Lie Groups and Lie Algebras, Chapters 4-6 - Nicolas Bourbaki

ELEMENTS OF MATHEMATICS
Lie Groups and lbie Algebras

Chapters 4-6
Springer
Springer
Berlin
Heidelberg
New York
Barcelona
Hong Kong
London
Milan
Paris
Tokyo
NICOLAS BOURBAKI
Lie Groups and

Lie Algebras
Chapters 4-6
' Springer
Originally published as
GROUPES ET ALGEBRES DE LIE
Hermann, Paris, i968
Translator
Andrew Pressley
Department of Mathematics
King's College London
Strand, London WC2R 2LS
United Kingdom
e-mail: anp@mth.kcl.ac.uk
Mathematics Subject Classification (2000 ):

22E10;22E15;22E60;22-02
Cataloging-in-Publication data applied for

Die Deutsche Bibliothek · CIP-Einheitsaufnahme
Bourbaki, Nicolas:
Elements of mathematics IN. Bourbaki. -Berlin; Heidelberg; New York;
London; Paris; Tokyo; Hong Kong: Springer
Einheitssacht.: £1ements de mathematique <engl.>
Lie groups and Lie algebras
Chapters 4/6. - 2002
ISBN 3-540-42650-7
ISBN 3-540-42650-7 Springer-Verlag Berlin Heidelberg New York
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned,
specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on
microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is
permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and
permission for use must always be obtained from Springer- Verlag. Violations are liable for prosecution under the
German Copyright Law.
Springer· Verlag Berlin Heidelberg New York
a member of BertelsmannSpringer Science+Business Media GmbH
http://www.springcr.de
©Springer-Verlag Berlin Heidelberg 2002
Printed in Germany
SPIN: 10659982 41/3142 db 5 4 3 2 1o

INTRODl.JCTION TO CHAPTERS IV, V
AND VI
The study of semi-simple (analytic or algebraic) groups and their Lie algebras
leads to the consideration of root systems, Co:r:eter groups and Tits systems.
Chapters IV, V and VI are devoted to these structures.
To orient the reader, we give several examples below.
I. Let g be a complex semi-simple Lie algebra and IJ a Cartan subalgebra
of g1 . A root of g with respect to 1J is a non-zero linear form a on 1J such
that there exists a non-zero element x of g with [h, x] = a(h)x for all h E IJ.
These roots form a reduced root system R in the vector space IJ* dual to
IJ. Giving R determines g up to isomorphism and every reduced root system
is isomorphic to a root system obtained in this way. An automorphism of g
leaving IJ stable defines an automorphism of IJ* leaving R invariant, and every
automorphism of R is obtained in this way. The Wey! group of R consists of
all the automorphisms of IJ" defined by the inner automorphisms of g leaving
1J stable; this is a Coxeter group.
Let G be a connected complex Lie group with Lie algebra g, and let I'
be the subgroup of G consisting of the elements h such that expc(27rih) = 1.
Let K be the root system in IJ inverse to R, let Q(K) be the subgroup of IJ
generated by R- and let P(K) be the subgroup associated to the subgroup
Q(R) of I)* generated by R (i.e. the set of h. E IJ such that >..(h) is an integer
for all ).. E Q(R)). Then P(K) ~ I' ~ Q(R-). Moreover, the centre of G
is canonically isomorphic to P(K)/ I' and the fundamental group of G to
I'/Q(K). In particular, I' is equal to P(K) if G is the adjoint group and I'
is equal to Q(K) if G is simply-connected. Finally, the weights of the finite-
dimensional linear representations of G are the elements of the subgroup of
IJ* associated to r.
II. Let G be a semi-simple connected compact real Lie group, and let g be its
Lie algebra. Let T be a maximal torus of G, with Lie algebra t, and let X be
the group of characters of T. Let R be the set of non-zero elements a of X
such that there exists a non-zero element .r of g with (Ad t).x = o(t)x for all
t E T. Identify X with a lattice in the real vector space V = X @z R; then R
is a reduced root system in V. Let N be the normaliser of T in G; the ac~ion
1 In this Introduction, we use freely the traditional terminology as well as the
notions defined in Chapters IV, V and VI.
VI TO CHAPTERS IV, V AND Vl
of Non T defines an isomorphism from the group N/T to the Weyl group of
R. We have P(R) ::> X ::> Q(R); moreover, X = P(R) if G is simply-connected
and X = Q(R) if the centre of G reduces to the identity element.
The complexified Lie algebra 9(C) of g is semi-simple and t(c) is a Cartan
subalgebra of it. There exists a canonical isomorphism from V(c) to the dual
of t(c) that transforms R into the root system of 9(C) with respect to t(c).
III. Let G be a connected semi-simple algebraic group over a commutative
field k. Let T be a maximal element of the set of tori of G split over k and
let X be the group of characters of T (the homomorphisms from T to the
multiplicative group). We identify X with a lattice in the real vector space
V = X ®z R. The roots of G with respect to T are the non-zero elements a of
X such that there exists a non-zero element x of the Lie algebra g of G with
(Ad t).x = a(t)x for all t E T. This gives a root system R in V, which is not
necessarily reduced. Let N be the normaliser and Z the centraliser of T in G
and let N(k) and Z(k) be their groups of rational points over k. The action
of N(k) on T defines an isomorphism from N(k)/Z(k) to the Wey! group of
R.
Let U be a maximal element of the set of unipotent subgroups of G, de-
fined over k and normalised by Z. Put P = Z.U. Then P(k) = Z(k).U(k) and
P(k) n N(k) = Z(k). Moreover, there exists a basis (a 1 , ... , O'.n) of R such
that the weights of T in U are the positive roots of R for this basis; if S de-
notes the set of elements of N(k)/Z(k) that correspond, via the isomorphism
defined above, to the symmetries s 0 , E W(R) associated to the roots ai, the
quadruple (G(k), P(k), N(k), S) is a Tits system.
IV. In the theory of semi-simple algebraic groups over a local field, Tits
systems are encountered whose Wey! group is the affine Wey! group of a.
root system. For example, let G = SL(n + 1, Qp) (with n 2: 1). Let B be
the group of matrices (aij) E SL(n + 1, Zp) such that a;,; E pZP for i < j,
and let N be the subgroup of G consisting of the matric~s having only one
non-zero element in each row and column. Then there exists a subset S of
N/(BnN) such that the quadruple (G,B,N,S) is a Tits system. The group
W = N/(B n N) is the affine Wey! group of a root system of type An; this is
an infinite Coxeter group.
Numerous conversations with J. Tits have been of invaluable assistance to ·ns
in the preparation of these chapters. We thank him ver·y cordially.
CONTENTS
INTRODUCTION TO CHAPTERS IV, V AND VI . . . . . . . . V

CONTENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . VII
CHAPTER IV COXETER GROUPS AND TITS SYSTEMS
§ 1. Coxeter Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
I. Length and reduced decompositions . . . . . . . . . . . . . . . . . . . . . 1
2. Dihedral groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
3. First properties of Coxeter groups . . . . . . . . . . . . . . . . . . . . . . 4
4. Reduced decompositions in a Coxeter group . . . . . . . . . . . . . . 5
5. The exchange condition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
6. Characterisation of Coxcter groups . . . . . . . . . . . . . . . . . . . . . . 10
7. Families of partitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
8. Subgroups of Coxeter groups . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
9. Coxeter matrices and Coxeter graphs . . . . . . . . . . . . . . . . . . . . 13
§ 2. Tits Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
1. Definitions and first properties . . . . . . . . . . . . . . . . . . . . . . . . . 15
2. An example . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3. Decomposition of G into double cosets . . . . . . . . . . . . . . . . . . 18
4. Relations with Coxeter systems . . . . . . . . . . . . . . . . . . . . . . . . . 19
5. Subgroups of G containing B . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
6. Parabolic subgroups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
7. The simplicity theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
Appendix. Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
L Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
2. The connected components of a graph . . . . . . . . . . . . . . . . . . . 27
3. Forests and trees . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Exercises for § 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
Exercises for § 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
VIII CONTENTS
CHAPTER V GROUPS GENERATED BY REFLECTIONS
§ 1. Hyperplanes, chambers and facets . . . . . . . . . . . . . . . . . . . . . . 61

1. Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2. Facets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3. Chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4. Walls and faces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
5. Intersecting hyperplanes ................... _· . . . . . . . . . . . 67
6. Simplicial cones and simplices . . . . . . . . . . . . . . . . . . . . . . . . . . 68
§ 2. Reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
1. Pseudo-reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
2. Reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3. Orthogonal reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4. Orthogonal reflections in a euclidean affine space . . . . . . . . . 73
5. Complements on plane rotations . . . . . . . . . . . . . . . . . . . . . . . . 74
§ 3. Groups of displacements generated by reflections . . . . . . . 76

1. Preliminary results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2. Relation with Coxeter systems . . . . . . . . . . . . . . . . . . . . . . . . . 78
3. Fundamental domain, stabilisers . . . . . . . . . . . . . . . . . . . . . . . . 79
4. Coxeter matrix and Coxeter graph of HI . . . . . . . . . . . . . . . . . 81
5. Systems of vectors with negative scalar products . . . . . . . . . . 82
6. Finiteness theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
7. Decomposition of the linear representation of W on T . . . . 86
8. Product decomposition of the affine space E . . . . . . . . . . . . . . 88
9. The structure of chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
10. Special points . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
§ 4. The geometric representation of a Coxeter group . . . . . . . 94

1. The form associated to a Coxeter group . . . . . . . . . . . . 94
2. The plane Es,s' and the group generated by (J 3 and (J's' 95
3. The group and representation associated to a Coxeter
matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4. The contragredient representation . . . . . . . . . . . . . . . . . . . . . . 97
5. Proof of lemma 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
6. The fundamental domain of W in the union of the
chambers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
7. Irreducibility of the geometric representation of a Coxeter
group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
8. Finiteness criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
9. The case in which BM is positive and degenerate . . . . . . . 105
CONTENTS IX
§ 5. Invariants in the symmetric algebra .................... . 108

1. Poincare series of graded algebras ...................... . 108
2. Invariants of a finite linear group: modular properties 110
3. Invariants of a finite linear group: ring-theoretic properties . 112
4. Anti-invariant elements ............................... . 117
5. Complements 119
§ 6. The Coxeter transformation ........................... . 121

1. Definition of Coxeter transformations .................. . 121
2. Eigenvalues of a Coxeter transformation: exponents . . . . . . . 122
Appendix: Complements on linear representations . . . . . . . . . . 129
Exercises for § 2. 133

CHAPTER VI ROOT SYSTEMS
§ 1. Root systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

1. Definition of a root system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
2. Direct sum of root systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 159
3. Relation between two roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160
4. Reduced root systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
5. Chambers and bases of root systems . . . . . . . . . . . . . . . . . . . . 166
6. Positive roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7. Closed sets of roots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
8. Highest root . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
9. Weights, radical weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
10. Fundamental weights, dominant. weights . . . . . . . . . . . . . . . . . 180
11. Coxeter transformations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
12. Canonical bilinear form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
§ 2. Affine Weyl group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

1. Affine Wey! group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
2. Weights and special weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
3. The normaliser of \Va . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
4. Application: order of the Weyl group . . . . . . . . . . . . . . . . . . . . 190
5. Root systems and groups generated by reflections . . . . . . . . . 191
x CONTENTS
§ 3. Exponential invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

l. The group algebra of a free abelian group . . . . . . . . . . . . . . . . 194
2. Case of the group of weights: maximal terms . . . . . . . . . . . . . 1%
3. Anti-invariant elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
4. Invariant elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
§ 4. Classification of root systems ..... .............. ........ 201

l. Finite Coxeter groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 201
2. Dynkin graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
3. Affine Weyl group and completed Dynkin graph . . . . . . . . . . 210
4. Preliminaries to the construction of root systems . . . . . . . . . 212
5. Systems of type B1 (l ~ 2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214
6. Systems of type C1 (l 2 2) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
7. Systems of type A1 (I 2 1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 217
8. Systems of type D1 (l 2 3) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
9. System of type F 4 . • • . . • . . . . . • • . • • • . . . . . . . • . . • • . . . • . • . 223
10. System of type Es . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
11. System of type E 7 . . . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . 227
12. System of type E6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
13. System of type G2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231
14. Irreducible non-reduced root systems . . . . . . . . . . . . . . . . . . . . 233
HISTORICAL NOTE (Chapters IV, V and VI) . . . . . . . . . . . . 249

BIBLIOGRAPHY .......................................... 255
INDEX OF NOTATION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259
INDEX OF TERMINOLOGY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 261
PLATE I. Systems of type At (l 2 1) . . . . . . . . . . . . . . . . . . . . . . . . 265
PLATE II. Systems of type Bi (l ~ 2) . . . . . . . . . . . . . . . . . . . . . . . 267
PLATE III. Systems of type Ct (l 2 2) 269
PLATE IV. Systems of type Dt (l ~ 3) 271
PLATE V. System of type E 6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 275
PLATE VI. System of type E 7 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 279
PLATE Vil. System of type E 8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
CONTENTS XI
PLATE VIII. System of type F 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

PLATE IX. System of type G 2 . . . . •••• . . . . . . . . . . . . . . . ••. . . . 289
PLATE X. Irreducible systems of rank 2 . . . . . . . . . . . . . . . . . . . . 291
Summary of the principal properties of root systems 293
CHAPTER IV
Coxeter Groups and Tits Systems
§ 1. COXETER GROUPS
In this section, W denotes a group written multiplicatively, with identity

element 1, and S denotes a set of generators of W such that S = s- 1 and
1 ~ S. Every element of W is the product of a finite sequence of elements of
S. From no. 3 onwards we assume that every element of S is of order 2.
1. LENGTH AND REDUCED DECOMPOSITIONS
DEFINITION 1. Let w E W. The length of w (with respect to S), denoted

by ls (w) or simply by l (w), is the smallest integer q ): 0 such that w is
the product of a sequence of q elements of S. A reduced decomposition of w
(with respect to S) is any sequences=- (s 1 , ... ; Sq) of elements of S such that
w '°-- s 1 .. . sq and q = l(w).
Thus 1 is the unique element of length 0 and S consists of the elements

of length 1.
PROPOSITION 1. Let w and w 1 be in W. We have the formuJas:
l(ww 1 ) ( l(w) + l(w'), (1)

l(w- 1 ) = l(w), (2)
1t(w) - l(w')I ( l(ww'- 1 ). (3)
Let (s 1 , ••• , sp) and (s~, ... , s~) be reduced decompositions of w and w'
respectively. Thus l(w) = p and l(w') = q. Since ww' =- s 1 •.. sps'1 •.. s~, we
have l(ww') ( p + q, proving (1). Since S =- s- 1 and w- 1 = s; 1 ... sj 1 , we
have l(w- 1 ) ( p = l(w). Replacing w by w- 1 gives the opposite inequality
l(w) ( l(w- 1 ), proving (2). Replacing w by ww'- 1 in (1) and (2) gives the
relations
l ( w) - l (w') ( l ( ww' - 1 ), (4)

l(ww'- 1 ) =- l(w'w- 1 ). (5)
2 COXETER GROUPS AND TITS SYSTEMS Ch. IV
Exchanging wand w' in (4) gives l(w') - l(w) ~ l(ww'- 1 ) by (5), proving
(3).
COROLLARY. Lets= (s 1 ,. . .,sp) ands'= (s~,. . .,s~) be two sequences

of elements of S such that w = s 1 ... Sp and w' = s~ ... s~. If the seq1J,ence
(s 1 , ... , Sp, s~ , ... , s~) is a reduced decomposition of ww', then s is a. reduced
decomposition of w and s' is one of w'.
By hypothesis, l(w) ~ p, l(w') ~ q and l(ww') = p + q. By (1), we must
have l (w) = p and l (w') = q, hence the corollary.
Remark In view of formulas (I) and (2), the formula d(w,w') = l(ww'- 1 )
defines a distance on W, invariant. under right translations.
2. DIHEDRAL GROUPS
DEFINITION 2. A dihedral gmup is a group generated by two distinct ele-

ments of order two.
Example. Let M be the multiplicative group {1, -1}, and let m. be an integer
~ 2 (resp. m = oo). Then M acts on the group Z/mZ (resp. on Z) by (-1).;.c =
-x, and the corresponding semi-direct product of M by Z/rnZ (resp. of M
by Z) is denoted by Dm. The elements of Dm are thus the pairs (c, x) with
€ = ±1 and x E Z/m.Z (resp. x E Z); the group law on Dm is given by the
formula
(€,x).(€ 1 ,x') = (€€ 1 ,€1 X + x'). (6)
We denote by i the class of 1 modulo m (resp. i = 1) and set
p = ( -1, 0), p' = (-1, l), 7l' = (1, l). (7)
Then p 2 = p' 2 = 1 and 7r = pp'. The formulas
1Tn = (l,ni), p1Tn = (-1,ni) (8)
show that Dm. is a dihedral group generated by {p, p'}.
PROPOSITION 2. Ass·u.me that S consi.sts of two distinct elements s nnd s'

of order 2.
(i) The subgroup P of W generated by p = ss' is noT"rnal, and W is the
semi-d'irect product of the subgroup T = {1, s} and P. Moreover, (W : P) = 2.
(ii) Let m be the order (finite or 'infinite) of p. Then m ~ 2 and \"f\T is
of order 2m. There is a v.niq·ue isomorphism cp from Dm to W such that
ip(p) = s and <p(p') = s'.
(i) We have sps- 1 = sss's =s's= p- 1 , and hence
(9)
§ 1. COXETER GROUPS 3
for every integer n. Since Wis generated by {s, s'}, and hence by {s,p}, the
subgroup P is normal. It follows that TP is a subgroup of W, and since TP
contains sands' = sp, we have W = TP = PUsP. To prove (i), it is therefore
enough to show that W i P. If W = P, the group W would be commutative,
so p 2 = s 2 s' 2 = 1. But then the only elements of W = P would be 1 and p,
contradicting the hypothesis that Vv contains at least three elements, namely
1, sands'.
(ii) Since sf s', we have pi 1 and so m ~ 2. Since Pis of order m and
(W : P) = 2, the order of W is 2m. If m is finite (resp. infinite), there exists
an isomorphism i.p' from Z/mZ (resp. Z) to P taking n top. Moreover, there
exists an isomorphism t.p 11 from M = {1, -1} to T taking -1 to s. The group
W is the semi-direct product of T and P. In view of the formulas (9) and
pn"' p- 1 = n-n, t.p 1 and t.p11 induce an isomorphism t.p from Dm to W such that
i.p(p) = s and cp( 7r) = p, and hence t.p(p') = s'. The uniqueness of <p follows
from the fact that Dm is generated by {p, p1 }.
Remark. Consider a dihedral group W of order 2m generated by two distinct

elements s and s' of order 2. Denote by Sq (resp. s~) the sequence of length q
whose odd (resp. even) numbered terms are equal to sand whose even (resp.
odd) numbered terms are equal to s', and let Wq (resp. w~) be the product
of the sequence Sq (resp. s~). We have
W2k = (ss')k, W2A:+1 = (ss')k s,

I
W2k = ( I
SS
)/.;
= (
SS
/)-k , I
W2k+ I = S S )k S
( I I
= (
SS
/)- k-1
S.
Ifs = (s 1 , •• . , sq) is a reduced decomposition (with respect to { s, s'}) of

an element w of W, then clearly s; i s.;+ 1 for 1 ~ i ~ q - 1. Hence, s =sq or
s = s~.
If m = oo, the elements ( ss')n and ( ss')"' s for n E Z are distinct. Hence,
the elements Wq (q ~ 0) and w~ (q > 0) are distinct, and ifs is a reduced
decomposition of wq (resp. w~) we necessarily have s = Sq (resp. s = s~).
It follows from this that l(wq) = l(w~) = q and that the set of reduced
decompositions of the elements of W consists of the sq and the s~. Moreover,
every element of W has a. unique reduced decomposition.
Suppose now that m is finite. If q ~ 2m, we have Wq = 'Wq- 2m and
w~ = w~_ 2 m; if m ~ q ~ 2m, we have Wq = w~m-q• w~ = W2m-q· Hence,
neither sq nor s~ are reduced decompositions if q > m. It follows that each
of the 2m elements of W is one of the 2m elements w 0 = w~, wq and w~ for
1 ~ q ~ m - 1, and Wm = w'm. These 2m elements are thus distinct and it
follows as above that l(wq) = l(w~) = q for q ~ m and that the set of reduced
.decompositions of elements of W consists of the Sq and the s~ for 0 ~ q ~ m.
Every element of W except Wm has a unique reduced decomposition; Wm has
two.
3. FIRST PROPERTIES OF COXETER GROUPS
Recall that from now on we assume that the elements of S are of order 2.
DEFINITION 3. (W, S) is said to be a Coxeter system if it satisfies the fol-

low·ing cond'ition:
(C) For s, s' in S, let m(s, s') be the order of ss' and let I be the set of
pairs (s, s') such that m(s, s') is finite. The generating set S and the relations
(ss')m(s,s') = 1 for (s, s') in I form a presentation 1 of the group W.
When (W, S) is a Coxeter system, one also says, by abuse of language,

that W is a Co.Teter grou.p.
Examples. 1) Let m be an integer:? 2 or oo and let W be a group defined by

a set of generators S = { s, s'} and the relations s 2 = s' 2 = 1 when m = oo,
s 2 = s' 2 = (ss')m = I when m is finite. Consider on the other hand the
dihedral group Dm (no. 2, Example) and the elements p and p' of Dm defined
by (7). Since p 2 = p' 2 = 1 and (pp')m = 1 when m is finite, there exists a
unique homomorphism f from W to Dm such that f(s) = p and f(s') = p'.
Since pp' is of order m., it follows that ss' is itself of order m. Hence, (W, S) is
a Coxeter system, Wis a dihedral group of order 2m and .f is an isomorphbm
(Prop. 2).
By transport of structure, it follows that every dihedral group is a Coxeter
group.
2) Let 6,, be the symmetric group of degree n, with n ~ 2. Let s; be the
transposition of i and i + l for 1 ~ i < n, and let S = { s1, ... , s,,_ 1 }. One can
show (§ 2, no.4, Example and § 1, Exerc. 4) that (6,,, S) is a Coxeter system.
3) For the classification of finite Coxeter groups, cf. Chap. 4, § 4.
Remark. Suppose that (W, S) is a. Coxeter system. There exists a homomorphism

E from W to the group {1,-1} characterized by E(s) = -1 for alls ES. We call
E(w) the signature of w; it is equal to (-1) 1('"l. The formula E(ww') = e:(w)c(w')
thus translates into l(ww') = l(w) + l(w') mod. 2.
1 This means that (W, S) has the following universal property: given a group G and
a map f from S to G such that (J(s)f(s'))rn.(s,s') = 1 for (s, s') in I, there exists
a homomorphism g from W to G extending f. This homomorphism is unique
because S generates W. An equivalent form of this definition is the following. Let
W be a group, f a homomorphism from W to W and h a map from S to W such
that f(h(s)) =sand (h(s)h(s'))"'(s.s') = 1 for (s, s') in Sand such that the h(s)
(for s E S) generate W. Then f is injective (and hence is an isomorphism from
W to W).
PROPOSITION 3. As.mme that (W, S) ·is a Coxeter system. Then, two ele-
ments s and s' of S are conj'ugate2 in W if and only ·~f the fol lowing condd·ion
is satisfied:
(I) There exists a finite sequ,ence (s 1 ,. .. , sq) of elements of S such that s 1 =
s, Sq= s' and SjSJ+l ·is of finite odd o·rder for 1 ~ j < q.
Let s and s' in S be such that p = ss' is of finite order 2n + 1. By (9),
sp-n = pn s hence
(10)
and s' is conjugate to s.

For all s in S, let As be the set of s' E S satisfying (I). With the hypotheses
in (I), the elements SJ and Sj+i are conjugate for 1 ~ j < q by the above,
hence every element s' of As is conjugate to s.
Let f be the map from S to M = {l, -1} equal to 1 on As and to -1
on S - As. Let s' and s" in S be such that s's" is of finite order m. Then
f (s') f (s") = 1 if s' and s" are both in A., or both in S - As. In the other
case, f(s')f(s") = -1, but mis even. Thus (.f(s')f(s"))"' = 1 in all cases.
Since (W,S) is a Coxeter system, there exists a homomorphism g from W to
M inducing f on S. Let s' be a conjugate of s. Since s belongs to the kernel
of g, so does s', hence f(s') = g(s') = 1 and finally s' E As· Q.E.D.
4. REDUCED DECOMPOSITIONS IN A COXETER GROUP
Suppose that (W, S) is a Coxeter system. Let T be the set of conjugates in

W of elements of S. For any finite sequences = (s 1 , ... , sq) of elements of S,
denote by <P(s) the sequence (t 1 , ... , tq) of elements of T defined by
tj = (s1 ... Sj_ i).sj(S1 ... Sj-1)- 1 for 1 ~ j ~ q. (11)

Then t 1 = s1 and s 1 ... sq = tqlq- 1 ... t 1 . For any element t E T, denote by
n(s, t) the number of integers j such that 1 ~ j ~ q and ti = t. Finally, put
R = {l, -1} x T.
Lemma 1. (i) Let w E W and t ET. The num/1er (-l)n(s,t) has the same
value Tf(w,t) for all sequences s = (s 1 ,. . .,sq) of elements of S such that
W = S1 ... Sq·
(ii) For w E W, let Uw be the map from R to itself de.fined by
Uw(E,t) = (E.Tf(W- 1,t),wtw- 1) (€ = ±l,t ET). (12)
The map w >--+ Uw is a homomorph·ism from W to the group of permutations

ofR.
2 Recall that two elements (resp. two subsets) of a group Ware said to be conjugate
·· if there exists an inner automorphism of W that transforms one into the other.
Fors ES, define a map Us from R to itself by
U 5 (e:,t) = (e:.(-1) 8•·',sts- 1 ) (e: = ±1,t ET), ( 13)
where Os,t is the Kronecker symbol. It is immediate that = IdR, which u;

shows that U 5 is a permutation of R.
Let s = (s 1 , ... , Sq) be a sequence of elements of S. Put w = sq ... s 1 and
Us = U 80 ••• Usl. We shall show, by induction on q, that
(14)
This is clear for q = 0, 1. If q > 1, puts'= (s 1 , ... ,sq-I) and

I
W =Sq-l···S1.
Using the induction hypothesis, we obtain
Us(e:,t) = U 50 (e:.(-l)n(s',tl,w'tw'- 1 )
= (£, (- l) n(s',t)+.5 ScJ:W

,, ,_, w t W -1) •
W )
But P(s) = (P(s'), w'- 1Sqw') and n(s, t) = n(s', t) + Ow'-'s q w' , t> proving
formula (14).
Now let s, s' E S be such that p = ss' has finite order m. Let s =
(s 1 , ... , s 2m) be the sequence of elements of S defined by Sj = s for j odd
and Sj = s' for j even. Then s 2 m ... s 1 = p-m = 1 and formula (11) implies
that
(15)
Since p is of order m., the elements ti, ... , tm are distinct and tj+m = tj for
1 ~ j :::;; m. For all t E T, the integer n(s, t) is thus equal to 0 or 2 and
(14) shows that Us= IdR· In other words, (UsUs')m = IdR. Thus, by the
definition of Coxeter systems, there exists a homomorphism w >---> Uw from
W to the group of permutations of R such that Us is given by the right-hand
side of (13). Then Uw =Us for every sequences= (s 1 , ... ,sq) such that
w =Sq ... s 1 and Lemma 1 follows immediately from (14).
Lemma 2. Lets= (s1, ... sq), <P(s) = (t 1 , ... ,tq) and w = s 1 •.• sq. Let
1
T w be the set of elements of T such that T/( w, t) = -1. Then s is a red·uced

decompos'ition of w ·if and only if the ti are distinct, and in that case T w =
{t 1 , ..• , tq} and Card (Tw) = l (w).
Clearly Tw C {t 1 ,. .. ,tq}· Takings to be reduced, it follows that
Card(Tw) :::;; l(w). Moreover, if the ti are distinct, then n(s, t) is equal to
1 or 0 according to whether t does or does not belong to {t 1 , ... , t 9 }. It
follows that Tw = {t1,. .. , tq} and that q = Card(Tw) :::;; l(w), which im-
plies that s is reduced. Suppose finally that ti = tj with i < .i· This gives
s; = 'USju- 1 , with u = si+ 1 .•• s.i_ 1 , hence
This shows that s is not a reduced decomposition of w.
Lemma 3. Let w E W ands E S be such that l(sw) :::;; l(w). For any sequence
s = (s 1 , ... , Sq) of elements of S with w = .s 1 ... Sq, there exists an integer j
such that 1 :::;; j ~ q and
SS1 ... Sj-1 = S1 ... S:j-tSj·
Let p be the length of w and w' = sw. By the Remark of no. 3, l(w') =
l(w) + 1 mod. 2. The hypothesis l(w') :::;; l(w) and the relation
jl(w) - l(w')I :::;; l(ww'- 1 ) = l(s) = 1

thus leads to l(w') = p - 1. Choose a reduced decomposition
(s~ , ... , s;,_ 1 )

ofw' and puts= (s,s;, .. .,s~_ 1 ) and <P(s') = (t'1 , ... ,t~). It is clear that
s' is a reduced decomposition of w and that t~ = s. The elements t~, ... , t~
being distinct by Lemma 2, we have n(s', s) = 1. Since w is the product of the
sequences, we have n(s, s) =: n(s', s) mod. 2 by Lemma 1, hence n(s, s) i 0.
Consequently, sis equal to one of the elements tj of the sequence <P(s), hence
the lemma.
Remark. The set T w defined in Lemma 2 consists of the elements of the

form w" sw"- 1 corresponding to the triples (w', w", s) E W x W x S such
that w = w"sw' and l(w') + l(w") + 1 = l(w).
5. THE EXCHANGE CONDITION
The "exchange condition" is the following assertion about (W, S):

(E) Let w E W and s E S be sv,ch that l (sw) :::;; l (w). For any reduced
decomposition (s 1 , ... , Sq) of w, there exists an integer j S'llch that 1 :::;; .i :::;; q
a.nd
SS1 ... Sj-1 = S1 ... Sj-lSj. (16)
We assume in this number that (W, S) satisfies (E). By Lemma 3, this is so if
(W, S) is a Coxeter system. The results of this number thus apply to Coxeter
systems.
PROPOSITION 4. Lets ES, w E Wands= (s 1 , ... ,sq) be a reduced

decomposition of w. Then one of the .following must hold:
a) l(sw) = l(w) + 1 and (s,s 1 , .. • ,.s 9 ) ·is a reduced decomposition ofsw.
b) l(sw) = l(w) - 1 and there exists an integer j such that 1 ::;; j ::;; q,
(s 1 , ... ,Sj-I, Sj+I, ... , Sq) is a reduced decomposition of sw and ~s, s1 , .. ., Sj-1,
s j +1 , ... , Sq) is a reduced decomposition of w.
Put w' = sw. By formula (3) of no. 1, we have
\l(w) - l(w')I::;; l(s) = 1.
We distinguish two cases:
a) l(w') > l(w). Then l(w') = q + 1 and w' = ss 1 •.. sq, so
(s, S1,. .. , Sq)
is a reduced decomposition of w'.
b) l(w')::;; l(w). By (E), there exists an integer j such that 1 ::;; .i::;; q and
(16) is satisfied. Then w = ss 1 ... Sj-ISj+I ... Sq and hence
I
W =SJ ... S,j-JSj+l · .. Sq.
Since q - 1::;; l(w')::;; q, it follows that l(w') = q- 1 and that (s 1 ,. . .,Sj-J,
sJ+ 1 , ... , Sq) is a reduced decomposition of w'.
Lemma 4. Let w E W have length q ~ 1, let D be the sel of reduced decompo-

sitions of w, and let F be a map from D to a set E. Assume that F(s) = F(s')
if the elements s = (s 1 , ... ,sq), s' = (s;, ... ,s~) o.f D satisfy one of the
following hypotheses:
a ) s 1 =s I1 orsq=sq.
I
b) There exists ands' in S such that s1 = sA: =s and Bk = s'i = s' for j
odd and k even.
Then F is constant.
A) Let s, s' E D and put t = (s~, s 1 , ... , Sq_i). We are going to show
that if F(s) -f. F(s') then t E D and F(t) -1- F(s). Indeed, w = s~ and s; ...
hence s;w = s2 ... s~ is of length < q. By Prop. 4, there exists an integer
.i such that 1::;; .i::;; q and the sequence u = (s;,s 1 , ... ,Sj-i,sJ+ 1 , ... ,sq)
belongs to D. We have F(u) = F(s') by condition a); if .i I q we would have
F(s) = F(u) for the same reason, and hence F(s) = F(s') contrary to our
hypothesi8. Thus j = q and hence t = u ED and F(t) = F(s') =I F(s).
B) Let s and s' belong to D. For any integer j with 0 ::;; .i ::;; q + 1 define
a sequence Sj of q elements of S as follows:
so = (s'1 , ... , s~)
s1 = (s1, ... , sq)
Sq+l-k = (s 1 ,s~, ... ,s 1 ,s;,s1,s2, ... ,sk)
(17)
for q - k even and 0 ::;; k ::;; q
Sq+l-k = (s~,s1, ... ,s1,s~,s1,s2, ... ,sk)
for q - k odd and 0 ::;; k ::;; q.
Denote by (Hj) the assertion "s.i E D, sH 1 E D and F(sj) i- F(sj+l )". By

A), (Hj) ~ (Hj+i) for 0 ~ j < q, and (Hq) is not satisfied by condition
b). Hence, (Ho) is not satisfied. Since s 0 = s' and s 1 = s, it follows that
F(s) = F(s').
PROPOSITION 5. Let M be a monoid {with unit element l} and f a map

from S to M. Fors, s' ES, let m(s, s') be the order of ss' and put
(f (s) f (s') )1 ifm(s,s') = 2l, l finite

a(s, s') = { v(s)f(s')) 1f(s) ifm(s,s') = 2l + 1, l .finite ( 18)
if m(s, s') = oo.
If a(s, s') = a(s', s) whenever sis' are ·in S, there exists a map g from W
to M such that
g(w) = f(si) ... f(sq) (19)
for all w E W and any reduced decomposition (s1, ... , Sq) of w.
For any w E W, let Dw be the set of reduced decompositions of w and
F w the map from Dw to M defined by
We are going to prove by induction on the length of w that F w is constant,

which will establish Prop. 5. The cases l(w) = 0, 1 being trivial, we assume
that q ~ 2 and that our assertion is proved for the elements w with l(w) < q.
Let w be of length q and s, s' E Dw; by Lemma 4 it suffices to prove that
Fw(s) = Fw(s') in cases a.) and b) of that lemma.
a) The formula
for w' = s 1 ... Sq-I and w" = s 2 ... Sq and the induction hypothesis show
that Fw(s) = Fw(s') if s1 = s~ or Sq= s~.
b) Suppose that there exist two elements sands' of S such that Sj = s~ =
s and Sk = sj = s' for j odd and k even. It suffices to treat the case s I- s'.
The sequences s and s' are then two distinct reduced decompositions of w in
the dihedral group generated by s and s'. By the Remark in no. 2, the order
m of ss' is necessarily finite and, in the notation of that remark, s = Sm and
s' = s~. Consequently, F w(s) = a(s, s') and F,,,(s') = a(s', s) and hence
Fw(s) = Fw(s').
6. CHARACTERISATION OF COXETER GROUPS
THEOREM 1. (W, S) is a Coxeter system if and only if it satisfies the ex-

change condition (E) of no. 5.
Lemma 3 of no. 4 shows that any Coxeter system satisfies (E).
Conversely, suppose that (E) is satisfied. Let G be a group and f a map
from S to G such that (.f(s)f(s'))m = 1 whenever sands' belong to Sand
ss' is of finite order m. By Prop. 5, there exists a map g from W to G such
that
g(w) = f(s1) ... f(sq) (20)
whenever w = s 1 ... s 9 is of length q. To prove that (W, S) is a Coxeter
system, it suffices to prove that g is a homomorphism, which is a consequence
of the formula
g(sw) = .f(s)g(w) for s ES, w E W
since S generates W. By Prop. 4 of no. 5, only two cases arc possible:
a) l(sw) = l(w) + 1: if (s 1 , ... , sq) is a reduced decomposition of w, then
( s, s 1 , . . . , Sq) is a reduced decomposition of sw, hence (21).
b) l(sw) = l(w) - 1: put w' = sw; then w = sw' and l(sw') = l(w') + 1.
By a), g(w) = f(s)g(sw) and hence f(s)g(w) = g(sw) since (f(s)) 2 = l.
7. FAMILIES OF PARTITIONS
Suppose that (W,S) is a Coxeter system. For any s ES, let Ps be the set of
elements win W such that l(sw) > l(w). We have the following properties:
(A) n Ps = {l}.
sES
Indeed, let w i 1 be in vV and let (81' ... , Sq) be a reduced decomposition

of w. Then q ,? 1 and (s 2 , ..• , sq) is a reduced decomposition of s 2 w, so
l(w) = q and l(siw) = q - 1. Hence, w rf. Ps,.
(B) For any s in S, the sets P s and sP s form. a partit'ion of W.
Let w E W and s E S. By Prop. 4 of no. 5, we must distinguish two cases:
a) l(sw) = l(w) + 1: then w E P 5 •
b) l(sw) = l(w) - 1: put w' = sw sow= sw'; then
l(w') < l(sw')
hence w' E P .. , that is w E sP 8 •

(C) Let s,s' be in Sand let w be in W. ff w E Ps and ws' E P 5 then
sw = ws'.
Let q be the length of w. From w E P 5 it follows that l(sw) = q + l;

and from ws' E Ps it follows that l(sws') = l(ws') - 1 ~ q. Since l(sws') =
l(sw) ± 1, we have finally that l(ws') = q + 1 and l(sws') = q.
Let (s 1 , ... , Sq) be a reduced decomposition of w and Sq+l = s'; then
( s 1 , ... , sq, Sq+ i) is a reduced decomposition of the element ws' of length
q + 1. By the exchange condition, there exists an integer j with 1 ~ j ~ q + 1
such that
SS1 ... Sj-1 = S1 ... Sj· (22)
If 1 ~ j ~ q, we would have sw = s 1 ... Sj-lSj+! ... Sq contradicting the
formula l(sw) = q + 1. Thus j = q + 1 and formula (22) can be written
sw = ws'.
Conversely, we have the following result:
PROPOSITION 6. Let (P s)sES be a fa:m:ily of subsets of W satisfying (C)

and the follow'ing conditions:
(A') 1 E P s for all s E S.
(B') The sets Ps and sP 5 are disjoint for alls E S.
Then, (W, S) is a Coxeter system and P 5 consists of the elements w of
W s·uch that l(sw) > l(w).
Let s E S and w E 'Vil. There are two possibilities:
a) w rf. P 8 • Let (s 1 , •.. , Sq) be a reduced decomposition of wand
Wj = S1 ... Sj
for 1 :,;; .i ~ q; also put w 0 = 1. Since wo E P, by (A') and since w = Wq is
not in P 8 , there exists an integer j with 1 ~ .i ~ q such that Wj-l E P s and
Wj = w1 _ 1 s.i does not belong to P 8 • By (C),
We have thus proved the formula
which implies that sw = s 1 ... Sj-!Sj+l ... sq and l(sw) < l(w).
b) w E P 8 : put w' = sw, so that w' rf. Ps by (B'). By a), we then have
l(sw') < l(w'), that is l(w) < l(sw).
Comparison of a) and b) proves that P s consists of those w E W such
· thitt l(sw) > l(w). The exchange condition follows from this as we have seen
.in:a), hence (W, S) is a Coxeter system (Th. 1 of no. 6).
8. SUBGROUPS OF COXETER GROUPS
In this number, we assume that (W, S) is a Coxeter system. For any subset
X of S, we denote by Wx the subgroup of W generated by X.
PROPOSITION 7. Let w be in W. There exists a subset Sw of S such that

{s 1 , ... , Sq} = Sw for any reduced decomposition ( s 1 , ... , Sq) of w.
Denote by M the monoid consisting of the subsets of S with the compo-
sition law (A, B) ,_, A U B; the identity element of M is 0. Put f (s) = { s}
for s E S. We are going to apply Prop. 5 of no. 5 to M and f. We have
a(s,s') = {s,s'} for s,s' in S if m(s,s') is finite, hence there exists a map
g : w >-> Sw from W to M such that g(w) = f(si) U · · · U f(sq), that is
Sw = {s 1 , ... , Sq} for any w E Wand any reduced decomposition (s 1 , ... , sq)
of w.
COROLLARY l. For any subset X of S, the s·ubgroup Wx of W consists of

the elements w of W such that Sw C X.
If w = s 1 ... Sq with s1, ... , sq in S, then w- 1 =sq ... s 1; hence
(23)
Prop. 4 of no. 5 shows that Ssw' C { s} U Sw' for s E S and w' E W, which
implies the formula
(24)
by induction on the length of w. By (23) and (24), the set U of w E W such
that Sw c Xis a subgroup of W; we have X c Uc Wx, hence U = Wx.
COROLLARY 2. For any subset X of S, we hn.ve Wx n S = X.

This follows from Cor. 1 and the formula Ss = { s} for s in S.
COROLLARY 3. The set S is a minimal generating set of W.

If X C S generates W, then W = Wx and hence X = Sn Wx = S by
Cor. 2.
COROLLARY 4. For any subset X of S and any w in Wx, the length of w

with respect to the genera.ting set X of W x is equal to ls (w).
Let (s 1 , ... , sq) be a reduced decomposition of w considered as an element
of W. We have w = s 1 ..• Sq and SJ EX for 1 ~ j ~ q (Cor. l); moreover, w
cannot be a product of q' < q elements of X c S by definition of q = ls(w).
THEOREM 2. (i) For any subset X of S, the pa.ir (Wx, X) is a. Coxeter

system.
(ii) Let (Xi)iEI be a family of S'Ubsets of S. rr x = n

iEI
xi, then Wx =
nwx.'
iEI
(iii) Let X and X' be two s·ubsets ofS. Then Wx C Wx' (resp. Wx = Wx')
if and only if X C X' (resp. X = X' ).
Every element of X is of order 2 and X generates W x. Let x E X and
w E Wx with lx(xw) ~ lx(w) = q. By Cor. 4 of Prop. 7, we have
ls(xw) ~ ls(w) = q.
Let x 1 , ... , Xq be elements of X such that w = ~x 1 ... Xq· Since (W, S) satisfies
the exchange condition (Th. 1 of no. 6), there exists an integer j such that
1 ~ j ~ q and xx 1 ... x 1 _ 1 = x 1 ... Xj_ 1 x 1 . Thus, (Wx, X) satisfies the
exchange condition and is therefore a Coxeter system (Th. 1 of no. 6). This
proves (i).
Assertions (ii) and (iii) follow immediately from Cor. 1 of Prop. 7.
9. COXETER MATRICES AND COXETER GRAPHS
DEFINITION 4. Let I be a set. A Coxeter matrix of type I is a symmetric

sq·uare matrix M (m.;1 )i,jEI whose entries are integers or too satisfying
the relations
m.;; =1 for all i E I; (25)

m.;j ? 2 for i,.i E I with i :j:. j. (26)
A Coxeter graph of type I is (by abuse of langv.age) a pair consisting of a
graph I'3 having I as its set of vertices and a map f from the set of edges of
this graph to the set consisting of +oo and the set of ·integers ? 3. I' is ca.lled
the underlying graph of the Co:r;eter graph (I', .f).
A Coxeter graph (I', .f) is associated to any Coxeter matrix M of type I

as follows:
the graph I' has set of vertices I and set of edges the set pairs {i, .i} of
elements of I such that m;1 ~ 3, and the map f associates to the edge {i, j}
the corresponding element mij of M.
It is clear that this gives rise to a b~jection between the set of Coxeter
~atrices of type I and the set of Coxeter graphs of type I.
To assist the reader in following our arguments, we often represent a Coxeter

· g~aph of type I by the diagram used to represent its underlying graph, and write
· either next to or above each edge {i,j} the number J({i,j}). We generally omit
· ~hese numbers if they are equal to 3.
;; ~ .· ..
~;~···Seethe appendix for the definition and properties of graphs used here.
If (W, S) is a Coxeter system, the matrix M = (m(s, s'))s, .. 'ES, where

m(8, s') is the order of ss', is a Coxeter matrix of type S which is called the
Coxeter matrix of (W, S): indeed, m(s, s) = 1 since s 2 = 1 for all s E S,
and m(s, s') = m(s', s) ~ 2 ifs -j s' since ss' = (s' s)- 1 is then -j 1. The
Coxeter graph (I', f) associated to M is called the Coxeter graph of (W, S).
We remark that two vertices s and s' of I' are joined if and only if s and s' do
not commute. For example, the Coxeter matrix of a dihedral group of order
· ( m1
2m 1s m)
1an d its
· C oxeter grap I1 1s
. represente dby
m
<>----<>
when m ~ 3 (or
if m. = 3) and by
0 0
when m = 2. *The Coxeter graph of the symmetric group 6n is represented

by
0-0---<lO-----<Q>- • • • -<00---<>0---00
(n - 1 vertices.).
We show later (Chap. V, § 4) that, conversely, any Coxet.er matrix is the matrix
of a Coxeter system.
A Coxeter system (W, S) is said to be irreducil1le if the underlying graph

of its Coxeter graph is connected (Appendix, no. 2) and non-empty. Equiv-
alently, S is non-empty and there exists no partition of S into two distinct
subsets S' and S" of S such that every element of S' commutes with every
element of S". More generally, let (I'i)iEI be the family of connected com-
ponents of I' (Appendix, no. 2) and let S,: be the set of vertices of I'i· Let
Wi = Ws, be the subgroup of W generated by Si (cf. no. 8). Then the
(Wi, Si) are irreducible Coxeter systems (no. 8, Th. 2) called the irred·uc'ible
components of (W, S). lVIoreover, the group \Vis the restricted direct product4
of the subgroups Wi for i E I. Indeed, this follows from the following more
general proposition:
4 A group G is the restricted direct prod1tct of a family (G;);EJ of subgroups if, for
any finite subset J of I, t.he subgroup GJ of G generated by the G,. for ·i E J is the
direct product of the G; for i E J and if G is the union of the GJ. Equivalently,
every element of G.,: commutes with every element of G; for ·i =I j and every
element of G can be writ.ten uniquely as a product fl g; with g; E G; and g; = 1
1.El
for all but. finitely many indices i. This last condition is equivalent to saying that
G is generated by the union of the G.; and that G; n GJ = {l} for all i E I and
all finite subsets J of I such that i ¢ J.
§ 2. TITS SYSTEMS 15
PROPOSITION 8. Let (Si)iEI be a partition of S such that every element

of S; commutes with every element of S_; if i ~ j. For all i E I, let W; be
the S7lbgro'UP generated by S;. Then W is the restricted direct product of the
family (W;)iEI.
It is clear that for all i E I the subgroup W~ ge':ierated by the union of
the W i for j # i is also generated by S~ = LJ S;. Tjms
i#j .
by Th. 2 of no. 8. Since W is generated by the union of the Wi, this proves
the proposition.
§ 2. TITS SYSTEMS
Jn this paragraph, the letters G, B, N, S, T, W have the meaning 'indicated in

no. 1 below.
1. DEFINITIONS AND FIRST PROPERTIES
Let G be a group and B a subgroup of G. The group B x B acts on G by

(b, b').g = bgb'- 1 for b, b' E B and g E G. The orbits of B x B on G are the
sets BgB for g E G, and are called the double cosets of G with respect to B.
They form a partition of G; the corresponding quotient is denoted by B\G/B.
If C and C' are double cosets, CC' is a union of double cosets.
DEFINITION 1. A Tds system is a. <ruadrv.ple (G, Il, N, S), where G is a

group, B and N a.re two sv./1gro·ups of G and S is a. subset of N/(B n N),
satisfying the following axioms:
(Tl) The set B u N generates G and B n N is a normal subgroup of N.
(T2) The set S generates the group W = N/ (B n N) a.nd consists of elements
of order 2.
(T3) sBw C BwB U BswB for s ES and w E W. 5
(T4) For alls E S, sBs <f_ B.
The group W = N/(B n N) is sometimes called the Weyl group of the Tits
~ystem (G,B,N,S).
5
Every element of W is a coset modulo B n N, and is thus a subset of G; hence
products such as BwB make sense. More generally, for any subset A of W, we
denote by BAB the subset U BwB.
wEA
Remarks. 1) We sha.11 see in no. 5 (Cor. of Th. 3) that, if (G, B, N) is given,

there exists at most one subset S of N / (B n N) such that ( G, B, N, S) is a Tits
system.
2) Let (G, B, N, S) be a Tits system, and let Z be a normal subgroup of G
contained in B. Let G' = G/Z, B' = B/Z, N' = N/(Z n N), and let S' be the
image of S in N' / (B' n N"). Then one sees immediately that ( G', B', N', S') is
a Tits system.
Throughout this paragraph, with (G, B, N, S) denoting a Tits system, we

set T = B n N and W = N /T. A double coset means a double cos et of G with
respect to B. For any w E W, we set C(w) = BwB; this is a double coset.
We are going to deduce several elementary consequences of the axioms
(Tl) to (T4). We denote by w, u/, ... elements of Wand bys, s', ... elements
of S. The following relations are clear:
C(l) = B, C(ww') c C(w).C(w'), C(w- 1 ) = C(w)- 1 . (1)
Axiom (T3) can also be written in the form
C(s).C(w) C C(w) u C(sw). (2)
lVIoreover, since C(sw) C C(s).C(w) by (1) and since C(s).C(w) is a union of

double cosets, there are only two possibilities:
C C _ { C(sw) if C(w) <f_ C(s).C(w)

(s). (w) - C(w) U C(sw) if C(w) C C(s).C(w). (3)
By (T4), BI C(s).C(s); putting w =sin (3) and using the relation s 2 = 1,

we obtain
C(s).C(s) =BU C(s). (4)
This formula shows that BU C(s) is a subgroup of G. Multiplying both
sides of (4) on the right by C(w), and using formula (3) and the relation
B.C(w) = C(w),
we obtain
C(s).C(s).C(w) = C(w) U C(sw). (5)
Taking the inverses of the sets entering into formulas (2), (3) and (5) and
then replacing w by w- 1 , we obtain the formulas
C(w).C(s) C C(w) U C(ws) (2')

C( ) C( ) = { C(ws) if C(w) <f__ C(w).C(s) (3')
w · 5 C(w) U C(ws) if C(w) C C(w).C(s)
C(w).C(s).C(s) = C(w) U C(ws). (5')
Lemma 1. Let s1, ... ,s 9 ES and let w E W. We have
C(s1 ... s 9 ).C(w) c U C(si, ... Si w),

(i1 , .... i,,) p
where (i 1 , ... ,ip) denotes the set of strictly increasing seqv.ences of integers
in the interval ( 1, q).
We argue by induction on q, the case q = 0 being trivial. If q ~ 1, we
have C(s 1 ... s 9 ).C(w) c C(s 1 ).C(s 2 .•. s 9 ).C(w). By the induction hypothe-
sis, C(s 2 ... s 9 ).C(w) is contained in the union of the C(sj 1 . . . sivw), where
2 ~.ii < ... < ]p ~ q.
By (T3), the set C(s 1 ).C(sj, ... SJv w) is contained in the union of the sets
C(s 1 sJ 1 • . • SJpw) and C(sj, ... SJpw). This proves the lemma.
2. AN EXAMPLE
Let k be a field, n an integer ~ 0, and (ei) the canonical basis of k 11 • Let

G = GL(n, k), let B be the upper triangular subgroup of G, and let N be the
subgroup of G consisting of the matrices having exactly one non-zero element
in each row and column. An element of N permutes the lines ke;; this gives
rise to a surjective homomorphism N -> 6 11 whose kernel is the subgroup
T = B n N of diagonal matrices, and allows us to identify W = N /T with
6 11 . We denote by s.i ( 1 ~ j ~ n - 1) the element of W corresponding to the
transposition of j and .i + 1; let S be the set of s.i. The quadruple (G, B, N, S)
is a Tits system. Indeed:
Axiom (Tl) follows from Cor. 2 of Prop. 14 of Algebra, Chap. II, § 10,
no. 13.
Axiom (T2) is proved in Algebra, Chap. I, Correction to p. 97.
Axiom (T4) is immediate.
It remains to verify axiom (T3), i.e.
SJBw C BwB lJ Bs7wB for 1 ~ j ~ n - 1, w E W,
or equivalently,
SjB C BB' lJ Bs.iB', with B' = wBw- 1 .

Let Gj be the subgroup of G consisting of the elements that fix the ei for
i i= j, .i + 1 and stabilize the plane spanned by e1 and e1 +1 ; this group is
i~omorphic to GL(2, k:). One checks that GjB = BGJ. Since s1 E Gj, we
have s.iB C BG1 , and it suffices to prove that
Gj c (B n Gj)(B' n Gj) lJ (nGj)Sj(B' n c.i)·
Identify Gj with GL(2,k); the group BnGj is then identified with the upper
triangular subgroup B 2 of GL(2, k), while the group B' n GJ is identified
18 COXETER GROUPS AND TITS SYSTEM.S Ch. IV
with B2 when w(j) < w(j + 1) and with the lower triangular subgroup B2
otherwise. In the first case, the formula to be proved can be written
this follows for example from the fact that B2 is the stabilizer of a point for
the action of GL(2, k) on the projective line P 1 (k), and acts transitively on
the complement of this point. In the second case, the formula to be proved
can be written -
GL(2, k) = B2B2 U B2sB2;
since B2 = sB 2 s, this follows from the preceding formula by multiplying on
the right by s.
3. DECOMPOSITION OF G INTO DOUBLE COSETS
THEOREM 1. We ha.ve G = BWB. The map w........, C(w) is a bijection from

W to the set B\G/B of do'Uble cosets of G with respect to B.
It is clear that BWB is stable under.rt-> :r- 1 , and Lemma 1 shows that
it is stable under the product. Since it contains B and N, it is equal to G.
It remains to prove that C(w) -=f- C(w') if w -=f- w'. For this, we shall prove
by induction on the integer q the following assertion:
(A 9 ) If w and w' are distinct elements of W such that ls(w) ~ ls(w') = q,
then C(w) I C(w').
(For the definition of ls (w), see § 1, no. 1.)
This assertion is clear for q = 0, since then w' = 1 and w -=f- 1, hence
C(w') =Band C(w) I B.
Assume that q ~ 1 and that w, w' sat.is(y the hypotheses of (A,,). There
exists s E S such that sw' is of length q - 1. We have
ls(w) > ls(sw') (6)
hence w I sw'. Moreover, sw I= sw'; by formula (3) of§ 1, no. 1, we have
ls(sw) ~ ls(w) - 1 ~ ls(sw') =.q - 1. (7)
By the induction hypothesis, C(sw') is distinct from C(w) and from C(sw);
from formula (2) it follows that
C(sw') n C(s).C(w) = 0. (8)
Since C(sw') C C(s).C(w'), we have finally that C(w) -=J C(w').
Remark. Axiom (T4) was not used in the preceding proof.

4. RELATIONS WITH COXETER SYSTEMS
THEOREM 2. The pair (W, S) is a Coxeter system. Moreover, for s ES and

w E W, the relations C(sw) = C(s).C(w) and ls(sw) > ls(w) are equivalent.
For any s E S, let P s be the set of elements w E W such that
C(s ).C(w) = C(sw).

We are going to verify that the P s satisfy conditions (A'), (B') and (C) of
§ 1, no. 7; the two assertions of the theorem will then follow from Prop. 6 of
§1, no. 7.
Condition (A') is clear.
We verify (B'). If P s and sP s had an element w in common, we would
have w E Ps and sw E P s, and hence
C(s).C(w) = C(sw), C(s).C(sw) = C(w).
It would follow that C(s).C(s).C(w) = C(w) and, by formula (5), this would
imply that C(w) = C(sw), which would contradict Th. 1.
We verify (C). Lets, s' ES and w, w' E W with w' = ws'. The assumption
that w E P s and w' t/. P s implies that
C(sw) = C(s).C(w) (9)

C(w') C C(s).C(w') (10)
by (3).
From (9) and the relation w = w's', it follows that
C(s)w's'B = C(sw). ( 11)
By formula (2'), C(w').C(s') c C(w') LJ C(w's'), which immediately implies

that
C(w')s'B c C(ws') lJ C(w). (12)
Since C(w') is a union of left cosets gB and since
C(s).C(w') = C(s)w'B,
formula (10) shows that C(s)w' meets C(w') and a fortiori that C(s)w's'B
meets C(w')s'B. It follows from formulas (11) and (12) that the double coset
C(sw) is equal to one of the double cosets C(ws') and C(w); since sw oJ w,
Th. 1 allows us to conclude that sw = ws'.
COROLLARY 1. Let w 1 , .•• ,w9 E Wand let w = w 1 ••• w 9 • fl

ls(w) = ls(wi) + · · · + ls(w 9 ),
then
C(w) = C(w1) ... C(w 9 ).
On taking reduced decompositions of the w.;, one is reduced to the case

of a reduced decomposition
W = S1 ... Sq, with Si E S.
If u = s2 ... Sq, then w = s1u and ls(s 1v.) > ls('u), so C(w) = C(s1).C(u) by
the theorem. The required formula follows from this by induction on q.
COROLLARY 2. Let w E W and let T,,. be the subset of W associated to w

by the procedure of Lemma 2of§1, no. 4. !ft E Tw, then
C(t) C C(w).C(w- 1 ).
If t E T w, there exist by definition elements w', w" E W and s E S such

that
w = w'sw" ls(w) = ls(w') + ls(w") + 1 and t = w'sw'- 1 .

By Cor. 1,
C(w).C(w- 1 ) = C(w').C(s).C(w").C(w"- 1 ).C(s).C(w'- 1 ).

Hence,
C(w).C(w- 1 ) =:i C(w').C(s).C(s).C(w'- 1 ).
By (4), C(s) C C(s).C(s). Hence,
C(w).C(w- 1 ) ~ C(w').C(s).C(w'- 1 ) ~ C(t).
COROLLARY 3. Let w E W and let Hw be the subgroup of G generated by

C(w).C(w- 1 ). Then:
a) For any reduced decomposition (s 1 , ... , sq) of w,
C(sj) C Hw for 1 ~ j ~ q.
b) The group Hw contains C( w) and is generated by C( w).

We prove a) by induction on j. Assume that C(sk:) is contained in Hw for
k < j. Let
t = (s1 ... Sj_i)sj(S1 ... Sj_i)- 1 .
The element t belongs to the subset T w of W defined in Lemma 2 of § 1,

no. 4. By Cor. 2, C(t) C Hw, and hence C(sj) C Hw.
Since C(w) = C(s 1 ) ... C(sq), cf. Cor. 1, we have C(w) C Hw, and b)
follows.
E1:ample. Th. 2, applied to the Tits system described in no. 2, shows that the
symmetric group 6n, with the set of transpositions of consecutive elements,
is a Coxeter gmup.
5. SUBGROUPS OF G CONTAINING B
For any subset X of S, we denote by Wx the subgroup of W generated by

X (cf. § l, no. 8) and by Gx the union BWxB of the double coscts C(w),
w E Wx. We have G 0 =Band Gs= G.
THEOREM 3. a) For any sv.bset X of S, the set Gx is a subgroup of G,

generated by U C(s).
sEX
b) The map X ,_.. Gx is a b·i.fcction from Sfl (S) to the set of subgroups of
G containing B.
c) Let(X;);EI beafam·ilyofsubsetsofX. !JX = nx;,
iEI
thenGx = n
iEI
Gx,.
d) Let X and Y be two subsets of S. Then Gx C Gv (resp. Gx = Gv) if
and only if X CY (resp. X = Y ).
It is clear that Gx = (Gx)- 1 ; Lemma 1 of no. 1 shows that Gx.Gx C Gx;
and hence a) follows, taking into account Cor. 1 of Th. 2.
The injectivity of X ,_.. Gx follows from that of X ........ Wx (§ 1, no. 8,
Th. 2). Conversely, let H be a subgroup of G containing B. Let Ube the set
of w E W such that C( w) C H. We have H = BUB since H is a union of
double cosets. Let X = Un S; we show that H = Cx. Clearly, Gx c H. On
the other hand, let u E U and let (s 1 , ... , sq) be a reduced decomposition
of u. Cor. 3 of Th. 2 implies that C(sj) c H, a.nd hence that Sj E X for
1 ~ .f ~ q. Thus, u E Wx, and since H is the union of the C(u) for u E U,
we have H C Gx, which proves b).
Assertions c) and d) follow from analogous properties of Wx (§ 1, no. 8,
Th. 2).
COROLLARY. The set S consists of the elements w E W such that w ::/= 1

and BU C(w) is a subgroup of G.
The elements w E W such that B LJ C(w) is a subgroup of G are those
for which there exists X C S with W x = { 1, w}. Moreover, if w I 1, we
necessarily have Card(X) = 1 , i.e. w E S.
Remark. 1) The above corollary shows that S is determined by (G, B, N);

for this reason, we sometimes allow ourselves to say that (G, B, N) is a Tits
system, or that (B, N) is a Tds system in G.
PROPOSITION 1. Let X be a subset of S and N' a subgroup of N whose

image in Wis equal to Wx. Then, (Gx,B,N',X) is a Tits system .
. We have Gx = BWxB = I3N'B, which shows that Gx is generated by
J":3UN'.
The verification of the axioms (Tl) to (T4) is now immediate.
R'ROPOSITJON 2. Let X, Y c S and w E W. We have
GxwGv = BWxwWvB.
Let s 1 , ... , s 9 EX and t 1 , ...• tq E Y. Lemma 1 shows that
and hence that

GxwGv c BWxwWvB.
The opposite inclusion is obvious.
Remark. 2) Denote by Gx \ G /Gv the set of subsets of G of-the form GxgGy,

g E G; and define Wx\ W /Wv analogously. The preceding proposition shows
that the canonical bijection w ,....... C( w) from W to B\ G /B defines by passage
to the quotient a bijection W x \ W /W v -+ Gx \ G /Gv.
PROPOSITION 3. Let X c S and g E G. The relation gBg- 1 c Gx ·implies

that g E Gx.
Let w E W be such that g E C(w). Since B is a subgroup of Gx, the
hypothesis gBg- 1 c Gx implies that C(w).C(w- 1 ) c Gx, and hence that
C(w) c Gx by Cor. 3 of Th. 2, so g belongs to Gx.
6. PARABOLIC SUBGROUPS
DEFINITION 2. A s1tbgroup of G ·is said to be parabolic ·~f" it contains a

confugate of B.
It is clear that every subgroup that contains a parabolic subgroup is

parabolic.
PROPOSITION 4. Let P be a s·ubgroup of G.

a.) P is parabolic if and only if there e.r:·ist.s a subset X of S such that P is
conj1tgate to Gx (cf. no. 5 for the definition of Gx).
b) Let X,X' CS and g,g' E G be s·uch that P = gGxg- 1 = g'Gx·g'- 1 .
Then, X = X' and g'g- 1 E P.
Assertion a) follows from Th. 3, b).
Under the hypotheses of b), we have
g -1 g'B g 1-1 g C 9 -1 .9'G X'.9 1-1 = G x,

and Prop. 3 shows that g- 1 g' E Gx. Hence, Gx· = Gx and X' = X by Th. 3,
b). Finally,
which proves b).

If the parabolic subgroup P is conjugate to Gx, where X C S, then P is
said to be of type X.
THEOREM 4. (i) Let P 1 and P2 be two parnbolic S'ubgrpups of G whose

intersection is parabolic and let g E G be such that gP 1 g- 1 C P2 . Then
g E P2 and P1 c P2.
(ii) Two parabolic subgroups whose intersection is parabolic are not con-
jugate.
(iii) Let Q1 and Q2 be two parabolic s·ubgroups of G contained in a .mbgro·up
Q of G. Then any g E G such that gQ 1 g- 1 = Q 2 belongs to Q.
(iv) Every parabolic subgroup is its own norma.lise.,B.
Assertion (i) follows from Props. 3 and 4, and implies (ii). Under the
hypotheses of (iii), we have gQ 1 g- 1 C Q, which implies that g E Q by (i).
Finally, (iv) follows from (iii) by taking Q 1 = Q 2 = Q.
PROPOSITION 5. Let P 1 and P2 be two parabolic su.bgroups of G. Then

P 1 n P2 contains a conjugate of T.
By first transformi11g P 1 and P 2 by an inner automorphism of G, we may
assume that B C P 1 . Let g E G be such that gBg- 1 C P 2 . By Th. l. there
exist n E N and b, b' E B such that g = bnb'. Since T is normal in N,
and
which proves the proposition.
7. SIMPLICITY THEOREMS
Lemma 2. Let H be a norm.al su.bgroup of G. There e:rists a subset X of S

such that BH = Gx and such that every element of X commv.tes with every
element of S - X.
Since BH is a subgroup of G containing B, there exists a unique subset X
of S such that BH = Gx (Th. 3) .
. Let s 1 E X and :9 2 E S- X; let n 1 and n 2 be representatives in N of s 1
and s2, respectively. Then n 1 E Gx = BH and there exists b E B such that
bn1 EH. Since His normal in G, the element h = n 2 bn 1n2, 1 of G belongs to
H. This means that
h E C(82).C(81).C(82).
If the length of 8 2 s 18 2 is equal to 3, Cor. 1 of Th. 2 implies that
61 .
H-H is a subgroup of a group G, the normaliser of H in G is the subgroup 'Jl(H)
consisting of the elements g of G such that gHg- 1 = H. A subgroup H' is said to
(l.onnalise H if H' C 'Jl(H), in which case HH' =- H'H is a subgroup of G in which
!Us normal.
and hence that h E H n C(s2s 1 s2). Since H n C(s2s 1s 2) is non-empty,

s2s 1s 2 E Wx. As (s2,s 1,s2) is a reduced decomposition, it follows that
s 2 E X, contrary to our assumption.
Thus ls(s2s 1s2) ~ 2; if ls(s2s1s2) = 1, then s1s2 ES and so (s 1s2) 2 =I,
or s 1 82 = 828 1 . If ls(82s1s2) = 2, property (E) of§ I, no. 5 implies that
s2s1 = s1s2, since 81 'I- s2. Q.E.D.
The following property of a group U enters into Th. 5 below:
(R) For any norm.al subgroup V of U distinct from U, the com.mutator sub-
group (cf. Algebra, Chap. I, § 6, no. 8) of U /V is distinct from U /V.
Every soluble group satisfies (R); in particular, every abelian group sat-
isfies (R). It can be shown that the symmetric group 6n satisfies (R.) (cf.
Exerc. 29).
THEOREM 5. Let Z be the intersection of the conjugates of B, let U be a

subgroup of B and let G 1 be the subgroup generated by the conjugates of U in
G. We make the following assumptions:
(1) U is normal in B and B =UT.
(2) U has property (R.).
(3) G 1 is eq1tal to its commutator subgroup.
(4) The Co.xeter system (W,S) is irreducible (cf.§ I, no. 9).
Then every .mbgro·up Hof G normalised by G 1 is e'ithcr contained in Z or
contains G 1 •
First we show that G = G 1T. The group G 1 T contains B and hence is
its own normaliser (Th. 4); but as N normalises G 1 and T, it also normalises
G 1T, so NC G 1 T. Since G i:;; generated by Band N, it follows that G = G 1 T.
Next, set
G'=G1H, B'=BnG', N'=NnG',

T' = T n G' = B' n N' and W' = N' /T'.
We have G = G'T since G' contain:;; G 1 , and hence N = N'T. The inclusion
of N' into N thus defines, on passing to the quotient, an isomorphism a :
W' ___. W. Let S' = a- 1 (S).
We now show that ( G', B', N', S') is a Tits system. Since G = BNB and
B =TU= UT, we have G = UNU. Since U is a subgroup of G', it follows
that G' = UN'U; :;;ince Uc B', this proves (Tl). Axiom (T2) is satisfied since
o is an isomorphism. Let w E Wand let w' = a- 1 (w) be the corresponding
element of W'. We have
BwB = BwB' = Bw'B', since B = B'T.

From this we conclude that G' n BwB = B'w'B', which means that the
inclusion of G' into G defines on passing to the quotient a bijection from
B'\G'/B' to B\G/B. Axiom (T3) follows immediately. Axiom (T4) follows

from B = B'T.
The subgroup His normal in G'. By Lemma 2 applied to (G',B',N',S'),
there exists a subset X' of S' such that B'H = G~, and every element of
S' -X' commutes with every element of X'. In view of assumption-(4), there
are only two possibilities:
a) X' = 0, i.e. B'H = B', so H c B' c B. If g E G, then g = g1 t with
9 1 E G1, t ET, and H C g1 Bg1 l since G 1 normalises H. Thus H C gBg- 1 ,
and since Z is the intersection of the gBg- 1 , we have H C Z.
b) X' = S', i.e. B'H = G'. Since G = G'T, we have
G = B'HT = HB'T = HB.
As B normalises U, every conjugate of U is of the form hUh- 1 with h E H.
Such a subgroup is contained in the group UH, hence G 1 c UH by the
definition of G 1 . Thus, we have the isomorphisms
U/(U n H) ~ UH/H = G1H/H ~ Gi,l(G1 n H).
By assumption (3), Gif(G 1 n H) is equal to its commutator subgroup. As-

sumption (2) now shows that the group U/(U n H), which is isomorphic to
Gif(G 1nH), reduces to the identity element. Hence G 1nH = G 1 and G 1 c H,
which completes the proof.
COROLLARY. Under the assumptions of Th. 5, the gmup Gif(G 1 n Z) 'is

either simple rwn-abel-ian or reduces to the identity element.
Th. 5 shows that Gif (G 1 nZ) is simple or reduces to the identity element.
On the other hand, assumption (3) implies that it is equal to its commutator
subgroup. Hence the corollary.
Remarks. 1) Assumptions (2), (3), (4) were not used in the proof that
(G', B', N', S') is a Tits system.
2) Suppose that Zn U = {1}. Since Zand U are normal in B, it follows
that every element of Z commutes with every element of U, and hence with
every element of G 1 . In view of the preceding corollary, it follows that G 1 n Z
is the centre of G 1 .
3) Assumption (3) is implied by the following condition:
(3') U is generated by the commutators b-Lu- 1 bu with u E U and
bEBnG 1 .
ftamples. 1) Let k be a field, n an integer ~ 0, G = GL(n, k), and let

· (G,B, N, S) be the Tits system <lescribed in no. 2. Let U be the strictly upper
.Jdangular subgroup of G, i.e. the subgroup of B consisting of the matrices
4\yhose diagonal entries are equal to 1. Condition ( 1) in Th. 5 is immediate,
26 COXETER GROUPS AND TITS SYSTEl'v!S Ch. IV
and so is (2) since U is soluble. Condition (4) is satisfied if n;;:: 2. One can
show (cf. Algebra, Chap. II,§ 10, Exerc. 13) that (3) is satisfied if n;;:: 3 and
Card(k);;:: 4. Under these conditions, we conclude that Gi/(G 1 nZ) is simple
and that G 1 n Z is the centre of G 1 (cf. Remark 2).
When k is commutative, G 1 = SL(n, k) (cf. Algebra, Chap, III,§ 8, no. 9).
*2) Let g be a simple Lie algebra over C, and let G be the group of inner
automorphisms of g (cf. Chap. III, § 6, no. 2, Prop. 2). By using Th. 5, one
can show that G is simple non-abelian .•
APPENDIX
GRAPHS
1. DEFINITIONS
DEFINITION l. A combinatorial graph (or simply a graph, 4 there is no

risk of confusion) is a. pair (A, S), where S is a. set and A is a sv.bset of ~(S)
consistfr1.g of sets with two elements.
Let I' = (A, S) be a graph. The elements of A arc called the edges and
those of S the vertices of I'; two vertices are said to be jo·ined if {x, y} is an
edge. A vertex is called terminal if it is joined to at most one vertex, and a
ramification po·int if it is joined to at least three vertices.
In accordance with the general definitions (Sets R., §8), an isomorphism.
from the graph I' to a graph I" = (A', S') is a bijection from S to S' that
takes A to A'. A graph I"= (A', S') is called a subgraph of I' if S' c Sand
A' c A; I' is called a full subgraph of I' if S' c S and A' = An ~(S). It is
clear that every subset of S is the set of vertices of exactly one full subgraph
of r.
To make the arguments easier to follow, we represent a graph by a diagram
having points corresponding to the vertices, two points being joined by a line
if and only if the vertices they represent are joined in the graph. For example,
the diagram
od
;---;--;<o e
represents a graph whose vertices are a, b, c, d, e and whose edges are {a, b},
{b,c}, {c,d} and {c,e}.
2. ,!CONNECTED COMPONENTS OF A GRAPH
P~t I' == (A, S) be a graph. If a and b are two vertices, a path joining a and
;b. is a sequence (Xo, ... , Xn) of vertices of I' with :ro = o., :z:n = b, the vertices
t~ and Xi+l being joined for 0 ~ i < n; the integer n ~ 0 is the length of the
·path · · sai'd to
,,.;"'--. : .Th e pa ti1 (Xo, ... , .Tn ) is · . . 1·t· x; r-t. x,; 1'f"z r-t. .J.· If' a
· b e m.1ect1vc
:.path (xo, ... , Xn) joining a and bis of minimal length among such paths, it is
necessarily injective; for if not, there would exist i and j with 0 ~ i <j ~ n
and xi = Xj and then the sequence
(xo, ... , x;, Xj+1, ... , :r:,.)
would be a path of length < n joining a and b.

The relation "there exists a path joining a and b" between two vertices a
and b of I' is an equivalence relation Ron the set S of vertices. The equivalence
classes of R are called the connected components of I'; and I' is said to be
connected if S has at most one connected component, that is if any two vertices
of r can be joined by at least one path.
PROPOSITION 1. Let I' = (A, S) be a graph and (S 0 )aEL its family of

connected components. Denote by I'a. the full subgraph of I' having Sa as its
set of vertices.
(i) For any a in L, the graph I'(> ·is connected.
(ii) 1f I"= (A', S') is a connected subgraph of I', there exists o: in L such
that I" C I'a·
(iii) If a I- /3, no element of S0 is joined in I' to any element of S13
(eq-uivalenlly. every edge of I' is an edge of one of the I'0 ).
(iv) Let (S~).XElvI be a partition of S such that, ·if Ai=µ, no element ofS~
is joined ·in r to any element of s;,; then each of the sets s~ is a union of
connected components of I'.
(i) Let ct be in L and a and b be in S0 • Then there exists a path
c = (:r: 0 , ... , x,.,.) joining a and b in I'. For any i with 0 ~ i ~ n, the path
(xo, ... , Xi) joins a to X; in r. so :r:; E Sn· Thus, c is a path in I'o. joining a
and b. It follows that I'C>. is connected.
(ii) Let I''= (A',S') be a non-empty connected subgraph of I', let a be
an element of S' and let 8 0 be the connected component of S containing a.
For any b in S', there exists a path c joining a. and b in I'', and a fortiori in
I'. It follows that S' c S 0 .
(iii) Given distinct elements a and (3 of L, and vertices a E Set and b E S,13,
there is no path joining a and b, and in particular no edge joining a and b.
(iv) Let a be in S~ and let So: be the connected component of I' containing
a. Then, for any bin Se>., there exists a path (:r:o, ... , Xn) joining a and bin
r. If i is an integer such that 0 ~ i < n and if Xi ES~, then X;+1 Es~ since
X; is joined to Xi+l· It follows by induction that X; Es~ for 0 ~ i ~ n, and
in particular that b = .rn is in S~. Thus, Set c S~.
COROLLARY 1. A graph r = (A, S) ·is connected if and only if there does

not exist a parl'ition (S', S") of S into two non-empty s1tbsets s·uch that no
clement of S' is joined in I' to any element of S".
Suppose that I' is not connected and let S' be one of its connected com-
ponents. The set S" = S - S' is non-empty by Prop. 1, (i) and no element of
S' is joined to any element of S" by Prop. 1, (iii).
APPENDIX GRAPHS 29
Suppose now that I' is connected and let (S', S") be a partition with the
stated property. By Prop. 1, (iv), the set S' contains at least one connected
component, so S' = S and S" = 0, a contradiction.
COROLLARY 2. A subset S' of S is o. union of connected components ~f and

only if no verte:z: belonging to S' is joined to any vertex belonging to S - S'.
The condition is sufficient by Prop. 1, (iv). It is necessary by Prop. 1,
(iii).
3. FORESTS AND TREES
Let I'= (A, S) be a graph. A circuit of I' is a sequence
of distinct vertices of I' such that n ~ 3, x; is joined to .-ri+l for 1 ::::; i < n
and Xn is joined to .7: 1 . I' is callee\ a forest if there is no circuit in I'; every
subgraph of I' is then also a forest. A connected forest is called a tree; the
connected components of a forest are thus trees.
PROPOSITION 2. Let I'= (A, S) be a forest having only a fin:ite number of

vertices.
(i) fl I' has at least one verte.r:, it has a terminal verte.r:.
(ii) If I' has at least t:wo vertices, there is a partition (S', S") of its set of
ver:tices into two non-empty subsets such that two distinct vertices that both
belong to S' or both belong to S" are never joined.
r
Suppose that has at least one vertex and let (.ro, ... 'Xn) be an injective
path of maximal length in I'. The vertex .1:0 cannot be joined to a vertex y
distinct from x 0 , x 1 , ... , Xn, since otherwise there would exist an injective
path in I' of length n + 1, namely (y, xo, ... , Xn)· The vertex x 0 is not joined
to any vertex :r., with 2 ::::; i ::::; n, since otherwise (x 0 , .-r 1, ... , .r.;) would be a
circuit in the forest r. Thus, Xo is terminal.
We shall prove (ii) by induction of the number m of vertices of I', the
case m = 2 being trivial. Suppose then that m ~ 3 and that assertion (ii)
is proved for graphs with m. - 1 vertices. Let a be a terminal vertex of I'
(cf. (i)). We a.pply the induction hypothesis to the full subgraph of I' whose
.vertices are the vertices x I a of I'. Thus, there exist two non-empty disjoint
subsets S~ and S~ of S with S~ u s;' = S- {a}, and such that two distinct
~ertices in S~ (resp. sn are never joined. Since a. is joined to at most one
yertex of I', we can suppose for example that it is not joined to any vertex
in S~. The partition (S~, S~ u {a.}) then has the required property. Q.E.D.
For any integer n ~ 1, denote by An the graph whose vertices are the
integers 1, 2, ... ,n and whose edges are the pairs {i,j} with i - j = ±1:
2 3 n-1 n
0-----0-----0· .......... ~o------o
A graph I' is said to be a cha-in of length m ~ 0 if it is isomorphic to Am+l·

This is equivalent to the existence in I' of an injective path (x 0 , ... , Xm)
containing all the vertices, the vertices Xi and .:r1 not being joined if \i- j\ > 1.
PROPOSITION 3. A graph is a chain 4 and only if its number of vertices is

finite and non-zero and it is a. tree with no ram·ification point.
Suppose that the graph I' is a chain (xo, ... , .:rm) with the properties listed
before the statement of Prop. 3. It is clear that any vertex of r is joined to
at most two vertices. If i < .i the path (Xi,. .. , xi) extracted from the path
(xo, ... '.:rm) joins Xi to Xj; thus, r is connected. Finally, let (.:rp,' ... '.'.l:Pn) be
a circuit in I', and let Pk be the smallest of the distinct integers p 1 , ... ,Pn·
There exist distinct indices ·i and .i such that xvk is joined to Xp, and Xp;:
this follows from the definition of a circuit. Since Pk < Pi and Pk < P.i, we
necessarily have Pi = Pi = Pk + 1, which is a contradiction since p 1 , ... , Pn
are distinct. Thus, there is no circuit in I'.
Conversely, let I' be a tree with no ramification point and with a finite
non-zero number of vertices, and let (.:r 0, ... , Xm) be an injective path of
maximal length in r. Denote by T the set of vertices other than Xo, ... 'Xm.
A vertex b E T cannot be joined to any vertex :x:;, for we would have either
a) i = 0, but then (b, x 0 , ... , Xm) would be an injective path of length m + 1
in r, or
b) i = m, but then (xo, ... , Xm, b) would be an injective path of length m + 1
in I', or
c) 0 1 and x;,x1 joined, there would be a circuit
(x;, X;+1, ... '.:Z:j) in r. Thus, r is a chain. Q.E.D.
EXERCISES
§1
1) a) Let (W, S) be a Coxeter system and let s 1 , ... , Sr be elements of S. Put

w = .s 1 ... s1 .. Show that if ls(w) < r, there exist two integers p and q with
1 ~ p < q :( r such that w = s1 ... Sp-1 Sp+l ... Sq-18q+1 ..• Sr. Show that
there is a strictly increasing sequence of integers .i (1), ... , j ( k) between 1 and
r such that (sj(l)' ... , si(k)) is a reduced decomposition of w.
b) Let (W, S) be a Coxeter system and X, Y, Z three subsets of S. Show that
Wx n (Wy.Wz) = (Wx n Wy ).(Wx n Wz)
(show that every element w E: Wv.'vVz has a. reduced decomposition
such tha.t si E Y and t.i E Z and use Cor. I of Prop. 7 of no. 8).
Show that
2) Let (W, S) be a Coxeter system and X a subset of S. Show that Wx is

normal in W if and only if every element of X commutes with every element
of S-X.
3) Let (W, S) be a Coxeter system and X, Y two subsets of S. Let a E W.

Show that there exists a unique element w E WxaWv of minimal length and
that every element
w' E WxaWv
9a,n be written in the form w' = xwy, with x E Wx,y E Wv and l(w') =
l,(x) + l(w) + l(y} (ta.kc an element of minimal length in Wxa.Wy and use
~xerc. l}. An element w E Wis said to be (X, Y)-reduced if it is the element
?f miriimal length in the double coset WxaWy.
:;:;'." Show that if w is (X, 0)-recluced then l(xw) = l(x) + l(w) for all x E Wx,
~.rid that every element of 'vV can be written uniquely in the form xw where
x E Wx and w is (X, 0)-reduced. Show that an element w E W is (X, 0)-

reduced if and only if l(xw) > l(w) for all x EX (write w in the form yw',
where y E Wx and w' i::; (X, 0)-reduced).
Show that w E Wis (X, Y)-reduced if and only if w is both (X, 0)-reduced
and (0, Y)-reduced.
4) Let n be an integer ~ 2. For any integer i with 1 ~ i ~ n - 1, denote

by s, the transposition of i and i + 1 in the set {l, 2, ... , n}, and by Hi the
set of w E 6 11 such that w- 1 (i) < w- 1 (i + l); put S = {s 1 , ... , s"_ i}. Show
that (6 11 , S) is a Coxeter system and that H;. is the set of w E 6 11 such that
l(w) < l(s(w) (use Prop. 6 of no. 7).
5) Let X be a non-empty set and W a set of permutations of X. Assume

given a set 9\ of equivalence relations on X, an element .r 0 E X and a map
cp : H ,__. SH from 9\ to W. Denote by 9\ 0 the set of H E 9\ such that
=
sH(.i; 0 ) x 0 mod. H' for all H' cl H in 9\, and by So the set of SH for H in
9\ 0 . We make the following assumptions:
(i) For any H E 9\, there are two equivalence classes modulo H that are
permuted by SH and s~ = l.
(ii) For all H E 9\ and all w E W, the transform w(H) of H by w is an
equivalence relation belonging to 9\ and Sw(H) = wsHw- 1.
(iii) For any 'UJ ;j l in W, the set of HE 9\ such that w(x 0 ) ¥ x0 mod. His
finite and meets 9\o.
a) Prove that (W, S0 ) is a Coxeter system (use Prop. 6 of no. 7).
b) Prove that the length ls 0 ( w) is equal to the number of elements H E 9\
such that
w(xo) ~ xo mod. H.
c) Let E be a finite set and X the set of total order relations on X. Denote by
W the group of permutations of E, acting in the obvious way on E. Let i and j
be distinct elements of E; say that two elements Rand R' of X are equivalent
mod. H; 1 if either R(i,j) and R'(i,j) or R(j,i) and R'(j,'i); and denote the
transposition of i and j by Sij· Let 9\ be the set of equivalence relations of
the form H,1 and <p the map from 91 to W defined by cp(Hij) = Sij; finally let
x 0 be an arbitrary element of X. Show that the objects thus defined satisfy
assumptions (i) to (iii), and recover the results of Exerc. 4.
6) Let E be a set with 6 elements and F the set of structures of the projective
line over the field F 5 on E. Denote by 6E the group of permutations of E;
for any CJ E 6s denote by (j the permutation of F induced by CJ by transport
of structure. Show that there exists a bijection u from E to F, and that the
map rY >---> u- 1 (ju is an outer automorphism of 6E (ifs is a transposition,
u- 1 su has three orbits of two elements).
§ 1. EXERCISES 33
7) Construct a group W and two subsets S and S' of W such that (W, S) and
(W, S') are isomorphic Coxeter systems, but such that there exists no inner
automorphism of W transforming S to S' (use Exerc. 4 arid Exerc. 6).
8) Construct a group Wand two subsets Sand S' of W such that (W, S) and
(W, S') are non-isomorphic Coxeter systems, one of them being irreducible
and the other not (for W take a dihedral group of order 12 generated by
{s, s'} where s and s' are of order 2 and s =f:. s', and put S = {s, s'} and
S' = {(ss') 3 ,s',s'(ss') 2 }).
9) Let (W, S) be a Coxeter system with matrix (m(s, s')), and let w+ be
the subgroup of W consisting of the elements of even length. Let so E S. Put
g 8 = sso. Show that the family (gs).sES-{so} and the relations g;'(s,so) = 1 for
m(s,so) i- oo and (g 5 g.~ 1 )m(s,s') = 1 for s,s' E S-{so} and m(s,s') =f:. oo,
form a presentation of w+. (Let H+ be the group defined by the above
presentation. Show that there exists an automorphism a of H+, with square
the identity, that transforms g_. to g; 1 for alls E S-{s0 }. If Ha is the
semi-direct product of {1, -1} and H+, relative to a, define mutually inverse
homomorphisms H0 --> W an cl W --> Ho:.) Show that if the elements of S
are conjugate (cf. Prop. 3), the group w+ is the comm·u.tator subgrov.p of W
(remark that the elements 9s are then commutators).
10) Let Un be the alternating group consisting of the permutations w E 6n

whose signature is equal to + 1. Show that U11. is the commutator subgroup
of 6 11 (use Exerc. 4 and 9). For any integer i with 1 ~ i ~ n - 2, put
Ui = SiSi+ 1 (in the notation of Exerc. 4). Show that the family (u;) and the
relations uf = 1, ·u? = 1 for i. ~ 2, (uiu.;+i) 3 = 1 for 1 ~ -i ~ n - 3, and
UiUj = UjUi for 1 ~ 'i ~ n - 4 and i + 2 ~ j ~ n - 2, form a presentation of
the group Un (use Exerc. 9).
*11) Let (W, S) be a Coxeter system. Let I'= be the graph whose set of
vertices is S, two vertices .s, .s' being joined by an edge if and only if m(s, s') =I
oo. Let Sa be the connected components of I'00 • Show that W can be identified
with the free product of the W s". In particular, every w E W can be written
uniquely as a product w 1 ... wh with w.; =f:. 1, w; belonging to W 5 0 and,
ai =I ai+ 1 for 1 ~ i ~ h - 1; show that the length of w is the sum of the 1
lengths of the wi .•
.12),Let (W, S) be a Coxeter system such that m(.s, s') is even ifs =I s' and
let X be a subset of S. Show that there exists a unique homomorphism fx
from W to Wx such that fx(s) = s for s EX and fx(s) = 1 for s E S-X.
Deduce that 'vV is the semi-direct product of Wx and the kernel of fx. Show
that if X c Y c S, there exists a unique homomorphism fx,Y from W y to
Wx such that fx = Fx,Y o fy and that W can be identified with a subgroup
of the projective system thus obtained from the Wx as X runs through the
filtered set of finite subsets of S.
, 13) Let (W, S) be a Coxeter system. For s, s' E S, define the sequence
a(s, s') by means of the following rules:
(i) if ss' is of infinite order, a(s, s') is the empty sequence;
(ii) if ss' is of finite order m the sequence a(s, s') is of length m and its even
(resp. odd) numbered terms are equal to s' (resp. s).
Denote by a(s, s 1 ) the product of the sequence a(s, s').
a) Show that the generating set S and the relations s 2 = 1 a11d a( s, s') =
a( s', s) form a presentation of the group W.
b) Let q be an integer ~ 1. Let Eq be the set of sequences of q elements of
S and let Rq be the smallest equivalence relation on Eq for which sequences
of the form (A,a(s,s'),B) and (A,a(s',s),B) (where s,s' ES and A and B
are sequences of clements of S) are equivalent. Let E~ be the set of sequences
s = (s 1 , ... , sq) such that w(s) = s 1 ... Sq is of length q. Show that sequences
s, s' E E~ are equivalent modulo Rq if and only if w(s) = w(s') (argue by
induction on q and apply Prop. 5 of no. 5).
c) Show that a sequences E Eq does not belong to E~ if and only ifs is equiv-
alent modulo Rq to a sequence in which two consecutive terms are equal. (Ar-
gue by induction on q and reduce to the case of a sequence (s 1 , ... , sq) which
does not belong to E~ but which is such that (.s 1 , ... ,s"_ 1 ) and (s 2 , ... ,sq)
belong to E;_
1; use Exerc. 1 to show that s 1 ... sq-I = s 2 ... .s 9 and apply
b).)
*14) Let (W, S) be a Coxeter system and let (r f) be its Coxeter graph. Let
k be an integer ?: 3. If a is an edge of r, put h(a) = f(a) if f(a) i oo
and fk.(a) =kif f(a) = oo. Let (W1;:, S) be a Coxeter system whose Coxeter
graph is equal to (I',h:) (Chap. V, §4, no. 3, Cor. of Prop. 4). Show that
there exists a unique homomorphism 'Pk from W to Wk inducing the identity
on S. Show that if k divides k', there exists a unique homomorphism 'Pk,k'
from Wk' to W, such that 'Pk = 'Pk,k' o 'Pl:'. Show that the homomorphism
(rpk) from W to the projective limit of the Wk is injective (use Exerc. 13) c)),
but in general not surjective (for example, in the case of the infinite dihedral
group) .•
15) 7 Let A be a set and e: a subset of ~(A). The elements of Q'. are called the
chambers of A and a subset F of a chamber C is called a facet, the codimens-ion
of F in C being the cardinal of C- F. A facet F is said to be a panel of C if
F is of codimension 1 in C. Two chambers C and C' arc said to be adjoining
7 Exercises 15 to 24, as well as Exercises 3 to 17 of § 2, are hitherto unpublished
and were communicated to us by J. Tits.
§ 1. EXERCISES 35
if they have a common panel F: then either C = C' or F =· C n C'. A gallery

of length n is a sequence I'= (Co, C1, ... , Cn) of n + 1 chambers such that
C; and C;+ 1 are adjoining for 0 :::;; i :::;; n - 1. Then Co and Cn are called the
ends of I'. The gallery I' is called injective if C; =I Ci+ 1 for.O:::;; i:::;; n -1 and
is called minimal if there is no gallery with the same ends and length < n.
The set A (together with <!'.) is a building if every element of A belongs to
at least one chamber and if any two chambers are the ends of a gallery. The
distance between two chambers C and C' is the length d(C, C') of a minimal
gallery with ends C and C'.
A sub-building of a building A is a subset D of A such that D together
with <!: n (JJ(D) is a building.
a) Show that if A is a building, a facet has the same codimension in every
chamber containing it; this allows us to speak of the codimension of a facet
and of a panel of A without reference to a particular chamber. A morphism
of a building B to A is a map f from B to A such that the restriction of f to
every chamber C of B is a bijection from C to a chamber .f(C) of A. Show
that the image under f of a facet of B is a facet of the same codimension.
b) A building is called an apartment if every panel is contained in exactly
two chambers. Show that if A is an apartment, every automorphism of A
(i.e. every permutation of A preserving <!'.) that leaves fixed all the points of a
chamber is the identity. More generally, let 'P be an endomorphism of A and
let C be a chamber of A such that ip(a) =a for all a EC. Let (C, C 1 , ... , Cn)
be a gallery of A. Show that, either the gallery (C, ip(C 1 ), ... , ip(Cn)) is not
injective, or ip(a) = a for all a belonging to the union of the Ci.
16) Let (W, S) be a Coxeter system. For s E S, denote by wrsJ the subgroup
Ws-{s} of W, by A the set of subsets of W of the form wW(s) for w E Wand
s E S, and by <!: the set of subsets of A of the form Cw = {wW(s) I s E S} for
w E W, which we call the chambers of A (Exerc. 15).
a) Show that the map w >--->Cw is a bijection from W onto<!'..
b) Show that two distinct chambers Cw and Cw' are adjoining if and only if
there exists s ES such that w' = ws. Deduce that A (together with Q:) is an
apartment (Exerc. 15), which we call the apartment of(W,S). Show that the
length of a minimal gallery with ends Cw and Gw' is equal to ls(w- 1w').
c) Let ;y- be the set of facets of A and let F E ~. Show that there exist a unique
subset X of S and an element w E W such that w W x =
.. .
n
'i. Then F is said
iEF
to be of type X. Show that the codimension of F is equal to the cardinal of
?C~ Show that the map .i : F >--->
' ·
n
i is a strictly decreasing biject'ion (with
iEF
respect to inclusion) from J to the set of subsets of W of the form wWx for
'!;.:E W and X C S. Show that every facet of type X contains a unique facet
of t;y_p.e Y for every Y such that X C Y c S.
d) Let W act on A be left translations and put C = Ce. Show that W acts
simply-transitively8 on Q'.. Let C 1 , ... , Cn be chambers of A. Show that the
following conditions are equivalent:
(i) the sequence I'= (Co = C, C 1 , ••• , Cn) is an injective gallery;
(ii) there exists a sequence S = (s1, ... , Sn) of elements of S such that C.i =
t.i(Cj_i) for 1 ~ j ~ n, where tj is the element of <P(s) defined in no. 4,
formula (11).
Show that if these conditions are satisfied, the sequence--s is unique and
Cn = s 1 ... sn(C). Show that the gallery I' is minimal if and only if the
sequence s(I') =sis a reduced decomposition of w = s 1 ... s 11 •
e) Let T be the union of the conjugates of S. Fort ET, the set Lt of points
of A invariant under t is called the wall defined by t. Show that Lt is a union
of panels and that a panel F is contained in Li if and only if j(F) is of the
form wW {s} with t = wsw- 1 . Deduce that, for any panel F, there is a unique
element t = t(F) E T such that F C L1.: L1 is callee\ the support of F.
Show that if w(Lt) = L1 (for w E W), then w = 1 or w = t.
f) Let w', w" E W. Put C' = w'(C), C" = w"(C) and let I' = (Co =
C',C 1 , ... ,Cn = C") be an injective gallery with ends C' and C". Let tj
be the element of T defining the support wall of the panel Cj n C.)_ 1 (for
1 ~ j ~ n). Show that the sequence lft(T) = (w'- 1 t_1 w')i~j~n coincides
with the sequence <P(s(w'- 1 (I')). For t E T, let n(r t) be the number of
times w'- 1 tw' appears in lft(I'). Deduce from Lemma 1 of no. 4 that the
number (-1 )n(I',t) depends only on C' and C" and not on I': we denote it by
q(C', C", t). Show that the relation 'f/(C', C", t) = 1 is an equivalence relation
between C' and C" that has two equivalence classes that are permuted by t.
Denote by <r+(t) the equivalence class containing C and by <!:-(t) the other.
Show that, for w E W and s E S, the chamber w(C) belongs to <r+(s) if
and only if l(sw) > l(w).
g) Let A+ ( t) (resp. A - ( t)) be the union of the chambers belonging to (!'.+ ( t)
(resp. ([-(t)) (fort ET). Show that A+(t) n A-(t) =Lt. (To show that
A +(t) n A-(t) c L1i reduce to the case t E S. If a E A +(t) n A-(t), put
a,= ww<sJ withs ES and w being (0,S-{s})-reduced (Exerc. 3). Ifw(C) E
'I-(t), then l(tw) < l(w) and w = ts 1 ... , sq with Sj ES and l(w) = q + 1.
Since a E A-(t), there exists :i: E w<sl such that wx(C) E <r+(t). Then
l(twx) = 1+l(w:i:)=1 + l(w) + l(x),
but since twx = s 1 ... SqX this is a contradiction. Thus, w(C) E l!:+(t) and
l(tw) = 1 + l(w). If x E W(s) is such that l(twx) < l(w:r:), deduce from
Exerc. 1 that t·wx = wx' with x' E w<sJ, and hence that ta = a and a E Li-)
8 Let H be a group and E a non-empty set on which H acts. H is said to act simply-
transitively on E if the map h ,_. h.x is a bijection from H onto .E for all x E E;
then E is said to be a pr·incipal homogeneous set for H.
§ l. EXERCISES 37
The subsets A+(t) and A-(t) are called the halves of A determined by
the wall Lt. Two points of A are said to be on the same side (resp. strictly on
opposite sides) of Lt if they belong (resp. do not both belong) to one of the
two halves. Any facet is contained in one of the halves determined by Lt. If
two facets are contained in different halves, they are said to be on opposite
sides of Li, or to be separated by L1 .
h) Let w E W. Show that l(w) is equal to the number of walls separating C
and w(C).
i) Show that the map tp that takes the half A+(t) (resp. A-(t)) to (1,t)
(resp. ( -1. t)) is a bijection from the set 9J1 of halves of A onto the set
R = {1, -1} x T (cf. 110. 4). With the notation of Lemma 1 of no. 4, show
that tp(w(M)) = Uw(<p(M)) for all w E Wand ME 911.
17) We retain the notation of Exerc. 16 and assume that W is fin'ite. Let
Sj be the set of walls of A. To any H E .fj, we associate the half H+ of A
determined by H containing C. Show that the elements of Sj can be numbered
in such a way that the map j 1-> n Ht is strictly decreasing on the interval
i~j
( 1, Card (Sj)). (Consider the family J' of intersections of the sets H + ordered by
inclusion and co11sider a strictly decreasing sequence (F 0 , ... , Fq) of elements
of J' of maximal length. For auy HE .fj, there exists an i such that H+ ~ Fj
for j :? i and H+ f; F;_ 1: show that Fi = H+ n F;_ i-)
18) Let A be an apartment (Exerc:. 15). A folding of A is an endomorphism Jr

of A such that rr 2 = 7T and such that every chamber of A is the image under
ri of 0 or 2 chambers.
a) Let 7T be a folding of A and C a chamber of A such that 7r(C) = C. For

any neighboming chamber C' of C, either rr(C') = C' or r.(C') = C; if a E C
we have -rr(a) = a. Show that, if (Co = C, C 1 , ... , C,,) is a gallery, either
rr(Ci) = C.;. for all i or (Co,rr(Ci), ... ,7r(Cn)) is not minimal (and in fact has
two consecutive equal chambers). Deduce that any minimal gallery whose
extremities are invariant under rr is invariant under 7T. lf (C = C 0 , C 1 , ••. , Cn)
is a minimal gallery and if -rr(Cn) =f. Cn, there exists an index i with 0 ~ i < n
such that rr(Cj) = Cj for 0 ~ j ~ ·i and rr(Cj) I C1 for ·i < .i ~ n.
b) Let C 1 and C 2 be two distinct neighbouring chambers and JT, 7T 1 two foldings
of A. Assume that 7r(Co) = C 1 and -rr'(C 1 ) = Co. Let C be a chamber, and
consider the following three conditions:
-rr(C) = C; (1)
d(C, Ci) < d(C, C2); (2)
7r (C)
1 i= C. (3)
Show that (1) ==? (2) ==? (3) and deduce that the three conditions are
·equivalent. Show that r. (resp. -rr') is the unique folding of A taking C 2 (resp.
Ci) to C 1 (resp. C 2) (assume that (2) is satisfied and let (C 1 , Cz, ... , C~ = C)
be a minimal gallery: show that 7r1 (Cj+ 1 ) is the unique chamber distinct from
7r'(Cj) and containing the panel 7r'(C'i n C}+ 1 )). Show that 7r(<!) and 7r 1('I)
form a partition of the set 'I of chambers of A and that 7r(a) = 7r 1 (a) = a
for all a E 7r(A) n 7r1 (A). Show that the map from A to itself that coincides
with Jr' on 7r(A) and with 7l' on 7r 1 (A) is an involutive automorphism of A. It
is called the reflection with respect to the panel C 1 n C2. Show that it is the
only non-trivial automorphism of A leaving fixed all the points of C 1 n C 2
(use Exerc. 15 b)).
c) Suppose that A is the apartment associated to a Coxeter system (W, S)
and retain the notation of Exerc. 16. Let C 1 and C 2 be two neighbouring
chambers and let t be the element of T such that the wall Lt is the support
of C 1 n C 2. Let Mj be the half of A determined by Lt and containing Cj (for
j = 1, 2). Show that the map 7l' defined by 7r(a) =a if a E M 1 and 7r(a) = t(a)
if a E M 2 is a folding of A such that 7r(C2) = C 1 and that the reflection with
respect to the panel C 1 n C2 is the map a i-+ t(a).
19) Let A be an apartment. Assume that, for any two distinct neighbouring
chambers C 1 and C2, there is a folding (Exerc. 18) of A taking C 1 to C 2 .
Let C be a chamber of A and (Ci)iEI the family of chambers neighbouring
C and distinct from C. Denote by si the reflection with respect to the panel
C n Ci (Exerc. 18 b)). Put S = {si \ i E I} and denote by W the group of
automorphisms of A generated by the Si·
a) Show that, for any chamber C', there exists w E W such that C' = w(C)
(argue by induction on d(C, C')).
b) Show that (W, S) is a Coxeter system (for i EI, put
Ps; = {w E W I w(C) C 7!'.;(A)},

where 7ri is the folding taking Ci to C, and show that the assumptions of
Prop. 6 are satisfied: to prove condition (C'), remark that if w E P s; and
WSj fl. P 8 , , then
w(C) n wsj(C) c 1t'i(A) n s.i1t'i(A).
Since w(C) and wsj(C) are neighbouring, it follows thats;= ws;w- 1 (Exerc.
18 b)).)
c) Let F be a facet of the chamber C. Show that the stabilizer WF of F
in W is generated by the si E S such that F c C n Ci (let w E W F with
ls(w) > 1 and let i EI be such that w = siw' with l(w') = l(w) -1; by Prop.
6, w' E P5 , , hence w(C) C Si7ri(A), F C 7ri(A) n sirri(A) and Si E WF)· In
particular, w(C) = C if and only if w = 1.
d) Show that the map a>---> W {a} is an isomorphism from the apartment A to
the apartment associated to (W, S) (Exerc. 16), compatible with the action
ofW.
§ 1. EXERCISES 39
20) Let A be a building and S a set. Then A is said to be numbered by S if

one is given a map f from A to S such that, for any chamber C of A, the
restriction off to C is a bijection from C to S. If F is a facet of A, .f(F) is
called the type of F. Let A be a numbered building. An endomorphism i.p of
A is called allowed if a and ip(a) are of the same type for all a EA.
a) Let i.p be an endomorphism of A. Show that, if there exists a chamber C of
A such that a and ip(a) are of the same type for all a E C, then i.p is allowed.
Show that if A is an apartment and 7r is a folding of A (Exerc. 18), then 7r is
an allowed endomorphism.
b) A subset D of A is called convex if, for all a E A- D, there exists an allowed
endomorphism i.p of A such that i.p(x) = x for all x ED and ip(a) /=a. Show
that any intersection of convex sets is convex and that, for any subset D of A,
there exists a smallest convex subset containing D: this is called the conve:i.:
hull of D and denoted by I'(D).
21) Let (W, S) be a Coxeter system and A the associated apartment (cf.
Exerc. 16, of which we retain the notation).
a) Show that there exists a unique numbering of A (called the canonical
numbering) for which the type of a facet F is that defined in Exerc. 16 c).
We shall always consider A to be equipped with this numbering.
b) Show that the allowed automorphisms of A are the operations of W.
c) Let D be a subset of A containing at least one chamber. Show that the
following conditions are cqnivalent:
(i) D is the intersection of the halves of A (Exerc. 16 g)) that contain D;
(ii) D is convex;
(iii) whenever two facets F 1 and F2 arc contained in D, the convex hull of
F 1 U F 2 is contained in D;
(iv) whenever a chamber C 1 and a facet Fare contained in D and (C1, ... , Cn)
is a gallery of minimal length such that F C C,., we have C; c D for 1 ~ i ~ n.
(To show that (iii)-==? (iv), use Exerc. 15 b). To show that (iv)~ (i), ar-
gue by contradiction. Let D' be the intersection of the halves of A containing
D; let a ED' - D, let Co be a chamber contained in D and let (C 0 , C 1 , ... , Cn)
be a gallery of minimal length such that a E Gn. Then Ci c D' for all j.
Show that there exists an integer j with 0 ~ j < n such that C.i c D and
qJ+I ¢:. D. Let M (resp. M') be the half of A determined by the support wall
9fthe panel CjnCJ+l and containing Cj (resp. Cj+d· Show that D ¢:. M. Let
.~ E D n (A- M) and let I' = (CJ, C~, ... , C~) be a gallery of smallest possible
length such that b EC~. Then C~, c D for 1 ~ k: ~ p and C~ c M'. Let 7r be
.~he;folding of A with image M' (Exerc. 18 c)): then 7r(C.i) = CJ+ 1 and the
¢allery n(I') is not injective (Exerc. 18 a)). If I'' = (CJ+ 1 , C~, ... , C~_ 2 , C~)
40 COXETER GROUPS AND TITS SYSTE!v!S Ch. IV
is the gallery obtained from rr(I') by suppressing one of the two equal con-
secutive chambers, the gallery (C1 1 CJ+ 1 , C~, ... , C~_ 2 , C~) is minimal by the
definition of I'. Deduce from (iv) that C1+ 1 c D. Contradiction.)
22) We retain the notation of Exerc. 16 and 21.

a) Lett ET and w E W. Show that the chambers C and w(C) are separated
by the wall Lt if and only if l(tw) < l(w) (use the folding determined by the
half A+ (t)).
b) Let w 0 E W. Show that the following conditions are equivalent:
l(ww 0 ) = l(wo) - l(w) for all w E W; (i)

l(tw 0 ) < l(wo) for all t ET; (ii)
whenever t E T, the chambers C and wo ( C) are separated by the wall L1 (iii)
(Use Exerc. 16 h) to show that (iii) ==? (i).)

Show that such an element wo is unique and exists if and only if W is
finite. It is then the element of greatest length of W and is characterized by
l(swo) < l(wo) for alls ES. (iv)
Moreover, w6 = l,woSwo =Sand l(wo) = Card(T).

(c) Assume that W is finite. Show that, for any chamber C0 , there exists
a unique chamber -C 0 such that C 0 U (-C 0 ) is not contained in any half
of A. Show that there exists a unique involutive automorphism <p of A (not
necessarily allowed such that <p(Co) = -Co for any chamber C 0 and that
<p(L) = L for any wall L of A. We put ip(a) = -a for a E A. If F is a facet,
the facet -F = cp(F) will be said to be opposed to F.
d) Let Co be a chamber of A and F a partition of Co. Show that the convex
hull of Co U ( - F) is the half of A determined by the wall L that is the support
of F and that contains Co.
23) We retain the notation of Exerc. 16. Let Aut(A) be the group of auto-
morphisms of the apartment A. Show that. if cp E Aut(A), then <p permutes
the walls of A, and <ptcp- 1 ET for all t ET (use Exerc. 18). Deduce that W
(identified with a subgroup of Aut(A)) is normal and that Aut(A) is the semi-
direct product of the subgroup E of automorphisms preserving the chamber
C by W. Show that the action of Aut(A) on W defines an isomorphism from
E to the group of automorphisms of the Coxeter system (W, S), or to that oJ
the Coxeter graph of (W, S) (cf. also Exerc. 19).
24) A structured build,ing is a building I equipped with a set U of sub-building:

satisfying the following conditions:
(SB 1) The sub-buildings A E U are apartments.
§ l. EXERCISES 41
(SB 2) Any two chambers of I are contained in at least one element of U.

(SB 3) Whenever A 1 , A2 EU are such that A 1 nA2 contains a chamber, there
exi::;ts an isomorphism from A 1 to A 2 leaving fixed the points of A 1 n A 2.
Let (I, U) be a structured building. The elements of U are called the apart-
ments of (I, U), or simply those of I.
a) Show that any two apartments of I are isomorphic. Let C be a chamber
of I and A an apartment of I containing C. Show that there exists a unique
endomorphism p of I (called the retraction with centre C of I onto A) such
that p(a) = a. for all a E A and that, for any apartment A' containing C, the
restriction of p t.o A' is an isomorphism from A' to A (remark that, by (SB
2) and Exerc. 15 b), for any apartment A' containing C, there exists a unique
isomorphism PA' of A' leaving all the points of C fixed). Show that p2 = p
and that p- 1 (C) = C.
b) Let A be an apartment of L Co a chamber and F a facet contained in A. Let
(C 0 , C 1 , ... , Cn) be a gallery of smallest possible length such that F C Gn..
Show that C; c A for 1 :::;; i :::;; n (argue by contra.cliction: if C; C A and
C;+J <t A, consider the retraction of I onto A with centre the chamber of A
distinct from G; and containing(\ n C;+il·
c) Let A be an apartment of I, Ca chamber of A, Fa panel of C, C' a chamber
of I and I'= (C 0 , ... , Cn = C') a gallery of smallest possible length such that
F c C 0 . Show that the retraction p of I onto A with centre C transforms I'
into a gallery
(C~, ... , C~ c;;: p(C'))
of smallest. possible length such that F c C~ (consider an apartment A'

containing C' and F and apply b) and the fact that the restriction of p to A'
is an isomorphism).
d) Let A be an apartment., C 1 and C 2 two distinct neighbouring chambers
contained in A, C' a chamber containing C 1 n C2 and distinct from C 1 and
C2 , a.nd A~ an apartment containing Ci and C' (for 'i = 1, 2). Let <p; (resp.
1/Ji) be the retraction of I onto A (resp. A;) with centre C; (resp. C') and let
p; be the restriction of t.p; o 1/J; to A. Let C be a chamber of A; show that if
,then P1(a) =a for all a E C and d(C, Ci) < d(C, C2) (consider a minimal
gallery I' with extremities C and C 1 and apply c) to show that p1 (I') is
·minimal; then use Exerc. 15 b)). Show that, if d(C, C 2 ) < d(C, C 1), then
)J1(C) =IC and PI(C) = p1 (C) (take a minimal gallery I' with extremities C
•and C2 and use c) to show that p 1 (I') is a gallery of smallest possible length
·having one extremity equal to p1 (C) and the other containing C 1 nC 2 ; deduce
':.that d(C 1 ,p 1(C 1 )):::;; d(C 2 ,p1(Ci))).
Show that Pi is a folding (Exerc. 19) of A (show that A= p 1 (A) U P2(A)

and define an involutive automorphism <J of A by putting <J(o.) = p 2(a) if
a E p1 (A) and <J(a) = p1 (a) if a E P2(A); show that if C is a chamber
contained in p 1 (A), then pj 1 (C) = {C,p2(C)}).
e) Let I be a spacious structured building, that is to say that every panel
is contained in at least three chambers. Show that there exists a Coxeter
system (W, S), unique up to isomorphism, such that the apartments of I are
isomorphic to the apartment Ao associated to (W, S) (use d) and Exerc. 19)).
Let A be an apartment of I and r.p an isomorphism from Ao to A. Show that
there exists a unique numbering of I, with values in S, such that the types of a
and r.p( a) are equal for all a E Ao (choose a chamber C of A and show, by using
b), that, if A' and A" are two apartments of I containing C, the numberings
of A' and A" extending that determined on C coincide on A' n A").
We shall say that the Coxeter system (W, S) and the numbering thus
obtained are adapted to I.
Show that the retractions introduced in o.) are allowed endomorphisms.
Show that a subset of I contained in an apartment A of I is convex in I if and
only if it is convex in A (Exerc. 20) .
.f) We retain the notation of e) and let 2l be the set of allowed isomorphisms
from Ao to the various apartments of I. Show that, if r.p, 'l/; E 2l and if C
is a chamber and F a facet of I contained in r.p(Ao) n 'lj;(A 0 ), there exists
an element w E W such that .,µ- 1 (C) = wr.p- 1 (C) and 1,b- 1 (F) = wr.p- 1 (F)
(consider an isomorphism A from ip(A 0 ) to 'ljJ(A 0 ) leaving fixed the points of
r.p(Ao) n 1/;(Ao) and apply Exerc. 21 b) to the automorphism 1/1- 1 o Ao <p- 1 of
Ao).
25) Let I be a set and let J" be the set of finite subsets of I. For any A E ~'
put E(A) = (-l)Card(A). Let G be an abelian group written additively, and
let r.p and 1,b be two maps from J to G. Show that the following properties are
equivalent:
r.p(A) = L
Be A
'l/;(B) for all A E J; (i)
1/;(A) = L c(A - B)r.p(B) for all A E ~- (ii)

BcA
26) Let (W, S) be a Coxeter system, with S finite. For any subset H of W,
denote by H(t) the formal power series wit.h integer coefficients defined by
H(t) = L t1(w).
·wEH
a) Assume that Card(S) = 2. Show that

W(t) = (1 + t - {m - tm+l )/(1 - t) if W is of finite order 2m,
W(t) = (1 + t)/(1 - t) if Wis infinite.
§ 1. EXERCISES 43
b) Assume that W is .finite. Let wo be the element of W of greatest length

(Exerc. 22) and let m = l(w 0 ). Show that
W(t) = tmW(C 1 )
(use Exerc. 22).
c) Let X be a subset of S. Denote by Ax the set of (X, 0)-reduced elements of
W (Exerc. 3) and Wx the subgroup of W generated by X. We know (Exerc. 3)
that an element w E W belongs to Ax if and only if l(xw) = l(w) + 1 for
all x E X, that every element w E W can be written uniquely in the form
w = v.v with u E Wx and v E Ax, and that in that case l(w) = l(u) + l(v).
Deduce the formula
W(t) = Wx(t).Ax(t).
d) Retain the above notation and denote by Bx the set of w E Ax such that
l(sw) = l(w) - 1 for alls E S-X. Show that
Ax(t) = L Bv(t).
XCYCS
Deduce that
Bx(t) = L E(Y-X)Ay(t) with E(Z) = (-lfard(Z)
XCYCS
(use Ex ere. 25).

e) Assume that W is finite and define m. and w 0 as in b). Show that B 0 = {Wo}
and that
tm = L E(Y) W(t)
Yes Wy(t)
(use c) and d)).
f) Assume that W is infinite. Show that B 0 = 0 and that
0= L E(Y) .
Yes Wv(t)
g) Show that the formal power series W(t) is a rational function oft (use
f) and argue by induction on Card(S)). Show that this rational function
vanishes whenever tis a root of unity, and that 1/W(oo) is an integer. Show
that wd-i) E Z[[t]].
44 COXETER GROUPS AND TITS SYSTE!v!S Ch. IV
§2
1) Let G be a group, B and N two subgroups of G and S a subset of

W = N/(B n N). For all w E W, put C(w) = BwB. Assume that conditions
(Tl) and (T2) of Def. 1 of no. 1 are satisfied, that whenever s E Sand w E W
at least one of the two relations C(s).C(w) = C(sw) and C(s).C(sw) = C(w)
is satisfied, and that BU C(s) is a subgroup of G for all s E S. Show that
condition (T3) is satisfied. Moreover, if the index of B in B U C(s) is ? 3,
then (G, B, N, S) is a Tits system.
2) Let G be a group, B and N two subgroups of G and S a subset of

W = N/(BnN). Let Z be a normal subgroup of G contained in B. Let B' and
N' be the canonical images of B and N in G' = G /Z. Show that the canonical
map from N to N' defines an isomorphism from W to W' = N' /(B' n N'). Let
S' be the image of S under this isomorphism. Show that (G', B', N', S') is a
Tits system if and only if (G, B, N, S) is one.
3) Let G be a group, Ba subgroup of G and (C(w))wEW the family of double

cosets of G with respect to B. Then B is called a Tits subgrov.p of G if there
exists a subset S of 'vV such that the following conditions are satisfied:
(1) the union of the C(s) for s ES generates G;
(2) for all s E S, the set BU C(s) is a subgroup of G and the index of B in
BU C(s) is~ 3;
(3) for all s E S and all w E W, there exists an element w' E \h/ such that
C(s).C(w) c C(w) U C(w').
From now on we assume that B is a Tits subgroup of G and that we are
given a subset S of W satisfying conditions (1), (2) and (3).
a) Show that C(s)- 1 = C(s) and that C(s).C(s) = BU C(s) for alls E S.
Show that, for alls E Sand all w E W. there exists an element w" E W such
that C(w).C(s) C C(w) U C(w").
b) If w E W, the length of w, denoted by l(w), is the smallest integer n ~ 0
for which there exist s 1 •... , Sn ES with C(w) C C(s 1 ) ... C(sn)· Show that
l(w) is finite for all w E W.
Let u,v E W with l(v.) < l(v) and lets ES. Show that, if C(v) C
C('u).C(s) (resp. C(v) c C(s).C(u)), then C(v) = C(ti).C(s) (resp, C(v) =
C(s).C(u)) (argue by induction on the length of u. If C(v) I- C('u).C(s), then
C(u).C(s) = C(u)UC(v). By using the induction hypothesis, show that there
exist t E: S and w E \!'./ such that
C(u) = C(t).C(w) with l(w) = l(v.) - 1.
From the relations C(v) C C(v).C(s) =- C(t).C(w).C(s) and C(t).C(w) =

C(u) I- C(v), deduce the existence of an element w' =f- w such that
§ 2. EXERCISES 45
C(w') c C(w).C(s) and C(v) c C(t).C(w'), so l(w') ·~ l(v) - 1 > l('u) - 1 =

l( w). The induction hypothesis now implies that
C(w') = C(w).C(s).
Moreover,
C(t).C(u).C(s) = C(t).C(t).C(w).C(s) = C(t).C(w).C(s) U C(w).C(s)

= C(u).C(s) U C(w') = C(u) U C(v) U C(w')
and also
C(t).C(u).C(s) = C(t).(C(u) u C(v)) :J C(t).C(t).C(w) :J C(w),
which is a contradiction since w i- u, v, w'.)

c) Show that, for all w E W and all s E S, there exists a unique clement,
denoted by s.w (resp, w * s), distinct from wand such that
C(s.w) c C(s).C(w) C C(w) U C(s.w)

(resp. C(w * s) c C(w).C(s) C C(w) U C(w * s)).
(Show by induction on l(w) that C(s).C(w) i- C(w). For this, write
C(w) = C(u).C(t)
with t ES, u E Wand l(w) = l(u) + 1. If C(s).C(w) = C(w), then
C(s).C(u).C(t) = C(u).C(t)
and multiplying on the right by C(t), we obtain that C(u) U C(w) =
C(s).C(u) U C(w). Since C(u) f C(s).C(u) by the induction hypothesis, we
have, by b),
C(.s).C(v,) = C(w),
and hence
C(v.) c C(s).C(s).C(v) = C(s).C(w) = C(w),
which is absurd).
d) Let s E S. Show that the map Ps : w f---7 s.w (resp. q5 : w ,_.. w * s) is a
permutation of W and that p; = Id (resp. q,~ = Id). Show that Ps o q1 = qt o Ps
for alls, t ES (study the product C(s).C(w).C(t) for w E Wand show that
(s.w)*t E {w,s.W,W*t,s.(w*t)}; show that (s.w)*t (/: {s.w,w*t} and that
~f.(s.w) * t = w then s.w = w * t and w = s.(w * t)).
e) Show that the group of permutations P (resp. Q) generated by the Ps (resp.
Qs) for s ES acts simply-transitively on W (to show that Pis transitive, use
(2) and argue as in Lemma 1 of no. l; to show that Pis simply transitive, use
d)), Deduce that W has a unique group structure such that the map p ,_.. p(e)
(where e denotes the element of W such that B = C(e)) is an isomorphism

from P to W. The map q .....-. q(e) is then also an isomorphism; moreover,
s.w = w * s = sw for alls ES and all w E Wand C(w)- 1 = C(w- 1 ).
f) Show that the pair (W, S) is a Coxeter system and generalize the results
of no. 4.
g) Let X be a subset of Sand Wx the subgroup of W generated by X. Show
that the union Gx of the C(w) for w E Wx is a subgroup of G and that
Th. 3 of no. 5 is still true. Show that B is a Tits subgroup-of Gx. Generalize
Props. 2 and 3 of no. 5, and Def. 2, Prop. 4 and Th. 4 of no. 6. Show that
S is the set of w E W such that BU C(w) is a subgroup of G distinct from
B. The Coxeter system (W, S) and the group W (called the Weyl gro·up of
(G, B)) thus depend only on (G, B).
h) Let N be a subgroup of G such that B n N is normal in N and such that
the intersection with N of every double coset C( w) with respect to B is a
double coset with respect to B n N. Show that the group N/(B n N) can be
identified with Wand that (G, B, N, S) is a Tits system.
4) Let (G,B,N,S) and (G',B',N',S') be two Tits systems with G = G' and
B = B', and with Weyl groups Wand W'. Let j be the bijection from \\T to
W' defined by the relation
BwB = B'j(w)B'.
Show that j is a group isomorphism between W and W' and that j(S) = S'.
5) Let E = (G, B, N, S) be a Tits system. Put T = B n N and let N be the
normaliser of N.
a) Let b EB n N. Show that bnb- 1 n- 1 EB n N for all n EN (put bn = n'b
and use Th. 1) and that b belongs to the intersection T of the conjugates
nBn- 1 for n EN. Show that T n N = T.
If T = T, E is said to be satv.rated.
b) Put N = N.T. Show that N is a subgroup of G containing T a.5 a normal
subgroup and that N n B = T. Show that the injection of N into N defines
an isomorphism from the Wey! group W of E to N/T.
c) Show that (G, B, N,j(S)) is a satmated Tits system, which is said to be
associated to E.
6) We use the notation of no. 2 and let No be the subgroup of N consisting

of the matrices all of whose entries are equal to 0 or l. Show that B n No ==
T n N0 = {1} and that the canonical map .i from N0 to W = N /T is an
isomorphism. Put So= F 1 (S). Show that (G, B, N0 , S0 ) is a Tits system and
that (G, B, N, S) is the associated saturated Tits system.
§ 2. EXEHCISES 47
7) Let G be a group acting on a set E. Then G is said to act doubly-transitively

on E if, whenever .x, y, x', y' E E are such that .1: :f y and x' :f y', there exists
an element g E G such that g.x = .x' and g.y = y'.
a) Let (G, B, N, S) be a Tits system whose Wey! group Wis of order 2. Show
that G acts doubly-transitively on G/B.
b) Let G be a group acting doubly transitively on a set E. Assume that
Carcl(E) ~ 3. Let e E E and let B be the stabilizer of e. Let x E E, with
x :f e, and let n E G be such that n(e) = .x and n(:z:) = e. Let N be the
subgroup of G generated by n. Let s be the canonical image of n in N/T.
Show that (G, B, N, {s}) is a Tits system whose Wey! group is of order 2.
8) Let (G, B, N, S) be a Tits system; put T = B n N and W = N/T. Let G

be a group containing G as a normal subgroup. Assume that for all h E G,
there exists g E G such that hBh- 1 = gBg- 1 and hNh- 1 = gNg- 1 . Let B
(resp. N) be the normaliser of B (m;p. N) in G; put r = B n N, N = r.N
and T = N (1 B.
a) Show that G = I'.G, B = I'.B, r r
n B = n G and t = (I' (1 B).T. The
groups fl = I'/(I' n B), G/G and B/B are thus canonically isomorphic. If
<J> c fl and if H is a snbgroup of G containing I' n B, we denote by 1>H the
union of the subsets <pH for 'P E <J>.
b) Show that T is normal in N (to show that n1n- 1 E N for n E N and
1' E I'n B, use Exerc. 5 a)), that N n T = T and that I' n T = I' n B.
The injection of N (resp. I') into N thus allows W (resp. Jl) to be identified
with a subgroup of vV = N/T. Show that fl normalises Sand that Wis the
semi-direct product of fl and W.
c) Show that, for all s ES and all ·u E \V.
BsBv.B c (BuB) U (RmB).
d) Show that the map ·u r-> BuB is a bijection from W to B\G/B (use Th. 1
and the fact that I' normalises B).
e) Let '13 be the set of pairs (<J>, X), where <!> is a subgroup of fl and X is
a subset of X normalised by <P. Put G(</>,X) = <l>Gx = B1>WxB (with the
notation of no. 5). Show that the map (<!>, X) f-t G(<J\X) is a bijection from '13
lo the set of subgroups of G containing B. Generalize assertions b) and c) of
Th. 3 and Prop. 2 of no. 5.
~··Show that the normaliser of G(<P,X) in G is the subgroup G(<P',X)• where
<P' is the set of elements of fl normalizing both is a maximal subgroup of fl;
(ii) acts transitively on S-X (which is non-empty).

Show that G(<P,X) is a ma.ximal s11bgroup in the set of ::;ubgroup::> of G not
containing G if and only if
(iii) X :f:. S, rJ) is the normaliser of X in [l and acts transitvcly on S- X.
g) Let commutes with every element of 'vV. Show that G'
is then normal in G, that the injection of N into N' defines an i8omorphism
j from W to W' = N'/T' and that (G',B',N',S') (with S' = j(S)) is a Tits
system.
9) Let (G, B, N, S) be a Tits system and X, Y, Z three subsets of S. Show that
Gx n (Gv.Gz) = (Gx n Gv ).(Gx n Gz)

(use Exerc. 1 of§ 1 and Prop. 2 of no. 3).
10) Let G be a group and B a Tits subgroup of G. We use the notation of

Exerc. 3. Fors ES, denote by G(s) the subgroup Gs-{ 8 } (Exerc. 3 g)). Let I
be the set of subsets of G of the form gG(s) (for g E G and s E S), and C!:: the
set of subsets of I of the form C 9 = {gG(s) I s E S} for g E G. The C9 are
called the chambers of I (§ 1, Exerc. 15). Let G act on I by left translations.
a) Let ~ be the set of facets of I (that is, the subsets of the chambers, cf.
§ 1, Exerc. 15). Let FE~. Show that there exist a unique subset X of Sand
an element g E G such that gGx = n
a; F is said to be of type X. Show
aEF
that F is then the set of gG(s) for s E S- X and is of codimension Card(X)
in any chamber containing it. Show that the map .i : F f-t a i8 a strictlyn
aEF
decreasing bijection, compatible with left translations, from ~
to the set of
subsets of G of the form gGx for g E G and X c S. Show that, if X C Y C S,
every facet of type X contains a unique facet of type Y.
b) Show that G acts transitively on the set C!:: of chambers and that the
stabilizer of the chamber C 9 (g E G) is equal to gBg- 1 ; the map g t---> C9
thus defines a bijection from G/B to C!::.
c) Show that two chambers C 9 and C 9 , (g, g' E G) are ne'ighbonring (§ 1,
Exerc. 15) if and only if there exists s ES ::;uch that 91 E g(B U BsB).
d) Let C 1 , ••• , Cn be chambers of I and put Co = Cc. Show that the following
conditions are equivalent:
(i) the sequence I'= (C 0 , C 1 , ... , Cn) is an injective gallery;
(ii) there exist a sequences= (s 1 , ... , sn) of elements of S and a sequence
( b1 , ... , b11 ) of elements of B such that Ci = b1 81 b2s2 .•. b.i si (Co) (where 81
denotes a given element of the double coset Bs.iB) for 1 ~ j ~ n.
§2. EXERCISES 49
Show that, if these conditions arc satisfied, the sequence s is unique: it is

then called the type of I' and denoted by s(r). Show that an injective gallery
is minimal if and only if its type is a reduced decomposition. Show that the
following conditions are satisfied:
(WI 1) For any two chambers C and C' of I, lhf.Te exists a uniqite element
t(C, C') of W such that the set of types of minimal galleries with extremities
C and C' is the set of reditced decompositions of t(C, C').
(WI 2) For any chamber C, the map C' ~ t(C, C') from the set of chambers
of I to W is S'urjective.
e) Show that two minimal galleries of the same type and with the same
extremities are identical. (Reduce to proving that, if (s 1 , ... , s,i) is a re-
duced decomposition and if bi, .. . , bn. b;, . .. , b~. arc elements of B such that
b1 s1 .. . bnsn E b;s1 ... b~sn.B, then b1.51 E Vis1B. For this, remark that if
s\ 1 b\ 1 b;.~ 1 ~ B, then this element would belong to Bs 1 B and we would have
b2 s2 ... bnsn E bs 1 b1 b;:s2 ... b~.snB with b, b' E B, contradicting Cor. l of Th. 2
of no. 4.)
f) Show that I equipped with <!'. is a building, said to be associated to the
pair (G, B). Show that there exists a unique 1wmberin9(§1, Exerc. 20) of I
for which the type of a facet is that defined in a).
g) Show that I is spacious, namely that every panel of I is contained in at
least three chambers (cf. § 1, Exerc. 24). Show that the following condition is
satisfied:
(G) Given a panel F, a chamber C, a gallery I' = (Co, .... Cn) .mch that
F c C.n and of smallest possible length, and chambers C' and C" containing
F and distinct from C,., there exists an element g E G such that g(Ci) =Ci
forO ~ i ~ n and g(C') = C".
(Reduce to the case Co = C; let u E S be such that F is of type S- {'n} and
let s be the type of I'. Show that (s, u) is a reduced decomposition. Take
h E G such that Cn == h(C); there exist b' and b" E B such that C' = hb'u(C)
and C" = hb"u(C). Then use Cor. l of Th. 2 of no. 4 generalize<l to the case
of a Tits subgroup (cf. Exerc. 3 f)) to show that there exists b E 13 such
that bhb'uB = hb"uf3; then b E hBuB1i- 1 B1i- 1 c (hBh- 1 ) u (hBuBh- 1 ). If
b E hBuBh- 1 , we would have BhB C Bh.BuB = BhuB (loc. cit.), which is
impossible. Hence b(G11 ) = Cn and e) implies that b(Ci) =Ci for a.II i.)
11) Let (W, S) be a Coxetcr system and I a building numbered by S (§ l,

Exerc. 20). If I'= (C 0 , ... , Cn) is an injective gallery, the type of I', denoted
by s(I'), is the sequence (s 1 , ... , sn) of elements of S such that the panel
(\_1 n C;. is of type S- {si} (for 1 ~ i. ~ n). If conditions (WI 1) and (WI 2)
of Exerc. 10 are satisfied, I is called a (W, S)-building.
, Let I be a spacious (W, S)-building (cf. Exerc. 10 g)) and let G be a group
ofallowed automorphisms of I satisfying the following condition:
(Go) Given three distinct chambers C, C' and C" containing the same panel,
there exists an clement g E G such that g(C) =C and g(C') = C".
Choose a chamber C of I and denote by B the stabilizer of C iu G.
a) Show that G acts transitively on the set of chambers of I.
b) Let C' and C" be two chambers. Show that t(C, C') = t(C, C") if and
only if there exist:::; b E B such that C" = b(C') (if t(C, C') = t(C, C") = w,
consider a reduced decomposition s of w alld a minimal gaJlery I'' (resp. I'")
with extremities C and C' (resp. C") and of type s. Argue by induction on
l(w) using (Go) and a)). Deduce that there exists a bijection w r---> B(w) from
W to B\G/B.
c) Let F be the facet of C of type S- {s}. Show that the stabilizer of F in
G is equal to BU B(s) (if g(F) = F, then t(C, g(C)) = 1 or s). Show that
BC B(s)B(s) (use (Go)).
d) Show that B is a Tits subgroup of G and that the Coxeter system of
(G, B) is canonically isomorphic to (W, S). (Let w E W and s E S be such
that ls(sw) = ls(w) + 1. Let g E B(w) and v E B(s); lets= (s 1 , ... ,sn)
be a reduced decomposition of w, (C = C 0 ,C 1 , .. .,Cn = g(C)) a minimal
gallery of types and choose an element.<;; E B(si); show, by using (Go), that
there exist elements b; EB such that G;. = bis 1 •.• b;s 1 (C). Put C'. = 11(Ci-i)
and show that the gallery (C,'u( C), u( C 1 ), ... , v.( Cn)) is of type ( s, s). and is
therefore minimal. Deduce that ug E B(sw) and that B(s)B(w) = B(sw). If
now l(sw) = l(w) - 1, put w' = sw: theu B(s)B(w') = B(w) and hence
B(s)B(w) = B(s)B(s)B(w') c B(w') u (B(s)B(w')) = B(sw) U B(w);
finally, since B c B(s)B(s), we also have B(w') = B(sw) c B(s)B(w).)

e) Show that there exists a unique isomorphism from the building a;;sociatcd
to (G, B) (Exerc. 10) to I, compatible with the action of G and taking tlw
canonical chamber Ce to C.
12) Let (G, B, N, S) be a Tits system, W = N/(B n N) its Weyl group, I the
(W, SJ-building associated to (G, B) (Exerc. 10) and C = Ce the canonical
chamber of I. Let A 0 be the apartment associated to the Coxctcr system
(W, S) (§ 1, Exerc. 16). For all g E G, let r.p 9 be the map from Ao to I that
takes a point wW(s) of Ao (w E W, s ES) to the point gwG(s) of I.
Show that, for all g E G, the map <p9 is an isomorphism of numbered
buildings from Ao to a subset of I that is the union of chambers gn(Ce)
for n E N. Show that I, equipped with the set ti of \Og(A 0 ) for g E G,
is a structured building (§ 1, Exerc. 24) (to prove (SB 2), remark that, if
g',g" E G, there exist b',b" EB and n EN such that 91 - 1 911 = l/nb";
then put g = g'b'n and show that g'(C) and g"(C) are contained in \Og(Ao)·
To prove (SB 3), reduce to the case of two apartments A' = 'Pe(Ao) and
§ 2. EXERCISES 51
A"= <l'b(A 0 ), with b EB, and show that the map a........, b(a) leaves the points
of A' n A" fixed by using Prop. 2 of no. 5). Show that the Coxeter system
(W,S) and the numbering of I are adapted to (1,11) (§ 1, Exerc. 24 e)) and
that the set J of allowed isomorphisms from Ao to the various elements of 11
is the set of <p9 for g E G.
13) We use the notation of Exerc. 24 of § 1: (I, 11) is a spacious structured

building equipped with a Coxeter system (W, S) and an adapted numbering
and J is the set of allowed isomorphisms from the apartment Ao associated
to (W, S) to the various apartments of (I, il). Further, let G be a group of
allowed automorphisms of I, preserving E: the group G then acts on J and
we assume that G acts transitively on J.
Denote by C a chamber of I, A an apartment of 11 containing C, <p an
allowed isomorphism from Ao to A taking the canonical chamber Ce of Ao
to C, B the stabilizer of C in G and N the stabilizer of A in G.
a) Show that, if A' and A" are two apartments belonging to 11, and containing
the same chamber, there exists g E G such that g(A') =A" and that g(a) =a
for all a E A' n A". Show that G acts transitively on the ::;et of pairs (A, C)
where A E 11 and C is a chamber of A.
b) Show that the map n........, <p- 1 on o <pis a surjective homomorphism from N
to W, with kernel B n N. We can thus identify N/(B n N) with W.
c) Show that conditions (WI 1) and (WI 2) of Exerc. 10 d) are satisfied (use
the following fa.ct: if an apartment of I contains two chambers C' and C", it
contains every minimal gallery with extremities C' and C" (§ 1, Excrc. 24 b))).
d) Show that condition (G) of Excrc. 10 g) (and a fortiori condition (Go) of
Exerc. 11) is satisfied (with the notation of (G), consider an apartment A'
(resp. A") of I containing Co and C' (resp. C"); show by using Exerc. 24 b)
of§ 1 that I' c A" n A" and use a)).
e) Show that (G, B, N, S) is a Tits system and that (I, U) is canonically iso-
morphic to the associated numbered structured building (Exerc. 12).
14) Let (G, B, N, S) be a Tits system and (I,11) the associated numbered
structured building (Exerc. 12). Show that G, considered as a group of au-
tomorphisms of I, satisfies the conditions of Exerc. 13. With the notation of
Exerc. 12, put A= 'Pe(A 0 ) and C = 'Pe(Ce); show that B is the stabilizer of
C in G. Let N be the stabilizer of A in G; show that N/(B n N) can be iden-
tified with W and that (G, B, N, S) is the saturated Tits system associated to
(G,B, N, S) (Exerc. 5).
15) We use the hypotheses and notation of Exerc. 10. Assume moreover that
the Weyl group W of (G, B) is finite and denote by w 0 the longest clement
of W (§ 1, Exerc. 22). Two chambers C and C' are said to be opposed if
:t(C, C') = w 0 •
a) Show that if C and C' are opposed, so are C' and C. Show that there
exists a chamber opposed to any given chamber C. Show that the stabilizer
of a chamber C in G acts transitively on the set of chambers opposed to C.
b) Let C and C' be two opposed chambers. Show that for all w E W, there ex-
ists a unique chamber Cw having the following property: if (s 1 , ... , sk) (resp.
(s'1 , ... , s;,.)) is a reduced decomposition of w (resp. of w' = wow- 1 ), there ex-
ists a minimal gallery (Co= C, C 1, ... , Gn = C') of type (s1, ... ,sk, s~, ... ,s/J
(with n = h + k) such that Gw = Ck (use Exerc. 22of~1 and Exerc. 10, d)
and e)). Show that Cw and Cwwo are opposed. .
c) Let 9J1 be the set of pairs of opposed chambers. For m = (C, C') E 9.n,
let Am be the union of the chambers Cw constructed above. Show that I
equipped with the set U of Am for m E 9J1 is a structured building (§ 1,
Exerc. 24) and that the Coxeter system (W,S) and the numbering of I are
adapted to (I, ti); define a canonical bijection from 9J1 to the set denoted by
J in Exerc. 24 of § 1. We identify these two sets.
d) Let m = (C, C') E 9J1 with C = Ce, and let N be the stabilizer of Am in
G. Show that N/(B n N) can be identified with W and that (G, B, N, S) is a
saturated Tits system (use Exerc. 13).
16) We retain the hypotheses and notation of Exerc. 15.

a) Let C and C' be two chambers. Show that there exists a chamber C"
opposed both to C and to C' (take C" opposed to C such that t(C, C') b of
greatest possible length; if t(C', C") "I w 0 , there exists a neighbouring cham-
ber C 1 of C" such that l(t(C',C 1 )) > l(t(C',C")); show by using condition
(G) of Exerc. 10 that we can assume that C 1 rf-. A(c,c'') and that C 1 is then
opposed to C).
b) Let a EU; show that there exists a unique involutive automorphism )A (not
necessarily allowed) that takes every chamber of A to the opposed chamber
(use Exerc. 22 c) of§ 1). Let F and F' be two facets of I; show that if F' =
)A(F) for some A EU containing F and F', then the same is true for a.II A Ell
containing F and F' (if F, F' E An A', with A, A' E ti, consider a chamber C
(resp. C') of A (resp. A') containing F (resp. F') an,d an A" E U containing
C and C' and use (SB 3)). Then F and F' are said to be opposed. Show that
two facets have a common opposed facet if a.nd only if they are of the same
type T and that a facet opposed to a facet of type T is of type w 0 Tw 0 1 .
c) Let Ao be the apartment associated to a Coxeter system (W, S) (§ 1, Exerc.
16) and let J be the set of allowed isomorphisms from Ao to the various
elements of ti. If o: is a half of A0 , with wall L (§ 1, Exerc. 16) and if cp E '.J,
we say that cp(a) is a semi-apartment of I, with wall cp(L). Show that cp(L)
then depends only on <p(a) and not on the pair (cp,o:). Let D 1 and Dz be
two distinct semi-apartments, with the same wall L: show that there exist
cp E J and a. wall Lo of Ao such that L = <p(Lo) and that the D; are the
§ 2. EXERCISES 53
images under <p of the two halves of A0 determined by Lo (choose a panel F

contained in L and two chambers C 1 and C 2 , one contained in D 1 and the
other in D 2 and such that C 1 contains F and Cz contains the panel opposed
to Fin D 1 (which is also opposed to Fin Dz) and consider the apartment of
I containing C 1 and C2).
17) We retain the hypotheses and notation of Exerc. 15 and 16. Choose an
element <p E J taking the canonical chamber C of A0 (§ 1, Exerc. 16) to the
canonical chamber Ce of I and, as in Exerc. 15 d), denote by N the stabilizer
of the apartment cp(Ao) E 11 in the group G. For any subset D of A 0 , denote
by Bo the subgroup of G leaving fixed all the points of cp(D): we have B =Be
and B n N = B Ao.
a) Let o: be a half of Ao containing C. Show that Ba acts transitively on
the semi-apartments of I distinct from cp( a) whose wall is the wall L of cp( o:)
(let X be such a semi-apartment, F a panel contained in L and C' a chamber
of cp(o:) containing F; show, by using Exerc. 16 c) and Exerc. 13 a), that.
there exists g E G such that g(X U cp(o:)) = cp(Ao) and g(C') = C'; show
that g(F) = F and deduce from Exerc. 22 c) of§ 1 that g E Bal· Deduce
that BL acts doubly transitively on the semi-apartments with wall L and
that (BL, B°', BL n N) is a Tits system whose Wey! group is of order 2 (cf.
Exerc. 7).
b) Let D 1 and Dz be two convex subsets of Ao such that C c D 1 c Dz
and that there exists a unique half o: of Ao with 0 1 C o: and D2 c/- o::
then D 1 = D 2 no: (§ 1, Exerc. 21 c)). Show that Bo, = BaB0 2 and that
Ban B02 = B n N (by considering a gallery of smallest possible length having
one extremity equal to C and the other containing a point a E Dz - D 1 ,
show that there exist two neighbouring chambers C~ and C~ of A0 such that
C~ c D 1 , Cz c D2, C~ cf. D 1 and C~ n c; contained in the wall L' of o:.
Put C, = cp(C~), F = cp(C~ n C2) and L =- cp(L'). Show that the convex hull
of D 1 Uc; is equal to Dz. Let b E Bo,: then b(Ci) = C 1 , so b(F) = F and
b(C2) i= C 1 . Deduce that the convex hull of b(Cz) UL is a semi-apartment X
distinct from cp( 0:) and with wall L, and that there exists b' E Ba such that
b1 (<p(Ao)) =XU <p(o:).
Show that b'b(a) =a for all a E cp(D 1 UC~) and that b'b E Bo 2 ).
c) Let (o:i)i(i(q be the halves of Ao containing C, numbered so that the map
.i I-+ n
l(i(j
o:,
s
~s strictly decreasing (cf. 1, Exerc. 17). Show that B = B 0 , . . • Bu 0 and that
if bi,b~ E B0c, with b1 ... bq =Vi ... b~, then b~ E b;(B n N) for 1,::;: i,::;: q.
18) We use the hypotheses and notation of no. 2. Let V, be the vector sub-
:Space of kn generated by c 1 ..•• , e; (for 1 ,::;: i,::;: n - 1).
a) Show that, for any subset X of S, the subgroup Gx (no. 5) consists of the
elements g E G such that g(Vi) =Vi for all i such that si 1- X.
b) Let I be the building associated to a Tits system (G, B, N, S) (Exerc. 10
and 12). Show that the map j that associates to the point gG(i) of I (where
G(i) denotes the subgroup Gs-{s;} of G, in other words the stabilizer of Vi)
the vector subspace g(Vi) is a bijection from I to the set~ of vector subspaces
of kn that are 1- {O} and f. kn, compatible with the action of G.
c) If E is a vector space, a fiag of E is a set of vector sub~paces of E that is
totally ordered by inclusion. Show that elements a1, ... , ak of I belong to the
same facet of I if and only if {.j(ai), ... ,j(ak)} is a flag of kn.
d) Show that G acts doubly-transitively on the set of I-dimensional vector
subspaces of k". Suppose that n ;?: 2 and let N 1 be the subgroup of G gen-
erated by the element that interchanges e 1 and e2 and leaves fixed the other
ei. Show that (G, G(l), N 1 ) is a Tits system whose Wey! group is of order 2.
e) Assume that k is commutative. Set
G' = SL(n, k), B' = G' n B, N' = G' n N and T' = N' n B' = T n G'.
Show that N' /T' can be identified with 6n. and that ( G', B', N', S) is a Tits
system (argue as in no. 2).
19) Let k be a commutative field of characteristic i 2 and let Q be the

quadratic form x 1 x 3 + :d on k:3. Show that the group SO(Q) acts doubly
transitively on the set of isotropic lines in k 3 . Compare the corresponding
Tits system (Exerc. 7) with that obtained in no. 2 for n = 2 (cf. Algebra.
Chap. IX, § 9, Exerc. 15).
20) We use the notation of no. 2, and assume that k is commutative. Consider
the following cases:
(Bt) n = 2l + 1 (with l;?: 1), k is of characteristic =I 2 and kn is equipped
with the quadratic form Q = X1Xn + · · · + x1X1+2 + xl+i;
(Ci) n = 2l (with l ;?: 1) and kn is equipped with the alternating form If>
such that <l>(ei, ei) = 0 for 1 ~ i ~ n, 1 ~ .i ~ n and i ~ .f, except when i ~ l,
.i = n + l - i, when <l>(ei, ej) = l;
(0 1) n = 2l (with l ;?: 2), k is of characteristic i 2 and kn is equipped
with the quadratic form Q = X1Xn + · · · + x1x1+1·
Denote by G 1 the special orthogonal group SO(Q) in cases (Bi) and (D1),
and the symplectic group Sp(<P) in case (Ci). Put
B1=G1nB, N1=G1nN and T1=G1nT=B1nN1.

a.) Show that the action of N 1 on the set of lines kei defines a homomorphism
from N 1 , with kernel T 1 , onto the subgroup W 1 of6n, allowing us to identify
W 1 with N i/T 1. Show that W 1 is the subgroup of 6 11 generated by
§ 2. EXERCISES 55
the Oj = SjSn- 1 for l ~ j <land by a1 = s1-1s1s1-i in case (B1);

the <T.i = s 1 sn-j for 1 ~ j <land by 01 = s1 in case (CL);
the 01 = iijSn-j for 1 ~ j <land by O'f = s1-is1s1-is1s1+ 1 s1 in case (D1).
b) Let Si be the set of O'j for 1 ~ j ~ l. Show that (G1,B 1,Ni.S 1 ) is a
Tits system (to prove that the subgroup H of G 1 generated by Bi and N 1
is the whole of G 1 , argue as in Algebra, Chap. II, § 10, no. 13), remarking
that H contains the lower triangular subgroup of G 1 and that, for any ~i E k
(2 ~ i ~ n), there exists a matrix b = (bi.i) E B 1 such that b11 = 1, b1; = ~i
for 2 ~ 't ~ n-1 and bin= ~n in case (C1), and bin= 0 in cases (BL) and (Di).
Then argue as in no. 2, introducing the subgroups G 1 .J = G 1 n (G1 , Gn-j)
for 1 ~ j < l and the subgroup Gu of elements of Ci leaving fixed
the ei for i f. l, l + 1, l + 2 and the subspace generated by e1, e1+ 1 and
e1 +2in case (Bi);
the ei for i =fl, l + 1 and the plane generated by e1 and e1 +1 in case (CL);
the e; for i < l - 1 or i > l + 2 and the two planes generated by e1_ 1 and
e1+ 1 and by e1 and e1+ 2 in case (DI).
Show that Gu can be identified, respectively, with GL(2, k), SL(2, k) or the
special orthogonal group of Exerc. 19).
*c) Show that the Coxeter graph of the group W 1 is of type (Bi), (C 1) or
(Di) in the three cases (Chap VI, § 4, no. 1),.
d) Show that, for any subset X of S 1 , the subgroup Gix consists of the
g E G 1 such that g(V;) = V; for all i such that Cl; tJ. X, except in case (D 1)
where the same assert.ion is valid provided that Yr- I denotes the subspace
generated by (:'l;···,e,._ 1 and e,.+ 1 . Deduce, as in Exerc. 18 b), that there
exists a bijection j from the building associated to (G 1 , B 1 ) to the set of
totally isotropic subspaces j. 0 in cases (Bi) and (Ci), and to the set of totally
isotropic subspaces of dimension f. 0 and f. r - 1 in case (Di). Show that
points a1, ... , ak ofl belong to the same facet if and only if {j (a 1 ), ... , .i (ad}
is a flag.
21) Let A be a discrete valuation ring (Commutative Algebra, Chap. VI, § 3,

no. 6), m its maximal ideal, 'Y a generator of m and K the field of fractions
of A. Let G be the group SL(2, K), I3 the subgroup of G consisting of the
'matrices ( ~ ~) such that a, b, d EA and c Em (with ad - be= 1) and N
rhe subgroup consisting of the matrices belonging to G and having only one
.non-zero entry in each row and column.
a) Show that T = B n N is normal in N and that W = N/T is an infinite

dihedral group generated by the classes s and s' of the matrices ( ~l ~)
and ( -~- l 6), respectively.
b) Show that (G,B,N,S) (with S = {s,s'}) is a Tits system.

c) Let H = SL(2, A) be the subgroup of G consisting of the matrices with
coefficients in A. Show that (H, B n H, N, {s}) is a Tits system. Compare with
Exerc. 18 e).
d) Let A be the completion of A and let G, B, N, T be the groups defined as
above but replacing A by A. Show that the injection of G into G defines an
isomorphism from the building I associated to (G, B) (Exerc. 10) to the build-
ing i associated to (G,B). Let (1,U) (resp. (I,ll)) be the stn.1.ctured building
ac;socia.ted to (C, B, N) (resp. (G, B, N)) (Exerc. 12): show that j(U) c ll, but
that j(U) i= ti if Ai= A (remark that the apartments fr (resp. il) correspond
bijectively to the conjugates of T (resp. T) by G (resp. G).
22) Let G be a group and B a subgroup of G.

a) Show that the following conditions are equivalent:
(i) B n gBg- 1 is of finite index in B for all g E G;
(ii) every double coset BgB with respect to B is a finite union of left cosets
with respect to B.
More precisely, show that, for all g E G, the index q9 of B n gBg- 1 in B is
equal to the number of left cosets with respect to B contained in the double
coset BgB. Show that q9h. ~ q_qqh for all g, h E G.
Assume from now on that conditions (i) and (ii) are satisfied and denote
by k a commutative ring. Fort E G/B (resp. t E B\G/B), denote by at the
map from G to k defined by at(g) = 0 if g ~ t and o. 1.(g) = 1 if g Et. Let L
(resp. H) be the k-module generated by the Ot fort E G/B (resp. t E B\G/B).
b) Show that there exists a unique linear form µ on L such that µ(at) = 1 for
all t E: G/B.
c) Let rp E Land 1/; EH. Show that, for all x E G, the map
e,c: y ........ rp(y)·t/.1(y- 1 x)
belongs to L and that the map rp * 1/; : x ,....... µ(Ox) belongs to L. Moreover,
if rp E H then <p * 110 E H. Show that the map (rp, 1/;) ,....... <p * 'lj; makes H into
an algebra over k, having as as its unit element, and makes L into a right
H-module.
The algebra H is called the Hecke algebra of G with respect to H and is
denoted by Hk(G, B).
§ 2. EXERCISES 57
d) Show that, fort, t' E B\G/B,
at * at' = I:
I"
m( t, t'; t")at",
where m(t, t'; t") is the number of cosets with respect to B contained in
t n gt'- l for all g E t".
e) Let G act on L by left translations. Show that the action of Hon L defines
an isomorphism from H to the commutant of the linear representation of G
on L thus obtained.
*f) Assume that G is finite and that the characteristic of k does not divide
the order of G. Show that Hk(G, B) is absolutely semi-simple over k (Algebra,
Chap VIII, § 7, no. 5) (use Maschkc's Theorem (Chap. V, Appendix) and
Prop. 3 of Algebra, Chap. VIII,§ 5, no. 1).*
g) Assume that G is a topological group and that B is a compact open
subgroup of G. Show that conditions (i) and (ii) are satisfied and that, when
k = R or C, the product i.p *'cf; is simply the convolution product relative
to the right Haar measure on G, normalised by the condition µ(B) = 1 (cf.
Integration, Chap. VIII, § 4, no. 5).
23) Let CW, S) be a Coxetcr system and k a commutative ring. Suppose that
we are given, for all s E S, two elements ,\, and µ 8 of k such that As = As'
and µ._, = µs' whenever s and s' are conjugate in W. Put E = k(W) and let
{ew} be the canonical basis of E.
a) Show that E has a unique algebra structure such that, for s E S and
wEW,
_ { e.,w if l(sw) > l(w)
e... ew - AsCw +µ 8 esw if l(sw) < l(w)
(introduce the endomorphism P s of E defined by the above formulas, where
e8 .ew is replaced by Ps(w), and the endomorphism Qs = jP,j- 1 , where j
denotes the automorphism of E defined by j (ew) = ew-,; show that P 5 Qt =
QtPs for s, t ES by remarking that the conditions l(swt) = l(w) and l(sw) =
l(wt) imply that sw = wt; then argue as in Exerc. 3 e)). The module E,
equipped with this algebra structure, will be denoted by Ek((A 5 ), (µ 8 )). Show
that E((O), (1)) is the group algebra k:[W] of W (Algebra, Chap. lll, § 2, no. 6).
b) Show that the family of generators (es)sES and the relations
e; = A5 e 8 +/.ls for s E S
(esetr = (et es)' for s, t E S such that st is of finite even order 2r
(e_.etr es = (e 1esY et for s, t E S such that st is of finite odd order 2r +1
:form a presentation of the algebra E (argue as in the proof of Th. 1 of no. 6
of§ 1).
58 COXETER GROUPS AND TITS SYSTEIVJS Ch. IV
24) Let G be a group and B a Tits subgroup of G (Exerc. 3, of which we use

the notation). Assume that. for alls ES, the double coset C(s) is the union
of a finite number q5 of left cosets with respect to B. We use the notation of
Exerc. 22 and put aw = ac(w) for all w E W.
a) Show that conditions (i) and (ii) of Exerc. 22 are satisfied. vVe can therefore
speak of the Hecke algebra Hk(G,B) (k being a commutative ring), of which
(aw)wEW is a basis.
b) Let s E S and w E W. Show that as* as = (qs - l)a; + q5 ; show that if
l(sw) > l(w) then as* aw= asw·
c) Show that the linear map from the algebra Ek((q 8 -1), (q 5 )) associated to
the Coxeter system (W, S) (Exerc. 23) to lh(G, B) that sends ew to aw for
all w E W is an isomorphism of algebras.
25) We use the notation of Exerc. 8. Assume in addition that, for all s E S,
the index q5 of B n gBg- 1 in B is finite for all g E BsB.
a) Show that the pair (G, B) satisfies conditions (i) and (ii) of Exerc. 22.
b) Show that, for all I E r' the map x f-> ')'X)'- l defines an automorphism (J
of the Hecke algebra Hk ( G. B) and that a depends only on the class w of I
in Q = I'/(I' n B).
c) Let k:[D] be the group algebra of Q and (ew) its canonical basis. Show that
the linear map j from k[f.?] '2h Hk(G, B) to Hk(G, B) defined by j(ew CZI a.swB)
= aswwB (in the notation of Exerc. 22), for w E Q and w E W, is bijective
and that
F 1(J(eu. (?) x)j(ew' Q9 y)) = Cww' 0 aw(:r)y
for w, w' E Jl and :r, y E Hk(G, B).
, 26) If E is an absolutely semi-simple algebra of finite rank over a com-

mutative field k, the n'Um.erical invariant of E is the sequence of integers
(n 1 , ... , nr) such that n 1 ~ · · · ~ nr > 0 and that k ®k Eis isomorphic, for
any algebraic closure k of k, to ni
Mn, (k).
Let V be an integral ring, K its field of fractions, --> TrE 0 ;k(xy) on E 0 is non-
degenerate. Show that E 0 and E 1 are absolutely semi-simple (cf. Algebra,
Chap IX, § 2, Exerc. 1).
b) Assume that E 0 and E 1 are absolutely semi-simple over k and K, respec-
tively. Show that E 0 and E 1 have the same numerical invariant (reduce to the
case where k and Kare algebraically closed); let (e;) be a basis of E over V and
§ 2. EXERCISES 59
let (Xi}__!)e indeterminates. Show that the characteristic polynomial of the cl-
ement I: X;ei of E1 ©K K[(X;)] (resp. Eo ®k k[(X;)]) is of the form P =
1.
P?j n
J
(resp. Q = fIQZ'•), where (n1 1 ••• ,nr) (resp. (m1 1 ••• ,ms)) is the numeri-
k
cal invariant of E 1 (resp. E 0 ), with dcg(Pj) = n.i (resp. dcg(Qk) = mk)·
Show that P; E V[(X;)] and that Q = <p(P). Deduce that there exist integers
c1 k ~ 0 such that
*27) Let (G, B, N, S) be a Tits system and let k be a commutative field. As-
sume that G is finite and that the characteristic of k divides neither the order
of G nor the order of the Wey! group W = N/(BnN). Show that the algebras
Hk(G, B) (Exerc. 22) and k[W] are absolutely semi-simple and have the same
numerical invariant, and hence are isomorphic when k is algebraically closed
(let Qs be the index of B n gBg- 1 in B for g E BsB, s E S. Consider the
algebra Eqxj((X(qs - 1)), (1 + X(qs - 1))) constructed as in Exerc. 23 from
the Coxeter system (W, S) and the polynomial ring k[X] and use Exerc. 26
a) and b), remarking that, by Maschke's theorem (Chap. V, Appendix), the
bilinear form Trk[WJ/d:ry) is non-degenerate) .•
28) Let G be a group, M a maximal subgroup of G and U a normal subgroup of

M, satisfying condition (R) of no. 7. Assume that G is equal to its commutator
subgroup, that it is generated by the union of the conjugates of U and that
the intersection of the conjugates of M reduces to the unit clement. Show
that G is simple. (Consider a normal subgroup N of G distinct from {l};
show that G = N.l\·1 and then that G = N.U.)
29) Let H be a simple non-abelian group. Let (J be an automorphism of

H of prime order p and let U be the semi-direct product of Z/pZ by H
corresponding to B. Show that, if() is not an inner automorphism, the only
normal subgroups of U are {l}, Hand U. Deduce that U is not soluble, but
that it satisfies condition (R) of no. 7. Apply this to show that the symmetric
group 6n satisfies condition (R).
CHAPTER V
Groups Generated by Reflections
§ 1. HYPERPLANES, CHAMBERS AND FACETS
In this section, E denotes a real affine space of finite dimension d and T

the space of translations of E (cf. Algebra, Chap. II, § 9). If a and b arc two
points of E, [ab] (resp. Jab[, resp. Jab]) will denote the closed segment (resp.
open segment, resp. segment open at a and closed at b) with extremities a, b.
The space T is provided with its unique separated topological vector space
topology, cf. Topological Vector Spaces, Chap. I, § 2, no. 3; it is isomorphic
to Rd. The space E is provided with the unique topology such that, for all
e E E, the map t ,....... e + t from T to E is a homeomorphism.
We denote by H a locally finite set of hyperplanes of E (General Topology,
Chap. I, § 1, no. 5).
1. NOTATIONS
Let H be a hyperplane of E. Recall that E - H has two connected components,

called the open half-spaces bounded by H. Their closures are called the closed
. half-spaces bounded by H. Let x, y E E. Then .7: and y are said to be strictly
on the same s·ide of H if they arc contained in the same open half-space
bounded by H, or equivalently, if the closed segment with extremities x and
y does not meet H; x and y are said to be on opposite sides of H if x belongs
to one of the open half-spaces bounded by H and y to the other. If x E E and
t E T, then x and t are said to be strictly on the same side of H if this is so
for x and h + t for all h E H.
Let A be a non-empty connected subset of E. For any hyperplane H of E
that does not meet A, DH (A) denotes the unique open half-space bounded
by H that contains A. If 1)1 is a set of hyperplanes of E, none of which meet
A, put
D\)'J(A) = n DH(A).
HE'.)1
(1)
If A consists of a single point a, we write DH (a) and D'l'l (a) instead of DH ( {a})
and D\)'J({a}).
62 GROUPS GENERATED BY REFLECTIONS Ch. V
2. FACETS
The set of points of E that do not belong to any hyperplane H of the set 5)
is open since f) is locally finite. More precisely, we have the following result:
PROPOSITION l. Let a be a point of E. There exists a connected open ne·igh-

bourhood of a that does not meet any hyperplane H that belongs to 5j and does
not pass through a. Moreover, there exist only finitely many hyperplanes th.at
belong to SJ and pass through a.
The set 'J1 of hyperplanes H such that H E f) and a ¢: H is locally finite
since it is contained in SJ. Hence, the set U of points of E that do not belong to
any of the hyperplanes of the set 'J1 is open. Since a E U, there is a connected
open neighbourhood of a contained in U. The remainder of the proposition
is clear.
Given two points.randy of E, denote by R{x,y} the relation
" For any hyperplane HE f), either x EH and y E H

or x and y are strictly on the same side of H."
Clearly, R is an equivalence relation on E.
DEFINITION l. A facet of E relative to f) is a.n equivalence class of the

eq1Livalence relation defined above.
PROPOSITION 2. The set of facets is locally .finite.

This is clear since SJ is locally finite.
Let F be a facet and o. a point of F. A hyperplane H E f) contains F if and
only if a E H; the set ~ of these hyperplanes is thus finite; their intersection
is an affine subspace L of E, which we shall call the affine support of F; the
dimension of L will be called the d·imension of F.
If 'J1 is the set of hyperplanes H E f) not containing F, then
F =Ln n DH (a).
HE'l'l
(2)
We shall prove that the closure of F is given by
F= Ln n DH(a).
HE':l'l
(3)
It is clear that the right-hand side contains the left-hand side. Conversely,
let x E Ln n
HE'l'l
DH (a). The open segment with extremities 0. and .7; is contained
in Land in each of the Dn(a) for H E 91, and hence in F. It follows that xis
in the closure of F, hence the formula.
PROPOSITION 3. Let F be a fa.eel and L its affine support.

§ 1. HYPERPLANES, CHAMBERS AND FACETS 63
(i) The set F is a convex open subset of the affine subspace L of E.

(ii) The closure of F is the union of F and facets of dimension strictly
smaller than that of F.
(iii) In the topological space L, the set F is the interior of its closure.
Since every open halfspace and every hyperplane are convex subsets of
E, formula (2) shows that F is the intersection of a family of convex subsets,
and hence is convex. On the other hand, let a be in F, and let Ube a convex
open neighbourhood of a in E that does not meet any of the hyperplanes in
the set 91 ofH E 5) such that a (j: H. For any HE 91, we thus have UC DH(a),
hence L n Uc F, so that Fis open in the topological space L.
Let b be a point of F - F, belonging to a facet F', and let 91' be the set of
hyperplanes H E 91 passing through b; put 91" = 91- 91'. For any H in 91",
we have bf-Hand b E DH(a), hence b E DH(a.) and DH(b) = DH(a); by the
definition of a facet. we thus have
F'=Ln n
HE'.l1'
Hn n
HE 1Yl"
DH(a), (4)
whereas (3) implies that
F = Ln n D1-1(a) n n
HE'.l1' HE'H"
DH(a), (5)
hence F' CF. We cannot have 91' = 0, for this would imply that F = F' by
(2) and (4), contrary to the hypothesis that b f- F and b E F'. The support
of F' is the set L' = L n n
H; we have a E L, but a f- H for H in 91', so
HE'.l1'
L' /:-Land finally dim L' <dim L, that is dim F' <dim F. This proves (ii).
Let H be in 91' and let D be the open half-space bounded by H and
distinct from DH(a); we have b E H n L, and it is immediate that D n Lis a
half-space of L bounded by the hyperplane H n L of L; consequently, every
neighbourhood of b in L meets D n L, and since D n L is disjoint from F by
(3), we see that the point b of F - F cannot be in the interior of F in the
topological space L. Since F is open in L we have (iii). Q.E.D.
COROLLARY. Let F and F' be two facets. ffF = F', the facets F and F' are
equal.
This follows from (iii).
PROPOSITION 4. Let F be a facet, and let L be an affine s·u.bspace of E

that is the intersection of hyperplanes belonging to f): denote by 91 the set
of hyperplanes H E f) that do not contain L. The following conditions are
~q·uivalent:
(i) There exists a .facet F' with S'upport L that meets F.
1. (ii) There exists a facet F' with s·u.pport L contained in F.
(iii) There exists a point x in L n F that does not belong to a.ny of the
hyperplanes of 91.
If these cond'itions are satisfied, L n D'Yl(F) is the unique facet with support
L contained in F.
(i) ==> (ii): Since Fis a union of facets (Prop. 3, (ii)), every facet that
meets F meets a facet contained in F, and so is equal to it.
(ii) ==> (iii): Every point x of F' satisfies (iii) since every hyperplane of
SJ containing x contains F', and hence L.
(iii) ==> (i): Let x be a point satisfying (iii) and let F' be the facet con-
taining x; it is clear that F' meets F. Let H E SJ; then x rt H if H E 91 and
clearly x E H if H rt 91; consequently the support of F' is the intersection of
the hyperplanes of SJ - 91, and is equal to L.
Finally, let F' be a facet with support L contained in F, and let x be a
point of F'; since no hyperplane of 91 c fJ passes through x, Prop. 1 shows
that there exists a convex open neighbourhood U of .T that does not meet
any hyperplane of 91. Since xis in the closure of F, Un F =f: 0; now 91 is the
set of hyperplanes H E fJ that do not contain F', and for all H in 91 we have
DH(x) = Dtt(U) = DH(U n F) = DH(F) and formula (2) implies that
F' = L n D'Y!(F). Q.E.D.
3. CHAMBERS
DEFINITION 2. A chamber of E relative to fJ (or simply a chamber ·if there

is no ambiguity regard·ing fJ) is a facet of E relative to .f) th.at is not contained
in any hyperplane belonging to fJ.
Let U be the open set in E consisting of the points that do not belong to
any hyperplane of fJ; since a hyperplane of .f) must contain any facet that it
meets, the chambers are the facets contained in U; every chamber is a convex
(hence connected) open subset of Eby Prop. 3, (i); since the chambers form a
partition of U, they are exactly the connected components of U. Every convex
subset A of U is connected, and thus contained in a chamber, which is unique
if A is non-empty. It is clear that the chambers are the facets with support
E, and Prop. 3, (iii) shows that every chamber is the interior of its closure.
Finally, let C be a chamber and A a non-empty subset of C; formulas (2) and
(3) imply that
C = n DH(A) = DfJ(A), -c = n -
DH(A) (6)
HEfJ HEfJ .
since DH(A) = Du(o.) for all a EA.
PROPOSITION 5. Let C be a non-empty S'ubset of E. Assu.me that there

exists a subset f)' of .f) with the following properties:
a) For any H E fJ', there exists an open half space Dr-1 bounded by H such
that c = n
HEfJ'
DH.
b) The set C does not meet any hyperplane belonging to f)- fj 1 .

Under these conditions, C is a chamber de.fined by fJ in E, and D 8
DH(C) for all HE f).
Properties a) and b) show that C is a convex subset of U; hence there is
a chamber C' with C C C'. Since C C DH, we have DH = DH ( C) for all H in
f) 1 , hence C =- Dw(C) :.J DSJ(C) since f) 1 Cf); we have Dn(CJ = C' by (6),
hence C :.J C'. Finally therefore, C = C'.
PROPOSITION 6. Every point of E is in the closure of al least one chamber.

If E reduces to a single point, this is cleai'. Otherwise, let a E E and let
Ht, ... , Hm be the hyperplanes of fJ containing a. Since fJ is locally finite,
there exists a neighbourhood V of a that does not meet any hyperplane of fJ
other than H 1 , ... , Hm. Lei D be a straight line passing through a. and not
contained in any of the Hi; if x E D, x fa, and x is sufficiently close to a,
the open segment Jax[ is contained in V and does not meet any of the H;.
Then ]ax[c U; since Ja:i:[ is connected, it is contained in a chamber C. hence
a EC.
PROPOSITION 7. Let L be an affine subspace of E and [2 a non-empt,y open
subset of L.
(i) There exists a point a in [2 f,fwt does not l>elong to any of the hyper-
planes of fJ that do not contain L.
(ii) If L is a hyperplane and L ¢ f), there e.xists a chamber that meets D.
(iii) ff Lis a hyperplane and LE f), there exists a point a ·in D that does
nol belong to any hyperplane H c,i L of fJ.
Denote by '.)1 the set of hyperplanes H with H E fJ and L <f. H, and by .£'.
the set of hyperplanes of the affine space L of the form L n H with H E '.)1, It
is clear that .£'. is a louilly finite set of hyperplanes in L, and Prop. 6 shows
that D meets a chamber I' defined by £ in L. If a is a point of r n rl, then
a~ H for all I-IE '.)1, hence (i).
Assume 11ow that L is a hyperplane; any hyperplane containing L is then
equal to it, so we may distinguish two cases:
a) L rt fj: thc11 '.)1 = f), and we have a rt H for all HE f); thus a belongs
to a chamber defined by fJ in E, hence (ii).
b) LE f): then '.)1 = fJ-{L}, hence (iii).
4. WALLS AND FACES
DEFINITION 3. Let C be a chamber of E. A face of C is a facet contained in

'the closure of C whose support is a hyperplane. A wall of C is a hyperplane
th.at is the support of a face of C.
Every wall of C belongs to fj. Prop. 4 shows that a hyperplane L E fJ is

a wall of C if and only if C f. DSJ-(L) (C). l\Ioreover, every wall of C is the
support of a single face of C.
PROPOSITION 8. Every hyperplane H belonging to fJ is the wall of at least

one chamber.
By Prop. 7, (iii), there exists a point a of H that does not belong to any
hyperplane H' -f. H of f); by Prop. 6, there exists a chamber C such that
a E C; Prop. 4 then shows that H is a wall of C.
PROPOSITION 9. Let C be a chamber and 9J1 the set of walls of C. Then

C = D!m(C) and every subset .C off) such that C = D.r.(C) contains 9J1. A
subset F of C is a facet if and only if it is a facet of E relative to the family
9J1.
a) Let .C be a subset of fJ such that C = D.r.(C). Consider a hyperplane L
belonging to f) but not to .C; let 01 be the set of hyperplanes H i= L belonging
to 5). Then .C C 01, hence C = DIJ1(C), and L does not meet D.:n(C). By
the implication (i) ==? (iii) in Prop. 4, the hyperplane L is not a wall of C.
Consequently, every wall of C belongs to ..C.
b) We assume that C = D£(C). Let H be a hyperplane belonging to .C
that is not a wall of C, and put ..C' = .C- {H}. By the implication (iii) ==? (i)
in Prop. 4, the convex set D.r.,(C) does not meet H, so D.i:,(C) c Du(C) and
C = D £.' ( C). If J is a finite subset of ..C that does not contain any wall of C,
we conclude by induction on the cardinal of J that C = D.r.-;v(C).
c) Let a be a point of C; clearly, CC D!m(a). Let a' be a point. of D!m(a);
since the closed segment [aa'] is compact, the set J' of hyperplanes H E fJ
that meet [aa'J is finite. Since a and a' are strictly on the same side of every
wall of C, no wall of C belongs to J'; by b), we have C = DSJ-;v(C). Since
a' E Ds:1 -~(a), we have a' E C. We have therefore proved that D9'Jl(a) C C,
which establishes the first part of the proposition.
d) To prove the last assertion of the proposition, it clearly suffices to show
that a subset F of C that is a facet of E relative to 9J1 is a facet of E relative
to f), or that every hyperplane H E fJ that meets F contains F. So let H be
a hyperplane that meets F but does not contain it. Since F is open in its
affine support, it is not completely on one side of H. It follows that C is not
completely on one side of H and hence that the hyperplane H does not belong
to f), which completes the proof.
Remarks. 1) It follows from formula (6) and Prop. 9 that the closure of a
chamber C is the intersection of the closed half-spaces that are bounded by
a wall of C and contain C.
2) Let F be a facet whose support is a hyperplane L; we shall show that F
is a face of two chambers. Let IJ1 be the set of hyperplanes H i= L belonging
to 5); put A = Dl)1 (F) and denote by D+ and o- the two open half-spaces
bounded by L. The set A is open and contains F C L, and since every
point of L is in the closure of D+ and o-, the sets c+ = A n D+ and
c- =Ano- arc non-empty; these are chambers. Moreover, the hyperplane
L meets D<n(F) = D'Yl(C+); Prop. 4 shows that L is a wall of c+ and that
F, which meets L n Drr1(F), is a face of C+; similarly, Fis a face of c-.

Finally, let C be a chamber of which F is a face, and suppose for example
that o+ = DL(C); by Prop. 4, the set D'Yl(C) meets F, and hence is equal to
D'Yl(F), and we have
c = D.!')(C) = DL(C) n D'Yl(C) = o+ n D'Yl(F) = c+.
5. INTERSECTING HYPERPLANES
Recall (Algebra, Chap. II, § 9·, no. 3) that two affine subspaces P and P' of E
are said to be para.IIel if there exists a vector t in T such that P' = t + P. It
is clear that the relation "P and P' are parallel" is an equivalence relation on
the set of affine subspaces of E.
Lemma 1. Any two non-parallel hyperplanes have non-empty ·intersection.

Let H and H' be two non-parallel hyperplanes, a E H and a.' E H'; there
exist two hyperplanes M and M' of the vector space T such that H = M + a
and H' = M' +a'; since Hand H' are not parallel, we have M ::f. M' and hence
T = M + M'; hence there exist u EM and u' EM' such that a' - a= u - u',
and the point v. + a = u' + a' belongs to H n H'.
Lemma 2. Let H and H' be two distinct hyperplanes of E, and f, f' two affine
functions on E such that H (resp. H') consists of the points a in E S'uch that
f(a) = 0 (resp. f'(a) = O). Finally, let 1 be a hyperplane of E. Assume that
one of the .following hypotheses is satisfied:
a) The hyperplanes H, H' and 1 are parallel.
b) The hyperplanes H and H' are not parallel, and H n H' c L.
Then there exist real numbers >., A', not both zero, such that L consists of
those points a E E at which the affine .function g = >..f + >.' .f' vanishes.
The lemma being trivial when L = H, we assume that there exists a point
a in 1 with a¢ H. Put>.= f'(a), >.' = -.f(a) and
g = >..f + >.'.f';
then >.' f. 0 since a ¢ H; moreover, since H ::f. H', there exists b E H such
that b ¢ H', so that f(b) = 0, f'(b) i= 0, and thus g(b) = - f(a)..f'(b) is non-
zero. The set 1 1 of points where the affine function g =/: 0 vanishes is then a
hyperplane of E; we have g(a) = 0, so a E 1 1 .
· a) Assume that H and H' are parallel: g and f both vanish at every point
of 11 n H, hence so does f' since >.' i= O; thus every point of 1 1 n H belongs
·to H'; but since H and H' are parallel and distinct, they are disjoint, so
11 n H = 0, and Lemma 1 shows that 1 1 is parallel to H. Since a E L and
a E 1 1 , we thus have L = 1 1 .
b) Assume that H and H' are not parallel: by Lemma l, there is a point c
in H n H'; we give E the vector space structure obtained by taking c as the
origin. Then H n H' is a vector subspace of E of codimension 2, and since

a tJ. H, the vector subspace M of E generated by HnH' and a is a hyperplane;
since HnH' c LnL 1 and a E LnL 1 , we have Mc LnL 1 , hence M = L = L 1 .
Q.E.D.
PROPOSITION 10. Let C be a chamber, let H and H' be two walls of C, and
let L be a hyperplane meeting DH(C) n Dw(C). Assume that H is d·istinct
.from H' and that one of the following conditions ·is satisfi,ed:
a) The hyperplanes I-1, H' and L are parallel. _
b) The hyperplanes H and H' are not parallel, and H n H' c L.
Then, L meets C.
Let b (resp. b') be a point of the face of C with support H (resp. H'); it is
immediate that every point of the segment [bb'] distinct from b and b' belongs
to C.
Introduce an affine function J that vanishes at every point of H and is such
that f(x) > 0 for x in DH(C); similarly, introduce an affine function f' having
an analogous property with respect to H'. By applying Lemma 2, we can find
numbers ,\ and ,\ 1 and an affine function g having the properties stated in the
lemma. We have (,\, ,\') i= (0, 0), and for every point :c of Ln D 8 (C) n Dw (C),
we have f(x) > 0, f'(x) > 0 and A.f(x) + >.'.f'(.'.C) = 0, so ,\)/ < 0. On the
other hand, we have g(b) = >.'.f'(b) and g(b') = ,\.f(b'), and since f(b') > 0,
J'(b) > 0, we have g(b)g(b') < 0. The points b and b' are thus strictly on
opposite sides of the hyperplane L, hence there exists a point c of L which
belongs to [bb'], and is distinct from b and b', hence that belongs to C.
6. EXAMPLES: SIMPLICIAL CONES AND SIMPLICES
a) Let a be a point of E, and (ei, ... , ed) a basis of T; every point of E can
thus be written uniquely in the form
(7)
with ~ 1 , ... , ~d being real numbers. Denote by e;the affine function on E
which, at any x E E, written in the form (7), takes the value ~i- Denote by
Hi the hyperplane formed by those x such that e~(x) = 0, and by .fJ the set
of hyperplanes H 1 , ..• , Hd. For every subset J of the set I= {l, 2, ... , d}, put
HJ = n
iEJ
Hi; for every sequence (€1: ... 'Ed) of nm~bers equal to 0, 1 or -1,
denote by F(£t, ... , Ed) the set of those x E E such that e~(:1:) is of sign ci
for all i in I (General Topology, Chap. 4, § 3, no. 2). It is immediate that the
facets defined by .lj in E are the sets F(c 1 , ... , cd) and that these sets are
pairwise distinct; the support of F(c 1 , ... , cd) is equal to HJ if J is the set of
i E I such that ci = O; in particular, the chambers are the sets of the form
F(c 1 , ... , cd) where each of the numbers£; is equal to 1 or -1.
The set C = F(l, ... , 1) formed by those x with e;(:r) > 0 for all i EI is
a chamber, called the open simpl·icial cone with vertex a defined by the basis
(e 1 , ••• , ed). Its closure consists of the points x such that e;(x) :;:;: 0 for all
i E I when d:;:;: l; otherwise, its closure is empty. For every subset J of I, let
CJ be the set of points :r of E such that e;(x) = 0 for i E J and e;(x) > 0 for
'i E I - .J. Then C.1 is a facet with support HJ, and it is an open simplicial cone
with vertex a in the affine space Hi; moreover, C = LJ CJ. In particular, the
Jc I
walls of C are the hyperplanes Hi for i E I, and the face of C contained in Hi
is equal to c{i}·
None of the sets H;, HJ, C, CJ and F(E 1 , ... , c:d) change if we replace the
basis (e 1, ... ,ed) by a basis (,\ 1 e 1 , ... ,,\ded) with A;> 0 for all i.
b) Suppose now that we are given an affinely free ::;ystem of points of E,
say (a 0 , a 1 , ..• , ad)· We know that every point of E can be written uniquely
in the form x = fo.a 0 + · · · + ~d.a.d, where (a, ... , ~d are real numbers with
fo+· · -+~d = 1 (Algebm, Chap. 2, § 9, no. 3). Define affine functions fo, ... , fd,
the function fi associating to the point x, written a::; above, the number ~i·
Denote by Hi the hyperplane of E defined by f; (x) = 0 and by Sj the set of
hyperplanes Ho,H 1 , ... ,Hd; finally, put I= {0,1, ... ,d}. The open simplex
with vertices ao, ... , ad is the set C of points x of E such that fi (:z:) > 0 for
all i E I; it is one of the chambers defined by Sj in E. The closure of C is the
set C of points x such that f;(x) :;:;: 0 for all i E I; it is the convex envelope
of the finite set {a.a, . .. , ad} and it is easy to see that the extreme points of
C are ao, ... , ad.
For any subset J of I distinct from I, put H.1 = n iEJ
H; and let C.J be the set
of points :r: of E such that f;(x) = 0 for i E J and f;(x) > 0 for i E I-J. The
set CJ is an open simplex in the affine space H1 with vertices the points ai
for i E 1-J; we have C 0 = C, C = U CJ and C.1 i C.1' if J =I .J'; moreover,
JI!
CJ is a facet with support H.1. In particular, the walls of C are H0 , ... , Hd
and c{i} is the face contained in H.;.
For any non-empty subset K of I, let BK be the set of points x of E such
that fi(:r:) > 0 for i E K and fi(x) < 0 for i E I- K. The sets BK are the
chambers defined by Sj in E and we have I31 = C. It is easy to see that C is
compact; on the other hand, if K is a subset of 1 distinct from I, of cardinal
p, the chamber BK contains the sequence of points x,.. defined for n :;:;: 2 by
for i EK
f;(Xn) = { (1 - pn)/(d + 1- p) for i E I- K
showing that BK is not relatively compact.

§ 2. REFLECTIONS
In this paragraph, K denotes a commutative field, assumed not to be a char-

acteristic 2 from no. 2 onwards. We denote by V a vector space over K.
1. PSEUDO-REFLECTIONS
DEFINITION 1. An endomorphism s of the vector space V is said to be a

pseudo-reflection if 1 - s is of rank 1.
Let s be a pseudo-reflection in V, and let D be the image of 1 - s. By

definition, D is of dimension l; thus, given a i= 0 in D, there exists a non-zero
linear form a* on V such that x - s(x) = (x, a.*).a for all x EV.
Conversely, given a i= 0 in V and a linear form a* :J 0 on V, the formula
Sa,a•(x) = x - (:1:,a*).a (x EV) (1)
defines a pseudo-reflection sa,o.·; the image of 1 - sa,a• is generated by a and

the kernel of 1 - sa,a• is the hyperplane of V consisting of those x such that
(x, a*) = 0. If V* is the dual of V, it is immediate that the transpose Sa• ,a of
sa,a• is the pseudo-reflection of V* given by the formula
Sa• ,a ( X·*) = X* - (X• , a ) .a • (x* EV*). (2)

If o. is a non-zero vector, a pseudo-reflection w'ith vector a is any pseudo-
refiection s such that a belongs to the image of 1 - s. The hyperplane of
a pseudo-re.fl,ection s is the kernel of 1 - s, the set of vectors :c such that
s(x) = x.
PROPOSITION l. Let G be a group and p an irreducible li:near representa-

tion of G on a vector space V; assume that there ex1:sts an element g of G
S'uch that p(g) is a pseudo-reflection.
(i) Every endomorph-ism of V commuting with p(G) is a homothety, and
p is absolutely irreducible.
(ii) Assume that V is finite dimensional. Let B be a non-zero bilinear form
on V invariant under p(G). Then B is non-degenerate, edher symmetric
or skew-symmetric, and every bil-inear form on V in;uariant under p(G) is
proportional to B.
Let v. be an endomorphism of V commuting with p( G). Let g be an element
of G such that p(g) is a pseudo-reflection and let D be the image of 1 - p(g).
Since D is of dimension 1 and u(D) c D, there exists a: in K such that u - a:. I
vanishes on D; the kernel N of u- a:. I is then a vector subspace of V invariant
under p(G) and is non-zero as it contains D; since p is irreducible, N = V
and u = o:.l. The second part of (i) follows from the first by Algebra, Chap.
VIII, § 13, no. 4, Cor. of Prop. 5.
§ 2. REFLECTIONS 71
Let N (resp. N') be the subspace of V consisting of those x such that

B(.r, y) = 0 (resp. I3(y,.r) = 0) for ally in V; since Bis invariant under p(G),
the subspaces N and N' of V are stable under p(G) and distinct from V since
B I= 0. Since p is irreducible, N = N' = 0 and B is non-degenerate.
Since V is finite dimensional, every bilinear form on V i::; given by the
formula
B'(x,y) = B(u(x),y) (3)
for some endomorphism tt of V. If B' is invariant under p(G), the endomor-
phism u commutes with p(G). Indeed, let :r, y be in V and let g be in G; since
B and B' are invariant under p(G), we have
B(u(p(g)(x)),y) = B'(p(g)(x),y) = B'(x,p(g- 1 )(y))

= B(u(x),p(g- 1 )(y)) = B(p(g)(u(.r)),y),
hence u(p(g)(x)) = p(g)(u.(x)) since Bis non-degenerate. By (i), there exist::;
a in K with u = o.1, so B' = a.B.
In particular, we can apply this to the bilinear form B'(x,y) = B(y,x);
then B(y,x) = a.B(x, y) = a 2 .B(y, x) for all x, yin V, and since B is non-
zero we have a 2 = 1, hence a = 1 or a = -1. Thus, B is either symmetric or
skew-symmetric.
2. REFLECTIONS
Recall that from now on, unless stated otherwise, the field K is assumed to
be of characteristic different from 2. A reflection in V is a pseudo-reflection
s such that s 2 = l; ifs is a reflection, we denote by v;the kernel of s - 1
and by v; that of s + 1.
PROPOSITION 2. Lets be an endomorphism. of V.

(i) Ifs is a reflection, V is the direct sum of the hyperplane v.; o.nd the
line v-;.
(ii) Conversely, assume that V is the direct sum. of a hyperplane H and a
l·ine D sv.ch that s(x) = x and s(y) = -y for.TE H and y ED. Thens is a
reflection and H = v;, D = V-;. Finally, D is the image of 1 - s.
(i) Ifs is a reflection, v; is a hyperplane. If x belongs to v; n v-;, then
x = s(x) = -:z:, so x = 0 since K is of characteristic f=. 2. Finally, for x in V,
the vector x' = s(x)+x (resp. x" = s(:i:)- .r) belongs to v; (resp. V;) since
s 2 = 1, and 2x = x' - x". Thus V is the direct sum of Vt and V_;, and V;
is necessarily of dimension 1 since v; is a hyperplane.
(ii) Under the stated hypotheses, every element of V can be written
11niquely in the form v = x+y with x EH and y E D and we have s(v) = x-y;
assertion (ii) follows immediately from this.
COROLLARY. If V is finite dimenS'ional, every reflection is of determ:inant

-1.
Let s be a reflection in V. Prop. 2, (i) shows that there exists a basis
(e1, ... ,en) of V such that s(ei) = C1, ... ,s(en-1) = en-1 and s(e 11 ) =-en,
hence det(s) = -1.
PROPOSITION 3. Let s be a reflection 'in V.

(i) A subspace V' of Vis stable under s if and only ifV;, C V' or V' c v;.
(ii) An endomorphism ·u of V commutes withs if and only 4v; and V;
are stable ·under u.
(i) If V' C Vt, then s(x) = x for all x in V', so s(V') C V'. As:;ume that
V; c V'; then, for any x in V', s(x) - x EV; c V', hence s(x) EV'; thus
s(V') c V'. Conversely, assume that s(V') CV'; if V' <t v-:;, there exists x
in V' with s(x) i- x; the non-zero vector a= s(x) - x belongs to the line V;,, ·
and hence generates this space; since a E V', we have V,:; c V'.
(ii) Assume first that ll commutes withs. If xis a vector such that s(x) =
c:.x (where E: = ±1), then s(u(x)) = u(s(x)) = c:.u(:r:), so v:
and V; are
stable under u. Conversely, assume that Vt and v; are stable under u; it is
clear that us - s·u vanishes on V.t and on v::;, and since V is the direct sum
of Vi and V;,, we have 'US - SU= Q.
COROLLARY. Two distinct reflections s and ·u commute if and only if

v;, c v; and v;, c v,:;,
If V.:; c V~ and V;,, c Vt, Prop. 3, (i) shows that V~ and V:;; are stable
under s, hence .s·u =us by Prop. 3, (ii).
Conversely, if SU = us, the subspace v; is stable under u by (ii); by (i),
there are two possible cases:
a) We have V;;, C V;: since they are both of dimension I, these spaces
are therefore equal, hence v:; cj_ V~; since V;!'" is stable, under s, we have
Vt. c Vt and so these two hyperplanes are equal But then s = u, contrary
to our assumptions.
b) We have V; c V~: the image of 1- sis thus contained in the kernel of
1-u, so (1-u).(1-s) = 0. Since sand u commute, we have (l-s).(l-1i) = 0,
in other words v;, c v;,
Remark. Let a =/= 0 be in V and let a* be a non-zero linear form on V: it

follows from formula ( 1) that
s;,a.(:c) =;I;+ ((a,a*) - 2)(x,a*).a.
and hence that sa,n.• is a reffoction if and only ~f (a, a.*) = 2. In this case, we
have Sa,a.• (a) = -a.
3. ORTHOGONAL REFLECTIONS
Assume that V is finite dimensional. Let B be a non-degenerate bilinear

form on V. By Algebra, Chap. IX, § 6, no. 3, Prop. 4, B is invariant under a
reflections in V if and only if the subspaces v; and V; of V are orthogonal
with respect to B; they are then non-isotropic. Moreover, for any non-isotropic
hyperplane H in V, there is a unique reflection s that preserves B and induces
the identity on H; this is the symmetry with respect to H, cf. Algebra, Chap.
IX, §6, no. 3. If a is a non-zero vector orthogonal to H, we have B(a, a) -f- 0
and the reHection s is given by the formula
B(x,a)
s (x ) = :r - 2---.o. for any x EV, (4)
B(a,a)
by Algebra, Chap. IX, § 6, no. 4, formula (6). The reflection s is also called
the orthogonal reflection wit.Ji. respect to H.
PROPOSITION 4. Assv.me that V is .finite dimensional. Let B be a non.-
degenerate symmetric bilinear form. on V, X a subspace of V and x 0 the
orthogonal complement of X with respect to B; .finally, lets be the orthogonal
reflect-ion with respect to a non-isotropic hyperplane H of V. The .following
conditions arc equivalent:
(i) X is stable under s.
(ii) x0 is stable under s.
(iii) H contains X or x 0 .
We have Vi = H, and by what we have said, v.; is the orthogonal
complement H 0 of H with respect to B. By Prop. 3, Xis stable under s if and
only if X C H or H° C X; but the relation H 0 c X is equivalent to x° C H by
Algebra, Chap. IX, § 1, no. 6, Cor. 1 to Prop. 4. This proves the equivalence
of (i) and (iii); that of (ii) and (iii) follows by interchanging the roles of X
and X 0 , since (X 0 ) 0 = X.
4. ORTHOGONAL REFLECTIONS IN A EUCLIDEAN

AFFINE SPACE
We retain the notation of the preceding number, and let E be an affine space
of which V is the space of translations. Giving the form B on V provides E
·. with the structure of a euclidean space (Algebra, Chap. IX, § 6, no. 6).
Let H be a non-isotropic hyperplane of E. The symmetry with respect
to H (Algebra, Chap. IX, § 6, no. 6) is also called the orthogonal reflection
with respect to H; we often denote it by SH. We have s~ = 1 and SH is the
unique displacement (loc. cit., Def. 3) of E, distinct from the identity and
leaving fixed the elements of H. The automorphism of V associated to SH
·is the orthogonal reflection with respect to the direction of H (which is a
~:non-isotropic hyperplane of V).
Every x in E can be written uniquely in the form .7: = h + v, with h E H

and v EV orthogonal to H; we have
SH ( h + V) = h - V.
PROPOSITION 5. Let H and H' be two parallel, non-isotropic hyperplanes
of E. There exists a unique vector v E V orthogonal to H and such that
H' = H + v. The displacement SH'SH is the translation by the vector 2v.
The existence and uniqueness of v are immediate. The automorphism of
V associated to sw SH is the identity; thus SH' SI-I is a translation. On the
other hand, let a EH'; then a - v EH and
SJ-I'SH(a - v) = SH'(a - v) =a+ v =(a - v) + 2v,

showing that swsH is the translation by the vector 2v.
COROLLARY. Let H and H' be two distinct, parallel, non-isotmpic hyper-

planes. If K is of characteristic zero (resp. p > 0, with p cf:- 2), the group
of displacements of E generated by SH and sw is an ·infinite dihedral group
(resp. a dihedral group of order 2p ).
Indeed, by Prop. 2 of Chap. IV, § 1, no. 2, it suffices to show that swsH
is of infinite order (resp. order 2p), which is dear.
Remark. We retain the notation of Prop. 5 and assume in addition that

K = R. Put s = SH and s' = sw. Let Hn be the hyperplane H + n.v
and let Cn be the set of points of E of the form a+ E.v with a E H and
n < E < n + 1. The C,,. are connected open sets forming a partition of
E-U Hn. They are therefore the chambers defined by the system fl =
n
(Hn )nEZ in E. The translation (s's) n transforms the chamber c = Co into
the chamber C2n, and since s(Co) = C_ 1 , we have (s' s)ns(C) = C 2,,_ 1 . It
follows that the dihedral group W generated by 8 and 8 1 permutes the cham-
bers Cn simply-transitively. Moreover, as we shall now show, if the chambers
C and w(C) are on opposite .<>ides of H (for w E W), we have l(sw) = l(w)- 1
(the lengths being taken with respect to S = {s,s'} (Chap. IV,§ 1, no. 1)).
Indeed, we then have w(C) = Cn for some n < 0. If n = - 2k, then w = (ss')k
and sw = (s' .<>)k- 1 s 1 , so l(w) = 2k and l(sw) = 2k - 1 (Chap. IV, § 1, no. 2,
Remark). Ifn = -2k- l, then w = (ss')"'s and sw = (s's)k, so l(w) = 2k+l
and l(sw) = 2k.
5. COMPLEMENTS ON PLANE ROTATIONS
In this no., V denotes a real vector space of dimension 2, provided with a

scalar product (that is, a non-degenerate, positive, symmetric bilinear form)
and an orientation. The measures of angles will be taken with respect to the
base 2rr; the principal measure of an angle between half-lines (resp. lines)
is thus a real number 8 such that 0 :::;; () < 27r (resp. 0 :::;; () < 7r) (General
Topology, Chap. VIII,§ 2, no. 3 and no. 6). By abuse of language, for any real
number 8, WC shall use 8 to denote an angle Whose measure is{) and denote
by Po the rotation with angle() (Algebra, Chap. IX, § 10, no. 3).
PROPOSITION 6. Let s be the orthogonal reflection with respect to a line D

of V. If Ll and Ll' are two half-lines starting at the origin (resp. two lines
passing through the origin) of V, we have
--- ---
(s(Ll), s(Ll')) :== - (Ll, Ll') (mod. 27r) (resp. (mod. 7r)).
Let u be a rotation transforming Ll into Ll'. Since sv. is an orthogonal

transformation of V of determinant -1, it is a reflection and thus (su) 2 = 1.
Consequently, u- 1 = sus- 1 transforms s(Ll) into s(Ll'), hence the proposi-
tion.
COROL~RY. Let D and D' be two lines of V and let() be a measure of the
angle (D, D'). Then sn,sn =PW·
We know that sn' so is a rotation since it is of determinant 1. Let L'i and
Ll' be two half-lines starting at the origin carried by D and D'. We have
(L'i,s;;n(L'i)) = (L'i,so,(Ll)) = (Ll,L'i') + (L'i',;;-(L'i))

---
= (Ll,L'i') + (so,(L'i'),so'(Ll))
- L'i) = 2(L'i,-Ll') (mod. 2r.),
= (L'i,---L'i') - (L'i',
hence the corollary.
Now let L'i and L'i' be two half-lines of V such that
L'i =I Ll' and Ll =I -Ll',
and let s and s' be the orthogonal reflections with respect to the lines ~nd
D' containing Ll and Ll'. Let 8 be the principal measure of the angle (D, D').
If 8 E 7TQ, denote by m the smallest integer ~ 1 such that mB E 7rZ. If
() ¢ 7rQ, put m = oo. Let W be the group generated bys and s'.
PROPOSITION 7. The gmu.p W is dihedral (Chap. IV, § 1, no. 2) of order

2m. It consists of the rotations Pzno and the products Pznos for n E Z. The
transforms of D and D' by the elements of W are the transforms of D by the
rotations Pno for n E Z.
The Corollary to Prop. 6 shows that s's is of order m, which gives the
first assertion. The elements of \V are thus of the form (s's)n = Pzno or
(s's)" s = Pznos; the last assertion follows from this, since D' = Po (D).
COROLLARY. Let C be the open angular sector formed by the union of the
·open half-lines L'i 1 starting at the origin and such that 0 < (L'i-;Lii) < e. Then
no transform of D or D' by an element of W meets C if and only if m is

finite and
e = 7r/m.
If m = oo, the image of the set ne (n E Z) is dense in R/27rZ (General
Topology, Chap. VII, § 1, Prop. 11); the union of the transforms of
D by the
elements of w is thus dense in v and meets c. lf m is finite and if e
= kn /m
with 1 < k < m, the integers k and m being re~ively prime, there exists an
=
integer h such that hk 1 mod. m; then (D,pM(D))-:= 7r/m (mod. n), and
PM(D) meets C. This shows that the condition is necessary. The converse is
immediate.
Remark. Assume that m is finite and that e = n /m. If n E Z, let Cn be the

union of the open half-lines "1 1 starting at 0 such that
ne < ("1, L1i) < (n + l)e.

The Cn for -m ~ n -;: : m are connected open subsets forming a partition of
E- U Dn (where Dn = Pno(D)). These are therefore the cha.rnbers determined
n
in Eby the system of m lines Dn (1 ~ n ~ m). We have C2k = P2ko(C) and
C2k-J = P2kos(C). Moreover, Cn = C if and only if n E 2mZ. Consequently,
the gro'U.p W perm1ltes the chambers Cn simply-trans'itively.
We show finally that, ·i;f w E W is such that the chambers C and w(C) a.re
on opposite sides of the line D, then l(sw) = l(w)-1 (the lengths being taken
with respect to S = {s,s'}). Indeed, the assumption implies that w(C) = C,.
with -m ~ n < 0. If n = -2k, we have w = (ss')k and sw = s'(ss')"'- 1 , so
l(w) = 2k and l(sw) = 2k-1 (Chap. IV,§ 1, no. 2, Remark). If n = -2k + 1,
we have
w = (ss')"'- 1s and sw = (s' s)k-l,
hence l(w) = 2k - l and l(sw) = 2k - 2. Q.E.D.
§ 3. GROUPS OF DISPLACEMENTS GENERATED

BY REFLECTIONS
In this paragraph, we denote by E a real affine space of finite dimension d,

and by T the space of translations of E. We assume that T is provided with a
scalar product (that is, a non-degenerate, positive, symmetric bilinear form),
denoted by (tit'). Fort ET, put lltll= (tit) 112. The function d(.7:,y) = llx-yll
is a. distance on E, which defines the topology of E (§ 1).
We denote by fJ a set of hyperplanes of E and by W the group of dis-
placements of the euclidean space E generated by the orthogonal reflections
§3. GROUPS OF DISPLACEMENTS 77
SH with respect to the hyperplanes H E fJ (§ 2, no. 4). We assume that the

following conditions are satisfied:
(Dl) For any w E W and any HE SJ, the hyperplane w(H) belongs to f);
(D2) The group W, provided with the discrete topology, acts properly on E.
Since E is locally compact, it follows from the Remark of § 4, no. 5 of
General Topology, Chap. IIL that condition (02) is equivalent to the following
condition:
(D'2) For any two compact subsets K and L of E, the set of w E W such that
w(K) meets L is .finite.
1. PRELIMINARY RESULTS
Lemma 1. The set of hyperplanes SJ is locally finite.

Indeed, let K be a compact subset of E. If a hyperplane H E fJ meets K,
the set s1-1(K) also meets K, since every point of Kn H is fixed by SH. The
set of H E SJ meeting K is thus finite by (D'2).
We can thus apply to E and SJ the definitions and results of§ 1. We shall
call the chambers, facets, walls, etc. defined in E by SJ simply the chambers,
facets, walls, etc. relative to W. Any displacement w E W permutes the
chambers, facets, walls, etc.
Lemma 2. Let C be a chamber.

(i) For any :c E E, there exists an element w E W sv.ch that w(:i:) EC.
(ii) For any chamber C', there is o:n. element w E W such that w(C') = C.
(iii) The gronp W is generated by the set of orthogonal reflections with
respect to the walls of C.
Let 9J1 be the set of walls of C and let W 9Jl be the subgroup of W generated
by the reftections with respect to the walls of C.
(i) Let x EE and let J be the orbit of x under the group W!Dt· It suffices
to prove that J meets C.
Let a be a point of C; there is a closed ball B with centre a meeting J;
since B is compact, property (D'2) of no. 1 shows that B n J is finite. Hence,
there exists a point y of J such that
d(a, y) ~ d(a, y') for all y' in J. (1)
We shall prove that y E C. For this, it suffices to show that if H is a wall of C

then y E Du(C) (cf. § 1, no. 4, Prop. 9). Since SH E W<JR, we have St1(Y) E J
and so (fig. 1)
d(a, y) 2 ~ d(a. SH(y)) 2 (2)
by (1). There exist b E H and two vectors t and u such that a = b + t and
Y = b + u, the vector u being orthogonal to H; then sH(Y) = b - u, and (2)
is equivalent to (t - ult - u) ~ (t +ult+ u), or to (tlu) ~ 0. This inequality

implies that y E DH(C).
Fig. 1
(ii) Let C' be a chamber and a' E C'. By what we have proved, there
exists w E Wrm such that w- 1 (a 1 ) EC; hence, the chamber C' meets w(C);
since w(C) is the union of w(C) and facets with empty interior (§ 1, no. 2,
Prop. 3), we have C' = w(C).
(iii) We have to prove that W = Wm and for this it suffices to prove
that sw E Wm for all H' E Sj. Now H' is a wall of at least one chamber
C' (§ 1, no. 4, Prop. 8); we have seen that there exists w E W!J.J1 such that
C' = w(C); consequently, there exists a wall H of C such that H' = w(H),
hence SH' = W.SH.w- 1 E w<JJI.
2. RELATION WITH COXETER SYSTEMS
THEOREM 1. Let C be a chamber and let S be the set of re.fiections with

respect to the walls of C.
(i) The pair (W, S) is a Co:reter system.
(ii) Let w E W and let H be a wall of C. The relation l(si-I'w) > l(w)
implies that the chambers C and w(C) are on the same side of H.
(iii) For any chamber C', there e1:ists a nniq1ie element w E W s·uch that
w(C) = C'.
(iv) The set of hyperplanes H such that SH E W is equal to fl.
Every element of S is of order 2 and Lemma 2 shows that S generates W.
For any wall I-I of C, denote by PH the set of elements w E 'vV such that the
chambers C and w(C) (which do not meet H) are on the same side of H. We
shall verify conditions (A'), (B') and (C) of Chap. IV, § 1, no. 7.
(A') 1 E Ptt: Trivial.
(B') Ptt and SH.PH are disjoint: Indeed, w(C) and s 8 w(C) are on opposite
sides of H, so if w(C) is on the same side of H as C, it is not on the same side
as s1-1w(C).
(C) Let w E PH and let H' be a wall of C such that wsH' tf. PH; th.en
wsw = Sf-IW: By assumption, w(C) is on the same side of H as C and
wsH'(C) is on the other side. Thus, wsH'(C) and w(C) are on opposite sides
§ 3. GROUPS OF DISPLACEMENTS 79
of H; hence, the chambers sw (C) and C are on opposite sides of the hy-
perplane w- 1 (H). Let a be a point of the face of C with support H'. The
point a= SH' (n) is in the closure of the two chambers C and sH' (C) that are
contained respectively in the two open half~spaces bounded by w- 1 (H); thus,
a E w- 1 (H), so H' = w- 1 (H). From this we deduce that sH' = w- 1 s1-1w,
hence WS[-!I = SH'W·
Assertions (i) and (ii) follow from this and Prop. 6 of Chap. IV,§ 1, no. 7.
Moreover, we have (loc. cit., condition (A))
n
HE9'Jl
Ptt = {l}. (3)
Lemma 2 shows that W acts transitively on the set of chambers. Moreover,

if w E W is such that w(C) = C, then w E PH for every wall H of C, hence
w = 1 by (3). This proves (iii).
Fiually, let H be a hyperplane such that SH E W. If H did not belong to
f), there would be at least one chamber C' meeting H (§1, no. 3, Prop. 7).
Every point of H n C' is invariant under SH, and thus belongs to the chambers
C' and sH(C'); thus, C' = Stt(C'), which contradicts (iii) since SH =F 1.
COROLLARY. Let E be a set of re.fl,ections generating W. Then every rc-

flect·i.on belong'ing to W is cor~jv.gate to an element of E.
Let f)' be the set of hyperplanes of the form w(H) with w E Wand HE f)
such that SH E E. Since W is generated by the family (sH)HEn' and since
f)' is stable under W, we can apply all the results of this no. to f)' instead
off); but Th. 1, (iv) shows that every reflection in Wis of the form SH with
H E f)', hence the corollary.
3. FUNDAMENTAL DOMAIN, STABILISERS
Recall (Integration, Chap. VII, § 2, no. 10, Def. 2) that a subset D of E is

called a fundamental doma'iri for the group W if every orbit of W in E meets
Din exactly one point. This is equivalent to the following pair of conditions:
a) For every x EE, there exists w E W sv.ch that w(x) ED.
b) If x, y E D and w E W are snch tha.t y = w(:c), then y = x (even
though we may have w =f. 1).
\Ne are going to prove the following three statements:
THEOREM 2. For any chamber C, the closure C of C is a fundamental

domain for the action of W on E.
PROPOSITION 1. Let F be a facet and C a. chamber such that F C C. Let
w E W. The following conditions are eqv.ivalent:
(i) w(F) meets F;
(ii) w(F) = F;
-(iii). w(F) = F;
(iv) w fixes at least one point of F;

(v) w fixes every point o.f F;
(vi) w fixes every point of F;
(vii) w belongs to the subgrov.p of W generated by the reflections with respect
to the walls of C conta·ining F.
For every subset A of E, denote by W(A) the subgroup of W con:;isting
of the elements that fix every point of A.
PROPOSITION 2. Let A be a non-empty subset of E, let fJA be the set of

hyperplanes H E fJ contain·ing A, let A' be the intersection of the H E fJA,
and let F be a facet of E open in A' (§ 1, no. 3, Prop. 7). Then W(A) =
W(A') = W(F), and W(A) is generated by the reflect·ions with respect to the
hyperplanes in fJA-
We first prove the following assertion:
(I) Let C be a chamber, let x and y be two points o.f C and let w E W be
siich that w( x) = y. Then x = y and w belongs to the subgroup W % where
!J1 is the set of walls of C contai.ning x.
We argue by induction on the length q of w (relative to the set S of
reflections with respect to the walls of C), the case q = 0 being obvious. If
q ~ 1, there exist a wall H of C and an element w' E W such that w = StJ'W 1
and l(w') = q - l. Since l(sttw) < l(w), the chambers C and w(C) are on
opposite sides of H by Th. 1 of no. 2. Thus, C n w(C) c H, soy E H. Thus
y = w'(x) and the induction hypothesis implies that :r = y aud w' E W<n.
Since y E H, it follows that H E !J'l, and hence that w = SttW 1 E w<Jl,
completing the proof of (I).
Proof of Theorem 2: this follows from (I) and Lemma 2.
Proof of Proposition 1: we know that two distinct facets are disjoint and
have distinct closures(§ 1, no. 2, Cor. of Prop. 3). The equivalence of (i), (ii)
and (iii) follows. On the other hand, it is clear that
(vii)===> (vi)===> (v) ===>(iv)===> (i)
and assertion (I) shows that (i) ===> (vii).
Proof of Proposition 2: let A" be the affine subspace of E generated by
A. Clearly, W(A) = W(A"). By Prop. 7 of§ 1, no. 3, there exists a point
x E A" that does not belong to any hyperplane H E fJ- fJA· Let F x be the
facet containing :z:: it is open in A" and Prop. 1 shows that
W(Fx) c W(A') c W(A) = W(A") c W({x}) = W(Fx),
hence W(A) = W(A') = W(Fx)· Replacing A by F, we also have
W(A) = W(F),
hence the proposition.
§ 3. GROUPS OF DISPLACE!VIENTS 81
Remarks. l) It follows from Th. 2 that, if C is a chamber and F a facet,

there ex·ists a uniqite facet F' conta·ined in C that is transformed into F by an
element of W.
2) It follows from Props. 1 and 2 that, for any non-empty subset A of E,
there exists a point a EE such that W(A) = W({a}); moreover, the group
W(A) is a Coxeter group (Th. 1).
3) Let C be a chamber of E and S the set of reflections with respect to
the walls of C. Let w E W and let (s 1 , ... , sq) be a reduced decomposition
of w with i'espect to S. If x E C is fixed by w, then si(x) = x for all j: this
follows from Prop. 1 above and Cor. 1 of Prop. 7 of Chap. IV, § 1.
4. COXETER MATRIX AND COXETER GRAPH OF W
Let C be a chamber, S = S(C) the set of orthogonal reflections with respect

to the walls of C and M = (m( s, s')) the Coxeter matrix of the Coxeter
system (W, S) (Chap. IV, § 1, no. 9): recall that m(s, s') is the order (finite
or infinite) of the element ss' of W (for s, s' E S). If C' is another chamber,
the unique element w E W such that w(C) = C' defines a bijection
S >--> f(s) =WSW-I
from S to S' = S(C'), and we have m(f(s), .f(s')) = m(s, s'). It follows that,
if W acts on the set X of pairs (C, s), where C is a chamber and s E S(C),
by w.(C, s) = (w(C), wsw- 1 ), each orbit i of W in X meets each of the
sets {C} x S(C) in exactly one point, which we denote by (C, si(C)). Thus,
if I is the set of orbits and i,j E I, the number mi.i = m(si(C), sj(C)) is
independent of the choice of chamber C. The matrix M(W) = (mij)i,JEI is a
Coxetcr matrix called the Coxetcr matri.1; of W. The Coxeter graph associated
to M(W) (Chap. IV,§ 1.. no. 9) is called the Coxeter graph of W.
Let C be a chamber. For any i E I, denote by Hi (C) the wall of C such
that si(C) is the reflection with respect to Hi(C) and by ei(C) the unit vector
orthogonal to H;(C) on the same side of H;.(C) as C. The map if--> Hi(C) is
called the canonically indexed family of walls of C.
PROPOSITION 3. Let C be a chamber and let i, .i E I with i j j. Put

s; = si(C), H; = H;(C), e; = ei(C) and define S.J, HJ and ej sim'ilarly.
(i) If H; and Hj are parallel, then m.;j = oo and ei = -er
(ii) If H; and H1 are not parallel, then m;j is .finite and
(eile.i) = - cos(7r/m;J)· (4)
(iii) We have (eileJ) ~ 0.

If Hi and Hj are parallel, sis.i is a translation (§ 2, no. 4, Prop. 5), so
mij = oo. Moreover, either e; = ej or ei = -eJ· Now, there exists a point
. a (resp. a') in the closure of C that belongs to H; (resp. H.i) but not to Hi
(resp. H;). Then (a' - ale;)> 0 and (a - a'le1) > 0, which excludes the case
e.; =ej and proves (i).
Assume now that Hi and Hj are not parallel. Choose an origin a E HinHj
and identify T with Eby the bijection t ,.._.. a+t. Let V be the plane orthogonal
to Hif\Hj and passing through a. Put I'= VnDH; (C)nDH; (C) (where DH(C)
denotes the open half-space bounded by Hand containing C (§1, no. 1)) and
let D (resp. D') be the half-line in V contained in H; n V (resp. Hj n V) and
in the closure of I'. For a suitable orientation of V, the set I' is the union of
the open half-lines L1 in V such that
0 < (D, L!) < (D, D').

Let W' be the subgroup of W generated by Si and SJ. For any w E W', the
hyperplanes w(H;) and w(H1) belong to S), contain H; nH1 and do not meet
C. It follows that they do not meet I' (§ 1, no. 5, Prop. 10). The Cor. of
Prop. 7 of§ 2, no. 5 thus implies (ii).
Finally, assertion (iii) follows immediately from (i) and (ii), since mij ~ 2
for i f=. .i.
Remark. Formula (4) is actually valid for all i, .i E I: in fact, 7f /rn;j =0 if

= oo, and if 1: = .i then m;j = 1 and (e;le;) = 1.
m;.i
5. SYSTEMS OF VECTORS WITH NEGATIVE SCALAR

PRODUCTS
Lemm.a 3. Let q be a positive qu.adral'ic for-m on a real vector space V and let
B be the associated symmetric bilinear form. Let a. 1 , . .. , an be elements of V
such that B(a;,a-;) ~ 0 fori f=.j.
(i) If c 1 , ... , Cn are real numbers su.ch that q (2( c;a,) = 0, then
(ii) If q is non-degenerate and ·~f' there exists a linear form f vn V such

that .f(a.;) > 0 for all i, the vectors a 1 , ... , an are linearly independent.
The relation B (a;, a;) ~ 0 for i f=. j immediately implies that
hence (i). If q is non-degenerate, the relation L c;a; =0 thus implies that

'
L lc;la; = O;
'
it follows that, for any linear form .f on V, we have L lc;lf(a;) = 0, and
hence c; = '
0 for all 1: if we also have f (a;) > 0 for all i. This proves (ii).
Lemma 4. Let Q = (Qij) be a real, symmetric, square matrix of order n such

that:
a) Qi.I :( 0 for i i= j;
b) there does not exist a parl'ition of { 1, 2, ... , n} ·into two non-empty
subsets I and J such that (i, j) E I x J implies Qij = O;
c) the q·uadratic form q(x 1 , ... , xn) = ~ qijXiXj on R" is posil'i'oe.
·1·1l
Then:
(i) The kernel N of q is of dimension 0or1. If dimN = 1, N is generated
by a vector all of whose coordinates are > 0.
(ii) The smalle8t eigenvalue of Q is of multiplicity 1 and a corresponding
eigenvector has all its coordinates > 0 or all its coordinates < 0.
Since q is a positive quadratic form, the kernel N of q is the set of isotropic
vectors for q (Algebra, Chap. IX, § 7, no. 1, Cor. of Prop. 2). Let a 1 , ••. , an
be the canonical basis of R". If L C-iUi E N, Lemma 3 shows that we also
have L
.,
jcija; E N, and hence L ~Jilc;j =
!
0 for all j. Let I be the set of i
such that c., f= 0. If j ¢ I, then QJi lei I :( 0 for i E I and Qji [c; I = 0 for
i rt I, so Qii = 0 for .i ¢ I and i E I. Assumption b) thus implies that either
I = 0 or I = { 1, ... , n}. Consequently, every non-zero vector in N has all
its coordinates i 0. If dim N ): 2, the intersection of N with the hyperplane
with equation Xi = 0 would be of dimension ): 1, contrary to what we have
just shown. The preceding argument also shows that, if dim N = 1, then N
contains a vector all of whose coordinates are > 0. This proves (i).
On the other hand, we know that the eigenvalues of Q are real (Algebra,
Chap. IX, § 7, no. 3, Prop. 5) and positive since q is positive. Let ). be the
smallest of them. The matrix Q' = Q- Al,, is then the matrix of a degenerate
positive form q' and the off-diagonal elements of Q' are the same as those of
Q. Consequently, Q' satisfies conditions a), b) and c) of the statement of the
lemma. Since the kernel N' of q' is the eigenspace of Q corresponding to the
eigenvalue >., assertion (ii) follows from (i).
Lemma 5. Let e 1 , •• • , e11 be vectors generating T such that:

a) (e;je 1 ) :( 0 for i =I .ii
b) there does not exist a partition of {l, ... , n} into two non-empty subsets
I and J such that (e; Ie j) = 0 for i E I and j E J.
Then there are two possibilities:
1) (e1, ... ,en) is a basis ofT;
2) n = dim T + l; there exists a family (c,)i,,;;;,,;;n of real numbers > 0
such that L c;e; = 0, and any family (c;)i,,;;i,,;;n of real numbers such that
Z::: c;e; = 0 is proportional to (c;)i,,;;i,,;;n.
' Put Q;j = (ei jc ). The matrix Q = (Q;j) then satisfies the hypotheses of
1
Lemma 4: conditions a) and b) of Lemma 4 are the same as conditions a)
and b) above, and c) is satisfied since~ q;(C;Xj =II L x;e; 11 2 . The kernel N
- . - i.) '·
of the quadratic form q on R", with matrix Q, is the set of (c 1 , ... , en) ER"
such that 2:= c;ei = 0. If N = {O}, the ei are linearly dependent and we are
i
in case 1). If dimN > 0, Lemma 4 (i) shows that we arc in case 2).
Lemma 6. Let (e 1 , ... , en) be a basis of T S'uch that (eilej) :;;:; 0 for i =I j.
(i) If x = 2:= cie.; E T is S'UCh that (xlei) ~ 0 for all i, then Ci ~ 0 for all i.
t
(ii) !f x and y arc two elements of T such that (xlei) ~ 0 and (ylc;) ~ 0
for all i, then (xly) ~ 0. If (xlei) > 0 and (yje.;) > 0 for all i, then (::i:jy) > 0.
Under the hypotheses of (i), assume that Ci < 0 for some i. Let f be the
linear form on T defined by f (ei) = 1 and
The vectors - .r, e 1 , ... , en then satisfy the hypotheses of Lemma 3 (ii) (taking
for q the metric form on T). We conclude that they are linearly independent,
which is absurd. Hence (i). Moreover, if x = 2:= c.;e; E T and if y E T, then
i
(xjy) = 2:= ci(eijy), so (ii)

i
follows immediately from (i).
6. FINITENESS THEOREMS
Lemma 'l. Let A be a set of ·un:it vectors in T. If there e.rists a real nmnber
>. < 1 ,mch that (a.la') :;;:; >. for a, a' E A and a =f o.', then the set A is finite.
For a, o.' E A such that a. f:. a.', we have
I\ a - a' 11 2 = 2 - 2(ala.') ? 2 - 2>..

Now, the unit sphere S of T being compact, there exists a. finite covering of
S by sets of diameter < (2 - 2>.) 1/ 2 and each of these sets contains at most
one point of A, hence the lemma.
Denote by U(w) the automorphism of T associated to the affine map
w E W from E to itself. We have
w(x + t) = w(x) + U(w).t for t ET and x EE.
This defines a homomorphism U from the group W to the orthogonal group

of T; the kernel of U is the set of translations belonging to W.
THEOREM 3. (i) The set of walls of a chamber is .finite.

(ii) The set of directions of hyperplanes belonging to .fj is .finite.
(iii) The group U(W) is finite.
Assertion (i) follows immediately from Prop. 3, (iii) and Lemma 7.
We prove (ii). Let C be a chamber and Wl the set of its walls. The facets
ofC (relative to Yj) are the same a:s those relative to Wl (§ 1, no. 4, Prop. 9).
Since 9J1 is finite, they are finite in number. Since a facet meets only finitely
many hyperplanes belonging to f), the set of hyperplanes belonging to f) and
meeting C is finite, hence so is the set A(C) of unit vectors in T orthogonal to
some hyperplane belonging to f) and meeting C. Consequently, there exists
a real number>.< 1 such that (ala')~>. for a,a/ E A(C) and a =fa'.
Let A be the set of unit vectors in T orthogonal to a hyperplane belonging
to f). Let a, a' E A with a =f a'. If a and a' are parallel, then a = -a' and
(ala')= -1. Otherwise, let HE f) (resp. H' E fJ) be such that a (resp. a') is
orthogonal to H (resp. H'). We have HnH' =f 0, and if x E HnH' there exists
an element w E W such that x E w(C). The vectors U(w).a and U(w).a' then
belong to A(C), we have
(ala')= (U(w).alU(w).a') ~ >.,

and the set A is finite by Lemma 7. Hence (ii).
Now let w E W be such that U(w).a =a for all a EA. Then V(w).t = t
for all t belonging to the subspace of T generated by A. On the other hand, if
t ET is orthogonal to A, we have U(s11).t = t for all HE f), hence U(w).t = t
and finally U(w) = 1. Since U(w)(A) = A for all w E W, we deduce that
U(W) is isomorphic to a group of permutations of the finite set A, hence (iii).
PROPOSITION 4. Let C be a chamber and 91 a set of walls of C. Let W'l'l

be the S'U.bgroup of W generated by the orthogonal re.ftections with respect to
the elements of 91. For H E 91, denote by eH the unit vector orthogonal to H
on the same side of H as C. The follow'ing condit'ions are equivalent:
(i) The gro-up W,n is .finite.
(ii) There exists a point of E invariant under every element of W 'Jl.
(iii) The hyperplanes belonging to 91 have non-empty intersection.
(iv) The fam:ily of vectors (eH)HE'l'l is free in T.
By property (D2) at the beginning of § 3, the stabiliser in W of every
point of Eis finite, so (ii) implies (i).
Since the group W'l'l is generated by the set of reflections with respect to
the hyperplanes belonging to 91, the fixed points of W 'l1 are the points of E
belonging to every hyperplane H E IJl, hence the equivalence of (ii) and (iii).
Assume that there exists a point a of E such that a E H for all H E 91 and
let t ET be such that a+ t E C. Since (eHlew) ~ 0 for H, H' E 91 such that
H =/= H' (Prop. 3), and since (tleH) > 0 for all HE 91, Lemma 3 (ii) implies
that the eH for H E 91 are linearly independent. Consequently, (iii) implies
(iv).
Suppose finally that the family (eH)HE':n is free. Let a be a point of E. For
any hyperplane H E 91, there exists a real number CH such that H consists
of the points a+ t of E with (tleH) =CH· Since the family (eH) is free, there
exists t E T such that ( tieJ.I) o=: cu for all H E 91, and the point a + t of E
belongs to all the hyperplanes HE 91. Thus, (iv) implies (iii).
Remarks. I) Since Wis generated by reflections with respect to the walls of

the chamber C, the preceding proposition gives a criterion for W to be finite.
We shall return to this question in no. 9.
2) Let F be a finite-dimensional real affine space and G a group of au-
tomorphisms of F. For all g E G, denote by U(g) the automorphism of the
space of translations V of F associated tog. Assume that the image U(G) is a
finite subgroup of GL(V); then V has a scalar product invariant under U(G)
(Integration, Chap. VII, § 3, no. 1, Prop. 1). If, in addition, G acts properly
on F when it is provided with the discrete topology, and if it is generated by
reflections, we can apply to G the results of this paragraph.
7. DECOMPOSITION OF THE LINEAR REPRESENTATION

OF WONT
Let I be the set of vertices of the Coxeter graph of W (no. 4) and let J be
a subset of I such that no vertex in .J is linked to any vertex in I - J. Ley C
be a chamber, s the canonical bijection from I to the set of reflections with
respect to the walls of C, and let W 1 ,c be the subgroup generated by the
image s(J). It follows from Chap. IV, § 1, no. 9, Prop. 8, that Wis the direct
prod11.ct of the two subgroups W1,c and W1-J,C· Let C' be another chamber
and s' the corresponding injection of I into W. We have seen (no. 4) that if
w E W transforms C into C', then s'(i) = ws(i)w- 1 for i EI. Since WJ,C is
normal in W, it follows that s'(i) E W1,c for all i E J. We deduce that the
S?tbgroup W1.c does not depend on C. We denote it simply by w_, from now
on.
The definition of W J,C clearly extends to an arbitrary subset J of I. But if there

exist a vertex in J and a vertex in I - J that are linked, then W J .c is not
normal and depends on the choice of C.
Let T~ be the subspace of T consisting of the vectors invariant under

every element of U (W J) and let T J be the subspace orthogonal to T~. Since
W 1 is a normal subgroup of W, it is clear that T~ is invariant under U(W),
and hence so is T J.
PROPOSITION 5. Let J 1 , ... , Js be the sets of vertices of the connected com-

ponents of the Co.Teter graph of W. For 1 !( p !( s, pitt
WP= WJp, Tp = T.ip, T~ = T~P' and To= n
l~)r~.s
T~.
(i) The gronp W is the direct product of the groups WP (1 ~ p !( s).

(ii) The space T is the orthogonal direct sv.m of the subspaces T 0 , T 1 , ...
. . . , Ts, which are all stable under U (W).
(iii) For all q sv.ch that 1 ~ q :;::; s, the subspace T~ of T consists of the
vectors invariant under U(Wq); it is the direct sum of the Tp for 1:;::; p ,:_: ; s
and pf- q.
§ 3. GROUPS OF DISPLACEIVIENTS 87
(iv) Let C be a chamber. The subspace Tp (for 1 ~ p ~ s) 'ts generated by

the vectors e.;(C) for i E JP (in the notation of no. 4).
(v) The representations of W 'tn the subspaces T P (1 ~ p ~ s) are abso-
lutely irreducible, non-trivial, and pairwise ineq'uivalent.
Assertion (i) follows from Chap. IV, § 1, no. 9. On the other hand, we
have already seen that the subspaces Tp arc invariant under U(W), and so is
To. Let C be a chamber; since WP is generated by the reflections si(C) for
i E JP, it is clear that T~ is the subspace orthogonal to the ei(C) for i E Jp,
hence (iv). Moreover, if i E Jp, .i E Jq with p I q, then mij = 2 since {i,j} is
not an edge of the Coxeter graph of W, so (eilej) = 0. Assertion (ii) is now
immediate. And assertion (iii) follows, since T~ is the orthogonal complement
of Tq·
Finally, let V be a subspace of Tp invariant under U(Wµ). For all i E JP,
either e; E V or ei is orthogonal to V (§ 2, no. 2, Prop. 3). Let A (resp. B)
be the set of i E .JP such that e; E V (resp. e; is orthogonal to V). Clearly
(eile.i) = 0 for i E A and j E B, and since Jr is connected, it follows that
either A = 0 and V = {O}, or A = JP and V = Tr. Consequently, the
representation of WP on T P is irreducible, hence absolutely irreducible (§ 2,
no. 1, Prop. 1). It is non-trivial by the very definition of Tp. Finally, the last
assertion in (v) follows immediately from (iii).
If the subspace T 0 of vectors in T invariant under U(W) reduces to {O},
then W is said to be essential; if the representation U of W on T is irreducible,
then W is said to be irreducible.
COROLLARY. Assume that WI {l}. Then W is irreducible ~f' and only ~f

it is essential and its Coxeter graph is connected,
Remark. Under the hypotheses of Prop. 5, the subspaces T P for 0 ~ p ~

s are the isotypical components of the linear representation U of W on T
(Algebra, Chap. VIII, §3, no. 4). It follows (Loe. cit., Prop. 11) that every
vector subspace V of T stable under W is the direct sum of the subspaces
V n Tp for 0 ~ p ~ s; moreover (loc. cit., Prop. 10), every endomorphism
commuting with the operators U(w) for w E W leaves stable each of the Tp
for 0 ~ p ~ s and induces on them a homothety for 1 ~ p ~ s. In particular,
the bilinear forms </> on T invariant under W are the bilinear forms given by
where if> 0 is a. bilinear form on To and ap (for 1 ~ p ~ s) is a real number:

indeed, such a form if> can be written uniquely in the form (t, t') 1--> (A(t)lt'),
where A is an endomorphism of T commuting with the U(w) for w E W.
8. PRODUCT DECOMPOSITION OF THE AFFINE SPACE E
We retain the notation of Prop. 5. For 0 ~ p ~ s, let Er be the set of orbits

of the group T~ in E, and let 'Pr be the canonical map from E to Er· The
action of T on E passes to the quotient; in particular, T P acts on Ep and
it is immediate (for example by taking an origin in E) that Ep is an affine
space admitting T P as its space of translations. Put E' = Eo x · · · x Es; this
is an affine space having T = To EB··· EB Ts as its space of translations. Let
<p : E ----+ E' be the product map of the 'Pri since <p commutes with the action
of T, this is a bijection and even an isomorphism of affine spaces. In what
follows, we identify E and E' by means of <p; the map <pp is then identified
with the canonical projection of E' onto Ep.
Since W leaves T~ stable, the action of W on E passes to the quotient
and defines an action of Won Ep (for 0 ~ p ~ s), and hence by restriction an.
action of WP on Ep (for 1 ~ p ~ s). On the other hand, let C be a chamber,
let i E I and let p be the integer such that i E JP. For any x E E, we have
s.;(C)(x) =x - .A.e.;(C) with .A ER.
Since ei E T~ for q I p, it follows that
ipq(w(x)) = ipq(:r) for x EE, w E WP, 0 ~ q ~sand qi p.

Consequently, if w = Wt ... Ws E W with Wp E WP for 1 ~ p ~ s, then
w((:co, ... , X8 )) = (xo, W1 (x1), ... , W (Xs))

5 (5)
for all Xp E Ep for 0 ~ p ~ s. In other words, the act'ion of W on E = E'

is exactly the product of the actions of the WP on Ep (we put W 0 = {1} ). It
follows that WP acts faithfully on Ev and that WP can be identified with a
group of displacements of the euclidean space Ep (the space T P of translations
of Er being provided, of course, with the scalar product induced by that on
T).
PROPOSITION 6. (i) The group WP is a group of d·isplacements of the

euclidean affine space Ep: it is generated by re.fl.ections; prov·ided with the
discrete topology, it acts properly on Ep; it is irreducible.
(ii) Let f:Jp be the set of hyperplanes H of Ep such, that SH E W p· The set
f:J consists of the hyperplanes of the form
H = Eo x E1 x · · · x Ep-t x Hr x Eµ+1 x · · · x E 5,
with p = 1, ... , s and Hp E fJp.

(iii) Every chamber C ·is of the form Eo x C 1 x · · · x C 8 , where for each p
the set Cp is a chamber defined ·in Ep by the set of hyperplanes f:Jp; moreover,
the walls of Cp are the hyperplanes ipp(Hi(C)) for ·i E Jp.
Let C be a chamber. Put Hi= Hi(C), ei = ei(C) ands;= si(C) for i EI
(in the notation of no. 4).
(i) Let i be in Jp; since ei E T P and T is the direct sum of the mutually
orthogonal subspaces T 0, T 1 , ... , Ts, the hyperplane of T orthogonal to ei is
of the form Li + T~, where L; is the hyperplane of T P orthogonal to e;. The
affine hyperplane H; of E is of the form L; + T;>
+ x, with x E E, and we have
H; = Eo x E 1 x · · · x Ep- 1 x H~ x Ep+ 1 x · · · x Es (6)
with H; = L; + C,Op(x) = cpp(H;). It is now immediate that Si acts in Ep by

the reflection associated to the hyperplane H; of Ep· Thus, the group WP is a
group of displacements generated by reflections in Ep; the verification of the
properness criterion (D'2) is immediate. Finally, Prop. 5, (v) shows that WP
is irreducible. This proves (i).
(ii) By the Cor. of Th. 1, the set S'Jp consists of the hyperplanes of the form
wp(H;) for i in Jp and Wp in W p· Further, if w = w 1 ... w 8 with 'Wp E WP for
all p, formulas (5) and (6) imply that
w(Hi) = Eo x E 1 x · · · x Ep-I x wp(H'.) x Ep+I x · · · x E.,, (7)
from which (ii) follows immediately.

(iii) Let i be in Jp; by formula (6), the open half-space D;. bounded by Hi
and containing C is of the form
D; = Eo x E1 x · · · x Ep-1 x D~ x Ep+1 x · · · x E .. ,
where D;, is an open half-space bounded by n D;; since

H; in Ev· Put Cp = iEJp
n D.;, it follows immediately that
c = iEI
C = Eo x C1 x · · · x Cs;
consequently, none of the sets Cp is empty, and since C does not meet any
hyperplane belonging to f), the set Cp does not meet any hyperplane belong-
ing to S'Jw Prop. 5 of § 1, no. 3 now shows that Cp is one of the chambers
defined by S'Jp in Ep. By using Prop. 4 of§ 1, no. 2, it is easy to see that the
walls of Cp are the hyperplanes H;
= <pp(H;) for ·i. E J1,.
9. STRUCTURE OF CHAMBERS
Let C be a chamber, let 9Jl be the set of walls of C, and for H E 9Jl let eH be
the unit vector orthogonal to H on the same side of H as C.
PROPOSITION 7. Assume that the group W is essential and finite. Then:

(i) There exists a unique point a ofE invariant ·undcrW.
(ii) The family (eH)HE!JJ'l is a basis of T.
(iii) The chamber C is the open simplicial cone with vertex a defined by
_the basis (e~)Hem ofT such that (culei1 ,) = Ottw·
(i) By Prop. 4 of no. 6, there exists a point a EE invariant under W. Let

t E T be such that t + a is invariant under W. For all w E W,
U(w).t +a= w(t +a) = t +a,
so U(w).t = t; since W is essential, this implies that t = 0, showing the
uniqueness of a.
(ii) Since Wis essential, T = T 1 in the notation of no. 7, and Prop. 5, (iv)
shows that the family (eH)HE!Ui generates the vector space T. The existence
of a point of E invariant under W shows that the family· (eH)HE<JJl is free
(no. 6, Prop. 4).
(iii) Let a be the unique point of E invariant under W. Since (e1-i)HE!JJI is
a basis of T, and since the scalar product is a non-degenerate bilinear form
on T, there exists a unique basis (c~1 )Hem ofT such that (cHlel1,) = OHw for
H, H' in 9J1. Every point x of E can be written uniquely in the form x = t +a
with t = L fo.e;1 and the fo real. Then :x: belongs to C if and only i( for
HE'.lJl
every hyperplane H E 9J1, x is on the same side of H as eH, or in other words
(tJeH) = (H is strictly positive. Hence (iii).
PROPOSITION 8. Assume that the gro·up W is essenl'io.l, irrcd11.cible and

infinite. Th.en:
(i) No point of E is ·invo:riant ·under W.
(ii) We have Card 9J1=dimT+1, and there exist real mtmbers c1-1 > Osuch
that L en .eH = 0. If the real numbers c~ are s·u.ch that L c~ .e 8 = 0, there
HE!JJI HE9Jl
exists a real nv:rnber ( such that c~ = ~CH for· all H in 9J1.
(iii) The chamber C is an open simplex.
Assertion (i) follows from Prop. 4. On the other hand, since Wis essential,
the vectors (e11)HE9Jl generate T. We have (eHJew) ~ 0 for H, H' E 9J1 and
H # H' (Prop. 3) and, since W is irreducible, there does not exist a partition
of 9J1 into two disjoint subsets 9J1' and IJR" such that H' E 9J1' and H" E 9J1"
imply that (eH' leH") = 0. We can thus apply Lemma 5 of no. 5, and case 1)
of that lemma is excluded; indeed, the eH are not linearly independent, since
W has no fixed point. Assertion (ii) follows.
We now prove (iii). Number the walls of C as H 0 , H 1 , ... , Hd and put
tm = eH,,,. By (ii), the vectors t 1 , ... , td form a basis of T, so the hyperplanes
H1, ... , Hd have a point ao in common, and there exists a basis (t;, ... , td) of
T such that (tmlt~J =Omni moreover, again by (ii) there exist real numbers
c1 > 0, ... , Cd > 0 such that
Since the vector to is orthogonal to the hyperplane H 0 , there exists a real

number c such that Ho is the set of points x = t + ao of E with (tlto) = -c.
Every point of E can be written uniquely in the form x = t + a0 with
t = ~1.t; + · · · + {d.td and {1, ... , {d real. The point :r: belongs to C if and
only if it is on the same side of Hm as tm for 0 :::.; m. :::.; d; this translates into
the inequalities (tlt 1 ) > 0, ... , (tltd) > 0, and (tlto) > -c, or equivalently
by 6 > 0, ... , t:d > 0, c 1t: 1 + · · · + C<1{<1 < r:. Since C is non-empty, c > 0.
Put am = a 0 + _s_ Cm
.t~ for 1 :::.; m :::.; d; then the chamber C consists of the
d
points of E of the form ao + I: A,.,,.. (am - ao) with A1 > 0, ... , Ad > 0 and
m=l
A1 +···+Ad < 1, so C is the open simplex with vertices a 0 , ... , ad. Q.E.D.
Remarks. I) Identify E with Eo x E 1 x · · · x Es and W with W 1 x · · · x Ws

as in no. 8. By Prop. 6, the chamber C is then identified with
where Cp is a chamber in Ep with respect to the set of hyperplanes f)p. By

Props. 7 and 8, each of the chambers C 1 , ... , Cs is an open simplicial cone
or an open simplex.
2) Assume that W is irreducible and essential. If H and H' are two walls
of C, then m1-11-1' = +oo if and only if eH = -e1-1' (Prop. 3). By Props. 7
and 8, this can happen only if H and H' are the only walls of C and E is of
dimension 1. Thus, the only case in which one of the rn1-1w is infinite is that
in which E is of dimension 1 and the group W is generated by the reflections
associated to two distinct points (cf. § 2, no. 4).
In the general case, the entries of the Coxeter matrix associated to W are
finite unless at least one of E 1 , ... , E., is of the preceding type.
10. SPECIAL POINTS
Let L be the set of translations belonging to W and let A be the set of t E T

such that the translation x r---+ t + x belongs to L. It is immediate that A
is stable under U(W) and that L is a normal subgroup of W. Since W acts
properly on E, the same holds for L, and it follows easily that A is a discrete
subgroup of T. For any point x of E, denote by W:r the stabiliser of.Tin W.
PROPOSITION 9. Let a E E. The following conditions are eqttivalent:

(i) We have W = W 0 .. L;
(ii) The restriction of the homomorph-ism U to Wa is an ·isomorphism
from Wa. to U(W);
(iii) For every hyperplane H E f), there exists a hyperplane H' E f) parallel
to H and s'Uch that a E H'.
It is clear that (i) {:}(ii), since Lis the kernel of U and L n Wa = {l}.
Assume (i) and let H E 5); then SH E W a·L so there exists a vector t E A
such that a= sH(a) + t; the vector tis orthogonal to H, and if H' = H + ~t
then sw(x) = stt(x) + t for all x EE (cf.§ 2, no. 4, Prop. 5). Since t E 11 and
SH E W, we have sw E W, and so H' E fj; we also have a= s~I'(a), and so
-a; EH'. Thus, (i) implies (iii).
Assume (iii). Let HE Sj; take H' as in (iii). Then sH'(a) =a, so SH' E Wa;
since H is parallel to H', the element w = swsH of W is a translation (§ 2,
no. 4, Prop. 5), sow EL; then SH= 8H W E W 0 .L. Since Wis generated by
1
the family (sH)HEfJ• it follows that W = Wa.L and hence (iii) implies (i).
DEFINITION l. A poi.nt a of E is specfol for W if it satisfies the eq1Livalent

conddions in Prop. 9.
It is clear that the set of special points of E is stable und~_r W.
PROPOSITION 10. There exists a special point for W.

By Prop. 6 of no. 8, we need only consider the case when W is essential.
The group U(W) of automorphisms of T is finite (cf. no. 6, Th. 3) and
U(sH) is an orthogonal reflection for every hyperplane H; moreover, U(W) is
generated by the family (U(sH))HE.fJ· By Prop. 7 of no. 9, there exists a basis
(e;)iEI of T such that the group U(W) is generated by the set of reflections
(s;)iEI defined by
si("t) = t - 2(tlei).e;.
The Cor. of Th. 1 of no. 2 shows that every reflections E U(W) is of the form
s = U(sH) with HE Sj. We can thus find in Sj a family of hyperplanes (H;);EJ
such thats; = U(sH,) for all i. Since the vectors e; are linearly independent,
there exists a. E E such that a E H; for all i E I. We have SH., E W a, so
U(W) = U(Wa), which means that W = Wa.·L since Lis the kernel of U.
Thus, a is special.
When W is finite and essential, there is only one special point for W, namely
the unique point invariant under W. Thus, the consideration of special points
is interesting mainly when W is infinite.
PROPOSITION 11. Assume that W is essential. Let a lie a special point for
W. The chambers relative to W a are the open simplicial cones with vertex
a. For every chamber C' relat-ive to W a, there exists a uniq·ue chamber C
relative to W contained in C' and such that a E C. The union of the w' (C)
for w' E W a is a closed neighbourhood of a in E. Every wall of C' is a. wall of
C. If W is infinite and 'irreducible, the walls of C are the walls of C' together
w1.th an affine hyperplane not parallel to the walls of C'.
Let f)' be the set of H E Sj containing a. The group W a is generated by
the SH for HE Sj' (no. 3, Prop. 2). The chambers relative to Wa. are the open
simplicial cones with vertex o. (no. 9, Prop. 7). Let C' be such a chamber and
let U be a non-empty open ball with centre a not meeting any element of
fJ - Sj'. Since a E C', there exists a point b in Un C'. Then b tf_ H for all H E f),
so b belongs to a chamber C relative to Sj. Since Sj' c f), we have C c C'. The
set Un C' does not meet any H E Sj and is convex, so Un C' c C; thus a E C.
Conversely, let C 1 be a chamber relative to Vv contained in C' and such that
a E C1 ; then C1 meets U and Un C 1 c Un C' = Un C; the chambers C

and C 1 , having a point in common, must coincide. For any w' E Vva, we have
w'(U) = U, so
Un w'(C) = w'(U n C) = w'(U n C') =Un w'(C');
since the union of the w' (C) for w' E W a is dense in E, the union of the
Un w' (C) = Un w' ( C') is dense in U, and the union of the w' (C) for w' E W a
thus contains U. Finally, if H is a wall of C', there exist a point c E U n H
and an open neighbourhood V c U of c such that V n C' is the intersection
of V and the open half-space bounded by H containing C'; since V n C' =
V n U n C' = V n U n C = V n C, it follows that H is a wall of C. If W is
infinite and irreducible, C is an open simplex (no. 9, Prop. 8) and so has one
more wall than the open simplicial cone C'.
COROLLARY. Assume that W is essential.

(i) If a E E is special, there exists a chamber C such that a is an extremal
po·int of C.
(ii) Jf C ·is a chamber, there exists an e:rtremal point of C that is special.
The first assertion follows from Prop. 11. The second follows from the
first and the fact that W acts transitively on the set of chambers.
However, an extremal point of C is not necessarily a special point for W (cf.

Chap. VI, Plate X, Systems B2 and G2).
Remark. 1) Assume that W is essential, irreducible and infinite and retain

the notation of Prop. 11. Since U is an isomorphism from Wa to U(W), it
follows that the Coxeter graph of the group of displacements U(W) (which
is generated by the reflections U (SH) for H E fJ) can be obtained from the
Coxeter graph of W by omitting the vertex i corresponding to the unique
wall of C that is not a wall of C'.
PROPOSITION 12. Assume that W is essential. Let a be a special point, let

L(a) be the set of its transforms ·u.nder the grov.p of translations L, and let C
be a. chamber. Then C meets L(a) in a ·uniqv.e point; this point 'tS an extremal
point of C.
There exists a chamber C 1 such that a is an extremal point of C1 (Cor.
of Prop. 11). Every chamber is of the form C = tw' ( C 1 ) with w' E W a and
t EL since W = Wa.L; thus tw'(a) = t(a) E L(a) is an extremal point of C.
On the other hand, C cannot contain two distinct points of L(a) since C is a
fundamental domain for W (no. 3, Th. 2).
Remark. 2) The set L(a) is contained in the set of special points, but in
general is distinct from it (cf. Chap. VI, § 2, no. 2 and Plates I to VI).
§ 4. GEOMETRIC REPRESENTATION OF A
COXETER GROUP
In this paragraph, all the vector spaces considered are real vector spaces.
1. FORM ASSOCIATED TO A COXETER GROUP
Let S be a set and let M = (m(s, s')s.s'ES) be a Coxeter matrix (Chap. IV,
§ 1, no. 9) of type S. Recall that this means:
(1) The elements of M are integers or +oo.
(2) M is symmetric.
(3) m(s, s) = 1 for alls.
(4) m(s, s') ~ 2 for s =Is'.
Let E = R(S), let (es)sES be the canonical basis of E, and let BM be the
bilinear form on E such that
7f
B1v1(es,e 8 )=-cos
1 ( ).
m. s,s'
The form BM is symmetric. It is called the assocfoted .form of the matrix
M. We have
BM(es,e.-) = 1 and BM(e 8 ,C 8 1 ) :( 0 if s =/ s'.

Lets ES and let fs be the linear form x ,___. 2BM(e 8 ,x). We denote by as the
pse·udo-refiectfon defined by the pair (e 8 ,f,) (cf. § 2, no. 1); since (e 8 , fs) = 2,
it is a reflection (§ 2, no. 2). We have
as(x) = x - 2B1v1(es X).es

1
and in particular
7r
as(e,1)=es 1+2cos ( ')"es.
m. s,s
Since e 8 is not isotropic for BM, the space Eis the direct sum of the line Res
and the hyperplane Hs orthogonal to e 8 • Since a, is equal to -1 on Res and
to 1 on H" it follows that as preserves the form BM. When S is finite and
BM is non-degenerate (a case to which we shall return in no. 8), it follows
that a 5 is an orthogonal refiection (§ 2, no. 3).
§ 4. GEOMETRIC REPRESENTATION OF' A COXETER GROUP 95
2. THE PLANE Es,s' AND THE GROUP GENERATED BY us

AND Us 1
In this number, we denote by s and s' two elements of S with s i= s'. Put
m = m(8, s') and denote by Es.s' the plane Res@ Re.5 1 •
PROPOSITION 1. The restriction of B1v1 to Es,s' is positive, and it is non-

degenerate if and only if m is finite.
Let z = xes +ye .. ,, with x, y ER, be an element of Es,s'· We have
BM(z, z) = x2 - 2xycos !!_ + y2

m
2
= (X - y. COS -7r) + y 2·27r
Sll1 - ,
·m m
which shows that B;v1 is positive on Es,s', and that it is non-degenerate there
if and only if sin :!;i i= 0. The proposition follows.
The reflections <7 s and <7s' leave Es,s' stable. We are going to determine
the order of the restriction of <7 8 <7 s' to Es,s'. We distinguish two cases:
a) m = +oo.
Let u = es+ e8 1. \Ve have BM('u, P.s) = Brv1(u, es') = 0, sou is invariant
under <7 8 and a 8 1. .Moreover,
<7s ( <7s 1 (es)) = <7s (es + 2e_,,) = 3es + 2es' = 2u + e."

hence
(<7 8 <7 5 )n(es)
1 = 2nu +es for all n E Z.
It fo11ows that the restriction of 17 5 <7 s' to Es,s' is of infinite order.
b) m is .finite.
The form I3H provides Es,s' with the structure of a rnclidean plane. Since
the scalar product of e5 and Cs' is equal to - cos :!!.. = cos (7r - :!!..
m.
) , we can
TYi.
orient Es,s' so that the angle between the half:. lines R+ es and R+ es' is equal
to 7r - :!;i. If D and D' denote the lines orthogonal to es and es',
- - 1T
(D',D) = 7r - (D,D') = - .
rn
Now, the restrictions as and as' of <7 s and cr s' to Es,s' are the orthogonal
symmetries with respect to D and D'. By the Cor. to Prop. 6 of§ 2, no. 5, it
follows that 0' 8 0 8 is the rotation with angle ;;: . In particular, its order ism.
1
We now return to E:
PROPOSITION 2. The subgrmtp of GL(E) generated by as and <70 is a 1
dihedral group of order 2m( s, s').

Since a 8 and <7s' are of order 2, and are distinct, it is enough to show that
their product <78 <7 5 1 is of order 1n(s, s'). When rn(s, s') is infinite, this follows
from a) above. When m(s, s') is finite, it follows from Prop. 1 that Eis the
direct sum of Es,s' and its orthogonal complement V s,s'; since as and as' are
the identity on V s,s', and since the restriction of a 5 a s' to Es.s' is of order
m(s, s') by b), the order of a 5 <J5 is indeed equal to m(s, s').
1
3. GROUP AND REPRESENTATION ASSOCIATED TO A

COXETER MATRIX
We retain the notation of the preceding numbers. Let w- = W(M) be the

group defined by the family of generators (gs)sES and the relations 1
, )m(s,s')
(g sYs' _ 1 for s, s' E S, m(s, s') :/- +oo.
- '
PROPOSITION 3. There exists a unique homomorphism
Cl : W --> GL(E)
such that a(g 8 ) = <J5 for alls E S. The elements of a(W) preserve the bilinear
form BM.
To prove the existence and uniqueness of Cl, it is enough to show that
(ClsCls' rn·<s,s') = 1 if m(s, s') :/- +oo. Now, ifs = s', this follows from the
fact that Cls is of order 2; ifs:/-- s', it follows from what we proved in no. 2.
Finally, since the reflections a,, preserve BM, so do the elements of <J(W).
Remark. 1) We shall see in no. 4 that Cl is injective; the group W can thus
be identified with the subgroup of GL(E) generated by the <J 5 •
PROPOSITION 4. a) The map s f---> 9s from S to W is in,jective.

b) Ea.ch of the g5 is of order 2.
c) Ifs, s' E S, 9s9s' is of order m(s, s').
Assertion a) follows from the fact that the composite map
S r--+ 9s >-7 as
from S to GL(E) is injective.

For b) (resp. c)), we remark that the order of 9s (resp. the order of 9s9s')
is at most 2 (resp. at most m(s, s')). Since we have seen in no. 2 that the
order of <J 5 (resp. of Clsas') is 2 (resp. m(s, s')), we must have equality.
In view of a), Scan be identified with a subset of W by means of the map
Sr--+ 9s·
1 This means that, if Ls denotes the free group on S, W is the quotient of Ls

by the smallest normal subgroup of Ls containing the elements (ss')m(.s.s'>, for
m(s, s') I +oo.
§ 4. GEOMETRIC REPRESENTATION OF A COXETER GROUP 97
COROLLARY. The pair (W, S) is a Coxeter system with matrix M.

This is simply the content of properties b) and c), together with the defi-
nition of W.
Remark. 2) We have thus shown that every Coxeter matrix corresponds to

a Coxeter group.
4. CONTRAGREDIENT REPRESENTATION
Let E* be the dual of E. Since W acts on Evia (J, it also acts, by transport
of structure, on E*. The corresponding representation
(J* : W ~ GL(E*)
is called the contragredient representation of (J. We have
(J*(w) = t(J(w- 1 ) for all w E W.
If x* EE* and w E W, we denote by w(x*) the transform of x* by (J*(w).

If s E S, denote by As the set of .r* E E* such that x* ( e5 ) > 0. Let C be
the intersection of the A 8 , s E S. When S is finite, C is an open s·implicial
cone in E* (§ 1, no. 6).
THEOREM 1 (Tits). If w E W and C n w(C) =J 0, then w = 1.

'vVe indicate immediately several consequences of this theorem:
COROLLARY 1. The group W acts simply-tro.ns'itively on the set of w(C)

forw E W.
This is clear.
COROLLARY 2. The representations (J and (J* are injective.

Indeed, if a*(w) = 1, then w(C) = C, so w = 1 by the theorem. The
injectivity of o follows from that of (J*.
COROLLARY 3. If S is· finite, o(W) is a d·iscrete subgroup of GL(E) (pro-

vided with its canonical Lie group structure}; similarly, (J* (W) is a discrete
subgroup of GL(E*).
Let x* E: C. The set U of g E GL(E*) such that g(x*) E C is a neigh-
bourhood of the identity element in GL(E*); by the theorem,
o*(W) n U = {l};
thus, (J*(W) is a discrete subgroup of GL(E*). By transport of structure, it
follows that o(W) is discrete in GL(E).
Proof of Theorem 1.
If w E W, denote by l(w) the length of w with respect to S (Chap. IV,
§ 1, no. 1).
We are going to prove the following assertions, where n denotes an integer
~ 0:
(Pn) Let w E: W, w'ith l(w) = n, ands ES. Then:
either w(C) C As;
or w(C) c s(A,) and l(sw) = l(w) - 1.
(Qn) Let w E W, with l(w) = n, and s,s' ES, sf:. s'. Let Ws,s' be !:he
s·ubgroup of W generated by s and s'. There exists u E vV s,s' such that
w(C) C u(As n A,,) and l(w) = l(u) + l(u- 1 w).
These assertions are trivial for n = 0. We prove them by induction on n,

according to the scheme
((Pn) and (Qn)) ==:? (Pn+1) and ((P,,+1) and (Q,i)) ==:? (Qn+d·
Proof that ((Pn) and (Qn)) ==:? (Pn+1).
Let w E W, with l(w) = n + 1, and s E S. We can write w in the form
w = s'w' withs' ES an<l l(w') = n. Ifs'= s, (Pn) applied tow' shows that
w'(C) C As, hence w(C) C s(As) and l(sw) = l(w') = l(w) - 1. Ifs' =/= s,
(Qn) applied tow' shows that there exists v. E Ws,.,' such that
w'(C) C ·u(A, n As') and l(w') = l(u) + l(u- 1 w').

We have w(C) = s'w'(C) C s'u(A, n As').
Lemma 1. Let s, s' E S, with s is', and let v E W s,s'. Then v(As n As') is
contained in either As or in s( A,), and in the second case l( sv) = l (v) - 1.
The proof will be given in no. 5.

We apply the lemma to the element v = s'u. There are two possibilities:
either
.> 1 u(As n As') C As and a fortiori w(C) C A 5 ,
or
s'u(As n As') C s(As) and a fortiori w(C) C s(A 5 ).
Moreover, in the second case, l(ss'u) = l(s'u) - 1. Hence
l(sw) = l(ss'w') = l(ss'u.u- 1 w') ~ l(ss'v.) + l(u- 1 w')

= l(s'u) + l(u- 1 w 1 ) - 1 ~ l(w) - 1,
and we know that this implies that l(sw) = l(w) - 1.

Proof that ((Pn+i) and (Qn)) ==::} (Qn+1).

Let w E W, with l(w) = n + 1, ands, s' ES, s =I= s'. If w(C) is contained
in As n A,,, condition ( Qn+ i) is satisfied with u = l. For if not, suppose for
example that w(C) is not contained in As. By (Pn+il, w(C) C s(As) and
l(sw) = n. By (Qn), applied to sw, there exists v E Ws.s' such that
sw(C) C v(As n As') and l(sw) = l(v) + l(v- 1 sw).
Then
w(C) C sv(As n As')
and
l(w) = 1 + l(sw) = 1 t- l(v) + l(v- 1 sw)
? l(sv) + l((sv)- 1 w):;::: l(w),
so the inequalities above must be equalities. It follows that (Qn+il is satisfied
with u = sv.
Proof of the theorem.
Let w E W, with w jc l. We can write win the form sw' withs ES and
l(w') = l(w) - 1. By (P 11 ), applied tow' and n = l(w'), we have w'(C) C As,
since the case w'(C) c s(A,) is excluded because l(sw') = l(w) = l(w') + 1.
So w(C) = sw'(C) C s(A,), and o;incc As and s(A 5 ) are disjoint, we have
C n w(C) = 0. Q.E.D.
5. PROOF OF LEMMA 1
Let E;,,, be the dual of the plane E,,,s' =Res (B Re 5 1 (no. 2). The transpose
of the injection Es,s' __, E is a surjection
p : E* __, E;,,,
that commutes with the action of the group W s,s'. It is clear that A 5 , As'
and As n As' a.re the inverse images under p of corresponding subsets of E;,s'
(considered as the space of the contragredient representation of the Coxeter
group W s,s'). Moreover, since the length of an clement of W s,s' is the same
with respect to {s,s'} and with respect to S (Chap. IV,§ 1, no. 8), we are
reduced finally to the case where S = {s,s'}: if m = m(s,s'), the group Wis
then a dihedral group of order 2m.
We now distinguish two cases:
a) m = +oo.
Let (E,E 1 ) be the dual basis of (e,,e 5 1). Then
S.E=-E+2E 1 , S 1 .E=E,
S.E 1 = E1 , S 1 .E 1 = 2E - E1 •
Let D be the a.ffine line of E* containing € and € 1 ; the formulas above

show that D is stable under s and s' and that the restriction of s (resp. s')
to D is the reflection with respect to the point € 1 (resp. c:). Let
e
be the affine bijection t .-.. @( t) = tc: + (1 - t )c:'. Let In be the image under
of the open interval )n, n + 1(, and let Cn be the union of the ,\In for ,\ > 0.
Then Co = C; moreover, by the Remark of § 2, no. 4, applied to the affine
space D, the In are permuted simply-transitively by W; hence so arc the Cn.
If v E W, v( C) is equal to one of the Cn, hence is contained in As if n ~ 0
and in s(A,) if n < 0. In the second case, Io and In are on opposite sides of
the point c-'; hence l(sv) = l(v) - 1 (lac. cit.).
b) m is finite.
The form BM is then non-degenerate (no. 2) so we can identify E* with E.
We have seen that E can be oriented so that the angle between the half-lines
R+es and R+es' is equal to K - ;;;. Let D (resp. D') be the half~line obtained
from R+es (resp. R+es') by a rotation of' 7f /2 (resp. -11 /2), cf. Fig. 2. The
chamber C is the set of x E E whose sea.Jar product with Cs and Cs' is > O; this
is the open a.ngula.r sector with origin D' and extremity D. By the Remark
of § 2, no 5, every element v of W transforms C into an angular sector that
is 011 the same side of D as C (i.e. is contained in As) or on the opposite side
(i.e. contained in s(A,)), and in the latter case
l(sv) = l(v) - 1,
which completes the proof of the lemma.
D D'
Fig. 2
6. FUNDAMENTAL DOMAIN OF W IN THE UNION OF THE

CHAMBERS
We retain the notation of no. 4. For s E S, denote by Hs the hyperplane of

E* orthogonal to es, by As the set of x* E E* such that (x*, e8 ) ? 0 and by C
the intersection of the As for s E S. For the weak topology cr(E*, E) defined
by the duality between E* and E (Topological Vector Spaces, Chap. II, §6,
no. 2), the As are closed half-spaces and C is a closed convex cone. Moreover,
C is the closure of C; indeed, if x* E C and y* E C, then x* + ty* E C for
every real number t > 0 and x* = limt-o(x* + ty*).
For X CS, put
Cx = ( n Hs) n ( n
sEX sES-X
As) .
We have Cx CC, C0 = C and Cs= {O}. The sets Cx, for XE B(S), form a
partition of C.
On the other hand, recall (Chap. IV, §I, no. 8) that Wx denotes the
subgroup of W generated by X. Clearly w(x*) = x* for w E Wx and x* E Cx.
PROPOSITION 5. Let X, X' c S and w, w' E W. If w(Cx) n w'(Cx') =/= 0,

then X = X', wWx = w'Wx' and w(Cx) = w'(Cx, ).
We are reduced immediately to the case w' = I. The proof is by induction
on the length n of w. If n = 0, the assertion is clear. If l(w) > 0, there exists
s ES such that l(sw) = l(w)-1 and then (cf. the end ofno. 4) w(C) C s(A 8 ),
hence w(C) C s(A 5 ). Since CC As, it follows that
C nw(C) c Hs.
Then s(x*) = x* for all x* EC n w(C), and a fortiori for all
x* E Cx' n w(Cx).
Consequently, the relation Cx' n w(Cx) I- 0 implies on the one hand
that Cx· n Hs i 0, and hence that s E X', and on the other hand that
Cx, n sw(Cx) :/; 0. The induction hypothesis then implies that X = X' and
swWx = Wx' = Wx, so sw E Wx and w E Wx since s E Wx. It follows
that wWx = Wx' and that w(Cx) = Cx = Cx'·
COROLLARY. Let X be a subset ofS and x* an element ofCx. The stabiliser

of x* in W is Wx.
Now let U be the union of the w(C) for w E W, and let ;J be the set of
subsets of U of the form w(Cx), with X C S and w E W. By the above, 3' is
a partition of U.
PROPOSITION 6. (i) The cone U is convex.

(ii) Every closed segment contained in U meets only finitely many elements
of 'J.
(iii) The cone C is a fundamental domain for the action of W on U.
To prove (iii), it is enough to show that, if x•, y* E C and w E W are
such that w(:r*) = y*, then ;r;* = y*. Now there exist two subsets X and Y of
S such that x* E Cx and y* E Cy; we have w(Cx) n Cy =J 0, and Prop. 5
shows that X = Y and w E Wx, which implies that x* = y*.
Now let x*, y* E U; we shall show that the segment .[x'y*J is covered
by .finitely many elements of 'J, which will establish both (i) and (ii). By
transforming .x* and y* by the same element of W, we can assume that
x* E C. Let w E W be such that y* E w(C). We argue by induction on the
length of w. For s E S, the relation w(C) </. As is equivalent to w(C) </.. As
and hence to l(sw) < l(w) (cf. no. 4). Prop. 7 of no. 8 of Chap. IV,§ 1 now
implies that there exist only finitely many s E S such that w(C) ¢ As. The
set T of s E S such that (y*, e8 ) < 0 is thus finite. On the other hand, the
intersection en [x*y*) is a closed segment [x* z*]. If z* = y*' that is if y• EC,
there exist subsets X and Y of S such that x* E Cx and y* E Cy. The open
segment ]x*y* [ is then contained in CxnY, so [x*y*) C Cx U Cy U CxnY.
If z* f. y*, there exists an s E T such that z* E Hs. Then w(C) cf.. A.,
and l(sw) < l(w). The induction hypothesis thus implies that the segment
[z*y*] = s([z*(s(y*))]) is covered by a finite number of elements of 'J, and
hence so is
[:z:*y*] = [x* z*] u [z*y*]
since [x* z*] C C.
7. IRREDUCIBILITY OF THE GEOMETRIC

REPRESENTATION OF A COXETER GROUP
We retain the notation of the preceding nos., and assume that S is finite.
PROPOSITION 7. Assnrn.e that (W, S) is irred·ucible (Chap. IV,§ 1, no. 9).

Let E 0 be the subspace of E orthogonal to E with respect to BM, The grD'up
W acts trivially on E 0 , and every subspace of E d·istinct from E and stable
under W ·is contained in E 0 ,
If x E E 0 , then o-s(x) = x - 2BM(e 5 , x)e, = x for all s E S. Since W is
generated by S, it follows that W acts trivially on E 0 ,
Let E' be a subspace of E stable under W. Let s, s' E S be two elements
that are linked in the graph I' of (W, S) (Chap. IV, § 1, no. 9); recall that
this means that m(s, s') ?: 3. Suppose that es E E'. Then O"s'(e.,) E E' and
since the coefficient of es' in a- s' ( e5 ) is non-zero, we have e 5 , E E'. Since I'
is connected, it follows that, if E' contains one of the e 5 it contains them all
and coincides with E. Except in this case, it follows from Prop. 3 of§ 2, no. 2
§4. GEOMETRIC REPRESENTATION OF A COXETER GROUP 103
that, for all s E S, E' i:;; contained in the hyperplane H,, orthogonal to e,,.
Since the intersection of the Hs is equal to E 0 , this proves the proposition.
COROLLARY. Ass1tme that (W, S) is irreducible. Then:

a) If BM is non-degenerate, the W-module E is absolutely simple.
b) If BM is degenerate, the 'v\T-module E ·is not sem·i-simple.
In case a), Prop. 7 shows that E is simple, hence also absolutely simple
(§ 2, no. 1, Prop. 1).
In case b), E 0 :f: 0, E :f: E 0 (since BM :f: 0), and Prop. 7 shows that E 0 has
no complement stable under W; thus, the W-module Eis not semi-simple.
8. FINITENESS CRITERION
We retain the notation of the preceding nos., and assume that S is finite.
THEOREM 2. The .following properties are eq·uivalent:

( 1) W is finite.
(2) The form BM is positive and non-degenerate.
(1) ==> (2). Let S = U Si be the decomposition of S into connected
!
components (Chap. IV,§ 1, no. 9), and let W =CT W; be the corresponding
decomposition of W. The space E can be '
identified
with the direct sum of
the spaces Ei = Rs,, and BM can be identified with the direct sum of the
corresponding forms Bl\.1,. We are thus reduced to the case when (W, S) is
irreducible. Since W is assumed to be finite, E is a semi-simple W-module
(Appendix, Prop. 2). By the Cor. to Prop. 5, it follows that E is absolutely
simple. Let B' be a positive non-degenerate form on E, and let B" be the sum
of its transforms under W. Since B" is invariant under W, it is proportional
to BM (§2, no. 1, Prop. l); since BM(e 5 ,c 5 ) = 1 for alls ES, the coefficient
of proportionality is> 0, and since B" is positive so is BM, which proves (2).
(2) ==> (1). If BM is positive non-degenerate, the orthogonal group O(BM)
is compact (Jntegrat-ion, Chap. VII, § 3, no. 1). Since a(W) is a discrete sub-
group of O(BM) (Cor. 3 of Th. 1), it follows that a(W) is finite, hence so is
W.Q.E.D.
The following result has been established in the course of the proof:
COROLLARY. ff (W, S) is irreducible and finite, E is an absolutely simple

W-module.
The finiteness criterion provided by Th. 2 permits the classification of all

finite Coxeter groups (cf. Chap. VI, § 4). We m;trict ourselves here to the
following preliminary result:
PROPOSITION 8. If W is finite, the graph of (W, S) is a forest (Chap. IV,

Appendix).
If not, this graph would contain a circuit (s 1 , ... , sn), n ~ 3. Putting

mi= m(s 1 ,si+i), 1 ~ i < n, and mn = m(sn,s 1 ), this means that mi~ 3
for all i. Let
x = es, + ... + es,,.
Then BM(x,x) = n + 2 Z:::::.BM(e 8 ,,c 5 J. Now 2
•<1
Ti Ti 1
BM(es,es+ 1 ) =-cos-~ -cos-~--?.,
' ' ni; 3 _
and similarly for BM (es,,, e5 i). Since the other terms in the sum are ~ 0, we
obtain
BM(x,x)~n-n=O,
contrary to the fact that BM is positive non-degenerate.
COROLLARY. If (W, S) is irred'Ucible a.nd finite, its gra.vh is a. tree.

Indeed, a connected forest is a tree.
Comparison with the res·ults of§ 3.
First of all let (W, S) be a finite Coxeter group. Denote by (xly) the form
BM(x, y); by Th. 2, this is a scalar product on E. For all s E S, let H 8 be
the hyperplane associated to the orthogonal reflection a 5 , and let fJ be the
family of hyperplanes w(H 5 ), for s E S, w E W. Let C 0 be the set of x E E
such that (xles) > 0 for alls ES. Finally, identify W (by means of a) with
a subgroup of the orthogonal group O(E) of the space E.
PROPOSITION 9. With the preceding nota.tions, W is the subgroup of O(E)

genera.led by the reflections with respect to the hyperplanes of f). It is an
essentia.l group (§ 3, no. 7) and Co is a. chamber of E relative to fJ.
The first assertion is trivial. On the other hand, if :r E E is invariant
under W, it is orthogonal to all the e8 , and hence is zero; this shows that
W is essential. Finally, the isomorphism E -. E* defined by B1vr transforms
C 0 to the set C of no. 4; the property (Pn) proved there shows that, for all
2 The roots of the equation z 3 - 1 = 0 are 1 and -J~iv'3. Thus cos~ = - ~ and
consequently cos j = ~. Note also that sin 2; = 1, hence
• 1T J3 71 511 J3 . 7r . 57T l
s111 3 = T, cos 6 = - cos 6 = 2 , sm 6 = sin 6 = 2.
Similarly, the roots of the equation z 2 - i = 0 are ± 1;}i, so

1T
cos - = s111 -4 = -V2
.'Ir
and consequently
. 37r
Sl\1 - = - COS 37r
-
V2 .
= -
4 2 4 4 2
§4. GEOMETRIC REPRESENTATION OF A COXETER GROUP 105
w E Wand alls E S, w(C 0 ) does not meet H 5 . It follows that Co is contained

in the complement U of the union of the hyperplanes of fJ, and since Co is
connected, open and closed in U, it is a. chamber of E relative to fl. Q.E.D.
We can apply to W and Co all the properties proved in § 3. In particnlar,
Co is a fundamental domain for the action of W on E (in other words, the
cone U defined in no. 6 is equal to the whole of E).
Conversely, let E be a finite dimensional real vector space, provided with
a scalar product (xly) and let W be an essential finite group of displacements
of E leaving 0 fixed; assume that W is generated by reffrctions. Let Co be a
chamber of E with respect to W (cf. § 3), and let S be the set of orthogonal
reflections relative to the walls of C 0 . Then (W, S) is a .finite Coxeter system
(§ 3, no. 2, Th. 1). Moreover, if s E S, denote by Hs the wall of C0 corre-
sponding to s, and denote by es the unit vector orthogonal to H 5 and on the
same side of Hs as C 0 . If (m(s, s')) denotes the Coxeter matrix of (W, S),
Props. 3 and 7 of § 3 show that
(csles') =- cos(7r/m(s, s'))
and that the e 5 form a basis of E. The natural representation of W on E can

thus be identified with the representation a of no. 3.
9. CASE IN WHICH BM IS POSITIVE AND DEGENERATE
In this number, we assume that Sis finite, that (W, S) is irreducible, and that
the form BM is positive and degenerate.
Lemma 2. The orthogonal complement E0 of E for BM is of dimension 1; it

is generated by an element v = L V 5 C8 with Vs > 0 for alls.
sES
This follows from Lemma 4 of§ 3, no. 5, applied to the matrix with entries
BM(e,, Cs').
Let v = L v 5 e 5 be the vector satisfying the conditions of Lemma 2 and
such that L ~s = 1, and let A be the affine hyperplane of E* consisting of the
s
y* EE* such that (v,y*) = 1. JfT denotes the orthogonal complement ofv in
E*, A has a natural structure of an affine space with space of translations T.
Moreover, the form BM defines by passage to the quotient a non-degenerate
scalar product on E/E 0 , hence also on its dual T; this gives a e·udidean
structure on the affine space A (Algebra., Chap. IX, § 6, no. 6).
Let G be the subgroup of GL(E) consisting of the automorphisms leaving
v and BM invariant; if g E G, the contragredient map 1g- 1 leaves A and T
stable, and defines by restriction to A a displacement i(g) of A (cf. § 3). It is
immediate that this gives an isomorphism from G to the group of displace-
ments of A. Moreover, the stabiliser Ga. of a point a of A can be identified
with the orthogonal group of the Hilbert space T and is thus compact. On
the other hand, G is a locally compact group countable at infinity and A is
a Bail'e space: it follows (Integration, Chap. VII, Appendix, Lemma 2) that
the map ·tj; : g f-4 g(a) defines a homeomorphism from G/Ga to A. Thus G
acts properly on A (General Topology, Chap. III,§ 4, no. 2, Prop. 5). Since W
is a subgroup of G, it can be identified with a group of displacements of A.
We are going to show that this group satisfies the assumptions of § 3. More
precisely:
PROPOSITION 10. The group W provided with the discrete topology acts
properly on A; it is generated by orthogonal reflections; it is infinite, irre-
ducible and essential (§ 3, no. 7). The intersection C n A is a chamber of A
for W. If Ls denotes the hyperplane of A formed by the intersection of A
with the hyperplane of E* orthogonal to es, the Ls for s E S are the walls of
CnA. ff E., is the unit vector ofT orthogonal to L, on the same side of Ls as
C n A, then (cslc 1 ) = - cos(11'/m(s, t)) (for s, t ES) and the Coxeter matri.T
of W (§ 3, no. 4) is ident~fied w·ith M.
By Cor. 3 of Th. 1, W is discrete in GL(E), and hence in G, and acts
properly on A. Let s E S. Since Card S ~ 2, the hyperplane of E* orthogonal
to es is not orthogonal to v and its intersection L,. with A is indeed a hyper-
plane. The displacement corresponding to s is thus a displacement of order 2
leaving fixed all the points of Ls: it is necessarily the orthogonal reflection as-
sociated to Ls. It follows that W is generated by orthogonal reflections. Th. 2
now shows that it is infinite and Prop. 7 that it is e::;sential and irreducible.
Since C is an open sirnplicial cone, whose walls arc the hyperplane::; with
equations (x*, es) = 0 (for s E S), the intersection C n A is a convex, hence
connected, open and closed subset of the complement of the union of the
L, in A. Moreover, C n A is non-empty, for if x* E C we have (x*, v) =
Lv 8 (x*,e,) > 0 and (x\v)- 1:c* E CnA. It follows that CnA is a chamber
s
of A relative to the system of the L8 • l\foreover, w(C n A) n Ls = 0 for all
w E W (cf. no. 4, property (P n)) and it. follows that C n A is a chamber of A
relative to the system consisting of the transforms of the Ls by the elements
of W; by the Cor. of Th. 1 of§ 3, no. 2, it follows that C n A is a chamber of
A relative to W.
Let a;be the vertex of the simplex C n A not in L8 • We have
for s, t E S and s -/: t, and
Let Es be the vector in T defined by the relations:
(csla; - a.;)= v; 1 for t ES, ti- s.

The vector cs is orthogonal to Ls and is on the same side of the hyperplane

L,, as C n A. Moreover
which shows that cs is the image of the class of es under the isomorphism
from E/E0 to T given by the quadratic form BM. It follows that
Consequently cs is a unit vector and the last assertion of Prop. 10 is proved.

The euclidean affine space A provided with the group W is called the
space associated to the Co.reter matri:c IVI and we denote it by Atv1.
Proposition 10 admits a converse:
PROPOSITION 11. Let W be a gro·up of d·i.splacements of a eu.ci'idean affine

space A, satisfying the assumptions of§ 3. Assume th.at Wis infinite, essential
and irreducible. Th.en the form BM attached to the Co.1:eter matri:c M of
W is positive degenerate and there exists a unique isomorphism from the
affine space Arv! associated to M to A, commuting with the action of W. This
isomorph.ism transforms the scalar product of Arv! into a multiple of the scalar
product of A.
Let Co be a chamber of A and let S be the set of orthogonal reflections
with respect to the walls of C 0 . If ''Is denotes the unit vector orthogonal to
the hyperplane N,, associated to the reflection sand on the same side of N 8
as Co (§3, Prop. 3), the form BM is such that BM(e 5 , P.i) = (1'/sl17t) for s. t ES.
It is thus positive. Since the 7/s are linearly dependent (§ 3, no. 9, Prop. 8), it
is degenerate.
We can thus apply the preceding constructions to M. With the same
notation as above, (cslct) = (11sl''lt) and there exists a unique isomorphism cp
of Hi! be rt spaces from T to the space of translations of A such that cp( t: 8 ) = 178 .
Let a and b be two distinct vertices of Co and s 0 the reflection in S such that
a ¢ N 80 . Put ,\ = ( 17 8 la - b) and let 'lj1 be the affine bijection from AM to A
defined by
It is then immediate that 1f;(Ls) = Ns for all s E S and that 1/1 transforms
the scalar product of AM into a multiple of that on A. It follows at once that
1/J commutes with the action of W. Finally, the uniqueness of 1/J is evident,
for a 8 for example is the unique point of AM invariant under the reflections
t E S, t :/: s.
§ 5. INVARIANTS IN THE SYMMETRIC ALGEBRA
1. POINCARE SERIES OF GRADED ALGEBRAS
Let K be a commutative ring with unit element, not reduced to 0. Let M be

a graded K-module of type Z, and Mn the set of homogeneous elements of
M of degree n. Assume that each lVIn is free and of fin:ite type. Then the rank
rkK(Mn) is defined for all n (Commutative Algebra, Chap. II, §5, no. 3).
DEFINITION 1. If there exists no E Z s·uch that Mn = 0 for n :::;; no, the

formal series I: rkK (.tvln)Tn, which is an element of Q( (T)), is called the
n;?;no
Poincare series of M and denoted by PM (T).
Let M' be another graded K-module of type Z, and (M~)nEZ its grading.
Assume that M~ is zero for n less than a certain number. Then
(I)
and, if M @KM' is provided with the total grading (Algebra, Chap. II, § 11,
no. 5),
(2)
PROPOSITION I. Let S = EBn>O Sn be a commutative graded K-algebra

with a system of generators (:c 1 , ~' ... , Xm) consisting of homogeneous and
algebraically independent elements. Let di be the degree of Xi, and assume
that di > 0 for all i. Then the Sn are free and of finite rank over K, and
m
Ps(T)= fl(l-Td')- 1 . (3)
1.=l
Indeed, S can be identified with the tensor product K[xi]@ · · ·@ K[xmJ,

provided with the total grading. The Poincare series of K[xi] is
2:: Tnd, = (l _ Td;)-1,
n~O
and it suffices to apply (2).

Under the assumptions of Prop. 1, we shall say that S is a graded polyno-
mial algebra over K.
COROLLARY. The degrees di are determined v.p to order by S.

m
Indeed, the inverse of Ps(T) is the polynomial N(T) = IT (1-Td; ), which
i= 1
is thus uniquely determined. If q is an integer ~ 1 and if ( E C is a primitive
qth root of unity, the multiplicity of the root (of N(T) is equal to the number
§ 5. INVARIANTS IN THE: SYMMETRIC ALGEBRA 109
of di that are multiples of Q. This number is zero for q sufficiently large. The
number of di equal to Q is thus determined uniquely by descending induction.
The integers di are called the characteristic degrees of S. The number of
them is equal to the transcendence degree of S over K when K is a field; we
shall also call it the transcendence degree of S over K in the general case. It
is the multiplicity of the root 1 of the polynomial N(T).
Let S = EBn~o Sn be a commutative graded K-algebra, and R = EBn~o Rn
a graded subalgebra of S. Assume that each Rn is free and of finite type,
and that the R-modv.lc S admits a finite bas·is consisting of homogeneous
elements z 1 , z2, ... , ZN of degrees Ji, fz, ... , fN. Then, if M denotes the graded
N
K-module L Kzj,
j=l
the graded K-module Sis isomorphic to R®K M, so each
Sn is free and of finite type and
(4)
PROPOSITION 2. Retain the preceding notation and assume that S and R

are graded polynomial algebras.
(i) R and S have the same transcendence degree r over K.
(ii) Let p 1 , ... , Pr (resp. Qi, ... , Qr) be the characteristic degrees of S (resp.
R). Then
(iii) Np1p2 ... p,. = q1Q2 ... q, ..

Formula (4) shows first of all that the multiplicity of the root 1 is the
same in the polynomials Ps(Tt 1 and PR(T)- 1 and, taking (3) into account,
proves both (i) and (ii).
It follows from (ii) that
bl ( + 1 T + T2 + ... + Tq; -1) = Ct T'j) iDl ( 1+ T + T2 + ... + p1 -1).
Putting T = 1 in this relation gives (iii).
Remark. Let S = K[X 1 , ... , Xn) be a graded polynomial algebra over K, di

the degree of X, and F (X 1 , ... , Xn) a homogeneous element of degree m of
S. Then
(5)
Indeed, it is immediate that the K-linear map D from S to S that transforms

every homogeneous element z of degree p into pz is a derivation of S. Thus
" aF n aF
mF(X1, ... 'Xn) = D(F(X1, ... 'Xn)) = L l D(Xi) ax i
i=
= L1 d;,Xi ax i .
i=
2. INVARIANTS OF A FINITE LINEAR GROUP: MODULAR

PROPERTIES
Let K be a commutative ring, Va K-module, and Ga group acting on V. We

know that every automorphism of V extends uniquely to an automorphism
of the symmetric algebra S = S(V), and thus G acts on this algebra. If x ES
and g E G, we denote by gg.x the transform of x by g. Let R be the subalgebra
sG of s formed by the elements invariant under G.
Assume that G is finite, V is of finite type. and K is noetherian. Then S is
an R-module of finite type, and R is a K-algebra of finite type (Commutative
Algebra, Chap. V, § 1, no. 9, Th. 2). Assume that S is integral and let N be
its field of fractions. The field of fractions L of R is the set of elements of N
invariant under G (Loe. cit., Car. of Prop. 23), so N is a Galois extension of
L. Every element of N can be written z/t with z E S and t E R ( loc. cit.,
Prop. 23). By Algebra, Chap. II, §7, no. 10, Car. 3 of Prop. 26, the rank of
the R-module S is [N : L]. Assume that G acts faithfully on V. The Galois
group of N over L can then be identified with G, so [N : L] = Card G; thus
rkR(S) = [N : L] = Card(G). (6)
For any graded algebra A = Ao EB A 1 EB· · ·EB An EB·· ·, denote by A+ the ideal
EBn>O An·
THEOREJVI 1. Let K be a commutative field, V a finite dimensional vector

space over K, S = S(V) the symmetric algebra of V, G a finite gro'up of auto-
rnorph-isms of V, and R the graded subalgebra of S consisting of the elements
invariant under G. Assume that G is generated by pseudo-reflections (§ 2,
no. 1) and that q = Card( G) is coprime to the characteristic of K. Then the
R-modv.le S has a basis consisting of q homogeneous elements.
a) Since every submodule of S/(R+S) is free over Ro= K, it is enough to
show (in view of Algebra, Chap. Il, § 11, no. 4, Prop.' 7) that the canon-
ical homomorphism from R+ ©R S to S is injective. For any R-rnodule
E, denote by T(E) the R-module Ker(R+ ®R E _, E) (*in other words,
T(E) = Tor~(R/R1-iE).). If E,E' are two R-modules and u is a homomor-
phism from E to E', the homomorphism 1 ® u from R+ ® E to R+ 0 E'
defiues by restriction to T(E) a homomorphism from T(E) to T(E') that we
denote by T(u). If u' is a homomorphism from E' to an R-module E", we
have T(v.' o u) = T(u') o T(u). Thus, if G acts R-linearly on E, then G acts
on T(E).
§ 5. INVARIANTS IN THE SYlvIMETRIC ALGEBRA 111
b) The group G acts R-linearly on S, and hence also on T(S). Moreover,

T(S) has a natural structure of graded S-module. We show first that, if g E G,
then g transforms every element ;i; of T(S) into an element congruent to :z:
modulo S 1 T(S). It is enough to do this when g is a pseudo-reflection. Then
there exists a non-zero vector v in V such that g(x) - x E Kv for all :i: E V.
Since V generates S, it follows that gs acts trivially on S/Sv. Thus, for any
y ES, there exists an element h(y) in S such that
gs(y) - y = h(y)v.
Since S is integral and v is non-zero, this element is determined uniquely by
y; it is immediate that h is an endomorphism of degree -1 of the R-modulc
S. Thus, gs - ls = mv o h, where mv denotes the homothety with ratio v in
S. Hence,
T(.qs) - lT(S) = T(gs - ls)= T(mv) o T(h),
the image of which is contained in vT(S), proving our assertion.
c) We show that any element of T(S) invariant under G is zero. Indeed,
let Q be the endomorphism of the R-module S defined by
Q(y) = q- 1 L gs(Y)
gEG
for ally E S. Then Q(S) = R. We can thus write Q = i o Q', where Q' is a
homomorphism from the R-module S to the R-module Rand 'i denotes the
canonical injection of R into S. Thus T(Q) = T(i) o T(Q') and T(Q') = 0
since T(R) = Kcr(R+ 0 R --t H.) = 0. Thus
0 = T(Q) = q- 1 L
gEG
T(.qs).
But q- 1 L T(gs) leaves fixed the clements ofT(S) invariant under G. These
gEG
elements are therefore zero.
d) Assume that T(S) -=J 0. There exists in T(S) a homogeneous element
u -=J 0 of minimum degree. By b), u is invariant under G. By c), ·u = 0. This
is a contradiction, so T(S) = 0. Q.E.D.
Remarks. l) It follows from Algebra, Chap. II, §11, no. 4, Prop. 7 that, if
(z 1 , ..• , zq) is a family of homogeneous elements of S whose canonical images
in S/(R+S) form a basis of S/(R+S) over K, then (z 1 , ... , zq) is a basis of S
over R.
2) Let g be a pseudo-reflection of V, whose order n ~ 2 is finite and co-
prime to the characteristic of K By Maschke's theorem (Appendix, Prop. 2),
V can be decomposed as DEB H, where His the hyperplane consisting of the
elements of V invariant under g and D is a line on which g acts by multipli-
cation by a primitive nth root of unity. When K = R, this is possible only
when n = 2, and g is then a reflection; in this case, the groups to which Th. 1
applies are the finite Coxeter groups. (For K = C, on the other hand, Th. 1
applies to certain groups that are not Coxeter groups.) 3
THEOREM 2. Retain ihe assumptions and notation of Theorem l.

(i) There ex·ists a graded vector subspace of S forming a complement to
R+S in S and stable under G.
(ii) Let U be such a complement. The canonical homomorphism from
U 0K R to Sis an isomorphism of G-modules, and the representation of Gin
U (resp. S) is isomorphic to the regular representation of G on K (resp. R).
Indeed, for any integer n ~ 0, the K-vector spaces Sn and (R+S) n Sn
are stable under G, and it follows from Maschke's theorem (Appendix, Prop.
2) that there exists a G-stable complement U,, of (R+S) n Sn in Sn. Then
I: Un is a G-stable complement of R+S in S, hence (i).
n;?:O
Let U be a graded vector subspace of S forming a complement of R+S
in S and stable under G. By Remark 1, every basis of the K-vector space U
is also a basis of the R-module S, and consequently is a basis of the field of
fractions N of S over the field of fractions L of R. Thus, the L-vector space N
can be identified with U ®KL. Since U is stable under G, this identification is
compatible with the action of G. The group algebra L[GJ of Gover L can be
identified with the algebra K[G] ®KL. The Galois extension N of L admits a
normal basis (Algebra, Chap. V, § 10, no. 9, Th. 6), which can be interpreted
as saying that N, considered as an L[GJ-module, is isomorphic to the module
of the regular representation of G over L. Since U is finite dimensional over K,
it follows from the Appendix, Prop. 1, that the K[G]-module U is isomorphic
to the regular representation of G over K. Our assertions follow immediately
from this.
3. INVARIANTS OF A FINITE LINEAR GROUP:

RING-THEORETIC PROPERTIES
THEOREM 3. We retain the assumptions and notation of Th. 1. In the set

of systems of genera.tors of the ideal R+ of R consist·ing of homogeneous
elements, choose a minimal element (a 1 , ... , 01). Let ki be the degree of ll'i.
Assume that the ki are coprime to the characteristic exponent of K. Then
l = dim V, the a;, generate the K-algebra R, and are algebra·ically independent
over K. In partic1;,lar, Risa graded K-a.lgebra of polynomials of transcendence
degree l over K.
The assumption made about the k; is superfluous, but is irrelevant for the
applications to finite Coxeter groups, since then K = R. Cf. no. 5, where we shall
give another proof of Th. 3.
3 A classification of such groups can be found in G. C. SHEPHARD and J. A.

TODD, Finite unitary reflection groups, Canadian J. of Maths., Vol. VI (1954),
p. 274-304.
§ 5. INVARIANTS IN THE SYMMETRIC ALGEBRA 113
Th. 3 follows from Prop. 2 (i), Th. 1 and the following lemma:
Lemma 1. Let K be a com.mutative field, S a graded K-algebra of polynomials,

and R a graded subalgebra of S of finite type such that the H.-module S admits
a basis (zA)>-Eil consisting of homogeneous elements. In the set of systems of
generators of the ideal R+ of R consisting of homogeneous elements, choose
a minimal element ( o: 1 , ... , as). Assume that, for o.ll i, the degree k; of o:,
is coprime to the characteristic exponent p of K. Then the a; generate the
K-algebra R and are algebraically independent over K.
By Algebra, Chap. II,§ 11, no. 4, Prop. 7, the assumption made about the
a; is equivalent to saying that they are homogeneous and that their images
in the K-vector space R+/(R+) 2 form a basis of this space. This condition
is invariant under extension of the base field; we can t.hus reduce to the case
where the latter is perfect.
The family (a 1 , ... , o:,) generates the algebra R by Commutative Algcbm
Chap. III, § 1, no. 2, Prop. 1. We argue by contradiction and assume that this
family is not algebraically independent over K.
1) We show first of all that there exist families
(/J;)1(i(s1 (yk)!(k(r, (d;k)J(i(s, l(k(1
of hornogeueous elements of S with the following properties:
/31 E R for all i, and the (3; are not all zero; (7)
deg Yk > 0 for all k: (8)
r
a; = I:
k=I
d;kYk for all i; (9)
s
L /3;d,k
i=l
= 0 for all k. (IO)
Let X 1 , ... , Xs be indcterminates and give K[X 1 , ... , X,] with the graded
algebra structure for which X, has degree k;. There exist non-zero homoge-
neous elements H(X1, ... ,Xs) of K[X 1 , .. .,Xs] such that H(a1,. .. ,a 5 ) =
O; choose H to be of minimum degree. If 8H/8X; f= 0, the polynomial
J:;(, (a 1 , ... , as) is a non-zero homogeneous element of R: ir p f= l, H is not
of the form Hi with Hi E K[X1, ... , Xs]· Take
8H
/3i = ki fJXi (a1, ... , as)·
Since K is perfect, the polynomials DH/ 8X, E K[X 1 , ... , X5 ] are not all zero
(Algebra, Chap. V, § 1, no. 3, Prop. 4); in view of the assumption made about
the k;, neither are the /Ji.
On the other hand, S can be identified with a graded algebra of poly-
nomials K[x 1 , ... , Xr] for appropriate indet.erminates .r 1 , ..• , Xr with suitable
degrees m; > 0. Let D1c be the partial derivative with respect to .Tk on S. Take
dik = ki 1 Dk(O!i). Then the equality (J.O) holds because its left-hand side is
Dk(H(0!1, ... , 0! 8 )). On the other hand, if we put Yi = 1n1X1, ... , Yr= m,.:rr,
the equality (9) follows from the equality (5) of no. 1.
2) Let b be the ideal of R generated by the /3i; there exists a subset J of
l={l,2, ... ,s}
such that (/3i)iEJ is a minimal system of generators of b. Then J i 0 since

b I 0. We shall deduce from (9) and ( 10) that, if i E J, a; is an R-linear
combination of the ll'J for j i= i, which will contradict the minimality of
(cx1, ... , O!s) and will complete the proof.
There exist homogeneous elements "/ji of R (i E J, j E I - J) such that
{3j = L
iEJ
"/jif3i
.
(j E I - J). (11)
Taking (11) into account, formula (10) gives
L. f3i
tE.l
(dik L. "/j·idjk) = o.
+ JE!.l (12)
Put
(13)
Thus
L
iEJ
f3iUik = 0. (14)
Write Uik = L Oik>..Z>.., where the O;k>.. belong to R. Relation (14) implies that
A.EA
L
iEJ
{3.;0;u = 0 for all k and A. If one of the 0-;k>.. had a non-zero homogeneous
component of degree 0, the preceding equality would imply that one of the /3;
(i E J) is an R-linear combination of the others, contradicting the minimality
of (/3;);EJ. Thus O;k>.. E R+ and consequently 'Uik E R+S for all i and k:. Thus,
s
there exist 'Uikh ES such that u;1c = L V-ikhO'h, in other words, by (13),
h=l
s
d;k . L "/jidjk = h=l
+ .JEl-.J L 'Uikho:h. (15).i.k
Multiply both sides of ( 15 );k by Yk and add for i in J fixed and k = 1, 2, ... , r;
in view of (9), we find that
s ,.
Ct'; L 'YjiO'j. = h=l
+ jEl-J L k=l
L UikhYkO'h·
Take the homogeneous components of degree k; of both sides. Since deg Yk >
0, O'; is an S-Jinear combination of the CXJ for .i ¥ i. Since S is free over Rand
0! 1 , ... , a 8 E R, it follows that a.; is an R-linear combination of the <l'.j with
j I i (Commutative Algebra, Chap. I, § 3, no. 5, Prop. 9 d) ).
COROLLARY. With the assumptions and notation of Th. 3, the product of

the characteristic degrees of R is Card(G).
Indeed, rkR(S) = Card(G) (formula (6), no. 2). The characteristic degrees
of S are equal to 1. Then the corollary follows from no. 1, Prop. 2 (iii).
Lemma 2. Let K be a commutative field, V a finite dimensional vector space

over K, S = EBn;;,o S11 the symmetric algebra of V, s an endomorphism of
V, and s(n) the canonical extension of s to Sn. Then, with T denoting an
indeterminate, we have in K[[T]]
00
L Tr(s(n))Tn = (dct(l - sT))- 1 .

-n.==0
Extending the base field if necessary, we can assume that K is algebraically

closed. Let (e 1 , .•. , er) be a basis of V with respect to which the matrix of s
is lower triangular, and let ,\ 1 , ... , Ar be the diagonal elements of this matrix.
. l1 respect to t lle b·cl.SIS
W 1t . · (ei(l) i(r)) i(I)+···+i(r)=n o f Sn, or·d ere d Iex1co-
.•. er .
1
graphically, the matrix of s(n) is lower triangular and its diagonal elements
,·i.(rJ Th
a.re t he pro d ucts A,;(1)
1 ... A.,. . . us
Tr(s(n)) = L A~(!) ... A~(r),

i(l)+· .. +i(r)=n
and consequently
r~O Tr(s(n))Tn = (.,~0 ,\~Tn) C~0 ,\~T") ... C,~0 ,\~Tn)

= (1 - -\1T)- 1 (1- -\2T)- 1 ... (1 - ArT)- 1
= (det(l - sT))- 1 .
Lemma 3. Let K, V and S be as in Lemma 2, G a finite gro-ap of automor-

phisms of V, q the order of G, and R the graded subalgebra of S consisting of
the elements invariant under G. Assume that K is of characteristfr 0. Then
the Poincare series of R is
q- 1 L (det(l - gT))- 1 .
gEG
Indeed, the endomorphism f = q- 1 L g(n) is a projection of S11 onto Rn,

gEG
so Tr(f) = dimKS~. Thus, the Poincare series of R is
and it suffices to apply Lemma 2.

PROPOSITION 3. With the assumptions and notation of Th. 3, Let H be the

set of pse·udo-refiect·ions belonging to G and distinct from 1. Assume that K
l
·is of characteristic 0. Then Card(H) =L (ki - 1).
i.=l
By Prop. 3 of the Appendix, we can assume that K is algebraically closed.
For any g E G, let )q (g),, , , , ..\1 (g) be its eigenvalues. Since every g E G is
diagonalizable (Appendix, Prop. 2), g = 1 if and only if all of the ,\i(g) are
equal to 1, and g E H if and only if the number of ,\i(g) equal to 1 is l - 1
(we then denote by ..\(g) the eigenvalue distinct from 1). Prop. 1 of no. 1 and
Lemma 3 prove that
l
q TI (1 -
·i=l
Tk, r l =
gEG
L (det(l - gT))- 1 (16)
in K[[T]], and hence in K(T). Consequently, we have in K(T)

(l-T)1-1 1 (l-T)1-1
q
l
=--+2=
1-T gEHl-..\(g)T
+ 2.::
g#l,gitHdet(l-gT)
,
I1 (1 - Tk,)
·1.=l
which can be written

l
q- TI (1 + T + ... + Tki - 1)
·1:=!
l
( 1 - T) TI (1 + T + ... + Tk, -1) (17)
i= l
1 (1 _ T)t-1
2.::
= gEH 1 - ..\(g)T
2.:: det(l
+ 9#1.g~H - gT)
l
It follows that q - TI(l + T + ··· + Tk'- 1 )
i= I
vanishes for T = 1, so
q = k1 k2 ... ki, which we knew already by the Cor. of Th. 3. This granted, let
l
Q(T) be the polynomial (1-T)- 1 ( q- .TI (1 + T + · · ·+ Tk, - 1 )). Differentiating
•=l
l
the equality (1 - T)Q(T) .fl (1 + T + · · · + T"''-
= q - 1=1 .
1) and putting T = 1,
we see that -Q(l) is the value for T = 1 of
l
= -L ((1+2T+···+(ki-l)Tk'- 2 ) TI(l+T+···+Tkj- 1 ))
i=l .i¥i
so
Q(l) = t
i=I
(ki - l)ki flk.i
2 rtfi
= (h kj) (t 1).
J=l i=l
k; -
2
Returning to ( 17), we have on the other hand
Q(1) = ( TI
l
j=l
kj ) ( I::
gEH
1
1 - A(g)
) .
Thus,
l
I:: ki - 1 = I:: 1 ( 18)
i=l 2 gEtl 1 - >..(g)
Now the elements of G that leave fixed the points of a given hyperplane leave
stable a complementary line of the hyperplane (Appendix, Prop. 2), and so
form a cyclic subgroup G' of G by Algebra, Chap. V, § 11, no. 1, Th. 1. Let
t be the order of G'; the values of >..(g) for g E G' are B, 02 , ... , et-i with e a
primitive tth root of unity. We have 1 ~ 0 ; + = 1, so l-b•-·
I:: gEG',g#l
l-l( ) -
9
~(t - 1) = !Card(H n G'). The equality (18) thus proves the proposition.
Rem.ark. When K = R, G is a Coxeter group and H is the set of reflec-

tions belonging to G, we know (§ 3) that the elements of H are in one-to-one
correspondence with the walls of V.
PROPOSITION 4. With the assumptions and notation of Th. 3, assume that

K is of characteristic i= 2. Then -1 E G if and only if the characteristic
degrees k 1 , ... , kt of R are all even.
Let f be the automorphism of the algebra S that extends the automor-
phism -1 of V. Then f (z) = ( -1 )degz z for all homogeneous z in S. Thus, if
-1 E G, every homogeneous element of odd degree of R is zero, and the ki
are all even. Conversely, if the k; are all even, every element of R is invariant
under f, and Galois theory shows that -1 E G.
4. ANTI-INVARIANT ELEMENTS
We retain the assumptions and notation of Th. 3, and assume that K is of

characteristic 0. An element z of S is said to be anti-invariant under G if
g(z) = (detg)- 1 z
for all g E G.
Let H be the set of pseudo-reflections belonging to G and distinct from
1. For all g E H, there exist e9 E V and f 9 E V* such that
g(:r) = x + f 9 (x)e 9 for all x E V.

PROPOSITION 5. (i) Let D be the element fl e9 of S. The elements of S

gEH
which are anti-·invariant under G are the elements of RD.
(ii) Identify S with the polynomial algebra K[X 1 , ... ,Xi] by choosing a
basis (X 1 , ... , X1) of V, and let (P1, ... , P1) be algebraically independent ho-
mogeneous elements of S generating the algebra R (Th. 3). Then the jacobian
J = det ( g~;) is of the form >.D, where >. E K*.
a) With the notation in (ii), we have
so, for all g E G,
g(J)( det g) dX1 I\ ... I\ dX1 = g(.J) d(gX 1 ) I\ ... I\ d(gXi)

= g(dP1 I\ ... I\ dP1) = dP1 I\ ... I\ dP1 = J dX 1 I\ ... I\ dX1,
hence J is anti-invariant under G. On the other hand, the field of fractions

N of S is a Galois extension of the field of fractions E of R (no. 2); if ,1 is
a derivation of E with values in an extension field f? of N, Ll extends to a
derivation of N with values in f? (Algebra, Chap. V, § 16. no. 4, Th. 3); since
the Pi are algebraically independent, it follows that dP 1 I\ . .. I\ dP 1 =f 0, hence
J =I 0.
b) Let z be an element of S anti-invariant under G. We show that z is
divisible by D in S. Let a be a. non-zero vector in V. The elements of G
that ieave Ka. stable leave stable a complementary hyperplane L (Appendix,
Prop. 2); an element of G leaving Ka. stable is 1 or a pseudo-reflection with
vector a if and only if it. induces 1 on L; the pseudo-reflections with vector a
belonging to G thus constitute, together with 1, a cyclic subgroup G' of G;
let t be its order. There exists a basis (X 1, ... , X1) of V such that a. = X 1 ,
X2 EL, ... , X1 EL, and z can be identified with a polynomial P(X 1 , ... , X1)
with coefficients in K. From g( z) = (detg )- 1 z for g E G', we see that X 1 only
appears in P(X 1 , ... , XL) with exponents congruent to -1 modulo t. Thus,
P(X 1 , ... ,X1) is divisible by x;-
1 = o.i- 1 . Now Dis, up to a scalar factor,
the product of the at-l for those a E V such that t > l, and these elements
of S are mutually coprime. Since S is factorial, z is divisible by D.
c) By a) and b), J is divisible by Din S. Now
l
deg J = L (k.;
i=l.
- 1) = Card(H)
(Prop. 3), so clegJ = degD, hence J = >.D with>. EK. Since J =f. 0, >.EK*.
This proves (ii).
d) Parts a) and c) of the proof show that D is anti-invariant under G.
Next, if y E R, it is clear that yD is anti-invariant under G. Finally, if z E S
is anti-invariant under G, we have seen in b) that there exists y ES such tha.t
z = yD. Since Sis integral, y E R.. This proves (i).
§ 5. INVARIANTS JN THE SYMMETRIC ALGEBRA ll9
* 5. COMPLEMENTS 4
Lemma 4. Let K be a comm1ttative .field, V a finite dimens'ional vector space

over K, G a finite group of automorphisms of V whose order q is invertible
in K, S the symmetric algebra of V, and R the subalgebro. of S consisting of
the elements invariant under G. A prime ideal B of he'ight 1 of S is ram~fied
over p = B () R (Commv,ta,tive Algebra) ~f' and only if there exist a non-zero
element a of V and a non-zero element f of V* such that B = Sa and the
psev,do-reflection Sa,J belongs to G. The decomposition group gz (B) is then
the subgroup of elements of G leaving Kn stable, and the inertia group 9T (B)
is the cyclic subgroup Ha of G cons'isting of the pseudo-reflect.ions of G with
vector a. The res'idue field S(B) of S at B is separable over the resid'ae .field
R(p) of R at p, and the ramification index e(B/p), equal to the coefficient
of B, augmented by 1, in the divisor div(Ds;R) of the different, is equal to
Card(Ha)·
To say that B is ramified over R means that its inertia group gT (B) does
not reduce to the identity, in other words that there exists g i 1 in G such
that g(z) = z (mod. B) for all z E S. Since S is a factorial ring, B is a
principal ideal Sa, and a must divide all the elements g(z) - z (z ES); now,
for z E V, these clements are homogeneous of degree 1 and are all non-zero
(since g i l); thus, a, must be homogeneous of degree Lin other words a E V.
Then there exists a linear form f on V such that g = sa.J, Conversely, if g is a.
pseudo-reflection s 0 ,,r different from 1, then g(z) = z (mod. Sa) for all z ES,
so g belongs to the inertia group of the prime ideal B = Sa. This proves the
first assertion of the lemma and the characterizations of Q2 (B) and gT (B).
Since q is coprime to the characteristic p of K (which is also that of S(B)),
the extension S(B) of R(p) is separable (Commutative Algebra, Chap. V, §2,
no. 2, Cor. of Prop. 5). The equality e(B/p) = Card(H 0 ) follows from this
(Commutative Algebra). Since e(B/p) is coprime top, the coefficient of B in
div(Ds;R) is e(B/p) - 1 ( Commutal'ive Algebra). This proves the lemma.
Lemma 5. Let K be a commutative field, S a graded polynomial algebra over

K, and Ra graded subalgebra ofS. Then Sis a.free graded R-module (Algebra,
Chap. II, § 11, no. 2) ~f and only ·~f the following two conditions a.re sa.tis,fi,ed:
a) R is a graded polynomial algebra over K;
b) if (0:' 1 , ... , ll:'s) is a system of generators of the K-a.lgebra R consisting of
nlgebraically independent homogeneous elements, this system is an S-regular
sequence5 .
4 In this number, we use results from chapters in preparation in ihe book on

Commutative Algebra. We indicate them by the symbol "Commutative Algebra".
5 This means that, for all i E { 1, 2, ... , s}, the canonical image of a:; in the ring
S/(SC\'1 + · · · + Sa,-iJ
·is not a zero divisor in this ring.
When S is an R-module of .finite type, b) is a consequence of a).

See Commutative Algebra for the proof.
THEOREM 4. Let K be a commutative field, V a finite dimensional vector

space over K, S the symmetric algebra of V, G a finite group of automor-
phisms of V, and R the subalgebra of S consisting of the elements ·invariant
under G. Assume that Q = Card G is invertible in K. The following condit'ions
are equivalent:
(i) G is generated by pseudo-reflections;
(ii) S is a free graded R-module;
(iii) R is a graded polynomial algebra over K.
The equivalence of (ii) and (iii) follows from no. 2 and Lemma 5. The
implication (i) ===? (ii) follows from Th. 1.
We show that (iii) ===? (i). Let G' be the subgroup of G generated by
the pseudo-reflections belonging to G, and let R' be the subalgebra of S
consisting of the elements invariant under G'. We have R C R' C S. By
Lemma 4, div(Ds;R) = cliv(Ds/R' ), so div(DR' /R) = 0. Assume then that R
is a graded polynomial algebra. Since this is also the case for R' (since G'
is generated by pseudo-reflections), Lemma 5 shows that the R.-module R'
admits a homogeneous basis (Q 1 , ... , Qm); let Qi= deg( Qi)· Put
d = det(TrR'/R(Q;Qj)), cf. Algebra, Chap. IX, §2.
The fact that div(DR' /R) is zero shows that div(d) = 0 ( Commv.tative Alge-
bra), which means that d belongs to K*. On the other hand TrR';R(Q;Qj)
is a homogeneous element of degree Qi + Qj, and d is homogeneous of degree
2 L Qi· Thus, L Qi = 0, i.e. qi = 0 for all i, which means that R' = R, and
i 1:
so G' = G by Galois theory. This proves that G is generated by pseudo-
reflections. Q.E.D.
Remarks. 1) Under the assumptions of Th. 4, the product of the character-

istic degrees of R is Q (formula (6) and Prop. 2 (iii)), so they are coprime to
the characteristic of K. This was asserted in no. 3.
2) If we do not assume that Card(G) is invertible in K, we still have the
implications (ii) {:::==> (iii) (cf. Lemma 5) and (ii) ===? (i) (cf. Exerc. 8); but
the implication (i) ===? (ii) is no longer true (Exerc. 9).
PROPOSITION 6. The assumptions and notation a.re those of Th. 4. Let H

be the set of pse-udo-reftections belonging to G a.nd distinct .from 1. Assume
that H generates G. For all g E G, put g(:r:) = :z: + f 9 (x)e 9 with e9 E V,
j~EV*.Put
D= CT e.9
gEH
ES.
(i) The different of S over R is the principal ideal SD.

§ 6. COXETER TRANSFORMATION 121
(ii) Identify S with the algebra K [X 1 , ... , Xi] by choosing a bas·is ( X1 , ... , X1)
of V, and let P 1 , ... , P1 be algebraically independent homogeneous elements
generating the algebra R. Then the jacobian J = clet ( g~~) is of the form >.D
where ,\ E K*.
l
(iii) L (deg(P;) - 1) = Card(H).
i=l
(iv) The set of anti-invariant elements of S is RD.
Assertion (i) follows from Lemma 4. Assertion (ii) follows from the fact
that SJ is the different of S over R. (Commutative Algebra.). Assertion (iii) is
obtained by writing down the fact that the homogeneous polynomials D and
J are of the same degree. The proof of (iv) is then the same as that given in
no. 4 (proof of Prop. 5, parts b) and d)).*
§ 6. COXETER TRANSFORMATION
In this para.graph, V denotes a real vector space of finite dimension l and W

denotes a finite subgroup of GL(V), generated by reflections and essential
(§ 3, no. 7). Provide V with a scalar product (xjy) invariant under W. Denote
by .lj the set of hyperplanes H of V ~mch that the corresponding orthogonal
reflection Stt belongs to W.
1. DEFINITION OF COXETER TRANSFORMATIONS
An ordered chamber relative to W is a pair consisting of a chamber C deter-

mined by .lj and a bijection i >-> H; from { 1, 2, ... , l} onto the set of walls of
C (cf. § 3, no. 9, Prop. 7).
DEFINITION L The Coxeter transformation determined by an ordered cha-

mber (C, (Hih~i~t) is the element c = stt, Stt 2 ••• Stt 1 of W.
PROPOSITION I. All Coxeter transformations are conjv.gate in W.

Since W permutes the chambers determined by .lj transitively (§ 3, no. 2,
Th. 1), we are reduced to proving the following: let (C, (H;) 1 ~;~i) be an
ordered chamber and rr E S1; then Stt, SH 2 ••• Stt 1 and Stt,,(l) SH,,, 2 , ••• sH,,ci>
are conjugate in W. Taking into account § 4, no. 8, Prop. 8, this will follow
immediately from the following lemma:
Lemma 1. Let X be a finite forest, and x >-> gx a map fmm X to a group I'
such tha.t 9x and 9y are conjugate whenever x and y are not linked in X. Let
T be the set of total orderings on X. For all ~ E T, let Pt, be the product in
I' of the sequence (gx )xEX defined by ~. Then the elements P<, are conjugate
. in r.
l) We proceed by induction on n =Card X. The case n = 1 is immediate,

so assume that n ;;:: 2. There exists in X a terminal vertex a (Chap. IV,
Appendix, no. 3, Prop. 2). Let b E X-{a} be a vertex linked to a if one
exists; if a is not linked to any vertex in X- {a}, let b in X - {a} be arbitrary.
In all cases, 9a commutes with 9x for x =/:- b. Let 'f) E T be such that a is the
largest element of X and b the largest element of X - {a}; we let ~ E T and
prove that PF., PTJ are conjugate.
2) Assume first that, for~' a is the largest element of X and b the largest
element of X- {a}. Let X' be the full subgraph X- {a}, .which is a forest.
Define a map x ...._. g~ from X' to I' by putting g~ = 9x if x =/:- b, 9~ = 9b9a·
Let ~', 'f/ 1 be the restrictions of ( 'f) to X'. The induction hypothesis applies,
so Pe and PTJ' are conjugate. But it is clear that Pc =Pt;., PTJ' = p,,, proving
the lemma in this case.
3) As:mme that a is the largest element of X for~- Let X 1 (resp. X 2 ) be.
the set of elements of X - {a} strictly larger (resp. smaller) than b; let ~i be
the restriction of ~ to Xi. Then
Pt;.= Pt;.,9bPt;.29a = Pt;.19b9aPf.2>

and this element is conjugate to Pt;. 2Pt;.,9b9a· We are thus reduced to case 2).
4) In the general case, let X3 (resp. X4 ) be the set of elements of X
strictly larger (resp. smaller) than a; let ~i be the restriction of~ to Xi. Then
Pi:, = Pt;. 3 9,.Pt;.., and this element is conjugate to Pi:,.P(, 3 9a· We are thus reduced
to case 3).
It follows from Prop. 1 that all the Coxeter transformations are of the
same order h = h(W). This number is called the Coxeter number of W.
Remark. Let W 1 , ... , Wm be essential finite groups acting in the spaces

V 1 , ... , V m and generated by reflections. Let C1 be a chamber relative to W1 .
Let W be the group W 1 x · · · x Wm acting in the space V 1 x · · · x Vm. Then
C 1 x · · · x Cm is a chamber relative to W. The Coxeter transformations of W
defined by Care the products c 1 c2 ... cm, where c1 is a Coxeter transformation
of W 1 defined by Cj.
2. EIGENVALUES OF A COXETER TRANSFORMATION
Since all Coxeter transformations are conjugate (no. 1, Prop. 1), they all have
the same characteristic polynomial P(T). Leth be the Coxeter number of W.
Then
P(T) = .IT
J=l
(T - i7rmj) ,
exp 2
h
where m 1 , m 2 , ... , mi are integers such that
DEFINITION 2. The integers m 1 , m 2 , ... , m1 ar·e called the exponents oJW.
Let C be a chamber determined by Sj, H 1 , ... , H 1 its walls, and put

s; = SH,. Denote by e; the unit vector orthogonal to Hi and on the same
side of H; as C. By Prop. 2 of the A ppenclix of Chap. IV, we can assume
that the H; arc numbered so that e 1, e2, ... , e, arc pairwise orthogonal and
er+!, er+-2· ... , e1 are pairwise orthogonal. Thens'= s 1s 2 ... s,. is the orthog-
onal symmetry with respect to the subspace
s 11 = Sr+is 1 -r 2 ... s1 is the orthogonal symmetry with respect to the subspace
V11 = Hr+ n Hr+ 2 n · · · n H,,

1
and c =s's" is a Coxeter transformation. Since (e 1 , ... , el) is a basis of V, V

is the direct sum of V' and V".
We deduce first that 1 is not an eigenval'ue of c. For if x E V is such that
c(x) = x, then s'(x) = s"(x), so x - s'(.r) = .T - s"(.:r) is orthogonal to V'
and V11 , and hence is zero. Thus, x = s'(x) = s"(x) EV' n V" = {O}.
Consequently,
(1)
The characteristic polynomial of c has real coefficients. Thus, for all j, the
power o f T - exp-
2ir.m' ' . equaI to t h at o f T - exp 2irr(h-m
P(T) 1s ; ) . H ence
1-,_J m h.
mJ + rn, -r 1 _ j = h ( 1 ~ j ~ l). (2)

Adding the equ<llities (2) we obtain
1
mi +rn.z+··· +m1 = -lh. (3)
2
Lemma 2. Assu,rne that W is irreducible and that dim V ~ 2. With the pre-
ceding notation, there c:rist two linearly independent vector-.5 z', z" s·uch that
(i) the plane P generated by z', z" is stable ·under s' ands";
(ii) s'IP and s"IP are orthogonal r·efiections with respect to Rz' and Rz";
(iii) z', z" E C, and P n C is the set of linear combinations of z', z" with
coefficients > 0.
Let (e 1 , ... , e1 ) be the basis of V such that (e' leJ) = O;J' Then C is the open
simplicial cone determined by the ei (§ 3, no. 9, Prop. 7). It is clear that V' is
generated by er+ 1 , ... , e1 and V" by e 1, ... , er. Let q be the endomorphism of
V such that q(e 1 ) = e 1 , ... ,q(e1 ) = e1• Its matrix with respect to (e 1 , ... ,e 1)
is Q = ((e;lej)). We have (e;lej) ~ 0 for i of. j (§3, no. 4, Prop. 3). Since
W is irreducible, there does not exist any partition {l, 2, ... , i} =- 11 Uh
such that (e;leJ) = 0 for i E 11 and j E: h. Thus (§ 3, no. 5, Lemma 4) Q
has an eigenvector (a 1 , ... , al) all of whose coordinates are > O; let a be the
corresponding eigenvalue. Put
z=a1e 1 + .. ·+a1e1,
z" = a1e 1 + · · · + arer E V 11 n C,
z' = a 1 -+ 1 er+l + · · · + a1e 1 EV' n C,
and let P be the plane generated by z' and z". Then P n C is the set of
linear combinations of z' and z" with coefficients> 0. The relation q(z) = az
l l
gives L ajej = r=l
;=l
L aajeJ; scalar multiplying by ek (where k ~ r) gives
l
ak + . L ai(ej\ek) = aak; thus
J=·t+l
l l
= - L a e1 + j=rtl
j=r+l J
L a..J eJ
l
= -z' L a.·e·.
+ j=r+l .I J
Thus, (a.- 1 )z 11 + z' is orthogonal to e 1 , ... , e1·, that is, to V". Hence, s 11 leaves
stable the plane generated by z 11 and (a - l)z 11 + z', that is, P. Similarly, s'
leaves P stable. Since z' E P n V' and z 11 E P n V", s'IP and s"IP are the
reflections with respect to Rz' and Rz 11 •
THEOREM 1. Assume that W is 'irreduci.ble. Then:

(i) m 1 = 1, m1 = h - 1.
(ii) Card(jj) = ~lh.
We retain the prece~g notation. The restriction of c = s' s 11 to P is the
rotation with angle 2(z",z') (§2, no. 5, Cor. of Prop. 6). Since c is of or-
der h, the h elements 1, c, ... , ch-l of W are pairwise distinct; the elements
s', s' c, ... , s' ch- l are thus pairwise distinct, and are distinct from the preced-
ing elements since ci IP is a rotation and s' cJ IP is a reflection. The set
{l , c, ... ' ch-1 s I ' s I c, ... ' s I ch-1}
1
is the subgroup W' of W generated by s' and s", and induces on P the group
W" generated by the orthogonal reflections with respect to Rz', Rz 11 • Tbe
transform of C by an element of W' is either disjoint from -C or equal to
-C. Thus, the transform of P n C by an element of W" is either disjoint
from -(P n C) or equal to -(P n C). Hence~r a suitable orientation of

P, there exists an integer m > 0 such that (z", z') = ~ (§ 2, no. 5, Cor. of
Prop. 7). Moreover, the sets g'(C), for g' E W', are pairwise &;joint; the
sets g"(P n C), for g" E W", are thus pairwise disjoint; so W" is of order
2h. Hence m = h. By definition, cJP is a rotation with angle ;~, and so has
eigenvalues exp 2 h.,,.,exp 2 i"(~-I). This proves that m 1 = l,m1 = h- l.
The transforms of Rz' and Rz" by W' are h lines D 1 , ... , D1i of P, and
the points of P- (D 1 U· · ·UD1i) are transforms by the elements of W' of points
of P n C. Thus, a hyperplane of 5j necessarily cuts P along one of the lines
Di, and consequently is a transform by an operation of W' of a hyperplane
of Sj containing Rz' or Rz".
Now, any HE Sj that contains Rz' is one of the hyperplanes H 1 , ... , Hr.
Indeed, let CH be the unit vector orthogonal to H and on the same side of H
as C. Then CH = >. 1 e 1 + · · · + >.1e1 with the A; all ?': 0 (§ 3, no. 5, Lemma 6,
(i)). Now 0 = (cHlz') = A1·+1ar+1 + · · · + A1a1, so
Ar+l = ... =A/= 0, and CH = A1C1 + ... + ArC,..
Suppose that two of the A; were non-zero, for example ,\ 1 and >. 2 ; since
e 1 , ... , er are pairwise orthogonal, we would have
si(eH) = -A1e1 + /\2e2 + · · · + ArCr

and the coordinates of s 1 (eH) would not be of the same sign, which is absurd
(Loe. cit.). Thus, CH is proportional to one of the vectors e 1 , .. .,e,., which
proves our assertion. Similarly, any H E Sj that contains Rz" is one of the
hyperplanes Hr+ 1, ... , H1.
The number of elements off) containing Rz' or Rz" is therefore l. If h is
even, Card(.fJ) is thus equal to ~l. If his odd, Card(f:>) is equal to"; 1 l+r, and
also to "; 1 l + (l - r); hence r = l -r, so r = &. and Card(Sj) = h;
1 l + ~ = ~l.
Remark. Retain the notation of the preceding proof. Let c' be the C-linear
extension of c to V ®a C, and c" the restriction of c' to P ©a C. From our
study of cJP, c" has an eigenvector x corresponding to the eigenvalue exp 2::',
and this eigenvector does not belong to any of the sets D ®a C, where D
denotes a line of P (since D is not stable under c). Now, for any H E Sj, we
have seen that H n P is a line; thus, :r iJ/. H ®a C.
COROLLARY. Let Ru be the set of ·an-it vectors of V orthogonal to an element

of Sj, ff W is irreducible,
L (xlu) 2 = h(xlx) (4)
·u.ERo
for all x EV.

Put f(x) = L (:riu) 2 . It is clear that f is a positive quadratic form
uERo
invariant under W, and non-degenerate since the Ci form a basis of V. Since
Wis irreducible, there exists a constant (3 such that f(x) = (J(xlx) (§ 2, no. 1,
Prop. l). If (xi)i~i~I is an orthonormal basis of V for the scalar product (.rly),
then
l l l
(Jl = I:: ,B(xilxi) = I:: f(xi) = i=l
I:: uERo
I:: (x;lu) 2
i=l. i=l
= I:: 1 = Card(Ro) = 2Card(f:J) = hl.
uERo
Thus (3 = h, which proves (4).
PROPOSITION 2. If W is irreducible and h is even, the unique element of

W that transforms C to -C is ch/ 2.
We use the notation in the proof of Th. 1. Since clP is a rotation through
an angle 2;:_, ch/ 2 transforms z' to -z', z" to - z", and hence z' + z" = z to
-z. Now z EC, so the chamber chl 2(C) is necessarily -C.
PROPOSITION 3. Assume that W is ·irreducible. Let ·u 1 , ... , ·u1 be homoge-

neous elements of the symmetric algebra S = S(V), algebraically independent
over R and generating the algebra of elements of S invariant under W (§ 5,
no. 3, Th. 3). If Pi is the degree of Uj, the e:ivonents of W are Pi -1, ... , PL -1.
Put V' = V 0R C, S' = S(V') = S 0R C, and extend the scalar product
on V to a hermitian form on V'. If c is a Coxeter transformation of W, there
exists an orthonormal basis (Xi)i~i~I of V' consisting of eigenvectors of c0 l
(Algebra, Chap. IX, § 7, no. 3, Prop. 4); moreover, we can assume that, for
1 ::;;; j ::;;; l, X 1 corresponds to the eigenvalue exp 2 i~;"J of c@ l. It is clear that
S' can be identified with the algebra C[X 1 ,. .. , X1], and that we can write
Uj ~ 1 = fj (X 1 , ... , X1), where fj is a homogeneous polynomial of degree
Pim C[X1, .. .,Xi]. Put Dj = axj and J = det(Dkfi). Recall (§5, no. 4,
Prop. 5) that J(X 1 , ... , Xt) is proportional to the product in S' of Card(Sj)
vectors Yk of Veach of which is orthogonal to a hyperplane off). Since we can
assume that X 1 t/. H 0 C for all HE f:J (Remark), the X 1 component of each
of the vectors Yk is non-zero, so J(l, 0, ... , 0) :f:. 0. The rule for expanding
determinants now proves the existence of a permutation o of {I, 2, ... , l}
such that (Dcr(j)fj)(l,O,. . .,0) :f:. 0 for all j. Since Dcr(.i).fj is homogeneous
of degree Pi - 1, the coefficient of X~ 1 - 1 Xa(;l in fj(X 1 , •.• , X1) is non-zero.
Now fj (X 1 , ... , X1) is invariant under c ® 1, and ·
(c ® 1) ( X Pi
1
-1. )
Xa(j) =
( 2i 7r ( PJ - 1 + ma(J) ) )
exph (
X p11 - l )
Xa(j) .
This proves that PJ - 1 + ma(j) = 0 (mod. h). Now h - ma(j) is an exponent

(formula (2)). Permuting the Uj if necessary, we can assume that Pj -1 =mi
(mod. h) for all j. Since Pi - 1 ~ 0 and m] < h, we have P.i - 1 =mi+ µih
with µ_ 1 an integer ~ 0. By § 5, Prop. 3, we see that
l l l
Card(SJ) =L (pj - 1) = L m 1 + h .i=l
L µJ·
j=l .i==l
l
Taking into account formula (3) and Th. l (ii), we obtain h L µ,j = 0, so
j=l
µJ = 0 for all j, and finally PJ - 1 = m1 for all j.
COROLLARY 1. If (mi)u;;,i;o;;l is the increasing seq1tence of e:z:ponents of W,

the order of Wis equal to (m 1 +l)(m2+1) ... (m.1 + 1).
This follows from the relations m 1 + 1 =Pi and § 5, Cor. of Th. 3.
COROLLARY 2. If c is a Co:ceter transformation of W,
exp (-2i11)
h
and exp (--2i1T)
h
-
are eigenvalues of c of multiplicity l,

Otherwise, there would exist two non-proportional homogeneous invari-
ants of degree 2 in S, and hence two non-proportional quadratic forms on V*
invariant under W, contrary to§ 2, no. 1, Prop. l.
COROLLARY 3. The homothety with ratio -1 o.f V belongs to W ~f and only

if all the exponents of W are odd. In that case, h is even and ch/ 2 = -1 for
any Coxeter transformation c of W.
The first assertion follows from § 5, no. 3, Prop. 4. Assume that the ex-
ponents of W are odd. Then h is even by formula (2), and
'
2Z1Tm ·) h/2
( expT = exp(i'ifmJ) = -1;
thus c"/ 2 = -1 since c is a semi-simple automorphism of V (Algebra, Chap.

IX, § 7, no. 3, Prop. 4).
APPENDIX
COMPLEMENTS ON LINEAR
REPRESENTATIONS
The following proposition generalizes Prop. 13 of Chap. 1, § 3, no. 8.
PROPOSITION 1. Let K be a commu.tativc field, A a K-algebra, and V o.nd

W two left A-mod'ules that are .finite dimensional vector spaces over K. If
there exists an extension L of K such that the (A 0K L )-modules V 0K L and
W 0K L are isomorphic, then the A-modules V and W are isomorphic.
a) Assume first that L is an extension of K of .finite degree n. Since V 0K L
and W 0K Lare isomorphic as (A 0K L)-modules, they are isomorphic as A-
modules; but, as A-modules, they are isomorphic to yn and wn, respectively.
Now V and W are A-modules of finite length; thus V (resp. W) is the direct
sum of a family (iv1;·') 1 ;:;;i;:;;p (resp. (N;j)i;:;;j;:;;t) of submodules such that the
M; (resp. Nj) are indecomposable, and that two of the Mi (resp. N1) with
distinct indices are not isomorphic (Algebra, Chap, VIII, § 2, no. 2, Th. 1).
Then yn (resp. Vv'n) is the direct sum of the Mr' (resp. N7 5
'); it follows
(Loe. cit.) that p = q and, after a suitable permutation of the N_i, that M.; is
isomorphic to Ni and nri is equal to ns.; for 1 ~ 'i ~ p. Thus, V is isomorphic
to W.
I>) Assume that K is an infinite field. The assumption implies that V
and W have the same dimension over K. Let (ei) 1 ;:;;;;:;;m and (c~)i;:;;;;:;;m be
bases of V and W over K, and (a>.) a basis of A over K. An isomorphism
u : V 0K L --> W 0K L is a bijective L-linear map and at the same time an
(A 0K L)-homomorphism, in other words it satisfies the conditions
a>,u(e;) = u(a>,e;) for all,\ and all i. (1)
Put O.>.Ci = ~r>.'iie:1, a>,e~ = ~~f~ijej, where the r>.ii and "Y~ij belong to K,
J J
and u(e;) = Lfoej, where the fo belong to L. The conditions (1) can be
:I
written
(2)
for all ,\, i, k. By assumption, the homogeneous linear equations (2) have a
solution (fo) E V 12 such that det( fo) i O. Since the coefficients of the
system (2) belong to K, we know (Algebra, Chap. II, § 8, no. 5, Prop. 6)
that this system also admits non-trivial solutions in Km''; let E be the vector
subspace of Mm (K) = Km 2 , not reduced to zero, formed by these solutions.

Let (c1)1~t~p be a basis ofE, and put (fo) = '2(r11c1 for any matrix (fo) EE;
then det(fo) is a polynomial P(1) 1 , .•. ,'fJp) with coefficients in K. On the
other hand, we know (lac. cit.) that the solutions of (2) in Lm 2 arc of the
form L (1c1, this time with (1 E L; for such a solution, det(fo) is equal to
I
P(( 1 , ... , (p)· Granting this, if we had P(77 1 , ... , 1]p) = 0 for all 1) 1 , .. . 1]p EK,
the coefficients of P would be zero since K is infinite; wc would then have
P(( 1 , ... , (p) = 0 for all ( 1 , ... , (p E L, which is contrary to our assumption.
We can therefore find a matrix (fo) E E such that det(fo) ¥ 0, and the
corresponding linear map V ~ W is an isomorphism.
c) General Case. Let fl be an algebraically closed extension of L, and Ko
the algebraic closure of K in fl. The a.5~:mmption implies that V ®1< fl and
W ®K fl are isomorphic (A ®K fl)-modules. Since Ko is infinite, part b) shows
that V ®K Ko and W ®K Ko are isomorphic (A ®K Ko)-modules. Retaining
the notation in b), the system (2) admits a solution (fo) E K 0' 2 such that
det(fo) ¥ 0. But the fo all belong to some algebraic extension K 1 of finite
degree over K. The (A®1<K 1 )-modules V ®K K 1 and W®K K 1 are isomorphic,
and the proof is completed by using a).
PROPOSITION 2 (Maschke). Let A be a ring with unit element, E o. left

A-mod·ule, F a direct summand of E, G a group of .finite order q, and p a
linear representation of G on E. Assume that q. l is invertible in A and that
F is stable ·under G. Then there exists a complement of F in E that is stable
under G.
Let p be a projection of E onto F. For all x E E, put
f(x) = q- 1 L p(s)- 1 p(p(s).x) .

.sEG
We have f(:r) E F and f(y) = y for all y E F, so .f is a projection of E onto

F. On the other hand, if t E G,
p(t)f(x) = q- 1 L p(sc 1 )- 1 p(p(s)x)

sEG
= q- 1 L p(s)- 1 p(p(st)x)
sEC
= f(p(t)x).
Thus f commutes with p(G), so Ker(f) is a complement of F stable under
G.
COROLLARY Let G be a .finite group of order q 1 and K a cornmutat'ive field

whose characteristic does not divide q. Then the group algebra of G O'uer K
is semi-simple.
Indeed, by Prop. 2, every module over this algebra is semi-simple.
APPENDIX COMPLEMENTS ON LINEAR REPRESENTATIONS 131
PROPOSITION 3. Let A be a commutative ring, M an A-module, G a finite

group acting on M, and A' an A-module. Assume that the order q of G
is invertible in A. Let MG be the set of elements of M invariant under G.
Then the canonical homomorphism from l\1G ©A A' to M ©A A' defines an
isomorphism from MG ©A A' onto the module (M ©A A')G of elements of
M ©A A' invariant under G.
Indeed, let Q be the projection of M onto MG defined by Q(x) =
q- 1 I: g( x) for all x E M. If i denotes the canonical injection of MG into
gEG
M, Q o i is the identity map of MG, so (Q © lA') o (i © lA') is the identity
map of MG ©A A'. Since Q © lA' = q- 1 I: (g © lA' ), the image of i © lA'
gEG
is (1\!I © A')c. On the other hand, i © lA' is injective by what has been said
before.
Remark. The preceding proposition applies in particular when A' is an A-

algebra. In that case, MG ©A A' is an A'-submodule of M ©A A'.
EXERCISES
§ 2.
1) Let K be a commutative ring with unit clement, let E be an A-module,

and let E* be its dual. Denote by tp the canonical homomorphism from E©E*
to End(E).
a) Any element distinct from l in End(E) of the form
Sx,y• = l -1.p(x©y*),
with x E E and y* E E*, is called a pseudo-reflection in E. Such an element s

is called a reflection if x and y* can be chosen so that (x, y*) = 2; show that
we then have s 2 = 1 and s(.r) = -x.
b) Let x E E, y* E E* be such that (x, y*) = l. Show that E is the direct
sum of the submod11le Kx generated by x and the orthogonal complement H
of y*. Show that Kx is free with basis x, and that s = s 2 x,y• is equal to 1 on
Hand to -1 on Kx.
2) With the notation of Exerc. 1, show that det(sx,y·) = l - (x, y*) if Eis a
free K-module of finite type.
~ 3) Let V be a complex Hilbert space with basis e 1 , ... , e1• For 1 ::::; i ::::; l, let
si be a unitary pseudo-reflection, with vector ei, such that si(ei) = ciei, with
Ci i= l; an element of V is invariant under Si if and only if it is orthogonal to
ei. Let W be the subgroup of GL(V) generated by the Si.
a.) Let i be an integer ? 1. Show that every element of/\ i V invariant under
Wis zero. (Argue by induction on l; if V' is the subspace of V generated by
e 1 , ... , e1_ 1 , and if e is a. non-zero vector orthogonal to V', any element of
/\'; V can be written in the form a+ (b A e), with a E /\i V' and b E /\i-l V';
if a+ (b A e) is invariant under W, a and b are invariant under the subgroup
W' generated by s 1 ,. • ., s1- i, and the induction hypothesis applies.)
b) Assume that W is .finite. Show, by using a), that, for any endomorphism
A ofV,
L det(A - w) = Card(W).det(A)
wEW
L <let(l-Aw) = Card(W).
wEW
Deduce that, for any A E End(V), there exists w E W such that Aw has no
non-zero fixed point.
c) Let I' be the graph whose set of vertices is (1, l), the arrows being the
sets {i, j} such that c; and ej are not orthogonal. Show that V is a simple
W-module if and only if I' is connected and non-empty.
d) Assume that Vis a simple W-module. Show that the W-modulcs /\iV
(0 :::::; i :::::; l) are simple. (Show that there exists an integer j such that the
graph I' - {.j} is connected. Argue by induction on l, and apply the induction
hypothesis to the subspace V' generated by the e;, i i j.) Show that these
modules are pairwise non-isomorphic 6 .
§ 3.
1) The notation and assumptions being those of no. 1, show that the chambers
relative to W are open simplices if and only if \V is infinite and irreducible.
Show that E/W is compact if and only if W is a product of infinite irreducible
groups.
2) Let V be a finite dimensional real vector space equipped with a scalar

product, F a finite subgroup of the orthogonal group of V generated by
reflections, A a discrete subgroup of V stable under F, and W the group of
displacements of V generated by F and the translations by a vector belonging
to A. Let .f) be the set of hyperplanes Hof V such that SH E F. Let R be the
set of elements of A orthogonal to an element of .f).
a.) W is generated by reflections if and only if R generates A.
b) In R 2 equipped with the scalar product ((x,y), (x', y')) ,.._.xx'+ yy', let
e1 = (1,0), e2 = (-~, ~), e3 = (-~,- ~), .1i =Rei,
F the dihedral group generated by the sd,, and A the discrete subgroup of
R 2 generated by the e;, a subgroup that is stable under F. Show that W is
not generated by reflections.
3) Let V be a finite dimensional real vector space and W a finite subgroup

of GL(V) generated by reflections. Show that every element of order 2 of W
is a product of pairwise commuting reflections belonging to W. (Argue by
induction on dim(V), and use Prop. 2.)
6 This exercise, hitherto unpublished, was communicated to us by R. Steinberg.
§3. EXERCISES 135
4) Let V be a finite dimensional real vector space, W a finite subgroup of

GL(V) generated by reflections, w an element of W, V' a vector subspace
of V stable under w, and k the order of the restriction w IV' of w to V'.
Show that there exists .T E W of order k, leaving V' stable, and such that
xlV' = w[V'. (Let W' be the set of elements of W leaving fixed the points
of V'. This group is generated by reflections, and w permutes the chambers
relative to W'; deduce that there exists h E W' such that wh leaves stable a
chamber of W'; show that we can take x = wh.)
~ 5) Let V be a finite dimensional real vector space, W a finite subgroup of

GL(V) generated by reflections, C a chamber of W and S the set of walls of
C. If JCS, let WJ be the subgroup of W generated by the SH for HE J, and
let c(J) = (-l)Car<l(Jl. Show that
I = L c:(J)
Card(W) .Jes Card(W.1) ·
(Let (:rjy) be a scalar product on V invariant under W, E the unit sphere of

V, andµ a positive measure on E invariant under W and of total mass 1. If
HES, let DH be the open half-space determined by Hand containing C. Let
E1-1 =DH (1 E. Then
(-C) n E = n (E -
HES
EH),
so
Car~(W) = µ(C nE) = l JJs ( 1- IPEH )dµ
=
Jes
n E1-1) .
I: c:(.J)tt ( HEJ
Conclude by remarking that n EH is the intersection with E of a chamber
HE.J
of WJ, and hence has measure equal to 1/Card(WJ).)
Recover formula (*) by means of Exerc. 26 e) of Chap. IV, §I (putt= 1
in the identity proved in the exercise in question).
6) a) Let K be a commutative field, V a vector space of finite dimension n

over K, ip a symmetric bilinear form on V, and N the kernel of ip. Assume
that dim N = 1. Show that the kernel of the extension of ip to /\ n- l V is of
dimension n - l.
b) Assume in addition that K =Rand that ip is positive. Let (e 1 , •.. , e,,) be
a basis of V, and a;1 = ip( e;, e1 ). Assume that a.;j ~ 0 for i i .i. Assume that
there is no partition I U J of {1, 2, ... , n} such that a;1 = 0 for i E I, j E J. Let
A;1 be the cofactor of a;.7 in the matrix (ai.i). Show that A;1 > 0 for all i and
j. (Let ( 1 e 1 +· · · +(nen be a vector generating N, all of whose coordinates are
> 0. Show that every row and column of ( A;J) is proportional to (( 1 , ... , ( 11 ).
Deduce that Aij = µ(i(j for some constantµ.. By considering the Aii, show
thatµ,> 0.)
c) Show that ( 1 , ... , (n are proportional to ~, ... , ~.
7) Let q(6, ... , ~n) = 2::

z,J
aij(i~j (atj = a)';) be a degenerate positive quadratic
form on Rn, such that aii ~ 0 for i i j. Assume that there is no partition
I U J of { 1, 2, ... , n} such that aij = 0 for i E I, j E J.
a) Show that, if we put (i = 0, we obtain a non-degenerate positive quadratic
form with respect to 6, ... , (i-1, ~i+ 1, ... , En·
b) Show that a.ii > 0 for all i. (Let (( 1 , ... , (n) be an element of the kernel of
q, with (1 > 0, ... , (n > 0. Use the equality a1i(1 + · · · + ani(n = 0.)
c) Show that if we replace one of the aij by some a~j < aij, the new form is
non-positive. (Use the equality 2::
O.ij(i(j = 0.)
i,J
8) Let (aij) be a real symmetric matrix with n rows and n columns.

n
a) Put 8k = L aik· Then, for all (1, ... , (n ER,
i=l
b) Let (1, ... ,(n E: R*. Put L(;aik = tk. Then

'
L a;kEi(k = L tk(~ - ~?: (i(kaik (~ - Ek)
2
i,k k (k 2 i.,k (;, (A:
(In a), replace(; by ~ and aik by (.;(kaik·)

c) If there exist numbers (1, ... , (n > 0 such that 2;: (;aik = 0 (k: =
' form La.;k(i(k
1,2,. .. ,n), and if%~ 0 for ·i =J- j, then the quadratic
i,k
is degenerate and positive. (Use b).)
d) Let ?:
ZJ
%Ei(j be a quadratic form on Rn such that % ~ 0 for 'i
.
i= j.
Assume that there is no partition I U J of { 1, 2, ... , n} such that q.;.j = 0 for
i E I, j E J. Then the form is non-degenerate and positive if and only if there
exist (1 > 0, ... , (n > 0 such that L (iQik = 0 (k = 1, ... , n).
i
§4. EXERCISES 137
§4.
In the exercises below, (W, S) denotes a Coxeter system. We assume that Sis
finite; its cardinal is called the rnnk of (W, S). We identify W with a subgroup
of GL(E) by means of CJ (cf. nos. 3 and 4).
1) Let E 0 be the orthogonal complement of E with respect to the form BM.

Show that E 0 is the radical of the W-module E (Algebra, Chap. VIII, § 6, no.
2), and that E/E0 is the direct sum of m absolutely simple, pairwise non-
isomorphic modules, where m is the number of connected components of the
graph of S.
2) a) Let I'w be the set of extremal generators of the cone w(C) and let A
be the union of the I'w for w E W. Show that A, equipped with the set
{I'wlw E W}, is a building (Chap. IV,§ 1, Exerc. 15). Show that the map j
from the apartment Ao associated to the Coxeter system (W, S) (Chap. IV,
§ 1, Exerc. 16) to A that transforms the point wW(s) of Ao (for w E W and
s ES) to the generator w(Re 5 ), is an isomorphism from Ao to A, compatible
with the action of W. Show that, if t = wBw- 1 , with w E Wand s ES, the
image under _j of the wall L 1 defined by t (resp. of a half of Ao defined by Lt)
(loc. cit.) is the set of elements of A contained in the hyperplane (resp. the
closed half-space) that is the transform by CJ* ( w) of the hyperplane e .• = 0
(resp. of one of the closed half-spaces e 5 ;? 0 or e 5 ~ 0).
b) Show that W is finite if and only if there exists an element w 0 E W such
that w 0 (C) = -C. This element w 0 is then unique and is the longest element
of W (use Exerc. 22 of Chap. IV, § 1). Show that in that case j (-a) = - j (a)
for all a E Ao (cf. Exerc. 22 c) of Chap. IV, § 1).
c) Show that W is finite if and only if the cone U formed by taking the union
of the cone closures w(C) for w E W is equal to the whole of E* (if W is
finite, use b) and the convexity of U. If U = E*, consider an element w E W
such that w(C) n (-C) -f- 0 and show that w(C) = - C).
d) Let H be a finite subgroup of W. Show that there exists a subset X of S
such that W x is finite and contains a conjugate of H. (Argue by induction
on Card(S). Let x E C and x = L h(x). Show, by using c) above, that Wis
hEH
finite if x = 0. If x -f- 0, there exist w E W and Y C S, Y -f- S, such that w(.'f)
belongs to Cv (in the notation of no. 6), and hence H C w- 1 Wvw; apply the
induction hypothesis to Y.)
3) Assume that (W, S) is irreducible.

a) Show that the commutant of the W-module E reduces to the homotheties.
b) Show that the centre of W reduces to {1} if Wis infinite or if Wis finite
and the longest element w0 (cf. Exerc. 2) is -f- -1. If W is finite and w 0 = -1,
the centre of W is { 1, wo}.
c) Show that any element w i- 1 of W such that wsw- 1 = S transforms C

to -C (show that w(e 5 ) = -ewsw-1, first for some s ES, and then for all
s ES). Deduce (Exerc. 2) that such an element exists only if W is finite and
that it is then equal to wo.
4) Assume that Card(S) = 3. Ifs ES, let a(s) = m(u,·v), where {u,v} =
S - {s}. Let A= I: l/a(s).
sES
a) If A > 1, show that B1vr is non-degenerate and positive (in which case W
is finite), ·-
b) If A= 1, show that BM is degenerate and positive.
c) If A < 1, show that BM is non-degenerate and of signature (2, 1) (cf.
Algebra, Chap. IX, § 7, no. 2).
Show that, in case a), the order q of W is given by the formula q =
4/(A - 1) (use Exerc. 5 of§ 3).
5) Let A be the subring of R generated by the numbers 2.cos(7r/m(s,s')).

Show that A is a free Z-module of finite type, and that the matrices of the
O"(w) for w E W have their coefficients in A. Deduce that the coefficients of
the characteristic polynomials of the O"(w) are algebraic integers.
, 6) a) Let m be an integer ~ 2, or +oo. Show that 4 cos 2 ~ E Z is equivalent

tomE {2,3,4,6,+oo}.
b) Let I' be a latt·ice in E, i.e. a discrete subgroup of E of rank dimE. Show
that, if r is stable under w, the integers m(s, s') for s =/: 8 1 all belong to the
set {2,3,'4,6,+oo}. (Observe that Tr(O"(w)) E Z for all w E W; apply this
result to w = ss', and use a) above.)
c) Assume that m(s, t) E {2,3,4,6,, +oo} for s =/: t ES. A family (xs)sES of
positive real numbers is called radical if it satisfies the following conditions:
m(.s, t) = 3 ~ Xs =.rt
m(s, t) =4~ Xs = J2.xt or Xt = J2.xs

m(s, t) = 6 =} x, = J3.xt or :i; 1 = v'3 ..r 5 .
m(s, t) = +oo ~ Xs =.rt or x., = 2xt or Xi = 2x 5 .
If (xs)sES is such a family, put 0: 8 = X 5 C5 • Show that
0" 5 (Cl'.t) = O:i - n(s, t)a,, with n(s, t) E Z.
Deduce that the lattice r with basis (a,)sES is stable under w.

d) With the assumptions in c), suppose that the graph of (W, S) is a forest.
Show that in that case there exists at least one radical family (x 5 ). (Argue
§ 4. EXERCISES 139
by induction on Card(S); apply the induction hypothesis to S- {s 0 }, where

so is a terminal vertex of the graph of S.)
e) \i\Tith the assumptions inc), suppose that the graph of (W,S) is a circuit.
Let n 4 (resp. n5) be the number of arrows {s, t} of this graph whose coefficient
m(s, t) is equal to 4 (resp. 6). Show that a radical family exists if and only
if n 4 and n6 are both even. If this condition is not satisfied, show that no
lattice of E is stable under W (if S = {s 1 , ... , Sn}, with Si linked to s;+ 1 for
1 ~ i ~ n - 1, and Sn linked to s 1 , put c = s 1 .. . sn and show that Tr(a(c))
is not an integer).
7) Assume that (W, S) is irreducible and that BM is positive.

a) Show that, for any subset T of S distinct from S, the group WT is finite
(use Th. 2 as well as Lemma 4 of § 3, no. 5).
b) Show that, if Card(S) ): 3, the m(s, s') are all finite.
c) Assume that W is infinite. Show that, if T C S, T i S, the group a(WT)
leaves stable a lattice in RT. Deduce that, if Card(S) ~ 3, the m(s, s'), sf. s',
all belong to the set {2, 3, 4, 6} (use Exerc. 6).
8) Let s E Sand w E W. Show that, if l(ws) > l(w), the element w(es) is
a linear combination with coefficients ;;::: 0 of the ei for t E S; show that, if
l(ws) < l(w), w(es) is a linear combination with coefficients ~ 0 of the et.
(Apply property (Pn) of no. 4 to w- 1 , and argue by polarity.)
9) Show that the intersection of the subgroups of W of finite index reduces

to the identity element (use Exerc. 5). Deduce that there exists a subgroup
of W of finite index that contains no element of finite order other than the
identity element. (Use Exerc. 2 d).)
, 10) Let G be a closed subgroup of GL(E) containing W. Assume that G

is unimodular (Integration, Chap. VII, § 1, no. 3). Let D be a half-line of E*
contained in C, and let Go be the stabiliser of D in G.
a) Let Ll be the set of elements g E G such that g(D) C C. Show that Ll is
open, stable under right multiplication by Go, and that the composite map
Ll -> G--+ W\G is injective, W\G denoting the homogeneous space of right
cosets of G with respect to W.
b) Letµ be a Haar measure on G. Show that, if µ(Ll) is finite, the subgroup
Go is compact. (Let K be a compact neighbourhood of the identity element
contained in Ll; show that there exist finitely many elements hi E Go such
t~at every set of the form Kh, with h E Go, meets one of the Kh;; deduce
that Go is contained in the union of the K- 1 .K.hi, and hence is compact.)
c) Let I/ be a non-zero positive measure on W\G invariant under G. Show
that, if v(W\G) < oo, then Go is compact.
, 11) Let H be the subset of Rn consisting of the points x = (x 0 , ... ,Xn-d

for which the form
B(x) = -x6 +xi+· .. + x;,_ 1

is< 0, and let PH be the image of Hin the projective space Pn_ 1 (R). Let
G be the orthogonal group of the form B.
a) Show that PH is a homogeneous space of G, that the stabiliser of a point
is compact, and that G acts properly on PH.
b) Let wand fl be the differential forms on H defined by the formulas
n-1
w = L
i=O
(-1rxidxo /\ · · · /\ dx;-1 /\ dx;+1 /\ · · · /\ dxn-1
fl= w
(-B(.r))n/2"
Show that fl is the inverse image under the canonical projection 7f : H - t PH
of a differential form fl on PH. Show that the positive measure v associated
to fl (Differentiable Varieties R, 2nd part) is invariant under G, and that it
is the only such measure, 11p to a scalar factor.
c) Let C be an open simplicial cone with vertex 0 in Rn (§ 1, no. 6). Assume
that C is contained in H, and denote by PC the image of C under 7r : H - t PH.
Show that, if n ~ 3, then v(PC) < oo (identify PH with the subspace of
n-1
Rn- l consisting of the (x 1 , ... , :cn-l) such that L x~ < 1, and determine
i=l
the measure corresponding to v on this subspace). Show that PC is relatively
compact in PH if and only if C is contained in H.
, 12) Assume that the form x.y = BM(x, y) is non-degenerate and that W
is infinite. Identify E with its dual E* by means of BM; in particular, denote
by (e;) the basis of E dual to the basis (e 8 ), and by C the interior of the
simplicial cone C generated by the e;. Let G be the orthogonal group of BM,
and letµ be a Haar measure on G; the group G is unimodular and contains
W.
a) Show that, if 11(W\G) < oo (where 11 is a non-zero positive measure
invariant under G), the form BM is of signature (n - 1, 1), with
n = dim(E) = Card(S),
and that x.x < 0 for all x E C.
(Let .T E C be such that x.x i 0, and let Lx be the hyperplane orthogonal
to x. Show, by using Exerc. 10, that the restriction of BM to Lx is either
positive or negative; show that the second case implies that B11:1 is of signature
(l,n- 1) and that this is impossible. Deduce that .r.x ~ 0 for all x EC, and
hence .T.X < 0 since C is open.)
§ 4. EXERCISES 141
b) Conversely, assume that BM is of signature (n - 1, 1) and x.x < 0 for all

x E C (in which case W, or (W, S), or the corresponding Coxeter graph, is
said to be of hyperbolic type). Let H be the set of x E E such that x.x < 0,
and let H+ be the connected component of H containing C. Show that H is
the disjoint union of H+ and H_ = -H+, and that H_ is contained in the
simplicial cone generated by (e5 ) sE s (use the fact that H _ is the polar of
H+). Show that H+ and H_ are stable under vV.
c) Retain the notation and assumptions of b). Let f be the linear form on E
defined by f(es) = 1 for alls ES. If x EH, put cp(:z:) = f(x) 2 /(.r.x). If PH
denotes the image of the cone H in the projective space P(E), the function
-t ip(w.x) and
w H f(w.x) attain their maximum for a value w 1 E W (use the fact that
G acts properly on PH. cf. Exerc. 11, and that W is discrete in G ); show
that these properties are equivalent to w 1 (x) E C. Deduce that C n 1-1+ is a
fundamental domain for the action of W in H+, and that the image of this
fundamental domain in PH has finite measure for the invariant measure v
on PH (use Exerc. 11). Conclude that 1/(W\G) < oo. Show that W\G is
compact if and only if C is contained in H+, i.e. if e; .e; < 0 for all s E S (in
which case W, or (W, S), or the corresponding Coxeter graph, is said to be
of compact hyperbolic type).
~ 13) Show that (W, S) is of hyperbolic type (cf. Exerc. 12) if and only if the
following two conditions are satisfied:
(Hi) The form BM is not positive.
(H 2 ) For any subset T of S distinct from S, the form BM(T) associated to
the Coxeter system (WT, T) is positive.
(If (W, S) is of hyperbolic type, we have seen that e;.e; ~ 0 for alls ES,
the notation being that of Exerc. 12. The restriction of Btv1 to the hyperplane
E( s) orthogonal to e; is thus ~ O; since E( s) is generated by the et for
t f s, (H2) follows. Conversely, assume that (H 1 ) and (H 2 ) are satisfied; let
x = ~ a 5 e 5 be an clement of E such that x ..x < O; let :i:+ (resp ..r_) be the
s
sum of the a 5 e8 for which a 8 is> 0 (resp. ~ O); show that either x+.X+ < 0
or x_.x_ < 0. IfV denotes the open simplicial cone generated by thee,, and
H the set of .x E E such that x.x < 0, deduce that there exists a connected
component Ho of H that meets V; by using (H2), show that H 0 does not
meet the walls of V, so Ho CV. Deduce that the form B~,1 (x,y) = x.y is
non-degenerate of signature (n - 1, 1), and that C is contained in - Ho, hence
the fact that (W, S) is of hyperbolic type.)
14) Show that (W, S) is of compact hyperbolic type (Exerc. 12) if and only
if the following two conditions arc satisfied:
(H 1 ) The form BM is not positive.

(HC) For any subset T of S distinct from S, the group WT is finite (i.e.
the form BM(T} is non-degenerate and positive).
(Use Exercs. 12 and 13.)
In particular, a Coxeter system of hyperbolic type of rank 3 is of compact
hyperbolic type if and only if the m(s, s') a.re all finite (cf. Exerc. 4).
, 15) *a) Show that the nine Coxeter graphs below 7 are of compact hyperbolic
type, and that they are, up to isomorphism, the only graphs of rank 4 with
this property (use the classification of Chap. VI, § 4): ..
CJ o D tJ CJ
b) The same question for rank 5, the list consisting of the following five graphs:
c) Show that there is no graph of compact hyperbolic type of rank ~ 6.*
16) *Show that any Coxeter graph of hyperbolic type of rank ~ 4 that has
an edge with coefficient 6 is isomorphic to one of the following eleven graphs
(which are non-compact and of rank 4):
O,__---<O>----<O,...._c6,,_0
*
7 In these graphs, any edge that has no label next to it is taken to have coefficient
3 (cf. Chap. IV, § 1, no. 9).
§ 4. EXERCISES 143
17) *Show that the hyperbolic Coxeter graphs of higher rank are the following
three graphs (which are of rank 10):
*
~ 18) Assume that (W, S) is of hyperbolic type, and that W leaves stable a
lattice I' of E. Let G be the orthogonal group of BM, and let G(I') be the
subgroup of elements g E G such that gI' = I'. Show that G(I') is a discrete
subgroup of G. Show that Wis a subgroup of finite index of G(T) (use the
fact that the measure of W\ G is finite).
*If (W, S) is actually of compact hyperbolic type, show that the corre-
sponding Coxetcr graph is isomorphic to one of the following (use Exercs. 4,
6 and 15):
::\} :\/4 and
Show that all the groups corresponding to the graphs of Exerc. 16 (with
the exception of the la.st four) leave stable a lattice (use the method of Exerc.
6) .•
19) Let (W, S) be the Coxetcr system corresponding to the graph
This is a system of compact hyperbolic type (cf. Exerc. 15). The coefficients
of the form BM with respect to the basis (es) belong to the subring A of the
field Q( J5) consisting of the elements of this field that are integers over Z.
a) Let a be the embedding of K into R that maps J5 to -JS. Show that
the transform a(BM) of the form BM by a is 11011-degenerate and positive.
b) Let G be the orthogonal group of BJVI, and let Ga be that of a(BM)· Let
GA be the subgroup of G consisting of the elements whose matrices with
respect to (es) have coefficients in A. Show that GA can be identified with a
discrete subgroup of G x Ga and then, using a), show that GA is a discrete
subgroup of G.
c) Show that W is a subgroup of GA of finite index.

d) Prove analogous results for the other graphs in Exerc. 15 (the field Q( VS)
sometimes being replaced by Q( J2), with the exception of the graph
treated in Exerc. 18, and of the graph8
, 20) For every subset { s, s'} of S such that m(s, s') = oo, let r(s, s') be a
real number ~ -1. Equip E with the bilinear form Br such that
7T
Br(e 8 , es')= BM(es, es') = - cos ( ') if m(s, s') =f. oo
m s,s
Br(es, es') = r(s, s') if m(s, s') = oo.
We define, as for BM, the reflections a 8 with vector es, leaving the form B,.
invariant.
a) Show that there exists a unique homomorphism ar : W -> GL(E) such
that O' r (gs) is equal to the reflection O's defined above.
b) Show that the assertions in Prop. 4, Th. 1, its Corollaries, and Lemma 1
remain true for a r.
§ 5.
1) Determine the algebra of symmetric invariants of a finite dihedral group

(for its canonical representation of dimension 2, cf. § 4, no. 2).
2) Let A be a principal ring, Ea free A-module of finite rank l, and G a finite

subgroup of GL(E). Assume the following:
(i) If q = Card(G), the element q.l of A is invertible.
(ii) G is generated by pseudo-reflections in E (i.e. by elements s such that
(s - l)(E) is a monogenic submodule of E, cf. § 2, Exerc. 1.)
8 As E. Vinberg has shown, the group W corresponding to this last graph is not a
subgroup of G of "arithmetic" type. The same situation arises for various other
graphs of hyperbolic type (non-compact, this time), notably for the last four in
Exerc. 16.
§ 5. EXERCISES 145
Let S(E) be the symmetric algebra of E, and let S(E)G be the subal-
gebra of S(E) consisting of the elements invariant under G. Show that, for
any homomorphism from A into a field k, S(E)G ® k can be identified with
S(E ® k)G. Deduce, by applying Th. 4, that S(E)G is a graded polynomial
algebra over A.
, 3) Let K be a field of characteristic zero, V a vector space of finite dimension

l over K, and G a. subgroup of GL(V) generated by pseudo-reflections. Put
q = Card(G); denote by S (resp. L) the symmetric algebra (resp. exterior
algebra) of V. If x EV, denote by :c (resp. x') its canonical image in S (resp.
L).
a) Let E = S ® L be the tensor product of S and L. Show that there exists a
unique derivation d on E such that dx = x' and dx' = 0 for all x E V.
b) Let SG be the algebra of symmetric invariants of G and let P 1 , ... , P1 be

homogeneous elements of Sc such that SG = K[P 1 , •.. , Pl]. For every subset
I= {i1, ... ,'ir} of (1,1) with i1 < ··· < ir, put
WJ = dPi 1 • • .dPir·
Show that the w1 are linearly independent over S, and that they belong to
the subalgebra EG consisting of the elements of E invariant under G. Deduce
that, for all w E EG, there exist a, c1 E sc, with a. f:. 0, such that
a.w = Lcrw1.
I
c) Show that every element of EG contained in S ® f\ 1 V is of the form

c.dP I ... dP1, with c E sG. (Apply Prop. 5.)
d) Show that the Wj form a basis of the sc-rnodule EG. (With the notation
in b), multiply both sides of the relation a.w = L c1w1 by an element WJ, and
I
apply c); deduce that, if I is the complement of .J, a. divides c1 ; hence 9 the
fact that the wr generate EG.)
e) Let Sn (resp. Lm) be the homogeneous component of S (resp. L) of degree
n (resp. m). Put
En,m = S.,. ® Lm, E~,m = EG n En,m, an,m =dim E~,m'
a(X, Y) = L O.n,mXnym.
n,m~O
By using d), prove the formula
a(X, Y) = TI 1 + Y.Xp,-1
i=l 1- XP;
where Pi = deg(Pi)·
9 For more details, see L. SOLOMON, Invariants of finite reflection groups, Nagoya
Math. Journal, v. XXII (1963), p. 57-64.
J) If g E G, let Trn,m(.9) be the trace of the automorphism of En,m defined

by g; put
Tr(X, Y)(g) = L Trn,m(g)XnYm.
n,m;;,o
Show that
Tr(X Y)( ) = det(l + y g).
' .9 det(l - Xg)
g) Let q = Card(G). Show that
~ L Tr(X, Y)(g) = a(X, Y),

q 9EG
i.e.
- L
q gEG
det(l + Yg)
det(l - Xg)
= n
i= 1
1 + Y.Xp,-1
1 - XP·
(Use the following result: if G acts on a finite dimensional vector space E,
the dimension of the space of elements of E invariant under G is equal to
~ I: TrE(g).)
gEG
What does this formula give for Y = O?
h) For any integer p ~ 0, let H,, be the set of elements g E G that have 1 as
an eigenvalue with multiplicity p. Let hp = Card(Hp)· Prove the formula
l l
I: h,, TP =
p=O
n (p.; -
i= 1
i + T).
(In the formula in g) above, replace Y by - 1 + T( 1 - X) and then put X =1

in the result. If g E Hp, the term Tr(X, Y)(g) becomes TP.)
a) Show that G 1 is a non-abelian simple group of order 168, containing 21

elements of order 2.
b) Show that the degrees of the irreducible complex representations of G 1 are
1, 3, 3, 6, 7, 8.
c) Let p: G 1 _. GL3(C) be an irreducible representation 10 of G 1 of degree
3. If y E G 1 is of order 2, show that Tr(p(y)) = -1. Deduce that -p(y) is a
reflection.
d) Let G be the subgroup of GL3(C) generated by the elements -p(y), for
y of order 2 in G 1 . Show that G is isomorphic to G 1 x { 1, - l}, and is thus
of order 336.
10 A detailed study of such a representation can be found in H. WEBER, Lehrbuch
der Algebra, Bd. II, Abschn. 15.
§ 5. EXERCISES 147
e) Show that the characteristic degrees k 1 , k 2 , k 3 of the algebra of symmetric

invariants of G are equal to 4, 6 and 14. (Use the relations IT k; = 336 and
L(k; - 1) = 2i.) '
"
J) Show that G is not a Coxcter group .•
5) *Let K be a field and S = K[X 1 , ... , Xn] a graded polynomial algebra

over K, generated by algebraically independent elements X;, homogeneous of
degrees > 0.
a) Let Y 1 , ... , Yn be homogeneous elements of S of degrees > 0, and let
R = K[Y 1 , ... , Y,,] be the subalgebra of S generated by these elements. Prove
the equivalence of the following properties:
(i) (Y 1 , ... , Y n) is an S-regular sequence.
(ii) S is integral over R
(iii) The ideal of S generated by Y 1 , ... , Yn is of finite codimension in S.
(iv) For any extension R of K, the system of equations Y.;(x 1 , ... , Xn) = 0
(1 ::::; 'i ::::; n, x; E K) has only the trivial solution (0, ... , 0).
( (i) {iii) follows from a theorem of Macaulay ( Cornm1ttative Algebra);
{:::=}
(iii) (iv) follows from the Zeros Theorem ( Cornrnv.tat·ive Algebra, Chap.
{:::=}
V, §3, no. 3, Prop. 2); (ii) {:::=} (iii) is easy.)

If these properties are satisfied, show that the Y; are algebraically inde-
pendent over K, and that S is a free R-module of rank equal to
I1 cleg(Y;)
I'I deg(X;) ·
'
b) Let G be a finite group of automorphisms of the graded algebra S, let sc be
its subalgebra of invariants, and let Yi, ... , Y n be elements of SG satisfying
conditions (i), ... , (iv) above. Show that SG = K[Y 1 , ... , Yn] if and only if
Ildeg(Y;)
Carcl(G) = {I.,
deg(X;) ..
~ 6) *Let n be an integer ,;::: 1, q a power of a prime number, K = Fq,

V = Kn, G = GL(n, K) and G 1 = SL(n, K). Identify G with the group
GL(V). Further, denote by S the algebra S(V) = K[X 1 , .•. , X,,] and R (resp.
Ri) the subalgebra of S consisting of the elements invariant under G (resp.
G1).
a) Let e = (e 1 , ... ,en) be a sequence of integers~ 0. Put
This is an element of S.
Show that
g.Le = det(g)Le for all g E G.
(Remark that, if g.Xi = LaijXj, then g.Xf = LaijX1qc.) In particular, Le
J J
belongs to R 1 .
b) Let j E [1, n]. Put
the product being over all families (ai.il.i<i~n of elements of K. Let

n
T= TI Zj.
j= 1
Show that T divides all the Le. Deduce that T = Len, where en =
(0, 1, ... , n - 1), and that Tis invariant under G1.
c) Denote by ei (1 :::;; i:::;; n -1) the sequence (0, 1, ... , i- 1, i + 1, ... , n), and
put Yi = Le)T, cf. b). Show that the Yi belong to R and that deg(Y;) =
qn _ qi.
d) Let S' = K[X 1 , ..• , Xn_iJ, and let T', Y~, ... , Y~,_ 2 be the elements of S'
defined in the same way as T, Y 1 , ... , Yn-l (but replacing n by n - 1). Let
f : S ___, S' be the homomorphism defined by
J(X.;) =Xi (1:::;; i:::;; n - 1), .f(Xn) = 0.
Show that
.f (T) = 0, .f (Y 1) = T'q(q-J), .f (Yi) = Y:~ 1 for 2 :::;; 'i :::;; n - l.
e) Show that the family (T, Y 1 , ... , Y n-i) satisfies condition (iv) of Exerc. 5.
(Let x = (x 1 , •.. , .rn) be a zero of the system (T, Y 1 , ... , Yn-l) in an exten-
sion R of K. Since x is a zero of T, the Xi satisfy at least one non-trivial
linear relation with coefficients in K. Transforming ;r; by an element of G 1 if
necessary, we can assume that Xn = 0. Conclude by applying d) and arguing
by induction on n.)
.f) Show that R1 = K[T, Y 1, ... , Yn-1L and that T, Yi, ... , Y n- l are alge-
braically independent ("Dickson's Theorem" - apply e) and Exerc. 5, remark-
ing that the order of G 1 is equal to deg(T) = Ldeg(Yi)).
t
§ 5. EXERCISES 149
, 7) *Let R be a regular local ring (Commutative Algebra., Chap. VIII, § 5,

no. 1), with maximal ideal m and residue field k. Let G be a finite group
of automorphisms of R, and let R' = RG be the subring of R consisting
of the elements invariant under G. This is a local ring with maximal ideal
m' = m n R'. Assume the following:
(i) R' is noetherian and R is an R' -module of finite type.
(ii) The composite R' --> R--> k is surjective.
Put V = m/m 2 ; this is a k-vector space. The action of G on R defines a
homomorphism€: G--> GL(V).
a) Let p be a prime ideal of R of height 1 (Commutative Algebra, Chap. VII,
§ 1, no. 6) and let s E G be such that s(p) = p and that s acts trivially on
R/p. Show that €(s) is a pseudo-reflection of V. (Remark that the image of
p in m/m 2 is of dimension 0 or 1.)
b) Show that, if R' is regular, the subgroup c:(G) of GL(V) is generated by
pseudo-reflections, (Let H be the subgroup of G generated by elements whose
image under€ is a pseudo-reflection, and let RH be the subring of R consisting
of elements invariant under H. Show, using a), that no prime ideal of R' of
height 1 is ramified in RH; using the fact that RH is integrally closed, deduce
by means of the Purity Theorem 11 that R' = RH, so H = G.)
c) Assume now that the order of G is coprime to the characteristic of k. Show
that € is injective.
Let (m~) be the filtration of R' induced by the m-adic filtration (mn) of
R, and let
i : gr(R') --> gr(R)
be the canonical homomorphism of the associated graded algebra of R' to
that of R (Commutative Algebra, Chap. III, § 2). Show that i is injective,
and that its image is the subring gr(R)G of gr(R) consisting of the elements
invariant under G.
d) Retain the assumptions and notation of c), and assume further that e(G) is
generated by pseudo-reflections. If l =dim V, let P 1 , ... , Pi be algebraically
independent homogeneous generators of the k-algebra gr(R)G (such elements
exist by Th. 4 and the fact that gr(R) can be identified with the symmetric
algebra of V); let p 1 , ... ,p1 be their degrees. By c), we can find Xi Em;,; such
that gr(xi) = P;. Show that the Xi generate the ideal m' and deduce that R'
is regular .•
, 8) *Let V be a finite dimensional vector space over a field K, let S be the

symmetric algebra of V, and let G be a finite subgroup of GL(V). Assume
11 Cf.
M. AUSLANDER, On the purity of the branch locus, Amer. J. of Math., v.
LXXXIV (1962), p. 116-125.
15.0 GROUPS GENERATED BY REFLECTIONS Ch. V
that the algebra sG of symmetric invariants of G is a graded polynomial

algebra.
a) Let u be an element of the dual V* of V, and let G,. be the subgroup
of G consisting of the elements leaving u invariant. Show that G,L is gener-
ated by pseudo-reflections. (The linear form u extends to a homomorphism
fu : S ~ K. If Su denotes the local ring of S with respect to the kernel of fu,
the ring Su is regular. Conclude by applying Exerc. 7 b) to G,,, considered as
a group of automorphisms of Su.)
In particular, G is generated by pseudo-reflections.
b) Let A be a subset of V*, and let GA= n Gu. Show that GA is generated
uEA
by pseudo-reflections. (Extending the base field if necessary, we can assume
that K is infinite. Show that in that case there exists an clement v of the
vector subspace of V* generated by A such that Gu = GA. Conclude by
applying a) to v.)*
9) Let V be a vector space of dimension 4 over a finite field K, of characteristic

different from 2, and let Q be a non-degenerate quadratic form on V of index
2 (Algebra, Chap. IX, § 4, no. 2). Let G = O(Q) be the orthogonal group of
this form; this is a finite group generated by reflections (loc. ed., § 6, no. 4,
Prop. 5).
a) Let E be a maximal totally isotropic subspace of V, and let GE be the
subgroup of G consisting of the elements g such that g(:1:) = x for all x E E.
Show that GE is isomorphic to the additive group of K, and contains no
pseudo-reflection.
b) Show that the algebra of symmetric invariants of G is not a graded poly-
nomial algebra (use the preceding exercise).
§ 6.
In the exercises below (with the exception of Exerc. 3), the assumptions and
notation are those of § 6.
1) Assume that Wis irreducible. Let c be a Coxetcr transformation of W, let

r be the subgroup of W generated by c, and let L1 be the set of unit vectors
r
orthogonal to an clement of fJ. Show that has l orbits in L1, and that each
orbit has h elements. (Argue ar; in the proof of Prop. 33 of Chap. VI, § 1,
no. 11.)
, 2) Assume that W is irreducible. Let C be a chamber relative to W,

(H 1 , ... , Hi) its walls, and ei a non-zero vector orthogonal to Hi. Assume
that e 1 , ... , e1. (resp. er+i, ... , e1) are pairwise orthogonal (cf. no. 2). For all
u E Z, define Hu. and s" by Hu = Hk if u =: k (mod. l) and Su = SH,..
§ 6. EXERCISES 151
a) Show that the elements of Sj are the s 1 s 2 ... s,, _ 1 H,, for it = 1, 2, ... , lh/2.
b) Let s' = s1 ... Sr and s" = Sr+l ... s1, so that c = s's" is the Coxeter
transformation associated to the ordered chamber C. Let w 0 be the element
of W that transforms C to -C (cf. § 4, Exerc. 2). Show that, if h is odd,
wo = s's"s' ... s"s'

~
= ..__,,_.
s"s's" ... s's"= s"c(h-IJ/ 2 = c(h-l)f 2s'.
h terms h tCl'lllS
c) Put S = {s 1 , ... , s1}. The pair (W, S) is a Coxeter system. If w E W,

denote by Is (w) the length of w with respect to S (Chap. IV, § 1, no. 1).
Show that
ls(s') = r, ls(s") = l - r. ls(e) = l.
Deduce that, with the a:;sumptions in b),
h-1 h-1
ls(w 0 ) ~ r + -·-
2 -1 and ls(wo) ~ (l - r) + -·-2 -l.
Show on the other hand that ls(w 0 ) = Card(Sj) = hl/2 (use Excrc. 22 of
Chap. IV, § 1). Deduce that r = l/2 and that (s 1 , s 2 , ... , s 11, 12 ) is a reduced
decomposition of w 0 .
d) Show that, if his even, (s 1 , ... , s 11i; 2 ) is a reduced decomposition of w 0 =
c"1 2 . (Same method.)
~ 3) Let K be a commutative riug, Ea free K-rnodule with basis (e 1 , ... ,e1),

and fi, ... , Ji clements of the dual of E. Put a;1 = fi (Cj). If 1 ~ i ~ l, let Si
be the pseudo-reflection se,,f, (§ 2, Exerc. 1). Then
Put c = s 1 ..• s 1, and z; = c(ei)·

a) For 1 ~ i, k ~ l, put
y'f = s1 ... sk(c;) and Yi = s1 ... si-l (e;) = y;:- 1 .
Thus yf = ei and yf = zi.

Show that
k-1 k
Y; - Y; = akiYk·
Deduce the formulas

e; L akiYk
= Yi + k<-i
z; = Yi -
k~i
L akiYk·
b) Let C be the matrix of c with respect to the basis (e;). Let U = (u.;1 ) and
V = (v;1 ) be the matrices defined by
if i < j if i < j
otherwise, otherwise.
The matrix I+ U is invertible with determinant l. Show that
C = (I - V) (I + U)- 1 .
Deduce that
det(,\I - C)-= det((.\ - l)I -t V + .\U),
in other words
det(-AI - C) =
(.\-l)+all .\a12 .\a13
a21 ,\ (,\ - 1 ) + a22 -Aa23
a31 a32 (,\ - 1) + a33
(,\ - 1) +au
c) Let I' = (I, S) be the graph whose set of vertices is I = [l, I] and whose set
S of arrows is the set of subsets {i, j} of I with 2 elements such that a;j i 0 or
aj, i 0. If a= {i,j} belongs to S, denote by 0- 0 the transposition of i and j
(considered as an element of the symmetric group S 1 ), and put a°'= - a; 1 aji·
Let ~ be the set of subsets of S consisting of the arrows whose vertices
are disjoint. If X E ~' denote by C(X) the set of i E I that are not vertices
of any arrow of X, and put a-x = IT <JO!, ax = IT a".
aEX O!EX
Let a E S1 , and let du be the term corresponding to a in the expansion
of the determinant of(.\ - l)I + V +-AU. Show that, if a is of the form ax,
with X E ~' then
du= axACard(X) IT (,\ - 1 + a,i)·
iEC(X)
Assume now that I' is a forest (Chap. IV, Appendix, no. 3). Show that, if
a E St is not of the form ax with X E ~' then du = 0. Deduce the formula
det(.\ - c) = L ax,\Carc\(X) IT (,\ - 1 +a,;).
XE~ iEC(X)
d) Consider the polynomial

x a12 a11
az1 x a21
P(X) =
a11 0.12 x
To the assumptions in c), we add that aii = 1 for all i. Show that in that
case
§ 6. EXERCISES 153
4) With the assumptions and notation of no. 1, denote by niJ the order of
SH,sH 1 , and put aii = -2cos;.j. Then a.ii= 2 and aiJ:::::; 0 if'i I= j, cf. §3.
Show 12 , by using the method of the preceding exercise, that

x a12 all
a21 x a21
l
TI (X - 2cos(nm;/h)),
i=l
ao a.12 x
where m 1 , ... , m1 are the exponents of W.
12 For more details, see: H. S. M. COXETER, The product of the generators of

a finite group generated by reflections, Duke Math. Journal., v. XVIII (1951),
p. 765-782.
CHAPTER VI
Root Systems
§ 1. ROOT SYSTEMS
In this paragraph, k denotes a field of characteristic z.ero. From no. 3 onwards,

we assume that k = R.
1. DEFINITION OF A ROOT SYSTEM
Lemma 1. Let V be a vector space over k, Ra .finite subset of V genemt'i11.g

V. For any a E R Stich that a -f- 0, there c:cists at most one reffoction s of V
such that s(o:) =-a and s(R) = R.
Let G be the group of automorphisms of V leaving R stable. Since R
generates V, G is isomorphic to a subgroup of the symmetric group of R,
and hence is finite. Let s,s' be reflections of V such that s(a) = s'(n) =-er,
s(R) = R, 8'(R) = R. Then t = ss' belongs to G, and hence is of finite order
m. On the other hand,
t(a) =a and t(x) = .r: mod. ka for all.TE V.

Hence, there exists a linear form f on V such that
t(x) = :c + f(.r:)o: for all x EV

and f (a) = 0. By induction on n, it follows that
tn(x) = x + nf(x)o: for all a: EV.
Taking n equal tom, we see that mf (x) =0 for all :r E V, so f = 0, t = 1

ands= s 1 .
DEFINITION 1. Let V be a vector space over k, and R a subset of V. Then

R is said to be a root system in V if the following conditions are satisfied:
(RS1) R is finite, does not contain 0, and generates V.

(RSn) For all o E R, there exists an element a:· of the dual V* of V
such that (o, o-) = 2 and that the reflection Sc.,<>· (cf. Chap. V, § 2) leaves R
stable.
156 ROOT SYSTEMS Ch. VI
(RSm) For all O'. ER, O'.-(R) ~ Z.
By Lemma 1, the reflection s0<,<> (and hence also the linear form O'.-) is
determined uniquely by O'., so (RSm) makes sense. We put s"'·"'- = s"'. Then
s"'(x) = x - (O'.',:r)O'. for all x EV.
The elements of R are called the roots (of the system considered). The
dimension of V is called the rank of the system.
The automorphisms of V that leave R stable are called the automorphisms
of R. They form a finite group denoted by A(R). The subgroup of A(R)
generated by the s"' is called the Weyl group of R and is denoted by W(R),
or simply by W.
Remark. 1) Let k' be an extension of k. Identify V canonically with a subset

of V © k' and V* with a subset of V* © k' = (V © k')*. Then, R is a root
system in V © k', and the o. - are the same as before.
Lemma 2. Let R be a root system in V. Let (xly) be a symmetric bil·inear

form on V, non-degenerate and invariant under W(R). Ident~fy V with V*
by means of this form. If O'. E R, then O'. is non-isotropic and
20'.
O'.-=--
(alct) ·
This follows from formula (4) of Chap. V, § 2, no. 3.
PROPOSITION 1. Let V Q (resp. VQ,) be the Q-vector subspace of V (resp.

V*) generated by the a (resp. the O'.-). Then V Q (resp. VQ,) is a Q-structure
on V (resp. V*) (Algebra, Chap. II,§ 8, no. 1). The restriction to VQ x VQ,
of the canonical bilinear form on V x V* gives an identification of each of the
spaces V Q, VQ, with the dv.al of the other. The set R is a root system in V Q,
If k = R, there exists a scalar product on V invariant under W(R) (In-
tegration, Chap. VII, § 3, no. 1, Prop. 1); Lemma 2 now shows that the a_
generate V*. By Remark 1, the O'.- again generate V* if k = Q. We now go
to the general case. Put E = VQ. By (RSm), each a· maps E to Q, and so
defines an element a of E*. It is immediate that R is a root system in E,
and that the element corresponding to a in E* is a. By what we said above,
the a generate the vector space E*. Consider the canonical homomorphism
i: E ©Q k --> V, and its transpose ti : V* -.. E* ©Q k. Since R. generates V,
ti is injective; but the image of ti contains the a, so ti is surjective. From
this we conclude finally that i and ti are isomorphisms. We can therefore
identify V with E © k, V* with E* © k, a- with ci, and VQ, with E*. Thus,
VQ (resp. VQ,) is a Q-structure on V (resp. V*). The restriction to VQ x VQ,
of the canonical bilinear form on V x V* can be identified with the canonical
bilinear form on E x E*, hence the proposition.
§ 1. ROOT SYSTEMS 157
Remarks. 2) Proposition 1 reduces the study of root systems to the case

k = Q. Remark 1 reduces it further to the study of root systems in the real
vector space VR = VQ ®QR. The Wey! groups associated to these different
systems are canonically identified.
3) Since the a- generate V*, the group W(R), considered as a subgroup
of GL(VR), is essential (Chap. V, § 3, no. 7). Moreover, the Cor. of Th. 1
of Chap. V, § 3, no. 2 shows that the only reflections belonging to W(R) are
the Sa.
PROPOSITION 2. The a· form a root system in V*, and cl - = a for all

a ER.
The a- satisfy (H.81) by Prop. 1. Since s 0 ,a· is an automorphism of the
vector space V equipped with the subset R, t(s 0 , 0 - ) - 1 leaves the set K of
the a· stable; but 1 (s 0 , 0 - ) - 1 = s 0 -, 0 , which proves that R: satisfies (RSu)
and that d =a. Finally, (o:-,{3) E Z for all a· E Kand /3 E R, so K satisfies
(RSm).
The set K is called the inverse root system of H.. The map a .-.. a· is
a bijection from R to K, called the canonical b·~jection from R to K. Note
that, if o:, /3 are elements of R such that a+ /3 E R, then (o + /3)" ¥-a-+ 13-
in general.
Since sa(a) =-a, axiom (RSn) shows that -R = R. Evidently (-a)"=
-a· and -1 E A(R) (but it is not always true that -1 E W(R)).
The equality t(s 0 , 0 · ) - 1 = s 0 -, 0 shows that the map u .-.. tu- 1 is an
isomorphism from the group W(R) to the group W(K). We identify these
two groups by means of this isomorphism; in other words, we consider W(R)
as acting both in V and in V*. Similarly for A(R).
PROPOSITION 3. For x, y E V, put
(xly) = L (a-,x)(o:-,y).
aER
Then (xly) is a non-degenerate symm.et1fr bilinear form. on V, invariant un-

der A(R). Forx,y E: VQ we have (xjy) E Q. The canon·ical e1:tension of(xly)
to
VR = Vq®qR
is non-degenerate and positive.
It is clear that (xly) is a symmetric bilinear form on V. If g E A(R),
(g(x)jg(y)) = L ( g(a:), x)tq(a·), y) = (x/y)

oER
1
since ( 1g)(R-) = R-. Ifx,y E VQ, then (xly) E Q by (RSrn). If z E VR, then
(zjz) = L (a·, z) 2 ;;::: 0, and (zlz) > 0 if z-::/: 0 by Prop. 1, so the canonical
°'ER
extension of (xly) to VR is positive and non-degenerate. The restriction of

(xly) to VQ is thus non-degenerate, and hence the form (x1y) on Vis non-
degenerate.
PROPOSITION 4. (i) Let X be a subset of R, let Vx be the vector snbspace of

v generated by x' and let v~ be the vector subspace of v· generated by the a -'
where a E X. Then V is the direct sum of V x and the orthogonal complement
o.fV~, V* is the direct sum of V~ and the orthogonal complement ofVx, and
v~ is ident~fied with the dual of v x.
(ii) H.nVx is a root system in Vx, and the canonical bijection.from RnVx
to its inverse root system is identified with the restriction of the map a ,_. a·
to Rn Vx.
By Remark 2, we can assume that k = R. Identify V with V' by means
of the symmetric bilinear form of Prop. 3. We have a· = (;~) for all a E R
(Lemma 2). Every vector subspace of V is non-isotropic, and the proposition
is now clear.
COROLLARY. Let V 1 be a vecto'f' subspace of V, and let V2 be I.he vector

sv.bspace generated by Rn V 1 . Then Rn V 1 ·is a rnot system in V2 .
This follows from (ii) applied to X = Rn V 1 •
For a, f3 E R., put
(o:,(J') = n(a:,/3). (1)
Then
n(o:, a) = 2 (2)
n(-o:,/3) = n(a, - (3) = -n(0'.,(3). (3)
By (RSm),
n(o:,{J) E Z. (4)
By the definition of n( 0:, /3),
sµ(a) = 0: -n(o,(3)(3. (5)
Formula (1) and Prop. 2 imply that
n(a,(3) = n(/3-,C<-). (6)
Let (xly) be a symmetric bilinear form on V, non-degenerate and invariant
under W(R) (Prop. 3). By Lemma 2,
2( 0:1/3)
n( a, (3) = (f3l/3) . (7)
It follows that
n(a., (3) = 0 ¢==;. n(/3, a) = 0 (01,8) = 0 ¢==;. 8"' and sµ commute. (8)
¢==;.
. n(/3, a) (Pl/3)
If (ol/3) :f: 0, then - (
n a,
(3) = -(-) .
a10
(9)
§ 1. ROOT SYSTEiV!S 159
2. DIRECT SUM OF ROOT SYSTEMS
Let V be a vector space over k that is the direct sum of a family (V;) 1 ~;~r
of vector spaces. Identify V* with the direct sum of the Vi. For all i, let R;
be a root system in V;. Then R = UR; is a root system in V whose inverse
system is R" = LJ R;·; the canonicai bijection from R to R. extends, for a.11
i, '
the canonical bijection from R; to R;.. The set R is called the d'irect sum
of the root systems Ri· Leto: E R;. If j -f. i, the kernel of a· contains V1 , so
Sa induces the identity on v.i; on the other hand, kn <;;; V;, so Sa leaves V.;
r
stable. These remarks show that W(R) can be identified with Il \.V(R;).
i=I
A root system R is said to be irreducible if R =/= 0 and if R is not the
direct sum of two non-empty root systems.
PROPOSITION 5. Let V be a vector space over k that is the direct sum of

vector spaces V 1 , ... , Vr. Let R be a root system in V. P'IJ,t R; = Rn V;. The
following three conditions a.re equivalent:
(i) the V.; are stable l/.nder \;!.,' ( R);
(ii) R<;;; V1UV2U···UV.r:
(iii) for all i, R; is a root system in V.;, and R is the direct snm of the R;.
(iii) ==> (i): this follows from wlrnt we said at the beginning of this mun-
ber.
(i) ==> (ii): assume that the V; are stable under W(R). Let a: E Rand
let H be the kernel of a« By Prop. 3 of Chap. V, § 2. no. 2, each V; is the
sum of a subspace of H and a subspace of kcr. Hence one of the V; contains
ko:, so 0: E V 1 UV 2 U · · · UV., ..
(ii)==> (iii): if condition (ii) is satisfied, R; generates V; for all i, so Riis
a root system in V; (Prop. 4). It is clear that R is the direct sum of the Ri·
COROLLARY. Let R be a. root system in V. The following condit-ions are

eqv.ivalent:
(i) R is irreducible;
(ii) the W(R)-rnodule V is S'imple;
(iii) the W(R)-module V is absolutely simple.
(ii) {=:::} (i): this follows from Prop. 5 and Maschke's Theorem (Chap. V,
Appendix, Prop. 2).
(iii) {=:::} (ii): this follows from Prop. 1 of Chap. V, § 2, no. 1.
PROPOSITION 6. Every root system R in V ·is the direct sum of a family

(~)iEI of irreducible root systems that is 'uniq·ue up to a b~jection of the index
set.
The existence of the R; is proved by induction on Card R: if R is non-
empty and not irreducible, R is the direct sum of two root systems R', R" such
that Card R' < Card R, Card R" < Card R, and the induction hypothesis
applies to R' and R". To prove uniqueness, it suffices to prove that, if R
is the direct sum of R' and R", every Ri is necessarily contained either in
R' or in R". Let V', V", v;, v;' be the vector subspaces of V generated by
R', R", R' n Ri, R" n Ri· Since the sum V' + V" is direct, the sum v; v;'
+ is
direct. Since Ri ~ R' U R", R; is the direct sum of the root systems R' n R;
and R" n Ri; hence either R' n R; = 0 or R" n Ri = 0, which proves the
assertion.
The Ri are called the irred·uC'ible components of R For any non-zero scalars
A;, the union of the .-\;Ri is a root system in V, whose inverse system is the
union of the \- 1 Ri·, and whose Wey! group is W(R).
PROPOSITION 7. Let R be a root system in V, (R;) the family of its ir-

reducible components, Vi the vector subspace of V generated by R;, B the
invariant symmetric bilinear form on V defined in Prop. 3, and B' a sym-
metric bilinear form on V invariant under W(R). Then the V; are pairwise
orthogonal with respect to B', and, for all i, the restrictions of B and B' to
vi are proportional.
If Vi EV;, Vj E Yj, i -:j:. j, and if w E W(Rj), then
B'(v;,w(vj)) = B'(v;,vi),
which shows that w(vJ) - VJ is orthogonal to v; with respect to B'. Since VJ

is irreducible for W(Rj), it is generated by the w( Vj) - Vj, and it is therefore
orthogonal to V;,
The fact that the restrictions of Band B' to each of the V; arc proportional
follows from Prop. 1 of Chap, V, fl 2, no. 1.
Remark. Choose a scalar product on VR invariant under W(R). It is then

possible to speak of the length of a root and the angle between two roots.
Prop. 7 shows that this angle is independent of the choice of scalar product,
as is the ratio of the lengths of two roots, provided they belong to the same
irreducible component of R.
3. RELATION BETWEEN TWO ROOTS
Recall that we ass·urne from now on that k = R. (We leave to the reader the
task of extending the definitions and results to the general case, by using the
method indicated in Remark 2 of no. 1.)
ThroughoiLt the fallowing, R denotes a root system in a vector space V;
and V is equipped with a scalar product (x, y) r--+ (xly) invariant ·under W(R),
cf. Prop. 3.
Let er., (3 E R. By formula (7) of no. 1,
n(o,(J)n((J,a) = 4cos 2 (et,{3) ~ 4. (10)

Thus, the integer n( n, /3)n(,8, o:) must take one of the values 0, 1, 2, 3, 4. In
view of Chap. V, § 2, no. 5, Cor. of Prop. 6, and of the footnote on the page
of Chap. V, § 4, no. 8, we see that the only possibilities are the following, up
to interchanging a and /3:
1) n(o:, /]) = n({3, o:) = O; (~) = ~; Sa.Sf:) of order 2;
2) n(a, ;3) = n({3, o.) = l; (Ct,/]) = i; SaSf3 of order 3;
I o: 11=11 /3 II;
3) n(o:,/3) = n((J,o:) = -1; (;)) = 2;; SaSf3 of order 3;
11~1=11/311;
4) n(a,(3) = 1, n(/3,n) = 2; (a,/]) = %; sas/3 of order 4;
I /3 II= v'2 I a II;
5) n(o,/3) = -1, n(/],a) = -2; (;)) = 3:; SaSf3 of order 4;
ll!jl= v'2 ii o.11;
6) n(o:,,8) = 1, n(/3,o:) = 3; (n,,6) = ~; Sasµ of order 6;
II f3 II= v3 II o: JI;
7) n(o:, /3) = -1, n(f], a) = -3; (;)) = 5;; SaSf3 of order 6;
II f3 II= V'3 ii a II;
8) n(a,/3) = n((J,o:) = 2; 0: = (3;
9) n(a,/3) = n((J,a) = -2; a= -/3;
10) n(a,/3) = 1, n({J,a) = 4; /3 = 2a;
11) n(o,(3) = -1, n(,8,o:) = -4; /3 = -2n.
In particular:
PROPOSITION 8. (i) ff' two roots are proportional, the .factor of proportion-
ality can only be ±1, ±~, ±2.
(ii) If a and f3 are two non-proportional roots, and if II a 11~11 /3 [[, then
n( a, (3) takes one o.f the valv.es 0, 1, -1.
If a root o: E R is such that ~o: ~ R, then o: is called an indivisible root.
THEOREJVI 1. Let o:, /3 be two roots.

(i) lfn(n,/3) > 0, a - /3 is a root unless a.= /3.
(ii) If n(a, /3) < 0, o: + /3 is a root unless n = -(3.
If n( a:, /3) > 0, the possibilities, by the list above, are the following:
1) n(a,/3) = l; then n -/3- s13(a) ER;
2) n(/3, a) = l; then (3 - n = sa(f3) ER, so o: - (3 E R;
3) (3 =a.
This proves (i), and (ii) follows by changing /3 to -/3.
162 ROOT SYSTEJVIS Ch. VI
COROLLARY. Let o and f3 be two roots.

(i) If (ol/3) > 0, ex - /3 is a root unless a= (3.
(ii) If (o:l,8) < 0, a+ (3 is a root unless o: = -/3.
(iii) If a - /3 t/. Ru {O} and a+ f3 t/. RU {O}, then (a1/3) = 0.
Assertions (i) and (ii) follow from Th. 1 and formula (7) of no. 1. Assertion
(iii) follows from (i) and (ii).
It is possible that o: + f3 ER, (a1/3) = 0 (cf. Plate X, System B 2 ). When
ex - f3 tf. RU {O} and o + (3 t/. Ru {O}, o: and f3 are s_aid to be strongly
orthogonal.
PROPOSITION 9. Let n and (3 be two non-proportional roots.

(i) The set I of integers j such that (3 + jo: is a root is an interval ( - q, p)
in Z containing 0.
(ii) Let S be the set of f3 +.in for j E I. Then,
s(\'.(S) = S and s 0 ((3 +po) = f3 - qn.
(iii) p-q = -n(/3,a).

Clearly, 0 E I. Let p (resp. - q) be the largest (resp. smallest) element of
I. If not all the integers in (-q,p) belong to I, there exist two integers r, s
in ( -q, p) with the following properties: s > r + 1, s E I, r E I, r + k tf. I for
1 ~ k: ~ s - r - 1. With the notation of the Cor. of Th. 1, (al/3 +so:) ~
0, (o:l/3 + rl\'.) ~ 0, which is absurd because
(od/3 +so:)~ 0 > (o:l/3 + rn).

This proves (i).
We have s(\'.((3+ jo:) = ;3-n(/3,a)a-ja = (3+ j'o: with .i' = - j - n((3,a).
Thus so:(S) ~ S and consequently s 0 (S) = S. Now .i ,__. ~ j - n(/3, a) is a
decreasing bijection from I to I. It follows that, j' = -q when j = p, so that
-q = -p - n((3, n). This proves (ii) and (iii).
The set S is called the a-chain of roots defined by (3, (3 - qn is its origin,
(3 +pa is its end, and p + q is its length.
COROLLARY. Let S be an a-chain of roots, o.nd "'/ the or'igin ofS. The length
of S is - n(1, o:); it is equal to 0, 1, 2 or 3.
The first assertion follows from Prop. 9, (iii), applied to (3 =/,and using
the fact that q = 0.
On the other hand, since / is not proportional to a, the list given at the
beginning of this no. shows that In(/, o:)I ~ 3, hence the corollary.
Remark. We retain the notation above. Then:

1) If the length of Sis 0, then (ab)= 0.
§ 1. ROOT SYSTErvJS 163
2) If the length of S is 1, then n( 7, a) = -1, and there are three cases:
n(a,-y) = -1, (o:la) = ('Yh), (ah)= -~(aJa), (~) = ~

n(a,-y) = -2, (ala)-= 2(-yh'), (ah)= -~(ala), (~) = '.?f
n(a,-y) = -3, (aJa) = 3(Jh), (al1) = -~(aJa), (;-:;) = ~
v+ a y y +a y Y +a
3) If the length of Sis 2, then n('Y, a)= -2, so

1 ~ 3rr
n(a,-y) = -1, (o:Ja) = 2hh), (o:h) = -(oio), (a,1) = 4·
~IX
/ t p-"\._
y y +a y + 2a
4) If the length of S is 3, theu n('Y, a) = -3, so

1 3 ~ 57r
n(a,7) = -1, (alo) = ;3(rh), (ah)= -2(0)0), (a,/')-= (j·
~a
/Jl3"{~
y y+ a y + ia y + 3 a
We shall see (Plate X, Systems A2 , B 2 , G2) that all these cases are actually
realised.
PROPOSITION 10. Let o:, /3 be two non-proportional roots sach that /3 + a

is a root. Let p, q be the inl.egers in Prop. 9. Then
(/3 + o:f,8 +a) q+1

(/31/J) p
Let S be the a-chain defined by /3, 1· its origin; its length l is :;:: 1 since
f3 +a is a root. The following cases are possible:
1) l = l; then f3 = -y,q = 0,p = 1, (/3 + u)/3 +a)= (/31/3).
2) l = 2, /3 = 1; then q = 0,p = 2, ((J + a)/3 + o:) = ~(,61,6).
3) l = 2, ,6 =I'+ o:; then q = l,p = 1, (,6 + a),6 +a)= 2(,8),6).
4) l = 3, /3=1; then q = 0,p = 3, ((3 + aJ(3 +a)= 1(.81(3).

5) l = 3, (3 = 'Y +a; then q = l,p = 2, (/3 + al/3 +a)= ((31,8).
6) l = 3, (3 ="I+ 2a; then q = 2,p = 1, ((3 + o:l/3 +a)= 3(/31/3).
In each case, the formula to be proved is satisfied.
PROPOSITION 11. Assume that R is irreducible. Let a and /3 be two roots

such that II a 11=11 (3 II· There exists g E W(R) such that g(a) = (3.
The transforms of o: by W(R) generate V (no. 2, Cor. of Prop. 5). Hence
there exists g E W(R) such that (g(a)J,B) "I 0. We assume from now on that
(a 1/3) i 0. By fonnula (9) of no. 1, n( a, ,B) = n((J, a). Replacing ,B if necessary
by Sfl (,B) = - (3, we can assume that n( a, ,B) > 0. Then, by the list at the
beginning of no. 3, either a = (3 (in which case the proposition is clear), or
n(a, (3) = n(,B, a) = 1; in that case
SaS(3Sa(f3) = SnS(3(/3 - a) = Sn(-/3 - 0: + /3) =a.
4. REDUCED ROOT SYSTEMS
A root system is said to be reduced if every root of the system is indivisible

(no. 3).
PROPOSITION 12. Ass·ume that R is irreduc·ible and reduced.

(i) The ratio ~~\~\ for o: E R, (3 E R must take one of the values
1,2,~,3,~.
(ii) The set of the (a10) for a E R has at most two elements.
Since R is irreducible, the transforms of a root by W(R) generate V (no. 2,
Cor. of Prop. 5). Hence, for any roots a, ,B, there exists a root (3 1 such that
(aj,B') "I 0 and (,B'l/3') = (/31/3). By formula (9) of no. 1 and the list of no. 3,
(f~i!;l takes one of the values 1, 2, 3,
~, ~ (recall that the system is assumed
to be reduced). By multiplying (xly) by a suitable scalar, we can assume that
(ajn) = 1 for certain roots and that the other possible values of (/31(3) for
(3 E R are 2 and 3. The values 2 and 3 cannot both be attained, since in that
case there would exist ,B E R, "( E R such that 1;:;~ = ~, contrary to what
we have seen above.
PROPOSITION 13. Assume that R is irreducible, non-reduced and of rank

;?: 2.
(i) The set Ro of indiV"isible roots is a root system ·i.n V; this system is
irreducible and reduced; and W(Ro) = W(R).
(ii) Let A be the set of roots a for which (ala) takes the smallest value A.
Then any two non-proportional elements of A are orthogonal.
(iii) Let B be the set of (3 E R su.ch that (/31/3) = 2A. Then B "I 0,
Ro = AU B, R =AU Bu 2A.
If o: E R - Ro, then ~u E R, but ~ Oa) rf- R (Prop. 8), so ~a E R 0 . This

proves that Ro satisfies (RSi). It is clear that, for all O' E R, sa,,,-(R0 ) = R 0 ,
so Ro satisfies (RSn) and (RSrn). Since a E R- Ila implies that ~a E Ro
and since Sa = sa; 2 , we have W(R) = W(Ro). Thus Ro is irreducible (Cor.
of Prop. 5), and it is evidently reduced.
Since R is not reduced, there exists O' E Ro such that 2a E R. Since
Ro is irreducible and dim V ?: 2, a cannot be proportional or orthogonal to
every root. Let (3 E R 0 be such that n((J, O') 'I 0 and (3 is not proportional
to O'. Changing (3 to -(3 if necessary, we can assume that n((J, a) > 0. Now
1n((3, a) = n((3, 2a) E Z, so n((3, a) E 2Z. From the list in no. 3, n((J, a) = 2,
((31,(3) = 2(alo:). Since Ro is reduced, Prop. 12 shows that, for all 'Y E R 0 ,
either hl'Y) = (ala:) or bh) = 2(afo:). Also, the above shows that, for all
'YER-Ro, the vector his an element of Ro such that (hlh) = (o:lo:).
Thus, A = (ala), B =f:. 0, Ro = AU B, and R ~ AU BU 2A; on the other
hand, if 'Y E A, there exists g E W(R) such that 'Y = g(a) (Prop. 11), so
2'Y = g(2a) E R; thus 2A ~ R and R = AU BU 2A. Finally, let 'Y, 'Y' be two
non-proportional elements of A. Then
n(2'Y, 'Y') = 2n('Y, 'Y') = 4n('Y, 2'Y') E 4Z, and /n('Y, 'Y')/ ::::;; 1
since 'Y and 'Y' have the same length, so n('Y, 'Y') = 0 and bl'Y') = 0.
PROPOSITION 14. Assume th.at R is irreducible and reduced, and that (aln)
takes the values A and 2-X for o: E R. Let A be the set of roots a such that
(adn) = .X. Assume that any two non-proportional elements of A are orthog-
onal. Then R 1 = RU 2A is an irreducible non-reduced root system and R is
the set of indivisible roots of R 1 .
It is clear that R 1 satisfies (RS1) and (RSm). We show that, if a., (3 E R 1 ,
then (a-,(:J) E z. This is clear if a.ER. Since (2at= 1a- a
for EA, it is also
immediate if a, /3 E 2A. Finally, assume that /3 E R and that a = 2'Y with
'YE A.
1) If'Y = ±/3, then (a-,/3) = ±~('Y-,'Y) = ±1.
2) If 'Y is not proportional to (3 and if (3 E A, the assumption on A implies
that ('Y~(3) = 0, so (o:-,/3) = 0.
3) If (3 E R - A, then ({:J/(3) = 2-X = 2('Yh ), so ((3, 'Y-) is equal to 0, 2 or
-2 by the list in no. 3. Thus ((3,o:-) = ~(f3,'Y-) E Z.
Thus R 1 is a root system in V, and the other assertions are clear.
5. CHAMBERS AND BASES OF ROOT SYSTEMS
For all a E R, let La be the hyperplane of V consisting of the points invariant

under Sa. The chambers in V determined by the set of the La (Chap. V, § 1,
no. 3) are called the chambers of R. The bijection V -> V* defined by the
scalar product (x/y) takes a to (a~f~-) for O'ER., hence La to La-, and hence
the chambers of R to those of W. If C is a chamber of R, the corresponding
chamber of K is denoted by c-. By Prop. 7 of no. 2, c- depends only on C

and not on the choice of (xly).
THEOREM 2. (i) The group W(R) acts simply-transitively on thP- set of

chambers.
(ii) Let C be a chamber. Then C is a fundamental domain for W(R).
(iii) C ·is an open simplicial cone (Chap. V, § 1, no. 6).
(iv) Let L 1 ,L 2 , ... ,L1 be the walls ofC. For all i, there exists a uniq·ue
·indivisible root ni sv.ch that L.; = L<>, and such that ai is on the same side of
Li as C.
(v) The set B(C) = {n1, ... , nt} is a basis of V.
(vi) C is the set of x E V such that (n,·, x) > 0 for all i (or, equivalenf;/y,
the set of x EV such that (xlo:i) > 0 for all i).
(vii) Let S be the set of the s<>;. The pair (W(R), S) is a Coxeter system
(Chap. IV,§ 1, no. 3).
Assertions (i) and (vii) follow from Chap. V, § 3, 110. 2, Th. 1. Assertion
(ii) follows from Chap. V, § 3, no. 3, Th. 2. Assertion (iv) is clear. The root
a; is orthogonal to L;, and a;· is identified with 2a:if(o:;lai). Since W(R) is
essential (no. 1, Remark 3), assertions (iii), (v) and (vi) follow from Chap. V,
§ 3, no. 9, Prop. 7.
Remarks. 1) Assertion (vii) shows in particular that W ( R) is generated by

the reflections 8a;.
2) If x, y E C,~en (xjy) > 0 (Chap. V, § 3, no. 5, Lemma 6), in other
words the angle (x, y) is acute.
3) Let m(a,(3) be the order of 8<>8f3 (n, (3 E B(C)). The matrix (m(a:, /J))
is identified with the Coxeter matrix (Chap. IV, § 1, no. 9) ~ (W, S). If
a i {J, Prop. 3 of Chap. V, § 3, no. 4 shows that the angle (a, (3) is equal
to Tt - _ · in particular, this angle is either obtuse or equal to rr, and
( nf3);
rn o:,
(aJ{J) ~ 0. By using the list in no. 3, it follows that m( a, {J) is equal to 2, 3, 4
or 6.
DEFINITION 2. A subset B of R is called a basis of R if there exists a

chamber C of R such that B = B(C). ff C is a chamber, B(C) is called the
basis of R defined by C.
Remarks. 4) Assertion (vi) of Th. 2 shows that the map C 1--> B(C) is a
bijection from the set of chambers to the set of bases. Consequently, W(R)
acts simply-transdively on the set of bases.
5) Let C be a chamber of R, and let B be the corresponding basis. If
a EB, put rp(a) = n· if 2a ~ Rand rp(a) = ~a:- if 2a E R. Then rp(B) is the
basis of R- defined by c-; this follows from the fact that the walls of c· are
the L"'·, for a EB.
DEFINITION 3. Let B be a basis of R. The Cartan matrix of R (relative to

B) is the matrix (n(o:, ,B))a,,BEB·
For all a: EB, n(a,o:) = 2. For o:,(J EB,
(o:lf:J) II a: II 7f
n(o:,{3) = 2(f31f3) = -2urrcos m(a,p)' (11)
where m(a:,,8) denotes as above the order of sus13. If n: f=- /3, n(a,fJ) = 0,
-1, -2 or -3 (cf. no. 3).
Remarks. 6) The Cartan matrix (n( a, /J)) should not be confused with the
Coxeter matrix (m(a:,,8)). Note in particular that the Cartan matrix is not
necessarily symmetric.
7) Canonical indexing. If B and B' are two bases of R, there exists a
unique element w E W such that w(B) = B'. \Ve have
n(w(a), w(/3)) = n(a,(3) and m(w(a), w(t3)) = m.(a:,(J)

for a, ,8 E B. Consequently, the Cartan and Coxeter matrices associated to B
can be obta.inecl from those associated to B' by composition with the bijection
a>---> w(a)
from B to B'.
The Cartan and Coxetcr matrices can actually be defined canonically in
the following way. Let X be the set of pairs (B, a), where B is a basis of R
and a E B. The group W acts in an obvious way on X and each orbit.of
W on X meets each of the sets {B} x B in exactly one point. If I is the set
of these orbits, each basis B admits a canonical inde:ring (ni)iEI· Moreover,
there exists a unique matrix N = ( nij) (resp. M = (mi.i)), of type I x I, such
that for any basis B, the Cartan (resp. Coxeter) matrix associated to B can
be obtained from N (resp. M) by composing with the canonical indexing of
B; it is called the canonical Cartan matri.x (resp. Coxeter matrix) of R.
PROPOSITION 15. Let B be a basis of R and a an indi'V'isible root. There

exist {3 EB and w E W(R) such that n = w(/3).
Let C be the chamber such that B = B(C). The hyperplane L0 is a wall
of a chamber C' of R, and there exists an element of W(R) that transforms
C' to C. We are therefore reduced to the case where L0 is a wall of C. Then o:
is proportional to an element f3 of R. Since a and f3 are indivisible, O' = ±,8.
If a= -(3, then a = Sf3((3), hence the proposition.
COROLLARY. Let R 1 and R.2 be two redv.ced root systems in vector spaces
V1 and V2, and let B 1 and B2 be bases of R1 and R2· Let f: B1--> B2 be a
bijection th.at transforms the Cartan matrix of R 1 to th.at of R2. Then there
exists an isomorphism F : V 1 --> V 2 that transforms R 1 to R2 and n to J(a)

for all a E B1.
Let F be the isomorphism from V 1 to V 2 that takes a to f (a) for all
a E B 1 . Then F transforms s 0 to SJ(a)• hence W(R 1) to W(R2) (Th. 2), and
hence R 1 to R2 (Prop. 15).
PROPOSITION 16. Let B be a basis of R, and G the subgroup of A(R)

consisting of the elements leaving B stable. Then W(R) is a normal subgroup
of A(R) and A(R) is the semi-direct product of G and W(_R).
If a E Rand t E A(R), then ts 0 C 1 = St(o.); since W(R) is generated
by the sa, we see that W(R) is a normal subgroup of A(R). By transport
of structure, A(R) transforms a basis of R to a basis of R. Since W(R) acts
simply-transitively on the set of bases, every clement of A(R) can be written
uniquely in the form 9192, where .91 E W(R) and 92 E G.
Remarks. 8) Let R 1, ... , Rp be root systems in vector spaces V 1 , ... , VP• R

the direct sum of the Ri in V =TI V;, C; a chamber of R;, and B; = B(G;).
It is immediate that C = TIC; is ' a chamber of Rand that B(C) = LJ B;. It
l 'l
follows from Th. 2 that all the chambers and bases of R are obtained in this
way.
6. POSITIVE ROOTS
Let C be a chamber of R, and let B(C) = {a 1, ... , at} be the corresponding

basis of R. The order relation on V (resp. V *) defined by C is the order
relation campatible with the vector space structure of V (resp. V*) for which
the elements ~ 0 a.re the linear combinations of the a; (resp. the an with
coefficients ~ 0. An element that is positive for one of these relations is said
to be positive for C, or positive for the basis B( C). These order relations are
also defined by c-, as one sees by identifying V with V* by using a scalar
product invariant under W(R). In view of Th. 2, no. 5, an element of V* is
~ 0 if and only if its values on C are ~ 0. An element x of V is ~ 0 if and
only if its values on c- are ~ 0, or, equivalently, if (xfy) ~ 0 for ally EC.
The elements of C are ~ 0 for C by Lemma 6 of Chap. V, § 3, no. 5.
But the set of elements ~ 0 for C is in general distinct from C (cf. Plate X,
Systems A2, B2, Gz).
THEOREM 3. Every root is a linear comb'ir1.ation wdh integer coefficients of

the same sign of elements of B(C). In particular, every root is either positive
or negative for C.
If a E R, the kernel La of a does not meet c-, so a is either > 0 on the
whole of c- or < 0 on the whole of c-, hence the second assertion. It remains
to show that a is contained in the subgroup P of V generated by B(C); we
can assume that o: is indivisible. Now the group P is clearly stable under the
s-y, for 1 E B(C), hence also under W(R) by Th. 2. Since o: is of the form
w(/3), with w E W(R) and f3 E B(C) (cf. Prop. 15), we have o: E P. Q.E.D.
Denote by R+ (C) the set of roots that are positive for C. Thus,
R = R+ (C) u (- R+ ( C))
is a partition of R.
COROLLARY. Let "/ be a linear combination of roots with integer coeffi-

cients, and a an indivisible root. If"/ is proportional to a, then"/ E Zo:.
By Prop. 15 of no. 5, C can be chosen so that o: E B(C). By Th. 3,
"/ = L nf3f3 with nf3 E Z.

{3EB(C)
Thus, if"/ is proportional to a, then "/ =n 0 0:, which proves the corollary.
Now let S be the set of reflections s<:> for n E B(C) and let T be the
union of the conjugates of S under W. For o: E B(C) and w E W, the
element t = ws 0 w- 1 of T is the orthogonal reflection Sf3 associated to the
root f3 = w(a); conversely, for any indivisible root (3, there exists an element
w E W such that a: = w- 1 ((3) E B( C) (Prop. 15) and Sf3 = ws 0 w- 1 E T.
It follows that a bijection '(j; from the set of indivisible roots to { ± 1} x T is
obtained by associating to an indivisible root f3 the pair (c:, Sf3), where c: = +l
if f3 is positive and c: = -1 if f3 is negative.
On the other hand, (W, S) is a Coxeter system (Th. 2) and the results of
Chap. IV, § 1, no. 4 can be applied. We have seen that, if w is an element
of W of length (with respect to S) equal to q, there exists a subset Tw of T,
with q elements, such that, if w = s 1 ... Sq with Si E S and if
(for 1 ~ i ~ q), then Tw = { t 1 , ... , tq}. Recall that we have also defined in
no. 4 of§ 1anumber17(w,t) (for w E Wand t ET) equal to +1 if t ~ Tw
and to -1 if t E Tw. Finally, recall that, if we define a map U w from the set
{±1} x T to itself by the formula
Uw(c:,t) = (c:77(w- 1 ,t),wtw- 1 ),
the map w >---> Uw is a homomorphism from W to the group of permutations
of the set {±1} x T (Chap. IV,§ 1, no. 4, Lemma 1).
PROPOSITION 17. Assume that R is reduced and let w E W and o: ER.

(i) We have 7/i(w(o:)) = Uw('tf;(o:)).
(ii) Assume that o: is positive. The root w(a) is negative if and only if
77(w- 1 , s 0 ) = -1,
in other words if Sa: E Tw-'.

(iii) We ha.very(w,sa) = -1 if and only if the chambers C andw(C) are

on opposite sides of the hyperplane L"'. In other words, the set Tw consists
of the reflections with respect to the walls separat·ing C and w(C).
Let f3 E B(C) and puts= sµ. Clearly Ts= {s} and consequently
U,(c,t)= {(E,sts- 1) ~f:t~s ( 12)

. (-E,8) 1f t = S.
On the other hand, let p = L n-y(Ph be a positive root. Put

-yEB{C) .
s(p) = L n-y(s(p))T.
-yEB(C)
If pi {3, there exists an element"( E B(C) with "f i {3, such that n-y(p) > 0,
and we have n-y(s(p)) = n-y(p) > 0 (no. 1, formula (5)). Hence s(p) is positive.
We deduce immediately that.
.t,( (- )) = { (E,SSpS- 1 ) if p # j3 ( 13)

9· s c .p ( -E,S ) 1'f -
p- f3 .
Comparison of (12) and (13) now shows that Us(-1{1("1)) = ·l{l(s("Y)) for all roots
"I and alls ES. Since S generates W, (i) follows.
On the other hand, saying that w(n:) is negative is equivalent to saying
that
'l/'(w(o:)) = (-1, WSaW- 1 ),
or, by (i), that Uw(4;(o:)) = (-l,ws"'w- 1 ). If in addition a is positive, then
·l{l(o.) = ( +1, Sn) and Uw(1f.1(a)) = (17(w- 1, Sn), ws,;w- 1 ), hence (ii).
Finally, by (ii), 11(w, s<>) = -1 if and only if one of the roots a and
w- 1 (a) is positive and the other negative. This is equivalent to saying that
(ajx).(w- 1 (0)1x) = (a1x).(n1w(x)) < 0 for all x EC, hence the first assertion
in (iii). The second assertion in (iii) follows immediately. ·
COROLLARY 1. Let fJ E B(C). The reffoction S(J permutes the positive roots
not proport'ional to (3.
vVc reduce immediately to the case in which R. is reduced. In that case,
our assertion follows from (ii) and the fact that Tsr; = Sfi·
COROLLARY 2. Ass1lTne that R is reduced. Let w E W, let q be the length of

w with respect to S (Chap. IV, § 1, no. 1), and let w = s 1 ... Sq be a reduced
decomposition of w. Let a 1 , ... , O'.q be the elements of B (C) correspond·ing to
s 1 ,. .. ,sq. Put
(}i=SqSq-1···8-i.+l(a.;), i=l, ... ,q.

The roots 8-; are > 0, pairwise distinct, w(8;) < 0, and every root a> 0 such
that w ((\') < 0 is equal to one of the ei.
§ l. ROOT SYSTEMS 171
Let X be the set of O'. > 0 such that w(a:) < 0. By (ii),
Card(X) = Card(Tw-1) = l(w- 1 ) = l(w) = q.
On the other hand, if n E X it is clear that there exists i E (1, q) such that
S;+1 ... sq(o:) > 0 and siS;+1 ... sq(a) < 0.

By Cor. 1, this implies that .s\s;+1 ... Sq(o:) = O:; and hence that a= e;. The
set X is thus contained in the set of O;. Since Carcl(X) = q, this is possible
only if X is equal to the set of B; and these are pairwise distinct. Hence the
corollary.
COROLLARY 3. Assume that R is reduced. There exists a unique longest

element wo in W. Its length ·is equal to the number of positive roots and w 0
transforms the chamber C to -C. We have w6
= 1 and l(wwo) = l(w 0 ) -l(w)
for all w E W.
It is clear that -C is a chamber. Hence there exists an element w 0 of W
that transforms C to -C. Then w 0 (a) < 0 for all positive roots a and the
first two assertions of Cor. 3 are immediate consequences of Cor. 2. We have
w6(C) = C, so w6 = 1. Finally, if w E W, the length l(w) (resp. l(ww 0 )) is
equal, by Prop. 17 (iii), to the number of walls separating C and w(C) (resp.
ww 0 (C) = -w(C)). Since w(C) and -w(C) are on opposite sides of every
wall, the sum l(w) + l(w 0 ) is equal to the total number of walls, that is to
l(wo).
PROPOSITION 18. Let x EV. The following three properties are equivalent:
(i) x EC;
(ii) x ~ s,,(x) for all C\' E B(C) (with respect to the order relation defined
by C};
(iii) .T ~ w(x) for all w E W.
Since s°'(x) = x - (x, o:")o: and since C is the set of elements x EV such
that (x, an
~ 0 for all a E B(C), the equivalence of (i) and (ii) is obvious. On
the other hancl, it is clear that (iii) ==} (ii). We show that (i) ==} (iii). Let
x EC, and let w E W. We argue by induction on the length l(w) of w. The
case l(w) = 0 is trivial. If l(w) ~ 1, w can be written in the form w = w's°',
with a E B(C) and l(w') = l(w) - 1. Then
x - w(.r) = x - w'(x) + w'(x - sa(x)).

The induction hypothesis shows that .r-w'(x) is positive. On the other hand,
w'(x - s 0 (x)) = w(sn(.r) - x) = -(x, a.·)w(a).
Now So: E Tw-t, and Prop. 17 (ii) shows that w(a) < 0. Hence the result.
COROLLARY. An element x E C if and only if :r: > w(x) for all w E W

such that w :/; 1.
PROPOSITION 19. Let (.Bi)i,,;i,,;n be a sequnce of positive roots for the cham-
ber C such that /31 + ,82 + · · · + .Bn is a root. Then there exists a permutation
7r E 6n such that, for all i E {1, 2, ... , n}, .Btr(l) + .Brr(2) + · · · + .B"(iJ is a root.
We argue by induction on n, the proposition being clear for n ~ 2. Put

n
.B = .81 + · · · + .Bn· Then L (.Bl.Bi)
i=l
> 0, so there e:xists an index k
= (.Bl.B)
such that (.Bl.Bk) > 0. If .B = .Bki then n = 1 since .Bi > 0 for all i. Otherwise
.B - .Bk is a root (no. 3, Cor. of Th. 1); it then suffices to apply the induction
hypothesis to fJ - .Bk = L .Bi·
if.k
COROLLARY l. Let a: E R+(C). Then a: E B(C) if and only ~la: is the sum
of two positive roots.
If a: is the sum of two positive roots, Th. 3 shows that a: E B(C). If
n
a:¢ B(C), Th. 3 shows that a:= L .Bk .with .Bk E B(C) for all k and n ~ 2.
k=l
n-1
Permuting the .Bk if necessary, we can assume that L is a root (Prop. 19),
k=l
n-l
and hence that a: is the sum of the positive roots L .Bk and ,B.,,.
k=l
COROLLARY 2. Let cp be a map from R to a abelian group I' having the

following properties:
1) cp(-a:) = -cp(a:) for a: ER;
2) ~la: ER, .BER are such that a:+.B ER, then cp(a+ .B) = cp(a:) + cp(,B).
Let Q be the subgroup of V generated by R. Then cp extends to a homomor-
phism from Q to r.
Let B be a basis of R. Let 1/.! be the unique homomorphism from Q to I'
that coincides with cp on B. It suffices to show that 1/J(a:) = cp(a:) when a: is
a positive root relative to B. We have a: = ,81 + · · · + .Bm with .Bi E B for all
i, and ,81 + · · · + .Bh ER for all h (Prop. 19). We show that 'tf;(a:) = cp(a:) by
induction on m. This is clear if m = 1. The induction hypothesis gives
1/J(.81 + · · · + .Bm-1) = cp(,81 + ••• + .Bm-1),

and we have 'lf1(.Bm) = cp(.Bm), hence 1/.!(a:) = cp(a:), which proves the corollary.
For any root a: = L nf3,B in R, denote by Y(a) the set of .B E B(C)
{3EB(C)
such that nf3 i 0. Moreover, observe that B(C) can be identified with the set
of vertices of the graph of the Coxeter system formed by W(R) and the sa.
(cf. Chap. IV,§ 1, no. 9 and Chap. V, §3, no. 2).
COROLLARY 3. a) Let a E R. Then Y(a) is a connected subset of B(C)

(Chap. IV, Appendix).
b) Let Y be a non-empty connected subset of B (C). Then L f3 belongs
/3EY
to R.
To prove a), we can assume that a is positive. We argue by induction on
Card(Y(a)), the assertion being trivial if Card(Y(a)) = l. By Prop. 19, there
exists /3 E B(C) such that a - f3 E R. Let p be the largest integer ?: 0 such
that/= a - p/3 E R. Since/ - /3 i Rand/+ p{J E R, (r){J) I 0 (Prop. 9);
thus /3 is linked to at least one element of Y(r). But Y(a) = Y(r) U {/3},
and Y(r) is connected by the induction hypothesis. Thus Y(a) is connected,
which proves a).
Now let Y be a non-empty connected subset of B(C); we show by induction
on Card(Y) that L /3 is a root. The case in which Card(Y) ::;:;; 1 is trivial.
/3EY
Assume that Card(Y) :;::: 2. Since X is a forest (Chap. V, § 4, no. 8, Prop. 8),
Y is a tree and has a terminal vertex /3 (Chap. IV, Appendix). The set
Y - {/3} is connected, and one of its element is linked to /3. By the induction
hypothesis, a= L /ER, and since (al/3) < 0, it follows that a+/3 ER
~1EY -{/3}
(Th. 1). Q.E.D.
7. CLOSED SETS OF ROOTS
DEFINITION 4. Let P be a snbset of R.

(i) Pis said to be closed if the conditions a E P,/3 E P,a + f3 ER imply
a+/3EP.
(ii) P is said to be parabol'ic if P is closed and if PU ( - P) = R.
(iii) Pis said to be symmetric if P = -P.
Lemma 3. Let C be a chamber of R and P a closed subset of R containing

R+(C) (in the notation of no. 6). Let E = B(C) n (-P), and let Q be the
set of roots that are linear combinations of elements of E with non-positive
integer coefficient. Then, P = R+ (C) U Q.
It is enough to show that P n ( - R+ ( C)) = Q. Let -a E Q. Then a is
the sum of n elements of E. We show, by induction on n, that -a E P.
This is clear if n = 1. If n > 1, then by Prop. 19 of no. 6 we can write
a = /3 + / with / E E and f3 the sum of n - 1 elements of E. By the
induction hypothesis, -/3 E P; since -/ E P and since P is closed, -a E P.
Thus, Qi; P n (-R+(C)). Conversely, let -a E P n (-R+(C)). Then a is
the sum of p elements of B(C). We show, by induction on p, that -a E Q.
This is clear if p = 1. If p > 1, then by Prop. 19, we can write a = f3 + /
with/ E B(C) and f3 a root that is the sum of p - 1 elements of B(C). Since
_-1 = /3 + (-a) and since P is closed, -1 E P, hence / E E. Moreover,
-/3 = / + (-a:) so -/3 E P since P is closed. By the induction hypothesis,

-/3 E Q, so -a:= -/3- / E Q. Thus, Pn (-R+(C)) <:.::; Q.
PROPOSITION 20. Let P be a subset of R. The following conditions are

equivalent:
(i) P is parabolic;
(ii) Pis closed and there exists a chamber C o.f R such that P ~ R+(C);
(iii) there exist a chamber C of R and a subset E of B(C) .mch that P
is the un'ion of R+ ( C) and the set Q of roots that are linear_combino.t'ions of
elements of E with non-positive integer coefficients. ·
(ii) ~ (iii): this follows from Lemma 3.
(iii) ==> (i): we adopt the assumptions and notatio11 of (iii). It is clear
that PU ( - P) = R. We show that, if a:, /3 E P are such that a: + /3 E R, then
a+ (3 E P. This is obvious if the root n + fJ is positive. Assume that a+ /3 is
negative. Then a+ /3 = L n-y/, with n-y ~ 0. But the coefficient of every
-yEI3(C)
element "/ of B(C)- E in o: or /3 is ~ O; hence n-y = 0 if/ E B(C)- E, so
a+ /3 E Q <:.::; P.
(i) ==> (ii): assume that P is parabolic. Let C be a chamber such that
Card(PnR+(C)) is as large as possible. Leto: E Il(C) and assume that ex E P,
so that -a E P. For all /3 E P n R+(C), /3 is not proportional to ex (for the
hypothesis /3 = 2n would imply that o: = 2a + (-a) E P since P is closed).
Thus so: (/3) E R+ (C) (no. 6, Cor. 1 of Prop. 17). If we put C' = s(\ (C),
then /3 = so:(s"'(/3)) E so:(R+(C)) = R+(C), so -a E P n R+(C') and hence
Carc\(P n R+(C')) > Card(P n R+(C)). This is absurd, since o: E P. Thus
B(C) <:.::; P, and consequently R+(C) <:.::; P by Prop. 19 and the fact that P is
closed.
COROLLARY 1. Let P be a subset of R. The following conditions are equiv-

alent:
(i) there exists a chamber C such that P = R+ (C);
(ii) Pis closed and {P, -P} is a partition ofR.
The chamber C s·uch that P = R+ (C) is then unique.
If P = R+(C), C- is the set of x* EV* such that (x*, x) > 0 for all x E P,
hence the uniqueness of C.
COROLLARY 2. Assume that V is equipped with the structure of an ordered

vector space such that, for this strnctnre, every root of R is either positive
or negative. Let P be the set of positive roots for this struci'ure. Then there
e.rists a v.nique chamber C of R such that P = R+ (C).
Indeed, P satisfies condition (ii) of Cor. 1.
This corollary applies in particular when the order being considered is
total, the condition on R then being automatically satisfied. Recall that such
§ I. ROOT SYSTEMS 175
an order can be obtained, for example, by choosing a.basis (e;)i"'i"'n of V

and taking the lexicographic order on V, so.<: = L ~ie; is ~ 0 if all the{; are
0, or if ~i > 0 for the smallei:;t index i such that ~;' i 0.
COROLLARY 3. A subset B of R is a basis of R if and only if the following

conditions are satisfied:
(i) the elements of B are linearly independent;
(ii) every root of R is a linear combination of elements of B in which the
coefficients are either all pos'itive or all negative:
(iii) every root of B is indivisible.
We already know that the conditions are necessary (no. 5, Th. 2, and
no. 6, Th. 3). Assume that conditions (i), (ii), (iii) are satisfied. Let P be the
set of roots that are linear combinations of elements of B with coefficients
~ 0. Since P satisfies condition (ii) of Cor. 1, there exists a chamber C such
that P = R+ (C); let B' = B (C), and let X and X' be the convex cones
generated by B and B'. Then
B <;;: P <;;: X and B' <;;: P <;;: X',
which shows that X and X' are both generated by P, and hence coincide. But
the half-lines generated by the elements of B (resp. by B') are the extreme
generators of X (resp. X'); since such a half-line contains only one indivisible
root, B = B'.
COROLLARY 4. Let B be a basis of R, B' a subset of B, V' the vector

s~tbspace of V generated by B', and R' = Rn V'. Then B' is a basis of the
root system B'.
This follows immediately from Cor. 3 and the Cor. of Prop. 4.
We call R' the root system generated by B'.
COROLLARY 5. Let B be a basis of R, A 1 , A2 , ... , Ar pairwise orthogonal

subsets of B, and A = A1 U A2 U · · · U A, .. Then every root o that is a linear
combination of elements of A is actually a linear combination of elements of
one of the A.;. In particular, ~f R ·is irreducible, there is no partition of B into
pairwise orthogonal subsets.
Let E 1 , ... , Er, Ebe the vector subspaces of V generated by A 1 , ... , Ar, A,
respectively. By Cor. 4, we can assume that E = V. Then, by Th. 2 (vii) of
no. 5, the Ei are stable under W(R), so R is the union of the Rn E; (no. 2,
Prop. 5).
COROLLARY 6. We adopt the hypotheses o.nd notation of Prop. 20. Let V 1

be the vector subspace of V generated by E. Then P n (-P) = Q u (-Q) =
Vi n R is a root system in V1 with basis E.
We have P n (-P) = (R+(C) u Q) n ((-R+(C)) u (-Q)) = Qu ( -Q). Th.

3 proves that Q U (-Q) = V 1 n R. Finally, E is a basis of the root system
V 1 n R by Cor. 4.
PROPOSITION 21. Let C (resp. C') be a chamber of R, E {resp. E') a subset

o.f B(C) (resp. B(C')), Q (resp. Q') the set of linear combinations of elements
of E (resp. E') with negative integer coefficients, and P = Q U R+(C) (resp.
P' = Q' UR+ (C')). If there exists an element of the W eyl group trans.forming
P to P', then there e.Tists an element of the Weyl group trans.forming C to C'
and E to E'. -
We reduce immediately to the case P = P'. Let V 1 be the vector subspace
of V generated by P n ( - P). Then E and E' arc bases of the root system
R 1 = P n (- P) in V 1 (Cor. 6 of Prop. 20). Hence there exists g 1 E W(Ri)
such that g 1 (E) = E'. It is clear that g 1 is induced by an element g of W(R)
that is a product of the symmetries Su with a E E. Let "( = L: Cf3{3 be
/3EB(C)
an element of P- R 1 . Then c,13 > 0 for at least one f3 E B(C)- E. Moreover,
if a E E, then Sa b) - 'Y E V 1 , so Sa b) has at least one coordinate > 0
with respect to B(C) (no. 1, formula (5)), hence sab) E R+(C) and finally
sub) E P-R 1 . It follows that P-R 1 is stable under the Sa, a E E, and
hence under g, so g(P) = P. We are thus reduced to proving the proposition
when P = P' and E = E'. In this case, Q = Q', so R+ (C) = P-Q = P-Q'
= R+ (C'), and hence C = C' (Cor. 1 of Prop. 20).
COROLLARY. Let P, P' be two parabolic subsets of R transformed into each

other by an element of the Weyl group. ff there exists a chamber C of R such
that R+(C) c;;; P and R+(C) ~ P', then P = P'.
This follows from Lemma 3 and Prop. 21 since the only element of W(R)
transforming C to C is 1, cf. no. 5, Th. 2.
PROPOSITION 22. Let P be a closed subset of R such that P n (- P) = 0.

Then there exists a chamber C of R such that P c;;; R+ (C).
1) In view of the Cor. of Th. 1, no. 3, the assumptions n E P, f3 E P,
(a1f3) < 0 imply that o: + f3 E P.
2) We show that no sum o: 1 + · · · + o:q (q ~ 1) of elements of P is zero.
We proceed by induction on q. The assertion being clear for q = 1, assume
that q ~ 2. If o: 1 + · · · + O'q = 0, then
- 0'1 = 0:2 + ... + O'.q,

so ( -o: 1 ja 2 + · · ·+o:q) > 0, hence there exists j E (2, q) such that (n 1 lnj) < 0.
By part 1) of the proof, a 1 + O'.j E P, and the relation (n 1 + <Y-j) + .L: .n; =0
i;il ,)
contradicts the induction hypothesis.
3) We show that there exists a non-zero element/ in V such that bin) ? 0

for all a E P. If not, the result of 1) would show that an infinite sequence
o: 1 , o: 2 , ... of elements of P could be found such that
f3i = 0:1 + · · · + O:; E P

for all i; there would exist two distinct integers i, j such that fJi = (Jj, which
would contradict the result of 2).
4) To prove the proposition, it is enough (Cor. 2 of Prop. 20) to show
that there exists a basis (o:k)i,;;;k,;;;t of V such that, for the lexicographic order
defined by this basis, every element of P is > 0. We proceed by induction on
l = dim V, and assume that the proposition is established for all dimensions
< l. Let/ E V be such that/ i 0 and ('ylo:) ? 0 for all a E P (cf. 3)). Let
L be the hyperplane orthogonal to/, and V' the subspace of L generated by
Rn L. Then R n L is a root system in V' and P n L is closed in R n L. By
the induction hypothesis, there exists a basis ((3 1 , ..• , /311) of V' such that the
elements of P n L are > 0 for the lexicographic order defined by this basis.
Then any basis of V whose first l' + 1 elements are /, (3 1 , ... , fit' and whose
remaining elements are in L has the required property.
PROPOSITION 23. Let P be a subset of R and V 1 (resp. I') the vector

subspace (resp. the subgroup) of V generated by P. The following conditions
are equivalent:
(i) P is closed and symmetric;
(ii) Pis closed, and Pis a root system in V 1 ;
(iii) I' n R = P.
Assume that these conditions are satisfied. For any n E P, let a 1- be the
restriction of a- to V 1 . Then the map a,......, 0: 1- is the canonical bijection from
the root system P to p-_
(iii) => (i): clear.
(i) => (ii): assume that P is closed and symmetric. First, P satisfies (RS 1 )
in V 1 . We show that, if a, (3 E P, then Sa ((3) E P. This is clear if 0t and {3
are proportional. Otherwise, s0 (/3) = f3 - n(/3, a)o: and f3 - pa E R for all
rational integers p between 0 and n(/3, a) (Prop. 9, no. 3), so
f3 - n(/3, 0t)0t E P
since P is closed and symmetric. Thus, s 0 , 01 -(P) = P, and P satisfies (RSu).
It is clear that P satisfies (RSrn). Thus, P satisfies (ii), and we have proved
the last assertion of the proposition at the same time.
(ii) =;. (iii): we show that, if condition (ii) is satisfied, then I' n R = P.
It is clear that P ~ I' n R. Let f3 E I' n R.. Since f3 E I' and P = -P,
f3 = 0t1 + 0t2 + · · · + Olk with o: 1 , ... , Olk E P. We shall prove that /3 E P. This
is clear if k = 1. We argue by induction on k. We have
k
L: (,Blai),
o < (,Gl,8) = i=l
so (,Bini)> 0 for some index i. If /3 =a;, then /3 E P. Otherwise, ,8 - o:; E Il
(Cor. of Th. 1, no. 3), so /3-a.i E P by the induction hypothesis, hence f3 E P
since P is closed.
The conditions of Prop. 23 can be realised with V 1 = V and yet P f R. For
example, this is the case when R is a system of type G2 and P a system of type
A2; cf. Plate X.
PROPOSITION 24. Let R' be the intersection of R with a vector subspace of

V, so that R' is a root system in the vector subspace V' that it generates (cf.
Cor. of Prop. 4, no. 1). Let B' be a basis of R'.
(i) There e.rists a baS'is of R containing B'.
(ii) R' is the set of elements of R that are linear combinations of elements
ofB'.
Assertion (ii) is clear. We prove (i). Let (c 1 , c 2 , ... , Et) be a basis of V such
that B' = (t:p+ 1 , Ep+ 2 , ... , c1). The lexicographic order on V corresponding
to this basis defines a chamber C of R. It is clear that every element of B' is
minimal in R+(C). Thus B' c;:_ B(C).
8. HIGHEST ROOT
PROPOSITION 25. Assume that R is irreducible. Let C be a chamber of R,

and let B( C) = {a. 1 , ... , 01} be the corresponding basis.
l l
(i) There exists a root Ci = L nia:i
i=l
such that, for every root L Pia.;,
i=l
we
have n1 ~ p 1 ,n2 ~ P2, .. .. n1 ~Pl· In other words, R has a largest element
for the ordering de.fined by C.
(ii) We have Ci EC.
(iii) We have (a I Ci)~ (a I a) for every root a.
(iv) For every positive root a' not proport'iona.l to Ci, we have n(a',Ci) = 0
or 1.
l l
1) Let a= L n;a;,
i=l
,8 = L p.;a.,.
i=l
be two maximal roots for the ordering
.
defined by C. We shall prove that a= {J, which will establish (i).
2) If (a lo:i) < 0 for some index i, it follows that either a+ O'.; E R or
a: = -a; (Cor. of Th. 1, no. 3), and both possibilities arc absurd by the
maximality of a. Thus (a I a;) ~ 0 for all i.
3) If () < 0, then a < -a, which is absurd. Thus n.; ~ 0 for all i. Let J
be the set of i such that n; > 0, and J' the complement of J in { 1, 2, ... , l}.
Then .J :f 0. If J' were non-empty, there would exist an i E J and an i' E J'
such that (a; I a:;') < 0 (Cor. 5 of Prop. 20, no. 7); we would then have
(o: IC\:i') = L
iEI
n;(a; I0'.; 1 ) <0
since (aj Ink) :;:;; 0 whenever j and k are distinct, which would contradict 2).
Thus J' = 0 and n; > 0 for all i.
4) We have (fJ Io:;) ? 0 for all i by 2). We cannot have (/JI o;) = 0 for all
i since /J =J 0. We deduce from 3) that
((1 I a) =L i
n; ((31 o:;) > 0.
If/ = o: - (3 E R, either a > f3 or f3 > 0: (Th. 3, no. 6), which contradicts

the maximality of a: and fJ. Thus o: = (3 (Cor. of Th. 1, no. 3).
5) By 2), ii E C. We shall prove that (o:' Io:') :;:;; (ii I&) for every c/ E R..
Since C is a fundamental domain for W(R), we can assume that a' EC. We
have ex - 0: 1 ;;:: 0, so (a - c/ Ix) ;;:: 0 for all :c EC. In particular (0: - ci I a) ;;:: 0
and (ii-0'10:')? 0, hence (Oil ii)? (0:'10:)? (o'la'). Thus n(o:',o:) must
be equal to 0, 1 or -1 if o:' is not proportional to 0:. If c/ ;;:: 0, then (0: Ia/) ;;:: 0
by 2), so n(o:', 0:) ;;:: 0 and n(o:', a) must be either 0 or l. Q.E.D.
Remark. The root
in (i) is said to be the highest rnot of R (with respect to C). Note that, by
(i), we haven; ;;:: 1 for all i.
9. WEIGHTS, RADICAL WEIGHTS
Let l = dim V. Denote by Q(R) the subgroup of V generated by R; the

elements of Q(R) are called the radical weights of R. By Th. 3 of no. 6, Q(R)
is a discrete subgroup of V of rank l, ancl every basis of R. is a basis of Q(R).
Similarly, the group Q(R") is a discrete subgroup of V* of rank l.
PROPOSITION 26. The set of x E V s1tch that (x, y*) E Z for all y* E
Q(R") (or, equivalently, for all y* E R") is a discrete subgroup G of V con-
taining Q(R). ff B' is a basis of R', the basis of V d11.al to B' is a basis of
G.
Let x E V. The following three properties are equivalent:

(i) (x,y*) E Z for ally* E Q(R:);
(ii) (x,y*) E Z for ally* EB';
(iii) the coordinates of x with respect to the dual basis of B' are in Z.
We deduce from this that the basis dual to B' is a basis of G. On the other
hand, (RSm) proves that R ~ G, so Q(R) ~ G.
The group G of Prop. 26 is denoted by P(R), and its elements are called the
weights of R. We can also consider the group P(R·) of weights of R'.
By Algebra, Chap. VII, 2nd edn., § 4, no. 8,
P(R)/Q(R), P(K)/Q(K)
are finite groups in duality over Q/Z, and hence are isomorphic. The common
order of these two groups is called the connection index of R (or of K).
If R is a direct sum of root systems R;, the group Q(R) (resp. P(R)) is
identified canonically with the direct sum of the Q(R;) (resp. P(Ri)).
PROPOSITION 27. Let R1 be a subset of R, Q 1 the subgro·up of Q(R) gen-

erated by R 1 , and W 1 the subgroup of W(R) generated by the Sa (a E R 1).
If p E P(R) and w E W1, then p - w(p) E Q1.
If w =Sa with a E R1, then
If w = S0: 1 Sa 2 ••• Sa,., with 0.1, ... , O.r E R1, it is still true that p- w(p) E Q1,
as we see by induction on r.
The group A(R) leaves P(R) and Q(R) invariant, and hence acts on the quo-
tient P(R)/Q(R). By Prop. 27, the group W(R) acts trivially on P(R)/Q(R).
Passing to the quotient, we see that the q·uot-ient group A(R)/W(R) (cf. Prop.
16, no. 5) acts canonically on P(R)/Q(R).
10. FUNDAMENTAL WEIGHTS, DOMINANT WEIGHTS
Assume that R is red·uced. Let C be a chamber of R, and let B be the corre-

sponding basis of R. Since R is reduced, B- = {a-}a:EB is a basis of K. The
dual bas·is (wa)aEB of Wis thus a basis of the group of weights; its elements
are called the fundamental weights (relative to B, or to C); if the elements
of B are denoted by (0' 1, ... , a1), the corresponding fundamental weights are
denoted by (w1,' .. 'wi).
Let :c E V. Then, x E C if and only if (.i:, a:') > 0 for all a E B. It follows
that C is the set of linear combinations of the Wa with coefficients ~ 0.
The entries n( O', (3) = (o, (3-) of the Cartan matrix are, for fixed o, the
coordinates of a with respect to the basis (w,e),BEB:
a= I:
,BEB
n(D',(3)w,e. (14)
The Cartan matrix is thus the transpose of the matrix of the canonical in-
jection
Q(R) ~ P(R)
with respect to the bases Band (wa)0:EB of the Z-modules Q(R) and P(R).
A weight w is said to be dominant if it belongs to C, in other words if
its coordinates with respect to (wa )aEB are integers ~ 0, or equivalently it'
g(w) ~ w for all g E W(R) (no. 6, Prop. 18). Since C is a fundamental domain
for W(R) (Th. 2), there exists, for any weight w, a unique weight w' such that
w' is a transform of w by W(R.).
We have
\wo, rn = (wo I (f32if3)) = bo{3
for o:, fJ E B ( c50 f3 denoting the Kronecker symbol), hence
In other words, w0 is orthogonal to f3 for f3 ¥ a . and its orthogonal projection

on~-~:-Ra: is ~o:. Since w0 E C, (wn IWf3) ~ 0 for er., /3 E B, i.e. the angle
(w 0 , w13) is acute or a right angle. The dominant weights are thew E V such
that 2(w I n)/(o: Io:) is an integer ~ 0 for all a: EB.
PROPOSITION 28. Let B be a basis of R, B' a subset of B, V' the vector

subspace of V gp_ncrated by B', R' = R n V' (which is a root system in V'),
R'' the inverse root system {which is identified with the canonical image of
R' in R'), V1 the orthogonal complement of R.1 • in V, and p the projection of
V onto V' parallel to V1 . Then, Q(R') = Q(R) n V', P(R') = p(P(R)). The
set of dominant weights of R.' -is the image under p of the set of dominant
weights of R..
Indeed, Q(R) is the subgroup of V with basis B, Q(R') is the subgroup of
V' with basis B' (no. 7, Cor. 4 of Prop. 20) from which Q(R') = Q(R) fl V' is
irnrnediate. lfw E P(R) and a ER', {p(w), a·)= {w, a:·) E Z, so p(w) E P(R'),
and hence p(P(R)) <;;;, P(R'). If w' E P(R'), w' ext.ends to a linear form won
V* vanishing on (B-B't; we have \w, a:·) E Z for all o EB, sow E P(R), and
w' = p(w); hence P(R') <;;;, p(P(R.)). Thus P(R') = p(P(R)), and the assertion
about dominant weights is proved in the same way.
PROPOSITION 29. Let p be hail the sum of the roots > 0.

(i) p = L w 0 ; this ·is an element of C.
oEB
(ii) sa(P) = p - o: for all l1' EB.
(iii) (2p I a) = (a I a) for all o EB.
Since R. is reduced, sa(R+ (C) - {a}) = R+(C) - {a} and s 0 (a) = -a:
for o E B (no. 6, Cor. 1 of Prop. 17), so s0 (2p) = 2p - 2c:x. Since sa(P) =
p - (p, a·), we see that
(p, a·) = 1 = {L Wf3, a·).

{3EB
Hence, p = ~ Wf3, and consequently p E C. Finally, (iii) is equivalent to

{p,on = i.
COROLLARY. Let a- be half the sum of the elements > 0 of R" (for B). For
all C\' E V, the sum of the coordinates of a with respect to the basis B is (o:, a-).
If o. ER, this s1tm is eq1wl to~ L n(a,,13).
/3ER+(C)
Interchanging the roles of R and R. above, we have (a:, a-) 1 for all
a E B, hence the corollary.
11. COXETER TRANSFORMATION
Let C be a chamber of R, let {a 1 , ... , a1} be the corresponding basis of R, and

let c = Saq ... s 01 • The element c of W is called the Co.'Eeter transformation
of W defined by C and the bijection i ........ ai (Chap. V, § 6, no. 1). Its order
his called the Coxeter number of W (or of R).
PROPOSITION 30. Assmne that R is ·irred·uC'ible. Let m be an integer be-

tween 1 and h - 1 and prime to h. Then exp( 2 i~m) is an eigenvalue of c of
multiplicity 1.
In particular, mis an exponent of W, cf. Chap. V, § 6, no. 2.
We fin;t prove a lemma:
Lemma 4. For all w E W, the characteristic polynomial of w has integer

weffi.C'ients.
We know (no. 6, Th. 3) that {a 1 , ... , 0: 1} is a basis of the subgroup Q(R)
of V generated by R. Since w leaves Q(R) stable, its matrix with respect
to {0: 1 , ... , o-. 1} has integer entries; hence its characteristic polynomial has
integer coefficients.
Let P be the characteristic polynomial of c. The above lemma shows that the
coefficients of P arc integers. By Chap. V, § 6, no. 2, Cor. 2 of Prop. 3, the
primitive hth root of unity z = exp( 2h") is a simple root of P. Every confugale
of z over Q is therefore also a simple root of P. But we know (Algebra., Chap.
V) that the primitive ht.h roots of unity are conjugate over Q. They are thus
all simple roots of P, which proves the proposition.
PROPOSITION 31. Assume that R is irreduC'ible and reduced, and let f3 =

n 1 a 1 + · · ·+n1a1 be the highest root of R (cf. no. 8). Then n 1 + · · · +n1 = h-1.
Let R+ be the set of positive roots relative to C. Then (no. 10, Cor. of
Prop. 29):
1
n1 + · · · + n1 = -2 L n(,13, a)
o:ER+
=l+~ L n(,13,a:)=l+ L (al/3)

2 o:ER+,o:#,6 oER+,o:#/3 (a I a)
By no. 8, Prop. 25 (iv), for all a: E R+ and C\'. -I ,13, n(o:,/3) = 0 or 1, so
n a, ,13) 2 -_ n ( a, /3) , t hat ..
( is 4(o: 1/3>2 _ 2(a IPl Hence.
(/3 I /3l2 - 1i3T7ff.
..
" (a I /3) 2
n1 + · · · + n1 + 1 = 2 + 2 o:ER+,a#fl
L, ( I )(/3 I /3)
a a
L (al/3)2 (/31/3)-1 L a I 2
= 2 o:ER+ (o I a)(/31 (3) = <>ER ( Tinlf (3) .
By Chap. V, § 6, no. 2, Cor. of Th. 1,
L,
'°"' (-1a-II I fJ) 2 = h(/31 f3)
aER Ia
so n1 + · · · + n1 + 1 = h.
PROPOSITION 32. Assume that R is irreducible, and that all roots ha.ve the
same length. Let a E: R. The n·u.mber of elements of R not orthogonal to a is
4h - 6.
Let R' be the set of roots not proportional to and not orthogonal to a.
By Chap. V, § 6, no. 2, Cor. of Th. 1,
(a I a) 2 +(a I - a) 2 + L (a I/3) 2 = h(n / a) 2 ,

.BER'
that is
If (3 E: R', then (a/ /3) =±~(a I 0:) by the list in no. 3. Hence
1 I I
4cardR = h - 2, Ca.rdR = 4h - 8,
and the number of roots not orthogonal to a ii; Card R' +2 = 4h - 6.
PROPOSITION 33. Assume that R is irreducible and red·uced. Put Sn, = Si,
and let I' be the subgroup of W generated by C = S1 ... S/.
(i) Let ()i = 81S1_ 1 ... si+1(oi) (i = l, ... ,l). Then, Bi> 0,c(Bi) < 0.
(ii) If a is a root > 0 such that c( o:) < 0, then o is equal to one of the Bi.
(iii) The farrt"ily (fii)i~i~l is a basis of Q(R).
(iv) Let fti be the orbit of B, under I'. The sets fli are pairwise disjoint,
they are all the orbits of I' on R, and each has h elements.
Observe first that (s 1 , ... , s1) is a reduced decompo.5if'ion of c (Chap. IV,
§ 1, no. 1) with respect to the set S of the Si· Indeed, otherwise there would
exist a subset X = S- {j} of l - 1 elements of S such that c E Wx, which
would contradict Cor. 2 of Prop. 7 of Chap. IV, § 1, no. 8.
Applying Cor. 2 of Prop. 17 of no. 6 to c gives assertions (i) and (ii).
Let Qi be the subgroup of Q(R) generated by the a 1 , j > i. It is immediate
that Qi is stable under the s1 , j > i, and that sj(o:;) =a; mod. Qi for .i > ·i.
Hence:
Bi= s1 ... si·H(a;) = ai mod. Q;.
In other words, there exist integers Cij such that
ei = o:i L: cijO:j.
+ j>i
Part (iii) follows immediately.
h-1
Finally, let o: be a root. The element L ck (a) is invariant under c, and
k=O
hence zero (Chap. V, § 6, no. 2). The ck(a) cannot therefore all have the
same sign, and there exists a k such that ck(a) > 0 and ck+ 1(a) < O. By (ii),
ck(a) is one of the e.;.
Thus every orbit of I' on R is one of the fli. Extend
(x J y) to a hermitian form on V ® C. Each orbit of r in R has at most h
elements, and there are at most l distinct orbits, by the above remarks. Now
(Chap. V, § 6, no. 2, Theorem 2, ii)), the cardinal of R is equal to hl, which
immediately implies (iv).
12. CANONICAL BILINEAR FORM
We have seen (no. 1, Prop. 3) that the symmetric bilinear form
(.1:,y) >--+ Bn.(x,y) = L (o.",x)(o.-,y)

<>ER
on V is non-degenerate and invariant under A(R). Interchanging the roles of R

and W, it follows that the symmetric bilinear form (x*, y*) >--+ Bw (:r*, y*) =
L (o.,x*)(o.,y*) on v· is non-degenerate and invariant under A(R).
a ER
The inverse form of Bw (resp. BR) on V (resp. V*) will be called the canonical
bilinear form on V (resp. V') and denoted by <l>n. (resp. <Pw). It is non-degenerate
and invariant under A(R). Let a be the isomorphism from v to v· defined by BR··
Then, for x E V and y E V:
<PR(x,y) = Bn.-(a(x),a(y)) =
oER
L (o.,a(x))(a,a(y)).
But (a:,a(x)) = BR·(a(a),a(x)) = <PR(o.,x). Hence,
<Pn(x,y) L <l>R(o:,:r)<P(o.,y).
= <>ER (16)
In view of Prop. 7 of no. 2, <f>R is the only non-zero symmetric bilinear form invariant
under W(R) which satisfies the identity (16).
For /3 E R, (16) gives
<l>R(/3, /3) "°""'

= <>ER 1
L- <PR(o., /3) 2 = -;<Pn.(/J, "°""' n(o:, /3) 2 ,
/3) 2 L-
•• oER
hence
(17)
Moreover, by Lemma 2 of no. 1, we have, for x, y E V:

§ 2. AFFINE WEYL GROUP 185
BR(x,y)= L: <PR(ER a, )'x) (ER
It follows that, if R is irreducible, there exists a constant l'(R) > 0 such that
{18)
By the definition of1(R), we have BR(x,y) = 41(R)<f>R(x,y), so
<Pw(x',y') = {41(R))- 1 Bn-(x',y*)
for x*, y* E V*. This proves that 1(R) = 1(W). On the other hand, for /3 E R.,
<Pw(/3-,fn = (4,,{R))- 1 oEL:ll (/3·,a) 2

= /'(R)-1 L; </>R{a,/]) 2
<>ER ih({J, {3) 2
so, by (16),
or finally
{19)
Further, if all the roots of R have the same length >. for <PR, (16) and (18) show
that
(20)
Moreover, if his the Coxeter number of W, the Cor. of Th. 1 of Chap. V, § 6, no.
2 shows that:
h<PR(x,.r) = L: ~) 2 for all x EV.
aER
(x, A
Comparing with (16), we deduce that
{21)
Finally, formula (19) shows that the roots of W have length >. for <Pw.
§ 2. AFFINE WEYL GROUP
In this paragraph (except in no. 5), we denote by Ra reduced root system

in a real vector space V. We denote by W the Wey! group of R; we identify
it with a group of automorphisms of the dual V* of V (§ 1, no. 1), and we
provide V* with a scalar product invariant under W. Let E be the affine
space underlying V*; for 11 E V*, we denote by t( v) the translation of E by
the vector v. Fina.Uy, we denote by P (resp. Q) the group of translations t(v)
whose vector v belongs to the group of weights P(W) (resp. to the group of
radical weights Q(K)) of the inverse root system R- of R.
1. AFFINE WEYL GROUP
For a E R and k E Z, let La,k be the hyperplane of E defined by:
10 ,k = {x E EI (a, x) = k}
and let Sa,k be the orthogonal reflection with respect to La,k· Then
sa,k(x) =x - ((a,x) - k)a· =s 0 ,o(x) + ka·

for all :i; E E. In other words,
SO'.,k = t(ko:") o S°' (1)
where Sa is the orthogonal reflection with respect to the hyperplane

La = La,o, i.e. the reflection associated to the root lY.
Formula (1) shows that Sn,k does not depend on the choice of scalar
product.
DEFINITION 1. The grov.p of affine transformations of E generated by the

reflections s 0 ,k for o: E R and k E Z is ca.lied the affine Weyl group of the
root system R and denoted by W a (R) (or simply by W ").
PROPOSITION 1. The gro1tp W a is the semi-direct product of W by Q.

Since W is generated by the reflections s 0 , it is contained in W a· On the
other. hand, t(on = s 0 , 1 o s 0 if a E R, which shows that Qi;;; W 0.
Sin!'.e W leaves Q(R:) stable (§ 1, no. 9), the group G of affine transfor-
mations generated by W and Q is the semi-direct product of W by Q. Now
G i;;; W 0 from above and s0 ,k E G for all a ER and k E Z by (1). It follows
that Wa = G.
PROPOSITION 2. The group W 0 , w"ith the discrete topology, a.cts properly

on E and permutes the hyperplanes L<>,k (for°' E R and k E Z).
Since Q(R:) is a discrete subgroup of V*, the group Q acts properly on
E. Hence, so does W 0 = W.Q, since Wis finite. Moreover, for o:,(3 ER and
k E Z:
sa(La,k) = 1 1 ,k with / = s13(0:) ER,
t(,8-) (Lo,k) = Let,k+n(a,,i3),
where n(o:,,8) = (,8",o:) is an integer, hence the second assertion.

We can thus apply the results of Chapter V, § 3 to W a acting on E. To
avoid any confusion with the chambers of the Wey! group W in V*, we shall
call the chambers determined by the system of hyperplanes La,k (for E R °'
and k E Z) in E alcoves. The group W a thus acts simply transitively on the
set of alcoves and the closv:re of a.n alcove is a fundamental domain for W a
acting on E (Chap. V, § 3, no. 2, Th. 1 and no. 3, Th. 2). It is clear that
the Wey! group W is identified with the canonical image U(Wa) of W 0 in
the orthogonal group of V* (cf. Chap. V, § 3, no. 6). It follows that W a

is essential (Chap. V, § 3, no. 7) and that W a is irreducible if and only if
the root system R is (§ 1, no. 2, Cor. of Prop. 5). If R is irreducible, every
alcove is an open simplex (Chap. V, § 3, no. 9, Prop. 8). In the general case.
the canonical product decomposition of the affine space E (Chap. V, § 3,
no. 8) corresponds to the decomposition of R into irreducible components. In
particular, the alcoves are products of open simplexes.
Note also that the Cor. of Th. 1 of Chap. V, § 3, no. 2 shows that the
s0<,k are the only reflections in W a.
2. WEIGHTS AND SPECIAL WEIGHTS
PROPOSITION 3. The special points (Chap. V, § 3, no. 10, Def. 1) of Wa

are the weights of R~.
Let x 0 E E and let o· E R. The hyperplane L parallel to Ker a and
passing through x 0 has equation (a:,x) = (a:,xo). To be equal to some L.13,k,
it is necessary on the one hand that o: and {3 are proportional, and so, since R
is reduced, that /3 = ±o:, and on the other hand that (a:, .-r 0 ) is an integer. It
follows immediately that .t:o is a special point of W" if and only if (a:, xo) E Z
for all a: ER, in other words, if and only if Xo E P(R') (§ 1, no. 9).
COROLLARY. (i) If w E P(W), there exists an alcove C such that w is an

extremal point of C.
(ii) If C is an alcove, C n Q(W) reduces to a single point and this i8 an
extremal point of C.
This follows from Prop. 3, in view of the Cor. of Prop. 11 of Chap. V,
§ 3, no. 10 and Prop. 12 of Chap. V, § 3, no. 10.
PROPOSITION 4. Let C' be a chamber of W.

(i) There exists a unique alcove C contained in C' sitch that 0 E C.
(ii) The union of the w(C) for w E W is a neighbourhood of 0 in E.
(iii) Every wall of C' is a wall of C.
This follows from Prop. 11 of Chap. V, § 3, no. 10.
Assume now that R is irreducible. Let (a;)iEI be a basis of R (§ 1, no. 5,
Def. 2), and let (Wi)iEI be the dual basis. The Wi are the fundamental weights
of R" for the chamber D' of W corresponding to the basis (o:;). Let
ii:= l:n;ai
iEI
be the highest root of R (§ 1, no. 8), and let J be the set of 'i E I such that
ni = 1.
PROPOSITION 5. Let C be the alcove contained in C' and containing 0 in
its clos'ure (Prop. 4).
(i) C is the set of x EE such that (ai, x) > 0 for all i E I and (a, x) < 1.
(ii) The set C n P(K) consists of 0 and the wi for i E J.
Let D be the set of x E E such that (&, x) < 1 and set C 1 = C' n D.
Since 0 E C, we have C <::;: D and hence C <::;: C 1 . We are going to show
that, for all a: E R and all k E Z, the sets C and C 1 are 011 the same side
of the hyperplane LCl,k· This will prove that C1 <::;: C and so will establish
assertion (i). If k = 0, the whole of the chamber C' is on one side of L0 , 0 ,
which establishes our assertion in this case. If k i= 0, we may, by replacing a
by -a, assume that k > 0. Then (a, x) < k on C, since 0-E C. On the other
hand, & - a is positive on C' (§ 1, no. 8, Prop. 25). Thus, for y E C 1, we
have (a, y) ~ (&, y) < 1 ~ k. Consequently, C and C 1 are on the same side
of La,k·
Now let w E P(R-). Then w = l;:piwi, with Pi E Z (§ 1, no. 10), and
w E C' ' positive. If w E C', then
if and only if the integers p; are w E C if and
only if (&,w) = °2:n;pi is~ 1, hence (ii).
'
COROLLARY. The alcove C is an open simplex with vertices 0 and the
wi/n;,
i EI.
This follows from (i).
3. NORMALISER OF W a
In this no., we assume that the chosen scalar product on V is invariant not
only under W but ·u.nder the whole of the group A(R). We identify A(R) and
A(R-).
Let G be the normaliser of W a. in the group of displacements of the affine
euclidean space E. If g is a displacement of E, and s is the orthogonal reflection
with respect to a hyperplane L, the displacement gsg- 1 is the orthogonal
reflection with respect to the hyperplane g(L). It follows that G is the set of
displacements of E that permute the hyperplanes La,k (for o: E Rand k E Z).
Now, the group of automorphisms of E is the semi-direct product of the
orthogonal group U of V* and the group T of translations. If u E U and
v EV*, the hyperplane La,k is transformed by g = uot(v) into the hyperplane
with equation
(u- 1 (a),x) = k + (a:,v).
Consequently, g E G if and only if, on the one hand tu permutes the roots, in
other words belongs to A(R), and on the other hand (a, v) E Z for all a E R,
that is v E P(K). In other words, the group G is the semi-direct product of
A(R) by P. Since Q <::;: P and W <::;: A(R), the quotient group G/Wa is the
semi-direct product of A(R)/W by P(K)/Q(R-); it is immediately checked
that the corresponding action of A(R)/W on P(W)/Q(R-) is the canonical
action (§1, no. 9).
We denote by W~ the subgroup of G formed by the semi-direct product of

W by P. This is a normal subgroup of G, and G/W~ is canonically isomorphic
to A(R)/W; moreover, the canonical map from P(K) to W~/Wa gives by
passing to the quotient an isomorphism from P(K)/Q(K) to W~/Wa.
Now let C be an alcove of E, and let Ge be the subgroup consisting of the
element8 g E G such that g(C) = C. Since W a is simply-transitive on the al-
coves, the group G is the semi-direct product of Ge by Wa. The corresponding
isomorphism from G /W a to Ge gives rise in particular to a canonical iso-
morphism from P(R-)/Q(R-) to the group I'c =Gen W~.
Assume that R is irreducible, and retain the notation of Prop. 5 of no. 2.
Put R.0 = R, and let Ri (i E I) be the root system generated by the a1 , for
j f: i. For i = 0 (resp. i EI), let w; be the unique element of W(Ri) (identified
with a subgroup of W) which transforms the positive roots of Ri relative to
the basis (a:J)Jr!i into negative roots(§ 1, no. 6, Cor. 3 of Prop. 17).
PROPOSITION 6. For all i E J, the element 'Yi = t(w;)wiwo belongs to I'c

and the map i ,........ 'Yi is a bijection from J to r c - {1}.
We remark first of all that the root Wi (a) is of the form
and hence is positive.

We show that, if i E J, then 'Yi E I'c. Indeed, let a E C and b = 'Yi (a). For
1 ~ j ~ l and j =I- i,
(b, a 1) = (w; + wiwo(a), 0:1) (2)

= (wo(a), w;(ni)) > 0
since w 0 (a) E -C' and w;(a:j) is negative. On the other hand,
(b, O:i) = 1 + (wo(a), w;(a:i)) ? 1 + (wo(a), a) >0 (3)

since w 0 (a) E -C', a-wi(o:i) takes negative values on -C', and (w 0 (a),a) >
-1. Finally,
(b, a)= n; + (wo(a), Wi(£i)) = 1 + (wo(a), Wi(o)) <1 (4)

since w 0(a) E -C' and Wi(ii) is a positive root. The relations (2), (3) and
(4) then imply that b E C, and hence that 'Yi E Tc. It is clear that the map
i ......, 'Yi is injective, since 'Yi (O) = Wi. Finally, let 'Y E I'c with 'Y I- 1, and put
'Y = tw with t E P and w E W. Then t f l since I'c nW = {l}. On the other
hand, t(O) = 'Y(O) E C n P(R") and Prop. 5 implies that there exists i E J
such that t(O) = wi· Then 'Yi 1'y(O) = 0, hence 'Y ='Yi since I'c n W = {l}.
This completes the proof.
COROLLARY. The (w;);o form a system of representatives in P(K) of the

non-zero elements of P(R")/Q(R-).
190 ROOT SYSTEMS Ch. Vl
Indeed, if we identify I'c with P(R")/Q(R"), the element 'Yi is identified

with the class of Wi mod. Q(R").
Remarks. 1) The map ry ~ ry(O) is a bijection from I'c to C n P(R").

2) The group G is also the normaliser of W a in the group of automor-
phisms of E provided only with its affine structure (cf. Exerc. 3).
4. APPLICATION: ORDER OF THE WEYL GROUP
Lemma 1. Let X be a locally compact space countable at infinity, G a group

acting continuo·usly a.nd properly on X, µ a measure on X invariant under
G, G' a subgroup of G, U and U' two open s·ubsets of X with finite non-zero
measure. Assume that the sU for s E G {resp. the s'U' for s' E G') are
pairwise disjoint and that the'ir union is of neglig·ible complement. Then G'
is of finite index in G and (G : G') = µ(U')/ µ(U).
Let (s>.)>.EA be a family of representatives of the right cosets of G' in G.
Let U 1 be the union of the s >. U. Then the s'U 1 , for s' E G', are pairwise dis-
joint and have union M = U sU. Let M' = U s'U'. The union of U' (resp.
sEG s'EG'
U 1 ) and a suitable subset of X- M' (resp. X- M) is a fundamental domain,
evidently µ-measurable, for G'. By Integration, Chap. VII, § 2, no. 10, Cor.
of Th. 4, µ(U') = µ.(Ui). This proves that Card 11 = (G : G') is finite, and
that /L(U') = (Card !l)µ(U).
PROPOSITION 7. Assume that R is in·educible. Let B = {0'. 1 , ... , a 1} l1e a

basis of R, f the connection index of R (§ 1, no. 9) and a = n1 n 1 + · · · + n10:1
the highest root of R (for the order defined by B ). Then the order of W is
equal to
(l!)n1n2 ... ntf.
Let (w 1 , ... , wt) be the basis of P(R") dual to B. By the Cor. of Prop. 5, the
open simplex C with vertices 0, n;- 1w 1 , ... , n1 1w1 is an alcove of E. Choose
a Haar measure µon the additive group V*. Let A be the set of elements of
V* of the form l, 1w1 + · · · + l,1w1, with 0 <Si < 1 for i = 1, ... , l. By Cor. 2
of Prop. 15 of Integration, Chap. VII, § 1, no. 10,
(5)
On the other hand, let A' be the set of elements of V* of the form
6a1- + · · · + l,101·,
with 0 <Si < 1 for i = 1, ... , l. Since (a 1·, •.. , al) is a basis of the Z-module
Q(R"), we can apply Lemma 1 with X = V*, G = Wa, G' = Q, U = C and
U' =A'. This gives
µ.(A')/µ(C) = (Wa: Q) = CardW. (6)

Finally, Lemma 1 can be applied again, taking X = V*, G = P, G' = Q,

U =A and U' =A'. This gives
µ(A')/µ(A) = (P: Q) = (P(W): Q(W)) = J. (7)
The proposition now follows by comparing formulas (5), (6) and (7).
5. ROOT SYSTEMS AND GROUPS GENERATED BY

REFLECTIONS
PROPOSITION 8. Let F be a real Hilbert space of fi,nite dimension l. Let

.lj be the set of affine hyperplanes of F and G the group generated by the
orthogonal reflections SH with respect to the hyperplanes H E .lj. Assume that
the conditions of Chap. V, § 3 are satisfied (i.e. that g(H) E f) for all HE .lj
and g E G, and that G acts properly on F ). Assume also that 0 is a. special
point for G and that the group T of translations belonging to G is of rank l.
Then there e.zists a unique reduced root system R in V = F* such that the
canonical isomorphism from F to V* transform.s G to the affine Weyl group
Wa ofR.
We remark first of all that the assumption on T implies that G is essential:
otherwise, the affine space F would decompose into a product F 0 x F 1 , with
dim F 1 < l, the group G being identified with a group of displacements acting
properly on F 1 (Chap. V, § 3, no. 8, Prop. 6), and T would not be of rank 1.
Let fJo be the set of H E Jj such that 0 E H. For H E Jjo, let .!jH be the
set of elements of .!j parallel to H. Since 0 is a special point, .f) is the union
of the .fJH for H E .lj 0. Since T is of rank l, there exists a v E F such that
the translation by the vector v belongs to T and v <t. H. The hyperplanes
H + kv fork E Z are pairwise distinct and belong to .f) 8 . Now let a be a unit
vector of F orthogonal to H: then H + (vja)a E JJ1-1 and since .f) is locally finite
(Chap. V, § 1, no. 1, Lemma 1), there is a smallest real number >. > 0 such
that H + >.a E .lj. We are going to show that .fJH is the set of hyperplanes
H + k>..a for k E Z. Indeed,
H' = H +>.a E .fJH
and the element SH' o SH of G is the translation by the vector 2>..a (Chap. V,
§ 2, no. 4, Prop. 5). Consequently, H + 2n>.o. = (sH'stt)"(H) and H + (2n +
l)>..a = (swsH)n(H') belongs to .fJH- On the other hand, if L E fJH, there
exists ~ E R such that L = H + ~>..a and there exists an integer n such that
either 2n <~ ~ 2n + 1, or 2n - 1 < ~ ~ 2n.
In the first case, (s 8 sH' )n(L) = H + (~ - 2n)>.a with

0 < (~ - 2n) >.. ~ >.
and the definition of >.. implies that ~ = 2n + 1; in the second case,

and the definition of >. implies that ~ = 2n.

It follows that if ll'H is the linear form on F such that
H' = {x E Fj (aH,x) = l},

the set 5JH is the set of hyperplanes L 0 H ,k = { x E F I (aH, x) = k} for k E Z,
and CXH and - ll'H are the only linear forms with this property.
Consequently, the proposition will be proved if we show that the set R of
elements of V of the form ±cxH is a reduced root system in V.
a) We prove condition (RS1): it is clear that R is finite (since 5) 0 is fi-
nite) and does not contain 0. Moreover, R generates V. Indeed, if x E F is
orthogonal to R, then x E H for all H E 5) 0 and the translation by the vector
x commutes with every element of G. Since G is essential, this implies that
:r = 0.
b) We prove (RSu). For v EV and r ER, put Lv,r = {x E FI (v,x) = r}
as above; if a E R, put Ha = 1 0 ,0, and let s 0 be the transpose of SH". There
exists a unique element a- E F orthogonal to Ha: and such that (a-,a) = 2.
Then SH" == S 0 ·,a and Sa = So:,a·· For /3 E R,
and there exist 'Y E Rand n E N* such that Ls"(/3),l = L,,,n. Then
SH" (L,,,i) = Lp,1/n
and so l/n E Z. Thus n = 1 and s 0 (/3) ='YE R. This proves (RS 11 ).

c) We prove (RSm). Let a ER and set H~ == L0 , 1 . Then
H~ =Ha+ (1/2)o-;
since the translation t(a-) by the vector a- is the product sH;,SH" (Chap. V,
§ 2, no. 4, Prop. 5), it belongs to T and a:-= t(a:-)(0) is a special point for G.
Consequently, for all /3 E R, there exists a hyperplane Lp,k passing through
ex-, with k an integer, which shows that (/3, o:") E Z, and proves (RSm).
d) Finally, it is clear that R is reduced, for if H, H' E 5) 0 , H =J. H', the
linear forms ll'H and ow are not proportional.
Remark. 1) The assumption that T is of rank l is satisfied in particular when

G is irredv.cible and infinite. Indeed, the vector space generated by vectors
corresponding to the translations in T is invariant under the canonical image
of G in the linear group of F. It is different from {O} if G is infinite and is
thus equal to the whole of F if G is infinite and irreducible.
A finite group generated by reflections is not always the Wey! group of a

root system. More precisely:
PROPOSITION 9. Let V be a real vector space of finite dimension l, and

let G be a finite subgroup of G L(V), generated by refiections and essential.
Give V a scalar product invariant under G. The following conditions are
equivalent:
(i) There exists a discrete subgroup of V of rank l that is stable under G.
(ii) There exists a Q-structure on V (Algebra, Chap. II,§ 8, no. 1, Def. 1)
invariant under G.
(iii) There exists a root system in V whose Weyl group is G.
(iv) There exists a discrete group G' of displacements ofV, acting properly
on V, and generated by refiections, such that G' is the semi-direct product of
G and a group of translations of rank l.
(ii) ==? (i): let V' c:; V be a Q-structure on V invariant under G. Let A
be a finite subset of V' generating the Q-vector space V'. Replacing A by
U s(A), we can assume that A is stable under G. Let B be the subgroup of
sEG
V generated by A. Then B is stable under G, of finite type and torsion-free,
so has a basis over Z which is both a basis of V' over Q and a basis of V over
R.
(iii) ==? (ii): this follows from Prop. 1 of§ 1, no. 1, for example.
(iv) ==? (iii): let G' be a group satisfying condition (iv). The group of
translations of G' is of rank l, and 0 is a special point for G' by Prop. 9 of
Chap. V, § 3, no. 10. Prop. 6 shows that there exists a reduced root system
Ro in V* such that G' is identified with Wa(R 0 ): the group G is then the
Wey! group of the inverse root system of R 0 .
(i) ==? (iv): assume that G leaves stable a discrete subgroup M of V, of
rank l. For any reflections E G, s(x) - x EM for all x EM, so the line Ds
orthogonal to H., meets l\/I; let a 5 , -a 5 be the generators of the cyclic group
D 5 n M; the set A of the O'.s and -as is st.able under G, hence generates a
subgroup M' of M stable under G; the discrete group M' is of rank l because
G is essential. Let G' be the group of affine transformations of V that is the
semi-direct product of G and the grnup of translations whose vectors belong
to M'. Let G~ be the subgroup of G' generated by the reflections of G'. We
shall show that G~ = G', which will complete the proof. First, G~ ~ G since
G is generated by reflections. On the other hand, for any reflections of G, let
ts be the translation with vector O'.s. The transformation so ts is a reflection,
and so ts E G'; thus t 5 is a product of two reflections of G'; this being true
for every reflection s of G, the translations whose vector belongs to M' are
all in G~.
DEFINITION 2. A group G satisfying the equivalent conditions of Prop. 9 is

called a crystallographic group.
Remark. 2) Let G be a finite group generated by reflections and essential.

Then G is crystallographic if and only if every element of its Coxeter matrix is
one of the ·integers l, 2, 3, 4, 6. Indeed, this condition is necessary by Remark 3)

of § 1, no. 5. The fact that it is sufficient will follow from the classification
of finite Coxeter groups given in § 4 (for a direct proof, see Chap. V, § 4,
Exerc. 6).
§ 3. EXPONENTIAL INVARIANTS
In this section, the letter A denotes a commutative ring, with a unit element,
and not reduced to 0.
1. GROUP ALGEBRA OF A FREE ABELIAN GROUP
Let P be a free Z-module of finite rank 1. We denote by A[P] the group

algebra of the additive group of P over A (Algebra, Chap. III,§ 2, no. 6). For
any p E: P, denote by eP the corresponding element of A[P]. Then (eP)pEP is
a basis of the A-module A[P], and, for any p,p' E P, we have
Lemma 1. Ass'Ume that A is factorial (Commutative Algebra, Chap. VII, § 3,

no. 1, Def. 1).
(i) The ring A[P] is .factorial.
(ii) If u, v are non-proportional elements of P, the elements 1 - eu, 1 - ev
of A[P] are relatively prime.
Let (p 1,p2 1 • • • ,p1) be a basis of P, and X 1,X2, ... ,X1 be indetermi-
nates. The A-linear map from A[X 1 , ... ,X1, X] 1 , ... ,X! 1 ] t.o A[P] that takes
X~· 1 X~ 2 ••• X~ 1 (where n 1 , n2, ... , n1 E Z) to en 1 r 1 +···+niri is an isomorphism
of rings. Now A[X 1 , ... , Xt] is a factorial ring ( Commu.tative Algebra, Chap.
VII,§ 3, no. 5), and A[X 1 ,. .. 1 X1,X] 1 ,. . .,X/ 1 ] is the ring of fractions of
A[X 1 , ... , X1J, hence is also factorial.
Let P' (resp. P") be the set of elements of P of which some multiple
belongs to Zu + Zv (resp. Zu). Then the groups P /P' and P' /P" are torsion-
free, so there exists a complement of P" in P' and a complement of P' in P.
Consequently, there exists a basis (z 1 , z2 , ... , z1) of the Z-module P and ra-
tional integers j, m, n such that u = jz 1 , v = mz 1 + nz 2 ,j > 0, n > 0. Putting
xi = eZ; for 1 ~ i ~ l, we then have l - eu = 1 - Xi' 1 - e0 = 1 - Xf' x~.
Let K be an algebraic closure of the field of fractions of A, so that A[P] can
be identified with a subring of the ring B = K [X 1 , ... , X1, X] 1 , ... , X! 1 J. For
any jth root of unity z, 1 - zX 1 is extremal in
moreover, the ideal generated by 1 - zX 1 contains no monomial in the Xi.

We conclude that the ideal (1 - zXi)B of B is a prime ideal of height 1
§ 3. EXPONENTIAL INVARIANTS 195
(Commutative Algebra, Chap. VII,§ 1, no. 6), hence that 1-zX 1 is extremal
in B. The extremal factors of 1 - X{ in B are thus of the form 1 - zX 1 . Now
none of these factors divide 1 - Xf' X2 in B (for the homomorphism f from
B to B such that f(Xi) = z- 1 ,f(Xi) =Xi for i ~ 2, satisfies
f(I - zX1) =0 and f(l - X~nX2) = l - z-mx2 =I 0).

Thus, 1 - X{ and I - Xi' X2 are relatively prime in B. Consequently, any
common divisor of 1 - X{ and I - X;"X2· in A[P] is invertible in B and so,
up to multiplication by an element of the form X~ 1 X~ 2 ••• X~', is equal to
an element a of A; in other words,a divides 1 in A, hence is invertible in A.
Thus, finally, 1 - X{ and 1 - X}X;'· are relatively prime in A[P].
2. CASE OF THE GROUP OF WEIGHTS; MAXIMAL TERMS
We retain the notations of the preceding number and let. R be a reduced root
system in a real vector space V. In the remainder of this section, we take for
P the group of weights of R (§ 1, no. 9). The group W = W(R) acts on P,
hence also on the algebra A[P]; we have w(eP) = ew(p) for w E Wand p E P.
Let C be a chamber of R (§ 1, no. 5) and let B = (a:;)i~i~l be the
corresponding basis of R. We provide V (and hence also P) with the order
structure defined by C. If p, p' E P, p ~ p' if and only if p - p' is a linear
combination of the O'i with positive coefficients.
DEFINITION 1. Let x = L .'EpeP be an clement of A[P]. The set S ofp E P

pEP
such that Xp =I 0 ·is called the support of x and the set X of maximal elements
of S is called the maxim.al support of x. A term XpeP with p E X is then called
a maximal term of.?:.
Lemma 2. Let x E A[P] and let (xpeP)pExbe the family of maximal te-rms of
x. Let q E P and let y E A[P] be such that eq is the un'ique maximal term of
y. Then, the family of ma.rimal terms of xy is (xpeP+9)pEX.
Put x = LxpeP,y = Lyre"and xy = Lzte 1 • Then r ~ q for all r E P
p r· t
such that Yr L XpYr·
f:. 0 and Zt = p+r=t
If t = p + q = p' + r with p E X and Xr'Yr =/= 0, then r ~ q, hence p' ~ p
and consequently p' = p. Thus Zp+q = XpYq = Xp I 0. This shows that X + q
is contained in the support of the product ."Cy.
On the other hand, if t = p' + r with .Tp'Yr f:. 0, there exists p E X such
that p' ~ p and we have t ~ p + q. The maximal support of xy is therefore
contained in X + q. Since no two elements of X + q are comparable, it follows
that X + q is exactly the maximal support of ."Cy and we have seen above that
Zp+q = -'Ep for p E X, which completes the proof of the lemma.
Remark. Since x # 0 means that the maximal support of x is non-empty,

Lemma 2 shows that x # 0 implies xy # 0 whenever y admits a unique
maximal term of the form e<J.
3. ANTI-INVARIANT ELEMENTS
We retain the notations of the preceding number. Denote by c:(w) the deter-
minant of the element w E W. Thus
c:(w) = (-l)l(wl,
the length l (w) being taken relative to the family of reflections s 0 ,.
DEFINITION 2. An clement x E A[P] is said to be anti-invariant ·under W

if
w(x) = (-l)l(w).x
for all w E W.
The anti-invariant elements of A[P] form an A-submodule of A[P]. For

any x E A[P], put
J(x) = L c:(w).w(x). (1)
tvEW
For x E A[P] and w E W, we have
w(J(x)) = L _c:(v).wv(x) =
vEW
c:(w) L
vEW
c:(v).v(.1:) = c:(w).J(x)
and J(.T) is anti-invariant. On the other hand, let q = Card(W). For any
anti-invariant element x of A[P], we have J(x) = q.x. It follows that, if q is
invertible in A, the map q- 1 J is a projection from A[P] onto the submodule
of anti-invariant elements.
Let w1 , .•• , Wt be the fundamental weights corresponding tci the cham-
ber C. The elements of P n C (resp. P n C) are the weights of the form
n 1 w1 + · · · + n1w1 with ni ~ 0 (resp. ni > 0) for 1 ~ i ~ l (§ 1, no. 10). On
the other hand,
p = W1 t · · · + Wt
is half the sum of the positive roots ( loc. cit.) so the elements of P n C are the
weights of the form p + p with p E P n C. Finally, if p E P n C, then w(p) < p
for all w # 1 (§ 1, no. 6, Cor. to Prop. 18) and eP is thus the unique maximal
term of J(eP).
PROPOSITION 1. If 2 is not a zero divisor in A, the elements J(eP) for

p E P n C form a basis of the module of anti-invariant elements of A[P].
The weights w(p) for w E Wand p E PnC are pairwise distinct. It follows
that the J(eP) for p E P n C are linearly independent.
3. EXPONENTIAL INVARIANTS 197
On the other hand, let x = Lp .rpeP be an anti-invariant clement of A[P].

If p 0 belongs to a wall, it is invariant under a reflection s E W and
It follows that 2xp 0 = 0, so Xp 0 = 0. Since every element that does not

belong to any wall can be written uniquely in the form w(p) with w E Wand
p E P n C, we thus have
X = L L Xw( ,ew(p)_ (2)
pEPnC wEW Pi
Since w(x) = Lxpew(p)

p
= E(w) LxpeP,
p
xw(p) = c;(w)xp and we deduce
from (2) that
which completes the proof.

Consider now the element d of the algebra A [ ~ P] defined by
d = fl (eo/2 - e-a/2) (3)

oE:R,a>O
= eP. fl (1 - e-")
aER,a>O
= e-P_ fl (e" - 1).

nER,a>O
Since p E P, d E A[P].
PROPOSITION 2. (i) The element d de.fined by (3) i,s an anti-invariant el-

ement of A[P]; its ·unique maximal term. (no. 2, Def. 1) is eP and d = J(eP).
(ii) For any p E P, the element J(eP) is divisible uniquely by d and the
quotient J(eP)/d is an elernent of A[P] invariant under W.
(iii) fl 2 is not a zero divisor in A, multiplication by d is a b(jection from
the set of elements of A[P] invariant under W to the set of anti-invariant
elements of A[P].
We know that, for 1 ( i ( l, the reflection si = Sry, leaves stable the set
of positive roots other than ai and that s1 (a;) =-a; (§ 1, no. 6, Cor. 1 of
Prop. 17). Hence,
s;(d) = (e-a,/2 - ea,/2)_ fl

aER,a>O,o,.Ca;
v·/2 - e-a/2)
= -d = c:(s;).d.
Since the S; generate W, this proves the first assertion in ( i). The second
assertion in (i) follows immediately from (3) and Lemma 2. noting that 1 is
the unique maximal term of 1 - c-a for a ER, a> 0.
Assume now that A = Z. By Prop. 1,
d= TI cpJ(eP) with Cp E Z. (4)

pEPnC
On the other hand, it is clear that
d= eP + L: c' eq.
q p and the coefficient of eP in d is zero by

(5). Thus, Cp = 0. Moreover, comparison of the coefficients of eP in (4) and
(5) shows that cp = 1 and hence that d = J(eP).
We continue to assume that A = Z. Let p E P, o E Rand M be a system of
representatives of the right cosets of W with respect to the subgroup { 1, So:}.
Then,
J(eP) = L c(w)ew(p) + L c(sC\:w)e 8 """(P).
wEM wEM
Now so:w(p) = w(p) - (a·, w(p))a = w(p) + nwa, with nw E Z. Thus,
J(eP) = 'L: c(w)ew(p)(l - e71 '"°').

wEM
If nw ;?: 0, it is clear that 1 - enw<> is divisible by 1 - e°' and this is also true
when nw < 0 since 1 - enwo: = -en"'°'(l - e-n'""'). Hence, .J(eP) is divisible
by 1 - e°' in Z[P].
By Lemma 1, Z[P] is factorial and the elements 1 - e0 for n E R and
a > 0 are mutually prime. It follows that J(eP) is divisible in Z[P] by the
product TI (1 - e°'), and hence also by d = e-P TI (1 - e°').
<>>0 C1:>0
Returning to the general case, by exten::;ion of scalars from Z to A, we
deduce from the above that d = J(eP) and that every element J(eP) is divisible
by d. Since eP is the unique maximal term of d, ·the Remark of no. 2 shows
that there exists a unique element y E A[P] such that J(eP) = dy and it
follows immediately that y is invariant under W, and hence that d and J(eP)
are anti-invariant. This proves (i) and (ii).
Finally, if 2 is not a zero divisor in A, the Remark of no. 2 and Prop.
imply (iii).
Remarks. l) If 2 is not a zero divisor in A, it is easy to check that dis the

unique anti-invariant element of A[P] with eP as its maximal term.
2) Lemma 2 of no. 2 shows that the unique maximal term of the quotient
J(eP)/d (for p E P n C) is eP-P.
§ 3. EXPONENTIAL INVARIANTS 199
4. INVARIANT ELEMENTS
Let A[P]w be the subalgebra of A[P] consisting of the elements invariant

under W. For p E P, denote by W.p the orbit of p under W, and let S(eP) =
L eq be the sum of the W-transforms of eP; this is a W-invariant element.
qEW.p
If p E P n C, w(p) ~ p for all w E W (§ 1, no. 6, Prop. 18) and eP is the
unique maximal term of S(eP).
Let .T = L XpeP E A[P]w; then .Tw(p) = .Tp for all p E P and all w E W.
p
On the other hand, every orbit of W in P meets P n C in exactly one point
(§ 1, no. 5, Th. 2). Hence,
x = ~ XpS(eP). (6)
PnC
We deduce:
Lemmo. 3. The S(eP) for p E P n C form a basis of the A-module A[P]w.
More generally:
PROPOSITION 3. For any p E P n C, let .xp be an element of A[P]w with

unique maximal term eP. Then the farn.ily (xp)pEPnc is a basis of the A-
module A[P]w.
We first prove a lemma.
Lemma 4. Let I be an ordered set satisfying the following condition:

(MIN) Every non-empty subset of I contains a minimal element.
Let E be an A-module, (ei)iEI a bas·is of E and (x;)iEl a family of elements
of E such that
for all i EI (with ai.i EA, the support of the family (a;i) being finite for all
i). Then, (x;);EI is a basis of E.
For any subset J of I, let EJ be the submodule of E with basis (e;.);EJ.
Let 6 be the set of subsets J of I with the following properties:
(a) Ifi' ~ i and ·i E J, then i' E J;
(b) (,r;);EJ is a basis of EJ.
It is immediate that 6, ordered by inclusion, is inductive and non-empty.
It therefore has a maximal element J. If J ::/=I, let i 0 be a minimal element of
I - J and put J' = JU { io}. Every element i E I such that i < io then belongs
to J: it follows that J' satisfies (a.). On the other hand, J' also satisfies (b):
indeed,
200 HOOT SYSTEl\IS Ch. VI
from which (b) follows. Ht'11cc .I' E 6. a nrnt.radiction. Thus, ,J = I and the
l<'mma is provt•d.
Wf' now prove Prop. :~. \Ve appl.v Lemma 4 with I = P n C. Let <J E I. and
Id; lq I)(' the set of p E l such that p ~ q. If p EI,,. tlw relations
q - p ~ 0. JI E C, <J E C
imply that
(q - PIP) ~ 0 and (q - plq) ~ 0,
a11<I he11ce that

(pip) ~ (plq) ~ (qlq).
The set I,1 is thus bo1tnd('(l. Since I is discrc•te. it. follows I.hat. J,1 is fi:nite. and it
is cbtr that I sat isfit•s the co11dit io11 (\II!\!). 011 the other hand. for nil p E I,
.
J ·p = ,.,,,, + ~
L
('1"1 ('. q
(/<!'
SU by (6).
The proposition now follows from Lennnas ~l and .t.
THEOHEi\I 1. Let w1 , ••.• w1 lw the .fundamental wrights rorresporuling to

the dw.mbrr C, and, .fin· I~ i ~ l, let .r; be an dnnent of A[P]w withe:;;, as
·its 11.ni<ruc ma:i:imal term. f,ct
<p: A[X1, .... Xi]~ A[PJ""

be the homomorphism from. lhe polynomial algebra A[X 1 , ••• , X,] to A[P]W
that lakes X; to :1:;. ThP-n. the m.a.p ;p is an isomorphism.
Lemma 2 implies that the image under;; of th<• monomial X't'' ... X~ 1 is
an clement with u11iq11t> maximal term <:"':;:;,+···+n,:;;,, Since 1•vt·r.v clement of
P n C can bt> writtP11 1111iqud~· iu tlw form 11. 1 ~ 1 +- · · · + n 1 ~1 , Prop. 3 shows
that t.he images lllldcr :.p of the llHlllOlllials 11 • . • x; x;"
arc a basis of A[P]w,
lwncc the tlworcm.
E:r:a.rnples. l) \Ve can take :r; = S(~::;:;, ).

2) B.v Rem.ark 2 of no. ;1. we ca11 take x; = .l(cP+:;:;,)/d (with thE' notation
in no. 3).
s 4. CLASSIFICATION OF ROOT SYSTE\IS 201
§ 4. CLASSIFICATION OF ROOT SYSTEMS

I. FINITE COXETER GROUPS
In thb section, we are goiug t.o detC'rrninc, 11p to isomorphism, all root sys-
tems. and cunsequent.ly all crystallographic groups (§ 2. no. !'i ). 1\forE' gener-
ally. we shall start by dctPnniniug all finit.c groups gt'Il<'rated by rC'flect ions in
a finite-dimensional rt'al vC'ctor spacc: this is equivalent (Chap. V, q 4, no. 8)
to dctPrmining nil finite Coxcter groups, or (Chap. V. 3 4. 110. 8, Th. 2)
to determiui11g all Coxeter matrices of finite order such that the associated
bilinear form is positive aud 11011-dCf.!/'nernte.
Let. 1\I = (m·ij )i ..iEI he a Coxct<>r matrix of finite order /. Put
<J;j = - ros(n'/mi.il·
RC'call that q;; = 1 and that % = <Jii is zero
or ::::; -1/2 for i =I j. Put
E = R a11d let (c;)iEI he the canonical basis of E. Denote b.v (:rlu) the
1
bilinear form on E associated to M (Chap. V, ~ 4. no. 1) and q the qua<lrntic

form .i: 1--. (:rl:r) on E. For :r = L~;:r;,
iEI
II .r II 2 = q(:r) '"' l/i;~i~j.

= . L.J
1,,1EI
Denote by (X . .f) the Co:r:ctcr .fJraJlh of l\I (Chap. TV. § I. no. 9). If a is an
edge of X. f (a) is called the order of a.
In the remainder of this nmnbcr. the Cox<>ter group W('.\1) <lefincd by
'.\I (Chap. V, § 4, 110. :n is assumed to be finite, so that <J is positive an<l
11011-dcgenernte and X is a fore8t (Chap. V, 3 4, 110. 8, Prop. 8). We also
assume that X is connf'.clcd (in other words that the Coxcter group W(M) is
frn:dm:ibfr). so l hat X is a tn:e.
From the condition that q is positive and non-degenerate, we shall obtain
comlit:ions on the mi.i that will enable us to list. all the possibilities for the
corresponding Coxctcr graphs: it will only rcmai11 to show that these possi-
bilit ics arc actually realised, in other words that the <.:0rrespo11diug groups
W('.\I) arc finitl'.
Lemma 1. For all i . .I: <J71· < l.

jfi
Let .J he the set of j E I snrh that <J;.i =/= 0. in other words snch that
{'i, .i} is an edge of X. If j . .i' E .J and j =I j', {.i, j'} is 11ot an edge (otherwise
i.j.j' would form a circuit). so (e;lci') = 0. Let F = L Rr·;. Then (c1 );EJ
. . .iEJ .
is an ortl1011ormal basis of F. The distanec d from c; t.o F is gi\'Pn by rl2 =
1- L(ciieJ) 2 = 1 - L <JT; = 1 - L qf1 . hence the kmma.
JE.I .1E.J . I~' .
Lemma 2. Any vcrfr:r: of X bclong8 to at mo8f 3 rrlgcs.

202 1100T SYSTEl\IS Ch. VI
l11de<•d. if i is linked to It other \'ertices. tlw relations <JlJ ? ~ for tlH'SC

other vcrtin·s implies that ~ < l h.v Lc•1mmt l. so h ,::; :I.
ll'mma :J. lf i lwlong.~ lo :J ed.qes. these <'<(<Jf'8 mr of order .'/.

If not. we wo11ld have. in view of the r<'lntio11 cos :T = '{!-,
whi«h is impossibl<.' (Lemma 1 ).
Lemm.a 4. If there e:r:ists an cdgr. of order? (). then l = 2.

l11dePd, let {i.,i} I><' such an edge~. If l > 2, one of I.he edg<'s i.j (say i)
would b<' linked to a third wrt<'X j'. since X is connected. 111 view of the
rel at ion cos ~ = '? W<' would haw
.......
2 l ( v'3 )2
L..<i;k.?:1+ -2 =l
ki1
which is irnpossible (Lemma l ).
lemma 5. A verte:.1: mnnot bdong to two distinct rdgl's of ordrr ? 4.

Lc·t i be such a vfftcx. \Ve would haV<'
Jr-1
S:. 1dJ ? ( ./{ )2 + ( '(}-) 2 = 1, which
is impossible (Lemma 1).
Let {i. j} be an edgi- of X. \Ve are going to define a new Coxctf'r graph,
which will lw ohtaim·cl from the graph of I\ I by identifying i and j. The set I' of
its v<~rtices is the quotient of I ohtaincd by identifying i and j. Put p = {i. j},
wliich is an clemPnt of I', and identify the elements of I distinct from i and
j with their canonical images in I'. Let k:. k' be two distinct p[cmcnts of I'.
Then, {k. k'} is an edge of the new graph in the following cases:
1) k and k' arc distinct from p a11d {/r, k'} is an edge of X: in this case,
the order of this edge is defined to be fllkk':
2) k = p. and one of the sets {'i. k'}. {j. k:'} is an edge of X: the order of
{p,k'} is defined to be lll;k' if {i.k'} is an edµ;c> ofX. and 111Jk' if {j.k'} is an
edge of X (these two possibilitic~are n111t11ally c·xdusivc bc•caus<' Xis a tr<'c).
Let :\I' = (m:.1k.iEI' be the JH'W Coxet<~r matrix thus defined. and put
q:J =- cos "~ . Then, for k i p. · t;,k 0 "" <J;k + </Jk. Thus, if (f,;) E R 1',
•J
L q~·k'f,kf,k' L <JH'f.kf,~,,' + f,~ -f,l

= k.k'EI - f.] - 2qijf,if,j (1)
k,k'El' .
= 2= qu·f.kf.k" - ( 1 + 2qij )f.~.

J.:.k:'E I
lemma fi. lf {i.j} i8 of order :l. \V(:\I') i8 o. finite Co:r:ctcr grO'ltp.

s 4. CLASSIFlCATIO:"J OF BOOT SYSTEi\IS
111ckcd. qij =-~.so (1) iH'f"OllWS
a11d (f.dkEI' ,___. L l/~·k'E.kE.·A, is a positin· 11011-dq?;<'IH'nttl' quadratic forn1.

k.k'EI'
It 11ow s11fHc<'s to apply Th. 2 of Chap. V. s ·l, 110. 8.
lemma 7. Wr: luwe mw of the fiillowing:
a) X hri.s a. uniqul' mmi;fim.tion point (Chap. IV, Appf'11<lix. 110. 1 ). rind
all fhe <~d.<fl'S of X are of onfrr .'I.
b) X is n diain awl has at mos/. one edge of onla ~ ,l.
\Ve argue by i11ductio11 011 /.
a) Ass11111c that X has a rnmificatio11 point i. Thf'n i IH'longs to 3 <'dµ;cs of
on lc•r :~. {i, ~: 1 } • { i. A· 2 }, { i. A::d ( L('llllllHs 2 and :l). If I = 4 t IH· lm11na is provc•cl.
If 1101. I hcu k 1. say. belongs to au edge dist incl from thosl' just 111e11tio11f'd
since X is c:omwctl'cl. l<kntify i a11d k 1 i11 the Coxct.er graph of:\·!. This µ;iws
a 1ww graph lo which the ill(luctio11 hypot h<'sis can be applied, i11 view of
Lemma 6. Now th!' imagP p of i is a ramification point of the new p;rnph
X'. Tims X' has 110 other ra111ificatio11 point aud all its cclg<'s an' of order
:3. Thus all the cdgf's of X are of order :t and i and k 1 arc its ouly possible
ramificatio11 poi11ts. But if ~: 1 were a rarnificatiou point. of X. p would belong
1o al least ,t edges in X', contrary t.o Lcrnma 2.
b) Assume that X has 110 ramification point. Then X is a chai11 (Chap.
IV. Appendix. 110. :3, Prop. :l). L<'I {i,.J} be a11 <'dgc of order~ 4. If I= 2,
t lw lc11111w is trivial. If not, then i, say. belongs to an edge { i, k} with k =/: j
(sineP X is eonnccted). This edg<' is of order :l (Lemma. 5 ). ldent.ify i and k
i11 the Coxet.cr graph of :\-!. By Lemma {). the induction hypot.hcsis C'an he
applied. Let p be the irna.gc of i in the new graph X'. 111 X'. {p,.i} is an edge
of order ? 4, so X' ha.<-; 110 oth<'r edg<· of order ? 4. and hem·<' {i,j} is the
011ly f'dge of order ? 4 in X.
i,, br vrrti<'es of X snrh that {i1.i2}.{i2.i:i}, ...

frmma 8. Let i 1.i 2 , •.••
,,
... , { i 11 _ 1, i,,} art' edges of order :J. Then q( L re,,.) = ~p(p + l ).
T"=i
We have (<';,Ir;,)= L(e;,.IF.;,.,) = -~.(P.;,lc;J = 0 ifs> r+ l. Thus

I' fl 71- I } 71-l
q( L I'<';,)= r=l
r~·I
L ,.'2 - 2 L -r(r + 1) = p 2 - 1·=!
r·~-1 2
L r.
By Theory of Sets. Chap. m. ~ 5. no. 8. Cor. to Prop. U. this is equal to
1 1
p2 - 2p(p - l) = 2J!(J! + 1).
204 ROOT SYSTEl\IS Ch. VI
lemma 9. As.rnrw: I/wt X i.~ a «ho.in with vertices I, 2, ... , I a.nd edges
{L2}.{2,:3} .... ,{/- Lt}.
(i) rt' one of fh<' f'.dgl's {2. :~}.{:I. 4}, .... {I - 2. / - I} is of order~ ·L this
rdgc is of order 4 mul the grnph is the fullowiny:
(ii) ff' the edye { 1. 2} is of order 5, the graph is one of the following:
We ean assume that I > 2 (Lemma 4). Assume that {i. i + 1} is of order
~ 4, with l :::;; i :::;; I - 1. Put
:r = e 1 + 2e2 + · · · + ie;, y = ('; + 2e1-1 +···+(I - i)!C;+1, and j =I - i.

B.v Lemma 8. II :r 11 2 = ~i(i + 1), II .II 11 2 = hU
+ 1). On tht> other lwnd,
(:rl:y) = ij(e;l<';+d = -ijcos ~with m = 4 or[> (LPmma 4). Now
which J!;ives
1 .. (.
-11 1 + !)('J + 1) > ·2·2
1..J cm; 27f
-
4 m
so
(i + l)(j + 1) > 4ij cos2 7r ~ 2ij.
m
This ~ives, first of alL ij - i - j - l < 0. or (i - 1)(.i - 1) < 2. If
1 16 cos 2 !!_. thus cos 2 !!_ < cos 2 ~

m m 5
and hence m = 4. This proves (i). If i = 1 and m = 5. then 2j + 2 > 4.i 3+/5 ,
or j V52- 1 < 2, j < J5 + 1 < 4, and sol= j + 1:::;; 4. This proves (ii).
1 The .=ith roots of l distinct from l are the solutions of z4 + z:J + z 2 + z + 1
0. Putting x = ~ (z + ~ ), this eql,lation becomes (2.r) 2 - 2 + 2x + 1 = 0, or
4x 2 + 2x - 1 = 0, so :1: = - t{V'>. H~nce,
2 c:os 2 ~ - l = cos 2 71' = V5 - 1 cos~ ~ = 3 + v'5 > ~ > 2_ cos ~ = 1+ v'5.

5 ,5 4 5 8 8 16' 5 4
~ 4. CLASSIFICATION OF' ROOT SYSTF:l\!S 205
Dcnmw. 10. ff X has a mm.ijimtion point i. th,. f1tll sub_qroph X - {i} is the
union of thn·c d1ains. and if 11 - 1, q - l. r - l rm: thr· lr-ngfhs of lh!'.se drnins,
th<' f.ripl<' {p. q. r} is c1r1rn.l. II.JI lo n 1w11nulo.tion. to 011.r: of the fripll'.s {I. 2. 2},
{l.2,:q. {1.2.-l}. {l.l.111} (forsomr·111 ~ l}.
Th<' wrt<'x i liclo11gs to;·~ cdg<'s (Lelllma 2). and thNr is 110 ot h<'r ramifira-
tio11 point (Lemma 7). so the full subgraph X- {i} is tli<' sm11 of :3 dwins X 1 ,
X:z. X: 1 each of whkh has a t<T111i11al wrtcx li11kPd to i in X. Let { i 1 • i 2 },
{i 2 .i;i} ..... {i 1,_ 1 .i1,} lw thl' edgt's of X 1 • {j 1 .):.!}. {,h.,h} ..... {j,,_ 1 .j,1}
thos<' of X2. and {k1.k2}. {k2.A:;i} ..... {k 1 ._1.~:,.} thos<' of X:i, with ·i1 ..h,
k 1 linked to i in X. W<' <"<lll asst Ill I<' that 71 ~ q ~ r ~ I. Pnt
.r = e;,, + 2ei 1, __ , + · · · + /H'i 1

y = c;,, + 2e;,,. , + · · · + q<'j,
Z = 1'.k, + 2f'k,. 1 + '' ' + 1'<'-k1 •
Sim·c all t.hP Pdgcs of X an· of or<l<'r ;3 (Lc'lllllU1 7), Lemma 8 giv<"s II ;i· 11 2=
bP(P + I), II JJ 11 2 = ~</(<J + 1), II z 11 2 = ~r(r +I). On t.}l(' othrr hand, <'i is
orthogonal to <';r <'ir ... , t'; 1,. so (r·;l.r) = p(eiic;,) = -1p:
similarly (e,ly) =
- ~<J, (c;iz) = -~r. The 1111it wctors II :r 11- .i:, II JI 11- :i;, II z 11- 1 z arc
1 1
11111tually ort hogo11nl. a11<l c; do<'S not b<'long to thP snbspan' F that they
µ;cncrntc: tlll' square of th(' clistcmc·1' fro111 c, to F is
x 2 ij ~ z 2
I - (1:,I~) - (f';I~)- - (f'ilN)
I p q r
=1----------
2µ+ l 2q+I 2r+l
I l I I 1 l 1 1 1
=l- - + - - - - - + - - - - - + - - - .
2 2p+l 2 2q+l 2 2r +l
Since this quantity is > 0. W<' have
(p+ J)-1+(q+1)-1+(r+1)-1>1. (4)
Henn· 3(r + 1)- 1 > l. so r < 2 aucl fi11ally r =I. Then (1) gives
(JI + l) - I + (I/ + l) - 1 > 2l (5)
he11n: 2(q + 0- 1 >~·so(/,,;;; 2. Finally. if I/= 2. (::i) giv!'S
I 1
(p + 1)- > (i' hence p::;;: 4.
THEORE.'\1 I. The graph of any irrr.du.ciblr. finite Co.rl'fer systr.m (W, S) is

·i8om01phir: to rm1: of the following:
20() HOOT SYSTEl\IS Ch. VI
A1 ~ ... --0---00 (l ~ 1 vertices)

B1 0----0.--.....0 - ... - o - -4- o (l ~ 2 1•ertie'l's)
/0
D1 C>----0--0- ... -<>---<> " ' - 0 (l ~ 4 l'ertices)
Ee 0------0------0
I
0
I
Es ~~r~0----0
<>--<>--~o
4
o~
p
o---<>(p = 5 or p ~ 7).
No two of Ow.~c grn,ph8 fL1'f'. isomorphic.

ludc<'d, let W = (m;.1) be th<' Coxctcr matrix of (\V.S). a11d let l =
Card(S). If 011<' of the llliJ is ~ 6. thn1 l = 2 (L<>1m11a -1) and the CoxC'tcr
graph of (\V,S) is of t.ypt> G2 or 12 (p) with p:;? 7. Ass11mc now that all the
m;.i ::;;; G.
a) If tlw 111.;.i arc not nll cq11al to :t th<· graph X of (W.S) is a chain arnl
t>xact ly 011<' of the mu is eq11a.I to 4 or 5 (Lemma 7). If 011<' of the miJ is cq11al
to 5. Lc11nua 9 shows that we hav<• one of llH' types H;i, H1 or 12 ( 5). If one
of the 111.;,; is equal to 4. Lc111111<1 9 shows that W<' lrnve 011<' of the typt'S 131 or
F1.
b) Ass mm• now that all the m;.1 arc <>q11al to :t If X is a chain, I he Cox~·tcr
graph is of type A1. If not, Lem1~1a HJ shows that it is of type E 6 , E1. Es or
o,.
The fact that no two of tlw Cuxetcr graphs listed are isomorphic is dear.
Conv<'rsdy:
THEOREM 2. The Co:1:d<T .<rm·u11s defirwd by the Cm·ctcr qraphs A,. 131 •... ,
12(/1) of Th. I arc finite. l
This is clear for 12 (p), tht> corqespo11di11µ; µ;n>HJ> being th<' dihedral group
of order 2p (Chap. IV. ~ L 110. 1J). Fin· H 4 tlw corresponding quadratic form
~ 4. Cl.ASSIFIC'ATION OF !WOT SYSTEl\IS 207
IS
2 2 2 2 7T
C:1 + (2 + (1 + (1 - (16 - (2(1 - 2(cos r: )6(1
,)
2
::::: (1 + (22 + t::'')i + (;" - 1+JG
(1(2 - (26 - --2-6(1
(I + (! 2 l + JG 2 :3 1 2 7 - 3 JG 2
= ((2 - - 2 - ) + (C:~ - - ..-,-(1) + 4(C:1 - 36) + 2-l (i·
Since 7 - :~JG is > 0. this forrn is positiw 11on-dc•g<>11rrntc. a11d tlw corre-
spo11di11g Coxl'tt>r group is ti nit<'. Tht• sam(' holds for that corresponding to
H:J· since it is isomorphic to a s11ligro11p of 1·lH' pn'C'<'di11g (Chap. IV, ~ 1,
no. 8).
For tllC' types A,, B1 .. ... G 2 . we shall co11st rnct. in :-\os. 5 to 13, root
s,vstcrns haYing tlw corrC'sponding groups as \V<'yl groups. \Vt• shall SC'<' that
these gro11ps an· not only finite. hnt f'l'lJ8f.nlloymphir: rn
2. no. :i).
2. DYNKIN GRAPHS
By abuse of language. we shall call a normed _qmph a pair (I'. .f) having the
following propCl'ti<'s:
l) r is a graph (cnllc·d tlw 11nderlyin.r1 graph of (r. .f)).
2) If E denotes lhe set of pain; (i.j) such that {i.j} is an edg<' of I' . .f is
a rnap from E to R such that f(i,j)f(j. i) = 1 for all (i.j) EE.
There is an obYio11s notion of isomorphism of normed graphs.
Let R he a rcduc0d root systt>m in a real vector space V. \Ve arc going
to associate to it n 11orn1ed graph (X. f). called tlw /Jynkin qmph of fL The
vertices of X will lw the d('U1e11ts of the set I of orbits of \V(R) iu the union
of the sets {B} x B (wlH'rP Bis the set of has('s of R). If~= (n;i)i.jEI (rC'sp.
'.\! = (m;J)i.JEI) is the cawmieal Cartan lllatrix (resp. the Coxetcr lllatrix) of
R (S 1. no. 5, Remark 7), two vcrtic<·s i and j of X are link(~d if a11Cl ouly if
11 iJ f= 0 and we then put
·. ') = Tl;j •
. f( l . .J
Tlji
Sin('c n;,; = 0 implies n;; = 0, this ddim•s a normed graph (X . .f).

Let (.rl11) lw a scalar product 011 V. i11vnriant 1111der \\'(R), irnd IPt
B = (n;);Ef IH' a basis of R. indexed ca11011icnll.v. Fornn1las (7) am\ (9) of
3 1. no. 1 show that wrtie<'s i a11d j of th~ grnph X arc• linkPcl if a11d only if
(n;)n;) i 0
and that
In view of the results of s

1, :\'os. :3 and f>. the only possibiliti0s an• the
following, np to interchanging i and j:
1) i and .i are not linked: nu = ni; = 0; 111;1 = 2;
2) f(i.j) = f(.i. i) =- l; nu = nJ; = -1; 111;.i =- :3;
:3) f(i.j) = 2. f(j, i) = 1/2; 11;.; = -2. 111; =-I; mij = 4;
4) f(i,.f) = :J, f(j, i) = 1/3: rlij = - :t 11ji = -1: Ill;, = 6.
We see from this that the Dynkin graph R determines the Cartan matrix
and the Coxctcr matrix of R, and hence dctcrmi1ws R up to isomorphism.
More precisely, the Cor. of Prop. 15 of§ 1. no. f> iu1plies the following result:
PHOPOSITlO:\' 1. Let R 1 and R 2 lie two reduced root systems in vector

spaces Vi n.nd V2. [,et I31 = (<t;)iEl, and Ih = (o;),EJ, be bases of R1 and
R.i, index<~d rn.nonically. Let >. be an ismrwrphism from the Dynkin graph of
R l to the Dynkin graph of R2. Then, there e:rists a unique isornorph:ism from
V 1 to V2 tra.nsfu111ii·11..<J R 1 into R2 and a; into n.x(i) fin· all i E I 1 •
It is clear that an autornorphism of H dcfi1ws an automorphism of the
Dynkin graph of R. and hence a homomorphism .,:; from the group A ( R.) to
the group of automorphisms of the Dynkin graph of IL
COROLLARY. The lunnomorphism. 9 defin<~s /1y pas.rnqe to thr quotient an

'isomor7Jhism .frnm the gronp A(R)/W(R) lo the group of antorrwrphisms of
the Dynkin graph of IL
Clearly, <p(g) = Id for ally E \V(R). 011 the other hand, Prop. I shows
that there exists an isomorphism ·1/-' from the group of isomorphisms of the
Dy11kin graph of R to the subgroup E of clements of A(R) leaving fixed a
given basis I3 of R, such that :.po·~J = Id. Since A(R) is the semi-direct product
of E and W(R) (s 1. no. 5. Prop. Hi), the corollary follows.
ln praeticc, the Dynkin graph (X, f) is rcpwscuted by a diagram composed
of nodes a11d bonds i11 the following way. The nodes correspond to the vertices
of X; two 11odes corresponding to two distinct wrticcs i and j are joined by
0, L 2 or :3 bonds in cases 1). 2). 3) and 1) above (up to interchanging i and
j). I\lorcowr. i11 cases :3) and 4). that is wlw11 f(i.j) > 1. or wll<'ll the roots
n 1 and OJ arc not orthogonal and not of the sa1111' length. an inequality sign
> is placed 011 the double or t.ripk bowl joi11ing the nodes corrcspo11ding to
i and j orieut.t>d towards the node corresponding to j (that. is, the shortest
root):
c=::::>:=
i j
(for f(i,j) = 2), ,
1. _1. (for f(i,j) = 3).
It is clear that the Dynkin graph (X, f) cau bc rccovf'rcd from this diagram.
We remark that the diagram asi;oriat(•d to the Coxctcr graph of W(R) can
be obtaiuecl from that associated to the Dynkiu graph of H by ket>ping the
no<les and the single bonds and rl'placiug the double (resp. triple) bonds hy
a bond smmounted by the number 4 (resp. 6). Conversely. given the Coxcter
~ 4. CLASSIFICATION OF ROOT SYSTE\IS 209
graph of W(R), th(' diagra111 associat.(~d t.o tlw Dy11ki11 grnph of H can be
n·C'ov1•n•d 11sing tl1c inw·rse of this procPd11n~. cxn~pt for the inequality signs
on the double or t.ripk bonds. Th. 1 t h11s giws im11wdiatdy the list of possibk
Dy11ki11 grnphs. !\fore prPciscly:
THEOREiVI :~. If R is an irrcducihlc rr.duced root. system. its Dynk'in graph i8

isomorphic to om: of I.he graphs l'l'.fJ'f'C81'lltcd by !hr: fi>llm11in!J din.yra.ms:
Az 0 - - 0 - 0 - ... -o-o (I ;:,, I vertices )

B1 0 - - 0 - - 0 - ... -<r---o= (l ;:,, 2 vertices)
Ct 0--0--0- · · · -o---o= (l ;:,, 3 vertices )
0
Dz 0 - 0 - - 0 - ... -o-o< (l ;:,, 4 vertices)
0
E6 0-0--0-0-----<>
E1 01-0-0---0
E8 0-0-1-0-o---<>---o
f 4
G2 ~
No two of these graphs a.rr. isom0171hic a.rid there !Lr<'. irrcdudhle reduced root
systems luzvin,q mch of fh<'m as their DynA:in yraph (np to isomorpl!'i.mt).
Tlw first. asst'rtio11 follows immcdiat.dy fro111 Th. l, i11 view of the preced-
ing remarks, t.hc fact that the CoxetPr gro11ps of the graph:; H:i, H 1 mHI I 2 (p)
(for JI = 5 and p ~ 7) art> not crystallographic, and till' fact that the two
possible irn•q11alitics for the doubk• (resp. trip It•) bond of the Dynkin graph
asso<:iat<'d to the Coxeter graph F 4 (n~sp. G 2 ) give isomorphic Dynkin graphs.
The second assNtion is clear a11<l the third will follow from the explicit con-
strnction of an incducihlP wduccd root system for each type. a construct.ion
that. will be carried out in Nos. 5 to 1:3.
Remo.r-b. l) Tlw graph A 1 reduces to a single nod<'; we denote• it also by

B 1 or C 1 . Thr. graph Il 2 = is also denoted by C 2 . The gm ph A3
o-<>---o is also denoted by D: 1 • Finall.v. D2 denotPs the graph consisting
of two unro11111'ct1'cl 11otks. (These c·o11vc11tio11s are deriwd from thl' properties
of the corresponding root systems, cf. nos. 5 to 8.)
2) If (X, .f) i:; the Dynkin graph of a n·d11eccl root. systPlll R. t.hc Dynkin
graph of the invcn.;(• systcn1 nm be identified with (X. 1- 1 ). In otlwr words,
tht' diagram associated to the Dy11ki11 graph of n· ca11 be obtai11<•d from that
associatl'd to the Dynkin graph of n by rcwrsing 1lw i11equalit.v signs. If R
is irreducible, WC SC(' that n is isomorphic to re 1111less R is of type B1 or C1,
in which case rr· is of t.ypc C 1 or B1•
210 HOOT SYSTEi\IS Ch. VI
3. AFFINE WEYL GROUP AND COMPLETED DYNKIN

GRAPH
Ld n l)(' an irn~dudbll' n•d11C'cd root sysl1•111 and IPt (X. f) lw its D.vnkin
graph. \Ve arc going to define a not her 11<m11cd graph (X . .f) that we shall ca II
the rompldcd Dynkin qm.ph of R. Tlw sd i of V<'rtic<'s of X co11sists of th<• set
I of vertic<'s of X and a vertex de11oted hy 0. not l)('lo11gi11g to I. To define .f,
choose a basis Il = (o;)iEI of R allll a scalar product (.rly) invariant 11ndf'r
W(R). L<>t n 0 be the negative of lh<' hiqhest rout. for the onkr dcfi11<•d by I3.
Two distinct vertices i.j E i are linked if and onlv if {n;lni) f 0 and we then
p11t
f(i,j) = (o,1n;).
(niln1)
It is in111wdiate that the graph X and the map .f thus defined do not depend
011 the ('hoicc of 13 or thl' scalar produd.
If tlw rank I of H is equal to I. th<'ll l = {i} and n 0 = -n: 1 : hence
f(O, i) = 1. If l ;? 2, no is 11ot proportional to any of the o; and (0 0 10;) is
'.( 0 (§ l. 110. 8. Prop. 25). For any pair {i,j) of disti11ct dcmcnts of L the
only possibilities an• those de11ott-d by 1). 2). ;3), ·!) in the pn•c<~ding 1111111her
(putting, for example. n. 0 ; = 11(n0. n,) aud 111 0; = order of sn 11 s,., for all i EI).
l 11 th<• case f ;? 2. I he compll'tcd D.vnkill graph is rqm~scut.cd hy a diagram
with thP same <:ouvcnt ious as i11 the prccPdiug lllllllher. somd.imcs i11dicati11g
hy dotted lines the boll(ls joining the wrtex () to th<' other wrticcs. \Ve
remark that the inequality sig11 > oil s11ch a ho11d. if it exists. is always
directed towards the v1~rtex distill(·( from 0. since' n 0 is a longc•st root (§ 1,
110. 8, Prop. 25). Tlw graph (X.f) is th1.• subgraph obtai1wd from (X./) by
deleting th<' wrt<'x 0.
The action of A(R) on (X. f) extends to an action oil (X. /). ll~aving 0
fixed. and \V(R) acts trivially 011 (X . . h
We retain the 11otat ions of ~ 2. Prop. 5 of § 2. 110. 2. t.ogetlwr with Th.
l of Chap. V, § J, 110. 2, shows that th<' Coxcter graph E of the affine Weyl
group W,, (R) can be obtainl'd from (X. f) by the same rules by which the
Coxctcr graph of W(B) is obtained from (X.f). Oil tlw other ham!. i<'t G
be the uormaliser of vV,,(H) (§ 2. no. ;3). To ally .<J E G corresponds an
automorphism ..p(y) of L' and ..p(.(J) = ld if .IJ E W., (ll). Conwrsel~·. given a.n
automorphism .>. of E there is. by Prop. 11 of Clwp. V, § 4. no. !J. a unique
element .<J = l/-'(.>.) prest•rviug a given akov<' (' alld such that ..p(!J) = .>.. Since
G is the semi-direct product of tll1c subgroup Ge of elements pn'Sl'rvi11g C
and Wa(R) rn 2. 110. :J). we dPdude that y defines by passage to the quotient
an isom017Jh:i8m from. G/Wa (m from Ge) tu Aut(E). It is immediate that
the composite of this isomorphisrn with the cauonical map fro111 A(R)/W(R)
to G/W,, coincides with the homomorphism from A(R)/W(R) to Aut(E)
iuducPd by the homomorphism frow A(H)/\V(R) t.o Aut(X. /) dditwd above.
By § 2. uo. :~, the group Aut(I:) is isomorphic to the semi-direct product of
~ 4. CLASSIFICATION OF ROOT SYSTE\IS 211
A(R)/W(R) hy P(W)/Q(W). and P(W)/Q(Ir) is isornorphic- to the group

s
l'c =- Gen w;, (with the 11otatio11 of 2. 110. :l): tlw clc111c•11t of Aut(E)
C'OJT('Spo11di11g to thP e!Pllll'Jlt. ')'; of r(' trnnsforms the \Trtcx () t.o the vertex
i of E.
Rnnark. It cau he shown t·h<lt. tlw canouical 1w1p
Aut(X. f)-. Aut(E)
is a.11 isomorph is111.
THEOREl\l 4. Let (\V. S) b(', 11.n inrdur?ihlr Co:relf'r system. with S .finifP. The
associated 1r1tadmtic form (Chap. V, ~ 4. no. l) is pos'iti11c and degfnnnlf if
o.nd ouly ~f the Co.1:rfrr .</mph <d (\V. S) is i.wmunph:ic to mw <~f the following:
A1 0_22._0
/o--o,_
Ai (l ;;;: 2) o·.... ··· ... ':a
(circuit with I + I vertices)
"'o--o/
B2 0__±_0_..1:_0
4
o/" ' 0---0- ... -o--o
Bi (l ;;;: 3) (I+ I l'ertices)
0'
Ci (l ;;;: 3) 4
0--0------0- . ..
4
--0--0--0 (I + I vertices)
0 /0
Di (t ;;;: 4) "'0-···-0 (I+ I vertice.I')
o/ "'o
0-----0-1-0---0
l
0
o-o---o-r-0-0---0
o-o-,-0-0-0-----o-o
4
0--0--0--0--0
o--o_Q_o
No two of lh<'sc Cruf'lcr gmphs are isomorphfr.

s s
By C'hap. V. '1, no. !.) and Prop. 8 of 2. no . .5. the Coxeter systems
whose quadrati« form is posit iw and ckge11crntc• arl' tlios<' whid1 c·orrespoml
212 HOOT SYSTE:\IS Ch. VI
to the affine \Ve.vi groups of irr<'d11ciblc reduced root systems. The theorem
therefore follows from th(' dct<'rmi11atio11 of the eo111pl<~tcd Dy11kin graphs
made in !HIS. f> to 1:~ below.
4. PRELIMINARIES TO THE CONSTRUCTION OF ROOT

SYSTEMS
Ll't. V b(' a real vector space of di1ue11sio11 I ;): 1 ('q11ippC'd with a scalar
product (.rly). La discrete s11bgro11p of V. A a li11it.<' set of lllllllbers > 0. and
R the set of n EL s11eh that (oln) EA. Ass11111<' that R generates V and that,
for auy pair (<L /i) of points of IL th(' rn1ml>n 2 ~;::;:~ is au iuteg<'r. Thell. R i8
a roof system in V. Ii1dc•<'<I. R dearly satisfies (RS1). Leto ER: let s 0 be the
1
orthogonal rcHc'<'tiou :r ,_., :r- 2((.1:i
no 1" )o: theu, if ·.1 ER. we' have 2(Pl "l) E Z,
n 1n:
so s,.(;:I) E L awl rnornovcr II s" (!:I) II= II t:I II· so s,.(!:I) E R: t.h11s. R sat isfks
(RS11) and (RS111). a11d is n•cl11cecl if A do<•s 11ot contain two u11mbcrs of the
form A awl 1A.
Now let V he a s11bspnc<' of E = R". L<'t (~ 1 ....• E,,) lH' the• rn11011icnl hasis
of E: we equip E with t\l(' sealar product for which this basis is ort ho11ormal
and ideutify E* with E (resp. v·
with V) by lllCHllS of this sc<ilnr prod11ct..
We define s11bgro11ps Lo. Lt, L 2 . L: 1 of E as follows:
I) Lo is t\l(' Z-rnodule with basis (E;). W<' haw (01;1) E Z for all o:.8 E Lo.
The vectors n E L 0 for which (rilo) = I an' th<' ±E; (J :::::; i:::::; n): those for
which (nlo) = 2 arP the ±E; ± E1 for i < j (the two si11:ns ±, in ±c:, ± Ej,
arP d1ose11 indq><'H<knt 1.v of 1•11ch oth1•r: W<' adopt au a11alogo11s convc·11t ion
t hro11gho11t the n•mai11cler of this sc•ct io11).
11.
2) L1 is tlw Z-s11lm1od11IP or Lo c:o11sisti11g of the .r = L f,;E:; E Lo sud1

i=I
that L"
i=l
~; is <'Wll; si11c<' f.; and t,;
.
have U1e same parity. this is eq11ivakut
to requiring that (.tl;r) is even. Let L~ be tlw submodule of L 1 gcnerntc<l
n n
by the t:; ± E;: we hav('

.
L E,;E; = (L f.;)c,,
i=I i~l
(mod. L;). and sine<> 2E,. =
(c: 1 + E,,) - (c-1 - E,,) E L'1 . it follow::; that L'1 = L 1 . Since L0 is generated by
L 1 and c- 1 • L 0 /L 1 is isomorphic to Z/2Z.
" n
3) L2 =Lo+ Z(~ LE;). It is clear that an ckment .i: = L f,;E; of Vis in
i=l i=l
L2 if and only if
2f., E Z. (; - f.J E Z for all i and j. (6)

n n
Since (~~:11 LE;)=~ for all k. 1t11<l sinec II~ LE:,
1.=I t=I
IJ 2 = 1· we have (n-1/3) E
~z for n, /i E L 2 if 11 is even. TIHf group L2/L 0 is isomorphic to Z/2Z.
q 4. CLASSIFICATION OF ROOT SYSTF:l\IS 213
" n
-1) L:1 = L, + ZO L c:,). Jf n is a 11111ltipl<' of ·1. L:1 is tll<' S<'t of L C,;E;
i=I ;=l
n
satisfying (G) and th<' rondition L (;
i=l
E 2Z: in this case. (nl::J) E Z for all
;1 E L:1-
rt.
H is immPdiate that th<' s11hgro11p of E associat<'d to L 0 (resp. L 1 • L2) is
Lo (resp. L2. L 1 ). Th<' s11bgro11p of E associated to L:i is the set of
11
:r = L C,;c:i
i=l
E L2
11 n
such that (;rH L €;)
1=1
E Z. that is. such that L Ci
1.=I
E 2Z: if n =0 (mo<l. 4),
t.his associated 1mbgroup is therefore L:i-
The abdian group L2/L 1 is of order 4. aud hence is il;omorphir to Z/4Z
or Z/2Z x Z/2Z (Algebra, Chap. VIL § 4. 110. G. Th. :3). If n is odd,
1 //
p(-
2
L c-d E L1
i=l
<=>fl= 0 (mod. 1)
so L2fL 1 is cydic of order 4. If /1. is <'VPll,
l //
v( ~ LE;) E L1 <=> p
2 1~1
= 0 (rno<l. 2)
so L2/L 1 • which contains two distiuct dcmcnts of or<lcr 2. is isomorphic to

Z/2Z x Z/2Z.
We shall use this 11ot:ation throughout the next nine nos. and in t.h(' plates.
For each type of Dynki11 graph i11 Th. :3, we shall dcsnibe cxplicit.ly:
(I) A root systE'm Rand the number of roots.
(II) A basis B of R. and th<' corresponding positive' roots. The basis B
will he indexed by t.he integprs 1..... l.
(III) Tlw Coxcter numlwr h (~ 1. no. ll).
(IV) The highest root ii (for the order defi11C'd by 13) and the complct.<>d
Dy11ki11 graph (no. :J). WP shall indicate ucxt. to each vertPx the eorresponding
root of B.
(V) Th<' i11versc system R·, the cano11ical bilinear form and the constant
, ( n) rn i. 110. 12).
(VI) The fundauw11tal weights relat.ivc to B (s L no. 10).
(VH) The snru of the positiv<' roots.
(VHJ) Tlw groups P(R). Q(R). P(R)/Q(R) and the conrl<'ctiou indC'x (§ 1,
HO. 9).
(IX) The exponents of \\'(R) (Chap. V. § fi. no. 2. Dcf. 2). Jn the casC's
A1. E3 1 • C 1 a11d D1 we determine the symmetric invariants.
(X) The order of \\'(R) (and in some cases its struct1m').
(XI) The group A(R)/W(ll). its H<'tion on t.lw Dynkin graph. and the
dt'111Pnt 11•0 of VV(ll) that trnnsfor111s B to - B.
(XII) The artion of P(R:)/Q(R-) 011 th(• c0111pktcd D\·11kin graph and the
act.ion of A(R)/W(H) ou P(W)/Q(tr).
For eaeh Dynki11 graph i11 Th. ~l. t hPst• data will lw collt•cted in Plat0s I
to IX. and ordered in a uniform way as above. \\'c also gi\'(':
(XIII) The Cmtan matrix, from which the' Dynkin graph is dcriv0d as
d<>s<Tibcd in no. 2.
5. SYSTEMS OF TYPE Bl (l :;::: 2)
(I) \Ve consider the group Lo in V = R 1 (no. 4). Let. R he the set of
n E Lo such that {olo) = 1 or (n1n) = 2. i11 other words the set of vectors
±E; {1 ~ i ~I) arnl ±c:; ± C:j (1 ~ i < j ~I). It is ckar that 11 p;PlH'rntPs V
and that 2(nld)/(nln) E Z for all 11 . .rJ E R. Thus H b a n·ehtcr•d root systpm
in V (no. 4). The number of roots is n = 21 + ,1'U; 1 > = 2l 2 .
(JI) Put
ft) = C:1 - E:2. 02 = E2 - E:3 ....• 0/-l =Et-I - c/. fl'./ =El·
Then
E;=n;+n;+1+···+01 (l~i~l)
E; + Ej = (o; + Oi+I + ... + 01) + (n, + r\'j+l + ... + n,) (1 ~ i <j ~ l)

C:; - Ej = <l; + O;+ l + · · · + Oj- l (1 ~ i <j ~ f).
Tims (11 1 . n 2 •.... ni) is a basis of R (~ I. no. 7. Cor. ;~of Prop. 20). ;\forcovcr,
II o; 11 2 = 2 for i < l, 11111 11 2 = 1. (r~;ln;+1) = -1for1~i~I-1. (o;lo;) = 0
for j > 'i + I: lhc Dy11ki11 graph of R is tlms of typ<• B 1• which shows that R
is irn_•<htcibl(>. Tlw positiw roots aw the c:, and the E; ± C:j (i < j).
(III) By Th. I (ii) of Chap. V, s
6. no. 2.
h = 11/l = 2l.
(IV) Leto = c: 1 + c:2 = n 1 + 2n 2 + 2n:1 + · · · + 2n,. wlii('h is a root. The
sum of its f'oordinatl's n_ ·lat ivP tfl the basis ( n;) is 2l - I = h - 1. I_n vif'w of
s
Prop. :H of 1. no. 11. <\ is the highest root of R. \Ve• have (<\.In;) = 0 for
i -f= 2 aud (iilo 2 ) = 1. Since n 2 s of knp;th 1 (n·sp. J2) when l = 2 (rcsp.
I :;::: :$). the rnmpleted D.vnkin g1" ph of H is as follows:
for l = 2
"'z "'i
~ ~ ... - =
~ "'1-1"'1
~ 4. CLASSIFICATION OF ROOT SYSTF::\IS 215
(V) Tlw form11la n · = (";1:,)

giws for W t.lw set of vcC'tors ±2E; ( l :::;; i :::;; I),
±E; ± Ej (I :::;; i < j :::;; l). The Dynki11 graph of W is obtairwd from that of R
h.r tlu' prnct•dme Pxplaincd in 110. 2. a11d we s1·c that R. is of type C1•
There an• -H - 2 roots 1101 orlhogoual to ;-J = E1. rn1mdy ±c 1 a11d ±E 1 ±:0,;
for 2:::;; .i:::;; /:for l'aC'h such root n. n(u . .'-1) = ±2. Formula (17) of~ 1,
no. 12 shows that. for <l>n, th<• sq11an· of the length of /J is (4/ - 2)- 1: hence
<l>n(.r. y) = (.rl11)/(tl - 2). Apply form11la (18) of~ 1. no. 12 with ;i; = y = (J.
This giws
2 + !(-H - 4) = "',·(R)- 1-
4 ' -If - 2
awl so 1(R) = (l + 1)(4/ - 2).

(VI) The fu11da111cntal weights w; (l:::;; i:::;; I) such that (w,ln.i) = 6;_; arc
P<tsily nilc11Int.cd. and Wl' find that
W; = cJ + Ez + · · · + E;
= 0-1 + 202 + · · · + (i - l)n;_ 1 + i(n1 + n;+1 + · · · + 01) (i <I)
1 l
W1 = 2(c1 + €2 + · · · + E1) = 2(n 1 + 202 + · · · + lo 1 ).
(VII) TIH' sum of tl1<• positiv<' roots is
2p = (2/ - l)c-1 + (21 + · · · + 3c1-1 + E1

- :J)c:z
= (2l - l)n1 + 2(2/ - 2)02 + · -· + i(2l - i)n; + · · · + /2n1.
(Vlll) \V<' have Q(ll) = L0 (no. -1). a11d P(H) is generated by Q(R) and
w1. IH'IW<' is equal to L:z (no. -1). Henn~. P(R)/Q(R) is ismuorphic to Z/2Z,
aud the comu·ction index is pq11al to 2.
(IX) aud (X) I11 R 1• the orthogonal n•H!'ction s,, -£, (if- j) iut<:'rchanges
=; and E J and leaves ::: k invariant. whell the illd<~x k is dist inct from i and
j. Tlw s,,-c, gc1wrntP a µ;ro11p G 1 isomorphic to thP s.nmnctric group 61.
Thl' orthogonal rdkctioll s,, tra11sforn1s E; to -E; and kavcs invariant the Ek
when the ind<~x I.: is distinct from i. The s,, generate a group G 2 isomorphic
to (Z/2Z) 1• The \Vey! group \V(H) is µ;<'11<·rated by G 1 and G 2 , a11d G 2 is
uonnal ill W(R). so W(R) is iso111orphir to the serni-dircct product of 6 1 hy
(Z/2Z) 1 • Its ord<'r is t.hc•rpfore 2 1.l!.
Tlw symmetric algchra S(R1 ) can he i1kntified canonically with th<' alge-
bra of polynomial functions P(E, 1..... f.1) 011 R 1• For such a polynomial to he
invariant under W(H). it is first of all 1wc·1•ssar,v that
for all choices of till' signs 011 the right-ham! side. so that
p ( E. l •.•.• E.1) = Q(~ 1'2 ..... E.12 )

when' Q is a poly11omi<1l, and further that Q is a syrnrnetrir polynomial;

nnd these conditions are sufficient. Conscq11ently (Algcbrn. Chap. V, App. I),
S(Rt)w(H) is the algebra gc11cratcd by the l polynomial functions
"\"""' 2 2 2
f; = ~ ~T(J)~T(2) • .. ~T(I) (L:(i:(l).
Tf\Ot
\furcover. the trnnscc11dcnc<' degr<'(' over R of t.he fidd of fractions of

S(R1)W(R) is /. so tlw t, ar<' algcbrnirally i11dcpe11dcnt. Sinn' th<' t, have
degree 2.4, ... , 21. we rnncludc t.hat th<' exponents of W(l1) (Chap. V. ~ 6,
no. 3. Prop. :~) are:
I. :1, 5, .... 21 - I.
(XI) The only automorphism of the Dynkin graph is t.hc identity clcm('Ht.
Thus, A(R) = W(R) and -1 E. W(R). Since -1 transforms I3 to -I3, we
conclude that 1110 = - 1.
(XII) The group P(W)/Q(W) is d11al to P(R.)/Q(R). and hcncC' is isomor-
phic to Z/2Z. Its noH-trivi<1l clement permutes the vertices corresponding to
rt 0 and n 1 and leav<'S the others fixed.
6. SYSTEMS OF TYPE Ct (l .? 2)
(I) The ('Xist('IHT of root systems of type C1 has been proved in no. 5,
since W<' haw seen that the inverse syst<·rn of a system of typ<' 13 1 is of type
Ct. A root systP111 of type Ct is tlms ohta.irn~d hv taking in R 1 tlw V<'Ctors
±2:=:; (1 :( i :( l). and ±c; ± EJ (1 :( i < .J :(I). The number of roots is 21 2 .
(ll) A basis of R can Ii<' obt.ai1H'd by taking th<' irnag(' und<'r th<' map
n r--+ (,;1;,) of l h<' basis of the systP111 consid<'r<'d i11 no. f>. \Ve obtain:
n1 =CJ - E2. n2 = c2 - E:i, .... <l[-1 =Et- J - E/. 1'/ = 2c1.
Thl' positiw roots are the 2c:; and the E, ± cJ (i < j).
(III) TIH' Coxctcr number is the same as for the inversl' syst<•m: h = 2l.
(IV) LP!(-~ = 2E1 = 2n 1 + 2n2 + · · · + 2nt _ 1 + 01. which is a root. Th<' snm
of its coordinat<'s rdat.ive to (n,) is 2[ - 1 = h - l. IIl'11cc. <\,is the highest.
root. WP have (ii10 1 ) = 0 for i ~ l, (iv[oi) = 2. Tims. t.lw cmnpld('d Dynkin
graph is
(V) w(' have already d<'l<'fll~IIWd W, which is of typt• I3,. By formula (10)
of 8 1, 110. 12 and by 110. 5 ( V). the square of th<' length of 2c:; for <PR is
thus, <Pn(.r,y) = (.cly)/4(1+1).

~ 4. CLASSIFICATION OF ROOT SYSTEJ\IS 217
We have 1(R) = 1(H:) =(I+ 1)(41 - 2).

(VI) The fundamental w<•ights arc ea.sily found:
l
= 01 + 2n2 + · · · + (i - l)n;_1 + i(n; + ll';+1 + · · · + -n1)
2
(i ~I).
(VII) The sum of tht• positive roots is
2p = 2lc 1 + (2l - 2)c2 + ··· + 4E1-1 + 2E1

= 2ln 1 + 2(2l - I )0::1 + · · · + i(2l - i + l )n; + · · ·
1
· · · + (l - 1)(/ + 2)o,_ 1 + 2l(l + l)n1.
(VIII) By 110. 4 and 110. 5 (VIII). Q(R) = L1 . P(H) = L0 : P(R)/Q(R) is

isomorphic to Z/2Z. and the cmmectio11 iwkx b 2.
(IX) a11d (X) These data depe11d only 011 \V(R). aud so are t.hc same ns
for type B1.
(XI) The same argument as in 110. G shows that A(H) = W(R) and
11.'(I = -1.
(XII) Tlw single non-identity <>lenw11t of P(W)/Q(W) defines the unique
11on-trivial automorphism of the completed Dynki11 graph: it intNchangcs the
vertices corrcsponcling t.o nj a11d n 1_.i for 0 ~ j ~ f.
7. SYSTEMS OF TYPE Al (l ~ 1)
l+l
(l) a11d (Ir) Let V lw the hyp<'rplanc in E = R 1+ 1 with equation L ~i = 0.
1=)
Replacing l by l +I i11 no. 5. we obtain a system R' of type B 1+ 1 in E with

basis
Since n 1 •.... o 1 generate V. R = R' n V is a root system in V with basis

(o 1 ••.•• o 1) (§ L no. 7, Cor. 4 of Prop. 20). By the calculation of the scalar
products in no. 5, it b immediate that R is of type A,. The clements of R arc
the c; - Ej (i 71 j. l ~ i ~ l + l. 1 ~ j ~ l + l). There are n = l(l + 1) roots.
The positive roots arc the E; - Ej where i < j.
= n/1 = l + l.
(III) We have h
(IV) Let <"i = E 1 - E1+ 1 = n 1 + n2 + · · · + n1. which is a root.. The sum of
its rnordiuat.es relative to (o;) b l = h. - 1. Hence ii is the highest root.
For l = l. ii= n1 so (ii-ln1) = 2; the Coxeter graph of the gro11p W (R) 0
00
o--o.
a,
For l ~ 2. (iiln;) = 0 for 0 < i <land (iiloi) = (r)1oi) = l. Hence t.hc

completed Dynki11 graph is:
(V) Identifying V with its dnal nsing th<' sralar product. we have n· =
= o for all o E R, so R: = R.
2" )
( o·)<\.'
For tlw form <Pn. tll(' l<'11gth of the roots is h- 112 = (l + 1)- 112 (§ 1,
no. 12): so <l>,.(.r, y) = (;1:1y)/2(l + l).
We hav<~ 1(R) = (l + 1) 2 (s I. 110. 12. fon1111la (20)).
(VI) Let (w;) 1,,;; 1 ,,;;1 be the family of fundamc11t.al weights. Put
l+I
w; = L
)= l
f,;jEJ·
.
with f.;j ER.
The conditions (w;ln.i) = 6;.i and w; EV µ;ive
f+I
f.;; - f.i.i+l = 1, f,;j - f.i.j+I = 0 for j ic i. L: fo = 0,
1~1
which easily IPad to
W; = E1 + ... + c; i
- --(E1 + ... + c/+1 )
l +1
l
= -((I - i + l)(n 1 + 2o 2 + · · · + (i - l)o;_i)
I+ 1
+ i((I - i + l)n; + (l - i)n;+1 + · · · + n1)).
(VII) The sum of th<' positive roots is
2p = f<:t + (/ - 2)E>,? + (l + · · · - (l - 2)c1 - 1€1+1

- 4)E;1
= ln1 + 2(1 - I )02 + · · · + i(l - i + I )n 1 + · · · + ln1.
(VIII) Introduce i11 E = R 1 +'
tlw subgroup Lo of no. 4. Let p he the
ort hogo11al projection of E ont,o V. l3y ~ 1. 110. 10, Prop. 28. we have
Q(R) = Q(R') nV ~ Lu n V. awl P(R) = p(P(R'));

in view of t.hc fact that t hl' !art hu1damcnt.al wci~ht of R' is orthogonal to
V. we have P(R) = p(Q(H')) = p(L 0 ). Thus. P(R) ts the gronp gc11ernted by
I+ l
tlw E:; - Ej and by p(Ei)
. .
=Ci _J ( / L: E;.
+ 1)- 1 i.=l so
f+I
P(R) = Q(H) + Z(c1 - (I+ l)"- 1 L c;).
1=l
CLA.'>SIFICATJO!'I OF ROOT SYSTE'\IS 2HJ
~owl+ l is the s1t1allcst intq.!;t'r 111 > () s11ch thi\t 111p(E; 1) E Q(R). Thus
P(f0/Q(B) is isornorpl1iC' to Z/(/ + l)Z ;111d the c01111cctio11 index is I+ 1.
(IX) n11d (X) For anv a11trnnorphis111 .<J of V. kt ;:(.11) he the a11tot11orphis111
of E that extends y and kavC's :c 1 + :C·2 + · · · + c/ invariant. If y is\ lw orthogonal
rdieetion 8 0 , - ~,IV. ;p(.q) is equal to s,, -c,. which i11tcrd1<1llgcs c:, a11d z 1 and
l('il\'PS fixed the E; with k distinct frorn i and j. Let
X -- {c~ I · ~c·2 • • • • • c-I ·t· l }·
TlH'll y f--> i.p(.<J)IX is an isomorphis1n from \\'(Tl.) to th<' symmetric group of

X. Th11s. \V ( n) is isomorphi(' to I h<' s~·m lll<'t ric ).!,l'Ollp 61 +I. and so is of ord('l'
(l + l )1.
The sn11rn<'tric algebra S(E) can lw identifi<'d ca11011icallv with th<' nlgc-
bra of pol.rno111ial functions P(~ 1 .~z ..... ~/ti) rn1 E. LC't G = ;:(\\"(R)). By
t hl' pn'ccdiug paragraph. the set S(E)c of cl<'llll'llts of S(E) i11varia11t \111-
d<·r C is the s('\. of s.1·111mdric pol~·n0111ials (Al_qcbro. Chap. V. App. I). (ltld
co11scqll<'1ttly (il1irl.) is the algchrn gencrnted by th<' f1111cti011s
(l ~ i ~I+ 1).
The alg<'hrn S(V) can he iclc-111 ificd with th<' restrict io11s to V oft lw poly-
nomial functions OJI E. If P E S(E)c. tl1t' restriction of P to V is clearly
i11vnria11t. ltll(\rr \\'(11). Coll\'('J'S('ly. if Q E S(V)\\'(HI. th<•rc ('Xists r E S(E)
<'Xl<'11di11g Q: rqibl('illg [' h.v ((/ + 1) 1)- 1 L y( P). which has t IH' sarn<' rest ric-
:1 E: (;
tio11 as P to V. W<' rn11 assntw' that PE S(E)c. Th11s. S(V)\\'(H) is g<'nnat<'d
by the s; =-<IV. Now 8 1 = 0. ~lorcov<'L the tra11sccnd<'11c" dqi,r~c over R of
th(' fidd of fractious of S(V) W( RJ isl, so the s; (2 ~ i. ~ I+ I) arc algebraically
i11drpt•11de11t. Since the s, HJ'(' of <\egret's 2. :~ ..... I+ I. 1llP <'XflOlWllts of vV(R)
arc
1.2.:l.. ... L.
(XI) For I= 1. A(R) == \V(R) = Z/2Z and 11·0 = - l.
For l? 2, let c: E A(R) I)(' t.h<' a11tolllorphistt1 that trn11sforms o, to 01+ l-i·
It is clear that t hr au tomorph isrn ind11ced by c is th<' 1wiqlH' no11-tri vial
nut.ornorphism of the Dy11ki11 grnph. Thc group A(H)/W(Il) is isomorphic
to Z/2Z. Since -1 is a11 clcm<'llt of A(R) which do<'s 11ot belong to \V(R)
by (IX) <111(! (X), we· s<'l' that A(R) is isomorphic to \\'(H) x Z/2Z. W<' have
U'o =- -::-.
(XII) ThC' gronp P(W)/Q(Ir) is cyclic of ord<'r l + 1 and nets 011 the
co111pld<'d Dvnkin graph b>· circular pcrn1ntntio11s. If I ? 2. tlw n11iqtt<' no11-
idc11tits t>IC'mc11t of A(R)/W(H) <trts on P(W)/Q(W) bv the automorphism
.r ~~ -:r.
220 ROOT SYSTFl\IS Ch. VI
8. SYSTEMS OF TYPE Dt (l ;? 3)
(I) Consi<kr in V = R 1 tll<' group Lo (no. 4). The set. R of n E Lo such

that (nln) = 2 consists of the wctors ±:.-; ± EJ (1 ~ i < j ~ l). lt is ckar
that R generates V and that 2(nl:1)/(n[n) E Z for nil o, 11 E R. Tints R is a
r<'duC'ed root syst<'lll in V (110. 4). The n11111lwr of roots is n = 21(1 - I).
(II) Put
The following formulas are immediate:
E; - Ej = O; + 01+1 + · · · + Oj-1 (i < j)

E; + Cj = O; + 01+1 + ... + Oj-1 + 2n, + 2ny+l + ...
. . . + 201-2 + rl'/-1 + 0:/ (i < .i ~ l - 2)
c;+E1-1 =0;+<>1.+1 +···+n1 (i</-1)
c;+c1=n;+o;+1+···+(%/-2+n1 (i<l-1)
E/-1 +Et = O/,
so (n 1 ...•• o 1 ) is a basis if R (~ 1. Ho. 2. C'or. 3 of Prop. 20). Fmthcr,

II o; 1[ 2 = 2 for all i. (n,l1t1 ) = 0 for i + 1 < j <'xcqJt for i = I - 2.j = I
i11 which rase (01-2101) = -L (o,[n;+ 1 ) = -1 for i ~I - 2. and finally
(n 1 _ 1 [ni) = -1: the Dynkin graph of R is thus of type 0 1 . The positivc roots
me the E; ± E1 for i < j.
(III) We have h = 11/l = 2(l - I).
(IV) LPt it = c 1 + c 2 = n: 1 + 2n 2 + · · · + 2n1_ 2 + n1- 1 + n1. which is a
root. The sum of its coordirtatt's relative to (n,) is
21 - :1 = h - 1.
Hence ii is the highest root.

If l = :1, we lmvc
If l;? 4. we have (ii-[ni) = 0 fo.r i fc 2 and (ii[n·2 ) = 1. HPnC'c the rnmpktc<l

Dynkin graph is:
(V) Since (o[n) = 2 for all o Ef R, we haw R- = R.

The kngth of tlw roots for <Pr is h- 11'2 = (2/ - 2)- 1 12 . Hcnce
<Pn(:r.y) = (.rl.11)/(4/ -1) and 1(R) = 1(1- 1) 2 .

s 4. CLASSIFICATION OF ROOT SYSTE!\IS 221
(VI) A cak11latio11 aualogo11s to that. in no. 7 gives tlw fillldamental

WC'ights:
"-''i =Et+ Ez + · · · + E;
= <.~1+2n2 + · · · + (i - l)n;_ 1 + i(o; + n;+1 + · ·· +ni- 2)
l
+ 2i(n1-1 +n:t)
for i < l - 1,
1
wt- I = -(c1 +E2+···+Et-2+E1-1 -E1)
2
l 1 1
= -(n1 + 2n2 + · · · + (l - 2)n1-2 + -lrr1- 1 + -(l - 2)01 ),
2 2 2
1
W/ = 2(c1 + E2 + · · · + Et-2 + E/-1 +Et)
l ' 1 1
= 2(n1 + 2n2 + · · · + (l - 2)n1-2 + 2(1- 2)n1-1 + 2tn1).
(VII) The sum of the positive roots is
2p = 2({ - 1)€1 + 2(/ - 2)c2 + · · · + 2Et-I

1~ i (i + 1) l (l - 1)
.~
l~-1
2(il - ,
2
)o; + ---(n1-1
2
+ rt1).
(VIII) The ±c:; ±c:.i generate L 1 (no. 4), so Q(n) = L 1 . Hence Q(K) = L 1
and rnnscqucntly P(R) = L2 (no. 4). l3y no. 4, P(fl)/Q(R) is isomorphic to
Z/4Z for l odd, and to Z/2Z x Z/2Z for l C'ven. In the first. case, P(fl)/Q(R)
is gcrwrat.cd by the canonical image of w 1 (and also by that of w1_ 1 ). I11 the
second case, P( R) /Q( R) is genC'rat.ed by the canouical images of w1 _ 1 and Wt.
In both cases the connection index is 4.
(IX) ctll(I (X) In R 1, the orthogonal reflection s,, -E, (i I j) interchanges
E, and C:J and leaves invariant the Ek with k distinct from i and j. The Sc:;-t:i
generate a group G 1 isomorphic to thl' syrnmetric group 6 1• On thC' other
hand, S;J = 8 0 ,-c:.,s<:,+£, transforms E; to -E;, c.i to -Ej a11d leaves invariant
the Ek with k distinct from i and j. The s;.i generate a group G 2 , the set
of antomorphisms 11, of the vector space R 1 such that u(E;) = (-l)v'E; with
l
CT (-1)"• = 1. The group G 2 i:; isomorphic to (Z/2Z) 1- 1• and G 2 is normal in
i=l
W(R), so W(R) is isomorphic to the semi-direct product of 6 1 by (Z/2Z) 1- 1•
Cow;equcntly. its order is 21- 1 .l!.
The polynomial functions f; of no. 5 an~ invariant under W(R). and so is
t = ~16 ... (1; moreover. f1 = t 2 . Let P((1 , .... (1) be a polynomial invariant
under W(R). Let ~~, ~~ 2 • , . ~r' he a monomial featuring in P su("h that 11; is
odd; then v; is odd for all j ht-cause the monomial ( -1 )''• +v, (~' (~2 . • . (;'1
features in 8;_j(P), so 11, +VJ ==:: 0 (mod. 2) and 111 = 1 (mod. 2). Thus
P =Pt +tP 2 • where all t.lw monomials featuring in Pt a11d P 2 haw only even
<'xpmwnts. Sinn' P is invariant 1111dcr th<' pernt11tatious of t.hc ~;. Pt awl P2
have t.he sat1H' property. a))(I so can he written as polynomials in t 1 . f'2, ... , f1.
This prows that. t 111• algPhra S(R1 ) 'N( n) is generated hy I 1 • t 2 , ... , f1 _ 1. t.
Moreover. the tnmsce11d<'ncc degre<' of the field of fractions of S(R1)\V(H)
is/. so t. 1 , t 2 .••.. 11_ 1 • t arc a.lgd1rnically indep<'ndt-nt. \V<' condudc that the
scq11c11cc of exponents. suitably ordered, is:
1. :i, 5, .... 2/ - 5. 21 - :u - 1.
Note that l - 1 appearn twice if I is even, and once if I is odd.

(XI) The a11tornorphis11ts of the Dy11ki11 grnph ar<' those of tlw 11nckrlyi11g
graph. Thus:
1) If l = :J, A(H)/W(R) is isomorphic to Z/2Z.
2) If l = 4, every permutation of the terminal wrticcs <kfincs an auto-
morphism of th<' graph, so A(R)/W(R) is isomorphic to 6:1.
:l) If l ? 5. the chains starting at the ramification point have length L 1,
and l - :i ? 2. The onl:v automorphism of the graph distinct from the idmtity
thus corresponds to the automorphism EE A(R) which interdmnges 01_ 1 and
n 1 and leaves fixed then; for 1 ~ i ~ l - 2. Thus A(R)/W(R) is isomorphic
to Z/2Z: moreovcL A(R) is the semi-direct product of the group G 1 ~ 61
<kfincd in (IX) by the gro11p G:i ('()Jlsist.iug of the automorphisms u of R 1
such that u.(c;) = ±t:; for all i.
If l is even, -1 E W(R), so 11· 0 = -1. If I is odd, -1 ¢ W(R), so A(R) =
W(R) x {1.-1} am! w0 = -E:.
(XII) For I t'VPn, P(W) /Q(W) has three clements of order 2, namely
w1, Wt-l and w1. Since w1 (resp. w 1_ 1) interchanii;es the vertices corresponding
to no and n1 (resp. n1_ 1), it i11tcrchangcs t.l1osc corresponding to ll"i and n,_ 1
(resp. cri) and also those corresponding to n 1 and n 1_J for
2 ~ j ~ l - 2.
\Ve have W1 = WtWL- l •

For l odd. P(Ir)/Q(Ir) has two elements of order 4. namely W1- t and Wt,
and one dcment of order 2, equal to w 1 • Indeed, w 1 interchanges the vertices
corresponding t.o no and o 1 , so it leaves fixed the vertices corresponding to
rt.; for 2 ~ j ~ l - 2 and is neccsstrily of ord<•r 2. Consequently Wt is of order 4
and transforms the vertex corres )()nding to o 0 (resp. o 1, resp. o 1 , resp. n 1_ i)
to that. correspo11cli11g to n1 (rcsr. n 1 . resp. o 1_ 1 , resp. n 0 ). and intcn:hanges
the vertices corresponding to n.i 1ml O/-J for 2 ~ j ~ I - 2. \Ve have w 1 = w[
a.ud W/- l = ;,vjl.
For I f= 4. the non-identity element of A(R)/W(R) interchanges the
vertices correspondi1111; to n 1_ 1 and o 1 • and <·011sc<111cntly interdiangcs the
elements w1_ 1 and Wt of P(W) 'Q(W). For l odd, the automorphism of
P(W)/Q(W) th11s obtained is the map .i:,....... -.r:.
3 4. CLASSIFICATION OF' ROOT SYSTE'.\IS 223
For I= ·1. A(H)/\V(R) ra11 lw identilic·d with the 12;ro11p of pt'rm11tatio11s

of { l. :l. 4} aud acts Ii.-· pcr11111tatio11s of the inclit'<'S on {w 1 • wa. (A..'-i }.
9. SYSTEM OF TYPE F 4
(I) Consider the gro11p L2 (no. 4) in R'1. Let R he the sl't of rl' E L2 s11ch
that (rl'ln) = 1 or (o/n) = 2; it C'Olltains the vectors
±E;, ±E; ± Ej (i < j), ~(±E1 ± E2 ± E3 ± E.1) 0
Conversdy. if n ER, the coordinates of n can ouly take the values 0. ±~. ±1
(since ( ~ )2 > 2); either these coordinates arc all intpgcrs, giving the vectors
±E;. ±c:; ±Er or they are all eq11al to±~. giving the vectors ~(±c 1 ± E2 ±
E;3±c4).
We show that. for n. /3 E R, we have 2(nli:J)/(olo) E Z. If n = ±E; or
n = ~(±€1 ± E2 ± E:i ± c:.1 ), then (nln) = 1 and we' have scc11 in 110. 4 that.
((tj/J) E ~z sinC'e n. (-J E L2 . If o: = ±Ei ±E.1· then (nln) = 2 awl wr have seen
in 110. 4 that (nl;J) E Z since n E L 1 and /:J E L2. Henre. Risa reduced root.
system in R 4 (no. 4). The number of roots is n = 8 -t- (~) 4 + 24 = 48.
(II) Give R 1 the lexicographir orckr defined by the httsis (c 1 ,t: 2 ,c:i,c 1 )
(S 1. no. 7). In particular, we have s 1 > c: 2 >Ea > s 1 . The positive roots arc
E;, E; ± Ej (i < j),

Tlw smallest root is n:i = E 4 . Among the roots belonging to Rc::1 + RE.1 but
not to Rc 4 • the smallest i:; (t 2 = E:i - E4 . Among the positive roots belonging
to Rc: 2 + Rc::3 + RE.1 but. not to Rc::1 + Rc:.1 • the smallest is n 1 = E2 - E3.
Among the positive roots uot. belonging to Rc2 + Rc3 + RE,i, the smallest is
o 4 = ~ (c 1 - E 2 - c;i - c 4 ). None of then; is a sum of 2 positive roots. Henre,
(n 1 .n 2 .n:1,o.i) is a basis of R (li L no. 6, Cor. I to Prop. 19). We have
II o, 11 2=11 02 11 2= 2, II n:i 11 2=11 o-i 11 2= 1, (01 led = (n2loa) = -1, (n:1ln1) =
-1, (n1103) = (n1 ln 1 ) = (02104) = 0. We scP that the Dynkin graph of R is
of type F.1 , and hence is irreducible.
(III) We have h = y = 12.
(IV) Let <°l· = E 1 + E 2 = 2o 1 + 302 + 40·:1 + 2n". The sum of the roordinates
of 0- with respect to (o:;) is 11 = h - 1. so (i. is the highest root. We have
(iilni) = 1. (iiln2) = (<tln:1) = (iiln.1) = 0.
The completed Dy11kin graph is
(V) Tiu' formula Cl- = (nn

2' ')
gives
1
· for ll" the set of vectors ±2E;.· ±E; ± EJ·,
±E 1 ± E 2 ± E:i ± E 1 . The Dynkin graph of W is ohtai1wd from that of R by
the procedmc explained in 110. 2. and we see that Jr is of type F 4 .
224 HOOT SYSTEl\IS Ch. VI
The roots not orthogonal to /i =

c: 1 are ±c: 1 , ±c: 1 ± c:1 (j ~ 2). and
~(±c: 1 ±c: 2 ±E:i ±c:-1): the nurnbcr n(oj/1) = 2(nj/i) is equal to ±2 for the
first 14 of these roots and t.o ± 1 for tlH' last 16: thus, for <PR, I he squar(' of
tlw lc11gth of fi is 4( l.:l.4 + 16.1)- 1 = tx;
hence
(:rJy)
<h(;r:,y) = 18"
We now apply formula(!~) of s l, no. 12, with x = y = /1: this gives
1 1/4 1
2 + 12 . -4 + 16. - 1 = "-(R)
' . · -18
so
(VI) Calculating the fundamental weights gives
W1 = c1 + c:2= 2n1 + :ln2 + 4n:i + 2n.1 =ii

w2 = 2c:1 + c:2 + c:3 = :kt1 + 6a2 + 8a3 + 4a4
1
W3 = 2(3c:1 + C:2 + c::i + c4) = 201 + 4n2 + 6a3 + 30!4
W4 = CJ = OJ + 20'2 + 3n:i + 2n.1.
(VII) The sum of the positiw roots is
(VIII) We have Q(R) = L 2 (no. 4), and P(R) = Q(R) by (VI). Hence, the
connection index is 1.
(IX) The family of exponents has 4 terms, and since h = 12, it must con-
tain the integers 1, .5, 7, 11, coprirne to 12 (s 1, no. 11, Prop. :JO); consequently,
these arc all the exponents of W(R).
(X) and (XI) The only automorphism of the Dynkin graph is the identity,
so A(R) = W(ll) and w 0 = -1. Let R' be the set of longest dcrnents of
R, that is, the ±c:; ± c:1 : R' is the root system of type D 4 constructed in
no. 8. Cl<'arly. every derrl<'nt of A(R) is an element of A(R'). Conversely, an
clement of A(R') leaves L 1 stable (since it is generated by R'), hence also its
associated L 2 , and hence also R. StW(R) = A(R) = A(R'). By no. 8, W(R)
is the semi-direct product. of 6:i by (R'), W(R') itself. being the semi-direct
product of 6 4 hy (Z/2Z)·l. The or er of W(R) is 3!4!2·l = 27 .3 2 .
§ t CLASSIFICATION OF ROOT SYSTEMS 225
10. SYSTEM OF TYPE E 8
(I) Co11sid<>r th0 group L: 1 (no. 4) in R 8. LPt R h<' the set of n E L:i such
that (nln) = 2: it cont.ai11s the vectors
8
(i=l
L l/(i) <'V<'n).
Couvcrsdy, if an element o E L; 3 is such that (0:!0:) = 2. its coordinates

must be among the valu<'s 0. ± ~. ± 1: by 110. 4. either these coordinates are
all iut.egers, giving the vectors ±E; ± EJ· or they arc all equal to±~ with an
i11tPger surn, givi11g the vef'tors
ll
witl1 L v(i)
i-=cl
evm.
We have seen (no. 'l) that (od/3) E Z for all n, /3 E L;1. Hence, R is a
reduced root syst('m. The 11u111lwr of roots is n = ( ~) .4 + 27 = 240.
(II) Let p be the vector (0,1,2,3,4.5.(i,23) of L;3. No clement of R. is
8
orthogoual top (this is clear for thP ±E; ±c:.i; if ~ .L (-1 )'"'(i) E; were ortho~<mal
1=!
Ii
top. Wt' would have Li.(- l)''(i+I) + 23(-1)''( 8 ) = 0, which is impossihle
i=I
6
sillC(' L i<
i=l
23). Hence (§ I, 110. 7, Cor. 2 of Prop. 20) the n E R such that
(nip) > 0 arc the positiw roots relative to a certain chamher. These roots
ar<> the ±E, + E.1 (i < j), and th('
7
witli L ll(i) eve11.
i=l
We hnve (ulp) E Z for all o ER (no. 4). and (nip) is equal
to l for the following roots:
1 l
01 = 2(t:1 + E8) - 2(cz + E;i + E4 + €5 + E(i + €7 ),
(.}~=CJ+ €2, <.l!.1 = c2 - CJ, 04 = E:i - E2, O's= E'4 - €3,
0() =Es - €4. <Y.7 = E!i - £;,. O'/l = E1 - c(;.
and these eight vectors form n basis of R 8 . By ~ 1. no. 6. Cor. I of Prop. 19,
(n1.<t2, .... 08) is a basis of R for which the positive roots arc those which
have het~ll defin<:'d above. \Ve have
226 ROOT SYSTE~IS Ch. VI
a11d (0 1 1nJ) =()for all other pairs of indices. H0ncc. th(' Dynkin graph of R
is of type E 8 . a11d H is irn~dncible.
(III) We have h = ~ = :30.
(IV) L!'t
<~ = E7 +Es = 201 + :1n2 + 4o:i + 6<1'1 + 5<1';, + 4<1'G + 3<1'7 + 2<1's,
which is a root. TIH' sum of its coordinates with n•speet to (n;) is 29 = h- 1. so
it is the highest root. It is orthogonal to all tlw n; except rt1<. and (<tlrvH) = 1.
Hence. the compkt<•d Dynkitt graph is:
o------o---o--0----0--0-----o-o
<Xi <X3 lex, ex, <Xs <X7 <Xa
(V) Since (nlo) = 2 for all(\' E n, WC hnv(' R' = R.

For <l>n, the squared length of tlw roots is L 110. 12). H0nre, fa rn
<l>n.(.r,y) = (xl.11)/60 and 1(R) = 900(~1. no. 12, fornmln (20)).
(VI) Calc11lating the fu11darn('11tal w0ights gives
W1 = 2E8 = 401 + 502 + 7n:; + lOrL1 + 8n5 + 6a5 + 4n7 + 2<1'8
l
W2 = 2(c1 + E2 + E:[ + c4 + Cfi + E(J + E7 + 5EH)
= 5n 1 + 8n2 + 10n:1 + !5n 4 + l2ns + 9n 6 + 6<1' 7 + 3ns
I
W3 = 2(-c:1 t E2 + E3 + E4 t c;, t f(i + E7 + 7c:s)
= 7n 1 + 1002 + 14n:1 + 20n 4 + Hin;;+ 12n5 + 8n7 + 4ns
W4 = c:J + E4 + E5 + E0 + E7 + f>E8
= 10n 1 + 1502 + 20n:i + 30n4 + 24<1';, + l8a 0 + 12n7 + 608
= 80 1 + 1202 + 1603 + 24rL1 + 200,, + l 5n 0 + l0n7 + 5a8

We; = E;, +Er; + E7 + :k8
= Go 1 + 902 + 12n:i + 1804 + 15ns + 1206 + 807 + 4ns
W7 = f 0 + E7 + 2E8
= 4n1 + 6n2 + 8n:1 + 204 + 1005 + 8aG + Gn7 + 3ns
W3 = E7 + E8
= 501 + 802 + lOn:i + 1504 + 12<t;; + 9nr, + 6n7 + 3ns =ii.
(VII) Half the sul!l of thP (l>ositive roots is the sum of the fu11damcntal
weights(§ 1. 110. 10. Prop. 29): I.his gives
p = E2 + 2E:i + 3c4 + 4E;, + 5c0 + 6c:7 + 23c-s

= 46o 1 +6802 + 9lo 3 + 135<1' 4 + lJOns + 84rrr, + 57ct7 + 2903.
\j 4. CL.'\SSIFICATION OF ROOT SYSTEl\JS 2'27
8
(\'III) Tlw gro11p Q(H.) is g1·11<•rat.ed by th<' E:; ± E.1 and ~ LE;, and is
1.=l
equal to L:i (110. ·l). Hc111:1~ P(R). which is associated to Q(K) =· Q(R) = L: 1.
is L: 1 (110. -1). The co11111•ction index is I.
(IX) Th<' family of exponents has 8 terms. and since It = 30, thC' i11tcµprs
I. 7. 11. I :t 17. I !J. 2:3. 29, copri11w to :lO. must feat me i11 this fa111ily; co11se-
q11c11tl_v. these arc th!' c·xporwnts of \V(R).
(X) From (IX) aud Chap. V. ~ 6. 110. 2. Cor. 1 of Prop. 3. it follows that
the ord1•r of W ( H) is
(Xl) The only antouwrphism of tlw Dynkin graph is the ide11titv sinr<' the
thr<'l' chains issuing from the ramiticatio11 point hav<' distinct: lengths. Hence,
A(H) = W(H) HIHI W11 = -1.
(I) and (rI) Let E = R 8 . and let H8 be the root system in E ronstrncted
in 110. I 0. Let V be t hi' h~·p1'rpla111• in E gem' rated by the roots n 1 •...• n 7 of
R 8 ; it is orthogounl to the eighth fu11de111JP11tal weight.,;= E7 + 2 8 of R 8 .
Let R = R8 n V. Then R is a red11cr.•d root system with basis ( n 1, .... n 7 ),
('f. ~ 1. 110. 7, Cor. 4 of Prop. 20: hc1icc, this systc•111 is of t_vpP E7. Its el1'nwnts
arc:
±(€7 - €8),
8
with L v( i)
i=I
odd.
The 1111rnber of roots is n = 2 + ( ~) .4 + 2° = 126. The positive roots arc
±€; + Cj (1 ::::; i < j::::; G). -E:7 + Eij,

6 (i
~(-€7 + f8+ i~ (-l)''(i)E;) with ;~ v(i) o<ld.
(IIT) We have h = y = 18.

(IV) Ll't ii = E>i - E7 = 201 + 202 + 303 + 4n.i + 311';, + 2<l'5 + r.:7, which
is a root. The sum of its roordi11ates with resp<'Ct to (o;) is 17 = h - 1.
It is t Iwrefore the' hight•st root. It. is orthogonal to <l; for 2 ::::; i ::::; 7. and
(iiln 1 ) = 1. The completed Dynkin graph is
«1 0!3 r·
O----o---0--0---0----0-0
0
«1 oc, oc,
oc,
228 ROOT SYSTEJ\IS ('h. VI
(V) Since (nlo) = 2 for all n E R. we have Ir= R.

For <f>n. the squared length of t lw roots is ~. so
<Pn(.r. y) = (:rty)/:lH. nnd 1(R) = 22 .3 1

(§ 1. 110. 12, formula (20)).
(VI) Cnlculati11g the f11nda11w11tal weights gives
w1 = E8 - E7 = 2(~1 + 2n2 + :ln3 + 40'-t + 3(\'.5 + 20'r, + C\'.7

1
V.:2 = 2k1 + E2 + E3 + E4 + €5 + C(; - 2c7 + 2c11)
l
= 2(401 + 70:2 + 80:1 + 12n-i + 90:;; + 8n<; + :ln 7 )
1
w:i = 2(-E1 + E2 + £:1 + E4 + E;, + E6 - 3E7 + 3Es)
= :ln 1 + 4n2 + 60-:i + 801 + 605 + 4nG + 20-7
W4 = £3 + €4 + £5 + E(; + 2(E8 - E7)
= 40:1 + 602 + 8n:i + l 2(L1 + 90-;; + onfi + 30-7
3
;,;,; = E4 + €5 + c(i + 2(€8 - E7)
1
= 2(un1 + 9n2 + 12n3 + 180:4 + 150-5 + lOO'G + 507)
W<> = E;, + Ee; - €7 + €8
= 2n 1 + :30·2 + 40:3 + Go 1 + 50';; + 4o-r. + 20'7
l
v.,•7 = E6 + 2(€8 - €7)
l
= 2(2n1 + 3n2 + 40'.:3 + 6n.1 + 5n;, + 4n6 + :3n 7 ).
7
(VII) The sum 2p of the positive roots is 2 L w; (~ 1. 110. 10, Prop. 29),
i=I
so
2p = 2t:2 + 4t::1 + Gc4 + 8€5 + lOE6 - l 7E7 + l 7E8
= :34o 1 + 49n 2 + 66rx:i + !J6o 4 + 75nri + 52rx 6 + 27n7.
(VIII) By 110. 10 (Vlll) and § L no. 10. Prop. 28. we have
Q(R) = Q(R8 ) nV = ~: 3 nV and P(R) = p(P(R 8 )) = p(L:i),
where p dcnot<•s I.he orthogon~d projection of E onto V. The group Q(R) has
basis (o 1••••• n 7 ): tlw group P(H) is )!;C1terated by Q(H) and
§ -1. CLASSIFICATIO:'-J OF ROOT SYSTEl\IS 229
We have w E P(RK). ~w €f_ P(RH). so 2p(nH) E Q(R) and p(o8 ) €f_ Q(H).
Tlrns, P( R) /Q( B) is iso1110rphic to Z/2Z and is gcncrat.cd by, for example.
tlic image of "-'7·
ThP co11m•ctio11 i11<kx is 2.
(IX) The scq11Pncc of expoiwuts of \.V ( R) has 7 t<'nns. The 1111mlwrs
I. 5. 7.11, I:~. 17. rnpri111e to h = 18. f(>at11n• in 1his scq11e11n'. Tlw la.-.;t Px-
po11c11t. m must be s11clt tha.t rn + m = 18 (Chap. V. § fi_ 110_ 2. forrnula (2)).
Hence, the SPfJIH.'llCC of cxponPnts is
l. 5. 7. 9, 11.13.17_
(X) From (IX) n11d Chap. V. § 6. 110. 2, Cor. l of Prop. :3. it follows that
thP order of W(R) is
(XT) Tlw only automorphism of the Dynkin graph is the identity. so

A(H.) = W(R) and w 0 = -1.
(XII) P(W)/Q(W) has only c>11c non-identity element. It interchanges the
vcrticc•s corrcspo11di11g to no and n 7. n 1 aml 110. <1:1 and 11'..i. and kaves f\' 2
and n. 1 fixed.
(l) and (II) L<>t E = R 8 , and It'! R8 IH' the root system in E constructed in
no. 10. Let. V he the vector subspace of E gl'twrnted by the roots u 1 , •••• f\' 6
of R1:1: this is the orthogonal compl('!llcnt of the plam• gencrnt.c•d by the last
two fimclamcntal weights ;;J = E 7 + c: 8 and rr = c- 6 + E 7 + 2c- 8 of R 8 .
Let R = H.8 n V. This is a reduced root system with basis (n 1 ••..• n 0 ),
and hence of type Eh. I ts elements are:
±E; ± Ej (1 ~ i<j ~fl),
l 5 - 5
±-(EH - E7 - E5+
2
L (- l)"(')E;)
;~1
with L
i=l
l/(i) <'V<'ll.
The lllHllber of roots is 11 = (~) .4 + 25 = 72. The positive roots arc
±€; + Ej (1 ~ i. < j ~ 5).
~(Es
2
- E7 - E<>+ t
t=I
(-l)"(i)E;) with t
i=l
v('i) even.
(lll) \Vp have• h = ~ = 12.

(IV) Let
- l
II= 2(€1 + €2 + E:! + E.J + E5 - c0 - €7 + cH)
= n 1 + 2n2 + 2n:i + :~n 1 + 2o 5 + n 6 ,
2:30 llOOT SYSTEl\IS Ch. VI
which is a root. The suw of its coorcli11ates with n•s1wct to (n;) is 11 = h - 1,

so ii is tlw highest root. It is orthogonal ton 1. 0;1. n. 1. o:-,. or;. awl (i•jn;!) = l.
Tlw co111pl<'lcd Dynki11 graph is
otz
(V) Since (njn) = 2 for all n E Il, we haven:= R.

For <l>n. the squared lcugth of the roots is so fl.
<Pn(:r. y) = (;r.jy)/24. and 1(R) = 14--1.
(VI) Calculating the fundamental weights gives:
2 1
W1 = 3(Es - E1 - En)= 3(4n1 + 3n2 + .'5n:i + 60 4 + 4n5 + 2ot>)
1
W2 = 2(€1 + E2 + E;1 + E.1 + E5 - c(; - E7 +Es)
= 01 + 20:2 + 2o:i + 30.1 + 2n5 + <Yn :::: <\:

5 1
W:i= 5(Ex-E1-E<i)+2(-E1 +E2+E:1+E1+E:;)
1
= 3(5n1 + Gn2 + l0n3 + 1204 + 805 + 4n6)
W.1 = E;i + E4 + €5 - E6 - + E1:1
E7
= 201 + :~02 + 4n;i + 6n 4 + 4or; + 2n"

2
W5 = 3(E8 - E7 - cfi) + €4 + E5
1 .
= 3(4n1 + 60:2 + 80:1+12n4 + lOas + 506)
1
W6 = 3(cs - €7 -E(i) +E;,
1 '
=3 (2rv1 + 30'.2 + 4n:1 + 6n.1 + 50·5 + 4n0).
6
(VII) Half the sum of the positive roots is 2: w;. so
i.=l
p = <:2 + 2£a + :~c:.1 + 41:5 + 4(Es - <:1 - <:1;)

= 801 + 1102 + l5n:i + 2ln.i + 150'.;; + 8n6.
(VIII) By 110. 10 (VIII) a11<l ~ ~, no. I 0, Prop. 28,
Q(R) = Q(R8) n V = L:i n V and P(R) = 71(P(l1x)) = p(L:1),

CLASSIFICATION OF ROOT SYSTE'.\IS 231
where' p denotes the orthogonal projection of E onto V. \V(• haYe
The group Q(R) has basis (o 1 ••••• n 1;). Tht> gro11p P(R) is generated by
Q(R) and p(n7 ). sine<' p(n1J) E P(R8 ) n V = Q(R8 ) n V = Q(R). We have
:~p(n1) E Q(11) and p(o7) ti- Q(R). Hc11ee. the group P(Il)/Q(R) is isomorphic
to Z/3Z: it is generated. for cxarnpl(', h:v t.hc image of we;.
The connection index is 3.
(IX) and (X) Ily g 2. no. 4. Prop. 7. the order of the Wey) group is
6!1.2.2.3.2.1.3 = 2 7 .3 1 .5. The seq11cnn~ of Pxponcnts has 6 terms bl't.wccn 1
awl 11. and contains t.hc integers l. 5. 7. 11 which are copriruc to 12. The
othPr exponents m.. m' are integers such that
m + m' = 12.
(m + l)(m' + 1)(1+1)(5 + 1)(7 + 1)(11+1) = 2 7 .34 .5,
s
in vi<•w of Chap. V. fi, no. 2. formula (2) and Cor. l of Prop. 3. The second
rdatio11 gives (m + l)(ni.' + 1) = 4f>. and since m + m' + 2 = 14. we obt.ain
111 = :I, m' = 8. Hence. I.he sequcm·e of cxpo11cnts is
1. ·'1. 5. 7. 8. l l.
(XI) a11d (XII) Si11('c t!te rnots all have th<' same length. the a11tornor-
phis111s of tlw Dynkin graph are thos<' of th<' 1111dcrlying graph. Apart from the
ideutity. there is only th<• automorphism c which transforms o I· n: 1 • n 4 • n;,. no,
n 2 i11to nc;. o 5 • n 4 , n: 1, o 1 • n 2 , rc•spectivcly. Ileuce, A(R)/vV(R) is isomorphic
to Z/2Z: since - l E W(R) (Chap. V. § 6. no. 2, Cor. 3 of Prop. 3), A(R)
is bomorphic to W(R) x {l, -1} and w0 can he idcutitiecl with -E. It fol-
lows that the non-i<lentit.y denwut. of A(R)/W(R) defines the automorphism
;r >---> -.t of P(R')/Q(W).
'.\Iorcovcr, P(W)/Q(W) has two 11011-ident.ity denwnts of order 3. They
defim· the two automorphisms of order :J oft.he con1pletcd Dynkin graph.
13. SYSTEM OF TYPE G 2
(I) L!'t E be the hyperplane in R:1 with cq11ation
~I+ 6 + 6 = 0.
Let R he the set of o E L 0 nV s11ch t.hat (nln) = 2 or (nlo) = 6. The elements
of Rare
±(€1 - E2). ± (c1 - E;i). ±(E2 - E:i). ±(2c1 - E2 - E:i),

± (2E2 - E1 - E;1), ±(2€3 - E1 - c2).
ROOT SYSTEl\IS Ch. VI
Th<'ll, H generates V, and 2l~~'./i) E Z for all o. {J E R: this is clear if (J =

±(.:; - c:.i) with if. j: if rJ = 2: 1 - ::: 2 - E;i for c•xamplc, W<' ha\'(' (nl/I) E :3Z,
a11d again our asscrtiou holds. He11cC'. H. is a reduct'd root syst1'm in V. The
1111mlwr of roots is 11 = 12.
(II) Put n 1 = E1 - E2. n2 = -2c- 1 + E2 + E;3. Then the roots are
±n:1, ±(01 + n2). ± (2n1 + 02), ±n2,

±(:301 + n·2). ± (:3n1 + 202).
Hence, (u1. n2) is a basis of R. We have II <J"1 11 2= 2, II n2 11 2= 6. (n1 ln-2) = -:3,

so R is a system of type G2. Th<' positive roots are n 1 . o 1 + n2. 201 + n2,
3n1 + n2, :3n1 + 20:2.
(III) We have h = ~ = u.
(IV) The highest root is ii = :10 1 + 2n 2 = -E1 - c2 + 2f3. We have
(ii.ln1) = 0, (iiln2) = :3. The cumplctt>d Dynkin graph is
(V) The invcrs<' system is the following set of vectors:

1
±n1. ±(01 + n2). ± (2n1 + n2), ±302,
I 1
±3(:ln1 +n2). ± 3(:101 +2n2).
Ther(' are 10 roots not orthogonal t.o <r 1 : W<' have n(/1. o i) = ± 1 for 4 oft hcsc
roots, n(i3,n 1 ) = ±3 for 4 others. and n(IJ,n 1 ) = ±2 for ,(J = ±n 1 . Hence, the
squared length of n 1 for <l>n is 4(4.l + 4.9 + 2.4)- 1 = f.J.
Hence, <Pn(x,y) =
(:rly) /24. We now apply formula ( 18) of ~ 1, no. 12. with .r = y = 1'.1' 1 ; this
givt's
l l l
2 + 4.- + 4.- = 1(R).-
4 ·1 12
so 1(R) = 48.
(VI) and (VII) Half the snm of the positive roots is
p =\501 + 31x2.
The fuudarncntal weights w1 a11<J w 2 are orthogonal to n 2 and n 1 , h<'nce
proportional to 2n 1 + n 2 and :~o 1 !- 2o 2 . V\'e have
Hc·•nce,
s 4. CLASSIFICATION OF HOOT SYSTE!\IS
(VIII) Q(R) is gcncratt>d by E 1 - E2 and E 1 - E:.i, for Pxarnple. ny (VI) and

(VIJ), P(R) = Q(R). Thi' comwrtion i11dcx is 1.
(IX) Tiu' family of expo11P11ts has 2 tPrms: since 1 and h - 1 = ,') are
exponents. they ar<:> the only mws.
(X) We have (n;,;x 2 ) = '.ir;. so W(R) is isomorphic to th<' dilwdral group
of order 12.
(XI) The ouly automorphism of th<' Dynkin diagram is th<' identity, so
A(R) = W(H) and wo = -1.
14. IRREDUCIBLE NON-REDUCED ROOT SYSTEMS
The irrcducihlc, no11-red11c<'d root systems ran be ohtained from the irre-
ducible. reduced systems by using Props. 1;~ and 14 of § l. no. 4. For each
integer I .)! 1. t.lww exists. up to isomorphism, a unique irreducible. non-
reduecd root systcm of rank/: let Rh<' a root system of typl' 13 1 • A the set
of shortest. roots of R; tak<' the u11io11 of Rand 2A. With the notation of no.
5. we obtain t.he 2l(l +I) vr~C"tors
±E;. ±2Ei· E; ± Ej (i < j).

EXERCISES
A11 the root systems considered below arc relative to rral \'Cctor spaces. \Ve
denote h.v (.rly) a scalar product iuvarinnt under tlH' \Ve.vi p;ronp (d'. 110. 3).
L) L<'f. R he a root. systc111, R = R 1 LJ R2 a part it ion of R. Assume that, if x, !J

arc two elcuwnts of H; alHl if J: + y (resp. :r - .'I) is a root. thcu ;r + !J E R;
(r('sp. :r: - y E R;). when' i = 1, 2. Thc11. R is the direct. sum of R 1 and R 2 .
(By usiug the• Cor. of Th. 1, no. 3, show that if :r E R 1 and lJ E R2 then
(;rly) = 0.)
2) Let H. be a root. system. o. and J two roots. If t is a scalar such that .iJ + tn E
R, tlwn 2t E Z. (For n(:'J + tn.n) = 11.(/i.o) + 2t.) If n is indivisible, then
t E Z. (If uot, show, by using Prop. 9, that thPre exists a root / orthogonal
ton· :mch that/+ 1o ER: then (r + 1njo) > 0. so 1n ER.)
:~) L<'t R be an irreducible uou-rednced root system in V. Th<'n' exists a non-

dcµ;enernte symmetric biliucar form <J> 011 V such that. if we ickntify V with
V* usiug <P. W<' have R = R-. (Use Prop. l:l.)
4) Let r he a free Z-module of fiuitc rank l, 1* the dual Z-mo<lulc, I a finite

set of indices, (:r;. xi);EJ a family of elements off' XI'* such that (:r;. :rj) = 2
for all i E I, s; = 8.,.,,,, •. Let. V be the n~al vcc:tor space I' ·:x!z R. whosC' dual
V* ca11 he identified w'ith I'* ?</z R. We also denote by .s; tlw reflt•rtion s; (XI 1
i11 V. Let H (r('sp. R') I)(' the set. of :r, (resp ..ri). Let E (resp. E') be the
subspace of V (r<'sp. V*) generated hy R (resp. B'). Assume that 8;(R) = R
for all i. Then, R is a root. system in E. E' is ca11011ica lly identified with the
dual of E aucl .r7 with J:;· for all i. Tf F (resp. F' is the subspace of V (resp.
V •) co11sisting of the poi11ts invariant under all the s; ( n·sp. 18; ), then I' n F
(resp. I'*nF') generates F (resp. F'). Ifwc put !'1 = (FnE)•.:1(lnF) (resp.
f't = (!'* n E') J; (I'* n F')), then T/I' 1 and T'/Tt an' iso111orphi(' finite
groups.
5) Let R hf' an irrcducihlc root s.vstcrn. For any /J E R,

HOOT SYSTEl\!S Ch. VI
awl consequently ;·(R) = 7(,.\R) for any non-zero scalar A. (Use formulas (17)
and ( 19) of 110. 12.)
()) a) Let A lw an abl'lian group of finite 1ype, and T a finite subset of A

not containing 0. Th('J"C exists a subgroup II of A of finitP index s11f'h that
H n T = 0. (Let t E T. l3y usi11g th<' strncture of abelia11 groups of finite
type, construct. a subgroup of A of finit(' index 1hat docs not contain t. Now
proceed by i11ductio11 on Card(T).)
b) LP!. R be a root system and Pa dosed syunuet.ric subset of R. There <'xist.s
a subgroup H of Q(R) of finite inctex such that P = H n R. (Pass to the
quotient by the suh11;roup ofQ(R) generated by P, and use a) and Prop. 23.)
7) The <·omwction index of a root system is equal to the cktermiuant of its

Cartan matrix. (Use formula (14) of no. 10. Show on the other hand that. in
a real vector space with a scalar product. if {E 1•... ,E,,),(E~ .. .. ,::;,)are two
bases such that (E;lcj) = 6;i, then <let(E, .... ,E,,)(E~ •... • E;,) > 0.)
8) Let R be a rcdun~d root system. C a chamlwr. 2a the :mm of the positive

<"lcm('nts of I"C. P'(R) the set. of ;i· E P(R) such that (;r:, a) E Z. Then,
Q(R) ~ P'(Il) <;;; P(R), and P(R)/P'(R) isoford<:'r 1 or2. Let (:J 1 , .... f3i) be
the basis of n· corresponding to C, and put. 2a = n 1 tli + · · · + 111.0c where the
n; are integers > 0. Then, P(R) = P'(R.) if and only if all the "fl; are f'ven.
, 9) Let B be a root system, Ca chamber. Put B(C) = {01, ... ,nt},
R+(C) = {<l:'1 ..... 01,<l:'1+1 •.... n"'}

(t.he <r; being pairwise distinct.), and o = u 1 + · · · + <l:' ••
a) For ·i = 1,2, ... ,.s, put E; = ±1. Let n* = E 1n 1 +· ·· +E.,n"'. If (n*ln;) > 0
for i = L ... , l, then o* = o and E; = 1 for all i. (Let"( (resp. 6) be the sum of
then; forwhichE; = 1 (resp. -1). Thf'n, (;·ln;)-(t5ln;) > 0.('Ylo;)+(t5ln;) =
(n;ln;). so
2("/ln;) > (n;ln;);
since 2(tlo;)/(<l;jo;) E Z, h1<1;) ~ (o;lo;) = (nln;), so "f - <YE C, /~a,
"( = n and 6 = 0.)
I>) For i = 1, 2 .... , s, let E; =1±1. The set of c:;o:; is the set of roots >0
~~~~~:;::'.·~~~;~~.~~::: :~,·:~~;jl\.ii :;;::':~.:;:.'.,;,;::;;:;~:;~:./7.:~;::·;;o~~:

set of a;. and hence n* into o* = LE;/t;O"(i)· where <1 E 6"'. Applying a),
show that n** = n, so /t;E; = 1.) '
~ 1. EXERCISES
10) Let o h<' th<' highest root of an irr<'ducibl<' rcd1wcd root system H rf'i<itivc
to SOlllC basis. Th(•n. (\'. is the highest root of rr if and 011 IY if all the rooh
arc of the sanw l<'l1gt 11.
I J) With the not at ion of Prop. :n of no. 11. show that fJ; + 01 is not a root
for any pair (i.j).
12) Let R IH' a root syst<'lll ill v of ra11k :? :L (n I ..... 0/) a basis of n. V'
(r<'sp. V") th<' subspace of V generated by the n; for i :? 2 (r<'sp. i ;? 3).
R' = Rn V', R" = Rn V". Let d (resp. d'. rl") lw the determinant of the
Cartan rnatrix of R (r<'sp. R', R"). Ass1111H' that n 1 is orthogonal to all th<'
n, <'XCcpt o 2 , and that I/ o 1 11=11 n2 II· Show that d = 2d' - d".
1:n Let R he a n·cluccd root svst<'lll. (n 1..... n 1) a basis of R and o =

r· 1 n 1 + · · · + r·101 a root. Then. (',(n;ln;)/(oln) E Z for all i. (Considt-r the
inverse systeni.)
lei) L<'t H b<' an irr<'d11eil1I<' root s.vst('lll. A the gtTatcst root length. S the set
of subsets of n ('()llSisti11g of pairwise orthogonal roots of length A. Then. ;my
two maximal <'lertH'nt.s of S 1m' transfornH•d into <'ach other by \V(H). (Use
Prop. 11. and Prop. l of Ch<1Jl. V. q:l. 110. :l.)
};>) Let Rhea root s~·stem of rnnk I.

o) - I E W(H) if and only if R ('Ontains l roots that <m' pairwis<' strongly
orthogonal. (To show that th<' condition is nccessary. argw• by induction on
l using Prop. 1 of Chap. V. fi 3, 110. :3.)
I>) If u· E \V(H.) is of ord<'r 2. th(•n· exists a sl't S of pairwis<' strongly orthog-

011al roots s11rh that 1c is the product of the N,. (n ES).
lG) Let H be a root systcni. [3 a basis of H. and H+ the set of positive' roots
rclatiw to this basis. For II' E \V(R). put Fw = R+ n 11'(-R + ). Show that
the niap "' r---. F,,. is a bijection froll1 \\i(H) to the set of subsets F of R+
such that F and R+ - F arc dosed. (Apply Cor. l of Prop. 20 to th<' subsPt
P = (R+ - F) u ( - F).)
, 17) LP! R I)(' a reduced root system and B a basis of R. For any sushet
.J of I3. kt W.i be the subgroup of \V(R) gell<'rnkd by the rdl<'ctious 8 11 for
n E .J. and let u·. 1 he tlH' lougcst ckme11t of ~·.J. Ld <P he t IH' set of 11'.J.
a) Show that. an ek111c11t ui E W(H) belongs to <P if and only if
W(B) c;:_ R+(B) U (-B).
(To show that the condition is sufficient. denote by .J 1 the set of n E B such
that w(o) E -B nnd p11t. .J = -11•(.J 1). If n E .J 1. w(o) E -.J and 11'.Jll'(n) E B;
HOOT SYSTE'.\IS Ch. VI
if 11 E 13 - .) 1. w1111(n) ER+: otherwise w(o) would h<' positiw rn1d 11•111 1 (0)
11q~ativc. which would i111ply that 1u(n) IH'lo11gs to th<' sulisyst<'llt g<'ll<'T'<ll<'d
liy th<' ,j E .J a11d that I\ hdo11gs to th<' s11 hsvst <'Ill generated bv tJt<' :1 E .J 1 ,
\\'hich is absmd. D<'dtl<'<' thill 11·.111· =- 1.)
18) Ld R Ii<' a root s.vste111 and l<'! Ph<' n parabolic s11lis<'t of R. Show that
th(' co111ph•n1<'11t of P i11 H is dos<'<l.
l!J) Let H be a root systcrn. and l<'t :r b(' <t non-zero cl<'m<'nt of Q(11) of
minimal lengt It. Show that :r E R.
~ :20) Let (n 1 ••••• n1) he a basis of an irr<'d11cibit' rcd11c<'d root system R. Let
rand JJ lw two i11tq.~ers. with 11) 2. s11ch that:
a) Leto= 1· 1111 + · · · + c1n1

he a root. Show that (11ln) = (rr 1 Jrr 1) (in other
words. that n is a long root) if and only if p divi<IPs c,.+ 1 ••••• 1·1.
b) Ld h he th<' Cox<'tcr number of W(R). Show that th<' nu111lwr of lon!!;est

(resp. shortest) roots of R is <'(]Ila! to hr (resp. h(I - r)).
21) L<'t H ht• a root system a11d I3 a basis of R. Let n. J E 13 and w E W(R) be
s11ch that 1i = w( n). Show that t her<' <'xist.s a s<'q11e11c<' n 1..... n,, of elements
of R and a S<'IJll<'IHT w 1 •...• 11•,, _ 1 of ('klllC'llts of \\' ( H) such that
(i) Ot =It. !\' 11 =Ji.
(ii) Ill= Wn-1 . .. ll't·
(iii) w;(n;) = 11;+ 1 for 1~i~11 - l.
(iv) For all i E (I. n - I). t h<'re <'xist 11; E I3 Sll<'h that 11•; lwlonµ;s to the
s11bµ;ro11p of vV(R) gen<'rnted by t IH' rC'flections s 0 , aud s:;,.
~ 22) \\iith the assnmpt ions and 11otatio11s of Prop. ;~;~ of no. 11, rknote by
w 1 ••••• Wt the hmdarnc11tal weigh ls of R.
a) Show that c- 1 (w;) = w; - f),. D<'d1tc<' tlmt l - r· maps P(B) to Q(R).
/1) Let f lH' the comu·ct ion index of R. a11d lt>t m 1, .... rn1 be th<' exponents
of \\i(B.). Show that
I I
f = det(l - c) = Il (1 - 1um 1 ) = 21 Il sin(m;rr/h)
.1=1 .
.1=l
with w = e2;ri/h.
EXERCISES 23!)
<') Let p be a pri111P !llllllh('!'. Write h in I h<' forlll h = p"H. with H not
divisil>ll' by /I. Show I hat 11 divides f if and 011ly if H divid('S one of thr lllj· 2
2:l) Let H bra rcduc·c•d root syste111. and Id X \w a s11hsct of P(Il). Say that
X is sa.turatcd if the following condition is satisfipd:
(S) For all 71 E X. n E H. and i E Z such that i is lwt\H'<'ll 0 and (p, n·),
fl - in EX.
o) Show that <'wry sat mated s1tl>sl't is st:ibk u11<ler thP \Vey! µ;roup \V of R.
h) Show that.. for a11y subset A of P(B). there exists a Slllallcst sat.urntc•d
s11hs!'t S(A) of P(H) co11tai11i11µ; A.
<') Lrt C be a rhamlH'r of H. am! let p E P( 11) n C he a dominant wright. Let
S(p) be the srnall<•st slllnrnt<•d subset of P(H) rnntai11i11g JI. Let E(J1) he the
set of dcmc'11ts p' E P(H) s11ch that
(i) p = p' rnod. Q(R).
(ii) For all w E W. w(f!') ~ II! (with wspcct to the order rdation defined by
C).
Show that E(p) is fi11it<'. co11tai11s p. and is siltmat<'<l. Conclude that. S(p)
is co11tairwd in E(p).
d) Show that. if o is a lo11p;c•st <·lcmc11t of IL S(o) =Ru {O}.
24) \Ve retain the notation and assumptions of the preceding exercise.
a) L<'t p E P(R). a11d h•t \V.p be t.he orbit of p under \V. Show that the
following couditiom; arc equivalent:
(i) S(p) = \V.p.
(ii) (p. n ·) = 0. 1 or -1 for all n E R.
(If (ii) is not satisfied. rn11st.rnd an element q E S(p) such that (qjq) < (pjp),
a11d henec q 1. \V .p.)
b) Let X he a 1101H•Jllpty saturated suhsH of P(R). Show that X contains an
clement. p satisfying conditions (i) a11d (ii) above. (Tak<'/! to he an l'i<'ment
of X of minimal b1gth.)
r·) Assunu• that R is irrcdurihh•. Let C he a d1amlwr of fl. let B = {o;} iEI
IJ<' tlw corresponding basis. and l<>t
.
r = '\"""
L.,., 11;0;
l
.
he the highest root of W. Let .I lw the set of i E I such that. ni = l. Let p -/= 0
be a dorninant w<•igbt. Show that co11ditio11s (i) and (ii) of a) an· <'q11ivalf'nt
to <'ach of the followiug rnnditious:
2 This exercise, hitherto unpuhlislwcl. was com1111111icated to us b.v R. Steinberg.
240 HOOT SYSTEi\IS Ch. VI
(iii) (p, -y*) = 1.

(iv) There exists i E .J such that pis equal t.o the corresponding fundamental
\Vl'ight.
A weight p sat isfyiug these couditious is calk~d minuscule. Show that every
non-empty saturated subset of P( R) contains 0 or a minuscule weight.
~ 2.
\Ve denote by Ila reduced root system in a real vector space V.
1) Let x EV and let. W(:z:) be the subgroup of W(R) consisting of the clements
u~ such that
w(.r) - :r E Q(H).
Show that W(;r) is grnwrnted by rdl.Pct.ions. (Use the affim• \Vey! group of
W.)
~ A:;s11111c that H b irreducible. Let {n 1 •... , o 1 } be a basis of R let (~ =

L1
n;o; be the highest root of R a11d let f be the connection index of H. Show
that .f - l is Pqual to t ht> n11111bcr of indices i such t.liat n; = 1.
:3) \Vith the uotation all<l assumptions of no. :t let 11. he an automorphism of
the affine space E that permutes the hyperplanes Ln,k (n E H. k E Z). Show
that ·u is a displa<"cment. (If ·u 0 is th<' linear map associated to u, show that
the transpose of u leaves R stable. and h<>rice hdongs to the group A(H).)
Dedncc that G is the normaliser of W" in the group of a11tomorphisms of the
affine space E.
4) Let C' he a chamber relative to W(R) in V*, and let C be the alcove
with vertex 0 contained in C'. Let Sa (resp. S) he the set of reflections in the
walb of C (resp. C'). The pairs (W".S,,) and (W,S) are Coxeter systems,
and S ~ S". Show that an element w E W 11 is (S. 0)-reduced (Chap. IV, §3,
Excrc. :l) if and 011ly if w(C) ~ C'.
, 5) Assume that R is irred11dl'1<>· <tll(l choose a dwmbcr C of R in V; dcuote

by B the corresponding basb of R.
a) Show that the rninusc11lc wqights (§ L Exerc. 24) of R form a system of
rcpresC'ntatives in P(R) of the Jnou-zero denwnts of P(R)/Q(R). (Apply to
R" the corollary of Prop. G.)
b) Let X be a saturntt>d subse~ of Q(R) (fi L ExPrc. 23). Ass11me that X
is non-empty and uot r<'duc:ed ~o {O}. Let p be a nou-zcro el<>meut of X of
minimal length. Show that p E I~. (In view of a), p does not satisfy condition
(ii) of Exerc. 24 of § 1; henc<', there exists n E R such that (p, n -) ~ 2, so
s3. EXERCISES 241
p - n E X. Since the \e11gt h of p - n is strictly less t hm1 that of fl. fl - n =0

so J> E R.)
c) L<>t p be a dominant. weight not belonging to Q(H). Show that t.he saturated
subset S(p) generated by p (cf. !j 1, Excrc. 2:J) co11taius a 1111iqut' mi1rnsc11le
weight. (Remark that S(p) b contained in a 11011-trivial class mod. Q(R):
conclude by applying a) aml Exe re. 2,1 <") of 1.) s
d) Let p b0 a dominant weight not belo11gi11g to Q(R). Prove that tlw followi11g
two properties are c•quivalPnt:
(i) pis minuseule.
( v) There is no dominant weight q f: p such that p - q is a linear combination

of elements of I3 with non-1wgative intrger coefficients.
(If 'I satisfies the conditions of (v), let p 1 be a minuscule weight belonging to
S(q). Tht'n. Jlt = p mod. Q(R). Pt <JI, and 11) shows that fJ is not 111i1111sc11le.
Hence, (i) ==? (v). Com·crsely. if pis not 1uinuscule. kt q E S(p)-W.p:
tram;formi11g q by an cl1'met1t of \V if necessary, we can assume that q E C;
b.,. Ex<"rc. 2:~ of~ l, q <fl. q f. µ. a11<l 11==p111od. Q(R). Hc•11ce, (v) ==? (i).)
s3.
The• not i1I ions and assumptions me t hosc> of nos. 2. :t .t
1) a) Sit ow that., for all i lwtwccn 1 awi l. t her<' <·xis ts a 11niq11c d<'rivat.ion
D; of A[PJ satisfying tlw following rn11dit ions:
a1) D; is A-linear.
a2 ) D;(e""1 ) = 6iJc"" 1 (6iJ being the Kro11cekcr symbol).
Ii) L<'t. (.r;) 1 ::;;;::;; 1 be a famil.v of d0mcnts of A[Pjw satisfying the condition of
Theorem 1. Show that
clet(D;(:r.i)) = d.
(Show that det (D; (.rJ)) is ant i-i11varia11t and has rnaxirnal term eP, hence the
result if 2 is not a zero di,·isor in A. Tn'at th(' gc11crnl ens<' by using the
principl<' of pcrma11c11cc of algebraic id011titics.) :i
2) Let P' be a subgroup of P(R) containi11g Q(R). Show that P' is stable
undt•r \V. Constrnct a11 example where the algebra A [Pjw is not isomorphic
to a polynomial algebra (take for R the product of two s.vst<·111s of rnnk l).
3 This ('Xercist', hit lwrto 1111publis]l('d. was commu11icalcd to us b.v fl. Steinherg.
242 HOOT SYSTEl\ IS Ch. VI
If R IS a root. syst<'lll, de11ot(' by w+ (R) tlw set of !'lcmc11t.s of \V(R) of

dl'l:cnni11a11t 1.
~ 1) L<'t R he a root system of t.ypc Es.

a) Show llwt if n. J E H arc co11grncnt. modulo 2Q{R), then ;J = ±o:.
h) Deduce from a) that. if 11• E \V(H) act.s trivi;illy on Q(R)/2Q{R). t.hcn
II'= ±1.
1') With the 11olatio11of110. 10. show I hat th<• q11adra.tic form ~{.rl.r) 011 Q(R)
ddiries by passage t.o tlw quotient a no11-dcg<'nenttc quadrat.ic form q8 on the
F:rv<'ct.or spa<:<' Q{ll)/2Q(R). Show lhat th<' pscudo-disni111i11aut of l/x {cf.
Algebra., Chap. IX, ~i 9. Ex<'rc. 9) is zero. and tlmt. l/s is of iwi<•x 4.
d) L<'t O{q8 ) be th<· orthogonal gro11p of Q(R)/2Q(R) for this form. Define,
b.v passing to the quoli<·11t. a ho1110111orphis111
h: W{R) _, O(qi;).
Show (b.v co111pari11µ; ord<•rs) that t lu~ s<'qll<'nce
l _, {1.-1} _, W(R) _!I:__. O(<bd _, l
is ('X;H:t..
('.)Show that. tlH' irnaµ;<· of W+(R) 1111dcr his t.hc s11bgroup o+(q8 ) of O(<Jid
defined ill Alyelm1.. Chap. IX, fj!). IJO. fi. DedUC(' tlrat w+(H)/{L-1} is a
si111pk no11-alH'lian group.
~ 2) Let R he a root systpm of typP Et;.

n) Put E = Q(R)/:W(H). This is an F:1-v<'ctor spaeP of dimension 5. With
th<' notation of 110. 12, show t.hat the scalar product (:rly) defines 011 Ea non-
dcgcncrntc sy111111ctrie bilinem form :.p. Show that any two distinct clcnlC'nt.s
of B have distinct images in E.
b) L<•t 0(9) be the orthogonal gr01j> of ..p. Thc11, O(p) = {l. -I} x SO(:;).
The spi11or norm defi11cs a surjcctivc homornorphisrn from SO(;y) to {l. - l}
whose kPrrwl is denott·d by so+( ). The group so+(.p) is si111ple of or-
der 25!.120. The quotient 0 (c.p) /SO+ ( c.p) is of type (2, 2). Dcd uce that 0 {:p)
contains a unique subgroup fl( 'P) o index 2, dist i11ct. from SO( :p) and not
('ont.ai11i11p; - 1.
<") Every clc111c11t of A(fl) <l<'fincs by passag<' to the quotient. an dcmcnt of

0(.,:;). Show (by ('()Ill paring orders) I hat I his giws a homomorphism from
A(R) to O(<p). The image of W(H) d11der this homomorphism is fl(c.p). and
~ 4. EXERCISES 243
that of w+(R) is so+(y). lfrrH'('. W(R) is an Pxtc11sion of Z/2Z by n simple

group of onl<•r 25920.
d) Ld. F = Q(H.)/2Q(R). This is 1111 F 2 -wctor space• of dimension f>. Show
that th<' quadratic for111 ~(.r/.r) ddii1c•s hy passage to the• q11otie11t. a non-
dege•11ernte quadn1t ic form fJ(i 011 F. of psP11do-discrimi11a11t. C'<(IHd to I. If
O(q 6 ) de11oles the eon·<·spoudi11g orthogo11al gro11p. ddi11c. by passi11g to the
q11otie11t, a ho1110111orphis111
h: W(R)--+ 0(11()).
Show that h is injective (note that - l ¢ \\'(R)). and tlw11 that it is an

isoworphism (compan· orders). Deduc.:c that thert' is a11 isomorphism from
w+(H) to o+(qG) (cf. Al9ebm. Chap. IX, s!J, 110. fi).
1·) B.v comparing c) a11d d). show' 1 that so+(~) is isomorphic to o+(qr,).
, :~) Let H lw a root systc111 of type· E1 .

a) Put E = Q(R)/2P(R). This is mr F2-\'<'ctor space of dimension (). With
th(' 11otatiou in !HJ. l l. show tltat the se:alar product (.r/y) clefiw·s a 11on-
dege1wrntP altcmating bilirH'ar form or1 E.
h) Ded tHT from 11) the exist<'11c.:e of a11 exact scqw•nce
1 ___. {l.-1} _, W(R) ~ Sp(6.F2)-+ l.
(Use the fact that Sp(G.F 2 ) is of order 2' 1.:3 4 .fi.7.)
<') Show that the restriction of h to w+ (R) is an isomorphism from w+ (R.)

to Sp((). F 2).
d) Give a sc•eortcl proof of /J) h,v 11si11g thl' quadratic form q7 011 th<' F2-V<'C'tor
spaec Q(H)/2Q(R) i11cltHTd by ~(;rl.r). IL'i well a~ the isomorphism
O(q7 ) __, Sp(G. F2)·
, 4) Let R bt' a11 irr<'d 11ei blc red nc·t•d root system in V. (II' 1 •.... n1) a basis of
R. 6 th1· highest root. Pnt. 1\ = n 111· 1 + · · · + 11101. \Ve are going to <il-tc•nninc
the maximal dosed. symmetric subsets of IL distinct. from R.
a) Let i E { l. 2.... , I}. Let R; ])(' thl' set of n E R which an• linear combina-
tions of th1· nJ for j cf. i. Show that R; is maximal if aJHI only if 11; = 1.
1 l\I. Ki\'ESEH has shown t.hat. all thP ··exceptional isomorphisms'' between the
finitf' classical µ;ronps ca11 lw obtained by nsinµ; a11 analogous met.hod. Cf. Ober
dit• A11snahrnc-lsornorphismcu zwisclw11 klassisch1•n Cruppen. HamlwrgFr Abh.,
Vol. XXXI (1!)67), p. i:rn-110.
IWOT SYSTE~lS Ch. VI
/J) Let i E {1, '2 . .... I} and ass11111c· that n; > I. Let S, h(' t ll<' sd of roots
I
L
.J= I
111 1 0 1
. .
with 111; = 0 (t11od. n, ). Sh<m· that S, is 1J1axir11al if aHd only if 11 1 is
prin1<'. (If 11; =ah with a> l, /J > l, rn11sider th<' s11hset S' or H consisting of
the roots L 111.10: 1 \\'ith 111 1 = ()
(111o<l. u). and show that S' strictly contains
.7
S,.) Show tlrnt the roots -rt. o 1 (j Ii) forn1 <1 basis of S, (which is of rank
l). Dcduc<' the Dy11ki11 graph of S;.
c) Every maxirnal dos('d. s,vmmctric s11bset of R is trnusforn1cd by an element
of W(R) into OllC or the suhsc'tS described in a) or /1). (Ld E he s11ch a subsPt.
Then E =Rn H. wh<'rt' H is a s11bgro11p of Q(R) of fiuit<' ind('x (81. Exerc.
6 b)); we can assurne that Q(H)/H is cyclic. Theu there exists 11* E: V* such
that Eis tlw S<'t of o E: R s\lch that (u*. n) E Z.
rl) List tlw maxirnal. dosed. symmetric subsets of Il for thE' differm1t types of
irrcd1t('ible rl'dll('ed root systPrns.
5) Let R be a root system. and P'(R) the s11bgrollp of P(R) introd11ced in

~ 1. Ex<'IT. 8. Show tlrnt P'(H) = P(H) if R is of typ<' A 1 with l <'Ven. or
B 1 with l ::::= o,:3 (mod. J). or D 1 with I = 0. l (mod. 4). or G 2 • or F 1 , or
E 6 . or E 8 . Show that P'(ll) = Q(H) if R is of t~·1w C 1• or B 1 with I = 1. 2
(mod. ·1). or E1. or A 1• If R is of type A1 with l odd aH<l > 1, P'(R)/Q(R)
is the 1wiq11<' subgroup of index 2 i11 the C'_vdi(' gro11p P(H)/Q(R). If H is of
t.VJ><' D1 with I = 2. :~ (111od. 4). P'(R)/Q(n) is the 11uiq11<' subgro11p of order
2 of P(B)/Q(R) stahl<' 1111dPr A(B).
~ ti) Let H be an irn'duciblc reduced root syst<'lll, (n 1 ..... ni) a hnsis of R,

fJ101 + · · · + p1n1 tlH' highest roo1. o 1n 1 +···+111n1 the su111 of the positive
roots. 111 1..... 1111 the <'XJ><rncuts of 'v\"(H). q its onl<-r . .f the co11w'ctio11 i11drx.
a) VPrih· that. in l~<t('h C'asc.
/lp1 Ji2 . .. µ1rn 1 m2 ... 1111 = o. 1 112 ... n1.

/1) Show that
.'Jlllt n1·2 ... 1111 = .fa1n2 ... a,.
(Use a) and Prop. 7 of 32. 110. :l.)
c) For anv positive root n =Lr;()';. Id c(n) =Lr;. Cakulatc in rnd1 C<lSC
I I
the polynolllial
P(I) = L r(n).
n>ll
I
Verify" that P(t) = t L(l + t-t- ··• + t"',- 1 ).
i= 1
-, For a proof that does 11ot 11sl· the ciarsification. sl'e: f3. KOST ANT. The principal
three-cli111011sional s11hgro11p and tlw f3Nti 1111mbffs of" <·ornplcx simple Lie g;ro11p,
Amer. ./. of Maths .. Vol. LXXXI (105!J). p. 07:l-l(J:l'.2.
EXERCISES 245
7) Let n b<' a11 irreducible wduccd root systf'm.

11.) Vcri(y that tlw cano11ical homomorphism from A(n)/W(R) to the group
of a11t.0111orphis111s of P(n)/Q(B) is injective.
I>) D<'d11cc that - l lwlo11gs to \V(R) if and oulv if Q(H) 2 2P(R).
8) L<'t H 1 aud R 2 lw two irrcd1wibl<' rcduc<•d root systems. Show that if

W(n 1 ) aud W(R 2 ) have the same order. R 1 is isomorphie to H 2 or n 2-. (Use
the dassification.) Do<'s this r<'sult still hold without the w;sumptio11 that. R 1
is irreducible?
!)) l.<>t R be a root syst<•m of rnnk /, and let p be a prime number dividing
the order of A(R). Show that J> ~I+ 1. (R<'d11e<' to the irrcdu<'ibh· ('m;c, and
use• tlw classification.)
10) 11.) Let (\V, S) he au irrcducibl<>. finite Coxder :-;ystcm. Put
W(t) = L tt(i..) (cf. Chap. IV. ~I. Exerc. 26.)

wE \\'
Ld 111 1 , ...• mt lw tlw c•xpm1cnts of \V (Chap. V, ~ G. 110. 2). Veri(v the

fon1111la
I
\V(t) = fl (I+ t + · · · + t"'')
i=l
for :-;111all values of l (use Th. I a11d Excrc. 2() of Chap. TV.§ 1) 6 .
h) L<'I n be a reduced. irrcdneiblc root system aud W (resp. (W,,) the \Vey!
group (resp. the affine Wey! group) of R. <'quipped with the Coxeter !!;ro11p
strnct11w dctt'nni1wd hv tlw choic<• of an alcow. Define \V(f) and the Pxpo-
ll<'llls 111; as nhovc am! p11t
W,,(t) = L 11<u·l.
wEW,,
VPrifv th<' formula
I 1 1 I + t + · · · + t"''
W,,(t) = \V(t) fl - - f l-----
i=I ] - /."'• i=l 1- f'"'
for s1t1all values of l (use Th. 2 am! Exr~rc. 2() of Chap. IV, § I )7.
i; For a proof of this formula that do~·s not use the classification and is valid for all
valtws of/, sec: L. SO LO :\ION, The ordC'rs oft lw finite C'lwvallcy groups, .Journal
o.f Alq<'bra.. Vol. Ill ( l96G). p. :m)-:J!):l.
7 For n proof of this formula that docs not use the classification aud is valid for all
vaht<'s of I. see: R. BOTT,A11 application oft he l\fors<' t.hcor~· t.o I he• topology of Lie
groups, !Juli. Soc. Muth. France. Vol. LXXXIV (L%G), p. 251-281; N. l\VAHORI
a11d II. i\IATSl"l\IOTO, 011 some Brnhat dcco111posilio11 and the structure of the
Heckt> rings of p-adk Chcvall<'Y groups, l'nhl. Math. Inst. Haute.~ Et. Sci... no. 25
(l!)(i5). p. fi-118.
2-W HOOT SYSTE:\IS C'h. VI
~ 11) LC'! (W. S) lH' a Cox('tt•r s_vst.Pm of type H:1 (cf. Th. l ).
s
a) \\iit.h tilt' notation of Chap. V. G. Ho. 2. Proof of L<'1111na 2. show t.hnt.
(.::'. z") = 71/5: d<'dll('<' that th<• Cox<'tPr 11t1111IH'r It of Wis <'<pial to Hl. and
that t.h(' <'XPOll('lltS of w an· 1. .') Hild 9.
h) Show. by usiug a). that Card(\\')= 120. a11d that th(' ll11111hl'r ofn·fkctions
in Wis 15.
r) R<•cm'<'l' tlH' formula C'md(W) = 120 by applying Excrc. f> of Chap. V. § 3.

d) Let A:; be the altnnating group of { 1. .... G}: if n.11. <-. d arc distinct ele-
lllellts of {I ..... fi}. denote hy (ah) th<' tra11sposition of a <tll(l 11. and (a/J)(('(l)
the product of the transpositions (uh) alld (('(/). L0t
1'1 = (14)(2:1). '"2 = (12)(15). 1';1 = (12)(:34).

Show that (r 1 1·2 )'' = (r 2 r:i): 1 = (r 11·: 1 ) 2 = I. D(•duce tlw <'xistcllc<' of" homo-
morphism f : 'vV ~ A; that takes S to { 1· 1, r2. r:i}: show that f is smj(•ctive.
<') Let € : w ~ {± 1 } bP the holllOlllOrphism II' ........ ( -1 ) 1( w). Show that
(f. €) : w A-, x { ± 1}
--+
is an isomorphism. (Use th<' fact tlwt the two groups hPillg consi<kred have
the sa11w ord('L)
~ 12) Let ( l. i, j. A·) he th<' callOHirnl basis of th<' fidd H of qllatcrnions, by

lll<'1ms of which H is ickntificd with R 1. Equip H with t ht• scalar product
~(.r/) + 1/f:). L<'t r he the 111ultiplicatiw group of q11aterniOllS of norm 1.
a) If a E I', the orthogonal refkct ioll s" ill H which transforms a to - a is

t lw map ./' r--+ - a;T·a.
/1) Let <J = eos f, - 4i

+(cos:\" )j E I'. and r = ~ (I+ i + j + k) E I'. Ld Q he
tlw s<'t of q11at.crnio11s obt.ai11('d from L q. r hy arbitrary P\'<'11 1wrn111tat ions
awl sign changes of tlw coordinat<'s. Thell Q is a subgroup uf J' of onkr 120.
<') Let W hP the subgroup of GL(H) g<~1wrnt(•cl ''-'' th(' s,, for o E Q. Show
t.hat \V l<'iwcs Q stable. is finite. irn'dncihk. and llo11-crystallogrnphic: deduce
that. W is of t.yp<' H 1 (<"f. Theorem I).
d) Show that W acts transitively 011 ~~ (usl' Prop. ;3 of Chap. IV, s1).
1•) Let a. 0 E Q. kt V he tht• vector s1~bspac:c~ of H orthogonal to n.o, mu\
let
\V 0 be the stabiliser of a 0 in \V. Shot that th<~ restrict.ion of W 11 to V is an
irr('d11cihlP group g<'IH'rntc•d h.v rd\c•C't°OllS (Chap. V. s:3. no. :3) and that it is
non-crystnllogra.phic. D<'dllC'<' that W is 11 Cox<'t<'r group of type H;1.
.f) Show that Carcl(W) = 2<i;3 2 52 (uscld). r·) and Ex<'rc. 11).
_q) Show that the Coxl't('r ll11t11bc•r of \l' is equal to :30. and that its exponents
arc 1, 11, rn. w.
EXERCISES 2H
l:l) LC'! V b<' th<' hyp<'rph111c i11 R 9 with <'qt111tio11 .r 1 + · · · + .l'<J = 0. Let R
IH' th<' sulisd of V <"Oltsistillg of th<' points
(2. 2. 2. -1. ·-1. - l. -·I. -- I. --1 ). (-2. -:2. -2. I. l. l. l. J. I),

(:J. -:l. IJ. 0. 0. 0. IJ. 0. O)
tog<'tlll'l' \\·ith t!tl' poi111s olitni111'cl frrn11 th!'sc hy IH'l'lllllting th<' c·oordi1111t<'s.
Sl1ow that B is 11 root S,\'St<'lll in \'of t.\'IH' Es.
l I) \Vith the IH>li1tio11 i11110. 7. shm1· that th<' 1111to111orphis111 of'R1+ 1 whid1
trn11sfon11s .: 1 to .:'2. :'.! to .::i· .... ..:, t.o E1+ 1 a11<l .:1+ 1 to .: 1 i11d11<"<'s 11 C'oxdcr
trn11sforn1<11io11 oft lw syst<'lll R of type A1.
I 0) D<'1<'1'lllinc' t lw 11ii1111sc11k \\"<•ights (li 1. Ex<'l'<" 21) for Pad1 typ<' of n•dnred,
irn'd1wihk rnot s.\·st<'lll. (One finds tll<' f1111da111P11tal wcight.s ...,· 1 ..........,for A1,
t IH' \l'<'ight ....,., for 131. t II<' weight ;.; 1 for C,. t ii<' W<'ights ...; 1. <-<-'!- 1. .,;t for D,. the
\l'Pights ;,; 1 n11cl .....·c; for Ee;. thl' \V<'ight ,,; 7 for E,. and 110111' l'or Es. F 1 a11d G 2.)
~Hi) L<'f (\\'. S) Ii<' Hll irn'dncilik. finite C'oxd!'r systc111. ancl kt /1 = C'arcl(S).
a) If(\\'. SJ is not. of typ<' F. 1. show that th<'l'<' <'Xists 11 s11bst'I X of S with

11 - I Pl<'lll<'llts Sll('h I hat (\V x. X) is of type' A,,_,.
h) ld!'11tif.\· \\'with a s11ligro11p of GL(Rs) h.v 111<'<\lls of tlH' cano11ical rcp-

res<'11t/\tio11 (Chap. V. ~I). Sho11· that th<')'(' r'xists n b11sis (c· 1..... <',,)of Rs
s11d1 that. for <'Wl',Y JH'n1111tatio11 rr E 6,,. th<' <111tomorphis!ll of' Rs that
trn11sl'om1s 1'; to 1·1T(i) for 1 ~ i ~ " h<'1011gs to\\' ('·B11rnsicle"s Tl1<'on'111'').
(\Vhc11 (W.S) is 11ot of typ<' F 1. 11s<' o): \l'h<'n it is of t.\'I><' F 1• n'111ark that.
\\' C"011tai11s a s11bgro11p of typ<' D4 (cf. 110. !J). which red1H'<'s th<• problem to
the pn'<'<'di11g ms<'.)
c) Let E he a subgrn11p of' t 11(' group of auto111orphis111s of (\V. S). Show
that th<' S\'llli-din'ct procl11ct E.\V of' E hy \V cmlH'ds i11 a c;u1011ical wa_\' i11
GL(RsJ. Show 1ln11 the s11bgro11p of GL(Rs) th11s dcfiu<'CI is g<'llf?nt1C'd by
rdl<'tl ious. l'Xcept in tlw followiug four cases:
(i) (\\'. S) is of typt' A,,. n ~ -1: tl1<' grn11p Eis of order 2.
(ii) (\\'. S) is of t,vpc D 1: tlu• grn11p Eis of orckr :~.
(iii) (\\'.SJ is of t.\'l><' F 1 : th<' group Eis of ord<'I' 2.

(iv) (\\". S) is of t.\'[H' E<;; the grn11p Eis of order 2.
Slul\\· that. i11 <'<IS<'S (i) a11cl (iv). E.\V = {±1} x \V. Sho\\' that.. i11 case (iii),
tll<' group E.\V does 11ot l<·aw· stai>ll' all.\" lattice i11 Rs.
d) L!'t \\' 1 IH' a finitP s11ligro11p of GL,, (R) gc11ernt<'cl h_\' rdlectio11s. nud
ass11111<· that \\' 1 is irn'd11cihlt• nud <'Ss<'11t ial. Ld G IH' a fi11it<' subgroup of
GL,,(R) contaiui11g W 1. Show that citlwr G is g<'llC'nttl'd b.\· reflectio11s or is
of th<' fonu I·:.\\'. whcr<' E n11<l \\'an' OJH' of the tq><'s (i). (ii). (iii). (iv) of c).
248 ROOT SYSTEl\IS C'h. VI
(Let W be the group generated by the r<'flections belonging to G. The group

G pcr11111tcs the charnh<'rs relative to V\'. Dcd11c<'. as in ~ 2. no. :~. that G is
of the form E.\V as above.)
HISTORICAL NOTE
(CHAPTERS IV, V AND VI)
(\'.I3. - The roman numerals in parentheses ref<'r to th<' hibliogrnphy at the

<'nd of this note.)
The groups considered in t hcsc chaptC'rs appeared in conn<>ctio11 with various

questions of Geometry. A11alysis and the Theory of Lie groups. soml'times
i11 tlH' form of permutation groups and sometimes in the form of groups of
displaccrncuts in euclidean or hyperbolic g<'omd.r.v, and these various points
of view have been unified only n•n•ntly.
The historical roots of the th('ory arc substantially C'arlier than thP intro-
ductiou of the concept of a group: indeed, tlwy arc found in the studiE•s of the
··regularity" or "symmetries" of geometrical fig11J"es. a11d not.ably in the deter-
mination of the regular polygons a11d polylwdra (which ccrtai11ly go<'s hack to
tlw Pythagoreans). which rnustitutcs I he crowning ad1icv<•111c11t. of t.lw Elr.-
ments of Euclid and om• of the most admirable creations of t.hl' Greek g<'nius.
LafPr, notably with the Arab authors of the high l\liddlt• Ages. and then
with Kepler. the lwginnings of the mat hl'llllltical theory of r<'gular "tilings''
of the plane or the sph<'rc by (not necessarily rqi;ular) congruent polygons
appear; this is 11ndo11htt>clly reiatl'd to th<' various typ<'s of dc•coration de-
vised hy the aucient and Arab civilisations (which can properly he consi<l-
cred as an aut hcnt.ic part of the rnatlwmatics dcvc•lopcd hy t.h<'s<' C'ivilisations
(XII)).
Around 1830-1840. studies in crystallography (H<'ssel. Bravais. l\fobi11s)
lead to th<' consideration of a problem that is actually the determination
of th(• finite gro11ps of displacernent.s i11 euclidean space of :3 dimensions. al-
t hough the authors q11oted above do not yet use• t hc languag<• of group theory;
that do\'s not really come into use until about 1860. and it is in the context
of the classification of groups that Jordan, in 18fi9 (VI). determines the dis-
cr<'l.l' groups of orie11tation-prcservi11g displacc'ment.s of R:i (and, more g<'n-
Prally. the do8ol subgroups of tlw group of orientation-preserving displace-
me11ts).
This st ream of ideas develops in rnauy directions nntil the fi11al years of the
HJt h century. The most sig11ifica11t of these developments arP the following:
1. Co11tin11ing a trend that appears very early int he theory of fin it<' groups,
"presentations'' of finite groups of displacements by generators and relations
250 lllSTOHIC.\L NOTE
of a situp](• tvpl' arc so11ght. Tl111s. 1lmll iIt Oil. in I K;)() ( V). pro\'C's that tlw finite
gro11ps of rot ntio11s of ('IH'lid<'aII spac<' R: 1 arc )!,('lHTat(•d h.\· tm> )!,('llPnttors
S. T sntisfvi11g tlw rdations S 1' =· T' 1 = (S'IT1 = I for appropriate v;1J11('s of p
awl 11.
2. Disn<'lt• gro11ps of displ<H'<'llH'llts lltay or rna.\· 11ot <·011t"11i11 rdkct ions.
Tu 1K:)2. l\fobi11s PSS('lltiall_\· dct1•rn1i11Ps tl1c finite gro11ps of displ<1<TllH'11ts in
spll<'rinil g<·o11t(•t ry .<Ji:111T11inl /Jy n'.fl1dio11s ( \d1ieh is l'q11iva le Ht Io I h<• same
probl('lll for tinitP gro11ps of <'1tcliclPan displa«Plll<'llts of R: 1 ): h1• finds that,
with I hP 1•xcPptio1t of t Ill' cyclic grn11ps. Sll('h a gro11p has as fuwla1t1eHtal
dmrnlill a sph('ri('al I ria Hglt• with sides of the fornt rr / p. rr /If. ;r /,.. wlu•n• />. q. r
are three illl('!!crs• .. > l such that lJJ + l,, + l,. > J (Ill) (d. ChaJ>. V. s::~ ·~,
Excrc .. 1). He also 11oti('cs that tiH'S(' gro11ps rn11h1iu all th<· fi11itc groups of
displac1•11u•11ts as s11hgro11ps.
:~. Tll<' lat tcr dcvP!op1ll<'Hts are extruded i1tto a rn•\\' an•a wlH'll the stud.\' of
··hiiugs .. of th<• co111pl<'x plallc or oft lw half-plane hy llH'ans of tig11res bo11nd<•d
by <"irnilnr ares h<~gi11s. followillg the work of ni<'lllatlll a11<l Schwarz Oil hy-
]>('l')!,<'Ollldric f111t('ti0Hs mid conformal rcpn•s('lltal ioHs: I<lci11 aml Poincare
1wtk<• it th<• fomtdntioll oft h<• t h<·<ir.\· of ··a11tm11orphic fuHctioHs·· a11d recog-
Hise i11 it (for t ht• Ci\S(' of ciff1tlar arC's ort 11og0Hal to a fixPd strnight. litw) a
problem eq11ivakut to tltc• del(!l'lltim1tioll of tlw dis<Tett- grnnps of displace-
HH'lll.s of th<• 11oll-('t1didcall hy1wrboli(' pln11<· (idcHtifi<'d with tlH' ··Poi11<"are
half-plaiw.. ) (X).
I. Th<' 11otious of n·gnlar pol_vh1•dro11 and of the tiling of R: 1 hv such
pol_vh<'drn ar<' <'Xt<•rnlcc l to all t hl' 1•1wlideall SJ><H'<'s R" b.\· Schlii!J i. iu work
that goPs baC'k to 1d>o11t 1850. but \\'hich was 01dv publislwd llllldt lat<'!' a.ud
which was ignored for a loHg time (lV): he dett'n11iu1•s c01t1plctc•ly t lw n•gnlar
··polytopes'" i11 each R". th<' group of displacp111<·11ts lc•ayiug i11\'ll.riallt snch a
polytop<'. aud a fn11dauH'lltal dmuaiu oft his group which. as iu t hi' cas<' 11 = :l
stndied h.\· l\Iiibi11s. is a "cl111mlwr'" whose projfftiou oil the sphen• S 11 _ 1 is
a splterirnl simplex. Hmv<'Vt'L h(• does Hot take 11p th(• i11wrs<' problem of
dct.ermiHing th<' fiHit1• groups of disphH·c•tn<•Hts g<'H<'I'at<•cl b.\' rdlcctions in
R": this prohl<'m will oul.v be solved lllud1 later. by C:oursat for /1 =··I (VII),
and for arhitrar.v /1 ill the work of E. Cartall (TX f)) and Coxd<•r (XIV). to
which W<' shall r<'t11rn lat<'r.
\\'it It the work of Killing and< · E. C'art11ll on Lie grn11ps. 11 new st.n·nm
of ideas IH'giHs around 18!)() that \\"ll dl'n·lop for 11 long period wi1 h011t. liHks
t.o till' pn•<'<'ding work. 111 the stud. of KilliHg (\'Ill) alld Cmta11 (IX a)) of
the str11ct11n· of c·o111pl1•x semisilllpl ·Li<' alg<'hrns. <:<'rtaill linear fon11s ;,;"on
a ·'Cartan snbrilg<'hrn'" h of' such a Li<• alg<'l>rn g im1m·diately play a basic
rol<': th<',\' HIT the ··root_s·· r<'iativP 1to h. so C'allcd ln•<·aus<' 1h<',\' appear ns
th<' roots of Ill<' drnrnet<'I'isti(' ('quation cfot(adg(.r) - T) = 0. ('Onsicl<•rt•cl as
fu11rtions of .r E ~- TIH' propcrti<'s ol\ lh<'s<! ··roots .. Pst11hlislH'cl Ii.\' Killin~ and
Cartan amo11nt to the assertioH tha~. ill the• lm1g11ag<' of Chnp. \'I. t-lwv form
a ·'recl11c1'cl root syst(•tn·· (l'f. Chap. VI. fj I. no. 1): thl'.\. t.h<'ll show that the
lllSTOIUCAI. !'\OTE 2!i l
dassilini tio11 of the co111plex s('lllisi nip le Lie alg<'hras red tl<'<'s to that of the
associa t cd ··root systt•rns··. w h i<"h i t.sdf wcl11cc::; lo Ill<' dd('l'lllim1 t io11 of <:<'rta in
matrices with i11t.eg<•r codfici<·uts (later called ··Carla11 nw trices··: cf. Chap.
s
VI. l. 110. !i). Killing a11d Cartan also show t lw t~xistcnce. for (•wry root. w,.,
of a11 invol11tivc p<'n11utatio11 S,. of th<• s<'t of roolsx: they 11se ill au essp11tial
way the tra11sfonm1tio11 C ::: S,, 1 S,,, ... S,.,. the proclt1<"l of the penrn1tations
associat <'d with I roots fon11i11g a. fm1darm'nt.al system (a I nrnsfonnat ion today
called a ··Coxl'tcr transformation··): they <'H'll ext<•JHI I his pernllltatiou to a
lill<'HI' trnusfornrnt.ion oft lw vector spa<·c· µ;cuerat<•d h.v tlw fuudalll<'lltal root:-;
...,·.,, (1 ,,:;; i ,,:;; l). a11d study its cigeuvahws ((VIII. II). p. 20; (IX a)). p. !i8).
llut ll<'ithcr Killing. uor Carta11 initially. SPClll to han• t honght of <"Ollsidc•ring
the group 9' g<'IH'rated by the S": and when Cartan. a little later (IX h)),
dl'lcrn1i11cs t lw Galois gronp 9 of the dmract.<'ristic <'quat ion
dct(aclg(:r) - T) =0
of a ·'gc11crnl clcnwnt"" ;J" E ~. he stndics it. initially without bringing in the
S,.: t hirt.v years lat<•r. uuder th(' infltH'IH"e of H. W<',d. lw pron•s (JX <")) thnt
9' is a normal subgroup of 9 awl dctcnni11es i11 all rn~cs thf' st.rndlll'<' of
tlH' q11otic11t group 9 /9' which (for a simple Lie algebra g) is of order l or
2. except. for I.VJ><' D.1 where it is isolllorphic to 6:1: at th<' sa1w· time he
iut.erpn•ts 9' as th<' gro11p i11d11ced by th(' inner a11tornorphisrns of a complex
sclllisimple Li<• algebra leaving fixed a Cartan snhalgdira 0 .
The work of J-1. \Vey!, to which we hm·<· just: allu<IP<I. iwrngnrat.cd the
).!;<'Olll<'lric interpretation of the group 9' (since calkd the ··\Vcyl group"' of
g): IH' had the idea of cousid<~ring the S0 as refi<'ct.io11s in the wctor space
of li11('ar forms 011 ~- i11 the same way as Killing and Cartan h<1d clonp for
!he transforrnatio11 C. It is also i11 the nwmoir (XIII) of H. Weyl that the
fuudament.al domain of t.lw ··affi11c \Vey\ group·· appears (but without the
link to the ··Wcyl group·· 9' !wing c:karly indicated): Weyl us<'s it to to prove
I.hat the funda11H'11t.al group of a compact sernisilllple group is finite. a crncial
point i11 his proof of t.he complete rcd11cibility of the linear rcprcsc11tat.ions
of a (:0111pl<'x semisirnpl<' Lie• algebra. A littl<' later. E. Cart.an c·0111plctes
tl1<! synthesis of the µ;lobal points of \'iew of JI. \Veyl. of his ow11 theory of
wal or complex scmisimple Lie algPlm-is. and of t ht> theory of ri<'mannian
sy111mdric spaces that. lw was then constrnctiug. Iu the 111e111oir (JX d)). he
c0111pletcs the determination of the fw1da111e11tal polytopes of the \Vey] group
and of the affitll' \Vey! group. and introduces the syste111s of w<'ights and
s
radical wt'ight.s (Chap. VI. 1. 110. !J): in (IX c)) he C'Xtcnds this discussion to
sym11wtric spaces. and not.ably e11co1111t<'rs llw first cxa111ples of 11011-rcd11ced
root sysl<'llls (Chap. VI, ~H. 110. I). Fi11ally. the article (IX f)) gives the first
~The not.at.ious v.J,. and S,. correspond rP:;[wclivcl,v tu tlH' 11ot'11tions n and s,. of
Chap. VI. !i l.
!J Tlw notati;rns 9 and 9' corn~spoml respect.ivcly to the notations A(l1) and \V(R)
of Chap. VI. ~ l. no. l.
252 HISTOUIC:AL l\OTE
proof that CV<'l',Y irwducihlc finite group gcrwratcd by rc·flcctions i11 R" has
a fundanwnt.al domain whose projection onto s,,_ 1 is a sph<'rical silllplex;
it. is also i11 this work that he proves th<• 11uiq1wness of t.he highest root
(for a11 arhitrnr.v h·xicogrnpltic ord<'ring 011 the root system) h,v g<'OJll<'trical
considcrnt ions.
A little later, van der Wa.crdc11 (XVI). starting from 1he llH'moir of H.
WeyL shows that t.hc classifkaticm of the cmnplcx semisi111plc Lil' algPlwas
is cquivalc111: to that of the n•d11ccd root. systems, which he carries 011t by
dc111cntary geometric co11siderntio11s (whcr\'as. with Killing and Cartan. this
classification is a result of complicated cakulations with dctcr111i11a11t.s). At
about tlH' same tirrw, Coxel.cr determines Pxplicitl.v all the irreducible finite
groups of E11clidca11 displacements which an~ ge11ernt<'d by refkctio11s (XIV
<')): this cmuplctcs the lll<'lllOir (IX d)) of Cartan. which had only det.er-
111i11Pd tlw ··crystallographic'· groups (i.<'. those associat<'d t.o a root syst<'m,
or having a11 cmheddi11g in m1 i11fi11itc disnete group of displan•111<•nts). The
followi11g year (XIV d)), Coxctcr shows that tlw finite groups g<'IH'rnted hy
retle<:tio11s are the 011ly fi11ite groups (up to isomorphism) acl111itti11g a pn•sen-
tatio11 by gcucrntors R; suhj<•ct. lo rclat.io11s of the form (R;Rj) 111 ' 1 = 1 (m; 1
i11t.C'gers), he1t('l' the 11a.me "Coxctcr·· groups since µ;ive11 t.o µ;roups (fi11itc or
11ot) admitting su('h a presentation.
Th<' first. link hctw<•cn the hvo streams of n'seard1 that we have ckscrihccl
above seems to have he<•n established by Cox<'ter (XIV bis). and then by
Witt (XVII). Tlwv ohsPrvP that. the irr<'ci11C"ihl<' infinite' groups of E11clid<"an
clispla('.('llle11ts genern.tcd h.v n'Hect ions correspond hijeC'l.ivdy ( 11p t.o isomor-
phism) to cornpkx simple' Li<' alg;cbrns. \Vitt p;iVl's a rn•w d<'t<·n11inatio11 of
discrete groups of this typ<'. and also C'Xl<•11<ls the t.heon•m of Coxc1cr (XIV
d)) referred t.o ahov<' hy charnct.C'rising th<' Cox<"l<'r p;ro11ps isomorphic to in-
finite cliscrd.<' groups of Euclid<'an displace111c11t.s. This result. and th<' fact
that. the analogous groups in hyperbolic geometry arc also Coxelc'r µ;ro11ps 10 ,
has led to the latt.<•r groups being stuclic~d dir<'ctly, initially e111phasising the
g<'o11wtric realisation ((XV). (XXV)). and tlH'll. following .I. Tits (XXV). in
a purdy algebraic fra111<'work.
Starting with tlw work of \Vit.t. thC' t.h<'ory of scmisirnplc· Lie groups and
that of discr<'t<• groups gen<~ra.IPd hy retlPctions conti1111c' to i11teract. <'X t wmcly
fr11it.full.v wi1 h C'ach ot lwr. 111 I !J.I I. Stifel (XVIll) remarks that th<• Wey I
groups are exactly the finite gr< ups w~rwrated by r<'fi1~ctions that leave a
latti('C' i11vmia11t. Cl1cvall<'y (XIX a)) a11C\ Harish-Chandra (XX 11.)) giw. in
1918-G 1, a priori. proofs of the bi cctivt' co1Tcsp0111lc'nc(• lwtwP1'1t "(T.vstallo-
graphi('" µ;roups and complex scm sim pk Lie nlgchrn.s: unt ii t.hcn. it had only
b1~e11 possible- to Vl'rify this cmTcs 011de11cc sC'paratcl.\' for each l,Y[W of simple
Lie algebra.
10 Thcsc~
groups, studit'd i11 depth i11 qtw c.:ase of dinwnsiou 2, have so far 0111.v been
considered incidentally in dimcnsiol1s :;,, :l.
HISTOHICAL NOTE
On the other ha11d. it is ohserwd aro1111d 1%0 that t.hP polynomials in-
variant 11mlN tlw WPyl group play au importa.ut. role in two an'as, the tlwory
of infinite-dimensional linear rcprescutatiow; (XX n)) mid i11 the topology of
Li<' groups. Coxd<'l' (XIV .f)) agai11 takes up the study of thl.' transformation
C formed by taking the product. of the fum!ammtal reflections of a finite
group \V 1?;<'11eratcd hy reficctious. HP observes (by lll('alls of a separate ex-
amination of each type) that. the algebra of poly11omials invariant under W
is generated hy algebraically imkpendcnt ell'rnents whos<' <fogrecs arc related
in a simple way to I.he eigenvalues of C. A priori proofs of these rPsults were
givc11 by Chevalley (XfX b)) in the fi.r:.;t area. and by Coleman (XXHI) and
Steinberg (XXfV) in the second.
\Vith the work of A. Borel on lirwar algebraic groups (XXJI), new devel-
opments begi11 which are to lead to a not.able enlargement of the thl'ory of
Lie groups. A. Borel ernphasises the importance of the maximal connected
soluble subgroups (since called "Borel subgroups") of a Lie group, and makes
them the principal tool for transporting a large part of the classical thl'ory to
algebraic groups owr an algebraically closed field (but wit.ho11t obtaining a
dassilicat.ion of the simple algebraic groups 11 ). The T3on'l subgroups (in the
case of real or complex classical gro11ps) had already arisen some y('ars earlier
i11 the work of Gelfand am! ~cumark on inifinitc-dirncnsional reprcst-ntations;
and in HIG,1. F. Bruhat had disC'owred the remarkable fact that, for the clas-
sical simple groups, the decomposition of the group into double cosets over a
Borel subgroup is imkxcd in a cano11i<:al way by the \Vey! group (XXI). This
res11lt. was subsequently ext.<'l1ded to all real and complex semisirnplc groups
by Barish-Chandra (XX b)). On the other hand, in 19G5. Chcvallcy (XIX c))
had succeeded in as::;ociating to every complex semisimple Lie algebra g and
to every commutative fi<'ld k. a group of matrices with coefficients in k hav-
ing a Bruhat. decomposition; and he used this last fact to show, with a small
rn1111lwr of except.ions, that the group thus defined was simple (in the sense of
the theory of abstract. groups). He thus "explained" the coinci<knce, already
11oticed by .Jordan and Lie, between the sirnpl() LiP groups (in the sense of
the theory of Lie groups) of typP A. B. C. D and the dassical simple groups
<kfined in a purely algebraic 1rnu11ier over an arbitrary field (a coincidence
which had until then only been externlcd to the excq>tional types G 2 and F 4
by Dickson (XI Ii) and c))). In particular. hy ta.king a finift: field k. the con-
struct ion of Chevalley provi<kd. for each type of complex simple Lie algebra,
a family of finite simple groups, contai11ing a large number of the then known
finite simple groups as well as three new series (corresponding to the simple
Lii~ algPbrns of types F 4 , E 7 and E 8 ). A little lat.er. by various methods, and
using modifications of those of Chcvallcy, several authors (Herzig. Suzuki,
11 An algebraic group of dimension > 0 is said to be sim.pli-: (i11 t.he se11se of algebraic
µ;eonu.'t.ry) if it docs not contain any 11onnal algebraic subgroup of <lirne11sion > 0
ot.her than it.self. It is said to be semi8implP if it is isogcnous to a product of
simple non-ahclian groups.
254 lllSTOHICAL NOTE
Rec. Steinberg and Tits) showed 011 the one hand that the other finite simple
groups k11own at the time could be obt.ai11ed in au analogous way, with the
exception of the alternating groups and the Mathieu groups, a.11d on the other
hand constrnctcd other series of new fi11ite simple groups (cf. (XXIX)).
At about the same time. Chcvalley (XIX d)) again took up the study of
linear algebraic groups and showed, 11sing the technique of Bruhat dccorn-
positio11s together with a key rcs11lt on the 11ormaliscr of a Borel subgroup,
that the theory of scmisimple linear algebraic groups over an a(qt>.braically
clos<'d field k of arbitrary characteristic 12 leads to essc11tially the same t.vpcs
as in the Killing-Cartan classification fork= C. After this, .J. Tits (XXV a)
and b)). analysing Chevalky's methods. is led to an axiomatised version of
th<' l3ruhat decomposition (the "B~-pairs"'), in a remarkably versatile form
which involves only the group strncturc: it is this notion that is now known
as H "Tits system''. All the simple groups (with the various meanings of the
term) we have discussed above are canonically equipped with Tits systems,
and Tits himself (XXV c)) has prove<l that the existrncc of such a system
in an abstract group G, togPlher with a few additional properties from pure
group theory, allows one to prove that G is simple, a theorem which covers
the majority oft he proofs given until then for these groups (cf. Chap. IV. § 2,
no. 6). On the other hand, in collaboration with A. Borel. he has genernlisPd
the results of Chcvallcy in (XIX d)) b.v showing the existence of Tits systems
in the group of rational points of a sc111isimple linear algebraic group over an
arbitrary field (XXVII).
All the Tits systems cncountcwcl in t hesl' q1wstions have a finite WPyl
group. Another category of examples was discovered by Iwahori and l\fat-
sumoto (XXVI); they have shown that. if, in Chcvall<·y's construction of (XIX
c)). k is a p-adic field. the group obtaimxl hns a Tits system whose Wey! group
is the affine Wcyl group of the <·0111plex semisimpk Li(' algebra with whkh
one started. This result has been extended by Bruhat and Tits (XXVIII) to
all scmisimple algebraic groups over a local fidd.
12 Thcexi:;tcnce of numerous "pathologie<iJ" simple Lie algebras over a field of char-

acteristic p > 0 could have led some/to doubt the universal character of the
Killing-Cartan da~sification.
BIBLIOGRAPHY
(I) J. HESSEL. Krysallomctrie oder Krystal/anomie und Krystallo_qraphie

( 1830. rcpr. in Ostwald's Klassiker, nos. 88 and 89. Leipzig (Te11bner),
18!)7)
(Il) A. Bravais, Memoires s11r lt>s polyMres de forrne syrnetriquc. Journal
de Math., (1 ), v. XIV ( 1849). p. 111-180.
(III) A. l\'IOB!lJS: a) Ueb<'r das Geset.z dn Symmetric der Krystalle und die
Anwendung dicses Gesetze auf die Eint.hcilung der Krystalle in Systcme,
J. fiir die reine tmd anqewanrlte Mnth., v. XLIII (1852). p. 36ii-:n4 (==
Gesammelte Wcrke. v. IL Leipzig (llirzf'I). 1886, p. :M9<360); Ii) Theorie
der symrnet.rische11 Figuren. Gesammeltr: Werke, v. II, Leipzig (Hirzel),
1886, p. 561- 708.
(IV) L. SCHLAFLI. Theori<' der vidfachcn Konti1111itiit. D1,nkschr. der Schw-
ciz. naturforsch. Gesr.llsduift. v. XXXVIll (1901), p. l-2:H (=- Ges.
math. Alihandl'Ungen, v. I, Basel (Birkhkiuser). 1950, p. Hi7-:J87).
(V) W. R. HAMILT0'.'11, :\lf'morandu111 respecting a nPw syst.l•rn of roots of
unity, Phil. Mag .. (1 ), v. XII ( 1856), p. 4,16.
(VI) C. JORDAN, Memoire sur les groupes de mouvcmC'nts. Ann. di Mat., v.
Il (1868-69), p. 167-215 and :!22-:M5 (=Oeuvres, v. IV. Paris (Cn11thi<'r-
Villars), 1964, p. 231-:~02).
(VII) E. COuRSAT, Sur les substitutions orthogouales ct. ks divisions regul-
ii>rPs de l'cspacc, Ann. Re. Norm. Sup .. (:l), v. VI (1889), p. !J-102.
(VIII) W. KILLI~G. Die ZusammC'nsl•tzung dcr stctigcn cndlichen Transfor-
mationsgruppeu: I) Math. Ann .. v. XXXI (1888), p. 2.52-290; II) ibid.,
v. xxxm (188!'.l), p. 1-48: III) ibid., v. xxxrv (1889), p. s1-122; IV)
i/1id., v. XXXVI (1890). p. 161-189.
(IX) E. CARTAN: a) Sur la st.ructme des groupes de transformations finis
et cont.inns (Thesis). Paris (Nony), 1894 ( = Oev.vm~ complktes, Paris
(Gauthier-Villars), 1952, v. I I. p. 137-287); b) Sur la reduction a sa forme
canonique de la structure d'11n groupe de transformations Fini ct cont.inn,
Amer. Journ. of Math., v. XVIII (1896). p. 1-46 (== Oeuvrr.s complCfFs,
Paris (Gauthicr-Villars), 1952, V. rI· p. 29:J-:353): c) Le principe de dualite
et la theorie des groupes simplPs et semi-simples, Rull. Sci. Math. .. v.
XLIX (1925), p. :J61-374 (= Oeuvres comp/Ries. Paris (Gauthier-Villars),
1%2, v. 11 , p. 555-568); d) La geometric des groupes simples. Ann. di
Mat., (11), v. IV ( 1927), p. 209-256 ( = Oeuvres c.ompl.Ctes. Paris (Gauthier-
Villars). 1952, v. 12, p. 7!):3-840); c) Sur certaincs forrnes rip111anniennes
rernarq1mbles des geornet.ries a groupe fondamental simple. Ann. Ee.
Norm. Sup., (~l), v. XLIV ( l!.127). p. :Hfi-467 ( = Oeuvres compli.!tcs, Paris
(Gauthier-Villars). 1952, v. h, p. 867-989): /) Cornplcment. au mt•moire
"Sur la geometric des groupes simples", Ann. di Mnt .. v, V (1!:128), p.
25G BIBLIOCRAPHY
253-WO ( = Oe-uvre8 comp/1•tes, Paris (Gauthier-Villars), 1952, v. '2, p.

lO!l:J-1010).
(X) R. FRICKE and F. KLEIN, Tlworie df:r rmtomorplwn Funktionen,
L<>ipzig (Teubner), 1897.
(XI) L. E. DICKSON: a) Theory of linear groups in an arbitrary field, Trans.
Amer. Math. Soc .. v. II (1901). p. 36:3-394: b) A 11ew system of simple
groups. Math. Ann., t. LX (1905), p. 137-150: c) A class of groups in
an arbitrary r<>alrn con11ectcd with the configuration of the 27 lines on
a cubic smface. Quart. Jouni. of Math .. v. XXXIII ( 1901 ), p. 145-173
and v. XXXIX (l!J08). p. 205-209.
(XII) A. SPEISER. Thcorie der Gruppen von endlii:hP.T Ordnuny. Berlin (Sprin-
ger), 1924.
(XIII) H. WEYL. Theorie der Darstellung kontinuerlicher halb-einfacher Grup-
pcn durch lineare Transformat.ioncn, Math. Zeitschr., v. XXIII (1925),
p. 271-309, v. XXIV (1926), p. 328<195 and 789-791 (= Selecta, Bascl-
Stutt.gart (13irkhiiuser), 1956, p. 262-366).
(XIV) H. S. l\'1. COXETER: a) Groups whose fundamental regions arc sim-
plexes, Jonrn. Lond. Math. Soc., v. VI (1931), p. 1:{2-1:{6; b) The poly-
topes with regular prismatic fiF;ures, Proc. Lonrl. Math. Soc., (2), v.
XXXIV (1932). p. 126-189: c) Discrete groups generated by reflections,
Ann. of Math., (2), v. XXXV (1934), p. 588-621: d) The complete enu-
meration of tinitc groups of the form Rf = (R .. R; )k'J = I, Journ.
Lond. Math. Soc:., v. X (19:35), p. 21-25; e) Regular polytopes, New York
(!vlacmillan), 1948 (2nd ed., 1963); f) The product of generators of a fi-
nite group generated by reflections, Duke Math. Joum., v. XVIII ( 1951),
p. 765-782.
(XIV bis) II. S. M. COXETER and IL WEYL, The 8tnu:turc and rrpresentations
of contin'Uons groups (Inst. for Adv. Study, mimeographed notes by N .
.Jacobson and R. Brauer, 1934-35): Appendix.
(XV) H. S. 1\1. COXETER. and W. 0 . .I. MOSER, Generators a.nd relations for
di8cretc: groups. Ergcb. der Math., Neuc Folge, Bd. 14, Berlin (Springer),
1957 (2nd ed., 1963).
(XVI) B. L. van der WAERDEN, Die Klassification <ler einfachen Lieschen
Gruppen, Math. Zeit.w:hr., v. XXXVII (1933). p. 44fi-462.
(XVII) E. WITT, Spiegelungsgmppen und Aufziihlung halbeinfacher Lieschen
Ringe, Abh. Math. Sem. Hamb. Univ., v. XIV (UMl), p. 289-322.
(XVIII) E. STIFEL, Ueber einP Beziehung zwischen geschlossenen Lie'sche Grup-
pen 11nd di:skontinuierlichen Bewegungsgruppcn euklidischer Riiume und
ihre Anwcnd1111g auf die Aufzahlung der cinfachen Lie'schen Gruppen,
Comm. Math. Hclv., v. XIV (1941-42), p. 350-380.
(XIX) C. CHEV ALLEY: a.) Sur la classification des algebres de Lie simples et
de leurs representations, C. R. Acarl. Sci .. v. CCXXVII (1948), p. 1136-
1138; b) Invariants of fin it. groups generated by reflections, Arner. Journ.
of Math., v. LXXVII (195 ), p. 778-782; c) Sur certains groupes simples,
Tohoku Math. Jo·urn., (2), v. VII (1!)55), p. 14-66; d) Classification des
groupes de Lie algebrique., 2 vols., Paris (Inst. H. Poincare), 1956-58.
(XX) HARISII CHANDRA: a) n some applications of the universal envelop-
ing algebra of a semi-sim le Lie algebra, Trans. Amer. Math. Soc .. v.
LXX (Hl5l), p. 28-96; b) n a lemma of Bruhat, .!own. de /lfoth., (9),
v. xxxv (1956), p. 20:l- 10.
(XXI) F. BR.UHAT. Representat'ons induites <les groupes de Lie semi-simples
complexes, C. R. Acad. Si., v. CCXXXVIII (1954), p. 437--139.
0113LIOGRAPllY 257
(XXJI) /\.. f30REL. Gro11pes lincaircs alp;0briq11es, Ann. of Malh., (2), v. LXTV
(1%6). p. 20-80.
(XXlll) A..J. COLEl\IAi'J, The Betti 1111111bers of the simple grn11ps, Can . .lourn..
of Math .. v. X (1!>58). p. :w>-:!5G.
(XXIV) R. STEINBEHG. Fiuit0 reflection groups. Trans. Amer. Math. Soc .. v.
X Cl (I !J59). p. 19:3-50·1.
(XXV) .I. TITS: a) Groupcs sirnpks et µ/0111t;tries associces. Prnc. Int. Congress
Math .. StockholnL 1962. p. 1!>7-221: Ii) Theori~me de Bruhat et sous-
groupt>s parabolique>s. C. R. Arn.d. Sci., v. CCLIV ( l!J62), p. 2910-2!)12;
c) Algebraic aud abstract si111pl0 g;ro11ps. Ann. of Maf,h., (2). v. LXXX
(1964), p. 313-329.
(XXVI) :'\I. IWAHORI and H. l\IATSC!\IOTO, On some f3ruhat decomposition
and the structure of the Hecke rings of p-adic Chevalley groups, 1'11/Jl.
Math. l.H.E.8 .. uo. 25 (19G5). p. 5-48.
(XXV!I) A. BOREL aud .!. TITS. Gro11pes reduct.ifs. Puhl. Math. I. H. E.S.. no.
27 ( l!JG5), p. 55-1.50.
(XXVIll) F. BlllJH/\.T and .I. TITS. Groupes algcbriques simples sur un corps
local, Proc. Conj. on loml Fidd8. p. 2:3-:36. Berlin (Springer), 1!)67.
(XXIX) H. CARTER, Simpl0 groups anrl sirnplP Lit' algebra~ • .!mmt. Dond. Mo.th.
So<:.,\'. XL (1965), p. rn:3-2,IO.
INDEX OF NOTATION
Tlw r<>for<'lw<' numbers indicate the drnptcr. paragraph and number. rcsp<'c-
tively.
A1 (root syst<'llJ of t.yp<'): V L .t, 1: VI. '1. 7 a11d Plate I.
A,: VI. 4. :t
A[P]: VI. :3, 1.
/\(R): VI. L 1.
ii (high<>st root): VI. I. 8: VI. I, 3.
no= -ii: vr. ,1, :3.
(n 1 . . . . . ni): VI. 1. 5.
B: IV. 2. l.
13(C) (basis d<'fitwd hy th<• chamh<'r C): VL 1. 5.
13i: (root syst<'lll of type): VI. 4. l: VI. .t. .')and Plate Tl.
13,: vr. 4. :t
131- 1 (bilinear form assodated to tll<' Coxct<'r matrix :\I): V. 4. 1.
C (d111111hcr): V. I. :3: VI. I, G.
1· (Coxct<'r transformation): V. (), 1: VI, I, 11.
C1 (root systC'111 of t,vp<'): VI. -I. I: VT. -1. G aml Plate Ill.
(~,: VT. -L :t
/;. /(:: VI. 2. :3.
1(B): VI. 1, 12.
ii= n
n>O
(c"l 2 - e-"1 2 ): vr. :3. :t
D1 (root syst<'lll of type): VI. .1. I: VT. -1. 8 and Plat<' IV.
i'>i: VI. 4. :3.
E: VI. -l. -1.
Eb. E 7 . E 8 (root syst<~m of type): VI. 1. 7: VI. ,J. 10: VI. 4. 11: VI. 4. 12 nncl
Plates V. VI. VII.
- - -
En. E7. EH: VI. t. :1 .
.0 I ..... E 11: \i"l. 4. -1.
re, (root syste111 of t_vpe): VJ. 4. l: VJ. ·1. !) HIJ<I Plat<' VHI.
F.1: VI. L :1.
G: VI. 2. :1.
G2 (root syst<'111 of't.\'Jw): VT. -1. l: VI. 4. 1:3 a11C! PIHi<· IX.
{;2: VI. •l, :3.
fJ: v. :3. I.
2fi() l:"<DEX OF NOTATION
h (Coxder m1mh1~r): V, 6, l; VI, 1, 11.

H3. lt1 (CoxctC'r systems of type): VI, 1, 1.
I2(P) (Coxct1•r system of typ<'): VL 4, 1.
.l(eP) = L det(u:)l'."'(l'l: VI. 3. :1.
wE\V
Lo. L1. L2. L:i (lattices in R"): VI. 4. 4.
I ( 111), ls ( 11') ( lcugt It of an element w): IV, l, 1.
rn(8, 81 ): IV, 1, 9.
N: IV. 2, 1.
P(R): Vl, 1. 9.
Q(R): VL 1. 9.
R (root system): VI. 1, 1.
W: VI, l, 1.
fl = 4 L n: VI. 1. 10: VI, 3, 3.
">0
S: IV, l. l.
S(cP) = L ,,q: VI. :3. 4.
</E \\l .p
S,,,: IV, L 8.
T=I3n~: IV. 2, l.
V: VI, l, 1: VI. 4, 4.
W = N/T: IV, 2, 1.
W: V, 3. l.
wo: VI. 1, 6.
W(R): VI, l, l.
w+(R): VI, ·1. Exercises.
Wx: IV. 1, 8.
<h: VI. I, 12.
(w 1 , •••• w1 ): VI. 1. 10.
INDEX OF TERMINOLOGY
The rc>forcnce n11r11t)('rs iurlicat.0 the C'hnptcr. paragraph and numh<'r (or. 0x-
c<'ptio11ally. <·xercisc). r<'sped i \'Ply.
Adjoining drnmbcrs: IV. l. Exercis<' 15.

Affine (\11/pyl group): VI. 2. I.
Akovc: VI, 2, 1.
Angle h<'twren two roots: VI. I, 2.
Anti-invariant: V. 5, 4 and VL 3, 3.
Apartment: IV. 1. ExcrcisP 15.
Arrow: IV. Appendix, l.
Associated (bilirwar form to a Cox<'lcr system): V. 11, 1.
Basis of a root system: VI, I. 5.
Building: IV, L Ex<•rcisc 15.
Canonical bilinear forru: VI. 1. 12.
Canonical Cartan matrix: VL 1. 5.
Cart.an matrix: VI, I, 5.
C'haiu: IV, Appendix. 3.
Chaiu of roots: VI. 1, 3.
Chamber: V. 1. ;3 and V. 3. 1.
Chamlwr of a building: IV, I, Exrrcisc 15.
Characteristic dcgn't's: V, 5, I.
Circuit: IV, App<•tulix. 3.
Closed set. of roots: VI. 1. 7.
Compact hyperbolic (Coxr.tr.r group of · typP): V. 1. Exercise 12.
Co111p0t1<•11ts (rnum·ct.<'d of a grnph): IV. Apr><'ndix. 2.
Components (irreducible of a Coxel<'r syste111): IV. 1. !J.
Cn11j11gat<' elements of a group: IV, I. :3.
Connected graph: IV. Appendix. 2.
Co1111e<·tion index: VI, I. !J.
Coutragrcdi<·nt. rl'pn•s<~ntat ion: V ..1. -1.
Cos<'ts (doubh•): IV, 2, 1.
Coxdcr graph: IV. l, 9.
Coxet.cr group: IV, l. :3.
Coxet<ff matrix: IV. I, !J.
:w2 !\'DEX OF n:H!\llf\OLOCY
Cox<'tt•r n1u11IH'r: V. G, and VL 1, 1 l.

Cox('((•r system: IV, I. :3.
CoxPtcr transformation: V. (L I nud VI. 1. IL
Crystallographic group: VI. 2. :>.
Dil1edrnl group: IV. I. '2.
Do111i11a11t W('ight: VI. 1. 10.
Dynkin grnph: VJ. -1. 2.
Essential group gc1wrat<'d hv rdiections: V, :3. 7.
Exd1a11g<' C'ondit ion: IV, I. :>.
Exponents of a finite Coxctc'l' group: V. G, 2.
Face of a chamlH'r: V. 1, -I.
Facd: V. 1. '2.
FundanH·ntal \V<'ight: VI, 1. 10.
Forest: IV. A ppcnd ix. :3.
Gallery: IV. l. Exercise 15.
Crnph: IV, Appendix. 1.
Grnph (co111pletcd D,vnkin): VI. l. :t
Graph (CoxdPr): IV. I. D.
Graph (Dynkin): VI. l. 2.
Half-space: V. 1. 1.
II PC' kc algcbrn: V. 2. ExetTiS(' 22.
Highest root: VI. 1. ~.
Hyp('J'holiC' (Cox<'t('r group of t.\'IH'): V .. t. Exercise 12.
Hyp<'rpla11c of a pscudo-rdl<'ction: V. 2. I.
Indivisible root: VI. 1. :1.
ltiverse s:vsl<'lll of roots: VI. I. l.
Irred11cible group gctH'rn!l'd by rcAcd.ions: V. :3. 7.
Irr<'d11cibl(' Coxl'tcr ,;_vst<'t11: IV. I. !J.
Irred11cible root s,vst(•111: VI. I. 2 .
.Join<~d V<'rtices of a graph: IV. Aprwndix, l.
L(•nglh of a path i11 a graph: IV. Appendix. 2.
Length of an el<•mcnt of a group: JV. 1. 1.
Length of a root: VI. 1. 2 .
.\li1111srnl(' weight: VJ. l. Exercise 2!.
l\ornrnlisl't': IV. 2. G.
Nomwd graph: VL 4. 2.
Opposed fac!'ts: IV. I. ExNcisc 18.
Path: IV. App<'ndix, 2.
Poincar0 series: V. G. I.
Polyno1ninl (graded algdm1): V. ~1 1.
Positive rnot: VI. l. (i.
Pn•s<•ntatioll ut' a group: IV. l. :l.
Psc11do-rdi<·ctirn1: V. 2, 1.
Hadical \\'(~ights: VI. I. !J.
11\DEX OF TEBl\II:\OLOGY
Ha111ification point of a graph: rv. Apprndix. I.

nauk of H mot S,\"Sl:<'lll: Vl. I. I.
H<·d11c<'d c[('('Olllposit iou: rv' 1. I.
RedU('<'d root systern: VI. I. -I.
Rdll'dion: V. 2. 2.
11.<'J>r<.'Sl'JLtatiou <lSSOciat<'d to H lill'tllll matrix: V. 4. 3.
11<'stridcd direct product: IV. I. !J.
Root sysl<•rn: VI. L I.
Sat mat <'d sd of weights: VI. I. Ex<'tTisl' 2:3.
Sign of an de111<'nt of a C'oxct<'r group: IV, 1. 3.
Simply-tnutsitivc group action: IV. 2. Ex<•rds<' :J.
Simplex: V. I. G.
Sirnplicial cor1<•: V. l. G.
Spacious building: JV. 1. Ex<'rcis<' 24.
Special point: V, :~. 10.
Strongly orthogonal roots: VL 1. :3.
Strnctured building: IV. 1. Ext'rrise 24.
S11bgrnph: IV. Appendix. 1.
Support of a facet: V. I, 2.
Terminal verlf'X of a grnph: IV, App0ndix. l.
Tits subgroup: IV. 2. ExPrcis<' :~.
Tits" theorl'lll: V. 4. 4.
Tre<•: IV. Ap1w11dix. :!.
Ve('tor of a ps<•11<lo-rc•flpct.ion: V. 2. 1.
V<•J't<'X of a graph: IV, Appendix. l.
\\'all of a d1atnbcr: V. I, . 1,
Weights: VI, l. 9.
Wcyl group of a root s.vst<'m: VI. 1, l.
\\"eyl group of a Tits syst<'llt: IV. 2. 1.
PLATE I
SYSTE1\1S OF TYPE A1 (I ~ l)
(I) V is the hyperplane of E = Rl+t consisting of the points the sum of

whose roordinates is zero.
Roots: E; - Ej (i -f-j, 1~i~l+L1~j~l+1).
Number of roots: n = l(l + 1).
(11) Basis: 01 = E1 - E2.<12 = E2 - E:i··· .. 01 = E1 - Et+I·
Positive roots: E; - EJ = L Ok (1 ~ i < .i ~ l + 1).
. i~k<j
(III) CoxPter number: h = l + 1.

(IV) Highest root: ii·= Et - Et+i = o 1 + Q2 + · · · + n1 =Wt + w1.
Completed Dy11ki11 graph (l ~ 2):
For l = J. the Coxetcr graph of th<' affine' \Vey] group is
00
0--0
(V) W = n,
<f> ( . ) - (.rjy) 1(R) = (! + 1) 2 •
R .T • .I} - 2(1 + I)'
(VI) F11ndamP11t al wt>ights:
i l+ I
W; =CJ + ... + E; - --
l+l.i=l
LE7
l
- -[(l
= 1 - i + l)(rY 1 + 2n 2 + · · · + (i - l)n;_i)
+ l
+ i((l - i + l)o; + (l - i)n;+1 + · · · + n,)].
2(i(j PLATE I
(VII) Sum of the positive roots:
2p =lei+ (l - 2)::2 + (l - 4)€:i + · · · - (l - 2}c1 - IEt+I

=lot + 2(l - I )02 + · · · + i(l - i + l)<v; + · · · + lcr1.
(VIII) Q(R): the set of we-tors whosl' C'(>0nli11ates arc integers with sum zero.
P(R): gc11eratl'd by Q(R) and E1 - (I+ 1)- 1 (E1 + E2 + · · · + E1+1).
P(R)/Q(R) is isomorphic to Z/(l + l)Z.
Conncetion index: l + I.
(IX) Expouents: 1. 2..... I.
(X) W(J1) = 61+ 1 , identified with the group of permutations of th<' E;.
Order of W(Il): (l + l)!
(XI) I = 1: A(R) = W(R): 11• 0 = -1.
l ~ 2: A(R) = W(R) x {I. -1} and 11 10 transforms er; t.o -n1+I-i·
(XII) The group P(W)/Q(W) is cyclic of oHkr (l + 1): it acts on the com-
ph~tcd Dynkin graph by circular permutations. If l ~ 2. the unique
11011-identity clc11w11t. of A (11) /\V(R) acts on P(R) /Q(R) by the auto-
morphism ;i· ,__. - J'.
(XIII) Cartan matrix (/ x /):
2 - l 0 () 0 0
- I 2 -1 0 0 0
() -1 2 -1 () ()
() () -1 2 0 0
() () () 0 -I 2
PLATE II
SYSTEl\JS OF TYPE B 1 (I ? 2)
(I) v = E = R 1.
Roots: ±E;, (1 ~ i ~ l), ±E; ±E.1 (1 ~ i <j ~ l).
N urnb<'r of roots: n = '21 2 .
(IT) Basis:n1 =E1 -E2.n2=E2 -E:i ..... n,_1 =E1-1-E1.a1=E1.
PositivP roots:
E, = L Ok (1 ~ i ~ l),
iO·(I
E; - E1
.
= L
1(J.-<.1
ni. (l ~ i < j ~I),
E; + E1 = L
1(k<J
o, + '2. L
.1( A·(I
nk (1 ~ i < j ~ l),
(III) CoxC'ter nurnlwr: h = 21.

(IV) Highest root:
ii-= Et + Ez = 01 + 202 + 2n:i + · · · + 2a1.

We have r1' = 2w2 if l = 2. 6 = w2 if I ? :3.
Completed D,vnkin graph:
for l = 2:
(V) W is the set of vP«tors
±2€ 1 (1 ~ i ~ /). ±E, ± E1 (1 ~ i <j ~ /).
(.r /y)
<l>B(.i:.y) = - 1- . 1(H) =(I+ 1)(41- 2).
4. - 2
268 PLATE II
(VI) F11ndal!l('lltal weights:
W1 = c\ + E2 (1 ~ i < /)
+ ··· + E;
·~n 1 +2n 2 +···+(i- l)n;_1 +i(n;+n;+1 +···+ni)

l
W/ = - (c1 + E2 +···+EL)
2
1
= :,z(n 1 + 2o 2 +···+In:,).
2p = (21- l)E1 + (21- ;~)c2 + ··· + :lc/--1 +Et

=- (21 - l)n 1 + 2('2/ - 2)n2 + · · · + i(21 - i)n; +- · · · + 12 cq.
I
(VIII) Q(R) = EB'.= 1 Zr:;, P(H) = EB'.= 1 Zc; + ZO LE,).
l=l
P(R)/Q(H) is isornorphic to Z/2Z, generated hy thr image of w1.
Connection index: 2.
(IX) Exponents: I. :3. :>, ... , 21 - I.
(X) W(H) is thP SPllli-dircct product of the group 61. ading by permu-
tations of the E;. by the gro11p (Z/'2Z) 1• acting by c; ,_.... (±l);E;. Its
order is 21.1!
(XI) ;\ ( R ) =-- \V ( ll): u-u = - l.
(XII) The 11niq1H' nm1-trivial Pl('!IWlll of P(W)/Q(K) defines the unique
non-trivial antornorphism of t lw com plct.('d Dynkin graph.
(XIII) Cartan matrix (Ix I):
2 - l () () 0 0
-I 2 -1 () () ()
() -1 2 -1 () ()
() () -1 2 () 0
() 0 0 0 2 -2
() () () () -1 2
PLATE III
SYSTE~IS OF TYPE C1 (l ~ 2)
(I) V = E = R 1.
Roots: ±2€;. (1 ~ i ~ l). ±E; ± Ej (1 ~ i <j ~ l).
N11111lwr of roots: 11 = 21 2 .
(II) Ila.sis: n1 = E1 - £2. oz = Ez - c:i, ... , n1-1 = Et-1 - c/, n1 = 2c1
Positive roots:
E; = E; -- Cj = . L .Ok (l ~ i < .i ~I),

,,k<J
E; + Ei
.
. L .o~. + 2 j'k<I
= ,,k<J L OA· + n1 (I ~ i < .i ~ I),
2c-; = L nk + nt (1 ~ i ~ l).
i'k
{V) Wis t!tc s<'t of vectors±.::;. ±€; ± Ej.
(;rl.11)
<Pn{.r.y) = 4 (l+ l)' ;(R) = (l+ l)(-ll--2).
(VI) Funda1tH'ntal W<'ight s:
0.J; =Et + E>2 + · · · + E; (J ~ i ~ /)

=n 1 +2n'2+···+{i- l)n;-1
1
+ i(n; + n;+i + ·· · + n,_, + 2n1).
270 PLATE 111
(VII) Sum of tl1c positiw roots:
2(! = 2/c1 + (2/ - 2k2 + · · · + 4Et-l + 2Et

=2/n 1 +2(21- J)n·2 +···+i(2/-i+l)ni+···
l
· · · + (f - 1)(/ + 2)01-t + 21(1 + l)n1.
(VIII) Q(R): tilt' set of poi11ts whos<> coordinates are intrgers with <'V<'ll sum.
P(R) = EB'.= l Zt:;.
P(R)/Q(R) is isolllorphic to Z/2Z. gcnPrnt.cd by the image of w 1 .
Com1<•ction indPx: 2.
(IX) Expu11c11ts: 1. :.LS ..... 21 - 1.
(X) \\'(R) is the scrni-din'cl product of thr gro11p 61. acting by pcrm11-
tntio11s of the E,. Irv the group (Z/2Z( acting by E; ........ (±1),E;. It.s
order is 2 1./!
(XI) A(R) = W(H): ll'rJ = -1.
(XII) The lllli(jlH' non-trivial <'lPment of P(W)/Q(W) <kfinf's thr unique
non-trivial automorphism oft llf' rnmpktcd Dynkin graph.
(XIIJ) Cartan matrix (l x I):
2 - l () () 0 0
-] 2 -1 () 0 0
() -1 2 -1 () ()
() {) -1 2 0 ()
() () () () 2 -1
() () () () -2 2
PLATE IV
SYSTEi\IS Of TYPE D1 (1 ? 3)
(I) V=E=R1•
Roots: ±c; ± EJ (1 ~ i < .i ~ !): (E,) t.}](' canonical basis of R 1.
:\'11111lwr of roots: 11 = '2/(l - I).
(II) Basis: n1 = E1 - Ez. n2 = <:1 - E:i ..... n1-1 =Et- 1 - !:1. n1 = E1-1 +Et.
Posit.ivc roots:
E; - EI L 111.-
= 1<k<j (J ~ i < .i ~ I).
.
E; + E1 = L Ok + n1 (I ~ i < l),
i~/..~/--2
E;+Ei= Lnk+'2 L rrk+ll'1-1+n1 (l~i<.i<l).

. i~/..·<J .1~k<I- J
(III) Cuxd<'r 11111uhcr: h = 21 - 2.

(IV) Hiµ;lwst root:
ii= E1 + E2 = n1 + 2r1'2 + · · · + 2a1-2 + n1-1 + 01.

WP have r~ = ..v 2 + w:i if l =
;~ and ii = ..v 2 if I ? •1.
Cornplt't<'d D.rnki11 grapl1 (I? ~):
(V) W =IL
(:r/v)
n (:t. y) = -l(l _ l) 1(R) = 4(1 - I ) 2 .
27'2 !'LATE IV
(VI) F111u\amm1tal weights:
W, =Et+ E2 + · · · + E;. (1 ~ i ~ f - 2)
= n-1 + 2n2 + · · · + (i - l)n; -1 + i(n1+n;.f1 + · · · + 01-J
l .
+ 21(01-1 + <lt)
l
W/-1 = 2(€1 +E2 + ··· + E/-2 +Et-I - Et)
l 1 1
=-(01 + 202 + ... + (l- 2)01-2 + -lCl't-1 + -(l - 2)<ti)
2 2 2
1
Wt= 2(€1 +c2 + ... +Et-2 +c1.1 +E1)
1 1 l
= 2(01 +202 + ·-· + (1-2)01-2 + 2(/- 2)n1-1+ 21n,).
(VII) S11111 of tlw posit iv<' roots:
2p = 2(/ - l)c 1 +2(1- 2)5:2 + · ·· + 2c1-1

= 2(1 - I )01 + 2(2/ - :3)02 + · · · + '2(il -
i(i + 1)) O; + · · ·
2
l(I - l)
· · · + - -2-(01-1 + ni).
(VIII) Q(R): th<' set. of points whose coordiuntcs IH"C int<'gcrs with <'V<'ll sum.
t I I.
P ( R) = ffi ZE; + Z ( - L E, ) .
w
i=l
'2 i=l
l odd: P(B)/Q(R) is iso111orphiC' t.o Zj4Z. gcncrat<'d hy tlw imag<' of

wt: W<' have ...v 1 = 2w1 and wt_ 1 = :3w1 mod. Q( H).
I ev<'n: P(R)/Q(H) is isomorphic to (Z/2Z) x (Z/2Z): t.lw thm' ele-
ments of order 2 an' till' imap;ps of w 1 , w1_ 1 aud w1.
C'OillH'd i011 illliPX: 4.
(IX) Expo11c11ts: I.:~. :>, .... 2/ - i>. 2/ - :U - l (the last. 11 ppcaring twin' if
l is <'V<'ll and om·1• if l is odd).
(X) \\'(R) is the scmi-dir('ct pn duct of th<' group 6 1• aC'ting by pcrm11ta-
tio11s of lhe E;. hy the grou (Z/2Z) 1 -- 1 . acting by E; ,__. (±l);E; with
0(± l); = 1. Its order is 21 1 .l!
I
(XI) l fc 4: A(H.)/W(Il) = Z/2Z ·tcting 011 the Dy11ki11 graph hy transposi-

tion of th(' nodes 01-1 and ~/.
I = 4: A(R)/\V(R) = 6:$· a :ting 011 th(' Dynki11 graph by permuting
t lil' 11od1·s n 1 . o;i am\ 0.1.
11 10 = -1 if l hi t>\'t'n: 11 10 =
\
-r
if l is odd. where E is the automorphism
t.hat iuterchanges n1.--J and n1 m1<l ll'aV<'s th1• othn o; fixed.
PLATE IV
(XII) Action of P(U-)/Q(W) = P(R)/Q(R) on t.)10 rnmpld.<•d Dy11ki11 µ;raph:

l odd: w/ tra11sf'on11s nu iuto n1. n1 into n 1. o 1 into n,_ 1 and n,_ 1 into
o 0 : it inl.('ffha11g('s n.i and n1-.1 for 2 ~ j ~I - 2.
I even: '-'-'I (n•sp. w1_ 1) i11tPrd1a11gl'S nu and n1 (resp. n 0 and n 1_ 1),
n 1 arnl o 1_ 1 (resp. n 1 and ni) awl intl'rchangl'S o; and n 1_J for
2 ~ j ~I - 2.
(XIII) Cartan uwtrix (Ix/):
2 -1 [) () () ()
-1 2 0 () (} ()
() () 2 -1 () ()
() () -1 2 -] -1
() 0 0 -1 2 ()
0 () () -] () 2
PLATE V
SYSTEi\f OF TYPE E6
(I) V is the subspace of E = R 8 comdsting oft hC' points whos<' coordinates

(f:;) are such that f:r; = 6 = -f:8·
Roots: ±E; ± EJ (1 ~ i < j ~ 5).
5 ~
±~(E8 - E7 - E(i + i~(-l)''(i)c;) with;~ v(i) ('Ven.
Number of mob: r1 = 7'2.

(II) Basis: n1 = ~(E1 +EH)-~(E2+E:1+E.1+.::,+Er;+E1). 02 =Et +E2.n:i =
E·2 - E).fl'I = E:l - E2 . n.~ = E4 - E:1.nr; =Er; - f4.
Positive roots: ±E; + E.i (1 ~ i <j ~ 5).
~(EH - E7 - Er,+ ;ti 5

(-l)''(i)c;) with
5
1~ v(i) even.
Positive roots with at least one CO<'fficicut. ?: 2 (we denote the root
• ) 1 (ICdt'f)n
<U:l'1 + f>0:2 +CO:! + (I 04 + l'll5 +.I On l).Y 1,. ':
01210 11210 01211 12210 11211 01221

1 1
12211 11221 12221 12321 12321
l 1 2
(III) Cox<'tcr number: It = 12.
(IV) Highest root.:
<} = ~(€1 + E2 + E;1 + E,1 + E5 - E6 - E1 + E8)

= n 1 + 2n2 + 2n:J + 3o 4 + 2n,, + 11'() = w2.
i:i·nw other positive roots can be obtained h.v applying Cor. 3 of Prop. 19 of Chap.
VL § 1. no. tl.
27ti PLATE V
Completed Dy11ki11 graph:
IXz
(V) R = W,
<Pn(;r,y) = (xl.11)/24, 1(R) = 144.
(VI) F11nda11u'11tal \Wights:
2 l
Wt = 3(ex - !:7 - .:o) = 3(4n1 + :~n2 + 5o:i +on_, + In;;+ 2nh)
l
W2 = 2(E1 +=:-2 +c:1 +.:1 +Er, -C:(i- E1 +Ex)
= n1 + 202 + 20:1 + :~n.1 + 2n;, + <t(i = <°i!

5 1
W:l = G(Ex - E1 - Eii) + 2(-Et + E2 + E;3 + E.1 + E;,)
1
= 3(501 + 602 + 10o:1 + 1211.1 + 80,; + 4ni;)
"'14 = E:1 + E.1 + E:;, - Er; - E7 + 5:8
= 201 + :~02 + 4n3 + Cin4 + 4<t:, + 21t5
2
W;, = 3(E8 - E7 - c(;) + .01 + €5
1 .
= 3(4n1 + 602 + 8n:i + 12n-i + 10n~, + 501;)
1
W(i = 3(c8 - E7 - c(i) + E;,
l
= 3(2n1 + :302 + 4o:i + 6n 4 + 5nr, + 4nr;).
2p = 2(::2 + 2:::i + :3::4 + 4c:r, + 4(cx - E7 - E'1>))

= 2(80:1 + 1102\+ 15n:i + 2101+151\'5 + 806).
(VIII) P(R)/Q(R) is isomorphic to lz/:3Z.
Conu<'ction index: :3.
(IX) Exponents: 1,4.5.7.8, 11.
(X) Ord('r of W(R): 27 .3 4 .5.
(XI) ~(R) =:- W(R) x { 1. -1 }: w 0 fransforms n1, 02. <1';1. 04. <Y;,. O(l, rPspec-
t1vdy. mt.o - 0 6 , -02. -o;,. -h 4 , -n~. -n 1 .
PLATE V 277
(XII) Tb0 non-identity Pl<'llH'llt of A(R)/W(R) defines tlH' automorphism

.r ,._, -.r of P(R)/Q(R).
The µ;roup of a11tornorphis111s of the compld<•d Dy11ki11 diaµ;nun is
isonwrphic to 6: 1: its dern<'Hts of order;~ an· iudnrcd by the two non-
trivial ck11H'11ts of P(W)/Q(W).
(XIII) Cartan matrix:
2 () -1 () () 0
() 2 () -1 0 ()
- I () 2 -1 () ()
() -] -] 2 -1 ()
() () 0 -] 2 -1
0 () () {) -1 2
PLATE VI
SYSTEi\f OF TYPE E7
(I) V is th<' hyperplane i11 E = R 8 orthogonal to c 7 + c8 •

R~>ots: ±E:; ± Ej (I ~ i ~ j ~ 6). ±(E1 - -"H),
G 8
±~(E1 - c- 8 + ;~(-l)''(i)c;) with L v(i)
i=l
odd.
'.\'urnbcr of roots: /1 = 12u.

(II} Ilasis:
Positive roots:
±c; + Ej (1 ~ i < j ~ 6). -E7 +EH,

b f>
~(<1+EH+;~(-l} 1 '(i)c,) with ;~v(i)odd.
Positive• roots containing o 7 a11d having at least one coefficient ~ 2
(wt~ deuotc tlw rnot acq + lm2 + cn:i + dn 1 + enr, + f nt; + gn7 hy
",.r1,. f y )11.
h •
012111 112111 012211 122111 112211

l l l l
012221 122211 112221 122221 12:~211
1 1 1 I 1
l n22 l 12:3211 12n21 I 2 :J 22 l 123321
2 l 2 2
11 Thc positive roots not containing o; come from Eo. The positive roots all of whose
coefficients arc ~ I can be obtained h,v applying Cor. :1 of Prop. l!l of Chap. VI,
S l. no. 0.
280 PL:'\TE VI
124:321 134:l21 2:l4321

2 2 2
(Ill) Coxeter munbcr: h = 18.
(IV) HighPst root:
Complct.Pd Dynkin graph:
0~3--r:)--«~--:,
00!2
(V) K = R.
<J>R(.r., JJ) = (:riy)/36,
(VI) Fun<lanwntal weights:
w1 = €8 - E7 = 2n1 + 2n2 + 3o::i + 4a4 + 3a5 + 206 + 0:1

1
W2 = 2(E1 + €2 + E:i + €4 + €;, + E(i - 2€7 + 2.::8)
1
= 2(4n1 + 71112 + 8o:a + 120,1 + !Ja5 + 806 + 3n1)
1 .
W;3 = 2(-€1 + E2 + €;3 + €4 -f-E5 +E6 - 3€7 +3.::8)
= 301 + 4n2 + 6o:i + 81}4 + 60::; + 4<'16 + 20:7
W1 = E;i + €4 + €5 + €6 + 2(E8 - €7)
= 4n1 + 602 + 803 + 1201 + 9a::; + 60:6 + 3a1
3
Wr; = E.t + €5 + €6 + 2(EB - €7)
1
= 2(6oq + 902 + 120-3 + 180: 4 + 15r}5 + 10n6 + 5<} 7 )
W5 = €5 + €6 - €7 + €8
= 20:1 + 302 + 4a3 + 60.1 + 5or; + 4ofi + 2a1

W7=Ei;+2(€8-€7)
1
l
= ~ (2n1 + 302 + 4t:i + 60-4 + 5a:, + 406 + 3n1 ).
(Vil) Sum of the positivc roots: \
2p = 2€2 + 4€3 + 6€4 ~ 8€5 + 10€6 - 17€7 + l 7Eg

= 3401 + 49<l'2 + 6(+:1 + 9604 + 7505 + 52a 6 + 2707.
PLATE VI 281
(VIIT) P(R)/Q(R) is isomorphic to Z/2Z.

Co1111cc:tio11 index: 2.
(IX) Expo11ents: 1,5,7.9.11,13,17.
(X) Order of W(R): 2 10 .a 1.5.7.
(XI) A(R) = W(R), wo = -1.
(XII) P(W)/Q(R') has 011ly <HI<' non-id<'11tity clC'ment: it ddlnes the unique
no11-trivial automorphism of tlw Dy11ki11 graph.
2 0 -1 () 0 () 0
0 2 () -1 () u ()
-1 0 2 -1 () 0 0
0 -I -1 2 -I 0 ()
() 0 () -1 2 -1 0
0 () 0 () - 1 2 -1
0 0 0 () () -1 2
PLATE VII
SYSTEl\I OF TYPE Ea
8 8
Roots: ±E; ± Ej (i < j). ~ L (- l)''(i)E;
1~1
with L v(i)
1=!
even.
~11111lwr of roots: 11 = 2-IO.
(II) Basis:
PositivP roots:
Positive roots contai11ing n 8 a11d having at least one coefficient ? 2(we

d1'notl• tlw root an 1 + bn'2 + cn:1 + <10.1 + rn!:i + fn 0 + .IJ0-7 + hnx hy
,,,.,f,.fqh)J!'>.
h •
0121111 0122111 1121111 0122211 1221111

l
1122111 1222111 1122211 0122221 1232111
I I l l 1
1222211 1122221 12:!2211 1232211 1222221
1 1 2 1
12:~2211 r n :~ 211 12:l2 2 2 l 1n:~21 i 1232221
2 2 2
1''The po:;it.ivP roots not containing tl~ couie fro111 E,. Tlw positive roots all of whose
coeffiei('lllS arc ,,;; l can lw obtained by npplyi11g C'or. :l of Prop. 19 of Chap. vr,
~I. no. ti.
28-1 PLATE Vil
1 2 :1 ;3 :3 2 l 124:3211 I 2 :l ;3 2 2 1 1 n :3 :3 2 1 l :3 -13 2 I l
2 2 I 2
l 2 -1:322 I I 2 ;3 3 :3 2 1 2 :3 I :12 1 l 1:14:3221 1213321
2 2 2 2 2
2 :14 ;3 2 2 1 1;34 :1 :12 J 124-1:321 n1 :1:J21 1 ;14 I ;3 21
2 2 2 2 2
l ;3 .'j 1 :.12 1 nLl:321 1 :1 .'j -1;32 1 2:3.'it:l21 23 fi ·I :3 21
2 2 ;3 2 3
24.14:321 2 4 5 ·1 :32 l 2 JG4:321 2-Hifi32 l
1 2165421
2 3 :1 :3 :1
2 4 6 ti 1 ;3 1 2 4 G.'"> ·I ;3 2
;3 :1
(III) Coxcter 11m11hcr: h = :m.
(IV) Highest root:
ii = E7 +Ex = 2rl'1 + :3n2 + 4n~1 + Go-l + 5o;, + 4nn + :3n 7 + 20-8
= W11.
Completed Dynkin graph is:
(V) W =IL
<Pn (:r, y) = (.i:ly)/GO, -y(R) = 900.
(VI) Fundamental weights:
w1 = 2c11 = 4n1 + Gn2 + 7o--:i + 100-.1 + 80,, + 6nn + -1n7 + 2o-s

1
W2 = 2k1 + E2 + E:1 + E.1 + + cfi + E7 + 5Ex)
E:;
= 5n 1 + 802 + Ulna+ lfin4 + 120:; + 9nn + 607 + 3<\'s

l
W:3 = 2(-E1 + E2 +Ea E4 + E:, + E<; + E7 + 7c:s)
= 7n 1 + 10n 2 + 14n 3 + 200-i + lf>n:, + 12nfi + 8n 7 + 4fl's
= lOo 1 + 1502 + 20< I :1 + :mn.1 + 24n:, + l 8nr; + 1207 + 6n8

PLATE Vil 28!)
Wr, = E4 + E;, + Et; + c; + 4E8

= 8nt + 12112 + IGn:1 + 21n1 + 20n,; + IGn1; + l0n7 + 508
w'r; = E,; + E"c; + E7 + :3.:8
= G11 1 + Dn 2 + l2o:1 + 180 1 + JGn:; + 12nn + 8n7 + 4n 8
w; = Ee; + E7 + 2E8
= 4n 1 + Go2 + 8n:1 + l2n 1 + LOn~ + 8rto + Gn; + 3ns
(VII) Sum of t lw positive roots:
2p=2(.:2 + 2.::1 + :~.:4 +·IE;;+ Gc-1; + Gc-7 + 2:~.oH)

= 2('Wn1 + G8n2 + 9101 + J;~!'in1 + l lOn;, + 84no + 0707 + 2!Jn8)·
(VIII) Q(R): tllf' sPt. of points with coordinat<'s ~; such that 2~; E Z,
8
~; - ~i. E Z. 2= ~'
i=l
E 2Z.
Comwct ioll index: l.
(IX) Exponents: l. 7, lJ. 1:3, 17. HJ. 2:l. 2!J.
(X) Order of W(H): 2 11 .:3".5 2 .7,
(XI) aud (XII) A(ll) = W(R). 11•0 = -1.
(XIII) Cart.au 111at.rix:
2 0 -1 () () 0 () ()
() 2 () -l () () () ()
-1 () 2 -1 fl 0 () ()
() -1 -] 2 -1 () () ()
() () () -1 2 - 1 () (}
0 () () () -] 2 --1 ()
() () () () () -1 2 -1
() () () () () () -1 2
PLATE VIII
SYSTEM OF TYPE F 4
(I) V=E = R 1•
Roots:
± E; . ( I ~ i ~ ·1). ± E; ± E j ( 1~ i <j ~ 4),
1
2(±E1 ±E2 ± E;i ± €4).
N umli<'r of roots: n = 48.
(II) Basis:
Positiv<> roots:
E, (1 ~ i ~ 4),
Positive roots havi11g at. btst 011c c:o('fficient ~ 2 (we dcnot<' the root
Wt I + fio2 + Cll':l + <f<L1 by (l /J ('(/) lfl:
0120 1120 0121 1220 1121 0122 1221 1122

l 2:l J 1222 12:32 1242 I :~-12 2:H2
(lJl) = 12.
['.; m11bcr of roots: h
(IV) Highest root: ii= E1 + E2 = 2n 1 + :kr2 + 4n:i + 20'1 = w,.
Completed Dynkin graph:
lGThc other positin' roots can he oht.1tirwd by applying Cor. ;J of Prop. 19 of C'hap.
VI. 3 I, no. 6.
288 PLXl'E VIII
(Y) W th<' set or wclors ±2.::,, ±c; ± E1 . ±Et± Ez ± E3 ± E4.
(.r Iul
<l>n(:r,y) = ----ul' 1(B) = 2.3 4 .
(VI) F11ndamcntal weights:
W1 =Et + E2 =-+ :kl'.2 + 4n3 + 204

2n1
w2 = 2E1 + E2 +Ea= :kr1 + 6n2 + 8n:1 + 404
1
W3 = 2(3c1 + E2 + E:1 + E1) = 201 + 4n2 + 60'3 + 3<Y4
w~ =- E1 =- c1:1 + 2n2 + :kv;1 + 2n1.
(VII) S11m of the positive roots:
(VIII) Q(R) = EBt=l ZE, + Z(~ LE,).

i=l
P(R) = Q(R).
Connection index: 1.
(IX) Exponents: L 5, 7, 11.
(X) and (XI) A(R) =- W(R), w 0 = -1.
- I 0
2 -2
-1 2
0 -1
PLATE IX
SYSTE:VI OF TYPE G 2
(I) V is the hyperplane in E = R 3 with equation ~ 1 +6 +6 = 0.

Root::;:
±(-"1 - c2). ± (c1 - 23), ±(c2 - c:3). ±(2c:1 - c:2 - c:;i),
± (2c2 - c: 1 - c:;i). ±(2c:1 - €1 - c2).
Number of roots: n = 12.

(II) Basi:s: n1 = c:1 - €2. n·2 = - 2€1 + E2 + €3.
Positive roots: 0-1. n t + n2. 2n 1 + (..\'2. 3n1 + n2, 301 + 202.
(IIf) Coxeter number: h = 6.
(IV) Highest root: 6: = -c: 1 - c2 + 20:3 = 301 + 2n2 = <..v'2·
Completed Dy11ki11 grnph:
111 llz
(V) R' is the set of vectors
(VI) Fundamental weights:
2p = 2(5n1 + 3nz).
2!)() PLATE IX
(VIII) P(IO = Q(Il).

C'outwcl io11 index: l.
(IX) ExpotH•uts: l. fl.
(X) W(H): thl' diltcdral group of ordcr 12.
(XI) aud (XH): A(H) = \\'(B). 111 1i = -1.
(XIII) C'art1111 matrix:
PLATE X
rrrnEDUCinLE SYSTEMS OF RANK 2
Tiu' first three diagrams abovP n•pr<'SCJlt the root. systems R of type A2. n2
aucl G 2 . Tlw shaclc'd rq!;iou n'pr<'s1'11ts t IH' drnmlier (' <"01T<>spo11cliug to the
basis (n 1 • n 2 ). Tia' liuc (.rid) = l. whcr<' ii d<'llOl<'S th<' highest root of the
292 PLATE X
n.-, is sltov-·11 dot t.cd, awl tlw clo11hl.v shadt'd n•giou n•pr<:'sents
invPrsc syst.crn
t.lu~
nkow of W with wrt<•x () <·ont.ai1wd in C.
The last diagram r<'prt'SP!lts t.hc 1111iq11<' irn•ducihle n011-rerlu.ad root sys-
tem of rank 2.
SUMMARY OF THE PRINCIPAL
PROPERTIES OF ROOT SYSTEMS
(In this snmmar,v Wf' restrict 011rsclvc's to the case of rcxl11rl'd root systems
and work owr the fip]d of real numbers.)
1) LP! V b<' a real V<'ctor spaC'<'. A n•d11cC'd root system in V is a subset

R of V with thl' followiug prop<'1til's:
(i) R is finite' and gcn<'rates V.
(ii) For all ( l E re then' exists o· E v· SU!'h th:-it (o. n") = 2. and such
that the 1w1p
8n : ./" ,_. ;r - (:i:. n·)n
from V to V trnnsfonns H to IL
(iii) For all n ER. n·(R) <::.= Z.
(iv) If o E R. thl'11 2n tJ_ R.
In view of (i). the <'iPinent n·. whose f!XistPnCf' is guaranteed by (ii), is
11niqt1C': thus (iii) mnk<'S Sl'llSt'. The map .~" is a rdl<'ction that kavcs fix<·d
the points of L" = Kcr(n·) awl tn111sfonr1s n to -n.
The de111e11ts of Han- !'allc·cl roots. TlH' dirnensio11 of V is called th<' rank
of the root syslt'm.
2) The group of autmnorphisms of V that lean' R st able is denoted by
A(R). The .'in (n E H) genernt.t• a s11bgro11p \\"(H) of A(R). rnll<'d thl' W<'yl
group of ll: this subgroup is uormal in A(R). The only reflections lwlouging
to \V(R) al"l' the 8 0 •
:l) The set n· of n· (for n ER) is a n~dncccl root syst.<'m in V*, called the
i11wrse s~';;lc•m of R. Th<' map n ,....... n· is a l>ij<Ttion. C'Hll<'d canonical. from R
to R-. \Ve have {Tr)·= R. a11d th<' ca11onic:al bijcdiou;; R ____, n·. rr· _, R are
n111tnal inverse;;. The map /1. ,_. 1 11- 1 dcfiues au i;;omorphism from \V(R) to
\V(R-) by means of which W<' identify tltcs<' two groups.
-I) Let\' be a r<'al vector ;;pac<' that is the din·ct s11m of vc'ctor s11bspaces
V 1 ••••• V ,.. For all i. let H; be a rt'duced root system in V;. Th@ th<' union
R oft lw R; is a root system in V. called the direct ;;um of the R;. Tllf' gro11p
\V(R) ca11 i>l' idt>11tificd witl1 tlu• prod11ct of the \V(R; ). If R i= 0 and H is not
the direct s11rn of two non-empt~· root syst<'ms. R is said to bP irred11cibk. This
i;; <>q11ivalc·nt to saying that \V(R) is irn'ducible. En•r.v rcd11c<'d root ;;ystem
2!H Sl."l\I;>.l!\llY OF f'RI!'iCIPi\L l'H.Ol'EIUIES
R is a din·C't sum of irreducible rC'duc<'d root. systems. uniquely determined

up to a (Wrmut.atio11, and called th(' irn•d11cihlP co111po11c11ts of rt
5) Let R h<' a reduced root syste111 in V. Then' t•xist scalar products
011 V iuvaria11t. 1u1<ler W(R). 111 the following. we~ denote by (:rly) suC'h a
scalar product. If V and V* ar<' idc11tified usi11g (:rl11), 2 ''>·
we haw• n· = 1.n1n:
.
The rcflc>ction .-;,, is the orthogo11al reflection which transforms n to -n. If
H is irreducible, t.hc Wcyl group acts transitively on the set of roots of a
given length. If R is irreducible, the scalar product (.1:ly) is unique up to
multiplication by a constant.
G) Let H. be a reduced root system. For o, /i E R, put
(o .•)") = n(a. d) E Z.
\Ve have
n(a.o) = 2.
s 11(n) = n - n(o,/J)/3,
. ( iJ) _ 2(nl/i)
II 0." - (!JI}) .
Th(' only possibilities arc th<' followi11g. up to interc:ha11ging n and ,i"J:
n(o. fi) = 11(,3. n) = O; (n./!) = i; .'lasr1 of order 2;

n(n-. i"J) = ri(,tJ. n) = 1: (11. ;J) = }; s0 s1'1 of order 3;
ll~l=ll/ill;
n(n.(i) = n(:3,n) = -l; (<l',,~'J) -_ 3,
2rr. s,.,'lfi of order 3;
11~1=1111 II;
n(o. m= l, n(f'i, o) = 2; (o,/3)=~: 8 0 .'lri of order 4;
I /i II= J2 1 o II;
n(<:i:, /i) = -1. n(/i. n) = -2: ;:-;,)
(u. ,_)
= 3rr.
4 , s 0 s3 of order 4;
11~1= v'2 I n II;
n(n,/J) = 1, n(/3.o) = 3; (nJi)=~; 8 0 .'id of order 6;
II ;:Ul= v'3 I n II:
n(n./i) = -1, n(fi,o) = -3: (er' /3) = 51~; sns 11 of order G;
I /111= v'3 ii n II:
n(<t./i)=n(tJ,n)=2; < =iJ;
n(o, ii) = n(/J, n) = -2: < = -/i;
7) L<'t n.J ER. If (nJ:3) > 0. n :i is a root unless o = ,d. If (ol;3) < 0,
o + fJis a root 1111less n = - iJ.
8) Let <L ;3 t><' two 11011-pr.oporti<~ial roots. The sc>t I of j E Z such that
,!3 + jn E R is an interval ( -q. p) of cont.aining 0. Wc> have
_ _ _ (.·J ) < +1 (!i + nl/i + n)

P <J - 11 ' • 0 ' P (!JI :i)
SU'.\f'.\IAHY OF l'HINCIPAL PHOPEHTIES 295
Ld She th<' set of .J + jn for j EI. Tlw11 8,,(S) =Sand s,,(:i + 7m) = d -q<~.
\Ve rnll S the' 11-chai11 of roots dcfiucd by _.i. ii·- qn its tll"igi11. d +po its <'11<!,
a)l(I JI+ <J its ll'11gth.
IfT is all n-ehai11 with origin/. the l<'ngth ofT is -11(~.. o).
n) Let x lw t lw 1111io11 of I hl' KPr I\ - (I\ E: n). Tiu.· COlllH'Ctcd compo1w11t.s of
V - X are called the d111rnbers of Hin V. They arc op<'n sirnplicial cones. The
\Vey! group acts simply-transitively 011 th<' set of chambers. If C is a chamber,
C is a fn11danw11tal domain for \V(B). v\'<' l1aw (.1ty) > 0 for .i:, y E C. The
hijl'd ion from V to V* eorT<'spondiug to (.1: Iy) defines a hijcct ion from the
S<'t of dwmh<'rs of ll ill vto the set of chambers of n-
in V*: \V(' de11ote by
c- the d1a11ilH'r that is the imagl' of C 1111dc•r this hij0rtion.
W) Let c be a chamber of n. Let LI· L-i ..... L, I)(' th<· Wil lls of c. For all
i. tlwn• exists a nuiqn<' root n; s11cl1 that L; = L,., and o; is 011 tlw same
side of L; as C. T\w fa111ily ( n 1..... n1) is a basis of V. and C is t II<' set of
.r EV s11d1 that (o~;.:r) > 0 for nil i. in other words such that (o;j:r:) > 0 for
all i. Th<'n {o 1 .•... n 1 } is called tht• basis B(C) of 11 ddii1Pd h.\· C. We have
(o;jn1 ) ~ 0 wlu•u i =/= j. Tlw group W(R) acts sirnply-trnnsitivt'l.v on the S<'t
of bas{'s. Every root is trn11sfonrll'd by solll<' dt•111p11t of \V(R) to an <'lenwnt
of I3(C). WP have {nj ..... n'i} = n(C-).
11) Put sn, = 8;. let Sh<' tlu• set" of the s;. and let m.; 1 lw th0 order of
8;-".i· Th0 pair ("W(H). S) is a Cox<'ll'r systP111 with Coxl'l<•r matrix (mij): in
other words. \.V(H.) is cl<~fiucd by thl' family of gl'IH'rntors (.~;) 1,;;;,;; 1 a11d the
rdatious (s;s_; )"'', = I. Two l'll'lll<'llt.s 8; and s1 an' conjugatl' i11 \V(H) if aud
011ly if there l'xist.s a st'<jlWllt:l' of i11dices (I 1. i'2, .... i,1 ) such that i 1 = -i. i'I = j
and each of the rtt.; 1 ; 1 ,., is Pqual to :3.
12) Ld 11;; = 11(11,.n,). Tlw matrix (11. 11 )i,;;;.1 ,;;1 is called the Cartan
matrix of 11.. It is i11dqH'nde11t. (11p to a permutation of 1. 2, .... l) of the
choice of C. We haw n;; = 2. n;j E {O, -1. -2. --:~} for i =/= j. If two root
systems haw the saim• Cartan matrix they arc isomorphic
13) Let G be the subgroup of A ( R) that lcaws I3( C) st1ihl<'. Thm A (R)
is the semi-direct product of G by W(H).
14) The order rclat ion on V ( l'<'sp. V*) defined b~· C is the order relation,
co111patiblc with tho w<'tor space strnct11n~ of V (resp. V* ). for which the
c•lt>m<'nts? 0 arc the li11Par comhinati011s of the o; (resp. n-;) with coefficients
? 0. These elements are callerl positive for C. or for B(C). These order rela-
tions are also defined by c-. An element of V is ? 0 if and only if its values on
c- arc ? 0. The set of dcment.s ? 0 for C contains C but is in general distinct
from C. Let J: E V. Then .i· E C if and only if ;r ? w(:r) for all ·w E W(R);
aud :r EC if and only if :I' > 11:(.r) for all w E W(R) disti11ct from 1.
15) Every root is eithf'r positiv<> or negative for C. We rlenotc b.v R+(C)
t.hc set. of roots that are positive for C. so that R = R+(C) u (-H+(C)) is
296 Slii\ll\IAHY OF PfUNCIPAL PllOl'EflTlES
a partition of R. The rcftcct.io11 s; transforms f~; to -n; awl pcrr1111tcs the

elements of R+ (C) distinct. from n; among themselves.
16) Let Il be a basis of R. Every positive• (r<'sp. negative) root for I3 is
a linear comhinatiou of elements of f3 with codtki<•nts that are integers ? 0
(m;p. ~ 0).
17) Let (i:/ 1 • d2 .... , .3 I)(' a seq1wnef' of mots that arc positive for C and
11 )
such that ;1 1 + · · · + /1,, is a root. TIH•rc exists a p1~nnu1·atio11 r. E 6,, such

that. for all i E {L 2 ..... n} ..N"(l) + 11"(2) + · · · + f17t(i) is a root.
18) Let ct E H+(C). Then n E B(C) if aud only if o cannot be \\'rittcn as
a snm of positive roots.
19) Let C be a chamber. ( o 1 , <12, •.. , oi) the C'OIT<'sponding basis. For any
:mbsPt .J of I= {L2, ... ,/}. le1 \V,1 IH' the s111Jgronp of W(R) gcneratt·d by
the s; such that i E .J. Let C,1 be t lw s<'t of li11Par combinations oft.he o 1 wit.Ji
j E .J and with codlicicnts > 0. so that C'.1 is a fa<'f'I of C.
Let.JC I. y E W(R). The following cowlitions ill'l' equivalent:
a) y leaws a point of C.i iuvariaut:
b) .<J leaves every point of ('.J invariant:
1') y !caws every point of C.1 invariaut;
d) y(C.1) =CJ:
<') q(Ci) = C.1;
.f) .<J E W1.
L<'t .J,.J' c {1.2,. ... /} aud y.g' E W(R). Th<' following coudit.ions arc
equivalent:
a) y(Ci) = .1/(C.1' ):
/1) .<1(C.1) nq'(C.1') f: 0:
c) yvVJ = y'\\'.1,:
d) .J = J' a11<l fl 1 E gW.1.
Let J1 . .J:i ..... .J,. c I and .J = .J 1 n ... n.J, .. Then W. 1 = W.1, n ... nw. 1,.
For ally E W(R). tlu•rp exists.JC I such that Cn,q(C) = C.1 and .IJ E \V.J.
20) L<'t r \)(' a suhs!'t 'of H. Tlw11 p is said to be clos!'d if till' ('Ollditions
n E P. ;IE P. 11 + :i E H imply n + d E P: mul P is said to lw parabolic if P
is dosPd and PI.) ( -P) = Tl. The following c·o11ditio11s are c·quivalcut:
a) P is parabolic: 1
Ii) Pis dosl'd and there exists a chamber C such that P :J R+(C);
c) thcrP exist a charnlwr C and as tbsPt I: of l3(C) such that Pis the u11ion
of R+ (C) and the set Q of roots that are liuear combinations of elements of
E with coefficiP11ts that arp integers ~ 0.
Assume that th<•se couditions are\ satisfied aud let V 1 be the vector sub-
space of V generated hy E. Then \
Pn (-P) = Qy (-Q) = V1 n n,
and P n (- P) is a root syst.l'm in V 1 with basis E.
SUMMARY Of' PRINClPAL PHOPEflTIES 2!)7
LPt P', C', E' li11ve analogo11s propnt.ies. If t her<' exists an !')pmc11t of
W(R) transforming P to P', there <·xists an <'k1uc11t of \\'(11) transfonning C
to C', E to E' and P to P'.
21) Let P be a subset of TI. The following ronrlitions are equivalent:

o) th('re exists a chainLH'r C s11ch that P = Il+ ( C):
b) Pis do:->cd, and {P. -P} is a partition of Il.
Th<' chamlwr C is then 11niq11c.
Assume that Vis cq11ipped with th<' strndurP of an ordered V<'dor space
s11ch that <'Very root is either positiv<' or negative. Ld R+ lw thf' set of
positive roots for this strnctnrc. There exists a 11niq11\' chamber C s11ch that
R+ =R+(C).
22) A subset n of 11 is " basis of R if and only if the clenwnts of I1 arc
linearly indep(•ndcnt and CY<'ry root is a linear COJllhinat ion of elcn1ents of f3
with codficit'11t.s that ar<' all ~ 0 or 11'! ( 0.
n) Let P lw a dosed subset of H. such that P fl (-P) = 0. Th<'rf' exists

a chamber C such that Pc 11+(C).
24) A subset P of n. is called sy1nmetric if P = - P. Let P lw a s11hsf't of
H. a11d l<'t V 1 (resp. /) IH' the vector s11l>spac<' (r<'sp. the rnlditiw s11l>gro11p)
of V gcncrnt<•d by P. TIH' following t·ouditions arc <'q11ivalc11t:
a) P is dosed and synHnd ric:
11) Pis dosed an<l Pis a root sys!<'m iu V 1:
e) rnR= P.
20) Ass11rnP that 11 is irr<'dtwil>k. Lct C })('a chamlwr: p11t.
Tlwn' exists a highest cknH'llt of R (for th<' onkr ddined hv C). that is. an
cl<'nl<'ll( (l = 111fl1 t · · · + 11/11/ of H Sllch that. for illl.V rnot
J11<l1 l-···+11101.
11 1 :;o: p 1..... 111? /ii· Th(')l 1ic E: C and II 11 ll:;o:ll n II for evPry root n.
2()) Th<' s11bgro11p of V g<'lH:l'Hl<'d \iv H is d<'11otPd by Q(l1): the <'l<•n1<'11ts
of Q(J1) an• called the radical W<'ights of n. The gro11p Q(H) is a discrete
s11bgrrn1p of V of rank I =dint V. Any basis of R is a basis of Q(H).
The s11hgro11p of V associated l.o Q(W) is dcnot<•d by P(H): the c!Pmeuts
of P(H.) an~ called tlw weights of H. The gro11p P(R) is a disnet<' subgro11p of
V of rank I contaiuiug Q(H.). The gro11ps P(J1)/Q(R). P(W)/Q(W) arc finite
and isomorphic: their order f is called the ('()JlHection illdPX of H. \\'ith the
11otatio11 of 20). the order of VV(R) is /!111112 ... 11tf.
The group A(H) kan~s P(H.). Q(R) stahk. and hc~nn· acts 011 P(R.)/Q(R).
TIJ(' group \V(R) ads trivially on P(H)/Q(It). so A(R)/W(R) acts on
P(R)/Q(n).
2!>8 SU!\L\!Al1Y OF PH!NC!l'AL P!lOl'EHTIES
27) Let C be a dmmb('r. L('t U = (o 1 , •••• n1) lw the cmT<'sponding basis

of R. The dual basis (w1 ..... (J.;'1) of (n"j .... ,11:1) is n basis of P(H). The
;,;; are ('a.lied th<' f1mdm11cntnl weights (for C. or for 13). Tlw s<'l of li1l<'ar
combinations of the w; with ('Odfici<'nls > () (n'sp. ;;? 0) is C (r<'sp. C).
TIH' li1war C'ombi1111tio11s of the ;,;; with intcg<'r co<'fficieuts ;;? 0 arc eallt-d
the dominant weights. Ev<'I")' <'lrn11<'11t of P(R) c1111 lw tra11sforn1P<i by s01111'
<'lc111P11t of \V(H) 1o a u11iq11c dominant \Wight. Tlw dominant w<'ights are
1lw l'lements....: of V such that. 2(("'1
'")J is au inl<'gcr;;? ()for all i.
n, n,
1
28) Let p == 4nEH.,(C)

L n. Thl'n. p = w 1 + · · · + w1 E C.
W) Lt>t. T he the group of translntious of v·• whose vectors belong to

Q(tr). The group of affint' tra11formatio11s of y• generated by T and W(R)
is the semi-direct prnd11ct of \V(R) by T. This group is called the affine Wey!
gro11p of H and is dP1uitPd by \V,, (R). It a('ts properly 011 V*. For o E R
and,\ E Z. ld s, .. >. be 1he map .i:* ,_. .r* - (.r*.n)n- + ,\o·: this is a11 affine
n~flpct ion. and the sl't. Ln.>. of its fixed points is definc•d by t lw equation
(J:'. n) = ,\: W<~ have Ln.A = Ln + i,\o ·. The s<>.A ar<' t.lw affine rdlPctio11s
belonging to W"(R). and they genct~atc the gro11p \V,,(R).
:lO) Let E h<' tlw 11nion of tll(' Lo.A for n E R a11d ,\ E Z. Th<> connect<-d
components of V* - E an' rnlled the alcoves of R. If n is irrcd11dblP. each
alcove is au OJl<'ll simplex: in gt'Jwrnl. an alcove is a product of ope11 sirnplices.
The gro11p \V,, (R) acts siu1ply-trn11sitivcly on t.he st't. of alcows. If C is an
akow, C is a fundarncntal domai11 for \V"(R). Ll't a 1 ••••• aq lw tlw rdicctions
in \ll/ 11 (H) concsponding to the walls of C: let /t;.i be the order of a,a.i. Then
\Va(R) is ddincd by the gPncrntors a; awl the relatious (a;a 1 )1'• 1 = 1.
:ll) If' p E P(W). tlwre <'xists m1 alcove C su('h that pi::; au <•xtrcmal point
of C. If C' is an ak:ove, there is a uuique radical weight that is an extremal
point of C'.
Let J:* E y•; the following conditions arc equivalent:
a) ;1:* E P(R"):
b) for all <l E R. t.lH' h\·1wrpla11c parnlld to Ln and passing through .r;• is
one of tlw Ln,A·
LPt C' bt' a chamber nf n-. Tlwrl'xists a uniqll<' akovc c contained in C'
and su~:h that (~ E C. Ass~1'.ne that is irrcd11cibk', au<l kt .J be the bigh<'st
root of R (for C'): then C 1s tlw set f :r* E C' such that (x*, ;J) < l.
:l2) Let She the symmetric alge a of V. sw the subalgebra consisting of

the ele11w11ts invariant under W = v\1( H), g the order of W. I -== di111 V. There
exist homogeneous. algebraically ind~>endent clements 11, Ji, ... , 11 that gen-
ernt<' sw. The sW -modules has a bas s consisting of .<J ho111ogeT1('0\IS elements.
Let a be the ideal of S generated hy he homogeneous clements of sw of de-
gree > 0: th(• l'Pprcsentat.io11 of w
Oil I a. induced by the !"('presentation of vV
SC~HdARY OF PR!i'\CIPAL PROPERTIES 29!)
on S hy passag1· to t Ii<• q11oti<~nt. is iso111orphic to the n•g11lar WJ>r('S<'ntation

of W (owr R).
:l:l) Ld. I I. 12 ..... r, I><· homoge1wo11s. algebraical 1.v independent ('l('lll('llts
generating sw. Their dcgn•<•s k 1 , kz . .... kt arc 11niq11ely detcnnirn'd ( 11p to
l
order) I>y R. We have .<J = J.: 1k2 ... k1. Th<' m11nb<'r of roots is 2 L (ki
1=1
- 1).
:M) An dcmcnt A of S is said to lw ant:i-i11variant. nnd<'r W if w(A) =

<kt(w).A for all u· E \\". Let R = R 1 U (-Rt) be a partitio11 of R, and put
7r = fl n. Th<' deme11t 7r of S is auti-invariant: the anti-iuvmiant demcnts
<rE HI
of Sare tht' elcmeuts of lhc form rrl. with I E sw.
;35) Let E be the group algd>nt Z[P] of the gro11p of weights P of R. If
fl E P. we denote by eP the correspo11ding element of E. We have FPc1/ = eP+P'
aud the 1·'' form a basis of E. There is an action of the gro11p W on E such
that w(eP) =
e"-'(JJ) if It.' E W and p E P. An Pleme11t z E E b said to
be anti-invariant if l!'(z) = dd(w).z for all It' E W. For all z E E, put
.J ( z) = L det ( w). w ( .:: ) . Let C h<' a charn her. The elem<'11ts .J (c1'), where
trEW
p E P n C, form a basis of the gro11p of anti-invariant elein('J)t.s of E. If p is
half th<' s11m of the positive roots, \VP have:
.J(eP) =el'
et>O
n (1 - c-") n (e"/ =
n>O
2 - ('-<>/ 2 ),
the products being tak<'n over the spt. of roots > 0.

36) With the notation of 35), put
z,, = .J(el'+P)/.J(d') for ]J E P.
The z1,, for p E PnC, form a basis of the group Ew of dC'rncnts of E invariant
under 'vV. If w 1 , •.•• w1 an• the fundamental wf'ight.s of ll. the elements zw,,
1 ~ i ~ l, are a lgehraicall_y independent and gem•rat<' the ring Ew.
;n) Let C be a diam her of R, (o 1 , .•. , lft) the corrn-;ponding basis. The
clement e = s 1 8 2 ... St of \V is called the Coxctcr trans format ion of R. The
conjugacy class of c does 11ot. dc•pend 011 C or on the m1111b<'ri11g of then,. The
order h of c is called the Coxcter numb<'r of R. The eigC'nrnhws of c are of
the form exp 2 ;~;";, wh<'n' the integers m 1 • rn. 2 .. .. , rn1 (called the exponents
of H) arc such that 1 ~ 111 1 ~ m 2 ~ • • · ~ m1 ~ h - l.
Assume that R is irrf'd11eiblc'. Then
rn1 = l, m1 = h- 1.
mJ+m1+i-J=h (l~j~l).
l 1
m 1 + m2 + · · · + m1 = 2th= 2card(R).
:100 SllMi\IAHY OF l'IU!\iCIPAL PHOI'ElfflES
Ewry m E { 1. 2.... , h - l} copri11w to h is equal to exactly one of th<' 1nJ.

Th<• 1111rnbcrs 111 1 +1. m 2 + l. .... m 1+1coi11cidc.11p to ord(•r. with the integers
dcuot<'d hy k 1 • k2 ..... A·, i11 :3:~). \Vith I he 11otat iou of 25). n 1 + · · · + n1 = h- l.
There exist I or hi ts of { I. r:. r· 2..... c" - 1 } 011 H. PH<'h of wh irh has h <'i<'Hl<'llts.
If h is <'\'PT!. c"l 2 trnnsforllls C to - C. \V<' have - 1 E \V if and 011ly if the
cxpo11e11ts of\.\' arc all odd: in that. ms<'. h is even and r:"l 2 = -1.
Druck: Strauss Offsetdruck, Miirlenbach

Verarbeitung: Schaffer, Grunstadt
N. Bourbaki, Lie Groups and Lie Algebras
Chapters 4-6
The purpo e of the Element. Qf Mathematics by Nicola Bourbaki

i to provide a formal y tematic pre entation of mathematic from
their beginning. Thi volume contains chapter 4 to 6 of the book on
Lie Groups and Lie Algebras. [t i devoted to root y tern , Coxeter
group and Tit y tern , which occur in the tudy of analytic or alge-
braic Lie group . It contain the following chapter :
1
-
t-J
, L lb"
, I
I
I. - L- ~['-
ISBN 978-3-540-69171-6
"
1
•• • I II
9 783540 691716
) .sp'ri n~er,.c·9m ~-!j-1' L- I - '- ' •
~
1
, I l-' 1
: . . 111

Lie Groups and Lie Algebras, Chapters 4-6 - Nicolas Bourbaki

Uploaded by

Copyright:

Available Formats

You might also like

Lie Groups and Lie Algebras, Chapters 4-6 - Nicolas Bourbaki

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lie Groups and Lie Algebras, Chapters 4-6 - Nicolas Bourbaki

Uploaded by

Copyright:

Available Formats

ELEMENTS OF MATHEMATICS

Lie Groups and lbie Algebras

Lie Groups and

Mathematics Subject Classification (2000 ):

Cataloging-in-Publication data applied for

ISBN 3-540-42650-7 Springer-Verlag Berlin Heidelberg New York

SPIN: 10659982 41/3142 db 5 4 3 2 1o

INTRODUCTION TO CHAPTERS IV, V AND VI . . . . . . . . V

CHAPTER IV COXETER GROUPS AND TITS SYSTEMS

CHAPTER V GROUPS GENERATED BY REFLECTIONS

§ 1. Hyperplanes, chambers and facets . . . . . . . . . . . . . . . . . . . . . . 61

§ 3. Groups of displacements generated by reflections . . . . . . . 76

§ 4. The geometric representation of a Coxeter group . . . . . . . 94

§ 5. Invariants in the symmetric algebra .................... . 108

§ 6. The Coxeter transformation ........................... . 121

Appendix: Complements on linear representations . . . . . . . . . . 129

Exercises for § 2. 133

CHAPTER VI ROOT SYSTEMS

§ 1. Root systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155

§ 2. Affine Weyl group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

§ 3. Exponential invariants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194

§ 4. Classification of root systems ..... .............. ........ 201

HISTORICAL NOTE (Chapters IV, V and VI) . . . . . . . . . . . . 249

PLATE VIII. System of type F 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 287

In this section, W denotes a group written multiplicatively, with identity

1. LENGTH AND REDUCED DECOMPOSITIONS

DEFINITION 1. Let w E W. The length of w (with respect to S), denoted

Thus 1 is the unique element of length 0 and S consists of the elements

PROPOSITION 1. Let w and w 1 be in W. We have the formuJas:

l(ww 1 ) ( l(w) + l(w'), (1)

l ( w) - l (w') ( l ( ww' - 1 ), (4)

COROLLARY. Lets= (s 1 ,. . .,sp) ands'= (s~,. . .,s~) be two sequences

DEFINITION 2. A dihedral gmup is a group generated by two distinct ele-

p = ( -1, 0), p' = (-1, l), 7l' = (1, l). (7)

Then p 2 = p' 2 = 1 and 7r = pp'. The formulas

1Tn = (l,ni), p1Tn = (-1,ni) (8)

show that Dm. is a dihedral group generated by {p, p'}.

PROPOSITION 2. Ass·u.me that S consi.sts of two distinct elements s nnd s'

Remark. Consider a dihedral group W of order 2m generated by two distinct

W2k = (ss')k, W2A:+1 = (ss')k s,

Ifs = (s 1 , •• . , sq) is a reduced decomposition (with respect to { s, s'}) of

3. FIRST PROPERTIES OF COXETER GROUPS

DEFINITION 3. (W, S) is said to be a Coxeter system if it satisfies the fol-

When (W, S) is a Coxeter system, one also says, by abuse of language,

Examples. 1) Let m be an integer:? 2 or oo and let W be a group defined by

Remark. Suppose that (W, S) is a. Coxeter system. There exists a homomorphism

and s' is conjugate to s.

4. REDUCED DECOMPOSITIONS IN A COXETER GROUP

Suppose that (W, S) is a Coxeter system. Let T be the set of conjugates in

tj = (s1 ... Sj_ i).sj(S1 ... Sj-1)- 1 for 1 ~ j ~ q. (11)

The map w >--+ Uw is a homomorph·ism from W to the group of permutations

Fors ES, define a map Us from R to itself by

U 5 (e:,t) = (e:.(-1) 8•·',sts- 1 ) (e: = ±1,t ET), ( 13)

where Os,t is the Kronecker symbol. It is immediate that = IdR, which u;

This is clear for q = 0, 1. If q > 1, puts'= (s 1 , ... ,sq-I) and

Using the induction hypothesis, we obtain

= (£, (- l) n(s',t)+.5 ScJ:W

T w be the set of elements of T such that T/( w, t) = -1. Then s is a red·uced

This shows that s is not a reduced decomposition of w.