Vous êtes sur la page 1sur 195

Representations of SL2 (Fq )

Algebra and Applications


Volume 13

Managing Editor:

Alain Verschoren
University of Antwerp, Belgium

Series Editors:

Alice Fialowski
Etvs Lornd University, Hungary
Eric Friedlander
Northwestern University, USA
John Greenlees
Sheffield University, UK
Gerhard Hiss
Aachen University, Germany
Ieke Moerdijk
Utrecht University, The Netherlands
Idun Reiten
Norwegian University of Science and Technology, Norway
Christoph Schweigert
Hamburg University, Germany
Mina Teicher
Bar-llan University, Israel

Algebra and Applications aims to publish well written and carefully refereed mono-
graphs with up-to-date information about progress in all fields of algebra, its clas-
sical impact on commutative and noncommutative algebraic and differential geom-
etry, K-theory and algebraic topology, as well as applications in related domains,
such as number theory, homotopy and (co)homology theory, physics and discrete
mathematics.

Particular emphasis will be put on state-of-the-art topics such as rings of differential


operators, Lie algebras and super-algebras, group rings and algebras, C -algebras,
Kac-Moody theory, arithmetic algebraic geometry, Hopf algebras and quantum
groups, as well as their applications. In addition, Algebra and Applications will also
publish monographs dedicated to computational aspects of these topics as well as
algebraic and geometric methods in computer science.
Cdric Bonnaf

Representations
of SL2(Fq )
Cdric Bonnaf
CNRS (UMR 5149)
Universit Montpellier 2
Institut de Mathmatiques et de Modlisation de Montpellier
Place Eugne Bataillon
34095 Montpellier Cedex
France
cedric.bonnafe@math.univ-montp2.fr

ISBN 978-0-85729-156-1 e-ISBN 978-0-85729-157-8


DOI 10.1007/978-0-85729-157-8
Springer London Dordrecht Heidelberg New York

British Library Cataloguing in Publication Data


A catalogue record for this book is available from the British Library

Library of Congress Control Number: 2010938218

Mathematics Subject Classification (2010): 20, 20-02

Springer-Verlag London Limited 2011


Apart from any fair dealing for the purposes of research or private study, or criticism or review, as per-
mitted under the Copyright, Designs and Patents Act 1988, this publication may only be reproduced,
stored or transmitted, in any form or by any means, with the prior permission in writing of the publish-
ers, or in the case of reprographic reproduction in accordance with the terms of licenses issued by the
Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to
the publishers.
The use of registered names, trademarks, etc., in this publication does not imply, even in the absence of a
specific statement, that such names are exempt from the relevant laws and regulations and therefore free
for general use.
The publisher makes no representation, express or implied, with regard to the accuracy of the information
contained in this book and cannot accept any legal responsibility or liability for any errors or omissions
that may be made.

Cover design: VTEX, Vilnius

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)


To my parents.
Preface

If our sole purpose were to calculate the character table of the finite group
G = SL2 (Fq ) (here, q is a power of a prime number p) by ad hoc methods, this
book would only amount to a few pages. Indeed, this problem was solved
independently by Jordan [Jor] and Schur [Sch] in 1907. The goal of this book
is rather to use the group G to give an introduction to the ordinary and
modular representation theory of finite reductive groups, and in particular
to Harish-Chandra and Deligne-Lusztig theories. It is addressed in particu-
lar to students who would like to delve into Deligne-Lusztig theory with a
concrete example at hand. The example of G = SL2 (Fq ) is sufficiently simple
to allow a complete description, and yet sufficiently rich to illustrate some
of the most delicate aspects of the theory.
There are a number of excellent texts on Deligne-Lusztig theory (see for
example Lusztig [Lu1], Carter [Carter], Digne-Michel [DiMi] for the theory
of ordinary characters and Cabanes-Enguehard [CaEn] for modular repre-
sentations). This book does not aim to offer a better approach, but rather to
complement the general theory with an illustrated example. We have tried
not to rely upon the above books and give full proofs, in the example of the
group G , of certain general theorems of Deligne-Lusztig theory (for example
the Mackay formula, character formulas, questions of cuspidality etc.). Al-
though it is not always straightforward, we have tried to give proofs which
reflect the spirit of the general theory, rather than giving ad hoc arguments.
We hope that this shows how general arguments of Deligne-Lusztig the-
ory may be made concrete in a particular case. At the end of the book we
have included a chapter offering a very succinct overview (without proof)
of Deligne-Lusztig theory in general, as well as making links to what has
already been seen (see Chapter 12).
Historically, the example of SL2 (Fq ) played a seminal role. In 1974, Drin-
feld (at age nineteen!) constructed a Langlands correspondence for GL2 (K),
where K is a global field of equal characteristic [Dri]. In the course of this
work, he remarks that the cuspidal characters of G = SL2 (Fq ) may be found
in the first -adic cohomology group (here,  is a prime number different

vii
viii Preface

from p) of the curve Y with equation xy q yx q = 1, on which G acts nat-


urally by linear changes of coordinates (we call Y the Drinfeld curve). This
example inspired Deligne and Lusztig (see their comments in [DeLu, page
117, lines 22-24]) who then, in their fundamental article [DeLu], established
the basis of what has come to be known as Deligne-Lusztig theory.
A large part of this book is concerned with unravelling Drinfelds ex-
ample. Our principal is to rather shamelessly make use of the fundamental
results of -adic cohomology (for which we provide an overview tailored to
our needs in Appendix A) to construct representations of G in characteristic
0 or . In order to efficiently use this machinery, we conduct a precise study
of the geometric properties of the action of G on the Drinfeld curve Y, with
particular attention being paid to the construction of quotients by various
finite groups.
Having completed this study we do not limit ourselves to character the-
ory. Indeed, a large part of this book is dedicated to the study of modular
representation theory, most notably via the study of Brous abelian defect
group conjecture [Bro]. This conjecture predicts the existence of an equiv-
alence of derived categories when the defect group is abelian. For the rep-
resentations of the group G in characteristic   {2, p}, the defect group is
cyclic, and such an equivalence can be obtained by entirely algebraic meth-
ods [Ric1], [Lin], [Rou2]. However, in order to stay true to the spirit of this
book, we show that it is possible, when  is odd and divides q + 1, to realise
this equivalence of derived categories using the complex of -adic cohomol-
ogy of the Drinfeld curve (this result is due to Rouquier [Rou1]).
For completeness we devote a chapter to the study of representations in
equal, or natural, characteristic. Here the Drinfeld curve ceases to be useful
to us. We give an algebraic construction of the simple modules by restric-
tion of rational representations of the group G = SL2 (Fq ), as may be done
for an arbitrary finite reductive group. Moreover, in this case the Sylow p-
subgroup is abelian, and it was shown by Okuyama [Oku1], [Oku2] (for the
principal block) and Yoshii [Yo] (for the nonprincipal block with full defect)
that Brous conjecture holds. Unfortunately, the proof is too involved to be
included in this book.

P REREQUISITES The reader should have a basic knowledge of the repre-


sentation theory of finite groups (as contained, for example, in [Ser] or [Isa]).
In the appendix we recall the basics of block theory. He or she should also
have a basic knowledge of algebraic geometry over an algebraically closed
field (knowledge of the first chapter of [Har], for example, is more than suf-
ficient). An overview of -adic cohomology is given in Appendix A, while
Appendix C contains some basic facts about reflection groups (this appendix
will only be used when we discuss some curiosities connected to the groups
SL2 (Fq ) for q {3, 5, 7}).
We have also added a number of sections and subsections (marked with
an asterisk) which contain illustrations, provided by G and Y, of related
Preface ix

but more geometric subjects (for example the Hurwitz formula, automor-
phisms of curves, Abyankhars conjecture, invariants of reflections groups).
These sections require a more sophisticated geometric background and are
not necessary for an understanding of the main body of this book.
Lastly, for results concerning derived categories, we will not need more
than is contained in the economical and efficient summary in Appendix A1
of the book of Cabanes and Enguehard [CaEn]. The sections and subsections
requiring some knowledge of derived categories are also marked with an
asterisk.

Besanon, France Cdric Bonnaf


April 2009
Acknowledgements

This book probably would have never seen the light of day had it not been
for the suggestion of Raphal Rouquier, who also offered constant encour-
agement whilst the project was underway. He also patiently answered nu-
merous questions concerning derived categories, and carefully read large
passages of this book. For all of this I would like to thank him warmly.
I would also like to thank Marc Cabanes for numerous fruitful conversa-
tions on the subject of this book.
This English version of the book is the translation, by Geordie Williamson,
of a French manuscript I had written. I want to thank him warmly for the
quality of his job, as well as for the hundreds of interesting remarks, com-
ments, critics, suggestions. It has been a great pleasure to work with him
during the translation process.

xi
Contents

Part I Preliminaries

1 Structure of SL2 (Fq ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3


1.1 Special Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.1 Bruhat Decomposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.2 The Non-Split Torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Distinguished Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3 Conjugacy Classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.1 Centralisers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.3.2 Parametrisation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Sylow Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 Sylow p-Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.4.2 Other Sylow Subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 The Geometry of the Drinfeld Curve . . . . . . . . . . . . . . . . . . . . . . . . . . . 15


2.1 Elementary Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Interesting Quotients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.1 Quotient by G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 Quotient by U . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 Quotient by q+1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3 Fixed Points under certain Frobenius Endomorphisms . . . . . . . 19
2.4 Compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Curiosities* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.5.1 Hurwitz Formula, Automorphisms* . . . . . . . . . . . . . . . . . 22
2.5.2 Abhyankars Conjecture (Raynauds Theorem)* . . . . . . 24

Part II Ordinary Characters

3 Harish-Chandra Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.1 Bimodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3.2 Harish-Chandra Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30

xiii
xiv Contents

3.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.2.2 Other Constructions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2.3 Mackey Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2.4 Restriction from GL2 (Fq ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34

4 Deligne-Lusztig Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1 Definition and First Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.1.2 The Character R (1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.1.3 Dimensions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.1.4 Cuspidality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.2 Mackey Formula . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.3 Parametrisation of Irr G . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Action of the Frobenius Endomorphism . . . . . . . . . . . . . . . . . . . . 46
4.4.1 Action on Hc1 (Y)e1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.2 Action on Hc1 (Y)e0 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.4.3 Action on Hc1 (Y)e Hc1 (Y)e 1 . . . . . . . . . . . . . . . . . . . . . 48

5 The Character Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51


5.1 Characters of Bimodules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.1.1 Calculation of Tr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.1.2 Calculation of Tr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.1.3 The Characters R( ) and R ( ) . . . . . . . . . . . . . . . . . . . . . . 53
5.2 Restriction to U . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.1 B-Invariant Characters of U . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.2 Restriction of Characters of G . . . . . . . . . . . . . . . . . . . . . . . 54
5.2.3 Values of . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.3 Character Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

Part III Modular Representations

6 More about Characters of G and of its Sylow Subgroups . . . . . . . . 63


6.1 Central Characters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2 Global McKay Conjecture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.2.1 Characters of N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.2.2 Characters of N  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.2.3 Characters of B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.2.4 Normalisers of Sylow 2-Subgroups . . . . . . . . . . . . . . . . . . 68
6.2.5 Verification of the Global McKay Conjecture . . . . . . . . . . 68

7 Unequal Characteristic: Generalities . . . . . . . . . . . . . . . . . . . . . . . . . . . 71


7.1 Blocks, Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1.1 Partition in -Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
7.1.2 Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
7.1.3 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Contents xv

7.2 Modular Harish-Chandra Induction . . . . . . . . . . . . . . . . . . . . . . . . 75


7.3 Deligne-Lusztig Induction* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

8 Unequal Characteristic: Equivalences of Categories . . . . . . . . . . . . . 85


8.1 Nilpotent Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.1.1 Harish-Chandra Induction . . . . . . . . . . . . . . . . . . . . . . . . . . 86
8.1.2 Deligne-Lusztig Induction* . . . . . . . . . . . . . . . . . . . . . . . . . . 87
8.2 Quasi-Isolated Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.2.1 Harish-Chandra Induction . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.2.2 Deligne-Lusztig Induction* . . . . . . . . . . . . . . . . . . . . . . . . . . 88
8.3 The Principal Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
8.3.1 The Case when  is Odd and Divides q 1 . . . . . . . . . . . . 89
8.3.2 The Case when  is Odd and Divides q + 1* . . . . . . . . . . . 90
8.3.3 The Case when  = 2* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
8.4 Alvis-Curtis Duality* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

9 Unequal Characteristic: Simple Modules, Decomposition


Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
9.1.1 Induction and Decomposition Matrices . . . . . . . . . . . . . . 97
9.1.2 Dimensions of Modules and Restriction to U . . . . . . . . . 98
9.2 Nilpotent Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
9.3 Quasi-Isolated Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
9.4 The Principal Block . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
9.4.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
9.4.2 The Case when  is Odd and Divides q 1 . . . . . . . . . . . . 103
9.4.3 The Case when  is Odd and Divides q + 1 . . . . . . . . . . . . 104
9.4.4 The Case when  = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

10 Equal Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109


10.1 Simple Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
10.1.1 Standard or Weyl Modules . . . . . . . . . . . . . . . . . . . . . . . . . . 110
10.1.2 Simple Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
10.1.3 The Grothendieck Ring of G . . . . . . . . . . . . . . . . . . . . . . . . . 117
10.2 Simple kG -Modules and Decomposition Matrices . . . . . . . . . . . 119
10.2.1 Simple kG -Modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
10.2.2 Decomposition Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
10.3 Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
10.3.1 Blocks and Brauer Correspondents . . . . . . . . . . . . . . . . . . 122
10.3.2 Brauer Trees* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123

Part IV Complements

11 Special Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129


11.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
11.2 The Case when q = 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
xvi Contents

11.2.1
Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
11.2.2
Character Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132
The Group SL2 (F3 ) as a Subgroup of SL2 (F ) . . . . . . . . . . 133
11.2.3
The Group SL2 (F3 ) as a Reflection Group of Rank 2 . . . 134
11.2.4
The Group PSL2 (F3 ) and the Isometries of the
11.2.5
Tetrahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
11.3 The Case when q = 5 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
11.3.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
11.3.2 Character Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
11.3.3 The Group SL2 (F5 ) as a Subgroup of SL2 (Fr ) . . . . . . . . . 137
11.3.4 The Group SL2 (F5 ) Z/5Z as a Reflection Group of
Rank 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
11.3.5 The Group PSL2 (F5 ), the Dodecahedron and the
Icosahedron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 141
11.4 The Case when q = 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142
11.4.1 Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
11.4.2 Character Table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
11.4.3 The Isomorphism Between the Groups PSL2 (F7 ) and
GL3 (F2 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
11.4.4 The Group PSL2 (F7 ) Z/2Z as a Reflection Group of
Rank 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145

12 Deligne-Lusztig Theory: an Overview* . . . . . . . . . . . . . . . . . . . . . . . . . 149


12.1 Deligne-Lusztig Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
12.2 Modular Representations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
12.2.1 Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
12.2.2 Modular Deligne-Lusztig Induction . . . . . . . . . . . . . . . . . . 155
12.2.3 The Geometric Version of Brous Conjecture . . . . . . . . . 155

Appendix A -Adic Cohomology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159


A.1 Properties of the Complex* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
A.2 Properties of the Cohomology Groups . . . . . . . . . . . . . . . . . . . . . . 160
A.2.1 General Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
A.2.2 Cohomology with Coefficients in K . . . . . . . . . . . . . . . . . . 161
A.2.3 The Euler Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
A.2.4 Action of a Frobenius Endomorphism . . . . . . . . . . . . . . . . 162
A.3 Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
A.3.1 The Projective Line . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
A.3.2 The One-Dimensional Torus . . . . . . . . . . . . . . . . . . . . . . . . . 164

Appendix B Block Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167


B.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
B.2 Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
B.2.1 Brauers Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
B.2.2 Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
Contents xvii

B.2.3 Equivalences of Categories: Methods . . . . . . . . . . . . . . . . 170


B.3 Decomposition Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
B.4 Brauer Trees* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

Appendix C Review of Reflection Groups . . . . . . . . . . . . . . . . . . . . . . . . . 177

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
List of Tables

1.1 Conjugacy classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

5.1 Values of Tr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.2 Values of Tr . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.3 Values of the characters R( ) and R ( ) . . . . . . . . . . . . . . . . . . . . . . 54
5.4 Character table of G = SL2 (Fq ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58

6.1 Central characters of G = SL2 (Fq ) . . . . . . . . . . . . . . . . . . . . . . . . . . . 64


6.2 Character table of N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.3 Character table of N  . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.4 Character table of B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

9.1 Decomposition matrix for A1 when  = 2 . . . . . . . . . . . . . . . . . . . . . 107

11.1 Character table of SL2 (F3 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132


11.2 Character table of SL2 (F5 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
11.3 Reflection groups having SL2 (F5 ) as derived group . . . . . . . . . . . 141
11.4 Character table of SL2 (F7 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

xix
General Notation

If E is a set, the cardinality of E (possibly infinite) will be denoted |E |. If


is an equivalence relation on E , we will denote by [E /] a set of represen-
tatives for the equivalence classes under . When this notation is used the
reader will have no difficulty verifying that the statement does not depend
on the choice of representative. In an expression of the form x[E /] f (x),
the reader will also be able to verify that the element f (x) depends only on
the equivalence class of x.
If is a group, we will denote by Z() its centre and D() its derived
subgroup. If E is a subset of , we denote by N (E ) (respectively C (E ))
its normaliser (respectively centraliser). If , we will denote by Cl ( )
its conjugacy class in and we set E = E 1 . Of course we have || =
|C ( )| | Cl ( )|. The order (possibly infinite) of will be denoted o( ). If
 , we will denote by [ ,  ] the commutator  1 1 .
If A is a ring, we will denote by A+ its underlying additive group and by

A the (multiplicative) group of invertible elements in A. The centre of A will
be denoted Z(A). We will denote by Amod the category of left A-modules of
finite type. Given another ring B, we denote by (A, B)bimod the category
of (A, B)-bimodules of finite type; that is, the category whose objects are
simultaneously a left A-module of finite type and a right B-module of finite
type satisfying (a m) b = a (m b) for all a A, b B and m M. Given a
commutative ring R, an R-algebra A, a commutative R-algebra R  and an A-
module M we will simplify the notation and denote by R  M the extension of
scalars R  R M: it is an R  A = R  R A-module. If in addition R is a field, we
will denote by K0 (A) the Grothendieck group of the category of A-modules
which are finite dimensional over R. If M is a left A-module which is finite
dimensional over R, we will denote by [M ] (or [M ]A if necessary) its class
in K0 (A). We denote by Irr A a set of representatives for the isomorphism
classes of simple (i.e. irreducible) A-modules which are finite dimensional
over R; it follows that K0 (A) is the free Z-module with basis ([S ])SIrr A .
We denote by Cb (A) the category of bounded complexes of A-modules,
K (A) the bounded homotopy category of A-modules and by Db (A) the de-
b

xxi
xxii General Notation

rived category of A-modules (see [CaEn, A1.2, A1.5 and A1.6]). After re-
placing A-modules by (A, B)-bimodules, we obtain the categories Cb (A, B),
Kb (A, B) and Db (A, B). If C and C  are two bounded complexes of A-
modules (or of (A, B)-bimodules), we will write C C C  (respectively C K
C  , respectively C D C  ) if C and C  are isomorphic in the category of com-
plexes (respectively in the homotopy category, respectively in the derived
category). We will denote by C [i] the complex shifted by i to the left [CaEn,
A1.2].
If M is a monoid and X is a subset of M , we denote by
X mon the sub-
monoid of M generated by X . If R is a commutative ring, the monoid alge-
bra of M over R will be denoted by RM . If K is a field and M is a finite di-
mensional KM -module, we use [M ]M to denote the class of M in K0 (KM )
if there is no possible ambiguity as to the field K.
If R is a ring and M is an R-module, we denote by GLR (M) the group
of its R-linear automorphisms. If n is a non-zero natural number, we will
denote by Matn (R) the R-algebra of n n square matrices with coefficients
in R and set GLn (R) = Matn (R) . The identity matrix will be denoted In .
Given M Matn (R), its transpose will be denoted t M.
If R is commutative, the R-algebra Matn (R) as well as the group GLn (R)
will be naturally identified with EndR (R n ) and GLR (R n ) respectively, using
the canonical identification of R n with the R-module of n 1 column vectors.
We will denote by SLn (R) the subgroup of GLn (R) consisting of matrices of
determinant 1. If K is a field, a matrix g GLn (K) will be said to be semi-
simple if it is diagonalisable over an algebraic closure of K. It will be said to
be unipotent if g In is nilpotent, that is, if (g In )n = 0.
Part I
Preliminaries
The representation theory of SL2 (Fq ) will not commence until Chapter 3.
Beforehand, in Chapters 1 and 2, we assemble the necessary preliminary
facts needed for the study of representations. In Chapter 1 we are interested
in the structure of the group SL2 (Fq ) (special subgroups, conjugacy classes,
structure of centralisers, Sylow subgroups and their normalisers). Chapter 2
begins the study of the geometric properties of the Drinfeld curve: we essen-
tially study certain interesting quotients as well as the action of the Frobe-
nius endomorphism.
Chapter 1
Structure of SL2 (Fq )

Notation. Let p be an odd prime number, F an algebraic clo-


sure of the finite field Fp with p elements, q a power of p,
Fq the subfield of F with q elements and G the group
 
x z
G = SL2 (Fq ) = { GL2 (Fq ) | xt yz = 1}.
y t

If n is a non-zero natural number, we denote by n the


group of the n-th roots of unity in F :

n = { F | n = 1}.

We will denote by Tr2 : Fq 2 Fq , + q and N2 : Fq2


Fq , 1+q the trace and norm respectively.

1.1. Special Subgroups

Before commencing, recall that

(1.1.1) |G | = q(q 1)(q + 1).

1.1.1. Bruhat Decomposition

Let B (respectively T , respectively U) denote the subgroup of G consisting


of upper triangular (respectively diagonal, respectively unipotent upper tri-
angular ) matrices. Then
B = T  U.

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 3


DOI 10.1007/978-0-85729-157-8_1, Springer-Verlag London Limited 2011
4 1 Structure of SL2 (Fq )

We have isomorphisms
 
a 0
d: Fq T , a diag(a, a 1
)=
0 a1

 
1x
and u: Fq+ U, x
.
01

Proposition 1.1.2. With the previous notation we have:


(a) The groups T and U are abelian and the group B is solvable.
(b) If a Fq and x Fq , then d(a)u(x)d(a)1 = u(a2 x).

Now set  
0 1
s= .
1 0
Then

(1.1.3) s 2 = I2 and s d(a) s 1 = d(a1 ).

In particular, s normalizes T . The Bruhat decomposition of G is the partition


of G into double cosets for the action of B:

(1.1.4) G = B BsB = B UsB.

Note also that

(1.1.5) B sB = Tand BsBsB = G .


 
ab
Proof (of 1.1.4 and 1.1.5). Let g = G \ B. Then c = 0 and therefore
cd
1
s u(a/c)g B. Hence g UsB. This proves 1.1.4.
The first equality of 1.1.5 is immediate. To establish the second equality
let X = BsBsB. Then X is stable by left and right multiplication by B and
s 2 = I2 X . It follows that X = B or X = G by 1.1.4. But su(1)s X \ B and
so X = G .  
In what follows we denote by N the group
T , s . The exact sequence

(1.1.6) 1 T N Z/2Z 1

is not split. In fact, all elements g N \ T satisfy g 2 = I2 . We will see later


that, if q > 3, then N is the normaliser of T (Corollary 1.3.2). It follows easily
from the Bruhat decomposition that


(1.1.7) G= UnU.
nN
1.1 Special Subgroups 5

1.1.2. The Non-Split Torus

The choice of a basis of the Fq -vector space Fq 2 induces an isomorphism



d : GLFq (Fq 2 ) GL2 (Fq ).

The group Fq2 acts on the Fq -vector space Fq 2 (which is of dimension 2) by


multiplication. We may therefore view it as a subgroup of GLFq (Fq 2 ). It is
then easy to verify that, if Fq2 , then

(1.1.8) Tr d ( ) = q + = Tr2 ( ) and det d ( ) = 1+q = N2 ( ).

(see Exercise 1.4 for an explicit construction of d and a proof). Therefore the
image T  of q+1 under d is contained in G .
The subgroup T (respectively T  ) of G will be called the split torus (respec-
tively the non-split torus, or anisotropic torus) of G . We have isomorphisms

(1.1.9) d : q1 = Fq T and d : q+1 T  .

The Frobenius automorphism F : Fq 2 Fq 2 , q is Fq -linear, and is


therefore an element of GLFq (Fq 2 ). Set s  = d (F ).We have

(1.1.10) s 2 = I2 and s  d ( )s 1 = d ( q ).

On the other hand,

(1.1.11) dets  = 1.

In fact, s  is of order 2 but does not commute with all the elements of GL2 (Fq )
by 1.1.10 and is therefore similar to the matrix diag(1, 1).
Fix an element 0 of Fq2 such that N2 (0 ) = 1 (the norm N2 : Fq2 Fq is
surjective by Exercise 1.4(a)) and set

s  = d (0 )s  .

Then s  G and, if q+1 ,

(1.1.12) s 2 = I2 and s  d ( ) s 1 = d ( q ) = d ( )1

In particular, s  normalizes T  . In what follows, the group


T  , s  will be
denoted N  . The exact sequence

(1.1.13) 1 T  N  Z/2Z 1

is not split. In fact, all the elements g N  \ T  satisfy g 2 = I2 . We will see


later that N  is the normaliser of T  (Corollary 1.3.2).
6 1 Structure of SL2 (Fq )

1.2. Distinguished Subgroups

It is easy to verify that

(1.2.1) Z(G ) = {I2 , I2 }.

To simplify notation, we set Z = Z(G ). We will show that, if q > 3, then G /Z


is simple. Beforehand we will need the following two lemmas:

Lemma 1.2.2. The group G is generated by U and sUs 1 . In particular, G =


U, s .

Proof. Set H =
U, sUs 1 . By 1.1.4, it is enough to show that s H and T H.
Two simple calculations show that

s = u(1) (su(1)s 1 ) u(1) H

and, if a Fq , then

d(a) = u(a) (su(a1 )s 1 ) u(a)s H,

which finishes the proof. 




Lemma 1.2.3. gBg 1 = Z .
g G

Proof. Set H = gBg 1 . It is clear that Z H. On the other hand, note that
g G
B s B = T . If we set u = u(1), then a calculation shows immediately that
u T T = Z . The result then follows. 

Theorem 1.2.4. We have:


(a) If q > 3, then D(G ) = G and Z is the only non-trivial normal subgroup of G .
In particular, G /Z is simple.
(b) If q = 3, the non-trivial normal subgroups of G are Z and N  . Moreover, G =
N   U and D(G ) = N  .

Proof. (a) Suppose that q > 3. We begin by showing that G = D(G ). As q > 3,
there exists a Fq such that a2 = 1. Now, if x Fq+ , it follows from Propo-
sition 1.1.2(b) that [d(a), u(x)] = u((a2 1)x). This shows that U D(G ) and
therefore that D(G ) = G by Lemma 1.2.2.
Let H be a non-trivial normal subgroup of G . If H is contained in B, then
H is contained in the intersection of the conjugates of B, that is H Z by
Lemma 1.2.3. If H is not contained in B, then set G  = HB. It is a subgroup
of G which strictly contains B, and so G  = G thanks to the Bruhat decom-
position 1.1.4. It follows that G /H B/(B H) and, as D(G ) = G , we obtain
that B/(B H) is equal to its derived group. As B is solvable, we must have
1.3 Conjugacy Classes 7

B H = B, and therefore H contains B and all of its conjugates. Therefore


G = HB = H by Lemma 1.2.2.
(b) is straightforward and will be shown in Chapter 11 (see Proposi-
tion 11.2.3), which treats particular cases related to small values of q. 


1.3. Conjugacy Classes

1.3.1. Centralisers

The following proposition describes the centralisers of certain elements of


our group G . As we will see in the following section, the list of elements
contained in the proposition is exhaustive up to conjugacy.

Proposition 1.3.1. Let g G . Then:


(a) If g {I2 , I2 }, then CG (g ) = G .
(b) If g U \ {I2 }, then CG (g ) = {I2 , I2 } U = ZU.
(c) If g = d(a) with a q1 \ {1, 1}, then CG (g ) = T .
(d) If g = d ( ) with q+1 \ {1, 1}, then CG (g ) = T  .

Proof. (a) is immediate, and (b) and (c) follow via elementary calculation.

Let us now show (d). Using the isomorphism d : GLFq (Fq 2 ) GL2 (Fq ), it
is enough to show
CGLFq (F 2 ) ( ) = Fq2 .
q

Firstly, it is clear that Fq2 is contained in the centraliser of . On the other


hand, fix g GLFq (Fq 2 ) such that g = g . Set 0 = g (1) Fq2 . We then have,
for all a and b in Fq ,

g (a + b ) = g (a) + bg ( ) = ag (1) + b g (1) = 0 (a + b )

as g and commute. But, as  {1, 1}, we have q = 1 = and therefore


Fq 2 = Fq Fq . It follows that g = 0 Fq2 . 


Corollary 1.3.2. We have:


(a) If q > 3, then CG (T ) = T and NG (T ) = N. If q = 3, then T = {I2 , I2 } = Z
and NG (T ) = CG (T ) = G .
(b) CG (T  ) = T  and NG (T  ) = N  .

Proof. (a) The case where q = 3 is immediate. Suppose therefore that q > 3.
It is clear that T CG (T ). Now choose g CG (T ). As q > 3, there exists an
element a Fq not equal to 1 or 1. As a consequence, g commutes with
d(a) T and therefore g T by Proposition 1.3.1(c).
8 1 Structure of SL2 (Fq )

For the second equality, note first that N NG (T ). On the other hand, let
g NG (T ). Then there exists b Fq such that g d(a) g 1 = d(b). This is only
possible if a = b or a = b 1 . If a = b, then g commutes with d(a) and therefore
belongs to T by Proposition 1.3.1(c). If a = b 1 , then sg commutes with d(a)
and therefore belongs to T , again by Proposition 1.3.1(c). The result follows.
(b) is shown in the same way after remarking that q+1 must contain an
element different from 1 and 1.  

1.3.2. Parametrisation

We denote by the relation on F defined by x y if y {x, x 1 }. The equiv-


alence classes of 1 and 1 contain a unique element, and all other classes
contain two elements. We also fix an element z0 Fq which is not a square
in Fq (which is possible as q is odd). Set
   
11 1 z0
u+ = and u = .
01 0 1

Theorem 1.3.3. The group G consists of q + 4 conjugacy classes. A set of represen-


tatives is given by

{I2 , I2 } {u+ , u , u+ , u } {d(a) | a [(q1 \ {1, 1})/ ]}

{d ( ) | [(q+1 \ {1, 1})/ ]}.


Proof. Denote by E the set

{I2 , I2 } {u+ , u , u+ , u } {d(a) | a [(q1 \ {1, 1})/ ]}

{d ( ) | [(q+1 \ {1, 1})/ ]}.


Let g and g  be two distinct elements of E . We will show that they are
not conjugate in G . In fact, in most cases they are not conjugate in GL2 (Fq )
(which can be deduced by comparing the eigenvalues over F). The only del-
icate case is to show that u+ and u (respectively u+ and u ) are not
conjugate in G (even though they are in GL2 (Fq )).
We will only prove this for u+ and u , the other case follows in the same
1
manner. Let us therefore suppose that there exists h G such that hu +h =
1 1
u . As Ker(u+ I2 ) = Ker(u I2 ) = Fq , h stabilises the line Fq . It
0 0
 
a b
follows that h B. If we write h = with a Fq and b Fq then a
0 a1
straightforward calculation shows that z0 = a2 , which is impossible.
It remains to show that every element of G is conjugate, in G , to an ele-
ment of E . For this, one could prove it using linear algebra. We have chosen
1.4 Sylow Subgroups 9

to prove it by a counting argument: it is enough to show that

| ClG (g )| = |G |.
g E

Using Proposition 1.3.1, we obtain

|G |
| ClG (g )| = |CG (g )|
g E g E
(q 1)(q + 1) q 3 q1
= 2+4 + q(q + 1) + q(q 1)
2 2 2
= |G |,

as expected. 


The results of this section are summarised in Table 1.1.

Table 1.1 Conjugacy classes

 
a
Representative I2 d(a) d ( )
0
a q1 \ {1} q+1 \ {1} {1}, a Fq

q 3 q 1
Number of classes 2 4
2 2

Order o(1) o(a) o( ) p o( )

q2 1
Cardinality 1 q(q + 1) q(q 1)
2

Centraliser G T T ZU

1.4. Sylow Subgroups

In the course of the description of Sylow subgroups of G , we will see that


some of the special subgroups introduced in Section 1.1 (B, U, T , N, T  , N  ...)
will occur as normalisers or centralizers of Sylow subgroups. The situation
is somewhat more complicated for Sylow 2-subgroups.
10 1 Structure of SL2 (Fq )

1.4.1. Sylow p-Subgroups

The following proposition is immediate.

Proposition 1.4.1. U is a Sylow p-subgroup of G . Moreover,

CG (U) = ZU and NG (U) = B.

1.4.2. Other Sylow Subgroups

Fix a prime number  not equal to p and dividing the order of G . We denote
by S (respectively S ) the -Sylow subgroup of T (respectively T  ). Note
that

(1.4.2) S and S are cyclic.

By 1.1.1,  divides (q 1)(q + 1). As gcd(q 1, q + 1) = 2 (recall that q is odd),


we have the following.

Theorem 1.4.3. Let  be a prime number different from p which divides the order
of G .
(a) If  odd and divides q 1, then S is a Sylow -subgroup of G . Moreover

CG (S ) = T and NG (S ) = N.

(b) If  is odd and divides q + 1, then S is a Sylow -subgroup of G . Moreover

CG (S ) = T  and NG (S ) = N  .

(c) If q 1 mod 4 (respectively q 3 mod 4), then


S2 , s (respectively
S2 , s  )
is a Sylow 2-subgroup of G .
(d) Let S be a Sylow 2-subgroup of G . Then

1 if q 1 mod 8,
CG (S) = Z and |NG (S)/S| =
3 if q 3 mod 8.

Proof. (a) If  is odd and divides q 1, then  divides neither q nor q + 1.


This shows that S is a Sylow -subgroup of G . Let g S \ {I2 }. Proposi-
tion 1.3.1(c) tells us that CG (g ) = T , which allows us to conclude easily that
CG (S ) = T . For the second equality, note that N NG (S ). On the other
hand, NG (S ) normalizes CG (S ) = T , and is therefore contained in N (as 
is odd and divides q 1, we have q > 3 and we can therefore apply Corol-
lary 1.3.2).
(b) is shown in the same way as (a).
1.4 Sylow Subgroups 11

(c) It is enough to remark that, if q 1 mod 4 (respectively q 3 mod 4),


then 4 does not divide q + 1 (respectively q 1). This allows us to determine
the 2-valuation of |G |.
(d) First suppose that q 3 mod 4. Set S =
S2 , s  . By (c), S is a Sylow 2-
subgroup of G . Now there exists g S2 \ {I2 , I2 } and therefore CG (g ) = T 
by Proposition 1.3.1(c). In particular, CG (S) T  . But, by 1.1.12, the only
elements of T  which commute with s  are I2 and I2 . The case where q 1
mod 4 is treated similarly.
It remains to calculate the normaliser of S.
If q 1 mod 8, we may suppose, by (c), that S =
S2 , s . Then S2 con-
tains an element of order 8, which we denote by t. Now t and t 1 are the
only elements of order 8 in S. Therefore, if g NG (S), then gtg 1 {t, t 1 },
which implies that g N. Remember that we hope to show that g S. After
multiplying by s, we may suppose that g T . But then gsg 1 S and hence,
as sgs 1 = g 1 , we have gsg 1 s 1 = g 2 S. Therefore g 2 S2 , which forces
g S2 . It follows that NG (S) = S.
If q 1 mod 8, the result is shown in the same way.
If q 3 mod 8, then |S| = 8 and, as I2 is the only element of order
2 in G , we can deduce that S is the quaternionic group of order 8. It is well
known that the group of outer automorphisms of S is isomorphic to S3 (via
the action on the three conjugacy classes of elements of order 4), therefore
NG (S)/SCG (S) = NG (S)/S is of order dividing 6. As it is, moreover, of odd
order, it can only be equal to 1 or 3. In order to show that it is of order 3,
we must construct an element of order 3 in the normaliser of S. We treat
only the case where q 5 mod 8, whereas the case where q 3 mod 8 is
relatively similar and left as an exercise (see Exercise 1.4).
So assume that q 5 mod 8. We may suppose that S =
S2 , s . Denote
by i an element of Fq for which i 2 = 1 (which exists as q 1 mod 4). Set
1+i
= and
2  

g= .
i i
It is simple to verify that g belongs to G , normalizes S and satisfies g 3 = I2 .
The proof of the theorem is finished.  

Exercises

1.1. Let k be the subfield of Fq generated by (Tr2 ( )) q+1 . Show that k = Fq .


 
(Hint: Set q  = |k| and show that, if q+1 , then 1 + 2 + q + q +1 = 0.)

1.2. Show that the map U B BsB, (u, b) usb is bijective.

1.3. Show that NG (B) = B.


12 1 Structure of SL2 (Fq )

1.4. The purpose of this exercise is to choose a basis over Fq of Fq 2 which


allows an easy and explicit construction of the matrices d ( ) for Fq2
and, in particular, to deduce a proof of 1.1.8.
(a) Show that Tr2 and N2 are surjective.
(b) For the rest of this exercise fix an element z Fq 2 such that z = 0 and
Tr2 (z) = 0 (i.e. z q + z = 0). Show that z  Fq and conclude that (1, z) is a

Fq -basis of Fq 2 . We construct the isomorphism d : GLFq (Fq 2 ) GL2 (Fq )
using the basis (1, z).
(c) Show that then s  = diag(1, 1).
(d) Let Fq2 . Show that

q + z( q )
2 2

d ( ) = .
q q +
2z 2
(e) Show that Tr d ( ) = Tr2 ( ), det d ( ) = N2 ( ) and that the characteristic
polynomial of d ( ) is (X )(X q ), where X is an indeterminate.
(f) Show that d ( ) is conjugate, in GL2 (Fq 2 ), to diag( , q ).
(g) Suppose that q 3 mod 8. Set S =
S2 , s  and denote by i an element of
q+1 such that i 2 = 1.
(g1) Show that S2 =
d (i) and calculate d (i).
(g2) Find a matrix g G such that gs  g 1 = d (i), g d (i)g 1 = s  d (i) and
show that g normalizes S and is of order 3 or 6.
(g3) Conclude that |NG (S)/S| = 3 (which completes the proof of Theo-
rem 1.4.3).

1.5. Denote by (respectively  ) the automorphism of T (respectively T  )


which sends an element to its inverse. Set M =
 T and M  =
  T  .
Let R be a commutative ring in which 2 is invertible. Show that the group
algebras RN and RM (respectively RN  and RM  ) are isomorphic.

1.6. Let H be a subgroup of T (respectively T  ) which is not contained in


Z . Show that CG (H) = T and NG (H) = N (respectively CG (H) = T  and
NG (H) = N  ).

1.7. Let n be a non-zero natural number and g GLn (F). Show that g is semi-
simple (respectively unipotent) if and only if its order is prime to p (respec-
tively a power of p). Remark: Note that g is of finite order as F = r  1 Fp r .

1.8. Show that u+ and u are conjugate in SL2 (Fq 2 ) (although they are not in
G = SL2 (Fq )). Show that they are also conjugate in GL2 (Fq ).
Show that two semi-simple elements of G are conjugate in G if and only
if they are conjugate in GL2 (Fq ).
1.4 Sylow Subgroups 13

1.9. Show that the number of unipotent elements of G is q 2 .

1.10. Show that the group G has q conjugacy classes of semi-simple ele-
ments.

1.11. Let r be a prime number and let P and Q be two distinct Sylow r -
subgroup of G . Show that:
(a) If r > 2, then P Q = {I2 }.
(b) If r = 2, then P Q = {I2 , I2 }.

1.12. Let P be a Sylow 2-subgroup of G . Show that P/D(P) Z/2Z Z/2Z.

1.13. Let K be a commutative field of characteristic different from 2 and


(e1 , e2 ) the canonical basis of K2 . Consider det : K2 K2 K, the alternating
bilinear (i.e.symplectic)
 form given by the determinant in the canonical ba-
0 1
sis. Let J = . Then J is the matrix of the symplectic form det in the
1 0
canonical basis. Finally, denote by Sp2 (K) the automorphism group of K2
which stabilise the symplectic form det.
(a) Show that Sp2 (K) = SL2 (K).
(b) Show that Sp2 (K) = {g GL2 (K) | t gJg = J}.
(c) Deduce that the automorphism of SL2 (K) defined by g t g 1 is inner
(and is induced by conjugation by J).
(d) Show that the automorphism of GL2 (K) given by g t g 1 is not inner.

1.14*. In this exercise we calculate the group of outer automorphisms of


G = SL2 (Fq ) and G = GL2 (Fq ). Let : Fq Fq , x x p be the Frobenius au-
tomorphism. We will also denote by the automorphism of G or G induced

by . Fix an element a0 Fq which is not a square and denote by the
automorphism of G induced by conjugation by diag(a0 , 1) G .

(a) Show that and are non-inner automorphisms of G .


Denote by Aut(G ) (respectively Inn(G ), respectively Out(G )) the group of
automorphisms of G (respectively inner automorphisms of G , respectively
outer automorphisms of G , i.e. Out(G ) = Aut(G )/ Inn(G )). If Aut(G ), we
will denote by its image in Out(G ). If g G , we set inn g : G G , h
ghg 1 .
(b) Show that and commute.
(c) Write q = p e . Show that 2 = e = Id
G.
(d) Let be an automorphism of G . Show that there exists g G such that
inn g stabilises B, U and T . (Hint: Use the fact that U is a Sylow p-
subgroup, that B is its normaliser and that all the Sylow p-subgroups of
G are conjugate in G .)
14 1 Structure of SL2 (Fq )

(e) Deduce that Out(G ) =



Z/2Z Z/eZ.
G
Denote by : G , g t g 1 . Recall that is not an inner automorphism
(see Exercise 1.13(d)).
of G
(f) Show that Out(G ) =

Z/2Z Z/eZ.
Chapter 2
The Geometry of the Drinfeld Curve

Let Y be the Drinfeld curve

Y = {(x, y ) A2 (F) | xy q yx q = 1}.

It is straightforward to verify that:


 
ab
G acts linearly on A2 (F) (via g (x, y ) = (ax +by , cx +dy ) if g = G)
cd
and stabilises Y ;
q+1 acts on A2 (F) by homotheties (via (x, y ) = ( x, y ) if q+1 )
and stabilises Y ;
the Frobenius endomorphism F : A2 (F) A2 (F), (x, y ) (x q , y q ) sta-
bilises Y.
Moreover, if g G and q+1 , then, as endomorphisms of A2 (F) (or Y),
we have

g = g,
g F = F g,
F = 1 F .

We can therefore form the monoid

G (q+1 
F mon )

which acts on A2 (F) and stabilises Y.


The purpose of this chapter is to assemble the geometric properties of
Y and the action of G (q+1 
F mon ) which allows us to calculate its -
adic cohomology (as a module for the monoid G (q+1 
F mon )). A large
part of this chapter is dedicated to the construction of quotients of Y by the
actions of the finite groups G , U and q+1 .

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 15


DOI 10.1007/978-0-85729-157-8_2, Springer-Verlag London Limited 2011
16 2 The Geometry of the Drinfeld Curve

2.1. Elementary Properties

The following proposition is (almost) immediate.


Proposition 2.1.1. The curve Y is affine, smooth and irreducible.
Proof. Y is affine because it is a closed subspace of the affine space A2 (F). It
is irreducible because the polynomial XY q YX q 1 in F[X , Y ] is irreducible
(See Exercise 2.1). It is smooth
 because
 the differential of this polynomial is
given by the 1 2 matrix Y q X q , which is zero only at (0, 0)  Y.  
Proposition 2.1.2. The group G acts freely on Y.
Proof. Let g G and (x, y ) Y be such that g (x, y ) = (x, y ). It follows that
1 is an eigenvalue of g and, after conjugating g by an element of G , we may
assume that there exists an a Fq such that
 
1a
g= .
01

Then x + ay = x and, as y = 0 (since (x, y ) Y), we conclude that a = 0. 



The next proposition is clear.
Proposition 2.1.3. The group q+1 acts freely on A2 (F) \ {(0, 0)} and therefore
also on Y.
Note however, that the group G q+1 does not act freely on Y: the pair
(I2 , 1) acts as the identity. (Even the quotient (G q+1 )/
(I2 , 1) does
not act freely, see Exercise 2.3.)

2.2. Interesting Quotients

We will now describe the quotients of Y by the finite groups G , U and q+1 .
In order to construct them we will use the following proposition, a proof of
which can be found in [Bor, Proposition 6.6]. (Note that the proposition is
far from optimal, but will be sufficient for our needs.)
Proposition 2.2.1. Let V and W be two smooth and irreducible varieties, : V
W a morphism of varieties, and a finite group acting on V. Suppose that the
following three properties are satisfied:
(1) is surjective;
(2) (v ) = (v  ) if and only if v and v  are in the same -orbit;
(3) There exists v0 V such that the differential of at v0 is surjective.
Then the morphism : V/ W induced by is an isomorphism of varieties.
2.2 Interesting Quotients 17

2.2.1. Quotient by G

The map
: Y A1 (F)
2 2
(x, y ) xy q yx q
is a morphism of varieties. It is q+1 
F mon -equivariant (for the action
of q+1 on A1 (F) given by z = 2 z and the action of F given by z
z q ). An elementary calculation shows that is constant on G -orbits. Even
better, if we denote by : Y/G A1 (F) the morphism of varieties obtained
by passing to the quotient, we have the following.

Theorem 2.2.2. The morphism of varieties : Y/G A1 (F) is a q+1 


F mon -
equivariant isomorphism.

Proof. The q+1 


F mon -equivariance is evident. In order to show that is
an isomorphism we must verify points (1), (2) and (3) of Proposition 2.2.1.
Choose a F. To show (1) and (2), it is sufficient to show that | 1 (a)| =
|G | (as G acts freely on Y by Proposition 2.1.2). After changing variables

(z, t) = (x, y /x), we have a bijection 1 (a) Ea , where

1 a
Ea = {(z, t) F F | t q t =
2
and t q t = 2 }.
z q+1 z q +1
2
As t q t = (t q t)q + (t q t), we obtain

1 1 1 a
Ea = {(z, t) F F | t q t = and q+1 + q 2 +q = q 2 +1 }.
z q+1 z z z
Or equivalently

1
Ea = {(z, t) F F | z q
2 1
az q1 + 1 = 0 and t q t = }.
z q+1

The polynomial z q 1 az q1 + 1 is coprime to its derivative, and therefore


2

has q 2 1 distinct non-zero roots. For each of these roots, there are q non-
1
zero solutions t to the equation t q t = q+1 . Therefore
z
| 1 (a)| = |Ea | = (q 2 1)q = |G |,

as expected.
We now turn to (3). Let v = (x0 , y0 ) Y. The tangent space Tv (Y) to
Y at v has equation y0q x x0q y = 0 and the differential dv : Tv (Y) F =
T (v ) (A1 (F)) is given by
2 2
dv (x, y ) = y0q x x0q y .
18 2 The Geometry of the Drinfeld Curve

Therefore, if (x, y ) Ker dv , then


2 2
y0q x x0q y = 0 and y0q x x0q y = 0.
2 2
The determinant of this system is y0q x0q +x0q y0q = (x0 y0q y0 x0q )q = 1, there-
fore Ker dv = 0. 


2.2.2. Quotient by U

The morphism
: Y A1 (F) \ {0}
(x, y ) y
is well-defined and is a morphism of varieties. It is q+1 
F mon -equivariant
(for the action of q+1 on A1 (F) \ {0} given by z = z and the action of
F given by z z q ). An elementary calculation show that is constant on
U-orbits. Even better, if we denote by : Y/U A1 (F) \ {0} the morphism
of varieties induced by passing to the quotient, we have the following.

Theorem 2.2.3. The morphism of varieties : Y/U A1 (F) \ {0} is a q+1 

F mon -equivariant isomorphism.

Proof. The q+1 


F mon -equivariance is evident. To show that is an iso-
morphism, we verify points (1), (2) and (3) of Proposition 2.2.1.
The surjectivity of is clear. We also have

(x, y ) = (x  , y  ) u U, (x  , y  ) = u (x, y ).

Indeed, if (x, y ) Y and (x  , y  ) Y are such that y = y  , then


 x q x  x  q x 
= ,
y y y y

x x x x
which shows that Fq . Now, if we set a = , then
y y
     
1a x x
= .
01 y y

This shows (2). Point (3) is immediate. 



2.3 Fixed Points under certain Frobenius Endomorphisms 19

2.2.3. Quotient by q+1

The morphism
: Y P1 (F) \ P1 (Fq )
(x, y ) [x : y ]
is well-defined and is G
F mon -equivariant morphism of varieties (for the
action of G induced by the natural action on P1 (F) and the action of F given
by [x; y ] [x q ; y q ]). An elementary calculation show that is constant on
q+1 -orbits. Even better, if we denote by : Y/q+1 P1 (F) \ P1 (Fq ) the
morphism of varieties induced by passage to the quotient, we have the fol-
lowing.
Theorem 2.2.4. The morphism of varieties : Y/q+1 P1 (F) \ P1 (Fq ) is a G

F mon -equivariant isomorphism.


Proof. The G
F mon -equivariance is evident. To show that is an isomor-
phism, we should verify points (1), (2) and (3) of Proposition 2.2.1, which is
straightforward.  

2.3. Fixed Points under certain Frobenius Endomorphisms

In order to get the most out of the Lefschetz fixed-point theorem (see Theo-
rem A.2.7(a) in Appendix A) we will need the following two results. Firstly,
note that, if q+1 , we have

(2.3.1) Y F = .

Indeed, (Y/q+1 )F = by Theorem 2.2.4. On the other hand, we have the


following.
Theorem 2.3.2. Let q+1 . Then

F2 0 if =
 1,
|Y | =
q3 q if = 1.

Proof. Let (x, y ) Y F . We then have


2

2 2
x = xq , y = yq and xy q yx q = 1.

As a consequence,
2 2
1 = (xy q yx q )q = x q y q y q x q = (x q y xy q ) = .

This shows that, if = 1, then Y F = .


2
20 2 The Geometry of the Drinfeld Curve

Therefore suppose that = 1. We are looking for the number of solu-


tions to the system
q2
x = x (1)
xy yx = 1 (2)
q q

2
y = y q (3)
However, if the pair (x, y ) satisfies (1) and (2), then it also satisfies (3). In-
1 + yx q
deed, if (x, y ) satisfies (1) and (2), then x = 0, y q = and therefore
x

2
 1 + yx q q 1 + y qxq
2
1 xy q yx q
yq = = q
= q
= = y .
x x x xq
It follows that it is sufficient to find the number of solutions to the system
given by equations (1) and (2). Now, x being non-zero, there are q 2 1 pos-
sibilities for x to be a solution of (1). As soon as we have fixed x, there are q
solutions to equation (2) (viewed as an equation in y ). Indeed, as an equation
in y , xy q yx q 1 has derivative x q = 0, and so this polynomial does not
admit multiple roots. This gives therefore (q 2 1)q solutions to equations
(1) and (2), and the theorem follows.  

R EMARK As G acts freely on Y, the set YF consists of a single G -orbit. 


2

2.4. Compactification

We will denote by [x; y ; z] homogeneous coordinates on the projective space


P2 (F). We view A2 (F) as the open subset of P2 (F) defined by

A2 (F) {[x; y ; z] P2 (F) | z = 0}.

We identify P2 (F) \ A2 (F) with P1 (F) (using the canonical isomorphism


[x; y ] [x; y ; 0]).Theaction of G (q+1 
F ) on A2 (F) extends uniquely
ab
to P2 (F): if g = , q+1 and [x; y ; z] P2 (F), then
cd

g [x; y ; z] = [ax + by ; cx + dy ; z],

[x; y ; z] = [ x; y ; z]

and F [x; y ; z] = [x q ; y q ; z q ].

Now let Y be the projective curve defined by

Y = {[x; y ; z] P2 (F) | xy q yx q = z q+1 }.


2.5 Curiosities* 21

The morphism
Y Y
(x, y ) [x; y ; 1]
is an open immersion and allows us to identify Y with Y A2 (F).
Proposition 2.4.1. The closed subvariety Y of P2 (F) is the closure of Y in P2 (F).
It is smooth and stable under the action of of G (q+1 
F ). Moreover,

Y \ Y P1 (Fq ),

with this isomorphism given by [x; y ] P1 (Fq ) [x; y ; 0].


Proof. The only point needing a little work is the smoothness. The points of
Y are smooth by Proposition 2.1.1. As G acts transitively on Y \Y = P1 (Fq )
G /B (as a G -set), it is enough to show that [1; 0; 0] is a smooth point of Y. For
this, let us consider the open subvariety defined by x = 0. In this open set
(again isomorphic to A2 (F), this time via the morphism (y , z) [1; y ; z]) Y is
defined by the equation y y q z q+1 = 0 and the differential at (0, 0) of this
polynomial is the 1 2 matrix

(1 0),

which is non-zero. 

We finish with a study of the quotient of Y by q+1 . Consider the mor-
phism
0 : Y P1 (F)
[x; y ; z] [x; y ].
It is well-defined, G -equivariant, and surjective. Moreover, it is constant on
q+1 -orbits and therefore induces, after passing to the quotient, a morphism
of varieties 0 : Y/q+1 P1 (F).
Theorem 2.4.2. The morphism of varieties 0 : Y/q+1 P1 (F) is a G
F mon -
equivariant isomorphism.
Proof. We omit the proof, as it follows the same arguments as those used in
the proof of Theorem 2.2.4.  

2.5. Curiosities*

Independent of representation theory, the Drinfeld curve has interesting ge-


ometric properties which we discuss briefly here: it has a large automor-
phism group and gives a solution to a particular case of the Abhyankars
Conjecture [Abh] about unramified coverings of the affine line in positive
characteristic.
22 2 The Geometry of the Drinfeld Curve

2.5.1. Hurwitz Formula, Automorphisms*

The group q+1 acts trivially on Y \ Y = P1 (Fq ). Also, as q+1 is of order


prime to p, the morphism 0 is tamely ramified: it is only ramified at the
points a P1 (Fq ) and ramification index at a is ea = q + 1. If we denote by
g(Y) the genus of Y, then

q(q 1)
(2.5.1) g(Y) =
2
as Y is a smooth plane curve of degree q + 1. Note also that 0 is a morphism
of degree deg 0 = q + 1. We can therefore verify the Hurwitz formula [Har,
Chapter IV, Corollary 2.4]

2g(Y) 2 = (deg 0 )(2 g(P1 (F)) 2) + (ea 1),


aP1 (Fq )

as g(P1 (F)) = 0.
We will now extend the group G q+1 to a bigger group G still acting
on Y (or Y). Set

G = {(g , ) GL2 (Fq ) Fq2 | det(g ) = 1+q }.

It is then straightforward to verify that,

(2.5.2) if (g , ) G and (x, y ) Y, then g ( x, y ) Y.

This defines for us an action of G on Y which extends naturally to an action


on Y. Set 

(I2 , 1) if q 3 mod 4,
D=

( 1 I2 , 1) if q 1 mod 4.
Then D is a central subgroup of G contained in the kernel of the action on Y
(and on Y). Even better, we have the following.
Lemma 2.5.3. The group G /D acts faithfully on Y (and Y).
Proof. Let (g , ) be an element of G which acts trivially on Y. Then (g , )
acts trivially on Y (as Y is dense in Y) and, after passing to the quotient
by {1} q+1 (which is a central subgroup of G ), we conclude that g acts
trivially on on P1 (F) (by Theorem 2.4.2). Therefore g is a homothety: g = I2 ,
with Fq .
Now, if (x, y ) Y, we have (g , ) (x, y ) = (x, y ), that is = 1. Therefore
= 1 . On the other hand, det(g ) = q+1 , which implies that 2 = q+1 or,
in other words, q+3 = 1. As q1 = 1, we collude that 4 = 1, which finishes
the proof.  
Let = D (G q+1 ) =
(I2 , 1) .
2.5 Curiosities* 23

Corollary 2.5.4. The group (G q+1 )/ acts faithfully on Y.

Denote by p1 : G GL2 (Fq ) and p2 : G Fq2 the canonical projections,


and i1 : q+1 G , (I2 , ) and i2 : G G , g (g , 1). The group G q+1
is contained in G and we set d : G Fq , (g , ) det(g ). We have a commu-
tative diagram

1 1

q+1 Fq2
i2 p2
i2 N2
d
1 G q+1 G Fq 1
i1 p1
i1 det

G GL2 (Fq )

1 1

in which all straight lines of the form 1 X G Y 1 are exact se-


quences (which follows essentially from the surjectivity of N2 ). In particular,

(2.5.5) |G | = q(q 2 1)2 .

It follows from Lemma 2.5.3 that



q(q 2 1)2

if q 3 mod 4,
2
| Aut Y| 



q(q 1)
2 2
if q 1 mod 4.
4
In particular, as soon as q  7, we have, by 2.5.1,

| Aut Y| > 84(g(Y) 1) = 42(q 2)(q + 1).

This illustrates the fact that the Hurwitz bound [Har, Chapter IV, Exer-
cise 2.5] is not valid in positive characteristic.
24 2 The Geometry of the Drinfeld Curve

2.5.2. Abhyankars Conjecture (Raynauds Theorem)*

It is not too difficult to show that if a finite group is the Galois group of
an unramified covering of the affine line A1 (F), then is generated by its
Sylow p-subgroups. The other implication was conjectured by Abhyankar
and shown by Raynaud in a very difficult work [Ray].

Raynauds theorem (Abhyankars conjecture). A finite group is the Galois


group of an unramified Galois covering of the affine line A1 (F) if and only if it is
generated by its Sylow p-subgroups.

E XAMPLE The morphism A1 (F) A1 (F), x x q x is an unramified Ga-


lois covering of A1 (F) with Galois group Fq+ . 

By Proposition 1.4.1 and Lemma 1.2.2, the group G = SL2 (Fq ) is generated
by its Sylow p-subgroups. By virtue of Raynauds theorem, G should be the
Galois group of an unramified covering of A1 (F). In fact, in this particular
case, the construction of such a covering is easy: the isomorphism Y/G
A1 (F) and the fact that G acts freely on Y (see Proposition 2.1.2) tells us that

(2.5.6) Y is an unramified Galois covering of A1 (F) with Galois group SL2 (Fq ).

Exercises

2.1. Show that the polynomial XY q YX q 1 in F[X , Y ] is irreducible (Hint:


By performing the change of variables (Z , T ) = (X /Y , 1/Y ) reduce the prob-
lem to showing that T q+1 Z q Z in F[Z , T ] is irreducible. View this as a
polynomial in T with coefficients F[Z ] and use Eisensteins criterion).
2.2*. Let F[X , Y ] a the polynomial ring in two variables, which we identify
with the algebra of polynomial functions on A2 (F). If g G , P F[X , Y ] and
v A2 (F), we set (g P)(v ) = P(g 1 v ).
(a) Show that this does indeed give an action of G via F-algebra automor-
phisms.
(b) Show that XY q1 X q and Y are algebraically independent and that
F[X , Y ]U = F[XY q1 X q , Y ].
2 2
(c) Show that XY q YX q divides XY q YX q .
2 2
XY q YX q
(d) Show that D1 = XY q YX q and D2 = are algebraically
XY q YX q
independent.
(e) Show that F[X , Y ]G = F[D1 , D2 ] (Dickson invariants).
(f) Use this to give another proof of Theorem 2.2.2.
2.5 Curiosities* 25

2.3. Denote by the subgroup of G q+1 generated by (I2 , 1). The pur-
pose of this exercise is to show that (G q+1 )/ does not act freely on Y.
To this end, choose q+1 \ {1, 1} and let v = (x, y ) A2 (F) be an eigen-
vector of d ( ) with eigenvalue .
(a) Show that xy q yx q = 0 (Hint: xy q yx q = x aFq (y + ax)).
(b) Let F be such that 1q = xy q yx q . Show that v Y.
(c) Show that (d ( ), 1 ) stabilises v Y.

2.4. Let Z = {(x, y ) A2 (F) | x q+1 + y q+1 + 1 = 0}. We keep the notation
F for the restriction to Z of the Frobenius endomorphism F of A2 (F). The
purpose of this exercise is to construct an isomorphism of Y and Z which
commutes with F 4 .
2
(a) Show that ZF = . Deduce that there does not exist an isomorphism of

varieties : Y Z such that F 2 = F 2 .
1
Let z Fq 2 \ Fq and d F be such that d q+1 = q .
z z
(b) Show that d Fq 4 .
 q q 
d z dz
(c) Let g = . Show that g GL2 (Fq 4 ) and that g (Z) = Y.
dq d

2.5. Denote by : Y/U Y/G the canonical projection. Set  =


1 : A1 (F) \ {0} A1 (F), so that the diagram


Y/U Y/G


A1 (F) \ {0} A1 (F)

commutes. Show that  (y ) = y q (y q + y ).


2

The author is indebted to G. Lusztig to whom this exercise is due.


Part II
Ordinary Characters
The purpose of the next three chapters is to calculate the character ta-
ble of G = SL2 (Fq ). To begin with, algebraic methods (in particular Harish-
Chandra induction) give roughly half of the irreducible characters (see
Chapter 3). The other half (the cuspidal characters) can be obtained in dif-
ferent ways (for example via ad hoc construction, as was done by Jordan
[Jor] and Schur [Sch] in 1907). In this book we take the approach of con-
structing the cuspidal characters using the -adic cohomology of the Drin-
feld curve (see Chapter 4). For this we will use results concerning -adic
cohomology contained in Appendix A as well as some geometric properties
obtained in Chapter 2.
Once the parametrisation has been obtained, the calculation of the char-
acter table is completed in Chapter 5: it raises geometric questions (trace on
-adic cohomology) as well as more arithmetic questions (Gauss sums).

N OTATION We fix a prime number  different from p and K , an algebraic


extension of the -adic field containing the ||-th roots of unity, for all finite
groups encountered in this book. This ensures that the algebra K is split
(by Brauers theorem [Isa, Theorem 8.4]).
If G is a finite group, we will denote by Irr its set of irreducible characters
over K (which we identify with Irr K ) and the Grothendieck group K0 (K )
will be identified with the character group Z Irr . If M is a K -module, we
therefore identify its isomorphism class [M ] with its character. We will de-
note by reg the character of the regular representation K of and by
,
the scalar product on K0 (K ) for which the set Irr is an orthonormal basis.
We denote by 1 (or 1) the trivial character of .
The group of linear characters of (with values in K ) will be denoted by
(so that Irr ). Given , we denote by K the K -module with

underlying vector space K and on which acts via the linear character .
If K0 (K ), we will denote by the virtual character defined by
( ) = ( 1 ) for all : is the character dual to (and is auto-dual if

= ). If M is a left K -module of finite type, its K -dual M can either be


viewed as a right K -module or as a left K -module (via the inverse map);
in the second case, we have [M ] = [M ] . If Irr , we will denote by e
(or e if necessary) the associated central primitive idempotent of K :

(1)
||
e = ( 1 ) .

Chapter 3
Harish-Chandra Induction

In this chapter we study Harish-Chandra induction, which associates to a


T -module the G -module obtained by first extending the T -module to a B-
module (letting U act trivially) and then inducing to G . This construction
allows us to obtain roughly half of the irreducible characters of G .

3.1. Bimodules

Let and  be two finite groups and let M be a (K , K  )-bimodule of


finite type. The dual bimodule M = Hom(M, K ) is naturally a (K  , K )-
bimodule. We define two functors
FM : K  mod K mod
V M K  V 

F K mod K  mod
M:
and
V M K V .

Because K and K  are semi-simple algebras, the bimodule M is projec-


tive both as a left K -module and as a right K  -module. It follows that the
functors FM and FM are left and right adjoint:

(3.1.1) HomK (V , FM V  ) HomK  ( FM V , V  )

and

(3.1.2) HomK (FM V  , V ) HomK  (V  , FM V )

We denote by FM : K0 (K  ) K0 (K ) and FM : K0 (K ) K0 (K  ) the Z-


linear maps induced by FM and FM respectively. They are characterised

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 29


DOI 10.1007/978-0-85729-157-8_3, Springer-Verlag London Limited 2011
30 3 Harish-Chandra Induction

by

(3.1.3) FM [V  ] = [M K  V  ] and
FM [V ] = [M K V ] .

Recall that, if V1 and V2 are two K -modules of finite type, then

(3.1.4)
[V1 ], [V2 ] = dimK HomK (V1 , V2 ).

Recall also that the Grothendieck group K0 (K ) is identified with the group
of (virtual) characters of . Under this identification, it follows from 3.1.1
and 3.1.4 that

(3.1.5)
, FM (  ) =
FM ( ),  

for all K0 (K ) and  K0 (K  ). If ( ,  )  , we denote by TrM ( ,  )


the trace of ( ,  ) on M. We have (see, for example, [DiMi, Proposition 4.5])

1
| |
(3.1.6) FM  ( ) = TrM ( ,  )  ( 1 )
 

and
1
||

(3.1.7) FM (  ) = TrM ( ,  ) ( 1 ).

3.2. Harish-Chandra Induction

3.2.1. Definition

In our group G , Harish-Chandra induction is defined using the bimodule


K [G /U]: by convention, K [G /U] is the K -vector space with basis G /U on
which G (respectively T ) acts by left (respectively right) translations on the
basis vectors. This is well-defined as T normalizes U. The dual of K [G /U]
may be naturally identified with K [U\G ], with the action of the group G
(respectively T ) given by right (respectively left) translations. In this way
we obtain two functors
RK : KT mod KG mod
V K [G /U] KT V

R : KG mod KT mod
K
and
W K [U\G ] KG W ,
3.2 Harish-Chandra Induction 31

called Harish-Chandra induction and restriction respectively. We denote by


R : K0 (KT ) K0 (KG ) and R : K0 (KG ) K0 (KT ), the induced Z-linear
maps.
An irreducible character g of G will be called cuspidal if there does not
exist a character of T q1 for which
, R( ) G = 0. In other words,
taking into account 3.1.5,

(3.2.1) is cuspidal if and only if R( ) = 0.

3.2.2. Other Constructions

In order to study these functors, it will be useful to interpret them using


classical induction and restriction. We will need the following notation: if V
(respectively ) is a KT -module (respectively a character of KT ), we will de-
note by VB (respectively B ) its restriction to B via the natural projection
B T . Note that [VB ]B = ([V ]T )B .

Proposition 3.2.2. Let V (respectively W ) be a KT -module (respectively a KG -


module). Then
RK V IndG B VB and RK W W U .

B VB = KG KB VB and RK (V ) = K [G /U] KT V . Set


Proof. We have IndG

1
eU = u.
|U| uU

Denote by U : KG K [G /U] the canonical morphism and U : K [G /U]


KG , gU geU . We have

U U (g ) = geU and U U (x) = x.

Set
: KG KB VB K [G /U] KT V
a KB v U (a) KT v

: K [G /U] KT V KG KB VB
and
a KT v U (a) KB v .

It is easy to verify that and are well-defined and morphisms of KG -


modules. It is also clear that = IdK [G /U]KT V . In order to show that
= IdKG KB VB , it is enough to note that a KB v = aeU KB v for all a KG
and v VB . This shows the first isomorphism.
Note that RK W = K [U\G ] KG W . Set
32 3 Harish-Chandra Induction

 : K [U\G ] KG W W U
a KG w U (a)w

 : W U K [U\G ] KG W
and
w U KG w .

Here, U (Ug ) = eU g for all g G . It is easy to verify that  and  are well-
defined and are mutually inverse isomorphisms of KT -modules.  

Corollary 3.2.3. Let be a character of T . Then R( ) = IndG


B B .

3.2.3. Mackey Formula

It follows easily from the Bruhat decomposition G = B BsB (see 1.1.4


and 1.1.5), from Corollary 3.2.3 and from the Mackey formula for classical
induction and restriction that, if , K0 (KT ), then


R( ), R( ) G =
, T +
, s T .

Here, s (t) = (s 1 ts). As s 1 ts = t 1 for all t T , we have, if ,


K0 (KT ),

(3.2.4)
R( ), R( ) G =
, T +
, T .

Note that, if is a linear character, then = 1 .


Let us now fix T = Irr T . By 3.2.4, we have:
If 2 = 1, then R( ) = R( 1 ) Irr G .
If = 0 (where 0 is the unique linear character of order 2 of T q1 ,
which exists because q is odd), then

R(0 ) = R+ (0 ) + R (0 ),

where R (0 ) Irr G and R+ (0 ) = R (0 ).


1G is a factor of R(1) and we denote by StG the other irreducible factor of
R(1) (the Steinberg character). We have

R(1) = 1G + StG , with deg 1G = 1 and deg StG = q.

If  { , 1 }, then
R( ), R( ) G = 0.
3.2 Harish-Chandra Induction 33

3.2.4. Restriction from GL2 (Fq )

In order to calculate the degrees of the characters R+ (0 ) and R (0 ) we will


show that they are conjugate under the action of G = GL2 (Fq ). Denote by B
(respectively T ) the subgroup of G consisting of upper triangular (respec-
tively diagonal) matrices. Then

(3.2.5) =T
B U =B
and G Bs
B.

, we denote by its restriction to B


If is a character of T via the surjec-
B
tion B T . We have

B B
ResB T
= (ResT )B .

We set
) = IndG .
R( B B

= G B,
As G we deduce from the Mackey formula for classical induction
and restriction that

G R( ) = R(ResT ).
ResG T
(3.2.6)

On the other hand, it follows from 3.2.5 that

(3.2.7) ), R(

R( ) =
, +
, s .
T T

In particular,

(3.2.8) is such that s = , then R(


if T ) is irreducible.

The following lemma is elementary.


of 0 , then s 0 = 0 .
Lemma 3.2.9. If 0 it an extension to T
The next corollary then follows immediately from 3.2.6 and 3.2.8.
of 0 , then R(
Corollary 3.2.10. If 0 is an extension to T 0 ) is an irreducible

character of G .
By 3.2.6 and Clifford theory [Isa, Theorem 6.2], we obtain the following.
Corollary 3.2.11. The irreducible characters R+ (0 ) and R (0 ) of G are conju-
.
gate under the action of G
In particular,

q +1
(3.2.12) deg R+ (0 ) = deg R (0 ) = .
2
Another proof of 3.2.12 will be given in Exercise 3.3.
34 3 Harish-Chandra Induction

3.2.5. Summary

We have therefore obtained:


one linear character 1G ;
one character of degree q, the Steinberg character StG ;
(q 3)/2 characters of degree q + 1 (the characters R( ), 2 = 1);
two characters R+ (0 ) and R (0 ) of degree (q + 1)/2 (and conjugate un-
).
der G
This yields (q + 5)/2 irreducible characters. As the number of conjugacy
classes (and therefore the number of irreducible characters) of G is q + 4, it
remains to construct (q + 3)/2 irreducible characters: these are the cuspidal
characters of G . Note also that

(3.2.13) R(regT ) = IndG 1G + StG + R+ (0 ) + R (0 ) + R( ).


U 1U =       2 =1
R(1) R(0 )

Exercises

denote the centre of G


3.1. Let Z .
= {diag(a, a) | a F }.
(a) Show that Z q
/G Z
(b) Show that |G | = 2.

3.2. Use the fact that R(1) is the character of the permutation representation
K [P1 (Fq )] in order to calculate its value on all conjugacy classes. Then verify
that, if g G ,

|CG (g )|p if g is semi-simple,
StG (g ) =
0 otherwise.

Here, ||p denotes the largest power of p which divides ||.


1 . Show that St is an irreducible character of
Denote by StG = R(1) G G
, then
, that ResG St = StG and that, if g G
G G G

|CG (g )|p if g is semi-simple,
StG (g ) =
0 otherwise.

3.3*. The goal of this exercise is to give a new proof of 3.2.12. If g G , we set

F (gU) = gusU K [G /U].


uU
3.2 Harish-Chandra Induction 35

(a) Show that this gives a well-defined K -linear endomorphism F of


K [G /U].
(b) Show that, if g G , x K [G /U] and t T , then

F (g x t) = g F (x) s t.

If T , we let V = RK (K ) (= K [G /U] KT K ).
(c) Show that F induces an isomorphism of KG -modules V V s .
Now fix T such that 2 = 1. By (c), F induces an automorphism of the
KG -module V which we will denote by F .
 
(d) Show that F2 = (1)q IdV + (a) F and that Tr(F ) = 0.
aFq
(e) Use (d) to deduce that (F1 q IdV1 )(F1 + IdV1 ) = 0 and that, if we set
I = Ker(F1 q IdV1 ) and S = Ker(F1 + IdV1 ), then V1 = I S is a decom-
position of V1 as a sum of irreducible KG -modules such that [I ] = 1G and
[S] = StG . 
(f) Deduce from (d) that, if we set V+0 = Ker(F0 0 (1)q IdV0 ) and

V0 = Ker(F0 + 0 (1)q IdV0 ), then V0 = V+0 V0 is a decomposi-
q+1
tion of V0 as a sum of irreducible KG -modules and that dimK V0 = .
2

R EMARK We have

1 if q 1 mod 4,
0 (1) =
1 if q 3 mod 4.

Consequently, the number 0 (1) is characterised by the property that


1 0 (1)q
0 (1)q 1 mod 4. Note that this implies that is an alge-
2
braic integer.

3.4. Let V be a finite dimensional left KT -module. We view the dual V


as a left KT -module. Show that we have an isomorphism of KG -modules
RK V (RK V ) . Hence deduce that R( ) = R( ) for every character
of T .
In particular, R(0 ) = R(0 ). Show that

R+ (0 ) if q 1 mod 4,
R+ (0 ) =
R (0 ) if q 3 mod 4.

Hint:
 Use Exercise 3.3 (and the remark that follows it, which implies that
0 (1)q is a real number if and only if q 1 mod 4).
Chapter 4
Deligne-Lusztig Induction

We will use the action of G q+1 on Y to construct a morphism between


the Grothendieck groups K0 (K q+1 ) and K0 (KG ). To this end, from now on
we will view the monoid q+1 
F mon as acting on the right on the Drinfeld
curve Y. It follows that the cohomology groups Hci (Y) inherit the structure
of (KG , K [q+1 
F mon ])-bimodules.
We will systematically use the results of Appendix A (which are refer-
enced as A.x.y).

4.1. Definition and First Properties

4.1.1. Definition

If is a character of q+1 , we set

R ( ) = (1)i [Hci (Y) K q+1 V ]G ,


i 0

where V is a K q+1 -module admitting the character . If is linear, we take


V = K . This defines a Z-linear map

R : K0 (K q+1 ) K0 (KG )

which sends a character of q+1 to a virtual character of G . The linear map


R will be called Deligne-Lusztig induction.
As the curve Y is affine and irreducible of dimension 1 it follows from
Theorem A.2.1(b) that Hci (Y) = 0 if i  {1, 2}. As a consequence,

R ( ) = [Hc1 (Y) K q+1 V ]G [Hc2 (Y) K q+1 V ]G .

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 37


DOI 10.1007/978-0-85729-157-8_4, Springer-Verlag London Limited 2011
38 4 Deligne-Lusztig Induction

On the other hand, the irreducibility of Y (see Proposition 2.1.1) and Theo-
rem A.2.1(c) tells us that

(4.1.1) [Hc2 (Y)]G q+1 = 1G q+1 .

The following result is then immediate.


Corollary 4.1.2. If is a non-trivial linear character of q+1 , then

R ( ) = [Hc1 (Y) K q+1 V ]G .

In this case, R ( ) is a character of G .


Proposition 4.1.3. Let K0 (K q+1 ). Then R ( ) = R ( ) = R ( ) .
Proof. Denote by : q+1 q+1 the homomorphism 1 . In order
to show the first equality, it suffices to show that R ( ) = R ( ), where
( ) = ( ( )). Denote by H i (Y) the (KG , K
c q+1 )-bimodule on which
the action of q+1 is twisted by . It is enough to show that the bimod-
ules Hci (Y) and Hci (Y) are isomorphic. However, the endomorphism F of
Y induces an K -linear automorphism of Hci (Y) (see Theorem A.2.7(c)). It is
easy to see (taking account of the commutation relations between F and the
elements of G and q+1 ) that F induces an isomorphism of (KG , K q+1 )-
bimodules Hci (Y) Hci (Y). The first equality is therefore proven.
We now turn to the second equality. Let g G . Then, by 3.1.6, we have

1
q + 1
R ( )(g ) = TrY (g , 1 ) ( ).
q+1

Therefore
1
q + 1
R ( ) (g ) = TrY (g 1 , 1 ) ( 1 ).
q+1

Now, by A.2.5, we have TrY (g , ) Z, therefore TrY (g , ) = TrY (g 1 , 1 ). It


follows that
1
q + 1
R ( ) (g ) = TrY (g , ) ( 1 ) = R ( )(g ),
q+1

as expected. 


4.1.2. The Character R (1)

We have, by definition,

R (1) = [Hc1 (Y)q+1 ]G [Hc2 (Y)q+1 ]G .


4.1 Definition and First Properties 39

Therefore, by 4.1.1 and A.2.3,

R (1) = [Hc1 (Y/q+1 )]G 1G .

Now Y/q+1 P1 (F) \ P1 (Fq ) (see Theorem 2.2.4). Therefore, by Theo-


rem A.2.6(a),

Hc (Y/q+1 )G = Hc (P1 (F))G Hc (P1 (Fq ))G .

Now, Hc (P1 (Fq ))G = [K [G /B]]G = 1G +StG and Hc (P1 (F)) = 21G (see A.3.1
and A.3.2). Hence

(4.1.4) R (1) = StG 1G

and therefore


1G if i = 2,
(4.1.5) [Hci (Y/q+1 )]G = StG if i = 1,


0 otherwise.

Although we have succeeded in calculating the character R (1), we still


have not obtained any new characters of G .

4.1.3. Dimensions

Let be a non-trivial element of q+1 . Then, by Theorem A.2.6(d),

TrY ( ) = TrY (1).

However Y = , therefore
TrY ( ) = 0.
We deduce from this that, as a character of q+1 , Hc (Y) is a multiple of the
character of the regular representation. So, for all (q+1 ) ,

(4.1.6) deg R ( ) = deg R (1) = q 1.

4.1.4. Cuspidality

The goal of this subsection is to show that, as soon as is a non-trivial linear


character of q+1 , the irreducible components of R ( ) are cuspidal. We first
start by a result which shows that Harish-Chandra induction is orthogonal
to Deligne-Lusztig induction.
40 4 Deligne-Lusztig Induction

Theorem 4.1.7. If and are characters of T q1 and q+1 respectively, then


R( ), R ( ) G = 0.

Proof. We may suppose that is a linear character. We have R (1) = 1G +


StG , R(1) = 1G + StG and therefore
R(1), R (1) G = 0. On the other hand,
if is a character of T not containing 1 as an irreducible factor then

R( ), R (1)
G = 0.
We may therefore suppose that = 1. In this case, R ( ) = [Hc1 (Y) KT
K ]G is a character (not only a virtual character) of G , therefore it suffices to
show the result when = regT . But,


R ( ), R(regT ) G =
R ( ), IndG
U 1U G = dimK (Hc (Y) K q+1 K ).
1 U

By A.2.3, we have

Hc1 (Y)U = Hc1 (Y/U) = Hc1 (A1 (F) \ {0}),

with the last equality following from Theorem 2.2.3. But, by A.3.4 and A.3.5,
we have
[Hc1 (A1 (F) \ {0})]q+1 = 1q+1 ,

and therefore dimK (Hc1 (Y)U K q+1 K ) = 0 as is a non-trivial linear char-


acter. 


Theorem 4.1.7 shows us that, if is a non-trivial linear character of q+1 ,


then the irreducible components of R ( ) are cuspidal. Moreover, equality
4.1.6 shows that R ( ) = 0. In order to obtain new irreducible characters of
G all that remains is to decompose R ( ).

4.2. Mackey Formula

The goal of this section is to prove the following theorem (compare 3.2.4).

Mackey formula. Let and be two elements of K0 (K q+1 ). We have

(4.2.1)
R ( ), R ( ) G =
, q+1 +
, q+1 .

R EMARK If is a linear character of q+1 , then = 1 . 

The rest of this section is dedicated to the proof of this Mackey formula.
We begin with a crucial geometric result determining the class of the mod-
ule Hc ((Y Y)/G ) in the Grothendieck group of q+1 q+1 -modules. We
(1) (2)
will need the following notation. We denote by q+1 (respectively q+1 ) the
q+1 q+1 -set with underlying set q+1 and such that, if q+1 and
4.2 Mackey Formula 41

( ,  ) q+1 q+1 , then ( ,  ) =  (respectively ( ,  ) = 1  ).


This defines two permutation K (q+1 q+1 )-modules. Then

(4.2.2) Hc ((Y Y)/G )q+1 q+1 = [K [q+1 ]]q+1 q+1 + [K [q+1 ]]q+1 q+1 .
(1) (2)

Proof (of 4.2.2). Set

Z = Y Y = {(x, y , z, t) A4 (F) | xy q yx q = 1 and zt q tz q = 1}.

Define
Z0 = {(x, y , z, t) Z | xt yz = 0}

and Z=0 = {(x, y , z, t) Z | xt yz = 0}.

Then Z0 and Z=0 are (G q+1 q+1 )-stable subvarieties of Z. By Theo-


rem A.2.6(a), we therefore have

( ) [Hc (Z/G )]q+1 q+1 = [Hc (Z0 /G )]q+1 q+1 + [Hc (Z=0 /G )]q+1 q+1 .

It is enough to show that the two terms on the right of ( ) correspond to the
two terms on the right of 4.2.2.
We begin with the first term. It is very easy to see that the morphism

q+1 Y Z0
( , x, y ) (x, y , x, y )

is an isomorphism of varieties. As Y/G A1 (F) (see Theorem 2.2.2), we


conclude that Z0 /G q+1 A1 , with the action of ( ,  ) q+1 q+1 on
( , x) q+1 A1 being as follows:

( ,  ) ( , x) = ( 1  , x).

Therefore, by Theorems A.2.1(c) and (f) and A.2.6(c), we have

Hc (Z0 /G ) = [K [q+1 ]]q+1 q+1 .


(2)
( )

We now study the quotient Z=0 /G . To this end, consider the variety

V = {(u, a, b) (A1 (F) \ {0}) A2 (F) | u q+1 ab = 1}

together with the morphism

: Z=0 V
(x, y , z, t) (xt yz, xt q yz q , x q t y q z).
42 4 Deligne-Lusztig Induction

A painstaking but easy calculation shows that does indeed take values
in V and that (g (x, y , z, t)) = (x, y , z, t) if (x, y , z, t) Z=0 and g G . To
show that induces an isomorphism : Z=0 /G V, all that remains it to
show the following four properties (see Proposition 2.2.1):
(a) is surjective.
(b) (x, y , z, t) = (x  , y  , z  , t  ) if and only if there exists g G such that
(x  , y  ) = g (x, y ) and (z  , t  ) = g (z, t).
(c) V and Z=0 are smooth varieties.
(d) If m Z=0 , then the differential dm is surjective.

Let us first prove (a) and (b). Taking into account Proposition 2.1.2, it is
enough to show that, if (u, a, b) V, then

| 1 (u, a, b)| = |G |.

Now, 1 (u, a, b) is the set of quadruples (x, y , z, t) satisfying




xy q yx q = 1 (1)




zt tz = 1
q q (2)
xt yz = u (3)



xt q yz q = a (4)


x q t y q z = b. (5)

The two equations (3) and (5), viewed as equations in t and z, form a system
of linear equations which, by (1), has determinant 1. As a consequence,
(x, y , z, t) 1 (u, a, b) if and only if (x, y , z, t) satisfy the system


xy q yx q = 1 (1)



q tz q = 1
zt (2)
z = ux bx
q (3 )



xt q yz q = a (4)


t = uy q by . (5 )

If we substitute z and t into (4), we obtain the equivalent system of equations


(because u = 0):

xy q yx q = 1 (1)



zt tz = 1
q q


(2)
z = ux q bx (3 )

q
q 2 yx q 2 = a + b (4 )

xy

uq
t = uy q by . (5 )
On the other hand, it is easy to verify that (1), (3), (4) and (5) imply (2). The
system is therefore equivalent to
4.2 Mackey Formula 43
q

xy yx q = 1 (1)


z = ux q bx (3 )
a + bq


2 2
xy q yx q = (4 )

uq

t = uy q by . (5 )

But, by Theorem 2.2.2, the number of couples (x, y ) satisfying (1) to (4) is
equal to |G |. Because a couple (z, t) is determined by (u, a, b) and (x, y ), we
indeed have | 1 (u, a, b)| = |G |, as expected.
We now turn to (c) and (d). Let m = (x0 , y0 , z0 , t0 ) Z=0 . The tangent space
T to Z=0 at m may be identified with

T = {(x, y , z, t) F4 | y0q x x0q y = 0 and t0q z z0q t = 0}.

Set (u0 , a0 , b0 ) = (m). The tangent space T  of the variety V at (m) may be
identified with

T  = {(u, a, b) F3 | u0q u b0 a a0 b = 0}.

Using these identifications it is easy to verify that the morphism dm : T


T  takes the following form:

dm (x, y , z, t) = (t0 x + x0 t y0 z z0 y , t0q x z0q y , x0q t y0q z).

It is enough to show that dm is injective. Now, if dm (x, y , z, t) = (0, 0, 0),


then in particular
q

y0 x x0q y = 0

t q z z q t = 0
0 0
t0q x z0q y = 0


q
x0 t y0q z = 0.
We indeed obtain x = y = z = t = 0 because, as m Z=0 , we have x0q t0q
y0q z0q = (x0 t0 y0 z0 )q = 0.
As (a), (b), (c) and (d) hold, the morphism : Z=0 V induces an iso-

morphism : Z=0 /G V and, under this isomorphism, the action of

( , ) q+1 q+1 is as follows:

( ,  ) (u, a, b) = (  u, 1 a, 1  b).

On the other hand, the torus F acts on V by the formula:

(u, a, b) = (u, a, 1 b),

and this action commutes with that of q+1 q+1 . As a consequence, by


Theorem A.2.6(e), we have
44 4 Deligne-Lusztig Induction

Hc (Z=0 /G )q+1 q+1 = Hc (V)q+1 q+1 = Hc (VF )q+1 q+1 .

Now VF = q+1 {0} {0}. Therefore

Hc (Z=0 /G )q+1 q+1 = [K [q+1 ]]q+1 q+1 .


(1)
( )

Equation 4.2.2 follows immediately from ( ), ( ) and ( ). 




In order to prove the Mackey formula, it will be sufficient to deduce some


algebraic consequences from the previous work, which was of a geometric
nature.

Proof (of the Mackey formula 4.2.1). We may and will suppose that and
are linear characters of q+1 . By Proposition 4.1.3, we have


R ( ), R ( ) G =
1G , R ( ) R ( ) G .

As a consequence,
   G

R ( ), R ( ) G = dimK Hc (Y) K q+1 K K Hc (Y) K q+1 K .

By Theorem A.2.6(b) and (c), we have


R ( ), R ( ) G = dimK Hc ((Y Y)/G ) K (q+1 q+1 ) K  ,

where  : q+1 q+1 K , ( ,  ) ( ) (  ). It therefore follows


from 4.2.2 that


R ( ), R ( ) G = dimK K [q+1 ] K (q+1 q+1 ) K 
(1)

(2)
+ dimK K [q+1 ] K (q+1 q+1 ) K  .

Set (1) = {( , 1 ) | q+1 } and (2) = {( , ) | q+1 }. Then


(i) q+1
K [q+1 ] = Indq+1
(i) 1 (i) .

Therefore, by Frobenius reciprocity,


q+1 q+1

R ( ), R ( ) G =
1 (1) , Resq+1
(1)  (1) +
1 (2) , Resq+1
(2)  (2) ,

and the result follows. 



4.3 Parametrisation of Irr G 45

4.3. Parametrisation of Irr G

The Mackey formula 4.2.1 together with 4.1.4 show that, if we denote by 0
the unique character of q+1 of order 2:
R (1) = 1G + StG .
R ( ) = R ( 1 ) Irr G if q+1
is such that 2 = 1.
R (0 ) = R+ (0 ) + R (0 ), where R (0 ) Irr G and R+ (0 ) = R (0 ).
  

If 2 = 1, 2 = 1 and  { , 1 }, then R ( ) = R ( ).
We obtain a formula analogous to 3.2.13:

(4.3.1) R (reg q+1 ) = 1G + StG + R+ (0 ) + R (0 ) + R ( ).


      2 =1
R (1) R (0 )

We have therefore obtained (q + 3)/2 cuspidal characters. In Chapter 3, we


obtained (q + 5)/2 non-cuspidal irreducible characters. As | Irr G | = q + 4, we
conclude that all the cuspidal characters of G have been obtained in the
cohomology of Y. Therefore we have

Irr G = {1G , StG } {R+ (0 ), R (0 ), R+ (0 ), R (0 )}


(4.3.2)
, 2 = 1}
{R( ) | q1 {R ( ) | q+1
, 2 = 1}.

If we denote by d the degree of R (0 ) , we have:

q 3  q + 1 2 q 1
|G | = 12 + q 2 + (q + 1)2 + 2 + (q 1)2 + d+
2
+ d2 .
2 2 2
Hence
(q 1)2
d+ + d = q 1 and 2
d+ + d2 = ,
2
which implies that

q1
(4.3.3) d+ = d = .
2
In the Exercises 4.3 and 4.4 we will give two other (more conceptual) proofs
of this last equality: these exercises imply that R+ (0 ) and R (0 ) are conju-
gate under G = GL2 (Fq ).
46 4 Deligne-Lusztig Induction

4.4. Action of the Frobenius Endomorphism

We will now study the eigenvalues of the action of F on Hci (Y). By Theo-
rem A.2.7(b), we have

(4.4.1) F = q on Hc2 (Y).

Taking this into account, in what follows we will only be interested in the
eigenvalues of the action of F on Hc1 (Y).
If is a linear character of q+1 , the KG -modules Hci (Y) K q+1 K and
i
Hc (Y)e are canonically isomorphic. Recall that

1
q + 1
e = ( 1 ) K q+1 .
q+1

The commutation relations between F and q+1 (and the fact that F is an
automorphism of Hci (Y) by Theorem A.2.7(c)) show that

(4.4.2) F (Hci (Y)e ) = Hci (Y)e 1 .

In particular, F stabilises Hc1 (Y)e1 and Hc1 (Y)e0 .

4.4.1. Action on Hc1 (Y)e1

The KG -module Hc1 (Y)e1 Hc1 (Y/q+1 ) is irreducible and therefore, by


Schurs lemma and the fact that the action of F commutes with that of
G , F acts on Hc1 (Y)e1 by multiplication by a scalar 1 . To calculate 1 , it
is enough to calculate the action of the Frobenius endomorphism F on
Hc1 (Y/q+1 ) = Hc1 (P1 (F) \ P1 (Fq )) (see Theorem 2.2.4). Now, the Lefschetz
fixed-point theorem (see Theorem A.2.7(a)) shows that
 F
0 = | P1 (F) \ P1 (Fq ) | = q 1 dimK Hc1 (Y/q+1 ) = q(1 1 ).

Hence

(4.4.3) 1 = 1.
4.4 Action of the Frobenius Endomorphism 47

4.4.2. Action on Hc1 (Y)e0

To simplify notation set V0 = Hc1 (Y)e0 . Denote by V0 the irreducible sub-


representation of V0 with character R (0 ). By Schurs lemma, F acts on
V0 by multiplication by a scalar . We would like to calculate .
Firstly, we have YF = and therefore, by the Lefschetz fixed-point for-
mula, we obtain

(q 1)(+ + )
0 = q q 1 Tr(F , Hc1 (Y)e ).
2 2 =1

But
Tr(F , Hc1 (Y)e ) = 0
2 =1

by 4.4.2. Hence

(4.4.4) = + .

To explicitly calculate + and , we will study the action of F 2 . As F 2


stabilises Hc1 (Y)e , it follows from Schurs lemma that F 2 acts on Hc1 (Y)e
by multiplication by a scalar (in fact, if = 0 , then this follows in fact
from 4.4.4 as + 2 = 2 ).

Theorem 4.4.5. Let (q+1 ) . Then

1 if = 1,
=
(1)q if = 1.

Proof. The equality 1 = 1 follows from 4.4.3. The Lefschetz fixed-point the-
orem shows that

|Y F | = q 2 q 1 (q 1) ( )
2

=1

for all q+1 . As a consequence, as 1 = 1, we have

|Y F | = q 2 1 (q 1)
2
( ) .

q+1

It then follows from Theorem 2.3.2 that



|Y F |
2
1 q2 if = 1,
(E ) ( ) = (q + 1)
q 1
=
q +1 if = 1.
q+1

The family of eigenvalues ( ) (q+1 ) is therefore a solution of the system


of linear equations (E ) q+1 . The determinant of this system is invertible
48 4 Deligne-Lusztig Induction

(being the determinant of the character table of a cyclic group) and therefore
it admits a unique solution. It remains only to verify that the solutions given
in the corollary are valid, which is routine. 

Corollary 4.4.6. We have = 0 (1)q.

4.4.3. Action on Hc1 (Y)e Hc1 (Y)e 1

Let q+1 be such that 2 = 1. Then F stabilises Hc1 (Y)e Hc1 (Y)e 1
and 2
 F acts as multiplication
 by (1)q. Therefore F has two eigenvalues,
(1)q and (1)q, each one having multiplicity q 1 (because,
as F (Hc1 (Y)e ) = Hc1 (Y)e 1 , the trace of F on the direct sum is zero).

Exercises

4.1* (Lusztig). In this exercise we show that



 R+ (0 ) si q 1 mod 4,
() R+ (0 ) =
R (0 ) si q 3 mod 4.

We will use Poincar duality, which is only possible using the (smooth) com-
pactification Y of Y. Denote by
, : Hc1 (Y)Hc1 (Y) Hc2 (Y) the perfect pair-
ing A.2.4.
(a) Show that, if = 1, then the KG -modules Hc1 (Y)e and Hc1 (Y)e are iso-
morphic (Hint: Use the open-closed exact sequence of Theorem A.2.1(d)).
To simplify notation set V0 = Hc1 (Y)e0 Hc1 (Y)e0 and denote by V0 the
irreducible submodule of V0 with character R (0 ).
(b) Show that
, induces a perfect KG -equivariant pairing between Hc1 (Y)e
and Hc1 (Y)e 1 . Deduce that R ( ) = R ( ).
We will denote by
, 0 the perfect pairing on V0 obtained by restriction
from
, .

1 if q 3 mod 4,
(c) Show that 0 (1) =
1 if q 1 mod 4.
(d) Suppose that q 1 mod 4. Show that V+0 is orthogonal to V0 . (Hint: Use
the F -equivariance of
, 0 and the knowledge of the eigenvalues of F ).
Deduced that
, 0 induces an isomorphism of KG -modules V+0 (V+0 )
(and therefore that R+ (0 ) = R+ (0 )).
4.4 Action of the Frobenius Endomorphism 49

(e) Suppose now that q 3 mod 4. Show that the restriction of


, 0 to
V+0 is zero (Hint: Use the F -equivariance of
, 0 and the knowledge of
the eigenvalues of F ). Deduce that
, 0 induces an isomorphism of KG -
modules V+0 (V0 ) (and therefore that R+ (0 ) = R (0 )).

R EMARK Part (c) shows that 0 (1) = 0 (1) (see the remark following
Exercise 3.3). 

4.2. Show that G has four irreducible characters of odd degree.

4.3. Denote by A = G /G Z(G


). Recall that A is of order 2 (see Exercise 3.1).
The group A acts on the conjugacy classes and on the irreducible characters
of G . In this exercise, we will give another proof of 4.3.3.
(a) Show that A stabilises the q semi-simple conjugacy classes of G and
permutes the other four without fixed points.
(b) Show that A stabilises the q characters 1G , StG , R( ) ( (q1 ) , 2 = 1)
and R ( ) ( (q+1 ) , 2 = 1). (Hint: Use the group G .)
(c) Deduce that A permutes the four characters R (0 ) and R (0 ) without
fixed points. (Hint: Apply some theorem of Brauer [Isa, Theorem 6.32].)
(d) Deduce that R+ (0 ) and R (0 ) are conjugate under G and that d+ =
d = (q 1)/2.

and deduce
4.4*. In this exercise, we define a Deligne-Lusztig induction for G
 
that the characters R+ (0 ) and R (0 ) are conjugate under G (which gives a
new proof of 4.3.3). Set

= {(x, y ) A2 (F) | (xy q yx q )q1 = 1}.


Y

F2 acts naturally on Y.
(a) Show that G We let F2 act on the right.
q q
=G
(b) Show that Y G Y = Y F2 . Here, A B denotes the quotient
q+1 q
of A B by the diagonal action of .
Set R ), [M ] i  0 (1)i [H i (Y)
 : K0 (K F2 ) K0 (K G M ] .
q F c KF G
q2 q2

F2
(c) Deduce from (b) that R  Resq = ResG R .
q+1 G
(d ) Show that

R  ( ) =
, +
, F .
 ( ), R
G F F
q2 q2

(Hint: Mimic the proof of the Mackey formula 4.2.1 for the group G .)
(e) Let 0 be an extension of 0 to Fq2 . Show that F 0 = 0 .
(f) Deduce that R  (0 ) is irreducible, that R (0 ) = ResG R
 (0 ), and that
G
R+ (0 ) and R (0 ) are conjugate under G .
(g) Show that d+ = d = (q 1)/2.
50 4 Deligne-Lusztig Induction

4.5. Let m be a non-zero natural number. Show that


Fm m
|Y | = |YF | + q + 1

and
m
q + 1 if m is odd,
Fm
|Y | = q m + 1 q(q 1)q m/2 if m 0 mod 4,

m
q + 1 (q 1)q m/2 if m 2 mod 4.
m
Verify that |YF | is divisible by |G |.

4.6. Let V = {(u, a, b) A3 (F) | u q+1 ab = 1} and denote by : Z V the


morphism defined by

(x, y , z, t) = (xt yz, xt q yz q , x q t y q z).

We also use the notation introduced in the proof of the Mackey formula (Z,
V, ,. . . ).
(a) Show that V is an open dense subset of V and that is a well-defined
extension of the morphism : Z=0 V.
(b) Show that V and Z are smooth.
(c) Show that is constant on G -orbits.
(d) Nevertheless, show that does not induce an isomorphism between
Z/G and V.
Chapter 5
The Character Table

Having parametrised the irreducible characters of the group G in the last


chapter, it is natural to turn to the question of determining their values on
the elements of G . For this, an important step is the calculation of the char-
acters of the bimodules K [G /U] and Hci (Y). For K [G /U], a little elementary
linear algebra is sufficient. For Hci (Y) it is necessary to invoke certain results
from Appendix A.
These calculations allow us to calculate the majority of the irreducible
characters of G , but it does not allow us to determine the values of R (0 )
and R (0 ). In the latter case, we study their restriction to U and utilise
certain elementary arithmetic results (on Gauss sums).

5.1. Characters of Bimodules

Denote by
Tr : G q1 K
(g , a) Tr((g , d(a)), K [G /U])

Tr : G q+1 K
and
(g , ) TrY (g , ).

The goal of this section is to calculate Tr and Tr .

5.1.1. Calculation of Tr

The bimodule K [G /U] is a permutation bimodule. Therefore, if (g , t) G


T , then
Tr((g , t), K [G /U]) = |{xU G /U | x 1 gx t 1 U}|.

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 51


DOI 10.1007/978-0-85729-157-8_5, Springer-Verlag London Limited 2011
52 5 The Character Table

From this we easily deduce the values of Tr(g , b), which are given in Ta-
ble 5.1.

Table 5.1 Values of Tr

 
x
I2 d(a) d ( )
0
{1} a q1 \ {1} q+1 \ {1} {1}, x Fq

b (q 2 1)b= (q 1)b{a,a1 } 0 (q 1)b=

In this and following tables, if P is a statement, then P has value 1 if P is


true and is 0 otherwise.

5.1.2. Calculation of Tr

Let (g , ) G q+1 . We write g = tu = ut, where t is of order prime to p


and u is of order a power of p (Jordan decomposition). By Theorem A.2.6(d),
we have
Tr (g , ) = TrY(t, ) (u).
Hence we must calculate the fixed points of (t, ) on Y: this is easy and the
result is given in the following lemma.

Lemma 5.1.1. We have




Y if 2 = 1 and t = I2 ,
(t, )
Y = q+1 v if 2 = 1 and t is conjugate to d ( ),


otherwise.

Here, v Y satisfies t v = 1 v .

Corollary 5.1.2. If is not an eigenvalue of g , then Tr (g , ) = 0.

Recall that u+ and u are representatives (in U) of the conjugacy classes


of non-trivial unipotent elements of G .

Lemma 5.1.3. Tr (u+ , 1) = Tr (u , 1) = (q + 1).

Proof. Set = Tr (u+ , 1). As u+ and u are conjugate under GL2 (Fq ), the ele-
ments (u+ , 1) and (u , 1) of G (see 2.5.1 for the definition of G ) are conjugate
under G . Therefore = Tr (u , 1). Now, taking into account 5.1.2, we have
5.2 Restriction to U 53


q 1 if g = 1,
2

R (reg q+1 )(g ) = if g is conjugate to u+ or u ,


0 otherwise.

Moreover, by Exercise 3.2,



q +1 if g = 1,
R(1)(g ) =
1 if g is conjugate to u+ or u .

But, by Theorem 4.1.7, we have


R(1), R (reg q+1 ) G = 0. In other words,

(q + 1)(q 2 1) + (q 2 1) = 0.

The result now follows easily. 




Thus we have obtained all necessary information to easily determine the


values of Tr (g ,  ), which are given in Table 5.2.

Table 5.2 Values of Tr

 
x
I2 d(a) d ( )
0
{1} a q1 \ {1} q+1 \ {1} {1}, x Fq

 (q 2 1)  = 0 (q + 1)  { , 1 } (q + 1)  =

5.1.3. The Characters R( ) and R ( )

Fix two linear characters and of q1 and q+1 respectively. Then the
values of R( ) and R ( ) are given in Table 5.3 (as follows, using 3.1.6 and
Tables 5.1 and 5.2).

5.2. Restriction to U

The information contained in Table 5.3 gives much of the character table
of G . The only characters which remain to be determined are R (0 ) and
R (0 ). For this, we will determine their restriction to U.
54 5 The Character Table

Table 5.3 Values of the characters R( ) and R ( )

 
x
I2 d(a) d ( )
0
{1} a q1 \ {1} q+1 \ {1} {1}, x Fq

R( ) (q + 1) ( ) (a) + (a1 ) 0 ( )

R ( ) (q 1) ( ) 0 ( ) ( )1 ( )

5.2.1. B-Invariant Characters of U

We fix a non-trivial linear character + of Fq+ . The morphism

Fq+ (Fq+ )
(5.2.1)
z (z + (zz  ))


is an isomorphism of groups. (It is enough to check injectivity, which follows


easily from the non-triviality of + .)
Denote by C the squares in Fq and let z0 Fq \ C . We set

+ : U K and : U K
u(z) + (cz) u(z) + (cz0 z).
cC cC

Taking into account Proposition 1.1.2(b), + and are B-invariant charac-


ters of U. Even better, one can easily verify the following lemma (recall that
is the subgroup of G
B = GL2 (Fq ) formed by upper triangular matrices).

Lemma 5.2.2. With notation as above, we have:


(a) (1U , + , ) is a Z-basis of K0 (KU)B .

(b) (1U , + + ) is a Z-basis of K0 (KU)B .

Note that

(5.2.3) + + = regU 1U .

5.2.2. Restriction of Characters of G

Firstly, we have the following.


5.2 Restriction to U 55

Proposition 5.2.4. Let and be linear characters of q1 and q+1 respectively.


Then
ResGU R( ) = regU +1U = 2 1U + + +


and U R ( ) = regU 1U = + + .
ResG

Proof. This follows from Table 5.3. 




Corollary 5.2.5. ResG


U StG = regU .

Denote by + = + u1 : U Fq . We have
 
1z
+ = + (z).
01

The previous proposition shows easily the next corollary.

Corollary 5.2.6.
R( ), IndG 
U + G =
R ( ), IndU + G = 1.
G

Up until now, the irreducible components of R (0 ) and R (0 ) of R(0 )


and R (0 ) have not yet been singled out. Corollary 5.2.6 gives us a possibil-
ity to do so.

Notation. From now on, we will denote by R+ (0 ) (respectively


R+ (0 )) the unique common irreducible component of R(0 ) (re-
spectively R (0 )) and IndG
U + .

Proposition 5.2.7. With the above choice, we obtain



U R (0 ) = 1U +
ResG and U R (0 ) = .
ResG

Proof. By construction, ResG U R+ (0 ) contains + and therefore contains

+ . As deg R+ (0 ) = deg + = (q 1)/2, we conclude the second equality.
We now turn to the first equality. Set = ResG U R (0 ) and write =
1U + + + . As R+ (0 ) and R (0 ) are conjugate under G , +
and are conjugate under B. Therefore + = . Moreover, + + = 2
by Proposition 5.2.4, therefore = 1. On the other hand, by construction,
+  1. As deg + = (q + 1)/2 = deg(1U + + ), we conclude that + = 1U +
+ . Similarly, = 1U + .  
56 5 The Character Table

5.2.3. Values of

In order to use the information gathered in the previous subsection, we will


calculate the values of the characters at the unipotent elements u . For
this we will need to use Gauss sums but, luckily, we will only need to con-
sider the simplest case. Let us temporarily define

= 0 (z)+ (z).
zFq

Then

(5.2.8) 2 = 0 (1)q.

Proof (of 5.2.8). We have

2 = 0 (zz  )+ (z + z  ).
z,z  Fq

Let z  = z 1 z  . We obtain

2 = 0 (z  )+ (z(1 + z  )).
z,z  Fq

Hence
 
2 = 0 (z  ) + (z(1 + z  ))
z  Fq zFq
 
= (q 1)0 (1) +

0 (z  ) + (z(1 + z  )) .
z  Fq \{1} zFq

But, if z  = 1, then

+ (z(1 + z  )) = 1 + + + (z(1 + z  )) = 1.
zFq zFq

Finally,

2 = (q 1)0 (1)

0 (z  ) = q 0 (1) 0 (z  ) = q0 (1),
z  Fq \{1} z  Fq

as expected. 

This allows us to choose a square root of 0 (1)q in K . We set

0 (1)q = 0 (z)+ (z).
zFq
5.3 Character Table 57

Then we have the following lemma.


 
1 0 (1)q 1 0 (1)q
Lemma 5.2.9. + (u ) = and (u ) = .
2 2
Proof. It is enough to note that + (u+ ) + (u+ ) = 1 (see 5.2.3) and that
+ (u+ ) (u+ ) = (by definition of ). 

5.3. Character Table

The previous section allows us to calculate the values of the characters


R (0 ) and R (0 ) at non-trivial unipotent elements. To complete the char-
acter table, we need only the following results.

Proposition 5.3.1. Let g G , q1 . Then:
and q+1
(a) R( )(g ) = (1)R( )(g ) and R ( )(g ) = (1)R ( )(g ).
(b) R (0 )(g ) = 0 (1)R (0 )(g ) and R (0 )(g ) = 0 (1)R (0 )(g ).
(c) If g is semi-simple, then R (0 )(g ) = 12 R(0 )(g ) and R (0 )(g ) = 12 R (0 )(g ).
Proof. (a) is clear and (b) follows immediately from (a). The last assertion fol-
lows from the fact that R+ (0 ) and R (0 ) (respectively R+ (0 ) and R (0 ))
are conjugate under G by Exercise 4.3 or 4.4. 
Having now collected all necessary information, the character table of
G is given in Table 5.4. To simplify notation in this table we have set q0 =
0 (1)q. In other words, q0 is the unique element of {q, q} such that q0 1
mod 4. In particular, as has already been pointed out in the remark following
Exercise 3.3, the numbers
1 q0
2
are algebraic integers.

Exercises

5.1. Show that, if (q1 ) and (q+1 ) , then

StG R( ) = IndG
T et StG R ( ) = IndG
T .

Conclude that
1 G 
StG = IndT 1T IndG
T  1T 
2

1 G 
and StG StG = IndT 1T + IndG
T  1T  .
2
58 5 The Character Table

Table 5.4 Character table of G = SL2 (Fq )

g I2 d(a) d ( ) u

{1} a Fq \ {1} q+1 \ {1} {1}, {}

q2 1
| ClG (g )| 1 q2 + q q2 q
2
o(g ) o( ) o(a) o( ) p o( )

CG (g ) G T T {I2 } U

1G 1 1 1 1

StG q 1 1 0

R( ), 2 = 1 (q + 1) ( ) (a) + (a)1 0 ( )

R ( ), 2 = 1 (q 1) ( ) 0 ( ) ( )1 ( )

(q + 1)0 ( ) 1 + q0
R (0 ), {} 0 (a) 0 0 ( )
2 2

(q 1)0 ( ) 1 + q0
R (0 ), {} 0 0 ( ) 0 ( )
2 2

5.2. Set
U .
+ = IndG +

Show that + is a multiplicity-free character of G and that


+ , + G = q + 1.

R EMARK The character + defined in this exercise is called a Gelfand-Graev


character of G . 

5.3. It was shown in Exercise 3.4 (respectively Exercise 4.1) that R+ (0 ) (re-
spectively R+ (0 )) is self-dual if and only if q 1 mod 4. Rediscover this
result by inspecting the character table.
= GL2 (Fq ).
5.4*. Calculate the character table of G
Part III
Modular Representations
In the next five chapters we study the modular representations of G . In
Appendix B we recall the necessary facts about modular representations of
abstract finite groups. For general reductive finite groups, the nature of
this study is radically different, depending on whether one is in unequal
characteristic (where one studies -modular representations, where  is a
prime number different from p) or in equal characteristic (where one studies
p-modular representations). Here p denotes the characteristic of the field of
definition of the group.
In unequal characteristic, the geometric methods of Deligne-Lusztig the-
ory remain as powerful as ever; this cohomology theory supplies represen-
tations not just over K , but over O and k as well (see the definition of O
and k in the notation below). It has also been observed that the family of
natural numbers d such that  divides d (q) and d (q) divides the order of
the group (d denotes the d-th cyclotomic polynomial) play a fundamental
role. In the case of our group G (whose cardinality is q1 (q)2 (q)) it will be
convenient to distinguish three cases:  = 2,  is odd and divides q 1,  is
odd and divides q + 1. The first case is the most subtle, because the Sylow
2-subgroup is non-abelian. The second case is the simplest and all O-blocks
of the group with non-trivial defect group are Morita equivalent to O-blocks
of N, or even to T for some of them (see Corollary 8.3.2). In the last case, all
non-principal O-blocks of G are Morita equivalent to blocks of N  , or even
of T  for some of them (see Propositions 8.1.4 and 8.2.2), while the princi-
pal O-block is Rickard equivalent to the principal O-block of N  . This can
be seen by algebraic methods, valid for all abstract finite groups (see [Ric1],
[Lin] and [Rou2]) but, in order to remain true to the spirit of this book and to
illustrate Deligne-Lusztig theory, we construct these equivalences with the
help of the cohomology Rc (Y, O) of the Drinfeld curve (with coefficients in
O). This construction was conjectured by Brou, Malle and Michel [BrMaMi,
Zyklotomische Heckealgebren, 1A] for an arbitrary reductive group, and
shown in the case of the group G = SL2 (Fq ) by Rouquier [Rou1] (see Corol-
lary 8.3.5). To be precise, Rouquier treats the case of the group GL2 (Fq ) but
his method can be easily adapted to the group SL2 (Fq ).
Finally, in Chapter 10, we give a brief account of the representations in
equal characteristic. The majority of this chapter is devoted to the construc-
tion of the simple modules. In accordance with the general theory of finite
reductive groups, all simple modules are obtained by restriction from ratio-
nal representations of the algebraic group G = SL2 (F).

N OTATION In the previous chapters, the irreducible characters of G have


been parametrised, using a prime number  different to p and an -adic field
K . Once this parametrisation has been obtained, the irreducible characters
may be viewed as KG -modules, where K is any sufficiently large field of
characteristic zero. We may also take K to be a finite extension of Qp .
In other words, if we do not make it explicit, we keep our field K , a suf-
ficiently large extension of Q , but we allow the possibility that  be equal to p.
We denote by O the ring of integers of K over Z and l denotes its maximal
ideal. Set k = O/l: it is a finite field of characteristic . If M is an O-module,
we denote by M its reduction modulo l, that is M = k O M. If m M, the
image of m in M will be denoted m.
Chapter 6
More about Characters of G and of its Sylow
Subgroups

The purpose of this chapter is to assemble some preliminary results (exclu-


sively concerning irreducible characters) which will be useful in the study of
modular representations. We describe central characters, congruences and
the character tables of normalisers of Sylow subgroups. We make use of this
information to verify the global McKay conjecture for G in all characteristics.

6.1. Central Characters

In order to determine the blocks of G , we will use Proposition B.1.5 applied


to Table 6.1 below, which gives the central characters of G .
Table 6.1 shows that we will need some results about congruences mod-
ulo l of cyclotomic numbers. To this end, denote by  (O) (respectively
 (O)) the group of roots of unity of O which are of order prime to  (re-
spectively of order a power of ) and denote by (O) the group of roots of
unity of O. Note that
(O) =  (O)  (O).
It is well known that

(6.1.1) the kernel of the canonical morphism (O) k is  (O).

We deduce the following result.

Lemma 6.1.2. Let a and b be two roots of unity in O such that a + a1 b + b 1


mod l. Then ab 1  (O) or ab  (O).
1 1
Proof. The hypothesis implies that a + a1 = b + b . Now, a a1 = b b = 1,
1
and therefore a {b, b }. The result then follows from 6.1.1. 


C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 63


DOI 10.1007/978-0-85729-157-8_6, Springer-Verlag London Limited 2011
64 6 More about Characters of G and of its Sylow Subgroups

Table 6.1 Central characters of G = SL2 (Fq )

g I2 d(a) d ( ) u

2 = 1 a q1 \ {1} q+1 \ {1} {1}, {}

q2 1
| ClG (g )| 1 q2 + q q2 q
2
o(g ) o( ) o(a) o( ) p o( )

CG (g ) G T T {I2 } U
q2 1
1G 1 q(q + 1) q(q 1)
2
StG 1 q+1 (q 1) 0
q1
R( ), 2 = 1 ( ) q( (a) + (a)1 ) 0 ( )
2
q + 1
R ( ), 2 = 1 ( ) 0 q( ( ) + ( )1 ) ( )
2

1 + q0
R (0 ), = 0 ( ) 2q 0 (a) 0 (q 1)0 ( )
2

1 + q0
R (0 ), = 0 ( ) 0 2q 0 ( ) (q + 1)0 ( )
2

Corollary 6.1.3. Let and be two linear characters of a finite abelian group
such that ( ) + ( )1 ( ) + ( )1 mod l for all . Then 1 or
is of order a power of .

6.2. Global McKay Conjecture

Here we verify the global McKay conjecture (stated in Appendix B) in order


to lay the groundwork for the study of more structural questions, like equiv-
alences of categories. To establish equivalences between certain blocks of G
and blocks of normalisers of Sylow subgroups, we will use Theorems B.2.5
and B.2.6. It is therefore necessary to understand the characters of such nor-
malisers. Recall that N (respectively N  , respectively B) is the normaliser of
a Sylow -subgroup when  is odd and divides q 1 (respectively  is odd
and divides q + 1, respectively  = p).
6.2 Global McKay Conjecture 65

6.2.1. Characters of N

We will always identify the groups T and q1 using the isomorphism d. If


T , we set
= IndN
T .

As s = 1 , a linear character of T is N-invariant if and only if 2 = 1.


Moreover, = 1 . Denote by 1N the trivial character of N, the linear
character of order 2 which is trivial on T , 0 the unique extension of 0

to N such that 0 (s) = 0 (1) (recall that s 2 = d(1)). Then Clifford
theory [Isa, Theorem 6.16] shows that

Irr N = {1N , , +0 , 0 } { | [T / ], 2 = 1}.

The character table of N is then easy to calculate and is given in Table 6.2
(we set s+ = s and s = sd(z0 ), where z0 is a non-square in Fq ).

Table 6.2 Character table of N

g I2 s d(a)

{1} {} a q1 \ {1}

q1
| ClN (g )| 1 2
2
o(g ) o( ) 4 o(a)

CN (g ) N
s T

1N 1 1 1

1 1 1

0 , {} 0 ( ) 0 (1) 0 (a)

, 2 = 1 2 ( ) 0 (a) + (a)1

In particular,

q +5
(6.2.1) | Irr N| = .
2
66 6 More about Characters of G and of its Sylow Subgroups

6.2.2. Characters of N 

The character table of N  is obtained in essentially the same way as for that
of N. If is a linear character of T  , we set

 = IndN
T .

As s = 1 , we conclude that  =  1 and that is N  -invariant if and
only if 2 = 1. Denote by 1N  the trivial character of N  ,  the unique linear
character of order 2 which is trivial on T  and 0 the unique extension of 0

with value 0 (1) on s  . Clifford theory [Isa, Theorem 6.16] shows that

Irr N  = {1N  ,  , +
0
, 0 } { | [(T  ) / ], 2 = 1}.

The character table of N  is then easily calculated and is given in Table 6.3
 = s  and s  = sd ( ), where is a non-square in
(we set s+ 0 0 q+1 ).

Table 6.3 Character table of N 

g I2 s d ( )

{1} {} q+1 \ {1}

q+1
| ClN  (g )| 1 2
2
o(g ) o( ) 4 o( )

CN  (g ) N
s T

1N  1 1 1

 1 1 1

0 , {} 0 ( ) 0 (1) 0 ( )

 , 2 = 1 2 ( ) 0 ( ) + ( )1

In particular,

q +7
(6.2.2) | Irr N  | = .
2
6.2 Global McKay Conjecture 67

6.2.3. Characters of B

Denote by 1+ (respectively 1 ) the trivial (respectively non-trivial) linear


character of Z . Recall that + is a non-trivial linear character of Fq+ and that
+ is the corresponding non-trivial linear character of U under the isomor-

phism u : Fq+ U. We denote by an irreducible component of the char-
acter defined in 5.2. If , {+, }, we set

B, = IndB
ZU 1  .

One may easily verify that

Irr B = { | T } {+,+
B
, +,
B
, ,+
B
, ,
B
}

and that the character table of B is given by Table 6.4 (recall that q0 =
0 (1)q).

Table 6.4 Character table of B

g I2 d(a) u 

{1} a Fq \ {1} ,  {+, }

q1
| ClB (g )| 1 q
2
o(g ) o( ) o(a) o( ) p

CB (g ) B T ZU

, T ( ) (a) ( )

(q 1)1 ( ) 1 +  q0
B, , , {} 0 1 ( )
2 2

In particular,

(6.2.3) | Irr B| = q + 3.
68 6 More about Characters of G and of its Sylow Subgroups

6.2.4. Normalisers of Sylow 2-Subgroups

As the Sylow 2-subgroups of G are non-abelian, we will be content to state


the following result, leaving its proof and the calculation of the character
table as an exercise (see Exercise 6.1).

(6.2.4) If S is a Sylow 2-subgroup of G , then | Irr2 NG (S)| = 4.

6.2.5. Verification of the Global McKay Conjecture

We can now verify the following result.

Proposition 6.2.5. If S is a Sylow -subgroup of G , then

| Irr G | = | Irr NG (S)|.

Proof. We may assume that  divides the order of G . Set

E = {R( ) | T , 2 = 1} and E  = {R ( ) | q+1



, 2 = 1}.

Note that
q 3 q 1
|E | = and |E  | = .
2 2
We have


{1 , St , R ( ), R ( )} if  = 2 and q 1 mod 4,
G G + 0 0


 
{1G , StG , R+ (0 ), R (0 )}
if  = 2 and q 3 mod 4,
Irr (G ) = {1G , StG , R+ (0 ), R (0 )} E if  is odd and divides q 1,



{1G , StG , R+ (0 ), R (0 )} E 

if  is odd and divides q + 1,


(Irr G ) \ {StG } if  = p.

It follows that


4 if  = 2,




q +5


if  is odd and divides q 1,
2
| Irr G | =




q +7

if  is odd and divides q + 1,

2


q + 3 if  = p.
6.2 Global McKay Conjecture 69

Moreover, when S is abelian, it follows from a theorem of Ito [Isa, Theorem


6.14] that | Irr NG (S)| = | Irr NG (S)| because  does not divide |NG (S)/S|. We
have therefore verified the global McKay conjecture thanks to 6.2.1, 6.2.2,
6.2.3 and 6.2.4.  

Exercises

6.1. Let S be a Sylow 2-subgroup of G . Calculate the character table of NG (S)


and verify 6.2.4.

6.2. Show that, if  divides |G | and is different from p, then StG is in the
principal -block of G .

R EMARK On the other hand, the Steinberg character StG is not in the
principal p-block of G (in fact, it is alone in its p-block by B.2.4). 

= GL2 (Fq ).
6.3. Verify McKays conjecture for G
6.4*. Determine, as a function of , the radical filtration of the kG -module
IndGB k.
Chapter 7
Unequal Characteristic: Generalities

Hypotheses. In this and the following two chapters, we assume


that  is a prime number different from p.

The purpose of this chapter is to assemble results valid in all unequal char-
acteristics. We determine, for example, the decomposition into O-blocks as
well as the Brauer correspondents. We also introduce modular (that is over
O, or even over Z ) and structural versions of Harish-Chandra and Deligne-
Lusztig induction. These are functors between categories, rather than being
simply maps between Grothendieck groups. The preliminary work on these
two functors will be useful in the next chapter, where we study the equiva-
lences of categories which they induce (which turn out to be either Morita
or derived equivalences).

7.1. Blocks, Brauer Correspondents

7.1.1. Partition in -Blocks

Using Table 6.1 (central characters) together with Corollary 6.1.3, we can
determine the partition of Irr G into -blocks. We will need the following
notation: we denote by T and T the maximal subgroups of T and T  of
order prime to , so that

T = S T and T  = S T .

These isomorphisms allow us to identify T and T  with S T and


S T 
 respectively. On the other hand, we identify T and q+1 via the

isomorphism d . If (respectively ) is a linear character of T (respec-

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 71


DOI 10.1007/978-0-85729-157-8_7, Springer-Verlag London Limited 2011
72 7 Unequal Characteristic: Generalities

tively T ), denote by B (respectively B ) the set formed by the irreducible


components of all R( ) (respectively R ( )), where S (respectively
S ). The following theorem describes the decomposition of Irr G into
-blocks (recall that  = p).

Theorem 7.1.1. If  = p, then:


(a) If T is such that 2 = 1, then B is an -block of G with defect group S .
(b) If T  
 is such that = 1, then B is an -block of G with defect group S .
2

(c) If  is odd and divides q 1, then:


(c1) {R+ (0 )} and {R (0 )} are -blocks of G with defect group 1 = S (note
that B 0 = {R+ (0 ), R (0 )}).
(c2) B1 and B0 are -blocks of G with defect group S (B1 being the principal
-block).
(d) If  is odd and divides q + 1, then:
(d1) {R+ (0 )} and {R (0 )} are -blocks of G with defect group 1 = S (note
that B0 = {R+ (0 ), R (0 )}).
(d2) B1 and B 0 are -blocks of G with defect group S (B1 being the principal
-block).
(e) If  = 2, then B1 B1 is the principal -block of G , with defect group a Sylow
2-subgroup of G (note that B1 B1 = {1G , StG }).

Proof. This follows from a careful inspection of Table 6.1, Corollary 6.1.3 and
the fact that q is invertible in O. 


7.1.2. Brauer Correspondents

Let (respectively ) be a linear character of T (respectively T ) such that


2 = 1 (respectively 2 = 1) and denote by A (respectively A ) the O-block
of G such that Irr KA = B (respectively Irr KA = B ). The principal O-
block of G will be denoted A1 or A1 . If moreover  is odd, we denote by A0
(respectively A0 ) the sum of the O-blocks of G satisfying Irr KA0 = B0
(respectively Irr KA0 = B 0 ).

R EMARK If  is odd and divides q 1, then A0 is an O-block with defect


group S while A0 is the sum of two O-blocks with defect group 1 = S .
If  is odd and divides q + 1, then A0 is the sum of two O-blocks with
defect group 1 = S while A0 is an O-block with defect group S .
If A is the sum of multiple O-blocks of G with the same defect group S,
we still use Brauer correspondent to refer to the sum of the associated Brauer
correspondents (i.e. a sum of O-blocks of NG (S)). 
7.1 Blocks, Brauer Correspondents 73

If (respectively ) is a linear character of T (respectively T ) such that


2 = 1 (respectively 2 = 1), we denote by b (respectively b ) the primitive
central idempotent of OT (respectively OT  ) equal to

1
b = (t) t 1
|T | tT 

1
|T | t 
(respectively b = (t  ) t 1 ).
T 

It is now straightforward to calculate the Brauer correspondents (the reader


is referred to Exercise 1.6 for the calculation of the normalisers of subgroups
of T and T  ). These are the object of the following theorem, where the corre-
spondents are given only for those O-blocks with non-central defect group.

Theorem 7.1.2. Let and be linear characters of T and T respectively. Let S
be a Sylow 2-subgroup of G .
(a) If  is odd and divides q 1, then:
(a1) If 2 = 1, then ONb is the Brauer correspondent of A .
(a2) If 2 = 1, then ON(b + b 1 ) is the Brauer correspondent of A .
(b) If  is odd and divides q + 1, then:
(b1) If 2 = 1, then ON  b is the Brauer correspondent of A .
(b2) If 2 = 1, then ON  (b + b 1 ) is the Brauer correspondent of A .
(c) If  = 2, then the principal O-block of NG (S) is the Brauer correspondent of
A1 = A1 . Moreover:
(c1) If q 1 mod 4 and = 1, then ON(b + b 1 ) is the Brauer correspon-
dent of A .
(c2) If q 3 mod 4 and = 1, then ON  (b + b 1 ) is the Brauer correspon-
dent of A .

R EMARK In statements (c1) and (c2) above, the condition = 1 (respec-


tively = 1) is equivalent to 2 = 1 (respectively 2 = 1): indeed, in (c), we
have  = 2, therefore and are of odd order. 

Proof. We will prove (a), the other cases are treated in a similar manner and
are left to the reader. Therefore suppose that  is odd and divides q 1. De-
note by i the primitive central idempotent in OG such that B = OGi . We
will show that 
b if 2 = 1,
BrS (i ) =
b + b 1 if 2 = 1.
74 7 Unequal Characteristic: Generalities

As CG (S ) = T , it is enough to calculate the coefficients of i on elements of


the group T . If t T , the coefficient of t in i is equal to
1  

1 1 + q 1 + (q + 1) (t) + (t 1 ) if = 1,

|G |

[S /]

=1

1  q +1 
2 0 (t) + (q + 1) 0 (t)( (t) + (t 1 )) if = 0 ,
|G |

2
[S /]

=1

q +1
|G | (t) + (t)

1
if 2 = 1.
S


This coefficient is therefore equal to



(q + 1)|S |

(t)tT if 2 = 1,

|G |



(q + 1)|S |
( (t) + (t 1 ))tT if 2 = 1,
|G |

which shows the desired result, because (q + 1)|S |/|G | = 1/q|T | 1/|T |
mod l (since q 1 mod l). 


7.1.3. Terminology

If (respectively ) is a linear character of T (respectively T ) such that


2 = 1 (respectively 2 = 1), we will say that the block A (respectively A )
is nilpotent. (The reader may verify that they are indeed nilpotent in the sense
of Brou and Puig [BrPu, definition 1.1].)
If  is odd, the (sum of) blocks A0 and A0 will be called quasi-isolated.
(One may verify that this agrees with the definition of a quasi-isolated semi-
simple element given in [Bon, 1.B] and the Jordan decomposition of blocks
of reductive groups in unequal characteristic [BrMi, Theorem 2.2].)
R EMARK The notion of nilpotent block is algebraic (as it is defined for any
block of any finite group) whereas the notion of quasi-isolated block is of
more geometric nature (as it is defined only for finite reductive groups and
requires Deligne-Lusztig theory). 
7.2 Modular Harish-Chandra Induction 75

7.2. Modular Harish-Chandra Induction

The functors of Harish-Chandra induction and restriction are defined by

R : Z T mod Z G mod
M Z [G /U] Z T M

R : Z G mod Z T mod
and
M Z [U\G ] Z G M.

If is a commutative Z -algebra, we denote by R and R the extension of


scalars to of the functors R and R respectively. For example, if = K , we
obtain the functors RK and RK defined in Section 3.2.

Proposition 7.2.1. The (Z G , Z T )-bimodule Z [G /U] is projective both as a left


Z G -module and as a right Z T -module.

Proof. Let eU = (1/q) uU u. Then eU is idempotent and belongs to Z U


Z G as q is invertible in Z . Now, Z [G /U] Z GeU as a (Z G , Z T )-bimodule.
The proposition follows, as Z G is free both as a Z G and Z T -module.  

Corollary 7.2.2. The functors R and R are adjoint.

We now turn to a study of the endomorphism algebra of the Z G -module


Z [G /U]. To this end we revisit a part of Exercise 3.3, this time working over
the ring Z rather than the field K . Firstly, note that the right action of T on
G /U induces a morphism of Z -algebras

Z T EndZ G Z [G /U]
t T (x x t 1 ).

We will view the algebra EndZ G Z [G /U] as a Z T -module via this mor-
phism.
On the other hand, denote by F the unique Z -linear endomorphism of
Z [G /U] such that
F (gU) = gusU
uU
 
0 1
for all g G . Recall that s = N NG (T ). It is easy to verify that, if
1 0
g G , t T and x Z [G /U], then

(7.2.3) F (g x t) = g F (x) s t.

In particular, F EndZ G Z [G /U]. Thanks to the endomorphisms induced


by T and F we may describe the endomorphism algebra EndZ G Z [G /U] as
a kind of quadratic extension of the group algebra Z T .
76 7 Unequal Characteristic: Generalities

Theorem 7.2.4. Set E = tT t Z T . Then:


(a) We have the following equality between elements of EndZ G Z [G /U]:

F 2 = qd(1) + F E .

Moreover, F t = s t F for all t Z T .


(b) If ,  Z T satisfies +  F = 0, then =  = 0.
(c) EndZ G Z [G /U] = Z T Z T F .
(d) If is a commutative Z -algebra, then EndG [G /U] = Z EndZ G Z [G /U].

Proof. (a) Let g G . Then

F 2 (gU) = gusvsU.
u,v U

Now, if a, b Fq , then

   
g d(1)U
  if b = 0,
1a 1b
g s sU = 1 a
01 01
g sd(b 1 )U if b = 0.
0 1

It follows that,
 
F 2 (gU) = qg d(1)U + gus t U.
uU tT

Hence the result.


We now turn to (b), (c) and (d). By using the isomorphism of (G , T )-
bimodules [G /U] GeU . The map

EndG GeU (eU GeU )opp


f f (eU )

is an isomorphism of -algebras. Now the set (eU neU )nN gives a -basis
of eU GeU by 1.1.7. Via this isomorphism, F corresponds to qeU seU . Now,
if t T and n N, we have (eU neU )(eU teU ) = eU nteU , which shows (b), (c)
and (d).  

In Chapter 8, we will use Theorem 7.2.4 to show that, if  is odd and


divides q 1, then EndZ G Z [G /U] Z N (see Theorem 8.3.1): this arithmetic
condition is related to Brous conjecture.
7.3 Deligne-Lusztig Induction* 77

7.3. Deligne-Lusztig Induction*

As G and q+1 act freely on the Drinfeld curve Y (see Propositions 2.1.2
and 2.1.3), it follows from Theorem A.1.5 that

(7.3.1) the complex Rc (Y, Z ) is left and right perfect,

that is, quasi-isomorphic to a bounded complex of (Z G , Z q+1 )-bimodules


projective as left and right modules. In order to maintain consistent nota-
tion we will identify q+1 and T  via the isomorphism d and view Y as a
variety equipped with an action of G (T  
F mon ), with F acting on T  as
conjugation by s  N  . In this way Rc (Y, Z ) will be viewed as a complex
of (Z G , Z T  )-bimodules, inducing a functor between the bounded derived
categories
R  : Db (Z T  ) Db (Z G )
C Rc (Y, Z ) Z T  C .
If is a commutative Z -algebra we denote by R the extension of scalars
to of the functor R  . On the other hand, as Y is a smooth curve, it follows
from A.1.4 that

(7.3.2) Hci (Y, Z ) is a free Z -module.

Recall also that, by Theorem A.2.1, we have

(7.3.3) Hc2 (Y, Z ) = Z (equipped with the trivial action of G )

and, by A.2.3, 7.3.2 and Theorem 2.2.2,

(7.3.4) Hc1 (Y, Z )G = 0.

From these properties we will deduce the following result.

Proposition 7.3.5 (Rouquier). There exist two projective Z G -modules P and Q


together with a morphism d : P Q satisfying the following properties.
(a) In the derived category Db (Z G ), we have an isomorphism
d
Rc (Y, Z ) D (0 P Q 0),

the term P being in degree 1.


(b) The Z G -module Q is the projective cover of Z (regarded as a trivial Z G -
module).
(c) The Z G -module P does not admit a subquotient isomorphic to Z (or, equiva-
lently, (KP)G = 0).

Proof. By 7.3.1 and 7.3.2, there exists projective Z G -modules P and Q to-
gether with a morphism of Z G -modules d : P Q such that Rc (Y, Z ) is
78 7 Unequal Characteristic: Generalities

quasi-isomorphic to the complex

d
C = (0 P Q 0).

Here, the terms P and Q are situated in degrees 1 and 2. By 7.3.3, we have
Q/ Im d Z . As Q is projective, we can write Q = P1 Q  , where P1 is the
projective cover of Z and Q  is contained in Im d.
The projectivity of Q  implies that the canonical surjective morphism
d (Q  ) Q  splits: we denote by Q  the Z G -submodule of P which is
1

the image of this section. It is isomorphic to Q  via the restriction of d.


On the other hand, the Z -module P/Q  is torsion-free: if x P satisfies
x Q  , then d(x) Q  and therefore d(x) Q  as Q/Q  is torsion-free. This
shows that x d 1 (Q  ) and in conclusion we remark that d 1 (Q  )/Q 
d 1 (Q  ) Ker d is torsion-free.
As P/Q  is torsion-free, the exact sequence of Z G -modules 0 Q 
P P/Q  0 splits over Z , and therefore splits as a sequence of Z G -
modules as Q  is projective, therefore relatively (G , 1)-projective and there-
fore relatively (G , 1)-injective [CuRe, Theorem 19.2]. Denote by P  a comple-
ment of Q  in P and denote by d  : P  P1 the composition of the morphism
P  Q P1 , the last arrow being the projection Q = P1 Q  P1 . The com-
plex
d
0 P Q 0
is therefore homotopic to a complex

d
0 P  P1 0.

This shows that we may (and will) assume that

Q = P1 .

All that remains is to verify the third assertion. We have an exact sequence
of Z G -modules
d
0 Hc1 (Y, Z ) P Q Hc2 (Y, Z ) 0,

which, after tensoring with the flat Z -algebra K , yields a short exact se-
quence
d
0 Hc1 (Y) KP
K
KQ Hc2 (Y) 0,
where dK : KP KQ is the canonical extension of d. Because the algebra KG
is semi-simple, the functor invariants under G is exact, which gives us a
final exact sequence

0 Hc1 (Y)G = 0 (KP)G (KQ)G Hc2 (Y)G = K 0.


7.3 Deligne-Lusztig Induction* 79

Now, (KQ)G = K as Q is the projective cover of Z , which shows property


(c). 

Encouraged by Proposition 7.3.5, we will study the endomorphism alge-
bra of Rc (Y, Z ), viewed as an object in the derived category Db (Z G ). We
will commence with the following important result, which will also prove
useful in the construction of equivalences of derived categories in the fol-
lowing chapter.
Corollary 7.3.6 (Rouquier). If i = 0, then

HomDb (Z G ) (Rc (Y, Z ), Rc (Y, Z )[i]) = 0.

Moreover, the Z -module EndDb (Z G ) Rc (Y, Z ) is torsion-free.


Proof. We keep the notation (P, Q, d) used in the statement of Proposi-
tion 7.3.5 and denote by C the complex

d
C = (0 P Q 0).

As P and Q are projective Z G -modules, we have

HomDb (Z G ) (Rc (Y, Z ), Rc (Y, Z )[i]) = HomKb (Z G ) (C , C [i]).

We begin with the first assertion. In virtue of the previous equality, it


suffices to show that
HomKb (Z G ) (C , C [i]) = 0
for all i = 0. As the result is evident for |i|  2, we assume that |i| = 1.
Suppose that i = 1. Then HomCb (Z G ) (C , C [1]) HomZ G (P, Q). There-
fore, let HomZ G (P, Q) and view as a morphism of complexes

d
C 0 P Q 0

d
C [1] 0 P Q 0.

Suppose first that is not contained in the image of d. Then induces a


surjection KP Hc2 (Y) K . It follows that (KP)G = 0, which is impossible
by Proposition 7.3.5(c).
Therefore Im Im d. The projectivity of P then allows one to construct
a morphism  : P P such that = d  . This shows that the morphism
of complexes is homotopic to zero.
Now suppose that i = 1. To give a morphism of complexes : C
C [1] is the same as giving a morphism of Z G -modules : Q P such
that Im Ker d and Im d Ker :
80 7 Unequal Characteristic: Generalities

d
C 0 P Q 0

d
C [1] 0 P Q 0.

Therefore factorises to give a morphism Q/ Im d P and hence to a mor-


phism Z P. This last morphism is necessarily zero by virtue of Proposi-
tion 7.3.5(c). Therefore = 0.
We now turn to the last assertion, concerning the absence of torsion in the
endomorphism algebra Rc (Y, Z ). So let EndCb (Z G ) (C ) be such that 
is homotopic to zero. We have to show that itself is homotopic to zero. Let
: P P and : Q Q be the morphisms of Z G -modules such that is
equal to the following morphism

d
C 0 P Q 0

d
C 0 P Q 0.

By the hypotheses, there exists a morphism of Z G -modules : Q P such


that  = d and  = d . As P and Q are torsion-free it is enough to
show that Im P.
Denote by : Q P/P the composition of : Q P with the canoni-
cal projection P P/P. Then is zero on Im d, and therefore factorises
to give a morphism : Hc2 (Y, Z ) P/P. If is non-zero, then P/P con-
tains an F G -submodule isomorphic to F . But P/P = F P is a projective
F G -module. Therefore if it contains the trivial module in its socle, then it
contains the projective cover of the trivial module as a direct summand. This
implies that P admits the Z G -module Q as a direct summand, and therefore
P G = 0, contradicting Proposition 7.3.5(c). Therefore = 0 and = 0, which
concludes the proof of the corollary.  
Denote by Hc (Y, Z ) the graded Z G -module i  0 Hci (Y, Z ) (in fact, we
have Hc (Y, Z ) = Hc1 (Y, Z ) Hc2 (Y, Z )). The previous proposition has the
following consequence.
Corollary 7.3.7 (Rouquier). The natural morphism of O-algebras

EndDb (Z G ) Rc (Y, Z ) Endgr
Z G Hc (Y, Z )

is injective (here, Endgr


? () denotes the algebra of graded endomorphisms).

Proof. By Corollary 7.3.6, the Z -algebra EndDb (Z G ) Rc (Y, Z ) injects natu-
rally into EndDb (Q G ) Rc (Y, Q ). But as Q G is semi-simple, this last algebra
is isomorphic to Endgr
Q G Hc (Y, Q ). The commutativity of the diagram
7.3 Deligne-Lusztig Induction* 81

EndDb (Z G ) Rc (Y, Z ) Endgr


Z G Hc (Y, Z )

EndDb (Q G ) Rc (Y, Q ) Endgr


Q G Hc (Y, Q )

implies the corollary. 




Having established these preliminary results, we now turn to the ques-


tion of determining the endomorphism algebra EndDb (Z G ) Rc (Y, Z ), using
similar techniques to those used in the determination of EndZ G Z [G /U]. The
right action of T  on Y commutes with the left action of G , and we therefore
have a canonical morphism of Z -algebras Z T  EndDb (Z G ) Rc (Y, Z ). Us-
ing this morphism we view EndDb (Z G ) Rc (Y, Z ) as a Z T  -module. On the
other hand, the Frobenius endomorphism F of the Drinfeld curve induces
an endomorphism, denoted F  , of the complex Rc (Y, Z ). As the action of
G commutes with that of F , this yields an element of EndDb (Z G ) Rc (Y, Z ).
It follows from the relations between the actions of F , G and T on Y that

(7.3.8) F  (g , t  ) = (g , t 1 ) F 

for all (g , t  ) G T  . The following theorem describes the structure of the


algebra EndDb (Z G ) Rc (Y, Z ) in an analogous fashion to the description of
EndZ G Z [G /U] in Theorem 7.2.4.

Theorem 7.3.9 (Rouquier). Set E  = t  T  t  OT  . Then:


(a) We have the following equality between elements of EndDb (Z G ) Rc (Y, Z ):

F 2 = qd (1) + F  E  .

Moreover, F  t  = s t  F  for all t  Z T  .
(b) If ,  Z T  satisfy +  F  = 0, then =  = 0.
(c) We have
EndDb (Z G ) Rc (Y, Z ) = Z T  Z T  F  .
(d) If is a commutative Z -algebra, then

EndDb (G ) Rc (Y, ) Z EndDb (Z G ) Rc (Y, Z ).

Proof. (a) By virtue of Corollary 7.3.7, it is enough to verify this equality on


the cohomology groups Hc1 (Y, Z ) and Hc2 (Y, Z ). But, by 7.3.2, it is enough
to verify this equality on the cohomology groups Hc1 (Y, K ) = Hc1 (Y) and
Hc2 (Y, K ) = Hc2 (Y). This is done using the calculation of the eigenvalues of F
and F 2 completed in Section 4.4. We have:
82 7 Unequal Characteristic: Generalities

On Hc2 (Y), q+1 acts trivially (therefore E  acts as multiplication by q + 1)


and F  acts as multiplication by q (see 4.4.1).
On Hc1 (Y)e1 , q+1 acts trivially (therefore E  acts as multiplication by q +
1) and F  acts as multiplication by 1 (see 4.4.3).
On Hc1 (Y)e0 , an element q+1 acts as multiplication by 0 ( ) (there-
fore E  acts as multiplication by 0) and F 2 acts as multiplication by
q 0 (1) (see 4.4.5).
If is a linear character of q+1 such that 2 = 1, then, on Hc1 (Y)(e +
e 1 ), an element q+1 acts as multiplication by ( ) (therefore E 
acts as multiplication by 0) and F 2 acts as multiplication by q (1)
(see 4.4.5).
The result follows immediately from these observations.
(b) If +  F  = 0, then this equality is still true on Hc1 (Y) and Hc2 (Y). If
T  , we denote by its linear extension to the group algebra KT  .
On Hc2 (Y), +  F  acts as multiplication by 1( )+q 1(  ). On Hc1 (Y)e1 , +
F  acts as multiplication by 1( ) 1(  ). As a consequence, 1( ) = 1(  ) = 0.


On the space Hc1 (Y)e0 , and  act as multiplication  ) and 0 (  )


by 0 (

respectively, while Fhas two eigenvalues,  0 (1)q and 0 (1)q.
Hence, 0 ( ) + 0 ( ) 0 (1)q = 0 ( ) 0 ( ) 0 (1)q = 0, which im-
plies that 0 ( ) = 0 (  ) = 0.
We now study the action of +  F  on Hc1 (Y)(e + e 1 ) if 2 = 1.
Let (v1 , ... , vr ) be a basis of Hc1 (Y)e . Then (F  (v1 ), ... , F  (vr )) is a basis of
Hc1 (Y)e 1 . In the basis (v1 , ... , vr , F  (v1 ), ... , F  (vr )) of Hc1 (Y)(e + e 1 ), the
matrix of F  has the form
 
0 (1)qIr
F 
Ir 0

while an element  of KT  has matrix

 diag( (  ), ... , (  ), (  ), ... , (  )).


    
r times r times

The equality +  F  = 0 therefore implies ( ) = (  ) = 0.


Hence we have shown that ( ) = (  ) = 0 for all linear characters of
T . Hence, =  = 0.


(c) Set A = EndDb (Z G ) Rc (Y, Z ) and A = Z T  Z T  F  A. We want to


show that A = A. Note first that, by Proposition 7.3.6, the O-modules A and
A are torsion-free. On the other hand, we claim that KA = K A. As K is flat
over Z , K A = EndDb (KG ) Rc (Y, K ), so it follows from (b) that it is enough to
show that
dimK EndDb (KG ) Rc (Y, K ) = 2 |T  |.
But
7.3 Deligne-Lusztig Induction* 83

dimK EndDb (KG ) Rc (Y, K ) =


[Hc1 (Y)]KG , [Hc1 (Y)]KG G
+
[Hc2 (Y)]KG , [Hc2 (Y)]KG G .

As
[Hc1 (Y)]KG , [Hc2 (Y)]KG G = 0, it is enough to show that


R (reg q+1 ), R (reg q+1 ) KG = 2 |T  |,

which follows immediately from the Mackey formula 4.2.1. We have there-
fore shown the claim, namely that KA = K A.
On the other hand, if we set, for K A,
1  
T ( ) = Tr( , Hc2 (Y)) Tr( , Hc1 (Y)) K.
|G |

The key to the proof of (c) is the following lemma.

Lemma 7.3.10. If A, then T ( ) Z . Moreover, the restriction of T to a


morphism T : A Z is a symmetrising form.

Proof (of Lemma 7.3.10). As the complex Rc (Y, Z ) is perfect as an object of


Db (Z G ) (see 7.3.1), it suffices to show that, if P is a projective Z G -module
and if EndZ G (P), then Tr( , KP)/|G | O. To show this we may assume
that P is indecomposable. Then there exists an idempotent e OG such that
P Z Ge and we may assume that P = Z Ge. If we denote by : Z G Z
the symmetrising form such that (g ) = g =1 for all g G , then it is easy to
verify that Tr( , KP) = |G | ( (e)). This shows the first claim.
We turn to the second claim. Denote by B the O-basis of A obtained by
the concatenation of (t  )t  T  and (t  F  )t  T  and set M = (T (bb ))b,b B . It
is enough to show that det M Z . Now, for all t1 , t2 T  , we have

0 if t1 t2 = 1,
 
T (t1 t2 ) = 1
if t1 t2 = 1,
q
 
T (t1 t2 F  ) = |Yt1 t2 F | = 0
and 
 1 2 0 if t1 t21 = I2 ,
T (t1 F  t2 F  ) = |Yt1 t2 F | =
1 if t1 t21 = I2 .
Indeed, the first equality follows from Table 5.2, while the second and third
follow from the Lefschetz fixed-point theorem combined with, respectively,

2.3.1 and Theorem 2.3.2. Hence, det(M) = (1/q)|T | O , as required.  

We now show why (c) follows from Lemma 7.3.10. Let (a1 , ... , ad ) be a
Z -basis of A and let (a1 , ... , ad ) be the dual Z -basis of A for the form T . Let
84 7 Unequal Characteristic: Generalities

a A. Then, by (), there exists 1 ,. . . , d K such that a = 1 a1 + + d ad .


By duality we obtain i = T (aai ). Now aai A as A is an algebra and so
i = T (aai ) O by Lemma 7.3.10, therefore a A. Hence A = A.
(d) follows immediately from the fact that Rc (Y, Z ) is perfect. 

In Chapter 8, we will use the previous theorem to show that, if  is odd
and divides q + 1, then EndZ G Rc (Y, Z ) Z N  (see Theorem 8.3.4).

Exercises

7.1. Complete the proof of 7.1.2.


7.2. We identify the Z G -modules Z [G /B] with the space of T -invariants
(for the right action) in Z [G /U]. Show that F restricts to an endomorphism
of Z [G /B]. We will denote by FB this restriction.
Denote by I the identity endomorphism of Z [G /B]. Show that FB2 = qI +
(q 1)FB and that

EndZ G Z [G /B] = Z I Z FB .

7.3 (Howlett-Lehrer, Dipper-Du). Denote by U the subgroup of G formed


by lower triangular matrices. Let R be a ring in which p is invertible. Iden-
tify the (RG , RT )-bimodules R[G /U] and R[G /U ]) with RGeU and RGeU
respectively. Show that the map

RGeU RGeU
a aeU

is an isomorphism of (RG , RT )-bimodules. (Hint: As both bimodules are


free over R of the same rank, it is sufficient to show surjectivity, which fol-
lows from the formula

qeU eU eU = eU + eU ueU
uU \{I2 }

and the fact that U \ {I2 } TUsU.)


7.4. Let R be a commutative ring in which p is invertible. If : T R is
a linear character, we denote by R the RT -module defined as follows: the
underlying R-module is R itself and an element t T acts as multiplication
by (t). Use the previous exercise (or the automorphism F , extended to the
ring R) to show that R[G /U] RT R R[G /U] RT R 1 .
7.5. Show that, if V is a Z T -module, then R(R(V )) V s V , where s V
denotes the Z T -module with underlying Z -module V but on which an el-
ement t T acts as s t = t 1 .
Chapter 8
Unequal Characteristic: Equivalences of
Categories

Hypothesis. In this chapter, as in the next and the previous


chapters, we assume that  is a prime number different from p.

The purpose of this chapter is to verify Brous abelian defect conjecture (see
Subsection B.2.2). In the case of non-principal blocks (which all have abelian
defect groups), the equivalences of categories predicted by Brous conjec-
ture are always Morita equivalences (see Sections 8.1 and 8.2). While it is
possible to obtain this result using Brauer trees and Brauers theorem B.4.2,
we give instead a concrete construction of these equivalences using Harish-
Chandra and Deligne-Lusztig induction. In the case of principal blocks,
treated in Section 8.3, the situation is more interesting. If  is odd and di-
vides q 1, then the principal block is Morita equivalent to its Brauer corre-
spondent and Harish-Chandra induction induces an equivalence. If  is odd
and divides q + 1, then the principal block is Rickard equivalent to its Brauer
correspondent and Deligne-Lusztig induction induces the required equiva-
lence. If  = 2, the situation is more complicated: when q 3 mod 8, then
the principal block of G is Rickard equivalent to its Brauer correspondent;
when q 1 mod 8, the derived category of the principal block is equiva-
lent to the derived category of an A -algebra. These two final results are due
to Gonard [Go].
The last section is dedicated to Alvis-Curtis duality, viewed as an end-
ofunctor of the homotopy category Kb (Z G ) or of the derived category
Db (Z G ). For an arbitrary finite reductive group it was shown by Cabanes
and Rickard [CaRi] that this duality is an equivalence of the derived cate-
gory. More recently, Okuyama [Oku3] improved this result by showing that
it was in fact an equivalence of the homotopy category (see also the work
of Cabanes [Ca] for a simplified treatment of Okuyamas theorem). We give
a very concrete proof of Okuyamas result in the case of our small group
G = SL2 (Fq ).

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 85


DOI 10.1007/978-0-85729-157-8_8, Springer-Verlag London Limited 2011
86 8 Unequal Characteristic: Equivalences of Categories

8.1. Nilpotent Blocks

8.1.1. Harish-Chandra Induction

Let be a linear character of T such that 2 = 1. We will show the following.

Proposition 8.1.1. Harish-Chandra induction RO induces a Morita equivalence


between A and OTb .

Proof. Denote by f the primitive central idempotent of OG such that A =


OGf . Set M = O[G /U]b . Then M is an (OG , OTb )-bimodule. Moreover,
the irreducible factors of the KG -module KM are the irreducible charac-
ters of KA . Moreover, f M = M and therefore M is in fact an (A , OTb )-
bimodule, which is projective as a left and right module by Proposition 7.2.1.
We would like to show that the functor M O Tb is an equivalence of cat-
egories.
By virtue of Brous theorem B.2.5, it is enough to show that the functor
KM KTb induces a bijection between Irr KTb and Irr KA . As Irr(KTb ) =
{ | S }, this amounts to showing that Irr KA = {R( ) | S }
which is precisely the definition of A . 

Corollary 8.1.2. The O-algebras A and ON(b + b 1 ) are Morita equivalent.

Proof. By Proposition 8.1.1, it is enough to show that the O-algebras A =


ON(b + b 1 ) and A = OTb are Morita equivalent. Set P = ONb . Then
P is an (A, A )-bimodule which is projective as a left and right module. We
would like to show that the functor P A is an equivalence of categories.
By virtue of Brous theorem B.2.5, it is enough to show that the func-
tor KP KA induces a bijection between Irr KA and Irr KA. However this
follows from the (easily verified) fact that Irr KA = { | S } and
[KP KA K ]KN = for all S . 


C OMMENTARY When S is not contained in Z (which occurs when  is


odd and divides q 1 or when  = 2 and q 1 mod 4), then NG (S ) = N
by Exercise 1.6 and ON(b + b 1 ) is the Brauer correspondent of A (by
Theorem 7.1.2). As a consequence, Corollary 8.1.2 shows that Brous con-
jecture (see Appendix B) is verified in a stronger form. Indeed, the predicted
equivalence of derived categories is induced by a Morita equivalence.
Another consequence of Theorem 8.1.1 is that

(8.1.3) A Matq+1 (OTb ) Matq+1 (OS ),

which agrees with structure theorems for nilpotent blocks (for the case of an
abelian defect group, see for example [BrPu, 1]). 
8.1 Nilpotent Blocks 87

8.1.2. Deligne-Lusztig Induction*

Let be a linear character of T  such that 2 = 1. We will show the following


result.
 induces a Morita equivalence
Proposition 8.1.4. Deligne-Lusztig induction RO
 
between A and OTb .

Proof. Denote by f the primitive central idempotent of OG such that A =


OGf . Set M  = Hc1 (Y, O)b . Then M  is an (OG , OT  b )-bimodule and the
irreducible factors of the KG -module KM  are the irreducible characters of
KA . Furthermore, f M  = M  and therefore M  is in fact an (A , OT  b )-
bimodule. We would like to show that the functor M  O T  b is an equiv-

alence of categories.
We first show that M  is projective as a left and right module. By 7.3.1, the
complex f Rc (Y, O)b is perfect as a left and right complex. Its i-th coho-
mology group is f Hci (Y, O)b , which is non-zero only when i = 1. Therefore
f Rc (Y, O)b is quasi-isomorphic to a complex of bimodules consisting of
one non-zero term f Hc1 (Y, O)b = M  occurring in degree 1. Therefore M  is
projective as a left and right module.
By virtue of Brous theorem B.2.5, it is enough to show that the functor
KM  KT  b gives a bijection between Irr KT  b and Irr KA . As Irr(KT  b ) =

{ | S }, this amounts to saying that Irr KA = {R ( ) | S },
which is precisely the definition of A .  

Corollary 8.1.5. The O-algebras A and ON  (b + b 1 ) are Morita equivalent.

Proof. The proof follows the same lines as that of Corollary 8.1.2, the goal
being to show that the (ON  (b + b 1 ), OT  b )-bimodule ON  b induces a
Morita equivalence.  

C OMMENTARY When S is not contained in Z (which occurs when  is


odd and divides q + 1 or when  = 2 and q 3 mod 4), then NG (S ) = N 
by Exercise 1.6 and ON  (b + b 1 ) is the Brauer correspondent of A (by
Theorem 7.1.2). We conclude that Corollary 8.1.5 shows Brous conjecture
(see Appendix B) in a stronger form. Indeed, the predicted equivalence of
derived categories is induced by a Morita equivalence.
Another consequence of Theorem 8.1.4 is that

(8.1.6) A Matq1 (OT  b ) Matq1 (OS ),

which agrees with structure theorems for nilpotent blocks (for the case of an
abelian defect group, see for example [BrPu, 1]). 
88 8 Unequal Characteristic: Equivalences of Categories

8.2. Quasi-Isolated Blocks

Hypothesis. In this and only this section we assume that  is


odd (and different from p).

This hypothesis is necessary for the existence of quasi-isolated blocks.

8.2.1. Harish-Chandra Induction

In this subsection we will show the following result.

Proposition 8.2.1. Harish-Chandra induction RO induces a Morita equivalence


between the O-algebras A0 and ONb0 .

Proof. Denote by q a root of q in O (recall that O is sufficiently large). Set
M = O[G /U]b0 . As in the proof of Proposition 8.1.1, M is an (A0 , OTb0 )-
bimodule. Moreover, by 7.2.3, M is stable under the endomorphism F . De-

note by S the restriction of F / q to M. As the element E = tT t acts as
multiplication by 0 on M, we have

S 2 = d(1)

by Theorem 7.2.4(a). Therefore, by 7.2.3, we can extend the right OT -module


structure on M to an ON-module structure, letting s act as the automor-
phism S . We may then view M as an (A0 , ONb0 )-bimodule.
As N/T is of order 2 and  is odd, it follows from Proposition 7.2.1 that M
is projective as a left and right module. By virtue of Brous Theorem B.2.5,
it is enough to show that EndA0 M ONb0 (via the canonical morphism).
This fact is a consequence of Theorem 7.2.4(c).  

C OMMENTARY If  is odd and divides q 1, then ONb0 is the Brauer


correspondent of A0 (see Theorem 7.1.2), and Proposition 8.2.1 shows that
Brous conjecture (see Appendix C) is true in a stronger form: the predicted
equivalence of derived categories is induces by a Morita equivalence. 

8.2.2. Deligne-Lusztig Induction*

In this subsection we will show the following result.

Proposition 8.2.2. Deligne-Lusztig induction R  induces a Morita equivalence be-


tween the Z -algebras A0 and ON  b 0 .
8.3 The Principal Block 89

Proof. The proof is similar to that of Proposition 8.2.1. First, we define


M  = Hc1 (Y, O)b 0 . It is an (A0 , OT  b 0 )-bimodule which is projective as a
left and right module by the same argument as in the proof of Proposi-
tion 8.1.4. We may then give it the structure of an (A0 , ON  b 0 )-bimodule,

letting s  act as F  / q (by 7.3.8 and Theorem 7.3.9(a)). This bimodule is
still projective as a left and right module, because N  /T  is of order 2 and 
is odd. It then results from Theorem 7.3.9(c) that EndA (M  ) ON  b 0 . We
0
may then conclude, thanks to Theorem B.2.5. 


C OMMENTARY If  is odd and divides q + 1, then ON  b 0 is the Brauer cor-


respondent of A0 (see Theorem 7.1.2), and Proposition 8.2.2 shows Brous
conjecture (see Appendix B) in a stronger form. Indeed, the predicted equiv-
alence of derived categories is induced by a Morita equivalence. 

8.3. The Principal Block

As described in the introduction to this chapter, the case of the principal


block is the most interesting. Here we give a proof of the results described
in the introduction when  is odd. When  = 2, we will content ourselves
with stating the results of Gonard [Go] without proof. In the richest case for
which we give a complete treatment, that is when  is odd and divides q + 1,
the equivalence predicted by Brous conjecture will be constructed using
the complex of cohomology Rc (Y, O). For this we will make crucial use
of the results of the previous chapter, most notably the description of the
algebra of endomorphisms of this complex.

8.3.1. The Case when  is Odd and Divides q 1

We begin by showing a result of a global nature which refines Theo-


rem 7.2.4.

Theorem 8.3.1. If  is odd and divides q 1, then EndZ G Z [G /U] Z N.

Proof. As q 1 mod  and  is odd, the polynomial X 2 q is split over F .



Therefore, by Hensels lemma, there exists an element + q of Z such that

+ q 1 mod Z and ( + q) = q. Set
2


+ q 1
1 1
S =
+ q
(F E ) (1 E ),
2 ( + q + 1)(q + 1)

and recall that E = tT t Z T . As  is odd and + q 1 mod Z , S be-

longs to EndZ G Z [G /U].
90 8 Unequal Characteristic: Equivalences of Categories

On the other hand a tedious but straightforward calculation shows that


S 2 = d(1) and S t = t 1 S for all t T . As a consequence, the right Z T -
module structure on Z [G /U] may be extended
to a right Z N-module struc-
+
ture by letting s act as S . Moreover, 1 ( +q+1)(q+1) E (OT ) because, if
q1


Z , then the inverse of the element 1 + E OT is 1 1+(q1) E OT .
Theorem 7.2.4(c) then shows that EndZ G Z [G /U] = Z T Z T S Z N, the
isomorphism being induced by the right action of Z N on Z [G /U].  
Corollary 8.3.2. If  is odd and divides q 1, then Harish-Chandra induction in-
duces a Morita equivalence between the O-algebras [T /] A and ON.
Proof. Set
A= A .
[T /]

It follows from 3.2.13 that the irreducible factors of K [G /U] are exactly the
elements of Irr KA. The result then follows from Theorem 8.3.1 and Brous
theorem B.2.5.  
C OMMENTARY When  is odd and divides q 1 the Morita equiva-
lence of the previous corollary induces an equivalence between A and its
Brauer correspondent, after cutting by block idempotents. Thanks to Corol-
lary 8.3.2, we rediscover when = 1 the results of Corollary 8.1.2 and Propo-
sition 8.2.1 (which were valid in all unequal characteristic) by a more direct
route. The reader may verify that the Morita equivalences constructed in
these corollaries agree with those constructed in Corollary 8.3.2. 
Corollary 8.3.3. If  is odd and divides q 1, then Harish-Chandra induction in-
duces a Morita equivalence between the principal O-block of G and that of N.
R EMARK Even when  is odd and divides q 1, the O-blocks A0 and
OTb0 are not Morita equivalent. Similarly, the O-blocks A1 and OTb1
are not Morita equivalent. There is even no equivalence of derived cate-
gories, because | Irr KA? | = |S | + 1 = |S | = | Irr KTb? | when ? {1, 0 } (see
Remark B.2.7). 

8.3.2. The Case when  is Odd and Divides q + 1*

We begin by showing a result of a global nature which refines Theo-


rem 7.3.9.
Theorem 8.3.4. If  is odd and divides q + 1, then EndDb (Z G ) Rc (Y, Z ) Z N  .
Proof. The proof is analogous to that of Theorem 8.3.1. Indeed, set

1 1  +
q 1
S = (F 
E ) (1 E  ),
+
q 2 ( q + 1)(q + 1)
+
8.3 The Principal Block 91

where we recall that E  = t  T  t  Z T  and + q denotes a square root

of q in Z such that + q 1 mod Z (this root exists because q 1
mod ).
One then shows that S 2 = d (1) and S  t  = t 1 S  for all t  T  .
Thanks to Theorem 7.3.9(c) this allows us to show that

EndDb (Z G ) Rc (Y, Z ) Z N  ,

the isomorphism being obtained by letting s  N  act through the automor-


phism S  . 


Corollary 8.3.5 (Rouquier). If  is odd and divides q + 1, then Deligne-Lusztig


induction induces a Rickard equivalence between [T  /] A and ON  .

Proof. The proof consists of two steps. In the first step we construct, using
the building blocks at our disposal, a complex of (OG , ON  )-bimodules. This
is the most delicate step. The second step consists of showing that this com-
plex verifies the conditions of Theorem B.2.6, which is almost a formality
using the results at our disposal.
First step: construction of the complex. By Appendix A (section A.1), the com-
plex Rc (Y, O) is a well-defined element in the category of complexes of bi-
modules Cb (OG , OT  
F mon ). Moreover, it is homotopic, in the category
Kb (OG , OT  ), to a bounded complex C of (OG , OT  )-bimodules of finite
type.
After eliminating direct factors homotopic to zero, we may suppose that

C is a complex of (OG , OT  )-bimodules without a direct factor homotopic to zero.

This property has the following consequence.

Lemma 8.3.6. In the algebra EndCb (O G ,O T  ) C the two-sided ideal of morphisms


homotopic to zero is contained in the radical of EndCb (O G ,O T  ) C .

Proof (of Lemma 8.3.6). Denote by A = EndCb (O G ,O T  ) C , I the two-sided


ideal of morphisms which are homotopic to zero, and R the radical of A . If
I is not contained in R, then (I + R)/R is a non-zero two-sided ideal in the
finite-dimensional semi-simple algebra A /R. In particular, (I +R)/R con-
tains a non-zero idempotent e of A /R. The theorem on lifting idempotents
[The, Theorem 3.1 (h)] implies that there exists an idempotent e I whose
image in A /R is e. It follows that eC is a direct factor of C and e : eC eC
is the identity, and therefore eC is homotopic to zero. This contradicts our
hypothesis. Hence I is contained in R as claimed.  

Denote by f : C Rc (Y, O) and f  : Rc (Y, O) C two morphisms of


complexes which are mutually inverse in the category Kb (OG , OT  ). Denote
by S  the element OT  
F mon which is equal to
92 8 Unequal Characteristic: Equivalences of Categories

 1  1  +
q 1
S = (F E ) (1 E  ),
+
q 2 ( q + 1)(q + 1)
+

by analogy with the proof of Theorem 8.3.4.


We then define S  : C C by

S  = f  S  f ,

where S  is viewed as a morphism S  : Rc (Y, O) Rc (Y, O). Then, if


(g , t  ) G T  , we have

(g , t  ) S  = S  (g , s t  ).

Moreover, as shown during the proof of Theorem 8.3.4, the element S 


2

d (1) in EndCb (O G ,O T  ) C is homotopic to zero. Set h = d (1)S 2 IdC . Then

S 2 = d (1)(1 + h),

where h EndCb (O G ,O T  ) C is homotopic to zero. By Lemma 8.3.6, h belongs


to the radical of EndCb (O G ,O T  ) C and we may therefore extract a square root

of 1 + h thanks to the standard formal series for 1 + X . This series con-
verges because  = 2. Therefore there exists h EndCb (O G ,O T  ) C which is
homotopic to zero and such that (1 + h )2 = 1 + h.
We now define  : C C by  = S  (1 + h ). Then 2 = d (1), which
shows that we can equip C with the structure of a complex of (OG , ON  )-

bimodules, letting s  act by  . The equality  t  = s t   may be easily
verified (for t  T  ).
Second step: verification of the criteria of Theorem B.2.6. Let us define

A = A
[T  /]

and denote by e  the primitive central idempotent of OG such that A =


OGe  . Set
C  = e C .
Then C  is a complex of (A , ON  )-bimodules which is quasi-isomorphic to
C because e  acts as the identity on cohomology groups by 7.3.2. By Corol-
lary 7.3.6, we have
HomDb (A ,O N  ) (C  , C  [i]) = 0
for all i = 0. Moreover, every irreducible character of KA is an irreducible
factor of H (K C  ) = Hc1 (Y) Hc2 (Y) and the natural morphism

ON  (EndDb (O G ) C  )opp
8.3 The Principal Block 93

is an isomorphism by Theorem 8.3.4. The proof of the corollary is now com-


plete by virtue of Theorem B.2.6. 


Corollary 8.3.7. If  is odd and divides q + 1, then Deligne-Lusztig induction in-


duces a Rickard equivalence between the principal O-block of G and that of N  .

R EMARK Even when  is odd and divides q + 1, the O-blocks A0 and
OT  b 0 are not Morita equivalent. Similarly, the O-blocks A1 and OT  b1
are not Morita equivalent. There is even no equivalence of derived cate-
gories, because | Irr KA? | = |S | + 1 = |S | = | Irr KT  b? | where ? {1, 0 } (see
Remark B.2.7). 

8.3.3. The Case when  = 2*

In this subsection we will content ourselves to mention without proof two


results of Gonard [Go]. Denote by S a Sylow 2-subgroup of G . Even though
S is not abelian, the following result remains valid.

Theorem 8.3.8 (Gonard). If  = 2 and q 3 mod 8 the principal O-blocks of


G and NG (S) are Rickard equivalent.

C OMMENTARY If  = 2 and q 3 mod 8, recall that |S| = 8 and that


|NG (S)| = 24 by Theorem 1.4.3. In particular, S/Z Z/2Z Z/2Z is abelian
and is a Sylow 2-subgroup of G /Z . As predicted by Brous conjecture, it is
possible to construct a Rickard equivalence between the principal blocks of
G /Z and NG /Z (S/Z ) = NG (S)/Z . Moreover, this equivalence may be lifted
to give the Rickard equivalence of Theorem 8.3.8.
To be precise, Gonard works with the group G = GL2 (Fq ), from which it
is easy to deduce Theorem 8.3.8. 

Set
A = EndDb (O G ) (O[G /U] Rc (Y, O)).
This algebra is an A -algebra (that is to say, an algebra equipped with higher
products mn : A n A verifying certain compatibility conditions). In his
thesis, Gonard [Go, 4.2] gives a complete description of this A -algebra.
For a review of A -algebras, the reader is referred to [Go, 4.1].

Theorem 8.3.9 (Gonard). If  = 2, then the derived category Db (OG ) is equiva-


lent to the derived category of the A -algebra A . This A -algebra satisfies

dimO A = 4(q + 1) and mn = 0 if n  4.


94 8 Unequal Characteristic: Equivalences of Categories

8.4. Alvis-Curtis Duality*

I would like to thank Marc Cabanes for suggesting this section to me, as well
as for having provided a simple argument for the proof of the main result
(Theorem 8.4.1).

Hypothesis. In this section we fix a commutative ring R in


which p is invertible.

R EMARK Q , Z , F , as well as K , O and k are rings in which p is invert-


ible. 

This hypothesis implies that the idempotent eU can be viewed as an ele-


ment of the group algebra RG . Recall that the (RG , RT )-bimodules R[G /U]
and RGeU are isomorphic. Denote by

: RGeU RT eU RG RG
a RT b ab.

It is a morphism of (RG , RG )-bimodules. We set

: RG RGeU RT eU RG
a ageU RT eU g 1
g [G /B]

the dual morphism. One may easily verify that is indeed a morphism of
(RG , RG )-bimodules (indeed, it is enough to verify that (g ) = g (1) =
(1)g for all g G ).
Denote by D the complex of (RG , RG )-bimodules
 
D = 0 RGeU RT eU RG RG 0

and set
 
D = 0 RG RGeU RT eU RG 0 ,
the dual complex. Even though this will not influence the principal result of
this section we suppose, in accordance with standard conventions, that the
non-zero terms of D (respectively D ) occur in degree 0 and 1 (respectively
1 and 0). If M is a (RG , RG )-bimodule, we also use the notation M for the
complex 0 M 0, where M is in degree 0.

Theorem 8.4.1. In the homotopy category Kb (RG , RG ), we have

D RG D K D RG D K RG .
8.4 Alvis-Curtis Duality* 95

Proof. To simplify notation set e = eU and M = RGe RT eRG . Furthermore,


we will write for the tensor product RT . We have
 f g 
D RG D = 0 M RG (RGe eRGe eRG ) M 0 ,

where  
f (a b) = ab a bge eg 1
g [G /B]

and g (x (a b c)) = (x) ab c.

The endomorphism F of the RG -module RGe defined by F (x) = qxese is


an isomorphism. Indeed, the calculation completed over Z in Theorem 7.2.4
remains valid here and shows that

F (F E ) = qd(1).

We will need the following lemma.

Lemma 8.4.2. The morphism

M M RGe eRGe eRG


(x x  ) (y y  ) x e x  + F 1 (y ) ese y 

is an isomorphism of (RG , RG )-bimodules.

Proof (of Lemma 8.4.2). By Bruhat decomposition 1.1.4, we have eRGe =


eRBe eR[BsB]e (as (RT , RT )-bimodules). As a consequence,
   
RGe eRGe = RGe eRBe RGe eR[BsB]e .

It is enough to show that the morphisms

RGe RGe eRBe RGe RGe eR[BsB]e


and
x x e y F 1 (y ) ese

are isomorphisms of (RG , RT )-bimodules. This is a straightforward conse-


quence of the fact that, in both cases, the inverse is given by the formula
a b ab. 


Using the isomorphism of Lemma 8.4.2, the complex D RG D is isomor-


phic (in the category Cb (RG , RG )) to the complex

f g
0 M RG M M M 0,

where
f  (m) = (m) m ( (m)) m
96 8 Unequal Characteristic: Equivalences of Categories

and g  (a m m ) = (a) m m .

This uses the fact that (a b) = a b + qase es 1 b.


Now consider
: RG M M RG M M
a m m a (m) m ( (a) m m ).

Then is an isomorphism of (RG , RG )-bimodules, which shows that the


complex D RG D is isomorphic (in the category Cb (RG , RG )) to the com-
plex
f  g 
0 M RG M M M 0,
where f  (m) = 0 m 0 and g  (a m m ) = m . In other words,

D RG D C RG C C [1],
Id
where C is the complex 0 M M
M 0 in which the non-zero terms are
concentrated in degrees 0 and 1. As C (and therefore C [1]) is homotopic to
zero, we deduce that D RG D is homotopic to RG .
The fact that D RG D is also homotopic to RG is shown in exactly the
same fashion.  
Corollary 8.4.3. The functor D RG induces an auto-equivalence of the homo-
topy category Kb (RG , RG ) (as well as of the derived category Db (RG , RG )).

Exercises

8.1. If T (respectively T  ), show that OTb OS (respectively


OT  b OS ).
Now suppose that  is odd. Show that the O-algebras ONb1 and ONb0
are isomorphic, and similarly for the O-algebras ON  b1 and ON  b 0 .
8.2. Show that, if  divides |G |/(q 1), then the algebras EndZ G Z [G /U] and
Z N are not isomorphic.
Show that, if  divides |G |/(q + 1), then the algebras EndDb (Z G ) Rc (Y, Z )
and Z N  are not isomorphic.

8.3*. If R is a field of characteristic different from p, the equivalence of homo-


topy categories D RG induces an isomorphism between the Grothendieck

groups that we will denote DR : K0 (RG ) K0 (RG ). Show that

DR [M ] = [RR ( RR M)] [M ].

Calculate DK .
Chapter 9
Unequal Characteristic: Simple Modules,
Decomposition Matrices

Hypothesis. In this chapter, as in the two previous chapters, we


suppose that  is a prime number different from p.

In this chapter we will determine the simple kG -modules and the decom-
position matrix Dec(OG ), as a function of the prime number . This study
will be carried out block by block, or more precisely by type of block (nilpo-
tent, quasi-isolated, principal). When the defect group is cyclic we will also
give the Brauer tree. We refer the reader to Appendix B for the definitions of
these concepts.
To carry out this study, we will use the equivalences of categories con-
structed in the previous chapter. When these equivalences are Morita equiv-
alences the simple modules correspond to one another and the decomposi-
tion matrices are preserved (as is the Brauer tree). When the equivalences
in question are genuine Rickard equivalences (which only occurs for the
principal block when  is odd and divides q + 1) the only property that is
conserved is the number of simple modules.

9.1. Preliminaries

9.1.1. Induction and Decomposition Matrices

By extension of scalars from to K to k, Harish-Chandra induction R (re-


spectively Deligne-Lusztig induction R  ) induces two functors RK and Rk
(respectively RK and Rk ). These functors induce linear maps, denoted R
and Rk (respectively R and Rk ) between the Grothendieck groups of T (re-

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 97


DOI 10.1007/978-0-85729-157-8_9, Springer-Verlag London Limited 2011
98 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

spectively T  ) and G :

R : K0 (KT ) K0 (KG ),
Rk : K0 (kT ) K0 (kG ),
R : K0 (KT  ) K0 (KG )
and Rk : K0 (kT  ) K0 (kG ).

Because the functor R (respectively R  ) is induced by an OG -module which


is O-free (respectively a complex of OG -modules which are O-free), the di-
agrams

R
K0 (KT ) K0 (KG )

(9.1.1) decO T decO G

Rk
K0 (kT ) K0 (kG )

and

R
K0 (KT  ) K0 (KG )

(9.1.2) decO T  decO G

Rk
K0 (kT  ) K0 (kG )

are commutative.

9.1.2. Dimensions of Modules and Restriction to U

In 5.2.1 we introduced a non-trivial linear character + of the additive group


+
Fq+ . We denote by k : Fq+ k the composition Fq+ O  k . If x Fq
we denote by x the linear character of U (with values in k ) defined by

x (u(z)) = k (xz)

for all z Fq . We set


1
ex = x (u1 )u kU.
q uU
9.2 Nilpotent Blocks 99

If a Fq , then

(9.1.3) d(a)ex d(a1 ) = ea2 x

(see Proposition 1.1.2(b)). Consequently, if M is a kG -module, then

(9.1.4) d(a)ex M = ea2 x M.

On the other hand, as the order of U is invertible in k, we have

(9.1.5) e0 M = M U and M = ex M.
xFq

The next proposition follows from these observations.

Proposition 9.1.6. Let M be a kG -module and let x and y be two elements of Fq


such that xy is not a square. Then

q 1
dimk M = dimk M U + (dimk ex M + dimk ey M).
2
In particular,

q 1
(9.1.7) dimk M dimk M U mod .
2

9.2. Nilpotent Blocks

Let (respectively ) be a linear character of T (respectively T ) such


that 2 = 1 (respectively 2 = 1). By Proposition 8.1.1 (respectively 8.1.4),
the bimodule O[G /U]b (respectively Hc1 (Y, O)b ) induces a Morita equiv-
alence between A and OTb (respectively between A and OT  b ). By ex-
tension of scalars we deduce that tensoring with k[G /U]b (respectively
Hc1 (Y, k)b ) induces a Morita equivalence between kA and kTb (respec-
tively kA and kT  b ). Now, kTb (respectively kT  b ) has only one simple
module k (respectively k ), that is to say the k-vector space k on which T

(respectively T  ) acts via the linear character T O  k (respectively

T  O  k ).
As a consequence,

Irr kA = {Rk k } = {IndG


B k }

and Irr kA = {Hc1 (Y, k) kT  k }.


100 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

Note that | Irr kA | = | Irr kA | = 1, as should be the case for any nilpotent
block. The decomposition matrices are given by

1 1
1 1

Dec(A ) = . and Dec(A ) = . ,
.. ..
1 1

where the number of lines of Dec(A ) (respectively Dec(A )) is |S | (respec-


tively |S |) because
Irr KA = {R( ) | S }

and Irr KA = {R ( ) | S }.

Brauer trees*. The only non-exceptional character of A (respectively A ) is


R( ) (respectively R ( )). The Brauer trees are therefore given by

TA i y

and

TA i y.

9.3. Quasi-Isolated Blocks

Hypothesis. In this and only this section we suppose that  is


odd (and different from p).

We recall that this hypothesis is necessary for the existence of quasi-isolated


blocks.

First, by Proposition 8.2.1 (respectively 8.2.2), the (A0 , OTb0 )-bimodule


O[G /U]b0 (respectively the (A0 , OT  b 0 )-bimodule Hc1 (Y, O)b 0 ) can be
extended to an (A0 , ONb0 )-bimodule (respectively to an (A0 , ON  b 0 )-
bimodule), and the latter induces a Morita equivalence. The action of the ele-

ment s of N (respectively s  of N  ) is given by F / q (respectively F  / q).
Now, the k-algebra kNb0 (respectively kN  b 0 ) admits two simple mod-
ules k0 (respectively k0 ) associated to the linear characters 0 (respectively
0 ) of N (respectively N  ) defined in Section 6.2. Via the Morita equivalence
they correspond to two simple modules of kA0 (respectively kA0 ) that we
9.4 The Principal Block 101

denote by R (0 ) (respectively R  ( )). By 9.1.1 (respectively 9.1.2), these


0
are the reductions modulo l of O-free OG -modules admitting R (0 ) (re-
spectively R (0 )) as characters. As

Irr KA0 = {R+ (0 ), R (0 )} {R(0 ) | [S / ], = 1}

and Irr KA0 = {R+ (0 ), R (0 )} {R(0 ) | [S / ], = 1},

we deduce that
Irr kA0 = {R+ (0 ), R (0 )},

that Irr kA0 = {R+


 
(0 ), R (0 )}

and that the decomposition matrices are given by



10 1 0
0 1 0 1

1
Dec(A0 ) = 1 1 and Dec(A0 ) = 1 .
.. .. .. ..
. . . .
11 11

Here, the first two lines of Dec(A0 ) (respectively Dec(A0 )) correspond to


the characters R (0 ) (respectively R (0 )). The number of lines of Dec(A0 )
(respectively Dec(A0 )) is equal to (|S | + 3)/2 (respectively (|S | + 3)/2).

Brauer trees*. The non-exceptional characters of A0 (respectively A0 ) are


R (0 ) (respectively R (0 )). Therefore, the Brauer trees are given by
i y i
TA0
and

TA i y i.
0

9.4. The Principal Block

9.4.1. Preliminaries

We identify Rk k with k[P1 (Fq )] and, via this identification, we denote by


StkG the k-vector subspace of k[P1 (Fq )] equal to
102 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

StkG = { l l | l = 0}.
lP1 (Fq ) lP1 (Fq )

We denote by v the element

v=
1
l k[P1 (Fq )].
lP (Fq )

Then kv and StkG are kG -submodules of k[P1 (Fq )], with kv being isomorphic
to the trivial module. We have

(9.4.1) decO G StG = [ StkG ]kG .

Moreover, as |P1 (Fq )| = q + 1, we deduce that

(9.4.2) kv StkG if and only if  divides q + 1.

We finish with the following result.

Proposition 9.4.3. If  = 2, then kv and StkG are the only two non-trivial kG -
submodules of k[P1 (Fq )].

Proof. We begin with some notation. If z Fq , we set lz = [z; 1] P1 (Fq ) and


l = [1; 0]. We then have

s lz = lz 1 and u(x) lz = lz+x

for all x Fq . Note that, with the usual conventions, the two previous for-
mulas make sense and are valid even when z {0, }.
If x Fq , we set
vx = k (xz)lz .
zFq

Then  
k[P1 (Fq )] = kv kvx .
xFq

The action of the subgroup U is entirely determined by the following for-


mulas:

() e0 v = v and ex vy = x=y vy

for all x, y Fq .
Let M be a non-trivial kG -submodule of k[P1 (Fq )]. If U acts trivially on M,
then G acts trivially on M because G is generated by the conjugates of U (see
Lemma 1.2.2). In this case, M = kv and the proposition holds. We may there-
fore suppose that U (i.e. G ) does not act trivially on M. As a consequence,
there exists an element x Fq such that
9.4 The Principal Block 103

(1) ex M = 0.

In other words, vx M by (). Again by (), we have

1  xy 
ey s vx = ey l +
q k z+
z
vy .
zFq

But, by Exercise 9.1 (which is easy) and the fact that  = 2, there exists y Fq
such that xy is not a square in Fq and
 xy 
k z+
z
= 0.
zFq

As a consequence,

(2) ey M = 0.

On the other hand, the injective morphism of kG -modules M  IndG


B k,
which, by adjunction, furnishes a non-trivial morphism of kB-modules
B M k. Hence,
ResG

(3) e0 M = 0.

It then follows from (1), (2), (3) and Proposition 9.1.6 that dimk M  q.
Now set M  = M StkG . Then dimk M   q 1  2 and again M  is a kG -
submodule of k[P1 (Fq )]. As its dimension is at least 2, G does not act trivially
on M  and the previous argument applies again to M  . We then obtain that
dimk M   q and, as M  StkG , we conclude that M  = StkG , that is to say StkG
M.  

9.4.2. The Case when  is Odd and Divides q 1

Hypothesis. In this and only this subsection we assume that 


is odd and divides q 1.

This case may be studied in a similar manner to the quasi-isolated block A0 .


As  does not divide q + 1, we deduce from 9.4.2 that

k[G /B] = StkG kv .


104 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

Hence the two simple modules of kA1 which correspond to the two simple
kN-modules k1 and k via the Morita equivalence are kv and StkG :

Irr kA1 = {k, StkG }.

As
Irr KA1 = {1G , StG } {R( ) | [S / ], = 1},
we conclude that the decomposition matrix is

10
0 1


Dec(A1 ) = 1 1 .
.. ..
. .
11

Here, the two first lines are indexed by 1G and StG and the number of lines
is (|S | + 1)/2.
Brauer trees*. The non-exceptional characters of A1 are 1G and StG , and the
Brauer tree is given by
i y i
TA1 .

9.4.3. The Case when  is Odd and Divides q + 1

Hypothesis. In this and only this subsection we suppose that 


is odd and divides q + 1.

In this case it follows from 9.4.2 that StkG is not simple. Indeed, it contains kv
k
as a submodule. Let us denote StG = StkG /kv .

Proposition 9.4.4. The kG -module k[G /B] is projective and indecomposable. The
k
kG -module StG is simple.
k
Proof. As  is odd, the simplicity of StG follows from Proposition 9.4.3. The
projectivity of k[G /B] follows because  does not divide the cardinality of B.
The fact that k[G /B] is indecomposable follows from Proposition 9.4.3 and
the fact that kv StkG . 

9.4 The Principal Block 105

As a consequence,
k
(9.4.5) Irr kA1 = {k, StG }.

As
Irr KA1 = {1G , StG } {R( ) | [S / ], = 1},
we conclude that the decomposition matrix has the form

10
1 1


Dec(A1 ) = 0 1 .
.. ..
. .
01

Brauer trees*. The non-exceptional characters of A1 are 1G and StG and the
Brauer tree is given by
y i i
TA1

9.4.4. The Case when  = 2

Hypothesis. In this and only this subsection we suppose that


 = 2.

Fix an O-free OG -module V such that KV admits the character R (0 ). Set


k
St = kV .

Then
k q 1
(9.4.6) dimk St =
2
and, by Proposition 5.2.7,
k
(9.4.7) dimk eU St = 0.

In order to show the last equality one uses the fact that |U| is invertible in k.
It then follows from Proposition 9.1.6 that
106 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

k
(9.4.8) St is a simple kG -module.
k k k k
Proposition 9.4.9. If  = 2, then Irr kA1 = {k, St+ , St }. Moreover, St+  St .
k k
Proof. The fact that St+  St follows immediately from considering the re-
striction to U and Proposition 5.2.7.
On the other hand, recall that Irr KA1 is the set of irreducible characters of
R( ) (for S2 ) and R ( ) (for S2 ). But, for such characters, it follows
from 9.1.1 and 9.1.2 that

decO G R( ) = decO G R(1) = [k ]kG + decO G StG

and decO G R ( ) = decO G R (1) = [k ]kG + decO G StG .

As
k k
decO G R (0 ) = [St+ ]kG + [St ]kG ,
the proposition then follows from 9.4.8 and the fact that the isomorphism
class of every simple kA1 -module appears in at least one decO G , where
Irr KA1 (because decO G [KA1 ]KG = [kA1 ]kG ). 


The calculations completed in the proof of the previous proposition al-


low us to compute the decomposition matrix of A1 (using also Proposi-
tion 5.2.7), which is given in Table 9.1 below. In this table, m = (|S2 | 2)/2
and m = (|S2 |2)/2 and 1 ,. . . , m (respectively 1 ,. . . , m ) give a set of rep-
resentatives for (S2 \ {1, 0 })/ (respectively (S2 \ {1, 0 })/ ). Note that,
if q 1 mod 4 (respectively q 3 mod 4), then m = 0 (respectively m = 0).

R EMARK 9.4.10 Note that there is no Brauer tree for A1 because the de-
fect group of A1 is a Sylow 2-subgroup of G and is therefore not cyclic (see
Theorem 1.4.3). 

= GL2 (Fq ) has only


R EMARK 9.4.11 When  = 2, the principal k-block of G
two simple modules, contrarily to the principal k-block of G = SL2 (Fq ). 

Exercises

9.1 (Katz). In this and only this exercise we suppose that  = 2. If a Fq , set
 a
Kl(a) = k z+ .
z
zFq

I would like to thank N. Katz for suggesting this exercise to me.


9.4 The Principal Block 107

Table 9.1 Decomposition matrix for A1 when  = 2.

k k k k
k St+ St k St+ St

1G 1 0 0 1G 1 0 0

StG 1 1 1 StG 1 1 1

R+ (0 ) 0 1 0 R+ (0 ) 0 1 0

R (0 ) 0 0 1 R (0 ) 0 0 1

R+ (0 ) 1 1 0 R+ (0 ) 1 1 0

R+ (0 ) 1 0 1 R+ (0 ) 1 0 1

R(1 ) 2 1 1 R (1 ) 0 1 1
.. .. .. .. .. .. .. ..
. . . . . . . .

R(m ) 2 1 1 R (m ) 0 1 1

Dec(A1 ) when q 1 mod 4 Dec(A1 ) when q 3 mod 4

The purpose of this exercise is to show that there exists an element a of Fq


which is not a square and such that Kl(a) = 0 (this fact is used in the proof of
Proposition 9.4.3). The value Kl(a) is called a Kloosterman sum. The following
exercise does not make use of the finer properties of such sums.
 a 
(a) Show that, if a Fq and b Fq , then Kl(ab2 ) = k b(z + ) .
z
zFq
a a
(b) Let a, z and z  be three elements of Fq such that z + = z  +  . Show
z z
that z = z  or zz  = a.
(c) From now on fix an element a0 of Fq which is not a square. Show that
a0
the fibres of the map : Fq Fq , z z + are of cardinality 0 or 2.
z
(d) If Fq we set
f ( ) = k (b ) Kl(a0 b 2 ).
bFq

Using (a), show that



1 q if | 1 ( )| = 0,
f ( ) =
q + 1 if | 1 ( )| = 2.
108 9 Unequal Characteristic: Simple Modules, Decomposition Matrices

(e) Using (c) and the fact that q 1 and q + 1 cannot both be zero as  = 2,
show that there exists an element a of Fq which is not a square and such
that Kl(a) = 0.

9.2. In this and only this exercise we suppose that  = 2. Fix a non-square a0 Fq
and consider C+ = {b 2 | b Fq } and C = a0 C+ . Set
 
Stk = kv kvx .
xC

Use the argument of the proof of Proposition 9.4.3 to show that Stk is a kG -
submodule of StkG and that the only non-trivial kG -submodules of k[P1 (Fq )]
are kv , Stk+ , Stk and StkG .
Conclude that k[G /B] is indecomposable and determine the structure of
the successive quotients Radi (k[P1 (Fq )])/ Radi+1 (k[P1 (Fq )]), for i  0.

9.3. In this and only this exercise, we suppose that  divides q + 1. Let H be the
algebra of endomorphisms of k[G /B]. Show that H k[X ]/
X 2 , where X
is an indeterminant. Deduce a new proof of the fact that k[G /B] is indecom-
posable.


9.4*. Calculate Dk : K0 (kG ) K0 (kG ) according to the values of . Recall
that Dk is defined in Exercise 8.3. (Hint: Use Exercise 8.3 and the fact that
decO G DK = Dk decO G .)
Chapter 10
Equal Characteristic

Hypothesis. In this and only this chapter we suppose that  = p.

In this chapter we study the representations of the group G in equal, or nat-


ural, characteristic. A significant part of this chapter will be dedicated to
the construction of the simple kG -modules. This classical construction gen-
eralises to the case of finite reductive groups. It turns out that the simple
kG -modules are the restrictions of simple rational representations of the
algebraic group G = SL2 (F). Having obtained this description, the determi-
nation of the decomposition matrices is straightforward.
We then determine the (very simple) partition into blocks as well as the
Brauer correspondents. There is only one block with trivial defect (which
corresponds to the Steinberg character StG ) and the two other O-blocks
(which both have U as their defect group recall that the normaliser of
U is B) are only distinguished by the action of the centre Z . As the group U
is abelian, Brous conjecture predicts an equivalence of derived categories
between the O-blocks of G and their Brauer correspondent. This result was
shown by Okuyama [Oku1], [Oku2], for the principal block and by Yoshii
[Yo] for the nonprincipal one (with defect group U) but the proof is too dif-
ficult to be included in this book.
To finish off, if q = p, then U is cyclic and we determine the Brauer trees
of the two blocks with defect group U.
R EMARK The methods used in this chapter are totally different from those
employed in the rest of this book. In particular, we use neither the geometry
of the Drinfeld curve, nor any cohomology theory with coefficients mod-
ulo p. All methods are entirely algebraic. 

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 109


DOI 10.1007/978-0-85729-157-8_10, Springer-Verlag London Limited 2011
110 10 Equal Characteristic

Terminology, notation. In this and only this chapter we


will denote by G the group SL2 (F) and B (respectively T,
respectively U) the subgroup of G formed by upper trian-
gular matrices (respectively diagonal, respectively upper
unitriangular matrices). Note that

B = T  U.

We will call a rational representation of G a morphism of alge-


braic groups G GLF (V ), where V is a finite dimensional
F-vector space. In this case we will often simply say that V
is a rational G-module, in which case the morphism is im-
plicit. We denote by K0 (G) the Grothendieck group of the
category of rational representations of G.
We will denote by : T F , diag(a, a1 ) a, so that
( )nZ is the set of rational representations of T. Simi-
n

larly, we identify K0 (T) with the group algebra Z[Z] writ-


ten exponentially: K0 (T) = nZ Z n . We will denote by
CarG : K0 (G) K0 (T) the morphism induced by restric-
tion. If V is a rational G-module, we denote by [V ]G
its class in K0 (G) and we will write CarG V instead of
CarG [V ]G in order to simplify notation.

10.1. Simple Modules

10.1.1. Standard or Weyl Modules

Denote by V2 the natural rational G-module of dimension 2, that is to say,


V2 = F2 equipped with the standard action of G = SL2 (F). If n is a natural
number, we denote by (n) the rational G-module defined by

(n) = Symn (V2 ),

where Symn (V2 ) denotes the n-th symmetric power of the vector space V2 .
These modules are called standard modules or Weyl modules. It is easy to verify
that

(10.1.1) dimF (n) = n + 1

and that

(10.1.2) (0) is the trivial G-module.


10.1 Simple Modules 111

We denote by (x, y ) the canonical basis of V2 , so that


   
ab ab
x = ax + cy and y = bx + dy .
cd cd

We have,
n
(n) = Fx ni y i .
i=0

As t x ni y i = n2i (t)x ni y i , we obtain the following formula:


n
(10.1.3) CarG (n) = n2i .
i=0

We will use the following preliminary results in the construction of the sim-
ple rational G-modules.
Lemma 10.1.4. (n)U = Fx n . If moreover 0  n  q 1, then (n)U = Fx n .
Proof. It is clear that x n (n)U (n)U . To show that (n)U Fx n , it is suf-
ficient, after replacing q by a power if necessary, to suppose that 0  n  q 1
and to show that (n)U Fx n . This will also show the second assertion.
Now let f be an element of (n)U . Let us write
n
f = i x ni y i .
i=0

Let Fq . Then u( ) f = f . Now,


 
n n
i ij  nj j
n
u( ) f = i x ni (y + x)i = i x y .
i=0 j=0 i=j j

Comparing coefficients of x n , we therefore have, for all Fq ,


n
i i = 0 .
i=0

As a consequence, the polynomial ni=1 i X i admits q distinct roots and is of


degree  q 1. It is therefore zero, which shows that 1 = = n = 0, and
hence f Fx n . 

Corollary 10.1.5. If V is a non-zero G-submodule of (n), then V contains x n .
If 0  n  q 1 and if V is a non-zero G -submodule of ResG
G (n), then V con-
tains x n .
Proof. After replacing q by a power, we may suppose that 0  n  q 1:
it is then enough to prove the second assertion. Let V be a non-zero G -
submodule of ResG G (n). Then V = 0 because U is a p-group [CuRe, Theo-
U

rem 5.24]. Therefore x V V by Lemma 10.1.4. 


n U 
112 10 Equal Characteristic

To complete this subsection we will give another construction of (n),


using the algebra F[G] of regular functions on G. Denote by : B F the
extension of to B which is trivial on U. If n Z, set

 (n) = {f F[G] | (g , b) G B, f (gb) = (b)n f (g )}.

Proposition 10.1.6. If  (n) = 0, then n  0. In this case, the G-modules (n)


and  (n) are isomorphic.
Proof. Equip V2 with the structure of a rational (G, B)-bimodule, letting B
act on the right via v b = (b)v for all b B and v V2 . We denote by V2
the dual of V2 , and let (x , y ) denote the dual basis of (x, y ). Then V2 is,
by duality, also equipped with the structure of a rational (G, B)-bimodule.
We identify the symmetric algebra Sym(V2 ) with the algebra F[V2 ] of regu-
lar (i.e. polynomial) functions on V2 . This algebra inherits the structure of
a (G, B)-bimodule. Via this action we may describe (n) as follows, when
n  0:

() (n) = {f Sym(V2 ) | b B, f b = (b)n f }.

Set
: Sym(V2 ) F[G]
f (g f (g y )).
If g , h, G and if f F[G], we set

(g f h)( ) = f (g 1 h1 ).

This equips F[G] with the structure of a (G, G)-bimodule. One may easily
verify that is a morphism of (G, B)-bimodules and that

()  (n) = {f F[G] | b B, f b = (b)n f }.

Set
F[G] (U) = {f F[G] | u U, f u = f }.
We will need the following result.
Lemma 10.1.7. is injective and its image is exactly F[G] (U) .
Proof (of Lemma 10.1.7). Let f Sym(V2 ) be such that (f ) = 0. Then f is zero
on the G-orbit of y . This orbit is dense in V2 , therefore f = 0. This shows
the injectivity .
On the other hand, the (G, B)-equivariance of implies that its image is
contained in F[G] (U) . For the other implication, we choose f F[G] (U) and
must show that f is in the image of . We denote by f the map defined on
V2 \ {0} by f(v ) = f (g ) when g G is such that v = g y . As f F[G] (U)
and U is the stabiliser of y in G, the function f is well-defined. We will show
that it is a regular function on V2 \ {0}.
10.1 Simple Modules 113

To this end, write V2 \ {0} = Ux Uy , where

Uv = {v V2 | v (v ) = 0}

for all v V2 . It is sufficient to show that the restriction of f to Uv is regular.


Fix g G such that g y = v . Set B = T  U , where U is the subgroup
of G formed by the lower unitriangular matrices. Then B stabilises the line
Fy and the map
v : B Uv
b gb y

is clearly an isomorphism of varieties. Moreover, if v Uv , then f(v ) =


f (g v1 (v )), and so f is regular on Uv .
We have therefore shown that f is a regular function on V2 \ {0}, and
therefore f is the restriction of a unique regular function f on V2 which, by
construction, verifies (f) = f .  

The proposition then follows immediately from Lemma 10.1.7 and the
equalities () and (). 


R EMARK The isomorphism F[G] (U) Sym(V2 ) is in fact an algebraic con-


sequence of the existence of a natural isomorphism of quasi-affine varieties
G/U V2 \ {0}, g U g y . 

10.1.2. Simple Modules

We denote by L(n) the G-submodule of (n) generated by x n . Denote by

Fp :  G   G 
ab ap b p

cd cp dp

the split Frobenius endomorphism of G over Fp . If i is a non-negative integer


and if V is a rational G-module, we denote by V (i) the rational G-module
with the same underlying space, but on which the element g G acts as
Fpi (g ) acts on V . For example V (0) = V . If v V , we will denote by v (i) the
corresponding element of the G-module V (i) . If g G, then

g v (i) = (Fpi (g ) v )(i) .

If m is a non-zero natural number, we will denote by (mi )i  0 the unique


sequence of elements of {0, 1, 2, ... , p 1} such that

m= mi p i .
i 0
114 10 Equal Characteristic

Note that the sequence (mi )i  0 becomes zero after a certain point. Let

I (n) = {m N | i  0, mi  ni }.

Note that I (n) {0, 1, 2, ... , n} and that 0, n I (n). Moreover, m I (n) if and
only if n m I (n). The following theorem describes the simple rational
G-modules as well as the simple FG -modules.

Theorem 10.1.8. With notation as above we have:


(a) L(n) = mI (n) Fx nm y m . In particular, if 0  n  p 1, then L(n) = (n).
(a) If 0  n  q 1, then L(n) is the G -submodule of ResG n
G (n) generated by x .
(b) L(n) is the unique simple G-submodule of (n).
(b) If 0  n  q 1, then ResG G L(n) is the unique simple FG -submodule of
ResG G (n).
(c) L(n) L(ni )(i) = (ni )(i) .
i 0 i 0
(d) If 0  m, n, then L(m) L(n) if and only if m = n.
(d)If 0  m, n  q 1, then ResGG L(m) ResG L(n) if and only if m = n.
G

(e) The family (L(n))n  0 is a set of representatives of the isomorphism classes of


simple rational G-modules.
(e) The family (ResGG L(n))0  n  q1 is a set of representatives of the isomorphism
classes of simple FG -modules.

Proof. We begin by proving (a) and (a). Let us temporarily denote by V (n)
the F-subvector space of (n) defined by

V (n) = Fx nm y m .
mI (n)
 
ab
Let g = G and let m I (n). Then
cd

(ax + cy )p (ni mi ) (bx + dy )p mi .


i i
g x nm y m = (ax + cy )nm (bx + dy )m =
i 0

As a consequence,

(ap x p + c p y p )ni mi (b p x p + d p y p )mi .


i i i i i i i i
() g x nm y m =
i 0

Expanding this product, we obtain a linear combination of monomials of the


form
x p (ni mi si ) y p si x p (mi ti ) y p ti ,
i i i i

i 0

where 0  si  ni mi and 0  ti  mi . Set m = i  0 (si +ti )p i . Then m I (n)


and
10.1 Simple Modules 115
 
x p (ni mi si ) y p si x p (mi ti ) y p ti = x nm y m ,
i i i i

i 0

which shows that indeed g x nm y m V (n). Hence V (n) is a G-submodule


of (n) containing x n and therefore V (n) contains L(n).
To complete the proof of (a) and (a), we may suppose that 0  n  q 1
(after possibly replacing q by a power). Now set

u : Fq+  G 
10

1

and denote by U the image of the morphism of groups u . It will be suffi-


cient for us to show that

(?) If 0  n  q 1, then V (n) is the U -submodule of (n) generated by x n .

Now, if Fq , then it follows from () that


 
ni
u ( ) x =
n
m x nm y m .
mI (n) i  0
m i

 
ni
If m I (n), then i  0 = 0, therefore it is sufficient for us to show that
mi
the matrix ( m )mI (n), Fq
is of rank |I (n)|, that is to say that the rows of this
matrix are linearly independent. Now, as 0  n  q 1, these rows are com-
plete rows of the matrix ( m )0  m  q1, Fq which is a square Vandemonde
matrix, and hence is clearly invertible.
The claims (b) and (b) follow immediately from Corollary 10.1.5 and
from (a).
We now turn to (c). If m I (n), we will denote by

vm = (x ni mi y mi )(i) L(ni )(i) = (ni )(i) .


i 0 i 0 i 0

The formula () shows that the F-linear map L(n) i  0 L(ni )(i) which
sends x nm y m to vm (for all m I (n)) and is in fact a morphism of G-modules.
Moreover it is an isomorphism.
To show (d) and (d) we may (after replacing q by a power), suppose
that 0  m  n  q 1. Now, if ResG
G L(m) ResG L(n), then the FT -modules
G

L(m) and L(n) are isomorphic. But T acts on L(m)U = Fx m via the char-
U U

T . Hence, ResT = ResT and therefore m n mod q 1 be-


acter ResT m T m T n

cause is injective on T Fq . As 0  m, n  q 1, this can only happen if
m = n or if m = 0 and n = q 1. Now dimF L(0) = 1 = q = dimF L(q 1). There-
fore m = n.
116 10 Equal Characteristic

(e) We denote by F(n) the vector space F on which B acts via the linear
character n . Then (F(n))nZ is a family of representatives of the isomor-
phism classes of simple rational B-modules. Indeed, if S is a simple rational
B-module, then S U = 0 (as U is a unipotent group [Bor, Theorems 4.4 and
4.8]) and is B-stable, therefore S = S U is a simple rational T-module.
Let V be a simple rational G-module. By the previous considerations
there exists n Z and a surjective B-equivariant morphism : ResG BV
F(n). Let
: V F[G]
v (g (g 1 v )).
Then is a morphism of G-modules (where F[G] is equipped with the struc-
ture of a G-module defined by (g f )(x) = f (g 1 x) for all g , x G and
f F[G]). As is a morphism of B-modules, the image of is contained in

 (n) = {f F[G] | (g , b) G B, f (gb) = n (b)f (g )}.

Moreover, : V  (n) is a non-zero morphism, and is therefore injective


as V is simple. As (n)  (n) (see Proposition 10.1.6), it follows from (b)
that V L(n).
The final statement (e) now follows immediately from (d) and from
the fact that, by Proposition B.3.2(b) and Exercises 1.7 and 1.10, we have
| Irr kG | = q. 


C OMMENTARY In the case of a general connected reductive group, the


classification of the simple rational modules by their highest weight is due
to Borel and Weil. In the case of our group G = SL2 (F), this corresponds to
part (e) of Theorem 10.1.8. On the other hand, the description (in part (c) of
Theorem 10.1.8) of the simple modules as tensor products of simple modules
twisted by the action of the Frobenius also generalises to all connected re-
ductive groups: this is the Steinberg tensor product theorem. For further details
concerning the representations of reductive groups in positive characteristic
the reader may consult the book of Jantzen [Jan].
A generalisation of part (e) of Theorem 10.1.8 gives a parametrisation of
the simple representations of finite reductive groups in equal characteristic:
see, for example, [Cartier, 6]. 

E XAMPLE 10.1.9 If r  0, then (p r 1) = L(p r 1). Indeed,


r 1
pr 1 = (p 1)pi
i=0

and therefore I (p r 1) = {0, 1, 2, ... , p r 1}. 


10.1 Simple Modules 117

10.1.3. The Grothendieck Ring of G

Theorem 10.1.8 shows that

(10.1.10) K0 (G) = Z [L(n)]G .


n0

On the other hand,

(10.1.11) CarG L(n) = n2m .


mI (n)

Set 
0 if n = 0,
n =
n + n if n > 0.
Then

(10.1.12) CarG L(n) = n2m n + Z m .


mI (n) 0  m<n
m  n/2
 
0 1
Now the element s = G normalises T and therefore acts on the
1 0
Grothendieck group K0 (T). One may verify that

K0 (T)s = Z n .
n0

This shows that

(10.1.13) The map CarG : K0 (G) K0 (T) is injective with image K0 (T)s .

Similarly, by 10.1.3, we have

(10.1.14) CarG (n) = n2m n + Z m .


0  m  n/2 0  m<n

Which shows that ([(n)]G )n  0 is also a Z-basis of the Grothendieck group


K0 (G). If m, n  0, we denote by m,n = [(n) : L(m)] the multiplicity of
L(m) as a factor in a Jordan-Hlder series of (n). These integers m,n are
uniquely determined by the equation

(10.1.15) CarG (n) = m,n CarG L(m).


m0

In particular, taking into account the formulas 10.1.12 and 10.1.14, we have

(10.1.16) m,n {0, 1}


118 10 Equal Characteristic

and


1 if m = n,
(10.1.17) m,n = 0 if m > n,


0 if m
 n mod 2.

Further to these equalities, it is not difficult to completely determine the


infinite matrix = (m,n )m,n  0 . For this we will need to define recursively
a set E (n) as follows:


{0} if 0  n  p 1,
 n n0 

E (n) = pE if n  p and n0 = p 1,

 p   
pE n n0 n0 + 1 + pE n n0 p
if n  p and n0  p 2.
p p

The decomposition matrix is described in the next proposition.

Proposition 10.1.18. Let m, n  0. Then m,n {0, 1} and m,n = 1 if and only
if m n 2E (n).

Proof. Let us first remark that:

If r E (n), then n 2r  0.

This follows from an elementary recurrence.


We will now show the proposition by induction on n. The crucial step
is given by the following equality, in which we set (1) = 0 to simplify
notation. We claim that, if n  0, then
 (1)
[(n)]G = [(n0 )]G [ nn
p
0
]G
()  (1)
+[(p n0 2)]G [ nnp0 p ]G .

We now show (). Write n = n0 + pn . We then have (n n0 )/p = n and


n0 n 

CarG (n0 ) F (n )(1) = n0 2i p(n 2j)
i=0 j=0
n0 n 
= n2(i+pj) .
i=0 j=0

Moreover
10.2 Simple kG -Modules and Decomposition Matrices 119

pn0 2 n 1

CarG (p n0 2) F (n 1)  (1)
= pn0 22i p(n 12j)
i=0 j=0
p1 n 1
= n2(i+pj) .
i=n0 +1 j=0

To show (), it is enough to remark that

{0, 1, 2, ... , n} = {i + pj | 0  i  n0 and 0  j  n }


{i + pj | n0 + 1  i  p 1 and 0  j  n 1}.

We now return to the proof of Proposition 10.1.18. This result is immedi-


ate if n  p 1. Arguing by induction, we may suppose that n  p and that
the result holds for all n < n. By () and the induction hypothesis, we have
(as L(i) L(j)(1) = L(i + pj) if 0  i  p 1)

[(n)]G = [L(n0 ) F L(n 2m)(1) ]G


mE (n )

+ [L(p n0 2) F  L(n 1 2m)(1) ]G


mE (n 1)





[L(n 2pm)]G if n0 = p 1,
mE (n )
=



[L(n 2(n0 + 1 + pm))]G if n0  p 2.
mE (n 1)

The proof of the proposition is now complete. 



E XAMPLE 10.1.19 If i, j {0, 1, ... , p 1}, then

[L(pj + i)]G if i = p 1 or j = 0,
[(pj + i)]G =
[L(pj + i)]G + [L(pj i 2)]G if i  p 2 and j  1. 

10.2. Simple kG -Modules and Decomposition Matrices

This section is primarily concerned with determining the decomposition


matrix of the finite group G modulo p.

10.2.1. Simple kG -Modules

Recall that k F as k is of characteristic p. On the other hand, O contains the


elements of the form + 1 , where is a (q + 1)-th root of unity (see the
120 10 Equal Characteristic

character table in Table 5.4). Therefore k contains all elements of the form
+ 1 , as traverses q+1 . By virtue of Exercise 1.1, this implies that

Fq k F.

The FG -module ResGG (n) is the extension of scalars to F of a kG -module


which we will denote by q (n): in fact, q (n) = Symn (V2k ), where V2k = k 2
is the natural representation of G = SL2 (Fq ) GL2 (k). Similarly, Fp restricts
to an automorphism of the finite group G and we may define q (n)(i) as
previously, and similarly for Lq (n) and Lq (n)(i) . By Theorem 10.1.8(e), the
morphism

{0, 1, 2, ... , q 1} Irr kG


(10.2.1)
n [Lq (n)]kG

is bijective.
On the other hand, by 10.1.17, we have

(10.2.2) [q (n) : Lq (m)] = m,n = [(n) : L(m)]

for all m, n {0, 1, 2, ... , q 1}, which shows that


 
(10.2.3) [q (n)]kG 0  n  q1 is a Z-basis of K0 (kG ).

To finish this subsection, note the following result.

Lemma 10.2.4. The simple kG -module Lq (q 1) is projective. If M is an OG -


module such that [KM ]KG = StG , then kM Lq (q 1).

Proof. By the assertion (?) in the proof of Theorem 10.1.8,

Lq (q 1) = kU x q1 .

As dimk Lq (q 1) = q by 10.1.1 and Theorem 10.1.8(c), we deduce an iso-


morphism of kU -modules ResG
U Lq (q 1) kU . Hence ResU Lq (q 1) is
G

a projective kU -module, hence Lq (q 1) is a projective kG -module (as U


is a Sylow p-subgroup of G ).
On the other hand, Lq (q 1)B = kx q1 = 0, and therefore there exists
an injection k ResG B Lq (q 1) (where k denotes the trivial kB-module).
Frobenius reciprocity then implies that there exists a non-zero morphism
IndGB k Lq (q 1). But Lq (q 1) is simple and hence this morphism is sur-
jective. The last assertion of the lemma follows.  
10.2 Simple kG -Modules and Decomposition Matrices 121

10.2.2. Decomposition Matrices

Recall that Dec(OG ) is the transpose of the matrix decO G : K0 (KG ) K0 (kG )
in the bases Irr KG and Irr kG . Taking into account 10.2.3, we denote by
Dec (OG ) the transpose of the matrix of the morphism decO G in the bases
Irr KG and (q (n))0  n  q1 .
If we set [q] the matrix (m,n )0  m,n  q1 (that is to say, the matrix
obtained from the infinite matrix by taking only the first q rows and
columns), then

(10.2.5) Dec(OG ) = Dec (OG ) t[q].

As we have already determined the matrix [q] in Proposition 10.1.18 it is


enough to determine the matrix Dec (OG ).
To this end, note that the canonical surjection O k admits a unique
section which we denote by : k  O (see 6.1.1). For H {T , T  } let

H : H O
t ( (t)).

We identify K0 (kG ) with the lattice of (virtual) Brauer characters [CuRe,


Definition 17.4 and Proposition 17.14]. Now a Brauer character of G
(which is defined on the p-regular elements of G , that is to say the set of
semi-simple elements) is totally determined by the restrictions ResGT and
ResG
T  .
For example, it follows from 10.1.3 that
n
(10.2.6) ResG
H [q (n)]kG = Hn2i
i=0

for all 0  n  q 1.
On the other hand, if H {T , T  }, then

Irr H = {Hi | 0  i  |H| 1}.

Examining the character table of G , we may easily verify that, if j Z, then

(10.2.7) ResG j j j
H R(T ) = H=T (H + H ) + 2 Hi
0  i  |H|1
ij mod 2

and

(10.2.8) ResG  j j j
H R (T  ) = H=T  (H + H ) + 2 Hi .
0  i  |H|1
ij mod 2
122 10 Equal Characteristic

Here, H=T is equal to 0 if H = T  and 1 if H = T (and, similarly, H=T  =


1 H=T ). We may easily deduce (again using the character table of G ) the
form of the matrix Dec (OG ):

Proposition 10.2.9. To simply notation set q (1) = Lq (1) = 0. Then:


(a) decO G 1G = [q (0)]kG = [Lq (0)]kG .
(b) decO G StG = [q (q 1)]kG = [Lq (q 1)]kG .
(c) If 0  i  q 1, then decO G R(Ti ) = [q (i)]kG + [q (q 1 i)]kG .
(d) If 1  i  q, then decO G R (Ti  ) = [q (i 2)]kG + [q (q 1 i)]kG .
(q1)/2
(e) decO G R (T ) = [q ((q 1)/2)]kG .
decO G R (T 
(q+1)/2
(f) )= [q ((q 3)/2)]kG .

10.3. Blocks

10.3.1. Blocks and Brauer Correspondents



Recall that d denotes the natural isomorphism of groups Fq T . We set

1 1
e+ = (d(1) + d(1)), G
e+ = e+ eStG and G
e = e = (d(1) d(1)).
2 2
Then it is straightforward to verify that
G G
(10.3.1) 1 = e+ + e + eStG

is a decomposition of 1 as a sum of pairwise orthogonal central idempotents


of OG . The following proposition is even better:
G , e G and e
Proposition 10.3.2. The central idempotents e+ StG are primitive.

Proof. Denote by 1+ (respectively 1 ) the trivial (respectively non-trivial)


linear character of Z . If Irr G , we set the unique linear character of Z
such that (zg ) = (z) (g ) for all (g , z) G Z . Let

B+ = { Irr G | = 1+ } \ {StG },

B = { Irr G | = 1 }

and BSt = {StG }.

One may easily verify, thanks to the table of central characters in Table 6.1,
that B+ , B and BSt are the p-blocks of G , which shows the result.  
10.3 Blocks 123

We set

A+ = OGe+
G
, A = OGe
G
and ASt = OGeStG .

Note that A+ is the principal O-block of G and that ASt is a block with trivial
defect. The straightforward calculation of the Brauer correspondence is left
to the reader.

Proposition 10.3.3. The group U is a defect group of A+ and A , and OBe? is a


Brauer correspondent of A? (for ? {+, }).

Proposition 10.3.4. We have

Irr kA+ = {[Lq (2n)]kG | 0  n  (q 3)/2},

Irr kA = {[Lq (2n + 1)]kG | 0  n (q 3)/2}

and Irr kASt = {[Lq (q 1)]kG }.

Brous conjecture was shown in this case by Okuyama [Oku1], [Oku2]


for the principal block and by Yoshii [Yo] for the nonpricipal one.

Theorem 10.3.5 (Okuyama, Yoshii). There exist Rickard equivalences

Db (A+ ) Db (OBe+ ) and Db (A ) Db (OBe ).

10.3.2. Brauer Trees*

As Brauer trees are only defined in the case of cyclic defect group we make
the following assumption which guarantees that the group U, which is the
defect group of A+ and A , is cyclic.

Hypothesis. In this and only this subsection we suppose that


 = p = q.

It follows from this hypothesis that

[q] = [p] = Ip ,

and therefore that


Dec(OG ) = Dec (OG )
124 10 Equal Characteristic

(see 10.2.5). Let R0 = 1G , Ri = R(Ti ) (if 1  i  (p 1)/2) and Ri = R (Ti  ) (if
1  i  (p + 1)/2). Up to the planar embedding, the Brauer tree of a block
with cyclic defect group is completely determined by its decomposition ma-
trix. We conclude from Proposition 10.2.9 that the Brauer trees of A+ and A
are as follows.
R0 R2 R2 R4 + exc
+
TA+ i i i i p p p i y

R1 R1 R3 R3
exc
TA i i i i p p p i y.

where
 

R(p1)/2 if p 1 mod 4, R(p1)/2 if p 1 mod 4,
+ = and exc
+
= 
R(p3)/2 if p 3 mod 4, R(p+1)/2 if p 3 mod 4,

and, similarly,
 
R(p3)/2 if p 1 mod 4, 
R(p+1)/2 if p 1 mod 4,

=  and exc =
R(p1)/2 if p 3 mod 4, R(p1)/2 if p 3 mod 4.

We can also calculate the Brauer trees for the Brauer correspondents A+
and A of A+ and A respectively. The exceptional characters of A+ (re-
spectively A ) are +,+
B and +,
B (respectively ,+
B and ,
B ). Using the

character table of the group B, we obtain the following Brauer trees.


4
T2 iT
i
S  p
S 
 p
S
Sy
TA T0 i  p
+
 C p
 C
 C p

i CC
Tp3 i p5
T
10.3 Blocks 125

5
T3 iT
i
S  p
S 
 p
S
Sy
TA T1 i  p

 C p
 C
 C p

i CC
Tp2 i p4
T

By virtue of Brauers theorem (see Appendix C), we may conclude that,


if ? {+, }, then

(10.3.6) the blocks A? and A? are not Morita equivalent.

R EMARK The blocks A+ and A are Morita equivalent because they have
the same Brauer tree (Brauers theorem) while the O-algebras A+ and A
are isomorphic (by restriction of the isomorphism OB OB, b T (b)b). 

Exercises

10.1* (Carter-Lusztig). Denote by K1 (kG ) the Grothendieck group of the


category of projective kG -modules, which we may view as a sublattice (and
even an ideal) of K0 (KG ). Show that

K1 (kG ) = K0 (KG ) StG .

10.2. If : T k is a linear character, we denote by k the kT -module


defined as follows: the underlying k-vector space is k itself and an element
t T acts as multiplication by (t). Show that, if = 1 , then the kG -
modules k[G /U] kT k and k[G /U] kT k 1 are not isomorphic.
Compare with Exercise 7.4.
10.3*. The purpose of this exercise is to calculate certain subalgebras of in-
variants.
(a) Let
F[G]UU = {f F[G] | (u, v ) U U, u f v = f }
and denote by : G F the regular function defined by
 
ab
= c.
cd
126 10 Equal Characteristic

Show that F[G]UU is generated by .


(b) Let

F[G]U U = {f F[G] | (u, v ) U U, u f v = f }
and denote by resT : F[G] F[T] the restriction morphism. Show that resT

induces an isomorphism F[G]U U F[T].
(c) Let
F[G] (G) = {f F[G] | g G, g f g 1 = f }.
Show that resT induces an isomorphism F[G] (G) F[T]s and that F[G] (G)
is generated by the trace function Tr : G F.
(d) Let
F[G] (T) = {f F[G] | t T, f t = f }.
 
ab
If g = G, we set
cd

f1 (g ) = ab, f2 (g ) = cd, d1 (g ) = ad and d2 (g ) = bc.

Show that F[G] (T) is generated by f1 ,f2 , d1 and d2 and that the kernel of
the canonical morphism

F[F1 , F2 , D1 , D2 ] F[G] (T)


P P(f1 , f2 , d1 , d2 )

is generated by F1 F2 D1 D2 and D1 D2 1 (here, F1 , F2 , D1 and D2 are


indeterminates).
Part IV
Complements
The goals of these two final chapters are totally opposite. In Chapter 11,
we consider the results obtained for SL2 (Fq ) when q {3, 5, 7} and point
out various curious facts that occur in these small cases. We will study
morphisms to SL2 (Fr ) as well as encountering exceptional reflection groups
of ranks 2 and 3, the Klein curve and certain exceptional isomorphisms.
In Chapter 12, we give a very brief summary (without proof) of Deligne-
Lusztig theory, whose goal is the study of representations of arbitrary finite
reductive groups using geometric methods. We explain to what extent the
results that we have seen for SL2 (Fq ) may be interpreted as a special case of
the general theory.
Chapter 11
Special Cases

In this chapter we will make explicit certain exotic properties of the groups
SL2 (Fq ) when q = 3, 5 or 7. These include exceptional isomorphisms, inclu-
sions as subgroups of SL2 (Fq ), and realisations as subgroups of reflection
groups. For a recollection of definitions, results about reflection groups, see
the Appendix C.

Notation. We will denote by Z the centre of G = SL2 (Fq ), Z the



centre of G = GL2 (Fq ), PGL2 (Fq ) = G /Z the projective general
linear group and PSL2 (Fq ) = G /Z the projective special linear
group. If n is a non-zero natural number, we denote by Sn the
symmetric group of degree n (i.e. the group of permutations of
the set {1, 2, ... , n}) and n : Sn {1, 1} the sign homomor-
phism. The alternating group of degree n (the kernel of the sign
homomorphism) will be denoted An .

11.1. Preliminaries

The linear character 0 of Fq , of order 2, is the Legendre character: 0 (z) is


equal to 1 if z is a square and is 1 otherwise. In particular, the linear char-
factorises to give a character of G
acter 0 det of G /Z
which we will denote
by det0 : PGL2 (Fq ) {1, 1}. One may easily verify that

(11.1.1) PSL2 (Fq ) = Ker(det0 ).

The action of PGL2 (Fq ) on the finite projective line P1 (Fq ) induces, after enu-
merating its elements 1 up to q + 1, a homomorphism of groups

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 129


DOI 10.1007/978-0-85729-157-8_11, Springer-Verlag London Limited 2011
130 11 Special Cases

q : PGL2 (Fq ) Sq+1 .

The character det0 may be expressed with the help of the sign homomor-
phism on Sq+1 :

(11.1.2) det0 = q+1 q .

denotes the subgroup of G


Proof. Recall that B formed by upper triangular

matrices. As G = B Bs B, the group G is generated by the conjugates of B.
then
As a consequence, it is enough to show that, if g B,

0 (det(g )) = q+1 q (g ),

where g denotes the image of g in PGL2 (Fq ). Write


 
ab
g= .
0c

Let = [1; 0] P1 (Fq ). Then g = and, if z Fq , then g [z; 1] = [(az +


b)/c; 1]. Denote by : Fq Fq , z (az + b)/c. It suffices to show that the
sign of the permutation is equal to 0 (ac).
To this end, fix a total order  on Fq and set

A=

(z  z) Fq .
z,z Fq
z<z 

Then the sign of is defined by

A =

( (z  ) (z)).
z,z Fq
z<z 

It follows that, in Fq , we have = (a/c)q(q1)/2 = (ac)q(q1)/2 c q(q1) =


(ac)(q1)/2 . Hence = 0 (ac) as expected. 


Hence the morphism q induces by restriction a morphism

q : PSL2 (Fq ) Aq+1 .

Recall that

(11.1.3) q and q are injective

and
q(q 2 1)
(11.1.4) |PGL2 (Fq )| = q(q 2 1) and |PSL2 (Fq )| = .
2
11.2 The Case when q = 3 131

11.2. The Case when q = 3

Hypothesis. In this and only this section we suppose that q = 3.

11.2.1. Structure

Recall that

(11.2.1) | SL2 (F3 )| = 24, |PGL2 (F3 )| = 24 and |PSL2 (F3 )| = 12,

which shows that the morphisms 3 and 3 defined in the previous sections
are isomorphisms, that is to say

(11.2.2) PGL2 (F3 ) S4 and PSL2 (F3 ) A4 .

Therefore SL2 (F3 ) is a non-trivial central extension of A4 by Z/2Z (which is


unique up to isomorphism) while PGL2 (F3 ) is a non-trivial central extension
of S4 (there are three such extensions as H 2 (S4 , Z/2Z) Z/2Z Z/2Z).
We now show the following result, stated in Theorem 1.2.4.

Proposition 11.2.3. We have G = N   U and N  = D(G ). The group N  is isomor-


phic to the quaternionic group of order 8. The only non-trivial normal subgroups of
G are Z and N  .

Proof. We have |N  | = 8 and |G | = 24, and therefore N  is a Sylow 2-subgroup


of G . It is therefore the inverse image, under 3 , of the Sylow 2-subgroup
of A4 . Now the Sylow 2-subgroup of A4 is normal in A4 (and even in S4 ),
therefore N  is normal in G . On the other hand, as N  U = {I2 } and |N  |
|U| = |G |, we have G = N   U.
This shows in particular that D(G ) N  . The equality N  = D(G ) and the
two last assertions of the proposition are left as an (easy) exercise.  

Let      
0 1 1 1 1 1
I= , J= and K= .
1 0 1 1 1 1
Then
N  = {I2 , I , J, K }
and the multiplication is given by

I 2 = J 2 = K 2 = I2 , IJ = JI = K , JK = KJ = I and KI = IK = J,

which shows again that N  is the quaternionic group of order 8. To simplify


notation let
132 11 Special Cases
 
11
u = u+ = .
01
One may easily verify that

U =
u , u 3 = I2 , u
I = J, u
J =K and u
K = I.

11.2.2. Character Table

The character table 5.4 specialises as follows. Let i denote an element of 4 of


order 4 (and note that q0 = 3), let j = (1 + 3)/2 (so that j is a third root
of unity in K ) and let i be a linear character of order 4 of 4 . The character
table of SL2 (F3 ) is given in Table 11.1.

Table 11.1 Character table of SL2 (F3 )

g I2 I2 d (i) u+ u u+ u

| ClG (g )| 1 1 6 4 4 4 4

o(g ) 1 2 4 3 3 6 6

CG (g ) G G T ZU ZU ZU ZU

1G 1 1 1 1 1 1 1

R+ (0 ) 1 1 1 j j2 j j2

R (0 ) 1 1 1 j2 j j2 j

R (i ) 2 2 0 1 1 1 1

R+ (0 ) 2 2 0 j 2 j j2 j

R (0 ) 2 2 0 j j 2 j j2

StG 3 3 1 0 0 0 0
11.2 The Case when q = 3 133

11.2.3. The Group SL2 (F3 ) as a Subgroup of SL2 (F )

Amongst the three characters of degree 2 of G , only R (i ) has rational val-


ues. Here we study the possible fields of definition of a representation ad-
mitting this character. For this, we commence by calculating the Frobenius-
Schur indicator of R (i ):

1
|G | g
R (i )(g 2 )
G
1
= (2 + 2 + 6 (2) + 4 (1) + 4 (1) + 4 (1) + 4 (1)) = 1.
24
As a consequence (see [Isa, Corollary 4.15]), R (i ) is not the character of a
real representation, even though it has rational (and hence real) values.

Proposition 11.2.4. Let R be a ring in which 2 is invertible and 1 is a sum of


two squares. Then exists an injective homomorphism : SL2 (F3 ) SL2 (R)
 there 
1 0
sending I2 to and such that, if g SL2 (F3 ), then
0 1

Tr( (g )) = R (i )(g ) 1R R.

Proof. Choose a and b in R such that a2 + b 2 = 1. Write


     
0 1 a b b a
(I ) = , (J) = , (K ) =
1 0 b a a b

and  
1 ba1 1ab
(u) = .
2 1 a b a b 1
One verifies easily that det (u) = 1, Tr (u) = 1 (and therefore (u)2 +
(u) + 1 = 0), that
 
1 0
(I )2 = (J)2 = (K )2 = ,
0 1

that
(I ) (J) = (J) (I ) = (K ),
and that
(u) (u) (u)
(I ) = (J), (I ) = (J) and (I ) = (J).

It follows that we can extend to a morphism of groups

: SL2 (F3 ) SL2 (R)


134 11 Special Cases

satisfying the conditions of the proposition. 




Corollary 11.2.5. Let R be a principal ideal domain of characteristic different from


2 and let K be its field of fractions. We assume that 1 is the sum of two squares
 exists an injective morphism of groups : SL2 (F3 ) SL2 (R)
in K. Then there
1 0
sending I2 to and such that, if g SL2 (F3 ), then
0 1

Tr( (g )) = R (i )(g ) 1R R.

Proof. By Proposition 11.2.4, there exists


 an injective
 morphism of groups
1 0
: SL2 (F3 ) SL2 (K) sending I2 to . The proposition follows
0 1
from Proposition B.3.1 (as R is a principal ideal domain). 

The corollary shows that, if K is a field of characteristic different from 2


in which 1 is a sum of two squares, then there exists a faithful irreducible
representation SL2 (F3 ) over K of dimension 2 admitting R (i ) as character.
In fact the converse is also true (see Exercise 11.4).

E XAMPLE 11.2.6 Denote by n a primitive nth root of unity in C. Then


Z[3 ], Z[4 ] and Z[5 ] satisfy the hypotheses of Corollary 11.2.5. Indeed, 1 =
32 + (32 )2 = 42 + 02 = (5 + 51 )2 + (52 52 )2 . 

E XAMPLE 11.2.7 In a finite field every element is the sum of two squares.
This is immediate in characteristic 2. If K is a finite field of characteristic
different from 2 and cardinality r , then, if we let C = {a2 | a K}, we have
|C | = (r + 1)/2. Therefore, if z K, the sets C and z C cannot be disjoint.
Therefore C + C = K and, in particular, 1 is the sum of two squares.
As a consequence, for all odd prime numbers , 1 is the sum of two
squares in F and, by Hensels lemma, 1 is the sum of two squares in
Z . We then obtain two injective group homomorphisms SL2 (F3 )  SL2 (Z )
and 3 : SL2 (F3 )  SL2 (F ). We may then deduce a new proof of The-
orem 1.4.3(d): if  3 mod 8, then S = 3 (N  ) is a Sylow 2-subgroup
of SL2 (F ). Therefore, the image of 3 is contained in NSL2 (F ) (S), which
shows that |NSL2 (F ) (S)/S|  3. 

11.2.4. The Group SL2 (F3 ) as a Reflection Group of Rank 2

A review of reflection groups is contained in Appendix C. Let V be a C-


vector space of dimension 2 and let : G GLC (V ) be an irreducible rep-
resentation of G whose character does not take on only rational values. If
u is an element of G of order 3, and if we denote by ,  the eigenvalues
of (u), then and  are third roots of unity such that +  {j, j 2 }.
11.2 The Case when q = 3 135

Consequently, { ,  } = {1, j} or {1, j 2 }. In either case, (u) is a reflection. As


G is generated by elements of 3, we obtain:

(11.2.8) (G ) G is a reflection group of rank 2.

Because N  is the derived group of G , we have (N  ) SLC (V ). As a conse-


quence, Ref( (G )) consists of (u), where u belongs to the set of elements of
order 3 in G . Hence
|Ref(G )| = 8.
If we denote by (d1 , d2 ) the sequence of degrees of (G ) (see the Shephard-
Todd-Chevalley theorem in Appendix C), then

d1 + d2 2 = |Ref( (G ))| = 8 and d1 d2 = |G | = 24,

and so d1 = 4 and d2 = 6.
The group (G ) is an exceptional complex reflection group denoted G4
in the classification of Shephard and Todd [ShTo] (see also the recent book
of Lehrer and Taylor [Leh]). The group G therefore admits a  presentation

11
G =< a, b | a = b = 1, aba = bab >, sending, for example, a to
3 3 and b
  01
1 0
to .
1 1

11.2.5. The Group PSL2 (F3 ) and the Isometries of the


Tetrahedron

Denote by Q[P1 (F3 )] the representation over Q of G with character IndG


B 1B .
Denote by H the hyperplane of Q[P1 (F3 )] equal to

H ={ a | a = 0}.
P1 (F3 ) P1 (F3 )

Then H is G -stable and the corresponding representation G GLQ (H )


admits StG as character, and factorises to give an injection

3 : PSL2 (F3 )  GLQ (H ).

Let us equip Q[P1 (F3 )] with a scalar product such that the set P1 (F3 ) pro-
vides an orthonormal basis. Denote by : Q[P1 (F3 )] H the orthogonal
projection and consider

= { ( ) | P1 (F3 )}.
136 11 Special Cases

Then is a regular tetrahedron in H upon which PSL2 (F3 ) acts (via 3 )


as isometries. One may easily verify that PSL2 (F3 ) is the group of oriented
isometries of the tetrahedron .

11.3. The Case when q = 5

Hypothesis. In this and only this section we suppose that q = 5.

11.3.1. Structure

We have

(11.3.1) | SL2 (F5 )| = 120, |PGL2 (F5 )| = 120 and |PSL2 (F5 )| = 60.

By Theorem 1.2.4, PSL2 (F5 ) is simple (of order 60), therefore

(11.3.2) PSL2 (F5 ) A5 .

This isomorphism may be seen as follows. Firstly, note that |N| = 8 and |G | =
8 15, and therefore N is a Sylow 2-subgroup of G . As q 5 mod 8, we have
|NG (N)/N| = 3, which shows that NG (N) is of index 5 in G . The action of G
on G /NG (N) (or, equivalently, the action of G on its five Sylow 2-subgroups)
induces a morphism G S5 having kernel Z and whose image, of order 60,
is necessarily equal to the alternating group A5 . We can also see NG (N) as
being the image of SL2 (F3 ) under the morphism 35 of Example 11.2.7.
It follows that G = SL2 (F5 ) is a non-trivial central extension of A5 (which
is unique up to isomorphism). Note that the isomorphism PSL2 (F5 ) A5
can be seen in other ways (see Exercise 11.1).

11.3.2. Character Table

The character table 5.4 specialises as follows. Let i denote an element of F5


such that i 2 = 1, j an element of 6 such that j 2 +j +1 = 0 (i.e. j is of order 3),
i a linear character of F5 of order 4, j a linear character of 6 of order 3
and denote by j the linear character of order 6 of 6 equal to 0 j . Then,
setting
1+ 5 1 5
= and = ,
2 2
the character table of SL2 (F5 ) is given in Table 11.2.
11.3 The Case when q = 5 137

Table 11.2 Character table of SL2 (F5 )

g I2 I2 d(i) d (j) d (j) u+ u u+ u

| ClG (g )| 1 1 30 20 20 12 12 12 12

o(g ) 1 2 4 3 6 5 5 10 10

CG (g ) G G T T T ZU ZU ZU ZU

1G 1 1 1 1 1 1 1 1 1

R+ (0 ) 2 2 0 1 1

R (0 ) 2 2 0 1 1

R+ (0 ) 3 3 1 0 0

R (0 ) 3 3 1 0 0

R (j ) 4 4 0 1 1 1 1 1 1

R (j ) 4 4 0 1 1 1 1 1 1

StG 5 5 1 1 1 0 0 0 0

R(i ) 6 6 0 0 0 1 1 1 1

11.3.3. The Group SL2 (F5 ) as a Subgroup of SL2 (Fr )

Let 5 be the fifth root of unity in K equal to + (1) (recall that the linear
character
+ : F5+ K was fixed in Section 5.2). We then have, by definition
of 5 (see 5.2.3),
1 + 5
5 + 51 = .
2
We remark that G admits two irreducible representations of degree 2 and in
both representations the centre Z acts non-trivially. The following proposi-
tion shows that we can realise these representations over the field Q(5 ).

Proposition 11.3.3. There exists an (injective) homomorphism of groups  :


SL2 (F5 ) GL2 (Q(5 )) with character R (0 ).

Proof. Denote by m the Schur indicator of R (0 ) over Q(5 ) (see [Isa, Defini-
tion 10.1]). The proposition amounts to showing that m = 1 (see [Isa, Corol-
lary 10.2 (e)]). But R (0 ) appears with multiplicity 1 in IndG
U + . As + has
138 11 Special Cases

values in Q(5 ) and has Schur index 1 over this field, it results from [Isa,
Lemma 10.4] that m = 1.  
R EMARK 11.3.4 The field Q(5 ) is the smallest field over which a rep-
resentation of SL2 (F5 ) admits R (0 ) as character. Indeed, it is impossible
to realise
such a representation over Q, because the character has values in
Q(5) = Q(5 + 51 ). Nor is it possible to realise such a representation over
Q( 5) because this field is contained in R and the Frobenius-Schur
indicator
of R (0 ) is 1 (as may be easily calculated). Now Q and Q( 5) are the only
proper subfields of Q(5 ), and the minimality of Q(5 ) follows. 

Because the ring Z[5 ] is a principal ideal domain, it follows from Proposi-
tion 11.3.3 and B.3.1 that there exists a morphism  : SL2 (F5 ) GL2 (Z[5 ])
such that Tr( (g )) = R (0 )(g ) for all g SL2 (F5 ). Because SL2 (F5 ) is per-
fect we have

(11.3.5) Im  SL2 (Z[5 ]).

If  is a prime number (possibly equal to p = 5) and if is a maximal ideal


of Z[5 ] containing , we denote by  the reduction modulo of +
 :

 : SL2 (F5 ) SL2 (Z[5 ]/ ).

Proposition 11.3.6. The representation  : SL2 (F5 ) SL2 (Z[5 ]/ ) is (abso-


lutely) irreducible. Moreover:
(a) If  = 2, then Ker  = {I2 }.
(b) If  is odd, then is injective.
Proof. The only normal subgroups of SL2 (F5 ) are {I2 }, {I2 } and SL2 (F5 ). As
Tr( (d (j))) = 1 = 0, we obtain that Ker  = SL2 (F5 ).
 
1 0
On the other hand,  (I2 ) = , which completes the proof of (a)
0 1
and (b). Now, if the representation were not (absolutely) irreducible, the im-
age of  would be contained, up to conjugation, in the subgroup of SL2 (Fr )
formed by upper triangular matrices (for some r  1). In particular, the im-
age would be solvable, which is not the case.  

Denote by R5 1 the ring of integers of Q( 5) (recall that R5 = Z[(1 +
5)/2] = Z[5 + 5 ]). As R+ (0 ) takes values in R5 , we have Tr( (g ))
R5 /( R5 ) for all g G . Consequently, it results from [Isa, Theorem 9.14(b)]
(with this result following from Wedderburns theorem that all finite di-
vision rings are commutative) that, after conjugating by an element of
GL2 (Z[5 ]/ ), we obtain an (absolutely) irreducible representation

5 : SL2 (F5 ) SL2 (R5 /( R5 )).

We can then deduce from Proposition 11.3.6 the following results.


11.3 The Case when q = 5 139

Corollary 11.3.7. With the above notation, we have:


(a) If  = 2, then R5 = 2R5 , Ker 5 = {I2 }, R5 /( R5 ) = F4 and 5
induces an isomorphism PSL2 (F5 ) SL2 (F4 ).
(b) If  > 2, then 5 is injective.

(c) If  = 5, then  R5 = 5R5 , R5 /( R5 ) = F5 and 5 is an isomorphism
SL2 (F5 ) SL2 (F5 ).
(d) If  1 mod 10, then R5 /( R5 ) = F and therefore SL2 (F5 ) is isomorphic
(via 5 ) to a subgroup of SL2 (F ).
(e) If  3 mod 10, then R5 /( R5 ) = F2 and therefore SL2 (F5 ) is isomorphic
(via 5 ) to a subgroup of SL2 (F2 ).

Proof. It is enough to determine the structure of the field R5 /( R5 ),that is


to say, to decide whether X 2 X 1 (the minimal polynomial of (1 + 5)/2)
admits a root in F or not. If  {2, 5} this is straightforward, and if   {2, 5}
we must determine whether 5 is a square modulo . By the law of quadratic
reciprocity, this is equivalent to determining if  is a square modulo 5. The
statements of the corollary then follow.  

E XAMPLE 11.3.8 Proposition 11.3.7(e) shows that SL2 (F5 ) is isomorphic


to a subgroup H of SL2 (F9 ). As H is of index 6, the left action of SL2 (F9 ) on
SL2 (F9 )/H induces a homomorphism SL2 (F9 ) S6 . As SL2 (F9 ) is equal to its
derived group and PSL2 (F9 ) is simple (see Theorem 1.2.4), one may compare
cardinalities to conclude that the above morphism induces an isomorphism

PSL2 (F9 ) A6 . 

E XAMPLE 11.3.9 Proposition 11.3.7(d) shows that SL2 (F5 ) is isomorphic


to a subgroup H of SL2 (F11 ). As H is of index 11, the left action of SL2 (F11 )
on SL2 (F11 )/H yields a homomorphism SL2 (F11 ) S11 . As SL2 (F11 ) is equal
to its derived group and PSL2 (F11 ) is simple (see Theorem 1.2.4), this mor-
phism induces an injection PSL2 (F11 )  A11 with image a transitive sub-
group. This morphism is exceptional by comparison with the natural ac-
tion of PSL2 (F11 ) on the set of 12 elements of P1 (F11 ). One may show that the
image of PSL2 (F11 )  A11 is contained in the Mathieu group M11 (a simple
subgroup of A11 of order 7920). This image, of index 12, then allows one to
construct a transitive action of the Mathieu group M11 on a set of order 12. 

11.3.4. The Group SL2 (F5 ) Z/5Z as a Reflection Group of


Rank 2

 : SL (F ) SL (Q( ))
Let us reconsider the irreducible representation + 2 5 2 5
C
and denote by 5 the subgroup of GL2 (Q(5 )) given by scalar transforma-
tions of order 1 or 5. Set W + = Im + . We have C W + = {1}. We may
5
therefore consider
140 11 Special Cases

W = 5C W + GL2 (Q(5 )).


We have

(11.3.10) D(W ) = W + .

Proposition 11.3.11. W is a subgroup of GL2 (Q(5 )) generated by reflections. All


its reflections are of order 5 and |Ref(W )| = 48. Furthermore, D(W ) = W + and
Z(W ) = 10 . The sequence of degrees of W is (20, 30) and W is the group denoted
G16 in the Shephard-Todd classification.

Proof. Amongst the elements of order 5 of W + , denote by E5+ the set of el-
ements whose eigenvalues are 52 and 52 (the others have eigenvalues 5
and 51 ). Set
R5 = {52 g | g E5+ }.
Then the elements of R5 are reflections of order 5 (with eigenvalues 1 and
5 , and determinant 5 ). We will show that

() W is generated by R5 .

Indeed, if W  denotes the subgroup of W generated by R5 and if + : W


W + denotes the canonical projection, then + (W  ) is the subgroup of W +
generated by E5+ . As E5+ is stable under conjugation, + (W  ) is a normal
subgroup of W + generated by elements of order 5, and hence is equal to
W + SL2 (F5 ) (see Theorem 1.2.4).
In particular, 120 divides |W  |, and W  is of index dividing 5 in W . On
the other hand the homomorphism det : W  5 is surjective, with kernel
W + W  . Therefore W + W  is a normal subgroup of W + of index dividing
5. Hence W + W  = W + by Theorem 1.2.4, that is to say W + W  and
|W  /W + | = 5. Hence W = W  , which shows (). The equalities D(W ) = W +
and Z(W ) = 10 now follow easily. Note that the order of a reflection is
equal to the order of its determinant, which shows that all reflections are of
order 5.
Now consider f : Ref(W ) W + , g det(g )2 g . The image f is indeed
contained in W + as det(g )5 = 1. On the other hand, if f (g ) = f (g  ), then
g = g  with 5 , which is only possible if = 1 or = det(g )1 (because
g and g  are reflections of order 5). Hence the fibres of f have cardinality at
most 2. Moreover, the image of f is contained in the set E5 of elements of
order 5 of W + . We conclude that, |Ref(W )|  2 |E5 | = 48.
If we denote by (d1 , d2 ) the sequence of degrees of W , then, by the
Shephard-Todd-Chevalley theorem (see Appendix C), we have

d1 d2 = |W | = 600, d1 + d2 2  48 and 10 divides d1 and d2 .

This forces (d1 , d2 ) = (20, 30) and |Ref(W )| = 48, as claimed. 



11.3 The Case when q = 5 141

E XAMPLE 11.3.12 In an analogous manner we can construct other reflec-


tion groups having W + as derived group. Let us view W + as a subgroup of
SL2 (Q( 60 )). Given an even divisor n of 60 let

Wn = W + n .

Then one can show, by similar arguments to those employed in the proof of
Proposition 11.3.11 (with some modifications . . . ), that

Wn is generated by reflections if and only if n {4, 6, 10, 12, 20, 30, 60}.
(n) (n)
If this is the case, denote by (d1 , d2 ) the sequence of degrees. These de-
grees together with the name of Wn in the Shephard-Todd classification are
given in Table 11.3. 

Table 11.3 Reflection groups having SL2 (F5 ) as derived group

(n) (n)
n |Wn | |Ref(Wn )| (d1 , d2 ) Z(Wn ) Standard name

4 240 30 (12, 20) 4 G22

6 360 40 (12, 30) 6 G20

10 600 48 (20, 30) 10 G16

12 720 70 (12, 60) 12 G21

20 1200 78 (20, 60) 20 G17

30 1800 88 (30, 60) 30 G18

60 3600 118 (60, 60) 60 G19

11.3.5. The Group PSL2 (F5 ), the Dodecahedron and the


Icosahedron

The two irreducible representations of SL2 (F5 ) of degree 3 can be realised


over the real numbers, as may be easily shown by calculating the Frobenius-
Schur indicator (which is 1 in both cases). Both representations contain I2
in their kernel, and therefore factorise to yield a representation of PSL2 (F5 )
with image contained in the special linear group (as SL2 (F5 ) is equal to its
derived group). Let us choose a homomorphism
142 11 Special Cases

: PSL2 (F5 ) SLR (V ),

where V is an R-vector space of dimension 3. We may fix an invariant scalar


product
, on V , so that gives a homomorphism to the group of oriented
isometries
: PSL2 (F5 ) SOR (V ,
, ).
Denote by E3 (respectively E3 ) the set of elements of order 3 (respectively 5)
of PSL2 (F5 ). We have

|E3 | = 20 and |E5 | = 24.

Every non-trivial element of SOR (V ,


, ) is the rotation around a well-
defined axis, and therefore the intersection of this axis with the unit sphere
S consists of two elements. Let

3 = {v S | g E3 , g v = v }

and 5 = {v S | g E5 , g v = v }.

If v 3 (respectively v 5 ), there exist two (respectively four) elements


of E3 (respectively E5 ) which fix it. In particular,

2 20 2 24
|3 | = = 20 and |5 | = = 12.
2 4
Proposition 11.3.13. The group PSL2 (F5 ) acts transitively on 3 and 5 .

Proof. Let v 3 . Denote by H the stabiliser, in PSL2 (F5 ), of v and let n = |H|.
Then H is isomorphic to a finite subgroup of the group of oriented isometries
of the orthogonal to v , which is of dimension 2. Hence H is cyclic. Moreover,
by construction, 3 divides n. The possible orders of the elements of PSL2 (F5 )
are 1, 2, 3, 4 or 5 and hence n = 3. Consequently, the orbit of v under PSL2 (F5 )
has cardinality 20 = |3 |. The result follows.
The transitivity of the action on 5 is shown in the same way.  

From Proposition 11.3.13 we may easily conclude the following corollary.

Corollary 11.3.14. 3 (respectively 5 ) is a regular dodecahedron (respectively an


icosahedron) and PSL2 (F5 ) is its group of oriented isometries.

11.4. The Case when q = 7

Hypothesis. In this and only this section we suppose that q = 7.


11.4 The Case when q = 7 143

11.4.1. Structure

We have

(11.4.1) | SL2 (F7 )| = 336, |PGL2 (F7 )| = 336 and |PSL2 (F7 )| = 168.

Recall that PSL2 (F7 ) is simple. However, contrary to the case of the groups
PSL2 (F3 ) and PSL2 (F5 ), PSL2 (F7 ) is not isomorphic to an alternating group.
We will show below (see Proposition 11.4.4) that

(11.4.2) PSL2 (F7 ) GL3 (F2 ).

11.4.2. Character Table

The character table 5.4 specialises as follows. Let j denote an element of 6


of order 3, j a linear character of order 3 of 6 , i a square root of 1 in 8 , i
a linear character of order 4 of8 , 8 an element of 8 of order 8, 8 a linear
character of 8 of order 8 and 2 = 8 (8 ) + 8 (8 )1 . We let

1 + 7 1 7
= and = .
2 2
The character table of SL2 (F7 ) is given in Table 11.4.

11.4.3. The Isomorphism Between the Groups PSL2 (F7 ) and


GL3 (F2 ).

By Table 11.4, the group SL2 (F7 ) admits two irreducible representations of
dimension 3 over K on which I2 acts trivially. These representations may
be realised in the -adic cohomology of the variety Y (with coefficients in a
sufficiently large extension of Q ) for all  = 7. In particular, we can take  = 2,
so that K is a finite extension of Q2 . Denote by O the ring of integers of K .
Proposition B.3.1, together with the fact that O is a principal ideal domain,
shows that there exists an injective homomorphism

: PSL2 (F7 )  GL3 (O)

such that

(11.4.3) Tr( (g )) = R+ (0 )(g )


144 11 Special Cases

Table 11.4 Character table of SL2 (F7 )

g I2 I2 d(j) d(j) d (i) d (8 ) d (83 ) u+ u u+ u

| ClG (g )| 1 1 56 56 42 42 42 24 24 24 24

o(g ) 1 2 3 6 4 8 8 7 7 14 14

CG (g ) G G T T T T T ZU ZU ZU ZU

1G 1 1 1 1 1 1 1 1 1 1 1

R+ (0 ) 3 3 0 0 1 1 1

R (0 ) 3 3 0 0 1 1 1

R+ (0 ) 4 4 1 1 0 0 0

R (0 ) 4 4 1 1 0 0 0

R (i ) 6 6 0 0 2 0 0 1 1 1 1

R (8 ) 6 6 0 0 0 2 2 1 1 1 1

R (83 ) 6 6 0 0 0 2 2 1 1 1 1

StG 7 7 1 1 1 1 1 0 0 0 0

R(j ) 8 8 1 1 0 0 0 1 1 1 1

R(j ) 8 8 1 1 0 0 0 1 1 1 1

for all g SL2 (F7 ) (we denote by g the image of g in PSL2 (F7 )). Recall that l
denotes the maximal ideal of O. We denote by

l : PSL2 (F7 ) GL3 (O/l)

the reductionmodulo l of .
Now (1+ 7)/2 is a root of the polynomial X 2 X +2 which splits mod-
ulo 2 and therefore (1 + 7)/2 Z2 . We conclude from 11.4.3 and the char-
acter table 11.4 that Tr(l (g )) F2 for all g PSL2 (F7 ). Furthermore, by [Isa,
Theorem 9.14(b)], after conjugating l by a suitable element of GL3 (O/ ),
we obtain a homomorphism

72 : PSL2 (F7 ) GL3 (F2 ).


11.4 The Case when q = 7 145

Proposition 11.4.4. The homomorphism 72 : PSL2 (F7 ) GL3 (F2 ) is an iso-


morphism of groups.

Proof. As |PSL2 (F7 )| = | GL3 (F2 )| = 168, it is enough to show that 72 is in-
jective. Because PSL2 (F7 ) is simple (see Theorem 1.2.4), it is enough to show
that Ker 72 = PSL2 (F7 ). But, by 11.4.3 and the character table 11.4, there
exists an element g PSL2 (F7 ) such that Tr(72 (g )) = 0. Therefore 72 (g )
is not the identify matrix, which completes the proof of the proposition.  

R EMARK 11.4.5 The group GL3 (F2 ) has a natural action on the set of 7
elements given by the projective plane P2 (F2 ). On the other hand, from the
point of view of PSL2 (F7 ), the most natural action is on the set of eight
elements given by the projective line P1 (F7 ). 

11.4.4. The Group PSL2 (F7 ) Z/2Z as a Reflection Group of


Rank 3

Let us fix an irreducible representation : PSL2 (F7 ) GL3 (C) with character
R+ (0 ). Denote by W + the image of . As is injective and PSL2 (F7 ) is equal
to its derived group, we have W + SL3 (C). Denote by 2C the subgroup of
GL3 (C) given by scalar transformations of order 1 or 2 and set

W = W + 2C PSL2 (F7 ) Z/2Z.

Then

(11.4.6) |W | = 336.

Let I denote the set of elements of W + of order 2. We have

(11.4.7) |I | = 21.

Set R = {s | s I }. Then

(11.4.8) |R| = 21.

Proposition 11.4.9. The elements of R are reflections, the group W is generated


by R and R = Ref(W ). The sequence of degrees of W is (4, 6, 14). The group W is
isomorphic to the complex reflection group G24 in the Shephard-Todd classification.

Proof. An element of order 2 in SL3 (C) necessarily has characteristic poly-


nomial (X + 1)2 (X 1). Hence, if we multiply such an element by 1 we
obtain a reflection (of order 2). Hence the elements of R are reflections.
Let W  =
R . Denote by : W W + the canonical projection. Then
(W  ) is a subgroup of W + containing I and is therefore normal and
146 11 Special Cases

non-trivial. The simplicity of PSL2 (F7 ) implies that (W  ) = W + . Hence


|W  | {168, 336}. Now, if |W  | = 168, then W  PSL2 (F7 ) and therefore
the morphism det is non-trivial on PSL2 (F7 ), which is impossible. Hence
|W  | = 336, and R generates W .
The group W is therefore generated by its reflections and R Ref(W ).
On the other hand, if s Ref(W ), then det(s) = 1, hence s W + and s
is an involution. We have therefore shown that Ref(W ) = R.
Now, denote by (d1 , d2 , d3 ) the sequence of degrees of W . Then, by the
Shephard-Todd-Chevalley theorem, d1 , d2 and d3 are even (as |Z(W )| = 2),
d1 + d2 + d3 3 = |Ref(W )| = 21 and d1 d2 d3 = |W | = 336. It is straightforward
to see that this forces (d1 , d2 , d3 ) = (4, 6, 14).
The last assertion is easily verified.  

Let us identify the algebra of polynomial functions on C3 with C[X , Y , Z ].


Proposition 11.4.9 shows that there exists a homogeneous polynomial P
C[X , Y , Z ]W of degree 4. This homogeneous polynomial defines a plane pro-
jective curve
C = {[x; y ; z] P2 (C) | P(x, y , z) = 0}.
One may verify that this plane projective curve is smooth. As the curve is of
degree d = 4, it has genus g = (d 1)(d 2)/2 = 3 [Har, Exercise I.7.2(b) and
Proposition IV.1.1]. Now, by a theorem of Hurwitz [Har, Exercise IV.2.5], we
have
| Aut C|  84 (g 1) = 168.
As the group PSL2 (F7 ) acts faithfully on on this curve, we deduce that

(11.4.10) Aut C PSL2 (F7 ) GL3 (F2 ).

Hence this curve provides an example in which the bound of Hurwitz is


attained. One may verify that in fact C is the Klein quartic (defined, for ex-
ample, in [Har, Exercise IV.5.7]).

Exercises

11.1. Suppose that q = 5. After numbering from 1 to 6 the points of P1 (F5 )


the action of G on P1 (F5 ) induces a homomorphism G
S6 , which we de-
note by . Set = Im PGL2 (F5 ).
(a) Let n be a non-zero natural number and let H be a subgroup of Sn of
index n. Denote by 0 : Sn Sn the morphism induced by the action of
Sn on Sn /H (after enumerating the elements of Sn /H from 1 up to n).
(a1) Show that 0 is an automorphism. (Hint: If n = 4, use the simplicity
of An and for n = 4, note that A4 does not have a subgroup of order 6.)
11.4 The Case when q = 7 147

(a2)Show that, if n is the number corresponding to H in Sn /H, then 0 (H)


is the stabiliser of n in Sn . Deduce that H Sn1 .
(a3) Show that if H is a transitive subgroup of Sn , then 0 is not an inner
automorphism.
(b) Deduce from (b) that S5 and that S6 has a non-inner automorphism.
(c) Deduce a new proof that PSL2 (F5 ) A5 (see 11.3.2).

11.2*. Let be a finite group. Show that:


(a) If is simple of order 60, then PSL2 (F5 ) A5 .
(b ) If is simple of order 168, then PSL2 (F7 ) GL3 (F2 ).
(c ) If is simple of order 360, then PSL2 (F9 ) A6 .
(d )If is simple of order 504, then PSL2 (F8 ) = SL2 (F8 ).
(e ) If is simple of order 660, then PSL2 (F11 ).
(f ) If is simple of order  1000, then is isomorphic to PSL2 (Fq ), for some
q {5, 7, 8, 9, 11}.

11.3. Consider
: SL 3 )
 2 (F F3+
ab
(a2 + c 2 )(ab + cd).
cd
Show that is a homomorphism of groups and that, if x F3 , then (u(x)) =
x. Show also that N  = Ker .

11.4. Let Q8 be the quaternionic group of order 8 and let K be a field of


characteristic zero. Suppose that there exists an irreducible representation
of Q8 over K of dimension 2. Show that 1 is a sum of two squares in K.
(Hint: Set Q8 = {1, I , J, K } (using standard notation) and fix an ir-
reducible representation
  : Q8 GL2 (K). Show that  we  may always sup-
0 1 ab
pose that (I ) = and that, if we set (J) = , then the relations
1 0 cd
(J)2 = (I )2 and ( (I ) (J))2 = (I )2 force d = a, b = c and a2 + b 2 = 1.)

R EMARK Denote by H the standard algebra of quaternions over a field


K of characteristic different from 2. That is H = K KI KJ KK , where
I 2 = J 2 = K 2 = 1, IJ = JI = K , JK = KJ = I and KI = IK = J. Then H
is split over K (that is, isomorphic to Mat2 (K)) if and only if 1 is a sum of
two squares in K. This result appears similar to that shown in Exercise 11.4
above. In fact, the two results are equivalent if we view the algebra H as a
quotient of the group algebra KQ8 by the ideal generated by z + 1, where z
is the unique non-trivial central element of Q8 .
11.5*. Let K be a field of characteristic different from 2. Show that SL2 (F5 )
admits an irreducible faithful representation of dimension 2 over K if and
only if 5 is a square in K and 1 is a sum of two squares in K.
148 11 Special Cases

11.6 (McKay correspondence). Let be a finite subgroup of SL2 (C) and nat
the character of the natural representation  SL2 (G). Define a graph as
follows: the set of vertices of G is Irr and two characters and  are con-
nected by an edge if is an irreducible factor of  nat . We denote by G the
full subgraph of G obtained by removing the vertex 1 .
(a) Show that nat is real valued (use Exercise 1.13 to show that nat = nat ).
 
(b) Deduce that, if , Irr , then is an irreducible factor of nat if and
only if  is an irreducible factor of nat .
Denote by A 4 (respectively A 5 ) the subgroup of SL2 (C) given by the im-
age of SL2 (F3 ) (respectively SL2 (F5 )) under the irreducible representation of
dimension 2 with character R (i ) (respectively R+ (0 )). Denote by S 4 the
normaliser of A 4 in SL2 (C). For background on Dynkin diagrams and root
systems, see [Bou, Chapter VI].
4.
(c) Calculate the character table of S
(d) Show that GA (respectively G , respectively GA
S
) is a graph of type
4 4 5
E6 (respectively E7 , respectively E8 ). Here, E? denotes the affine Dynkin
diagram associated to the Dynkin diagram of type E? .
(e) Show that GA 4 (respectively GS 4 , respectively GA 5 ) is a Dynkin diagram
of type E6 (respectively E7 , respectively E8 ).
(f) Generalise the above to all finite subgroups of SL2 (C).

11.7. Show that the group of outer automorphisms of A6 is isomorphic to


Z/2Z Z/2Z. (Hint: Use Example 11.3.8 and Exercise 1.14.)
Chapter 12
Deligne-Lusztig Theory: an Overview*

This chapter gives a very succinct overview of Deligne-Lusztig theory.


We will recall some of the principal results of the theory (including the
parametrisation of characters and partition into blocks) with the goal of con-
necting this general theory with what we have seen for SL2 (Fq ). This chapter
requires some knowledge of algebraic groups, for which we refer the reader
to [Bor] or [DiMi].
For more details on the subjects covered in this chapter the reader is re-
ferred to the books [Lu1], [Carter], [DiMi] or [CaEn]. We also recommend
the magnificent and foundational article on the subject, written by Deligne
and Lusztig in 1976 [DeLu].

Hypotheses and notation. In this and only this chapter we fix


a reductive group G defined over a field F as well as a Frobe-
nius endomorphism F : G G which equips G with a rational
structure over the finite field Fq . Denote by B an F -stable Borel
subgroup of G, T an F -stable maximal torus of B and U the
unipotent radical of B. The Weyl group of G relative to T will be
denoted W ; by definition, W = NG (T).
Furthermore, we return to the assumption that  is a prime
number different from p.

The finite reductive group is the group GF of fixed points of F over G:

GF = {g G | F (g ) = g }.

The object of Deligne-Lusztig theory is the study of ordinary and modular


representations (in unequal characteristic) of the finite group GF . Recall one
of the fundamental theorems concerning algebraic groups defined over a
finite field.

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 149


DOI 10.1007/978-0-85729-157-8_12, Springer-Verlag London Limited 2011
150 12 Deligne-Lusztig Theory: an Overview*

Langs theorem. If H is a connected algebraic group and if F : H H is a Frobe-


nius endomorphism of H, then the morphism H H, h h1 F (h) is an unrami-
fied Galois covering with group HF ; in particular, it is surjective.

12.1. Deligne-Lusztig Induction

If w W , we fix a representative w of w in NG (T) and denote by wF : T


T, t w F (t)w 1 . Then wF is a Frobenius endomorphism T. The Deligne-
Lusztig variety is the variety

Y(w ) = {g U G/U | g 1 F (g ) Uw U}.

The action of GF by left translations on G/U stabilises Y(w ) as does the ac-
tion of TwF by right translations on G/U (recall that T normalises U). Hence
Y(w ) comes equipped with the structure of a (GF , TwF )-variety and its coho-
mology groups Hci (Y(w )) inherit the structure of a (K GF , K TwF )-bimodule.
We define
Rw : K0 (K TwF ) K0 (K GF )
[M ]K TwF (1) i
[Hc (Y(w )) K TwF
i
M ]K GF
i 0

and
R : K0 (K GF ) K0 (K TwF )
w
[M ]K GF (1) [Hci (Y(w )) K GF M ]K GF
i
i 0

The morphisms Rw and R are called Deligne-Lusztig induction and restric-


w
tion respectively. One may easily verify (thanks to Langs Theorem, see Ex-
ercise 12.1) that these morphisms depend only on w and not on the choice
of a representative w . They are adjoint (for the standard scalar product on
the Grothendieck groups).
If w W and if is a linear character of TwF , the (virtual) character Rw ( )
of GF is called a Deligne-Lusztig character. One of the first fundamental re-
sults of this theory is the Mackey formula [DeLu, Theorem 6.8].

Mackey formula. Let w and w  be two elements of W and let and  be two

linear characters of TwF and Tw F respectively. Then

(12.1.1)
Rw ( ), Rw  (  ) GF = |{x W | xwF (x)1 = w  and  = x}|.

R EMARK Note that two Deligne-Lusztig characters can be orthogonal


and yet have common irreducible factors. Indeed, such characters are vir-
tual characters. 
12.1 Deligne-Lusztig Induction 151

The next corollary follows from Mackey formula by computing the norm
of the difference of the two virtual characters involved.

Corollary 12.1.2. If x and w are two elements of W , then

Rx 1 wF (x) = Rw x ,
1 wF (x)
where x : (Tx ) (TwF ) , x 1 .

The second fundamental result is the following [DeLu, Corollary 7.7].

Theorem 12.1.3. If is an irreducible character of GF , then there exists w W


and a linear character of TwF such that
Rw ( ), GF = 0.

It follows that, in order to parametrise the irreducible characters of GF , it


is enough to decompose the Deligne-Lusztig characters. This enormously
difficult work was completed by Lusztig in 1984 (for groups with connected
centre, which constitutes the largest part of the work). An important notion
used to achieve this parametrisation is that of geometric conjugacy. In order to
define this, let us introduce an integer n0  1 such that the Frobenius endo-
morphisms (w F )n0 and F n0 of G agree for all w W and induce the identify
on W (such an n0 always exists). Denote by
n0
Nw : T F TwF
t t wF t (wF )n0 1 t

the norm map.


Let (G, F ) denote the set of couples (w , ), where w W , and is a
linear character of TwF . If (w , ) and (w  ,  ) are two elements of (G, F ),
we say that (w , ) and (w  ,  ) are geometrically conjugate if the characters
Nw and  Nw  of TF 0 are conjugate under W . If this is the case we write
n

(w , ) (w  ,  ). A geometric series is an equivalence class for the relation .


The proof of the following theorem may be found in [DeLu, Theorem 6.1].

Theorem 12.1.4. Let (w , ) and (w  ,  ) be two elements of (G, F ) such that


Rw ( ) and Rw  (  ) contain a common irreducible factor. Then (w , ) (w  ,  ).

If S is a geometric series, we set

E (GF , S ) = { Irr GF | (w , ) S ,
Rw ( ), GF = 0}.

The set of characters E (GF , S ) is called a Lusztig series. Combining Theo-


rems 12.1.3 and 12.1.4, we obtain


(12.1.5) Irr GF = E (GF , S ).
S (G,F )/
152 12 Deligne-Lusztig Theory: an Overview*

Once one has obtained this initial description, all that remains to do is to
parametrise the characters in a given Lusztig series. This was completed by
Lusztig in a number of very long articles and a book [Lu2].
If (w , ) (G, F ), denote by W (w , ) the stabiliser, in the group W , of the
n
linear character Nw of TF 0 . Then W (w , ) is a wF -stable subgroup of W .
If the centre of G is connected, then W (w , ) is a subgroup of W generated
by reflections.

Theorem 12.1.6 (Lusztig). Let S be a geometric series and let (w , ) S . If the


centre of G is connected, then E (GF , S ) is in bijection with a subset which only
depends on the pair (W (w , ), wF ), and not on the group G or the cardinality q of
the finite field.

E XAMPLE 12.1.7 If W (w , ) = 1, then |E (GF , S )| = 1 and

E (GF , S ) = { (w )Rw ( )},

where (w ) is the sign of w . 

Via the bijection of Theorem 12.1.6, Lusztig gives formulas for the degrees
of the characters, for their multiplicities in all Deligne-Lusztig characters etc.
He also deduces an analogous result for groups with non-connected centre,
but for this one needs further techniques (see [Lu3]) which we will not re-
call here. In the group G = SL2 (Fq ) the non-connectedness of the centre is
responsible, for example, both for the decomposition of the characters R(0 )
and R (0 ) and for the existence of quasi-isolated blocks.
We call a unipotent character of GF any element of E (GF , S1 ), where S1
is the geometric series {(w , 1) | w W }. In this case W (w , 1) = W and we
conclude from Theorem 12.1.6 that

(12.1.8) |E (GF , S1 )| only depends on the pair (W , F ), not on q.

E XAMPLE 12.1.9 Let us suppose that G = SL2 (F) and that F : G G is the
natural split Frobenius endomorphism
   q q
ab a b
F = q q .
cd c d

Then GF is our favourite finite group G = SL2 (Fq ). Moreover, W = {1,s},


where s denotes the class of s in NG (T)/T. We have

TF = T and Y(1) = GF /UF = G /U

therefore Hc (Y(1)) = K [GF /UF ] by Theorem A.2.1(c), which shows that the
Deligne-Lusztig induction R1 is nothing but the Harish-Chandra induction
R defined in Section 3.2.
12.1 Deligne-Lusztig Induction 153

Let us now turn to to the Deligne-Lusztig induction map associated to s.


On the other hand, it is a well-known geometrical fact that the morphism of
varieties
G/U A (F) \ {(0, 0)}
2

ab
(a, c)
cd
is an isomorphism. Via this isomorphism we have

TsF q+1 and Y(s) Y,

the isomorphism being compatible with the actions of G , q+1 and F . As a


consequence,
Rs = R .
The map R is therefore none other (up to a sign) than the Deligne-Lusztig in-
duction associated to the non-trivial element of the Weyl group. The reader
may verity that the above Mackey formula 12.1.1 is then a condensed ver-
sion of the various Mackey formulas (3.2.4, 4.1.7 and 4.2.1) obtained for our
group SL2 (Fq ). Moreover,

E (G , S1 ) = {1, StG },

which allows one to verify 12.1.8.


Furthermore, (1, 1) (s, 1), (1, 0 ) (s, 0 ) and

W if 2 = 1,
W (w , ) =
1 if 2 = 1.

One may then verify the above claims concerning Lusztig series. 

We finish this section by recalling some geometric properties of the vari-


eties Y(w ).

Proposition 12.1.10. Let w W . Then:


(a) The variety Y(w ) is quasi-affine, smooth, and purely of dimension l(w ) =
dim Bw B/B.
(b) The group TwF acts freely on Y(w ).
(c) The stabilisers, in GF , of elements of Y(w ) are p-groups.

R EMARK It is conjectured that Deligne-Lusztig varieties are always affine


(and not only quasi-affine) but for the moment this is only known when
q is large enough [DeLu, Theorem 9.7] and in certain particular cases. The
smallest examples where it is not known if the varieties Y(w ) are affine occur
in the group of type G2 , when q = 2 and l(w ) = 3. 
154 12 Deligne-Lusztig Theory: an Overview*

12.2. Modular Representations

Recall that

 is a prime number different from p.

Because of its use of -adic cohomology and geometric methods, Deligne-


Lusztig theory is particularly well adapted to the study of representations
in unequal characteristic. Here we will recall some facts which illustrate this
phenomenon.

12.2.1. Blocks

If (w , ) (G, F ), we denote by  the  -part of . A pair (w , ) is called


-regular if =  . We denote by  (G, F ) the set of -regular elements of
(G, F ). If S is a geometric series, we denote by

S = {(w ,  ) | (w , ) S }.

Then S is also a geometric series. A geometric series S is called -regular


if S = S . If S is an -regular geometric series, we denote by


E (GF , S ) = E (GF , S  ).
S  (G,F )/
S  =S


The following result is due to Brou and Michel [BrMi, Theorem 2.2].

Theorem 12.2.1 (Brou-Michel). If S is an -regular geometric series, then


E (GF , S ) is a union of -blocks.

In other words, if we set


()
eS = e ,
E (GF ,S )

then Theorem 12.2.1 of Brou and Michel may be translated as


()
(12.2.2) eS OGF .
12.2 Modular Representations 155

12.2.2. Modular Deligne-Lusztig Induction

If w W , we denote by Rw and Rw the functors of Deligne-Lusztig induction


and restriction:
Rw : Db (OTwF ) Db (OGF )
M Rc (Y(w ), O) O TwF M

Rw : Db (OGF ) Db (OTwF )
M Rc (Y(w ), O) O GF M.
These functors are well-defined and adjoint to one another. Indeed, the com-
plex of bimodules Rc (Y(w ), O) is perfect as a left and right module by
virtue of Proposition 12.1.10 and Theorem A.1.5.
If {K , O, k}, we denote by Rw and Rw the extension of scalars
of the functors Rw and Rw respectively. For example, the functor K Rw
(respectively K Rw ) induces the morphism Rw (respectively Rw ) between
Grothendieck groups.
The Morita equivalence of Theorems 8.1.1 and 8.1.4 may be generalised
to the case of arbitrary finite reductive groups in the following way [Bro,
Theorem 3.3].

Theorem 12.2.3 (Brou). Let (w , )  (G, F ) such that W (w , ) = 1 and


let S be the geometric series which contains it. Then the cohomology group
l(w ) () ()
Hc (Y(w ), O) induces a Morita equivalence between OTwF e and OGF eS .
()
In Theorem 12.2.3, e denotes the primitive central idempotent of OTwF
()
such that Irr K TwF e .

12.2.3. The Geometric Version of Brous Conjecture

We will conclude this overview with a geometric version of Brous conjec-


ture (the general conjecture is given in Appendix B). For this, we make the
following hypotheses, which allow us to considerably simplify the exposi-
tion.
156 12 Deligne-Lusztig Theory: an Overview*

Hypotheses and notation. In this and only this subsection we


suppose that:
(1) F acts trivially on W (if G is semi-simple, this is equivalent
to saying that F is a split Frobenius endomorphism of G).
(2)  does not divide |W |.
(3) Denote by d the order of q modulo : we suppose that d is a
regular number for W in the sense of Springer [Spr, 4].

Brous conjecture (geometric version). There exists an element w W of or-


der d and regular in the sense of Springer (see [Spr, 4]), as well as a complex of
(OGF , ONGw F (T))-bimodules Cw such that:
GF (N w F (T))opp
G
(a) ResGF (TwF )opp )
Cw D Rc (Y(w ), O).
(b) Cw induces a Rickard equivalence between the principal blocks of GF and
NGw F (T).

R EMARK Note that, under the previous assumptions, if S denotes a Sylow


-subgroup of GF , then NGF (S) is isomorphic to NGw F (T). 

This version has been shown in very few cases. The case where  divides
q 1 was shown by Puig (see, for example, [CaEn, Theorem 23.12]):

Theorem 12.2.4 (Puig). If d = 1 (that is to say if  divides q 1), then the ge-
ometric version of Brous conjecture is true. In this case the induced equivalence
is even a Morita equivalence, because the Deligne-Lusztig variety in question is of
dimension zero.

The next result was proved by Rouquier and the author [BoRo, Theo-
rem 4.6].

Theorem 12.2.5. If GF = GLn (Fq ), SLn (Fq ) or PGLn (Fq ) and if d = n, then the
geometric version of Brous conjecture is true.

E XAMPLE In the case where GF = SL2 (Fq ), we have already seen Puigs
theorem (see Corollary 8.3.3) and Theorem 12.2.5 (see Corollary 8.3.7). 
Very recently, Olivier Dudas has shown in his thesis [Du] the final result
of this chapter.

Theorem 12.2.6 (Dudas). If G is of type Bn , Cn and d = 2n, then the geometric


version of Brous conjecture is true.
12.2 Modular Representations 157

Exercises

12.1. Let w and w be two representatives of the same element w W in


NG (T). Show that the (GF , TwF )-varieties Y(w ) and Y(w ) are isomorphic.
Deduce that the morphisms Rw and Rw (as well as the functors Rw and
R ) do not depend on a choice of representative w of w . (Hint: Use Langs
w
theorem.)

12.2. Verify the assertions of Example 12.1.9. Use the Lang map G G, g
g 1 F (g ) and Langs theorem to show that Y/G A1 (F).
Appendix A
-Adic Cohomology

The tale topology, the fundamentals of which were developed in the Smi-
naire de Gomtrie Algbrique du Bois-Marie in the 1960s (see, for example,
[SGA1], [SGA4], [SGA4 12 ]) allows one to associate functorially to any alge-
braic variety a complex and corresponding cohomology groups (with co-
efficients in , which can be K , O or O/ln , for n  1) enjoying numerous
remarkable properties. In this appendix we recall those properties which
are essential to our goal of studying the representations of SL2 (Fq ) (func-
toriality, perfectness, exact sequences, bounds, Knneth formula, Poincar
duality, traces, the Lefschetz fixed-point theorem, . . . ). We refer the reader to
[SGA4 12 ] for all results stated without reference.

A.1. Properties of the Complex*

The complex of -adic cohomology inherits an action of the monoid of endo-


morphisms of the variety. Even better, it can be realised as a complex of mod-
ules for this monoid (see the works of Rickard [Ric3] and Rouquier [Rou3]).
Therefore, in this appendix we work under the following hypotheses. Let
V be a quasi-projective algebraic variety defined over F and let be a monoid
acting on V via endomorphisms. We also set to be one of K , O or O/ln (for
n  1).

E XAMPLE We can take V = Y (the Drinfeld curve) and = G (q+1 

F mon ). 

By [Rou3], there exists a bounded complex Rc (V, ) of -modules


whose cohomology groups, which we denote Hci (V, ), are, as -modules,
the cohomology groups with compact support of the variety V (with coeffi-
cients in the constant sheaf ). To simplify notation we denote by Hci (V) the
K -module Hci (V, K ).

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 159


DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011
160 A -Adic Cohomology

Recall that

(A.1.1) Rc (V, ) LO Rc (V, O).

Here, LO denotes the derived functor of the tensor product. In particular, as


K is O-flat,

(A.1.2) Rc (V) K O Rc (V, O)

and

(A.1.3) Hci (V, K ) K O Hci (V, O).

Moreover, by [SGA4 12 , Arcata, III, 3],

(A.1.4) if V is a smooth curve, then Hci (V, O) is torsion-free.

Another essential property is given in the following proposition.

Theorem A.1.5 (Rickard, Rouquier). Suppose that is a finite group. Then:


(a) Rc (V, ) is homotopic to a bounded complex of -modules of finite type.
(b) If the stabilisers of points of V are of order invertible in , then Rc (V, ) is
homotopic to a bounded complex of projective -modules.

A.2. Properties of the Cohomology Groups

A.2.1. General Properties

We begin by recalling certain properties of cohomology groups with coeffi-


cients in the general ring .

Theorem A.2.1. Let d = dim V and let I (V) denote the set of irreducible compo-
nents of V of dimension d. Then:
(a) Hci (V, ) is a -module of finite type.
(b) Hci (V, ) = 0 if i < 0 or if i > 2d. If V is affine and purely of dimension d, then,
moreover, Hci (V) = 0 if i < d.
(c) If is a group we have an isomorphism of -modules Hc2d (V) [I (V)].
(d) If U is a -stable open subvariety of V with closed complement Z, then we have
a long exact sequence of -modules

Hci (U, ) Hci (V, ) Hci (Z, ) Hci+1 (U, )

(e) If is contained in a connected algebraic group acting regularly on V, then


acts trivially on Hci (V, ).
A.2 Properties of the Cohomology Groups 161

if i = 2d,
(f) Hci (Ad (F), ) =
0 otherwise.

Proof. (a), (b), (c), (d) and (f) are shown in [SGA4 12 ]. (e) is shown in [DeLu,
Proposition 6.4].  

A.2.2. Cohomology with Coefficients in K

We will now state some properties which hold when = K .

Knneth formula. If V is another variety on which acts, then we have an


isomorphism of K -modules
r
(A.2.2) Hcr (V V ) Hci (V) K Hcr i (V )
i=0

for all r  0.
If is a finite subgroup of which is normalised by (i.e. = for all
), then the quotient monoid / acts on the quotient variety V/ and
we have an isomorphism of (/)-modules

(A.2.3) Hci (V/) Hci (V) .

Poincar duality. If V is irreducible, projective and smooth of dimension d,


then we have, for all i {0, 1, 2, ... , 2d}, a perfect -equivariant duality

(A.2.4) Hci (V) Hc2di (V) Hc2d (V).

R EMARK Note that, under the above hypotheses, dimK Hc2d (V) = 1 (see
Theorem A.2.1 (c)). However, it is possible for not to act trivially on
Hc2d (V). For example, a Frobenius endomorphism over Fq acts as multipli-
cation by q d (see Theorem A.2.7(b) below). 

A.2.3. The Euler Characteristic

It is often useful to consider the Euler characteristic of V. If one takes into


account the action of the monoid , the Euler characteristic becomes not just
a number, but an element in the Grothendieck group K0 (K ) of the category
of K -modules of finite dimension over K defined by

Hc (V) = (1)i [Hci (V)] .


i 0
162 A -Adic Cohomology

We use the notation Hc (V) if we wish to emphasise the monoid .


If End(V) or , we denote by TrV ( ) the alternating sum

TrV ( ) = (1)i Tr( , Hci (V)).


i 0

The function TrV : End(V) K is called the Lefschetz character of V. Recall


that, if is an automorphism of V of finite order, then

(A.2.5) TrV ( ) Z.

Theorem A.2.6. Let d = dim V. Then:


(a) If U is an open -stable subvariety of V and if Z denotes its closed complement,
then Hc (V) = Hc (U) + Hc (Z). In particular, TrV = TrU + TrZ .
(b) If is a finite subgroup of normalised by , then Hc (V/)/ = Hc (V) .
(c) If V is an algebraic variety upon which also acts, then Hc (V V ) = Hc (V)
Hc (V ).
(d) If s and u are two invertible elements of such that su = us, s is of order prime
to p and u is of order a power of p, then

TrV (su) = TrVs (u).

(e) If is a finite group and S is a torus acting on V and commuting with the
action of , then Hc (V) = Hc (VS ).

Proof. (a), (b) and (c) may be found in [SGA4 12 ], while (d) is shown in [DeLu,
Theorem 3.2] and (e) is shown, for example, in [DiMi, Proposition 10.15].  

A.2.4. Action of a Frobenius Endomorphism

Suppose that V is defined over the finite field Fq , with associated Frobenius
endomorphism F . Then F acts on the cohomology groups Hci (V). We assem-
ble here some classical results concerning this action.

Theorem A.2.7. We have:


(a) TrV (F ) = |VF | (Lefschetz fixed-point theorem).
(b) If V is irreducible of dimension d, then F acts on Hc2d (V) as multiplication by
qd .
(c) The eigenvalues of F on Hci (V) are algebraic integers of the form q j/2 , where
j is a natural number such that 0  j  i, and is an algebraic number, all of
whose complex conjugates are of norm 1. In particular, F is an automorphism of
the K -vector space Hci (V).
(d) If V is projective and smooth, then the eigenvalues of F on Hci (V) are alge-
braic integers, all of whose complex conjugates have norm q i/2 .
A.3 Examples 163

Proof. (a) and (b) are shown in [SGA4 12 ]. (c) and (d) have been shown twice
by Deligne [De1], [De2]: these results constitute the last difficulty in resolv-
ing the celebrated Weil conjectures, which are analogues for algebraic vari-
eties of the Riemann hypothesis for number fields.  

A.3. Examples

A.3.1. The Projective Line

We identify A1 (F) with the open subvariety of P1 (F) equal to {[x; y ]


P1 (F) | y = 0} and set = [1; 0]. By Theorem A.2.1(d), we have an exact
sequence

0 Hc0 (A1 (F), ) Hc0 (P1 (F), ) Hc0 (, )

Hc1 (A1 (F), ) Hc1 (P1 (F), ) Hc1 (, )


Hc2 (A1 (F), ) Hc2 (P1 (F), ) Hc2 (, ) 0.
Using Theorem A.2.1(f) for d = 0 and d = 1, we obtain an exact sequence

0 0 Hc0 (P1 (F), )

0 Hc1 (P1 (F), ) 0


Hc2 (P1 (F), ) 0 0.
We conclude:

if i = 0 or 2,
(A.3.1) Hci (P1 (F), ) =
0 otherwise.

On the other hand, the connected group GL2 (F) acts regularly on P1 (F),
therefore, by Theorem A.2.1(e),

(A.3.2) GL2 (F) acts trivially on Hci (P1 (F)).

To conclude, we can keep track of the action of the Frobenius endomorphism


F : P1 (F) P1 (F), [x; y ] [x q ; y q ] in the above exact sequences. Using The-
orem A.2.7(b), we obtain

(A.3.3) F = 1 on Hc0 (P1 (F)) and F = q on Hc2 (P1 (F)).

In particular, we have verified the Weil conjectures (Theorem A.2.7(c)) and


the Lefschetz fixed-point theorem (Theorem A.2.7(a)) which says that
164 A -Adic Cohomology

|P1 (Fq )| = q + 1,

which is no surprise.

A.3.2. The One-Dimensional Torus

Denote by U = A1 (F) \ {0}. By Theorem A.2.1(d), we have an exact sequence

0 Hc0 (U, ) Hc0 (A1 (F), ) Hc0 (0, )

Hc1 (U, ) Hc1 (A1 (F), ) Hc1 (0, )


Hc2 (U, ) Hc2 (A1 (F), ) Hc2 (0, ) 0.
Using Theorem A.2.1(f) for d = 0 and d = 1, we therefore obtain an exact
sequence
0 Hc0 (U, ) 0
Hc1 (U, ) 0 0
Hc2 (U, ) 0 0.
We conclude:

if i = 1 or 2,
(A.3.4) Hci (A1 (F) \ {0}, ) =
0 otherwise.

On the other hand, the connected group F acts regularly (by multiplication)
on A1 (F) \ {0}, therefore, by Theorem A.2.1(e),

(A.3.5) F acts trivially on Hci (A1 (F) \ {0}).

To conclude, we can keep track of the action of the Frobenius endomorphism


F : A1 (F) A1 (F), x x q in the above exact sequence. Using Theorem
A.2.7(b), we obtain

(A.3.6) F = 1 on Hc1 (A1 (F) \ {0}) and F = q on Hc2 (A1 (F) \ {0}).

The Lefschetz fixed-point theorem (Theorem A.2.7(a)) tells us that

|A1 (Fq ) \ {0}| = q 1,

which is again no surprise.


A.3 Examples 165

Exercises

A.1. Determine the cohomology of Pn (F).

A.2. Let a1 ,. . . , an be pairwise distinct points of A1 (F). Calculate the coho-


mology of A1 (F) \ {a1 , ... , an }.

A.3. In statement (c) of Theorem A.2.7, the algebraic number is not neces-
sarily an algebraic integer. This may occur and when it does it considerably
restricts the values of . Indeed, show that an algebraic integer (over Z) all of
whose complex conjugates are of norm 1 is a root of unity.
Appendix B
Block Theory

Fix a finite group . In this appendix we recall the essential results (that
is, essential for our purposes) of block theory, where the object of study is
the representations of the algebras O and k. As in the rest of this book we
assume that the algebras K  and k are split for all groups  met in this
appendix. We denote by O k, a a the reduction modulo l, and extend
this notation to O-modules.
For further details and developments the reader may consult one of the
many references on this subject: [Alp], [CuRe, Chapters 2 and 7], [Isa, Chap-
ter 15], [NaTs], [Ser, Part 3], [The].

B.1. Definition

Set = O or = k. A -bloc of is an indecomposable direct summand of


the (, )-bimodule . If we decompose the (, )-bimodule as a
direct sum of indecomposable factors in two different ways

= A1 Ar = A1 As ,

and write 1 = e1 + + er = e1 + + es , where ei Ai and ej Aj . Then, the
next proposition follows from [The, Corollary 4.2].

Proposition B.1.1. With the above notation we have:


(a) r = s and there exists a permutation of {1, 2, ... , r } such that Ai = A (i) for
all i (and therefore ei = e (i) ).
(b) ei is a primitive central idempotent of and Ai = ei ; furthermore, Ai is a
-algebra with identity ei .
(c) If i = j, then ei ej = ej ei = 0.
(d) The -algebras and A1 Ar are isomorphic.

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 167


DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011
168 B Block Theory

The previous proposition shows that the set of -blocks of is well-


defined. The following proposition (which follows from results about lifting
idempotents [The, Theorems 3.1 and 3.2]) shows that it does not depend too
much on .

Proposition B.1.2. If A is an O-block of , then A is a k-block of . If e is a primi-


tive central idempotent of O, then e is a primitive central idempotent of k.
The reduction modulo l induces a bijection between the O-blocks and k-blocks of
. It also induces a bijection between the primitive central idempotents of O and
of k.

The isomorphism of O-algebras O A1 Ar of Proposition B.1.1


induces an isomorphism K KA1 KAr and hence induces a partition

(B.1.3) Irr K = Irr(KA1 ) Irr(KAr ).

We call an -block of a subset of Irr K of the form Irr(KA), where A is an


O-block of . If is an irreducible character of and if a Z(O), we denote
by (a) the scalar by which a acts on an irreducible representation afford-
ing as character (by virtue of Schurs lemma). Then : Z(O) O is a
morphism of O-algebras which will be called the central character associated

to . If X is a subset of , we set X = X O. Recall that (Cl ( )) [/]
is a O-basis of Z(O). Moreover, recall [Isa, Theorem 3.7] that

 ( ) | Cl ( )|
(B.1.4) (Cl ( )) = O.
(1)

The following proposition gives a characterisation of the partition into


-blocks of Irr K using central characters (see, for example, [Isa, Theo-
rem 15.18]).

Proposition B.1.5. Let and  be two irreducible characters of . Then the fol-
lowing are equivalent:
(1) and  are in the same -block.
(2) (a)  (a) mod l for all a Z(O).
( ) | Cl ( )|  ( ) | Cl ( )|
(3) mod l for all .
(1)  (1)

B.2. Brauer Correspondents

B.2.1. Brauers Theorems

If D is a subgroup of , we define
B.2 Brauer Correspondents 169

BrD : Z(k) Z(kN (D))


a C (D) a .

Then [Isa, Lemma 15.32] we have the following.


Proposition B.2.1. If D is an -subgroup of , then BrD is a morphism of k-
algebras.
When D is an -subgroup of , BrD is called the Brauer morphism. If A is
an O-block, we call the defect group of A any -subgroup D of such that
BrD (A) = 0 and which is maximal with this property. Then [Isa, Lemma
15.33].
Proposition B.2.2. If D and D  are two defect groups of an O-block of , then D
and D  are conjugate in .
Proposition B.2.2 justifies the abuse of language committed by referring
to the defect group of A. The following proposition relates the -valuation
of the degree of a character to that of the order of the defect group of the
corresponding -block [Isa, Theorem 15.41].
Proposition B.2.3. If D is the defect group of A and if Irr(KA), then

|D| (1)
O.
||

Moreover, there exists at least one character Irr(KA) such that

|D| (1)
O .
||

E XAMPLE B.2.4 Let an irreducible character of such that ||/ (1) O .


Then e Z(O) and therefore e is a primitive central idempotent of O
(because it is already primitive in K ). As a consequence, is alone in its
-block. Moreover, Proposition B.2.3 shows that the defect group of Oe is
trivial. 

For a proof of the following theorem the reader is referred to [The, Corol-
lary 37.13] or [Isa, Theorem 15.45].
Brauers first main theorem. Let D be an -subgroup of . The map BrD induces
a bijection between the set of O-blocks of with defect group D and the set of O-
blocks of N (D) with defect group D.
If A is an O-block of with defect group D and if A is the O-block of
N (D) associated to A by the bijection of Brauers first main theorem, we
say that A is the Brauer correspondent of A. It is characterised as follows: A is

the unique O-block of N (D) such that BrD (A)A = 0. If we denote by e (re-
spectively e  ) the primitive central idempotent of O (respectively ON (D))
170 B Block Theory

such that A = Oe (respectively A = ON (D)e  ), then e  is characterised by


the property that BrD (e) = e  .
We call the principal block of the unique O-block A of such that 1
Irr(KA). We denote the principal block of by B0 ().

Brauers third main theorem. Let S be a Sylow -subgroup of . Then S is the


defect group of B0 () and B0 (N (S)) is the Brauer correspondent of B0 ().

B.2.2. Conjectures

One of the central problems in block theory is to relate the representations


in an O-block to those of its Brauer correspondent. There exist diverse con-
jectures in this direction. We will be content state the following.

McKay conjecture (global). Denote by Irr () the set of characters of of degree


prime to  and let S be a Sylow -subgroup of . Then
 
| Irr (K )| = | Irr K N (S) |.

In case the defect group is abelian, Brou proposed the following conjec-
ture. It is of a much more structural nature, but there does not exist a version
if the defect group is not abelian.

Brous conjecture. Let A be a block of with defect group D and let A be its
Brauer correspondent. If D is abelian, then the derived categories Db (A) and Db (A )
are equivalent as triangulated categories.

This is not the place to enter into an exhaustive description of the nu-
merous variants and refinements of these conjectures. We remark however
that if the Sylow -subgroup of is abelian, then Brous conjecture implies
McKays conjecture. Note also that these conjectures (and their variants)
have been verified in a very large number of cases.

B.2.3. Equivalences of Categories: Methods

In order to obtain Morita equivalences or derived equivalences for our


group SL2 (Fq ) we will make use of some general results. Fix A and A as
two sums of O-blocks of finite groups and  . A fundamental property of
blocks of finite groups is that these algebras are symmetric [The, 6]. This fact
will considerably simplify our work in what follows.

Morita equivalences. Fix an (A, A )-bimodule M. We will recall some criteria


used to verify that the functor M A : A mod Amod is an equiva-
B.2 Brauer Correspondents 171

lence of categories (then called a Morita equivalence between A and A ). The


following theorem is shown in [Bro, Theorem 0.2].
Theorem B.2.5 (Brou). Suppose that the bimodule M is projective both as a left
A-module and as a right A -module. Then the following properties are equivalent:
(1) The functor M A : A mod Amod is a Morita equivalence.
(2) The functor KM KA : KA mod KAmod is a Morita equivalence.
(3) Every irreducible character of KA is an irreducible component of the KG -module
KM, and the map Irr KA Irr KA, [V ]KA [KM KA V ]KA is well-defined
and bijective.
(4) Every irreducible character of KA is a factor of KM and the natural map KA
EndKA (KM)opp is an isomorphism.
(5) Every irreducible character of KA is a factor of KM and the natural map A
EndA (M)opp is an isomorphism.
Rickard equivalences. Let C be a complex of (A, A )-bimodules. We recall
some criteria which may be used to verify that the functor

C LA : Db (A ) Db (A)

is an equivalence of categories (then called a Rickard equivalence between A


and A ). Recall first that, if C = M[i] (that is to say, the complex with one non-
zero term M in degree i) and if M induces a Morita equivalence between
A and A , then C induces a Rickard equivalence between A and A .
We denote by H (C ) the (A, A )-bimodule iZ H i (C ). Note that, as K is
O-flat, we have
H (K C ) = KH (C ).
The following theorem is part of the folklore of the subject, but we have not
been able to find a satisfactory reference (note however, that the proof below
is very strongly influenced by [Ric4, Theorem 2.1]).
Theorem B.2.6. Suppose that C is perfect, both as a complex of left A-modules and
of right A -modules, and that

HomDb (A) (C , C [i]) = 0

for all i = 0. Then the following properties are equivalent:


(1) The functor C A is a Rickard equivalence between A and A .
(2) All irreducible characters of KA are factors of H (K C ) and the natural map
A EndDb (A) (C )opp is an isomorphism.
Proof. Firstly, it is clear that (1) implies (2). It remains to show that (2) implies
(1). Therefore suppose that all irreducible characters of KA occur as factors of
H (K C ) and that the natural map A EndDb (A) (C )opp is an isomorphism.
First step: reduction to the case where A is an O-block. Let e be the central idem-
potent of O such that A = Oe and let e = e1 + +en be a decomposition of
172 B Block Theory

e into a sum of primitive central idempotents. Set Ai = Oei = Aei . Then the
action of ei on C is an element of EndDb (A) (C ), therefore there exists an ele-
ment ei of A such that the endomorphism induced by right multiplication
by ei is equal (in the derived category) to that induced by left multiplication
by ei . Moreover, ei is idempotent and central (because this endomorphism
commutes with the action of A ). Let Ai = A ei . Then
n n n n
A = Ai , A = Ai and C = ei C = C ei .
i=1 i=1 i=1 i=1

It is now sufficient to show that ei C = C ei induces a Rickard equivalence


between Ai and Ai . It is easy to verify that ei C satisfies the same hypotheses
as C and so we can (and will) suppose that A is an O-block of O.
Second step: adjunction. Firstly, after replacing C by a homotopic complex, we
may suppose that C is a complex of (A, A )-bimodules which are projective
as left and right modules. As the algebras A and A are symmetric, we have
isomorphisms

C C HomCb (A) (C , A) C HomCb (A ) (C , A ).

As a consequence, the functors C A and C A between the categories of


complexes Cb (A) and Cb (A ) are left and right adjoint to each other. The unit
and counit of these adjunctions induce natural morphisms


A C A C A

of complexes of (A, A)-bimodules. Tensoring with C we obtain natural mor-


phisms of complexes of (A, A )-bimodules

 
C C A C A C C .

It follows from general properties of adjoint functors that  (respectively  )


is a split injection (respectively split surjection) [McL, Theorem IV.1.1]. We
may therefore write C A C A C C C0 , where C0 is a complex of (A, A )-
bimodules projective as left and right modules. By assumption, the natural
morphism A C A C is an isomorphism in the derived category (that is
a quasi-isomorphism) and therefore C A C A C D C . As a consequence,
C0 is acyclic and  and  are quasi-isomorphisms. In particular,   is a
quasi-isomorphism.
On the other hand, is an endomorphism of the (A, A)-bimodule A
and therefore there exists an element z in the centre Z(A) of A such that
(a) = za for all a A. Suppose that is not an isomorphism. Then z
is an element in the radical of Z(A) (as A is a block, Z(A) is a local ring). In this
case   cannot be a quasi-isomorphism (as C is non-zero by hypothesis).
We have therefore shown that is an isomorphism.
B.3 Decomposition Matrices 173

Third step: conclusion. We may therefore write C A C = A C1 , where C1


is a complex of (A, A)-bimodules, projective as left and right modules. Now,
we have an isomorphism (in the derived category)

C A C A C A C D C A A A C D C A C .

Therefore
A D A C1 C1 C1 A C1 .
Therefore C1 is acyclic, which shows that

C A C D A

and completes the proof of the theorem. 




R EMARK B.2.7 If C induces a Rickard equivalence between A and A , then:


(a) The complex kC induces a Rickard equivalence between kA and kA .
(b) The complex K C induces a Rickard equivalence between KA and KA .
(c) The centres of A and A are isomorphic.
(d) | Irr KA| = | Irr KA | and | Irr kA| = | Irr kA |. 

B.3. Decomposition Matrices

We begin with a classical result in representation theory [CuRe, Propositions


23.16 and 16.16].

Proposition B.3.1. Let R be a principal ideal domain and A an R-algebra which


is of finite type and free as an R-module. Denote by K the field of fractions of R.
Let V be a KA-module of finite type. Then there exists an A-stable R-lattice M
of V (so that V = K R V ). Moreover, for all maximal ideals m of R, the class
[M/mM ]A/mA in the Grothendieck group K0 (A/mA) does not depend on the choice
of the A-submodule M of V such that V = K R M.

If A is an O-algebra which is free and of finite type as an O-module,


Proposition B.3.1 shows that we can define a decomposition map

decA : K0 (KA) K0 (kA)

as follows: if V is a KA-module, we may find in V an A-stable O-lattice M


(so that V = KM) and we set

decA ([V ]KA ) = [kM ]kA .

We denote by Dec(A) the decomposition matrix of A, that is to say, the trans-


pose of the matrix of the map decA in the canonical bases ([V ]KA )V Irr KA and
174 B Block Theory

([S ]kA )SIrr kA . Its rows are indexed by Irr KA, while its columns are indexed
by Irr kA.
In the case of group algebras and their blocks, we recall the following
results (see [The, Theorems 42.3 and 42.8] or [Isa, Corollary 2.7 and 15.11]).

Proposition B.3.2. We have:


(a) | Irr K | is equal to the number of conjugacy classes of elements of .
(b) | Irr k| is equal to the number of conjugacy classes of -regular elements of .
(c) decO = decA and Dec(O) = Dec(A).
A{O -blocks of } A{O -blocks of }

E XAMPLE B.3.3 If A is an O-block of with trivial defect group and if


V denotes the unique (up to isomorphism) simple KA-module (see Exam-
ple B.2.4), then, for all A-stable O-lattices M of V , kM is the unique simple
kA-module. As a consequence, Dec(A) = (1).
In particular, if is an  -group, then all O-blocks of have trivial defect
group and hence the decomposition map decO induces a bijection between
Irr K and Irr k and the decomposition matrix Dec(O) is the identity. 

B.4. Brauer Trees*

Hypothesis. We fix an O-block A of with defect group D. In


this section only we suppose that the group D is cyclic.

For the definitions and results referred to without reference in this section,
we refer the reader to [HiLu]. To each block of a group with cyclic defect
group is associated a graph, called the Brauer tree (the graph is in fact a tree).
We recall its construction.
Let A denote the Brauer correspondent of A. The primitive central idem-

potent e  of kN (D) such that A = kN (D)e  is in fact an element of kC (D).
We denote by d the number of primitive central idempotents of kC (D) oc-
curring in the decomposition of e  . Recall that d divides p 1. The set Irr KA
decomposes as follows:

Irr KA = {1 , ... , d } { | },

where the ( ) are the exceptional characters of KA. We then set exc =
and
VA = {1 , ... , d , exc }.
The Brauer tree TA of A is then the pointed graph (that is a graph with a
distinguished vertex, called the exceptional vertex) defined as follows:
B.4 Brauer Trees* 175

The set of vertices of TA is VA . The exceptional vertex is exc .


We join two distinct elements and  in VA if +  is the character of a
projective indecomposable A-module.

The following facts are classical:


(a) TA is a connected tree.
(b) If is the character of an indecomposable projective A-module, then
there exists two distinct elements and  of VA such that = +  .
In particular, the isomorphism classes of projective indecomposable A-
modules are in bijection with the edges of the Brauer tree TA .
In depictions of the Brauer trees given in this book the exceptional vertex
will be represented by a round black circle yand the non-exceptional ver-
tices will be represented by a round white circle i.

R EMARK B.4.1 The Brauer tree of a general finite group is equipped with
an extra piece of data, its planar embedding, that is to say its representation
in the plane. This data is fundamental, but requires much more information
to obtain. As almost all the Brauer trees which occur in this book have only
one planar embedding we will not preoccupy ourselves with this issue. 

A fundamental theorem in the theory of blocks of groups with cyclic de-


fect group is the following.

Theorem B.4.2 (Brauer). Let and  be two finite groups and let A and A be
two O-blocks with cyclic defect groups of the same order. Then A and A are Morita
equivalent if and only if the Brauer trees TA and TA are isomorphic (as pointed
graphs equipped with a planar embedding).

E XAMPLE B.4.3 Suppose that = A5 , the alternating group of degree 5,


and that  = 5. Recall that Irr A5 = {1 , 3 , 3 , 4 , 5 }, where the index denotes
the degree of the character. If A the principal O-block of , then Irr KA =
{1 , 3 , 3 , 4 }, with defect group D =< (1, 2, 3, 4, 5) > a Sylow 5-subgroup
of and VA = {1 , 4 , 3 + 3 }. The Brauer tree TA is therefore

TA y i i

On the other hand, the reader may verify that the Brauer tree of the principal
block A of the normaliser of D in (that is to say of the Brauer correspon-
dent of A) is
i y i
TA .
where Irr KA = {1, , 2 , 2 }, and 1 and (respectively 2 and 2 ) are the
two characters of KA of degree 1 (respectively 2). The characters of degree
2 are the exceptional characters.
176 B Block Theory

By virtue of Brauers Theorem B.4.2, the blocks A and A are not Morita
equivalent: in fact, TA and TA are not isomorphic as pointed graphs, even
though they are as graphs.
However Brous conjecture, as verified by Rickard [Ric2], asserts that
the derived categories Db (A) and Db (A ) are equivalent as triangulated cat-
egories. 
Appendix C
Review of Reflection Groups

Fix a field K of characteristic zero as well as a K-vector space V of finite


dimension n. If g GLK (V ), we say that g is a reflection if Ker(g IdV ) is of
codimension 1. If is a finite subgroup of GLK (V ), we denote by Ref() the
set of its reflections. We say that is a reflection group on V if it is generated
by the reflections that it contains.
The K-algebra of polynomial functions on V will be denoted K[V ]. This
algebra inherits an action of GLK (V ). A K-algebra of finite type is called poly-
nomial if it is isomorphic to an algebra K[V  ], where V  is a finite dimensional
K-vector space. If is a finite subgroup of GLn (K), the algebra of invariants
K[V ] is a K-algebra of finite type. The Shephard-Todd-Chevalley theorem
characterises the reflection groups in terms of their algebras of invariants
[Bou, chapitre V, 5].

Shephard-Todd-Chevalley theorem. Let be a finite subgroup of GLK (V ).


Then is a reflection group if and only if the invariant algebra K[V ] is polynomial.
If this is the case, then there exists n homogeneous and algebraically independent
polynomials p1 ,. . . , pn K[V ] such that K[V ] = K[p1 , ... , pn ]. If di = deg(pi ) and
if the pi are chosen so that d1  d2   dn , then:
1 Tr( ) n
1
(a) =
|| det(1 X ) i=1 1 X di
(here, X is an indeterminate).

(b) The sequence (d1 , ... , dn ) does not depend on the choice of the pi . We call it the
sequence of degrees of .
(c) || = d1 d2 dn and |Ref()| = ni=1 (di 1).
(d) If moreover is an irreducible subgroup of GLK (V ), then it is absolutely irre-
ducible, its centre is cyclic and |Z()| = pgcd(d1 , d2 , ... , dn ).

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 177


DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011
References

[Abh] S. A BHYANKAR, Coverings of algebraic curves, Amer. J. Math. 79 (1957), 825


856.
[Alp] J.L. A LPERIN, Local representation theory, Cambridge Studies in Advanced
Mathematics 11, Cambridge University Press, Cambridge, 1986, x + 178pp.
[Bor] A. B OREL, Linear algebraic groups, deuxime dition, Graduate Texts in Math-
ematics 126, Springer-Verlag, New York, 1991, xii + 288pp.
[Bon] C. B ONNAF, Quasi-isolated elements in reductive groups, Comm. Algebra 33
(2005), 23152337.
[BoRo] C. B ONNAF & R. R OUQUIER, Coxeter orbits and modular representations,
Nagoya J. Math. 183 (2006), 134.
[Bou] N. B OURBAKI, Groupes et algbres de Lie, chapitres IV, V et VI, Hermann, Paris,
1968.
[Bro] M. B ROU, Isomtries de caractres et quivalences de Morita ou drives,
Publ. Math. I.H..S. 71 (1990), 4563.
[BrMaMi] M. B ROU, G. M ALLE & J. M ICHEL, Reprsentations unipotentes gnriques des
groupes rductifs finis, Astrisque 212 (1993).
[BrMi] M. B ROU & J. M ICHEL, Blocs et sries de Lusztig dans un groupe rductif
fini, J. Reine Angew. Math. 395 (1989), 5667.
[BrPu] M. B ROU & L. P UIG, A Frobenius theorem for blocks, Invent. Math. 56 (1980),
117128.
[Ca] M. C ABANES, On Okuyamas theorems about Alvis-Curtis duality, preprint
(2008), arXiv:0810.0952.
[CaEn] M. C ABANES & M. E NGUEHARD, Representation theory of finite reductive groups,
New Mathematical Monographs 1, Cambridge University Press, Cambridge,
2004, xviii + 436pp.
[CaRi] M. C ABANES & J. R ICKARD, Alvis-Curtis duality as an equivalence of de-
rived categories, Modular representation theory of finite groups (Charlottesville,
VA, 1998), 157174, de Gruyter, Berlin, 2001.
[Carter] R.W. C ARTER, Finite groups of Lie type. Conjugacy classes and complex characters,
Pure and Applied Mathematics, John Wiley & Sons (New York), 1985, xii +
544pp.
[Cartier] P. C ARTIER, Reprsentations linaires des groupes algbriques semi-simples
en caractristique non nulle, Sminaire Bourbaki, Vol. 8, Exp. No. 255, 179188,
Soc. Math. France, Paris, 1995.
[CuRe] C.W. C URTIS & I. R EINER, Methods of representation theory, with applications to
finite groups and orders, vol. I, rdition de loriginal de 1981, Wiley Classics
Library, New York, 1990. xxiv + 819pp. Methods of representation theory, with

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 179


DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011
180 References

applications to finite groups and orders, vol. II, rdition de loriginal de 1987,
Wiley Classics Library, New York, 1994. xviii + 951pp.
[De1] P. D ELIGNE, La conjecture de Weil I, Publ. Math. I.H..S. 43 (1974), 273307.
[De2] P. D ELIGNE, La conjecture de Weil II, Publ. Math. I.H..S. 52 (1980), 137252.
[DeLu] P. D ELIGNE & G. L USZTIG, Representations of reductive groups over finite
fields, Ann. of Math. 103 (1976), 103161.
[DiMi] F. D IGNE & J. M ICHEL, Representations of finite groups of Lie type, London Math-
ematical Society Student Texts 21, Cambridge University Press, 1991, iv +
159pp.
[Dri] V.G. D RINFELD, Elliptic modules (Russian), Mat. Sb. (N.S.) 94 (1974), 594627.
[Du] O. D UDAS, Gomtrie des varits de Deligne-Lusztig, dcompositions, cohomolo-
gie modulo  et reprsentations modulaires, Ph. D. thesis, Universit de Franche-
Comt, Besanon, 2010.
[Go] B. G ONARD, Catgories drives de blocs dfaut non ablien de GL2 (Fq ),
Ph.D. Thesis, Universit Paris VII (2002).
[HiLu] G. H ISS & K. L UX, Brauer trees of sporadic groups, Oxford Science Publications,
The Clarendon Press, Oxford University Press, New York, 1989, x + 526pp.
[Jan] J.C. J ANTZEN, Representations of algebraic groups, Second edition, Mathematical
Surveys and Monographs, 107, American Mathematical Society, Providence,
RI, 2003, xiv + 576pp.
[Jor] H.E. J ORDAN, Group-characters of various types of linear groups, Amer. J.
Math. 29 (1907), 387405.
[Har] R. H ARTSHORNE, Algebraic geometry, Graduate Texts in Mathematics 52,
Springer-Verlag, New York-Heidelberg, 1977, xvi + 496pp.
[Isa] M. I. I SAACS, Character theory of finite groups, AMS Chelsea Publishing, Provi-
dence, RI, 2006, xii + 310pp.
[Leh] G.I. L EHRER & D.E. TAYLOR, Unitary reflection groups, Cambridge University
Press, Australian Mathematical Society Lecture Note Series 20.
[Lin] M. L INCKELMANN, Derived equivalence for cyclic blocks over a p-adic ring,
Math. Z. 207 (1991), 293304.
[Lu1] G. L USZTIG, Representations of finite Chevalley groups, CBMS Regional Confer-
ence (Madison 1977), CBMS Regional Conference Series in Mathematics 39,
American Mathematical Society, Providence, 1978, v + 48pp.
[Lu2] G. L USZTIG, Characters of reductive groups over finite fields, Annals of Math.
Studies 107 (1984), Princeton University Press, xxi + 384pp.
[Lu3] G. L USZTIG, On the representations of reductive groups with non-connected
center, Astrisque 168 (1988), 157166.
[McL] S. M AC L ANE, Categories for the working mathematician, Graduate Texts in Math-
ematics 5, Springer-Verlag, New York-Berlin, 1971, ix + 262pp.
[NaTs] H. N AGAO & Y. T SUSHIMA, Representations of finite groups, Academic Press,
Inc., Boston, MA, 1989, xviii + 424pp.
[Oku1] T. O KUYAMA, Derived equivalences in SL(2, q), preprint (2000).
[Oku2] T. O KUYAMA, Remarks on splendid tilting complexes, dans Representation
theory of finite groups and related topics (Kyoto, 1998), Surikaisekikenkyusho
Kokyuroku 1149 (2000), 5359.
[Oku3] T. O KUYAMA, On conjectures on complexes of some module categories related
to Coxeter complexes, preprint (2006), 25pp.
[Ray] M. R AYNAUD, Revtements de la droite affine en caractristique p > 0 et con-
jecture dAbhyankar, Invent. Math. 116 (1994), 425462.
[Ric1] J. R ICKARD, Derived categories and stable equivalence, J. Pure and App. Alg.
61 (1989), 303317.
[Ric2] J. R ICKARD, Derived equivalences for the principal blocks of A4 and A5 ,
preprint (1990).
[Ric3] J. R ICKARD, Finite group actions and tale cohomology, Publ. Math. I.H.E.S 80
(1994), 8194.
References 181

[Ric4] J. R ICKARD, Splendid equivalences: derived categories and permutation


modules, Proc. L.M.S. 72 (1996), 331-358.
[Rou1] R. R OUQUIER, Some examples of Rickard complexes, Proceedings of the Con-
ference on representation theory of groups, algebras, orders, Ann. St. Univ.
Ovidius Constantza 4 (1996), 169173.
[Rou2] R. R OUQUIER, The derived category of blocks with cyclic defect groups, dans
Derived equivalences for group rings, 199220, Lecture Notes in Math. 1685,
Springer, Berlin, 1998.
[Rou3] R. R OUQUIER, Complexes de chanes tales et courbes de Deligne-Lusztig, J.
Algebra 257 (2002), 482508.
[Sch] I. S CHUR, Untersuchungen ber die Darstellung des endlichen Gruppen
durch geborchene lineare Substitutionen, J. fr Math. 132 (1907).
[Ser] J.-P. S ERRE, Reprsentations linaires des groupes finis, Troisime dition, Her-
mann, Paris, 1978, 182pp. English translation: Linear representations of fi-
nite groups, Graduate Texts in Mathematics 42, Springer-Verlag, New York-
Heidelberg, 1977, x + 170pp.
[SGA1] A. G ROTHENDIECK, Revtements tales et groupe fondamental (SGA1), Lecture
Notes in Mathematics 224, Springer, 1971.
[SGA4] M. A RTIN, A. G ROTHENDIECK ET J.-L. V ERDIER, Thorie des topos et coho-
mologie tale des schmas (SGA4), Lecture Notes in Mathematics 269, 270, 305,
Springer, 19721973.
[SGA4 12 ] P. D ELIGNE, Cohomologie tale (SGA4 21 ), Lecture Notes in Mathematics 569,
Springer, 1977.
[ShTo] G.C. S HEPHARD & J.A. T ODD, Finite unitary reflection groups, Canadian J.
Math. 6 (1954), 274304.
[Spr] T.A. S PRINGER, Regular elements of finite reflection groups, Invent. Math. 25
(1974), 159198.
[The] J. T HVENAZ, G -algebras and modular representation theory, Oxford Mathemat-
ical Monographs, Oxford Science Publications, The Clarendon Press, Oxford
University Press, New York, 1995, xxviii + 470pp.
[We] A. W EIL, Numbers of solutions of equations in finite fields, Bull. Amer. Math.
Soc. 55 (1949), 497508.
[Yo] Y. Y OSHII, Brous conjecture for the nonprincipal block of SL(2, q) with full
defect, J. Algebra 321 (2009), 24862499.
Index

Symbols [E / ] xxi
|E | xxi
1 28 E (n) 118
(A, B)bimod xxi E (GF , S ) 151
A+ xxi E (GF , S ) 154
A xxi F 15
A+ , A 124 Fp 113
A , A 72 F(n) 116
Amod xxi F, Fp , Fq 3
a 167 FM , FM 29
An 129 F[G] (U) 112
B 3 F 75
B 33 FM , FM 29
b , b 73 G 3
B 110, 149 GLR (M) xxii
B , B 72
G 33
Cb (A), Cb (A, B) xxi G 110, 149
C (E ) xxi GF 149
C [i] xxii g(C) 22
C C C  , C K C  , C D C  xxii G 22
CarG 110 GLn (R) xxii
Cl ( ) xxi Hc (V) 161
d 4 Hc (Y, Z ) 80
d 5 Hc (C ) 171
D() xxi Hci (V, ), Hci (V) 159
Db (A), Db (A, B) xxi I (n) 114
D 22 Irr A xxi
Dec(A) 173 Irr 28
decA 173 In , I 2 xxii
det0 129 K 28
D,D 94 K 28
e+ , e 122 Kb (A), Kb (A, B) xxi
eU 31 K0 (A) xxi
e , e 28 K0 (G) 110
()
eS 154 L(n) 113
E xxi Lq (n) 120
Endgr
Z G Hc (Y, Z ) 80  28

C. Bonnaf, Representations of SL2 (Fq ), Algebra and Applications 13, 183


DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011
184 Index

l 61 V2 110
(mi )i  0 113 VB 31
Matn (R) xxii V 37
tM xxii [V ]G 110
[M ]M xxii VA 174
[M ], [M ]A xxi W 149
N 4 W (w , ) 152
N 5 w 150
N2 3 (x, y ) 111
N (E ) xxi Y 15
Nw 151 Y(w ) 150
o( ) xxi Y 20
O 61 Z 6
p 3 Z 34
PGL2 (Fq ), PSL2 (Fq ) 129 Z(A) xxi
q 3 Z() xxi
q0 57 151
R M xxi 8
R0 , Ri , Ri 124
, , 28
RM xxii 0 32
R 33 B 31
R+ (0 ), R (0 ) 45 B 33
R+ (0 ), R (0 ) 32 28
RK , RK 31 + 54
Rw , Rw . 150 65
Ref() 177  66
reg 28 +0 , 0 65
Rc (V, ) 159 B, , ,
B 67
R 77
+0 , 0 66
R, R 75
22
RK , RK 30
 (n) 112
R , R 75
(n) 110
Rw , Rw 155
q (n) 120
S , S 10
, 94
SLn (R) xxii
[q] 121
s 4
m,n 117
s  ,s  5
110
StG 32
n 129
S1 152
112
S 154
[ ,  ] xxi
Sn 129
T 3 28
T , T 71 : Y A1 (F) 17

: Y/G A1 (F) 17
T 33
n 3
T 110, 149
(O ),  (O ),  (O ) 63
TA 174
(G, F ) 151
TrV 162
 (G, F ) 154
Tr, Tr 51
168
Tr2 3
q 130
U 3
q 130
u+ , u 8
U 110, 149 : Y P1 (F) \ P1 (Fq ) 19

u 4 : Y/q+1 P1 (F) \ P1 (Fq ) 19
V (i) , v (i) 113 0 : Y P1 (F) 21
Index 185


0 : Y/q+1 P1 (F) 21 character 150
+ 55 induction 37, 77, 150
1 46 restriction 150
47 theory 149
0 45 variety 150
 154 Dickson invariants 24
+ , 54 Drinfeld curve 15
: Y A1 (F) \ {0} 18 dual

: Y/U A1 (F) \ {0} 18 character 28
module 28
A
E
Abhyankars conjecture 24 Euler characteristic 161
exceptional
B character 174
vertex 174
-block 167
-block 168 F
Brauer
correspondence 168 formula
correspondent 169 Knneth 161
first main theorem 169 Mackey 32, 40, 150
morphism 169
third main theorem 170 G
tree 174 geometric conjugacy 151
Brous conjecture 170 geometric series 151
Bruhat decomposition 4 reflection group 177
C H

character Harish-Chandra
Deligne-Lusztig 150 induction 30, 75
category restriction 30, 75
derived xxii Hurwitz
homotopy xxii bound 23
of complexes xxii formula 22
character
central 168 I
cuspidal 31
induction
dual 28
Deligne-Lusztig 37, 77, 150
exceptional 174
Harish-Chandra 30, 75
Steinberg 32
unipotent 152 K
conjecture
McKay 64 Knneth formula 161
cuspidal, character 31
L
D
-regular 154
Lang theorem 149
decomposition
Lusztig series 151
map 173
matrix 173 M
degrees 177
Deligne-Lusztig Mackey formula 32, 40, 150
186 Index

McKay conjecture 64, 170 module 110


module representation 110
dual 28 Raynauds theorem 24
rational 110 reflection 177
standard 110 restriction
Weyl 110 Deligne-Lusztig 150
Morita equivalence 170 Harish-Chandra 30, 75
Rickard equivalence 171
N

nilpotent block 74 S

P semi-simple xxii
standard module 110
polynomial algebra 177
principal block 170
U
Q
unipotent xxii
quasi-isolated block 74

R W

rational Weyl module 110