ) = α(a)W (ei ) for all a in C ∞ (T2 ). Using Lemma
3.4.13 and the above observations, we deduce that any CQG which has a unitary repre-
sentation commuting with the Dirac operator is a quantum subgroup of C(T2 ) ∗ C(T).
On the other hand, C(T2 ) ∗ C(T) has a unitary representation commuting with D,
given by the formulae ( 3.4.22 ) - ( 3.4.25 ) taking q12 = q21 = 0, q11 = q22 = q 0 where
q 0 is the generator of C(T) and z1 , z2 to be the generator of C(T2 ). This completes the
proof. 2
Chapter 3: Quantum group of orientation preserving Riemannian isometries 102

2
Remark 3.4.15. The canonical grading on C(T ! ) is given by the operator (id ⊗ γ) on
0 1
L2 (T2 ⊗ C2 ) where γ is the matrix .
−1 0
The representation of C(T2 ) ∗ C(T) clearly commutes with the grading operator and
hence is isomorphic with QISO(C(T
^ 2 ), L2 (T2 ⊗ C2 ), D, γ).

Remark 3.4.16. This example shows that the conditions of Theorem 3.2.22 are not
^+ .
necessary for the existence of QISO

3.5 QISO+ for zero dimensional manifolds

3.5.1 Inductive limit construction for quantum isometry groups
In this section we use the limiting construction for an inductive system of compact
quantum groups ( Lemma 1.2.25 ) and give an application for quantum isometry groups
which is fundamental for the results of the next section.
The next theorem connects the inductive construction done in Lemma 1.2.25 with
some specific quantum isometry groups.

Theorem 3.5.1. Suppose that A is a C ∗ -algebra acting on a Hilbert space H and that
D is a (densely defined) self adjoint operator on H with compact resolvent, such that
D has a one-dimensional eigenspace spanned by a vector ξ which is cyclic and separat-
ing for A. Let (A∞ ∗
n )n∈IN be an increasing net of a unital -subalgebras of A and put
A∞ = n∈IN A∞ ∞ is dense in A and that for each a ∈ A∞ the commu-
S
n . Suppose that A
tator [D, a] is densely defined and bounded. Additionally put Hn = A∞ n ξ, let Pn denote
the orthogonal projection on Hn and assume that each Pn commutes with D. Then
each (A∞n , Hn , D|Hn ) is a spectral triple satisfying the conditions of Theorem 3.2.13,
there exist natural compatible CQG morphisms π ^+ (A∞ , H , D| ) →
: QISO
m,n m m Hm
^+ (A∞ , H , D| ) (n, m ∈ IN , m ≤ n) and
QISO n n Hn

^+ (A∞ , H, D) = lim QISO

QISO ^+ (A∞ , H , D| ).
n n Hn
n∈IN

^+ by QISO+ .
Similar conclusions hold if we replace everywhere above QISO
^+ only, since the proof for
Proof: We prove the assertion corresponding to QISO
QISO+ follows by very similar arguments. Let us denote QISO ^+ (A∞ , H , D ) by
n n n
Sn and the corresponding unitary representation (in Hn ) by Un . Let us denote the
category of compact quantum groups acting by orientation preserving isometries on
(A∞ ∞
n , Hn , D|Hn ) and (A , H, D) respectively by Qn and Q.
103 QISO+ for zero dimensional manifolds

Since Un is a unitary which commutes with Dn ≡ D|Hn and hence preserves the
eigenspaces of Dn , it restricts to a unitary representation of Sn on each Hm for m ≤ n.
In other words, (Sn , Un |Hm ) ∈ Obj(Qm ), and by the universality of Sm there exists a
compact quantum group morphism, say, πm,n : Sm → Sn such that (id ⊗ πm,n )Um |Hm =
Un |Hm .
Let p ≤ m ≤ n. Then we have (id ⊗ πm,n πp,m )Up |Hp = Un |Hp . It follows by the
uniqueness of the map πp,n that πp,n = πm,n πp,m , that is (Sn )n∈IN forms an inductive
system of compact quantum groups satisfying the assumptions of Lemma 1.2.25. Denote
by S∞ the inductive limit CQG obtained in that lemma, with πn,∞ : Sn → S denoting
the corresponding CQG morphisms. The family of formulas U |Hn := (id ⊗ πn,∞ ) ◦ Un
combine to define a unitary representation U of S∞ on H. It is also easy to see from
the construction that U commutes with D. This means that (S∞ , U ) ∈ Obj(Q), hence
there exists a unique surjective CQG morphism from S := QISO ^+ (A∞ , H, D) to S
∞
identifying S∞ as a quantum subgroup of S.
The proof will now be complete if we can show that there is a surjective CQG
morphism in the reverse direction, identifying S as a quantum subgroup of S∞ . This
can be deduced from Lemma 1.2.25 by using the universality property of the inductive
^+ (A∞ , H, D)
limit. Indeed, for each n ∈ IN the unitary representation, say Vn , of QISO
restricts to Hn and commutes with D on that subspace, thus inducing a CQG morphism
^+ (A∞ , H , D ) into S. The family of morphisms (ρ )
ρn from Sn = QISO n n n n n∈IN satisfies
the compatibility conditions required in Lemma 1.2.25. It remains to show that the
induced CQG morphism ρ∞ from S∞ into S is surjective. By the faithfulness of the
representation V of QISO^+ (A∞ , H, D), we know that the span of matrix elements
corresponding to all Vn forms a norm-dense subset of S. As the range of ρn contains the
matrix elements corresponding to Vn = V |Hn , the proof of surjectivity of ρ∞ is finished.
2

The assumptions of the theorem might seem very restrictive. In the next section
however we will describe a natural family of spectral triples on AF -algebras, constructed
in [16], for which we have exactly the situation as above.

3.5.2 Quantum isometry groups for spectral triples on AF algebras

We first recall the construction of natural spectral triples on AF algebras due to
E. Christensen and C. Ivan ( [16] ). Let A be a unital AF C ∗ -algebra, the norm closure
of an increasing sequence (An )n∈IN of finite dimensional C ∗ -algebras. We always put
A0 = C1A , A∞ = ∞
S
n=1 An and assume that the unit in each An is the unit of A.
Suppose that A is acting on a Hilbert space H and that ξ ∈ H is a separating and
cyclic unit vector for A. Let Pn denote the orthogonal projection onto the subspace
Chapter 3: Quantum group of orientation preserving Riemannian isometries 104

Hn := An ξ of H and write Q0 = P0 = PCξ , Qn = Pn − Pn−1 for n ∈ IN . There exists a

(strictly increasing) sequence of real numbers (αn )∞
n=1 such that the self adjoint operator
∞
P
D = n∈IN αn Qn yields a spectral triple (A , H, D). Due to the existence of a cyclic
and separating vector the quantum group of orientation preserving isometries exists by
Theorem 3.2.22.
In [16], the following fact was also observed:

Proposition 3.5.2. If A is infinite-dimensional and p > 0 then one can choose (αn )∞ n=1
in such a way that the spectral triple is p-summable. For this reasons, the spectral triple
should be thought of as 0-dimensional noncommutative manifolds.

Note that for each n ∈ IN by restricting we obtain a (finite-dimensional) spec-

tral triple (An , Hn , D|Hn ). As we are precisely in the framework of Theorem 3.5.1,
to compute QISO+ (A∞ , H, D) we need to understand the quantum isometry groups
QISO+ (A∞ n , Hn , D|Hn ) and embeddings relating them. To simplify the notation we
will write Sn := QISO+ (An , Hn , D|Hn ).
We begin with some general observations.

Lemma 3.5.3. Let QU An ,ωξ denote the universal quantum group acting on An and
preserving the (faithful) state on An given by vector ξ (see [60]). There exists a CQG
morphism from QU An ,ωξ to Sn .

Proof: The proof is based on considering the spectral triple given by (An , Hn , Dn0 ),
where Dn0 = Pn −P0 . It is then easy to see that QISO+ (An , Hn , Dn0 ) is isomorphic to the
universal compact quantum group acting on An and preserving ωξ . On the other hand
universality assures the existence of the CQG morphism from QISO+ (An , Hn , Dn0 ) to
Sn . 2

Lemma 3.5.4. Assume that each An is commutative, An = Ckn , n ∈ IN . There exists

a CQG morphism from QU kn to Skn , where QU kn denotes the universal quantum group
acting on kn points ( [60]).

Proof: We observe that for any measure µ on the set {1, . . . , kn } which has full
support there is a natural CQG morphism from QU kn to QU Ckn ,µ . In case when µ is
uniformly distributed, we simply have QU Ckn ,µ = QU kn , as follows from Lemma 1.2.33.
2

Let αn : An → An ⊗ Sn denote the universal action (on the n-th level). Then we
have the following important property, being the direct consequence of the Theorem
3.5.1. We have
αn+1 (An ) ⊂ An ⊗ Sn+1 (3.5.1)
105 QISO+ for zero dimensional manifolds

(where we identified An with a subalgebra of An+1 ) and Sn is generated exactly by these

coefficients of Sn+1 which appear in the image of An under αn+1 . This in conjunction
with the previous lemma suggests the strategy for computing relevant quantum isometry
groups inductively. Suppose that we have determined the generators of Sn . Then Sn+1
is generated by generators of Sn and these of the QU An ,ωξ , with the only additional
relations provided by the equation ( 3.5.1 ).
This will be used below to determine the concrete form of relations determining Sn
for the commutative AF algebras.
Before stating the next result, we fix some notations. Let An be a sequence of
commutative finite dimensional C ∗ algebras as above. Let An = C(Xn ) where Xn =
{x1 , x2 , ..., xm }. Dualizing the embedding from An to An+1 , there is a surjective map,
say fn+1,n from Xn+1 to Xn . Let li denote the cardinality of the set {x ∈ Xn+1 :
fn+1,n (x) = xi }. Thus the embedding of An into An+1 is determined by the sequence
{li : i = 1, 2, ..., m}. We note that the cardinality of Xn+1 equals m
P
i=1 li . Moreover, a
basis of An+1 is given by {ei,ri : ri ∈ {1, 2, ...li }} where ei,ri is the indicator function
of an element y of Xn+1 such that fn+1,n (y) = xi and y is the ri th element in Xn+1
−1
belonging to fn+1,n {xi }.

Lemma 3.5.5. Let A be a commutative AF algebra. Suppose that An is isomorphic

to Cm and the embedding of An into An+1 is given by a sequence (li )m 0
i=1 . Let m =
Pm
i=1 li . Suppose that the ‘copy’ of QU m in Sn is given by the family of projections
ai,j ( i, j ∈ {1, . . . m} ) and that the ‘copy’ of QU m0 in Sn+1 is given by the family of
projections a(i,ri ),(j,sj ) (i, j ∈ {1, . . . , m}, ri ∈ {1, . . . , li }, sj ∈ {1, . . . , lj }). Then the
formula ( 3.5.1 ) is equivalent to the following system of equalities:

li
X
ai,j = a(i,ri ),(j,sj ) (3.5.2)
ri =1

for each i, j ∈ {1, . . . , m}, sj ∈ {1, . . . , lj }.

Proof: We have (for the universal action α : An → An ⊗ Sn )

m
X
α(eei ) = eej ⊗ ai,j ,
j=1

where by eei we denote the image of the basis vector ei ∈ An in An+1 . As eej =
Plj
rj =1 e(j,sj ) ,

li li X lj
m X
X X
α(eei ) = α(ei,ri ) = e(j,sj ) ⊗ a(i,ri ),(j,sj ) .
ri =1 ri =1 j=1 sj =1
Chapter 3: Quantum group of orientation preserving Riemannian isometries 106

On the other hand we have

lj
m X
X
α(eei ) = e(j,sj ) ⊗ ai,j ,
j=1 sj =1

and the comparison of the formulas above yields exactly ( 3.5.2 ). 2

One can deduce from the above lemma the exact structure of generators and relations
between them for each Sn associated with a commutative AF algebra. To be precise, if
An = Ckn for some kn ∈ IN , then the quantum isometry group Sn is generated as a unital
C ∗ -algebra by the family of self adjoint projections ni=1 {aαi ,βi : αi , βi = 1, · · · , ki }
S

such that for each fixed i = 1, . . . , n the family {a(αi ,βi ) : αi , βi = 1, · · · , ki } satisfies
the relations of QU kn and the additional relations between a(αi ,βi ) and a(αi+1 ,βi+1 ) for
i ∈ {1, . . . , n − 1} are given by the formulas ( 3.5.2 ), after suitable reinterpretation of
indices according to the multiplicities in the embedding of Cki into Cki+1 .
Chapter 4

Quantum isometry groups for

Rieffel deformed manifolds

^+ by proving
In this chapter, we give a general scheme for computing QISOL and QISO R
L ^ +
that QISO , ( respectively QISO ) of a deformed noncommutative manifold coincides
R
with (under reasonable assumptions) a similar deformation of QISOL , ( respectively
^+ ) of the original manifold.
QISO R

4.1 Deformation of spectral triple

We recall from Chapter 1 the generalities of CQG s and Hopf algebras, in particular,
the dense unital Hopf ∗-subalgebra S0 of a CQG S generated by the matrix elements
of the irreducible unitary representations, the Sweedler convention for CQG action, as
well as the convolutions f / c, c . f and f  g for functionals f, g on S and c in S.
Moreover, given an action γ : B → B ⊗ S of the compact quantum group (S, ∆) on a
unital C ∗ -algebra A, the dense, unital ∗-subalgebra of A on which γ becomes an action
by the Hopf ∗-algebra S0 is going to be denoted by A0 .
A word of caution: The algebra A0 should not be confused with the Rieffel deformed
∗
C algebra AJ in the case J = 0, for which we simply write A.
Let (S, ∆S ) be a compact quantum group. We also adopt the convention of calling
a vector space M an S co-module if it is an algebraic S0 co-module in the sense of
definition 1.2.13.

Before introducing the set up in which we are going to work, we prove the following
well known fact for the sake of completeness.

Proposition 4.1.1. Let E be a Banach space and G a second countable Lie group with
a strongly continuous action α on E such that kαg (x)k = kxk for all g in G and for all

107
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 108

x in E. Then E ∞ = {e ∈ E : g → αg (e) is C ∞ } is norm dense in A.

Proof: For a compactly supported continuous function f on G and a in E, we will

R
denote by α(f )(a) the norm convergent integral G f (h)αh (a)dh where dh denotes a left
invariant Haar measure on G. Then, it can be seen that αg (α(f )a) = G f (g −1 h)αh (a)dh.
R

Thus, for f in Cc∞ (G), α(f )(a) is in E ∞ . Now, for any  > 0, we choose a small enough
neighbourhood U of identity of G, such that kαg (a) − ak ≤  for all g in U. Next, we
choose f in Cc∞ (G) with f ≥ 0, G f dh = 1 and supp(f ) ⊆ U. Then,
R

kα(f )(a) − ak
Z Z

= f (g)αg (a)dg − a f (g)dg

ZG G


= f (g)(αg (a) − a)dg

G
Z
≤ f (g) kαg (a) − ak dg
G
≤ .

This shows that E ∞ is dense in E. 2

Lemma 4.1.2. Let A be a C ∗ algebra with a strongly continuous action α of G as

above. Then A∞ is closed under holomorphic functional calculus. Let φ be a positive
linear map from A∞ to another C ∗ algebra B. Then, for any self adjoint element x in
A∞ , kφ(x)k ≤ kxk φ(1).

Proof: The first fact is quite well known. We refer to [52] for a proof. For the second
part, let x be a self adjoint element of A∞ . Then, y = (1 + ) kxk − x is a positive
and invertible element ( since its spectrum does not contain zero ) of A∞ , which being
closed under holomorphic functional calculus, is closed under taking square root of an
1 1 1
invertible element. Thus, y 2 belongs to A∞ and therefore φ(y) = φ((y 2 )∗ y 2 ) ≥ 0. This
proves the Lemma. 2

Let (A, Tn , β) be a C ∗ dynamical system, that is, A is endowed with a strongly

continuous action of Tn by ∗ automorphisms. Moreover, π0 : A → B(H) be a faithful
representation, where H is a separable Hilbert space.
Let A∞ be the smooth algebra corresponding to the Tn action β.
Assume now that we are given a spectral triple (A∞ , π0 , H, D) of compact type.
Suppose that D has eigenvalues {λ0 , λ1 , .........} and Vi denotes the (finite dimensional)
eigenspace of λi and let S00 denote the linear span of {Vi : i = 0, 1, 2, ..}.
Suppose, furthermore, that there exists a compact abelian Lie group T fn , with a
covering map γ : T fn → Tn . The Lie algebra of both Tn and T fn are isomorphic with
109 Deformation of spectral triple

Rn and we denote by e and ee respectively the corresponding exponential maps, so that

e(u) = e(2πiu), u ∈ Rn and γ(e e(u)) = e(u). By a slight abuse of notation we shall
n
denote the R -action βe(u) by βu .
Assumption:
There exists a strongly continuous unitary representation Vg̃ , g̃ ∈ T fn of T
fn on H such
that
(a) Vg̃ D = DVg̃ for all g̃,
(b) Vg̃ π0 (a)Vg̃ −1 = π0 (βg (a)), where a belongs to A, g̃ belongs to T
fn , and g = γ(g̃).

We shall now show that we can ‘deform’ the given spectral triple along the lines
of [19]. For each J, the map πJ : A∞ → Lin(H∞ ) (where H∞ is the smooth subspace
corresponding to the representation V and Lin(V) denotes the space of linear maps on
a vector space V) defined by
Z Z
πJ (a)s ≡ a ×J s := βJu (a)βev (s)e(u.v)dudv

extends to a ∗-representation of the C ∗ -algebra A∞ in B(H) where β fv = Vee(v) ( which

∞
clearly maps H into H ). ∞

We can extend the action of Tn on the C ∗ subalgebra A1 of B(H) generated by

π0 (A), {eitD : t ∈ R} and elements of the form {[D, a] : a ∈ A∞ } by βg (X) = VgeXVge−1
for all X in A1 where by an abuse of notation, we denote the action by the same symbol
β. Let A∞ 1 denote the smooth vectors of A1 with respect to this action. We note that
for all a in A∞ ∞
1 , [D, a] belongs to A1 .

Lemma 4.1.3. β is a strongly continuous action (in the C ∗ -sense) of Tn on A1 and

hence for all X in A1 ∞ , πJ (X) defined by
Z Z
πJ (X)s = βJu (X)βev (s)e(u.v)dudv

is a bounded operator.
Proof: We note that β is already strongly continuous on the C ∗ algebra generated
by π0 (A), {eitD : t ∈ R}. Thus it suffices to check the statement for elements of the
form [D, a] where a belongs to A∞ .
To this end, fix any one parameter subgroup gt of Tn such that gt goes to the
identity of Tn as t → 0. Let Tt0 , Tet denote the group of normal ∗-automorphisms on
B(H) defined by Tt0 (X) = Vgt XVg−1 and Tet (X) = eitD Xe−itD . As Vgt and D commute,
t
so do the generators of T 0 and Tet . In particular, each of these generators leave the
t
domain of the other invariant. Note also that A∞ is in the domain of the both the
generators, and the generator of Tet is given by [D, ·] there. Thus, for a in A∞ , we have
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 110

a, [D, a] belong to Dom(Ξ) (where Ξ is the generator of Tt0 ), and Ξ([D, a]) = [D, Ξ(a)]
belongs to B(H).
Using this, we obtain

0
Z t
Ts0 (Ξ([D, a]))ds ≤ t kΞ([D, a])k .

Tt ([D, a]) − [D, a] =
0

The required strong continuity follows from this. Then applying Proposition 1.3.5 to
the C ∗ algebra A1 and the action β, we deduce that πJ (X) is a bounded operator. 2

Lemma 4.1.4. For each J, (A∞

J , πJ , H, D) is a spectral triple, that is, [D, πJ (a)] belongs
∞
to B(H) for all a in AJ .
RR RR
Proof: [D, πJ (a)](s) = D fv (s)e(u.v)dudv −
βJu (a)β βJu (a)β
fv (Ds)e(u.v)dudv.
Using ( 1.3.1 ) and closability of D, we have
RR RR
D βJu (a)β
fv (s)e(u.v)dudv = D(βJu (a)βfv (s))e(u.v)dudv.
As D commutes with V, the above expression equals
Z Z Z Z
D(βJu (a)βv (s))e(u.v)dudv −
f βJu (a)Dβfv (s)e(u.v)dudv.

So we have
Z Z
[D, πJ (a)](s) = [D, βJu (a)]β
fv (s)e(u.v)dudv
Z Z
−1 f
= VJu
f [D, a]VJu
f βv (s)e(u.v)dudv
= πJ ([D, a]),

which is a bounded operator by Lemma 4.1.3. 2

4.2 Some preparatory results

In this section, we prove some preparatory results which will be needed in the next two
sections. Let Tn , Tfn , β, β,
e γ be as in the previous subsection. By abuse of notation,
we will use the symbols β and βe for the corresponding comodule maps also. Let γ ∗ , γ∗
be the canonical maps induced by γ from C(Tn ) → C(T fn ) → Lie(Tn )
fn ) and Lie(T
respectively. Moreover, from now on, we will identify A∞ ∞
J with πJ (A ) and often write
π0 (a) simply as a.
Assumption
111 Some preparatory results

2a. Let (Q,e ∆) be a CQG and Q a Woronowicz C ∗ subalgebra of Q. e Let there exist
unital ∗-subalgebra A0 ⊆ A, which is norm dense in every AJ , such that α is an action
: α(π0 (A0 )) ⊆ π0 (A0 ) ⊗ Q0 . Let S0 be a vector subspace of H ( not necessarily closed )
such that there is a map α e : S0 → S0 ⊗alg Qf0 making it into an algebraic Q
f0 co module.
Moreover, (id ⊗ πQe )e α = β.
e
2b. C(T fn ) is a quantum subgroup of Q, e the quotient map being denoted by π e .
Q
2c. α α(s) for a in A0 , s in S0 .
e(as) = α(a)e

We recall that we shall denote by η the canonical homomorphism from Rn to Tn given

by η(x1 , x2 , ......, xn ) = (e(x1 ), e(x2 ), ....., e(xn )). Moreover, we define Ω(u) := eve(u) ◦πQ ,
Ω(u)
e := evẽ(u) ◦ πQe , for u in Rn , where evx (respectively evx̃ ) denotes the state on
C(Tn ) ( respectively, on C(T e n ) ) obtained by evaluation of a function at the point x
(respectively x̃).

We now make some observations.

Lemma 4.2.1. 1. From assumption 2c., it follows that adαe = α.

2. (id ⊗ πQe )ad(eα) = adβe.
3. βex = (id ⊗ Ω(x))e
e α.
4. βx = (id ⊗ Ω(x))α.
5. (γ ∗ )−1 ◦ πQe is a surjective C ∗ homomorphism from Q to C(Tn ) identifying C(Tn )
as a quantum subgroup of Q.

Proof : By using ( 1.2.1 ), we have

adαe (a)s
α−1 (s)
e(a ⊗ id)e
= α
e(a ⊗ id)(s(1) ⊗ κ(s(2) )
= α
e(as(1) ) ⊗ κ(s(2) )
= α
α(s(1) ) ⊗ κ(s(2) )
= α(a)e
α ⊗ id)(id ⊗ κ)e
= α(a)(e α(s)
α−1 ⊗ id)(s)
α ⊗ id)(e
= α(a)(e
= α(a)s,

where we have used Sweedler notations. This proves 1.

Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 112

2. follows from 1. and the fact that πQe is a homomorphism.

βex (h)
e e(x)) = (id ⊗ πf
= β(h)(e Q )e e(x)) = (id ⊗ evee(x) π
α(h)(e eQ )e
α(h)
= (id ⊗ Ω(x))e
e α(h).

Therefore, βex = (id ⊗ Ω(x))e

e α. Similarly, 4. follows from 2.
We now prove 5. Let us denote by γ ∗ the dual map of γ, so that γ ∗ : C(Tn ) → C(T en)
is an injective C ∗ -homomorphism. It is quite clear that (id ⊗ πQe ) ◦ α(A0 ) ⊆ Im(id ⊗ γ ∗ ),
hence it follows that πQe (Q0 ) ⊆ Im(γ ∗ ). Thus, πQ := (γ ∗ )−1 ◦ πQe is a surjective CQG
morphism from Q to C(Tn ), which identifies C(Tn ) as a quantum subgroup of Q. 2

For a fixed J, we shall work with several multiplications on the vector space A0 ⊗alg
Q0 . We shall denote the counit and antipode of Q f0 by  and κ respectively. Let us
define the following operation :
Z
xy = e(−u.v)e(w.s)(Ω(−Ju)
e / x . (Ω(Jw)))(
e Ω(−v)
e / y . Ω(s))dudvdwds,
e
R4n

where x, y belong to Q f0 . Then  is a bilinear maps, and will be seen to be associative

multiplication later on.
We note that when x is in Q0 , f in C(Tn ), γ ∗ (f )(e
e(u) = f (γ(e
e(u)) = f (e(γ∗ (u))) =
f (e(u)) ( as γ is a covering map ). Using this, we have

(Ω(u)
e ⊗ id)∆(x)
= (evee(u) πQe ⊗ id)∆(x)
= (evee(u) γ ∗ πQ ⊗ id)∆(x)
(as γ ∗ πQ = πQe )
= (eve(u) πQ ⊗ id)∆(x)
= (Ω(u) ⊗ id)∆(x)

and thus when x belongs to Q0 ,

(Ω(u)
e ⊗ id)∆(x) = (Ω(u) ⊗ id)∆(x). (4.2.1)

Moreover, we define bilinear maps •, •J , by setting (a ⊗ x) • (b ⊗ y) := ab ⊗ x 

y, (a ⊗ x) •J (b ⊗ y) := (a ×J b) ⊗ (x  y), for a, b in A0 , x, y in Q
e0 .
113 Some preparatory results

Lemma 4.2.2. For x in Q

f0 , we have

Ω(u)
e / (Ω(v)
e / x) = (Ω(u)
e  Ω(v))
e / x.

For x in Q0 , we have

Ω(u) / (Ω(v) / x) = (Ω(u)  Ω(v)) / x.

fn ), hence, we have
Proof : We will denote by ∆Tfn the coproduct on C(T

(πQe ⊗ πQe )∆ = ∆Tfn πQe (4.2.2)

Moreover, we note that as Tn is a commutative group, f  g = g  f for any two

functionals f and g on C(Tn ).

Ω(u)
e / (Ω(v)
e / x)
= (Ω(u)
e ⊗ id)∆(Ω(v)
e / x)
= (Ω(u)
e ⊗ id)∆(Ω(v)(x
e (1) ).x(2) )

= (Ω(u)
e ⊗ id)∆(x(2) )Ω(v)(x
e (1) )

= (Ω(v)
e ⊗ Ω(u)
e ⊗ id)(x(1) ⊗ x(2)(1) ⊗ x(2)(2) )
= (Ω(v)
e ⊗ Ω(u)
e ⊗ id)((id ⊗ ∆)∆(x))
= (Ω(v)
e ⊗ Ω(u)
e ⊗ id)((∆ ⊗ id)∆(x))
= (Ω(v)
e ⊗ Ω(u))∆(x
e (1) ) ⊗ x(2)

= (evη(v) ⊗ evη(u) )(πQe ⊗ πQe )∆(x(1) ) ⊗ x(2) ,

which by ( 4.2.2 ) equals

(evη(v) ⊗ evη(u) )∆Tn πQe (x(1) ) ⊗ x(2)

= (evη(v)  evη(u) )(πQe (x(1) ) ⊗ x(2) )
= (evη(u)  evη(v) )(πQe (x(1) ) ⊗ x(2) )
= (Ω(u)
e ⊗ Ω(v))∆(x
e (1) ) ⊗ x(2)

= (Ω(u)
e  Ω(v))(x
e (1) ) ⊗ x(2)

= ((Ω(u)
e  Ω(v))
e ⊗ id)∆(x)
= (Ω(u)
e  Ω(v))
e / x.

The second part follows from this and using ( 4.2.1 ). 2

Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 114

Lemma 4.2.3. The map  satisfies

Z Z
(Ω(Ju) / x)  (Ω(v) / y)e(u.v)dudv =
e e (x . (Ω(Ju)))(y
e . Ω(v))e(u.v)dudv,
e
R2n R2n

for x, y in Q
f0 . When x, y are in Q0 , we have
Z Z
(Ω(Ju) / x)  (Ω(v) / y)e(u.v)dudv = (x . (Ω(Ju)))(y . Ω(v))e(u.v)dudv.
R2n R2n

Proof : The expression in the left hand side equals

Z
0 e 0 ) / y)e(u0 .v 0 )du0 dv 0
(Ω(Ju
e ) / x)  (Ω(v
Z Z
0
= { e(−u.v)e(w.s)(Ω(−Ju)
e / (Ω(Ju
e ) / x) . Ω(Jw))
e
R2n R4n
(Ω(−v)
e e 0 ) / y) . Ω(s))
/ (Ω(v e dudvdwds}e(u0 .v 0 )du0 dv 0
Z
0 e 0 − v) / y . Ω(s))
= (Ω(J(u
e − u)) / x) . Ω(Jw))(
e Ω(v e
R6n
e(u .v )e(−u.v)e(w.s)dudvdwdsdu0 dv 0
0 0
Z Z
= e(w.s)dwds{ e(u0 .v 0 )e(−u.v)dudvdu0 dv 0
R2n R4n
0 e 0 − v) / ys )},
(Ω(J(u
e − u)) / xw )(Ω(v

where xw = x . Ω(Jw),
e ys = y . Ω(s).
e
The proof of the lemma will be complete if we show
Z
e(u0 .v 0 )e(−u.v)(Ω(J(u
e 0 e 0 − v) / ys )dudvdu0 dv 0 = xw .ys .
− u)) / xw )(Ω(v
R4n

By changing variable in the above integral, with z = u0 − u, t = v 0 − v, it becomes

R
R4n e(−u.v)e((u + z).(v + t))φ(z, t)dudvdzdt
R
= R4n φ(z, t)e(u.t + z.v)e(z.t)dudvdzdt, where

φ(z, t) = (Ω(J(z))
e / xw )(Ω(t)
e / ys ).

By taking (z, t) = X, (v, u) = Y, and F (X) = φ(z, t)e(z.t), the integral can be written
as
Z Z
F (X)e(X.Y )dXdY
= F (0) ( by Proposition 1.3.1)
= (Ω(J(0))
e / xw )(Ω(0)
e / ys )
= xw .ys ,
115 Some preparatory results

since

Ω(J(0))
e / xw = (evη(0) πQe ⊗ id)∆(xw ) = (Tfn ◦ πQe ⊗ id)∆(xw ) = ( ⊗ id)∆(xw ) = xw

and similarly Ω(0) fn ).

e / ys = ys , where  fn denotes the counit of the quantum group C(T
T
This proves the claim and hence the first part of the Lemma. The second part
follows from this and ( 4.2.1 ). 2

Lemma 4.2.4. We have for a in A0 , s in S0 ,

e(βeu (s)) = s(1) ⊗ (id ⊗ Ω(u))(∆(s

α e (2) )), (4.2.3)

α(βu (a)) = a(1) ⊗ (id ⊗ Ω(u))(∆(a(2) )). (4.2.4)

Proof : βeu = (id ⊗ evu ◦ π

e)e
α. We have

βeu (s)
= (id ⊗ Ω(u))e
e α(s)
= (id ⊗ Ω(u))(s
e (1) ⊗ s(2) )

= s(1) (Ω(u))(s
e (2) ).

This gives,

α
e(βeu (s))
= α
e(s(1) )Ω(u)(s
e (2) )

= (id ⊗ id ⊗ Ω(u))(e
e α(s(1) ) ⊗ s(2) )
= (id ⊗ id ⊗ Ω(u))((e
e α ⊗ id)e
α(s))
= (id ⊗ id ⊗ Ω(u))((id
e ⊗ ∆)e
α(s))
= s(1) ⊗ (id ⊗ Ω(u))∆(s
e (2) ).

Proceeding in a similar way, we obtain βu (a) = a(1) (Ω(u))(a(2) ) for all a in A0 and
hence α(βu (a)) = a(1) ⊗ (id ⊗ Ω(u))(∆(a(2) )) for all a in A0 . 2

Lemma 4.2.5. For all s in S0 , a in A0 , we have

Z Z
e(a ×J s)) = a(1) s(1) ⊗ (
α (a(2) . Ω(Ju))(s(2) . Ω(v))e(u.v)dudv).
e (4.2.5)
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 116

For a, b in A0 , we have
Z Z
α(a ×J b) = a(1) b(1) ⊗ ( (a(2) . Ω(Ju))(b(2) . Ω(v))e(u.v)dudv). (4.2.6)

Proof : Using the notations and definitions in section 1.3, we note that for any
f : R2 → C belonging to IB(R2 ) and fixed x in E( where E is a Banach algebra ), the
function F (u, v) = xf (u, v) belongs to IB E (R2 ) and we have
Z Z
x( f (u, v)e(u.v)dudv)
XZ Z
= x (lim (f φp )(u, v)e(u.v)dudv)
L
p∈L
XZ Z
= lim x (f φp )(u, v)e(u.v)dudv)
L
p∈L
Z Z
= x f (u, v)e(u.v)dudv.

Then,

e(a ×J s)
α
Z Z
= αe( βJu (a)βev (s)e(u.v)dudv)
Z Z
= αe( a(1) (Ω(Ju))(a(2) )s(1) (Ω(v))(s
e (2) )e(u.v)dudv)
Z Z
= αe((a(1) s(1) ) (Ω(Ju))(a(2) )(Ω(v))(s
e (2) )e(u.v)dudv)
Z Z
= α(a(1) )eα(s(1) ) (Ω(Ju))(a(2) )(Ω(v))(s
e (2) )e(u.v)dudv

( by assumption 2.c )
Z Z
= α(a(1) )(Ω(Ju))(a(2) )eα(s(1) )(Ω(v))(s
e (2) )e(u.v)dudv
Z Z
= α(a(1) Ω(Ju)(a(2) ))eα(s(1) Ω(v)(s
e (2) ))e(u.v)dudv
Z Z
= α(βJu (a))eα(βev (s))e(u.v)dudv
Z Z
= (a(1) ⊗ (id ⊗ Ω(Ju))(∆(a(2) )))(s(1) ⊗ (id ⊗ Ω(v)))(∆(s
e (2) ))

e(u.v)dudv
( using Lemma 4.2.4)
Z Z
= a(1) s(1) ⊗ (a(2) . Ω(Ju))(s(2) . Ω(v))e(u.v)dudv.
e

2
117 Some preparatory results

Lemma 4.2.6. For s in S0 , a in A0 ,

Z Z
α(a) •J α
e(s) = a(1) s(1) ⊗ ( (Ω(Ju) / a(2) )  (Ω(v)
e / s(2) )e(u.v)dudv. (4.2.7)

For a, b in A0 ,
Z Z
α(a) •J α(b) = a(1) b(1) ⊗ { (Ω(Ju) / a(2) )  (Ω(v) / b(2) )e(u.v)dudv}. (4.2.8)

Proof : We have

α(a) •J α
e(s)
= (a(1) ⊗ a(2) ) •J (s(1) ⊗ s(2) )
= a(1) ×J s(1) ⊗ (a(2)  s(2) )
Z Z
= βJu (a(1) )βev (s(1) )e(u.v)dudv ⊗ (a(2)  s(2) ).

Let  be the counit of Q.

e So we have (id ⊗ )α = id and (id ⊗ )e
α = id. This gives,

α(a) •J α
e(s)
Z Z
= (id ⊗ )α(βJu (a(1) ))(id ⊗ )e
α(βev (s(1) )e(u.v)dudv ⊗ (a(2)  s2 ).

RR
Note that by Lemma 4.2.4, (id ⊗ )(α(βJu (a(1) ))(id ⊗ )(e α(βev (s(1) )))
e(u.v)dudv
RR
= (id ⊗ )(a(1)(1) ⊗ (id ⊗ Ω(Ju))(∆(a(1)(2) )))(id ⊗ )(s(1)(1) ⊗ (id⊗
Ω(v))(∆(s(1)(2) )))e(u.v)dudv
e
RR
= (id ⊗ )(a(1)(1) ⊗ (a(1)(2) . Ω(Ju))(id ⊗ )(s(1)(1) ⊗ (s(1)(2) . Ω(v)))e(u.v)
dudv
RR
= a(1)(1) s(1)(1) (a(1)(2) . Ω(Ju))(s(1)(2) . Ω(v))e(u.v)dudv.
e
Using the fact that f   =   f = f for any functional on Q0 , one has (a(1)(2) .
Ω(Ju)) = Ω(Ju)(a(1)(2) ) and (s(1)(2) . Ω(v))
e = Ω(v)(s
e (1)(2) ), from which it follows that

α(a) •J α
e(s)
Z Z
= a(1)(1) s(1)(1) Ω(Ju)(a(1)(2) )Ω(v)(s (1)(2) )e(u.v)dudv ⊗ (a(2)  s(2) )
e
Z Z
= (id ⊗ Ω(Ju) ⊗ id)(a(1)(1) ⊗ a(1)(2) ⊗ a(2) ) • (id ⊗ Ω(v)
e ⊗ id)(s(1)(1) ⊗
s(1)(2) ⊗ s(2) )e(u.v)dudv
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 118

Z Z
= (id ⊗ Ω(Ju) ⊗ id)(a(1) ⊗ ∆(a(2) )) • (id ⊗ (Ω(v)
e ⊗ id))(s(1) ⊗ ∆(s(2) ))
e(u.v)dudv
Z Z
= {a(1) ⊗ (Ω(Ju) ⊗ id)∆(a(2) )} • {s(1) ⊗ (Ω(v)
e ⊗ id)∆(s2 )}e(u.v)dudv
Z Z
= a(1) s(1) ⊗ ((Ω(Ju)
e ⊗ id)∆(a2 ))  (Ω(v)
e ⊗ id))∆(s(2) )e(u.v)dudv
( by (4.2.1))
Z Z
= a(1) s(1) ⊗ (Ω(Ju)
e / a(2) )  (Ω(v)
e / s(2) )e(u.v)dudv,

where we have used the relation (α ⊗ id)α = (id ⊗ ∆)α to get a(1)(1) ⊗ a(1)(2) ⊗ a(2) =
a(1) ⊗ ∆(a(2) ) and similarly s(1)(1) ⊗ s(1)(2) ⊗ s(2) = s(1) ⊗ ∆(s(2) ). 2

Combining Lemma 4.2.3, Lemma 4.2.5 and Lemma 4.2.6, we conclude the following.

Lemma 4.2.7. For a in A0 , s in S0 , we have

α(a) •J α e(a ×J s).

e(s) = α (4.2.9)

For a, b in A0 , we have
α(a) •J α(b) = α(a ×J b). (4.2.10)

We shall now identify  with the multiplication of a Rieffel-type deformation of

Q( Qe ). We discuss the case for Q, e that of Q being similar. Since Qe has a quantum
subgroup isomorphic with C(T fn ), we can consider the following canonical action χ of
R2n on Qe ( as in ( 1.3.8 ) ) given by

χ(s,u) = (Ω(−s)
e ⊗ id)∆(id ⊗ Ω(u))∆.
e

Now, let Je := −J ⊕J, which is a skew-symmetric 2n×2n real matrix, so one can deform
Q
e by defining the product of x and y (x, y belonging to Q
f0 , say) to be the following:
Z Z
χJ(u,w)
e (x)χv,s (y)e((u, w).(v, s))d(u, w)d(v, s).

We claim that this is nothing but  introduced before.

Lemma 4.2.8.

x  y = x ×Je y for all x, y ∈ Q

f0 .
119 g + of a Rieffel deformed noncommutative manifold
QISO R

Proof : Let us first observe that

χJ(u,w)
e (x)
= (Ω(Ju)
e ⊗ id)∆(id ⊗ Ω(Jw))∆(x)
e

= Ω(Ju)
e / x . Ω(Jw),
e

and similarly χ(v,s) (y) = Ω(−v)

e / y . Ω(s).
e
Thus, we have

xy
Z
= (Ω(−Ju)
e / x . Ω(Jw))(
e Ω(−v)
e / y . Ω(s))e(−u.v)e(w.s)dudvdwds
e
4n
ZR
0 0
= (Ω(Ju
e ) / x . Ω(Jw))(
e Ω(−v)
e / y . Ω(s))e(u
e .v)e(w.s)du0 dvdwds
Z R4n Z
= χJ(u,w)
e (x)χ(v,s) (y)e((u, w).(v, s))d(u, w)d(v, s),
R2n R2n

which proves the claim. 2

Let us denote by Q e e ( Q e ) the C ∗ algebra obtained from Q

e ( Q ) by the Rieffel
J J
deformation w.r.t. the matrix Je described above. We recall from subsection 1.3.1 that
the coproduct ∆ on Q f0 ( Q0 ) extends to a coproduct for the deformed algebra as well
and (Qe e, ∆) ( (Q e, ∆) ) is a compact quantum group.
J J

4.3 ^+ of a Rieffel deformed noncommutative mani-

QISOR
fold
4.3.1 Derivation of the result
In this subsection, our set up is as in section 4.1 so that we have spectral triples on A∞
J
for each J.

Lemma 4.3.1. Suppose that (Q, e U ) belongs to Obj(Q(A, H, D)), and there exists a
unital ∗-subalgebra A0 ⊆ A which is norm dense in every AJ such that
αU (π0 (A0 )) ⊆ π0 (A0 ) ⊗alg Q0 , where Q ⊆ Qe is the smallest Woronowicz C ∗ subal-
gebra such that αU (A0 ) ⊆ π0 (A0 ) ⊗ Q, and Q0 is the Hopf ∗-algebra obtained by matrix
coefficients of irreducible unitary (co)-representations of Q. Also, let S0 = span{as :
a ∈ A0 , s ∈ S00 }, Then we have the following:
(a) U (S0 ) ⊆ S0 ⊗alg Qe0 .
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 120

e := U |S0 : S0 → S0 ⊗alg Q
(b) α e0 makes S0 an algebraic Q
e0 co-module, satisfying

α α(s) for all a in A0 , s in S0 .

e(π0 (a)s) = αU (a)e

Moreover, if (C(T e in Q(A, H, D) such that V (.⊗id)V ∗ = β,

fn ), V ) is a sub object of Q
then C(Tn ) is a quantum subgroup of Q.

Proof: U commutes with D and hence preserves the eigenspaces of D which shows
that U preserves S00 . Then, U e (s⊗1) ⊆ (A0 ⊗Q0 )(S00 ⊗ Q
e (as⊗1) = α(a)U f0 ) ⊆ S0 ⊗ Q
f0 .
Thus, the first assertion follows.
The second assertion follows from the definition of α
e and αU . The third assertion
follows as in Lemma 4.2.1. 2

Remark 4.3.2. From the definitions of A0 and S0 , it follows that

(i) π0 (A0 )S0 ⊆ S0 ,
(ii) βg (A0 ) ⊆ A0 for all g.

Let us now fix the object (Q, e U ) as in the statement of Lemma 4.3.1. We recall
that using the identification of Q0 as a common vector-subspace of all QJe, we shall
sometimes denote this identification map from Q0 to QJe by ρJ .

Let us consider the finite dimensional unitary representations U (i) := U |Vi , where Vi
is the eigenspace of D corresponding to the eigenvalue λi . By Corollary 1.3.14, we can
(i)
view U (i) as a unitary representation of QJe as well, and let us denote it by UJ . In this
way, we obtain a unitary representation UJ on the Hilbert space H, which is the closed
linear span of all the Vi ’s. It is obvious from the construction (and the fact that the
linear span of Vi ’s, that is S0 , is a core for D) that UJ D = (D ⊗ I)UJ . Let αJ := αUJ .
With this, we have the following:

Lemma 4.3.3. For a in A0 , we have αJ (a) = (α(a))J ≡ (πJ ⊗ ρJ )(α(a)), hence in

particular, for every state φ on QJe, (id ⊗ φ) ◦ αJ (AJ ) ⊆ A00J .

Using the equation ( 4.2.9 ), we have for all s in S0 , a in A0 ,

αJ (a)UJ (s)
= UJ (πJ (a)s)
e(a ×J s)
= α
= α(a) •J α
e(s)
= (α(a))J UJ (s),
121 g + of a Rieffel deformed noncommutative manifold
QISO R

from which we conclude by the density of S0 in H that αJ (a) = (α(a))J belongs to

πJ (A0 ) ⊗ QJe. The lemma now follows using the norm-density of A0 in AJ . 2

Corollary 4.3.4. (Q e e, UJ ) is an orientation preserving isometric action on the spectral

J
triple (A∞
J , H, D).

We shall now show that if we fix a ‘volume-form’ in terms of an R-twisted structure,

then the ‘deformed’ action αJ preserves it.

Lemma 4.3.5. Suppose, in addition to the set-up already assumed, that there is an
invertible positive operator R on H such that (A∞ , H, D, R) is an R-twisted Θ-summable
spectral triple, and let τR be the corresponding ‘volume form’. Assume that αU preserves
the functional τR . Then the action αUJ preserves τR too.

Proof : Let the (finite dimensional) eigenspace corresponding to the eigenvalue λn of

D be Vn . As U commutes with D, there exists subspaces Vn,k of Vn and an orthonormal
basis {en,k e n,k
j }j for Vn,k such that the restriction of U to Vn,k is irreducible. Write U (ej ⊗
P n,k e ∗ (en,k ) = P en,k ⊗ tn∗ .
1) = e ⊗ tn . Then, U
i i i,j j i i j,i
Then H will be decomposed as H = ⊕n≥1, k Vn,k .
Let R(en,i n,s
P
j )= s,t Fn (i, j, s, t)et .
By hypothesis, U e (. ⊗ id)U e ∗ preserves the functional τR (·) = T r(R ·) on ED where ED
is as in Proposition 3.2.7, that is the weakly dense ∗ subalgebra of B(H) generated by
the rank one operators |ξ >< η| where ξ, η are eigenvectors of D. Thus, (τR ⊗id)(U e (X ⊗
id)Ue ∗ ) = τR (X).1Q for all X in ED .
Then, for a in ED , we have:

(τR ⊗ h)(U eJ∗ )

eJ (a ⊗ 1)U
X D n,i E
= ej ⊗ 1, U eJ∗ (Ren,i ⊗ 1)
eJ (a ⊗ 1)U
j
n,i,j
X D E
= e ∗ (en,i ⊗ 1), (a ⊗ 1)U
U e ∗ (Fn (i, j, s, t)en,s ⊗ 1)
J j J t
n,i,j,s,t
D E
n ∗ n ∗
Fn (i, j, s, t) en,i n,s
X
= k ⊗ (t j,k ) , (a ⊗ 1)(e l ⊗ (tt,l )
n,i,j,s,t,k,l
D E
Fn (i, j, s, t) en,i n,s
hJ ((tnj,k ) ×J (tnt,l )∗ )
X
= k , ael
n,i,j,s,t,k,l
D E
Fn (i, j, s, t) en,i n,s
X
= k , ael h0 (tnj,k tnt,l ∗ )
n,i,j,s,t,k,l

= (τR ⊗ h)(U e ∗)
e (a ⊗ 1)U

= τR (a).1

where hJ ((tnj,k ) ×J (tnt,l )∗ ) = h0 (tnj,k tnt,l ∗ ) as deduced by using Lemma 1.3.10.

Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 122

Thus (τR ⊗ h)U fJ ∗ = τR (a).1

fJ (a ⊗ id)U
Let (τR ⊗ h)UfJ (X ⊗ id)UfJ ∗ = (τR ∗ h)(X). As U fJ (· ⊗ id)UfJ ∗ keeps ED invariant, we
can use Sweedler notation: U fJ ∗ = a(1) ⊗ a(2) , with a, a(1) belonging to ED , a(2)
fJ (a ⊗ 1)U
belonging to Q
f˜, to have
J

(τR ∗ h ⊗ id)U fJ ∗
fJ (a ⊗ id)U

= (τR ⊗ h)(U fJ ∗ ) ⊗ a(2)

fJ (a(1) ⊗ id)U

= (τR ⊗ h ⊗ id)(αU ⊗ id)αU (a)

= (τR ⊗ h ⊗ id)(id ⊗ ∆)αU (a)
= (τR ⊗ id)(id ⊗ (h ⊗ id)∆)αU (a)
= (τR ⊗ id)(id ⊗ h(.).1)αU (a)
= (id ⊗ h(.).1)(τR ⊗ id)αU (a)
= (id ⊗ h(.).1)τ (a).1
= τ (a).1.

Thus,

(τR ⊗ id)(U fJ ∗ )
fJ (a ⊗ 1)U

= (τR ∗ h ⊗ id)(U fJ ∗ ) = (τR ∗ h)(a(1) )a(2)

fJ (a ⊗ 1)U

= (τR ⊗ h ⊗ id)(a(1)(1) ⊗ a(1)(2) ⊗ a(2) ) = (τR ⊗ h ⊗ id)(id ⊗ ∆Je)

∗
fJ (a ⊗ 1)U
(U fJ )

= τR (a(1) )(h ⊗ id) ◦ ∆Je(a(2) ) = τR (a(1) )h(a(2) ).1QJe

= (τR ⊗ h)(a(1) ⊗ a(2) ) = (τR ∗ h)(a).1QJe = τR (a).1QJe.

+
Remark 4.3.6. If QISOR (A∞ , H, D) ( QISO+ (A∞ , H, D), if it exists ) has a C ∗
action, then from the definition of a C ∗ action, we get a subalgebra A0 as in Lemma
4.3.1. Thus, the assumptions of section 4.2 are satisfied so that the conclusions in
+
that subsection hold for QISOR (A∞ , H, D) ( QISO+ (A∞ , H, D)).. Similarly, the con-
+
clusions of Lemma 4.3.1 and the subsequent Lemmas hold for QISOR (A∞ , H, D) (
QISO+ (A∞ , H, D)).

For any two compact quantum groups (S1 , U S1 ) and (S2 , U S2 ) in Q0 (AJ , H, D), we
write S1 < S2 if S1 is a sub object of S2 in the category Q0 (AJ , H, D).
123 g + of a Rieffel deformed noncommutative manifold
QISO R

Lemma 4.3.7. If G1 , G2 be two CQG s such that G1 < G2 in the category Q0 (AJ , H, D),
If (G1 )Je and (G2 )Je make sense, then (G1 )J < (G2 )J in the category Q0 (AJ , H, D).

Proof : From Corollary 4.3.4, we see that (Gi )Je is an object in the category
Q0 (A 0
J , H, D). Let π2 be the morphism from G2 to G1 in the category Q and π1 be the
morphism from G1 to Tn in the same category. Let ∆i , ×iJ , χi denote respectively the
coproducts, products and R2n action on (Gi )J , i = 1, 2.
As the quantum group structure is not altered under Rieffel deformation, to prove
the Lemma, it is enough to show that π2 is a homomorphism from (G2 )J to (G1 )J .
In any CQG (Q, ∆), f, g linear functionals on Q and for all a in Q0 , (f ⊗
id)∆(id ⊗ g)∆(a) = (f ⊗ id)∆(a(1) )g(a(2) ) = (f ⊗ id)(a(1)(1) ⊗ a(1)(2) )g(a(2) ) =
f (a(1)(1) )g(a(2) )a(1)(2) = (f ⊗ id ⊗ g)(a(1)(1) ⊗ a(1)(2) ⊗ a2 ) = (f ⊗ id ⊗ g)(∆(a(1) ) ⊗ a(2) ) =
(f ⊗ id ⊗ g)(∆ ⊗ id)∆(a).
Hence,
(f ⊗ id)∆(id ⊗ g)∆ = (f ⊗ id ⊗ g)(∆ ⊗ id)∆. (4.3.1)

Moreover, we will also need the equation

(π2 ⊗ π2 )∆2 = ∆1 π2 , (4.3.2)

which holds as π2 is a morphism of CQG s, G2 → G1 .

Let λ, ρ be as in ( 1.3.6 ) and ( 1.3.7 ) and a belongs to (G2 )J .
Then,

π2 χ2(s,u) (a)
= π2 (λη(−s) ρη(u) )(a)
= π2 (evη(−s) (π1 ◦ π2 ) ⊗ id)∆2 (id ⊗ evη(u) (π1 ◦ π2 ))∆2 (a))
= π2 (evη(−s) (π1 ◦ π2 ) ⊗ id ⊗ evη(u) ◦ (π1 ◦ π2 ))(∆2 ⊗ id)∆2 (a)
( by ( 4.3.1 ) )
= (evη(−s) (π1 ◦ π2 ) ⊗ π2 ⊗ evη(u) ◦ (π1 ◦ π2 ))(∆2 ⊗ id)∆2 (a)
= (evη(−s) π1 ⊗ id ⊗ evη(u) π1 )(π2 ⊗ π2 ⊗ π2 )(∆2 ⊗ id)∆2 (a)
= (evη(−s) π1 ⊗ id ⊗ evη(u) π1 )((π2 ⊗ π2 )∆2 ⊗ π2 )∆2 (a)
= (evη(−s) π1 ⊗ id ⊗ evη(u) π1 )(∆1 π2 ⊗ π2 )∆2 (a)
( using (4.3.2) )
= (evη(−s) π1 ⊗ id ⊗ evη(u) π1 )(∆1 ⊗ id)(π2 ⊗ π2 )∆(2) (a)
= (evη(−s) π1 ⊗ id ⊗ evη(u) π1 )(∆1 ⊗ id)∆1 (π2 (a))
= (evη(−s) π1 ⊗ id)∆1 (id ⊗ evη(u) π1 )∆1 (π2 (a))
= (λη(−s) ρη(u) )π2 (a)
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 124

= χ1(s,u) π2 (a).

Thus, for all (s, u) in R2n , π2 χ2(s,u) = χ1(s,u) π2 .

Therefore, for all a, b in (G2 )J ,

π2 (a ×2J b)
Z Z
= π2 ( χ2Ju (a)χ2v (b)e(u.v)dudv)
Z Z
= π2 (χ2Ju (a))π2 (χ2v (b))e(u.v)dudv
Z Z
= χ1Ju (π2 (a))χ1v (π2 (b))e(u.v)dudv

= π2 (a) ×1J π2 (b).

where the third step is permissible by Proposition 1.3.2.

This proves that π2 is indeed a homomorphism. 2

+ +
Theorem 4.3.8. 1. If QISOR (A∞ ∞ ∗
J , H, D) and (QISOR (A , H, D))Je have C actions
on A and AJ respectively, we have

^+ (A∞ , H, D) ∼
QISO ^+ (A∞ , H, D))
= (QISO
R J R Je

+
(A∞ ∼ + ∞
QISOR J , H, D) = (QISOR (A , H, D))Je.
+
^ (A∞ , H, D) and QISO
2. If moreover, QISO ^+ (A∞ , H, D) both exist and have C ∗
J
actions on A and AJ respectively, then
 
+ +
^ (A∞ ∼ ^ (A∞ , H, D) ,
QISO J , H, D) = QISO
Je

QISO+ (A∞ ∼ + ∞
J , H, D) = (QISO (A , H, D))Je.

Proof : We prove 1 only. From Corollary 4.3.4 and Lemma 4.3.5 we see that
^+ (A, H, D) is an object of Q0 (A , H, D). Thus,
QISO R Je R J

^+ (A , H, D) in Q0 (A , H, D).
^+ (A, H, D)) < QISO
(QISO R Je R J R J

^+ (A, H, D)) ) < (QISO

So, by Lemma 4.3.7, ((QISO ^+ (A , H, D))
R Je −Je R J −Je
+ +
in Q0 (A, H, D), hence QISO (A, H, D) < (QISO (A , H, D)) .
R
^
R
^
R J −Je
125 g + of a Rieffel deformed noncommutative manifold
QISO R

Replacing A by A−J , we have

^+ (A , H, D)
QISO R −J
^+ ((A ) , H, D)
< QISO R −J J −Je
(in Q0 R (A−J , H, D))
∼ ^+ (A, H, D)
= QISO (in Q0 R (A−J , H, D)) ∼ ^+ (A, H, D)) .
= (QISO
R −Je R −J
f

^+ (A , H, D) < (QISO
Thus, QISO ^+ (A, H, D)) in Q0 (A , H, D) which implies
R J R Je R J
^ + ∼ ^ + 0
QISOR (AJ , H, D) = (QISOR (A, H, D))Je in Q R (AJ , H, D). 2

4.3.2 Computations
Fix a real number θ, and then we recall from subsection 1.1.1 that the C ∗ algebra Aθ
is the universal C ∗ algebra generated by two unitaries U and V such that U V = λV U ,
where λ := e2πiθ . We also recall from section 1.3 that Aθ is a Rieffel type deformation
of C(T2 ) by using 2 2
! the canonical action of R on T and the skew symmetric matrix
0 − 2θ
J = θ
. It is well-known (see [17]) that the set {U m V n : m, n ∈ Z} is an
2 0
orthonormal basis for L2 (Aθ , τ ), where τ denotes the unique faithful normalized trace
on Aθ given by, τ ( amn U m V n ) = a00 . We will denote the GNS space L2 (Aθ , τ ) by H0 .
P

Let Afin
θ be the unital ∗-subalgebra generated by finite complex linear combinations of
U V , where m, n belong to Z, and d1 , d2 be the maps on Afin
m n m n
θ defined by d1 (U V ) =
mU m V n , d2 (U m V n ) = nU m V n .
We consider the spectral triple obtained from the classical spectral triple on T2 as
described in section 4.1.
+ +
^ (A∞ , H, D) = QISO
Theorem 4.3.9. QISO ^ (C ∞ (T2 )) = C(T2 ) ∗ C(T), and
θ
QISO+ (A∞ + ∞ 2 2
θ ) = QISO (C (T )) = C(T ).

Proof: We use Theorem 4.3.8 and recall that QISO+ (C ∞ (T2 )) = C(T2 ) ( Theorem
3.4.14 ) which is generated by z1 and z2 , say. QISO+ (C ∞ (T2 )) contains C(T2 ) itself as a
quantum subgroup. Hence, by Theorem 4.3.8, QISO+ (A∞ θ ) is the CQG obtained from
the Rieffel deformation 2
of C(T ) via 4
  the action of R and the skew symmetric matrix
θ
0 −2 0 0
 θ 
 2 0 0 0 
J = J ⊕ −J = 
e   so that J(r e 1 , r2 , r3 , r4 ) = (− θ r2 , θ r1 , θ r4 , − θ r3 ).
2 2 2 2
 0 0 0 2θ  
0 0 − 2θ 0
For f1 , f2 in C ∞ (T2 ), r = (r1 , r2 , r3 , r4 ) in R4 , r0 = (r10 , r20 , r30 , r40 ) in R4 , (t1 , t2 ) in T2 ,
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 126

the deformed product is given by

Z Z
0 0
(f1 ×Je f2 )(t1 , t2 ) = χJ(r
e 1 ,r2 ,r3 ,r4 ) (f1 )(t1 , t2 )χ(r10 ,r20 ,r30 ,r40 ) (f2 )(t1 , t2 )e(r.r )drdr .

Here, for f in C ∞ (T2 ),

χ(r1 ,r2 ,r3 ,r4 ) f (t1 , t2 )

= (evη(−(r1 ,r2 )) ⊗ id)∆(id ⊗ evη(r3 ,r4 ) )∆(f )(t1 , t2 )
= f (η(−(r1 , r2 ))(t1 , t2 )η(r3 , r4 ))
= f ((e(−r1 ), e(−r2 ))(t1 , t2 )(e(r3 ), e(r4 )))
= f (e(r3 − r1 )t1 , e(r4 − r2 )t2 ).

Therefore,

z1 ×Je z2
Z Z
θ θ θ θ
= z1 (e( r4 + r2 )t1 , e(− r1 − r3 )t2 )z2 (e(r30 − r10 )t1 , e(−r20 + r40 )t2 )
2 2 2 2
0 0
e(r.r )drdr
Z Z
θ θ
= e( r4 + r2 )t1 e(−r20 + r40 )t2 e(r1 .r10 )e(r2 .r20 ).e(r3 .r30 )e(r4 .r40 )drdr0
2 2
Z Z Z Z
θ 0 0 0 θ
= t1 t2 e( r2 )e(−r2 )e(r2 .r2 )dr2 dr2 e( r4 )e(r40 )e(r4 .r40 )dr4 dr40
2 2
Z Z
e(r1 .r10 )dr1 dr10 e(r3 .r30 )dr3 dr30
RR

Z Z Z Z
θ θ
= t1 t2 e(− r2 )e(r20 )e((−r2 ).(−r20 ))dr2 dr20 e(− r4 )e(−r40 )e(−r4 . − r40 )
2 2
0
dr4 dr4 .1.1.

( by Proposition 1.3.1 )
Similarly,

z2 ×Je z1
Z Z
θ θ
= e(− r1 − r3 )t2 e(r30 − r10 )t1 e(r1 .r10 )e(r3 .r30 )dr1 dr10 dr3 dr30 .1.1
2 2
Z Z Z Z
θ θ
= t1 t2 e(− r1 )e(−r10 )e(r1 .r10 )dr1 dr10 e(− r3 )e(r30 )e(r3 .r30 )dr3 dr30
2 2
= z1 ×Je z2 .

This proves the theorem.

2
127 QISOL of a Rieffel deformed noncommutative manifold

4.4 QISOL of a Rieffel deformed noncommutative mani-

fold
4.4.1 Derivation of the main result
In this subsection, our set up is as in section 4.1 so that we have spectral triples on A∞ J
for each J such that the spectral triple on A∞ satisfies all the assumptions mentioned
in chapter 2 for ensuring the existence of QISOL where L is the Laplacian coming from
the spectral triple. As QISOL has a C ∗ action on A, we get a subalgebra A0 as in
Lemma 4.3.1. Thus, the assumptions of section 4.2 are satisfied so that the conclusions
in that subsection hold for QISOL (A∞ , H, D).
We recall from Chapter 2 that A0 will denote the ∗-algebra generated by complex
linear (algebraic, not closed) span A∞ 0 of the eigenvectors of L which has a countable
discrete set of eigenvalues each with finite multiplicities, by assumptions for the existence
of QISOL . By the same assumptions, A∞ 0 is a subset of A
∞ and is norm-dense in A.

Here, we make the following additional assumptions.

Assumptions (i) A0 is dense in A∞ w.r.t. the Frechet topology coming from the
action of Tn .
( ii ) n≥1 Dom(Ln ) = A∞ .
T

(iii) L commutes with the Tn -action β, hence C(Tn ) can be identified as a sub
object of QISOL in the category Q0L .
Let π denote the surjective map from QISOL to its quantum subgroup C(Tn ), which
is a morphism of compact quantum groups. We denote by α : A → A ⊗ QISOL the
action of QISOL on A, and note that on A0 , this action is algebraic, that is, it is
an action of the Hopf ∗-algebra Q0 consisting of matrix elements of finite dimensional
unitary representations of Q. We have (id ⊗ π) ◦ α = β.
We recall from Proposition 1.3.8 that A∞ = A∞ J as topological spaces, that is they
coincide as sets and the corresponding Frechet topologies are also equivalent. In view of
this, we shall denote this space simply by A∞ , unless one needs to consider it as Frechet
algebra, in which case the suffix J will be used.

Lemma 4.4.1. For F in B B(H) (R2 ) ( notation as in section 1.3 ) and a trace class
operator W,
Z Z Z Z
T r( F (u, v)W e(u.v)dudv) = T r(F (u, v)W )e(u.v)dudv.

Proof : From the definition of oscillatory integral, we have πJ (F (u, v)) =

P R
p∈L (F (u, v)φp )(u, v)e(u.v)dudv ( notations as in section 1.3 ). Let {Ln }n be a
sequence of subsets of L such that it increases to L. Then, as the above sum is strongly
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 128

P R
convergent, limn ( p∈Ln (F (u, v)φp )(u, v)e(u.v)dudv) converges in SOT. We deduce
the result by using Lemma 1.1.5. 2

Proposition 4.4.2. Let LJ denote the Laplacian from the spectral triple (A∞
J , H, D).
∞
Then LJ coincides with L on A ⊆ AJ .

Proof : We recall that from the proof of Lemma 4.1.4, we have [D, πJ (a)] =
πJ ([D, a]). Denoting the inner product on Ω1J (A∞ ) by h., .iJ and letting a, b in A∞ , Lim
as in subsection 1.5.2, we have by using Lemma 4.4.1

hLJ a, biJ
T r([D, πJ (a)]∗ [D, πJ (b)]e−tD )
2

= Limt→0+
T r(e−tD2 )
T r(πJ ([D, a]∗ [D, b])e−tD )
2

= Limt→0+
T r(e−tD2 )
βJu ([D, a]∗ [D, b])βfv e(u.v)dudve−tD2 )
RR
T r(
= Limt→0+
T r(e−tD2 )
T r(βJu ([D, a]∗ [D, b])βfv e−tD2 )e(u.v)dudv
RR
= Limt→0+
T r(e−tD2 )
( by Lemma 4.4.1)
RR ∗ −tD2 )e(u.v))dudv
T r(VJu
f ([D, a] [D, b])VJu f −1 βv e
= Limt→0+
T r(e−tD2 )
RR ∗ −tD2 V
T r(VJu
f ([D, a] [D, b])βv e f −1 )e(u.v)dudv
Ju
= Limt→0+ −tD 2
T r(e )
RR ∗ −tD 2
T r(([D, a] [D, b])βfv e )e(u.v))dudv
= Limt→0+ 2
T r(e−tD )
[D, a]∗ [D, b]β
fv e−tD2 e(u.v)dudv)
RR
T r(
= Limt→0+
T r(e−tD2 )
which by Proposition 1.3.1, equals
T r([D, a]∗ [D, b]e−tD )
2

= Limt→0+
T r(e−tD2 )
= hLa, bi .

. 2
Thus, the quantum isometry group QISOL (AJ ) is the universal compact quantum
group acting on AJ , with the action keeping each of the eigenspaces of L invariant. Note
that the algebraic span of eigenvectors of LJ coincides with that of L, that is A∞
0 , which
∞ ∞
is already assumed to be Frechet-dense in A = AJ , hence in particular norm-dense
in AJ . Moreover, all the results of section 4.2 hold for QISOL (AJ ).
129 QISOL of a Rieffel deformed noncommutative manifold

We now state and prove a criterion, to be used later, for extending positive maps
defined on A0 .

Lemma 4.4.3. Let B be another unital C ∗ -algebra equipped with a Tn -action, so that
we can consider the C ∗ -algebras BJ for any skew symmetric n × n matrix J. Let φ :
A∞ → B ∞ be a linear map, satisfying the following :
(a) φ is positive w.r.t. the deformed products ×J on A0 and B ∞ , that is φ(a∗ ×J a) ≥ 0
( in BJ∞ ⊂ BJ ) for all a in A0 , and
(b) φ extends to a norm-bounded map (say φ0 ) from A to B.
Then φ also have an extension φJ as a k kJ -bounded positive map from AJ to BJ
satisfying kφJ k = kφ(1)kJ .

Proof : We can view φ as a map between the Frechet spaces A∞ and B ∞ , which is
clearly closable, since it is continuous w.r.t. the norm-topologies of A and B, which are
weaker than the corresponding Frechet topologies. By the Closed Graph Theorem, we
conclude that φ is continuous in the Frechet topology. Since A∞ = A∞ J and B
∞ = B∞
J
as Frechet spaces, consider φ as a continuous map from A∞ J to B ∞ , and it follows
J
by the Frechet-continuity of ×J and ∗ and the Frechet-density of A0 in A∞ J that the
positivity (w.r.t. ×J ) of the restriction of φ to A0 ⊂ A∞ J is inherited by the extension
∞ ∞ ∞ ∞
on A = AJ . Indeed, given a in AJ = A , choose a sequence an belonging to A0
such that an → a in the Frechet topology. We have φ(a∗ ×J a) = limn φ(a∗n ×J an ) in
the Frechet topology, so in particular, φ(a∗n ×J an ) → φ(a∗ ×J a) in the norm of BJ ,
which implies that φ(a∗ ×J a) is a positive element of BJ since φ(a∗n ×J an ) is so for
each n. Next, by Lemma 4.1.2, we note that A∞ is closed under holomorphic functional
calculus as a unital ∗-subalgebra of AJ (the identity of A∞ J is same as that of A),
∞
and hence, by Lemma 4.1.2, for any self adjoint
∗
element
∗
A , ∗ kφ(x)k
x in ≤ kxk φ(1).

∞ y+y y−y y+y y−y∗
Thus, for any y in A , kφ(y)k = φ( 2 + i 2i ) ≤ φ( 2 ) + φ( 2i ) ≤

∗

y−y∗
( y+y + 2i )φ(1) ≤ 2 kyk φ(1). Thus, φ is bounded on A
∞ and hence admits a

2
bounded extension (say φJ ) on AJ , which will still be a positive map, so in particular
the norm of φJ is same as kφJ (1)k. 2

Now we note that due to the assumption (iii), (QISOL )Je makes sense.

Remark 4.4.4. By Lemma 4.2.7 along with Lemma 4.2.8, α is a homomorphism from
AJ to AJ ⊗(QISOL (A))Je and hence (QISOL (A))Je is an object in the category QL (AJ ).

Theorem 4.4.5. If the Haar state is faithful on Q, then α : A0 → A0 ⊗ Q0 extends

to an action of the compact quantum group QJe on AJ , which is isometric, smooth and
faithful.
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 130

Proof : We have already seen in ( 4.2.10 ) that α is an algebra homomorphism

from A0 to A0 ⊗alg Q0 (w.r.t. the deformed products), and it is also ∗-homomorphism
since it is so for the undeformed case and the involution ∗ is the same for the deformed
and undeformed algebras. It now suffices to show that α extends to AJ as a C ∗ -
homomorphism. Let us fix any faithful imbedding AJ ⊆ B(H0 ) (where H0 is a Hilbert
space) and consider the imbedding QJe ⊆ B(L2 (hJ )) ( by virtue of Lemma 1.3.13 ).
By definition, the norm on AJ ⊗ QJe is the minimal (injective) C ∗ -norm, so it is equal
to the norm inherited from the imbedding AJ ⊗alg QJe ⊆ B(H0 ⊗ L2 (hJ )). Let us
consider the dense subspace D ⊂ H0 ⊗ L2 (hJ ) consisting of vectors which are finite
P
linear combinations of the form i ui ⊗ xi , with ui belonging to H0 , xi belonging to
Q0 ⊂ L2 (hJ ). Fix such a vector ξ = ki=1 ui ⊗ xi and consider B := A ⊗ Mk (C), with
P

the Tn -action β ⊗ id on B. Let φ : A∞ → B ∞ be the map given by

 k
φ(a) := (id ⊗ φ(xi ,xj ) )(α(a)) ,
i,j=1

where φ(x,y) (z) := h(x∗ ×Je z ×Je y) for x, y, z in Q0 . By Remark 1.3.11, φ(xi ,xj ) extends
to Q as a bounded linear functional. Note that the range of φ is in B ∞ = A∞ ⊗ Mk (C)
since we have (id ⊗ φ(x,y) )α(A∞ ) ⊆ A∞ by Proposition 2.1.2, using our assumption (ii)
that n≥1 Dom(Ln ) = A∞ .
T

Since α maps A0 into A0 ⊗alg Q0 and h = hJ on Q0 , it is easy to see that for a in

A0 , φ(a∗ ×J a) is positive in BJ . As φ(xi ,xj ) extends to Q as a bounded linear functional,
φ extends to a bounded linear (but not necessarily positive) map from A to B. Thus,
the hypotheses of Lemma 4.4.3 are satisfied and we conclude that φ admits a positive
extension, say φJ , from AJ to BJ = AJ ⊗ Mk (C). Thus, we have for a in A0 ,

k
X
hui , φ(a∗ ×J a)uj i
i,j=1
X X
≤ kak2J hui , φ(1)uj i = kak2J hui , uj ih(x∗i ×Je xj )
ij ij

X k
X
= kak2J hui ⊗ xi , uj ⊗ xj i = kak2J k ui ⊗ xi k2 .
ij i=1

This implies

k
X
kα(a)ξk2 = hξ, α(a∗ ×J a)ξi = hui , φ(a∗ ×J a)uj i ≤ kak2J kξk2
i,j=1

for all ξ in D and a in A0 , hence α admits a bounded extension which is clearly a

C ∗ -homomorphism. 2
131 QISOL of a Rieffel deformed noncommutative manifold

For any two objects S1 and S2 in Q0L ), we write S1 <L S2 if S1 is a sub-object of S2

in the category Q0L .
Lemma 4.4.6. If G1 , G2 be two CQG s such that G1 <L G2 in the category
Q0L (A, H, D). If (G1 )Je and (G2 )Je make sense, then (G1 )Je <L (G2 )Je in the category
Q0L (AJ , H, D).
Proof : By Remark 4.4.4, (Gi )Je are objects in the category Q0L (AJ ), i = 1, 2. The
rest of the proof is the same as that of Lemma 4.3.7 and hence we omit it. 2

Theorem 4.4.7. If the Haar state on QISOL (A) is faithful, we have the isomorphism
of compact quantum groups:

(QISOL (A))Je ∼
= QISOL (AJ ).

Proof : By Theorem 4.4.5, we have seen that (QISOL (A))Je also acts faithfully,
smoothly and isometrically on AJ , which implies,

(QISOL (A, H, D))Je < QISOL (AJ , H, D) in Q0L (AJ , H, D).

The rest of the proof is same as that of Theorem 4.3.8 by using Lemma 4.4.6 and hence
we omit it. 2

4.4.2 Computations
The case of the noncommutative tori

We recall ( chapter 1 ) that the noncommutative n-tori Tnθ is the universal C ∗ algebra
generated by n unitaries U1 , U2 , ..., Un such that Ui Uj = e(θij )Uj Ui , i, j, = 1, 2, ...n
where θ ≡ ((θij )) is a n × n skew symmetric matrix. We also recall that Tnθ is a Rieffel
deformation of C(Tn ) via the Rn action on C(Tn ) ( section 1.3 ) and the matrix J = 2θ .
We consider the isospectral deformation of the classical spectral triple on C(Tn )
which will give a spectral triple on Tnθ . so that the corresponding Laplacian L is given
by L(U1m1 ...Unmn ) = −(m21 + ...m2n )U1m1 ...Unmn , and it is also easy to see that all the
assumptions in chapter 2 required for defining QISOL (Tnθ ) are satisfied.
Theorem 4.4.8. Using Theorem 4.4.7, we conclude that QISOL (Tnθ ) is a Rieffel de-
formation of QISOL (C(Tn )).

Next, we will use Theorem 4.4.7 to compute the exact structure of QISOL (Aθ ).
Our target is to show that QISOL (A 1 ) is a commutative C ∗ algebra which will give an
2
example of a non commutative C ∗ algebra with QISOL a commutative CQG.
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 132

We have seen ( Theorem 2.2.17 ) that QISOL (C(Tn )) ∼ = C(Tn >
(Zn2 >
Sn )).
In particular, QISOL (C(T2 )) ∼ = C(T2 >
(Z22 >
S2 )) where the definition of the
two semi-direct products are described below. Let the generators of the first, second
and the third copy of Z2 in Z22 >
Z2 be denoted by (1, 0, 0), (0, 1, 0), (0, 0, 1). Therefore,
as a set (Z22 >
Z2 ) equals {(γ1 , γ2 , γ3 ) : γi ∈ {0, 1}, i = 1, 2, 3} and T2
(Z22 >
Z2 ) =
{(z1 , z2 , γ1 , γ2 , γ3 ) : z1 , z2 ∈ S 1 , γi ∈ {0, 1}, i = 1, 2, 3}. The action of Z2 on Z22 is
given by (0, 0, 1)(x, y) = (y, x) ( ( x, y ) belongs to T2 ) and the action of Z22 >
Z2 on
T2 ( denoted by ◦ ) is given by

(1, 0, 0) ◦ (z1 , z2 ) = (z1 , z2 ), (0, 1, 0) ◦ (z1 , z2 ) = (z1 , z2 ), (0, 0, 1) ◦ (z1 , z2 ) = (z2 , z1 ).

Lemma 4.4.9. Let (z1 , z2 ), (z10 , z20 ), (z100 , z200 ) belong to T2 . Then we have the following
group multiplication formulae in T2
(Z22 >
Z2 ) :

(z1 , z2 , 0, 0, 0)(z10 , z20 , 1, 0, 1)(z100 , z200 , 0, 0, 0) = (z1 z10 z200 , z2 z20 z100 , 1, 0, 1), (4.4.1)

(z1 , z2 , 0, 0, 0)(z10 , z20 , 0, 0, 1)(z100 , z200 , 0, 0, 0) = (z1 z10 z200 , z2 z20 z100 , 0, 0, 1), (4.4.2)

(z1 , z2 , 0, 0, 0)(z10 , z20 , 0, 0, 0)(z100 , z200 , 0, 0, 0) = (z1 z10 z100 , z2 z20 z200 , 0, 0, 0). (4.4.3)

Proof : Using the definition of semi direct product , we have

(z1 , z2 , 0, 0, 0)(z10 , z20 , 1, 0, 1)(z100 , z200 , 0, 0, 0)

= (z1 z10 , z2 z20 , 1, 0, 1)(z100 , z200 , 0, 0, 0)
= (z1 z10 , z2 z20 )((1, 0, 1) ◦ (z100 , z200 ), 1, 0, 1)
= (z1 z10 , z2 z20 )(((1, 0), 0)((0, 0), 1) ◦ (z100 , z200 ), 1, 0, 1)
= ((z1 z10 , z2 z20 )(z200 , z100 ), 1, 0, 1)
= (z1 z10 z200 , z2 z20 z100 , 1, 0, 1).

Thus, we have ( 4.4.1 ).

Similarly,

(z1 , z2 , 0, 0, 0)(z10 , z20 , 0, 0, 1)(z100 , z200 , 0, 0, 0)

= (z1 z10 , z2 z20 , 0, 0, 1)(z100 , z200 , 0, 0, 0)
= ((z1 z10 , z2 z20 )((0, 0, 1) ◦ (z100 , z200 )), 0, 0, 1)
= ((z1 z10 , z2 z20 )(z200 , z100 ), 0, 0, 1)
= (z1 z10 z200 , z2 z20 z100 , 0, 0, 1),
133 QISOL of a Rieffel deformed noncommutative manifold

which proves ( 4.4.2 ).

( 4.4.3 ) follows similarly and therefore we omit the proof. 2

QISOL (C(T2 )) = C(T2 >
(Z22 >
Z2 )) which as a C ∗ algebra is isomorphic to

⊕i=1,2,...8 C(T2 ). We will denote the generators of QISOL (T2 ) by {A(γ1 ,γ2 ,γ3 ) , B(γ1 ,γ2 ,γ3 ) :
(γ1 , γ2 , γ3 ) ∈ Z32 }, more precisely, we fix the following convention:

A(γ1 ,γ2 ,γ3 ) (z1 , z2 , γ1 , γ2 , γ3 ) = z1 , B(γ1 ,γ2 ,γ3 ) (z1 , z2 , γ1 , γ2 , γ3 ) = z2 .

Now, T2 sits as a subgroup of T2 >
Z23 as {(z1 , z2 , 0, 0, 0) : zi ∈ S 1 }.

Hence, the action of T2 on T2 >
Z23 with which we are concerned with is given
by the group multiplication in T2 >
Z23 , that is, the action of (z1 , z2 ) ((zi ∈ S 1 )) on
(z10 , z20 , γ1 , γ2 , γ3 )(∈ T2 >
Z23 ) is given by (z1 , z2 , 0, 0, 0)(z10 , z20 , γ1 , γ2 , γ3 ).
The action of R4 on QISOL (C(T2 )) as prescribed by Wang ( 1.3.8 ) is given by

α(s,u) = (Ω(−s) ⊗ id)∆(id ⊗ Ω(u))∆

where s, u belong to R2 .
If s = (s1 , s2 ), u = (u1 , u2 ) belong to R2 and z1 , z2 belong to S 1 , we have
α(s,u) f (z1 , z2 , γ1 , γ2 , γ3 ) = f ((e(−s1 ), e(−s2 ), 0, 0, 0)(z1 , z2 , γ1 , γ2 , γ3 )(e(u1 ), e(u2 ),
0, 0, 0)).  
0 2θ 0 0
 θ 
 −2 0 0 0 
Moreover, J = −J ⊕ J = 
e 
θ 
.
 0 0 0 − 2 
0 0 2θ 0
Hence, J(s e 1 , s2 , t1 , t2 )t = ( θ s2 , − θ s1 , − θ t2 , θ t1 )t where t stands for the transpose of
2 2 2 2
a matrix.
Let s = (s1 , s2 ), t = (t1 , t2 ), s0 = (s01 , s02 ), t0 = (t01 , t02 ) belong to R2 u = (s, t), v =
(s0 , t0 ).

Lemma 4.4.10. {A(γ1 ,γ2 ,γ3 ) , B(γ1 ,γ2 ,γ3 ) } are unitaries for the multiplication ×Je and
for (γ1 , γ2 , γ3 ) belonging to Z32 .

Proof : We give the proof for A(1,0,1) , the proof for the rest being exactly similar.

A∗(1,0,1) ×Je A1,0,1 (u1 , u2 , 1, 0, 1)

Z Z
∗
= χJ(u)
e (A(1,0,1) )(u1 , u2 , 1, 0, 1)χv (A1,0,1 )(u1 , u2 , 1, 0, 1)

e(u.v)dudv
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 134

Z Z
= χ( θ s2 ,− θ s1 ,− θ t2 , θ t1 ) A∗(1,0,1) (u1 , u2 , 1, 0, 1)χ(s01 ,s02 ,t01 ,t02 ) A(1,0,1) (u1 ,
2 2 2 2

u2 , 1, 0, 1)e(u.v)dudv
Z Z
θ θ θ θ
= A∗(1,0,1) ((e(− s2 ), e( s1 ))(u1 , u2 , 1, 0, 1)(e(− t2 ), e( t1 )))
2 2 2 2
A(1,0,1) ((e(−s01 ), e(−s02 ))(u1 , u2 , 1, 0, 1)(e(t01 ), e(t02 )))e(u.v)dudv

which by ( 4.4.1 ) equals

Z Z
θ θ θ θ
A∗(1,0,1) (e(− s2 )u1 e(− t1 ), e( s1 )u2 e(− t2 ))A(1,0,1) (e(−s01 )u1 e(−t02 ),
2 2 2 2
e(−s02 )u2 e(t01 ))e(u.v)dudv
Z Z
θ θ
= ... e( s2 )e( t1 )e(−s01 )e(−t02 )e(s1 .s01 + t1 .t01 + s2 .s02 + t2 .t02 )ds1 ds01 ds2 ds02 dt1
2 2
dt01 dt2 dt02
Z Z Z Z
0 0 0 θ
= e(−s1 )e(s1 .s1 )ds1 ds1 e( s2 )e(s2 .s02 )ds2 ds02
2
Z Z
θ
e( t1 )e(t1 .t01 )dt1 dt01 e(−t02 )e(t2 .t02 )dt2 dt02
RR
2

which by Proposition 1.3.1 equals

1.1.1.1 = 1

Similarly,

A(1,0,1) ×Je A∗1,0,1 (u1 , u2 , 1, 0, 1) = 1

Remark 4.4.11. It can be easily checked that

A∗γ1 ,γ2 ,γ3 ×Je Aγ1 ,γ2 ,γ3 (u1 , u2 , γ10 , γ20 , γ30 ) = 0

if (γ1 , γ2 , γ3 ) 6= (γ10 , γ20 , γ30 ).

Similar is the case with Bγ1 ,γ2 ,γ3 .

Lemma 4.4.12. A000 ×Je B000 = B000 ×Je A000 .

135 QISOL of a Rieffel deformed noncommutative manifold

Proof :

(A000 ×Je B000 )((u1 , u2 , 0, 0, 0))

Z Z
= χJu
e (A000 )((u1 , u2 , 0, 0, 0))χv (B000 )((u1 , u2 , 0, 0, 0))e(u.v)dudv
Z Z
= ..... χJ(s e 1 ,s2 ,t1 ,t2 ) A000 ((u1 , u2 , 0, 0, 0))χ(s01 ,s02 ,t01 ,t02 ) (B000 )((u1 , u2 , 0, 0, 0))

e(u.v)dsdtds0 dt0
Z Z
= ... χ( θ s2 ,− θ s1 ,− θ t2 , θ t1 ) A000 ((u1 , u2 , 0, 0, 0))χ(s01 ,s02 ,t01 ,t02 ) (B000 )((u1 , u2 , 0, 0, 0))
2 2 2 2
0 0
e(u.v)dsdtds dt
Z Z
θ θ θ θ
= ... A000 [(e(− s2 ), e( s1 ))(u1 , u2 , 0, 0, 0)(e(− t2 ), e( t1 ))]
2 2 2 2
0 0 0 0 0 0
B000 [(e(−s1 ), e(−s2 ))(u1 , u2 , 0, 0, 0)(e(t1 ), e(t2 ))]e(u.v)dsdtds dt

which by ( 4.4.3 ) equals

... A000 [e(− 2θ s2 )u1 e(− 2θ t2 ), e( 2θ s1 )u2 e( 2θ t1 ), 0, 0, 0]B000 [(e(−s01 )u1 e(t01 )), (e(−s02 )
R R

u2 e(t02 )), 0, 0, 0]e(u.v)dsdtds0 dt0

Z Z
θ θ
= e(− s2 )u1 e(− t2 )e(−s02 )u2 e(t02 )e(s1 s01 )e(s2 s02 )e(t1 t01 )e(t2 t02 )dsdtds0 dt0
....
2 2
Z Z Z Z
θ θ
= u1 u2 e(− s2 )e(−s02 )e(s2 .s02 )ds2 ds02 e(− t2 )e(t02 )e(t2 .t02 )dt2 dt02
2 2
Z Z Z Z
e(s1 .s01 )ds1 ds01 e(t1 .t01 )dt1 dt01

which by Corollary 1.3.4 and Proposition 1.3.1

θ θ
= u1 u2 e(− )e( ).1.1
2 2
= u1 u2 .

Similarly,

(B000 ×Je A000 )((u1 , u2 , 0, 0, 0))

Z Z
θ θ θ θ
= .... B000 [(e(− s2 )u1 e(− t2 )), (e( s1 )u2 e( t1 )), 0, 0, 0)]
2 2 2 2
A000 [(e(−s1 )u1 e(t1 )), (e(−s2 )u2 e(t2 ))0, 0, 0)]e(u.v)dsdtds0 dt0
0 0 0 0
Z Z
θ θ
= ... e( s1 )u2 e( t1 )e(−s01 )u1 e(t01 )e(u.v)dsdtds0 dt0
2 2
Z Z Z Z
θ θ
= u1 u2 e( s1 )e(−s01 )e(s1 .s01 )ds1 ds01 e( t1 )e(t01 )e(t1 .t01 )dt1 dt01
2 2
Z Z Z Z
e(s2 s02 )ds2 ds02 e(t2 t02 )dt2 dt02 ,
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 136

which by Corollary 1.3.4 and Proposition 1.3.1

= u1 u2 .

Therefore, A000 ×Je B000 = B000 ×Je A000 . 2

Lemma 4.4.13. A001 ×Je B001 = e(−2θ)B001 ×Je A001

Proof : Proceeding exactly as in the previous Lemma, we obtain

(A001 ×Je B001 )(u1 , u2 , 0, 0, 1)

Z Z
θ θ θ θ
= .... A001 [(e(− s2 ), e( s1 ))(u1 , u2 , 0, 0, 1)(e(− t2 ), e( t1 ))]
2 2 2 2
B001 [(e(−s01 ), e(−s02 ))(u1 , u2 , 0, 0, 1)(e(t01 ), e(t02 ))]e(u.v)dsdtds0 dt0 ,

which by ( 4.4.2 ) equals

Z Z
θ θ θ θ
.... A001 [e(− s2 )u1 e( t1 ), e( s1 )u2 e(− t2 ), 0, 0, 1]
2 2 2 2
B001 [e(−s01 )u1 e(t02 ), e(−s02 )u2 e(t01 ), 0, 0, 1]e(u.v)dsdtds0 dt0
Z Z
θ θ
= u1 u2 ... e(− s2 )e( t1 )e(−s02 )e(t01 )e(u.v)dsdtds0 dt0
2 2
Z Z Z Z
θ 0 0 0 θ
= u1 u2 e(− s2 )e(−s2 )e(s2 .s2 )ds2 ds2 e( t1 )e(t01 )e(t1 .t01 )dt1 dt01
2 2
Z Z Z Z
0 0
e(s1 .s1 )ds1 ds1 e(t2 .t02 )dt2 dt02

which by Corollary 1.3.4 and Proposition 1.3.1 equals

θ θ
u1 u2 e(− )e(− )
2 2
= u1 u2 e(−θ).

Similarly, using Corollary 1.3.4 and Proposition 1.3.1

(B001 ×Je A001 )(u1 , u2 , 0, 0, 1)

Z Z
θ θ θ θ
= .... B001 [e(− s2 )u1 e( t1 ), e( s1 )u2 e(− t2 ), 0, 0, 1]
2 2 2 2
A001 [e(−s01 )u1 e(t02 ), e(−s02 )u2 e(t01 ), 0, 0, 1]e(u.v)dsdtds0 dt0
Z Z
θ θ
= u1 u2 .... e( s1 )e(− t2 )e(−s01 )e(t02 )e(s1 .s01 + s2 .s02 + t1 .t01 + t2 .t02 )dsdtds0 dt0
2 2
θ θ
= u1 u2 e( )e( )
2 2
= u1 u2 e(θ).
137 QISOL of a Rieffel deformed noncommutative manifold

This proves the Lemma. 2

Remark 4.4.14. Proceeding similarly, one can prove A0,1,1 ×Je B0,1,1 = B0,1,1 ×Je
A0,1,1 , A1,1,0 ×Je B1,1,0 = B1,1,0 ×Je A1,1,0 , A1,0,1 ×Je B1,0,1 = B1,0,1 ×Je A1,0,1
and A0,0,1 ×Je B0,0,1 = e(−2θ)B0,0,1 ×Je A0,0,1 , A1,0,0 ×Je B1,0,0 = e(−2θ)B1,0,0 ×Je
A1,0,0 , A1,1,1 ×Je B1,1,1 = e(−2θ)B1,1,1 ×Je A1,1,1 .

Let us now consider a C ∗ algebra B, which has eight direct summands, four of which
are isomorphic with the commutative algebra C(T2 ), and the other four are irrational
rotation algebras. More precisely, we take

B = ⊕8k=1 C ∗ (Uk1 , Uk2 ),

where for odd k, Uk1 , Uk2 are the two commuting unitary generators of C(T2 ), and for
even k, Uk1 Uk2 = e(2πiθ)Uk2 Uk1 , that is they generate A2θ .

Now we are in a position to describe QISOL (Aθ ) explicitly.

Theorem 4.4.15. QISOL (Aθ ) is isomorphic with B = C(T2 ) ⊕ A2θ ⊕ C(T2 ) ⊕ A2θ ⊕
C(T2 ) ⊕ A2θ ⊕ C(T2 ) ⊕ A2θ .

Proof : Define φ : B → QISOL (Aθ ) by

φ(U11 ) = A000 , φ(U12 ) = B000 , φ(U21 ) = A010 , φ(U22 ) = A010 , φ(U31 ) =
A011 , φ(U32 ) = B011 , φ(U41 ) = A001 , φ(U42 ) = B001 , φ(U51 ) = A110 , φ(U52 ) =
B110 , φ(U61 ) = A111 , φ(U62 ) = B111 , φ(U71 ) = A101 , φ(U72 ) = B101 , φ(U81 ) =
A100 , φ(U82 ) = B100 .
From Lemma 4.4.10 - Lemma 4.4.13 and Remark 4.4.14, it is clear that φ is indeed
a ∗ isomorphism.
2

Remark 4.4.16. In particular, we note that if θ is taken to be 1/2, then we have a com-
mutative compact quantum group as the quantum isometry group of a noncommutative
C ∗ algebra.

The case of θ deformed spheres

We can apply Theorem 4.4.7 on Sθn which are obtained by Rieffel deformation of
C(S n ) as described in section 1.3. We will consider the isospectral deformation of
the classical spectral triple on C(S n ). Since we have proved in Theorem 2.2.2 that
QISOL (S n ) ∼= C(O(n)), it will follow that QISOL (Sθn ) ∼
= Oθ (n), where Oθ (n) is the
compact quantum group discussed in subsection 1.3.1 as the θ-deformation of C(O(n)).
Chapter 4: Quantum isometry groups for Rieffel deformed manifolds 138
Chapter 5

Quantum isometry groups of the

Podles spheres

In this chapter, we compute quantum group of orientation preserving isometries for

2 . We do it for two different families of spectral
spectral triples on the Podles spheres Sµ,c
triples, one constructed by Dabrowski et al in [24] and the other by Chakraborty and
Pal in [14] for c > 0. We get completely different kinds of quantum group of orientation
preserving isometries for the two families; for the former, it is SOµ (3), whereas, for the
latter it is C ∗ (Z2 ∗ Z∞ ) where Z∞ denotes countably infinite copies of the group of
integers.

5.1 The Podles Spheres

The Podles spheres were discovered in [43]. We will also need the equivalent descriptions
of the Podles spheres as given in [24], [37] and [50]. Hence, we give the descriptions one
by one.

5.1.1 The original definition by Podles

2 is the universal C ∗ algebra generated by
Let c belongs to R. The Podles’ sphere Sµ,c
elements e−1 , e0 , e1 such that :

e∗i = e−i , i = −1, 0, 1,

2
(1 + µ2 )(e−1 e1 + µ−2 e1 e−1 ) + e0 2 = ((1 + µ2 ) µ−2 c + 1)I,

e0 e−1 − µ2 e−1 e0 = (1 − µ2 )e−1 ,

(1 + µ2 )(e−1 e1 − e1 e−1 ) + (1 − µ2 )e0 2 = (1 − µ2 )e0 ,

139
Chapter 5: Quantum isometry groups of the Podles spheres 140

e1 e0 − µ2 e0 e1 = (1 − µ2 )e1 .

Let
−1
B = e1 , A = (1 + µ2 ) (1 − e0 ). (5.1.1)

Then we have an alternate description of the Podles spheres, that is, the universal
C∗ algebra generated by elements A and B satisfying the relations:

A∗ = A, AB = µ−2 BA,

B ∗ B = A − A2 + cI, BB ∗ = µ2 A − µ4 A2 + cI.

These are the relations which we are going to exploit for getting homomorphism
conditions while computing the quantum group of orientation preserving isometries.

2 by O(S 2 ).
Notation : We will denote the co-ordinate ∗ algebra of Sµ,c µ,c

5.1.2 The Podles spheres as in [24]

−µn −n
We take µ in (0, 1) and t in (0, 1]. For n belonging to R, let [n]µ = µµ−µ−1 .
2 ∗
Then Sµ,c be the universal C algebra generated by elements x−1 , x0 , x1 satisfying
the relations:
x−1 (x0 − t) = µ2 (x0 − t)x−1 , (5.1.2)

x1 (x0 − t) = µ−2 (x0 − t)x1 , (5.1.3)

− [2]x−1 x1 + (µ2 x0 + t)(x0 − t) = [2]2 (1 − t), (5.1.4)

− [2]x1 x−1 + (µ−2 x0 + t)(x0 − t) = [2]2 (1 − t), (5.1.5)

where c = t−1 − t, t > 0.

2 is given by
The involution on Sµ,c

x∗−1 = −µ−1 x1 , x∗0 = x0 . (5.1.6)

Setting

1 − t−1 x0 1
A= 2
, B = µ(1 + µ2 )− 2 t−1 x−1 , (5.1.7)
1+µ
2 is the same as the Podles’ sphere as in [43].
one obtains ( [24] ) that Sµ,c
141 The Podles Spheres

5.1.3 The Podles spheres as in [37]

This description is taken from [37], page 124.
Let α0 , β be elements of C such that (α0 , β) 6= (0, 0). We denote by χq,α0 ,β the
algebra with three generators x−1 , x0 , x1 and the following defining relations:

x20 − qx1 x−1 − q −1 x−1 x1 = β.1, (5.1.8)

(1 − q 2 )x20 + qx−1 x1 − qx1 x−1 = (1 − q 2 )α0 x0 , (5.1.9)

x−1 x0 − q 2 x0 x−1 = (1 − q 2 )α0 x−1 , (5.1.10)

x0 x1 − q 2 x1 x0 = (1 − q 2 )α0 x1 . (5.1.11)
−1
Let ρ2 = α2 (β − α2 ) .
Then for q and ρ real, the involution is defined by x∗−1 = −q −1 x1 , x∗0 = x0 , x∗1 =
−qx−1 .
Moreover, from page 125 of [37], it follows that for ρ belonging to C, χq,α0 ,β can be
realized as a ∗ subalgebra of SUµ (2) via:

− 21 2 1
− 21 2
x−1 = (1 + q 2 ) a + ρ(1 + q −2 ) 2 ac − q(1 + q 2 ) c , (5.1.12)

x0 = ba + ρ(1 + (q + q −1 )bc) − cd, (5.1.13)

− 12 2 1
− 12 2
x1 = (1 + q 2 ) b + ρ(1 + q 2 ) 2 db − q(1 + q 2 ) d . (5.1.14)

where a, b, c, d are as in subsection 1.2.5.

2 as defined above is the same as χ 0 with q = µ, α0 = t, β =
Proposition 5.1.1. Sµ,c q,α ,β
2 −2 2 2
t + µ (µ + 1) (1 − t).
µ2 −µ−2 µ2 +1
Proof : We note that [2]µ = µ−µ−1
= µ .
From ( 5.1.5 ), we have

− [2]µ2 x1 x−1 + x20 + (tµ2 − t)x0 − µ2 t2 = [2]2 µ2 (1 − t). (5.1.15)

Adding ( 5.1.4 ) with ( 5.1.15 ), we obtain

−[2]x−1 x1 − [2]µ2 x1 x−1 + (µ2 + 1)x20 − (1 + µ2 )t2 = [2]2 (1 − t)(1 + µ2 ).
2
Using [2]µ = 1+µµ , we get this to be
2
2 1+µ2 2 (1+µ2 )
− 1+µ
µ x−1 x1 − µ µ x1 x−1 + (1 + µ2 )x20 − (1 + µ2 )t2 = µ2
(1 − t)(1 + µ2 ).
Hence,

2
x20 − µx1 x−1 − µ−1 x−1 x1 = µ−2 (1 + µ2 ) (1 − t) + t2 = β.1.
Chapter 5: Quantum isometry groups of the Podles spheres 142

Thus, we have derived ( 5.1.8 ).

Multiplying (5.1.4) by µ2 , we have

− [2]µ2 x−1 x1 + µ4 x20 + t(1 − µ2 )µ2 x0 − t2 µ2 = [2]2 µ2 (1 − t). (5.1.16)

Subtracting (5.1.15) from (5.1.16) gives

µ2 + 1 2 µ2 + 1 2
− µ x−1 x1 + (µ4 − 1)x20 + t(1 − µ2 )µ2 x0 + µ x1 x−1 − t(µ2 − 1)x0 = 0,
µ µ

and hence,
(1 − µ2 )x20 + µx−1 x1 − µx1 x−1 = t(1 − µ2 )x0 .

which proves ( 5.1.9 ).

Next, (5.1.2) gives
x−1 x0 − µ2 x0 x−1 = (1 − µ2 )tx−1

which is (5.1.10).
Finally, (5.1.3) gives µ2 x1 x0 − µ2 tx1 = x0 x1 − tx1 , that is, (5.1.11) is obtained.
Thus, we obtain the relations of χq,α0 ,β from the relations of Sµ,c 2 for q = µ, α0 =
2
t, β = t2 + µ−2 (µ2 + 1) (1 − t).
Similarly, ( 5.1.10 ) is same as ( 5.1.2 ), ( 5.1.11 ) is same as ( 5.1.3 ). Subtracting
( 5.1.9 ) from ( 5.1.8 ) gives ( 5.1.4 ) and adding ( 5.1.8 ) with µ−2 times ( 5.1.9 ) gives
( 5.1.5 ). Thus, we get back the relations of Sµ,c 2 from the relations of χ 0 .
q,α ,β 2
2

Thus, combining ( 5.1.12 ) - ( 5.1.14 ) with the correspondence ( 1.2.26 ) and using
Proposition 5.1.1, we have expressions of x−1 , x0 , x1 in terms of SUµ (2) elements:

µα2 + ρ(1 + µ2 )αγ − µ2 γ 2

x−1 = 1 , (5.1.17)
µ(1 + µ2 ) 2

x0 = −µγ ∗ α + ρ(1 − (1 + µ2 )γ ∗ γ) − γα∗ , (5.1.18)

µ2 γ ∗2 − ρµ(1 + µ2 )α∗ γ ∗ − µα∗2
x1 = 1 , (5.1.19)
(1 + µ2 ) 2
µ2 t2
where ρ2 = (µ2 +1)2 (1−t)
.

5.1.4 The description as in [50]

We need the quantum group Uµ (su(2)) for this description.
143 The Podles Spheres

Let
1 −1 − 1
Xc = µ 2 (µ−1 − µ) c 2 (1 − K 2 ) + EK + µF K for all c ∈ (0, ∞),

X0 = 1 − K 2 .

One has ∆(Xc ) = 1 ⊗ Xc + Xc ⊗ K 2 .

Then ( as in Page 9, [50] ) we have the following :

2
O(Sµ,c ) = {x ∈ O(SUµ (2)) : x / Xc = 0}

2 ):
where / is as in subsection 1.2.5. The following is a basis of the vector space O(Sµ,c

{Ak , Ak B l , Ak B ∗ m , k ≥ 0, l, m > 0}.

So, any element of O(Sµ,c 2 ) can be written as a finite linear combination of elements of

the form Ak , Ak B l , Ak B ∗ l .
Let ψ be the densely defined linear map on L2 (SUµ (2)) defined by ψ(x) = x / Xc .
l /E = µαl l l αlk l l /
From [51], ( Page 5 ), we recall that vj,k k−1 vj,k−1 , vj,k /F = µ vj,k+1 , vj,k
l . for some constants αl .
K = µk vj,k j

Lemma 5.1.2. The map ψ is closable and we have Sµ,c2 ⊆ Ker(ψ) where ψ is the closed

2 denotes the Hilbert subspace generated by S 2 in L2 (SU (2))

extension of ψ and Sµ,c µ,c µ
2 ) =
( the G.N.S space corresponding to the Haar state on SUµ (2)). Moreover, O(Sµ,c
T T
O(SUµ (2)) Ker(ψ) = O(SUµ (2)) Ker(ψ).
l / (1 − K 2 ) = (1 − µ2k )v l , v l / EK = µk αl
Proof : We note that vj,k l l
j,k j,k k−1 vj,k−1 , vj,k /
1
l belongs to Dom(ψ) with (ψ ∗ )(v l ) = µ 2 (µ−1 − µ) c− 2 . −1 1
l
F K = µk αkl vj,k+1 . Hence, vj,k j,k
(1 − µ2k )vj,k
l + µk α l v l
k−1 j,k+1 + µ k+1 αl v l
k j,k−1 . Hence, O(SUµ (2)) ⊆ Dom(ψ ∗ ) implying
that ψ is closable, hence Ker(ψ) is closed. The lemma now follows from the observation
2 ) = Ker(ψ) ⊆ Ker(ψ).
that O(Sµ,c 2

5.1.5 Haar functional on the Podles spheres

We recall from [50], page- 33 that for all bounded complex Borel function f on σ(A),
∞
X ∞
X
2n 2n
h(f (A)) = γ+ f (λ+ µ )µ + γ− f (λ− µ2n )µ2n . (5.1.20)
n=0 n=0
1 1
where λ+ = 12 + (c + 14 ) 2 , λ− = 1
2 − (c + 14 ) 2 , γ+ = (1 − µ2 )λ+ (λ+ − λ− )−1 , γ− =
(1 − µ2 )λ− (λ− − λ+ )−1 .
Chapter 5: Quantum isometry groups of the Podles spheres 144

Lemma 5.1.3. {x−1 , x0 , x1 } is a set of orthogonal vectors.

Proof : From ( 5.1.17 ) and ( 5.1.18 ), we note that x∗−1 x0 belongs to span{α∗2 γ ∗ α,
α∗2 , α∗2 γ ∗ γ, α∗2 γα∗ , γ ∗ α∗ γ ∗ α, γ ∗ α∗ , γ ∗ α∗ γ ∗ γ, γ ∗ α∗ γα∗ , γ ∗3 α, γ ∗2 , γ ∗3 γ, γ ∗2 γα∗ }.
Further, using ( 1.2.10 ) - ( 1.2.14 ), we note that this span equals span{α∗ γ ∗ , α∗ γ ∗2 γ,
α∗2 , α∗2 γ ∗ γ, α∗3 γ, γ ∗2 , γ ∗3 γ, α∗ γ ∗ , αγ ∗3 , γ ∗3 }.
Then h(x∗−1 x0 ) = 0 follows by using ( 1.2.15 ). Similarly, one can prove h(x∗0 x1 ) =
h(x∗1 x−1 ) = 0. 2

Lemma 5.1.4.
1 2 (1 − µ2 )(λ3+ − λ3− )
h(A) = , h(A ) = .
1 + µ2 (λ+ − λ− )(1 − µ6 )
Proof : Recalling ( 5.1.20 ), we have

h(A)
∞
X ∞
X
= γ+ λ+ µ4n + γ− λ− µ4n
n=0 n=0
(1 − µ2 )(λ2+ − λ2− )
=
(λ+ − λ− )(1 − µ4 )
λ+ + λ−
=
1 + µ2
1
= .
1 + µ2

Similarly, putting f (x) = x2 in ( 5.1.20 ), we have

h(A2 )
∞
X ∞
X
= γ+ λ2+ µ6n + γ− λ2− µ6n
n=0 n=0
(1 − µ2 )(λ3+ 3
− λ− )
= .
(λ+ − λ− )(1 − µ6 )

−1
Proposition 5.1.5. h(x∗−1 x−1 ) = h(x∗0 x0 ) = h(x∗1 x1 ) = t2 (1 − µ2 )(1 − µ6 ) [µ2 +
t−1 (µ4 + 2µ2 + 1) + t(−µ4 − 2µ2 − 1)].

Proof :
t2 (1+µ2 ) ∗
From ( 5.1.7 ) we have x∗−1 x−1 = µ2
B B and hence, using Lemma 5.1.4, we
145 Spectral triples on the Podles spheres

obtain

h(x∗−1 x−1 )
t2 (1 + µ2 )
= [h(A) − h(A2 ) + (t−1 − t).1]
µ2

1 − µ6 − (1 − µ4 )(t−1 − t + 1) + (t−1 − t)(1 + µ2 )(1 − µ6 )

= ,
(1 + µ2 )(1 − µ6 )
from which we get the desired result.
Similarly, the second equality is derived from h(x∗0 x0 ) = t2 − 2t2 (1 + µ2 )h(A) +
2
(1 + µ2 ) t2 h(A2 ).
From ( 5.1.6 ), x1 = −µx∗−1 and hence, x∗1 x1 = t2 (1 + µ2 )(µ2 A − µ4 A2 + c.I),
from which h(x∗1 x1 ) is obtained and can be shown to be equal to the same value as
h(x∗−1 x−1 ) = h(x∗0 x0 ).
2

5.2 Spectral triples on the Podles spheres

5.2.1 The spectral triple as in [24]
2 discussed in [24].
We now recall the spectral triples on Sµc
1 1
Let s = −c− 2 λ− , λ± = 12 ± (c + 14 ) 2 .
For all j belonging to 12 IN ,
uj = (α∗ − sγ ∗ )(α∗ − µ−1 sγ ∗ )......(α∗ − µ−2j+1 sγ ∗ ),
wj = (α − µsγ)(α − µ2 sγ)........(α − µ2j sγ),
u−j = E 2j . wj ,
u0 = w0 = 1,
1 1 1
y1 = (1 + µ−2 ) 2 (c 2 µ2 γ ∗2 − µγ ∗ α∗ − µc 2 α∗2 ),
l = F l−k . (y l−|j| u ) −1 .

Nkj 1 j
Define

l l l−|j| 1
vk,j = Nk,j F l−k . (y1 uj ), l ∈ IN 0 , j, k = −l, −l + 1, ......l. (5.2.1)
2

Let MN be the Hilbert subspace of L2 (SUµ (2)) with the orthonormal basis {vm,N
l :
l = |N | , |N | + 1, ........, m = −l, .......l}.
Set
H = M− 1 ⊕ M 1 ,
2 2
Chapter 5: Quantum isometry groups of the Podles spheres 146

2 on H by
and define a representation π of Sµ,c

l
π(xi )vm,N = αi− (l, m; N )vm+i,N
l−1
+ αi0 (l, m; N )vm+i,N
l
+ αi+ (l, m; N )vm+i,N
l+1
, (5.2.2)

where αi− , αi0 , αi+ are some constants.

2 ) with S 2 .
We will often identify π(Sµ,c µ,c
Finally by Proposition 7.2 of [24], the following Dirac operator D gives a spectral
2 ), H, D) which we are going to work with :
triple (O(Sµ,c

l l
D(vm,± 1 ) = (c1 l + c2 )v
m,∓ 1
, (5.2.3)
2 2

where c1 , c2 are elements of R, c1 6= 0.

5.2.2 SUµ (2) equivariance of the spectral triple

From [24], we see that the vector spaces ν± l l : m = −l, ....l} are (2l + 1)
1 = span{v
2
m,± 12
dimensional Hilbert spaces on which the SUµ (2) representation is unitarily equivalent
to the standard lth unitary irreducible representation of SUµ (2), that is, if the represen-
l
P l
tation is denoted by U0 , then U0 (vi,± 1) = vj,± 1 ⊗ tli,j where tli,j denotes the matrix
2 2
elements in the lth unitary irreducible representation of SUµ (2).
We now recall Theorem 3.5 of [32].
n
Proposition 5.2.1. Let R0 be an operator on H defined by R0 (vi,± −2i∓1 v n
1) = µ .
i,± 1
2 2
2
Then Tr(R0 e−tD ) < ∞ ( for all t > 0) and one has

(τR0 ⊗ id)(U f0 ∗ ) = τR (x).1,

f0 (x ⊗ 1)U
0

2
for all x in B(H), where τR0 (x) = Tr(xR0 e−tD ).
n
We define a positive, unbounded operator R on H by R(vi,± −2i v n
1) = µ .
i,± 1
2 2

Proposition 5.2.2. αU0 preserves the R-twisted volume. In particular, for x in π(Sµ,c 2 )
τR (x) 2
and t > 0, we have h(x) = τR (1) , where τR (x) := Tr(xRe−tD ), and h denotes the re-
2 , which is the unique SU (2)-
striction of the Haar state of SUµ (2) to the subalgebra Sµ,c µ
2
invariant state on Sµ,c .

Proof : It is enough to prove that τR is αU0 -invariant. Let us denote by P 1 , P− 1

2 2
the projections onto the closed subspaces generated by {vi,l 1 } and {vi,−
l
1 } respectively.
2 2
2
Moreover, let τ± be the functionals defined by τ± (x) = Tr(xR0 P± 1 e−tD ). Now observ-
2
2
ing that R0 , e−tD and U0 commute with P± 1 and using Proposition 5.2.1, we have, for
2
147 Spectral triples on the Podles spheres

x belonging to B(H),

(τ± ⊗ id)(αU0 (x))

= (Tr ⊗ id)(U f0 ∗ (R0 P 1 e−tD2 ⊗ id))
f0 (x ⊗ 1)U
± 2

= (Tr ⊗ id)(U f0 ∗ (R0 e−tD2 ⊗ id))

f0 (xP 1 ⊗ 1)U
± 2

= (τR0 ⊗ id)αU0 (xP 1 )

2

= τR0 (xP± 1 )
2

= τ± (x).1,

that is τ± are αU0 -invariant.

2
Thus, x 7→ Tr(xR0 P± 1 e−tD ) is invariant under αU0 . Moreover, since we have
2
RP± 1 = µ± R0 P± 1 , the functional τR coincides with µ−1 τ+ +µτ− , hence is αU0 -invariant.
2 2
2

Theorem 5.2.3. (SUµ (2), U0 ) is an object in Q0R (D).

Proof :
X
l l l
(D ⊗ id)U0 (vi,± 1 ) = (D ⊗ id) vj,± 1 ⊗ ti,j
2 2
X
l l
= (c1 l + c2 ) vj,∓ 1 ⊗ ti,j
2
l
= (c1 l + c2 )U0 (vi,∓ 1)
2
l
= U0 D(vi,± 1 ).
2

Thus, the above spectral triple is equivariant w.r.t. the representation U0 and it
preserves τR by Proposition 5.2.2, which completes the proof. 2

5.2.3 The CQG SOµ (3) and its action on the Podles sphere
Here we recall the CQG SOµ (3) as described in [44].
SOµ (3) is the universal unital C ∗ algebra generated by elements M, N, G, C, L sat-
isfying :
 
 L∗ L = (I − N )(I − µ−2 N ), LL∗ = (I − µ2 N )(I − µ4 N ), G∗ G = GG∗ = N 2 , 

 

∗ M = N − N 2 , M M ∗ = µ2 N − µ 4 N 2 , C ∗ C = N − N 2 ,
 



 M 



∗ 2 4 2 4
 
 CC = µ N − µ N , LN = µ N L, GN = N G, 


 M N = µ2 N M, CN = µ2 N C, LG = µ4 GL, 


2 2
 



 LM = µ M L, M G = µ GM, CM = M C, 




 ∗ 4 ∗ 2 −1 ∗ −1
LG = µ G L, M = µ LG, M L = µ (I − N )C, N = N. ∗ 

(5.2.4)
Chapter 5: Quantum isometry groups of the Podles spheres 148

This CQG can be identified with a Woronowicz subalgebra of SUµ (2) by taking:

N = γ ∗ γ, M = αγ, C = αγ ∗ , G = γ 2 , L = α2 ,

where α, γ are as in subsection 1.2.4.

2 , that is the action obtained by restricting
The canonical action of SUµ (2) on Sµ,c
2 , is actually a faithful action of SO (3).
the coproduct of SUµ (2) to the subalgebra Sµ,c µ
With respect to the ordered basis {x−1 , x0 , x1 }, this action on the subspace gener-
ated by them is given by the following SOµ (3)-valued 3 × 3-matrix( [37] ):
 1 
(1+q 2 ) 2
a2 q ab b2
1 1
 
(1+q 2 ) 2 (1+q 2 ) 2
q −1 )bc
 

 q ac I + (q + q bd 

1
(1+q 2 ) 2
c2 q cd d2

( where a, b, c, d are as in subsection 1.2.5 ).

Recalling the correspondence in ( 1.2.26 ) and Proposition 5.1.1, the matrix in the
α, γ notation is :
 1 
α2 −µ(1 + µ−2 ) 2 αγ ∗ µ2 γ ∗2
 (1 + µ−2 ) 21 αγ I − µ(µ + µ−1 )γ ∗ γ −µ(1 + µ−2 ) 21 γ ∗ α∗  .
 
 
1
γ 2 −2
(1 + µ ) 2 γα ∗ α ∗2

Finally, in the symbols M, N, G, C, L, the above matrix is :

 1 
L −µ(1 + µ−2 ) 2 C µ2 G∗
 1 1 
Z1 =  −2 2 I − µ(µ + µ−1 )N −µ(1 + µ−2 ) 2 M ∗ 
 (1 + µ ) M . (5.2.5)
1
G (1 + µ−2 ) C ∗
2 L∗

2 , the Woronowicz subalgebra of SU (2) generated by the

As x−1 , x0 , x1 generates Sµ,c µ
elements of the form < ξ ⊗ 1, ∆U (a)(η ⊗ 1) > is SOµ (3) ( where ξ, η are elements of H, a
belongs to O(Sµ,c 2 ) and < ·, · > is the SU (2)-valued inner product of H ⊗ SU (2) ).
µ µ

5.3 Quantum Isometry Groups of the Podles sphere

^+ (S 2 ) with respect to the spectral triple given in [24] and
Here we will compute QISO R µ,c
show that it is isomorphic with SOµ (3).
Let (Q̃, U ) be an object in the category Q0R (D) and Q be the Woronowicz C ∗
subalgebra of Q̃ generated by < (ξ ⊗ 1), αU (a)(η ⊗ 1) >Q̃ , for ξ, η belonging to H, a
2 (where < ·, · > is the Q̃ valued inner product of H ⊗ Q̃). We shall
belongs to Sµ,c Q̃
149 Quantum Isometry Groups of the Podles sphere

denote αU by φ from now on.

The computation has the following steps:
Step 1. We prove that φ is ‘linear’, in the sense that it keeps the span of
{1, A, B, B ∗ } invariant.
Step 2. We shall exploit the facts that φ is a ∗-homomorphism and preserves the
2 , that is the restriction of the Haar state of SU (2).
canonical volume form on Sµ,c µ
Step 3. We will compute the antipode of Q and apply it to get some more relations.
Step 4. We will use the above steps to identify Q as a subobject of SOµ (3) which
will finish the proof.

Remark 5.3.1. The first step does not make use of the fact that α preserves the R-
twisted volume, so linearity of the action follows for any object in the bigger category
Q0 (D).

We now note down some useful facts for later use.

Lemma 5.3.2. We observe :

l
1. The eigenspace of D corresponding to (c1 l + c2 ) and −(c1 l + c2 ) are span{vm, 1 +
2
l l l
vm,− 1 : −l ≤ m ≤ l} and span {v − vm,− 1 : −l ≤ m ≤ l} respectively.
2
m, 1 2 2

2. The eigenspace of |D| corresponding to the eigenvalue (c1 . 21 +c2 ) is span{α, γ, α∗ ,

γ ∗ }.

Proof : 1. follows from ( 5.2.3 ). To prove 2., we note that by 1, it is sufficient to

1 1 1 1
identify span{v−2 1 , 1 , v 12 , 1 , v−2 1 ,− 1 , v 12 ,− 1 }.
2 2 2 2 2 2 2 2
Using ( 5.2.1 ) and ( 1.2.27 ), we have:
1
v−2 1 , 1
2 2
1 1
= N−2 1 , 1 F . (y10 u 1 ) = N−2 1 , 1 F . (α∗ − sγ ∗ )
2 2 2 2 2
1 1 1
−1
2
= N− 1 , 1 (γ + µ sα) = N 2
− 12 , 12
γ + µ−1 sN−2 1 , 1 α.
2 2 2 2

1
v 12 , 1
2 2
1 1
= N 12, 1 F 0 . (y10 u 1 ) = N 12, 1 (α∗ − sγ ∗ )
2 2 2 2 2
1 1
∗ ∗
= N 2
1 1
,
α − sN 2
1 1
,
γ .
2 2 2 2
Chapter 5: Quantum isometry groups of the Podles spheres 150

1
v−2 1 ,− 1
2 2
1 1
= N−2 1 ,− 1 F . (y10 u− 1 ) = N−2 1 ,− 1 F . (E . w 1 )
2 2 2 2 2 2
1 1
= N− 1 ,− 1 F . (E . (α − µsγ)) = N− 1 ,− 1 F . (−µγ ∗ − µsα∗ )
2 2

2 2 2 2
1 1 1
= N− 1 ,− 1 (α − µsγ) = N− 1 ,− 1 α − µsN− 1 ,− 1 γ.
2 2 2

2 2 2 2 2 2

1
v 12 ,− 1
2 2
1 1
= N 12,− 1 F 0 . (y10 u− 1 ) = N 12,− 1 (E . (α − µsγ))
2 2 2 2 2
1 1 1
= N 2
1
,− 12
(−µγ ∗ − µsα∗ ) = − µN 12,− 1 γ ∗ − µsN 12,− 1 α∗ .
2 2 2 2 2

Combining these, we have the result. 2

l l−1 l l+1
Lemma 5.3.3. 1. π(A)vm,N ∈ span{vm,N , vm,N , vm,N },
l l−1 l l+1
π(B)vm,N ∈ span{vm−1,N , vm−1,N , vm−1,N },
l−1 l+1
π(B ∗ )vm,N
l ∈ span{vm+1,N l
, vm+1,N , vm+1,N }.
l−k l−k+1 l+k
2. π(Ak )(vm,N
l ) ∈ span{vm,N , vm,N , ......., vm,N }.
0 0 l−m −n 0 0 l−(n0 +m0 −1) 0
l+n +m 0
3. π(Am B n )(vm,N
l ) ∈ span{vm−n 0 ,N , vm−n0 ,N , ......., vm−n 0 ,N }.

l−s−r l−s−r+1 l+s+r

4. π(Ar B ∗s )(vm,N
l ) ∈ span{vm+s,N , vm+s,N , .......vm+s,N }.

Proof : Using ( 5.1.7 ) and ( 5.2.2 ), we have

l
π(A)vm,N
1 t−1
= v l
m,N − [α− (l, m; N )vm,N
l−1
+ α00 (l, m; N )vm,N
l
+ α0+ (l, m; N )vm,N
l+1
].
1 + µ2 1 + µ2 0

l l l−1 l+1
Thus, π(A)vm,N belongs to span{vm,N , vm,N , vm,N }.
∗
Similarly, using the expressions for B and B from ( 5.1.7 ) and then using ( 5.2.2
) just as above, we get the required statements for π(B) and π(B ∗ ). This proves 1.
Repeated use of 1. now yields 2.,3. and 4. 2

5.3.1 Linearity of the action

For a vector v in H, we shall denote by Tv the map from B(H) to L2 (SUµ (2)) given by
Tv (x) = xv ∈ H ⊂ L2 (SUµ (2)). It is clearly a continuous map w.r.t. the SOT on B(H)
151 Quantum Isometry Groups of the Podles sphere

and the Hilbert space topology of L2 (SUµ (2)).

For an element a in SUµ (2), we consider the right multiplication Ra as a bounded
linear map on L2 (SUµ (2)). Clearly the composition Ra Tv is a continuous linear map
from B(H) (with SOT) to the Hilbert space L2 (SUµ (2)). We now define

T = Rα∗ Tα + µ2 Rγ Tγ ∗ .
Lemma 5.3.4. For any state ω on Q̃ and x belonging to Sµ,c 2 , we have T (φ (x)) =
ω
2 2
φω (x) ≡ R1 (φω (x)) ∈ Sµ,c ⊆ L (SUµ (2)), where φω (x) = (id ⊗ ω)(φ(x)).
Proof : It is clear from the definition of T (using αα∗ + µ2 γγ ∗ = 1) that T (x) = x ≡
2 ⊂ B(H), where x in the right hand side of the above denotes the
R1 (x) for x in Sµ,c
2 as a vector in L2 (SU (2)). Now, the lemma follows by noting
identification of x in Sµ,c µ
that for x belonging to Sµ,c2 , φ (x) belongs to (S 2 )00 , which is the SOT closure of S 2 ,
ω µ,c µ,c
and the SOT continuity of T discussed before. 2

Let
0 0 0 0
V l = Span{vi,±
l
1 , −l ≤ i ≤ l , l ≤ l}.
2

l
As Span{vi,± 1 , −l ≤ i ≤ l}, is the eigenspace of |D| corresponding to the eigenvalue
2
c1 l + c2 , Ũ and Ũ ∗ keep V l invariant for all l.
Lemma 5.3.5. There is some finite dimensional subspace V of O(SUµ (2)) such that
1 1
2
Rα∗ (φω (A)vj,± 1 ), Rγ (φω (A)v
2
j,± 21
) belong to V for all states ω on Q̃.
2
The same holds when A is replaced by B or B ∗ .
Proof : We prove the result for A only, since a similar argument will work for B and
B∗.
1 1
2
We have φ(A)(vj,± e e ∗ 2 1 ⊗ 1).
1 ⊗ 1) = U (π(A) ⊗ 1)U (v
2
j,± 2
1 1
Now, e ∗ (v 2 1
U ⊗ 1) belong to V ⊗ Q̃, and then using the definition of π as well
2
j,± 2
1
e ∗ (v 2 1 ⊗ 1) belong to Span{v l0 1 : −l0 ≤ j ≤
as the Lemma 5.3.3, we get (π(A) ⊗ 1)U j,± j,± 2 2
3 3 1
l0 , l0 ≤ 3
2 } ⊗ Q̃ = V ⊗ Q̃. Again, Ũ keeps V ⊗ Q̃ invariant, so Rα∗ (φω (A)(vj,± 1 )) belong
2 2 2

2
3 1 3
to Span{vα∗ , v ∈ V }. Similarly, Rγ (φω (A)(vj,± 1 )) belong to Span{vγ : v ∈ V 2 }. So,
2 2

2
3
the lemma follows for A by taking V = Span{vα∗ , vγ : v ∈ V 2 } ⊂ O(SUµ (2)). 2

1
Since α, γ ∗ are in Span{vj,±
2
1 }, we have the following immediate corollary:
2

Corollary 5.3.6. There is a finite dimensional subspace V of O(SUµ (2)) such that
for every state (hence for every bounded linear functional) ω on Q̃, we have T (φω (A))
belongs to V. Similar conclusion holds for B and B ∗ as well.
Chapter 5: Quantum isometry groups of the Podles spheres 152

Proposition 5.3.7. φ(A), φ(B), φ(B ∗ ) belong to O(Sµ,c

2 )⊗
alg Q.

Proof : We give the proof for φ(A) only, the proof for B, B ∗ being similar. From the
Corollary 5.3.6 and Lemma 5.3.4 it follows that for every bounded linear functional ω
T 2 T
on Q̃, we have T (φω (A)) belongs to V Sµ,c ⊂ O(SUµ (2)) Ker(ψ) ( by Lemma 5.1.2 )
2 ), where V is the finite dimensional subspace mentioned
T 2 T
and hence V Sµ,c = V O(Sµ,c
2 ) is a finite dimensional subspace of O(S 2 )
T
in Corollary 5.3.6. Clearly, V O(Sµ,c µ,c
implying that there must be finite m, say, such that for every ω, T (φω (A))belongs to
Span{Ak , Ak B l , Ak B ∗l : 0 ≤ k, l ≤ m}. Denote by W the (finite dimensional) subspace
of B(H) spanned by {Ak , Ak B l , Ak B ∗l : 0 ≤ k, l ≤ m}. Since for every state (and hence
for every bounded linear functional) ω on Q̃, we have T (φω (A)) = R1 (φω (A)) ≡ φω (A).1,
it is clear that φω (A) is in W for every ω in Q̃∗ . Now, let us fix any faithful state ω on
the separable unital C ∗ -algebra Q̃ and embed Q̃ in B(L2 (Q, ω)) ≡ B(K). Thus, we get a
canonical embedding of L(H ⊗ Q̃) in B(H ⊗ K). Let us thus identify φ(A) as an element
of B(H ⊗ K), and then by choosing a countable family of elements {q1 , q2 , ...} of Q̃
which is an orthonormal basis in K = L2 (ω), we can write φ(A) as a weakly convergent
series of the form ∞ ij ij
P
i,j=1 φ (A) ⊗ |qi >< qj |. But φ (A) = (id ⊗ ωij )(φ(A)), where
ωij (·) = ω(qi∗ · qj ). So we have φij (A) belongs to W for all i, j, and hence the sequence
Pn ij
i,j=1 φ (A) ⊗ |qi >< qj | belongs to W ⊗ B(K) converges weakly, and W being finite
dimensional ( hence weakly closed ), the limit, that is φ(A), must belong to W ⊗ B(K).
In other words, if A1 , ..., Ak denotes a basis of W, we can write φ(A) = ki=1 Ai ⊗ Bi
P

for some Bi in B(K).

We claim that each Bi must belong to Q̃. For any trace-class positive operator
P
ρ in H, say of the form ρ = j λj |ej >< ej |, where {e1 , e2 , , ...} is an orthonormal
P
basis of H and λj ≥ 0, j λj < ∞, let us denote by ψρ the normal functional on
B(H) given by x 7→ Tr(ρx), and it is easy to see that it has a canonical extension
P
ψ̃ρ := (ψρ ⊗ id) on L(H ⊗ Q̃) given by ψ̃ρ (X) = j λj < ej ⊗ 1, X(ej ⊗ 1) >Q̃ , where X
is in L(H ⊗ Q̃) and < ·, ·, >Q̃ denotes that Q̃ valued inner product of H ⊗ Q̃. Clearly, ψ̃ρ
is a bounded linear map from L(H ⊗ Q̃) to Q̃. Now, since A1 , ..., Ak in the expression of
φ(A) are linearly independent, we can choose trace class operators ρ1 , ..., ρn such that
ψρi (Ai ) = 1 and ψρi (Aj ) = 0 for j 6= i. Then, by applying ψ̃ρi on φ(A) we conclude that
Bi belongs to Q̃. But by definition of Q as the Woronowicz subalgebra of Q̃ generated
by < ξ ⊗ 1, φ(x)(η ⊗ 1) >Q̃ , with η, ξ belonging to H, we must have Bi belongs to Q. 2

Proposition 5.3.8. φ keeps the span of 1, A, B, B ∗ invariant.

Proof : We prove the result for φ(A). The proof for the others are exactly similar.
Using Proposition 5.3.7, we can write φ(A) as a finite sum of the form k≥0 Ak ⊗
P
0 0
Qk + m0 ,n0 ,n0 6=0 Am B n ⊗ Rm0 ,n0 + r,s,s6=0 Ar B ∗s ⊗ Rr,s
0 .
P P

Let ξ = vml .
0 ,N0
153 Quantum Isometry Groups of the Podles sphere

l
We have U (ξ) belongs to Span{vm,N , m = −l, ......l, N = ± 12 }. Let us write
X
e (ξ ⊗ 1) = l l
U vm,N ⊗ q(m,N ),(m0 ,N0 ) ,
m=−l,....l,N =± 21

l
where q(m,N ),(m0 ,N0 ) belong to Q. Since αU preserves the R-twisted volume, we have :

X
l l∗
q(m, N ), (m0 , N 0 ) q(m,N ), (m0 , N 0 ) = 1. (5.3.1)
m0 ,N 0

l 0 0 0 0
It also follows that U (Aξ) belong to Span{vm, N , m = −l , ........l , l = l − 1, l, l + 1, N =
± 21 }.
Recalling Lemma 5.3.3, we have φ(A)U e (ξ ⊗ 1) = P k l ⊗
k, m=−l,....l,N =± 21 A vm, N
l m0 n0 l l
P
Qk q(m, N ), (m0 ,N0 ) + m0 , n0 ,n0 6=0, m=−l,....l,N =± 12 A B vm,N ⊗ Rm0 , n0 q(m,N ), (m0 ,N0 )
+ r,s, s6=0, m=−l,...l, N =± 1 Ar B ∗s vm,N l 0 ql
P
⊗ Rr,s (m,N ), (m0 ,N0 ) .
2
0 0
Let m0 denote the largest integer m such that there is a nonzero coefficient of
0 0 l−m0 −n0
A B n , n0 ≥ 1 in the expression of φ(A). We claim that the coefficient of vm−n00 ,N
m in
φ(A)U e (ξ ⊗ 1) is Rm0 ,n0 q l
0 (m,N ), (m0 ,N0 ) .
l−m0 −n0
Indeed, the term vm−n00 ,N can arise in three ways: it can come from a term of the
00 00 00 00
form Am B n vm,N l or Ak vm,N
l or Ar B ∗s vm.N
l for some m , n , k, r, s.
00 00
In the first case, using Lemma 5.3.3, we must have l − m00 − n0 = l − m − n + t, 0 ≤
00 00 00
t ≤ 2m and m − n0 = m − n implying m = m00 + t, and since m00 is the largest integer
0 0
such that such that Am0 B n appears in φ(A), we only have the possibility t = 0, that is
0
l−m −n 0 0 0
vm−n00 ,N appears only in Am0 B n .
In the second case, we have m − n0 = m implying n0 = 0. - a contradiction. In the
last case, we have m−n0 = m+s so that −n0 = s which is only possible when n0 = s = 0
which is again a contradiction.
l−m0 −n e (Aξ ⊗ 1) is zero if m0 ≥ 1 ( as n0 6= 0 ). It now
Now, coefficient of vm−n00 ,N in U 0
follows from the above claim, using Lemma 5.3.3 and comparing coefficients in the
e (Aξ ⊗ 1) = φ(A)U e (ξ ⊗ 1), that Rm0 ,n0 q l 0
equality U 0 (m,N ), (m0 ,N0 ) = 0 for all n ≥ 1, for all
m, N when m00 ≥ 1. Now varying (m0 , N0 ), we conclude that the above holds for all
(m0 , N0 ). Using (5.3.1), we conclude that
l l∗ 0
P
Rm00 ,n0 m0 ,N 0 q(m,N ), (m0 ,N 0 ) q(m, N ),(m0 , N 0 ) = 0 for all n ≥ 1,
that is, Rm00 , n0 = 0 for all n0 ≥ 1 if m00 ≥ 1. Proceeding by induction on m00 , we
deduce Rm0 , n0 = 0 for all m0 ≥ 1, n0 ≥ 1.
Similarly, we have Qk = 0 for all k ≥ 2 and Rr, 0
s = 0 for all r ≥ 1, s ≥ 1.
Thus, φ(A) belongs to span{1, A, B, B , B , .......B n , B ∗2 , .....B ∗m }. But the coeffi-
∗ 2
l−n0
cient of vm−n 0 , N in φ(A)U (ξ ⊗1) is R0,n0 . Arguing as before, we conclude that R0, n0 = 0
e
0
for all n ≥ 2. In a similar way, we can prove R0, 0 0
n0 = 0 for all n ≥ 2. 2
Chapter 5: Quantum isometry groups of the Podles spheres 154

In view of the above, let us write:

φ(A) = 1 ⊗ T1 + A ⊗ T2 + B ⊗ T3 + B ∗ ⊗ T4 , (5.3.2)

φ(B) = 1 ⊗ S1 + A ⊗ S2 + B ⊗ S3 + B ∗ ⊗ S4 , (5.3.3)

for some Ti , Si in Q.

5.3.2 Homomorphism conditions

In this subsection, we shall use the facts that φ is a ∗-homomorphism and it preserves
the R-twisted volume to derive relations among Ti , Si in ( 5.3.2 ), ( 5.3.3 ).

Lemma 5.3.9.
1 − T2
T1 = ,
1 + µ2
−S2
S1 = .
1 + µ2
Proof : We have the expressions of A and B in terms of the SUµ (2) elements from the
−1
equations ( 5.1.17 ), ( 5.1.18 ) and ( 5.1.19 ). From these, we note that h(A) = (1 + µ2 )
and h(B) = 0. By recalling Proposition 5.2.2, we use (h ⊗ id)φ(A) = h(A).1 and
(h ⊗ id)φ(B) = h(B).1 to have the above two equations. 2

Lemma 5.3.10. T1∗ = T1 , T2∗ = T2 , T4∗ = T3 .

Proof : Follows by comparing the coefficients of 1, A and B respectively in the

equation φ(A∗ ) = φ(A). 2

We shall now assume that µ 6= 1. The case µ = 1 will be discussed separately.

Lemma 5.3.11.

2
S2∗ S2 + c(1 + µ2 )2 S3∗ S3 + c(1 + µ2 ) S4∗ S4

2
= (1 − T2 )(µ2 + T2 ) − c(1 + µ2 )2 T3 T3∗ − c(1 + µ2 ) T3∗ T3 + c(1 + µ2 )2 .1, (5.3.4)

−2S2∗ S2 +(1+µ2 )S3∗ S3 +µ2 (1+µ2 )S4∗ S4 = (µ2 +2T2 −1)T2 −µ2 (1+µ2 )T3 T3∗ −(1+µ2 )T3∗ T3 ,
(5.3.5)
∗ ∗ 4 ∗ 2 4 ∗ ∗
S2 S2 − S3 S3 − µ S4 S4 = −T2 + µ T3 T3 + T3 T3 , (5.3.6)

S2∗ S4 + S3∗ S2 = −(µ2 + T2 )T3∗ + T3∗ (1 − T2 ), (5.3.7)

S2∗ S3 + µ2 S4∗ S2 = −T2 T3 − µ2 T3 T2 , (5.3.8)

S4∗ S3 = −T32 . (5.3.9)

155 Quantum Isometry Groups of the Podles sphere

Proof : It follows by comparing the coefficients of 1, A, A2 , B ∗ , AB and B 2 in

the equation φ(B ∗ B) = φ(A) − φ(A2 ) + cI and then substituting S1 , T1 , T2∗ , T4 by
−S2 1−T2 ∗
1+µ2
, 1+µ 2 , T2 , T3 respectively by using the relations in Lemma 5.3.9 and Lemma

5.3.10. 2

Lemma 5.3.12.

2 2
−S2 (1 − T2 ) + c(1 + µ2 ) S3 T3∗ + c(1 + µ2 ) S4 T3

2
= −µ2 (1 − T2 )S2 + cµ2 (1 + µ2 ) T3 S4 + cµ2 (1 + µ2 )2 T3∗ S3 , (5.3.10)

S2 −2S2 T2 +(1+µ2 )(µ2 S3 T3∗ +S4 T3 ) = µ2 S2 −2µ2 T2 S2 +µ4 (1+µ2 )T3 S4 +µ2 (1+µ2 )T3∗ S3 ,
(5.3.11)
2 2
− S2 T3 + S3 (1 − T2 ) = −µ T3 S2 + µ (1 − T2 )S3 , (5.3.12)

− S2 T3∗ + S4 (1 − T2 ) = µ2 (1 − T2 )S4 − µ2 T3∗ S2 , (5.3.13)

S2 T3 + µ2 S3 T2 = µ2 (T2 S3 + µ2 T3 S2 ), (5.3.14)

S3 T3 = µ2 T3 S3 , (5.3.15)

S4 T3∗ = µ2 T3∗ S4 . (5.3.16)

Proof : It follows by equating the coefficients of 1, A, B, B ∗ , AB, B 2 and B ∗ 2 in the

equation φ(BA) = µ2 φ(AB) and then using Lemma 5.3.9 and Lemma 5.3.10. 2

Lemma 5.3.13.

S2 S2∗ + c(1 + µ2 )2 S3 S3∗ + c(1 + µ2 )2 S4 S4∗

= µ2 (1 − T2 )(1 + µ2 T2 ) + c(1 + µ2 )2 T3 T3∗ + c(1 + µ2 )2 T3∗ T3 + c(1 + µ2 )2 .1, (5.3.17)

−2S2 S2∗ + µ2 (1 + µ2 )S3 S3∗ + (1 + µ2 )S4 S4∗

= µ2 (1 + µ2 )T2 − 2µ4 (1 − T2 )T2 − µ6 (1 + µ2 )T3 T3∗ − µ4 (1 + µ2 )T3∗ T3 , (5.3.18)

− S2 S4∗ − S3 S2∗ = µ2 (1 + µ2 )T3 − µ4 (1 − T2 )T3 − µ4 T3 (1 − T2 ), (5.3.19)

S2 S2∗ − µ4 S3 S3∗ − S4 S4∗ = −µ4 T22 + µ8 T3 T3∗ + µ4 T3∗ T3 , (5.3.20)

S2 S4∗ + µ2 S3 S2∗ = −µ4 T2 T3 − µ6 T3 T2 , (5.3.21)

S3 S4∗ = −µ4 T32 . (5.3.22)

Proof : The Lemma is proved by equating the coefficient of 1, A, B, A2 , AB, B 2

Chapter 5: Quantum isometry groups of the Podles spheres 156

in the equation φ(BB ∗ ) = µ2 φ(A) − µ4 φ(A2 ) + c.1 and then using Lemma 5.3.9 and
Lemma 5.3.10. 2

5.3.3 Relations from the antipode

Now, we compute the antipode, say κ of Q. e
To begin with, we recall from Lemma 5.1.3 and Proposition 5.1.5 that {x−1 , x0 , x1 }
is a set of orthogonal vectors with same norm.

Lemma 5.3.14. If x0−1 , x00 , x01 is the normalized basis corresponding to {x−1 , x0 , x1 },
then from ( 5.3.2 ) and ( 5.3.3 ) we obtain
−1
φ(x0−1 ) = x0−1 ⊗ S3 + x00 ⊗ −µ−1 (1 + µ2 ) 2 S2 + x01 ⊗ −µ−1 S4 ,
1 1
φ(x00 ) = x0−1 ⊗ −µ(1 + µ2 ) 2 T3 + x00 ⊗ T2 + x01 ⊗ (1 + µ2 ) 2 T4 ,
− 21
φ(x01 ) = x0−1 ⊗ −µS4∗ + x00 ⊗ (1 + µ2 ) S2∗ + x01 ⊗ S3∗ .

Proof : As x−1 , x0 , x1 have same norm, it follows that x0i = Kxi , where K =
kxi k−1 , i = {−1, 0, 1}.
Now, using ( 5.1.7 ) and ( 5.3.3 ), we have

φ(x0−1 )
1
Kt(1 + µ2 ) 2
= φ(B)
µ
1
Kt(1 + µ2 ) 2 1 − t−1 x0 µx−1 µ(−µ−1 x1 )
= [1 ⊗ S1 + ⊗ S2 + 1 ⊗ S3 + 1 ⊗ S4
µ 1 + µ2 t(1 + µ2 ) 2 t(1 + µ2 ) 2
S2 S4
= Kx−1 ⊗ S3 + Kx0 ⊗ − 1 + Kx1 ⊗ −
µ(1 + µ2 ) 2 µ

( by Lemma 5.3.9 )

− 21
= x0−1 ⊗ S3 + x00 ⊗ −µ−1 (1 + µ2 ) S2 + x01 ⊗ −µ−1 S4 .

By similar calculations, we get the second and the third equations. 2

Hence, φ keeps the span of the orthonormal set {x0−1 , x00 , x01 } invariant. More-
over, φ is kept invariant by the Haar state h of SUµ (2). Therefore, we have a unitary
representation of the CQG Q e on span {x0 , x0 , x0 }.
−1 0 1
Using T4 = T3∗ from Lemma 5.3.10, the unitary matrix, say Z corresponding to φ
and the ordered basis {x0−1 , x00 , x01 } is given by :
157 Quantum Isometry Groups of the Podles sphere

 p 
S3 −µ 1 + µ2 T3 −µS4∗
∗
−S
√ 2 √ S2
 
Z= T2 . (5.3.23)
 µ 1+µ2 2
1+µ 
p
−1
−µ S4 1 + µ2 T3∗ S3∗

Recall that ( cf [41] ), the antipode κ on the matrix elements of a finite-dimensional

unitary representation U α ≡ (uαpq ) is given by κ(uαpq ) = (uαqp )∗ . Hence, the antipode is
given by :

S2∗
κ(T2 ) = T2 , κ(T3 ) = , κ(S2 ) = µ2 (1 + µ2 )T3∗ ,
µ2 (1 + µ2 )
S2
κ(S3 ) = S3∗ , κ(S4 ) = µ2 S4 , κ(T3∗ ) = ,
1 + µ2
κ(S2∗ ) = (1 + µ2 )T3 , κ(S3∗ ) = S3 , κ(S4∗ ) = µ−2 S4∗ .

Now we derive some more equations by applying κ on the equations obtained by

homomorphism condition.

Lemma 5.3.15.

2
µ4 (1 + µ2 ) T3∗ T3 + cµ2 (1 + µ2 )2 S3∗ S3 + cµ2 (1 + µ2 )2 S4 S4∗

= µ2 (1 − T2 )(µ2 + T2 ) − cS2 S2∗ − cS2∗ S2 + cµ2 (1 + µ2 )2 .1, (5.3.24)

3 2 2
−2µ4 (1 + µ2 ) T3∗ T3 + µ2 (1 + µ2 ) S3∗ S3 + µ4 (1 + µ2 ) S4 S4∗

= µ2 (1 + µ2 )T2 (µ2 + 2T2 − 1) − µ2 S2 S2∗ − S2∗ S2 , (5.3.25)

4 2 2
µ4 (1 + µ2 ) T3∗ T3 − µ2 (1 + µ2 ) S3∗ S3 − µ6 (1 + µ2 ) S4 S4∗

2
= −µ2 (1 + µ2 ) T22 + µ4 S2 S2∗ + S2∗ S2 , (5.3.26)

2 2
µ2 (1 + µ2 ) S4 T3 + µ2 (1 + µ2 ) T3∗ S3 = −S2 (µ2 + T2 ) + (1 − T2 )S2 , (5.3.27)

−S22
S4 S3 = − . (5.3.28)
µ2 (1 + µ2 )2
Proof : The relations follow by applying κ on ( 5.3.4 ), ( 5.3.5 ), ( 5.3.6), ( 5.3.7 )
and ( 5.3.9 ) respectively. 2
Chapter 5: Quantum isometry groups of the Podles spheres 158

Lemma 5.3.16.

− µ2 (1 − T2 )T3∗ + cS2 S3∗ + cS2∗ S4 = −µ4 T3∗ (1 − T2 ) + cµ2 S4 S2∗ + cµ2 S3∗ S2 , (5.3.29)

S3 S2 = µ2 S2 S3 , (5.3.30)

S2 S4 = µ2 S4 S2 , (5.3.31)

− S2∗ T3∗ + (1 − T2 )S3∗ = −µ2 T3∗ S2∗ + µ2 S3∗ (1 − T2 ), (5.3.32)

− S2 T3∗ + (1 − T2 )S4 = µ2 S4 (1 − T2 ) − µ2 T3∗ S2 , (5.3.33)

T3 S2 + µ2 S3 T2 = µ2 (T2 S3 + µ2 S2 T3 ). (5.3.34)

Proof : The relations follow by applying κ on ( 5.3.10 ), ( 5.3.15 ), ( 5.3.16 ), ( 5.3.12

), ( 5.3.13 ) and ( 5.3.14 ) respectively. 2

Lemma 5.3.17.

2
µ4 (1 + µ2 ) T3 T3∗ + cµ2 (1 + µ2 )2 S3 S3∗ + cµ2 (1 + µ2 )2 S4∗ S4

= µ4 (1 − T2 )(1 + µ2 T2 ) + cS2 S2∗ + cS2∗ S2 + cµ2 (1 + µ2 )2 .1, (5.3.35)

µ2 S22
S3 S4 = − , (5.3.36)
(1 + µ2 )2
2 2
− µ2 (1 + µ2 ) S4∗ T3∗ − µ2 (1 + µ2 ) T3 S3∗ = µ2 (1 + µ2 )S2∗ − µ4 S2∗ (1 − T2 ) − µ4 (1 − T2 )S2∗ ,
(5.3.37)
2 2 ∗ ∗ 2 2 2 ∗ 2 ∗ 4 ∗
(1 + µ ) S4 T3 + µ (1 + µ ) T3 S3 = −µ S2 T2 − µ T2 S2 . (5.3.38)

Proof : The relations follow by applying κ on (5.3.17), (5.3.22), ( 5.3.19 ) and (5.3.21)
respectively. 2

Remark 5.3.18. It follows from ( 5.3.28 ) and ( 5.3.36 ) that µ4 S4 S3 = S3 S4 .

5.3.4 Identification of SOµ (3) as the quantum isometry group

Motivated by ( 5.2.5 ) and ( 5.3.23 ) we are led to state and prove the following statement
:
The map SOµ (3) → Q sending M, L, G, N, C to −(1 + µ2 )−1 S2 , S3 , − µ−1 S4 , (1 +
µ2 )−1 (1 − T2 ), µT3 respectively is a ∗ homomorphism ( See Proposition 5.3.30 ).
To prove this, it is enough to show that all the relations of SOµ (3)( as in ( 5.2.4 ) )
+ 2 ), H, D) via the above map are satisfied.
when translated to relations of QISOR (O(Sµ,c
Hence, we list the relations one by one.
L∗ L = (I − N )(I − µ−2 N ) gives
159 Quantum Isometry Groups of the Podles sphere

1−T2 −2 1−T2 ) = µ2 +T2 µ2 (1+µ2 )−(1−T2 )

S3∗ S3 = (1 − 1+µ 2 )(1 − µ 1+µ2 1+µ2
( µ2 (1+µ2 )
),
which implies

2
µ2 (1 + µ2 ) S3∗ S3 = (µ2 + T2 )(µ2 (1 + µ2 ) − (1 − T2 )).

1−T2 4 1−T2
LL∗ = (1 − µ2 N )(1 − µ4 N ) gives S3 S3∗ = (1 − µ2 ( 1+µ2 ))(1 − µ ( 1+µ2 )) implying

2
(1 + µ2 ) S3 S3∗ = (1 + µ2 T2 )((1 + µ2 ) − µ4 (1 − T2 )).

S4∗ S4 S4∗ (1−T2 )2

G∗ G = GG∗ = N 2 gives − µ (− µ ) = (− Sµ4 )(− µ ) = (1+µ2 )2
implying

µ2 (1 − T2 )2
S4∗ S4 = S4 S4∗ = .
(1 + µ2 )2

M ∗ M = N − N 2 gives
S2∗ S2 1 − T2 (1 − T2 )2
= − ,
(1 + µ2 )2 1 + µ2 (1 + µ2 )2
S2∗ S2 (1+µ2 )(1−T2 )−(1−T2 )2
which means (1+µ2 )2
= (1+µ2 )2
implying

S2∗ S2 = (1 − T2 )(µ2 + T2 ).

S2 S2∗ 1−T2 4 1−T2 2

M M ∗ = µ2 N − µ4 N 2 gives (1+µ2 )2
= µ2 ( 1+µ 2 ) − µ ( 1+µ2 ) , which implies

S2 S2∗ = µ2 (1 − T2 )(1 + µ2 T2 ).

1−T2 1−T2 2
C ∗ C = N − N 2 gives µT3∗ µT3 = 1+µ2
− ( 1+µ2 ) , which implies

2
(1 − T2 )(µ2 + T2 ) = µ2 (1 + µ2 ) T3∗ T3 .

2
CC ∗ = µ2 N −µ4 N 2 gives (µT3 )(µT3 )∗ = µ2 ( 1+µ
1−T2 4 1−T2 2 ∗
2 )−µ ( 1+µ2 ) , which implies µ T3 T3 =

µ2 (1+µ2 )(1−T2 )−µ4 (1−T2 )2 2

(1+µ2 )2
. Therefore, (1+µ2 )(1−T2 )−µ2 (1 − T2 )2 = (1 + µ2 ) T3 T3∗ . Thus,

2
(1 − T2 )(1 + µ2 T2 ) = (1 + µ2 ) T3 T3∗ .

1−T2 4 1−T2
LN = µ4 N L gives S3 ( 1+µ2 ) = µ ( 1+µ2 )S3 , implying

S3 (1 − T2 ) = µ4 (1 − T2 )S3 .

GN = N G gives − Sµ4 ( 1+µ

1−T2 1−T2 S4
2 ) = ( 1+µ2 )(− µ ), which implies S4 (1 − T2 ) = (1 − T2 )S4 .
Chapter 5: Quantum isometry groups of the Podles spheres 160

Thus,
S4 T2 = T2 S4 .
S2 1−T2 2 1−T2 S2
M N = µ2 N M gives (− 1+µ 2 )( 1+µ2 ) = µ ( 1+µ2 )(− 1+µ2 ), implying

S2 (1 − T2 ) = µ2 (1 − T2 )S2 .

1−T2 2 1−T2
CN = µ2 N C gives µT3 ( 1+µ 2 ) = µ ( 1+µ2 )µT3 , which implies

T3 (1 − T2 ) = µ2 (1 − T2 )T3 .

LG = µ4 GL gives S3 (−µ−1 S4 ) = µ4 (−µ−1 S4 )S3 , that is,

S3 S4 = µ4 S4 S3 .

S2 S2
LM = µ2 M L gives S3 (− 1+µ 2
2 ) = µ (− 1+µ2 )S3 , that is,

S3 S2 = µ2 S2 S3 .

S2 S4 S4 S2
M G = µ2 GM gives (− 1+µ 2
2 )(− µ ) = µ (− µ )(− 1+µ2 ), that is,

S2 S4 = µ2 S4 S2 .

S2 S2
CM = M C gives (µT3 )(− 1+µ 2 ) = (− 1+µ2 )(µT3 ), that is,

T3 S2 = S2 T3 .

S4∗ S4∗
LG∗ = µ4 G∗ L gives S3 (− µ ) = µ4 (− µ )S3 , that is,

S3 S4∗ = µ4 S4∗ S3 .

S2 2 −1 S (− S4 ), that is,
M 2 = µ−1 LG gives (− 1+µ 2) = µ 3 µ

µ2
S3 S4 = − S22 .
(1 + µ2 )2

S∗ 1−T2
M ∗ L = µ−1 (I − N )C gives − 1+µ
2
2 S3 = µ
−1 (1 −
1+µ2
)µT3 , that is,

−S2∗ S3 = (µ2 + T2 )T3 .

161 Quantum Isometry Groups of the Podles sphere

(1−T2 )∗ 1−T2
N ∗ = N gives 1+µ2
= 1+µ2
, that is,

T2∗ = T2 .

Thus, we are led to prove the following lemmas, in some of which we will need
µ2 6= 1. The case µ = 1 will be dealt separately.

Lemma 5.3.19. S2∗ S2 = (1 − T2 )(µ2 + T2 ).

Proof : Subtracting the equation obtained by multiplying c(1 + µ2 ) with ( 5.3.5 )

from ( 5.3.4 ), we have

2
(1 + 2c(1 + µ2 ))S2∗ S2 + c(1 + µ2 ) (1 − µ2 )S4∗ S4

2 2
= (1 − T2 )(µ2 + T2 ) − c(1 + µ2 )(µ2 + 2T2 − 1)T2 + c(1 + µ2 ) (µ2 − 1)T3 T3∗ + c(1 + µ2 ) .1.
(5.3.39)
2 2
Again, by adding ( 5.3.4 ) with c(1 + µ ) times ( 5.3.6 ) gives

2 2
(1 + c(1 + µ2 ) )S2∗ S2 + c(1 − µ4 )(1 + µ2 ) S4∗ S4 .

2 2 2
= (1 − T2 )(µ2 + T2 ) − c(1 + µ2 ) T22 + c(1 + µ2 ) (µ4 − 1)T3 T3∗ + c(1 + µ2 ) .1. (5.3.40)

Subtracting the equation obtained by multiplying (µ2 + 1) with ( 5.3.39 ) from ( 5.3.40
) we obtain

2 2
−(µ2 + c(1 + µ2 ) )S2∗ S2 = (1 − T2 )(µ2 + T2 ) − c(1 + µ2 ) T22

2 2
−(1 + µ2 )(1 − T2 )(µ2 + T2 ) − cµ2 (1 + µ2 ) .1 + c(1 + µ2 ) (µ2 + 2T2 − 1)T2 .

2
The right hand side can be seen to equal −(µ2 + c(1 + µ2 ) )(1 − T2 )(µ2 + T2 ). Thus,
S2∗ S2 = (1 − T2 )(µ2 + T2 ). 2

Lemma 5.3.20.
2
µ2 (1 + µ2 ) T3∗ T3 = (1 − T2 )(µ2 + T2 ), (5.3.41)
2
(1 + µ2 ) T3 T3∗ = (1 − T2 )(1 + µ2 T2 ), (5.3.42)

S2 S2∗ = µ2 (1 − T2 )(1 + µ2 T2 ). (5.3.43)

Proof : Applying κ on Lemma 5.3.19, we obtain ( 5.3.41 ).

Unitarity of the matrix Z ( ( 2, 2 ) position of the matrix Z ∗ Z ) gives µ2 (1 +
µ2 )T3∗ T3 + T22 + (1 + µ2 )T3 T3∗ = 1.
Chapter 5: Quantum isometry groups of the Podles spheres 162

Using ( 5.3.41 ) we deduce −(1 + µ2 )2 T3 T3∗ = (T2 − 1)(1 + µ2 T2 ). Thus we obtain (

5.3.42 ).
Applying κ on ( 5.3.42 ), we deduce ( 5.3.43 ). 2
−2 2
Lemma 5.3.21. S4∗ S4 = S4 S4∗ = (1 + µ2 ) µ (1 − T2 )2 .
3
Proof : Adding ( 5.3.25 ) and ( 5.3.26 ), we have : −µ4 (1 + µ2 ) (1 − µ2 )T3∗ T3 +
2
µ4 (1 + µ2 ) (1 − µ2 )S4 S4∗ = −µ2 (1 + µ2 )(1 − µ2 )T2 (1 − T2 ) − µ2 (1 − µ2 )S2 S2∗ .
Using µ2 6= 1, we obtain,

3 2
−µ4 (1 + µ2 ) T3∗ T3 + µ4 (1 + µ2 ) S4 S4∗ = −µ2 (1 + µ2 )T2 (1 − T2 ) − µ2 S2 S2∗ .

Now using ( 5.3.41 ) and ( 5.3.43 ), we reduce the above equation to

2
µ4 (1 + µ2 ) S4 S4∗
= −µ2 (1 − T2 )(T2 + µ2 T2 + µ2 + µ4 T2 ) + µ2 (1 + µ2 )(1 − T2 )(µ2 + T2 )
= µ6 (1 − T2 )2 .

Thus,

µ6
S4 S4∗ = (1 − T2 )2
µ4 (1 + µ2 )2
µ2
= (1 − T2 )2 .
(1 + µ2 )2

µ2
Applying κ, we have S4∗ S4 = (1+µ2 )2
(1 − T2 )2 .
µ2
Thus, S4∗ S4 = S4 S4∗ = (1+µ2 )2
(1 − T2 )2 . 2

2
Lemma 5.3.22. µ2 (1 + µ2 ) S3∗ S3 = (µ2 + T2 )[µ2 (1 + µ2 ) − (1 − T2 )].

Proof : Using Lemma 5.3.19 in ( 5.3.4 ), we have

S3∗ S3 + T3∗ T3 + T3 T3∗ + S4∗ S4 = 1. (5.3.44)

The lemma is derived by substituting the expressions of T3∗ T3 , T3 T3∗ and S4∗ S4 from
(5.3.41), ( 5.3.42 ) and Lemma 5.3.21 in the equation ( 5.3.44 ). 2
2
Lemma 5.3.23. (1 + µ2 ) S3 S3∗ = (1 + µ2 T2 )(1 + µ2 − µ4 (1 − T2 )).

Proof : By unitarity of the matrix Z, in particular equating the ( 1, 1 ) th entry

of ZZ ∗ to 1 we get S3 S3∗ + µ2 (1 + µ2 )T3 T3∗ + µ2 S4∗ S4 = 1. Then the Lemma follows by
using ( 5.3.42 ) and Lemma 5.3.21 in the above equation. 2
163 Quantum Isometry Groups of the Podles sphere

Lemma 5.3.24. −S2∗ S3 = (µ2 + T2 )T3 .

Proof : By applying ∗ and then multiplying by µ2 on ( 5.3.7 ) we have µ2 S2∗ S3 +

µ2 S4∗ S2 = −µ2 T3 (µ2 + T2 ) + µ2 (1 − T2 )T3 . Subtracting this from ( 5.3.8 ) we have
(1 − µ2 )S2∗ S3 = −T2 T3 − µ2 T3 T2 + µ2 T3 (µ2 + T2 ) − µ2 (1 − T2 )T3 which implies −S2∗ S3 =
(µ2 + T2 )T3 as µ2 6= 1. 2

Lemma 5.3.25. S2 (1 − T2 ) = µ2 (1 − T2 )S2 .

Proof : Applying κ to Lemma 5.3.24 and then taking adjoint, we have

2
µ2 (1 + µ2 ) T3∗ S3 = −(µ2 + T2 )S2 . (5.3.45)

Adding ( 5.3.37 ) and ( 5.3.38 ) and then taking adjoint, we get ( by using µ2 6= 1 )

2
µ2 (1 + µ2 ) T3 S4 = µ4 (1 − T2 )S2 . (5.3.46)

Moreover, ( 5.3.27 ) gives

2 2
µ2 (1 + µ2 ) S4 T3 = −S2 (µ2 + T2 ) + (1 − T2 )S2 − µ2 (1 + µ2 ) T3∗ S3 .

Using ( 5.3.45 ), the right hand side of this equation turns out to be S2 (1 − T2 ).
Thus,

2
(1 + µ2 ) S4 T3 = µ−2 S2 (1 − T2 ). (5.3.47)

Again, application of adjoint to the equation ( 5.3.37 ) gives :

2 2
µ2 (1 + µ2 ) S3 T3∗ = −µ2 (1 + µ2 ) T3 S4 − µ2 (1 + µ2 )S2 + µ4 (1 − T2 )S2 + µ4 S2 (1 − T2 ).

Using ( 5.3.46 ), we get

2
(1 + µ2 ) S3 T3∗ = −S2 (1 + µ2 T2 ). (5.3.48)

Using ( 5.3.45 ) - ( 5.3.48 ) to the equation ( 5.3.11 ), we obtain :

S2 −2S2 T2 −(1+µ2 )−1 µ2 S2 (1+µ2 T2 )+µ−2 (1+µ2 )−1 S2 (1−T2 ) = µ2 S2 −2µ2 T2 S2 +
(1 + µ2 )−1 µ6 (1 − T2 )S2 − (1 + µ2 )−1 (µ2 + T2 )S2 .
This gives
µ2 (1 + µ2 )[(S2 − S2 T2 ) − (µ2 S2 − µ2 T2 S2 )] − µ2 (1 + µ2 )(S2 T2 − µ2 T2 S2 ) − µ4 S2 −
µ6 S2 T2 + S2 (1 − T2 ) − µ8 (S2 − T2 S2 ) + µ4 S2 + µ2 T2 S2 = 0.
Thus, µ2 (1 + µ2 )[S2 (1 − T2 ) − µ2 (1 − T2 )S2 ] + S2 (1 − T2 ) − µ2 (S2 − T2 S2 ) + µ6 [S2 (1 −
T2 ) − µ2 (1 − T2 )S2 ] − µ6 (1 − T2 )S2 + µ4 S2 (1 − T2 ) + µ2 (S2 (1 − T2 ) − µ2 (1 − T2 )S2 ) = 0.
Chapter 5: Quantum isometry groups of the Podles spheres 164

On simplifying,(µ6 + 2µ4 + 2µ2 + 1)(S2 (1 − T2 ) − µ2 (1 − T2 )S2 ) = 0, which proves

the lemma as 0 < µ < 1. 2

Lemma 5.3.26.
T3 (1 − T2 ) = µ2 (1 − T2 )T3 , (5.3.49)

S3 S4∗ = µ4 S4∗ S3 . (5.3.50)

Proof : The equation ( 5.3.49 ) follows by applying κ on Lemma 5.3.25 and then
taking ∗.
We have S4∗ S3 = −T32 from ( 5.3.9 ). On the other hand we have S3 S4∗ = −µ4 T32
from ( 5.3.22 ). Combining these two, we get ( 5.3.50 ). 2

Lemma 5.3.27. S4 T2 = T2 S4 .

Proof : Subtracting ( 5.3.33 ) from ( 5.3.13 ) we get the required result. 2

Lemma 5.3.28. T3 S2 = S2 T3 .

Proof : By applying ∗ on ( 5.3.32 ) and then subtracting it from ( 5.3.12 ) we obtain

S2 T3 − T3 S2 = 0. 2

Lemma 5.3.29. S3 (1 − T2 ) = µ4 (1 − T2 )S3 .

Proof : By adding ( 5.3.12 ) with ( 5.3.14 ) we obtain

S3 (1 − T2 ) + µ2 S3 (T2 − 1) = µ2 (µ2 − 1)T3 S2 .

Thus, using µ2 6= 1,
S3 (1 − T2 ) = −µ2 T3 S2 . (5.3.51)

Moreover, by applying ∗ on ( 5.3.32 ), we obtain

µ2 (1 − T2 )S3 = µ2 S2 T3 − T3 S2 + S3 (1 − T2 ).

Thus,
µ4 (1 − T2 )S3 = µ4 S2 T3 − µ2 T3 S2 + µ2 S3 (1 − T2 ).

Hence, to prove the Lemma it suffices to prove:

S3 (1 − T2 ) = µ4 S2 T3 − µ2 T3 S2 + µ2 S3 (1 − T2 ).

After using T3 S2 = S2 T3 obtained from Lemma 5.3.28 we get this to be the same as
(1 − µ2 )S3 (1 − T2 ) = µ2 (µ2 − 1)T3 S2 . This is equivalent to S3 (1 − T2 ) = −µ2 T3 S2 ( as
µ2 6= 1 ) which follows from ( 5.3.51 ). 2
165 Quantum Isometry Groups of the Podles sphere

Proposition 5.3.30. Assume µ 6= 1. The map SOµ (3) → Q sending M, L, G, N, C to

−(1+µ2 )−1 S2 , S3 , −µ−1 S4 , (1+µ2 )−1 (1−T2 ), µT3 respectively is a ∗ homomorphism.

Proof : Now, we note that the proof of this Lemma reduces to verification of the
relations on Q as derived in Lemmas 5.3.19 - 5.3.29 along with the following equations
:

S3 S4 = µ4 S4 S3 , (5.3.52)

S3 S2 = µ2 S2 S3 , (5.3.53)

S2 S4 = µ2 S4 S2 , (5.3.54)
µ2
S3 S4 = − S22 , (5.3.55)
(1 + µ2 )2
which follow from Remark 5.3.18, ( 5.3.30 ), ( 5.3.31 ), ( 5.3.36 ) respectively. 2

+
Theorem 5.3.31. For µ 6= 1, QISOR 2 ), H, D) ∼ SO (3).
(O(Sµ,c = µ

Proof : We have seen in Theorem 5.2.3 that SUµ (2) is an object in Q0R (D) and
SOµ (3) is the corresponding maximal Woronowicz subalgebra for which the action is
+
faithful. Thus, SOµ (3) is a quantum subgroup of QISOR (D). Now, Proposition 5.3.30
+
implies that QISOR (D) is a quantum subgroup of SOµ (3), thereby completing the
proof. 2

Remark 5.3.32. We observe that in the proof of Theorem 5.3.31, the only place where
the structure of D was used was in Proposition 5.3.8 and there we used the fact that the
2 ), H, |D|),
unitary commutes with |D| . Thus, if we replace this spectral triple by (O(Sµ,c
everything remains same and we deduce that

+
QISOR 2
(O(Sµ,c ), H, |D|) ∼ +
= QISOR 2
(O(Sµ,c ), H, D) ∼
= SOµ (3).

5.3.5 The quantum isometry group in the case µ = 1

As we had mentioned earlier, some of the Lemmas which were required for the proof of
Theorem 5.3.31 needed the condition µ 6= 1. In this subsection, we will give the proof
for µ = 1 case.
To begin with, we prove some of the Lemmas in the case µ = 1 which needed µ 6= 1
previously. We note that the proofs of Lemmas 5.3.19, 5.3.20, 5.3.25, 5.3.26, 5.3.27,
5.3.28 go through even in the case µ = 1. Therefore, we can use these Lemmas here.

Lemma 5.3.33. S3 (1 − T2 ) = (1 − T2 )S3 .

Chapter 5: Quantum isometry groups of the Podles spheres 166

Proof : From ( 5.3.34 ) , we obtain T3 S2 + S3 T2 = T2 S3 + S2 T3 .

Using T3 S2 = S2 T3 , from Lemma 5.3.28, we have S3 T2 = T2 S3 which proves the
Lemma.
2

Lemma 5.3.34. S2 (1 − T2 ) = (1 − T2 )S2 .

Proof : From ( 5.3.15 ) and ( 5.3.16 ), we have respectively S3 T3 = T3 S3 and

S4 T3∗ = T3∗ S4 . From ( 5.3.41 ) and ( 5.3.42 ), we see that in the case µ = 1, T3 is
normal. Hence, S3 T3∗ = T3∗ S3 and S4 T3 = T3 S4 . Using these in ( 5.3.10 ), we have
S2 (1 − T2 ) = (1 − T2 )S2 . 2

Lemma 5.3.35. S3 S3∗ = S3∗ S3 .

Proof : Multiplying by 4 the equation ( 5.3.20 ), we have

4S2 S2∗ − 4S3 S3∗ − 4S4 S4∗ = −4T22 + 4T3 T3∗ + 4T3∗ T3 .

Again, from ( 5.3.25 ), we have

16T3∗ T3 − 4S3∗ S3 − 4S4 S4∗ = −4T22 + S2 S2∗ + S2∗ S2 .

Subtracting this from the previous equation, we have

4S2 S2∗ − 4(S3 S3∗ − S3∗ S3 ) − 16T3∗ T3 = 4T3 T3∗ + 4T3∗ T3 − S2 S2∗ − S2∗ S2 .

Again using Lemma 5.3.19, ( 5.3.41 ), ( 5.3.42 ) and ( 5.3.43 ) in this equation, we
obtain
−4(S3 S3∗ − S3∗ S3 ) = 4(1 − T22 ) − 6(1 − T22 ) + 2(1 − T22 ) = 0,

which implies S3∗ S3 = S3 S3∗ . 2

Lemma 5.3.36. S4∗ S4 = S4 S4∗ .

Proof : From ( 5.3.5 ), we have

−2S2∗ S2 + 2S3∗ S3 + 2S4∗ S4 = 2T22 − 2T3 T3∗ − 2T3∗ T3 .

Again, from ( 5.3.18 ), we obtain

−2S2 S2∗ + 2S3 S3∗ + 2S4 S4∗ = 2T2 − 2(1 − T2 )T2 − 2T3 T3∗ − 2T3∗ T3 .

Subtracting this from the previous equation and using Lemma 5.3.19, ( 5.3.43 ) and
Lemma 5.3.35 we get 2(S4∗ S4 − S4 S4∗ ) = 0 implying S4∗ S4 = S4 S4∗ . 2
167 Quantum Isometry Groups of the Podles sphere

+ 2 ), H, D) is commutative as a C ∗ algebra. As
Now we prove that QISOR (O(S1,c
+ 2 ), H, D) is generated by T , T , S , S , S , it is enough to show that
QISOR (O(S1,c 2 3 2 3 4
+ 2
these elements belong to the centre of QISOR (O(Sµ,c ), H, D).
+ 2 ), H, D).
Lemma 5.3.37. T2 , T3 , S2 , S3 , S4 , belong to the centre of QISOR (O(Sµ,c
Proof : T2 is self adjoint. From ( 5.3.41 ) and ( 5.3.42 ) we note that T3 is normal in
the case µ = 1. From Lemma 5.3.19 and ( 5.3.43 ), we deduce that S2 is normal in the
case µ = 1. Similarly, from Lemma 5.3.35 and Lemma 5.3.36 we obtain that S3 and S4
are normal. Hence, it is enough to show that the elements T2 , T3 , S2 , S3 , S4 commute
among themselves.
Now, from ( 5.3.49 ), Lemma 5.3.34, Lemma 5.3.33, Lemma 5.3.27 we get that T2
commutes with T3 , S2 , S3 , S4 respectively. From Lemma 5.3.28 and ( 5.3.15 ) we have
that T3 commutes with S2 , S3 respectively. Now taking adjoint on the equation ( 5.3.16
) we obtain that T3 commutes with S4∗ implying that T3 commutes with S4 . From (
5.3.30 ) and ( 5.3.31 ), we have that S2 commutes with S3 and S4 respectively. Finally,
S3 S4 = S4 S3 follows from Remark 5.3.18. 2

From Lemma 5.3.37, we deduce the following.

+ 2 ), H, D) is commutative as a C ∗ algebra and hence
Theorem 5.3.38. QISOR (O(S1,c
coincides with the classical compact quantum group of orientation preserving isometries
of the sphere, that is, C(SO(3)).
Remark 5.3.39. Our characterization of SOµ (3) as the quantum isometry group of a
noncommutative Riemannian manifold generalizes the classical description of the group
SO(3) as the group of orientation preserving isometries of the usual Riemannian struc-
ture on the 2-sphere. It may be mentioned here that in a very recent article ( [54]), P.
M. Soltan has characterized SOµ (3) as the universal compact quantum group acting on
the finite dimensional C ∗ -algebra M2 (C) such that the action preserves a functional ωµ
defined in [54]. In the classical case, we have three equivalent descriptions of SO(3):
(a) as a quotient of SU (2), (b) as the group of (orientation preserving) isometries of
S 2 , and (c) as the automorphism group of M2 . In the quantum case the definition of
SOµ (3) is an analogue of (a), so the characterization of SOµ (3) obtained in this paper
as the quantum isometry group, together with Soltan’s characterization, completes the
generalization of all three classical descriptions of SO(3).

5.3.6 ^+ (D)
Existence of QISO
For the above spectral triple we have been unable to settle the issue of the existence
^+ (D) which is the universal object ( if it exists ) in the category Q0 (D) men-
of QISO
tioned in subsection 3.2.2. Nevertheless, we now show that if it exists, the Woronowicz
Chapter 5: Quantum isometry groups of the Podles spheres 168

subalgebra QISO+ (D) must be SOµ (3). In particular, the universal object in the sub-
category of CQG s acting by orientation preserving isometries and containing SOµ (3)
as a quantum subgroup exists.

Lemma 5.3.40. If QISO^+ (D) exists, its induced action on S 2 , say α0 , must preserve
µ,c
2
the state h on the subspace spanned by {1, A, B, B , AB, AB , A2 , B 2 , B ∗ }.
∗ ∗

Proof : Let W0 = C.1, W 1 = Span{1, A, B, B ∗ },

2
W 3 = Span{1, A, B, B ∗ , AB, AB ∗ , A2 , B 2 , B ∗2 }.
2
We note that the proof of Proposition 5.3.8 and the Lemmas preceding it do not
use the assumption that the action is volume preserving, so the proof of Proposition
5.3.8 goes through verbatim implying that α0 keeps invariant the subspace spanned
by {1, A, B, B ∗ } and hence it preserves W 3 as well. Let W 3 = W 1 ⊕ W 0 be the
2 2 2
orthogonal decomposition w.r.t. the Haar state (say h0 ) of QISO+ (D). Since SOµ (3)
is a sub-object of QISO+ (D), there is a CQG morphism π from QISO+ (D) onto SOµ (3)
satisfying (id⊗π)α0 = ∆, where ∆ is the SOµ (3) action on Sµ,c2 . It follows from this that

any QISO+ (D)-invariant subspace (in particular W 0 ) is also SOµ (3)-invariant. On the
other hand, it is easily seen that on W 3 , the SOµ (3)-action decomposes as W 1 ⊕ W 00 ,
2 2
(orthogonality w.r.t. h, the Haar state of SOµ (3))), where W 00 is a five dimensional
irreducible subspace.
We claim that W 0 = W 00 , which will prove that the QISO+ (D)-action α0 has the
same h-orthogonal decomposition as the SOµ (3)-action on W 3 , so preserves C.1 and its
2
3
h-orthogonal complements. This will prove that α0 preserves the Haar state h on W 2 .
We now prove the claim. We observe that V := W 0 W 00 is invariant under the
T

SOµ (3)-action but due to the irreducibility of ∆ on the vector space W 0 or W 00 , it has
to be zero or W 0 = W 00 . Now, dim(V) = 0 implies dim(W 0 ) + dim(W 00 ) = 5 + 5 > 9 =
dim(W 3 ) which is a contradiction unless W 0 = W 00 .
2
2

Theorem 5.3.41. If QISO ^+ (D) exists, then we have QISO+ (D) ∼ = SOµ (3). In par-
0
ticular, the universal object in the subcategory of Q (D) with objects containing SOµ (3)
as a sub- object, exists.

Proof : In Lemma 5.3.40, it was noted that Proposition 5.3.8 follows as before. We
observe that the other Lemmas used to prove Theorem 5.3.31 require the conclusion of
Lemma 5.3.40 as the only extra ingredient. 2
169 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

2
5.4 The spectral triple by Chakraborty and Pal on Sµ,c ,c >
0
Now, we shall consider another class of spectral triples on the Podles spheres and show
that they give rise to completely different quantum groups of (orientation preserving)
isometries. Indeed, for these spectral triples, we have been able to prove the existence
^+ and identify it with the CQG C ∗ (Z ∗ Z∞ ) where Z∞ denotes countably
of QISO 2
infinite copies of the group of integers.
In this section, we will work with c > 0.

5.4.1 The spectral triple

2 introduced and studied in [14].
Let us describe the spectral triple on Sµ,c
Let H+ = H− = l2 (IN {0}), H = H+ ⊕ H− .
S

Let {en , n ≥ 0} be the canonical orthonormal basis of H+ = H− and N be the

operator defined on it by N (en ) = nen .
We recall the irreducible representations π+ and π− : H± → H± as in [14].

π± (A)en = λ± µ2n en , (5.4.1)

1
π± (B)en = c± (n) 2 en−1 , (5.4.2)

where
1
1 1 2 2
e−1 = 0, λ± = ± (c + ) , c± (n) = λ± µ2n − (λ± µ2n ) + c. (5.4.3)
2 4
!
0 N
Let π = π+ ⊕ π− and D = .
N 0
2 , π, H, D) is a spectral triple.
Then (Sµ,c
! that the eigenvalues of D are {n : n ∈ Z} and eigenspace
We note ! is spanned
en en
by corresponding to the positive eigenvalue n and for the negative
en −en
eigenvalue −n.

Lemma 5.4.1.
1
π+ (B ∗ )(en ) = c+ (n + 1) 2 en+1 ,
1
π− (B ∗ )(en ) = c− (n + 1) 2 en+1 .
* ! !+
P en en0 P 1
Proof : π(B)( n cn ), = n cn c+ (n) 2 hen−1 , en0 i = cn0 +1
0 0
1 1
0 0
P P
c+ (n + 1) 2 = n cn c+ (n + 1) 2 hen , en0 +1 i = n cn
Chapter 5: Quantum isometry groups of the Podles spheres 170

1
* ! !+
c+ (n0 + 1) 2 en0 +1
 
1 en
en , c+ (n0 + 1) en0 +1
P
2 = n cn , .
0 0
1
Hence, π+ (B ∗ )(en ) = c+ (n + 1) 2 en+1 .
1
Similarly, π− (B ∗ )(en ) = c− (n + 1) 2 en+1 . 2
!
en
Lemma 5.4.2. If Pn , Qn denote the projections onto the subspace generated by
0
!
0 2 ).
and respectively, then Pn , Qn belong to π(Sµ,c
en

Proof : We claim that for all n 6= 0, c+ (n) and c− (n) are distinct.
2 2
Let c+ (n) = c− (n). Therefore, λ+ µ2n − (λ+ µ2n ) + c = λ− µ2n − (λ− µ2n ) + c. This
implies (λ+ + λ− )µ2n = 1. Thus, µ2n = ! 1 and so n has to be 0. ! !
e n 1 en−1 en
Now, for all n ≥ 1, π(B ∗ B) = c+ (n) 2 π(B ∗ ) = c+ (n) .
0 0 0
! !
0 0
Similarly, for all n ≥ 1, π(B ∗ B) = c− (n) .
en en
Hence, for all n ≥ 1, c+ (n) and c!− (n) is a discrete ∗
! distinct set of eigenvalues of B B
en 0
with eigenspace spanned by and respectively. Hence, the eigenpro-
0 en
jections corresponding to these eigenvalues belong to the C ∗ (B ∗ B) ⊆ π(Sµ,c 2 ). Hence,

Pn , Qn belong to π(Sµ,c 2 ) for all n ≥ 1.

! ! ! !
e0 e0 0 0
Moreover, π(A) = λ+ and π(A) = λ− .
0 0 e0 e0
Thus, by the same reasons as above, P0 , Q0 belong to π(Sµ,c 2 ). 2
!
2 )
00 X11 X12
Lemma 5.4.3. π(Sµ,c ={ belong to B(H ⊕ H) : X12 = X21 = 0}.
X21 X22

Proof : It suffices to prove that the commutant π(Sµ,c 2 )0 is the Von Neumann algebra
!
c1 I 0
of operators of the form { for some c1 , c2 in C. We use the fact that π+
0 c2 I
and π− are irreducible representations.
!
X11 X12 2 )0 . Using the fact that X commutes with
Let X = ∈ π(Sµ,c
X21 X22
π(A), π(B), π(B ∗ ), we have: X11 belongs to π+ (Sµ,c 2 )0 ∼ C and X
= 22 belongs to
2 0 ∼
π− (Sµ,c ) = C, so let X11 = c1 I, X22 = c2 I for some c1 , c2 .
Moreover,
X12 π− (A) = π+ (A)X12 , (5.4.4)

X12 π− (B) = π+ (B)X12 , (5.4.5)

171 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

X12 π− (B ∗ ) = π+ (B ∗ )X12 . (5.4.6)

Now, ( 5.4.5 ) implies X12 e0 belongs to Ker(π+ (B)) = Ce0 . Let X12 e0 = p0 e0 .
1 1
We have, π+ (B)(X12 e1 ) = c−2 (1)X12 e0 = p0 c−2 (1)e0 , that is, π+ (B)(X12 e1 ) belongs
to Ce0 .
Since it follows from the definition of π+ (B) that π+ (B) maps span {ei : i ≥ 2} to
(Ce0 )⊥ = span{ei : i ≥ 1}, X12 e1 must belong to span{e0 , e1 }.
Inductively, we conclude that for all n, X12 (en ) belongs to span{e0 , e1 , ......en }.
1
Using the definition of π± (B ∗ )en along with ( 5.4.6 ), we have c−2 (1)X12 e1 =
1
1 2 (1)
2 c+
p0 c+ (1)e1 , that is, X12 e1 = p0 1 e1 .
2 (1)
c−
We argue in a similar way by induction that X12 en = c0n en for some constants c0n .
Now we apply ( 5.4.6 ) and ( 5.4.5 ) on the vectors en and en+1 to get
1 1
c0n c+
2 (n+1) c0n c−
2 (n+1)
c0n+1 = 1 and c0n+1 = 1 .
2 (n+1)
c− 2 (n+1)
c+
Since c+ (n + 1) 6= c− (n + 1) for n ≥ 0, we have c0n = 0.
Hence, c0n = 0 for all n implying X12 = 0.
It follows similarly that X21 = 0. 2

5.4.2 Computation of the quantum isometry group

Let (Q, e ∆, U ) be an object in the category Q0 (D), with α = αU and the corresponding
Woronowicz C ∗ subalgebra of Q e generated by {< (ξ ⊗ 1), α(x)(η ⊗ 1) > , ξ, η ∈ H, x ∈
Q̃
2 } is denoted by Q. Assume, without loss of generality, that the representation U is
Sµ,c
faithful. ! !
en en
Since U commutes with D, it preserves the eigenvectors and .
en −en
! !
en en
Let U = ⊗ qn+ ,
en en
! !
en en
U = ⊗ qn− ,
−en −en
for some qn+ , qn− in Q.
e
! !
en en
Lemma 5.4.4. For all n ≥ 0, α(A) = ⊗ 14 {λ+ µ2n (1 + qn+ qn−∗ ) +
0 0
!
−∗ − − 0
2n + 2n +∗ 2n
λ− µ (1 − qn qn ) + λ+ µ (1 + qn qn ) − λ− µ (qn qn − 1)} + +∗ ⊗ 41 {λ+ µ2n (1 +
en
qn+ qn−∗ ) + λ− µ2n (1 − qn+ qn−∗ ) − λ+ µ2n (1 + qn− qn+∗ ) + λ− µ2n (qn− qn+∗ − 1)}.
Chapter 5: Quantum isometry groups of the Podles spheres 172

! !
0 en
For all n ≥ 0, α(A) = ⊗ 41 [λ+ µ2n (1 − qn+ qn−∗ ) + λ− µ2n (1 + qn+ qn−∗ ) +
en 0
!
− − 0
2n +∗ 2n
λ+ µ (qn qn − 1) − λ− µ (qn qn + 1)] ++∗ ⊗ 14 [λ+ µ2n (1 − qn+ qn−∗ ) + λ− µ2n (1 +
en
qn+ qn−∗ ) − λ+ µ2n (qn− qn+∗ − 1) + λ− µ2n (qn− qn+∗ + 1)].
! !
en en−1 1 1
+ 1
For all n ≥ 1, α(B) = ⊗ 41 [(c+ (n) 2 +c− (n) 2 )qn−1 qn+∗ +(c+ (n) 2 −
0 0
!
1
− +∗
1 1
+ −∗
1 1
− −∗ 0
c− (n) 2 )qn−1 qn +(c+ (n) 2 −c− (n) 2 )qn−1 qn +(c+ (n) 2 +c− (n) 2 )qn−1 qn ]+ ⊗
en−1
1 1 1 1 1 1
+ − +
1
4 [(c+ (n) + c− (n) 2 )qn−1
2 qn+∗ − (c+ (n) 2 − c− (n) 2 )qn−1 qn+∗ + (c+ (n) 2 − c− (n) 2 )qn−1 qn−∗ −
1 1
−
(c+ (n) 2 + c− (n) 2 )qn−1 qn−∗ ].
! !
0 en−1 1
+ 1 1
For all n ≥ 1, α(B) = ⊗ 14 [(c+ (n) 2 +c− (n) 2 )qn−1 qn+∗ +(c− (n) 2 −
en 0
!
1
+ −∗
1 1
− +∗
1 1
− −∗ 0
c+ (n) 2 )qn−1 qn +(c+ (n) 2 −c− (n) 2 )qn−1 qn −(c+ (n) 2 +c− (n) 2 )qn−1 qn ]+ ⊗
en−1
1 1 1 1 1 1
+ + −
1
4 [(c+ (n) + c− (n) 2 )qn−1
2 qn+∗ + (c− (n) 2 − c+ (n) 2 )qn−1 qn−∗ − (c+ (n) 2 − c− (n) 2 )qn−1 qn+∗ +
1 1
−
(c+ (n) 2 + c− (n) 2 )qn−1 qn−∗ ].
! !
en en+1 1
+ 1
For all n ≥ 0, α(B ∗ ) = ⊗ 14 [(c+ (n + 1) 2 +c− (n + 1) 2 )(qn+1 qn+∗ +
0 0
!
− −∗
1 1
+ −∗ − 0 1
+∗
qn+1 qn )+(c+ (n + 1) 2 −c− (n + 1) 2 )(qn+1 qn +qn+1 qn )]+ ⊗ 41 [(c+ (n + 1) 2 +
en+1
1 1 1
+ − + −
c− (n + 1) 2 )(qn+1 qn+∗ − qn+1 qn−∗ ) + (c+ (n + 1) 2 − c− (n + 1) 2 )(qn+1 qn−∗ − qn+1 qn+∗ )].
!
en
Proof : One has, α(A)( )
0
!
e (A ⊗ 1)U e ∗( e n
=U ⊗ 1)
0
! !
1e ∗ e n e n
= 2 U (π(A) ⊗ 1)U e ( ⊗1+ ⊗ 1)
en −en
! !
e n e n
= 12 U
e (π(A) ⊗ 1)[ ⊗ qn+∗ + ⊗ qn−∗ ]
en −en
! !
π+ (A)e n π+ (A)e n
= 12 U
e[ ⊗ qn+∗ + ⊗ qn−∗ ].
π− (A)en −π− (A)en
By using ( 5.4.1 ), ! we get this to be equal to!
2n λ+ µ2n en
e [ λ+ µ en
= 12 U ⊗ q +∗ + ⊗ qn−∗ ]
n
λ− µ2n en −λ− µ2n en
173 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

! !
en 0
= 12 U
e[ ⊗ λ+ µ2n (qn+∗ + qn−∗ ) + ⊗ λ− µ2n (qn+∗ − qn−∗ )]
0 en
! !
1e en 2n +∗ −∗ 2n +∗ −∗ en
= 4U[ ⊗{λ+ µ (qn +qn )+λ− µ (qn −qn )}+ ⊗{λ+ µ2n (qn+∗ +
en −en
qn−∗ ) − λ− µ2n (q
! n
+∗ − q −∗ )}]
n !
en e n
= ⊗ 41 qn+ {λ+ µ2n (qn+∗ +qn−∗ )+λ− µ2n (qn+∗ −qn−∗ )}+ ⊗ 41 qn− {λ+ µ2n (qn+∗ +
en −en
qn−∗ ) − λ− µ2n! (qn+∗ − qn−∗ )}
en
= ⊗ 14 {λ+ µ2n (1 + qn+ qn−∗ ) + λ− µ2n (1 − qn+ qn−∗ ) + λ+ µ2n (1 + qn− qn+∗ ) −
0
!
− 0
2n +∗
λ− µ (qn qn − 1)} + ⊗ 14 {λ+ µ2n (1 + qn+ qn−∗ ) + λ− µ2n (1 − qn+ qn−∗ ) − λ+ µ2n (1 +
en
qn− qn+∗ ) + λ− µ2n (qn− qn+∗ −! 1)}.
0
Similarly, α(A)
en
!
e (π(A) ⊗ 1)U ∗ 0
=U e ( ⊗ 1)
en
! !
e ∗ ([ e n e n
= 12 Ue (π(A) ⊗ 1)U − ] ⊗ 1)
en −en
! !
π + (A)e n π+ (A)e n
= 12 Ue[ ⊗ qn+∗ − ⊗ qn−∗ ]
π− (A)en −π− (A)en
! !
λ+ µ2n en λ + µ 2n e
n
1e
= 2U[ ⊗ qn+∗ − ⊗ qn−∗ ]
λ− µ2n en −λ− µ2n en
! !
e n en
= 14 Ue[ ⊗{λ+ µ2n (qn+∗ −qn−∗ )+λ− µ2n (qn+∗ +qn−∗ )}+ ⊗{λ+ µ2n (qn+∗ −
en −en
qn−∗ ) − λ− µ2n! (qn+∗ + qn−∗ )}]
en
= ⊗ 41 [λ+ µ2n (1 − qn+ qn−∗ ) + λ− µ2n (1 + qn+ qn−∗ ) + λ+ µ2n (qn− qn+∗ − 1) −
0
!
0
λ− µ2n (qn− qn+∗ +1)]+ ⊗ 41 [λ+ µ2n (1−qn+ qn−∗ )+λ− µ2n (1+qn+ qn−∗ )−λ+ µ2n (qn− qn+∗ −
en
1) + λ− µ2n (qn− qn+∗ + 1)].
As the proof of the others are exactly similar, we omit the proofs.

Lemma 5.4.5. We have:

qn+ qn−∗ = qn− qn+∗ for all n, (5.4.7)

1 1 1 1
− −
+
(c+ (n) 2 + c− (n) 2 )(qn−1 qn+∗ − qn−1 qn−∗ ) + (c+ (n) 2 − c− (n) 2 )(qn−1
+
qn−∗ − qn−1 qn+∗ ) = 0

for all n ≥ 1, (5.4.8)

Chapter 5: Quantum isometry groups of the Podles spheres 174

1 1 1 1
− −
+
(c+ (n) 2 + c− (n) 2 )(qn−1 qn+∗ − qn−1 qn−∗ ) + (c+ (n) 2 − c− (n) 2 )(qn−1 +
qn+∗ − qn−1 qn−∗ ) = 0

for all n ≥ 1, (5.4.9)

1 1 1 1
− −
+
(c+ (n + 1) 2 + c− (n + 1) 2 )(qn+1 qn+∗ − qn+1 qn−∗ ) = (c+ (n + 1) 2 − c− (n + 1) 2 )(qn+1 qn+∗ −

+
qn+1 qn−∗ ) for all n. (5.4.10)

Proof : Since α(A) maps π(Sµ,c 2 ) into its double commutant, we conclude by using
!
00
2 ) given in Lemma 5.4.3 that the coefficient of 0
the description of π(Sµ,c in
en
!
en
α(A) must be 0, which implies ( by Lemma 5.4.4 )
0
λ+ [1 + qn+ qn−∗ − (1 + qn− qn+∗ )] + λ− [1 − qn+ qn−∗ + qn− qn+∗ − 1] = 0.
Hence,
(λ+ − λ− )(qn+ qn−∗ − qn− qn+∗ ) = 0. Hence, (qn+ qn−∗ − qn− qn+∗ ) = 0.
Proceeding in a similar
! way, ( ! 5.4.8 ),( 5.4.9 ),!( 5.4.10 ) follow!from the facts that
0 en−1 0 en
coefficients of , and in α(B) ,
en−1 0 en+1 0
! !
0 en
α(B) and α(B )∗ ( respectively ) are zero. 2
en 0

Corollary 5.4.6. We have

∞
X 1
α(A) = APn ⊗ {λ+ (1 + qn+ qn−∗ ) + λ− (1 − qn+ qn−∗ )}
2λ+
n=0

∞
X 1
+ AQn ⊗ {λ+ (1 − qn+ qn−∗ ) + λ− (1 + qn+ qn−∗ )}.
2λ−
n=0

∞
X 1 1 1
−
α(B) = BPn ⊗ +
[(c+ (n) 2 + c− (n) 2 )(qn−1 qn+∗ + qn−1 qn−∗ )
4c+ (n)
n=1

∞
1
−1 X 1 1 1
+(c+ (n) 2 − c− (n) 2 )(qn−1 +
qn+∗ + qn−1 qn−∗ )] + BQn ⊗ [(c+ (n) 2 + c− (n) 2 ).
4c− (n)
n=1
175 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

1 1
− −
+
(qn−1 qn+∗ + qn−1 qn−∗ ) − (c+ (n) 2 − c− (n) 2 )(qn−1
+
qn−∗ + qn−1 qn+∗ )].
! ! ! !
en π+ (A) 0 en π+ (A)en
Proof : We note that π(A) = = =
0 0 π− (A) 0 0
!
en
λ+ µ2n .
0
0 1
en
π(A)@
! B C
A
en 0
Thus, = λ+ µ2n
.
0
0 1
0
π(A)@
! B C
A
0 en
Similarly, = λ− µ2n
.
en

Now, using ( 5.4.7 ),

! !
en en 1
α(A) = ⊗ {λ+ µ2n (1 + qn+ qn−∗ ) + λ− µ2n (1 − qn+ qn−∗ )}
0 0 2

!
en 1
= π(A) ⊗ {λ+ (1 + qn+ qn−∗ ) + λ− (1 − qn+ qn−∗ )}.
0 2λ+

Similarly,
! !
0 0 1
α(A) = π(A) ⊗ {λ+ (1 − qn+ qn−∗ ) + λ− (1 + qn+ qn−∗ )}.
en en 2λ−

Thus, α(A) = ∞
P 1 + −∗ + −∗
P∞
n=0 APn ⊗ 2λ+ {λ+ (1 + qn qn ) + λ− (1 − qn qn )} + + n=0 AQn ⊗
1 + −∗ + −∗
2λ− {λ+ (1 − qn qn ) + λ− (1 + qn qn )}. ! !
en 0
By similar considerations from α(B) , α(B) , the result follows. 2
0 en

fn = Pn + Qn . Then for each vector v in H, α(P

Lemma 5.4.7. Let P fn v ⊗ 1.
fn )v = P

2 ) ( Lemma 5.4.2 ).
Proof : To start with, we recall that Pn and Qn belong to π(Sµ,c
fn belongs to π(S 2 ).
Hence, P µ,c

!
en
α(P
fn )
en
!
fn ⊗ 1)U e∗ en
= U
e (P
en
Chapter 5: Quantum isometry groups of the Podles spheres 176

!
en
= U
e( ⊗ qn+∗ )
en
!
en
= ⊗1
en
!
fn ⊗ 1)( en
= (P ⊗ 1).
en

For k 6= n,
!
e∗
fk ⊗ 1)U en
U
e (P
en
!
en
= U fk ⊗ 1)(
e (P ⊗ qn+∗ )
en
= 0 !
fk ⊗ 1)( en
= (P ⊗ 1).
en

Similarly,
!
en
α(P
fn )
−en
!
en
= U
e( ⊗ qn−∗ )
−en
!
en
= ⊗1
−en
!
fn ⊗ 1)( en
= (P ⊗ 1),
−en

and for ! !
en en
k 6= n, α(P
fk ) fk ⊗ 1)(
= 0 = (P ⊗ 1).
−en −en
Combining all these, we get the required result. 2

Proposition 5.4.8. As a C ∗ -algebra, Q e is generated by the unitaries {q + }n≥0 , and

n
the self-adjoint unitary y0 = q0−∗ q0+ . Moreover, Q is generated by unitaries zn =
+
qn−1 qn+∗ , n ≥ 1 along with a self adjoint unitary w0 .
177 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

Proof : Replacing n + 1 by n in ( 5.4.10 ) we have,

1 1 1 1
−∗ −∗
+∗
(c+ (n) 2 +c− (n) 2 )(qn+ qn−1 −qn− qn−1 )−(c+ (n) 2 −c− (n) 2 )(qn− qn−1
+∗
−qn+ qn−1 ) = 0 for all n ≥ 1.
(5.4.11)
Subtracting ( 5.4.11 ) from the equation obtained by applying ∗ on ( 5.4.8 ), we have :
1 1
+∗ −∗
2(c+ (n) 2 − c− (n) 2 )(qn− qn−1 − qn+ qn−1 ) = 0 for all n ≥ 1. Now, from the proof of Lemma
1 1
5.4.2, (c+ (n) − c− (n) ) 6= 0 for all n ≥ 1. This implies :
2 2

−∗
qn− qn−1
+∗
= qn+ qn−1 for all n ≥ 1. (5.4.12)

Using ( 5.4.12 ) in ( 5.4.11 ), we have

−∗
+∗
qn+ qn−1 = qn− qn−1 for all n ≥ 1. (5.4.13)

Let yn = qn−∗ qn+ .

Then, using ( 5.4.7 ), we have qn−∗ qn+ = qn+∗ qn− . Hence, yn = yn∗ . Moreover, as yn is a
product of unitaries, it is a self adjoint unitary.
Now, from ( 5.4.12 ), we have qn− = qn+ yn−1∗ for all n ≥ 1. Therefore,

qn− = qn+ yn−1 for all n ≥ 1. (5.4.14)

−∗ +
Next, from ( 5.4.13 ), we obtain qn−∗ qn+ = qn−1 qn−1 for all n ≥ 1 implying

yn = yn−1 for all n ≥ 1. (5.4.15)

From ( 5.4.14 ) and ( 5.4.15 ) and the faithfulness of the representation U , we conclude
e is generated by {qn+ }n≥0 and y0 .
that Q
Now we prove the second part of the proposition.
Using Lemma 5.4.7, we note that for all v in H, α(AP fk )v = α(A)(P fk v ⊗1) = APk v ⊗
1 + −∗ + −∗ 1 + −∗ + −∗
2λ+ {λ+ (1 + qk qk ) + λ− (1 − qk qk )} + AQk v ⊗ 2λ− {λ+ (1 − qk qk ) + λ− (1 + qk qk )}.
Therefore, α(AP fk ) = APk ⊗ 1 {λ+ (1 + q + q −∗ ) + λ− (1 − q + q −∗ )} + AQk ⊗ 1 {λ+ (1 −
2λ+ k k k k 2λ−
qk+ qk−∗ ) + λ− (1 + qk+ qk−∗ )}.
Now, APk and AQk being distinct elements, there exist linear functional φ such that
φ(APk ) = 1, φ(AQk ) = 0 and vice versa. Hence, (φ ⊗ id)α(AP fk ) = λ+ (1 + q + q −∗ ) +
m m
+ −∗ + −∗ + −∗
λ− (1 − qm qm ) belongs to Q. Similarly, λ+ (1 − qm qm ) + λ− (1 + qm qm ) belongs to Q
for all m.
Subtracting, we get qm + q −∗ belongs to Q.
m
Using the expression of α(B) in a similar way, we have
1 1 1 1
+ − − +
(c+ (n) 2 + c− (n) 2 )(qn−1 qn+∗ + qn−1 qn−∗ ) + (c+ (n) 2 − c− (n) 2 )(qn−1 qn+∗ + qn−1 qn−∗ )
belongs to Q for all n ≥ 1.
Chapter 5: Quantum isometry groups of the Podles spheres 178

1 1 1 1
+ − + −
and (c+ (n) 2 + c− (n) 2 )(qn−1 qn+∗ + qn−1 qn−∗ ) − (c+ (n) 2 − c− (n) 2 )(qn−1 qn−∗ + qn−1 qn+∗ )
belongs to Q for all n ≥ 1.
Adding and subtracting, we have

−
+
qn−1 qn+∗ + qn−1 qn−∗ ∈ Q for all n ≥ 1, (5.4.16)

−
qn−1 +
qn+∗ + qn−1 qn−∗ ∈ Q for all n ≥ 1. (5.4.17)

Recalling ( 5.4.15 ), we have qn− = qn+ yn−1 = qn+ y0 . Using this in ( 5.4.16 ), we obtain

−
+
qn−1 qn+∗ + qn−1 qn−∗
+
= qn−1 +
qn+∗ + qn−1 y0 y0∗ qn+∗
+ +
= qn−1 qn+∗ + qn−1 qn+∗
+
= 2qn−1 qn+∗ .

Similarly, using qn− = qn+ y0 in ( 5.4.17 ), one has

−
qn−1 +
qn+∗ + qn−1 qn−∗
+ + +
= qn−1 y0 qn+∗ + qn−1 y0 qn+∗ = 2qn−1 y0 qn+∗ .

+ +
Hence, we conclude that qn−1 qn+∗ and qn−1 y0 qn+∗ are in Q for all n ≥ 1.
Let
+
zn = qn−1 qn+∗ ,
+
wn = qn−1 y0 qn+∗ ,

for all n ≥ 1.
Then, we observe that

zn∗ wn = qn+ y0 qn+∗ = qn+ qn−∗ .

Moreover,

∗
q0+ q0−∗ = q0+ (q0+ y0∗ ) = q0+ y0 q0+∗ = q0+ y0 q1+∗ q1+ q0+∗ = w1 z1∗ .

Thus for all n ≥ 0, qn+ qn−∗ belong to C ∗ ({zn , wn }n≥1 ).

− ∗
For all n ≥ 2, qn−1 qn−∗ = qn−1
+ ∗
yn−2 ∗
(qn+ yn−1 +
) = qn−1 y0∗ y0 qn+∗ = qn−1
+ +
qn+∗ = qn−1 qn+∗ = zn ,

∗
q0− q1−∗ = q0+ y0 (q1+ y0 ) = q0+ y0 y0∗ q1+∗ = q0+ q1+∗ = z1 .
179 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

Finally,

−
qn−1 +
qn+∗ = qn−1 y0∗ qn+∗ = wn

and

q0− q1+∗ = q0+ y0 q1+∗ = w1 .

Now, from the expressions of α(A) and α(B), it is clear that Q is generated by
∗ + ∗ − ∗ − ∗ + ∗
qn+ qn− , qn−1 qn+ + qn−1 qn− , qn−1 qn+ + qn−1 qn− . By the above made observations, these
belong to C ∗ ({zn , wn }n≥1 ) which implies that Q is a C ∗ subalgebra of C ∗ ({zn , wn }n≥1 ).
Moreover, from the definitions of zn , wn it is clear that C ∗ ({zn , wn }n≥1 ) is a C ∗ subal-
gebra of Q.
Therefore, Q ∼ = C ∗ ({zn , wn }n≥1 ).
+ +∗
In fact, a simpler description is possible by noting that zn wn+1 = qn−1 qn+∗ qn+ y0 qn+1 =
+ +∗ + +∗ + +∗ ∗
qn−1 y0 qn+1 = qn−1 y0 qn qn qn+1 = wn zn+1 and so, wn+1 = zn wn zn+1 which implies
{wn }n≥1 is a subset of C ∗ ({zn }n≥1 , w1 ).
Defining w0 = w1∗ z1 , we note that z1 = q0+ y0 q1+∗ q1+ y0∗ q1+∗ = w1 (z1∗ w1 ) which implies
w1∗ z1 = z1∗ w1 . Thus, w0 is self adjoint. It is a unitary as it is a product of unitaries.
Thus Q ∼ = C ∗ {{zn }n≥1 , w0 }. 2

Lemma 5.4.9. ∆(qn± ) = qn± ⊗ qn± ,

∆(y1 ) = y1 ⊗ y1 .

Proof : We use the! fact that U is a unitary

! representation.
!
en en e n
(id ⊗ ∆)U = (id ⊗ ∆)( ⊗ qn+ ) = ( ⊗ ∆(qn+ ).
en en en
! !
en en
U(12) U(13) ( )= ⊗ qn+ ⊗ qn+ .
en en
Hence, ∆(qn+ ) = qn+ ⊗ qn+ .
Similarly, ∆(qn− ) = qn− ⊗ qn− .
∗ ∗ ∗ ∗ ∗
Moreover, ∆(y1 ) = ∆(qn− qn+ ) = (qn− ⊗ qn− )(qn+ ⊗ qn+ ) = qn− qn+ ⊗ qn− qn+ = y1 ⊗ y1 .
2

Let us now consider the quantum group Se ∼ = C ∗ (Z2 ∗ Z∞ ), where Z∞ = Z ∗ Z ∗ · · ·

denotes the free product of countably infinitely many copies of Z. By the Remarks 1.1.7,
1.1.4 and 1.1.3, Se ∼
= C(Z2 ) ∗ C(T) ∗ C(T) ∗ · · · , and let us denote by rn+ the generator
of n th copy of C(T) and by y the generator of C(Z2 ).
The coproduct ∆0 on Se is given by ∆0 (rn+ ) = rn+ ⊗ rn+ , ∆0 (y) = y ⊗ y.
Chapter 5: Quantum isometry groups of the Podles spheres 180

Define
! !
en en
V = ⊗ rn+ .
en en

! !
en en
V = ⊗ rn+ y.
−en −en

Lemma 5.4.10. V commutes with D and V is a unitary representation of Se , that is

e ∆0 , V ) is an object in Q0 (D).
(S,

Proof !
: As the eigenspaces
! corresponding to distinct eigenvalues of D are spanned
en en
by and , V commutes with D.
en −en
The fact that V is a representation follows from the proof of Lemma 5.4.9.
To prove that V is a unitary, it is enough to check the following:
* ! !+ * ! !+ * ! !+
en em en em en em
V ,V = , .1, V ,V =
en em en em en −em
* ! !+ * ! !+ * ! !+
en em en em en em
, .1, V ,V = , .1,
en −em −en −em −en −em

The proofs being similar, we prove only the first equation.

* ! !+ * ! ! +
en em en e m
V ,V = ⊗ rn+ , +
⊗ rm
en em en em
* ! !+
en em
= , ⊗ rn+∗ rm
+,
en em
* ! !+
en em
which, if n 6= m equals 0 = , .1,
en em
* ! !+
en em
and if n = m, equals 1 = , .1. 2
en em

Setting rn− = rn+ y, we observe that rn− is a unitary and satisfies :

−∗
rn− rn−1
+∗
= rn+ rn−1 for all n ≥ 1. (5.4.18)

−∗
+∗
rn+ rn−1 = rn− rn−1 for all n ≥ 1. (5.4.19)
−∗ + + − +∗ −
Using rn− = rn+ rn−1 rn−1 ( from 5.4.18 ) in ( 5.4.19 ) we have rn−1 = rn−1 rn−1 rn−1 .
181 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

This implies
∗
rn+ rn− = rn− rn+∗ for all n. (5.4.20)

Moreover, taking ∗ on ( 5.4.18 ) and ( 5.4.19 ) respectively, we get the following:

−
+
rn−1 rn−∗ − rn−1 rn+∗ = 0 for all n ≥ 1. (5.4.21)

−
+
rn−1 rn+∗ − rn−1 rn−∗ = 0 for all n ≥ 1. (5.4.22)

Thus, the equations ( 5.4.7 ) - ( 5.4.10 ) in Lemma 5.4.5 are satisfied with qn± ’s replaced
by rn± ’s and hence it is easy to see that there is a C ∗ -homomorphism from Se to Q̃
sending y, rn+ to y0 and qn+ respectively, which is surjective by Proposition 5.4.8 and
is a CQG morphism by Lemma 5.4.9. In other words, (S, e ∆0 , V ) is indeed a universal
object in Q0 (D). It is clear that the maximal Woronowicz subalgebra of Se for which
+ ∗ ∗
the action is faithful, that is QISO+ (D), is generated by rn−1 rn+ , n ≥ 1 and r0+ yr1+ ,
so again isomorphic with C ∗ (Z2 ∗ Z∞ ).
This is summarized in the following:

Theorem 5.4.11. The universal object in the category Q0 (D), that is QISO
^+ (D) exists
and is isomorphic with C ∗ (Z2 ∗ Z∞ ). Moreover, the quantum group QISO+ (D) is again
isomorphic with C ∗ (Z2 ∗ Z∞ ).

Remark 5.4.12. This example shows that QISO+ in general may not be matrix quan-
tum group, that is may not have a finite dimensional fundamental unitary, even if the
underlying spectral triple is of compact type. This is somewhat against the intuition
derived from the classical situation, since for a classical compact Riemannian manifold
the group of isometries is always a compact Lie group, hence has an embedding into the
group of orthogonal matrices of some finite dimension.

We end this chapter by noting that α gives an example where the quantum group
of orientation preserving isometries does not have a C ∗ action. Before that, we recall
some useful properties of the so called Toeplitz algebra from [26].

Proposition 5.4.13. Let τ1 be the unilateral shift operator on l2 (IN ) defined by τ1 (en ) =
en−1 , n ≥ 1, τ (e0 ) = 0. Then the Toeplitz algebra C ∗ (τ1 ). is the C ∗ algebra generated
by τ1 , on l2 (IN ). It contains all compact operators and moreover, the commutator of any
two elements of C ∗ (τ1 ) is compact.

Let τ be the operator on H defined by τ = τ1 ⊗ id.

Lemma 5.4.14. B = τ |B| .

Chapter 5: Quantum isometry groups of the Podles spheres 182

Proof : We note that

! !
en 1 e n
|B| ( ) = (A − A2 + cI) 2 ( )
0 0
! !
q e n 1 en
= λ+ µ2n − λ2+ µ4n + c( ) = c+ (n) 2
0 0
! ! !
en 1 en−1 en
and hence τ |B| ( ) = c+ (n) ( 2 ) = B( ).
0 0 0
! !
0 0
Similarly, τ |B| ( ) = B( ). This completes the proof of the Lemma.
en en
2

Lemma 5.4.15.
∗
X
+
α(τ ) = τ (Pn + Qn ) ⊗ rn−1 rn+ ,
n≥1

where rn± are the elements of QISO

^+ (D) introduced before.

Proof : For all n ≥ 1, we have

! !
en e∗
e (τ ⊗ id)U en
α(τ )( )=U
0 0
! !
1e e ∗[ en en
= U (τ ⊗ id)U + ]
2 en −en
! !
1e en en
= U (τ ⊗ id)[ ⊗ rn+∗ + ⊗ rn−∗ ]
2 en −en
! !
1e en 0
= U (τ ⊗ id)[ ⊗ (rn+∗ + rn−∗ ) + ⊗ (rn+∗ − rn−∗ )]
2 0 en
! !
1e en−1 0
= U[ ⊗ (rn+∗ + rn−∗ ) + ⊗ (rn+∗ − rn−∗ )]
2 0 en−1
! !
1e en−1 en−1
= U[ ⊗ (2rn+∗ ) + ⊗ 2rn−∗ ]
4 en−1 −en−1
! ! !
1 en−1 en−1 − en−1
= [ +
⊗ rn−1 rn+∗ + ⊗ rn−1 rn−∗ ] = +
⊗ rn−1 rn+∗ .
2 en−1 −en−1 0
! !
0 0 +
Similarly, α(τ ) = ⊗ rn−1 rn+∗ for all n ≥ 1.
en en−1
183 2 ,c > 0
The spectral triple by Chakraborty and Pal on Sµ,c

! !
e0 0
Moreover, α(τ ) = α(τ ) = 0.
0 e0
+ +
rn+∗ + n≥1 τ Qn ⊗ rn−1 rn+∗ = n≥1 τ (Pn + Qn ) ⊗
P P P
Thus, α(τ ) = n≥1 τ Pn ⊗ rn−1
+
rn−1 rn+∗ 2

Theorem 5.4.16. The ∗-homomorphism α is not a C ∗ action.

Proof : We begin with the observation that each of the C ∗ algebras π± (Sµ,c 2 ) is

nothing but the Toeplitz algebra. For example, consider C := π+ (Sµ,c 2 ). Clearly, T =

π+ (B) in an invertible operator with the polar decomposition given by, T = τ1 |T |, hence
τ1 belongs to C. Thus, C contains the Toeplitz algebra C ∗ (τ1 ), which by Proposition
5.4.13 contains all compact operators. In particular, C ∗ (τ1 ) must contain π+ (A) as well
as all the eigenprojections Pn of |π+ (B)| so it must contain the whole of C. Similar
arguments will work for π− (Sµ,c2 ).

Thus, τ = τ1 ⊕ τ1 = π(B)|π(B)|−1 belongs to π(Sµ,c 2 ). If α is a C ∗ action, then for

an arbitrary state φ on QISO+ (D) we must have αφ (τ ) ≡ (id ⊗ φ) ◦ α(τ ) is in π(Sµ,c 2 ),

hence αφ (τ )P+ must belong to C = π+ (Sµ,c 2 ), where P denotes the projection onto
+
H+ . By Proposition 5.4.13, this implies that [αφ (τ )P+ , τ1 ] must be a compact operator.
We claim that for suitably chosen φ, this compactness condition is violated, which will
complete the proof of the theorem.
To this end, fix an irrational number θ and consider the sequence λn = e2πinθ of
complex number of unit modulus. We note that the linear functionals which send the
generator of C(Z2 )( which is y ) to 1 and the generator of the n-th copy of C(T) ( which
+
is rn−1 rn+∗ by Proposition 5.4.8 ) to λn are evaluation maps and hence homomorphisms.
Using Remark 1.1.6, we have a unital ∗-homomorphism φ : QISO+ (D) = C(Z2 ) ∗
∞
C(T)∗ → C which extends the above mentioned homomorphisms. Hence, αφ (τ ) =
P
n λn τ (Pn + Qn ). Moreover, we see that
!
en
[αφ (τ )P+ , τ1 ]
0
! !
en−1 en−1 +
= (id ⊗ φ)α(τ ) − τ (id ⊗ φ)[ ⊗ rn−1 rn+∗ ]
0 0
!
en−2
= (λn−1 − λn ) , n ≥ 2.
0
! !
0 0
Similarly, [αφ (τ )P+ , τ1 ] = (λn−1 −λn ) . Hence, the above commutator
en en−2
cannot be compact since λn − λn−1 does not go to 0 as n → ∞. 2
Chapter 5: Quantum isometry groups of the Podles spheres 184

Corollary 5.4.17. The subcategory of Q0 (D) consisting of objects (Q̃, U ) where αU is

a C ∗ action does not have a universal object.

Proof: By Theorem 5.4.16, the proof will be complete if we can show that if a
universal object exists for the subcategory (say Q01 ) mentioned above, then it must
be isomorphic with QISO^+ (D). For this, consider the quantum subgroups Q eN , N =
^+ (D) generated by r+ , n = 1, ..., N and y. Let πN : QISO
1, 2, ..., of QISO ^+ (D) → Q eN
n
be the CQG morphism given by πN (y) = y, πN (rn+ ) = rn+ for n ≤ N and πN (rn+ ) = 1
for n > N.
We claim that (Q eN , UN := (id ⊗ πN ) ◦ V ) is an object in Q0 ( where V denotes the
1
unitary representation of QISO^+ (D) on H ). To see this, we first note that for all N,
(id ⊗ πN )α(A) = N
P 1 + +∗ + +∗
PN
n=0 APn ⊗ 2λ+ {λ+ (1 + rn yrn ) + λ− (1 − rn yrn )} + n=0 AQn ⊗
1 + +∗ + +∗
P∞ 1
2λ− {λ+ (1 − rn yrn ) + λ− (1 + rn yrn )} + n=N +1 APn ⊗ 2λ+ {λ+ (1 + y) + λ− (1 − y)} +
P ∞ 1
n=N +1 AQn ⊗ 2λ− {λ+ (1 − y) + λ− (1 + y)}.
Among the four summands, the first two clearly belong to A ⊗ Q eN . Moreover, the
PN
sum of the third and the fourth summand equals A(1 − n=1 Pn ) ⊗ 2λ1+ {λ+ (1 + y) +
λ− (1 − y)} + A(1 − N 1
P
n=1 Qn ) ⊗ 2λ− {λ+ (1 − y) + λ− (1 + y)} which is an element of
A⊗Q eN .
We proceed similarly in the case of B, to note that it is enough to show that for all
N,
∞ 1 1 ∞ 1 1
X c+ (n) 2 + c− (n) 2 X (c+ (n) 2 − c− (n) 2 )y
BPn ⊗ + BPn ⊗
2c+ (n) 2c+ (n)
n=N +2 n=N +2

∞ 1 1 ∞ 1 1
X c+ (n) 2 + c− (n) 2 X (c+ (n) 2 − c− (n) 2 )y
+ BQn ⊗ − BQn ⊗
2c− (n) 2c− (n)
n=N +2 n=N +2

belongs to A ⊗ Q g N . The norm of the second and the fourth term can be eas-
− 12
ily seen to be finite. The first term equals 12 B(1 − N
P +1 2
n=1 Pn )[(A − A + cI) +
1
−1 2
cI) { λλ− ( λλ−
2
(A − A2 + +
A− + cI} ] ⊗ 1 and therefore belongs to A ⊗ Q
+
A) g N . The
third term can be treated similarly.
Thus, there is surjective CQG morphism ψN from the universal object, say G, e of
Q01 to Q eN )N ≥1 form an inductive system of objects in Q0 (D), with the
eN . Clearly, (Q
^+ (D), and the morphisms ψN induce a surjective CQG
inductive limit being QISO
morphism (say ψ) from Ge to QISO^+ (D). But Ge is an object in Q0 (D), so must be a
^+ (D). This
quantum subgroup of the universal object in this category, that is, QISO
gives the CQG morphism from QISO ^+ (D) onto G, e which is obviously the inverse of ψ,
and hence we get the desired isomorphism between Ge and QISO^+ (D). 2
Bibliography

[1] Banica, T.: Quantum automorphism groups of small metric spaces, Pacific J.
Math. 219(2005), no. 1, 27–51.

[2] Banica, T.: Quantum automorphism groups of homogeneous graphs, J. Funct.

Anal. 224(2005), no. 2, 243–280.

[3] Banica, T.; Bichon, J.: Quantum groups acting on 4 points. J. Reine Angew.
Math. 626 (2009), 75–114.

[4] Banica, T.; Bichon, J.; Collins, B.: Quantum permutation groups: a survey.
Noncommutative harmonic analysis with applications to probability, 13–34, Ba-
nach Center Publ., 78, Polish Acad. Sci. Inst. Math., Warsaw, 2007.

[5] Banica, T.; Moroianu, S.: On the structure of quantum permutation groups.
Proc. Amer. Math. Soc. 135 (2007), no. 1, 21–29 (electronic).

[6] Bhowmick, J. and Goswami, D.: Quantum isometry groups : examples and
computations, Comm. Math. Phys. 285 ( 2009 ), no. 2, 421-444.

[7] Bhowmick, J. : Quantum Isometry Group of the n tori, to appear in Proc. Amer.
Math. Soc., arXiv 0803.4434.

[8] Bhowmick, J., Goswami, D.: Quantum group of orientation preserving Rieman-
nian Isometries, arXiv:0806.3687.

[9] Bhowmick, J., Goswami, D., Skalski, A.: Quantum Isometry Groups of 0- Di-
mensional Manifolds, to appear in Trans. AMS, arXiv:0807.4288.

[10] Bhowmick, J. and Goswami, D.: Quantum isometry groups of the Podles
Spheres, arXiv:0810.0658.

[11] Bichon, J.: Quantum automorphism groups of finite graphs, Proc. Amer. Math.
Soc. 131(2003), no. 3, 665–673.

185
BIBLIOGRAPHY 186

[12] Chakraborty, P. S., Goswami, D. and Sinha, Kalyan B.: Probability and geom-
etry on some noncommutative manifolds, J. Operator Theory 49 (2003), no. 1,
185–201.

[13] Chakraborty, P. S. and Pal, A.: Equivariant spectral triples on the quantum
SU (2) group, K. Theory 28(2003), 107–126.

[14] Chakraborty, P. S. and Pal, A.: Spectral triples and associated Connes-de
Rham complex for the quantum SU(2) and the quantum sphere, Comm. Math.
Phys.,240 (2003), No. 3, 447-456.

[15] Chari, V; Pressley, A.: “ A guide to Quantum Groups”, Cambridge University

Press ( 1994 ).

[16] Christensen, E.; Ivan, C. : Spectral triples for AF C ∗ -algebras and metrics on
the Cantor set, J. Operator Theory 56 (2006), no. 1, 17–46.

[17] Connes, A.: “Noncommutative Geometry”, Academic Press, London-New York

(1994).

[18] Connes, A.: Cyclic cohomology, quantum group symmetries and the local index
formula for SUq (2), J. Inst. Math. Jussieu 3(2004), no. 1, 17–68.

[19] Connes, Alain; Dubois-Violette, Michel. : Noncommutative finite-dimensional

manifolds. I. Spherical manifolds and related examples. Comm. Math. Phys.
230 (2002), no. 3, 539–579.

[20] Connes, A.: On the Spectral Characterization of Manifolds, preprint,

arXiv:math/0810.2088.

[21] Connes, A.: Compact metric spaces, Fredholm modules, and hyperfiniteness,
Ergodic Theory Dynam. Systems 9 (1989), no. 2, 207–220.

[22] Connes, Alain; Marcolli, Matilde.: “Noncommutative geometry, quantum fields

and motives”, American Mathematical Society Colloquium Publications, 55
American Mathematical Society, Providence, RI; Hindustan Book Agency, New
Delhi, 2008.

[23] Dabrowski, L., Landi, G., Sitarz, A., van Suijlekom, W. and Varilly, Joseph C.:
The Dirac operator on SUq (2), Comm. Math. Phys. 259(2005), no. 3, 729–759.

[24] Dabrowski, L., D’Andrea, F., Landi, G., Wagner, E.: Dirac operators on all
Podles quantum spheres, J. Noncomm. Geom. 1 (2007) 213-239.

[25] Dabrowski, L.: Spinors and Theta Deformations, preprint, arXiv:0808.0440.

187 BIBLIOGRAPHY

[26] Davidson, Kenneth R.: “C ∗ algebras by examples”, Hindustan Book Agency.

[27] Enock, M., Schwartz, J-M.: “Kac algebras and Duality of Locally Compact
Groups”, Springer Verlag, Berlin Heidelberg New York London Paris Tokyo
Hong Kong Barcelona Budapest (1991).

[28] Friedrich, T.: ”Dirac operators in Riemannian Geometry ”, Graduate Studies in

Mathematics 25, American Math. Soc., Providence, Rhode Island ( 2000 ).

[29] Fröhlich, J.; Grandjean, O.; Recknagel, A.: Supersymmetric quantum theory
and non-commutative geometry, Comm. Math. Phys. 203 (1999), no. 1, 119–
184.

[30] Goswami, D.: Quantum Group of isometries in Classical and Non Commutative
Geometry, Comm. Math. Phys. 285 ( 2009 ), no. 1, 141-160.

[31] Goswami, D.: Twisted entire cyclic cohomology, JLO cocycles and equivariant
spectral triples, Rev. Math. Phys. 16 ( 2004 ), no.5, 583-602.

[32] Goswami, D.: Some Noncommutative Geometric aspects of SUq (2),

arXiv:0108003v4.

[33] Helgason, S.:“ Topics in Harmonic analysis on homogeneous spaces ”,

Birkhauser, Boston. Basel. Stuttgart (1981).

[34] Helgason, S.: “ Differential Geometry, Lie Groups, and Symmetric Spaces ”,
Academic Press, New York San Francisco London ( 1978 ).

[35] Kassel, Christian.: “ Quantum groups”, Graduate Texts in Mathematics, 155

Springer-Verlag, New York, ( 1995 ).

[36] Kato, T. : “Perturbation Theory for Linear Operators”, Springer Verlag, Berlin
Heidelberg New York (1980).

[37] Klimyk, A. and Schmudgen, K. : “Quantum Groups and their Representations”,

Springer, New York, 1998.

[38] Kustermans, J.; Vaes, S.: Locally compact quantum groups. Ann, Sci. Ecole
Norm. Sup. (4) 33 (2000), no. 6, 837–934.

[39] Lance, E.C.: “Hilbert C ∗ modules: A toolkit for Operator Algebraists ” London
Mathematical Society Lecture Note Series, Vol.210, ( Cambridge: Cambridge
University Press, 1995 ).
BIBLIOGRAPHY 188

[40] Landi, Giovanni.: “ An Introduction to Noncommutative Spaces and their Ge-

ometries ”‘, Lecture notes in Physics monographs, Springer Berlin, Heidelberg,
Volume 51, ( 1997 ).

[41] Maes, A. and Van Daele, A.: Notes on compact quantum groups, Nieuw Arch.
Wisk. (4) 16 (1998), no. 1-2, 73–112.

[42] Masuda, T.; Nakagami, Y.; Woronowicz, S.L.: A C ∗ -algebraic framework for
quantum groups. Internat. J. Math. 14 (2003), no. 9, 903–1001.

[43] Podles, P.: Quantum Spheres, Letters in Mathematical Physics 14 ( 1987 ) 193
- 202.

[44] Podles, P.: Symmetries of quantum spaces. Subgroups and quotient spaces of
quantum SU (2) and SO(3) groups, Comm. Math. Phys. 170 (1995), 1-20.

[45] Rennie, A, Varilly, J.C.: Reconstruction of manifolds in non commutative geom-

etry : arxiv math/0610418.

[46] Rieffel, Mark. A. : Deformation Quantization for actions of Rd , Memoirs of the

American Mathematical Society, November 1993. Volume 106, Number 506.

[47] Rieffel , Mark. A. : Compact Quantum Groups associated with toral subgroups
. Contemp . Math.145(1992), 465–491.

[48] Rosenberg, S.: “The Laplacian on a Riemannian Manifold”, Cambridge Univer-

sity Press, Cambridge (1997).

[49] Rudin, W.: “Functional Analysis”, Tata-McGraw-Hill, New Delhi (1974).

[50] Schmudgen, K. and Wagner, E. : Representation of cross product algebras of

Podles quantum spheres, math.QA/0305309.

[51] Schmudgen, K. and Wagner, E. : Dirac operators and a twisted cyclic cocycle
on the standard Podles’ quantum sphere, J. Reine Angew. Math. 574 (2004),
219–235.

[52] Sinha, K.B; Goswami, D.: Quantum Stochastic Processes and Noncommutative
Geometry, Cambridge Tracts in Mathematics, 169, Cambridge University Press.

[53] Soltan, P. M.: Quantum families of maps and quantum semigroups on finite
quantum spaces, preprint, arXiv:math/0610922.

[54] Soltan, P. M.: Quantum SO(3) Groups and Quantum Group Actions on M2 ,
arXiv: 0810.0398v1.
189 BIBLIOGRAPHY

[55] Srivastava, S. M.: “A course on Borel sets”, Graduate Texts in Mathematics,

180, Springer-Verlag, New York (1998).

[56] Takesaki, M.: “ Theory of Operator Algebras I”’, Springer Verlag, New York,
Heidelberg, Berlin.

[57] Van Daele, Alfons.: The Haar measure on a compact quantum group, Proc.
Amer. Math. Soc. 123 (1995), 3125-3128.

[58] Van Daele, Alfons.: Discrete Quantum Groups, Journal of Algebra, 180 ( 1996
), 431-444.

[59] Wang, S.: Free products of compact quantum groups, Comm. Math. Phys. 167
(1995), no. 3, 671–692.

[60] Wang, S.: Quantum symmetry groups of finite spaces, Comm. Math. Phys.
195(1998), 195–211.

[61] Wang, S.: Structure and isomorphism classification of compact quantum groups
Au (Q) and Bu (Q), J. Operator Theory 48 (2002), 573–583.

[62] Wang, S.: Deformation of Compact Quantum groups via Rieffel’s Quantization,
Comm. Math. Phys. 178( 1996 ), 747-764.

[63] Wang, S.: Ergodic actions of universal quantum groups on operator algebras.
Comm. Math. Phys. 203 (1999), no. 2, 481–498.

[64] Wang, S.: Rieffel type discrete deformation of finite quantum groups. Comm.
Math. Phys. 202 (1999), 291-307.

[65] Varilly, J.C.: “An Introduction to Noncommutative Geometry”, EMS Series of

Lectures in Mathematics (2006).

[66] Woronowicz, S. L.: Compact matrix pseudogroups, Comm. Math. Phys.

111(1987), no. 4, 613–665.

[67] Woronowicz, S. L.: “Compact quantum groups”, pp. 845–884 in Symétries quan-
tiques (Quantum symmetries) (Les Houches, 1995), edited by A. Connes et al.,
Elsevier, Amsterdam, 1998.

[68] Woronowicz, S. L.: Pseudogroups, pseudospaces and Pontryagin duality, Pro-

ceedings of the International Conference on Mathematical Physics, Lausane
(1979), Lecture Notes in Physics 116, pp. 407-412.