Timmermann_titelei 31.1.

2008 16:36 Uhr Seite 1

S E
M M
E S

S E
M
E S
M
E M S
Timmermann_titelei 31.1.2008 16:36 Uhr Seite 2

EMS Textbooks in Mathematics

EMS Textbooks in Mathematics is a book series aimed at students or professional mathematici-

ans seeking an introduction into a particular field. The individual volumes are intended to provide
not only relevant techniques, results and their applications, but afford insight into the motivations
and ideas behind the theory. Suitably designed exercises help to master the subject and prepare
the reader for the study of more advanced and specialized literature.

Jørn Justesen and Tom Høholdt, A Course In Error-Correcting Codes

Markus Stroppel, Locally Compact Groups
Peter Kunkel and Volker Mehrmann, Differential-Algebraic Equations
Dorothee D. Haroske and Hans Triebel, Distributions, Sobolev Spaces, Elliptic Equations
Timmermann_titelei 31.1.2008 16:36 Uhr Seite 3

Thomas Timmermann

An Invitation to
Quantum Groups
and Duality
From Hopf Algebras to Multiplicative
Unitaries and Beyond
S E
M M
E S

S E
M
E S
M
European Mathematical Society
Timmermann_titelei 31.1.2008 16:36 Uhr Seite 4

Author:

Thomas Timmermann
Fachbereich Mathematik und Informatik
Westfälische Wilhelms-Universität Münster
Einsteinstraße 62
48149 Münster
Germany
E-Mail: [email protected]

2000 Mathematical Subject Classification (primary; secondary): 16W30, 46-01, 46L55; 22A22,
22D25, 22D35, 46L10

Key words: Hopf algebras, duality, multiplicative unitaries, quantum groups, quantum groupoids

The Swiss National Library lists this publication in The Swiss Book, the Swiss national bibliography,
and the detailed bibliographic data are available on the Internet at http://www.helveticat.ch.

ISBN 978-3-03719-043-2

This work is subject to copyright. All rights are reserved, whether the whole or part of the material
is concerned, specifically the rights of translation, reprinting, re-use of illustrations, recitation,
broadcasting, reproduction on microfilms or in other ways, and storage in data banks. For any kind of
use permission of the copyright owner must be obtained.

© 2008 European Mathematical Society

Contact address:
European Mathematical Society Publishing House
Seminar for Applied Mathematics
ETH-Zentrum FLI C4
CH-8092 Zürich
Switzerland
Phone: +41 (0)44 632 34 36
Email: [email protected]
Homepage: www.ems-ph.org

Typeset using the author’s TEX files: I. Zimmermann, Freiburg

Printed on acid-free paper produced from chlorine-free pulp. TCF °°
Printed in Germany

987654321
 To my parents
Bettina and Werner Timmermann
Preface

The original aim of this book was to present some results of my PhD thesis on
quantum groupoids in the setting of C  -algebras. But soon I realized that it would
be more useful to provide an introduction to the mathematical background of the
thesis than to focus on my own special results.
This book is not written by an expert and does not aim at experts; rather, it is
addressed to graduate students and non-experts from other fields. It shall provide
an introduction to quantum groups in the setting of C  -algebras and von Neumann
algebras, and enable the reader to proceed to advanced topics and research articles.
Roughly, I tried to write the book that I missed when I started to learn the theory.
Much of the material presented in this book is scattered over many research
articles and was not yet covered in introductory texts. I tried to select the most
important approaches, to present the main results of several foundational articles in
a coherent manner and from one common perspective, and to explain the context
and the interrelations of the individual approaches. Apart from the last chapter,
which summarizes some of the main results of my PhD thesis, little in this book is
original. The presentation and the choice of topics is, of course, strongly influenced
by my personal view and limited by my personal knowledge. Several omissions
had to be made in order to finish this book in finite time and space.
Naturally, it is difficult to avoid misprints and minor mistakes – I hope that only
few serious errors remained. Certainly, there are places where the presentation
could be improved, where references should be added, or where other corrections
could and should be made. I am grateful for every hint, correction, or comment
that is send to the author or the publisher. An up-to-date table of corrections can be
found at the following web address:
http://www.math.uni-muenster.de/timmermt/quantum-groups.html
The suggestion to write this book came from my PhD supervisor Joachim Cuntz,
who also provided the contact to Manfred Karbe of the EMS Publishing House.
I would like to take this opportunity to thank Joachim Cuntz for this initiative, and
for his generous support during the last years.
I would like to thank the EMS Publishing House, in particular Manfred Karbe
and Irene Zimmermann, for the friendly cooperation, and Stefaan Vaes for com-
ments on the book and many helpful suggestions. Following his advice, I included
the examples of quantum groups presented in Chapter 6 and Section 8.4 – much to
the benefit of the reader and the book, I think.
This book was written at the SFB 478 “Mathematische Strukturen in der Mathe-
matik” in Münster. I would like to thank the SFB, and the Deutsche Forschungsge-
meinschaft who funds this SFB, for support, and for the ideal working environment.
viii Preface

Furthermore, I would like to thank the members of our research group “Funktional-
analysis, Operatoralgebren und nichtkommutative Geometrie”, in particularAlcides
Buss, Siegfried Echterhoff, and Walther Paravicini, for many interesting suggestions
and discussions.
Finally, I would like to thank my wife Kristina Timmermann for her kind support
during the last years, and my father Werner Timmermann, who carefully read the
manuscript and improved it by innumerable helpful suggestions, corrections, and
hints.

Münster, January 2008 Thomas Timmermann

Contents

Preface vii
Introduction xv

Part I From groups to quantum groups 1

1 Hopf algebras 3
1.1 Motivation: Pontrjagin duality . . . . . . . . . . . . . . . . . . . 3
1.2 The concept of a Hopf algebra . . . . . . . . . . . . . . . . . . . 5
1.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.2 Examples related to groups . . . . . . . . . . . . . . . . . 6
1.3 Axiomatics of Hopf algebras . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Coalgebras and bialgebras . . . . . . . . . . . . . . . . . 12
1.3.2 Convolution . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.3 Properties of the antipode . . . . . . . . . . . . . . . . . 18
1.3.4 Another characterization of Hopf algebras . . . . . . . . . 22
1.3.5 Hopf -algebras . . . . . . . . . . . . . . . . . . . . . . 26
1.4 The duality of Hopf algebras . . . . . . . . . . . . . . . . . . . . 28
1.4.1 The duality of finite-dimensional Hopf algebras . . . . . . 28
1.4.2 Dual pairings of Hopf algebras . . . . . . . . . . . . . . . 31
1.4.3 The restricted dual of a Hopf algebra . . . . . . . . . . . 35

2 Multiplier Hopf algebras and their duality 40

2.1 Definition of multiplier Hopf algebras . . . . . . . . . . . . . . . 40
2.1.1 Multipliers of algebras . . . . . . . . . . . . . . . . . . . 41
2.1.2 Multiplier bialgebras . . . . . . . . . . . . . . . . . . . . 42
2.1.3 Multiplier Hopf algebras . . . . . . . . . . . . . . . . . . 44
2.2 Integrals and their modular properties . . . . . . . . . . . . . . . 47
2.2.1 The concept of an integral . . . . . . . . . . . . . . . . . 47
2.2.2 Existence and uniqueness . . . . . . . . . . . . . . . . . 52
2.2.3 The modular element of an integral . . . . . . . . . . . . 54
2.2.4 The modular automorphism of an integral . . . . . . . . . 56
2.3 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
2.3.1 The duality of regular multiplier Hopf algebras . . . . . . 58
2.3.2 The duality of algebraic quantum groups . . . . . . . . . 63
x Contents

3 Algebraic compact quantum groups 65

3.1 Corepresentations of Hopf -algebras . . . . . . . . . . . . . . . 66
3.1.1 Definition and examples . . . . . . . . . . . . . . . . . . 66
3.1.2 Reformulation of the concept of a corepresentation . . . . 69
3.1.3 Construction of new corepresentations . . . . . . . . . . . 75
3.2 Corepresentation theory and structure theory . . . . . . . . . . . . 80
3.2.1 Decomposition into irreducible corepresentations . . . . . 80
3.2.2 Schur’s orthogonality relations . . . . . . . . . . . . . . . 83
3.2.3 Characterization of compact quantum groups . . . . . . . 86
3.2.4 Characters of corepresentations . . . . . . . . . . . . . . 87
3.2.5 Modular properties of the Haar state . . . . . . . . . . . . 89
3.3 Discrete algebraic quantum groups . . . . . . . . . . . . . . . . . 91

Part II Quantum groups and C  -/von Neumann bialgebras 95

4 First definitions and examples 97

4.1 C  -bialgebras and von Neumann bialgebras . . . . . . . . . . . . 97
4.2 Bialgebras associated to groups . . . . . . . . . . . . . . . . . . . 100
4.3 Approaches to quantum groups in the setting of
von Neumann algebras and C  -algebras . . . . . . . . . . . . . . 104

5 C  -algebraic compact quantum groups 107

5.1 Definition and examples . . . . . . . . . . . . . . . . . . . . . . 107
5.2 Corepresentations of C  -bialgebras . . . . . . . . . . . . . . . . 111
5.2.1 Unitary corepresentations of C  -algebraic compact
quantum groups . . . . . . . . . . . . . . . . . . . . . . . 111
5.2.2 Corepresentation operators of C  -bialgebras . . . . . . . 113
5.2.3 Constructions related to corepresentation operators . . . . 116
5.3 Corepresentation theory and structure theory . . . . . . . . . . . . 121
5.3.1 Decomposition into irreducible corepresentations . . . . . 122
5.3.2 Schur’s orthogonality relations . . . . . . . . . . . . . . . 125
5.3.3 Characterization of C  -algebraic compact quantum groups 127
5.4 The relation to algebraic compact quantum groups . . . . . . . . . 128
5.4.1 From C  -algebraic to algebraic CQGs . . . . . . . . . . . 128
5.4.2 From algebraic to C  -algebraic CQGs . . . . . . . . . . . 130

6 Examples of compact quantum groups 135

6.1 Compact matrix quantum groups . . . . . . . . . . . . . . . . . . 136
6.2 The compact quantum group SU .2/ . . . . . . . . . . . . . . . . 142
6.2.1 Definition and first properties . . . . . . . . . . . . . . . 142
6.2.2 Corepresentations and their weights . . . . . . . . . . . . 145
6.2.3 Corepresentations and differential calculi . . . . . . . . . 149
 Contents xi

6.2.4 Modular properties of the Haar state . . . . . . . . . . . . 152

6.3 Products of compact quantum groups . . . . . . . . . . . . . . . . 154
6.4 The free unitary and the free orthogonal quantum groups . . . . . 159

7 Multiplicative unitaries 166

7.1 The concept of a multiplicative unitary . . . . . . . . . . . . . . . 167
7.1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.1.2 Definition and examples . . . . . . . . . . . . . . . . . . 168
7.2 The legs of a multiplicative unitary . . . . . . . . . . . . . . . . . 172
7.2.1 Definition and first properties . . . . . . . . . . . . . . . 172
7.2.2 Well-behaved multiplicative unitaries . . . . . . . . . . . 175
7.2.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 179
7.2.4 The dual pairing, counit, and antipode of the legs . . . . . 184
7.3 Classes of well-behaved multiplicative unitaries . . . . . . . . . . 189
7.3.1 Regular multiplicative unitaries . . . . . . . . . . . . . . 189
7.3.2 Manageable and modular multiplicative unitaries . . . . . 197

8 Locally compact quantum groups 203

8.1 The concept of a locally compact quantum group . . . . . . . . . 203
8.1.1 Weights . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.1.2 Locally compact quantum groups in the setting of
von Neumann algebras . . . . . . . . . . . . . . . . . . . 205
8.1.3 The modular automorphism group of a weight . . . . . . . 207
8.1.4 Reduced C  -algebraic quantum groups . . . . . . . . . . 210
8.2 Additional prerequisites . . . . . . . . . . . . . . . . . . . . . . . 213
8.3 Main properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
8.3.1 The multiplicative unitary . . . . . . . . . . . . . . . . . 217
8.3.2 The antipode and modular properties . . . . . . . . . . . 218
8.3.3 The duality of locally compact quantum groups . . . . . . 222
8.3.4 Passage between the different levels . . . . . . . . . . . . 225
8.4 Examples of locally compact quantum groups . . . . . . . . . . . 227
8.4.1 C  -algebras generated by unbounded elements . . . . . . 229
8.4.2 The quantum groups E .2/ and Ey .2/ . . . . . . . . . . 233
8.4.3 The quantum az C b group . . . . . . . . . . . . . . . . 242

Part III Selected topics 249

9 Coactions on C  -algebras, reduced crossed products, and duality 251

9.1 Actions of groups and Takesaki–Takai duality . . . . . . . . . . . 252
9.2 Coactions of C  -bialgebras on C  -algebras . . . . . . . . . . . . 256
9.3 Weak Kac systems . . . . . . . . . . . . . . . . . . . . . . . . . 263
xii Contents

9.3.1 Balanced multiplicative unitaries . . . . . . . . . . . . . . 263

9.3.2 Weak Kac systems . . . . . . . . . . . . . . . . . . . . . 267
9.3.3 Examples of weak Kac systems . . . . . . . . . . . . . . 268
9.4 Reduced crossed products and dual coactions . . . . . . . . . . . 274
9.4.1 The reduced crossed product of a coaction of A.V / . . . . 274
9.4.2 The dual coaction of a coaction of A.V / . . . . . . . . . . 277
y /. . . . . . . . . .
9.4.3 The dual coaction of a coaction of A.V 279
9.4.4 Comparison with the reduced crossed product of an action 280
9.5 Kac systems and the Baaj–Skandalis duality theorem . . . . . . . 282
9.5.1 Kac systems . . . . . . . . . . . . . . . . . . . . . . . . 282
9.5.2 The Baaj–Skandalis duality theorem . . . . . . . . . . . . 286

10 Pseudo-multiplicative unitaries on Hilbert spaces 289

10.1 The relative tensor product of Hilbert modules . . . . . . . . . . . 291
10.1.1 Hilbert modules over von Neumann algebras . . . . . . . 291
10.1.2 Outline of the construction . . . . . . . . . . . . . . . . . 293
10.1.3 Bounded elements of a Hilbert module . . . . . . . . . . 296
10.1.4 Construction of the relative tensor product . . . . . . . . . 300
10.1.5 Properties of the relative tensor product . . . . . . . . . . 302
10.2 Hopf–von Neumann bimodules . . . . . . . . . . . . . . . . . . . 307
10.2.1 The fiber product of von Neumann algebras . . . . . . . . 307
10.2.2 Hopf–von Neumann bimodules . . . . . . . . . . . . . . 312
10.3 Pseudo-multiplicative unitaries on Hilbert spaces . . . . . . . . . 314
10.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 314
10.3.2 The legs of a pseudo-multiplicative unitary . . . . . . . . 316
10.3.3 The pseudo-multiplicative unitary of a groupoid . . . . . . 323

11 Pseudo-multiplicative unitaries on C  -modules 328

11.1 Pseudo-multiplicative unitaries on C  -modules . . . . . . . . . . 329
11.1.1 The flipped internal tensor product of C  -modules . . . . 329
11.1.2 Definition and examples . . . . . . . . . . . . . . . . . . 330
11.1.3 Obstructions to the construction of the legs . . . . . . . . 335
11.2 Semigroup grading techniques on right C  -bimodules . . . . . . . 337
11.2.1 Homogeneous operators and C  -families . . . . . . . . . 337
11.2.2 Homogeneous elements of right C  -bimodules . . . . . . 342
11.2.3 Examples related to groupoids . . . . . . . . . . . . . . . 347
11.3 Hopf C  -families . . . . . . . . . . . . . . . . . . . . . . . . . . 350
11.3.1 The internal tensor product of C  -families . . . . . . . . 350
11.3.2 Morphisms of C  -families . . . . . . . . . . . . . . . . . 354
11.3.3 Hopf C  -families . . . . . . . . . . . . . . . . . . . . . . 358
11.4 The legs of a decomposable pseudo-multiplicative unitary . . . . . 359
11.5 Coactions of Hopf C  -families . . . . . . . . . . . . . . . . . . . 365
 Contents xiii

12 Appendix 369
12.1 C  -algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
12.2 C  -modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
12.3 Von Neumann algebras . . . . . . . . . . . . . . . . . . . . . . . 375
12.4 Slice maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
12.5 Auxiliary results . . . . . . . . . . . . . . . . . . . . . . . . . . . 381

Bibliography 385
Symbol Index 397
Index 401
Introduction

The aim of this book is to give an introduction to the duality of quantum groups and
to quantum groups in the setting of C  -algebras and von Neumann algebras.
Roughly, a Hopf algebra or quantum group is the natural generalization of a
group within the setting of non-commutative geometry: following the general prin-
ciple of non-commutative geometry, the underlying space of the group is replaced
by an algebra, and the group operations are replaced by additional structure maps
on this algebra.
In the setting of C  -algebras and von Neumann algebras, the term “quantum
group” refers to generalizations of locally compact groups. In other fields of mathe-
matics, the term “quantum group” is usually applied to a wide range of mathematical
objects, which are studied by quite different methods. Therefore it seems appropri-
ate to give an overview before we outline the approach adopted in this book.

Hopf algebras and quantum groups in the algebraic setting

Initially, Hopf algebras were studied in a purely algebraic setting. The first examples
appeared in the following situations:
Algebraic topology. In the study of the cohomology ring H  .G/ of a compact
Lie group G, Hopf investigated the map  W H  .G/ ! H  .G/ ˝ H  .G/ induced
by the multiplication map G  G ! G, and used algebraic properties of this
map to determine the structure of H  .G/. More generally, if X is “a group up to
homotopy”, more precisely, a Hopf space, then H  .X / is a Hopf algebra, and this
algebraic structure can be used to show that X has the same cohomology like a
product of spheres [66].
Affine algebraic groups. A basic principle in algebraic geometry is to describe
an affine space X in terms of the algebra of regular functions O.X /. Now, almost
by definition, algebraic group structures on X correspond bijectively with Hopf
algebra structures on O.X /. Applications of the Hopf algebra point of view are
given, for example, in [1].
Representation theory of groups. Further natural examples of Hopf algebras are
the group algebra kG of a finite group G and the universal enveloping algebra U.g/
of a Lie algebra g.
These examples fall into two classes: the first two examples of Hopf algebras
are commutative, and the last examples satisfy a cocommutativity condition that is,
in some sense, dual to commutativity. In particular, both classes are closely related
to classical groups.
xvi Introduction

The theory of Hopf algebras received strong new impulses when new classes of
examples were constructed that were neither commutative nor cocommutative:
Deformations. One of the most influential developments in the theory of Hopf
algebras was the introduction of q-deformations of universal enveloping algebras
associated to certain Lie algebras. First examples were constructed by Faddeev and
the Leningrad school in connection with work on the quantum inverse scattering
method; later, Drinfeld and Jimbo produced a q-deformed Hopf algebra for every
semisimple complex Lie algebra [37], where the deformation is related to a certain
Poisson structure on the initial Lie algebra. These Drinfeld–Jimbo Hopf algebras
and their representation theory are very well understood, see, for example, [23],
[24], [68], [79], [80], [84], [103], [140].
Knot invariants and the Yang–Baxter equation. There exists an intriguing con-
nection between physics, low-dimensional topology, and the corepresentation the-
ory of certain Hopf algebras. The starting point is that the category of corepresenta-
tions of a Hopf algebra carries a natural tensor product, very much like the category
of representations of a group. This tensor product is symmetric only if the Hopf al-
gebra is cocommutative. But for certain Hopf algebras called braided or triangular,
there exists a braiding, which is an isomorphism cV;W W V ˝ W ! W ˝ V , natural
in the corepresentations V and W . The coherence constraints on such a braiding
can be related to planar braid diagrams and to the quantum Yang–Baxter equation
known from physics. In particular, one can construct knot invariants and solutions
of the Yang–Baxter equation out of braided Hopf algebras. Conversely, solutions
of the quantum Yang–Baxter equation give rise to bialgebras, and, in special cases,
to Hopf algebras. A very nice account of these topics can be found in [79].
Unlike the first commutative and cocommutative examples of Hopf algebras,
the new examples listed above are no longer directly related to classical groups;
therefore they are usually called quantum groups.
Algebraic quantum groups and their duality. An algebraic framework for the
study of quantum groups and their duality was developed by Van Daele [174],
[177]. In his theory, a quantum group is a non-unital Hopf algebra equipped with
an integral, which is an analogue of the Haar measure of a locally compact group,
and to every such quantum group, one can associate a dual quantum group.

Quantum groups in the setting of C  -algebras and von Neumann

algebras
A major motivation for the introduction of quantum groups in the setting of
C  -algebras and von Neumann algebras was the generalization of Pontrjagin duality
to non-abelian locally compact groups:
 Introduction xvii

Kac algebras and generalized Pontrjagin duality. For every locally compact
abelian group G, the set of characters G y is a locally compact abelian group again,
and the Pontrjagin–van Kampen theorem says that G y
y Š G (see Section 1.1). For
a non-abelian locally compact group, a generalized dual can no longer be defined
in the form of a group, and one has to look for a larger category (of “quantum
groups”) that includes both locally compact groups and their generalized duals.
This problem was solved by Vainerman and Kac [167], [168], and by Enock and
Schwartz [47]: they defined the notion of a Kac algebra, which is a von Neumann
algebra equipped with similar structure maps like a Hopf algebra, and constructed
for every Kac algebra A a dual Kac algebra A, y such that Ay
y Š A. An important rôle
in their theory is played by the analogue of the Haar measure of a locally compact
group, which is part of the structure of a Kac algebra. A C  -algebraic counterpart
of the theory was developed by Vallin and Enock [49], [170].
The concept of a Kac algebra, however, turned out to be too restrictive to include
all interesting examples of quantum groups in the setting of C  -algebras:
Compact quantum groups. Woronowicz developed a general theory of compact
quantum groups in the setting of C  -algebras [193], [202], which contains examples
that do not satisfy all axioms of a Kac algebra. This theory is very appealing:
the definition of a compact quantum group is concise, the existence of a Haar
measure on every compact quantum group can be deduced from the axioms, and
the corepresentation theory of every such quantum group is very similar to the
representation theory of a compact group.
A new perspective on quantum groups in the setting of C  -algebras and von
Neumann algebras was introduced by Baaj and Skandalis:
Multiplicative unitaries. Examples of multiplicative unitaries were used for a
long time in the theory of quantum groups, till Baaj and Skandalis put them center-
stage, formulated an abstract definition, and gave a comprehensive treatment [7].
Roughly, a multiplicative unitary simultaneously encodes a quantum group and the
dual of that quantum group; conversely, to every “reasonable” quantum group, one
can associate a multiplicative unitary.
Finally, comprehensive theories of locally compact quantum groups were de-
veloped, which cover all known examples:
Locally compact quantum groups / weighted Hopf algebras. The theories devel-
oped by Vaes and Kustermans [91], [93], and Masuda, Nakagami, and Woronowicz
[110], seem to give a definite answer to the question “What is a locally compact
quantum group in the setting of C  -algebras/von Neumann algebras?”.
xviii Introduction

Organization of the book

The aim of this book, as stated above, is to give an introduction to the duality of
quantum groups, and to quantum groups in the setting of C  -algebras and von
Neumann algebras.
One possible approach would be to start immediately with a study of locally
compact quantum groups, which form the most general framework. For someone
who is familiar with Hopf algebras and with the high-level analytic techniques used
in the theory of locally compact quantum groups, this is probably the best choice.
In this book, however, we shall adopt another approach, which may be better suited
for graduate students and researchers from other fields.

Contents of the book. Part I of this book provides an introduction to quantum

groups in a purely algebraic setting. After a review of Hopf algebras and their duality
(Chapter 1), we discuss the duality of algebraic quantum groups developed by Van
Daele (Chapter 2). This theory provides a very nice model for the generalizations
of Pontrjagin duality that will be considered in Part II and yields many fundamental
formulas. Finally, we investigate algebraic compact quantum groups (Chapter 3).
This class of quantum groups can be studied not only in an algebraic, but also in a
C  -algebraic setting, and will serve us as a bridge for the passage to the setting of
C  -algebras.
In Part II, we turn to quantum groups in the setting of C  -algebras and von
Neumann algebras. First, we discuss the problems that arise in the definition of
a Hopf C  -algebra or Hopf–von Neumann algebra, consider examples related to
locally compact groups, and list the existing approaches (Chapter 4). Then, we
present Woronowicz’s theory of C  -algebraic compact quantum groups (Chapter 5),
which is particularly accessible and close to the algebraic setting, and consider
important examples (Chapters 6). Closely related to quantum groups in the setting
of C  -algebras and von Neumann algebras are multiplicative unitaries, which are
studied subsequently (Chapter 7). Part II ends with an overview of the theory of
locally compact quantum groups and some examples (Chapter 8). We focus on
motivation, which can often be found in the theory of algebraic quantum groups,
and explain the central analytic tools, but do not give any proof.
Part III of this book is devoted to selected topics. First, we discuss coactions
of quantum groups on C  -algebras, reduced crossed products, and a generalization
of the Takesaki–Takai duality theorem for group actions (Chapter 9). The crossed
product construction and the duality theorem make essential use of multiplicative
unitaries, and are due to Baaj and Skandalis. Next, we give an introduction to
measurable quantum groupoids, or, more precisely, to pseudo-multiplicative uni-
taries on Hilbert spaces (Chapter 10). In particular, we present the relative tensor
product of Hilbert spaces, which is also known as Connes’ fusion. Finally, we
 Introduction xix

summarize some results of the author’s thesis on pseudo-multiplicative unitaries on

C  -modules and quantum groupoids in the setting of C  -algebras (Chapter 11).
Frequently used notation and important terms used in this book are listed in
separate indices, and some background is compiled in a short appendix.

Prerequisites. This book should be accessible to graduate students and researchers

from other fields of mathematics. For Part I, no special background is needed. Part II
assumes some familiarity with Hilbert spaces, C  -algebras, and von Neumann
algebras, as summarized in the appendix; at some points, we use the language of
C  -modules, which is also summarized in the appendix. Part III contains advanced
topics and is addressed to readers with some background in the field of C  -algebras
or von Neumann algebras.

Logical dependence of the chapters. The logical dependence of the individual

chapters of this book is sketched in the following diagram:
Parts I and II Part III
3 5 6 9


















1 2 4 7 8 7 ?? 11
??
??
?^^^^
10

The dotted lines indicate that a chapter provides examples or motivation for the
developments in a subsequent chapter, without that an understanding of the first
chapter is needed for an understanding of the second one.

Preliminaries and notation

Let us fix some notation and terminology.
As usual, the letters N; Z; R; C denote the sets of natural, integer, real, and
complex numbers, respectively. We put T WD fz 2 C W jzj D 1g. The letter k will
stand for an arbitrary field.
We adopt the following convention. A sesquilinear form on a complex vector
space H is a map h  j  i W H  H ! H that is conjugate-linear in the first variable
and linear in the second variable. In particular, we apply this convention to inner
products on Hilbert spaces.
The domain of definition of a map  is denoted by Dom./, and the image is
denoted by Im./. The identity map on a set X is denoted by idX or shortly by id.
We denote the set of all bounded linear operators from a Hilbert space H1 to a
Hilbert space H2 by L.H1 ; H2 /, and the subset of all compact linear operators by
xx Introduction

K.H1 ; H2 /. Furthermore, we use the ket-bra notation: for every element  of a

Hilbert space H , we define maps

ji W C ! H;  7! ; and hj W H ! C;  7! hji:

Considering C as a Hilbert space, we have ji 2 K.C; H /; hj 2 K.H; C/, and
ji D hj.
Given a subset X of a vector space V , we denote by span X  V the linear span
of X . If V is a topological vector space, we denote by span X and ŒX  the closed
linear span of X. We say that X  V is linearly dense in Y  V if span X D Y .
In Part II and III, we denote the algebraic tensor product by the symbol “ˇ” to
distinguish it from the minimal tensor product of C  -algebras.
 Part I

From groups to quantum groups

Chapter 1
Hopf algebras

This chapter gives a brief introduction to Hopf algebras. We focus on examples

related to groups, on the axiomatics, and on the duality of Hopf algebras. The
contents is standard and can be found in nearly every text on Hopf algebras or
quantum groups, for example, in [23], [29], [79], [80], [111], [140]; two classical
references are [1], [145].

1.1 Motivation: Pontrjagin duality

One of the motivations to study Hopf algebras is the question how the Pontrjagin
duality of locally compact abelian groups can be extended to non-abelian groups.
Let us briefly recall classical Pontrjagin duality. To each locally compact abelian
group G, one can associate a dual group G, y which is locally compact and abelian
again, as follows. As a set, Gy consists of all characters on G, that is, of all contin-
uous group homomorphisms G ! T ; the group operation is given by pointwise
multiplication of characters, and the topology is the topology of uniform conver-
gence on compact subsets. The following standard examples are well-known from
Fourier analysis:
y Š T;
Z 1
Z=nZ Š Z=nZ; y Š R;
R y Š Z:
T
y by evaluation at x.
Evidently, each element x 2 G determines a character on G
Denote this character by evx .
Theorem (Pontrjagin duality). For every locally compact abelian group G, the
map G ! Gy
y given by x 7! evx is an isomorphism of topological groups.
This result can be proven via two different strategies:
• The classical approach [112], [122], [190] uses the structure theory of abelian
groups. First, the duality is verified for special classes of groups like the
examples listed above or all discrete and compact abelian groups. Then,
the contravariant functor G 7! G y is shown to be compatible with direct
sums, extensions, limits, and colimits. The structure theory says that these
operations, applied to the special classes of groups considered before, yield
all locally compact abelian groups, and so the theorem follows.
• The modern approach [20], [33], [150, ChapterVII, Section 3] uses the Fourier
transformation and functional analysis in varying levels of abstraction.
4 Chapter 1. Hopf algebras

Over the last decades, much work has been spent on generalizations of Pontrjagin
duality to larger classes of groups, in particular to non-abelian locally compact
groups. For such a group, characters retain too little information – they only capture
abelian quotients of the group. A natural solution is to consider higher-dimensional
representations as well. But then, the dual of a group can no longer be equipped
with a group structure, and one has to look for a new category of “generalized
groups” that contains locally compact groups and their duals. Roughly, one can
distinguish two approaches:
“The dual of a group is its representation theory”. This approach can be com-
pared to Grothendieck’s idea to replace a space by its category of sheaves; here, one
replaces a group G by the category of all representations of G on vector spaces, the
morphisms in this category being the intertwiners of representations. Depending
on the context, additional requirements on the representations and intertwiners may
be necessary, for example, continuity. Equipped with the natural tensor product of
representations, this category becomes a symmetric monoidal or symmetric tensor
category; the tensor subcategory of all one-dimensional representations corresponds
precisely to the group of characters. One early achievement of this approach is the
Tannaka–Krein duality theorem [22], [25], [62], [151], which says that a compact
group can be reconstructed from the category of its representations. For a survey
on this approach, see [73].
“The dual of a space is its function algebra”. The idea is to encode the underlying
space of a group G by some “coordinate algebra” A of functions on G, and the
multiplication G  G ! G by a comultiplication A ! A ˝ A. This leads to the
notion of a Hopf algebra, which, in various variants, is the central topic of this book.
The dual of the group is encoded by the group algebra which can be thought of as
the coordinate algebra of the dual group. For an abelian group, this interpretation
can be made precise using the Fourier transform.
Of course, both approaches are intimately related. Roughly, the (representation
theory of the) coordinate algebra of a group encodes the underlying space of the
group, and the (representation theory of the) group algebra encodes the representa-
tion theory of the group.
For locally compact groups, the second approach has successfully been pursued
in the setting of von Neumann algebras and C  -algebras [47], [91], leading to a
generalization of Pontrjagin duality which covers all locally compact groups. This
generalization is outlined in Chapter 8.
 1.2. The concept of a Hopf algebra 5

1.2 The concept of a Hopf algebra

1.2.1 Definition
We fix the following notation. The letter k will always stand for a field. Given an
algebra A, we denote by m W A ˝ A ! A the multiplication map a ˝ b 7! ab, and,
if A is unital, by  W k ! A the unit map  7! 1A .

Definition 1.2.1. A Hopf algebra is a unital algebra A .over k/, equipped with

i) a unital homomorphism  W A ! A ˝ A, called the coproduct or comultipli-

cation, which is coassociative in the sense that the square

A
 /A˝A
 id˝ (1.1)
 
A˝A /A˝A˝A
˝ id

commutes, that is, . ˝ idA / ı  D .idA ˝ / ı ;

ii) a homomorphism W A ! k, called the counit, which makes the diagram

A˝Ao /A˝A
 
A
˝ id id˝ (1.2)
 
k˝A /Ao A˝k
Š Š

commute, that is, . ˝ idA / ı  D idA D .idA ˝ / ı ;

iii) a linear map S W A ! A, called the antipode, which makes the diagram

A˝Ao /A˝A
 
A
S˝ id ı id ˝ S (1.3)
  
A˝A /Ao A˝A
m m

commute, that is, m ı .S ˝ idA / ı  D  ı D m ı .idA ˝ S / ı .

We shall always use the letters ; , and S to denote the comultiplication,

counit, and antipode of a Hopf algebra. When several Hopf algebras A; B; : : : are
considered simultaneously, we index the structure maps by the respective Hopf
algebras and write A ; A ; SA ; B ; B ; SB ; : : : .
6 Chapter 1. Hopf algebras

Definition 1.2.2. A morphism of Hopf algebras A and B is a unital algebra homo-

morphism F W A ! B that is compatible with the structure maps in the sense that
the squares

A A SA
A /A˝A A /k A /A

F F ˝F F F F
    
B / B ˝ B; B / k; B /B
B B SB

commute, that is, B ı F D .F ˝ F / ı A ; B ı F D A , and SB ı F D F ı SA .

Remark 1.2.3. We shall see in Remark 1.3.23 and Proposition 1.3.17 that the
definition of a Hopf algebra and the definition of a morphism of Hopf algebras can
be weakened.

1.2.2 Examples related to groups

Let G be a group. We denote by k.G/ the algebra of all k-valued functions on G,
where the addition and multiplication are defined pointwise. The structure maps
of G, that is,

the multiplication the inclusion of the unit the inversion

G  G ! G; feg ,! G; G ! G;
.x; y/ 7! xy; e 7! e; x 7! x 1 ;

induce the following algebra homomorphisms:

 W k.G/ ! k.G  G/; W k.G/ ! k; S W k.G/ ! k.G/;

(1.4)
..f //.x; y/ WD f .xy/; .f / WD f .e/; .S.f //.x/ WD f .x 1 /:

Equipped with these structure maps, k.G/ almost forms a Hopf algebra – the only
defect is that the target of the map  is k.G  G/ and not k.G/ ˝ k.G/. We
can identify k.G/ ˝ k.G/ with a subspace of k.G  G/, but the image of  is
not contained in this subspace unless G is finite. However, each unital subalgebra
A  k.G/ that satisfies .A/  A ˝ A and S.A/  A is a Hopf algebra with
respect to the restrictions of the maps ; , and S . This can be verified by direct
calculations, or by comparing diagrams (1.1)–(1.3) with diagrams that express the
group axioms.

Example 1.2.4 (The function algebra of a finite group). Let G be a finite group.
Then the tensor product k.G/ ˝ k.G/ can be identified with k.G  G/, and the
 1.2. The concept of a Hopf algebra 7

algebra k.G/, equipped with the maps , , and S defined in (1.4), forms a Hopf
algebra.
Let us rewrite the structure maps of this Hopf algebra in terms of a canonical
basis. For each x 2 G, we define a function ıx 2 k.G/ by
(
1; x D y;
ıx .y/ WD ıx;y D
0; otherwise:

Then the family .ıx /x2G is a basis of k.G/, and for each x 2 G,
X
.ıx / D ıy ˝ ız ; .ıx / D ıx;e ; S.ıx / D ıx 1 :
y;z2G
yzDx

Example 1.2.5 (The algebra of representative functions of a group). Let G be a

topological group. Recall that a continuous complex-valued function f on G is
representative if and only if it satisfies the following equivalent conditions:

i) the linear span of all left- and right-translates of f , that is, of all functions of
the form f .z  x/ W y 7! f .zyx/, where x; z 2 G, has finite dimension;
ii) the linear span of all right-translates of f , that is, of all functions of the form
f .  x/ W y 7! f .yx/, where x 2 G, has finite dimension;
iii) the linear span of all left-translates of f , that is, of all functions of the form
f .z  / W y 7! f .zy/, where z 2 G, has finite dimension;
iv) there exists a continuous representation
of G on some finite-dimensional
complex vector space V and elements v 2 V ,  2 V 0 D Homk .V; k/ such
that f .x/ D .
.x/v/ for all x 2 G.

Let us prove the equivalence of these conditions. Clearly, i) implies ii) and iii). If f
satisfies condition ii), then right translation defines a continuous representation

of G on the space V WD spanff .  x/ j x 2 Gg; in detail,

is given by
.x/g WD
g.  x/ for all g 2 V and x 2 G. Since f .x/ D .
.x/f /.e/ D .
.x/f / for
all x 2 G, the function f satisfies condition iv). A similar argument shows that
iii) implies iv). Finally, assume that f has the form described in iv), and denote
by C .
/ the space of all functions of the form x 7! .
.x/w/, where w 2 V
and 2 V 0 . Then dim C .
/  .dim V /2 < 1, and for all x; z 2 G, the
function f .z  x/ belongs to C .
/ because it can be written in the form f .z  x/ D
.
.z  x/v/ D .
.  /w/, where D .
.z// and w D
.x/v. Therefore, f
satisfies condition i).
Denote the space of all continuous representative functions on G by Rep.G/.
A moment of consideration shows that this space is a unital subalgebra of C.G/.
We show that .Rep.G//  Rep.G/ ˝ Rep.G/ and S.Rep.G//  Rep.G/, which
8 Chapter 1. Hopf algebras

implies that Rep.G/, equipped with the restrictions of the maps , , and S defined
in formula (1.4), forms a Hopf algebra.
So, consider a function f 2 Rep.G/ of the form f .x/ D .
.x/v/ as in
a basis .wi /i of V and denote by . i /i the associated
condition iv) above. Choose P
dual basis of V 0 . Then w D i wi i .w/ for each w 2 V , and hence
X
..f //.x; y/ D f .xy/ D .
.x/
.y/v/ D .
.x/wi /  i .
.y/v/
i
P
for all x; y 2 G. Therefore .f / D i f1;i ˝ f2;i , where f1;i D .
.  /wi / 2
Rep.G/ and f2;i D i .
.  /v/ 2 Rep.G/ for all i, and .f / 2 Rep.G/˝Rep.G/.
Next, consider the contragredient representation
0 of
, which is the representation
on V 0 given by
0 .x/ D .
.x/1  / for all 2 V 0 and x 2 G. Denote by
00
evv 2 V the functional given by 7! .v/. Then

.S.f //.x/ D f .x 1 / D .
.x 1 /v/ D evv .
0 .x// for all x 2 G;

and hence S.f / 2 Rep.G/.

Let us add two remarks:
i) For every group G and every field k, one can define a Hopf algebra
Repk .G/  k.G/ of representative functions which are defined by condi-
tions i)–iv) as above, but without any continuity assumption on f or on the
representation
.
ii) If G is compact, then Rep.G/ is dense in C.G/ by the Peter–Weyl-Theorem
[22, Chapter III], [62, Chapter 7], so that the Hopf algebra Rep.G/ is suffi-
ciently large to encode the group G.
Example 1.2.6 (The coordinate algebra of a matrix group). Let G be one of the
groups SLn .k/; SOn .k/, or Spn .k/, where n 2 N. For i; j D 1; : : : ; n, denote
by uij W Mn .k/ ! k the function that maps each matrix to its .i; j /th entry. Con-
sider the unital subalgebra O.G/  k.G/ generated by the restrictions of these
functions. If the characteristic of k is 0, this algebra can easily be described
in terms of generators and relations; for P example, O.SLn .k// is isomorphic to
kŒU11 ; : : : ; Unn =.det 1/, where det D  sgn. /  U1.1/ : : : Un.n/ , the sum
being taken over all permutations of the set f1; : : : ; ng.
The spaces .O.G// and S.O.G// are contained in O.G/ ˝ O.G/ and O.G/,
respectively, which implies that the algebra O.G/, equipped with the restrictions
of the maps ; , and S defined in (1.4), forms a Hopf algebra. Indeed, in terms of
the generators uij , the comultiplication and the counit are given by
X
.uij / D uik ˝ ukj and .uij / D ıi;j for all i; j;
k
 1.2. The concept of a Hopf algebra 9
P
respectively, because uij .cd / D k uik .c/ukj .d / and uij .1n / D ıi;j for all
c; d 2 Mn .k/. Using Cramer’s rule, one can also express S.uij / in terms of the
generators ukl , where k; l D 1; : : : ; n.
The groups SLn .k/, SOn .k/, and Spn .k/ are particular examples of linear
algebraic groups or affine algebraic groups. Every affine algebraic group over k
can be encoded by a Hopf algebra:
Example 1.2.7. (The coordinate ring of an affine algebraic group) Let G be an
affine algebraic group over an algebraically closed field k, that is, an affine algebraic
variety over k whose set of points is equipped with the structure of a group, such
that the group operations are morphisms of algebraic varieties (over k). Then the
coordinate ring O.G/, which is the algebra of polynomial functions on the variety G,
is a Hopf algebra with respect to the operations given in formula (1.4), see [19],
[65], [143].
Almost by definition, the assignment G 7! O.G/ defines a contravariant equiv-
alence between the category of affine algebraic groups over k and the category of
commutative Hopf algebras over k that are finitely generated and reduced, see [1,
Chapter 4], [65], or [145, Chapter XV].
Let us consider two elementary examples. To simplify the discussion, we as-
sume that the characteristic of k is 0. Let n 2 N.
• The coordinate Hopf algebra O.kn / of the additive group kn is isomorphic
the polynomial algebra kŒX1 ; : : : ; Xn , where Xi corresponds to the i th co-
ordinate function xi W .c1 ; : : : ; cn / 7! ci . In terms of the generators Xi , the
comultiplication, counit, and antipode are given by

.Xi / D Xi ˝ 1 C 1 ˝ Xi ; .Xi / D 0; S.Xi / D Xi for all i;

because ..xi //.c; d / D xi .c C d / D ci C di D xi .c/ C xi .d /, .xi / D

xi .0/ D 0, and .S.xi //.c/ D xi .c/ D ci for all c; d 2 kn .
• The multiplicative group kn WD .k n f0g/n can be identified with the affine
algebraic variety f.c; d / 2 kn  kn j ci di D 1 for all i g; its coordinate
Hopf algebra O.kn / is isomorphic to the algebra of Laurent polynomials
kŒX1 ; X11 ; : : : ; Xn ; Xn1 , where the comultiplication, counit, and antipode
are given by

.Xi / D Xi ˝ Xi ; .Xi / D 1; S.Xi / D Xi1 for all i:

The Hopf algebras presented next do not consist of functions on a group:

Example 1.2.8 (The group algebra of a discrete group). Let G be a discrete group.
Recall that the group algebra kG of G is the vector space of all finitely supported
10 Chapter 1. Hopf algebras

k-valued
P functions on G, equipped with the convolution product .f  g/.x/ WD
yzDx .y/g.z/. Equivalently, kG can be defined as the universal algebra gener-
f
ated by a family of elements .Ux /x2G , subject to the relation Uy Uz D Uyz for all
y; z 2 G.
The group algebra kG is a Hopf algebra with respect to the maps

 W kG ! kG ˝ kG; W kG ! k; S W kG ! kG;
.Ux / WD Ux ˝ Ux ; .Ux / WD 1; S.Ux / WD Ux 1 ;

where  and are multiplicative, and S is antimultiplicative. Indeed, straightfor-

ward calculations show that for these maps, the diagrams (1.1)–(1.3) commute.
Let us assume that G is abelian and k D C. Then the Hopf algebra kG D CG
y In that case, CG is isomorphic to
can be described in terms of the dual group G:
the Hopf algebra Rep.G/ y introduced in Example 1.2.5. Let us sketch the proof.
y with the group
First, we can identify G
˚ Q  Y
.zx /x2G 2 x2G T j zx zy D zxy for all x; y 2 G  T:
x2G
Q
By Tychonoff’s Theorem, x2G T is compact, and hence G y is compact as well.
y
Since G is abelian, each of its continuous finite-dimensional representations is
equivalent to a direct sum of continuous one-dimensional representations. Such
y and hence, by Pontrja-
representations correspond bijectively with characters on G,
y is generated by the family
gin duality, with elements of G. Thus the algebra Rep.G/
of functions .evx /x2G given by evx ./ D .x/. To show that these functions are
linearly independent, we use the Haar measure O on G: y the linear independence
R R
O
follows easily from the equation Gy evx evy d  D Gy evx 1 y D 0 for all x ¤ y,
R R
which can be deduced from the relation evz ./ Gy evz d O D Gy evz .  /d O D
R
O y y
y evz d ; z 2 G;  2 G. Thus .evx /x is a basis of Rep.G/. Now the assignment
G
Š
Ux 7! evx defines an isomorphism of Hopf algebras CG  ! Rep.G/,y because

.evx evy /./ D evx ./ evy ./ D .x/.y/ D .xy/ D evxy ./;

..evx //.; / D evx ./ D ./.x/ D .x/.x/ D evx ./ evx ./;
.S.evx //./ D evx .1 / D .x/1 D .x 1 / D evx 1 ./;
.evx / D evx .e/ D 1 y and x; y 2 G:
for all ;  2 G

In particular, the Hopf algebra CZn is isomorphic to the Hopf algebra Rep.T n /.
Moreover, if the characteristic of k is 0, then the Hopf algebra kZn is isomorphic
to the Hopf algebra O.kn / (see Example 1.2.7) via the map Uk 7! X1k1 : : : Xnkn ;
where k D .k1 ; : : : ; kn / 2 Zn .
 1.3. Axiomatics of Hopf algebras 11

Example 1.2.9 (The universal enveloping algebra of a Lie algebra). Let g be a

Lie algebra over k. Recall that the universal enveloping algebra U.g/ of g is
the universal unital algebra generated by elements of g, subject to the relation
xy  yx D Œx; y for all x; y 2 g. Equivalently, U.g/ can be characterized
by the following universal property: the algebra U.g/ contains g as a subspace,
and for every unital algebra A and every linear map F W g ! A that satisfies
F .Œx; y/ D F .x/F .y/  F .y/F .x/ for all x; y 2 g, there exists a unique unital
algebra homomorphism U.g/ ! A that extends F . This universal property implies
that the linear maps
 W g ! U.g/ ˝ U.g/; W g ! k; S W g ! g;
.x/ WD x ˝ 1 C 1 ˝ x; .x/ WD 0; S.x/ WD x;

extend to unital algebra homomorphisms

 W U.g/ ! U.g/ ˝ U.g/; W U.g/ ! k; S W U.g/ ! U.g/op ;

where U.g/op denotes the opposite algebra of U.g/, that is, the algebra obtained by
reversing the multiplication. Indeed, for all x; y 2 g,
Œ.x/; .y/ D Œx ˝ 1 C 1 ˝ x; y ˝ 1 C 1 ˝ y
D Œx; y ˝ 1 C 1 ˝ Œx; y D .Œx; y/;
Œ .x/; .y/ D 0 D .Œx; y/;
ŒS.x/; S.y/ D .y/.x/  .x/.y/ D Œx; y D S.Œx; y/:
Routine calculations show that these maps turn U.g/ into a Hopf algebra. For
further details, see [65, Chapter XVI], [56, Chapter 3], or [181, Sections 3.2, 3.4].
Let us give an almost trivial example. Consider kn as a Lie algebra with trivial
Lie bracket given by Œc; d  D 0 for all c; d 2 kn . A comparison with Example 1.2.7
shows that if the characteristic of k is 0, then U.kn / is isomorphic to the Hopf
algebra O.kn /.
Example 1.2.10. Important examples of Hopf algebras which attracted much atten-
tion over the last decades are q-deformations of the coordinate Hopf algebra O.G/
of a semisimple complex Lie group G, and of the universal enveloping algebra
U.g/ of a semisimple complex Lie algebra g. Comprehensive accounts of these
q-deformed Hopf algebras can be found in many books, for example, in [23], [24],
[68], [79], [80], [84], [103], [140]; see also the introduction to Chapter 6.

1.3 Axiomatics of Hopf algebras

To obtain a better understanding of the concept of a Hopf algebra, we shall take
a closer look at the different algebraic structures that appear in the definition. We
12 Chapter 1. Hopf algebras

begin with the comultiplication, which we first consider separately and then in
combination with the multiplication of the underlying algebra. Next, we turn to the
antipode and deduce a list of standard relations which are permanently used in the
theory of Hopf algebras. Finally, we characterize Hopf algebras in terms of two
natural maps which are important for generalizations of Hopf algebras in the setting
of non-unital algebras, of C  -algebras, and of von Neumann algebras.

1.3.1 Coalgebras and bialgebras

At the beginning, the theory of Hopf algebras may be difficult to learn because it
is based not only on the language of algebras and modules, but also on the less
familiar language of coalgebras and comodules. For example, a Hopf algebra is
a unital algebra and simultaneously a counital coalgebra. These structures are
compatible in a sense that is made precise by the concept of a bialgebra, and the
existence of an antipode can be characterized in terms of a convolution product
which merges the two structure maps. In the following paragraphs, we explain
these statements and the terms involved.
An algebra A can be considered as a vector space equipped with a linear map
m W A ˝ A ! A; a ˝ b 7! ab, which is associative in the sense that the square

A˝A˝A
m˝id /A˝A

id˝m m (1.5)
 
A˝A /A
m

commutes. Each element a 2 A gives rise to a linear map  W k ! A;  7! a,

and the element a is the unit for the multiplication if and only if the diagram
Š Š
k˝Ao A /A˝k

˝id id ˝ (1.6)
 
A˝A /Ao A˝A
m m

commutes. Reversing all arrows in this description of an algebra, we obtain the

definition of a coalgebra:
Definition 1.3.1. A coalgebra .over k/ is a vector space A equipped with a linear
map  W A ! A ˝ A called the coproduct or comultiplication that is coassociative
in the sense that .˝idA /ı D .idA ˝ /ı, or, equivalently, that diagram (1.1)
on page 5 commutes.
Let .A; / be a coalgebra. A linear map W A ! k is a counit for .A; / if
. ˝ idA / ı  D idA D .idA ˝ / ı , that is, if the diagram (1.2) on page 5
commutes. A coalgebra is called counital if it has a counit.
 1.3. Axiomatics of Hopf algebras 13

A morphism of coalgebras .A; A / and .B; B / is a linear map F W A ! B

that satisfies B ı F D .F ˝ F / ı A . The map F is counital if A and B have
counits A and B , respectively, and if B ı F D A .
If .A; / is a coalgebra and the comultiplication  is understood, we freely
speak of A itself as a coalgebra.
Remarks 1.3.2. i) Every coalgebra has at most one counit. Indeed, if 1 and 2 are
counits for a coalgebra .A; /, then

1 D 1 ı .idA ˝ 2 / ı  D . 1 ˝ 2 / ı  D 2 ı . 1 ˝ idA / ı  D 2 :

ii) Given coalgebras .A; A / and .B; B /, we can construct the following new
coalgebras:
Coopposite coalgebra. Denote by † W A ˝ A ! A ˝ A the flip map a ˝ b 7!
b ˝ a. Then .A; A /cop WD .A; † ı A / is a coalgebra, called the coopposite
coalgebra of .A; A /. Evidently, a linear map W A ! k is a counit for
.A; A / if and only if it is a counit for .A; A /cop . The coalgebra .A; A / is
called cocommutative if .A; A /cop D .A; A /, that is, if † ı A D A .
Direct sum. Denote by A˚B the composition of the map A ˚B W A˚B !
.A ˝ A/ ˚ .B ˝ B/ with the natural inclusion .A ˝ A/ ˚ .B ˝ B/ ,!
.A ˚ B/ ˝ .A ˚ B/. Then .A ˚ B; A˚B / is a coalgebra. If A and B possess
counits A and B , respectively, then the map .a; b/ 7! A .a/ C B .b/ is a
counit for .A ˚ B; A˚B /.
Tensor product. Denote by A˝B the composition of the map A ˝B W A˝
Š
B ! A˝A˝B ˝B with the isomorphism A˝A˝B ˝B  ! A˝B ˝A˝B
given by a1 ˝ a2 ˝ b1 ˝ b2 7! a1 ˝ b1 ˝ a2 ˝ b2 . Then .A ˝ B; A˝B /
is a coalgebra. If A and B possess counits A and B , respectively, then the
map a ˝ b 7! A .a/ B .b/ is a counit for .A ˝ B; A˝B /.
For calculations in coalgebras, the following Sweedler notation or Sigma nota-
tion is very useful.

Pa coalgebra and a 2 A. Then .a/ 2 A ˝ A can be

Notation 1.3.3. Let .A; / be
written in the form .a/ D i a1;i ˝ a2;i , where a1;i ; a2;i 2 A. To simplify the
notation, we suppress the summation index i and write
X X
.a/ D a1;i ˝ a2;i DW a.1/ ˝ a.2/ :
i

Here, the subscripts “.1/” and “.2/” indicate

P the order of the factors in the tensor
product; thus, for example, †..a// D a.2/ ˝ a.1/ . We extend this notation to
14 Chapter 1. Hopf algebras

iterated applications of  as follows. Since  is coassociative, the elements

X X
.idA ˝/..a// D a.1/ ˝ .a.2/ / D a.1/ ˝ .a.2/ /.1/ ˝ .a.2/ /.2/

and
X X
. ˝ idA /..a// D .a.1/ / ˝ a.2/ D .a.1/ /.1/ ˝ .a.1/ /.2/ ˝ a.2/
P
are equal. We write this element as a.1/ ˝ a.2/ ˝ a.3/ . More generally, consider
the maps .n/ W A ! A ˝    ˝ A .n C 1 factors/, inductively defined by

.0/ WD idA and .nC1/ WD ..n/ ˝ idA / ı  for n 0:

By coassociativity, every map A ! A ˝    ˝ A that is obtained by n successive

applications of  to one factor of the intermediate tensor product A ˝    ˝ A
coincides with .n/ . We write
X
.n/ .a/ DW a.1/ ˝    ˝ a.nC1/ :

Examples 1.3.4. i) Let .A; A / and .B; B / be coalgebras. Then for all a 2 A,
b 2 B,
X
†..a// D a.2/ ˝ a.1/ ;
X X
A˚B ..a; b// D .a.1/ ; 0/ ˝ .a.2/ ; 0/ C .0; b.1/ / ˝ .0; b.2/ /;
X
A˝B .a ˝ b/ D a.1/ ˝ b.1/ ˝ a.2/ ˝ b.2/ :

ii) In Sweedler notation, the axioms for the counit and the antipode of a Hopf
algebra A take the form
X X
.a.1/ /a.2/ D a D a.1/ .a.2/ /
and
X X
S.a.1/ /a.2/ D . .a// D a.1/ S.a.2/ / for all a 2 A:

A combination of these axioms yields the following useful formula:

X X
S.a.1/ /a.2/ ˝ a.3/ D S..a.1/ /.1/ /.a.1/ /.2/ ˝ a.2/
X X
D . .a.1/ // ˝ a.2/ D 1A ˝ .a.1/ /a.2/ D 1A ˝ a:

Next, we consider algebra structures and coalgebra structures that are compatible
in a natural sense.
 1.3. Axiomatics of Hopf algebras 15

Lemma 1.3.5. Let A be a vector space equipped with the structure of an algebra
and of a coalgebra. Then the following conditions are equivalent:

i) The comultiplication  W A ! A ˝ A is an algebra homomorphism.

ii) The multiplication m W A ˝ A ! A is a morphism of coalgebras.
iii) The following diagram commutes:

A˝A
m /A

˝
  (1.7)
A ˝ A ˝ A ˝VA VVVVV
id ˝ †˝ id VV+ 
A˝A˝A˝A / A ˝ A.
m˝m

Proof. Condition iii) is equivalent to i) and ii) because the multiplication and co-
multiplication of A ˝ A are given by

id ˝† ˝ id m˝m
.A ˝ A/ ˝ .A ˝ A/ ! A ˝ A ˝ A ˝ A ! A ˝ A
and
˝ id ˝† ˝ id
A ˝ A ! A ˝ A ˝ A ˝ A ! .A ˝ A/ ˝ .A ˝ A/

respectively. 

Definition 1.3.6. A bialgebra .over k/ is a vector space A equipped with the struc-
ture of an algebra and a coalgebra such that diagram (1.7) commutes. We shall
usually not mention the multiplication map m explicitly and refer to the pair .A; /
consisting of the algebra A and the comultiplication  as a bialgebra.
A bialgebra is called unital if it is unital as an algebra and the comultiplication is
a unital algebra homomorphism; it is called counital if it is counital as a coalgebra
and the multiplication is a counital morphism of coalgebras.
A morphism of bialgebras .A; A / and .B; B / is a linear map F W A ! B
that is a morphism of algebras and of coalgebras. It is called unital/counital if it is
unital/counital as a map of algebras/coalgebras.

Remarks 1.3.7. i) Often, bialgebras are assumed to be unital and counital. We

explicitly state these assumptions whenever we impose them.
ii) For a unital/counital bialgebra .A; /, the compatibility conditions between
the unit and the comultiplication / between the counit and the multiplication amount
16 Chapter 1. Hopf algebras

to the commutativity of the following squares:


k /A A˝A
˝ /k˝k
.
Š  m Š (1.8)
   
k˝k /A˝A A / k:
˝ 

In particular, .A; / is counital if and only if the counit is multiplicative.

ii) Given bialgebras .A; A / and .B; B /, we can construct the following new
bialgebras:

Opposite and coopposite bialgebra. Reversing the multiplication, the comul-

tiplication, or both, of .A; A /, we obtain three new bialgebras. More pre-
cisely, denote by † W A˝A ! A˝A the flip, and by Aop the opposite algebra
of A. Then the pairs .A; A /op WD .Aop ; A /, .A; A /cop WD .A; † ı A /,
.A; A /op;cop WD .Aop ; † ı A / are bialgebras again.

Direct sum and tensor product. The vector spaces A ˚ B and A ˝ B are bial-
gebras with respect to the usual algebra structure and the coalgebra structure
defined in Remark 1.3.2 ii).

iv) Evidently, every Hopf algebra is a unital and counital bialgebra. Furthermore,
every unital and counital bialgebra .A; / admits at most one antipode that turns it
into a Hopf algebra: if S1 and S2 are antipodes for A, then
X X
S1 .a/ D .a.1/ /S1 .a.2/ / D S2 .a.1/ /a.2/ S1 .a.3/ /
X
D S2 .a.1/ / .a.2/ / D S2 .a/ for all a 2 A:

Remarks 1.3.2 i) and 1.3.7 iv) imply that the counit and the antipode of a Hopf
algebra are uniquely determined by the comultiplication. Therefore, we shall refer
to a pair .A; / as a Hopf algebra if A is a Hopf algebra with comultiplication .

1.3.2 Convolution
Let .A; / be a coalgebra and B an algebra over k. Then the space of linear maps
Homk .A; B/ carries an important convolution product, defined by

f  g WD mB ı .f ˝ g/ ı A for all f; g 2 Homk .A; B/;

that is,
X
.f  g/.a/ WD f .a.1/ /g.a.2/ / for all f; g 2 Homk .A; B/; a 2 A:
 1.3. Axiomatics of Hopf algebras 17

The convolution product is associative because

X
..f  g/  h/.a/ D f .a.1/ /g.a.2/ /h.a.3/ / D .f  .g  h//.a/
for all f; g; h 2 Homk .A; B/ and a 2 A. Thus Homk .A; B/ becomes an algebra.
Particularly interesting is the case B D k: The convolution product turns A0 D
Homk .A; k/ into an algebra. When we speak of A0 as an algebra, we always refer
to this algebra structure.
Remarks 1.3.8. i) If .A; / is a finite-dimensional bialgebra or Hopf algebra, then
A0 can also be equipped with the structure of a bialgebra or Hopf algebra, see
Theorem 1.4.1.
ii) In general, the convolution algebra Homk .A; B/ need not be unital. But if
.A; / has a counit A and B has a unit B , then the composition B ı A is a unit
for the convolution algebra Homk .A; B/. Indeed,
X X
..B ı A /  f /.a/ D 1B A .a.1/ /f .a.2/ / D f . A .a.1/ /a.2/ / D f .a/
and similarly .f  .B ı A //.a/ D f .a/ for all f 2 Homk .A; B/ and a 2 A.
iii) Every algebra homomorphism F W B ! C induces an algebra homomor-
phism
F W Homk .A; B/ ! Homk .A; C /; f 7! F ı f:
Indeed, for all f; g 2 Homk .A; B/ and a 2 A,
X
..F f /  .F g//.a/ D F .f .a.1/ //F .g.a.2/ //
X
D F .f .a.1/ /g.a.2/ // D .F .f  g//.a/:
Likewise, if .D; D / is a coalgebra and G W D ! A is a morphism of coalgebras,
then
G  W Homk .A; B/ ! Homk .D; B/; g 7! g ı G;
is an algebra homomorphism. The assignments ..A; /; B/ 7! Homk .A; B/ and
.F; G/ 7! F ı G  D G  ı F define a bifunctor from the categories of coalgebras
and algebras to the category of algebras.
Using the convolution product, we can characterize Hopf algebras among bial-
gebras as follows:
Remark 1.3.9. Let .A; / be a unital and counital bialgebra and S W A ! A a
linear map. Then diagram (1.3) on page 5 commutes if and only if S is inverse to
the identity map idA in the convolution algebra Homk .A; A/, that is, if S  idA D
 ı D idA S. In particular, the bialgebra .A; / can be equipped with the
structure of a Hopf algebra if and only if the identity map idA is invertible in the
convolution algebra Homk .A; A/.
18 Chapter 1. Hopf algebras

Let us turn to another important convolution product. Assume that .A; / is a

coalgebra. For each a 2 A and f 2 A0 , we define
X
f  a WD .idA ˝f /..a// D a.1/ f .a.2/ /
and
X
a  f WD .f ˝ idA /..a// D f .a.1/ /a.2/ :

Lemma 1.3.10. If .A; / is a bialgebra, then the maps

A0  A ! A; .f; a/ 7! f  a; A  A0 ! A; .a; f / 7! a  f;

turn A into a bimodule over the algebra A0 .

Proof. First, we show that the map .f; a/ 7! f  a turns A into a left module over
A0 . For each h 2 A0 , put .h/ WD .id ˝ h/ ı  W A ! A. We need to show that
.f  g/ D .f / .g/ for all f; g 2 A0 . But

.f  g/ D .id ˝.f  g// ı  D .id ˝f ˝ g/ ı .id ˝ / ı 

D .id ˝f ˝ g/ ı . ˝ id/ ı 
D .id ˝f / ı  ı .id ˝ g/ ı  D .f / .g/:

A similar argument shows that the map .a; f / 7! a  f turns A into a right
module over A0 . Finally, these maps turn A into a bimodule because for all f; g 2 A0
and a 2 A,
X
.f  a/  g D f .a.1/ /a.2/ g.a.3/ / D f  .a  g/: 

Remark 1.3.11. If .A; / is a coalgebra with counit , then  a D a D a  for

all a 2 A; if .A; / is a Hopf algebra, then S  a D . .a// D a  S for all a 2 A.

1.3.3 Properties of the antipode

The antipode of a Hopf algebra satisfies several fundamental relations that are not
obvious from the definition. To some extent, the antipode of a Hopf algebra behaves
like the inversion of a group: the inversion of a group is antimultiplicative, and the
antipode of a Hopf algebra is both antimultiplicative and anticomultiplicative.
Proposition 1.3.12. The antipode of a Hopf algebra .A; / is a unital and counital
morphism .A; / ! .A; /op;cop of bialgebras, that is, the following conditions
hold:

i/ S ı m D m ı † ı .S ˝ S /; ii/ S ı  D ;
iii/  ı S D .S ˝ S/ ı † ı ; iv/ ı S D :
 1.3. Axiomatics of Hopf algebras 19

Equivalently, for all a; b 2 A,

i/0 S.ab/ D S.b/S.a/; ii/0 S.1A / D 1A ;

X X
iii/0 .S.a//.1/ ˝ .S.a//.2/ D S.a.2/ / ˝ S.a.1/ /; iv/0 .S.a// D .a/:

Proof. i) Consider A˝A as a coalgebra (see Remark 1.3.2 ii)) and Homk .A˝A; A/
as an algebra with respect to the convolution product (see Section 1.3.2). We claim:

.S ı m/  m D  ı ı m D m  .m ı † ı .S ˝ S //: (1.9)

Since ı m W A ˝ A ! k is the counit of A ˝ A, the map  ı ı m W A ˝ A ! A is

the unit in Homk .A ˝ A; A/ (see Remark 1.3.8 ii)). Thus, equation (1.9) implies
that m is invertible with respect to the convolution product, and that the inverse of
m is S ı m D m ı † ı .S ˝ S/. Let us prove (1.9). For all a; b 2 A,

X
..S ı m/  m/.a ˝ b/ D S.m.a.1/ ˝ b.1/ //  m.a.2/ ˝ b.2/ /
X
D S.a.1/ b.1/ /a.2/ b.2/
X
D S..ab/.1/ /.ab/.2/ D . .ab//;
X
.m  .m ı † ı .S ˝ S ///.a ˝ b/ D m.a.1/ ˝ b.1/ /  m.S.b.2/ / ˝ S.a.2/ //
X
D a.1/ b.1/ S.b.2/ /S.a.2/ /
X
D a.1/ . .b//S.a.2/ / D . .a//. .b//:

ii) We have 1A D . ı /.1A / D .m ı .S ˝ id/ ı /.1A / D S.1A /1A .

iii) Consider A ˝ A as an algebra and Homk .A; A ˝ A/ as an algebra with
respect to the convolution product. We claim:

  . ı S/ D  ı  ı D ..S ˝ S / ı † ı /  : (1.10)

Since .1A / D 1A ˝ 1A , the map  ı  W k ! A ˝ A is the unit map of A ˝ A.

Therefore  ı  ı is the unit in Homk .A; A ˝ A/ (see Remark 1.3.8 ii)). Thus,
equation (1.10) implies that  is invertible with respect to the convolution and that
20 Chapter 1. Hopf algebras

the inverse of  is  ı S D .S ˝ S/ ı † ı . Let us prove (1.10). For all a 2 A,

X
.  . ı S //.a/ D .a.1/ /.S.a.2/ //
X
D .a.1/ S.a.2/ // D . ı  ı /.a/;
X
...S ˝ S/ ı † ı /  /.a/ D .S.a.2/ / ˝ S.a.1/ //  .a.3/ ˝ a.4/ /
X
D S.a.2/ /a.3/ ˝ S.a.1/ /a.4/
X
D 1A ˝ .a.2/ /S.a.1/ /a.3/
X
D 1A ˝ S.a.1/ /a.2/ D . ı  ı /.a/:

iv) For all a 2 A,

X
.S.a// D .S.a.1/ // .a.2/ /
X
D .S.a.1/ /a.2/ / D . ı  ı /.a/ D .a/: 

Remarks 1.3.13. i) In the proof above, the equations .S ı m/  m D  ı ı m and

  . ı S / D  ı  ı can also be deduced from Remark 1.3.8 iii): since m is a
morphism of coalgebras and  is an algebra homomorphism,

.S ı m/  m D m .S /  m .idA / D m .S  idA / D m . ı / D  ı ı m;
  . ı S/ D  .idA /   .S / D  .idA S / D  . ı / D  ı  ı :

ii) Assuming some familiarity with the Sweedler notation, we can verify condi-
tions i)0 and ii)0 in Proposition 1.3.12 by direct calculations:
X
S.a/S.b/ D S.a.1/ /S.b.1/ / .b.2/ / .a.2/ /
X
D S.a.1/ /S.b.1/ / .b.2/ a.2/ /
X
D S.a.1/ /S.b.1/ /.b.2/ a.2/ /.1/ S..b.2/ a.2/ /.2/ /
X
D S.a.1/ /S.b.1/ /b.2/ a.2/ S.b.3/ a.3/ /
X
D S.a.1/ /a.2/ S.ba.3/ / D S.ba/
 1.3. Axiomatics of Hopf algebras 21

and
X X
S.a.2/ / ˝ S.a.1/ / D.S.a.2/ / ˝ S.a.1/ //.. .a.3/ ///
X
D .S.a.2/ / ˝ S.a.1/ //.a.3/ S.a.4/ //
X
D .S.a.2/ /a.3/ ˝ S.a.1/ /a.4/ /.S.a.5/ //
X
D .. .a.2/ // ˝ S.a.1/ /a.3/ /.S.a.4/ //
X
D .1A ˝ S.a.1/ /a.2/ /.S.a.3/ // D .S.a//:

The previous proposition implies that for every Hopf algebra, the square of the
antipode is a morphism of the Hopf algebra. But unlike the inversion of a group,
the antipode of a Hopf algebra need not be involutive, that is, S 2 need not be equal
to the identity – the antipode need not even bijective.
Proposition 1.3.14. For every Hopf algebra .A; /, the following conditions are
equivalent:
i) The antipode S of .A; / is bijective.
ii) The bialgebra .A; /op is a Hopf algebra.
iii) The bialgebra .A; /cop is a Hopf algebra.
If these conditions hold, then S 1 is the antipode of .A; /op and of .A; /cop .
The proof involves the following evident statement:
Lemma 1.3.15. Let .A; / be a Hopf algebra and T W A ! A a linear map. Then
the following conditions are equivalent:
i) The bialgebra .A; /op is a Hopf algebra with antipode T .
ii) m ı † ı .T ˝ id/ ı  D  ı D m ı † ı .id ˝T / ı .
P P
iii) a.2/ T .a.1/ / D . .a// D T .a.2/ /a.1/ for all a 2 A.
iv) m ı .id ˝T / ı † ı  D  ı D m ı .T ˝ id/ ı † ı .
v) The bialgebra .A; /cop is a Hopf algebra with antipode T . 
Proof of Proposition 1.3.14. i) ) ii), iii): Suppose that S is invertible. We use
Lemma 1.3.15 to show that S 1 is the antipode of .A; /op and of .A; /cop . By
Proposition 1.3.12,
X X
a.2/ S 1 .a.1/ / D S 1 .a.1/ S.a.2/ // D S 1 .. .a/// D . .a//
22 Chapter 1. Hopf algebras
P
and similarly S 1 .a.2/ /a.1/ D . .a// for all a 2 A.
ii), iii) ) i): Let .A; /op or .A; /cop be a Hopf algebra with antipode T .
Then by Proposition 1.3.12 and Lemma 1.3.15,
X X
S.T .a// D .a.2/ /S.T .a.1/ // D a.3/ T .a.2/ /S.T .a.1/ //
X X
D a.2/ T .a.1/ /.1/ S.T .a.1/ /.2/ / D a.2/ .T .a.1/ // D a
and similarly T .S.a// D a for all a 2 A. 
Note that for every Hopf algebra .A; /, the bialgebra .A; /op;cop is a Hopf
algebra with the same antipode as .A; /.
Corollary 1.3.16. For every commutative or cocommutative Hopf algebra .A; /,
we have S 2 D idA .
Proof. In both cases, S 1 and S are antipodes for .A; /op and .A; /cop . By
Remark 1.3.7 iv), S D S 1 . 
Proposition 1.3.17. Let .A; A / and .B; B / be Hopf algebras and F W A ! B
a unital and counital morphism of bialgebras. Then F ı SA D SB ı F , that is, F
is a morphism of Hopf algebras.
Proof. Consider the convolution algebra Homk .A; B/. By Remark 1.3.8 ii), its
unit is B ı A D B ı B ı F D F ı A ı A , and by Remark 1.3.8 iii),
.SB ı F /  F D .SB  idB / ı F D B ı B ı F
D F ı A ı A D F ı .idA SA / D F  .F ı SA /:
Therefore F is invertible with respect to the convolution, and its inverse is SB ıF D
F ı SA . 

1.3.4 Another characterization of Hopf algebras

By Remark 1.3.9, Hopf algebras can be characterized as those bialgebras for which
the identity map is invertible with respect to the convolution product. Now, we
present another characterization of Hopf algebras which is particularly well suited
for generalizations to non-unital algebras and C  -algebras. This characterization
was given by Van Daele [174], who used it for the definition of multiplier Hopf
algebras (see Chapter 2).
Let .A; / be a unital bialgebra. Consider the linear maps
T1 WD .id ˝m/ ı . ˝ id/ W A ˝ A ! A ˝ A;
X
a ˝ b 7! a.1/ ˝ a.2/ b D .a/.1A ˝ b/;
(1.11)
 1.3. Axiomatics of Hopf algebras 23

and

T2 WD .m ˝ id/ ı .id ˝/ W A ˝ A ! A ˝ A;

X
a ˝ b 7! ab.1/ ˝ b.2/ D .a ˝ 1A /.b/:
(1.12)

These maps will play a central rôle in later chapters.

Theorem 1.3.18. A unital bialgebra is a Hopf algebra if and only if the associated
maps T1 and T2 are bijective.

This theorem shows that the existence of a counit and of an antipode can be
expressed solely in terms of the bialgebra structure. Before we proceed to the
proof, we consider an analogous result concerning semigroups for motivation:

Proposition 1.3.19. Let  be a semigroup. Denote by m W    !  the

semigroup operation, and by  W  !    the diagonal embedding. Then  is
a group if and only if the following two maps are bijective:

ˆ1 WD .id m / ı .  id / W    !   ; .x; y/ 7! .x; xy/;

ˆ2 WD .m  id / ı .id  / W    !   ; .x; y/ 7! .xy; y/:

Proof. If  is a group, then the maps ‰1 ; ‰2 W   !   given by ‰1 .x; y/ WD

.x; x 1 y/ and ‰2 .x; y/ WD .xy 1 ; y/ are inverse to ˆ1 and ˆ2 , respectively.
Conversely, assume that the maps ˆ1 and ˆ2 are bijective. First, we show that
 has a unit. Fix x 2 . By assumption, the map lx W  !  given by y 7! xy is
bijective. Hence there exists an element e 2  such that x D xe. Then zxe D zx
for all z 2 , and by surjectivity of the map rx W  ! ; z 7! zx, we get y D ye
for all y 2 . On the other hand, xy D xey for all y 2 , and using injectivity of
the map lx , we find y D ey for all y 2 . Thus e is a unit.
Finally, by surjectivity of lx and rx , the element x has a right and a left inverse,
which necessarily coincide. Since x 2  was arbitrary,  is a group. 

Remark 1.3.20. For a finite semigroup , the maps ˆ1 and ˆ2 are bijective if
and only if they are injective, and this holds if and only if  has the following
cancellation property: if x; y1 ; y2 2  satisfy xy1 D xy2 , then y1 D y2 , and if
x1 ; x2 ; y 2  satisfy x1 y D x2 y, then x1 D x2 .

Now we return to the maps T1 and T2 introduced above and take one step towards
the proof of Theorem 1.3.18:

Lemma 1.3.21. Let .A; / be a Hopf algebra. Then the maps T1 and T2 defined
in (1.11) and (1.12), respectively, are bijective.
24 Chapter 1. Hopf algebras

Proof. Consider the maps

R1 WD .id ˝m/ ı .id ˝S ˝ id/ ı . ˝ id/ W A ˝ A ! A ˝ A;

X
a ˝ b 7! a.1/ ˝ S.a.2/ /b D ..id ˝S /..a///.1 ˝ b/;
and
R2 WD .m ˝ id/ ı .id ˝S ˝ id/ ı .id ˝/ W A ˝ A ! A ˝ A;
X
a ˝ b 7! aS.b.1/ / ˝ b.2/ D .a ˝ 1/..S ˝ id/..b///:

We show that R1 is inverse to T1 , and a similar calculation shows that R2 is inverse

to T2 : for all a; b 2 A,
X 
R1 .T1 .a ˝ b// D R1 a.1/ ˝ a.2/ b
X X
D a.1/ ˝ S.a.2/ /a.3/ b D a.1/ ˝ .a.2/ /b D a ˝ b
and
X 
T1 .R1 .a ˝ b// D T1 a.1/ ˝ S.a.2/ /b
X X
D a.1/ ˝ a.2/ S.a.3/ /b D a.1/ ˝ .a.2/ /b D a ˝ b: 

Before we prove the reverse implication, let us collect some useful relations
between the maps T1 , T2 , the multiplication m, and the comultiplication :
X
. ˝ id/ ı T1 D .id ˝T1 / ı . ˝ id/ W a ˝ b 7! a.1/ ˝ a.2/ ˝ a.3/ b;
X
T1 ı .id ˝m/ D .id ˝m/ ı .T1 ˝ id/ W a ˝ b ˝ c 7! a.1/ ˝ a.2/ bc;
X (1.13)
.id ˝/ ı T2 D .T2 ˝ id/ ı .id ˝/ W a ˝ b 7! ab.1/ ˝ b.2/ ˝ b.3/ ;
X
T2 ı .m ˝ id/ D .m ˝ id/ ı .id ˝T2 / W a ˝ b ˝ c 7! abc.1/ ˝ c.2/ :

The following proposition completes the proof of Theorem 1.3.18:

Proposition 1.3.22 ([174]). Let .A; / be a unital bialgebra. If the maps T1 and T2
defined in (1.11) and (1.12) are bijective, then .A; / is a Hopf algebra.

Proof. We need to construct a counit and an antipode for .A; /. Let us start with
the counit. The proof of the previous lemma shows that if .A; / were a Hopf
then we could express the counit in terms of the map T1 as follows:
algebra,P
.a/ D a.1/ S.a.2/ / D m.T11 .a ˝ 1A // for all a 2 A. So, consider the map

E W A ! A; a 7! m.T11 .a ˝ 1A //:
 1.3. Axiomatics of Hopf algebras 25

We show that the image of E is contained in k  1A . Using the first relation in (1.13),
we find that for all a; b 2 A,
.id ˝E/..a ˝ 1A /.b// D .a ˝ 1A /  .id ˝m/..id ˝T11 /..b/ ˝ 1A //
D .a ˝ 1A /  .id ˝m/.. ˝ id/.T11 .b ˝ 1A ///
D .a ˝ 1A /  T1 .T11 .b ˝ 1A // D ab ˝ 1A :
Since T2 is surjective, elements of the form .a ˝ 1A /.b/ span A ˝ A. Therefore,
the calculation above shows that the image of E is contained in k  1A .
Define W A ! k by E.a/ D .a/  1A for all a 2 A. We show that is a
counit. Let b 2 A. By the calculation above, .id ˝ /..b// D b. Using the
second equation in (1.13), we find
X
. ˝ id/..b// D m.T11 .b.1/ ˝ 1A //  b.2/
D m..id ˝m/.T11 .b.1/ ˝ 1A / ˝ b.2/ //
D m.T11 .b.1/ ˝ b.2/ // D m.b ˝ 1A / D b:
It remains to show that is a homomorphism. The previous results and multiplica-
tivity of  imply
X X
.a.1/ / .b.1/ /a.2/ b.2/ c D abc D .a.1/ b.1/ /a.2/ b.2/ c for all a; b; c 2 A:
P
Because T2 is surjective, we can replace a.1/ ˝ b.1/ ˝ a.2/ b.2/ c by a0 ˝ b 0 ˝ 1A ,
where a0 ; b 0 2 A are arbitrary. Thus we find .a0 / .b 0 / D .a0 b 0 / for all a0 ; b 0 2 A.
Next, we construct the antipode. The proof of the previous lemma shows that
if .A; / were a Hopf algebra,Pthen we could express the antipode S in terms of
T1 and as follows: S.a/ D .a.1/ /S.a.2/ / D . ˝ id/.T11 .a ˝ 1A // for all
a 2 A. So, consider the map
S W A ! A; a 7! . ˝ id/.T11 .a ˝ 1A //:
Let a 2 A. From the relation .id ˝m/ ı .T1 ˝ id/1 D T11 ı .id ˝m/, we deduce
X X
S.a.1/ /a.2/ D . ˝ id/.T11 .a.1/ ˝ 1A //  a.2/
X
D . ˝ id/.T11 .a.1/ ˝ a.2/ // D . ˝ id/.a ˝ 1A / D .a/1A :

Since . ˝ id/ ı T11 D .id ˝T1 /1 ı . ˝ id/ and .id ˝ / ı  D id,
X X
a.1/ S.a.2/ / D a.1/  . ˝ id/.T11 .a.2/ ˝ 1A //
X
D .m ı .id ˝ ˝ id/ ı .id ˝T11 //.a.1/ ˝ a.2/ ˝ 1A /
D .m ı .id ˝ ˝ id/ ı . ˝ id/ ı T11 /.a ˝ 1A /
D m.T11 .a ˝ 1A // D .a/1A : 
26 Chapter 1. Hopf algebras

Remark 1.3.23. In the proof of Lemma 1.3.21, we did not use the fact that the
counit is an algebra homomorphism. By Theorem 1.3.18, this assumption can be
dropped from the definition of a Hopf algebra because it is implied by the remaining
assumptions.

1.3.5 Hopf -algebras

Hopf -algebras are Hopf algebras equipped with a conjugate-linear involution that
is compatible with the bialgebra structure in a natural way:

Definition 1.3.24. An involution on a complex vector space A is a map  W A ! A,

a 7! a ; that is conjugate-linear and involutive in the sense that
N  ; .a / D a
.a C b/ D a C b  ; .a/ D a for all a; b 2 A and  2 C:

A complex vector space with a fixed involution is also called a -vector space.
A linear map  W A ! B of -vector spaces is -linear if .a / D .a/ for
all a 2 A.
A -algebra is a complex algebra A equipped with an involution such that
.ab/ D b  a for all a; b 2 A. A -coalgebra is a complex Pcoalgebra .A; /,
where A is equipped with an involution such that .a / D a.1/  ˝ a.2/  for
all a 2 A. A -bialgebra is a complex bialgebra .A; /, where A is a -algebra
and .A; / a -coalgebra. A -bialgebra that is a Hopf algebra is called a Hopf
-algebra.
A morphism of -algebras/-coalgebras/-bialgebras/Hopf -algebras is a
-linear morphism of the underlying algebras/coalgebras/bialgebras/Hopf algebras.
A morphism of -algebras is also called a -homomorphism.

An important class of Hopf -algebras – the class of algebraic compact quantum

groups – is studied in detail in Chapter 3, and analogues of Hopf -algebras in the
setting of C  -algebras and von Neumann algebras are discussed in Part II.

Remarks 1.3.25. i) Note the following asymmetry in the definition of -algebras

and -coalgebras: for a -algebra A, the involution reverses the multiplication and
can be considered as a homomorphism A ! Aop , whereas for a -coalgebra .A; /,
the involution does not reverse the comultiplication but is a coalgebra morphism
.A; / ! .A; /. P 
ii) Given
P a -coalgebra .A; / and an element a 2 A, the expressions
P a .1/ ˝
 
a .2/ and a.1/  ˝ a.2/  coincide; hence we shortly write a.1/ ˝ a.2/ .

Examples 1.3.26. i) Let G be a finite group. Then the Hopf algebra C.G/ defined
in Example 1.2.4 is a Hopf -algebra with respect to the involution f 7! f  given
by f  .x/ WD f .x/ for all x 2 G and f 2 C.G/.
 1.3. Axiomatics of Hopf algebras 27

ii) For a topological group G, the same formula as above turns the Hopf algebra
Rep.G/ defined in Example 1.2.5 into a Hopf -algebra; note that Rep.G/ is closed
under this involution.
iii) Let G be a discrete group. Then the Hopf algebra CG defined in Exam-
ple 1.2.8 is a Hopf -algebra with respect to the involution Ux 7! Ux 1 ; x 2 G.
iv) The group SU2 .C/ is equal to the subgroup of SL2 .C/ that consists of all
matrices A that satisfy A D A1 . In terms of matrix elements, the last relation
reads
     1  
˛N N ˛ ˇ ˛ ˇ ı ˇ
D D D :
ˇN ıN  ı  ı  ˛
Motivated by this equation, we equip the Hopf algebra O.SL2 .C// defined in Ex-
ample 1.2.6 with the involution given by
u11  WD u22 ; u12  WD u21 ; u21  WD u12 ; u22  WD u11 :
It is easy to check that this involution turns O.SL2 .C// into a Hopf -algebra.
This Hopf -algebra corresponds to the group SU2 .C/ and will be denoted by
O.SU2 .C//.
In a Hopf -algebra, the behavior of the multiplication and comultiplication
with respect to the involution is prescribed by the definition. For the counit and
antipode, we obtain the following relations:
Proposition 1.3.27. The counit of a counital -coalgebra is -linear.
Proof. Let .A; / be a -coalgebra with counit . Then the map  W A ! C given
by  .a/ D .a / for all a 2 A is a counit for A as well, since
X X
.id ˝  /..a // D   
a.1/ .a.2/ / D .a.1/ .a.2/ // D a

and similarly .  ˝ id/..a // D a for all a 2 A. By Remark 1.3.2 i), D  .


Proposition 1.3.28. The antipode of a Hopf -algebra is bijective and satisfies
S ı  ı S ı  D id.
Proof. Let .A; / be a Hopf -algebra. Then the map S  W A ! A given by
S  .a/ D S.a / for all a 2 A is an antipode for the bialgebra .A; /op , since
X X

a.2/ S  .a.1/

/D 
a.2/ S.a.1/ /
X
D .S.a.1/ /a.2/ / D . .a// D . .a //
P   
and likewise S .a.2/ /a.1/ D . .a // for all a 2 A. By Proposition 1.3.14,
S  D S 1 . 
28 Chapter 1. Hopf algebras

Corollary 1.3.29. The antipode of a Hopf -algebra is -linear if and only if it is

involutive in the sense that S 2 D id.
Proof. By the previous proposition, the map S ı  is invertible, and

 ı S D S ı  , .S ı / ı . ı S/ D .S ı / ı .S ı / , S 2 D id : 

1.4 The duality of Hopf algebras

The concept of a bialgebra and of a Hopf algebra has an intrinsic symmetry that
gives rise to a duality which can be considered as an analogue of Pontrjagin duality.
In the finite-dimensional case, this duality behaves very nicely. In the infinite-
dimensional case, additional concepts and stronger assumptions are needed, and a
satisfying duality theory will only be achieved in Chapter 2.

1.4.1 The duality of finite-dimensional Hopf algebras

Let us fix some notation. Consider a vector space V . We denote by V 0 D
Homk .V; k/ the dual space, and by V W V ! V 00 the natural embedding given
by .V .v//.f / WD f .v/ for all v 2 V; f 2 V 0 . Furthermore, we consider V 0 ˝ V 0
as a subspace of .V ˝V /0 via the embedding given by .f ˝g/.v ˝w/ D f .v/g.w/
for all v; w 2 V and f; g 2 V 0 . It is easy to see that this embedding, like the embed-
ding V , is an isomorphism if and only if V has finite dimension. Finally, recall that
every linear map of vector spaces F W V ! W induces a dual map F 0 W W 0 ! V 0
by composition, that is, F 0 .f / D f ı F for all f 2 W 0 .
Theorem 1.4.1. i) Let .A; A / be a coalgebra. Then the dual space A0 is an algebra
with respect to the multiplication
.A /0
mA0 W A0 ˝ A0 ,! .A ˝ A/0 ! A0 ; .fg/.a/ D .f ˝ g/..a//:

The algebra A0 is unital if and only if .A; A / is counital, and in this case, the unit
of A0 coincides with the counit of .A; A /.
ii) Let A be a finite-dimensional algebra. Then the dual space A0 is a coalgebra
with respect to the comultiplication
.mA /0
A0 W A0 ! .A ˝ A/0 Š A0 ˝ A0 ; .A0 .f //.a ˝ b/ D f .ab/:

The coalgebra .A0 ; A0 / is counital if and only if A is unital, and in this case, the
counit of .A0 ; A0 / coincides with evaluation at the unit of A.
iii) Let .A; A / be a finite-dimensional bialgebra. Then A0 , equipped with the
multiplication and comultiplication defined above, is a bialgebra. If .A; A / is a
 1.4. The duality of Hopf algebras 29

Hopf algebra, then .A0 ; A0 / is a Hopf algebra with antipode SA0 D .SA /0 . The
Š
natural isomorphism A W A  ! A00 of vector spaces is an isomorphism of bialgebras
or Hopf algebras, respectively.
iv) Let .A; A / be a Hopf -algebra. Then A0 is a -algebra with respect to
the involution given by f  .a/ WD f .S.a/ / for all a 2 A; f 2 A0 . If .A; A / is a
finite-dimensional Hopf -algebra, then .A0 ; A0 / is a Hopf -algebra again, and
Š
! A00 is an isomorphism of Hopf -algebras.
the natural isomorphism A W A 
Proof. i)–iii) All statements follow easily from the symmetry of the commutative
diagrams that express the axioms for the structure maps involved: If we apply the
dualization functor A 7! A0 , F 7! F 0 , to the commutative diagrams
(1.1); (1.2); (1.5); (1.6); (1.7); (1.8); (1.3) for A
and use the isomorphism .A ˝ A/0 Š A0 ˝ A0 in ii), iii), we find that the diagrams
(1.5); (1.6); (1.1); (1.2); (1.7); (1.8); (1.3) for A0
commute. Let us give an example. If we apply the dualization functor to the right
square of diagram (1.3), we obtain
.A /0 mA0
AO 0 o AO 0 o
A
A / A˝A .A˝A/0 A0 ˝A
O
0
O
A ıA id ˝SA 7! .A ıA /0 .id ˝SA /0
A0 ıA0 id ˝SA0

 
Ao / .A˝A/0 A0 / A0 ˝A0 ,
mA A˝A A0 .mA /0 A0

which is the right square of diagram (1.3) for A0 .

iv) Let us show that the involution defined in iv) turns A0 into a -algebra. We
denote by  W C ! C the complex conjugation. Then f  D  ı f ı  ı SA for all
f 2 A0 , and from Propositions 1.3.28 and 1.3.12, we find
.f  / D  ı . ı f ı  ı SA / ı  ı SA D f
and
.fg/ D  ı .f ˝ g/ ı A ı  ı SA
D .. ı g ı  ı SA / ˝ . ı f ı  ı SA // ı A D g  f 
for all f; g 2 A0 . Now, assume that A has finite dimension, and let f 2 A0 . By
definition and by Proposition 1.3.12,
A0 .f  / D f  ı mA D  ı f ı  ı SA ı mA
D  ı f ı mA ı .. ı SA / ˝ . ı SA //:
30 Chapter 1. Hopf algebras
P
Inserting the relation f ı mA D A0 .f / D f.1/ ˝ f.2/ , we find
X X
A0 .f  / D . ı f.1/ ı  ı SA / ˝ . ı f.2/ ı  ı SA / D f.1/  ˝ f.2/  :

Thus .A0 ; A0 / is a -coalgebra. By iii), it is also a Hopf algebra, and hence a Hopf
-algebra.
Š
! A00 of Hopf algebras
Finally, let us show that the natural isomorphism A W A 
is -linear. By definition and by Proposition 1.3.28,

..a// .f / D .a/.SA0 .f / / D .SA0 .f / /.a/

D .SA0 .f //.SA .a/ / D .f ı SA ı  ı SA /.a/ D f .a /
D ..a //.f /

for all a 2 A and f 2 A0 . 

The following two examples explain the relation between the duality of finite-
dimensional Hopf algebras and the Pontrjagin duality of finite abelian groups.
Example 1.4.2. Let G be a finite group and consider the associated Hopf algebra
k.G/ defined in Example 1.2.4. Denote by .ıx /x2G the canonical basis of k.G/, and
by ."x /x2G the dual basis of k.G/0 , determined by "x .ıy / D ıx;y for all x; y 2 G.
We compute the structure maps of the Hopf algebra k.G/0 . To distinguish between
the counit, comultiplication, and antipode of k.G/ and k.G/0 , we index the structure
maps of k.G/0 .
• The product of two elements "x ; "y 2 k.G/0 is determined by
X
."x  "y /.ız / D ."x ˝ "y /..ız // D "x .ıx 0 /  "y .ıy 0 / D ıxy;z ;
x 0 ;y 0 2G
x 0 y 0 Dz

whence "x "y D "xy . If e 2 G denotes the unit, then "e is the unit of k.G/0 .
• The coproduct of an element "z 2 k.G/0 is determined by

.k.G/0 ."z //.ıx ˝ ıy / D "z .ıx  ıy / D ız;x  ız;y ;

so k.G/0 ."z / D "z ˝ "z .

P
• The counit of k.G/0 acts by evaluation at the unit 1k.G/ D z ız , whence
k.G/0 ."z / D 1 for all z 2 G.
• The antipode Sk.G/0 , applied to an element "x 2 k.G/0 , acts by

.Sk.G/0 ."x //.ıy / D "x .S.ıy // D "x .ıy 1 / D ıx;y 1 ;

and hence Sk.G/0 ."x / D "x 1 .

 1.4. The duality of Hopf algebras 31

Comparing with the definition of the Hopf algebra kG in Example 1.2.8, we find
that the map k.G/0 ! kG given by "x 7! Ux for all x 2 G is an isomorphism
of Hopf algebras. If k D C, then this is also an isomorphism of Hopf -algebras,
because "x  D "x 1 for all x 2 G:
"x  .ıy / D "x .S.ıy // D "x .ıy 1 / D ıx;y 1 :

Example 1.4.3. Let G be a finite abelian group. By the previous example, C.G/0 Š
CG. Moreover, the Fourier transform F W CG ! C.G/ y given by .F Ux /./ WD
y
.x/ for all  2 G and x 2 G is an isomorphism of Hopf algebras: this is just
Š
a special case of the isomorphism CG  ! Rep.G/ y explained in Example 1.2.8,
y y y
and since G is finite, Rep.G/ D C.G/. Summarizing, we find that the Pontrjagin
duality of finite abelian groups and the duality of finite-dimensional Hopf algebras
fits into the following scheme:
 dual /G
G group y

C.G/  / C.G/
dual
y
Hopf algebra

The duality established in Theorem 1.4.1 iii) does not easily extend to infinite-
dimensional bialgebras or Hopf algebras: For an infinite-dimensional algebra A,
the inclusion A0 ˝ A0  .A ˝ A/0 is strict. Therefore, the dual of the multiplication
map .mA /0 W A0 ! .A ˝ A/0 need not define a comultiplication on A0 – its image
may fail to be contained in the subspace A0 ˝ A0  .A ˝ A/0 . This problem can be
addressed in several ways. We can

• consider dual pairings of A with other Hopf algebras;

• look for a subalgebra Aı  A0 that satisfies .mA /0 .Aı /  Aı ˝ Aı ;
y /,
• consider non-unital Hopf algebras .A; y for which the image of  y is con-
y y y y
tained in a space of multipliers M.A ˝ A/ that is larger than A ˝ A.

In the following sections, we discuss the first and second approach; the third one
involves several additional concepts and is presented in Chapter 2.

1.4.2 Dual pairings of Hopf algebras

Let .A; A / be a finite-dimensional bialgebra. In Theorem 1.4.1, we saw that the
dual space A0 of A is a bialgebra again. The relations between the structure maps
32 Chapter 1. Hopf algebras

of .A; A / and .A0 ; A0 / can be expressed conveniently in terms of the natural
pairings

.  j  / W A  A0 ! k; .  j  / W .A ˝ A/  .A0 ˝ A0 / ! k;
.ajf / WD f .a/; .a1 ˝ a2 jf1 ˝ f2 / WD f1 .a1 /  f2 .a2 /;

as follows: for all a; a1 ; a2 2 A and f; f1 ; f2 2 A0 ,

X
.ajf1 f2 / D .ajmA0 .f1 ˝ f2 // D .A .a/jf1 ˝ f2 / D .a.1/ jf1 /.a.2/ jf2 /
and
X
.a1 a2 jf / D .mA .a1 ˝ a2 /jf / D .a1 ˝ a2 jA0 .f // D .a1 jf.1/ /.a2 jf.2/ /:

Furthermore, if .A; A / is a Hopf algebra (or Hopf -algebra), then so is .A0 ; A0 /,
and in that case, the unit, counit, antipode (and involution) of .A; A / and .A0 ; A0 /
are related by similar equations. These relations motivate the following definition:
Definition 1.4.4. A dual pairing between two Hopf algebras (Hopf -algebras)
.A; A / and .B; B / is a bilinear map .  j  / W A  B ! k, written .a; b/ 7! .ajb/,
that satisfies
X X
.ajb1 b2 / D .a.1/ jb1 /.a.2/ jb2 /; .a1 a2 jb/ D .a1 jb.1/ /.a2 jb.2/ /;

.aj1B / D A .a/; .1A jb/ D B .b/; .SA .a/jb/ D .ajSB .b// (1.14)

.and .ajb  / D .SA .a/ jb/; .a jb/ D .ajSB .b/ /; respectively/

for all a; a1 ; a2 2 A and b; b1 ; b2 2 B. The dual pairing is called perfect or non-

degenerate if for each non-zero a0 2 A and each non-zero b0 2 B, there exist
a 2 A and b 2 B such that .a0 jb/ ¤ 0 and .ajb0 / ¤ 0.
Remarks 1.4.5. i) Let .  j  / W A  B ! k be a dual pairing of Hopf algebras. Then
each a 2 A defines a linear map .aj  / W B ! k; b 7! .ajb/, and the map A ! B 0
given by a 7! .aj  / is a unital homomorphism of algebras. Similarly, we obtain a
unital homomorphism B ! A0 ; b 7! .  jb/. The pairing is perfect if and only if
these homomorphisms are injective.
ii) For every finite-dimensional Hopf algebra .A; A /, the canonical pairing
between A and A0 is a perfect dual pairing of Hopf algebras. If .B; B / is another
Hopf algebra and .  j  / W A  B ! k is a perfect dual pairing, then B has finite
dimension and the homomorphisms A ! B 0 , a 7! .aj  /; and B ! A0 , b 7! .  jb/;
are isomorphisms of Hopf algebras.
iii) Let .A; A / and .B; B / be Hopf -algebras and .  j  / W A  B ! k a dual
pairing of Hopf algebras. Then .ajb  / D .SA .a/ jb/ for all a 2 A and b 2 B
 1.4. The duality of Hopf algebras 33

if and only if .a jb/ D .ajSB .b/ / for all a 2 A and b 2 B. Indeed, if the first
condition holds, then by Proposition 1.3.28,

.a jb/ D .SA .SA .a/ /jb/ D .SA .a/ jSB .b// D .ajSB .b/ /

for all a 2 A; b 2 B, and the reverse implication follows similarly. Thus, one
of these conditions may be omitted in the definition of a dual pairing of Hopf
-algebras.
Let us consider several examples of dual pairings.
Example 1.4.6. There exists a dual pairing O.SLn .C//  U.sln .C// ! C, de-
termined by .uij jX / WD uij .X / for all i; j D 1; : : : ; n and X 2 sln .C/. Here,
uij W Mn .C/ ! C denotes the .i; j /th coordinate function as in Example 1.2.6, and
the Lie algebra sln .C/ is identified with the space of n  n-matrices with vanishing
trace. A detailed discussion of this dual pairing for n D 2 can be found in [79,
Section V.7]; a generalization is given in Example 1.4.14.
Example 1.4.7. Let G be a compact connected Lie group with Lie algebra g. There
exists a dual pairing between the Hopf algebra of representative functions Rep.G/
and the universal enveloping algebra U.g/, which can be described as follows.
First, observe that the Hopf algebra Rep.G/ is contained in C 1 .G/. Indeed,
every continuous group homomorphism of Lie groups is automatically smooth [22,
Proposition I.3.12], and therefore every continuous finite-dimensional representa-
tion and every representative function of G is smooth.
Next, recall that every X 2 g determines a left-invariant vector field on G
and hence also a first-order differential operator DX W C 1 .G/ ! C 1 .G/. The
map X 7! DX extends to an injective unital algebra homomorphism D W U.g/ ,!
EndC .C 1 .G//, ! 7! D! ; explicitly, the differential operator D! associated to an
element ! D X1 : : : Xn 2 U.g/ is given by

.D! f /.y/ D .DX1 : : : DXn f /.y/

ˇ
@n ˇ
D ˇ f .y  exp.t1 X1 / : : : exp.tn Xn //;
@t1 : : : @tn ˇ t1 ;:::;tn D0

see [181, Section 3.4].

Denote by e 2 G the unit element. Then the bilinear map

.  j  / W U.g/  Rep.G/ ! C; .!jf / WD .D! f /.e/;

is a perfect dual pairing of Hopf algebras. Let us prove this assertion.

i) Denote by 1U and 1Rep the units and by U and Rep the counits of U.g/
and Rep.G/, respectively. Since the embedding D W U.g/ ,! EndC .C 1 .G// is
unital, .1U jf / D .D1U f /.e/ D f .e/ D Rep .f / for all f 2 Rep.G/. Moreover,
34 Chapter 1. Hopf algebras

.DX1 :::Xn /1Rep D 0 for all X1 ; : : : ; Xn 2 g, whence .!j1Rep / D .D! 1Rep /.e/
equals U .!/ for all ! 2 U.g/. P
ii) We show that .!1 !2 jf / D .!1 jf.1/ /.!2 jf.2/ / for all !1 ; !2 2 U.g/ and
f 2 Rep.G/. Since g and 1U generate U.g/ as an algebra, it suffices to prove
X
.X1 : : : Xn jf / D .X1 jf.1/ / : : : .Xn jf.n/ / (1.15)

for all X1 ; : : : ; Xn 2 g and f 2 Rep.G/. By definition,

ˇ
@n ˇ
.X1 : : : Xn jf / D ˇ f .exp.t1 X1 / : : : exp.tn Xn //:
@t1 : : : @tn ˇ t1 ;:::;tn D0

We insert the relation

X
f .x1 : : : xn / D ..n1/ .f //.x1 ; : : : ; xn / D f.1/ .x1 / : : : f.n/ .xn /

with xi WD exp.ti Xi / and find that .X1 : : : Xn jf / is equal to

ˇ X
@n ˇ
ˇ f.1/ .exp.t1 X1 // : : : f.n/ .exp.tn Xn //
ˇ
@t1 : : : @tn t1 ;:::;tn D0
X X
D .DX1 f.1/ /.e/ : : : .DXn f.n/ /.e/ D .X1 jf.1/ / : : : .Xn jf.n/ /:
P
iii) Let us show that .!jgh/ D .!.1/ jg/.!.2/ jh/ for all ! 2 U.g/ and g; h 2
Rep.G/. The Leibniz rule implies that for every X 2 g,

.Xjgh/ D .DX .gh//.e/ D .DX g/.e/  h.e/ C g.e/  .DX h/.e/

D .X jg/.1U jh/ C .1U jg/.X jh/:
P
By definition of the comultiplication on U.g/, X ˝1U C1U ˝X D X.1/ ˝X.2/ .
Hence the equation above can be rewritten as follows:
X
.X jgh/ D .X.1/ jg/.X.2/ jh/ for all X 2 g:

Using relation (1.15) with f D gh, or the Leibniz rule for higher order derivatives
of the product gh, it is easy to see that the equation above remains valid if we
replace X 2 g by an element ! D X1 : : : Xn 2 U.g/.
iv) For all X 2 g and f 2 Rep.G/,
ˇ
@ ˇˇ
.XjS.f // D ˇ f .exp.tX /1 /
@t tD0
ˇ
@ ˇˇ
D ˇ f .exp.tX // D .X jf / D .S.X /jf /:
@t tD0
 1.4. The duality of Hopf algebras 35

Using Proposition 1.3.12 and equation (1.15), it is easy to deduce that .!jS.f // D
.S.!/jf / for all ! 2 U.g/.
v) Let ! 2 U.g/, and assume that .!jf / D .D! f /.e/ D 0 for all f 2 Rep.G/.
We show that then ! D 0. Since the operator D! is left-invariant and the space
Rep.G/ is closed under left translation, it follows that D! .Rep.G// D 0. We show
that then also D! .C 1 .G// D 0, and this implies ! D 0. The space C 1 .G/ carries
a natural locally convex topology, and D! is continuous with respect to this topology
[32, Chapter XVII]. So, it suffices to prove that Rep.G/ is dense in C 1 .G/. The
group G acts continuously (in a suitable sense) on C 1 .G/ via left translations, and
by a generalization of the Peter–Weyl theorem [22, III, Theorem 5.7], the subspace
of C 1 .G/ spanned by finite-dimensional G-invariant subspaces is dense. Every
such subspace is contained in Rep.G/ and hence Rep.G/ is dense in C 1 .G/.
vi) Let f ¤ 0 be a representative function on G. We show that .!jf / ¤ 0
for some ! 2 U.g/. Denote by V  Rep.G/ the linear span of all right translates
of f , and by
W G ! Aut.V / the representation given by
.x/g WD g.  x/ for all
x 2 G and g 2 V . Note that V has finite dimension and that
is smooth. Recall
that the Lie algebra of Aut.V / is End.V /. Differentiating
at e 2 G, we obtain
a homomorphism of Lie algebras g ! End.V / which extends to a unital algebra
homomorphism D
W U.g/ ! End.V /. It is easy to check that D
.X /g D DX g
for every X 2 g and g 2 V . Choose some x 2 G such that f .x/ ¤ 0. Since G
is compact and connected, there exists an P X 2 g such that exp.X / D x [22, IV,
Theorem 2.2]. For each n 2 N, put !n WD nkD1 X k =kŠ 2 U.g/. Then

lim D
.!n / D exp.D
.X // D
.exp.X // D
.x/
n!1

in Hom.V / (with respect to every norm topology), and therefore

lim .!n jf / D lim .D

.!n /f /.e/ D .
.x/f /.e/ D f .x/ ¤ 0:
n!1 n!1

Consequently, .!n jf / ¤ 0 for some n 2 N.

1.4.3 The restricted dual of a Hopf algebra

For every unital algebra A, there exists a largest subspace Aı  A0 for which the
map A0 D .mA /0 W A0 ! .A ˝ A/0 defines a comultiplication. Elements of this
subspace can be characterized as follows:

Lemma 1.4.8. Let A be a unital algebra and f 2 A0 . Then the following conditions
are equivalent:

i) A0 .f / 2 A0 ˝ A0 ,
ii) ker f contains a left ideal of A that has finite codimension,
36 Chapter 1. Hopf algebras

iii) ker f contains a right ideal of A that has finite codimension,

iv) ker f contains an ideal of A that has finite codimension.
P
Proof. i) ) ii): Write A0 .f / D i gi ˝ hi with gi ; hi T 2 A0 , where the gi are
linearly independent and the hi are non-zero. Then J WD i ker hi is a left ideal
of A with finite codimension and contained in ker f :
• AJ  J : if b 2 A and c 2 J , then
X X
0D gi .ab/hi .c/ D f .abc/ D gi .a/hi .bc/ for all a 2 A;
i i

and the linear independence of the gi implies hi .bc/ D 0 for all i ;

• J has finite codimension inTA, because ker hi  A has codimension 1 for
each i and the intersection i ker hi is finite;
P
• f .J / D 0 since f .a/ D f .1A  a/ D i gi .1A /hi .a/ D 0 for each a 2 J .
i) ) iii): The proof is similar to the proof of the implication i) ) ii).
ii) ) iv): Let J  A be a left ideal of finite codimension that is contained in
ker f . Then the map
W A ! Homk .A=J / given by
.a/.b C J / WD ab C J
is an algebra homomorphism. Since A=J and hence also Homk .A=J / have finite
dimension, the kernel I WD ker
has finite codimension in A. Finally, the relation

.I / D 0 implies I  J  ker f .
iii) ) iv): Again, the proof is similar to the proof given above.
vi) ) i): Let I  A be an ideal of finite codimension that is contained in ker f .
Denote by
W A ! A=I the quotient map and by
0 W .A=I /0 ! A0 its transpose.
We show that A0 .f / belongs to the space

.
0 ˝
0 /..A=I /0 ˝ .A=I /0 /  A0 ˝ A0 :

Since A0 .f / vanishes on A ˝ I C I ˝ A, it defines an element

g 2 ..A ˝ A/=.A ˝ I C I ˝ A//0 Š ..A=I / ˝ .A=I //0 :

Since A=I has finite dimension, ..A=I /˝.A=I //0 is isomorphic to .A=I /0 ˝.A=I /0
and g can be considered as an element of .A=I /0 ˝ .A=I /0 . It is easy to see that
A0 .f / D .
0 ˝
0 /.g/, and this relation completes the proof. 
Definition 1.4.9. The restricted dual of a unital algebra A is the subspace Aı 
A0 consisting of all linear maps that satisfy the equivalent conditions given in
Lemma 1.4.8.
Proposition 1.4.10. i) Let A be a unital algebra. Then the space Aı satisfies
A0 .Aı /  Aı ˝Aı . Put Aı WD A0 jAı . Then .Aı ; Aı / is a counital coalgebra,
where the counit is given by f 7! f .1A /.
 1.4. The duality of Hopf algebras 37

ii) Let .A; A / be a unital bialgebra. Then Aı  A0 is a subalgebra and

.A ; Aı / is a counital bialgebra. If .A; A / is a Hopf algebra, then so is .Aı ; Aı /.
ı

iii) For every Hopf -algebra .A; A /, the formula f  .a/ WD f .S.a/ / defines
an involution on Aı that turns .Aı ; Aı / into a Hopf -algebra.

Proof.
P i) First, we show that A0 .Aı /  Aı ˝Aı . Let f 2 Aı and write A0 .f / D
0
i gi ˝ hi with gi ; hi 2 A , where the hi are linearly independent. By assumption
on the hi , we can choose for every j an element aj 2 A such that hi .aj / D ıi;j
for all i; j . Then
X X
gj .ab/ D gi .ab/hi .aj / D f .abaj / D gi .a/hi .baj / for all a; b 2 A;
i i
P
and consequently A0 .gj / D i gi ˝ hi .  aj / 2 A0 ˝ A0 and gj 2 Aı . Therefore,
A0 .f / 2 Aı ˝ A0 , and a similar argument shows that A0 .f / 2 A0 ˝ Aı . We
conclude that A0 .f / 2 Aı ˝ Aı for every f 2 Aı , so that A0 restricts to a map
Aı W Aı ! Aı ˝ Aı . This map is coassociative because the multiplication mA is
associative. The assertion concerning the counit is evident.
ii) Suppose that A is a unital bialgebra. Then Aı  A0 is a subalgebra because

A0 .fg/ D A0 .f /A0 .g/ 2 .A0 ˝ A0 /  .A0 ˝ A0 /  A0 ˝ A0 for all f; g 2 Aı :

The multiplication and comultiplication on Aı are compatible because the comul-

tiplication and multiplication on A are compatible.
Assume that .A; A / is a Hopf algebra. Denote by SAı WD .SA /0 jAı W Aı ! A0
the restriction of the transpose of SA . We claim that SAı .Aı /  Aı . By Proposi-
tion 1.3.12, we have for all f 2 Aı and a; b 2 A

.A0 .f ı SA //.a ˝ b/ D f .SA .ab//

D f .SA .b/SA .a//
D .A0 .f //.SA .b/ ˝ SA .a//;

and hence
X
A0 .f ı SA / D .f.2/ ı SA / ˝ .f.1/ ı SA / 2 A0 ˝ A0 :

The claim follows. Now we conclude as in Theorem 1.4.1 that .Aı ; Aı / is a Hopf
algebra with antipode SAı .
iii) A similar calculation as in ii) shows that for every f 2 Aı , the functional
f 2 A0 given by a 7! f .S.a/ / belongs to Aı . The compatibility of this involu-


tion with the Hopf algebra structure on Aı is verified as in Theorem 1.4.1 iv). 
38 Chapter 1. Hopf algebras

Dual pairings and restricted duals of Hopf algebras are related to each other as
follows:

Proposition 1.4.11. i) For every Hopf algebra .A; A /, the pairing A  Aı ! k

given by .a; f / 7! f .a/ is a dual pairing of Hopf algebras.
ii) If .  j  / W A  B ! k is a .perfect/ dual pairing of Hopf algebras, then the
induced maps A ! B 0 ; a 7! .aj  /, and B ! A0 ; b 7! .  jb/, restrict to .injective/
morphisms of Hopf algebras A ! B ı and B ! Aı .

Proof. Statement i) is obvious from the definitions, and statement ii) follows easily
from Remark 1.4.5 i) and Lemma 1.4.8. 

Example 1.4.12. Let G be a discrete group. We determine .kG/ı .

Evidently, .kG/0 can be identified with k.G/: each  2 .kG/0 corresponds to
the function z 7! .Uz /, and each function f 2 k.G/ corresponds to the linear
map Uz 7! f .z/. We claim that this identification induces an isomorphism

.kG/ı Š Repk .G/

of Hopf algebras, where Repk .G/ denotes the Hopf algebra of representative
k-valued functions on G, see the last remarks of Example 1.2.5. In fact, we only
prove that the spaces .kG/ı and Repk .G/ coincide; the compatibility of the structure
maps of .kG/ı and Repk .G/ follows from similar calculations as in Example 1.2.8
and Example 1.4.2.
Assume that f 2 k.G/ Š .kG/0 vanishes on an ideal I  kG that has finite
codimension. Then left multiplication defines a representation
of G on the finite-
dimensional vector space V WD .kG/=I , in formulas,
.x/.Uy C I / WD Uxy C I .
Denote by  2 V 0 the map given by Ux C I 7! f .x/. Then f .x/ D .Ux C I / D
.
.x/.Ue C I // for all x 2 G, so f 2 Repk .G/.
Conversely, assume that f 2 Repk .G/ has the form f .x/ D .
.x/v/,
where
is a representation of G on some finite-dimensional vector space V and
 2 V 0 , v 2 V . Then the functional in .kG/0 corresponding to f vanishes on the
kernel of the representation
Q W kG ! Homk .V /, Ux 7!
.x/, which is an ideal
in kG of codimension less than or equal to dim Homk .V /.

Remark 1.4.13. For every unital algebra A, the space Aı is the largest subspace of
A0 on which A0 defines a comultiplication, as can be seen from Lemma 1.4.8 i).
Unfortunately, the space Aı may be quite small – there exist groups G for which
.kG/ı Š Repk .G/ D k , see [1, Exercise 2.5] or [18, 2.7].

Example 1.4.14. If G is a semisimple connected simply connected affine algebraic

group over an algebraically closed field k of characteristic 0 and g is the Lie algebra
of G, then there exists a perfect dual pairing U.g/  O.G/ ! k that induces an
 1.4. The duality of Hopf algebras 39

isomorphism O.G/ Š U.g/ı , see [65, Theorem 3.1]. However, U.g/ 6Š O.G/ı –
for any affine algebraic group G over k,

U.g/ Š f 2 O.G/0 j .mn1 / D 0 for some n > 0g;

where m1 D ff 2 O.G/ j f .1/ D 0g [1, p. 198]. A concrete example of this

pairing is given in Example 1.4.6.
Chapter 2
Multiplier Hopf algebras and their duality

A multiplier Hopf algebra is a non-unital generalization of a Hopf algebra, where the

target of the comultiplication is no longer the twofold tensor product of the underly-
ing algebra, but an enlarged multiplier algebra. A motivation for this generalization,
the pertaining definitions, and the main example are presented in Section 2.1.
A remarkable feature of multiplier Hopf algebras is that they admit a nice duality
which extends the (incomplete) duality of Hopf algebras presented in Section 1.4.
This duality is based on left- and right-invariant linear functionals called integrals,
which are analogues of the Haar measures of a group. These integrals and their
modular properties are discussed in Section 2.2, and an account of the duality theory
is given in Section 2.3.
The theory of multiplier Hopf algebras was developed by Van Daele; all results
presented in this chapter are taken from the articles [174], [177].

2.1 Definition of multiplier Hopf algebras

Our primary examples of Hopf algebras were algebras of functions on suitable
groups, where the comultiplication, counit, and antipode were defined as the trans-
poses of the group multiplication, of the inclusion of the unit, and of the group
inversion, respectively, as in equation (1.4). Recall that the functions belonging to
such a Hopf algebra had to be chosen carefully and in accordance with the flavor
of the groups, as, for example,
• arbitrary functions if the groups are finite,
• representative functions if the groups are compact, or
• polynomial functions if the groups are affine algebraic.
Now assume that G is an infinite discrete group. The discussion in Section 1.2.2
showed that the algebra k.G/ of all functions on G is too large to carry the comul-
tiplication . A natural replacement would be the subalgebra kfin .G/  k.G/ of
functions with finite support. But unfortunately, the image .kfin .G//  k.G G/
is not contained in the subspace kfin .G/ ˝ kfin .G/ Š kfin .G  G/, so that  does
not restrict to a comultiplication on kfin .G/.
To solve this problem, we extend the notion of a morphism of algebras by
allowing a morphism from an algebra A to an algebra B to take values in an
enlarged multiplier algebra M.B/. In the situation above, the multiplier algebra
M.kfin .G/ ˝ kfin .G// will be equal to k.G  G/.
 2.1. Definition of multiplier Hopf algebras 41

2.1.1 Multipliers of algebras

The concept of a multiplier Hopf algebra is based on the notion of a multiplier of
an algebra:

Definition 2.1.1. Let A be an algebra. A left / right multiplier of A is a linear map

T W A ! A that satisfies T .ab/ D T .a/b / T .ab/ D aT .b/ for all a; b 2 A.
A multiplier of A is a pair .Tl ; Tr / consisting of a left multiplier Tl and a right
multiplier Tr satisfying bTl .a/ D Tr .b/a for all a; b 2 A. The set of all multi-
pliers of A is denoted by M.A/. It is an algebra with respect to the multiplication
.Tl ; Tr /  .Sl ; Sr / WD .Tl ı Sl ; Sr ı Tr /, as one can easily check.

Remark 2.1.2. For every commutative algebra A, a map T W A ! A is a left

multiplier if and only if it is a right multiplier.

When working with multipliers, it is convenient to restrict to algebras and ho-

momorphisms that are non-degenerate in the following sense:

Definition 2.1.3. i) An algebra A is non-degenerate if

(a) for every a 2 A, a ¤ 0, we have Aa ¤ 0 and aA ¤ 0, and

(b) the linear span of AA is equal to A.

ii) Let A and B be non-degenerate algebras. A homomorphism  W A ! M.B/

is non-degenerate if the linear span of .A/B and the linear span of B.A/ both
are equal to B.

We shall frequently use the following properties of multipliers:

Proposition 2.1.4. Let A and B be non-degenerate algebras.

i) For every element a 2 A, the maps b 7! ab and b 7! ba define a multiplier,

and the induced map A ! M.A/ embeds A as an ideal in M.A/.

From now on, we consider A as an ideal of M.A/ as in i).

ii) aT D Tr .a/ and T a D Tl .a/ for all a 2 A and T D .Tl ; Tr / 2 M.A/.

iii) If A is unital, then T 1A D T D 1A T 2 A for each T 2 M.A/, and
M.A/ D A.
iv) Let A be a -algebra. For every T 2 M.A/, the formulas T  a WD .a T /
and aT  WD .T a / ; a 2 A, define a multiplier T  2 M.A/. The involution
T 7! T  turns M.A/ into a -algebra.
v) The tensor product A˝B is non-degenerate and one has a natural embedding
M.A/ ˝ M.B/ ,! M.A ˝ B/.
42 Chapter 2. Multiplier Hopf algebras and their duality

vi) Every non-degenerate homomorphism  W A ! M.B/ extends uniquely to

a homomorphism M.A/ ! M.B/. If A and B are -algebras and  is a
-homomorphism, then the extension is a -homomorphism again.
Proof. Statements i)–iv) are easy to check. P
v) Consider an element x 2 A ˝ B. Write x D i ai ˝ bi , where the bi are
linearly independent. Assume that x.c ˝ d / D 0 for all c 2 A and d 2 B. Then
X
.f ˝ id/.x.c ˝ d // D f .ai c/bi d D 0 for all c 2 A; d 2 B; f 2 A0 ;
i
P
and using the fact that B is non-degenerate, we conclude that i f .ai c/bi D 0 for
all c 2 A, d 2 B, f 2 A0 . Since the bi are linearly independent, we must have
f .ai c/ D 0 for all i and all c 2 A, f 2 A0 . But A is non-degenerate, so ai D 0
for all i and hence x D 0.
A similar argument shows that x D 0 if .c ˝ d /x D 0 for all c 2 A and d 2 B.
Finally, it is clear that .A ˝ B/.A ˝ B/ is linearly dense in A ˝ B.
vi) Let T 2 M.A/. Every extension Q W M.A/ ! M.B/ of  must satisfy
X X X X
Q /.ai /bi D
.T .T ai /bi and Q /D
bi .ai /.T bi .ai T /
i i i i
(2.1)
P
for all ai 2 A and bi 2 B. If i .ai /bi D 0, then
X X X
.c/ .T ai /bi D .cT ai /bi D .cT / .ai /bi D 0 for all c 2 A;
i i i
P
and since B and  are non-degenerate, i .T ai /bi D 0. This and a similar
Q / 2 M.B/. It is easy to check
argument show that (2.1) defines a multiplier .T
Q / is a homomorphism and that Q is a -homomorphism
that the assignment T 7! .T
if  is a -homomorphism. 

2.1.2 Multiplier bialgebras

Motivated by the discussion in the introduction to this section, we generalize the
notion of a bialgebra as follows:
Definition 2.1.5. A multiplier bialgebra is a non-degenerate algebra A equipped
with a non-degenerate homomorphism  W A ! M.A ˝ A/ such that
i) the following subsets of M.A ˝ A/ are contained in A ˝ A  M.A ˝ A/:
.A/.1 ˝ A/; .A/.A ˝ 1/; .A ˝ 1/.A/; .1 ˝ A/.A/I
 2.1. Definition of multiplier Hopf algebras 43

ii)  is coassociative in the sense that the following diagram commutes:

A
 / M.A ˝ A/

 id ˝
 
M.A ˝ A/ / M.A ˝ A ˝ A/.
˝id

If A is a -algebra and  a -homomorphism, then .A; / is called a multiplier

-bialgebra.
A morphism of multiplier (-)bialgebras .A; A / and .B; B / is a non-degen-
erate (-)homomorphism F W A ! M.B/ that satisfies B ı F D .F ˝ F / ı A ,
that is, the following square commutes:

A
F / M.B/

A B
 
M.A ˝ A/ / M.B ˝ B/.
F ˝F

Remarks 2.1.6. i) In condition i) above, the symbol 1 denotes the unit in M.A/, and
the spaces A˝1 and 1˝A are considered as subsets of M.A/˝M.A/  M.A˝A/.
ii) In the diagrams above, the homomorphisms  ˝ id; id ˝ W A ˝ A !
A ˝ A ˝ A and F ˝ F W A ˝ A ! B ˝ B have been extended to the respective
multiplier algebras.
iii) Given multiplier bialgebras .A; A / and .B; B /, we can construct new
multiplier bialgebras .A; A /op , .A; A /cop , .A; A /op;cop and equip A ˚ B and
A ˝ B with the structure of multiplier bialgebras in a similar way as in the case of
bialgebras, compare Remark 1.3.7 iii).
The Sweedler notation 1.3.3 can be extended to multiplier bialgebras, but then it
has to be used with much more care. Although we shall rarely use it, we explain this
notation in detail because it illustrates the difference between multiplier bialgebras
and ordinary bialgebras.
Notation 2.1.7. Let .A; / be a multiplier bialgebra and aP2 A. In general, the
multiplier .a/ 2 M.A ˝ A/ can not be written as a sum i a1;i ˝ a2;i , where
a1;i ; a2;i 2 M.A/. Nevertheless, we write this multiplier as a formal sum
X
.a/ DW a.1/ ˝ a.2/ :

As for ordinary bialgebras, we extend this notation to iterated applications of the co-
multiplication as follows. Since  is non-degenerate, we can define non-degenerate
44 Chapter 2. Multiplier Hopf algebras and their duality

homomorphisms .n/ W A ! M.A˝nC1 / by

.0/ WD idA and .nC1/ WD ..n/ ˝ idA / ı  for n 0:

For each n 2 N, we write the multiplier .n/ .a/ 2 M.A˝nC1 / as a formal sum
X
.n/ .a/ DW a.1/ ˝    ˝ a.nC1/ :

We think of the multiplier .n/ .a/ as having nC1 legs which are represented by the
symbols a.1/ ; : : : ; a.nC1/ , and treat the formal sums introduced above like ordinary
sums of elementary tensors in a tensor product of algebras. Thus we write
X X
.a/.b ˝ 1/ D a.1/ b ˝ a.2/ ; .b ˝ 1/.a/ D ba.1/ ˝ a.2/ ;
X X
.a/.1 ˝ b/ D a.1/ ˝ a.2/ b; .1 ˝ b/.a/ D a.1/ ˝ ba.2/

for all a; b 2 A. The definition of a multiplier bialgebra implies that these four
expressions belong to A ˝ A; thus, the formal sums above stand for finite sums of
elementary tensors in A ˝ A. We say that in the products in the first line above, the
first leg a.1/ of .a/ is covered by b, whereas in the products in the second line,
the second leg a.2/ of .a/ is covered by b.
P
More generally, every product of .n/ .a/ D a.1/ ˝  ˝a.nC1/ with elements
of A, where at least n legs of .n/ .a/ are covered by elements of A, belongs to
A˝nC1 and is equal to a finite sum of elementary tensors in A˝nC1 .
A drawback of the Sweedler notation for multiplier bialgebras
P is that it may
be difficult to see whether certain products of formal sums a.1/ ˝    ˝ a.nC1/
belong to an ordinary tensor product or not.
Example 2.1.8. Let .A; / be multiplier bialgebra and a; b; c; d 2 A. The sum
X
b.1/ a.1/ ˝ ca.2/ b.2/ ˝ a.3/ d ˝ b.3/
P
P element of A ˝ A ˝ A:
is a well-defined PFirst, note that a.1/ ˝ ca.2/ 2 A ˝ A.
Therefore, a.1/ ˝ ca.2/ ˝ aP .3/ d D a.1/ ˝ .ca.2/ /.1 ˝ d / 2 A ˝ A ˝ A,
and hence the first two legs of b.1/ ˝ b.2/ ˝ b.3/ are covered by elements of A.

2.1.3 Multiplier Hopf algebras

Multiplier Hopf algebras form a special class of multiplier bialgebras:
Definition 2.1.9. A multiplier (-)bialgebra .A; / is a multiplier Hopf (-)algebra
if the linear maps T1 ; T2 W A ˝ A ! A ˝ A given by
T1 .a ˝ b/ D .a/.1 ˝ b/ and T2 .a ˝ b/ D .a ˝ 1/.b/
 2.1. Definition of multiplier Hopf algebras 45

are bijective. A multiplier Hopf algebra .A; / is regular if the multiplier bialgebras
.A; /op and .A; /cop are multiplier Hopf algebras.
A morphism of multiplier Hopf (-)algebras is simply a morphism of the un-
derlying multiplier (-)bialgebras.

This definition is similar to the characterization of Hopf algebras among bial-

gebras given in Section 1.3.4.
Before we consider examples of multiplier Hopf algebras and additional struc-
ture maps like the counit and antipode, let us list some easy observations:

Remarks 2.1.10. i) Every Hopf algebra is a multiplier Hopf algebra by Theo-

rem 1.3.18. In particular, the notion of a regular Hopf algebra is well defined,
and by Proposition 1.3.14, a Hopf algebra is regular if and only if its antipode is
bijective. We shall see an analogue of this statement for multiplier Hopf algebras
in Proposition 2.1.12 iii).
ii) If .A; / is a multiplier Hopf algebra, then also .A; /op;cop is a multiplier
Hopf algebra, and .A; /op is a multiplier Hopf algebra if and only if .A; /cop is
a multiplier Hopf algebra.
iii) Every multiplier Hopf -algebra is regular.

Example 2.1.11. Suppose that G is a discrete group. Then M.kfin .G/˝kfin .G// Š
k.G  G/, and the algebra kfin .G/ equipped with the comultiplication

 W kfin .G/ ! k.G  G/ Š M.kfin .G/ ˝ kfin .G//; .f /.x; y/ D f .xy/;

is a regular multiplier Hopf algebra. Let us explain this in some more detail.
The natural embedding k.G/ ˝ k.G/ ,! k.G  G/ identifies kfin .G/ ˝ kfin .G/
with kfin .GG/, and it is easy to see that M.kfin .GG// is isomorphic to k.GG/.
The isomorphism M.kfin .G/ ˝ kfin .G// Š k.G  G/ thus obtained identifies a
multiplier T with the function fT defined by

fT .x; y/ D  , T .ıx ˝ ıy / D .ıx ˝ ıy / for all x; y 2 G:

The map  is evidently coassociative, and the maps T1 and T2 considered

in Definition 2.1.9 are bijective because they are the transposes of the bijections
ˆ1 ; ˆ2 W G  G ! G  G considered in Proposition 1.3.19.
If k D C, then kfin .G/ is a multiplier Hopf -algebra with respect to pointwise
conjugation of functions.

Multiplier Hopf algebras are similar to Hopf algebras in many respects. For
example, they also possess a counit and an antipode:

Proposition 2.1.12. Let .A; / be a multiplier Hopf algebra.

46 Chapter 2. Multiplier Hopf algebras and their duality

i) There exists a unique non-degenerate homomorphism W A ! k, called the

counit of .A; /, such that the following diagram commutes:

M.A ˝ A/ o / M.A ˝ A/
 
A
_
˝id id ˝ (2.2)
  
M.k ˝ A/ / M.A/ o M.A ˝ k/.
Š Š

If .A; / is a multiplier Hopf -algebra, then is a -homomorphism.

ii) There exists a unique linear map S W A ! M.A/, called the antipode of
.A; /, such that

mM.A/ ..S ˝ idA /..a/.1 ˝ b/// D .a/b;

(2.3)
mM.A/ ..idA ˝S /..a ˝ 1/.b/// D a .b/ for all a; b 2 A:

Here, mM.A/ denotes the multiplication in M.A/. Moreover, S is a morphism

of multiplier Hopf algebras .A; / ! .A; /op;cop .
iii) A multiplier Hopf algebra .A; / is regular if and only if the antipode S is a
Š
linear isomorphism A 
! A.
iv) If .A; / is a multiplier Hopf -algebra, then S ı  ı S ı  D idA .
v) Let F W A ! M.B/ be a morphism of multiplier Hopf algebras. Then SB ı
F D F ı SA .

Proof. The construction of and S proceeds along the same lines as in the proof
of Proposition 1.3.22 and differs mainly in more involved notation; for details, see
[174, Sections 3–5]. The remaining assertions about and S follow similarly as
for Hopf algebras; for details, see [174, Proposition 5.2] and [177, Proposition 2.8].


Remark 2.1.13. In extended Sweedler notation, equation (2.3) takes the form
X X
S.a.1/ /a.2/ b D .a/b; ab.1/ S.b.2/ / D a .b/ for all a; b 2 A:

Theorem 2.1.14. A multiplier bialgebra .A; / is a regular multiplier Hopf algebra

if and only if there exist a homomorphism W A ! k and a bijective linear map
S W A ! A such that diagram (2.2) commutes and equation (2.3) holds.

Proof. This follows from the previous proposition and a similar argument as in
Lemma 1.3.21, see also [177, Proposition 2.9]. 
 2.2. Integrals and their modular properties 47

2.2 Integrals and their modular properties

The concept of an integral on a multiplier Hopf algebra is the natural analogue
of the Haar measure of a locally compact group and fundamental to the duality
theory of multiplier Hopf algebras presented in the next section. In contrast to
locally compact groups, which always have an invariant measure, multiplier Hopf
algebras need not possess integrals. Uniqueness, however, holds under very mild
assumptions.
This section is organized as follows. First, we introduce the concept of an
integral, give some examples, and list some easy properties that can be found in
most books on Hopf algebras, as, for example, in [1], [80], [111], [145]. Next, we
show that integrals are unique and faithful, closely following the original article
[177] of Van Daele. Like the Haar measure of a locally compact group, every
integral on a multiplier Hopf algebra enjoys several modular properties. These are
studied in the last sections; again, we follow [177].

2.2.1 The concept of an integral

To motivate the definition of left and right integrals, let us reformulate the concept
of a Haar measure in terms of Hopf algebras.
Let G be a locally compact group with left Haar measure  and let A  C.G/
be some Hopf algebra of functions on G with comultiplication, counit, and antipode
as in equation (1.4) on page 6. If the functions in A are integrable, that is, if A is
contained in L1 .G; /, the Haar measure  defines a linear map
Z
 W A ! C; f 7! f .y/d .y/: (2.4)
G

Left-invariance of  amounts to the fact that for each function f 2 A, the function
F on G defined by
Z
F .x/ WD f .xy/d .y/ for all x 2 G
G

satisfies F .x/ D .f / for all x 2 G. We replace the multiplication of G by the

comultiplication of A, using the relation
X
f .xy/ D ..f //.x; y/ D f.1/ .x/f.2/ .y/ for all x; y 2 G;

and obtain
X Z 
F D f.1/ f.2/ .y/d .y/ D .id ˝/..f //:
G
48 Chapter 2. Multiplier Hopf algebras and their duality

Thus, the invariance condition F .x/ D .f /, x 2 G, takes the form

.id ˝/..f // D .f /1A :

Now, let .A; / be an arbitrary multiplier Hopf algebra. Given a linear map
 W A ! k and an element a 2 A, we define .id ˝/..a// 2 M.A/ by
X
..id ˝/..a///b WD .id ˝/..a/.b ˝ 1// D a.1/ b.a.2/ /;
X
b..id ˝/..a/// WD .id ˝/..b ˝ 1/.a// D ba.1/ .a.2/ /:

Similarly we define . ˝ id/..a// 2 M.A/.

Definition 2.2.1. Let .A; / be a multiplier Hopf algebra. A linear map  W A ! k

is

• left-invariant if .id ˝/..a// D .a/1M.A/ for all a 2 A, and

• right-invariant if . ˝ id/..a// D .a/1M.A/ for all a 2 A.

If  is non-zero and left- or right-invariant, we call it a left or right integral on

.A; /, respectively. A map that is a left and a right integral is briefly called an
integral.
We call .A; / a multiplier Hopf algebra with integrals if there exist a left and
a right integral on .A; /, and unimodular if every left integral on .A; / is a right
integral and vice versa.

Recall that a linear map  W A ! k on an algebra A is called

• faithful if .aA/ ¤ 0 and .Aa/ ¤ 0 for every non-zero a 2 A,

• positive if A is a -algebra and .a a/ 0 for all a 2 A, and

• normalized if A is unital and .1A / D 1.

We call a positive normalized functional on a unital -algebra a state.

Remarks 2.2.2. i) Let .A; / be a multiplier Hopf algebra. It is easy to see that a
functional  2 A0 is left-invariant if and only if for all f 2 A0 and a; b 2 A,

.f ˝ /..a/.b ˝ 1// D f .b/.a/; .f ˝ /..b ˝ 1/.a// D f .b/.a/:

Likewise,  is right-invariant if and only if for all f 2 A0 and a; b 2 A,

. ˝ f /..a/.1 ˝ b// D .a/f .b/; . ˝ f /..1 ˝ b/.a// D .a/f .b/:

 2.2. Integrals and their modular properties 49

ii) For a Hopf algebra .A; /, the invariance of linear maps can be characterized
in terms of the convolution product defined in Section 1.3.2: Let  2 A0 . Then

 is left-invariant , f   D f .1A / for all f 2 A0 ;

 is right-invariant ,   f D f .1A / for all f 2 A0 :

This follows from the relations

f ..id ˝/..a/// D .f  /.a/ and f .. ˝ id/..a/// D .  f /.a/;

which hold for all a 2 A.

iii) If a Hopf algebra has a left integral, then its antipode is bijective [29, Theo-
rem 5.2.3, Corollary 5.4.6], [145, Corollary 5.1.7].
iv) Let .A; / be a multiplier Hopf -algebra and  a non-zero positive func-
tional on A.
(a)  is -linear: Since A is non-degenerate, it suffices to show that .a b/ D
.b  a/ for all a; b 2 A. Now the inequalities ..a C b/ .a C b// 0,
.a a/ 0, .b  b/ 0 imply =.a b/ C =.b  a/ D 0, and replacing b
by ib, we find <.a b/  <.b  a/ D 0.
(b) If A is unital, then there exists a real number r > 0 such that r is normal-
ized: The Cauchy–Schwarz inequality implies j.a/j2  .1/.a a/ for all
a 2 A, whence .1/ > 0.
(c)  is faithful if and only if .a a/ > 0 for all non-zero a 2 A: The Cauchy–
Schwarz inequality implies that for each a 2 A, the functional .a  / or
.  a/ vanishes on A if and only if .aa / D 0 or .a a/ D 0, respectively.
Examples 2.2.3. Let us consider the Hopf algebras introduced in Section 1.2.2:
i) Let G be a compact group with normalized
R Haar measure . Then the map
 W Rep.G/ ! C given by f 7! G f d  is a positive faithful normalized
integral.
ii) Let G be a discrete group. Then the map  W kfin .G/ ! k given by .ıx / D 1
for all x 2 G is a faithful integral; it corresponds to integration with respect
to the counting measure on G. For k D C, this integral is positive.
iii) Let G be a discrete group. Then the map  W kG ! k given by .Ue / D 1
and .Ux / D 0 for x 2 G; x ¤ e; is a faithful normalized integral. For
k D C, this integral is positive.
iv) Consider the additive group kn , where n 2 N and k 2 fR; Cg. The Hopf
algebra O.kn / does not contain any non-zero function that is integrable with
50 Chapter 2. Multiplier Hopf algebras and their duality

respect to the Haar measure on kn , and therefore formula (2.4) does not define
an integral on O.kn /. In fact, the Hopf algebra O.kn / does not possess any
left or right integral. We prove this assertion for n D 1, the general case can
be treated similarly. Assume that  W O.k/ ! k is left-invariant. For every
m 2 N,
!
Xm
m
.X m / D X mk ˝ X k
k
kD0
and !
X
m
m
1O.k/  .X m / D .id ˝/..X m // D X mk .X k /:
k
kD0

Thus .X k / D 0 for all k < m and m 2 N, that is,  D 0. A similar

argument shows that every right-invariant functional vanishes.
v) Consider the multiplicative group kn , where n 2 N and k 2 fR; Cg. Again,
the Haar measure on kn does not yield a left or right integral on the Hopf
algebra O.kn / because the only integrable function in O.kn / is 0.
Nevertheless, the Hopf algebra O.kn / does possess a normalized integral 
for every field k, given by 1O.kn / 7! 1 and X1k1 : : : Xnkn 7! 0 for all
.k1 ; : : : ; kn / 2 Zn n f0g. With respect to the natural isomorphism O.kn / Š
kZn (see the end of Example 1.2.8), this map corresponds to the integral
described in ii). For k D C, we have an isomorphism O.Cn / Š Rep.T n /
(see the end of Example 1.2.8), and the map  corresponds to integration over
T n with respect to the Haar measure as in i).
The following result is fundamental:
Proposition 2.2.4. Every left integral and every right integral on a regular multi-
plier Hopf algebra is faithful.
Proof. To simplify the presentation, we restrict ourselves to the case of a Hopf
algebra; see [177, Proposition 3.4] for the general case. So, let  be a left integral
on a regular Hopf algebra .A; / and a 2 A. Assume that .Aa/ D 0. Then

0 D c.ba/ D .id ˝/..c ˝ 1/.b/.a// for all c; b 2 A:

Since the map T2 W A ˝ A ! A ˝ A; c ˝ d ! .c ˝ 1/.d /, is surjective,

P replace .c ˝ 1/.b/ by an arbitrary element d ˝ e 2 A ˝ A, and find
we can
0 D da.1/ .ea.2/ / for all d; e 2 A. We put d D 1A , apply , and get
X
0D a.1/ ˝ a.2/ .ea.3/ / for all e 2 A:
 2.2. Integrals and their modular properties 51
P P
Write a.1/ ˝ a.2/ ˝ a.3/ D i a1;i ˝ a2;i ˝ a3;i , choosing the elements a2;i
linearly independent. Then 0 D a1;i .ea3;i / for all e 2 A and all i . Replacing e
by f S.a2;i / for each i, where f 2 A is fixed but arbitrary, and summing again, we
obtain
X X
0D a1;i .f S.a2;i /a3;i / D a.1/ .f .a.2/ // D a.f / for all f 2 A:
i

This shows a D 0. A modification of this argument shows that .aA/ D 0 implies

a D 0. The proof of the assertion concerning right integrals is similar. 

Corollary 2.2.5. Every left or right integral  on a multiplier Hopf -algebra

.A; / satisfies .a a/ > 0 for all a 2 A; a ¤ 0.

Proof. Combine Proposition 2.2.4 with Remark 2.2.2 iv) (c). 

Before we consider the question of uniqueness and existence of left and right
integrals, let us clarify the relation between these two notions. Recall that for
a locally compact group, the inversion of the group interchanges the left with the
right Haar measure, and that for a compact group, the left and the right Haar measure
coincide. Similar results hold for multiplier Hopf algebras:

Proposition 2.2.6. i) Let .A; / be a regular multiplier Hopf algebra with a

left/right integral . Then  ı S and  ı S 1 are right/left integrals on .A; /.
ii) On every Hopf algebra, there exists at most one normalized left/right inte-
gral , and this  is simultaneously a right/left integral and satisfies  D  ı S D
 ı S 1 .

Proof. i) Let  be a left integral on .A; /. By Proposition 2.1.12 iii), the antipode
S is a linear isomorphism of A, whence  ı S is non-zero. Furthermore,

S.. ı S ˝ id/..a/.1 ˝ b/// D . ˝ id/..S ˝ S /..a/.1 ˝ b///

D .id ˝/..S.b/ ˝ 1/.S.a/// D S.b/.S.a//

for all a; b 2 A. An application of S 1 shows . ı S ˝ id/..a// D .S.a//, so

 ı S is right-invariant. The proofs of the other assertions are similar.
ii) Let .A; / be a Hopf algebra with a normalized left integral . By Re-
mark 2.2.2 iii) and Proposition 2.1.12 iii), .A; / is regular, so WD  ı S is a
Q we have Q D because
right integral. For every normalized left integral ,

Q
.a/ D Q
.1A .a// D. Q
˝ /..a// Q .a/1A / D
D . .a/; a 2 A:

The proofs of the other assertions are similar. 

52 Chapter 2. Multiplier Hopf algebras and their duality

For left and for right integrals on multiplier Hopf -algebras, it is natural to de-
mand positivity. Unfortunately, the correspondence between left and right integrals
established above need not preserve positivity. Nevertheless, the following result
holds:

Proposition 2.2.7 ([94, Theorem 9.9]). A multiplier Hopf -algebra has a positive
left integral if and only if it has a positive right integral.

Comments on the proof. This result seems to be accessible only via operator theory
and a fair amount of work; therefore we do not include the proof. Roughly, one
proceeds as follows. Given a multiplier Hopf -algebra .A; / with positive left
integral , one constructs a GNS-representation .H; ƒ;
/ for .A; / and considers
the unbounded conjugate-linear map G on H given by ƒ.a/ 7! ƒ.S.a/ /. Let
G D IM 1=2 be the polar decomposition of G. Then I is an antiunitary involutive
map and one can show that for each a 2 A, the operator I
.a /I is contained in

.A/. Therefore, one can define a map R W A ! A, called the unitary antipode of
.A; /, by the formula
.R.a// D I
.a /I . Finally, the map  ı R turns out to
be a positive right-invariant functional on .A; /.

Definition 2.2.8. An algebraic quantum group is a multiplier Hopf -algebra with

a positive left integral and a positive right integral.

2.2.2 Existence and uniqueness

The existence of integrals on multiplier Hopf algebras can be shown under special
assumptions only, for example

• every finite-dimensional Hopf algebra has a left integral and a right integral
[177, Proposition 5.1];

• a Hopf algebra has a normalized integral if and only if it is cosemisimple,

which, roughly, means that the dual algebra A0 is semisimple [29, Exer-
cise 5.5.9], [80, Section 11.2], [145, Chapter XIV].

A comprehensive characterization of all Hopf algebras that possess left integrals can
be found in [29, Section 5.3]. Hopf -algebras with positive integrals, also called
algebraic compact quantum groups, are studied in Chapter 3, and further examples
of such Hopf -algebras are given in Chapter 6.
In contrast to existence, uniqueness of integrals already holds for all regular mul-
tiplier Hopf algebras. This class includes all multiplier Hopf -algebras and all Hopf
algebras that have a left integral, see Remark 2.2.2 iii) and Proposition 2.1.12 iii).
The first main step towards the proof of uniqueness is the following result which
will be of individual interest later on.
 2.2. Integrals and their modular properties 53

Proposition 2.2.9. Let .A; / be a regular multiplier Hopf algebra with a left
integral  and a right integral . Then
f .  a/ j a 2 Ag D f.  a/ j a 2 Ag and f .a  / j a 2 Ag D f.a  / j a 2 Ag:
In particular, these spaces do not depend on the choice of  and .
Proof. We show that for each a 2 A there exists some c 2 A such that .  a/ D
.  c/; the rest of the proof is similar and uses the regularity of .A; /. So, let
a 2 A and choose b 2 A such that .b/ D 1. We apply Remark 2.2.2 i) to  and
f WD , and find
.xa/ D .b/.xa/ D . ˝ /..xa/.b ˝ 1// for all x 2 A:
Since the map T1 W A ˝P A ! A ˝ A; c ˝ d 7! .c/.1 ˝ d /, is surjective, we can
write .a/.b ˝ 1/ D i .ci /.1 ˝ di / with ci ; di 2 A. Using Remark 2.2.2 i)
again, but this time for and f WD , we find
X
.xa/ D . ˝ /..x/.ci /.1 ˝ di //
i
X X
D . .xci /di / D .xci .di //
i i
P
for all x 2 A. Thus .  a/ D .  c/, where c WD i ci .di /. 
Theorem 2.2.10. Let .A; / be a regular multiplier Hopf algebra with integrals.
Then the space of all left-invariant functionals and the space of all right-invariant
functionals on A both have dimension 1.
Proof. Let 1 and 2 be two left integrals and a right integral on .A; /. Choose
a; b 2 A such that .ab/ D 1. Let x 2 A. Since the map T2 W A ˝ A !
P˝ A; c ˝ y ! .c ˝ 1/.y/, is surjective, we can choose ci ; yi 2 A such that
A
i .ci ˝ 1/.yi / D .1 ˝ x/.a/. Then

1 .x/ D .ab/1 .x/ D . ˝ 1 /..1 ˝ x/.a/.b//

X 
D. ˝ 1 / .ci ˝ 1/.yi /.b/ (2.5)
i
X
D .ci /1 .yi b/:
i

By Proposition 2.2.9, there exists some d 2 A such that 1 .  b/ D 2 .  d /. Re-

versing the transformations in (2.5), we obtain
X
1 .x/ D .ci /2 .yi d / D .ad /2 .x/ for all x 2 A: 
i
54 Chapter 2. Multiplier Hopf algebras and their duality

Remark 2.2.11. Let .A; / be a multiplier Hopf -algebra with a positive left
integral . Then there exists a number z 2 C with jzj D 1 such that the right
integral z ı S is positive. This follows immediately from Propositions 2.2.6, 2.2.7
and Theorem 2.2.10.

2.2.3 The modular element of an integral

An integral on a multiplier Hopf algebra has similar modular properties like the Haar
measure of a locally compact group: every left integral  is related to the associated
right integral  ı S via a modular multiplier which is an analogue of the modular
function of a locally compact group. Let us recall this modular function. Consider
a locally compact group G with left Haar measure  and denote by i W G ! G
the inversion x 7! x 1 . Then the image 1 WD i ./ of the left Haar measure
under i is a right-invariant measure, and there exists a strictly positive continuous
function ıG on G – called the modular function of G – such that
Z Z Z
1 1
f .y/d.y/ D f .y /ıG .y /d .y/ D f .z/ıG .z/d 1 .z/ (2.6)
G G G

for all positive measurable functions f on G. The function ıG is a homomorphism

from G to the multiplicative group of positive real numbers, that is,

ıG .xy/ D ıG .x/ıG .y/; ıG .e/ D 1; ıG .x 1 / D ıG .x/1 for all x; y 2 GI

(2.7)

here, e 2 G denotes the unit as usual.

Let us translate equation (2.6) into the language of integrals on Hopf algebras.
Suppose that A  L1 .G/ is a multiplier Hopf algebra with structure maps as in
equation (1.4) on page 6 and that  is the left integral on .A; / given by integration
with respect to  as in equation (2.4) on page 47. Then equation (2.6) takes the
form

.f / D . ı S /.f ıG / for all f 2 A:

Here, we assume for the moment that ıG belongs to M.A/. The relations in (2.7)
1
are equivalent to the relations .ıG / D ıG ˝ ıG ; .ıG / D 1, and S.ıG / D ıG .
Every multiplier Hopf algebra with integrals has a modular element ı that be-
haves very much like ıG . There is, however, one small change:
1
The conventional choice of ı corresponds to ıG instead of ıG .

The first step towards the construction of ı is the following lemma, which is
also of individual interest:
 2.2. Integrals and their modular properties 55

Lemma 2.2.12. Let .A; / be a regular multiplier Hopf algebra with a left inte-
gral  and a right integral . Then for all a; b; x 2 A,
. ˝ S /..x ˝ 1/.a// D . ˝ id/..x/.a ˝ 1//
and
.S ˝ /..b/.1 ˝ x// D .id ˝/..1 ˝ b/.x//:
In Sweedler notation, these formulas can be rewritten as follows:
X X
.xa.1/ /S.a.2/ / D .x.1/ a/x.2/ ;
X X
S.b.1/ /.b.2/ x/ D x.1/ .bx.2/ /:

Proof. We only prove the first equation, the second one follows similarly. To
simplify notation, we assume that .A; / is a Hopf algebra; for the general case,
see [177, Proof of Proposition 3.11]. Let x; a 2 A. Since is right-invariant,
X X
.xa.1/ /S.a.2/ / D .x.1/ a.1/ /x.2/ a.2/ S.a.3/ /
X X
D .x.1/ a.1/ /x.2/ .a.2/ / D .x.1/ a/x.2/ : 

Proposition 2.2.13. Let .A; / be a regular multiplier Hopf algebra with a left
integral .
i) There exists an invertible multiplier ı 2 M.A/ such that
. ˝ id/..a// D .a/ı for all a 2 A: (2.8)
ii) The extensions of ; , and S to M.A/ act on ı as follows:
.ı/ D ı ˝ ı; .ı/ D 1; S.ı/ D ı 1 : (2.9)
iii) .S.a// D .aı/ for all a 2 A.
Proof. To simplify notation, we assume that .A; / is a Hopf algebra; the general
case is treated similarly, see [177, Propositions 3.8–3.10].
i) It suffices to show that the map . ˝ id/ ı  factorizes through . It fol-
lows from Remark 2.2.2 ii) that for every ! 2 A0 , the convolution product   !
is left-invariant, and by Theorem 2.2.10, this product vanishes on ker . Now
. ˝ id/..ker // D 0 results from the equation
!.. ˝ id/..ker /// D .  !/.ker / D 0 for all ! 2 A0 :
P
ii) We apply  and to the equation .a/ı D .a.1/ /a.2/ and obtain
X X
..a/ı/ D .a.1/ /a.2/ ˝ a.3/ D .a.1/ /ı ˝ a.2/ D ı ˝ .a/ı;
X
..a/ı/ D .a.1/ / .a.2/ / D .a/
56 Chapter 2. Multiplier Hopf algebras and their duality

for all a 2 A. Consequently, .ı/ D ı ˝ ı and .ı/ D 1. Furthermore,

S.ı/ı D m..S ˝ id/..ı/// D .ı/1A D 1A

and similarly ıS.ı/ D 1A ; in particular, ı is invertible.

iii) By right-invariance of  ı S and by the second equation in Lemma 2.2.12,
X X
.S.a//.b/ D .S.a.1/ //.a.2/ b/ D .b.1/ /.ab.2/ / D .b/.aı/

for all a; b 2 A. 

Remarks 2.2.14. Let .A; / and , ı be as above.

i) If is a right integral on .A; /, then .id ˝ /..a// D .a/ı 1 for all
a 2 A. This follows easily from equations (2.8), (2.9) and the fact that is
a multiple of  ı S.

ii) .S 2 .a// D .S.a/ı/ D .S.ı 1 a// D .ı 1 aı/ for all a 2 A.

2.2.4 The modular automorphism of an integral

In general, an integral  on a multiplier Hopf algebra .A; / need not be tracial
in the sense that .ab/ D .ba/ for all a; b 2 A. However, there exists some
control on the non-commutativity of A with respect to , namely, there exists an
automorphism  of A that satisfies .ab/ D .b .a// for all a; b 2 A. This
modular automorphism is an analogue of the modular automorphism of a weight
on a von Neumann algebra, see Section 8.1.3.
The construction of this modular automorphism proceeds in several steps.

Lemma 2.2.15. For every P Hopf algebra W A˝A ! A˝A

.A; /, the maps R; T P
given by T .a ˝ b/ WD a.1/ ˝ bS 2 .a.2/ / and R.a ˝ b/ WD a.1/ ˝ bS.a.2/ /
are inverse to each other.

Proof. By Proposition 1.3.12,

X X
T .R.a ˝ b// D T .a.1/ ˝ bS.a.2/ // D a.1/ ˝ bS.a.3/ /S 2 .a.2/ /
X X
D a.1/ ˝ bS.S.a.2/ /a.3/ / D a.1/ ˝ b .a.2/ / D a ˝ b

and similarly R.T .a ˝ b// D a ˝ b for all a; b 2 A. 

Proposition 2.2.16. Let  be a left integral on a regular multiplier Hopf algebra

.A; /. Then the subspaces f.  a/ j a 2 Ag and f.b  / j b 2 Ag of A0 coincide.
 2.2. Integrals and their modular properties 57

Proof. To simplify notation, we assume that .A; / is a Hopf algebra; the general
case is proved similarly, see [177, Proposition 3.11]. We show that for every a 2 A,
there exists a b 2 A such that .  a/ D .b  /, the converse follows from a similar
argument. Let be a right integral on .A; /. By Proposition 2.2.9, there exists
an a0 2 A such that .  a/ D .  a0 /. Two applications of Lemma 2.2.12 show
that for all c; d; x 2 A,
X X X
.xc.1/ /.dS.c.2/ // D .x.1/ c/.dx.2/ / D .S.d.1/ /c/.d.2/ x/:

Since the map R defined in the previous lemma is surjective, the composition
.id ˝/ ı R W A ˝ A ! A is surjective as well. Hence we can write a0 in the form
X X
a0 D .id ˝/.R.ci ˝ di // D ci.1/ .di S.ci.2/ // for some ci ; di 2 A:
i i

P above, applied to c WD ci and d WD

The equation
0
di for each i individually, shows
that b WD i .S.di.1/ /c/di.2/ satisfies .  a / D .b  /. 
Let  be a left integral on a regular multiplier Hopf algebra .A; /. Since the
map  ıS 2 is left-invariant, there exists a unique scalar  2 k such that  ıS 2 D  ,
see Theorem 2.2.10. This element  figures in the next theorem:
Theorem 2.2.17. Let .A; / be a regular multiplier Hopf algebra with a left inte-
gral . Then there exists an automorphism  of A such that

.ax/ D .x .a// for all a; x 2 A:

Furthermore,
 ı  D ;  ı  D .S 2 ˝  / ı ;  .ı/ D  1 ı:

Proof. By Proposition 2.2.16, there exists for every a 2 A an element  .a/ 2 A

such that .ax/ D .x .a// for all x 2 A, and this element is unique because 
is faithful (Proposition 2.2.4). The assignment  W A ! A; a 7!  .a/, is injec-
tive because  is faithful, surjective by Proposition 2.2.16, and a homomorphism
because

.x.ab// D .abx/ D .bx .a// D .x .a/ .b// for all a; b; x 2 A:

Dropping x in the relation above, we obtain . .c// D .c/ for all c 2

span A2 D A.
Let us prove that  ı  D .S 2 ˝  / ı . Since functionals of the form
! WD .b  /, where b 2 A, separate the elements of A (Proposition 2.2.4), it is
enough to show that for all such functionals ! and for all a 2 A,

.id ˝!/.. .a/// D .id ˝!/..S 2 ˝  /..a///;

58 Chapter 2. Multiplier Hopf algebras and their duality

that is,
.id ˝/..1 ˝ b/. .a/// D .id ˝/..1 ˝ b/..S 2 ˝  /..a////:

Two applications of Lemma 2.2.12 show that the left-hand side is equal to

.S ˝ /..b/.1 ˝  .a/// D .S ˝ /..1 ˝ a/.b//

D .S 2 ˝ /..a/.1 ˝ b// D .id ˝/..1 ˝ b/..S 2 ˝  /..a////:

Finally, by definition of  and by Remark 2.2.14 ii),

 .a/ D .S 2 .a// D .ı 1 aı/ D .aı .ı 1 //

for all a 2 A. Therefore, ı .ı 1 / D  and  1 ı D  .ı/. 

Remarks 2.2.18. i) For every right integral on a regular multiplier Hopf algebra
.A; /, there exists a modular automorphism  0 of A such that .ax/ D .x 0 .a//
for all a; x 2 A, and this  0 satisfies

 0 D S 1 ı  1 ı S D ı .  /ı 1 and  ı  0 D . 0 ˝ S 2 / ı :

This follows easily from the fact that is a multiple of  ı S D .  ı/.

ii) For a multiplier Hopf -algebra, the modular automorphism of a left or right
integral need not be a -homomorphism.

2.3 Duality
Let .A; / be a regular multiplier Hopf algebra with integrals. Then one can
y /
associate to .A; / a dual regular multiplier Hopf algebra .A; y with integrals and
y y
y /
show that the bidual .A; y is naturally isomorphic to .A; /. This duality extends
to algebraic quantum groups and constitutes an algebraic analogue of Pontrjagin
duality, compare also with the discussion in Example 1.4.3. These results and
constructions are due to Van Daele; we follow his article [177].

2.3.1 The duality of regular multiplier Hopf algebras

Let .A; / be a regular multiplier Hopf algebra with a left integral  and a right
integral . Put

Ay WD f.  a/ j a 2 Ag  A0 :

By Theorem 2.2.17 and Proposition 2.2.9,

Ay D f.a  / j a 2 Ag D f .a  / j a 2 Ag D f .  a/ j a 2 Ag:
 2.3. Duality 59

In particular, Ay does not depend on the choice of ; of course, this follows from
Theorem 2.2.10 as well.
We equip Ay with a multiplication and a comultiplication that are dual to the
comultiplication and multiplication, respectively, of .A; /. Let us begin with the
multiplication. Given elements !1 D .  a1 / and !2 D .  a2 / of A, y we can
define a linear map !1 !2 W A ! k by the formula

.!1 !2 /.x/ WD .!1 ˝ !2 /..x// D . ˝ /..x/.a1 ˝ a2 //: (2.10)

Lemma 2.3.1. Equation (2.10) defines an associative and non-degenerate multi-

y
plication on A.
Proof. Let !1 D .  a1 / and !2 D .  a2 / as above. We have to show that
the map !1 !2 belongs to A. y Since .A; / is regular, we can write a1 ˝ a2 D
P
i .ci /.di ˝ 1/ with suitably chosen ci ; di 2 A. Then
X X
.!1 !2 /.x/ D . ˝ /..xci /.di ˝ 1// D .xci /.di / D .xb/
i i
P
for all x 2 A, where b D y
i ci .di /. Hence !1 !2 2 A. Moreover, using
regularity of .A; /, we see that AyAy spans A.y
Associativity of the multiplication follows easily from coassociativity of . Let
us show that the multiplication is non-degenerate. If !1 !2 D 0 for all !2 , then

0 D .!1 !2 /.b/ D . ˝ /..b/.a1 ˝ a2 // for all b; a2 2 A:

Since the map T1 W A ˝ A ! A ˝ A; b ˝ a2 7! .b/.1 ˝ a2 /, is surjective, we

can replace .b/.1 ˝ a2 / by arbitrary elements c ˝ d 2 A ˝ A. Then we obtain
.ca1 /.d / D 0 for all c; d 2 A, and by Proposition 2.2.4, also a1 D 0 and
!1 D 0. 
y W Ay ! M.A˝
Next, we turn to the comultiplication  y A/.
y The obvious formula
y
..!//.x ˝ y/ WD !.xy/ only defines a map  y W Ay ! .A ˝ A/0 , and it is not
immediately clear that the image of this map can be identified with a subspace of
M.Ay ˝ A/.
y Therefore we take another approach:
y W Ay ! M.Ay ˝ A/
Lemma 2.3.2. There exists a homomorphism  y such that
y 2 / 2 Ay ˝ Ay and .!
i) .!1 ˝ 1/.! y 2 /.1 ˝ !1 / 2 Ay ˝ Ay for all !1 ; !2 2 A,
y

ii) for all x; y 2 A,

y 2 //.x ˝ y/ D .!1 ˝ !2 /..x/.1 ˝ y//;
..!1 ˝ 1/.! (2.11)
y 2 /.1 ˝ !1 //.x ˝ y/ D .!2 ˝ !1 /..x ˝ 1/.y//:
..! (2.12)
60 Chapter 2. Multiplier Hopf algebras and their duality

Before we give the proof, let us briefly comment on equation (2.11) and (2.12).
If the linear map 2 A0 belonged to A, y then it would be the unit, and inserting
for !1 in (2.11) and (2.12) would yield
 
. ˝ !2 /..x/.1 ˝ y//
y
..!2 //.x ˝ y/ D D !2 .xy/:
.!2 ˝ /..x ˝ 1/.y//
Proof. The proof uses the same techniques as the proofs given in the previous
section; therefore we only indicate the main steps. One shows that:
y equations (2.11) and (2.12) define elements of Ay ˝ A;
• for all !1 ; !2 2 A, y
• for each !2 2 A, y the assignments !1 ˝ 1 7! .!1 ˝ 1/.! y 2 / and 1 ˝ !1 7!
y 2 /.1 ˝ !1 / extend to a right multiplier .!
.! y 2 /r and a left multiplier
y 2 /l on Ay ˝ A;
.! y
• .! y 2 /l form a multiplier .!
y 2 /r and .! y 2 / 2 M.Ay ˝ A/
y because
y 2 /l .1 ˝ !3 // D ..!1 ˝ 1/.!
.!1 ˝ 1/  ..! y 2 /r /  .1 ˝ !3 /

y
for all !1 ; !3 2 A;
• the map  y W Ay ! M.Ay ˝ A/
y thus obtained is an algebra homomorphism.
The proofs of the individual steps are lengthy but do not require any substantial new
idea. For the first step, see the proof of Proposition 2.3.5; for further details, see
[177, Section 4]. 
Proposition 2.3.3. The pair .A; y /
y is a regular multiplier Hopf algebra. Its an-
tipode S and counit O are given by .Sy.!//.a/ D !.S.a// and
y

O ..  a// D .a/ D O ..a  //; O . .  a// D .a/ D O . .a  //

for all a 2 A.
Proof. Again, the proof uses the same techniques as the proofs of the preceding
results; for details, see [177, Section 4]. 
Remark 2.3.4. If .A; / is a Hopf algebra, then the counit O on the dual multiplier
y /
Hopf algebra .A; y is given by O .!/ D !.1A / for all ! 2 A. y This follows
immediately from Proposition 2.3.3.
Next, we identify integrals on the multiplier Hopf algebra .A;y /.
y We reserve
the symbol O for left-invariant and the symbol O for right-invariant functionals.
Proposition 2.3.5. The maps O W Ay ! k and O W Ay ! k given by
O .!/ WD .a/ for ! D .  a/; O
.!/ WD .a/ for ! D .a  /
y /,
are a right and a left integral on .A; y respectively.
 2.3. Duality 61

O right-invariance of O follows similarly.

Proof. We only prove left-invariance of ,
Consider elements !1 D .a1  / and !2 D .a2  / of A. y First, we compute the
y 2 /. For all x; y 2 A,
product .!1 ˝ 1/.!
y 2 //.x ˝ y/ D .
..!1 ˝ 1/.! ˝
/..a1 ˝ a2 /.x/.1 ˝ y//:
P
Since .A; / is regular, we can write a1 ˝ a2 D i .1 ˝ ci /.bi / with suitably
chosen bi ; ci 2 A. Then
X
y 2 //.x ˝ y/ D
..!1 ˝ 1/.! . ˝ /..1 ˝ ci /.bi x/.1 ˝ y//
i
X
D .bi x/ .ci y/:
i
P
y 2/ D
Hence .!1 ˝ 1/.! .bi  / ˝ O
.ci  /, and by definition of ,
i
X
O
.id ˝/..! y
1 ˝ 1/.!2 // D .bi  / .ci /:
i

Inserting the relation

X X
bi .ci / D .id ˝ /..1 ˝ ci /.bi // D .id ˝ /.a1 ˝ a2 / D a1 .a.2/ /;
i i

O
we find .id ˝/..! y
1 ˝ 1/.!2 // D
O 2 /. Consequently,
.a1  / .a2 / D !1 .!
y 2 // D 1 y  .!
.id ˝O /..! O 2 /. Since !2 2 Ay was arbitrary, O is left-invariant.
M.A/

Summarizing, we obtain the following duality result:
Theorem 2.3.6. If .A; / is a regular multiplier Hopf algebra with integrals, then
y /
the dual .A; y is a regular multiplier Hopf algebra with integrals. 
Example 2.3.7. Let G be a discrete group and consider the multiplier Hopf algebra
kfin .G/. Recall that the counting measure on G yields an integral  on kfin .G/

2
(Example 2.2.3 ii)). Similar calculations as in Example 1.4.2 show that the dual
algebra kfin .G/ is unital and that the linear map
2
kfin .G/ ! kG; .  ıx / 7! Ux ;

is an isomorphism of Hopf algebras. The dual integral O D O is given by

O ..  ıx // D .ıx / D ıx;e for all x 2 G and corresponds to the integral on
kG defined in Example 2.2.3 iii).
b
Likewise, kG Š kfin .G/ as multiplier Hopf algebras, and again, one can easily
b
verify that the integral on kfin .G/ corresponds to the dual integral on kG defined
in Proposition 2.3.5.
62 Chapter 2. Multiplier Hopf algebras and their duality

The pattern observed in the previous example is typical: The bidual .A;y y
y /
y of
a regular multiplier Hopf algebra .A; / with integrals is naturally isomorphic to
.A; /. The main step towards the proof of this assertion is the following lemma.
Lemma 2.3.8. Let !1 ; !2 2 A, y where !2 D .  a2 / for some a2 2 A. Then
O .!1 !2 / D !1 .S 1 .a2 //.

Proof. To simplify notation, we assume that .A; / is a Hopf algebra; the general
case is proved similarly, see [177, Lemma 4.11]. Let !1 D .  a1 / 2 A,y where
a1 2 A, and choose ci ; di 2 A such that
X XX
a1 ˝ a2 D .ci /.di ˝ 1/ D ci.1/ di ˝ ci.2/ : (2.13)
i i
P
By the proof of Lemma 2.3.1, !1 !2 D .  b/, where b WD i ci .di /, and by the
definition of O (Proposition 2.3.5),
X X
O .!1 !2 / D .b/ D .ci /.di / D . .ci /di /:
i i
P
Now we insert the relation . .ci // D S 1 .ci.2/ /ci.1/ (see Proposition 1.3.14)
in the right-hand side, use equation (2.13), and obtain
XX
O .!1 !2 / D .S 1 .ci.2/ /ci.1/ di / D .S 1 .a2 /a1 / D !1 .S 1 .a2 //: 
i

Theorem 2.3.9. Let .A; / be a regular multiplier Hopf algebra with integrals.
Then the map  W A ! .A/y 0 given by ..a//.!/ WD !.a/ for all a 2 A, ! 2 Ay takes
values in Ay
y and defines an isomorphism of multiplier Hopf algebras .A; / Š
y /.
y
.A; y
y

Proof. First, we show that the map  is injective. Assume that .a/ D 0 for some
a 2 A. Then .ab/ D ..a//..  b// D 0 for all b 2 A, and since  is faithful
(Proposition 2.2.4), a D 0.
Next, we show that the image of  is A.y
y Every element of Ay y is of the form
y Furthermore, since the antipode S is bijective, every ! 2 Ay
O .  !/, where ! 2 A.
can be written in the form ! D .  S.a// for some a 2 A. If ! is written in that
form, then O .  !/ D .a/: by the previous lemma,
y
O .!1 !/ D !1 .S 1 .S.a/// D !1 .a/ D ..a//.!1 / for all !1 2 A:

Thus the map  restricts to an isomorphism of vector spaces A Š A. y

y Straight-
forward but lengthy calculations show that this isomorphism is compatible with the
multiplication and comultiplication on A and A,y
y see [177, Theorem 4.12]. 
 2.3. Duality 63

2.3.2 The duality of algebraic quantum groups

The duality of regular multiplier Hopf algebras with integrals extends to a duality
of algebraic quantum groups as follows.
Lemma 2.3.10. Let .A; / be a multiplier Hopf -algebra. Then the formula
!  .a/ WD !.S.a/ /, where a 2 A and ! 2 A, y defines an involution on Ay that
y /
turns .A; y into a multiplier Hopf -algebra.

Proof. This follows from similar calculations as in the proof of Theorem 1.4.1. 
The following result is an analogue of the Plancherel/Parseval identity:
Theorem 2.3.11. Let .A; / be a multiplier Hopf -algebra with a positive left
y /
integral , and denote by O the right integral on .A; y defined in Proposition 2.3.5.
Then
O ..  a1 / .  a2 // D .a1 a2 / for all a1 ; a2 2 A: (2.14)

In particular, O is positive. Likewise, if is a positive right integral on .A; /,

y /
then the left integral O on .A; y defined in Proposition 2.3.5 is positive.

Proof. We only prove equation (2.14). Given a1 ; a2 2 A, we apply Lemma 2.3.8

to !1 D .  a1 / and !2 D .  a2 /, insert the definition of !1 , and find
O ..  a1 / .  a2 // D O .!1 !2 / D !1 .S 1 .a2 //
D !1 .S.S 1 .a2 // / D !1 .a2 / D .a2 a1 / D .a1 a2 /: 
Remark 2.3.12. The classical Parseval/Plancherel identity has the form
Z Z
f .x/g.x/dx D .F f /./.F g/./d for all f; g 2 L2 .G/;
G y
G

where G is a locally compact abelian group, G y its dual group, F W L2 .G/ ! L2 .G/y
the Fourier transformation, and the integrals are taken with respect to Haar measures
with appropriate scaling.
Summarizing, we obtain the following duality theorem for algebraic quantum
groups:
Theorem 2.3.13. Let .A; / be an algebraic quantum group.
y /
i) The dual .A; y is an algebraic quantum group.

ii) The map  W A ! .A/ y 0 given by ..a//.!/ WD !.a/ for all a 2 A, ! 2 Ay

y
takes values in Ay and defines an isomorphism of algebraic quantum groups
.A; / Š .A;y y
y /.
y
64 Chapter 2. Multiplier Hopf algebras and their duality

Proof. i) This follows directly from Theorem 2.3.6, Lemma 2.3.10, and Theo-
rem 2.3.11.
ii) The map  is an isomorphism of multiplier Hopf algebras by Theorem 2.3.9,
and -linear by a similar calculation as in the proof of Theorem 1.4.1. 
Chapter 3
Algebraic compact quantum groups

Algebraic compact quantum groups form a class of quantum groups that is partic-
ularly well understood. They can be approached from several directions:
First, their theory can be formulated either in terms of Hopf -algebras or in
terms of C  -bialgebras. One can pass back and forth between these two levels,
which provide two descriptions of the same underlying objects. In this chapter, we
focus on the algebraic setting. In Chapter 5, we switch to the setting of C  -algebras
and explain in detail the close relation between algebraic and C  -algebraic compact
quantum groups.
Second, algebraic compact quantum groups can be characterized either by the
existence of a positive integral or in terms of their corepresentation theory. Co-
representations of Hopf algebras are analogues of group representations, and from
a categorical perspective, the corepresentation theory of compact quantum groups
is very similar to the representation theory of compact groups. If a positive integral
exists, then it can be used to prove many properties of the corepresentations in a
similar way as the Haar measure is used in the theory of group representations.
Conversely, if the corepresentations of a Hopf -algebra satisfy certain natural
properties, then the existence of a positive integral can be deduced.
Our starting point is the following definition:

Definition. An algebraic compact quantum group or compact algebraic quantum

group is a unital algebraic quantum group, that is, a Hopf -algebra with a positive
integral. The unique normalized positive integral on an algebraic compact quantum
group is called its Haar state and denoted by h.
A morphism of algebraic compact quantum groups is simply a morphism of the
underlying Hopf -algebras (see Definition 1.3.24).

Remark. Uniqueness of a normalized positive integral follows from iv) (b) in Re-
mark 2.2.2 and Proposition 2.2.6.

In the literature, algebraic compact quantum groups are sometimes also called
compact quantum group algebras or simply compact quantum groups. In an early
article [40], Effros and Ruan called them discrete quantum groups, but this terminol-
ogy has become standard for the dual of a compact quantum group, see Section 3.3.

Examples. We already met the following examples of algebraic compact quantum

groups:
66 Chapter 3. Algebraic compact quantum groups

• the Hopf -algebra C.G/ of functions on a finite group G (see Examples 1.2.4,
1.3.26 i), 2.2.3 i));

• more generally, the Hopf -algebra Rep.G/ of representative functions on a

compact group G (see Examples 1.2.5, 1.3.26 ii), 2.2.3 i));

• the group algebra CG of a discrete group G (see Examples 1.2.8, 1.3.26 iii),
2.2.3 iii)).

A detailed discussion of further important examples can be found in Chapter 6.

Our study of algebraic compact quantum groups proceeds as follows. First, we

introduce the concept of a corepresentation and reformulate it in several equivalent
ways (Section 3.1). Then, we develop the theory of corepresentations (Section 3.2)
and characterize algebraic compact quantum groups in terms of their corepresen-
tations. Finally, we dicuss the duals of compact quantum groups, namely, discrete
algebraic quantum groups (Section 3.3).
Much of the theory developed in this chapter carries over to ordinary Hopf
algebras with normalized integrals – these are precisely the cosemisimple Hopf
algebras [29], [80], [145]. However, our primary interest lies on Hopf -algebras
and their C  -algebraic variants. Most of the results presented in the first two
sections can also be found in [80, Section 11]. The original reference are the two
articles [193] and [202] by Woronowicz, but also the article [40] by Effros and
Ruan.

3.1 Corepresentations of Hopf -algebras

The concept of an algebra, dualized, leads to the concept of a coalgebra. Similarly,
the notion of a module over an algebra, dualized, leads to the notion of a comodule
over a coalgebra. We focus on particular comodules over Hopf -algebras, called
corepresentations, that are relevant for our discussion of algebraic compact quantum
groups in the next section. Most of the following definitions, constructions, and
results apply in wider generality.

3.1.1 Definition and examples

Throughout this section, let .A; / be a Hopf -algebra. In the next definition, we
use the following construction: Given a vector space V equipped with a Hermitian
inner product h  j  iV , we define an A-valued sesquilinear inner product on V ˝ A
by the formula

hv ˝ ajw ˝ biA WD hvjwiV a b for all v; w 2 V; a; b 2 A:

 3.1. Corepresentations of Hopf -algebras 67

Definition 3.1.1. A corepresentation of a Hopf -algebra .A; / on a complex

vector space V is a linear map ı W V ! V ˝ A that satisfies

.ı ˝ idA / ı ı D .idV ˝/ ı ı and .idV ˝ / ı ı D idV :

Let ı and ıQ be corepresentations of .A; / on vector spaces V and Vz , respec-

tively. We call

• a subspace W  V invariant with respect to ı if ı.W /  W ˝ A;

• ı irreducible if V contains no non-trivial invariant subspace;
• ı unitary if hı.v/jı.w/iA D hvjwiV  1A for all v; w 2 V ;
• a linear map T W V ! Vz an intertwiner from ı to ıQ if ıQ ı T D .T ˝ idA / ı ı;
• ı and ıQ equivalent, written ı ' ı,
Q if ı and ıQ admit an invertible intertwiner;
• C .ı/ WD spanf.f ˝ id/.ı.v// j v 2 V; f 2 V 0 g  A the space of matrix
elements of ı.

We denote the space of all intertwiners from ı to ıQ by Hom.ı; ı/.

Q

Remark 3.1.2. If ı and ıQ are equivalent corepresentations, then C .ı/ D C .ı/, Q

and ı is irreducible if and only if ıQ is irreducible. The proof is straightforward.

We extend the Sweedler notation (see Notation 1.3.3) to corepresentations as

follows:

Notation 3.1.3. Let ıP be a corepresentation on a vector space V . Given v 2 V ,

we can write ı.v/ D i v0;i ˝ v1;i , where v0;i 2 V and v1;i 2 A. To simplify
notation, we suppress the summation index i and write
X X
ı.v/ D v0;i ˝ v1;i DW v.0/ ˝ v.1/ :
i

Define ı .n/ W V ! V ˝ A˝n for each n 2 N by

ı .0/ WD idV and ı .nC1/ WD .ı .n/ ˝ idA / ı ı for n 0:

Note that ı .nC1/ D .idV ˝.n/ / ı ı for all n 2 N. We write

X
ı .n/ .v/ DW v.0/ ˝ v.1/ ˝    ˝ v.n/ :

Example 3.1.4. Let ı be a corepresentation on a Hermitian vector space V and

denote by h  j  iAPthe A-valued inner product on V ˝ A defined above. Then
hı.v/jı.w/iA D hv.0/ jw.0/ iV  v.1/  w.1/ for all v; w 2 V .
68 Chapter 3. Algebraic compact quantum groups

The following example shows how group representations can be expressed in

terms of corepresentations.
Example 3.1.5. Let G be a compact group and consider the Hopf -algebra Rep.G/
of representative functions on G (see Examples 1.2.5, 1.3.26 ii)). For every finite-
dimensional vector space V , continuous representations of G on V correspond
bijectively with corepresentations of Rep.G/ on V :
Let
be a continuous representation of G on V . Then for every v 2 V , the
function ı.v/ W G ! V given by x 7!
.x/v belongs to C.GI V /. We identify
C.GI V / with V ˝ C.G/ and regard ı.v/ as an element of V ˝ C.G/. Then ı.v/
is contained in the subspace V ˝ Rep.G/  V ˝ C.G/ and the map v 7! ı.v/
is a corepresentation. Let us prove the first claim.PChoose a basis .vi /i of V , and
denote by .i /i the dual basis of V 0 . Then ı.v/ D i vi ˝ fi for some fi 2 C.G/,
and
X X
fi .x/vi D .idV ˝ evx /.ı.v// D
.x/v D i .
.x/v/vi for all x 2 G;
i i

were evx W C.G/ ! C denotes the evaluation at x 2 G as usual. Thus ı.v/ is

contained in V ˝ Rep.G/. The map ı W v 7! ı.v/ is a corepresentation since for
all x; y 2 G, the composition
.idV ˝ evx ˝ evy / ı .ı ˝ idA / ı ı D .idV ˝ evx / ı ı ı .idV ˝ evy / ı ı
D
.x/
.y/
is equal to
.idV ˝ evx ˝ evy / ı .idV ˝/ ı ı D .idV ˝.evx ˝ evy // ı ı
D .idV ˝ evxy / ı ı D
.xy/:
Conversely, let ı be a corepresentation of Rep.G/ on V . Then the map
W G !
Hom.V / given by
.x/v WD .id ˝ evx /.ı.v// is a representation of G – this follows
from the two equations displayed above and the relation
.e/ D .idV ˝ eve / ı ı D
.idV ˝ / ı ı D idV , where e denotes the unit of G.
The matrix elements of a representation
appear in the associated corepresen-
tation ı as follows: If we identify Hom.V / with Mn .C/ via the basis .vi /i and
consider P
W G ! Hom.V / as an element of C.GI Mn .C// Š Mn .C.G//, then
ı.vj / D i vi ˝ fij for all i; j if and only if
corresponds to the matrix .fij /i;j .
Finally, if V is Hermitian, then
is unitary if and only if ı is unitary: for all
v; w 2 V , the element hı.v/jı.w/iRep.G/ 2 Rep.G/ corresponds to the function
x 7! h
.x/vj
.x/wi.
Example 3.1.6. The comultiplication  is a corepresentation of the Hopf -algebra
.A; / on the vector space A. This corepresentation is called the regular corepresen-
tation of .A; /. If .A; / has a positive integral h, then the regular corepresentation
 3.1. Corepresentations of Hopf -algebras 69

can be equipped with a unitary structure as follows: The map .a; b/ 7! h.a b/ de-
fines a Hermitian inner product on A because h is positive and faithful, and for all
a; b 2 A,

h.a/j.b/iA D .id ˝h/..a /.b// D h.a b/1A D hajbi1A

because h is left-invariant.
The space of matrix elements C./ of the regular corepresentation  is equal
to A because every element a 2 A can be written as a D . ˝ id/..a//.
Every non-zero corepresentation ı on some vector space V admits a non-zero
intertwiner to the regular corepresentation. Indeed, for every f 2 V 0 , the map
Tf W V ! A given by Tf .v/ WD .f ˝ id/.ı.v// is an intertwiner from ı to 
because the following diagram commutes:

Tf

V /V ˝A /+ A
ı f ˝id

ı id ˝ 
  f ˝id ˝ id 
V ˝A
ı˝id /V ˝A˝A /2 A ˝ A.

Tf ˝id

Note that Tf ¤ 0 for some f 2 V 0 because ı.V / ¤ 0.

3.1.2 Reformulation of the concept of a corepresentation

In Example 3.1.5, we observed a correspondence between corepresentations of a
Hopf -algebra Rep.G/ and representations of the group G. In general, corepre-
sentations of a Hopf -algebra .A; / can be described in many equivalent ways, in
particular, also in terms of representations of the dual -algebra A0 . These different
descriptions are the topic of this subsection.
Let us fix some notation. Recall that A0 is a -algebra with respect to the
operations

.fg/.a/ D .f ˝ g/..a//; f  .a/ D f .S.a/ /;

where f; g 2 A0 and a 2 A (Theorem 1.4.1). Denote by † W A˝A ! A˝A the flip

a˝b 7! b˝a. Given a vector space V and an element X 2 Hom.V /˝A, we define
XŒ12 ; XŒ13 2 Hom.V /˝A˝A by XŒ12 WD X ˝1A and XŒ13 WD .id ˝†/.XŒ12 /.
In statement i) of the following proposition, we use the bilinear map

.Hom.V / ˝ A/  .V ˝ A/ ! V ˝ A; .f ˝ a/.v ˝ b/ WD f .v/ ˝ ab:

70 Chapter 3. Algebraic compact quantum groups

If we embed A in Hom.A/ via left multiplication, we obtain an embedding

Hom.V / ˝ A ,! Hom.V ˝ A/, and the map above corresponds to the evalua-
tion map Hom.V ˝ A/  .V ˝ A/ ! V ˝ A.
Proposition 3.1.7. Let V be a vector space with basis .vi /i and let ı W V ! V ˝ A
be a linear map. Define
• a map
W A0 ! Hom.V / by
.f /v WD .idV ˝f /.ı.v//;
P
• a family a D .aij /i;j of elements of A by ı.vj / D i vi ˝ aij ; note that for
each j , there exist only finitely many i such that aij ¤ 0;
P
• if V has finite dimension, an element X 2 Hom.V / ˝ A by X WD i;j eij ˝
aij , where eij 2 Hom.V / denotes the map vk 7! vi  ıj;k .
In the following statements, conditions involving X are to be taken into account
only if V has finite dimension. Qualifiers like “for all i; j ” are omitted, and the
sums formed in (a3), (b3), (d3), (e3), and (e4) are finite.
i)
.f / D .id ˝f /.X / for all f 2 A0 and ı.v/ D X.v ˝ 1A / for all v 2 V .
ii) The following conditions are equivalent:
P
.a1/ .ı ˝ idA / ı ı D .idV ˝/ ı ıI .a3/ .aij / D k aik ˝ akj I
.a2/
is an algebra homomorphism; .a4/ .id ˝/.X / D XŒ12 XŒ13 :

Assume that (a1)–(a4) hold.

iii) The following conditions are equivalent:
.b1/ ı is injective; .b5/ X is invertible;
.b2/
is non-degenerate: .b6/ X 1 D .id ˝S /.X /I
if
.A0 /v D 0, then v D 0; .c1/ .idV ˝ / ı ı D idV I
.b3/ a is invertible:
P 2 A such
there is bij P .c2/
. / D idV I
that k bik akj D ıi;j D k aik bkj I
.c3/ .aij / D ıi;j I
.b4/ as (b3), but with bij D S.aij /I .c4/ .id ˝ /.X / D idV :

Assume that (b1)–(c4) hold, so that ı is a corepresentation.

iv) Let ıQ be a corepresentation on a vector space Vz with basis .vQ k /k and let
T W V ! Vz be a linear map. Define a family of complex numbers .tkj /k;j
P
Q aQ D .aQ kl /k;l , Xz similarly as
; a; X .
by T vj D k vQ k tkj for all j , and
;
Then the following conditions are equivalent:
Q P P
.d1/ T intertwines ı and ı; .d3/ i tki aij D laQ kl tlj ;
.d2/ T intertwines
and
; Q .d4/ .T ˝ 1A /X D Xz .T ˝ 1A /.
 3.1. Corepresentations of Hopf -algebras 71

v) If V is Hermitian and .vi /i orthonormal, then the following conditions are

equivalent:
P 
.e1/ ı is unitary; .e4/ k aki akj D ıi;j ;
.e2/
is a -homomorphism; .e5/ S.aij / D aj i  ;
P 
P   
.e3/ k aki akj D ıi;j D k aik aj k ; .e6/ X X D idV ˝1A D XX .

vi) Let W  V be a subspace and e 2 Hom.V / an idempotent such that eV D W .

Then the following conditions are equivalent:

.f1/ ı.W /  W ˝ A; .f4/ e

.f /e D
.f /e for all f 2 A0 ;
.f2/
.A0 /W  W ; .f5/ .e ˝ 1/X.e ˝ 1/ D X.e ˝ 1/.
.f3/ .e ˝ idA / ı ı ı e D ı ı e;

vii) C.ı/ D spanfaij j i; j g D f.! ˝id/.X / j ! 2 Hom.V /0 g D spanf!.

.f // j
! 2 Hom.V /0 ; f 2 A0 g.
Proof. In the following calculations, all sums taken over infinite index sets have
only finitely many non-zero summands.
i) For all f 2 A0 and all j ,
X X

.f /vj D .id ˝f /.ı.vj // D vi f .aij / and .id ˝f /.X / D eij f .aij /:

i i;j

The first assertion follows, and the second one follows from the relation
X
X.vj ˝ 1A / D vi ˝ aij D ı.vj / for all j:
i

We now give the main steps of the proofs of the assertions ii)–vi).
(a1) , (a2): For all f; g 2 A0 and v 2 V ,

.f /
.g/v D
.fg/v , .
.f / ˝ g/.ı.v// D .id ˝fg/.ı.v//
, .id ˝f ˝ g/..ı ˝ id/.ı.v/// D .id ˝f ˝ g/..id ˝/.ı.v///:

(a1) , (a3): Compare coefficients in

X
.id ˝/.ı.vj // D vi ˝ .aij /;
i
X X
.ı ˝ id/.ı.vj // D ı.vk / ˝ akj D vi ˝ aik ˝ akj :
k i;k
72 Chapter 3. Algebraic compact quantum groups

(a3) , (a4): Compare coefficients in

X
.id ˝/.X / D eij ˝ .aij /;
i;j
X X
XŒ12 XŒ13 D eik elj ˝ aik ˝ alj D eij ˝ aik ˝ akj :
i;j;k;l i;j;k

(c1) , (c2): Evident from the definition. P

(c1) , (c3): Consider the coefficients in .idV ˝ /.ı.vj // D i vi .aij /.
(c2) , (c4): Immediate from i).
(b1) , (b2): For every v 2 V ,

0 D ı.v/ , 0 D .id ˝f /.ı.v// D

.f /v for all f 2 A0 :

(b2) , (c2): This follows from the fact Pthat is the unit of A0 .
(b3) ) (b1): Consider an element v D i i vi 2 V . If 0 D ı.v/, then
X X
0 D ı.v/ D i ı.vi / D vk ˝ i aki ;
i k;i
P
so i i aki D 0 for all k, and by invertibility of .aij /i;j also i D 0 for all i .
(b4) ) (b3): Trivial.
(c3) , (b4): The axioms for the antipode and (a3) imply
X X
aik S.akj / D .aij / D S.aik /akj :
k k

(b3) , (b5)
P and (b4) , (b6): Evidently, .bij /i;j is inverse to .aij /i;j if and
only if Y WD i;j eij ˝ bij is inverse to X :
X X
XY D eij ˝ aik bkj and YX D eij ˝ bik akj :
i;j;k i;j;k

(d1) , (d2): Condition (d1) holds if and only if for all f 2 A0 and v 2 V ,
Q v// D .idV ˝f /..T ˝ idA /.ı.v/// D T ..idV ˝f /.ı.v///:
.idV ˝f /.ı.T

Here, the left- and right-hand side are equal to

.fQ /T v and T
.f /v, respectively.
(d1) , (d3): Compare coefficients in
X X
Q .vj // D
ı.T Q vQ l /tlj D
ı. vQ k ˝ aQ kl tlj ;
l k;l
X X
.T ˝ id/.ı.vj // D T vi ˝ aij D vQ k ˝ tki aij :
i k;i
 3.1. Corepresentations of Hopf -algebras 73

(d2) , (d4): By i), we have for all f 2 A0

Q /T D T
.f / , .id ˝f /.Xz .T ˝ 1A // D .id ˝f /..T ˝ 1A /X /:

.f

(e1) , (e4): This follows from the calculation

X X
hı.vi /jı.vj /iA D hvl jvk iV ali  akj D aki  akj :
k;l k

(e2) , (e5): By definition of the involution on A0 , we have for all f 2 A0

hvi j
.f /vj iV D h
.f  /vi jvj iV
X X
, hvi jvk iV f .akj / D f  .ali /hvl jvj iV
k l

, f .aij / D f  .aj i /
, f .aij / D f .S.aj i / /:

(e3) ) (e4): Trivial. (e4) ) (e3): This follows from the invertibility of a,
see (b3).
(e3) , (e5): By (b4), the family a D ..aj i / /i;j is inverse to a if and only if
S.aij / D .aj i / for all i; j .
(e3) , (e6): This follows from a similar calculation as in (b3) , (b5).
(f1) , (f2): ı.W /  W ˝ A if and only if for all f 2 A0 ,

.id ˝f /.ı.W //  W; that is,

.f /W  W:

(f1) , (f3) and (f2) , (f4): Evident.

(f4) , (f5): Immediate from i).
vii) These relations follow easily from the definition. 
The preceding proposition motivates the following definitions:
Definition 3.1.8. Let I be some index set. A family .aij /i;j 2I of elements of A is
called a corepresentation matrix if fi 2 I j aij ¤ 0g is finite for each j and
X
.aij / D aik ˝ akj and .aij / D ıi;j for all i; j 2 I:
k

Let a D .aij /i;j and aQ D .aQ kl /k;l be corepresentation matrices. We call

P 
• the matrix a unitary if k aki akj D ıi;j for all i; j 2 I ;
• a family .tki /k;i of complex numbers P from a to aQ if the set
Pan intertwiner
fk j tki ¤ 0g is finite for each i and i tki aij D l aQ kl tlj for all k; j ;
• a and aQ equivalent, written a ' a,
Q if they admit an invertible intertwiner;
74 Chapter 3. Algebraic compact quantum groups

• C.a/ D spanfaij j i; j 2 I g the space of matrix elements of a.

We denote the space of all intertwiners from a to aQ by Hom.a; a/.
Q
Definition 3.1.9. Let V be a finite-dimensional vector space. An operator X 2
Hom.V / ˝ A is called a corepresentation operator if
.id ˝/.X / D XŒ12 XŒ13 and .id ˝ /.X / D idV :

Let X and Xz be corepresentation operators on finite-dimensional vector spaces V

and Vz , respectively. We call
• a subspace W  V invariant if X.e ˝ 1A / D .e ˝ 1A /X.e ˝ 1A / for every
idempotent e 2 Hom.V; W /;
• X irreducible if V contains no non-trivial invariant subspace;
• X unitary if V is Hermitian and XX  D idV ˝1A D X  X ;
• a linear map T W V ! Vz an intertwiner from X to Xz if .T ˝ 1A /X D
z ˝ 1A /;
X.T
• X and Xz equivalent, written X ' Xz , if they admit an invertible intertwiner;
• C.X/ D f.! ˝ id/.X / j ! 2 Hom.V /0 g the space of matrix elements of X .
We denote the space of all intertwiners from X to Xz by Hom.X; Xz /.
Let us summarize the main results of Proposition 3.1.7:
i) For every vector space V with a fixed basis .vi /i2I , we have a bijective corre-
spondence between corepresentations ı on V and corepresentation matrices
.aij /i;j 2I , prescribed by the relation
X
ı.vj / D vi ˝ aij for all j 2 I:
i

ii) For every finite-dimensional vector space V , we have a bijective correspon-

dence between corepresentations ı and corepresentation operators X on V ,
prescribed by the relation
X.v ˝ 1A / D ı.v/ for all v 2 V:

iii) These correspondences preserve unitarity, intertwiners, spaces of matrix ele-

ments, and in case ii) also invariant subspaces and irreducibility.
From now on, we freely identify corepresentations, corepresentation matrices, and
corepresentation operators by means of the bijective correspondences i) and ii).
We call a corepresentation matrix irreducible if it corresponds to an irreducible
corepresentation.
 3.1. Corepresentations of Hopf -algebras 75

Remarks 3.1.10. i) Let X be a corepresentation operator on a finite-dimensional

Hermitian vector space V . If we identify elements of A with left multiplication
operators in Hom.A/ and consider X as an element of Hom.V ˝ A/, we can
express the space of matrix elements C .X / in terms of maps of the form

jwi ˝ 1 W A ! V ˝ A; a 7! w ˝ a; hvj ˝ 1 W V ˝ A ! A; u ˝ a 7! hvjuia;

where v; w 2 V , as follows: C .X / D spanf.hvj ˝ 1/X.jwi ˝ 1/ j v; w 2 V g.

ii) If C is the space of matrix elements of some corepresentation, then .C / 
C ˝ C ; if the corepresentation is unitary, then S.C / D C  . This follows from i)
and Proposition 3.1.7 (a3), (e5).
iii) If ı is a corepresentation on some vector space V and
W A0 ! Hom.V / is
the associated representation defined in Proposition 3.1.7, then

ker
D ff 2 A0 j f .a/ D 0 for all a 2 C.ı/g:

Indeed,
.f / D 0 if and only if for all v 2 V and  2 V 0 ,

0 D .
.f /v/ D . ˝ f /.ı.v// D f .. ˝ id/ı.v//:

iv) Let V and W be finite-dimensional vector spaces. If X 2 Hom.V / ˝ A

is a corepresentation operator and T 2 Hom.V; W / is invertible, then the element
.T ˝ 1/X.T ˝ 1/1 2 Hom.W / ˝ A is a corepresentation operator as well.

3.1.3 Construction of new corepresentations

Corepresentations admit several standard constructions – direct sum, tensor product,
and conjugation – which turn the category of corepresentations of a Hopf -algebra
into a monoidal category [79], [104]. We explain these constructions and describe
them in terms of corepresentations, corepresentation matrices, and corepresentation
operators. For most of the statements, the proofs are straightforward and therefore
will be omitted.

Direct sum. Let .ı˛ /˛ be a family of corepresentations on vector spaces .V˛ /˛ .

Then the map
M L
M M 

˛ ı˛
ı˛ W V ! .V˛ ˝ A/ Š V˛ ˝ A
˛ ˛ ˛ ˛

L again, called the direct sum of .ı˛ /˛ . L

is a corepresentation This is the unique core-
presentation on ˛ V˛ for which the inclusion ˇ W Vˇ ,! ˛ V˛ is an intertwiner
for each ˇ.
76 Chapter 3. Algebraic compact quantum groups

The direct sum construction is functorial: for every second family of corepre-
sentations .ıQ˛ /˛ , there exists a map
Y   M

˛
Hom.ı˛ ; ıQ˛ / ! Hom  
˛
ı˛ ; ıQ˛ ; .T˛ /˛ 7!
˛
 T˛ WD
˛
T˛ :
˛

L then ˛ ı˛ is unitary with respect to the canonical Her-

If each ı˛ is unitary,
mitian structure on ˛ V˛ (note that we take algebraic direct sums).
Clearly,
  X
C 
ı˛ D
˛
C .ı˛ /:
˛

0
For each ˛, denote by
˛ W A ! Hom.V
L ˛ / the representation associated to ı˛ .
Then the representation
W A ! Hom. ˛ V˛ / associated to ˛ ı˛ is given by
0

M

.f /.v˛ /˛ D .
˛ .f /v˛ /˛ for all f 2 A0 and .v˛ /˛ 2 V˛ :
˛

For each ˛, let I˛ be some ` index set and u˛ D .u˛ij /i;j 2I˛ an I˛  I˛ -co-
representation matrix. Put I WD ˛ I˛ (disjoint union). Then the I  I -matrix
(
u˛ i; j 2 I˛ ;
˛

u˛ WD .uij /i;j 2I ; where uij WD ij
0 otherwise;

is a corepresentation matrix, called the direct sum of .u˛ /˛ . If each u˛ is the

corepresentation matrix corresponding to ı˛ and some basis .ei˛ /i of V˛ , then ˛ u˛
Lthe corepresentation matrix corresponding to ˛ ı˛ and the basis .ei /i2I of
˛
is
˛ V˛ . L
Assume that ˛ V˛ has finite dimension and that .X˛ /˛ is a family of corepre-
sentation operators on the spaces .V˛ /˛ . Then there exists a unique corepresentation
operator
M 
X˛ 2 Hom
˛ ˛
V˛ ˝ A

L
such that for each ˇ, the inclusion ˇ W Vˇ ,! ˛ V˛ is an intertwiner from Xˇ to
˛ X˛ . This operator ˛ X˛ is the image of the family .X˛ /˛ under the canonical
map
Y Y  M 
.Hom.V˛ / ˝ A/ Š Hom.V˛ / ˝ A ,! Hom V˛ ˝ A:
˛ ˛ ˛

If each X˛ is the corepresentation operator corresponding to ı˛ , then ˛ X˛ is the

corepresentation operator corresponding to ˛ ı˛ .
 3.1. Corepresentations of Hopf -algebras 77

Tensor product. Let ıV and ıW be corepresentations on vector spaces V and W ,

respectively. Then the map
X
ıV  ıW W V ˝ W 7! .V ˝ W / ˝ A; v ˝ w 7! v.0/ ˝ w.0/ ˝ v.1/ w.1/ ;

is a corepresentation again, called the tensor product of ıV and ıW .

The tensor product construction is functorial: for every second pair of corepre-
sentations ıVz and ıWz , there exists a map

Hom.ıV ; ıVz /  Hom.ıW ; ıWz / ! Hom.ıV  ıW ; ıVz  ıWz /;

.S; T / 7! S  T WD S ˝ T:

If ıV and ıW are unitary, then ıV  ıW is unitary with respect to the canonical

Hermitian structure on V ˝ W . Indeed, for all v; vQ 2 V and w; wQ 2 W ,
˝ ˛
.ıV  ıW /.v ˝ w/j.ıV  ıW /.vQ ˝ w/ Q A
X
 
D hv.0/ jvQ .0/ iV hw.0/ jwQ .0/ iW w.1/ v.1/ vQ .1/ wQ .1/
Q V hwjwi
D hvjvi Q W 1A D hv ˝ wjvQ ˝ wi
Q .V ˝W / :

Clearly,

C .ıV  ıW / D span C.ıV /C.ıW /:

Let I and J be two index sets and v D .vij /i;j 2I and w D .wk;l /k;l2J two
corepresentation matrices. Then the .I  J /  .I  J /-matrix

v  w D .vij wkl /.i;k/;.j;l/

is a corepresentation matrix again. If v, w are the corepresentation matrices cor-

responding to ıV , ıW and some bases .ei /i2I , .fk /k2J of V , W , respectively,
then v  w is the corepresentation matrix corresponding to ıV  ıW and the basis
.ei ˝ fk /.i;k/ of V ˝ W .
Assume that V and W have finite dimension and that X and Y are corepresen-
tation operators on V and W , respectively. Then the operator

X  Y WD XŒ13 YŒ23 2 Hom.V / ˝ Hom.W / ˝ A  Hom.V ˝ W / ˝ A

is a corepresentation operator again. Here we have YŒ23 D idV ˝Y and XŒ13 D

.† ˝ idA /.idW ˝X /, where † W Hom.W / ˝ Hom.V / ! Hom.V / ˝ Hom.W /
denotes the flip. If X and Y are the corepresentation operators corresponding to ıV
and ıW , then X  Y is the corepresentation operator corresponding to ıV  ıW .
The tensor product is associative in a natural sense. If the Hopf -algebra
.A; / is commutative, then the tensor product is commutative – in that case, the
78 Chapter 3. Algebraic compact quantum groups

Š
natural isomorphism V ˝ W  ! W ˝ V intertwines ıV  ıW and ıW  ıV , and
ıV  ıW ' ıW  ıV . If .A; / is not commutative, this relation need not hold.
However, there exist interesting examples of non-commutative Hopf -algebras for
which the bifunctors
.ıV ; ıW / 7! ıV  ıW and .ıV ; ıW / 7! ıW  ıV
are naturally equivalent. A natural equivalence between these bifunctors is called a
braiding if it satisfies some additional coherence conditions, and Hopf -algebras
that possess a braiding are called braided. An interesting discussion of braided
Hopf -algebras and their relation to knot invariants is given in [79].
The tensor product is distributive: for each pair of families of corepresentations
.ı˛ /˛ and .ıQˇ /ˇ , there exists a natural isomorphism
   
˛
ı˛  ˇ
ıQˇ ' .ı˛  ıQˇ /:
˛;ˇ

Conjugation. Let ıV be a corepresentation on a vector space V . Denote by Vx

the conjugate vector space of V and by v 7! vN the canonical conjugate-linear
isomorphism. Then the map
X
ıV W Vx ! Vx ˝ A; vN 7! 
v.0/ ˝ v.1/ ;

is a corepresentation again, called the conjugate of ıV .

For every corepresentation ıW on a vector space W , there exists a conjugate-
linear map
Hom.ıV ; ıW / ! Hom.ıV ; ıW /; T 7! Tx ;

where Tx vN D T v for each T 2 Hom.V; W / and v 2 V . Moreover, a subspace

V0  Vx is invariant for ıV if and only if V0  V is invariant for ıV ; in particular,
ıV is irreducible if and only if ıV is irreducible.
If ıV is unitary, then ıV need not be unitary with respect to the natural Hermitian
structure on Vx .
Evidently,
C .ıV / D C .ıV / :
Let I be an index set and v D .vij /i;j 2I a corepresentation matrix. Then
vN WD .vij  /i;j 2I
is a corepresentation matrix again. If v corresponds to ıV and some basis .ei /i
ei /i of Vx . If v is unitary, then
of V , then vN corresponds to ıV and the basis .S
vN D .S.vj i //i;j 2I by Proposition 3.1.7 v).
 3.1. Corepresentations of Hopf -algebras 79

If V has finite dimension and X is a corepresentation operator on V , then the

image of X under the conjugate-linear map
Hom.V / ˝ A ! Hom.Vx / ˝ A; T ˝ a 7! Tx ˝ a ;
is a corepresentation operator again, which we denote by X x . If X corresponds to
x
ıV , then X corresponds to ıV , as one can easily check.
Conjugation is compatible with direct sums and tensor products in the following
sense: For every family of corepresentations .ı˛ /˛ and for each pair of corepresen-
tations ıV , ıW , there exist natural isomorphisms
 
 ˛
ı˛ ' ı˛ ; .ıV  ıW / ' ıW  ıV :
˛

The following result is called Frobenius reciprocity:

Proposition 3.1.11. Let ıU , ıV , ıW be corepresentations on spaces U , V , W ,
where ıU is unitary and finite-dimensional. Then the natural isomorphisms
Š
U W Hom.Ux ˝ V; W / 
! Hom.V; U ˝ W /;
Š
U ! Hom.V; W ˝ Ux /
W Hom.V ˝ U; W / 
restrict to isomorphisms
Š
Hom.ıU  ıV ; ıW / 
! Hom.ıV ; ıU  ıW /;
Š
Hom.ıV  ıU ; ıW / 
! Hom.ıV ; ıW  ıU /:
Proof. We only prove the assertion concerning U ; for U , the proof is similar.
Let u, v, w be the corepresentation matrices corresponding to ıU , ıV , ıW and bases
.ei /i , .fk /k , .gr /r of U , V , W , respectively, where .ei /i is orthonormal. With
respect to the bases .S ei ˝ fk /.i;k/ , .gr /r , .fk /k , .ei ˝ gr /.i;r/ of Ux ˝ V , W , V ,
U ˝ W , respectively, the map U is given by
.tr.i;k/ /r;.i;k/ 7! .tr.i;k/ /.i;r/;k :
Now .tr.i;k/ /r;.i;k/ 2 Hom.uN  v; w/ , .tr.i;k/ /.i;r/;k 2 Hom.v; u  w/ because
X X X
wrs ts.j;l/ equals tr.i;k/ .uN  v/.i;k/.j;l/ D tr.i;k/ uij vkl for all r; j; l
s i;k i;k

if and only if
X X X
uij wrs ts.j;l/ D .u  w/.i;r/.j;s/ ts.j;l/ equals tr.i;k/ vkl for all r; i; l:
s;j s;j k


80 Chapter 3. Algebraic compact quantum groups

3.2 Corepresentation theory and structure theory

The corepresentations of an algebraic compact quantum group enjoy the following
nice properties:
1. Every corepresentation is equivalent to a direct sum of irreducible finite-
dimensional unitary corepresentations (Theorem 3.2.1).
2. Two irreducible corepresentations either admit no non-zero intertwiner or
they are equivalent and the space of intertwiners has dimension one (Propo-
sition 3.2.2).
3. The quantum group is spanned by the matrix elements of irreducible corepre-
sentations and these matrix elements satisfy certain orthogonality relations
with respect to the Haar state (Proposition 3.2.6 and 3.2.9).
4. The modular properties of the Haar state can be described in terms of a
one-parameter family of characters on the quantum group which can be con-
structed out of certain intertwiners of irreducible corepresentations (Theo-
rem 3.2.19).
5. Those Hopf -algebras that are algebraic compact quantum groups can be
characterized in terms of corepresentations (Theorem 3.2.12).
All of these results are proved in the following sections. For later applications, we
state some results in a slightly wider generality.

3.2.1 Decomposition into irreducible corepresentations

In this subsection, we show that the category of corepresentations of an algebraic
compact quantum group can be described in terms of its irreducible corepresenta-
tions. Throughout this subsection, .A; / is a Hopf -algebra.
The main step is the following theorem:
Theorem 3.2.1. Let ı be a corepresentation of .A; / on a vector space V .
i) If .A; / is an algebraic compact quantum group, then ı is equivalent to a
unitary corepresentation.
ii) If ı is unitary and W  V is an invariant subspace, then the orthogonal
complement W ?  V is invariant again.
iii) Every element v 2 V is contained in some finite-dimensional invariant sub-
space of V . In particular, V has finite dimension if ı is irreducible.
iv) If for every finite-dimensional invariant subspace W  V , the restriction
ıjW is equivalent to a unitary corepresentation, in particular, if .A; / is an
algebraic compact quantum group, then ı is equivalent to a direct sum of
finite-dimensional irreducible unitary corepresentations.
 3.2. Corepresentation theory and structure theory 81

Proof. i) Choose some Hermitian inner product hjiV on V . Since the Haar state h
of .A; / is positive and faithful, the formula .a; b/ 7! h.a b/ defines a Hermitian
inner product on A. By a standard argument, the formula hv ˝ ajw ˝ bi WD
hvjwiV h.a b/ defines a Hermitian inner product on V ˝ A. Because ı is injective,
we can define a second Hermitian inner product h  j  i0V on V by the formula
X
hvjwi0V WD hı.v/jı.w/i D 
hv.0/ jw.0/ iV  h.v.1/ w.1/ / for all v; w 2 V:

The associated A-valued inner product h  j  iA0 on V ˝ A is given by

X X
hı.v/jı.w/iA0 D hv.0/ jw.0/ i0V v.1/

w.1/ D 
hv.0/ jw.0/ iV h.v.1/ 
w.1/ /v.2/ w.2/

for all v; w 2 V . Using the fact that h is right-invariant, we find

X
hı.v/jı.w/iA0 D 
hv.0/ jw.0/ iV h.v.1/ w.1/ /1A D hvjwi0V 1A :

Therefore, ı is unitary with respect to the inner product h  j  i0V .

ii) By Proposition 3.1.7 vi), W  V is invariant for ı if and only if it is invariant
for the associated -representation
W A0 ! Hom.V /. Now the claim follows
from a standard argument.
iii) Denote by
W A0 ! Hom.V / the representation associated to ı in Propo-
sition 3.1.7. Since
. / D idV , the subspace
.A0 /v  V contains v. P Evidently,

.A0 /v is invariant for

, so by Proposition 3.1.7 vi) also for ı. If ı.v/ D i vi ˝ai ,
where vi 2 V and ai 2 A, then
.A0 /v is contained in the linear span of the vi .
Therefore,
.A0 /v has finite dimension.
iv) This follows from ii) and iii) by an application the Lemma of Zorn. 

The next step is an analogue of the Lemma of Schur [22, Chapter II, Lemma 2.1],
[62, Chapter VII, Theorem 27.9] for intertwiners of corepresentations:

Proposition 3.2.2. Let ıV and ıW be corepresentations on vector spaces V and W ,

respectively.

i) For every T 2 Hom.ıV ; ıW /, the subspaces ker T  V and Im T  W are

invariant.

ii) Assume that ıV and ıW are irreducible. Then either dim Hom.ıV ; ıW / D 1
and ıV ' ıW , or Hom.ıV ; ıW / D 0.

Proof. i) Let T 2 Hom.ıV ; ıW / and denote by

V and
W the representations
of A0 associated to ıV and ıW , respectively. Then
W .A0 /T D T
V .A0 / by
Proposition 3.1.7 iv), whence
W .A0 / Im T  Im T and
V .A0 / ker T  ker T .
Now the claim follows from Proposition 3.1.7 vi).
82 Chapter 3. Algebraic compact quantum groups

ii) By i), every non-zero intertwiner from ıV to ıW is bijective. If S and T are

non-zero intertwiners, then det.S  T /, considered as a polynomial in , must
vanish at some 0 2 C. Since S 0 T is not bijective, we must have S 0 T D 0.
Note that the determinant det.S  T / is well defined because the dimension of V
and W is finite by Theorem 3.2.1 iii). 
Proposition 3.2.3. Two irreducible unitary corepresentations are equivalent if and
only if they admit a unitary intertwiner.
Proof. Let ı be an irreducible corepresentation on a vector space V . We only need
to show that up to multiplication by a constant factor, there exists only one Hermitian
inner product on V with respect to which ı is unitary. Given such an inner product
h  j  iV , denote by F W Vx ˝ V ! C the linear map given by vN ˝ w 7! hvjwiV ,
and by  W C ! C ˝ A; 1 7! 1 ˝ 1A , the trivial corepresentation. Then for all
v; w 2 V ,

hvjwiV 1A D F .vN ˝ w/1A D  .F .vN ˝ w//

and
X
hı.v/jı.w/iA D 
F .v.0/ ˝ w.0/ / ˝ v.1/ w.1/ D .F ˝ id/..ıN  ı/.vN ˝ w//:

Consequently, ı is unitary with respect to h  j  iV if and only if F 2 Hom.ıN 

ı;  /. But by Frobenius reciprocity (Proposition 3.1.11), dim Hom.ıN  ı;  / D
dim Hom.ı; ı/ D 1, and the claim follows. 
Using the preceding results, we can describe the category of corepresentations
of an algebraic compact quantum group as follows:
Corollary 3.2.4. Let .A; / be an algebraic compact quantum group and .ı˛ /˛2I
a maximal family of pairwise inequivalent irreducible corepresentations of .A; /.
i) Every corepresentation of .A; / is equivalent to a direct sum k ı˛k , where
.˛k /k is some family of elements of I .
ii) Let .˛k /k and .ˇl /l be families of elements of I . Then
  Y   Y M
Hom k

ı˛k ;
l
ıˇl Š
k
Hom ı˛k ; 
l
ıˇl Š
k
C:
l s.t.

˛k Dˇl

This corollary shows that the corepresentation category of an algebraic com-

pact quantum group, regarded as an ordinary category, has a very simple struc-
ture. The interesting and non-trivial information is contained in the monoidal
structure of this category, that is, in the tensor product. Note that the conjuga-
tion of corepresentations can be reconstructed from the monoidal structure: If ıV
 3.2. Corepresentation theory and structure theory 83

and ıW are irreducible finite-dimensional corepresentations, then ıW ' ıV if and

only if dim Hom.; ıV  ıW / ¤ 0, where  denotes the trivial corepresentation.
This follows from Frobenius reciprocity (Proposition 3.1.11) and Schur’s Lemma
(Proposition 3.2.2 ii)).

3.2.2 Schur’s orthogonality relations

The matrix elements of irreducible corepresentations satisfy an analogue of Schur’s
orthogonality relations known from the representation theory of compact groups
[22, Chapter II, Theorem 4.5, 4.6], [62, Chapter VII, Theorem 27.19]. To prove
these relations, we shall use the following lemma:

Lemma 3.2.5. Let .A; / be a Hopf -algebra with a normalized integral h, let
ıV and ıW be corepresentations on finite-dimensional vector spaces V and W ,
respectively, and let R 2 Hom.V; W /. Denote by X and Y the corepresentation
operators corresponding to ıV and ıW , respectively, and define S; T 2 Hom.V; W /
by

S WD .id ˝h/.Y 1 .R ˝ 1/X / and T WD .id ˝h/.Y .R ˝ 1/X 1 /:

Then S; T 2 Hom.ıV ; ıW /. If R 2 Hom.ıV ; ıW /, then S D T D R.

Proof. Since Y and X are corepresentation operators, the composition

Y 1 .S ˝ 1/X D Y 1  .id ˝h/.Y 1 .R ˝ 1/X /  X

1 1
D .id ˝h ˝ id/.YŒ13 YŒ12 .R ˝ 1 ˝ 1/XŒ12 XŒ13 /

can be rewritten in the form

.id ˝h ˝ id/..id ˝/.Y 1 /  .R ˝ 1 ˝ 1/  .id ˝/.X //:

By right-invariance of h, this composition is equal to

.id ˝h/.Y 1 .R ˝ 1/X / ˝ 1 D S ˝ 1:

Thus .S ˝1/X D Y .S ˝1/. A similar calculation shows that .T ˝1/X D Y .T ˝1/;

here, one has to use left-invariance of h. Proposition 3.1.7 iv) implies S; T 2
Hom.ıV ; ıW /. If R 2 Hom.ıV ; ıW /, then Y 1 .R˝1/X D R˝1 D Y .R˝1/X 1
by Proposition 3.1.7 iv), and hence S D R D T . 

Combining Schur’s Lemma 3.2.2 with Lemma 3.2.5, we obtain the following
orthogonality relations:
84 Chapter 3. Algebraic compact quantum groups

Proposition 3.2.6. Let .A; / be a Hopf -algebra with a normalized integral h,

and let ıV and ıW be inequivalent irreducible corepresentations of .A; / on vector
spaces V and W , respectively. Then for all a 2 C .ıV / and b 2 C .ıW /,

h.S.b/a/ D 0 D h.bS.a//: (3.1)

If ıV and ıW are unitary, then h.b  a/ D 0 D h.ba / for all a 2 C .ıV / and
b 2 C .ıW /.
Proof. By Theorem 3.2.1 iii), V and W have finite dimension. Let X and Y
be the corepresentation operators associated to ıV and ıW , respectively, and let
a D .h0 j ˝ id/X.ji ˝ id/ 2 C.X / and b D .h 0 j ˝ id/Y .ji ˝ id/ 2 C.Y /, where
; 0 2 V and ;  0 2 W (see Remark 3.1.10 i)). Then by Proposition 3.1.7 iii),

h.S.b/a/ D h 0 jSi; where S D .id ˝h/.Y 1 .jih0 j ˝ id/X /;

h.bS.a// D h 0 jT i; where T D .id ˝h/.Y .jih0 j ˝ id/X 1 /;

and by Proposition 3.2.2 and Lemma 3.2.5, S; T 2 Hom.ıV ; ıW / D f0g. Thus we

have proved (3.1). The second assertion follows from (3.1) and Remark 3.1.10 ii).

Corollary 3.2.7. Let .A; / be a Hopf -algebra with a normalized integral h.
Then h.C.ı// D 0 for every non-trivial irreducible corepresentation ı. 
Combining the preceding results with the fact that every integral is faithful
(Proposition 2.2.4), we find:
Corollary 3.2.8. Let .A; / be a Hopf -algebra with a normalized integral h.
Then for each pair of corepresentations ıV and ıW ,

Hom.ıV ; ıW / D 0 , h.C .ıV /S.C .ıW /// D 0

, h.S.C .ıV //C .ıW // D 0 , Hom.ıW ; ıV / D 0: 

The next result describes expressions of the form h.b  a/, where b and a are
matrix elements of the same irreducible corepresentation:
Proposition 3.2.9. Let .A; / be a Hopf -algebra with a normalized integral h,
and let a 2 Mn .A/ be an irreducible unitary corepresentation matrix, where n 2 N,
such that the conjugate aN is equivalent to a unitary corepresentation matrix.
i) The matrix at WD .aj i /i;j 2 Mn .A/ is invertible and its inverse at WD .at /1
is a corepresentation matrix.
ii) There exists a unique intertwiner Fz 2 Hom.a; N at / such that Tr Fz D
Tr.Fz 1 / > 0, and this Fz is invertible and positive definite.
 3.2. Corepresentation theory and structure theory 85

iii) For all i; j; k; l,

ıj;l
h.S.aj i /akl / D h.aij  akl / D .Fz 1 /ik ;
Tr.Fz 1 /
ıi;k z
h.aij S.alk // D h.aij akl  / D Fj l :
Tr Fz

iv) The elements .aij /i;j are linearly independent.

Proof. i) By assumption, there exists an invertible T 2 Mn .C/ such that b WD

N 1 is unitary. Then at D aN  D T  b  .T 1 / is invertible and
T aT

at D .aN  /1 D T  b.T  /1 D T  T a.T

N  T /1 : (3.2)

Remark 3.1.10 iv) implies that at is a corepresentation matrix.

ii) Uniqueness of Fz follows from the fact that aN is irreducible (see Section 3.1.3)
and from Proposition 3.2.2 ii). We prove existence. Let T be as in i). Since T
is invertible, T  T is positive definite. Choose  such that Fz WD T  T satisfies
Tr Fz D Tr Fz 1 > 0. By equation (3.2), Fz 2 Hom.a; N at /.

iii) First, note that h.S.aj i /akl / D h.aij akl / and h.aij S.alk // D h.aij akl  /
by Proposition 3.1.7 v). P
Denote by .ei /i the standard basis of Cn , and by Y WD i;j jei ihej j ˝ aij and
P
Yx WD i;j jei ihej j ˝ aij  the corepresentation operators associated to a and a, N
respectively. Note that the operator .Yx  /1 corresponds to the matrix at . Straight-
forward calculations show that
(
 hej jSik el i; where Sik WD .id ˝h/.Y  .jei ihek j ˝ 1/Y /;
h.aij akl / D (3.3)
hei jTj l ek i; where Tj l WD .id ˝h/.Yx .jej ihel j ˝ 1/Yx  /:

By Lemma 3.2.5, Sik 2 Hom.a; a/ and Tj l 2 Hom.at ; a/, N and by Proposi-

z 1
tion 3.2.2 and ii), Sik D ik id and Tj l D j l F for some ik ; j l 2 C. Hence

j l .Fz 1 /ik D hei jTj l ek i D h.aij  akl / D hej jSik el i D ik ıj;l

for all i; j; k; l. This equation shows that j l D 0 for j ¤ l and that  WD jj
does not depend on j . We can read off  D 1= Tr.Fz 1 / from the equation
X X X
n Tr.Fz 1 / D hei jTjj ei i D hej jSi i ej i D Tr Si i D .Tr ˝h/.Y  Y / D n:
i;j i;j i

Thus, we have proved the first equation in assertion iii). The proof of the second
one is similar.
86 Chapter 3. Algebraic compact quantum groups
P
iv) Assume that 0 D k;l akl kl for some kl 2 C. Then for all i; j ,
X 1 X
0D h.aij  akl /kl D .Fz 1 /ik kj :
k;l
Tr.Fz 1 / k

Since Fz 1 is invertible, we must have kj D 0 for all k; j . 

Remark 3.2.10. If .A; / is an algebraic compact quantum group and a 2 Mn .A/
is a corepresentation matrix, then aN is equivalent to a unitary corepresentation matrix
by Theorem 3.2.1 i) and Proposition 3.1.7 v).
For later use, we note the following result:
Proposition 3.2.11. Let ı be an irreducible corepresentation of a Hopf -algebra
.A; /. Then the corepresentation jC.ı/ is equivalent to ı n for some n 2 N.
Proof. Denote by V the underlying vector space of ı and let .fi /i be a basis of V 0 .
Then the maps Ti W V ! A given by v 7! .fi ˝id/.ı.v//
P intertwine ı and the regular
corepresentation  (see Example 3.1.6), and i Im Ti D C.ı/. If Ti is not 0, then
it is injective and its image is invariant by Proposition 3.2.2. P Hence, jIm
Ti
is
equivalent to ı, in particular, it is irreducible. Since Im Ti \ j ¤i Im Tj is an
invariant subspace of jIm Ti , the intersection is either equal to 0 or toLIm Ti . Now
an easy inductive argument shows that C .ı/ is equal to the direct sum j 2J Im Tj
for some subset of indices J . 

3.2.3 Characterization of compact quantum groups

Algebraic compact quantum groups can be characterized in terms of their corepre-
sentations as follows:
Theorem 3.2.12. The following conditions on a Hopf -algebra .A; / are equiv-
alent:
i) .A; / has a positive integral.
ii) Every finite-dimensional corepresentation of .A; / is equivalent to a unitary
corepresentation.
iii) A is spanned by the matrix elements of its irreducible finite-dimensional
unitary corepresentations.
iv) As an algebra, A is generated by the matrix elements of all finite-dimensional
corepresentations that are equivalent to unitary corepresentations.
v) Let .u˛ /˛ be a maximal family of pairwise inequivalent irreducible unitary
corepresentation matrices, where u˛ D .u˛ij /i;j . Then .u˛ij /˛;i;j is a basis
of A.
 3.2. Corepresentation theory and structure theory 87

Proof. i) ) ii): This is Theorem 3.2.1 i).

ii) ) iii): By Theorem 3.2.1 iv), there exists a family .ı˛ /˛ of finite-dimensional
irreducible
P unitary corepresentations such that  ' ˛ ı˛ , and A D C ./ D
˛ C.ı˛ /.
iii) ) iv): Obvious.
iv) ) iii): Let us call a corepresentation unitarizable if it is equivalent to a
unitary corepresentation. If ıV and ıW are unitarizable finite-dimensional corepre-
sentations, then also ıV  ıW and ıV  ıW are unitarizable, and C.ıV / C C.ıW / D
C .ıV  ıW / and C .ıV /C .ıW /  C .ıV  ıW /. Hence, the algebra generated by
the matrix elements of finite-dimensional unitarizable corepresentations is equal
to the vector space spanned by the matrix elements of such corepresentations. By
Theorem 3.2.1 iv), every such corepresentation is equivalent to a direct sum of
irreducible unitary corepresentations.
iii) ) v): Let the family .u˛ /˛ be given.
P From Proposition 3.2.11 and Propo-
sition 3.2.2, one easily deduces C .u˛ / \ ˇ ¤˛ C .uˇ / D 0 for all ˛.
We show that .A; / has a normalized integral h. Observe that the unit 1A
is the matrix element of the trivial corepresentation C ! C ˝ A,  7!  ˝ 1A .
Therefore we can define a linear map h W A ! C by 1A 7! 1 and u˛ij 7! 0 if u˛ does
not correspond to the trivial corepresentation. Since .u˛ij / 2 spanfu˛kl ˝ u˛mn j
k; l; m; ng for every ˛, the map h is a normalized integral.
Finally, we show that for each ˛, the irreducible corepresentation matrix u˛ D
˛
.uij /i;j is equivalent to a unitary one, and then the claim follows from Proposi-
tion 3.2.9 iv). By Corollary 3.2.8 and Proposition 3.2.2, it suffices to show that
P //Cˇ.u // ¤ 0 for some ˇ. But this follows from the assumption that
˛ ˇ
h.S.C.u
A D ˇ C.u / and the fact that h is faithful (Proposition 2.2.4).
v) ) i): The proof of the implication iii) ) v) shows that .A; / has a nor-
malized integral h and that for each ˛, the corepresentation matrix u˛ is equiv-
alent to a unitary one. We prove that h is positive. For each ˛, denote by
Fz˛ 2 Hom.u˛ ; .u ˛ t
P / / the intertwiner constructed in Proposition 3.2.9. Consider
an element a D ˛;i;j ˛ij u˛ij 2 A. By Proposition 3.2.6 and 3.2.9,

X X 1
h.a a/ D ˛ij h.u˛ij  uˇkl / ˇkl D ˛ij .Fz˛1 /ik ˛kj :
˛;i;j;ˇ;k;l ˛;i;j;k
Tr.Fz˛1 /

Since each Fz˛ is positive definite, so is each Fz˛1 . Hence h.a a/ > 0. 

3.2.4 Characters of corepresentations

Let .A; / be an algebraic compact quantum group with Haar state h. An important
tool in the study of corepresentations of .A; / are the associated characters:
88 Chapter 3. Algebraic compact quantum groups

Definition 3.2.13.PThe character of a corepresentation matrix a 2 Mn .A/ is the

element .a/ WD niD1 ai i 2 A.

Recall that the equivalence classes of finite-dimensional corepresentations of

.A; / form a semiring with respect to the direct sum and the tensor product (see
Section 3.1.3). Frequently, it is more convenient to work with rings instead of
semirings. Forming the characters of corepresentations, we can embed the semiring
of (equivalence classes of) finite-dimensional corepresentations into a subring of A:

Proposition 3.2.14. Let a 2 Mm .A/ and b 2 Mn .A/ be corepresentation matrices

of .A; /. Then

.a  b/ D .a/ C .b/; .a  b/ D .a/.b/;

N D .a/ D S..a//; ..a// D m;
.a/
m D n and a ' b ) .a/ D .b/:

Proof. All assertions follow easily from the definitions, so we only prove .a/ N D
.a/ D S..a//
Pm and ..a// D m. We may assume
Pm that a is unitary,
Pm  and then
..a// D iD1 .ai i / D m and S..a// D iD1 S.ai i / D iD1 ai i D .a/ N
by Proposition 3.1.7 iii), v). 

The following results show that the map a 7! .a/ is an embedding:

Proposition 3.2.15. If a, b are irreducible corepresentation matrices, then

(
  1; a ' b;
h..a/ .b// D h..a/.b/ / D
0; otherwise.

Proof. This follows directly from Proposition 3.2.6 and 3.2.9. 

Corollary 3.2.16. Let a be a corepresentation matrix of .A; /, and let .u˛ /˛ be
a maximal family of pairwise inequivalent irreducible corepresentation matrices of
.A; /. For each ˛, put n˛ WD h..u˛ / .a//. Then
X
a'  ˛
.u˛ /n˛ ; .a/ D
˛
n˛ .u˛ /;
X
dim Hom.a; a/ D n2˛ D h..a/ .a//:
˛

In particular, for every other corepresentation matrix b of .A; /, we have a ' b
if and only if .a/ D .b/. 
 3.2. Corepresentation theory and structure theory 89

3.2.5 Modular properties of the Haar state

Let .A; / be an algebraic compact quantum group. The modular properties of the
Haar state of .A; / can conveniently be expressed in terms of a one-parameter
family .fz /z2C of characters. These characters are related to certain intertwiners
which are constructed in the following proposition.
Given a matrix c D .cij /i;j 2 Mn .A/, we put S.c/ WD .S.cij //i;j .
Proposition 3.2.17. Let a 2 Mn .A/ be an irreducible unitary corepresentation
matrix, where n 2 N. Then S 2 .a/ is an irreducible corepresentation matrix and
there exists a unique invertible intertwiner F 2 Hom.a; S 2 .a// such that Tr F D
Tr F 1 > 0. This F is positive definite, and the intertwiner Fz 2 Hom.a;N at / of
x z
Proposition 3.2.9 is equal to F D F . In particular, Tr F D Tr F .
t

Proof. By Theorem 3.2.1 i), there exists an invertible matrix T 2 Mn .C/ such that
N 1 is unitary. By Proposition 3.1.7 v),
b WD T aT

aN D S.a/t and Tx aTx 1 D bN D S.b/t D T t S.a/

N t T t D T t S 2 .a/T t :

Here, T t WD .T t /1 D .T 1 /t . As in Proposition 3.2.9 i), ii), we conclude that

S 2 .a/ is an irreducible corepresentation matrix and that F is unique and equal to
T t Tx D Tx  Tx for some  > 0. The proof of Proposition 3.2.9 i), ii) shows
that Fz D Q  T  T for some Q > 0, so Fz D Fx=.
Q Since Tr F D Tr F 1 and
Tr Fz D Tr Fz , we must have =
1 Q D 1. Finally, Tr F D Tr Fx D Tr Fz because F
is positive. 

Now we turn to the one-parameter family .fz /z2C . We will need the following
concept from function theory: An entire (that is, holomorphic on C) function g is
of exponential growth on the right half-plane if there exist C > 0 and d 2 R such
that jg.z/j  C e d Re.z/ for all z 2 C with Re.z/ > 0.
Lemma 3.2.18. If g1 and g2 are entire functions of exponential growth on the right
half-plane such that g1 .n/ D g2 .n/ for all n 2 N, then g1 D g2 .
Proof. This is a generalization of Carlson’s Theorem [64, Theorem 11.3.3], see
[192, p. 228]. 

Theorem 3.2.19. Let .A; / be an algebraic compact quantum group. Then there
exists a family .fz /z2C of characters on A such that for all z; z 0 2 C and all
a; b 2 A, the following conditions are satisfied:
i) The function z 00 7! fz 00 .a/ is entire and of exponential growth on the right
half-plane.
ii) f0 D and fz  fz 0 D fzCz 0 .
90 Chapter 3. Algebraic compact quantum groups

iii) fz .1A / D 1; fz .S.a// D fz .a/, and fz .a / D fzN .a/.

iv) S 2 .a/ D f1  a  f1 .

v) h.ab/ D h.b.f1  a  f1 // for all a; b 2 A.

The family .fz /z2C is uniquely determined by the conditions i), ii), v).

Proof. We define the family .fz /z as follows. Let .u˛ /˛ be a family of irreducible
unitary corepresentation matrices as in Theorem 3.2.12 v). For each ˛, denote by
F˛ 2 Hom.u˛ ; S 2 .u˛ // the intertwiner constructed in Proposition 3.2.17 and put
fz .u˛ij / WD .F˛z /ij for all z 2 C and all i; j .
We prove the assertions in the order i), ii), iv), v), vi), iii), where “vi)” denotes
the claim that each fz is a character and that the family .fz /z is uniquely determined
by condition v).
i) This is an immediate consequence of the definition of the family .fz /z .
ii) By construction, f0 .u˛ij / D ıi;j D .u˛ij / and
X
.fz  fz 0 /.u˛ij / D .fz ˝ fz 0 /..u˛ij // D fz .u˛ik /fz 0 .u˛kj /
k
X 0 0
D .F˛z /ik .F˛z /kj D .F˛zCz /ij D fzCz 0 .u˛ij /:
k

iv) The relation S 2 .u˛ / D F˛ u˛ F˛1 implies

X
S 2 .u˛ij / D .F˛ /ik u˛kl .F˛1 /lj D .f1 ˝ id ˝f1 /..2/ .u˛ij // D f1  u˛ij  f1 :
k;l

v) By Proposition 3.2.6, we may assume a 2 C .u˛ / and b 2 C.u˛ / for some ˛.

Using Proposition 3.2.9 iii) and the relation Fz D F t (Proposition 3.2.17), we find
X
h..u˛kl / .f1  u˛ij  f1 // D h..u˛kl / .F˛ /im u˛mn .F˛ /nj /
m;n
X ıl;n
D .F˛ /im .F˛1 /mk .F˛ /nj
m;n
Tr F ˛

ıi;k
D .F˛ /lj D h.u˛ij u˛kl  /:
Tr F˛

vi) First, we show that fz is a character for each z. By v), the modular auto-
morphism  of h (Proposition 2.2.17) is given by a 7! f1  a  f1 . Assertion ii)
implies f2 .a/ D .f1   f1 /.a/ D .f1  a  f1 / for all a 2 A, that is, f2 D ı  .
Thus f2 is a character. Since the convolution of characters is a character again,
 3.3. Discrete algebraic quantum groups 91

f2k D f2      f2 (k times) is a character for all k 2 N. Now an application of

Lemma 3.2.18 shows that fz is a character for each z 2 C.
Next, we prove that the family .fz /z is uniquely determined by conditions i),
ii), v). By Proposition 2.2.17, condition v) uniquely determines the automorphism
a 7! f1  a  f1 , and therefore also the character f2 . Next, f2k is uniquely
determined for all k 2 N by condition ii), and Lemma 3.2.18 implies that the
family .fz /z is uniquely determined by the family .f2k /k2N .
iii) By construction, fz .1A / D 1 for each z 2 C, and by ii) and vi),
X
fz .S.a// D fz .S.a.1/ //  f0 .a.2/ /
X
D fz .S.a.1/ //  fz .a.2/ /fz .a.3/ /
X
D fz .S.a.1/ /a.2/ /fz .a.3/ / D fz .1A /fz .a/:

Finally, by Proposition 3.1.7 v),

fz ..u˛ij / / D fz .S.uj˛i // D fz .uj˛i / D .F˛z /j i D .F˛zN /ij D fzN .u˛ij /: 

For each z 2 C, define z;z 0 W A ! A by a 7! fz  a  fz 0 . The preceding
theorem implies:
Corollary 3.2.20. For all z; z 0 2 C, the map z;z 0 is an algebra automorphism
of A, and for all w; w 0 2 C,
0;0 D id; z;z 0 ı w;w 0 D zCw;z 0 Cw 0 ;
h ı z;z 0 D h; z;z 0 ı  D  ı z;
N zN 0 ;
z;z 0 ı S D S ı z 0 ;z ;  ı z;z 0 D . w;z 0 ˝ z;w / ı ;
S 1 D 1;1 ı S: 

3.3 Discrete algebraic quantum groups

A classical result says that the Pontrjagin dual of a compact abelian group is discrete
and that the Pontrjagin dual of a discrete abelian group is compact. Similarly, the
dual of an algebraic compact quantum group is a discrete algebraic quantum group
and vice versa. Like the compact ones, discrete algebraic quantum groups can be
characterized in several equivalent ways. They were studied by Van Daele in [176]
and [177, Section 5], and by Effros and Ruan in [40]; beware that Effros and Ruan
use non-standard terminology.
We take the following definition as our starting point:
Definition 3.3.1. An algebraic quantum group is discrete if, as a -algebra, it is
isomorphic to an algebraic direct sum of matrix algebras, where a matrix algebra
means a -algebra of the form Mn .C/ for some n 2 N .
92 Chapter 3. Algebraic compact quantum groups

Theorem 3.3.2. Let .A; / be an algebraic quantum group and .A; y /

y its dual
.see Section 2.3/. Then the following conditions are equivalent:
i) .A; / is compact.
ii) There exists an element h 2 Ay such that !h D O .!/h D h! for all ! 2 A.
y
y /
iii) .A; y is discrete.

Proof. i) ) ii): Denote by h the Haar state of .A; /. Since A is unital, h D h.  1A /
y and by Remark 2.2.2 ii) and Remark 2.3.4,
belongs to A,

!h D !  h D !.1A /h D O .!/h D h  ! D h! y
for all ! 2 A:

ii) ) i): Let h 2 Ay be such that !h D O .!/h D h! for all ! 2 A. y We show

that h is a left integral on .A; /. Let  be some left integral on A, let a 2 A, and
put T WD .id ˝h/..a// 2 M.A/. Then for all b 2 A,

.T b/ D . ˝ h/..a/.b ˝ 1// D ..  b/h/.a/ D O ..  b//h.a/ D .b/h.a/:

Since  is faithful (Proposition 2.2.4), T D 1M.A/ h.a/. The claim follows.

By definition of A,y we can write h D h.  e1 / D h.e2  / with some e1 ; e2 2 A.
Then h.e2 ab/ D h.ab/ D h.abe1 / for all a; b 2 A, and since h is faithful (Propo-
sition 2.2.4), it follows that e2 a D a and b D be1 for all a; b 2 A. Thus e1 D e2
is a unit of A. L
iii) ) ii): Assume that Ay Š i Mni .C/ as -algebras, where .ni /i is some
family of natural numbers. Since the counit O W Ay ! C is a non-zero homomor-
phism and each summand Mni .C/ is a simple algebra, there exists precisely one
index i0 such that O corresponds to the projection .xi /i 7! xi0 , and ni0 D 1. Now
the element h 2 Ay that corresponds to the family .xi /i given by xi0 D 1 and xi D 0
for i ¤ i0 obviously satisfies !h D O .!/h D h! for all ! 2 A. y
i) ) iii): By Theorem 3.2.12, there exists a family of irreducible unitary matrix
corepresentations u˛ of .A; / such that .u˛ij /˛;i;j is a basis of A. For each ˛,
denote by Fz˛ 2 Hom.u˛ ; .u˛ /t / the intertwiner constructed in Proposition 3.2.9
and put
X
!ij˛ WD .Tr Fz˛ / .Fz˛ /ik h..u˛kj /  / 2 Ay for all i; j:
k

y By Proposi-
Since each Fz˛ is invertible, the family .!ij˛ /˛;i;j forms a basis of A.
tion 3.2.9 iii),

!ij˛ .uˇkl / D ı˛;ˇ  ıi;k  ıj;l for all ˛; ˇ; i; j; k; l:

 3.3. Discrete algebraic quantum groups 93
P
From this equation and the relations .u˛ij / D k u˛ik ˝ u˛kj and S.u˛ij / D uj˛i
(see Proposition 3.1.7), it is easy to deduce that
ˇ
!ij˛ !kl D ı˛;ˇ  ıj;k  !il˛ and .!ij˛ / D !j˛i for all ˛; ˇ; i; j; k; l:

Thus Ay is isomorphic to an algebraic direct sum of matrix algebras. 

Remark 3.3.3. The preceding result can be strengthened as follows. Call a mul-
tiplier Hopf -algebra .A; / discrete if, as a -algebra, it is isomorphic to an al-
gebraic direct sum of matrix algebras. Van Daele [176] showed that every discrete
multiplier Hopf -algebra has a positive left integral and a positive right integral,
and therefore is a discrete algebraic quantum group.
 Part II

Quantum groups in the setting of

C -algebras and von Neumann
algebras
Chapter 4
First definitions and examples

In Part I of this book, we regarded Hopf algebras as purely algebraic objects. This
point of view is adequate for the study of quantum analogues of discrete, compact,
or affine algebraic groups. In Part II, we want to study quantum analogues of general
locally compact groups; therefore we need to consider topological variants of Hopf
algebras. More precisely, the topological aspects of locally compact groups and of
their quantum analogues will be covered by a theory based on C  -algebras, and the
measurable aspects will be covered by a theory based on von Neumann algebras.
Throughout Part II of this book, we shall focus on the setting of C  -algebras.
To find a definition of a Hopf C  -algebra or Hopf–von Neumann algebra that
assumes few axioms but covers many examples is not an easy task. Some of the
related difficulties and the main existing approaches are summarized in Chapter 4.
The classical examples of Hopf C  -algebras and Hopf–von Neumann algebras
associated to locally compact groups are also discussed in this chapter.
Particularly accessible and well understood is the theory of C  -algebraic com-
pact quantum groups developed by Woronowicz [193], [202]. Much of the core-
presentation theory of algebraic compact quantum groups carries over to the C  -
algebraic setting, and the two classes of quantum groups are closely related. These
topics form the contents of Chapter 5 .
General locally compact quantum groups were introduced and studied by Vaes
and Kustermans [88], [91], [158], and by Masuda, Nakagami, and Woronowicz
[110]. The theories developed by these two groups are very satisfying in terms
of the results that can be proved and the examples that are covered; however, the
details of the theories are highly intricate. We outline the approach of Vaes and
Kustermans in Chapter 8, but focus on motivation and give no proofs.
Fundamental to almost all approaches to Hopf algebras and generalized Pontr-
jagin duality in the setting of C  -algebras or von Neumann algebras is the concept
of a multiplicative unitary. Examples of such unitaries were used in various proofs
and constructions for a long time till Baaj and Skandalis put them center-stage, for-
mulated an abstract definition, and gave a comprehensive treatment [7]. We discuss
multiplicative unitaries in Chapter 7; they will reappear in Chapters 8, 9, 10, and 11.

4.1 C  -bialgebras and von Neumann bialgebras

Notation. A short summary on C  -algebras and von Neumann algebras as well
as standard references can be found in the appendix. We shall use the following
notation.
98 Chapter 4. First definitions and examples

Given a C  -algebra A, we denote by M.A/ the multiplier algebra of A; this is a

C -algebra again. If  W A ! M.B/ is a -homomorphism of C  -algebras that is


non-degenerate in the sense that span .A/B D B, we denote the unique extension
to a -homomorphism M.A/ ! M.B/ by  again. Given C  -algebras A1 ; A2 ,
we denote by A1 ˝ A2 the minimal tensor product of A1 and A2 .
Given a von Neumann algebra M , we denote by M its predual which is the
space of all normal linear functionals on M . The von Neumann-algebraic tensor
product of von Neumann algebras M1 and M2 will be denoted by M1 x̋ M2 .
From now on, we denote the purely algebraic tensor product of vector spaces
by “ˇ” instead of “˝”.

C  -bialgebras and von Neumann bialgebras. The concept of a comultiplication

or a bialgebra carries over to the setting of C  -algebras and von Neumann algebras
easily:
Definition 4.1.1. A C  -bialgebra is a C  -algebra A equipped with a non-degen-
erate -homomorphism  W A ! M.A ˝ A/ called the comultiplication such that
i)  is coassociative in the sense that . ˝ idA / ı  D .idA ˝/ ı ,
ii) .A/.1 ˝ A/ and .A/.A ˝ 1/ are contained in A ˝ A.
A C  -bialgebra .A; / is called bisimplifiable if each of the sets .A/.1 ˝ A/
and .A/.A ˝ 1/ is linearly dense in A ˝ A.
A morphism of C  -bialgebras .A; A / and .B; B / is a non-degenerate -ho-
momorphism F W A ! M.B/ that satisfies B ı F D .F ˝ F / ı A .
Remarks 4.1.2. i) In the definition above, the non-degenerate homomorphisms
 ˝ idA ; idA ˝ W A ˝ A ! A ˝ A ˝ A and F ˝ F W A ˝ A ! B ˝ B have been
extended to the multiplier C  -algebra M.A ˝ A/.
ii) Given C  -bialgebras .A; A / and .B; B /, one can construct new C  -bial-
gebras .A; A /op , .A; A /cop , .A; A /op;cop , and .A ˚ B; A˚B /, .A ˝ B; A˝B /
in much the same way as in the case of bialgebras and multiplier bialgebras, see
Remarks 1.3.7 iii).
Definition 4.1.3. A von Neumann bialgebra is a von Neumann algebra M equipped
with a normal unital -homomorphism  W M ! M x̋ M called the comultiplica-
tion such that . x̋ idM / ı  D .idM x̋ / ı .
A morphism of von Neumann bialgebras .M; M / and .N; N / is a normal
-homomorphism F W M ! N that satisfies N ı F D .F x̋ F / ı M .
Remark 4.1.4. Given von Neumann-bialgebras .M; M / and .N; N /, one can
construct new C  -bialgebras .M; M /op , .M; M /cop , .M; M /op;cop , and
.M ˚ N; M ˚N /, .M ˝ N; M ˝N / in much the same way as in the case of
bialgebras and multiplier bialgebras, see Remarks 1.3.7 iii).
 4.1. C  -bialgebras and von Neumann bialgebras 99

Convolution. Let .A; / be a C  -bialgebra. Then we can define a convolution

product on A0 , very much as in the case of bialgebras (see Section 1.3.2): Given
f; g 2 A0 , the functional f ˝ g on A ˝ A0 extends uniquely to a strictly continuous
functional on M.A˝A/ (Corollary 12.1.2), and the convolution product of f and g
is the functional

f  g WD .f ˝ g/ ı  2 A0 :

The convolution product is associative: given f; g; h 2 A0 , the products .f  g/  h

and f .gh/ both are equal to the composition of the strictly continuous functional
f ˝ g ˝ h on M.A ˝ A ˝ A/ with the map .id ˝/ ı  D . ˝ id/ ı .
For each f 2 A0 , we can define slice maps id ˝f; f ˝ id W M.A˝A/ ! M.A/
that are norm-continuous and strictly continuous on bounded subsets (see Sec-
tion 12.4). We put

f  a WD .id ˝f /..a//; a  f WD .f ˝ id/..a//

for all f 2 A0 , a 2 A. The multipliers f  a and a  f belong to A by condition ii)

of Definition 4.1.1 and Proposition 12.4.3. Similarly as in the case of bialgebras,
the maps .f; a/ 7! f  a and .a; f / 7! a  f turn A into a (Banach) bimodule
over A0 .
Similarly, if .M; / is a von Neumann bialgebra, then the space of normal linear
functionals M carries an associative convolution product, given by

f  g WD .f x̋ g/ ı  2 M ;

and M carries a natural structure of a bimodule over M .

From bialgebras to Hopf algebras. If we want to define analogues of Hopf alge-

bras in the setting of C  -algebras or von Neumann algebras, and extend Pontrjagin
duality to such generalized Hopf algebras, the following problems arise:

• In many examples, the counit and the antipode are unbounded and densely
defined only. In those cases, it is difficult to make sense of the axioms
. ˝ id/ ı  D id D .id ˝ / ı  and m ı .S ˝ id/ ı  D  ı D m ı
.id ˝S /ı that characterize Hopf algebras – how can the maps ˝id; id ˝
and S ˝ id; id ˝S , which are unbounded and densely defined, be extended
to the image of ?

• Further problems arise from the axiom m ı .S ˝ id/ ı  D  ı D m ı

.id ˝S / ı , because the multiplication map m W A ˇ A ! A need not
extend to the minimal C  -tensor product A ˝ A.
100 Chapter 4. First definitions and examples

• To generalize Pontrjagin duality, we need an analogue of the Haar measure of

a locally compact group. But in contrast to the situation of locally compact
groups, it seems to be very difficult to deduce the existence of an invariant
integral on a C  -bialgebra or von Neumann bialgebra from a reasonable set
of axioms.

So, at the moment, we have to leave the answer to the question “What is a Hopf
C  -algebra or a Hopf–von Neumann algebra?” open. A satisfying answer that was
proposed by Vaes and Kustermans will be presented in Chapter 8.

4.2 Bialgebras associated to groups

To every locally compact group, one can associate several C  -bialgebras and von
Neumann bialgebras:

Example 4.2.1. Let G be a locally compact group. Then the C  -algebra C0 .G/,
equipped with the comultiplication

 W C0 .G/ ! Cb .G  G/ Š M.C0 .G/ ˝ C0 .G//; .f /.x; y/ WD f .xy/;

is a C  -bialgebra. It is bisimplifiable, as can be seen from the equations

..f /.1 ˝ g//.x; y/ D f .xy/g.y/; ..f /.g ˝ 1//.x; y/ D f .xy/g.x/;

where f; g 2 C0 .G/ and x; y 2 G, and from the fact that the maps G G ! G G
given by .x; y/ 7! .xy; y/ and .x; y/ 7! .x; xy/, respectively, are homeomor-
phisms.
The C  -bialgebra C0 .G/ has a well behaved counit and antipode: the -homo-
morphism W C0 .G/ ! C given by .f / WD f .e/, where e denotes the unit of G,
and the -homomorphism S W C0 .G/ ! C0 .G/ given by .Sf /.x/ WD f .x 1 / for
all x 2 G and f 2 C0 .G/ satisfy the counit identity and the antipode identity
known from Hopf algebras.

Example 4.2.2. To every locally compact group G, one can associate a full/uni-
versal group C  -bialgebra C  .G/ and a reduced group C  -bialgebra Cr .G/ as
follows.
Let us begin with the full group C  -algebra C  .G/. Denote by  the left Haar
measure and by ı the modular function of G. Then the space L1 .G; / is a Banach
-algebra with respect to the multiplication and involution given by
Z
.f  g/.x/ D f .y/g.y 1 x/d .y/; f  .x/ D f .x 1 /ı.x/1 : (4.1)
G
 4.2. Bialgebras associated to groups 101

This Banach -algebra has a bounded approximate unit. Indeed, a routine argument
shows that if .V / is a neighborhood basis of e 2 G, and if .f /R is a net of non-
negative measurable functions on G such that supp f  V and G f d  D 1 for
all , then .f / is a bounded approximate unit.
The group C  -algebra C  .G/ is the enveloping C  -algebra of L1 .G; /, that
is, the completion of L1 .G; / with respect to the norm
kf k WD supfk .f /k j is a  -representation of
L1 .G; / on some Hilbert space H g:
If G is discrete, then for each x 2 G, the element Ux 2 L1 .G; / given by
y!7 ıx;y is a unitary element of C  .G/, and the family .Ux /x2G is linearly dense
in C  .G/. In the general case, the map
Ux W L1 .G; / ! L1 .G; /; .Ux f /.y/ D f .x 1 y/;
is a multiplier for each x 2 G; if G is discrete, we simply recover the family
.Ux /x2G defined before. For each x 2 G, the multiplier Ux 2 M.L1 .G; //
extends to a multiplier Ux 2 M.C  .G// (see Corollary 12.5.2). Evidently,
Ux Uy D Uxy and .Ux / D Ux 1 for all x; y 2 G:
One can show that the linear spanRof the family .Ux /x is strictly dense in M.C  .G//.
Using the relation fg D G f .y/Uy g d .y/, which holds for all f; g 2
L1 .G; /, one can show that
Z
f D f .x/Ux d .x/ for every f 2 L1 .G; /: (4.2)
G

Here, the integral of a (suitable) function h W G ! E  , where E is a Banach

 
Rspace and E its bidual, is defined to be the element of E given by  7!
G .h.x//d.x/.
The C  -algebra C  .G/ has the following universal property. Recall that a
strongly continuous unitary representation of G on a Hilbert space H is a homomor-
phism
from G to the group of unitary operators on H such that for every  2 H ,
the map G ! H given by x 7!
.x/ is continuous. For every such representa-
tion
, there exists a unique non-degenerate -homomorphism C  .G/ ! L.H /
whose extension to M.C  .G// maps Ux to
.x/ Rfor every x 2 G. Indeed, the
map
Q W L1 .G; / ! L.H / given by
.f Q / WD G f .x/
.x/d .x/ defines a
-representation of L1 .G; / and hence a representation of C  .G/ that has the
desired property.
The universal property of C  .G/ can be used to construct non-degenerate
-homomorphisms
 W C  .G/ ! M.C  .G/ ˝ C  .G//; W C  .G/ ! C; S W C  .G/ ! C  .G/op ;
102 Chapter 4. First definitions and examples

such that the extensions of these maps to M.C  .G// act as follows:

.Ux / D Ux ˝ Ux ; .Ux / D 1; S.Ux / D Ux 1 :

Equation (4.2) shows that for each f 2 L1 .G; /  C  .G/,

Z
.f / D .Ux ˝ Ux /f .x/d .x/;
ZG
.f / D f .x/d .x/;
G

and
Z
S.f / D f .x/Ux 1 d .x/
ZG Z
1 1
D f .x /Ux d .x /D f .x 1 /ı.x/1 Ux d .x/;
G G

that is, .S.f //.x/ D f .x 1 /ı.x/1 for all x 2 G.

It is easy to check that the C  -bialgebra .C  .G/; / is bisimplifiable.
Let us turn to the reduced group C  -algebra Cr .G/. By definition, this is
the C  -subalgebra of L.L2 .G; // generated by the left regular representation
L W L1 .G; / ! L.L2 .G; //,
Z
.L.f //.x/ D f .y/.y 1 x/d .y/ for f 2 L1 .G; /;  2 L2 .G; /; x 2 G:
G

The representation L extends to a representation of C  .G/, and this extension yields

a surjective -homomorphism L W C  .G/ ! Cr .G/. We put

Lx WD L.Ux / 2 M.Cr .G//  L.L2 .G; // for each x 2 G:

The quotient map L W C  .G/ ! Cr .G/ is an isomorphism if and only if G

is amenable, that is, if and only if there exists a state  on L1 .G; / which is
left-invariant in the sense that .f / D .f .x  // for all f 2 L1 .G/; x 2 G
[119, Theorem 4.21], [121, Theorem 7.3.9]; here, f .x  / denotes the function
y 7! f .xy/.
There exists a -homomorphism r W Cr .G/ ! M.Cr .G/ ˝ Cr .G// such
that r .Lx / D Lx ˝ Lx for all x 2 G, and the following diagram commutes:

C  .G/
 / M.C  .G/ ˝ C  .G//
L L˝L
 
Cr .G/ / M.C  .G/ ˝ C  .G//.
r r r
 4.2. Bialgebras associated to groups 103

The construction of r involves a multiplicative unitary and is given in Exam-

ple 7.2.13. Like .C  .G/; /, also .Cr .G/; r / is a bisimplifiable C  -bialgebra.
The counit W C  .G/ ! C factorizes to a -homomorphism r W Cr .G/ ! C
if and only if the group G is amenable. The “if” part is obvious, for the “only if”
part, see [119, Proof of Theorem 4.21].
Finally, the antipode S W C  .G/ ! C  .G/op factorizes to a -homomorphism
Sr W Cr .G/ ! Cr .G/op , that is, there exists a commutative diagram

C  .G/
S / C  .G/op
L op
 L
Cr .G/ / Cr .G/op .
Sr

This can be seen as follows: The formula .I /.x/ WD .x/, where  2 L2 .G; /
and x 2 G, defines a conjugate-linear isometric map I on L2 .G; /, and a short
calculation shows that Sr .L.f // D IL.f / I for all f 2 L1 .G; /.
Every continuous homomorphism of locally compact groups  W G ! H in-
duces a morphism of C  -bialgebras
R  W C  .G/ ! M.C  .H // by the formula
Ux 7! U
.x/ , that is, f 7! G U
.x/ f .x/d .x/. This follows easily from the
universal property of C  .G/. However, there need not exist a -homomorphism
;r W Cr .G/ ! M.Cr .H // that makes the following diagram commute:


C  .G/ / M.C  .H //
LG L
  H
Cr .G/ / M.C  .H //;

;r r

here, LG and LH denote the respective quotient maps.

In the abelian case, the C  -bialgebras introduced above are related via Pontr-
jagin duality as follows (compare also with Example 1.4.3). Let G be a locally
compact abelian group G. Then G is amenable [119, Proposition 0.15], whence
Cr .G/ Š C  .G/. For each x 2 G, denote by evx W G y ! C the function given by
y
 7! .x/. Note that evx belongs to Cb .G/ Š M.C0 .G//. y

y
Proposition 4.2.3. Let G be a locally compact abelian group with dual group G.
Š
There exists an isomorphism of C  -bialgebras ˆ W C  .G/  y such that
! C0 .G/
ˆ.Ux / D evx for all x 2 G.

Proof. By the universal property of C  .G/, there exists a -homomorphism

y whose extension to M.C  .G// maps Ux to evx for each
ˆ W C  .G/ ! C0 .G/
x 2 G. The universal property of the C  -algebra C  .G/ furthermore implies that
104 Chapter 4. First definitions and examples

its spectrum, that is, its space of characters or continuous one-dimensional repre-
y Since
sentations, can be identified with the space of characters of G, that is, with G.
ˆ is induced by this identification, ˆ is an isomorphism. The calculation

..ˆ ˝ ˆ/..Ux ///.; 0 / D .ˆ.Ux //./  .ˆ.Ux //.0 /

D .x/0 .x/ D .0 /.x/;

..ˆ.Ux ///.; 0 / D .ˆ.Ux //.0 / D .0 /.x/; y x 2 G;

; 0 2 G;

shows that  ı ˆ D .ˆ ˝ ˆ/ ı . 

Let us briefly consider the setting of von Neumann algebras:

Example 4.2.4. Let G be a locally compact group with left Haar measure . Then
L1 .G; /, equipped with the comultiplication given by .f /.x; y/ D f .xy/ for
all x; y 2 G and f 2 L1 .G; /, is a von Neumann bialgebra.
The group von Neumann algebra L.G/  L.L2 .G; // is the von Neumann
algebra generated by the unitaries Lx 2 L.L2 .G; // defined in Example 4.2.2.
There exists a normal unital -homomorphism L.G/ ! L.G/ x̋ L.G/ such that
Lx 7! Lx ˝ Lx for all x 2 G, and this -homomorphism turns L.G/ into a
von Neumann bialgebra. The proof of this assertion proceeds via a multiplicative
unitary, see Example 7.2.13

4.3 Approaches to quantum groups in the setting of

von Neumann algebras and C  -algebras
The three guises of a quantum group
In the setting of C  -algebras and von Neumann algebras, a quantum group usu-
ally appears in several guises: as a full/universal C  -bialgebra, as a reduced C  -
bialgebra, and as a von Neumann bialgebra. The first two C  -bialgebras, however,
may coincide. Furthermore, a quantum group may be of an algebraic origin in the
sense that it can be described by a (multiplier) Hopf -algebra.
This multitude of bialgebras associated to one (quantum) group appeared already
in Example 4.2.2 and 4.2.4: Every locally compact group G gives rise to the C  -
bialgebra C0 .G/ and the von Neumann bialgebra L1 .G/, and, dually, to the full
group C  -algebra C  .G/, the reduced group C  -algebra Cr .G/, and the group

2
von Neumann algebra L.G/. If the group G is compact (or discrete), then the
multiplier Hopf -algebras Rep.G/ and Rep.G/ (or Cc .G/ and CG, respectively),
provide further descriptions of the same underlying quantum group and its dual.
The universal C  -bialgebra, the reduced C  -bialgebra, and the von Neumann
bialgebra (as well as the multiplier Hopf -algebra, if present) of a quantum group
 4.3. Approaches to quantum groups 105

provide equivalent views on one and the same underlying object, and one can pass
back and forth between these different points of view. The distinction between the
reduced and the universal C  -bialgebra amounts to the choice whether a quantum
group is studied in terms of its regular representation or in terms of all of its repre-
sentations. Naturally, the “reduced theory” is closer to the von Neumann algebraic
setting, and in this book, we focus on that “reduced theory”.

Existing approaches
In the setting of C  -algebras and von Neumann algebras, several approaches to
quantum groups of varying levels of generality and technical complexity have been
developed:
Kac algebras. The first satisfactory extension of Pontrjagin duality to all locally
compact groups was given in the framework of Kac algebras developed by Enock
and Schwartz [47] and by Kac and Vainerman [167], [168] in the seventies. This
theory was formulated in the setting of von Neumann algebras; later, it was extended
to the setting of C  -algebras by Vallin and Enock [49], [170]. The existence of a
Haar weight – the analogue of a Haar measure – is postulated as an axiom.
After the development of the theory, many examples of quantum groups were
found that fit into this framework [31], [48], [77], [78], [96], [106], [130], [132],
[165], [173], [208].
A severe limitation of the theory of Kac algebras is that the antipode is assumed
to be bounded and to commute with the involution.
Compact quantum groups. The theory of compact quantum groups developed
by Woronowicz [193], [194], [202] is perhaps most easily accessible and closest
to the purely algebraic setting. For C  -algebraic compact quantum groups, the
Haar weight is bounded, that is, a state, and its existence can be deduced from a
few natural axioms. Inside every C  -algebraic compact quantum group, one can
identify a unique dense algebraic compact quantum group, and one can pass back
and forth between the algebraic and the C  -algebraic level. We discuss C  -alge-
braic compact quantum groups in Chapter 5.
The first example of a compact quantum group – the famous quantum group
SU .2/ introduced by Woronowicz (see Section 6.2) – showed that the antipode
of a compact quantum group need not be bounded and need not commute with the
involution. In particular SU .2/ is not a Kac algebra.
Multiplicative unitaries. A fundamental tool for the study of quantum groups
in the setting of C  -algebras and von Neumann algebras, in particular in relation
with Pontrjagin duality, are multiplicative unitaries. Their theory was developed
by Baaj and Skandalis [7]; an important contribution was made by Woronowicz
[201]. Roughly, every Hopf -algebra, C  -bialgebra, and von Neumann bialgebra
106 Chapter 4. First definitions and examples

equipped with a positive integral or a Haar weight gives rise to a multiplicative

unitary; conversely, under certain regularity assumptions, a multiplicative unitary
gives rise to a pair of C  -bialgebras and von Neumann bialgebras. Both transitions,
applied subsequently, can be used to construct the Pontrjagin dual of a suitable C  -
bialgebra or von Neumann bialgebra. Moreover, multiplicative unitaries facilitate
the transition between the different guises of a quantum group.
Locally compact quantum groups / weighted Hopf C  -algebras. The most com-
prehensive and at the same time technically most demanding approaches are the
theory of locally compact quantum groups developed by Vaes and Kustermans [91],
[93] and the theory of weighted Hopf C  -algebras developed by Masuda, Nakagami
and Woronowicz [110]. In both theories, the existence of a left Haar weight is as-
sumed. Additionally, the first theory assumes the existence of a right Haar weight,
whereas the second theory assumes the existence of an antipode. The theory of lo-
cally compact quantum groups involves fewer axioms; here, the antipode, like the
counit, is constructed out of the Haar weight. Both theories include a nice gener-
alization of Pontrjagin duality and make substantial use of multiplicative unitaries.
We give a survey on locally compact quantum groups in Chapter 8.
Hopf C  -algebras. Vaes and Van Daele propose a definition of a Hopf C  -alge-
bra [164] which is based on the characterization of Hopf algebras given in The-
orem 1.3.18. Out of the comultiplication, Vaes and Van Daele construct a counit
and an antipode on a subset of the Hopf C  -algebra which is dense in all known
examples.
Chapter 5
C  -algebraic compact quantum groups

C  -algebraic compact quantum groups are particularly well understood:

1. They can be defined in terms of a few simple axioms.

2. From these axioms one can deduce the existence of a Haar state which is the
analogue of the Haar measure of a group and of the Haar state of an algebraic
compact quantum group (Section 5.1).
3. Using the Haar state, one can show that their corepresentation theory is es-
sentially the same as the representation theory of compact groups and the
corepresentation theory of algebraic compact quantum groups (Sections 5.2
and 5.3).
4. Like algebraic compact quantum groups, they can be characterized in terms
of their corepresentations (Section 5.3).
5. Using the corepresentation theory, one can identify a unique dense algebraic
compact quantum group inside every C  -algebraic compact quantum group.
This algebraic compact quantum group answers most questions concerning
the counit, the antipode, and the modular properties of the Haar state (Sec-
tion 5.4, but see also Example 8.1.22 and 8.3.7).
6. Every algebraic compact quantum group can be completed so that one obtains
a C  -algebraic compact quantum group. In general, several completions
exist, but one can always identify a minimal and a maximal one (Section 5.4).

The original references for C  -algebraic compact quantum groups are the articles
[193], [202] by Woronowicz; detailed accounts can also be found in [80], [105].
Throughout this section, we use the notation and definitions introduced in Sec-
tion 4.1 and the background on C  -algebras summarized in Section 12.1.

5.1 Definition and examples

The definition of a C  -algebraic compact quantum group is very brief:

Definition 5.1.1. A C  -algebraic compact quantum group is a unital bisimplifiable

C  -bialgebra.
A morphism of C  -algebraic compact quantum groups is just a morphism of
the underlying C  -bialgebras (see Definition 4.1.1).
108 Chapter 5. C  -algebraic compact quantum groups

More explicitly, a C  -algebraic compact quantum group is a unital C  -algebra

A equipped with a unital -homomorphism  W A ! A ˝ A such that

i)  is coassociative in the sense that . ˝ idA / ı  D .idA ˝/ ı , and

ii) each of the sets .A/.1A ˝ A/ and .A/.A ˝ 1A / is linearly dense in A ˝ A.

If .A; / is a C  -algebraic compact quantum group and the comultiplication  is

understood, we freely speak of A itself as a C  -algebraic compact quantum group.

Examples 5.1.2. i) For every compact group G, the C  -bialgebra C.G/ introduced
in Example 4.2.1 is a C  -algebraic compact quantum group.
ii) For every discrete group G, the C  -bialgebras C  .G/ and Cr .G/ introduced
in Example 4.2.2 are C  -algebraic compact quantum groups.

Every commutative C  -algebraic compact quantum group is of the same form

as in Example 5.1.2 i):

Proposition 5.1.3. Let .A; A / be a C  -algebraic compact quantum group. Then

Š
! C.G/ of C  -bialge-
there exists a compact group G and an isomorphism ˆ W A 
bras.

Proof. By the Gelfand theorem, there exist a compact space G, a continuous map
Š
m W G  G ! G, and an isomorphism ˆ W A  ! C.G/ such that the map G WD
.ˆ ˝ ˆ/A ˆ1 is given by .G .f //.x; y/ D f .m.x; y// (here, we identify
C.G/ ˝ C.G/ with C.G  G/).
Since A is coassociative, the map m is associative and turns G into a commu-
tative semigroup. The map G  G ! G  G given by .x; y/ 7! .m.x; y/; y/ is
injective because G .C.G//.1 ˝ C.G// is linearly dense in C.G  G/ and

.f ˝ g/.m.x; y/; y/ D G .f /.x; y/g.y/ D .G .f /.1 ˝ g//.x; y/

for all x; y 2 G and f; g 2 C.G/. Likewise, the map G  G ! G  G given by

.x; y/ 7! .x; m.x; y// is injective. Therefore, G is a compact semigroup with the
cancellation property. A classical result says that every such semigroup is a group,
see [63, Theorem II.9.16] or [105, Proposition 3.2]. 

For a C  -algebraic compact quantum group, the analogue of the Haar measure
of a compact group is an invariant state. The precise definition is as follows; for
motivation, see Section 2.2.

Definition 5.1.4. Let .A; / be a C  -algebraic compact quantum group. A state h

on A is left-invariant if .idA ˝h/..a// D 1A h.a/ for all a 2 A, and right-invariant
if .h ˝ idA /..a// D 1A h.a/ for all a 2 A.
 5.1. Definition and examples 109

Remark 5.1.5. Let .A; / be a C  -algebraic compact quantum group. Then a state
h on A is left-invariant / right-invariant if and only if gh D g.1A /h / hg D hg.1A /
for each g 2 A0 . This follows from the relations g ı .idA ˝h/ ı  D g  h and
g ı .h ˝ idA / ı  D h  g.
The following result is crucial for everything that follows:
Theorem 5.1.6. Every C  -algebraic compact quantum group has a left-invariant
and right-invariant state, and every state that is left- or right-invariant coincides
with this state.
This result was first proved by Woronowicz [193, Theorem 4.2], [202, Theo-
rem 1.3] under an additional separability assumption which was later removed by
Van Daele [175]. Note that in the purely algebraic setting, an analogous statement
does not hold.
The proof depends on two lemmas. Let .A; / be a C  -algebraic compact
quantum group.
Lemma 5.1.7. For every state on A, there exists a state h on A such that h  D
 h D h.
Proof. Let be a state on A. For each k 2 N, denote by k the k-fold convolution
product of . Since the unit ball of A0 is weak--compact, the sequence .hn /n given
by hn WD . C 2 C    C .n1/ C n /=n has a weak--accumulation point, h,
say. It is easy to see that for all n 2 N,
1 .nC1/
 hn D hn  D hn C .  /:
n
Consequently,  h D h  D h. 
Lemma 5.1.8. Let h and be states on A such that h  D  h D h. If ! 2 A0
satisfies 0  !  , then h  ! D !  h D !.1A /h.
Proof. Assume that ! 2 A0 satisfies 0  !  . We show that !  h D !.1A /h,
and a similar argument shows that h  ! D !.1A /h.
Denote by Lh˝ WD fq 2 A ˝ A j .h ˝ /.q  q/ D 0g the left ideal related to
the state h ˝ , and define Lh˝! similarly. Then Lh˝  Lh˝! because !  ,
and Lh˝!  ker.h ˝ !/ by the Cauchy–Schwarz inequality. Define ‰L W A ! A
by
‰L .a/ WD h  a  1A h.a/ for all a 2 A:
We shall show that .id ˝‰L /..A//  Lh˝ . Combining the relations 1A ˝
A  Œ.A ˝ 1A /.A/ and .A ˝ 1A /Lh˝  Lh˝  ker.h ˝ !/, we then find
1A ˝ ‰L .A/  Œ.id ˝‰L /..A ˝ 1A /.A//
D Œ.A ˝ 1A /.id ˝‰L /..A//  Œ.A ˝ 1A /Lh˝   ker.h ˝ !/;
110 Chapter 5. C  -algebraic compact quantum groups

and this relation implies !  h D !.1A /h because for all a 2 A,

0 D .h ˝ !/.1A ˝ ‰L .a// D !.‰L .a//

D !.h  a/  !.1A /h.a/ D .!  h/.a/  !.1A /h.a/:

So, let us prove that .id ˝‰L /..A//  Lh˝ . Given c 2 A, put d WD h  c.
Then

q WD .id ˝‰L /..c//

D .id ˝ id ˝h/..2/ .c//  .id ˝h/..c// ˝ 1A D .d /  d ˝ 1A

and .h ˝ /.q  q/ D X  Y  Y  C Z, where

X D .h ˝ /..d  d // D .h  /.d  d /;
Y D .h ˝ /..d  ˝ 1A /.d // D h.d  .  d // D h.d  .  h  c// D h.d  d /;
Z D .h ˝ /.d  d ˝ 1A / D h.d  d /:

The assumption implies X D Z D Y D Y  , so .h ˝ /.q  q/ D 0 and q 2 Lh˝ .

The proof is finished. 
Proof of Theorem 5.1.6. For each positive functional ! on A, define K! to be the
set of states h on A that satisfy !  h D h  ! D !.1A /h. Then K!  A0 is
compact with respect to the weak--topology, and non-empty by Lemma 5.1.7.
By Lemma 5.1.8, K!1 C!2  K!1 \ K!2 for all positive functionals !1 ; !2 .
By compactness, the intersection of all K! is non-empty and contains a left- and
right-invariant state h.
Finally, if h0 is another left- or right-invariant state, then h0 D h  h0 D h or
h D h0  h D h, respectively (see Remark 5.1.5).
0

Definition 5.1.9. The unique left- and right-invariant state of a C  -algebraic com-
pact quantum group is called its Haar state and denoted by h. A C  -algebraic
compact quantum group is reduced if its Haar state is faithful.
For C  -algebraic compact quantum groups associated to groups, the Haar states
are easily identified:
Examples 5.1.10. i) Let G be a compact group. Then theRHaar state of the C  -alge-
braic compact quantum group C.G/ is given by f 7! G f d , where  denotes
the normalized Haar measure of G. Evidently, this Haar state is faithful.
ii) Let G be a discrete group. Then the Haar states of the C  -algebraic compact
quantum groups C  .G/ and Cr .G/ are given by Ux 7! ıx;e and Lx 7! ıx;e for all
x 2 G, respectively, where e 2 G denotes the unit, and .Ux /x and .Lx /x denote
the canonical generators as in Example 4.2.2. Note that these Haar states are tracial

in the sense that h.ab/ D h.ba/ for all a; b 2 C.r/ .G/.
 5.2. Corepresentations of C  -bialgebras 111

Let us show that the Haar state of Cr .G/ is faithful. In terms of the standard
basis ."x /x2G of l 2 .G/, it is given by T 7! h"e jT "e i. The map Cr .G/ ! l 2 .G/
given by T 7! T "e is injective because T "y D TRy 1 "e D Ry 1 T "e for all
y 2 G, where Ry 1 2 L.l 2 .G// denotes the right shift "x 7! "xy . Therefore,
h"e jT  T "e i D hT "e jT "e i ¤ 0 whenever T ¤ 0.
Since the Haar state of C  .G/ factorizes through the quotient map L W C  .G/ !

Cr .G/, it is faithful if and only if this quotient map is faithful, that is, an isomor-
phism. This happens if and only if G is amenable [119, Theorem 4.21], [121,
Theorem 7.3.9].

5.2 Corepresentations of C  -bialgebras

Like corepresentations of Hopf -algebras, corepresentations of C  -bialgebras can
be described in several equivalent ways:
• Unitary corepresentations can elegantly be described in the language of C  -
modules. To simplify the presentation, we explain this approach for C  -al-
gebraic compact quantum groups only instead of general C  -bialgebras.
• Non-unitary corepresentations are more conveniently studied in terms of co-
representation operators or corepresentation matrices. We explain the con-
nection to the unitary corepresentations mentioned above but consider general
C  -bialgebras.
As in the setting of Hopf -algebras, one can associate to every corepresentation
of a C  -bialgebra a representation of the dual algebra, construct new corepresen-
tations out of given ones, and associate to every C  -algebraic compact quantum
group a particular regular corepresentation. We present the pertaining definitions
and constructions, frequently referring to Section 3.1 for motivation.
We shall use the notation and definitions introduced in Section 4.1, the slice
maps discussed in Section 12.4, and the background on C  -modules summarized
in Section 12.2. As before, we use the symbol “ˇ” to denote algebraic tensor
products and the symbol “˝” to denote minimal tensor products of C  -algebras
and completed internal or external tensor products of C  -modules.

5.2.1 Unitary corepresentations of C  -algebraic compact quantum

groups
Recall that a unitary corepresentation of an algebraic compact quantum group
.A0 ; 0 / on a finite-dimensional Hilbert space H0 is a linear map ı0 W H0 !
H0 ˇ A0 that satisfies

.ı0 ˇ id/ ı ı0 D .id ˇ 0 / ı ı0

112 Chapter 5. C  -algebraic compact quantum groups

and
hı0 ./jı0 ./iA0 D hji  1A0 for all ;  2 H0 :
Here, the A0 -valued inner product h  j  iA0 on H0 ˇ A0 was given by h ˇ bj ˇ
aiA0 D hjib  a for all ;  2 H0 and a; b 2 A0 .
This definition can be adapted to the present setting as follows. Let .A; / be
a C  -algebraic compact quantum group and H a Hilbert space.
• We consider H and A as C  -modules over C and A, respectively, and form
the tensor product of C  -modules H ˝ A (see Section 12.2). This is a C  -
module over A. It is a completion of H ˇ A; its structure maps are given
by

h ˝ bj ˝ ai D hjib  a; . ˝ b/a D  ˝ ba

for all ;  2 H , a; b 2 A. Similarly, we define a C  -module H ˝ A ˝ A

over A ˝ A.
• By definition, the map idH ˇ  W H ˇ A ! H ˇ .A ˝ A/ satisfies

h.idH ˇ /.y/j.idH ˇ /.x/i D .hyjxi/ for all x; y 2 H ˇ A

and extends to an isometric linear map idH ˝ W H ˝ A ! H ˝ .A ˝ A/.

• If ı W H ! H ˝ A is a linear map that satisfies hı./jı./i D hji  1A for
all ;  2 H , then the map ı ˇ idA W H ˇ A ! .H ˝ A/ ˇ A satisfies

h.ı ˇ idA /.y/j.ı ˇ idA /.x/i D 1A ˝ hyjxi for all x; y 2 H ˇ A

and extends to an isometric linear map ı ˝ idA W H ˝ A ! H ˝ A ˝ A.

Definition 5.2.1. A unitary corepresentation of a C  -algebraic compact quantum
group .A; / on a Hilbert space H is a linear map ı W H ! H ˝ A that satisfies
the following conditions:
i) hı./jı./i D hji  1A for all ;  2 H ;
ii) the set ı.H /A is linearly dense in H ˝ A;
iii) .idH ˝/ ı ı D .ı ˝ idA / ı ı, that is, the following diagram commutes:

H
ı /H ˝A
ı ı˝id
 
H ˝A / H ˝ A ˝ A.
id ˝
 5.2. Corepresentations of C  -bialgebras 113

One can extend this definition to the case of a non-unital C  -bialgebra .A; /
by replacing the C  -module H ˝ A by the multiplier C  -module M.H ˝ A/ D
L.A; H ˝A/ and considering linear maps ı W H ! M.H ˝A/ that satisfy obvious
analogues of the conditions i)–iii) above. Moreover, one can consider intertwiners,
invariant subspaces, and irreducibility for unitary corepresentations, very much like
in the setting of Hopf -algebras. However, it turns out to be more convenient to
work with corepresentation operators instead of unitary corepresentations.

5.2.2 Corepresentation operators of C  -bialgebras

Before we can define corepresentation operators of a general C  -bialgebra, we need
to collect some preliminaries and fix some notation. Let .A; / be a C  -bialgebra
and H a Hilbert space.

• The C  -algebra K.H / ˝ A embeds naturally in LA .H ˝ A/, and the image

of this embedding is KA .H ˝ A/ [69, Lemma 1.2.7]. Thus

LA .H ˝ A/ Š M.KA .H ˝ A// Š M.K.H / ˝ A/: (5.1)

From now on, we shall use these identifications without further notice.

• Since  is non-degenerate, the -homomorphism

id ˝ W K.H / ˝ A ! K.H / ˝ M.A ˝ A/ ,! M.K.H / ˝ A ˝ A/

(5.2)

is non-degenerate and extends to the multiplier algebra (5.1).

• If .A; / is a C  -algebraic compact quantum group, then the -homomor-

phism (5.2) can be expressed in terms of the map idH ˝ W H ˝ A !
H ˝ A ˝ A: with respect to the natural isomorphisms M.K.H / ˝ A/ Š
LA .H ˝ A/ and M.K.H / ˝ A ˝ A/ Š LA˝A .H ˝ A ˝ A/,

..id ˝/.X //. ˝ 1.A˝A/ / D .idH ˝/.X. ˝ 1A // (5.3)

for all  2 H and X 2 LA .H ˝ A/.

• Given X 2 M.K.H / ˝ A/, we define XŒ12 ; XŒ13 2 M.K.H / ˝ A ˝ A/

by

XŒ12 WD X ˝ 1 and XŒ13 WD .id ˝†/.XŒ12 /;

where † W A ˝ A ! A ˝ A denotes the flip as usual. This is a particular

example of the leg notation that will be used extensively in Chapter 7.
114 Chapter 5. C  -algebraic compact quantum groups

Proposition 5.2.2. Let .A; / be a C  -algebraic compact quantum group and let
H be a Hilbert space.
i) Let ı be a unitary corepresentation of .A; / on H . Then the map

X W H ˇ A ! H ˝ A;  ˇ a 7! ı./a;

extends to a unitary operator X 2 LA .H ˝A/, and XŒ12 XŒ13 D .id ˝/.X /.

ii) If a unitary X 2 LA .H ˝ A/ satisfies XŒ12 XŒ13 D .id ˝/.X /, the map

ı W H ! H ˝ A;  7! X. ˝ 1A /;

is a unitary corepresentation.
Proof. i) For all elementary tensors y D  ˝ b and x D  ˝ a in H ˇ A,

hX.y/jX.x/i D hı./bjı./ai D b  hı./jı./ia D b  hji1A a D hyjxi:

Thus X extends to an isometric linear map H ˝ A ! H ˝ A. Since X.H ˇ A/ D

span ı.H /A is dense in H ˝ A, the extension is unitary. Finally, the relation
.ı ˝ idA / ı ı D .idH ˝/ ı ı implies

XŒ12 XŒ13 . ˝ 1.A˝A/ / D .ı ˝ idA /.ı.//

D .idH ˝/.ı.//
(5.4)
D .idH ˝/.X. ˝ 1A //
D ..id ˝/.X //. ˝ 1.A˝A/ /

for all  2 H , whence XŒ12 XŒ13 D .id ˝/.X /.

ii) Since X is unitary,

hı./jı./i D hX. ˝ 1A /jX. ˝ 1A /i D h ˝ 1A j ˝ 1A i D hji  1A

for all ;  2 H , and

Œı.H /A D Œ.X.H ˝ 1A //A D X.H ˝ A/ D H ˝ A:

Thus the map ı satisfies conditions i) and ii) of Definition 5.2.1. A similar calculation
like (5.4) shows that ı also satisfies condition iii). 
Propositions 5.2.2 and 3.1.7 motivate the following definition:
Definition 5.2.3. Let .A; / be a C  -bialgebra. A .unitary/ corepresentation
operator of .A; / on a Hilbert space H is an invertible (unitary) operator X 2
M.K.H / ˝ A/ Š LA .H ˝ A/ that satisfies XŒ12 XŒ13 D .id ˝/.X /. If H has
finite dimension, we call X finite-dimensional.
 5.2. Corepresentations of C  -bialgebras 115

Let X and Y be corepresentation operators on Hilbert spaces H and K, respec-

tively.
An intertwiner from X to Y is an operator T 2 L.H; K/ that satisfies the
equation Y.T ˝ idA / D .T ˝ idA /X . We denote the space of all intertwiners from
X to Y by Hom.X; Y /. We call X and Y equivalent, written X ' Y , if X and Y
admit an invertible intertwiner.
A subspace L  H is invariant for X if the orthogonal projection pL W H !
L  H satisfies X.pL ˝ 1A / D .pL ˝ 1A /X.pL ˝ 1A /. The corepresentation
operator X is irreducible if there exists no invariant subspace besides 0 and H .
The space of matrix elements of X is the space
C .X / WD span f.!; ˝ id/.X / j ;  2 H g  M.A/;
where !; W K.H / ! C is given by T 7! hjT i and the slice map !; ˝ id is
defined as in Section 12.4.
Remarks 5.2.4. i) For every C  -algebraic compact quantum group .A; / and
every Hilbert space H , there exists a bijective correspondence between unitary co-
representations and unitary corepresentation operators of .A; / on H , see Propo-
sition 5.2.2.
ii) The equation Y .T ˝ 1A / D .T ˝ 1A /X figuring in the definition of an
intertwiner should be considered as an equation in LA .H ˝ A; K ˝ A/ or in
M.K.H ˚ K/ ˝ A/.
iii) If H is a finite-dimensional Hilbert space and A is unital, then we have
M.K.H /˝A/ D K.H /˝A D L.H /˝A. Generally, the space L.H /˝M.A/ Š
M.K.H // ˝ M.A/ embeds in M.K.H / ˝ A/ as a strict subspace. For an illus-
tration of the difference between corepresentation operators in M.K.H // ˝ M.A/
and corepresentation operators in M.K.H / ˝ A/, see the end of Example 5.2.5.
Corepresentations are related to group representations as follows:
Example 5.2.5. Let G be a locally compact group. Then corepresentation operators
of the C  -bialgebra C0 .G/ (see Example 4.2.1) on a Hilbert space H correspond
bijectively with strictly continuous representations of G on H :
The C  -algebra M.K.H / ˝ C0 .G// can be identified with the C  -algebra of
norm-bounded and strictly continuous functions G ! M.K.H // Š L.H / [2,
Corollary 3.4]. Explicitly, each multiplier X 2 M.K.H / ˝ C0 .G// corresponds
to the function

W G ! M.K.H // Š L.H /; x 7! .id ˝ evx /.X /;
where evx W C0 .G/ ! C denotes evaluation at x 2 G as usual. The relations

.x/
.y/ D .id ˝ evx /.X /  .id ˝ evy /.X / D .id ˝ evx ˝ evy /.XŒ12 XŒ13 /;

.xy/ D .id ˝ evxy /.X / D .id ˝ evx ˝ evy /..id ˝/.X //; x; y 2 G;
116 Chapter 5. C  -algebraic compact quantum groups

show that X is a corepresentation operator if and only if

is a representation.
Evidently, X is unitary if and only if
is unitary. Furthermore, a subspace K  H
is invariant for X if and only if it is invariant for
; in particular, X is irreducible
if and only if
is irreducible.
Note that every strictly continuous representation is strongly continuous. Con-
versely, it is easy to see that every strongly continuous unitary representation is
strictly continuous.
Finally, let G be compact and let X 2 M.K.H /˝C.G// be a corepresentation
operator. Then X is contained in L.H / ˝ C.G/ if and only if the associated
representation
is norm-continuous, because L.H / ˝ C.G/ corresponds to the
C  -algebra of norm-continuous functions G ! L.H /.

5.2.3 Constructions related to corepresentation operators

In this section, we introduce several constructions related to corepresentation oper-
ators that are analogues of the purely algebraic constructions for corepresentations
of Hopf -algebras discussed in Section 3.1.2 and 3.1.3.
Throughout this section, let .A; / be a C  -bialgebra.

Corepresentation matrices. Let X be a corepresentation operator of .A; / on

an n-dimensional Hilbert space H with orthonormal basis .ei /niD1 , where n 2 N.
Note that then X 2 M.K.H / ˝ A/ D K.H / ˝ M.A/. The choice of the basis
.ei /i defines a natural identification H Š Cn and an isomorphism
K.H / ˝ M.A/ Š K.Cn / ˝ M.A/ D Mn .C/ ˝ M.A/ Š Mn .M.A//: (5.5)
With respect to this isomorphism, X corresponds to the matrix
a D .aij /i;j 2 Mn .M.A//; where aij D .!ei ;ej ˝ id/.X / for all i; j:
As in the setting of Hopf -algebras (Section 3.1), we find
X
.aij / D aik ˝ akj for all i; j; (5.6)
k
C .X / D span faij j i; j g; (5.7)
and X is unitary if and only if the matrix a is unitary, that is, if
X X
aki  akj D ıi;j  1A D aik aj k  for all i; j: (5.8)
k k

Definition 5.2.6. A corepresentation matrix of .A; / is an invertible matrix a 2

Mn .M.A//, where n 2 N, that satisfies
P equation (5.6). The matrix a is irreducible
if the corepresentation operator i;j eij ˝ aij on Cn is irreducible, where .eij /i;j
denotes the standard matrix units.
 5.2. Corepresentations of C  -bialgebras 117

Representation of the dual algebra. To every corepresentation operator of

.A; /, one can associate a representation of the dual algebra A0 . The definition of
this representation involves slice maps, which are discussed in Section 12.4.

Proposition 5.2.7. i) Let X be a corepresentation operator of .A; / on a Hilbert

space H . Then the map

X W A0 ! M.K.H // Š L.H /; f 7! .id ˝f /.X /; (5.9)

is a homomorphism, that is,

X .f  g/ D
X .f /
X .g/ for all f; g 2 A0 .
ii) Let X, H and
X be as above and let K be a subspace of H . Then K is
invariant for X if and only if it is invariant for
X .
iii) Let X and Y be corepresentation operators of .A; / on Hilbert spaces H
and K, respectively, and let T 2 L.H; K/. Then T 2 Hom.X; Y / if and only if
T
X .f / D
Y .f /T for all f 2 A0 .

Proof. i) The proof consists of a calculation with slice maps; comments on such
calculations can be found in Section 12.4. For all f; g 2 A0 ,

.id ˝f /.X /  .id ˝g/.X / D .id ˝f ˝ g/.XŒ12 XŒ13 /

D .id ˝f ˝ g/..id ˝/.X // D .id ˝fg/.X /:

ii), iii) The proofs are similar to the proofs of the equivalences (f4),(f5) and
(d2),(d4) in Proposition 3.1.7. 

In general, the algebra A0 is not equipped with a natural involution: the formula
f  .a/ WD f .S.a/ / known from the setting of Hopf -algebras makes no sense
if .A; / has no antipode S. Hence we can not ask whether the representation
X
associated to a corepresentation operator X is a -homomorphism. But if .A; /
is a C  -algebraic compact quantum group and X is unitary, then
.A0 / contains a
useful non-degenerate C  -subalgebra:

Proposition 5.2.8. Let .A; / be a C  -algebraic compact quantum group with

Haar state h, and let X be a corepresentation operator of .A; / on a Hilbert
space H . Put A0h WD fh.  a/ j a 2 Ag  A0 .

i) The space CX WD Œ
X .A0h /  L.H / is a non-degenerate C  -algebra and
X 2 M.CX ˝ A/.

ii) A subspace K  H is invariant for X if and only if it is invariant for CX .

iii) If p 2 L.H / is a projection and X.p ˝ 1A / D .p ˝ 1A /X.p ˝ 1A /, then

.p ˝ 1A /X D X.p ˝ 1A /.
118 Chapter 5. C  -algebraic compact quantum groups

Proof. i) Writing out the definition, we find

CX D span f.id ˝h/.X.idH ˝a// j a 2 Ag:

First, we show that ŒCX CX  D CX ; this implies that CX is a C  -algebra. By

left-invariance of h,

CX ˝ 1A D Œ.id ˝ id ˝h/..id ˝/.X.idH ˝A///;

and hence

ŒCX CX  D Œ.id ˝h/..idH ˝A /X  .CX ˝ 1A //

D Œ.id ˝h ˝ h/..idH ˝A ˝ 1A / .X  ˝ 1A / ..id ˝/.X // .idH ˝.A///:

We insert the relation .X  ˝ 1A /..id ˝/.X // D XŒ12 XŒ12 XŒ13 D XŒ13 , move

idH ˝A ˝ 1A to the right of XŒ13 , and find

ŒCX CX  D Œ.id ˝h ˝ h/.XŒ13 .idH ˝A ˝ 1A / .idH ˝.A///:

Since Œ.A ˝ 1A /.A/ D A ˝ A,

ŒCX CX  D Œ.id ˝h ˝ h/.XŒ13 .idH ˝A ˝ A// D Œ.id ˝h/.X.idH ˝A// D CX :

Put K WD K.H /. Since K ˝ A  M.K ˝ A/ is an ideal and X 2 M.K ˝ A/

is unitary, X.K ˝ A/ D K ˝ A and

ŒCX K D Œ.id ˝h/.X.K ˝ A// D Œ.id ˝h/.K ˝ A/ D K:

Consequently, CX is non-degenerate.
Let us prove that X belongs to M.CX ˝ A/. By definition of CX ,

ŒX.CX ˝ A/ D Œ.id ˝ id ˝h/.XŒ12 XŒ13 .idH ˝A ˝ A//:

We replace XŒ12 XŒ13 by .id ˝/.X / and A ˝ A by Œ.A/.A ˝ 1A /, and find

ŒX.CX ˝ A/ D Œ.id ˝ id ˝h/..id ˝/.X.idH ˝A///  .idH ˝A/:

Since h is left-invariant, this is equal to

Œ..id ˝h/.X.idH ˝A// ˝ 1A /  .idH ˝A/ D CX ˝ A:

Similarly, one shows that .CX ˝ A/X  CX ˝ A. Thus X 2 M.CX ˝ A/.

ii), iii) Let K  H be some subspace, and denote by p the orthogonal projection
onto K. If .p ˝ 1A /X.p ˝ 1A / D X.p ˝ 1A /, then pcp D cp for all c 2 CX , and
since CX is self-adjoint, also pc D cp for all c 2 CX . Conversely, if cp D pc for
all c 2 CX , then also X.p ˝ 1A / D .p ˝ 1A /X because X 2 M.CX ˝ A/. 
 5.2. Corepresentations of C  -bialgebras 119

The regular corepresentation. To every C  -algebraic compact quantum group,

one can associate a regular corepresentation. This corepresentation plays an im-
portant rôle in the characterization of compact quantum groups.
To construct this corepresentation, we need to fix some notation and collect
some preliminaries. Let .A; / be a C  -algebraic compact quantum group with
Haar state h and associated GNS-representation .Hh ; ƒh ;
h / (see Section 12.1).
Then the map ƒh ˇ idA W A ˇ A ! Hh ˇ A extends to a continuous linear map
ƒh ˝ idA W A ˝ A ! Hh ˝ A such that
.ƒh ˝ idA /.x/ D .
h ˝ idA /.x/ .ƒh .1A / ˝ 1A /;
and
h.ƒh ˝ idA /.x/j.ƒh ˝ idA /.y/i D .h ˝ idA /.x  y/ for all x; y 2 A ˝ A:
Theorem 5.2.9. Let .A; / be a C  -algebraic compact quantum group with Haar
state h and associated GNS-representation .Hh ; ƒh ;
h /.
i) The map ıh W Hh ! Hh ˝ A given by ƒh .a/ 7! .ƒh ˝ idA /..a// is a
unitary corepresentation.
Denote by Xh 2 M.K.Hh / ˝ A/ the corepresentation operator corresponding to
ıh as in Proposition 5.2.2.
ii) Xh .
h .a/ ˝ 1A / D ..
h ˝ idA /..a///Xh for all a 2 A.
iii) C .Xh / D A.
Proof. i) The map ıh W ƒh .A/ ! Hh ˝ A given by ƒh .a/ 7! .ƒh ˝ id/..a//
extends to Hh and satisfies condition 5.2.1 i) because h is right-invariant:
hıh .ƒh .b//jıh .ƒh .a//i D .h ˝ id/..b  /.a// D h.b  a/1A for all a; b 2 A:
The map ıh satisfies condition 5.2.1 ii) because .A/.1 ˝ A/ is dense in A ˝ A:
Œıh .ƒh .A//A D Œ.ƒh ˝ id/..A//  A
D Œ.ƒh ˝ id/..A/.1 ˝ A// D Œƒh .A/ ˝ AI
and condition 5.2.1 iii) because  is coassociative:
.idHh ˝/ ı ıh ı ƒh D .ƒh ˝ id ˝ id/ ı .id ˝/ ı 
D .ƒh ˝ id ˝ id/ ı . ˝ id/ ı  D .ıh ˝ id/ ı ıh ı ƒh :
ii) By definition of Xh , we have for all a; b 2 A
Xh .
h .a/ƒh .b/ ˝ 1A / D ıh .ƒh .ab// D ..
h ˝ id/..a///ıh .ƒh .b//
D ..
h ˝ id/..a///Xh .ƒh .b/ ˝ 1A /:
iii) Since Xh .ƒh .A/ ˝ 1A / D ıh .ƒh .A// and A is bisimplifiable,
C.Xh / D Œ.h ˝ id/..A ˝ 1/.A// D Œ.h ˝ id/.A ˝ A/ D A: 
120 Chapter 5. C  -algebraic compact quantum groups

Definition 5.2.10. The regular corepresentation/regular corepresentation operator

of a C  -algebraic compact quantum group .A; / is the corepresentation ıh /the
corepresentation operator Xh defined in Theorem 5.2.9.

Construction of new corepresentation operators. The corepresentations of ev-

ery C  -bialgebra admit a direct sum and a tensor product which turn the category
of all corepresentations into a monoidal/tensor category [79], [104] and, more pre-
cisely, into a concrete monoidal/tensor W  -category [195] and strict monoidal/ten-
sor C  -category [102]. Corepresentations of C  -algebraic compact quantum groups
additionally admit a conjugation. All these constructions are similar to the corre-
sponding constructions for corepresentations of Hopf -algebras (see Section 3.1.3);
therefore we only give a brief summary.

Direct sum. Let .X˛ /˛ be a family of corepresentation operators on Hilbert spaces

.H˛ /˛ . Then there exists a unique corepresentation operator
M

˛
X˛ 2 M K
˛
H˛ ˝ A ;

L the direct sum of .X˛ /˛ , such that for each ˇ, the natural inclusion ˇ W Hˇ ,!
called
˛ H˛ is an intertwiner from Xˇ to ˛ X˛ .
The direct sum construction is functorial: For every second family of corepre-
sentation operators .Y˛ /˛ on Hilbert spaces .K˛ /˛ , there exists a natural map
Y
M

˛
Hom.X˛ ; Y˛ / ! Hom  
˛
X˛ ;
˛
Y˛ ; .T˛ /˛ 7!
˛
T˛ :

Clearly, ˛ X˛ is unitary if each X˛ is unitary, and

[
C ˛
X˛ D span C .X˛ /:
˛

Tensor product. Let X and Y be corepresentation operators on Hilbert spaces H

and K, respectively. Then the operator
X  Y WD XŒ13 YŒ23 2 M.K.H / ˝ K.K/ ˝ A/ D M.K.H ˝ K/ ˝ A/
is a corepresentation operator again, called the tensor product of X and Y . Here,
YŒ23 D idH ˝Y and XŒ13 D .† ˝ idA /.idK ˝X /, where † W K.K/ ˝ K.H / !
K.H / ˝ K.K/ denotes the flip as usual.
The tensor product construction is functorial: for every second pair of corepre-
sentation operators Xz and Yz , there exists a map
Hom.X; Xz /  Hom.Y; Yz / ! Hom.X  Y; Xz  Yz /; .S; T / 7! S ˝ T:
 5.3. Corepresentation theory and structure theory 121

It is easy to see that X  Y is unitary if X and Y are, and that

C .X  Y / D span C.X /C .Y /:

If A is commutative, then the natural isomorphism H ˝K Š K ˝H intertwines

X  Y and Y  X ; in particular,

X  Y ' Y  X:

If A is not commutative, this relation need not hold. However, there may exist a
braiding for .A; /, that is, a natural equivalence between the bifunctors .X; Y / 7!
X  Y and .X; Y / 7! Y  X that satisfies some additional coherence properties
(see [79]).

Conjugation. Let X be a corepresentation operator on a finite-dimensional Hilbert

space H . Denote by H x the conjugate Hilbert space of H , and by  7! N the canon-
Š
ical conjugate-linear isomorphism H  ! H x ; thus, N D N and hj N D hji
N i
for all ;  2 H and  2 C. For each T 2 L.H /, the map N 7! T  defines an
operator Tx 2 L.H x /, and the map T 7! Tx is a conjugate-linear -isomorphism
L.H / ! L.H /. x
Denote by j W L.H / ˝ M.A/ ! L.H x / ˝ M.A/ the map given by T ˝ S 7!
x  x
T ˝ S , and consider the operator X WD j.X / 2 L.H x / ˝ M.A/. If
X
XD jei ihej j ˝ aij ;
i;j

where .ei /i is some orthonormal basis of H , then

X
Xx D jS ej j ˝ aij  :
ei ihS
i;j

Since  is a -homomorphism, X x satisfies X

x Œ12X
x Œ13 D .id ˝/.X x /. It is not
immediately clear whether the operator X x is invertible. We shall prove that if
.A; / is a C  -algebraic compact quantum group, then X x is invertible and hence a
corepresentation operator (see Corollary 5.3.10).
x / as before, and then C.X
Of course, we can still define C .X x / D C .X / .

5.3 Corepresentation theory and structure theory

The corepresentation theory of C  -algebraic compact quantum groups is very simi-
lar to the representation theory of compact groups and to the corepresentation theory
of algebraic compact quantum groups:
122 Chapter 5. C  -algebraic compact quantum groups

1. Every corepresentation operator is equivalent to a direct sum of irreducible

finite-dimensional unitary corepresentation operators (Theorem 5.3.3).
2. Two irreducible corepresentations either do not admit a non-zero intertwiner,
or they are equivalent and the space of intertwiners has dimension one (Propo-
sition 5.3.4).
3. The subspace spanned by the matrix elements of irreducible corepresentations
is dense in the quantum group, and these matrix elements satisfy certain
orthogonality relations with respect to the Haar state (Propositions 5.3.7,
5.3.8).
4. Moreover, C  -algebraic compact quantum groups can be characterized in
terms of corepresentations (Theorem 5.3.11).
Several of these results follow from similar arguments as in the setting of algebraic
compact quantum groups (see Section 3.2), but sometimes, we have to use new
methods and ideas.
Throughout this section, let .A; / be a C  -algebraic compact quantum group
with Haar state h.

5.3.1 Decomposition into irreducible corepresentations

In this subsection, we show that corepresentation operators and intertwiners between
them can be described in terms of irreducible corepresentation operators.
As a tool, we first construct intertwiners between corepresentation operators
by an averaging procedure (cf. Lemma 3.2.5). Let X and Y be corepresentation
operators on Hilbert spaces H and K, respectively, and let R 2 L.H; K/. We
identify M.K.H / ˝ A/; M.K.K/ ˝ A/, and L.H; K/ ˝ A with subspaces of
M.K.H ˚ K/ ˝ A/, and use the slice map

id ˝h W M.K.H ˚ K/ ˝ A/ ! M.K.H ˚ K// Š L.H ˚ K/

(see Section 12.4) to define operators S; T 2 L.H; K/ as follows:

S WD .id ˝h/.Y 1 .R ˝ 1A /X /; T WD .id ˝h/.Y .R ˝ 1A /X 1 /: (5.10)

Lemma 5.3.1. Let X and Y be corepresentations operators on Hilbert spaces H

and K, respectively, let R 2 L.H; K/, and define S; T as above. Then S; T 2
Hom.X; Y /. If R 2 Hom.X; Y /, then S D R D T . If R is compact, so are S
and T .
Proof. The first statements follows from similar calculations as in Lemma 3.2.5. If
R is compact, then Y 1 .R ˝ 1A /X and Y .R ˝ 1A /X 1 belong to K.H; K/ ˝ A 
K.H ˚ K/ ˝ A, so that S and T are compact as well. 
 5.3. Corepresentation theory and structure theory 123

Remark 5.3.2. We can equip L.H; K/ with a strict topology by identifying it with
a subspace of M.K.H ˚ K//. Then the assignments R 7! S and R 7! T defined
in (5.10), considered as maps L.H; K/ ! L.H; K/, are strictly continuous on
bounded subsets: The assignment R 7! R ˝ 1A is strictly continuous on bounded
subsets, and left and right multiplication by Y , Y 1 , X , or X 1 , respectively, as
well as the slice map id ˝h are strictly continuous.

Every corepresentation operator is a direct sum of finite-dimensional irreducible

ones (cf. Theorem 3.2.1):

Theorem 5.3.3. Let X be a corepresentation operator on a Hilbert space H .

i) X is equivalent to a unitary corepresentation operator.

ii) If X is unitary, then the orthogonal complement of every invariant subspace
of H is invariant again.
iii) If H is non-zero, then it contains a non-zero invariant subspace of finite
dimension. In particular, X is finite-dimensional if it is irreducible.
iv) X is equivalent to a direct sum of finite-dimensional irreducible unitary co-
representation operators.

Proof. i) Put T WD .id ˝h/.X  X / 2 L.H /. Then T is invertible: X  X >  1

for some > 0 because X and hence also X  X are invertible, and then T >  1
because id ˝ h is positive. Consider the operator

Y WD .T 1=2 ˝ 1A /X.T 1=2 ˝ 1A /:

This is a corepresentation operator which is equivalent to X . Furthermore,

X  .T ˝ 1A /X D .id ˝ h ˝ id/.XŒ13
 
XŒ12 XŒ12 XŒ13 /
D .id ˝ h ˝ id/..id ˝ /.X  X //
D .id ˝ h/.X  X / ˝ 1A D T ˝ 1A ;

so Y  Y D .T 1=2 ˝ idA /X  .T ˝ idA /X.T 1=2 ˝ idA / D 1, that is, Y is unitary.

ii) This claim follows from the fact that a subspace is invariant if and only if it
is invariant for the C  -subalgebra of L.H / defined in Proposition 5.2.8.
iii) By i), we may assume that X is unitary. Let .u˛ /˛ be an approximate unit for
K.H /. Since conjugation by X is an automorphism of K.H / ˝ A, the elements
X.u˛ ˝ 1A /X  form an approximate unit for K.H / ˝ A. By Lemma 5.3.1, the
elements .id ˝ h/.X.u˛ ˝ 1A /X  / belong to the C  -algebra

D WD K.H / \ Hom.X; X /;
124 Chapter 5. C  -algebraic compact quantum groups

and since id ˝h is continuous on K.H / ˝ A, these elements form an approximate

unit for K.H /. Therefore, D  K.H / is non-degenerate and contains a non-zero
projection P . Now PH  H is a finite-dimensional invariant subspace.
iv) This follows from i)–iii) by a straightforward application of Zorn’s Lemma.
Alternatively, one could choose a maximal family .p˛L /˛ of pairwise orthogonal

minimal projections in the C -algebra D; then H Š ˛ p˛ H since D is non-
degenerate, and the restriction of X to p˛ H is irreducible for each ˛ because p˛ H
is a minimal invariant subspace. 

We have the following analogue of Schur’s Lemma (cf. Proposition 3.2.2):

Proposition 5.3.4. Let X and Y be corepresentation operators on Hilbert spaces

H and K, respectively.

i) For every T 2 Hom.X; Y /, the subspaces ker T  H and Im T  K are

invariant.

ii) If X and Y are irreducible, then either Hom.X; Y / has dimension 1 and
X ' Y , or Hom.X; Y / D 0.

Proof. The proof is essentially the same as the proof of Proposition 3.2.2, the only
difference is that for i), we use the representations
X and
Y associated to X
and Y , respectively (see Proposition 5.2.7): If T 2 Hom.X; Y /, then T
X .A0 / D

Y .A0 /T , whence ker T  H is invariant for

X .A0 / and X , and Im T  K is
invariant for
Y .A0 / and Y . 

As in the setting of algebraic compact quantum groups, the preceding results

show that the category of all corepresentation operators of .A; /, regarded as an
ordinary category, has a very simple structure (cf. Corollary 3.2.4):

Corollary 5.3.5. Let .X˛ /˛2I be a maximal family of pairwise inequivalent irre-
ducible corepresentations operators of .A; /.

i) Every corepresentation operator of .A; / is equivalent to a direct sum

k X˛k , where .˛k /k is some family of elements of I .
ii) Let .˛k /k and .ˇl /l be families of elements of I . Then

Y
Y M
Hom k
X˛k ; l
Xˇl Š
k
Hom X˛k ; l
Xˇl Š
k
C;
l s.t.
˛k Dˇl

Q L
where and denote the direct product and the l 2 -sum of Banach spaces,
respectively. 
 5.3. Corepresentation theory and structure theory 125

The interesting and non-trivial information of the category of all corepresenta-

tion operators is contained in the monoidal structure, that is, in the tensor product.
Finally, we show that every irreducible corepresentation is contained in the
regular corepresentation:
Proposition 5.3.6. Let X be an irreducible corepresentation operator. Then there
exists an injective intertwiner T 2 Hom.X; Xh /, where Xh denotes the regular
corepresentation operator (see Theorem 5.2.9).
Proof. Denote by H the underlying Hilbert space of X , and by .aij /i;j the corepre-
sentation matrix of X with respect to some orthonormal basis .ei /i . For a fixed i ,
consider the map Ti W H ! Hh given by ej 7! ƒh .aij / for all j . The calculations
X
Xh .Ti ˝ 1A /.ej ˝ 1A / D Xh .ƒh .aij / ˝ 1A / D ƒh .aik / ˝ akj ;
k
X X
.Ti ˝ 1A /X.ej ˝ 1A / D .Ti ˝ 1A /.ek ˝ akj / D ƒh .aik / ˝ akj
k k

show that TP i 2 Hom.X; Xh /. By Theorem 5.3.3, we may assume that X is uni-

tary. Then i aij  aij D 1A for each j , and hence there exist i0 and j such that
h.ai0 j  ai0 j / ¤ 0, that is, Ti0 ej D ƒh .ai0 j / ¤ 0. In particular, Ti0 is non-zero.
By Proposition 5.3.4, Ti0 must be injective. 

5.3.2 Schur’s orthogonality relations

In this subsection, we study the matrix elements of irreducible corepresentation
operators and prove an analogue of Schur’s orthogonality relations known from the
representation theory of compact groups [22, Chapter II, Theorems 4.5, 4.6], [62,
Chapter VII, Theorem 27.19], and from the corepresentation theory of algebraic
compact quantum groups (cf. Section 3.2.2).
Proposition 5.3.7. Let X and Y be inequivalent irreducible corepresentation op-
erators. Then h.b  a/ D 0 D h.ba / for all a 2 C.X / and b 2 C.Y /.
Proof. The proof of Proposition 3.2.6 carries over without modifications. 
To obtain a better understanding of expressions of the form h.b  a/, where b
and a are matrix elements of the same irreducible corepresentation operator, we
need to consider particular intertwiners (cf. Proposition 3.2.9).
P
Proposition 5.3.8. Let Y D i;j eij ˝ aij be an irreducible unitary corepresen-
tation operator on Cn , where n 2 N and .eij /i;j 2 Mn .C/ denote the standard
matrix units.
i) Yx and Yx  WD .Yx  /1 are irreducible corepresentation operators on Cn .
126 Chapter 5. C  -algebraic compact quantum groups

ii) There exists a unique intertwiner Fz 2 Hom.Yx ; Yx  / that satisfies Tr Fz D

Tr Fz 1 > 0, and this operator is positive and invertible.
iii) For all i; j; k; l,
ıj;l ıi;k z
h.aij  akl / D .Fz 1 /ik and h.aij akl  / D Fj l ;
Tr.Fz 1 / Tr Fz

where the matrix representations of Fz and Fz 1 are taken with respect to the
standard basis of Cn .
iv) The elements .aij /i;j are linearly independent.
Proof. i) First, we modify the proof of Proposition 5.3.6 and show that Yx is equiva-
lent to the restriction of the regular corepresentation operator Xh (see Theorem 5.2.9)
to an invariant subspace. This implies that Yx is a corepresentation operator.
Denote by .ei /i the standard units of Cn . Fix some i , and consider the map

T W Cn ! Hh ;eSj 7! ƒh .aij  /:
P
Equation (5.6) on page 116 shows that .aij  / D ij aik  ˝ akj  for all i; j , and
as in the proof of Proposition 5.3.6, we deduce

Xh .T ˝ 1A / D .T ˝ 1A /Yx : (5.11)
P
Since Y is unitary, j aij aij  D 1A , and hence h.aij0 aij0  / ¤ 0 for some j0 .
So, T ej0 D ƒh .aij0  / ¤ 0 and T ¤ 0. We show that T is injective and that its
image is invariant. Denote by
Xh and
Y the representations associated to Xh
and Y , respectively (see Proposition 5.2.7), and by
Yx W A0 ! L.Cn / the map
given by f 7! .id ˝f /.Yx /. For each f 2 A0 , define fx 2 A0 by a 7! f .a /.
P P
Then
Yx .f / D i;j eSij f .aij  / D i;j eij fN.aij / D
Y .fN/ for all f 2 A0 , and
combining this relation with equation (5.11), we find

Xh .A0 /T D T
Yx .A0 / D T
Y .A0 /:

Using Proposition 5.2.7, we conclude that Im T  Hh is invariant for Xh and

ker T  Cn is invariant for Y . Since Y is irreducible and T ¤ 0, the map T must
be injective. Thus, Yx is equivalent to a restriction of Xh to the invariant subspace
Im T ; in particular, it is a corepresentation operator.
An easy calculation shows that Yx  is a corepresentation operator:

YxŒ12 YŒ13 D .YxŒ12 YxŒ13 / D ..id ˝/.Yx // D .id ˝/.Yx  /:
 x 

The fact that Yx  is irreducible will follow from statement ii).

 5.3. Corepresentation theory and structure theory 127

ii) Put G z WD .id ˝h/.Yx  Yx / 2 L.Cn /. Evidently, G z is positive, and as in

Proposition 3.2.1, we conclude: Yx is invertible ) Yx Yx is invertible ) Yx  Yx > 1


for some > 0 ) G z > 1 ) G z is invertible. By Lemma 5.3.1, G

z 2 Hom.Yx ; Yx  /;
in particular, Yx ' Yx . Now Fz is some uniquely determined multiple of G.
 z
iii), iv) This follows as in Proposition 3.2.9. 

Propositions 5.3.4, 5.3.7 and 5.3.8 imply:

Corollary 5.3.9. For all corepresentation operators X and Y ,

Hom.X; Y / D 0 , h.C .Y / C .X // D 0
, h.C .Y /C .X / / D 0 , Hom.Y; X / D 0: 

Recall that we associated to every corepresentation operator X on a Hilbert

x 2 M.K.H
space H a conjugate operator X x / ˝ A/. Now we can prove:

Corollary 5.3.10. The conjugate of a corepresentation operator is a corepre-

sentation operator again.

Proof. Every corepresentation operator is equivalent to a direct sum of irreducible

ones, and for every irreducible corepresentation operator, the conjugate is a core-
presentation operator by Proposition 5.3.8. 

5.3.3 Characterization of C  -algebraic compact quantum groups

Like algebraic compact quantum groups, C  -algebraic compact quantum groups
can be characterized in terms of their corepresentations:

Theorem 5.3.11. The following conditions on a unital C  -bialgebra .A; / are

equivalent:

i) .A; / is a C  -algebraic compact quantum group.

ii) The subspace of A spanned by the elements of all irreducible unitary corepre-
sentation matrices a 2 Mn .A/ .n 2 N/ whose conjugate aN D .aij  /i;j is
invertible, is dense in A.
iii) The -subalgebra of A generated by the elements of all corepresentation
matrices a 2 Mn .A/ .n 2 N/ whose conjugate aN is invertible, is dense in A.
iv) The subalgebra of A generated by the elements of all corepresentation ma-
trices is dense in A.

Proof. i) ) ii): By Theorem 5.3.3 iv), the regular corepresentation operator is

equivalent to a direct sum of finite-dimensional irreducible unitary corepresenta-
tion operators, and by Theorem 5.2.9 iii), its space of matrix elements is equal to A.
128 Chapter 5. C  -algebraic compact quantum groups

Therefore the matrix elements of finite-dimensional irreducible unitary corepre-

sentations span a dense subspace of A. Furthermore, if a 2 Mn .A/ is a unitary
irreducible corepresentation matrix, then the matrix aN is invertible by Proposi-
tion 3.2.9 i).
ii) ) iii): Trivial.
iii) ) iv): If a 2 Mn .A/ is a corepresentation matrix and aN is invertible, then
also aN is a corepresentation matrix.
iv) ) i): We have to show that .A/.1A ˝ A/ and .A/.A ˝ 1A / are linearly
dense in A ˝ A. If a 2 Mn .A/ is a corepresentation matrix with inverse b, then
X X X
.aik /.1A ˝ bkj / D ail ˝ alk bkj D ail ˝ ıl;j 1A D aij ˝ 1A
k k;l l

for all i; j . Moreover, if

X X
c ˝ 1A D .c1;i /.1A ˝ c2;i / and d ˝ 1A D .d1;j /.1A ˝ d2;j /;
i j

for some c1;i ; c2;i ; d1;j ; d2;j 2 A, then

X
cd ˝ 1A D .c1;i /.1A ˝ c2;i /.d ˝ 1A /
i
X X
D .c1;i /.d ˝ 1A /.1A ˝ c2;i / D .c1;i d1;j /.1A ˝ d2;j c2;i /:
i i;j

Now, the assumption and the equations above imply that .A/.1A ˝ A/ is linearly
dense in A ˝ A. A similar argument shows that .A/.A ˝ 1A / is linearly dense in
A ˝ A. 

5.4 The relation to algebraic compact quantum groups

C  -algebraic and algebraic compact quantum groups are closely related. In this
section, we explain how one can associate to every C  -algebraic compact quantum
group an algebraic one, and how one can associate to an algebraic compact quantum
group a maximal and a minimal C  -algebraic one.
Notation. We shall abbreviate the phrase “compact quantum group” by the acronym
“CQG” whenever it seems convenient.

5.4.1 From C  -algebraic to algebraic CQGs

For every C  -algebraic compact quantum group, the subspace spanned by the ma-
trix elements of finite-dimensional corepresentation operators forms an algebraic
compact quantum group:
 5.4. The relation to algebraic compact quantum groups 129

Theorem 5.4.1. Let .A; / be a C  -algebraic compact quantum group. Denote

by A0  A the subspace spanned by the matrix elements of all finite-dimensional
corepresentation operators.
i) A0 is dense in A, the image .A0 / is contained in the algebraic tensor
product A0 ˇ A0 , and .A0 ; 0 / is an algebraic compact quantum group,
where 0 WD jA0 .
z 0 / is an algebraic compact quantum group such that Az0  A is a
ii) If .Az0 ; 
dense -subalgebra and  z 0 D j z , then .Az0 ; 
z 0 / D .A0 ; 0 /.
A0

We call .A; /0 WD .A0 ; 0 / the algebraic CQG associated to .A; /.
iii) Every morphism of C  -algebraic CQGs .A; A / ! .B; B / restricts to a
morphism .A; A /0 ! .B; B /0 of algebraic CQGs.
Proof. i) A0 is dense in A by Theorem 5.3.11 ii), and a -algebra because
x/
C .X/ C C .Y / D C .X  Y /; C .X /C .Y /  C.X  Y /; C .X / D C.X

for all finite-dimensional corepresentation operators X and Y . The unit 1A belongs

to A0 because it is the matrix element of the trivial corepresentation. Equations
(5.6) and (5.7) on page 116 show that .A0 / is contained in A0 ˇ A0 . Thus
.A0 ; 0 / is a bialgebra.
Let us show that .A0 ; 0 / is a Hopf algebra. Choose a maximal family .u˛ /˛
of pairwise inequivalent irreducible unitary corepresentation matrices of .A; /.
By Theorem 5.3.3 iv), Proposition 5.3.7, and Proposition 5.3.8 iii), iv), the family
.u˛ij /˛;i;j is a basis of A0 . Consider the linear maps

0 W A0 ! C; u˛ij 7! ıi;j ; and S0 W A0 ! A0 ; u˛ij 7! .uj˛i / :

Equations (5.6) and (5.8) imply that 0 and S0 satisfy the axioms for the counit and
for the antipode of a Hopf algebra. Therefore .A0 ; 0 / is a Hopf -algebra. It is
an algebraic compact quantum group because the restriction of h to A0 is a positive
integral.
ii) By Theorem 3.2.12 iii), Az0 is spanned by the components of its finite-
dimensional corepresentation matrices; since these matrices are also corepresenta-
tion matrices of .A; /, the space Az0 is contained in A0 . If there exists an irreducible
unitary corepresentation matrix .aij /i;j of A whose components are not contained
in Az0 , we obtain a contradiction to the assumption that Az0 is dense in A from
Proposition 5.3.7 and 5.3.8 iii).
iii) Let .B; B / be another C  -algebraic compact quantum group and let

W A ! B be a morphism of C  -bialgebras. Since

is non-degenerate, it is
unital. For every corepresentation matrix .aij /i;j 2 Mn .A/, the image under
,
130 Chapter 5. C  -algebraic compact quantum groups

.
.aij //i;j 2 Mn .B/, is a corepresentation matrix as well. Therefore,
.A0 / is con-
tained in B0 , that is,
restricts to a unital morphism of -bialgebras
0 W A0 ! B0 .
From the construction of the counit and antipode in i), it is easy to see that
0 pre-
serves the counit and the antipode. 

Example 5.4.2. i) For every compact group G, the C  -bialgebra C.G/ introduced
in Example 4.2.1 is a C  -algebraic CQG, and the associated algebraic CQG is the
Hopf -algebra of representative functions Rep.G/ introduced in Example 1.2.5.
ii) For every discrete group G, the C  -bialgebras C  .G/ and Cr .G/ introduced
in Example 4.2.2 are C  -algebraic CQGs, and the associated algebraic CQG is both
times the group Hopf -algebra CG introduced in Example 1.2.8.

The preceding theorem shows that every C  -algebraic compact quantum group
.A; / is the completion of some unique algebraic compact quantum group .A0 ; 0 /
which has the same category of finite-dimensional corepresentations. This result
provides a partial answer to the question for the antipode and for the counit of
.A; /: at least, both maps are well defined on A0 . If .A; / is reduced, then the
antipode of .A0 ; 0 / can be extended to a closed (possibly unbounded) linear map
on A and can be described in terms of a unitary antipode and a scaling group, see
Example 8.3.7. Moreover, in that case, the modular automorphism of the Haar state
of .A0 ; 0 / (Theorem 2.2.17) extends to a modular automorphism of the Haar state
of .A; /, see Example 8.1.22.

5.4.2 From algebraic to C  -algebraic CQGs

An algebraic compact quantum group can have several different completions that
are C  -algebraic compact quantum groups, see Example 5.4.2 ii). However, there
always exist a maximal and a minimal one. First, we describe the maximal com-
pletion:

Theorem 5.4.3. Let .A; / be an algebraic compact quantum group. Then

a 7! kaku WD supfp.a/ j p is a C  -seminorm on Ag

is a C  -norm on A. Denote by Au the corresponding completion of A.

i) Au is a unital C  -algebra, and the comultiplication  extends to a -homo-

morphism u W Au ! Au ˝ Au that turns Au into a C  -algebraic CQG. The
counit of A extends to a character u on Au that satisfies . u ˝ id/ ı u D
id D .id ˝ u / ı u .

ii) .A; / is the algebraic compact quantum group associated to .Au ; u /.

We call .A; /u WD .Au ; u / the universal C  -algebraic CQG of .A; /.
 5.4. The relation to algebraic compact quantum groups 131

iii) Every morphism of algebraic CQGs .A; A / ! .B; B / extends uniquely

to a morphism of C  -algebraic CQGs .A; A /u ! .B; B /u .

Proof. First, we show that kaku is finite for each a 2 A. Let p be a C  -seminorm
on A. Then for each unitary corepresentation matrix P .aij /i;j 2 Mn .A/ and all
i; j , we have p.aij /2 D p.aij  aij /  1 because i aij  aij D 1A . Since A is
spanned by the elements of unitary corepresentation matrices (Theorem 3.2.12 iii)),
it follows that kaku is finite for each a 2 A.
Let us show that the seminorm k  ku is a norm. Denote by h the Haar state
of .A; /, and by H the completion of A with respect to the inner product given
by hbjci WD h.b  c/ for all b; c 2 A. Let .aP ij /i;j be a corepresentation matrix of
.A; /, and fix i; j . Since h is positive and k akj  akj D 1A ,

kaij bkH
2
D h.b  aij  aij b/  h.b  1A b/ D kbkH
2
for all b 2 A:

Therefore the map b 7! aij b extends to a bounded operator

.aij / 2 L.H /. Since
A is spanned by the elements of unitary corepresentation matrices,
extends to a
-homomorphism
W A ! L.H /, where
.a/b D ab for all a; b 2 A. Since h
is faithful (Proposition 2.2.4), kaku k
.a/k k
.a/1A k D h.a a/ > 0 for all
non-zero a 2 A.
i) The maps a 7! k.a/k.Au ˝Au / and a 7! j .a/j are C  -seminorms on A.
Therefore they are dominated by k  ku , and  and extend to -homomorphisms
u and u as claimed. By Theorem 3.2.12 iii) and 5.3.11 ii), .Au ; u / is a C  -
algebraic compact quantum group. Finally, the equation concerning u follows
from the corresponding equation for and the density of A in Au .
ii) This follows immediately from Theorem 5.4.1 ii).
iii) Let .B; B / be an algebraic compact quantum group and let
W A ! B
be a morphism of Hopf -algebras. By definition of the norm k  ku on Au , the

-homomorphism A  ! B ! Bu extends to a -homomorphism
u W Au ! Bu .
Since
preserves the unit and the comultiplication, so does
u . 

Remark 5.4.4. The antipode of .A; / need not be bounded with respect to the
norm k  ku and need not extend to Au .

Now we turn to the minimal completion:

Theorem 5.4.5. Let .A; / be an algebraic compact quantum group. Denote by

.H; ƒ;
/ the GNS-representation for the Haar state hu of .Au ; u / D .A; /u ,
and put Ar WD
.Au /.

i)
restricts to an embedding A ,! Ar , and the comultiplication  extends to
a -homomorphism r W Ar ! Ar ˝ Ar that turns .Ar ; r / into a reduced
C  -algebraic compact quantum group.
132 Chapter 5. C  -algebraic compact quantum groups

ii) .A; / is the algebraic compact quantum group associated to .Ar ; r /.

We call .A; /r WD .Ar ; r / the reduced C  -algebraic CQG of .A; /.

Proof. i) The restriction of hu to A is a positive normalized integral, and hence

equal to the Haar state h of .A; / (Proposition 2.2.6 ii)). Since h is faithful
(Proposition 2.2.4), k
.a/k2 k
.a/ƒ.1A /k2 D h.a a/ > 0 for all a 2 A.
To show that  extends to a -homomorphism r W Ar ! Ar ˝ Ar , we intro-
duce an auxiliary multiplicative unitary: The map

V0 W ƒ.A/ ˇ ƒ.A/ ! ƒ.A/ ˇ ƒ.A/;

ƒ.b/ ˇ ƒ.c/ 7! .ƒ ˇ ƒ/..c/.b ˇ 1//;

is bijective because A is a Hopf -algebra, and isometric since

hV0 .ƒ.b/ ˇ ƒ.c//jV0 .ƒ.b 0 / ˇ ƒ.c 0 //i

D .h ˇ h/..b  ˇ 1/.c  c 0 /.b 0 ˇ 1//
D h.b  b 0 /h.c  c 0 / D hƒ.b/ ˇ ƒ.c/jƒ.b 0 / ˇ ƒ.c 0 /i

for all b; c; b 0 ; c 0 2 A. Therefore, V0 extends to a unitary V on H ˝ H . Now

V .1 ˝
.a// D .
˝
/..a//V for all a 2 AI

indeed, the operators on the left-hand side and on the right-hand side are given
by ƒ.b/ ˇ ƒ.c/ 7! .ƒ ˇ ƒ/..ac/.b ˇ 1// for all b; c 2 A. Therefore the
-homomorphism

r W L.H / ! L.H ˝ H /; T 7! V .1 ˝ T /V  ;

extends  as desired. By Theorems 3.2.12 iii) and 5.3.11 ii), .Ar ; r / is a C  -al-
gebraic compact quantum group.
Finally, let us show that the Haar state hr on Ar is faithful. Note that the restric-
tion of hr to A is equal to h. Denote by  the modular automorphism of h (see Theo-
rem 2.2.17 or 3.2.19 v)). Let x 2 Ar ; x ¤ 0. We have to show that hr .x  x/ > 0.
Since ƒ.A/ is dense in H , we find b; c 2 A such that hƒ.c/jxƒ.b/i > 0. By the
Cauchy–Schwarz inequality, there exists a constant Cb;c > 0 such that

jhƒ.c/j
.a/ƒ.b/ij D jh.c  ab/j D jh.ab .c  //j
 Cb;c h.a a/ D Cb;c hr .
.a a//

for all a 2 A. But then also 0 < jhƒ.c/jxƒ.b/ij  Cb;c hr .x  x/, so 0 < hr .x  x/.
ii) This follows immediately from Theorem 5.4.1 ii). 
 5.4. The relation to algebraic compact quantum groups 133

Example 5.4.6. i) Let G be a discrete group. Then the group algebra CG carries
the structure of an algebraic CQG (see p. 66), and the associated universal and the
associated reduced C  -algebraic CQG are the C  -bialgebras C  .G/ and Cr .G/,
respectively, introduced in Example 4.2.2.
ii) Let G be a compact group. Then the Hopf -algebra of representative func-
tions Rep.G/ carries the structure of an algebraic CQG (see p. 66), and the associ-
ated universal and the associated reduced C  -algebraic CQG are both equal to the
C  -bialgebra C.G/ introduced in Example 4.2.1.

Remark 5.4.7. The counit of .A; / need not be bounded with respect to the
norm of Ar and need not extend to Ar : If G is a discrete group and A D CG, then
Ar D Cr .G/, and extends to Ar if and only if G is amenable. In particular, this
example shows that the assignment A 7! Ar does not extend to a functor, because
C can be considered as an algebraic CQG and as a morphism of algebraic CQGs.

Using the preceding constructions, we can associate to every C  -algebraic CQG

.A; / the following three CQGs:

• an algebraic CQG .A0 ; 0 /,

• the universal C  -algebraic CQG .Au ; u / of .A0 ; 0 /,

• the reduced C  -algebraic CQG .Ar ; r / of .A0 ; 0 /.

We call .A; /u WD .Au ; u / and .A; /r WD .Ar ; r / the universal and the
reduced CQG associated to .A; /.

Proposition 5.4.8. Let .A; / be a C  -algebraic CQG. Denote by u , , r the

canonical embeddings of A0 in Au , A, Ar , respectively. Then there exists a com-
mutative diagram

A0 QQQ
u llll QQrQ
ll QQQ
v ll
l 
//A / /( Ar ,
Au

where the horizontal maps are surjective morphisms of C  -bialgebras.

Proof. Denote by k  ku ; k  k and k  kr the norms on Au ; A and Ar , respectively.

Evidently, it suffices to show that

kr .a/kr  k.a/k  ku .a/ku for all a 2 A0 :

The inequality on the right-hand side follows immediately from the definition of
k  ku . To prove the inequality on the left-hand side, let us use the notation of
134 Chapter 5. C  -algebraic compact quantum groups

Theorem 5.4.5. Denote by h and h0 the Haar states of A and A0 , respectively. For
all a; b 2 A0 ,

.b/ .a a/.b/  .b/ .b/k.a/k2 in A

and, since h0 D h ı  by Proposition 2.2.6,

kr .a/ƒ.b/k2 D jh0 .b  a ab/j D jh..b  /.a a/.b//j

 jh0 .b  b/j  k.a/k2 D kƒ.b/k2  k.a/k2 :

Therefore, kr .a/k  k.a/k for all a 2 A0 . 

The previous result motivates the following terminology:
Definition 5.4.9. We call a C  -algebraic CQG .A; / universal if the natural sur-
jection Au ! A is an isomorphism.
We call a CQG .A; / amenable if the associated universal and reduced
C  -algebraic CQGs coincide.
Let us summarize the contents of Theorems 5.4.1, 5.4.3 and 5.4.5 in categorical
terms. Evidently, the class of all algebraic compact quantum groups forms a cate-
gory with respect to morphisms of Hopf -algebras, and the class of all universal
C  -algebraic compact quantum groups forms a category with respect to morphisms
of C  -bialgebras.
Corollary 5.4.10. i) The assignment .A0 ; 0 / 7! .Au ; u / defines a natural equiv-
alence between the category of algebraic CQGs and the category of universal
C  -algebraic CQGs.
ii) The assignment .A0 ; 0 / 7! .Ar ; r / defines a bijective correspondence
.up to isomorphism/ between all algebraic CQGs and all reduced C  -algebraic
CQGs. 
Chapter 6
Examples of compact quantum groups

The general theory of compact quantum groups is complemented by a rich supply

of examples. The most important sources of examples are the following:

q-deformations of compact semisimple Lie groups. In [37] and [70], Drinfeld

and Jimbo associated to every semisimple complex Lie algebra g a Hopf algebra
Uq .g/ which is a deformation of the universal enveloping algebra U.g/ depending
on a formal parameter q. These q-deformations attracted much attention [23], [24],
[68], [79], [80], [84], [103], [140], but were studied from a different perspective
than that of compact quantum groups. The first contact point for the two approaches
was the compact quantum group SU .2/, which was introduced independently by
Soibelman and Vaksman [169] and by Woronowicz [194]; see also [133]. Later,
Levendorski and Soibelman [100] and Rosso [134] constructed many more exam-
ples of compact quantum groups out of the q-deformations of Drinfeld and Jimbo;
see also [84].

Universal/free compact quantum groups. Another important class of examples

are the free unitary and the free orthogonal quantum groups, which were introduced
by Wang [185], [186] and generalized by Wang and Van Daele [179], and the
quantum permutation groups introduced by Wang [189]. A detailed study of these
quantum groups was given by Banica [12], [13], [14].

General constructions. There exist several general constructions that produce new
compact quantum groups out of given ones, for example

• the free product, tensor product, and crossed product constructions of Wang
[186], [187],

• the deformation of compact groups by toral subgroups and its generalization

to compact quantum groups, which were given by Rieffel [131] and Wang
[188], respectively.

The examples of the first and second kind listed above belong to the class of
compact matrix quantum groups. We begin with a review of this class (Section 6.1)
and thereafter discuss the following examples: the compact quantum group SU .2/
of Woronowicz (Section 6.2); free products and tensor products of compact quantum
groups (Section 6.3); and the free unitary and the free orthogonal quantum group
(Section 6.4).
136 Chapter 6. Examples of compact quantum groups

Throughout this chapter, we use the results on compact quantum groups that
were obtained in Chapters 3 and 5; in particular the correspondence between alge-
braic and C  -algebraic compact quantum groups described in Section 5.4 and the
correspondence between corepresentations, corepresentation operators, and core-
presentation matrices described in Sections 3.1 and 5.2.

6.1 Compact matrix quantum groups

Compact matrix quantum groups are analogues of compact Lie groups. They in-
clude many examples of compact quantum groups like the q-deformations of com-
pact semisimple Lie groups and the universal compact quantum groups mentioned
above. Their theory was developed by Woronowicz [193]; it preceded the theory
of general compact quantum groups.
Let us fix some notation. Throughout this section, we treat the algebraic and
the C  -algebraic setting in parallel and use the symbol “˝” to denote the algebraic
tensor product of -algebras and the minimal tensor product of C  -algebras. It
will always be clear from the context which tensor product we refer to. Given
a -algebra A and a matrix u D .uij /i;j 2 Mn .A/, we put uN WD .uij  /i;j and
ut D .uj i /i;j ; thus, uN  D ut .
Roughly, a compact matrix quantum group is a compact quantum group that is
generated by the entries of one fundamental corepresentation matrix:

Definition 6.1.1. A C  -algebraic compact matrix quantum group is a unital C  -

algebra A equipped with a -homomorphism  W A ! A ˝ A and a unitary u 2
Mn .A/, where n 2 N, such that
P
i) .uij / D k uik ˝ ukj for all i; j ;

ii) uN is an invertible matrix;

iii) the elements uij .1  i; j  n/ generate A (as a C  -algebra).

An algebraic compact matrix quantum group is a Hopf -algebra .A; / together

with a unitary u 2 Mn .A/, where n 2 N, such that

i) u is a corepresentation matrix;

ii) uN is equivalent to a unitary corepresentation matrix;

iii) the elements uij .1  i; j  n/ generate A (as a -algebra).

In both cases, u is called the fundamental corepresentation .matrix/ of .A; ; u/.

 6.1. Compact matrix quantum groups 137

Remarks 6.1.2. i) In the definition above,  is uniquely determined by A and u

because of conditions i) and iii). Therefore one can also refer to the pair .A; u/ as
a compact matrix quantum group.
ii) Let .A; ; u/ be an algebraic compact matrix quantum group. Then the counit
and the antipode of the Hopf -algebra .A; / are determined by .uij / D ıi;j and
S.uij / D uji for all i; j ; see Proposition 3.1.7 iii), v).

Definition 6.1.3. We call two compact matrix quantum groups .A; A ; u/ and
.B; B ; v/ similar and write .A; A ; u/ ' .B; B ; v/ if there exist a -isomor-
phism f W A ! B, an n 2 N, and a matrix T 2 GLn .C/ such that u 2 Mn .A/,
v 2 Mn .B/, and v D T .f .uij //i;j T 1 .

Notation. We abbreviate the phrase “compact matrix quantum group” by the

acronym “CMQG” whenever it seems convenient, and omit the qualifiers “alge-
braic” and “C  -algebraic” when we want to refer to both cases.

The terminology introduced in Definition 6.1.1 is justified because of the fol-

lowing result:

Proposition 6.1.4. i) If .A; ; u/ is a CMQG, then .A; / is a CQG and u, uN are

corepresentation matrices of .A; /.
Let .A; A ; u/ and .B; B ; v/ be compact matrix quantum groups, where u 2
Mn .A/ and v 2 Mn .B/ for some n 2 N.
ii) Let f W A ! B be a -homomorphism such that f .uij / D vij for all i; j .
Then f is a morphism of compact quantum groups.
iii) If .A; A ; u/ ' .B; B ; v/ as CMQGs, then .A; A / Š .B; B / as CQGs.

Proof. i) In the algebraic case, the assertion follows from Theorem 3.2.12 iv). In
the C  -algebraic case, conditions i) and iii) of Definition 6.1.1 imply that  is
coassociative, so that .A; / is a C  -bialgebra. Furthermore, condition i) implies
that u is a corepresentation matrix of .A; /. Now the assertion follows from
conditions ii), iii) and Theorem 5.3.11 iii).
ii) We have .f ˝ f / ı A D B ı f because A is generated by the uij and
X
.f ˝ f /.A .uij // D f .uik / ˝ f .ukj /
k
X
D vik ˝ vkj D B .vij / D B .f .uij //
k

for all i , j .
138 Chapter 6. Examples of compact quantum groups

iii) Let f W A ! B and T be as in Definition 6.1.3. Then

X
.f ˝ f /.A .uij // D f .uil / ˝ f .ulj /
l
X
1 1
D Tim vmp Tpl ˝ Tlq vq n Tnj
m;p;l;q;n
X (6.1)
1
D Tim vmk ˝ vk n Tnj
m;k;n
X
1
D Tim B .vmn /Tnj D B .f .uij //
m;n

for all i; j , and hence .f ˝ f / ı A D B ı f . 

The next results characterize those compact quantum groups which can be re-
garded as compact matrix quantum groups:

Lemma 6.1.5. Let .A; / be a compact quantum group with a unitary corepre-
sentation matrix u 2 Mn .A/, where n 2 N. Then the following conditions are
equivalent:

i) Every irreducible corepresentation of .A; / is contained in some iterated

N and the trivial corepresentation.
tensor product of u, u,
ii) The matrix entries of u generate A.

Proof. This follows from the fact that every -monomial in the matrix entries of u
appears as a matrix coefficient of some tensor product of the corepresentations u
N and from Theorem 3.2.1 / 5.3.3 and Corollary 3.2.8 / 5.3.9.
and u, 

Proposition 6.1.6. For every algebraic/C  -algebraic compact quantum group

.A; /, the following conditions are equivalent:

i) There exists a unitary corepresentation matrix u 2 Mn .A/ for some n 2 N

such that every irreducible corepresentation of .A; / is contained in some
N and the trivial corepresentation.
iterated tensor product of u, u,
ii) The following -algebra is finitely generated: A / the subalgebra A0  A of
matrix coefficients of finite-dimensional corepresentations.
iii) There exists a unitary u 2 Mn .A/ for some n 2 N such that .A; ; u/ is an
algebraic/C  -algebraic compact matrix quantum group.

Proof. i) ) ii) This is evident.

ii) ) i) If A / A0 is finitely generated as a -algebra, then there exist finite-
dimensional unitary corepresentation matrices u1 ; : : : ; uk whose matrix entries
 6.1. Compact matrix quantum groups 139

generate A / A0 as a -algebra (see Theorem 3.2.12 v)), and we can choose u

to be the block matrix with diagonal matrix blocks u1 ; : : : ; uk .
iii) , i) This follows directly from Lemma 6.1.5. 

Remark 6.1.7. Compact matrix groups can be reconstructed from their category
of finite-dimensional corepresentations. This generalization of the Tannaka–Krein
duality theory was proved by Woronowicz, see [195].

Algebraic and C  -algebraic compact matrix quantum groups are related as fol-
lows. Recall from Section 5.4 that one can associate

• to every C  -algebraic CQG .A; / an algebraic CQG .A0 ; 0 /,

• to every algebraic CQG .A; / a universal and a reduced C  -algebraic CQG
.Au ; u / and .Ar ; r /, respectively.

Proposition 6.1.8. i) If .A; ; u/ is a C  -algebraic CMQG, then the coefficients

of u belong to A0 , and .A0 ; 0 ; u/ is an algebraic CMQG.
If .A; ; u/ is an algebraic CMQG, then .Au ; u ; u/ and .Ar ; r ; u/ are
C  -algebraic CMQGs.
ii) If .A; A ; u/ and .B; B ; v/ are similar C  -algebraic CMQGs, then the
algebraic CMQGs .A0 ; A;0 ; u/ and .B0 ; B;0 ; v/ are similar.
If .A; A ; u/ and .B; B ; v/ are similar algebraic CMQGs, then the C  -alge-
braic CMQGs .Au ; A;u ; u/ and .Bu ; B;u ; v/ are similar. 

Let us consider some examples of compact matrix quantum groups. We begin

with examples related to classical groups:

Example 6.1.9. Let G  Un .C/ be a closed subgroup, where n 2 N. For

i; j D 1; : : : ; n, denote by uij W C.G/ ! C the function x 7! xij . Then the
C  -bialgebra C.G/ together with the matrix u D .uij /i;j is a C  -algebraic com-
pact matrix quantum group. P Indeed, condition i) of Definition 6.1.1 holds because
.uij /.x; y/ D uij .xy/ D k uik .x/ukj .y/ for all x; y 2 G, condition ii) holds
because uuN t D .uu / D 1 (here we use commutativity of C.G/), and condition
iii) holds by the Stone–Weierstrass Theorem.
The Hopf -algebra Rep.G/ of all representative functions on G contains all uij ,
and together with the matrix u, it is an algebraic compact matrix quantum group.
The fact that condition ii) is satisfied in this case is well-known; this follows also
from the fact that Rep.G/ is the algebraic compact quantum group associated to
C.G/ (see Proposition 6.1.8 and Example 5.4.2 i)).
Let H  Un .C/ be another closed subgroup. It is easy to see that the following
conditions are equivalent:

i) .C.G/; G ; uG / ' .C.H /; H ; uH /;

140 Chapter 6. Examples of compact quantum groups

ii) .Rep.G/; G ; uG / ' .Rep.H /; H ; uH /;

iii) H D T GT 1 for some T 2 GLn .C/.
Here, we indexed the comultiplications and the fundamental corepresentations of
C.G/ and C.H / by “G” and “H ”, respectively, to avoid ambiguities.
Note that every compact Lie group G can be embedded as a closed subgroup
into Un .C/ for some n 2 N and thus gives rise to an example of a compact matrix
quantum group.
Example 6.1.10. Let G be a discrete group that is finitely generated by elements
x1 ; : : : ; xn . Then the group Hopf -algebra CG together with the diagonal matrix
u D diag.Ux1 ; : : : ; Uxn / 2 Mn .CG/ is an algebraic CMQG, and the universal
group C  -bialgebra C  .G/ together with the matrix u as well as the reduced group
C  -bialgebra Cr .G/ together with the matrix diag.Lx1 ; : : : ; Lxn / are C  -algebraic
CMQGs.
Every commutative compact matrix quantum group has the form described in
Example 6.1.9:
Proposition 6.1.11. Let .A; ; u/ be an algebraic/C  -algebraic compact matrix
quantum group, where u 2 Mn .A/, n 2 N. Assume that A is commutative.
Then there exist a closed compact subgroup G  Un .C/ and an isomorphism
Š Š
ˆW A ! Rep.G/ / ˆ W A 
! C.G/ such that .ˆ.uij //.x/ D xij for all x 2 G and
i; j D 1; : : : ; n.
Proof. By Proposition 6.1.8 and Example 5.4.2 i) and 5.4.6 ii), it suffices to consider
the case that .A; ; u/ is a C  -algebraic CMQG. Denote by Ay the spectrum of A,
that is, the space of all -homomorphisms A ! C, and consider the map

‰ W Ay ! Mn .C/;  7! ..uij //i;j :

This map is an embedding because the matrix entries of u generate A, it is a

homeomorphism onto a closed subset because Ay is compact, and its image, which
we denote by G, is contained in Un .C/ because u is unitary.
y Then 00 WD . ˝ 0 / ı  2 A,
Let ; 0 2 A. y and ‰./‰.0 / D ‰.00 /
because
X
.‰./‰.0 //ij D .uik /0 .ukj / D . ˝ 0 /..uij // D ‰.00 /ij for all i; j:
k

In particular, GG  G. Thus G is a compact semigroup with cancellation and

hence a group (see, for example, [63, Theorem II.9.16] or [105, Proposition 3.2]).
By the Gelfand Theorem (see Section 12.1), the map ˆ W A ! C.G/ given by
.ˆ.a//.x/ WD .‰ 1 .x//.a/ is an isomorphism, and .ˆ.uij //.‰.// D .uij / D
.‰.//ij for all ; i; j . The claim follows. 
 6.1. Compact matrix quantum groups 141

Finally, we give an example of a compact matrix quantum group that is neither

commutative nor cocommutative. Its construction uses the following easy obser-
vation:
a -algebra, n
2 N, and u 2 Mn .A/ is unitary, then also
Lemma 6.1.12. If A isP
the matrix n .u/ WD k uik ˝ ukj i;j 2 Mn .A ˝ A/ is unitary. 
Example 6.1.13. For each n 2 N, one can define a quantum permutation group on
n letters as follows. Given a -algebra A, call a unitary matrix u 2 Mn .A/ magic
if for each i and j ,
i) uij is a projection,
P
ii) the projections ui1 ; : : : ; ui n are orthogonal and k uik D 1, and
P
iii) the projections u1j ; : : : ; unj are orthogonal and k ukj D 1.
Note that uN D u for every magic unitary matrix u.
Denote by As .n/ the universal C  -algebra with elements uij .1  i; j  n/
such that u D .uij /i;j is a magic unitary matrix. A routine calculation shows that
X 
n .u/ WD uik ˝ ukj 2 Mn .As .n/ ˝ As .n//
i;j
k

is a magic unitary matrix again. By the universal property of As .n/, there exists a
unique -homomorphism  W As .n/ ! As .n/ ˝ As .n/ such that .As .n/; ; u/ is
a C  -algebraic compact matrix quantum group. This is the quantum permutation
group on n letters. It was introduced by Wang [189] and thereafter studied by
Banica and Collins [14], [15].
The quantum permutation group is related to the group S.n/ of permutations
of n letters as follows. Denote by Acs .n/ the universal commutative C  -algebra
with elements ucij .1  i; j  n/ such that uc D .ucij /i;j is a magic unitary
matrix. Then there exists a unique -homomorphism c such that .Acs .n/; c ; uc /
is a C  -algebraic CMQG, and a surjective morphism of C  -algebraic CMQGs
As .n/ ! Acs .n/, given by uij 7! ucij . By Proposition 6.1.11, .Acs .n/; c ; uc / is
isomorphic to the C  -algebraic CMQG associated to the group

f..ucij //i;j 2 GLn .C/ j  is a character on Acs .n/g

P P
D fx 2 GLn .C/ j xij 2 f0; 1g; k xik D 1; k xkj D 1 for all i; j g
D fx 2 GLn .C/ j there is
2 S.n/ with xij D ıi; .j / for all i; j g Š S.n/:
142 Chapter 6. Examples of compact quantum groups

6.2 The compact quantum group SU .2/

The quantum group SU .2/ is probably the fundamental example of a compact
quantum group that is neither commutative nor cocommutative. It was introduced
independently by Soibelman and Vaksman [169] as the dual of the q-deformed
enveloping Lie algebra Uq .su2 /, and by Woronowicz [194] as a deformation of the
compact Lie group SU2 .C/ or, more precisely, of the C  -bialgebra C.SU2 .C//. A
comparison between these approaches can be found in [133].
We shall follow the approach of Woronowicz and present the construction of the
compact quantum group SU .2/, describe the corepresentation theory, and deduce
some results on the Haar state.
For further information, we refer to the following literature: A gentle introduc-
tion to the q-deformed enveloping Lie algebra Uq .su2 /, the Hopf algebra SU .2/,
and their interrelation can be found in [79]; for a comprehensive treatment, see [24],
[80]. An introduction to q-deformations, their relation to Poisson–Lie groups, and
to functions on q-deformed quantum groups is given in [24], [84]. The harmonic
analysis on SU .2/ and relations to q-special functions were studied in many pa-
pers, see, for example, [83], [108], [109], [169]; comprehensive treatments can be
found in [80], [183].
Note that in the literature, the deformation parameter  is usually denoted by q,
and the quantum group SU .2/ is denoted by a variety of different symbols. Com-
mon notation for the underlying algebraic compact quantum group includes, for
example, SUq .2/, Oq .SU2 /, CŒSU2 q .

6.2.1 Definition and first properties

Recall that the compact Lie group SU2 .C/ consists of all matrices of the form
 
˛ N
g.˛;/ WD ; where ˛;  2 C and ˛˛ N C N  D 1:
 ˛N

Define a; c 2 C.SU2 .C// by a.g.˛;/ / WD ˛ and c.g.˛;/ / WD . Then the C  -bi-

algebra C.SU2 .C// together with the unitary matrix
 
a c 
u WD 2 M2 .C.SU2 .C/// (6.2)
c a

is a C  -algebraic compact matrix quantum group (Example 6.1.9). The corre-

sponding algebraic compact quantum group is the Hopf -algebra of represen-
tative functions Rep.SU2 .C//  C.SU2 .C//. As a C  -algebra or -algebra,
respectively, C.SU2 .C// and Rep.SU2 .C// are isomorphic to the universal uni-
tal commutative C  -algebra or -algebra generated by elements a; c that satisfy
 6.2. The compact quantum group SU .2/ 143

a a C c  c D 1. A short calculation shows that the counit and antipode of

Rep.SU2 .C// are given by

.a/ D 1; .c/ D 0; S.a/ D a ; S.a / D a; S.c/ D c; S.c  / D c  :

Definition 6.2.1. For  2 Œ1; 1, denote by SU .2/ the universal unital C  -alge-
bra generated by elements a; c subject to the condition that the following matrix is
unitary:
 
a c 
u WD :
c a

If u is as above, then
 
a c
u D ;
c a
 
 a a C c  c a c  C c  a
u uD ;
ca C ac 2 cc  C aa
  
aa C 2 c  c ac   c  a
uu D :
ca  a c cc  C a a

Hence u is unitary if and only if a and c satisfy

a a C c  c D 1; aa C 2 c  c D 1;
(6.3)
c  c D cc  ; ac D ca; ac  D c  a:

Lemma 6.2.2. Let A be a -algebra and v 2 M2 .A/. If  2 Œ1; 1 n f0g, then

the following conditions are equivalent:
 0 0

i) v D ac 0 c
a 0 for some a0 ; c 0 2 A;
 
N 1 , where F D
ii) v D F vF 0 1
1 0 .
 
v22 
v21

N 1 D
Proof. A short calculation shows F vF 
1 v12 
v11
. 

Proposition 6.2.3. For each  2 Œ1; 1 n f0g, there exists a unique -homomor-
phism  W SU .2/ ! SU .2/ ˝ SU .2/ such that .SU .2/; ; u/ is a C  -alge-
braic compact matrix quantum group, and  is determined by

.a/ D a ˝ a  c  ˝ c; .c/ D c ˝ a C a ˝ c: (6.4)

144 Chapter 6. Examples of compact quantum groups
P
Proof. The matrix v 2 M2 .SU .2/ ˝ SU .2// given by vij D k uik ˝ ukj is
unitary (Lemma 6.1.12) and has the form
 0 
a c 0 a0 D a ˝ a  c  ˝ c;
vD 0 0 ; where 0
c a c D c ˝ a C a ˝ c:

The universal property of SU .2/ yields a -homomorphism

 W SU .2/ ! SU .2/ ˝ SU .2/

such that .a/ D a0 and .c/ D c 0 . Thus, condition i) of Definition 6.1.1 is satis-
fied. Condition ii) – invertibility of uN – follows from Lemma 6.2.2, and condition
iii) is trivially satisfied. 

Definition 6.2.1 and Proposition 6.2.3 yield an entire family .SU .2//2.0;1 of
compact matrix quantum groups which can be considered as a continuous defor-
mation of SU1 .2/ Š C.SU2 .C// [17]. From now on, we fix :

Assumption. The deformation parameter  is some fixed element of Œ1; 1 n f0g.

Remark 6.2.4. Let .A; A / be a C  -bialgebra with a unitary corepresentation

matrix v 2 M2 .A/. Denote by .e1 ; e2 / the standard basis of C2 and put ekl WD
ek ˝el 2 C2 ˝C2 for k; l D 1; 2. Then the element det WD e12 e21 2 C2 ˝C2
can be considered as a -deformed determinant, and the following conditions are
equivalent:
 0 0

i) v D ac 0 c
a 0 for some a0 ; c 0 2 A;

ii) the corepresentation ı W C2 ˝ C2 ! C2 ˝ C2 ˝ A corresponding to v  v

preserves det in the sense that ı.det  / D det ˝1.

Indeed, writing ı and det in terms of the basis e11 ; e12 ; e21 ; e22 , one can check:
0 10 10 1 0 1
v11 v12 v11 v12 0 0
B v11 v12 C Bv21 v22 CB 1 C B 1 C
ii) , B
@v21
CB
A@
CB C D B C
v22 v11 v12 A @A @A
v21 v22 v21 v22 0 0
0 10 1 0  
1 0 1
v11 v12 0 v11 v21 0
Bv21 v22 CB 1 C B   CB C
,B C B C D B  v11 v21 C B 1 C , i):
@ v11 v12 A @A @v12 
v22 A @A
 
v21 v22 0 v12 v22 0
 6.2. The compact quantum group SU .2/ 145

Denote by SU0 .2/  SU .2/ the -subalgebra generated by a and c, and
by 0 the restriction of  to SU0 .2/. Then .SU0 .2/; 0 ; u/ is the algebraic
compact matrix quantum group associated to .SU .2/; ; u/ (Proposition 6.1.8).
By Remark 6.1.2 ii), the counit and the antipode of .SU0 .2/; 0 / are given by

0 .a/ D 1; 0 .c/ D 1;
(6.5)
S0 .a/ D a ; S0 .a / D a; S0 .c/ D c; S0 .c  / D 1 c  :
 

For each k 2 Z and m; n 2 N, put

(
ak c m c n ; k 0;
akmn WD
a.k/ c m c n ; k < 0:

Proposition 6.2.5. The family .akmn /k;m;n is a basis of SU0 .2/.

Proof. Using equation (6.3), one can check that the family .akmn /k;m;n spans
SU0 .2/. We show that this family is linearly independent. Let H be a Hilbert
space with orthonormal basis .er;s /r2N;s2Z , and define a0 ; c 0 2 L.H / by
p
a0 er;s WD 1  2r er1;s ; c 0 er;s WD r er;sC1 for all r 2 N; s 2 Z:

Straightforward calculations show that a0 and c 0 satisfy the relations (6.3). Hence
there exists a -homomorphism
W SU .2/ ! L.H / mapping a to a0 and c to c 0 .
A short calculation shows that for all k 2 Z, m; n 2 N, and r k,
 jkj1
Y q 

.akmn /er;0 D 1 2rCsgn.k/2l r.mCn/ erk;nm :
lD0
P
Consider a non-trivial finite linear combination x D k;m;n kmn akmn 2 SU0 .2/.
We claim that
.x/ ¤ 0. Indeed, if k; m; n are chosen such that kmn ¤ 0 and
k 0 m0 n0 D 0 whenever m0 C n0 < m C n, then
1
lim herk;nm j
.x/er;0 i D kmn ¤ 0: 
r!1 r.mCn/

6.2.2 Corepresentations and their weights

The corepresentation theory of the compact quantum group SU .2/ is very similar
to the representation theory of the compact Lie group SU2 .C/. It can be determined
either via a differential calculus or via harmonic analysis and special q-functions.
We shall follow the first approach, which is the original one of Woronowicz [194];
for the second approach, see, for example, [80, Section 4.3].
146 Chapter 6. Examples of compact quantum groups

In this subsection, we formulate the main result of Woronowicz in terms of

weight functions of corepresentations and draw some immediate conclusions con-
cerning tensor products and characters of irreducible corepresentations. The proof
of the main result, which involves the differential calculus mentioned above, is
given in the next section.
Let us start with the definition of the weight function associated to a corepre-
sentation of SU .2/. This definition involves an analogue of the maximal torus of
the compact Lie group SU2 .C/:

Proposition 6.2.6. i) Denote by z 2 C.T / the identity function. There exists a

unique morphism of compact quantum groups
T W SU .2/ ! C.T / such that

T .a/ D z and
T .c/ D 0.
ii) For each  2 T , there exists a unique -homomorphism  W SU .2/ ! C
such that  .a/ D  and  .c/ D 0. For all ;  0 2 T , one has    0 D   0 .

Proof. Immediate from the definition of SU .2/ and equation (6.4). 

Let v be a corepresentation operator of SU .2/ on a finite-dimensional Hilbert

space H . Then .id ˝
T /.v/ is a corepresentation operator of C.T /, which corre-
sponds to a representation

v W T ! L.H /;  7! .id ˝ /.v/;

(see Example 5.2.5). Since every representation of T is a direct sum of represen-

tations of the form  7!  k , where k 2 Z, the space H decomposes into a direct
sum
M
H Š Hk ; where Hk WD f 2 H j v ./ D  k  for all  2 T g:
k2Z

For each k 2 Z, we call Hk the kth weight space of v. If Hk ¤ 0, we call k a

weight of v. The maximal k 2 Z that is a weight for v is called the highest weight
of v. We shall see that up to unitary equivalence, v is completely determined by
the weight function

Wv W Z ! N; k 7! dim Hk :

Lemma 6.2.7. Let v and w be finite-dimensional corepresentation operators of

.SU .2/; /. Then for all k 2 Z,
X
Wvw .k/ D Wv .k/ C Ww .k/; Wvw .k/ D Wv .p/Ww .q/:
pCqDk
 6.2. The compact quantum group SU .2/ 147

Proof. For each k 2 Z, denote by Hk , Kk , .H ˚K/k , and .H ˝K/k the kth weight
space of v, w, v  w, and v  w, respectively. The first equation above follows
from the evident relation .H ˚L K/k D Hk ˚ Kk . The second one P will follow
once we proved .H ˝ K/kP D pCqDk Hp ˝ Kq . Writing v D r fr ˝ vr 2
L.H / ˝ SU .2/ and w D s gs ˝ ws 2 L.K/ ˝ SU .2/, we find that for each
 2 T,
X X
vw ./ D .fr ˝ gs / .vr ws / D fr  .vr / ˝ gs  .ws / D v ./ ˝ w ./:
r;s r;s
L
The relation pCqDk Hp ˝ Kq  .H ˝ K/k follows easily. Summing over k and
comparing dimensions, we find that this inclusion is an equality for each k. 
The key to the corepresentation theory of SU .2/ is the following result:
Theorem 6.2.8. Up to equivalence, there exists for each n 2 N precisely one
irreducible corepresentation matrix u.n/ 2 MnC1 .SU .2//, and the associated
weight function W.n/ WD Wu.n/ is given by
(
1; k 2 fn; 2  n; : : : ; n  2; ng;
W.n/ .k/ D (6.6)
0; otherwise:
The proof of this theorem is outlined in the next section.
Corollary 6.2.9. i) Up to equivalence, every finite-dimensional corepresentation
of .SU .2/; / is completely determined by its weight function.
ii) For each n 2 N, the comultiplication  restricts to a corepresentation ı.n/
on V.n/ WD spanfak c l j k C l D ng  SU0 .2/ that is equivalent to u.n/ .
iii) u.1/ ' u.
iv) u.m/  u.n/ ' u.jmnj/  u.jmnjC2/      u.mCn2/  u.mCn/ for all
m; n 2 N.
Proof. i) Since every corepresentation is equivalent to a direct sum of irreducible
ones (Theorem 3.2.1) and the weight function behaves additively (Lemma 6.2.7),
it suffices to show that the coefficients dn of a finite linear combination W D
P
n dn W.n/ can be reconstructed from W . But equation (6.6) implies that dn D
W .n/  W .n C 2/ for all n 2 N.
ii) Equation (6.4) shows that .V.n/ /  V.n/ ˝ SU .2/ and

.id ˝
T /..ak c l // D ak c l ˝ z kl for all k; l 2 N:
Consequently, ı.n/ is a corepresentation with weight function W.n/ .
iii) This follows immediately from ii) and equation (6.4).
iv) It suffices to check that the weight functions of the left- and of the right-hand
side are equal, and this can easily be verified using Lemma 6.2.7. 
148 Chapter 6. Examples of compact quantum groups

The characters .n/ WD .u.n/ / of the corepresentations u.n/ (n 2 N) enjoy the

following properties:

Proposition 6.2.10. i) .0/ D 1, .1/ D a C a , and for all n; m 2 N,

.m/ .n/ D .jmnj/ C .jmnjC2/ C    C .mCn2/ C .mCn/ : (6.7)

ii) For each n 2 N,

nC1
 
sin t a C a
.n/ D 21
; where t WD 2 arccos : (6.8)
sin 2 t 2

iii) The unital homomorphism CŒX  ! SU .2/ given by X 7! .1/ is injective.
iv) For each n 2 N, there exists a unique fn 2 CŒX  of degree n such that
.n/ D fn ..1/ /. The family .fn /n2N is a basis of CŒX .

Proof. i) .1/ D .u/ D a C a because u.1/ ' u. The decomposition formula

for the product .m/ .n/ follows from Corollary 6.2.9 iv) and Proposition 3.2.14.
ii) For each n 2 N, put s.n/ WD sin..n C 1/t =2/. We have to show that
s.n/ =s.0/ D .n/ for all n 2 N. Using the relations

sin.2y/ D 2 sin.y/ cos.y/; sin.x ˙ y/ D sin.x/ cos.y/ ˙ cos.x/ sin.y/

for x D .n C 1/t =2, y D t =2, and inserting 2 cos.t =2/ D a C a D .1/ , we find

s.1/ D 2s.0/ cos.t =2/ D s.0/ .1/ ; s.nC1/ C s.n1/ D 2s.n/ cos.t =2/ D s.n/ .1/ :

On the other hand, equation (6.7) implies .nC1/ C.n1/ D .n/ .1/ for all n 2 N.
Hence .n/ D s.n/ =s.0/ for all n 2 N.
iii) By Proposition 6.2.5, the powers n.1/ D .a C a /n .n 2 N/ are linearly
independent.
iv) Equation (6.7) shows that spanf.n/ j n 2 Ng D spanfn.1/ j n 2 Ng. This
relation implies the existence of polynomials fn 2 CŒX  such that fn ..1/ / D .n/
for all n 2 N and spanffn j n 2 Ng D CŒX . Uniqueness of .fn /n follows
from iii), and linear independence from the fact that the family ..n/ /n is linearly
independent (see Proposition 3.2.15). 

Proposition 6.2.11. For each f 2 C0 .R/,

Z 2 p
1
h.f ..1/ // D f .x/ 4  x 2 dx:
2
2
 6.2. The compact quantum group SU .2/ 149

Proof. By the Stone–Weierstrass Theorem, it suffices to prove the assertion for all
the functions fn 2 CŒX  (n 2 N) defined in Proposition 6.2.10 iv). By Proposi-
tion 3.2.15,
h.fn ..1/ // D h..0/ .n/ / D ı0;n :
We substitute t D 2 arccos.x=2/, x D 2 cos.t =2/, use equation (6.8), and get
Z 2 Z 0

1 p 1 sin nC1 t

fn .x/ 4  x dx D
2 1
 2j sin 12 t j   sin 12 t dt
2
2
2 2
2 sin 2 t
Z
1 0

D sin nC1 2
t sin 12 t dt D ı0;n : 

2

6.2.3 Corepresentations and differential calculi

In this subsection, we outline a proof of Theorem 6.2.8, following the approach
of Woronowicz [194]. The basic idea behind his approach is to use a differential
calculus for SU .2/ in a similar way as one uses the Lie algebra su2 to determine
the irreducible representations of SU2 .C/.
Let us briefly recall the classical situation. Every continuous representation
of
SU2 .C/ is smooth and determined by the associated representation D
of the Lie
algebra su2 . The complexification of this Lie algebra is spanned by the matrices
     
1 0 0 1 0 0
H WD ; X WD ; Y WD ;
0 1 0 0 1 0
which satisfy ŒH; X  D 2X, ŒY; H  D 2Y , ŒY; X  D H , see [55, Lecture 26], [61,
Section 2.9]. Thus
is determined by the infinitesimal generators A0 WD D
.X /,
A1 WD D
.H /, A2 WD D
.Y /, and the preceding commutator relations and the
unitarity of
imply
ŒA1 ; A0  D 2A0 ; ŒA2 ; A1  D 2A2 ; ŒA2 ; A0  D A1 ;
(6.9)
A1 D A1 ; A0 D A2 :
The analysis of these relations, which is straightforward and explained in almost
every book on Lie groups or Lie algebras, leads directly to the classification of all
irreducible representations of SU2 .C/.
Let us now turn to the quantum group SU .2/ and construct a replacement
for the Lie algebra su2 . The basic ingredients are algebra homomorphisms f0 , f1 ,
f2 W SU0 .2/ ! C and linear maps 0 ; 1 ; 2 W SU0 .2/ ! C, which are determined
by
f0 .a/ D f2 .a/ D 1 ; f0 .a / D f2 .a / D ; f1 .a/ D 2 ; f1 .a / D 2 ;
fk .c/ D fk .c  / D 0 for k D 0; 1; 2;
150 Chapter 6. Examples of compact quantum groups

and
0 .c  / D 1 ; 1 .a / D 2 ; 1 .a/ D 1; 2 .c/ D ;
k .x/ D 0 for all other cases where k 2 f0; 1; 2g; x 2 fa; a ; c; c  g;
k .xy/ D k .x/fk .y/ C 0 .x/k .y/ for all x; y 2 SU0 .2/:
For the precise construction of these maps, see [194].
Theorem 6.2.12. Denote by  the free left module over SU0 .2/ with three gener-
ators !1 , !2 , !3 .
i)  is a bimodule over SU0 .2/ with respect to the right multiplication given
by !k x WD .fk  x/!k for all k D 0; 1; 2 and x 2 SU0 .2/.
P
ii) The map d W SU0 .2/ !  given by x 7! k .k  x/!k is a derivation, that
is, d.xy/ D d.x/y C xd.y/ for all x; y 2 SU0 .2/. Moreover, d.x/ D 0 if
and only if x 2 C1.
iii) There exists a unique involution on  such that .xd.y/z/ D z  d.y  /x  for
all x; y; z 2 SU0 .2/. Moreover, !0 D !2 , !1 D !1 , !2 D 1 !0 .
Proof. See [194, Section 2]. 
The next step is the construction of an n-th order differential calculus for arbitrary
n 2 N. Let  0 WD SU0 .2/ and put  ˝n WD  ˝SU0 .2/    ˝SU0 .2/  (n factors )
L1
for n 1. Then  ˝ WD nD0 
˝n
carries a natural structure of a -algebra.
˝
Denote by S   the -ideal generated by the elements
!0 ˝ !0 ; !1 ˝ !1 ; !2 ˝ !2 ;
!2 ˝ !0 C  !0 ˝ !2 ; !1 ˝ !0 C 4 !0 ˝ !1 ; !2 ˝ !1 C 4 !1 ˝ !2
2

(here, we omitted the subscripts at the tensor symbols), and put  ^ WD  ˝ =S .

Since S is spanned by homogeneous elements,  ^ is a graded -algebra again, that
is,  ^ is a direct sum of the subspaces  ^n D  ˝n =. ˝n \ S /, where n 2 N, and
 ^n  ^m   ^.nCm/ , . ^n / D  ^n for all n; m 2 N. Evidently,  ^0 D SU0 .2/
and  ^1 D . We denote the multiplication in  ^ by the symbol “^”, and call
elements ! 2  ^n , where n 2 N, homogeneous of degree j!j D n.
Theorem 6.2.13. i)  ^ is graded-commutative in the sense that ! ^ ! 0 D
0
.1/j!jj! j ! 0 ^ ! for all homogeneous !; ! 0 2  ^ .
ii) The map d W SU0 .2/ !  ^1 extends uniquely to a -linear map d W  ^ !
 such that d ı d D 0 and for all n 2 N and all homogeneous !; ! 0 2  ^ ,
^

d. ^n /   ^.nC1/ ; d.! ^ ! 0 / D d.!/ ^ ! 0 C .1/j!j ! ^ d.! 0 /:

iii) d.!0 / D !0 ^!1 =.2 C1/, d.!1 / D !0 ^!2 , d.!2 / D !1 ^!2 =.2 C1/.
 6.2. The compact quantum group SU .2/ 151

Now we apply the differential calculus that we have just developed to the study
of corepresentations. To every corepresentation matrix v 2 Mn .SU0 .2//, where
n 2 N, we associate infinitesimal generators
Avk WD .k .vij //i;j 2 Mn .C/; where k D 0; 1; 2;
P
and an element Av WD k Avk !k 2 Mn ./.
For each complex vector space V , let us identify Mn .V / with Mn .C/ ˝ V .
Proposition 6.2.14. Let v 2 Mn .SU0 .2// Š Mn .C/ ˝ SU0 .2/ be a unitary co-
representation matrix.
i) .id ˝d /.v/ D vAv , .id ˝d /.v  / D Av v  , and Av Av D .id ˝d /.Av /,
.Av / D Av .
ii) Let w 2 Mm .SU0 .2// be a corepresentation matrix and T 2 Mm;n .C/.
Then T v D wT , TAv D Aw T .
Proof. i) These equations follow, in the order in which they are listed, from the
relations
X X X
dvij D .k  vij /!k D vil k .vlj /!k D .vAvk /ij !k for all i; j;
k k;l k

0 D .id ˝d /.v  v/ D .id ˝d /.v  /v C v  .id ˝d /.v/ D .id ˝d /.v  /v C Av ;

0 D .id ˝d /2 .v/ D .id ˝d /.v/Av C v.id ˝d /.Av / D v.Av Av C .id ˝d /.Av //;
.Av / v  D .vAv / D .id ˝d /.v/ D .id ˝d /.v  / D Av v  :
ii) First, wT v  2 Mm;n .C/ if and only if TAv D Aw T because both conditions
are satisfied if and only if the following expression is 0:
.id ˝d /.wT v  / D .id ˝d /.w/T v  C wT .id ˝d /.v  / D w.Aw T  TAv /v  :
But if wT v  2 Mm;n .C/, then T D .id ˝ /.w/T .id ˝ /.v/ D .id ˝ /.wT v  / D
wT v  because .id ˝ /.v/ D 1 and .id ˝ /.w/ D 1. 
Theorem 6.2.15. Let v 2 MnC1 .SU0 .2// be a unitary corepresentation matrix.
Then the infinitesimal generators Av0 ; Av1 ; Av2 satisfy
Av2 Av0  1 Av0 Av2 D Av1 ;
.Av0 / D Av2 ;
2 Av1 Av0  2 Av0 Av1 D .1 C 2 /Av0 ;
.Av1 / D Av1 :
2 Av2 Av1  2 Av1 Av2 D .1 C 2 /Av2 ;
If v is irreducible, there exists a basis .n ; nC2 ; : : : ; n2 ; n / of CnC1 such that
Av0 k D ckC1 kC2 ; Av1 k D dk k ; Av2 k D ck k2 for all k;
p
where dk D .2k  1/=.2  1/, ck D .k  n /.n  2k /=.2  1/
if jj < 1, and dk D 2k, ck D ..n C 1/.n  k C 1//1=2 if jj D 1.
152 Chapter 6. Examples of compact quantum groups

Proof. The relations for the infinitesimal generators follow easily from Proposi-
tion 6.2.14. A careful analysis of these relations, which can be found in [194], leads
to the explicit form given in the second part of the theorem. 

Note that for  D 1, the relations on the infinitesimal generators given above
coincide with the relations given in equation (6.9).
The explicit form of an irreducible corepresentation given above allows us to
compute the associated weight function and thereby prove the key result of the
preceding section:

Proof of Theorem 6.2.8. Let v 2 MnC1 .SU0 .2// Š MnC1 .C/ ˝ SU0 .2/ be an
irreducible unitary corepresentation matrix. For each weight k of v, denote
P by pk the
orthogonal projection onto the kth weight space. Then .id ˝
T /.v/ D k pk ˝z k .
Assume that jj < 1. Define a linear map  W CŒz; z 1  ! C by z k 7!
2k
.  1/=.2  1/. We claim that 1 D  ı
T . This relation implies
X X
Av1 D .id ˝1 /.v/ D pk .z k / D pk .2k  1/=.2  1/;
k k

and comparing with the description of Av1 given in Theorem 6.2.15, we find
(
1; k 2 fn; 2  n; : : : ; n  2; ng;
Wv .k/ D dim.Im pk / D
0; otherwise:

So it suffices to prove that 1 D  ı

T . We verify this equation for the basis
elements .akpq /k;p;q (see Proposition 6.2.5). Let k 2 Z and p; q 2 N. By
[194, Equation (2.5)], 1 .akpq / D .f1 .akpq /  0 .akpq //=.2  1/. Using
the multiplicativity of f1 and 0 , we easily find f1 .akpq / D ıp;0 ıq;0 2k and
0 .akpq / D ıp;0 ıq;0 . Therefore,

1 .akpq / D ıp;0 ıq;0 .2k  1/=.2  1/ D ıp;0 ıq;0 .z k / D .

T .akpq //:

The case jj D 1 is treated similarly. 

6.2.4 Modular properties of the Haar state

The results on the corepresentation theory obtained in the preceding sections al-
low us to describe the Haar state of SU .2/ and the modular characters .fz /z2C
introduced in Theorem 3.2.19:
Proposition 6.2.16. i) The automorphisms .fz /z2C are determined by

fz .a/ D jjz ; fz .a / D jjz ; fz .c/ D 0; fz .c  / D 0 for all z 2 C:

 6.2. The compact quantum group SU .2/ 153

ii) For all z 2 C,

fz  a D jjz a D a  fz ; fz  a D jjz a D a  fz ;
fz  c D jjz c; c  fz D jjz c; fz  c  D jjz c  ; c   fz D jjz c  :

Proof. i) By equation (6.5), the matrix S02 .u/ WD .S02 .uij //i;j is equal to
 
a 2 c 
:
2 c a

A short calculation shows that the matrix F WD diag.jj1 ; jj/ satisfies u D

F 1 S02 .u/F . Moreover, Tr F > 0. By the proof of Theorem 3.2.19, fz .uij / D
.F z /ij for i; j D 1; 2. The claim follows.
ii) Combine i) and equation (6.4). 
The next proposition describes the values of the Haar state on the basis elements
.akmn /k;m;n , which were defined by
(
ak c m c n ; k 0;
akmn WD .k/ m n
a c c ; k < 0:

Theorem 6.2.17. The Haar state h of SU .2/ is given by

h.akmn / D ık;0 ım;n .1  2 /=.1  2mC2 /: (6.10)

For every function f 2 C. .c  c//,

1
X
h.f .c  c// D .1  2 / 2k f .2k /: (6.11)
kD0

Proof. Let k 2 Z, m; n 2 N, and z 2 C. Using Proposition 6.2.16 ii) and

multiplicativity of fz , we find

fz  akmn D jj.kCmn/z akmn ; akmn  fz D jj.kmCn/z akmn :

On the other hand,

h.fz  akmn / D .h ˝ fz /..akmn // D fz .1/h.akmn / D h.akmn /

and similarly h.akmn  fz / D h.akmn /. Consequently, h.akmn / D 0 unless k D 0

and m D n.
It remains to compute the value of h on the elements bm WD c m c m for all
m 2 N. By Theorem 3.2.19 v),

2mC2 h.aa bm / D 2mC2 h.a bm .f1  a  f1 // D 2m h.a bm a/;

154 Chapter 6. Examples of compact quantum groups

and from equation (6.3), one easily deduces

2mC2 aa bm D 2mC2 bm  2mC4 bmC1 ; 2m a bm a D bm  bmC1 :

Combining these relations, we find h.bm /.2mC2  1/ D h.bmC1 /.2mC4  1/.

Using the relation h.b0 / D h.1/ D 1, we obtain equation (6.10). This equation
implies that equation (6.11) holds whenever f has the form f .x/ D x m for some
m 2 N. Since such functions are linearly dense in C. .c  c//, equation (6.11)
holds for all f 2 C. .c  c//. 

6.3 Products of compact quantum groups

The free product and the tensor product of compact quantum groups were intro-
duced by Wang [185], [186], [187]. We briefly outline these general constructions,
describe the irreducible corepresentations of a product in terms of the irreducible
corepresentations of the two factors, and compare the algebraic and C  -algebraic
versions.

Free products of compact quantum groups

Let A and B be unital algebras. The free .unital/ product of A and B is a unital
algebra A  B with embeddings A W A ,! A  B and B W B ,! A  B such that
for each unital algebra C and each pair of unital homomorphisms f W A ! C and
g W B ! C , there exists a unique unital homomorphism f  g W A  B ! C such
that .f g/ıA D f and .f g/ıB D g, that is, the following diagram commutes:
A B
A FF / A  B o x
B
FF xxg
f
FFF f g xx x
FF x
"  |xx
C.

Routine arguments show that A  B is uniquely determined up to isomorphism, that

Aop  B op Š .A  B/op , and that A  B carries a unique structure of a -algebra
such that the embeddings A and B become -homomorphisms whenever A and B
are -algebras.
The free product of unital C  -algebras is defined similarly – just replace algebras
and homomorphisms by C  -algebras and -homomorphisms.
We adopt the following notation: for each n 2 N, we denote by QA W Mn .A/ !
Mn .A  B/ the map given by .uij /i;j 7! .A .uij //i;j . Likewise, we define QB .
Throughout this subsection, we treat the algebraic and the C  -algebraic setting
in parallel and use the symbol “˝” to denote the algebraic tensor product of -
algebras and the minimal tensor product of C  -algebras.
 6.3. Products of compact quantum groups 155

Theorem 6.3.1. Let .A; A / and .B; B / be algebraic or C  -algebraic compact

quantum groups. Then there exists a unique comultiplication  on A  B such that
.A; A /  .B; B / WD .A  B; / is a compact quantum group and the maps A
and B are morphisms of compact quantum groups.
Proof. By the universal property of A  B, there exists a unique -homomorphism
 W A  B ! .A  B/ ˝ .A  B/ such that the following diagram commutes:
A B
A QQQQ /AB o
mB
QQQ mmm
mmm
QQA
.A ˝A /ı
QQQ(  ˝B /ıB
.Bm
 vmmmm
.A  B/ ˝ .A  B/.
The coassociativity of A and B immediately implies that  is coassociative.
If .A; A / and .B; B / are algebraic compact quantum groups, then .A  B; /
is a Hopf -algebra. Indeed, if A , SA and B , SB denote the counit and the antipode
of .A; A / and .B; B /, respectively, then the maps WD A  B W A  B ! C
and S WD .Aop ı SA /  .B op ı SB / W A  B ! Aop  B op Š .A  B/op are a counit
and an antipode for .A  B; /, as one can easily check.
Let us show that .A  B; / is a compact quantum group. The C  -algebra/-
algebra A  B is generated by A .A/ and B .B/, and by Theorem 3.2.12 or 5.3.11,
respectively, A and B are generated by the entries of unitary corepresentation ma-
trices. But if u is a unitary corepresentation matrix of .A; A / or .B; B /, then
QA .u/ or QB .u/, respectively, is a unitary corepresentation matrix of .A  B; /.
Applying Theorem 3.2.12 or Theorem 5.3.11 again, we find that .A  B; / is a
compact quantum group. 
Corollary 6.3.2. If .A; A ; u/ and .B; B ; v/ are compact matrix quantum groups,
then so is .A  B; ; QA .u/  QB .v//. 
The Haar state of .A  B; / turns out to be the free product of the Haar states of
.A; A / and .B; B /. The free product of states was introduced by Voiculescu; it is
a fundamental concept in free probability theory [184]. Let us briefly recall how it
is defined. If  and are states on -algebras/C  -algebras A and B, respectively,
then there exists a unique state   on A  B, called the free product of  and ,
such that
i) .  / ı A D  and .  / ı B D ;
ii) if c1 ; : : : ; cn are elements of ker  or ker and no adjacent elements belong
to the same algebra A or B, then c1 : : : cn 2 ker.  /;
see [184, Chapter 1]. In the algebraic setting, the existence of   is easy to
show; in the setting of C  -algebras, one can define a free product of the GNS-
constructions for  and , which yields a GNS-construction for   . The image
156 Chapter 6. Examples of compact quantum groups

of A  B under the GNS-representation for   is called the reduced free product

of .A; / and .B; /; it is denoted by .A; /  .B; /.
Proposition 6.3.3. Let .A; A / and .B; B / be compact quantum groups with
Haar states hA and hB , respectively. Then the Haar state of the compact quantum
group .A  B; / is equal to the free product hA  hB .
Proof. We only need to show that h WD hA  hB is left-invariant. Let .u˛ /˛
and .v ˇ /ˇ be maximal families of non-trivial pairwise inequivalent irreducible
corepresentation matrices of .A; A / and .B; B /, respectively. Denote by W 
A  B the subset of all elements of the form

w D wi11j1 : : : winnjn 2 A  B; (6.12)

where n 1, each wimmjm is a matrix entry of some QA .u˛ / or QB .v ˇ /, and no
adjacent factors are taken from the same algebra A or B.
By Theorem 3.2.12 or 5.3.11, the unit 1 2 A  B and the set W span (a
dense subspace of) A  B. Evidently, h.1/ D 1 D .id ˝h/..1//. We show that
h.w/ D 0 D .id ˝h/..w// for each w 2 W , and then the claim follows.
ˇ
By Corollary 3.2.7, hA .u˛ij / D 0 D hB .vkl / for all ˛; i; j and ˇ; k; l. Combin-
ing this relation with condition ii) in the definition of the free product of states, we
get h.w/ D 0 for each w 2 W . Moreover, if w is as in (6.12), then
X
.id ˝h/..w// D .wi11k1 : : : winnkn /  h.wk11 j1 : : : wknn jn / D 0: 
k1 ;:::;kn

The irreducible corepresentations of a free product are easily determined. Given

a compact quantum group .A; /, let us call a maximal family of pairwise inequiva-
lent non-trivial unitary irreducible corepresentation matrices a representative family.
Proposition 6.3.4. Let .A; A / and .B; B / be compact quantum groups with
representative families .u˛ /˛ and .v ˇ /ˇ , respectively. Then a representative family
for .A  B; / is given by all corepresentation matrices of the form

w 1      w n ; (6.13)

where n 1, each w m belongs to .QA .u˛ //˛ or .QB .v ˇ //ˇ , respectively, and no
two adjacent factors are taken from the same family.
Proof. Denote by hA and hB the Haar states of the compact quantum groups .A; A /
and .B; B /, and recall from Proposition 6.3.3 that the Haar state h of .A  B; /
is equal to the free product hA  hB .
We show by induction on n that for every corepresentation matrix w as in (6.13),
the associated character .w/ satisfies h..w/ .w// D 1; then w is irreducible
by Corollary 3.2.16.
 6.3. Products of compact quantum groups 157

For n D 1, the assertion is evident because w D w 1 is irreducible. So, let

n > 1 and assume that the assertion holds for n  1. By Proposition 3.2.14, .w/ D
1 : : : n , where k WD .w k / for k D 1; : : : ; n, and by Corollary 3.2.7 and
Proposition 3.2.15, h.k / D 0 D h.k / for k D 1; : : : ; n1, and h.1 1 1/ D 0.
Using condition ii) in the definition of the free product h D hA  hB , we find
h.n : : : 2 .1 1  1/2 : : : n / D 0, and inserting the induction hypothesis, we
obtain

h..w/ .w// D h.n : : : 2 1 1 2 : : : n / D h.n : : : 2 2 : : : n / D 1:

Thus every corepresentation of the form given in (6.13) is irreducible. Evidently,

the entries of the corepresentation matrices given in (6.13) include all elements of
A  B of the form given in (6.12). Therefore these matrix entries and the unit
1 2 A  B span (a dense subspace of) A  B. From Corollary 3.2.8 or 5.3.9, it
follows that every non-trivial irreducible corepresentation matrix of A  B has, up
to equivalence, the form given in (6.13). 
Let us briefly clarify the relation between the algebraic and the C  -algebraic
free product. Recall from Section 5.4 that to every CQG .A; /, one can associate
• an algebraic CQG .A0 ; 0 / D .A; /0 ,
• a universal and a reduced C  -algebraic CQG .Au ; u / D .A; /u and
.Ar ; r / D .A; /r , respectively.
Proposition 6.3.5. Let .A; A / and .B; B / be compact quantum groups.
i) ..A; A /  .B; B //0 Š .A; A /0  .B; B /0 .
ii) ..A; A /  .B; B //u Š .A; A /u  .B; B /u .
iii) Denote by hA and hB the Haar states of .A; A /r and .B; B /r , respectively.
Then .A  B/r Š .Ar ; hA /  .Br ; hB /.
Proof. Assertion i) follows easily from Proposition 6.3.4, ii) from the definitions,
and iii) from Proposition 6.3.3. 
Corollary 6.3.6. Let .A; A / and .B; B / be CQGs, and denote by hA , hB the
Haar states of Ar , Br , respectively. There exists a comultiplication r on .Ar ; hA /
.Br ; hB / such that ..Ar ; hA /  .Br ; hB /; r / is a reduced C  -algebraic CQG and
the natural maps A; B ! .Ar ; hA /  .Br ; hB / are morphisms of CQGs. 

Tensor products of compact quantum groups. To each pair of compact quantum

groups, one can associate a tensor product which is a compact quantum group again.
We briefly summarize this construction; the main results can be proved by similar
techniques as used in the case of a free product.
158 Chapter 6. Examples of compact quantum groups

We adopt the following notation. Let A and B be algebras/C  -algebras. We

denote by A ˝ B the algebraic tensor product/the maximal C  -tensor product, and
.max/
by A ; B W A ! A ˝ B the natural maps given by a 7! a ˝ 1B and b 7! 1A ˝ b,
.max/ .max/ .max/
respectively. As before, the minimal tensor product of C  -algebras C and D is
denoted by C ˝ D. For each n 2 N, we denote by QA W Mn .A/ ! Mn .A ˝ B/
.max/
the map given by .uij /i;j 7! .A .uij //i;j . Similarly we define QB . A representative
family for a compact quantum group is a maximal family of pairwise inequivalent
non-trivial unitary irreducible corepresentation matrices.

Proposition 6.3.7. Let .A; A / and .B; B / be compact quantum groups.

i) There exists a unique comultiplication  on A ˝ B such that

.max/

.A; A / ˝ .B; B / WD .A ˝ B; /
.max/ .max/

is a compact quantum group and A and B are morphisms of CQGs.

ii) Denote by hA and hB the Haar states of .A; A / and .B; B /, respectively.
Then the Haar state of .A ˝ B; / is hA ˝ hB .
.max/ .max/

iii) If .u˛ /˛ and .v ˇ /ˇ are representative families for .A; A / and .B; B /,
respectively, then .QA .u˛ /  QB .v ˇ //˛;ˇ is a representative family for .A ˝
.max/
B; /.

iv) If .A; A ; u/ and .B; B ; v/ are CMQGs, then .A ˝ B; ; QA .u/  QB .v//
.max/
is a CMQG.

v) In the notation introduced before Proposition 6.3.5,

(a) ..A; A / ˝ .B; B //0 Š .A; A /0 ˝ .B; B /0 ;

.max/ .max/

(b) ..A; A / ˝ .B; B //u Š .A; A /u ˝ .B; B /u ;

.max/ .max/

(c) Ar ˝ Br Š .A ˝ B/r . 
.max/

Corollary 6.3.8. Let .A; A /, .B; B / be CQGs. Then there exists a unique
comultiplication r on Ar ˝ Br such that .Ar ˝ Br ; r / is a C  -algebraic CQG
and the natural maps A; B ! Ar ˝ Br are morphisms of CQGs. 
 6.4. The free unitary and the free orthogonal quantum groups 159

6.4 The free unitary and the free orthogonal quantum groups
The free unitary quantum groups are universal compact matrix quantum groups
which contain every other compact matrix quantum group as a quantum subgroup,
very much like the unitary groups contain every other compact Lie group as a Lie
subgroup. They were introduced by Wang [185], [186] and generalized by Wang
and Van Daele [179]. Wang and Van Daele also introduced the free orthogonal
quantum groups, whose definition was slightly modified by Banica [12]. In this
section, we shall only give a brief overview of the definition and some fundamental
properties.

Definition and first properties. The free unitary and the free orthogonal quantum
groups are defined as follows:
Definition 6.4.1. i) For each F 2 GLn .C/, we denote by Au .F / the universal
unital C  -algebra generated by elements uij (1  i; j  n) such that the matrices
N 1 are unitary.
u D .uij /i;j and F uF
ii) We call a matrix F 2 GLn .C/ admissible if F Fx 2 R1. For each admissible
F 2 GLn .C/, we denote by Ao .F / the universal unital C  -algebra generated by
elements vij (1  i; j  n) such that the matrix v D .vij /i;j is unitary and
v D F vFN 1 .
N 1 is unitary if and only
Remarks 6.4.2. i) In Definition 6.4.1 i), the matrix F uF
if

ut .F  F /u.F
N  F /1 D 1 D .F  F /u.F
N  F /1 ut :

In particular, Au .F / D Au .F 0 / whenever F  F D F 0 F 0 . Therefore, some au-

thors denote the C  -algebra Au .F / defined above by Au .Q/, where Q D F  F .
N 1 if and only
ii) In Definition 6.4.1 ii), the unitary matrix v satisfies v D F vF
  1 t 
if v D .F / v F , and this relations holds if and only if

v t F  v.F  /1 D 1 D F  v.F  /1 v t :

iii) Evidently, Ao .F / D Ao .F / for each admissible F and each  2 C,  ¤ 0.

The C  -algebras Au .F / and Ao .F / are compact matrix quantum groups:
Proposition 6.4.3. i) For each F 2 GLn .C/, the C  -algebra Au .F / carries a
unique comultiplication u that turns .Au .F /; u ; u/ into a C  -algebraic compact
matrix quantum group.
ii) For each admissible F 2 GLn .C/, the C  -algebra Ao .F / carries a unique
comultiplication o that turns .Ao .F /; o ; v/ into a C  -algebraic compact matrix
quantum group.
160 Chapter 6. Examples of compact quantum groups
P

Proof. i) For each w 2 Mn .Au .F //, put n .w/ WD k wik ˝ wkj i;j . By the
universal property of Au .F /, it suffices to show that n .u/ and F n .u/F 1 are
unitary. But by Lemma 6.1.12, n .u/ and n .F uF N 1 / are unitary, and a similar
N 1 / D F n .u/F 1 .
calculation as in equation (6.1) shows that n .F uF
ii) The proof is similar as for i). 
Remark 6.4.4. Let .A; ; w/ be a compact matrix quantum group and assume
that w D F wF N 1 for some F 2 GLn .C/. Then w D F F wF N 1 F 1 D
x x 1 x
.F F /w.F F / . If F F 62 C1, then by Schur’s Lemma (Proposition 5.3.4), w
is reducible. Moreover, it is easy to see that F Fx 2 C1 implies F Fx 2 R1. These
observations motivate the restriction to admissible F 2 GLn .C/ in the definition
of Ao .F /.
The quantum groups Au .F / are universal in the following sense:
Proposition 6.4.5. Let .A; ; w/ be a C  -algebraic compact matrix quantum
group, where w 2 Mn .A/, n 2 N.
i) There exist an F 2 GLn .C/ and a -homomorphism f W Au .F / ! A such
that f .uij / D wij for all i; j . This f is surjective and a morphism of compact
matrix quantum groups.
ii) If w is irreducible and w ' w,
N then there exist an admissible F 2 GLn .C/
and a -homomorphism f W Ao .F / ! A such that f .vij / D wij for all i; j .
This f is surjective and a morphism of compact matrix quantum groups.
Proof. i) By Theorem 5.3.3 i), wN is equivalent to a unitary corepresentation matrix.
Hence there exists an F 2 GLn .C/ such that F wF N 1 is unitary. Now the existence
of f follows from the universal property of Au .F /.
ii) The proof is similar to the proof of i), see also Remark 6.4.4. 
Definition 6.4.6. For each F 2 GLn .C/, we call .Au .F /; u ; u/ the free unitary
C  -algebraic quantum group with parameter F . For each admissible F 2 GLn .C/,
we call .Ao .F /; o ; v/ the free orthogonal C  -algebraic quantum group with pa-
rameter F .
The next result clarifies to which extent .Au .F /; u ; u/ and .Ao .F /; o ; v/
depend on F . To avoid ambiguities, let us index the generators uij , vij and the
comultiplications u , o by F .
Proposition 6.4.7. i) Let F 2 GLn .C/ and F 0 D F U for some U 2 Un .C/. Then
0 F0 F0
.Au .F /; Fu ; u / ' .Au .F /; u ; u /.
F

u ; u / depends only on the list of eigenvalues

ii) Up to similarity, .Au .F /; F F

.with multiplicity/ of F F .
iii) Let F 2 GLn .C/ be admissible and U 2 Un .C/. Then F 0 WD UF U t is
0 F0 F0
admissible and .Ao .F /; Fo ; v / ' .Ao .F /; o ; v /.
F
 6.4. The free unitary and the free orthogonal quantum groups 161

Proof. i) Since the matrices Ux , uF , F 0 uF 0 .F 0 /1 are unitary, so are w WD Ux uF Ux 

0 0

and F wF N 1 D F U uF U  F 1 . Consequently, we can define a -homomorphism

0

f W Au .F / ! Au .F 0 / such that f .uF ij / D wij for all i; j . By similar arguments,

we can define a -homomorphism Au .F 0 / ! Au .F / that is inverse to f . There-
fore, f is an isomorphism.
ii) This follows directly from i).
iii) The matrix F 0 is admissible because F 0FS0 D UF U t Ux FxU  D UF FxU  2
0
R1. Moreover, the matrix w WD U  v F U is unitary and satisfies F wF N 1 D
0
U  F 0 v F F 01 U D U  v F U D w. Consequently, we can define a -homo-
0

morphism f W Ao .F / ! Ao .F 0 / such that f .vijF / D wij for all i; j . By similar

arguments, we can define a -homomorphism Ao .F 0 / ! Ao .F / that is inverse
to f . The claim follows. 
For each admissible matrix F 2 GL2 .C/, the compact matrix quantum group
.Ao .F /; o ; v/ coincides with the compact matrix quantum group .SU .2/; ; u/
introduced in Section 6.2:
 
Proposition 6.4.8. For each  2 Œ1; 1 n f0g, the matrix F WD 01 01 is
admissible and .SU .2/; ; u/ D .Ao .F /; o ; v/. Moreover,

f.Ao .F /; o ; v/ j F 2 GL2 .C/ admissibleg

D f.SU .2/; ; u/ j  2 Œ1; 1 n f0gg:

Proof. This follows from Lemma 6.2.2, Proposition 6.4.7 iii), and elementary linear
algebra, see [13, Proposition 7]. 

The corepresentation theory of the free orthogonal quantum group. The co-
representation theory of the free orthogonal quantum group .Ao .F /; o / is very
similar to the corepresentation theory of .SU .2/; /. It was determined by Banica
[12]. We shall outline the main results and the steps of the proofs.
We fix an admissible F 2 GLn .C/, where n 2 N. To determine the corepre-
sentation theory of .Ao .F /; o ; v/, we shall use the results on the corepresentation
theory of the C  -algebraic compact matrix quantum group .SU .2/; ; u/ that were
obtained in Section 6.2. From now on, u denotes the fundamental corepresentation
matrix of SU .2/.
Proposition 6.4.9. dim Hom.v r ; v r /  dim Hom.ur ; ur / for all r 2 N.
Proof. We outline the main steps of the proof; for details, see [12]. Let us fix some
notation. We put H WD Cn and define
• Ip WD idH ˝p for all p 2 N;
P
• E W C ! H ˝ H by E.1/ WD i;j Fj i ei ˝ ej ;
162 Chapter 6. Examples of compact quantum groups

• Vp;q WD Ip ˝ E ˝ Iq W H ˝.pCq/ ! H ˝.pC2Cq/ for all p; q.

Step 1. Denote by Z the category with objects H ˝r (r 2 N) and morphism
sets Z.H ˝r ; H ˝s /  L.H ˝r ; H ˝s / consisting of all finite linear combinations

of compositions of maps of the form Ip , Vp;q , Vp;q , where p; q 2 N. This category

carries the structure of a concrete monoidal W -category [195], and as such cor-
responds to a compact matrix quantum group [195, Theorem 1.3] which is given
by the same relations as Ao .F /. Hence, Hom.ur ; us / D Z.H ˝r ; H ˝s / for all
r; s 2 N.
Step 2. Using Step 1, one shows that Hom.ur ; ur / is generated (as a -alge-
bra) by the Ir (r 2 N) and by the elements fs WD Is1 ˝ P ˝ Irs1 , where
s D 1; : : : ; r  1 and P WD EE  =kE.1/k2 . These elements satisfy
i) fs2 D fs D fs for all 1  s  r  1,
ii) fs f t D f t fs for all 1  s; t  r  1 with js  t j 2,
iii) ˇfs f t fs D fs for all 1  s; t  r  1 with js  t j D 1,
ˇ D kE.1/k4 =kF Fxk2 . These are just the relations defining the Temperley–Lieb
Algebra Aˇ;r , see [71]. Consequently, dim Hom.ur ; ur /  dim Aˇ;r . It is well
known that the dimension of Aˇ;r is the Catalan number Cr D .2r/Š=.rŠ.r C 1/Š/
[71, Aside 4.1.4].
Step 3. By Corollary 3.2.16 and Proposition 6.2.11,
Z 2 p
r r 1
dim Hom.v ; v / D hSU ..v/ .v/ / D
r r
x 2r 4  x 2 dx;
2
2
where hSU denotes the Haar state of .SU .2/; /. A routine calculation shows that
this integral is equal to the Catalan number Cr ; see, for example [12, Remarque
after Corollaire 1]. Summarizing, we obtain
dim Hom.ur ; ur /  dim Aˇ;r D Cr D dim Hom.v r ; v r /: 
Denote by .u.r/ /r2N the irreducible corepresentation matrices of .SU .2/; /
given in Theorem 6.2.8, and by ..r/ /r2N the associated characters. By Proposi-
tion 6.2.10 iii), the subalgebra CŒ.u/  SU .2/ generated by .u/ D .1/ is just
the polynomial algebra. Therefore we can define a unital homomorphism
ˆ W CŒ.u/ ! Ao .F /; .u/ 7! .v/:
Theorem 6.4.10. For each r 2 N, there exists an irreducible corepresentation
matrix v.r/ of .Ao .F /; o / such that .v.r/ / D ˆ..r/ /.
Proof. For r  1, the assertion is satisfied with v.0/ D 1 and v.1/ D v. We prove
by induction on r that there exists an irreducible corepresentation matrix v.r/ such
that v.r1/  v ' v.r2/  v.r/ and .v.r/ / D ˆ..r/ / for all r 2.
 6.4. The free unitary and the free orthogonal quantum groups 163

Assume that the assertion holds for all s < r. Using Frobenius reciprocity
(Proposition 3.1.11), the relation v ' v,
N and the induction hypothesis for r  1, we
find

Hom.v.r2/ ; v.r1/  v/ Š Hom.v.r2/  v; v.r1/ / ¤ 0:

Since v.r2/ is irreducible, it follows from Theorem 5.3.3 that v.r1/ v ' v.r2/ 
v.r/ for some corepresentation matrix v.r/ , and by Proposition 3.2.14 and 6.2.10 i),

.v.r/ / D .v.r1/ /.v/  .v.r2/ / D ˆ..r1/ .1/  .r2/ / D ˆ..r/ /:

We show that Hom.v.r/ ; v.r/ / has dimension one, and this implies that v.r/ is
irreducible. An inductive application of the formula .s/ .1/ D .s1/ C .sC1/
for s < r (see Proposition 6.2.10 i)) shows that there exist numbers a.r; s/ 2 N
such that
X
.ur / D r.1/ D .r/ C a.r; s/.s/ :
s<r

Applying ˆ, we get
X
.v r / D .v/r D .v.r/ / C a.r; s/.v.s/ /: (6.14)
s<r
P
By Corollary 3.2.16, dim Hom.ur ; ur / D 1 C s<r a.r; s/2 , and using irre-
ducibility of v.s/ for s < r and Proposition 6.4.9, we find
X
dim Hom.v.r/ ; v.r/ /  dim Hom.v r ; v r /  a.r; s/2
s<r
X
r r
 dim Hom.u ;u / a.r; s/2 D 1: 
s<r

Remarks 6.4.11. i) For n D 2, we have v.r/ 2 MrC1 .Ao .F //, see Proposi-
tion 6.4.8 ii) and Theorem 6.2.8. For n 3, one can show by induction that
v.r/ 2 Mz .Ao .F // for z WD .x rC1  y rC1 /=.x  y/, where x and y are the
solutions of the equation X 2  nX C 1 D 0.
ii) The proofs of Proposition 6.4.9 and Theorem 6.4.10 show that for each r 2 N,
the algebra Hom.v r ; v r / is the Temperley–Lieb algebra Aˇ;r .
Knowing the characters of the corepresentation matrices v.r/ , we can draw the
following conclusions:
Corollary 6.4.12. i) v.0/ D 1, v.1/ D v, and for all r; s 2 N,

v.r/  v.s/ ' v.jrsj/  v.jrsjC2/      v.rCs2/  v.rCs/ :

164 Chapter 6. Examples of compact quantum groups

ii) Every irreducible corepresentation matrix of .Ao .F /; o / is equivalent to

v.r/ for some r 2 N.
iii) The unital homomorphism CŒX  ! Ao .F / given by X 7! .v/ is injective.
iv) .v/ is semi-circular with respect to the Haar state h of .Ao .F /; o / in the
sense that
Z 2 p
1
h.f ..v/// D f .x/ 4  x 2 dx for each f 2 C0 .R/:
2
2
v) .Ao .F /; o / is not amenable if n > 2.
Proof. i) Clearly, .v.0/ / D 1 and .v.1/ / D ˆ..u// D .v/. Using the relation
.v.t/ / D ˆ..t/ / for t D jr  sj; : : : ; r C s and Proposition 6.2.10 i), we find

.v.r/ /.v.s/ / D .v.jrsj/ / C .v.jrsjC2/ / C    C .v.rCs2/ / C .v.rCs/ /:

This relation, Proposition 3.2.14, and Corollary 3.2.16 imply the decomposition
formula for v.r/  v.s/ .
ii) By Lemma 6.1.5, every irreducible corepresentation of .Ao .F /; o / is con-
tained in v r for some r 2 N, and formula (6.14) shows that v r is a direct sum
of copies of the corepresentations v.s/ for s D 0; : : : ; r.
iii) By Proposition 3.2.15, the family ..v.r/ //r is linearly independent, and
equation (6.14) implies that so is the family ..v/r /r .
iv) Using the Stone–Weierstrass Theorem, we see that it suffices to prove the
assertion for f 2 C.Œ2; 2/ of the form x 7! x r , where r 2 N. Denote by
hSU the Haar state of SU .2/, and let a.r; 0/ be as in formula (6.14). Then by
Corollary 3.2.16 and Proposition 6.2.11,

h..v/r / D dim Hom.v.0/ ; v r / D a.r; 0/

Z 2 p
r 1
D dim Hom.u.0/ ; u / D hSU ..u/ / D
r
xr 4  x 2 dx:
2
2

v) Assume that n > 2 and that .Ao .F /; o / is amenable. Then the Haar
state h is faithful, and iv) implies that the spectrum of .v/ is contained in the
interval Œ2; 2. Hence n  .v/ is invertible. On the other hand, if .Ao .F /; o /
is amenable, the counit 0 of the corresponding algebraic compact quantum group
extends to a -homomorphism W Ao .F / ! C, and .n  .v// D n  n D 0 by
Proposition 3.2.14, contradicting the invertibility of n  .v/. 

The corepresentation theory of the free unitary quantum group. The corepre-
sentation theory of the free unitary quantum group .Au .F /; u / was determined
by Banica [13]. We shall only state the main results; for the proofs see [13].
 6.4. The free unitary and the free orthogonal quantum groups 165

Let us start with some preliminaries. We denote by N  N the free coproduct

of the monoid N with itself, by ˛; ˇ 2 N  N the two canonical generators, and by
e 2 N  N the neutral element. Thus N  N can be identified with the set of all
words in ˛ and ˇ, where e corresponds to the empty word and the multiplication is
given by concatenation. Finally, we denote by x 7! xN the unique antimultiplicative
involution on N  N defined by eN WD e, ˛N WD ˇ, ˇN WD ˛.
We fix an F 2 GLn .C/, where n 2 N, and consider the free unitary quantum
group .Au .F /; u ; u/.
Theorem 6.4.13. i) There exists a family of pairwise inequivalent irreducible co-
representation matrices .ux /x2NN of .Au .F /; u / such that
X
ue D 1; u˛ D u; uˇ D u; N u Sx D uxN ; ux  uy D uab
a;b;c 2 NN
xDac; yDcbN

for all x; y 2 N. Every irreducible corepresentation matrix of .Au .F /; u / is

equivalent to ux for some x 2 N  N.
ii) Denote by ChX; X  i the free unital -algebra with generators X and X  .
Then the unital -homomorphism ChX; X  i ! Au .F / given by X 7! .u/ is
injective.
iii) .u/ is semi-circular with respect to the Haar state h of .Au .F /; u / in the
sense that
Z 2 p
1
h.f ..u/// D f .x/ 4  x 2 dx for each f 2 C0 .R/:
2
2

iv) Assume that F Fx 2 R1. Denote by Au .F /r and Ao .F /r the reduced C  -

algebraic quantum groups associated to Au .F / and Ao .F /, respectively, by ho;r
and hT the Haar states of Ao .F /r and C.T /, respectively, and by z 2 C.T / the
canonical generator. Then there exists an isomorphism of C  -algebraic compact
quantum groups Au .F /r ! .C.T /; hT /  .Ao .F /r ; ho;r / such that uij 7! z  vij
for all i; j .
Chapter 7
Multiplicative unitaries

Multiplicative unitaries are fundamental to the theory of quantum groups in the set-
ting of C  -algebras and von Neumann algebras, and to generalizations of Pontrjagin
duality. Roughly, a multiplicative unitary is one single map that encodes all struc-
ture maps of a quantum group and of its generalized Pontrjagin dual simultaneously.
In some more detail, the most important features of multiplicative unitaries are the
following:
• To every “reasonable” quantum group, one can associate a multiplicative
unitary (see Examples 7.1.4, 7.1.6, 7.1.7, and Theorems 8.3.1, 8.3.18).
• Out of every multiplicative unitary, one can construct a dual pair of von Neu-
mann bialgebras and (if the unitary is well-behaved) a dual pair of “reduced”
C  -bialgebras (see Section 7.2). Moreover, one can define a dual pair of
“universal” C  -bialgebras, using the concept of a representation and a core-
presentation of a multiplicative unitary [7, Appendice].
These two constructions facilitate the transition
• between the three guises of a quantum group (see Sections 4.3, 8.3.4 and
Theorems 5.4.5, 7.2.14);
• from a quantum group (in form of a C  -bialgebra/von Neumann bialgebra)
to the dual quantum group. This is the first key step in generalized Pontrjagin
duality (see Section 8.3.3). The second step in that duality – the identification
of the bidual – can also be achieved via the associated multiplicative unitaries
(see Theorem 8.3.15 and its proof).
Further important features of multiplicative unitaries are:
• Some examples of quantum groups are most easily constructed via a multi-
plicative unitary (see, e.g., Section 8.4).
• If a multiplicative unitary is modular, then one can construct from it an-
tipodes on the associated “reduced” C  -bialgebras (see Section 7.3.2). Un-
der favorable circumstances, one can also construct Haar weights on these
C  -bialgebras (see [205] and [67, Section 1.4]).
• Multiplicative unitaries play a central rôle in the construction of reduced
crossed products for coactions, in the construction of dual coactions, and in
the proof of Baaj–Skandalis duality (see Chapter 9).
 7.1. The concept of a multiplicative unitary 167

In the duality theory of locally compact groups and of crossed products, first
examples of multiplicative unitaries were used for some time, before Baaj and Skan-
dalis introduced the general definition and developed a beautiful and rich theory
[7]. Building on their work, Woronowicz introduced the class of manageable mul-
tiplicative unitaries [201], which is particularly well adapted to operator-algebraic
approaches to quantum groups.

7.1 The concept of a multiplicative unitary

Multiplicative unitaries are easily defined in terms of one short equation, but without
a discussion of examples, the motivation for and the implications of this short
equation are difficult to grasp. Therefore let us first consider an algebraic variant
of a multiplicative unitary.

7.1.1 Motivation
Let .A; / be a bialgebra. Consider the maps T1 ; T2 W A ˇ A ! A ˇ A given by

T1 .a ˇ b/ D .a/.1 ˇ b/ and T2 .a ˇ b/ D .a ˇ 1/.b/;

respectively. By Theorem 1.3.18, .A; / is a Hopf algebra if and only if T1 and T2

are isomorphisms, and in that case, the counit and the antipode of .A; / can be
reconstructed from T1 or T2 . If .A; / is a Hopf algebra, we can even reconstruct
its multiplication and comultiplication from the map T1 – thus all structure maps
of an arbitrary Hopf algebra can be encoded by one single map. Let us explain in
detail how this reconstruction works.
For arbitrary a 2 A; f 2 A0 , consider the linear map tf;a W A ! A given by

tf;a .b/ WD .f ˇ id/.T1 .a ˇ b// for all b 2 A:

Inserting the definition of T1 , we find that tf;a .b/ D .f ˇ id/..a//b for all b 2 A.
Denote by
W A ! Homk .A/ the left regular representation, that is,
.c/b D cb
for all c; b 2 A. Then tf;a D
..f ˇ id/..a///, and

spanftf;a j a 2 A; f 2 A0 g D
.A/ Š A as algebras;

because A is spanned by elements of the form .f ˇ id/..a//, where a 2 A and

f 2 A0 (simply let f D ).
The comultiplication  can be recovered from T1 as well: the formula

T1 ..c ˇ 1/.a ˇ b// D .ca/.1 ˇ b/ D .c/T1 .a ˇ b/

shows that .
ˇ
/..c// D T1 .
.c/ ˇ id/T11 for all c 2 A.
168 Chapter 7. Multiplicative unitaries

Implicitly, we recovered simultaneously with the Hopf algebra .A; / also the
dual algebra A0 . However, this algebra can also be reconstructed directly from the
map T1 in a similar way as A: For arbitrary b 2 A; f 2 A0 , consider the linear
0
map tf;b W A ! A given by
0
tf;b .a/ WD .id ˇ f /.T1 .a ˇ b// for all a 2 A:
0
Inserting the definition of T1 again, we find that tf;b .a/ D .id ˇ f .  b//..a// D
0
f .  b/  a for all a 2 A. For each g 2 A , denote by .g/ 2 Homk .A/ the operator
0
given by a 7! g  a. Then tf;b D .f .  b//. The map W A0 ! Homk .A/; g 7!
.g/; is an algebra homomorphism by Lemma 1.3.10 and easily seen to be injective.
We have
0
spanftf;b j b 2 A; f 2 A0 g D .A0 / Š A0 as algebras;

because A0 is spanned by elements of the form f .  b/, where f 2 A0 and b 2 A

(simply let b D 1A ).
The comultiplication on A0 can only be defined in form of the dual A0 W A0 !
.A ˇ A/0 of the multiplication map of A. However, we can extend the homomor-
phism ˇ W A0 ˇ A0 ! Homk .A ˇ A/ to .A ˇ A/0 by the formula
X
.. ˇ /.!//.a ˇ b/ WD a.1/ ˇ b.1/ !.a.2/ ˇ b.2/ /;

and the calculation

X
.id ˇ .f //.T1 .a ˇ b// D a.1/ ˇ a.2/ b.1/ f .a.3/ b.2/ /
X
D T1 .a.1/ ˇ b.1/ f .a.2/ b.2/ //
X
D T1 .a.1/ ˇ b.1/ A0 .f /.a.2/ ˇ b.2/ //

shows that . ˇ /.A0 .f // D T11 .id ˇ .f //T1 for all f 2 A0 .

Let us add that this reconstruction procedure extends to algebraic quantum
groups after straightforward modifications.

7.1.2 Definition and examples

The definition of a multiplicative unitary involves the following leg notation:
Notation 7.1.1. Let H be a Hilbert space and T 2 L.H ˝H /. We define operators
TŒ12 ; TŒ23 ; TŒ13 2 L.H ˝ H ˝ H / by the formulas

TŒ12 WD T ˝ idH ; TŒ23 WD idH ˝T;

TŒ13 WD †Œ12 TŒ23 †Œ12 D †Œ23 TŒ12 †Œ23 ;
 7.1. The concept of a multiplicative unitary 169

where † 2 L.H ˝ H / denotes the flip  ˝  7!  ˝ . Thus, the lower indices i ,

j on TŒij  indicate the two factors of the tensor product H ˝ H ˝ H on which T
acts.
More generally, given an operator T 2 L.H ˝k /, an n k, and indices 1 
i1 <    < ik  n, we define an operator TŒi1 :::ik  2 L.H ˝n / by letting T act on
the factors at the positions i1 ; : : : ; ik . Clearly, this notation can also be extended to
tuples of indices .i1 ; : : : ; ik / that are pairwise distinct but not necessarily ordered,
for example, for k D n D 2, we put TŒ21 WD †T †.
Of course, the notation introduced above can also be applied to operators on
algebraic tensor products of vector spaces and, more generally, to morphisms in
any tensor category.
Definition 7.1.2. A multiplicative unitary on a Hilbert space H is a unitary V 2
L.H ˝ H / that satisfies
VŒ12 VŒ13 VŒ23 D VŒ23 VŒ12 : (7.1)
Remarks 7.1.3. i) Equation (7.1) holds if and only if the following diagram com-
mutes:
H ˝ H ˝ HU
VŒ12iiiii4 UUUUVŒ23
UUUU
ii
iiii UU*
H ˝ H 9˝ H H ˝ BH ˝ H
99 
99  (7.2)
VŒ23 999  VŒ12
9 

H ˝H ˝H / H ˝ H ˝ H.
VŒ13
Therefore, (7.1) is called the pentagon equation.
ii) For every Hilbert space H , the identity idH ˝H is a multiplicative unitary
on H .
iii) Given a multiplicative unitary V on a Hilbert space H , we can construct the
following new unitaries:
Opposite unitary. The operator V op WD †V  † is a multiplicative unitary,
called the opposite of V . It will play an important rôle later on.
Unitary transformation. If K is a Hilbert space and U 2 L.H; K/ a unitary,
then the operator W WD .U ˝U /V .U  ˝U  / 2 L.K ˝K/ is a multiplicative
unitary on K. Two multiplicative unitaries V and W are called equivalent if
they are related this way.
Tensor product. If W is a multiplicative unitary on some Hilbert space K, then
the operator V  W WD VŒ13 WŒ24 2 L.H ˝ K ˝ H ˝ K/ is a multiplicative
unitary on H ˝ K.
170 Chapter 7. Multiplicative unitaries

Let us consider some examples of multiplicative unitaries:

Example 7.1.4. Let G be a locally compact group with left Haar measure  and
right Haar measure 1 . For  D ; 1 , we identify L2 .G; / ˝ L2 .G; / with
L2 .G  G;   /, where    denotes the product measure. Then the operators

VG 2 L.L2 .G; 1 / ˝ L2 .G; 1 //; WG 2 L.L2 .G; / ˝ L2 .G; //;
.VG /.x; y/ WD .xy; y/; .WG /.x; y/ WD .x; x 1 y/;

respectively, are multiplicative unitaries. Indeed, VG and WG are isometric because

1 and  are invariant with respect to right or left translations, and their images are
dense because the maps .x; y/ 7! .xy; y/ and .x; y/ 7! .x; x 1 y/ are homeomor-
phisms of G  G. The pentagon equations for VG and WG amount to associativity
of the multiplication in G and can be verified by straightforward calculations, for
example,

.VGŒ23 VGŒ12 /.x; y; z/ D .VGŒ12 /.x; yz; z/ D .x.yz/; yz; z/

and
.VGŒ12 VGŒ13 VGŒ23 /.x; y; z/ D .VGŒ13 VGŒ23 /.xy; y; z/
D .VGŒ23 /..xy/z; y; z/ D ..xy/z; yz; z/

for all  2 L2 .G  G  G; 1 1 1 / Š L2 .G; 1 /˝3 and all x; y; z 2 G.
op op
Short calculations show that the multiplicative unitaries VG and WG act as follows:

.VG /.x; y/ D .x; yx 1 /

op op
and .WG /.x; y/ D .yx; y/:

The multiplicative unitaries that arise from groups as in the previous example
can be characterized as follows.
Example 7.1.5. A multiplicative unitary V is called commutative if VŒ13 commutes
with VŒ23 , and cocommutative if VŒ12 commutes with VŒ13 . One easily verifies
that a multiplicative unitary V is commutative/cocommutative if and only if V op is
cocommutative/commutative.
For every locally compact group G, the operator VG is commutative and the
operator WG is cocommutative. Baaj and Skandalis showed that for every commu-
tative multiplicative unitary V , there exist a locally compact group G and a Hilbert
space K such that V is equivalent to VG  idK˝K [7, Theorem 2.2], [8].
To every algebraic quantum group, one can associate two multiplicative uni-
taries:
Example 7.1.6. Let .A0 ; 0 / be an algebraic quantum group with positive
right integral . Recall that is faithful (Corollary 2.2.5). Denote by H the
 7.1. The concept of a multiplicative unitary 171

GNS-space for , that is, the completion of A0 with respect to the inner product
hajbi WD .a b/. Then the map
T1 W A0 ˇ A0 ! A0 ˇ A0 ; a ˇ b 7! .a/.1 ˇ b/;
extends to a multiplicative unitary VA0 on H :
• T1 is isometric with respect to the inner product because is right invariant:
for all a; b; c; d 2 A0
hT1 .a ˇ b/jT1 .c ˇ d /i D . ˇ /..1 ˇ b  /0 .a c/.1 ˇ d //
D .a c/ .b  d / D ha ˇ bjc ˇ d iI

• the image of T1 is dense in H ˝ H because it is equal to A0 ˇ A0 ;

• VA0 satisfies the pentagon equation: for all a; b; c 2 A0 ,
.VA0 /Œ12 .VA0 /Œ13 .VA0 /Œ23 .a ˇ b ˇ c/
D ..2/
0 .a//.1 ˇ 0 .b//.1 ˇ 1 ˇ c/
D .VA0 /Œ23 .VA0 /Œ12 .a ˇ b ˇ c/:

If  is a positive left integral on .A0 ; 0 /, then the inverse of the map

op
T2 W A0 ˇ A0 ! A0 ˇ A0 ; a ˇ b 7! .b/.a ˇ 1/;
op
extends to a multiplicative unitary WA0 on the GNS-space H
for . Indeed, T2
op
is isometric because  is left-invariant, T2 has dense image because .A0 ; 0 /
is regular, and the pentagon equation for WA0 is a consequence of the equation
op op op op op
.T2 /Œ23 .T2 /Œ13 .T2 /Œ12 D .T2 /Œ12 .T2 /Œ23 which is easily verified.
Example 7.1.7. Let .A; / be a C  -algebraic compact quantum group with Haar
op
state h and associated GNS-space Hh , and define T1 ; T2 W A ˝ A ! A ˝ A by the
same formulas as above. Then T1 and .T2 /1 extend to multiplicative unitaries
op

VA and WA on Hh ; the unitary VA was used already in the proof of Theorem 5.4.5.
This example is related to the preceding ones as follows:
• Consider the algebraic compact quantum group .A0 ; 0 / associated to .A; /
(see Theorem 5.4.1). The restriction h0 WD hjA0 is a left- and a right-invariant
integral on .A0 ; 0 /, and clearly Hh D Hh0 ; VA D VA0 ; WA D WA0 .
• If G is a compact group and A D C.G/, then Hh D L2 .G; /, where 
denotes the normalized Haar measure of G, and VA D VG , WA D WG ,
because
.T1 .g ˝ h//.x; y/ D ..g/.1 ˝ h//.x; y/
D ..g//.x; y/  .1 ˝ h/.x; y/ D g.xy/h.y/
172 Chapter 7. Multiplicative unitaries

and
op
.T2 .g ˝ h//.x; y/ D ..h/.g ˝ 1//.x; y/ D g.x/h.xy/
for all x; y 2 G and g; h 2 C.G/.
• If G is a discrete group and A D Cr .G/, then Hh D l 2 .G/ and VA D WG ,
op
WA D WG . This can be deduced from the relations
T1 .Lx ˝ Ly / D .Lx /.1 ˝ Ly / D Lx ˝ Lxy ;
op
T2 .Lx ˝ Ly / D .Ly /.Lx ˝ 1/ D Lyx ˝ Ly for all x; y 2 G:

Remark 7.1.8. Those multiplicative unitaries that arise from C  -algebraic compact
quantum groups were characterized by Baaj and Skandalis, see [7, Paragraphe 4].

7.2 The legs of a multiplicative unitary

Out of every multiplicative unitary, we shall construct a dual pair of von Neu-
mann bialgebras by a similar procedure like that presented in Section 7.1.1. If the
multiplicative unitary satisfies an additional regularity or manageability condition,
we can additionally construct a dual pair of bisimplifiable C  -bialgebras that are
weakly dense in the aforementioned von Neumann bialgebras. These von Neumann
bialgebras and C  -bialgebras are the legs of the multiplicative unitary.
Moreover, we construct a densely defined counit and antipode on these bialge-
bras, but it seems to be difficult to relate these maps to the usual axioms for the
counit and the antipode of a Hopf algebra.

7.2.1 Definition and first properties

Let V be a multiplicative unitary on a Hilbert space H . We define normal -homo-
morphisms  y V ; V W L.H / ! L.H ˝ H / and subspaces Ay0 .V /; A0 .V / 
L.H / such that – under favorable conditions – suitable completions of Ay0 .V / and
y V or V , turn out to be C  -bialgebras
A0 .V /, equipped with the comultiplications 
or von Neumann bialgebras, respectively. These constructions are motivated by and
should be compared with the reconstruction procedure discussed in Section 7.1.1.
To simplify notation, we shall frequently denote the identity map idH by 1, thinking
of it as the unit in L.H /.

y V and V . Consider the following normal -homo-

The comultiplications 
morphisms:
y V W L.H / ! L.H ˝ H /; V W L.H / ! L.H ˝ H /;

(7.3)
T 7! V  .1 ˝ T /V; T 7! V .T ˝ 1/V  :
 7.2. The legs of a multiplicative unitary 173

y and  for 
We shall usually omit the subscript “V ” and write  y V and V ,
respectively. Denote by “ x̋ ” the von Neumann-algebraic tensor product. The
normal -homomorphisms
y x̋ id;  x̋ id W L.H ˝ H / ! L.H ˝ H / x̋ L.H / D L.H ˝ H ˝ H /

and
y id x̋  W L.H ˝ H / ! L.H / x̋ L.H ˝ H / D L.H ˝ H ˝ H /
id x̋ ;

can easily be expressed in terms of V , for example,

y x̋ id/.R/ D .V  ˝ 1/.1 ˝ R/.V ˝ 1/
. for all R 2 L.H ˝ H /:

In succinct leg notation, we have for all R 2 L.H ˝ H /:

y x̋ id/.R/ D V  R V ;
. 
. x̋ id/.R/ D VŒ12 RŒ13 VŒ12 ;
Œ12 Œ23 Œ12
(7.4)
y
.id x̋ /.R/ 
D VŒ23 RŒ13 VŒ23 ; 
.id x̋ /.R/ D VŒ23 RŒ12 VŒ23 :

y x̋ id/ ı 
Lemma 7.2.1. . y D .id x̋ /
y ı
y and . x̋ id/ ı  D .id x̋ / ı .

y x̋ id/..T
Proof. Let T 2 L.H /. In leg notation, . y // takes the form

y x̋ id/..T
y // D V  .T
y / V  
. Œ12 Œ23 Œ12 D VŒ12 VŒ23 TŒ3VŒ23 VŒ12 :

  
By the pentagon equation, this is equal to VŒ23 VŒ13 VŒ12 TŒ3VŒ12 VŒ13 VŒ23 . But
 
VŒ12 TŒ3VŒ12 D TŒ3VŒ12 VŒ12 D TŒ3 , whence

y x̋ id/..T
. y // D V  V  T V V
Œ23 Œ13 Œ3 Œ13 Œ23
 y y .T
y //:
D VŒ23 .T /Œ13 VŒ23 D .id x̋ /.

The assertion concerning  follows from a similar calculation. 

The algebras Ay0 .V / and A0 .V /. The subspaces Ay0 .V / and A0 .V / are obtained
from V by the application of slice maps, which are reviewed in Section 12.4. Denote
by L.H / the set of all normal linear functionals on L.H /. We shall be interested
in the spaces

Ay0 .V / WD faO ! j ! 2 L.H / g; A0 .V / WD fa! j ! 2 L.H / g;

where for each ! 2 L.H / ,

aO ! WD .id x̋ !/.V / 2 L.H /; a! WD .! x̋ id/.V / 2 L.H /:

174 Chapter 7. Multiplicative unitaries

In the following paragraphs, we shall frequently encounter calculations that

involve slice maps. In particular, we use (analogues of) formulas (12.2)–(12.5) and
relations like

.id x̋ !/.X /.id x̋ /.Y / D .id x̋ ! x̋ /.XŒ12 YŒ13 /

D .id x̋  x̋ !/.XŒ13 YŒ12 /;
.id x̋ !/.X /. x̋ id/.Y / D . x̋ id x̋ !/.XŒ23 YŒ12 /;

which hold for all X; Y 2 L.H ˝ H / and !;  2 L.H / ; see also Remark 12.4.5.
Lemma 7.2.2. Ay0 .V / and A0 .V / are subalgebras of L.H /, and each of the sets
Ay0 .V /H; Ay0 .V / H; A0 .V /H and A0 .V / H is linearly dense in H .
Proof. First, we show that Ay0 .V / is an algebra. Let !; ! 0 2 L.H / . Then

aO ! aO ! 0 D .id x̋ !/.V /  .id x̋ ! 0 /.V / D .id x̋ ! x̋ ! 0 /.VŒ12 VŒ13 /:


We replace VŒ12 VŒ13 by VŒ23 VŒ12 VŒ23 using the pentagon equation, and find

aO ! aO ! 0 D .id x̋ ! x̋ ! 0 /.VŒ23 VŒ12 VŒ23


/ D .id x̋ ! x̋ ! 0 /..id x̋ /.V //

(see formula (7.4)). The linear functional ! 00 WD .! x̋ ! 0 / ı  is normal, and

aO ! aO ! 0 D .id x̋ ! 00 /.V / D aO ! 00 2 Ay0 .V /:

Next, assume that  2 H is orthogonal to the set Ay0 .V /H . We show that  D 0,

and this implies that the set Ay0 .V /H is linearly dense in H . For every ;  2 H ,
put !; WD hj  i. Then !; 2 L.H / , and by assumption,

0 D hjaO !; #i D hj.id x̋ !; /.V /#i D h ˝ jV .# ˝ /i

for all ; ; # 2 H . Since V is surjective, this relation implies h ˝ j# 0 ˝  0 i D 0

for all ; # 0 ;  0 2 H , and hence  D 0.
The proofs of the remaining assertions are similar. 
The algebras Ay0 .V / and A0 .V / need not be closed with respect to the involution
(see, for example, [9, Remark 4.5]). To describe the adjoints of elements aO ! 2
Ay0 .V / and a! 2 A0 .V /, we use the following simple result:
Lemma 7.2.3. i) For every ! 2 L.H / , the linear map !  W L.H / ! C given by
T 7! !.T  / is normal. The assignment ! 7! !  defines an involution, that is, a
conjugate-linear involutive map, on L.H / .
ii) ..! x̋ id/.R// D .!  x̋ id/.R / and ..id x̋ !/.R// D .id x̋ !  /.R /
for each R 2 L.H ˝ H /.
 7.2. The legs of a multiplicative unitary 175

Proof. This follows from standard arguments and calculations, see also formu-
la (12.4). 
Example 7.2.4. For all ;  2 H , one has .!; / D !; because .!; / .T / D
hjT  i D hjT i for each T 2 L.H /.
The preceding lemma implies that for each ! 2 L.H / ,
.aO ! / D .id x̋ !  /.V  / and .a! / D .!  x̋ id/.V  /: (7.5)

In general, these elements need not belong to Ay0 .V / or A0 .V /, respectively.

The left-right symmetry. The constructions introduced so far are symmetric in a

sense that is made precise in the following lemma. As before, † 2 L.H ˝ H /
denotes the flip ˝ 7!  ˝. Note that Ad† .S ˝T / D T ˝S for all S; T 2 L.H /
since †.S ˝ T /†. ˝ / D †.S ˝ T / D T  ˝ S  for all ;  2 H .
Lemma 7.2.5. We have
Ay0 .V op / D A0 .V / ; A0 .V op / D Ay0 .V / ;
y V op D Ad† ıV ;
 V op D Ad† ı yV :

Proof. The first and third equation follow from the fact that
.id x̋ !/.V op / D .id x̋ !/.†V  †/ D .! x̋ id/.V  /
D ..!  x̋ id/.V // D .a!  /
for every ! 2 L.H / (see equation (7.5)) and
y V op .T / D V op  .1 ˝ T /V op D †V †.1 ˝ T /†V  †

D †V .T ˝ 1/V  † D Ad† .V .T //
for every T 2 L.H /. The remaining assertions follow similarly. 

7.2.2 Well-behaved multiplicative unitaries

We shall primarily be interested in the following classes of multiplicative unitaries:
Definition 7.2.6. A multiplicative unitary V on a Hilbert space H is
i) well-behaved if the following conditions are satisfied:
(a) the following spaces are C  -subalgebras of L.H /:
y / WD k  k-closure of Ay0 .V /; A.V / WD k  k-closure of A0 .V /;
A.V
176 Chapter 7. Multiplicative unitaries

(b) the restrictions of the -homomorphisms  y and  defined in equa-

y / and A.V /, respectively, turn A.V
tion (7.3) to A.V y / and A.V / into
bisimplifiable C  -bialgebras,
(c) V 2 M.A.V y / ˝ A.V //;

ii) weakly well-behaved if the following spaces are von Neumann algebras:

Ayw .V / WD w-closure of Ay0 .V /; Aw .V / WD w-closure of A0 .V /;

where “w-closure” denotes the closure with respect to the weak operator
topology.
y /, Ayw .V / and A0 .V /, A.V /, A! .V /
Informally, we call the spaces Ay0 .V /, A.V
together with the restrictions of the homomorphisms  y and , respectively, the left
leg and the right leg of V . In Section 7.3.1, we discuss two classes of (weakly)
well-behaved multiplicative unitaries – the (weakly) regular and the manageable
ones.

Reformulation in terms of ket-bra-operators. For later use in Section 7.3, we

y /, Ayw .V / and A.V /, Aw .V / in terms
reformulate the definition of the spaces A.V
of ket-bra operators. More precisely, we combine the ket-bra notation with the leg
notation as follows. For each ;  2 H , we define operators
ji˝id
jiŒ1 W H Š C ˝ H ! H ˝ H;  0 7!  ˝  0 ; (7.6)
and
id ˝ji
jiŒ2 W H Š H ˝ C ! H ˝ H; 0 7! 0 ˝ : (7.7)

Evidently, the adjoints hjŒ1 WD jiŒ1  and hjŒ2 WD jiŒ2  are given by

hjŒ1 .0 ˝  0 / D hj0 i 0 and hjŒ2 .0 ˝  0 / D 0 hj 0 i (7.8)

for all 0 ;  0 2 H , respectively.

In terms of these maps, the operators aO ! and a! that comprise the spaces Ay0 .V /
and A0 .V / can be rewritten as follows. For a functional ! 2 L.H / of the form
! D ! 0 ; D h 0 j  i, where ;  0 2 H ,

aO ! D .id x̋ ! 0 ; /.V / D h 0 jŒ2 V jiŒ2 ;

and for a functional ! 2 L.H / of the form ! D !0 ; D h0 ji, where ; 0 2 H ,

a! D .!0 ; x̋ id/.V / D h0 jŒ1 V jiŒ1 :

 7.2. The legs of a multiplicative unitary 177

y / with
Lemma 7.2.7. i) The set fh 0 jŒ2 V jiŒ2 j ;  0 2 H g is linearly dense in A.V
respect to the norm and in Ayw .V / with respect to the weak operator topology.
ii) The set fh0 jŒ1 V jiŒ1 j ; 0 2 H g is linearly dense in A.V / with respect to
the norm and in Aw .V / with respect to the weak operator topology.
0
Proof. P i) Let ! 2 P
L.H / . Then there exist sequences
P 0 .n /n and .n /n in H that
0 2
satisfy n kn k ; n kn k < 1, and !.T / D n hn jT n i for each T 2 L.H /.
2

Thus ! can be approximated in norm by finite linear combinations of maps of the

form ! 0 ; , where ;  0 2 H . Since k id x̋ k  kk for each  2 L.H /
(Proposition 12.4.4), .id x̋ !/.V / can be approximated in norm by finite linear
combinations of elements of the form h 0 jŒ2 V jiŒ2 , where ;  0 2 H .
ii) The proof is similar to the proof of i). 

The setting of von Neumann algebras. Evidently, every well-behaved multi-

plicative unitary is weakly well-behaved. But the definition of the term “weakly
well-behaved” seems to be much weaker than the definition of the term “well-
behaved”: It only involves an analogue of condition i)(a) and not i)(b) or i)(c). We
shall see that in the setting of von Neumann algebras, analogues of these conditions
are automatically fulfilled.
Given a Hilbert space K and a subset C  L.K/, we denote by C 0 WD fS 2
L.K/ j S T D T S for all T 2 C g the commutant of C .
Proposition 7.2.8. Let V be a multiplicative unitary on a Hilbert space H . Con-
sider the following von Neumann algebras:
Byw .V / WD .Ay0 .V / C Ay0 .V / /00 ; Bw .V / WD .A0 .V / C A0 .V / /00 :

y and  defined in equation (7.3) restrict to normal -homomor-

i) The maps 
phisms
y W Byw .V / ! Byw .V / x̋ Byw .V /;
  W Bw .V / ! Bw .V / x̋ Bw .V /:

ii) .Byw .V /; /
y and .Bw .V /; / are von Neumann bialgebras.

iii) V 2 Byw .V / x̋ Bw .V /.
For the proof of this proposition, we use the following lemma.
Lemma 7.2.9. For each X 2 L.H ˝ H /,
X 2 .Byw .V / ˝ 1/0 , ŒXŒ12 ; VŒ13  D 0;
X 2 .1 ˝ Byw .V //0 , ŒXŒ12 ; VŒ23  D 0;
X 2 .1 ˝ Bw .V //0 , ŒXŒ23 ; VŒ13  D 0;
X 2 .Bw .V / ˝ 1/0 , ŒXŒ23 ; VŒ12  D 0:
178 Chapter 7. Multiplicative unitaries

In particular, Byw .V /0 D fT 2 L.H / j ŒT ˝ 1; V  D 0g and Bw .V /0 D fT 2

L.H / j Œ1 ˝ T; V  D 0g.
Proof. We shall only prove the first equivalence; the others follow similarly. By
definition, we have for all ! 2 L.H /
X.aO ! ˝ 1/ D .id x̋ id x̋ !/.XŒ12 VŒ13 /;
.aO ! ˝ 1/X D .id x̋ id x̋ !/.VŒ13 XŒ12 /:
Now the implication “)” follows immediately from the fact that maps of the form
id x̋ id x̋ !, where ! 2 L.H / , separate the elements of L.H ˝ H ˝ H /.
 
Conversely, if XŒ12 VŒ13 D VŒ13 XŒ12 , then XŒ12VŒ13 D VŒ13 XŒ12 and hence

X.aO ! ˝ 1/ D .id x̋ id x̋ !  /.XŒ12VŒ13


/
D .id x̋ id x̋ !  /.VŒ13

XŒ12/ D .aO ! ˝ 1/X
for all ! 2 L.H / (compare equation (7.5)). 
Proof of Proposition 7.2.8. We prove the statements concerning Byw .V /; replacing
V by V op , we obtain the corresponding statements for Byw .V op / D Bw .V /.
i) We only need to show that for each ! 2 L.H / and X 2 .Byw .V / x̋ Byw .V //0 ,
y aO ! / commutes with X . By the pentagon equation,
the operator .
y aO ! / D V  .1 ˝ aO ! /V D .id x̋ id x̋ !/.V  V V /
. Œ12 Œ23 Œ12
(7.9)
D .id x̋ id x̋ !/.VŒ13 VŒ23 /:

On the other hand, X 2 .1 ˝ Byw .V //0 \ .Byw .V / ˝ 1/0 , so that XŒ12 commutes
with VŒ13 VŒ23 by Lemma 7.2.9. Therefore,
y aO ! / D .id x̋ id x̋ !/.X V V /
X . Œ12 Œ13 Œ23
y aO ! /X:
D .id x̋ id x̋ !/.VŒ13 VŒ23 XŒ12 / D .
ii) This follows from i) and Lemma 7.2.1.
iii) We only need to show that V commutes with .Byw .V / x̋ Bw .V //0 . But if
X 2 .Byw .V / x̋ Bw .V //0 , then X 2 .Byw .V / ˝ 1/0 \ .1 ˝ Bw .V //0 , so that XŒ13
commutes with VŒ12 and VŒ23 by Lemma 7.2.9, and then XŒ13 also commutes with
 
VŒ13 D VŒ12 VŒ23 VŒ12 VŒ23 . 
Corollary 7.2.10. Let V be a weakly well-behaved multiplicative unitary. Then
y and  restrict to normal -homomorphisms
the maps 
y W Ayw .V / ! Ayw .V / x̋ Ayw .V / and  W Aw .V / ! Aw .V / x̋ Aw .V /;


and .Ayw .V /; /
y and .Aw .V /; / are von Neumann bialgebras.
 7.2. The legs of a multiplicative unitary 179

Proof. The assumption implies that Ayw .V / and Aw .V / are equal to the von Neu-
mann algebras Byw .V / and Bw .V / defined in the proposition above. Now the claim
follows from that proposition. 

The left/right-symmetry again. The apparent symmetry in the definition of the

left and of the right leg of a (weakly) well-behaved multiplicative unitary V is
related to the symmetry between V and the opposite unitary V op as follows:
Proposition 7.2.11. If V is a well-behaved multiplicative unitary, then the opposite
multiplicative unitary V op is well-behaved and

.A.V y V op / D .A.V /; V /cop ;

y op /;  y /; 
.A.V op /; V op / D .A.V y V /cop :

If V is a weakly well-behaved multiplicative unitary, then the opposite multiplicative

unitary V op is weakly well-behaved and
y V op / D .Aw .V /; V /cop ; .Aw .V op /; V op / D .Ayw .V /; 
.Ayw .V op /;  y V /cop :

Proof. This follows immediately from Lemma 7.2.5. 

Remark 7.2.12. The proposition above shows that the symmetry between the two
legs of a (weakly) well-behaved multiplicative unitary V is not perfect. If we would
replace either  by Ad† ı or y by Ad† ı, y then the passage between the left and
the right leg would amount to the passage between V and V op . This twist becomes
relevant when we consider coactions of C  -bialgebras and crossed products, see
Sections 9.3 and 9.4.

7.2.3 Examples
Let us determine the legs of some of the multiplicative unitaries introduced in
Section 7.1.2.
From the multiplicative unitary associated to a group, we recover the C  -bi-
algebras and von Neumann bialgebras discussed in Section 4.2:
Example 7.2.13. Let G be a locally compact group and consider the associated
multiplicative unitary WG (see Example 7.1.4). Recall that WG acts on the Hilbert
space L2 .G; / ˝ L2 .G; / Š L2 .G  G;   / via .WG /.x; y/ D .x; x 1 y/
for all x; y 2 G and  2 L2 .G  G;   /, where  denotes the left Haar measure
of G as usual. We show that
y G / Š C0 .G/
A.W and A.WG / D Cr .G/ as C  -bialgebras: (7.10)

This implies that Ayw .WG / Š L1 .G/ and Aw .WG / D L.G/ as von Neumann
bialgebras. Moreover, it is easy to see that WG 2 M.C0 .G/ ˝ Cr .G//, so that
180 Chapter 7. Multiplicative unitaries

WG is well-behaved (and weakly well-behaved); see also Example 7.3.4 iii) and
Theorem 7.3.11. For the definition of the C  -bialgebras C0 .G/, Cr .G/ and the
von Neumann bialgebras L1 .G/, L.G/, see Example 4.2.1, 4.2.2, and 4.2.4.
Let us prove (7.10). We compute the operator aO ! D .id x̋ !/.WG /, where
! D ! 0 ; D h 0 j  i for some  0 ;  2 H :
Z
0
.aO ! /.x/ D .h jŒ2 WG jiŒ2 /.x/ D  0 .y/ .WG . ˝ //.x; y/ d .y/
Z G

D  0 .y/.x/.x 1 y/ d .y/ D f .x/.x/

G

for all  2 L2 .G; / and x 2 G, where the function f W G ! C is given by

Z
f .x/ D  0 .y/.x 1 y/ d .y/ D h 0 jLx i for all x 2 G:
G

Here, Lx 2 L.L2 .G; // denotes the left translation operator  7! .x 1  / as
usual. If we approximate  0 ;  in L2 -norm by elements of Cc .G/, the function
f gets approximated in sup-norm by elements of Cc .G/. Therefore f belongs
to C0 .G/. Denote by
M W C0 .G/ ! L.L2 .G; // the representation given by
multiplication operators. The calculation above shows that aO ! D
M .f /, and by
Lemma 7.2.7,
y G / 
M .C0 .G// Š C0 .G/:
A.W
If we replace  by an approximate unit for the convolution algebra Cc .G/ 
L1 .G; / (see Example 4.2.2) and choose  0 2 Cc .G/, then f converges in sup-
norm to x0 . Therefore, the inclusion above is an equality.
y G / Š C0 .G/ as C  -bialgebras. For all f 2 C0 .G/,
Let us show that A.W
y WG .
M .f // D .
M ˝
M /.C .G/ .f //;
 0

because
.WG .1 ˝
M .f //WG /.x; y/ D ..1 ˝
M .f //WG /.x; xy/
D f .xy/ .WG /.x; xy/ D f .xy/.x; y/
for all x; y 2 G and  2 L2 .G  G;   / Š L2 .G; / ˝ L2 .G; /.
To determine A.WG /, we compute the operator a! D .! x̋ id/.WG /, where
! D !0 ; D h0 j  i for some ; 0 2 H :
Z
.a! /.y/ D .h0 jŒ1 WG jiŒ1 /.y/ D 0 .x/.WG . ˝ //.x; y/ d .x/
Z G

D 0 .x/.x/ .x 1 y/ d .x/ D .L.g//.y/

G
 7.2. The legs of a multiplicative unitary 181

for all  2 L2 .G; / and y 2 G, where g D x0  (pointwise product). Using

Lemma 7.2.7 and the relation L2 .G; /L2 .G; / D L1 .G; /, we conclude

A.WG / D k  k-closure of fL.f / j f 2 L1 .G; /g D Cr .G/

as C  -algebras.
Let us show that A.WG / D Cr .G/ as C  -bialgebras. As above, we denote by
Lz 2 M.Cr .G//; z 2 G; the left translation operators. The operators

WG .Lz / D WG .Lz ˝ 1/WG and Cr .G/ .Lz / D Lz ˝ Lz

are equal for all z 2 G, because

.WG .Lz ˝ 1/WG /.x; y/ D ..Lz ˝ 1/WG /.x; x 1 y/

D .WG /.z 1 x; x 1 y/
D .z 1 x; z 1 xx 1 y/ D ..Lz ˝ Lz //.x; y/

for all x; y 2 G and  2 L2 .G  G;   / Š L2 .G; / ˝ L2 .G; /.

From the multiplicative unitary associated to an algebraic quantum group (see

Example 7.1.6), we obtain a C  -algebraic completion of the algebraic quantum
group and a C  -algebraic completion of its dual:

Theorem 7.2.14. Let .A0 ; 0 / be an algebraic quantum group with positive right
integral . Then the associated multiplicative unitary V WD VA0 is well-behaved.
In the notation of Example 7.1.6, we have:

i) For every a 2 A0 , the map A0 ! A0 given by d 7! ad extends to a bounded

linear operator
.a/ on H . The map
W A0 ! L.H /; a 7!
.a/, is an
injective -homomorphism, and A.V / is the norm-closure of
.A0 /.

ii) The space 0 .A0 / C A0 ˇ A0  M.A0 ˇ A0 / is a -subalgebra, the

-homomorphism
ˇ
W A0 ˇ A0 ! L.H ˝ H / extends uniquely to
a -homomorphism on this subalgebra, and

.
ˇ
/.0 .a// D V .
.a/ ˝ 1/V  D V .
.a// for all a 2 A0 :

Denote by .Ay0 ; 
y 0 / the dual algebraic quantum group of .A0 ; 0 /.

iii) For every aO 2 Ay0 , the map A0 ! A0 given by d 7! .id ˇ a/.

O 0 .d // extends
to a bounded linear operator .a/O on H . The map W Ay0 ! L.H /; aO 7!
O is an injective -homomorphism, and A.V
.a/, y / is equal to the norm-closure
of .Ay0 /.
182 Chapter 7. Multiplicative unitaries

y 0 .Ay0 / C Ay0 ˇ Ay0  M.Ay0 ˇ Ay0 / is a -subalgebra, the

iv) The space 
-homomorphism ˇ W Ay0 ˇ Ay0 ! L.H ˝ H / extends uniquely to a
-homomorphism on this -subalgebra, and
y 0 .a//
. ˇ /. O D V  .1 ˝ .a//V
O y V . .a//
D O for all aO 2 Ay0 :

Proof. We first prove the assertions i)–iv) and then show that V is well-behaved.
i) Let us determine the operator a! D .! x̋ id/.V / for a functional ! of the
form ! D !b;c D hbj  ci, where b; c 2 A0 . For all d 2 A0 ,
a! d D hbjŒ1 V jciŒ1 d D hbjŒ1 V .c ˝ d /
X X
D hbjc.1/ ic.2/ d D .b  c.1/ /c.2/ d;
P
that is, a! d D ad , where a WD .b  c.1/ /c.2/ . Since the map T2 (see Defini-
tion 2.1.9 on page 44) is surjective, the map A0 ˇ A0 7! A0 given by b ˇ c 7!
P
.b  c.1/ /c.2/ D . ˇ id/..b  ˇ 1/0 .c// is surjective as well.
The existence of the map
follows directly, and the density of
.A0 / in A.V /
follows by Lemma 7.2.7. Evidently,
is a -homomorphism. It is injective,
because A0 is non-degenerate and is faithful (Corollary 2.2.5): For every non-
zero a 2 A0 , there exists a d 2 A0 such that ad ¤ 0, and then k
.a/d k2 D
h
.a/d j
.a/d i D .d  a ad / ¤ 0.
ii) The first assertion follows easily from the fact that A0 ˇ A0  M.A0 ˇ A0 /
is an ideal, and the remaining assertions follow from the relations
V .b ˝ c/ D 0 .b/.1 ˇ c/ and V .
.a/b ˝ c/ D 0 .a/0 .b/.1 ˇ c/;
which hold for all a; b; c 2 A0 .
iii) Let us compute the operator aO ! D .id x̋ !/.V /, where ! D !b;c D hbj  ci
with b; c 2 A0 . For all d 2 A0 ,
X
aO ! d D hbjŒ2 V jciŒ2 d D d.1/ hbjd.2/ ci
X
D d.1/ .b  d.2/ c/ D .id ˇ a/.
O 0 .d //;

where aO D .b   c/ 2 .A0 /0 . Using the modular automorphism  0 of (see

Remarks 2.2.18), it is easy to see that the linear span of the subset f .b   c/ j
b; c 2 A0 g  .A0 /0 coincides with Ay0 .
y /
The existence of the map follows directly, and the density of .Ay0 / in A.V
follows by Lemma 7.2.7. Next, we show that is a -homomorphism. Let a; O
O b 2 Ay0
and c; d 2 A0  H . Since aO bO D .aO ˇ b/O ı 0 (see equation (2.10) on page 59),

O b/d
.a/ . O D ..id ˇ a/ O ı 0 /.d /
O ı 0 ı .id ˇ b/
O
D .id ˇ aO ˇ b/..2/ O
0 .d // D .a
O b/d:
 7.2. The legs of a multiplicative unitary 183

Define aL 2 Ay0 by a.e/

L O  / for all e 2 A0 . Then
WD a.e
X X
  
h .a/cjd
O iD .c.1/ O .2/ / D
d /a.c .c.1/ L .2/
d /a.c /:

Inserting Lemma 2.2.12 and the relation aL ı S0 D aO  (Lemma 2.3.10), we find

X X
 
.c.1/ L .2/
d /a.c /D .c  d.1/ /a.S
L 0 .d.2/ // D hcj .aO  /d i:

Consequently,
is a -homomorphism.
Finally, since the set .A0 ˇ 1/0 .A0 / spans A0 ˇ A0 , and since

O D .id ˇ a/..e

.e/ .a/d O ˇ 1/0 .d // for all e 2 A0 ;

the -homomorphism is injective.

iv) The first assertion follows easily from the fact that Ay0 ˇ Ay0  M.Ay0 ˇ Ay0 /
is an ideal, and the remaining assertions follow from the relations V .b ˝ c/ D
0 .b/.1 ˇ c/ and
.2/
.id ˝ .a//V
O .b ˝ c/ D .id ˇ id ˇ a/.
O 0 .b/.1 ˇ 0 .c///;

which hold for all aO 2 Ay0 and b; c 2 A0 .

To complete the proof that V is well-behaved, we need to show that V belongs
y / ˝ A.V //. Let a; b; c; d 2 A0 , and put aO WD .  a/. Then
to M.A.V
X
V . .a/O ˝
.b//.c ˝ d / D V .c.1/ .c.2/ a/ ˝ bd /
X
D c.1/ ˇ c.2/ bd .c.3/ a/:
P P
Choose bi0 ; ai0 2 A0 such that b ˇ a D i .ai0 /.1/ bi0 ˇ .ai0 /.2/ . Then
X XX
c.1/ ˇ c.2/ bd .c.3/ a/ D c.1/ ˇ c.2/ .ai0 /.1/ bi0 d .c.3/ .ai0 /.2/ /
i
XX
D c.1/ .c.2/ ai0 / ˇ bi0 d
i
X
D . .aO i0 / ˝
.bi0 //.c ˝ d /;
i

where aO i0 D .  ai0 /. Consequently, V . .Ay0 / ˇ

.A0 //  .Ay0 / ˇ
.A0 /, and a
similar argument shows that V  . .Ay0 / ˇ
.A0 //  .Ay0 / ˇ
.A0 /. 
In Example 7.3.4 iv) and Theorem 7.3.11, we shall see again that the multiplica-
tive unitary V studied above is well-behaved. A detailed analysis of the preceding
example can be found in [94].
184 Chapter 7. Multiplicative unitaries

7.2.4 The dual pairing, counit, and antipode of the legs

We now consider additional structural properties of the legs of a multiplicative
unitary and construct a densely defined dual pairing, counit, and antipode. Up to
minor modifications, the dual pairing satisfies all equations that characterize a dual
pairing of Hopf -algebras. Thus, this dual pairing gives a precise meaning to the
vague idea that the two legs of a multiplicative unitary are dual to each other.
Before we introduce the pairing, let us recall some terminology. Let X and
Y be complex vector spaces. We say that a bilinear map .  j  / W X  Y ! C is
non-degenerate if for every non-zero x0 2 X and every non-zero y0 2 Y , there
exist y 2 Y and x 2 X such that .x0 jy/ ¤ 0 and .xjy0 / ¤ 0.
Proposition 7.2.15. Let V be a multiplicative unitary on a Hilbert space H . Then
the bilinear map .  j  / W Ay0 .V /  A0 .V / ! C given by
.aO ! ja / WD . x̋ !/.V / D !.a / D .aO ! / (7.11)
is separately  -weakly continuous, non-degenerate, and extends to bilinear maps
.  j   W Ay0 .V /  Aw .V / ! C and Π j  / W Ayw .V /  A0 .V / ! C;
(7.12)
.aO ! ; a/ 7! !.a/; O a / 7! .a/;
.a; O
that are  -weakly continuous in the second or first variable, respectively. For all
!; ! 0 2 L.H / ; a 2 Aw .V /, and ;  0 2 L.H / ; aO 2 Ayw .V /,
.aO ! aO ! 0 ja D .! x̋ ! 0 /..a// and Œaja y a//:
O  a 0 / D . x̋  0 /.. O (7.13)

Proof. Evidently, formula (7.11) defines a bilinear map .  j  / W Ay0 .V /  A0 .V / !

C that is separately  -weakly continuous. This bilinear map is non-degenerate
because the natural pairing L.H /  L.H / ! C is non-degenerate, and it extends
uniquely to Ay0 .V /Aw .V / and Ayw .V /A0 .V / because of the separate continuity.
The first formula in (7.13) follows immediately from the proof of Lemma 7.2.2,
and the second formula can be proved by similar calculations. 
Next, we construct a counit and an antipode on the legs of a weakly well-behaved
multiplicative unitary V . In the formulation of the following proposition, we use the
pairings defined in equation (7.12), the fact that the identity operator idH belongs
to Ayw .V / and Aw .V /, and equation (7.5) from page 175.
Proposition 7.2.16. Let V be a weakly well-behaved multiplicative unitary on a
Hilbert space H .
i) There exist algebra homomorphisms
O W Ay0 .V / ! C; aO ! 7! .aO ! j idH  D !.idH /;
W A0 .V / ! C; a 7! ŒidH ja / D .idH /:
 7.2. The legs of a multiplicative unitary 185

ii) There exist conjugate-linear algebra antihomomorphisms

Sy W Ay0 .V / ! Ayw .V /; aO ! D .id x̋ !/.V / 7! .id x̋ !/.V  / D .aO !  / ;

S W A0 .V / ! Aw .V /; a D . x̋ id/.V / 7! . x̋ id/.V  / D .a  / :

For all !;  2 L.H / ,

y aO ! / ja / D .aO ! ja   and .aO ! jS.a /  D ŒaO !  ja /:
ŒS.

y Ay0 .V //  A.V
If V is well-behaved, then S. y / and S.A0 .V //  A.V /.

Proof. We only prove the claims concerning O and Sy; for and S , the corresponding
assertions follow similarly.
i) Equation (7.13) shows that for all !; ! 0 2 L.H / ,

.aO ! aO ! 0 j idH  D .! x̋ ! 0 /..idH // D !.idH / ! 0 .idH / D .aO ! j idH   .aO ! 0 j idH :

Therefore, O is a homomorphism.
ii) First, we find from equation (7.5) that

.! x̋ id/.V  / D .a!  / 2 Ay0 .V /  Ayw .V / D Ayw .V /

for all ! 2 L.H / . If V is well-behaved, we can replace Ayw .V / by A.Vy / in this

equation.
Let us show that the assignment Sy W aO ! 7! .id x̋ !/.V  / is well defined. By
equation (7.5), we have for all !;  2 L.H /

.aO ! ja   D !.a  / D !..  x̋ id/.V  // D   ..id x̋ !/.V  //:

Assume that aO ! D 0. Then the left-hand side of the equation above, and hence also
the right-hand side, is zero. Since functionals of the form   , where  2 L.H / ,
separate the elements of L.H /, we must have .id x̋ !/.V  / D 0. Therefore, Sy is
well defined. Moreover, the calculation above shows that

.aO ! ja   D ..id x̋ !/.V  / / D .Sy.aO ! / / D ŒSy.aO ! / ja /:

Finally, let us prove that Sy is an antihomomorphism. Given !; ! 0 2 L.H / ,

put ! 00 WD .! x̋ ! 0 / ı . Then aO ! aO ! 0 D aO ! 00 by the proof of Lemma 7.2.2, and
using the definition of Sy and the pentagon equation, we find
y aO ! 0 /S.
S. y aO ! / D .id x̋ ! ˝ ! 0 /.V  V  /
Œ13 Œ12
D .id x̋ ! ˝ ! 0 /.VŒ23 VŒ12
 
VŒ23 /
D .id x̋ ! ˝ ! 0 /..id x̋ /.V  //
D .id ˝! 00 /.V  / D Sy.aO ! 00 / D Sy.aO ! aO ! 0 /: 
186 Chapter 7. Multiplicative unitaries

The antipodes Sy and S introduced above are unbounded and therefore difficult to
handle. If V is manageable or modular, then Sy and S extend to closed maps and can
be described in terms of objects that are more tractable: a unitary antipode, which
is an involutive -antiautomorphism, and a scaling group, which is a one-parameter
group of automorphisms, see Section 7.3.2.
y S constructed above to the
It seems to be difficult to relate the maps O , and S,
usual axioms for the counit and the antipode of a Hopf algebra; several obstacles
were discussed in the beginning of Chapter 4. In the following remark, we indicate
some possible interpretations of these axioms.
Remark 7.2.17. Let V be a weakly well-behaved multiplicative unitary on a Hilbert
space H . Consider the maps O and Sy constructed in Proposition 7.2.16.
In order to formulate an analogue of the counit identity . 0 ˇ id/ ı 0 D id D
.id ˇ 0 / ı 0 of a Hopf algebra .A0 ; 0 / for the map O , we need to extend the
maps O ˇ id and id ˇ O , which are initially defined on the algebraic tensor product
Ay0 .V / ˇ Ay0 .V / only, to .
y Ay0 .V //. To do so, we approximate O pointwise by
normal linear functionals as follows. Since Aw .V / contains idH and A0 .V / is dense
Aw .V / with respect to the -weak topology, there exists a net . / in L.H / such
that the elements a WD a converge -weakly to idH . Then

O .aO ! / D .aO ! j idH  D lim .aO ! ja / D lim v .aO ! / for all ! 2 L.H / :

This equation suggests that we should extend the maps O ˇ id and id ˇ O to the
image of  y by approximating them pointwise by maps of the form  x̋ id and
id x̋  , respectively. Thus we claim that for all ! 2 L.H / ,
 -weakly -weakly
y aO ! // ! aO !
.v x̋ id/.. and y aO ! // ! aO ! :
.id x̋ v /..
!1 !1

y aO ! / D .id x̋ id x̋ !/.VŒ13 VŒ23 /

Let us prove this claim. Inserting the relation .
(see equation (7.9) on page 178) into the left-hand sides, we find
y aO ! // D . x̋ id x̋ !/.VŒ13 VŒ23 / D .id x̋ !/..1 ˝ a /V /;
. x̋ id/..
y aO ! // D .id x̋  x̋ !/.VŒ13 VŒ23 / D .id x̋ !/.V .1 ˝ a //:
.id x̋  /..

Now the claim follows from the fact that .1 ˝ a /V and V .1 ˝ a / converge
 -weakly to V as  tends to infinity and that id x̋ ! is  -weakly continuous.
Consider the antipode axiom m0 ı .S0 ˇ id/ ı 0 D 0 ı 0 D m0 ı .id ˇ S0 / ı
0 of a Hopf algebra .A0 ; 0 /. The relations .y aO ! / D .id x̋ id x̋ !/.VŒ13 VŒ23 /
y
and S ..id x̋ ! /.V // D .id x̋ ! /.V /; ! 2 L.H / , suggest to think of the
0 0  0

undefined expressions .Sy ˝ id/..y aO ! // and .id ˝Sy/..y aO ! // as being equal to

 
.id x̋ id x̋ !/.VŒ13 VŒ23 / and .id x̋ id x̋ !/.VŒ13 VŒ23 /;
 7.2. The legs of a multiplicative unitary 187

respectively, for all ! 2 L.H / . These identifications furthermore suggest to think

of the expressions m..Sy ˝ id/.. y aO ! /// and m..id ˝Sy/..
y aO ! ///, which are not
defined again, as being equal, respectively, to
.id x̋ !/.V  V / and .id x̋ !/.V V  /:
Now, since V is unitary, the last two operators are equal to .id x̋ !/.idH ˝H / D
idH !.idH / D idH O .aO ! /. Of course, these symbolic identifications and calcula-
tions are not precise and can only serve as a motivation.
We conclude this subsection with two examples:
Example 7.2.18. Consider the legs of the multiplicative unitary WG associated to
a locally compact group G (see Example 7.1.4 and 7.2.13). We calculate the dual
pairing and the maps O ; Sy and ; S for operators of the form
aO ! D .id x̋ !/.WG /; where ! D ! 0 ; for some ;  0 2 L2 .G; /;
and
a D . x̋ id/.WG /; where  D !0 ; for some ; 0 2 L2 .G; /:
The calculations in Example 7.2.13 showed that
aO ! D
M .f /; where f 2 C0 .G/ is given by f .x/ D h 0 jLx i for all x 2 G;
a D L.g/; where g 2 L1 .G; / is given by g.x/ D 0 .x/.x/ for all x 2 G:
Here,
M denotes the representation by multiplication operators and L denotes left
convolution.
The dual pairing .  j  / W Ay0 .WG /  A0 .WG / ! C is given by
Z
0
.aO ! ja / D .aO ! / D h j
M .f /i D g.x/f .x/d .x/:
G

The maps O and are given by

Z
0 0
O .aO ! / D !.id/ D h ji D f .e/; .a / D .id/ D h ji D g.x/d .x/I
G

here, e 2 G denotes the unit. Finally, similar calculations as in Example 7.2.13

show that
Sy.aO ! / D
M .fQ/; where fQ 2 C0 .G/ is given by fQ.x/ D h 0 jLx 1 i D f .x 1 /;
S.a / D L.g/;
Q where gQ 2 L1 .G; / is given by g.x/
Q D g.x 1 /ı.x/1 I
here, ı denotes the modular function of G.
A comparison with Example 4.2.2 shows that the maps O , Sy and , S are given
by the same formulas like the counit and the antipode of the C  -bialgebras C0 .G/
and Cr .G/, respectively.
188 Chapter 7. Multiplicative unitaries

Example 7.2.19. Consider the legs of the multiplicative unitary V WD VA0 asso-
ciated to an algebraic quantum group .A0 ; 0 / with positive right integral (see
Example 7.1.6 and Theorem 7.2.14). We calculate the dual pairing and the maps O ,
Sy and , S for operators of the form
aO ! D .id x̋ !/.V /; where ! D !c;d for some c; d 2 A0 ;
and
a D . x̋ id/.V /; where  D !e;f for some e; f 2 A0 :
The calculations in the proof of Theorem 7.2.14 showed that
aO ! D .a/;
O where aO D .c   d / 2 Ay0 ;
X
a D
.a/; where a D .e  f.1/ /f.2/ 2 A0 I
for the definition of
and , see Theorem 7.2.14.
The dual pairing .  j  / W Ay0 .V /  A0 .V / ! C is given by
.aO ! ja / D !.a / D hcj
.a/d i D .c  ad / D a.a/I
O

thus it extends the natural pairing Ay0  A0 ! C. The map O is given by

O .aO ! / D !.id/ D hcjd i D .c  d /:
Using the modular automorphism of (Remark 2.2.18 i)) and the explicit formula
for the counit O0 of .Ay0 ; 
y 0 / given in Proposition 2.3.3), it is easy to see that

O Thus the map O corresponds to the counit O0 of Ay0 . Likewise,
.c d / D O0 .a/.
the map defined in Proposition 7.2.16 corresponds to the counit 0 of .A0 ; 0 /:
.a / D .id/ D hejf i D .e  f / D 0 .a/:
y aO ! / D .aO !  / and S.a / D .a  / . Using the relations !  D
Let us compute S.

!d;c ,  D !f;e and the calculations in the proof of Theorem 7.2.14, we find

S. O  ; where bO D
y aO ! / D .b/ .d   c/;
X
S.a / D
.b/ ; where b D .f  e.1/ /e.2/ :

Denote by Sy0 and S0 the antipode of .Ay0 ; 

y 0 / and .A0 ; 0 /, respectively. By
y
definition of the involution on A0 ,
bO  .x/ D .d  S0 .x/ c/ D .c  S0 .x/d / D a.S
O 0 .x// for all x 2 A0 ;

and hence Sy. .a//

O D .Sy0 .a//.
O Using positivity of and Lemma 2.2.12, we find
b D . ˝ id/..e  /.f ˝ 1// D . ˝ S0 /..e  ˝ 1/.f // D S0 .a/;
and therefore S.
.a// D
.S0 .a//.
 7.3. Classes of well-behaved multiplicative unitaries 189

7.3 Classes of well-behaved multiplicative unitaries

Evidently, it is desirable to have simple general criteria that tell whether a given
multiplicative unitary is (weakly) well-behaved or not. The first such criterion –
regularity – was introduced by Baaj and Skandalis [7]; variants were studied in [4],
[5], [44]. Some examples of multiplicative unitaries are regular, but other important
examples are not (see Example 7.3.4 v)) – regularity is too restrictive to cover all
multiplicative unitaries that arise from C  -algebraic quantum groups.
In [201], Woronowicz introduced a manageability condition that is particularly
well adapted to C  -algebraic quantum groups. In particular, the multiplicative
unitary of every locally compact quantum group is manageable, see Section 8.3 or
[91], [110]. For some examples of quantum groups, it is easier to construct a mul-
tiplicative unitary which is not manageable but satisfies a more general modularity
condition that was introduced by Soltan and Woronowicz in [142].
In this section, we discuss all these criteria and show that each of them is
sufficient for the well-behavior of the multiplicative unitary under consideration.

7.3.1 Regular multiplicative unitaries

Let V be a multiplicative unitary on a Hilbert space H . We consider the space
C0 .V / WD f.! x̋ id/.V †/ j ! 2 L.H / g  L.H /;
where † 2 L.H ˝ H / denotes the flip  ˝  7!  ˝  as usual. Put
C .V / WD k  k-closure of C0 .V /; Cw .V / WD w-closure of C0 .V /;
where w-closure denotes the closure with respect to the weak operator topology.
Note that .! x̋ id/.V †/ D ..id x̋ !/ ı Ad† /.V †/ D .id x̋ !/.†V / for all
! 2 L.H / .
Definition 7.3.1. A multiplicative unitary V on a Hilbert space H is called
• regular if K.H / D C .V /,
• semi-regular if K.H /  C .V /,
• weakly regular if L.H / D Cw .V /.
Remark 7.3.2. If V is regular, then it is also semi-regular. Moreover, if V
is semi-regular, then it is also weakly regular, because in that case, L.H / D
w-closure of K.H /  w-closure of C .V / D Cw .V /.
Before we explain the relevance of the conditions introduced above, let us refor-
mulate the definition of the spaces C.V / and Cw .V / in terms of ket-bra operators
and consider some examples. Recall the maps
jiŒ1 ; jiŒ2 W H ! H ˝ H and hjŒ1 ; hjŒ2 W H ˝ H ! H; where  2 H;
190 Chapter 7. Multiplicative unitaries

that were introduced in the formulas (7.6)–(7.8) on page 176. In terms of these maps,
we can rewrite an operator .! x̋ id/.V †/ 2 C0 .V /, where ! D !0 ; D h0 j  i
with 0 ;  2 H , as follows:

.! x̋ id/.V †/ D h0 jŒ1 V †jiŒ1

D h0 jŒ1 V jiŒ2
D h0 jŒ2 †V jiŒ2 D .id x̋ !/.†V /:

Lemma 7.3.3. Let V be a multiplicative unitary on a Hilbert space H . Then the

set fh0 jŒ1 V jiŒ2 j 0 ;  2 H g  L.H / is linearly dense in C .V / with respect to
the norm and in Cw .V / with respect to the weak operator topology.
Proof. The proof is virtually the same as the proof of Lemma 7.2.7. 
Examples 7.3.4. i) For every Hilbert space H , the operator idH ˝H is regular. This
follows from the relation

.h0 jŒ1 idH ˝H jiŒ2 / D h0 jŒ1 . ˝ / D h0 ji for all ; ; 0 2 H:

ii) If V is a (weakly) regular multiplicative unitary on a Hilbert space H , then

its opposite V op D †V  † is (weakly) regular again. Indeed, by Lemma 7.2.3, we
have for every ! 2 L.H /

.! x̋ id/.V op †/ D .! x̋ id/.†V  / D ..!  x̋ id/.V †// ;

and hence C0 .V op / D C0 .V / .
iii) The multiplicative unitaries VG and WG associated to a locally compact
group G (see Example 7.1.4) are regular. We prove this for WG ; for VG , the proof
is similar. Let 0 ;  2 L2 .G; /. The operator h0 jŒ1 WG jiŒ2 is determined by
Z
0
.h jŒ1 WG jiŒ2 /.y/ D 0 .x/.WG . ˝ //.x; y/d .x/
Z G

D 0 .x/.x/.x 1 y/d .x/ for all  2 L2 .G; /:

G
0
Thus, h jŒ1 WG jiŒ2 is an integral operator with kernel K0 ; W GG ! C given by
K0 ; .y; x/ D 0 .x/.x 1 y/. It is easy to see that K0 ; belongs to L2 .GG; /,
and the space of all integral operators with kernel in L2 .G  G;   / is a dense
subspace of K.L2 .G; // [28, II, Proposition 4.7]. Combining these observations
with Lemma 7.3.3, we find that C .WG / is contained in K.L2 .G; //. A standard
argument shows that the linear span of functions of the form K0 ; , where 0 ;  2
L2 .G; /, is dense in L2 .G  G;   /, and therefore, C.WG / D K.L2 .G; //.
iv) The multiplicative unitaries VA0 and WA0 associated to an algebraic quantum
group .A0 ; 0 / (see Example 7.1.6) are regular. We prove this for V WD VA0 ; the
 7.3. Classes of well-behaved multiplicative unitaries 191

proof for WA0 is similar. Let us use the notation introduced in Example 7.1.6 and
Theorem 7.2.14. An operator of the form hcjŒ1 V jbiŒ2 , where b; c 2 A0 , acts on
an element a 2 A0 as follows:
X X
hcjŒ1 V jbiŒ2 a D hcjŒ1 V .a ˝ b/ D hcja.1/ ia.2/ b D .c  a.1/ /a.2/ b:

Using the modular automorphism  0 of (see Remark 2.2.18 i)), we can rewrite
the right-hand side in the form
X
.a.1/  0 .c  //a.2/ b D . ˇ id/.0 .a/. 0 .c  / ˇ b//:

Since the map A0 ˇ A0 ! A0 ˇ A0 ; c 0 ˇ bP 0

7! .c 0 /.1 ˇ b 0 /, is surjective (see
Definition 2.1.9), we can write  .c / ˇ b D i 0 .ci0 /.1 ˇ bi0 / with ci0 ; bi0 2 A0 .
0 

We insert this equation into the expression above, use right-invariance of , and
find
X X
hcjŒ1 V jbiŒ2 a D . ˇ id/.0 .aci0 /.1 ˇ bi0 // D .aci0 /bi0 :
i i

Define ci00 2 A0 by ci00  WD  01 .ci0 /. Then .aci0 / D .ci00  a/ D hci00 jai, and
X
hcjŒ1 V jbiŒ2 a D hci00 jaibi0 :
i

This equation shows that the operator hcjŒ1 V jbiŒ2 is compact. By Lemma 7.3.3,
C .V /  K.H /. Reversing the preceding transformations, we find that this
inclusion is an equality.
v) The bicrossed product construction [9] yields examples of multiplicative
unitaries that are semi-regular but not regular, and examples that are not even semi-
regular, see [9, Theorem 3.11 and Section 4].
vi) The multiplicative unitary of the quantum group E .2/ is semi-regular but
not regular, see [4], [5] or Section 8.4.2.
Semi-regularity, regularity, and weak regularity have interesting implications
on the legs of a multiplicative unitary. Most importantly, every (weakly) regular
multiplicative unitary is (weakly) well-behaved. The proof of this result, which is
given at the end of this subsection, and the proofs of the other results discussed in
the remainder of this subsection all use similar techniques:
Given a multiplicative unitary V on a Hilbert space H , we perform purely alge-
braic manipulations on operators between threefold tensor products of the Hilbert
spaces H and C like, for example, H ˝ H ˝ H , C ˝ H ˝ H , H ˝ C ˝ C. Of
course, we can always neglect the factor C wherever it occurs, but the algebraic
manipulations will be easier to follow if we do not make such identifications during
the calculations. To streamline the presentation, we adopt the following notation:
192 Chapter 7. Multiplicative unitaries

• We extend the leg notation to operators between different spaces in a similar

way as in formulas (7.6)–(7.8) on page 176; for example, we denote the
operator

idC ˝ji ˝ idH W C ˝ C ˝ H ! C ˝ H ˝ H

by jiŒ2 and its adjoint by hjŒ2 . The symbols jiŒ2 and hjŒ2 will also denote
the operators idH ˝ji ˝ idH , idH ˝ji ˝ idC , idC ˝ji ˝ idC and their
respective adjoints; from the context, it will be clear which of these operators
we refer to.
• We put

jH i WD fji j  2 H g D K.C; H /
and
hH j WD fhj j  2 H g D K.H; C/;

and apply the leg notation to the spaces hH j and jH i in the obvious way.
• We indicate identifications of tensor products that involve C as a factor and
identifications of operators on such tensor products by the symbol “
”; for
example, H ˝ C
H
C ˝ H and L.H ˝ C; C ˝ H /
L.H /.

Let us illustrate this notation by some examples that will be used later on.
 
• K.H /
K.H ˝ C; C ˝ H / D jH iŒ2 hH jŒ1 .

• By Lemma 7.3.3,
     
C .V /
hH jŒ1 V jH iŒ2
hH jŒ1 V †jH iŒ1
hH jŒ2 †V jH iŒ2 :

In particular, V is regular if and only if

   
hH jŒ1 V jH iŒ2 D jH iŒ2 hH jŒ1 : (7.14)

 
y /
hH jŒ2 V jH iŒ2  L.H ˝ C/
L.H /.
• By Lemma 7.2.7, A.V

The following results explain the relevance of Definition 7.3.1. In the proofs,
we permanently use Lemma 12.5.3. First, we study the space C.V /.

Lemma 7.3.5. Let V be a multiplicative unitary on a Hilbert space H . Then the set
C .V /C.V / is linearly dense in C.V /; in particular, C .V / and Cw .V / are algebras.
Furthermore, C .V /H and C .V / H are linearly dense in H .
 7.3. Classes of well-behaved multiplicative unitaries 193

Proof. With respect to the identification L.H /

L.H ˝ C ˝ C; C ˝ C ˝ H /,
 
ŒC .V /C.V /
hH jŒ2 VŒ23 jH iŒ3 hH jŒ1 VŒ12 jH iŒ2
 
D hH ˝ H jŒ12 VŒ23 VŒ12 jH ˝ H iŒ23
 
D hH ˝ H jŒ12 VŒ12 VŒ13 VŒ23 jH ˝ H iŒ23
   
D hH ˝ H jŒ12 VŒ13 jH ˝ H iŒ23 D hH jŒ1 VŒ13 jH iŒ3
C .V /:

Moreover,

ŒC .V /H  D ŒhH jŒ1 V .H ˝ H / D ŒhH jŒ1 .H ˝ H / D H;

ŒC .V / H  D ŒhH jŒ2 V  .H ˝ H / D ŒhH jŒ2 .H ˝ H / D H: 

Proposition 7.3.6. If V is a semi-regular multiplicative unitary, then C.V / is a

C  -algebra.
For the proof of this proposition, we use the following result:
Lemma 7.3.7. Let H be a Hilbert space and C  L.H / a norm-closed subalgebra
such that C  C  C and C C   C . Then C is a C  -algebra.
Proof. We only need to prove C   C . For every c 2 C , there exists a sequence of
polynomials .pn /n such that c D limn cpn .c  c/, and then c  D limn pn .c  c/c  2
ŒC  C C    C . 
Proof of Proposition 7.3.6. Let H denote the underlying Hilbert space of V . By
definition,
 
ŒC .V /C .V / 
hH jŒ1 VŒ13 jH iŒ3 hH jŒ2 VŒ12

jH iŒ1
 L.C ˝ H ˝ C; C ˝ C ˝ H /:
   
Since V is semi-regular, jH iŒ3 hH jŒ2  hH jŒ2 VŒ23 jH iŒ3 . Thus
 
ŒC .V /C .V /   hH jŒ1 VŒ13 hH jŒ2 VŒ23 jH iŒ3 VŒ12

jH iŒ1
 

D hH ˝ H jŒ12 VŒ13 VŒ23 VŒ12 jH ˝ H iŒ13 :
 
By the pentagon equation, VŒ13 VŒ23 VŒ12 D VŒ12 VŒ23 . Therefore
 
ŒC .V /C .V /   hH ˝ H jŒ12 VŒ12
VŒ23 jH ˝ H iŒ13
 
D hH ˝ H jŒ12 VŒ23 jH ˝ H iŒ13
 
D hH jŒ2 VŒ23 jH iŒ3
C .V /:

A similar calculation shows that ŒC .V / C .V / is contained in C.V /. Now the

assertion follows from Lemma 7.3.5 and Lemma 7.3.7. 
194 Chapter 7. Multiplicative unitaries

For a general multiplicative unitary V , the space C .V / need not be closed under
y / and A.V / are closed under involution, as we
involution. If it is, then also A.V
shall see in the next lemma, but there exist multiplicative unitaries for which this is
not the case [9, Remark 4.5].
Lemma 7.3.8. Let V be a multiplicative unitary that satisfies C .V / D C .V /.
y / and A.V / are C  -algebras.
Then A.V
Proof. Let H denote the underlying Hilbert space of V . By Lemma 7.2.2, Ay0 .V /
y / D A.V
and A0 .V / are algebras; thus we only need to show that A.V y / and

A.V / D A.V /. Let us prove the first equation. By assumption, the space
 
By WD hH jŒ2 V .1 ˝ C .V //V  jH iŒ2  L.H ˝ C/

is self-adjoint. Inserting the definition of C .V / and using the pentagon equation,

we find
 
By
hH jŒ3 VŒ13 hH jŒ2 VŒ23 jH iŒ3 VŒ12

jH iŒ2
 

D hH ˝ H jŒ23 VŒ13 VŒ23 VŒ12 jH ˝ H iŒ23
 

D hH ˝ H jŒ23 VŒ12 VŒ23 jH ˝ H iŒ23
   
D hH ˝ H jŒ23 VŒ12
jH ˝ H iŒ23
hH jŒ2 VŒ12  y / :
jH iŒ2 D A.V

Since By was self-adjoint, so is A.V

y /. The assertion concerning A.V / can be proved
similarly. Alternatively, apply the preceding proof to V op , and use Example 7.3.4 ii)
and Lemma 7.2.5. 
y / and
Corollary 7.3.9. Let V be a semi-regular multiplicative unitary. Then A.V

A.V / are C -algebras.
Proof. Combine Proposition 7.3.6 with Lemma 7.3.8. 
The following result is due to Enock [44, Proposition 3.12], who proved it in a
much more general setting, see Theorem 10.3.18 in Section 10.3.
Theorem 7.3.10. If V is a multiplicative unitary that satisfies Cw .V / D Cw .V /,
then V is weakly well-behaved. In particular, every weakly regular multiplicative
unitary is weakly well-behaved.
Proof. Similar calculations as in the proof of Lemma 7.3.8 show that Ayw .V / is
closed under involution Рsimply let Π  denote the weakly closed linear span instead
of the norm closed linear span, and use Lemma 12.5.3. Since Ayw .V / acts non-
degenerately on H (Lemma 7.2.2), it is a von Neumann algebra. Replacing V
by V op , which is regular by Example 7.3.4 ii), we find that also Aw .V / is a von
Neumann algebra; here, we use Lemma 7.2.5. 
 7.3. Classes of well-behaved multiplicative unitaries 195

The main result of this subsection is the following theorem of Baaj and Skandalis
[7, Théorème 3.8]:
Theorem 7.3.11. Every regular multiplicative unitary is well-behaved.
We divide the proof into several steps. From now on, let V be a regular multi-
plicative unitary on a Hilbert space H . To simplify notation, we put Ay WD A.V
y /.
Lemma 7.3.12. Ay is a C  -algebra.
Proof. This follows immediately from Corollary 7.3.9, but let us give an alternative
proof that is more direct. By definition,
 
ŒAyAy 
hH jŒ3 VŒ13 jH iŒ3 hH jŒ2 VŒ12

jH iŒ2  L.H ˝ C ˝ C/:
Since V is regular, we can replace jH iŒ3 hH jŒ2 by hH jŒ2 VŒ23 jH iŒ3 in the middle:
 
ŒAyAy 
hH jŒ3 VŒ13 hH jŒ2 VŒ23 jH iŒ3 VŒ12

jH iŒ2
 

D hH ˝ H jŒ23 VŒ13 VŒ23 VŒ12 jH ˝ H iŒ23 :
 
We use the pentagon equation to replace VŒ13 VŒ23 VŒ12 by VŒ12 VŒ23 , and obtain
 
ŒAyAy 
hH ˝ H jŒ23 VŒ12

VŒ23 jH ˝ H iŒ23
 

D hH ˝ H jŒ23 VŒ12 jH ˝ H iŒ23
Ay :

This equation implies Ay D Ay and AyAy  A.

y 
Lemma 7.3.13. i) Œ.Ay ˝ hH j/V  D ŒAy ˝ hH j  L.H ˝ H; H ˝ C/.
ii) ŒV .Ay ˝ jH i/ D ŒAy ˝ jH i  L.H ˝ C; H ˝ H /.
Proof. i) By definition,
 
Œ.Ay ˝ hH j/V 
hH jŒ2 VŒ12 jH iŒ2 hH jŒ3 VŒ13
 L.H ˝ C ˝ C; H ˝ C ˝ H /;
and by the pentagon equation,
 
Œ.Ay ˝ hH j/V 
hH ˝ H jŒ23 VŒ12 VŒ13 jH iŒ2
 

D hH ˝ H jŒ23 VŒ23 VŒ12 VŒ23 jH iŒ2
 

D hH ˝ H jŒ23 VŒ12 VŒ23 jH iŒ2 :
We move hH jŒ3 to the right of VŒ12 , use regularity of V , and find
 
Œ.Ay ˝ hH j/V 
hH jŒ2 VŒ12 hH jŒ3 VŒ23

jH iŒ2
 
D hH jŒ2 VŒ12 jH iŒ2 hH jŒ3
ŒAy ˝ hH j:

ii) Multiply the equation in i) by V  on the right, take adjoints, and identify Ay
with Ay using Lemma 7.3.12. 
196 Chapter 7. Multiplicative unitaries

Lemma 7.3.14. Œ.Ay ˝ 1/.

y Ay / D ŒAy ˝ Ay  D Œ.
y Ay /.1 ˝ Ay /.
y and Ay together with the pentagon equation imply
Proof. The definition of 
 
y Ay / D V  .1 ˝ A/V
. y
V  hH j V jH i V
Œ12 Œ3 Œ23 Œ3 Œ12
 

D hH jŒ3 VŒ12 VŒ23 VŒ12 jH iŒ3
 
D hH jŒ3 VŒ13 VŒ23 jH iŒ3  L.H ˝ H ˝ C/:
We insert this relation into the left- and right-hand side of the equation that we want
to prove, apply the previous lemma, and obtain
 
Œ.Ay ˝ 1/.
y Ay /
AyŒ1 hH jŒ3 VŒ13 VŒ23 jH iŒ3
 
D AyŒ1 hH jŒ3 VŒ23 jH iŒ3
ŒAy ˝ Ay 
and
 
y A/.1
Œ. y ˝ Ay /
hH jŒ3 VŒ13 VŒ23 jH iŒ3 AyŒ2
 
D hH jŒ3 VŒ13 jH iŒ3 AyŒ2
ŒAy ˝ Ay : 

Proof of Theorem 7.3.11. If V is a regular multiplicative unitary, then A.Vy / is a

 y y 
C -algebra by Lemma 7.3.12, and .A.V /; / is a bisimplifiable C -bialgebra by
Lemma 7.2.1 and Lemma 7.3.14. Replacing V by V op , we obtain the respective
y op / D A.V /, see also Lemma 7.2.5.
statements for A.V
It remains to show that V 2 M.A.Vy / ˝ A.V //. Lemma 7.3.13 implies that
y / ˝ K.H /, and a similar argument shows that V is a
V is a multiplier of A.V
 
multiplier of K.H / ˝ A.V /. Therefore, VŒ13 D VŒ12 VŒ23 VŒ12 VŒ23 belongs to
M.A.Vy / ˝ K.H / ˝ A.V //. The claim follows. 
For completeness, we cite the following result of Baaj [5, Théorème 3.12]. The
proof involves similar techniques as presented above. Recall that a symmetry in a
C  -algebra is a self-adjoint unitary.
Theorem 7.3.15. Let V be a multiplicative unitary on a Hilbert space H and
assume that
i) V is semi-biregular in the sense that K.H / is contained in
span f.! x̋ id/.V †/ j ! 2 L.H / g
and
span f.id x̋ !/.V †/ j ! 2 L.H / gI

ii) V is balanced in the sense that there exists a symmetry U 2 L.H / such that
†.U ˝ 1/V .U ˝ 1/† is a multiplicative unitary.
Then V is well-behaved.
 7.3. Classes of well-behaved multiplicative unitaries 197

7.3.2 Manageable and modular multiplicative unitaries

The manageability condition and the more general modularity condition were in-
troduced by Woronowicz [201] and by Soltan and Woronowicz [142], respectively.
They are particularly well adapted to unitaries associated to quantum groups and can
be considered as modifications of the regularity condition discussed in the previous
section.
We formulate the manageability and the modularity condition and show that
every modular multiplicative unitary is well-behaved, building on the proofs given
in the preceding section. Moreover, we discuss the polar decomposition of the
antipode for a modular multiplicative unitary, but omit the proof. All results are
taken from [201] and [142].
The manageability and the modularity condition involve the conjugate of a
Hilbert space and unbounded self-adjoint operators; a standard references for the
latter are, for example, [28], [75], [136]. We denote the conjugate of a Hilbert space
H by H x and the canonical conjugate-linear isomorphism H ! H N thus,
x by  7! ;
N
N D  and hj N
N i D hji for all ;  2 H and  2 C. For each T 2 L.H /,
the map N 7! T  defines an operator Tx 2 L.H x /, and the map T 7! Tx is a
conjugate-linear -isomorphism L.H / ! L.H x /.
The precise formulation of the manageability and the modularity condition may
appear a bit mysterious at first glance; therefore we begin with a motivation. Let V
be a multiplicative unitary on a Hilbert space H . Then the regularity condition can
be expressed in terms of the map

ˆ W B.H / ! B.H /; ! 7! .! x̋ id/.V †/:

Indeed, V is regular if and only if ˆ.B.H / / is a dense subset of K.H /. Assume

that H has finite dimension. Then H x ˇ H can be identified with B.H / via
N ˇ 
hj  i, and with B.H / via N ˇ 
jihj. With respect to these
identifications, ˆ corresponds to the linear map Vz W H x ˇH !H x ˇ H given by
X X
Vz .x0 ˇ / D S00i ˇ i00 , h0 jŒ1 V jiŒ2 D ji00 ih00i j:
i i

If 0 ; ; i00 ; 00i 2 H are as above and ;  0 2 H , then

X
h0 ˇ  0 jV . ˇ /i D h 0 jh0 jŒ1 V jiŒ2 i D h 0 ji00 ih00i ji
i
X
D hN ˇ  0 jS00i ˇ i00 i D hN ˇ  0 jVz .x0 ˇ /i:
i

The existence of such a relation, modified by an additional scaling operator, is the

key idea of modularity and manageability:
198 Chapter 7. Multiplicative unitaries

Definition 7.3.16. A multiplicative unitary V on a Hilbert space H is modular if

y and Q on H
there exist positive self-adjoint (not necessarily bounded) operators Q
and a unitary operator Vz on H x ˝ H such that
y D ker Q D f0g;
i) ker Q
y ˝ Q/V D Q
ii) V  .Q y ˝ Q;

iii) for all ; 0 2 H and  2 Dom.Q1 /,  0 2 Dom.Q/,

h0 ˝  0 jV . ˝ /i D hN ˝ Q 0 jVz .x0 ˝ Q1 /i: (7.15)

y D Q.
We call V manageable if we can choose Q
The opposite of a modular or manageable multiplicative unitary is modular or
manageable again:
Proposition 7.3.17. Let V be a modular/manageable multiplicative unitary on a
y Vz as in the definition above.
Hilbert space H and Q; Q;
S
i) Vz  .Q S
y ˝ Q1 /Vz D Q
y ˝ Q1 .

ii) For all ;  0 2 H and 0 2 Dom.Q y 1 /,  2 Dom.Q/,y

˝ ˇ ˛
h0 ˝  0 jV . ˝ /i D Q y ˝  0 ˇVz .Q
y 1 0 ˝ / :

iii) The multiplicative unitary V op D †V  † is modular/manageable.

Proof. i) Let t 2 R. Since V  .Q y ˝ Q/V D Q y ˝ Q, the unitary Q
y it ˝ Qit
commutes with V . Therefore the left-hand side of equation (7.15) does not change
when we replace 0 ;  0 ; ;  by Q y it 0 , Qit  0 , Q
y it , Qit , and
˝ ˇ ˛
Qy it  ˝ Qit Q 0 ˇVz .Qy it 0 ˝ Qit Q1 / D hN ˝ Q 0 jVz .x0 ˝ Q1 /i

for all ; 0 2 H and  2 Dom.Q1 /,  0 2 Dom.Q/. Consequently, Vz commutes

with Q y it ˝Qit D .Q/ x
y
y it ˝Qit and, since t 2 R was arbitrary, also with Q˝Q 1
.
1 0
ii) For  2 Dom.Q / and  2 Dom.Q/, the formula follows directly from i)
and equation (7.15), and for arbitrary ;  0 2 H by a simple continuity argument.
y
iii) The relation V  .Q˝Q/V y
D Q˝Q y op D Q˝ Q,
implies .V op / .Q˝ Q/V y
0 0
and by ii), we have for all ;  2 H and  2 Dom.Q y /,  2 Dom.Q/
1 y

hV op . 0 ˝ 0 /j ˝ i D h0 ˝  0 jV . ˝ /i

˝ ˇ ˛
D Qy ˝  0 ˇVz .Qy 1 0 ˝ /
˝ ˇ ˛
D †Vz †.N ˝ Q y 1 0 /ˇx0 ˝ Q
y :
 7.3. Classes of well-behaved multiplicative unitaries 199

e
We put V op WD †Vz †, conjugate the equation above, and find
˝
h ˝ jV op . 0 ˝ 0 /i D x0 ˝ Q
ˇ
e
y ˇV op .N ˝ Q
˛
y 1 0 / : 

The first main result concerning modular unitaries is the following theorem of
Woronowicz and Soltan [201, Theorem 1.5], [142, Theorem 2.3]:
Theorem 7.3.18. Every modular multiplicative unitary is well-behaved.
Before we give the proof, let us state the second main result concerning modular
unitaries.
For each well-behaved multiplicative unitary V , we defined an antipode
S W A0 .V / ! A.V /, where A0 .V / D f.! x̋ id/.V / j ! 2 L.H / g  A.V /,
by

S..! x̋ id/.V // D .! x̋ id/.V  / for each ! 2 L.H / ;

see Section 7.2.4. If V is modular, much more information about this antipode is
available. Most importantly, it can de described in terms of a -antiautomorphism
R W A.V / ! A.V / called the unitary antipode, and of a one-parameter group of
automorphisms . t / t of A.V / called the scaling group. We only state the result and
refer to [201, Theorem 1.5] and [142, Theorem 2.3] for the proof. For background
on one-parameter groups and their analytic generators, see Section 8.1.3.
Theorem 7.3.19. Let V be a modular multiplicative unitary on a Hilbert space H
y Vz as in Definition 7.3.16.
and Q, Q,
i) The map S W A0 .V / ! A.V / extends to a closed linear map S W Dom.S / 
A.V / ! A.V /.
ii) Dom.S/  A.V / is a subalgebra and S.ab/ D S.b/S.a/ for all a; b 2
Dom.S/.
iii) S.Dom.S // D Dom.S / and S.S.a/ / D a for all a 2 Dom.S /.
iv) The map S admits the following polar decomposition: S D R ı i=2 , where
R is a -antiautomorphism of A.V / and i=2 is the analytic generator of a
one-parameter group  D . t / t2R of -automorphisms of the C  -algebra
A.V /.
v) R commutes with  t for all t 2 R; in particular, Dom.S / D Dom.i=2 /.
vi) R and  are uniquely determined by S .
vii)  ı  t D . t ˝  t / ı  for all t 2 R, and  ı R D † ı .R ˝ R/ ı , where
† W A ˝ A ! A ˝ A denotes the flip map.
200 Chapter 7. Multiplicative unitaries

viii) Let Vz and Q be the operators related to V as in Definition 7.3.16. Then

(a)  t .a/ D Q2it aQ2it for all t 2 R, a 2 A.V /.
(b) Denote by V >˝R the image of V under the extension of the -antihomo-
morphism A.Vy / ˝ A.V / ! L.H x / ˝ A.V /; X ˝ Y 7! Xx  ˝ R.Y /,
to the multiplier algebra. Then Vz  D V >˝R .
Remark 7.3.20. Replacing the unitary V by its opposite V op , we obtain a closed
extension of the map Sy W Ay0 .V / ! A.V
y / introduced in Proposition 7.2.16 with a
similar polar decomposition as above.
In the remainder of this subsection, we prove Theorem 7.3.18. We fix a mod-
ular multiplicative unitary V on a Hilbert space H and choose Q, y Q, Vz as in
Definition 7.3.16.
The strategy of the proof is to establish a “Q-weighted regularity” for V
(Lemma 7.3.23) and to follow the proof of Theorem 7.3.11, replacing usual regular-
ity by the modified version. The formulation of this Q-weighted regularity involves
Hilbert–Schmidt operators; two standard references are [75, Section 2.7] and [113,
Section 2.4]. Put
H Q .H / WD f 2 L.H / j H  Dom.Q1 / and
Q1 is of Hilbert–Schmidt classg:
This is a Banach space with respect to the Q-weighted Hilbert–Schmidt-norm

1=2
k kQ WD kQ1 k2 D Tr.  Q2 / I
here, k  k2 denotes the usual Hilbert–Schmidt norm, and Tr denotes the usual trace.
The space H Q .H / can be described as follows:
Lemma 7.3.21. An operator 2 L.H / belongs to H Q .H / if and only if there
x ˝ H such that
exists a vector ‰. / 2 H
h 0 j i D hN ˝ Q 0 j‰. /i for all  0 2 Dom.Q/;  2 H:
x ˝H .
The correspondence 7! ‰. / is a linear isometric bijection H Q .H / ! H
Proof. This follows easily from the well-known bijective correspondence between
Hx ˝ H and the space of all Hilbert–Schmidt-class operators on H , see [75, Propo-
sition 2.6.9]. 
Example 7.3.22. For each 0 2 H and  2 Dom.Q1 /, the operator WD
jih0 j belongs to H Q .H /: its image is evidently contained in Dom.Q1 /, and
Q1 jih0 j D jQ1 ih0 j is a Hilbert–Schmidt operator. Since
h 0 j i D h 0 jih0 ji D hN ˝ Q 0 jx0 ˝ Q1 i for all  0 2 Dom.Q/;  2 H;
 7.3. Classes of well-behaved multiplicative unitaries 201

the condition of the previous lemma is satisfied with ‰. / D x0 ˝ Q1 .

The linear span of all operators of the form jih0 j, where 0 2 H and  2
Dom.Q1 /, is dense in H Q .H / because H x ˇ Dom.Q/ is dense in H x ˝ H.

The next lemma shows that V satisfies a Q-weighted regularity condition:

Lemma 7.3.23. For each 0 2 H and  2 Dom.Q1 /, the operator WD

h0 jŒ1 V jiŒ2 belongs to H Q .H /, and the linear span of all operators of this
form is dense in H Q .H /.

Proof. Let 0 ; ; be as above. Then for all  0 2 Dom.Q/ and  2 H ,

h 0 j i D h0 ˝  0 jV . ˝ /i D hN ˝ Q 0 jVz .x0 ˝ Q1 /i:

This relation shows that the condition of Lemma 7.3.21 is satisfied with ‰. / D
Vz .x0 ˝ Q1 / 2 H
x ˝ H . The linear span of all operators of this form is dense in
H Q .H / because Vz is unitary and Hx ˇ Dom.Q/ is dense in H x ˝ H. 

Given a subset X  H Q .H /  L.H /, we denote by ŒX Q  H Q .H /

and ŒX  L.H / the closure of the linear span of X , taken with respect to the
Q-weighted Hilbert–Schmidt norm or the usual operator norm. Example 7.3.22
and Lemma 7.3.23 imply the following analogue of equation (7.14) (page 192):
 
Œj Dom.Q1 /ihH jQ D H Q .H / D hH jŒ1 V j Dom.Q1 /iŒ2 Q : (7.16)

The following proposition allows us to adopt the proof of Theorem 7.3.11 to

Theorem 7.3.18:

Proposition 7.3.24. i) Let  0 2 Dom.Q/ and 2 H Q .H /. Then the operator

T WD h 0 jŒ2 V .1 ˝ / 2 L.H ˝ H; H ˝ C/ satisfies kT k  kQ 0 k  k kQ .
ii) ŒhH jŒ2 V .1 ˝ X / D ŒhH jŒ2 V .1 ˝ ŒX Q / for each subset X  H Q .H /.

Proof. i) Let ; 0 2 H and  2 Dom.Q1 /. Then  2 Dom.Q1 / and

h0 jT . ˝ /i D h0 ˝  0 jV . ˝ /i D hN ˝ Q 0 jVz .x0 ˝ Q1 /i:

Since kVz k D 1 and kQ1 k  kQ1 k2 D k kQ , the equation above implies

jh0 jT . ˝ /ij  kk  kQ 0 k  k0 k  k kQ  kk:

Since Dom.Q1 / is dense in H , the inequality extends to all  L 2 H . Using an

orthonormal basis .ei /i of H and the decomposition H ˝ H D i Cei ˝ H , it
is easy to deduce that jh0 jT ij  kQ 0 k  k0 k  k kQ  kk for all 0 2 H and
 2 H ˝ H.
202 Chapter 7. Multiplicative unitaries

ii) By statement i), Œh 0 jŒ2 V .1 ˝ X / D Œh 0 jŒ2 V .1 ˝ ŒX Q / for every  0 2

Dom.Q/, and since Dom.Q/ is dense in H ,

ŒhH jŒ2 V .1 ˝ X / D ŒhDom.Q/jŒ2 V .1 ˝ X /

D ŒhDom.Q/jŒ2 V .1 ˝ ŒX Q /
D ŒhH jŒ2 V .1 ˝ ŒX Q /: 

Proof of Theorem 7.3.18. The proof proceeds along the same line as the proof of
Theorem 7.3.11; let us briefly indicate the necessary modifications. The first step
is to show that Ay D A.V
y / is a C  -algebra. Since Dom.Q1 / is dense in H ,

ŒAyAy  D ŒhH jŒ3 VŒ13 jH iŒ3 hH jŒ2 VŒ12


jH iŒ2 
 1 

D hH jŒ3 VŒ13 j Dom.Q /iŒ3 hH jŒ2 VŒ12 jH iŒ2 :

By Proposition 7.3.24 ii) and equation (7.16), we can replace the middle term
j Dom.Q1 /iŒ3 hH jŒ2 by hH jŒ2 VŒ23 j Dom.Q1 /iŒ3 :
 
ŒAyAy  D hH jŒ3 VŒ13 hH jŒ2 VŒ23 j Dom.Q1 /iŒ3 VŒ12

jH iŒ2 :

Now we replace Dom.Q1 / by H again and proceed as in Lemma 7.3.12.

The rest of the proof of Theorem 7.3.11 carries over similarly. 
Chapter 8
Locally compact quantum groups

The theory of locally compact quantum groups developed by Kustermans and Vaes
[87], [88], [91], [158] provides a comprehensive framework for the study of quantum
groups in the setting of C  -algebras and von Neumann algebras. It includes a far-
reaching generalization of the classical Pontrjagin duality of locally compact abelian
groups that covers all locally compact groups.
The theory developed by Kustermans and Vaes builds on work of Kac, Vainer-
man, Enock, Schwartz, Baaj, Skandalis, Van Daele, Woronowicz, and many others.
It is technically demanding, and the proofs of the main results are long and involved.
Therefore we only present a survey and refer the reader to the original articles [87],
[91], [93], [158] for details. Other surveys are [88], [89], [92]. In our presentation,
we shall profit from the material developed in the preceding chapters of this book,
in particular in Chapters 2 and 7, which motivates many constructions in the theory
of locally compact quantum groups. Some background on C  -algebras and von
Neumann algebras used in this chapter is summarized in the appendix.
Every locally compact quantum group appears in several guises:
• as a locally compact quantum group in the setting of von Neumann algebras,
• as a reduced C  -algebraic quantum group, and
• as a universal C  -algebraic quantum group,
see also Section 4.3. The first and second of these variants have a similar flavor;
they will form the topic of this chapter. For a discussion of the third variant, see
[87].

8.1 The concept of a locally compact quantum group

A locally compact quantum group is a von Neumann bialgebra or C  -bialgebra
equipped with a left and a right Haar weight. These Haar weights are analogues
of the left and the right Haar measure on a locally compact group and of the left
and the right integral of an algebraic quantum group. As indicated in Chapter 4,
these Haar weights are absolutely fundamental for the development of the theory.
Unfortunately, up to now, the existence of Haar weights on a given C  -bialgebra or
von Neumann bialgebra can not be deduced from a reasonable set of assumptions
but has to be postulated as an axiom. The situation improves if one starts from a
multiplicative unitary, see [205] and [67, Section 1.4]. A fundamental tool for the
treatment of the Haar weights is the celebrated Tomita–Takesaki theory.
204 Chapter 8. Locally compact quantum groups

We begin this section with a brief introduction to weights on C  -algebras and

von Neumann algebras. Then we define locally compact quantum groups in the
setting of von Neumann algebras and summarize the main results of the Tomita–
Takesaki theory that are fundamental to the theory of locally compact quantum
groups. In the setting of C  -algebras, similar results do not hold in general and
have to be incorporated into the definition of the Haar weights in form of addi-
tional assumptions. The section closes with the definition of reduced C  -algebraic
quantum groups.

8.1.1 Weights
The prototypical example of a Haar weight on a von Neumann bialgebra or C  -
bialgebra is integration of functions on a locally compact group G with respect to
the Haar measure . Unless G is compact, not every element of the von Neumann
algebra L1 .G; / or the C  -algebra C0 .G/ is integrable, and the integral is well
defined only on a dense subspace or on the cone of positive functions. Likewise,
a weight on a C  -algebra or von Neumann algebra is first defined on the cone of
positive elements and then extended to a linear map on the subset of integrable
elements. The precise definition subsumes von Neumann algebras as special C  -
algebras:

Definition 8.1.1. Let A be a C  -algebra. Denote by AC the set of all positive

elements of A. A weight on A is a function  W AC ! Œ0; 1 satisfying

i) .a C b/ D .a/ C .b/ for all a; b 2 AC ;

ii) .ra/ D r.a/ for all r 2 Œ0; 1/ and a 2 AC .

The weight  is called faithful if .a/ ¤ 0 for each non-zero a 2 AC .

Let  be a weight on a C  -algebra A. We use the following standard notation:

• M
C WD fa 2 AC j .a/ < 1g is the set of all positive -integrable elements;

• N
WD fa 2 A j .a a/ < 1g is the set of all -square-integrable elements;

• M
WD span M
C D span N
 N
is the set of all -integrable elements.

It is easy to check that N

is a left ideal and that M
is a -subalgebra of A [121,
Lemma 5.1.2]. Moreover,  extends uniquely to a linear functional on M
[121,
Lemma 5.1.2], which we denote by  again.
The theory of weights subsumes classical integration theory and is generally
considered as non-commutative integration theory.
 8.1. The concept of a locally compact quantum group 205

Example 8.1.2. Let X be a locally compact space with a Borel measure . Then in-
tegration with respect to  defines a weight on the von Neumann algebra L1 .X; /,
which we denote by  again:
Z
.f / WD f d for all f 2 L1 .X; /C :
X

Evidently, M D L1 .X; / \ L1 .X; / and N D L2 .X; / \ L1 .X; /.

Unbounded weights are difficult to handle. To retain some control, one usually
imposes some of the following conditions:

Definition 8.1.3. A weight  on a C  -algebra A is

• lower semi-continuous if the subset fa 2 AC j .a/  g  A is closed for

every  2 RC ;
• densely defined if M
C is dense in AC (or, equivalently, if N
or M
is dense
in A);
• proper if  is non-zero, densely defined, and lower semi-continuous.

A weight  on a von Neumann algebra M is

• normal if the subset fa 2 M C j .a/  g  M is  -weakly closed for

every  2 RC ;
• semi-finite if M
C is dense in M C (or, equivalently, if N
or M
is dense in
M ) with respect to the  -weak topology;
• n.s.f. if  is normal, semi-finite, and faithful.

8.1.2 Locally compact quantum groups in the setting of

von Neumann algebras
A locally compact quantum group in the setting of von Neumann algebras is sim-
ply a von Neumann bialgebra equipped with a left invariant and a right invariant
n.s.f. weight. The definition involves the cone of positive normal linear function-
als on a von Neumann algebra M , which we denote by MC , and slice maps (see
Section 12.4).

Definition 8.1.4. Let .M; / be a von Neumann bialgebra. A weight  on M is

• left-invariant if ..! x̋ id/..a/// D !.1/.a/ for all a 2 M

C ; ! 2 MC ;

• right-invariant if ..id x̋ !/..a/// D !.1/.a/ for all a 2 M

C ; ! 2
MC .
206 Chapter 8. Locally compact quantum groups

A locally compact quantum group .in the setting of von Neumann algebras/ is a
von Neumann bialgebra .M; / that has a left-invariant n.s.f. weight and a right-
invariant n.s.f. weight.
For a motivation of the invariance conditions, see Section 2.2.1. As in the case
of algebraic quantum groups, uniqueness of the Haar weights can be shown once
the existence is settled; however, the proof is much more involved and depends on
a fair amount of the theory of locally compact quantum groups.
Theorem 8.1.5. For every locally compact quantum group in the setting of von
Neumann algebras, the left-invariant n.s.f. weight and the right-invariant n.s.f.
weight are unique up to positive constants.
Proof. This is stated in [158, Section 1.14]; a proof is given in the setting of C  -
algebras, and that proof carries over to the setting of von Neumann algebras. 
Definition 8.1.6. The invariant n.s.f. weights of a locally compact quantum group
in the setting of von Neumann algebras (which are uniquely determined up to some
constant) are called its left and its right Haar weight, respectively.
Classical examples of locally compact quantum groups are the von Neumann
bialgebras associated to a locally compact group:
Example 8.1.7. For every locally compact group G, the von Neumann bialgebras
L1 .G/ and L.G/ introduced in Example 4.2.4 are locally compact quantum groups.
Let us describe their Haar weights.
The left Haar weight  and the right Haar weight of L1 .G/ are given by
Z Z
.f / D f d ; .f / D f d1 for all f 2 L1 .G/C ;
G G
1
where  and  denote the left and the right Haar measure of G, respectively.
Invariance of  and of follows immediately from the translation invariance of 
and 1 , see also Section 2.2.1.
The von Neumann bialgebra L.G/ has a Haar weight O that is both left- and
right-invariant; it is given by
8
ˆ
ˆ .kf k2 /2 ; there exists f 2 L2 .G; / such that
ˆ
< R
O  .T /.y/ D G f .x/.x 1 y/d .x/
.T T / D
ˆ
ˆ for all  2 Cc .G/; y 2 G;
:̂
1; otherwise,
for every T 2 L.G/C , see [121, Section 7.2.7] or [148, Section VII.3]. A short
calculation shows
O
.L.g// D g.e/ for all g 2 Cc .G/ such that L.g/ 2 L.G/C :
 8.1. The concept of a locally compact quantum group 207

8.1.3 The modular automorphism group of a weight

Every n.s.f. weight on a von Neumann algebra has a modular automorphism group
that measures the deviation of the weight from being a trace. This automorphism
group is reviewed in the following paragraphs. Standard references are, for example,
[21], [76], [121], [144], [147], [150].
The Haar weights  and of a locally compact quantum group .M; / need
not be traces, that is, we can neither expect .a b/ D .ba / for all a; b 2 N

nor .a b/ D .ba / for all a; b 2 N . Recall that for every left integral 0
on an algebraic quantum group .A0 ; 0 /, there exists a modular automorphism 0
such that 0 .a b/ D .b0 .a // for all a; b 2 A0 (Theorem 2.2.17). A deep and
fundamental result of the celebrated Tomita–Takesaki theory says that every n.s.f.
weight on a von Neumann algebra admits a similar modular automorphism group.
This modular automorphism group is essential for the development of a satisfying
theory of locally compact quantum groups.
Let us proceed to the precise definitions and statements. The modular automor-
phism group of an n.s.f. weight is a particular instance of the following concept:
Definition 8.1.8. A one-parameter group of -automorphisms on a C  -algebra A
is a family ˛ D .˛ t / t2R of -automorphisms of A that satisfies ˛s ı ˛ t D ˛sCt for
all s; t 2 R. The one-parameter group ˛ is called
• norm-continuous if for every a 2 A, the map R ! A given by t 7! ˛ t .a/ is
continuous;
• strongly continuous if A is a von Neumann algebra and for every a 2 A,
the map considered above is continuous with respect to the strong operator
topology on A.
Remark 8.1.9. The term “strongly continuous one-parameter group” is frequently
also used for one-parameter groups of automorphisms of general Banach spaces
that are norm-continuous in a similar sense as above.
We shall be interested in analytic extensions of one-parameter groups. These
extensions involve analytic functions that are defined on horizontal strips on the
complex plane of the form

I.z/ WD fy 2 C W j=yj  j=zjg  C; where z 2 C:

Given a C  -algebra/von Neumann algebra A, let us call a function f W I.z/ ! A

norm-regular/strongly regular if it is
i) analytic on the interior of I.z/, that is, for every y0 in the interior of I.z/, the
limit limy!y0 .f .y/  f .y0 //=.y  y0 / exists w.r.t. the norm topology;
ii) norm-bounded on I.z/;
208 Chapter 8. Locally compact quantum groups

iii) norm-continuous/ -strongly- continuous on I.z/.

For the definition of the  -strong- topology, see Section 12.3. From the Identity
Theorem of complex analysis, it follows that two functions f; g W I.z/ ! A that
are norm-regular/strongly regular and coincide on R  I.z/ must coincide on I.z/.
Definition 8.1.10. Let ˛ be a norm-continuous/strongly continuous one-parameter
group of automorphisms on a C  -algebra/von Neumann algebra A. For each z 2
C n R, put

Dom.˛z / WD fx 2 A j there exists f W I.z/ ! A norm-regular/strongly

regular such that f .t / D ˛ t .x/ for all t 2 Rg;

and define ˛z W Dom.˛z / ! A by ˛z .x/ WD f .z/, where f is as above. Note that

the function f is uniquely determined and ˛z is well defined.
T The family .˛z /z2C is called the analytic extension of ˛, and the elements of
z2C Dom.˛z / are called analytic.

Proposition 8.1.11. Let ˛ be a norm-continuous/strongly continuous one-para-

meter group of automorphisms on a C  -algebra/von Neumann algebra A, and let
y; z 2 C.
i) The analytic elements are dense in A with respect to the norm/ -strong-
topology.
ii) Dom.˛z /  A is a subalgebra and ˛z W Dom.˛z / ! A is an algebra homo-
morphism.
iii) Dom.˛z / D Dom.˛zN / and ˛zN .a / D ˛z .a/ for all a 2 Dom.˛z /.
iv) ˛z ı ˛ t D ˛ t ı ˛z D ˛zCt for all t 2 R.
v) Dom.˛y ı ˛z / D Dom.˛yCz / \ Dom.˛z / and ˛y .˛z .a// D ˛yCz .a/ for
all a 2 Dom.˛y ı ˛z /. If y and z lie on the same side of the real axis, then
˛y ı ˛z D ˛yCz .
vi) If y 2 I.z/, then Dom.˛z /  Dom.˛y /.
vii) ˛z is injective, Im ˛z D Dom.˛z /, and ˛z 1 D ˛z .
viii) ˛z is closed with respect to the norm topology/ -strong- topology.
For the definition of closed maps, see Section 8.2.
Proof. Statements ii)–viii) follow without much work from the definition, see the
Preprint [86]. For the proof of statement i), see also [121, Section 8.12]. 
Remark 8.1.12. Every norm-continuous/strongly continuous one-parameter group
˛ on a C  -algebra/von Neumann algebra is uniquely determined by the map ˛i ,
which is called the .analytic/ generator of ˛.
 8.1. The concept of a locally compact quantum group 209

Given the preceding definitions, we can formulate the main result of Tomita–
Takesaki theory:
Theorem 8.1.13. Let  be an n.s.f. weight on a von Neumann algebra M . There
exists a unique strongly continuous one-parameter group of automorphisms  on M
such that
i)  leaves  invariant, that is,  ı  t D  for all t 2 R;

ii) .x  x/ D  i=2 .x/i=2 .x/ for all x 2 Dom.i=2 /.
Furthermore, .ax/ D .xi .a// in the following two situations:
a) a 2 Dom.i / and x 2 M
; in that case, ax and xi .a/ belong to M
;
b) x 2 N
\ N
 ; a 2 N
 \ Dom.i /, and i .a/ 2 N
.
The one-parameter group  is called the modular automorphism group of .
For an illustration, consider the following simple example:
Example 8.1.14. Let  be an n.s.f. weight on the von Neumann algebra Mn .C/,
where n 2 N. Elementary linear algebra shows that there exists a positive definite
matrix ı 2 Mn .C/ such that
.x/ D Tr.xı/ D Tr.ı 1=2 xı 1=2 / D Tr.ıx/ for all x 2 Mn .C/C ;
where Tr W Mn .C/ ! C denotes the usual trace. Note that  extends to a positive
linear map on Mn .C/ by the same formula as above. We claim that the modular
automorphism group  of  is given by
 t .x/ D ı it xı it for all x 2 Mn .C/; t 2 R:

Note that ı it is unitary for each t 2 R, so that  t is a -automorphism.

Let us prove the claim. The weight  is invariant with respect to  because
. t .x// D .ı it xı it / D Tr.ı it xı it ı/ D Tr.xı it ıı it / D Tr.xı/ D .x/
for all x 2 Mn .C/. Evidently, the analytic extension of  is given by z .x/ D
ı iz xı iz for all x 2 Mn .C/ and z 2 C. Therefore, i=2 .x/ D ı 1=2 xı 1=2 and

.i=2 .x/i=2 .x/ / D .ı 1=2 xı 1=2  ı 1=2 x  ı 1=2 /

D Tr.xıx  / D Tr.x  xı/ D .x  x/
for all x 2 Mn .C/. These equations show that  satisfies conditions i) and ii) of
Theorem 8.1.13. Furthermore, .ax/ D .xi .a// for all a; x 2 Mn .C/:
.ax/ D Tr.axı/ D Tr.xıa/ D Tr.xıaı 1 ı/ D Tr.xi .a/ı/ D .xi .a//:
210 Chapter 8. Locally compact quantum groups

8.1.4 Reduced C  -algebraic quantum groups

For the Haar weights of a reduced C  -algebraic quantum group, the existence of a
modular automorphism group has to be assumed.
Definition 8.1.15. A proper weight  on a C  -algebra A is a KMS-weight if there
exists a norm-continuous one-parameter group  on A such that conditions i) and
ii) of Theorem 8.1.13 hold. As in the setting of von Neumann algebras, the one-
parameter group  is called a modular automorphism group for .
References for KMS-weights on C  -algebras are, for example, [90], [91], [121],
[150], [158].
Every KMS-weight on a C  -algebra satisfies the analogue of the second part of
Theorem 8.1.13:
Theorem 8.1.16. Let  be a KMS-weight on a C  -algebra A. If  is faithful,
then the modular automorphism group of  is uniquely determined. Moreover,
.ax/ D .xi .a// in the following two situations:
a) a 2 Dom.i / and x 2 M
; in that case, ax and xi .a/ belong to M
;
b) x 2 N
\ N
 ; a 2 N
 \ Dom.i /, and i .a/ 2 N
.
Given a C  -algebra A, we denote by AC the cone of all positive linear func-
tionals on A.
Definition 8.1.17. A reduced C  -algebraic quantum group is a C  -bialgebra
.A; / that satisfies the following conditions:
i) the following two sets are linearly dense subsets of A:
f.! ˝ id/..a// j ! 2 AC ; a 2 Ag; f.id ˝!/..a// j ! 2 AC ; a 2 AgI

ii) there exists a faithful KMS-weight  on A which is left-invariant in the sense

that ..! ˝ id/..a/// D !.1/.a/ for all ! 2 AC and a 2 M
C ;
iii) there exists a KMS-weight on A which is right-invariant in the sense that
..id ˝!/..a/// D !.1/ .a/ for all ! 2 AC and a 2 MC .

Remarks 8.1.18. i) In conditions ii) and iii), ! ˝ id and id ˝! denote slice maps
as defined in Proposition 12.4.1. To define !.1/, one extends ! to the unitization
(or multiplier algebra) of A, see Corollary 12.1.2, and finds (as for any positive
functional) !.1/ D kwk. Finally, note that .! ˝ id/..a// and .id ˝!/..a//
belong to A by condition i), so that we can apply  or , respectively.
ii) The KMS-condition on the weights in ii) and iii) can be replaced by a condition
that first seems to be weaker, but turns out to be equivalent: the weights  and
need only be approximate KMS-weights [91].
 8.1. The concept of a locally compact quantum group 211

iii) If .A; / is a bisimplifiable C  -bialgebra, then condition i) above holds. This

follows easily from Proposition 12.4.3. Conversely, every reduced C  -algebraic
quantum group is a bisimplifiable C  -bialgebra, as we shall see in Proposition 8.3.2.
iv) For locally compact quantum groups in the setting of von Neumann algebras,
an analogue of condition i) is automatically satisfied, see [93, Proposition 1.4] or
[158, Proposition 1.14.5].
v) The KMS-weight in ii) turns out to be faithful, see Remark 8.3.9.
As in the setting of von Neumann algebras, one can show:
Proposition 8.1.19. For every reduced C  -algebraic quantum group, the faithful
left-invariant KMS-weight and the right-invariant KMS-weight are unique up to
positive constants.
Proof. The proof involves a large amount of the theory of locally compact quantum
groups, see [91, Theorems 7.14, 7.15] or [158, Theorems 1.10.1, 1.10.2]. 
Definition 8.1.20. The invariant KMS-weights of a reduced C  -algebraic quantum
group (which are uniquely determined up to some constant) are called its left and
its right Haar weight, respectively.
Classical examples of reduced C  -algebraic quantum groups are the C  -bialge-
bras associated to a locally compact group:
Example 8.1.21. For every locally compact group G, the C  -bialgebras C0 .G/
and Cr .G/ introduced in Example 4.2.2 are reduced C  -algebraic quantum groups.
The Haar weights on C0 .G/ coincide with the restrictions of the Haar weights on
L1 .G/; trivially, they are traces. The Haar weight on Cr .G/ coincides with the
restriction of the Haar weight on L.G/. Its modular automorphism group can be
described as follows [150, VII, Proposition 3.1]: Denote by ı the modular function
of G (see Section 2.2.3), and consider the function ı it for each t 2 R as a unitary
multiplication operator on L2 .G; /. Then  t .x/ D ı it xı it for all x 2 Cr .G/
and t 2 R.
Example 8.1.22. Every reduced C  -algebraic compact quantum group .A; / is
a reduced C  -algebraic quantum group. To prove this, we only need to show that
the Haar state h of .A; / is a KMS-state. Denote by
• .H; ƒ;
/ the GNS-construction for h;
• .u˛ /˛ a maximal family of pairwise inequivalent irreducible unitary corepre-
sentation matrices of .A; /;
• .A0 ; 0 / the Hopf -algebra of matrix elements of finite-dimensional core-
presentations, that is, A0 D spanfu˛kl j ˛; k; lg;
• .fz /z2C the family of characters on A0 defined in Theorem 3.2.19;
212 Chapter 8. Locally compact quantum groups

• .z0 /z2C the family of automorphisms

P (not necessarily -automorphisms) of
A0 given by z0 .a/ D fiz .a.1/ /a.2/ fiz .a.3/ /; in the notation of Corol-
lary 3.2.20, z0 D iz;iz .
We show that the family . t0 / t2R extends to a norm-continuous one-parameter
group  on A and that  is a modular automorphism group for h. The extension is
implemented by a one-parameter group of unitaries on H which is constructed as
follows. Consider the map

r0 W ƒ.A0 / ! ƒ.A0 /; ƒ.a/ 7! ƒ.i

0
.a//:

Let ˛ be arbitrary. Then H˛ WD spanfƒ.u˛kl / j k; lg has finite dimension and

r0 .H˛ /  H˛ because
X
0
i .u˛kl / D f1 .u˛km /u˛mn f1 .u˛nl / for all k; l
m;n

by equation (5.6). By Corollary 3.2.20, we have for all a; b 2 A0

hƒ.b/jr0 ƒ.a/i D h.b  i

0
.a// D .h ı i
0
/.i0 .b  /a/
D h.i
0
.b/ a/ D hr0 ƒ.b/jƒ.a/i:

Thus r0 jH˛ is self-adjoint.

L
Since H D ˛ H˛ (Proposition 3.2.6, 3.2.9), there exists a unique self-adjoint
operator r on H such that rjH˛ D r0 jH˛ for all ˛. A short calculation shows that
r iz
.a/r iz D
.z0 .a// for all a 2 A0 ; z 2 C. Now we can conclude:
• Since r it is unitary for each t 2 R, we can extend for each t 2 R the
-automorphism  t0 of A0 to a -automorphism  t of A by the formula

. t .a// D r it
.a/r it .
• Since A0 is dense in A and s0 ı  t0 D sCt
0
for all s; t 2 R, we have
s ı  t D sCt for all s; t 2 R.
• Since for each b 2 A0 , the map C ! C; z 7! fz .b/; is holomorphic, the
map R ! A0 ,! A given by t 7!  t0 .a/ D  t .a/ is norm-continuous for
each a 2 A0 . Since each  t has norm 1, it follows that for each a 2 A, the
map R ! A given by t 7!  t .a/ is norm-continuous.
Thus we obtain a norm-continuous one-parameter group  D . t / t2R on A.
Let us show that  satisfies conditions i) and ii) of Theorem 8.1.13. By Corol-
lary 3.2.20, we have h. t .a// D h.a/ for all a 2 A0 and hence for all a 2 A. Let
a 2 Dom.i=2 /. By Proposition 8.1.11,

h.i=2 .a/i=2 .a/ / D h.i=2 .a / i=2 .a // D hr 1=2 ƒ.a /jr 1=2 ƒ.a /i:
 8.2. Additional prerequisites 213

We need to show that this is equal to h.a a/ D hƒ.a/jƒ.a/i. Since the map

r 1=2 and the map ƒ.A L0 / ! ƒ.A0 /; ƒ.a/ 7! ƒ.a /; are compatible with the
decomposition H D ˛ H˛ , we may assume that a 2 A0 . But for a 2 A0 ,
Corollary 3.2.20 implies

h.i=2 .a/i=2 .a/ / D h.i=2

0 0
.a/i=2 .a/ / D h.i=2
0 0
.a/i=2 .a //
D h.i=2
0
.a /i=2
0
.a// D h.a a/:

More generally, every algebraic quantum group gives rise to a reduced C  -

algebraic quantum group:

Example 8.1.23. Let .A0 ; 0 / be an algebraic quantum group with dual .Ay0 ; 
y 0 /.
In Example 7.1.6 and Theorem 7.2.14, we associated to .A0 ; 0 / a well-behaved
y /; /
multiplicative unitary V whose right and left leg .A.V /; / and .A.V y were
y y 0 /, re-
completions of the algebraic quantum group .A0 ; 0 / and its dual .A0 ; 

spectively. Kustermans and Van Daele showed that these C -bialgebras are reduced
C  -algebraic quantum groups [94].

Further examples of locally compact quantum groups are discussed in Sec-

tion 8.4.

8.2 Additional prerequisites

Before we can proceed in the theory of locally compact quantum groups, we need
to review several preliminary concepts and tools related to unbounded linear maps
and weights.

Closed and densely defined operators. We will often encounter linear maps
that are not continuous and not everywhere defined, as, for example, the analytic
extensions of one-parameter groups. Let us fix some related terminology.
Given topological vector spaces E and F , a densely defined linear map from E
to F is a linear map T W Dom.T / ! F whose domain of definition Dom.T / is a
dense subspace of E. We write T W Dom.T /  E ! F in order to indicate the
domain and range of T . Such a map T is called closed if it satisfies the following
equivalent conditions:

i) the graph G.T / WD f.e; T .e// j e 2 Dom.T /g is a closed subspace of E ˚F ;

ii) if .e / is a net in Dom.T / that converges to some element e 2 E and the

net .T .e // converges to some element f 2 F , then e 2 Dom.T / and
T .e/ D f .
214 Chapter 8. Locally compact quantum groups

A subspace E0  Dom.T / is called a core for T if the subset f.e; T .e// j e 2

E0 g  G.T / is dense. Evidently, every densely defined closed map is completely
determined by the restriction to some core.
The composition of densely defined linear maps is defined like the usual com-
position of partially defined maps; it need not be densely defined. Furthermore, the
composition of closed maps need not be closed.

Constructions with weights ([90], [91], [150], [158]). We need to define several
constructions with weights such as formation of tensor products or slice maps and
extension to multiplier algebras. The general idea is that all constructions that be-
have nicely for positive linear functionals on C  -algebras or normal positive linear
functionals on von Neumann algebras can be carried over to lower semi-continuous
or normal weights, respectively, by means of the following approximation result:

Lemma 8.2.1. A weight  on a C  -algebra A is lower semi-continuous if and only

if

.x/ D sup !.x/ for all x 2 AC ; where F

D f! 2 AC j !   on AC g:
!2F

A semi-finite weight  on a von Neumann algebra A is normal if and only if the

condition above holds with AC replaced by AC
 in the definition of F
.

Let  and be proper weights on C  -algebras A and B, respectively, and put

G
WD .0; 1/  F
 F
. The set G
is naturally ordered; therefore it can be used
as the index set of a net.

• The extension of  to the C  -algebra M.A/ is the weight N defined by .x/

N WD
C
N
sup!2F !.x/ for all x 2 M.A/ . Here, !N denotes the unique strictly
continuous extension of a functional ! 2 AC to M.A/, see Corollary 12.1.2.

• The tensor product of  and is the weight  ˝ on A ˝ B defined by

. ˝ /.x/ D supf.! ˝  /.x/ j ! 2 F

;  2 F g

for all x 2 .A ˝ B/C . This weight is proper.

• The slice map  ˝ idB on M.A ˝ B/ is defined as follows. Put

x C WD fx 2 M.A ˝ B/C j s-lim .! ˝ id/.x/ exists in M.B/g

M
˝id
!2G

and
xC ;
. ˝ idB /.x/ WD s-lim .! ˝ id/.x/ for all x 2 M
˝id
!2G
 8.2. Additional prerequisites 215

where “s-lim” denotes the strict limit. Furthermore, let

˚ 
x
˝id WD span M
M x C ; Nx
˝id WD x 2 M.A ˝ B/ j x  x 2 M
xC

˝id
˝id :

x
˝id ! M.B/.
Then  ˝ idB extends to a linear map  ˝ idB W M
Likewise, one can define a slice map idA ˝ .
Similarly, one defines tensor products and slice maps of n.s.f. weights on von Neu-
mann algebras.

GNS-construction for weights ([90], [91], [150], [158]). For a locally compact
quantum group, the Hilbert space associated to the left Haar weight plays a funda-
mental rôle. This space is a particular example of the following construction:
Definition 8.2.2. Let  be a weight on a C  -algebra A. A GNS-construction for  is
a triple .H
; ƒ
;

/ consisting of a Hilbert space H

, a linear map ƒ
W N
! H

with dense image, and a representation

W A ! L.H
/ such that for all a; b 2 N

and c 2 A,

hƒ
.b/jƒ
.a/i D .b  a/ and

.c/ƒ
.b/ D ƒ
.cb/:

It is easy to see that for each weight, there exists a GNS-construction, and that
this construction is unique up to a unitary transformation. In general, the GNS-map
ƒ
of a weight  is unbounded, but one still has some control:
Proposition 8.2.3. i) Let  be a lower semi-continuous weight a C  -algebra with
GNS-construction .H
; ƒ
;

/. Then the map ƒ

is closed with respect to the
norm, and the representation

is non-degenerate.
ii) Let  be a normal and semi-finite weight on a von Neumann algebra M with
GNS-construction .H
; ƒ
;

/. Then the map ƒ

is closed with respect to the
 -weak topology on M and the weak topology on H
, and the representation

is normal and unital.

The Tomita–Takesaki theory [147] provides a detailed description of the GNS-
construction of an n.s.f. weight:
Theorem 8.2.4. Let  be an n.s.f. weight on a von Neumann algebra M with
GNS-construction .H
; ƒ
;

/ and modular automorphism group  .

i) There exists a unique closed conjugate-linear operator T on H
such that
ƒ
.N
\ N
 / is a core for T and T ƒ
.x/ D ƒ
.x  / for all x 2 N
\ N
 .
ii) The operator r WD T  T is strictly positive. There exists a unique antiunitary
J on H
such that T D J r 1=2 . One has J D J  , J 2 D 1, and J r it J D
r it , J r t J D r t for all t 2 R.
216 Chapter 8. Locally compact quantum groups

iii) J

.M /J D

.M /0 .

iv) ƒ
. t .x// D r it ƒ
.x/ for all x 2 N
and t 2 R.

v) Jƒ
.x/ D ƒ
.i=2 .x/ / for all x 2 N
\ Dom.i=2 /.

vi) If x 2 N
and y 2 Dom.i=2 /, then xy 2 N
and

ƒ
.xy/ D J

.i=2 .y// Jƒ
.x/:

For KMS-weights on C  -algebras, the following partial analogue holds:

Theorem 8.2.5. Let  be a KMS-weight on a C  -algebra A with GNS-construc-
tion .H
; ƒ
;

/ and modular automorphism group  . Then conditions i)–ii) and

iv)–vi) of Theorem 8.2.4 hold, and J

.A/J 

.A/0 .
The operators J and r above are called the modular conjugation and the mod-
ular operator of , respectively. Usually, the modular operator is denoted by the
symbol , which we reserve for the comultiplication.

8.3 Main properties

For every locally compact quantum group, one can construct:
• a counit in the form of a densely defined unbounded operator; since it only
plays a minor rôle in the theory, we will not discuss it;

• an antipode in the form of a densely defined unbounded operator and a polar

decomposition of this antipode into a scaling group and a unitary antipode;

• a modular element that relates the right to the left Haar weight;

• a dual locally compact quantum group – this generalizes Pontrjagin duality

to the class of all locally compact quantum groups;

• a manageable multiplicative unitary which plays a fundamental rôle for the

development of the theory.
Furthermore, one can associate to every locally compact quantum group in the set-
ting of von Neumann algebras a reduced C  -algebraic quantum group and vice
versa. These transitions establish a bijective correspondence between the two
classes of bialgebras, see also Section 4.3. We focus on reduced C  -algebraic
quantum groups; the setting of von Neumann algebras is very similar.
All constructions discussed in this section are technical and involved; therefore
we only indicate some main ideas. The interested reader who is willing to spend
 8.3. Main properties 217

some time and energy is referred to the original article [91] and to the thesis [158]
for proofs and details.
Throughout this section, let .A; / be a reduced C  -algebraic quantum group
with left Haar weight  and right Haar weight . For motivation, we shall fre-
quently consider an algebraic quantum group .A0 ; 0 / with left integral 0 and
right integral 0 . Let us stress that the algebraic quantum group and the reduced
C  -algebraic quantum group are not assumed to be related in any way. As before,
we denote the algebraic tensor product by “ˇ”.

8.3.1 The multiplicative unitary

The multiplicative unitary of a locally compact quantum group plays a central rôle
in the theory. Most importantly, it facilitates the transition between the setting of
C  -algebras and the setting of von Neumann algebras, and the construction of the
dual of a locally compact quantum group.
To motivate the construction of the multiplicative unitary, we recall the defi-
nition of the multiplicative unitary WA0 of an algebraic quantum group .A0 ; 0 /
(Example 7.1.6). The underlying Hilbert space of WA0 is the GNS-space for the
left integral on .A0 ; 0 /, and the adjoint .WA0 / is given by

a ˇ b 7! 0 .b/.a ˇ 1/; where a; b 2 A0 : (8.1)

How can this construction be adapted to the reduced C  -algebraic quantum

group .A; /? The underlying Hilbert space should now be the GNS-space H
of
the left Haar weight  of .A; /, and formula (8.1) should be rewritten as

ƒ
.a/ ˝ ƒ
.b/ 7! .ƒ
˝ ƒ
/..b/.a ˝ 1//; where a; b 2 N
:

To make sense of the right-hand side, we need to extend the map

ƒ
ˇ ƒ
W N
ˇ N
! H
ˇ H
 H
˝ H

to the subspace .N

/.N
˝ 1/ of A ˝ A. This can be done as follows. Consider
the weight  ˝  on A ˝ A. Evidently, N
ˇ N
 N
˝
, and one can show that
• the map ƒ
ˇ ƒ
extends uniquely to a closed linear map ƒ
˝ƒ
W N
˝
!
H
˝ H
for which N
ˇ N
is a core;
• .N
/.N
˝ 1/  N
˝
; this follows from the left-invariance of .
Theorem 8.3.1. Let .A; / be a reduced C  -algebraic quantum group with left
Haar weight  and associated GNS-construction .H
; ƒ
;

/. Then there exists

a multiplicative unitary W on H
such that

W  .ƒ
.a/ ˝ ƒ
.b// D .ƒ
˝ ƒ
/..b/.a ˝ 1// for all a; b 2 N
:
218 Chapter 8. Locally compact quantum groups

Proof. The detailed proof is given in [91, Theorem 3.16, Proposition 3.18] and
[158, Theorem 1.3.1]; be prepared for a long reading. Using the left-invariance of
the left Haar weight , it is not difficult to show that the formula above defines an
isometry W  . The hard part of the proof is to show that W  has dense image; this
step involves the right Haar weight. The pentagon equation for W follows easily
from the coassociativity of the comultiplication . 
y /; 
Recall that every multiplicative unitary gives rise to a left leg .A.W y W / and
a right leg .A.W /; W / (see Section 7.2).
Proposition 8.3.2. The representation

defines an isomorphism of the C  -bi-

y /; 
algebras .A; / and .A.W y W /:

y /;

.A/ D span f.id x̋ !/.W / j ! 2 L.H

/ g D A.W
.

/..a// D W  .1 ˝

.a//W D  y W .

.a// for all a 2 A:

Proof. See [91, Equation (4.2), Proposition 3.17] or [158, Proposition 1.3.4]. 
Later, we shall see that W is manageable (Proposition 8.3.10) and that its right
leg .A.W /; W / is a reduced C  -algebraic quantum group again, namely, the
coopposite of the generalized Pontrjagin dual of .A; / (Section 8.3.3).

8.3.2 The antipode and modular properties

At first sight, it may be difficult to guess which construction known from the theory
of algebraic quantum groups can be used to define the antipode of a reduced C  -
algebraic quantum group. However, the antipode S0 and the left integral 0 of an
algebraic quantum group .A0 ; 0 / are related by the equation

S0 ..id ˇ 0 /.0 .b  /.1 ˇ a/// D .id ˇ 0 /..1 ˇ b  /0 .a//

for all a; b 2 A0 (see Lemma 2.2.12).

To adapt this formula to the reduced C  -algebraic quantum group .A; /, we
use the slice map idA ˝ defined on page 214. To apply this slice map to products
of the form .b  /.1 ˝ a/ and .1 ˝ b  /.a/, we need to impose restrictions on b
and a. The following relations imply that it suffices to assume a; b 2 N
:

.N
/  Nx id ˝
; M.A/ ˇ N
 Nx id ˝
; .Nx id ˝
/Nx id ˝
 M
x
id ˝
:

The first of these inclusions can be deduced from the left-invariance of , and the
second and third inclusion are obvious.
Alternatively, we can identify .A; / with the left leg of the multiplicative
unitary W (Theorem 8.3.1) and define an antipode as in Proposition 7.2.16. Both
approaches turn out to be equivalent:
 8.3. Main properties 219

Theorem 8.3.3. Let .A; / be a reduced C  -algebraic quantum group with left
Haar weight , right Haar weight , and multiplicative unitary W . There exists a
unique closed densely defined linear map S W Dom.S /  A ! A that satisfies the
following conditions:
i) spanf.id ˝/..b  /.1 ˝ a// j a; b 2 N
g  A is a core for S and
S..id ˝/..b  /.1 ˝ a/// D .id ˝/..1 ˝ b  /.a// for all a; b 2 N
:

ii) spanf. ˝ id/..b  ˝ 1/.a// j a; b 2 N g  A is a core for S and

S.. ˝ id/..b  ˝ 1/.a/// D . ˝ id/..b  /.a ˝ 1// for all a; b 2 N :

iii) f.id x̋ !/.W / j ! 2 L.H

/ g 

.A/ Š A is a core for S and

S..id x̋ !/.W // D .id x̋ !/.W  / for all ! 2 L.H
/ :

Proof. See [91, Proposition 5.24, Corollary 5.35, Proposition 8.3] or [158, Propo-
sitions 1.6.4, 1.6.17, 1.8.6]. 
The preceding theorem characterizes the antipode, but the actual construction
and the polar decomposition of the antipode are given in the next theorem. The
starting point of this construction is an operator G on H
which, roughly, satisfies
Gƒ
.c/ D ƒ
.S.c  // for suitable c 2 N
. To define G without reference to the
antipode S , we insert the equation of condition 8.3.3 ii) into the desired equation
Gƒ
.c/ D ƒ
.S.c  //: for suitable a; b 2 N , the operator G acts by
ƒ
.. ˝ id/..b  /.a ˝ 1/// 7! ƒ
.. ˝ id/..a /.b ˝ 1///:
To be able to apply the GNS-map ƒ
to the image of the slice map ˝ id on
both sides above, we need to impose restrictions on the elements a and b. If
a; b 2 N
 N , then also a; b 2 N because N is a left ideal, and hence the slice
map ˝ id can be applied in both sides of the equation above (see the discussion
before Theorem 8.3.3). Using a Fubini-type theorem and left-invariance of , one
can furthermore show that in this case, the images . ˝ id/..a /.b ˝ 1// and
. ˝ id/..b  /.a ˝ 1// belong to N
.
Theorem 8.3.4. Let .A; / be a reduced C  -algebraic quantum group with right
Haar weight and left Haar weight , and let .H
; ƒ
;

/ be a GNS-construction
for .
i) There exists a unique closed densely defined conjugate-linear operator G on
H
such that spanfƒ
.. ˝ id/..b  /.a ˝ 1/// j a; b 2 N
 N g  H
is
a core for G and
Gƒ
.. ˝ id/..b  /.a ˝ 1/// D ƒ
.. ˝ id/..a /.b ˝ 1///
for all a; b 2 N
 N . This operator satisfies G 2 D id.
220 Chapter 8. Locally compact quantum groups

ii) The operator N WD G  G is strictly positive. There exists a unique anti-

unitary I on H
such that G D I N 1=2 . Moreover, I D I  , I 2 D 1, and
I NI D N 1 .
iii) There exists a unique norm-continuous one-parameter group  of -auto-
morphisms of A such that

. t .a// D N it

.a/N it for all t 2 R; a 2 A.

iv) There exists a unique -antiautomorphism R of A such that

.R.a// D
I

.a/ I for all a 2 A.

v) The operator S WD Ri=2 satisfies the conditions of Theorem 8.3.3.
Proof. See [91, Propositions 3.22, 5.11, 5.20] or [158, Propositions 1.4.1, 1.4.3,
1.4.14]. 
Let us point out the parallel between the operators N , I and the one-parameter
group  obtained above with the modular operator, the modular conjugation, and
the modular automorphism group of a weight ! in Tomita–Takesaki theory: The
former are constructed from the map G in a similar way like the latter from the map
ƒ! .a/ 7! ƒ! .a /, where a 2 N! \ N! , and its polar decomposition.
Definition 8.3.5. The one-parameter group  and the maps R and S are called the
scaling group, the unitary antipode, and the antipode of .A; /, respectively.
Remark 8.3.6. We shall see in Proposition 8.3.10 that the multiplicative unitary
W of .A; / is manageable. Therefore, an antipode for .A; / and a polar decom-
position can also be obtained from Theorem 7.3.19. Comparing Theorem 7.3.19
with Theorem 8.3.3 and 8.3.4, we see that the antipode and its polar decomposition
constructed above differ from the antipode and the polar decomposition constructed
in Theorem 7.3.19 only in the sign for the parameter of the scaling group. However,
the constructions in this paragraph can not be avoided because they are used in the
proof of manageability of W .
Example 8.3.7. Let us compute the scaling group and unitary antipode of a reduced
C  -algebraic compact quantum group .A; /. We shall use the same notation as in
Example 8.1.22. In particular, .A0 ; 0 / denotes the associated algebraic compact
quantum group and S0 its antipode.
First, we determine G. Evidently, ƒ.A0 /  Dom.G/. By Lemma 2.2.12,
.h ˇ id/..b  /.a ˇ 1// D S0 ..h ˇ id/..b  ˇ 1/.a/// for all a; b 2 A0 ;
and consequently Gƒ.a/ D ƒ.S0 .a // for all a 2 A0 as expected. Moreover,
Remark 3.1.10 iii) implies that GH˛  H˛ .
Let us determine G  . Using the relation h ı S0 D h (Proposition 2.2.6 ii)), we
find
hƒ.a/jGƒ.b/i D h.a S0 .b  // D h.S0 .b  S01 .a ///
D h.b  S01 .a // D hƒ.b/jƒ.S01 .a //i
 8.3. Main properties 221

for all a; b 2 A0 . Since G is conjugate-linear, the equation above implies G  ƒ.a/ D

ƒ.S01 .a // for all a 2 A0 .
Using the relation  ı S0 ı  D S01 (Proposition 1.3.28), we find
Nƒ.a/ D G  Gƒ.a/ D ƒ.S01 .S0 .a / // D ƒ.S02 .a// for all a 2 A0 :
For each z 2 C, define z0 W A0 ! A0 by a 7! fiz  a  fiz . Then S02 D i 0

(Theorem 3.2.19 iv)) and hence Nƒ.a/ D ƒ.i0 .a// for all a 2 A. Since z0 z00 D
0
0 for all z; z 2 C (Corollary 3.2.20),
0
zCz

N it ƒ.a/ D ƒ. t0 .a// for all a 2 A0 ; t 2 R;

and since z0 is an algebra automorphism for all z 2 C,

. t .a// D N it
.a/N it D
. t0 .a// D
.fit  a  fit /
for all a 2 A0 , t 2 R. Moreover, the unitary antipode of .A; / is given by
R.a/ D S.i=2 .a// D S.f1=2  a  f1=2 / for all a 2 A0 :
By now we have associated quite a number of structure maps to the reduced
C  -algebraic quantum group .A; /. Evidently, it is useful to collect as many
relations between these maps as possible. Let us denote by 
and  the modular
automorphism groups of  and , respectively.
Proposition 8.3.8. i)  , R and S commute in the sense that S ı R D R ı S and
 t ı R D R ı  t ;  t ı S D S ı  t for all t 2 R;
ii)  , 
and  commute in a similar sense as above;
iii) R2 D idA and S 2 D i ;
iv) S is injective and S 1 D i=2 ı R;
v) for all x; y 2 Dom.S /, we have xy 2 Dom.S / and S.xy/ D S.y/S.x/;
vi) for all x 2 Dom.S /, we have S.x/ 2 Dom.S / and S.S.x/ / D x;
vii)  ı R D † ı .R ˝ R/ ı , where † denotes the extension of the flip
a ˝ b 7! b ˝ a;
viii) for all t 2 R,
 ı  t
D . t ˝  t
/ ı ;  ı  t D . t ˝ t / ı ;

. t ˝  t / ı  D  ı  t D . t
˝ t / ı I
ix)  ı R is a right Haar weight and ı R a left Haar weight;
x) there exists a scaling constant  > 0 such that

 ı  t D  t ; ı  t D  t ;  ı  t D  t ; ı  t
D  t :
222 Chapter 8. Locally compact quantum groups

Proof. See [91, Propositions 5.22, 5.23, 5.26, 6.8] or [158, Corollary 1.4.18, Propo-
sitions 1.4.20, 1.4.21, Theorem 1.8.1]. 

Remark 8.3.9. By statement ix) and Proposition 8.1.19, the right Haar weight of
a reduced C  -algebraic quantum group is faithful.

Like every locally compact group and every algebraic quantum group, the re-
duced C  -algebraic quantum group .A; / has a modular element ı which relates
the left and the right Haar weight [91, Section 7], [158, Section 1.9]. This modular
element is an unbounded multiplier of the C  -algebra A, more precisely, an affili-
ated element (see Section 8.4.1). Roughly, ı is the Radon–Nikodym derivative of
with respect to , and D .ı 1=2  ı 1=2 /. The following relations hold:

.ı/ D ı ˝ ı; R.ı/ D ı 1 ;

and for all t 2 R; a 2 A,

 t .ı/ D ı;  t
.ı/ D  t ı D  t .ı/;  t .a/ D ı it  t
.a/ı it I

see [91, Proposition 7.12] or [158, Proposition 1.9.11]. For the precise definition
of the expressions above, see Section 8.4.1.

8.3.3 The duality of locally compact quantum groups

To every locally compact quantum group, one can associate a dual locally compact
quantum group. Furthermore, one can show that the bidual – the dual of the dual –
is naturally isomorphic to the initial locally compact quantum group. This is a
far-reaching generalization of Pontrjagin duality which covers all locally compact
groups. The most difficult step in this generalized Pontrjagin duality is the con-
struction of the Haar weights on the dual of a locally compact quantum group. The
bidual can then quite easily be identified via the associated multiplicative unitary.
Let us turn to the details.
As a C  -bialgebra, the dual of the reduced C  -algebraic quantum group .A; /
is constructed out of the multiplicative unitary W defined in Theorem 8.3.1: It is
y /
the pair .A; y WD .A.W /; W /cop , in detail,

Ay WD span f.! x̋ id/.W / j ! 2 L.H

/ g D A.W / (8.2)
and
y a/
. O WD †W .aO ˝ 1/W  † D .Ad† ıW /.a/ y
O for all aO 2 A; (8.3)

where † 2 L.H ˝ H / denotes the flip  ˝  7!  ˝ .

Proposition 8.3.10. The multiplicative unitary W is manageable.

 8.3. Main properties 223

Proof. See [91, Proposition 6.10] or [158, Proposition 1.8.3]. The operators Q
y figuring in Definition 7.3.16 are related to the scaling group  and the scaling
and Q
constant  of .A; /: Q D Q y D P 1=2 , where P is the strictly positive operator on
H defined by P ƒ
.a/ D  1=2 ƒ
. t .a// for all a 2 N
and t 2 R.
it

y /
Corollary 8.3.11. .A; y is a bisimplifiable C  -algebra.

Proof. This follows from Proposition 8.3.10 and Theorem 7.3.18. 

To prove that the C  -bialgebra .A;y /
y is a reduced C  -algebraic quantum group,
we need to construct a left and a right Haar weight on it. It turns out to be easier to
identify the associated GNS-constructions instead of the Haar weights themselves.
These GNS-constructions can be obtained from the densely defined dual pairing
of the C  -algebras A D A.W y / and Ay D A.W / introduced in Proposition 7.2.15
by an approach that is motivated by the Plancherel Theorem for algebraic quantum
groups (Theorem 2.3.11).
Let us first illustrate the approach for an algebraic quantum group .A0 ; 0 / with
left integral 0 . Because of differing conventions, the left integral 0 gives rise to a
right integral O 0 on the dual multiplier Hopf -algebra .Ay0 ;  y 0 /, whereas the left
O
Haar weight  of .A; / will give rise to a left Haar weight  on .A; y /.
y
The Plancherel Theorem 2.3.11 states that
O 0 .0 .  a1 / 0 .  a2 // D 0 .a1 a2/ for all a1 ; a2 2 A0 :

We can rewrite this equation in terms of

• the GNS-map ƒ
0 W A0 ! H
0 of 0 ,
• the GNS-map ƒ O 0 W Ay0 ! H O 0 of O 0 , and
• the natural pairing .  j  / W Ay0  A0 ! C given by .a;
O b/ 7! a.b/
O
as follows. For i D 1; 2, put aO i WD 0 .  ai / and i WD ƒ
0 .ai /. Then

hƒ O 0 .aO 1 /jƒ O 0 .aO 2 /i D O 0 .aO 1 aO 2/ D 0 .a1 a2/ D hƒ

0 .a1 /jƒ
0 .a2 /i D h1 j2 i;

and for i D 1; 2, the vector i can be characterized in terms of aO i by the relation

hƒ
0 .b/ji i D hƒ
0 .b/jƒ
0 .ai /i D 0 .b  ai / D .aO i jb  / for all b 2 A0 :

Thus, the GNS-map ƒ O 0 can be constructed out of the dual pairing Ay0  A0 ! C
and the GNS-map ƒ
0 without knowledge of O 0 : put H O WD H
0 and 0

ƒ O 0 .a/ O  / D hƒ
0 .b/ji
O WD  , .ajb for all b 2 A0 :
224 Chapter 8. Locally compact quantum groups

Let us adapt this approach to the reduced C  -algebraic quantum group .A; /.
y
We define a dual pairing on a dense subspace of AA. Forget that Ay0 denoted an al-
gebraic quantum group before, and put Ay0 WD f.! x̋ id/.W / j ! 2 L.H
/ g  A. y
Consider the pairing
.  j   W Ay0  A ! C; ..! x̋ id/.W /jb WD !.

.b//:

This is just the dual pairing of A0 .W / D Ay0 and A.Wy / D

.A/ Š A defined in
Proposition 7.2.15; in particular, it is well defined. Put
y WD faO 2 Ay0 j there is  2 H
with .ajb
D0 .ƒ/ O   D hƒ
.b/ji for all b 2 N
g;
y W D0 .ƒ/
and consider the map ƒ y ! H
given by

y a/
ƒ. O   D hƒ
.b/ji for all b 2 N
:
O WD  , .ajb

Theorem 8.3.12. i) The map ƒ y extends uniquely to a closed densely defined linear
map ƒy W Dom.ƒ/y  Ay ! H
for which D0 .ƒ/ y is a core.
ii) There exists a unique KMS-weight O on Ay such that N O D Dom.ƒ/

y and
.a y
O  b/ D hƒ.a/j y
ƒ.b/i H for all a; b 2 N
O . Moreover, there exists a unique
representation

O W Ay ! L.H
/ such that .H
; ƒ; y
O / is a GNS-construction

O
for .
y
iii) The weight O is faithful and left-invariant with respect to .
Proof. See [91, Propositions 8.13, 8.14, 8.15] or [158, Proposition 1.11.12, Theo-
rems 1.11.13, 1.11.14]. 
A right Haar weight on .A;y /
y can be constructed by a similar procedure or by
means of the unitary antipode of .A; /. Thus, we arrive at the following important
theorem.
Theorem 8.3.13 ([91, Theorem 8.20], [158, Theorem 1.11.19]). Let .A; / be a
y /
reduced C  -algebraic quantum group. Then the C  -bialgebra .A; y defined by

(8.2) and (8.3) is a reduced C -algebraic quantum group.
y /
Definition 8.3.14. The reduced C  -algebraic quantum group .A; y is called the
reduced dual of .A; /.
The structure maps of .A; / and of .A; y /
y are related by many equations, see
[91, Section 8] or [158, Sections 1.11, 1.13].
The next theorem identifies the reduced dual .A; y y
y /
y of .A;y /.
y This result is
a far-reaching generalization of the classical Pontrjagin duality of locally compact
abelian groups and may be considered as the most important result in the theory of
locally compact quantum groups.
 8.3. Main properties 225

Theorem 8.3.15. Let .A; / be a reduced C  -algebraic quantum group. Then the
y y
y /
C  -bialgebras .A; y and .A; / are isomorphic.
In contrast to Theorem 8.3.13, Theorem 8.3.15 follows quite easily: One need
y y
y /
not construct an explicit isomorphism between the C  -bialgebras .A; y and .A; /,
but only compares the associated multiplicative unitaries.
y associated to the reduced dual
Proposition 8.3.16. The multiplicative unitary W
y /
.A; y coincides with †W  †.
Proof. This result follows without much work from Theorem 8.3.12, see [91, Propo-
sition 8.16] or [158, Proposition 1.11.15]. 
Proof of Theorem 8.3.15. By the previous proposition, the multiplicative unitary
y
associated to .A; y
y /
y is equal to W y
y D †W y  † D W . By Proposition 8.3.2, the
y /
y
C  -bialgebras .A; y and .A; / can be identified with the left legs of W
y y
y and W ,
respectively, and therefore they are isomorphic. 

8.3.4 Passage between the different levels

The theory of locally compact quantum groups in the setting of von Neumann
algebras is very similar to the theory of reduced C  -algebraic quantum groups;
both theories describe the same objects from different points of view. The equiv-
alence of both theories can be considered as a generalization of Weil’s theorem
[190, Appendice] which establishes a bijective correspondence between measurable
groups equipped with translation-invariant measures and locally compact groups.
In the following section, we sketch how this correspondence extends to locally com-
pact quantum groups and how one can pass back and forth between the setting of
C  -algebras and the setting of von Neumann algebras.
Let .A; / be a reduced C  -algebraic quantum group. We shall associate to
.A; / a locally compact quantum group .A; z /
z in the setting of von Neumann alge-
bras. As before, we denote by .H
; ƒ
;

/ the GNS-construction for the left Haar

weight , and by W the multiplicative unitary associated to .A; / (Theorem 8.3.1).
Since W is manageable (Proposition 8.3.10), it is well-behaved (Theorem 7.3.18),
and hence the pair .A;z /
z given by

Az WD

.A/00 D Ayw .W /; z WD 
 y W W x 7! W  .1 ˝ x/W;

is a von Neumann bialgebra (Corollary 7.2.10). The left Haar weight  of .A; /
z /
can be extended to a left Haar weight Q on .A; z as follows. Put F
D f! 2 A j
C
!.a/  .a/ for all a 2 AC g. One shows:
• for every ! 2 F
, there exists a unique !Q 2 L.H
/ such that !.

Q
.a// D
!.a/ for all a 2 A;
226 Chapter 8. Locally compact quantum groups

Q
• the formula .x/ D sup!2F !.x/;
Q x 2 AzC , defines an n.s.f. weight on Az
z
which is left-invariant with respect to .
A similar procedure, applied to the right Haar weight of .A; /, yields a right Haar
z /.
weight on .A; z Thus, we arrive at the following result:

Theorem 8.3.17 ([158, Section 1.14]). The von Neumann bialgebra .A; z /
z is a
locally compact quantum group in the setting of von Neumann algebras.
Conversely, let .M; / be a locally compact quantum group in the setting of von
Neumann algebras. We shall associate to .M; / a reduced C  -algebraic quantum
group .A; /. This can be done by a similar procedure as above:
• identify .M; / with the left leg of a multiplicative unitary W in a similar
way as in Theorem 8.3.1 and Proposition 8.3.2,
• take the C  -bialgebra .A; / corresponding to the left leg of the multiplicative
unitary W , and
• restrict the Haar weights of .M; / to .A; /.
Let us state the main steps precisely. Denote by .H
; ƒ
;

/ the GNS-con-
struction for the left Haar weight  of .M; /.
Theorem 8.3.18. There exists a multiplicative unitary W 2 L.H
˝ H
/ such
that W  .ƒ
.a/ ˝ ƒ
.b// D .ƒ
˝ ƒ
/..b/.a ˝ 1// for all a; b 2 N
. This
multiplicative unitary is manageable.
Proof. See [93, Theorem 1.2] or [158, Theorem 1.14.2]; manageability is proved
similarly as in the setting of C  -algebras. 
Here, the right-hand side of the equation characterizing W  has to be defined
carefully; this can be done as in the setting of C  -algebras, see the discussion before
Theorem 8.3.1. By Theorem 7.3.18, the unitary W is well-behaved, whence the
pair .A; / given by
y /;
A WD span f.! x̋ id/.W / j ! 2 L.H
/ g D A.W
 WD  y W W a 7! W  .1 ˝ a/W;

is a bisimplifiable C  -bialgebra. Since A is a C  -subalgebra of

.M / Š M , we
can restrict the Haar weights of .M; / from

.M /C to AC .
Theorem 8.3.19 ([93, Proposition 1.6], [158, Theorem 1.14.7] ). The restriction
of the left/right Haar weight of .M; / from

.M /C to AC is a faithful KMS-
weight which is left-invariant/right-invariant with respect to . In particular, the
C  -bialgebra .A; / is a reduced C  -algebraic quantum group.
 8.4. Examples of locally compact quantum groups 227

The transition from a locally compact quantum group in the setting of von
Neumann algebras to the associated reduced C  -algebraic quantum group and the
reverse transition preserve the associated multiplicative unitaries. Thus we find:
Theorem 8.3.20. Theorem 8.3.17 and 8.3.19 set up a bijective correspondence .up
to isomorphism/ between all locally compact quantum groups in the setting of von
Neumann algebras and all reduced C  -algebraic quantum groups.
Proof. See [93, Section 1.2] or [158, Section 1.14]. 

8.4 Examples of locally compact quantum groups

Like the general theory of locally compact quantum groups, the construction of
examples has a long history, is highly non-trivial, and often involves special tech-
niques. The development of the general theory and the construction of examples
happened in parallel, often influencing and stimulating each other.
Presently, the following examples are known. First, there exist two general
constructions that produce new quantum groups out of given ones:
• the bicrossed product construction [7], [157], [9], [162], [10], [166], and
• the quantum double and the double crossed product construction [7], [110],
[11].
In a purely algebraic framework, these constructions were introduced by Kac [74]
and Drinfeld [37], respectively, and studied by many authors; see, for example,
[107, Chapters 6, 7]. We shall not discuss these constructions here.
Second, there exist several quantum versions of classical groups, like
• the quantum group E .2/ [197], [198], [196], [180], [4], [5], [117], [118],
[67],
• the quantum az C b group [204], [127], [141], [178], [205], [124],
• the quantum ax C b group [207], [126], [178], [205], [203], [135],
f 1/ [81], [82],
• the quantum group SU.1;
• the quantum group GL2 .C/ [125], [123].
The construction of these examples requires a lot of hard work. Therefore we shall
only give an overview over the first two examples and refer to the original literature
for proofs and further details.
Roughly, the construction of the examples listed above can be described by the
following recipe:
1. Start with a classical group G of matrices, look at the Hopf -algebra of
polynomial functions on G, and describe this -algebra in terms of generators
and relations.
228 Chapter 8. Locally compact quantum groups

2. Deform the relations by some complex number , consider the -algebra A0

generated by the generators and the deformed relations, and try to define a
comultiplication 0 W A0 ! A0 ˇ A0 such that .A0 ; 0 / becomes a Hopf
-algebra.
3. Represent the generators of A0 by (possibly) unbounded closed operators on
a Hilbert space H , or, equivalently, construct a (universal) C  -algebra A with
generators (suitably interpreted) and relations as A0 .
4. Construct a comultiplication  on A that agrees with 0 on the generators.
5. Find a left and a right Haar weight on .A; /.

Sometimes, one can proceed after Steps 1–3 as follows:

40 . Construct a Hopf -algebra .Ay0 ; y 0 / that is dual to .A0 ; 0 / by applying

Steps 1–3 to a dual of the Hopf -algebra of polynomial functions on G (with
the same parameter  and Hilbert space H ).
50 . Using a magic formula that involves special functions of the operators in-
troduced in Steps 3 and 40 , construct a (manageable/modular) multiplicative
unitary W whose right (or left) leg is the desired C  -bialgebra .A; /.
60 . Construct the Haar weights on .A; / from the unitary W , see [205] and [67,
Section 1.4].

The first approach was developed for E .2/ and SUf  .1; 1/ in [197], [4] and [81],
respectively. The second approach is particularly elegant; it was developed for the
quantum az C b- and ax C b-group in [204], [127], [141] and [207], [126], [205],
respectively, and chosen in [67] for E .2/.
Let us add several comments on the individual steps:

• The deformation of the relations in Step 2 can sometimes be motivated by the

analysis of a natural coaction of .A0 ; 0 /, see, for example, the case E .2/
(Section 8.4.2).
• The unbounded operators constructed in Step 3 do (of course) not belong to
the C  -algebra A and are only affiliated with it, see Section 8.4.1.
• The construction of the comultiplication  in Step 4 is often the most dif-
ficult step and requires a detailed analysis of tensor products of unbounded
operators.
• Usually, the relations on the generators of A0 obtained in Step 2 have to be
complemented by additional spectral conditions depending on .
• Steps 4, 5, and 50 often involve advanced usage of special functions, which
itself constitutes a nice but non-trivial piece of mathematics.
 8.4. Examples of locally compact quantum groups 229

8.4.1 C  -algebras generated by unbounded elements

In the following sections, we want to describe C  -algebras of “functions on locally
compact quantum groups that vanish at infinity” in terms of generators, thought of as
“quantum coordinate functions”, and relations. In general, these generators will be
unbounded because the quantum groups are non-compact; in particular, they can not
belong to the C  -algebra itself. Therefore we have to discuss unbounded elements
that are affiliated with a C  -algebra and clarify in which sense such elements can
generate a C  -algebra.

Affiliated elements of C  -algebras. The affiliation relation was introduced by

Baaj [3] and Woronowicz [197]. We shall follow the treatment in [197] but omit
most of the proofs.
Let us start with an informal motivation. Briefly, an element T should be
affiliated with a C  -algebra A if bounded continuous functions of T belong to
M.A/. In particular, we expect z.T / 2 M.A/ for the function
z W C ! C;  7! .1 C jj2 /1=2 :
Since 1  jz./j2 D .1 C jj2 /1 for all  2 C, we should have
z.T / D T .1  jz.T /j2 /1=2 :
Definition 8.4.1. Let A be a C  -algebra and T W Dom.T /  A ! A a densely
defined linear map. We say that T is affiliated with A and write T ˜ A if there exists
a zT 2 M.A/ such that kzT k  1 and for all x; y 2 A,
x 2 Dom.T / and y D T x
,
there exists a 2 A with x D .1  zT zT /1=2 a and y D zT a:
In this case, the element zT , which is uniquely determined by T , is called the
z-transform of T . The set of all affiliated elements of A is denoted by A˜ .
Remarks 8.4.2. i) Beware that the sum and the product of two affiliated elements
of a C  -algebra need not be densely defined.
ii) There exists the notion of an affiliated element of a von Neumann algebra
which differs from the notion introduced above. If M is a von Neumann algebra
and T is an affiliated element of M in the C  -algebraic sense, then T 2 M (see
Proposition 8.4.3 iii)).
Proposition 8.4.3. Let A be a C  -algebra and T ˜ A.
i) T is a closed linear map, Dom.T /  A is a right ideal, and T .ab/ D T .a/b
for all a 2 Dom.T / and b 2 A.
230 Chapter 8. Locally compact quantum groups

ii) If kT k < 1, then T 2 M.A/.

iii) If A is unital, then T 2 A.

Proof. Statement i) follows easily from the definition; for ii) and iii), see [197,
Proposition 1.3]. 

Example 8.4.4. Let X be a locally compact space. It is easy to see that for each
g 2 C.X/, pointwise multiplication of functions by g defines an affiliated element
of the C  -algebra C0 .X /, and every affiliated element of the C  -algebra C0 .X / is
obtained this way.
Non-degenerate -homomorphisms can be extended to affiliated elements. The
precise formulation of this result involves the concept of a core (see Section 8.2).
Proposition 8.4.5 ([197, Theorem 1.2]). Let A; B be C  -algebras,  W A ! M.B/
a non-degenerate -homomorphism, and T ˜ A.
i) There exists a .T / ˜ B such that .Dom.T //B is a core of .T / and
.T /..a/b/ D .T a/b for all a 2 Dom.T / and b 2 B.
ii) The z-transforms of T and .T / satisfy z
.T / D .zT /.
iii) If C is a C  -algebra and W B ! M.C / is a non-degenerate -homo-
morphism, then . ı /.T / D ..T //.
Each affiliated element has an adjoint:
Proposition 8.4.6 ([197, Theorem 1.4]). Let A be a C  -algebra and T ˜ A.
i) There exists a unique T  ˜ A such that for all a; b 2 A,

a 2 Dom.T  / and b D T  a , a .T x/ D b  x for all x 2 Dom.T /:

ii) The z-transforms of T and T  satisfy z.T  / D .zT / .

iii) If B is a C  -algebra and  W A ! M.B/ is a non-degenerate -homo-
morphism, then .T  / D .T / .
We shall be interested primarily in affiliated elements that are normal in the
following sense:
Lemma 8.4.7 ([197, pages 406, 407]). Let A be a C  -algebra and T ˜ A. Then zT
is normal if and only if Dom.T / D Dom.T  / and .T a/ .T a/ D .T  a/ .T  a/
for all a 2 Dom.T /.
Definition 8.4.8. Let A be a C  -algebra and T ˜ A. We call T normal if zT is
normal.
 8.4. Examples of locally compact quantum groups 231

Multiplication by the identify function idC 2 C.C/ defines a normal affiliated

element m of C0 .C/ (see Example 8.4.4) which is universal in the following sense:
Proposition 8.4.9 ([197, Theorem 1.5]). Let A be a C  -algebra and T ˜ A. Then
there exists a unique non-degenerate -homomorphism T W C0 .C/ ! M.A/ such
that T .m/ D T .
Definition 8.4.10. Let A be a C  -algebra, T ˜ A normal, and T W C0 .C/ ! M.A/
as above. Then the spectrum of T is the set
 .T / WD f 2 C j f ./ D 0 for all f 2 ker T g  C:
Proposition 8.4.11 ([197, Theorem 1.6]). Let A be a C  -algebra, T ˜ A normal,
and define mT ˜ C0 . .T // by .mT f /./ D f ./. Then there exists a unique
injective non-degenerate -homomorphism T W C0 . .T // ! M.A/ such that
T .mT / D T .

Using the -homomorphism T, we can define the functional calculus for T :

Notation 8.4.12. Let A be a C  -algebra, T ˜ A normal, and T as above. For
each f 2 Cb . .T //, we define f .T / 2 M.A/ by f .T / WD T .f /.
To describe the action of the comultiplication on the generators of our quantum
groups, we need to form tensor products of affiliated elements:
Proposition 8.4.13 ([206, Theorem 6.1]). Let A; B be C  -algebras and S ˜ A,
T ˜ B. Then there exists a unique S ˝ T ˜ A ˝ B such that the algebraic tensor
product Dom.S / ˇ Dom.T /  A ˝ B is a core for S ˝ T and .S ˝ T /.a ˝ b/ D
Sa ˝ T b for all a 2 Dom.S /, b 2 Dom.T /. Moreover, .S ˝ T / D S  ˝ T  .
For later use, we note the following simple result:
Proposition 8.4.14 ([197, Proposition 0.2]). Let A be a C  -algebra and a 2 M.A/.
If aA and a A are dense in A, then there exists a unique unitary u 2 M.A/ such
that a D ujaj.

C  -algebras generated by affiliated elements. Given a C  -algebra A and ele-

ments T1 ; : : : ; Tn 2 A, it is clear what we mean when we say that A is generated
by T1 ; : : : ; Tn . When T1 ; : : : ; Tn are no longer elements of the C  -algebra A but
only affiliated with it, the interpretation of this statement is no longer obvious.
Woronowicz introduced the following definition [200, Definition 3.1]:
Definition 8.4.15. Let A be a C  -algebra and T1 ; : : : ; Tn ˜ A. We say that A is
generated by T1 ; : : : ; Tn if for each Hilbert space H , each non-degenerate C  -al-
gebra B  L.H /, and each non-degenerate -homomorphism
W A ! L.H /,
the relation
.T1 /; : : : ;
.Tn / ˜ B implies
.A/B D B, that is,
restricts to a
non-degenerate -homomorphism A ! M.B/.
232 Chapter 8. Locally compact quantum groups

Remark 8.4.16. Given a C  -algebra A and elements T1 ; : : : ; Tn ˜ A, it is not clear

whether there exists a C  -subalgebra B  M.A/ such that T1 ; : : : ; Tn ˜ B and
such that B is generated by T1 ; : : : ; Tn .
The following result illustrates the concept introduced above:
Proposition 8.4.17. Let A, B be C  -algebras,  W B ,! M.A/ a non-degenerate
embedding, and S1 ; : : : ; Sn ˜ B. If .S1 /; : : : ; .Sn / generate A, then .B/ D A.
Proof. Let
be a faithful non-degenerate representation of A on a Hilbert space H .
Then
..S1 //; : : : ;
..Sn // ˜
..B// and hence
.A/
..B// D
..B//.
On the other hand, A.B/ D A. Since
is faithful, the claim follows. 
If one wants to find out whether a C  -algebra is generated by certain affiliated
elements, the following criterion is useful:
Proposition 8.4.18 ([200, Theorem 3.3]). Let A be a C  -algebra, T1 ; : : : ; Tn ˜ A,
and assume that
i) T1 ; : : : ; Tn separate the representations of A: if
1 and
2 are non-degen-
erate representations of A on a Hilbert space H , then
1 D
2 if and only if

1 .Tj / D
2 .Tj / for j D 1; : : : ; n;
ii) there exist r1 ; : : : ; rk 2 f.1 C Tj Tj /1 ; .1 C Tj Tj /1 j j D 1; : : : ; ng 
M.A/ such that r1 : : : rk 2 A.
Then A is generated by T1 ; : : : ; Tn .
The following examples illustrate this criterion in simple cases:
Examples 8.4.19. i) Let A be a unital C  -algebra and T1 ; : : : ; Tn 2 A. It is easy
to see that if the space of all -polynomials in 1 and T1 ; : : : ; Tn is dense in A, then
the conditions of Proposition 8.4.18 hold and A is generated by T1 ; : : : ; Tn in the
sense of Definition 8.4.15.
Conversely, if A is a C  -algebra generated by elements T1 ; : : : ; Tn ˜ A and
kT1 k; : : : ; kTn k < 1, then A is unital, T1 ; : : : ; Tn 2 A, and the space of all
-polynomials in 1 and T1 ; : : : ; Tn is dense in A. Indeed, in that case, T1 ; : : : ; Tn 2
M.A/ (Proposition 8.4.3 ii)), and if B  M.A/ denotes the closure of the space of
all -polynomials in 1 and T1 ; : : : ; Tn , then B D A by Proposition 8.4.17.
ii) Let X be a locally compact space and T1 ; : : : ; Tn 2 C.X /. Then the
C  -algebra C0 .X / is generated
P by T1 ; : : : ; Tn if and only if T1 ; : : : ; Tn separate the
points of X and limx!1 niD1 jTi .x/j2 D 1. The “if” part follows easily from
Proposition 8.4.18; for the “only if” part, see [200, Example 2 after Theorem 3.3].
Further results and applications related to the presentation of C  -algebras in
terms of generators and relations can be found in [200].
 8.4. Examples of locally compact quantum groups 233

8.4.2 The quantum groups E .2/ and Ey .2/

The quantum group E .2/ was constructed by Woronowicz via the first approach
outlined in the introduction to this section: He starts from a double cover of the
group of motions of the Euclidean plane, deforms the associated Hopf -algebra of
polynomials functions, and constructs a C  -bialgebra E .2/ with similar genera-
tors and relations like that Hopf -algebra. These relations are complemented by
additional spectral conditions on the generators which ensure that the comultipli-
cation is well defined [197].
Woronowicz also studied the corepresentation theory of the C  -bialgebra E .2/,
constructed the Pontrjagin dual Ey .2/ as a C  -bialgebra [196], and studied its co-
representation theory in a joint work with Van Daele [180].
Another fundamental contribution was made by Baaj, who found the Haar
weights on the C  -bialgebras E .2/ and Ey .2/ [4], [5]; see also [118]. In partic-
ular, his results imply that E .2/ and Ey .2/ are locally compact quantum groups.
Moreover, Baaj studied the multiplicative unitary associated to E .2/ and showed
that this unitary is not regular but only semi-regular.
A comprehensive description of the locally compact quantum groups E .2/ and
Ey .2/ was finally given by Jacobs in his PhD thesis [67].1
Let us add that the quantum groups E .2/ and SU .2/ are related via a con-
traction procedure studied in [117], [199].
Throughout this subsection, the parameter  is an element of .0; 1/.

The classical group, its Hopf -algebra, and the deformed Hopf -algebra.
Consider the group E.2/  GL2 .C/ consisting of all matrices of the form
 
v n
g.v;n/ D ; where v 2 T ; n 2 C:
0 vN
E.2/ acts on C via g.v;n/ z WD v 2 z C vn, and fg.1;0/ ; g.1;0/ g is the kernel of this
action. Thus, E.2/ is the unique connected double cover of the group of rotations
and dilations of the Euclidean plane.
The -subalgebra A0  C.E.2// generated by the coordinate functions
g.v;n/ 7! v and g.v;n/ 7! n is isomorphic to the universal unital commutative
-algebra generated by a unitary v and an element n. It carries the structure of a
Hopf -algebra; the structure maps can be read off from equation (8.5) below for
 D 1.
Denote by A0; the universal unital -algebra generated by elements v; n such
that
v is unitary; n is normal; vn D nv: (8.4)
1
In [67], the generators v, n of E .2/ are denoted by c, d .
234 Chapter 8. Locally compact quantum groups

Then A0; can be equipped with the structure of a Hopf -algebra, where

0 .v/ D v ˇ v; 0 .n/ D v ˇ n C n ˇ v  ;

0 .v/ D 1; 0 .n/ D 0; (8.5)
   
S0 .v/ D v ; S0 .v / D v; S0 .n/ D n=; S0 .n / D n =:

The deformation of A0 into A0; can be related to a natural deformation of the

-algebra CŒz; z   into a “non-commutative quantum plane” C Œz; z   as follows.
For a Hopf -algebra .B0 ; B0 /, the analogue of a group action on a -algebra C
is a left coaction on C , which is a -homomorphism ı W C ! B0 ˇ C that satisfies
.id ˇ ı/ ı ı D .B0 ˇ id/ ı ı (see also Section 9.2). The action of E.2/ on C
induces a (unital) right coaction of A0 on CŒz; z   via

z 7! v 2 ˇ z C vn ˇ 1: (8.6)

Denote by C Œz; z   the universal unital -algebra generated by an element z subject

to the relation z  z D 2 zz  . One can easily check that if B0 is a -algebra with a
unitary v and a normal element n, then formula (8.6) defines a -homomorphism
B0 ! B0 ˇ C Œz; z   if and only if n nv D 2 vn n and nv 2 D 2 v 2 n. In
particular, formula (8.6) defines a right coaction of .A0; ; 0 / on C Œz; z  .

Commutation relations, spectral conditions, and the C  -bialgebra E .2/. The

following discussion involves actions of groups on C  -algebras and associated
crossed products, which are reviewed in Section 9.1.
Assume that H is a Hilbert space and that v; n ˜ L.H / satisfy the relations in
(8.4). Then

vf .n/v  D f .vnv  / D f .n/ for all f 2 C0 .C/; (8.7)

whence .n/ D  .n/. Given a subset Y  C satisfying Y D Y , denote by

˛ W Z ! Aut.C0 .Y // the action given by ˛k .f / D f .k  /. Equation (8.7) and the
universal property of the crossed product imply that there exists a -homomorphism

W C0 . .n// Ì˛ Z ! L.H /; f Ì Uk 7! f .n/v k :

These considerations show that a universal C  -algebra A with elements v; n ˜ A

satisfying (8.4) can be constructed as follows. Denote by m ˜ C0 .C/ the multipli-
cation operator given by .mf /.z/ D zf .z/, and put

A WD C0 .C/ Ì˛ Z; v WD 1 Ì U1 2 M.A/; n WD m Ì 1 ˜ AI

here, .  / Ì 1 W C0 .C/ ,! A; f 7! f Ì 1, denotes the natural inclusion.

 8.4. Examples of locally compact quantum groups 235

Unfortunately, A can not be equipped with a comultiplication  such that

.v/ D v ˝ v and .n/ D v ˝ n C n ˝ v  :

This fundamental fact was discovered by Woronowicz [197]. The precise statement
of his result, which is given below, involves irreducible representations of A. Stan-
dard arguments show that each irreducible representation
of A on a Hilbert space
H has one of the following two forms:
i) .
.n// D 0, dim H D 1, and
.v/ D z for some z 2 T ; or
ii) .
.n// D tZ for a non-zero t 2 C, and there exists an orthonormal basis
.ek /k such that
.n/ek D tk ek and
.v/ek D ekC1 for all k 2 N.
Theorem 8.4.20 ([197, Theorem 3.1]). Let
1 and
2 be infinite-dimensional ir-
reducible representations of A. Put N WD
1 .v/ ˝
2 .n/ C
1 .n/ ˝
2 .v/ and
†i WD .
i .n//T , where i D 1; 2.
i) If †1 ¤ †2 , then N is closed but not normal and has no normal extension.
ii) If †1 D †2 , then N is closeable, its closure Nz is normal, and  .Nz / D †1 .
The theorem above implies that the desired comultiplication can only be defined
if we supplement the defining relations (8.4) by a restriction on the spectrum  .n/,
which amounts to replacing A by the quotient  .n/ Ì˛ Z. The usual choice is
x ;
 .n/   where  WD fk z j k 2 Z; z 2 T g  C: (8.8)

x / the multiplication operator .mf /.z/ D zf .z/. Then similar

Denote by m ˜ C0 .
arguments as above show that
x / Ì˛ Z; v WD 1 Ì U1 2 M.E .2//; n WD m Ì 1 ˜ E .2/
E .2/ WD C0 .

is the universal example of a C  -algebra with affiliated elements v; n satisfying

the algebraic relations (8.4) and the spectral condition (8.8). Moreover, E .2/
is generated by v; n in the sense of Definition 8.4.15, as can easily be seen from
Proposition 8.4.18. Note that f Ì Uk D f .n/v k for all f 2 C0 . x / and k 2 Z.

Theorem 8.4.21 ([197, Theorems 3.3, 3.4]). There exists a unique non-degenerate
-homomorphism  W E .2/ ! M.E .2/ ˝ E .2// such that .v/ D v ˝ v and
.n/ D v ˝ n C n ˝ v  . Moreover, .E .2/; / is a C  -bialgebra.

The Haar weight and the antipode of E .2/. A left- and right-invariant Haar
weight on the C  -bialgebra .E .2/; / was found by Baaj [4], [5]; we briefly
outline his construction.
236 Chapter 8. Locally compact quantum groups

There exists a positive contraction

x / Ì˛ Z ! C0 .
W E .2/ D C0 . x /; f Ì Uk 7! f ık;0 :

Indeed, if we put C WD C0 .x / and denote by .ek /k the standard basis of l 2 .Z/
and by j0 2 LC .C; l .Z/ ˝ C / the map c 7! e0 ˝ c, then coincides with the
2

map C Ì˛ Z D C Ì˛;r Z  LC .l 2 .Z/ ˝ C / ! LC .C /, T 7! j0 Tj0 .

x / by
Define a weight  on C0 .

X Z 1
.f / WD 2n f .n e 2 it /dt x /; f 0:
for f 2 C0 .
n2Z 0

Theorem 8.4.22. The composition  WD  ı is a left and a right Haar weight on

.E .2/; /, and .E .2/; / is a unimodular locally compact quantum group. The
modular automorphism group  of  is uniquely determined by  t .v/ D 2it v
and  t .n/ D n for all t 2 R.

Proof. The assertions concerning  and . t / t are proved in [5, Proposition 4.1,
Théorème 4.2], see also [67, Section 2.7]. The fact that .E .2/; / is a locally
compact quantum group is proved in [67, Sections 2.4 and 2.5]. 
P
A short calculation shows that for each x D k fk .n/v k 2 E .2/,

X Z 1
.x  x/ D 2n jfk .nk e 2 it /j2 dt:
0
k;n

Next, we describe the unitary antipode, the scaling group, and the antipode
of .E .2/; /. For all k; n 2 Z, put Sk WD fz 2  x j jzj D k g and define
hk;n W Sk ! C by z 7! z . Put
n

x / j f is constant on every Sk g;
D1 WD ff 2 C0 .
x / j there exists n 2 Z such that f jS D hk;n
D2 WD ff 2 C0 . k

or f jSk D 0 for all k 2 Zg:

Theorem 8.4.23. i) The unitary antipode R of .E .2/; / is uniquely determined

by R.v/ D v  and R.n/ D n.
ii) The scaling group  of .E .2/; / is uniquely determined by  t .v/ D v and
 t .n/ D 2it n for all t 2 R.
iii) The -subalgebra D  E .2/ generated by the set D1 [ D2 is a core for S .

Proof. See [67, Section 2.6]. 

 8.4. Examples of locally compact quantum groups 237
P
Short calculations show that for each x D k fk .n/v k 2 E .2/ and t 2 R,
X X
R.x/ D fk .k n/v k ;  t .x/ D fk .2it n/v k ;
k k

and if x 2 D, then
X
S.x/ D fk .k1 n/v k :
k

In particular, the antipode S is unbounded, S 2 ¤ id, and S ı  ¤  ı S [67,

Proposition 2.6.32].

The classical dual group, its Hopf -algebra, and the deformed Hopf -algebra.
The dual of the locally compact quantum group .E .2/; / turns out to be a defor-
mation of the group E.2/y  GL2 .C/ that consists of all matrices of the form
 1 
a 0
gO .a;b/ D ; where a 2 .0; 1/; b 2 C:
b a

y
The action of E.2/ on C given by gO .a;b/ z WD a2 z C ba identifies E.2/ y with the
group generated by all translations and dilations of the Euclidian plane.
y ˙  GL2 .C/ the group consisting of all matrices gO .a;b/ as
Denote by E.2/
above, but with a 2 R n f0g arbitrary. The -subalgebra Ay0  C.E.2/ y ˙ / generated
by the coordinate functions gO .a;b/ 7! a and gO .a;b/ 7! b is isomorphic to the universal
unital commutative -algebra generated by a self-adjoint invertible element a and
an element b, and carries the structure of a Hopf -algebra, where

y 0 .a/ D a ˇ a; 
 y 0 .b/ D b ˇ a1 C a ˇ b;

O0 .a/ D 1; O0 .b/ D 0; (8.9)
Sy0 .a/ D a 1
; Sy0 .a / D .a /1 ; Sy0 .b/ D b; Sy0 .b  / D b  :


Denote by Ay0; the universal unital -algebra generated by elements a and b

satisfying

a is self-adjoint and invertible; b is normal; ab D ba: (8.10)

This is a Hopf -algebra with respect to the comultiplication and counit given in
(8.9); the antipode is given by

Sy0 .a/ D a1 ; Sy0 .a / D .a /1 ; Sy0 .b/ D b=; Sy0 .b  / D b  =:
238 Chapter 8. Locally compact quantum groups

There exists a dual pairing .  j  / between the Hopf -algebras .A0; ; 0 / and
.Ay0; ; 
y 0 /, given by

.vja/ D 1=2 ; .nja/ D 0; .vjb/ D 0; .njb/ D 1:

Commutation relations, spectral conditions, and the C  -bialgebra Ey .2/.

When we shift from Hopf -algebras to C  -algebras, it is appropriate to replace
(8.10) by
a > 0 is self-adjoint; b is normal;
(8.11)
a; jbj strongly commute; ab D baI
(recall that two self-adjoint operators are said to commute strongly if all of their
spectral projections commute). Moreover, one imposes a spectral condition on the
joint spectrum  .a; jbj/:
x ;
.a; jbj/   where  WD f.s ; r / j s 2 12 Z; r 2 12 Z; r C s 2 Zg:
(8.12)
Let us note that in [180] and [196], the generator a is replaced by the element
1
N D 2 log a, so that a D  2 N .
Assume that H is a Hilbert space and that a; b ˜ L.H / satisfy (8.11), (8.12).
Let b D ujbj be the polar decomposition, and put H0 WD uH D .ker b/? . Then
ug.a; jbj/u D g.uau ; jbj/ D g.1 a; jbj/ in L.H0 / for each g 2 C0 . /:
x  / given by ˇk .g/.x; y/ D g.k x; y/, and
Denote by ˇ the action of Z on C0 .
observe that ˇ preserves C0 . /  C0 .x  /. Put

Ey .2/ WD C0 . / Ìˇ Z C C0 .

x  / Ìˇ 1  C0 .
x  / Ìˇ Z:

The equation above and the universal property of the crossed product imply that
there exists a -homomorphism
C0 . / Ìˇ Z ! L.H0 /; g Ì Uk 7! g.a; jbj/uk ;

which naturally extends to a -homomorphism Ey2 ./ ! L.H /.

Denote by mx ; my ˜ C0 . x  / the multiplication operators given by

.mx g/.x; y/ WD xg.x; y/; .my g/.x; y/ WD yg.x; y/:

The preceding arguments show that
Ey .2/ as above; a WD mx Ì 1 ˜ Ey .2/; b WD .1 Ì U1 /.my Ì 1/ ˜ Ey .2/
is the universal example of a C  -algebra with affiliated elements a; b satisfying the
algebraic relations (8.11) and the spectral condition (8.12). Proposition 8.4.18 can
be used to show that Ey .2/ is generated by a; b in the sense of Definition 8.4.15.
 8.4. Examples of locally compact quantum groups 239

Theorem 8.4.24. There exists a unique non-degenerate -homomorphism

y W Ey .2/ ! M.Ey .2/ ˝ Ey .2// such that .a/
 y D a ˝ a and .b/y D b˝
1 y y 
a C a ˝ b. .E .2/; / is a C -bialgebra and as such isomorphic to the coop-
posite of the dual of the locally compact quantum group .E .2/; /. In particular,
y is a locally compact quantum group.
.Ey .2/; /

Proof. The existence of the comultiplication follows from results in [198]. The fact
that .Ey .2/; /
y cop is the Pontrjagin dual of .E .2/; / was first proved (not in the
setting of locally compact quantum groups) in [196], see also [180]. For the proof
of the last assertions, see [67, Sections 2.4, 2.5, 2.8]. 

The Haar weights and the antipode of Ey .2/. The Haar weights of .Ey .2/; / y
were first constructed by Baaj [4], [5]. They can be described as follows. Similarly
as for E .2/, there exists a positive contraction

O W Ey .2/ D C0 .
x  / Ìˇ Z ! C0 .
x  /; g Ì Uk 7! gık;0 :

x  / by
Define weights 
O and  O on C0 .
X X

O .g/ WD rCs g.s ; r /;  O .g/ WD rs g.s ; r /:
.s ;r /2 .s ;r /2

Theorem 8.4.25. i) The left Haar weight O and the right Haar weight O on
.Ey .2/; /
y are .up to a positive factor/ given by O D  O ı O and O D  O ı .

O

O O
ii) The modular automorphism groups  and  of O and O , respectively, are
O O O O
uniquely determined by  t
.a/ D a,  t
.b/ D 2it b and  t .a/ D a,  t .b/ D
2it b for all t 2 R.
y is given by ıO D a4 .
iii) The modular element of .Ey .2/; /

Proof. See [5, Proposition 4.20, Théorème 4.21] or [67, Section 2.7]. 
P
A short calculation shows that for each x D k gk .a; jbj/uk ,
X X
O  x/ D
.x 2.rsCk/ jgk .s ; r /j2 ;
k2Z .s ;r /2
X X
O .x  x/ D 2.rCsk/ jgk .s ; r /j2 :
k2Z .s ;r /2 

The antipode of .Ey .2/; /

y can be described as follows:
240 Chapter 8. Locally compact quantum groups

Theorem 8.4.26. i) The unitary antipode R y of .Ey .2/; /

y is uniquely determined
1
by R.a/ D a and R.b/ D b.
y is uniquely determined by Ot .a/ D a and
ii) The scaling group O of .Ey .2/; /
Ot .b/ D 2it b for all t 2 R.
iii) The -algebra D y  Ey .2/ generated by the set fg.a; jbj/uk ; h.a; jbj/ j
g 2 C0 . /; k 2 N n f0g; h 2 C0 . x  /g is a core for Sy.

Proof. See [67, Section 2.6]. 

P
Short calculations show that for each x D k gk .a; jbj/uk 2 Ey .2/, t 2 R,
X X
y
R.x/ D gk .k a1 ; jbj/.u/k ; Ot .x/ D 2itk gk .a; jbj/uk ;
k k

y then
and if x 2 D,
X
y
S.x/ D ./k gk .k a1 ; jbj/uk :
k

In particular, the antipode Sy is unbounded, Sy2 ¤ id, and Sy ı  ¤  ı Sy [67,

Proposition 2.6.58].

The multiplicative unitary of E .2/ and Ey .2/. We close this subsection with
some brief remarks on the multiplicative unitary of .E .2/; /. Denote by H
the
GNS-space for the Haar weight  on .E .2/; /, and by W 2 L.H
˝ H
/ the
multiplicative unitary associated to .E .2/; / (see Theorem 8.3.1).

Theorem 8.4.27. W is semi-regular but not regular.

Proof. See [5, Proposition 4.7] or [67, Proposition 2.8.23]. 

Recall that we can identify E .2/ D A.W y / and Ey .2/ D A.W / with C  -
subalgebras of L.H
/ and consider W as an element of M.E .2/ ˝ Ey .2// 
L.H
˝ H
/ (see Section 8.3 and Theorem 7.3.18). As such, W can elegantly be
described in terms of the generators v; n ˜ E .2/ and a; b ˜ Ey .2/. This description
involves the quantum exponential function F W  x ! C, defined by
‚Y 1C
1 2k
zN
; x n f2k j k 2 Ng;
z2
F .z/ D 1C 2k z (8.13)
kD0
2k
1; z 2 f j k 2 Ng;
 8.4. Examples of locally compact quantum groups 241

and the bicharacter  W Z=2  T ! T , defined by

 .k=2 ; z/ WD z k for all k 2 Z; z 2 T : (8.14)

Note that the product in (8.13) converges since 0 <  < 1. These functions were
introduced by Woronowicz; they play a central rôle in the study of the quantum
groups E .2/ and Ey .2/ and of the associated multiplicative unitary. It is easy to
see that

• jF .z/j D 1 and F .z/

N D F .z/ for all z 2  , and

• F is continuous.

To explain the terminology “quantum exponentional function”, we cite the following

result; for further details, see [198].

Theorem 8.4.28 ([198]). Let H be a Hilbert space and let R and S be normal
operators on H with polar decompositions R D U jRj, S D V jSj. Assume that
x ,
i) .R/;  .S /  
ii) V and U commute,
iii) jRj and jSj strongly commute, that is, all of their spectral projections com-
mute,
iv) U  jSjU D jS j and V jRjV  D jRj.

Then R C S admits a normal extension R u S, and F .R u S / D F .R/F .S /.

The multiplicative unitary W 2 M.E .2/ ˝ Ey .2// can be described in terms
of the functions F ,  and the operators vn ˝ ab; v ˝ 1; 1 ˝ a ˜ E .2/ ˝ Ey .2/
as follows:
x .
Lemma 8.4.29. vn ˝ ab is normal and  .vn ˝ ab/  

Theorem 8.4.30. W D F .vn ˝ ab/ .v ˝ 1; 1 ˝ a/ in M.E .2/ ˝ Ey .2//.

Proof. See [67, Section 2.8.2, in particular Proposition 2.8.11, Lemma 2.8.18,
Lemma 2.8.20]. 

A variation of the formula for W given above first appeared in [196]. An

instructive motivation for this formula can be found in [67, Section 2.1].
242 Chapter 8. Locally compact quantum groups

8.4.3 The quantum az C b group

The quantum az C b group was constructed by Woronowicz in [204] via the second
approach outlined in the introduction to this section: He starts from the classical
az C b group, deforms the associated Hopf -algebra of polynomial functions, and
formulates an appropriate set of relations and spectral conditions on the generators of
the quantum azCb group. Here, the deformation parameter  either is a root of unity
or an element of .0; 1/. From a representation of the generators as operators on a
Hilbert space, he directly constructs a manageable multiplicative unitary whose right
leg is the quantum az C b group (on the level of C  -bialgebras). This construction
involves an analogue of the quantum exponential function encountered in the study
of E .2/ [198]. The Pontrjagin dual of the quantum az C b group turns out to be
isomorphic to the quantum az C b group again.
The Haar weights on the quantum az C b group were found by Van Daele
[178], see also [127], [205]. He observed that in the root of unity case, the scaling
constant is not equal to 1 – the quantum az C b group was the first example for this
phenomenon. In his PhD thesis [141], Soltan constructed the quantum az Cb group
for new values of the deformation parameter , extending the results of Woronowicz
and Van Daele.

The classical group, its Hopf -algebra, and the deformed Hopf -algebra.
The az C b group is the group of affine transformations of the plane C. It can
identified with the subgroup G  GL2 .C/ consisting of all matrices of the form
 
a b
g.a;b/ WD ; where a; b 2 C; a ¤ 0I
0 1

the action on the plane C being given by g.a;b/ z D az C b.

The -subalgebra A0  C.G/ generated by the coordinate functions g.a;b/ 7!
a and g.a;b/ 7! b is isomorphic to the universal unital commutative -algebra
generated by an invertible element a and an element b, and carries the structure of
a Hopf -algebra, where

0 .a/ D a ˇ a; 0 .b/ D a ˇ b C b ˇ 1;
‡
0 .a/ D 1; 0 .b/ D 0;
(8.15)
S0 .a/ D a1 ; S0 .b/ D a1 b;
S0 .a / D .a /1 ; S0 .b  / D .a /1 b  :

Fix a non-zero  2 C, and denote by A0; the universal unital -algebra

generated by elements a; b satisfying

a is normal and invertible; b is normal; ab D 2 ba; ab  D b  a: (8.16)

 8.4. Examples of locally compact quantum groups 243

The -algebra A0; can be equipped with the structure of a Hopf -algebra, where
the structure maps are given by equation (8.15) again. Moreover, there exists a dual
pairing .  j  / of the Hopf -algebra A0; with itself such that
.aja/ D 2 ; .bja/ D 0; .ajb/ D 0; .bjb/ D t;
where t 2 C can be chosen arbitrarily.

The commutation relations in the setting of C  -algebras. When we shift from

Hopf -algebras to C  -algebras, it is appropriate to replace (8.16) by
a; b are normal; ker a D f0g;

Phase.a/bPhase.a/ D b; (8.17)
it
jaj bjaj
it
D  b for all t 2 R:
it

Here, Phase.a/ denotes the unitary in the polar decomposition of a, see Proposi-
tion 8.4.14.
Assume that H is a Hilbert space and that a; b ˜ L.H / satisfy (8.17). For each
pair .k; t/ 2 Z  R, define a character .k;t/ on C WD C n f0g by

.k;t/ .z/ WD Phase.z/k jzjit for all z 2 C :

Note that the map Z  R ! C y  , .k; t / 7! .k;t/ , is a group isomorphism. Now,

the last two relations in (8.17) imply
.k;t/ .a/b.k;t/ .a/ D kCit b for all .k; t / 2 Z  R: (8.18)
As in the analysis of E .2/, we deduce the existence of an action  W Z  R !
Aut.C0 ..b///, .k;t/ .f / D f .kCit  /, and of a -homomorphism
M.C0 . .b// Ì .Z  R// ! L.H /; f Ì U.k;t/ 7! f .b/.k;t/ .a/:
Denote by m ˜ C  .ZR/ the element corresponding to the identity map on C
with respect to the isomorphism C  .ZR/ Š C0 .C / of Proposition 4.2.3, so that
.k;t/ .m / D U.k;t/ for all .k; t / 2 ZR. Define m ˜ C0 .C/ by .mf /.z/ D zf .z/.
Then the considerations above show that a universal C  -algebra with affiliated
elements a; b satisfying (8.17) is given by
C0 .C/ Ì .Z  R/; a WD 1 Ì m ; b WD m Ì 1: (8.19)

Spectral conditions, the group  , and the C  -algebra A. Similarly as in the

case of the quantum group E .2/, the C  -algebra in (8.19) can not be equipped
with a comultiplication  such that
.a/ D a ˝ a; .b/ D a ˝ b C b ˝ 1:
244 Chapter 8. Locally compact quantum groups

To remedy this defect, one imposes additional spectral conditions on the generators
a; b of the form
x ;
.a/; .b/   where  WD fkCit j .k; t / 2 Z  Rg  C: (8.20)

The shape of  and the analysis of the relations (8.17) and (8.20) depends very
much on the value of . The following cases have been studied:
Case .1/.  2 .0; 1/ ([204], [127], [178], [205]);
Shere,  is a union of circles
whose radii form a geometric progression:  D k2Z k T ;
Case .2/.  D exp. 2 i
/, where N 2 2N ([204], [178], [205]); here,  is a union
N S 1
of straight lines dividing C into N sectors:  D N k
kD0 exp.2
i N /RC ;

Case .3/.  D exp. 1 /, where < < 0 and = 2 2Z n f0g ([141], [205]);
here,  is a union of N logarithmic spirals dividing C into N sectors:  D
SN 1 k it
kD0  f j t 2 Rg.

From now on, we assume that  has one of the forms listed above. In all
three cases,  is a multiplicative subgroup of C and self-dual. Explicitly, the
self-duality can be expressed in terms of a so-called Fresnel function
‚
itk ; case (1);
.k 2 t 2 /=2
˛ W  ! T ; ˛.kCit / WD  ; case (2);
.kCi t /2
e i= 2 ; case (3);

and the function

˛.xy/
 W    ! T ; .x; y/ 7! :
˛.x/˛.y/
Straightforward calculations show that  is symmetric a bicharacter,

 .x; y/ D  .y; x/;  .x; yz/ D  .x; y/ .x; z/ for all x; y; z 2  ;

y given by x 7!  .x;  / is a group isomorphism, and that

that the map  7! 

 .x; kCit / D Phase.x/k jxjit D .k;t/ .x/ for all .k; t / 2 Z  R; x 2  :

(8.21)

The universal C  -algebra with affiliated elements a; b satisfying (8.17) and

(8.20) can be constructed as follows. Define
x //;
 W  ! Aut.C0 . x /;
m ˜ C0 . m ˜ C  .Z  R/
 8.4. Examples of locally compact quantum groups 245

(similarly) as above, and denote by  W C  .Z  R/ ! M.C  . // the -homo-

morphism induced by the homomorphism  W Z  R !  , .k; t / 7! kCit . Then
similar considerations as in the preceding paragraph show that the triple
x / Ì  ;
A WD C0 . a WD 1 Ì  .m / ˜ A; b WD m Ì 1 ˜ A

has the desired universality property. Moreover, A is generated by a, a1 , b in the

sense of Definition 8.4.15 [204, Proposition 4.1], [141, Proposition 6.1]. Note that
equation (8.21) implies

UkCi t D  .U.k;t/ / D  ..k;t/ .m // D .k;t/ . .m // D  . .m /; kCit /

for all .k; t/ 2 Z  R. Therefore, a ˜ A can also be characterized by the relation

1 Ì Ux D  .a; x/ for all x 2  :

Quantum exponential function, a multiplicative unitary, and the C  -bialgebra

structure on A. The comultiplication on A is defined via a multiplicative unitary
W as in the second approach outlined in the introduction.
Similarly as in the case of E .2/, the definition of W involves a (continuous)
quantum exponential function F W  x ! T , which is characterized by

1
Y 1 C 2k z

˚
case (1), (3): F .z/ D x n f2k j k 2 Ng;
if z 2 
1 C 2k z
kD0

f0 .r/ Y 1 C 2k r
l=2
; 2 j l; r > 0;
1Cr
kD1 1 C  r
2k
case (2): F .l r/ D
Y
.l1/=2
1 C 2kC1 r
f0 .r/ ; 2 − l; r > 0;
kD0 1 C 2kC1 r
 Z 1 
1 log.1 C t N=2 /
where f0 .z/ D exp dt ;

i 0 t C z 1
see [204, Section 1] and [141, Proposition 2.3]. In all cases, F satisfies an ana-
logue of Theorem 8.4.28 which explains the terminology [204, Theorem 2.6], [141,
Theorem 3.5].
Let us call a non-degenerate representation
of A on a Hilbert space H proper
if it satisfies the following equivalent conditions:
i)
.b/ ˜ L.H / is invertible,
ii) ker
.b/ D 0,
246 Chapter 8. Locally compact quantum groups

iii) the restriction of

to C0 . / Ì   A is non-degenerate.

Theorem 8.4.31. Let

be a proper non-degenerate representation of A on a Hilbert
space H . Put a0 WD
.a/ and b0 WD
.b/.

i) There exists a proper non-degenerate representation

O W A ! L.H / such
that
(
1 b 1 a0 ; case .1/, .3/;
O

.a/ D b0 ;
.b/
O D 0 1
a0 b0 ; case .2/:

Put aO 0 WD
.a/
O and bO0 WD
.b/.
O

ii) W WD F .bO0 ˝ b0 / .aO 0 ˝ 1; 1 ˝ a0 / is a modular multiplicative unitary.

iii) We have

A.W / D
.A/; W .a0 / D a0 ˝ a0 ; W .b0 / D a0 ˝ b0 C b0 ˝ 1;
y / D
.A/;
A.W O  y W .bO0 / D bO0 ˝ aO 0 C 1 ˝ bO0 :
y W .aO 0 / D aO 0 ˝ aO 0 ; 

Proof. The proofs are very similar in all cases and scattered over several papers.
Case (1): i) is contained in [204, Theorem A.1]; for ii), iii) see [127, Section 3].
Case (2): i) follows from [204, Propositions 2.1, 2.2], see also [178, Lemma 3.9];
ii) is stated in [142, Section 5], see also [204, Proposition 3.2]; iii) follows similarly
as in case (3), see also [204, Sections 6, 7] and [204, Proposition 3.2].
Case (3): i) follows from [141, Theorem 3.2, Corollary 3.3], ii) [141, Proposi-
tion 5.2, Corollary 5.7], iii) [141, Propositions 5.2, 6.2, 6.3]. 

Now we can choose a non-degenerate faithful representation

of C0 . / Ì
 on some Hilbert space H , extend it to a faithful representation
of A 
M.C0 . / Ì  /, apply Theorem 8.4.31, and find:

Corollary 8.4.32. There exists a .unique/ non-degenerate -homomorphism

 W A ! M.A ˝ A/ such that .a/ D a ˝ a and .b/ D a ˝ b C b ˝ 1.
Moreover, .A; / is a bisimplifiable C  -bialgebra. 

The antipode and the Haar weights of A. From the modular multiplicative
unitary W constructed above, we obtain a unitary antipode R and a scaling group
. t / t (see Theorem 7.3.19):

Theorem 8.4.33. i) The unitary antipode R of .A; / is the unique -antiautomor-

phism R W A ! A satisfying R.a/ D a1 , R.b/ D a1 b.
 8.4. Examples of locally compact quantum groups 247

ii) The scaling group2 . t / t of .A; / is given by  t .x/ D jaj2it xjaj2it ; in

particular,  t .a/ D a and  t .b/ D 2it b for each t 2 R.
Proof. Case (1): [127, Section 3].
Case (2): [204, Proposition 3.3, Section 6]; Woronowicz proves the assertions
starting from a modified manageable multiplicative unitary instead of W , but the
argument can be adapted to W .
Case (3): [141, Section 6.2]. 
A right Haar weight on .A; / was first found by Van Daele [178]:
Theorem 8.4.34. i) There exists a right Haar weight on .A; / such that for all
c 2 A of the form c D g.a/f .b/, where g 2 C0 . /, f 2 C0 . x /,
Z Z

.c c/ D jg.x/j d .x/
2
jf .x/j2 jxj2 d .x/I
 x


here,  denotes the Haar measure of  .

ii) The modular automorphism group . t / t of is given by

 t .x/ D jbj2it xjbj2it I

in particular,  t .a/ D 2it a and  t .b/ D b for each t 2 R.

Proof. i) See [205, Theorem 3.1, Equation (3.8)]; the original references are [178,
Theorems 4.5, 5.11].
ii) This follows from [205, Equation (3.7)]; for cases (1), (2), see also [178,
Theorem 4.4, Proposition 5.6]. 
A left Haar weight on .A; / is given by  WD ı R. Consequently, we find:
Theorem 8.4.35. .A; / is a locally compact quantum group.
The quantum az C b group was the first example of a locally compact quantum
group with scaling constant not equal to 1 (in cases (2), (3)):
Proposition 8.4.36. ı  t D j4it j ; in particular, the scaling constant  of
.A; / is not 1.
Proof. Case (1): [178, Proposition 4.12]; case (2): [178, Section 5]; case (3) (and
also cases (1), (2)): [205, Section 3]. 

2
Here, we adopt the sign convention for  chosen by Woronowicz, which differs from the convention
chosen by Kustermans and Vaes.
 Part III

Selected topics
Chapter 9
Coactions on C  -algebras, reduced crossed
products, and duality

The topic of this chapter are coactions of C  -bialgebras on C  -algebras. Such

coactions generalize actions of groups on one side and Fell bundles on groups on
the other side; in particular, they provide a unified approach to these two concepts.
We review group actions in Section 9.1 and introduce the concept of a coaction in
Section 9.2.
Like actions of a group, every coaction of a C  -bialgebra can be encoded
by a crossed product. The construction of this crossed product involves not
only the C  -bialgebra of the coaction but also a second C  -bialgebra that is
dual to the initial one in a suitable sense. In the classical setting of a group ac-
tion, such a pair of dual C  -bialgebras appears as follows. An action of a locally
compact group G on a C  -algebra C corresponds to a coaction of the C  -bialgebra
C0 .G/ on C , and the associated (reduced) crossed product C Ì.r/ G can be con-

sidered as a twisted tensor product of C and the C  -bialgebra C.r/ .G/. For the
convenience of the reader, we review the reduced crossed product construction in
Section 9.1.
As a framework for the construction of reduced crossed products for coactions
of C  -bialgebras, Baaj and Skandalis introduced the concept of a Kac system [6].
A Kac system consists of a regular multiplicative unitary V and an additional sym-
metry, subject to a number of axioms. The multiplicative unitary V gives rise to
y /; /
two C  -bialgebras .A.V y and .A.V /; / (see Chapter 7), which are dual to
each other in a suitable sense. Now the reduced crossed product for a coaction of
y /; /
the C  -bialgebra .A.V y or .A.V /; / on a C  -algebra C can be considered
as a twisted tensor product of C and the C  -algebra A.V / or A.V y /, respectively.
These concepts and constructions are explained in Sections 9.3 and 9.4.
The main result of this chapter is the Baaj–Skandalis duality theorem which
generalizes Takesaki–Takai duality. Let us briefly recall the latter. For every action
˛ of a locally compact abelian group G on a C  -algebra C , the associated crossed
product C Ì˛;r C carries a dual action ˛O of the dual group G, y and the Takesaki–
y
Takai duality theorem identifies the bidual action ˛OO of G
y Š G on C Ì˛;r G Ì˛;r y
O G
with a stabilization of the initial action ˛ on C . In particular, the iterated reduced
crossed product is just a stabilization of C . Thus, Takesaki–Takai duality extends
the Pontrjagin duality of locally compact abelian groups to actions of such groups
on C  -algebras. A short summary is given in Section 9.1.
252 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

The duality theorem of Baaj and Skandalis extends the result of Takesaki and
Takai as follows: In the framework of Kac systems, the reduced crossed product
of a coaction of a C  -bialgebra carries a dual coaction of the dual C  -bialgebra,
and the theorem of Baaj and Skandalis identifies the dual of that coaction (that is,
the bidual coaction) with a stabilization of the initial coaction. On the level of the
underlying C  -algebras, this means that the iterated reduced crossed product is a
stabilization of the initial C  -algebra. These constructions and results are presented
in Sections 9.4 and Section 9.5.
The definitions and results presented in Sections 9.3–9.5 are taken from the
fundamental article of Baaj and Skandalis [7]. The presentation, however, has been
adapted to the purpose of this book.
Let us briefly comment on the setting of the theory presented in this chapter.
The most comprehensive framework for the generalization of group actions, the
crossed product construction, and Takesaki–Takai duality to quantum groups in the
setting of C  -algebras or von Neumann algebras is the theory of locally compact
quantum groups introduced in Chapter 8. In that framework, one can work on the
level of
• locally compact quantum groups in the setting of von Neumann algebras,
• reduced C  -algebraic quantum groups, or
• universal C  -algebraic quantum groups.
Similarly, one can consider
• coactions of von Neumann bialgebras on von Neumann algebras and associ-
ated crossed products,
• coactions of C  -bialgebras on C  -algebras and associated reduced crossed
products, and
• coactions of C  -bialgebras on C  -algebras and associated full/universal
crossed products.
In the first setting, a rich theory with many important results, applications, and
examples was developed by Vaes [159], [160], [161] and Vaes and Vainerman [163],
building on work of Enock and Schwartz [41], [46] and Enock and Nest [45]. The
theory presented in this chapter fits into the second setting. A full/universal crossed
product construction was introduced by Vergnioux [182].

9.1 Actions of groups and Takesaki–Takai duality

This section gives a brief review on group actions on C  -algebras and associated
reduced crossed products. The definitions and results presented here are classical,
therefore we omit the proofs. Reference are, for example, [121], [191].
 9.1. Actions of groups and Takesaki–Takai duality 253

Throughout this section, G denotes a locally compact group with left Haar
measure  and modular function ı.
Definition 9.1.1. A .strongly continuous/ action of G on a C  -algebra C is a group
homomorphism ˛ W G ! Aut.C /; x 7! ˛x , such that for every c 2 C , the map
G ! C given by x 7! ˛x .c/ is norm-continuous. We call C a G-C  -algebra if ˛
is understood.
Let ˛ and ˇ be actions of G on C  -algebras C and D, respectively. A non-
degenerate -homomorphism W C ! M.D/ is called covariant with respect to ˛
and ˇ if .˛x .c// D ˇx . .c// for all c 2 C and x 2 G. Here, ˇx is extended to a
-automorphism of M.C / for each x 2 G.
Actions on C  -algebras are usually studied in terms of covariant representations:
Definition 9.1.2. Let ˛ be an action of G on a C  -algebra C . A .non-degenerate/
covariant representation of .C; ˛/ is a triple .H;
; u/ consisting of a Hilbert
space H , a (non-degenerate) representation
of C on H , and a unitary repre-
sentation u of G on H such that

.˛x .c// D u.x/
.c/u.x/ for all c 2 C; x 2 G: (9.1)

Remark 9.1.3. In the situation above, the representation u induces a strongly con-
tinuous action Adu of G on the C  -algebra K.H / via

Adu .x/ W T 7! u.x/T u.x/ for all x 2 G; T 2 K.H /;

and condition (9.1) is equivalent to the condition that the -homomorphism

W C !
L.H / Š M.K.H // is covariant with respect to ˛ and Adu .
Let ˛ be an action of G on a C  -algebra C . Then the space Cc .GI C / of C -
valued continuous functions on G with compact support is a -algebra with respect
to the operations
Z
.f  g/.x/ WD f .y/˛y .g.y 1 x//d .y/; f  .x/ WD ˛x .f .x 1 //  ı.x/1 :
G

We shall define the reduced crossed product of the action ˛ in terms of a left
regular representation of Cc .GI C /, using the language of C  -modules (for a short
summary, see Section 12.2). We consider the Hilbert space L2 .G; / and the C  -
algebra C as C  -modules over C and C , respectively, and form the tensor product
L2 .G; / ˝ C , which is a C  -module over C again. This C  -module can also be
described as follows. It is easy to see that the space Cc .GI C / is a pre-C  -module
over C with respect to the operations
Z
hf jgi WD f .x/ g.x/d .x/ and .f c/.x/ WD f .x/c; where f; g 2 C; c 2 C:
G
254 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Denote the completion of this pre-C  -module by L2 .GI C /. Then the natural
embedding of the algebraic tensor product Cc .G/ ˇ C into Cc .GI C / extends to
Š
an isomorphism of C  -modules L2 .G; / ˝ C 
! L2 .GI C /, as one can easily
check.

Proposition 9.1.4. Let ˛ be an action of G on a C  -algebra C . Then there exist

non-degenerate injective -homomorphisms

C ! LC .L2 .GI C //; c 7! c Ìr 1;

Cr .G/ ! LC .L2 .GI C //; L.h/ 7! 1 Ìr L.h/;
Cc .GI C / ! LC .L2 .GI C //; f 7! L˛ .f /;

such that for all x 2 G and g 2 Cc .GI C /,

..c Ìr 1/g/.x/ D ˛x 1 .c/g.x/;

Z
..1 Ìr L.h//g/.x/ D h.y/g.y 1 x/d .y/;
ZG

.L˛ .f /g/.x/ D ˛x 1 .f .y//g.y 1 x/d .y/:

G

Proof. This can be checked by straightforward calculations; see, for example, [121],
[191]. 

Definition 9.1.5. Let ˛ be an action of G on a C  -algebra C . The associated

reduced crossed product is the C  -subalgebra C Ì˛;r G  LC .L2 .GI C // given
by

C Ì˛;r G WD span fL˛ .f / j f 2 Cc .GI C /g:

The reduced crossed product can also be constructed without reference to C  -

modules (see, for example, [121]), but the definition given above is more direct and
more convenient.

Proposition 9.1.6. Let ˛ be an action of G on a C  -algebra C . Then

C Ìr 1  M.C Ì˛;r G/; 1 Ìr Cr .G/  M.C Ì˛;r G/;

span ..C Ìr 1/.1 Ìr C  .G/// D C Ì˛;r G D span ..1 Ìr C  .G//.C Ìr 1//:

Proof. Again, this can be checked by straightforward calculations and arguments.



The reduced crossed product has the following simple functorial property:
 9.1. Actions of groups and Takesaki–Takai duality 255

Proposition 9.1.7. Let ˛ and ˇ be actions of G on C  -algebras C and D, respec-

tively, and let W C ! M.D/ be a non-degenerate covariant -homomorphism.
Then there exists a unique non-degenerate -homomorphism

Ìr id W C Ì˛;r G ! M.D Ìˇ;r G/

such that . Ìr id/.L˛ .f // D Lˇ . ı f / for all f 2 Cc .GI C /.

Proof. Uniqueness is evident, let us prove existence. The C  -module L2 .GI D/ Š

L2 .G; / ˝ D can be identified with the internal tensor product L2 .G; / ˝ C ˝
D Š L2 .GI C / ˝ D because C ˝ D Š D, and the map

LC .L2 .GI C // ! LD .L2 .GI C / ˝ D/ Š LD .L2 .GI D//;

T 7! T ˝ idD ;

is a -homomorphism (see Section 12.2). An easy calculation shows that L˛ .f /˝

idD D Lˇ . ı f / for all f 2 Cc .GI C /. 

In the remaining part of this section, we focus on the case that G is abelian.
Then the reduced crossed product C Ì˛;r G carries a natural dual action ˛O of the
dual group G,y and the assignment C 7! C Ì˛;r G extends to a functor from G-
 y  -algebras.
C -algebras to G-C

Proposition 9.1.8. Let ˛ be an action of G on a C  -algebra C , where G is

abelian. For every continuous character  on G, there exists an automorphism
˛O  of Cc .GI C / such that

.˛O  .f //.x/ D .x/f .x/ for all x 2 G; f 2 Cc .GI C /:

This automorphism extends to an automorphism of the C  -algebra C Ì˛;r G. The

y on C Ì˛ G.
assignment  7! ˛O  defines a strongly continuous action of G

Proof. See [121], [191] 

Definition 9.1.9. Let ˛ be an action of G on a C  -algebra C , where G is abelian.

y on C Ì˛;r G constructed above is called the dual action of ˛.
The action ˛O of G

The construction of the dual action evidently is functorial in the following sense:

Proposition 9.1.10. Let ˛ and ˇ be actions of G on C  -algebras C and D, re-

spectively, where G is abelian. Let W C ! M.D/ be a non-degenerate covariant
-homomorphism. Then the -homomorphism Ìr id W C Ì˛;r G ! M.D Ìˇ;r G/
O
is covariant with respect to ˛O and ˇ.
256 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

The main result in this section is the Takesaki–Takai duality theorem. Roughly,
this theorem extends Pontrjagin duality from groups to actions: Given a locally
compact abelian group G, we constructed a functor from the category of G-C  -
y  -algebras. Replacing G by G,
algebras to the category of G-C y we get a functor
y
from G-C  y
y
-algebras to G-C 
-algebras, and by Pontrjagin duality, we can identify
Gy
y with G. Thus we obtain two functors

.C;˛/7!.C Ì˛;r G; ˛/
O
category of / category of
G-C  -algebras o y ˇO /
y
G-C  -algebras.
.DÌˇ;r G; .D;ˇ /

The Takesaki–Takai duality theorem says that these functors induce an equivalence
of categories up to equivariant Morita equivalence. We shall only give a less abstract
and more concrete formulation of the duality theorem: For every G-C  -algebra C ,
the associated bidual C Ì˛;r G Ì˛;r y
O G is naturally covariantly isomorphic to a stabi-
lization of C . This stabilization is a tensor product of C with K.L2 .G; //, where
G acts on K.L2 .G; // via the right regular representation W G ! L.L2 .G//,
given by . .x//.y/ D .yx/ for all y 2 G;  2 L2 .G; /; x 2 G.

Theorem 9.1.11 (Takesaki–Takai duality). Let ˛ be an action of G on a C  -alge-

bra C , where G is abelian. Then there exists an isomorphism

C Ì˛;r G Ì˛;r y
O G Š C ˝ K.L .G; //
2

that is covariant with respect to ˛OO and ˛  Ad , respectively, where the action
˛  Ad is given by .˛  Ad /x .c ˝ T / D ˛x .c/ ˝ x1 T x for all c 2 C; T 2
K.L2 .G; //.

A proof is given in [121, Theorem 7.9.3], see also [150, Chapter X]. We shall
prove a more general result – the Baaj–Skandalis duality theorem – in Section 9.5.2.

9.2 Coactions of C  -bialgebras on C  -algebras

The central concept in this chapter is that of a coaction:

Definition 9.2.1. A .right/ coaction of a C  -bialgebra .A; / on a C  -algebra

C is a non-degenerate -homomorphism ı W C ! M.C ˝ A/ which satisfies the
following two conditions:

i) ı.C /.1 ˝ A/  C ˝ A,
 9.2. Coactions of C  -bialgebras on C  -algebras 257

ii) .ı ˝ idA / ı ı D .idC ˝/ ı ı, that is, the following diagram commutes:

C
ı / M.C ˝ A/

ı ı˝idA
 
M.C ˝ A/ / M.C ˝ A ˝ A/.
idC ˝

We also refer to the pair .C; ı/ as a (right) coaction of .A; /. A coaction .C; ı/ is
an .A; /-C  -algebra if

iii) ı is injective and

iv) ı.C /.1 ˝ A/ is linearly dense in C ˝ A.

Let .C; ıC / and .D; ıD / be coactions of .A; /. A non-degenerate -homo-

morphism W C ! M.D/ is covariant with respect to ıC and ıD if ıD ı D
. ˝ idA / ı ıC , that is, if the following diagram commutes:

C / M.D/

ıC ıD
 
M.C ˝ A/ / M.D ˝ A/.
˝idA

Remarks 9.2.2. i) In condition ii) of the definition above, the -homomorphisms

ı˝idA and idC ˝ are extended to the multiplier C  -algebra M.C ˝A/. Likewise,
in the last paragraph, ıD is extended to M.D/ and ˝idA is extended to M.C ˝A/.
ii) Evidently, the class of all coactions of a fixed C  -bialgebra .A; / forms a
category with respect to covariant -homomorphisms.
iii) Left coactions can be defined similarly as right coactions, and left coactions of
a C  -bialgebra .A; / correspond bijectively with right coactions of the coopposite
C  -bialgebra .A; /cop defined in Remark 4.1.2 ii). Since left coactions will not
play a rôle in this book, we shall refer to right coactions simply as coactions.
iv) For every C  -algebra C and C  -bialgebra .A; /, one can define a trivial
coaction of .A; / on C by the formula c 7! c ˝ 1 for all c 2 C . Clearly, the
trivial coaction turns C into an .A; /-C  -algebra.
v) For every C  -bialgebra .A; /, the map  W A ! M.A ˝ A/ is a right
coaction (and a left coaction) of .A; / on A itself. This is called the right (or left)
regular coaction.

There exist several standard constructions to produce new coactions out of given
ones:
258 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Remark 9.2.3. Let .A; A / and .B; B / be C  -bialgebras. Recall that A ˝ B is

a C  -bialgebra with respect to the -homomorphism A˝B given by

A ˝B
A ˝ B ! M.A ˝ A/ ˝ M.B ˝ B/
id ˝†˝id
,! M.A ˝ A ˝ B ˝ B/ ! M.A ˝ B ˝ A ˝ B/I

here, † denotes the flip A ˝ B ! B ˝ A; a ˝ b 7! b ˝ a.

Given a coaction .C; ıC / of .A; A / and a coaction .D; ıD / of .B; B /, we
can form the following new coactions:
External direct sum. The composition of the map

ıC ˚ ıD W C ˚ D ! M.C ˝ A/ ˚ M.D ˝ B/

with the extension of the non-degenerate -homomorphism

.C ˝ A/ ˚ .D ˝ B/ ! M..C ˚ D/ ˝ A ˝ B/;
..c ˝ a/; .d ˝ b// 7! .c; 0/ ˝ a ˝ 1 C .0; d / ˝ 1 ˝ b;

is a coaction of .A ˝ B; A˝B / on C ˚ D, which we denote by ıC ˚D .

External tensor product. The composition

ıC ˝ıD
C ˝ D ! M.C ˝ A/ ˝ M.D ˝ B/
id ˝†˝id
,! M.C ˝ A ˝ D ˝ B/ ! M.C ˝ D ˝ A ˝ B/

is a coaction of .A ˝ B; A˝B / on C ˝ D, which we denote by ıC ˝D . Here,

† W A ˝ D ! D ˝ A denotes the flip a ˝ d 7! d ˝ a again.
Push-forward. If  W A ! M.B/ is a morphism of C  -bialgebras, then the
pair .C; .idC ˝/ ı ıC / is a coaction of .B; B /.
Internal direct sum. Assume that .A; A / D .B; B /. Then the composition
of the map

ıC ˚ıD
C ˚ D ! M.C ˝ A/ ˚ M.D ˝ A/ ,! M..C ˝ A/ ˚ .D ˝ A//
Š
with the extension of the isomorphism .C ˝A/˚.D ˝A/  ! .C ˚D/˝A is
a coaction of .A; A / on C ˚D. If A is commutative, this internal direct sum
coincides with the push-forward of the external direct sum .C ˚ D; ıC ˚D /
along the multiplication map mA W A ˝ A ! A; a1 ˝ a2 7! a1 a2 .
 9.2. Coactions of C  -bialgebras on C  -algebras 259

Internal tensor product. Assume that .A; A / D .B; B / and that A is

commutative. Then the push-forward of the external tensor product .C ˝
D; ıC ˝D / along the multiplication map mA W A ˝ A ! A is a coaction of
.A; A / on C ˝ D. If A is not commutative, then in general, such an internal
tensor product can not be defined.

Actions of a group G and coactions of C0 .G /. Coactions generalize actions in

the following sense: For every locally compact group G, there exists a bijective
correspondence between G-C  -algebras and .C0 .G/; /-C  -algebras. To explain
this correspondence, we need to recall a few facts and some notation.
Let X be a locally compact space and C a C  -algebra. Then the C  -algebra
Cb .X/ of bounded continuous functions on X can be identified with the multiplier
algebra M.C0 .X //, and the C  -algebra Cb .XI C / of bounded continuous C -valued
functions on X can be identified with a C  -subalgebra of M.C ˝ C0 .X // [2,
Corollary 3.4]. For each x 2 X , we denote by evx W Cb .X / ! C the evaluation
at x. The non-degenerate -homomorphism id ˝ evx W C ˝ C0 .X / ! C extends
to a -homomorphism M.C ˝ C0 .X // ! M.C /, which we denote by id ˝ evx
again.

Theorem 9.2.4. Let G be a locally compact group and C a C  -algebra.

i) Let ˛ be an action of G on C . Then the map ı0 W C ! Cb .GI C / given by

.ı0 .c//.x/ WD ˛x .c/ for all c 2 C; x 2 G;

is a -homomorphism. Denote by ı the composition of ı0 with the canonical

embedding Cb .GI C / ,! M.C ˝C0 .G//. Then ı is a coaction of .C0 .G/; /
and the pair .C; ı/ is a .C0 .G/; /-C  -algebra.

ii) Let ı be a coaction of .C0 .G/; / on C such that

(a) ı is injective, or
(b) ı.C /.1 ˝ C0 .G// is linearly dense in C ˝ C0 .G/.

Then for each x 2 G, the map

˛x WD .id ˝ evx / ı ı W C ! M.C /

is an automorphism of C , and the map ˛ W x 7! ˛x is an action of G on C .

iii) The constructions in i) and ii) define a bijective correspondence between the
set of all actions of G on C and the set of all coactions of .C0 .G/; / on C
that turn C into a .C0 .G/; /-C  -algebra.
260 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

iv) Let .C; ˛/, .D; ˇ/ be actions of G and .C; ıC /, .D; ıD / the corresponding
coactions of .C0 .G/; /. Then a non-degenerate -homomorphism W C !
M.D/ is covariant with respect to ˛ and ˇ if and only if it is covariant with
respect to ıC and ıD .
v) The assignment .C; ˛/ 7! .C; ı/ constructed in i) and the assignment  7! 
define an equivalence of the category of G-C  -algebras and the category
of .C0 .G/; /-C  -algebras, where the morphisms are all non-degenerate
covariant -homomorphisms.
Some of the arguments of the following proof were used already in Exam-
ple 5.2.5, where we related representations of a locally compact group G to corepre-
sentations of the C  -bialgebra .C0 .G/; /.
Proof. i) Evidently, ı0 is a -homomorphism. We claim that .ı ˝ id/ ı ı D
.id ˝/ ı ı. By construction,
.id ˝ evx ˝ evy / ı .ı ˝ id/ ı ı D .id ˝ evx / ı ı ı .id ˝ evy / ı ı D ˛x ı ˛y ;
.id ˝ evx ˝ evy / ı .id ˝/ ı ı D .id ˝ evxy / ı ı D ˛xy for all x; y 2 G:
(9.2)
Now the claim follows from the fact that ˛x ı ˛y D ˛xy for all x; y 2 G and that
maps of the form id ˝ evx ˝ evy separate the elements of M.C ˝C0 .G/˝C0 .G//.
Clearly, ı.C /.1 ˝ C0 .G//  C ˝ C0 .G/, so ı is a coaction.
Next, we show that .C; ı/ is a .C0 .G/; /-C  -algebra. Evidently, the map ı
is injective; in fact, .idC ˝ eve / ı ı D idC , where e 2 G denotes the unit. It
remains to show that the set ı.C /.1 ˝ C0 .G// is linearly dense in C ˝ C0 .G/, or,
equivalently, that the set ı0 .C /C0 .G/ is linearly dense in C0 .GI C /. The equation
.ı0 .˛x 1 .c///.x/ D ˛x .˛x 1 .c// D c shows that for every x 2 G and every
c 2 C , there exists an element in ı0 .C / that takes the value c at the point x. Now
the density that we need to prove follows from a standard argument.
ii) The inclusion ı.C /.1˝C0 .G//  C ˝C0 .G/ implies that ı.C / is contained
in the subspace Cb .GI C / of M.C ˝ C0 .G//. Thus, the image of ˛x is contained
in C for every x 2 G, and the map G ! C given by x 7! ˛x .c/ is continuous for
every c 2 C .
The relation .ı ˝ id/ ı ı D .id ˝/ ı ı and equation (9.2) imply that ˛x ı ˛y D
˛xy for all x; y 2 G. Let us show that for each x 2 G, the -homomorphism ˛x
is an automorphism. The relations ˛y D ˛yx 1 ı ˛x and ˛y D ˛x ı ˛x 1 y , and
similar relations with x and y exchanged, show that
ker ˛x D ker ˛y ; Im ˛x D Im ˛y for all x; y 2 G:
Thus it suffices to show that ker ˛e D 0 and that Im ˛e D C . Since the map ˛e is
idempotent, these two conditions are equivalent. If assumption (a) is satisfied, then
 9.2. Coactions of C  -bialgebras on C  -algebras 261
T
ker ˛e D x2G ker ˛x D ker ı D 0; if assumption (b) is satisfied, then

Im ˛e D span .id ˝ eve /.ı.C /.1 ˝ C0 .G/// D C  eve .C0 .G// D C:

Summarizing, we find that the map ˛ W G ! Aut.C /; x 7! ˛x , is an action.

iii) This follows immediately from the definition of ı0 in i) and of ˛x in ii).
iv) We have . ˝ id/ ı ıC D ıD ı  if and only if

.id ˝ evx / ı . ˝ id/ ı ıC D .id ˝ evx / ı ıD ı  for all x 2 G;

that is, if and only if  ı ˛x D ˇx ı  for all x 2 G.

v) Obvious from iii) and iv). 
Remark 9.2.5. The previous result implies that every coaction .C; ı/ of the C  -bi-
algebra .C0 .G/; / that satisfies one of the conditions ii)(a) and ii)(b) automatically
satisfies both conditions and is a .C0 .G/; /-C  -algebra.
The preceding result can be extended to groupoids, see Theorem 11.5.7.

Fell bundles on a group G and coactions of Cr .G /. The concept of a coaction

generalizes not only the concept of an action, but also the concept of a Fell bundle
on a group: To every Fell bundle F on a locally compact group G, one can associate
a coaction of Cr .G/ on the reduced C  -algebra of that Fell bundle. Let us briefly
outline this construction. Standard references on Fell bundles are [53], [54].
Let G be a locally compact group with left Haar measure  and associated right
Haar measure 1 .
Recall that a continuous Banach bundle on G is a topological space F with an
open continuous surjection p W F ! G such that
i) for each point x 2 G, the fiber Fx WD p 1 .x/ is endowed with a Banach
space structure;
ii) the maps C  F ! F and F .2/ ! F given in each fiber by scalar multipli-
cation and addition, respectively, are continuous; here, F .2/ denotes the set
f.x; y/ 2 F  F j p.x/ D p.y/g;
iii) the map F ! R given by the norm on each fiber is continuous;
iv) for each point x 2 G and each neighborhood W of the zero 0x over x in
F , there exist a number > 0 and a neighborhood U of x such that the set
fe 2 p 1 .U / j kek < g is contained in W .
A Fell bundle on G is a continuous Banach bundle p W F ! G together with a
continuous multiplication map F  F ! F and an involution  W F ! F such
that for all a; a1 ; a2 ; a3 2 F ,
262 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

i) p.a1 a2 / D p.a1 /p.a2 / and p.a / D p.a/1 ;

ii) for all x; y 2 G, the map Fx  Fy ! Fxy induced by the multiplication is
bilinear, and the map Fx ! Fx 1 induced by the involution is conjugate-
linear;
iii) .a1 a2 /a3 D a1 .a2 a3 /; .a1 a2 / D a2 a1 , and .a / D a;
iv) ka1 a2 k  ka1 kka2 k; ka ak D kak2 D ka k2 , and a a 0.
Here, Fx D p 1 .x/  F for each x 2 G. Evidently, the fiber Fe over the unit
e 2 G is endowed with the structure of a C  -algebra.
To the Fell bundle F , one can associate a reduced C  -algebra Cr .F / as follows.
Denote by c .F / the space of continuous sections of F with compact support. This
space is a pre-C  -module over the C  -algebra Fe , where
Z
hji D .x/ .x/d1 .x/; .a/.x/ D .x/a for all ;  2 c .F /; a 2 Fe :
G

We denote the associated C  -module by L2 .GI F /. Each section f 2 c .F /

defines a left convolution operator LF .f / 2 LFe .L2 .GI F // via
Z
.LF .f //.x/ D f .y/.y 1 x/d .y/ for all  2 c .F /; x 2 G;
G

and each a 2 F defines a left convolution operator LF .a/ 2 LFe .L2 .GI F // via

.LF .a//.x/ D a.p.a/1 x/ for all  2 c .F /; x 2 G:

The reduced C  -algebra of F is the C  -subalgebra Cr .F /  LFe .L2 .GI F //

generated by the operators LF .f /, where f 2 c .F /. It is easy to see that for
each a 2 F , the operator LF .a/ belongs to M.Cr .F //.
Recall that for each x 2 G, we defined a multiplier Lx 2 M.Cr .G// (see
Example 4.2.2).
Theorem 9.2.6. i) Let F be a Fell bundle on a locally compact group G. There
exists a unique coaction ıF of Cr .G/ on Cr .F / whose extension to M.Cr .F //
satisfies

ıF .LF .a// D LF .a/ ˝ Lp.a/ for all a 2 F:

This coaction turns Cr .F / into a Cr .G/-C  -algebra.

ii) Assume that G is discrete and that ı is an injective coaction of Cr .G/ on
a C  -algebra C . Then .C; ı/ is a Cr .G/-C  -algebra. There exists a unique Fell
bundle F on G such that

Fx D fc 2 C j ı.c/ D c ˝ Lx g for all x 2 G;

 9.3. Weak Kac systems 263

and such that the multiplication and involution on F are induced by the multi-
plication and involution on C . Moreover, the map c .F / ! C given by f 7!
P  Š
x2G f .x/ extends to an isomorphism Cr .F / 
! C . With respect to this isomor-
phism, the coaction ıF corresponds to the coaction ı.

Proof. The coaction in i) is constructed by means of an auxiliary unitary, similarly

as the comultiplication on Cr .G/; see, for example, [97]. For the proof of ii), see
[6, Corollaire 15]. 

The preceding result can be extended to groupoids, see Theorem 11.5.7.

9.3 Weak Kac systems

The rest of this chapter is devoted to reduced crossed products for coactions of C  -
bialgebras. The proper framework for the construction of such crossed products are
weak Kac systems. A weak Kac system consists of a well-behaved multiplicative
unitary V and an additional symmetry U . The multiplicative unitary V gives
y /; /
rise to two C  -bialgebras .A.V y and .A.V /; /, called the legs of V (see
Section 7.2), and for coactions of these C  -bialgebras, we shall construct reduced
crossed products. The additional symmetry U serves two purposes:
y /; /
• to overcome the asymmetry between the left leg .A.V y and the right leg
.A.V /; / with respect to right coactions (as opposed to both right and left
coactions);
• to define a “right-regular” representation x 7! AdU .x/ of each leg that com-
mutes with the natural (“left-regular”) representation on the Hilbert space
underlying the multiplicative unitary.

These two purposes are addressed in Sections 9.3.1 and 9.3.2, where we introduce
the notion of a balanced multiplicative unitary and a weak Kac system. Examples
of weak Kac systems are discussed in Section 9.3.3.
We shall use the notation and concepts introduced in Chapter 7; in particular,
the leg notation (see Notation 7.1.1) and slice maps (see Section 12.4).

9.3.1 Balanced multiplicative unitaries

A balanced multiplicative unitary is a multiplicative unitary V equipped with an
additional symmetry U that allows us to identify the left and the right leg of V with
the right and the left leg of two auxiliary multiplicative unitaries V{ and Vy . Then, the
rôles of the right leg and the left leg get switched when we pass from V to V{ or Vy ,
so that we obtain a perfect symmetry between “right” and “left”. This symmetry
264 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

will be used in the construction of reduced crossed products for coactions of the
legs of V in Section 9.4. Let us turn to the precise definition.
Let H be a Hilbert space and U 2 L.H / a symmetry, that is, a self-adjoint
unitary. As before, we denote by † 2 L.H ˝ H / the flip  ˝  7!  ˝ .
For each T 2 L.H ˝ H /, put
T{ WD †.1 ˝ U /T .1 ˝ U /†; Ty WD †.U ˝ 1/T .U ˝ 1/†: (9.3)
The maps T 7! T{ and T 7! Ty are automorphisms of order four. Indeed, the
squares of these maps are given by T 7! .U ˝ U /T .U ˝ U / D Ad.U ˝U / .T /
because
†.1 ˝ U /†.1 ˝ U / D U ˝ U D †.U ˝ 1/†.U ˝ 1/;
and the square of the map T 7! Ad.U ˝U / .T / is equal to the identity because
.U ˝ U /2 D idH ˝H . Furthermore, the relation †.1 ˝ U /†.U ˝ 1/ D idH ˝H
shows that the maps T 7! T{ and T 7! Ty are inverse to each other.
Definition 9.3.1. A balanced multiplicative unitary on a Hilbert space H is a pair
.V; U / consisting of a multiplicative unitary V on H and a symmetry U on H such
that the unitaries V{ and Vy defined in equation (9.3) are multiplicative.
The preceding definition is due to Baaj [5].
Remarks 9.3.2. i) In succinct leg notation,
V{ D †UŒ2 V UŒ2 † D UŒ1 †V †UŒ1 D UŒ1 VŒ21 UŒ1 ;
Vy D †UŒ1 V UŒ1 † D UŒ2 †V †UŒ2 D UŒ2 VŒ21 UŒ2 :
ii) Since Vy D Ad.U ˝U / .V{ /, the unitary Vy is multiplicative if and only if the
unitary V{ is multiplicative.
iii) The remarks preceding the definition above imply that iterated applications
of the automorphisms T 7! T{ and T 7! Ty to V yield the operators
{ y
V{
{ D Vy ; V{ y
{ D Ad.U ˝U / .V / D V;
y y
V{ D V:
y

iv) If .V; U / is a balanced multiplicative unitary, then also .V{ ; U /, .Vy ; U /, and
.Ad.U ˝U / .V /; U / are balanced multiplicative unitaries. This follows immediately
from iii).
v) If .V; U / is a balanced multiplicative unitary, then also .V op ; U / is a balanced
multiplicative unitary. Indeed,
1
.V op / D .V{ /op ; Q
.V op / D .Vy /op :
These relations follow from straightforward calculations, for example,
1
.V op / D UŒ2 ††V  ††UŒ2 D UŒ2 V  UŒ2 D ††UŒ2 V  UŒ2 †† D .V{ /op :
 9.3. Weak Kac systems 265

Let .V; U / be a balanced multiplicative unitary and assume that V is

well-behaved (Definition 7.2.6). Then we can associate to V two C  -bialgebras
y /; /
.A.V y and .A.V /; /, called the left and the right leg of V . Now we show that
this left and right leg of V can be identified with the right and the left legs of the
auxiliary unitaries V{ and Vy . Before we make this precise, let us consider the legs
of the unitary V{
{ D Ad.U ˝U / .V / D Vyy:

Lemma 9.3.3. Let V be a well-behaved multiplicative unitary on a Hilbert space

H and let U be a symmetry on H .
y // and AdU .A.V // are C  -bialgebras with re-
i) The C  -algebras AdU .A.V
spect to the comultiplication given by

AdU .a/ y a//

O 7! Ad.U ˝U / .. O and AdU .a/ 7! Ad.U ˝U / ..a//;

y / and a 2 A.V /, respectively.

where aO 2 A.V

ii) The operator V{{ D Ad.U ˝U / .V / D Vy

y is a well-behaved multiplicative uni-

tary, and the following C -bialgebras are isomorphic or equal, respectively:

y / Š AdU .A.V
A.V y V{
y // D A. { /; A.V / Š AdU .A.V // D A.V{
{ /:

Proof. Straightforward. 
Let us turn to the legs of the unitaries V{ and Vy :
Proposition 9.3.4. Let .V; U / be a balanced multiplicative unitary and assume
that V is well-behaved. Then also the multiplicative unitaries V{ and Vy are well-
behaved, and the following C  -bialgebras are equal:
y V{ / D AdU .A.V //; A.V{ / D A.V
A. y /; y Vy / D A.V /; A.Vy / D AdU .A.V
A. y //:

The proof of this proposition uses the following lemma.

Lemma 9.3.5. VŒ13 VŒ23 V{Œ12 D V{Œ12 VŒ13 and VyŒ23 VŒ12 VŒ13 D VŒ13 VyŒ23 .

Proof. Let us prove the first equation. We insert the relation V{ D UŒ1 VŒ21 UŒ1
(see Remark 9.3.2 i)) into the pentagon equation V{Œ12 V{Œ13 V{Œ23 D V{Œ23 V{Œ12 and
obtain

UŒ1 VŒ21 UŒ1  UŒ1 VŒ31 UŒ1  V{Œ23 D V{Œ23  UŒ1 VŒ21 UŒ1 :

Since UŒ1 commutes with V{Œ23 , we can cancel UŒ1 everywhere, and find

VŒ21 VŒ31 V{Œ23 D V{Œ23 VŒ21 :

266 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Now we conjugate both sides of this equation by the automorphism †Œ23 †Œ12 W ˝
 ˝  7!  ˝  ˝  of H ˝ H ˝ H , that is, we renumber the legs of the operators
according to the permutation .1; 2; 3/ 7! .3; 1; 2/, and obtain the first equation
stated in the lemma. The proof of the second equation is similar. 

Proof of Proposition 9.3.4. We only prove the equation A.V{ / D A.V y /; the re-
maining equations can be proved by similar calculations or can be deduced from
the former one using Lemma 9.3.3, Remark 9.3.2 v), and Proposition 7.2.11. De-
note by H the Hilbert space underlying V . Let ! 2 L.H / and put ! 0 WD ! ıAdU .
Then

.! x̋ id/.V{ / D .! x̋ id/.†UŒ2 V UŒ2 †/

D .id x̋ !/.UŒ2 V UŒ2 / D .id x̋ ! 0 /.V /:

Since the map L.H / ! L.H / given by ! 7! ! ı AdU is bijective, it follows

that A0 .V{ / D Ay0 .V /, and hence also A.V{ / D A.V
y /. In particular, A.V{ / is a

C -algebra.
y V on an element of the form
Let us determine the action of the maps V{ and 
aO WD .id x̋ ! /.V /, where ! 2 L.H / . The previous lemma implies
0 0

O D V{ .aO ˝ 1/V{  D .id x̋ id x̋ ! 0 /.V{Œ12 VŒ13 V{Œ12

V{ .a/ 
/
D .id x̋ id x̋ ! 0 /.VŒ13 VŒ23 /;

and the pentagon equation for V implies

y V .a/
 O D V  .1 ˝ a/V
O D .id x̋ id x̋ ! 0 /.VŒ12

VŒ23 VŒ12 /
D .id x̋ id x̋ ! 0 /.VŒ13 VŒ23 /;

O D
that is, V{ .a/ y V .a/.
O Since A0 .V{ / and Ay0 .V / are dense in A.V{ / and A.V
y /,
O D
respectively, it follows that V{ .a/ y V .a/ y / D A.V{ /. Therefore,
O for all aO 2 A.V
{ 
.A.V /; V{ / is a bisimplifiable C -bialgebra, and equal to .A.V y /; 
y V /.
To prove that V{ and Vy are well-behaved, it only remains to show that V{ 2
M.A. y V{ / ˝ A.V{ // and Vy 2 M.A. y Vy / ˝ A.Vy //. But since V 2 M.A.V
y / ˝ A.V //,

V{ D †U2 V U2 † 2 †U2 M.A.V

y / ˝ A.V //U2 †
y // D M.A.
D M.AdU .A.V // ˝ A.V y V{ / ˝ A.V{ //;

and a similar calculation shows that Vy 2 M.A.

y Vy / ˝ A.Vy //. 
 9.3. Weak Kac systems 267

9.3.2 Weak Kac systems

In typical examples of balanced multiplicative unitaries, conjugation by the sym-
metry U facilitates the transition between the left-regular and the right-regular rep-
resentation of the legs of V . A natural condition to impose is that for each leg, these
two representations commute. This condition leads to the following definition:
Definition 9.3.6. We call a balanced multiplicative unitary .V; U / a weak Kac
system if V is well-behaved and VŒ23 VyŒ12 D VyŒ12 VŒ23 and VŒ12 V{Œ23 D V{Œ23 VŒ12 .
Remark 9.3.7. In [182], Vergnioux introduced a notion of a weak Kac system that
includes an additional condition which is needed for the construction of full crossed
products. For our purposes, the weaker form given above is sufficient.
The preceding definition is motivated by the following observation:
Lemma 9.3.8. Let V be a multiplicative unitary on a Hilbert space H and U a
symmetry on H . Then

VŒ23 VyŒ12 D VyŒ12 VŒ23 , ŒA.V

y /; AdU .A.V
y // D 0;
VŒ12 V{Œ23 D V{Œ23 VŒ12 , ŒA.V /; AdU .A.V // D 0:

Proof. We only prove the first equivalence; the second one follows similarly. Since
maps of the form ! x̋ id x̋ ! 0 W L.H ˝H ˝H / ! L.H /, where !; ! 0 2 L.H / ,
separate the elements of L.H ˝ H ˝ H /, the relation VŒ23 VyŒ12 D VyŒ12 VŒ23 holds
if and only if for all !; ! 0 2 L.H / , the operator

.! x̋ id x̋ ! 0 /.VŒ23 VyŒ12 / D .id x̋ ! 0 /.V /  .! x̋ id/.Vy /

is equal to

.! x̋ id x̋ ! 0 /.VyŒ12 VŒ23 / D .! x̋ id/.Vy /  .id x̋ ! 0 /.V /:

By definition and by a similar calculation as in the proof of Proposition 9.3.4,

elements of the form .id x̋ ! 0 /.V / and .! x̋ id/.Vy /, where !; ! 0 2 L.H / , are
y / and AdU .A.V
dense in A.V y //, respectively. Hence, the operators above are equal
0 O D AdU .b/
for all !; ! 2 L.H / if and only if aO AdU .b/ O aO for all a;
O bO 2 A.V
y /. 

The notion of a weak Kac system is symmetric in the following sense:

Proposition 9.3.9. If .V; U / is a weak Kac system, then also .V{ ; U /, .Vy ; U /, and
.V op ; U / are weak Kac systems.
Proof. If .V; U / is a weak Kac system, then
• V{ , Vy , and V op are well-behaved by Proposition 9.3.4 and 7.2.11;
268 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

• .V{ ; U /, .Vy ; U /, and .V op ; U / are balanced multiplicative unitaries by Re-

marks 9.3.2 iv) and v);
• .V{ ; U /, .Vy ; U /, and .V op ; U / satisfy the last condition in Definition 9.3.6 by
Lemma 9.3.8 and Propositions 9.3.4, 7.2.11. 

Definition 9.3.10. Let .V; U / be a weak Kac system. Then .V{ ; U /, .Vy ; U /, and
.V op ; U / are called the predual, the dual, and the opposite weak Kac system of
.V; U /, respectively.

9.3.3 Examples of weak Kac systems

To every locally compact group, to every compact quantum group, and, more gen-
erally, to every locally compact quantum group, one can associate a weak Kac
system:

Example 9.3.11. Let G be a locally compact group with left Haar measure . In
Example 7.1.4, we defined a well-behaved multiplicative unitary WG on the Hilbert
space L2 .G; / via

.WG /.x; y/ D .x; x 1 y/ for all x; y 2 G;  2 L2 .G  G;   /I

here, we identified L2 .G; / ˝ L2 .G; / with L2 .G  G;   /. Now we use the

group inversion on G to construct a symmetry U on L2 .G; / such that .WG ; U /
is a weak Kac system.
Denote by 1 the right Haar measure associated to the left Haar measure ,
and by ı D d =d 1 the modular function of G (see Section 2.2.3). Then, it is
easy to check that the formula

.U /.x/ WD .x 1 /ı.x/1=2 for all x 2 G;  2 L2 .G; /;

defines a symmetry U on L2 .G; / – use equation (2.8) from page 55 and the fact
that ı is a homomorphism.
To prove that .WG ; U / is a balanced multiplicative unitary, we determine the
unitary W yG D †UŒ1 WG UŒ1 †. For all x; y 2 G and  2 L2 .G  G;   /,

yG /.x; y/ D .WG UŒ1 †/.y 1 ; x/  ı.y/1=2

.W
D .UŒ1 †/.y 1 ; yx/  ı.y/1=2
D .yx; y/  ı.y/1=2 ı.y 1 /1=2 D .yx; y/:
op
A comparison with the formula for WG given at the end of Example 7.1.4 shows that
WyG D W op . Therefore, WyG is a multiplicative unitary and .WG ; U / is a balanced
G
multiplicative unitary.
 9.3. Weak Kac systems 269

{G is given by
A similar calculation as above shows that the unitary W
{G /.x; y/ D .xy; y/  ı.y/1=2
.W for all x; y 2 G;  2 L2 .G  G;   /;

{G is equivalent to the multiplicative unitary VG on L2 .G; 1 / (see

and that W
Š
! L2 .G; 1 / given
Example 7.1.4) with respect to the identification L2 .G; / 
by  7! ı .
1=2

Let us check that the pair .WG ; U / forms a weak Kac system. Straightforward
calculations show that for all  2 L2 .G  G  G;     / and x; y; z 2 G,

.WyGŒ12 WGŒ23 /.x; y; z/ D .yx; y; y 1 z/ D .WGŒ23 W

yGŒ12 /.x; y; z/;
{GŒ23 /.x; y; z/ D .x; x 1 yz; z/ı.z/1=2 D .W
.WGŒ12 W {GŒ23 WGŒ12 /.x; y; z/:

y G / and of
For completeness, we also determine the representation AdU of A.W
2
A.WG / on L .G; /. By Example 7.2.13,
y G / D
M .C0 .G//, where
M W C0 .G/ ! L.L2 .G; // denotes the
• A.W
representation via multiplication operators, and
• A.WG / D L.C  .G// D Cr .G/, where L W C  .G/ ! L.L2 .G; // de-
notes the left regular representation.

Simple calculations show that

• AdU ı
M D
M ı , where  denotes the automorphism of C0 .G/ given by
..g//.x/ D g.x 1 / for all x 2 G; g 2 C0 .G/;
• AdU ıL is the right regular representation of C  .G/ on L2 .G; / given by
Z
..AdU ıL/.f //.x/ D f .y/ı.y/1=2 .xy/d .y/
G

for all  2 L2 .G; /, x 2 G, f 2 L1 .G; /.

Example 9.3.12. To every compact quantum group, one can associate a weak Kac
system as follows. Let us start from an algebraic compact quantum group .A0 ; 0 /;
given a C  -algebraic compact quantum group .A; /, simply choose .A0 ; 0 /
to be the associated Hopf -algebra of matrix coefficients of finite-dimensional
corepresentations (see Theorem 5.4.1).
Denote by h0 the Haar state of .A0 ; 0 /, and by H the completion of A0 with
respect to the inner product given by hajbi WD h0 .a b/ for all a; b 2 A0 . In
Example 7.1.6, we constructed a multiplicative unitary V D VA0 on H such that

V .a ˇ b/ D 0 .a/.1 ˇ b/ for all a; b 2 A0  H:

270 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Now we use the antipode S0 of A0 and the characters .fz /z2C introduced in Theo-
rem 3.2.19 to construct a symmetry U on H such that .V; U / is a weak Kac system.
0
Pz 2 C, denote by z;z the automorphism of A0 given by a 7! fz 
Given z; 0

a  fz 0 D fz 0 .a.1/ /a.2/ fz .a.3/ /. We claim that the bijection

U0 W A0 ! A0 ; a 7! 1;0 .S0 .a//;

extends to a symmetry U on H . By Corollary 3.2.20, U0 D 1;0 ı S0 D S0 ı 0;1

and

U0 ı U0 D 1;0 ı S0 ı S0 ı 0;1 D 1;0 ı 1;1 ı 0;1 D idA0 :

It remains to show that h0 .U0 .a/ U0 .b// D h0 .a b/ for all a; b 2 A0 . Using
Corollary 3.2.20 and the relation  ı S0 D S01 ı  from Proposition 1.3.28, we
find

 ı U0 D  ı 1;0 ı S0 D 1;0 ı  ı S0 D 1;0 ı S01 ı  D 0;1 ı S0 ı :

We insert this relation into the expression h0 .U0 .a/ U0 .b//, use Theorem 3.2.19 v),
Corollary 3.2.20, and the identity h0 ı S0 D h0 from Proposition 2.2.6, and get

h0 .U0 .a/ U0 .b// D h0 . 0;1 .S0 .a // 1;0 .S0 .b///

D h0 . 1;0 .S0 .b/S0 .a /// D h0 .S0 .a b// D h0 .a b/:

Consider the unitary Vy D †UŒ1 V UŒ1 †. We claim that Vy is equal to the

multiplicative unitary W D WA0 constructed in Example 7.1.6, whose adjoint is
given by

W  .a ˇ b/ D 0 .b/.a ˇ 1/ for all a; b 2 A0  H:

In particular, this claim implies that .V; U / is a balanced multiplicative unitary. To

prove this claim, we need to compute the composition 0 ıU0 . By Corollary 3.2.20
and Proposition 1.3.12,

0 ı U0 D 0 ı 1;0 ı S0 D .id ˇ 1;0 / ı 0 ı S0

D † ı . 1;0 ˇ id/ ı .S0 ˇ S0 / ı 0 D † ı .U0 ˇ S0 / ı 0 :

The relation W D †UŒ1 V UŒ1 † is equivalent to W  †UŒ1 V D †UŒ1 . The

operator W  †UŒ1 V acts on an element a ˇ b 2 A0 ˇ A0  H ˝ H as follows:

V X †.U0 ˇ1/ X
a ˇ b 7! a.1/ ˇ a.2/ b 7! a.2/ b ˇ U0 .a.1/ /
W X
7! 0 .U0 .a.1/ //.a.2/ b ˇ 1/:
 9.3. Weak Kac systems 271

We insert the formula for 0 ı U0 derived above, use the defining properties of the
antipode and counit (see also Example 1.3.4 ii)), and find that for all a; b 2 A0 ,
X
W  †UŒ1 V .a ˇ b/ D S0 .a.2/ /a.3/ b ˇ U0 .a.1/ /
D b ˇ U0 .a/ D †UŒ1 .a ˇ b/:

Thus Vy D W , and the claim is proved.

The operator V{ D †UŒ2 V UŒ2 † acts on an element a ˇ b 2 A0 ˇ A0 as
follows:
.1ˇU0 /† V X
a ˇ b 7! b ˇ U0 .a/ 7! b.1/ ˇ b.2/ U0 .a/
†.1ˇU0 / X
7! U0 .b.2/ U0 .a// ˇ b.1/ :

Since S0 is an antihomomorphism and 1;0 is a homomorphism, the composition

U0 D 1;0 ı S0 is an antihomomorphism. Hence,
X X
V{ .a ˇ b/ D U02 .a/U0 .b.2/ / ˇ b.1/ D aU0 .b.2/ / ˇ b.1/

for all a; b 2 A0  H .
Straightforward calculations show that the pair .V; U / is a weak Kac system:
X
VyŒ12

VŒ23 .a ˇ b ˇ c/ D b.1/ a ˇ b.2/ ˇ b.3/ c
D VŒ23 VyŒ12

.a ˇ b ˇ c/;
X
VŒ12 V{Œ23 .a ˇ b ˇ c/ D a.1/ ˇ a.2/ bU0 .c.2/ / ˇ c.1/
D V{Œ23 VŒ12 .a ˇ b ˇ c/

for all a; b; c 2 A0  H .
y / and of A.V /
For completeness, let us determine the representation AdU of A.V
on H . From Theorem 7.2.14, we know that
y / D span .Ay0 /, where Ay0 is the dual of A0 defined in Section 2.3, and
• A.V P
W Ay0 ! L.H / is the -homomorphism given by .a/b O D O .2/ /
b.1/ a.b
for all aO 2 Ay0 and b 2 A0  H ;
• A.V / D span
.A0 /, where
W A0 ! L.H / is the -homomorphism given
by
.a/b D ab for all a; b 2 A0 .
The representation AdU ı
is easily computed: since U0 is an antihomomorphism
and U02 D id,

U0
.a/U0 b D U0 .aU0 .b// D bU0 .a/ for all a; b 2 A0 :
272 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Next, we consider the representation AdU ı . Combining the formula for 0 ı U0

derived above with the relations U0 ı S0 D .U0 /2 ı 0;1 D 0;1 ; U0 D S0 ı 0;1 ,
and the identity
X X X
b.1/ ˇ 0;1 .b.2/ / D b.1/ ˇ f1 .b.2/ /b.3/ D 1;0 .b.1/ / ˇ b.2/ ;

we find that for all aO 2 Ay0 and b 2 A0 ,

X
O 0b D
U0 .a/U O 0 .b/.2/ /
U0 .U0 .b/.1/ /a.U
X
D O 0 .b.1/ //
U0 .S0 .b.2/ //a.U
X
D O 0 . 0;1 .b.1/ ///
0;1 .b.2/ /a.S
X
D .aO ı S0 ı 1;1 /.b.1/ /b.2/ :

Example 9.3.13. To every locally compact quantum group, one can associate a
weak Kac system as follows. Let us start from a reduced C  -algebraic quantum
group .A; / with left Haar weight  and right Haar weight ; for a locally compact
quantum group in the setting of von Neumann algebras or a universal C  -algebraic
quantum group, the construction is completely analogous.
Let .H;
; ƒ/ be a GNS-construction for the left Haar weight . In Section 8.3.1,
we introduced a multiplicative unitary WA WD W on H , given by

WA .ƒ.a/ ˝ ƒ.b// D .ƒ ˝ ƒ/..b/.a ˝ 1// for all a; b 2 N

:

Consider the operator U WD  i=8 JI on H , where

•  is the scaling constant of .A; /, see Proposition 8.3.8;
• J is the modular conjugation of  on H , see Theorem 8.2.4;
• I denotes the conjugate-linear isometry on H constructed in Theorem 8.3.4.
We show that .WA ; U / is a weak Kac system.
Before we begin the proof, let us note that I is equal to the modular conjuga-
tion Jy associated to the left Haar weight O on the dual .A;y /
y and the GNS-map
y
ƒ W N
O ! H constructed in Theorem 8.3.12, see [93, Corollary 2.9] or [158,
Corollary 1.13.12].
Let us prove that U is a symmetry. By construction, the maps J and I are
conjugate-linear, isometric, and satisfy J  D J; I  D I; J 2 D idH D I 2 . More-
over, by [93, Corollary 2.12] or [158, Corollary 1.13.15], IJ D JyJ D  i=4 J Jy D
 i=4 JI . Combining these relations, we find that the operator U is linear, isometric,
and satisfies U 2 D  i=4 JIJI D IJ 2 I D idH .
 9.3. Weak Kac systems 273

z
The unitary WA D †UŒ2 WA UŒ2 † can be described in terms of a certain GNS-
map  W N ! H for the right Haar weight of .A; /, defined in [91, Nota-
tion 7.13] and [158, Notation 1.9.12], as follows. By [158, Section 1.13], there
exists a multiplicative unitary VA on H such that

VA ..a/ ˝ .b// D . ˝ /..a/.1 ˝ b// for all a; b 2 N ;

and by [93, Corollary 2.2, Proposition 2.15] or [158, Corollary 1.13.3, Proposi-
tion 1.13.18],

.I ˝ J /WA .I ˝ J / D WA ; .I ˝ I /†WA †.I ˝ I / D VA :

Combining these relations, we find

z
W D †.1 ˝ IJ /W .1 ˝ JI /† D .I ˝ I /†.I ˝ J /W .I ˝ J /†.I ˝ I /
A A A
D .I ˝ I /†WA †.I ˝ I / D VA :

z
In particular, WA is multiplicative and .WA ; U / is a balanced multiplicative unitary.
The unitary W cA D †UŒ1 WA UŒ1 † does not coincide with the multiplicative
unitary of the reduced dual .A; y /,
y which is usually denoted by W y .1 Using [93,
Proposition 2.15] or [158, Proposition 1.13.18], one can show that the opposite
cA /op of W
.W cA is equal to the unitary V y given by
A

y a/
VAy.. O ˝ . O D .
y b// y ˝ /.
y .y a/.1
O O
˝ b// O bO 2 N O I
for all a;

here,  y /.
y W N O ! H is a certain GNS-map for the right Haar weight O of .A; y
 y A / and AdU .A.W
Next, consider the C -algebras A.W y A //. By Proposition 8.3.2,
y A / D
.A/, and by Theorem 8.3.4 iv) and Theorem 8.2.4, the C  -algebra
A.W

y A // D U
.A/U D JI
.A/IJ D J
.R.A// J D J
.A/J
AdU .A.W (9.4)

commutes with
.A/ D A.W y A /.
A similar argument shows that the C  -algebras A.WA / and AdU .A.WA // com-
mute – use the relation A.WA / D Ay (see Section 8.3.3), the symmetry in the
definition of U apparent from the equation U D  i=4 J Jy, and the biduality Theo-
rem 8.3.15.
Summarizing, we find that .WA ; U / is a weak Kac system as claimed.

1 y for the
Here, notation does not match nicely; in the present context, the notation WAy instead of W
y /
multiplicative unitary of .A; y may be preferable.
274 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

9.4 Reduced crossed products and dual coactions

The concept of a weak Kac system introduced in the previous subsection provides
the right framework for the construction of reduced crossed products, which is the
topic of this subsection. We proceed as follows. Throughout this subsection, we
fix a weak Kac system .V; U / on a Hilbert space H and consider coactions of the
y /; /.
right leg .A.V /; / and the left leg .A.V y
Given a coaction .C; ı/ of .A.V /; /, we define the reduced crossed product
y / and construct a dual coaction ıO of .A.V
C Ìı;r A.V y /; / y on C Ìı;r A.V
y /, using the
auxiliary multiplicative unitary V{ introduced in the previous subsection. We show
that ıO turns C Ìı;r A.V
y / into an .A.Vy /; /-C
y 
-algebra and that the assignment
.C; ı/ 7! .C Ìı;r A.V O is functorial.
y /; ı/
Next, we carry over these constructions and results to coactions of .A.Vy /; /.
y
y y
It is tempting to simply exchange .A.V /; / and .A.V /; / in the preceding defi-
nitions, but we have to replace the weak Kac system .V; U / by its dual .Vy ; U / or
predual .V{ ; U /.
Finally, we explain how these constructions for coactions relate to the corre-
sponding constructions for group actions presented in Section 9.1.

9.4.1 The reduced crossed product of a coaction of A.V /

y / of a coaction .C; ı/ of .A.V /; /
We define the reduced crossed product C Ìı;r A.V
 
using the language of C -modules. The C -algebra C of the coaction can be con-
sidered as a C  -module over itself, and the Hilbert space H of the weak Kac
system can be considered as a C  -module over C. Thus we can form a tensor
product C ˝ H , which is a C  -module over C (see Section 12.2).
Definition 9.4.1. Let .V; U / be a weak Kac system on a Hilbert space H and .C; ı/ a
coaction of .A.V /; /. The reduced crossed product of .C; ı/ is the C  -subalgebra
y /  LC .C ˝ H / generated by the subset
C Ìı;r A.V
y //  LC .C ˝ H /:
ı.C /.1 ˝ A.V (9.5)

y / for C Ìı;r A.V

If the coaction ı is understood, we shortly write C Ìr A.V y /.

Remarks 9.4.2. i) In formula (9.5), we use the extension of the non-degenerate

ı
embedding C ˝ A.V / ,! LC .C ˝ H / to define the composition C !  M.C ˝
A.V // ,! LC .C ˝ H /.
y / can also be defined without reference to C  -
i) The C  -algebra C Ìı;r A.V
modules, using a non-degenerate faithful representation of C on some Hilbert space
K: If
W C ! L.K/ is such a representation, then C Ìı;r A.V y / is isomorphic to
y //.
the C  -subalgebra of L.K ˝ H / generated by the set .
˝ id/.ı.C //  .1 ˝ A.V
 9.4. Reduced crossed products and dual coactions 275

To see this, consider K as a C  -module over C. Then the internal tensor product
.C ˝ H / ˝ K is a Hilbert space that is isomorphic to K ˝ H via .c ˝ / ˝ 

.c/˝ (see Section 12.2). Now the map Ind W LC .C ˝H / ! L..C ˝H /˝

K/ Š L.K ˝ H / is an injective -homomorphism (see Section 12.2) that maps
y // onto .
˝ id/.ı.C //  .1 ˝ A.V
ı.C /.1 ˝ A.V y //, and C Ìı;r A.V y / onto the

C -algebra generated by that subset.
Like the crossed product of a group action, the reduced crossed product of a
coaction can be considered as a twisted tensor product. In the present setting, it is
difficult to make this precise, but see the following example:
Example 9.4.3. Consider the weak Kac system .V; U / associated to an algebraic
compact quantum group .A0 ; 0 / (see Example 9.3.12). Recall that A0 embeds in
A.V / as a dense -subalgebra and that the comultiplication  of A.V / extends the
comultiplication 0 of A0 (see Theorem 7.2.14).
Assume that .C; ı/ is a coaction of .A.V /; / and that there exists a dense
-subalgebra C0  C such that ı.C0 / is contained in the algebraic tensor product
C0 ˇ A0 . Denote by ı0 W C0 ! C0 ˇ A0 the restriction of ı to C0 .
The reduced crossed product C Ìı;r A.Vy / can be considered as the completion
of a twisted tensor product C0 Ì Ay0 that is spanned by products of two -algebras
C0 Ì 1 Š ı.C0 / and 1 Ì Ay0 Š Ay0 . Here, .Ay0 ;  y 0 / denotes the dual of .A0 ; 0 /
as usual. Moreover, the twist governing the commutation relations between C0 Ì 1
and 1 Ì Ay0 can be described in terms of ı0 as follows. Recall from Example 9.3.12
that
• the underlying Hilbert space H of .V; U / is the completion of A0 with respect
to the inner product induced by the Haar state;
y / D span .Ay0 /, where W Ay0 ! L.H / is the -homomorphism given
• A.V P
O D b.1/ a.b
by .a/b O .2/ / for all aO 2 Ay0 and b 2 A0  H ;
• A.V / D span
.A0 /, where
W A0 ! L.H / is the -homomorphism given
by
.a/b D ab for all a; b 2 A0 .
y /  LC .C ˝H / is generated by products
By definition, the C  -algebra C Ìı;r A.V
of the form .c Ì 1/.1 Ì a/,O where
c Ì 1 WD .id ˇ
/.ı.c//; 1 Ì aO WD idC ˇ .a/;
O

and c 2 C0 ; aO 2 Ay0 . Let us compute the product

.1 Ì a/.c
O Ì 1/ D .idC ˇ .a//
O  .id ˇ
/.ı.c// 2 LC .C ˝ H /:
We apply this operator to an element d ˇ b 2 C0 ˇ A0  C ˝ H , and obtain
X
.1 Ì a/.c
O Ì 1/.d ˇ b/ D c.0/ d ˇ .a/
.c
O .1/ /b:
276 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Here, we use the extended Sweedler notation for ı0 similarly

P as it was defined in
Notation 3.1.3 for corepresentations, writing ı0 .c/ D c.0/ ˇ c.1/ 2 C0 ˇ Ay0 .
By definition of the representations
; and of the comultiplication of Ay0 ,
X
O
.a/
.c .1/ /b D O .2/ b.2/ /
c.1/ b.1/ a.c
X
D O .2/  //b:

.c.1/ / .a.c

Therefore,
X
.1 Ì a/.c
O Ì 1/.d ˇ b/ D ..c.0/ Ì 1/ .1 Ì a.c
O .1/  ///.d ˇ b/:

y / is the completion of the space

Thus, the reduced crossed product C Ìı;r A.V
y y
C0 Ì A0 D span.C0 Ì 1/.1 Ì A0 /, and this space can be considered as a twisted
tensor product of the -algebras C0 Ì 1 and 1 Ì Ay0 , where the twist is given by
X
.1 Ì a/.c
O Ì 1/ D O .1/  // for all c 2 C0 ; aO 2 Ay0 :
.c.0/ Ì 1/ .1 Ì a.c

In Definition 9.4.1, the reduced crossed product of a coaction was defined in

terms of a generating subset. Frequently, it is useful to know that this subset is
linearly dense in the crossed product:
Proposition 9.4.4. Let .V; U / be a weak Kac system and .C; ı/ a coaction of
y / D span .ı.C /.1 ˝ A.V
.A.V /; /. Then C Ìı;r A.V y ///.

Proof. Let us begin with some preliminaries. We denote by H the Hilbert space
underlying the weak Kac system .V; U /. For every  2 H , we define an operator
jiŒ3 WD idC ˝ idH ˝ji 2 LC .C ˝ H; C ˝ H ˝ H /; c ˝  7! c ˝  ˝ ;
and put hjŒ3 WD jiŒ3 . Moreover, we put jH iŒ3 WD fjiŒ3 j  2 H g, hH jŒ3 WD
jH i , and Ay WD A.V
Œ3
y /; A WD A.V /. By Lemma 7.2.7,
 
1 ˝ Ay D hH jŒ3 .1 ˝ V /jH iŒ3  LC .C ˝ H /:
y
To prove the assertion, it suffices to show that the space Œ.1 ˝ A/ı.C / is con-
y
tained in Œı.C /.1 ˝ A/. By the equation above,
y
Œ.1 ˝ A/ı.C / D ŒhH jŒ3 .1 ˝ V /jH iŒ3 ı.C /
 
D hH jŒ3 .1 ˝ V /.ı.C / ˝ 1/jH iŒ3 :

As V .a ˝1/V  D .a/ for all a 2 A, we have .1˝V /.ı.c/˝1/ D ı .2/ .c/.1˝V /

for all c 2 C , where ı .2/ WD .id ˝/ ı ı D .ı ˝ id/ ı ı. Therefore,
 
y
Œ.1 ˝ A/ı.C / D hH jŒ3 .ı .2/ .C //.1 ˝ V /jH iŒ3 :
 9.4. Reduced crossed products and dual coactions 277

Since A acts non-degenerately on H , we can replace hH jŒ3 by hH jŒ3 AŒ3 . Inserting

the relation
ŒAŒ3 ı .2/ .C / D Œ.ı ˝ id/..1 ˝ A/ı.C //  Œ.ı ˝ id/.C ˝ A/ D Œı.C / ˝ A;
we find
 
y
Œ.1 ˝ A/ı.C /  hH jŒ3 .ı.C / ˝ A/.1 ˝ V /jH iŒ3 :
Now we move AŒ3 to the left, replace hH jŒ3 AŒ3 by hH jŒ3 again, and find
 
y
Œ.1 ˝ A/ı.C /  hH jŒ3 .ı.C / ˝ 1/.1 ˝ V /jH iŒ3 D Œı.C /.1 ˝ A/: y 
Corollary 9.4.5. Let .V; U / be a weak Kac system and let .C; ı/ be a coaction of
.A.V /; /. Then the maps
C ! LC .C ˝ H /; c 7! ı.c/; y / ! LC .C ˝ H /; aO 7! 1 ˝ a;
and A.V O
y / onto non-degenerate C  -subalgebras of M.C Ìı;r A.V
send C and A.V y // and
extend to -homomorphisms
y // and M.A.V
M.C / ! M.C Ìı;r A.V y // ! M.C Ìı;r A.V
y //: 

9.4.2 The dual coaction of a coaction of A.V /

y /; /
Given a coaction .C; ı/ of .A.V /; /, we construct a coaction of .A.V y on the
y
reduced crossed product C Ìı;r A.V /, combining the trivial coaction on ı.C / with
y /:
the right regular coaction on A.V
Theorem 9.4.6. Let .V; U / be a weak Kac system and .C; ı/ a coaction of
.A.V /; /. Then the formula
ı.c/.1 ˝ a/ y a//
O 7! .ı.c/ ˝ 1/.1 ˝ . O

defines a coaction ıy of .A.V

y /; /
y on C Ìı;r A.V
y / that turns C Ìı;r A.V
y / into an
y /; /-C
.A.V y 
-algebra.
Proof. Denote by H the Hilbert space underlying .V; U / and put A WD A.V /,
y /. First, let us show that the map ıO is well defined. We identify
Ay WD A.V
y ˝ Ay  LC .C ˝ H / ˝ L.H /
.C Ìı;r A/
with a non-degenerate C  -subalgebra of LC .C ˝ H ˝ H / and extend this iden-
tification to multiplier algebras. Consider the unitary V{ D †UŒ2 V UŒ2 †. By the
proof of Lemma 9.3.8 and by Proposition 9.3.4,
V{ .a ˝ 1/V{  D a ˝ 1 for all a 2 A; V{ .aO ˝ 1/V{  D .
y a/ y
O for all aO 2 A;
278 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

and therefore,
O
ı.x/ D .1 ˝ V{ /.x ˝ 1/.1 ˝ V{  / y D C Ìı;r A:
for all x 2 Œı.C /.1 ˝ A/ y

It is easy to see that the map ıO is a coaction. Let us prove that it turns C Ìı;r Ay
into an .A; y /-C
y 
-algebra. The equation above shows that ıO is injective, and since
.A/  .1 ˝ A/ is linearly dense in Ay ˝ A,
y y y O Ìı;r A/
y the product ı.C y  .id y
˝A/
C Ìr Ay
is linearly dense in
y A//.1
Œ.ı.C / ˝ 1/.1 ˝ . y y
˝ 1 ˝ A/
D Œ.ı.C / ˝ 1/.1 ˝ Ay ˝ A/
y D .C Ìı;r A/
y ˝ A:
y 

Definition 9.4.7. Let .V; U / be a weak Kac system. The dual of a coaction .C; ı/
of .A.V /; / is the coaction .C Ìı;r A.V O of .A.V
y /; ı/ y /; /
y defined above.

The construction of the dual coaction defines an assignment from the class of all
y /; /-C
coactions of .A.V /; / to the class of all .A.V y 
-algebras. This assignment
extends to a functor:

Theorem 9.4.8. Let .V; U / be a weak Kac system.

i) Let .C; ıC / and .D; ıD / be coactions of .A.V /; / and  W C ! M.D/ a

non-degenerate covariant -homomorphism. Then the formula

ıC .c/.1 ˝ a/
O 7! ıD ..c//.1 ˝ a/
O

defines a non-degenerate -homomorphism

y / ! M.D Ìr A.V
 Ìr id W C Ìr A.V y //

which is covariant with respect to the dual coactions ıOC and ıOD .
O and  7!  Ìr id defines a functor
y /; ı/
ii) The assignment .C; ı/ 7! .C Ìı;r A.V
y /; /-
from the category of coactions of .A.V /; / to the category of .A.V y
C  -algebras.

Proof. i) Denote by H the Hilbert space underlying .V; U /. We consider D as

a C  -module over itself and identify the internal tensor product .C ˝ H / ˝
D
with the C  -module D ˝ H via the map .c ˝ / ˝
d 7! .c/d ˝ . The
-homomorphism

Ind
W LC .C ˝ H / ! LD ..C ˝ H / ˝
D/ Š LD .D ˝ H /;
T 7! T ˝
idD
Ind
.T /;
 9.4. Reduced crossed products and dual coactions 279

y /  LC .C ˝H / as follows:
(see Section 12.2) acts on the C  -subalgebra C Ìr A.V

Ind
.ıC .c/.1 ˝ a//
O D . ˝ id/.ıC .c/.1 ˝ a//
O
D ıD ..c//.1 ˝ a/
O y /:
for all c 2 C; aO 2 A.V

Consequently,  Ìr id D Ind
jC Ìr A.V y / is well defined. This -homomorphism
y / and D Ìr
evidently is covariant with respect to the dual coactions on C Ìr A.V
y
A.V /.
ii) The proof of this statement is straightforward. 

y /
9.4.3 The dual coaction of a coaction of A.V
The constructions presented above carry over to coactions of the left leg .A.Vy /; / y
of a weak Kac system .V; U / as follows. It is not appropriate to simply exchange
y /; /
the C  -bialgebras .A.V /; / and .A.V y in all definitions. Rather, we have to
replace the weak Kac system .V; U / by its predual .V{ ; U / and use the identifications
A.V{ / D A.V
y / and A.y V{ / D AdU .A.V // Š A.V / established in Proposition 9.3.4;
alternatively, we could use the dual .Vy ; U / in a similar way.
Definition 9.4.9. Let .V; U / be a weak Kac system on a Hilbert space H and .C; ı/
y /; /.
a coaction of .A.V y The reduced crossed product of .C; ı/ is the C  -algebra
C Ìı;r A.V /  LC .C ˝ H / generated by the subset

ı.C /.1 ˝ AdU .A.V ///  LC .C ˝ H /:

If the coaction ı is understood, we shortly write C Ìr A.V / for C Ìı;r A.V /.

Theorem 9.4.10. Let .V; U / be a weak Kac system.
y /; /.
i) Let .C; ı/ be a coaction of .A.V y

(a) C Ìı;r A.V / D span .ı.C /.1 ˝ AdU .A.V ////.

(b) The formula

ı.c/.1 ˝ AdU .a// 7! .ı.c/ ˝ 1/  .1 ˝ Ad.U ˝1/ ..a///

defines a coaction ıO of .A.V /; / on C Ìı;r A.V / that turns C Ìı;r A.V /

into an .A.V /; /-C  -algebra.
y /; /
ii) Let .C; ıC / and .D; ıD / be coactions of .A.V y and  W C ! M.D/ a
non-degenerate covariant -homomorphism. Then the formula

ıC .c/.1 ˝ AdU .a// 7! ıD ..c//.1 ˝ AdU .a//

280 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

defines a non-degenerate -homomorphism

 Ìr id W C Ìr A.V / ! M.D Ìr A.V //

which is covariant with respect to the coactions ıOC and ıOD .

O and  7!  Ìr id defines a functor
iii) The assignment .C; ı/ 7! .C Ìı;r A.V /; ı/
y /; /
from the category of coactions of .A.V y to the category of .A.V /; /-
C  -algebras.
Proof. All assertions follow easily from Proposition 9.4.4, Theorem 9.4.6, and The-
orem 9.4.8, if one replaces the weak Kac system .V; U / by its predual .V{ ; U / and
uses the identifications of the legs of these weak Kac systems listed in Proposi-
tion 9.3.4. 
Definition 9.4.11. Let .V; U / be a weak Kac system. The dual of a coaction .C; ı/
y /; /
of .A.V y is the coaction .C Ìı;r A.V /; ı/O of .A.V /; / defined above.

9.4.4 Comparison with the reduced crossed product of a group

action
The reduced crossed product and the dual coaction introduced above are related to
the corresponding constructions for a group action as follows.
Let G be a locally compact group. By Theorem 9.2.4, every action .C; ˛/ of G
corresponds to a unique coaction .C; ı/ of the C  -bialgebra C0 .G/. The reduced
crossed product C Ì˛;r G for an action .C; ˛/ was defined in Definition 9.1.5.
To construct the reduced crossed product for the corresponding coaction .C; ı/,
we use the weak Kac system .WG ; U / introduced in Example 9.3.11. Since the
legs of this weak Kac system are given by A.W y G / D
M .C0 .G// Š C0 .G/ and

A.WG / D Cr .G/, we can consider ı as a coaction of A.W y G / and form the reduced

crossed product C Ìı;r Cr .G/ as in Definition 9.4.9. We shall see that the crossed
products C Ì˛;r G and C Ìı;r Cr .G/ are naturally isomorphic.
If the group G is abelian, we can also compare the dual action and the dual coac-
tion on the respective crossed products: In this case, the C  -bialgebra Cr .G/ is iso-
morphic to C0 .G/y (Proposition 4.2.3), so that the dual coaction of Cr .G/ Š C0 .G/ y
 y
on C Ìı;r Cr .G/ corresponds to an action of the dual group G (Theorem 9.2.4).
We will see that the isomorphism C Ìı;r Cr .G/ Š C Ì˛;r G referred to above
y on C Ì˛;r G.
identifies this action with the dual action of G

In the next theorem, we identify the C -bialgebra C0 .G/ with A.W y G / via the
isomorphism
M .
Theorem 9.4.12. Let ˛ be an action of a locally compact group G on a C  -alge-
bra C . Denote by ı W C ! M.C ˝C0 .G// the coaction of C0 .G/ that corresponds
to the action ˛.
 9.4. Reduced crossed products and dual coactions 281
id ˝U
i) Conjugation by the isomorphism C ˝ L2 .G; / ! C ˝ L2 .G; / Š
L2 .GI C / induces an isomorphism
Š
C Ìı;r Cr .G/ 
! C Ì˛;r G; ı.c/.1 ˝ AdU .g// 7! c Ìr g: (9.6)

ii) Assume that G is abelian. With respect to the isomorphism (9.6), the dual
y on C Ì˛;r G corresponds to the dual coaction of the C  -bialgebra
action of G
y on C Ìı;r Cr .G/.
Cr .G/ Š C0 .G/
Proof. i) Denote the canonical isomorphism C ˝ L2 .G; / Š L2 .GI C / by ‡ . By
Proposition 9.1.6 and 9.4.4,
C Ìı;r Cr .G/ D Œı.C /.1 ˝ Cr .G//; C Ì˛;r G D Œ.C Ìr 1/.1 Ìr Cr .G//:
Therefore it suffices to show that for all c 2 C and g 2 Cr .G/,
Ad‡.id ˝U / .ı.c// D c Ìr 1; Ad‡.id ˝U / .1 ˝ AdU .g// D 1 Ìr g: (9.7)
The second equation is evident; let us prove the first one. The final remarks in
Example 9.3.11 imply
Ad.id ˝U / .ı.c// D .id ˝.
M ı  //.ı.c//;
where  denotes the -automorphism of C0 .G/ given by .f /.x/ D f .x 1 / for
all f 2 C0 .G/; x 2 G. Let h 2 L2 .GI C /; x 2 G. Then
.Ad‡.id ˝U /.ı.c//h/.x/ D .id ˝ evx ı /.ı.c//  h.x/ D .id ˝ evx 1 /.ı.c//  h.x/;
and by Theorem 9.2.4, .Ad‡.id ˝U / .ı.c//h/.x/ D ˛x 1 .c/h.x/. Comparing with
the definition of c Ìr 1 (see Proposition 9.1.4), we obtain the first equation in (9.7).

ii) First, note that Cr .G/ D C  .G/ because G is abelian. Let us identify C  .G/
with C0 .G/y as in Proposition 4.2.3. Denote by ˇ the action of G y on C Ìı;r C  .G/
that corresponds to the dual coaction ıO of C .G/ Š C0 .G/.
 y Let  2 G. y By
O
Theorem 9.2.4, ˇ D .id ˝ ev / ı ı. The extension of this automorphism to the
multiplier algebra acts as follows:
ıO
ˇ W ı.c/.1 ˝ AdU .Ux // 7! .ı.c/.1 ˝ AdU .Ux /// ˝ Ux
id ˝ ev
7! ı.c/.1 ˝ AdU .Ux //.x/ for all c 2 C; x 2 G:
On the other hand, the extension of the automorphism ˛O  to the multiplier algebra
M.C Ì˛ G/ is given by
˛O  W c Ì Ux 7! c Ì Ux .x/ for all c 2 C; x 2 G:
Comparing the formulas above, we find that with respect to the isomorphism (9.6),
the action ˇ and the coaction ıO correspond to the action ˛.
O 
282 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

9.5 Kac systems and the Baaj–Skandalis duality theorem

We now come to the main result of this chapter – the Baaj–Skandalis duality theorem.
In the preceding section, we associated to every coaction of the right leg of a weak
Kac system .V; U / a reduced crossed product and a dual coaction of the left leg
of .V; U /. Conversely, to every coaction of the left leg, we associated a reduced
crossed product and a dual coaction of the right leg. Summarizing, we constructed
a pair of functors as follows:

category of .C;ı/ 7! .C Ìr A.V O

y /; ı/ category of
/
coactions of .A.V /; /
7!
Ìr id y y
.A.V /; /-algebras
S T

category of O
.C Ìr A.V /; ı/  .C;ı/ category of
o
.A.V /; /-algebras
Ìr id 
y /; /:
coactions of .A.V y

Naturally, we would like to know how these functors are related. The Baaj–
Skandalis theorem says that under favorable conditions – if .V; U / is a Kac system –
these two functors induce an equivalence of the category of .A.V /; /-C  -algebras
y /; /-C
with the category of .A.V y 
-algebras up to equivariant Morita equivalence.
To prove this assertion, one studies the composition of these two functors, that is,
the iteration of the reduced crossed product construction.
We proceed as follows. First, we introduce the notion of a Kac system and
consider some examples. Then, we determine the composition of the two functors
above. We do not discuss equivariant Morita equivalence; for related concepts, see
[6].

9.5.1 Kac systems

A Kac system is a special balanced multiplicative unitary. Recall that to every
balanced multiplicative unitary .V; U /, we associated two auxiliary multiplicative
unitaries

V{ D †.1 ˝ U /V .1 ˝ U /† and Vy D †.U ˝ 1/V .U ˝ 1/†:

Definition 9.5.1. A balanced multiplicative unitary .V; U / is called a Kac system

if
i) the multiplicative unitaries V , V{ , Vy are regular, and
ii) .†.1 ˝ U /V /3 D 1.
Remarks 9.5.2. i) The unitary Vy is regular if and only if V{ is regular because
Vy D Ad.U ˝U / .V{ /.
 9.5. Kac systems and the Baaj–Skandalis duality theorem 283

ii) In leg notation, the equation .†.1˝U /V /3 D 1 takes the form .†UŒ2 V /3 D 1.
Conjugating by † or V , we find that this condition is equivalent to the relation
.UŒ2 V †/3 D 1 and to the relation .V †UŒ2 /3 D 1.
iii) A multiplicative unitary V on a Hilbert space H is called irreducible [7,
Définition 6.2] if there exists a symmetry U on H such that .V; U / is a balanced
multiplicative unitary and .†.1 ˝ U /V /3 D 1.
Condition ii) in Definition 9.5.1 may appear mysterious at first sight. However,
we shall see in (the proof of) Proposition 9.5.3 and 9.5.5 that this condition is closely
related to the Duality Theorem 9.5.11.
Proposition 9.5.3. Every Kac system is a weak Kac system.
The proof of this result uses the following lemma.
Lemma 9.5.4. Let .V; U / be a balanced multiplicative unitary.
i) ŒVŒ23 ; VŒ12 V{Œ12 †Œ12  D 0 and ŒVŒ12 ; †Œ23 VyŒ23 VŒ23  D 0.

ii) .†UŒ2 V /3 D 1 if and only if Vy V V{ D UŒ1 †.

Proof. i) We only prove the first equation, the second one follows similarly. By
Lemma 9.3.5, V{Œ12 VŒ13 D VŒ13 VŒ23 V{Œ12 . We multiply this equation by VŒ12 on
the left and by †Œ12 on the right, use the pentagon equation for V , and obtain

VŒ12 V{Œ12 †Œ12 VŒ23 D VŒ12 V{Œ12 VŒ13 †Œ12

D VŒ12 VŒ13 VŒ23 V{Œ12 †Œ12 D VŒ23 VŒ12 V{Œ12 †Œ12 :

ii) Rearranging the factors in the product UŒ1 UŒ2 .†UŒ2 V /3 UŒ2 †, we find

UŒ1 UŒ2 .†UŒ2 V /3 UŒ2 † D †UŒ1 V UŒ1 †  V  †UŒ2 V UŒ2 † D Vy  V  V{ :

Thus, .†UŒ2 V /3 is equal to 1 if and only if UŒ1 † D UŒ1 UŒ2 UŒ2 † is equal to
Vy V V{ . 
Proof of Proposition 9.5.3. By part i) of the previous lemma, VŒ23 commutes with
VŒ12 V{Œ12 †Œ12 , and by part ii), VŒ12 V{Œ12 †Œ12 D VyŒ12

UŒ1 . Hence VŒ23 commutes
y  y y
with V , and since VŒ12 is unitary, also with VŒ12 . A similar argument shows that
Œ12
VŒ12 commutes with V{Œ23

and hence also with V{Œ23 . 
The second main property of a Kac system – in addition to being a weak Kac
system – is stated in the following proposition:
Proposition 9.5.5. Let .V; U / be a Kac system on a Hilbert space H . Then
y / D K.H /.
span A.V /A.V
284 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Proof. We combine the leg notation and ket-bra notation as in Sections 7.2 and 7.3;
see formulas (7.6)–(7.8) on page 176. By conditions i) and ii) of Definition 9.5.1,
 
K.H / D hH jŒ2 V  jH iŒ1 and V  D †UŒ2 V †UŒ2 V †UŒ2 :
We insert the second equation into the first one, use U K.H /U D K.H /, and
obtain
   
K.H / D hH jŒ2 †UŒ2 V †UŒ2 V †UŒ2 jH iŒ1 D hH jŒ1 V †UŒ2 V jH iŒ2 :
y /. On the left-hand side we ob-
We multiply this equation on the right by A.V
y y
tain ŒK.H /A.V / D K.H / because A.V / acts non-degenerately on H . On the
y / D ŒjH iŒ2 A.V
right-hand side, ŒV jH iŒ2 A.V y / by Lemma 7.3.13. Hence, by
Lemma 7.2.7,
y /
K.H / D ŒhH jŒ1 V †UŒ2 jH iŒ2 A.V
y / D ŒA.V /A.V
D ŒhH jŒ1 V jH iŒ1 A.V y /: 
The preceding proposition has the following partial converse:
Proposition 9.5.6. Let V be a multiplicative unitary on a Hilbert space H and U
a symmetry on H such that VŒ23 VyŒ12 D VyŒ12 VŒ23 and VŒ12 V{Œ23 D V{Œ23 VŒ12 .
y /A.V / D K.H /, then V is regular.
i) If span A.V
ii) If Vy ; V{ are multiplicative and the commutants of A.V /[ A.V
y / and of A.V /[
y y {
AdU .A.V // both are only scalars, then UŒ2 †V V V is scalar.
Proof. The proof uses similar techniques as presented in this section and in Sec-
tions 9.3 and 7.3; for details, see [7, Proposition 6.9]. 
We end this section with some examples and standard constructions:
Example 9.5.7. For every locally compact group G, the weak Kac system .WG ; U /
constructed in Example 9.3.11 is a Kac system. Indeed, the multiplicative unitaries
WG and W yG D .WG /op are regular by Example 7.3.4 iii) and ii), and from the
calculations in Example 9.3.11, we find
.W {G /.x; y/ D .WG W
yG WG W {G /.yx; y/
{G /.yx; .yx/1 y / D .yx  x 1 ; x 1 /  ı.x 1 /1=2
D .W
„ ƒ‚ … „ ƒ‚ …
Dx 1 Dy
and
.UŒ1 †/.x; y/ D .†/.x 1 ; y/  ı.x/1=2 D .y; x 1 /  ı.x 1 /1=2
yG WG W
for all  2 L2 .G  G;   / and x; y 2 G. Thus W {G D UŒ1 †, and by
Lemma 9.5.4, .WG ; U / is a Kac system.
 9.5. Kac systems and the Baaj–Skandalis duality theorem 285

Example 9.5.8. For every algebraic compact quantum group .A0 ; 0 /, the weak
Kac system .V; U / constructed in Example 9.3.12 is a Kac system. Let us prove
this claim. The multiplicative unitaries V D VA0 and Vy D WA0 are regular
by Example 7.3.4 iv). We show that V V{ D Vy  UŒ1 †, and by Lemma 9.5.4,
this equation implies that .V; U / is a Kac system. By Proposition 1.3.12 and
Corollary 3.2.20,

0 ı U0 D 0 ı S0 ı 0;1 D † ı .S0 ˇ S0 / ı 0 ı 0;1

D † ı .S0 ˇ S0 / ı . 1;1 ˇ 0;1 / ı 0 D † ı .S01 ˇ U0 / ı 0 :

Combining this equation with the calculations in Example 9.3.12, we find

X X
V V{ .a ˇ b/ D V .aU0 .b.2/ / ˇ b.1/ / D a.1/ U0 .b.3/ / ˇ a.2/ S01 .b.2/ /b.1/

for all a; b 2 A0 . Since S01 is the antipode of .A0 ; 0 /op (see Proposition 1.3.14),
we can simplify this expression and find that for all a; b 2 A0 ,
X
V V{ .a ˇ b/ D a.1/ U0 .b/ ˇ a.2/ D Vy  .U0 .b/ ˇ a/ D Vy  UŒ1 †.a ˇ b/:

Like the notion of a weak Kac system, the notion of a Kac system is highly
symmetric:
Proposition 9.5.9. If .V; U / is a Kac system, then also .V{ ; U /, .Vy ; U /, and
.V op ; U / are Kac systems.
Proof. If .V; U / is a Kac system, then
• .V{ ; U / and .Vy ; U / are balanced multiplicative unitaries by Remark 9.3.2 iv);
• the multiplicative unitaries V{ ; Vy , and V{
{ D Ad U ˝U .V / D Vy
y are regular by
assumption;
• .V{ †UŒ2 /3 D .†UŒ2 V /3 D 1 and .†UŒ2 Vy /3 D .V †UŒ2 /3 D 1 by Re-
mark 9.5.2 ii).
Remark 9.5.2 ii) implies that .V{ ; U / and .Vy ; U / are Kac systems. Moreover,
• .V op ; U / is a balanced multiplicative unitary by Remark 9.3.2 v);
z b
• V op and V op are equal to Vy op and V{ op , respectively, by the same remark, and
by Example 7.3.4 ii), the latter unitaries and V op are regular;
• .UŒ2 V op †/3 D .UŒ2 †V  /3 D ..V †UŒ2 / /3 D 1 by Remark 9.5.2 ii).
Again, Remark 9.5.2 ii) implies that .V op ; U / is a Kac system. 
Definition 9.5.10. Let .V; U / be a Kac system. Then .V{ ; U /, .Vy ; U /, and .V op ; U /
are called the predual, the dual, and the opposite Kac system of .V; U /, respectively.
286 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

9.5.2 The Baaj–Skandalis duality theorem

The Baaj–Skandalis duality theorem is the main result of this chapter. Its content
was described already in the introduction to this section. Given the results collected
in the previous subsections, the proof is not difficult.
Our formulation of the duality theorem involves the following notation. Given
a Hilbert space H and a coaction .C; ı/ of a C  -bialgebra .A; /, we define a
coaction ıŒ13 of .A; / on C ˝ K.H / by the formula

ıŒ13 .c ˝ T / D Ad.1˝†/.ı.c/ ˝ idH /  .1 ˝ T ˝ 1/ for all c 2 C; T 2 K.H /:

Here, Ad.1˝†/ W M.C ˝ A ˝ K.H // ! M.C ˝ K.H / ˝ A/ denotes the isomor-

phism induced by c ˝ a ˝ T 7! c ˝ T ˝ a.

Theorem 9.5.11 (Baaj–Skandalis duality). Let .V; U / be a Kac system on a Hilbert

space H .

i) Let .C; ı/ be an .A.V /; /-C  -algebra and ıO the dual coaction of .A.V
y /; /
y
y
on C Ìı;r A.V /. There exists a natural isomorphism

y / Ì O A.V / Š C ˝ K.H /
C Ìı;r A.V ı;r

y / Ì O A.V /
which identifies the bidual coaction of .A.V /; / on C Ìı;r A.V ı;r
with the map Ad.1˝V / ııŒ13 W C ˝ K.H / ! M.C ˝ K.H / ˝ A.V //.

y /; /-C
ii) Let .C; ı/ be an .A.V y 
-algebra and ıO the dual coaction of .A.V /; /
on C Ìı;r A.V /. There exists a natural isomorphism

C Ìı;r A.V / Ìı;r y

O A.V / Š C ˝ K.H /

y /; /
which identifies the bidual coaction of .A.V y on C Ì A.V / Ì O A.V
y /
ı;r ı;r
with the map Ad.1˝V{ / ııŒ13 W C ˝ K.H / ! M.C ˝ K.H //.

Proof. i) To simplify notation, put Ay WD A.V y / and A WD A.V /. By Defini-

tions 9.4.1, 9.4.7, 9.4.9, Proposition 9.4.4, and Theorem 9.4.10 i)(a), the iterated
crossed product C Ìr Ay Ìr A is equal to
y A//.1
Œ.ı.C / ˝ 1/.1 ˝ . y ˝ 1 ˝ AdU .A//  LC .C ˝ H ˝ H /:

y and by the proof of Lemma 9.3.8, conjugation by 1 ˝ V

By definition of  and 

maps this C -algebra isomorphically onto
y ˝ 1 ˝ AdU .A//;
Œı .2/ .C /.1 ˝ 1 ˝ A/.1
 9.5. Kac systems and the Baaj–Skandalis duality theorem 287

where ı .2/ D .id ˝/ ı ı D .ı ˝ id/ ı ı. Since ı is injective, the formula

ı .2/ .c/.1 ˝ 1 ˝ T / 7! ı.c/.1 ˝ T /; where c 2 C; T 2 L.H /;

defines an isomorphism of this C  -algebra onto

Œı.C /.1 ˝ Ay  AdU .A//  LC .C ˝ H /:

Now ŒAy  AdU .A/ D ŒA.V{ /  A.

y V{ / by Proposition 9.3.4, and since .V{ ; U / is a
Kac system (Proposition 9.5.9), ŒA.V{ /  A.
y V{ / D K.H / (Proposition 9.5.5). Thus

C Ìr Ay Ìr A Š Œı.C /.1 ˝ K.H //:

We make use of the fact that A  L.H / is non-degenerate and the assumption
Œı.C /.1 ˝ A/ D C ˝ A, and find

C Ìr Ay Ìr A Š Œı.C /.1 ˝ A  K.H //

D Œ.C ˝ A/.1 ˝ K.H // D C ˝ K.H /:

This isomorphism identifies an element of the form

y a//.1
.ı.c/ ˝ 1/.1 ˝ . O ˝ 1 ˝ AdU .a// 2 C Ìr Ay Ìr A (9.8)
with
ı.c/.1 ˝ aO  AdU .a// 2 C ˝ K.H /: (9.9)

The bidual coaction of .A; / on C Ìr Ay Ìr A maps the element (9.8) to

y a/
.ı.c/ ˝ 1 ˝ 1/.1 ˝ . O ˝ 1/.1 ˝ 1 ˝ Ad.U ˝1/ ..a///:

Put W WD UŒ1 V UŒ1 D †Vy †. The map Ad.1˝W / ııŒ13 sends the element (9.9)
to

.1 ˝ W /.Ad.1˝†/.ı .2/ .c//  .1 ˝ aO  AdU .a/ ˝ 1//.1 ˝ W  /

D .ı.c/ ˝ 1/.1 ˝ aO ˝ 1/.1 ˝ Ad.U ˝1/ ..a///:

In this equation we used the following relations:

• Ad†Vy † .Ad† ..b/// D b ˝ 1 for all b 2 A by Proposition 9.3.4;

• UŒ1 V UŒ1 and Ay ˝ 1 commute by the proof of Lemma 9.3.8;

• AdUŒ1 V UŒ1 .AdU .a/ ˝ 1/ D Ad.U ˝1/ ..a// by definition of .

288 Chapter 9. Coactions on C  -algebras, reduced crossed products, and duality

Therefore, the bidual coaction of .A; / on C Ìr Ay Ìr A corresponds to the map

Ad.1˝W / ııŒ13 . With respect to the isomorphism
Ad
.1˝U /
C Ìr Ay Ìr A Š C ˝ K.H / ! C ˝ K.H /;

the bidual coaction corresponds to Ad.1˝V / ııŒ13 , because

Ad.1˝U ˝1/ ı Ad.1˝W / ııŒ13 ı Ad.1˝U / D Ad.1˝V / ııŒ13 :

ii) The proof is similar as in i), simply replace .V; U / by the predual .V{ ; U /. 
Chapter 10
Pseudo-multiplicative unitaries on Hilbert spaces

In parts I and II of this book, we discussed several approaches to quantum groups.

Most of these approaches can be extended to quantum groupoids, which simul-
taneously generalize quantum groups on one side and classical groupoids on the
other side. The next two chapters provide an introduction to some techniques and
basic concepts related to quantum groupoids in the setting of von Neumann algebras
and C  -algebras.
First steps towards a theory of quantum groupoids in the setting of von Neumann
algebras were taken by Vallin and Enock, who generalized the concept of a von
Neumann bialgebra and of a multiplicative unitary [50], [51], [171], [172]. Building
on their work and on the theory of locally compact quantum groups of Kustermans
and Vaes [91], Lesieur developed a general theory of measurable quantum groupoids
[99]. A main feature of that theory is a generalization of Pontrjagin duality: to
every measurable quantum groupoid, Lesieur associates a dual such that the bidual
is naturally isomorphic to the initial quantum groupoid.
A major motivation for the introduction of quantum groupoids in the theory of
von Neumann algebras comes from the theory of subfactors. A central problem,
raised by Ocneanu, is to describe the symmetries of a given subfactor, that is, of
an inclusion of factors, in terms of an associated generalized Galois group and to
formulate a Galois theory for inclusions of factors [114], [115]. Let us briefly
explain the problem. Given a factor N1 with a suitable action of a group G, we
obtain two inclusions of factors N0 ,! N1 ,! N2 , where N0 WD N1G is the fixed
point algebra and N2 WD N1 Ì G is the crossed product. The question is whether
each inclusion of factors N0 ,! N1 can be described in a similar way. Building
on earlier results by various authors [30], [101], [146], Enock and Nest showed in
[45] that for every irreducible depth 2 inclusion N0 ,! N1 , there exists a locally
compact quantum group M and an action of M on N1 such that N0 is equal to the
fixed point algebra N1M and the inclusion N1 ,! N2 given by the basic construction
[72] is isomorphic to the inclusion N1 ,! N1 Ì M . If the irreducibility assumption
is dropped, then M is no longer a quantum group but a quantum groupoid [43],
[50], [51], [99]. Thus, quantum groupoids arise naturally in the study of subfactors.
Like the theory of locally compact quantum groups (or even more so), the theory
of measurable quantum groupoids is technically demanding and involved. Let us
briefly outline the main concepts and their rôle in the theory. For background on
groupoids, see Section 10.3.3.
290 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Hopf–von Neumann bimodules. A Hopf–von Neumann bimodule is a von Neu-

mann algebra M equipped with a comultiplication that is modeled on the multipli-
cation map of a groupoid.
Recall that a groupoid consists of a set of arrows G, a set of units G 0 , a range and a
source map r; s W G ! G 0 , and a multiplication G s  r G ! G – the composition
G0
of arrows. The multiplication is only defined on the fiber product G s  r G D
G0
f.x; y/ 2 G  G j s.x/ D r.y/g  G  G because two arrows can be composed
only if the source of the left one matches up with the range of the right one.
Likewise, a Hopf–von Neumann bimodule consists of a von Neumann algebra
M , a von Neumann algebra N with a representation and an antirepresentation
r; s W N ! M , and a comultiplication  W M ! M s  r M . Here, the target of
N
the comultiplication is no longer the ordinary tensor product M x̋ M as in the case
of a von Neumann bialgebra, but a fiber product of von Neumann algebras that is
defined in terms of a relative tensor product of Hilbert modules. The notion of a
Hopf–von Neumann bimodule was introduced by Vallin [171].
Operator-valued weights. Similar to a locally compact quantum group in the
setting of von Neumann algebras, a measured quantum groupoid is a Hopf–von
Neumann bimodule equipped with a left and a right Haar weight. Recall that for
a groupoid G, the proper analogue of a Haar measure is a Haar system, which
is a family of measures indexed by the elements of G 0 . Integration with respect
to such a Haar measure does not define a linear functional Cc .G/ ! C but a
linear map Cc .G/ ! Cc .G 0 /. Likewise, the Haar weights of a measured quantum
groupoid .N; M; r; s; / are operator-valued weights that take values in the base N .
Operator-valued weights on von Neumann algebras were introduced by Haagerup
[59], [60]; further references are [144], [150].
Pseudo-multiplicative unitaries. The preceding chapters showed that multiplica-
tive unitaries play a central rôle in the theory of locally compact quantum groups.
In the theory of measurable quantum groupoids, this rôle is taken by pseudo-
multiplicative unitaries. These unitaries do not act on an ordinary tensor product
of Hilbert spaces but on a relative tensor product of Hilbert modules. The concept
of a pseudo-multiplicative unitary was introduced by Vallin [172].
The relative tensor product of Hilbert modules. The relative tensor product
of Hilbert modules is an analogue of the tensor product of modules over a ring,
where the modules are Hilbert spaces, the ring is a von Neumann algebra, and the
relative tensor product is a Hilbert space again. If the underlying von Neumann
algebra is commutative, the relative tensor product corresponds to the fiberwise
tensor product of fields of Hilbert spaces. The construction in the general case is
much more involved; it was introduced by Connes and is frequently called Connes’
fusion.
 10.1. The relative tensor product of Hilbert modules 291

We shall not attempt to give an overview of the theory of measured quantum

groupoids, but provide an introduction to some of the concepts listed above. First,
we give a fairly detailed introduction to the relative tensor product of Hilbert mod-
ules (Section 10.1). This construction is then used to define the fiber product of von
Neumann algebras and Hopf–von Neumann bimodules (Section 10.2). Finally, we
discuss pseudo-multiplicative unitaries on Hilbert spaces (Section 10.3) and study
examples related to groupoids (Section 10.3.3).

10.1 The relative tensor product of Hilbert modules

The relative tensor product of Hilbert modules is an analogue of the tensor product
of modules over a ring and of the internal tensor product of C  -modules over a
C  -algebra. Roughly, this construction starts from a von Neumann algebra N
and Hilbert spaces H and K with an antirepresentation and a representation of N ,
respectively, and produces a new Hilbert space H ˝ K by factoring out the actions
N
of N on H and K. The precise construction is fairly involved.
Before we turn to the details, let us add some bibliographical remarks. For
Hilbert spaces over a commutative von Neumann algebra, the relative tensor prod-
uct was defined by Sauvageout. Connes generalized the construction to the non-
commutative case; unfortunately, his manuscript was never published. Original
references for the construction are [27], [138]; alternative references are [52] and
[150, Section IX.3].
Throughout this section, let N be a von Neumann algebra.

10.1.1 Hilbert modules over von Neumann algebras

We adopt the following terminology:

Definition 10.1.1. A left/right Hilbert N -module is a Hilbert space H equipped

with a non-degenerate normal representation/antirepresentation of N . We write
x / x for the action of an element x 2 N on a vector  2 H .
Let N1 and N2 be von Neumann algebras. A Hilbert N1 -N2 -bimodule is a
Hilbert space H equipped with the structure of a left Hilbert N1 -module and a right
Hilbert N2 -module such that .x1 /x2 D x1 .x2 / for all x1 2 N1 ; x2 2 N2 ,  2 H .
A morphism of left Hilbert N -modules H1 and H2 is an operator T 2 L.H1 ; H2 /
that satisfies T x D xT  for all x 2 N and  2 H1 . We denote the set of all
such morphisms by LN .H1 ; H2 /. Similarly, we define morphisms of right Hilbert
N -modules and of Hilbert N1 -N2 -bimodules.

We denote by N op the opposite von Neumann algebra of N and by N !

N op ; x 7! x op , the canonical antiautomorphism (see Section 12.3). Clearly, left
292 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Hilbert N -modules correspond bijectively with right Hilbert N op -modules, and

right Hilbert N -modules correspond bijectively with left Hilbert N op -modules.
The fundamental example of a Hilbert N -N -bimodule is the GNS-space asso-
ciated to an n.s.f. weight. This bimodule plays a central rôle in the construction of
the relative tensor product.
Example 10.1.2. Let be a normal semi-finite faithful weight on N with GNS-
construction .H ; ƒ ;
/. Then the space H carries the following structure of
a Hilbert N -N -bimodule:
The representation
turns H into a left Hilbert N -module. Using the mod-
ular conjugation J for (see Theorem 8.2.4), we can also define an antirepresen-
tation
N ! L.H /; z 7! J
.z/ J : (10.1)
This map is linear because J is conjugate-linear, an antihomomorphism because
J 2 D idH , and normal and non-degenerate because
is normal and non-
degenerate. Furthermore, this antirepresentation commutes with
(see Theo-
rem 8.2.4). Thus, H becomes a Hilbert N -N -bimodule.
While
amounts to left multiplication, the antirepresentation (10.1) amounts
to right multiplication up to a twist by the modular automorphism group  of :

.x/ƒ .y/ D ƒ .xy/ and J
.z/ J ƒ .y/ D ƒ .yi=2 .z//

for all x 2 N; y 2 N , and z 2 Dom.i=2 / (see Theorem 8.2.4 vi)).

The preceding construction yields two Hilbert N op -N op -bimodules:
• we can consider H as a Hilbert N op -N op -bimodule via x op y op WD yx for
all x; y 2 N and  2 H , or
• we can define an opposite weight op on N op by op .x op / D .x/ for all
positive elements x 2 N , and consider the associated N op -N op -bimodule
H op .
It turns out that these two Hilbert N op -N op -bimodules are isomorphic: the relation
op
..y op / y op / D .yy  /; y 2 N , implies
N op D fy op j y  2 N g D .N  op
/ ;
and straightforward calculations show that the space H and the maps
ƒ op W N op ! H ; 7 J ƒ .y  /;
y op !

op W N op ! L.H /; x op !7 J
.x  /J ;
form a GNS-construction for op . Evidently, the representation
op corresponds
to the antirepresentation (10.1), and by symmetry, it follows that the representation
 10.1. The relative tensor product of Hilbert modules 293

corresponds to the antirepresentation N op ! L.H op / given by x op 7! J op 

op .x op /  J op .

The category of Hilbert modules over a von Neumann algebra has sufficiently
many morphisms:

Proposition 10.1.3. Let be an n.s.f. weight on N . Consider the associated GNS-

space H as a right Hilbert N -module via the antirepresentation (10.1). For every
right Hilbert N -module H and every element  2 H , there exists a partial isometry
u 2 LN .H ; H /, such that uu  D .

Proof. Given  2 H , consider the positive normal linear functional ! on N given by

!.y/ WD hjyi for all y 2 N . By [58, Lemma 2.10] or [150, IX, Theorem 1.2 iv)],
there exists a vector  2 H such that !.y/ D hjyi for all y 2 N . By the choice
of , the map u W H ! H given by u.y/ WD y for all y 2 N and u./ WD 0 for
all  2 .N /? is a well-defined partial isometry that satisfies uu  D . 

In the situation above, the set LN .H ; H /H is equal to H . If we replace H

by an arbitrary right N -module, a similar result holds:

Proposition 10.1.4. Let H1 and H2 be Hilbert spaces, considered as right N -

modules via non-degenerate normal -antihomomorphisms
1 W N ! L.H1 / and

2 W N ! L.H2 /. If ker
1  ker
2 , then LN .H1 ; H2 /H1 is linearly dense
in H2 .

Proof. By [35, I.4.4, Théorème 3], [137, Proposition 2.7.4], or [149, IV, The-
orem 5.5], the -antihomomorphism
2 has the following form: there exist a
Hilbert space H , a projection p 2 .
1 .N / ˝ 1/0  L.H1 ˝ H /, and a unitary
U 2 L.p.H1 ˝ H /; H2 /, such that

2 .x/ D AdU .p.
1 .x/ ˝ 1// for all x 2 N:

Consider H1 ˝ H and p.H1 ˝ H / as right N -modules via the antirepresentation

x 7!
1 .x/ ˝ 1 and its restriction. Then U 2 LN .p.H1 ˝ H /; H2 / and p 2
LN .H1 ˝ H; p.H1 ˝ H //, so that it suffices to show that LN .H1 ; H1 ˝ H /H1
is linearly dense in H1 ˝ H . But for every  2 H , the map  7!  ˝  defines an
element in LN .H1 ; H1 ˝ H / with image H1 ˝ C. The claim follows. 

10.1.2 Outline of the construction

The construction of the relative tensor product is quite involved. Therefore we first
consider the special case where the underlying von Neumann algebra is commutative
and outline the general approach before we turn to the details.
294 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

The commutative case. Let N be a commutative von Neumann algebra on a

separable Hilbert space. Then there exist a compact metrizable space X and a
positive Borel measure  on X with support X such that N is normally isomorphic to
L1 .X; / [35, 1.7.3, Théorème 1]. Every Hilbert module over N corresponds to a
measurable field of Hilbert spaces over .X; /, and the relative tensor product of two
such modules is formed by taking the fiberwise tensor product of the corresponding
fields. Let us explain the details; standard references are [35, Chapitre II], [149,
Section IV.8].

Definition 10.1.5. A measurable field of Hilbert spaces on .X; / is a pair H D

..Hx /x ; .H//
Q consisting of a family of Hilbert spaces .Hx /x2X and a subset
.H/  x2X Hx subject to the following conditions:

i) for all ;  2 .H/, the function x 7! h.x/j.x/i is measurable;

Q
ii) if  2 x2X Hx and for each  2 .H/, the function x 7! h.x/j.x/i is
measurable, then  2 .H/;
iii) there exists a sequence .n /n in .H/ such that the sequence .n .x//n is
linearly dense in Hx for each x 2 X .

The elements of .H/ are called R the measurable sections of H. A measurable

section  is square-integrable if X k.x/k2 d.x/ < 1. The direct integral of H,
R˚
denoted by X Hx d.x/, is the space of all square-integrable sections of H, where
two sections are identified if they coincide -almost everywhere.
R˚
The direct integral
R X Hx d.x/ above is a Hilbert space with respect to the inner
product hji WD X h.x/j.x/id.x/, and a Hilbert module over L1 .X; / with
respect to pointwise multiplication of sections by functions.

Theorem 10.1.6. For every separable Hilbert module H over L1 .X; /, there
R˚
exists a measurable field of Hilbert spaces H on .X; / such that H Š X Hx d.x/
as Hilbert L1 .X; /-modules.

Proof. See [35, II.6.2, Théorème 2]. 

The tensor product of fields of Hilbert spaces is easily constructed:

Proposition 10.1.7. Let H and K be measurable fields of Hilbert spaces on .X; /.
Then there exists a unique measurable field of Hilbert spaces H ˝ K on .X; / such
that

i) .H ˝ K/x D Hx ˝ Kx for all x 2 X , and

ii) for all  2 .H/;  2 .K/, the section x 7! .x/ ˝ .x/ belongs to
.H ˝ K/.
 10.1. The relative tensor product of Hilbert modules 295

Proof. See [35, II.1.8, Proposition 10]. 

The relative tensor product of L1 .X; /-modules can now be defined as follows:
for each pair of measurable fields of Hilbert spaces H and K, we put
Z ˚  Z ˚  Z ˚
Hx d.x/ ˝ Kx d.x/ WD .H ˝ K/x d.x/:
X L1 .X;/ X X

Straightforward but tedious verifications show that the Hilbert space on the right-
hand side does not depend on the precise choice of the fields H and K but only on
the direct integrals that appear on the left-hand side.

The general case. Let us consider the general case where N is an arbitrary von
Neumann algebra and H is a right and K a left N -module. A first attempt at
the definition of the relative tensor product H ˝ K might be to look for an inner
N
product on the algebraic tensor product H ˇ K. The analogy with the internal
N
tensor product of C  -modules suggests to define an N -valued inner product h  j  iN
on H and consider the sesquilinear form on the algebraic tensor product H ˇ K
given by

h0 ˇ  0 j ˇ i WD h 0 jh0 jiN i; where ; 0 2 H; ;  0 2 K:

Then, H ˝ K should be the Hilbert space obtained from that sesquilinear form by
N
the standard procedure, that is, one factorizes out the null space of the form and
completes the quotient with respect to the induced norm.
The approach sketched above does not work in general and has to be modified.
Indeed, the spaces H ˇ K and H turn out to be too large to carry the desired inner
N
products: the space H has to be replaced by a dense subspace of elements that are
bounded with respect to the action of N . The precise definition involves the choice
of a weight on N but turns out to be essentially independent of that choice.
For illustration, let us return to the commutative case considered above: there,
H and K correspond to measurable fields of Hilbert spaces H and K on .X; /,
and the elements of H and K correspond to square-integrable sections of H and K,
respectively. In general, the fiberwise tensor product of such sections is not square-
integrable but integrable. However, the fiberwise tensor product of a section that is
essentially bounded with a square-integrable section is always square-integrable. If
R˚
; 0 2 X Hx d.x/ Š H are essentially bounded in the sense that the functions
x 7! k.x/k and x 7! k0 .x/k belong to L1 .X; /, then also the function x 7!
h0 .x/j.x/i belongs to L1 .X; / Š N , and one can define h0 jiN to be that
function.
296 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

10.1.3 Bounded elements of a Hilbert module

In this subsection, we introduce the space of bounded elements of a right Hilbert
N -module and construct an N -valued inner product on that space. As explained
above, the definition of that inner product is the crucial step in the construction
of the relative tensor product. To define the space of bounded elements of a right
Hilbert N -module H and the N -valued inner product thereon, we choose an n.s.f.
weight on N and identify certain elements of H – the op -bounded elements –
with morphisms between the GNS-space H and H . Before we turn to the details,
let us note that throughout this subsection, we use the N -N -bimodule structure on
H and the notation introduced in Example 10.1.2.

Definition 10.1.8. Let be an n.s.f. weight on N and K a left Hilbert N -module.

An element  2 K is -bounded if the map

R ./ W ƒ .N / ! K; ƒ .y/ 7! y;

is bounded. In this case, we denote the extension to a bounded operator H ! K

by R ./ again. We denote the set of -bounded elements of K by D.K; /.
Let H be a right Hilbert N -module. An element  2 H is op -bounded if the
map

L ./ W ƒ op .N op / ! H; ƒ op .y op / 7! y;

is bounded. In this case, we denote the extension to a bounded operator H ! H

by L ./ again. We denote the set of op -bounded elements of H by D.H; op /.

Remarks 10.1.9. i) Evidently, an element  of a left Hilbert N -module K is

-bounded if and only if there exists a constant C 0 such that kyk2 
C kƒ .y/k2 D C .y  y/ for all y 2 N . Likewise, an element  of a right
Hilbert N -module H is op -bounded if and only if there exists a constant C 0
such that kyk2  C kƒ op .y op /k2 D C .yy  / for all y 2 N  .
ii) If we consider a left Hilbert N -module as a right Hilbert N op -module, we
obtain two notions of op -bounded elements; however, it is easy to see that these
two notions coincide. A similar remark applies to right Hilbert N -modules, which
can also be considered as left Hilbert N op -modules.
iii) In [150], Takesaki writes D.H; / instead of D.H; op /, and D 0 .K; /
instead of D.K; /.

We shall focus on right Hilbert N -modules, on op -bounded elements, and on

the maps L ./ introduced above. These maps are morphisms of right N -modules,
and this fact can be used to construct an N -valued inner product on D.H; op /:

Lemma 10.1.10. Let H be a right Hilbert N -module and an n.s.f. weight on N .

 10.1. The relative tensor product of Hilbert modules 297

i) Consider H as a right Hilbert N -module. Then L ./ 2 LN .H ; H / and

L .0 / L ./ 2
.N / for all ; 0 2 D.H; op /.
ii) The map h  j  i W D.H; op
/  D.H; op
/ ! N given by

h0 ji WD x , L .0 / L ./ D

.x/

is well defined, sesquilinear, and positive definite in the sense that for each
non-zero  2 D.H; op /, the element hji 2 N is positive and non-zero.
Proof. i) Let ; 0 2 D.H; op /, let y 2 N  , and let x 2 N . By definition,
ƒ op .y op /x D
op .x op /ƒ op .y op / D ƒ op ..yx/op /, and hence

L ./.ƒ op .y op /x/ D .yx/ D .y/x D .L ./ƒ op .y op //x:

Consequently, L ./ 2 LN .H ; H / and L .0 / L ./ 2 LN .H /. By Theo-

rem 8.2.4, LN .H / D .J
.N /J /0 is equal to
.N /.
ii) The map h  j  i is well defined because
is injective. Let  2 D.H; op /.
Evidently, hji D L ./ L ./ is positive. Assume that hji D 0. Then
also L ./ D 0 and

hy  ji D hjyi D hjL ./ƒ op .y op /i D 0 for all  2 H; y 2 N 

:

Since N is weakly dense in N and HN D H , we must have  D 0. 

Remark 10.1.11. Let K be a left Hilbert N -module and an n.s.f. weight on N .
Considering K as a right Hilbert N op -module, we obtain an N op -valued inner
product h  j  i op on D.K; /. Since L op ./ D R ./ for all  2 K,

h 0 ji op D x op , R . 0 / R ./ D
op .x op / for all ;  0 2 D.K; /:

For later applications to the relative tensor product of Hilbert N -modules, we

collect several properties of the maps L ./ and of the N -valued product h  j  i
constructed above:
Proposition 10.1.12. Let H be a right Hilbert N -module and an n.s.f. weight
on N .
i) For each  2 D.H; op
/ and x 2 Dom.i=2 /, we have x 2 D.H; op
/
and L .x/ D L ./
.i=2 .x//.

ii) For every  2 D.H; op /, every right Hilbert N -module K, and every
T 2 LN .H; K/, we have T  2 D.K; op / and L .T / D T L ./. Fur-
thermore, h0 jT i D hT  0 ji for every 0 2 D.K; op /.
iii) D.H ; op
/ D ƒ .N / and L .ƒ .y// D
.y/ for all y 2 N .
298 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

iv) D.H; op
/ D LN .H ; H /ƒ .N / is dense in H .
v) For all ; 0 2 D.H; op
/,
0
h ji 2 N ; ƒ .h0 ji / D L .0 / ; .h0 ji / D h0 ji:

Proof. i) Let y 2 N . By Theorem 8.2.4 vi) and Proposition 8.1.11 iii),

.i=2 .x//ƒ op .y op / D
.i=2 .x  // J ƒ .y  /
D J ƒ .y  x  / D ƒ op ..xy/op /;
and therefore
L .x/ƒ op .y op / D xy D L ./
.i=2 .x//ƒ op .y op /:
ii) The first assertions follow from the relation

T L ./ƒ op .y op / D T .y/ D .T /y D L .T /ƒ op .y op / for all y 2 N ;
and the last one from the relation
L .0 / T L ./ D .T  L .0 // L ./ D L .T  0 / L ./:
iii) This is a standard result from the theory of Hilbert algebras, see [150, Proof
of Theorem VII.2.6].
iv) Statements ii) and iii) imply LN .H ; H /ƒ .N /  D.H; op /. Let us
prove the reverse inclusion. Given  2 D.H; op /, we choose a partial isometry
u 2 LN .H ; H / such that uu  D  (see Proposition 10.1.3). Then u  2
D.H ; op / D ƒ .N / by iii) and  D u.u / 2 LN .H ; H /ƒ .N /. Thus,
D.H; op / D LN .H ; H /ƒ .N /. By definition of H , this space is dense in
LN .H ; H /H , and by Proposition 10.1.3, also in H .
v) Let ; 0 2 D.H; op /. By Lemma 10.1.10 i) and by ii), L .0 /  2
D.H ; op /, and by iii), L .0 /  D ƒ .y/ for some y 2 N . Therefore,
L .0 / L ./ D L .L .0 / / D L .ƒ .y// D
.y/ 2
.N /
and ƒ .h0 ji / D ƒ .y/ D L .0 / . Thus we have proved the first and
second assertion. Let us prove .hji / D hji. Choose a partial isometry u 2
LN .H ; H / such that uu  D  (see Proposition 10.1.3). By iii), u  D ƒ .x/
for some x 2 N , and
.x/ D L .u / D u L ./. Then x  x D hji
because

.x  x/ D L ./ uu L ./ D L ./ L .uu / D L ./ L ./;
and
.hji / D .x  x/ D kƒ .x/k2 D kk2 D hji:
Using polarization, we get .h0 ji / D h0 ji for all ; 0 2 D.H; op
/. 
 10.1. The relative tensor product of Hilbert modules 299

Remark 10.1.13. In the situation of the preceding proposition, the linear span of
the set fL ./L .0 / j ; 0 2 D.H; op /g is a weakly dense ideal in LN .H /,
see [26, Proposition 3] or [150, IX.3, Lemma 3.9].
The preceding definitions and results can be rewritten in a suggestive symbolic
calculus. We briefly summarize this calculus, but restrict to a symbolic level and
do not make anything mathematically precise. Further details can be found in [27,
V, Appendix B].
Remark 10.1.14. Before we begin, let us stress that all calculations in this remark
are purely formal symbolic manipulations. We shall use the sign “
” to indicate
that two expressions are equal on a symbolic level.
Consider the Hilbert N -N -bimodule H associated to an n.s.f. weight on N
(see Example 10.1.2). We put
y 1=2
W
ƒ .y/ for y 2 N ;
1=2
1=2
x W
i=2 .x/ for x 2 Dom.i=2 /;

and treat the symbols 1=2 and 1=2 like invertible self-adjoint operators. Then
the N -N -bimodule structure of H takes the form
1=2
x.y 1=2
/z
xy. 1=2
z / 1=2

.xyi=2 .z// 1=2

for all x 2 N , y 2 N , z 2 Dom.i=2 /. A comparison with Example 10.1.2

shows that this symbolic calculation correctly describes the actions of N .
For the GNS-map of the opposite weight op , we put

1=2
y W
y op . op 1=2
/
ƒ op .y op / for y 2 N :
Then the N -N -bimodule structure of H takes the form
1=2
x. 1=2
y/z
1=2
. x 1=2
/yz
1=2
i=2 .x/yz

for all x 2 Dom.i=2 /, y 2 N  , z 2 N . Again, one can show that this symbolic
calculation correctly describes the actions of N .
Given a Hilbert space H , we put
0  W
h0 ji for ; 0 2 H:
Then the defining property of the modular automorphism group  (see Theo-
rem 8.1.13 ii)) can be read off from the following symbolic calculations. Given
x 2 N \ Dom.i=2 /, put z WD 1=2 x 1=2
i=2 .x/. Then

.x  x/
hx 1=2
jx 1=2
i
1=2 
x x 1=2

z 1=2 1=2
z

h 1=2
zj 1=2
zi
op
..z / .z //
.zz /
.i=2 .x/i=2 .x/ /:
op  op 
300 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Moreover, in this notation, .x/

1=2 x 1=2 for each x 2 M ; this relation
will be used below.
Now consider a right Hilbert N -module H . Let us write
1=2 1=2  1=2 
 W
L ./;  W
. /
L ./ for  2 D.H; op
/:

Then the action of L ./ takes the simple form

1=2 
L ./ƒ op .y op /
  1=2
y
y for all  2 D.H; op
/; y 2 N :

If we identify N with
.N /, the inner product h  j  i can be written as follows:

h0 ji
L .0 / L ./
1=2 0
  1=2
for ; 0 2 D.H; op
/:

Now we can “prove” some of the results of Proposition 10.1.12 by the following
symbolic calculations:
1=2 1=2 1=2
L .T x/
.T x/
T . 1=2
x /
T L ./i=2 .x/;
1=2
L .ƒ .y//
.y 1=2
/
y

.y/;

ƒ .h0 ji /
. 1=2 0
  1=2
/ 1=2

. 1=2 0
 /
L .0 / ;

.h0 ji /
1=2
. 1=2 0
  1=2
/ 1=2

0 
h0 ji;

where ; 0 2 D.H; op
/; T 2 LN .H; K/; x 2 Dom.i=2 /, and y 2 N .

10.1.4 Construction of the relative tensor product

Equipped with the concepts introduced above, we are ready to define the relative
tensor product of Hilbert modules.
Recall that to every sesquilinear form on a complex vector space, one can asso-
ciate a Hilbert space by a standard procedure: one factorizes out the null space of
the form and completes the quotient with respect to the induced norm.
Lemma 10.1.15. Let H be a right and K a left Hilbert N -module, and let be
an n.s.f. weight on N . Consider H as a left and K as a right Hilbert N op -module.
i) Formula (10.2) defines a positive sesquilinear form on D.H; op
/ ˇ K:

h0 ˇ  0 j ˇ i WD h 0 jh0 ji i: (10.2)

ii) Formula (10.3) defines a positive sesquilinear form on H ˇ D.K; /:

h0 ˇ  0 j ˇ iop WD h0 jh 0 ji op i: (10.3)

 10.1. The relative tensor product of Hilbert modules 301

iii) The restrictions of h  j  i and h  j  iop to D.H; op / ˇ D.K; / coincide,

and for all ;  0 2 D.H; op / and ; 0 2 D.K; /,

h0 ˇ  0 j ˇ i D hƒ op .hj 0 i op /jƒ .h0 ji /i D h0 ˇ  0 j ˇ iop :

iv) The Hilbert spaces associated to the positive sesquilinear forms considered
in i)–iii) are naturally isomorphic.
Proof. i), ii) This follows from a standard argument (see [95, Chapter 4]).
iii) It suffices to prove the two equalities in the formula above. We only treat
the first one; the proof of the second one is similar. By definition,

h0 ˇ  0 j ˇ i D h 0 jh0 ji i D h 0 j.h0 ji /op i:

Now .h0 ji /op D L op ./ƒ .h0 ji / by definition of L op ./, and hence

h0 ˇ  0 j ˇ i D hL op ./  0 jƒ .h0 ji /i:

Inserting the relation L op ./  0 D ƒ op .hj 0 i op / from Proposition 10.1.12 v),

we obtain the desired equality.
iv) This follows immediately from iii) and the fact that D.H; op / and D.K; /
are dense in H and K, respectively (see Proposition 10.1.12 iv)). 
Definition 10.1.16. The relative tensor product of a right Hilbert N -module H and
a left Hilbert N -module K relative to an n.s.f. weight on N is the Hilbert space
H ˝ K associated by the standard procedure to the sesquilinear forms defined in
Lemma 10.1.15 i)–iii). Given elements  2 D.H; op / and  2 D.K; /, we write
 ˝  for the image of  ˇ  under the natural map D.H; op / ˇ D.K; / !
H ˝ K. Likewise, we define the element  ˝  if either  2 H or  2 K.

If N is commutative, the relative tensor product of Hilbert N -modules corre-

sponds to the fiberwise tensor product of measurable fields of Hilbert spaces:
Proposition 10.1.17. Let N D L1 .X; /, where X is a compact metrizable space
and  is aR positive Borel measure on X . Denote also by the weight on N given
by f 7! X f d. Let H and K be measurable fields of Hilbert spaces on .X; /
R˚ R˚
and consider H D X Hx d.x/ and K D X Kx d.x/ as Hilbert N -modules
via pointwise multiplication of sections by functions.
i) An element  2 H belongs to D.H; op / if and only if the function x 7!
k.x/k belongs to L2 .X; / \ L1 .X; /.
ii) For all ; 0 2 D.H; op
/, the inner product h0 ji is equal to the function
x 7! h0 .x/j.x/i.
302 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces
R˚
iii) H ˝ K Š X Hx ˝ Kx d.x/.

Proof. Straightforward. 
In the general case, the actions of N on the factors of the relative tensor product
do not behave as one might expect. Unlike the usual transformation rule x ˇ  D
N
 ˇ x that holds in an algebraic tensor product H ˇ K for all  2 H;  2 K,
N N
x 2 N , we have the following relation:
Proposition 10.1.18. Let H be a right and K a left Hilbert N -module, and let
be an n.s.f. weight on N . Then for all  2 D.H; op /,  2 K, x 2 Dom.i=2 /,

x ˝  D  ˝ i=2 .x/: (10.4)

Proof. By Proposition 10.1.12 i), h0 jxi D h0 ji i=2 .x/, and hence

h0 ˝  0 jx ˝ i D h 0 jh0 ji i=2 .x/i D h0 ˝  0 j ˝ i=2 .x/i

for all 0 2 D.H; op

/ and  0 2 K. 
Remark 10.1.19. We extend the notation introduced in Remark 10.1.14 and put
1=2 1=2
0 ˝  W
0 ˝  and  ˝ 0 W
 ˝ 0
N N

for 0 2 D.H; op /,  2 K and  2 H , 0 2 D.K; /. Then the following

symbolic calculation “proves” Proposition 10.1.18:
1=2 1=2 1=2
0 x ˝ 0
0 . 1=2
x / ˝ 0
N N
1=2 1=2

0 i=2 .x/ ˝ 0
0 ˝ i=2 .x/0 :
N N

10.1.5 Properties of the relative tensor product

The relative tensor product has all functorial properties that one would naturally
expect.
Lemma 10.1.15 implies that the construction of the relative tensor product is
symmetric in the following sense:
Proposition 10.1.20. Let H be a right and K a left Hilbert N -module, and let be
an n.s.f. weight on N . Consider H as a left and K as a right Hilbert N op -module.
Then the map
D.H; op
/ ˝ D.K; / ! D.K; / ˝ D.H; op
/;  ˝  7!  ˝ ;
op op
 10.1. The relative tensor product of Hilbert modules 303
Š
is well defined and extends to a unitary † W H ˝ K 
! K ˝ H. 
op

Evidently, the relative tensor product commutes with direct sums:

Proposition 10.1.21. Let .H / be a family of right and .K / a family of left
Hilbert N -modules,
L and let beLan n.s.f. weight on N . Consider the direct
sums H WD H and K WD  K as Hilbert N -modules in the obvious
way. Then the family of natural inclusions H ˝ K ,! H ˝ K induces a unitary
L Š
; .H ˝ K / ! H ˝ K.

Proof. Straightforward. 
The relative tensor product admits the following natural maps:
Proposition 10.1.22. Let H be a right and K a left N -module, and let be an
n.s.f. weight on N . For every 0 2 D.H; op / and 0 2 D.K; /, the maps
j0 i Œ1 W K ! H ˝ K;  7! 0 ˝ ; j0 i Œ2 W H ! H ˝ K;  7!  ˝ 0 ;

are linear operators of norm

kj0 i Œ1 k  kh0 j0 i k1=2 ; kj0 i Œ2 k  kh0 j0 i op k1=2 :
If the representation of N on K/on H is injective, the left/right inequality above is
an equality. The adjoints h0 j Œ1 WD j0 i Œ1 and h0 j Œ2 WD j0 i Œ2 are given
by
0
h0 j Œ1 . ˝  0 / D h0 j0 i  0 ; h0 j Œ2 .
00
˝  00 / D h0 j 00 i op 00 ;

where 0 2 D.H; op
/,  0 2 K and 00 2 H ,  00 2 D.K; /.
Proof. Straightforward. 
We shall show that the construction of the relative tensor product is bifunctorial.
First, we consider a special case.
Proposition 10.1.23. Let H be a right and K a left N -module, and let be an
n.s.f. weight on N .
i) For each S 2 LN .H /, there exists an operator S ˝ 1 2 L.H ˝ K/
such that .S ˝ 1/. ˝ / D S ˝  for all  2 H and  2 D.K; /.
The map LN .H / ! L.H ˝ K/ given by S 7! S ˝ 1 is a normal non-
degenerate -homomorphism. If the representation of N on K is injective,
this -homomorphism is injective.
304 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

ii) For each T 2 LN .K/, there exists an operator 1 ˝ T 2 L.H ˝ K/ such

that .1 ˝ T /. ˝ / D  ˝ T  for all  2 D.H; op
/ and  2 K. The map
LN .K/ ! L.H ˝ K/ given by T 7! 1 ˝ T is a normal non-degenerate
-homomorphism. If the antirepresentation of N on H is injective, this -
homomorphism is injective.
Proof. By Proposition 10.1.20, it suffices to prove i). First, we show that for each
S 2 LN .H /, the map S ˝ 1 is well defined and bounded. Let i 2 D.H; op /
and i 2 K, where i D 1; : : :P 10.1.12 ii), S i 2 D.H; op /
; n. By PropositionP
for all i , and by definition, k i Si ˝ i k D i;j hi jhS i jS j i j i. Now
2

.hS i jS j i /i;j  kSk2 .hi jj i /i;j in Mn .N / because

.L .i / S  SL .j //i;j  kSk2 .L .i / L .j //i;j in Mn .

.N //;

and hence
X 2 X X 2

S i ˝ i  kSk2  hi jhi jj i j i D kSk2  i ˝ i :
i i;j i

Consequently, S ˝ 1 is well defined and bounded.

Using Proposition 10.1.12 ii), we find that the map
W LN .H / ! L.H ˝ K/
given by S 7! S ˝ 1 is a -homomorphism; in particular, it is norm-continuous. To
prove that
is normal, we show that for every normal functional ! 2 L.H ˝ K/ ,
the composition ! ı
is normal. By [149, II.3, Theorem 2.6], functionals of the
form ! 0 ; D h 0 j  i, where ;  0 2 H ˝ K, are linearly dense in L.H ˝ K/ .
Evidently, each such functional can be approximated in norm by linear combinations
of functionals of the form ! D h0 ˝  0 j  . ˝ /i, where ; 0 2 H; ;  0 2
D.K; /. Since
is norm-continuous and L.H / is closed in L.H / [149, II.3,
Theorem 2.6], it is enough to show that for each such functional !, the composition
! ı
is normal. But

.! ı
/.S / D h0 ˝  0 jS ˝ i D h0 jh 0 ji op S i D h0 jS 00 i

for all S 2 LN .H /, where 00 WD h 0 ji op . Hence ! ı

is normal.
Finally, we assume that the representation of N on K is injective, and show
that
is injective. Let S 2 LN .H /; S ¤ 0. Since D.H; op / is dense in
H (Proposition 10.1.12 iv)), there exists an  2 D.H; op / such that S  ¤ 0.
 10.1. The relative tensor product of Hilbert modules 305

By Lemma 10.1.10 ii), hSjSi ¤ 0, and since the representation of N on K

is injective, there exists a  2 K such that kS  ˝ k2 D hjhS jS i i ¤ 0.
Therefore,
.S / ¤ 0. 
Combining the previous result with Proposition 10.1.21, we obtain:
Corollary 10.1.24. Let be an n.s.f. weight on N .
i) Let H1 , H2 be right and K1 , K2 left Hilbert N -modules. For all S 2
LN .H1 ; H2 / and T 2 LN .K1 ; K2 /, there exists an operator
S ˝ T 2 L.H1 ˝ K1 ; H2 ˝ K2 /;  ˝  7! S  ˝ T :

ii) The assignments .H; K/ 7! H ˝ K and .S; T / 7! S ˝ T define a bifunctor

from the categories of right and left Hilbert N -modules to the category of
Hilbert spaces.
Proof. Assertion i) follows easily from Proposition 10.1.21 and 10.1.23; simply
consider H WD H1 ˚ H2 as a right and K WD K1 ˚ K2 as a left Hilbert N -module,
and S and T as elements of LN .H / and LN .K/, respectively. Statement ii) is an
immediate consequence. 
The relative tensor product is associative:
Proposition 10.1.25. Let N1 and N2 be von Neumann algebras with n.s.f. weights
1 and 2 , respectively, and let K1 be a right Hilbert N1 -module, H a Hilbert
N1 -N2 -bimodule, and K2 a left Hilbert N2 -module. Then
i) K1 ˝ H is a right Hilbert N2 -module via .1 ˝ /x WD 1 ˝ x;
1 1 1

ii) H ˝ K2 is a left Hilbert N1 -module via y. ˝ 2 / WD y ˝ 2 ;

2 2 2

iii) .K1 ˝ H / ˝ K2 Š K1 ˝ .H ˝ K2 / via .1 ˝ / ˝ 2 $ 1 ˝ . ˝ 2 /

1 2 1 2 1 2 1 2
op
for all 1 2 D.K1 ; 1 /; 2 2 D.K2 ; 2 /,  2 H.
Proof. i), ii) This follows immediately from Corollary 10.1.24.
iii) By Lemma 10.1.15, .K1 ˝ H / ˝ K2 and K1 ˝ .H ˝ K2 / are equal to
1 2 1 2

op op
Œ.D.K1 ; 1 / ˝ H / ˝ D.K2 ; 2 / and ŒD.K1 ; 1 / ˝ .H ˝ D.K2 ; 2 //;
1 2 1 2

op
respectively. Let 1 2 D.K1 ; 1 /, 2 2 D.K2 ; 2 /,  2 H . Several applications
of Lemma 10.1.15 show that in .K1 ˝ H / ˝ K2 and in K1 ˝ .H ˝ K2 /,
1 2 1 2

k.1 ˝ / ˝ 2 k2 D h1 ˝ j1 ˝ h2 j2 i op i D hjh1 j1 i 1

h2 j2 i op i
2 2
1 2 1 1
306 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

and
k1 ˝ . ˝ 2 /k2 D h ˝ 2 jh1 j1 i 1
 ˝ 2 i D hjh2 j2 i op h1 j1 i 1
i:
2
1 2 2 2

Since H is an N1 -N2 -bimodule, these expressions are equal, so that the identifica-
tion given in statement ii) is a well-defined isometric isomorphism. 

The relative tensor product absorbs the standard Hilbert N -N -bimodule H

introduced in Example 10.1.2, that is, H is a unit for the relative tensor product:

Proposition 10.1.26. Let be an n.s.f. weight on N .

i) For every right Hilbert N -module H , we have H ˝ H Š H as right Hilbert


N -modules via  ˝ ƒ op .y op / $ y for all  2 D.H; op
/, y 2 N .

ii) For every left Hilbert N -module K, we have H ˝ K Š K as left Hilbert

N -modules via ƒ .y/ ˝  $ y for all  2 D.K; /, y 2 N .

Proof. i) The map  ˝ ƒ op .y op / 7! y is well defined and isometric because for


all  2 D.H; op
/ and y 2 N ,

k ˝ ƒ op .y op /k2 D hƒ op .y/jhji ƒ op .y/i

D hL ./ƒ op .y/jL ./ƒ op .y/i D hyjyi D kyk2 :

The map is surjective because .D.H; op /N  /? D .H N  /? D .HN /? D 0;

here, we used that D.H; op / is dense in H (Proposition 10.1.12 iv)), that N  is
weakly dense in N , and that N acts non-degenerately and normal on H .
Finally, the map is a morphism of right N -modules because ƒ op .y op /x D

op .x op /ƒ op .y op / D ƒ op ..yx/op / for all y 2 N  and x 2 N .

ii) Consider K as a right N op -module and use i) and Proposition 10.1.20. 

Finally, the relative tensor product is essentially independent of the choice of

the weight on N :

Theorem 10.1.27. Let and  be n.s.f. weights on N . Then the bifunctors

.H; K/ 7! H ˝ K; .S; T / 7! S ˝ T and .H; K/ 7! H ˝ K; .S; T / 7! S ˝ T

from the categories of right and left Hilbert N -modules to the category of Hilbert
spaces are naturally isomorphic.

The proof involves the Radon–Nikodym derivative of n.s.f. weights and can be
found in [150, IX.3, Theorem 3.21].
 10.2. Hopf–von Neumann bimodules 307

Remarks 10.1.28. i) The natural isomorphism H ˝ K Š H ˝ K is not given by

 ˝  $  ˝ . In the notation of Remark 10.1.19, it has the form  1=2 ˝  $

N
1=2
.d =d/1=2 ˝ , where d =d denotes the Radon–Nikodym derivative of
N
with respect to .
ii) The natural isomorphisms H ˝ K Š H ˝ K and K ˝ H Š K ˝ H are


op op

compatible with the flip maps † and †

defined in Proposition 10.1.20.

Notation 10.1.29. In the remainder of this chapter, we neglect the precise choice
of the n.s.f. weight and denote the bifunctor .  / ˝ .  / by .  / ˝ .  /.
N

10.2 Hopf–von Neumann bimodules

Hopf–von Neumann bimodules generalize von Neumann bialgebras just like group-
oids generalize groups, and groupoids just like von Neumann bialgebras generalize
groups, see the introduction to this chapter. The definition is due to Vallin [171]
and involves the fiber product of von Neumann algebras which was introduced by
Sauvageot [139].

10.2.1 The fiber product of von Neumann algebras

The fiber product of von Neumann algebras is an analogue of the algebraic tensor
product of algebras over a common subalgebra. Its definition involves the relative
tensor product of Hilbert modules.
We fix a von Neumann algebra N and adopt the following terminology:

Definition 10.2.1. A left/right von Neumann N -module is a von Neumann al-

gebra M equipped with a unital normal -homomorphism/-antihomomorphism
N ! M.
Let N1 and N2 be von Neumann algebras. A von Neumann N1 -N2 -bimodule is
a von Neumann algebra M equipped with the structure of a left N1 -module and a
right N2 -module such that the images of N1 and N2 in M commute.
A morphism of left von Neumann N -modules M1 ; M2 is a non-degenerate nor-
mal -homomorphism
W M1 ! M2 that satisfies x
.y/ D
.xy/ for all x 2 N
and y 2 M1 . Similarly, we define morphisms of right von Neumann N -modules
and of von Neumann N1 -N2 -bimodules.

Remarks 10.2.2. i) Evidently, left von Neumann N -modules correspond bijec-

tively with right von Neumann N op -modules, and right von Neumann N -modules
correspond bijectively with left von Neumann N op -modules.
308 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

i) If M is a von Neumann algebra on a Hilbert space H and M is a left/right

N -module, then the composition N ! M ,! L.H / turns H into a left/right
Hilbert N -module.
Throughout this section, we apply Notation 10.1.29 and write “˝” instead of
N
“˝” in all constructions that involve the relative tensor product. As usual, we denote
the commutant of a set X of operators on some Hilbert space by X 0 .
The definition of the fiber product of von Neumann algebras involves the fol-
lowing auxiliary construction:
Definition 10.2.3. Let H1 be a right and H2 a left Hilbert N -module. The relative
tensor product of von Neumann algebras M1  LN .H1 / and M2  LN .H2 / is
the von Neumann algebra

M1 x̋ M2 WD fx ˝ y j x 2 M1 ; y 2 M2 g00  L.H1 ˝ H2 /:
N N N

The fiber product of von Neumann modules is defined as follows:

Definition 10.2.4. Let M1 and M2 be von Neumann algebras on Hilbert spaces
H1 and H2 , respectively. Assume that M1 is a right and M2 a left N -module, and
consider H1 and H2 as N -modules as in Remark 10.2.2 ii). Then the fiber product
of M1 and M2 over N is the von Neumann algebra

M1  M2 WD .M1 0 x̋ M2 0 /0  L.H1 ˝ H2 /:
N N N

Remarks 10.2.5. i) Strictly speaking, the von Neumann algebras constructed in

Definition 10.2.3 and 10.2.4 are well defined up to spatial isomorphism only, see
Theorem 10.1.27.
ii) For i D 1; 2, consider Hi as a left Hilbert Mi -module. Evidently, Mi 0 D
LMi .Hi /  LN .Hi / for i D 1; 2, so that the relative tensor product M1 0 x̋ M2 0 is
N
well defined. Moreover,

M1  M2 D .LM1 .H1 / x̋ LM2 .H2 //0 D L.M1 0 x̋ M2 0 / .H1 ˝ H2 /:

N N N N

Š
iii) The flip map †N W H1 ˝ H2 
! H2 ˝ H1 (see Proposition 10.1.20)
N N op
Š
induces an isomorphism Ad†N W L.H1 ˝ H2 / 
! L.H2 ˝ H1 / which restricts
N N op
Š
to an isomorphism Ad† W M1  M2 
! M2 op M1 .
N N
iv) If N is commutative and the maps N ! Mi ; i D 1; 2, take values only
in the center Z.Mi /, then M1 and M2 are direct integrals of measurable fields of
 10.2. Hopf–von Neumann bimodules 309

von Neumann algebras, and their fiber product M1  M2 is the direct integral of
N
the fiberwise tensor product of these fields. For details on fields of von Neumann
algebras and their fiberwise tensor product, see [35, Chapitre II].
In general, the fiber product M1  M2 of von Neumann algebras can not be
N
described in terms of elementary tensor products x1 ˝ x2 , where x1 2 M1 and
N
x2 2 M2 . More precisely, let H1 ; H2 and M1 ; M2 be as in the definition above.
Then an elementary tensor x1 ˝ x2 as above is well defined if and only if xi 2
N
LN .Hi / for i D 1; 2 (see Proposition 10.1.23). In that case, x1 ˝ x2 belongs to
N
M1  M2 :
N

Lemma 10.2.6. Let H1 , H2 and M1 , M2 be as in Definition 10.2.4. Then

.M1 \ LN .H1 // x̋ .M2 \ LN .H2 //  M1  M2 :

N N

Proof. For i D 1; 2, put MiN WD Mi \ LN .Hi /. Then Mi 0  .MiN /0 and

M1N x̋ M2N  .M1 0 x̋ M2 0 /0 D M1  M2 : 

N N N

As one should expect, the fiber product of bimodules is a bimodule:

Proposition 10.2.7. Let M1 and M2 be von Neumann algebras on Hilbert spaces
H1 and H2 , respectively. Assume that M1 is an N1 -N -bimodule and M2 an N -N2 -
bimodule, where N1 and N2 are some von Neumann algebras. Then M1  M2 is a
N
von Neumann N1 -N2 -bimodule via x1 7! x1 ˝ 1 for all x1 2 N1 and x2 7! 1 ˝ x2
N N
for all x2 2 N2 .
Proof. For i D 1; 2, denote by i the fixed -(anti)homomorphism Ni ! Mi .
From Proposition 10.1.23, we obtain normal unital -(anti)homomorphisms

N1 ! LN .H1 / \ M1 ! LN .H1 ˝ H2 /; x1 7! 1 .x1 / ˝ 1;

N N
N2 ! LN .H2 / \ M2 ! LN .H1 ˝ H2 /; x2 7! 1 ˝ 2 .x2 /;
N N

and by Lemma 10.2.6, the images of these maps are contained in M1  M2 . 

N

The fiber product is associative:

Proposition 10.2.8. Let M1 , M2 , M3 be von Neumann algebras on Hilbert spaces
H1 , H2 , H3 , respectively. Assume that M1 is a right N1 -module, M2 an N1 -N2 -
bimodule, and M3 a left N2 -module, where N1 and N2 are some von Neumann
310 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

algebras. Then the natural isomorphism .H1 ˝ H2 / ˝ H3 Š H1 ˝ .H2 ˝ H3 /

N1 N2 N1 N2
induces an isomorphism

.M1  M2 /  M3 Š M1  .M2  M3 /:
N1 N2 N1 N2

Proof. The isomorphism of the Hilbert spaces identifies .M1 0 x̋ M2 0 / x̋ M3 0 with

N1 N2
M1 0 x̋ .M2 0 x̋ M3 0 /, and the iterated fiber products are the respective commutants
N1 N2
of these von Neumann algebras. 
As one should expect, the fiber product is bifunctorial:
Proposition 10.2.9. Let H1 , H2 and M1 , M2 be as in Definition 10.2.4. Let
i
.i D 1; 2/ be a non-degenerate normal representation of Mi on a Hilbert space Ki
i
and consider Ki as a Hilbert N -module via the map N ! Mi ! L.Ki /.
i) There exists a unique normal unital -homomorphism

1 
2 W M1  M2 !
1 .M1 / 
2 .M2 /
N N N

such that for all T1 2 LM1 .H1 ; K1 /, T2 2 LM2 .H2 ; K2 /, x 2 M1  M2 ,

N

.
1 
2 /.x/ ı .T1 ˝ T2 / D .T1 ˝ T2 / ı x: (10.5)
N N N

ii) .
1 
2 /.M1  M2 / D
1 .M1 / 
2 .M2 /.
N N N
iii) If
1 and
2 are injective, so is
1 
2 .
N

Proof. i) By Proposition 10.1.4, elements of the form .T1 ˝ T2 /, where Ti 2

N
LMi .Hi ; Ki / and  2 H1 ˝ H2 , are linearly dense in K1 ˝ K2 . Therefore,
1 
2
N N N
is uniquely determined by condition (10.5). To prove existence, we consider the
special cases K1 D H1 ;
1 D idM1 and K2 D H2 ;
2 D idM2 separately; then
it is easy to see that the composition
1 
2 WD .id .M1 / 
2 / ı .
1  idM2 /
N N N
satisfies condition (10.5). We treat the case K2 D H2 ;
2 D idM2 ; the other case
is similar.
By [35, I.4.4, Théorème 3], [137, Proposition 2.7.4], or [149, IV, Theo-
rem 5.5],
1 can be written as follows: there exist a Hilbert space H0 , a pro-
jection p 2 .1 x̋ M1 /0  L.H0 ˝ H1 /, and a unitary U 2 L.p.H0 ˝ H1 /; K1 /,
such that

1 .z/ D AdU .p.1 ˝ z// for all z 2 M1 :
 10.2. Hopf–von Neumann bimodules 311

We consider H0 ˝H1 and p.H0 ˝H1 / as right Hilbert N -modules via the right N -
module structure of M1 and the action of M1 on H1 ; then p and U are morphisms
of right N -modules. Furthermore, we use the identification .H0 ˝ H1 / ˝ H2 Š
N
H0 ˝.H1 ˝ H2 / and omit the parentheses. We define
1  idM2 as the composition
N N
of the following three normal -homomorphisms:
• amplification by H0 , given by M1  M2 ! L.H0 ˝ H1 ˝ H2 /; x 7!
N N
idH0 ˝ x,
• induction by the projection p ˝ idH2 , given by
N

C idH0 x̋ .M1  M2 / ! L.p.H0 ˝ H1 / ˝ H2 /;

N N
idH0 ˝ x 7! .p ˝ idH2 /.idH0 ˝ x/I
N

note that this is a normal -homomorphism because p ˝ idH2 is contained in

N

L.H0 / x̋ M1 0 x̋ M2 0 D L.H0 / x̋ .M1  M2 /0 D .1 x̋ .M1  M2 //0 I

N N N

• conjugation by U ˝ idH2 2 L.p.H0 ˝ H1 / ˝ H2 ; K1 ˝ H2 /.

N N N

Thus we put

.
1  idM2 /.x/ WD Ad.U ˝1/ ..p ˝ 1/.1 ˝ x// for all x 2 M1  M2 :
N N N N

Evidently, this composition is non-degenerate and normal, and a straightforward

calculation shows that it satisfies condition (10.5).
ii) It suffices to prove ..
1 
2 /.M1  M2 //0 D
1 .M1 /0 x̋
2 .M2 /0 . Let us
N N N
show that the right-hand side is contained in the left-hand side. By condition (10.5),
we have for all x 2 M1  M2 ; Ti 2 LMi .Hi ; Ki /, Si 2
i .Mi /0
N

.S1 ˝ S2 / ı .
1 
2 /.x/ ı .T1 ˝ T2 / D .S1 T1 ˝ S2 T2 / ı x
N N N N
D .
1 
2 /.x/ ı .S1 ˝ S2 / ı .T1 ˝ T2 /:
N N N

Since elements of the form .T1 ˝ T2 /, where Ti 2 LMi .Hi ; Ki / and  2 H1 ˝ H2 ,
N N
are linearly dense in K1 ˝ K2 (Proposition 10.1.4), the equation above implies that
N
S1 ˝ S2 commutes with .
1 
2 /.x/. Thus,
1 .M1 /0 x̋
2 .M2 /0 is contained in
N N N
..
1 
2 /.M1  M2 //0 .
N N
312 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Let us prove the reverse inclusion. For i D 1; 2, put Yi D LMi .Hi ; Ki /.

Consider the von Neumann algebra
0 100
M1 0 ˝ M2 0 Y1 ˝ Y2
A WD @ N N A  L..H1 ˝ H2 / ˚ .K1 ˝ K2 //:
Y1 ˝ Y2
1 .M1 /0 ˝
2 .M2 /0 N N
N N

Here, we think of elements of the Hilbert space .H1 ˝ H2 / ˚ .K1 ˝ K2 / as column

N N
vectors with upper entry in H1 ˝ H2 and lower entry in K1 ˝ K2 , and of elements
N N
of A as 2-by-2-matrices in the obvious way. It is easy to see that
( ! ˇ )
x ˇ
0
A D ˇ
.
1 
2 /.x/ ˇ x 2 M1 N  M2 ;
N
and that !
0
..
1 
2 /.M1  M2 //0  A00 D A:
N N

Therefore, ..
1 
2 /.M1  M2 //0 
1 .M1 /0 x̋
2 .M2 /0 .
N N N
iii) For i D 1; 2, assume that
i is injective, and denote by
i1 W
i .Mi / ! Mi
its inverse. Then
11 
21 is inverse to
1 
2 . 
N N

Using the previous proposition, we can extend the fiber product to a bifunctor:
Given a morphism
1 W M1 ! N1 of right von Neumann N -modules and a mor-
phism
2 W M2 ! N2 of left von Neumann N -modules, we can define a normal
-homomorphism
1 
2 W M1  M2 ! N1  N2 . Then, we find:
N N N

Corollary 10.2.10. The assignments

.M1 ; M2 / 7! M1  M2 and .
1 ;
2 / 7!
1 
2
N N

define a bifunctor from the categories of right and left von Neumann N -modules to
the category of von Neumann algebras. 

10.2.2 Hopf–von Neumann bimodules

Informally, the concept of a Hopf–von Neumann bimodule has already been de-
scribed in the introduction to this chapter. Let us now give the precise definition.
Notation 10.2.11. Let M and N be von Neumann algebras. If we want to emphasize
that we consider M as a left/right N -module via a specific -homomorphism/-
antihomomorphism
, we write M or M for that N -module, respectively.
 10.2. Hopf–von Neumann bimodules 313

Definition 10.2.12. A Hopf–von Neumann bimodule is a tuple .N; M; r; s; / con-

sisting of a von Neumann algebra N called the base, a von Neumann algebra M ,
a -homomorphisms r W N ! M called the range map, a -antihomomorphism
s W N ! M called the source map, and a -homomorphism  W M ! Ms  r M
N
called the comultiplication, such that
i) r, s,  are normal, injective and unital;
ii) .s.x// D 1 ˝ s.x/ and .r.x// D r.x/ ˝ 1 for all x 2 N ;
N N
iii)  is coassociative in the sense that .  idM / ı  D .idM  / ı  as maps
N N
M ! Ms  r Ms  r M , that is, the following diagram commutes:
N N

M
 /M  M
s r
N

   id
  N
Ms  r M / Ms  r Ms  r M .
N id   N N
N

Remarks 10.2.13. Let .N; M; r; s; / be a Hopf–von Neumann bimodule.

i) The maps r and s turn M into an N -N -bimodule. Since no other (anti)repre-
sentations of N on M occur at the moment, we may omit the indices s and r
in the fiber product and write M  M for Ms  r M .
N N
ii) The maps N ! M  M given by x 7! r.x/˝1 and x 7! 1˝s.x/ turn M  M
N N N N
into an N -N -bimodule (Proposition 10.2.7). Condition 10.2.12 ii) says that
the map  W M ! M  M is a morphism of N -N -bimodules. Therefore the
N
maps   id and id   in condition iii) are well defined (Proposition 10.2.9).
N N
iii) In condition 10.2.12 i), the assumption that  is unital can be omitted, because
this follows from the assumption that r and s are unital and from condition
10.2.12 ii).
iv) The tuple .N op ; M; s; r; Ad† ı/ is a Hopf–von Neumann bimodule again,
called the coopposite of .N; M; r; s; /. Here, Ad†N denotes the natural
isomorphism Ms  r M ! Mr op s M (see Remark 10.2.5 iii)).
N N

Examples 10.2.14. The main examples of Hopf–von Neumann bimodules arise

from
• locally compact groupoids, see Section 10.3.3 or [171];
• inclusions of factors, see [42], [43], [44], [50], [51].
314 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Further examples of Hopf–von Neumann bimodules are discussed in the thesis of

Lesieur [99]. In most examples, the construction of a Hopf–von Neumann bimodule
proceeds via a pseudo-multiplicative unitary, see the next section.

10.3 Pseudo-multiplicative unitaries on Hilbert spaces

The concept of a pseudo-multiplicative unitary on a Hilbert space, introduced by
Vallin [172], plays a similar fundamental rôle in the theory of measurable quantum
groupoids like the concept of a multiplicative unitary in the theory of locally compact
quantum groups.

10.3.1 Definition
A pseudo-multiplicative unitary acts on relative tensor products of a Hilbert space,
taken with respect to several representations:
Assumption 10.3.1. Throughout this section, let N be a von Neumann algebra with
an n.s.f. weight , and let H be a Hilbert space with
• a non-degenerate injective normal representation ˛ of N , and
• non-degenerate injective normal antirepresentations ˇ; ˇO of N ,
O / commute pairwise.
such that ˛.N /, ˇ.N /, ˇ.N
The Hilbert space H can be considered as a module in various ways. To distin-
guish between the different possibilities, we use the following notation:
Notation 10.3.2. We denote by ˛ H the left Hilbert N -module given by H and ˛,
and by Hˇ , HˇO the right Hilbert N -modules given by H and ˇ or ˇ,O respectively.
op
If we consider ˛ H , Hˇ , HˇO as Hilbert N -modules, we write H˛ , ˇ H , ˇO H ,
respectively. We combine these notations to denote bimodule structures on H ,
and use a similar notation to denote module or bimodule structures on an arbitrary
Hilbert space with a given (anti)representation.
O we can form the following relative tensor products:
Given H and ˛, ˇ, ˇ,

HˇO ˝ ˛ H; Hˇ ˝ ˛ H; H˛ ˝ ˇO H; H˛ ˝ ˇ H:
op op

A pseudo-multiplicative unitary on H is a unitary

V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H
op

that satisfies the pentagon equation VŒ12 VŒ13 VŒ23 D VŒ23 VŒ12 known from the
definition of a multiplicative unitary. In the present context, this equation requires
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 315

some explanation, for example, the individual factors VŒij  do not act on one fixed
space but on several intermediate relative tensor products. Therefore, several inter-
twining conditions have to be imposed on V and ˛; ˇ; ˇO to ensure that each factor
VŒij  occurring in the pentagon equation is well defined.
To describe these intertwining relations, we use the following leg notation for
O / commute, we can define (anti)rep-
(anti)representations: Since ˛.N /, ˇ.N /, ˇ.N
resentations

˛Œ1 ; ˇŒ1 ; ˇŒ2 ; ˇOŒ2 on HˇO ˝ ˛ H; ˛Œ1 ; ˇOŒ1 ; ˇOŒ2 ; ˇŒ2 on Hˇ ˝ ˛ H;

ˇOŒ1 ; ˇŒ1 ; ˇŒ2 ; ˛Œ2 on H˛ ˝ ˇO H; ˇŒ1 ; ˇOŒ1 ; ˇOŒ2 ; ˛Œ2 on H˛ ˝ ˇ H I

op op

for example, the first four (anti)representations are given by

˛Œ1 .x/ D ˛.x/ ˝ 1; ˇŒ1 .x/ D ˇ.x/ ˝ 1;

ˇŒ2 .x/ D 1 ˝ ˇ.x/; ˇOŒ2 .x/ D 1 ˝ ˇ.x/:

O

Lemma 10.3.3. If a bounded linear operator V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H satisfies

op

V ˛Œ1 .x/ D ˛Œ2 .x/V; VˇŒ2 .x/ D ˇOŒ1 .x/V;

(10.6)
VˇŒ1 .x/ D ˇŒ1 .x/V; V ˇOŒ2 .x/ D ˇOŒ2 .x/V for all x 2 N;

then all operators in the following diagram are well defined:

H˛ ˝ ˇ HˇO ˝ ˛ H
ii i ii4 op UUUU
UUU
i iiiiV ˝1 1 ˝ V UUUUUU
iii op *
HˇO ˝ ˛ HˇO ˝ ˛ H H˛ ˝ ˇ H˛ ˝ ˇ H
:: op B

op
:: 1˝V V ˝ 1 
:: op 
::  (10.7)
 
HˇO ˝ ˛Œ2 .H˛ ˝ ˇ H / .HˇO ˝ ˛ H /O ˛Œ1 ˝ ˇ H
op op

1˝† †Œ23
 V ˝1
HˇO ˝ ˛ Hˇ ˝ ˛ H / .H˛ ˝op ˇ H /ˇOŒ1 ˝ ˛ H .

Here, †Œ23 denotes the isomorphism given by .1 ˝ 2 / ˝ 3 7! .1 ˝ 3 / ˝ 2 .

op op
316 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Proof. This follows immediately from the fact that the relative tensor product is
symmetric (Proposition 10.1.20), associative (Proposition 10.1.25), and bifunctorial
(Corollary 10.1.24). 

We extend the leg notation to the operators that occur in diagram (10.7) and
write

VŒ12 for V ˝ 1 and V ˝ 1I VŒ23 for 1 ˝ V and 1 ˝ V I

op op

VŒ13 for †Œ23 .V ˝ 1/.1 ˝ †/:

Thus each of the symbols VŒ12 and VŒ23 simultaneously denotes two different
operators, which have in common that they act like V on the first and second or on
the second and third factor of different relative tensor products. Now the diagram
(10.7) commutes if and only if V satisfies the pentagon equation VŒ12 VŒ13 VŒ23 D
VŒ23 VŒ12 .

Definition 10.3.4. A unitary V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H is pseudo-multiplicative

op

if it satisfies the intertwining conditions (10.6) and if diagram (10.7) commutes.

Remark 10.3.5. Let V W HˇO ˝ ˛ H ! H˛ ˝ ˇH be a pseudo-multiplicative

op

unitary. Then also the composition

† V †
V op WD †V  † W Hˇ ˝ ˛ H 
! H˛ ˝ ˇ H ! HˇO ˝ ˛ H 
! H˛ ˝ ˇO H
op op

is a pseudo-multiplicative unitary, as one can easily check; here, the rôles of ˇ and ˇO
get reversed. This unitary is called the opposite of V .

10.3.2 The legs of a pseudo-multiplicative unitary

To every pseudo-multiplicative unitary, one can associate two Hopf–von Neumann
bimodules – the left and the right leg of that unitary. This construction generalizes
the construction presented in Section 7.2, where we associated to every multiplica-
tive unitary two von Neumann bialgebras. Throughout this subsection, we consider
a fixed pseudo-multiplicative unitary

V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H;
op
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 317

where N; ; H , and ˛; ˇ; ˇO are as in Assumption 10.3.1.

O /0 . Consider the von Neumann

The von Neumann bimodules ˇ.N /0 and ˇ.N
algebras

ˇ.N /0 D Lˇ.N / .H /  L.H / and O /0 D L O .H /  L.H /:

ˇ.N ˇ .N /

Since ˛, ˇ, ˇO commute pairwise, we can regard

• ˇ.N /0 as a von Neumann N -N -bimodule via ˛; ˇO W N ! ˇ.N /0 , and
O /0 as a von Neumann N op -N op -bimodule via ˇ; ˛ W N ! ˇ.N
• ˇ.N O /0 .

In particular, we can form the fiber products

ˇ.N /0  ˇ.N /0 D .ˇ.N / N̋ ˇ.N //0  L.HˇO ˝ ˛ H /

and
ˇ.N O /0 D .ˇ.N
O /0  ˇ.N O //0  L.H˛ ˝ ˇ H /;
O / N̋ ˇ.N
op op op

and consider these fiber products as a von Neumann N -N -bimodule and a von
Neumann N op -N op -bimodule, respectively, via the maps

˛Œ1 ; ˇOŒ2 W N ! ˇ.N /0  ˇ.N /0

and
O /0  ˇ.N
ˇŒ1 ; ˛Œ2 W N ! ˇ.N O /0 :
op

The comultiplications  y and . Like a multiplicative unitary, the pseudo-multi-

plicative unitary V gives rise to two maps y and  which define the comultiplication
on the left and on the right leg of V , respectively.
Proposition 10.3.6. The maps
y D
 y V W ˇ.N /0 ! L.H O ˝ ˛ H /; y 7! V  .1 ˝ y/V;
ˇ op

and
O /0 ! L.H˛ ˝ ˇ H /; z 7! V .z ˝ 1/V  ;
 D V W ˇ.N
op

define injective morphisms of von Neumann bimodules

O /0 ! ˇ.N
y W ˇ.N /0 ! ˇ.N /0  ˇ.N /0 and  W ˇ.N
 O /0
O /0  ˇ.N (10.8)
op
318 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

and satisfy
y  id/ ı 
. y D .id  /
y ı ;
y . op id/ ı  D .id op / ı :

y and  are injective normal -homomorphisms because the

Proof. The maps 
0
maps ˇ.N / ! L.H˛ ˝ ˇ H / and ˇ.N O /0 ! L.H O ˝ ˛ H / given by y 7!
op ˇ
1 ˝ y and z 7! z ˝ 1, respectively, are injective normal -homomorphisms
op

(Proposition 10.1.23) and because V is unitary.

y
Let us show that .ˇ.N /0 / is contained in ˇ.N /0  ˇ.N /0 . Using the intertwin-
ing relations (10.6), we find that for all x1 ; x2 2 N and y 2 ˇ.N /0 , the operators
y
.y/ and ˇ.x1 / ˝ ˇ.x2 / commute:
O 2 / ˝ y/V
.ˇ.x1 / ˝ ˇ.x2 //V  .1 ˝ y/V D V  .ˇ.x1 /ˇ.x
op op


D V .1 ˝ y/V .ˇ.x1 / ˝ ˇ.x2 //:
op

y
Thus, .ˇ.N /0 /  .ˇ.N / N̋ ˇ.N //0 D ˇ.N /0  ˇ.N /0 . A similar argument shows
O /0 / is contained in ˇ.N
that .ˇ.N O /0 .
O /0  ˇ.N
op

The intertwining relations (10.6) immediately imply that for all x 2 N ,

y
.˛.x// D ˛Œ1 .x/; O
y ˇ.x//
. D ˇOŒ2 .x/;
.˛.x// D ˛Œ2 .x/; .ˇ.x// D ˇŒ1 .x/:

Therefore, y and  are morphisms of von Neumann bimodules as claimed.

The coassociativity equations follow from the pentagon equation for V ; the
proof is similar as in the case of a multiplicative unitary (see Lemma 7.2.1). 

The spaces Ayalg .V / and Aalg .V /. The legs of V are defined in terms of operators
on H that are constructed similarly as in the case of a multiplicative unitary in
Lemma 7.2.7. Let us begin with the left leg. By Proposition 10.1.22, we can define
for each  2 D.˛ H; / and  0 2 D.ˇ H; op / the following operators:
O
jiˇ ;˛
Œ2
W H ! HˇO ˝ ˛ H; j 0 i˛;ˇop Œ2 W H ! H˛ ˝ ˇ H;
op

 7!  ˝ ;  7!  ˝  0 ;
op

and
O
aO . 0;/ WD h 0 j˛;ˇop Œ2 V jiˇ ;˛
Œ2
W H ! HˇO ˝ ˛ H ! H˛ ˝ ˇ H ! H;
op
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 319

where h 0 j˛;ˇop Œ2 D .j 0 i˛;ˇop Œ2 / . The operator aO . 0;/ is characterized by the relation

h 0 jaO . 0;/ i D h 0 ˝  0 jV . ˝ /i for all ;  0 2 H; (10.9)

op

and can be written in the form aO . 0;/ D .id  ! 0; /.V /, where id  ! 0; denotes the
generalized slice map
O
id  ! 0; W L.HˇO ˝ ˛ H; H˛ ˝ ˇ H / ! L.H /; X 7! h 0 j˛;ˇop Œ2 X jiˇ ;˛
Œ2
:
op

Symmetrically, we define for each  2 D.HˇO ; op

/ and 0 2 D.H˛ ; /

O
jiˇ ;˛
Œ1
W H ! HˇO ˝ ˛ H; j0 i˛;ˇop Œ2 W H ! H˛ ˝ ˇ H;
op

0
 7!  ˝ ;  7!  ˝ ;
op

and
O
a.0;/ WD h0 j˛;ˇop Œ2 V jiˇ ;˛
Œ1
W H ! HˇO ˝ ˛ H ! H˛ ˝ ˇ H ! H;
op

where h0 j˛;ˇop Œ2 D .j0 i˛;ˇop Œ2 / . Similarly as aO . 0;/ , the operator a.0;/ is charac-
terized by the equation

h 0 ja.0;/ i D h0 ˝  0 jV . ˝ /i for all ;  0 2 H;

op

and can be written in the form a.0;/ D .!0;  id/.V /, where

O
!0;  id W L.HˇO ˝ ˛ H; H˛ ˝ ˇ H / ! L.H /; X 7! h0 j˛;ˇop Œ2 X jiˇ ;˛
Œ1
:
op

We consider the spaces

Ayalg .V / WD spanfaO . 0;/ j  2 D.˛ H; /;  0 2 D.ˇ H; op

/g  L.H /
and
Aalg .V / WD spanfa.0;/ j  2 D.HˇO ; op
/; 0 2 D.H˛ ; /g  L.H /:

The intertwining relations of V imply the following inclusions:

Lemma 10.3.7. i) ˇ.NO /Ayalg .V /˛.N / D Ayalg .V /  ˇ.N /0 .

O /0 .
ii) ˛.N /Aalg .V /ˇ.N / D Aalg .V /  ˇ.N
320 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Proof. i) Let  2 D.˛ H; /,  0 2 D.ˇ H; op

/, and x; y; z 2 N . We show that

O
ˇ.x/ˇ.y/  aO . 0;/  ˛.z/ D aO .# 0;#/  ˇ.x/; where # D ˇ.y/; # 0 D ˛.z/  0 ;

and this implies the assertion. Note that # 2 D.˛ H; / and # 0 2 D.ˇ H; op
/ by
Proposition 10.1.12. Let ;  0 2 H . By equation (10.9),

O
h 0 jˇ.x/ˇ.y/ O  /ˇ.x  / 0 jaO . 0 ;/ ˛.z/i
 aO . 0 ;/  ˛.z/i D hˇ.y
O  /ˇ.x  / 0 ˝  0 jV .˛.z/ ˝ /i;
D hˇ.y
op

and the intertwining relations (10.6) imply that this is equal to

h 0 ˝ ˛.z/  0 jV .ˇ.x/ ˝ ˇ.y//i D h 0 jaO .# 0;#/ ˇ.x/i:

op

ii) The proof is similar to the proof of i). 

The subspaces Ayalg .V / and Aalg .V / of L.H / are non-degenerate in the follow-
ing sense:

Lemma 10.3.8. The sets Ayalg .V /H; Ayalg .V / H and Aalg .V /H; Aalg .V / H are
linearly dense in H .

Proof. The proof is essentially the same as in the case of a multiplicative unitary,
see Lemma 7.2.2. 

The left-right symmetry. As in the setting of multiplicative unitaries, the preced-

ing definitions are symmetric in the following sense:

Lemma 10.3.9. We have

Ayalg .V op / D Aalg .V / ; Aalg .V op / D Ayalg .V / ;

y V op D Ad† ıV ;
 V op D Ad† ı yV :

Proof. The proof is essentially the same as in the case of a multiplicative unitary,
see Lemma 7.2.5. 

Hopf–von Neumann bimodules associated to V . Let us study various closures

of the spaces Ayalg .V / and Aalg .V /. First, consider the von Neumann algebras

Byw .V / WD .Ayalg .V / C Ayalg .V / /00 and Bw .V / WD .Aalg .V / C Aalg .V / /00 :

 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 321

Proposition 10.3.10. i) We have .˛.N / [ ˇ.NO //  Byw .V /  ˇ.N /0 and .ˇ.N / [

O /.
˛.N //  Bw .V /  ˇ.N 0

O /0 that
ii) The commutant .Byw .V //0 is equal to the set of all T 2 ˛.N /0 \ ˇ.N
0
satisfy .T ˝ 1/V D V .T ˝ 1/, and the commutant .Bw .V // is equal to the set of
op

all T 2 ˇ.N /0 \ ˛.N /0 that satisfy .1 ˝ T /V D V .1 ˝ T /.

op

O and Bw .V /
Consider Byw .V / as a von Neumann N -N -bimodule via ˛ and ˇ,
as a von Neumann N -N -bimodule via ˇ and ˛.
op op

y and  restrict to normal -homomorphisms

iii) The maps 
y W Byw .V / ! Byw .V /  Byw .V / and  W Bw .V / ! Bw .V /  Bw .V /:
 op

Proof. Assertion i) follows easily from Lemma 10.3.7 and Lemma 10.3.8, and the
proof of assertions ii) and iii) is essentially the same as in the case of a multiplicative
unitary, see Lemma 7.2.9. 
The preceding proposition immediately implies the following main result:
Theorem 10.3.11. Let V W HˇO ˝ ˛ H ! H˛ ˝ ˇH be a pseudo-multiplicative
op

O /
unitary. Then the tuples .N; Byw .V /; ˛; ˇ; y and .N op ; Bw .V /; ˇ; ˛; / are Hopf–
von Neumann bimodules. 
The closures of the spaces Ayalg .V / and Aalg .V / with respect to the norm or
the weak operator topology are more difficult to analyze than in the setting of
multiplicative unitaries. For the proofs of the following statements, we refer to the
literature. Put
y / WD k  k-closure of Ayalg .V /;
A.V A.V / WD k  k-closure of Aalg .V /;
Ayw .V / WD w-closure of Ayalg .V /; Aw .V / WD w-closure of Aalg .V /;
where “w-closure” denotes the closure with respect to the weak operator topology.
y / and A.V / (and
Proposition 10.3.12 ([44, Proposition 3.6]). The spaces A.V
y
hence also Aw .V / and Aw .V /) are non-degenerate subalgebras of L.H /.
As for multiplicative unitaries, we use the following terminology:
Definition 10.3.13. A pseudo-multiplicative unitary V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H
op

is weakly well-behaved if Ayw .V / and Aw .V / are von Neumann algebras.

By Proposition 10.3.12, a pseudo-multiplicative unitary V is well-behaved as
soon as Ayw .V / D Ayw .V / and Aw .V / D Aw .V /. In that case, Ayw .V / D
Byw .V / and Aw .V / D Bw .V /. Thus, Theorem 10.3.11 implies:
322 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Corollary 10.3.14. If V W HˇO ˝ ˛ H ! H˛ ˝ ˇH is a weakly well-behaved

op

O /,
pseudo-multiplicative unitary, then .N; Ayw .V /; ˛; ˇ; y .N op ; Aw .V /; ˇ; ˛; /
are Hopf–von Neumann bimodules. 
Let us mention two classes of weakly well-behaved pseudo-multiplicative uni-
taries: the weakly regular ones, and the manageable ones. For a discussion of
manageability, see [99].

Weak regularity. The concept of regularity was carried over from multiplicative
unitaries to pseudo-multiplicative unitaries by Enock [44]. The definition involves
operators of the following form. Given ; 0 2 D.˛ H; / D D.H˛ ; /, we can
consider the composition
O
c.0;/ WD h0 j˛;ˇop Œ1 V jiˇ ;˛
Œ2
W H ! HˇO ˝ ˛ H ! H˛ ˝ ˇ H ! H;
op

which is characterized by the equation h 0 jc.0;/ i D h0 ˝  0 jV . ˝ /i for all

;  0 2 H . Let
op

Calg .V / WD spanfc.0;/ j 0 ;  2 D.H˛ ; /g  L.H /;

and denote by C .V / and Cw .V / the closure of Calg .V / with respect to the norm
topology and the weak operator topology, respectively.
Lemma 10.3.15. i) Calg .V /  ˛.N /0 .
ii) The sets Calg .V /H and Calg .V / H are linearly dense in H .
Proof. The proof of i) is similar to the proof of Lemma 10.3.7, and the proof of ii)
is similar to the proof of the last statement of Lemma 7.2.2. 
Recall that to each  2 D.˛ H; / D D.H˛ ; /, we associated an operator
R ./ D L op ./ W H ! H (Definition 10.1.8). This operator occurs in the
following definition.
Definition 10.3.16. A pseudo-multiplicative unitary V W HˇO ˝ ˛ H ! H˛ ˝ ˇ H
op

is weakly regular if Cw .V / D ˛.N /0 , and regular if

C .V / D span fL op ./L op .0 / j 0 ;  2 D.H˛ ; /g:

The following results generalize Lemma 7.3.5 and Theorem 7.3.10, however,
the proofs involve new techniques.
Proposition 10.3.17 ([44, Proposition 3.10]). The space C.V / .and hence also
Cw .V // is a subalgebra of L.H /.
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 323

Theorem 10.3.18 ([44, Proposition 3.12]). Let V W HˇO ˝ ˛ H ! H˛ ˝ ˇH be

op

a pseudo-multiplicative unitary. If Cw .V / D Cw .V / , then V is weakly well-

behaved.

10.3.3 The pseudo-multiplicative unitary of a groupoid

The prototypical example of a pseudo-multiplicative unitary is the unitary associated
to a locally compact groupoid. After a review of some preliminaries, we construct
this unitary and determine the associated Hopf–von Neumann bimodules. The
original references are [171], [172].

Preliminaries. Let us briefly recall some concepts related to groupoids and inte-
gration over groupoids. Standard references are [120], [129].
A groupoid is a small category in which every morphism is invertible. Equiv-
alently, a groupoid consists of a set of morphisms or arrows G, a set of objects or
units G 0  G, two maps r; s W G ! G 0 called the range and source map, and a
composition map Gs;r 2
! G, where Gs;r 2
D f.x; y/  G  G j s.x/ D r.y/g,
subject to several conditions (see [129, Definition 1.1]). A topological groupoid is
a groupoid equipped with a topology on its set of morphisms for which the inver-
sion and the composition are continuous; then also the range and source map are
continuous.
Let G be a topological groupoid that is locally compact, Hausdorff, and second
countable. To perform translation-invariant integration on G, we need an analogue
of the Haar measure of a locally compact group. The precise definition of this
analogue involves the fibers of the range and the source map, which we denote by
G u WD r 1 .u/ and Gu WD s 1 .u/ for each u 2 G 0 . A left Haar system on G is a
family of Borel measures  D .u /u2G 0 such that
i) for each u 2 G 0 , u is a regular Borel measure on G u with support G u ;
R
ii) for each f 2 Cc .G/, the function G 0 ! C, u 7! G u f d u , is continuous;

iii) for each x 2 G and f 2 Cc .G r.x/ /,

Z Z
f .y/ d  .y/ D
r.x/
f .xy 0 / d s.x/ .y 0 /:
G r.x/ G s.x/

Denote by i W G ! G the inversion map x 7! x 1 . For each u 2 G 0 , define the

measure 1 1
u on Gu as the push-forward of  along i , that is, u D i . /. The
u u
1 1
family  D .u /u2G 0 is called the right Haar system associated to . Haar
systems need not exist and need not be unique, but in many examples, they arise
naturally.
324 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Let  be a Radon measure on the unit space G 0 , and define a measure  on G

via
Z Z Z
f d WD f .x/ d u .x/ d.u/ for all f 2 Cc .G/:
G G0 Gu

The push-forward of  via the inversion map i is denoted by  1 ; evidently,

Z Z Z
f d 1 D f .x/d 1
u .x/ d.u/:
G G0 Gu

The measure  is called quasi-invariant if  and  1 are equivalent. There always

exist sufficiently many quasi-invariant measures [129, I, Proposition 3.6].

Hilbert modules associated to a groupoid. Let G be a topological groupoid that

is locally compact, Hausdorff, and second countable. Furthermore, let  be a left
Haar system on G and  a quasi-invariant Radon measure on G 0 , and define  and
 1 as above. From now on, we denote the range and the source map of G by
rG and sG , respectively, and reserve the letters r and s for certain representations
constructed out of rG and sG .
The measure  onR G 0 defines an n.s.f. weight on the von Neumann algebra
1
L .G 0 ; / via f 7! G 0 f d. Let us denote this weight by  again.
The Hilbert space L2 .G; / can be considered as a Hilbert module over the
Neumann algebra L1 .G 0 ; / via the -homomorphisms
r W L1 .G 0 ; / ! L1 .G; / and s W L1 .G 0 ; / ! L1 .G; /;
.r.f //.x/ WD f .rG .x//; .s.f //.x/ WD f .sG .x//;
and the representation L1 .G; / ! L.L2 .G; // given by multiplication opera-
tors. We denote by L2 .G; /r and L2 .G; /s the Hilbert L1 .G 0 ; /-module given
by L2 .G; / and the representation r or s, respectively.
By construction, the Hilbert module L2 .G; /r is a direct integral:
Z ˚
L .G; /r D
2
L2 .G u ; u /d.u/:
G0

The Hilbert module L .G; /s can be described similarly. Denote by D WD

2

d=d 1 the Radon–Nikodym derivative of  with respect to  1 . Then

Z Z Z
f .x/d.x/ D f .x/D.x/d 1
u .x/d.u/ for all f 2 L .G; /
1
G G0 Gu

and
Z ˚
L2 .G; /s D L2 .Gu ; Du 1
u /d.u/; where Du D DjGu :
G0
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 325

For later use, we identify the bounded elements of the Hilbert modules L2 .G; /r
and L2 .G; /s . By direct calculations or an application of Proposition 10.1.17, we
find:
Lemma 10.3.19. i) An element  2 L2 .G; R /r is bounded with respect to  if and
only if the function on G 0 given by u 7! G u j.x/j2 d u .x/ belongs to L1 .G 0 ; /.
For ;  0 2 D.L2 .G; /r ; /, the product h 0 ji 2 L1 .G 0 ; / is given by u 7!
R
0 u
G u  .x/.x/d  .x/.
ii) An element  2 L2 .G; / R s is bounded with respect to  if and only if the
function on G given by u 7! Gu j.x/j D.x/d 1
0 2 1
u .x/ belongs to L .G ; /.
0
0 0 1
For ;  2 D.L .G; /s ; /, the product h ji 2 L .G ; / is given by u 7!
2 0
R 1
0 
Gu  .x/.x/D.x/d u .x/.

The pseudo-multiplicative unitary. We construct a pseudo-multiplicative unitary

WG W L2 .G; /s ˝ r L2 .G; / ! L2 .G; /r ˝ r L2 .G; /I (10.10)

 

in the notation of Definition 10.3.4,

N D L1 .G 0 ; / D N op ; H D L2 .G; /;
DD op
; ˛ D r; ˇ D r; ˇO D s:

The relative tensor products that appear in (10.10) can conveniently be described
2 2
in terms of the measure s;r on Gs;r given by
Z Z Z Z
2
f ds;r WD f .x; y/ d sG .x/ .y/ d u .x/ d.u/
2
Gs;r G0 Gu G sG .x/

2
and the measure r;r 2
on Gr;r D f.x; y/ 2 G 2 j rG .x/ D rG .y/g given by
Z Z Z Z
g dr;r WD
2
g.x; y/ d u .y/ d u .x/ d.u/;
2
Gr;r G0 Gu Gu

where f 2 Cc .Gs;r
2
/ and g 2 Cc .Gr;r
2
/:
Lemma 10.3.20. For k D s; r, respectively, the map

ˆk;r W L2 .G; /k ˝ r L2 .G; / ! L2 .Gk;r

2 2
; k;r /

given by
.ˆk;r . ˝ //.x; y/ WD .x/.y/;


where  2 L2 .G; / and  2 D.L2 .G; /r ; /, is an isometric isomorphism.

326 Chapter 10. Pseudo-multiplicative unitaries on Hilbert spaces

Proof. We prove the assertion for k D s; the case k D r is similar. Let  2

L2 .G; /;  2 D.L2 .G; /r ; /, and put  WD  ˝ . By Lemma 10.3.19 i),


kk2 D hjs.hji /i

Z Z
D .x/.hji .sG .x///.x/d u .x/ d.u/
G0 Gu
Z Z Z
D j.x/j2 j.y/j2 d sG .x/ .y/ d u .x/ d.u/
G 0 G u G sG .x/
Z
D j.ˆs;r .//.x; y/j2 ds;r
2
.x; y/
2
Gs;r

D kˆs;r ./k2 :

This calculation implies that ˆs;r is well defined and isometric, and it is easy to see
that this map has dense image. 

Alternatively, the preceding result can also be deduced from Proposition 10.1.17.
From now on, we identify L2 .G; /k ˝ r L2 .G; / with L2 .Gk;r 2 2
; k;r / (k D r; s)

via the map ˆk;r .
Proposition 10.3.21. The map

WG W L2 .G; /s ˝r L2 .G; / ! L2 .G; /r ˝r L2 .G; /

 

given by

.WG /.x; y/ D .x; x 1 y/ for all  2 L2 .Gs;r

2 2
; s;r /; .x; y/ 2 Gr;r
2
;

is a pseudo-multiplicative unitary.
Proof. The fact that WG is a unitary follows easily from Lemma 10.3.20 and from the
left-invariance of the Haar system . To verify the pentagon equation, one describes
the iterated relative tensor products of L2 .G; / that occur in the pentagon diagram
(10.7) in terms of measures on iterated fiber products of G, similarly as it was done
in Lemma 10.3.20 for simple relative tensor products, and uses a similar calculation
as in Example 7.1.4. 

Let us identify the legs of WG . Recall that the von Neumann algebra of G is
the von Neumann algebra LG  L.L2 .G; // generated by operators of the form
L.g/, where g 2 L1 .G; /,
Z
.L.g//.x/ D g.y/.y 1 x/d rG .x/ .y/ for all x 2 G;  2 L2 .G; /:
G rG .x/
 10.3. Pseudo-multiplicative unitaries on Hilbert spaces 327

Proposition 10.3.22. Ayw .WG / D L1 .G; / and Aw .WG / D LG; in particular,

the pseudo-multiplicative unitary WG is weakly well-behaved.
Proof. This follows from similar calculations and arguments as in Example 7.2.13,
where G was a group. 
By Theorem 10.3.11, we obtain Hopf–von Neumann bimodules
y
.L1 .G 0 ; /; L1 .G; /; r; s; / and .L1 .G 0 ; /; LG; r; r; /:

Let us describe the comultiplications

y W L1 .G; / ! L1 .G; /s  r L1 .G; /;
 f 7! WG .1 ˝ f /WG ;
 
and
 W LG ! LGr  r LG; L.g/ 7! WG .L.g/ ˝ 1/WG :
 

Using the isomorphisms of Lemma 10.3.20, we identify L1 .G; /s  r L1 .G; /


and LGr  r LG with von Neumann algebras on the Hilbert space L2 .Gk;r
2 2
; k;r /Š

L2 .G; /k ˝ r L2 .G; /, where k D s or k D r, respectively.


Proposition 10.3.23. i) For all f 2 L1 .G; /;  2 L2 .Gs;r

2 2
; s;r /; .x; y/ 2 Gs;r
2
,

y //.x; y/ D f .xy/.x; y/:

..f

ii) For all g 2 L1 .G; /;  2 L2 .Gr;r

2 2
; r;r /, .x; y/ 2 Gr;r
2
,
Z
..L.g///.x; y/ D g.z/.z 1 x; z 1 y/d rG .x/ .z/:
G rG .x/

Proof. Again, this follows from similar calculations as in Example 7.2.13, where
G was a group. 
Finally, we note:
Proposition 10.3.24. WG is regular.
Proof. This follows from similar calculations as in Example 7.3.4 iii); a detailed
proof can be found in [44, Proposition 4.8]. 
Chapter 11
Pseudo-multiplicative unitaries on C  -modules

In contrast to the theory of measurable quantum groupoids, a general theory of

locally compact quantum groupoids in the setting of C  -algebras does not exist
yet. Such a theory should be based on C  -algebraic analogues of the following von
Neumann-algebraic concepts:

• pseudo-multiplicative unitaries on Hilbert spaces,

• Hopf–von Neumann bimodules, and
• operator-valued weights.

Such C  -algebraic analogues of the first and second concepts have recently been
introduced and studied in [154], [155], [156].
In this book we restrict ourselves to a special class of “decomposable” locally
compact quantum groupoids or, more precisely, to a special class of “decompos-
able” pseudo-multiplicative unitaries on C  -modules. For classical groupoids, this
implies a restriction to groupoids that are r-discrete or extensions of r-discrete
groupoids by group bundles.
We proceed as follows. First, we introduce pseudo-multiplicative unitaries on
C  -modules (Section 11.1). When we try to define the legs of such a pseudo-mul-
tiplicative unitary, we immediately encounter several problems that we can only
solve under a certain decomposability assumption.
Tailored to this decomposability assumption, we develop a general calculus of
homogeneous operators on C  -bimodules (Section 11.2) and introduce C  -fami-
lies of such operators, which generalize C  -algebras. Next, we define internal
tensor products and morphisms of C  -families, which enter the definition of a Hopf
C  -family (Section 11.3).
The legs of a decomposable regular pseudo-multiplicative unitary can then be
constructed in the form of Hopf C  -families (Section 11.4). Moreover, one can
define coactions of Hopf C  -families, and relate coactions of the Hopf C  -families
associated to a decomposable groupoid to actions and Fell bundles of that groupoid
(Section 11.5).
The results and concepts presented in this chapter are part of the author’s thesis
[152]. For further details, we refer to this thesis, the article [153], and forthcoming
publications.
 11.1. Pseudo-multiplicative unitaries on C  -modules 329

11.1 Pseudo-multiplicative unitaries on C  -modules

Pseudo-multiplicative unitaries on C  -modules are natural generalizations of mul-
tiplicative unitaries and analogues of pseudo-multiplicative unitaries on Hilbert
spaces. The definition and the main example – the pseudo-multiplicative unitary
of a locally compact groupoid – are given in the first sections. Thereafter, we dis-
cuss the question how to define the legs of such a pseudo-multiplicative unitary,
and outline a strategy for a partial solution that will be implemented in the fol-
lowing sections. Let us note that a general solution based on a better notion of a
pseudo-multiplicative unitary was recently proposed in [154].

11.1.1 The flipped internal tensor product of C  -modules

The definition of a pseudo-multiplicative unitary on a C  -module involves the
internal tensor product of C  -modules, which is reviewed in Appendix 12.2, and
the flipped internal tensor product of C  -modules, which is defined as follows.
Let A and B be C  -algebras, let E be a C  -module over B with a fixed repre-
sentation A ! LB .E/, and let F be a C  -module over A. Then one can form the
internal tensor product F ˝A E, whose structure maps are given by

h0 ˝A  0 j ˝A i.F ˝A E / D h 0 jh0 jiF iE and . ˝A /b D  ˝A b

for all ; 0 2 F , ;  0 2 E, and b 2 B. It will be convenient to denote the internal

tensor product by “”; thus, for example, F  E D F ˝A E.
Similarly, we can construct a flipped internal tensor product E  F : We equip
the algebraic tensor product E ˇ F with the structure maps h 0 ˇ 0 j ˇ i WD
h 0 jh0 jiF iE and . 0 ˇ 0 /b WD  0 b ˇ 0 , factor out the null space of the semi-
norm  7! khjik, form the completion with respect to the induced norm on the
quotient, and obtain a C  -module over B which we call the flipped internal tensor
product E  F . Thus, E  F is the closed linear span of elements   , where
 2 E and  2 F , and

h 0  0 j  i D h 0 jh0 jiF iE and .  /b D b  

for all ;  0 2 E, ; 0 2 F , and b 2 B. Evidently, we have a unitary map

Š
†W F  E 
! E  F;    7!   :

The flipped internal tensor product E F can be interpreted as follows. The right
C  -B-module E and the right C  -A-module F can be considered in a canonical
way as a left C  -B op -module and a left C  -Aop -module, respectively, and the
representation of A on E can be considered as an antirepresentation of Aop on E.
Then, the flipped internal tensor product E  F is just the natural internal tensor
330 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

product of the left C  -modules E and F . However, we consider this internal tensor
product as a right C  -module again.
If we want to emphasize that the (flipped) internal tensor product F  E or
E  F is formed with respect to a fixed representation
W A ! LB .E/, we denote
these tensor products by F  E or E  F , respectively.

11.1.2 Definition and examples

Recall that a multiplicative unitary on a Hilbert space H is a unitary V W H ˝ H !
H ˝H that satisfies the pentagon equation VŒ12 VŒ13 VŒ23 D VŒ23 VŒ12 . We extend
this concept, replacing H by a C  -module E with representations ˇ; O ˇ.

O ˇ/ over B is a full C  -B-module E with

Definition 11.1.1. A C  -trimodule .E; ˇ;
two commuting nondegenerate faithful representations ˇ; O ˇ of B on E.

O ˇ/ be a C  -trimodule over B. Then we can form (flipped) internal

Let .E; ˇ;
tensor products E  ˇO E, E  ˇ E and EˇO  E, Eˇ  E. Roughly, a pseudo-
multiplicative unitary on E is a unitary

V W EˇO  E ! E  ˇ E

that satisfies the pentagon equation VŒ12 VŒ13 VŒ23 D VŒ23 VŒ13 . Similarly as in the
case of a pseudo-multiplicative unitary on a Hilbert space, the individual factors VŒij 
act on iterated internal tensor products. Therefore, several intertwining conditions
on V and ˇ; ˇO have to be assumed to ensure that each factor VŒij  is well defined.
Since ˇ and ˇO commute, we can define representations

ˇŒ1 ; ˇOŒ1 ; ˇŒ2 on E  ˇO E; ˇŒ1 ; ˇOŒ1 ; ˇOŒ2 on E  ˇ E;

ˇŒ1 ; ˇŒ2 ; ˇOŒ2 on EˇO  E; ˇOŒ1 ; ˇOŒ2 ; ˇŒ2 on Eˇ  EI

for example, the first three representations are given by

ˇŒ1 .b/ WD ˇ.b/  1; ˇOŒ1 .b/ WD ˇ.b/

O  1; ˇŒ2 .b/ WD 1  ˇ.b/

for all b 2 B.

Lemma 11.1.2. Let V 2 LB .EˇO  E; E  ˇ E/ and assume that for all b 2 B,

VˇŒ2 .b/ D ˇOŒ1 .b/V; VˇŒ1 .b/ D ˇŒ1 .b/V; V ˇOŒ2 .b/ D ˇOŒ2 .b/V: (11.1)
 11.1. Pseudo-multiplicative unitaries on C  -modules 331

Then all operators in the following diagram are well defined:

E  ˇ EˇO  E
llll5 TTTT
TTTT
l l lllV 1 1 V TTT)
l
EˇO  EˇO  E E  ˇE  ˇE
66 B
66 
661V V 1 
66 
  (11.2)
EˇO  .E  ˇ E/ .EˇO  E/ O  ˇE

1† †Œ23

EˇO  Eˇ  E V 1
/ .E  ˇ E/ˇOŒ1  E.

Here, †Œ23 denotes the isomorphism given by .1  2 /  3 7! .1  3 /  2 .

Proof. This follows from Proposition 12.2.1. 

We extend the leg notation to the operators occurring in diagram (11.2), and
write

VŒ12 for V  1 and V  1I VŒ23 for 1  V and 1  V I

VŒ13 for †Œ23 .V  1/.1  †/:

Then diagram (11.2) commutes if and only if VŒ12 VŒ13 VŒ23 D VŒ23 VŒ12 .

O ˇ/ be a C  -trimodule over B. We call a unitary

Definition 11.1.3. Let .E; ˇ;
V 2 LB .EˇO  E; E  ˇ E/ pseudo-multiplicative if it satisfies equation (11.1) and
if diagram (11.2) commutes.

Evidently, this definition generalizes the notion of a multiplicative unitary on

Hilbert spaces and translates the concept of a pseudo-multiplicative unitary on
a Hilbert space into the language of C  -modules. For commutative B, Defini-
tion 11.1.3 subsumes the following special cases:

i) If ˇ.b/ D b D ˇ.b/O for all  2 E and b 2 B, then V is a continuous

field of multiplicative unitaries as defined by Blanchard [17].

O
ii) If ˇ.b/ D b for all  2 E and b 2 B, then V is a pseudo-multiplicative
unitary in the sense of O’uchi [116].
332 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Remark 11.1.4. Let .E; ˇ; O ˇ/ and V W E O E ! E ˇ E be as in Definition 11.1.3.

ˇ
O is a C  -trimodule over B, and the unitary
Then .E; ˇ; ˇ/
† V †
V op WD †V  † W Eˇ  E 
! E  ˇ E ! EˇO  E 
! E  ˇO E
is pseudo-multiplicative, as one can easily check. The unitary V op is called the
opposite of V .
Let us consider the fundamental example of a pseudo-multiplicative unitary,
discussed already in [116] – the unitary associated to a locally compact groupoid.
For background on groupoids and Haar systems, see Section 10.3.3.
Example 11.1.5. Let G be a locally compact, second countable, Hausdorff groupoid
with left Haar system . We denote its unit space by G 0 , its range map by rG , its
1
source map by sG , and put G u WD rG .u/ for each u 2 G 0 .
Let B WD C0 .G 0 /. Denote by L2 .G; / the C  -module over B associated to G
and ; this is the completion of the pre-C  -module Cc .G/, whose structure maps
are given by
Z
h 0 ji.u/ D  0 .x/.x/ d u .x/ and .f /.x/ D .x/f .rG .x//
Gu

for all u 2 G 0 ; x 2 G; ;  0 2 Cc .G/, f 2 B. The C  -module L2 .G; /

corresponds to a continuous field of Hilbert spaces [34], [38], [57] over G 0 with
fiber L2 .G u ; u / at u 2 G 0 . Define representations r; s W B ! LB .L2 .G; // by
.r.f //.x/ WD f .rG .x//.x/ and .s.f //.x/ WD f .sG .x//.x/

for all x 2 G;  2 Cc .G/; f 2 B. Then .E; ˇ; O ˇ/ WD .L2 .G; /; s; r/ is a


C -trimodule over B.
The (flipped) internal tensor products EˇO  E and E  ˇ E can be described as
follows. For k D r; s, put Gk;r 2
WD f.x; y/ 2 G  G j kG .x/ D rG .y/g. Consider
2
Cc .Gs;r 2
/ and Cc .Gr;r / as pre-C  -modules over B via the structure maps
Z Z
h 0 ji.u/ WD  0 .x; y/.x; y/ d sG .x/ .y/d u .x/ for Cc .Gs;r
2
/;
Gu G sG .x/
Z Z
h 0 ji.u/ WD  0 .x; y/.x; y/ d u .y/d u .x/ 2
for Cc .Gr;r /;
Gu Gu

.f /.x; y/ WD .x; y/f .rG .x// for both;

2
and denote by L2 .Gs;r / and L2 .Gr;r
2
/ the respective completions. Then we have
natural isomorphisms EˇO  E Š L2 .Gs;r 2
/ and E  ˇ E Š L2 .Gr;r
2
/, similarly as
in Lemma 10.3.20.
 11.1. Pseudo-multiplicative unitaries on C  -modules 333

Now the map W0 W Cc .Gs;r

2
/ ! Cc .Gr;r
2
/ given by

.W0 /.x; y/ WD .x; x 1 y/ for all .x; y/ 2 Gr;r

2
;  2 Cc .Gs;r
2
/;

extends to a pseudo-multiplicative unitary WG W EˇO  E ! E  ˇ E. Indeed, the

map W0 is a linear isomorphism because it is the transpose of a homeomorphism
2
Gr;r ! Gs;r
2
, and it extends to a unitary because
Z Z
hW0 jW0 i.u/ D .x; x 1 y/.x; x 1 y/ d u .y/d u .x/
u u
ZG ZG
D .x; y 0 /.x; y 0 / d sG .x/ .y 0 /d u .x/ D hji.u/
Gu G sG .x/

for all u 2 G 0 and  2 Cc .Gs;r

2
/. To check that WG satisfies the pentagon equation,
one can use a similar calculation as in the case where G is a group, see Example 7.1.4.
The pseudo-multiplicative unitary WG is related to the pseudo-multiplicative
unitary constructed in Section 10.3.3 as follows. Let  be a quasi-invariant Radon
1
Rmeasure on G . Denote also by  the weight on L .G ; / given by f 7!
0 0

G 0 f d, and consider the pseudo-multiplicative unitary

WG W L2 .G; /s ˝ r L2 .G; / ! L2 .G; /r ˝ r L2 .G; /

 

constructed in Proposition 10.3.21. Let B act on L2 .G 0 ; / by multiplication

operators. Then

L2 .G; /k ˝ r L2 .G; / Š L2 .Gk;r

2 2
; k;r / Š L2 .Gk;r
2
/  L2 .G 0 ; /


for k D s; r, and these isomorphisms identify WG with WG  idL2 .G;/ , as one
can easily check.
The following example is a C  -algebraic analogue of a pseudo-multiplicative
unitary on Hilbert spaces considered by Lesieur [99, Section 7.6].
Example 11.1.6. Let B be a unital C  -algebra, C  Z.B/ a C  -subalgebra con-
taining 1B , and  W B ! C a faithful conditional expectation, that is, a faithful pos-
itive C -linear map such that jC D idC . We associate to  a pseudo-multiplicative
unitary W as follows.
First, consider B as a pre-C  -module over C via the inner product ha0 jai WD
 .a0 a/ and via right multiplication, and denote by B the completion. Next,
consider B as a right C  -B-B-bimodule in the natural way, and denote by E WD
B  B the internal tensor product over C . Thus E is generated by elements a  b,
where a; b 2 B, and ha0  b 0 ja  bi D b 0  .a0 a/b, .a  b/b 0 D a  bb 0 for all
a; b; a0 ; b 0 2 B.
334 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Routine arguments show that there exist representations ˇ;O ˇ W B ! LB .E/

O /.a  b/ D b a  b and ˇ.b /.a  b/ D a  b b for all a; b; b 0 2 B;
such that ˇ.b 0 0 0 0

O ˇ/ is a C  -trimodule.
here, we use .B/  Z.B/. Evidently, .E; ˇ;
We claim that there exist unitaries
X W EˇO  E ! B  B  B; .a  b/  .c  d / 7! da  c  b;
Y W E  ˇ E ! B  B  B; .a  b/  .c  d / 7! a  c  bd:
Indeed, for x WD .a  b/  .c  d / and y WD .a  b/  .c  d / as above,
kXxk2 D khbjhcjhdajdaicibik D kb  .c   .a d  da/c/bk
D kb  .a d   .c  c/da/bk
D kha  bjhc  d jc  d ia  bik D kxk2
and
kYyk2 D khbd jhcjhajaicibd ik D kd  b   .c   .a a/c/bd k
D kd  .c  c/b   .a a/bd k
D khc  d jc  ha  bja  bid ik D kyk2 I
here, we use .B/  Z.B/ and  .e .f // D  .e/ .f / for e; f 2 B. Now consider
the unitary W WD Y  X W EˇO  E ! E  ˇ E. Explicitly,
W ..a  b/  .c  d // D .da  b/  .c  1/ for all a; b; c; d 2 B; (11.3)
as can be seen from the relation
Y..da  b/  .c  1// D da  c  b D X..a  b/  .c  d //:
The following calculations show that W satisfies the intertwining conditions in
equation (11.1) and that diagram (11.2) commutes: for a; b; c; d; e; f; g 2 B,
O
.a  b/  .c  d /
 ˇŒ1 .e/ˇŒ2 .f /ˇŒ2 .g/ / .a  eb/  .gc  f d /
_ _
W
W

  
.da  b/  .c  1/ / .f da  eb/  .gc  1/
ˇŒ1 .e/ˇOŒ1 .f /ˇOŒ2 .g/
and
W Œ12 W Œ23
/ .da  b/  .c  1/  .e  f / 
_ 
.a  b/  .c  d /  .e  f / .da  b/  .f c  1/  .e  1/
_ O
WŒ23 W Œ12
 W Œ13 _
 / ..a  b/  .f c  d //  .e  1/:
.a  b/  ..f c  d /  .e  1//
 11.1. Pseudo-multiplicative unitaries on C  -modules 335

11.1.3 Obstructions to the construction of the legs

Given a pseudo-multiplicative unitary V W EˇO  E ! E  ˇ E, it is natural to
ask whether we can define a left and a right leg of V in a similar way as it was
done in Section 7.2 and Section 10.3.2 for every multiplicative unitary and every
pseudo-multiplicative unitary on a Hilbert space. The constructions in Section 7.2
and Section 10.3.2 suggest the following plan:

1. Define C  -algebras A.V y / and A.V / of operators on E, spanned by com-

positions of the form aO . 0;/ D Œ 0 jŒ2 V jiŒ2 and a.0;/ D h0 jŒ1 V jŒ1 ,
respectively, where ;  0 and ; 0 are suitable elements of E, and the maps
jiŒ2 ; j 0 Œ2 D .Œ 0 jŒ2 / and jŒ1 ; j0 iŒ1 D .h0 jŒ1 / have the form

jiŒ2 W E ! EˇO  E;  7!   ; j 0 Œ2 W E ! E  ˇ E;  7!    0 ;

jŒ1 W E ! EˇO  E;  7!   ; j0 iŒ1 W E ! E  ˇ E;  7! 0  :

y /  A.V
2. Define fiber products A.V y / and A.V /  A.V / as C  -algebras of

operators on the C -module EˇO  E or E  ˇ E, respectively.

3. Define comultiplications  y W A.V

y / ! A.V y /  A.Vy / and  W A.V / !
y a/
A.V /  A.V / via the formulas . O D V  .1  a/V
O and .a/ D V .a  1/V  ,
respectively.

If we try to carry out this plan, we encounter the following problems:

y / and A.V / as C  -subalgebras
1. At a first attempt, we might try to define A.V
y
of LB .E/. But then A.V / and A.V / would have to commute with the right
multiplication of B on E, and this excludes many examples:
• For the pseudo-multiplicative unitary WG associated to a locally compact
groupoid G with left Haar system  (Example 11.1.5), the right leg
A.WG / should correspond to the left regular representation of G on
L2 .G; /. But in general, this left regular representation, which is an
integrated form of left multiplication on G, does not commute with the
module multiplication, which is the transpose of the range map of G.
• If V were a pseudo-multiplicative unitary on a Hilbert space, then the
right module multiplication would correspond to the representation ˛,
and Lemma 10.3.7 indicates that the legs of V need not commute with ˛.
Therefore we need a substitute for the C  -algebra LB .E/ that is large enough
to accommodate A.V y / and A.V /.
y / and A.V / can not be defined as C  -subalgebras of LB .E/
The fact that A.V
is related to the problem that the maps j 0 Œ2 and jŒ1 that we wanted to use for
336 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

the definition of the operators aO . 0;/ and a.0;/ are not well-behaved. Indeed,
j 0 Œ2 need not commute with the right module multiplication because

j 0 Œ2 .b/ D b   0 D   ˇ.b/ 0 ¤    0 b D .   0 /b D .j 0 Œ2 /b

unless ˇ.b/ 0 D  0 b for all b 2 B, which is a severe restriction on the choice

O
of  0 . Likewise, jŒ1 is not adjointable unless ˇ.b/ D b for all b 2 B.
y /  A.V
2. It is not clear how to define A.V y / and A.V /  A.V /. In general,
these fiber products can not be spanned by operators of the form

aO 1  aO 2 W    7! aO 1   aO 2  and a1  a2 W    7! a1   a2 ;

y / and a1 ; a2 2 A.V /, because such opera-

respectively, where aO 1 ; aO 2 2 A.V
y /  A.V
tors need not be well defined. If we put A.V y / WD .A.V y /0 /0
y /0  A.V
0 0 0
and A.V /  A.V / WD .A.V /  A.V / / as in the setting of von Neumann
algebras (see Section 10.2.1), we obtain spaces that are too large. Note that
it is also not clear in which ambient C  -algebra the respective commutants
should be taken.

The strategy for a special case. We can solve the problems outlined above and
define the legs of a pseudo-multiplicative unitary on C  -modules only in a special
case, using the following strategy:
The problems listed in Step 1 originate from the fact that the operators involved
are not necessarily adjointable. We consider operators on the C  -module E that are
not strictly adjointable, but adjointable up to a twist by some partial automorphism 
of B (Section 11.2.1). We demand that these operators also commute with the repre-
sentation ˇ or ˇO up to a twist by some partial automorphism of B. Such operators
form graded families L.ˇO E/ D .L .ˇO E//; and L.ˇ E/ D .L .ˇ E//; , which
carry an involution and are closed under multiplication and addition of homoge-
neous elements. Under suitable assumptions, we can then define the left and the
right leg of V in the form of subfamilies A.V y /  L. O E/ and A.V /  L.ˇ E/
ˇ
(Section 11.4).
The assumption that we need to impose is the following. The non-adjointability
of the operators considered in Step 1 is related to the discrepancy between the
right module multiplication on E and the representations ˇO and ˇ. Similarly as
for operators on E, we consider elements of E that intertwine the representation ˇO
or ˇ, respectively, and the right module multiplication up to a twist by some partial
automorphism  of B (Section 11.2.2). These elements form families H.ˇO E/ D
.H .ˇO E// and H.ˇ E/ D .H .ˇ E// , and the assumption that we need to impose
is that these families are linearly dense in E. If this condition is satisfied, we call
the pseudo-multiplicative unitary V decomposable.
 11.2. Semigroup grading techniques on right C  -bimodules 337

11.2 Semigroup grading techniques on right C  -bimodules

In this section, we develop a general calculus of homogeneous operators on C  -bi-
modules and of homogeneous elements of C  -bimodules. Moreover, we introduce
the concept of a C  -family, which can be thought of as a generalized C  -algebra
of homogeneous operators on a C  -module. These concepts will be used in Sec-
tion 11.4 to construct the legs of a decomposable pseudo-multiplicative unitary on
a C  -module.
Before we start, let us fix some terminology and notation.
Recall that a partial automorphism of a C  -algebra C is a -isomorphism
 W Dom./ ! Im. /, where Dom. / and Im. / are closed ideals of C . Compo-
sitions and inverses of partial automorphisms are partial automorphisms again; thus
the set PAut.C / of all partial automorphisms of C forms an inverse semigroup. We
denote the inverse of a partial automorphism  by   . Let ;  0 2 PAut.C /. We
say that  0 extends  and write  0  if Dom. /  Dom. 0 / and  0 jDom./ D  .
We put

 ^  0 WD maxf 00 2 PAut.C / j  00  ;  00   0 gI

thus  ^  0 D jI D  0 jI , where I  Dom. / \ Dom. 0 / is the largest ideal on

which  and  0 coincide.
Throughout this section, let A and B be C  -algebras.
Definition 11.2.1. A right C  -A-B-bimodule is a C  -module E over B with a fixed
non-degenerate -homomorphism A ! LB .E/. Given right C  -A-B-bimodules
E and F , we put

LB
A
.E; F / WD fT 2 LB .E; F / j aT  D T a for all a 2 A;  2 Eg:

11.2.1 Homogeneous operators and C  -families

We consider maps of right C  -bimodules which almost preserve the bimodule
structure:
Definition 11.2.2. Let E; F be right C  -A-B-bimodules and 2 PAut.A/,  2
PAut.B/. We call a map T W E ! F a . ;  /-homogeneous operator if
i) Im.T /  span Im. /F and T a D .a/T  for all a 2 Dom. /,  2 E, and
ii) there exists a map S W F ! E such that hSF jEi  Dom. / and hjT i D
.hS ji/ for all  2 E,  2 F .
Let us collect some first properties of homogeneous operators:
Proposition 11.2.3. Let E, F , ,  and T , S be as above.
338 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

i) T and S are linear and bounded, and kT k D kS k.

ii) T .b/ D .T / .b/ for all b 2 Dom. / and  2 E.
iii) There exists a partial automorphism 0 2 PAut.B/ such that whenever T
is . 0 ;  0 /-homogeneous for some 0 2 PAut.A/ and  0 2 PAut.B/, then
0   0 .
iv) S is uniquely determined by T .
v) Let .u / and .v / be approximate units of the ideals Dom. / and Dom. /,
respectively. Then lim T .u / D T  D lim T .v / for all  2 E.
Proof. i) The proof is completely analogous to the case of ordinary adjointable
operators on C  -modules.
ii) This relation follows from the fact that for all ;  2 E and b 2 Dom. /,

hjT .b/i D  .hSjbi/ D  .hSjib/ D  .hS ji/ .b/ D hj.T / .b/i:

iii) Put J WD ŒhF jTEi  B. Then J is a subspace of Im. /, and an ideal

in B because BJ  ŒhFBjTEi and J Im. /  ŒhF jT .EB/i by ii). Denote
by 0 the restriction of  to   .J /. Assume that T is also . 0 ;  0 /-homogeneous
for some 0 2 PAut.A/ and  0 2 PAut.B/, and that S 0 satisfies condition ii) in
Definition 11.2.2 for T and . 0 ;  0 /. Then

.hS jib/ D hjT .b/i D  0 .hS 0 jib/ for all ;  2 E; b 2 B;

and hence .  .a/b/ D  0 . 0 .a/b/ for all a 2 J; b 2 B. Let .u / be an

approximate unit for J and d 2 J . The last relation and the inclusion J  Im. 0 /
imply

d D lim  0 . 0 .d / 0 .u // D lim  .  .d / 0 .u //:

Therefore,   .d / D lim   .d / 0 .u / 2  0 .J /. Now

d D lim  .  .u /  .d // D lim  0 . 0 .u /  .d // D  0 .  .d //:

Consequently, 0   0 .
iv) Similarly as in the case of ordinary adjointable operators, one finds that S is
uniquely determined by T and  . But by ii), S is independent of  .
v) This follows from standard arguments. 
The preceding proposition justifies the following definition:
Definition 11.2.4. Let E, F be right C  -A-B-bimodules and let T W E ! F be a
. ; /-homogeneous operator, where 2 PAut.A/;  2 PAut.B/. Then the map
S W F ! E in Definition 11.2.2 ii) is called the adjoint of T and denoted by T  .
 11.2. Semigroup grading techniques on right C  -bimodules 339

For later use, we note the following simple example:

Example 11.2.5. Let 2 PAut.B/, and consider Dom. /; Im. /  B as right
sub-C  -bimodules of B. Then 2 L .Dom. /; Im. //. Indeed, condition i)
in Definition 11.2.2 is easily checked, and for condition ii), note that for all b 2
Dom. / and c 2 Im. /,

hcj .b/i D c  .b/ D .  .c  /b/ D .h  .c/jbi/:

Remark 11.2.6. In the situation of Definition 11.2.2, one can consider

E.;/ WD ŒDom. /E Dom. /  E and F .;/ WD ŒIm. /F Im. /  F

as right C  -Dom. /-Dom. /-bimodules, where the structure maps of E.;/ are
inherited from E, and the structure maps of F .;/ are twisted by and  in
a straightforward way. Then every . ;  /-homogeneous operator T W E ! F re-
stricts to an operator T.;/ 2 LDom./
Dom./
.E.;/ ; F .;/ /, whose adjoint is a restriction
of T  .
The preceding remark shows that homogeneous operators generalize ordinary
operators on right C  -bimodules only slightly. The point is that we shall consider
entire families of homogeneous operators:
Notation 11.2.7. Let E, F be right C  -A-B-bimodules and 2 PAut.A/,  2
PAut.B/. We denote the set of all . ;  /-homogeneous operators from E to F by
L .E; F / and put L .E/ WD L .E; E/. The strict topology on L .E; F / is the
topology given by the family of seminorms T 7! kT k and T 7! kT  k, where
 2 E and  2 F . Finally, we put L.E; F / WD .L .E; F //; and L.E/ WD
.L .E//;
The family of all homogeneous operators has the following properties:
Proposition 11.2.8. Let E, F , G be right C  -A-B-bimodules and ; 0 2 PAut.A/,
;  0 2 PAut.B/.
i) L .E; F / is a closed subspace of the space of all bounded linear maps from
E to F , and complete with respect to the strict topology.
0 0
ii) L 0 .F; G/L .E; F /  L 0 .E; G/.

N  , kT  k D kT k D kT  T k1=2 ,
iii) L .E; F / D L  .F; E/, and .T / D T
0
.ST / D T  S  for all  2 C, T 2 L .E; F /, S 2 L 0 .F; G/.
0
iv) Lid
id .E; F / D LB .E; F /, and for each pair of partial identities 2 PAut.A/,
A
0
2 PAut.B/, the space L .E/ is a C  -subalgebra of LB .E/.
A
340 Chapter 11. Pseudo-multiplicative unitaries on C  -modules
 
v) L .E; F / is a right C  -L  
   .F /-L   .E/-bimodule.
0
vi) L .E; F /  L 0 .E; F / if  0 and    0 .
Proof. Most of these assertions generalize facts about ordinary operators on right
C  -bimodules and can be proved in a similar way by the help of Proposition 11.2.3.
0
Therefore we only prove ii). Let T 2 L .E; F /; T 0 2 L 0 .F; G/. By Defini-
tion 11.2.2 i) and Proposition 11.2.3 v),

ŒT 0 TE  ŒT 0 Im. /F   Œ 0 .Dom. 0 / \ Im. //G D ŒIm. 0 /G

and T 0 T b D 0 . .b//T 0 T  for all b 2 Dom. 0 /;  2 E. Moreover, by Defini-

tion 11.2.2 ii) and Proposition 11.2.3 v), hT 0 GjTEi  Dom. 0 / \ Im. / and

hT  T 0 GjEi D   .hT 0 GjTEi/    .Dom. 0 / \ Im. // D Dom. 0  /:

Finally, hjT 0 T i D  0 .hT 0 jT i/ D . 0  /.hT  T 0 ji/ for all  2 E;  2 G.

0
Therefore, T 0 T 2 L 0 .E; G/ and .T 0 T / D T  T 0 . 
We adopt the following notation:
Notation 11.2.9. Let E, F be right C  -A-B-bimodules, and let C D .C /; be a
family of closed subspaces C  L .E; F /, where 2 PAut.A/;  2 PAut.B/.
• Given a family D D .D /; of closed subspaces D  L .E; F /, we write
D  C if and only if D  C for all 2 PAut.A/;  2 PAut.B/.

• We define a family C  L.F; E/ by .C / WD .C  / for all ; .
• We put ŒCE WD span fT  j T 2 C ; 2 PAut.A/;  2 PAut.B/;  2 Eg.
• Let G be a right C  -A-B-bimodule and D  L.F; G/ a family of closed
subspaces. The product ŒDC  L.E; G/ is the family given by
00 0
ŒDC 00 WD span fT 0 T j T 0 2 D 0 ; T 2 C ; 0  00 ;  0    00 g

for all 00 2 PAut.A/;  00 2 PAut.B/. Clearly, the product .D; C/ 7! ŒDC

is associative.
Similarly, we define families ŒDT ; ŒS C  L.E; G/ for operators T 2
0
L .E; F / and S 2 L 0 .F; G/, where ; 0 2 PAut.A/; ;  0 2 PAut.B/.
• By a slight abuse of notation, we define a family Lid .E; F /  L.E; F / by

 WD L .E; F /;
.Lid .E; F //id .Lid .E; F // WD 0 for ¤ id :
id

Similarly, we define a family Lid .E; F /  L.E; F /.

 11.2. Semigroup grading techniques on right C  -bimodules 341

In the notation introduced above, we have ŒL.F; G/L.E; F /  L.E; G/ and

L.E; F / D L.F; E/ for all right C  -A-B-bimodules E; F; G.
Definition 11.2.10. Let E be a right C  -A-B-bimodule. We call a family of
closed subspaces C  L.E/ a C  -family on E if ŒCC  C, C  C, and C11 
C22 whenever 1  2 and 1  2 . We call a C  -family C non-degenerate if
ŒCE D E.
Remarks 11.2.11. Let C be a C  -family on a right C  -A-B-bimodule E.
i) For each pair of partial identities 0 2 PAut.A/ and 2 PAut.B/, the space
0 
C  Lid
id .E/ D LB .E/ is a C -algebra because
A

0 0 0 0 0 0 0 0
.C / D C D C and C C  C D C :

ii) For each 2 PAut.A/ and  2 PAut.B/, the space C is a C  -module over

the C  -algebra C  because
   
.C / C D C  C  C  and C C   C
 
  D C :



Likewise, C is a left C  -module over the C  -algebra C 
  and a C -
 
bimodule over C  
   and C   .
    id 
id C  D C D C Cid for each ;  ; this follows from ii) (see Sec-
iii) ŒCid
tion 12.2).
iv) The C  -family C is non-degenerate if and only if the C  -algebra Cidid 
LB
A
.E/ is non-degenerate in the usual sense. This follows easily from iii).
To every C  -family, one can associate a multiplier C  -family:
Definition 11.2.12. Let C be a C  -family on a right C  -A-B-bimodule E. The
multiplier family of C is the family M.C/  L.E/ given by

M.C/ D fT 2 L .E/ j ŒT C  C and ŒCT   Cg

for all 2 PAut.A/ and  2 PAut.B/. Evidently, this is a C  -family.

Remark 11.2.13. Remark 11.2.11 iii) implies that for all 2 PAut.A/;  2
PAut.B/,

M.C/ D fT 2 L .E/ j T Cid

id  C and Cid T  C g:
 id 

In Section 11.4, we shall construct for every decomposable regular pseudo-

multiplicative unitary a left and a right leg in the form of C  -families. In principle,
we could embed these C  -families in some ambient C  -algebras, which would
342 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

bring us back into more familiar terrain. However, until recently, it was not clear
how to define a fiber product, which is necessary to define the analogue of a Hopf–
von Neumann bimodule, on the level of the ambient C  -algebras. On the level of
C  -families, an internal tensor will be defined in Section 11.3 quite easily. For
recent developments in the setting of C  -algebras, see [154], [155], [156].

11.2.2 Homogeneous elements of right C  -bimodules

We consider elements of right C  -bimodules that almost intertwine left and right
multiplication:
Definition 11.2.14. Let E be a right C  -B-B-bimodule and  2 PAut.B/. An
element  2 E is  -homogeneous if  2 ŒE Dom. / and b D .b/ for all
b 2 Dom./. The set of all -homogeneous elements of E is denoted by H .E/.
We call E decomposable if the family H.E/ WD .H .E// is linearly dense in E.
Note that B can be regarded as a C  -module over B in a natural way, and that
left multiplication turns B into a right C  -B-B-bimodule. Thus we can speak of
homogeneous elements of B; these will be studied later.
Let E be a right C  -B-B-bimodule. For each  2 E, we define maps
ji W B ! E; b 7! b; j W B ! E; b 7! b:
Then ji has an adjoint hj D ji W  7! hji, and kjik D kk [95, p. 12–13].
Proposition 11.2.15. Let  2 E and  2 PAut.B/. Then the following conditions
are equivalent:
i/  2 H .E/; ii/ ji 2 Lid .B; E/; iii/ j 2 Lid
  .B; E/:

If i)–iii) hold, then kjk D kk and Œj WD j is given by  7! .hji/.
Proof. i) ) ii), iii): Assume that i) holds. To prove ii), we only need to show that ji
satisfies condition i) of Definition 11.2.2. But by assumption, Im ji  ŒIm. /E
and ji.bb 0 / D bb 0 D .b/jib 0 for all b 2 Dom. /; b 0 2 B. Let us prove iii).
Evidently, j commutes with left multiplication. By assumption, hji 2 Dom. /
for all  2 E, so that the map Œj W  7! .hji/ is well defined. Let .u / be an
approximate unit of Im. /. Then
hjjbi D limhju bi D limhji  .u b/ D   ..hji/ b/ D   .hŒjjbi/

for all  2 E, b 2 B. Hence iii) holds. Moreover, we may assume ku k  1 for

all , and then kk D lim kju k  kjk. The reverse inequality is evident.
ii) ) i): If ii) holds, then  2 ŒB D ŒIm ji  ŒIm. /E, and c D .c/ for
each c 2 Dom. / because cb D jicb D .c/.jib/ D .c/b for each b 2 B.
iii) ) i): This follows from a similar argument like the implication “ii) ) i)”.

 11.2. Semigroup grading techniques on right C  -bimodules 343

Let E be a C  -module over A and F a right C  -A-B-bimodule. For each

 2 E and  2 F , we define maps

jiŒ1 W F ! E  F;  7!   ; jŒ2 W E ! E  F;  7!   :

Then jiŒ1 has an adjoint hjŒ1 D jiŒ1 W    0 7! hji 0 , and kjik D kk if the
representation A ! LB .F / is injective [95, Lemma 4.6].

Proposition 11.2.16. Let E; F be right C  -B-B-bimodules and  2 PAut.B/.

i) If  2 H .E/, then jiŒ1 2 Lid .F; E  F /.

ii) Let  2 H .F /. Then jŒ2 2 Lid


.E; E  F /, and ŒjŒ2 WD jŒ2  is given
0 0
by    7! .hj i/. If E is full, then kjk D kk.

Proof. The proof is similar to that of Proposition 11.2.15; we only sketch the main
steps for ii). Let  2 H .F /. For all ;  0 2 E and  0 2 F ,

h 0   0 j  i D h 0 jh 0 jii D   ..h 0 ji/h 0 ji/ D   .h 0 .hj 0 i/ji/:

For  0 D ,  0 D , this equation shows kjŒ2 k2  k.hji/kkk2 , and hence

kjŒ2 k  kk. If E is full, this inequality is an equality. Finally, the equation
above shows that the formula for ŒjŒ2 defines a bounded map E  F ! E, and
that h 0   0 jjŒ2 i D   .hŒjŒ2 . 0   0 /ji/ for all ;  0 2 E and  0 2 F . 

Next, we collect several useful formulas concerning homogeneous elements.

Let E and F be right C  -B-B-bimodules. For ;  0 2 PAut.B/, put

H .E/  H 0 .F / WD span f   j  2 H .E/;  2 H 0 .F /g  E  F:

Proposition 11.2.17. Let ;  0 ; ; 2 PAut.B/. Then:

i) H .E/ D ŒH 0 .E/ Dom. /  H 0 .E/ if    0 .

ii) hH .E/j H 0 .E/i  H   0 .B/.

iii) For each  2 E, the set f 0 2 PAut.B/ j  2 H 0 .E/g either is empty or has
a minimal element.

iv) L .E; F / H .E/  H   .F / and H .A/ H .E/ H .B/  H .E/.

v) The space I WD ŒhH .E/j H .E/i is an ideal in Z.B/, and H .E/ is a

right C  -Z.B/-I -bimodule. In particular, H .E/I D H .E/.

vi) If E is full and decomposable, then B is decomposable and the ideal of Z.B/
spanned by all I 00 , where  00 2 PAut.B/, is non-degenerate in B.
344 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

vii) H .E/ \ H 0 .E/ D H.^ 0 / .E/.

viii) H .E/  H 0 .F /  H  0 .E  F /.
Proof. We only prove assertions iii), iv), v), iiv); the others follow from straight-
forward calculations or can be deduced from Proposition 11.2.8 and 11.2.15.
iii) Given  2 E, apply Propositions 11.2.3 iii) and 11.2.15 to j.
iv) Let T 2 L .E; F / and  2 H .E/. Choose approximate units .u / ,
.v / , .w / of Dom. /; Im. /; Dom. /, respectively. By Proposition 11.2.3,
T  D lim T .u v w / D lim T .  .u v /w / D lim .T / .  .u v /w /:
;; ;; ;;

Since .. .u v /w //;; is an approximate unit for Dom.    /, the equation



above implies T  2 ŒF Dom.    /. Moreover, for all b 2 Dom.    /,

.T /b D T .  .b// D T .  .b// D ..    /.b//T :
Therefore, T  2 H   .F /. The second inclusion in assertion iv) follows similarly.
v) The assumptions imply that B is contained in the closure of
X X X
hH 0 .E/j H .E/i D I 0 hH 0 .E/j H .E/i  I 0 H 0  .B/I
; 0 ; 0 ; 0

here, we used ii) and iv). The claims follow.

vii) By i), H.^ 0 / .E/  H .E/ \ H 0 .E/. Conversely, if  2 H .E/ \
H 0 .E/ and  00 2 PAut.B/ is minimal with  2 H 00 .E/ (see iii)), then  00  
and  00   0 , whence  00   ^  0 and  2 H. ^ 0 / .E/. 
The preceding proposition suggests the following notation:
Notation 11.2.18. Let E be a right C  -B-B-bimodule, and let E D .E / and
E0 D .E0 / be families of closed subspaces E  H .E/ and E0  H .E/, where
 2 PAut.B/.
• We write E0  E if E0  E for all  2 PAut.B/.
• We define a family ŒhE0 jEi  H.B/ by
ŒhE0 jEi 00 D span fh 0 ji j  2 E ;  0 2 E0 0 ; ;  0 2 PAut.B/;  0    00 g:

• Given a family C  L.E; F /, where F is a right C  -B-B-bimodule, we

define a family ŒCE  H.F / by
ŒCE D span fS j S 2 C ;  2 E 0 ; ;  0 ;  2 PAut.B/;  0     g:
Similarly, we define a family ŒSE  H.F / for each homogeneous operator
SW E ! F.
 11.2. Semigroup grading techniques on right C  -bimodules 345

• Given a right C  -B-B-bimodule F and a family F  H.F /, we define a

family ŒE  F  H.E  F / by

ŒE  F 00 WD span f   j  2 E ;  2 F 0 ; ;  0 2 PAut.B/;   0   00 g:

Let E, F be right C  -B-B-bimodules. For each  00 2 PAut.B/, put

00
Kid .E; F / WD span fjih 0 j j  2 H .F /;  0 2 H 0 .E/;   0   00 g:

Proposition 11.2.19. Let E, F be right C  -B-B-bimodules.

i) If E or F is decomposable, then Kid .E; F / D KB .E; F / \ Lid .E; F / for

each  2 PAut.B/.

ii) If E and F are decomposable, then E  F is decomposable and H.E  F / D

ŒH.E/  H.F /.

Proof. See [153, Propositions 3.16, 3.17]. 

Let E be a C  -A-module, F a right C  -B-B-bimodule, and

W A ! LB
B
.F /

a -homomorphism. Then E  F is a right C -B-B-bimodule via the represen-
tation B ! LB .E  F /; b 7! id  b (use Proposition 12.2.1). Given a family
F  H.F /, we define a family ŒE  F  H.E  F / by

ŒE  F WD span f   j  2 E;  2 F g:

Proposition 11.2.20. If F is decomposable, then E  F is decomposable and

H.E  F / D ŒE  H.F /.

Proof. See [153, Proposition 3.18]. 

Homogeneous elements of C  -algebras. Let us collect some useful properties

of homogeneous elements of C  -algebras.

Proposition 11.2.21. Let b 2 H .B/,  2 PAut.B/, and denote by Ib  B the

ideal generated by b  b. Then:

i) b is normal and b  b is central.

ii) There exists a unitary u 2 M.Ib / such that b D u.b  b/1=2 .

iii) With u as in ii), the map Adu W Ib ! Ib is the minimal partial automorphism
of B with respect to which b is homogeneous.

iv) .b/ D b; in particular, b 2 Dom.  / and   .b/ D b.

346 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Proof. i) The positive elements b  b and bb  are central by Proposition 11.2.17 ii),
whence bb   bb  D b  bbb  D b  b  b  b. Consequently, bb  D b  b.
ii) Put D WD spec.b/ n f0g. For n 1, define fn 2 C0 .D/ by fn .z/ WD z=jzj if
jzj 1=n, and fn .z/ WD nz if jzj  1=n. Then .fn /n converges in M.D/ strictly
to a unitary, and functional calculus shows that the sequence .fn .b//n converges in
M.Ib / strictly to some unitary u. Denote by idD 2 C0 .D/ the identity map. Then
limn fn j idD j D idD in C0 .D/, and hence u.b  b/1=2 D limn fn .b/j idD .b/j D
idD .b/ D b.
iii) Evidently, b 2 Ib and bd D u.b  b/1=2 d D udu u.b  b/1=2 D Adu .d /b
for all d 2 Ib , so b 2 HAdu .B/. If b 2 H 0 .B/ for some  0 2 PAut.B/, then
Ib  Dom. 0 / because b 2 Dom. 0 /, and Adu   0 by Proposition 11.2.17 iii).
iv) .b/ D Adu .b/ D u.u.b  b/1=2 /u D u.b  b/1=2 D b by iii) and because
.b b/ is central. The relations b 2 Dom.  / and b D   .b/ follow.
 1=2

Proposition 11.2.22. Let ;  0 ; 2 PAut.B/. Then:
i) bc D .cb/ and cb D   .bc/ for all b 2 H .B/; c 2 B.
ii) H .B/ D H .B/ \ Dom. ^ id/.
iii) .H .B/ \ Dom. //  H .B/.
iv) H 0 .B/ H .B/  H 0  .B/ and H .B/ D H  .B/.
v) B is decomposable if and only if the inclusion Z.B/  B is non-degenerate.
In particular, every unital C  -algebra is decomposable.
Proof. i) Let b 2 H .B/, c 2 B, and let .u / be an approximate unit of Dom. /.
Then

bc D lim bu c D lim .u c/b D .c  .b// D .cb/

by Proposition 11.2.21 iv), and similarly cb D   .bc/.

ii) This follows from Proposition 11.2.21 iv).
iii) Combine Example 11.2.5 with Proposition 11.2.17 iv).
iv) Straightforward.
v) If B is decomposable, then ŒBZ.B/ D B by Proposition 11.2.17 vi). Con-
versely, assume ŒZ.B/B D B. For each unitary u 2 M.B/ and each b 2 Z.B/,
the product bu is contained in HAdu .B/ because buc D .ucu /bu for all c 2 B.
By [113, Remark 2.2.2], each element of B can be written as a sum of four unitaries
in M.B/. Therefore B is decomposable. 
To every C  -bimodule E, we associate a C  -family O.E/ as follows:
Proposition 11.2.23. Let A, B be C  -algebras and E a right C  -A-B-bimodule.
 11.2. Semigroup grading techniques on right C  -bimodules 347

i) Let a 2 H .A/, 2 PAut.A/, and b 2 H .B/,  2 PAut.B/. Then the map

oa;b W E ! E,  7! ab, is a . ;   /-homogeneous operator with adjoint
.oa;b / D oa ;b  .

ii) For all 2 PAut.A/,  2 PAut.B/, put

O .E/ WD span foa;b j a 2 H .A/; b 2 H  .B/g:

Then O.E/  L.E/ is a C  -family.

Proof. i) Let a; b as above. Then oa;b satisfies condition i) of Definition 11.2.2

because Im.oa;b /  aE  Im. /E and

oa;b a0  D aa0 b D .a0 /ab D .a0 /oa;b 

for all a0 2 Dom. /;  2 E. Further, by Proposition 11.2.22 i), iv) and 11.2.21 iv),
a 2 H .A/,  .b  / D b  2 H  .B/, hoa ;b  EjEi  bhEjEi  Dom.  /,
and

hjabi D ha jib D   .bha ji/ D   .ha b  ji/

for all ;  2 E. The claim follows.

ii) Obvious from i) and Proposition 11.2.22 iv). 

Definition 11.2.24. Let E be a right C  -A-B-bimodule, where A and B are de-

composable. A family C  L.E/ is called an O.E/-module if ŒO.E/C  C,

and a non-degenerate O.E/-module if additionally C D ŒO 
  .E/C  for all
2 PAut.A/ and  2 PAut.B/.

Remark 11.2.25. The C  -family O.E/ defined above is interesting primarily if

A and B are decomposable. However, we can consider E as a right C  -M.A/-
M.B/-bimodule via the identification E Š A  E  M.B/, and M.A/ and M.B/
are decomposable by Proposition 11.2.22 v).

11.2.3 Examples related to groupoids

Let G be a locally compact, second countable, Hausdorff groupoid with left Haar
system , and consider the C  -module L2 .G; / over the C  -algebra B WD C0 .G 0 /.
Recall that the range and source map of G induce representations r; s W B !
LB .L2 .G; //. To avoid confusion with these representations, we denote the range
and the source map of G by rG and sG , respectively. As before, we write r L2 .G; /
or s L2 .G; / to indicate whether we consider L2 .G; / as a right C  -bimodule via
the representation r or s.
348 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Homogeneous elements of L2 .G; /. The right C  -bimodule r L2 .G; / is de-

composable: Evidently,

Hid .r L2 .G; // D L2 .G; /:

The right C  -bimodule s L2 .G; / is decomposable if the groupoid G satisfies the

following decomposability condition:
Definition 11.2.26. We call an open subset U  G homogeneous if for all x; y 2 U ,
we have rG .x/ D rG .y/ if and only if sG .x/ D sG .y/. We call the groupoid G
decomposable if it is equal to the union of its open homogeneous subsets.
Remarks 11.2.27. i) If U; V  G are open homogeneous subsets, then also U 1
and U V D fxy j .x; y/ 2 Gs;r 2
\ .U  V /g are open and homogeneous.
ii) Recall that an open subset U  G is called a G-set if the restrictions
rjU W U ! r.U / and sjU W U ! s.U / are homeomorphisms and r.U /; s.U /  G 0
are open. Moreover, recall that G is r-discrete if and only if it is the union of open
G-sets [129, Proposition 2.8]. Evidently, every G-set is homogeneous, and if G is
r-discrete, then it is decomposable.
iii) It is easy to see that every extension of an r-discrete groupoid by a bundle
of groups is decomposable. In fact, it is not difficult to show that a groupoid is
decomposable if and only if it is an extension of an r-discrete groupoid by a bundle
of groups [152, Proposition 3.4].
Let us prove that s L2 .G; / is decomposable if G is decomposable. Since
the range and the source map of G are open [129, I, Proposition 2.4], every open
homogeneous subset U  G defines a homeomorphism qU W sG .U / ! rG .U / by
sG .x/ 7! rG .x/, which induces partial automorphisms

qU  W C0 .sG .U // ! C0 .rG .U // and qU W C0 .rG .U // ! C0 .sG .U //

of the C  -algebra B D C0 .G 0 /.
Lemma 11.2.28. Let U  G be open and homogeneous. Then

Cc .U /  HqU .s L2 .G; //:

Proof. Clearly, Cc .U /  L2 .G; /C0 .rG .U //, and for all  2 Cc .U /; f 2

C0 .rG .U //, and x 2 U ,

.f /.x/ D .x/f .rG .x// D .x/f .qU .sG .x/// D .s.qU .f ///.x/: 

Combining this lemma with a partition of unity argument, we find:

Proposition 11.2.29. If the groupoid G is decomposable, then the right C  -bimod-
ule s L2 .G; / is decomposable. 
 11.2. Semigroup grading techniques on right C  -bimodules 349

The left regular representation on L2 .G; /. We shall see that if G is decom-
posable, then the left regular representation on L2 .G; / gives rise to a C  -family.
To simplify the discussion, we impose the following condition:

Definition 11.2.30. We call the left Haar system  unimodular if for every open
homogeneous subset U  G, every function f 2 Cc .U /, and every z 2 U ,
Z Z
f .y/d rG .z/
.y/ D f .y 1 /d sG .z/ .y/:
G rG .z/ G sG .z/

Example 11.2.31. If G is r-discrete, then the family of counting measures is a

unimodular left Haar system.

For a function f 2 Cc .G/, we denote by L.f / W L2 .G; / ! L2 .G; / the

map given by left convolution with f ,
Z
.L.f //.x/ WD f .y/.y 1 x/d rG .x/ .y/ for all x 2 G;  2 Cc .G/:
G rG .x/

Moreover, we denote by f  2 Cc .G/ the function given by f  .x/ WD f .x 1 / for

all x 2 G.

Proposition 11.2.32. Assume that G is decomposable and that the left Haar system
 is unimodular.

i) For every open homogeneous subset U  G and every f 2 Cc .U /, we

qU  
qU  .s L .G; //; L.f / 2 LqU  .r L .G; //, and L.f / D
have L.f / 2 Lid 2 2

L.f /.

ii) Let f 2 Cc .U / and g 2 Cc .V /, where U; V  G are open homogeneous

subsets. Then L.f /L.g/ D L.h/, where h 2 Cc .U V / is given by
Z
h.x/ D f .y/g.y 1 x/d rG .x/ .y/ for all x 2 G:
G rG .x/

iii) The family of closed subspaces Cr .G/  L.r L2 .G; // given by

Cr .G/ D span fL.f / j f 2 Cc .U /; U  G open homogeneous;

qU   and qU    g

for all ;  2 PAut.B/ is a non-degenerate C  -family and a non-degenerate

O.r L2 .G; //-module.
350 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Proof. i) It is easy to see that L.f / commutes with the representation s, that the
image of L.f / is contained in r.C0 .rG .U ///L2 .G; /, and that L.f /r.b/ D
r.qU  .b//L.f / for each b 2 C0 .sG .U //.
Let ;  2 Cc .G/. Then the inner products hjL.f /i and hL.f  /ji, consid-
ered as functions on G 0 , vanish outside rG .U / and sG .U /, respectively. For each
z 2 U,
Z Z
hjL.f /i.rG .z// D .x/f .y/.y 1 x/d rG .z/ .y/d rG .z/ .x/:
G rG .z/ G rG .z/

Since  is left-invariant and unimodular, this iterated integral is equal to

Z Z
.yx 0 /f  .y 1 /.x 0 /d sG .y/ .x 0 /d rG .z/ .y/
r G .z/ s G .y/
G
ZG Z
D .y 1 x 0 /f  .y/.x 0 /d sG .z/ .x 0 /d sG .z/ .y/
G sG .z/ G sG .z/

D hL.f  /ji.sG .z//:


Since sG .z/ D qU .rG .z//, it follows that hjL.f /i D qU .hL.f  /ji/. Thus
assertion i) is proved.
ii), iii) Statement ii) follows from a routine calculation, and statement iii) follows
from i), ii), and Remark 11.2.27 i). 

11.3 Hopf C  -families

In this section, we introduce the internal tensor product of C  -families and the
notion of a morphism of C  -families. These concepts are needed for the definition
of a Hopf C  -family, which is given afterwards. Throughout this subsection, let A,
B, C be C  -algebras.

11.3.1 The internal tensor product of C  -families

Let E be a right C  -A-B-bimodule and F a right C  -B-C -bimodule. Then we can
form the internal tensor product E  F , which is a right C  -A-C -bimodule. Given
C  -families C  L.E/ and D  L.F /, we shall define an internal tensor product
C  D  L.E  F /, which is a C  -family again. This internal tensor product will
be generated by operators of the form
S  T W    7! S   T ;
where S and T are suitable operators of the C  -families C and D, respectively. The
relation
S.b/  T  D .S  T /.b  / D .S  T /.  b/ D S   T b
 11.3. Hopf C  -families 351

shows that the map S  T is well defined only if S intertwines the right multipli-
cation on E in a way that matches up with the way in which T intertwines the left
multiplication on F . In the next definition and lemma, we formulate and investigate
this compatibility condition:
Definition 11.3.1. We call two partial automorphisms ;  2 PAut.B/ compatible
and write   if    id and    id.
Lemma 11.3.2. Let ;  2 PAut.B/ such that   .
i)     ;
ii) .a/ D  .a/ for all a; b 2 Dom. / \ Dom. /;
iii) .Dom. / \ Dom. // D Im. / \ Im. / D  .Dom. / \ Dom. //;
iv) .ab/ D .a/ .b/ D  .ab/ for all a 2 Dom. /; b 2 Dom. /;
v) if 0 ;  0 2 PAut.B/ and 0   0 , then 0    0 .
Proof. Assertions i) and ii) follow immediately from the definition.
iii) By ii), .Dom. / \ Dom. // D  .Dom. / \ Dom. // is contained in
Im. / \ Im. /. To obtain the reverse inclusion, replace ;  by  ;   .
iv) Let .u / be an approximate unit for Dom. / \ Dom. /. By iii), . .u //
is an approximate unit for Im. / \ Im. /, and for all a 2 Dom. /; b 2 Dom. /,
.ab/ D lim .abu / D lim .a/ .bu /

D lim .a/ .bu / D lim .a/ .b/ .u / D .a/ .b/:

v) Since 0  0  id and    id, the products . 0 /.  0 / D . 0  0 /  and

. 0 / . 0 / D 0 .   / 0 are restrictions of   and 0  0 , respectively. 
In general, compatibility is not transitive: the automorphism of the ideal f0g is
compatible with every other partial automorphism of B.
Proposition 11.3.3. Let E1 ; E2 be right C  -A-B-bimodules and F1 ; F2 right C  -
B-C -bimodules. Furthermore, let
• S 2 LSS .E1 ; E2 /, where S 2 PAut.A/; S 2 PAut.B/,
• T 2 LTT .F1 ; F2 /, where T 2 PAut.B/; T 2 PAut.C /.
If S  T , then there exists an operator S  T 2 LST .E1  F1 ; E2  F2 / such
that
.S  T /.  / D S  T  for all  2 E1 ;  2 F1 : (11.4)
Moreover, kS  T k  kS kkT k and .S  T / D S   T  .
352 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Proof. To simplify notation, we put E WD E1 ˚ E2 ; F WD F1 ˚ F2 , and consider

S and T as elements of LSS .E/ and LTT .F /, respectively, in the natural way. Let
; 0 2 E and ;  0 2 F . Then

h0   0 jS  T i D h 0 jh0 jSiT i D h 0 jS .hS  0 ji/T i:

Let .u / be an approximate unit for Dom. T /. Then the homogeneity of T and S ,

Proposition 11.2.3 v), and Lemma 11.3.2 iv) imply

S .hS  0 ji/T  D lim T .u /S .hS  0 ji/T 

D lim T u hS  0 ji D T hS  0 ji:

We insert this relation into the equation above, and find

h0   0 jS   T i D h 0 jT hS  0 jii D T .hT   0 jhS  0 jii/

(11.5)
D T .hS  0  T   0 j  i/:

Let us show that formula (11.4) defines a bounded map S  T . By equation

(11.5), we have for all i 2 E, i 2 F , where i D 1; : : : ; n,
X 2 X

Si  T i D hS  S i  T  T i jj  j i
i i;j

Now S  S 2 LB .E/, T  T 2 LB C .F /, and by Proposition 12.2.1, the operators

S  S  1; 1  T  T; S  S  T  T 2 LC .E  F / are well defined. Since

S  S  T  T D .S  S  1/.1  T  T / D .1  T  T /.S  S  1/;

we have

kS  T k2  kS  S  T  T k  kS  S  1kk1  T  T k  kS k2 kT k2 :

Evidently, the image of S  T is contained in Im. S /.E  F /, and

.S  T /a.  / D Sa  T 
D S .a/S   T  D S .a/.S  T /.  /

for all  2 E;  2 F , and a 2 Dom. S /. Replacing S and T by their adjoints, we

obtain a bounded map S   T  W E  F ! E  F , and formula (11.5) shows that
S  T is . S ; T /-homogeneous with adjoint S   T  as claimed. 

Next, we introduce the internal tensor product of C  -families. For later appli-
cations, we state the definition in a slightly wider generality.
 11.3. Hopf C  -families 353

Notation 11.3.4. Let E1 ; E2 be right C  -A-B-bimodules and F1 ; F2 right C  -B-

C -bimodules. Furthermore, let C  L.E1 ; E2 / and D  L.F1 ; F2 / be families of
closed subspaces. Then the internal tensor product C  D  L.E1  F1 ; E2  F2 /
is the family given by

.C  D/ WD span fS  T j S 2 CS ; T 2 DT ; S ; T 2 PAut.B/; S  T g

for all 2 PAut.A/ and  2 PAut.C /.

The internal tensor product of families does not always behave as one might
naïvely expect:
Remark 11.3.5. Let E be a right C  -A-B-bimodule, F a right C  -B-C -bimodule,
and A; C  L.E/, B; D  L.F / families of closed subspaces. Then

Œ.A  B/.C  D/  ŒAC  ŒBD:

This inclusion may be strict and fail to be an equality. As a simple example, consider
the case where all spaces comprising the families C and D are 0 except for C11 and
D22 , where 1 2 PAut.A/; 2 2 PAut.C /, and 1 ; 2 2 PAut.B/ are fixed and not
compatible. Then C  D D 0 D C  D, but C C  D D need not be 0.
The internal tensor product of C  -families is a C  -family again:
Proposition 11.3.6. Let E be a right C  -A-B-bimodule, F a right C  -B-C -bi-
module, and let C  L.E/ and D  L.F / be C  -families.
i) C  D  L.E  F / is a C  -family.
ii) If C and D are non-degenerate, so is C  D.
iii) If A, B, C are decomposable, C is a .non-degenerate/ O.E/-module, and
D is a .non-degenerate/ O.F /-module, then C  D is a .non-degenerate/
O.E  F /-module.
iv) M.C/  M.D/  M.C  D/.
Proof. This follows easily from the definitions and from Lemma 11.3.2. 

It is easy to see that the internal tensor product is associative:

Proposition 11.3.7. Let A, B, C , D be C  -algebras, let E be a right C  -A-B-
bimodule, F a right C  -B-C -bimodule, and G a right C  -C -D-bimodule. Fur-
thermore, let B  L.E/, C  L.F /, D  L.G/ be C  -families. Then the natural
isomorphism .E  F /  G Š E  .F  G/ induces an isomorphism of C  -families
.B  C/  D Š B  .C  D/. 
354 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Remark 11.3.8. The internal tensor product is not quite associative on the level
of operators: It may happen that R; S, and T are operators such that the internal
tensor products R  S and .R  S/  T are well defined while the internal tensor
products S  T and R  .S  T / are not defined. This phenomenon occurs, for
example, if R D 0, but it is not restricted to such trivial cases.
The constructions introduced above can easily be adapted to the flipped internal
tensor product of right C  -bimodules and give rise to a flipped internal tensor
product of homogeneous operators and C  -families.

11.3.2 Morphisms of C  -families

It seems to be difficult to find a natural notion of a morphism of C  -families that
makes the internal tensor product bifunctorial. Therefore we adopt a pragmatic
approach:
Definition 11.3.9. Let C and D be C  -families on right C  -A-B-bimodules. A
family of linear maps  W C ! D is a family  D . /; of linear maps  W C !
D , defined for all 2 PAut.A/;  2 PAut.B/. We call a family of linear maps
W C ! D
• A0 -B 0 -extendible, where A0 and B 0 are C  -algebras, if for each right C  -
A0 -A-bimodule X and each right C  -B-B 0 -bimodule Y , there exists a linear
map
YX W .L.X /  C  L.Y //id
id ! .L.X /  D  L.Y //id
id

such that YX .R  S  T / D R   .S /  T for all R 2 Lid 0 .X /; S 2

 0
C ; T 2 Lid .Y /, where  ; 2 PAut.A/, ; 2 PAut.B/,   ,   0 ;
0 0 0

• extendible if  is A0 -B 0 -extendible for every C  -algebra A0 and B 0 ;

• injective if each component  is injective;
• a morphism if  is extendible and YX always is a -homomorphism.
We call a morphism  W C ! M.D/ non-degenerate if Œ.C/D D D.
Let B, C, D be C  -families on right C  -A-B-bimodules. The composition of
two families of linear maps  W B ! C and W C ! D is the family ı W B ! D
given by . ı / WD  ı  for all ; .

Remark 11.3.10. i) .L.X /  C  L.Y //id id and .L.X /  D  L.Y //id are C -
id
A0 A0
subalgebras of LB 0 .X  E  Y / and LB 0 .X  F  Y /, respectively.
ii) Clearly, the composition of (extendible) families of linear maps/of morphisms
is a (extendible) family of linear maps/a morphism again, and the collection of all
C  -families on right C  -A-B-bimodules and all (extendible) families of linear
maps/all morphisms forms a category.
 11.3. Hopf C  -families 355

To determine whether a family of linear maps is extendible, it suffices to check

that it is C-C-extendible:

Proposition 11.3.11. Let  W C ! D be a family of linear maps between C  -

families. Then  is extendible if and only if it is C-C-extendible.

Proof. Assume that  is C-C-extendible, and let A0 , B 0 be C  -algebras. We show

that  is A0 -B 0 -extendible. Let X 0 be a right C  -A0 -A-bimodule and Y 0 a right
C  -B-B 0 -bimodule. Denote by X the C  -module X 0 , considered as a right C  -C-
A-bimodule via multiplication by scalars. Choose a faithful representation of B 0
on a Hilbert space H and put Y WD Y 0  H . For G D E; F and B D C; D, the
A0 0 0 C 0
embedding LB 0 .X  G  Y / ,! LC .X  G  Y  H /, T 7! T  idH , maps
0 0 id
.L.X /  B  L.Y //id to .L.X /  B  L.Y //id . Restricting the map YX (which
id
0
exists by assumption), we obtain the desired map YX0 . 

For the study of morphisms of C  -families, the following embedding result is

useful:

Theorem 11.3.12. There exist a right C  -A-A-bimodule I A, a right C  -B-B-


bimodule I B, and operators V 2 Lid .I A/, W 2 Lid .I B/, defined for each
 2 PAut.A/, 2 PAut.B/, such that for each right C  -A-B-bimodule E, the
maps

 W L .E/ ! LB
A
.I A  E  I B/; T 7! V  T  W ;

defined for each 2 PAut.A/;  2 PAut.B/, satisfy

 0 0
k .T /k D kT k;  .T / D   .T  /;  .T / 0 .T 0 / D  0
 0 .T T /

0
for all T 2 L .E/, T 0 2 L 0 .E/, ; 0 2 PAut.A/, ;  0 2 PAut.B/.

Proof. See [153, Theorem 5.10]. 

Morphisms of C  -families behave in many respects as one should expect:

Proposition 11.3.13. Let  W C ! D be a morphism of C  -families. Then

0 0 
0 .c 0 / .c/ D 0  .c 0 c/;  .c/ D  .c  /; k .c/k  kck;
0
 .c/ D 0 .c/ if . ;  /  . 0 ;  0 /
0
for all c 2 C , c 0 2 C 0 , ; 0 2 PAut.A/, ;  0 2 PAut.B/. In particular,

idid W Cid
id ! Did is a -homomorphism of C -algebras.
id
356 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Proof. This follows from the existence of a -homomorphism IIB

A
that makes the
diagram below commute for all 2 PAut.A/ and  2 PAut.B/:
 / Œ.C/  .L.I A/  C  L.I B//id
C  id




IA
  IB



D / Œ.D/  .L.I A/  D  L.I B//id . 
id

Remarks 11.3.14. i) A morphism  W C ! D of C  -families is injective if and

only if the component idid is injective because

k .c/k2 D k .c/  .c/k D k  .c  c/k D kidid .c  c/k

for all c 2 C and all ; .

i) A morphism  W C ! M.D/ of C  -families is non-degenerate if and only
if the natural map idid W Cid
id ! M.D/id ! M.Did / is a non-degenerate -homo-
id id

morphism of C -algebras. This follows from Remark 11.2.11 iii).
The internal tensor product of C  -families is bifunctorial:
Proposition 11.3.15. Let  W A ! C and W B ! D be extendible families of lin-
ear maps/.non-degenerate/ morphisms of C  -families on right C  -A-B-bimodules
and right C  -B-C -bimodules, respectively. Then there exists an extendible family
of linear maps/.non-degenerate/ morphism

 W AB!CD
such that
0
.  / 0 .a  b/ D  .a/   0 .b/
0
for all a 2 A ; b 2 B 0 , 2 PAut.A/; ; 0 2 PAut.B/;  0 2 PAut.C /,   0 .
Proof. If we can prove the assertion for the case that B D D; D idB and for the
case that A D C;  D idA , then we can simply put   WD .  id/ ı .id  /.
We treat the first case, the second one is similar.
Let 2 PAut.A/ and  0 2 PAut.C /. Denote by F the right C  -bimodule on
which B acts. If ; 0 2 PAut.B/ and   0 , then the diagram
0  /  0 ..A  B/ 0 /  .L.I A/  A  L.F  I C //id
A  B 0    id
 0


 id
IA
   F IC
0   0
C  B 0 /  0 ..C  B/ 0 /  .L.I A/  C  L.F  I C //id
  id

commutes, and we can insert a unique linear map .  id/ 0 W .A  B/ 0 ! .C  B/ 0

that does not depend on ; 0 such that the diagram still commutes.
 11.3. Hopf C  -families 357

The family ..  id/ 0 /; 0 thus defined is extendible. Indeed, let X be a right
C -C-A-bimodule and Y a right C  -C -C-bimodule. Then F  Y is a right C  -


B-C-bimodule, and the linear map

FXY W .L.X /  A  L.F  Y //id

id ! .L.X /  C  L.F  Y //id
id

restricts to a linear map

Y W .L.X /  A  B  L.Y //id ! .L.X /  C  B  L.Y //id

.  id/X id id

which has the desired properties. If  is a morphism, then FXY and hence also
Y are -homomorphisms, so that   id is a morphism.
.  id/X 
Non-degenerate morphisms of C  -families can be extended to multipliers:
Proposition 11.3.16. Let  W C ! M.D/ be a non-degenerate morphism of C  -
families. If the C  -family D is non-degenerate, then  extends uniquely to a mor-
phism M.C/ ! M.D/.
Proof. Uniqueness follows, once existence is proved, by a standard argument. De-
note by F the underlying right C  -bimodule of D. Choose an approximate unit
.u / for the C  -algebra Cidid such that 0  u  1 for all .
For each 2 PAut.A/,  2 PAut.B/, we construct an extension N  W M.C/ !
M.D/ of  as follows. Let c 2 M.C/ . Since  and D are non-degenerate, the net
. .cu // converges strictly to some N  .c/ 2 L .F / (see Proposition 11.2.8 i)).
Now N  .c/ 2 M.D/ because
 id id id    
N  .c/Did N
id D   .c/ id .Cid /Did   .cCid /Did  D ;
id id 

N 
id   .c/  D .
and likewise Did
We show that the family N W M.C/ ! M.D/ thus defined is a morphism.
Let X be a right C  -C-A-bimodule and Y a right C  -B-C-bimodule. By as-
sumption on , the -homomorphism YX is non-degenerate and extends to a -ho-
momorphism

YX W M..L.X /  C  L.Y //id

id / ! M..L.X /  M.D/  L.Y //id /:
id

  0
0 0
For all R 2 Lid
 0 .X /, S 2 M.C/ , T 2 Lid .Y /, where  ; 2 PAut.A/, ; 2
PAut.B/, and  0  ,   0 , the operators YX .R  S  T / and R  N  .S /  T are
equal because they coincide with the strict limit of the net .R   .S u /  T / .
Hence, YX restricts to a -homomorphism

N YX W .L.X /  M.C/  L.Y //id

id ! .L.X /  M.D/  L.Y //id :
id

358 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

We are primarily concerned with the following examples of morphisms:

Examples 11.3.17. i) An inclusion of C  -families is a morphism.
Let C be a C  -family on a right C  -A-B-bimodule E.
ii) Let F be a right C  -A-B-bimodule and V 2 LB A
.E; F / an isometry. Then
AdV .C/ WD ŒV CV   L.F / is a C -family, and the formula c 7! VcV 
 

defines an isomorphism AdV W C ! AdV .C/. If C is a (non-degenerate)

O.E/-module, then AdV .C/ is a (non-degenerate) O.F /-module; if V is
unitary and C non-degenerate, then AdV .C/ is non-degenerate.
iiii) Let F be a C  -module over C and
W C ! LB .E/ a -homomorphism such
that
.C / commutes with each operator in C. Consider F  E as a right
C  -A-B-bimodule via a.  / WD   a for all a 2 A,  2 F ,  2 E. By a
slight abuse of notation, we denote by 1  C  L.F  E/ the internal tensor
product of C with the C  -family generated by the identity operator on F .
Then 1  C is a C  -family, and the map T 7! 1  T defines a non-degenerate
morphism C ! 1  C. If
.hF jF i/  LB .E/ is non-degenerate, then this
morphism is injective. If the C  -family C is non-degenerate, then 1  C is
non-degenerate.

11.3.3 Hopf C  -families

The notion of a Hopf C  -family is the straightforward generalization of a bisimplifi-
able C  -bialgebra (Definition 4.1.1). For our applications to pseudo-multiplicative
unitaries, we define two variants which differ only in the usage of the flipped or
non-flipped internal tensor product:
Definition 11.3.18. A Hopf C  -family over a C  -algebra B is a non-degenerate
C  -family A on a right C  -B-B-bimodule equipped with a non-degenerate mor-
phism  W A ! M.A  A/ called the comultiplication such that
i) Œ.A/.1  A/ D A  A D Œ.A/.A  1/,
ii) .id / ı  D .  id/ ı , that is, the following diagram commutes:

A
 / M.A  A/

 id 
 
M.A  A/ / M.A  A  A/.
id

y
A flipped Hopf C  -family over a C  -algebra B is a non-degenerate C  -family A
on a right C  -B-B-bimodule equipped with a non-degenerate morphism  yWAy!
M.A y called the comultiplication such that
y  A/
 11.4. The legs of a decomposable pseudo-multiplicative unitary 359

y A/.1
i) Œ. y DA
y  A/ yA
y D Œ. y A
y A/. y  1/,
y ı
ii) .id / y D .
y  id/ ı .
y
Remark 11.3.19. Condition i) implies that the morphisms  and  y are non-
y 
degenerate; therefore we can extend the morphisms id ; id and id ; y id
to M.A  A/ and M.A y respectively.
y  A/,

11.4 The legs of a decomposable pseudo-multiplicative

unitary
O ˇ/ be a C  -trimodule over a C  -algebra B and V W E O  E ! E  ˇ E
Let .E; ˇ; ˇ
y /; /
a pseudo-multiplicative unitary. We shall construct legs .A.V y and .A.V /; /,
similarly as for a multiplicative unitary (see Section 7.2.1) or a pseudo-multipli-
cative unitary on a Hilbert space (see Section 10.3.2). The construction will be
interesting, however, only if the right C  -bimodule ˇ E or the right C  -bimodule
ˇO
E is decomposable. If that is the case and if the unitary V satisfies a certain
y /; /
regularity condition, then .A.V y and .A.V /; / are (flipped) Hopf C  -families.
For detailed proofs, we refer to the article [153].

y / and A.V /. The families A.V

The families A.V y / and A.V / are spanned by
operators of the following form:
Lemma 11.4.1. Let ;  2 PAut.B/.
i) Let  2 H .ˇ E/,  0 2 H .ˇ E/. Then we have homogeneous operators
jiŒ2 V j 0 
Œ2
ˇO
E ! ˇŒ2 .EˇO  E/ ! ˇO .E  ˇ E/ 
! ˇO E;
.;id/-hmg. .id;id/-hmg. Œ1 .id;/-hmg.

where jiŒ2  D    and j Œ2  D    for all  2 E. Put Œ 0 jŒ2 WD j 0 Œ2 .

0 0

The composition aO . 0;/ WD Œ 0 jŒ2 V jiŒ2 belongs to L .ˇO E/ and satisfies

h 0 jaO . 0;/ i D  .h 0   0 jV .  /i/ for all ;  0 2 E:

ii) Let  2 H.ˇO E/, 0 2 H .ˇO E/. Then we have homogeneous operators

jŒ1 V j 0 i
Œ1
ˇE ! ˇŒ2 .EˇO  E/ ! ˇO .E  ˇ E/ ! ˇ E;
.id;/-hmg. .id;id/-hmg. Œ1 .;id/-hmg.

where jŒ1  D    and j iŒ1  D    for all  2 E. Put h0 jŒ1 WD j0 iŒ1 .
0 0

The composition a.0;/ WD h0 jŒ1 V jŒ1 belongs to L .ˇ E/ and satisfies

h 0 ja.0;/ i D h0   0 jV .  /i for all ;  0 2 E:
360 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Proof. All assertions follow easily from Proposition 11.2.16, equation (10.6), and
Proposition 11.2.8. 
y /  L. O E/ and A.V /  L.ˇ E/ by
We define families A.V ˇ

y / WD span faO . 0;/ j  2 H .ˇ E/;  0 2 H .ˇ E/g  L . O E/

A.V   ˇ
and
A.V / WD span fa.0;/ j  2 H  .ˇO E/; 0 2 H .ˇO E/g  L .ˇ E/

for each ;  2 PAut.B/.

Applying the ket-bra notation to families of homogeneous elements, we can
y / and A.V / as follows. Define jˇ Ei  Lid .B; ˇ E/
rewrite the definition of A.V
and jˇ E  Lid .B; ˇ E/ by
0 0
jˇ Eiid WD fji j  2 H .ˇ E/g; jˇ Eid
 WD fj  j  2 H  .ˇ E/g

(see Proposition 11.2.15). Put hˇ Ej WD jˇ Ei and Œˇ Ej WD jˇ E . Replacing ˇ E by

ˇO
E, we similarly define jˇO Ei, hˇO Ej, jˇO E, ŒˇO Ej. To all of these families, we apply
the leg notation just like to individual ket-bra operators. Then
   
y / D Œˇ EjŒ2 V jˇ EiŒ2 and A.V / D h O EjŒ1 V j O EŒ1 :
A.V ˇ ˇ

For each  2 PAut.B/, b 2 H .B/, we have an .id;   /-homogeneous operator

(see the proof of Proposition 11.2.23)

˛.b/ W E ! E;  7! b:

Lemma 11.4.2. Let b 2 B, ;  0 2 ˇO E, ; 0 2 ˇ E be homogeneous. Then

O
aO . 0;/ ˇ.b/ D aO . 0;b/ ; aO . 0;/ ˛.b/ D aO . 0 b ;/ ; aO . 0;/ ˇ.b/ D ˇ.b/aO . 0;/ ;

O
ˇ.b/a O
. 0;/ D a. 0;/ ˇ.b/; ˛.b/a. 0;/ D a. 0;b/ ; ˇ.b/a.0;/ D a.0 b ;/ :

Proof. This follows from similar calculations as in the proof of Lemma 10.3.7, see
[153, Lemma 4.3]. 
O
To shorten the notation, we denote the family H.B/ by B. Define ˇ.B/ 
Lid .ˇO E/ and ˛.B/  Lid .ˇO E/ by

O  O
ˇ.B/id WD fˇ.b/ j b 2 H .B/g;  WD f˛.b/ j b 2 H  .B/g;
˛.B/id

and similarly define ˇ.B/  Lid .ˇ E/; ˛.B/  Lid .ˇ E/.

Given a right C  -bimodule F and a family C  L.F /, we denote by C0  L.F /
the family of all homogeneous operators that commute with all operators of C.
 11.4. The legs of a decomposable pseudo-multiplicative unitary 361

y /˛.B/ D ŒA.V
Proposition 11.4.3. i) ŒA.V O
y /ˇ.B/ D A.V y / is
y /  ˇ.B/0 . If A.V

a C -family, then it is a non-degenerate O.ˇO E/-module.
O
ii) Œ˛.B/A.V / D Œˇ.B/A.V / D A.V /  ˇ.B/0
. If A.V / is a C  -family,
then it is a non-degenerate O.ˇ E/-module.
Proof. This follows easily from Lemma 11.4.2, see [153, Proposition 4.4]. 
y / and A.V / are non-degenerate in the following sense:
The families A.V
y / E D E if ˇ E is decomposable.
Proposition 11.4.4. i) ŒA.V
ii) ŒA.V /E D E if ˇO E is decomposable.
y / H. O E/ D H. O E/ and
iii) If ˇ E and ˇO E are decomposable, then ŒA.V ˇ ˇ
ŒA.V / H.ˇ E/ D H.ˇ E/.
Proof. See [153, Proposition 4.5]. 
y / and A.V / are closed under multiplication. The proof
Next, we show that A.V
involves the following observation. If ˇ E is decomposable, then

ŒV ŒH.ˇ E/  E D ŒV H.ˇŒ1 .EˇO  E// (Proposition 11.2.20)

D H.ˇŒ1 .E  ˇ E// (Equation (10.6)) (11.6)
D ŒH.ˇ E/  H.ˇ E/ (Proposition 11.2.19 ii)).

y /A.V
Proposition 11.4.5. i) ŒA.V y / D A.V y / if ˇ E is decomposable.
ii) ŒA.V /A.V / D A.V / if ˇO E is decomposable.

y /A.V
Proof. We only prove assertion i). By definition, ŒA.V y /  L. O E/ is the
ˇ
family of closed subspaces spanned by all compositions of the form

jiŒ2 V Œ 0 jŒ2 jiŒ2 V Œ 0 jŒ2

aO . 0;/ aO . 0;/ W E !E Ǒ E !E ˇ E !E !E Ǒ E !E ˇ E !E;

where ;  0 ; ;  0 2 ˇ E are homogeneous. Moving Œ 0 jŒ2 to the left and jiŒ2 to the
right, we can write aO . 0;/ aO . 0;/ in the form

j  iŒ2 VŒ13 VŒ12 Œ 0   0 jŒ2

E !E Ǒ .E ˇ E /!.E Ǒ E /ˇ E !E ˇ E ˇ E !E:

Using the pentagon equation (10.7) and Proposition 11.2.19 ii), we find that the
y /A.V
product ŒA.V y / is equal to the family spanned by all compositions

j!iŒ2 VŒ23 VŒ12 VŒ23 Œ! 0 jŒ2
E !E Ǒ .E ˇ E /!E Ǒ E Ǒ E !E ˇ E Ǒ E !E ˇ E ˇ E 
!E;
362 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

where !; ! 0 2 ˇŒ1 .E  ˇ E/ are homogeneous. Now equation (11.6) implies that

y /A.V
ŒA.V y / is equal to the family spanned by all compositions

j#iŒ2 idE jiŒ2 VŒ12 idE h0 jŒ2 Œ# 0 jŒ2

E !E Ǒ E !E Ǒ E Ǒ E !E ˇ E Ǒ E !E ˇ E !E;
where #; # 0 2 ˇ E are homogeneous and ; 0 2 E are arbitrary. Since
O 0 ji/;
.id h0 jŒ2 /VŒ12 D V .id h0 jŒ2 /; .id h0 jŒ2 /.id jiŒ2 / D id ˇ.h
the composition above is equal to

j#iŒ2 ˇOŒ2 .h0 ji/V Œ# 0 jŒ2

E !E Ǒ E !E ˇ E !E;

O 0 ji/#. Note that # 00 2 ˇ E is homogeneous

that is, to aO .# 0;# 00 / , where # 00 D ˇ.h
because ˇO commutes with ˇ. Using the fact that E is full and that ˇO is non-
degenerate, we find that ŒA.V y /A.V y / is equal to the family spanned by all operators
0 00
aO .# 0;# 00 / , where # ; # 2 ˇ E are homogeneous. This is A.V y /. 

The comultiplications y and . Denote by B y  L. O E/ and B  L.ˇ E/ the

ˇ
 y / and A.V /, respectively. By Lemma 11.4.1, B
C -families generated by A.V y and
O
B commute with ˇ.B/ and ˇ.B/, respectively, so that we can define morphisms
y ! L. O .E  ˇ E//; aO 7! 1  a;
B O B ! L.ˇŒ1 .EˇO  E//; a 7! a  1;
ˇ Œ2

(see Example 11.3.17 iii)). Composing with conjugation by V  or V , respectively,

we obtain morphisms (see Example 11.3.17 ii) and equation (10.6))
y D
 y ! L. O .E O  E//;
yV W B aO 7! V  .1  a/V;
O
ˇ ˇ Œ2

 D V W B ! L.ˇŒ1 .E  ˇ E//; a 7! V .a  1/V  :

Proposition 11.4.6. i) If ˇ E is decomposable and Œ. y B y / D B

y /.1  B y B
y D
y B
Œ. y /.B y ; /
y  1/, then .B y is a flipped Hopf C -family.


ii) If ˇO E is decomposable and Œ.B/.1  B/ D B  B D Œ.B/.B  1/, then

.B; / is a Hopf C  -family.
Proof. We only prove assertion i); the proof of assertion ii) is similar. Let us
make the assumptions stated in i). By Proposition 11.4.4, the C  -family B y is
non-degenerate, and by the second assumption,  y is a non-degenerate morphism
y ! M.B
B y By /. It remains to show that  y  , where
y is coassociative. Let aO 2 B
y a/
;  2 PAut.B/. By definition . 
O D V .1  a/V O , and hence
. y a//
y  id/.. 
O D VŒ12 
VŒ23 O Œ23 VŒ12 ;
.1  1  a/V
 11.4. The legs of a decomposable pseudo-multiplicative unitary 363

where VŒ23 VŒ12 W EˇO  EˇO  E ! E ˇ EˇO  E ! E  ˇ E  ˇ E. We squeeze

in conjugation VŒ12 and find
y  id/..
. y a// 
O D VŒ12 
VŒ23 
O Œ12
VŒ12 ..1  1/  a/V VŒ23 VŒ12 ;

where VŒ12 .VŒ23 VŒ12 / W EˇO  EˇO  E ! E  ˇ E  ˇ E ! .EˇO  E/ 

ˇ E. By the pentagon equation (10.7), VŒ12 VŒ23 VŒ12 is equal to the composition
VŒ13 VŒ23 W EˇO  EˇO  E ! EˇO  .E  ˇ E/ ! .EˇO  E/  ˇ E. Therefore,

y  id/..
. y a// 
O D VŒ23 
VŒ13 y .
O Œ13 VŒ23 D .id /.
..1  1/  a/V y a//:
O 
When we pass from V to V op , the legs get switched as follows:
Proposition 11.4.7. We have
y op / D A.V / ; 
A.V y V op D Ad† ıV ;
y / ; V op D Ad† ı
A.V op / D A.V yV :

Proof. The proof is similar to the proof of Lemma 7.2.5. 

Regularity and decomposability conditions on V . Evidently, it is desirable

to have simple criteria that tell whether the legs of a given pseudo-multiplicative
unitary are Hopf C  -families or not. We shall see that the regularity condition
known from multiplicative unitaries (Section 7.3.1) can be adapted to that purpose,
but has to be refined by additional assumptions.
First, let us explain how the regularity condition can be adapted to the present
setting. As before, let V W EˇO  E ! E  ˇ E be a pseudo-multiplicative unitary.
For each ; 0 2 E, we have adjointable operators
jiŒ2 W E ! EˇO  E;  7!   ; j0 iŒ1 W E ! E  ˇ E;  7! 0  :

Put h0 jŒ1 WD j0 iŒ1 . The regularity condition involves operators of the form

c.0 ;/ WD h0 jŒ1 V jiŒ2 2 LB .E/

and the space
 
C .V / WD span fc.0 ;/ j ; 0 2 Eg D hEjŒ1 V jEiŒ2 :
Definition 11.4.8. We call a pseudo-multiplicative unitary V W EˇO  E ! E  ˇ E
semi-regular if KB .E/  C .V /, and regular if KB .E/ D C.V /.
y /; /
To ensure that the left leg .A.V y or the right leg .A.V /; / of V is a Hopf
C  -family, we need to impose stronger assumptions than regularity alone. The first
step towards this refinement is the following observation:
364 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Lemma 11.4.9. i) c.0 ;/ 2 Lid .ˇ E/ for all  2 E, 0 2 H  .ˇ E/,  2 PAut.B/.

ii) c.0 ;/ 2 Lid .ˇO E/ for all  2 H .ˇO E/, 0 2 E,  2 PAut.B/.
Proof. i) If ; 0 ;  are as in i), then we have homogeneous operators
jiŒ2 V h 0 jŒ1
ˇE ! ˇŒ1 .EˇO  E/ ! ˇŒ1 .E  ˇ E/ ! ˇ E
.id;id/-hmg. .id;id/-hmg. .;id/-hmg.

by Proposition 11.2.16 and equation (11.1), and the composition c.0 ;/ is .; id/-
homogeneous by Proposition 11.2.8.
ii) If ; 0 ;  are as in ii), then we have homogeneous operators
jiŒ2 V h 0 jŒ1
ˇO
E ! ˇO .EˇO  E/ ! ˇO .E  ˇ E/ ! ˇO E;
.;id/-hmg. Œ2 .id;id/-hmg. Œ2 .id;id/-hmg.

and the claim follows as above. 

y /  Lid .ˇ E/ and C.V /  Lid . O E/ by
We define families C.V ˇ

y / WD span fc.0 ;/ j  2 E; 0 2 H  .ˇ E/g;

C.V id

C.V /id WD span fc.0 ;/ j  2 H .ˇO E/; 0 2 Eg

for all  2 PAut.B/. In the notation introduced
 before Lemma 11.4.2, we have
y / D hˇ EjŒ1 V jEiŒ2 and C.V / D hEjŒ1 V j O EiŒ2 .
C.V ˇ
y / and C.V /, the C  -
The following definition involves the C  -families C.V
families Kid .ˇ E/ and Kid .ˇO E/ that were defined before Proposition 11.2.19, the
family F.ˇ E/  H.ˇ E/ defined by
F .ˇ E/ WD span f1 3 .h3 j2 i/ j i 2 Hi .ˇ E/; i 2 PAut.B/; 1 2 3   g
 H .ˇ E/;
and the family F.ˇO E/  H.ˇO E/ which is defined similarly.
Definition 11.4.10. We call a pseudo-multiplicative unitary V W EˇO  E ! E  ˇ E
decomposably left regular if the following conditions hold:
i/ ˇ E is decomposable, y / D Kid .ˇ E/;
ii/ C.V iii/ F.ˇ E/ D H.ˇ E/:
We call V decomposably right regular if the following conditions hold:
i/ ˇO E is decomposable, ii/ C.V / D Kid .ˇO E/; iii/ F.ˇO E/ D H.ˇO E/:
Theorem 11.4.11. Let V W EˇO E ! E ˇ E be a pseudo-multiplicative unitary. If
y /; /
V is decomposably left / right regular, then .A.V y / .A.V /; / is a flipped /ordi-
nary Hopf C  -family.
Proof. See [152]. 
 11.5. Coactions of Hopf C  -families 365

The legs of the pseudo-multiplicative unitary of a groupoid. Let us determine

the legs of the pseudo-multiplicative unitary WG associated to a decomposable
groupoid G (see Example 11.1.5). We use the same notation as in Example 11.1.5
and in Proposition 11.2.32, and denote by
W C0 .G/ D B ! LB .L2 .G; // the
representation given by multiplication operators.
Proposition 11.4.12. Let G be a locally compact, second countable, Hausdorff
groupoid with left Haar system . If G is decomposable, then the pseudo-multipli-
cative unitary WG is decomposably left regular and decomposably right regular.
Proof. See [152]. 
Proposition 11.4.13. Let G be a locally compact, second countable, Hausdorff
groupoid with left Haar measure .
y G / is a C  -family, and
i) The family A.W
y G /  A.W
A.W y G /id D
.C0 .G// for all ;  2 PAut.B/:
 id

For all f 2 C0 .G/;  2 L2 .G; /s  L2 .G; / and .x; y/ 2 Gs;r

2
,

y
..
.f ///.x; y/ D f .xy/.x; y/:

y G /; /
In particular, .A.W y is a Hopf C  -family.

ii) If G is decomposable, then A.WG / is a C  -family, and

A.WG / D span fL.f / j f 2 Cc .U /; U  G open homogeneous;

qU   and qU    g

for all ;  2 PAut.B/. For every open homogeneous subset U  G and all
f 2 Cc .U /,  2 L2 .G; /  r L2 .G; /, .x; y/ 2 Gr;r
2
,
Z
..L.f ///.x; y/ D f .z/.z 1 x; z 1 y/d rG .x/ .z/:
G rG .x/

Proof. This follows from Theorem 11.4.11, Proposition 11.4.12, and similar cal-
culations as in the case where G is a group (see Example 7.2.13). 

11.5 Coactions of Hopf C  -families

This section gives a brief overview on coactions of Hopf C  -families, which were
introduced and studied in [152]. We state the pertaining definitions and explain
how such coactions simultaneously generalize actions of groupoids on C  -algebras
366 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

and Fell bundles on decomposable groupoids. Let us mention that one can define
reduced crossed products for coactions of Hopf C  -families and prove an analogue
of the Takesaki–Takai–Baaj–Skandalis duality theorem in a similar way as it was
done for coactions of C  -bialgebras in Chapter 9; see [152].
Throughout this section, let A and B be a C  -algebra.
The concept of a coaction carries over to Hopf C  -families as follows:

Definition 11.5.1. Let .A; / be a Hopf C  -family on a right C  -B-B-bimodule

and C a non-degenerate C  -family on a right C  -A-B-bimodule.
A .right/ coaction of .A; / on C is a non-degenerate morphism ı W C ! M.C 
A/ such that Œı.C/.1  A/  C  A and .ı  id/ ı ı D .id / ı ı. We call .C; ı/
an .A; /-C  -family if ı is injective and Œı.C/.1  A/ D C  A.
Let D be a non-degenerate C  -family on a right C  -A-B-bimodule, and let
ıC and ıD be coactions of .A; / on C and D, respectively. A non-degenerate
morphism  W C ! M.D/ is called covariant if ıD ı  D .  id/ ı ıC .

Remark 11.5.2. Similarly as above, one can define left coactions of Hopf C  -fami-
lies and right and left coactions of flipped Hopf C  -families.

To consider coactions of Hopf C  -families on C  -algebras, we need to con-

struct an internal tensor product of C  -families with C  -algebras. The necessary
definitions are straightforward. Nevertheless, for Hopf C  -families, it seems to be
more natural to consider coactions on C  -families than coactions on C  -algebras.

Definition 11.5.3. A C  -B-algebra is a C  -algebra C equipped with a non-

degenerate -homomorphism B ! M.C /. A morphism between C  -B-algebras
C and D is a non-degenerate -homomorphism
W C ! M.D/ that satisfies

.bc/ D b
.c/ for all b 2 B and c 2 C . Given a C  -B-algebra C , we define a
family H.C / D .H .C // 2PAut.B/ by

H .C / WD fc 2 ŒC Dom. / j cb D .b/c for all b 2 Dom. /g;  2 PAut.B/:

Let C be a C  -B-algebra. Then we can consider C as a right C  -B-C -

bimodule, and each element c 2 H .C /, where  2 PAut.B/, can be considered as
a .; id/-homogeneous operator via left multiplication. Thus we can identify H.C /
with a subfamily of Lid .C /. Evidently H.C /  Lid .C / is a C  -family.

Definition 11.5.4. Let C be a C  -family on a C  -A-B-bimodule E, and let C be a

C  -B-algebra. The internal tensor product C  C is the subspace of LC .E  C /
given by

C  C WD span fc 0  c j c 0 2 C ; c 2 H .C /; where
2 PAut.A/; ;  2 PAut.B/;    g:
 11.5. Coactions of Hopf C  -families 367

The internal tensor product constructed above has the following properties:

Proposition 11.5.5. Let C; E and C be as in the preceding definition.

i) The space C  C is a C  -algebra.

ii) If B is decomposable and C is a non-degenerate O.E/-module, then the
natural representation A ! LC .E  C / turns C  C into a C  -A-algebra.
iii) For every morphism of C  -B-algebras
W C ! M.D/, there exists a non-
degenerate -homomorphism

id 
W C  C ! M.C  D/

such that .id 

/.c 0  c/ D c 0 
.c/ for all c 0 2 C , c 2 H .C /, where
2 PAut.A/, ;  2 PAut.B/;    . If B is decomposable and C is a
non-degenerate O.E/-module, then id 
is a morphism of C  -A-algebras.
iv) Let D be a C  -family on some right C  -A-B-bimodule F . Assume that A and
B are decomposable and that C and D are non-degenerate O.E/- or O.F /-
modules, respectively. Then every non-degenerate morphism of C  -families
W C ! M.D/ induces a morphism of C  -A-algebras

 id W C  C ! M.D  C /

such that .  id/.c 0  c/ D .c 0 /  c for all c 0 2 C ; c 2 H .C /, where

2 PAut.A/, ;  2 PAut.B/;   .
v) The internal tensor product is associative in the natural sense.

Proof. See [152]. 

Similarly as above, one can define a flipped internal tensor product C  C which
has analogous properties. In the following definition, we use this flipped internal
tensor product because we want to focus on right coactions:

Definition 11.5.6. Let .A; / be a Hopf C  -family on a right C  -B-B-bimodule E.

Assume that B is decomposable and that A is a non-degenerate O.E/-module.

i) A .right/ coaction of .A; / on a C  -B-algebra C is a morphism of C  -

B-algebras ı W C ! M.C  A/ such that ı.C /.1  A/  C  A and
.ı  id/ ı ı D .id / ı ı.
ii) A morphism of C  -B-algebras
W C ! M.D/ is called covariant with
respect to coactions ıC and ıD of .A; / on C and D, respectively, if ıD ı
D
.
 id/ ı ıC .
368 Chapter 11. Pseudo-multiplicative unitaries on C  -modules

Coactions, groupoid actions, and Fell bundles on groupoids. Coactions of

Hopf C  -families generalize groupoid actions and Fell bundles on decomposable
groupoids, very much like coactions of C  -bialgebras generalize group actions and
Fell bundles on groups (see Section 9.2). Let us briefly outline these relations.
Consider a locally compact, second countable, Hausdorff groupoid G with
left Haar system , and assume that G is decomposable. The associated pseudo-
multiplicative unitary WG (Example 11.1.5) gives rise to a flipped Hopf C  -family
y G /; /
.A.W y and a Hopf C  -family .A.WG /; / (Proposition 11.4.13, 11.4.12).
Note that the Hopf C  -family .A.WG /; / can also be considered as a flipped
Hopf C  -family: We have a natural isomorphism

L2 .G; /  r L2 .G; / Š L2 .G; /r  L2 .G; /;

which induces an isomorphism A.WG /  A.WG / Š A.WG /  A.WG /.

The concepts of a group action and of a Fell bundle generalize to groupoids
easily, see, for example, [98] and [85].
Theorem 11.5.7. Let G be a groupoid as above.
y G /; /
i) Injective coactions of the flipped Hopf C  -family .A.W y on a C  -
C0 .G /-algebra C correspond bijectively with continuous actions of the
0

groupoid G on C .
ii) Assume that G is r-discrete. Then injective coactions of the flipped Hopf
C  -family .A.WG /; / on C  -C0 .G 0 /-algebras correspond bijectively .up
to isomorphism/ with upper semi-continuous Fell bundles on G.
Proof. See [152]. 
Chapter 12
Appendix

12.1 C  -algebras
Standard references on C  -algebras are, for example, [34], [113], [121], [137],
[149].

Basic definitions and facts. A -algebra is a complex algebra A with a conjugate-

linear antiautomorphism  W A ! A called the involution of A. A -homomorphism
between two -algebras is a homomorphism of complex algebras that intertwines
the respective involutions.
A Banach algebra is an algebra A equipped with a norm k  k such that A
is complete and kabk  kak  kbk for all a; b 2 A. A Banach -algebra is a
Banach algebra A that is simultaneously a -algebra such that kak D ka k for all
a 2 A. A Banach -algebra A is a C  -algebra if ka ak D kak2 for all a 2 A. A
C  -.semi/norm on a -algebra A is a (semi)norm that satisfies ka k D kak and
ka ak D kak2 for all a 2 A.
An approximate unit of a Banach algebra A is a net .u / of elements of A that
satisfies lim u a D a D lim au for all a 2 A.
Let A be a C  -algebra. An element a 2 A is central if ab D ba for all b 2 A, a
partial isometry if aa a D a, an isometry if a a D 1, a unitary if a a D aa D 1,
self-adjoint if a D a , a projection if a2 D a D a , positive if it can be written in the
form a D b  b for some b 2 A. The natural order on A is given by a b W, a  b
is positive.
Every C  -algebra contains an approximate unit consisting of positive elements
which have norm less than 1.
By an ideal of A we mean a two-sided ideal in the algebraic sense that is closed
with respect to the involution and with respect to the norm; thus, every ideal is a
C  -algebra again.
If A is non-unital, then the unitization of A, denoted by AC , is the C  -algebra
with underlying vector space A ˚ C and multiplication, involution, and norm given
by

N
.a; /  .b; / WD .ab C b C a; /; .a; / WD .a ; /;
k.a; /k WD supfkac C ck j c 2 A; kck  1g for all a; b 2 A; ;  2 C:

The algebra A is contained in AC as a two-sided ideal.

370 Chapter 12. Appendix

The multiplier algebra M.A/ of A defined in Section 2.1.1 is a C  -algebra

with respect to the operator norm kT k WD supfkT ak j a 2 A; kak  1g. The
strict topology on M.A/ is the topology defined by the seminorms a ; a (a 2 A),
where a .T / D kT ak and a .T / D kaT k for all a 2 A and T 2 M.A/. The
C  -algebra A is dense in M.A/ with respect to the strict topology: if .u / is a
bounded approximate unit for A and T 2 M.A/, then .T u / is a net in A that
converges strictly to T .
Every -homomorphism of C  -algebras is automatically norm-decreasing. Let
A and B be C  -algebras. A -homomorphism
W A ! M.B/ is called non-
degenerate if
.A/B is linearly dense in B. In that case,
extends uniquely to a
strictly continuous -homomorphism M.A/ ! M.B/.
For each unitary U 2 A, we define AdU W A ! A by a 7! UaU  ; this is a
-automorphism.

Commutative C  -algebras and Gelfand duality. The category of all commuta-

tive C  -algebras with non-degenerate -homomorphisms is equivalent to the cat-
egory of all locally compact Hausdorff spaces with proper continuous maps. This
correspondence is established as follows:
Let X be a locally compact Hausdorff space. Then the algebra C0 .X / of
complex-valued functions on X vanishing at infinity, equipped with the supre-
mum norm, is a commutative C  -algebra. The multiplier algebra M.C0 .X // is
canonically isomorphic to the C  -algebra Cb .X / of all bounded continuous func-
tions on X. A map  W X ! Y between locally compact Hausdorff spaces induces
a non-degenerate -homomorphism   W C0 .Y / ! M.C0 .X // via .  f /.x/ D
f ..x//. The image of   is contained in C0 .X / if and only if  is proper.
Conversely, let A be a commutative C  -algebra. A character on A is a non-zero
-homomorphism A ! C. Denote by Ay the set of all characters of A endowed
with the weakest topology that makes all functions of the form  7! .a/, where
a 2 A and  2 A, y continuous. Then Ay is a locally compact Hausdorff space, and
the Gelfand–Naimark theorem says that the natural map A ! C0 .A/ y is a natural
isomorphism. Each -homomorphism  W A ! B of commutative C  -algebras
induces a proper continuous map   W By ! Ay via  7!  ı .

Representations. For every Hilbert space H , the space L.H / of all bounded
linear operators on H is a C  -algebra, and the subspace K.H /  L.H / of all
compact linear operators is an ideal. Moreover, L.H / D M.K.H //.
A .bounded/ representation of a -algebra A on a Hilbert space H is a -
homomorphism
W A ! L.H /. The representation
is called faithful if
is
injective, and non-degenerate if
.A/H is linearly dense in H . Note that if A is a
C  -algebra, then every -homomorphism A ! L.H / is automatically bounded.
 12.1. C  -algebras 371

Positive functionals and states. Let A be a C  -algebra. We denote by A0 the space

of continuous linear functionals A ! C. A linear map  W A ! B of C  -algebras
is called positive if it preserves the order. A positive map is always continuous.
More precisely, if  2 A0 is positive, then kk D lim j.u /j < 1 for every
approximate unit .u / of A. In particular, kk D .1/ if A is unital. A state on a
C  -algebra A is a positive linear functional of norm one.
By the Hahn–Banach theorem, there exists for each a 2 A a  2 A0 such that
.a/ D kak and kk D 1. For each ! 2 A0 , there exist positive functionals
!1 ; : : : ; !4 2 A0 such that ! D !1  !2 C i.!3  !4 /.

GNS-construction. A GNS-construction for a state  on a C  -algebra A consists

of a Hilbert space H , a linear map ƒ W A ! H with dense image, and a represen-
tation
W A ! L.H /, such that hƒ.a/jƒ.b/i D .a b/ and
.a/ƒ.b/ D ƒ.ab/
for all a; b 2 A. It is easy to see that the GNS-construction is unique up to a uni-
tary transformation. For every GNS-construction .H; ƒ;
/, there exists a unique
vector  2 H , called the cyclic vector, such that ƒ.a/ D
.a/ for all a 2 A;
moreover, kk D 1 and hj
.a/i D .a/ for all a 2 A. If .u / is a bounded
approximate unit for A, then .
.u // converges in norm to  for each  2 H .
A particular GNS-construction for  can be obtained as follows. By the Cauchy–
Schwarz inequality, the set N WD fa 2 A j .a a/ D 0g  A is a subspace (even
a left ideal). Denote by H the completion of the quotient A=N with respect to
the inner product ha C N jb C N i WD .a b/. Then H is a Hilbert space, the
map ƒ W A ! H; a 7! a C N , has dense image, and there exists a representation

W A ! L.H / such that .H; ƒ;

/ is a GNS-construction.
The universal GNS-representation for a C  -algebra A is obtained as follows.
L each state  on A, choose a GNS-representation .H
; ƒ
;

/. Put H WD
For

H
, and define
W A ! L.H / by .
.a//
D

.a/
for every a 2 A;  D
.
/
2 H , and every state . Then
is called the universal GNS-representation
for A. This representation is always faithful; in particular, A is isomorphic to

.A/  L.H /.

Enveloping C  -algebra of a -algebra. Let A be a -algebra. If for every a 2 A,

the supremum

jaj WD supfk
.a/k j
W A ! B is a -homomorphism; B a C  -algebrag

is finite, then j  j W a ! jaj is a C  -seminorm on A. The completion of the quotient

A=fa 2 A j jaj D 0g with respect to the norm induced by j  j is a C  -algebra,
called the enveloping C  -algebra of A. Denote this C  -algebra by Ae and the
natural map A ! Ae by qe . By definition, Ae and qe have the following universal
property: For every -homomorphism
W A ! B, where B is a C  -algebra, there
372 Chapter 12. Appendix

exists a unique -homomorphism

e W Ae ! B such that
D
e ı qe . Note that
for each a 2 A,

jaj D supfk
.a/k j
is a bounded representation
of A on some Hilbert space H g:

The minimal tensor product. The minimal tensor product of two C  -algebras A
and B is defined as follows. Denote by
A W A ! L.HA / and
B W B ! L.HB / the
universal GNS-representations of A and B, respectively. Then the minimal tensor
product A ˝ B is the C  -subalgebra of L.HA ˝ HB / generated by all operators
of the form
A .a/ ˝
B .b/, where a 2 A and b 2 B.
One has a canonical inclusion M.A/ ˝ M.B/  M.A ˝ B/.
The minimal tensor product is functorial, that is, for each pair of -homo-
morphisms i W Ai ! Bi of C  -algebras, where i D 1; 2, there exists a -homo-
morphism 1 ˝ 2 W A1 ˝ A2 ! B1 ˝ B2 such that .1 ˝ 2 /.a1 ˝ a2 / D
1 .a1 / ˝ 2 .a2 / for all ai 2 Ai .
For each state  on A and each state on B, there exists a unique product state
 ˝ on A ˝ B such that . ˝ /.a ˝ b/ D .a/ .b/ for all a 2 A and b 2 B.
More generally, for each  2 A0 and 2 B 0 , there exists a unique functional
0
 ˝ 2 .A ˝ B/ such that . ˝ /.a ˝ b/ D .a/ .b/ for all a 2 A, b 2 B,
and k ˝ k  kkk k [149, Section IV.4, page 208].

Complements on linear functionals. The following results are well known, but
it seems useful to include proofs.

Proposition 12.1.1. Let A be a C  -algebra. Then each of the sets

f.  b/ j  2 A0 ; b 2 Ag; f.a  / j  2 A0 ; a 2 Ag;

f.a  b/ j  2 A0 ; a; b 2 Ag

is equal to A0 .

Proof. We only prove the assertion concerning the first set; the remaining assertions
follow similarly. The map A  A0 ! A0 , .b; / 7! .  b/, turns A0 into a Banach
module over A. We shall show that AA0 is linearly dense in A0 , and then the Cohen
Factorization Theorem [62, Theorem 32.22], [36] implies AA0 D A0 .
Let  2 A0 be a state with GNS-construction .H; ƒ;
/ and cyclic vector .
Choose an approximate unit .u / for A. Since lim
.u / D , the functional
.  u / D hj
.  /
.u /i converges in norm to  D hj
.  /i as  tends to
infinity. Using the fact that every functional can be written as a linear combination
of states, we find that AA0 is dense in A0 . 
 12.2. C  -modules 373

Corollary 12.1.2. Let A be a C  -algebra and  2 A0 . Then  extends uniquely

Q D kk. If  is posi-
to a strictly continuous functional Q 2 M.A/0 of norm kk
Q
tive/a state, so is . 

Proposition 12.1.3. Let A be a C  -algebra and T 2 M.A/. Then kT k D

Q / j ! 2 A0 ; k!k  1g.
supf!.T

Proof. Evidently, kT k supf!.TQ / j ! 2 A0 ; k!k  1g. Let .u / be an

approximate unit for A such that ku k < 1 for all . Then kT u k  kT kku k 
kT k for all  and lim T u a D T a for all a 2 A, whence lim kT u k D kT k.
Given > 0, choose an index  and an ! 2 A0 of norm 1 such that jkT u kkT kj <
=2 and kT u k D !.T u /. Then kT k < kT u k C =2 D !.T u / C =2 
Q / C .
!.T 

12.2 C  -modules
An excellent reference on C  -modules is [95]; other sources are, for example, [16,
Section II.7], [57, Section 2.5], and [128, Chapter 2].

Definition and basic facts. Let A be a C  -algebra. A .right/ pre-C  -module

over A or a .right/ pre-C  -A-module is a complex vector space E equipped with a
right module structure over A and a sesquilinear map h  j  i W E  E ! A such that

hji D hji; hjai D hjia; hji 0 for all ;  2 E; a 2 A:

The assignment kk WD khjik1=2 defines a seminorm on E. Evidently, k  k is

a norm if and only if hji ¤ 0 for all  2 E;  ¤ 0. If k  k is a norm and E is
complete with respect k  k, then E is called a .right/ C  -module over A or a .right/
C  -A-module.
To every pre-C  -module E over A, one can associate a C  -module over A as
follows. The subset N WD f 2 E j kk D 0g is easily seen to be a right submodule,
and the quotient E=N is a pre-C  -module over A with respect to the structure maps
h C N j C N i WD hji and . C N /a WD a C N . The completion of E=N with
respect to the norm is a C  -module over A.
If E is a C  -module over A, then EA is linearly dense in E. Indeed, if .u /
is a bounded approximate unit for A, then for all  2 E,
!1
k  u k2 D kh  u j  u ik D .1  u /hji.1  u / ! 0:

The Cohen Factorization Theorem [62, Theorem 32.22], [36] implies that one even
has EA D E.
A C  -module E over A is called full if the set hEjEi is linearly dense in A.
374 Chapter 12. Appendix

A basic example of a C  -module is the C  -algebra A itself, with inner product

hajbi WD a b and the obvious right module structure. Another basic example:
every Hilbert space can be considered as a C  -module over C.

Operators on C  -modules. Let E and F be C  -modules over a C  -algebra A.

An adjoint of a map T W E ! F is a map S W F ! E that satisfies hjT i D
hS ji for all  2 F and  2 E. If it exists, the adjoint of T is unique; it is
denoted by T  . In this case, T and T  are bounded, A-linear in the sense that
T .a/ D .T /a; T  .a/ D .T  /a for all  2 E;  2 F; a 2 A, and T and
T  satisfy kT k2 D kT  T k D kT  k2 . The space of all adjointable operators
from E to F is denoted by LA .E; F /. If E D F , this space is also denoted by
LA .E/. Equipped with the natural operations and the operator norm, LA .E/ is a
C  -algebra. The space LA .E; F /, equipped with the inner product hSjT i WD S  T
and the right module structure given by composition, is a C  -module over LA .E/.
For each pair of elements  2 E and  2 F , the map jihj W E ! F given
by  7! hji defines an adjointable operator E ! F . Its adjoint is given by the
map jihj W F ! E. An operator T 2 LA .E; F / is called compact if it can be
approximated in norm by linear combinations of such elementary operators. The
set of all compact operators is denoted by KA .E; F /; if E D F , this space is
also denoted by KA .E/. The composition of a compact operator with an arbitrary
operator is compact again, and KA .E/ is an ideal in LA .E/. Moreover, LA .E/ D
M.KA .E//. The space KA .A; F / can be identified with F via jihaj
a for
all  2 F; a 2 A.
The strict topology on LA .E; F / is the topology generated by the family of
seminorms  ;  2 E, and  ;  2 F , where  .T / D kT k and  .T / D kT  k
for all  2 E;  2 F; T 2 LA .E; F /. The subspace KA .E; F / is strictly dense in
LA .E; F /.
A representation of a C  -algebra B on E is a -homomorphism
W B !
LA .E/. The representation is called faithful if it is injective, and non-degenerate
if
.B/E is linearly dense in E.

Amplification. Let E be a C  -module over a C  -algebra A, and let H be a

Hilbert space. Then the algebraic tensor product E ˇ H is a pre-C  -module over
A with respect to the structure maps
h0 ˇ  0 j ˇ i WD h0 jiE  h 0 jiH and . ˇ /a WD a ˇ 
for all ; 0 2 E, ;  0 2 H , a 2 A. The completion of this pre-C  -module is
a C  -module over A, which we denote by E ˝ H . Similarly, we can define a
C  -A-module H ˝ E.
For each S 2 LA .E/ and T 2 L.H /, there exists an operator S ˝ T 2
LA .E ˝ H / such that .S ˝ T /. ˝ / D S  ˝ T  for all  2 E;  2 H . The
 12.3. Von Neumann algebras 375

map LA .E/ ˇ L.H / ! LA .E ˝ H / given by S ˇ T 7! S ˝ T extends

to a -homomorphism LA .E/ ˝ L.H / ! LA .E ˝ H /, which restricts to a
Š
-isomorphism KA .E/ ˝ K.H / 
! KA .E ˝ H /.

Internal tensor product. Let E and F be C  -modules over C  -algebras A and B,

respectively. Assume that we have fixed a representation A ! LB .F /. Then the
algebraic tensor product E ˇ F is a pre-C  -module over B via

h0 ˇ  0 j ˇ i WD h 0 jh0 jiE iF and . ˇ /b WD  ˇ b

for all ; 0 2 E, ;  0 2 F , b 2 B. The associated C  -module over B is denoted

by E ˝A F and called the internal tensor product of E and F . For each  2 E
and  2 F , we denote by  ˝A  the image of  ˇ  under the natural map
E ˇ F ! E ˝A F .
If A D C and the representation of A D C on F is given by scalar multiplication,
then F is a Hilbert space and F ˝C E is equal to the C  -module F ˝ E introduced
above.
If E D A and the representation of A on F is non-degenerate, then there exists
Š
an isomorphism A ˝A F  ! F; a ˝A  7! a.
We shall frequently use the following result [39, Proposition 1.34]:
Proposition 12.2.1. For i D 1; 2, let Ei be a C  -module over a C  -algebra A and
Fi a C  -module over a C  -algebra B with a fixed representation A ! LB .Fi /.
Let S 2 LA .E1 ; E2 / and T 2 LB .F1 ; F2 /. If T a D aT  for all a 2 A and
 2 F1 , then there exists a unique operator S ˝A T 2 LB .E1 ˝A F1 ; E2 ˝A F2 /
such that .S ˝A T /. ˝A / D S ˝A T  for all  2 E1 and  2 F1 . Moreover,
.S ˝A T / D S  ˝A T  . 
The internal tensor product of C  -modules is associative in an obvious sense.

12.3 Von Neumann algebras

Standard references on von Neumann algebras are, for example, [35], [75], [76],
[137], [149], [150].

Topologies on L.H /. Let H be a Hilbert space. Then L.H / can be equipped

with the several locally convex topologies:
• The family of seminorms ; given by ; .T / WD jhjT ij, where ;  2 H ,
defines the weak .operator/ topology on L.H /.
• The family of seminorms  given by  .T / WD kT k, where  2 H ,
defines the strong .operator/ topology on L.H /.
376 Chapter 12. Appendix

• The family of seminorms  ;  , where  is given by  .T / WD kT  k

for each  2 H , defines the strong- .operator/ topology on L.H /.
• The family of seminorms Q ; given by Q ; .T / WD jhj.T ˝ id/ij, where
;  2 H ˝ l 2 .N/, defines the -weak .operator/ topology on L.H /.
• The family of seminorms Q  given by Q  .T / WD k.T ˝ id/k, where  2
H ˝ l 2 .N/, defines the -strong .operator/ topology on L.H /.
• The family of seminorms Q  ; Q  , where Q  is given by Q  .T / WD k.T  ˝
id/k for each  2 H ˝ l 2 .N/, defines the  -strong- .operator/ topology
on L.H /.

These topologies are related as follows:

norm   -strong-   -strong   -weak

S S S

strong-  strong  weak;

where “” means that the left-hand side is finer/stronger than the right-hand side.

Definition and basic facts. Let H be a Hilbert space. The commutant of a subset
X  L.H /, usually denoted by X 0 , is the set of all T 2 L.H / that commute with
every element of X . Evidently, X 0 is an algebra, and if X  D X , then X 0 is a
-algebra.
By von Neumann’s double commutant theorem, the following conditions on a
C  -algebra A  L.H / containing idH are equivalent: (i) A is equal to the double
commutant A00 , (ii) A is closed with respect to the weak topology. If these conditions
hold, then A is called a von Neumann algebra.
A linear functional on a von Neumann algebra is called normal if it is  -weakly
continuous. The space of all normal linear functionals on a von Neumann algebra
A is denoted by A . If A is a von Neumann algebra on a Hilbert space H , then
every normal linear functional ! 2 A can be written in the form
X
!.x/ D hn jxn i for all x 2 A;
n
P
whereP.n /n and .n /n are sequences of vectors in H satisfying n kn k2 < 1
and n kn k2 < 1. Conversely, if ! has the form above, then it is normal.
For each element x of a von Neumann algebra A, there exists a normal functional
! 2 A such that kxk D !.x/ and k!k D 1.
More generally, a map A ! B of von Neumann algebras is called normal if it
is continuous with respect to the -weak topologies on A and B, respectively.
 12.4. Slice maps 377

Tensor product. Let H and K be Hilbert spaces. The tensor product of von
Neumann algebras N  L.H / and M  L.K/, which we denote by N x̋ M , is
the von Neumann algebra on H ˝K generated by the operators S ˝T 2 L.H ˝K/,
where S 2 N and T 2 M . In particular, one easily finds L.H / x̋ L.K/ D
L.H ˝ K/. More generally, .N x̋ M /0 D N 0 x̋ M 0 [149, IV, Theorem 5.9].
The tensor product of von Neumann algebras is functorial with respect to normal
-homomorphisms: For each pair of normal -homomorphisms i W Mi ! Ni of
von Neumann algebras, where i D 1; 2, there exists a normal -homomorphism
1 x̋ 2 W M1 x̋ M2 ! N1 x̋ N2 such that .1 x̋ 2 /.x1 x̋ x2 / D 1 .x1 / x̋ 2 .x2 /
for all x1 2 M1 and x2 2 M2 .
For each  2 N and 2 M , there exists a unique functional  x̋ 2
.N x̋ M / such that . x̋ /.a x̋ b/ D .a/ .b/ for all a 2 N , b 2 M , and
k x̋ k  kkk k [149, Section IV.5]. If  and are positive or states, then so
is  x̋ .

Abstract von Neumann algebras. A C  -algebra A is called a W  -algebra if it

is isomorphic as a Banach space to the dual of some other Banach space A . In
this case, A is uniquely determined (up to isomorphism) and called the predual
of A. Every von Neumann algebra is a W  -algebra, where the predual is the space
of all normal linear functionals. Conversely, for every W  -algebra, there exists a
-isomorphism onto a von Neumann algebra that is continuous with respect to the
weak--topology.
In particular, for every von Neumann algebra N , the opposite algebra N op is a

W -algebra. Somewhat sloppily, we refer to N op as a von Neumann algebra again.

12.4 Slice maps

Slice maps are an important tool for the theory of quantum groups in the setting
of C  -algebras and von Neumann algebras. We shall describe in detail how these
slice maps are constructed and collect some frequently used formulas.

Slice maps in the setting of -algebras. Let A and B be unital -algebras with
linear maps  W A ! C, W B ! C. Then we can define linear slice maps

 ˇ id W A ˇ B ! B; aˇb!
7 .a/b;
(12.1)
id ˇ W A ˇ B ! B; aˇb!7 a .b/:

One easily verifies that for all a; a0 2 A, b; b 0 2 B and x 2 A ˇ B,

. ı . ˇ id//.x/ D . ˇ /.x/ D . ı .id ˇ //.x/ (12.2)

378 Chapter 12. Appendix

and
b.. ˇ id/.x//b 0 D . ˇ id/..1 ˇ b/x.1 ˇ b 0 //;
(12.3)
a..id ˇ /.x//a0 D .id ˇ /..a ˇ 1/x.a0 ˇ 1//:

If  and are -linear, so are  ˇ id and id ˇ . In general,

.. ˇ id/.x// D .N ˇ id/.x  /; where N W A ! C; a 7! .a /;

(12.4)
..id ˇ /.x// D .id ˇ N /.x  /; where N W B ! C; b 7! .b  /:

Given another -algebra C , we can define a linear slice map

 ˇ id ˇ W A ˇ C ˇ B ! C; a ˇ c ˇ b 7! .a/c .b/;

and for all x 2 A ˇ C , y 2 C ˇ B,

. ˇ id/.x/  .id ˇ /.y/ D . ˇ id ˇ /..x ˇ 1/.1 ˇ y//;

(12.5)
.id ˇ /.y/  . ˇ id/.x/ D . ˇ id ˇ /..1 ˇ y/.x ˇ 1//:

When we switch the order of the factors in the tensor product A ˇ C ˇ B, we

obtain new slice maps that satisfy analogues of equation (12.5).

Slice maps in the setting of C  -algebras. In the setting of C  -algebras, one

often uses slice maps that are defined on the minimal tensor product A ˝ B of
C  -algebras A and B and on the multiplier algebra M.A ˝ B/.

Proposition 12.4.1. Let A and B be C  -algebras and  2 A0 , 2 B 0 . Then

 ˇ id and id ˇ extend uniquely to linear maps

 ˝ id W M.A ˝ B/ ! M.B/; id ˝ W M.A ˝ B/ ! M.A/

that are norm-continuous and strictly continuous on bounded subsets. We have

k ˝ id k D kk and k id ˝ k D k k. If  or is positive, so is  ˝ id or id ˝ ,
respectively.

Remark 12.4.2. The algebraic tensor product A ˇ B is strictly dense in M.A˝B/

because it is norm-dense in A˝B and because A˝B is strictly dense in M.A˝B/.
Moreover, the unit ball of the algebraic tensor product A ˇ B is strictly dense in the
unit ball of M.A˝B/. Consequently,  ˝id and id ˝ are uniquely determined by
their restrictions to A ˇ B. Furthermore, it follows that the equations (12.2)–(12.5)
and many more formulas satisfied by  ˇ id and id ˇ extend to ˝id and id ˝ ,
respectively. Finally, we find . ˝ id/.A ˝ B/  B and .id ˝ /.A ˝ B/  A.
 12.4. Slice maps 379

Proof of Proposition 12.4.1. We only prove the assertions concerning  ˝ id; for
id ˝ , the arguments are similar. Uniqueness of  ˝ id was already observed
above. To prove existence, we consider two cases.
The case where  is a state: Let .H
; ƒ
;

/ be a GNS-construction for 
with cyclic vector 
, and let .HB ; ƒB ;
B / be the universal GNS-construction
for B. By definition of A ˝ B, there exists a non-degenerate -homomorphism

W A ˝ B ! L.H
˝ HB / such that
.a ˝ b/ D

.a/ ˝
B .b/ for all a 2 A,
b 2 B. Extend
to the multiplier algebra M.A ˝ B/ and consider the map

ˆ W M.A ˝ B/ ! L.HB /; T 7! h
jŒ1 ı
.T / ı j
iŒ1 ;

where j
iŒ1 W HB ! H
˝ HB is given by  7! 
˝ , and h
jŒ1 D j
iŒ1 .
We have ˆ.A ˝ B/ 
B .B/ because for all a 2 A, b 2 B,

ˆ.a ˝ b/ D h
j

.a/
i
B .b/ D .a/
B .b/:

Choose an approximate unit .u / for A. Then lim

.u /
D 
. If .T / is a
norm-bounded net in M.A ˝ B/ that converges strictly to some T 2 M.A ˝ B/,
and if b 2 B, then

lim ˆ.T /
B .b/ D lim h
jŒ1 ı
.T / ı j

.u /
iŒ1 ı
B .b/
 ;

D lim h
jŒ1 ı
.T /.

.u / ˝
B .b// ı j
iŒ1 :
;

Using the relation

.T /.

.u / ˝
B .b// D
.T .u ˝ b// and the fact that
.T / converges strictly to T , we find

lim ˆ.T /
B .b/ D lim h
jŒ1 ı
.T /.

.u / ˝
B .b// ı j
iŒ1 D ˆ.T /
B .b/:


A similar argument shows that lim

B .b/ˆ.T / D
B .b/ˆ.T /. Approximating
any T 2 M.A ˝ B/ strictly by a norm-bounded net in A ˝ B, we find that
ˆ.M.A ˝ B// 
B .M.B// and that ˆ W M.A ˝ B/ !
B .M.B// is strictly
continuous on norm-bounded subsets.
Since
B W M.B/ !
B .M.B// is an isomorphism, we can define  ˝ id WD

B1 ı ˆ. By construction,  ˝ id is positive and has norm less than or equal to

k
k2 k
k D 1 D kk.
General case: Let  2 A0 be arbitrary. Then  is a linear combinations of states,
and all assertions except for the norm estimates follow easily from the special case
considered above. Using Proposition 12.1.3, we find

k ˝ id k D supfk! ı . ˝ id/k W ! 2 B 0  M.B/0 ; k!k  1g:

But for ! as above, k! ı . ˝ id/k D k ˝ !k  kkk!k D kk, and hence

k ˝ id k  kk. 
380 Chapter 12. Appendix

Proposition 12.4.3. Let A, B be C  -algebras, T 2 M.A ˝ B/,  2 A0 , 2 B 0.

i) If .A ˝ 1/T .A ˝ 1/  A ˝ B, then . ˝ id/.T / 2 B.
ii) If .1 ˝ B/T .1 ˝ B/  A ˝ B, then .id ˝ /.T / 2 A.
Proof. We only prove i), the proof of ii) is similar. By Proposition 12.1.1, we may
assume that  has the form !.a1  a2 /, where ! 2 A0 and a1 ; a2 2 A. But then
. ˝ id/.T / D .! ˝ id/..a1 ˝ 1/T .a2 ˝ 1// 2 .! ˝ id/.A ˝ B/  B. 

Slice maps in the setting of von Neumann algebras. In the setting of von Neu-
mann algebras, slice maps can be defined for normal functionals:
Proposition 12.4.4. Let A and B be von Neumann algebras and  2 A , 2
B . Then the maps  ˇ id and id ˇ extend uniquely to norm-continuous
normal linear maps  x̋ id W A x̋ B ! B and id x̋ W A x̋ B ! A. We have
k x̋ id k D kk and k id x̋ k D k k. If  or is positive, so is  x̋ id or
id x̋ , respectively.
Remark 12.4.5. Since A ˇ B is dense in A x̋ B with respect to the  -weak topol-
ogy, the extensions  x̋ id and id x̋ are uniquely determined by their restrictions
to A ˇ B. Moreover, it follows that the equations (12.2)–(12.5) and many more
formulas satisfied by  ˇ id and id ˇ extend to  ˝ id and id ˝ , respectively.
Proof of Proposition 12.4.4. We only prove the assertions concerning  x̋ id.
Uniqueness follows from the observation made above. We prove existence. Denote
by H and K the underlying Hilbert spaces of A and B, respectively. Since  is
normal, there exist sequences .n /n and .n /n of vectors in H such that
X X
C WD kn k2 < 1; C WD kn k2 < 1;
n n
X
.a/ D hn jan i for all a 2 A:
n

For each x 2 A x̋ B, the sesquilinear map x W K  K ! C given by

X
x .;  0 / WD hn ˝ jx.n ˝  0 /i
n

is bounded: using the relation jhn ˝ jx.n ˝  0 /ij  kn kkn kkxkkkk 0 k and
the Cauchy–Schwarz inequality, we find
kx .;  0 /k  C C  kxk  kk  k 0 k for all ;  0 2 H:
Hence we can define  x̋ id W A x̋ B ! L.K/ by
hj. x̋ id/.x/ 0 i WD x .;  0 / for all x 2 A x̋ B;
 12.5. Auxiliary results 381

and k x̋ id k  C C . Moreover,

k x̋ id k D supfk! ı . x̋ id/k W ! 2 B ; k!k  1g;

and for ! as above, k! ı . x̋ id/k D k x̋ !k  kkk!k D kk. Hence,

k x̋ id k  kk.
We claim that  x̋ id is normal. Let P2 B . Then there exist sequences
P .n /n
and .n0 /n of vectors
P in K such that C  WD m km k 2
< 1, C 0 WD
m k 0
m k <1
0
and .b/ D m hm jbm i for all b 2 B. The relations
X X
0 2
kn ˝ m k2  C C < 1; kn ˝ m k  C C 0 < 1;
n;m n;m
X
0
. ı . x̋ id//.x/ D hn ˝ m jx.n ˝ m /i for all x 2 A x̋ B
n;m

show that the composition ı . x̋ id/ is normal. Since 2 B was arbitrary,

the claim follows.
Finally, a similar construction as in the proof of Proposition 12.4.1 shows that
 x̋ id is positive if  is a normal state, and consequently,  x̋ id is positive when-
ever  is positive. 

12.5 Auxiliary results

Proposition 12.5.1. Let
W A ! B be a bounded homomorphism of Banach
algebras. Assume that

i)
is non-degenerate in the sense that span
.A/B D B D span B
.A/;

ii) A has a bounded approximate unit, that is, a norm-bounded net .u / of
elements in A such that lim u a D a D lim au for all a 2 A.

Then
extends uniquely to a homomorphism
Q W M.A/ ! M.B/ such that

Q /
.a/ D
.T a/ and
.a/
.T

.T Q / D
.aT / (12.6)

for all T 2 M.A/, a 2 A.

Proof. Let T 2 M.A/ and put v WD

.T u / for all . Then the net .v / is
norm-bounded because kv k  k
kkT kku k for all , and

lim v
.a/ D lim
.T .u a// D
.T a/;
 
lim
.a/v D lim
..aT /u / D
.aT /
 
382 Chapter 12. Appendix

for all a 2 A. Since .v / is norm-bounded and span

.A/B D B D span B
.A/,
it follows that for every b 2 B, the nets .v b/ and .bv / converge in norm.
Clearly, the assignments

Q /b WD lim v b

.T and Q / WD lim bv ; b 2 B;
b
.T
 

Q / 2 M.B/ of norm
define a multiplier
.T

Q /k  sup kv k  k
kkT k sup ku k:
k
.T
 

The map T 7!
.T
Q / is obviously linear and satisfies condition (12.6). Furthermore,
by assumption i), the relation

Q /
.T

.S Q /
.a/ D
.S/
.T
Q a/ D
.S T a/ D
.S
Q T /
.a/; a 2 A;

Q
implies that
.S/ Q /b D
.ST

.T Q /b for all S; T 2 M.A/ and b 2 B, and similarly
Q /
.T
b
.S Q / D b
.ST
Q /. Therefore,
Q is a homomorphism. 
Corollary 12.5.2. Let
W A ! B be a bounded -homomorphism of Banach
-algebras that satisfies conditions i) and ii) of Proposition 12.5.1. Then
extends
to uniquely to a -homomorphism
Q W M.A/ ! M.B/ such that (12.6) holds.
Proof. The fact that the extension
Q provided by the previous proposition is a
-homomorphism follows easily from (12.6) and condition i). 
Lemma 12.5.3. Let V1 ; : : : ; Vn and W be topological vector spaces, and let
F W V1      Vn ! W be a multilinear map that is separately continuous in
each component. Suppose that for each i D 1; : : : ; n, we are given a closed sub-
space Yi  Vi , and a subset Xi  Yi that is linearly dense. Then

span fF .x1 ; : : : ; xn / j xi 2 Xi for all ig

D span fF .y1 ; : : : ; yn / j yi 2 Yi for all i g:

Proof. Replacing Xi by span Xi , we may assume each Xi is a subspace. Put

U WD span fF .x1 ; : : : ; xn / j xi 2 Xi for all i g:

We prove by induction that for every k D 0; : : : ; n,

F .y1 ; : : : ; yk ; xkC1 ; : : : ; xn / 2 U

for all yi 2 Yi , i D 1; : : : ; k, and xj 2 Xj , j D k C 1; : : : ; n. Note that for

k D n, this statement implies the claim. For k D 0, the assertion is trivially
satisfied. We assume that the statement holds for k and prove it for k C 1. Given
 12.5. Auxiliary results 383

yi 2 Yi ; i D 1; : : : ; k C 1, and xj 2 Xj ; j D k C 2; : : : ; n, we can approximate

ykC1 by some net .xkC1; / in XkC1 , and

F .y1 ; : : : ; yk ; xkC1; ; xkC2 ; : : : ; xn / 2 U for every 

by assumption. Since U is closed and F is continuous in the .k C 1/st variable, we

also have F .y1 ; : : : ; yk ; ykC1 ; xkC2 ; : : : ; xn / 2 U . 
Bibliography

[1] E. Abe, Hopf algebras. Cambridge Tracts in Math. 74, Cambridge University Press,
Cambridge 1980. xv, 3, 9, 38, 39, 47
[2] C. A. Akemann, G. K. Pedersen, and J. Tomiyama, Multipliers of C  -algebras.
J. Funct. Anal. 13 (1973), 277–301. 115, 259
[3] S. Baaj, Multiplicateurs non bornés. Thèse 3ème cycle, Université Paris VI, Paris
1980. 229
[4] S. Baaj, Représentation régulière du groupe quantique E .2/ de Woronowicz. C. R.
Acad. Sci. Paris Sér. I Math. 314 (1992), 1021–1026. 189, 191, 227, 228, 233, 235,
239
[5] S. Baaj, Représentation régulière du groupe quantique des déplacements de Worono-
wicz. Astérisque 232 (1995), 11–48. 189, 191, 196, 227, 233, 235, 236, 239, 240,
264
[6] S. Baaj and G. Skandalis, C  -algèbres de Hopf et théorie de Kasparov équivariante.
K-Theory 2 (1989), 683–721. 251, 263, 282
[7] S. Baaj and G. Skandalis, Unitaires multiplicatifs et dualité pour les produits croisés
de C  -algèbres. Ann. Sci. École Norm. Sup. (4) 26 (1993), 425–488. xvii, 97, 105,
166, 167, 170, 172, 189, 195, 227, 252, 283, 284
[8] S. Baaj and G. Skandalis, Unitaires multiplicatifs commutatifs. C. R. Acad. Sci. Paris
Sér. I Math. 336 (2003), 299–304. 170
[9] S. Baaj, G. Skandalis, and S. Vaes, Non-semi-regular quantum groups coming from
number theory. Comm. Math. Phys. 235 (2003), 139–167. 174, 191, 194, 227
[10] S. Baaj, G. Skandalis, and S. Vaes, Measurable Kac cohomology for bicrossed prod-
ucts. Trans. Amer. Math. Soc. 357 (2005), 1497–1524. 227
[11] S. Baaj and S. Vaes, Double crossed products of locally compact quantum groups.
J. Inst. Math. Jussieu 4 (2005), 135–173. 227
[12] T. Banica, Théorie des représentations du groupe quantique compact libre O.n/. C. R.
Acad. Sci. Paris Sér. I Math. 322 (1996), 241–244. 135, 159, 161, 162
[13] T. Banica, Le groupe quantique compact libre U.n/. Comm. Math. Phys. 190 (1997),
143–172. 135, 161, 164
[14] T. Banica, Symmetries of a generic coaction. Math. Ann. 314 (1999), 763–780. 135,
141
[15] T. Banica and B. Collins. Integration over quantum permutation groups. J. Funct.
Anal. 242 (2007), 641–657. 141
[16] B. Blackadar, Operator algebras. Encyclopaedia Math. Sci. 122, Springer-Verlag,
Berlin 2006. 373
386 Bibliography

[17] É. Blanchard, Déformations de C  -algèbres de Hopf. Bull. Soc. Math. France 124
(1996), 141–215. 144, 331
[18] R. J. Blattner, M. Cohen, and S. Montgomery, Crossed products and inner actions of
Hopf algebras. Trans. Amer. Math. Soc. 298 (1986), 671–711. 38
[19] A. Borel, Linear algebraic groups. 2nd ed., Grad. Texts in Math. 126, Springer-Verlag,
New York 1991. 9
[20] N. Bourbaki, Éléments de mathématique. Fasc. XXXII. Théories spectrales. Chap. 1
et 2: Algèbres normées. Groupes localement compacts commutatifs, Hermann, Paris
1967. 3
[21] O. Bratteli and D. W. Robinson, Operator algebras and quantum statistical mechan-
ics. Vol. 1: C  - and W  -algebras, symmetry groups, decomposition of states, Texts
Monogr. Phys., Springer-Verlag, New York 1979. 207
[22] T. Bröcker and T. tom Dieck, Representations of compact Lie groups. Grad. Texts in
Math. 98, Springer-Verlag, New York 1995. 4, 8, 33, 35, 81, 83, 125
[23] K. A. Brown and K. R. Goodearl, Lectures on algebraic quantum groups. Adv. Courses
Math. CRM Barcelona, Birkhäuser, Basel 2002. xvi, 3, 11, 135
[24] V. Chari and A. Pressley. A guide to quantum groups. Cambridge University Press,
Cambridge 1994. xvi, 11, 135, 142
[25] C. Chevalley. Theory of Lie groups I. Princeton Math. Ser. 8, Princeton University
Press, Princeton, N. J., 1946. 4
[26] A. Connes, On the spatial theory of von Neumann algebras. J. Funct. Anal. 35 (1980),
153–164. 299
[27] A. Connes, Noncommutative geometry. Academic Press Inc., San Diego 1994. 291,
299
[28] J. B. Conway, A course in functional analysis. 2nd ed., Grad. Texts in Math. 96,
Springer-Verlag, New York 1990. 190, 197
[29] S. Dăscălescu, C. Năstăsescu, and Ş. Raianu, Hopf algebras, Monographs Textbooks
Pure Appl. Math. 235, Marcel Dekker, New York 2001. 3, 49, 52, 66
[30] M.-C. David, Paragroupe d’Adrian Ocneanu et algèbre de Kac. Pacific J. Math. 172
(1996), 331–363. 289
[31] J. De Cannière, Produit croisé d’une algèbre de Kac par un groupe localement compact.
Bull. Soc. Math. France 107 (1979), 337–372. 105
[32] J. Dieudonné, Treatise on analysis. Vol. III, Pure Appl. Math. 10-III, Academic Press,
New York 1972. 35
[33] J. Dieudonné, Treatise on analysis. Vol. VI, Pure Appl. Math. 10-VI, Academic Press,
New York 1978. 3
[34] J. Dixmier, Les C  -algèbres et leurs représentations. Cahiers Scientifiques, Fasc.
XXIX, Gauthier-Villars & Cie , Paris 1964. 332, 369
 Bibliography 387

[35] J. Dixmier, Les algèbres d’opérateurs dans l’espace hilbertien (algèbres de von Neu-
mann). 2nd ed., Cahiers Scientifiques, Fasc. XXV, Gauthier-Villars, Paris 1969. 293,
294, 295, 309, 310, 375
[36] R. S. Doran and J. Wichmann, Approximate identities and factorization in Banach
modules. Lecture Notes in Math. 768, Springer-Verlag, Berlin 1979. 372, 373
[37] V. G. Drinfeld, Quantum groups. In Proc. Internat. Congr. Math. (Berkeley, Calif.,
1986 ), Vol. 1, Amer. Math. Soc., Providence, RI, 1987, 798–820. xvi, 135, 227
[38] M. J. Dupré and R. M. Gillette, Banach bundles, Banach modules and automorphisms
of C -algebras. Pitman Res. Notes Math. Ser. 92, Pitman Advanced Publishing Pro-
gram, Boston 1983. 332
[39] S. Echterhoff, S. Kaliszewski, J. Quigg, and I. Raeburn, A categorical approach to im-
primitivity theorems for C  -dynamical systems. Mem. Amer. Math. Soc. 180 (2006),
no. 850. 375
[40] E. G. Effros and Z.-J. Ruan, Discrete quantum groups I: The Haar measure. Internat.
J. Math. 5 (1994), 681–723. 65, 66, 91
[41] M. Enock, Produit croisé d’une algèbre de von Neumann par une algèbre de Kac.
J. Funct. Anal. 26 (1977), 16–47. 252
[42] M. Enock, Quantum groupoids and pseudo-multiplicative unitaries. In Locally com-
pact quantum groups and groupoids (Strasbourg, 2002), IRMA Lect. Math. Theor.
Phys. 2, Walter de Gruyter, Berlin 2003, 17–47. 313
[43] M. Enock. Inclusions of von Neumann algebras and quantum groupoïds III. J. Funct.
Anal. 223 (2005), 311–364. 289, 313
[44] M. Enock, Quantum groupoids of compact type. J. Inst. Math. Jussieu 4 (2005),
29–133. 189, 194, 313, 321, 322, 323, 327
[45] M. Enock and R. Nest, Irreducible inclusions of factors, multiplicative unitaries, and
Kac algebras. J. Funct. Anal. 137 (1996), 466–543. 252, 289
[46] M. Enock and J.-M. Schwartz, Produit croisé d’une algèbre de von Neumann par une
algèbre de Kac, II. Publ. Res. Inst. Math. Sci. 16 (1980), 189–232. 252
[47] M. Enock and J.-M. Schwartz, Kac algebras and duality of locally compact groups.
Springer-Verlag, Berlin 1992. xvii, 4, 105
[48] M. Enock and L. Vainerman, Deformation of a Kac algebra by an abelian subgroup.
Comm. Math. Phys. 178 (1996), 571–596. 105
[49] M. Enock and J.-M. Vallin, C -algèbres de Kac et algèbres de Kac. Proc. London
Math. Soc. (3) 66 (1993), 619–650. xvii, 105
[50] M. Enock and J.-M. Vallin, Inclusions of von Neumann algebras, and quantum
groupoïds. J. Funct. Anal. 172 (2000), 249–300. 289, 313
[51] M. Enock and J.-M. Vallin, Inclusions of von Neumann algebras and quantum
groupoïds II. J. Funct. Anal. 178 (2000), 156–225. 289, 313
[52] T. Falcone, L2 -von Neumann modules, their relative tensor products and the spatial
derivative. Illinois J. Math. 44 (2000), 407–437. 291
388 Bibliography

[53] J. M. G. Fell and R. S. Doran, Representations of  -algebras, locally compact groups,

and Banach  -algebraic bundles. Vol. 1: Basic representation theory of groups and
algebras, Pure Appl. Math. 125, Academic Press, Boston, MA, 1988. 261
[54] J. M. G. Fell and R. S. Doran, Representations of  -algebras, locally compact groups,
and Banach  -algebraic bundles. Vol. 2: Banach  -algebraic bundles, induced rep-
resentations, and the generalized Mackey analysis, Pure Appl. Math. 126, Academic
Press, Boston 1988. 261
[55] W. Fulton and J. Harris, Representation theory. Grad. Texts in Math. 129, Springer-
Verlag, New York 1991. 149
[56] V. V. Gorbatsevich, A. L. Onishchik, and E. B. Vinberg, Foundations of Lie theory and
Lie transformation groups. Springer-Verlag, Berlin 1997. 11
[57] J. M. Gracia-Bondía, J. C. Várilly, and H. Figueroa, Elements of noncommutative
geometry. Birkhäuser Advanced Texts, Birkhäuser, Boston 2001. 332, 373
[58] U. Haagerup, The standard form of von Neumann algebras. Math. Scand. 37 (1975),
271–283. 293
[59] U. Haagerup, Operator valued weights in von Neumann algebras, I. J. Funct. Anal.
32 (1979), 175–206. 290
[60] U. Haagerup, Operator valued weights in von Neumann algebras, II. J. Funct. Anal.
33 (1979), 339–361. 290
[61] B. C. Hall, Lie groups, Lie algebras, and representations. Grad. Texts in Math. 222,
Springer-Verlag, New York 2003. 149
[62] E. Hewitt and K. A. Ross, Abstract harmonic analysis. Vol. II: Structure and analysis
for compact groups. Analysis on locally compact abelian groups, Grundlehren Math.
Wiss. 152, Springer-Verlag, Berlin 1970. 4, 8, 81, 83, 125, 372, 373
[63] E. Hewitt and K. A. Ross, Abstract harmonic analysis. Vol. I: Structure of topological
groups, integration theory, group representations, 2nd. ed., Grundlehren Math. Wiss.
115, Springer-Verlag, Berlin 1979. 108, 140
[64] E. Hille, Analytic function theory. Vol. II, Ginn and Company, Boston 1962. 89
[65] G. P. Hochschild, Basic theory of algebraic groups and Lie algebras. Grad. Texts in
Math. 75, Springer-Verlag, New York 1981. 9, 11, 39
[66] H. Hopf, Über die Topologie der Gruppen-Mannigfaltigkeiten und ihre Verallge-
meinerungen. Ann. of Math. (2) 42 (1941), 22–52. xv
[67] A. Jacobs, The quantum E.2/ group. PhD thesis, Katholieke Universiteit Leuven,
Leuven 2005. Available at http://hdl.handle.net/1979/154 166, 203, 227, 228, 233,
236, 237, 239, 240, 241
[68] J. C. Jantzen, Lectures on quantum groups. Grad. Stud. Math. 6, Amer. Math. Soc.,
Providence, RI, 1996. xvi, 11, 135
[69] K. K. Jensen and K. Thomsen, Elements of KK-theory. Birkhäuser, Boston 1991. 113
[70] M. Jimbo, A q-difference analogue of U.g/ and the Yang-Baxter equation. Lett. Math.
Phys. 10 (1985), 63–69. 135
 Bibliography 389

[71] V. F. R. Jones, Index for subfactors. Invent. Math. 72 (1983), 1–25. 162
[72] V. Jones and V. S. Sunder, Introduction to subfactors, London Math. Soc. Lecture Note
Ser. 234, Cambridge University Press, Cambridge 1997. 289
[73] A. Joyal and R. Street, An introduction to Tannaka duality and quantum groups. In
Category theory (Como, 1990), Lecture Notes in Math. 1488, Springer-Verlag, Berlin
1991, 413–492. 4
[74] G. I. Kac, Extensions of groups to ring groups. Mat. Sb. (N.S.) 76 (1968), 473–496;
English transl. Math.-USSR Sb. 5 (1968), 451–474. 227
[75] R. V. Kadison and J. R. Ringrose, Fundamentals of the theory of operator algebras.
Vol. I: Elementary theory, Grad. Stud. Math. 15, Amer. Math. Soc., Providence, RI,
1997. 197, 200, 375
[76] R. V. Kadison and J. R. Ringrose, Fundamentals of the theory of operator algebras.
Vol. II: Advanced theory, Grad. Stud. Math. 16, Amer. Math. Soc., Providence, RI,
1997. 207, 375
[77] B.-J. Kahng, Non-compact quantum groups arising from Heisenberg type Lie bialge-
bras. J. Operator Theory 44 (2000), 303–334. 105
[78] B.-J. Kahng, Haar measure on a locally compact quantum group. J. Ramanujan Math.
Soc. 18 (2003), 384–414. 105
[79] C. Kassel, Quantum groups, Grad. Texts in Math. 155, Springer-Verlag, New York
1995. xvi, 3, 11, 33, 75, 78, 120, 121, 135, 142
[80] A. Klimyk and K. Schmüdgen, Quantum groups and their representations. Texts
Monogr. Phys., Springer-Verlag, Berlin 1997. xvi, 3, 11, 47, 52, 66, 107, 135, 142,
145
[81] E. Koelink and J. Kustermans, A locally compact quantum group analogue of the
normalizer of SU.1; 1/ in SL.2; C/. Comm. Math. Phys. 233 (2003), 231–296. 227,
228

[82] E. Koelink and J. Kustermans, Quantum SU.1; 1/ and its Pontryagin dual. In Locally
compact quantum groups and groupoids (Strasbourg, 2002), IRMA Lect. Math. Theor.
Phys. 2, Walter de Gruyter, Berlin 2003, 49–77. 227
[83] T. H. Koornwinder, Representations of the twisted SU.2/ quantum group and some
q-hypergeometric orthogonal polynomials. Nederl. Akad. Wetensch. Indag. Math. 51
(1989), 97–117. 142
[84] L. I. Korogodski and Y. S. Soibelman, Algebras of functions on quantum groups:
Part I. Math. Surveys Monogr. 56, Amer. Math. Soc., Providence, RI, 1998. xvi, 11,
135, 142
[85] A. Kumjian, Fell bundles over groupoids. Proc. Amer. Math. Soc. 126 (1998),
1115–1125. 368
[86] J. Kustermans, One-parameter representations on C -algebras. Technical report,
Odense Universitet, 1997. Available at http://arxiv.org/abs/funct-an/9707009 208
[87] J. Kustermans, Locally compact quantum groups in the universal setting. Internat.
J. Math. 12 (2001), 289–338. 203
390 Bibliography

[88] J. Kustermans, Locally compact quantum groups. In Quantum independent increment

processes I. Lecture Notes in Math. 1865, Springer-Verlag, Berlin 2005, 99–180. 97,
203
[89] J. Kustermans and L. Tuset, A survey of C  -algebraic quantum groups, Part II. Irish
Math. Soc. Bull. 44 (2000), 6–54. 203
[90] J. Kustermans and S. Vaes, Weight theory for C -algebraic quantum groups. Technical
report, Katholieke Universiteit Leuven and University College Cork, 1999. Available
at http://arxiv.org/abs/math/9901063v1 210, 214, 215
[91] J. Kustermans and S. Vaes, Locally compact quantum groups. Ann. Sci. École Norm.
Sup. (4) 33 (2000), 837–934. xvii, 4, 97, 106, 189, 203, 210, 211, 214, 215, 217, 218,
219, 220, 222, 223, 224, 225, 273, 289
[92] J. Kustermans and S. Vaes, The operator algebra approach to quantum groups. Proc.
Natl. Acad. Sci. USA 97 (2000), 547–552. 203
[93] J. Kustermans and S. Vaes, Locally compact quantum groups in the von Neumann
algebraic setting. Math. Scand. 92 (2003), 68–92. xvii, 106, 203, 211, 226, 227, 272,
273
[94] J. Kustermans and A. Van Daele, C  -algebraic quantum groups arising from algebraic
quantum groups. Internat. J. Math. 8 (1997), 1067–1139. 52, 183, 213
[95] E. C. Lance, Hilbert C  -modules. A toolkit for operator algebraists. London Math.
Soc. Lecture Note Ser. 210, Cambridge University Press, Cambridge 1995. 301, 342,
343, 373
[96] M. B. Landstad, Quantizations arising from abelian subgroups. Internat. J. Math. 5
(1994), 897–936. 105
[97] M. B. Landstad, J. Phillips, I. Raeburn, and C. E. Sutherland, Representations of
crossed products by coactions and principal bundles. Trans. Amer. Math. Soc. 299
(1987), 747–784. 263
[98] P.-Y. Le Gall, Théorie de Kasparov équivariante et groupoïdes I. K-Theory 16 (1999),
361–390. 368
[99] F. Lesieur, Groupoïdes quantiques mesurés: axiomatique, étude, dualité, exem-
ples. PhD thesis, Universite d’Orléans, Orléans 2003. Available at http://www.univ-
orleans.fr/mapmo/publications/lesieur/these.php 289, 314, 322, 333
[100] S. Levendorski and Y. Soibelman, Algebras of functions on compact quantum groups,
Schubert cells and quantum tori. Comm. Math. Phys. 139 (1991), 141–170. 135
[101] R. Longo, A duality for Hopf algebras and for subfactors. I. Comm. Math. Phys. 159
(1994), 133–150. 289
[102] R. Longo and J. E. Roberts, A theory of dimension. K-Theory, 11 (1997), 103–159.
120
[103] G. Lusztig, Introduction to quantum groups. Progr. Math. 110, Birkhäuser, Boston
1993. xvi, 11, 135
[104] S. Mac Lane, Categories for the working mathematician. 2nd ed., Grad. Texts in
Math. 5, Springer-Verlag, New York 1998. 75, 120
 Bibliography 391

[105] A. Maes and A. Van Daele, Notes on compact quantum groups. Nieuw Arch. Wisk. (4)
16 (1998), 73–112. 107, 108, 140
[106] S. Majid, Hopf–von Neumann algebra bicrossproducts, Kac algebra bicrossproducts,
and the classical Yang–Baxter equations. J. Funct. Anal. 95 (1991), 291–319. 105
[107] S. Majid, Foundations of quantum group theory. Cambridge University Press, Cam-
bridge 1995. 227
[108] T. Masuda, K. Mimachi, Y. Nakagami, M. Noumi, and K. Ueno, Representation of
quantum groups. In Mappings of operator algebras (Philadelphia, PA, 1988), Progr.
Math. 84, Birkhäuser, Boston 1990, 119–128. 142
[109] T. Masuda, K. Mimachi, Y. Nakagami, M. Noumi, and K. Ueno, Representations of
the quantum group SUq .2/ and the little q-Jacobi polynomials. J. Funct. Anal. 99
(1991), 357–386. 142
[110] T. Masuda, Y. Nakagami, and S. L. Woronowicz, A C  -algebraic framework for
quantum groups. Internat. J. Math. 14 (2003), 903–1001. xvii, 97, 106, 189, 227
[111] S. Montgomery, Hopf algebras and their actions on rings. CBMS Regional Conf. Ser.
in Math. 82, Amer. Math. Soc., Providence, RI, 1993. 3, 47
[112] S. A. Morris, Pontryagin duality and the structure of locally compact abelian groups.
29 London Math. Soc. Lecture Note Ser. 29, Cambridge University Press, Cambridge
1977. 3
[113] G. J. Murphy, C  -algebras and operator theory. Academic Press, Boston 1990. 200,
346, 369
[114] A. Ocneanu, Quantized groups, string algebras and Galois theory for algebras. In
Operator algebras and applications, Vol. 2: Mathematical physics and subfactors,
London Math. Soc. Lecture Note Ser. 136, Cambridge University Press, Cambridge
1988, 119–172. 289
[115] A. Ocneanu, Operator algebras, topology and subgroups of quantum symmetry – con-
struction of subgroups of quantum groups. In Taniguchi Conference on Mathematics
Nara ’98, Adv. Stud. Pure Math. 31, Math. Soc. Japan, Tokyo 2001, 235–263. 289
[116] M. O’uchi, Pseudo-multiplicative unitaries on Hilbert C  -modules. Far East J. Math.
Sci. (FJMS) 2001, Special Volume, Part II, 229–249. 331, 332
[117] A. Pal, On some quantum groups and their representations. PhD thesis, In-
dian Statistical Institute, New Delhi 1995. Available at http://www.isid.ac.in/~arup/
papers/thesis.pdf 227, 233
[118] A. Pal, Haar measure on Eq .2/. Pacific J. Math. 176 (1996), 217–233. 227, 233
[119] A. L. T. Paterson, Amenability. Math. Surveys Monogr. 29, Amer. Math. Soc., Provi-
dence, RI, 1988. 102, 103, 111
[120] A. L. T. Paterson, Groupoids, inverse semigroups, and their operator algebras. Progr.
Math. 170, Birkhäuser, Boston 1999. 323
[121] G. K. Pedersen, C  -algebras and their automorphism groups. London Math. Soc.
Monogr. 14, Academic Press, London 1979. 102, 111, 204, 206, 207, 208, 210, 252,
254, 255, 256, 369
392 Bibliography

[122] L. Pontrjagin, Topological groups. Princeton Math. Ser. 2, Princeton University Press,
Princeton 1939. 3
[123] W. Pusz, Quantum GL.2; C/ group as double group over “az C b” quantum group.
Rep. Math. Phys. 49 (2002), 113–122. 227
[124] W. Pusz and P. M. Sołtan, Functional form of unitary representations of the quantum
‘az C b’ group. Rep. Math. Phys. 52 (2003), 309–319. 227
[125] W. Pusz and S. L. Woronowicz, A quantum GL.2; C/ group at roots of unity. Rep.
Math. Phys. 47 (2001), 431–462. 227
[126] W. Pusz and S. L. Woronowicz, A new quantum deformation of ‘ax Cb’group. Comm.
Math. Phys. 259 (2005), 325–362. 227, 228
[127] W. Pusz and S. L. Woronowicz, On a quantum group of unitary operators. The quantum
‘az C b’ group. In Twenty years of Bialowieza: a mathematical anthology, World Sci.
Monogr. Ser. Math. 8, World Scientific Publishing, Singapore 2005, 229–261. 227,
228, 242, 244, 246, 247
[128] I. Raeburn and D. P. Williams, Morita equivalence and continuous-trace C  -algebras.
Math. Surveys Monogr. 60, Amer. Math. Soc., Providence, RI, 1998. 373
[129] J. Renault, A groupoid approach to C -algebras. Lecture Notes in Math. 793, Springer-
Verlag, Berlin 1980. 323, 324, 348
[130] M. A. Rieffel, Some solvable quantum groups. In Operator algebras and topology
(Craiova, 1989), Pitman Res. Notes Math. Ser. 270, Longman Sci. Tech., Harlow
1992, 146–159. 105
[131] M. A. Rieffel, Compact quantum groups associated with toral subgroups. In Repre-
sentation theory of groups and algebras, Contemp. Math. 145, Amer. Math. Soc.,
Providence, RI, 1993, 465–491. 135
[132] M. A. Rieffel, Non-compact quantum groups associated with abelian subgroups.
Comm. Math. Phys. 171 (1995), 181–201. 105
[133] M. Rosso, Comparaison des groupes SU.2/ quantiques de Drinfel’d et de Woronowicz.
C. R. Acad. Sci. Paris Sér. I Math. 304 (1987), 323–326. 135, 142
[134] M. Rosso, Algèbres enveloppantes quantifiées, groupes quantiques compacts de ma-
trices et calcul différentiel non commutatif. Duke Math. J. 61 (1990), 11–40. 135
[135] M. Rowicka, Exponential equations related to the quantum ‘ax C b’ group. Comm.
Math. Phys. 244 (2004), 419–453. 227
[136] W. Rudin, Functional analysis. 2nd ed., McGraw-Hill, New York 1991. 197
[137] S. Sakai, C  -algebras and W  -algebras. Ergeb. Math. Grenzgeb. 60, Springer-Ver-
lag, Berlin 1971. 293, 310, 369, 375
[138] J.-L. Sauvageot, Sur le produit tensoriel relatif d’espaces de Hilbert. J. Operator
Theory 9 (1983), 237–252. 291
[139] J.-L. Sauvageot, Produits tensoriels de Z-modules et applications. In Operator alge-
bras and their connections with topology and ergodic theory (Buşteni, 1983), Lecture
Notes in Math. 1132, Springer-Verlag, Berlin 1985, 468–485. 307
 Bibliography 393

[140] S. Shnider and S. Sternberg, Quantum groups: From coalgebras to Drinfeld algebras.
International Press, Cambridge, MA, 1993. xvi, 3, 11, 135
[141] P. M. Sołtan, New quantum “az C b” groups. Rev. Math. Phys. 17 (2005), 313–364.
227, 228, 242, 244, 245, 246, 247
[142] P. M. Sołtan and S. L. Woronowicz, A remark on manageable multiplicative unitaries.
Lett. Math. Phys. 57 (2001), 239–252. 189, 197, 199, 246
[143] T. A. Springer, Linear algebraic groups. 2nd ed., Progr. Math. 9, Birkhäuser, Boston
1998. 9
[144] Ş. Strătilă, Modular theory in operator algebras. Editura Academiei, Bucharest; Aba-
cus Press, Tunbridge Wells, Kent, 1981. 207, 290
[145] M. E. Sweedler, Hopf algebras. W. A. Benjamin, New York 1969. 3, 9, 47, 49, 52, 66
[146] W. Szymański, Finite index subfactors and Hopf algebra crossed products. Proc. Amer.
Math. Soc. 120 (1994), 519–528. 289
[147] M. Takesaki, Tomita’s theory of modular Hilbert algebras and its applications. Lecture
Notes in Math. 128, Springer-Verlag, Berlin 1970. 207, 215
[148] M. Takesaki, Duality and von Neumann algebras. In Lectures on operator algebras,
Tulane Univ. Ring and Operator Theory Year, 1970–1971, Vol. II, Lecture Notes in
Math. 247, Springer, Berlin 1972, 665–786. 206
[149] M. Takesaki, Theory of operator algebras I. Encyclopaedia Math. Sci. 124, Springer-
Verlag, Berlin 2002. 293, 294, 304, 310, 369, 372, 375, 377
[150] M. Takesaki, Theory of operator algebras II. Encyclopaedia Math. Sci. 125, Springer-
Verlag, Berlin 2003. 3, 207, 210, 211, 214, 215, 256, 290, 291, 293, 296, 298, 299,
306, 375
[151] T. Tannaka, Sôtai genri (Japanese) [The principle of duality]. Iwanami Shoten, Tokyo
1951. 4
[152] T. Timmermann, Pseudo-multiplicative unitaries and pseudo-Kac systems on C  -
modules. PhD thesis, Universität Münster, Münster 2005. Available at http://www.
math.uni-muenster.de/sfb/about/publ/heft394.ps 328, 348, 364, 365, 366, 367, 368
[153] T. Timmermann, Pseudo-multiplicative unitaries on C*-modules and Hopf C*-fami-
lies I. J. Noncommut. Geom. 1 (2007), 497–542. 328, 345, 355, 359, 360, 361
[154] T. Timmermann, C  -pseudo-multiplicative unitaries. SFB 478 - Geometrische Struk-
turen in der Mathematik, Münster, Technical Report No. 481, September 2007. Avail-
able at www.arxiv.org/abs/0709.2995. 328, 329, 342
[155] T. Timmermann, C  -pseudo-Kac systems and duality for coactions of concrete Hopf
C  -bimodules. SFB 478 - Geometrische Strukturen in der Mathematik, Münster,
Technical Report No. 484, October 2007. Available at www.arxiv.org/abs/0709.4617.
328, 342
[156] T. Timmermann, Finite-dimensional Hopf-C  -bimodules and C  -pseudo-multipli-
cative unitaries. SFB 478 - Geometrische Strukturen in der Mathematik, Münster,
November 2007. Available at www.arxiv.org/abs/0711.1420. 328, 342
394 Bibliography

[157] S. Vaes, Examples of locally compact quantum groups through the bicrossed prod-
uct construction. In XIIIth International congress on mathematical physics (London,
2000), International Press, Boston 2001, 341–348. 227
[158] S. Vaes, Locally compact quantum groups. PhD thesis, Katholieke Universiteit Leu-
ven, Leuven 2001. Available at http://wis.kuleuven.be/analyse/stefaan/PhD.html 97,
203, 206, 210, 211, 214, 215, 217, 218, 219, 220, 222, 223, 224, 225, 226, 227, 272,
273
[159] S. Vaes, The unitary implementation of a locally compact quantum group action.
J. Funct. Anal. 180 (2001), 426–480. 252
[160] S. Vaes, A new approach to induction and imprimitivity results. J. Funct. Anal. 229
(2005), 317–374. 252
[161] S. Vaes, Strictly outer actions of groups and quantum groups. J. Reine Angew. Math.
578 (2005), 147–184. 252
[162] S. Vaes and L. Vainerman, Extensions of locally compact quantum groups and the
bicrossed product construction. Adv. Math. 175 (2003), 1–101. 227
[163] S. Vaes and L. Vainerman, On low-dimensional locally compact quantum groups.
In Locally compact quantum groups and groupoids (Strasbourg, 2002), IRMA Lect.
Math. Theor. Phys. 2, Walter de Gruyter, Berlin 2003, 127–187. 252
[164] S. Vaes and A. Van Daele, Hopf C -algebras. Proc. London Math. Soc. (3) 82 (2001),
337–384. 106
[165] L. Vainerman, 2-cocycles and twisting of Kac algebras. Comm. Math. Phys. 191
(1998), 697–721. 105
[166] L. Vainerman, The bicrossed product construction for locally compact quantum
groups. Bull. Kerala Math. Assoc. 2005, Special issue, 99–136. 227
[167] L. Vainerman and G. I. Kac, Nonunimodular ring groups and Hopf-von Neumann
algebras. Dokl. Akad. Nauk SSSR 211 (1973), 1031–1034; English transl. Soviet.
Math. Dokl. 14 (1973), 1144–1148. xvii, 105
[168] L. Vainerman and G. I. Kac, Nonunimodular ring groups and Hopf-von Neumann
algebras. Mat. Sb. (N.S.) 94 (1974), 194–225; English transl. Math. USSR Sb. 23
(1974), 185–214. xvii, 105
[169] L. L. Vaksman andY. S. Soibelman, Algebra of functions on the quantum group SU.2/.
Funktsional. Anal. i Prilozhen. 22 (1988), 1–14; English transl. Funct. Anal. Appl. 22
(1988), 170–181. 135, 142
[170] J.-M. Vallin, C -algèbres de Hopf et C -algèbres de Kac. Proc. London Math. Soc.
(3) 50 (1985), 131–174. xvii, 105
[171] J.-M. Vallin, Bimodules de Hopf et poids opératoriels de Haar. J. Operator Theory 35
(1996), 39–65. 289, 290, 307, 313, 323
[172] J.-M. Vallin, Unitaire pseudo-multiplicatif associé à un groupoïde. Applications à la
moyennabilité. J. Operator Theory 44 (2000), 347–368. 289, 290, 314, 323
 Bibliography 395

[173] A. Van Daele, Quantum deformation of the Heisenberg group. In Current topics in op-
erator algebras (Nara, 1990), World Scientific Publishing, Singapore 1991, 314–325.
105
[174] A. Van Daele, Multiplier Hopf algebras. Trans. Amer. Math. Soc. 342 (1994), 917–932.
xvi, 22, 24, 40, 46
[175] A. Van Daele, The Haar measure on a compact quantum group. Proc. Amer. Math.
Soc. 123 (1995), 3125–3128. 109
[176] A. Van Daele, Discrete quantum groups, J. Algebra 180 (1996), 431–444. 91, 93
[177] A. Van Daele, An algebraic framework for group duality. Adv. Math. 140 (1998),
323–366. xvi, 40, 46, 47, 50, 52, 55, 57, 58, 60, 62, 91
[178] A. Van Daele, The Haar measure on some locally compact quantum groups. Technical
report, Katholieke Universiteit Leuven, Leuven 2001. Available at http://arxiv.org/
abs/math/0109004 227, 242, 244, 246, 247
[179] A. Van Daele and S. Wang, Universal quantum groups. Internat. J. Math. 7 (1996),
255–263. 135, 159
[180] A. Van Daele and S. L. Woronowicz, Duality for the quantum E.2/ group. Pacific
J. Math. 173 (1996), 375–385. 227, 233, 238, 239
[181] V. S. Varadarajan, Lie groups, Lie algebras, and their representations. Grad. Texts in
Math. 102, Springer-Verlag, New York 1984. 11, 33
[182] R. Vergnioux, KK-théorie équivariante et opérateur de Julg-Valette pour les
groupes quantiques. PhD thesis, Universite Paris VII, Paris 2002. Available at
http://www.institut.math.jussieu.fr/theses/2002/vergnioux/ 252, 267
[183] N. Ja. Vilenkin and A. U. Klimyk, Representation of Lie groups and special functions.
Volume 3: Classical and quantum groups and special functions, Math. Appl. 75 (Soviet
Series), Kluwer Academic Publishers, Dordrecht 1992. 142
[184] D. V. Voiculescu, K. J. Dykema, and A. Nica, Free random variables. CRM Monogr.
Ser. 1, Amer. Math. Soc., Providence, RI, 1992. 155
[185] S. Wang, General constructions of compact quantum groups. PhD thesis, University
of California, Berkeley 1993. 135, 154, 159
[186] S. Wang, Free products of compact quantum groups. Comm. Math. Phys. 167 (1995),
671–692. 135, 154, 159
[187] S. Wang, Tensor products and crossed products of compact quantum groups. Proc.
London Math. Soc. (3) 71 (1995), 695–720. 135, 154
[188] S. Wang, Deformations of compact quantum groups via Rieffel’s quantization. Comm.
Math. Phys. 178 (1996), 747–764. 135
[189] S. Wang, Quantum symmetry groups of finite spaces. Comm. Math. Phys. 195 (1998),
195–211. 135, 141
[190] A. Weil, L’intégration dans les groupes topologiques et ses applications. Hermann,
Paris 1940. 3, 225
396 Bibliography

[191] D. P. Williams, Crossed products of C  -algebras. Math. Surveys Monogr. 134, Amer.
Math. Soc., Providence, RI, 2007. 252, 254, 255
[192] S. L. Woronowicz. On the purification of factor states. Comm. Math. Phys. 28 (1972),
221–235. 89
[193] S. L. Woronowicz, Compact matrix pseudogroups. Comm. Math. Phys. 111 (1987),
613–665. xvii, 66, 97, 105, 107, 109, 136
[194] S. L. Woronowicz, Twisted SU.2/ group. An example of a noncommutative differential
calculus. Publ. Res. Inst. Math. Sci. 23 (1987), 117–181. 105, 135, 142, 145, 149,
150, 152
[195] S. L. Woronowicz, Tannaka-Krein duality for compact matrix pseudogroups. Twisted
SU.N / groups. Invent. Math. 93 (1988), 35–76. 120, 139, 162
[196] S. L. Woronowicz, Quantum E.2/ group and its Pontryagin dual. Lett. Math. Phys.
23 (1991), 251–263. 227, 233, 238, 239, 241
[197] S. L. Woronowicz, Unbounded elements affiliated with C  -algebras and noncompact
quantum groups. Comm. Math. Phys. 136 (1991), 399–432. 227, 228, 229, 230, 231,
233, 235
[198] S. L. Woronowicz, Operator equalities related to the quantum E.2/ group. Comm.
Math. Phys. 144 (1992), 417–428. 227, 239, 241, 242
[199] S. L. Woronowicz. Quantum SU.2/ and E.2/ groups. Contraction procedure. Comm.
Math. Phys. 149 (1992), 637–652. 233
[200] S. L. Woronowicz, C  -algebras generated by unbounded elements. Rev. Math. Phys.
7 (1995), 481–521. 231, 232
[201] S. L. Woronowicz, From multiplicative unitaries to quantum groups. Internat. J. Math.
7 (1996), 127–149. 105, 167, 189, 197, 199
[202] S. L. Woronowicz, Compact quantum groups. In Symétries quantiques (Les Houches,
1995), North-Holland, Amsterdam 1998, 845–884. xvii, 66, 97, 105, 107, 109
[203] S. L. Woronowicz, Quantum exponential function. Rev. Math. Phys. 12 (2000),
873–920. 227
[204] S. L. Woronowicz, Quantum “az C b” group on complex plane. Internat. J. Math. 12
(2001), 461–503. 227, 228, 242, 244, 245, 246, 247
[205] S. L. Woronowicz, Haar weights on some quantum groups. In Group 24: Physical and
mathematical aspects of symmetries, Proceedings of the 24th international colloquium
on group theoretical methods in physics, Paris, 15–20 July 2002, Conference Series
173, Institute of Physics Publishing, Bristol 2002, 763–772. 166, 203, 227, 228, 242,
244, 247
[206] S. L. Woronowicz and K. Napiórkowski, Operator theory in the C  -algebra frame-
work. Rep. Math. Phys. 31 (1992), 353–371. 231
[207] S. L. Woronowicz and S. Zakrzewski, Quantum ‘ax C b’ group. Rev. Math. Phys. 14
(2002), 797–828. 227, 228
[208] T. Yamanouchi, Bicrossproduct Kac algebras, bicrossproduct groups and von Neu-
mann algebras of Takesaki’s type. Math. Scand. 71 (1992), 252–260. 105
Symbol Index

hj bra-operator associated to an element , xvi

ji ket-operator associated to an element , xvi
.A; /cop coopposite of a coalgebra, bialgebra,
C  -bialgebra .A; /, 13, 16, 43, 98
ŒX closed linear span of a set X , xvi

Aop opposite of an algebra A, 16, 98

a.i/ Sweedler notation, 13, 43, 67
A predual of a von Neumann algebra A, 377
.A.w/ .V /; / right leg of a (pseudo)multiplicative unitary V ,
175–176, 321
AdU conjugation by the unitary U , 370
y
.Ay.w/ .V /; / left leg of a (pseudo)multiplicative unitary V ,
175–176, 321

C the set of complex numbers, xv

CD internal tensor product of C  -families, 353
y /, C Ìr A.V /
C Ìr A.V reduced crossed product of a coaction, 274, 279
C Ì˛;r G reduced crossed product of an action, 254
C.ı/ space of matrix elements of a corepresentation ı,
67
C.X/ space of matrix elements of a corepresentation
operator X , 74, 115
Cr .G/ reduced group C  -algebra of a group G, 102
C  .G/ full group C  -algebra of a group G, 100
op
D.H; / -bounded elements of a right Hilbert module H ,
296
D.K; / -bounded elements of a left Hilbert module K,
296
 comultiplication, 5, 12, 42, 44, 98
ı1  ı2 ; u1  u2 ; X1  X2 direct sum of corepresentation
(operators/matrices), 75–76, 120
ı1  ı2 ; u1  u2 ; X1  X2 tensor product of corepresentation
(operators/matrices), 77–78, 120–121
.n/ iterates of the comultiplication , 14, 44
y V ; V
 comultiplications on the legs of a (pseudo)multi-
plicative unitary V , 172, 317, 362
398 Symbol Index

Dom./ domain of definition of a map , xv

E  F , E  F flipped internal tensor product of C  -modules,

329
E ˝A F , E  F , E  F internal tensor product of C  -modules, 330, 375
counit, 5, 12, 45
O ; counits on the legs of a multiplicative unitary, 184
 unit map of an algebra, 5

f  g; f  a; a  f convolution product, 16–18, 99

G0 unit space of a groupoid, 323

y
G Pontrjagin dual of a locally compact
abelian group G, 3

jH i; hH j spaces of ket-bra operators, 192

H ˝K relative tensor product of Hilbert modules, 301
H
GNS-space for the weight , 215
H.E/ family of all homogeneous elements of a right
C  -bimodule, 342

id identity map, xv
id ˇ f; f ˇ id algebraic slice map, 377
id ˝;  ˝ id C  -algebraic slice map, 378
id x̋ ;  x̋ id von Neumann-algebraic slice map, 380
Im./ image of a map , xv

k some field, xv
KA .E; F / set of compact operators on C  -modules, 374
k.G/ algebra of all k-valued functions on a group G, 6
K.H1 ; H2 / set of compact linear operators on Hilbert spaces,
xvi
kG group algebra of a group G, 9

LA .E; F / set of adjointable operators on C  -modules, 374

L ./ left multiplication by the -bounded element ,
296
L.E; F / family of homogeneous operators on right
C  -bimodules, 339
L.G/ von Neumann algebra of a group G, 104
L.H1 ; H2 / set of bounded linear operators on Hilbert spaces, xv
 Symbol Index 399

ƒ
GNS-map for the weight , 215
op
ƒ op fixed GNS-map for the opposite weight , 292

m multiplication map of an algebra, 5

M.A/ multiplier algebra of an algebra A, 41
M.C/ multiplier C  -family of a C  -family C, 341
M1  M2 fiber product of von Neumann modules, 308
N
M
space of -integrable elements, 204

N the set of natural numbers, xv

N x̋ M tensor product of von Neumann algebras, 377
N
space of -square-integrable elements, 204

O.G/ coordinate algebra of a matrix group or affine algebraic group G,

8, 9
ˇ algebraic tensor product, 111

PAut.A/ inverse semigroup of partial automorphisms of A, 337

R the set of real numbers, xv

Rep.G/ algebra of representative functions on a group G, 7

S antipode, 5, 45, 219

Sy; S antipodes on the legs of a multiplicative unitary, 185, 199
span X linear span of the set X , xvi
span X closed linear span of a set X , xvi

T the set of complex numbers of modulus 1, xv

T ˜A affiliated element of a C  -algebra A, 229

Ux canonical generator of a group algebra, or multiplier of a group

C  -algebra, 10, 101

v.i/ Sweedler notation, 67

VŒij  leg notation, 169

V{ Vy auxiliary unitaries of a balanced multiplicative unitary V , 264

Z the set of integers, xv

Index

-algebra, -coalgebra, -bialgebra, tensor product, 16

-Hopf algebra, 26 unital, 15
-homomorphism, 26, 369 bounded element of a Hilbert module,
covariant, 253, 257 296
-linear map, 26
C  -(semi)norm, 369
.A; /-C  -algebra, 257 C  -algebra, 369
action, 259 C  -algebraic compact quantum group,
dual, 255, 281 107, 127
strongly continuous, 253 amenable, 134
affiliated element of a C  -algebra, 229 reduced, 110, 132–134, 211
algebraic quantum group, 52 universal, 130–134
compact, 65, 86, 128–134 C  -algebraic quantum group
discrete, 91 compact, see C  -algebraic com-
dual, 63–64 pact quantum group
analytic extension, 208 full/universal, 104, 203
antipode reduced, see reduced
of a C  -bialgebra, 99 C  -algebraic quantum
of a Hopf -algebra, 27 group

of a Hopf algebra, 5, 17–22 C -bialgebra, 98
of a multiplier Hopf algebra, 46 coopposite, 98
of a reduced C  -algebraic direct sum, 98
quantum group, 219, 220 of functions on a group, 100, 101,
on the legs of a multiplicative 211
unitary, 184–187, 199–200 opposite, 98
approximate unit, 369 tensor product, 98
C  -family, 341
Baaj–Skandalis duality, 252, 286 non-degenerate, 341

balanced multiplicative unitary, 196, C -module, 373
264, 267, 282 full, 373
Banach algebra, 369 character of a corepresentation, 88
bialgebra, 15 closed map, 213
coopposite, 16 CMQG, see compact matrix quantum
counital, 15 group
direct sum, 16 coaction, 256
dual, 28 dual, 278, 280, 286
opposite, 16 left, 257
402 Index

regular, 257 finite-dimensional, 114

right, 256 intertwiner, 115
trivial, 257 irreducible, 67, 115
coalgebra, 12 matrix, 73, 116
cocommutative, 13 matrix elements, 67, 115
coopposite, 13 operator, 74, 114
counital, 12 regular, 68, 119, 120
direct sum, 13 tensor product, 77, 120
dual algebra, 28 unitary, 67, 112
tensor product, 13 counit
cocommutative coalgebra, 13 of a -coalgebra, 27
compact matrix quantum group, 136 of a C  -bialgebra, 99
compact quantum group, xvii of a coalgebra, 12, 13
C  -algebraic, see C  -algebraic of a Hopf algebra, 5
compact quantum group of a multiplier Hopf algebra, 46
algebraic, see algebraic quantum on the legs of a multiplicative
group, compact unitary, 184–187
compact matrix quantum group, covariant
136 -homomorphism, 253, 257
free product, 155 representation, 253
tensor product, 158 CQG, see compact quantum group
comultiplication crossed product
of a Hopf C  -family, 358 reduced of a coaction, 274, 279,
of a C  -bialgebra, 98 286
reduced of an action, 254, 280
of a bialgebra, 15
of a coalgebra, 12 decomposable
of a Hopf algebra, 5 groupoid, 348
of a von Neumann bialgebra, 98 right C  -bimodule, 342
on the legs of a (pseudo-)multi- densely defined map, 213
plicative unitary, 172, 317, discrete quantum group, see algebraic
362 quantum group, discrete
Connes’ fusion, see relative tensor dual
product of Hilbert modules action, 255, 281
convolution algebra, 17 algebra of a coalgebra, 28
convolution product, 16, 99 algebraic quantum group, 63–64
coproduct, see comultiplication coaction, 278, 280, 286
core, 214 coalgebra of a finite-dimensional
corepresentation, 67, 112 algebra, 28
character, 88 finite-dimensional bialgebra, 28
conjugate, 78, 121 finite-dimensional Hopf algebra,
direct sum, 75, 120 28, 32
 Index 403

Kac system, 285 G-C  -algebra, see action

locally compact abelian group, 3 GNS-construction
multiplier Hopf algebra, 58–64 for a state, 371
reduced C  -algebraic quantum for a weight, 215
group, 224, 225 group algebra, 9, 27, 30, 31, 38, 49
restricted dual of a unital group C  -algebra
algebra, bialgebra, full, 100
Hopf algebra, 36, 37 of a discrete group, 108, 110
weak Kac system, 268 reduced, 102, 211, 262
dual pairing of a discrete group, 108, 110
of Hopf algebras, 32, 38 group representation, 68, 101, 115
of the legs of a multiplicative group von Neumann algebra, 104, 206
unitary, 184 groupoid, 323
duality decomposable, 348
Baaj–Skandalis duality, 286
of algebraic quantum groups, Haar state
62–64 of a C  -algebraic compact
of finite-dimensional bialgebras quantum group, 110
and Hopf algebras, 28 of an algebraic compact
of locally compact abelian groups, quantum group, 65
3 Haar system
of multiplier Hopf algebras, left, 323
58–64
right, 323
of reduced C  -algebraic
Haar weight, 206, 211
quantum groups, 225
Hilbert module, 291
Pontrjagin duality, 3, 31, 225
standard, 292
Takesaki–Takai duality, 256
homogeneous
enveloping C  -algebra, 371 element of a right C  -bimodule,
342
Fell bundle, 261 operator on a right
fiber product of von Neumann C  -bimodule, 337
modules, 308 subset of a groupoid, 348
flipped internal tensor product, 329 Hopf C  -family, 358
free orthogonal quantum group, 160 Hopf algebra, xv, 5, 23
free product of compact quantum groups, dual, 28
155 of functions on a finite group, 6,
free unitary quantum group, 160 26, 30
full C  -algebraic quantum group, 203 of polynomial functions on a
full C  -module, 373 matrix group, 8
functional of polynomial functions on an
left-/right-invariant, 48 algebraic group, 9, 38
404 Index

of representative functions of a of an algebraic compact

group, 7, 27, 33, 38, 49, 68 quantum group, 89
Hopf C  -algebra, 106 modular automorphism group of
Hopf–von Neumann bimodule, 313 a weight, 209, 210
integral, 48 modular conjugation, 216
faithful, 48–50 modular element
normalized, 48, 49, 51 of a multiplier Hopf algebra, 55
positive, 48, 49, 52 of a reduced C  -algebraic
internal tensor product quantum group, 222
of (C  -)families, 353 modular function of a group, 54
of homogeneous operators, 351 modular operator, 216
intertwiner, 115 morphism
Kac algebras, xvii, 105 of -algebras, -coalgebras,
Kac system, 282, 286 -bialgebras,
of a compact quantum group, 269, Hopf--algebras, 26
285 of C  -bialgebras, 98
of a group, 268, 284 of bialgebras, 15
dual, opposite, predual, 285 of coalgebras, 13
weak, see weak Kac system of Hilbert modules, 291
ket-bra notation, xx of Hopf algebras, 6, 22
KMS-weight, 210, 216 of multiplier bialgebras, 43
left-invariant of multiplier Hopf algebras, 45
functional on a multiplier Hopf of von Neumann modules, 307
algebra, 48 multiplicative unitary, xvii, 169, 267
weight on a C  -bialgebra, 210 antipode on the legs, 184–187,
weight on a von Neumann 199–200
bialgebra, 205 balanced, 196, 264, 267, 282
leg notation, 168, 192 commutative, cocommutative,
legs 170
of a decomposable pseudo-multi- counit on the legs, 184–187
plicative unitary, 359 dual pairing of the legs, 184
of a multiplicative unitary, irreducible, 283
172–176 legs, 172–176
of a pseudo-multiplicative manageable, 198
unitary, 316–322 modular, 198
linearly dense, xx of a C  -algebraic compact
locally compact quantum group, 206 quantum group, 171–172
matrix elements, 67, 115 of a compact quantum group, 269
modular automorphism of a group, 170, 179–181, 187,
of a multiplier Hopf algebra, 57 190, 268
 Index 405

of a reduced C  -algebraic decomposably regular, 364

quantum group, 217, 222 legs, 359
of an algebraic quantum group, of a groupoid, 332–333, 365
170–171, 181–183, 188, 190 on a Hilbert space, 316
opposite, 169 (weakly) regular, 322
regular, 189, 195 legs, 316–322
semi-regular, 189, 194, 196 of a groupoid, 326, 333
weakly regular, 189, 194 weakly well-behaved, 321, 323
weakly well-behaved, 176, 194
well-behaved, 175, 195, 196, 199
multiplier q-deformation, xvi, 11, 135
of an algebra, 41–42 quantum group
multiplier C  -algebra, 370 C  -algebraic compact,
multiplier bialgebra, 42 see C  -algebraic compact
multiplier family, 341 quantum group
multiplier Hopf algebra, 44 algebraic, see algebraic quantum
discrete, 93 group
dual, 58–64 algebraic compact, see algebraic
of functions on a discrete group, quantum group, compact
45, 49 discrete, see algebraic quantum
regular, 45, 46 group, discrete
unimodular, 48 locally compact quantum group
with integrals, 48 in the setting of von Neu-
mann algebras, 206
non-degenerate reduced C  -algebraic,
-homomorphism of C  -algebras, see reduced C  -algebraic
370 quantum group
C  -family, 341 quantum groupoid, 290
algebra, 41
morphism of algebras, 41 reduced C  -algebraic quantum group,
203, 210
one-parameter group, 207
antipode, 219
analytic extension, 208
dual, 224
partial automorphism, 337 modular element, 222
compatibility, 351 multiplicative unitary, 217, 222
pentagon equation, 169, 316, 331 scaling group, 220
Plancherel theorem, 63, 223 unitary antipode, 220
Pontrjagin duality, 3, 31, 225 reduced crossed product
pre-C  -module, 373 of a coaction, 274, 279
pseudo-multiplicative unitary of an action, 254, 280
on a C  -module, 331 regular corepresentation, 68, 119, 120
406 Index

regular multiplicative unitary, on a C  -algebra, 371

see multiplicative unitary, strict topology, 370, 374
regular Sweedler notation, 13, 43
relative tensor product for comodules or corepresenta-
of Hilbert modules, 301 tions, 67
commutative case, 294–295, 301
of von Neumann modules, 308 Takesaki–Takai duality, 251, 256
representation, 370, 374 Tomita–Takesaki theory, 207–209,
covariant, 253 215–216
representative functions of a group,
see Hopf algebra unit map, 5
restricted dual, 36 unitary antipode
right C  -bimodule, 337 of a reduced C  -algebraic
right-invariant quantum group, 220
functional on a multiplier Hopf on the legs of a multiplicative
algebra, 48 unitary, 199–200
weight on a C  -bialgebra, 210 universal C  -algebraic quantum
weight on a von Neumann group, 203
bialgebra, 205 universal enveloping algebra, xvi, 11,
33, 38, 135
scaling constant of a reduced
C  -algebraic quantum group, von Neumann bialgebra, 98
221 of functions on a group, 104, 206
scaling group von Neumann module, 307
of a reduced C  -algebraic
quantum group, 220 weak Kac system, 267, 274, 279, 283
on the legs of a multiplicative of a compact quantum group,
unitary, 199–200 269, 285
Schur’s lemma, 81, 124 of a group, 268, 280, 284
Schur’s orthogonality relations, of a locally compact quantum
84–86, 125–127 group, 272
semi-regular multiplicative unitary, predual, dual, opposite, 268
see multiplicative unitary, weakly regular multiplicative unitary,
semi-regular 189
Sigma notation, see Sweedler weight, 204
notation densely defined,
slice map, 377 lower semi-continuous,
on a C  -algebra, 378 proper,
on a von Neumann algebra, 380 extension to multiplier algebra,
with respect to a weight, 214 214
state faithful, 204
on a -algebra, 48 GNS-construction, 215
 Index 407

Haar weight, 206, 211 right-invariant, 205, 210

KMS-weight, 210, 216 slice map, 214
left-invariant, 205, 210 tensor product, 214
normal, semi-finite, n.s.f., 205,
214