Author: Isaacs M.  

Tags: mathematics   algebra   group theory  

ISBN: 0-12-374550-0

Year: 1976

Text
                    CHARACTER THEORY OF
FINITE ROUPS
f. MARTIN ISAACS


Character Theory of Finite Groups
Pure and Applied Mathematics A Series of Monographs and Textbooks Editors Samuel Eilenberg end Hymen Beee Columbis University, Nsw York RECENT TITLES James W. Vick. Homology Theory: An Introduction to Algebraic Topology E. R. Kolchin. Differential Algebra and Algebraic Groups Gerald J. Janusz. Algebraic Number Fields A. S. B. Holland. Introduction to the Theory of Entire Functions Wayne Roberts and Dale Varberg. Convex Functions A. M. Ostrowski. Solution of Equations in Euclidean and Banach Spaces, Third Edition of Solution of Equations and Systems of Equations H. M. Edwards. Riemann's Zeta Function Samuel Eilenberg. Automata, Languages, and Machines: Volume A and Volume B Morris Hirsch and Stephen Smale. Differential Equations, Dynamical Systems, and Linear Algebra Wilhelm Magnus. Noneuclidean Tesselations and Their Groups Francois Treves. Basic Linear Partial Differential Equations William M. Boothby. An Introduction to Differentiable Manifolds and Riemannian Geometry Brayton Gray. Homotopy Theory: An Introduction to Algebraic Topology Robert A. Adams. Sobolev Spaces John J. Benedetto. Spectral Synthesis D. V. Widder. The Heat Equation Irving Ezra Segal. Mathematical Cosmology and Extragalactic Astronomy J. Dieudonne. Treatise on Analysis: Volume II, enlarged and corrected printing; Volume IV Werner Greub, Stephen Halperin, and Ray Vanstone. Connections, Curvature, and Cohomology: Volume III, Cohomology of Principal Bundles and Homogeneous Spaces I. Martin Isaacs. Character Theory of Finite Groups In preparation James R. Brown. Ergodic Theory and Topological Dynamics Clifford A. Truesdell. A First Course in Rational Continuum Mechanics: Volume 1, General Concepts K. D. Stroyan and W. A. J. Luxemburg. Introduction to the Theory of Infinitesimals Melvyn Berger. Nonlinearity and Functional Analysis: Lectures on Nonlinear Problems in Mathematical Analysis B. M. Puttaswamaiah and John D. Dixon. Modular Representations of Finite Groups
CHARACTER THEORY OF FINITE GROUPS I. MARTIN ISAACS Department of Mathematics University of Wisconsin Madison, Wisconsin 1976 ACADEMIC PRESS New York San Francisco London A Subsidiary of Harcourt Brace Jovanovich, Publishers
Copyright © 1976, by Academic Press, Inc. all rights reserved. no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. ACADEMIC PRESS, INC. Ill Fifth Avenue, New York, New York 10003 United Kingdom Edition published by ACADEMIC PRESS, INC. (LONDON) LTD. 24/28 Oval Road, London NW1 Library of Congress Cataloging in Publication Data Isaacs, Irving Martin, (date) Character theory of finite groups. (Pure and applied mathematics, a series of monographs and textbooks; 69) Bibliography: p. 1. Finite groups. 2. Characters of groups. I. Title. II. Series. QA3.P8 [QA171] 510'.8s [512.24] ISBN 0-12-374550-0 75-19652 AMS (MOS) 1970 Subject Classifications: 20C15, 20C05, 20C25, and 20C20 PRINTED IN THE UNITED STATES OF AMERICA
Contents PREFACE vii ACKNOWLEDGMENTS ix NOTATION xi 1. Algebras, modules, and representations l Problems • • 2. Group representations and characters 13 Problems 29 3. Characters and integrality 33 Problems 44 4. Products of characters 47 Problems 59 5. Induced characters 62 Problems 73 6. Normal subgroups 78 Problems 95 7. T.I. sets and exceptional characters Problems 99 121
vi Contents 8. Brauer's theorem 126 Problems 141 9. Changing the field 144 Problems 156 10. The Schur index 160 Problems 171 11. Projective representations 174 Problems 194 12. Character degrees 198 Problems 216 13. Character correspondence 219 Problems 237 14. Linear groups 240 Problems 260 15. Changing the characteristic 262 Problems 285 Appendix Some character tables 287 Bibliographic notes 292 References 295 INDEX 297
Preface Character theory provides a powerful tool for proving theorems about finite groups. In fact, there are some important results, such as Frobenius' theorem, for which no proof without characters is known. (Until fairly recently, Burnside's p"<f theorem was another outstanding example of this.) Although a significant part of this book deals with techniques for applying characters to "pure" group theory, an even larger part is devoted to the properties of characters themselves and how these properties reflect and are reflected in the structure of the group. The reader will need to know some basic finite group theory: the Sylow theorems and how to use them and some elementary properties of permutation groups and solvable and nilpotent groups. A knowledge of additional topics such as transfer and the Schur-Zassenhaus theorem would be helpful at a few points but is not essential. The other prerequisites are Galois theory and some familiarity with rings. In summary, the content of a first-year graduate algebra course should provide sufficient preparation. Chapter 1 consists of ring theoretic preliminaries, and Chapters 2-6 and 8 contain the basic material of character theory. Chapter 7 is concerned with one of the more important techniques for the application of characters to group theory. The emphasis in all chapters except 1,9, 10, and 15 is on characters over the complex numbers rather than on modules and representations over other fields. In Chapter 9, irreducible representations over arbitrary fields are considered; and in Chapter 10, this is specialized to subfields of the complex numbers. Chapter 15 is an introduction (and only that) to Brauer's Vll
vm Preface theory of blocks and " modular characters " and so is concerned with connections between complex characters and characteristic ^-representations. The remaining chapters concern more specialized topics. Chapter 12 deals with the connections between the set of degrees of the irreducible characters and the structure of a group. Chapter 13 is essentially an exposition of a paper of Glauberman, and Chapter 14 contains some "classical" results on complex linear groups and a small sampling of more recent developments. Following each chapter there is a selection of problems. In addition to some routine exercises, these include examples, further results, and extensions and variations of theorems in the text. It is hoped that the reader will find that these problems enhance his understanding (and enjoyment) of character theory.
Acknowledgments At various times during the preparation of this book I was supported by research grants from the National Science Foundation and a research fellowship from the Sloan Foundation. These organizations deserve (and hereby receive) my sincere thanks. Finally, I thank my teacher, Richard Brauer, both for introducing me to character theory and for his large part in developing the subject which I find so fascinating. IX
This Page Intentionally Left Blank
Notation Mn(F) Xy Ay A° +.!■ M(V) ndV) J(A) C\(g) £„ XT o(g) Til) Xh det/ cox [x,y] v„(z) cp x 9 9° 9° Ga /g(9) algebra of n x n matrices over F linear transformation of the A-module V, induced by x e A {xv\x eA} regular A-module internal direct sum see Definition 1.12 see Lemma 1.13 the Jacobson radical, Problem 1.4 conjugacy class of g symmetric group of degree n transpose of the matrix X the order of g see Definition 2.26 restriction of / to H see Problem 2.3 the algebra homomorphism Z(C[G]) -» C induced by / x~ 1y~ 1xy, the commutator see Lemma 4.4 and the discussion preceding it see Definition 4.20 see Definition 5.1 induced character stabilizer of a in permutation representation inertia group, Definition 6.10 xi
Xll Notation o(x) determinantal order of /, equals the order of det / in group of linear characters Qk the field Q(e27"lk) Op(G), Op (G) minimal normal subgroups of p-power index and index prime to p R[£f] ring of /^-linear combinations of y £ Irr(G) R[_£T\° {9e/?[^]|9(l) = 0} P(G,je) Z[{lHG|He^}] 9*> 9*>gP> 9P' see the discussion following Lemma 8.18 3E£ see the discussion at beginning of Chapter 9 F(x) the field generated over F by the values of / rrifix) the Schur index, Definition 10.1 Z(G, A), B(G, A), H(G, A) see the discussion preceding Theorem 11.7 M(G) the Schur multiplier, Definition 11.12 A the group of linear characters of A Ch(G|9), Irr(G|9) see the discussion preceding Definition 11.23 coreG(H) f] HB for g e G cd.(G) {z(l)|zeIrr(G)} \{x) vanishing-off subgroup, <x e G | x(x) ^ 0> d.l.(G) derived length of G F(G) Fitting subgroup of G b(G) max(c.d.(G)) Op(G), Op(G) maximal normal subgroup of order a power of p, prime to p Irrs(G) the set of S-invariant / e Irr(G) <J>(G) Frattini subgroup of G IBr(G) the set of irreducible Brauer characters of G dXv decomposition numbers, Definition 15.9 <£,, the projective character associated with q> e IBr(G) B1(G) the set of p-blocks of G eB, XB the idempotent and algebra homomorphism of Z(F[G]) associated with B e B1(G) (5( Jf) the set of defect groups for the class Jf Jf the sum of the elements of Jf in F[G] a^Jf) coefficient of Jf in eB 8(B) the set of defect groups for B e B1(G) d(B) defect of BeBl(G)
1 Algebras, modules, and representations Character theory provides a means of applying ring theoretic techniques to the study of finite groups. Although much of the theory can be developed in other ways, it seems more natural to approach characters via rings (or more accurately, algebras). The purpose of this chapter is to provide the reader with the ring theoretic prerequisites needed in the rest of the book. Many of the results in this chapter are true in more general contexts than those considered here. Nevertheless, an effort has been made to avoid excess generality and to prove only that which will be needed later. (1.1) definition Let F be a field and let A be an F-vector space which is also a ring with 1. Suppose for all c e F and x,ye A, that (cx)y = c(xy) = x(cy). Then A is an F-algebra. We mention some examples of algebras over a field F: (a) M„(F) is the algebra of n x n matrices over the field F. (b) Let V be an F-vector space. Then End(l/), the set of F-linear transformations of V, is an F-algebra under the following conventions. If x, ye End(l/), then xy is defined by (v)xy = ((v)x)y and if c e F, then ex is defined by (v)(cx) = (cv)x. Of course, (v)(x + y) = (v)x + (v)y. This is a good place to digress briefly to discuss some notational conventions which will be used throughout this book. In writing scalar multiplication in a vector space, the scalar may be written on either side of the vector 1
2 Chapter 1 to which it is applied. Similarly, functions are written on whichever side is convenient, but the rule for function composition is always " fg means do / first, then g." Because of this rule, in those situations where function composition is important [as in example (b)], it will usually be convenient to write functions on the right. We now return to another example of an algebra, the one of primary importance for us: the group algebra. (c) Let G be a finite group. Then F[G] is the set of "formal" sums {Z»eG«99 \ag e F}- The structure of an F-vector space is given to F[G] in the obvious way and the element of F[G] for which ag = 1 and ah = 0 if h j= g is identified with g. This identification embeds G into F[G] and in fact G is a basis for F[G]. One result of this identification is to give a new meaning for £ agg. We may now view this expression not only as a formal sum, but also as an actual sum, a linear combination of the basis vectors. Finally, to define multiplication on F[G], we multiply the basis vectors according to their group multiplication and extend linearly to all of F[G]. It is routine to check that this defines the structure of an F-algebra on F\_G]- The construction of F\_G] suggests a general method of constructing algebras which should be mentioned. Let A be a finite dimensional F-algebra with F-basis vu ..., v„. We have then v^j = £ cijkvk, where cijk e F are the multiplication constants of A with respect to the basis {vt}. It is clear that these constants determine the algebra, so that any n-dimensional algebra may be specified by prescribing n3 constants cljk e F. Of course, only a small subset of all possible sets of constants define an algebra since most sets of constants define multiplications that turn out to be nonassociative. From now on, the word "algebra" in this book will mean a finite dimensional algebra. We make a few observations and definitions before going on to prove anything. Let A be an F-algebra. Then F • 1 = {c\ \c e F} is a subalgebra of A contained in the center Z(y4) since (c\)x = c(\x) = c(x\) = x(c\) for x e A. It is sometimes convenient to identify F with F • 1 and thus to view F as a subalgebra of A. If / is a left or right ideal of A as a ring and xe I and c e F, we have ex = (cl)x = x(cl)e I and / is a subspace. (If we had not required algebras to contain 1, then ideals would not automatically be subspaces.) If / is an ideal (this means two-sided), then A/I has the structure of an F- algebra in a natural manner. If A and B are F-algebras and q> is a ring homomorphism from A to B with <p(l) = 1, it is not necessarily true that cp is an F-linear transformation. (1.2) definition Let A and B be F-algebras. Suppose that <p: A -» B satisfies
Algebras, modules, and representations 3 (a) <p(xy) = <p(x)<p(y) for all x, y e A; (b) <p(l)=l; (c) cp is an F-linear transformation. Then <p is an algebra homomorphism or an F-homomorphism. (1.3) definition Let A be an F-algebra and let V be a finite dimensional F-vector space. Suppose for every v e V and x e A that a unique vx e V is defined. Assume for all x, y e A, v, w e V, and ceF that (a) (d + w)x = vx + wx, (b) u(x + y) = vx + vy, (c) (wc)>> = v(xy), (d) (cu)x = c(tuc) = v(cx), (e) t>l = v. Then F is an A-module. Let F be an A-module. Each x e A defines a map xv: V -» K by t> i—> dx. By (a) and part of (d) of the definition of a module, xve End(F). By (b), (c), (e), and part of (d), the map x \-* xv is an algebra homomorphism A -» End( V). Its image is denoted by Av. Some important examples of modules are the following. If A £ End(K), then 7 is an A-module in a natural way. If A = M„(F\ then the row space of dimension n over F is an A-module under matrix multiplication. If A is any algebra, then A itself is an A-module under right multiplication. This module is called the regular A-module and is denoted by A°. If V is an A-module and W £ V is an A-invariant subspace, then W is a submodule of V. Thus the right ideals of A are exactly the submodules of A°. If W is a submodule of V, then the space J^W becomes an A-module in the usual manner. Note that if / is a proper ideal of A, then the objects A/I, A°/I, and (A/If are all defined and all different, being respectively an algebra, an A-module, and the regular (A//)-module. However, A°/l is an A-module which is annihilated by / (i.e., if v e A°/I and xel, then vx = 0), and so it may be viewed as an (A//)-module. As such it becomes (A/If. If V and W are A-modules, a linear transformation cp: V -» W is an A-homomorphism if (p(vx) = (p(v)x for all v e V and x e A. An A-isomorphism is an A-homomorphism which is one-to-one and onto and, as is the usual situation, if V and W are A-isomorphic, they are exactly the same "as seen by A." For instance, in this situation they are annihilated by exactly the same set of elements of A. The set Hom^K W) of A-homomorphisms from V to W has the structure of an F-space by (ccp)(v) = c((pv) for ce F and (<p + 9)(t>) = (p(v) + 9(d). In addition Hom/1(K V) is a ring [remember, <p9 is defined by (v)(p& = (u<p)9], and in fact Hom^K V) is an F-algebra. It is exactly the centralizer
4 Chapter 1 of Av in End(F) and is denoted EA(V). If E = EA(V), then V is an E-module and Ef/iV) 3 Av. Later in this chapter we shall find a sufficient condition for equality here. (1.4) definition Let V be a nonzero ,4-module. Then V is irreducible if its only submodules are 0 and V. It is obvious that if cp e Hom^CV, W) then ker cp and im (p are sub- modules of V and W, respectively. The following important lemma is now immediate. (1.5) lemma (Schur) If V and W are irreducible A -modules, then every nonzero element of Hom/1(K W) has an inverse in Hom/1(VK V). An immediate consequence of Schur's lemma is that if V is an irreducible A-module, then EA(V) is a division algebra, i.e., every nonzero element is invertible. (1.6) corollary Let F be algebraically closed, A an F-algebra, and V an irreducible A-module. Then EA(V) = F-1, the set of scalar multiplications on V. Proof Clearly, F ■ 1 £ EA(V). Let 9 e EA(V). Then 9 is a linear transformation of a finite dimensional vector space V over F and so has an eigenvalue L Then 9 — 11 e EA(V) and is not invertible. Thus 9 — 11 = 0 and 9 = 11 e F • 1 as claimed. | To justify the study of irreducible modules we remark that for certain algebras, every module is a direct sum of irreducible ones, and thus to know all irreducibles is to know all modules for these algebras. It happens that group algebras over fields of characteristic zero are among these fortunate algebras. This will follow from Maschke's theorem which will be proved shortly. (1.7) definition Let V be an A-module. Suppose for every submodule W £ V, there exists another submodule U £ V such that V = W + U (the dot indicating that the sum is direct). Then V is a completely reducible module. Observe that irreducible modules are completely reducible as are all modules over fields (i.e., vector spaces). (1.8) definition An algebra A is semisimple if its regular module, A0 is completely reducible. (1.9) theorem (Maschke) Let G be a finite group and F a field whose characteristic does not divide \G\. Then every F[G]-module is completely reducible.
Algebras, modules, and representations S Proof Let V be an F[G]-module with submodule W. Let U0 be an F-subspace of V which is complementary to W, i.e., V = W + U0. Let cp be the projection map of V onto W with respect to U0. Now cp is a linear transformation which is not necessarily an F[G]-homomorphism. The object of the remainder of the proof is to modify (p in order to create, an F[G]-homo- morphism. Define 9: V -+ W by Clearly, 9 is F-linear. To show that 9 is an F[G]-homomorphism, we compute |G| 9EG I O | g since hg runs over G as g does for fixed /j. If w e H< we have wg eW for all g e G and thus (p(wg) = wg. It follows that 9(w) = w. Now let U = ker 9, an F[G]-submodule of V. We have 9(d) e VK so that 9(9)d)) = 9(d) and 9(d - 9(d)) = 9(d) - 9(d) = 0. Thus v = 9(d) + (d - 9(d)) eW+U and V = W + U. Finally, if w e W n U, we have w = S(w) = 0, so that V = W + V and the proof is complete. | A consequence of Maschke's theorem is that F[G~] is semisimple if char(F)^|G|. The converse of this statement is also true and the reader is referred to the problems for a proof. (1.10) theorem Let V be an ^module. Then V is completely reducible iff it is a sum of irreducible submodules. Proof Suppose V = Y,K where the Va are irreducible. Let W £ V. By finite dimensionality, choose U £ V maximal such that W n U = 0. We claim that W + U = V. Otherwise, we may choose Va <£. W + U and thus (W+U)nVa = 0 by the irreducibility of Fa. It follows that W r\ (U + Va) = 0 and this violates the maximality of U. Thus V = W + U and V is completely reducible. Conversely, suppose V is completely reducible and let S be the sum of all of the irreducible submodules of V. If S < V, we may write V = S + T with T t6 0. By finite dimensionality, T contains an irreducible submodule which is a contradiction since T n, S = 0. | (1.11) lemma Let V be an ,4-module and suppose V = Y,K where the Va are irreducible submodules. Then V is the direct sum of some of the V^s. Proof Choose W £ V maximal with the property that W is the direct sum of some Vx's. If W < V, then Vx £ W for some a. Since Va is irreducible,
6 Chapter 1 we have W r\ Va = 0 and Wi = W + Vx > W. This violates the maximality of W and we conclude that W = V as desired. | We see from the previous two results that the completely reducible modules are exactly the direct sums of irreducible modules. It follows that in order to know all modules for a group algebra over a field of characteristic 0, it suffices to know all irreducible modules. (1.12) definition Let V be a completely reducible ,4-module and let M be an irreducible A-module. The M-homogeneous part of V, denoted M(V), is the sum of all those submodules of V which are isomorphic to M. Observe that if M s N, then M(V) = N(V). As will be shown shortly, if M and N are nonisomorphic irreducible /4-modules, then M(V) n, N(V) = 0. If V has no submodules isomorphic to M, then M(V) = 0. (1.13) lemma Let V = £• Wi be a direct sum of ,4-modules with W( irreducible for all i. Let M be any irreducible ,4-module. Then (a) M(HisanE/1(F)-submoduleof V; (b) M(V) = li{Wt\Wl*Mh (c) The number nM(V) of Wt which are isomorphic to M is an invariant of K independent of the given direct sum decomposition. Proof (a) Let 9 eEA(V). We need to show M(V)$ £ M(V). It suffices to show that if W £ V and W s M, then W& £ M(V). This is sufficient because M(V) is the sum of such W. If W9 = 0, there is nothing to prove; if W9 t6 0, then since W9 is a homomorphic image of the irreducible module W, we have W& S W S M. Thus W9 £ M(7). (b) Clearly, £ {V^ | Wt = M} £ M(V). Let ^ be the projection map of F onto Wt. Now let W£ K W^ M. If nj(W) ± 0, then nfW) = Wj and W £ W). Thus TtjiW) £ X {Wil Wt s M} for all j. However, W £ £, n/W) and hence W £ ^{H^| Wt S M}. It follows that M(V) £ £ Ml Wt S M}. (c) By (b), we have dim M(F) = nM(V) dim M and it is immediate that nM(V) is an invariant. | By part (b) of Lemma 1.13 it follows that a completely reducible module is the direct sum of its M-homogeneous components for distinct M. In particular, M(V) n N(V) = 0 if M ik N. We wish to discuss "all irreducible /4-modules." This poses some difficulties. First, we really mean "all isomorphism classes of irreducible A -modules." It would be convenient to have a representative set of /4-modules. By this we mean a set Jl{A) of irreducible /4-modules with the property that every irreducible y4-module is isomorphic to exactly one element of Jt{A\ Since modules are external to the algebra, it is not clear how one can find a
Algebras, modules, and representations 7 representative set for a given algebra A. To do this we need to produce a set of A -modules large enough to contain copies of all irreducibles. The next lemma shows that the set of homomorphic images of A° suffices. (1.14) lemma Let A be an F-algebra. Then every irreducible ,4-module is isomorphic to a factor module of A°. \{A is semisimple, then every irreducible A -module is isomorphic to a submodule oi A0. Proof Let V be an irreducible A-module, choose 0 ^ v e V, and define ft A -» V by 9(x) = vx. Then 9 is clearly F-linear and 9(xy) = vxy = 9(x)y so that 9 e HomA(A°, V). Now, v e im 9 £ V and hence im 9 = V since V is irreducible. Let W = ker 9. Then V = A°/W. If A is semisimple, then A0 = W + U and /T/W s I/. The proof is complete. | Now fix a representative set M{A) of irreducible A-modules. By part (b) of Lemma 1.13, we have V = £• MejnA)M(V) for every completely reducible A-module V. Suppose A is a semisimple algebra. We can then apply the above to A0 and write A° = £ • M(A°). It turns out that M(A°) is actually a two-sided ideal of A. For notational convenience, we write M(A) for M(A°) in what follows. (1.15) theorem (Wedderburn) Let A be a semisimple algebra and let M be an irreducible A-module. Then (a) M(A) is a minimal ideal of A; (b) if W is irreducible, then it is annihilated by M(A) unless W s M, (c) the map x i-> xM is one-to-one from M(A) onto y4M £ End(M); (d) Jt{A) is a finite set. Proof If x e A, the map 9*: y i—> x_y satisfies 9* e E^A0). Therefore, by Lemma 1.13(a), xM(A) = M(A)$X £ M(A) and M(A) is a left ideal. Since M(A) is a submodule of A0, it follows that it is an ideal of A. Minimality will follow after (c) is proved. If W is an irreducible A-module with W yk M, then W(A) r\ M(A) = 0 by Lemma 1.13(b). Since W(A) and M(A) are ideals, we have W(A)M(A) = 0. By Lemma 1.14, A0 has a submodule W0 s W and W0 £ W(A), so that M(A) annihilates WQ. Since W = W0, they have the same annihilator in A and (b) follows. By (b), it follows that xw = 0 if x e M(A) and M ^ W.lt now follows from the direct sum decomposition A = £ • M€^u)W(^)tnat f°r y e ^> we have ^ = xM, where x is the component of y in M(A). We conclude that the map xbxm maps M(A) onto AM. If x e M(A) and xM = 0, then it follows from (b) that x annihilates every irreducible, and hence every completely reducible A -module. Thus x= lx£y4°x = 0 and (c) is proved.
8 Chapter 1 To show the minimality of M(A), let I < M(A) be an ideal of A. Now M(A) is a sum of submodules isomorphic to M and thus there exists M0 £ M(A), M0 = M,M0^/.AsM0n/< M0 and M0 = M is irreducible, we have M0 n 7 = 0. Thus M07 £ M0 n J = 0 and 7 annihilates M0 and hence also M. Therefore, if x e 7, we have xM = 0. However, x h-» xm is one-to-one for x e M(y4) and thus x = 0. We conclude that 7 = 0 and the minimality of M(A) is proved. Finally, M(A) =£ 0 for every M by Lemma 1.14, and yet A = £• M€.*u)W(^4) is finite dimensional. It follows that |^(X)| is finite and the proof is complete. | Observe that each M(A) is actually an algebra, its unit element being the component of 1 in M(A) under the decomposition A = £ • M(A). Since the map x i—> xM is an algebra homomorphism from A to y4M, it follows from (c) of the theorem that the restriction of this map to M(A) is an algebra isomorphism from M(A) onto Au. Since M1(A)M2(A) = 0 for Mj % M2, it follows that every ideal of the algebra M(A) is in fact an ideal of A. We conclude from the minimality of M(A) as an ideal of A, that M(A) is a simple algebra, i.e., it has no nontrivial proper ideals. Thus the preceding theorem asserts (among other things) that a semisimple algebra is a direct sum of simple algebras. This is a somewhat more usual statement of Wedderburn's theorem. (It is also true that every simple algebra is semisimple. This follows from Problem 1.5.) To review the situation now: group algebras over fields of characteristic zero are semisimple; semisimple algebras are direct sums of ideals M(A) and M(A) is naturally isomorphic to Au. What remains is to study the simple algebras Au. This is the purpose of the "double centralizer" theorem which follows. The proof we give here is based on an idea of M. Rieffel. It should be remarked that the hypothesis that A is semisimple is actually superfluous since the theorem is really about AM which is automatically semisimple when M is completely reducible. See Problem 1.6. (1.16) theorem (Double Centralizer) Let A be a semisimple algebra and let M be an irreducible A-module. Let D = EA(M). Then ED(M) = AM. Proof It is no loss to replace M by an isomorphic module, and so by Lemma 1.14, we may assume that M £ A°. Let I = M{A) so that M £ 7. It is clear that AM £ ED(M) so we prove the reverse inclusion. Let 9 e ED(M) so that (ma)9 = (m9)a for a e D. If m e M, define am: M -» A by (x)am = mx. Since m e M is a right ideal of A, we have mx e M and am: M -> M. If a e A and x e M, we have (xa)oim = m(xa) = (mx)a = (xa.m)a. It follows that am e EA(M) = D. Thus for m,ne M, we have (*) (mn)S = (nam)9 = (n9)am = m(n9>).
Algebras, modules, and representations 9 Now fix n e M with n#0 and let e be the unit element of /. We have AnA £ / and by the minimality of the ideal /, we have e e I = AnA and e = £ ainbi for suitable ah bt e A. Urns M, we have m = me = m £ «,-«£>,• = X(ma;)(nfc,). Since mat e M and nbt e M, Equation (*) yields for all m e M that m9 = X {{mad(nbd)9 = I NIW = m £ a,.((nfc;)9). Thus 9 = mm e y4M where u = £ aj((nfcj)9). The proof is complete. | (1.17) corollary Let A be a semisimple algebra over an algebraically closed field F and let M be an irreducible A-module. Then (a) AM = End(M); (b) dim(AM) = dim(M(A)) = dim(M)2; (c) nM(A°) = dim(M). Furthermore, if M{A) is a representative set of irreducible ,4-modules, then (d) dim(A) = XM€^M>dim(M)2, (e) dim(Z(A)) = |uT(X)|. Proof By Corollary 1.6, EA(M) = F ■ 1 for irreducible M and thus y4M = EF.i(M) = End(M). By elementary linear algebra, End(M) s M^F), the algebra of d x d matrices, where d = dim M. Thus dim AM = dim(End(M)) = d1. Since M(A) s Au, we have (b). Now M(A) is the direct sum of nM(A°) isomorphic copies of M, thus d2 = dim M(A) = nu(A°) dim M = dnM(y4°), and (c) follows. Since A = £ • Me.*M)Af(/4), (d) is immediate from (b). Finally, let ZM = HM(A)\ Now Z{AM) = AM n E^M) = Au n F • 1 = F • 1. Thus dim ZM = dim(ZUM)) = 1. Clearly, £ • M,M(A)ZM <= Z(A) and dim(£ ZM) = l-^(4)|. However, if z e Z{A), write z = £ aM, aM e M(A). If w e M(y4), then waM = wz = zu = aMu since the distinct M(A)'s annihilate each other. Thus aM e ZM and £ ZM = Z{A). The proof is complete. | Although the word "representations" constitutes one third of the title of this chapter, so far nothing has been said about them. In fact, that isn't really true, because as we shall see, representations are just a different way of looking at modules. (1.18) definition Let A be an F-algebra. A representation of A is an algebra homomorphism 3E: A -» M„(F). The integer n is the degree of 3E. Two representations 3E, 9) of degree n are similar if there exists a nonsingular n x n matrix P, such that 3E(a) = P~l<!0(a)P for all a e A.
10 Chapter 1 Clearly, similarity is an equivalence relation among representations. Also, if 9) is a representation of degree n and P is any nonsingular n x n matrix, then the formula 3E(a) = P~ ll$)(a)P defines a new representation 3E. It is easy to build modules from representations and representations from modules. If 3E is a representation of degree n of the F-algebra A, let V be the n-dimensional row vector space over F. If v e Fand X is any n x n matrix over F, then vX e V. Define va = vH{a) for aeA.lt is routine to check that this gives the structure of an ,4-module to V. Conversely, if M is an A-module, choose an F-basis for M and let 3E(a) be the matrix of au with respect to this basis. It is now easy to check that 3E is a representation. Note that a different choice of basis might give a different representation (and usually does). Starting with a representation 3E, constructing the module V as above, and then choosing the appropriate basis for V and constructing the corresponding representation will result in the original representation 3E. Suppose V and W are A-modules and 9e Hom^K W). How does this situation look from the representation point of view? Choose bases 6Y and 6W for V and W. Since 9 is a linear transformation from V to W, it has a matrix P with respect to the given bases. Note that P is m x n where m = dim V and n = dim W. Now, the fact that (va)S = (v$)a for all veV and a e A yields 3E(a)P = Ptyla), where 3E and 9) are the representations given by V and W with respect to the bases 6V and 6W. If 9 is a module isomorphism, then P is nonsingular and the above matrix equation yields that 3E and 9) are similar representations. In particular, the different representations arising from a given module with different choices of basis are all similar. The above reasoning may be reversed to show that if the representations arising from V and W with respect to bases 6V and Sw are similar, then V and W are in fact isomorphic modules. It follows that there is a natural one-to-one correspondence between isomorphism classes of A-modules and similarity classes of representations of A. If V is an /4-module and W < V is a proper nonzero submodule, choose a basis 6W for W and extend this to 6V, a basis for V. Number 6V so that the last m vectors are £w, where m = dim W. Let 3E be the representation of A corresponding to V with respect to the basis 6V and let 9) be the representation corresponding to W with respect to the basis 6W. It is then easy to see for a e A that 3E(a) has the form (3(a) U(aj Furthermore, 3 is a representation corresponding to V/W. Note that U is a function from A into (n — m) x m matrices (where n = dim V), but U is not a representation.
Problems 11 The representation 3E is said to be in reduced form and one similar to 3E isreducible. Thus the irreducible representations correspond to the irreducible modules. If there exists a submodule U £ V in the above situation, with V = W + U, then the basis for W may be extended to V by adjoining to it a basis for U. When this is done, the result is that U(a) = 0 for all aeA.lt follows from this discussion that if 3E is any representation corresponding to a completely reducible module, then 3E is similar to a representation in block diagonal form, where each of the blocks is an irreducible representation. Problems (1.1) Let V be an A-module. Show that V is completely reducible iff the intersection of all of the maximal submodules of V is trivial. Hint To prove "if," embed V into a sum of irreducible modules. Recall that our definition of "module" requires finite dimensionality. Note This problem is "dual" to Theorem 1.10 which implies that V is completely reducible iff it is the sum of its minimal submodules. Theorem 1.10 is true much more generally than we have proved it. It holds for arbitrary modules over rings. Problem 1.1, however, is not valid in this greater generality. A counterexample is the regular module of the ring of integers Z. (1.2) Let 3E and 9) be representations of an F-algebra, A. A nonzero matrix P is said to intertwine 3E and 9) if PH(a) = ^(afP for all a e A. Assume 3E and 9) are irreducible. (a) If P intertwines 3E and 9), show that P is square and nonsingular. (b) Assume that F is algebraically closed and that P and Q both intertwine 3E and 9). Show that Q = XP for some X e F. (1.3) Show that an algebra A is semisimple iff every A-module is completely reducible. (1.4) Let A be an algebra. For A-module V, let d{V) = {a e A\ Va = 0}. Let J(A) = C\mzjha)<s4(M), where Jt(A) is a representative set of irreducible A-modules. Show (a) s#(V) is an ideal of A for all V. (b) VJ(A) < V for every nonzero A-module V. (c) J(A)n = 0 for some integer n. (d) If / is a right ideal of A and Im = 0 for some m, then / £ J(A). Note The ideal J(A) is called the Jacobson radical of A; jtf(V) is the annihilator of V and a right ideal / with I" = 0 is said to be nilpotent.
12 Chapter 1 (1.5) Prove that the following are equivalent for the algebra A. (a) J(A) = 0. (b) A has no nonzero nilpotent right ideals. (c) A has no nonzero nilpotent ideals. (d) A is semisimple. Hint If V is irreducible, then sf(V) is an intersection of maximal right ideals of A. (1.6) Let A be an algebra and V a completely reducible A-module. Show that the algebra Av is semisimple. Note A consequence of Problem 1.6 is that the hypothesis that A is semisimple in the Double Centralizer Theorem 1.16 may be dropped since this theorem is really about AM for an irreducible module M. (1.7) Let A be an algebra and V an irreducible A-module. Show that \Jf(Ay)\ = 1. (1.8) Let G be a group, HsGa subgroup and F a field with characteristic prime to | G: H |. Let V be an F[G]-module with submodule W. Suppose that there exists U0 £ V, an F[H]-submodule such that V = W + U0. Show that there exists an F[G]-submodule U £ V with V = W + U. Note This generalization of Maschke's Theorem 1.9 is due to D. G. Higman. (1.9) Let G be a group and F a field of characteristic p. Suppose p\\G\ and show that F[G] is not semisimple. Hint (I9eG0)2=O. (1.10) Let M be an A-module. Show that M is completely reducible iff MJ(A) = 0. Hint A/J(A) is semisimple.
Group representations and characters Let G be a finite group and let F be a field. Suppose 3E is a representation of F[G] with degree n. Since 3E is an algebra homomorphism, X(l) = /, the identity matrix. It follows for g e G that 3E(g) is nonsingular and 3E(g)_1 = X(g~'). If we restrict the function 3E to G £ F[G], we obtain a group homomorphism from G into the general linear group GL(n, F), that is, the multiplicative group of nonsingular n x n matrices over F. (2.1) definition Let F be a field and G a group. Then an F-representation of G is a homomorphism 3E: G -» GL(n, F) for some integer n. We have seen that a representation of F[_G] determines an F-representation of G by restriction. Conversely, an F-representation 3E0 of G determines a representation 3E of F[G] by linear extension. That is, We shall usually use the same symbol to denote both an F-representation of G and the corresponding representation of F[G~\. Also, the adjectives "similar" and "irreducible" will be applied to F-representations of G as if they were the corresponding representations of F[G~\. Some caution is necessary here since if F £ E, a larger field, and 3E is an F-representation of G, then 3E is automatically an E-representation. It is entirely possible, however, that 3E is irreducible as an F-representation, and yet is reducible as an E-representation. We will explore this situation in some depth in Chapter 9. One further triviality which should be mentioned now is the following. If JV o G and 3E is an F-representation of G with JV £ ker 3E, then there is a 13
14 Chapter 2 unique F-representation X of G/N defined by X(Ng) = X(g). This formula can also be used to define the representation 3E if J is given. Note that 3E is irreducible iff J is. We shall often fail to distinguish between 3E and X. The trouble with representations is that they contain too much information. If 3E is an F-representation of G of degree n, then for each element of G we have n2 entries in X(g). Some of this data is clearly redundant because it distinguishes between similar representations. The idea behind character theory is to throw away most of the information and to save just enough to be useful. This is done by calculating the traces (that is, the sums of the diagonal entries) of the matrices in question. Recall that if A and B are any two n x n matrices over a field, then tr(y4B) = tr(BA). (2.2) definition Let 3E be an F-representation of G. Then the F-character X of G afforded by 3E is the function given by 1(g) = tr X(g). As is the case with F-representations of G, we may view F-characters as functions on all of F[G]. Note that if the characteristic char(F) ± 0, then the constant function 0 is an F-character. On the other hand, if char(F) = 0, then 0 is definitely not an F-character because x(l) = deg 3E, where 3E is an F-representation of G which affords /. In this case, we say that x(l) is the degree of /. Most F-characters of a group G are not homomorphisms of any kind. However, if 1 is a homomorphism from G into the multiplicative group of F, then 3E(g) = (1(g)) is an F-representation of G of degree 1 which affords X as its character. Characters of degree 1 are called linear characters. In particular, the function lc with constant value 1 on G is a linear F-character. It is called the principal F-character. (2.3) lemma (a) Similar F-representations of G afford equal characters, (b) Characters are constant on the conjugacy classes of a group. Proof UP is nonsingular then tr(P~ 1AP) = tr(P ■P~1A) = tr(A). Both (a) and (b) follow from this observation. To see (b), observe that X(h~ lgh) = X(h)~ lX(g)X(h) if 3E is a representation of G, and hence tr(X(h~1gh)) = tr(X(g)). | We make one further general observation. If 3E and 9) are F-representations of G, then is also an F-representation. Since tr 3(#) = tr X(g) + tr ?)(#), it follows that the set of F-characters of G is closed under addition.
Group representations and characters IS We now restrict our attention to the special case that the field F = C, the complex numbers. We emphasize, however, that the subfield consisting of the algebraic elements in C would work exactly as well, and in fact with only minor modifications, most of what follows works for any algebraically closed field of characteristic not dividing \G\. Let us establish some notation. Fix a finite group G and choose a representative set of irreducible C[G]-modules, Jt(C[G]) = {Mu ..., MJ. Choose a basis in each M, and let 3E; be the resulting representation of C[G]. Let Xt be the character afforded by 3Ej. It follows that the set Irr(G) = {*i> • • • > Xk) is the set of all irreducible C-characters of G (that is, characters afforded by irreducible representations). Henceforth, the word "character" will mean C-character unless otherwise stated. Since sums of characters are characters, it follows that / = £?= j n,x, is a character whenever the n, are nonnegative integers which are not all zero. Conversely, if / is any character of G afforded by a representation 3E corresponding to a module K we can decompose V into a direct sum of irreducible modules. It follows that / is the sum of the corresponding irreducible characters. We have, in fact, / = £ nMi(V)x,. Corollary 1.17(d) asserts that dim(C[G]) = X?=1(dim M,-)2. Since dim C[G] = \G\ and dim M, = deg £, = x,(l), we obtain the fundamental formula |G|= £>,-(l)2. i= 1 It seems natural at this point to ask how we can determine the integer k purely group theoretically, without looking at representations. By Corollary 1.17(e), we have k = |^(C[G])| = dim Z(C[G]). (2.4) theorem Let Jf i, jf2»• • •■» ^r be the conjugacy classes of a group G. Let X, = £«*-, * e C[G]. Then the X, form a basis for Z(C[G]) and if KtKj = Y,fl,jv^v> then the multiplication constants a,JV are nonnegative integers. Proof It is clear that the K; lie in Z(C[G]). Moreover, they are linearly independent because they are sums of disjoint sets of elements. If z = £ agg e Z(C[G]) and h e G, we have z = h~ 1zh = £ aggh. Comparing the coefficients of gh on both sides, we obtain agh = ag. In other words, the coefficients ag have the constant value a, for all g e if;. It follows that z = X a(Ki and thus the X, span Z(C[G]). To find aiJV, pick g e Jfv. Then aijv is the coefficient of g in KtKj. From the definition of multiplication in a group algebra, this is | {(x, y)\x e Jf h y e Jfy, xy = g}\. Since ai]v is the cardinality of a set, it is a nonnegative integer. |
16 Chapter 2 (2.5) corollary The number k of similarity classes of irreducible representations of G is equal to the number of conjugacy classes of G. (2.6) corollary The group G is abelian iff every irreducible character is linear. Proof Let k be the number of classes of G. Then k = \ G | iff G is abelian. Now \G\ = Yj=i Z;(l)2 and x,-(l) > 1 for all i. It follows that k = \G\ iff Xi(l) = 1 for all i. The proof is complete. | We have not yet proved that the & are distinct. To see this, we introduce a little more notation. From the results of Chapter 1 we have the direct sum C[G]= i-M;(C[G]). i=i Let 1 = £ et with et e M;(C[G]). Since M/C[G]) annihilates the module Mi if i t6 j, we have 3E;(e;) = 0 in this case. It follows that 3£,<e,) = 3E;(1) = /. Therefore, ^(e,) = 0 if i ^ j and x,(e,) = /;(1) =£ 0 and we conclude that the /,- are distinct as functions on C[G]. Thus the /,- are also distinct as functions onG. We may now restate two of the earlier results in a slightly more convenient form. (2.7) corollary Let G be a group. Then | Irr(G) | equals the number of conjugacy classes of G and I Z(D2 = |G|. Xelrr(G) For certain very small groups, the information contained in Corollary 2.7 is sufficient to determine the irreducible character degrees. For instance, if G = Z3, the symmetric group on three symbols, then G has exactly three conjugacy classes and \G\ = 6. It follows that the /;(1) are 1,1, and 2. Actually, the preceding argument with the e,'s yields more than the fact that the Xis are distinct. A class function on a group G is a function, cp: G -» C which is constant on conjugacy classes. All characters are class functions. (2.8) theorem Every class function <p of G can be uniquely expressed in the form 9= Z axK Xelrr(G) where ax e C. Furthermore, q> is a character iff all of the ax are nonnegative integers and (^#0.
Group representations and characters 17 Proof The set of class functions of G forms a vector space over C whose dimension is the number of classes of G. We claim that Irr(G) is a basis for this space. Since |Irr(G)| = k = number of classes, it suffices to show that if £ atXi = 0, then each a; = 0. This is immediate by evaluation at e;. The second statement has already been proved. | If i = Yj= i «iZ; is a character, then those & with nt > 0 are called the irreducible constituents of /. In general, if ip is a character such that / — \p is also a character or is zero, then \p is called a constituent of /. An important consequence of Theorem 2.8 is the following. (2.9) corollary Let 3E and 9) be C-representations of a group G. Then 3E and 9) are similar iff they afford equal characters. Proof We already know that similar representations afford equal characters. Let V and W be C[G]-modules corresponding to 3E and 9) respectively. Then 3E affords the character £ nMi(V)x,- and 9) affords £ nMi(W)Xi- If these characters are equal, it follows that nMi{V) = nMi(W) for all i. Therefore, V and W are isomorphic to identical direct sums of irreducible C[G]- modules and thus are isomorphic and 3E and 9) are similar as required. | For a particular group G, the irreducible characters are usually presented in a character table: a (square) array of complex numbers whose rows correspond to the /, and whose columns correspond to the classes Jft. An example of a character table is the accompanying one for the symmetric group Z4 on four symbols. g: 1 (12) (1 2)(3 4) (12 3 4) (12 3) icto)i \Cl(g)\ Xi ll X.3 n Xs 24 1 1 1 2 3 3 In this table, the classes are denoted by writing a representative element g in its cyclic notation. (The reader is reminded that in a symmetric group, two elements are conjugate iff they have the same cycle structure.) The sizes of the centralizer C(g) and of the corresponding conjugacy class Cl(g) are
18 Chapter 2 given for convenience, although they are not, properly speaking, a part of the table. Further character tables are given in the Appendix (page 287). It is rather difficult to describe the process by which these tables are constructed. Usually, various combinations of ad hoc arguments and general theorems are necessary. As the student learns some of these theorems, he is urged to try to construct some character tables on his own. The important point is that it is very much easier to construct a character table than it is to construct representations. Theorem 2.8 tells us that every class function is a linear combination of irreducible characters. For example, if G = E4 and (p(g) is defined to be the number of points moved by g e G, then cp is a class function and is a linear combination of %lt /2, • • •, Xs • Thus the row (0,2,4,4, 3) is a linear combination of the five rows of the given table. It turns out that there is an easy method for computing the coefficients, practically by inspection from the table. This is done by using the so called "orthogonality relations," which we are about to derive. These relations are also extremely useful in the construction of character tables. Before we leave Z4 we should comment on the fact that all of the character values, which a priori are only known to be complex numbers, in fact turn out to be integers. (Of course the xf(l), being the degrees of representations, are always positive integers.) It is not always true that all character values are ordinary integers, although frequently a large fraction of them are. It is true, however, for all symmetric groups. All of this should become clear later. The key to the orthogonality relations is to compute explicitly the coefficients of the group elements in the e,'s in terms of the characters. To do this we use the character p of G afforded by a representation corresponding to the regular module C[G]°. This regular character p will be computed in two ways. (2.10) lemma IfgeGandg ^ 1, then p(g) = 0. Also p(\)= \G\. Proof We must choose a basis for C[G]° to obtain a corresponding representation. We simply take G, in some ordering, as the basis and let 5R be the corresponding representation. If 5R(g) = (ay), then ay = 0 unlessgtg = gr in which case ay = 1. Since p(g) is the number of gt satisfying g,g = gh the lemma follows. | Since p is a character of G, it may be expressed as an integer linear combination of the Xi- We do this explicitly. (2.11) LEMMA P = JJ-iZl(l)Zl-
Group representations and characters 19 Proof If V is any C[G]-module, it may be decomposed as a direct sum of irreducibles. The character afforded by a corresponding representation is£nM,(P0Xi- Now by Corollary 1.17(c) we know that nMi(C[G]°) = dim M; = x,(l). The result follows. | It is suggested that the reader use Lemma 2.11 and the character table of Z4 to compute p explicitly for this group, and check the result against Lemma 2.10. (2.12) theorem e, = (1/IGI) L,eG U\)l,\g-l)g. Proof Write et = Y,ag9- By Lemma 2.10, we have p(etg~l) = ag\G\. Lemma 2.11 thus yields ag\G\ = Y.xmj{eig-1). j Since (0 if i*j, wehavex/e.g-1) = xt(g~ '^y, where the Kronecker (5,7is0or 1 depending on whether i and j are unequal or equal. We now have aB\G\ = xi(l)Xi(g-1) and the result follows. | (2.13) theorem (Generalized Orthogonality Relation) The following holds for every he G. ^Ixi(gh)xj(9~1) = ziM. Proof The e{ lie in trivially intersecting ideals of C[G] and thus etej = 0 if i t6 j. Since 1 = £ e-p multiplication by et yields et2 = et. We now substitute the formula of Theorem 2.12 into the equation e,ej = (5l7e; and compare the coefficients of the group elements on both sides. The coefficient of a fixed h e G on the right hand side is (d 0/| G |)/j( 1 )x;(h ~1) and that on the left-hand side is |G| gsG The result now follows by equating these expressions and substituting h for/T1. |
20 Chapter 2 By taking h = 1 in Theorem 2.13 we obtain (2.14) corollary (First Orthogonality Relation) rprr Exited-1) = <V |G| 9€G Since the expression %(g~') has come up several times, we digress for a while to discuss its connection with %(g) and some related questions. (2.15) lemma Let 3E be a representation of G affording the character / and let g e G. Let n = o(g\ the order of g. Then (a) 3E(g) is similar to a diagonal matrix diagta,..., ef); (b) e," = 1; (c) X(0) = le,and|x(g)| <;*(l); (d) X(g~1) = M- Proof The restriction of 3E to the cyclic group <g> is a representation of <g> and hence it is no loss to assume G = <g>. By Maschke's theorem and other results of Chapter 1, it follows that 3E is similar to a representation in block diagonal form, with irreducible representations of G appearing as the diagonal blocks. Since G = <g> is abelian, Corollary 2.6 asserts that its irreducible representations have degree 1, and thus 3E is similar to a diagonal representation. Now (a) follows, and we may assume that 3E is diagonal. We have / = 3E(g") = X(gf = diag(e1",..., e/). Therefore (b) is proved. It follows that |e,| = 1 and |£e,| ^ Yje,-| = / = *0). It is clear that x(g) = £ e, so that (c) follows. Now 3E(gf_1) = 3E(g)_1 = diag(e1_1,..., Ej-'1) _so that Xig"1) = Ze._1- Since |e,| = 1, we have e,"1 = e^ and i(g~l) = i(g). The proof is complete. | Combining Corollary 2.14 and Lemma 2.15(d), we obtain I"! 06G This suggests the following definition. (2.16) definition Let (p and 9 be class functions on a group G. Then 0,9] = -!- 2>(3)%) |G| gsG is the inner product of q> and 9. Some of the obvious properties of this "inner product" are (a) [>,8] = E5T£I; (b) [<p, <p] > 0 unless <p = 0;
Group representations and characters 21 (c) [£-!<?! + c2(p2, 9] = CjE^j, 9] + c2[<p2, 9]; (d) O, CjSj + c292] = c7[<p, 8,] + c2[_q>, S2]. Therefore, [ , ] has all of the properties usually used to define an inner product in linear algebra and analysis. (In fact, this makes the space of class functions into a finite dimensional Hilbert space.) We know that Irr(G) is a basis for the space of class functions and it is the content of the orthogonality relation that it is, in fact, an orthonormal basis, that is, [Zi.Zy] = <V This yields the promised method for expressing an arbitrary class function in terms of the irreducible characters; for if [<p, /,] = q, then (p = £ c,Xj. Another application of the inner product is to determine instantaneously whether or not a given character is irreducible. (2.17) corollary Let x and \ji be (not necessarily irreducible) characters of G. Then [/, i/f] = [i/f, /] is a nonnegative integer. Also / is irreducible iff [*. X] = 1- Proof We have / = £ n,-/,- and \p = £ »>,•/,-, with all n{ and m; non- negative integers. Then [/, t/f] = £ n,mj = |>, /] and [*, *] = £ n;2. The result is now immediate, since / is irreducible exactly when one nt = 1 and all other nt = 0. | The following "second orthogonality relation" is derived from the first and so imposes no new necessary condition for an array of complex numbers to be a character table. Nevertheless, it is often extremely useful in the construction of character tables and in the extraction of information from them. (2.18) theorem (Second Orthogonality Relation) Let g,heG. Then I x(g)W) = o XeIrr<G1 if g is not conjugate to h in G. Otherwise, the sum is equal to |C(g)|. Proof Let gu g2,..., gk be representatives of the conjugacy classes of G. Let X be the k x k matrix whose (i,j) entry is Xi(9j)- (In other words, X is the character table, viewed as a matrix.) Let D be the diagonal matrix .with entries (5iy | Jf; | where Jf,- is the conjugacy class Cl(g,). The first orthogonality relation asserts that \g\s,]= Y.xte)xjg)= El^vlzitev)JW gsG v=l
22 Chapter 2 This system of k2 equations may be replaced by the single matrix equation |G|/ = XDXT, where / is the identity matrix and the superscript denotes transpose. Since a right inverse for a square matrix is necessarily also a left inverse, this yields |G|/ = DXTX. We now write this as a system of equations and obtain \G\8„ = l.\jrl\xJgdx*(gj)- V Since | G | /1 Jf; | = | Qsi) I, this yields I xtiX)=\Qg,)\&i}, *eIrr(G) which is the desired result. | In the character table for Z4 that was given earlier, the size of each con- jugacy class was given. We see now that this information is derivable from the body of the table. It was given only for ease of computation of inner products. As a check on our results so far, observe what happens if we take h = 1 in the second orthogonality relation. The reader should recognize what results as a combination of Lemmas 2.10 and 2.11. Let £ £ C be the field of algebraic numbers. By Lemma 2.15, all of the character values %(g) e E for / e Irr(G) and g e G. How is Irr(G), which is a set of functions G -» E related to the set of irreducible £-characters of G, which we denote by Irr£(G)? In fact these sets are equal. Suppose 3E is an irreducible £-representation of G which affords / e Irr£(G). Now X may be viewed as a C-representation and thus / is a character (that is, a C-character) of G. Since the entire development of character theory up to this point would have been the same over £ as over C and since / e Irr£(G), it follows that [/, /] = 1. However / is a C-character of G and it follows from Corollary 2.17 that / elrr(G). Thus Irr£(G) £ Irr(G). We also have that both Irr^G) and Irr(G) have the same cardinality, namely that of the set of conjugacy classes of G. Therefore Irr^G) = Irr(G). The point of this digression is to suggest that there is something "absolute" about a character table. It is not entirely an artifact of our choice of the particular field C. Another consequence of this argument which is sometimes useful is that if i e Irr(G), then / is afforded by an £-representation of G. This type of consideration will be discussed much more fully in Chapter 9. Until then, we resume our convention that "character" means "C-character."
Group representations and characters 23 A great deal of information about a group can be recovered from its character table. In particular, all of the normal subgroups of G can be found. A normal subgroup is a union of conjugacy classes and the word "found" in the preceding sentence means that those sets of conjugacy classes whose unions form subgroups can be listed. In particular, the orders of all normal subgroups and the inclusion relations among them can be determined. (2.19) lemma Let 3E be a C-representation of G which affords the character /. Then g e ker 3E iff %{g) = x(l). Proof If g e ker 3E, then X(g) = I = 3E(1) and %(g) = /(l). Conversely, by Lemma 2.15, xdd) = gj + • • • + ef, where e; is a root of unity and/ = /(l). Since \e-t\ = 1, the equation %(g) =f forces e; = 1 for all i. Now 3E(g) is similar to diagta,... ,ef) = I and therefore X(g) = I and the proof is complete. | (2.20) definition Let x be a character of G. Then ker / = {g e G \ %(g) = (2.21) lemma Let x be a character of G with / = £ n,Z; for /, e Irr(G). Then ker / = f]{ker Xi\nt > 0}. Also f){ker Xi\ I < i < k} = I. Proof Since \xiig)\ ^ Z,(l) by Lemma 2.15, xig) = x(l) forces Xi(g) = X,(l) whenever n-t ^ 0. The reverse inclusion is trivial and the first assertion follows. To prove the second statement, consider the regular character p. By Lemma 2.10, ker p = 1 and the result follows. | The normal subgroups N{ = ker /; can be found by inspection from the character table of a group G. We claim that every normal subgroup is the intersection of some of the JV,'s and thus can be found from the character table. To see this, let N o G and let 5R be the regular representation of the group G/N so that ker 5R = N/N. Now view 5R as a representation of G with kernel N and let / be the corresponding character of G. Then N = ker 5R = kwZ=n{tfil[X.Z.]*0}. Given a normal subgroup NofG (where "given" means listing the classes Jf; which it contains), we may calculate | N \ from the character table using |AM = £{|Jf,||jr,£N}. (Recall that |Cl(g)| = |G: Qg)\ and |C(g)| = Xx€irr<G) \x(g)\2 so tnat \$r-,\ is determined from the character table.) It follows from this discussion that G is simple iff ker / = 1 for all non- principal x e Irr(G) and therefore simplicity of a group can be easily determined from its character table.
24 Chapter 2 The group G is solvable iff it has a chain of normal subgroups, 1 = M0 £ M( £ • • • £ M„ = G such that | M,: Mf_ , | is a prime power for all i, 1 < i < n. Since the M, can be located and their orders determined from the character table of G, it follows that the table determines solvability or non- solvability of G. This may be a good place to remark that the character table of G does not determine G up to isomorphism. In fact if p is any prime, there are two non- isomorphic nonabelian groups of order p3. These two groups have identical character tables. Let N o G. It seems natural to ask whether the character tables of N and of G/N can be calculated from that of G. The answer is "no" for N but "yes" for G/N. We have already observed that there is a one-to-one correspondence between representations of G/N and representations of G with kernel containing N. Furthermore, under this correspondence, irreducible representations correspond to irreducible representations. This situation may be interpreted in terms of characters as follows. (2.22) lemma Let N <i G. (a) If x is a character of G and N c ker /, then / is constant on cosets of N in G and the function x on G/N defined by x(Ng) = x(0) is a character of GIN. (b) If x is a character of G/N, then the function / defined by x(g) = t{Ng) is a character of G. (c) In both (a) and (b), / e Irr(G) iff x e Irr(G/N). Usually, we shall identify / and %■ Under this identification, we have \xx(G/N) = {x e Irr(G)| N £ ker /}. To demonstrate what is happening here, let us consider the example G = Z4 and N the normal subgroup of order 4. The classes of G which are contained in N are the identity and Cl(( 1 2)(3 4)). Referring to the character table on p. 17, we see that the irreducible characters /,• of G with N £ ker /,- are / (, Xi ■, and X3 and so under above identification, we haveIrr(G/JV)={x„x2,x3}. In order to write the character table for G/N we need to know its con- jugacy classes. If Jf is a class of G, then Jf, its image in G/N, is a class of G/N. However, distinct classes of G may have equal images in G/N. This poses no problem since if g, hsG, then g and h are conjugate in G = G/N iff x(g) = x(h) for all / e Irr(G/./V). (The second orthogonality relation, applied to G/N proves this.) The part of the character table of G = E4 corresponding to the characters of G/N is
Group representations and characters 25 xr- Xi- Xi- ■*\ l l 2 .w\ l -l 0 ■*\ l l 2 ■W\ 1 -1 0 ■^'s 1 1 -1 Observe that columns 1 and 3 are identical, as are columns 2 and 4. Deleting repeats, we obtain the character table 1 1 1 1-1 1 2 0-1 for G/N. (In this case G/N sl3.| The preceding discussion provides a second method for computing | N \ from the character table of G. Namely, by using the fact that |G:A/| = X{*(l)2|zelrr(G/JV)} = I{/(D2|/elrr(G) and JVsker*}. By Corollary 2.7, a group is abelian iff all of its irreducible characters are linear. It follows that given N o G, the character table of G determines whether or not G/N is abelian. There is no known way to determine from the table whether or not N is abelian. (2.23) corollary Let G be a group with commutator subgroup G'. Then (a) G'= fliker/.|/eIrr(G),/(l) = 1}; (b) | G: G' | = the number of linear characters of G. Proof If / is a linear character of G, then /. is a homomorphism into the abelian multiplicative group of C. It follows that G' £ ker X. Since G/G' is abelian, all / e Irr(G/G') are linear and thus Irr(G/G') = {/eIrr(G)|/(l) = 1}. (This equality, of course, depends on the identification of characters of G/G' with characters of G.) Finally, for any JV <i G, we have N = f){ker %\% e Irr(G) and N £ ker /} and hence (a) follows. The number of linear characters of G is equal to the total number of irreducible characters of the abelian group G/G' and hence equals |G/G'|. The proof is now complete. | The information in Corollary 2.23(b) is useful for finding the set of character degrees of a group G. For instance, if G is a nonabelian group of order 27, then \G: G'\ = 9 and G has exactly 11 conjugacy classes. By the
26 Chapter 2 preceding, G has exactly nine linear characters and two nonlinear irreducible characters / and \p. We have 27 = |G| = 9 + z(l)2 + ^(l)2. Since the only way that 27 — 9 = 18 can be written as a sum of two squares is 32 + 32, it follows that jrfl) = 3 = i/<l). Before leaving the discussion of character tables and factor groups, we mention an amusing result. The following could be proved without characters but it is somewhat tricky to do so. (2.24) corollary Let g e G and N <i G. Then | CG/N(Ng) \ < \ CG(g) \. Proof From the second orthogonality relation, we have \CGIN(Ng)\= £ |z(Ng)|2 = £{lx(0)l2|xeIrr(G),JVckerz} XeIrr(G/N) < E \X(9)\2 = \CG(g)\. | Xelrr(G) We now discuss the connections between characters and the center of a group. (2.25) lemma Let 3E be an irreducible C-representation of G of degree n. Suppose A is an n x n matrix over C which commutes with 3E(g) for all g e G. Then A = al for some a e C. Proof Let M be the n-dimensional row space over C so that M is an irreducible C[G]-module via m-a = m3E(a) for meM and aeC[G]. Let 9: M -» M be defined by m9 = mA. Then 9eEC[G](M)=Cl and 9 = a • 1 for some a e C. The result follows. | (2.26) definition Let x be a character of G. Then Z(/) = {g e G11 x(g) | = Z(l)}- If H £ G and 3E is a representation of G, then its restriction to H, denoted 3EH, is a representation of H. Similarly, the restriction Xh °f a character / of G to H is a character of H and we can write Xh = Z n^ l/(€lrr(H) for suitable integers n^. Note that if Xh e Irr(H), then / e Irr(G). Of course, the converse of this statement is false.
Group representations and characters 27 (2.27) lemma Let x be a character of G and let Z = Z(x) and/ = x(l). Let 3E be a representation of G which affords /. Then (a) Z = {geG\X(g) = el for some eeC}; (b) Z is a subgroup of G; (c) Xz = ft- f°r some linear character k of Z; (d) Z/ker / is cyclic; (e) Z/ker / c Z(GAer z). Furthermore, if x £ Irr(G), then (f) Z/ker z = Z(GAer z). Proof By Lemma 2.15, H(g) is similar to diag^,..., ef), with |e;| = 1, 1 < i < f. Since x(g) = X fy, it follows that |/(g)| =/ iff all £j are equal. Since the only matrix similar to el is el itself, conclusion (a) follows. Define the function k: Z -> C by H{g) = 1(g)/ for geZ. It follows for g,heZ that 3E(g/i) = k(g)k(h)I and hence Z is a subgroup and k is a homomor- phism (linear character) of Z. We have that xio) = P\o) for g e Z and (b) and (c) have been proved. Clearly, ker x = ker k and thus Z/ker x is isomorphic to the image of k, a finite multiplicative subgroup of the field C. This subgroup is necessarily cyclic and (d) follows. Also, ker x = ker 3E and 3E(Z) £ Z(3E(G)) and (e) is an immediate consequence. Finally, if g(ker x) e Z(G/ker xl then X(g) e Z(3E(G)). If x e Irr(G), then by Lemma 2.25, we conclude that X(g) = el for some e e C. Now (f) follows from (a) and the proof is complete. | (2.28) corollary Let G be a group. Then Z(G)=f){Z(z)lzeIrr(G)}. Proof Since (TAG) ker z)/ker x £ Z(G/ker /), it follows from Lemma 2.27(f) that Z(G) £ Z(/). Conversely, suppose g e Z(/) for every / e Irr(G). It follows that g(ker(/)) e Z(G/ker /) and thus for any xeG, the commutator [0,*] = g'^-tgxekerx- Thus [#, x] e f){ker x\xe !"•((?)} = i and g commutes with x. Since xeG was arbitrary, we have geZ^G). | It is apparent from Corollary 2.28 that Z(G) can be located from the character table of G. It follows that it can be determined from the table whether or not G is nilpotent. This is done by finding Z(G), then finding the character table of G/Z(G), and iterating this process. The sequence of subgroups of G which results is the upper central series and G is nilpotent iff this sequence reaches G.
28 Chapter 2 Some information about character degrees can be obtained using Lemma 2.27(c). We need a lemma first. (2.29) lemma Let H c G and let / be a character of G. Then [Zh,Zh]£|G:H|[x,Z] with equality iff %(g) = 0 for all g e G — H. Proof We have \H\[Xb,Xh]= I\x(h)\2< l\x(9)\2 = \G\h,Xl hell (EC since \x(g)\2 Sr 0 for geG — H. Equality thus holds iff x(g) = 0 for all geG — H. The result follows. | (2.30) corollary Let x e Irr(G). Then /(l)2 < | G: Z(/)|. Equality occurs iff x vanishes on G — Z(/). Proof By Lemma 2.27(c), we have XzM = X(W and thus [_XzM, Xuxd = X(lKl A] = z(l)2. Therefore X(l)2<|G:Zto|[z,Z] = |G:Z(Z)| with equality iff / vanishes on G — Z(/). | We already knew, of course, that /(I)2 < | G | for / e Irr(G). We now have the slight improvement that /(l)2 < \G: Z(G)|. Equality can occur here, and when it does, TXy) = Z(G) and / vanishes on G — Z(G). It has been conjectured that only in a solvable group is it possible to have x(l)2 = \G: Z(G)| with x £ Irr(G). As of this writing, the question is still open. Observe that the nonabelian groups of order 27 which were discussed earlier give examples where equality occurs. (2.31) theorem Suppose that / e Irr(G) and that G/Z(/) is abelian. Then |G:Z(z)| = z(l)2. Proof It suffices to prove that / vanishes on G — Z(/). Let geG — Z(/). Then by Lemma 2.27(f), we have that there exists h e G with g~1h~1gh$ ker /. However, since GjTXx)is abelian, we haveg~ lh~lgh = ze 2Ij). Now, if 3E is a representation of G which affords /, then 3E(z) = el and e j= 1 since z £ ker 3E. We have X(gz) = X(g)X(z) = eX(g) and thus /(gz) = e/(g). However, gz = h~ lgh and so x(gz) = x(g\ Since ex(g) = x(d) and e ^ 1, we have x(g) = 0 as desired. | A character / of G is said to be faithful if ker / = 1. Every group has a faithful character, namely its regular character p; but not every group has a faithful irreducible character.
Problems 29 (2.32) theorem (a) If G has a faithful irreducible character, then Z(G) is cyclic. (b) If G is a p-group and Z(G) is cyclic, then G has a faithful irreducible character. Proof (a) Let / e Irr(G) be faithful. By Lemma 2.27(f), Z(G) = Zfa) and by part (d) of that lemma, Z(/) is cyclic. (b) Since G is a p-group, it follows that if 1 ^ N <i G, then N n Z(G) t^ 1. Now let Z be the unique subgroup of order p in the cyclic group Z(G), so that Z £ N for every nontrivial normal subgroup iV of G. Since P){ker xl/eIrr(G)} = 1, it follows that Z £ ker / for some /elrr(G). We conclude that ker / = 1 and the proof is complete. | Problem 2.19 provides an example to show that the full converse of Theorem 2.32(a) is not true. Problems (2.1) (a) Let 3E be an irreducible F-representation of G over an arbitrary field. Show that Y^gea %(g) = 0 unless 3E is the principal representation. (b) Let H £ G and g e G be such that all elements of the coset Hg are conjugate in G. Let / be a C-character of G such that [/H, 1H] = 0. Show that X(g) = 0. Hint (b) Compute the trace of £,,€ H H(hg), where 3E affords /. In the following, all characters are over C. (2.2) (a) Let % be a character of G. Show that / is afforded by a representation 3E such that all entries of 3E(g) for all g e G lie in some field F £ C with |F: 0| < oo. (b) Let e = e2nil", where n= \G\ and let / be a character of G. (Note that X(g)eQ[e] for all ge G by Lemma 2.15.) Let a be an automorphism of the field Q[e] and define /": G -» C by /"(g) = /(g)". Show that /" is a character and that f e Irr(G) iff / e Irr(G). (2.3) Let i be a character of G. Define det /: G -» C as follows. Choose 3E affording / and set (det X)(g) = det X(g). Show that det / is a uniquely defined linear character of G. (2.4) (a) Let G be a nonabelian group of order 8. Show that G has a unique nonlinear irreducible character /. Show that x(l) = 2, x(z) = —2, and X(x) = 0, where z e G - {1} and x e G - G'.
30 Chapter 2 (b) If G S £>8, show that det / ¥= 1G- (c) If G S £>8, show that det z = 1G. Hint Show that ker(det /) contains all elements of order 4. Use Lemma 2.15. Note Although D8 and Q8 have identical character tables, the map det: Irr(G) -» Irr(G) is not the same for both groups. (2.5) (a) Find a real representation of Ds which affords the charaoter / of Problem 2.4(a). (b) Show that this cannot be done for the group Q8. (2.6) Let x, i> be characters of G. Define/xtft/G--♦ 'C'"byi(^)(gr) = .jigMg). (a) If 1^(1) = 1, show that xi> is a character. (b) If tA(l) = 1, show that rfi e Irr(G) iff % e'Irr(G). (c) If ip = x (that is, ip(g) = /[^.and^l) > U,^how'that # £ irr(G). Note In Chapter 4 we will showtthat. x^ is always a character. (2.7) Let G be abelian and write G -= Hrr(G). (a) Show that G is an abelian ..group under the multiplication of Problem 2.6. (b) If H s G, let H1 = {1 e G|H^iker 1}. Show<that 1 is a bijection from the set of subgroups of G onto the-setof subgroupstof((j. (c) Show that G ^G. Hints There is a natural isomorphism of;G onto G. lUsetthis for(b). For (c), use the fundamental theorem of abelian1 groups. (2.8) Let x be a faithful character of G. Show that H £ G is abelian: iff every irreducible constituent of Xh is linear. (2.9) (a) Let / be a character of an abelian group A. Show E IzWi2 > \a\x{\\ xeA (b) Let A £ G with A abelian and \G:A\ = n. Show that *(1) <, n for allzelrr(G). (2.10) Suppose G = (J"= j Ah where the y4; are abelian subgroups of G and A{ n Aj = 1 if i t6 y. (a) Let/eIrr(G). Show that if/(l) > l,then/(l) > \G\/(n - 1). (b) IfGisnonabelian^henl/ljl <, n — 1 for each i and n - 1 ^ (|G|)1/2. Hints For (a), bound £9€G |/(g)|2 using Problem 2.9(a). For (b), use Problem 2.9(b).
Problems 31 (2.11) Let g e G. Show that g is conjugate to g l in G iff %(g) is real for all characters / of G. Note An element of a group which is conjugate to its inverse is called a real element. If G has any real elements other than 1, then G must necessarily have even order. (2.12) Let | G | = n and let g e G. Show that i{g) is rational for every character i of G iff g is conjugate to gm for every integer m with (m, n) = 1. Hints Let e be a primitive nth root of 1 in C and let E = Q[e]. Let <§ be the Galois group of E over Q. Given (m, n) = 1, show that there exists a e 'S with %(gm) = /(g)" for all g e G and all characters /. Conversely, for every er e 'S, there is an m such that this formula holds. (2.13) Let | G' | = p, a prime. Assume that G' £ Z(G). Show that X(1)2 = |G:Z(G)| for every nonlinear / e Irr(G). (2.14) Let H c G' n Z(G) be cyclic of order n and let m be the maximum of the orders of the elements of G/H. Assume that n is a prime power and show that | G | > n2m. Hints Choose /elrr(G) with Hnker / = 1. Let k = det / (as in Problem 2.3). We have /H = #(l)/z with fi e Irr(H). Using k and fi, show that Z(l) ^ n. Finish the proof using Problem 2.9(b). Note The assumption on n can be removed using Corollary 5.4. (2.15) Let x £ Irr(G) be faithful and suppose H £ G and /H e Irr(H). Show that CG(H) = Z(G). (2.16) Let H £ G and let / be a (possibly reducible) character of G which vanishes on G — H. Assume either that H = 1 or that G is abelian. Show that |G:H| divides/(l). Hint Let 1 be an irreducible constituent of %H. Under either hypothesis, find fi e Irr(G) with fiH = k. Compute \j, /*] and conclude | G: H11 {jH, 1]. ./Vote A natural common generalization of the situations H = 1 and G is abelian is H c Z(G). Is the result true under the hypothesis H £ Z(G)? (2.17) Let A < G be abelian and assume there exists / e Irr(G) with x(l) = | G: A |. Show that G has a nontrivial normal abelian subgroup. Hint Show that / vanishes on G — A.
32 Chapter 2 (2.18) Let A o G and suppose A = CG(a) for every a j= I, aeA. Assume further that G/A is abelian. Show that G has exactly (\A\ - 1)/|G: X| nonlinear irreducible characters and that these all have degree equal to \G:A\ and vanish on G - A. Hints Let k = | Irr(G)|. By counting classes, show that k£l+(\A\- l)/\G:A\ + (\G\-\A\)/\A\. Using characters, show that k>\G:A\ + (\G\-\G:A\)/\G:A\2. Use Problem 2.9(b). Note The group G of Problem 2.18 is a special case of a Frobenius group. The character theory of such groups will be discussed more fully later. Observe that although the hypotheses of 2.18 are very special, this situation does arise frequently. Some examples are AA and nonabelian groups of order pq where p and q are primes with p\(q — 1). (2.19) Let£= <x1,x2,x3,x4> be an elementary abelian group of order 16. Let P = <y> be cyclic of order 3. Let P act on E by Xi = *2i x2 = *1*2> -*3 = x4i x4 = *3*4- Let G be the semidirect product E x P. Show that Z(G) = 1 but that G does not have a faithful irreducible character. Hint The smallest possible degree for a faithful character of E is 4. (2.20) Let 3E and 9) be irreducible C-representations of G and define the functions a,J(g) and b^g) by H{g) = (a,jg)) and 9)(gf) = {bi}{g)). Write SPqrs= ZflP«(^«(9-1)- Show that Sp,„ = 0 if 3E and 9) are not similar. If 3E = ?), show that Spqrs = 0 unless p = s and q = r in which case Sp„p = | G |/deg 3E. Hint Let P„ = £,.<; X(g)£,r?)(g~') where Ev is the (deg X) x (deg 9)) matrix with all entries zero except the (q, r) entry which equals one. Note that XPqr = P,r9). Use Schur's lemma. For the last statement use Lemma 2.25 and compute tr(P,r). Note The results of this problem are called the Schur relations. They can be used to give another proof of the orthogonality relations.
Characters and integrality One of the most celebrated applications of character theory to pure group theory is Burnside's theorem which asserts that a group with order divisible by at most two primes is solvable. The proof of this theorem (and much of the rest of character theory) depends on properties of algebraic integers. We begin by establishing some of the most basic of these properties. (3.1) definition An algebraic integer is a complex number which is a root of a polynomial of the form x" + an-.!*"-1 + ■■■ + a0, where af e Z for 0 < i < n — 1. (3.2) lemma The rational algebraic integers are precisely the elements ofZ. Proof If a e Z, then a is a root of the polynomial x — a and thus is an algebraic integer. Conversely, let r/s be an algebraic integer with r,seZ. We may assume that (r, s) = 1. We have (r/sf + a„^(r/sf-1 +-.. + ao = 0. Now multiply by s" and rearrange terms to obtain r" = -s(a„_1r""1 + a„_2sr""2 + ••• + aos""1). We conclude that s\r". However, since (r, s) = 1, this yields s = +1 and r/s e Z as desired. | 33
34 Chapter 3 Frequently, the word "integer" is used to mean an algebraic integer, and the elements of Z are referred to as " rational integers." One of the most important properties of the set of algebraic integers is that it is a ring. In other words, sums and products of integers are integers. This fact seems surprising from the definition, but it is not hard to prove indirectly. This we proceed to do. (3.3) lemma Let X = {<*!,..., at} be a finite set of algebraic integers. Then there exists a ring S satisfying (a) ZzSz C: (b) XsS; (c) there exists a finite subset, Y of S such that every element of S is a Z-linear combination of elements of Y. Proof The integer a,- satisfies an equation of the form «"' = ». where f is a polynomial of degree n{ — 1 with coefficients in Z. Let Y = {ai'otj2 • • • afcfc|0 <, rt < nt - 1} and let S be the set of all Z-linear combinations of elements of Y. Using the equation a."' = /((a,), any power of a.{ may be written as a Z- linear combination of 1, a,-, a,2,..., a"'~'. It follows from this that the product of any two elements of Y lies in S and hence S is a ring. All of the properties claimed for S are now clear. | Condition (c) of the above lemma may be paraphrased by saying that S is finitely generated as a Z-module. We now prove a strong converse to Lemma 3.3. (3.4) theorem Let S be a ring with Z £ S £ C. Suppose that S is finitely generated as a Z-module. Then every element of S is an algebraic integer. Proof Let se S and let Y = {yu ...,y„}zS have the property that every element of S is a Z-linear combination of elements of Y. We then have Wi = Z auyj j for all i, with ayeZ. Let A be the matrix (ay) and let v be the column, colfj!,..., y„). Then Av = sv and thus s is a root of the polynomial f(x) = det(x/ - A). It follows that s is an algebraic integer and the proof is complete. |
Characters and integrality 3S (3.5) corollary Sums and products of algebraic integers are algebraic integers. Proof Let a and /? be algebraic integers. By Lemma 3.3, there exists a ring S with Z z S £ C such that a, /? e S and S is finitely generated as a Z-module. Since a + /? and a/? e S, it follows from Theorem 3.4 that they are algebraic integers. | (3.6) corollary Let / be a character of a group G. Then %(g) is an algebraic integer for all g e G. Proof By Lemma 2.15, we know that %(g) = gj + ■■■ + ef, where the gj are roots of a polynomial of the form x" — 1, and therefore are algebraic integers. The result now follows. | We can now see the reason for the assertion made in Chapter 2 that all of the entries in the character table of the symmetric group £„ lie in Z.lfge £„ and m is relatively prime to o(g), then gm and g have identical cycle structures and therefore these elements are conjugate in £„. It follows from Problem 2.12 that %(g) is rational for all / e Irr(£„). Since %(g) is an algebraic integer, Lemma 3.2 yields that %(g) e Z as claimed. Let G be a group and / e Irr(G). We wish to define a function co depending on /, from the center of the group algebra C[G] into C. Let 3E be any representation which affords /. If z e Z(C[G]), then we may conclude from Lemma 2.25 that X(z) = el for some g e C. Observe that since the only matrix similar to el is el itself, the complex number g does not depend on the choice of the particular representation affording /. We now define co by setting co(z) = g. In other words 3E(z) = co(z)I for all z e Z(C[G]). We shall often write co = cox in order to emphasize the dependence of co on /. Since 3E is an algebra homomorphism, it is easy to see that co is also a homomorphism. In particular, co is C-linear and hence to determine co on Z(C[G]), it suffices to calculate its values on a basis. Such a basis is given by the class sums for»the conjugacy classes of G. Let Jf be a class with sum K e C[G] and let g e Jf. Calculation of traces in the equation 3E(X) = co(K)I yields xHMK) = X(K) = X x(x) = \^\x(g)
36 Chapter 3 and thus Zte)|JT| coJK) = Z(D Note that it follows from this formula that the functions a>x are determined by the character table of G. (3.7) theorem Let /elrr(G) and let K be a class sum in C[G]. Then cox(K) is an algebraic integer. Proof Let Jf u ..., Jf k be the classes of G, with corresponding class sums Ku...,Kk. By Theorem 2.4, we have KtKj = £v aiJvKv where a0-v e Z. Since co = cox is an algebra homomorphism from Z(C[G]) to C, we have V Let S be the set of all Z-linear combinations of the co(X(). It follows that S is closed under multiplication. Since eo(l) = 1, it follows that Z z S z C and Theorem 3.4 applies. All of the elements of S are therefore algebraic integers and the proof is complete. | It should be emphasized that the fact that x(g)|Cl(g)|/x(l) is an algebraic integer does not follow from the fact that %(g) is integral since division of an integer by an integer does not usually result in an integer. We proceed now toward Burnside's solvability theorem. The essence of the argument is contained in the next result. (3.8) theorem (Burnside) Let / e Irr(G) and let Jf be a conjugacy class of G with g e Jf. Suppose that (/(1), | Jf |) = 1. Then either g e Z(/) or else 1(9) = 0. Proof We know that x(gf)|Jf |/x(l) is an algebraic integer. Since (/(l), |jf |) = 1, we may choose rational integersu and v so that m/(1) + v\Jf\= 1. Thus x(9)(i - «x(i)) = VM\*\ z(D zd) is an algebraic integer. Since u%(g) is also integral, it follows that a = x(g)/x(l) is an algebraic integer. Suppose that g£Z(x), so that \iig)\ < z(l) and |a|<l. Now let n = o[g) and let E be the splitting field for the polynomial x" — 1 over Q in C so that a e E. Let ^ be the Galois group of £ over Q. Since x(g) is a sum of x( 1) roots of unity, so is x(g)a for each a e <§. It follows that | /(gf)" | ^ /(1)
Characters and integrality 37 and I a" I <. 1 for a e 0. We have then n< < i. For each a e CS, a." satisfies the same rational polynomials that a satisfies and hence is integral. Therefore /? = ]~[ a" is an algebraic integer. However, /? is clearly fixed by all a e <& and therefore /? e Q by elementary Galois theory. It follows from Lemma 3.2 that /? e Z. Since | /? | < 1, we have /? = 0 and hence a." = 0 for some er. Therefore 0 = a = i(g)li(V) and /(g) = 0. The proof is complete. | (3.9) theorem Let G be a nonabelian simple group. Then {1} is the only conjugacy class of G which has prime power size. Proof Suppose geG, \Cl(g)\ = p", and g * 1. Let /elrr(G), x * lc- Then ker / = 1 since G is simple and Z{/) = Z(G) = 1 since G is nonabelian. Thus if pJ(x(l), then x(g) = 0 by Theorem 3.8. Now 0 = P(g) = I x(Mg) = i + Z x(Dx(0)- X€lrr(G) X€lrr(G);pU(l) We have — 1 = pa, where the sum being taken over / e Irr(G) where p | /(1). It follows that a = — 1/p is an algebraic integer and this violates Lemma 3.2. | (3.10) theorem Let \G\ = p"qb, where p and q are primes. Then G is solvable. Proof Use induction on \G\. We may assume |G| > 1 and choose a maximal proper normal subgroup N. If N > 1, then by the inductive hypothesis, N and G/N are solvable and thus G is solvable and the result follows. Suppose then N = 1, so that G is simple. Let P?^ 1 be a Sylow subgroup of G. We may choose geZ(P), g # 1. Then |C%)| = |G:C(g)| divides |G: P\, which is a prime power. It now follows from Theorem 3.9 that the simple group G is abelian and the proof is complete. | We now obtain some strong results about the degrees of the irreducible characters of a group G. The fact is that /(1) 11G: Z(/) | for / e Irr(G). We shall first prove the weaker statement that the irreducible character degrees divide the group order. This proof is much less complicated and serves to motivate the stronger proof.
38 Chapter 3 (3.11) theorem Let x e Irr(G). Then /(1) 11G |. Proof From the first orthogonality relation we have \G\= ZxigMg'1). geG We wish to rewrite this equation in terms of cox. Let Jf 1; Jf 2, • • •» ^k be the classes of G, with class sums Kt and representative elements gr We have then k k \g\= Y\^i\x(g-M9i'l)= IxVMKMgr1), ■=i i=i where co = cox. This yields |G|/z(l) = 2>(K,M?f1), which is an algebraic integer. Since |G|//(1) is rational, it lies in Z and the result follows. | (3.12) theorem Let x e Irr(G). Then /(1) 11G: Z(/) |. Proof Since / may be viewed as a character of G/ker /, it is no loss to assume that ker / = 1. Under this assumption, Z(G) = Z{y). For x, yeG, define x = y if there exists zeZ = Z(G) such that x is conjugate to yz. It is easy to check that = is an equivalence relation and thus partitions G into equivalence classes. We claim that | x(x) | is constant as x runs over one of these classes. To see this, observe that Xz = xWh where X is a faithful linear character of Z, and that x(yz) = Kz)x(y) for z e Z and yeG. If x = y, then xg = yz for some z and x(x) = X(z)x{y). Since | l(z) | = 1, the claim follows. Let #!, #2* • • •, ^r be those ( = )-classes on which / does not vanish. We have then |G|= Ilx(9)l2= i^iWxig-M2, geG i=l where the gt are representatives for the #,. We claim 1^1 = |Cl(gf,-)||Z|. Clearly, every x e #,- is of the form yz where y e Cl(gt) and z e Z. It suffices to show that all of these elements yz are distinct. Suppose that _y1z1 = y2z2, yu y2 e Cl(g{), and zu z2 e Z. Then X(.yiM(zi) = Zt^MO^) and /(yj = x(y2) = x(gf,-) j= 0. Thus l(zt) = l(z2) and hence zt = z2 since 1 is faithful on Z. Thus _yt = y2 and the claim is established. We have now \G\ = Zl^llz^)!2 = Y\Cl(gi)\x(9iMgi-l)\Z\ = YJx(iMKi)x(gr1)\z\,
Characters and integrality 39 where Kt = ^eci^)* an(* a = cox. It follows that |G:Z|/jrfl)= t^K-Mdr1), ■ =i an algebraic integer which is a rational number. The result now follows. | As a combined application of Theorems 3.8 and 3.12, we prove the following. (3.13) theorem Let G have a faithful irreducible character of degree p", where p is a prime and suppose that a Sylow p-subgroup of G is abelian. Then p" is the exact power of p dividing | G: Z(G) |. Proof Let / be the given faithful character of G. By Theorem 3.12, pa = Z(l) divides |G:Z(G)|. Let P e Sylp(G) and let xeP. Thus P £ C(x) and hence (/(l), |Cl(x)|) = 1. By Theorem 3.8, /(*) = 0 if x i Z(/) = Z(G). Let Z = P n Z(G) so that / vanishes on P — Z. Now by Problem 2.16, we conclude that |P:Z||p". Since P/Z s PZ(G)/Z(G), which is a Sylow subgroup of G/Z(G), the result follows. I The above result is typical of a number of theorems about "complex linear groups," that is, groups of nonsingular matrices over C. In these theorems, one is given the degree n of a finite linear group G and the object is to control the structure of G in terms of n, often under certain additional assumptions such as the irreducibility of G. In Theorem 3.13 we are given a group having a faithful representation of known degree, and this is obviously equivalent to being given a linear group of that degree. If G is a complex linear group, let S be the subgroup consisting of the elements of G which are scalar matrices (that is, of the form el). Observe that if G is irreducible, then S = Z(G). The group GlS is called the collineation group associated with G. Frequently, the object of a theorem about linear groups of given degree is to obtain information about the associated collineation group. Theorem 3.13 is of this nature. The reason for this situation may be seen from the following. Let G be a linear group of degree n and let C be a group of n x n scalar matrices. Let G* = GC and let S* be the scalar subgroup of G*. The linear group G* may be much larger than G but G/S s G*/S*. We have already seen several situations in which a character value is forced to be zero. We shall now prove that every nonlinear irreducible character vanishes somewhere. We begin with a preliminary result.
40 Chapter 3 (3.14) lemma Let G be a cyclic group and let / be a (possibly reducible) character of G. Let S = {g e G | G = <#>} and assume that /(s) ^ 0 for all seS. Then Ilx(s)|2>|S|. seS Proof Let n = |G| and let £ be the splitting field for the polynomial x" — 1 over Q in C. Let ^ be the Galois group of £ over Q. If a e <§ and e is an nth root of 1, then e" = em for some m e Z, (m, n) = 1. Now /(s) = et + • • • + ef where e," = 1 and hence /(s)" = e^ + • • • + efm = /(s™). The group ^ is abelian and therestriction of complex conjugation to £ is an element of <&. It follows that a" = (a)" for all a e £ and a e 0, and thus la"!2 = aV = a'a" = (| a I2)". Therefore (|x(s)|2r = |x(sm)|2, where (m, n) = 1 and m depends only on a. Observe that if s e S and (m, n) = 1, then sm e S. Also, the map x h-» xm is one-to-one on G and therefore effects a permutation of S. It follows that Flses I X(s) I2 's invariant under ^ and hence is rational. Since it is an algebraic integer, it must lie in Z, and since / does not vanish on S, we have ni*(s)i2*i. seS for any positive real numbers ru r2,..., rfc, we lln>(Un)llk ^Ilx(s)|2>l I J I seS and the proof is complete. | (3.15) theorem (Burnside) Let /elrr(G) with *(1) > 1. Then i(g) = 0 for some g e G. Proof Partition G into equivalence classes by calling two elements of G equivalent if they generate the same cyclic subgroup of G. Assume %(g) ^ 0 for all g e G. Then by Lemma 3.14, we have 1\X(S)\2^\S\ seS for every equivalence class S. Sum this inequality over all equivalence classes of nonidentity elements to obtain IIX(0)|2>|G|-1 9*1 Now we use the fact that have and we conclude that
Characters and integrality 41 and thus |G|= Ilx(0)l2>|G|-l+x(l)2. geG This forces /(l) <, 1 which contradicts the hypothesis. | The next topic we shall discuss does not, strictly speaking, depend on algebraic integers. Nevertheless it seems appropriate to include it here. Let G and H be finite groups and suppose C[G] = C[H], where this is a C-algebra isomorphism. What can we infer about the relationship between G and HI Clearly, \G\ = |H| and there exists a degree-preserving one-to-one correspondence between Irr(G) and Irr(H). We cannot conclude, however, that G and H are isomorphic or even that they have identical character tables. Indeed, if G is abelian, it follows from the results of Chapter 1 that C[G] is the direct sum of | G | copies of C. Thus if G and H are abelian and \G\ = \H\, then C[G] s C[H]. The situation becomes more interesting if we make the weaker assumption that Z[G] = Z[H], where Z[G] represents the group ring of G over Z and may be identified with the ring of Z-linear combinations of elements of G in C[G]. It is conjectured that if Z[G] £ Z[H] for finite G and H, then G = H. The best result in this direction that has been proved as of this writing is due to A. Whitcomb. It asserts that if G is metabelian and Z[G] sZ[H],thenG sH. If Z[G] = Z[H], then it is clear that we may view H as a multiplicative subgroup of Z[G] £ C[G] and that H spans C[G] over C. In particular, IH| > |G| and hence by symmetry \H\ = \G\ and H is a basis for C[G]. (3.16) definition Let H £ C[G] be such that H is a multiplicative group which is a basis for C[G]. Suppose that every element of H is a Z-linear combination of elements of G. Then H is an integral group basis in C[G]. If H is an integral group basis in C[G], we identify C[G] with C[H] in the natural manner. An important result used in studying the isomorphism problem is due to G. Glauberman. Most of the rest of this chapter is devoted to its proof and some consequences. We use the notation Z Jf to denote the sum of the elements of the conjugacy class Jf, computed in the group algebra. (3.17) theorem (Glauberman) Let H be an integral group basis in C[G]. Then there exists a one-to-one correspondence between the sets of conjugacy classes of H and G such that if if corresponds to Jf, then | if | = |Jf| and Ii?= ±ZjT.
42 Chapter 3 We need a lemma. (3.18) lemma Let a, /? e Z(C[G]) and view the characters of G as being defined on all of C[G]. Define /elrr(G) Then (a) <ctat + c2a2, /?> = c^a,, 0> + c2<a2, 0>; (b) </?,a> = <^>; (c) <K„ Ky> = 0 if i*j; (d) <K„K,> = |G||jr,|; where the jf, are the classes of G and Xf = EjT,. Proof Statements (a) and (b) are immediate since characters are linear functions. To prove (c) and (d), let x e jf t and y e Jf j. Then x(Kj) = | Jf f | x(x) and x(Xj-) = |Jf/|x(}0- The result is now immediate from the second orthogonality relation. | Proof of Theorem 3.17 Identify C[H] with C[G] so that k = dim(Z(C[G])) is the common number of classes of H and G. Let ift, Jz?2, ..., ifk be the classes of H and jf t, Jf2,..., Jf t the classes of G. Write X; = ZjT, and L,- = Ei?,. Since the X,- are a basis for Z(C[G]), we may write U = Z c0K;. j Since L( is a Z-linear combination of elements of G, we conclude that all Ctj-eZ. The form < , > denned in Lemma 3.18 depends only on the algebra C[G] = C[H] and not on the particular groups G and H. This is because the irreducible characters, defined on C[G], are simply the traces of the irreducible representations of the algebra. We therefore have (1) \H\\X,\ = <LltLi> = S(c|.J.)2|G||JfJ.|. j Sum (1) over i to obtain (2) |H|2 = |G|Xfe(cy)2W.|. Since the Lt span Z(C[G]) but no proper subset of {Kj} spans this space, it follows that for each j, there exists i with cis =£ 0. Thus £,- (cy)2 > 1. Since | H | = | G |, Equation (2) yields |G| = Z (% (ctA\JTj\ > X |JT;| = \G\
Characters and integrality 43 and thus we have equality and £, (cu)2 = 1 for all j. Therefore, for each j there is a unique i with ctJ ^ 0. This defines a map{jfj} -» {if,}. Since for each i there exists 7 with cy ,£ 0, this map is onto and hence is one-to-one. Because all nonzero c,7 = ± 1, we have Lt = ±Kj when ££, and jf, correspond. Finally, Equation (1) yields |H|| if f| = I G||jf j I and thus I iff I = \Jfi\ and the proof is complete. | (3.19) corollary Let H be an integral group basis in C[G]. Then there exists an integral group basis H* in C[G] such that H = H* and the class sums of H* equal the class sums of G. Proof Let 3 be the linear extension of the principal character of G to C[G]. (This is usually called the augmentation map.) Note that 8(£ agg) = £ ag and 5 is an algebra homomorphism C[G] -» C. If/i e H, then c5(/i)e Zand since 5(h)5(h~l) = (5(1) = l,wehavec5(/i) = ±1. Let /1* = 5(h)h e C[G] for /ieH and note that c5(/i*) = c5(/i)2 = 1. Put H* = {h*\heH} and note that H* is a group and the map hi->/i* is an isomorphism. Clearly, H* is an integral group basis in C[G]. Let if be a class of H* so that Y,S£ = ±ZJf for some class Jf of G by Theorem 3.17. Since 8(h*) = 1 for h* e H*, we have (5(Zif) = |if |. Since (5(ZJf) = I Jf I, the ambiguous sign above must be positive and the proof is complete. | (3.20) corollary Let H be an integral group basis in C[G]. Then G and H have identical character tables. Proof By Corollary 3.19, we may assume that there exists a one-to-one correspondence between the classes of H and the classes of G such that if ^£ and Jf correspond, then Y,<£ = £Jf and | if | = | Jf |. Let / be the trace of an irreducible representation of C[H] = C[G]. Let if and Jf correspond and let /1 e if and geJf. It suffices to show that /(/i) = /(g). However tih)\2\ = x(Zif) = z(ZJf) = *(0)|jr| and since | if | = | Jf |, the result follows. | We have seen that if Z[G] s Z[H], then C[G] has an integral group basis isomorphic to H. The converse is true but is less obvious. (3.21) theorem Let H be an integral group basis in C[G]. Then Z[G] = Z[H] in C[G]. Proof Here, Z [H] denotes the ring of Z-linear combinations of elements of H. (It is isomorphic to the abstract integral group ring.) Since H £ Z[G], we have Z[H] £ Z[G]. We show that G £ Z[if]. WriteG = {g,|l < i <. n}andH = {h(\l < i < n},wheren = \G\ = \H\.
44 Chapter 3 We have h; = £, a^gj for ai} e Z. We shall show that the matrix A = (au) has an inverse with entries in Z and this will complete the proof. Write hf1 = Z.' ^ij9i~' and B = (^"'A an integer matrix. Let p be the character of the regular representation of C[G] = C[H] so that p(hihf~l) = ndij. Now express /i,/ij~ ' as a linear combination of elements of G and observe that the coefficient of 1 is £v ahbvj. It follows that ndfj = P(hihf') = n £ a,vbv> v and thus AB = /, the identity matrix. The result now follows. | Problems (3.1) Let a be an algebraic integer and suppose that /(a) = 0, where /(x)e 0[x] is irreducible and monic. Show that/(x) e Z[x]. (3.2) Let G be a group, g e G, and let / be a character of G. Suppose \%(g) | = 1. Show that %(g) is a root of unity. Hint Let £ £ C be the splitting field for x" - 1 over Q. For integral a e E, with | a | = 1, let/a e Z[x] be the polynomial of Problem 3.1. Show that only finitely many polynomials can arise this way. Do this by bounding the degree and the coefficients offa. (3.3) Show that no simple group can have an irreducible character of degree 2. Hint Problem 2.3 is relevant. (3.4) Let G be a simple group and suppose / e Irr(G) with /(l) = p, a prime. Show that a Sylow p-subgroup of G has order p. Hint If the Sylow p-subgroup P is nonabelian, then Z(P) s Z(/). (3.5) Suppose A £ G is abelian and \G:A\ is a prime power. Show that G' < G. (3.6) Let G be a p-group and suppose / e Irr(G). Show that x(l)2||G: Z(/)|. (3.7) Let /elrr(G) be faithful and suppose /(l) = p" for some prime p. Let P e Sylp(G) and suppose that CC(P) £ P. Show that G' < G. Hint Let Q £ CC(P) be a p'-subgroup, Q ± 1. Show that Q n Z(x) # 1 and consider det /. (3.8) Let x be a (possibly reducible) character of G which is constant on G — {1}. Show that / = a\G + bpG, where a,beZ and p is the regular character of G. Show that if G ¥= ker /, then x(l) > | G | - 1.
Problems 45 Hint First show that / = a\a + bpG for some a, b e C. (3.9) Let Jfu ..., Jfk be the conjugacy classes of a group G and let K,,..., Kk be the corresponding class sums. Choose representatives gt e Jf, and let a,jv be the integers defined by V Show that ip \G\ X£{G) jrfl) • Hint Use the eo^ and the second orthogonality relation. Notes This formula shows that the a1JV can be calculated from the character table. Therefore, the character table of G can be used to answer questions such as: Is an element g e jf „ the product of an element of Jf, with one of Jf fl The fact that the a1JV are nonnegative rational integers imposes another necessary condition on an array of complex numbers that the array be a character table for some group. Since the K{ are a basis for Z(C[G]), it follows that the character table of G determines Z(C[G]) up to algebra isomorphism. (3.10) We write [x, y] for the commutator x~ ly~lxy of x and y in a group G. (a) Let g e G and fix x e G. Show that g is conjugate to [x, y] for some ^ e G iff y lx(*)|2x(g)/Q X€lrr(G) Z(l) (b) Show that g = [x, _y] for some x, y e G iff X€lrr(G) Z(l) Note We already knew that the character table of a group determines the commutator subgroup. Problem 3.10 says more than this, since the commutator subgroup usually does not consist entirely of commutators. (3.11) Let g e G be a commutator. Suppose m e Z, (m, o{g)) = 1. Show that gm is a commutator. Hint See the hint to Problem 2.12. (3.12) Let x e Irr(G) and g,heG. Show Zfo)Z(*) = 7^7 IzW- |G| 2€C
46 Chapter 3 Hint Let 3E afford / and let X, and K, be the class sums in C[G] which contain g and h. Use the fact that 3E(X,)3E(XJ) = 3E(X1XJ). (3.13) {Brauer) Let Ku ..., Kk be the class sums in C[G]. Suppose there exists c e C such that £ X, = cflK,. Show G = G'. Hint Let lc#xeIrr(G). Show that oix(Kj) = 0 for some i. (3.14) (Brauer) In the notation of the previous problem, show that if G = G', then there exists c e Q such that £ Kt = cf]K.- Hint To show that two elements a, be Z(C[G]) are equal, it suffices to prove that a>x(a) = ax(b) for all / e Irr(G). (3.15) (Thompson) Let £ be a Galois extension of Q with Galois group ^. Let a e £ be an algebraic integer with the property that a" is real and positive forallere^. AtheoremofSiegel(^nn.o/Mat/i.46 (1945) p. 303, Theorem III) asserts that if a =£ 1, then Use this to show that if / e Irr(G) then %(x) is either zero or a root of unity for more than a third of the elements xeG. Hints Mimic the proof of Theorem 3.15. Use Problem 3.2. (3.16) (Burnside) Let |G| be odd and suppose/elrr(G) is not principal. Show that x ¥= X- Hint Using orthogonality, show that if/ =£ lcand^ = x,then/(l) = 2a for some algebraic integer a. (3.17) (Burnside) Let |G| be odd and suppose that G has exactly k con- jugacy classes. Show that \G\ = kmodl6. Hint If n is an odd integer, then n2 = 1 mod 8.
4 Products of characters Let x and \\i be characters of G. The fact that / + \p is a character is a triviality. We may define a new class function xi> °n G by setting (x^){g) = x(g#(g). It is true but somewhat less trivial that /i/f is a character. [If either / or \ji is linear, this is Problem 2.6(a).] Let V and W be C[G]-modules. We shall construct a new C[G]-module V ® W called the tensor product of V and W. Choose bases {vu ...,vn} for 7 and {wu ..., wm} for W. Let V ® W be the C-space spanned by the mn symbols vt (g) vv,-. [More precisely, V ® W is the set of formal sums of the form £ a^0>j ® wj), with ay e C] U veV and w e W, suppose t> = £ a-^ v ® w = £ a.-b/iv ® Wj) eV ®W. and w = >, ftj-Wj-. We define Note that not every element of V ® W has the form v ® w for v e V and weW (except in the special case that n or m = 1). We define an action of G on V ® W by setting (v, ® Wj)g = vtg ® Wjg and extending this by linearity to all of V ® W. The reader should check that live V, w e W, and g e G, then (v ® w)g = vg ® wg. It follows that (xg^ = ^192) for x e V ® W and gt e G. Next we give V ® W the structure of a C[G] -module by extending the action of G by linearity in C[G]. In other words, for x e V ® W we define It is routine to check that this really makes V ® W into a C[G]-module. 47
48 Chapter 4 A few words of caution are in order here. If a e C[G], it is not necessarily true that (v-t ® wjp. = t^a ® Wjtx. It is for this reason that we first defined the action of G on V ® W and then extended it to C[G]. If A. is an arbitrary algebra with modules V and W, it is not generally possible to define the structure of an /1-module on V ® W. We have not yet shown that the C[G]-module V ® W is determined (up to isomorphism) by V and W, independently of the choice of bases. This is, in fact, not hard to prove for group algebras over any field. We shall leave the general situation to the problems. (4.1) theorem Let C[G]-modules V and W afford characters / and \ji, respectively. Choose bases in V and W and construct V ® W. Then V ® W affords the character x*P and is independent of the choice of bases. Proof Let {i>,| 1 < i < n} and {wr\ 1 < r <, m} be bases for V and W, respectively, and let g e G. Write n m vi9 = Hauvj and wrg = Y,brsWs, j=l -S=l with Ay, brse C. Then %(g) = YH= i «i,and \ji(g) = JJ*=, b„. Let 9 be the character afforded by V ® W. Since (vt ® wr)g = X flyMfj ® vvs), we have %) = Z flii ftrr = Z aii Z ftrr = /(g)^). i, r i r We now know that the character afforded by V ® W is independent of the choice of bases and the result follows by Corollary 2.9. | (4.2) corollary Products of characters are characters. Corollary 4.2 provides another necessary condition for an array of complex numbers to be a character table. It asserts that the inner product of any row with the product of any two rows is a nonnegative integer. In this connection, the following formula is relevant: ExiiA, /2] = j-gj Z Zi(s#(0)/2(i) = Lzi, fe]- In general, products of irreducible characters are not irreducible. For instance, if / e Irr(G), then te Irr(G); nevertheless, lc is a constituent of XX since [XL lc] = [LK\g\ = !•
Products of characters 49 (4.3) theorem (Burnside-Brauer) Let / be a faithful character of G and suppose %(g) takes on exactly m different values for g e G. Then every \p e Irr(G) is a constituent of one of the characters (xf for 0 < j < m. Note In the special case x = p, the regular character of G, we have m = 2. The theorem is already known to be true in this case. Proof Let at, a2,..., am be the distinct values taken on by /. Let Gi = {geG\y>(g) = *i}. Assume a.^ = /(l) so that Gt = ker / = {1}. Fix \\i elrr(G) and let ft = ZseGi #?)• Now for./ ^ 0, we have |G| 1 = 1 If t/^ is not a constituent of xJ for any y", 0 < j < m, we have E(«^« = o, o<y<m. i The determinant of this system of m equations in the m "unknowns" /Sj is the so-called "Vandermonde determinant" and is equal to ± J^f^</a,- — a,-) ^ 0. It follows that all ft = 0. On the other hand, jSt = i^(l) ^ 0 and this contradiction proves the theorem. | Let g e G and let n > 0 be an integer. We ask how many nth roots g has in G. Let $n(g) = \{heG\h'' = g}\. Observe that if (| G |, n) = 1, we may choose an integer m such that nm = 1 mod | G |. Thus if h" = k", then h = h"m = A:nm = fc and we have 9„(g) < 1 for all g eG. Since the map hh-»/i" is one-to-one on G, it must be onto and it follows that 8n(g) = 1 for all g e G. The situation becomes considerably more interesting if we drop the assumption that n is prime to | G |. Since 9„ is clearly a class function on G, we may write #n = Z VnMX< X€lrr(G) where v„(/) is a uniquely determined complex number. (4.4) LEMMA Vn(x) = ( 1/| G\)Y,geGX(g").
50 Chapter 4 Proof By the orthogonality relations we have \V\gzG Since 9nfef)Hg) = L,tCr)l»=9 H^j, we have l«l (l€C Finally, replace hby h'1 to obtain the desired result. | Let cp be any class function of G and let n be a positive integer. We define a new function <p<n) by <p(n)(gf) = <p{g"). Note that <p<n) is a class function. Lemma 4.4 asserts that v„(/) = [x(n), lc] for / e Irr(G). It is not, in general, true that x(n) is a character for % e Irr(G), however, it is always a difference of two characters (see problems) and thus v„(/) e Z. (4.5) theorem (Frobenius-Schur) Let / e Irr(G). Then (a) x<2) is a difference of characters; (b) v2(X)= 1,-l,or0; (c) v2(x) t6 0 iff / is real valued. Proof Let V be a C[G]-module which affords / and let vu v2,---,v„ be a basis for V. Let W = V (8> V and define the linear map *: W -> W by (vt ® Vj)* = Vj® v{. Let Ws= {weW\w* = w} and WA = {weW\w* = -w}. These subspaces are called the symmetric and antisymmetric parts of W, respectively. If w e W, we have w + w*eH's and:w — w* eWA. Since w + w* w — w* it follows that W = Ws + WA. . We claim that if w e W and g e G, then (wg)* = w*g. It suffices to check this as w runs over the basis v( ® v}, and thus we need to show that (v,g ® Vjg)* = Vjg ® vtg. In fact, it is true that (x ® y)* = y ® x for all x, yeV. This is seen by expanding x and y in terms of the vt.
Products of characters 51 It now follows that Ws and WA are C[G]-submodules of W and we have X = Xs + Xa> where xs and Xa are tne characters afforded by Ws and H^. We compute Xa using the basis Wjj = t)f ® Vj - Vj <g> l>„ i < j, for WA. Suppose v-,g = Y. Qirvr- We have ">ij9 = Z (airaJs ~ ajrais)Vr ®VS= Y,(airajs - fljrfl«SKs. r,s r<s Therefore n(g)= Hauajj- aaaiy This yields ZXaIq) = Ha>>ajj ~ \Zauan = ( Z «»)( Z ajj) ~ Z aiJa)i- i*j i*j \i J\ j J i,J Let 3E be the representation corresponding to the basis {vt} of V, so that X(g) = (Ay) = M. We have 2Zi4(0) = (tr M)2 - tr(M2) = X(g)2 - X(g2) since M2 = 3E(gr2). This yields z(2)(gf) = x(g)2 - 2z/1(gr) and thus z<2) = z2 - 2^, a difference of characters, as claimed. Now v2(z) = [z<2), lc] = [z2, lc] - 21x4, lc]. If z is not real valued, then 0 = [z, z] = Lz2, lc] and thus [z^, lc] = 0 since Xa is a constituent of Z2. It follows that v2(z) = 0 in this case. If z is real valued, then 1 = [z2, lc] tnus Ix*, lc] = 0 or 1 and v2(z) = 1 or -1. The proof is now complete. | (4.6) corollary Let G have exactly t involutions. Then i + f= Z v2(z)z(i), X€lrr(C) where v2(z) = 0 if z 7^ Z and v2(z) = ± 1 if z = Z- Proof This is immediate since 92(1) = 1 + t. | If N <i G, we have identified Irr(G/iV) with a subset of Irr(G). The following lemma shows that v„(z) is well defined under this identification. (4.7) lemma Let N <i G and let z e Irr(G) with N £ ker z- Let v„(x) be as above and let v„(z) be the corresponding number computed in G/N. Then v„(z) = i>„(z)-
52 Chapter 4 Proof We have = i7q Z xte") = vn(xl I (4.8) lemma Let x £ Irr(G) and let X be a linear character of G with X" = lc. Then vn(x) = v„(l/). Proof Recall that Xx e Irr(G) by Problem 2.6(b). Now v„(ix) = r^ Z to)(0") = r^ Z %"k(0n) = t4 Z xig") = v„(z) |G|9€C |G| g \G\ since 1(9") = 1(g)" = X"{g) = \. | As an application of the Frobenius-Schur theorem, we prove the following. (4.9) theorem (Alperin-Feit-Thompson) Let G be a 2-group containing exactly t involutions. If t = 1 mod 4, then either G is cyclic or | G: G' | = 4. Proof If G is abelian, then it clearly must be cyclic. We suppose G is not abelian and show that | G: G' | = 4. If Z(G) is not cyclic, choose K s Z(G) elementary abelian of order 4. The set {x e G\x2 = 1} is a union of cosets of X and hence 4|(r + 1), a contradiction. Therefore, Z(G) is cyclic and G contains the unique minimal subgroup Z of order 2. Since G' > 1, we have Z s G'. Also G/Z is not cyclic since G is not abelian. If G/Z satisfies the hypotheses of the theorem, then |G:G'| = |(G/Z):(G/Z)' | = 4 by induction and we are done. We may therefore assume that the number of involutions in G/Z is not = 1 mod 4. We have Z v2(z)x(l) = t + 1 =2 mod 4 xelrr(G) and Z v 2(x)xW # 2 mod 4, Z€lrr(C);ZSkerz where the second statement follows via Lemma 4.7. We conclude that (*) Z v2(Z)Z(l)#0mod4. Z€lrr(C);Z«kerz Now let C be the group of linear characters X of G which satisfy X2 = lc. For x e Irr(G), we have Xx e Irr(G) and since Z £ ker X, we conclude that
Products of characters S3 Z £ ker x iff Z £ ker(l/). Therefore C permutes {/ e Irr(G)|Z £ ker /} and partitions this set into orbits 6t. By Lemma 4.8, v2(x) is constant on each orbit asisx(l). Since 16X | is a power of 2, we conclude from Equation (*) that there exists X e Irr(G) such that (i) Z £ ker /, (ii) v2(Z) * 0, (iii) *(1)|0|£2, where 0 is the orbit containing /. Since Z £ ker /, we have ker z = 1. Since G is not abelian, z(l) > 1 and hence z(l) = 2 and 101 = 1. Therefore, Ax = X f°r a^ ^ e C. Now let F = <I>(G), the Frattini subgroup. If g e G - F, then there exists k e C for which 1(g) =£ 1 and it follows that xid) = 0. By Lemma 2.29, we now have 4 = Z(D2>[ZF,ZF] = |G:F|. Since G is not cyclic, \G:F\ > 4 and we have equality above. This forces XF = 2//, where fi is a faithful linear character off. Since v2(z) ¥= 0, we know that / is real valued. Therefore, /z is also, and so \F\ < 2. Thus \G\ <, 8 and hence G s D8 or Q8. In either case, \G: G'\ = 4 as desired. | A theorem of O. Taussky [Theorem III 11.9(a) in Huppert] asserts that the only nonabelian 2-groups G for which | G: G' | = 4 are the dihedral, semi- dihedral, and generalized quaternion groups. In each of these groups, the number of involutions is in fact = 1 mod 4. As another application of Corollary 4.6, we prove some results of Brauer and Fowler that were originally obtained in a different way. (4.10) lemma Let a,, a2,..., a„ be arbitrary real numbers. Then Proof The well-known Schwarz inequality asserts that (2>&i)2 s: (2><2)(2>2) for real a, and bt. The lemma follows by setting all bt = 1. |
S4 Chapter 4 (4.11) theorem Assume \G\ = g is even and that G contains exactly t involutions. Let a = (g — l)/r. Then (a) There exists xeG,xJ= 1, with | G: C(x) \ < a2. (b) There exists real / e Irr(G), ](# le, with x(l) < a. Proof By Corollary 4.6, 0 < (g - l)/a = t < Z Z(l), where ^ = {zeIrr(G)|z = z and X * U)- In particular, s = \Sf\ =£ 0. Now t2<(lx(l)Y<sI>(l)2<S(0-l) by Lemma 4.10. It follows that g — 1 < sol2 and thus sa2>Xx(l)2- Therefore, /(l) < a for some / e y, proving (b). Since s < k — 1, where k = |Irr(G)| is the total number of conjugacy classes of G, we have (k - l)a2 > g - 1 and hence some nonidentity class of G has size <a2. This proves (a). | Recall that an element x e G is said to be real if x is conjugate in G to x~'. It is a fact (which will be proved later) that the number of classes of real elements of G is equal to the number of real irreducible characters. Assuming this, it is immediate from the inequality sa2 > g — 1 in the above proof that statement (a) can be strengthened to guarantee that a real xeG exists with x ± 1 and |G: C(x)\ <, a2. (4.12) corollary Let n be a positive integer. There exist at most finitely many simple groups containing an involution with centralizer of order n. Proof Let G be such a group with \G\ = g. Then G contains at least g/n involutions and hence a < n in the notation of Theorem 4.11. By (a) of that theorem, there exists xeG with 1 < | G: C(x) \ < n2. Therefore, G is isomorphic to a subgroup of the alternating group A„2_ 1. The result follows. |
Products of characters SS (4.13) corollary Let G have even order, g > 2. Then G contains a proper subgroup of order > (g)113. Proof We use induction on \G\. First assume Z(G) = Z > 1. If |G: Z\ is even, then there exists H/Z < G/Z with \G:H\ = \(G/Z):(H/Z)\<\G/Z\2'3<g2'3 and the result follows. If | G: Z | is odd, then G has a central Sylow 2-subgroup and thus has a normal 2-complement by a transfer theorem of Burnside. (Or see Theorem 5.6.) It follows that G has a subgroup of index 2 and the result follows. Now assume Z(G) = 1 and let a be as in Theorem 4.11. Then 1 < |G:C(x)| < a2 for some xeG and we are done if a2 < g213. Assume, then, that a > g113 and let t be an involution in G. Then l<|G:C(T)|<(0-l)/a<g/a<g2'3 and the result follows. | We now wish to discuss the question of which real / e Irr(G) satisfy v2(z) = + 1- The answer is interesting but does not seem to be very important in the applications of the Frobenius-Schur theorem. The solution to this problem involves some matrix theory. Let U and V be C[G]-modules with bases {uu ...,«„} and {vu ..., vm}, respectively. An element w e U ® V is uniquely of the form W = £ fli/Mi ® Vj) 'J and this defines the n x m matrix (ay). We write M(w) = (ay). For g e G, we compute M(wg). Let 3E and 9) be representations of G corresponding to U and V respectively, with respect to the given bases. Write *(0) = (Ks), <D(g) = (cpq). Then («,• ® t>j)0 = u,g ® Vjg = £ feisCj,(Ms ® i>,). Thus W9 = Z bisaijCJq(us ® vq) •', j, s,«
56 Chapter 4 and, therefore, M(wg) = mT M(wW(g\ (4.14) theorem Let/elrr(G) be real valued and let 3E be a representation which affords /. Then there exists a nonzero matrix M such that X(g)TMX(g) = M for all geG. Furthermore, for any such matrix, MT = v2(/)M. Proof Let V be a C[G]-module corresponding to 3E and write W = V ® V. Let M: W -> M„(C) be as previously stated so that M(wg) = X(0)TM(w)X(0) for w e W and geG. Now W affords the character x2 and \j2, lc] = [/, /] = 1 and hence there is a one-dimensional space of G-fixed points, W0 £ W. The matrices M satisfying X(g)JMX(g) = M are precisely the M(x) for xeW0. In the proof of Theorem 4.5, we had a map *: W -» W and we decomposed W = Ws + W^, where Ws = {we W|w* = w} and H^ = {weW|w* = — w}. Also, the proof of Theorem 4.5 showed that v2(x) = 1 iff [Xa< lc] = 0, where /^ is the character afforded by W^. We must have one of the following two situations: (i) W0^Wa,[Xa,Ig] = Iv2(x)=-U (ii) W0 £ Ws, Ixa, 1C] = 0, v2(X) = 1. Thus if x e W0, we have x* = v2(/)x. Now, for we W we clearly have M(w*) = M(w)T. Thus for xe W0 we have M(x)T = v2(x)M(x) and the result follows. | (4.15) corollary Suppose / e Irr(G) is afforded by a real representation. Then v2(x) = + 1. Proof Let 3E be an R-representation which affords / and let M= XX(/i)TX(/i). heG It is clear that MJ = M and 3E(gr)T M3E(gr) = M for all geG. All that remains is to show that M^O. For any n x n real matrix X, is it clear that the diagonal entries of XJX are ^ 0, and at least one of them is >0 unless X = 0. The result now follows. |
Products of characters S7 By Problem 2.5(b), we know that it is possible that a real-valued irreducible character / is not afforded by any real representation. Such characters are exactly those for which v2(z) = — 1. The proof of this seems to require some nontrivial matrix theory. If M is a square, complex matrix, we write M* = MT. and say M is unitary if MM* = /and Mis normal if MM* = M*M. A standard theorem of linear algebra asserts that if M is normal, then there exists a unitary matrix, U, such that U~ lMU is diagonal. (4.16) lemma Let D be a diagonal matrix. Then D = E2 for some diagonal matrix E such that every matrix which commutes with D also commutes with E. Proof Suppose a,, ...,a„ are distinct complex numbers and /?,,..., /?„ e C are arbitrary. Then there exists a polynomial / such that/(a;) = /?,-. We may therefore choose / such that /(a)2 = a for every diagonal entry a of D. Then E = f(D) has the desired properties. | (4.17) theorem Let 3E be a C-representation of a group, G. Then 3E is similar to a representation 9) such that 9)(g) is unitary for all g e G. Proof Let M = £«,6C X(g)*X(gl Then M* = M and M is normal. We may choose a unitary matrix U such that U~ lMU is diagonal. Since U~i = I/*, we have (U-iX(g)U)*=U-1X(g)*U and tr'Ml/ = £ *i(ff)*3Eite). where 3E, = l/~'3El/. It is therefore no loss to assume that M is diagonal. The diagonal entries of 3E(g)*3E(g) are all real and nonnegative and since £( 1) = /, it follows that the diagonal entries of M are positive. We may therefore write M = P2 where P is a nonsingular real diagonal matrix. Since y.{g)*MH{g) = M for all g e G, we have (P-1X(&)*P)(PX(g)P-') = /. Since (PX(g)P-1)* = P_13E(gf)*P, it follows that 9) = P3EP"1 is the desired representation. | (4.18) lemma Let P*P = I and PT = P for a square matrix P. Then P = QQ~' for some matrix Q. Proof Since P is unitary, it is normal and we may write U~ iPU = D = E2, where U is unitary, D and E are diagonal, and every matrix which commutes with D also commutes with E. Now DJ = D, PJ = P, and UJ = [7 "'.This yields Z) = DT = (l/-'Pl/)T = Cr'PlJ = TJ-^UDV-^U.
58 Chapter 4 Therefore, U ' U commutes with D and hence also with E. It follows that E = Cr'l/ElT'l/and UEU'1 = UEU'1. From P*P = /, we obtain / = D*D = DD and it follows that the diagonal entries of D and hence of E have absolute value 1 and E = £_1. Now let Q = UEU-1. Then QQ-1 = (0£tr')(t/£t/-1) = (I/EIT1)2 = UDU'1 = P and the proof is complete. | (4.19) theorem Let / e Irr(G) and suppose v2(/) = 1. Then / is afforded by some real representation of G. Proof By Theorem 4.17, / is afforded by a representation X, such that X(g) is unitary for all g e G. Since / = /, X and X are similar and we may choose P such that P~XJP = 3E. Now Since v2(/) = 1, it follows that PT = P by Theorem 4.14. From the equation X(g)P = PX(g), we obtain P*X(g)* = 3E(gf)*P* and since X(g)* = 3E(gr', this yields P*X(g)'1 = 3E(g)-'P*. Thus 3EP* = P*J and 3EP*P = P*X~P = P*P3E. By Schur's lemma, it follows that P*P = a.1 for some a. Clearly, a is real and positive and we may therefore replace P by P/oc1/2 and assume P*P = /. Now Lemma 4.18 applies to yield P = QQ~' for some Q and we have QQ1X = IQQ1 or Q- '3EQ = Q~ XXQ. In other words, 9) = Q_ '3EQ is a real representation which affords / and the proof is complete. | We may summarize the situation as follows for / e Irr(G): (a) v2(x) = 0 iff/is not real; (b) v2(x) = 1 iff x is afforded by a real representation; (c) v2(z) = — 1 iff/ is real but is not afforded by any real representation. One further remark should be made. If z e Irr(G) and v2(/) = — 1, then X(l) must be even. This may be seen as follows. The nonzero matrix M in Theorem 4.14 is nonsingular by Schur's lemma. If v2(/) = — 1 then MT = -Mand det(M) = det(MT) = det(-M) = (-1)*(1) det M. It follows that (-1)*(1) = 1 and /(l) is even.
Problems 59 We close this chapter with a discussion of the characters of direct products. (4.20) definition Let G = H x K and let <p and 9 be class functions on H and K, respectively. Define / = cp x y by %(hk) = (p(h)S(k) for h e H and keK. Observe that cp x y is a class function of G. If cp is a character of H, then under the isomorphism H s G/X, there is a corresponding character q> of G with X ^ ker 0 and <p(/i/c) = (p(h). Similarly, if y is a character of K, there is a corresponding character y of G, with §(hk) = y(/c). It follows that q> x y = $y is a character of G. (4.21) theorem LetG = H x K. Then the characters <p x y for <p e Irr(H) and y £ Irr(K) are exactly the irreducible characters of G. Proof Let <p, <p, e Irr(H) and y, y, e Irr(K). Let / = cp x y and /, = <p, x y,. Then [/, Zi] = iTT, E X(0)Z^)" = 777^7 £ (pQWkfaWAk) \(j\gsG \H\\K\ hBB:kEK = (j^ £/{h)V^)(\k\ X ^W)) = l>, <P.][S, Si]- It follows that the <p x y are all distinct and irreducible. Now £ (? x y)(i)2 = X<p(i)29(D2 = (X<p(i)2YxW ^elrr(H).,Selrr(K) ^.S \^ /\s = |H||K| = |G|, and thus the cp x y are all of Irr(G). | Problems (4.1) Let / and \ji be characters of G. Express det(/i/0 in terms of det / and deti/f. Hint It suffices to assume that G is cyclic. (4.2) (Garrison) Let Jf be a class of G which is not contained in any proper normal subgroup. Let K be the corresponding class sum in C[G] and let m be the number of distinct values of cox(K) for / e Irr(G). Show that every element of G can be written as a product of fewer than m elements of Jf.
60 Chapter 4 Hint Mimic the proof of Theorem 4.3 using the second orthogonality relation. (4.3) Let G = H x K. Let <p e Irr(H) and 9 e Irr(X) be faithful. Show that cp x 9 is faithful iff (| Z(H) |, | Z(K) |) = 1. (4.4) Suppose G = HK with H <= QK). (a) Let /elrr(G). Show that Xh = S(l)<p and Xk = <p(l)9 for some <p e Irr(H) and 9 e Irr(X). (b) Let <p e Irr(H) and 9 e Irr(X) and suppose that q>HnK and 9HnK have a common constituent. Show that there exists a unique / e Irr(G) with Xn= 9U)<pandxK = <p(l)9. Hint G is a homomorphic image of H x K. (4.5) Let G = H x A where A is abelian and let n > 0 with (|H|, n) = 1. Show that x(n) e Irr(G) for every / e Irr(G). (4.6) Let n > 0 and assume that x(n) e Irr(G) for every / e Irr(G). Show that G = H x A, where A is abelian and (|H|, n) = 1. Hints (a) Let rf = (| G |, n). Show that it is no loss to assume that (|G|/d,n)=l. (b) Let A = f)^irr(G) ker x(n). Show that A = {geG\gn = 1} and \A\ = d. (c) Let H = f){ker zlze Irr(G), z(n) = 1G}. Show \G : H\ = d. (4.7) Let V be a C[G]-module with basis {vu ...,v„} and let p be a prime. Let W = F ® • • • ® FQ? times) and define a on W so that (xt ® • • • ® xpf = x2 ® x3 ® • • • ® xp ® Xi for all x, e 7. Let e be a primitive pth root of 1 in C. Let Wv = {weWlw" = ew}. Show that Wt is a C[G]-submodule of W which affords the character Xi = (Xp — X<p))/P> where / is the character afforded by V. Conclude that x(m) is a difference of characters for all m > 1. Hints Let "W = {t>fl ® • • • ® vip \ not all i3 are equal}. For we if, let w = w + e~lw* + e~2w"2 + ••• + e~<p-1V"'. Now <a> permutes niT into orbits of size p. Let 1^0 be a set of representatives of these orbits. Then {vv| w e "W^\ is a basis for Wt. (4.8) Let G be a p-group with Frattini factor group of order > p2a~l. Show that the number of elements of order p in G is = — 1 mod p". Hint Mimic the proof of Theorem 4.9. Use the fact that vp(x) is an integer (Problem 4.7).
Problems 61 (4.9) If/e Irr(G), then |v2(/)| <, 1. Show that no absolute bound on v„(x) exists for any n > 2. Hint There exist groups of prime exponent p#2 and of exponent 4 with irreducible characters of arbitrarily large degree. (4.10) Let V, W, Vu and Wt be F[G]-modules for an arbitrary field F. Fix a particular F-basis in each of these modules and construct V ®W and Vt ® Wt as was done for C[G]. Now suppose a: V -» Vt and /?: W -* Wx are module homomorphisms. Show that there exists a module homomorphism y.V ®W -^ViQWi such that y(v (g) w) = <x(v) (g) 0(w) for t> e 7 and weW. Conclude that V ® W is determined up to module isomorphism independently of the choice of bases. (4.11) Let G be simple and let S e Syl2(G) be elementary abelian of order q. Suppose S = CG(x) for all x, 1 # x e S. Show that v2(x) = 1 for all / e Irr(G) and that x(x) = /(l) - q for / ^ 1G and xeS. Hint First show that G contains exactly | G | /q involutions and that the remaining elements of G have odd order. This is done without characters using the fact that if x, y are nonconjugate involutions, then |Z«x, y»| is even. Note This problem is continued as Problem 5.20 where it is proved that \G\=q(q- l)(q+ 1). (4.12) Let x, «A, 9 e Irr(G). Show that [#, 9] < 9( 1).
Induced characters Let H £ G be a subgroup. Given a character / of G, we obtained the character Xh °f # by restriction. In this chapter we study an approximately dual process, where characters of G are induced from characters of H. (5.1) definition Let H £ G and let <p be a class function of H. Then <pG, the induced class function on G, is given by I" I xeG where <p° is defined by <p°(h) = q>(h) ifheH and <p°(y) = 0 if y £ H. Observe that <pG is really a class function on G and that <pG(\) = \G: H\q>(\). Another useful formula for (pG(g) is obtained by choosing a transversal T for the right cosets of H in G (that is, a set of representatives for these cosets). It is then easy to see that 9G(g)= ?,<p0(tgt-1). (5.2) lemma {Frobenius Reciprocity) Let H £ G and suppose that cp is a class function on H and that 9 is a class function on G. Then [>, Sfl] = l>G> 8]. Proof We have [>G, 9] = t^ E <PG(0)%) = r^rj ^ E E 9°(xgx- l)W\ |G| 9eG |w| \"\ gsG xeG 62
Induced characters 63 Setting y = xgx l and observing that 9(g) = 9(y), we obtain I O I \H\ yBGxBG \tI\yeB (5.3) corollary Let H £ G and suppose <p is a character of H. Then <pG is a character of G. Proof Let / e Irr(G). Then Xh is a character of H and thus [<pG, /] = [<P> Z«] is a nonnegative integer. This proves the result since cpG j= 0. | (5.4) corollary Let H £ G and suppose <p e Irr(H). Then there exists X e Irr(G) such that <p is a constituent of Xh ■ Proof Take / to be an irreducible constituent of (pG. Then 0 * [>G, z] = L>, Zfl] and the result follows. | (5.5) corollary Let G be abelian and suppose H £ G. Then every 1 e Irr(H) is the restriction of some fi e Irr(G). Induction of characters is closely related to the transfer map from a group into an abelian factor group of a subgroup. Many of the results provable by transfer can also be proved using induced characters. We give one example of this now. (5.6) theorem Let G have an abelian Sylow p-subgroup. Then |G'nZ(G)| is not divisible by p. Proof Assume the contrary and choose U £ G' n Z(G) of order p, with U £ P e Sylp(G). Let X e Irr(I/), X * lv, and choose fi e Irr(P) with /% = 1. Now write V°= X «** xelrr(G) and observe that fiG( 1) = /z( 1) | G: P | = | G: P | is prime to p. Therefore, there exists x £ Irr(G) such that ax =£ 0 and p^ 1). We have 0 * [>G z] = D*, ZJ and thus X is a constituent of Xv Since 1/ £ Z(G), we have Xv = z(*M and (det z)t; = ^*(1)- (See Problem 2.3.) Since U e G', we have (det x)v = lu and thus XxW = lv. Since pJfxi^X ^ ^ li/> and 1^1 = P> this is a contradiction and proves the theorem. |
64 Chapter 5 Explicit computation of induced characters is extremely useful for the construction of character tables, despite the fact that cpG is usually reducible even if cp is irreducible. [Note that if <p is reducible, then <pG is necessarily reducible since (cpi + <p2)G = <PiG + (P2°-l Given H £ G, cp a character of H and g e G, an efficient way to compute cpa(g) explicitly is to choose representatives, xu ..., xm for the classes of H contained in Cl(g) in G and to use the formula cpG(g) = \CG(g)\l j^- where it is understood that (pG(g) = 0 if H n Cl(g) = 0. This formula is immediate from the definition of <pG since as x runs over G, xgx~i = Xj for exactly | CG(g) | values of x. We shall now construct the character table of G = A5. There are five conjugacy classes; one each of elements of order 1, 2, and 3 and two of elements of order 5. It is convenient to label these classes 1, 2, 3, 5l5 and 52 in the table. Let Xi = 1g- We have: Class Jf; 1 2 3 5t 52 \C(x)\iorxeJf: \X\ 60 4 3 5 5 1 15 20 12 12 11111 Now let H = y44 £ G and compute (1H)G. We have (1H)G: 5 12 0 0 using the above formula. [For instance, if o(g) = 3, there are two classes in H of elements conjugate in G to g. If xt and x2 are representatives of these classes, we have lH(Xi) = 1 and |CG(g)| = 3 = |CH(xj)| and thus (lH)G(gf) = 2 as listed.] Now [(1H)G, 1G] = [1h, 1h] = 1 by Frobenius reciprocity and thus (1H)G — 1G is a character. We call this character /4. Thus n = iUf- iG: 4 0 i -l -l. Direct computation now yields [/4, /4] = 1 and we have found an irreducible character. Next, let 1 be a nonprincipal linear character of H = A4 so that l(x) = 1 if o(x) = 1 or 2 and l(x) = co = e2nil3 for one class of elements of order 3 in H and X(x) = a2 for the other class. We get X5 = 1G: 5 1 -1 0 0 since co + co2 = -1. We compute [/5, x5] = 1 and so Xs e Irr(G).
Induced characters 65 Now let K be a cyclic subgroup of G of order 5 and let /z e Irr(K), fi =£ lK. The class 5t of G contains a pair of inverse elements of K, say x and x~i. Choose fi so that fi{x) = e = e2nil5. Since x2 and x~2 lie in the class 52 we have HG: 12 0 0 e + e4 e2 + e3. Computation yields [/iG, /5] = 1 and thus HG - Xs- 7 -1 1 e + e4 e2 + e3 is a character. Further computation yields that [/iG — Xs, /4] = 1 and hence Z3 = V° - Xs ~ U- 3 - 1 0 e + e4 + 1 e2 + e3 + 1 is a character and [/3, x3] = 1 so that /3 is irreducible. Finally, replacing fi above, by v e Irr(X) with v(x) = e2 yields ip3: 3 - 1 0 e2 + e3 + 1 e + e4 + 1 is irreducible and we have found all five irreducible characters of G. We now consider the module theoretic interpretation of induced characters. (5.7) definition Let V be a F[G]-module. Suppose V = Wf + • • • + Wk where the Wt are subspaces of V which are transitively permuted by the action of G. Then V = Wv + ■ ■ ■ + Wk is an imprimitivity decomposition of V. If V is irreducible and has no proper imprimitivity decomposition (that is, with k > 1), then V is a primitive F[G]-module. We emphasize that if V = Wt -j- • • • 4- W* is an imprimitivity decomposition of V with /c > 1, then the Wt are definitely not F[G]-submodules of V. In this situation, let H = {g eG\Wth = WJ, then Wt is a F[H]-module. (5.8) theorem Let V = Wt + ■■ ■ + Wkbe an imprimitivity decomposition of the F[G]-module V. Assume FgC, Let H be the stabilizer of Wv Suppose V affords the character x of G and Wv affords the character 9 of H. Then x = 9G- Proof Let T be a right transversal for H in G and write W = Wv Then Wt runs over the Wt as t runs over T and V = £ • Wt. Choose a basis, wt,..., wm for W so that {wj 111 < i <, m, t e T} is a basis for V. Let g e G. We compute xig) using this basis.
66 Chapter 5 For te T, the contribution of the basis vectors wtt to %(g) will be zero unless Wtg = Wt, that is, tgt~l e H. Assuming tgt~1 = heH, write w-th = 5] a/;w; so that £|"=1 ait = 9(rgr~')- Now, (Wjt)9 = Wj/if = 5] fluw;f an(* hence the contribution of w{ t to /(g) is a^. Therefore, the total contribution of the w,t, 1 < i < m, is 9(tgt~'). It follows that x(g) = ^(eT 9°(fgr') = $G(g) as desired. | (5.9) theorem Let H c g and let W be an F[H]-module. Then there exists an F[G]-module Shaving an imprimitivity decomposition V = £• Wt, where H is the stabilizer of Wt and Wt s W as F[H]-modules. Proof Let V be the external direct sum of | G: H | copies of W as an F-vectorspace. Let T be a right transversal for H in G and assume 1 e T. Let W ® t, for t e T, denote subspaces of V, each isomorphic to W, such that ^ = E"<<et(W® t)- Fix isomorphisms a,: W-> W ® t and let w (g) t denote at(w) so that W ® t = {w ® t \ w e W}. Now for weW and g e G, define w ® ge V as follows. Write g = ht, with fi£fl and te T, and set w ® g = vWi ® t. Note that for we W, xeH, and g e G, we have wx ® g = w ® xg. (The usual notation for what we have constructed is V = W ® f[H]F[G].) We now let G act on V by defining (w ® t)g = w ® tg for g e G. Observe that (w ® gj^j = w ® g^ for all gu g2e G. Extending this definition by linearity gives V the structure of an F[G]-module and V = £ -W®t is an imprimitivity decomposition. Clearly the stabilizer of W ® 1 is H and the map at: W -^ W ® 1 is an isomorphism of F[H]-modules. The proof is complete. | If s e Ht, then W ® s = W ® t in the above construction and it is reasonably clear that the module V is independent of the choice of coset representatives. We write V = WG. Note, however, that we have not yet proved that WG is the unique (up to F[G]-isomorphism) F[G]-module which satisfies the conclusion of Theorem 5.9. In the special case that F = C, let W afford the character 9. Then by Theorem 5.8, it follows that any C[G]-module which satisfies the conclusion of 5.9 affords / = 9G. This proves uniqueness in this case. The general case is left to the problems. Note that Theorems 5.8 and 5.9 provide an alternate proof of the fact that induced characters are characters. Another consequence of these results is that if i e Irr(G), then / is afforded by a primitive module iff / =£ 9G for any character 9 of a proper subgroup of G. Such / are called primitive characters.
Induced characters 67 (5.10) definition Let x be a character of G. Then / is monomial if / = AG, where X is a linear character of some (not necessarily proper) subgroup of G. The group G is an M-group if every / e Irr(G) is monomial. Note that if / is monomial, then by Theorems 5.8 and 5.9, / is afforded by a module which has an imprimitivity decomposition into subspaces of dimension 1. It follows that i is afforded by a representation X with the property that each row and column of X(g) has exactly one nonzero entry for each g e G. (5.11) lemma Let 9 be a character ofHsG. Then ker(9G) = f] (ker 3)x. xeG Proof Let / = 9G. Then g e ker / iff Yj$°(xgx-1)= £3(1). xbG xbG Since |9°(xgx-1)| < 9(1), we conclude that geker / iff ^(xgx"1) = 9(1) for all x e G. This happens iff g e (ker 9-)* for all x and the proof is complete. | (5.12) theorem Let G be an M-group and let 1 = ft < f2 < ■ ■ ■ < fk be the distinct degrees of the irreducible characters of G. Let / e Irr(G) with Z(l) = fi- Then G(i) £ ker /, where G(i) denotes the ith term of the derived series of G. Proof If i = 1, then / is linear and G' £ ker /. Assume i > 1 and work by induction on i. If ifre Irr(G) and t/f(l) < /(l), then lA(l) = f} for j < i and G(.--DcGU)£ker^. Since G is an M-group, we may choose H < G and linear 1 e Irr(H) with X = 1G. Now [(1H)G 1G] = [1«> 1«] = 1 and thus (1H)G is not irreducible. If \p is any irreducible constituent of (1H)G, then \p( 1) < (1H)G(1) =\G:H\ = XG(l) = /(l) and thus G(i_ u £ ker i>. It follows that G(i_ u £ ker((lH)G) £ H by Lemma 5.11. Now G<0 £ H' £ ker 1 and since G<0 -a G, we have G<0 £ f) (ker Xf = ker / and the proof is complete. | (5.13) corollary (Taketa) Let G be an M-group. Then G is solvable. It is not the case that all solvable groups are M-groups. [The smallest counterexample is SL(2, 3) of order 24.] In Chapter 6 we shall discuss some sufficient conditions. There is no known characterization of M-groups other than in terms of characters.
68 Chapter 5 We shall now discuss some of the connections between character theory and the theory of permutation groups. Suppose the group G acts on the finite set Q. In other words, for each a e Q and g e G, an element a9 e Q is defined such that a1 = a and (a9)*1 = a9*1 for a e Q and g,heG. For a e Q, we shall use the notation Ga = {g e G|a9 = a} so that \G:Ga\ = \6\, where & is the orbit containing a. Let G act on Q. For g e G, let %(g) = | {a e Q | a9 = a} |. The nonnegative integer valued function / is called the permutation character associated with the action. To see that / really is a character, let V be a C-space with a basis identified with fi. Let G act on V by permuting the basis. This makes V into a C[G]-module called a permutation module. It is clear that V affords /. (5.14) lemma Let G act transitively on Q, let a e Q and let H = Ga. Then (1H)G is the permutation character of the action. Proof Let V be the permutation module with basis fi. Then V = E • 0efiC/? is an imprimitivity decomposition for V. The result is now immediate from Theorem 5.8. | (5.15) corollary Let G act on Q with permutation character /. Suppose Q decomposes into exactly k orbits under the action of G. Then [/, 1G] = k. Proof Write Q = (J*= t (9{ where the 0j are orbits. Let Xt be the permutation character of G on &x so that / = £ xf. For aefi);, we have %t = (lGa)Gby Lemma 5.14. Thus [*„ 1G] = [(1G/, 1G] = [1G„, 1GJ = 1. Therefore, [/, 1G] = £ [/i, 1G] = /c and the proof is complete. | Corollary 5.15 can also be proved as follows. Let &x be as above with ae0».Then|0,| = |G|/|GJ. Now let <f = {(a,g)|aefi,geG, a9 = a} and compute | y | two ways. We have Ez(0) = m= Eigj = £ £igj = E i M = £|Gi = *iGi and the result follows. (5.16) corollary Let G act transitively on Q with permutation character /. Suppose that aefl and that Ga has exactly r orbits on Q. (One of these is {a}.) Then [*, *] = r. Proof We have r = LXg., IgJ = [X, (1g„)g] = ih Z\ by 5.15, 5.14 and Frobenius reciprocity. |
Induced characters 69 The integer r in Corollary 5.16 is called the rank of the transitive action. Recall that G is doubly transitive (or 2-transitve) on Q if |Q| > 2 and Ga is transitive on Q — {a}, that is, if r = 2. (5.17) corollary Let G act on Q with permutation character /. Then the action is 2-transitive iff / = 1G + \ji where \ji e Irr(G) and \\i j= 1G. Corollary 5.17 is often useful for finding irreducible characters. For instance, it shows that the symmetric group E„ has an irreducible character of degree n — 1 for n > 2, and that this character restricts irreducibly to A„ for n > 4. We now consider the question of finding (not necessarily normal) subgroups of a group G from information about its characters. Suppose we wish to prove that G has a subgroup H of fairly small index. This could be done if the character (1H)G could be recognized as such. Because (1H)G is a transitive permutation character (of the action on the set of right cosets of H), there are a number of necessary conditions it must satisfy. (5.18) theorem Let H £ G and let / = (1H)G. Then (a) Z(D||G|; (b) [z,£|£W)for,MIrr(G); (c) [x, lc] = 1; (d) x(g) is a nonnegative rational integer for all g e G; (e) xig) < X(9m) for g e G and integers m; (f) X(g) = Oifo(g)Jf\G\/x(l); (g) x(g)\Cl(g)\/xW is integral for all g e G. Proof Most of these assertions are immediate from the fact that z is a transitive permutation character. Statement (b) follows since [X, -A] = E(1h)G, •/'] = [!«, -Ah] ^(l)- Statement (f) follows since |H| = |G|/z(l) and if o{g)Jf\H\, then no conjugate of g lies in H and so (lH)G(g) = 0. To prove (g), let Q be the set of right cosets of H so that x is the character of the action of G on Q. Let X = Cl(g) and let y = {(a, x)| a e Q, x e Jf, and ax = a}. Since / is constant on Jf, we have \Jf\x(9) = \y\= ElXnGJ. Since all Ga are conjugate in G, | Jf n GJ = m is independent of a and we have | Jf |xig) = m\0.\ = m/( 1), proving the result. | The necessary conditions given in Theorem 5.18 are definitely not sufficient to guarantee that z is a permutation character. For example, the
70 Chapter 5 simple group M22, of order 443,520, has an irreducible character of degree 55. The character obtained by adding the principal character to it satisfies the conditions (a)-(g) and yet M22 has no subgroup of index 56. The following result can often be used to find proper subgroups of a group. (5.19) theorem (Brauer) Let / be a character of G with [/, 1G] = 0. Let A, B £ G and suppose Then A and B generate a proper subgroup of G. Proof Let U be a C[G]-module which affords / and let V, W and Y be the subspaces of fixed points of U under A, B and A r\ B respectively. Then V ^ Y,W£ 7,and dim V + dim W = [jA, 1J + [/B, 1B] > [xAnB, lAnB] = dim Y. It follows that V r\ W ^ 0 and thus {A, B} has nontrivial fixed points on U. Since [/, 1G] = 0, we conclude that (A, B} < G. | An example showing how Theorem 5.19 can be used is the following proof that A6 is the unique simple group ofits order. [Since | A61 = |PSL(2, 9)|, it follows that A6 s PSL(2,9).] As is often the case when trying to identify a particular group up to isomorphism, there is a great deal of tedious and detailed work involved. Since the result hardly seems to be worth that much effort, the proof which follows is somewhat more sketchy than most in this book. (5.20) theorem Let G be simple and suppose \G\ = 360 = 23 • 32 • 5. ThenG s A6. Proof Suppose G0 Z G with | G: G01 = k > 1. Then G is isomorphic to a subgroup of the alternating group Ak. Therefore k > 6 and the result follows if we can find G0 with k = 6. The only divisors of 360 which are = 1 mod 3 are 1, 4, 10, and 40. Let PeSyl3(G) and N = NG(P). We conclude that \G:N\ = 10 or 40. By Burnside's transfer theorem, N > P and hence | G: N \ = 10. Now suppose P <=: M <G. We claim M £ N. If not, then 1 < | M: M n N \ <, 10 and hence | M: M n N \ = 4 or 10 by Sylow's theorem. Since the Sylow 3-subgroups of G generate all of G, we have | Syl3(M)| < 10 and thus \M: M n N\ = 4 and \G:M\ divides 10. Thus \G:M\ = 10 and M r\ N = P. Now application of Burnside's theorem to M yields K<M with | X | = 4. Thus 811 NG(X) | and P £ NG(X) so that | G: NG(X) | < 5. This contradiction shows M £ N as claimed.
Induced characters 71 Now let P ± Pt e Syl3(G) and set D = P r\ Pv Then Pt £ N(£>) so that N(£>) £ N. However, P s N(£>) and we conclude that N(£>) = G so thatD = 1. Next we claim that G contains no element of order 6. If x is an even permutation on 10 points and o(x) = 6, then x must contain a 2-cycle and x2 fixes at least two points. Since no element of G of order 3 lies in two distinct Sylow 3-subgroups, the claim follows. In particular, N/P acts faithfully on P. Since the automorphism group of a cyclic group of order 9 has order 6, it follows that P is elementary abelian. Furthermore, each involution in N inverts all elements of P. Now P = C(y) for each l#yeP and it follows that N has exactly six conjugacy classes and thus exactly two nonlinear irreducible characters each of degree 4. If 1G j= % e Irr(G), then %N must have one of these as a constituent since N' £ ker /. Therefore x(l) > 4. Let %e Irr(G) with /(l) odd. Since 4^/(1), it follows that Xn has a linear constituent X and thus x is a constituent of XG. Therefore, /(l) < 1G(1) = 10 and if / t6 1g we have z(l) = 5 or 9. We will show that G has an irreducible character of degree 5. From Zxeirr(G) X(l)2 — 360 = 0 mod 4, it follows that there exist at least three nonprincipal irreducible characters of odd degree. It suffices to show that there is at most one of degree 9. In fact we show that if / e Irr(G) and /(l) = 9, then / is a constituent of (lN)G. Indeed, / is a constituent of Xa for some linear character X of N. Thus XG = / + y., where /z(l) = 1 and thus fi = 1G. Since [fiN, 1] t6 0, it follows that X = 1^ as claimed. Now fix i e Irr(G) with /(l) = 5. Then %N has a linear constituent and so [Xp> lp] t6 0. Let v be a nonprincipal linear constituent of %P. Since every element of P is real, %P is real valued and {jP, v] = [/P, v]. Let 1/ = ker v, so that | U\ = 3. It follows that \jv, 1^] > 3 and hence by Theorem 5.19, any two conjugates of U generate a proper subgroup of G. Note that we can find ten conjugates of U such that no two of them centralize each other. Let V be a conjugate of U which does not centralize U, and let H = (U, V} < G. If \G: H\ = 6, we are done, so assume that |H| < 60. Now U e Syl3(H) or else P S H and thus H £ N. This would imply V £ P which is not the case. Since I/, V e Syl3(H), it follows that 3 x 4 or 3 x 10 divides |H| and thus | H\ = 12, 24, or 30. If | H | = 30, then there exists S<fl with | S \ = 5. Therefore, | G: N(S) | = 1 mod 5 and divides \G:H\= 12. Thus |G: N(S)| = 6. We may assume that this does not occur and hence |H| = 12 or 24. The only group of order 24 which is generated by elements of order 3 has a normal subgroup of order 8 and thus contains elements of order 6. We conclude |H| = 12 and there exists K<H, elementary of order 4. Next, we claim \jH, 1H] = 2. To see this, we use the fact that the unique
72 Chapter 5 nonlinear irreducible character 9 of H has degree 3 and is real. It follows that [#i/> It/] = 1 and thus Xh = 9 + lt + 12 where (Xt)D = 1^. It follows that 1; = 1H and the claim is established. Also [xv, lu] = 3. Now choose W conjugate to U with W £ H and W not commuting with U. Let L= (U, W). We may assume |L| < 60 and thus \L\ = 12 and H r\L = U. Since [/H, 1H] = 2 = [/L, 1 J, Theorem 5.19 applies again to show R = <H, L> < G. Since R £ N, it follows that 9^1/? |. Reasoning as before, \R\ ^ 24. Since 12||/?|, this forces \R\ = 60 and the proof is complete. | It is often convenient to be able to write a character of a group in terms of characters induced from linear characters of subgroups of a specified type. For example, an important theorem of Brauer (which we will prove in Chapter 8) asserts that every character is a Z-linear combination of characters induced from linear characters of nilpotent subgroups. For now we consider only rational valued characters and cyclic subgroups. We prove the following. (5.21) theorem (Artin) Let/be a rational valued character of G. Then y an , g *■ ^|N(H):H| " ' where H runs over the cyclic subgroups of G and aH e Z. The proof of Artin's theorem depends on a well-known result of algebraic number theory, namely that the cyclotomic polynomials are irreducible over Q. In other words, for each positive integer n, all of the primitive nth roots of unity are conjugate over Q. (This fact is also needed for Problem 3.11.) (5.22) lemma Let i be a rational valued character of G and let x, yeG with <x> = <y>. Then /(*) = x(y). Proof Let n = o(x) = o(y) and let e be a primitive nth root of unity. We have y = xm, with (m, n) = 1 and thus em is a primitive nth root of unity and em = e" for some automorphism a of Q[e]. Now %(x) = Y.f=l ei' where each e; is an nth root of unity and hence is a power of e. We have x(y) = x(xl = E *r = £ e," = x(x)a. Since x(x) is rational, we have x{x)a = %(x) and the proof is complete. | Proof of 5.21 Define an equivalence relation = on G by x = y if <x> and <>>> are conjugate in G. It follows from Lemma 5.22 that / is constant on the equivalence classes under = and thus x is a Z-linear combination of the characteristic functions of these classes.
Problems 73 Let #!, ^2,..., (€m be the distinct (= )-classes of G and let 4>, be the characteristic function of <€t so that <&;(*) = 1 if x e #, and <J)((x) = 0, otherwise. Choose representatives xte <€t and let Hf = <x(> and nt = \Ht\. We have (1) |«f,| = |G:N(Hl)|<»(fil). We prove by induction on nt that (2) \nnd\^>i = Y.aJ^H1)G i for suitable as e Z where a,- = 0 unless H} is conjugate to a subgroup of Hf. The result will then follow since \N(Ht)\ divides |N(H,)| when a} ¥= 0. If n, = 1, then ^ = {1} and |G|<D( = (1H()G and (2) holds in this case. Suppose Hi > 1. Write (1h,)g = 5>,4>; and compute the coefficients b3 as follows using Frobenius reciprocity and the fact that [(D;, <Dk] = <5;k|<^.|/|G|: b;|^.|/|G| = [(1H()G, <&j] = [lfl„ (*/)*,]. Now (<D;)H. = 0 unless H; is conjugate to a subgroup of Hj. In that case, (<D;)Hi takes on the value 1 on <p{n}) elements of Ht and vanishes elsewhere. Thus [1H(, (<bj)nj = <P(nj)/ni and we conclude that bJ=\G\<p(nj)/nt\Vj\ = \N(HJ)\/nl using (1). Therefore, n,(WG = ElN(H;)|<Di, where j runs over the set of subscripts for which H3 is conjugate to a subgroup of Hj. Solving for |N(H()|<I>; and applying the inductive hypothesis yields (2) and proves the theorem. | (5.23) corollary Let x be any rational valued character of G. Then IG | X = Eh an l»°> wnere H runs over tne cyclic subgroups of G and aH e Z. (5.1) Let H £ X £ G and suppose that <p is a class function of H. Show that (9Kf = <PG- (5.2) Let H, K £ G with H/C = G and suppose <p is a class function of H. Show that (9G)K = (9HnK)K.
74 Chapter 5 (5.3) Let H £ G and suppose <p is a class function of H and i/f is a class function of G. Show that {(p\liH)G = ip6!/'. Note Problems (5.1)-(5.3) will be used frequently. (5.4) Let b(G) = max{x(l)|/e Irr(G)}. If H £ G, show that b(H) < b(G) < \G:H\b(H). Note Since H is abelian iff b(H) = 1, the inequality b(G) <\G:H\b(H) generalizes Problem 2.9(b). (5.5) Let F be an arbitrary field and let H £ G. Let W be an F[H]-module. Show that there is a unique (up to F[G]-isomorphism) F[G]-module V such that V has an imprimitivity decomposition V = £ • Wf, where H is the stabilizer of Wl and Ws W^ as F[H]-modules. Notes By Theorem 5.9, V s WG. In the following problem, if V is an F[G]-module and H s G, we use the notation FH to denote the module V viewed as an F[H]-module. (5.6) (Mackey) Let H, K £ G and let T be a set of double coset representatives so that G = (J HtK is a disjoint union. Let W be an F[H]-module for an arbitrary field F. Show that (WG)K= l-(WH.nKf. Note Mackey's theorem generalizes Problem 5.2. (5.7) Let H s G and suppose \ji is a character of H. Let X £ G and assume (tAG)K e Irr(X). Show that HK = G. Note Problem 5.7 is an immediate consequence of Problem 5.6. It can, however, be done without using modules. The key step is to show that K^g)k, (ipHnicf] * 0 and then conclude that \G: H\ <\K:K r\ H\. (5.8) Let / be a monomial character of G and suppose K c g with Xk £ Irr(X). Show that %K is a monomial character of K. (5.9) Suppose that F is a subfield of C and let H £ G. Suppose that 9 is a character of H which is afforded by an F-representation. Show that 9G is afforded by an F-representation. Note It follows that every monomial character of G is afforded by a O [^-representation, where e is a primitive nth root of unity, n being the
Problems 7S exponent of G (that is, the least common multiple of the orders of the elements of G). Clearly, an arbitrary character of G has values in Q[e]. In Chapter 10 we shall prove the theorem of Brauer which asserts that every character of G is afforded by a Q[e]-representation. (5.10) Let H s G and view C[H] s C[G]. Let / be a right ideal of C[H] and let J = 7C[G]. Suppose that I affords 9 (viewing I as a C[H]-module). Show that J affords 9G. (5.11) Let N <i G and /elrr(G). Suppose that [/N, 1N] # 0. Show that N £ leer/. Hint Use Lemma 5.11. (5.12) Let cpG = / e Irr(G) with <p e Irr(H). Show that Z(x) £ H. (5.13) (L. Solomon) Show that each row sum in a character table is a nonnegative rational integer. Hint Let G act on G by conjugation. Consider the permutation character. Note The column sums are also integers. They can be negative. (5.14) Let G be nonabelian and let/ = min{x(l)|xeIrr(G), x(l) > l}.Show (a) If|G'| < /thenG's Z(G), (b) if | [G, G'] | <, f, then G' is abelian, (c) ifHs Gand|G:H| </,thenG'£H. Hint For (a). If x e G, then | Cl(x) | :£ | G' |. (5.15) Let HsG and suppose <p is a character of H with det cp = 1H. Let X = (pG and show (det x)2 = 1G. (5.16) Let H be a maximal subgroup of G and let / = (1H)G. Let \p be a nonprincipal irreducible constituent of /. Show ker \ji = ker /. (5.17) Let H <=, G and let / = (1H)G. Fix a positive integer n. For g e G, let m(#) = [/<»>> !<»>] and define 9(g) = nm<9). Show that 9 is a character of G. Hint Let Q be the set of right cosets of H in G. Fix a set S with | S | = n and let A be the set of all functions Q -> S. Then G acts on A. Note This result provides another necessary condition to add to the list in Theorem 5.18.
76 Chapter 5 (5.18) Let G be a group. Write a(n) = |{geG\o(g) = n}\. Then the polynomial /(*) = ( 1/| G|)£a(m)x'Gi/m m takes on integer values whenever xeZ. Hint Use Problem 5.17. (5.19) Let G be doubly transitive on fi and let H £ G with \G: H\ < |fi|. Show that H is transitive on Q. (5.20) Assume the notation and hypotheses of Problem 4.11 and show that \G\ = q(q - l)(q + 1). Hints Let N = N(S) and show \N\ = q(q- 1). Let Xelrr(N) with 1 ^ In and S £ ker 1. Let xeS, x 76 1. Show lG(x) = 1(1). Write 1G = E^irr(o axx and conclude 4D = E «,(Z(1) - 1) = IG: N|l(l) - q £ flr Finish the problem by showing that if ax =£ 0, then x(l) > q + 1 so that Y.ax ^ W)- To show /(l) > <j + 1, compute [/s, Is] and use the fact that since fe, 1] > 0 and / is real, then [/s, ls] ^ 2. Note Using the result of this problem, it is not very hard to prove that G = SL(2, q), a theorem originally due to Brauer, Suzuki, and Wall. (5.21) Let P(G) denote the set of all Z-linear combinations of characters of of G of the form 1HG for H £ Q. Show that P(G) is a ring. (5.22) Let / be a rational valued character of G and let nG(x) be the least positive integer such that nG(x)x £ P(G). Let N = ker / so that / may be viewed as a character of G/N. Show that nG(x) = nGIN(x). Note The existence of nG(x) is guaranteed by Corollary 5.23. Hint If nx = £%(1h)g, then nx = £ Ah(1\h)g- Show this by writing (1H)G = (1,vh)g + £<H), where £m is a character of G with the property that N is not contained in the kernel of any of its irreducible constituents. (5.23) Let x be any character of G. Show that \G\x = la,XG, where 1 runs over the set of linear characters of cyclic subgroups and ax e Z. Hint Use Problem 5.3.
Problems 77 (5.24) Let G be a doubly transitive permutation group on Q and let a, /? e Q, with a#^. Let q> e Irr(Ga) and assume that (pG f e Irr(Ga^). Show that [<pG, <pG] <; 2. Hint Use Problem 5.6. (5.25) Assume that every Sylow subgroup of G has a faithful irreducible character. Show that G has one also. Hint Let S be the product of all of the minimal normal subgroups of G. (This is called the socle of G.) Then S is a direct product of simple groups. Find 9 e Irr(S) such that ker 9 contains no nontrivial normal subgroup of G. (5.26) (Passman) Let X £ G - {1} be a subset and let n = \X\. Assume for all primes p < n, that a Sylow p-subgroup of G is cyclic. Show that there exists i e Irr(G) such that X n ker(/) = 0. Hints Reduce to the case that G is abelian by observing that it suffices to assume that G = <AT> and that if [x, y] $ ker(x), then both x and y £ ker /. In the abelian case, let N s G be maximal such that N r\ X = 0. Show that G/N is cyclic. Note If n = 2, the hypothesis on Sylow subgroups is vacuously satisfied.
Normal subgroups Let x e Irr(G) and H s G. In general, very little can be said about the restriction in • The situation is quite different if H is normal in G. Let H <i G. If 9 is a class function of H and g e G, we define 99: H -> C by 9('(/i) = $(ghg~l). We say that 9" is conjugate to 9 in G. (6.1) lemma Let H o G and let <p, 9 be class functions of H and x, y e G. Then (a) (p* is a class function; (b) {ffY = qf>\ (c) [>*,9*] = [>,9]; (d) [Zh, <P*] = [in, <p] for class functions % of G; (e) <?)•* is a character if <p is. Proof (a) If /i and k are conjugate in H, then so are /i9 and kg for all geG. Take g = x_1. (b) Compute. [Note that had we defined Bg(h) to be &(hg), then (b) would fail.] (c) The two sums are identical since xhx~l runs over H as h does. (d) We have (i„f = i„. Apply (c). (e) Let X afford cp and define 3E* by Xx(h) = X(xhx~l). Check that Xx is a representation which affords cpx. | It follows from Lemma 6.1 that G permutes Irr(H) by g: 9i-> 99. Note that H acts trivially and thus G/H permutes Irr(H).
Normal subgroups 79 (6.2) theorem (Clifford) Let H <i G and let / e Irr(G). Let 9 be an irreducible constituent of Xh and SUppOSC *? — i?i, i?2 > < • • •> **t are the distinct conjugates of 9 in G. Then Xh = e£9i, i=l where e = [/„, 9]. Proo/ We compute (9G)H. For h e H, we have I«I*eG IWI*eG since xhx~1eH for all x e G. Thus |H|(9G)H = ^£fi 9* and hence if <p e Irr(H) and <p i {9;}, we have 0 = [£ 9*, <p] and therefore [(9G)H, <p] = 0. Since x is a constituent of 9G by Frobenius reciprocity, it follows that [.Xh' 9] — 0- Thus all irreducible constituents of Xh are among the 9f and Xh = p= i [Zh. 9i]9.- However, [*„, 9f] = [&,, 9] by Lemma 6.1(d) and the proof is complete. | Theorem 6.2 is so important that we digress to consider another proof in a more general, module theoretic setting. If X is an F-representation of H <i G for an arbitrary field F, we define the conjugate representation 3E9 for g e G by X9(h) = X(ghg~l). Note that 3E9 is an F-representation which is irreducible iff X is. Also, if 3Et and 3E2 are F-representations of H, then Xtg and 3E2* are similar iff Xt and 3E2 are. (6.3) definition Let H <i G and let Wt and W2 be F[H]-modules. Then Wt is conjugate to W2 if there exist bases for the Wf such that the corresponding F-representations of H are conjugate. (6.4) lemma Let Vbe an F[G]-module and let H <G. Suppose W £ V is an F[/TJ-submodule. (a) If g e G, then Wg is an F[H]-submodule of V and is conjugate to W. (b) If M is an F[H]-module conjugate to W, then M ^ Wg for some #eG. (c) If U s F is an F[H]-submodule isomorphic to W, then Ug = Wg as F[H]-modules. Proof We have (Wg)/i = W(ghg~ v)g =■ Wg since ghg~1 e H. Thus Wg is an F[H]-submodule. Let wl5 w2, ..., w„ be a basis for W. Then wvg, ..., w„g is a basis for Wg. Let 3E and 9) be the F-representations of H corresponding to W and Wg respectively, with respect to these bases. We claim that 9) = 3E».
80 Chapter 6 If X(ghg l) = (ai}), we have w{ghg-1 =Z<*ijWj } and thus (wtg)h = £ aij(Wj9l j Therefore, 9)(/i) = (ay) = Xe(h), establishing the claim. The proof of (a) is now complete. If M is conjugate to W, then M corresponds to the representation X9 for some geG. Since Wg corresponds to the same representation, we have Wg s M, proving (b). Finally, if U = W then with respect to a suitable basis, U corresponds to X. Thus Ug and Wg both correspond to X° and (c) follows. | (6.5) theorem {Clifford) Let H -a G and let 7 be an irreducible F[G]- module. Let W be any irreducible F[/T]-submodule of V. Then (a) 7 = Y,' Wf where the Wt are irreducible F[H]-submodules of V. (b) Each Wj is of the form Wgx for some gteG and thus is conjugate to W. (c) In the notation of Lemma 1.13, nw(V) = nM(V) for every F[H]- module M conjugate to W. Note Another way of saying (c) is that each isomorphism class of F[H]- modules conjugate to W is represented equally often among the Wt. Proof of Theorem 6.5 Clearly, £9eG Wg is G-invariant and hence V= Yw3 geG by the irreducibility of V. The F[H]-modules Wg are conjugate to W and hence are irreducible. By Lemma 1.11, it follows that V is the direct sum of some of the Wg. This proves (a) and (b). Now let M be an F[H]-module conjugate to W. Since dim(M) = dim( W), it follows from Lemma 1.13(b) that it suffices to show that dim(M(V)) = dim(W(V)) in order to prove that nM(V) = nw(V). Since M is conjugate to W, it follows by Lemma 6.4(b) that M = Wg for some geG. Now 6.4(c) yields W(V)g £ M(V) and M(*0g-1 £ W(V). Thus W(V)g = M(V) and the result follows. | (6.6) corollary Let H -a G and let 7 be an irreducible F[G]-module for some field F Then when viewed as an F[H]-module, V is completely reducible. Proof Use Theorems 1.10 and 6.5. |
Normal subgroups 81 Note that Corollary 6.6 is independent of Maschke's theorem and is valid even if char(F) divides \H\. Also note that Theorem 6.2 is a consequence of Theorem 6.5. We now return to the study of C-characters and obtain some consequences of Theorem 6.2. (6.7) corollary Let H <i G and suppose that / e Irr(G) and [/H, 1H] ^ 0. Then H £ ker /. Proof We have Xh = e £ 9h where 9t = 1H and the 9f are conjugate. Since (\Hf = 1H for all g e G, we have Xh = z(l)l« and H £ ker /. | Corollary 6.7 can also be obtained as a consequence of Lemma 5.11 (see Problem 5.11) since /is a constituent of (1H)G and ker(lH)G = f]HB = H. An interesting class of problems in character theory arises from considering how the structure of a group G and the set {/(1) | / e Irr(G)} are related. We give one result of this type as an application of Theorem 6.2. Many more results like this will be found in Chapter 12. (6.8) lemma Let H <i G and suppose x £ Irr(G) and 9 e Irr(H) with [XH,9]*0.Then9(l)|x(l). Proof Since 99(1) = 9(1), we conclude that /(l) = et9(l) where Xh = e Y!i= i fy as in Theorem 6.2 with 3 = 3^ I (6.9) theorem Suppose x(l) is a power of the prime p for every / e Irr(G). Then G has a normal abelian p-complement. Proof If N <i G, then by Lemma 6.8,9( 1) is a power of p for all 9 e Irr(N). Working by induction on | G |, we see that it suffices to find a normal subgroup N of index p, since the normal abelian p-complement for N will be one forG. We have \G\ = \G:G'\+ I /(I)2. xeIrr(G):;(<l)>l If G is abelian, the result is trivial. Otherwise, some / e Irr(G) is nonlinear and thus p 11G |. Since the last sum is divisible by p, we see that p 11G: G' |. It follows that G has a normal subgroup of index p and the proof is complete. | The converse of Theorem 6.9 is true and will be proved later in this chapter.
82 Chapter 6 (6.10) definition Let H <i G and let 9 e Irr(H). Then 7G(9)={0eG|9» = 9} is the inertia group of 9 in G. Since 7G(9) is the stabilizer of 9 in the action of G on Irr(H), it follows that it is a subgroup and that/G(9) 2 77. Also | G : 7G(9) | is the size of the orbit of 9 and so in the formula i=l of Theorem 6.2, we have t = \G: /G(9)|. In particular, t divides \G:H\. It turns out that e divides \G:H\ also, but that is much more difficult to show. We shall prove some special cases in this chapter and the general result will be proved in Chapter 11, using projective representations. The following result is of fundamental importance in the character theory of normal subgroups. (6.11) theorem Let H -a G, 9 e Irr(H), and T = 7G(9). Let s/ = Welrrmirjfo, 9] * 0}, £ = {/e Irr(G)|[/H, 9] * 0}. Then (a) If \jiestf, then i/fG is irreducible; (b) The map \j/\-npG is a. bijection of si onto 3$\ (c) If t/fG = /, with xjiestf, then i/f is the unique irreducible constituent of Xt which lies in si; (d) If >AG = x, with tf, e.«/, then [>„, 9] = [zfl, 9]. Proof Let \fiesi and choose any irreducible constituent / of \jiG. Then \ji is a constituent of /T and since 9 is a constituent of \pH, we conclude / e 34. Let 9 = 9t, 92 9, be the distinct G-conjugates of 9. Then t = | G: T| and xH = e £',_ i fy f°r some integer e. Since 9 is invariant in T, we conclude from Theorem 6.2 that \pH = /9 for some/. Since \p is a constituent of xT, we have/ < e. Therefore et9(l) = z(l) < tfrG(l) = ^(1) = /t9(l) < et9(l) and hence equality holds throughout. In particular, x(l) = \jiG{\) and we conclude that / = i/fG so that (a) follows. Also [Zh,9] = <> = / = [>„, 9]
Normal subgroups 83 and (d) is proved. Statement (c) follows from the last equality since if i/f t e j/, \pi 76 \p, and \pt is a constituent of xT, then a contradiction. The map in (b) is well defined by (a) and its image lies in Si by (d). It is one- to-one by (c). To prove that it maps onto SH, let / e 38. Since 9 is a constituent of Xh , there must be some irreducible constituent i/f of %T with [i/fH, 9] ^ 0. Thus i(/6j/ and x is a constituent of \\iG. Therefore / = \\iG and the proof is complete. | (6.12) corollary Let /elrr(G) be primitive. Then for every Af<G, XN is a multiple of an irreducible character of N. Proof We have %N = e£j=1 9j, where the #,- are distinct and t = |G:/G($i)|. By Theorem 6.11, i = i>° for some i/felnt/o^)). Since / is primitive, it follows that /G(#i) = Gandr = l.Thusx,v = e9t and the proof is complete. | The converse of Corollary 6.12 is false. For instance the irreducible character of A5 of degree 5 is imprimitive and yet the conclusion of the corollary is (trivially) valid for this character. Characters which are multiples of an irreducible are called homogeneous characters. Irreducible characters whose restriction to every normal subgroup is homogeneous are often called quasi-primitive characters. (6.13) corollary Suppose G has a faithful primitive character and let A <i G be abelian. Then A £ Z(G). Proof Let / e Irr(G) be primitive and faithful. Then Xa = ek for some linear character k. Thus A £ Z(/) = Z(G). | (6.14) corollary Every nilpotent group is an M-group. Proof Let G be nilpotent and let / e Irr(G). Choose H £ G minimal, such that there exists \p e Irr(H), with / = \pa. By transitivity of induction (Problem 5.1), ij/ is a primitive character of H. It follows that \p is a faithful primitive character of H/ker(i/f) = H and thus every normal abelian subgroup of H is central. Since every nilpotent group contains a normal self-centralizing subgroup, we conclude that H is abelian. Then \p is linear and the proof is complete. I We shall soon give a much more general sufficient condition for a group to be an M-group. The preceding result is included here simply to illustrate the usefulness of Theorem 6.11.
84 Chapter 6 (6.15) theorem (ltd) Let A o G be abelian. Then x(l) dividies | G: A \ for all x e Irr(G). Proof Let A e Irr(A), with [ja, ^] # 0, and let T = 7G(1). Then for some t^elrr(T), we have x = i>G and \jiA = el. Thus A £ Z(i/f) and hence t/f(l) divides | T: A | by Theorem 3.12. Since /(1) = | G: T | i/f( 1), we conclude that X( 1) divides | G: A | and the proof is complete. | Note that the converse of Theorem 6.9 is immediate from Theorem 6.15. Therefore, the purely group theoretic condition that G has an abelian normal p-complement is precisely equivalent to the condition that /(1) is a power of p for all x £ Irr(G). Suppose N <i G and 9e Irr(N). Write 9G = £ e^,- for x,e Irr(G) and e( > 0. Then (x^ = et £}=1 9Jf where 9 = 9t 9, are the distinct conjugates of 9 in G. Let T = 7G(9). By Theorem 6.11 it follows that 9T = £ e,i/fj with t/f j e Irr(T) and i/^6 = /;. It follows that for the purpose of investigating the nature of the integers eh it suffices to assume T = G, that is, that 9 is invariant in G. For the remainder of this discussion we assume that 9 is invariant. We have 9G = £ elXl and (Z|)K = Cj9. Therefore, Xj(1) = e,-9(l) and \G: N\8(l) = 9G(1) = 5>jZj(l) = J e,29(l), so that $>.2 = |G:/V|. It was remarked earlier (without proof) that ef divides \G: N\. (Note that if N is abelian then ef = x,-(l) and in this case we know that et\\G: N\ by Theorem 6.15.) We see that in some respects, the integers et behave like character degrees for G/N. If they were really character degrees, then one of them, say elt would equal 1. That would mean that (Xi)N = 9. In other words, that 9 is extendible to G. (Note that if 9 is extendible then it is automatically invariant.) A consequence of the next result is that if some ex = 1, then the eK are exactly the degrees of the irreducible characters of G/N. (6.16) theorem Let N <i G and let <p, 9 e Irr(N) be invariant in G. Assume cpS is irreducible and that 9 extends to / e Irr(G). Let Sf = {P e Irr(G) | [<pG, ff] * 0} and 3T = {if, e Irr(G) | [(<p9)G, £| * 0}. Then 01-» fo defines a bijection of y onto ^".
Normal subgroups 85 Proof We have (cpG)N is a multiple of (p and hence ((pG)N = \ G: N \ cp by comparing degrees. Thus \_<pG, <pG] = \_q>, ((pG)N~\ = \G:N\\_<p,<p\ = \G:N\. Similarly, [(<p9)G, (<p9)G] = | G: A/1. Also, by Problem 5.3, (<p9)G = (pGx- Now write <pG = £,eJ, e,/J. We have |>G, <pG] = |G: N| = |>Gz, <pG/] and hence Since [fo, y/] ^ 0 and [fix, Pxi ^ 1 and all e0 > 0, this forces [fix, YX] = 0 if fi t6 y and [/?/, /?/] = 1. Thus the fix are distinct irreducible characters for distinct fi. These are all of the irreducible constituents of q>Gx = (<p9)G. The proof is complete. | What is probably the most important special case of Theorem 6.16 is when (p = In . (6.17) corollary (Gallagher) Let N o G and let / e Irr(G) be such that Xn = 9 e Irr(N). Then the characters fix for /? e lrr(G/N) are irreducible, distinct for distinct fi and are all of the irreducible constituents of 9G. Proof This is exactly Theorem 6.16 in the case (p = lN. | Note that in Corollary 6.17 we have identified Irr(G/A0 with a subset of Irr(G). In this situation, we see that the ex are exactly the fi(\) for fi e lrr(G/N). There are other situations where we have control of the ef's. The following "going down" theorem is useful in studying the characters of solvable groups. (6.18) theorem Let X/L be an abelian chief factor of G. (That is, X,L<i G and noM<G exists with L< M < K.) Suppose 9 e Irr(X) is invariant in G. Then one of the following holds: (a) 9LeIrr(L); (b) 9L = ecp for some (p e Irr(L) and e2 = |X: L\; (c) 9L = Yj= i <Pi where the <pte Irr(L) are distinct and t = \K\L\. Proof Let cp be an irreducible constituent of 9L and let T = IG((p)- Since 9 is invariant in G, every G-conjugate of q> is a constituent of 9L and hence is X-conjugate to <p. It follows that \G:T\ = \K:K r\T\ and hence KT = G. Since X/L is abelian, K r\ T o KT = G and thus either K r\ T = K or K r\T = L. If X n T = L, then 9L = e £j=! <?;, where t = |X: L|, <p! = <p, and the <p{ are distinct. Thus 9(1) = e\K: L|<p(l). Since 9 is a constituent of <pK, we have 9(1) < |X: L|<p(l) and therefore e = 1. This is situation (c).
86 Chapter 6 Now assume K r\ T = K so that <p is invariant in K and 9L = e<p for some e. Let 1 e Irr(X/L). Since 1 is linear, 19 e Irr(X). Also (19)L = 9L = e<p. Suppose that all of the characters 19 are distinct as 1 runs over Irr(X/L). Each of these \K: L\ characters is an irreducible constituent of cpK with multiplicity e and we have e|K:L|9(l)<«/(l) = |K:L|<p(l). Therefore, e>(l) = eS(l) <: <p(l) and e = 1. This is situation (a). In the remaining case, 19 = /z9 for some X, fieIrr(X/L) with 1 j= fi. Let 1/ = ker(l/I). Then L ^ U < K and 9 vanishes on K — U. Since 9 is invariant in G, it follows that 9 vanishes on K — Ua for all g e G. Since D»eG Ue = L, we conclude that 9 vanishes on X — L. By Lemma 2.29 we have |X:L| = |X:L|[9,9] = [9L)9L] = e2 and the proof is complete. | (6.19) corollary Let N<iG with \G:N\=p, a prime. Suppose 1 e Irr(G). Then either (a) Xn is irreducible or (b) Xn = E?= i &i> where the 9f are distinct and irreducible. Proof TakeK = GandL = Nand apply Theorem 6.18. Case (b) of that theorem cannot occur since p is not a square. | (6.20) corollary Let N<iG and suppose \G:N\=p, a prime. Let 9 e Irr(iV) be invariant in G. Then 9 is extendible to G. Proof Let / be an irreducible constituent of 9G. Then Z\ = e9 for some e. Comparison with Corollary 6.19 yields e = 1. | Actually, Corollary 6.20 would still be true if the hypothesis that | G: N \ is prime were weakened to G/N cyclic. This stronger result will be proved in Chapter 11. (6.21) definition Let N o G and let / e Irr(G). Then / is a relative M- character with respect to N if there exists H with N s H s G and t/f e Irr(H) such that iiG = x and iAjv e Irr(iV). If every / e Irr(G) is a relative M-character with respect to N, then G is a relative M-group with respect to N. Note that / e Irr(G) is a relative M-character with respect to 1 iff it is a monomial character, and G is a relative M-group with respect to 1 iff it is an
Normal subgroups 87 M-group. Also, it is clear that if G is a relative M-group with respect to JV, then G/N is an M-group. The converse of the last statement is false. If G = SL(2,3) and JV = Z(G), then G/N = A 4, which is an M-group. However, G is not a relative M-group with respect to JV since G has an irreducible character of degree 2 and has no subgroup of index 2 as would be required by the definition. (6.22) theorem Suppose JV <a G and G/N is solvable. Suppose, furthermore, that every chief factor of every subgroup of G/N has nonsquare order. Then G is a relative M-group with respect to JV. Note The hypotheses on G/N, above, are automatically satisfied if G/N is nilpotent or supersolvable since then all chief factors of subgroups have prime order. Proof of Theorem 6.22 Let / e Irr(G). We must show that / is a relative M-character with respect to JV. If %N is irreducible, this is clearly the case and hence we assume Xn reduces. Let K -o G be minimal such that K 2 JV and Xk is irreducible. Then K > JV and we may choose L <i G, L 2 JV such that K/L is a chief factor of G. In particular, K/L is abelian of nonsquare order. We apply Theorem 6.18 to the chief factor K/L and the G-invariant character xKeIrr(X). Since xl is reducible and \K:L\ is nonsquare, we conclude that Xl = S=i •?■> wnere trie <?;eIrr(L) are distinct and t = \K:L\ > 1. Let T = IG((pi) 2 L 2 JV. By Theorem 6.11, we conclude that x = 4>G for some \ji e Irr(T). Since T/N < G/N, we may apply induction on | G: JV| to conclude that T is a relative M-group with respect to JV and that \ji = 9-T for some 9 e Irr(H) where JV s H s T and SN e Irr(JV). Now x = ^° = (8Tf = 9G by Problem 5.1 and the proof is complete. | Note that Corollary 6.14 follows immediately from this result as does the somewhat more general fact that all supersolvable groups are M-groups. We prove a still more general sufficient condition. (6.23) theorem Let JV o G and suppose that all Sylow subgroups of JV are abelian. Assume that G is solvable and is a relative M-group with respect to JV. Then G is an M-group. Proof Let / e Irr(G). Since / is a relative M-character with respect to JV, choose HsG with JV s H, \\i e Irr(H), \j/N e Irr(JV) and \pa = x- Now choose U s H, minimal such that there exists 9 e lrr(U), with 9H = \p. By Problem 5.1, 9G = (9G)H = \jiG = x and it suffices to prove that 9 is linear. Let M = U n,N. Since (&NV)tl = \p and \jiNV is irreducible, it follows that \jiNV = $NV and thus (Sw% = ^ e Irr(JV). By Problem 5.2, ($M)N = (&NV)N e Irr(JV) and hence 9M e Irr(M).
88 Chapter 6 By the minimality of U and Problem 5.1, 9 is a primitive character of U. Let K = ker_9, V =_U/K and M = MK/K. Then 9 is a faithful primitive character of U and M has all of its Sylow subgroups abelian. Now let Z = Z(M) <a L7. If Z < M, let AjZ be a chief factor of {J with A <= M. Then X/Z is a p-group for some prime p. Let P e Sylp(y4). Then P is abelian and A = PZ, so that A is abelian. Since A <i I/, we conclude from Corollary 6.13 that A s Z(E7). This contradicts A > Z. We conclude that Z = M, and hence M is abelian. Since 9M is irreducible, so is 9MK. Thus 9MK e Irr(M) and hence 9(1) = 1. The proof is complete. | We introduce some notation. If / is a character of G, let det / = 1 be the uniquely defined linear character of Problem 2.3. Now write o(%) = o(X), the order of X as an element of the group of linear characters of G. (We call o(%) the determinantal order of /.) Since \G: ker X\ = o(X), it follows that o(%) divides \G\. Now let JV o G and suppose 9 e Irr(iV) is invariant in G. We wish to find sufficient conditions that 9 be extendible to G. If (| G: N |, 9(1)) = 1, then 9 is extendible iff det 9 is extendible. We shall prove this now under the additional hypothesis that G/N is solvable and defer the general proof to Chapter 8. As will be seen, there is a gain in replacing the problem of extending 9 by that of extending the linear character, det 9. (6.24) lemma Let N <i G and suppose 9 e Irr(iV) is extendible to G and that (| G: N \, 9( 1)) = 1. Let X = det 9 and let fi be an extension of X to G. Then there exists a unique extension / of 9 to G such that det / = fi. Proof Let >/ be an extension of 9 to G and let v = det r\ so that vN = X. Let fiv = a. so that txN = lN and hence a|G:iV| = 1G. Write/ = 9(1) so that det(afc>/) = abf det rj = a.bfv for beZ. Since (/, \G: N\) = 1, we can choose b e Z such that ft/ = 1 mod|G:iV| and thus a?f = a. Let / = a.brj. Then det i = <xbfv = av = y.. Since <xN = \N, we have /^ = >/w = 9 as desired. To prove uniqueness, suppose %0 is an extension of 9 with det(/0) = fi. By Corollary 6.17, we have %0 = //? for some fi e lrr(G/N). Since /0(1) = 9(1) = /(l), we have 0(1)= land fi = det(z0) = p det x = PfH and/?' = lG.Since(/, |G:iV|) = 1, this forces /? = lGand/0 = /.The proof is now complete. | (6.25) theorem Let N <a G and suppose G/N is solvable. Let 9 e Irr(iV) be invariant in G and suppose (9(1), | G: N\) = 1. Then 9 is extendible to G iff det 9 is extendible to G.
Normal subgroups 89 Proof If 9 is extendible, then obviously so is det 9. We prove the converse. Let y. be an extension of 1 = det 9 to G. We work by induction on | G: JV|. (We may assume G > JV.) Let G0 be a maximal normal subgroup of G with G0 ^ JV. Since X extends to /zGo, the inductive hypothesis guarantees that 9 is extendible to G0. By Lemma 6.24, there is a unique extension, Xo e Irr(G0) with det(x„) = Hg0- We claim Xo is invariant in G. For g e G, we have fo0)% = 99 = 9 and det((x0)») = det(x0)9 = 0zG/ = /zGo. The uniqueness of /„ now yields (xoY = Xo and establishes the claim. Since G/N is solvable, | G: G01 is prime and by Corollary 6.20, Xo is extendible to G. The proof is now complete. | We now discuss some sufficient conditions for extending linear characters. A version of the following theorem is true without the assumption of linearity. The general result will be derived from this special case in Chapter 11. (6.26) theorem Let JV<G and suppose X is a linear character of JV which is invariant in G. For each prime p | o(l), choose Hp £ G with Hp/N e Sylp(G/JV), and assume that X is extendible to Hp. Then X is extendible toG. Note If p Jf | G: JV |, then Hp = N and 1 is automatically extendible to Hp. It is only necessary, therefore, to check the hypothesis for primes dividing |G:JV|. Proof Let m = o(X). For each p | m, we may choose lp, a power of 1, such that X = Y\ Xp and o(Xp) is a power of p. We shall show that Xp is extendible to fip e Irr(G). Then fi = J~[ fip is an extension of 1. Since lp is a power of 1, which is extendible to Hpi it follows that Xp is extendible to Hp. Also lp is invariant in G. We now see that it is no loss to assume that m is a power of p. Let v be an extension of X to Hp. Since p Jf\G: Hp\ and v(l) = 1, it follows that p/fv^l) and hence there exists an irreducible constituent / of vG with p^(l).Letz(l)=/. We have \_xH , v] ^ 0 and hence \_xN, X~\ ^ 0. Since X is invariant in G, we conclude that Xn = /^ and thus (det x)w = J/. Let (5 = det /. Since p Jf f and m is a power of p, we may choose b e Z with/ft = 1 mod m. Then ((5% = Xfb = X and the proof is complete. | (6.27) corollary Let JV <i G and suppose X is a linear character of JV which is invariant in G. Assume (|G: JV|, o(l)) = 1. Then X has a unique extension /z to G with the property that (| G: JV |, o(fi)) = 1. In fact, o{y) = o(l).
90 Chapter 6 Proof The existence of an extension is immediate from Theorem 6.26 since the hypotheses are trivially satisfied. Let v be an extension of X and choose beZ, with b\G:N\ = 1 mod (o(l)). Let fi = vfc|G:JV|. Then fxN = Xb\G-.N\ = x In particuiar) 0^ > 0(X). On the other hand, (voW)N = lN, so that v0<A) e Irr(G/JV) and n«W _ ivoWbyG:N\ _ j Therefore, o(y) = o(X). For uniqueness, suppose t is an extension of X with (o(t), \G: N\) = 1. Then (o(/zf), \ G: N |) = 1 and (fit )N = lN so that /zf e Irr( G/JV). It follows that \ix = 1G and n = t as desired. | (6.28) corollary Let JV <i G with G/N solvable and suppose 9 e Irr(iV) is invariant in G. Assume (|G: JV|, o(9)9(1)) = 1. Then 9 has a unique extension, x g Irr(G) with (| G: JV|, o(z)) = 1. In fact o(%) = o(9). Proof By Corollary 6.27, let y. be the unique extension of X = det 9 to G with (|G: JV|, o(/z)) = 1. By Theorem 6.25 and Lemma 6.24, let x be the unique extension of 9 to G with det x = V- Then o(x) = o[/j.) = o(X) = o(9). If Xo extends 9 and (| G: JV |, o(x0)) = 1, let /z0 = det(z0). Thus (|G:JV|,o(/z0))=l and hence /z0 = fi by the uniqueness of /z. Then z0 = z by the uniqueness of /. This completes the proof. | Note that since o(9) and 9(1) divide |JV|, the condition that (|G:JV|, o(9)9(l))= 1 in Corollary 6.28 will be automatic if (|G: JV|, |JV|) = 1. The hypothesis of solvability in Corollary 6.28 will be removed in Chapter 8. We shall give one further extendibility criterion now. It can be used when G/N is a p-group for p^2. Unlike the previous results, this condition is independent of character degree. (6.29) definition Let x be a character of G and let p be a prime. Then x is p-rational if there exists an integer r with pjfr, such that x(g) £ Q[e2"'/r] for every g e G. We use the notation Qk = Q[e2"i/k] for 1 < keZ. As is well known, Qk n Q, = Qd, where d = (k, I). By Lemma 2.15, it follows that /(#) e Q|G| for every character x of G and every g e G. Write \G\ = n = p"m, with p Jfm. It follows that x is p-rational iff all of its values lie in Qm. Let yp(G) denote the Galois group 9(QJQm). If / is a character of G, let x" be defined by x"ig) = lig)" for a e 9(QJQ). Then / is p-rational iff/" = X for all a e ^p(g) [Note that by Problem 2.2, /" is necessarily a character. In particular, @P(G) permutes Irr(G).]
Normal subgroups 91 From Galois theory we have 9(QJQm) = 9(Qpa/Q), where as above n = mp" and p Jf m. This isomorphism is the restriction map. It follows if p 76 2 and a > 0, that e2nilp is not invariant under 9(QJQm) and hence if p#2 and X is a linear character of G with p | o(l), then 1 is not p-rational. Also, when p 7^ 2, we have <S^G) s 0(OP„/O) is cyclic. We have one more general remark. If H £ G, then Q|H| £ Q|G| and so x" is defined for characters / of H with a e ^P(G). It follows that / is p-rational ifiFz" = z for all (xe^G). (6.30) theorem Let JV <G with G/JV a p-group, p#2. Suppose 9 e Irr(JV) is invariant in G and p-rational. Then 9G has a unique p-rational irreducible constituent /. Furthermore, Xn = & Proof Use induction on \G:N\. We suppose |G:JV| > 1. Let K<G with JV £ X and |K : JV| = p. By Corollary 6.20, 9 can be extended to rj e Irr(X). Since p ¥= 2, ^P(K) is cyclic and we choose a generator er. Now >/" e lrr(X) and (>/% = 9" = 9 since 9 is p-rational. By Corollary 6.17, we have rf = rjX for some (linear) X e lrr(K/N). If X = lK, then rf = r\ and >/ is p-rational. Suppose that this is not the case. Then p = o(X) and since p ¥= 2v/e have X" ^ X and so 1" = lm for some m e Z, with m # 1 mod p. Choose b e Z with (1 — m)ft = 1 mod p and let ^ = X\ Since AN = ljv, we have \pN = rjN = 9. Also, mb + 1 = b mod p and thus ji' = (A)" = Xmb+lti = Xbrj = \j/ and \j/ is p-rational. Thus in any case, 9 has a p-rational extension, \p e Irr(X). If <p e Irr(X) is any p-rational extension of 9, then cp = ip/j. for some unique fi e lrr(K/N) by Corollary 6.17. We have cp = cp" =,(^n)a = ^"n" = #" and thus n" = fi. Since p # 2, it follows that p Jf o(n) and thus fi = lK. Therefore <p = \p. Now let g e G. Then W = (\p"Y = \pe and 9» = 9 so that \pe is a p- rational extension of 9 to K and by the preceding paragraph, we have \jig = ip and \ji is invariant in G. Since \G:K\ < \G: JV|, the inductive hypothesis guarantees that ipG has a unique p-rational irreducible constituent / and that Xk = "A s° that Xn = & Let x0 be any p-rational irreducible constituent of 9G. We show Xo = X- Let cp be an irreducible constituent of (x0)k- Since (x0)N is a multiple of 9, it follows that cpN = 9and<p = t/n> for some v e Irr(X/JV). Since K/N £ Z(G/JV), we have v° = vforgeG and thus cpe = \peve = \pv = cp and cp is invariant in G. It follows that (xo)k = e<P for some integer e and thus cp is p-rational since cp(k) = (l/e)x0(k) for keK. Therefore cp = \ji and Xo is a constituent of i/fG. It follows that Xo = X and the proof is complete. |
92 Chapter 6 We can use some of our results on character extendibility to prove Tate's theorem. This result, which was originally proved in an entirely different way, serves as a "booster" for transfer theory, as will be explained. We define Op(G) to be the unique minimal normal subgroup of G such that G/Op(G) is a p-group (for the fixed prime p). Similarly, let AP(G) be the unique minimal normal subgroup of G such that G/AP(G) is an abelian p- group.[Note that AP(G) = G'Op(G).] Let P e Sylp(G) and let JV 2 P. Then JV is said to control p-transfer if JV/AP(JV) s G/AP(G), or equivalently, JV n AP(G) = AP(JV). Several of the standard transfer theorems assert that under suitable hypotheses, certain subgroups JV control p-transfer. [Usually, JV = N(W) for some subgroup, W <a P.] Tate's theorem guarantees that whenever JV/AP(JV) ^ G/AP(G), then also JV/Op(JV) z G/Op(G). (6.31) theorem (Tate) Let PeSylp(G) and JV 2 P. Suppose that JV n AP(G) = Ap(JV).ThenJV n Op(G) = Op(JV). Proof (Thompson) Since JV 2 P, we have JVOp(G) = G. Let U = N r\ Op(G) so that U -a JV and N/U S G/Op(G) is a p-group. Thus U 2 Op(JV). Assume U > Op(N) and choose V <a JV, 7 2 Op(JV) such that t//7 is a chief factor of JV. Let 1 be a nonprincipal linear character of U/V. Since U/V is a chief factor of the p-group JV/Op(JV), we conclude that [//Ks Z(JV/F) and hence 1 is invariant in JV. Let 9 = kOP(G). If/ is an irreducible constituent of 9 and x e JV, then since A* = A, we have [X, 9] = [Xv. *] = l(Xx)u, X*] - [_(xx)v, X] = If, 8]. Since P £ JV, it follows that we may write A as A runs over the sums of the orbits of Irr(Op(G)) under the action of P. Now 9(1) = \Op(G):U\ = | G: JV | is prime to p. It follows that for some A, we have p Jfa^ and p Jf A(l). Write A = £^0 / where 0 is an orbit. Since /(l)
Normal subgroups 93 is constant for / e 6, we conclude that pjf\&\. Since 0 is an orbit under the p-group P, we have 10 \ = 1 and A = /, where / e Irr(Op(G)), / is invariant underP,p^x(l)andp^[9,x]. Since POp(G) = G and / is invariant under P, it follows that / is invariant in G. We claim that / is extendible to G. By Corollary 6.28, it suffices to show that p X Z(l)o(z)- Now Op(G) has a normal subgroup, ker(det /) = K, such that |Op(G):K| = o(x) and Op(G)/K is abelian. Since Op(Op(G)) = Op(G), it follows that Op(G) has no nontrivial p-factor group and thus p Jf o(x). Since p X Z(l)» Corollary 6.28 does apply. Let i/f be an extension of / to G and write <pelrr(N) We have E K&v, W = D/V i] = Lxv, i~] = Ez, 9], which is prime to p. We may therefore choose q> e lrr(N) with p ^ [pj,, 1]. In particular, \_(pv, l~] =£ 0 and since 1 e Irr(t/) is invariant in JV, we conclude that <pD = el and thus <p(\) = e = \_(pv, 1] is prime to p. Since V £ ker I and <pj, = el, we conclude that 7 £ ker <p and <p e lrr(N/V). Since JV/7 is a p-group, <p(l) is a power of p and hence <p(l) = 1. Now JV/ker <p is abelian and thus AP(JV) £ ker q>. However, U = N r\ Op(G) £ JV r\ AP(G) = AP(JV) and therefore 1/ £ ker (p. Since <?„ = el, it follows that I = \v. This contradicts the choice of I and completes the proof of the theorem. | We discuss one further topic in this chapter. Let JV <i G so that G permutes Irr(JV). It is also true that G permutes the set of conjugacy classes of JV and it is natural to consider the relationship between these two actions. It is not the case that they are necessarily permutation isomorphic although they are closely related. (6.32) theorem (Brauer) Let A be a group which acts on Irr(G) and on the set of conjugacy classes of G. Assume that %(g) = x'ig") for all / e Irr(G), aeA and geG; where g" is an element of Cl(g)a. Then for each ae A, the number of fixed irreducible characters of G is equal to the number of fixed classes. Proof Let /, and Jf; be the irreducible characters and conjugacy classes of G for 1 < i, j < k. Choose g} e Jf 3 and write g" = gs if Jf," = Jts. Let X = (Xiigj)), the character table of G, viewed as a matrix.
94 Chapter 6 For as A, let P(a) = (py), where py = 0 unless Xi" = Xj> in which case ptJ = 1. Similarly, define Q(a) = (<jy), where qtJ = 1 if Jf," = Jf j and is zero otherwise. The (w, v) entry of the matrix P(a)X is since only when /,■ = Xu is pui j= 0. Similarly, the (u, v) entry of the matrix XQ(a) is Yxuidj)Qjv = Xuidav~') j since only when gj = g"„~l is <?;„ ,£ 0. The hypothesis of the theorem now implies that P(a)X = XQ(a). Thus Q(a) = X~lP(a)X since the orthogonality relations guarantee that X is nonsingular. We conclude that tr P(a) = tr Q(a). Since tr P(a) is the number of / e Irr(G) which a fixes and tr Q(a) is the number of a-fixed conjugacy classes, the proof is complete. | (6.33) corollary Under the hypotheses of Theorem 6.32 the numbers of orbits in the actions of A on the irreducible characters and conjugacy classes of G are equal. Proof Apply Corollary 5.15 to the result of Theorem 6.32. | We may apply Theorem 6.32 to obtain information about the characters of "Frobenius groups." (6.34) theorem Let N <i G and assume that CG(x) £ N for every 1 t6 x e N. Then (a) For cp e Irr(iV), with cp ^ \N, we have IG((p) = N and <pG e Irr(G). (b) For x e Irr(G) with N £ ker /, we have x = 9° f°r some q> e Irr(iV). Proof Let <peIrr(iV), ^ 1^. To show that <pGeIrr(G), it suffices by Theorem 6.11 to show that IG((p) = N. In order to prove (a), therefore, it suffices to show that no element g e G — N can normalize any nontrivial conjugacy class of N and then apply Theorem 6.32. Suppose then, geG — N normalizes Jf, a class of N. Let x e Jf. Then xaeJf and thus xa = x" for some neN. Therefore gn~leC(x). Since gn~l $N and xeN, the hypothesis yields x = 1 and thus Jf = {1}. This proves (a). Now let x £ Irr(G) with N £ ker /. Choose an irreducible constituent cp ofxN with cp =£ lN. Then x is a constituent of <pN which is irreducible and so cpN = x- The proof is now complete. |
Problems 95 It turns out that the groups G satisfying the hypotheses of Theorem 6.34 and for which N < G are exactly the "Frobenius groups" which are discussed at some length in the next chapter. Problems (6.1) Let JV -a G and 9 e Irr(JV). Show that 9G e Irr(G) iff 7G(9) = JV. (6.2) Let N <i G and suppose G/N is abelian. Let C be the group of linear characters of G/N so that C acts on Irr(G) by multiplication. (See Problem 2.6.) Let 9 e Irr(JV). Show that where/is an integer and the & e Irr(G) constitute an orbit under C. Hint Let / e Irr(G). Then (%N)G = p%, where p is the regular character of G/N. (6.3) Let N -a G and let / e Irr(G) and 9 e Irr( JV) with \_%N, 9] ± 0. Show that the following are equivalent: (a) Z„ = e9, withe2 = \G:N\; (b) x vanishes on G — JV and 9 is invariant in G; (c) x is the unique irreducible constituent of 9G and 9 is invariant in G. JVote In the situation of this problem, we say that / and 9 are fully ramified with respect to G/N. (6.4) Define Jf(G) = {H £ G | cpa e Irr(G) for some cp e Irr(H)}. Let H(G) = C\Jf(G). Show that if G is an M-group, then H(G) is abelian. JVote By Problem 5.12, Z(G) £ H(G) for all groups, G. Of course, H(G) is characteristic in G. (6.5) Let H(G) be as in the preceding problem. Show that H(G) centralizes N/N' for every JV <i G and conclude that if G is solvable, then H(G) is nil- potent. Note In fact we may conclude that if G is solvable, then H(G)' £ Z(H(G)). This is so because a group which is nilpotent of class > 2 necessarily has a characteristic abelian noncentral subgroup; namely, the next to the last nontrivial term in its lower central series. Hint Ucpe Irr(JV) with N o G, then H(G) £ IG(cp). Use this to show that [JV, H(G)] £ JV'.
96 Chapter 6 (6.6) Let G be solvable and assume that every / e Irr(G) is quasi-primitive. Show that G is abelian. (6.7) Let JV <G and assume that G/N is solvable. Let / e Irr(G) and 9 e Irr(JV), with [*„, 8] * 0. Show that x( 1 )M 1) divides \G:N\. Note The conclusion of this problem is valid even if G/N is not solvable. (6.8) Suppose that G has exactly one nonlinear irreducible character. Show that G' is an elementary abelian p-group. (6.9) (Dornhoff) Let G be an M-group and suppose JV -o G with (|JV|,|G:JV|) = 1. Show that JV is an M-group. Hint Let 9eIrr(JV). Find H £ G and lelrr(H), 1(1) = 1 such that [(!*%, 9] * 0 and XG e Irr(G). Use Problem 6.7 to show that lra(l) = 9(1) and then use Problem 5.8. (6.10) Let JV«a G with (|G: JV|, |JV|) = 1. Suppose that every subgroup of G/N is an M-group. Show that G is a relative M-group with respect to JV. (6.11) Let A o G with A abelian. Suppose / e Irr(G) is a monomial character. Show that / is a relative M-character with respect to A. Hint Write / = 1G, with k a linear character of H <=, G. Then every irreducible constituent of (kAnll)A has multiplicity 1. Note Since Problem 6.11 is false if A is not abelian, it does not follow (and is not true) that an M-group is necessarily a relative M-group with respect to every normal subgroup. (6.12) Let K/L be an abelian chief factor of G. Let <p e Irr(L) and T = IG((p). Assume that KT = G. Show that one of the following occurs: (a) <pKelrr(K); (b) <pK = e9 for some 9 eIrr(X), where e2 = \K:L\. (c) cpK = Y!t=i &h where the 9;eIrr(K) are distinct and t = \K:L\. Hint Use Problem 6.2 and the ideas in the proof of Theorem 6.18. (6.13) Show that the number of real classes of a group G is equal to the number of real valued / e Irr(G). Note It is easy to see that if | G | is odd, then 1 is the only real element. It follows that Problem 6.13 generalizes Problem 3.16. (6.14) Let F be a field with Q £ F £ C. Say that g e G is an F-element if X(g) e F for every / e Irr(G). Let G be a p-group with p^2 and show that the
Problems 97 number of classes of F-elements of G is equal to the number of / e Irr(G) with values in F. Hint Let e be a primitive | G | th root of unity. Let ^ be the Galois group of F[e] over F. Define actions of <& on Irr(G) and on the set of classes of G. Note Problem 6.14 is false without some assumption on G. Counterexamples with G a 2-group and G of odd order have been found by Thompson and Dade respectively. (6.15) Let F be an arbitrary field and let N-a G. View F[iV] £ f[G]. Show that J(FrjV]) £ J(F[G]). (See Problem 1.4.) (6.16) Let P be a p-group and Q a <j-group, where p and q are distinct odd primes and P £ Aut(Q). Show that \P\ < \\ Q| by carrying out the following steps. (a) It suffices to assume that Q is elementary abelian, that is, is an F\_P]- module with F = GF(q). (b) It suffices to assume that Q is an irreducible F[P]-module. (c) We may assume that P is not cyclic, so that by theorems on p-groups, it follows that there exists A <a P with A elementary abelian of order p2. (d) Q is imprimitive as an F[P]-module and we may choose H £ P, \P:H\ = p, with Q = £ • f= i Wf, each W{an F[H]-module with all | Wt\ equal. (e) Let N{ be the kernel of the action of Hon Wt. Then iPi<Pnitf--^-i- (f) Complete the proof by induction. Hint Clifford's theorem (6.5) is used in (d). Show that F[A] cannot have a faithful irreducible module. Notes The result of Problem 6.16 may be stated as a theorem in arithmetic. If Q is elementary abelian of order q", then | Aut(Q) | = f] "r/ (qn — q') and hence this number is not divisible by p" if p" > \q". The induction in the previous step (f) would not go through to prove the weaker theorem that | P | < \Q\. (6.17) Let N o G with G/N cyclic. Let 9 e Irr(iV) be invariant in G and assume that (9{l),\H:N\)= 1. Show that 9 is extendible to G. Hint Show that G/ker(det 9) is abelian. Note The result of Problem 6.17 is true without the assumption that (9(l),|G:iV|)= 1.
98 Chapter 6 (6.18) Let N <a G and suppose G = NH with N n H = 1. Let 9 e Irr(JV) be invariant in G and assume (9( 1), | G: JV |) = 1. If H is solvable, show that 9 is extendible to G." Note The result of Problem 6.18 is true without the assumption that H is solvable. However, even if H is abelian, the condition (9(1), | G: N\) = 1 cannot be removed. (6.19) Let N o G with G/N a p-group and let 9 e Irr(JV) be invariant and p- rational and assume p Jf o(9) and p ^ 9(1). Let / be the extension of 9 to G with p X o(x) (which exists by Corollary 6.28.) Show that / is p-rational. (6.20) (Thompson) Let EP(G) denote the minimal normal subgroup of G such that G/EP(G) is an elementary abelian p-group. Suppose N £ G with p Jf | G: N\ and EP(JV) = EP(G) n JV. Show that Op(N) = Op(G) n JV. Hints Sharpen the proof of Theorem 6.31 as follows. Choose 1 so that 1" = lv. Let <$ = 0(O|G|/OP) and let A e Sylp(0). Now A permutes Irr(H) for all H £ G and X is invariant under A. Now proceed with the proof of Theorem 6.31 but arrange matters so that x, i> and q> are invariant under A. (6.21) Let 1 = H0 < H[ < • • ■ < H„ = G. Assume that Hj/H,-_ t is non- abelian. Show that there exists / elrr(G) with /(l) > 2". Hint Use Corollary 6.17.
T.I. sets and exceptional characters Suppose we know that G has a subgroup H with certain specified properties and assume that we have some information about how H is embedded in G. How can we draw conclusions about G? We might try inducing the irreducible characters of H to G. Usually this gives little information since if | H | is much smaller than | G |, the characters 9G tend to have large numbers of irreducible constituents with large multiplicities, even if 9 e Irr(H). In certain situations, however, one can find a difference of two characters 9t — 92 of H where ($t — 92)G = Xi — Xi and Xi afld Xi are under control. In these situations, information about Irr(G) can be obtained. The earliest example of the use of this technique is in the proof of Frobenius' theorem. We shall give this before discussing any of the later refinements of these ideas. In Theorem 6.34 we considered groups G having a normal subgroup JV such that CG(x) £ JV for all 1 j= x e JV. It follows immediately using Sylow's theorem and the fact that nontrivial p-groups have nontrivial centers that (|JV|, |G:JV|) = 1. By the Schur-Zassenhaus theorem we conclude that there exists H s G with NH = G and JV r\H = 1. In this situation, let g e G — H. Write g = xn with x e H and 1 j= n e JV. If y e H r\ Ha, then y e H" and y = h" for some he H. Since y e H, we have [h, n] = h~ 1h" = h~lye H. Since JV <a G, we have [h, n] e H r\ N = 1 and h e C(n) £ JV. Thus h = 1 and y = 1. We conclude that H r\ HB = 1. (7.1) definition LetH £ G, with 1 < H < G. Assume that H r\ Hg = 1 whenever g e G — H. Then H is a Frobenius complement in G. A group which contains a Frobenius complement is called a Frobenius group. 99
100 Chapter 7 We have proved in the foregoing that groups satisfying the hypotheses ofTheorem 6.34 with 1 < JV < G are Frobenius groups. Frobenius' theorem is the converse of this. (7.2) theorem (Frobenius) Let G be a Frobenius group with complement H. Then there exists JV*«a G with HN = G and H r\ JV = 1. The fact that the group JV of Theorem 7.2 satisfies the condition that CG(x) s JV for all 1 t6 x e JV is not difficult to prove and is left to the reader. Before beginning the proof of Theorem 7.2, we mention the curious fact that it is trivial to find JV. What is hard is to prove that JV is a subgroup. (7.3) lemma Let H be a Frobenius complement in G. Let n = (g- uh*)u{1}. Then|JV| = \G:H\. If M<c G with M n H = 1, then M £ JV. Proof Since H = NG(H), there are \G: H\ distinct subgroups of the form Hx. These contain exactly |G: H|(|H| — 1) nonidentity elements. The remaining elements of G constitute the set JV. We have \N\ = \G\ - |G:H|(|H| - 1) = \G\ - \G\ + \G:H\ = \G:H\. IfM<GandMnH=l, then M n Hx = 1 for all xeG and thus M ^ N. The proof is complete. | We mention that except for some special cases, no proof of Theorem 7.2 is known which does not use characters. (7.4) lemma Let H be a Frobenius complement in G. Let 9 be a class function of H which satisfies 9(1) = 0. Then (9G)H = 9. Proof Let 1 ¥= h e H. Then 9G(/I) = (l/|tf|)E9V*"1)- xeG lf$°(xhx-1) ^ 0,then I * xhx~l eH n Hx~l andxeH.Then 9°(x^_l) = 9(h). We have 9G(/I) = (l/|tf|)£9(/I) = 9(/I). xeH Since 9G( 1) = | G: H \ 9( 1) = 0, the proof is complete. | The proof of Theorem 7.2 may be motivated as follows. Assuming the theorem is true, let / e Irr(G) with JV £ ker /. Then Xh e Irr(H). Now given Xh = (p e Irr(H) we try to find / e Irr(G). We do this for each <p e lrr(H) and check that (~)ker x is the desired normal subgroup.
T. I. sets and exceptional characters 101 Proof of Theorem 7.2 Let \H j= (pelrr(H) and write 9 = cp — <p(l)lH so that 9(1) = 0. Now [9G, 9G] = [9, (9G)H] = [9, 9] by Lemma 7.4. Thus [9G 9G] = 1 + <p(l)2. Now [9G, 1G] = [9, 1H] = -<p(l). We may therefore write 9G = q>* — <p(l)lG, where <p* is a class function of G, \_q>*, 1G] = 0, and 1 + <p(l)2 = \_(p*,(p*l + <p(l)2,sothat \_q>*, <p*] = 1. Since 9 is a difference of characters, so is 9G and hence <p* is a difference of characters also. Since [cp*, cp*2 = 1, it follows that ± q>* e Irr(G). Furthermore, if he H, then <p*(/j) = 9G(/z) + <p(l) = 9(/j) + <p(l) = <p(h). In particular, <p*(l) > 0 and thus cp* e Irr(G). For every nonprincipal cp e Irr(H), we have now chosen an extension, <p*eIrr(G).LetM = f)„ ker <p*.Ifx eM n H,then<p(x) = <p*(x) = <p*(l) = <p(l) for all (p elrr(H) and thus x = 1. By Lemma 7.3, M £ N. Conversely, if g e G lies in no conjugate of H, then <?*(#)-<p(l)=9G(g) = 0 and g e ker cp*. It follows that M = N and hence the normal subgroup M satisfies |M| = |G:H|. We have \MH\ = \M\\H\ = |G:H||H| = \G\ and the result follows. | The normal subgroup whose existence is guaranteed by Theorem 7.2 is called the Frobenius kernel of G. By Lemma 7.3, it is uniquely determined by H. We mention without proof that in fact a Frobenius group has a unique conjugacy class of complements and a unique kernel. An entirely equivalent version of Frobenius' theorem may be stated as follows. (7.5) corollary Let G be a transitive permutation group on Q with character /. Assume /(g) < 1 for all g eG with g j= 1. Then the set {g e G | xig) = 0} u {1} is a transitive normal subgroup. Proof Let a e Q and assume | Q| > 2. If there exists any geG,g j= 1 with X(g) t6 0, then Ga is a Frobenius complement. By Theorem 7.2. and Lemma 7.3, {geG|x(g) = 0} u {1} = JV is the Frobenius kernel. It is transitive since NGX = G. | We shall now discuss some of the ways that the ideas in Frobenius' proof have been extended and used in other contexts. We need to introduce some terminology. (7.6) definition Let XcGbea subset. Then X is a T.I. set (trivial intersection set) if for every g eG, either I9 = IorIsnIs {1}.
102 Chapter 7 (7.7) lemma Let X be a T.I. set in G and let <p and 9 be class functions on JV = NG(X). Assume that cp and 9 vanish on JV - X and that 9(1) = 0. Then 9G(x) = 9(x) for all x eX and [9G, <pG] = [9, <p]. JVoqf Let x e X. Then 9G(x) = (1/1 JV|) £yeG 9P(yxy-l). If 9°(yxy-') * O.then^'eXnl'"' srndyxy'1 # 1. It follows that yeJV and ^(yxy'1) = 9(x). The first statement now follows. Wehave[9G,<pG] = [(3%, <p]. Since <p vanishes on JV - Xand(9% - 9 vanishes on X, we conclude that [(9% - 9, <p] = 0 and hence [9G, 9G] = [9, <p] as claimed. | Because of the requirement that 9(1) = 0, the preceding lemma cannot be applied if 9 is a character. We usually take 9 to be a difference of two characters. Such a difference is called a generalized character. Note that 9 is a generalized character of G iff [9, /] e Z for all / e Irr(G). Also, the set of generalized characters of G is a ring. We are now ready to consider groups whose Sylow 2-subgroup P is generalized quaternion. If | P \ > 16, we shall prove the theorem of Brauer and Suzuki which asserts (among other things) that such a group cannot be simple. This theorem is also true if | P \ = 8 but it is more difficult to prove in that case. We shall use the following facts about a generalized quaternion group P: (a) P has a cyclic subgroup of index 2; (b) \P:P'\ = 4; (c) |Z(P)| = 2; (d) noncyclic subgroups of P are themselves generalized quaternion; (e) P contains a unique involution. We shall need to use the result of Problem 3.9 in the proof. (7.8) theorem (Brauer-Suzuki) Let P e Syl2(G) be generalized quaternion with \P\ > 16. Then there exists JV<i G with |JV| odd and such that G/N has a normal subgroup of order 2. Note that possibly JV = 1 in 7.8. We first prove the following weaker version. The full result will then follow easily. (7.9) theorem Let PeSyl2(G) be generalized quaternion with \P\ > 16. Then there exists M <i G such that | M | is even and G/M is nonabelian. Proof Let H s P be cyclic with \P: H\ = 2. We have FsH and |P'| = |P|/4 > 4, so that F > Z(P) and hence Cp(P') = H. Let C = CG(P') and JV = NG(P'). Now P e Syl2(JV) and C <i JV so that H = PnCe Syl2(C). Since a Sylow 2-subgroup of C is cyclic, it follows (for instance from Burn- side's transfer theorem) that C has a normal 2-complement K and K <i JV.
T. I. sets and exceptional characters 103 Now N/C is isomorphic to a subgroup of Aut(P'). Since P' is a cyclic 2-group, it follows that N/C is a 2-group. We therefore conclude that JV = KP. Since C = KHwehave|JV: C\ = 2. Let U £ F with \F: U\ = 2. Let X = C- UK. We claim that X is a T.I. set and N = N(X). Now C/K s H is cyclic and UK/K is its unique subgroup of order equal to | U \. It follows for y e C that y e X iff o{yK) in C/K exceeds | U \. We conclude that y e X iff (211/1) | o(y). Now |P'| = 2|t/| and P'-a C. Since P' contains all elements of C of order 21 U \, it follows that F £ <x> for all x e X. If x e X n X9, then P' £ <x> and (P'f £ <x>. Since |P'| = |(P7I, we have F = (F)B and geN. Since clearly N £ N(X), it now follows, as claimed, that X is a T.I. set and N = N(X). Since | P' | > 4, we also have 41 o(x) for every x e X. Now C/l/K is cyclic of order 4. Let X be a linear character of C with ker 1 = UK. Let 9 = 1N - (lcf. Since ker 1N = UK £ ker(lcf, we conclude that 9 vanishes on t/X and in particular 9(1) = 0. Clearly, 9 vanishes on N — C and hence 9 vanishes on N — X. We may therefore apply Lemma 7.7 with cp = 9 to conclude that [9G, 9G] = [9, 9]. To compute [9, 9], observe that (lc)^ = In + fi where ker y. = C. We claim that AN e Irr(iV). Otherwise, XN is a sum of linear characters and F £ N' £ ker 1" = UK, which is not the case. Thus 9 = 1N - fi - lN and [9, 9] = 3. Now [9G, 9G] = 3 and [9G, 1G] = [9, 1„] = -1. It follows that 9G= ±Zi±Z2- 1g> where %u ite Irr(G) are not principal. Since 9G(1) = 9(1) = 0, we conclude that the signs above are opposite and without loss we may write 9G = Zi - X2 ~ 1G- Since 9G(g) = 0 unless g is conjugate to an element of X, we conclude (1) Xi(Q) - 12(9) = 1 if 4Jfo(g). Since P has a unique involution, it follows that G has a unique conjugacy class of involutions. Let Jf t be this class. Define the class function <p on G by (p(9) = \{(x,y)\x,yeJfu xy = g}\. lf(p(g) ¥= 0, then g = xy for involutions x and y and hence g* = yx = g~l. \io(g) is even, let a be the involution in <g>. Then <x, er> is an abelian, non- cyclic group of order 4. Since P has no such subgroup, neither does G by Sylow's theorem. We conclude that q>(g) = 0 if o(g) is even. Thus <p(9)(Xi(g) - Xiia) - 1) = 0 for all geG. Therefore (2) [cp, Iv-li- IgJ = 0.
104 Chapter 7 In C[G], let Kv be the class sum corresponding to the conjugacy class Jf v. Then in the notation of Problem 3.9, ^1^1 = Lflllv^v> andifge Jfv, then (pig) = allv. By Problem 3.9 we have for g e Jf v and x e Jf t that «»(ff) = fliiv = (l-^il2/|G|) £ )(W2i/z(l). Xelrr(G) Since /(x) and <p(#) are real, we may rewrite this equation as (\G\/\Jt1\2)(p= £ (ZW2/Z(D)Z Xelrr(G) and (|G|/|jf1|2)[<p,z] = zW2/z(l) for all / e Irr(G). We conclude from Equation (2) that ZiW2/Zi(i) - Z2W2/z2(i) = i- From Equation (1) we have /2M = ZiW — 1 since 4 Jf o(x) and also /2(1) = Xi(l) — 1. Substitution into the preceding equation yields ZiW2/Zi(l) - (ZiW - l)2/(Zi(l) - 1) = I- Simplifying this, we obtain (ZiW - Zi(l))2 = 0 and we conclude that x e ker Xi- Since o(x) = 2, we have |ker Xi\ is even. Now, ^(l) = 1 + /2(1) ^ 2 and thus G/ker /t is nonabelian. The proof is now complete with M = ker/!. | Proof of Theorem 7.8 Let U be the (normal) subgroup generated by all of the involutions in G. If U has a cyclic Sylow 2-subgroup, then U has a normal 2-complement JV and JV <i G. In this case, U/N <i G/JV and U/N is a cyclic 2-group. The result follows. Assume then that the Sylow 2-subgroups of U are noncyclic. We derive a contradiction. Since 8| | U\, we may choose V, with U £ V £ UP such that | 7: U | < 2 and 1611 V \. Now Theorem 7.9 applies to V and we may choose M<F with 7/M nonabelian and | M | even. Since a Sylow 2-subgroup of V contains a unique involution, it follows that all involutions of V are conjugate. Since M<F contains an involution, it follows that M contains all of the involutions of V and hence U £ M. Thus 17: M | < | V: C/1 <2 and this contradicts V/M being nonabelian and completes the proof. |
T. I. sets and exceptional characters 105 By the Brauer-Fowler theorem (Corollary 4.12), there are at most finitely many nonisomorphic simple groups which contain an involution whose centralizer is isomorphic to some given group C. Much work has been done in recent years to find all of the simple groups corresponding to various specific C. A key step is to find all possible orders of these simple groups and this often involves character theoretic techniques related to those in this chapter. As an illustration of this we prove the following. (7.10) theorem Let G = G' and suppose reG is an involution with Cg(t) dihedral of order 8. Then \G\ = 168 or 360. We need a lemma. (7.11) lemma (Thompson) Let SeSyl2(G) and suppose M^S with | S: M | = 2. Let t e S be an involution which is not conjugate in G to any element of M. Then t £ G'. Proof Let G act by right multiplication on Q = {Mg\geG}. Then |Q| = 21G: S|. Now if Mgx = Mg, then gxg~l e M, which is not the case. Thus t has no fixed points on Q and since \G:S\ is odd, it follows that t induces an odd permutation on Q. Therefore, there exists A <i G with | G: A | = 2 and i^A. The result follows. | Proof of Theorem 7.10 Let D = Cg(t) and D £ SeSyl2(G). Then Z(S) £ Z(D) = <t> and hence S £ C(t) = D. Thus D e Syl2(G). Let M £ D be cyclic of order 4 so that t is the unique involution in M. Since G = G', Lemma 7.11 guarantees that every involution in G is conjugate to t. Now M is a T.I. set in G since if M n M" # 1, then teM* and thus t = t* and x e D. Since M<Dwe have M = Mx. This argument also shows that D = NG(M). Let 1 be a faithful linear character of M and let 9 = (1M - Xf. Since kD is irreducible, it follows that [9, 9] = 3. Also, 9(1) = 0 and 9 vanishes on D - M. Thus by Lemma 7.7 we have [9G, 9G] = 3 and also (9% = 9. Since 9G(1) = 9(1) = 0, it follows that we may write 9G = 1G + / - i//, where /, \ji e Irr(G). Calculation in D yields 4 = 9(t) = 9g(t) and we have (1) 0=1 + z(l) - Ml), 4=1+ jrft) - ^t). Let Jf denote the (unique) conjugacy class of involutions in G and define the class function cp by 9(g) = \{(x,y)\x,yejf,xy = g}\.
106 Chapter 7 If xy = g for involutions x and y, then gx = g~l and conversely, if x e Jf, x 76 g, and g* = g"l, then y = xg is an involution. It follows that cp(g) = \{xejf\x ¥= g, gx = g~l}\- If 1 ¥= geM and gx = g~\ then ix = 1 and xeD. We conclude that cp(g) = 4 for 1 =£ geM. Reasoning as in the proof of Theorem 7.9 and using Problem 3.9, we have 9 \g\ s£(G)m)q' Since | Jf | = | G | /8 this yields rQG -, |G|f. yix)2 m2' Also, [9G, cp] = [(1M - X), <Pm\- Since <p has the constant value 4 on M - {1}, this yields [9G,<p] = (1 + i) + (2) + (1 - i) = 4. We conclude that XW2 -AW2 (2) 2s = \G 1 + z(i) 'A(i) Write a = x(l) and ft = /(t) so that t/^(l) = a + 1 and i/^(t) = ft - 3 by Equations (1). By the second orthogonality relation we have 8 = I C(t) I > 1 + X(x2) + ^(t)2 = 1 + ft2 + (ft - 3)2. Since ft e Z, we conclude that ft = 1 or 2. Assume ft = 1. Equation (2) yields 28 = |G|[1 + (1/a) - (4/(a + 1))] and \G\ = 28a(a + l)/(a - l)2. Now 2\a(a + 1) but 24/|'|G| and we conclude that 23\(a — 1). Therefore 22 X a(a + 1) and since 2311G | we have 24 X (« — !)• No odd prime divisor of a — 1 can divide 28a(a + 1) and we conclude that a — 1 is a power of 2 and thus a = 9. This yields |G| = 22 x 9 x 10 = 360. Now assume ft = 2. Equation (2) yields 28 = |G|[1 + (4/a) - (l/(a + 1))] and \G\ = 28a(a+ l)/(a + 2)2. Reasoning exactly as above, we conclude that a + 2 = 8 and \G\ = 22 x 6 x 7 = 168. The proof is complete. | We mention that GL(3, 2) s PSL(2, 7) is the unique group G of order 168 with G = G' and A6 s PSL(2, 9) is the unique one of order 360. We go back to the problem of obtaining information about the irreducible characters of a group from information about a subgroup. Let
T. I. sets and exceptional characters 107 y £ Irr(JV). We introduce the notation Z[5^] for the set of Z-linear combinations of elements of Sf. (Thus Z[Irr(JV)] is the set of generalized characters ofJV.) Suppose JV £ G and that we can find a map *: Z[y~] -► Z[Irr(G)] such that * is Z-linear and that |>*, 9*] = |>, 9] for all cp, 9 e Z\_y]. (Such a map is called a //near isometry.) In that case we have [/*, /*] = 1 for / e y and thus + x* e Irr(G). Write e(x) = + 1 so that e(/)x* e Irr(G). The characters e(/)X* are called the exceptional characters associated with y and*.They are in one-to-one correspondence with y. How might such a map * be constructed? An easy example of a linear map 1\_y\ -» Z[Irr(G)] is the induction map 9 -» 9G. This map, however, is rarely an isometry on Z[5^]. By Lemma 7.7, we see that there are situations when induction is an isometry on Z[^]° = {9 e ~l\_y~\ 19(1) = 0}. This occurs, for instance, if JV = N(X), where X is a T.I. set and y = {/e Irr(iV) | / vanishes on JV - X}. The problem then is to extend a linear isometry from Z[^]° to all of Z[^]. (7.12) definition Let JV £ G and 9> £ Irr(JV) with \y\>2. Suppose t:Z[^]° -» Z[Irr(G)]° is a linear isometry. We say that (y, t) is coherent if t can be extended to a linear isometry * defined on Z[5^]. If t is the induction map and (5^, t) is coherent, we simply say y is coherent. It should be emphasized that even in this case, the map * usually is not induction. The prototypical example of coherence is where JV is a Frobenius complement in G. There, Irr(JV) is coherent; and the proof of that fact is the essence of the proof of Frobenius' theorem. If (y, t) is coherent, the map * is not always uniquely defined. Nevertheless the set of exceptional characters e(x)x* is uniquely determined by (y, t). (7.13) lemma Let (y, t) be coherent and let * be a linear isometry which extends t. Then there exists e = + 1 such that ex* e Irr(G) for all / e y. The function f:Sf-> Irr(G) defined by /(/) = ex* is one-to-one. The image off is {ip e Irr(G)| [9T, t/0 * 0 for some 9 e Z[^]°}. Proof For / e y, [/*, /*] = 1, and we may choose e(/) = + 1 so that e(x)X*e'lrr(G). We claim e(x) = e(£) for all ZeSf. We have 9 = *(1)£ - £(l)xeZ[,ST and thus Z(l)^-^(l)Z*=9IeZ[Irr(G)]°. Evaluation at 1 yields 0 = x(l)£*(l) - £(l)z*(l) a"d hence £*(1) and z*(l) have the same sign. Thus e(x) = £(£) as claimed. The foregoing also shows that [9T, /(/)] ^ 0 for some 9 e Z[^]°. Conversely, suppose ^elrr(G) and [9T, ^] # 0, where 9eZ[^]°. Then
108 Chapter 7 9 = Yxesr ax* and & = X ax**- Therefore 0 # [/*, t/f] = e[/(z), i/f] for some xey and hence/(z) = \p. To show that/is one-to-one, it suffices to show that * has trivial kernel. This follows since if 9* = 0, then 0 = [9*, 9*] = [9, 9] and hence 9 = 0. | Suppose N £ G, if ^ Irr(iV), |^| > 1 and t: Z[^]° -»Z[Irr(G)]° is a linear isometry. We seek conditions on if sufficient to guarantee that (if, t) is coherent. One such condition is that all z e if have equal degrees. Although this is not hard to prove directly, we shall derive it as a corollary of the following more general result of Feit. The main point of Feit's theorem may be summarized as follows. List the degrees z(l) for z e if in increasing order. Assume that the smallest of these divides all of the others. Then if is coherent if the degrees do not increase too rapidly. (7.14) theorem (Feit) Let N £ G, if £ Irr(JV) and let v. I\_ifJ -> Z[Irr(G)]° be a linear isometry. Suppose if = if0 u {/}, where (if$, t) is coherent. Assume that there exists \p e if$ such that i/f(l)|z(l) and zd) < Y^r, X £(D2- 2^(1) st^o Then (if, t) is coherent. Proof Let *: Z[^0] -» Z[Irr(G)] be a linear isometry which extends t on Z[y0]°. We shall define z* so that the map * can be extended to all of Z[^] by linearity. Let /(l) = dip(l) so that z - dip e l\_ifj. Define A by (i) (z - #r = a - #* + x &<{*, where [A, <!;*] = 0 for all £,eif0. Of course, A is a (possibly 0) generalized character of G and all b^ e Z. We shall show that [A, A] = 1 and all b4 = 0. Now i + d2 = to - #), (Z - #)] = [(Z - #r, (z - dm and thus 1 + d2 = [A, A] + £ V + (ft, - rf)2- Therefore (2) [A,A] + XV = 1 + 2<tfV Furthermore, since t maps into Z[Irr(G)]°, we have (/ - #)T(1) = 0 and thus (3) 0 = 5>^*(1)-#*(1) + A(1). i
T. I. sets and exceptional characters 109 For £ e &0 let S = t/,(l)£ - £(l)t/f e Z[^0]°. Then [e\ (x - dm = [s, a - #)] = dm. Therefore «(D = [HT, E M*] - rf[Sr, ^*] + [E\ A] = 0(l)be - £(1)6, + #(1) + 0 since ST = ^(1)^* — £(l)t/'*. We conclude that Ml)fce = £(1)V Now define /z = ^/^(l). We have then (4) fc« = tf(D for all <* e 5V Also, since ST(1) = 0, we have W)TO = fll)lTO and we may write (5) TO = ofll) for all £ e ^0 and fixed a#0, By hypothesis we have (6) 2#(1)2 < £ £(1)2. Suppose /j^O. Then by (4) we obtain 2<v = M2(2#(i)2) < ti2 E £(i)2 = E V and therefore i + 2rfV^EV- Now (2) yields [A, A] +2db/ <2dbr Since b^, is an integer j= 0 (since ^#0) and [A, A] > 0, we obtain A = 0 and *V = 1- Thus A(l) = 0 and (3), (5), and (4) yield #(l) = X^(l) = Ml£(l)2- Since b^, = l,n= 1/^(1) and thus #(1)2 = ££(1)2- This contradicts (6) and proves /z = 0.
no Chapter 7 It now follows that all bi = 0 and (1) yields (X - #)T = A - #*. From (2) we have [A, A] = 1 and we extend * to Z[5^] linearly by defining Z* = A. We now show that * is a linear isometry on Z[5^] and that it extends t. If 9 e Z[^]°, with [9, /] = a, we can write 9 = a(x - #) + <p, where <p e Z[^0]°. Then 9* = a(z - #)* + <p* = a(x - #)T + <px = F and thus * extends t. To show that * is an isometry, it suffices to show [/z*, v*] = [/i, v] for fi, v e y. If fi, v e Sf0, we already know this. Since [A, £*] = 0 for ^ey0 and [A, A] = 1, the result now follows. | (7.15) corollary Let N c G and ¥ £ Irr(iV) with |^| > 2. Suppose t:Z^]° -► Z[Irr(G)] is a linear isometry. Suppose that all xe ^ have equal degrees. Then (Sf, t) is coherent. Proof Use induction on n = \Hf\. Let Xu Xi e -^ be distinct. Then [(Zi - Xi)\ (Zi ~ lif] = E(Zi - lil (Zi ~ XzTi = 2 and since (xi — Z2)T(1) = 0, we may write (xi — XiY = a — P where a, /? elrr(G) are distinct. If n = 2, define xt* = a and /2* = /? and extend by linearity to Z[^]. In this case Z[^]° = {a(xt - z2)|aeZ} and thus * agrees with t on Z[^]°. If n = 3, let x3 e <f - {xu Xi)- Reasoning as above, (xi - XsY = fi - v where fi, v e Irr(G) and ^v, Also [(Zi - to)', (Zi - XsYl = 1 and we conclude that either y. = a or v = /? but not both. If /z = a, define Zi* = ^ Xi* = A and Xz* = v. If v = /3, define *!* = -0, *2* = -a, and X3* = — /z. In either case, extend * linearly to Z[5^] and check that * agrees with t on Xi — Xi and on Zi — Xz and that * is an isometry. Since ZDST = {a(Xl - Xi) + HXi - X3)\a,beZ}, the result follows in this case. Suppose n > 3. Write & = £f0 u {/}, where 3 < | ^01 = " — 1. By the inductive hypothesis, (£f0, t) is coherent. Choose ipeSf0 and observe that
T. I. sets and exceptional characters 111 ^(1)|Z(D since ^(l) = z(l). Now and thus (Sf, t) is coherent by Theorem 7.14. | The most common applications of coherence are in the situation of a "tamely imbedded " subgroup as defined in the paper of Feit and Thompson on the solvability of groups of odd order. For the purposes of this book, we give a considerably more restrictive definition which, nevertheless, is important in applications. (7.16) definition Let K £ G. Assume (a) K is a T.I. set in G; (b) K<NG(K); (c) CG(fc) £ K for all 1 ¥= k e K. Then K is a T.I.F.N. subgroup of G. The "F.N." in the foregoing definition stands for "Frobenius normalizer." Note that if N = NG(JC), where K is T.I.F.N. in G, then N is a Frobenius group and K is the Frobenius kernel. By Theorem 6.34, the irreducible characters of N are thus of two types, those with kernel containing K and those induced from nonprincipal q> e hr(K). If y = {/ e Irr(JV) | K £ ker /}, then the object of what follows is to prove that y is coherent. Results of Feit and Sibley do, in fact, prove this in most cases. Note that if K £ G satisfies condition (a) of Definition 7.16, then (c) is equivalent to CN(k) £ K for all l^keK, where N = NG(K). An important theorem of Thompson (not involving characters) asserts that Frobenius kernels are necessarily nilpotent. The nilpotence of T.I.F.N. subgroups will be assumed in what follows. (The reader may simply consider this to be an additional hypothesis in Definition 7.16.) A more elementary fact about a T.I.F.N. subgroup K of G is that (\G:K\, \K\) = 1, that is, K is a Hall subgroup. To see this, let p||X|, let PeSylp(X) and suppose P £ SeSylp(G). Then Z(S) £ C(P) £ K by part (c) of 7.16. Now Z(S) ^ 1 and so S £ C(Z(S)) £ K, again by 7.16(c). Thus PX\G:K\. We mention some situations where T.I.F.N. subgroups arise. If PeSylp(G),|P| = pandP = CG(P) < NG(P), then Pis T.I.F.N. This happens, for instance in permutation groups of degree p and in the groups PSL(2, p). A doubly transitive permutation group is a Zassenhaus group if some nonidentity element fixes two points but none fixes three. Let G be a Zassenhaus group on a set Q and let a, /? e Q with a j= /?. Then Ga is a Frobenius
112 Chapter 7 group with complement G^. (We are assuming | Q | > 2 to avoid triviality.) Let K be the Frobenius kernel of Ga. The nonidentity elements of K are exactly those elements of G which fix a and no other point. It follows easily that K is T.I.F.N. in G and Ga = NG(K). (7.17) lemma Let K be T.I.F.N. in G and let JV = NG(K) and <f = {ZeIrr(JV)|K £ ker /}. Then induction defines a linear isometry Z[^]° -+ Z[Irr(G)]°. Also, if 9 e Z[^]°, then (9G)N = 9. Proof UieHf, then x = £N for some £ e Irr(X) by Theorem 6.34. Thus % vanishes on JV - K. If cp, 9 e Z[^]°, it follows that <p, 9 vanish on N - K and Lemma 7.7 yields \jpG, 3G] = [<p, 9] and the first assertion is proved. Also, (9G)K = 9K by Lemma 7.7. Since 9 vanishes JV - K, it suffices to show that 9G vanishes on JV — K in order to prove that (9G)N = 9. However, no element of JV — K can be G-conjugate to an element of K since if xeN - K and o(x)||K|, then K < <X, x> and <X, x> = X<x> so that |<X, x>: K| is not relatively prime to \K\. This contradicts the fact that (| G: K |, | K |) = 1. It now follows from the definition of 9G and the fact that 9 vanishes on JV — K that 9G also vanishes on JV — K and the proof is complete. | (7.18) corollary (Brauer-Suzuki) Let K be an abelian T.I.F.N. subgroup of G. Let JV = NG(X)ande = | JV: K |. Let ^ = {/eIrr(JV)|K £ ker/}. Then either (a) \Sf\ = \,\K\ = \ + e and K is an elementary abelian p-group or (b) y is coherent and there exists a one-to-one function/: y -»Irr(G) and e = ± 1 such that (z - if = e(f(x) - /(0) for all z, £ e ^. Proof All of the orbits of nonprincipal irreducible characters of K under the action of JV have size e. Thus /(l) = e for all / e ^ and \Sf\ = (\K\ — l)/e. U\Sf\>2, then Corollary 7.15 yields that Sf is coherent. In this case, (b) follows from Lemma 7.13. If \y\ = 1, then \K\ — 1 = e and the nonidentity elements of K are all conjugate in JV. It follows that they all have the same prime order and (a) follows. | The theorems of Feit and Sibley generalize Corollary 7.18 by dropping the assumption that K is abelian. If | K | - 1 ^ e, they prove that either Sf is coherent or K is a 2-group. Before proceeding with the proofs, we discuss some implications of coherence in the T.I.F.N. situation. (7.19) lemma Let X£G be T.I.F.N. and let JV = NG(X). Suppose 3C £ {/eIrr(JV)|X £ ker /} and that 9C is coherent. Let * denote an iso-
T. I. sets and exceptional characters 113 metry Z\_3T\ -> Z[Irr(G)] such that * extends induction on Z[#"]°. Let « = Ez6arZ(l)Z-Then (a) If \pelrr(G) and \p$ {±y*\is3C}, then i//N = aa. + 9, where a is rational and [9, Z] = 0 for all Z e 9C. (t>) ((x*)n ~ X)/x( 1) is independent of Z e 3C. (c) (x*)n - X = a<* + # for x g #\ where a is rational and [9, £] = 0 for alUeST. Proof (a) Let Z, £ e SC. Then [>K,Z(1K; - £(1)Z] = DMzdK; - £(Dz)G] = [>, Z(D£* - flDzM = 0. Thus x(1)|>n, £] = ^(1)[^, z] and hence |>N, z]/z(l) = a is independent of Z e #\ Therefore, \j/N = aa. + 9, where [9, Z] = 0 for all Z e #\ Thus (a) is proved. (b) Let xA e X. Then A = z(l)£ - £(l)z e Z[^° and so (AG)W = A by Lemma 7.17. However, AG = A* = Z(l)£* — £(l)z* and thus Z(l)£ - £(1)Z = z(l)«*)K - {(l)(z*)K and «D((Z*)W - Z) = zU)(tf*)w - 0. Thus (b) follows. (c) Let £,»/ e 3C and write A = £{\)r\ - >/(l)£. Then [(Z*)w, A] = [z*. AG] = [/*, A*] = [z, A] so that [(/*)n - /, A] = 0. It now follows as in (a) that (x*)N - Z = «a + #, where a = [(z*)K - z, £]/£(l) and K, 9] = 0 for £ e ST. | (7.20) theorem Let X c G be T.I.F.N. and let N = NG(X). Suppose y={^e Irr(iV)|X £ ker Z} is coherent and let S £ Irr(G) be the corresponding set of exceptional characters. Let * be an isometry Z[5^] -» Z[<f] which extends induction on Z\_Sf]° and let e = +1 so that ez* e $ for Z e y. Then (a) S={i//e Irr(G) | ^ is not constant on K - {1}}. (b) If \ji = ex* e S, then i/fK - eZx is constant on K — {1}. This constant has the form mZ(l)/|./V:K| for some m e Z which is independent of the choice ofZe^. (c) If g e G is not conjugate to an element of K — {1}, then t/K^Vi/Kl) is independent of the choice ofipeS. (d) If z e ^, then Z*( l)/z( 1) is independent of the choice of Z.
114 Chapter 7 Proof We apply Lemma 7.19 with 3C - £f. Then a = pN - pNjK where pN and pNjK are regular characters. In particular, a. is constant on X - {1}. Also, if 9 is a generalized character of JV and [9, /] = 0, for all X e S?, then X is in the kernel of every irreducible constituent of 9 and hence 9K is constant. Now, if \p $ 8, then 7.19(a) yields \pN = aa. +9 and i/f is constant on X - {1}. If \p e S, then 7.19(c) yields eipN = x + <*<* + & where &// = X*- However, / e if cannot be constant on X - {1}, otherwise, by Problem 3.8, every nonprincipal irreducible character of X is a constituent of Xk- This would force \9'\ = 1 which is not the case. Thus \pN is not constant on X — {1} and (a) is proved. The first part of (b) is immediate from 7.19(c). To evaluate the constant, choose ZeSP with £(1) = |JV:X|, that is, £, = XN for some nonprincipal linear 1 e Irr(X). Now (£*)K — £K is a generalized character of X with constant value m on X - {1}. It follows that m e Z. If / e Sf is arbitrary, 7.19(b) yields that m'/x(l) = m/£(l) where m' is the constant value taken on by (X*)k ~ Xk on X - {1}. Since {(1) = \N;K\, we have tri = mx(l)/|JV:X| and (b) follows. Let x, £ e &> so that £(l)z* - z(l)£* = (£(l)z - X(1)£)G, which vanishes on elements g e G not conjugate to elements of X - {1}. Thus £(l)x*(g) - xO)£*(ff) = 0 and (c) follows. Finally, (d) is immediate from 7.19(b). The proof is complete. | We now begin work toward the theorems of Feit and Sibley. (7.21) lemma Let JV be a Frobenius group with Frobenius kernel K. Suppose | JV:X | is even. Then X is abelian. Proof Let t e JV be an involution. Then t $ X since (|X|,JV:X|)=1 and thus x' ¥= x for x e X - {1}. Now map X -» X by x i-> x~ 'x'. This map is one-to-one since if x_1x' = y~V, thenyx-1 = (yx~ly and thus yx~l = 1. Therefore the map is onto and every xeX has the form u~lu'. Thus x' = (u~vux)x = (u')~1u = x"1. We now have (xy)' = (xyy1 = y~ix~l = y'x' = (yxf. Thus xy = yx for x, y e X and the result follows. | (7.22) lemma Let P e Sylp(G) with p ^ 2 and suppose P is T.I.F.N. in G and is nonabelian. Let JV = NG(P) and let 1 < Z £ Z(P) with Z <\ JV. Let 3C = {x e Irr(JV)|Z £ ker /}. Then SC is coherent. JVote By Corollary 7.15, the lemma is also true if P is abelian except in the degenerate case that \3C\ = l.InthatcaseZ = Pand|P| = |JV:P| + 1. Proof of Lemma 7.22 Let e = | JV: P |. By Lemma 7.21, e is odd and thus |JV| is odd. Since Z > 1, 3C ¥=0. Let £ e 3C have minimum possible degree
T. I. sets and exceptional characters 115 andlet^0 = {xeSC\x(\) = {(1)}. Since \N\ is odd, $± £ and thus | ^01 > 2. By Corollary 7.15, st0 is coherent. Now define sets #", for i > 0 by Sft = ^"i-! u {/,•}, where the x, are chosen from <Sf— ^oSothat/.O) < z(+i(l). Thus f0cf1c...cft = | We use induction on i to show that SC{ is coherent. Now the degrees of the / e Sf are all of the form ep" and so £(l)|x(l) for all xe3C.lt follows by Theorem 7.14 that in order to show that Sfi is coherent, it suffices to show for i > 1 that 2£(l)Zj(l)< X X(D2- We have \N\= Z z(l)2 = |N:Z|+ XZ(D2 xelrr(JV) XEiT and so |P:Z| divides £Z6ir Z(l)2- Thus e2|P:Z| divides £ *(1)2. If/ e #, then / = S" for some 9 e Irr(P) and 9(1)2 < | P: Z | by Corollary 2.30. Thus x(l)2 = e29(l)2 divides e2|P:Z| and hence /((I)2 divides ^.y x(l)2. Also,Zi(l)2IZj<l)2forj > j and we conclude that/((I)2 divides ^ZEa.(_1 *(1)2. In particular, since 2£(1) < p£(l) < #,(1), we have 2«l)Zi(l) < Z,(D2 < I Z(D2 and the result follows. | (7.23) lemma Let N £ G and ^ c Irr(N). Let a: Z[^]° -» Z[Irr(G)]° be a linear isometry and let S£, <W £ ^ with #n <^ = 0 and (3C, a) and (<^, a) coherent. Let a and t be isometries on Z[^] and Z[<^], respectively, where er and x are as implied by coherence. Then (a) if, rf] = b for all / e SC and i/e?'. (b) Suppose x0 e#" and >/0 e <& with (ZoOfoo - »7o(l)Zo)a = ZoOfoo1 - r,0(l)x0a. Then 3C u <^ is coherent. Proof For /z, v e ST, write A(/z, v) = /z(l)v - v(l)/z e Z[^]°. Let 5, e = ± 1 so that (5/" and erjz e Irr(G) for ](ef and t] e<W. To prove (a), suppose If, 1*] * 0. Then btf = 9 = eif for some 9 e Irr(G). Pick Xi e 3T - {/} and (|[6f- {>/} and write t/f = «5xiCT and ^ = £>//. Then ^S^ and
116 Chapter 7 we have 0 = [A(z, Xl\ A(i,, ijj] = [A(z, XiT, Afo, i^f] = rf[(zdWf-Zi(l)%(f(l)«-fi(l)0)] = e«5(Zi(D'7i(l) + Zd¥l)[^a). Since [i/f, <!;] ^ 0, this is impossible, and (a) is proved. Now fix Xo e #" and >/0 e <^ and assume A(/0, rj0f = /0(l)'7ot - fo(l W- Define * on 3C u 'W by/* = /"for/e^andf/* = if for >/ e "^ and extend * by linearity to Q\_3C u <¥]. It follows from (a) and the fact that a and t are isometries, that * is an isometry and *: 1\_3C u <¥] -» Z[Irr(G)]. It thus suffices to show that * agrees with the linear extension of a to Q[_SC u W]°. However, * agrees with a on Q[^"]°, Q[^]° and A(/0, rj0). A dimension count shows that these span Q\_3C u W]° over Q and the result follows. | (7.24) theorem (Sibley) Let P e Sylp(G) be T.I.F.N. in G with p # 2. Let N = NG(P) and 9> = {/eIrr(JV)|P £ ker/}. Let e = \N:P\. Then one of the following occurs. (a) \Sf\ = 1, P is elementary abelian, and \P\ = e + 1. (b) ^ is coherent. Proof If P is abelian, this is included in Corollary 7.18. Assume that P is nonabelian so that e is odd by Lemma 7.21, and thus |JV| is odd. Let Z = P'nZ(P)> landlet<^= {i\ s^\Z g ker^.Let^- = {/e^F = ker/}. Then <^ is coherent by Lemma 7.22 and every ^ef satisfies /(l) = e. Since / ,£ / for / e #", we have #" coherent by Corollary 7.15. Since Z £ P', we have SC r\ <W = 0. Most of the proof is devoted to showing that SC u <W is coherent. Let *: l\dC] -» Z[Irr(G)] be a linear isometry which extends induction on J.\_9T]°. Similarly, let t be an isometry on I\W] which extends induction. Since SC r\ <W = 0, Lemma 7.23 yields [/*, if] = 0 for / e SC and >/ e <^. Let a = (1/e) £ ,6^ >/(l)>/. Since e|>/(l) for >/ e ^ a is a character of JV. For %e3£,x* ¥= ±i]z for any ij e<W and Lemma 7.19(a) yields (X*)n = #* + aza, where 9Z is a generalized character of JV such that \_9X, if] = 0 for r\ e 'W and az e Q. Since all x(l) are equal for / e 3C, Lemma 7.19(b) yields that (X*)n ~ X = $x ~ X + ax<* is independent of / e SC. It follows that 9X — x = A and ax = a are independent of x £ 3C and (1) (X*)n = X + A + ace. The next several paragraphs are devoted to proving that a e Z.
T. I. sets and exceptional characters 117 Since Z is contained in the kernel of / e 3C and of every irreducible constituent of A, we have for z e Z that x(z) = x(l) and A(z) = A(l) and hence Equation (1) yields Z*(l) - z*(z) = a(a(l) - a(z)). We also have pN = pNjZ + ea. where pN and pNjZ are regular characters. For z eZ — {1} we thus have \N\ = pN(l)- pN(z) = e(a(l) - a(z)) and hence (2) x*(l)-X*(z) = a\N\/e = a\P\. We now compare j(*(l) and x*(z) in a different way. Let Jf 0, jf 1;... be the conjugacy classes of G, numbered so that (a) Jf0 = {1}; (b) Xtr\Z ±0 iff i<r; (c) Jf;nP^0ifn<s. Note that since P is a T.I. set with N = N(P), we have Jf, n P is a conjugacy class of N for i < s and Jf; n P c z <i N for i < r. Write X; e C[G] for the class sum corresponding to Jf, and KtKj = X fly/i^/i. where ay„ e Z. Now fix i, _/ < r. Let t/f e Irr(G) and let co(KJ = t/f(x) | X ^ | /t/f( 1) for x e Jf^ as in Chapter 3. Let R <=, C be the ring of algebraic integers so that (o(K^ e R. Write t/^(l) = wp' with pjfm, and let <j = \P\/p'. We have eo(KjMK/) = Zflf^K/J- For /z > s, we claim that <y(K,J e <?/?. From the T.I.F.N. property and the fact that Jflir\P=0, it follows that \P\ divides |jfj and thus m(a>(K^lq) e R. Since also q(a>(K^lq) e /? and (w, <j) = 1, it follows that (x^K^/q e R as claimed. Thus eo(K,)eo(Ky) = £ ai)MKd mod «*■ /i=0 Now let r < p. < s and x e Jf^ n P. Let C = Cp(x) and let Q = {(w, v)\ueJth veJfj, uv = x}. Then a,^ = |I1| and C acts on Q by (m, t>)c = (if, vc). If c e C - {1} and (m, vf = (u, v), then u, v e CG(c) £ P and so w e Jf, n P c z and similarly ceZ. However wt> = x e Jf^ and Jf^ nZ=0. This contradiction shows that all orbits of C on fi have size | C | and thus | C| divides alJtl.
118 Chapter 7 We have CG(x) ^ p and so CG(x) = C and \P:C\ divides | jfj. It now follows that miau^aiK^/q) e /? and since q(ai}lloi(K^lq) e /?, we have a^miK^) e 4/? and r oiKiMKj) = X ai}MKd mod ^/?- /i=0 Now suppose t/f e Irr(G) is such that i/f is constant on Z — {1}. Then all ai(KJ are equal (to a, say) for 1 < fi < r and ca(K0) = 1. Also write aH = S=i fly,/, so that co2 = ayo + aya) mod qR. Since | JV | is odd, the nonidentity elements of Z are not conjugate in JV (and hence not in G either) to their inverses. Thus all0 = 0. Assume Jf 2 is the class of inverses of Jf t so that a12o = I Jfil = |G|/|P|. We thus have (3) ana> = a2 = \G\/\P\ + al2oi mod qR. In the special case that i> = 1G, we have co = \G\/\P\ and q = \P\ so that flii|G|/|P| = |G|/|P| + fl12|G|/|P|mod|P| and atl = I + a12 mod |P|. Now (3) yields (1 + al2)co = \G\/\P\ + al2co mod qR and (4) (o = \G\/\P\ mod qR. We apply this to the character i/f = e/* e Irr(G) with e = + 1 and ^ef. Note that \\i is constant on Z - {1} by (2) and thus (4) applies. We have for z e Z - {l}that Z*(z)|G:P|/z*(l) = co = \G:P\ mod qR. Since \P\ divides <j/*(l), this yields Z*(z)|G:P| = z*(l)|G:P|mod|P|K. Now (2) yields \G:P\-\P\a = \G:P\(x*(l) - x*(z))e\P\R and hence \G:P\aeR. Since also \P\a = x*(l) - z*(z)e/? and (|G:P|, |P|) = 1, we have ae/?n<Q = Zas desired. Now let x, Xi e % and >/ e ^. Note that x(l) = e divides >/(l) and we write c = r](l)/e. Let <p = c* - r\ e Z[^]°. We have [<PG, Zi*] = l>, (Zi*)w] = |>, Zi + A + fla].
T. I. sets and exceptional characters 119 Since [/, a] = 0 = [r\, xt + A], this yields [<PG. Zi*] = c[x, /i + A] - fl[ij, a]. However, [>/, a] = c by definition of a and since a e Z, we have c| [<pG, /t*]. This calculation also shows that [>G, Z*] = c + WG, Zi*] for xt t6 /. This yields q>G = c(\ + b)X* + cb X Z* ~ r> where ft e Z and [T, £*] = 0 for all £ e X. We also have [>G, «pG] = [<p, <p] - 1 + c2 and thus there are two possibilities: (i) b = 0 and [T, T] = 1; or (ii) fc = -l,|aT| = 2and[r,r| = 1. In situation (ii), we can replace* by** where /** = — Z2*and/** = — Zi* for 9C = {xu Xi)- The result of this change is to put us into situation (i). We thus assume that (ex — rjf = ex* — T. Now let i]l e 'W — {r\\ so that fid) = 19, f(l>h - fi(Df] = l>G. '/(D'/it - 71(1)71 = [r, fftdjff1 - frtitoii. Thus again we have two possibilities: (i) T = if; or (ii) T = -riSm<\ri(\) = ril(\). If T ¥= rf, then rjtz = -T for every i^e*- {>/} and thus \<&\ = 2 and <^ = {>/, >/!} with >/(l) = ^(l). In this situation, we can redefine t and thus we may assume that (i) occurs. Thus (Q - if = Q* - 1Z and hence #T u <W is coherent by Lemma 7.23(b). To complete the proof, we observe that £,6» l(l)2 = \N\ - \N\Z\ and so is divisible by \P\Z\ and hence by e2\P:Z\. Let i// e 9> - (9C u <^). Then as in Lemma 7.22, we have \p = 9N, with 9 e Irr(P) and 9( l)2 < | P: Z |. Thus iP(l)2 = e2B(l)2<e2\P:Z\< ££(1)2- 4e»
120 Chapter 7 Let ](el so that x(l) = e and \p(l) > px(l) > 2/(1) since \ji £ J and so 9(1) ^ p. Thus 2jrflMl) < ^(l)2 < X £(1)2. Since #" u <& is coherent and x(l)|t/^(l) for all ii sif, repeated application of Theorem 7.14 yields that if is coherent and the proof is complete. | (7.25) theorem (Feit-Sibley) Let K £ G be T.I.F.N. with JV = NG(K), e=\N:K\ and ^ = {/ e Irr(JV)|K £ ker /}. Then one of the following occurs. (a) \Sf\ = 1,\K\ = e + \ and K is an elementary abelian p-group. (b) K e Syl2(G) and | K: K' \ < 4e2. (c) if is coherent. Proof If K is abelian, either (a) or (c) occurs by Corollary 7.18. Assume then, that K is nonabelian and so e is odd. If \K:K'\ = e + 1, then K/K' is an elementary abelian p-group and thus K is a p-group. Since e + 1 is even p = 2 and (b) follows. Assume therefore, that \{xeif\K' c ker /}| > 2. This set is thus coherent by Corollary 7.15. Let L<i N with L £ K' minimal such that X = {/ e ^ | L £ ker /} is coherent and assume that if is not coherent so that L > 1. Let M o N be such that L/M is a chief factor of N. Since {ze^|M£ker/} is not coherent, repeated application of Theorem 7.14 yields \ji e Irr(iV) such that 2^(1) > XX(D2 = \N:L\ - \N:K\ = e[\K:L\ - 1) and M £ ker i/f. Let Z/M = Z(K/M). Then Z n L > M and hence LqZ. Also ^ = 9N for some 9eIrr(X) and 5(1)2<|K:Z| so that iA(l)2 < e2|X:Z|.Thus 4e4\K:Z\ > 4eV(D2 ^e\\K:L\ - l)2. Now write a = |K:L|andfc = \Z: L|.We have b\K : Z\ = \K:L\ = a and 4e2a = 4>2fc|K:Z| > b(\K:L\ - l)2 = fc(a - l)2. If 4e2 < b(a - 2), this would yield ba(a — 2) > ft(a - l)2, which is not the case. Thus b(a — 2) < 4e2. If Z = L, then Z(K/M) = L/M is a p-group for some prime and thus K/M is a p-group. Since M £ X', it follows that K is a p-group. By Theorem 7.24, we have p = 2. Also, a — 2 < 4e2 so that \K:K'\ < \K:L\ = a < 4e2 + 1.
Problems 121 Since | K : K' | is a power of 2 and e > 1 is odd, we have \K:K'\ < 4e2 and (b) holds. If Z > L, then we must have e\(\Z: L\ - 1) and b > e + 1. Also, K/M is nonabelian and thus Z < K and e|(|X : Z\ — 1). Thus a > (e + l)b > (e + l)2 > e2 + 2. Thus 4e2 > b(a — 2) > (e + l)e2 > 4e2 since e > 3. This contradiction proves the theorem. | We remark that situation (b) can actually occur with y not coherent. The inequality | K : K' \ < 4e2 can be sharpened. In fact, the groups in which (b) occurs have been classified. Problems (7.1) Let N <i G, H £ G, with NH = G and N n H = 1. Show that the following are equivalent: (a) CG(n) £ iV for all 1 *neN; (b) C„(n)= 1 for all 1 *neN; (c) CG(h) £tf for all 1 *h eH; (d) Every x e G — N is conjugate to an element of H; (e) If 1 ¥= h e H, then h is conjugate to every element of Nh. (f) H is a Frobenius complement in G. iVofe Problem 7.1 does not involve characters. It is included to acquaint the reader with some elementary properties of Frobenius groups. Much deeper information is known. (7.2) (Wielandt) Let H £ G with M <i H and suppose that H n Hx £ M whenever x £ H. Show that there exists N<i G with NH = G and N n H = M. Hint Note that H - M is a T.I. set. Mimic the proof of Frobenius' theorem. (7.3) Let H £ G and £ e Irr(tf). Suppose (£ - £(1)1H)G = 9 and [9, 9] = 1 + £(1)2. Show that there exists No G with N n, H = ker £, and every xeG — N conjugate to some element of H. (7.4) Let H < G and suppose induction to G is an isometry on Z[Irr(#)]°. Show that H is a Frobenius complement in G. (7.5) Let N £ G and y s Irr(iV). Assume that induction to G is an isometry on Z[^]° and \<f\ > 2. Suppose ^0 = {£ e ^ is extendible to G} ^0.
122 Chapter 7 Let d = g.c.d.(K(l)|^e^0}) and assume d|£(l) for all ZeSf. Show that y0 = y, y is coherent, and that every £,e y has a unique extension to G. (7.6) Let Pe Sylp(G) with \P\ = p, CG(P) = P, and |NG(P): P| = e. Show that G has at most e + (p — l)/e irreducible characters with degree not divisible by p. Hints Note that P is T.I.F.N. in G if e > 1. If e < p - 1, let S be the corresponding set of exceptional characters of G. Show Y^xes IzWI2 > p — e, where 1 ^ x e P. If *(x) = 0, then p | *(1). Note It is a consequence of some of R. Brauer's deep results in block theory that in the situation of Problem 7.6, the upper bound is always attained. (7.7) In the situation of the preceding problem, assume that G has exactly e + (p - l)/e irreducible characters of p'-degree. Let P = <x>. If / e Irr(G) is not exceptional (in particular if e = p — 1), show that /(x) = — 1, 0, or 1 and that /(l) = /(x) mod p. If e < p — 1 and $ is the corresponding set of exceptional characters, let a = Y.xssl{x)- Show that a = ±1 and x(l) = — eer mod p for / e <f. Note The results of Problem 7.7 together with the equation Z z(DzW = o Xelrr(G) and the facts that x(l)||G| and X^1)2 = \G\ provide a powerful tool for computing the degrees of the irreducible characters of a group which satisfies the hypotheses of Problem 7.6. (7.8) In the situation of Lemma 7.13, assume that \Sf\>2. Prove that *, e, and/are uniquely determined by (Sf, t). (7.9) Let K c G be T.I.F.N. Let / be a (possibly reducible) character of G which is constant on K - {1}. Assume that y = {\p e Irr(NG(X)) | K £ ker \j/} is coherent. Show that the multiplicity with which each exceptional character appears in / is proportional to its degree. (7.10) Let K be T.I.F.N. in G and assume that y = {xlie Irr(N(X))|X £ ker xfi} is coherent. Let &" be the corresponding set of exceptional characters of G. Let M = f)XEr ker /. Show that M n K = 1. (7.11) In the situation of Problem 7.10, show that either M = ker / for all X e F or else MK <a G.
Problems 123 Hint Suppose / e 2T and U = ker x > M. Then U/M is a rc-group where n is the set of prime divisors of | K | and so U £ MK. Use the Frattini argument. (7.12) Let K be T.I.F.N. in G and let e = |N(X):K\. Suppose that G has a faithful irreducible character of degree < 2e. Show that K is abelian. (7.13) In the situation of Problem 7.12, assume that K -?6 G. Show that K is an elementary abelian p-group. Hint Let k be a linear character of K with | K : ker 11 = p. Show that there exists faithful /e Irr(G) with /(l) < 2e and [/K, 1] ^ 0. Use Problem 7.11. (7.14) Let G be a simple Zassenhaus group in which the stabilizer of two points has even order e. Let the degree of G (the number of points permuted) be k + 1. Show that k is a prime power. Hint Count involutions to prove that G has a nonprincipal irreducible character of degree < 2e. Note Problem 7.14 is true without the assumption that e is even. This is a theorem of Feit and the case where e is odd is proved in his book. (7.15) Let N £ G and if £ Irr(JV). Suppose that v. I\_ifJ - Z[Irr(G)]° is a linear isometry. Assume that | if \ > 3 and that for every /, \p e if, we have (X(l)t/f - ii{\)tf = a9 - b<p for some <p, 9 e Irr(G) with a, be Z. Show that {if, t) is coherent. Hint Suppose /, \j/, r\ e if are distinct. Write (*( W - 4>{\)xy = a9- b<p, (n(\)xl> - ^W = Cfi - dv with a, b,c,d> 0. Then exactly one of 9 = y. or cp = v holds. (7.16) Let JV £ G, ^ £ lrr(JV) and suppose t: Z[^]° - Z[Irr(G)]° is a linear isometry. Let ifu if2^if such that if^ c\if2 ^0. Assume either that \SfY\ > 3 or that ifY = {\ji, /} with \p(l) = *(1). Show that (^! u if 2, t) is coherent. Hint Let/J: y, -► Irr(G) and e, = ± 1 be such that £;/; defines an appropriate extension of x on Z[^,]. If 1^1 > 3 for i = 1, 2, use the hint for Problem 7.15 to show that et = e2 and/^x) = /2(z) f°r Ke^i c\if2. (7.17) Let X £ G be T.I.F.N. with JV = NG(X) and ^={ZeIrr(N)|KgkerZ}. Suppose that \\i is an exceptional character of G corresponding to /. Show that the values of \ji and / generate the same field over Q.
124 Chapter 7 (7.18) Let Ku K2 c G be T.I.F.N. with Nt = NG(Kf) and ^{zelrrMIK.gkerz} and assume that the ^,- are coherent with St the corresponding sets of exceptional characters. Suppose SY n 82 *t 0- Show that St = S2 and Xt is conjugate to K2. Hint To show that K2 = K^, it suffices to show that x° e K2 for some xeKux¥= 1. Use the fact that the K( are nilpotent to prove this. (7.19) Suppose \G\ is odd and that C^x) is abelian for every 1 ^ xeG. Assume that G is neither abelian nor a Frobenius group and show that every nonprincipal / e Irr(G) is exceptional for some T.I.F.N. subgroup. Note A group with all C(x) abelian for 1 =£ x e G is called a CA-group. Problem 7.19 is a step in M. Suzuki's determination of all simple CA-groups. (7.20) Let K c G be T.I.F.N. with |X| even. Let N = NG(X) < G and y = {i/f e Irr(iV)|X £ ker i/f}. Assume that y is coherent. By carrying out the following steps, show that G has a nontrivial normal subgroup of odd order. (a) Show that G has a unique conjugacy class of involutions and that JV contains exactly e = | N: K | involutions. (b) Let il/eSf and let \ji* be the corresponding exceptional character of G. For 1 t6 x e K we have tA*(x) = et/^x) + a^, where a^eZis independent of x. Show that V2OA*) = ev2(W + (tfr*(l) - #(1) - <i,)/|K| where v2 is as in Theorem 4.5. (c) Show that (ip*)K = ipK + a^,lK. (d) Show that there exists \ji e & with ker(i/f*) t6 1. (e) Complete the proof. Hints (a) See the hint for Problem 4.11. (b) Let s/t = {xeG|x # 1, x2 = 1}; si2 = {x e GIx2 t6 1, x is conjugate to an element of K}; s/3 = G - (rfi u si2\ Compute £t/f*(x2) separately on each of s/u si2, and s/3. Note that (c) Let»v = (^*(l)-#(l)-fl#)/|K|. Use (b) to study the behavior of m^ as i/f varies over £f and conclude that m^ = 0. Use the fact that m^ is proportional to \ji(\). (d) Note that K is not elementary abelian.
Problems 125 Notes A corollary of Problem 7.20 is that a simple Zassenhaus group of odd degree has degree 1 + 2". This is another special case of Feit's theorem. There do exist simple groups with T.I.F.N. subgroups of even order. If G = SL(2, 2"), with n > 2, then the Sylow 2-subgroup K of G is T.I.F.N. It is elementary abelian of order 2". In this case, coherence fails because |N(X):X| = 2" - 1 and the set Sf contains only one character. If G = Sz(2") with odd n > 1 [the Suzuki simple group of order (2" - 1)(22")(22" + 1)], then again the Sylow 2-subgroup is T.I.F.N. Here, case (b) of Theorem 7.25 holds. (7.21) (Sibley) In the situation of Theorem 7.20, let e = \N:K\ and k = |K\. If m is as in 7.20 (b), show that -eKsfk + 1) < m < el(yfk - 1). In particular, if K is not an elementary abelian p-group, show that m = 0. Hints Let 1 =£ x e K and use the inequality |CG(x)|> 5>(x)iAW, ill bS where S is the set of exceptional characters. Derive that 2m + e > m2(k - \)/e. (7.22) Let K £ G be T.I.F.N. with N = N(K) and <?={xeln(N)\K<£kerx} coherent. Let \\i e Irr(G) with K £ ker \ji. (a) If \ji is nonexceptional, show that |X| < \ji(\) + 1 - [»/>*, 1K]- (b) If \ji is exceptional, show that | K \ < t/f(l)2. Hints (b) Reduce to the case that \jiN = / + A, where / e y and either A = 0 or K c ker A. If A ^ 0, appeal to Problem 7.21. If A = 0, pick an irreducible constituent, r\ =£ lGof\pij/ and apply part (a) to n.
Brauer's theorem There are many ways in which class functions of a group can be constructed. For instance, we could define 9(g) = |NG«g»| or 9(g) = 1 if g2 is conjugate to some fixed x e G, and 9(g) = 0 otherwise. It is perhaps too much to expect that such arbitrarily defined functions will turn out to be characters. (For instance, in the second example given, 9(1) = 0 if x # 1.) One might hope, however, that a well chosen 9 will be a generalized character, that is, 9 e Z[Irr(G)], or at least that it is an R-linear combination of irreducible characters for some specified ring R with Z £ R £ C. (We denote the set of these R-generalized characters by /?[Irr(G)].) How can one decide if a given class function 9 is an R-generalized character? Of course, if one knows Irr(G), the answer is easy: simply check that [9, %] e R for all / e Irr(G). A more typical situation is that one has enough information about some family 3f of subgroups of G so that it can be shown that 9H e /?[Irr(H)] for every H eJf. The main point of Brauer's theorem is that for suitable families 3f the last relationship is sufficient to guarantee that 9e/?[Irr(G)]. For instance, this will be true if Jf is the collection of all nilpotent subgroups of G. (8.1) definition Let R be a ring with Z £ R £ C and let ^f be a family of subgroups of G. (a) 3>tR(G, tf) is the set of class functions 9 of G such that 9H e R [Irr(H)] forallHe^f. (b) JR(G, Jtf) is the set of /^-linear combinations of characters t/^G for tAelrr(H), He#. 126
Brauer's theorem 127 If R = Z, we delete the subscripts and write 9t(G, M) and J(G, tf). (8.2) lemma Let W be a collection of subgroups of G and let R be a ring with Z £ /? £ C. Then (a) ./(G, JT) £ A(G, ■#") £ K[Irr(G)] £ ^«(G, Jf). (b) 3tR(G, Jtf) is a ring in which JR(G, Jf) is an ideal. (Addition and multiplication are pointwise.) Proof The containments in (a) are all obvious from the definitions. To prove (b) observe that Z[Irr(H)] is a ring for every He# Since Z £ R, it follows that K[Irr(H)] is a ring. If <p, 9 e ^R(G, ^f), then (<p% = <pH3He/?[Irr(tf)] and hence q>9 e 3%R(G, #?), which is, therefore, a ring. To prove that JR(G, df) is an ideal, we use the fact that if a is a class function of H £ G and /? is a class function of G, then (a/?H)G = aG/J. This is immediate from the definition of induction (and appears as Problem 5.3). Let cpeJR(G, Jf) and 9e@R(G, Jf). Then cp = YMefWanf with tA(H)e/?[Irr(H)]. Now <p3 = £ 0/WG9 = I (•/'(hA)6 Since \pm9HeR[ln(H)], it follows that (p9eJR(G,J^) and the proof is complete. | It follows from Lemma 8.2 that in order to prove that JR(G, W) = /?[Irr(G)], it suffices to show that 1G e J(G, Jff). Furthermore, if this can be done for some family Jf, it follows that JR(G, Jf) = /?[Irr(G)] = St AG, Jf) for all R with Z £ R £ C. (8.3) definition (Brauer) A group E is p-elementary (where p is a prime) if E is the direct product of a cyclic group and a p-group. We say that E is elementary if it is p-elementary for some prime. (8.4) theorem (Brauer) Let Z £ R £ C, where R is a ring. (a) A class function 9 of G is an R-generalized character iff BE e R[lrr(EJ] for every elementary E £ G. (b) Every / e Irr(G) is a Z-linear combination of characters of the form 1G for linear characters 1 of elementary subgroups of G. Let S be the set of elementary subgroups of G. Statement (a) of Brauer's theorem is exactly the assertion that /?[Irr(G)] = 3fcR(G, $). This part of the result is often called the "characterization of characters."
128 Chapter 8 Every E e $ is nilpotent and hence is an M-group by Corollary 6.14. If cp e Irr(£), then q> = XE, where X is a linear character of some E0 £ E. Since E0 e $ and cpG = 1G, it follows that J(G, S) is exactly the set of Z-linear combinations of XG for linear X e Irr(£) with E e i. Therefore, statement (b) of Brauer's theorem amounts to the assertion that Z[Irr(G)] = J(G, S). This is often called the "theorem on induced characters." By Lemma 8.2, both parts of Brauer's Theorem 8.4 will be proved when we show that 1G e J(G, i\ We work toward this goal. The proof given here is based on an idea of B. Banaschewski. It is suggested that the reader also consult the Brauer-Tate paper for another approach. (8.5) lemma (Banaschewski) Let S be a nonempty finite set and let R be a ring of Z-valued functions defined on S (with pointwise addition and multiplication). If the function ls with constant value 1 does not lie in /?, then there exists x e S and a prime p, such that p divides/(x) for every/ e R. Proof For each xeS, let Ix = {f(x)\f eR}. Then Ix is an additive subgroup of Z. If for some x e S, we have Ix < Z, then Ix £ (p) for some prime and the result follows. Assume then, thatIx = Z for every xeS. For each x, we may therefore choose/* e R with/*(x) = 1. Thus/* — ls vanishes at x and Ekes (/* — *s) = 0- Expanding this product yields an expression for ls as a linear combination of products of the functions fx. Thus ls e R. | To obtain a ring to which we can apply Lemma 8.5, we consider permutation characters of G, that is, characters of the form (1H)G for subgroups H£G. (8.6) lemma Let H, K £ G. Then (1H)G(1K)G = X av(\v)G for subgroups U £ H and integers av > 0. Proof Write 9 = (1K)G so that (1H)G(1K)G = 9(1 H)G = (9„f by Problem 5.3. By Lemma 5.14, 9 is the permutation character of G acting on the set of right cosets of K. It follows that 9H is a permutation character of H and hence 9a = Xfltf(li/)H for subgroups U £ H and integers av > 0. Since ((l vff = (l Dft the result follows. | (8.7) corollary The set of Z-linear combinations of characters of G of the form (1H)G is a ring P(G). Let ^f be a collection of subgroups of G with the property that if K £ H e JV, then Ke/. Let P(G, Jt?) denote the set of Z-linear combinations of characters of the form (1H)G with H eJf. Then P(G, tf) is an ideal of P(G). To apply Corollary 8.7, we define a class of groups more general than elementary groups.
Brauer's theorem 129 (8.8) definition A group H is p-quasi-elementary if H has a cyclic normal p-complement for the prime p. We say that H is quasi-elementary if it is p- quasi-elementary for some p. Note that subgroups of p-quasi-elementary groups are themselves p- quasi-elementary. Thus in the notation of Corollary 8.7, P(G, Jf) is closed under multiplication, where Jf is either the set of all quasi-elementary subgroups of G or is the set of all p-quasi-elementary subgroups of G for some fixed prime p. (8.9) lemma Let x e G and let p be a prime. Then there exists a p-quasi- elementary subgroup, H c G, such that (lH)G(x) is not divisible by p. Proof Let C be the p-complement in the group <x>. (Possibly C = 1 or C = <x>.) Let JV = NG(C), so that x e JV. Since <x>/C is a p-group, we may choose H/C e Sylp( JV/C) with <x> ^ h. Then C is a normal p-complement for H and H is p-quasi-elementary. Now (1H)G is the permutation character of G acting on the right cosets of H, and so (l„f(x) = \{Hy\Hyx = Hy,yeG}\. If Hyx = Hy, we have xyl e H and hence Cy~' c h. However, C is the unique p-complement in H and hence Cy~' = C and y e JV. We therefore need to count the number of fixed points in the action of <x> on the cosets of H in JV. Since C <i JV, and C £ H, we see that C is in the kernel of the action of JV on the cosets of H in JV. Since <x>/C is a p-group, it follows that the number of nonfixed cosets is divisible by p and hence (lH)G(x) = \N:H\ mod p. By the choice of H, pX\ N: H | and the proof is complete. | (8.10) theorem (L. Solomon) Let ^f be the set of quasi-elementary subgroups of G and Wp the set of p-quasi-elementary subgroups for some prime p. Then (a) lGeP(G,#). (b) mlG e P(GT ^f p) for some m e Z with pjfm. Proof By Corollary 8.7, P(G, Jf) is a ring of Z-valued functions on G. If 1G £ P(G, ^f), then by Lemma 8.5, there exists x e G and a prime p with p | <p(x) for all (p e P(G, JF). This contradicts Lemma 8.9 and so (a) is proved. For (b) let R = {<p + np\G\<p e P(G, Jt?p), n e Z}. Then R is a ring. If there exists x e G and a prime q, with <j | <p(x) for all (peR, then since plG e R, we have q = p and thus we have a contradiction to Lemma 8.9. By Lemma 8.5, we conclude that lGe/? and hence (1 — np)lGeP(G, ^fp) for some neZ. This completes the proof. | Only part (a) of Theorem 8.10 is needed to prove Brauer's theorem; however, part (b) is useful for certain refinements of the result. Recall that we
130 Chapter 8 need \GsJ(G, $). The method of proof is essentially to use 8.10 to reduce the problem to quasi-elementary groups. We need a lemma. (8.11) lemma Suppose G = CP where C o G, P is a p-group and p/HC|. Let 1 be a linear character of C which is invariant in G and suppose Q(P)£ ker 1. Then 1 = lc. Proof Let X = ker k. Since 1 is linear, it takes on distinct constant values on the distinct cosets of K in C. Since k is invariant under the action of P, it follows that P normalizes each coset Kx for xeC. In the conjugation action of P on Kx, the number of moved points is divisible by p. Since pJf\K\, we have pJf\Kx\ and thus Kx r\ CC(P) ^ 0. Since CC(P) £ X by hypothesis, we conclude that Kx = K and thus K = C and 1 = lc as claimed. | Proof of Theorem 8.4 As was remarked following the statement of the theorem, it suffices to prove that 1G e J(G, $). We use induction on | G |. We may therefore suppose that lHe<f(H, SH) whenever H < G, where SH is the set of elementary subgroups of H. Thus Z[Irr(H)] = J(H, SH) for these H by Lemma 8.2. By transitivity of induction (Problem 5.1), it follows for <peJ{H,SH), that <pG e J(G, SH) £ J(G, $). Thus for all H<G and <p e Irr(H), we have <pG e J{G, S) and it will suffice to show that 1G is a Z- linear combination of characters induced from proper subgroups. By Solomon's Theorem 8.10(a) we are done if G is not quasi-elementary and so we may assume that G has the cyclic normal p-complement C. Let P e Sylp(G) and Z = CC(P). Since we may clearly assume that G is not elementary, we have Z < C and E = PZ < G. Write (1£)G = 1G + E, where S is a possibly reducible character of G. We shall show that every irreducible constituent of S is induced from a proper subgroup and thus 1G = (1£)G - S e J(G, S) as desired. Let i be an irreducible constituent of S. Now CE = G and C n E = Z so that lc + Sc = ((1£)G)C = (lz)c by Problem 5.2. Thus 1 = E(lz)C, lc] = [lc + Sc, lc] and hence [Sc, lc] = 0. Thus [jc, lc] = 0. Let k be an irreducible constituent of %c. Then k ^ lc. However, Z <i G and so Z £ ker((l£)G) and hence Z £ ker /. We therefore have Z £ ker k and thus by Lemma 8.11, k is not invariant in G. Let T = IG(k) < G. By Theorem 6.11, / = \pG for some \ji e Irr(T). The result now follows. | The following is a useful special case of part (a) of Brauer's theorem.
Brauer's theorem 131 (8.12) corollary Let x be a class function of G. Suppose (a) Xe is a generalized character for every elementary E £ G; (b) [/,/] = 1; (c) Z(D>0. Then / e Irr(G). Proof Using (a) and Theorem 8.4(a), we have l = Y,az^ for <!; e Irr(G) and ai e Z. From (b) we conclude that at most one ai^=Q and / = + ^ for some £, e Irr(G). Then (c) yields the result. | Before going on to reap some of the harvest of applications of Brauer's theorem, we digress briefly to derive the "local" version of the result. In the Brauer-Tate proof, this is obtained as an intermediate result. (8.13) corollary Let Sp be the set of p-elementary subgroups of G for some fixed prime p. Then m\GeJ(G,Sp) for some m e Z with pjfm. Proof By Theorem 8.10(b), there exists m e Z with pjfm such that m 1G is a Z-linear combination of characters of the form (1H)G for p-quasi-elementary H £ G. Let SH denote the set of elementary subgroups of H. If H is p-quasi- elementary, then SH £ Sp and by transitivity of induction, we have (lH)Ge^(G,^H)£^(G,^p). The result now follows. | We shall use Brauer's theorem to remove the solvability hypothesis from Theorem 6.25 on extending characters. To do this we need part (c) of the following result. [Part (b) will be used in Chapter 13.] (8.14) lemma Let N<iG and let zelrr(G) with Xn = # e Irr(iV). Let \ji e Irr(G). For each coset Ng of JV in G compute r\(Ng) = (\/\N\)YJ -AWzW- xsNg We have (a) If [\pN, 9] # 0, then i\ e Irr(G/N). (b) If !>„, 9] = 0, then tj(Ng) = 0 for all g. (c) If x = iA, then rj(Ng) = 1 for all g.
132 Chapter 8 Proof Conjugates of Ng in G/N all have the form Ngh = h~l(Ng)h for h e G. It follows that r\ is a class function on G/N. We may thus write r\ = £ b^, where £ runs over Irr(G/JV) and ft,, e C. Now *>< = [*. fl = TTt^t I *NgWg) |G.JV| NS6G/jv l"l lefi where in the last expression, we view t, e Irr(G) and use the fact that as such, ^ is constant on cosets of JV. We thus have b4 = \j//, /<!;]. Now by Corollary 6.17, x£ e Irr(G) and the characters x£ are distinct for distinct <!; e Irr(G/JV). Therefore r\ = Y.'b^ satisfies rj = 0 unless \\i = x£ for some £, e Irr(G/JV) in which case r\ = £. By Corollary 6.17, \ji = x£ for some £ iff [t/^, 9] t6 0. The result now follows. | (8.15) theorem (Gallagher) Let JV o G with ,9 e Irr(JV) invariant in G. Suppose (9(1), |G : N\) = 1. Then 9 is extendible to G iff det .9 is extendible. Proof The "only if" part of the assertion is trivial. Assume X = det 9 is extendible to \i e Irr(G). If JV s H £ G with H/JV solvable, then 9 is extendible to H by Theorem 6.25. By Lemma 6.24, there is a unique extension X(H), such that det(/(H)) = fiH. Define the function / on G as follows. If g e G, let H = <JV, g} so that H/JV is cyclic. Set x(g) = X(H)(q)- We shall use Corollary 8.12 to show that / e Irr(G). First we establish that / is a class function. If g, x e G, let H = <JV, g}, so that Hx = <JV, g*>. Define \p on H* by i/f(/i*) = x(H)(/i) for /i e H. Since the map /i h» A* is an isomorphism and /(H) e Irr(H), we have \p e ln(Hx). Also, (deti/f)(/i*) = (det(/(H)))(/i) = /z(/i) = fi(hx), so that det i// = /zH*. Furthermore, if neJV, we have \ji(n) = Xml"* ) = #(«*') = 9x(n) = 9(n). It now follows from the uniqueness of z(H*,, that x(H*) = \\i and hence *(»*) = Xw*)(9x) = ^(9X) = Xwis) = X(g) and / is a class function. Next, let E s G be elementary and set H = JV£ so that H/JV s £/(JV n E) is nilpotent and x(H) is defined. We shall show that the restriction Xh = Xan and thus Xe = (X{H))e which is a character of E. If g e H, then <JV, g> = K £ H. Clearly, (x(H))K is an extension of 9 to K and det((/(H))K) = fiK so that (x(H))K = X(Ky Therefore, xio) = fojfo) = X<hM hence la = Xw as claimed, and Xe is a character. Next, we compute [/, /]. For each coset Ng of JV in G, we have I IzWI2 = I l/(H,WI2,
Brauer's theorem 133 where H = <JV, g} = <JV, x> for all xeNg. Since (x(H))n = 9eIrr(iV), Lemma 8.14(c) yields £„„, lx(H)(x)|2 = |JV| and it follows that X«g IzWI2 = |G:JV|.|JV| = |G|andrjf,Z] = l. Finally, /(l) = 9(1) > 1 and hence / e Irr(G) by Corollary 8.12. Clearly, %N = 9 and the proof is complete. | (8.16) corollary Let JV <a G and 9 e Irr(iV) with 9 invariant in G. Suppose (|G: JV|, o(9)9(l)) = 1. Then 9 has a unique extension, /e Irr(G) with (| G: N |, oft)) = 1. In fact, oft) = o(9). In particular, this holds if (| G: N |, | N |) = 1. Proof This is immediate from Corollary 6.27 and Theorem 8.15. (See the proof of Corollary 6.28.) | We have already seen several conditions sufficient to guarantee that a character value is zero. The following is a powerful one. (8.17) theorem Let /elrr(G) and suppose pjf( | G|/z(l)) for some prime p. Then %(g) = 0 whenever p | o{g). Proof Define 9 on G by 9(g) = x(g) if pJfo(g) and 9(g) = 0 if p | ofcr). We shall use Brauer's theorem to show that 9 is a generalized character. Let £cGbe elementary. Since E is nilpotent, we may write E = P x Q, where P is a p-group and pjf\ Q \. If x e E and p^x), we have x e Q and so 9 vanishes on E — Q and 9Q = %Q. Let t/f e Irr(£). We have IE| [9£, tfr] = £ z(x)iAW = |Q| [ft, ^]. xbQ Since | £ | = | P11Q |, we conclude that | P | [9£, i//~\ e Z. Now let co = <yz be the algebra homomorphism Z(C[G]) -» C associated with / so that co(X) = x(g)| Jf |//(1), where Jf is the conjugacy class containing g and K = EJf e C[G]. We shall write co(gr) for the algebraic integer co(K\ so that X(g) = z(l)co(g)/|X| = x(l)co(g)|CG(g)|/|G| Now |£| [S*. -A] = Z zWiAW = t^t Z co(x)lA(x")|CG(x)|. Since P Q CG(x) for x e Q, we have ^~ [S*. *] = -J^jjfl [9£, £| = £ co(x)i/Kxl|CG(x): P|,
134 Chapter 8 which is an algebraic integer. Since \P\ [9E, \p~\ e Z, we have [S£, \p~\eQ and thus (| G11Q | /z( 1)) [9E, tA] e Z. Since | G11Q | //( 1) e Z and is relatively prime to |P|, we conclude that [$£, i/f] e Z and hence $£ is a generalized character of£. Brauer's theorem now yields that 9 is a generalized character of G. In particular [9, z]eZ. Now [9, z] = (1/|G|) £ {lz(0)l2lpM0)}, and so 0 < [3, *] < [/, /] = 1. We conclude that [9, /] = [/, /] and 0 = [(z-.9),z]=^X{lz(9)l2lp|o(g)}. The result now follows. | A character i e Irr(G) is said to have p-defect zero if p does not divide |G|//(1). In Chapter 2 we discussed the question of what information can be obtained about a group from a knowledge of its character table. Since D8 and Q8 have the same table, one cannot determine the orders of the elements in the various classes. Nevertheless, it is possible to determine the sets of prime divisors of these orders. To do this, we shall apply Theorem 8.4(a) in a situation where R =£ Z. If geG, we shall use the notation n(g) to denote the set of prime divisors ofofef). (8.18) lemma Let geG and let n be a set of primes. Then there exist uniquex,yeG with (a) g = xy = yx; (b) n(x) ^ n and n(y) n % = 0. Furthermore, x, y e <g>. Proof Write o(g) = mn such that every prime divisor of m is in n and no prime divisor of n is in n. Then (m, n) = 1 and we have km + In = 1 for some k, leZ. Let x = gin and y = gkm so that g = xy = yx and x, ye<g>. Since xm = 1 = /, (b) follows. Now suppose g = uv = vu with n(u) ^ n and n(v) r\n = 0. Then u c C(g) c C(x) and hence n(xu~l) c n(x) u 7t(w) c n. Similarly, n(y~vv) c 7t(t>) u n(y), so that 7r()>~ lv) n 7t(xu~') = 0. Since ^_1t> = xw_1, we have xw_1 = 1 = y~ 'u and uniqueness follows. | If g, n, x, and y are as in Lemma 8.18, we write x = g„ and y = g„,. If n = {p}, we write x = gp and y = gp,.
Brauer's theorem 13S (8.19) lemma Let | G | = mn with (m, n) = 1 and let n be the set of prime divisors of m. Let g e G, with g = #„-, and define 9 on G by ,9(x) = {n 'f *"' iS conJu8ate to & (0 if xK> is not conjugate to g. Let /? be the ring of algebraic integers in Q[e], where e is a primitive nth root of 1. Then 9 is an /^-generalized character. Proof Let E £ G be elementary. It suffices to show that 9E e /?[Irr(£)]. Since E is nilpotent, we may write E = H x K, where H is a rc-group and K is a rc'-group. Then for xe E, we have x = uv,ue H,veK and thus w = x„ and v = x„.. If g is not conjugate to any element of X, then 9£ = 0 and there is nothing to prove. Suppose then, that gug2,...,graKths distinct elements of K which are conjugate to g. Then the elements xeE with x„. conjugate to g are exactly the elements of the cosets Hgh 1 < i <r. Let i/f e Irr(£). By Theorem 4.21, we have \ji = £ x « for some £ e Irr(H) and w e Irr(K). We have IE | [3£, *] = £ £ 9(ugtWUg~d = n £ ^) £ W\ i- 1 ueH i= 1 usH Since £„eH £(m) = 0 unless £ = 1H, we have either [9E, i/f] = 0 or |£|[9£,.A] = n|H|£^7). i= 1 Now |£| = |X||H|and |K ||n and thus [S£,i/f] is a multiple of £ «(#() which lies in R. The result now follows. | (8.20) theorem Let R be the ring of algebraic integers in Q[e], where e is a primitive |G|th root of 1. Let pe Z be a prime and let / be a maximal ideal of R with pel. Let x, yeG. Then xp. and yp. are conjugate in G iff X(x) = X(y) mod J for every / e Irr(G). Proof Write \G\ = p"n, with p^n, and define 9 on G by 9(g) = n if gP' is conjugate to xp. and 9(g) = 0, otherwise. By Lemma 8.19, 9 is an R-generalized character of G. Suppose /(x) = %(y) mod / for every / e Irr(G). It follows that 9(x) = 9(y) mod /. Now 9(x) = n. If 9(y) = 0, then n = 0 mod / and n e I. Since pel and p^n, it follows that 1 = (p,n)el and this is a contradiction. Thus 9(y) = n and hence yp* is conjugate to xp. as desired. Conversely, we may suppose xp. = g = yp- and we show that /(x) = %(g) = x(y) mod /. Write x<x> = £ 1;, where the X{ are linear characters. Since g e <x>, it suffices to show that X(g) = Mx) mod / for linear 1.
136 Chapter 8 Now x = gh, where h = xp e <x> and l(x) = l(g)l(/i). Let l(/i) = 8 so that (5pm = 1 for some m. For a e R, let a* denote its image in R/I which is a field of characteristic p. We have 0* = (8*ym - 1* = (8* - l*)pm and thus ,5* - 1* = 0* and 8 = 1 mod /. It follows that l(x) = Mg) mod / and the result follows. | Note that by Theorem 8.20, it follows that if xp- and yp. are conjugate and X e Irr(G), then /(x) — %(y) lies in every maximal ideal of R which contains p. (8.21) theorem (G. Higman) Let X Y, Jf2, ..., Jft be the conjugacy classes of G. Then the character table of G determines the sets of primes n^ = n(g^ lor g^X^. Proof Let Irr(G) = {/; 11 < i < k}. We are given the complex numbers ain = XiiSii) f°r 0/ie -*V ^ 1 e •%"*>tnen v is tne unique integer 1 < v < k, such that aiv e Z and a,v > \aifl\ for all i, /z and hence we know nv = 0. For notational simplicity, we now assume 1 e jf t. Next we compute £f (aa)2 = |G| and let /? be the ring of algebraic integers in Q[e], where e is a primitive |G | th root of 1. For each prime, p11G |, we choose a maximal ideal of R, Ip 2 pR. Construct the equivalence relation ~p on {/i\ 1 < fi < k} by setting fi ~p v if a;/i = aiv mod 7P for all i, 1 < i < k. By Theorem 8.20, /z ~p v iff the elements of Jf ^ and Jfv have conjugate p'-parts. In particular, if (i~pv, then n^ u {p} = 7tv u {p}. Furthermore, given fi, there exists v, with p$nv, such that /z ~ p v. To see this, take g e JT^ and choose v so that gp, e Jf v. Let n be a set of prime divisors of G. Using induction on | n \, we show how to construct the set y, = {/zl^ = ?c}. We have ^0 = {1}. If 7t j= 0, write 7t = 7t0 u {p}, where p$n0. By the preceding remarks, it follows that ■5^= {\i\\ii^no and /z~pv for some vg^J. The proof is now complete. | Transfer theory is a tool commonly used for producing normal subgroups of a group, especially normal p-complements or normal 7t-comple- ments. Many of these transfer theorems can also be proved using characters, and in particular using Brauer's theorem. An example of a typical transfer theorem is the following: Let H be a Hall subgroup of G (that is, (| H \, \ G: H \) = 1) and suppose that H is nilpotent and that if x, y e H are any two elements which are conjugate in G, then x and y are already conjugate in H. Then H has a normal complement in G. In the above situation, let n be the set of prime divisors of | H \. Using the assumption that H is nilpotent, it is not too hard to prove that if U £ G is a
Brauer's theorem 137 nilpotent rc-subgroup, then U is conjugate to a subgroup of H. Because of this, the following result is a generalization of the above transfer theorem. (8.22) theorem (Brauer-Suzuki) Let H be a Hall subgroup of G and suppose that whenever two elements of H are conjugate in G, then they are already conjugate in H. Assume for every elementary subgroup E £ G, that if | E111H |, then E is conjugate to a subgroup of H. Then there exists K<G with HK = G and H n K = 1. Proof Let n be the set of prime divisors of |H|, so that by hypothesis, every elementary rc-subgroup of G is conjugate to a subgroup of H. Let 9 be a class function of H. We define a class function 5 on G as follows. For g e G, <#„> is an elementary rc-group and hence g„ is conjugate to some element x e H. Set §(g) = 9(x) and observe that this is well defined since if g„ is also conjugate to y e H, then x and y are conjugate in H by hypothesis and so 9(x) = %). For x e H, let /x(x) denote the number of elements g e G with g„ conjugate to x. Let xu x2 x, be representatives for the conjugacy classes of H. If 9 and cp are class functions of H, we have (*) & 9\ = yir t SixMxdiAxd. \G\ (=i Now let R be a ring with ZsKcC and suppose that 9 is an /?-gener- alized character of H. We claim that S is an /^-generalized character of G. To prove this, let E £ G be elementary. We show that (5)£ is an /^-generalized character of E. Write E = U x V, where U is a 7t-group and V is a 7t'-group. We may assume that U £ H. If w e 1/ and veV, then (w)„ = w and hence §(uv) = 9(u). Since U £ H, we may write 9^ = £ a^ for t/f e Irr(t/) and a^ e /?. It follows that (9)£ = 5> fli/rC/' x V)> which is an R-generalized character of E. Thus 5 is an R- generalized character of G by Brauer's theorem as claimed. Now let R be the ring of algebraic integers in Q[e], where e is a primitive |H|th root of 1. For 1 < i < t, define 9, = £V€irr(H) q>(x^q>, so that 9;(xj) = 0 if i t6 j and 9;(xj) = | CH(x.) | by the second orthogonality relation. Also, 9, is an /^-generalized character of H and hence [5;, 1G] e R and [S«. lG] = ^ i W/4*j) = Mx'),|Cf(Xi)l. luU=i |G| We conclude that /x(x,)|CH(x,)|/|G| is a positive rational number which is an algebraic integer. It follows that it is a positive rational integer and thus /Axd\C^x,)\/\G\ > 1
138 Chapter 8 and fi(x{) > |G|/|CH(x,)| for all i. We now have Therefore, equality holds throughout and /z(xj) = |G|/|CH(x,)|. Thus Equation (*) yields Now suppose / e Irr(H). By taking R = Z, we conclude that % is a generalized character of G. Since [J?, %~] = \j, /] = 1 and %(\) = x(\) > 0, it follows that % e Irr(G). Let K = fifkerOOIZ e Irr(H)}. Then X -c G. If u e H n X, then z(w) = jf(M) = jj(l) = /(I) for all /elrr(H) and thus u = 1. On the other hand, if P^tc, PeSylp(G), and veP, then t>„ = 1 and %(v) = z(l) = x(l) for all /elrr(H), so that t>eX. Thus P s K and hence |X| > \G:H\. It follows that HK = G. | To demonstrate the power of Theorem 8.22, we show how easily Burn- side's transfer theorem follows from it. (8.23) corollary Let P e Sylp(G) and suppose P £ Z(N(P)). Then there exists K~a G with KP = G and K r\ P = 1. Proof If E £ G is a p-group, then E is conjugate to some subgroup of P by Sylow's theorems. It therefore suffices to assume x, y e P are conjugate in G and show that x and y are conjugate in P. We have then, y = xa for some g e G. Since P is abelian, P £ Cg(j>) and Pg £ CG(x9) = CG(y). Thus P»c = P for some c e CG(x9) and hence gc e N(P) £ C(x) by hypothesis. Now y = x° = (xe)c = x°c = x and the proof is complete. | We also remark that Theorem 8.22 provides an alternate proof of Frobenius' Theorem 7.2. It is routine to check that a Frobenius complement necessarily satisfies the hypotheses of Theorem 8.22. Our next results depend on Corollary 8.13 where we restrict attention to p-elementary subgroups for some fixed p. (8.24) theorem (Dade) Let N <a G with G/N a p-group. Let 9 e Irr(iV) be invariant in G. Then there exists a p-elementary subgroup E £ G, and
Brauer's theorem 139 <p e Irr(£ n JV), such that (a) NE = G; (b) <p is invariant in E; (c) [9£nJV,<p]isprimetop. Proof By Corollary 8.13, we have (*) m\G = £ a^G, where \p runs over irreducible characters of p-elementary subgroups, a^ e Z and p^m. If £ is p-elementary and i/f e Irr(£), we compute [/?, (i/^n] for G-invariant class functions p of JV. Now (/?<% = \G:N\P and (j3£JV)n = \EN:N\p. Since 0G and 0£JV both vanish on £JV - N, we have (0G)£JV = |G:£JV|0£Jv.Now where the latter equality follows since \jiG = (\liEN)G. This yields [fi, W%] = \G:EN\IPEN, rNl = \G:EN\[fi, (*"%] = \G:EN\[p,WEnN)Nl where we have used Problem 5.2 to obtain the last equality. We apply this with /? = 95. For each \ji in Equation (*), write \ji e Irr(£^) for p-elementary £^ s G. We obtain m = [95, my = £ a^[95, (,/,%] = 5>*|G:JVEJ[8S,WrE,nN)w]- Since p^m, there exists ^ such that p does not divide | G : NE | [95, (i/f£ n A,)iV], where we have written E = £^. Since G/N is a p-group and pjf\ G: JVE |, we conclude that NE = G, and (a) follows. Now write L = E r\ N so that [55, (^f] = [$L§L, tfrj = [9L, ^L9J is prime to p. Since 9 is invariant in G we have 9L is invariant in £ and we may write 9L = £A eA A, where A runs over sums of orbits of the action of E on Irr(L) and eA e Z. Write y = \\iL 9L so that y is invariant in £. We have [A. Ml] = 5>aIA 7] A is prime to p. Choose A such that p/feA[A, }>] and write A = <pt + • • • + <p„ where the <pt e Irr(L) and are conjugate under £.
140 Chapter 8 Since y is invariant under E, all [<p;, y] are equal and [A, y] = t\_<pu y]. However t is a divisor of|£:L| = |G:JV| and so is a power of p. Since pjf\_\ y], we have t = 1 and <pt = q> is invariant in £. This is (b). Finally, [#l> <p] = eA is prime to p and the proof is complete. | An interesting special case is when N = G. (8.25) corollary Let x £ Irr(G) and let p be a fixed prime. Then there exists p-elementary E £ G and <p e Irr(£) with [jE, <p] ^ 0 mod p. Examples exist which show that in Corollary 8.25, E cannot always be taken to be a p-group. Dade's Theorem 8.24 has a nice application to the question of character extendibility. (8.26) theorem Let JV -a G with G/N a p-group. Let P e Sylp(G) and assume that P n JV £ Z(P) and that every linear character of P r\ JV is extendible to P. Then every G-invariant irreducible character of JV is extendible to G. Proof Let 9eIrr(JV) be invariant in G. By Theorem 8.24, choose a p-elementary subgroup £ £ G, with £JV = G, and q> e Irr(£ n JV) such thatp^rA^.p], We have E = Q x C, where Q is a p-group and C is cyclic. We may assume that C is a p'-group so that EnN = (QnN)xC and we write (p = a. x X where a eIrr(Q n JV) and 1 elrr(C). We may assume Q £ P so that Q n, N ^ P n N, which is abelian. It follows that a is extendible to P n JV and thence to P by hypothesis. Therefore a has an extension a e Irr(Q) and <p = a x A is an extension of <p to £. Now since EN = G, we have (<pG)N = (q>EnN)N = 9N and hence which is prime to p. Therefore, there exists an irreducible constituent / of cpG with pJf\_XN, #]• Since 9 is invariant in G, it follows that xn = e& with pjfe. Now e divides | G: JV | by Problem 6.7 and since G/N is a p-group it follows that e = 1 and / extends 9. (Note that Problem 6.7 follows immediately from Corollary 6.19 by induction.) | We remark that in the notation of Theorem 8.26, the hypothesis on P will automatically be satisfied if P is abelian. More generally, it suffices for F n JV = 1. It has been mentioned that if JV -a G and 9 e Irr(JV) is invariant in G, then it suffices to show that 9 is extendible to the inverse images in G of the Sylow subgroups of G/N in order to prove that 9 is extendible to G. It is for this reason that it is useful to obtain results like Theorem 8.26 which give sufficient conditions for extendibility when the factor group is a p-group.
Problems 141 Problems (8.1) Let Z £ R £ C, where R is a ring and the additive group (Z, +) is a direct summand of (R, +). Let W be a family of subgroups of G. Show that A(G, JT) n Z[Irr(G)] = ./(G, jf). Hint Write /? = Z 4- S. If 9 = £ a^G, where all a^, e S, then [,9, *] e S for all / e Irr(G). Note The Brauer-Tate proof of Brauer's theorem uses this fact applied to R = Z[e], where e is a root of unity. (8.2) Let s# be the collection of abelian subgroups of the group G. Show that the following statements are equivalent: (a) \GeJ(G,sf)\ (b) «(G,j/) = Z[Irr(G)]; (c) every Sylow subgroup of G is abelian. Hint If a Sylow p-subgroup of G is nonabelian, consider the function 9 = (l/p)p, where p is the regular character of G. (8.3) Let i be a character of G. For g e G and prime p, write g = gpgP' in the notation introduced following Lemma 8.18. Define the class function xP by Ipig) = X(9P\ Show that (xp)N is a character for every nilpotent JV £ G. Show that if G is not nilpotent then xP 1S not a character of G for some X e Irr(G) and some prime p. Note In fact, xP is a character for every / e Irr(G) iff G has a normal Sylow p-subgroup. Of course, by Brauer's theorem, xP is always a generalized character. (8.4) Let x and \ji be characters of G and suppose \G\ = mn, where (m, n) = 1. Let where the sum is taken over those g e G such that o(g) \ m. Show that a/m e Z. (8.5) Let /elrr(G) and suppose |G| = mn with (m, n) = 1. Assume that X(x) = 0 for all 1 # x e G such that x" = 1. Suppose y e G and /" ^ 1. Show that x(y) = 0. (8.6) Let G and H be groups with classes Jf, and if, respectively and irreducible characters /, and t/^, respectively. Assume that whenever gejft and /i e if, we have x,<0) = \jifti) for all i and j. (In short, G and H have identical character tables.) Let x e Z(P) where P e Sylp(G). Suppose x e Jf,. Let y e if (. Show o(x) = o(y).
142 Chapter 8 Hint If o(y) > o(x) and z e H, with <y> = <z> and yp = zp, show that z e if, and conclude that p 11 if,-1. See the hint for Problem 2.12. (8.7) Suppose G and H have identical character tables and that G is solvable and has an abelian Sylow p-subgroup. Show that G and H have isomorphic Sylow p-subgroups. Hint If G has no nontrivial normal p'-subgroup, then its Sylow p- subgroup is normal. Use Problem 8.6. (8.8) (Dade) Let N <a H £ G. Assume the following: (a) (|tf:JV|,|G:tf|)=l. (b) If x, x9eH for some g eG, then there exists heH and neN with x9 = xhn. (c) If £ £ G is elementary and (|£|, |G:H|) = 1, then EB £ H for some 3 e G. Show that there exists M <i G such that HM = G and H r\M = N. Hint Mimic the proof of Theorem 8.22. Consider only those class functions 9 of H which are constant on cosets of JV. Note Many of the important "transfer" theorems follow from Problem 8.8, for instance, the Focal Subgroup Theorem: LetPeSylp(G)andletJV = {x~lxg\xeP,xBeP,g eG>.Then AP(G) n P = N. Here, AP(G) denotes the normal subgroup of G minimal such that the factor group is an abelian p-group. Problem 8.8 yields M <i G with MP = G and M r\ P = N. It follows that M 2 AP(G). The reverse inclusion follows from N £ G' £ AP(G). (8.9) Show that a group G is quasi-elementary iff 1G cannot be written in the form £ %(1h)g for proper subgroups H, with aH e Z. Hint For "only if": Assume 1G = £ aH(lH)G, where the H run over representatives for the conjugacy classes of proper subgroups. Let C be a cyclic normal p-complement for G. Choose H0 minimal, such that pJf(aHo\G:H0\), and let 1 be a faithful linear character of C/(C r\ H0). Consider the numbers [(aH(lH)G)c, 1]. Note Problem 8.9 essentially says that Theorem 8.10(a) is the best possible. (8.10) Show that the integers m in Theorem 8.10(b) and Corollary 8.13 can be taken to be the p'-part of | G |.
Problems 143 (8.11) Let x £ Irr(G) be a character of defect zero with respect to the prime p. Let Q^G, with pJf\Q\ and let PeSylp(CG(Q)). Show that (l/|P|)xQ is a character of Q. (8.12) Let P oG, where P is a p-group, and let / e Irr(G/P) be of defect zero in G/P with respect to p. Let 9 e Irr(P) be invariant in G. Define the function \ji on G by #7) = 0 if g„{P *l>(g) = HgP)x(gP) if gP£P- (a) Show that \ji is a generalized character of G. (b) If G = PCG(P), show that ip e Irr(G). (8.13) Let N o G be a Hall 7t-subgroup. Define the function v on N by v(x) = | {y e CG(x)\y is of 7t'-order} |. Show that v is a generalized character of N. Hint If 9 is a G-invariant character of N (which is not necessarily irreducible), define 5 on G by kg) = %*)• Show that 9 is a generalized character of G and compute [9, 1G].
9 Changing the field So far we have restricted our attention almost exclusively to characters, representations, and modules over the complex numbers. In this chapter we digress from our study of the properties of Irr(G) to consider irreducible group representations over arbitrary fields. (In prime characteristic, this falls far short of the general case, since not all representations are completely reducible.) In particular, if F £ E is a field extension we explore the connections between the irreducible ^-representations and F-representations of a group. In prime characteristic, we shall see that this situation is (surprisingly) under somewhat better control than it is in characteristic zero, which will be more fully considered in Chapter 10. Let F c E and let 3E be an F-representation of a group G. Then 3E maps G into a group of nonsingular matrices over F that, of course, are also non- singular over E. We may, therefore, view 3E as an F-representation of G. As such we denote it by 3E£. (The superscript thus merely indicates a change in point of view.) If 3Et and 3£2 are similar F-representations, then 3Et£ and 3E2£ are similar and it follows that if 3E corresponds to the F[G]-module V, then there is a uniquely defined (up to isomorphism) £[G]-module VE that corresponds to 3E£. (For those readers familiar with tensor products, we remark that VE = V ® FE.) We shall not, however, need to refer to VE again, since it is usually easier to work with the representation 3E£. The F-representation 3E may be extended by linearity to obtain a representation of F\_G~\, which we shall continue to call 3E. Under this convention, the £[G]-representation 3E£ is an extension of the F[G]-represen- tation 3E. 144
Changing the field 145 0 1 0 0 1 0 0 0 0 0 0 1 °\ o\ -7 0/ X(b) = 0 ' 0 1-1 ^ o 0 0 0 -1 1 0 0 0 0^ 1 0 0 If 3E£ is irreducible, then clearly so is 3E. However, 3E£ may well reduce, even if 3E is irreducible. To illustrate what can happen, we consider two examples, both for the field extension RcC. Let G = <g> be cyclic of order 3 and let 3E be the R-representation defined by Then 3EC affords the character k + I, where 1 is a faithful linear character of G and so 3EC is reducible. Since k is not real valued, it is not afforded by any R-representation and it follows that 3E is irreducible over U. Now let QB = <a, b} be the quaternion group of order 8, where o(a) = 4 = o(b). Define X(a) = It is not hard to check that this defines a representation of Q8. (In fact this is the representation whose module is the quaternion algebra H = IR + Ui + Uj + Uk with respect to the basis {1, i,j, k}, where a = i and b = j act by right multiplication.) Now 3EC affords the character 2/, where / e Irr(Q8) and X(l) = 2. By Problem 2.5(b), / is not afforded by any R-representation of QB and thus 3E is irreducible over U. (This can also be deduced from the fact that H is a division algebra.) (9.1) definition Let 3E be an F-representation of G. Then 3E is absolutely irreducible if 3E£ is irreducible for every field £2F. (9.2) theorem Let 3E be an irreducible F-representation of G with degree n. The following are equivalent. (a) 3E is absolutely irreducible. (b) 3E£ is irreducible for every finite degree extension E 2 F. (c) The centralizer of 3E(G) in the matrix ring Mn(F) consists of scalar matrices. (d) X(FlG-]) = Mn(F). Proof That (a) implies (b) is trivial. Now assume (b) and let M e M„(F), with MX(g) = H(g)M for all geG. Let £ be a finite degree extension of F, chosen so that M has an eigenvalue, keE. Since 3E£ defines an irreducible representation of £[G] and M - kl is a singular matrix centralizing its image, it follows from Schur's Lemma 1.5 that M - kl = 0. Thus (c) follows. That (c) implies (d) is immediate from the Double Centralizer Theorem 1.16. Finally, assume (d) and let L s F. Since every M e Mn(L) is an L-linear
146 Chapter 9 combination of matrices in Mn(F), it follows that 3E(L[G]) = Mn(L). Thus for every L representation 9) similar to 3E, we have 9)(L[G]) = M„(L) and so 9) cannot be in reduced form. Thus 3E is irreducible over L and the proof is complete. | (9.3) definition The field F is a splitting field for G if every irreducible F-representation of G is absolutely irreducible. (9.4) corollary If F is algebraically closed, then F is a splitting field for every group. Proof Since F has no proper finite degree extension, condition (b) of Theorem 9.2 holds for every irreducible F-representation of G. | Suppose F is a splitting field for G and E 2 F. Then every irreducible F-representation 3E determines an irreducible F-representation 3E£. If {£,} is a set of representatives for the similarity classes of irreducible F-represen- tations of G, then {3E,£} is a complete set of representatives for the irreducible £-representations of G. In order to prove this, we need to discuss the " irreducible constituents " of a possibly reducible representation. If V is an F[G]-module, then a composition series for V is a chain of submodules V=V0>Vl> — >Vn = Q such that each Vt_ JVj is an irreducible module. The modules V{_ JVi are the factors of the series. The Jordan-Holder theorem asserts that the factors of any two composition series for V are the same up to isomorphism (and counting multiplicities). An irreducible module W isomorphic to a factor of some (and hence all) composition series for V is called an irreducible constituent of V. If W is isomorphic to an irreducible submodule of V, then W is a constituent of V. We call it a bottom constituent in that case. Similarly any irreducible homomorphic image of V is a constituent. These are the top constituents of V. If V is completely reducible [e.g., if char(F)/|'|G|], then every irreducible constituent of V is both a top and a bottom constituent. (Caution: the converse is false.) All of the preceding remarks may be translated into the language of representations. If 3E is an F-representation of G corresponding to the F[G~\- module V, then 3E is similar to a representation 3 m triangular block form /3ite) * \ \ o ' 3„(0)/ where the irreducible representations 3; correspond to the factors Vi_1/Vi.
Changing the field 147 Representations similar to the 3; are the irreducible constituents of 3E; those similar to 3i being top constituents and those similar to 3„ being bottom constituents. (Of course some of the other irreducible constituents of 3E may also be top or bottom constituents since 3 is not necessarily the only representation in triangular block form to which 3E is similar.) Note that the character afforded by a representation is the sum of the characters of all of the irreducible constituents (counting multiplicities). We emphasize that by the Jordan-Holder theorem, the representation 3E has only finitely many irreducible constituents (up to similarity), namely those which appear in any single triangular block representation similar to 3E. (9.5) corollary Let F be a field and G a group. (a) Every irreducible F-representation of G is a top constituent of the regular F-representation 5R. (b) There exist only finitely many similarity classes of irreducible F-representations of G. (c) If E 2 F and 9) is an irreducible F-representation of G, then 9) is a constituent of 3E£ for some irreducible F-representation 3E. Proof Statement (a) is immediate from Lemma 1.14 and (b) follows from (a) via the Jordan-Holder theorem. Let 3 be any F-representation of G. The irreducible constituents of 3£ rnay be found by taking the irreducible constituents 3E; of 3 and then finding the irreducible constituents of the 3E,£. Now (c) follows by applying this remark to 3 = 5R. | The following result is a useful tool for establishing the similarity of two F-representations of G and for other purposes. (9.6) theorem Let 3E be an irreducible representation of F\_G] and let a e F[G]. Then there exists b e F[G1 such that X(b) = 3E(a) and 9)(b) = 0 for every irreducible F[G]-representation 9) which is not similar to 3E. Proof Let {3E,} be a set of representatives for the similarity classes of irreducible F[G]-representations. Let It= {xe F[G] |3E,(x) = 0} so that J{ is an ideal of F[G] and in the notation of Problem 1.4, J(F[G]) = f)/{. Let A = F\_G]/(f]Ii), so that each 3E( may be viewed as a representation of the algebra A. As such, the 3E,- are irreducible and pairwise nonsimilar. In particular, J(A) = 0 and by Problem 1.5, A is semisimple. By Theorem 1.15, A has minimal ideals M{, such that X/M,) = 0 if j j= i and 3E,(Mj) = 3E,(y4) = 3Ej(F[G]). Now suppose 3E = 3Et. Choose b in the inverse image in F\_G] of Mt with X(b) = 3E(a). The result follows. | (9.7) corollary Let 3E and 9) be irreducible F-representations of G. Suppose F £ E and that 3E£ and 9)£ have a common irreducible constituent. Then 3E is similar to ?).
148 Chapter 9 Proof Let 3 be an irreducible constituent of 3EE and of 9)£. If 3E and 9) are not similar, view them as representations of F[G] and choose b e F[G] with X(b) = 3E(1) and 9)(b) = 0. It follows that if U is any £[G]-representation similar to 3E£, then \X(b) is the identity matrix and hence 3(b) is the identity matrix. A similar argument, working with 9)£, shows that 3(b) = 0 and this contradiction proves the result. | Note that Corollary 9.7 asserts that the representation 3E in Corollary 9.5(c) is unique up to similarity. (9.8) corollary Let F be a splitting field for G and let {3E,} be a set of representatives for the similarity classes of irreducible F-representations. Suppose £2F, Then £ is a splitting field and {3E,£} is a set of representatives for the irreducible ^-representations of G. Proof Since the 3E; are absolutely irreducible, the 3E£ are absolutely irreducible. They are pairwise nonsimilar by Corollary 9.7. Finally, suppose 9) is any irreducible £-representation. By Corollary 9.5(c), 9) is a constituent of 3E,£ for some i. Since X,£ is irreducible, it is similar to 9) and the proof is complete. | (9.9) theorem Let £ be a splitting field for G and let F £ £. Then F is a splitting field iff every irreducible ^-representation of G is similar to 9)£ for some F-representation 9). Proof Suppose F is a splitting field. Then by Corollary 9.8, every irreducible £-representation is as desired. Conversely, let 3 be any irreducible F-representation and let 3E be an irreducible constituent of 3£. By hypothesis, there exists an irreducible F-representation 9), such that 9)£ is similar to 3E and hence 9)£ and 3£ have an irreducible constituent in common. By Corollary 9.7,9) is similar to 3, and thus 3£ is absolutely irreducible. It follows that 3 is absolutely irreducible since the only F-matrices which centralize all 3(9) are scalar. The proof is complete. | (9.10) corollary Let F be any field and G a group. Then some finite degree extension of F is a splitting field for G. Proof Let F be the algebraic closure of F, so that F is a splitting field. Let {3Ej} be a set of representatives for the similarity classes of irreducible F-representations. By Corollary 9.5(b), | {3E,} | < oo and hence only finitely many elements of F occur as entries in any of the matrices 3tt(g) for g e G. Adjoin all of these elements to F so as to obtain the field E. Since F is algebraic over F, it follows that | E: F\ < oo. Since each 3E, may be viewed as an F-representation of G, it follows from Theorem 9.9 that £ is a splitting field. |
Changing the field 149 (9.11) corollary Let Q £ F s C. Then F is a splitting field for G iff every / e Irr(G) is afforded by some F-representation. Proof If F is a splitting field and / e Irr(G), choose a C-representation 3E that affords /. By Theorem 9.9,3E is similar to 9)c for some F-representation % Then 9) affords /. Conversely, suppose that every / e Irr(G) is afforded by some F-representation. Let 3E be an irreducible C-representation and let 9) be an F- representation which affords the same character. Then by the linear independence of Irr(G) and using the fact that C has characteristic zero, it follows that 3E is the unique irreducible constituent of 9)c and occurs with multiplicity 1. Thus 9)c is similar to 3E. The result now follows from Theorem 9.9. | We shall now discuss the character theory of a group over an arbitrary splitting field. The following result is actually true without the assumption that £ is a splitting field. We shall prove the more general fact later. (9.12) lemma Let £ be a splitting field for G. Then the characters of non- similar irreducible ^-representations of G are nonzero, distinct, and linearly independent over E. Proof Let {3E,} be a set of representatives for the similarity classes of irreducible ^representations, and let /, be the character afforded by 3E,. View Xi as being defined on all of £[G]. Since 3Ef(£[G]) is a full matrix ring over E, we may choose aieE\_G] with x,(a,) = 1. By Theorem 9.6, we may assume that //a,) = 0 if i =£ j. The result is now immediate. | If £ is a splitting field for G, we shall use the notation Irr^G) to denote the set of characters of the (absolutely) irreducible ^representations of G. The point of the next result is that in some sense Irr^G) is not as dependent on the particular field £ as the notation would indicate. (9.13) lemma Let £ be a splitting field for G and let / e Irr^G). Suppose K £ £ is a subfield which contains %(g) for all g e G and let F 2 K be another splitting field for G. Then / e IrrF(G). Proof We may replace F by a K-isomorphic copy and assume that £, F £ L for some field L. By Corollary 9.8, L is a splitting field for G and IrrF(G) = IrrL(G) = Irr^G). The result follows. | The next result is of great importance in studying representations in prime characteristic. Its proof depends on Wedderburn's theorem which asserts that finite division rings are commutative. As the quaternion group of order 8 shows, Theorem 9.14 would be false in characteristic zero.
ISO Chapter 9 (9.14) theorem Let 3E be an absolutely irreducible ^representation of G, where £ is of prime characteristic. Suppose that 3E affords the character / and that %(g) e F for all g e G, where F is some subfield of E. Then X is similar to 9)£ for some absolutely irreducible F-representation 9). Proof First we consider the crucial special case that | E \ < oo. Working by induction on | E \, it is no loss to assume that F is a maximal subfield of E. Let A be the set of all F-linear combinations of the matrices X(g) for g e G. Then A is an F-subalgebra of the matrix ring M„(E), where n is the degree of 3E. If a e Z(A), then a is a matrix over E which commutes with all X(g). Since 3E is absolutely irreducible, a is a scalar matrix and thus F • 1 £ Z{A) £ £ • 1. It follows that Z(A) is a finite integral domain, and hence is a field. Since F is a maximal sufield of E, we conclude that either Z(A) = F • 1 or Z(A) = E • 1. If Z(A) = £• 1, then A = £• A = 3E(£[G]). Since 3E is absolutely irreducible, we conclude that A = M„(£) and thus if a e £ — F, we can choose a e A with tr(a) = a. However, a is an F-linear combination of matrices of the form X(g), each of which has trace in F. This contradiction shows that Z(A) = F-l. Next, we claim that the F-algebra A is algebra isomorphic to the full matrix algebra Mk(F) for some integer k. To show this, we first establish that A is semisimple. By Problem 1.5, it suffices to show that A has no nonzero nilpotent ideals. If / is a nilpotent ideal of A, then £ • / is a nilpotent ideal in £ ■ A = 3E(£[G]) = M„(£). It follows that 1 = 0. (Any nonzero matrix has a multiple which is not nilpotent.) By Theorem 1.15, the semisimple F-algebra A is a direct sum of minimal ideals. Since dimF(Z(y4)) = 1, it follows that there is only one direct summand and hence in the notation of 1.15, A = M(A) s AM for some irreducible A-module M. Let D = EA(M), so that D is a division ring by Schur's Lemma 1.5. Since dimF(M) < oo and \F\ < oo, we have |D| < oo and thus D is commutative by Wedderburn's theorem. By the Double Centralizer Theorem 1.16, we have AM = ED(M) and thus D s Z(AM) s F. Since D contains the scalar multiplications, 1 • F, of F on M, we conclude that D = IF. Thus if k = dimF(M), then A s AM = Ed(M) = EF(M) ^ Mk(F) and we have an F-algebra isomorphism 9: A -► Mk(F). Thus ?) = 3E9: F[_G^\ ->• Mk(F) is a representation of F[G]. It is absolutely irreducible since it maps onto Mk(F). We claim that 9)£ is similar to 3E. To see this, let 3 be an irreducible F- representation such that 3E is a constituent of 3£- If 31S not similar to ?), use Theorem 9.6 to choose b eF[G] such that 3(b) = 0 and ?)(£>) ^ 0. Since 3E is a constituent of 3£, we have X(b) = 0 and thus ty(b) = (3E(ft))9 = 0. This contradiction shows that 9) is similar to 3 and thus 3E is a constituent of 9)£. Since 9) is absolutely irreducible, the claim follows and the theorem is proved when £ is finite.
Changing the field 151 We now consider the general case where £ may be infinite. Since we may replace £ by a larger field, it is no loss to assume that E is algebraically closed. Let K £ E be the prime field and let L 2 K be a splitting field with \L:K\ < oo (by Corollary 9.10). Since E is algebraically closed, we may assume L <=, E. Since L is a splitting field, 3E is similar to 3£ for some L- representation 3 and 3 affords / which takes values in L n F. Since \L\ < oo, the first part of the proof yields an absolutely irreducible (L n F)-representation 9), such that 9)L is similar to 3 and hence 9)£ is similar to 3E. Now 9)F is the desired F-representation and the proof is complete. | Theorem 9.14 remains true if X is irreducible but not absolutely irreducible. This will be proved in Corollary 9.23. By the exponent of G we mean the least positive integer n, such that g" = lforallgeG.Clearlyn||G|. (9.15) corollary Let G have exponent n and assume the polynomial x" — 1 splits into linear factors in the field F. If F has prime characteristic, then it is a splitting field for G. Proof Let E 2 F be a splitting field and let / e Irr^G). Then %(g) is a sum of nth roots of unity and hence %(g) e F. The result follows by Theorems 9.14 and 9.9. | An entirely different proof shows that Corollary 9.15 also holds in characteristic zero. This result of R. Brauer depends on his theorem on induced characters and will be proved in Chapter 10. Let E be any field, a an automorphism of E, and 3E an £-representation of G. We can apply a to every entry in the matrix H(g) for every g e G. What results is a new representation, denoted 3E", which may or may not be similar to 3E. Similarly, if 9: G -» E is a function, we write 9-"(g) = {S(g))a. Suppose 3E affords the £-character / and that / takes values in F £ E. If t e Aut(F), it is not obvious that /T is an F-character of G. (9.16) lemma Let £ be a splitting field for G and let / e Irr^G). Suppose 1(g) e F £ £ for all g e G and let t e Aut(F). Then /* e Irr£(G). Proof Let E be an algebraic closure for £ and let F £ £ be an algebraic closure for F. Then Irr£(G) = Irr£(G) = Irrf(G) by Corollary 9.8. Since t is extendible to an automorphism of F, the result follows. | Let K £ £ be a field extension and suppose / is an £-character of G. We write X(j) to denote the subfield of £ generated by K and the character values X(g) for g e G. Note that K(%) is contained in a splitting field for a polynomial
152 Chapter 9 of the form x" — 1 over K. Since this polynomial yields a Galois extension with an abelian Galois group, it follows that K(x) is a finite degree Galois extension of K and the Galois group ^(K(x)/K) is abelian. Assume that £ is a splitting field for G and that F £ E. If /, \\i e Irr£(G), we say that / and i/f are Galois conjugate over F if F(x) = F(\p) and there exists a e g(F(x)/F) such that x" = \ji. It is clear that this defines an equivalence relation on Irr£(G). (9.17) lemma Let £ be a splitting field for G and let / e Irr£(G). Let Sf be the equivalence class of / with respect to Galois conjugacy over F where F £ E. We have (a) If K £ E, K(x) = K, and a e &(K/K n F), then x" e if. (b) If F0 £ F and \ji e if, then \ji is Galois conjugate to / over F0. (c) \y\ = \F(x):F\. Proof (a) By 9.16, x" e Irr£(G). Now (K r\ F)(x) 2 K r\ F is a normal extension and so (X n F)(x) £ X is invariant under a. It is thus no loss to assume that K = (K n F)(/) and so X £ F(/) and no proper subfield of F(x) contains both F and K. It follows by Galois theory that restiction maps g(F(x)/F) onto <S{KjK r\ F) and so x" = f for some t e 9(F(x)/F). Now f(Zt) = fa)andsoz" = zte^. (b) This is immediate by applying (a) to F0 and taking K = F(x). (c) We have that if is the orbit of / under &(F(x)/F). By definition of F(x), the stabilizer of / in this group is trivial and so \!f\ = |^(F(/)/F)| = \F(X):F\. | Now let F be any field and let 3E be an irreducible F-representation of G. Let E 2 F be a splitting field for G. What does 3E£ look like? We shall prove that it is completely reducible; that all irreducible constituents occur with equal multiplicity; that the characters of these constituents constitute a Galois conjugacy class over F and that the common multiplicity is 1 except possibly when F has characteristic zero. (9.18) lemma Let F £ E with \E:F\ = n< oo and let V be an irreducible £[G]-module corresponding to the £-representation 3E. Then V may be viewed as an F[G]-module and as such let it correspond to the F-representation 3- Then (a) deg 3 = n deg 3E. (b) 3 has a unique (up to similarity) irreducible constituent. It is the F-representation 9) such that 3E is a constituent of ?)£. (c) If 3E affords the £-character / and F(x) = F, then 3 affords nx- Proof We may certainly view V as an F-space. Let v1,v2,---,vm be an £-basis for V and let eu e2,. ■ ■, e„ be an F-basis for E. It is routine to check
Changing the field 1S3 that {vte}} is an F-basis for V and so V is finite dimensional over F, 3 is defined, and (a) follows. Let 9) be a (unique up to similarity) irreducible F-representation of G such that 3E is a constituent of 9)£. By Theorem 9.6, choose b e F[G~\ c £[G], such that ?)(£>) = 9)(1) and 9)0(fe) = 0 for every irreducible F-representation 9)0 not similar to 9). Since 3E is a constituent of 9)£ and ?)(£>) is an identity matrix, then so is X(b) an identity matrix. Thus b acts as an identity on V and hence 3(^) is an identity matrix and ?),(£>) ,£ 0 for every irreducible constituent 9)f of 3- Thus every 9), is similar to 9) and (b) is proved. For (c), let g e G and write v$ = £; Vjoty, where a(j- e £ and x(g) = £, ««■ Now write v for 1 ^ /z, v ^ n and /?y/iV e F. Let 3 afford the character \ji. Then (v.ejg = £ fjev^(J> and i/%) = £ &,„„. We have and since we are assuming that £; a„ = /(#) e F, we conclude that I a« = I &;„„ for each /z, 1 < fi < n. The result now follows. | (9.19) corollary Let F c E with |£:F| = n < oo. Let 3E be an irreducible F-representation of G and let 9) be an irreducible F-representation such that 3E is a constituent of 9)£. Then deg 9) divides n(deg 3E). Proof Let 3 be an F-representation obtained by viewing an F[G]- module corresponding to 3E as an F[G]-module. By Corollary 9.7 and Lemma 9.18(b), we conclude that 9) is the unique irreducible constituent of 3 and so deg 9) divides deg 3- The result now follows from 9.18(a). | The following corollary is what survives of Theorem 9.14 when the hypothesis of prime characteristic is dropped. (9.20) corollary Let 3E be an absolutely irreducible F-representation of G which affords the character /. Let F £ £ be such that F(#) = F. Then there exists an irreducible F-representation 9), such that 3E is the unique (up to similarity) irreducible constituent of 9)£. In particular, 9) affords m/ for some integer m.
154 Chapter 9 Proof If F has prime charactersitic, then by Theorem 9.14, there exists 9) such that 9)£ is similar to 3E and there is nothing to prove. Assume then that char(F) = 0. It will suffice to show for some positive integer n that the character nx is afforded by some F-representation 3- This is sufficient since by the linear independence of IrrL(G) for a splitting field L 2 E, it follows that every irreducible constituent of 3L affords / and so is similar to 3EL. We may thus take 9) to be any irreducible constituent of 3 and quote Corollary 9.7 to complete the proof. To produce 3> let K 2 F be a splitting field with \K :F\ = n < oo. By Lemma 9.13, /eIrrK(G). Let 3 be the F-representation which results by taking a K[G]-module which affords / and viewing it as an F[G]-module. Then 3 affords nx by Lemma 9.18(c) and the result follows. | Of course it follows in the above situation that if 9)0 is any irreducible F-representation such that 3E is a constituent of (9)0)£, then 9)0 is similar to 9) and hence 3E is the unique irreducible constituent of (9)0)£- (9.21) theorem Let F s E, where £ is a splitting field for G. Let 9) be an irreducible F-representation of G. Then (a) The irreducible constituents of 9)£ all occur with equal multiplicity m. (b) Ifchar(F)^ 0, then m = 1. (c) The characters /, e Irr£(G) afforded by the irreducible constituents of 9)£ constitute a Galois conjugacy class over F and so the fields F(xi) are all equal. (d) Let L = F(x). The irreducible constituents of 9)L occur with multiplicity 1. (e) If 3 is any irreducible constituent of 9)L then 3£ has a unique irreducible constituent. Its multiplicity is m. (f) 9)L and 9)£ are completely reducible. Proof Let 3E be an irreducible constituent of 9)£ and suppose 3E affords X e Irr£(G). Let L = F(x) and let 3 be an irreducible constituent of ?)L such that X is a constituent of 3£. By Corollary 9.20, 3E is the unique irreducible constituent of 3£. Let m be its multiplicity, so that 3 affords the character m/. If char(£) =£ 0, then Theorem 9.14 yields m = 1 and 3 is absolutely irreducible. Let x = Xu li > ■ ■ ■ > In be the distinct Galois conjugates of / over F so that n = | L: F | by Lemma 9.17(c). For a e @(L/F), form the L-representation 3" defined by 3<r(sf) = 3idf- As a runs over @(L/F) we obtain n representations 3" affording the characters mx„ 1 <i < n. Since m = 1 when char(£) ^ 0, the mxi are distinct in all cases and thus the 3" are pairwise nonsimilar. Also,
Changing the field 155 each (3")£ has a unique irreducible constituent and it occurs with multiplicity m. This follows from the characters if char(£) = 0 and it holds in prime characteristic since then all of the 3" are absolutely irreducible. We claim that the n = \ L: F | representations 3" are exactly the irreducible constituents of 9)L and that each occurs with multiplicity 1. Since the irreducible constituents of (3<r)£ and (3I)£ are nonsimilar if a j= t, statements (a)-(e) will follow when the claim is established. Since 3 is a constituent of 9)L and C^L)a = 9)L for a e <&(L/F), it follows that 3" is a constituent of 9)L for every a. Therefore n(deg 3) ^ deg 9) since n = \&(L/F)\. By Corollary 9.19, deg 9) divides n(deg 3) and we thus have equality. Therefore the 3" are the only irreducible constituents of 9)L and each has multiplicity 1 as claimed. All that remains is to show that 9)L and 9)£ are completely reducible. This follows from Maschke's Theorem 1.9ifchar(£) = 0 so we assume char(£) j= 0. We may assume without loss that 3 is a bottom constituent of 9)L (since some constituent certainly is). Since (9)L)or = 9)L, it follows that 3" is a bottom constituent for every ae&(L/F). Now let V be an L[G]-module corresponding to 9)L and let W be the sum of all of the irreducible submodules of V. By Theorem 1.10, W is completely reducible and it suffices to show W = V. Since every irreducible constituent of 9)L is a bottom constituent, every composition factor of V occurs as a composition factor of W. Since the composition factors of V all have multiplicity 1, V/W is necessarily trivial. The proof for ?)£ is similar since 3£ is irreducible in this case. | We obtain some consequences now. The first generalizes Theorem 9.14. (9.22) corollary Let F be any field. Then the characters of nonsimilar irreducible F-representations of G are nonzero, distinct and linearly independent over F. Proof Let E 2 F be a splitting field for G. By Theorem 9.21, the characters of nonsimilar irreducible F-representations of G are nonzero multiples of sums of disjoint subsets of Irr£(G). The result follows from 9.12 | (9.23) corollary Let F £ E be fields of prime characteristic. Let 3E be an irreducible F-representation of G which affords the character /. Let ?) be an irreducible F-representation such that 3E is a constituent of ?)£. Then deg 9) = \F(x) :F\ deg 3E. In particular, if F(x) = F, then 3E is similar to ?)£. Proof Let L 2 E be a splitting field for G and let C e IrrL(G) be the character of an irreducible constituent of HL. Let & and !T be the Galois conjugacy classes of £ over E and F respectively. Since E has prime characteristic, it suffices by Theorem 9.21 to show that \2T\ = \F(x): F\ \y\.
156 Chapter 9 By Lemma 9.17(b), we have ^cf and thus z = £ if takes on values in F(Q and so F(x) £ F(Q. Since |.T | = \F(Q: F| by 9.17(c), we must show that 1^1 = 1 F(0: F(Z) |. Let <S = 0(F(O/F(z))- If * e 0, then I if = Z = f = I ff tieSf tisSf and it follows from the linear independence of IrrL(G) and Lemma 9.16 that 0 permutes if. Since only the identity of 0 can fix £, this yields \if\ > \<S\ = | F(0: F(z) |. However, F(x) ££n F(Q and thus \if\ = |£(Q:£| = |F(Q:£nF(0| < |F(0:F(z)l and the proof is complete. | Problems (9.1) Let fc£be fields and let Fbea finite dimensional £-space. Let W £ V be an F-subspace. We write V = W*F E provided V = WE and dim£(F) = dimF(W). (a) If V = W *F E and U £ W is an F-subspace, show that UE = U*FE. (b) If 3E is an F-representation of G, show that dimF(3E(F[G])) = dim£(X£(£[G])). Note The symbol *F denotes an internal tensor product. (9.2) Let 3E be an F-representation of G and let E 2 F. Let V be an £[G]- module corresponding to 3E£. (a) Show that V = W *F E for some G-invariant, F-subspace W £ V such that the F[G]-module W corresponds to X. (b) Let U £ W be an F[G]-submodule. Let U and W/I/ correspond to the F-representations ^) and 3> respectively. Show that l/F, is an F[G]- submodule of V which corresponds to ^)£ and that V/UE corresponds to 3£- (9.3) Let W be an F[G]-module and let F £ E with |£ : F\ = n < oo. Let W£ be an £[G]-module corresponding to X£ where 3E is an F-representation corresponding to W. Now view WE as an F-space. Show that the resulting F[G]-module is isomorphic to the direct sum of n copies of W. Note It follows from the Krull-Schmidt theorem that if V and W are F[G]-modules and V®V@---@V=*W®W®---®W, where each direct sum has n terms, then V = W. We conclude via Problem 9.3 that if 3E and ^) are F-representations of G and 3E£ is similar to ^)£, then 3E is similar to % provided \E:F\ < oo.
Problems 1S7 (9.4) Let F £ E be a field extension of possibly infinite degree. Let 3E and ^ be F-representations of G and assume 3E£ and tyE are similar. (a) Show that there exist matrices Ml,M2,...,Mk over F and elements eu e2, ■ ■ ■, ek e E such that X(g)Mt = M^g) for all g e G and 1 < i < k and such that eiMt + • • • + ekMk is nonsingular over E. (b) If | F | > deg 3E, show that 3E is similar to $). Hint [For (b)] Letf(xu x2, ..., xk) be a nonzero polynomial over F and assume that the degree off in x, is < \F\ for each i. Then there exists aua2,..., ak eF such that/(a,, a2,..., ak) =£ 0. (9.5) Under the hypotheses of Problem 9.4, show that 3E and ^ are similar without assuming anything about | F | or | E: F |. Hint Combine the results of Problem 9.4(b) and the note following Problem 9.3. (9.6) Let F £ E and let V be an irreducible £[G]-module which affords the character /. Assume that E = F(%). Show that V is irreducible when viewed as an F[G]-module. (9.7) Let F have prime characteristic and let 3E be an irreducible F-repre- sentation of G. Let D be the centralizer of 3E(G) in the matrix algebra M„(F) where n = deg 3E. Show that D is a field. Hint Let E 2 F be a splitting field and consider the centralizer of 3E£ in Mn(E). Use Theorems 9.21 and 9.2. (9.8) In the situation of Problem 9.7, let / be the character of an irreducible constituent of XE where E 2 F is a splitting field for G. Show that the natural isomorphism between F and F • 1, the scalar matrices in M„(F), extends to an isomorphism F(%) = D. Hint Let L = F(%) and let V be an L[G]-module affording /. View V as an F[G]-module. Then D =* EF[G](V). Use Problem 9.6. (9.9) Let 3E be an F-representation of G and assume 3E£ is completely reducible for some E 2 F. Show that 3E is completely reducible. Hint Consider 3E(J(F[G])). Use Problem 1.10. (9.10) Let 3E be a completely reducible F-representation of G and let E 2 F. Show that 3E£ is completely reducible. Note The analogous statement for algebras other than group algebras does not hold, in general. (9.11) Let F££ and view F[_G^\ <= £[G]. Show that J(E[GJ) = J(F[G]) *F E (in the notation of Problem 9.1).
158 Chapter 9 Hint To obtain J(E[G]) £ (J(F[GJ)E, use Problems 9.2(b), 9.10, and 1.10 to argue that £[G]°/(J(F[G]))£ is a completely reducible £[G]-module. (9.12) If 3E is an F-representation of G let s(3E) denote dimF(3E(F[G])). Let ^Du ..., H)t be the distinct irreducible constituents of 3E. Show that I^,)<S(X) with equality if 3E is completely reducible. (9.13) Let 3E be an irreducible F-representation of G. Let ^) be an irreducible constituent of 3E£, where E 2 F is a splitting field. Show that ms(3E) = deg(3E) degC?)), where s(3E) is as in Problem 9.12 and m is the multiplicity of ^) as a constituent of 3E£. (9.14) Let 3E be an irreducible F-representation of G and let D be the cen- tralizer of 3E(F[G]) in the matrix ring M„(F), where n = deg(3E). Let d = dimF(D). Show that s(3E) = n2/d. Hint Let V be an F[G]-module corresponding to 3E so that V becomes a vector space over the division ring D. Express s(3E) = dimF(3E(F[G])) in terms of dim^K) and d. Use Theorem 1.16. Note Applying Problem 9.14 in the situation of Problem 9.13 we obtain mn2/d = ms(3E) = n degC^)and mn = d degC?)). Since n = wdegC?))|F(x): F|, where / e Irr£(G) is afforded by ^), we obtain-rf = m2 \F(%) :F\. Compare this with Problem 9.8. This formula can be used to generalize Problem 9.8 to the characteristic zero case. (9.15) Let 3E be an irreducible F-representation of G and let / be the character afforded by an irreducible constituent ^) of 3E£, where E 2 F is a splitting field for G. Let D be the centralizer of 3E(F[G]) as in Problem 9.14. Show that the isomorphism FsF'lcfl extends to an isomorphism F(X) = Z(D). Hints Let L = F(%) and let 3 be the irreducible constituent of 3EL such that ^) is a constituent of 3£. Let V be an L[G]-module corresponding to 3- Let D0 = EL[G](V) and £>! = EF[G](V). Check that D0^DV^D. Use the note following Problem 9.14 to show that D0 = Dt. Let Z = Z(D0) 2 L • 1. Observe that D0 = EZ[G](V). Let 2B be the Z- representation corresponding to V viewed as a Z[G]-module, and compute the multiplicity of ^) as a constituent of 2B£. Now use the note following Problem 9.14 again to show Z = L • 1. (9.16) Let 3E be an irreducible F-representation of G and let E 2 F. Show that g e G lies in ker(3E) iff g e kerC^) for every irreducible constituent ^) of 3E£.
Problems 159 (9.17) Let F have prime characteristic p. Show that Op(G) (the largest normal p-subgroup of G) is the intersection of the kernels of the irreducible F-representations of G. (9.18) Let k = char(F). (a) If k t6 2, show that Qe has a unique (up to similarity) nonlinear irreducible F-representation. (b) If k ¥= 0, show that F is a splitting field for Q8. (c) If k = 0, show that F is a splitting field for Q8 iff — 1 is a sum of two squares in F. (9.19) Let F c E, where £ is a splitting field for G. Does there necessarily exist a splitting field K such that F £ X £ E and |X: F| < oo? Consider zero and nonzero characteristic separately. Hint Consider F = Q and E = Q(a, /?) £ C, where a, /? are suitable transcendentals. Take G = QB. (9.20) Let G be cyclic of order n and let F have characteristic not dividing n. Let £ = F(e), where e is a primitive nth root of unity. Let m = | E: F \. Show that every faithful irreducible F-representation of G has degree m and that there are exactly q>{n)/m similarity classes of such representations. Note If \F\ = q < oo, then m can be characterized as the least positive integer such that n \(qm — 1). (9.21) Let G = GL(n, pe) be the full group of nonsingular n x n matrices over GF(pe). Show that G has an irreducible representation of degree ne over GF(p). (9.22) Let F £ £ be fields of prime characteristic and let 3E be an ^representation of G. Assume that 3E affords a character / such that F(%) = F. Do not assume 3E is irreducible. (a) Show that / is afforded by some F-representation. (b) Find an example where 3E is not similar to ^)£ for any F-representation ^.
The Schur index The main question considered in this chapter is the following. If x e Irr(G), then for which fields F c c is % afforded by an F-representation? If F c c is not one of these fields, we wish to measure the extent to which x fails to be afforded over F. This and the results of Chapter 9 (and, in particular, Theorem 9.21) suggest the following definition. (10.1) definition Let F £ E, where E is any splitting field for G. Let X e Irr£(G). Choose an irreducible ^-representation 3E which affords / and an irreducible F-representation ^ such that 3E is a constituent of tyE. Then the multiplicity of 3£ as a constituent of tyE is the Schur index of / over F. It is denoted by mF(/). Note that given / as above, representations 3E and ^) do exist and are unique up to similarity and so mF(x) is well defined. Actually, mF(x) does not really depend on E. If L 2 F(x) is another splitting field, then / e IrrL(G) by Lemma 9.13. A routine argument shows that mF(x) is the same when computed in E or in L. By Theorem 9.21(b), Schur indices are always trivial (that is, equal to 1) in prime characteristic. For this reason we now restrict our attention to the characteristic zero case. In fact, it is really no loss to confine ourselves to subfields of the complex numbers, C. Many important results about Schur indices appear to depend on deep facts about division algebras and number theory. Nevertheless, as this chapter will demonstrate, much can be done by elementary means. In particular, some of the results of Chapter 8 will prove invaluable. 160
The Schur index 161 We collect some facts. (10.2) corollary Let x e Irr(G) and F £ C. We have (a) mF(x) = mF(x)(x). (b) Let y be the Galois conjugacy class of / over F. Then mF(x)(£ if) is the character of an irreducible F-representation of G. (c) If S is the character of any F-representation, then mF(x) divides [S.Z]. (d) mF(x) is the smallest integer m such that mx is afforded by an F(x)- representation. (e) mF(x) is the unique integer m such that mx is afforded by an irreducible F(/)-representation. (f) If F £ E £ C, then m£(/) divides mF(/). (g) If F £ £ £ C and | £: F\ = n < oo, then mF(x) divides nm^x)- (h) mF(x) divides/(I). Proof Part (a) follows from Theorem 9.21(e) and part (b) is a consequence of 9.2 l(a, c). Let ^) be the F-representation in (b). Then ^ is the unique (up to similarity) irreducible F-representation whose character \ji satisfies [>> X] ^ 0- Since mF(/) = \_\ji, /], part (c) follows. To prove (d) and (e), we may assume (by (a)) that F = F(x\ Now y = {/} and the representation ^ of the preceding paragraph affords m/. Parts (d) and (e) are now immediate. Now let F £ E £ C. Any character afforded by an F-representation is also afforded by an F-representation. Now (f) follows from (b) and (c). Assume \E:F\ = n < oo. Then n0 = \E(x):F(x)\ = \E: E n F(x)\ which divides n. Since m£(/)x is afforded by an £(x)-representation, we conclude from Lemma 9.18(c) that n0 m£(/)x is afforded by an F(/)-representation. Now (c) and (a) yield (g). Finally, let p be the character of the regular F-representation of G. Then [p> Xl = Z(l) and (h) follows from (c). The proof is complete. | A useful method for obtaining information about Schur indices is to use induced representations. Let F £ C and H £ G. If 9 is a character of H which is afforded by an F-representation of H, then 9G is afforded by an F-representation of G. (See Theorems 5.8 and 5.9.) This idea underlies much of the remainder of the chapter. We use it to prove the following celebrated result. (10.3) theorem (Brauer) Let G have exponent n and let F = Q(e2*iln). Then F is a splitting field for G and every / e Irr(G) is afforded by an F- representation.
162 Chapter 10 Proof The two conclusions are equivalent by Corollary 9.11. Let /elrr(G). Since F(%) = F, it suffices by 10.2(d) to show that mF(x) = 1. By Brauer's Theorem 8.4 we may write z = ZMG, where 1 runs over linear characters of subgroups of G and ax e Z. Now a linear k e Irr(H) is obviously afforded by an F-representation of H and hence kG is afforded by an F-representation of G. By 10.2(c), we have mF(/)| [1G, /] and since i = Dr. z] = 2>i[*G, z], X it follows that mF(x) = 1 and the proof is complete. | We wish to exploit the ideas in the above proof in order to obtain some more delicate information about the Schur index. (10.4) lemma Let H s G and F s C. Suppose \}i e Irr(H) and / e Irr(G). Then mF(x) divides mF(t/0|F(x, t/0: F(x)\ \jpa, *]. Proof We may replace F by E = F(%). This is so since m£(/) = mF(x) by 10.2(a) and m£(t/^) | mF(i^) by 10.2(f) (and of course F(%, ijj) = E(\li)). Thus we assume F = F(x). Let & be the Galois conjugacy class of \ji over F so that mF(i/f) £ ^ is afforded by an F-representation of G. Therefore mF(%) divides by 10.2(c). Now if ne£f, then \ji = n" for some er e ^(F(t/^)/F). We have X = X" and [fG,z] = [(fGr.zn = [^G.z] and thus mF(/) divides mF(i/>) | ^ | [\pa, /]. Since | V \ = \ F(\p): F |, the proof is complete. | The next theorem is a variation on a result of Brauer and Witt. It allows us to analyze mF(x) one prime at a time in terms of certain sections of a group with sharply defined properties. (A section of G is a factor group, H/K, where K -a H £ G.) (10.5) definition Let F £ C. Then (H, X, 9) is an F-triple provided (a) H is a group, X<H,X = CH(X); (b) 9elrr(H) is faithful; (c) the irreducible (linear) constituents of Sx are Galois conjugate over F(9).
The Schur index 163 (10.6) corollary Let (H, X, 9) be an F-triple and let 1 be a linear constituent of &x- Then (a) lis faithful and X is cyclic; (b) I„(k) = X, k" = 9 and F(S) £ F(l); (c) Sx is afforded by an irreducible F(9)-representation; (d) H/X s 0(F(1)/F(9)). Proof By part (c) of the definition, all linear constituents of $x have the same kernel. Thus ker k c ker 9 = 1 and (a) follows. Since k takes on distinct values at distinct elements of X, it follows that if kh = k, then h e C(X) = X. Thus X = Ij^k) and kG is irreducible by Theorem 6.11(a). Now (b) follows. Also Sx is the sum of 9(1) = \H:X\ distinct linear characters. These must constitute the full Galois conjugacy class of k over F(9) since if a e ^(F(l)/F(9)), then \_k", 9*] = \_k, 9*] ^ 0. Since mm)(k) = 1, Corollary 10.2(b) asserts that 9jf is afforded by an irreducible F(9)-representation and (c) is proved. Finally, for each heH, there exists a unique ah e ^(F(k)/F(9)) such that I""' = X"\ Now l""'k"' = (k'-f1 = (kk")ah = kaka\ Therefore, akh = akah and we have a homomorphism, a: H -> ^(F(l)/F(9)), defined by a(h) = ah. Now ker a = IH{k) = X. Since it was shown above that the H-conjugates of k constitute a full Galois conjugacy class over F(9), it follows that a maps onto <g(F(k)/F(S)) and the proof is complete. | (10.7) theorem Let /elrr(G) and F £ C. Assume p" divides mF(x) for some prime p. Then there exists an F-triple (H, X, 9) such that (a) H is a section of G; (b) pa|mF(9); (c) H/X is a p-group; (d) pJ|F(Z,8):F(z)|. Proo/" By Lemma 9.17(b), it follows that if (H, X, 9) is an F-triple for some E 2 F, then it is an F-triple. By 10.2(f, a) we may therefore replace F by F(x) and so we assume F = F(x). Use induction on \G\. If there exists K < G and t/f eIrr(X) such that pJf bl>a, X] and pJf\F(\l/):F\, then by Lemma 10.4 it follows that p" \ mF(i/0 and we may apply the inductive hypothesis to K with respect to ip and obtain an F-triple (H, X, 9) which satisfies (a), (b), and (c) and such that pjf\ F(\p, 9): F(t/01. Since also pJf\F(\ji): F |, we conclude that pJf\F(9):F\ and the proof is complete in this case. We therefore assume that no such pair (K, \ji) exists. By Solomon's Theorem 8.7 we can write 1G = Z ao.(^o)G' where aQeZ and Q runs over ^f, the set of quasi-elementary subgroups of G. It follows that 1 = ZlG = Z aQx(lQf = X aQ(xQf
164 Chapter 10 and thus 1 = [*,*] = Z «q[(Xq)g. *]■ Choose QeJP such that p does not divide \_(Xq)g, x\ = [Jq, Xq]- If "A is any irreducible constituent of xQ, then since F(x) = F, every element of the Galois conjugacy class of \ji over F has equal multiplicity as a constituent of XQ and we may write Xq = X ^aA. where 6AeZ and A runs over sums of Galois conjugacy classes over F in Irr(Q). Thus p does not divide £ 2>a2[A, A] and we choose A such that pJfbA2\_A, A]. Write A = £ !?, where y is a Galois conjugacy class over F and let i/^ey. Then ftA = [/Q, i/f] = [/, i/fG] is not divisible by p and [A, A] = |!? \ = \ F(\ji): F | is not divisible by p. By the result of the second paragraph of the proof, we conclude that G = Q is quasi-elementary. Thus G has a cyclic normal ^-complement C for some prime q. By Theorem 6.15 we have x(l) is a power of q. Now if a = 0, the result is trivial, taking H = X = 1, and so we assume a > 0 and p|mF(x). Since mF(x)|/(l) by 10.2(h), it follows that q = p and G/C is a p-group. Let X o G, with X 2 C be chosen maximal such that X is abelian. By the inductive hypothesis applied to G/ker /, we may assume that / is faithful. We shall show that (G, X, x) is an F-triple to complete the proof. We have X £ CG(X) -a G. If X < CG(X), then since G/X is a p-group, there exists U^G with X £ U £ CG(X) and | U : X \ = p. Thus U is abelian and this contradicts the choice of X. Thus part (a) of Definition 10.5 is established. Now let X be a linear constituent of Xx and let S = {g e G | X9 is Galois conjugate to X over F}. If st, s2 e S, there exist e^, er2 e ^(F(l)/F) with Is' = 1"'. It follows that Is'52 = X"2"' and thus sis2 e S and S is a subgroup of G. Let T = 7G(1) £ S. By Theorem 6.11(b), there exists a unique >/ e Irr(T) such that rja = x and [r\x, 1] ^ 0. Let \p = rjs so that t/^G = / and \p e Irr(S). We claim that F(\p) = F. Let ae^(F(iii, X)/F). Then /" = x and so [1", Xx] = E^, X*] * 0. Thus X" = X? for some geG and since X" is Galois conjugate to X over F, we have g g S. Since Galois conjugate characters have the same inertia groups, we have T~a S and so fa")9"' e Irr(T). Now ((>7T",)G = (>7'r)G = X'r = X and [iwr'h, i] = [(*")*, ^] = Kriix, n * o. By the uniqueness of r\ we conclude that (rja)B"' = >/ and so •A" = (ri")S = {r\B? = ris = ^
The Schur index 165 since g e S. Since F(i/f, X) is a Galois extension of F, it follows that \ji takes values in F and thus F(i/f) = F as claimed. By the result of the second paragraph of the proof applied to (S, \ji), it follows that S = G. Thus condition (c) of Definition 10.5 is satisfied and the proof is complete. | Before proceeding to our major applications of Theorem 10.7 we derive two easier consequences which demonstrate its power. These are included in Theorem 10.9. (10.8) lemma Let H c g and suppose H has a complement in G. Let (pelrr(H) and suppose q>G = /elrr(G). Then mF(x) divides <p(l) for every F£ C. Proof Let U c G with UH = G and U n H = 1. Since (1[,)G is afforded by an F-representation, we have mF(x) | [(1 vf, /]. Now [(1 vf, /] = [(1 vf, <pG] = [((1i/)g)h> <?]■ However, ((1[,)g)h = (1i/m«)h = P«, the regular character of H. Thus [(1[;)G, xl = lPh> <p] = <P(1) and the result follows. | (10.9) theorem Let /elrr(G) and assume p|mF(x) for some F £ C and prime p. Then the Sylow p-subgroups of G are not elementary abelian and prrifix) divides \G\. Proof Let p" be the p-part of mF(x) so that a > 0. Since mF(x)|/(l), the second statement will follow if pfl+1 divides | G |. Let (H, X, 9) be the F-triple whose existence is guaranteed by Theorem 10.7. Thus 9 = 1H for some linear lelrr(X) by Corollary 10.6(b). Since mF(9) > 1, we conclude from Lemma 10.8 that X is not complemented in H. Since H/X is a p-group, it follows that a Sylow p-subgroup of H is not elementary abelian. Also, if PeSylp(H), then P n X ^ 1 and so 9(1) = |H:X| < |P|. Since pfl<mF(9) < 9(1), we have f < |P| and so pfl+1||H|. The result now follows. | There is an important case when we can decide whether or not mpiS) = 1 for an F-triple (H, X, 9). If H/X is cyclic, the problem "reduces" to a question in field theory. If E 2 F is a Galois field extension and a e E, we define ^£/F(a) = riffeiW) a"- ^ne image of this norm map is clearly contained in F. (10.10) theorem Let (H, X, 9) be an F-triple for some F e c and assume H = XC, where C is a cyclic group. Let 1 be a linear constituent of 9* and write E = F(l) and K = F(S). Let m = \ X n C | and let e be a primitive mth root of unity in C. Then e e K. Also, mf{9) = 1 iff e lies in the image of NEjK. Proof Since X r\ C ^ Z(H) and 9 e Irr(H) is faithful, we can choose a generator y of X n C, such that 9(y) = e9(l) and thus eeF(9) = K. Write | X: X n C | = s and \H: X\ = \C: X r\ C\ = t and choose generators x and c for X and C such that xs = y = c\ Note that X(xf = X(y) = e. By Corollary
166 Chapter 10 10.6 we have t = 9(l) = \9(E/K)\ = \E:K\. Write xc = xr for some integer r. By Definition 10.5 there exists a e <g(E/K) such that Xc~' = X". It follows that the orbits of X under C and under <er> are identical. Since C is transitive on the t distinct linear constituents of Sx, we conclude that |<<x>| > t = \9(E/K)\ and thus <g(E/K) = <<x>. Therefore, NEIK(z)= art"2-/"1 for ae E. Also, X(x)r = X(xr) = Xc~\x) = X(x)°. Let V be the (unique up to isomorphism) irreducible K[X]-module which affords Sx- (See Corollary 10.6(c).) Now mF(9) = 1 iff 9 is afforded by a X[H]-module (by Corollary 10.2(d)) and this happens iff there exists a X-linear action of C on V such that for v e V we have (a) v ■ c' = ve; (b) ((vc~l)-x)-c = vxr. (The sufficiency of this condition involves an easy generators and relations argument on H.) View E as an £[X] module affording X so that /? ■ u = fiX(u) for u e X. With this same action of X on E, we now view £ as a X[X]-module of dimension | E : K | = t. By Lemma 9.18(b) it follows that V is an irreducible constituent of the X[X]-module E. (This is because X is a constituent of tyE where ^ is a K-representation corresponding to V.) Since dimK(F) = 9(1) = t = dimK(E), it follows that V = E as X[X] modules. We conclude that mF(9) = 1 iff there exists a X-linear transformation c of £ such that for all 0 e £ (a') B •£' = /?£• (b') ((B-c-l)X(x))-c = B(X(x)f. Now suppose that mF(9) = 1 and let c be as above. Write d = X(x), a. = 1 ■ c and let n > 0 be an integer. We show that ((5") ■ c = ad"' by induction on n. This is clear for n = 0. For n > 0, the inductive hypothesis and (b') yield (dn) • c = (((a(5(n-1)r) ■ £-1)8) ■ c = a(5(n- i)rdr = tx8"r as claimed. Since d" = 5" we have (dn) ■ c = a((5")or. Since E = K(X) and d = X(x), the d" span E over K. The map /? -» olB" for Be E is X-linear and agrees with c on the 8". We conclude that B-c = a.B" for all 0 e £. Now (a') yields e = 1 • (c)' = aa'a"2 ■ ■ ■ a"'"' = JV£/K(a) as desired.
The Schur index 167 Conversely, suppose NEjK(a) = e for some a e E. Define c on E by /? ■ c = a/?" and observe that c is X-linear. Check that (a') and (b') are satisfied. Thus mF(9) = 1 and the proof is complete. | (10.11) lemma Let p be a prime and let 1 ^hZ. Assume that p\(k — 1) and if p = 2, assume 4|(/c — 1). Write f(e) = (kT - l)/(/c - 1) for 0 < e e Z. Then pe is the exact power of p dividing/(e). Proof Since/(0) = 1, we assume e > 1 and use induction on e. Since k""' = 1 + (k - \)f(e - 1), we have kT=l+ p(k - \)f(e - 1) + ^-^ (k - l)2f(e ~ l)m + (k- Iffie - l)n for suitable integers m, n. Thus f(e) = f(e - 1) p+P(pl)(/c_ 1)m + (/c_ 1}: -] and it suffices to show that the second and third terms in the brackets are divisible by p2. Since p\(k — 1), this is clear if p j= 2. If p = 2, the hypothesis that 4\(k - 1) yields the result. | We next give our principal result about the Schur index. Note that it generalizes Brauer's Theorem 10.3. (10.12) theorem (Goldschmidt-Isaacs) Let / e Irr(G), where G has exponent n and let e be a primitive nth root of unity in C. Let F ^ C and assume that ^(F(e)/F(x)) has a cyclic Sylow p-subgroup P. Then pjfmpix) except possibly when p = 2, P > 1 and ^/--l $ F(%). Proof We may replace F by F(%) and assume F(%) = F. Suppose p | mF(x) and let (H, X, 9) be the F-triple whose existence is given by Theorem 10.7 so that p|mF(9). Let 1 be a linear constituent of Sx so that F(%) = F £ F(9) c F(l) <= F(e). By 10.6(d), H/X S 0(F(1)/F(9)) which is a section of 0(F(e)/F(x)). Since H/X is a p-group, it must be cyclic. Also, since 9(1) > 1, we have H > X and we conclude that P > 1. There exists a cyclic p-group C with XC = H. Let |C| = pc and \C r\ X\ = p" and write E = F(l) and X = F(3). Since mF(9) # 1, Theorem 10.10 yields that no primitive pfcth root of unity lies in the image of NE/K, but that K does contain such a root. Since NEjK(l) = 1, we have b > 0 and K contains a primitive pth root of unity.
168 Chapter 10 Let y be a primitive pcth root of unity in C. Then | K(y): K | divides pc b and K(y) £ F(e). Also | E: K | = | H: X \ = pc~b and E £ F(e). Since 0(F(e)/K) is abelian and has a cyclic Sylow p-subgroup, its subgroups of p-power index are linearly ordered by inclusion. The fundamental theorem of Galois theory now yields that K(y) £ E (since \K(y) :K\ < \ E : K |) and so y e E. We compute NE/K(y). Write t = pc~b. If y e X, then NE/K(y) = y\ which is a primitive pfcth root of unity, a contradiction. Thus y £ K. Write ^(E/K) = <er>, where o(<r) = t and y" = yk for some integer k ^ 1. Then y"' = yk' and W£/K(}') = }7V2---}''lt"1 = y", where q = (k' — l)/(/c — 1). If pc_fc is the exact p-power divisor of q, then yq is a primitive pfcth root of unity which is not the case. By Lemma 10.11 then, pjf(k- l)ifp^2and4^(fc- 1) if p = 2. Let 8 e K be a primitive p"th root of unity with a chosen as large as possible. Then 0 < b < a. Since y $ K, we have a < c and 8 e <y> so that 8 = 8" = 8k and p"\(k — 1). Since a > 0, we have p = 2 and a = 1 so that ■v/—1 £ X. The proof is complete. | (10.13) corollary (Fong) Let G have exponent n = mp", where p is prime and pjfm. Suppose F £ C and F contains a primitive mth root of unity. Let i e Irr(G). Then mF(x) = 1 unless p = 2 and v*^^ £ -F> m which case mF(x) < 2. Proof Ifp = 2andy^T^F, let £ = F[V^] so that \E:F\ = 2 and rrifix) | (2mE(x)) by Corollary 10.2(g). Thus it suffices to assume that sf--l e F if p = 2 and to prove mF(x) = 1. Let e be a primitive nth root of unity in C. The hypotheses on F now guarantee that ^(F(e)/F) is cyclic. If q is a prime divisor of mF(/), then Theorem 10.12 yields q = 2 and ,/-^l $ F. Thus p ¥= 2,4Jfm, and mF(/) is a power of 2. We have 4Jfn and thus a Sylow 2-subgroup of G is elementary abelian and 2J(mF{x) by Theorem 10.9. The result follows. | (10.14) corollary (Roquette) Let G be a p-group and /elrr(G). Let F £ C. Then mF(x) = 1 unless p = 2 and s/-\ $F in which case mF(x) < 2. Proof Immediate from 10.13. | Note that taking G to be the quaternion group of order 8 shows that the exceptional cases in the three preceding results can, in fact, occur. (10.15) corollary (L. Solomon) Let k be the product of the distinct prime divisors of \G\ and let F s C. Assume that F contains a primitive (2k)th root of unity. Then mF(/) = 1 for all / e Irr(G).
The Schur index 169 Proof Note that if 211G |, then ^f — 1 e F. Let e be a primitive nth root of unity, where n is the exponent of G. Since every prime divisor of n divides k, the hypothesis on F guarantees that ^(F(e)/F) is cyclic. If / e Irr(G) and p is a prime divisor of mF(x), then Theorem 10.12 yields p = 2 and y/--l $ F. Since mF(x) | x(1) we conclude that 211G | and we have a contradiction. | Most of the above results are directed to showing that Schur indices are small. We have not yet seen an example to show that mF(x) can be greater than 2. In fact every positive integer can occur as a Schur index. Here, we shall settle for an indication of how to prove that every prime can occur. We need two facts from number theory (a) Let F = Q(e), where e is a root of unity. Let R = Z[e] and let / be a proper ideal of the ring R. Let a e F. Then a = u/v for some u, v e R with not both u,vel. (b) (Dirichlet) Let a,beZ with (a, b) = 1. Then there exists a prime of the form ak + b for some k e Z. Let p be a prime. By (b) above, choose a prime q = p2k + (p + 1) for some k. Then p\(q — 1) but p2Jf(q — 1). We now construct the semidirect product G = QP, where Q is cyclic of order q and P is cyclic of order p2 and acts nontrivially (but not faithfully) on Q. There exists faithful / e Irr(G) with X(l) = p. We claim that mQ(x) = p. (10.16) theorem Let G = XP, where X -o G is cyclic of order pq,P *&G is cyclic of order p2 and |G| = p2<? for primes p and <j such that p2Jf(q — 1). Then there exists faithful / e Irr(G) such that mQ(x) = p. Proof Note that X = CG(X). Let 1 be a faithful linear character of X and let / = 1G. Then /e Irr(G), /(l) = p, and (G, X, /) is a Q-triple. Since ™Q(x)\p by 10.2(h), it suffices to show that mQ(x) ¥= 1. Let /C = Q(/), £ = Q(l), and co e K be a primitive pth root of unity. By Theorem 10.10, it suffices to show that co is not in the image of NEjK. Let v be a primitive qth root of unity in E and write e = cov so that e is a primitive (p<j)th root and £ = Q(e). Let R = Z[e] and let I 3 qRbea. maximal ideal of R. Since /?// is a field of characteristic q, the only qth root of unity in R/I is 1. Thus if a e ^(E/K), then v = 1 = v" mod / and since coeK, we have e" = cov" = cov = e mod /. Since R = Z[e], it follows that r" = r mod / for all reR. Since | E : X | = p, we conclude that NEjK(r) = rp mod / for all re/?. Suppose, by way of contradiction, that co = NEjK(a) for some a e E. By Fact (a), write a = ft/c, where b,ceR but not both b,cel. We conclude that
170 Chapter 10 bp = NEIK(b) = NE/K{c)NEIK{a) = cvm mod /. Since not both b, eel, we conclude that co* = (b*/c*f, where *: R -» R/I is the natural homomorphism. Now 1 + x + x2 + ■ ■ ■ + xp~l = l\?=i (x - ^ and setting x = 1, we see that 1 — co divides p in R. Since p* j= 0 in R/I, we conclude that co* ^ 1. Thus b*/c* is a primitive p2 root of unity in R/I. Since p|(<? — 1), there exists a e Z such that a* is a primitive pth root of unity in R/I. Since (e*)p = 1, we have e* = (a*f for some k and hence eeZ + I. Thus /? = Z[e] = Z + / and /?// S Z/(Z n I) = Z/qZ. Since P2^(<? — 1)» 2/qZ does not contain a primitive p2 root of unity and this contradiction completes the proof. | Suppose F c E c c, with | E: F\ = n < oo and let / e Irr(G) with F(%) = F. By 10.2(g), rrifix) divides nm^x). In particular, if / is afforded by an E- representation, then m^x) = 1 and so mF(x) divides \E:F\. This suggests the question of finding minimal fields £2F over which / is afforded. (10.17) theorem Let x e hr(G) and F £ C with F(x) = F. Let m = mF(x). Then there exists E 2 F such that / is afforded by an ^-representation and \E:F\ = m. Proof Let V be an irreducible F[G]-module which affords mx (by Corollary 10.2(b)). Let D = EF[G](V), the centralizer ring of V, so that D is a division ring by Schur's lemma. Let E be any maximal subfield of D. Then E acts on F and V may be viewed as an £-space. Since E commutes with the action of G, we may view V as an £[G]-module. Now EE[G](V) is the centralizer of E in D. The maximality of E thus yields EE[G](V) = E and thus the £[G]-module V corresponds to an absolutely irreducible ^-representation X by Theorem 9.2(a, c). Let 3 be the F-representation corresponding to the F[G]-module V. By Lemma 9.18, X is a constituent of 3£- Since 3 affords mx and 3E is absolutely irreducible, we conclude that X affords /. Finally, by 9.18(a) we conclude that m = \E:F\ and the proof is complete. | We can, of course, find an F-isomorphism of E into C in the above situation, and so the result remains true if we add the condition that E ^ C. However, it is not always true that if F c L £ C, where L is a splitting field for G and / e Irr(G) with F(x) = F, then there exists a field E with F £ E £ L, | E: F\ = mF(x) and such that / is afforded by an £-representation. We close this chapter by stating a few more facts about the Schur index. The proofs of some of these seem to require a fairly deep knowledge of number theory and division algebras. (a) (Fein) If /e Irr(G), then mu(x) divides n\_x", 1g] for every positive integer n.
Problems 171 (b) (Brauer-Speiser) (Corollary of (a)) If / e Irr(G) is real valued, then mQ(x) < 2. (c) (Fein-Yamada) If /e Irr(G), then mu(x) divides the exponent of G and mQ(x)2 divides \G\. Problems (10.1) Let H £ G, 9 e Irr(H) and / e Irr(G). Suppose F £ C. (a) If Xh = $> show that mF(x)|w,(9) and mF(9) < |G: H|mF(x). (b) If 9G = x, show that mF(S)\mF(x) and mF(x) < \G:H\mF(S). (10.2) Let N o G and <p e Irr(JV). Let F £ C and put H = IG((p). Let 9 be an irreducible constituent of <pH and let / = 9G. Show that mF(x) divides |NG(H):H|mF(9). Hint Consider {g e G | cpg and <p are Galois conjugate over F(/)}. (10.3) Let /elrr(G) and F £ C. Define lF(x) to be the greatest common divisor of the integers nx = \ F(x, ^) ■' F(x) | [iG, xl as X runs over linear characters of subgroups of G. Show that mF(/) | /F(/) but that examples exist where mF(/) < /F(/). Hint For the example, take G = Q8. (10.4) Let /e Irr(G) and F £ C. We say that / is F-semiprimitive if there does not exist H < G and \ji e Irr(H) such that iiG = x and F(\j/) = F(x). (a) If x is F-semiprimitive and N <i G, show that the irreducible constituents of Xn are Galois conjugate over F(x). (b) Let G be nilpotent and assume s/~—l e F if | G | is even. Show that /e Irr(G) is F-semiprimitive iff/(l) = 1. (c) In the situation of (b) and the notation of Problem 10.3, show that 1F(X)= lforallxGlrr(G). Hint A noncyclic p-group in which every normal abelian subgroup is cyclic is necessarily a 2-group and contains a cyclic maximal subgroup. Note Since mF(x) \ lF(x), this problem provides an alternate (and more elementary) proof of Corollary 10.14. Since lF(x) can, in general, exceed mF(x), this result strengthens Corollary 10.14 slightly. (10.5) Let F £ C be a field in which — 1 is a sum of two squares. Let G be a 2-group and /e Irr(G). Show mF(x) = 1. Hint Reduce to the case where x(l) = 2 and Q8 £ G.
172 Chapter 10 Note This strengthens Corollary 10.14 in a different direction and suggests a strengthening of Theorem 10.12. In fact, B. Fein has proved that the exceptional case in Theorem 10.12 can only occur if — 1 is not a sum of two squares in F. (10.6) Let ze Irr(G) and F s C. Let H = G x G x ■ ■ • x G be a direct product of copies of G and \et\p = xxXx'"xXe Irr(H). Show that mF(i/f) divides mF(x). Note In fact if H is the product of mF(/) copies of G, then mF(\ji) = 1. Conversely, if mF(\p) = 1 and H is the product of n copies of G, then mF(x) \ n provided \F: Q\ < oo. (10.7) (Fein) Let H be the product of n copies of G and let / and \ji be as in Problem 10.6. Show that mQ(i/0 divides \jn, 1G]. Note Problem 10.7 and the note preceding it prove that mu(x) divides n\_f, 1G] for every n > 0 and / e Irr(G). (10.8) Let H = G x G and let /elrr(G). Let i> = x x /elrr(H). Show that mQ(\p) = 1. (10.9) Let G = Q8 x Z3 and let x e Irr(G) be faithful. Show that mQ(x) = 1. (10.10) Let G = H x X and \ji e Irr(H) and 9 e Irr(X). Let * = \p x 9. Let F£ C. (a) Show that mF(/) divides mF(t/^)mF(9). (b) Show that equality occurs in (a) provided (mF(i/f), 9(1)|F(9): F\) = 1 and(mF(9),^(l)|F(^):F|)=l. (c) Let p, q be primes such that pjf(q — 1) and qjf(p — 1). Show that pq occurs as a Schur index. (10.11) Let x e Irr(G) and p | mu(x) for some odd prime p. Show that G contains an element of order pq, where <j is some prime such that p\(q — 1). (10.12) Let p be an odd prime and let P be a nonabelian p-group of order p3 and exponent p. It is possible to find H s Aut(P) such that H s Q8 and CP(H) = Z(P) = Cp(t), where t is the involution in H. Let G = PH, the semi- direct product. Let 9 e Irr(P) with 9(1) = p. (a) Show that 9 can be extended to 9e Irr(G) such that %) = ± 1 and §(r) = p mod 4. (b) Show that [£„, /] = (p - 9»)/4, where * e Irr(tf), *(1) = 2. (c) Conclude that — 1 is a sum of two squares in Q(e2"ilp) if p = 3 or 5 mod 8.
Problems 173 Hints Obtain § via Corollary 6.28. Compute %) by working in JV = <P, t>- Note that |CN(f)| = 2p and that the p - 1 Galois conjugates of &N are all equal at t. If p = 3 or 5 mod 8, show that mF(x) = 1, where F = 0(e2*ilp). Note Conversely, if — 1 is a sum of two squares in 0(e2*ilp), then p ^ 7 mod 8. Some primes = 1 mod 8 can occur, however. (10.13) Let PeSylp(G) be abelian of exponent p". Show that paJfmQ(x) for x g Irr(G). Hint If X -o G is cyclic and G/X is a p-group, then there exists F £ G such that V n X = 1 and \G: VX\ < p". (10.14) Show that the following two statements can be added to the conclusions of Theorem 10.7 provided p =£ 2. (e) If Y is the p-complement in X, then CH(Y) = X. (f) If U e Sylp(X) and P e SylP(H), then U £ <D(P), the Frattini subgroup. Hints If CP(Y) > U, show that H has a noncyclic normal abelian p- subgroup. If M £ P and MU = P, then 9My is irreducible. (10.15) Suppose mF(x) = z(l) for some / e Irr(G) and F £ C. (a) If H £ G, show that all irreducible constituents of Xh are Galois conjugate over F(x) and that mF(9) = 9(1) for each such constituent 9. (b) If / t6 1G, show that G is not simple. (c) If 2^/(1) or >/^T g F, show that G/ker / is solvable. Hint Use Lemma 10.4. (10.16) Suppose that mF(x) = z(l) for all / e Irr(G) with F £ C. Show that every subgroup of G is normal.
Projective representations Let JV -o G and suppose 9 e Irr(iV) is invariant in G. For each irreducible constituent % of 9G we have that Xn = e(/)$» where e(/), the ramification, is a positive integer. In Chapter 6 we obtained some information about these mysterious integers. If 9 is extendible to G [which is equivalent to saying that some e(x) = 1], then the e(x)'s are exactly the degrees of the irreducible characters of G/N. (See Corollary 6.17.) We shall see that, in general, the e(x)'s are the degrees of irreducible "projective representations" of G/N. (11.1) definition Let G be a group and F a field. Let X: G -»GL(n, F) be such that for every g,heG, there exists a scalar a.(g, h)eF such that X(g)X(h) = X(gh)z(g, h). Then X is a projective F-representation of G. Its degree is n and the function a: G x G -» F is the associated factor set of 3E. Note that the "factor set" a has nonzero values and is uniquely determined by X. Both of these observations follow from the fact that the matrices X(g) are nonsingular. Let Z(n, F) s GL(n, F) be the group of nonzero scalar matrices. [Note that Z(n, F) = Z(GL(n, F)).] By definition, PGL(n, F) = GL(n, F)/Z(n, F) is the projective general linear group. If X is a projective F-representation of G of degree n, then the composition of X with the canonical homomorphism GL(n, F) -» PGL(n, F) is a homomorphism G -» PGL(n, F). Conversely, if 7T.- G -»PGL(n, F) is any homomorphism, we can define a projective representation X of G by setting 3E(gf) equal to any element of the coset n(g) of Z(n, F) in GL(n, F). 174
Projective representations 17S Before proceeding with the study of projective representations, we digress to give an instance of how they can arise in the study of ordinary representations. (11.2) theorem Let JV<G and suppose ^) is an irreducible C-represen- tation of JV whose character is invariant in G. Then there exists a projective C-representation X of G such that for all n e JV and g e G we have (a) X(n)=y(n); (b) X(ng) = X(n)X(g); (c) X(gn) = X(g)X(n). Furthermore, if X0 is another projective representation satisfying (a), (b), and (c), then X0(g) = X(g)/i(g) for some function fi: G -* C *, which is constant on cosets of JV. Proof For geG and neN, write V(gng~ l) = tye(ri). Since ^ affords a G-invariant character, we conclude that ^) and tya are similar representations ofJV. Now choose a transversal T for JV in G (that is, a set of coset representatives). Take 1 e T. For each t e T, choose a nonsingular matrix P, such that P,tyPt~l = ^)'. Take P± = I. Since every element of G is uniquely of the form nt for n e JV and t e T we can define 3E on G by 3E(nt) = '^(n)P,. Properties (a) and (b) are immediate and (c) follows since X(nt)X(m) = V(»)PtV{m) = WW(m)Pt = ^(nrmr1)?, = X(ntmt~v ■ t) = X(nt • m). Properties (a), (b), and (c) yield mVin) = *(gn) = X{gng-l-g) = Wsng-1)^) and X{gmny%grl = V{gng-1) for all g e G and n e JV. If A is any nonsingular matrix such that A^(ri)A-' = ^(gng-1) for all neN, then A ~ lX(g) commutes with all ^(n) for n e JV and thus /I" lX(g) is a scalar matrix by Corollary 1.6. If X0 also satisfies (a), (b), and (c), we may take A = X0(g) and conclude that X0(g) = X(g)y.(g) for some y.(g) e C *. Also X(g)X(hW(n)X(hy1X(g)-1 = X(gmhnh~')X(gYl = ^(ghnh-'g-1). Comparing this with X(ghmn)X(ghy' = y(ghnh-'g-1)
176 Chapter 11 yields X(g)X(h) = X(gh)a.(g, h) for some a(g, /i)eC and thus X is a projective representation. All that remains now is to check that fi is constant on cosets of N. We have X(n)X(g)f4g) = X0(n)X0(g) = X0(ng) = X(ng)y.{ng). Since X(ri)X(g) = X(ng) is nonsingular, the result follows. | Of course, the above argument does not really require that ^ be a complex representation. Using Theorem 9.2 and Lemma 9.12, any absolutely irreducible representation will do. We begin our study of projective representations by considering the associated factor sets. (11.3) lemma Let a: G x G -» F* be the associated factor set of a projective F-representation of G. Then u(xy, z)a(x, y) = <x.(x, yz)cc(y, z) for all x,y,ze G. Proof Let X be the projective representation. We have X(x)X(y)X(z) = X(xy)X(z)a.(x, y) = X(xyz)a.(xy, z)a.(x, y). Also X(x)X(y)X(z) = X(x)X(yz)a(y, z) = X(xyz)a(x, yz)a(y, z). The result now follows because all of the matrices are nonsingular. | (11.4) definition Let A be a possibly infinite abelian group and let G be any group. Then an A-factor set of G is a function a: G x G -» A such that a(xy, z)cc(x, y) = a(x, yz)a(y, z) for all x, y, z e G. Thus the factor set of a projective F-representation is an F * -factor set where F * denotes the multiplicative group of F. Conversely, we shall show that every F * -factor set is associated with a projective F-representation. To do this we introduce the "twisted group algebra." Let G be a finite group and F a field. Let a be an F* -factor set of G. Let FX\_G] be the F-vectorspace with basis {g\g e G}. (That is, there is a specified basis of Fa[G] which is in one-to-one correspondence with G.) Define multiplication in F"[_G] hyg-h = g~ha{g, h) and extend via the distributive law. To establish that the multiplication thus defined is associative, it suffices to check it on the basis elements g for g e G. That it holds there is immediate from the
Projective representations 177 definition of a factor set. The finite dimensional algebra F*[G~\ is the twisted group algebra with respect to a. Note that if a is the trivial F * -factor set, that is, a{g, h) = 1 for all g, h, we can identify F*[G] with F[G]. In order to see that F*\_G] has a 1, we need the following. (11.5) lemma Let a be an /1-factor set of G. Then a(l, x) = a(l, 1) = <x.(x, 1) for all x g G. Proof We have a(l-l,x)a(l, 1)= a(l, l-x)a(l,x). Canceling a(l, x) yields a(l, 1) = a(l,x) for xeG. That a(l, 1) = a(x, 1) follows symmetrically. | Let a be an F*-factor set of G and let v = a(l, 1)"l e F. Now (vl)g = vga(l, g) = g and similarly g{v\) = g. Thus vT is the unit element in Fa[G]. It is now immediate that the elements g e F*[_G~i for g e G all have inverses. Now let a be an F * -factor set of G and let ^ be any representation of the algebra F"\_G]. Define X(g) = ty(g). Then X(g) is nonsingular and x^xw = mW(Ii) = W • K) = y(gK*(g, h)) = *(gh)*(g, h) so that X is a projective representation of G with factor set a. Conversely, if X is a projective F-representation of G with factor set a, we can define a representation ^ of Fa[G] by setting ty(g) = X(g) and extending by linearity. In other words, the projective F-representations of G having factor set a are in a natural one-to-one correspondence with the representations of the twisted group algebra F*[G]. The situation is analogous to the connection between ordinary representations and the ordinary group algebra. Exactly as in the case with ordinary representations, we define two projective representations X and ^) to be similar if ^) = P~ lXP for some non- singular matrix P. Also X is irreducible if it is not similar to a projective representation in the form Note that similar projective representations have equal factor sets and correspond to similar representations of the appropriate twisted group algebra. Also, irreducible projective representations correspond to irreducible representations of the algebra. Since every finite dimensional algebra has irreducible representations, we have proved the following. (11.6) corollary Let a be an F * -factor set of G. Then G has irreducible projective F-representations with factor set a.
178 Chapter 11 The set of /1-factor sets of G forms a group under pointwise multiplication. In the language of group cohomology, this group is denoted Z2(G, A), the group of "2-cocycles." If fi: G -» A is an arbitrary function, we can define 8(fi): G x G -» A by d(fi)(g,h) = fi(g)fi(h)fi(gh)-\ It is routine to check that 8(fi) is a factor set. Note that 8 is a homomorphism from the group of /1-valued functions on G (with pointwise multiplication) into Z2(G, A). The image of 8 is the subgroup B2(G, A) c Z2(G, A) which is called the group of "2-coboundaries." The factor group Z2(G, A)/B2(G, A) can be identified with the second cohomology group H2(G, A). (More generally in cohomology theory, one assumes that G acts on A. Here we are considering only the "trivial action" case.) The superscript 2 above refers to the fact that we are discussing functions of two variables on G. Since that will be the only situation considered here, we write Z(G, A) for the group of A -factor sets B(G, A) for the image of d and we define H(G, A) = Z(G, A)/B(G, A). We say that two /1-factor sets of G are equivalent if they are congruent mod B(G, A). Thus H(G, A) is the set of equivalence classes of /1-factor sets onG. If X is a projective F-representation on G with factor set a, and y.:G -» F * is any function, define ^ = Xfi by ty(g) = X(g)(j(g). It is trivial to check that ^ is a projective representation of G with factor set /? = o«5(/z). Thus X and 9) have equivalent factor sets. If X and ^ are projective F-representations of G, we say that X and ^) are equivalent if ^ is similar to X/i for some function fi: G -> F*. This is easily seen to define an equivalence relation which preserves irreduciblility. We can now give a necessary and sufficient condition for an invariant irreducible character of a normal subgroup to be extendible. (11.7) theorem Let JV -a G and let 9 e Irr(iV) be invariant in G. Let ^ be a representation affording 9 and let X be a projective representation of G satisfying conditions (a), (b), and (c) of Theorem 11.2. Let a be the factor set of X. Define p e Z(G/N, C *) by 0(gJV, WV) = ofo, h). Then /? is well-defined and its image /? e H(G/N, C *) depends only on 9. Also, 9 is extendible to G iffjg= 1. Proof For m,neN and g.heG, we have a(gfn, hm)X(gnhm) = X(gn)X(hm) = X(g)X(nhm)
Projective representations 179 using 11.2(b) and (c). Furthermore, X(g)X(nhm) = X(g)X(hnhm) = X(g)X(h)X(nhm) = a(0, h)X(gh)X(nhm) = a(g, h)X(ghnhm) by 11.2(c). Since X(gnhm) = X(ghnhm) is nonsingular, we have a.(gn, hm) = a{g, h) and /? is well defined. That /? is a factor set is clear. If X0 were chosen in place of X, we have X0 = X/j. where fi: G -* C * is constant on cosets of JV so that we can define v(gN) = fi{g). If X0 has factor set a0, then tx0(g, h) = x(g, h)n{g)n(h)n(ghyi = p(gN, hN)v(gN)v(hN)v(ghN)-1 and hence /? is independent of the choice of X. If PtyP~l were chosen in place of ty, we can replace X by PXP~l which leaves a, /? and ft unchanged. Thus /? is uniquely determined once 9 is given. If 9 is extendible to G, we can choose X and ^) so that 3E is a representation. Thus a = 1 and hence /? = 1. Conversely, if /? = 1, there exists v: G/N -> C * such that /JfoJV, MV) = vigNXhNXghNy1. Define /z: G -» Cx by fi(g) = v(gN). Now define X0(g) = 3E(g)/z(g)_1 so that X0 is a projective representation of G with factor set cc0(g, h) = cc(g, h)y.{gY ^(h)' lf4gh) = 1 and thus X0 is a representation of G. Furthermore, since X(l) = ^)(1) = /, we have l = a(l,l) = /z(l)/z(l)/z(l)-1 = /z(l). Thus /z(n) = v(l) = /z(l) = 1 for all neN and X0(n) = 3E(n)/z(n)_1 = ^(n). Thus X0 is an extension of ^) and the proof is complete. | A word of caution about Theorem 11.7 is appropriate. In the notation of the theorem, if a e B(G, C *) so that a = 1 in H(G, C *), it does not follow that 9 is extendible to G. For instance, take G = QB, N = Z(G), and 9 the non- principal linear character of JV. Then 9 is not extendible to G and yet H(G, Cx) = 1. (See Problem 11.18.) The theory of projective representations is closely related to that of central extensions. (11.8) definition A central extension of a group G is a (possibly infinite) group T together with a homomorphism n of T onto G such that ker 7t £ Z(r).
180 Chapter 11 (11.9) lemma Let (r, 7t) be a central extension of G with A = ker n. Let X be a set of coset representatives for A in T and write X = {xg\g e G}, where n(xg) = g. Define a:GxG->/l by xgxh = a(g, h)xgh. Then aeZ(G,A). Furthermore, the equivalence class of a is independent of the choice of X. Proof That a is an /1-factor set follows by computing xgxhxk two ways, using the associative law in r. If Y = {yg} is another set of coset representatives, then y9 = fi(g)xg for some fi(g) e A. Now yBy>, = KdMh)xgxh = fi(g)n(h)tx(g, h)xgh = V(g)v(hMghylcc(g,h)ygh and the result follows. | If A and U are abelian groups and 1 e Hom(A, U), then for each ae Z(G, A), we define 1(a) by 1(a) (g, h) = l(a(g, h)). It is routine to check that 1(a) eZ(G,t/). (11.10) corollary Let (r, 7t) be a finite central extension of G and let A, X = {xg} and a be as in Lemma 11.9. Let 9) be an (ordinary) F-represen- tation of T such that the restriction tyA is the scalar representation XI for some leHom(A,F*). Define X(g) = ^(x9) for geG. Then X is a projective F- representation of G with factor set 1(a). Furthermore, V(y) = my)My) for all jieT, where /z: T -» F * is the function defined by fi(y) = Myx^). Also X is irreducible iff ^) is and the equivalence class of X is independent of the choice of coset representatives X. Proof We have X(g)X(h) = WxM*J = ^W0> hK*) = M9, hWgh) and X is a projective representation with factor set 1(a). If y e r, we have y = axg, where g = n(y) and ae A. Thus SK)0 = SfoWfl) = X(n(y))l(yx«tt- In particular, 3E(G) and ^)(r) span the same vectorspace of matrices over F and the assertion about irreducibility follows. Finally, if Xt is the projective representation determined by an alternate choice of coset representatives, we have *i(nO0) = V(y)^iy)'1 = sWjOMOmiOO" ' and X and 3Et are equivalent. |
Projective representations 181 Note that if ^) is an irreducible C-representation (or any absolutely irreducible F-representation), then the condition that tyA be a scalar representation in Corollary 11.10 is automatically satisfied. Henceforth we shall consider only C-representations. (11.11) definition Let (r, n) be a finite central extension of G. Let X be a projective C-representation of G. We say that X can be lifted to T if there exists an ordinary representation ^ of T and a function /z: T -» C * such that $(*) = X(tcW)/zW for all x g T. Furthermore, (r, 7t) has the projective lifting property for G if every projective C-representation of G can be lifted to T. Note that if X is lifted to the representation ^ on T, then ^ is necessarily a scalar representation and by Corollary 11.10 we can construct a projective representation 3Et of G by choosing coset representatives for ker n in T. Then XMx))^) = V(x) = X(n(xMx) and hence Xt is equivalent to X. Thus if (r, rc) has the projective lifting property for G, then all projective C-representations of G are equivalent to ones obtained via the construction in Corollary 11.10. We shall prove a theorem of Schur which asserts that every finite group has a finite central extension with the projective lifting property. The point of Schur's theorem is that it allows us to apply what we know about ordinary representations to the study of projective representations. For instance, in the situation of Corollary 11.10, if ^) is irreducible and F = C, then we have deg X = deg ^ divides | T: A | = | G | by Theorem 3.12. Thus a consequence of Schur's theorem is that the degree of every irreducible projective C-representation of G divides \G\. (11.12) definition The Schur multiplier of G is the group H(G, C"). It is denoted M(G). For a finite abelian group A we use the notation A to denote the group \xx{A). If (r, n) is a central extension of G with A = kern finite, we construct a homomorphism n: A -» M(G) as follows. Choose a set X of coset representatives for A in T and write X = {xg\geG}, where n(xg) = g. Let ocgZ(G, A) be defined by xgxh = <x.(g, h)xgh as in Lemma 11.9. For XeA define rj(l) = i(a), where 1(a) g Z(G, C) is defined by 1(a) (g, h) = l(a(g, h)) and the bar denotes the canonical map Z(G, £*)-> H(G, C*) = M(G). Note that r\ is a homomorphism. The map >/: /?->M(G) is independent of the choice of coset representatives X. This follows since another choice would yield a factor set /? g Z(G, A), which is equivalent to a by Lemma 11.9. Since iz^'e B(G, A), it
182 Chapter 11 follows that l(a)l(/3)- l=Jia.fT^) e B(G, C *). Thus 1(a) and l(/3) are equivalent in Z(G, C") and 1(a) = Wl We shall call the unique homomorphism n: A-> M(G) constructed above the standard map. (11.13) theorem Let (r, n) be a finite central extension of G and let rj be the associated standard map. Let X be a projective C-representation of G with factor set y. Then X can be lifted to r iff y lies in the image of n. In particular, (r, n) has the projective lifting property iff n maps onto M(G). Proof. Let A = kern and let X = {xg\g e G} be a set of coset representatives for /I in T with n{xg) = g. Write xgxh = a(g, /i)xg)l so that a e Z(G, /I). Now suppose y = n{X) for some Xe A. Then 1(a) is equivalent to y and we have X(oig,h))=y(g,h)t4g)fi(h)fi(ghrl for some function fi: G -* C *. Define ^ on T by ^(ax9) = X(a)X(g)fi(g) for a e 4 and ge G. We have mxM*-) = X(g)*(hMgMh) = X(gh)y(9, %(<?)/#) = X(a(g, h))X(ghMgh) = ^(oi(g, h)xgh) = V(xgxh). Since ty(axg) = X(a)ty(xg), it follows that ^ is a representation. Thus ^) lifts Xtor. Conversely, if X can be lifted, we have ty(x) = X(n(x))n(x) for some representation ^ of T and function fi: Y -> C. Thus X(l) = /z(l)~ M|)(l) and for every aeA we have ^(a) = X(l)fi(a) = /z(a)/z(l)-1,?)(l) and so 1(a) = H(a)/i(l)~i is a linear character of A. Now write v(g) = fi{xg). We have l(a(g, ^))^)(x9,) = ^)(x9)^)(x,) = X(g)X(h)v(g)v(h). Thus l(a(#, h)Wgh)X{gh) = y(g, h)v(g)v(h)X(gh) and 1(a) is equivalent to y. Therefore tfl) = m = y and the proof is complete. | Given an arbitrary finite group G we shall show that M(G) is finite and construct a central extension (r, ri) of G with A = ker rc such that the standard map n:A-> M(G) is an isomorphism. This will thus be a group with the projective lifting property for G with smallest possible order, namely | G11 M(G) |. Such a group T is called a Schur representation group for G. An abelian group Q is divisible if for every x e Q and positive integer n there exists yeQ, with/ = x. For instance, F * is divisible if F is algebraically closed.
Projective representations 183 (11.14) lemma Let A be an (infinite) abelian group and let Q c A with Q divisible. Assume | A : Q | < oo. Then Q is complemented in A. Proof Use induction on | A : Q\. We may assume A > Q and choose ae A - Q.Letn = o(aQ) in A/Q and let u = a" e Q. Let ve Q with t>" = u by divisibility and let ft = av~l so that ft" = 1. Since aQ = bQ, it follows that n = o(bQ) in A/Q and thus <b_> n Q = 1. Now let ,4 = 4/<fc> so 0 = Q(b}/(by satisfies 0 s Q and \A: 0| = |/1: 8<ft>| < |/I: Q\. By the inductive hypothesis, Q is complemented in A and thus there exists B ^ A with B n Q<£>> = <£>> and QB = A. Now gnB = Qn Q<£>> nB=2n <£>> = 1 and the proof is complete. | The above result remains true without the assumption that | A: Q | is finite. We will not need that more general fact, however. (11.15) theorem Let F be an algebraically closed field and G a finite group. Then H(G, F *) is finite and each of its elements has order dividing \G\. Furthermore, B(G, F*) is complemented in Z(G, F"). Proof First, we argue that B(G, F *) is divisible. If 0 e B(G, F *) and n is a positive integer, write /? = d(/i) for some function fi: G -» F *. For each g e G, choose v(g)e F* such that v(g)n = fi(g). Then c5(v)n = (5(^) = p. We can thus apply Lemma 11.14 when we show that \H(G, F*)| < oo. Let a e Z(G, F *) and define tAo) = EI a(0>;,(:)- xeG For fixed g, /i e G, we have a(g, hx)a.(h, x) = o{gh, x)a.(g, h). Taking the product over all x e G yields fi(g)ti(h) = ii{gh)z(g, h)M and thus a|G| e B(G, F *). This shows that H(G,F*) has exponent dividing | G |. Now let t/ = {aeZ(G,Fx)|a|G| = 1}. For aeZ(G,Fx), let /4 = <B(G, Fx), a>. By the result of the previous paragraph \A:B(G, F*)\ divides \G\. Thus by Lemma 11.14, B(G, F*) is complemented in A and the complement is a subgroup of U. Thus a e B(G, Fx )t/ and hence B(G,F*)U = Z(G,F*). Now every element of U is a function from G x G into {ye F|>>|G| = 1}. Since this is a finite set, it follows that | U | < oo and thus |tf(G,Fx)| = |B(G,Fx)t/:B(G,Fx)| < \U\ < oo and the proof is complete. | Next we need a general method for constructing central extensions.
184 Chapter 11 (11.16) lemma Let A be an arbitrary abelian group and let ae Z(G, A). Then there exists a central extension (r, n) of G with kern = A and such that a set of coset representatives {xg\g e G} exists with n(xg) = g and xgxh = Proof Let T = G x A as a set and define (g, a)(h, b) = (gh, a.(g, h)ab). That this multiplication is associative follows from the fact that a is an A- factor set. Let a(l, l)~l = zeA. Then (1, z)(g, a) = (g, a(l,g)za) = (g, a) by Lemma 11.5. Also (g-\a-ia(g-1,gr1z)(g,a) = (l,z) and hence T is a group with 1 = (1, z). Clearly n: F -> G defined by n(g, a) = g is a homomorphism with ker 7t = /?"= {(1, fl)|ae i4}. That A £ Z(G) follows since a(l, g) = a(g, 1) by Lemma 11.5. Since (1, za)(l,zb) = (1, zab), it follows that a<->(l,za) defines an isomorphism A ^ A. We identify /I and A via this isomorphism. Let X = {(g, l)\geG}. Then n maps X one-to-one and onto G and so X is a set of coset representatives for A in T. Now (g, 1)0, 1) = (gh, a(g, h)) = (1, zafo, fc))(gfc, 1) by Lemma 11.5. Since (1, za(g, h)) = a(g,h) under our identification, the proof is complete. | (11.17) theorem (Schur) Given G, there exists a finite central extension (r, n) which has the projective lifting property for G. Furthermore, (r, n) can be chosen such that ker n = A = M(G) and the standard map A -> M(G) is an isomorphism. Proof Let M be a complement for B(G, C *) in Z(G, C *) by Theorem 11.15. Let A = M. Define <x(g, h)e A by <x(g, h)(y) = y(g, h) for y e M. It is clear that in fact a(gr, h)eM = A. Next (a(g/i, /c)a(g, /i))(y) = y(gh, k)y(g, h) using the definition of multiplication in M. Similarly, (a(#, ^/c)a(^, /c))(y) = y(g, hk)y(h, k) and since y runs over a set of factor sets, it follows that a e Z(G, A). Now let (r, 7t) and X = {xg\g e G} be as in Lemma 11.16 so that xgxh = a.(g, h)xgh. Let >/: A -» M(G) be the standard map. We show that rj maps onto. For y eM(G) = Z(G, C*)/B(G, Cx), there exists yeM r\y since M complements B(G, Cx) in Z(G, Cx). Now define X on A = M to be the evaluation map at y. Note that Xe A. Now !(%, h)) = a(g, ^)(y) = y(g, h)
Projective representations 185 so that 1(a) = y. Therefore as desired. By Theorem 11.13, it follows that T has the projective lifting property for G. Also, \A\ = \A\>\r,(A)\ = \M(G)\ = \M\ = \A\ and so rj must be one-to-one and A = M(G). However A = A by Problem 2.7(c) and the proof is complete. | (11.18) corollary Let X be an irreducible projective C-representation of G. Then deg(3E) divides \G\. Proof See the discussion preceding Definition 11.12. | Before going on to exploit Schur's theorem, we give another result which is useful in the computation of M(G). We have defined a Schur representation group of G to be a minimal central extension with the projective lifting property. The next theorem will yield another characterization. (11.19) theorem Let (r, n) be a finite central extension of G with A = ker n and let r\: A -> M(G) be the standard map. Let A0 = A n F. Then ker rj = {AeA\A0^ ker 1}. In particular, r\ is one-to-one iff A <=, F. Proof Let X = {xg\g e G} be as usual and write xgxh = a.(g, h)xgh with a e Z(G, A). Suppose 1 e ker rj. Then 1 = rj(X) = 1(a) and so 1(a) e B(G, C *). Thus Mg,h)) = fi(gMh)fi(ghyl for some function fi: G -* C *. Now define 1 on T by \axg) = l(a)/*(g) and check that l(xg)l(xh) = Uxgxh). (This is essentially the same calculation as in the proof of Theorem 11.13.) It follows that X is a linear character of T which extends 1. Now F ^ ker 1 and so A0 c ker 1. Conversely, let A0 c ker 1. Then 1 is extendible to l0e Irr(/IF/F) by l0(a:>c) = 1(a) for x e F. Since T/F is abelian, 10 is extendible to lt e Irr(r/F). Now h(xg)^i(xh) = Xi(xgxh) = l(cc(g, /j))1i(x9A).
186 Chapter 11 Define ^g) = li(x9). Then and the result follows. | (11.20) corollary Let (r, 7t) be a finite central extension of G with A = ker n. (a) If A £ F, then A is isomorphic to a subgroup of M(G). (b) Assume |/1| = |M(G)|. Then /I £ F iff r has the projective lifting property for G. In this case M(G) = A. Proof (a) follows since A = A and (b) is immediate from Theorems 11.19 and 11.13. | Thus we see that r is a Schur representation group for G iff T/A = G for some A <= Z(F) such that \A \ = \M(G)\ and A <= r. (11.21) corollary Let p be a prime divisor of |M(G)|. Then a Sylow p-subgroup of G is noncyclic. Proof Let T be a Schur representation group for G (which exists by Theorem 11.17). We may assume that T/A = G with A c Z(r)- Let P//1 e Sylp(G) and assume P/A is cyclic. Since /I is central, it follows that P is abelian and thus r has an abelian Sylow p-subgroup. By Theorem 5.6, pJC\r nZ(r)|. Since A <= F by Corollary 11.20, we have pJf\A\. Thus pi|M(G)|. | The next corollary is quite important. It can, of course, be proved without all of the complex machinery which we have developed. (11.22) corollary Let Af<G with G/N cyclic and let 9eIrr(N) be invariant in G. Then 9 is extendible to G. Proof By Corollary 11.21, M(G/N) = H(G/N,C) is trivial. The result now follows from Theorem 11.7. | The above theory, together with some rather technical computations yields a tool which is very useful when studying the characters of groups with normal subgroups. Let N <i G and let 9 e lrr(N) be invariant in G. Under these hypotheses we say that (G, N, 9) is a character triple. The analysis of this situation is much easier if 9 is linear. We shall use a Schur representation group for G/N to replace (G, N, 9) by another character triple (I\ A, X) in which r/A = G/N and X is linear. Of course, this would not be of much value unless we knew that the character theory of r was somehow closely tied to
Projective representations 187 the character theory of G. For instance, we would want lr and 9G to have the same numbers of irreducible constituents with the same ramifications. We define below the notion of "isomorphism" of character triples. This rather complicated definition makes precise some of the ways in which the character theory of two character triples can be related. We first introduce some notation. If (G, N, 9) is a character triple, let Ch(G 19) denote the set of (possibly reducible) characters / of G such that Xn is a multiple of 9. Let Irr(G|9) be the irreducible characters among these, so that Irr(G|9) is the set of irreducible constituents of 9°. Note that if N c H c g, then (H, N, 9) is a character triple and Xh e Ch(H 19) whenever / e Ch( G19). If r. U -» V is an isomorphism of groups and (pelrr(U), let (pTelrr(V) denote the corresponding character, so that q>x(ux) = q>(u). (11.23) definition Let (G, N, 9) and (r, M, cp) be character triples and let v. G/N -» T/M be an isomorphism. For N £ H c G, let Hr denote the inverse image in T of i{HjN). For every such H, suppose there exists a map aH: Ch(H|9) -> Ch(HT|<p) such that the following conditions hold for H, K withN c K c He Gand/, t/^eCh(H|9). (a) o„(x + ip) = a„(x) + (TH(^); (b) [x,./0 = [>h(xW'/')]; (c) ^(Zk) = (erH(x)V; (d) <rH(Z/?) = <7H(z)/T for /? e Irr(tf/JV). Let a denote the union of the maps erH. Then (t, a) is an isomorphism from (G, N, 9) to (r, M, <p). Note that if (t, a) is an isomorphism from (G, N, 9) to (r, M, <p), then erH is determined by its restriction to Irr(H 19). (This follows from (a).) By (b) it follows that aH maps Irr(H|9) one-to-one into \xx{Hx\q>). Therefore, to construct an isomorphism (t, a) it suffices to define erH on Irr(H 19) to be one- to-one, then extend the definition by (a) and check that (c) and (d) hold for Xelrr(tf|9). (11.24) lemma Let (t, a): (G, N, 9) -* (r, M, cp) be an isomorphism of character triples. Then aH is a bijection of Ch(H 19) onto Ch(HT | <p) for all H withN c H c G. Furthermore, x(l)/9(l) = (7H(x)(l)/<p(l)forallxeCh(H|9). Proof If OalXi) = oH(x2) for /(eCh(H|9), we have [/;, i/^] = ["ateX <th('/')] is independent of i for all \ji e Irr(H 19). It follows that Xi = li and hence erH is one-to-one. For /eCh(H|9) write e(x) = z(l)/9(l) and similarly set e(rj) = rj(l)/(p(l) for rieCh(Hr\(p). Note that aN(9)elrr(M\(p) and so cr^(9) = <p. We have
188 Chapter 11 Xn = e(z)9 and rjM = e(r])(p and thus e(<rH(x))<P = (oh(X))m = On(Xn) = ^(e(z)9) = e(x)<p and thus e(aH(x)) = e(x) as claimed. By Frobenius reciprocity, we have 9H= Z e(X)x zelrr(H|8) and comparing degrees yields £e(x)29(l) = |H:JV|S(1) so that £e(x)2 = \H:N\ where / runs over Irr(H|9). Similarly £ e(rf)2 = \Hr: M\ = \H:N\ for rj e Irr(HT | cp). Since aH maps Irr(H 19) one-to-one into Irr(HT | cp\ we have \H:N\ = Ze(Xf = £e(o„(X))2 < I<W2 = \H:N\. It follows that every r\ e Irr(HT | <p) is of the form oH{x) for some / e Irr(H 19). The result now follows. | (11.25) corollary Isomorphism is an equivalence relation on character triples. Proof. The reflexive and transitive properties are obvious. If (T,(7):(G,JV,9)^(r,M,<p) is an isomorphism, then aH: Ch(H|9)-> Ch(K|<p) is one-to-one and onto where K = HX. We can, therefore, define a"1 by (<x~')K = {oH)~l for M £ K £ T where H = KT~'. It is routine to check that (T-',(j-1):(r,M,<p)^(G,iV,9) is an isomorphism. | A nearly trivial example of an isomorphism of character triples is given by the following result. (11.26) lemma Let (G, N, 9) be a character triple and let /z: G -» T be an onto homomorphism with ker y. £ ker 9. Let M = /z(iV) and let q> e Irr(M) be the character corresponding to 9 e Irr(iV/ker /z). Then (G, JV, 9) and (r, M, <p) are isomorphic character triples. Proof We have r. G/N -» T/M defined naturally from /z. For N £ H £ G and / e Ch(H 19) we obtain ker fi £ ker / and we may view / as a character of H/ker fi. Let oH{x) be the corresponding character of (i(H) = H/ker fi. Check that (t, er) is the desired isomorphism. | Another example of an isomorphism of character triples is provided by the following.
Projective representations 189 (11.27) lemma Let (G, JV, cp) be a character triple and let r\ e Irr(G) be such that r]N(p = 9e Irr(JV). For JV £ H £ G, define aB: Ch(H\cp) -+ Ch(H|9) by atM>) = </"7h ■ Let i: G/N -> G/N be the identity map. Then (i, (J): (G, JV, p) -»(G, JV, 9) is an isomorphism of character triples. Proof It is clear that aH does map Ch(H | q>) to Ch(H 19) and that properties (a), (c), and (d) of Definition 11.23 hold. It thus suffices to show that aH maps Irr(H | <p) one-to-one into Irr(H 19). This follows from Theorem 6.16. | Given invariant X, 9eIrr(JV) for JV-a G, with /(l) = 1, it follows that (G, JV, X) and (G, JV, 9) are isomorphic if 91"i is extendible to G. (11.28) theorem Let (G,N, 9) be a character triple and let (r, n) be a finite central extension of G/N having the projective lifting property. Let A = ker n. Then (G, JV, 9) is isomorphic to (r, A, X) for some Xe A. Proof By Theorem 11.7, the triple (G,N,9-) determines an element peH(G/N, C*) = M(G/N). Since T has the projective lifting property for G/N, we can find X e A such that rj(X) = fi~l where r\ is the standard map (Theorem 11.13). Now let G* £ G x T be defined by G* = {(g, x)\g = n(x)}, where g = gN, the image of g in G/N. Note that G* is a subgroup of G x T. Let L = N x A and observe that L -a G*. Define 9* and 1* on L by &*(n, a) = &(n) and X*(n, a) = 1(a). Note that 9*, 1* e Irr(L) are invariant in G*. We have projection homomorphisms fiG: G* -> G and /zr: G* -> T. These maps are onto and ker (iG = 1 x /I s ker 9* and ker /zr = JV x Is ker 1*. It follows by Lemma 11.26 that (G*, L, 9*) is isomorphic to (G, JV, 9) and (G*, L, 1*) is isomorphic to (r, A, X). By Corollary 11.25, it suffices to show that (G*, L, X*) and (G*, L, 9*) are isomorphic and by Theorem 11.27 we will be done when we show that 9*(1*)~l is extendible to G*. Let 9) be a representation of JV affording 9 and let 3E be a projective representation of G as in Theorem 11.2. Let a be the factor set of G belonging to 3E and let /? be the corresponding factor set of G/N as in Theorem 11.7. Let {xg\g e G/N} be a set of coset representatives for A in T with n(xg) = g. Take xi = 1. Write xsxh = y(g, h)xsh so that y e Z(G/N, A). Since >/(!) = /T \ we have X(y)peB(G/N,C*) and hence ^(7(ff,«))ate,*) = v(g)-1v(«)"1v(ff«) for some function v: G/N -» C *.
190 Chapter 11 Every element of G* is uniquely of the form {g, axs) for g e G and a e A. Define 3 on G* by 3(ff,axJ) = X(g)l(a)"1v(g). We compute 3(0, axs)3(h, bxh) = HgKMpbY'"fo. MaMfi) and 3tefc, aby(g, h)xsh) = l£{gh)X(ab)- lMy{g, R))" 1v(^). Since these two expressions are equal, we conclude that 3 is a representation ofG*. Now y(l, 1) = 1 and a(l, 1) = 1 and it follows that v(n) = v(l) = 1 for neN. Therefore 3(n, a) = 9)(n)l(a)"l and 3L affords S*(A*)~K The proof is now complete. | We now consider some applications. (11.29) corollary Let JV -a G and /elrr(G). Let 9eIrr(JV) be a constituent of Xn . Then /(l)/9( 1) divides |G:JV|. Proof Let T = 7G(9), the inertia group, and let \jie\xx(T) such that \pG = x and i/^ = e9 (Theorem 6.11). Since x(l) = |G: T|i/<1), it suffices to show that i/f(l)/9(l) divides | T: JV|. Let (r, A, X) be a character triple isomorphic to (T, JV, 9) with 1 linear. Let C e Irr(r|l) correspond to \p e Irr(T|9). Then t/<l)/9(l) = C(l)/l(l) = C(l) by Lemma 11.24. Since A £ Z(Q we have C( 1) divides \r-.A\ = \T:N\by Theorem 3.12. The result follows. | Note that by repeated application of this result we can weaken the hypothesis that JV is normal in G and assume only that JV is subnormal, that is, that there exist subgroups JV, such that JV -a JV, -a ci JVt <a G. In particular, if JV is subnormal and abelian then since 9(1) = 1 we obtain the following generalization of Ito's Theorem 6.15, (11.30) corollary Let N S: G be subnormal and abelian. Then x(l) divides | G: JV | for every / e Irr(G). (11.31) corollary Let JV-a G and let 9eIrr(JV) be invariant in G. Suppose for every Sylow subgroup P/N of G/N that 9 is extendible to P. Then 9 is extendible to G. Proof Let (r, A, X) be a character triple isomorphic to (G, JV, 9) and with 1 linear. If JV £ H s G and /I £ X £ r are such that H and X correspond,
Projective representations 191 then 9 is extendible to H iff X is extendible to K. This follows from Lemma 11.24 since extendibility of 9 is equivalent to existence of / e Irr(H|9) with X(l)/9(1) = 1 and similarly for extendibility of X. Thus X is extendible to the inverse image in T of every Sylow subgroup of Y/A. By Theorem 6.26, we conclude that X is extendible to r and hence 9 is extendible to G, as desired. | We can use Corollary 11.31 to obtain an alternate proof of Gallagher's Theorem 8.15, a proof independent of the results of Chapter 8. We are assuming that (G, N, 9) is a character triple in which (9( 1), | G: N |) = 1 and that det(9) is extendible to G. By Theorem 6.25, 9 is extendible to H for every H with JVcHsG and H/N solvable. The result now follows from Corollary 11.31. The following combines several of our extendibility criteria. (11.32) theorem Let N o G and 9eIrr(iV), with 9 invariant in G. Assume for every prime divisor p of (9( 1 )o(9), | G : N |) that a Sylow p-subgroup of G is abelian or if p ^ 2, that 9 is p-rational. Then 9 is extendible to G. Proof By Corollary 11.31, it suffices to assume that G/N is a p-group for some prime p. If a Sylow p-subgroup of G is abelian, then 9 is extendible to G by Theorem 8.26. If p ^ 2 and 9 is p-rational, then 9 is extendible by Theorem 6.30. In the remaining case, (|G:JV|, o(9)9(l)) = 1 and Corollary 6.28 yields the result. | As another application of the theory of projective representations, we prove a theorem of T. Berger about the characters of solvable groups. Recall that x £ Irr(G) is said to be quasi-primitive if %N is homogeneous (that is, a multiple of an irreducible) for every JV o G. Primitive characters are necessarily quasi-primitive and, in general, the converse is false. (11.33) theorem (Berger) Let G be solvable and suppose /elrr(G) is quasi-primitive. Then / is primitive. In order to prove Berger's theorem, we shall exploit the fact that there exists a central extension of G with the projective lifting property. We break the proof into several intermediate results. The first of these is independent of projective representations. (11.34) theorem Let L^H<G with LoG,H maximal and G/L solvable. Let 9 e Irr(H) such that 9G = / and 9L are irreducible. Then there exists M <i G such that Xm is not homogeneous and M ^ L. Proof We may assume without loss that L = coreG(H), the largest normal subgroup of G contained in H. Let K/L be a chief factor of G so that
192 Chapter 11 K/L is an abelian p-group and X £ H.ThusG = XHbythemaximalityofH. Since K/L is abelian, we have H n Ko K and since HnK<Hwe conclude that H nK-^G and thus H r\ K = L. Let C = CG(K/L). Then C<aG and so CnH-^H. However, K £ N(C n H) and hence C n H ~^G. The maximality of L forces L = CnH.We have C = K(CnH) = X. If X = G then H = L <a G and since 9G e Irr(G) we have 7G(y) = L by Problem 6.1 and thus Xl is not homogeneous. We assume, therefore, that K < G and let M/K be a chief factor of G so that M/K is an abelian <j-group for some prime q. If q = p, then (K/L) n Z(M/L) ,£ 1 and the minimality of K/L yields X/L £ Z(M/L). This contradicts CG(K/L) = K and we conclude We suppose that Xl and /M are homogeneous and write Xm = &t\ for some if e Irr(M). Now /(1)/5(1) = | G: H | = | K : L | is a power of p. Since 5L e Irr(L), it is a constituent of r\L and hence >/( l)/y(l) is an integer which divides /(l)/5( 1). Therefore >/(l)/y(l) is a power of p. Since M/X is a <j-group for <j j= p, we conclude that >/KeIrr(X) by Theorem 6.18 or Corollary 11.29. Since Xl is homogeneous, it is a multiple of 5L and thus >/L is also. Let R = M r\H and let t be any irreducible constituent of >/R. Then [tl, 5 J ^ 0 and hence t = yR/? for some /? e Irr(/?/L) by Corollary 6.17. Since M = KR and X n /? = L, we can find y e Irr(M/X) such that yR = /?. Now *M = (W)" = (SRyRf = (»Rfy. However, since MH = G and M n H = R, we have (V = (y°)M = Zm = erj and thus tm = er\y. Since >/ is a constituent of iM by Frobenius reciprocity, we have [rj, rjy~\ ^ 0. Since >/KeIrr(X), Corollary 6.17 yields y = 1M and hence /? = 1R and t = yR. Thus >/R is homogeneous and since (SR)M = er\ we have rjR = e3R and ij(1) = ey(l). This yields y(l)|M:/?| = ((9R)M)(1) = e2y(l) and [tjr^r] = e2 = \M:R\. It follows that r\ vanishes on M - R by Lemma 2.29. As r\ is G-invariant, it vanishes on M — Ra for all g e G and hence vanishes on M — f]Ra = M — L. In particular, >/ vanishes on M — X and so \M : X | = [>/K, >/K] by Lemma 2.29. This contradicts rjK e Irr(X) and completes the proof. | (11.35) lemma Let (t, a): (G, N, y) -> (T, M, cp) be an isomorphism of character triples and let JV £ H £ G. Suppose i/f is a character of H. Then ipeCh(H\&) iff tAGeCh(G|y). Also, if ipeCh(H\&l we have crG(^G) =
Projective representations 193 Proof The first assertion is immediate from Frobenius reciprocity. To prove that <rG('/'G) = (<JH('/'))r when \ji e Ch(H 19), it suffices to check that for all /e Irr(r|<p). Since aG maps Irr(G|9) onto Irr(r|<p) by Lemma 11.24, we have i = crG(^) for some £, e Irr(G 19). Now i°Gw% °Gm = bi>G, n = W, «fl] and the result follows. | (11.36) lemma Let Z £ Z(r) be such that T has the projective lifting property for G = T/Z with respect to the natural homomorphism n: T -» G. Suppose K< r, H g r, M = T and H n K 2 Z. Let 9e Irr(H/Z) and assume HnKs Z(9). Then there exists / e Irr(r) such that K £ Z(/) and XH = 19 for some linear 1 e Irr(H). Proof Let 9) be a representation of H which affords 9. Let L = H n X <i H. Since L £ Z(9), 9)L is a scalar representation and we can apply Corollary 11.10 to H/(L n ker 9) and construct a projective representation 3E of H/L such that 9)(/i) = X(hL)/i(h) for some function /z: H -> C\ Now let t: G -» H/L be the composite of the natural maps G = T/Z ^ T/K = HK/K s H/L so that x(hkZ) = hL for heH and fceX, Let 3E* be the projective representation of G obtained by composing t with 3E so that X*(hkZ) = 3E(/iL) = Now lift 3E* to the representation 3 of T so that 3M = 3E*(xZ)v(x) for x e r, where v: T -» C *. This yields 3(hk) = y(h)fi(hriv(hk) for he H and keK. Let 3 afford /. Restricting the above to H, we have 3W = (9(h)i4hy1v(h) and hence l(/i) = /z(/i)_1v(/i) defines a linear character lelrr(H) and Xh = ^1. Thus Zelrr(r). For /c e K we have 3(/c) = 9)(l)/z(ir1v(/c)) which is a scalar matrix. Thus K £ Z(#) and the proof is complete. | The next result is a generalization of Theorem 11.34 which includes Berger's Theorem 11.33.
194 Chapter 11 (11.37) theorem Let L^H <G with L~^G and G/L solvable. Let 9 e Irr(H), with 9G = x e Irr(G). Then there exists M oG such that M 2 L and xm is not homogeneous. Proof We may assume that there do not exist subgroups L0 £ H0 < G and 90 e Irr(H0) such that L < L0 -a G and (90)G = /. We may also assume that H is maximal in G. We have L = coreG(H). We assume that %L is homogeneous and write %L = acp with <p e Irr(L) so that (G, L, <p) is a character triple and 9 e Irr(H | <p). It follows from Lemma 11.35 that the hypotheses of the theorem remain invariant if (G, L, <p) is replaced by an isomorphic triple, and similarly for the conclusion. Therefore, by Theorem 11.28, we may assume that q>{\) = 1. Thus L £ Z(x) and in particular, L £ Z(9). Let (r, 7t) be a finite central extension of G having the projective lifting property. Let Z = kern and identify G with T/Z via n. For subgroups 1/ £ G, let I/* denote the inverse image of 1/ in T so that t/*/Z = U and G* = T. Let X/L be a chief factor of G so that X/L is abelian and K £ H. Thus KH = GandKnH<KHso that X n H = L. We have 9 e Irr(H*/Z) and L* £ Z(9). Also X*H* = T and X* n H* = L* 2 Z. Lemma 11.36 thus applies and yields r\ e Irr(r) such that h\H* = 9 for some linear k e Irr(H*) andX* £ Z(rj). We have Z = 9r = (l^H.)r = Fri and hence lr e Irr(r). Since 1L. e Irr(L*), Theorem 11.34 applies and we can find M ~3. L* with M<T and (Ar)M not homogeneous. If M = L*, let vt and v2 be distinct irreducible constituents of (lr)M- Let y. be an irreducible constituent of rjL*. Since L* £ X* £ Z(>/) we have /z(l) = 1 and thus /zvt and fiv2 are distinct irreducible constituents of (rjXr)L* = Xl*- This is a contradiction since /L* is homogeneous. Therefore, M > L*. Since (lr)M is not homogeneous, Theorem 6.11 yields a subgroup T S M with T < T and \}i e Irr(T) such that i/fr = lr. Thus z = r\lT = #r = (riT\l/)r. Since Z £ ker i we have Z £ Ysx{y\t$) and thus in G, % is induced from the proper subgroup T/Z < G. Also, T/Z 2 M/Z > L and this contradicts the first sentence of the proof. It follows that %L is not homogeneous and the proof is complete. | Problems (11.1) Let a, /? e Z(G, F *), where F is a field. If a and /? are equivalent, show that F'lG] s f*[G].
Problems 195 (11.2) Let aeZ(G, A) and let geG. Show that a(x, y) = a(y, x) for x, Note In general, xy = yx does not imply that a(x, y) = <x(y, x). (11.3) Let a e Z(G, A) and let x,yeG commute. Assume that A is divisible. Show that a(x, y) = a(y, x) iff the restriction of a to <x, y) is equivalent to the trivial factor set. Hint Use Lemma 11.16. (11.4) Let a e Z(G, A). Say that geG is a-special if a(g, c) = a(c, g) for every c e CG{g). Show that if g is a-special, then so is every conjugate of g in G. Hint Let (r, 7t) be as in Lemma 11.16 with respect to G, A, and a. Then 7t(x) is a-special iff 7t(Cr(x)) = CG(n(x)\ Note If a and /? are equivalent, then geG is a-special iff it is /^-special. Thus one can speak of a-special elements for a e H(G, A). (11.5) Let a e Z(G, A) and let X be a conjugacy class of G. Show that X consists of a-special elements (see Problem 11.4) iff there exists a function \i: jf -»/I such that l4gWg, h) = /i(ghMh, gh) for all g e X and h e G. Hint See the hint for Problem 11.4. Consider the conjugacy class of xg in T, where xg is the coset representative of A in r corresponding to g. (11.6) Let (r, 7t) be a finite central extension of G with ker n = A. Choose a set of coset representatives for A in r and let a e Z(G, A) be as in Lemma 11.9. Let X e A. For each / e Irr(r 11), choose a representation 9) affording / and construct the projective C-representation X as in Lemma 11.10. Show that this defines a bijection of Irr(r | X) onto the set of similarity classes of irreducible projective C-representations of G with factor set 1(a). (11.7) Let a e Z(G, C *). Show that Ca [G] is semisimple. Hint Show that £ (dim M)2 = | G |, where M runs over a representative set of irreducible Ca[G]-modules. Use Problem 11.6. (11.8) Let a e Z(G, F *). Show that dim Z(Fa[G]) is equal to the number of conjugacy classes of a-special elements in G. Hint Use Problem 11.5. (11.9) Let (G, N, 9) be a character triple. If g e G/N, say that g is ^-special if 9 is extendible to <iV, g, c> for all c e G with [g, c] e N. Check that this is well
196 Chapter 11 defined. Let peH(G/N,C*) be as in Theorem 11.7. Show that geG/N is 9-special iff it is /^-special. (See the note following Problem 11.4.) Note It is clear (without appeal to Problem 11.4) that conjugates of 9-special elements in G/N are 9-special. (11.10) (Gallagher) Let(G, N, 9) be a character triple. Show that | Irr(G 19) | is the number of 9-special conjugacy classes of G/N. Hint Several of the previous problems are relevant. (11.11) The character triple (G,N,9) is fully ramified if there exists X eIrr(G|9) such that (x(l)/9(l))2 = \G:N\. Suppose (G, JV, 9) is fully ramified and G > N. Show that no cyclic subgroup of G/N can be its own cen- tralizer in G/N. Hint Use Problem 11.10. (11.12) Let (G,N,9-) be a character triple. For x,yeG with \_x,y~\eN, define the complex number <x, y^ as follows: Let \ji be an extension of 9 to H = (N, y). Write ipx = ipl for X e ln(H/N) and put «x, y» = X(y). Show that C> is uniquely defined and that (a) «x, y» = «xn, >>m» for n, m e N\ (b) (xlX2, y» = «*!, y»«x2>>» whenever [xu yl [x2, y'] e N; (c) «*, y» = 9([x, y]) if 9 is linear. Note By (a) we can view <,> as being defined on commuting pairs of elements of G/N. (11.13) Let (G, N, 9) be a character triple. If N £ H £ G, i/f eCh(H|9) and g = gN e G/N we define \pB e Ch(HB | 9) by \pB(hB) = \p(h). Check that this is well defined. We say that the isomorphism. (T,(7):(G,JV,9)^(r,M,<p) is strong provided (o„(\li))m = aH9(\jiB) for all g e G/N, all H with N £ H £ G and all i/f e Ch(H 19). Show that the isomorphism constructed in Theorem 11.28 is strong. (11.14) In the notation of Problem 11.12, show that <x,.y> = (.y, x»-1 and «x, yty2^ = «x, y^ix, y2» if [x, y{\, [x, y2] e N. Hint Viewing (,)as defined on commuting pairs of elements of G/N, it is invariant under strong isomorphisms of character triples. (See Problem 11.13.)
Problems 197 (11.15) Let (G, N, 9) be a character triple. (a) If g e G and <g, x> j= 1 for some x e G, show that /(#) = 0 for all ZeIrr(G|9). (b) If g e G/N, show that g is 9-special iff <g, x^ = 1 for all x e G such that <g, x> is defined. See Problems 11.12 and 11.9 for definitions. (11.16) Let £ be elementary abelian of order p". Show that M(E) is elementary abelian of order pn{n~1)/2. Hint Consider a Schur representation group for E to get \M(E)\ < pn<n-1'/2. For equality, let {x,|l < i < n} be a generating set for E. Define ay: E x E -* C by a.i$\ x/*, Y[ */*) = e"'bj> where e is a primitive pth root of unity. (11.17) Let G = A5. Show that |M(G)| = 2. Note In fact, | M(A„) \ = 2 for all n > 4 except for n = 6, 7, where the Schur multipliers have order 6. (11.18) Let G = CH, where H and C are cyclic, C<G,andHnC = Z(G). Show that M(G) = 1. Hint If T is a Schur representation group for G, show that \G'\ = |F|. (11.19) Let (r\, 7^) and (r2, n2) be Schur representation groups for G. Define r £ r\ x T2 by r = {(*, y) | tc^x) = n2(y)} and show that Ft s F s F2. Hint Let /lj = ker nj and define 7t: r -» G by 7t(x, y) = 7ti(x), so that ker n = /lt x 42 and (F n) is a central extension of G. Use Theorem 11.19 to show that |F| < \r\ |. Note that |F: F n (/4t x A2)\ = |G'|. iVote One cannot conclude that r\ s T2 as the two nonabelian groups of order p3 show. If G = G' then rt = Tj and r\ = F2 in this case.
Character degrees Let c.d.(G) denote* the set {x(l)|xeIrr(G)}, where G is a group. In this chapter we consider what can be said about G when c.d.(G) is known. We have already seen in Theorem 6.9 that if every/ 6 c.d.(G) is a power of the prime p then G has an abelian normal p-complement. The first results in this chapter consider a weaker hypothesis, namely that p divides / for every / e c.d.(G) with/ > 1. (12.1) theorem Fix a prime p. Write ^(G) = {xeIrr(G)|p/Cx(l) and PMX)}- Let s(G) = £y(G) X(l)2- Then |Op(G)| = s(G)modp. Proof Let N = Op(G). Since Op(JV) = N, N can have no irreducible character with determinantal order divisible by p. Thus Irr(iV) = <?(N) u {tAeIrr(JV)|p|t/<l)} and hence \N\ = £,6lrr(K)M)2 = s(N) mod p. It will therefore suffice to show that s(N) = s(G) mod p. Now G acts on Sf(N) and the resulting orbits are of p-power size. Let Sf0 = {\jieSf(N)\\j/ is G-invariant}. Since all characters in an orbit have equal degrees, it follows that s(N) = ^£j,0 t/^(l)2 mod p. We show that restriction defines a one-to-one map of ^(G) onto Sf0 and this will complete the proof. If /e^(G) then (z(l), \G:N\) = 1 and so xweIrr(iV) by Problem 6.7. Thus xNeSf0. Conversely, if \\i e ^0 then by Corollary 6.28, \ji is extendible to G and a unique extension of \ji lies in ^(G). The result now follows. | * The initials c.d. stand for "character degrees." 198
Character degrees 199 (12.2) corollary (Thompson) Suppose p|x(l) for every nonlinear X e Irr(G), where p is a prime. Then G has a normal p-complement. Proof In the notation of Theorem 12,1, all / e £f(G) are linear and have kernels which contain Op'(G). It follows that ^(G) = Irr(G/G'Op'(G)) and s(G)= |G:G'Op'(G)|. Thus PJfs(G) and hence by 12.1, p^|Op(G)|. Thus Op(G) is a normal p-complement for G. | The following lemma is very useful for inductive proofs of theorems giving information about G when c.d.(G) is known. (12.3) lemma Let G be solvable and assume that G'is the unique minimal normal subgroup of G. Then all nonlinear irreducible characters of G have equal degree/and one of the following situations obtains: (a) G is a p-group, Z(G) is cyclic and G/Z(G) is elementary abelian of order/2. (b) G is a Frobenius group with an abelian Frobenius complement of order/ Also, G' is the Frobenius kernel and is an elementary abelian p-group. Proof If Z(G) t^ 1, we must have that Z(G) is a cyclic p-group and G' £ Z(G) with | G' | = p. Every / e Irr(G) with #(1) > 1 is faithful and satisfies X(l)2 = | G: Z(G)| by Theorem 2.31. If x, y e G, then since [x, y] e Z(G), we have [xp, y] = [x, y~]p = 1 since | G' | = p. Thus x" e Z(G) for all x e G. This completes the proof in situation (a). Now suppose that Z(G) = 1. Certainly, G' is an elementary abelian p-group for some prime p. Choose q \ \ G |, a prime different from p and let QeSyl,(G). Then G'QoG and it follows by the Frattini argument that G = G'N where N = NG(Q). Now N n, G' -a N since G'^aG and N r\G' <G' since G' is abelian. Thus N n G' <a G. Now Q ^ G and so N < G and N ^ G'. By the minimality of G', it follows that N n G' = 1 and thus N is abelian. If 1 t^ x e N, then N normalizes CG.(x) since N centralizes x. Since CG.(x) o G' we have CG(x)<a NG' = G. If CG(x) ¥= 1, then x centralizes G' and it follows that x eZ(G), a contradiction. It follows that CG(x) = 1 and thus G is a Frobenius group with kernel G' and complement N by Problem 7.1. We conclude that all nonlinear / e Irr(G) are of the form XG for linear X e Irr(G'). Thus | G: G' | = | N | is the common degree of all nonlinear irreducible characters of G. The proof is complete. | The way Lemma 12.3 is applied in practice is the following. Given non- abelian G, let K <i G be maximal such that G/K is nonabelian. Then (G/K)'
200 Chapter 12 is the unique minimal normal subgroup of G/K. Thus either G/K is non- solvable or else satisfies the hypotheses of Lemma 12.3. If/is a character ofG, we introduce the notation V(x) = (geG\x(g) ^ 0>. Then \(x), the vanishing-off subgroup, is the smallest subgroup, KsG such that x vanishes on G — V. Of course, V(/) <a G. The relevance of \(x) to character degrees is this. Suppose / e Irr(G) and V(z) £ JV«a G. Then by Lemma 2.29, we have [xn, %N] = \G:N\. Now write Xn = e £j= i 9( where the 9( e Irr(JV) are distinct and of equal degree (Theorem 6.2). We have |G:JV| = [xn,Xn1 = e2t and /(l) = et9(l), where 9 is one of the 9,. It follows that |G: JV|9(1)2 divides /(l)2. In particular, IG: JV | < x( I)2 and /(1) is divisible by every prime divisor of | G: JV |. The following result is useful when K -a G and G/K satisfies the conditions of case (b) of Lemma 12.3. (12.4) theorem Let K o G be such that G/K is a Frobenius group with kernel N/K, an elementary abelian p-group. Let \ji e lrr(N). Then one of the following holds. (a) |G:JV|i/f(l)ec.d.(G). (b) V(i/0 s K and thus \N:K\ divides \}i(l)2. Proof For 1 e Irr(JV/K), let T(l) denote 7G(i/fl), so that T(l) 2 N. If T(l) = JV for some 1, then (li/0G e Irr(G) and hence \G:N\ip(l)e c.d.(G). We suppose then, that T(l) > JV for all k Let W = \(ip)K so that K £ W £ JV and let S = NG(W) 2 JV. Since IG(\I/) normalizes V(t^), we have 7G(i/f) £ S. However, V(i/f) = \(X\ji) for all 1 e Irr(JV/X) since the 1 are linear, and it follows that T(l) £ S for all X. Now view JV/X as an F[S/JV]-module, where F is the field with p elements. Since G/K is a Fro benius group, we have | G/N | divides | N/K \ — 1 and thus pJf\G/N\. In particular, pJf\S/N\ and hence N/K is completely reducible as an F[S/JV]-module by Maschke's Theorem 1.9. Since W/K is a submodule, we can write N/K = (W/K) x (U/K), where S normalizes U. Now suppose W > K so that 1/ < JV. Let leIrr(JV/t/) with 1 ^ 1^. Since G/K is a Fro benius group, we have 7G(1) =(JV and thus 7S(1) = JV. Therefore, |Irr(JV/t/)| > 1 + |S/JV|. We claim that there exist distinct l,fie Irr(JV/t/) with T(l) n T(/z) > JV. If not, then since JV < T(l) £ S for all 1 e Irr(JV/t/) we have \S/N\ - 1 £ X(|T(1)/JV| - 1) :> |Irr(JV/t/)| > |S/JV|, where the sum runs over X e Irr(JV/t/). This contradiction shows that 1, /z e Irr(JV/t/) exist with T(l) n T(/z) > JV and 1 ^ /z, as claimed. Now let x e (T(l) n T(/z)) - JV and write v = 1/7. Then A* = /z*v* and ipX = (\pX)x = \pxXx = (\px/ix)vx = (\pfiYvx = ip/ivx
Character degrees 201 and thus \jiv = \jivx and we have \(\p) £ ker(vv*). Also U £ ker(vv*) and thus JV = V(t/^)t/ £ ker(vv*) so that vv* = 1^ and v* = v. Since X ± fi, we have v ^ \N and 7G(v) = JV. This contradicts x £ JV and thus proves that W = K and \(\p) £ K as desired. That this implies |JV: X||i^(l)2 follows from the remarks preceding the statement of the theorem. | In particular, in the situation of Theorem 12.4, we have for \ji e Irr(JV) that either | G: JV|i/f(l) e c.d.(G) or plt/^l). This most important consequence of 12.4 could have been obtained with somewhat less work than was needed to prove the full strength of the theorem. (12.5) theorem Let c.d.(G) = {1, m}. Then at least one of the following occurs. (a) G has an abelian normal subgroup of index m. (b) m = pe for a prime p and G is the direct product of a p-group and an abelian group. Proof Suppose there exists K<G such that G/K satisfies conclusion (b) of Lemma 12.3. Let N/K = (G/K)' so that \G: JV| = m and N/K is a p-group. Since G/K is a Frobenius group, m\(\N/K\ — 1) and so pjfm. We claim that JV is abelian. Let \ji e Irr(JV) and let / be an irreducible constituent of \pG. Then x(l) = 1 or m and i/f(l)|x(l). In particular, p^^l). By Theorem 12.4, mi/f(l)ec.d.(G) and thus i/f(l) = 1. This establishes the claim and situation (a) of the theorem holds in this case. We now suppose that no K < G as above exists. Let n be the set of prime divisors of m. By Corollary 12.2, G has a normal p-complement for every p e n. If | n \ > 1, then G has no irreducible character of p-power degree and hence G/Op(G) is abelian for all p. It follows that G' is a 7t'-group. Since the elements of c.d.(G') divide elements of c.d.(G), we conclude that c.d.(G') = {1} and G' is abelian. In particular, G is solvable. Now let K -a G be maximal such that G/K is nonabelian. Thus G/K satisfies the hypotheses of Lemma 12.3. By assumption, we are in case (a) of the lemma and this contradicts m not being a prime power. We may now assume that n = {p}. Let A be the normal p-complement of G so that A is abelian. Let k e \xx(A) and let T = IG(ty- If i> is any irreducible constituent of 1G, then by Clifford's theorem, \G:T\ divides \p(l), and thus | G: T | <, m. Now (1T)G is not irreducible and has degree <m. It follows that all of its irreducible constituents are linear and thus G' £ ker((lT)G) £ T. (Note that we have just done Problem 5.14(c).) Now if D = f|;uirru) IM, then A £ Z(£) for all £eIrr(D) and thus A £ Z(£>). Thus D = A x P for P e Syl^D). Since G' £ D, we have D -a G and hence P -a G. If G' £ P then [G, /4] £ A r\ P = 1 so that A £ Z(G)
202 Chapter 12 the result follows. Otherwise, let K 2 P, K <i G with K maximal such^ that G/K is nonabelian. Since G is solvable, the hypotheses of Lemma 12.3 are satisfied. By assumption we are in case (a) of the lemma and G/K is a p-group. Thus K 2 Op(G) = A. We have K 2 AP = D 2 G' and this contradicts G/K being nonabelian and completes the proof. | (12.6) corollary If | c.d.(G) | = 2, then G' is abelian. Proof Let c.d.(G) = {1, m). If A -a G with |G : /4| = m then G/4 can have no nonlinear irreducible characters and so is abelian. Thus if case (a) of Theorem 12.5 holds, we are done. In case (b) of 12.5, G is nilpotent and hence is an M-group by Corollary 6.14. The result follows by Theorem 5.12. | In the case that m is a prime we can sharpen Theorem 12.5 to give a necessary and sufficient condition on G that c.d.(G) = {1, m). The following results prove more than is needed for this and will be used again. Theorem 12.7 is actually a generalization of Theorem 6.16. (12.7) theorem Let JV o G and suppose 9t, 92 e Irr(JV) are invariant in G and 9^2 e Irr(JV). Let /, be an irreducible constituent of (9j)G for i = 1, 2 and let \ji be an irreducible constituent of X\li- Then W)Zi(l) St /2(l)9i(l)2- Proof We have 0 =£ [Zifo, <A] = Lfo, "AZi] and thus Xi is a constituent of tftfi. Also (za)w = to(l)/S2(l))S2 and it follows that Z2W2O) =S L^OMtfi)*] = [92(Zi)iv, >M. However, (x\)n = (Zi(l)/#i(l))#i and this yields Z2O) < zi(i) r9 9 , -, Now 9^2 is the unique irreducible constituent of 1/^ and thus [9i$2, *Pn] = t/^(l)/91(l)92(l)- Substitute this in the above and simplify to obtain the result. I (12.8) corollary Let JV o G and suppose /? e Irr(G) with JV £ Z(/J). Let Selrr(JV). Then there exists an integer b such that £>9(l)ec.d.(G) and b2 > /?(l)t, where t = |G:/G(S)|. Proof Let T = 7^9) and let y be an irreducible constituent of fiT. Then /? is a constituent of yG and so /?(1) ^ y(l)r. Also, JV £ Z(y) and we write yN = y(lU. Let ^ be an irreducible constituent of 9-T and let r\ be an irreducible
Character degrees 203 constituent of £y. Since 91 e Irr(iV), Theorem 12.7, applies and so £(1>7(1) > y(l)9(l)2. Thus ZG(l)rjG(l) > y(l)t29(l)2 > j8(l)t9(l)2. Now T = 7G(9) = 7G(91) and hence £G, rjG e Irr(G) by Theorem 6.11. Let / be whichever of £G, r\G has larger degree. Then /(1) = b&( 1) for some integer b and ft29(l)2 = Z(l)2 > ^(1)^(1) > Al)t9(l)2 and the result follows. | (12.9) corollary Let c.d.(G) = {1, p}, where p is a prime. Then there exists abelian A <i G with | G : /I | = p or p2. Proof By Theorem 12.5, we may assume that G is a p-group. Let K<G be maximal such that G/K is nonabelian and let Z/K = Z(G/X). By Lemma 12.3 it follows that |G : Z\ = p2. Let 0eIrr(G/K) with 0(1) = p and let 9eIrr(Z). Then since Z <= Z(0), Corollary 12.8 yields an integer b such that b2 > 0(1) = p and 69(1) e c.d.(G). This forces 9(1) = 1. Thus Z is abelian and the proof is complete. | (12.10) lemma Let A ^ Gbe abelian and let b = max(c.d.(G)). Then (l/|^|)XlCG(a)|>|G|/ft. aeA Proof Since | CG(a) I = £* I *(fl) 12 for X e Irr(GX we have (l/l^l)ZlCG(a)| = (l/|^|)XZlz(a)l2 = Z[^,zJ- " X a x However, iA is the sum of x(l) linear characters and hence [%A, %A~] > /(l). Thus (l/l^l)Z|CG(a)|^Xx(l)- a X Furthermore, \G\ = £z Z(l)2 ^ b £z Z(l) and thus £z /(l) > \G\/b and the result follows. | (12.11) theorem Let G be nonabelian and let p be a prime. Then cd.(G) = {1, p} iff one of the following holds. (a) There exists abelian A <i G with \G: A\ = p. (b) |G:Z(G)| = p3. Proof If (a) holds, then /(1) | p for every / e Irr(G) by Ito's Theorem 6.15. Since G is nonabelian, c.d.(G) = {1, p}.If |G :Z(G)| = p3, then x(l)2 < p3 by
204 Chapter 12 Corollary 2.30 and x(l)|p3 by 6.15 (or 3.12). Again it follows that c.d.(G) = {l,p}. Conversely, suppose c.d.(G) = {1, p} and assume that (a) is false. We may assume that G is a p-group. By Corollary 12.9, there exists abelian A <i G with | G: A \ = p2. By Lemma 12.10, we have (l/|,4|)X|CG(a)|>|G|/p. aeA Now A acts by conjugation on G — A and by Corollary 5.15, the number of orbits of this action is (i/|/H)E(|Cri(fl)|-|/i|)£(|G|/p)-M|. asA It follows that the average size a of these orbits satisfies *<(\G\-\A\)/((\G\/p)-\A\) = p+l<p2 and so A has an orbit of size 1 or p on G — A, Since G does not have an abelian subgroup of index p, A = CG(A) and thus there exists x e G — A in an orbit of size p. Let K = (A, x>. Then Z(X) = CA(x) has index p in A and index p3 in G. We shall show that Z(G) = Z(X) to complete the proof. Let Z = Z(X). If / e Irr(G) and %K is irreducible, then Z £ Z(/) and [G, Z] c ker /. On the other hand, if %K reduces, then all irreducible constituents are linear and K' c ker /. Suppose [G, Z] > 1. We also have K' > 1 but K' n[G,Z]^ ker / for every / e Irr(G) so that K' r\[G,Z] = 1. Now X'[G, Z] is the direct product of two nontrivial groups and so has an irreducible character 9 with K' £ ker 9 and [G, Z] £ ker 9. Let / be an irreducible constituent of 9G. Then K' £ ker / and [G, Z] £ ker /. This contradiction shows that [G, Z] = 1 and completes the proof. | We remark that the last several sentences of the proof could be replaced by an appeal to Problem 5,26. Next we refine Theorem 12.5 in a somewhat different direction. Namely, if c.d,(G) = {1, m} and G has no abelian normal subgroup of index m, then the nilpotence class of G is < 3. (12.12) lemma Let A<i G with A abelian and G/A cyclic. Then \A\ = \G'\\AnZ(G)\. Proof Let G/A = (Ag} and let a: A -> A be defined by o(a) = a~la". Then a is a homomorphism and ker a = CA(g) = A n Z(G). Let / be the image of a. Then g e N(7) and so I <i G. Since G = (A, g) and g centralizes A mod 7, it follows that G/I is abelian and G' s /. Clearly, I £ G' and hence |/l| = |ker(T||/| = |/lnZ(G)||G'| and the proof is complete. |
Character degrees 205 (12.13) lemma Let c.d.(G) = {1, m} and suppose A, B £ G are abelian of index m with A ± B. Then \G'\ < m and G' £ Z(G). Proof If X £ G with | G: X | < m, then G'sK and hence K<G by Problem 5.14(c). In particular, A <aG. Let A < K £ G, then K<G and thus K'<G. If X' < G', let 1G. # /z e Irr(G'/K') and let / be an irreducible constituent of fiG. Now K' £ ker / and so %K has a linear constituent 1 and /(I) < 1G(1) = |G: X| < m. Thus *(1) = 1 and G' £ ker /. It follows that G' £ ker fi, contradicting the choice of fi. Therefore K' = G'. Since A ^ B, let b e B - A and let X = </l, b). Thus K//1 is cyclic and \G'\ = \K'\ = \A:A nZ(X)| by Lemma 12.12. However, A n Z(X) 2 /4 n B and so |-4;/ln Z(X)| <|/lMnB|<|G:B| = m and the first assertion is proved. For g e G, the conjugacy class of g is contained in the coset gG' and so has size < \G'\ < m. Thus \G:C(g)\ < m and G' £ C(g) by the first sentence of the proof. Since g is arbitrary, we have G' £ Z(G) and the proof is complete. | (12.14) theorem Let c.d.(G) = {1, m} and suppose that G has no abelian normal subgroup of index m. Then [G, G'] £ Z(G), that is, G is nilpotent of class ^3. Proof By Theorem 12.5 we may assume that G is a p-group. If G has a faithful irreducible character / then / = 1G for linear 1 e Irr(X) and K £ G since G is an M-group. We have | G: K | = /(l) = m and so X o G by Problem 5.14(c). Thus all irreducible constituents of Xk are linear and K' £ ker / = 1, a contradiction. Thus G has no faithful irreducible character and hence Z(G) is not cyclic. Let Zt, Z2, Z3 £ Z(G) be distinct subgroups of order p. If G/Zt is of nil- potence class <3, then [G, G, G, G] ^ Zt. If this happens for two distinct Z;, we conclude that [G, G, G, G] = 1 and we are done. Assume then that G/Zt and G/Z2 (say) do not have class < 3. Working by induction on \G\, it follows that there exist A, B<i G with |G: A\ = m = | G: B| and /4' £ Zt and B' £ Z2. Let Z = Zt Z2 so that AZ/Z and BZ/Zare abelian. Since G/Z is nonabelian, we have m e c.d.(G/Z) and so | G: AZ\ >m and | G: BZ| > m. It follows that Z £ /I n B. UA/Z = B/Z, then A' £ Zt n Z2 = 1 and /I is abelian, a contradiction. Thus /4/Z ^ B/Z and Lemma 12.13 applies to yield G'Z/Z £ Z(G/Z). Thus [G', G] £ Z £ Z(G) and the proof is complete. |
206 Chapter 12 We now consider groups G for which |c.d.(G)| = 3. Although the information obtained is not as detailed as when |c.d.(G)| = 2, we do prove a result analogous to Corollary 12.6. (12.15) theorem Let |c.d.(G)| = 3. Then G" = 1, that is, G is solvable of derived length < 3. Before proceeding with the proof of this result we make some general remarks. As the group A5 shows, we cannot conclude from |c.d.(G)| = 4 that G is solvable. It has been conjectured by G. Seitz that for solvable G, d.l.(G) < |c.d.(G)|. (Here, d.l.(G) is the derived length of G; the smallest integer k for which G(k\ the kih commutator subgroup of G, is trivial.) By Theorem 5.12, this conjecture holds for M-groups. It has also been proved when | cd.(G) | = 4 (S. Garrison) and when | G | is odd (T. Berger). It is known that for any solvable group, d.l.(G) < 3|c.d.(G)|. (12.16) lemma Let JV-aGwith(|JV|,|G:iV|)= 1. Then |c.d.(JV)|<|c.d.(G)|. Also if G/N is supersolvable and N is solvable with d.l.(iV) < | c.d.(iV) |, then d.l.(G)< |cd.(G)|. Proof Let n be the set of prime divisors of | JV |. If / e Irr(G), let 9 be an irreducible constituent of %N. By Corollary 11.29, /(1)/9(1) divides |G:JV|. Since 9(1) divides |JV|, it follows that 9(1) is exactly the rc-part of x(l). Since every 9 e Irr(iV) arises this way, we see that c.d.(JV) is exactly the set of rc-parts of the elements of c.d.(G). The first assertion follows. Now assume that N is solvable and G/N is supersolvable. It follows that G/N' is an M-group by Theorems 6.22 and 6.23 and thus d.l(G/N') < \c.d.(G/N')\ by Theorem 5.12. We may assume that N' > 1 and observe that d.l.(G) <; d.l.(G/JV') + d.l.(JV') < |c.d.(G/JV')| + d.l.(iV) - 1 < |c.d.(G/JV')| + |c.d.(JV)| - 1, where the last inequality follows from the assumption that d.l.(N) < \ c.d.(N) \. Now every/ ec.d.(G/iV') divides |G:iV| by Ito's Theorem 6.15 and thus the 7t-part off is trivial. We conclude from the first part of the proof that I c.d.(JV) | < | c.d.(G) | - | c.d.(G/iV') | + 1 and the result follows. | Proof of Theorem 12.15 Let c.d.(G) = {1, m, n}. If (m, n) ^ 1, then G has a proper normal p-complement N for some prime p by Corollary 12.2. By Lemma 12.16, |c.d.(N)| < 3 and so N is solvable and d.l.(N) <, | c.d.(JV) | by induction if | cd.(iV) | = 3 and by Corollary 12.6 if | cd.(iV) | < 3. In this case we are done by Lemma 12.16.
Character degrees 207 Assume now that (n, m) = 1. Suppose there exists K<G with G/K solvable and nonabelian. We may assume that G/K satisfies the hypotheses of Lemma 12.3, If G/K is a p-group then (say) n is a power of p. Let z e Irr(G) with %(1) = m. Since pjfm, we have %K e Irr(X) and thus by Gallagher's theorem (Corollary 6.17), //? e Irr(G) for /? e lrr(G/K), This is a contradiction since mn $ cd.(G). Therefore G/K is a Frobenius group with kernel N/K = (G/K)', where N/K is a p-group and \G : N\ = n (say). Suppose \ji e lrr(N) with i/f(l) > 1. Since ni/f( 1) £ c.d.(G), we conclude from Theorem 12.4 that p | \j/{ 1). Let z be an irreducible constituent of \jiG. Then plz(l) and since pjfn, we have z(l) = m. Thus Xn is irreducible since (|G:JV|, x(l)) = 1. We conclude that %N = \ji and \p(l) = m. Thus c.d.(JV) = {1, m} and dl(JV) < 2 by Corollary 12.6. Since G/JV is abelian, we have d.l.(G) < 3 as desired. Suppose now that no such K exists. If/ elrr(G) with z(l) ^ l,thenG/kerz is nonabelian and thus is nonsolvable. However | c.d.(G/ker /) | < 3 and working by induction on \G\ we must have ker / = 1. Therefore, every nonlinear irreducible character of G is faithful. Now let n < m and let %eIrr(G) with z(l) = «• By Theorem 4.3, each irreducible character of G is a constituent of z' for suitable integers t. Let t be minimal such that z' has an irreducible constituent \p of degree m. Thus Ez£, <A] t6 0 for some irreducible constituent £ of z'~ '• Now <j;(l) j= 1 or else z£ is irreducible forcing z£ = tA which is not the case. Also ^(1) # m by the minimality of t. Thus £(1) = n. Since [zi, t/f] = 0 for linear 1, we conclude that [1, i/^] = 0 and \p% has no linear constituents. Thus a b •AZ = E & + E fj' i= 1 j= 1 where £t, >/; e Irr(G), £,-(1) = n, >/j{1) = m and a > 1 since £ is one of the <!;,. Comparing degrees yields mn = an + bm and since (m, n) = 1, we have m\a. Since ft > 0 and a > 0, this yields a = m and ft = 0 and t/^z = E?= i &• We claim that each ^ is of the form 1;Z for some linear character, Xt. It suffices to find a linear constituent of xii ■ Suppose (for some i) that x£t has no linear constituent. Then r s j= 1 k=l where Zj, iK £Irr(G), z/1) = » and t/^(l) = w. Thus n2 = rn + sm and n\s. However, 0 j= [t/^z, ^] = [i/f, z£/] and \ji is one of the i/ft. Thus s > 1 and hence s > n. Since r > 0, this yields n2 > nm and n > m, a contradiction. We now have & = l,z for linear 1( and thus f^ = E £i = X E ^i an(* hence ('/'g')(Zg') = wzG'-
208 Chapter 12 Since ker \j/ = 1, we have il/(x) ¥= m for x ¥= 1 and thus x(x) = 0 for all x e G' - {1}. It follows that [/G-, 1G.] ^ 0 and thus G' £ ker x- Thus contradiction completes the proof. | Another technique for studying character degrees is based on the following lemma. (12.17) lemma (Garrison) Let Hs G and let 9e\xx(H). Suppose for every irreducible constituent / of 9G that Xn = $• Then \(9) <i G. Proof We have (9G)„ is a multiple of 9 and so (9G)H = |G: H|9. Let heH with 9(/j) ^ 0 and let S = {x eG|/i*eH}. By definition of 9G we have (l/\H\)ZHhx) = SG(h)=\G:H\$(h)- xeS However, since 9G(/i*) = 9G(h), it follows for x e S that 9(h*) = 9(/j). This yields \S\9(h) = \G\9(h) and since 9(h) ± 0, we have \S\ = \G\ and S = G. Thus ft'eH for all g eG and 9(hg) = 9(h) ± 0. Thus G leaves the generating set for \(9) invariant under conjugation and the result follows. | (12.18) lemma Let x g Irr(G) and let ker / < JV o G. Then ker x < N n \(x). Proof We may assume ker / = 1. If V(/) n JV = 1 then / vanishes on N — {1} and so [/^, 1N] ^0. This forces JV £ ker /, a contradiction. | In the following, F(G) denotes the Fitting subgroup of G, the (unique) largest normal nilpotent subgroup. (12.19) theorem (Broline-Garrison) Let/elrr(G)andX = ker/.Either of the following conditions guarantees the existence of t^elrr(G) with t/f(l)>X(l)andkert/f <K: (a) K£F(G); (b) K = F(G), G/K is solvable and K < G. Proof Suppose H is a maximal subgroup of G and that KH = G. Then Xh = 9 e lrr(H). If \\i is an irreducible constituent of 9G such that \jiH ^ 9, then i/'h reduces and t/^(l) > 9(1) = /(l). Also, we cannot have (ker \j/)H = G or else \jiH would be irreducible. The maximality of H thus yields ker \ji £ H and since 9 is a constituent of i/fH, we have ker \ji £ ker 9 = H r\K < K. Thus the result follows in this situation and we may suppose that whenever KH = G for a maximal subgroup HsG, the character 9 = Xh satisfies the hypotheses of Lemma 12.17. In particular, \(9) -o G.
Character degrees 209 Now we choose L <a G as follows. If K £ F(G), take L = K. If K £ F(G), then we are in situation (b) and F(G) = K < G. Here, G/K is solvable and we take L > K with L/K an elementary abelian chief factor of G. Thus L is not nilpotent and we choose a nonnormal Sylow subgroup P of L. Note that in the case K = F(G) and L > K, we must have L = KP. By the Frattini argument, G = LNG(P) = KNG(P) and NG(P) < G. Let H 2 NG(P) be a maximal subgroup of G. Thus G = HX and we let 9 = /H. Then V(9) -a G and thus V(3) nL<G. Now K r\H = ker 9 £ V(9) and so in the case K = L, we have P ^ K r\H ^ L r\ V(9). In the case that L/K is a chief factor of G, we have (L n H)/(ker 9)isachieffactorofHandthusL n H £ V(3) by Lemma 12.18. Thus in any case, P s L n V(S)<GandP is Sylow in L n V(S). The Frattini argument now yields G = (L n V(9))NG(P) £ H. This is a contradiction and proves the theorem. | (12.20) corollary Let /elrr(G). If either x(l) = max(c.d.(G)) or ker/ is minimal among kernels of irreducible characters of G then ker / is nil- potent. (12.21) corollary (Garrison) Let G be solvable and let | c.d.(G) | = n. Then there exist JV( -a G with 1 = N0 £ Ni £ • • • £ N„ = G such that Ni+ JNi is nilpotent for 0 < i < n. Proof Let Nl = F(G), the Fitting subgroup. Then c.d.(G/JV!) £ c.d.(G) and if JVt < G, then cd^G/iVJ does not contain the largest element of c.d.(G) by Theorem 12.19. In this case, \c.d,(G/Nl)\ < n and the result follows by induction on | G |. | Suppose A £ G is abelian. By Problem 5.4 (or 2.9(b)), | G : A \ is an upper bound for c.d.(G). Conversely, suppose we know max(c.d.(G)). Can we conclude that there exists an abelian subgroup with bounded index in G? We can, although it is certainly not true that there necessarily exists abelian A £ G with \G:A\ = max(c.d.(G)). We use the notation b(G) = max(c.d.(G)). Note that if H £ G and \ji e Irr(H), then \ji is a constituent of 1h f°r some / e Irr(G), and thus ^(l)<Z(l)<ft(G) and hence b(H) < b(G). (12.22) lemma Let b(G) = b. Then there exists xeG- {1} such that |G:CG(x)|<fc2.
210 Chapter 12 Proof Let k = | Irr(G) | - 1. Then |G|-l = £z(l)2</cft2, where the sum runs over nonprincipal / e Irr(G). Let m = min{|G:CG(x)||xeG,x ¥= 1}. Then | G | — 1 > mk since each of the k nonidentity conjugacy classes has size > m. We now have kb2 > mk and the result follows. | (12.23) theorem Let b(G) = b. Then G has an abelian subgroup of index <(ft!)2. Proof Use induction on b. If b = 1, the result is trivial and if b = 2, then c.d.(G) = {1, 2} and we are done by Corollary 12.9. Assume then, that b > 3. Let K <i G be maximal such that G/K is nonabelian. It follows that b(K) <, b/2 since if \ji e Irr(X) with \ji(\) > b/2, then necessarily \ji = %K for some zelrr(G). By Corollary 6.17, we have 0xeIrr(G) for 0eIrr(G/K). Since G/K is nonabelian, we may choose /? with 0(1) > 2 and this yields 0(1)2(1) > ft, a contradiction. By Lemma 12.3, we have three main cases to consider namely: G/K is non- solvable, G/K is a Frobenius group and G/K is a p-group. Case I G/K is nonsolvable. Since b[G/K) < ft, Lemma 12.22 yields x e G/K such that x ¥= 1 and |(G/K):CG/K(x)|<ft2. Let C/K = CG/K(x). If b(C) < b - 1, then there exists abelian A s C with \C:A\ <((b- l)!)2andso|G:/4| = \G:C\\C:A\ < (ft!)2 and we are done. Assume that b(C) = b and let \ji e Irr(C) with t/^(l) = b. Then every irreducible constituent / of \jiG satisfies /(l) = b and thus %c = \p. It follows by Lemma 12.17 that V(t/0 <G and thus XV(tA) <iG. If KV(t/0 > K, then G/KV(t/0 is abelian and soC<G since KV(t/0 s C. If Z/K = Z(C/K) then Z <i G. Also, x e Z/K and so Z > X. It follows that G/Z is abelian and G/K is solvable, a contradiction. We conclude that V(t^) c x and hence \C:K\ < tA(l)2 < b2 by the remarks preceding Theorem 12.4. Thus \G:K\ <b\ By the inductive hypothesis, K has an abelian subgroup of index <(rj>/2]!)2. For b > 5, we have ([ft/2]!)2ft4 < (ft!)2 and the result follows. Assume then, that ft < 4. If 2 £ c.d.(G/X), then c.d.(G/X) s {1, 3,4} which contradicts the nonsolvability of G/K by Theorem 12.15. Thus we may choose 9 e Irr(G/X) with 9(1) = 2. We have G/ker 9 is nonabelian and hence ker 9 = K. Let M/K = (G/K)'. Then M' £ X = ker 9 and hence 9M is
Character degrees 211 irreducible. Pick y e M/K of order 2. Since y e ker(det(9)), it follows that yeZ(9) = Z(G/K) = 1, a contradiction. This completes case 1. Case 2 G/K is a Frobenius group. Let N/K = (G/K)', the Frobenius kernel. By Lemma 12.3, \G:N\e cd.(G) and so |G :N\ < ft. If b(N) < ft/2, then application of the inductive hypothesis to N yields an abelian subgroup A £ N with \N :A\ < ([ft/2]!)2. Since ft([ft/2]!)2 < (ft!)2 we are done. Suppose then, that ft(N) > ft/2 and let \j/e lrr(N) with t/^(l) > ft/2. Then ip(l)\G:N\^c.d.(G) and by Theorem 12.4 it follows that t^(l)2 > \N:K\ and hence | G: K \ < ft3. Since ft > 3, we have ([ft/2] !)2ft3 < (ft!)2 and we are done in case 2 by applying the inductive hypothesis to K. Case 3 G/K is a p-group. Let Z/K = Z(G/K). By Lemma 12.3, \G:Z\ = 0( l)2, where 0eIrr(G) and Z £ Z(0). In particular, | G: Z | < ft2.Ifft(Z) > ft/2, pick i// e Irr(Z) with \p( 1) > ft/2. Then \p = Xz f°r some / e Irr(G). It follows from Theorem 12.7 or 6.16 that /?(l)t/f(l) ec.d.(G) and this is a contradiction since /?(1)^(1) > ft. Thus ft(Z) < ft/2 and the result follows as in previous cases since ([ft/2]!)2ft2 < (ft!)2. The proof is now complete. | It is apparent in the above proof that in cases 2 and 3, the inequalities obtained are far from being best possible. The limiting factor in this proof is the nonsolvable case 1. If one assumes that G is solvable it is possible to obtain a better bound. It is not known what the best possible bound is either in the general situation or for solvable groups. The following result provides a tool which can be used to find an abelian subgroup in G of index <ft(G)4 when G has an abelian normal subgroup with nilpotent factor group. (Compare Theorem 12.24 with part (a) of Theorem 12.19.) (12.24) theorem (Broline) Let /e Irr(G) and let K = ker / and F/K = F(G/K). Suppose F is not nilpotent. Then there exists \ji e Irr(G) with ker \ji < K. Proof Let Q e Syl,(F) with Q -^ F and let H 2 NG(Q) be a maximal subgroup. Now QK <i G since F/K is nilpotent and normal in G/K and hence HK = G by the Frattini argument. Thus 9 = in e Irr(H). Since ker(9G) <=, H, we have ker(9G) £ K. Let \ji be an irreducible constituent of 9G with K £ ker \p and let L = ker \ji. If L £ H, then L £ ker 9 = H n K < K and we are done. Suppose then, that L £ H so that LH = G and \jiH is irreducible. Then \jiH = 9 and L r\ H = ker 9 = X n H. Write JV = ker 9. Now KnL<GandK>KnL2KnH.It follows that (K n,L)H < KH = G and thus K n, L s H. Therefore K n L = K n, H = N
212 Chapter 12 We have \F:FnH\ = \G:H\ = \L:N\ = \KL:K\. Since Q £ F r\ H and QeSyl,(F), it follows that qJC\KL/K\. Since XQ/X -a G/X, we conclude that the commutator [XL/X, XQ/X] = 1 and thus [L, Q] £ X. Since L -a G, we have [L, Q] £ X n L = N and hence L £ N(JVQ). However, NQ = (H n, K)Q = H r\KQ~aH since XQ -a G. Therefore, G = LH <= N(JVQ) and JVQ -a G. The Frattini argument yields G = JVN(Q) £ H, a contradiction. | (12.25) lemma Let G" = 1 and let m = |G:F(G)|. Then m < b(G) and b(F(G)) <L b(G)/m. Proof Let / e Irr(G) be such that ker / is minimal. Since G is a relative M-group with respect to G' which is abelian, there exists H 2 G' and linear X e Irr(H) such that kG = %. Also, H 2 ker /. Since H -a G, all irreducible constituents of Xh are linear and thus H/ker / is abelian. Thus H/ker / £ F(G/ker /) and hence H is nilpotent by Theorem 12.24. In particular,// £ F(G)andm < \G:H\ = /(l) < b(G). Now let 9eIrr(F(G)) and let \\i be an irreducible constituent of 9G. Choose x £ Irr(G) with minimal ker / £ ker \ji and let H be as above with | G: H\ <L b(G), H/ker / abelian and H £ F(G). We have W £ ker / £ ker \p and thus t/^H has linear constituents. Thus 9H has linear constituents and hence 9(1) < | F(G): H \ = | G: H \ jm < ft(G)/w and the proof is complete. | (12.26) theorem Suppose U -a G is abelian and G/U is nilpotent. Then there exists abelian A £ G such that | G: A | < fe(G)4. Proof Use induction on b = b(G). We may assume b > 1 and so G is not abelian. First, suppose that every nilpotent factor group of G is abelian. Then G' £ U, G" = 1 and Lemma 12.25 applies. Since G is not nilpotent, F(G) < G and b(F(G)) <, b/m < b, where m = \G: F(G)| <; ft. By the inductive hypothesis, there exists abelian A £ F(G) with |F(G):/1| < (b/m)4 and hence \G:A\< b*/m3 < b4. Now suppose that G does have a nonabelian nilpotent factor group. Choose X <a G, maximal such that G/X is nonabelian and nilpotent. By Lemma 12.3, G/X is a p-group and |G: Z\ = /2, where Z/X = Z(G/X) and Z = Z(0) for some 0 elrr(G) with 0(1) = /. By Corollary 12.8, we conclude that b(Z) ^ b/f1'2 < b. By the inductive hypothesis, there exists abelian A sZ, with |Z:/4|^ft(Z)4^ft4//2- Thus |G:/4| = \G:Z\\Z:A\ £ b4 and the proof is complete. |
Character degrees 213 Next we discuss the "p-structure" of G, where p is a prime which is in some sense large when compared to b(G). The objective here is to show that | G: Op(G) | is not divisible by too large a power of p. We prove that ifft(G) < p, then pjf\G: Op(G)\ and that if b(G) < p3/2, then p2X\ G: Op(G)|. (12.27) lemma Let G act transitively on a set Q with |Q| > 1. Let H = Ga for some a e Q. Then the average size of the orbits of H on Q — {a} is < b(G). Proof Let 9 = (1H)G, the permutation character of G on Q. Write $ = 1g + X flx& where the sum runs over the nonprincipal irreducible constituents of 9. The number of orbits t of H on Q — {a} is given by t = Now |fi - {a}| = 9(1) - 1 = 2Xx(l) < b Xaz, where ft = fc(G). Thus the average orbit size s is given by s = (I «*x(i))/(I ax2) < (b x <g/(E«,) = *>• I (12.28) corollary If G has more than one Sylow p-subgroup, then there exist distinct P, Q e Sylp(G) such that b(G) > |NG(P): NG(P) n NG(Q)| > |P:P n Q|. Proof Apply Lemma 12.27 to the conjugation action of G on Sylp(G). Let P e Sylp(G) and H = NG(P). Then some orbit of H on Sylp(G) - {P} must have size <b(G). Let Q be in such an orbit. Then b(G) >\H: NH(Q)| and we have the first inequality. Also \H:N„(Q)\>\P:NP(Q)\ and since NP(Q) = P n Q, the result follows. | (12.29) theorem Let p be a prime and let b(G) < p. Then G has a normal abelian Sylow p-subgroup. Proof Let P e Sylp(G). If P ^ G, then by Corollary 12.28, we can find eeSylp(G) such that Q ^ P and |P:P n Q\ < b(G) < p. This is a contradiction and shows P <i G. That P is abelian follows since every/ e c.d.(P) is a power of p satisfying / < p. The proof is complete. | (12.30) lemma (Burnside) Let P e Sylp(G) and let X, Y £ P be normal subsets of P which are conjugate in G. Then X and Y are conjugate in NG(P). Proof Suppose 7 = X9 so that P £ N(Y) and P9 £ N(X)9 = N(X9) = N(Y). By Sylow's theorem in N(Y), we have P9" = P for some weN(Y) and gu e N(P). However, X9" = Y" = Y and the proof is complete. | (12.31) lemma LetHs Gwith|G:H| = p, a prime. Then Op'(H) <i G.
214 Chapter 12 Proof Let K be the kernel of the action of G on the right cosets of H (acting by right multiplication). Then K ^ H and K<G. Also \G:K\ divides p! and so \H: K | divides (p - 1)! and is prime to p. Thus Op'(H) £K<H and so Op'(H) = Op'(K) -a G since K<G. | (12.32) theorem Suppose b(G) < p3/2 for some prime p. Then p2^|G:Op(G)|. Proof We may assume that Op(G) = 1. Let PeSylp(G) and assume \P\ > p2.LetJV = NG(P). By Corollary 12.28, choose Q ± P, with Q e Sylp(G) such that p3/2>ft(G)>|iV:Nw(0|>|P:PnQ| and set D = P n Q. Then |P: D\ = p. Let M = NG(D) 2 P. Now P £ N(Q) and so N„(Q) < PN„(Q) £ JV n M. It follows that p3/2 > IJViNjrfQ)! >p\N:Nr\M\ and |JV: N n M| < p1/2. Note that M < G. We consider the action (by right multiplication) of M on fi = {Mx | x e G}. Suppose r of the orbits of M on fi — {M} have size < p2 and s have size > p2. Let fi0 be the union of the r smaller orbits. We claim (*) If Mx e fi0, then DDX = DXD and x e MNM. Assuming (*), let us complete the proof. We have |fi — {M} | > p2s so that Lemma 12.27 yields p3/2 > b(G) > \Q - {M}\/(r + s)> p2s/(r + s). If s = 0, then MxeQ0 for every xeG — M and hence DDX = DXD for all x e G by (*). Thus <£>* | x e G> is a p-group that is normal in G, a contradiction. Thus s > 0 and p1/2 < (r + s)/s < 1 + r. Now if Mx eO0, then x e MNM because of (*) and hence the orbit of Mx under M contains an element of the form Mn for n e N. The number of distinct Mn for neN is | iV: N n M | < p1/2 and thus at most p1/2 M-orbits of Q contain an element of the form Mn. These include the trivial orbit {M} and so we have 1 + r < p1/2. This contradicts a previous inequality. We now work to establish (*). Let Mx e Q0 so that x $ M and D ^ Dx. If either of D or Dx normalizes the other, then DDX = DXD is a p-group properly containing D. Since \P:D\ = p it follows that DDX e Sylp(G) and |D£>*: D\ = p so that £> <i £>£>* and DDX ^ M. Since also P c m we have P = (DDT for some m e M by Sylow's theorem. Now Dm, Dxm <i P and
Character degrees 21S hence Lemma 12.30 yields Dm = Dxmn for some n e JV. Thus xmnm l e N(D) = M and x e MNM as desired. The remaining case is where D £ Mx and Dx £ M. Let W = M n Mx so that WD and WDX are groups. Since Mx e Q0, we have \M :M n Mx\ < p2 and hence \DW:W\ < \M:W\ < p2 and |DW:W| = p. Similarly |D*W: w\ = p. By Lemma 12.31, Op'(W) is normal in both WD and WD*. Write X = Op'(W) and L = NG(X). Since |M:W| < p2 and |DW: W| = p it follows that |M:DW| < p and DW contains a full Sylow p-subgroup of M and hence of G. It follows that DK/K and DXK/K are Sylow subgroups of L/K. Also, D*X ^ DX since DK^M and D* £ M. Thus |Sylp(L/X)| > 1 and b(L/K) > p by Theorem 12.29. By Corollary 12.8, b(L) > ab(K) for some a with a2 > b(LIK) > p. Thus p3/2 > b(L) > pll2b(K) and b(K) < p. Thus X has a normal Sylow p- subgroup U by Theorem 12.29. Now UD e Sylp(M) and UD <i WD since 1/ is characteristic in X <i WD. Thus |M:NM(t/D)| < |M :DW\ < p and it follows that UD <i M. Thus | Sylp(M) | = 1 which is a contradiction since P,Qe Sylp(M). This completes the proof. | We close this chapter by considering the opposite of the stituation with which we began. Suppose no/ e cd.(G) is divisible by the prime p. A sufficient condition for this to happen is that G has a normal abelian Sylow p-subgroup. (This follows by Ito's Theorem 6.15.) It is conjectured that this condition is also necessary. The next result shows that to prove the conjecture, it would suffice to check simple groups. (12.33) theorem Suppose G does not have a normal abelian Sylow p- subgroup and that no element of c.d.(G) is divisible by p. Then G has a non- abelian simple composition factor S of order divisible by p such that no element of c.d.(S) is divisible by p. Proof If N <i G and t/^eIrr(iV), choose /elrr(G) with [zN,^] ^ 0. Then t/^(l)|x(l) and so p)(\ji(\). In particular, if JV e Sylp(G), then JV is necessarily abelian and thus G does not have a normal Sylow p-subgroup. Also, if G is simple, the result is trivial and we assume G is not simple. Let JV be a maximal normal subgroup of G. Working by induction on | G |, we may assume that JV has a normal Sylow p-subgroup P. Then G/P does not have a normal Sylow p-subgroup and if P > 1 we complete the proof by applying the inductive hypothesis to G/P. Suppose then, that P = 1. We must have p 11G: JV | and since G/N is simple we can take S = G/N unless G/N is abelian, that is, | G/N \ = p. We suppose this is the case. Let QeSylp(G)sothat|Q| = p.
216 Chapter 12 If \j/ e lrr(N), then i/fG cannot be irreducible since p | \jiG(\). Thus /G(i/0 > JV and hence 7G('/') = G and \ji is invariant in G. Thus Q acts trivially on Irr(G). It follows by Brauer's Theorem 6.32 that Q acts trivially on the set of con- jugacy classes of JV. If jf is a class of JV then Q permutes jf. Since pjf\ N\, we have pX\tf\ and thus Q fixes an element of Jf. It follows that C = CN(Q) meets every conjugacy class of JV. Thus JV = [jxeN Cx and this yields |JV| - 1 < |JV:N^(C)|(|C| - 1) < \N:C\(\C\ - 1) = \N\ - \N:C\. Thus | JV: C | < 1 and C = JV. We conclude that Q o G, a contradiction. | (12.34) corollary {ho) Let G be solvable. Then G has a normal abelian Sylow p-subgroup iff every element of c.d.(G) is relatively prime to p. Problems (12.1) The normal subgroups JV,- o G are a S_y/ow tower if 1 = JV0 c JVt c ... c Nk = G and JV,+ i/JV; is a Sylow subgroup of G/N; for each i, 0 < i < k. Suppose for every m,ne c.d.(G), either m \ n orn | m. Show that G has a Sylow tower. (12.2) Suppose that every/ ec.d.(G) is a power of the integer m. Assume that m is not a prime power. Show that there exists abelian A<=,G with | G: A \ = b(G) and that such an A is necessarily normal in G. Hint This generalizes part of Theorem 12.5. Mimic the proof of that theorem. (12.3) Suppose that G is solvable and that for every m, nec.d.(G) with m t6 n, we have (m, n) = 1. Show that |c.d.(G)| < 3. (12.4) Let G be solvable with b(G) = b > 1 and suppose that G has no factor group which is a nonabelian p-group. Show that there exists L ^ G and an integer r with 2 <r < b such that b(L) < b/r and | G: L | < br. (12.5) Let G be solvable with b(G) = b. Show that G has an abelian subgroup of index <kbloi2{b) for a suitable constant k independent of b. (12.6) Let c.d.(G) = {1, pe}, where p is a prime and e > 1. If a Sylow p- subgroup of G is nonabelian, show that G is nilpotent. Hint Use the fact that abelian Frobenius complements are cyclic to show that if G is not nilpotent, then there exists abelian H <i G with \G:H\ = pe and G/H cyclic. Now let G/K be as in Lemma 12.3(a) and consider HK. (12.7) Let G be solvable with b = b(G). Let p be a prime.
Problems 217 (a) Show that there exists proper K<G and integer e > 0 such that pel4b(K) <b and pe+1Jf\G:K\. (b) Show that if pf 11G : Op(G) |, then pf < b4. (12.8) Let G" = 1 and assume that all Sylow subgroups of G are abelian. Show that b(G) = \ G : A | for some abelian A o G. (12.9) Let G be solvable and suppose that all Sylow subgroups of G are abelian. Show that | G: F(G)| < b(G)2. Hints F(G) = CG(F(G)). Let leIrr(F(G)) be such that |G:/G(1)| is maximal. Let T = IG(1) and s = \G: T\. Show that b(T/F(G)) < b/s and let F/F(G) = F(T/F(G)). Show b(F) < s. Use Corollary 11.32. (12.10) Suppose that every/ e c.d.(G) divides pe where p is a prime. Show that there exists abelian A ^ G with index dividing p4e. (12.11) Suppose 0eIrr(G) with 0(1) = / and that Z = Z(0) satisfies |G:Z| = f2. LetZ <= H £ G. (a) If | G: H | > / show that fc(tf) < fc(G). (b) If | G: H| = /and / e Irr(H) with /(l) = ft(G), show that \(x) <= Z. (12.12) (a) Suppose A £ G is abelian and |G: CG(A)| > b(G). Show that there exists A0 ^ A such that | A : A0 \ < b(G)2 and CG(A0) > CG(A). (b) If G is a p-group, improve part (a) to read \A : A01 < b(G), Hint Use Lemma 12.10. (12.13) Show that there exists a function/defined on positive integers such that for any group G if b(G) = b, then there exists H £ G with | G: H\ <b and |tf:Z(tf)| </(&). Hint Use repeated applications of Problem 12.12. (12.14) Let G be solvable and let p be a prime. Suppose p2Jff for all / e c.d.(G). Show that either Op(G) > 1 or a Sylow p-subgroup of G is abelian. Hint In a minimal counterexample, let M be a minimal normal subgroup. Show that | Op(G/M)| = p. Now show that Op(G/M) is a direct factor of a Sylow subgroup of G/M. Produce the other factor by considering 7G(1) for suitable X e Irr(M). (12.15) Let N<i G with G/N a p-group and p ^ 2. Let Selrr(iV) be invariant in G. Suppose that every irreducible constituent of 9G has degree 5Sp9(l). Show that b(G/N) < p. Hints Extend 9 to §eIrr(H) with JVcJf and \G:H\ = p. For <p e lrr(H/N) with <p(l) = p, consider (<p9)G. Conclude that there exists linear
218 Chapter 12 fi e lrr(H/N) such that (px = cp/i for x e G, with /z independent of the choice of cp. Use this and the hypothesis p ,£ 2 to show that cp is invariant in G. Note If p = 2, then Problem 12.15 is actually false. (12.16) Let G be a p-group and suppose b(G) = p2. If p =£ 2, show that n{kerz|z(i) = p2}=l. Hint Use Problem 12.15. Note Passman has conjectured that if G is any p-group with b(G) = pe and e < p, then P){ker x|/(l) = pe} = 1. He has proved this when G has class 2.
Character correspondence We have already seen several examples of the following situation (for instance, Theorems 6.11 and 6.16). We are given H, a subgroup or factor group of G. Subsets if £ Irr(H) and 3" c Irr(G) are specified and it is proved that there exists a "natural" one-to-one correspondence between if and 3". Here, the word "natural" is intended to mean that the correspondence is uniquely described by some general rule and thus more is being said than merely that |if \ = \3~\. We shall not attempt to give a precise definition of naturalness. Most of this chapter is devoted to the study of a particular character correspondence which was discovered by G. Glauberman. We introduce some notation. Let S and G be groups such that S acts on G. [That is, we are given a homomorphism S -» Aut(G).] In this situation, we can construct the semidirect product r of G by S so that G <i T, S £ r, GS = T, G n S = 1 and the given action of S on G is the action by conjugation in T. (In fact these properties characterize the semidirect product.) If x is a character of G and seS, then as usual we define the character Xs of G by xs(gs) = 1(g)- Then S permutes Irr(G). We write Irrs(G) = {x e Irr(G) \f = X for all s e S}. (13.1) theorem (Glauberman) For every pair of groups (G, S) such that S is solvable and acts on G and (|G|,|S|)= 1, there exists a uniquely defined one-to-one map n(G, S): Irrs(G) -»Irr(CG(S)). These maps satisfy the following properties: (a) If T <i S and B = CG(T), then n(G, T) maps Irrs(G) onto Irrs(B). (b) In the situation of (a), n(G, S) = n(G, T)n(B, S/T). 219
220 Chapter 13 (c) Suppose S is a p-group and C = CG(S). Let / e Irrs(G) and i/f = (x)7t(G, S). Then i/f is the unique irreducible constituent of Xc such that In the situation of (a) in the theorem, note that B is S-invariant since T <i S. Thus S acts on B and in fact S/T acts on B. Therefore n(B, S/T) is defined on Irrs/T(B) = Irrs(B) which is the image of Irrs(G) under n(G, T). Also, n(B, S/T) maps to the irreducible characters of CB(S/T) = CB(S) = CG(S). Thus the equation in (b) makes sense. Also note that if T = S in (a), then B = CG(S) and Irrs(B) = Irr(B). Thus this special case of (a) asserts that n(G, S) always maps Irrs(G) onto Irr(CG(S)). There is enough information in the statement of Theorem 13.1 to determine n uniquely. Thus suppose that n0(G, S) is defined whenever (G, S) satisfies the hypotheses of 13.1 and that n0 satisfies (a), (b), and (c). We claim that n0(G, S) = n(G, S). By 13.1(c), this is certainly the case if S is a p-group so we may work by induction on | S | and assume that S has composite order. Let T <i S have prime index and let B = CG(T). Then n(B, S/T) = n0(B, S/T) and n(G, T) = n0(G, T) and 13.1(b) yields n(G, S) = n0(G, S). The preceding argument suggests how to construct the map n(G, S); namely, prove that if S is a cyclic p-group and / e Irrs(G), then %c does have a unique irreducible constituent /? such that [/c, /?] $ 0 mod p, where C = CG(S). Define n(G, S) in this case by d)n(G, S) = (i. For general solvable S, define n(G, S) by working along a composition series for S, There are numerous technical difficulties with this approach, not the least of which is to show that the map constructed is independent of the composition series. The key to overcoming these difficulties is to find a uniform definition for n(G, S) for all cyclic S. Following Glauberman, this is what we shall do. We establish some notation which will be used repeatedly. (13.2) hypothesis Let S act on G and suppose (|G|, \S\) = 1. Let C = Cc(S) and let T be the semidirect product T = GS. (13.3) lemma Assume the situation in 13.2 and let / e Irrs(G). Then there exists a unique extension % of / to T such that (o{%), \S\) = 1. Also, % is the unique extension such that S £ ker(det #). Proof The first statement is just Corollary 8.16. Since o(#s) divides both \S\ and o(%), we have o(#s) = 1 and S c ker(det #). If i> is an extension of/ with S £ ker(det $), then 0(1^) = o(det \ji) and divides | Y: ker(det \ji) \ which is prime to | S \. Thus \p = % and the proof is complete. | We call the character % of Lemma 13.3 the canonical extension of /. For positive integers n, we write Q„ = 0(e), where e is a primitive nth root of unity. If (m, ri) = 1, it is well known from Galois theory that Q„ n Qm
Character correspondence 221 = Q. Thus if a e Q„ is invariant under the Galois group ^(Qmn/Qm), then aeO. If M<i H with \M\ = m and \H :M\ = n such that (m, n) = 1, we shall use the notation &(H/M) to denote 0(Onm/OJ. Note that &(H/M) permutes Irr(H) (by Problem 2.2 or Lemma 9.16). Suppose 9 e Irr(M) is extendible to H. Since ^(H/M) fixes 9, it permutes the set of extensions of 9 to H. In the situation of 13.2, if / e Irrs(G), and % is the canonical extension of X to T, then for t e^(T/G), (xY is an extension of / and o(xz) = o(x). Thus f = % and hence f has values in Q|G| and %s has values in Q|G| n Q|S| = Q. We have thus proved the following corollary. (13.4) corollary In the notation of Lemma 13.3, %s is rational valued. The following strengthens this. (13.5) lemma Assume Hypothesis 13.2. Let /eIrrs(G) and let % be the canonical extension of / to T. Note that CS = C x S. For each irreducible constituent /? of Xc there exists a (possibly reducible) character \jip of S such that #cs = Y,t (P x •M- This equation uniquely determines the \jip. Also, the \pp are rational valued. Proof We have Irr( CS) = {^x <p\pelrr(C), <peIrr(S)}. Write %cs = Yji.q, apJiP * <?)• Set \\it = £„ a$v9 and observe that \pp = 0 unless \jc, PI ¥=■ 0. Thus #cs = Yj (P x ^f\ where the sum runs over those /?eIrr(C), which are constituents of Xc- Also, this equation uniquely determines the \jips. If t e 0(r/G) then (%)z = X and pz = p. Thus xcs = ((mcs = KP*(hn Thus \jip is invariant under 0(r/G) and hence has values in Q(G|. Since its values also lie in Q|S|, the result follows. | (13.6) theorem Assume Hypothesis 13.2 and that S is cyclic. Then for each x £ Irrs(G), there exist unique /? e Irr(C) and e = +1 such that %(cs) = eP(c) for all c e C and all generators s of S, where % is the canonical extension of x to T. Also, p is a constituent of Xc and the map / h-» $ is one-to-one. Proof Write %cs = Y P x ^fi as m Lemma 13.5 and fix a generator s of S. Write 9(c) = #(cs) for c e C. Then 9 = Jj, »A^(s)jS. In particular, 9 is a class function on C. We claim that [9, 9] = 1. Let T be a set of representatives for the right cosets of C in G. If t1; t2 e T and x e (Cs)'1 n (Cs)'2, then there exist c1; c2e C with qV' = (c^)'1 = x = (c2s)H = c2t2st2.
222 Chapter 13 We thus have two factorizations of x into products of commuting elements of orders dividing \G\ and \S\. By Lemma 8.18, s'1 = s'2 and thus t^-1 eC(s) = C since S = <s>. Thus rt = t2. Hence the sets (Cs)' are disjoint for distinct teT and \\J,(Cs)'\ = \T\\C\ = \G\ = \Gs\. Since s's'^G, we have (Cs)' £ Gs and hence Gs = (J, (Cs)' is a disjoint union. We have |C|[M]= II2WI2= II2(*')I2 for all teT.lt follows that |T||C|[M]= II2WI2- *eGs By Lemma 8.14(c), the latter sum equals | G\ = | T11 C\ and thus [9, 9] = 1 as claimed. We now have l = [9,9] = £l'/'^)|2. However, i/^(s) is a rational algebraic integer by Lemma 13.5 and so lies in Z. Thus i/^(s) is nonzero for some unique /? and for that /?, i/^(s) = e = +1. This yields %(cs) = e/?(c) for ceC. This equation clearly determines e and /? uniquely. We now show that /? is independent of the choice of the generator s of S, If S = <s0>, then s0 = sm for some m with (m, \S\) = 1. Thus there exists an automorphism a of the field Q|S| such that X"(s) = X(sf = l(s0) for all 1 e Irr(S). Therefore 0 j= i/^s)" =» ^p(s0) and hence replacing s by s0 yields the same /?. Finally, suppose /l5 /2 e Irrs(G) determine the same character /? so that j?i(cs) = e;f}(c) for c e C and i = 1,2. Since Gs = (J(Cs)', it follows that ft^gs) = £1^2 Xi{9s) f°r all g e G and thus by Lemma 8.14(b), it follows that Xl = (Xi)g = (l2)G = #2 and the map /i-> /? is one-to-one. The proof is complete. | (13.7) definition Assume Hypothesis 13.2 and that S is cyclic. Construct maps y(G, S): Irrs(G) -+ Irr(C) and e(G, S): Irrs(G) -> {-1, 1} by (z)y(G, S) = fi and (x)e(G, S) = e = +1 where #(cs) = e0(c) for ceC, <s> = S, and % is the canonical extension of / to Y. It will turn out that the map n(G, S) of Theorem 13.1 equals y(G, S) when y(G, S) is defined, that is, for cyclic S. The map y(G, S): Irrs(G) -»Irr(C) of Definition 13.7 is one-to-one by 13.6. It is also onto. One way to prove this is to show that |Irrs(G)| = |Irr(C)|.
Character correspondence 223 Since S is cyclic, it follows from Brauer's Theorem 6.32 that | Irrs(G) | is equal to the number of S-invariant conjugacy classes of G. It is true that each S- invariant class of G intersects C nontrivially and, in fact, the interesction is a single conjugacy class of C. It follows that the number of S-invariant classes of G is equal to the total number of conjugacy classes of C, and hence equals |Irr(C)|. Thus y(G. S) maps onto. The assertion about S-invariant classes of G in the preceding paragraph follows from the Schur-Zassenhaus theorem. We digress from the discussion of characters in order to give a proof. (13.8) lemma (Glauberman) Let S act on G with (|S|, |G|) = 1. Assume that one of S or G is solvable. Let S and G both act on a set Q such that (a) (a • g) ■ s = (a • s) • gs for all a e Q, g e G, and seS. (b) G is transitive on Q. Then S fixes a point of Q. Proof Let T = GS, the semidirect product. For gs e T and a e Q, define a • (gs) = (a • g) ■ s. Condition (a) above guarantees that this is an action. Pick a efi and let H = Ta. Since |fi| = \G:G r\H\ = |r:H|by(b), it follows that \H: G n H\ = \S\. By the existence part of the Schur- Zassenhaus theorem, let T be a complement for G n H in H. Then \T\ = \S\ and T is a complement for G in T. Now the conjugacy part of the Schur-Zassenhaus theorem yields S = Tx for some x e T. Thus S £ H* and S fixes a • x e Q. The proof is complete. | (13.9) corollary In the situation of Lemma 13.8, the set of S-fixed points of Q is an orbit under the action of CG(S). Proof If a e Q is fixed by S and c e CG(S), then (a • c) • s = (a • s) • cs = a • c and a • c is S-fixed. Now suppose a, /? e Q are S-fixed. Let X = {g e G|a • g = /?}. Then X is a left coset of Gfi and is S-invariant. Let Gp act on X by right multiplication. Note that G^ is S-invariant and is transitive on X. For xeX,geGfi and s e S, we have (x ■ g) ■ s = (xgf = xsgs = (x • s) • gs and Lemma 13.8 applies to the actions of S on G0 and S and Gp on X. Thus S fixes a point xeX. Then x e CG(S) and a • x = /?. The proof is complete. | (13.10) corollary Assume Hypothesis 13.2 and that at least one of G or S is solvable. Then Jf i-^ Jf n C defines a bijection from the set of S-invariant conjugacy classes of G onto the set of conjugacy classes of C. Proof Let Jf be an S-invariant class of G. The conjugation action of G on Jf is transitive. For keJt,geG, and s e S, we have (k-g)-s = (g- lkgf = (g' lyk°gs = (k-s). g\
224 Chapter 13 By Lemma 13.8 we conclude that Jf n C ¥= 0 and by 13.9, Jf n C is a class of C. Since the classes of G are disjoint, the map Jf i-> Jf n C is one-to-one. If ceC, then the G-class of c is S-invariant. It follows that the map jf i-> Jf n C is onto. | (13.11) corollary The map y(G, S) of Definition 13.7 maps Irrs(G) onto Irr(C). Assume Hypothesis 13.2 and that S is solvable. Let if be a composition series for S with y:l = So<S!< xSk = S. Let Q = CG(S,) so that G= C0 2 Ct 2 ••■ 2 Ct= C. Also, Si+1 £ Nr(C,-) for 0 < i < k and we view Si+l/S; as acting on Cf. Since S;+ l/Si is cyclic, y» = y(C„s,+1/S|) is defined for each i, 0 ^ j < k. We have y,:IrrSl + 1(C,)-In(Q + i). Each yt is one-to-one and maps onto Irr(Ci+1) and so we can define yrl:hr(Ci+i)^lrr(Cd. (13.12) definition Assume Hypothesis 13.2 with S solvable. Let if be a composition series for S and use the above notation. Put 9C{G, if) = (\xx(Q)y;}vy-k}2 ■ ■ ■ 7l~ ly0~l £ Irr(G) and let n(G, if): SC(G, if) -»Irr(C) be given by n(G, if) = y0yi • • • Ik-1- Thus 3C(G, if) is the largest set on which the composite function yoTi • • • yt_! is defined. Since each yt is one-to-one, n(G, if) is one-to-one and by construction of 3C(G, if), we see that n(G, if) maps SC(G, if) onto Irr(C). Our objective now is to show that 3C(G, if) = Irrs(G), that n(G, if) is independent of the choice of if and that for cyclic S, n(G, if) = y(G, if). We obtain these results by considering the case that S is a p-group. (13.13) lemma Let Z s R s C, where R is a ring containing the values of all ^ e Irr(G). Let 9 be a generalized character of G with values in an ideal / of R. Assume / n Z £ pZ for some prime p not dividing \G\. Then p divides [9, <J] for all £ 6 Irr(G).
Character correspondence 225 Proof We have | G | [8, £| = £ %)£(#) e / n Z. Thus p 11G | [8, £] and the result follows since pjf \ G \. | (13.14) theorem Assume Hypothesis 13.2 and that S is a p-group. Let /eIrrs(G). Then there exists a unique /?eIrr(C), with \jc, P] ^ 0 mod p. Furthermore (a) [XoP] = ±lmodp; (b) if S is cyclic, then 0 = (x)y(G, S) and (x)e(G, S) = [&, j8] mod p. (c) If ^ is a composition series for S, then 3C{G, if) = Irr^G) and (x)7t(G, if) = /?. In particular, n(G, if) is independent of the choice of if. Proof First assume that S is cyclic and let /? = (x)y(G, S) and e = (x)e(G, S) so that %(cs) = e/?(c) for c e S and S = <s>, where # is the canonical extension of / to T. Let R be the ring of algebraic integers in Q|r| and let / be a maximal ideal of R with p e /. In the notation of Theorem 8.20, we have (cs)p, = c and thus that theorem yields /(c) = %{c) = %{cs) mod /. We therefore have /(c) = e/?(c) mod / for all c e C and thus Xc — e/? is a generalized character of C with values in /. Since 1 £ / and p e /, we have I n I = pZ and since p^|C|, Lemma 13.13 yields [/c - e/?, £] = 0 mod p for all £ e Irr(C). It follows that [_Xc, £1=0 mod p for £ ^ 0 and [/c, 0] = e mod p. When S is cyclic, this proves everything but (c). If \S\ = p, then ^: 1 <a S and ^(G, if) = Irrs(G) and tc(G, S) = y(G, S). In particular, part (c) of the theorem holds when \S\ = p. Now assume \S\ > p and drop the assumption that S is cyclic. Work by induction on \S\. Let if: 1 = S0 <i • • • <i Sk = S and write T = St_ t and 5": 1 = S0 <i • ■ • <i St_! = T. Let B = CG(T). Then tc(G, if) = n(G, S~)y(B, S/T). Also, 3C(G, if) is the inverse image in 3C{G, 3~) of Irrs/T(B) = Irrs(B) under the map n(G, 3T); %(G, 3T) -+ Irr(B). By the inductive hypothesis applied to T, we have / e Irrs(G) £ IrrT(G) = 3C(G, 9~) and Xb = P® ± £, where 9 is a character of B or is zero and Z = (x)n(G, 3-). UseS, we have [/B, £] = [_(f)B, £] = [/B, « and thus [.Xb, F\ ^ 0 mod p. Thus £ = <f and £ e Irrs(5). In particular, x e %(G, if). If 0 e Irr(C), we have \lc, PI = E(P8 ± Oc « = ± Kc, « mod p
226 Chapter 13 and since £, e Irrs/T(B) and S/T is cyclic, it follows from the first part of the proof that there is a unique /?eIrr(C) with \_£c, /?] ^ 0 mod p, namely 0 = (£)y(B, S/T). Thus fi = (z)tc(G, if) is unique in Irr(C) such that [Xc> 0] ^ 0 mod p and in fact [/c, 0] = ± [<^c, 0] = ± 1 mod p. We already have Irrs(G) £ 3C(G, if). Now suppose tA e %(G, if) £ #"(G, 3T) = IrrT(G). Let >/ = (\p)n(G, 3~) so that >/eIrrs(B). Let seS. Then 0 ^ |>B, >/] = [(^b, yf\ = [(i/Ob. >/] mod p. Since T <i S and \p e IrrT(G), it follows that \lis e IrrT(G) = 3C(G, 9~) and thus r\ = (^s)n(G, £T) since W)B, rf\ £ 0 mod p. Since 7t(G, 5") is one-to-one, we have \p = \j/s and thus \p e Irrs(G). We have now shown that 3C(G, if) = Irrs(G) and have given a description of the map n(G, if) which is independent of if. This completes the proof. | Assume Hypothesis 13.2 and that S is cyclic so that y(G, S): Irrs(G) -> Irr(C) is defined. Suppose errr^r-! is an isomorphism and that er(G) = Gi and ct(S) = Si. Then a(C) = C} = C^St) and y(Gu St): IrrSl(Gi) -»Irr(d) is defined. Because y(G, S) is uniquely defined, independently of any arbitrary choices, it is clear that if (x)y(G,S) = (S, . then (XiMGi,Sl) = p1, where Xi and /?! correspond to / and /? via the isomorphism, a. That is, Xi(9a) = X(g) and jS^c") = P(c) for g e G and c e C. (Recall that the computation of (x)y(G, S) requires choosing a generator s of S, but that the result is independent of this choice.) An important special case of this invariance under isomorphism of y(G, S) is when a e Aut(r) and G" = G and S" = S. In that case we have (Xa)y(G, S) = ((xMG, S))°. (13.15) lemma Assume Hypothesis 13.2 and let To S with T cyclic and B = CG(T). Then y(G, T) maps Irrs(G) onto Irrs(B). Proof Since Irrs(G) £ IrrT(G), y(G, T) is defined on Irrs(G). Let H = GT o r. If s e S, then s defines an automorphism of H with Gs = G and Ts = T. Therefore, by the above discussion, we have (f)y(G, T) = ((z)y(G, T)f for all x e Irr7(G). It is immediate that y(G, T) maps Irrs(G) into Irrs(B). Since y(G, T) maps onto Irr(B) and is one-to-one, the result follows. |
Character correspondence 227 (13.16) theorem Assume Hypothesis 13.2 and that S is cyclic. Let p be a prime and let T be the p-complement in S. Let B = CG(T). Then y(G, S) = y(G,TMB,S/T) on Irrs(G). Proof If pJf\S\, then T = S, B = C, and y(C, 1) is the identity map on Irr(C) (as is clear from Definition 13.7). The result is thus trivial in this case and we assume p 11S |. In particular, pjf \ G \. Let x e Irrs(G) £ IrrT(G) and let £ = (z)y(G, T) e Irr(B). Let 0 =(z)y(G, S). By Lemma 13.15, £ e Irrs(B) = Irrs/T(B). We must show that fi = (£)y(B, S/T). Let p0 = (£)y(B,S/T). Since S/T is a p-group, Theorem 13.14 yields £c = P<P + £o^o> where <p is a character (or is zero) and e0 = +1. Let R be the ring of algebraic integers in Q|r| and let / be a maximal ideal containing p. We have then £(c) = e0fi0(c) mod / for c e C. By definition of y(G, T), we have %(bt) = e£(b), where T = <t>, 6eB, e = +1 and 2 is the canonical extension of % to r. Applying this to c e C £ B yields #(ct) = e£(c) = ee0)So(c) mod 1 for any generator, t of T. Now let S = <s> and let t = (s)p, in the notation of Theorem 8.20. Then <t> = T and (cs)p, = ct for c e C. Thus by 8.20, we have %(cs) = %(ci) mod / and thus t(cs) = eeopo(c)modl. By definition of y(G, S), we have %(cs) = 5f](c) for c e C with d = ± 1. Thus )S(c) = 8ee0(i0(c) mod I for all ceC. By Lemma 13.13, [)S - (5ee0/?<,,/?] is divisible by p and thus /? = )S0 as desired. | (13.17) corollary Assume Hypothesis 13.2 and that S is cyclic with \S\ = pq, where p and <j are primes. Let Sf be a composition series for S. Then 3C(G, if) = Irrs(G) and n(G, if) = y(G, S). Proof If p' = q, the result is immediate from Theorem 13.14(b) and (c). Assume then that p ^ q. Write if: 1 <i T <i S. We may assume | T | = q. Now Lemma 13.15 yields that SC(G, if) = Irrs(G). Theorem 13.16 asserts that n(G, if) = y(G, T)y(B, S/T) = y(G, S) where B = CG(T). The proof is complete. | (13.18) theorem Assume Hypothesis 13.2 with S solvable. Let if and 9~ be composition series for S. Then SC(G, if) = 3C(G, 3~) and n(G, if) = n(G, $~).
228 Chapter 13 Proof Use induction on the composition length k of S, If' k = 1 then y = 3~ and there is nothing to prove so assume k > 1. Write y:i = s0<--<st = s, Let £f* and ^"* be the composition series for Sk _ t and Tt _ 1 respectively, obtained by deleting S from y and ^". Consider the case that St_ t = Tt_ t. Let B = CG(St_ t). Then Irrs(B) = (In(QMB.S/Slk_1r1andso %(G,<?) = (lrrs(B))n(G,y*rl and SC(G,3-) = (lxxs(B))n(G,3-*rx. By the inductive hypothesis, n(G, £f*) = n(G, 5"*) and hence 3C(G, Sf) = SC(G, F). Also n(G, ST) = n(G, <?*)y(B, S/Sk-1) = n(G, <T*)y(B, S/Tk_ t) = n(G, T). Assume now that Sk-t =£ Tt_t and let M = St_! n Tk_v Let ^ be a composition series for M and extend ^ to composition series y° and 3~° for S which run through St_ t and Tt_ 1; respectively. By the preceding paragraph, 3C{G, y°) = 3C{G, <f) and n(G, Sf°) = n(G, <f\ We may thus replace Zf by y° and similarly replace 9~ by 9~°. We may now assume that S; = Tt for i < k — 2 and that Jt: 1 = S0<i aSt_2 = M. Let F: 1 -a St_ t/M -a S/A#, F: 1 -a Tt_ t/M <a S/A# and let D = CG(M). It follows that %(G, <f) = (%(D, &)MG, JtY' and 3C(G, $~) = (3C(D, F))n(G, Jt)' \ Also n(G, <f) = n(G, Jt)n(D, 3s) and n(G, $~) = n(G, Jt)n(D, W\ Therefore, it suffices to prove that 3C{D, &) = 3C{D, F) and n(D, F) = n(D, F), We may therefore assume that M = 1 and£> = G.ThusS = Sk-t x Tk_t is abelian of order pq for primes p and q. If p = q, then 3C(G, <f) = Irrs(G) = 3C(G, F) and n(G, <f) = n(G, 9~)
Character correspondence 229 by Theorem 13.14. If p ^ q, then G is cyclic and Corollary 13.17 yields the result. The proof is complete. | Assume Hypothesis 13.2 with solvable S and choose a composition series y for S. By Theorem 13.18, 3C(G, Zf) is independent of the particular composition series and so we can write 3C(G, S) = 9C{G, Sf). Similarly, we can write n(G, S) = n(G, y\ Then 3C(G, S) and n(G, S) are unambiguously defined. We have SC(G, S) £ Irr(G) and n(G, S) is a bijection of 3C(G, S) onto Irr(C). As in the discussion preceding Lemma 13.15, suppose a: r -» r\ is an isomorphism and let Gt = a(G), St = a(S\ and C\ = a{C) = Cq^S^). Then a induces bijections Irr(G) -» \rr(G{) and Irr(C) -»Ir^CJ. Since #"(G, S) is uniquely defined, we have 3C(GU St) is the image of #"(G, S) in Ir^GJ and 0 = (z)tc(G, S) implies that ft = (xMGu Si), where Zl(g') = fa) and ft^") = 0(c). In particular, if a e Aut(r) and G" = G and S" = S, then er leaves SC(G, S) setwise invariant and (x")n(G, S) = ((z)tc(G, S))" for / e SC(G, S). (13.19) corollary Assume Hypothesis 13.2, with S solvable, and let SC(G, S) and n(G, S) be as above. Then SC(G, S) = Irrs(G). Also, if T <i S and B = CG(T), then (a) n(G, T) maps Irrs(G) onto Irrs(B); (b) n(G, S) = n(G, T)n(B, S/T). Proof If T <i S and B = CG(T), use a composition series for S, which runs through T in order to construct n(G, S). Then (b) is immediate and %(G, S) = (3C(B, S/TMG, T)~l. Thus (a) will follow once we prove the first statement. We show that 3C(G, S) = Irrs(G) by induction on the composition length k of S. If k = 1 then SC(G, S) = (Irr(C))y(G, S)~' = Irr^G). Suppose then, that k > 1 and let T <i S with 1 < T < S. Let H = GT <i T. For s e S, we have Gs = G and Ts = T and hence by the discussion preceding the statement of the corollary, we see that (f)n(G, T) = ((x)n(G, T)f for x g SC(G, T) = IrrT(G). Since n(G, T) is a bijection from IrrT(G) onto Irr(B), where B = CG(T), we see that n(G, T) carries the S-invariant characters in IrrT(G) onto Irrs(B). Since Irrs(G) £ IrrT(G), it follows that Irrs(G) = (Irrs(B))7t(G, 7T1 = 3C(B, S/T)n(G, T)~l = 3C{G, S\ where the second equality is by the inductive hypothesis applied to S/T. | (13.20) definition Assume Hypothesis 13.2 with S solvable. Then the Glauberman map is the map n(G, S): Irr^G) -»Irr(C) constructed above.
230 Chapter 13 We have now completed the proof of Theorem 13.1. The Glauberman map satisfies conditions (a) and (b) of 13.1 by Corollary 13.19. It satisfies condition (c) by Theorem 13.14(c). Also, by Theorem 13.14(a), we have the following. (13.21) corollary Let n(G, S) be the Glauberman map with S a p-group. Let C = CG(S) and let xeIrrs(G) and 0 = (x)n(G, S). Then [zc, 0] = ±1 mod p. There is one further loose end. (13.22) corollary Assume Hypothesis 13.2 with S cyclic. Then n(G, S) = y(G, S). Proof Use induction on | S \. Let p 11S \. If S is a p-group, the result follows from Theorem 13.14(b). Assume that S is not a p-group and let T be the p- complement in S and B = CG(T). Then n(G, T) = y(G, T) and n(B, S/T) = y(B, S/T) by the inductive hypothesis. The result now follows from Corollary 13.19(b) (or Theorem 13.1(b)) and Theorem 13.16. | Let S act on G. Then S permutes Irr(G) and S permutes the set C1(G) of conjugacy classes of G. By Brauer's Theorem 6.32, the permutation characters of S on Irr(G) and C1(G) are equal and it is natural to ask if these actions are permutation isomorphic. That is, does there exist a bijection a: Irr(G) -> C1(G) such that a(/s) = a(/)s for all / e Irr(G) and s e S? In general, the answer is no. However, if (| G |, | S |) = 1 and S is solvable, it follows via Glauberman's Theorem 13.1 that the actions of S on Irr(G) and C1(G) are permutation isomorphic. (13.23) lemma Let the group S permute two sets Q and A. Suppose that for every T £ S, the number of fixed points of T on Q equals the number on A. Then Q and A are permutation isomorphic. Proof We prove the existence of a bijection a: Q -> A such that a(eo • s) = a(eo)• s for all coeQ and seS by induction on |Q|. (Note that taking T = 1 yields |fi| = |A|.) Let T s S be maximal such that T has a fixed point on Q. (Possibly T = S.) Let T fix co e Q and 1 e A. By the maximality of T, we have S^ = T = Sx. Let 6a be the orbit of co and Ox the orbit of 1 under S. Write Q = ff.uQ, and A = &x u At where the unions are disjoint. Map a0: 6a -> 6X by a0(co • s) = Is and check that a0 is well-defined, one-to-one, onto and that a0(v• s) = a0(v)• s for all ve(9m and seS, Since every H s S has equal numbers of fixed points on Om and 6X, it follows that H has equal numbers of fixed points on fit and At. By the
Character correspondence 231 inductive hypothesis, there exists a permutation isomorphism at: fit -» At. Now define a on Q by combining a0 and al. | (13.24) theorem LetSactonGwithSsolvableand(|G|,|S|)= 1. Then (a) S fixes the same numbers of irreducible characters and conjugacy classes of G. (b) The actions of S on Irr(G) and C1(G) are permutation isomorphic. Proof It suffices to prove (a) since (b) follows via Lemma 13.23 by application of (a) to all subgroups of S. Since n(G, S) maps Irrs(G) one-to-one and onto Irr(C), it follows that |Irrs(G)| = |Irr(C)| = |C1(C)|. By Corollary 13.10, intersection defines a bijection from the set of S-fixed conjugacy classes of G onto C1(C). The result now follows. | Theorem 13.24 becomes false if the hypothesis that (|G|, \S\) = 1 is dropped. (See Problem 13.16 for an example.) Suppose we continue to assume that (|G|,|S|)= 1, but drop the assumption that S is solvable. If S is nonsolvable, then 2||S| by the Feit- Thompson theorem and thus 2xK IGI - Thus G is solvable, again by Feit- Thompson. In this situation, where Hypothesis 13.2 is satisfied with G solvable of odd order, it is possible to construct a natural character correspondence from Irrs(G) onto Irr(C) by a method entirely different from Glau- berman's. (And thus Theorem 13.24 remains valid without the hypothesis that S is solvable.) We shall describe the map Irrs(G) -» Irr(C) in this case but without giving the proof since the only known proofs are too long to include. The key step is the following. (13.25) theorem Assume Hypothesis 13.2 and that G is solvable of odd order. Let C £ H £ G. Suppose that there exist S-invariant normal subgroups, K and L of G such that (a) L s K and K/L is abelian; (b) G = KC; (c) H = LC. Then for each x £ Irrs(G), there exists a unique \p e lrrs(H) such that [jH, i/f] is odd. The map / h-» \ji is a bijection from Irrs(G) onto Irrs(H). Proof Omitted. |
232 Chapter 13 To construct the correspondence between Irrs(G) and Irr(C) we construct a chain of subgroups G = C0 > Ct > ■ ■ ■ > Ck = C and apply Theorem 13.25 to obtain maps Irrs(Cj)-> lrr^Cl+l). The composition of these maps is the desired correspondence. Assume Hypothesis 13.2 and that G is solvable of odd order. We consider the set Jf of S-invariant subgroups H with C s H £ G. For H ejf, define H* = [H, SJC. Since [H, S] -a HS it follows that [H, S]' -a HS and thus H* e JT. If H > C, then [H, S] > 1 and [H, S] > [H, S]' by solvability. Since [H*, S] £ [H, S]', it follows that H* < H. We now define the subgroups C,- e jf by C0 = G and C»+1 = (C,)* and we have the desired chain of subgroups. We must now check that the hypotheses of Theorem 13.25 are satisfied by taking H = G*. Let K = [G, S] and L = [G, S]'. Then K and L are normal S-invariant subgroups of G and K/L is abelian. That G = KC follows fairly easily from Glauberman's Lemma 13.8. Condition (c) of Theorem 13.25 is automatic from the definition of G*. It has been conjectured for every group G and prime p that if JV = NG(.P) for P e Sylp(G), then the numbers of irreducible characters of p'-degree of G and of JV are equal. (For simple groups G, this conjecture is due to McKay.) Using Glauberman's Theorem 13.1, we prove a result which includes the special case of this conjecture when G has a normal p-complement. (13.26) theorem Let G = KH with X<i G, H solvable, and (|H|, |X|) = l.LetJV = NG(H)andput <r={zeIrr(G)|(|tf|,/(l))=l} and ^=faeIrr(iV)|(|H|,irtl))=l}. Then there exists a uniquely defined bijection of SC onto <&. Proof We have N = (N n K)H and the commutator [JVnX, H]sHnX = l so that JV n K = C = CK(H) and JV = C x H. It follows that <W={px l|/JeIrr(C),leIrr(tf), 1(1)= 1}. Now if x g 3C, let 9 be an irreducible constituent of %K. Then | G: 7G(^)I divides both *(1) and |G:X| = |H|. Thus JG(9) = G and 9eIrrH(X). Let
Character correspondence 233 9 be the canonical extension of 9 to G. Then x = H for some unique £eIrr(G/K) by Gallagher's theorem (Corollary 6.17). Since (x(l), |H|) = 1, we must have £(1) = 1. Also, / uniquely determines 9 by Xk = & and thus determines 9 and £. We can now map 3C to <& by /1-» {(9)n(K, H)) x £H. Since restriction defines a bijection of Irr(G/X) onto Irr(H) and since &; e S£ for every 9 e IrrH(X) and linear £ e Irr(G/K), it follows that we have mapped SC onto $/. The map is one-to-one since the r\ e <W are uniquely of the form Suppose S acts on G and that JV <i G is S-invariant. Also, assume that (|G:JV|,|S|) = 1. We consider a pair of " dual" questions: (a) Let x e Irrs(G). Does there exist 9 e Irrs(iV) with [/N, 9] ^ 0? (b) Let 9 e Irr^N). Does there exist / e Irrs(G) with [9G, /] # 0? Both questions can be answered in the affirmative. By the Feit-Thompson theorem, at least one of S and G/N is solvable and we assume this. Also, note that if S is a p-group, both facts can be proved relatively easily by counting arguments. (13.27) theorem Let S act on G and leave JV <i G invariant. Assume that (| S |, | G: JV |) = 1 and that one of S or G/N is solvable. Let / e Irrs(G). Then XN has an S-invariant irreducible constituent. Proof Let fi = {9eIrr(N)|[^, 9] ^ 0}. Then G/N permutes fi transivively and since / is S-invariant, S permutes Q. Also, the action of S on G induces an action on G/N. Let 9 e Q, s e S, and g e G. For x e JV, we have (9T(xs)=9"(x) = 9(9*0-1) and (9TV) = SVxV)"1) = %^-1)- The hypotheses of Glauberman's Lemma 13.8 are thus satisfied and the result follows. | The situation of question (b) is more difficult and interesting. First we consider the case where G/N is solvable. (13.28) theorem Let S act on G and leave JV <i G invariant. Assume that (|S|, |G: JV|) = 1 and that G/N is solvable. Let 9eIrrs(JV). Then 9G has an S-invariant irreducible constituent. Proof First assume that G/N is abelian, and let A be the group of linear characters of G/N. Let fi = {/ e Irr(G) | [9G, /] ¥= 0}. If / e fi and k e A, then
234 Chapter 13 #1 e fi since (%1)N = Xn. This defines an action of A on Q. We claim that this action is transitive. Let /, \ji efi. Then 0# \_xN, \pN~] = [(xN)G, t/f] and \ji is an irreducible constituent of (xNf = fel^)G = x(\Nf. However, (lNf = ^a^ and thus (Xn)g = Z Z^- Since #1 e Irr(G) for each XeA,it follows that \ji = %X for some k and A is transitive on Q as claimed. Since 9 is S-invariant, it follows that S permutes Q and also S acts on the group A since (l/z)s = ls/zs. If / e Q, 1 e A, and s e S, we clearly have (ztf = W and the hypotheses of Glauberman's Lemma 13.8 are satisfied since \A\ = \G:N\ and thus (| X |, | S |) = 1. The result follows in this case. We now assume that G/N is nonabelian and work by induction on | G: N\. Let M/N = (G/N)'. Then M <i G is S-invariant and M < G since G/JV is solvable. By the inductive hypothesis, 9M has an S-invariant irreducible constituent \ji and by the first part of the proof, \jiG has an S-invariant irreducible constituent. The result now follows since 9G = (9M)G. | To handle the case that G/N is not solvable, we appeal to the Glauberman correspondence, Theorem 13.1. We first restrict attention to the situation where(|G|,|S|)= 1. (13.29) theorem Assume Hypothesis 13.2 and that S is solvable. Let JVo T with N £ G. Let / e Irrs(G) and 9 e Irrs(JV). Write £ = (x)n(G, S) and q> = (S)n(N, S). Then [9G, Z] * 0 iff [>c, £] * 0. Proof We first consider the case that S is a p-group. Then [/c, £] ^ 0 mod p and every irreducible constituent of Xc other than £, occurs with multiplicity divisible by p. Since XcnN = (Xc)cnN> it follows that (1) [XCnN'fpl = DfoflKcnN.^modp. Now write Xn = X ^a A, where A runs over sums of orbits of the action of S on Irr(iV). If A = £ 0 where (9 is such an orbit, then [ACnJV, cp] = ICIfocnw.p] forijeffi If |0|> 1, then p||0| and [ACnJV, cp] = 0 mod p. Thus [Zcniv.<p]= Z [*iv>'7][>7cniv><p]mocip. i,eIrrs(JV) However, for >/ e lrrs(N), we have \_rjc n N, <p] = 0 mod p unless cp = (rj)n(N, S), that is, unless r\ = 9. Thus (2) [XcnN, <l>] = [.Xn, 9][9CnN, <p] mod p. Since [jc, Q p 0 p [9CnJV, <p], comparison of Equations (1) and (2) yields that \_Xn> 9] = 0 mod p iff \_iCr,N> <P\ = 0 mod p. Since JV <i G and
Character correspondence 23S pjf | G | it follows that [Zw, 9] = 0 mod p iff [zK, 9] = 0. Similarly, [£c n N, <p] = 0 mod p iff [iJCfllv, <p] = 0. The result thus follows if S is a p-group. To complete the proof, we use induction on | S \. We may choose T <i S with S/T a nontrivial p-group. Let B = CG(T). By Theorem 13.1, we have n(G, S) = n(G, T)n(B, S/T) and similarly, n(N, S) = n(N, T)n(B n N, S/T). Now [9G, jf] * 0 iff [((9)7t(iV, T))B, (x)n(G, T)] ^ 0 by the inductive hypothesis. The result now follows by the first part of the proof. | (13.30) corollary Let S act on G with S solvable and (| S \, \ G \) = 1. Let N <i G be S-invariant. Let 9 e lrrs(N). Then 9G has an S-invariant irreducible constituent. Proof Let C = CG(S) and cp = ($)n(N, S) so that cp e Irr(C n N). Let £ be an irreducible constituent of <pc and let / = (<p)n(G, S)~1. Then / e Irrs(G) and [9G, /] j= 0 by Theorem 13.29. The proof is complete. | To complete our analysis of question (b), we need to consider the case where S is solvable and (| S \, | G |) ¥= 1. Of course, we continue to assume that (\S\, \G:N\) = 1. First, we observe that it is no loss to assume that 9 is invariant in G since if U = IG(9), then S leaves U invariant. Now, if we can find an S-invariant irreducible constituent \\i of 9V, then t/^GeIrrs(G) as desired. We shall finish the proof by an appeal to the theory of projective representations of Chapter 11. The following proves slightly more than we need. (13.31) theorem LetJVs G«a r with JV«a Tand (|r:G|, |G:JV|) = 1. Assume that one of Y/G or G/N is solvable. Let 9 e Irr(iV) be invariant in T. Then 9G has some T-invariant irreducible constituent. Proof By Theorem 11.28, we can find a character triple (Tl,Nl,9l) and an isomorphism (t, a): (T, N, 9) -> (Tu Nu 9t) with 9t(l) = 1. Let Gi = G1 so that the isomorphism v. F/N -> r^/iV! carries G/N to G^iV*!. Suppose we can find i/^ elntG!^) with i/^ invariant in rt. Let t^eIrr(G|9) with <tg(1/') = "Ai- We claim that \ji is invariant in r. To see this, let / be an irreducible constituent of t/^r so that /elrr(r|9). Let Xi = °r(x)- Then (/i)Gl = eipi for some integer e and it follows that /G = ei/f. Thus ip is invariant as desired. The argument of the preceding paragraph shows that it is no loss to assume that 9 is linear. We may thus factor 9 = X/j. where (o(X), \G:N\)= 1 and (o(/z), | T: G |) = 1 and X and y. are powers of 9. Thus X and y. are invariant in T. By Corollary 6.27, X has a unique extension, X e Irr(G) such that o(X) = o(X). Because of the uniqueness, it follows that X is invariant in T. Suppose we can find a T-invariant irreducible constituent \ji of fiG. Then \jil e Irr(G) is T-invariant and 9 = fiX is a constituent of (^1)^,.
236 Chapter 13 We may therefore assume that 9 = fi, that is, that 9 is linear and (o(9), |T: G|) = 1. Also, we may replace T by T/ker 9 and thus assume that 9 is faithful. Thus |iV| = o(9) is relatively prime to |T: G| and hence (|G|, |T: G|) = 1. By the Schur-Zassenhaus theorem, we can find S £ r with SG = T and S n G = 1. Thus S acts on G and leaves N and 9 invariant. Also, one of S or G/iV is solvable. Now Corollary 13.30 or Theorem 13.28 yields an S-invariant irreducible constituent of 9G. The result follows. | To close the chapter, we obtain some further information in a special case of Theorem 13.6. (13.32) theorem Assume Hypothesis 13.2 and that S is cyclic. Also suppose that C = CG(x) for all x e S - {1}. Let / e Irrs(G) and 0 = (x)n(G, S). Let I be the canonical extension of / to r. Then there exists e = +1 and fi e Irr(S) with fi2 = ls such that (a) %{cx) = £P(c)(i(x) for all x e S - {1}; (b) Zc = IS19 + e/?, where 9 is a character of C or is zero; (c) Crfl)-e0(l))/|S| = *eZ. (d) Xs = kps + e.fi(V)fi, where ps is the regular character. (e) fi ¥= ls iff IS| is even and k is odd. 'Proof If 1 < T s S then n(G, S) = n(G, T) = y(G, T) and hence %(cx) = + Ac) for all x e S - {1} where the sign is independent of c. It follows that in the notation of Lemma 13.5, we have t/^(x) = ± 1 for all x j= 1 and iji^x) = 0 for x ^ 1 and 0 ^ 9 e irr(C). Write i/f = i//0. We work to express \ji in terms of Irr(S). Let lu X2 e Irr(S) and compute [(*i--U£l = (i/ls|) Z ±(i1W-i2w). JceS; jc?s 1 Since Ut(x) — A2(x)| < 2 for xeS, this yields IQ^i - ^2X«A:i < 2(|S| - 1)/|S| < 2. Therefore, the multiplicities with which lt and X2 occur in \ji differ by at most 1. It follows that \p = aps ± fi where a e Z and fi is a sum of distinct linear characters of S with /z(l) < |S|/2. Also, /z(x) = ±tA(x) = ±1 for 1 ^ x eS and hence /z(l) = [>,/z] = (l/|S|)Ml)2 + |S|-l) and (i(l)2 - |S|/*(l) + \S\ - 1 = 0. It follows that /z(l) = 1 and /zelrr(S). Define e = +1 by the equation \p = aps + e/z. (Note that if \S\ = 2, then
Problems 237 neither e nor fi is uniquely defined.) Now xicx) = 0(<#(x) = e/^M*) f°r l#x£S and (a) is proved. Also fi(x) = ± 1 for x e S so that fi2 = ls. Since all i//v for <p ^ /? vanish on S — {1}, each is a multiple ofps and hence Up x </v)c is a multiple of | S\ <p. Since (/? x i/^)c = a| S| /? + e/?, statement (b) follows and (c) is immediate from (b). Also (d) now follows. We now prove (e). Since \i2 = ls, it follows that fi = ls if | S | is odd. Suppose then, that \S\ is even and let xeIrr(S) with x2 = ls but x ^ ls. (This uniquely defines x.) Now det(ps) = n^irns) ^ = t. By (d) it follows that ls = det(zs) = tV (1). Since 2| \S\, we have 2J((J(1) and therefore fi = xk. Statement (e) now follows and the proof is complete. | (13.33) corollary Let (r, N, 9) be a character triple and let N s G <i T. Suppose T/G is cyclic and that r/N is a Frobenius group with kernel G/N. Assume that 9G = ex for zelrr(G) (and thus e2 = \G:N\). Then |T:G| divides e — e for some e = ± 1. Proof We may replace (r, N, 9) by an isomorphic character triple and assume 9(1) = 1. Write 9 = pk, where (o(l), \G:N\) = 1 and (o(/z), |T: G|) = 1 and 1 and fi are powers of 9 and thus invariant in T and 1 is extendible to G by Corollary 6.27. Let v be an extension of I. Then ^ = (l9)« = (Vjv9)G = v9G = e(vz). Since v/eIrr(G) we may replace 9 by fi and assume (o(9), |T: G|) = 1. We may also assume that ker 9 = 1 so that N c Z(H and (|G|, |r:G|) = 1. Let S be a complement for G in T. If 1 ^ x e S, then CG/A,(x) = 1 and it follows that N = CG(x) and we are in the situation of Theorem 13.32. (Note that x £ Irrs(G) since / is the unique irreducible constituent of 9G.) The result now follows from 13.32(c). | Problems (13.1) Assume Hypothesis 13.2 with S solvable. Prove the following facts by using the statement of Theorem 13.1, but do not appeal to any of the results used in constructing the Glauberman map. (a) If x e Irrs(G) then (x)7t(G, S) is a constituent of Xc- (b) If p = (x)n(G, S), then Q(X) = QW). (c) If C = G, then n(G, S): Irr(G) -> Irr(G) is the identity map. (13.2) In the situation of Theorem 13.1, let 0 = (x)n(G, S) for / e Irrs(G). Show that x( 1) divides |G: C|j8(l). Hint In the case that S is cyclic, consider a = eoj as in Chapter 3.
238 Chapter 13 (13.3) In the situation of Theorem 13.1, let JVo G be S-invariant with NC = G. Let 9 e Irrs(JV). Show that IM&MN, S)) = U») n C. (13.4) Let S act on G and leave JV <i G invariant. Assume that (| S |, | G: JV |) = 1 and that one of S or G/N is solvable. Suppose CGIN(S) = 1. If / e Irrs(G), show that there exists a unique i/f e Irrs(JV) such that [x^, i/f] ^ 0. Write (X)5 = \ji. Show that 5 maps Irrs(G) onto Irrs(JV). (13.5) Assume Hypothesis 13.2 and that S is solvable. Let C £ JV £ G with jv<ir. (a) Show that CG/JV(S) = 1. (b) Let 5: Irrs(G) -»Irrs(iV) be as in Problem 13.4. Show that 5 is one- to-one. (c) Show that dn(N, S) = n(G, S). Hint (For (a)) Use Glauberman's Lemma 13.8. (13.6) In the situation of Theorem 13.1, assume that G is solvable. Let X e Irrs(G) and /} = (i)k(G, S). Show that 0(1) divides *(1). Hints Let JV < G be normal, S-invariant and maximal such. Then either NC = G or N 2 C. Use Problem 13.3 or 13.5(c) and induction on | G |. (13.7) In Problem 13.4, assume that G/N is solvable. Show that 8 is one-to- one. Hint Use induction on | G: JV |. (13.8) Assume Hypothesis 13.2 and that G is nilpotent. Show that |Irrs(G)| = |Irr(C)|. Do not assume Theorem 13.25. Hint Use Problem 13.7. (13.9) Assume Hypothesis 13.2 and that G is solvable. Show that C > 1 iff | Irrs(G)| > 1. Do not assume Theorem 13.25. (13.10) Let JVo T with N ^G-oT and (|T: G|, |G: JV|) = 1. Assume that one of Y/G or G/N is solvable. Let K/N be a complement for G/N in T/JV and assume CGjN(K/N) = 1. Let 9 e Irr(JV) be invariant in K. Show that there exists a unique T-invariant / e Irr(G) with [9G, /] ^ 0. Hints Use the argument of Theorem 13.31 to reduce to the case (| G |, | T: G |) = 1. Use Problem 13.5(b) if T/G is solvable. (13.11) Let S be solvable and let H be a group with (\S\, \H\) = 1. Let G = HxHx-xH where there are \S\ factors and let S act on G by
Problems 239 permuting the factors regularly. Let C = CG(S) and note that C = {(h,h,...,h)\heH} = H. Also, if / e Irrs(G), then / = 9x9x---x9 for some 9 e Irr(H). If (x)n(G, S) = p, show that p((h, h,...,h)) = 9(/j|s|). (13.12) Let JV <i G and suppose 9 e Irr(JV) is invariant in G. Suppose 9G = ex for some / e Irr(G). (Thus (G, JV, 9) is a fully ramified character triple and e2 = | G: JV|.) Assume that S is solvable, acts on G, leaves JV and 9 invariant and that (|G: JV|, |S|) = 1. Let C/N = CG/N(S). Show that Xc = ft with £ e Irr(C) and/2 = |G:C|. Hwf Use the technique of the proof of Theorem 13.31 to reduce to the case that (| G |, | S\) = 1 and 9(1) = 1 with 9 faithful. In this case, C = CG(S). Use Theorem 13.1. (13.13) Let S act on G and leave JV -a G invariant. Assume that (| S \, \ G: N \) = 1 and that S acts trivially on G/N. Let 9eIrr(JV) and /elrr(G) with [Xiv> #] t6 0- Show that 9 is S-invariant iff / is S-invariant. Do not assume the Feit-Thompson theorem. Hint In showing that 9 e Irrs(JV) implies / e Irrs(G), it suffices to assume that G/N is cyclic. (13.14) In the situation of Theorem 13.1, let JV<i G be S-invariant with NC = G. Let 9 e Irrs(JV) and <p = (9)tc(JV, S). Let / = 7G(9) (so that I n C = Ic(<P) by Problem 13.3). If \ji is an irreducible constituent of 9', show that (ii>G)n(G, S) = (W)n(l, S)f. (13.15) Let E be elementary abelian of order p2 for p j= 2 and let S be dihedral of order 2p. (a) Define an action of S on E such that Irrs(£) = {1E} but C^S) > 1. (b) Define an action of S on E such that CE(S) = 1 but |Irrs(£)| > 1.
Linear groups Suppose x is a faithful character of G. In this chapter we are concerned with drawing conclusions about G when given information about /. For instance, we already know that if / is irreducible, then Z(G) is cyclic and that if all irreducible constituents of / are linear, then G is abelian. Another, less trivial example which we have seen is Theorem 3.13. A faithful F-representation of G with degree n is an isomorphism of G with a linear group of degree n over F; in other words, a subgroup of GL(n, F). For our purposes, we will restrict attention to finite linear groups. (It should be pointed out, however, that stating that an infinite group is a linear group imposes a type of finiteness condition on it, that is, it guarantees that the group is not "too badly" infinite.) We shall also restrict attention to complex linear groups. Thus from now on, a "linear group" is a finite subgroup of GL(n, C) for some n. A group is thus isomorphic to a linear group of degree n iff it has a faithful character of degree n. We say that a linear group is irreducible if the identity map is an irreducible representation. (14.1) theorem (BUchfeldt) Let G be a linear group of degree n and let 7t = {p|p is prime, p > n + 1}. Then G has an abelian Hall rc-subgroup. We need a lemma. Recall that a character / is p-rational (where p is prime) if its values lie in Qr for some r with pjfr. (This is Definition 6.29.) (14.2) lemma Let p 76 q be primes such that G has no element of order pq. Then each / e Irr(G) is either p-rational or ^-rational. 240
Linear groups 241 Proof Let /elrr(G) and suppose that z is neither p-rational nor q- rational. Then in the notation of the discussion following Definition 6.29, there exists a e 0P(G) and t e &q(G) with f # X ^ t- Also, by Problem 2.2(b), Z",ZreIrr(G). If g eG and pJfo(g), then %(g)e Qm, where \G\ - pam and pjfm. Thus by definition of 0p(G), we have /(g)" = x{g). Similarly, if qJfo(g), then z(#) = x(g)- Since G has no element of order pq, it follows for every g e G that either P^o(9)or 4^o(g) and we conclude that l(x-xa)Ax-xl']=o. Since [/, z1] = 0 = [z°, /], we obtain 0 = [z, z] + Df, ZT = 1 + [Z*. Z1] > 1 and this contradiction proves the result. I The following easy lemma is a special case of a more general result due to Schur which we will prove later. (14.3) lemma Let x be a faithful p-rational character of G and assume that p divides \G\. Then z(l) > p - 1. Proof Since z is p-rational, it has values in Qr for some r with pjfr. Let P^ G with |P| = p. Then zPhas values in Qr n Qp = Q. Let <S = <S(QP/Q). Then ^ fixes zP and thus permutes the linear constituents of zP. Since z is faithful, zP has a nonprincipal linear constituent X and l(x) t6 1, where 1 t6 x e P. Since ^ is transitive on the p — 1 primitive pth roots of 1, it follows that the images of X under ^ take on p — 1 different values at x and thus there are at least p — 1 different characters in the orbit of X under <§. Since each is a constituent of xP, the result follows. | An observation that is often useful when working with linear groups is that if {Xj} is a family of normal subgroups of G with f]Kt = l,thenGcanbe isomorphically embedded in the direct product U(G/Kd via g^(...,gKt,...). Proof of Theorem 14.1 Use induction on n and for groups with degree n, induct on | G |. Let z be a faithful character of G with z( 1) = n. Suppose that z is reducible and write z = Zi + Xi- Let Kt = ker Xt so that G/Kt is isomorphic to a linear group of degree z,(l) < n. By the inductive hypothesis, G/Kt has an abelian Hall 7t,-subgroup where nt = {p\p > 1 + z.-(l)}- Since n c m, it follows that G/X,- has an abelian Hall 7t-subgroup H/K;. If H,- < G, then the inductive hypothesis yields an abelian Hall 7t-subgroup H of Ht. Since no prime in n divides | (G/X,): (H,/X,) | = | G: H; |, it follows that H is a
242 Chapter 14 Hall 7t-subgroup of G and we are done if Ht < G. We may thus assume that H, = G and G/Kt is an abelian rc-group for i = 1, 2. However, Kt n K2 = ker xi ^ ker /2 = ^er / = 1 and thus G is isomorphic to a subgroup of the abelian rc-group, (G/K]) x (G/K2)- The result follows in this case. We now assume that / is irreducible. If H £ G is a rc-subgroup, let 9 be an irreducible constituent of %H. Then 8(1) divides |H| and so a prime divisor p of 9(1) satisfies n < p <. 9(1) < n and this contradiction shows that 9(1) = 1. Thus Xh is a sum of linear characters and since / is faithful, it follows that H is abelian. Suppose there exists M <i G with \G:M\ = perc. Let H be a Hall n- subgroup of M, which exists by the inductive hypothesis. If H <i G, let P e Sylp(G). Then //P is a Hall rc-subgroup of G and we are done. Suppose then, H~fi G and choose a Sylow subgroup Q of H with Q~& G. Then Q is Sylow in M and thus G = MNG(Q) by the Frattini argument. Now H is abelian and so H £ NG(Q). We have |G :NG(Q)| = |M :NM(Q)| which divides \M\H\. Thus |G: NG(Q)| involves no primes from n. Since NG(Q) < G, it has a Hall 7t-subgroup which is one for G. We may thus assume that no such M exists. Next,supposeZ £ Z(G)with \Z\ = pen.ThenXz = X( 1Mwitho(^) = P- Thus (det(/))z = 1*(1) # lz since /(l) < p. Therefore, p|o(x) and it follows that p 11G : ker(det /) |. This yields a normal subgroup of index p, a contradiction. (Note that we have reproved part of Theorem 5.6 here.) Thus no such Z exists. Now let H £ G be a 7t-subgroup of maximum possible order. Suppose H is not a Hall 7t-subgroup. Then there exists qen with q \ \ G : H \. In particular, H ¥= 1 or else H < QeSyl,(G) which violates the maximality of H. Let x e H have prime order pen and let C = CG(x). Then C < G by the previous paragraph and thus C has a Hall 7t-subgroup K by the inductive hypothesis. Since H is abelian, we have H £ C and thus | H | < | K \. The maximality of |H | yields \H\ = \ K \ and H is a Hall 7t-subgroup of C. In particular, qJ(\C:H\ and it follows that q \ \ G : C \. LetxePe Sylp(G). Then P is abelian and so P £ C(x) = C and pj( \ G: C \. Thus p t6 q. Since p can be any prime divisor of \H \, we have q^|H \. Therefore, q)(\C\ since H is a Hall 7t-subgroup of C. We conclude that x centralizes no element of order q in G. We claim that G contains no element of order pq. Otherwise, there exist commuting y, zeG with o(y) = p and o(z) = q. However, H contains a full Sylow p-subgroup of G since C does and H is a Hall 7t-subgroup of C. Thus H contains a conjugate of y which we may suppose to be x. Since x centralizes no element of order q, this is a contradiction and proves the claim. Lemma 14.2 now yields that/ is r-rational for r = por<j.Thusx(l) > r — 1 by Lemma 14.3. This is a contradiction since ren and the proof is complete. |
Linear groups 243 (14.4) corollary Suppose x(l) is prime for every / e Irr(G) with x(l) > 1. Then G is solvable. Proof The hypothesis is inherited by factor groups and by normal subgroups and so we may assume that G is simple. Let /elrr(G) with minimal /(l) > 1. By Problem 3.3, /(l) = p > 2. Let n = {q\q is prime, q > p + 1}. Since G is simple, / is faithful and Theorem 14.1 yields an abelian Hall rc-subgroup A £ G. Let 1 ¥= a e A. Then A £ CG(a) and the only prime divisors of | G: CG(a) | are < p + 1 and hence <p. If \ji e Irr(G) and t/^(l) £ {1, p}, then (.A(l),|G:CG(a)|)= 1 and by Burnside's Theorem 3.8, we have \p(a) = 0 or a e Z{\p). Since G is simple, Z(t/f) = 1 and hence i/f(a) = 0. Now 0 = £ flDfti) = 1 + p £ £(a) «€lrr(G) <SeIrr(G);4(l) = p and hence 1/p is an algebraic integer. This contradiction completes the proof. | In the situation of Corollary 14.4, Problem 12.3 applies since G is solvable. It follows that |c.d.(G)| < 3 and thus G" = 1 by Theorem 12.15. Our next results concern finding normal subgroups of a linear group of degree n whose order is divisible by a prime which is large when compared with n. (14.5) theorem (D. L. Winter) Let G be a solvable irreducible linear group of degree n. Suppose that a Sylow p-subgroup of G is not normal. Then n is divisible by a prime power q > 1 such that q = — 1, 0, or 1 mod p. Proof Let / e Irr(G) be faithful with /(l) = n. Suppose G is a counterexample to the theorem with minimum possible order. We argue first that every proper normal subgroup of G has index divisible by p and has a normal Sylow p-subgroup. Let M <i G be proper and let 9t,..., 9r be the distinct irreducible constituents of Xm ■ Letm be the common degree of the fy. Then m \ n and hence no prime power q > 1 with q = -1, 0, or 1 can divide m. Since |M\ < \G\, it follows that M/ker fy has a normal Sylow p-subgroup. Now (~)ker 9, £ ker / = 1 and thus M is isomorphic to a subgroup of (M/ker St) x ■ • ■ x (M/ker 9r).
244 Chapter 14 It follows that M has a normal Sylow p-subgroup which is necessarily normal in G. Since G does not have a normal Sylow p-subgroup, we conclude that p divides | G: M | as claimed. Now let K be a maximal normal subgroup of G and let P e Sylp(K). Then P <iG and \G:K\ = p. Since G is not a p-group, we have P < K and we choose L^P such that K/L is a chief factor of G. Write | K : L \ = a and let SeSylp(G). Now LS < G and pJc \ G: LS \. Thus LS ^G G. Since K(LS) = G, it follows that CK/L(S) < K/L. However, CK/L(S) -a G/L and we conclude that CK/L(S) = 1. Since S is a p-group, it follows that a = 1 mod p. We claim that Xl reduces. Otherwise, Xls is irreducible and the minimality of G yields S <i LS. Since p/fa(l) it follows that Xs is a sum of linear constituents and hence S is abelian. Thus S £ CG(P) -a G and hence pjf\ G : CG(P)|. By the first part of the proof, we conclude that CG(P) = G and P £ Z(G). Therefore, L = P x N where JV = Op.(L) <i G. Since S <i LS and S n JV = 1, it follows that S £ CG(JV) <a G and thus pjf \ G : CG(N) |. Therefore, CG(JV) = G and JV c Z(G). Thus L = NP ^ Z(G) and since /L e Irr(L), we have n = 1 and G is abelian. This is a contradiction and proves that Xl reduces, as claimed. Since pJfx(l\ we have Xk e Irr(K). Since /L ^ Irr(L) there are two possibilities by Theorem 6..18. Either Xl is a sum of |K : L| = a distinct irreducible constituents or else Xl = ei>, with "A e Irr(L), and e2 = a. In the first case, a > 1 is a prime power dividing n and a = 1 mod p, a contradiction. In the second case, e > 1 is a prime power dividing n and e2 = a = 1 mod p. Thus e = ± 1 mod p and the proof is complete. | (14.6) corollary (ho) Let G be a solvable linear group of degree n and let p > n + 1 be a prime. Suppose that a Sylow p-subgroup of G is not normal. Then G is irreducible, p = n + 1, and n is a power of 2. Proof If G is irreducible, then by Winter's Theorem 14.5, there exists a prime power q > 1 that divides n such that q = — 1, 0, or 1 mod p. Thus p— l>n><j>p— 1 and hence n = p — 1 is a prime power. We conclude that n is a power of 2 and the proof is complete in this case. If G is reducible, write / = Xi + li - where / is a faithful character of G of degree n. Let Kt = ker Xi so that K± c\K2 = \ and G is isomorphic to a subgroup of (G/Ki) x (G/K2). Each G/Kt is isomorphic to a solvable linear group of degree x,(l) < n. Working by induction on n, it follows that each G/Kt has a normal Sylow p-subgroup and hence so does their direct product. Since G does not have a normal Sylow subgroup we have a contradiction and the proof is complete. | The hypothesis that G is solvable in Corollary 14.6 can easily be relaxed to the assumption that G is "p-solvable." We say that G is p-solvable if there
Linear groups 245 exist subgroups Nt <i G with 1 = N0 c JVt c . •. c Nk = G and such that each factor Ni+ l/Ni is either a p-group or of order prime to p. (14.7) corollary Let G be a p-solvable linear group of degree n < p — 1 and suppose that a Sylow p-subgroup of G is not normal. Then n = p — 1 is a power of 2 and G is irreducible. Proof Assume either that G is reducible, n < p — 1 or that n is not a power of 2. Use induction on \G\. Let M be the next to last term in a p- solvable series for G so that M < G, G/M is either a p-group or a p'-group and M is p-solvable. By the inductive hypothesis, M has a normal Sylow p- subgroup P and thus we may assume that G/M is a p-group. Let Q/P e Sylq(M/P) for some prime q ^ p. The Frattini argument yields MNG(Q) = G and thus |G:NG(Q)| = |M:NM(Q)|. Since P £ Q <= NM(Q) and P e Sylp(M), we conclude that p/f | G : NG(Q) |. Let S e Sylp(NG(Q)). Thus SeSylp(G). Now SQ is solvable and hence S <i SQ by Corollary 14.6. Since | G : Q | is not divisible by <j, we conclude that qJf\G: NG(S)|. Since |G:NG(S)| is independent of the choice of S e Sylp(G) and q =£ p is arbitrary, the result follows. | It is conjectured that Winter's Theorem 14.5 holds for all p-solvable groups. The conjecture is known to hold for n < 2p + 1. Unlike the ease with which Corollary 14.7 follows from Corollary 14.6, the facts for n > p seem quite deep. Even the special case where G has a normal p-complement is open if n is significantly larger than 2p. The results which have been obtained all seem to depend heavily on the Glauberman correspondence, and in particular on Theorem 13.14. The proofs are too complicated to give here. Now we drop all hypotheses of solvability or p-solvability. Suppose G is a linear group of degree n and a Sylow p-subgroup of G is not normal. How big can p be? The first bound was established by Blichfeldt and the best possible bound, p <, 2n + 1, was proved by Feit and Thompson. The Feit-Thompson proof depends on a very deep result of Brauer which gives the bound under the additional assumption that p^ld. Here we shall prove p < (n + l)2 and also show how the Feit-Thompson reduction to the case p2Jf\ G\ works. We need some preliminary results. (14.8) lemma Let H £ G be abelian and let y be the set of nonprincipal irreducible constituents of (1H)G. Compute a = min{/(l)/[x, (lH)G]lze^}- Then t/^(l) > a — 1 for every nonlinear ijj elrr(G).
246 Chapter 14 Proof Let \ji e Irr(G) with \ji(X) > 1. Write i/^ = \G + £ axx, where the sum runs over nonprincipal x £ Irr(G). Since H is abelian, we have •A(l) < [>«, -Ah] = E-Ah-Ah, lfl] = [-/"A, (1h)°] = 1+1 afo, (l„)Gj\ However, [/, (1H)G] <, Z(l)/a for / e S? and this yields Ml) < 1 + (l/a)£az/(l) < 1 + (l/a)(^(l)2 - 1). Thus o#(l) - 1) < i/<l)2 - 1 and since i/f(l) > 1 we obtain a < i/f(l) + 1 as desired. | (14.9) lemma Let N <i G and suppose g e G with JV n C(gr) = 1. Let / be a character of G such that \jN, ljv] = 0. Then %(g) = 0. Proof Let g = gN e G/JV. Then CG/N(g) 3 CG(g)N/N and hence |CG/JV(0)| > |CG(9)JV/JV| = \CG(g)\. Thus I l«%)|2> I K(9)l2- i^eIrr(G/N) ^elrr(G) Viewing Irr(G/JV) £ Irr(G), we conclude that £(g) = 0 for £ e Irr(G) with JV ^ ker <!;. The result follows. | Note that the above proof is essentially a repetition of the proof of Corollary 2.24. Also, if g and JV are as in the Lemma 14.9, then all elements of the coset Ng are conjugate in G and thus the result follows from Problem 2.1(b). (14.10) theorem (Feit-Thompson) Let H £ G be abelian with H n Z(G) = 1. Assume for every geG — Z(G) that CG(g) has nontrivial intersection with at most one conjugate of H in G. Let / e Irr(G) with H £ ker /. Then (a) Z(1)2>|H|[ZH, 1H]2. Also, if z(l) > 1 and H is contained in no proper normal subgroup of G, then (b) (l + z(l))2>|H|. Proof Let JV = NG(H) and X = {xeG- Z(G)|C(x)nH > 1}. Note that H - {1} £ X and that JV £ N(X). We claim that X is a T.I. set and that JV = N(X). In particular, X £ JV. Let g e G be such that X r\ X9 ^ 0. Choose x e X with x9 e X. Then C(x) n H > 1 and so C(x9) n HB > 1. Also, C(x9) nH>l and hence H = Hg by hypothesis. Thus g e JV and X = Xg and the claim is established. We have \G\ = \G\[x, Z] = I l*(0)l2 > Z(l)2 + IG: JV| £ |Z(x)|2 geG JceX
Linear groups 247 and thus (i) IlxWI2<|N|. xeX Now write Xn = a + P, where [aH, 1H] = 0 and fiH = 0(1)1H. (This is possible since H <i JV.) Since H £ ker /, we have a ^ 0 but we allow the possibility that /? = 0. Write Z = Z(G). Then £Z€Z |/(z)|2 = |Z|/(1)2 and since Z r\X =0, we conclude from (1) that (2) |Z|/(1)2 + |JV|> £ lz(x)|2= £ |a(x)+ /?(*) |2. By Lemma 14.9, we have a(y) = 0 for y e JV — (X u Z) and thus (3) X |a(x) + 0(x)|2 = |JV|[a,a]+2|JV|[a,/?]+ £ |0(x)|2. Since [a, /3] = 0 and [a, a] > 1, (2) and (3) yield \Z\X(l)2 + \N\ > \N\ + £ |/?(x)|2. xeX u Z Because HZ ^ X u Z, we obtain (4) |Z|Z(1)2 > £ |/?(x)|2. *eHZ Write Xz = Z(1M- Then /?z = /?(1)1 and since /?H = /?(1)1H, we have /?(*«) = 0(l)A(z) for he H and z e Z. Thus | 0(x) |2 = 0(1)2 for x e HZ and (4) yields /?(l)2|tf||Z|<|Z|x(l)2. Thusx(l)2 > 0(l)2\H\ and since 0(1) = [/H, lH], the proof of (a) is complete. Now assume that no proper normal subgroup of G contains H. Let Sf be the set of nonprincipal irreducible constituents of (1H)G. If ip eSf, then H £ ker ip and (a) yields i/<l)2 > \Jia, \„~\2\H\ and Itfl1'2 < a = min{t/<l)/[tMlH)G]|</>e.n Lemma 14.8 then gives for /(l) > 1 that Z(l) + 1 >a> l#|1/2 and (b) follows. | (14.11) theorem (Feit-Thompson-Blichfeldt). Let G be a linear group of degree n and let p be a prime. Assume one of the following. (a) Every subgroup of G of order not divisible by p2 has a normal Sylow p-subgroup and p > n + 1. (b) p>(n + l)2. Then G has a normal Sylow subgroup.
24S Chapter 14 Proof Use induction on \G\. We may thus suppose that every proper subgroup of G has a normal Sylow p-subgroup. Let P e Sylp(G) and assume P -56 G. Since P has a faithful character of degree n < p and this character can have no nonlinear irreducible constituents, we conclude that P is abelian. If P n Z(G) ^ 1, then by Theorem 5.6, we cannot have P n Z(G) £ G' and thus p||G: G'|. It follows that there exists N <i G with |G:JV| = p. Since JV has a normal Sylow p-subgroup, we conclude that G is p-solvable. Then G violates Corollary 14.7 since n < p — 1 and P ~fi G. We conclude that P n Z(G) = 1. Now if 1 # y e P n P9, then P, P9 e Sylp(C(y)) and this forces P = PB since C(y) < G. Thus P is a T.I. set. Now suppose that xeG — Z(G). Then C(x) has a unique Sylow p-subgroup S and if S > 1, then S is contained in a unique conjugate of P. It follows that C(x) has nontrivial intersection with at most one conjugate of P. IfP £ M -a GwithM < G,thenP <M and thus P -a G, a contradiction. We have now shown that P satisfies the hypotheses of Theorem 14.10(b) and thus \P\ < (i/^l) + l)2 for every nonlinear ifr e Irr(G). Since G is nonabelian, the given faithful character must have some nonlinear irreducible constituent and we conclude that \P\ < (n + l)2. Thus p <(n + l)2, contradicting hypothesis (b). We are therefore in the situation of hypothesis (a) and in particular, p2||G|. Thus p2 < \P\ < (n + l)2 and p < n + 1 which is our final contradiction. | As was mentioned before, Brauer proved that if G is a linear group of degree n and p2Jf\G\ where p > 2n + 1, then a Sylow p-subgroup of G is normal. It then follows from Theorem 14.11(a) that the hypothesis p2X IG | is unnecessary. There is a great deal of information known about linear groups of degree n in which a Sylow p-subgroup is not normal and n < p <2n + 1. This all requires deep results from Brauer's "modular character" theory and we will not discuss it further here. There are many ways in which the structure of a group is limited in terms of its degree as a linear group. For example, there exist integer valued functions/, such that for linear groups G of degree n we have: (a) If G is solvable, then d.l.(G) < /t(n). (b) If G is p-solvable and pe divides | G: Op(G)|, then e < f2(n). (c) If G is a p-group and | G: 3>(G) | = pe, then e <; /3(n). A much more general result of this type is Jordan's theorem.
Linear groups 249 (14.12) theorem (C.Jordan) Let G be a linear group of degree n. Then there exists abelian A <i G such that |GM| ^(n!)12""t"'+1)+1) where 7t(/c) denotes the number of primes <, k. In fact, the existence of the functions/,,/2, and /3 mentioned above follows easily from Jordan's theorem, although the best known bounds for /i,/2, and/3 are very much better than those which can be derived from the inequality in Theorem 14.12 (or from any other known bound for Jordan's theorem). There is no reason to suppose that the index of an abelian normal subgroup of maximum possible order in a linear group of degree n can be anywhere nearly as large as (n!)12n('t(''+1)+1). (For instance, if n = 2, the "correct" bound is 60 rather than 2 ■ 126.) The good bounds for the functions j\ are proved by methods independent of Jordan's theorem. We begin work now on a proof of Theorem 14.12 which is due to Fro- benius. As the reader will see, this proof has an unusual geometric flavor. Recall that a square complex matrix U is said to be "unitary" if UJ = U"1. Note that the unitary n x n matrices form a subgroup of GL(n, C). Also, if U is unitary, then there exists unitary V such that V~lUV is diagonal. (More generally, this holds for all normal matrices U, that is, those which satisfy UU* = U*U, where U* = UJ.) Also, the diagonal matrix VlUV is unitary since both U and V are. Write V1UV = diag(lt,..., kn) so that the Xj are the eigenvalues of U. Since V~lUV is unitary, it follows that kfl = X] and this proves that the eigenvalues of a unitary matrix lie on the unit circle. (14.13) lemma Let A and B be n x n complex unitary matrices. Assume that the eigenvalues of B lie in the interior of some arc of length n on the unit circle. Suppose that A commutes with A~ iB~ iAB. Then A commutes with B. Proof We may conjugate A and B by a unitary matrix so as to diagon- alize B and hence we may assume that B = diag(ft1,..., b„). Since a permutation matrix is unitary, we may rearrange the bj in any desired order by conjugating both A and B by an appropriate permuation matrix. We may thus write b} = emj where &t < 92 < ■ ■ ■ < 9„ < 9t + n. Write A = (a^). We shall show that a^ = Qiib^^ bv. This immediately yields that AB = BA. Now write C = A~lB~lAB. We have A"1 = A1 and B~1 = B so that C = A1 BAB. Also, since AC = CAv/e have A1 BAB = C = AC A-1 = B-lABA~v = BAB A1 Now evaluate the diagonal entry, cmm of C. We have li v
250 Chapter 14 and thus We compare imaginary parts and obtain £ |a„J2 sin(9m - 9,) = I |amv|2 sin(9v - 9J. /i V Thus for all m, 1 < m < n, we have (*) £(|amv|2 + |avJ2)sin(9v-9J = 0. V We now use (*) to prove that if 9; j= 9k, then aJk = 0. Suppose this is false and choosey minimal such that there exists k with 9,- < 9k and either aJk ^ 0 or akj ¥= 0. Take m = j in (*). If v < j, then either 9V = 9; in which case sin(9v — 9;) = 0 or 9V < 9j in which case avj = 0 = aJV by the minimality of j. Thus (*) yields Z(|flJ-vl2 + |flvjl2)sin(9v-9j) = 0. Now for v > j we have 9^ < 9V < 9; + n and thus sin(9v — 9^) > 0. Also, Ifl/vl2 + l«vjl2 ^ 0- We conclude that for each value v > j, either Ifljvl2 + |flVJ-l2 = 0 or sin(9v — 9j) = 0. Now 0 < 9k — 9j- < n and hence sin(9k — 9;) > 0. Therefore |a;k|2 + |akJ|2 = 0 and thus ajk = 0 = akj. This is a contradiction and completes the proof. | (14.14) lemma Let A and B ben x n complex unitary matrices and suppose that the eigenvalues of A lie in an arc of length a < n on the unit circle. Then the eigenvalues of A" iB~ iAB lie on the unit circle between — a and a. Proof Certainly A" iB~ iAB is unitary and so its eigenvalues lie on the unit circle. Let a be an arc of length a which contains the eigenvalues of A and write C = B~ lAB. We need to find the eigenvalues of A" lC. Let 1 be an eigenvalue of A" iC and let the column vector x be a corresponding eigenvector so that A" 1Cx = he and thus Cx = A Ax and 3cTCx = M^x1 Ax). Thus it suffices to show that arg(3cTy4x)ea and arg(3cTOc) ea, where arg(a) = a/|a| for 0 j= aeC. Since both A and C are unitary with eigenvalues in a it suffices to show that arg(yJUy)ea whenever y#0 and U is unitary with eigenvalues in a. Let U and y be such and let V be unitary with V~lUV = D, a diagonal matrix. Let z = V~ly so that yJUy = zJVJUVz = zJDz
Linear groups 251 sinceFT= V '.NowifD = diag(rflf..., d„)andz = co\(zu ..., z„),wehave zJDz = 1\zt\2dt. Since the dt lie in a and \zi\1 > 0, it follows that £ |Zj|2rfj lies in the infinite wedge which is the union of all rays from the origin through points of a. Since a < n and some |z,|2 is nonzero, it follows that £ |z(|2dj ^ 0 and arg(£ I Zj |2rfj) e a as desired. | (14.15) theorem (Frobenius) Let G be a linear group, let A, B e G and suppose that the eigenvalues of A and £ lie in arcs of length a and /? respectively, on the unit circle. Then (a) If /? < n and A commutes with the commutator [A, B], then LA, B] = 1. (b) If a < 7t/3 and ft < n, then [A, B] = 1. Proof By Theorem 4.17 we can conjugate all of the elements of G by a matrix such that the resulting matrices are all unitary. We may thus suppose that A and B are unitary matrices. Now (a) is just a restatement of Lemma 14.13. To prove (b), conjugate by a unitary matrix which diagonalizes A. We may thus assume that A is diagonal. Construct elements Bt e G by setting B0 = B and B, = \_A, B^J for i > 1. By Lemma 14.14, the eigenvalues of Bt lie between — n/3 and 7t/3 on the unit circle for i > 1 and thus for each i > 0, they lie in the interior of some arc of length n. Thus by part (a) we see that if Bi+1 = 1 for any i > 0, then Bj = 1. It thus suffices to prove that Bt = 1 for some i. If M is any complex matrix, we define 9(M) = tr(MTM) so that if M = (m^), we have 9(M) = £,„ |m,J2. Thus 9(M) > 0 and 9(M) = 0 iff M is a zero matrix. If U is unitary, we have 9(UM) = tr(MTUTUM) = tr(MTM) = 9(M). Now we compute (*) 9(Bi+1 - 1) = 9(B,A(Bl+1 - 1)) = BiB^A-'Br'AB, - 1)) = 9(/4Bi - BtA) = 9iA(Bt - 1) - (B, - 1)A). Now write B, — 1 = (ft^v). Since A is diagonal, we can write A = diag(alf a2,..., a„), where all of the a} lie inside some arc of length n/3. The (fi, v) entry of A(B{ — 1) - (B; - 1)A is a^fc^v - t^a, = ^(a,* - «v)- Note that \all — av| < 1. Equation (*) now yields that 9(Bi+1 - 1) = 11VI2K - «vl2 ^ 11VI2 = S(B, - 1), t*t V P> V where in fact the inequality is strict unless all & = 0.
252 Chapter 14 If no Bt = 1, we therefore have 9(B0 - 1) > 9{Bi - 1) > 9(B2 - 1) > ■ ■ ■. Since G is a finite group, there are only finitely many Bt and this is a contradiction. We conclude that Bt = 1 for some i and the proof is complete. | Observe that we have used our standing assumption that the groups we consider are finite twice in the above proof. The first time was in the appeal to Theorem 4.17 which is false for infinite groups. This use of finiteness could have been avoided by assuming that G was contained in the group of unitary matrices. The second application of finiteness in the last paragraph of the proof is more fundamental. For instance, part (b) of the theorem is not valid for the full unitary group. (14.16) theorem Let G be a linear group of degree n. Then there exists abelian A <i G such that \H:H nA\ < 12" for every abelian subgroup H ^ G. Proof Since G is a finite group, the eigenvalues of every element of G lie on the unit circle. If g e G, let a.(g) be the length of the shortest closed arc which contains the eigenvalues of g. Note that a is a class function on G. Put A = <a e G| a(a) < 7t/3>. Then /1<G and A is abelian since the generators of A commute by Theorem 14.15(b). Now let H c G with H abelian. By Maschke's Theorem 1.9 we may conjugate all of the matrices in G so as to assume that all h e H are diagonal matrices. Partition the unit circle into twelve half-open arcs, abQ2 o12, each of length n/6. Each element heH determines a function fh from {i\ 1 < i < n} into {j\ 1 < j < 12} by writing h = diag(alf..., a„) and setting fh(i) = j if a.ea,-. Note that there are at most 12" different functions fh that can be obtained this way. Suppose x,yeH with fx = fy- Write x = diag(alf..., a„) and y = diag(/?lf..., /?„). Then for each i, a; and /?,- lie in the same a,- and hence a,/?, lies strictly between — n/6 and n/6 on the unit circle. Thus xy~l eA and it follows that elements in distinct cosets of A n H in H determine different functions. Thus \H :A n H\ < 12" and the proof is complete. | (14.17) lemma Let P be a linear p-group of degree n. Then there exists abelian A <i P such that | P : A \ divides n! Proof By Corollary 6.14, P is an M-group and hence if \ji e Irr(P), then there exists H c p with \P: H\ = t/^(l) and such that t/^H has a linear constituent. Choose such a subgroup H^ for each \p e Irr(P).
Linear groups 253 Let x be a faithful character of P of degree n and let Q be the set of all right cosets of all H^ for irreducible constituents \p of /. Then | Q | < n and G acts on Q by right multiplication. Let A be the kernel of this action. Thus | P: A | divides n! To show that A is abelian, it suffices to show that Xa is a sum of linear characters. However, for each irreducible constituent \p of / we have A £ H^, and thus i//A has a linear constituent. Since A <i G, it follows that all irreducible constituents of i/^ are linear and the result follows. | Proof of Theorem 14.12 Let A -a G be as in Theorem 14.16 so that A is abelian and \H :H n A\ < 12" for every abelian subgroup H £ G. We claim that |G:X| < (n!)12n('t(''+1)+1). Factor |G:X| into prime powers. For each prime p let rp = \G:A\P. By Blichfeldt's Theorem 14.1, there exists an abelian Hall subgroup H0 of G for the set of primes >n + 1. We have \G : A\ = \G :H0A\\H0A inland it follows that I! rp = | H0 A :A\ = \H0:H0nA\< 12". p>n+l For p < n + 1, let P eSylp(G) and let Hp = P be abelian with |P: Hp| dividing n! (using Lemma 14.17). Then \P : Hp\ < (n!)p and rp = IPAMI = \PA:HPA\\HPA:A\ = |P:PnHpA||Hp :Hpn A| <\P:Hp\\2n<{n%W. Thus EI rP ^ (n (" Op )12""("+ ° = (n !)12"'t"'+ !>. p<n+l \p / The result now follows. | We can also use Frobenius' Theorem 14.15 in a different way. (14.18) theorem Let p > 7 and let G be an irreducible linear group of degree n <2p with n^p. If the Sylow p-subgroups of G are nonabelian, then p2Jf\ G : Op(G)|. Thus in any case, G/Op(G) has abelian Sylow p-subgroups. Proof Let PeSylp(G) be nonabelian so that P'nZ(P)^l. Let U S F nZ(P)with|l/| = p. Let x £ Irr(G) be the character of the given faithful representation so that /(l) = n. Since P is not abelian, zPhas some nonlinear irreducible constituent \ji. We must have t/^(l) = p. Since /(l) < 2p, we conclude that xP = i> + A, where A is a sum of linear characters. Since U £ P', we have U £ ker A and since U £ Z(P), we also have \jiv = pfi for some linear fi e lrr(U). Thus Xv = W + (n — P)lu- Since / is faithful, we have n ¥= lv
254 Chapter 14 Now pick ueU, with fi(u) = e2lti/p, and note that 2n/p < n/3 since p > 7. Therefore, u has only two distinct eigenvalues, namely 1 and e2*'/p, and these lie in an arc of length < 7t/3 on the unit circle. By Frobenius' Theorem 14.15(b) it follows that u commutes with all of its conjugates in G and thus A = (ua\g e G> <i G is abelian. Write Xa = e Yj= i ^i>wnere trie i; are distinct linear characters of A and et = n. Since 1/ = <u> £ A, we conclude that e divides \%v, /z] = p. Since also e | n, we have e = 1 and t = n. Now let K be the kernel of the permutation action of G on {!,}. Then IG: K | divides n! and hence p2Jf\ G: K |. If 9 is any irreducible constituent of xK, then 9^ is the sum of an orbit of the action of K on {1J. By definition of K, it follows that 9(1) = 1. We conclude that K is abelian and so the Sylow p-subgroup of K is contained in Op(G). The result follows. | We now present the promised generalization of Lemma 14.3. (14.19) theorem (Schur) Let G have a faithful p-rational character of degree n and let P e Sylp(G) with \P\ = f. Then a < t ln/{p - l)p'] < np/{p - l)2. i = 0 The brackets above denote the greatest integer function and thus the "infinite" sum has only finitely many nonzero terms. If n < p — 1 then all terms in the sum are zero and the result that a = 0 is Lemma 143. Note that G is really irrelevant in Theorem 14.19; the result is really about P. Also, a p-rational character of a p-group is necessarily rational valued since the values lie in Qpa r\ Qr where pjfr. Suppose x £ Irr(P) is faithful and we wish to bound \P\. Obviously, knowledge of /(l) is not sufficient since even if z(l) = 1, P could be an unboundedly large cyclic group. However, | P \ is bounded in terms of /(1) and the degree of the field extension Q(/) 2 Q. If z eZ(P) with o(z) = p, then x(z) = /(l)e, where e is a primitive pth root of 1. Thus e e <Q>(/) and hence p — 1 divides | Q(x): QI ■ Since |Qp«:Q| = (p- l)pfl_1, it follows that if P # 1, then | Q(x): Q | is of the form (p — l)p' for some integer t > 0. (14.20) theorem Let IPI = p", where p is a prime and a > 0. Let / e Irr(P) be faithful with *(1) = / and | Q(x) :Q\ = {p- l)p'. Then a < (f - l)/(p - 1) + (t + l)f. Proof We use induction on/. Iff = 1, then P is cyclic and since / is faithful, it follows that Q(/) = Qp« and thus t = a — 1. In this case, the inequality is actually equality.
Linear groups 255 Now suppose / > 1. Since P is an M-group, there exists a maximal subgroup H and 9 e Irr(H) with 9P = x- Thus \P : H\ = p and xH = £f= i 9;, where the 9, are the distinct conjugates of 9 under P. Note that the fields 0(9,) are all equal and that since 8P = /, we have Q(x) £ 0(9). Suppose Q(x) = 0(9). Let K,- = ker 9;. Thus f^i = 1 and H is iso- morphically contained in (H/KJ x ■ ■■ x (H/Kp). The |H:K,| are all equal, say to pb. Thus a < pft + 1. Since 9(1) = f/p, the inductive hypothesis yields b < ((//p) - l)/(p - 1) + (t + I)f/p and hence a < pb + 1 <; (/ - p)/(p - 1) + (t + l)f + 1 = (/- l)/(p - 1) + (t + l)f as required. The remaining case is where Q(/) < 0(9). Consider the Galois group 0 = 0(O(9)/Q(x)) so that |9| = |0(9): 0(/)|. In particular, <S ± 1 is a p- group. Since ^ fixes /, it permutes the irreducible constituents 9; of /H. Since the stabilizer of 9 in ^ is trivial, the orbit of 9 under ^ has size | # |. It follows that \<S\ = p and |0(9): 0| = (p - l)p'+1. Also, the 9, are ^-conjugate and hence all ker 9, are equal. Since f)ker 9, = 1, we conclude that ker 9 = 1. We can now apply the inductive hypothesis to 9 e Irr(H) and obtain a - 1 < ((f/p) - l)/(p - 1) + (t + 2)f/p = (f- l)/(p - 1) + (t + \)f/p <(/-l)/(p-D + (t + D/-l and the proof is complete. | We note that if p ^ 2, it is always possible to choose H in the above proof so that 0(x) = 0(9). The inequality in Theorem 14.20 is best possible. Proof of Theorem 14.19 Define the function a on positive integers by a(/c)= I[/c/(p-l)p']. f = 0 Since \_k/(p — l)p'] < k/(p — l)p' and this inequality is strict for large i, we have a(/c)<(/c/(p-l))f>-' = /cp/(p-l)2 i = 0 and the second inequality in the statement of the theorem follows. Since \_x + y] > |>] + [>], we have a.(k + I) > a(k) + a(/). We know that Xp is rational valued, where / is the given character of G. If Xp = Xi + li > where the /,- are rational valued, then working by induction on n we have a; < a(/j(l)), where |P/ker Xi\ = p"'- By the usual argument, a ^ at + a2
256 Chapter 14 and so a < a(Zl(l)) + afc2(l)) < a(Zl(l) + *2(1)) = a(n) as desired. We may now suppose that Xp is not the sum of two rational characters. It follows that Xp = |j=i £,■> where the <!;,■ e Irr(P) constitute an orbit under the Galois group 0 = 0(OP«/O). Let £ = £t. Then ker £,- = ker £ for all i and thus 1 = f)ker £t = ker £ and ^ is faithful. The stabilizer in <§ of £ is 0(OP«/O(£)) and it follows that the orbit size r = |0(£): Q| = (p - l)p' for some t > 0. Let/ = £(1) = £,(1) for all i so that n = (p - l)p'f. Now a(n) = I + p + p2 + ■■■ + p'f = (f- l)/(p - 1) + / + pf + ■ ■ ■ + p'f >(f - 1)/(P - 1) + (t + 1)/ > a, where the last inequality is by Theorem 14.20. | As an application of Schur's Theorem 14.19, we prove the following. (14.21) theorem Let G be a p-solvable linear group of degree n. Let P e Sylp(G) and | P : Op(G) | = p". Then a<, I[n/(p-l)p']. i = 0 The proof of Theorem 14.21 depends on some standard facts about p- solvable groups. The first is nearly trivial, namely that subgroups and factor groups of p-solvable groups are p-solvable. The second fact is a special case of what is often called "Lemma 1.2.3" since that was the designation in the paper of P. Hall and G. Higman where it first appeared. (14.22) lemma (Hall-Higman) Let G be p-solvable with Op(G) = 1. Then CG(Op.(G)) £ Op.(G). Proof Let H = Op.(G) and C = CG(Op.(G)). Then C<GandCis p-solvable. Let D = Op.(C) so that D <i G and thus D = HUH ^ C, then C/D is a non trivial p-solvable group and Op.(C/D) = 1. Thus Op( C/D) > land we let E/D = Op(C/D) and P e Sylp(£) so that P > 1. Now D = H = C(C) c N(P) and thus P <i P£> = £ <i G. It follows that P <i G. Since P > 1 and Op(G) = 1, this is a contradiction and the proof is complete. | Proof of Theorem 14.21 Let U = Op(G) and H/U = Op.(G/U). By the Schur-Zassenhaus Theorem, U is complemented in H and we may choose a complement K. Let JV = NG(K). Since G permutes the set of complements for U in H and H is transitive on this set, it follows that G = NH = NU. Let S e Sylp(JV) and put G0 = KS. We have N n 1/ £ S £ G0 and |G:l/|p = |JV:iVnl/|p=|G0:JVnt/|p.
Linear groups 2S7 We claim that JV n U = Op(G0) and thus it suffices to assume G = G0. To prove the claim, note that JVn(/<JV and so JV n U <i G0 and thus JV n t/ £ F = Op(G0). Since V -a G0, K -a G0 and FnK = l, we have V £ C(X) and thus VU/U £ QKU/U) = C(H/t/). Since Op(G/t/) = 1, Lemma 14.22 yields C(H/U) £ H/U and hence Ft//t/ £ H/U and Ft/ £ H. Since Ft/ is a p-group and U e Sylp(H), we have VU = U and thus V £ t/. Therefore, 7 £ JV n U and the claim is established. We may now assume that G0 = G so that K is a normal p-complement for G. Let / be a faithful character of G with /(l) = n. For each irreducible constituent 9 of Xk, let 9 be the canonical extension of 9 to /G(9) and let 9* = (9)G e Irr(G). Note that any field automorphism of Q|G| which fixes 9 also fixes 9* and it follows that 9* is p-rational. Choose a set y of representatives for the G-orbits of irreducible constituents of Xk and define \p = £9€y 9*. Then n = X{\)> X 9(1)|G:/G(9)|= £ 9*(1) = <«1). Also /K and t/^K have the same sets of irreducible constituents and so ker \p r\ K = ker x n K = 1. Thus ker »// is a p-group and ker \ji £ Op(G). Now application of Schur's Theorem 14.19 to G/ker \ji yields the desired result. | Write ap(/c) = Yj^o W(P ~ 1)p']- With further work one can replace the inequality a < ap(n) in Theorem 14.21 by a < ap(/zn), where fi = f if p = 2 and fi= (p — l)/p if p j= 2 is not of the form 2e + 1. With these improvements, equality can be obtained for all n. If G is solvable, this improvement is due to J. D. Dixon. The general p-solvable case was done by D. L. Winter and depends on the results of Chapter 13. Suppose G is a primitive linear group of degree n. In other words, the identity map is a primitive representation in the sense of Chapter 5. By Corollary 6.13, it follows that every abelian normal subgroup of G is central and thus by Jordan's Theorem 14.12 there are only finitely many possiblities for the group G/Z(G). These have been explicitly enumerated for certain small values of n. We give a sample of this type of result although this proof is not typical. (14.23) theorem Let G be a primitive linear group of degree 2. Then |G:Z(G)| = 12, 24, or 60. Proof Write Z = Z(G) and let /elrr(G) be faithful with /(l) = 2. If H £ G is nonabelian, then Xh e Irr(H) and it follows that Z(H) £ Z(/) = Z. Thus if g e G — Z, then C(g) is abelian.
255 Chapter 14 Let Sf be the set of maximal abelian subgroups of G. Then clearly G = \}S/> and Z = f)Sf. If A, Be & with Ar\B > Z, then C(A r\ B) is abelian and contains both A and B. Therefore A = B and thus G is a disjoint union: G = Z u \J(A- Z). For A e 9> we have \j>A, %A~] = 2 and [*z, Xz] = 4. Thus £ |Z(X)|2 = 2|A|-4|Z| = 2|A-Z|-2|Z|. jce A — Z This yields |G|= IlzWi2= £lz(*)|2 + I I IzWI2 jceG xeZ AsSf xeA-Z = 4\Z\ + 2%j\A-Z\-2\<?\\Z\ AeS? = 4\Z\ + 2(\G\ - \Z\) - 2s\Z\ where s = \Sf\. We thus obtain | G : Z| = 2s - 2. Now y is a union of conjugacy classes of subgroups. lfAeSf, we want to compute JV = NG(y4). If A < JV then JV is nonabelian and %N e Irr(JV). However, %A = X + fi with X =£ fi. It follows that | JV : IN(X) \ = 2 and the restriction of i to IN(X) is reducible. Thus 7^(1) is abelian and hence A = IN(X). We conclude that for all A eSf, we have |N(X) :A\ < 2. Also A -^ G. We now consider the group G/Z = G of order 2s — 2. Let 9~ = {A/Z\AeSf} so that 9" partitions G_and \2T\ = s. Also, 9~ is a union of conjugacy classes of subgroups of G. Write ST = &t u 3~2 u ■ ■ ■ u 3~T, where the 9~i are the distinct conjugacy classes. Let at be the common size of the subgroups in ^"( and let t{= \&~t\. IfBe^„ then |G:N(B)| = |G:B|/2or |G:N(B)| = \G:B\ and hence (1) t( = (s - l)/fl( or t, = 2(s - l)/a,. We also have (2) £'< = *. (3) ^ti(ai-l) = 2(s-l)-l, and all t( > 1 and a,- > 1. We may assume that a^ < a2< ■■■ < ar and that if flj = ai+1 then tt > ti+1. If a! > 2 then £ tj(a; - 1) > 2 £ tt = 2s which contradicts (3). Thus at = 2 and rt = s — 1 or (s — l)/2. However, if rt = s — 1, then (2) yields
Linear groups 259 t2 = 1, a contradiction. Thus rt = (s — l)/2. If a2 = 2, then t2 = (s — l)/2 forcing t3 = 1, which is impossible. Thus a2 S; 3. If a2 > 4, then 2(s - 1) - 1 = X ti(«i - 1) > (s - l)/2 + 3 £ t, i=2 = (s - l)/2 + 3(s - (s - l)/2) = 2(s - 1) + 3, a contradiction. Thus a2 = 3 and f2 = (s — l)/3 or 2(s — l)/3. Suppose t2 = 2(s — l)/3. Then s>tl + t2 = (s- l)/2 + 2(s - l)/3 = 7(s - l)/6 > s - 1 and hence s = 7(s — l)/6 which yields s = 7 and | G : Z | = 12. We now suppose that t2 = (s — l)/3. Since tt + t2 = 5(s — l)/6 < s, we have r > 3. If a3 > 6, Equation (3) yields 2(s - 1) - 1 = £ trffl. - 1) = (s - l)/2 + 2(s - l)/3 + £ ^(a,. - 1) i=3 > 7(s - l)/6 + 5(s - tt - t2) = 7(s - l)/6 + 5(s - l)/6 + 5, a contradiction. Thus a3 = 3, 4, or 5. If a3 = 3 = a2 then t3 < t2 and so f3 = (s — l)/3. This yields s>tl + t2 + t3 = (s- l)/2 + (s - l)/3 + (s - l)/3 and again s = 7(s — 1 )/6 and | G : Z | = 12. Suppose a3 = 4. If f 3 = (s — 1 )/2, we have s>(s- l)/2 + (s - l)/3 + (s - l)/2 = 4(s - l)/3 > s - 1 and thus s = 4. This yields | G | = 6, which is impossible since a3 must divide | G\ by Lagrange's theorem. Thus t3 = (s — l)/4 which yields s>(s- l)/2 + (s - l)/3 + (s - l)/4 = 13(s - 1)/12 > s - 1. Thus s = 13(s - 1)/12 and s = 13 and | G : Z | = 24 in this case. The remaining case is a3 = 5. If t3 = 2(s — l)/5, this yields s>(s- l)/2 + (s - l)/3 + 2(s - l)/5 = 37(s - l)/30 > s - 1. since this has no integer solution, we have t3 = (s — l)/5 and s > 31(s - l)/30 > s - 1 and s = 31. This gives | G : Z| = 60 and the proof is complete. |
260 Chapter 14 We mention that all three cases of Theorem 14.23 can occur. If G = SL(2, 3), we get G/Z =* A4 of order 12. If G = GL(2, 3), we get G/Z S Z4 of order 24 and if G = SL(2, 5), we get G/Z S A5 of order 60. In fact, A4, E4, and A5 are the only possibilities for G/Z. Problems (14.1) Let G be a linear group of degree n. Show that there exists another linear group G* of degree n such that G/TXG) S G*/Z(G*) and det(g*) = 1 for g* e G*. Do this in such a way that G* is irreducible iff G is irreducible. Hint Let / be the n x n identity matrix and let S = {a/|a e C, a" = det(g) for some g e G}. Consider GS £ GL(n, C). (14.2) Let x £ Irr(G) be p-rational and faithful and assume that Z(l) < k(p - 1) for some k < p. Show that the Sylow p-subgroups of G are elementary abelian of order <pk. (14.3) Let G be an irreducible p-solvable linear group of degree p" > 1. Let N = Op.(G) and U/N = Op(G/N). Show that U is nonabelian. (14.4) Let G be a linear group of degree n and suppose that n is not divisible by any prime power q > 1 such that q = 1, 0, or — 1 mod p ,where p is some prime. Assume that G has a solvable irreducible normal subgroup. Show that a Sylow p-subgroup of G is normal. Hint Let P0 e Sylp(S) where S is normal, irreducible, and solvable and consider C = CG(P0) <i G. (14.5) Let G have a normal p-complement N and assume that a Sylow p- subgroup of G is not normal. Suppose that G is a linear group of degree p — 1. Show that N/(N n Z(G)) is a 2-group and thus iV is nilpotent. Hint Let P e Sylp(G). Show that CN(P) £ Z(N). (14.6) Let G be a linear group of degree p — 1, where p is a prime. Suppose there exists p-solvable M <i G such that M does not have a normal Sylow p-subgroup. Show that G is solvable. (14.7) Let G be an irreducible linear group of degree p + 1 where p is an odd prime. Let P e Sylp(G) and suppose P^i G. Assume that G has a normal p-complement, N. Show that p + 1 is a power of 2. Hints Let G be a minimal counterexample and let q be an odd prime divisor of | N: CN(P) \. Let Q e Syl,(JV) and show that NN(Q) is abelian.
Problems 261 (14.8) Replace the hypothesis that G has a normal p-complement in Problem 14.7 by the weaker condition that G is p-solvable. Draw the same conclusion. Hints Let U = Op(G) and M/U = Op.(G/U). It is no loss to assume that G/M is a p-group and that Op'(G) = G. If U £ Z(G), let T = IG(1), where X is a linear constituent of %v and / is the given faithful character of G. Show that Xt has an irreducible constituent of degree p and that M nT<M. (14.9) Let G be a linear group which is generated by two elements, x and y. Suppose that each of x and y have only two distinct eigenvalues. Show that the irreducible constituents of G (that is, of the given faithful representation) have degree at most 2. Hint If V is a nonzero vectorspace over C and Vu V2, V3, and V± are subspaces such that V = Vt + V} whenever i ^ j, then there exists a subspace U = 7 such that dim 1/ = 2 and 1/ n V- ^ 0 for all i. (14.10) Let G be a solvable linear group of degree n. Show that G has a nilpotent normal subgroup of index < n\. (14.11) Let p be a prime. Show that for every integer n > 0, there exists a p-solvable linear group G of degree n with | P : Op(G) | > pa, where P e Sylp(G) and
Changing the characteristic Some of the deepest and most powerful results in group representation theory involve "modular" representations, that is, representations over fields of prime characteristic. These are important for at least two reasons. First, if K, H o G, with K ^ H, and H/K is an elementary abelian p-group, then H/K may be viewed as an F[G]-module, where F is the field of order p. In this situation, the representation theory can give direct structural information about G. Perhaps an even more important way in which the characteristic p representations of G are relevant is that they can give new information about the characteristic zero situation. This is especially the case when | G | is divisible by p since then it is possible to obtain results which relate the p-subgroups of G with the properties of Irr(G). Our emphasis in this chapter will be on the relationship between the absolutely irreducible characteristic p representations of G and Irr(G). Following R. Brauer, who was the originator of this theory, we shall focus our attention on characters rather than on modules or representations. (In fact, we shall not even mention the indecomposable but not irreducible modules which seem to be crucial for many of the deeper results.) The objective of this chapter is to familiarize the reader with the principal definitions and the most basic results of the theory. We do not attempt to give a comprehensive treatment of the subject and we shall not prove every fact that is mentioned. We establish some notation which will remain fixed throughout this chapter. Let R be the full ring of algebraic integers in C and let p be a prime. We construct a particular field F of characteristic p by choosing a maximal 262 15
Changing the characteristic 263 ideal M 2 pR of R and setting F = R/M. (Note that there is a certain amount of arbitrariness here since M is not uniquely determined.) Let * denote the natural homomorphism /? -> F. Since pi* = p* = 0, it follows that char(F) = p as claimed. (15.1) lemma Let U = {eeC|em = 1 for some meZ with pjfm}. Let/?, F and * be as above. Then (a) 1/cR; (b) * maps U isomorphically onto F *; (c) F is algebraically closed and algebraic over its prime field. Proof Clearly U £ Rand so * is defined on U. If ae U - {1}, then a is a primitive nth root of 1 for some n > 1 with/^n and hence n-l 1 + X + ■ ■ ■ + X"-1 = (x" - \)/(x - 1) = [] (* - **)■ i=l Setting x = 1, we conclude that 1 — a divides n in R. If a* = 1, then (1 — a)* = 0 and thus n* = 0. Since p* = 0 and (p, n) = 1, it follows that 1* = 0, a contradiction. Thus * maps U isomorphically into F*. If a e F, then a = a* for some a e R and there exists monic/ e Z[x], with f(a) = 0. Let K £ F be the prime subfield. Then 0 # /* e K[x] and/*(a) = /(a)* = 0. Thus F is algebraic over K. To complete the proof, let E be an algebraic extension of F. Then [/*cf c £* and it suffices to show that £* £ I/* in order to conclude that I/* = F* and that F is algebraically closed. Let f} eE". Then 0 is algebraic over F and hence over K and thus (}m — 1, where m = |K(/?)| — 1. Now pjfm and so I/* contains m roots of xm — 1. Thus /Set/* and the proof is complete. | Continuing with the above notation, let 3E be an F-representation of a group G. Let y be the set of p-regular elements of G, that is, elements of order not divisible by p. We define a function q>: y -> C as follows. Let xe^ and let eu£2,...,£feF* be the eigenvalues of 3E(x), counting multiplicities. (Thus/ = deg 3E and £ et = t/^(x) where t/f is the F-character afforded by 3E.) For each i, there exists a unique m, e 1/ such that (m,)* = e,-. Let cp(x) = £ m( . The function <p: ^ -» R £ C is called the Brauer character of G afforded by 3E. Note that similar F-representations afford equal Brauer characters and that Brauer characters are constant on conjugacy classes. Both of these statements follow from the fact that 3E(x)andP~ lX(x)P have the same eigenvalues. Also, if x e y, then q>{x~l) = q>(x) since the eigenvalues of 3E(x~l) are the reciprocals of those of H(x) and for u e U we have (u)* = (u ~l)* = («*) ~'.
264 Chapter 15 (15.2) lemma Let 3E be an F-representation of G which affords the Brauer character cp and the F-character \ji. For g e G, let x = gp^ e if. Then cp(x)* Proof We have g — xy, where x, y e <g>, pJfo(x) and o(y) is a power of p. Replace 3E by a similar representation so as to assume that X(g) is in upper triangular form. Since X(g) = X(x)X(y) it follows that the eigenvalues a, of X(g) can be factored, a, = e,<5f, where e^ ..., ef are the eigenvalues of 3E(x) and (dif^ = 1. Since o(y) is a power of p, we have 8{ = 1 and a, = e,. Thus \p(g) = \p(x). That \p(x) = cp(x)* is immediate from the definition of cp. | Lemma 15.2 provides one reason why we only bother to define Brauer characters on p-regular elements: this is sufficient to reconstruct the full F-character afforded by 3E. Some words of caution are appropriate here. Given a group G and a function, cp: if -> C, where if is the set of p-regular elements of G, it is not always meaningful to ask if cp is a Brauer character unless the ideal M ^ Ris specified or some other additional information is given. Examples exist where cp is a Brauer character with respect to some choice of M and is not one when some other maximal ideal is chosen. Also, if a is an automorphism of the complex numbers and cp is a Brauer character of G, then cp" need not also be a Brauer character where cp" is defined by cp"(x) = cp(x)a for xtif. (15.3) lemma Let cp be a Brauer character of G. Then cp, the complex conjugate function, is also a Brauer character. Proof Let 3E be an F-representation affording cp, where F = R/M, as usual. For g e G, define 9)(g) = 3E(gT 1)T and observe that 9) is an F-representation of G. If £!,..., ef are the eigenvalues of H(g\ then g! ~\ ..., e^~' are the eigenvalues of H(g)~1 = X(g~l) and hence of 9)(g). The result now follows. | Let 3Et,..., 3Er be a set of representatives for the similarity classes of irreducible F-representations of G and let cp{ be the Brauer character afforded by 3E,-. We say that the cpt are irreducible Brauer characters and we write IBr(G) = {cpt}. When we use this notation, it is understood that a particular prime p and maximal ideal M have been fixed. It is routine to prove that sums of Brauer characters are Brauer characters and that every Brauer character is of the form £ n,-<Pj, where the nx e Z are nonnegative and not all zero. To prove that the cpt are linearly independent we need the following fact from algebraic number theory. Although it is not terribly deep, we omit the proof.
Changing the characteristic 265 (15.4) lemma Let <*!,..., ameC be algebraic over Q and let /be a proper ideal of R, the ring of algebraic integers. Suppose that not all a.t = 0. Then there exists /? e C such that /?a( e R for all i but not all /?a( e /. Proof Omitted. | (15.5) theorem The irreducible Brauer characters q>t are distinct and linearly independent over C. Also, if £ u-i<p{x) = 0 mod M for all x e ^ with a, e R, then all a, = 0 mod M. Proof We prove the second statement first. Since the 3E, are absolutely irreducible, we have X<(F[G]) = Mfi{F), where/J = deg 3E(. It follows that we can choose b{ e F[G], with \jif]bi) = 1, where i/^ is the character afforded by Xt. Also, by Theorem 9.6, we may suppose that il/£bj) = 0 if i =£ )■ We have £ u-i<p{x) = 0 mod M for all x e y and thus £ oi-*^,{g) = 0 for all g e G by Lemma 15.2. It follows that £, a^ip^bj) = 0 for ally, and thus a(* = 0 for all i. This proves the assertion. Now let £ be the algebraic closure of Q in C and suppose that £ a(<p( = 0 with a(e£. If not all ol{ = 0, we apply Lemma 15.4 and choose /? with all /fa, e R but not all /to, e M. Since £ (fiv-^<Pi = 0, this contradicts the first part of the proof. Thus all a, = 0 and the cpt are linearly independent over E. Since the q>t have values in E, it follows by elementary linear algebra that they remain linearly independent over any extension field of E. | The principal reason that Brauer characters are important is that they provide a link which connects Irr(G) with the characteristic p representations ofG. (15.6) theorem Let x be an ordinary character of G and let % denote the restriction of / to Zf, the set of p-regular elements of G. Then % is a Brauer character of G (for any choice of M). In order to prove Theorem 15.6, we need to consider a somewhat larger ring than R. As always, we assume that we have fixed a particular maximal ideal, M 2 pR. Let R = {a/0|a, $ eR, 0$ M} £ C and observe that R is a ring and R 2 R. Let Si = {a//? | a e M, /? 6 R — M}. Then M is an ideal of R and every element of R — M has an inverse in R. It follows that M is the unique maximal ideal of R. We call R a ring of local integers for the prime p. We extend the homomorphism *: R -> F to R by defining (a//?)* = a*//?*. Note that if? is the kernel of this extension and M = At n /?. (15.7) lemma (Nakayama) Let R and .A? be as in the preceding and let V be a finitely generated R-module. Suppose V = VAi + U for some sub- module U £ K Then 1/ = K
266 Chapter 15 Proof Since V is finitely generated, we can choose vuv2,- ■. ,vneV such that V = Yj vt M + U.Do this with the minimal possible n. If n = 0, then U = V and we are done. Suppose n > 1 and write n v „ = £ vimi + "> i=l where mteM and «e[/. Thus n-l Since m„ e M, we have I — m„eR —M and thus 1 — m„ is invertible in R and t>„ e YjZl ViAl + U. It follows that V = Y]=l vtM+ U and this contradicts the minimality of n and completes the proof. | Suppose 3E is a C-representation of G with the property that all entries in the matrices H(g) for g e G lie in R. We can construct an F-representation 3E* of G by setting 3E*(gf) = 3E(g)*, that is, we apply * to every entry of X(g). The following includes Theorem 15.6. (15.8) theorem Let 3E be a C-representation of G. Then there exists a C- representation 9) similar to 3E such that all entries in ?)(g) lie in R for all g e G. If 9) is any such representation, then the F-representation 9)* affords the Brauer character % where / is the ordinary character afforded by 3E (and by$). Proof Let E be the algebraic closure of Q in C. By the results of Chapter 9, every C-representation of G is similar to one of the form 3C for some £- representation 3- It therefore suffices to assume that 3E is an £-representation and to produce a similar £-representation ?) with entries in R. Let V be an £[G]-module corresponding to 3E and let vu ...,v„ be an £-basis for V. Let W be the £-span of the finite set {vtg\ 1 < i < n,geG} so that W is a finitely generated R-module which is G-invariant. Let t: W -> W/WM be the natural homomorphism and view W/Whl as an F-vector space via (w-r)a* = (woc)t for aeR. Let {vVj-r} be an F-basis for W/Wffi so that (£ WjR)z = W/WM and W = WM + £ WjR~. By Nakayama's Lemma 15.7, we have W = £ w; R and thus the wj span V over £ since W contains an £-basis for V. We claim that the w} are linearly independent over £. Suppose that £ Wj-a; = 0 with Zj e£ and not all a,- = 0. By Lemma 15.4, we can multiply by a suitable /? e £ and assume that all a, e /? £ /? but that not all a;* = 0. Now 0 = (£ Wj-0;)t = £ (wyT)a^* and this contradicts the linear independence of the W;T and proves the claim.
Changing the characteristic 267 Now let 9) be the F-representation of G corresponding to V with respect to the basis {w;} so that 9)(#) = (a;j), when w,g = £ WjOij. However, w,g e W = £ wj^ and so all au e R. This completes the proof of the first assertion. Now for p-regular g e G, we need to compute the eigenvalues of V*(g) = Wq)*- Let / be the characteristic polynomial of 9)(gr) so that f(x) = det(x/ - 9)(#)). Then /e/?[x] and /* eF|>] is the characteristic polynomial of 9)(g)*. Write/(x) = Y[(x - 1{) and note that the eigenvalues 1; lie in J? c r. Then/*(x) = Y\(x — A,*) and hence the eigenvalues of 9)*(g) are the X;*. Thus %{g) = £ X-, is the value of the Brauer character afforded by 9)* at the element g. The proof is complete. | In the foregoing proof, it is not the case that the character / uniquely determines the F-representation 9)* up to similarity. It is possible that / is afforded by another C-representation 3 with entries in R such that ?)* and 3* are not similar over F. Of course, / does uniquely determine the Brauer character % and we may write % = £ n9cp where cp runs over IBr(G) and 0 < n e Z. The coefficients n^ are uniquely determined because of the linear independence of IBr(G) and since nv is the multiplicity of a particular irreducible F-representation as a constituent of 9)*, it follows that 9)* and 3* have the same irreducible constituents with the same multiplicities. Since 9)* and 3* need not be completely reducible, it does not follow that they are similar. (15.9) definition Let x e Irr(G) and let % be the restriction of / to the p- regular elements of G. Write «>ElBr(G) The uniquely defined nonnegative integers dM are the decomposition numbers of G for the prime p. We view the decomposition numbers as forming a |Irr(G)| x |IBr(G)| matrix, called the decomposition matrix. Although the decomposition numbers are not even defined until the maximal ideal M is chosen, it is a fact (whose proof we omit) that the decomposition matrix of G for the prime p is uniquely determined up to permutations of the rows and columns. (15.10) theorem The decomposition matrix (dxill) has linearly independent columns. Also, IBr(G) is a basis for the space of C-valued functions defined on p-regular elements of G and constant on conjugacy classes. Proof Let V be the Space of p-regular class functions and let W c y be the span of IBr(G). Let U be the span of the columns of (dX(/1) so that dim U < |IBr(G)| = dim W < dim V,
268 Chapter 15 where the equality follows from Theorem 15.5. The theorem will follow when we show that dim V < dim U. It therefore suffices to find a one-to-one linear map from the dual space V of V into U. The elements of U are columns indexed by / e Irr(G). For a e V, define u = col(a(#)). To show that ueU, observe that a(#) = a(^ dx<p<p) = Y^ dx<pa.((p). Thus u is a linear combination of the col(rf^) for <p e IBr(G). We map 9 -> U by a i-> col(a(#)). If a(#) = 0 for all / e Irr(G), we claim that a = 0. To see this, let 9 e V and extend 9 to a class function 9t of G. Write 9t = £ ax% so tnat ^ = X flx# and a(9) = 0 as desired. The result now follows. | (15.11) corollary The number of conjugacy classes of p-regular elements of G is equal to | IBr(G) |. This is also the number of similarity classes of irreducible ^-representations of G for every splitting field K of characteristic p. Proof The first assertion is immediate from Theorem 15.10. To prove the second statement, we may replace K by its algebraic closure by Corollary 9.8. Assuming that K is algebraically closed, it contains an isomorphic copy of F which is an algebraic closure for Zp. By Corollary 9.8 again, we may assume that F = K. The result now follows from the fact that nonsimilar irreducible F-representations afford distinct Brauer characters. | (15.12) corollary If q> e IBr( G), then there exists / e Irr(G) with dX(p =£ 0. Note that for each / e Irr(G), it is trivial that there exists q> e IBr(G) with d„± 0 since 0 * *(1) = £ dX9cp(l). The interesting case of this theory is when p 11G |. The reason for this is given by the following. (15.13) theorem Suppose pJf\G\. Then IBr(G) = Irr(G). Proof In this case, the group algebra F\_G] is completely reducible and thus by Corollary 1.17 we have \G\ = dim F[G] = £ (deg 3E,)2, where the 3Ej are a set of representatives for the similarity classes of irreducible F-representations. If 3E; affords (p; e IBr(G), then deg 3E, = <p;(l) and hence x «ku' = igi= x x(i)2 = ife<w>(i)Y «>eIBr(G) xelrr(G) X \d> J = X <p(1M1)I^„. If q> t6 (J-, then ^ dX(pdxll > 0 and if (p = fi, then ^ dx<pdxtl > 1 by Corollary 15.12. It follows that these inequalities are all equalities. We now have ^ (dXll,)2 = 1 and hence for each q>, there is a unique / such that dM ¥= 0, and in fact dxv = 1. Since ^ dXvdxll = 0 if cp ^ /z, it
Changing the characteristic 269 follows that for each / e Irr(G), there is a unique cp with dx<p ^ 0. We conclude that % = % = q> and the result follows. | A necessary condition that a function 9: £f -> C be a Brauer character is that for each subgroup H £ G with pJf\H\, the restriction 9H is a Brauer character and hence is an ordinary character of H. This is something which can be checked. It is trivial that every q> e IBr(G) is a C-linear combination of the Brauer characters % for / e G. This may be seen by extending q> to a class function of G. (15.14) theorem Let <peIBr(G). Then q> is a Z-linear combination of ttlzelntG)}. Proof Extend <p to a class function 9 of G by setting 9(g) = q>(gP'). It suffices to show that 9 is a generalized character of G. We appeal to Brauer's Theorem 8.4(a). Let E £ G be elementary and write E = P x Q, where P is a p-group and pX\Q|. If geE, write g = xy with xeP and y eQ. Then 9(g) = <p(y) and so 9£ = 1P x <pQ. By Theorem 15.13, q>Q is a character of Q and hence 9£ is a character of E. It follows that 9 is a generalized character and the proof is complete. | We digress to show another way in which Theorem 15.13 can be used. (15.15) lemma Let E be an algebraically closed field of characteristic not dividing | N |, where N is a group. Let H act on N and suppose that CH(n) = 1 for all 1 ¥= neN. Let ?) be a nonprincipal irreducible £-representation of N and write ty\nh) = 9)(n) for n e N and heH. Then the representations 9)*1 are pairwise nonsimilar for heH. Proof Let K be the algebraic closure in E of the prime subfield of E. Then $ = 3£ f°r some irreducible K-representation 3 of N. It suffices to show that the 3*1 are pairwise nonsimilar by Corollary 9.7 and hence we may assume that K = E. If char(£) = 0, then (up to isomorphism) E £ C and the representations (9)c)'1 are pairwise nonsimilar by Theorem 6.34 and Problem 7.1 which proves this case. Suppose char(£) = p. Choose a maximal ideal, M 2 pR of R and let F = R/M as usual. Then F = E and we may assume F = E. Let ?) afford q>eIBr(iV) = Irr(iV). By Theorem 6.34, the characters <ph for heH are all distinct and the result follows. | The following result is often quite useful.
270 Chapter 15 (15.16) theorem Let G be a Frobenius group with kernel N and let K be a field with characteristic not dividing \N\. Let V be a K[G]-module and suppose that CV(N) = 0. Let H be a Frobenius complement for G. Then V has a basis which is permuted by H with orbits of size \H\. Also, if H0 £ H, then dim CV(H0) =\H:H0\ dim CV(H). Proof The second statement follows from the first since if b is a basis permuted by H, then dim CV(H0) is equal to the number of orbits of the action of H0 on b. Since each orbit of H0 on b has size \H0\, the assertion follows. To prove the first statement we argue that it suffices to assume that K is algebraically closed. Let 3E be a K-representation of G corresponding to V and let E 2 K. The condition that CV(N) = 0 is equivalent to the restriction 3E,v having no principal constituent and this property is inherited by (XE)N by Theorem 9.6. The conclusion of the theorem is equivalent to 3EH being similar to a representation 9) in block diagonal form with each block being the regular representation of H. If we can prove that (3EH)£ and 9)£ are similar, then 3EH and 9) are similar by Problem 9.5. It follows that we may replace K by any extension field and thus we assume that K is algebraically closed. Now let Jt be a representative set of irreducible K[iV]-modules. Since H acts on the set of similarity classes of ^-representations of JV, we can define a corresponding action of H on Jt. Let Jt0 £ Jt be a set of representatives for the H-orbits of nonprincipal K[iV]-modules in Jt. In the notation of Definition 1.12, let W = X M{V) £ V. MeJ0 We claim that V = £ -hBll Wh. This will suffice to prove the result since we obtain a basis for V by choosing any basis for W and taking all H-trans- lates. To prove the claim, observe that Jt = [jhBH (Jt0)h u {I/}, where U is a principal module. By Lemma 15.15, this union is disjoint. Since VN has no principal constituent, Lemma 1.13 yields V=I-hJ I M(V)\ We*)1 / The result now follows from Lemma 6.4. | We now resume consideration of the general case where p can divide \G\. We introduce the concept of "blocks" which is at the heart of Brauer's theory. For each / e Irr(G), we have the algebra homomorphism cox: Z(C[G]) -» C as in Chapter 3. The function cox is determined by its values on the conjugacy
Changing the characteristic 271 class sums Kt which form a basis for Z(C[G]). Furthermore, the values cox(Ki) lie in R. For /, \ji e Irr(G), write / ~ \ji if co^K)* = co^Kf)* for all i. This clearly establishes an equivalence relation on Irr(G). (15.17) definition Ap-ft/oc/cofGisasubsetB £ Irr(G) u IBr(G) such that (a) B r\ Irr(G) is an equivalence class under the relation ~ defined above. (b) B r\ IBr(G) = {q> e lBr(G)\dXll, ± 0 for some /efln Irr(G)}. From its definition, it appears that the equivalence relation ~ on Irr(G) depends on the choice of the maximal ideal M. In fact, this is not true. (15.18) theorem Let x, i> £ Irr(G). Then / and \ji lie in the same p-block iff cox(K) — co^(K) lies in every maximal ideal of R which contains pR for every class sum K. Proof The "if" statement is clear. Assume K~ i> and fix K. Let a = cox(K) — oi^,(K). We shall show that a" e pR for some integer n and the result will follow. Let 0 = #(Q|G|/Q), the Galois group, and let a e<&. Let e be a primitive | G|th root of unity so that e" = em for some m, with (m,\G\) = 1. Let g e G be in the class with sum equal to K and let L be the sum of the class containing gm. We have x(gf = X(gm). Also \C(g)\ = \C(gm)\ since <g> = <<f>. It follows that cox(K)a = cox(L) and similarly co^(K)a = co^(L). It follows that of = cox(L) — co^(L) e M since / ~ \ji. Let f(x) = f\aEe (x - a") e(Q n /?)[*] = Zpc]. Since /eM for all a e ^, all of the coefficients of/except for the leading one lie in M n Z = pZ. Thus 0 = /(a) = a" mod pR, where n = | # |. The proof is now complete. | Although it is clear from the definition that every / e Irr(G) lies in a unique p-block, the analogous statement about IBr(G), though true, is not so obvious. To prove it, we relate the p-blocks of G to the F-algebra, Z(F[G]). We extend the map *: R -> F to a ring homomorphism *: £[G] -» F[G] by setting g* = g for g e G. (Here, R[_G] is simply the R-span of the group elements in C[G].) Since the class sums K{ form a basis for Z(C[G]) and their images KfeFlG] form a basis for Z(F[G]), it follows that Z(R[G]) = R[G] n Z(C[G]) maps onto Z(F[G]) via *. Now suppose x e Irr(G). Then cox maps Z(^[G]) to R and we can define cox*: Z(F[G]) -^ F by setting cox*(z*) = cox(z)* for z e Z(R[G]). This is well defined since if zu z2 e Z(R[G]), with zt* = z2*, then zt - z2 has coefficients in M and thus (eo^Zi - z2))* = 0. It is trivial to check that cox*: Z(F[G]) -» F is an algebra homomorphism and that if /, \\i e Irr(G), then eo/ = co^* iff X~ i>\ that is, iff / and \ji lie in the same p-block.
272 Chapter 15 Let B1(G) be the set of p-blocks of G. If BeBl(G), let kB = cax* for X e B r\ Irr(G). Thus B >-* kB is a one-to-one map from B1(G) into the set of algebra homomorphisms Z(F[G]) -» F. (15.19) theorem Let cp e IBr(G). Then cp lies in a unique p-block B. If <i) is an irreducible F-representation which affords cp, then ^(u) = AB(u)7 for all ueZ(F[G]). Proo/ By Corollary 15.12, dw ^ 0 for some /elrr(G). Let BeBl(G), with / e B. Then <p e B. Let 9) afford <p. We will be done when we show that sJ)(u) = kB(u)I for all u e Z(F[G]) since this equation uniquely determines kB and thus uniquely determines B. Let X be a C-representation which affords / and which has entries in R (Theorem 15.8). Thus X* affords the Brauer character % and by the linear independence of IBr(G), it follows that 9) has multiplicity dx<p > 0 as a constituent of 3E*.Now letueZ(F[G])and write u = z*forsomezeZ(£[G]). Then X*(u) = X(z)* = (ojx(z)I)* = oi*(u)l = lB(u)I. Since <i) is a constituent of X*, the result follows. | We define a graph with Irr(G) as its vertex set by linking /, \\i e Irr(G) iff there exists q> e IBr(G) such that dw and d^^ are both nonzero. This is called the Brauer graph. If / and \ji are linked by cp, it follows that / and \\i lie in the same p-block, namely the unique one which contains cp. This proves the following. (15.20) corollary Let B be a p-block of G. Then B r\ Irr(G) is a union of connected components of the Brauer graph. We shall see that in fact B n Irr(G) is a single connected component. One of the principal benefits of considering blocks in various applications of the theory is that in certain circumstances we can replace equations like Xelrr(G) (which arise from the second orthogonality relation) by equations like Z x(x)xJy) = o, Xelrr(G) nB where B is a p-block. In particular, this holds whenever xp and yp are not conjugate in G. We shall prove a weak form of this "block orthogonality." For each cp e IBr(G) we define %= 1 dx<pX. Xelrr(G)
Changing the characteristic 273 The O's are called projective characters of G. (There is no connection with projective representations in the sense of Chapter 11, but there is a connection with projective modules in ring theory. The O's are also called "principal indecomposable characters" in the literature.) (15.21) lemma Let si ^ irr(G) be a union of connected components of the Brauer graph and let ^ = {<peIBr(G)| dx<p ^ 0 for some /e si). Let x, y e G with pjfo(x). Then EzMxO0 = Z (plx)®^). Proof For / e si, we have X(x) = Z fiU<P(x) since dxtl = 0 for all y. e IBr(G) — 28. Also, if <p e ^, then since d^ = 0 for all £ e Irr(G) - j/. Now Z zMztF) = Z «U<pMx(F) = Z pM^OO and the proof is complete. | (15.22) corollary For each <peIBr(G) we have %(y) = 0 if p\o(y). Furthermore, | P\ divides <5„(1) where P e Sylp(G). Proof Let x e G be p-regular and let p | o(y). By the second orthogonality relation we have Z Z(x)zO0 = O. Xelrr(G) By Lemma 15.21 with si = Irr(G) and ^ = IBr(G), we conclude that Z cp(xW& = 0. «>ElBr(G) Since this holds for all p-regular x e G, the linear independence of IBr(G) yields Qjly) = 0 for all q>. The second statement follows since \P\ [(Ov)P, 1P] = Ov(l). | (15.23) corollary (Weak Block Orthogonality) Let x, y e G with pJfo(x) and p | o(>>). Let B be a p-block of G. Then Z X(*)ZOO = 0. XeBnlrr(G)
274 Chapter 15 Proof Apply Lemma 15.21 with jsi = Irr(G) n B and Si = IBr(G) r\ B. Since O^tv) = 0 for all q>, the result is immediate. | Before going on to develop more of the theory of blocks we digress to give an application of what we have already done. (15.24) theorem (Brauer) Let G be a simple group of order p"qbr where p, q, r are distinct primes. Let S e Sylr(G). Then S = CG(S). The p-block containing 1G is called the principal block. (15.25) lemma Let G be simple and let B be the principal p-block of G. Suppose x £ B r\ Irr(G) and x(l) is a power of p. Then i = la- Proof Let g e G and let K e C[G] be the sum of the elements in Cl(g), the conjugacy class containing g. Then co^K) = \Cl(g)\ and since / and 1G lie in the same p-block, we have i \ x(g)ICl(g)| (*) —-7JT— = ^(K)= |C%)|modM. Now let P e Sylp(G). If P = 1, then since /(1) divides | G |, we have /(1) = 1 and by simplicity / = 1G. Assume then, that P > 1 and take g e Z(P) — {1}. Then pJf\Cl{g)\ and so |C%)| #0modM since M n Z = pZ. Thus (*) yields %(g) =£ 0. However (/(l), |C%)|)= 1 and Burnside's Theorem 3.8 yields that g e Z(/). Since g ¥= 1 and G is simple, it follows that / = 1G. | Proof of Theorem 15.24 Suppose CG(S) > S. Then there exists xeG of order pr or qr. Say o(x) = pr and let B be the principal p-block of G. By Corollary 15.23, we have -1= Z zd)zW+ I z(i)zW. We claim that the second sum is zero. If /eB n Irr(G) and / j= 1G and 4^Z(1)> then r|/(l) by Lemma 15.25. Since r\o(x), we have z(x) = 0 by Theorem 8.17 and the claim follows. We conclude that — l/q is an algebraic integer and this contradiction completes the proof. | For i e Irr(G), let ex e Z(C[G]) be the idempotent corresponding to /. By Theorem 2.12 we have 7^7 IW)g. |G| gBG
Changing the characteristic 27S Note that in general, ex $ R\_G~\ (where R is a ring of local integers for p) and so we cannot find idempotents in F[G] simply by applying * to ex. Since ex e^ = 0 if x ¥= ip, sums of the various e^'s are also idempotents and we shall consider when such a sum lies in R[G]. Note that if si <=, Irr(G), then the set si can be recovered from s = Yxe* ex smce ^ = ixe^rr(G)\sex ¥= 6}. (15.26) theorem (Osima) Let si be a connected component of the Brauer graph and let/ = Yxe* ex. Write/ = Y agg. Then (a) ag = (l/\G\)^x^X(DM; (b) a9e£forallgeG; (c) ag = 0 if p\o(g). Proof Statement (a) is immediate from the formula for ex in Theorem 2.12. By Lemma 15.21, we have ag = (1/|G|) YjI> ^ ss VW^JsX where 3b = {<p e \Br(G)\dx<p ¥= 0 for some / esi}. If p\o(g), then O^fer) = 0 by Corollary 15.22 and (c) follows. Now assume that pjfoig). Lemma 15.21 then yields ag = (l/\G\)£%(im). However, \P\ divides <B„(1) where Pe Sylp(G) and thus <5„(1)/|G| ek~. Since cp(g) e R, it follows that ageR and (b) is proved. | (15.27) theorem The connected components of the Brauer graph are exactly the sets Irr(G) n B for p-blocks B. Furthermore, every set si c Irr(G) such that Yxe* exeR\_G] is a union of sets of the form Irr(G) n B. Proof We prove the second statement first. If / e Irr(G) is afforded by 3E, then 5.(ex) is the identity matrix and 3E(e^) = 0 if \p =£ %. Thus oix(ex) = 1 and eo^) = Ofori/f ^ /.Write/ = Y* e^. It follows that/ej^ iff co^/) = 1 and otherwise cox(f) = 0. Thus / e si iff cox(f)* ¥= 0. Now if/ e R[G], then all oix(f)* are equal as / runs over Irr(G) n B for a block B. The assertion follows. Now let si be a connected component of the Brauer graph. By Theorem 15.26, Y* exe ^[G] and tnus -^ is a union of sets of the form Irr(G) n B. Since each Irr(G) n B is a union of connected components of the Brauer graph, the result follows. | We now have three different characterizations of the sets B n Irr(G). In addition to their definition as equivalence classes under ~, they are also the connected components of the Brauer graph and they are the minimal nonempty subsets si £ Irr(G) such that Yxe* exe K[G]. The following is a strengthening of Theorem 15.14.
276 Chapter 15 (15.28) lemma Let B be a block of G and let <p e IBr(G) n B. Then q> is a Z-linear combination of Brauer characters of the form # for / e Irr(G) n B. Proof By Theorem 15.14, there exist bxeZ such that q> = Y.x^i"iO) bx% Thus q> = 9B + 90, where 9B = X M and 9«= I bxt *eIrr(G)nB *eIrr(G)-B Now we express 9B and 90 in terms of IBr(G) using the decomposition numbers dxll for fi e IBr(G). If / e B, then dxll = 0 when fi $ B and hence 9B is a linear combination of fi e B. Similarly, if / £ B, then dxil = 0 when fieB and hence 90 is a linear combination of /z £ B. The equation <p = 9B + 90 and the linear independence of IBr(G) now yield q> = 9B and the proof is complete. | (15.29) theorem Let B be a p-block of G. Then |BnIrr(G)| > |BnIBr(G)|. Let / e B n Irr(G). Then the following are equivalent. (a) |BnIrr(G)| = |BnIBr(G)|. (b) p^(|G|/z(l)). (c) B n Irr(G) = {/}. Also in this case, B n IBr(G) = {#}. Proof Let £> = (d^) be the decomposition matrix and let DB be the submatrix corresponding to the rows and columns indexed by elements of B. For each q> e B, the part of the corresponding column of D outside of DB consists of zeros. Since the columns of D are linearly independent by Theorem 15.10, it follows that the columns of DB are linearly independent and thus DB has at least as many rows as columns. Now assume (a). Then DB is a square nonsingular matrix and we let DB~l = (avX). For fixed / e Irr(G) n B, we have and so / is a linear combination of the Ov and hence vanishes on elements of order divisible by p (Corollary 15.22). If PeSylp(G), then \P\Ixp, lp] = Z(l) and (b) follows. Assuming (b) we have ^ = Ml)/|G|)X»ei?[G] gsG and so B n Irr(G) = {/} by Theorem 15.27.
Changing the characteristic 277 Finally assume (c). Then 0 < |IBr(G) n B\ < |Irr(G) n B\ = 1 and (a) follows. Also, if IBr(G) r\ B = {cp}, then cp = b% for some ft e Z by Lemma 15.28. Thus % = dx<pcp = dx<pbf and dx<pb = 1. It follows that dX(p = 1 and the proof is complete. | Note that Theorem 15.29 provides an alternate proof of Theorem 8.17. Also observe that if pjf \ G |, then each p-block contains a unique irreducible character. We use Theorem 15.26 to obtain further connections between the p-blocks of G and Z(F[G]). For each B eBl(G), writefB = XxeBnirriG) ex. This Osima idempotent lies in Z(R[G]) and we let eB = fB* e Z(F[G]). (15.30) theorem We have the following. (a) lB(eB) = 1 and XB.(eB) = 0 for blocks B ± B'. (b) The eB are idempotents and eB eB. = 0 for B =£ B'. (c) eB is an F-linear combination of class sums of p-regular classes. (d) ZeB=l. (e) If kB(z) = 0 for all B e B1(G), then z is nilpotent. (f) The kB are all of the algebra homomorphisms Z(F[G]) -» F. (g) Every idempotent of Z(F[G]) is a sum of some of the eB. Proof (a) For / e Irr(G), we have oix(fB) =1 if / e B and otherwise oix(fB) = 0. If / e B', then a>x* = kB. and (a) follows.- (b) Since/B/B. = dBB,fB we have eBeB, = 5BB,eB. Since X^eB) = 1, we have eB j= 0. (c) Immediate from Osima's Theorem 15.26(c). (d) Z fB = lex =1 and thus 2>B=1* = 1. (e) Let 9) be any irreducible F-representation of G and let ?) afford cp e IBr(G) n B. Then 9)(z) = lB(z)7 = 0 by Theorem 15.19. Thus z is in the Jacobson radical, J(F\_GJ) and hence is nilpotent (Problem 1.4). (f) Let 1: Z(F[G]) -» F be an algebra homomorphism. Then ker k has codimension 1 and so Z(F[G]) = ker 1 + F ■ 1 and k is determined by its kernel. If k j= kB, then ker kB £ ker k and we can choose zB e ker 1B with k(zB) ¥= 0. Assuming 1 £ {1B}, let z = f] zB. Then lB(z) = 0 for all B and hence z is nilpotent by (e). However, l(z) = f] k(zB) =£ 0. This is a contradiction. (g) Let e e Z(F[G]) be an idempotent. Then e = e £ eB = £ eeB and it suffices to show that either eeB = 0 or eeB = eB. Suppose eeB ^ 0. Then since eeB is not nilpotent and kB,(eeB) = 0 for B' j= B, we have lB(eeB) ¥= 0 and thus lB(eeB) = 1 = lB(eB). Thus lB.(eB(l - e)) = 0 for all B' eBl(G) and since eB(l — e) = (eB(l — e))2,wehaveeB(l — e) = 0and the result follows. |
278 Chapter 15 We now discuss some of the connections between block theory and the collection of p-subgroups of G. We work in the situation of Theorem 15.30. If Jf is a conjugacy class of G we consider the Sylow p-subgroups of CG(x) for x e jf. These are called the p-defect groups for jf. They constitute a conjugacy class of p-subgroups of G. The collection of p-defect groups for jf is denoted 8( Jf). We use the notation jf for the sum in F\_G] of the elements of the class Jf so that the £ form a basis for Z(F[G]). If B e B1(G), write eB = Y^a^yf)^ so that aB is a uniquely determined function from the set of classes of G into F. In fact, aB(Jf) = ((1/|G|) X^imonB X(l)X(g))* for # e jf by Theorem 15.26(a). By 15.30(c) we have aB(Jf) = 0 if Jf does not consist of p-regular elements. Since 1 = lB(eB) = £ aB(Jf)lBpf),it follows that for each B e B1(G), there exists at least one class Jf such that aB(Jf) =£ 0 and 1B( jf) ^ 0. We call such a class a defect class for B. (15.31) theorem (Min-Max) Let Jf be a defect class for B e B1(G) and let £> e <5(jf). Let if be any class of G. (a) If aB(Jz?) t6 0, then D contains a defect group for if. (b) If 1B(^) t6 0, then D is contained in a defect group for if. To prove the min-max theorem, we define the Brauer homomorphism f}P. Let P c G be a p-subgroup. Let N = NG(P) and C = CG(P). We map /?P:Z(F[G])^Z(F[iV])by jceX n C and extend by linearity. Since Jf n C is a union of classes of JV, we do have fr(jr)eZ(F[JV]). (15.32) lemma The map 0P: Z(F[G]) -> Z(F[iV]) is an algebra homomorphism. Proof If suffices to check that 0P(i^) = PA30PA&) for classes Jf, if. For ceC, let j/ = {(x, y)\xejf,ye£',xy = c} and •^o = {(*> y)|xe jf nCyeif nC,xy = c}. Then | sf |* is the coefficient of c in /?P(Jf J?) and | j/0 |* is the coefficient of c in Pp{jf)Pp(^). It thus suffices to check that \sf\ = \sf0\ mod p. Since P c C(c), it follows that P acts on d by (x, yf = (xu, y") for us P. Then j/0 is exactly the set of fixed points of s# under the action of P. The result follows. |
Changing the characteristic 279 (15.33) lemma Let P be a p-subgroup of G and let (tP be the corresponding Brauer homomorphism. Let z = Yjax-X". Then /?P(z) j= 0 iff P is contained in some D e 8( Jf) with a^ ,£ 0. Proof The sets jf n C(P) are disjoint for distinct classes Jf and thus the nonzero elements of the form /?p(jf) are linearly independent. It follows that pP(z) ± 0 iff fiP(£) ± 0 for some Jf with ax * 0. Now 0p(jO ^ 0 iff Jf n C(P) # 0 and this happens iff P £ C(x) for some x e jf. Since P c C(x) iff P is contained in some Sylow p-subgroup of C(x), the result follows. | (15.34) lemma Let B e B1(G) and let if be a class of G with aB(S£) ^ 0. Let P e (5(i?) and N = NG(P). Then there exists b e Bl(iV) such that XB = pPXb. WIG]) -> Z(F[JV]) - F. Proof By Lemma 15.33, /?P(eB) ¥= 0. Since /?P is a homomorphism, e = /Sp(eB) is a nonzero idempotent inZ(F[JV]) and so is not nilpotent. Thus there exists b eBl(JV) with Xb(e) # 0. Let fi = pPXb. Then /z(eB) = lfc(e) # 0 and /z is an algebra homomorphism Z(F[G]) -> F. Since fi(eB) ¥= 0, we have H = XB and the proof is complete. | Proof of Theorem 15.31 If aB(<£) # 0, let Pec5(if). Then 1B = 0Plfc for some fceBl(JV(P)) by Lemma 15.34. Since Jf is a defect class, we have 0 ¥= Xg(Jf) = Xb(fiP(Jf)) and so $P(tf) ^ 0. Thus P is contained in some defect group of jf by Lemma 15.33. Now (a) follows. Now apply Lemma 15.34 to Jf. Since aB(Jf) #0we have XB = pDXb for some b e B1(N(£>)). Suppose 1B(J^) # 0. Then 0D(Jzf) # 0 and hence D is contained in a defect group of if. (15.35) definition Let B be a p-block of G. Then the p-defect groups of the defect classes of B are called defect groups of B. The set of these is denoted 8(B). (15.36) corollary Let B e B1(G). Then 8(B) is a single conjugacy class of subgroups. Proof Let Jf t and Jf2 be defect classes for G. It suffices to show that 8(Jti) = 8(Jf2). Let £),-e (5(Jf,). By Theorem 15.31, each of the £>,• contains a conjugate of the other. The result follows. | Which p-subgroups of G can be defect groups for blocks? It is a fact (which we shall not prove) that a defect group for a block is necessarily of the form P r\ Q for some P, Q eSylp(G). We prove the weaker assertion that Op(G) is contained in every defect group of a block.
280 Chapter 15 (15.37) lemma Let P = Op(G). Then Pgker 9) for every irreducible F-representation 9) of G. Proof Since P has a unique p-regular class, the principal representation of P is its unique irreducible F-representation by Corollary 15.11. If 9) is an irreducible F-representation of G, then the restriction 9)P is completely reducible by Corollary 6.6. The result follows from these two facts. | (15.38) theorem Let Jf be a class of G and assume that Jf n C(Op(G)) = 0. Then jf is nilpotent. Proof Let P = Op(G) act on Jf by conjugation and let 6 be an orbit of this action. Then \G\ > 1 and so p||0|. Let xe&. If ye&, then y = xu = x[x, u] for some ueP. Since P o G, we have \_x, u]eP and so (9 £ xP. Now let 9) be an irreducible F-representation of G so that P £ ker 9) by Lemma 15.37 and thus 9) has the constant value 9)(x) on the coset xP. Therefore Y,yse Viy) = WWW = 0 since p||0|. We thus have 9)(jf) = 0 for all irreducible 9) and hence jf e J(F[G]), the Jacobson radical. It follows that jf is nilpotent, | (15.39) corollary Every defect group for a p-block of G contains Op(G). Proof Let B e B1(G) and let Jf be a defect class for B. Then lB(jf) # 0 and hence jf is not nilpotent. Thus Jf n C(Op(G)) ¥= 0 and it follows that Op(G) £ D for every D e 5(jf). I (15.40) corollary Let G be p-solvable with Op.(G) = 1. Then G has a unique p-block. Proof Let B, B'eBl(G). We claim that lB(eB.) = aB-({l}). Since this is independent of B, the result will follow. We have lB(eB) = XaB-(X')^Jt), where the sum runs over classes Jf. Since 1B(1) = 1, it suffices to show that if Jf ^ {1}, then either aB-(jf) = 0 or lB(jf) = 0. Let P = Op(G). If lB(jf) ^ 0, then Jf n C(P) ^0. However, the Hall- Higman "Lemma 1.2.3" yields C(P) £ Pand thus Jf c P. Since Jf 76 {1}, the elements of Jf are not p-regular and thus aB( Jf) = 0. The proof is complete. I If B e B1(G), let D be a defect group for B and write | D | = pd. We call d the defect of B and write d = d(B). (This is well defined by Corollary 15.36.) We show how to compute d(B) from a knowledge of B n Irr(G) or B n IBr(G). We mention that if m, n e Z with pjfn, then m/n e R. Thus if p | m, we have m/n epK Q: Ai. (15.41) theorem Let IGI = p"m with /^m and let B e Bl( G) with d(B) = rf. Then p"~d is the largest power of p which divides all x(l) for / e Irr(G) n B.
Changing the characteristic 281 Proof Let X be a defect class for B and let gsjf. Then |Jf| = | G |/1 C(g)| and so f~d is the p-part of | JtT | since |D \ = pd for D e Sylp(C(g)). Let / e B n Irr(G). We have o * iB(jf) = (c»z)*(jf) = (zte)|jr|/z(i))*. Since x(g)e7? and i{g)\ jf\/x(l)eR - U, we conclude that |jf|//(l)^M and hence the p-part of | Jf | cannot exceed that of x(l). Thus pa~d divides /(l). The coefficient of g in the Osima idempotent/B is given by fl. = U/|G|) Z xWxW) XeBnlrr(G) and thus fl. = d/|G|) Z %{\M9) ,j>EBnIBr(G) by Lemma 15.21 (since g is p-regular). We have 0 ^ a^Jf) = a* and ag$ id. However, p" I <!>,,( 1) by Corollary 15.22 and thus %(l)/\G\eR for all cpeIBr(G). We conclude that cp(g)4Al for some cp e IBr(G) r\ B. Now <p(g) is a Z-linear combination of x(g) for / e Irr(G) n B by Lemma 15.28_and thus xTgit® for some / eIrr(G) r\ B, Now j^)| jf |//(1) = a e/? and xig) = a/(l)/| jf |. It follows that z(l)/| Jf | £j<? and so the p-part of *(1) cannot exceed p"~d. The proof is complete. | (15.42) corollary In the situation of Theorem 15.41, pa~d is the largest power of p which divides all <p(l) for cp e IBr(G) r\ B. Proof In fact {/(l)|x elrr(G) n B} and {cp(\)\cp e IBr(G) n B} have the same greatest common divisor. This follows since each /(l) is a Z-linear combination of the <p(l) using decomposition numbers and each <p(l) is a Z-linear combination of the x(l) by Lemma 15.28. | In connection with Corollary 15.42, we mention that <p(l) need not divide |G|for<peIBr(G). In the situation of Theorem 15.41, if / e B n Irr(G), then the p-part of/(l) can be written in the form p"~d+h, where h > 0. The integer h is called the height of/. Brauer has conjectured that all / e Irr(G) n B have height zero iff a defect group of B is abelian. Note that if d(B) = 1 and / e B has positive height, then p" divides /(l) and thus B n Irr(G) = {/} by Theorem 15.29. This forces d(B) = 0, a contradiction. Thus if d(B) = 1, then all characters in B have height zero. Suppose H c G and b e B1(H). Let X = lb: Z(F[H]) -> F be the corresponding central homomorphism. We construct a linear map 1G:Z(F[G]) -> F
282 Chapter 15 by setting \xeX~nH J It may be that kG is an algebra homomorphism, in which case XG = XB for some unique B e B1(G). When this happens we say that ba is defined and we write ba = B. The block ba is called the induced block. (15.43) lemma Let b e B1(H) for H £ g and suppose bG is defined. Then every defect group for b is contained in a defect group for ba. Proof Let jf be a defect class for bG and let De8(b). We have 0 t6 (lb)G( jf) = lfc(X £\ where if runs over the classes of H contained in Jf. In particular, Jf n H # 0 and there exists Jz? = Jf n H such that lfc(^) ^ 0. By the Min-Max Theorem 15.31, there exists Pec5(if) with D c P. Now P e Sylp(CH(x)) for some x e if £ jf and thus there exists S e Sylp(CG(x)), with S^P. Thus DsSe8(Jf) = 8(bG). | The Brauer homomorphism can be used to give a sufficient condition for induced blocks to be defined. (15.44) lemma Let P s G be a p-subgroup and let C(P) £ H c N(P). Then ftG is denned for all ft eBl(H). If b eBl(H) and B e B1(G), then ftG = B iff 1B = /Jpl;,, where /?P is the Brauer homomorphism. Proof The image of 0P: Z(F[G]) -+ Z(F[N(P)]) actually lies in Z(F[H]). Let 1: Z(F[H]) -> F be an algebra homomorphism and let tx = pPk Z(F[G]) -> Z(F[H]) -> F so that /z is an algebra homomorphism. We claim that y. = XG. Let C = C(P) c H and let Jf be a class of G. Write £* where w = X^^nc*- Then lG(jf) = l(w + t>)and /z(jf) = Mfi^jt)) = l(u). We must therefore show that X(v) = 0. Now v is a sum of elements of the form &, where if is a class of H such that £? r\ C = 0. Since P<Hwe have C(Op(H)) c C and it follows that g is nilpotent by Theorem 15.38. Thus X(&) = 0 and hence l(t>) = 0 and kG = fi as claimed. If 1 = lfc, then 1G = (ipl is an algebra homomorphism and ftG = B is defined, where B is the unique block such that 1B = /?P1. | Brauer's first and second "main theorems" concern induced blocks. (15.45) theorem (First Main) Let D £ G be a p-subgroup and let JV = N(D). Then ft i-> ftG is a bijection of {fteBl(JV)|Z)ec5(ft)} onto {BeBl(G)\De8(B)}.
Changing the characteristic 283 (15.46) lemma Let P be a p-subgroup of G. Let C = C(P) and JV = N(P). Then Jf h-» jf n C is a bijection of the set of classes of G with p-defect group P onto the set of classes of JV with p-defect group P. Proof Suppose P e 8(jT) and x, y e jf n C. Then P e Sylp(C(x)) and P e Sylp(C(y)). Write y = x9. Then P, P9 e Sylp(C(y)) and hence P = P9C for some c e C(y). Then gceN and y = xgc. It follows that if = jf n C is a class of JV. Clearly P e (5(if) and the map Jf i-> jf n C is one-to-one. If if is a class of N with P e (5(if), let jf be the unique class of G which contains ^£. Now if £ C since if n C ^ 0 and C <i JV. If x e if, let Pcje Sylp(CG(x)). If S > P, then P < NS(P) = S n JV £ C„(x). Since S n JV is a p-group, this contradicts P e (5(if). We conclude that P = Se 8( Jf) and jf i-> if. The proof is complete. | Proof of Theorem 15.45 Write a = {b e Bl(JV) | D e 3(b)}. libeSi, then bG is defined by Lemma 15.44. Let bG = BeB1(G). Since De8(b), we have D £ P for some P e 8(B) by Lemma 15.43. We claim that D = P. Let if be a defect class of b and let jf 2 if be a class of G. Then XnC = ^£ and £> e 8(3f) by Lemma 15.46. By Lemma 15.44 we have lB(jf) = kb(pD(£)) = h(&) * 0. By the Min-Max Theorem 15.31, it follows that D contains some defect group of B. Thus Pg £ D £ P for some g e G and hence D = P e 8(B) as desired. Thus block induction maps SS into {B e B1(G)| £> e 8(B)}. Now let B eBl(G) with De8(B). Let jf be a defect class for B. Then aB(jf) ^ 0 and D e 8(Jf). Thus 1B = pDlb for some b e Bl(JV) by Lemma 15.34 and hence bG = B by Lemma 15.44. We must show that ft e Si. Let P e (5(b) so that D £ P by Corollary 15.39 since £> -a JV. By Lemma 15.43, P is contained in some defect group for B and we have D £ P £ D9 for some g e G. ThusD = Pe«5(fc)andfce;#. Finally, let bu b2e@ with bf = B = b2G. Let if be a class of JV with D 2 P e (5(if). If £> > P, then by the min-max theorem, kbl(&) = 0 = kbl(&). If D = P, then by Lemma 15.46, if = C n jf for some class Jf of G. Then Xbl( <£) = Xbl(Pd#)) = ^bW for i = 1, 2 and hence the lfci agree on all if for classes ^£ with defect group contained in D. By the min-max theorem, it follows that Xb2(ebl) = Xb~(ebl) = 1 and thus b1 = b2. | To state Brauer's "second main theorem" we need to introduce "generalized decomposition numbers." (15.47) lemma Let n e G with o(n) = pe and let C = CG(7t). For / e Irr(G) and (p e IBr(C), there exist unique dnx<p e R r\ Qpe such that X(xn)= X d"x<p(p(x) «>ElBr(C) for all p-regular xeC.
284 Chapter 15 Proof Write Xc = Z ^"A witn a<i>e z and "A e Irr(Q- We have \jiZ(C) = t/Kl)/v for some linear ^ e Irr(Z(C)). Then i/^(x7c) = il/(x)(i^(n) and therefore i^Elrr(C)i(j>ElBr(C) and so we take dnM = ^^imo^fi^nld^. Uniqueness follows from the linear independence of IBr(C). | The algebraic integers d"x<p are called generalized decomposition numbers. Note that if n = 1, then C = G and d\v = dXv. Also, if g e G is arbitrary, we can take n = gp and x = gp. and thus express %(g) in terms of generalized decomposition numbers and Brauer characters. Note that if b is a block of C = C(n) for a p-element n then bG is defined by Lemma 15.44. (15.48) theorem (Second Main) Let /elrr(G) and <peIBr(C), where C = CG(7t) for some p-element n e G. Let / eBeB1(G) and ipebeB1(C). Thenrf^ = OiffcG^B. Proof Omitted. | Theorem 15.48 is extremely powerful and useful. We give a few consequences. (15.49) corollary Let x £ Irr(G) and g e G. Suppose gp is not contained in any defect group for the p-block containing /. Then x(g) = 0. Proof Let n = gp and write g = nx, where x e C(n) is p-regular. Then X(g) = %,d*„cp(x). If d\„ ± 0 then cpebeB1(C(tc)) and x ebG, Let Q ed(b). Then 7t e Op(C(7t)) c Q by Corollary 15.39. Also, Lemma 15.43 yields P e d(bG) with Q1^ P and thus rc e P, a contradiction. Thus all d"x<p = 0 and the result follows. | Note that Corollary 15.49 generalizes Theorem 8.17 since if/ has p-defect zero, then the subgroup 1 is the unique defect group for the p-block containing x- If 9 is a class function of G and B e B1(G), write 9B = XxeirrionB [#> l\l so that 9 = Xbebko 9b. (15.50) theorem Let 9 be a class function of G and let n e G be a p-element. Suppose 9(7uc) = 0 for all p-regular x e C(7t). Then 9B(7uc) = 0 for all such x andallBeBl(G). Proof We have 0=9(tcx)= X [9,Z]Z(^)= X [9,X]^W XElrr(G) ^eIrr(G);veIBr(C)
Problems 285 for all p-regular x e C(n) = C. The linear independence of IBr(C) yields ZxEimG) [8, xW» = 0 for each cp e IBr(C). If cp e IBr(C) n ft with ft e Bl(C), Theorem 15.48 yields d"„ = 0 for x i bG and thus £, e i„(C, n * W x¥% = 0. Nowletj^= \J{b nIBr(C)|fceBl(C),fcG = B}. Then for each cp e j/ we have Yxzb [9, x]</Jv = 0 and thus X e B; ij> e jrf (15.51) corollary (Block Orthogonality) Let g, h e G be such that gp and /ip are not conjugate in G. Then I X(<« = 0 *Elrr(G)nB for every p-block B. Proof Write 9 = Xxeirr(G) X(h)X and let n = gp. If x e C(n) is p-regular, then n = (nx)p is not conjugate to hp in G and hence S(nx) = 0 by the second orthgonality relation. Thus 0 = 9B(g) = ^eB xidixW) bv Theorem 15.50. | Problems (15.1) Let D be the decomposition matrix for G. The matrix DT D = C is the Cartan matrix. (The rows and columns of C are indexed by IBr(G).) For cp, 9eIBr(G), let yvS = (1/|G|) Y.xey <P(x)Wx), where ¥ is the set of p-regular elements of G. Define the matrix r = (y^). Show that T = C"l. Hint Let X0 be the part of the character table corresponding to the p-regular classes and let Y be the Brauer character table. Then X0 = DY, (15.2) (a) Let cp, 9 e IBr(G) lie in different p-blocks. Show that I <p(x)%) = 0. xey (b) Let /, t/' e Irr(G) lie in different p-blocks. Show that Y#ey XM^M = 0. (15.3) Let | G | = p"m with pjfm and let \ji e Irr(G) u IBr(G). Define the class function, S^ by 5^(x) = pV(x) if pXo{x) and S^(x) = 0 if p | o(x). Show that S^ is a generalized character of G. Conclude that det(C) is a power of p, where C is the Cartan matrix as in Problem 15.1. (15.4) (a) Show that {<I>J<p eIBr(G)} is a basis for the space of class functions on G which vanish on G — ¥. (b) If cp, fi e IBr(G), show that (1/1G |) £, E y <D,(x)£M = «5W. (15.5) (a) Show that the product of two Brauer characters is a Brauer character.
286 Chapter 15 (b) If cp, ji e IBr(G), define S on G by S(x) = Ov(x)fi(x) for p-regular x; S(x) = 0 when p\o(x). Show that S is a nonnegative integer linear combination of {Ov | v e IBr(G)}. (c) If q>, /zelBr(G), show that O^O^ is a nonnegative integer combination of {<5V| v e IBr(G)}. (15.6) Let ^ be the collection of classes of G which are defect classes for blocks of defect d. Show that | {B eBl(G)\d(B) = d) \ < \<€\. Hint Let A c Z(F[G]) be the span of the Jf for Jf e<#. Then the restrictions of the algebra homomorphisms 1B to A are linearly independent for those B with defect d. (15.7) Let N = Op.(G). Show that if /, \ji e Irr(G) lie in the same p-block, then Xn and ^Pn have the same irreducible constituents. If n denotes the number of classes of G contained in N, conclude that |B1(G)| > n. (15.8) Let P e Sylp(G). Show that the number of p-blocks of G with defect group P is equal to the number of p-regular classes of N(P) contained in C(P). Hint Use Problems 15.6 and 15.7. (15.9) Let N be a normal p-complement for G and let !T be the set of orbits of the action of G on Irr(iV). For each BeBl(G), let 0(B) denote {SGlrr(AT)|[xN, 9] ¥= 0 for some /eBn Irr(G)}. Show that B^(9(B) is a bijection of B1(G) onto F. (15.10) In the situation and notation of Problem 15.9, show that 3(B) = U»6«(*>syM'G(% Hints If /elrr(G), Selrr(iV), and [%N, 9] ^ 0, then co^K) = co^K), where K is any conjugacy class sum of G for a class contained in N. If Jf is a defect class for B and P e Sylp(/G(9)) for some 9 e (9(B), show that P fixes one of the classes of N contained in Jf and conclude that P is contained in a defect group for B.
Appendix Some character tables In the following tables, each conjugacy class is denoted by the order of its elements. If there are more than one class of elements of a given order, they will be distinguished by subscripts. 1. G = Z4 \G\ = 24 = 23 x 3 Class |C(s)| |Clto)| Xi Xi X3 Xa Xs 1 24 1 1 1 2 3 3 2i 4 6 1 -1 0 1 -1 22 8 3 1 1 2 -1 -1 3 3 8 1 1 -1 0 0 4 4 6 1 -1 0 -1 1 Note Class 2t is the class of transpositions. 287
2. G = SL(2, 3) \G\ = 24 = 23 x 3 Class \Qg)\ |C%)| Xi Xl Xi Xa X5 X6 Xi 1 24 1 1 1 1 3 2 2 2 2 24 1 1 1 1 3 -2 -2 -2 4 4 6 1 1 1 -1 0 0 0 3i 6 4 1 CO CO2 0 -1 -co -co2 32 6 4 1 CO2 01 0 -1 -co2 — CO 6i 6 4 1 CO CO2 0 1 CO CO2 62 6 4 1 CO2 CO 0 1 CO2 CO Irrational Entries co = e2*'13. 3. G = A5 S PSL(2, 5) S SL(2, 4) |G| = 60 = 22 x 3 x 5 Class \C(g)\ \Cl(g)\ Xi Xi Xi Xa Xs 1 60 1 1 4 5 3 3 2 4 15 1 0 1 -1 -1 3 3 20 1 1 -1 0 0 5i 5 12 1 -1 0 <*i a2 52 5 12 1 -1 0 a2 <*i Irrational Entries at = (1 + y/5)/2 = 1 + s + e4 and a2 = (1 1 + e2 + e3, where e = e2*"5.
Some character tables 289 4. G = PSL(2, 7) = GL(3, 2) |G| = 168 = 23 x 3 x 7 Class |C(3)| |Cl(g)| Zi Xi Xi n Xs x<> 1 168 1 1 6 7 8 3 3 2 8 21 1 2 -1 0 -1 -1 4 4 42 1 0 -1 0 1 1 3 3 56 1 0 1 -1 0 0 7i 7 24 1 -1 0 1 a a 72 7 24 1 -1 0 1 a a Irrational Entries a = (— 1 + iy/l)/2 = e + e2 + e4, where e = e2*'n. 5. G = A6 =£ PSL(2, 9) |G| = 360 = 23 x 32 x 5 Class |Cte)i |Clto)| Xi Xi Xi X* Xs X6 Xi 1 360 1 1 5 5 9 10 8 8 2 8 45 1 1 1 1 -2 0 0 4 4 90 1 -1 -1 1 0 0 0 3i 9 40 1 2 -1 0 1 -1 -1 32 9 40 1 -1 2 0 1 -1 -1 5i 5 72 1 0 0 -1 0 «1 a2 52 5 72 1 0 0 -1 0 a2 ai Irrational Entries at = (1 + y/S)/2 = 1 + e + e4 anda2 = (1 — y/S)/2 = 1 + e2 + e3, where e = e2*"5.
290 Appendix 6. G = SL(2, 8) \G\ = 504 = 23 x 32 x 7 Class |C(s)| icito)i X\ Xi X3 X* Xs Xe Xi x% X9 ational *'n. P! 1 504 1 1 8 7 7 7 7 9 9 9 Entries = -(8 2 8 63 1 0 — 1 -1 -1 -1 ai = 3 9 56 1 -1 -2 1 1 1 0 0 0 = e + + n p2 = 9i 9 56 1 -1 1 Pi Pi P3 0 0 0 e6, a2 -(«52 92 9 56 1 -1 1 Pi Ps Pi 0 0 0 93 9 56 1 -1 1 h Pi Pi 0 0 0 7i 7 72 1 1 0 0 0 0 1*1 a2 <*3 = e2 + e5 and a3 + 81), and /?3 = - 72 7 72 1 1 0 0 0 0 a2 <*3 ai = e3 + -(<54 + 73 7 72 1 1 0 0 0 0 <*3 «1 a2 e4, where (55), where 7. G = PSL(2, 11) \G\ = 660 = 22 x 3 x 5 x 11 Class |C(0)| |C%)| Xi Xi Xi X* Xs X6 Xi x% 1 660 1 1 10 10 11 12 12 5 5 2 12 55 1 2 -2 -1 0 0 1 1 3 6 110 1 1 1 -1 0 0 -1 -1 6 6 110 1 -1 1 -1 0 0 1 1 5i 5 132 1 0 0 1 «i a2 0 0 52 5 132 1 0 0 1 a2 «i 0 0 Hi 11 60 1 -1 -1 0 1 1 P P 112 11 60 1 -1 -1 0 1 1 P P Irrational Entries at = (— 1 + ■y/S)/2 = e + e4, a2 = (— 1 — y/5)/2 = e2 + e\ where e = e2*1'5. /? = (-1+ ujli)/! = 8 + 83 + 84 + 85 + 89, where 8 = e2*'111.
Some character tables 291 . G = \G\ Class \C(g)\ icite)i Xi ii Xi n Xs Xf> Xl Xs X9 Xn Mn = 7920 = 1 7920 1 1 10 11 10 10 16 16 44 45 ,: 55 = 24 2 48 165 1 2 3 -2 -2 0 0 4 -3 -1 x 32 x 5 x 4 8 990 1 2 -1 0 0 0 0 0 1 -1 8i 8 990 1 0 -1 a a 0 0 0 -1 1 11 82 8 990 1 0 -1 a a 0 0 0 -1 1 3 18 440 1 1 2 1 1 -2 -2 -1 0 1 6 6 1320 1 -1 0 1 1 0 0 1 0 -1 5 5 1584 1 0 1 0 0 1 1 -1 0 0 Hi 11 720 1 -1 0 -1 -1 P P~ 0 1 0 112 11 720 1 -1 0 -1 -1 P P 0 1 0 Irrational Entries a = ujl. 0 = (-1 + iy/ll)/2 = e + e3 + e4 + e5 + e9, where e= e2"111.
Bibliographic notes There exist several excellent bibliographies in group theory and character theory (for instance in the books by Huppert [26], Dornhoff [15], and Curtis and Reiner [12]) and so, instead of giving a comprehensive list of publications, we shall only mention some of the items which are directly relevant to the various chapters of this book. General Some other books on characters and representations are Dornhoff [15], Feit [16], and Curtis and Reiner [12]. Each of these has a point of view somewhat different from the others and from this book. As a reference on group theory we mention Huppert [26] which also has an extensive chapter on characters. Finally, we come to Burnside [10]. This classic, although somewhat difficult for the modern reader, contains a wealth of material. Chapter 1 Further relevant information on rings and algebras can be found in Curtis and Reiner [12] and Herstein [25]. Chapter 2 Other methods exist for obtaining the basic results about characters such as the orthogonality relations. For instance, instead of using the central idempotents of C[G], Feit [16] and Dornhoff [15] use a matrix approach which results in additional information, namely the Schur relations which appear here as Problem 2.20. Chapter 3 There is, of course, a large literature in algebraic number theory. A reference for those parts of the subject most relevant to group theory is the appropriate chapter in Curtis and Reiner [12]. The proof of 292
Bibliographic notes 293 Theorem 3.12 given here was discovered (independently) by G. Glauberman and the author. For another proof, see (4.2) of [16]. Whitcomb's result on isomorphism of integer group rings occurs in Whitcomb [37]. Chapter 4 The Brauer-Fowler proof [8] of Theorem 4.11(a) does not depend on characters. They obtain a slightly better bound which is of the same order of magnitude. Chapter 5 There is a great deal more to be said about the relationship between permutation groups and representation theory. For instance, see Chapter V of Wielandt [38] and Sections V.20 and V.21 of Huppert [26]. Chapter 6 Clifford's results appear in Ref. [11]. This paper includes significantly more than Theorem 6.5; it also has part of Theorem 6.11 and bears heavily on Chapter 11. The " going down " Theorem 6.18 and its dual Problem 6.12 appears in Isaacs [28], however Corollary 6.19 goes back to Burnside's book [10]. A result on relative M-groups which is more general than Theorem 6.22 occurs in Price [35]. Theorems 6.22 and 6.23 give sufficient conditions for a group to be an M-group which generalize a result of Huppert (Satz V.18.4 of Huppert [26]). The connections between the extendibility of 9 and det(9) are due to Gallagher [19]. A proof of a version of Tate's Theorem 6.31 without characters appears as Satz IV.4.7 of Huppert [26] and Thompson's proof (including Problem 6.20) is found in Ref. [36]. Chapter 7 A proof of Theorem 7.8 for the case that |P| = 8 that does not depend on "modular characters" has recently been discovered by Glauberman [22]. For a proof of Theorem 7.10 without characters, see Bender [2]. A large fraction of the known applications of character theory to "pure" group theory are either directly or indirectly related to the content of this chapter. We mention as examples Sections 28 and 32 of Feit [16]. Chapter 8 Brauer's Theorem 8.4 occurs in Brauer and Tate [9] and in some of Brauer's earlier papers. Banaschewski's Lemma 8.5 appears (in a somewhat more complicated form) in Ref. [1]. Dade's Theorem 8.24 and its consequence Theorem 8.26 appears in more general form in Ref. [13]. Chapter 9 An alternate source for much of this material is Curtis and Reiner [12]. Chapter 10 The more standard version of Theorem 10.7 is due to Brauer and Witt (Theorem 70.28 of [12] or Yamada's notes [41]). Theorem 10.12 appears in Goldschmidt and Isaacs [24]. What amounts to a special case of Theorem 10.16 occurs in Burnside [10] as Exercise 8 on page 319. That every integer can occur as a Schur index was proved by Brauer [5] using groups
294 Bibliographic notes similar to those of Theorem 10.16. A great deal of further information can be found in Yamada [41]. Chapter 11 Much of the theory relating projective representations with the properties of a character triple was originated by Clifford [11]. Our more character theoretic setting of Clifford's work occurs in Isaacs [31]. Berger's Theorem 11.33 appears in more general form in Ref. [3]. Chapter 12 Much of this material is the work of Passman and the author and appears in Isaacs and Passman [33, 34], and Isaacs [29]. Also relevant are Isaacs [32], Berger [4], and Garrison [20]. Chapter 13 This chapter is taken almost entirely from Glauberman's paper [21]. Parts of Theorems 13.6 and 13.14 also appear in Isaacs [27]. A proof of Theorem 13.25 can be found in Isaacs [31]. Chapter 14 The literature on linear groups is very extensive and we mention just a sample. Dixon's book [14] is a good reference. For information on solvable and p-solvable linear groups, see Winter [39, 40] and Isaacs [30]. We also mention the lecture notes by Feit and Sibley [18] for results without solvability hypotheses. Chapter 15 Brauer's papers [6] and [7] are good sources for further reading on blocks and Brauer characters. There is also a chapter on the subject in Curtis and Reiner [12]. The material is treated from a different point of view in Part B of Dornhoff [15] and in Feit's notes [17]. Goldschmidt's notes [23] provide a development of the subject along lines similar to those used here.
References 1. B. Banaschewski, On the character rings of finite groups, Canad. J. Math. 15 (1963), 605-612. 2. H. Bender, Finite groups with large subgroups, Illinois J. Math. 18 (1974), 223-228. 3. T. R. Berger, Primitive solvable groups, J. Algebra 33 (1975), 9-21. 4. T. R. Berger, Characters and derived length in groups of odd order, J. Algebra (to be published). 5. R. Brauer, Gruppen linearer Substitutionen II, Math. Z. 31 (1930), 733-747. 6. R. Brauer, Zur Darstellungstheorie der Gruppen endlicher Ordnung I, II, Math. Z. 63 (1956) 406-444; 72 (1959), 25-46. 7. R. Brauer, Some applications of the theory of blocks of characters I-V, J. Algebra 1 (1964), 152-167; 1 (1964), 307-334; 3 (1966), 225-255; 17 (1971), 489-521; 28 (1974), 433-460. 8. R. Brauer and K. A. Fowler, Groups of even order, Ann. Math. (2) 62 (1955), 565-583. 9. R. Brauer and J. Tate, On the characters of finite groups, Ann. Math. (2) 62 (1955), 1-7. 10. W. Burnside, Theory of groups of finite order (2nd ed. 1911), reprinted by Dover, New York, 1955. 11. A.H.Clifford, Representations induced in an invariant subgroup, Ann. Math. (2)38(1937), 533-550. 12. C. W. Curtis and I. Reiner, Representation theory of finite groups and associative algebras, Wiley (Interscience), New York, 1962. 13. E. C. Dade, Isomorphisms of Clifford extensions, Ann. Math. (2) 92 (1970), 375-433. 14. J. D. Dixon, The structure of linear groups, Van Nostrand-Reinhold, Princeton, New Jersey, 1971. 15. L. Dornhoff, Group representation theory, Dekker, New York, 1971. 16. W. Feit, Characters of finite groups, Benjamin, New York, 1967. 17. W. Feit, Representations of finite groups, Mimeographed notes, Yale Univ., 1969. 18. W. Feit and D. A. Sibley, Finite linear groups of relatively small degree, Mimeographed notes, Yale Univ., 1974. 19. P. X. Gallagher, Group characters and normal Hall subgroups, Nagoya Math. J. 21 (1962), 223-230. 29S
296 References 20. S. C. Garrison, On groups with a small number of character degrees, Ph.D. Thesis, Univ. of Wisconsin, Madison, 1973. 21. G. Glauberman, Correspondences of characters for relatively prime operator groups, Canad. J. Math. 20 (1968), 1465-1488. 22. G Glauberman, On groups with a quaternion Sylow 2-subgroup, Illinois J. Math. 18 (1974), 60-65. 23. D. M. Goldschmidt (to be published). 24. D. M. Goldschmidt and I. M. Isaacs, Schur indices in finite groups, J. Algebra 33 (1975), 191-199. 25. I. N. Herstein, Noncommutative rings, Cams monograph 15, Math. Assoc. Amer., 1968. 26. B. Huppert, Endliche Gruppen I, Springer-Verlag, Berlin, 1967. 27. I. M. Isaacs, Extensions of certain linear groups, J. Algebra 4 (1966), 3-12. 28. I. M. Isaacs, Fixed points and characters . .. , Canad. J. Math. 20 (1968), 1315-1320. 29. I. M. Isaacs, Groups having at most three irreducible character degrees, Proc. Amer. Math. Soc. 21 (1969), 185-188. 30. I. M. Isaacs, Complex /(-solvable linear groups, J. Algebra 24 (1973), 513-530. 31. I. M. Isaacs, Characters of solvable and symplectic groups, Amer. J. Math. 95 (1973), 594-635. 32. I. M. Isaacs, Character degrees and derived length of a solvable group, Canad. J. Math. 27 (1975), 146-151. 33. I. M. Isaacs and D. S. Passman, A characterization of groups in terms of the degrees of their characters I, II, Pacific J. Math. 15 (1965), 877-903; 24 (1968), 467-510. 34. I. M. Isaacs and D. S. Passman, Finite groups with small character degrees and large prime divisors II, Pacific J. Math. 29 (1969), 311-324. 35. D. Price, A generalization of M-groups, Ph.D. Thesis, Univ. of Chicago, 1971. 36. J. G. Thompson, Normal/(-complements and irreducible characters, J. Algebra 14 (1970), 129-134. 37. A. Whitcomb, The group ring problem, Ph.D. Thesis, Univ. of Chicago, 1971. 38. H. Wielandt, Finite permutation groups, Academic Press, New York, 1964. 39. D. L. Winter,/^-solvable linear groups of finite order, Trans. Amer. Math. Soc. 157 (1971), 155-160. 40. D. L. Winter, On the structure of certain /(-solvable linear groups II, J. Algebra 33 (1975), 170-190. 41. T. Yamada, The Schur subgroup of the Brauer group, Springer-Verlag, Berlin, 1974.
Index Abelian group, 16, 30, 182-183 extension of characters, 63 Abelian subgroup and character degree, 30, 84, 190, 209-212, 216-217 of linear group, 240, 249 Absolutely irreducible representation, 145- 146 character of, 149-156 Action of group on classes and characters, 93-94, 230-231 number of orbits, 68 /7-group on ^-group, 97 Algebra, 1 Algebra homomorphism, 3 Algebraic integer, 33-40, 44, 135, 265 Alperin, J., 52 Antisymmetric part of tensor square, 50, 56 Artin, E., 72 Augmentation map, 43 Automorphism of field, see Field automorphism, Galois group, Galois conjugate characters Banashewski, B., 128 Berger, T., 191, 206 Blichfeldt, H. F., 240, 247 Block, 271-286 defect of, 280 defect group of, 279 number of with maximal defect, 286 numbers of characters in, 276 Block orthogonality, 273, 285 Bottom constituent, 146-147 Brauer character and character over F, 264 definition, 263 linear independence, 265 from ordinary character, 266-267 Brauer-Fowler theorem, 54-55 Brauer graph, 272, 275 Brauer homomorphism, 278-279, 282 Brauer, R., 46, 49, 122 characterization of characters, 127 first main theorem, 282 second main theorem, 284 theorem on actions on classes and characters, 93 theorem on groups of order p"qbr, 274 theorem on induced characters, 127 theorem on splitting fields, 161 theorem yielding proper subgroups, 70 Brauer-Speiser theorem, 171 Brauer-Suzuki theorem, on coherence, 112 giving normal ^-complement, 137 on quaternion Sylow subgroups, 102 297
298 Index Brauer-Suzuki-Wall theorem, 76 Brauer-Witt theorem, 162 Broline, D., 208, 211 Burnside-Brauer theorem, 49 Burnside, W., 36, 40, 46, 213 p'q" theorem, 37 transfer theorem, 138 C Canonical extension, 220 Cartan matrix, 285 Center of character, 26-28, 75 of group, 27, 29, 63 of group algebra, 9, 15, 45-46, 277, see also Homomorphism, Multiplication constants Central extension, 179-182, 184-186, 195 Centralizer ring of module, 3-4, 8 of representation, 26, 145, 157-158 Character, 14, see also Irreducible character afforded over subfleld of C, 22, 29-30, 58, 74, see also Schur index and field automorphism, see Galois group Character correspondence, see Correspondence of characters Character degree, see Degree of character, Set of irreducible character degrees Character table, 17, 21, 45, 48, 287-291 construction for As, 64-65 of factor group, 24 fields generated by rows and columns, 96-97 information from, 22-25, 27, 36, 45, 134-136, 141-142 and isomorphism of groups, 24 row sum, 75 Character triple, 186-189, 195-197, 237 Character value, 20, 35, 44,46, 122 constant on G - {1}, 44 equals zero, 28-29, 36, 40, 133, 246, 284 equals zero off subgroup, 28, 31, 95, see also Vanishing-off subgroup exceptional character, 113, 123 rational, 31 real, 31 symmetric groups, 17-18,35 Characterization of characters, 127 Chief factor, 85, 87, 96 Class function, 16, 126 of direct product, 59 of form x"", 50, 60 induced, 62 Clifford, A. H., 79-80 Coboundary, 178 Cocycle, 178 Coherence, 107-125 Collineation group, 39 Commutator, 45 Commutator subgroup, 25, 63, 75, 96, 204-205 Completely reducible module, 4-7, 12, 80, 146 Completely reducible representation, 11, 154, 157-158 Complex linear group, 240-261, see also Faithful character /7-solvable, 245, 256, 260-261 solvable, 243-244, 260-261 Composition series, 146 Conjugacy class in factor group, 24 number of, 16, 46 number of/7-regular, 268 prime power size, 37 in simple group, 37 in symmetric group, 17 Conjugacy class sum in C[G], see Center of group algebra, Homomorphism, Multiplication constants in Z[G], 41-44 Conjugate class function, 78-79 module and representation, 79-80 Constituent of character, 17 of representation, 146-147 Control of transfer, 92 Correspondence of characters and coprime action, 219-239 and inertia group, 82 and products, 84-85 Cyclic center and faithful character, 29, 32 Cyclic subgroups and induced characters, 72,76 Cyclotomic polynomial, 72
Index 299 D Dade, E. C, 97, 138, 142 Decomposition matrix, 267, 285 Decomposition number, 267 Defect class for block, 278 Defect group of a block, 279 of a class, 278 Defect of a block, 280 Defect zero character, 134, 143, 276, 284 Degree of character, 14, 37-38, see also Set of irreducible character degrees and abelian subgroup, 30, 84, 190 and center, 28, 31,38, 44 and derived series, 67 in group of order 27, 25 in M-group, 67 in nilpotent group, 28 in/7-block, 274, 280-281 prime power, 39, 44, 274 in £3, 16 when/>2,f|G|, 122 Degree of projective representation, 174 Degree of representation, 9 Derived series, 67, 202, 206 Determinant of character, 29, 59, 75, 88 and extendibility, 88-90, 132-133 Determinantal order, 88-90, 133, 198 Difference of characters, 50, 60, 99, see also Generalized character Dihedral group, 30, 53, 105 Direct product, 59-60, 172 Dirichlet, P. G. L., 169 Divisible abelian group, 182 Division algebra, 4, 149 maximal subfleld of, 170 Dornhoff, L., 96 Double centralizer theorem, 8 Double transitivity, 69, 76-77, 111 E Elementary group, 127 End(K), 1 Equivalent factor sets and projective representations, 178 Even order group, 54-55 Exceptional character, 107, 113, 122-125 Exponent of group, 151, 161, 167 Extendible character, 63, 84-86, 88-91, 97-98, 132-133, 140, 178, 186, 190-191 F F-triple, 162-163 Factor group centralizer in, 26 characters of, 24, 51, 76, 84-85 representation of, 14 Factor set, 174, 176-180, 183, 194-195 Faithful character, 28-30, 32, 77, 83, 123, see also Complex linear group of direct product, 60 of /7-power degree, 39, 44 Fein, B., 170-172 Feit-Sibley theorem, 120 Feit-Thompson theorem on abelian T.I. subgroup, 246 on odd order groups, 111, 231 Feit-Thompson-Blichfeldt theorem, 247 Feit, W., 52, 108, 123, 125 Field automorphism, 29, 151, see also Galois group, Galois conjugate characters Field extension and representations, 144-159 and Schur index, 161, 170 Field generated by character values, 151 and corresponding representation, 150, 153, 159 and order of /7-group, 254 Field restriction and modules, 152-154, 156-157 Fitting subgroup, 208-209, 211, 217 Focal subgroup theorem, 142 Fong, P., 168 Frattini subgroup, 60, 173 Frobenius complement, 99, 107, 121 Frobenius group, 99-101, 121, 200, 237 representations of, 94, 269-270 Frobenius kernel, 100-101, 111, 114 Frobenius reciprocity, 62 Frobenius-Schur theorem, 50-52, 58 Frobenius theorem on existence of kernel, 100-101 on linear groups, 251 Fully ramified character, 95-96, 196, 237, 239
300 Index G Gallagher, P. X., 85, 132, 191, 196 Galois conjugate characters, 152, 154, 161-162, 173 Galois group, 31, 36, 46, 90-91, 165, 221, see also Field automorphism Garrison, S., 59, 206, 208-209 General linear group, 13 Generalized character, 102, 107, 141, 143, see also Difference of characters over a ring, 126, 135 Generalized decomposition numbers, 283-284 Generalized orthogonality relations, 19 Generalized quaternion group, 53, 102 GL(n,F), 13, 174 Glauberman, G.,41, 219 Glauberman map, 219, 229 Glauberman's lemma, 223 Going down theorem, 85 Goldschmidt-Isaacs theorem, 167 Group algebra, 2 Group ring over Z, 41, 43 H Hall-Higman lemma, 256 Hall subgroup, 111,240 character degrees of, 206 and extendibility of characters, 90, 133 generalized character of, 143 and M-group, 96 normal complement for, 136-137 Height of a character, 281 Higman, D. G., 12 Higman, G., 136 Homogeneous character, 83 Homogeneous part of module, 6 Homomorphism of algebras, 3 of center of group algebra, 35-36, 59, 271-272, 277 of modules, 3 of tensor product, 61 I IBr(G), 264 Ideal, 2, 75, 127, 135, 224 Idempotents, 16, 19, 274-277 Imprimitivity decomposition, 65-66 Induced block, 282 Induced character, 63-77 center of, 75 and cyclic subgroups, 72, 76 explicit computation, 64 from inertia group, 82 kernel of, 67 module interpretation, 65 and right ideal of group algebra, 75 Induced class function, 62, 73-74 Induced module, 66, 74 Induction and restriction, 62 Inertia group, 82, 95 Inner product, 20-21 Integral group basis, 41-44 Intertwining matrix, 11 Invariant character, 84-85,91, 138-139, 175, 178, 233-236, see also Extendible character, Character triple Involution, centralizer in simple group, 54, 105 as fixed point free automorphism, 114 number of, 51-54 Irr(G), 15 Irreducible Brauer character, 264 number of, 268 Irreducible character, 15 over arbitrary field, 149-156 number of, 16, 232 Irreducible constituent, see Constituent Irreducible linear group, 240 Irreducible module, 4 Irreducible projective representation, 177 Irreducible representation, 11 of cyclic group, 159 of group, 13 structure over arbitrary field, 154 Isometry, 107-110, 113, 115, 121 Isomorphism of character triples, 187-189, 192, 196 Isomorphism of integer group rings, 41 Ito, N., 84, 190,216,244 J Jacobson radical, 11-12, 97, 157 Jordan-H61der theorem, 146 Jordan's theorem, 249
Index 301 K Kernel of character, 23-24, 75, 77, 81, 208-209, 211 of induced character, 67 Krull-Schmidt theorem, 156 L Lifting a projective representation, 181-182 Linear character, 14 and abelian group, 16, 30 action of group of, 95 and commutator subgroup, 25 extendibility, 89 Linear group, see Complex linear group, General linear group Linear isometry, see Isometry Local integers, 265 M M-group, 67, 86-87, 95-96 solvability and derived length, 67 sufficient condition for, 83, 87 McKay, J., 232 Mackey, G. W., 74 Maschke's theorem, 4 Mathieu group M22, 70 Maximal subgroup, 75, 191 Min-max theorem, 278 Module, 3 and corresponding representation, 10 finitely generated over Z, 34 and normal subgroup, 79-80 Monomial character, 67, 74, 86, 96 Multiplication constants, 2 of center of group algebra, 15,45 N Nakayama's lemma, 265 Nilpotent group, 27, 83, 111, 171,212 character degree in, 28 Nilpotent ideal, 11 Norm map, 165 Normal matrix, 57, 249 Normal subgroup, 78-98 from character table, 23 in linear group, 243-245, 249, 260 prime index, 86 restriction of character to, 79 Normals-complement, 137-138 O Odd order group, 46, 231 Order of character, see Detertninantal order Orthogonality relations, 18-22, 273, 285 Orthonormality of Irr(G), 21 Osima idempotent, 277 Osima, M., 275 P /7-block, see Block /7-defect group, see Defect group /7-defect zero, 134, 143, 276, 284 /7-elementary group, 127, 131, 138,140 /7-group, 29, 60, 91, 96-98, 168, 218 /7-quasi-elementary group, 129, 131 /(-rational character, 90, 98, 191, 240-241, 254, 260 /(-regular element, 263 /7-solvable group, 244-245, 256, 260, 280 Partitioned group, 30 Passman, D. S., 77, 218 Perfect group, 46 Permutation character, 68, 93, 101 necessary conditions for, 69, 75 ring of, 76, 128 Permutation group, 68, 76-77, 101,111 Permutation isomorphism, 230 Permutation module, 68 PGL(n,F), 174 Primitive character, 66, 83, 191 Primitive linear group, 257 Primitive module, 65 Principal block, 274 Principal character, 14 Principal indecomposable character, see Projective character Product of characters, 30, 47-49, 59-61, 84-85, 202, 285-286
302 Index Projective character, 273, 285 Projective general linear group, 174 Projective lifting property, 181-182, 193 Projective representation, 174-197 Quasi-elementary group, 129-130, 142 Quasi-primitive character, 83, 96, 191 Quaternion algebra, 145 Quaternion group, 30, 145, 159, 168, 171, see also Generalized quaternion group Quotient group, see Factor group R /^-generalized character, 126, 135 Radical, see Jacobson radical Ramification, 174 Rank of permutation group, 69 Rational character, 31, 72-73, 76, see also /7-rational character Rational integer, 34 Real character, 46, 50, 54-58, 96 Real element, 31,54, 96 Real representation, 30, 56-58 Regular character, 18 Regular module, 3 Regular representation, 18, 147 Relative M-character, 86, 96 Relative M-group, 86, 96 Representation, 9, see also Irreducible representation and corresponding module, 10 of group, 13 over R, 30, 56-58 Representation group, see Schur representation group Representative set of modules, 6 Restriction, 26 connection with induction, 62 to normal subgroup, 79 Rieffel, M., 8 Roots of unity, 20, 44, 46, 90-91, 220-221, 263 and splitting field, 161, 165-168 Roquette, P., 168 Schur, I., theorem on faithful /7-rational character, 254 theorem on lifting projectives, 181, 184 Schurindex, 160-173 and abelian Sylow subgroup, 165, 173 and direct products, 172 and exponent of group, 161, 167-168 and field extension, 161, 170 Schur multiplier, 181-186, 197 Schur relations, 32 Schur representation group, 182, 185-186, 197 Schur's lemma, 4 Schur-Zassenhaus theorem, 99, 223 Schwarz inequality, 53 Second orthogonality relation, 21 Section of group, 162 Seitz, G., 206 Semidihedral group, 53 Semidirect product, 219 Semiprimitive character, 171 Semisimple algebra, 4, 7-9, 12 Set of irreducible character degrees, 81, 198-218, 243 and abelian subgroup, 209-212, 216-217 cardinality three, 206 cardinality two, 201-205, 216 and derived length, 202, 206 divisible by p, 199 and Fitting length, 209 maximum, 74, 98, 203, 209-214, 216-218 minimum nonlinear, 75 and/7-structure, 213-217 power of/7, 81, 84 prime to/7, 215-216 Sibley, D., 116, 120, 125 Siegel, C. L., 46 Similar projective representations, 177 Similar representations, 9, 144, 147, 157 characters of, 14, 17 Simple algebra, 8 Simple group and centralizer of involution, 54 characters of, 23,44 conjugacy classes of, 37 order pftfr, 274 order 360, 70
Index 303 SL(2, q), 76 Socle, 77 Solomon, L., 75, 129, 168 Solvable group, 24, 95, 142, 199, 209, 212, 216-217, see also M-group character correspondence in, 231 complex linear, 243-244 and extendible characters, 88, 90, 98 quasi-primitive characters of, 96, 191, 194 sufficient conditions for, 37, 67, 206, 243 Special element, 195-196 Splitting field, 146, 148-149, 154, 159, 161 Standard map, 182 Strong isomorphism of character triples, 196 Subnormal subgroup, 190 Supersolvable group, 87, 206 Suzuki group, 125 Suzuki, M., 102, 112, 124 Sylow subgroup, 77, 114, 116, 122, 138, 142 all abelian, 87, 217 and character degrees, 213-215 and extendible character, 89, 190-191 of Galois group, 98, 167-168 in linear group, 39, 44, 243-245, 247, 253, 260 and Schur multiplier, 186 strongly self-centralizing, 61, 76 Sylow tower, 216 Symmetric group, 17-18, 35, 69 Symmetric part of tensor square, 50, 56 T T.l. set, 101-102, 107, 111,246 T.I.F.N. subgroup, 111-125 Taketa, K., 67 Tamely imbedded subgroup, 111 Tate, J., 92, 128 Taussky, O., 53 Tensor power, 60 Tensor product, 47-48, 50, 55, 61, 156 and induced module, 66 Thompson, J. G., 46, 52, 92, 97-98, 105, 111, 199 Top constituent, 146-147 Trace of matrix, 14 Transfer, 63, 92, 136, 138, 142 Transitivity of induction, 73 Transversal, 62 Trivial intersection set, see T.I. set Twisted group algebra, 176-177, 194-195 U Unitary matrix, 57, 249-250 V Values of characters, see Character value Vandermonde determinant, 49 Vanishing-off subgroup, see also character value, 200, 208 W Weak block orthogonality, 273 Wedderburn, J. H. M. theorem on finite division rings, 149 theorem on semisimple algebras, 7 Whitcomb, A., 41 Wielandt, H., 121 Winter, D. L., 243, 257 Y Yamada, T., 171 Z Zassenhausgroup, 111, 123, 125 Zeros of characters, see Character value, Vanishing-off subgroup