0 Votes +0 Votes -

6 vues195 pagesMathematics

Jun 07, 2017

© © All Rights Reserved

PDF, TXT ou lisez en ligne sur Scribd

Mathematics

© All Rights Reserved

6 vues

Mathematics

© All Rights Reserved

- Lie Algebras and Representation Theory
- Caesar's Cipher
- LESSON PLAN IN TRANSLATING VERBAL TO MATHEMATICAL SENTENCE.docx
- Linear
- group
- Dr. P.J's Introduction to Abstract Algebra (Suryavanshi's collection)
- G031038047.pdf
- Abstract Algebra
- Mathematics Non Mathematics
- a
- Dror Bar-Natan and Zsuzsanna Dancso- Finite type invariants of W-knotted Objects: From Alexander to Kashiwara and Vergne
- 5047755-Equilibrium-and-travelingwave-solutions-of-plane-Couette-flow-.pdf
- [W. K. Nicholson, M. F. Yousif,] Quasi-Frobenius R(B-ok.org)
- BTU Syllabus
- Math Syllebus
- GRE Prob
- 1829451007
- DM 10M
- An Elementary Theory of the Category of Sets
- Isi Mtech Cs 08

Vous êtes sur la page 1sur 195

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.

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

DOI 10.1007/978-0-85729-157-8

Springer London Dordrecht Heidelberg New York

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

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.

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

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.

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.

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

1.1 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

2.1 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

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

5.1 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

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

7.1 Blocks, Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

7.1.1 Partition in -Blocks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

7.1.2 Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

7.1.3 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

Contents xv

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

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

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

10.1 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

11.1 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

12.1 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

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

B.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

B.2 Brauer Correspondents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

B.2.1 Brauers Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

B.2.2 Conjectures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170

Contents xvii

B.3 Decomposition Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

B.4 Brauer Trees* . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

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

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183

List of Tables

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

6.2 Character table of N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

6.3 Character table of N . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

6.4 Character table of B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

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

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 )

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

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

n = { F | n = 1}.

Fq , 1+q the trace and norm respectively.

of upper triangular (respectively diagonal, respectively unipotent upper tri-

angular ) matrices. Then

B = T U.

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

(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

of G into double cosets for the action of B:

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

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

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

multiplication. We may therefore view it as a subgroup of GLFq (Fq 2 ). It is

then easy to verify that, if Fq2 , then

(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 .

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 ).

(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 .

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

T , s will be

denoted N . The exact sequence

(1.1.13) 1 T N Z/2Z 1

later that N is the normaliser of T (Corollary 1.3.2).

6 1 Structure of SL2 (Fq )

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

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

and, if a Fq , then

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.

(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

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.

1.3.1. Centralisers

our group G . As we will see in the following section, the list of elements

contained in the proposition is exhaustive up to conjugacy.

(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

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 ,

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

(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

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

tatives is given by

Proof. Denote by E the set

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

| ClG (g )| = |G |.

g E

|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.

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

q2 1

Cardinality 1 q(q + 1) q(q 1)

2

Centraliser G T T ZU

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 )

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

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

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.

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

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

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

12 1 Structure of SL2 (Fq )

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).

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.

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

1.10. 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 }.

(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.

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 .

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 )

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

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 .

G (q+1

F mon )

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 .

DOI 10.1007/978-0-85729-157-8_2, Springer-Verlag London Limited 2011

16 2 The Geometry of the Drinfeld Curve

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

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.)

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.

F mon -

equivariant isomorphism.

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

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

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.

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 ).

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

2.3 Fixed Points under certain Frobenius Endomorphisms 19

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

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.

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 = .

following.

Theorem 2.3.2. Let q+1 . Then

F2 0 if =

1,

|Y | =

q3 q if = 1.

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 ) = .

2

20 2 The Geometry of the Drinfeld Curve

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.

2

2.4. Compactification

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

[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

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

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

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 ),

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*

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

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]

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

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

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

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

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,

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

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].

group of an unramified Galois covering of the affine line A1 (F) if and only if it is

generated by its Sylow p-subgroups.

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

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

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

Y/U Y/G

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

2

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).

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

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

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

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 .

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:

and

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

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 ] .

(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 ( ),

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 ).

3.2.1. Definition

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

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,

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 .

module). Then

RK V IndG B VB and RK W W U .

Proof. We have IndG

1

eU = u.

|U| uU

KG , gU geU . We have

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 .

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.

B B .

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 .

K0 (KT ),

(3.2.4)

R( ), R( )G =

, T +

, T .

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 ),

1G is a factor of R(1) and we denote by StG the other irreducible factor of

R(1) (the Steinberg character). We have

If { , 1 }, then

R( ), R( )G = 0.

3.2 Harish-Chandra Induction 33

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.

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)

(3.2.7)
), R(

R(
) =

,
+

, s
.

T T

In particular,

if T ) is irreducible.

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

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

U 1U = 2 =1

R(1) R(0 )

Exercises

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.

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

uU

3.2 Harish-Chandra Induction 35

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.

1 0 (1)q

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

2

braic integer.

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

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).

4.1.1. Definition

i 0

V = K . This defines a Z-linear map

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

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,

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

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

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

follows that

1

q + 1

R ( ) (g ) = TrY (g , ) ( 1 ) = R ( )(g ),

q+1

as expected.

We have, by definition,

4.1 Definition and First Properties 39

rem A.2.6(a),

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

and therefore

1G if i = 2,

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

0 otherwise.

have not obtained any new characters of G .

4.1.3. Dimensions

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 ) ,

4.1.4. Cuspidality

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

R( ), R ( )G = 0.

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

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 ,

acter.

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 ( ).

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

(4.2.1)

R ( ), R ( )G =

, q+1 +

, q+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

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)

Define

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

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 )

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).

(2)

( )

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

: 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 |.

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 )

(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

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

identified with

T takes the following form:

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).

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

(1)

( )

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 .

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

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 .

(i) q+1

K [q+1 ] = Indq+1

(i) 1 (i) .

q+1 q+1

R ( ), R ( )G =

1 (1) , Resq+1

(1) (1) +

1 (2) , Resq+1

(2) (2) ,

4.3 Parametrisation of Irr G 45

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:

2 =1

R (1) R (0 )

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

(4.3.2)

, 2 = 1}

{R( ) | q1 {R ( ) | q+1

, 2 = 1}.

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

We will now study the eigenvalues of the action of F on Hci (Y). By Theo-

rem A.2.7(b), we have

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

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

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) = + .

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

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

2

( ) .

q+1

|Y F |

2

1 q2 if = 1,

(E ) ( ) = (q + 1)

q 1

=

q +1 if = 1.

q+1

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.

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

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

, 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).

). 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

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

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 |.

morphism defined by

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

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).

Denote by

Tr : G q1 K

(g , a)
Tr((g , d(a)), K [G /U])

Tr : G q+1 K

and

(g , )
TrY (g , ).

5.1.1. Calculation of Tr

T , then

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

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.

x

I2 d(a) d ( )

0

{1} a q1 \ {1} q+1 \ {1} {1}, x Fq

true and is 0 otherwise.

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.

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 .

of non-trivial unipotent elements of G .

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.

q +1 if g = 1,

R(1)(g ) =

1 if g is conjugate to u+ or u .

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

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

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

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) =

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

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 ( )

Fq+ (Fq+ )

(5.2.1)

z
(z
+ (zz ))

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

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).

(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 Restriction to U 55

Then

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

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

ResG

U StG = regU .

Denote by + = + u1 : U Fq . We have

1z

+ = + (z).

01

Corollary 5.2.6.

R( ), IndG

U + G =

R ( ), IndU + G = 1.

G

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

ity to do so.

R+ (0 )) the unique common irreducible component of R(0 ) (re-

spectively R (0 )) and IndG

U + .

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

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.

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

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 ).

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

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

g I2 d(a) d ( ) u

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 +

+ , + G = q + 1.

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).

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

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.

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

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.

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

g I2 d(a) d ( ) u

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 .

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

T , we set

= IndN

T .

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

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 ).

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 ).

g I2 s d ( )

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

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 .

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).

g I2 d(a) u

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

the following result, leaving its proof and the calculation of the character

table as an exercise (see Exercise 6.1).

, 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

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

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

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).

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

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-

DOI 10.1007/978-0-85729-157-8_7, Springer-Verlag London Limited 2011

72 7 Unequal Characteristic: Generalities

components of all R( ) (respectively R ( )), where S (respectively

S ). The following theorem describes the decomposition of Irr G into

-blocks (recall that = p).

(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

(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.

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 ).

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

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

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 .

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

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

(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

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

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.

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.

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

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.

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.

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

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

F 2 = qd(1) + F E .

(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].

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

We now turn to (b), (c) and (d). By using the isomorphism of (G , T )-

bimodules [G /U] GeU . The map

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).

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

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

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

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),

(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

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

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.

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

7.3 Deligne-Lusztig Induction* 79

(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

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).

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.

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.

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.

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 )

? () 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

Z G Hc (Y, Z )

Q G Hc (Y, Q )

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

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.

(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

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

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.

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

r times r times

Hence we have shown that ( ) = ( ) = 0 for all linear characters of

T . Hence, = = 0.

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

[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 |

morphism T : A Z is a symmetrising form.

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

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

7.2. 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 .

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

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 }

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

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 ).

DOI 10.1007/978-0-85729-157-8_8, Springer-Verlag London Limited 2011

86 8 Unequal Characteristic: Equivalences of Categories

between A and OTb .

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 .

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 .

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

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

result.

induces a Morita equivalence

Proposition 8.1.4. Deligne-Lusztig induction RO

between A and OTb .

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 .

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.

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

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

odd (and different from p).

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)

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).

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.

tween the Z -algebras A0 and ON b 0 .

8.3 The Principal Block 89

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.

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.

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.

rem 7.2.4.

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

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).

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

phism S .

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

homotopic to zero is contained in the radical of EndCb (O G ,O T ) C .

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.

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)

+

We then define S : C C by

S = f S f ,

(g , t ) G T , we have

(g , t ) S = S (g , s t ).

2

S 2 = d (1)(1 + h),

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 /]

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

plete by virtue of Theorem B.2.6.

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).

results of Gonard [Go]. Denote by S a Sylow 2-subgroup of G . Even though

S is not abelian, the following result remains valid.

G and NG (S) are Rickard equivalent.

|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].

lent to the derived category of the A -algebra A . This A -algebra satisfies

94 8 Unequal Characteristic: Equivalences of Categories

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).

which p is invertible.

ible.

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.

: 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.

D RG D K D RG D K RG .

8.4 Alvis-Curtis Duality* 95

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]

an isomorphism. Indeed, the calculation completed over Z in Theorem 7.2.4

remains valid here and shows that

F (F E ) = qd(1).

(x x ) (y y ) x e x + F 1 (y ) ese y

eRBe eR[BsB]e (as (RT , RT )-bimodules). As a consequence,

RGe eRGe = RGe eRBe RGe eR[BsB]e .

and

x x e y F 1 (y ) ese

quence of the fact that, in both cases, the inverse is given by the formula

a b ab.

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 .

Now consider

: RG M M RG M M

a m m a (m) m ( (a) m m ).

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

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.

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

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

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-

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 ).

is O-free (respectively a complex of OG -modules which are O-free), the di-

agrams

R

K0 (KT ) K0 (KG )

Rk

K0 (kT ) K0 (kG )

and

R

K0 (KT ) K0 (KG )

Rk

K0 (kT ) K0 (kG )

are commutative.

+

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)

1

ex = x (u1 )u kU.

q uU

9.2 Nilpotent Blocks 99

If a Fq , then

(9.1.5) e0 M = M U and M = ex M.

xFq

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

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,

B 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

tively |S |) because

Irr KA = {R( ) | S }

R( ) (respectively R ( )). The Brauer trees are therefore given by

TA i y

and

TA i y.

odd (and different from p).

blocks.

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

0

are the reductions modulo l of O-free OG -modules admitting R (0 ) (re-

spectively R (0 )) as characters. As

we deduce that

Irr kA0 = {R+ (0 ), R (0 )},

(0 ), R (0 )}

10 1 0

0 1 0 1

1

Dec(A0 ) = 1 1 and Dec(A0 ) = 1 .

.. .. .. ..

. . . .

11 11

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).

R (0 ) (respectively R (0 )). Therefore, the Brauer trees are given by

i y i

TA0

and

TA i y i.

0

9.4.1. Preliminaries

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 )

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

Proposition 9.4.3. If = 2, then kv and StkG are the only two non-trivial kG -

submodules of k[P1 (Fq )].

l = [1; 0]. We then have

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

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.

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.

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.

is odd and divides q 1.

As does not divide q + 1, we deduce from 9.4.2 that

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 :

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 .

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

= 2.

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

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 ).

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).

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

9.4 The Principal Block 107

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

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

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

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.

DOI 10.1007/978-0-85729-157-8_10, Springer-Verlag London Limited 2011

110 10 Equal Characteristic

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.

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

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.

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

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

and that

10.1 Simple Modules 111

ab ab

x = ax + cy and y = bx + dy .

cd cd

We have,

n

(n) = Fx ni y i .

i=0

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

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

n

i i = 0 .

i=0

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

n U

112 10 Equal Characteristic

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

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:

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

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

Uv = {v V2 | v (v ) = 0}

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

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 ().

sequence of the existence of a natural isomorphism of quasi-affine varieties

G/U V2 \ {0}, g U g y .

Fp : G G

ab ap b p

cd cp dp

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

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.

(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

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

i i

g x nm y m = (ax + cy )nm (bx + dy )m =

i 0

As a consequence,

i i i i i i i i

() g x nm y m =

i 0

form

x p (ni mi si ) y p si x p (mi ti ) y p ti ,

i i i i

i 0

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

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

cient for us to show 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

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

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

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.

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].

r 1

pr 1 = (p 1)pi

i=0

10.1 Simple Modules 117

n0

mI (n)

Set

0 if n = 0,

n =

n + n if n > 0.

Then

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

(10.1.13) The map CarG : K0 (G) K0 (T) is injective with image K0 (T)s .

0 m n/2 0 m<n

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

m0

In particular, taking into account the formulas 10.1.12 and 10.1.14, we have

118 10 Equal Characteristic

and

1 if m = n,

(10.1.17) m,n = 0 if m > n,

0 if m

n mod 2.

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

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).

If r E (n), then n 2r 0.

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 .

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

{i + pj | n0 + 1 i p 1 and 0 j n 1}.

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)

mE (n )

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)

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.

matrix of the finite group G modulo p.

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.

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

(10.2.1)

n [Lq (n)]kG

is bijective.

On the other hand, by 10.1.17, we have

(10.2.3) [q (n)]kG 0 n q1 is a Z-basis of K0 (kG ).

module such that [KM ]KG = StG , then kM Lq (q 1).

Lq (q 1) = kU x q1 .

morphism of kU -modules ResG

U Lq (q 1) kU . Hence ResU Lq (q 1) is

G

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

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

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)).

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

(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

1 H=T ). We may easily deduce (again using the character table of G ) the

form of the matrix Dec (OG ):

(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

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

of OG . The following proposition is even better:

G , e G and e

Proposition 10.3.2. The central idempotents e+ StG are primitive.

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 }

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.

Brauer correspondent of A? (for ? {+, }).

for the principal block and by Yoshii [Yo] for the nonpricipal one.

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.

= p = q.

[q] = [p] = Ip ,

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

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

if ? {+, }, then

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

category of projective kG -modules, which we may view as a sublattice (and

even an ideal) of K0 (KG ). Show that

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

(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

Show that F[G] (T) is generated by f1 ,f2 , d1 and d2 and that the kernel of

the canonical morphism

P P(f1 , f2 , d1 , d2 )

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.

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

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

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

DOI 10.1007/978-0-85729-157-8_11, Springer-Verlag London Limited 2011

130 11 Special Cases

The character det0 may be expressed with the help of the sign homomor-

phism on Sq+1 :

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 ),

ab

g= .

0c

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

A =

( (z ) (z)).

z,z Fq

z<z

(ac)(q1)/2 . Hence = 0 (ac) as expected.

Recall that

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

11.2.1. 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

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.

phic to the quaternionic group of order 8. The only non-trivial normal subgroups of

G are Z and N .

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,

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.

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.

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

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.

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.

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).

134 11 Special Cases

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.

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).

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).

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.

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

G is generated by elements of 3, we obtain:

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

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

Tetrahedron

B 1B .

Denote by H the hyperplane of Q[P1 (F3 )] equal to

H ={ a | a = 0}.

P1 (F3 ) P1 (F3 )

admits StG as character, and factorises to give an injection

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

as isometries. One may easily verify that PSL2 (F3 ) is the group of oriented

isometries of the tetrahedron .

11.3.1. Structure

We have

(11.3.1) | SL2 (F5 )| = 120, |PGL2 (F5 )| = 120 and |PSL2 (F5 )| = 60.

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).

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

| 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

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 ).

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

of Z[5 ] containing , we denote by the reduction modulo of +

:

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

11.3 The Case when q = 5 139

(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 ).

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.

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 .

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.

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

We have

(11.3.10) D(W ) = W + .

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 .

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

11.3 The Case when q = 5 141

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.

(n) (n)

n |Wn | |Ref(Wn )| (d1 , d2 ) Z(Wn ) Standard name

Icosahedron

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

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

, ) 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 }.

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.

icosahedron) and PSL2 (F5 ) is its group of oriented isometries.

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

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.

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

such that

144 11 Special Cases

| 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

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

11.4 The Case when q = 7 145

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 ).

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

Then

(11.4.6) |W | = 336.

(11.4.7) |I | = 21.

Set R = {s | s I }. Then

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.

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

|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.

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

attained. One may verify that in fact C is the Klein quartic (defined, for ex-

ample, in [Har, Exercise IV.5.7]).

Exercises

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

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).

(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 .

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.)

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).

Z/2Z Z/2Z. (Hint: Use Example 11.3.8 and Exercise 1.14.)

Chapter 12

Deligne-Lusztig Theory: an Overview*

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].

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.

GF = {g G | F (g ) = g }.

representations (in unequal characteristic) of the finite group GF . Recall one

of the fundamental theorems concerning algebraic groups defined over a

finite field.

DOI 10.1007/978-0-85729-157-8_12, Springer-Verlag London Limited 2011

150 12 Deligne-Lusztig Theory: an Overview*

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.

T, t w F (t)w 1 . Then wF is a Frobenius endomorphism T. The Deligne-

Lusztig variety is the variety

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

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}|.

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.

Rx 1 wF (x) = Rw x ,

1 wF (x)

where x : (Tx ) (TwF ) ,
x 1 .

and a linear character of TwF such that

Rw ( ), GF = 0.

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

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

The proof of the following theorem may be found in [DeLu, Theorem 6.1].

Rw ( ) and Rw ( ) contain a common irreducible factor. Then (w , ) (w , ).

E (GF , S ) = { Irr GF | (w , ) S ,

Rw ( ), GF = 0}.

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.

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.

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

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

where s denotes the class of s in NG (T)/T. We have

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

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

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 },

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.

eties Y(w ).

(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.

(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*

Recall that

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

-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 }.

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].

E (GF , S ) is a union of -blocks.

()

eS = e ,

E (GF ,S )

()

(12.2.2) eS OGF .

12.2 Modular Representations 155

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].

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 .

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*

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].

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).

-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.

version of Brous conjecture is true.

12.2 Modular Representations 157

Exercises

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.

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).

F mon ).

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 ).

DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011

160 A -Adic Cohomology

Recall that

K is O-flat,

and

(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.

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

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].

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

then we have, for all i {0, 1, 2, ... , 2d}, a perfect -equivariant duality

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).

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

i 0

162 A -Adic Cohomology

If End(V) or , we denote by TrV ( ) the alternating sum

i 0

that, if is an automorphism of V of finite order, then

(A.2.5) TrV ( ) Z.

(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

(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].

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.

(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

P1 (F) | y = 0} and set = [1; 0]. By Theorem A.2.1(d), we have an exact

sequence

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

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),

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

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.

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),

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}).

A.3 Examples 165

Exercises

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

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].

(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.

DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011

168 B Block Theory

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 .

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.

induces an isomorphism K KA1 KAr and hence induces a partition

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)

-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)

If D is a subgroup of , we define

B.2 Brauer Correspondents 169

a C (D) a .

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.

||

|D| (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

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 ().

defect group of B0 () and B0 (N (S)) is the Brauer correspondent of B0 ().

B.2.2. Conjectures

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.

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.

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.

used to verify that the functor M A : A mod Amod is an equiva-

B.2 Brauer Correspondents 171

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

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

(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

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

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

phisms of complexes of (A, A )-bimodules

C C A C A C C .

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

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

(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 |.

23.16 and 16.16].

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.

Proposition B.3.1 shows that we can define a decomposition map

(so that V = KM) and we set

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]).

(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 }

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.

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

We join two distinct elements and in VA if + is the character of a

projective indecomposable A-module.

(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.

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).

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

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].

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 ).

DOI 10.1007/978-0-85729-157-8, Springer-Verlag London Limited 2011

References

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

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

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

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

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

- Lie Algebras and Representation TheoryTransféré parJulian Bravo
- Caesar's CipherTransféré parRyanJuleCruz
- LESSON PLAN IN TRANSLATING VERBAL TO MATHEMATICAL SENTENCE.docxTransféré parGlaiza Asentista Mangaron Lazaga
- LinearTransféré parKiel Light
- groupTransféré parAnonymous lIcnP1rr
- Dr. P.J's Introduction to Abstract Algebra (Suryavanshi's collection)Transféré parDeepak Suryavanshi
- G031038047.pdfTransféré parVishal Bhatt
- Abstract AlgebraTransféré parGunnar Calvert
- Mathematics Non MathematicsTransféré parRakesh Khanna
- aTransféré parDeddy Armanto
- Dror Bar-Natan and Zsuzsanna Dancso- Finite type invariants of W-knotted Objects: From Alexander to Kashiwara and VergneTransféré parGmonicon
- 5047755-Equilibrium-and-travelingwave-solutions-of-plane-Couette-flow-.pdfTransféré pargoharmahmoodkhokhar
- [W. K. Nicholson, M. F. Yousif,] Quasi-Frobenius R(B-ok.org)Transféré parleblanc.aza
- BTU SyllabusTransféré parAbhi Tailor
- Math SyllebusTransféré parBiswa Lbd
- GRE ProbTransféré parjfdinatale
- 1829451007Transféré pargabriel
- DM 10MTransféré parRajesh
- An Elementary Theory of the Category of SetsTransféré parRodrigo Rafael Gomes
- Isi Mtech Cs 08Transféré parapi-26401608
- list_ia_0.pdfTransféré parmile1966
- Jam SyllebusTransféré parSubhajit Sen
- Closed Graph Theorems and Baire Spaces (Moors)Transféré parJuan Luis Fuentes Maguiña
- 11543320Transféré parJdu
- AaaaQuestionTransféré parManabendra Giri
- 8. Crypto Math - Part 2Transféré parabhishek reddy
- 2007180Transféré parPedro A. Marrone G.
- postma.pdfTransféré parFelipe Troncoso
- Course Guide Math 17-bTransféré parJames Albert Aquino
- bbm%3A978-0-387-33430-1%2F1Transféré parKarn Kumar

- Category Theory for the Working MathematicianTransféré parsyed hamed hussain
- Stanley I. Grossman - Elementary Linear Algebra-Wadsworth (1987)Transféré parDaniél Cárdenas
- Navas DiableretsTransféré parDaniél Cárdenas
- EZetaTransféré parJonel Pagalilauan
- Hartshorne Algebraic Geometry SolutionsTransféré parΠΤΟΛΕΜΑΙΟΣ ΗΡΑΚΛΕΙΔΗΣ
- Galois TheoryTransféré parCarolina Torres Lozano Ψ
- debruijn.pdfTransféré parDaniél Cárdenas
- ANTe6Transféré parDaniél Cárdenas
- Analytic Number Theory v6Transféré parMichael Yu
- Linear GuestTransféré parManish kumar
- MilneTransféré parDaniél Cárdenas
- [J.S. Milne] Fields and Galois Theory (2015)Transféré parBiginch
- Ram Murty M._ Esmonde J. Problems in Algebraic Number Theory (2ed._ Springer_ 2005)(353s)Transféré paras_5k
- Theory of Groups and Quantum Mechanics - Hermann Weyl - 2edTransféré parDaniél Cárdenas
- Complex Analysis 6 Exercises Solution - Mathematics - Prof Yum-Tong SiuTransféré parDaniél Cárdenas
- CHAP09 Finitely Generated Abelian GroupsTransféré parDaniél Cárdenas
- Milnor J. Topology From the Differentiable Viewpoint (Princeton, 1965)Transféré pardeadchildsta
- Construction of Non Separable Hilbert SpaceTransféré parDaniél Cárdenas
- M.N. Huxley-The Distribution of Prime Numbers_ Large Sieves and Zero-Density Theorems (Oxford Mathematical Monographs)-Oxford University Press (1972)Transféré parDaniél Cárdenas

- VZ09 - SCADA Ready Utilizing Rockwell Software Products to Realize Specific Requirements of a Supervisory Control and Data Acquisition System_lpwTransféré parisaavedracastro
- Coca-cola (2)Advertising&Promotional StrtegyTransféré parAbhishek Singh
- Biology Questions and AnswersTransféré parbusybeefreedom
- Wisconsin Highways 1835-1945Transféré parHistory
- Methods and Issues of Studying the Brain in PsychologyTransféré parKevin Brewer
- Www.nagarjunauniversity.ac.in Ugsyllabus ComTransféré parAshok Kumar U
- HegelTransféré parTaqbir Talha
- Bodyweight Exercise Cardio IntervalsTransféré parFrank Porcelli
- GEA_Group_enTransféré parAntonio Rubio
- Explanations on the Concept of IEC61439Transféré parffilep
- My God Ahura MazdaTransféré parКатя Катарина
- scipy-refTransféré parqwertioptgb
- 3 Phase Alternating Current MotorTransféré parVinci Valcu Parne
- The Robots Are Not ComingTransféré parMark Essenburg
- 6136_01_que_20080618Transféré parShahidullah Chowdhury
- New ECOSTransféré parAhmed Hassan Abo Asem
- Manual VespaTransféré parAlfonso Catron
- Membrane vs CriptogiardiaTransféré paredza aria wikurendra
- Substation TRX - Dry Type.docTransféré parGilberto Mejia
- Precast Concrete for Building SystemsTransféré parmukundmgk
- Penjajuća oplata K.pdfTransféré parMala Mu
- Chapter 1 Rate of ReactionTransféré parnurulnisa86
- Final PaperTransféré parDanica Saluna
- Durga SaptaslokeeTransféré paradith24
- Final Dx ResultsTransféré parzysheai
- Cpmlex PISTransféré parBhavesh Patel
- TDW58.Tut Model GrbTransféré parfunctio1
- Emergencies Part2Transféré parosama-alali
- Lecture 5- Form TolerancesTransféré parOleg Ushakov
- WFRP 3rd Ed - The Book of the AsurTransféré parTyvaldr Raud Valisson

## Bien plus que des documents.

Découvrez tout ce que Scribd a à offrir, dont les livres et les livres audio des principaux éditeurs.

Annulez à tout moment.