Académique Documents
Professionnel Documents
Culture Documents
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.
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
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
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.
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
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
Part IV Complements
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
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
xix
General Notation
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 )
n = { F | n = 1}.
We have isomorphisms
a 0
d: Fq T , a
diag(a, a 1
)=
0 a1
1x
and u: Fq+ U, x
.
01
Now set
0 1
s= .
1 0
Then
(1.1.6) 1 T N Z/2Z 1
(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 .
(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
(1.1.13) 1 T N Z/2Z 1
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
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
1.3.1. Centralisers
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
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
| 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
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
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
Exercises
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 }.
g = g,
g F = F g,
F = 1 F .
G (q+1
F mon )
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.
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
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
2.2.2. 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.
(x, y ) = (x , y ) u U, (x , y ) = u (x, y ).
x x x x
which shows that Fq . Now, if we set a = , then
y y
1a x x
= .
01 y y
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
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 = .
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
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.
2.4. Compactification
[x; y ; z] = [ x; y ; z]
and F [x; y ; z] = [x q ; y q ; z q ].
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 ),
(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*
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]
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
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
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].
By Proposition 1.4.1 and Lemma 1.2.2, the group G = SL2 (Fq ) is generated
by its Sylow p-subgroups. By virtue of Raynauds theorem, G should be the
Galois group of an unramified covering of A1 (F). In fact, in this particular
case, the construction of such a covering is easy: the isomorphism Y/G
A1 (F) and the fact that G acts freely on Y (see Proposition 2.1.2) tells us that
(2.5.6) Y is an unramified Galois covering of A1 (F) with Galois group SL2 (Fq ).
Exercises
2.3. 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
Y/U Y/G
A1 (F) \ {0} A1 (F)
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
(1)
||
e = ( 1 ) .
Chapter 3
Harish-Chandra Induction
3.1. Bimodules
F K mod K mod
M:
and
V
M K V .
and
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 ( ),
1
| |
(3.1.6) FM ( ) = TrM ( , ) ( 1 )
and
1
||
(3.1.7) FM ( ) = TrM ( , ) ( 1 ).
3.2.1. Definition
R : KG mod KT mod
K
and
W
K [U\G ] KG W ,
3.2 Harish-Chandra Induction 31
1
eU = u.
|U| uU
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 .
: 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.
R( ), R( )G =
, T +
, s T .
(3.2.4)
R( ), R( )G =
, T +
, T .
R(0 ) = R+ (0 ) + R (0 ),
If { , 1 }, then
R( ), R( )G = 0.
3.2 Harish-Chandra Induction 33
(3.2.5)
=T
B
U
=B
and G
Bs
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,
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
Exercises
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.
3.3*. The goal of this exercise is to give a new proof of 3.2.12. If g G , we set
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.
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
4.1.1. Definition
R : K0 (K q+1 ) K0 (KG )
On the other hand, the irreducibility of Y (see Proposition 2.1.1) and Theo-
rem A.2.1(c) tells us that
1
q + 1
R ( )(g ) = TrY (g , 1 ) ( ).
q+1
Therefore
1
q + 1
R ( ) (g ) = TrY (g 1 , 1 ) ( 1 ).
q+1
as expected.
We have, by definition,
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.
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
R( ), R ( )G = 0.
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 ,
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
(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}
( ) [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 )
( , ) ( , x) = ( 1 , x).
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 |.
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 )
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
( , ) (u, a, b) = ( u, 1 a, 1 b).
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 ,
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 .
R ( ), R ( )G =
1 (1) , Resq+1
(1) (1) +
1 (2) , Resq+1
(2) (2) ,
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:
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
Hence
(4.4.3) 1 = 1.
4.4 Action of the Frobenius Endomorphism 47
(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) = + .
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
(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
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
R EMARK Part (c) shows that 0 (1) = 0 (1) (see the remark following
Exercise 3.3).
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
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
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 |.
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
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
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
Here, v Y satisfies t v = 1 v .
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)(q 2 1) + (q 2 1) = 0.
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 ))
+ : U K and : U K
u(z)
+ (cz) u(z)
+ (cz0 z).
cC cC
Note that
(5.2.3) + + = regU 1U .
and U R ( ) = regU 1U = + + .
ResG
Denote by + = + u1 : U Fq . We have
1z
+ = + (z).
01
Corollary 5.2.6.
R( ), IndG
U + G =
R ( ), IndU + G = 1.
G
5.2.3. Values of
= 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
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 +
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).
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 .
6.2.1. Characters of N
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
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
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
Exercises
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
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
Proof. This follows from a careful inspection of Table 6.1, Corollary 6.1.3 and
the fact that q is invertible in O.
1
b = (t) t 1
|T | tT
1
|T | t
(respectively b = (t ) t 1 ).
T
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 .
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
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
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
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.
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.
F 2 = qd(1) + F E .
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
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).
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
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
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
d
C = (0 P Q 0).
d
C 0 P Q 0
d
C [1] 0 P Q 0.
d
C 0 P Q 0
d
C [1] 0 P Q 0.
d
C 0 P Q 0
d
C 0 P Q 0.
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
(7.3.8) F (g , t ) = (g , t 1 ) F
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
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 |
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
Exercises
EndZ G Z [G /B] = Z I Z FB .
RGeU RGeU
a
aeU
qeU eU eU = eU + eU ueU
uU \{I2 }
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 ).
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
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.
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
S 2 = d(1)
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).
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
S = f S f ,
S 2 = d (1)(1 + h),
A = A
[T /]
ON (EndDb (O G ) C )opp
8.3 The Principal Block 93
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).
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].
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).
: 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
where
f (a b) = ab a bge eg 1
g [G /B]
F (F E ) = qd(1).
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 .
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
DR [M ] = [RR ( RR M)] [M ].
Calculate DK .
Chapter 9
Unequal Characteristic: Simple Modules,
Decomposition Matrices
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 T ) and G :
R : K0 (KT ) K0 (KG ),
Rk : K0 (kT ) K0 (kG ),
R : K0 (KT ) K0 (KG )
and Rk : K0 (kT ) K0 (kG ).
R
K0 (KT ) K0 (KG )
Rk
K0 (kT ) K0 (kG )
and
R
K0 (KT ) K0 (KG )
Rk
K0 (kT ) K0 (kG )
are commutative.
x (u(z)) = k (xz)
If a Fq , then
(9.1.5) e0 M = M U and M = ex M.
xFq
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
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
TA i y
and
TA i y.
we deduce that
Irr kA0 = {R+ (0 ), R (0 )},
TA i y i.
0
9.4.1. Preliminaries
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 )].
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
() 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.
(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.
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 .
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
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 ).
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).
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
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
(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
B = T U.
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
We have,
n
(n) = Fx ni y i .
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
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}
The proposition then follows immediately from Lemma 10.1.7 and the
equalities () and ().
Fp : G G
ab ap b p
cd cp dp
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.
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
As a consequence,
i 0
i 0
u : Fq+ G
10
1
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
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
(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
Set
0 if n = 0,
n =
n + n if n > 0.
Then
K0 (T)s = Z n .
n0
(10.1.13) The map CarG : K0 (G) K0 (T) is injective with image K0 (T)s .
In particular, taking into account the formulas 10.1.12 and 10.1.14, we have
and
1 if m = n,
(10.1.17) m,n = 0 if m > n,
0 if m
n mod 2.
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.
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
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.
is bijective.
On the other hand, by 10.1.17, we have
Lq (q 1) = kU x q1 .
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
H : H O
t
( (t)).
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
10.3. Blocks
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
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.
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.
[q] = [p] = Ip ,
(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
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
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
Show that F[G] (T) is generated by f1 ,f2 , d1 and d2 and that the kernel of
the canonical morphism
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.
11.1. Preliminaries
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
The character det0 may be expressed with the help of the sign homomor-
phism on Sq+1 :
0 (det(g )) = q+1 q (g ),
A=
(z z) Fq .
z,z Fq
z<z
A =
( (z ) (z)).
z,z Fq
z<z
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
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,
U =
u, u 3 = I2 , u
I = J, u
J =K and u
K = I.
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
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.
Tr( (g )) = R (i )(g ) 1R R.
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).
Tr( (g )) = R (i )(g ) 1R R.
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.
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
H ={ a | a = 0}.
P1 (F3 ) P1 (F3 )
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
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).
| 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 ).
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
: 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
(11.3.10) D(W ) = W + .
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 .
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
3 = {v S | g E3 , g v = v }
and 5 = {v S | g E5 , g v = v }.
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.
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
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
| 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
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 ).
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
Exercises
11.3. Consider
: SL 3 )
2 (F F3+
ab
(a2 + c 2 )(ab + cd).
cd
Show that is a homomorphism of groups and that, if x F3 , then (u(x)) =
x. Show also that N = Ker .
11.6 (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).
GF = {g G | F (g ) = g }.
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
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}|.
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 .
E (GF , S ) = { Irr GF | (w , ) S ,
Rw ( ), GF = 0}.
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.
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
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
ab
(a, c)
cd
is an isomorphism. Via this isomorphism we have
E (G , S1 ) = {1, StG },
One may then verify the above claims concerning Lusztig series.
Recall that
12.2.1. Blocks
S = {(w , ) | (w , ) S }.
The following result is due to Brou and Michel [BrMi, Theorem 2.2].
()
eS = e ,
E (GF ,S )
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].
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.
Exercises
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.
F mon ).
Recall that
and
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
Proof. (a), (b), (c), (d) and (f) are shown in [SGA4 12 ]. (e) is shown in [DeLu,
Proposition 6.4].
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
R EMARK Note that, under the above hypotheses, dimK Hc2d (V) = 1 (see
Theorem A.2.1 (c)). However, it is possible for not to act trivially on
Hc2d (V). For example, a Frobenius endomorphism over Fq acts as multipli-
cation by q d (see Theorem A.2.7(b) below).
(A.2.5) TrV ( ) Z.
(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.
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
On the other hand, the connected group GL2 (F) acts regularly on P1 (F),
therefore, by Theorem A.2.1(e),
|P1 (Fq )| = q + 1,
which is no surprise.
On the other hand, the connected group F acts regularly (by multiplication)
on A1 (F) \ {0}, therefore, by Theorem A.2.1(e),
(A.3.6) F = 1 on Hc1 (A1 (F) \ {0}) and F = q on Hc2 (A1 (F) \ {0}).
Exercises
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
= 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].
( ) | Cl ( )|
(B.1.4) (Cl ( )) = O.
(1)
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
|D| (1)
O.
||
|D| (1)
O .
||
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
B.2.2. Conjectures
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.
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
A C A C A
C C A C A C C .
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
([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]).
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
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.
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).
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
(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 ).
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
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
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
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