For more FREE Organic Chemistry Material visit www.examville.com
.Full description
engleski
this is organic chemistry lab that seeks to identify different functional groups such as amide ether etcFull description
Descripción: 5th Edition of Focus Groups: A practical guide for applied research
Military Review Magazine article, translation of a Russian article, r.e fighting in Chechnya
Full description
Book on Representation Theory from the French mathematician Serre in 1977.Full description
marginalized groupsFull description
5th Edition of Focus Groups: A practical guide for applied researchFull description
Deskripsi lengkap
Group theory, maths, graphs
GIAN-CARLO ROTA, Edilor ENCYCLOPEDIA OF MATHEMATICS AND ITS APPLICATIONS Volume 16
Section: Algebra P. M. Cohn and Roger Lyndon, Section Editors
The Representation Theory of the Symmetric Group GordonJames
AdalbertKerber
Sidney Sussex College Cambridge, Great Britain
University of Bayreuth Bayreuth, Federal Republic of Germany
Foreword by
P. M. Cohn University of London, Bedford Col1ege
Introduction by
G. de B. Robinson University of Toronto
....
••
1981
Addison-Wesley Publishing Company Advanced Book Program Reading, Massachusetts London' Amsterdam- Don Mills. Ontario· Svdnev- Tokyo
Library of Congress Cataloging in Publication Data
James. G. D. (Gordon Douglas). 1945The representation theory of the symmetric group. (Encyclopedia of mathematics and its applications; v. 16. Section. Algebra) Bibliography: p. Includes index. I. Symmetry groups. 2. Representations of groups. I. Kerber. Adalbert. II. Title. III. Series: Encyclopedia of mathematics and its applications; v. 16. IV. Series: Encyclopedia of mathematics and its applications. Section. Algebra. QA171.J34 512'.53 81-12681 ISBN 0-201-13515-9 AACR2
American Mathematical Society (MOS) Subject Classification Scheme (1980): 20C30 Copyrightc 1981 by Addison-Wesley Publishing Company. Inc. Published simultaneously in Canada. All rights reserved. No part of this publication may be reproduced. stored in a retrieval system. or transmitted. in any form or by any means. electronic. mechanical. photocopving. recording. or otherwise. without the prior written permission of the publisher. Addison- Wesley Publishing Company. Inc .. Advanced Book Program. Reading. Massachusetts 01867. U.S.A.
Manufactured in the United States of America
Contents Editor's Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Section Editor's Foreword Introduction by G. de B. Robinson Preface List of Symbols
XIII
xv xvii xxi xxiii
Chapter 1 Symmetric Groups and Their Young Subgroups . . . . . . . . . . . 1 1.1 Symmetric and Alternating Groups 1.2 The Conjugacy Classes of Symmetric and Alternating Groups 1.3 Young Subgroups of S, and Their Double Cosets 1.4 The Diagram Lattice 1.5 Young Subgroups as Horizontal and Vertical Groups of Young Tableaux Exercises
1
8 15 21 29 33
Chapter 2 Ordinary Irreducible Representations and Characters of Symmetric and Alternating Groups . . . . . . . . . . . . . . . . . . .34
2.1 The Ordinary Irreducible Representations of S, 2.2 The Permutation Characters Induced by Young Subgroups 2.3 The Ordinary Irreducible Characters as Z-linear Combinations of Permutation Characters 2.4 A Recursion Formula for the Irreducible Characters 2.5 Ordinary Irreducible Representations and Characters of An 2.6 S, is Characterized by its Character Table 2.7 Cores and Quotients of Partitions 2.8 Young's Rule and the Littlewood-Richardson Rule 2.9 Inner Tensor Products Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Chapter 3 Ordinary Irreducible Matrix Representations of Symmetric Groups 3.1 A Decomposition of the Group Algebra QSn into Minimal Left Ideals 3.2 The Seminormal Basis of QSn ..•••••••............ 3.3 The Representing Matrices va
34 38 45 58 65 72 75 87 95 100 101 101 109 115
viii
Contents
3.4
Chapter 4 4.1 4.2 4.3 4.4
The Orthogonal and the Natural Form of (a 1 Exercises
126 131
Representations of Wreath Products
132
Wreath Products The Conjugacy Classes of GwrS Representations of Wreath Products over Algebraically Closed Fields Special Cases and Properties of Representations of Wreath Products Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
132 138
Ii
•••••••••••••••••••
Chapter 5 Applications to Combinatorics and Representation Theory 5.1 5.2 5.3 5.4 5.5
Chapter 6 6.1 6.2 6.3
Chapter 7 7.1 7.2 7.3
Chapter 8 8.1 8.2 8.3 8.4
The P61ya Theory of Enumeration Symmetrization of Representations Permutrization of Representations Plethysms of Representations Multiply Transitive Groups Exercises
146 155 161
162 163 184 202 218 227 237
Modular Representations . . . . . . . . . . . . . . . . . . . . . . ... 240 The p-block Structure of the Ordinary Irreducibles of SIIand All; Generalized Decomposition Numbers The Dimensions of a p-block; u-numbers; Defect Groups Techniques for Finding Decomposition Matrices Exercises
240 254 265 292
Representation Theory of Sn over an Arbitrary Field . . . . . .294 Specht Modules The Standard Basis of the Specht Module On the Role of Hook Lengths Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
294 301 306 318
Representations of General Linear Groups
319
Weyl Modules The Hyperalgebra Irreducible GL(m. F)-modules over F Further Connections between Specht and Weyl Modules Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
320 327 334 341 346
Contents
IX
Appendix I: Tables
348
I.A I.B I.C I.D
Character Tables 348 Class Multiplication Coefficients 356 Representing Matrices 368 Decompositions of Symmetrizations and Permutrizations 380 I.E Decomposition Numbers . . . . . . . . . . . . . . . . . . . . . . . .. 413 I.F Irreducible Brauer Characters 430 I.G Littlewood-Richardson Coefficients 436 I.H Character Tables of Wreath Products of Symmetric Groups 442 I.I Decompositions of Inner Tensor Products 451 Appendix II: Notes and References
459
II.A II.B II.C II.D Index
459 460 468 468 507
Books and Lecture Notes Comments on the Chapters Suggestions for Further Reading References
Editor's Statement A large body of mathematics consists of facts that can be presented and described much like any other natural phenomenon. These facts, at times explicitly brought out as theorems, at other times concealed within a proof, make up most of the applications of mathematics, and are the most likely to survive change of style and of interest. This ENCYCLOPEDIA will attempt to present the factual body of all mathematics. Clarity of exposition, accessibility to the non-specialist, and a thorough bibliography are required of each author. Volumes will appear in no particular order, but will be organized intosections, each one comprising a recognizable branch of present-day mathematics. Numbers of volumes and sections will be reconsidered as times and needs change. It is hoped that this enterprise will make mathematics more widely used where it is needed, and more accessible in fields in which it can be applied but where it has not yet penetrated because of insufficient information. GIAN-CARLO
xi
ROTA
Foreword The theory of group representation has its roots in the character theory of abelian groups, which was formulated first for cyclic groups in the context of number theory (Gauss, Dirichlet but already implicit in the work of Euler), and later generalized by Frobenius and Stickelberger to any finite abelian groups. For an abelian group all irreducible representations (over C) are of course l-dirnensional and hence are completely described by their characters. The representation theory of finite groups emerged around the turn of the century as the work of Frobenius, Schur, and Burnside. While it applied in principle to any finite group, the symmetric group S" was a simple but important special case; -simple because its characters and irreducible representations could already be found in the rational field, important because every finite group could be embedded in some symmetric group. Moreover, the theory can be applied whenever we have a symmetric group action on a linear space. Perhaps the simplest example is the case of a bilinear formf(x, y). No theory is required to decomposefinto a symmetric part: sex, Y)=f(x, Y)+f(Y, x) and an antisymmetric part: a(x, Y)=f(x, y)
s( x. y, z) = ~ f( ax, ov, oz ). a
a(x, Y, z)= ~ sgnaf( ax, oy, oz }, a
where a ranges over all permutations of x, y, z: No other linear combination of f 's is only multiplied by a scalar factor by the S3-action (such a factor would have to be I or sgn a, because every permutation is a product of transpositions), but we can find pairs of linear combinations spanning a 2-dimensional S3-module, e.g.
+f( x ) +f(
p = f( x, y, z)
y, x , z) - f( z , y, x) - f( z , x , y)
q =f( z . y,
y, z . x ) - f( x. y, z ) - f( x , z , y)
Here p is obtained by 'symmetrizing' x, y and 'anti symmetrizing' x, Xlll
Z,
and q
Foreword
XIV
is obtained by interchanging x, z in p. If (x, y) denotes the transposition of x, y etc., then we have
(x, y)=p, (x, z)q=p,
(x, z )p=q, (y, z )q=q.
(y, z )p= -p-q,
(x, y )q= -p-q,
Thus we obtain the representation
(x,y)--(~
-1 ) -1 '
(X,z)--(~ ~),
(X,y,z)--(~
-1 ) -1 '
(x,z,y)--
-1 ( -1
-1
(y, z)-- ( -1
~ ),
~ ),
which is an irreducible 2-dimensional representation of S3' It was Alfred Young's achievement to find a natural classification of all the irreducible representations of S, in terms of 'Young tableaux', which are essentially the different ways of fully symmetrizing and antisymmetrizing. The n-symbols permuted are arranged in a diagram so that rows are symmetrized and columns anti symmetrized. In the above example we symmetrized x, y and anti symmetrized x, z; this is indicated by the tableau x
y
z Young's derivation via tableaux was even more direct than Frobenius' and Schur's earlier method, using bialternants or S-functions, although these functions are useful in formulating combinations of representations such as plethysm. There have been many accounts of the theory, from various points of view, and often" the original sources have been hard to follow. It is good to have a general treatment, -by two authors who have both made substantial original contributions, -which combines the best of previous accounts, and systematizes and adds much that is new. After a clear exposition of Young's approach (in modern terms) they present an improved version of Specht modules giving a characteristic-free treatment and leading to a practical algorithm for estimating dimensions. The applications to combinatorics include Polya's enumeration theory, and also the less well known work of Redfield, and there is a separate chapter on the connection with representations of the general linear groups. The comprehensive treatment, with helpful suggestions for further .reading, very full references, various tables of characters, as well as the interesting historical introduction by G. de B. Robinson, will all help to make 'James-Kerber' the standard work on the subject. P. M.
COHN
4r#ru
Introduction In this introduction to the work of James and Kerber I should like to survey briefly the story of developments in the representation theory of the symmetric group. Detailed references will not be possible, but it seems worthwhile to glance at the background which has aroused so much interest in recent years. The idea of a group goes back a long way and is inherent in the study of the regular polyhedra by the Greeks. It was Galois who systematically developed the connection with algebraic equations, early in the nineteenth century. Not long after, the geometrical relationship between the lines on a general cubic surface and the bitangents of a plane quartic curve aroused the interest of Hesse and Cayley, with a significant contribution by Schlafli in 1858 [I, Chapter IX].* Jordan in his Traite des Substitutions, 1870 [2], and Klein in his Vorlesungen uber das Ikosaeder, 1884 [3], added new dimensions to Galois's work. The first edition of Burnside's Theory of Groups of Finite Order appeared in 1897, just at the time when Frobenius's papers in the Berliner Sitzungsberichte were changing the whole algebraic approach. With the appearance of Schur's Thesis [4] in 1901, the need for a revision of Burnside's work became apparent. Burnside began his preface to the second edition, which appeared in 1911 [5], with the comment: "Very considerable advance in the theory of groups of finite order has been made since the appearance of the first edition of this book. In particular the theory of groups of linear substitutions has been the subject of numerous and important investigations by several writers .... " His preface concludes with the remark: "lowe my best thanks to the Rev. Alfred Young, M.A., Rector of Birdbrook, Essex, and former Fellow of Clare College, Cambridge, who read the whole of the book as it passed through the Press. His careful criticism has saved me from many errors and his suggestions have been of great help to me." Alfred Young was born in 1873 and graduated from Cambridge in 1895. His first paper, "The irreducible concomitants of any number of binary quartics," appeared in the Proceedings of the London Mathematical Society in 1899. It had been refereed by Burnside, who told him to read the works of Frobenius and Schur; unfortunately Young knew no German, so it was not till after the war that he was able to incorporate their ideas in his important QSA series. 'References will be found at the end of this Introduction.
XVll
XVlll
Introduction
My own contact with Alfred Young began in 1929. I graduated from the University of Toronto in 1927 and was much interested in geometry, owing largely to the presence on our staff of Jacques Chapel on from Paris. I was fortunate in obtaining a small scholarship at St. John's College, Cambridge, where my first supervisor was M. H. A. Newman. Under his guidance I began to read topology. No group theory was taught in Toronto or Cambridge in those days, but its significance in topology fascinated me. Soon this became apparent to Newman, and he arranged for me to be transferred to Alfred Young as a graduate student. Young came in to Cambridge once a week to lecture. He and his wife stayed at the Blue Boar Hotel, just across the street from St. John's, where I would go to visit him. The geometrical aspects of group theory continued to interest me, and I attended Baker's tea party every week. This was where I met Donald Coxeter and several other geometers to whom I refer in the Introduction to Young's Collected Papers, published in Toronto [6], 1977. After earning my Ph.D. in Cambridge in 1931, I returned to the University of Toronto. My work on the symmetric group continued, and with the cooperation of J. S. Frame and Philip Hall, yielded the dimension formulae for the irreducible representations of S, and GL d over the real field. Richard Brauer was on staff in Toronto 1935-1948, and his interest in representation theory was responsible for much of the development which took place in those years and later, while he was at Ann Arbor 1949-1952, and Harvard 1952-1978. In 1958 I was invited to lecture at the Australian universities, and my Representation Theory of the Symmetric Group appeared in 1961. In 1968 my wife and I went to Christchurch, New Zealand, for three months, and it was during this period that I became interested in the application of group theory to physics. W. T. Sharp in Toronto had obtained his Ph.D. with Wigner in Princeton, and I made contact with Wybourne in New Zealand and with Biedenharn in the U.S.. and attended a seminar in Bochum in West Germany in 1969. It was there that I met Adalbert Kerber and many other interesting people. Not long after, I was in touch with Gordon James, who got his Ph.D. in Cambridge with J. G. Thompson. Then when the representation theory gathering was held in Oberwolfach in 1975 I had a chance to talk with many group theorists whose writings I had read, but had never met. Afterwards my wife and I paid a brief visit to the Kerbers in historic Aachen. It was in the autumn of 1975 that Gordon James came to spend a year at the University of Toronto. He and Kerber had begun to work on this book and we had many conversations; Kerber was largely interested in wreath products, while James had begun writing his considerable number of papers on modular theory. A number of errors had appeared in the decomposition matrices at the end of my book [7], and James has done much to improve their construction in this volume.
References
XIX
In April of 1976 Foata organized another gathering of group theorists in Strasbourg. He did a beautiful job, exploiting the charm of the city and its university to bring together a large number of speakers [8] on various aspects and applications of the symmetric group. Having been invited by Professor McConnell, I gave a repeat performance of my Strasbourg talk in Dublin. This was my first visit to Ireland, and it gave me much pleasure to see J. L. Synge, who had been on our staff in Toronto for many years. It was in June 1978 that T. V. Narayana of the University of Alberta in Edmonton arranged a gathering at the University of Waterloo. He had become involved in Young's work, and his volume Lattice Path Combinatorics with Statistical Applications was published in our Exposition Series in 1979. The proceedings of Young Day has just appeared [9] with an introduction by J. S. Frame and a paper generalizing the hook-formulae for °n(2).
The last gathering in Oberwolfach which I attended was in January 1979. This book was well on its way but we all regretted that publication would be so long delayed. Its content contributes much to complete the picture, from the point of view of the representation theory of Sn' but there remains the question raised by Frame's work: -Could there be a degree formula for the irreducible representations of S, or GL d over a finite field? Future research may provide the answer. In conclusion, let me refer to The Theory of Partitions [10] by Andrews, which has appeared in this series. Professor Rota's comment is worth quoting: "Professor Andrews has written the first thorough survey of this many-sided field. The specialist will consult it for the more recondite results, the student will be challenged by many deceptively simple facts, and the applied scientist may locate in it a missing identity to organize his data." When Young's Collected Papers came out, Andrews wrote a most interesting review of his work [11], listing 121 papers based on the original ideas of this remarkable man. The accompanying portrait of Alfred Young was sent to me by Professor Garnir. It has given me much pleasure to work with the authors of this book, and I wish its readers every satisfaction, as well as the best of luck in further developing these ideas.
References 1. G. Miller, H. Blichfeldt, and L. Dickson, Finite Groups, Dover Publications, New York, 1961. 2. C. Jordan, Traite des Substitutions. Paris, 1870. 3. F. Klein, Vorlesungen uber des Ikosaeder, B. G. Teubner, Leipzig, 1884. 4. I. Schur, Inaugural-Dissertation, Berlin, 1901. 5. W. Burnside, Theory of Groups of Finite Order, Cambridge University Press, Cambridge, 1911.
xx
Introduction
6. A. Young. Collected Papers. University of Toronto. 1977. 7. G. de B. Robinson. Representation Theorv of the Symmetric Group. University of Toronto. 1961. R. D. Foata. COn/hi/wtolre et Representatum du Group Svmetrique, Strasbourg 1976. Springer-Verlag (579). 1977. 9. T. V. Narayana. Young-Day Proceedings- Waterloo 1978. Marcel Dekker (57). 191'0. 10. G. E. Andrews. The Theorv of Partitions, Addison Wesley. Advanced Book Program. 1976. 1 I. G. E. Andrews. Bull. Am. Math. Soc .. l. 989-997 (1979). G. DE B. ROBINSON
Preface The purpose of this book is to provide an account of both the ordinary and the modular representation theory of the symmetric groups. The range of applications of this theory is vast, varying from theoretical physics through combinatorics to the study of polynomial identity algebras, and new uses are still being found. So diverse are the questions which arise that we feel justified in hoping the reader might find that some part of our text inspires him to undertake research of his own into one of the many unsolved problems in this elegant branch of mathematics. There are several different ways of approaching symmetric group representations. and while we have tried to illuminate parts of the theory by giving more than one description of it. we have made no effort to cover every view of the subject. The ordinary representation theory of the symmetric groups was first developed by Frobenius, but the greatest contribution to the early material came from Alfred Young. Since Young's main interest lay in quantitative substitutional analysis, it is difficult for a modern mathematican to understand his papers. The reader is referred to the book Substitutional Analysis by D. E. Rutherford for a pleasant account of a great part of Young's work. Both Frobenius's and Young's collected works are now available. We include an account of the group algebra and its idempotents, along the lines pursued by Young, since the symmetric group is one of the very rare cases where many aspects of general representation theory can be described explicitly. This also helps us to motivate the introduction of many combinatorial structures which turn out to be useful. The combinatorics are themselves a fascinating and fruitful basis for further study. and they continue to inspire many research papers. The development of the modular representation theory of symmetric groups was started by T. Nakayama. who derived some p-modular properties of symmetric groups S,/, for n<2p. and stated a conjecture about the p-block structure of symmetric groups SII of arbitrary degree n. This conjecture was proved jointly by R. Brauer and G. de B. Robinson. Robinson and his coworkers developed these methods rapidly to study the decomposition numbers of the symmetric groups. The situation as it was in 1961 is described in Robinson's book. Later. these methods were combined by the present authors with modular results derived from W. Specht's alternative approach to ordinary representation theory. Specht showed how to derive representations by considering XXI
XXII
Preface
submodules of a polynomial ring F [x I' ... , X n j, and this method yielded interesting results without referring to the characteristic of the field F. Using modules isomorphic to those of Specht, we explain another approach to the ordinary representations, at the same time extracting information about the modular theory. There is a more recent approach to a characteristic-free representation theory of Sn starting with a basis of a certain intertwining space which was originally derived for invariant-theoretical purposes by G.-c. Rota. We shall not present this method, as we anticipate that it will be described in another book. The main application which we cover involves the representations of an arbitrary group, using symmetry operators on tensor space, but we also discuss combinatorics, wreath products, and permutation groups. The standard reference for symmetric functions is the book of D. E. Littlewood, but I. G. Macdonald's recent volume gives an up-to-date account of many of the results. Both authors are greatly indebted to Professor Robinson and the University of Toronto for their generous hospitality during several visits, and they wish to record their gratitude to Professor Robinson, without whose continued enthusiasm, encouragement, and advice the book would not have been written. In conclusion, we would like to express our thanks for the important help and criticism we received from so many colleagues and from our students, whom we have taken pleasure in working with and who have given us so much useful advice. We mention in particular H. Boerner, R. W. Carter, M. Clausen, N. Esper, H. K. Farahat, A. Golembiowski, M. Klemm, W. Lehmann, A. O. Morris, J. Neubuser, H. Pahlings, M. H. Peel, F. Sanger, D. Stockhofe, J. Tappe, K.-J. Thiirlings and B. Wagner. G. D. JAMES A. KERBER
List of Symbols
Symbol
Meaning
Definition on p.
The number of i-cycles occurring in the cycle notation of 7T a( 7T) The cycle type (a l( 7T),... , a n( 7T» of 7T The type of (f; 7T) a(f;7T) af(m) The p-residue of the node which m replaces in tf The alternating group on {I, ... , n} An The alternating representation of Sh ASh The k th Bell number Bk c( 7T) The number of pairwise disjoint cyclic factors of 7T The field of complex numbers C The conjugacy class of S, consisting Cn of permutations with cycle partition IX c-: The two conjugacy classes of An arising from splitting class C" of S, CA( 7T) The conjugacy class in An of the permutation 7T The centralizer in An of the permutation 7T CA(7T) C s ( 7T) The conjugacy class in S, of the permutation 7T Cs ( 7T) The centralizer in S, of the permutation 7T char S, The ring of generalized characters of S, Cyc(H) The cycle index of H Cyc(H, D) The generalized cycle index of H with respect to X D P61ya insertion of p in H Cyc(Htp) The depth of IX dn A p-modular decomposition number of IX d~,f3 dr(r, s) The axial distance between rand s in tr diagG* The diagonal subgroup of G* The determinant of a Gram matrix for Sf3 det/3 Df3 Sf3I( Sf3n Sf3.L) for /3p-regular The p-modular decomposition matrix of S, D~.p a i(7T)
XXII!
7 10
139 307 7 17 228 5
12 12 II II II II
39 170 17I
170 99 243 123 134 313 299 244
List of Symbols
XXIV
(#D
r
DO[a] D6 nD ,
ea
(ja
ea I
eAai e'J et,
et
The representation of Gwr H arising naturally from the representation D of G The symmetrization of D by [a] The nth permutrization of D by D, The essentially idempotent element '\ -a~x a of 0 S, arising from the tableau t a The primitive idempotent element (l/Ka)e a of OSn arising from the tableau (a The essentially idempotent element arising from the standard tableau t,et The primitive idempotent element arising from the standard tableau tt A particular element of U et'1T,~
GwrH [G]f{ Gf{
The poly tabloid arising from t The projection of 0 nV onto the homogeneous component of type [a] A certain subgroup of Gwr H The element of the Weyl module corresponding to the tableau T The dimension of the ordinary irreducible representation [a] of S, The number of different ways of reaching [a] by removing rim q-hooks from [a] An element of Gwr H A field W a/ (J-ya n Wa-l-) The v th cycle product of ( /; '1T) The base group of Gwr H The set of all mappings from n into G The wreath product of G by H The exponentiation of G by H The power of G by H
G[H]
~[HwrG]
Ea Ef{ ET
r /qa
Cf: '1T) F Fa gv(f; '1T) G* Gn
A certain element of FS n G(i! GL(m, F) The general linear group Gaussian polynomial Gm.n
h~1
The number of nodes in the hook Hi! A certain "hook length" .
L.p E If"P
The horizontal group of (a The horizontal group of {a The (i, j)-hook of [a] Intertwining number The identity mapping of { 1,2,... , n} The simple two sided ideal of OSn corresponding to the partition a The identity representation of S~ The leg length of the hook Hij tf)n;.oL(n) W(I)18I... 181 W(I) (n times) eyaL(n) The left ideal of OSn generated by {I, 2,... , m + n}
17
4
er
{m+ I, ... , m+n}
Mil
{m} The matrix whose (i, k) entry is
Mf3
The FSIl-module spanned by ,B-tabloids
n na n~
~ n 1k ] n(k) n(k) ~
~o
p(k; m, n) p(n) p'(n) p*(n) P(n)
r», r», p(k)
pm II
us, Ts, [a k ]) {l,2, ... , n}
The set of dissections n" of n A dissection of n One of a set of disjoint subsets of n forming a dissection n~ of n The set of all k-subsets of n The set of all k-tuples over n The set of all injective k-tuples over
The number of proper partitions of kinta ,;;;m parts, each one being ,;;;n The number of proper partitions of n The number of p-regular proper partitions of n The number of p-regular classes of Sn The set of proper partitions of n
225 10 243 243 23
The permutation groups on n1kl,n(k),n(k), respectively, corresponding to the permutation group P on n The representation of S, afforded by 0 nv
228 150
n
List of Symbols
xxvi
Q 'n(g)
IR Rm Rn(g) R~J RSn
st[a]
sP SA
Sn Sn S(k,j)
SA(n) SP(n) fa f
a*
fa,
t[a] {t} [T] u~A
U U UF
va
V(ta) VI 'Va
'\(/ Q9nV wa WT
wa 7L.
Zn
a' [a]
The field of rational numbers The number of nth roots of g The real number field ~m-l
;=1
(i, m)
The set of nth roots of g The part of the rim of [a] associated with Hi) The regular representation of S, The set of standard a-tableaux The Specht module for the partition {3 A Young subgroup corresponding to A The symmetric group on {I, 2, ... , n} The symmetric group on f2 A Stirling number of the second kind The set of self-associated partitions of n The set of split partitions of n A Young tableau The tableau obtained by removing n from the standard tableau t" The ith standard a-tableau The set of a-tableaux The tabloid containing t The equivalence class of T A u-number with respect to p The universal enveloping algebra The matrix of u-numbers The hyperalgebra The vertical group of t" The vertical group of t" The vertical group of t ~"E va(sgn 'IT)'IT ~"E v,(sgn 'IT)'IT V0 ... 0 V (n factors) The tensor corresponding to the first standard a-tableau The tensor corresponding to the tableau T The Weyl module The ring of integers The character table of S, The partition associated with a The Young diagram associated with a; also, the corresponding equivalence class of representations
The matrix representation afforded by Young's natural form of [a] The reduction of Young's natural representation modulo p idGL(vp[a] The q-core of [a] The q-quotient of [a] The cycle partition of the permutation '17 Skew diagram L~= ,Ia;- fJ;1 The representation ([ a]; [fJ]) of Smwr Sn (a ~ m, fJ~ n) The outer tensor product of [a] and [fJ] (a~ m, fJ~ n) (this is an irreducible represen tation of SmX Sn) [a] # [ fJIt Sm+ n ( a ~ m, fJ~ n ) Inner tensor product of [a] and [fJ] ( a ~ n, fJ~ n) (this is in general a reducible representation of Sn) The plethysm of [a] and [fJ], which means (a; fJ)i Smn The defect of the block containing [a] The determinant module The difference product
L{=lf~
The seminormal unit corresponding to tt A element of the seminormal basis of OSn The character of [a] The value of ~a on the conjugacy class of elements with cycle partition f3 n!/r' The number of which Ais a composition or partition The composition obtained from Aby replacing Aiwith Ai - l The exponent of p dividing an integer The natural form of [a] The character of IS a i s, The value of ~a on the conjugacy ciass of elements whose cycle partition is fJ The matrix whose (i, j) entry is ~;;~ The image of Q' under the permutation '17 The permutation such that tt='17i~tZ
The permutation on n[k], n(k>,n(k), respectively, corresponding to the permutation 7T of n The seminormal form of [a 1 The ith partial sum of the parts of a The (i, k)-coefficient function arising from a" Bilinear form on MP The natural bilinear form on LIJ) A bilinear form on L" The class function whose value at g is XD(g") ~
sgn7T~?--id-(I'-id)01T
1TES N The orthogonal form of [a 1 A set Lexicographic order for partitions Total order on the set of standard a-tableaux Total order on the set of tabloids Dominance order for partitions Is a neighbor of Is isomorphic to Has the same irreducible constituents as Is similar to The identity mapping of {I, 2, ... , n} The cardinality of a set The group generated by The F-vector space spanned by The F-vector space whose basis is Is a (proper) partition of Is a type of Is an improper partition of The result of inducing a representation The restriction of a representation Outer tensor product of representations
228 114 26 114 297
330 331 204 49 127 23 107
303 23
24 244 2 4
9
11 15
The Representation Theory of the Symmetric Group
CHAPTER 1
Symmetric Groups and Their Young Subgroups
In this chapter we introduce much of the notation we shall use later on, and prove some basic results on symmetric groups. Many of the ideas concern partitions a of nonnegative integers n, and the corresponding Young subgroups Sa of the symmetric group Sn' Combinatorial structures such as Young diagrams, the diagram lattice, and Young tableaux are related to partitions, and they will help us in the next chapter to construct the ordinary irreducible representations of Sn'
1.1 Let
Symmetric and Alternating Groups
n denote a set.
A bijective mapping 7T of n onto itself, for short
having the property that {w IwEn and 7T( w) =I=- w} is finite, is called a permutation of n. The order In I of n is called the degree of the permutation 7T. If both 7T and p are permutations of n, then their composition is denoted by 7Tp and defined by
It is again bijective, keeps almost all the points fixed, and therefore is also a permutation. If a set P of permutations of n forms a group under this composition, we call P a permutation group and say that P is acting on n. InI is then called the degree of P. ENCYCLOPEDIA OF MATHEMATICS and Its Applications, Gian-Carlo Rota (ed.). Vol. 16: G. D. James and A. Kerber, The Representation Theory of the Symmetric Group ISBN 0-201-13515-9 Copyright 1981 by Addison- Wesley Publishing Company, Inc., Advanced Book Program. All rights reserved. No part of this publication may be reproduced. stored in a retrieval system, or transmitted, in any form or by any means. electronic, mechanical. photocopying, recording. or otherwise. without the prior permission of the publisher.
2
Symmetric Groups and Their Young Subgroups 1.1
The set of all the permutations of Q, i.e. the set 1.1.1
S!1 : = { 17 117: Q >----l7 Q and 17( W
h'" w for finitely many wE [2}
is a permutation group, it is called the symmetric group on Q. The elements of Q are called points. If [2' C Q, we denote by image under 17 E S!1:
17[Q']
its
When defining permutation groups, the nature of the points on which they act is irrelevant in a sense to be described next. Two permutation groups, say P on Q and P' on Q', i.e. subgroups of S!1 and S!1' (for short: P';;;;;'S!1' P'';;;;;'S!1') are called similar if and only if there exists a bijection f: [2 >---7+ Q' and an isomorphism >: pc::::. P' such that the following holds: 1.1.2
(This means that by renaming the elements of P by > and the points of [2 by f, we obtain P'.) If 1.1.2 holds, then we write p=.p'.
It is easy to check that two symmetric groups are similar if and only if they are of the same degree: 1.1.3
Hence it is only the degree that really matters. This shows that if n: = IQI EN 0: = {O, 1,2,3, ... }, we may assume (up to similarity) that Q =n, where n:={l .... ,n}.
1.1.4
(In particular, 0= 0, the empty set.) The symmetric group on n is denoted by Sn: 1.1.5 So consists of one element only (as does Sl)' An easy induction shows that
the following is true: 1.1.6
A permutation
17 E
Sn is written down in full detail by putting the images
1.1 Symmetric and Alternating Groups
3
'77'(i) in a row under the points i E n. say '77'=( '77'll)
'77'tn) ).
This will sometimes be abbreviated by
'77'=( '77'(i)). Hence. for example,
~
2 2
~),U
2 I
~ ).(j
2 2
i).
(~
2 3
~), (i
2 3
i).(;
2 I
~) }.
S3 ={(
The points I, ... , n which form the first row need not be written in their natural order; e.g.
(~
~) =( ~
2 3
I 2
With this in mind, we call a permutation of the form 1.1.7 cyclic or a cycle. In order to emphasize r, we also call 1.1.7 an r-cycle. A shorter notation for the cycle 1.1.7 is
1.1.8 where the points which are cyclically permuted are put together in round brackets. For example
(~
2 3
;)=(2,3)(1).
Commas which separate the points may be omitted if no confusion can arise (e.g. if n";;; 10). and I-cycles can be left out, if it is clear which n is meant. Hence we may write just 1.1.9
(i 1 •• ·i r )
Symmetric Groups and Their Young Subgroups 1.1
4
instead of 1.1.7 or 1.1.8. This cycle can also be expressed as 1.1.10 The identity mapping id n , which consists of I-cycles only, will be denoted by I or Is,,:
n) .
I ... Is n : = ( I . . . n =ld n ·
Thus e.g.
S3 = { I , (1
2) , ( I
3) , ( I
3) , (2
2
3) , (1
3
2) }.
The notation 1.1.9 for the cyclic permutation 1.1.7 is not uniquely determined, for the following is true: ( i 1 ···i r)=(i 2 ···i ri)= "'=(i ri 1 .. ·i r- I ) . I
1.1.11
This means that a cycle which arises from 1.1.9 by cyclically permuting the points, describes the same permutation as does 1.1.9. 2-cycles, i.e. permutations which just interchange two points, are called transpositions. S3 for example contains the transpositions (12), (13), and (23). The order of a cycle (i, ... i r)' i.e. the order of the cyclic subgroup
«i
1 ••
·i r
)
generated by the cycle, is equal to its length: 1.1.12
1(0," ·ir)I=r.
The inverse of a cycle is easily obtained by reversing the sequence of the points: 1.1.13
Two cycles 77 = (i 1 •.• i r ) and p = (JI •• "is) are called disjoint if the two sets of points which are not left fixed by 77 and p are disjoint sets. Disjoint cycles 77 and p are commuting permutations, i.e. 77p = p77. Each permutation can be written as a product of pairwise disjoint cycles, e.g.
(~
2 5
3 2
4 6
5 3
6 4
7 I
~) =(1
7)(2 5 3)(6 4)(8).
The set of disjoint cycles is uniquely determined by the given permutation.
5
1. I Symmetric and Alternating Groups
For a general 1T E Sn let c( 1T) be the number of disjoint cyclic factors including I-cycles, let I v (1 ,,;;; v,,;;; c( 1T» be their lengths, and choose for each v an elementiv of the v-th cyclic factor. Then c(or )
1.1.14
1T=
II
Uv1T(jJ·· '1T 1,-I(jJ).
v=l
This notation becomes uniquely determined if we choose the iv so that the following holds:
1.1.15
(i)\fI,,;;;v";;;c(1T). sEll (ii)\fI,,;;;v
(}v";;;77 s (jv))'
(jv
If this is true for 1.1.14. then 1.1.14 is called the cycle notation for 1T. So much for notation. We now consider subsets which generate Sn' Because of 1.1.16
each cycle, and therefore each 1TESn , too, can be written as a product of transpositions. Hence Sn is generated by its subset of transpositions (if no transpositions occur in Sn' then n,,;;;I, but So and S, are generated by 0). But except for n,,;;;2, we do not need every transposition. For if I";;;j
(j, k+ l)=(k, k+ I)(j, k )(k, k+ 1), so that (j, k + 1) can be obtained from (j, k) with the aid of the transposition (k, k+ 1) of successive points. This shows that S" is generated by the transpositions (k, k + 1) of successive points, 1,,;;; k < n. Another system of generators of S" is {(12), (1 ... n )}. This is true because, for 0,,;;; r";;; n - 2, we have
(1 .. ·n )'(12)(1' . 'n )-r = (r+ 1, r+2). Hence we have proved the following: 1.1.17 LEMMA.
Sn = «( 12), (23), ... , (n - 1, n )
=«(12),(1·· ·n). We would now like to introduce the sign of a permutation. In order to do this we define for n EN: = { 1,2, ... } the difference product ~" by
6
Symmetric Groups and Their Young Subgroup.s 1. 1
and for
n~2
we put
~n: = and define an action of
Since
7T
7T
II
(j-i) EZ.
I~i
E Sn on ± ~ n by
E Sn is a bijection of n, we have for n ~ 2
Defining the sign of
7T
by
1.1.18
if
n~2,
and putting sgn 1So : = sgn 1s, : = 1z ,
we have for each permutation
7T
sgn 7T E { 1, - 1}. Since for each
7T,
P E Sn we have
the symmetric group Sn acts on Q: = {~n' - ~ n} in such a way that for each wEQ we have lw=w and 7T(pW)=(7Tp)W. This shows that the action defines a permutation representation of Sn on Q, i.e. a homomorphism Sn -+ SQ' Since 7T~ n = - ~ n when 7T is a transposition, the image of this permutation representation is SQ if and only if n ~ 2, while it is {l su} if and only if n < 2. This proves the following lemma: 1.1.19 LEMMA. The mapping sgn: 7TI-...sgn 7T is a homomorphism of Sn into the multiplicative group {I, -I}. It is surjective if and only if n~2.
The representation of Sn afforded by this homomorphism is called the alternating representation. The kernel of the homomorphism sgn is denoted by An: 1.1.20
For example A 3 = (l,(1
2
3),(1
3 2)}.
1.1 Symmetric and Alternating Groups
7
This subgroup An of Sn is called the alternating group on n. The permutations ?TESn , which are elements of An' are called even permutations, while the elements of Sn\A n are called odd permutations. The homomorphism theorem yields 1.1.21
(i) IAol=IA11=1, (ii) \fn~2(IAnl =n !/2).
F or cycles (i 1 ... i r) we have because of 1.1.16
More generally, ?T E Sn is even if and only if the number of cyclic factors of ?T which are of even length is even. Thus, if ?T E Sn' we have 1.1.23. ?TEA n
= n-c(?T) is even.
For if we denote by ai(?T) the number of i-cycles occurring in the cyclenotation of ?T, then n = ~ Ja i(?T), while c( ?T) = ~;a;( ?T), so that
n-c('IT)= L(i-I)a;(?T)
=L(j-l)aj(?T)
(modulo 2)
j
(modulo 2)
if the last two sums are taken over those j where j is even. This yields 1.1.23 by an application of 1.1.22. We can rephrase this as follows: If ?TESn , then 1.1.24
Before we leave the alternating group in order to consider the conjugacy classes of Sn' we should make the following remark: 1.1.25. An is the commutator subgroup of Sn' Even more: each element of An is itself a commutator.
A commutator in Sn' is an element of the form ?Tp?T - I p -I. Hence, as sgn?T = sgn?T - I and sgn p = sgn p - I each commutator is contained in An' Therefore the commutator subgroup S~ of Sn' which is generated by all these elements, is also contained in An: S~ ,,;;A n. In order to prove An ";;S~ we
8
Symmetric Groups and Their Young Subgroups 1.2
verify the second half of the statement 1.1.25. To this end we note that for each i such that 2i + I,,;;; n we have (1, ... ,2i+ 1)= (1, ... , i+ I)(i+ I ..... 2i+ 1)
so that (l, ... ,2i+ 1) is of the form pap-Ia-I where p: =(1. ... , i+ 1) and a is a suitable element of S2i+l (apply 1.2.1). Also. for i";;;j
(l, .... 2i )(2i+ I, .... 2i+2j)= (1, ... , i+j+ I)(2i. i+j+ I, ... ,2i+2j), so this element is a commutator in S2i+2j' Similar remarks hold for arbitrary cycles of odd length and for each pair of disjoint cycles of even lengths. Thus each even permutation, since it consists of cycles with odd lengths together with an even number of cycles with even lengths (so that we can pair them off), is of form pap - la - 1 and is therefore a commutator. This completes the proof of 1.1.25. Since for any finite group G and its commutator subgroup G', the commutator factor G/ G' is isomorphic to the group of one-dimensional characters of Gover C, we get as an immediate corollary of 1.1.25: 1.1.26. The only homomorphisms of Sn into the multiplicative group of Care and '7T......sgn'7T.
'7T...... l c
1.2
The Conjugacy Classes of Symmetric and Alternating Groups
We shall describe the conjugacy classes of SII' In order to do this, we first of all note how p'7Tp - I is obtained from '7T. Since
1.2.1
P'7Tp
-1_ (
-
p(i)) ,
i )( i )(P(i))_ ( P( i ) '7T ( i ) i p'7T ( i )
we get p'7Tp - I from '7T by an application of p to the points in the cyclic factors of the same '7T. For if
then by 1.2.1 we have
We notice that under this process of applying p to the points, the brackets remain invariant, so that the cyclic factors of p'7Tp -I (in cycle notation) are of the same lengths as those of '7T.
12 Conjugacy Classes of Symmetric and Alternating Groups
9
On the other hand, let 7T and a be permutations which are both products l:S:;v:S:;c(7T), say of C(7T) cyclic factors of the same lengths
This shows that two permutations are conjugate if and only if they have the same cycle structure. In order to make this more precise, we introduce the notion of a partition of n. A sequence of nonnegative integers
is called a (proper) partition of n if and only if it satisfies
1.2.2
Y;
(ii) ~
(X;=n.
;=1
The (Xi are called the parts of abbreviated by
(x.
The fact that
(X
IS
a partition of
n
IS
(X~n.
(X~n, then by 1.2.2 (ii), there is an h such that take the liberty of shortening (X as follows:
If
(Xi
=0 for all i>h. We may
10
Symmetric Groups and Their YDung Subgroups 1,2
(normally, we choose h such that ah>O, ah+1 =0). We list below the partitions of the first few nonnegative integers, using this convention: n=O
The number p( n) of partitions of n grows rapidly with n. E.g. p(O)= 1, p(l)= 1, p(4)=5, p(1O)=42, p(20)=627, p(50) =204226, p(lOO) = 190569292.
A table for p(n) (n,.;;;;IOO) can be found in the book of G. E. Andrews
[1976]. The following notation is useful in the case when several nonzero parts of a are equal, say a i parts are equal to i, l,.;;;;i";;;;n:
If a i =0, then i a , is usually left out. For example
(3,2,1 2 )=(3,2,1,1,0,0, ... ). If now 'fT is an element of Sn' then the ordered lengths a;( 'fT), 1";;;;i";;;;c( 'fT), of the cyclic factors of 'fT in cycle notation form a uniquely determined partition of n, which we call the cycle partition of 'fT, and which we denote by a( 'fT):
1.2.3 The corresponding n-tuple consisting of the multiplicities of parts of a( 'fT), i.e.
1.2.4 is called the cycle type of 'fT. Correspondingly we call a : = (a I' ... , an) a type of n if and only if
1.2.5
(i)
'""I";;;;/''';;;;n v
(ii)
~ia, =n.
(a ; E'""''''0' )
1.2 Conjugacy Classes of Symmetric and Alternating Groups
II
This will be abbreviated by aHn.
At the beginning of this section we showed that two elements 7T and p of Sn are conjugate if and only if they have the same cycle structure. Using the notion of cycle partition and cycle type and denoting by
the conjugacy class of 7T E Sn' this reads as follows:
From we obtain 1.2.7 for each 7TESn • This together with 1.2.6 shows that every permutation in Sn is conjugate to its inverse. Groups in which each element is a conjugate of its inverse are called ambivalent, so that we have proved the following 1.2.8 LEMMA. Sn is ambivalent.
This is not true for the alternating groups. A 0' A I' and A 2 are ambivalent, but A 3 is not ambivalent. There are, in fact, only a very few alternating groups which are ambivalent. We would like to determine these ambivalent alternating groups, for ambivalent groups have a real character table. This problem leads us to a description of the conjugacy classes of An' Since An is a normal subgroup of Sn' a conjugacy class C S of Sn is either contained in An' or contained in Sn\A n. 1.1.22 helps us to determine whether C S is contained in An or not, and hence it remains to check whether the Sn-class C S C;A n is also an An-class or splits into several A n -classes. Now the order of a conjugacy class C S ( 7T) is equal to the index of the centralizer Cs ( 7T) of 7T in Sn' Since for the centralizer CA( 7T) of 7T in An we have 1.2.9
which is a subgroup of index";;; 2 in Cs ( 7T), either C s ( 7T) c; A n is an A n -class itself, i.e.
Symmetric Groups and Their Young Subgroups 1.2
12
or C s ( 7T) splits into exactly two An -classes of the same order 1C s ( 7T )1/2. Furthermore C s( 7T) splits (for n> 1) if and only if CA( 7T) = Cs ( 7T), i.e. if and only if Cs ( 7T) contains no odd permutations. Cs ( 7T) contains odd permutations if 7T contains a cyclic factor of even length (which is then an odd element of e 7T» or if 7T contains two cyclic factors of the same odd length, say
s(
so that
Conversely. if 7T consists of cyclic factors of pairwise different odd lengths, then by 1.2.1 Cs ( 7T) is generated by these cyclic factors, and hence is contained in An. This proves the following to be true: 1.2.10 LEMMA. C S(7T) splits into two An-classes of equal order if and ollly if n> 1 and the nonzero parts of a( 7T ) are pairwise different and odd. In all other cases C S( 7T) does not split.
We denote the conjugacy class of elements with cycle partition a by
If this class splits into two conjugacy classes of An. we denote these by
where the notation is fixed by assuming that 1.2.11
In order to complete the classification of ambivalent alternating groups, we must determine which CO:'" are ambivalent, i.e. contain the inverse of each of their elements. 1.2.12 LEMMA. A o, AI' A 2 • As, A 6 • A IO ' and A I4 are the on~}' ambivalent alternating groups.
Proof Let
13
1.2 Conjugacy Classes of Symmetric and Alternating Groups
be a product of disjoint cycles of odd lengths in cycle notation. We can form
and call this permutation the standard conjugator. It satisfies
(i) We notice first that ~ is even if and only if the number of cyclic factors of 1T whose lengths are congruent 3 modulo 4 is even. (ii) If the standard-conjugator is even for each element in a splitting class, then A n is ambivalent. The only splitting classes of An' nE{O,1,2,5,6,1O,14}, are the classes corresponding to the partitions (5)f-5,
(5,1)f-6,
(13,1)f-14,
(7,3)f-10,
(11,3)f-14,
(9,1)f-1O, (9,5)f-14.
°
In each of them the number of components congruent to 3 modulo 4 is or 2, so this number is even. Hence the standard conjugator is even in each case, by (i). This shows that the groups mentioned in the statement are in fact ambivalent. (iii) If the standard conjugator ~ of 1T is odd, then the A" -class CA( 1T) in question is not ambivalent. For if 0' E A n satisfies 0'1T0' - I = 1T- 1, then
which contradicts Cs ( 1T) = CA( 1T), since 0' - 1~ E S"\A,,. (iv) It remains to show that for each n~{O, 1,2,5,6, 10, 14} there are partitions (X with pairwise different and odd parts (Xi *0 such that the number of (Xi satisfying (Xi =3 (4) is odd. (a) (b) (c) (d)
n=4k, kEN: (4k-1, 1) has the desired properties. n=4k+ 1, 2",;kEN: (4k-3,3,1) satisfies the conditions. n=4k+2, 4",;kEN: Consider (4(k-l)-3,5,3, I). n=4k+ 3, kENo: Take (4k+ 3).
•
1.2.13 COROLLARY. The An-classes Cac'c are ambivalent if and only if the number of parts (Xi of (X with the property (Xi =3 (4) is even.
We conclude this section with an evaluation of the order of its conjugacy class C s ( 1T).
1T E
Sn and of
Symmetric Groups and Their YDung Subgroups 1.2
14
If in cycle notation
and if rEN, then since disjoint cycles commute, we have
so that the order of '1T, i.e. the order 1< '1T >I of the cyclic subgroup <'1T > of Sn which is generated by '1T, is the least common multiple of the lengths Iv of the cyclic factors of '1T. In terms of the cycle partition a( '1T) of '1T this reads 1.2.14
The conjugacy class C a which consists of the elements with cycle-partition a and cycle-type a is obtained as follows. Its elements arise from a system of a; empty cycles of length i, 1,,;;; i";;; n, i.e. from ... ( ..... ) ... ( ..... ) ...
I<-i-I a;
\<-i->\ cycles
by inserting the points 1, ... , n in all the n! possible ways. Since cyclic permutation of points does not change a cycle and since disjoint cycles commute, each element of c a arises just II JGia;! times in this process. This proves that for'1TES n:
IC S ( '1T)1 =n l/II,iG,(?T)a;( '1T )!, (ii) ICs ('1T)!=IIJ ,(?T)a;('1T)!. (i)
1.2.15
LEMMA.
G
We postpone a detailed description of the centralizer Cs ( '1T) of '1T, since it requires the concept of wreath product. The preceding results on conjugacy classes allow us to draw the first conclusions concerning the representation theory of Sn and An. 1.2.8 for example implies that all the ordinary characters of Sn are real-valued, while by 1.2.12 the following is true for alternating groups: 1.2.16 THEOREM. A a, AI' A z , As, A 6 , AlQ, and A l4 are the on~v alternating groups, the character table of which is a matrix over the real number field ~.
1.3 Young Subgroups of Sn and Their Double eosets
15
But the character table of Sn is even a matrix over the ring I of integers. For there exists a general theorem (see e.g. Huppert [1967, I, Chapter V, (13.7)]) which says that if G is a finite group and for every gE G and m E ~ which satisfies (m, IGI)= 1, gm is conjugate to g, then the character table of G is a matrix over I. Sn satisfies the conditions of this theorem; it is clear that for every '7TESn and each mE~ which satisfies (m, n!)= 1,
a( '7T m ) =a( '7T), since the mth power of each k-cycle, obtain
k~n,
is a k-cycle again. Thus we
1.2.17 THEOREM. The character table of Sn is a matrix over I, for each nE~o'
This follows also from the fact that there exists a I-basis of permutation characters for the character ring of Sn' These permutation characters are induced from the so-called Young subgroups of Sn' which will be introduced next.
1.3
Young Subgroups of Sn and Their Double Cosets
We now introduce the so-called Young subgroups of Sn' These subgroups are named after Alfred Young (1873-1940), to whom we are indebted for large parts of the ordinary representation theory of Sn' The permutation representations induced from Young subgroups playa vital role in both the ordinary and the modular theory of Sn' We have seen that the conjugacy classes of Sn are indexed by partitions, so it is to be expected that subgroups of Sn useful in representation theory should also be related to partitions. Indeed, we shall obtain a Young subgroup of Sn for each partition of n, but we want to define a wider class of Young subgroups. In order to do this we first introduce a concept which generalizes the notion of a partition. A sequence of nonnegative integers
is called an improper partition of n if and only if 00
1.3.1
~ A,=n.
,== I
The fact that A is an improper partition of n will be abbreviated by
Symmetric Groups and Their Young Subgroups 1J
16
Note that we require the parts of A to belong to No. Furthermore, 1.3.1 tells us that A, is eventually zero. We adopt the convention of writing
when A, =0 for all i>h. The only difference between a proper and an improper partition is that we do not require the parts of an improper partition to be nonincreasing. By IAI we indicate the natural number of which A is a proper or improper partition:
IAI: = L: A" i=1
Let A=(A I' A2' ... ) be such an improper partition of n, and let n~,
l";;;i,
denote subsets of n which are pairwise disjoint, and which satisfy
so that we have the following dissection of n: oc
It'
n= \.:) I~
I
"
If now S/' denotes the subgroup of Sn consisting of the A,! elements which leave each point of n~ fixed, then the product oc
1.3.2
SA: =
II S/' i=1
is a subgroup of S", which is isomorphic to the (finite) direct product
SA is called a Young subgroup corresponding to AF n or, more explicitly, the Young subgroup corresponding to rf: = {rrj, rf2 , ... }. It is important, and clear, that all the subgroups SA which correspond to A are conjugate to each other in S". Special cases are the Young subgroups Sa which correspond to a proper partition a~n. Obviously each SA' AF n, is of the form Sa' a~n.
We shall be interested in representations of S" which are induced by certain one-dimensional representations of Young subgroups of S".
1.3 Young Subgroups of Sn and Their Double Cosets
17
If F denotes a field, then there are two trivial one-dimensional representations of S>." AF n, over F, at hand. The first one is the identity representation IS>., of S>.,. Its representation space is the one-dimensional vector space F 1 over F, where each element 7T E S>., is mapped onto the identity element id F I of the general linear group G L( F 1) of F,:
1.3.3 The second trivial one-dimensional representation of S>." AF n, over F is the alternating representation AS>., of S>." where 7T is mapped onto ± id F I ' depending on the sign of 7T: 1.3.4 If t.t is another improper partition of n, then we can form IS>." IS"., and AS". and induce these representations into Sn' obtaining the representations
of Sn' We would like to evaluate the intertwining numbers
in order to get an idea about common irreducible constituents of these induced representations of Sn" Denoting by ~ the restriction of representations to subgroups, Mackey's intertwining number theorem (d. Curtis and Reiner [1962, (44.5)]) tells us the following: 1.3.5
i(IS>.,i Sn' IS". i Sn) = ~ i( ISd s>., n 7TS".7T - I ,ISJ") ~ s>., n 7TS".7T -}) s""s~
The sum has to be taken over the double cosets S>.,7TS". of S>., and S". in Sn and ISJ"l( 7Tp7T -}): = IS/p), pES".. This shows that the intertwining number of IS>., i Sn and IS". i Sn is equal to the number of double cosets S>.,7TS". of S>., and S". in Sn' For the second intertwining number we obtain analogously 1.3.6
Now both I(SAn1TS,.1T-1) and A(SAn1TS,.1T-1), the identity representation and the alternating representation of the intersection SA n 1TS,.1T -I, are irreducible. Since the intersection is a direct product of symmetric groups, the two irreducible representations are equal if and only if SA n1TS,.1T -) is trivial or the characteristic char F is equal to 2. Therefore if char F=2, 1.3.7
otherwise. S~'1TS#
s~n'1Ts#'1T-1 ={l}
1.3.5 and 1.3.7 show that the desired intertwining numbers can be expressed as numbers of specific double cosets of Young subgroups. This suggests that we should examine more closely the double cosets SA1TS,.. The following lemma will turn out to be crucial in this context:
Proof
(i) =:
If pES A1TS,., say P=01TT, oESA, TES,., then for each k,
so that for each i and k we have:
This yields the statement, since (ii) ¢=: The assumption
0
is a bijection.
implies that for a fixed i the subsets Ii; n1T[ntl and the subsets rr; np[ntl form two dissections of Ii; into subsets which can be collected into pairs
of subsets of equal order. Hence for each i there exist 'Vk
0i
(o;[n~ n1T[nt]] =n~ np[nt]).
E S;A.
which satisfy
19
1.3 Young Subgroups of Sn and Their Double eoselS
The product a:= a 1a2'" ES A of such permutations ai then satisfies the equations
'Vk
(a'1T[n~]=p[n~]).
•
Thus there exist TES!' such that p=a'1TT, as stated.
This shows that the double coset S A'1TS!' is characterized by the numbers 1 ~i, k.
Now assume that Ai =IJ-i =0 if i>n. Then we have an injective mapping
from the set of double casets S A'1TS!' into the set of all the n X n matrices over No. It is clear that the image of f is exactly the set of nXn matrices (Zik) which satisfy
1.3.9
n
(ii) ~
n Zik =IJ-k
~
and
Zik
=A i ·
k=1
i=1
Summarizing, we have proved the following important result: 1.3.10 THEOREM. If A=(A" ... , An} and IJ-=(IJ-" ... , IJ- n } are improper partitions of n with corresponding Young subgroups SA and S!', then the mapping
establishes a bijection between the set of double cosets of SA and S!' in Sn and the set ofnXn matrices (Zik) over No which satisfy n
~ i=1
n Zik =IJ-k
and
~
Zik
=A,.
k=1
For the number of these double eosets we obtain in particular: 1.3.11 COROLLARY. The number of double cosets S A'1TS!' is equal to the number of n X n matrices over No with prescribed vector A= (A I' ... , An) as vector of row sums and prescribed vector IJ- = (IJ-, , ... , IJ- n) as vector of column sums.
20
Symmetric Groups and Their Young Subgroups 1.3
If we restrict our attention to double cosets S/-,'7TSp. with the trivial intersection property
1.3.12
s/-,n'7TSp.'7T- 1 ={I},
and restrict f to this subset, we obtain as the image of this restriction the set of n X n 0-1 matrices with A as row-sum vector and Il as column-sum vector: 1.3.13 COROLLARY. The number of double cosets S/-,'7TSp. with the trivial intersection property 1.3.12 is equal to the number of n X n 0-1 matrices with row sums Ai and column sums Il k'
For numerical purposes it is useful to notice that the number of n X n matrices over 1\1 0 with row sums Ai and column sums Ilk is equal to the coefficient of
in the formal power series
while the number of nXn 0-1 matrices with row sums Ai and column sums Ilk is equal to the coefficient of x/-,yp. in the formal power series
This yields 1.3.14 COROLLARY. The number of double cosets S/-,'7TSp. is equal to the coefficient of x/-,yp. in
while the number of such double cosets with the trivial intersection property is equal to the coefficient of x/-,yp. in
1.4 The Diagram Lattice
2\
If we take for example l\: = (2, 1,0) and Jl: = (I, I, I), then we have to consider ( I - XI Y I ) -
Hence there exist just 3 double-cosets St2,I.O)7TS(I.I, I) in S3 (which is of course trivial from Stl. l , I) = {l}, but we wanted to demonstrate the numerical method of examining a generating function). An application of 1,3.5 yields
This shows how the formal power series
II (l+xiJ\)
i, k
enable us to evaluate the intertwining numbers
of representations of Sn which are induced from one-dimensional representations of Young-subgroups. Since intertwining numbers can often be interpreted in terms of the multiplicities of irreducible constituents common to the representations involved, such a method for calculating intertwining numbers is very useful.
1.4
The Diagram Lattice
In the last section we considered pairs of representations IS}, 1Sn and AS!, iSn of Sn and expressed their intertwining numbers i(/S}, lSn' AS!, lSn) in terms of numbers of double cosets and of numbers of 0-1 matrices. In order to apply these results we shall now take for l\ and fJ. specific pairs of partitions a and f3 of n. It is our aim to show that certain pairs (a, f3) of partitions of n have the property 1.4.1
i(IS" iSn, ASIJ lSn)= 1.
In the case when the characteristic of the groundfield F does not divide n!, 1.4.1 means that these two induced representations have a uniquely determined irreducible constituent in common, and that this constituent is
22
Symmetric Groups and Their Young Subgroups 1.4
contained in both of the induced representations with multiplicity 1. It will in fact turn out that we can obtain in this way a complete system of ordinary irreducible representations of Sn' In order to do this we keep the partition a = (at, ... , a h )~h
fixed. This partitIOn a can be illustrated by the corresponding Young diagram [a], which consists of n nodes X placed in rows. The i th row of [a] consists of a, nodes, 1~i, and all the rows start in the same column:
[a]: =
1.4.2
X X
X X
X
X
X
at nodes
X
a 2 nodes
ah nodes
X
The partition a : = (3,2, 12 ) for example can be visualized by X
321 2 ]=X
[ , ,
X
X X
X
X
(we write [3,2,1 2 ] instead of [(3,2,1 2 )]). Recalling that a, ~a,+t, 1~i, we see that the lengths a; of the columns of [a] form another partition a' of n: , ), a '·. -- (a'I' a2""
1.4.3
where a; : = ~ 1. j aj~i
This partition a' is called the partition associated with a. Correspondingly [a'] is called the Young diagram associated with [a]. [a'] arises from [a] by simply interchanging rows and columns, i.e. by reflecting [a] in its main diagonal: e.g.
2
~X
[3,2,1 ]=; X
/~ yields
X X X X [(3,2,1 2 )']= X X =[4,2,1]. X
Partitions a and Young diagrams [a] where a = a' are called self-associated. We aim to characterize the partitions f3~n which satisfy for a given a~n the inequality
I4 The Diagram Lattice
23
This can be done in terms of a certain partial order on the set 1.4.4
P(n): = {yly~n}
of all the partitIOns of n. The partial order will not be the natural lexicographic order .;;;;, which is defined as follows:
It is clear that .;;;; is a total order, so that the order diagram is always linear. The partial order ~ which we have in mind is defined in terms of the partial sums i
~yp I
of the parts YP of the partitions in question: 1.4.6 In the case when this holds we say that f3 dominates a and call ~ the dominance order. (When '\, JlF n, '\~Jl is defined similarly.) It is easy to see that the dominance order differs from the lexicographic order on pen) if and only if n~6. The order diagram of (P(6),~) is
(6) (5,1) (4,2) (3 2 ) (3,2,1)
1.4.7
(2 3 )
(2 2 , )2) (2,1 4 )
(1 6 ) It is obvious that the following is true for all a, f3~n:
1.4.8
a~f3
=
a';;;;f3.
It will be useful to have a characterization of partitions a and f3 of n which
24
Symmetric Groups and Their Young Subgroups 1.4
are neighbors with respect to 1.4.9
:~
a
a situation which we denote by a
[a
We shall prove that this can be expressed very nicely in terms of the corresponding Young diagrams [a] and [{3]. In fact we shall show that a
x
Moreover this step has to be as small a step as possible, which means either i=j-l or i
has the form
x X
X
j
More formally this reads as follows: 1.4.10 THEOREM. If a=(a l ,a2'00,) and {3=({3\,{32'00.) are partitions of n, then a
(1) Assuming a
j:
**
= min{t l
(Xv
=
{3v. i
1.4 The Diagram Lattice
25
so that obviously l~i
and
Also
aJ
> /3j;:;;' /3J+ 1 ;:;;'aj + I' Therefore
Hence in any case, whether i> 1 or not, we obtain
so that y = /3 by assumption, and [/3J arises from [aJ by moving exactly one node upwards from the end of the j th row of [a J to the end of the i th row of raj. If ii=j-l and a i i=aJ , then iaJ • We can therefore put
We notice that i j (a. = Y. /3.).
=
Hence if i=j-l, a
26
Symmetric Groups and Their Young Subgroups 1.4
In the case when i =1=j - 1, we obtain from (iii) that
By a
for v=l=k, l.
But this can hold only if k = i and l = j; for otherwise
or
would contradict a i = ... = a j . Therefore y = {3 in this case too.
•
This can be used in the proof of 1.4.11 LEMMA. a,,;;;J{3
~
{3',,;;;Ja'.
Proof If a,,;;;J{3, then there exists a chain of partitions aPf-n, satisfies
O~v~r,
which
1.4.10 yields that a i
Thus we obtain the chain
which shows {3',,;;;Ja'. The converse is true by symmetry. 104.11 means that the mapping 1.4.12
-':P(n)----;7 P(n):m....a'
is an order antiisomorphism. The dominance order is described in terms of the partial sums 1.4.13
•
27
1.4 The Diagram Lattice
of the parts of a. Using these expressions we can define for any a, /3l-n 1.4.14
a/\/3:=y,
and 1.4.15
where a/ : = min {at, af },
1 ~ i,
aV/3:= (a'/\/3')'.
It is not difficult to see that a V /3 is the supremum and a /\ /3 the infimum of a and /3 with respect to ~. Thus: 1.4.16. The dominance order
~
induces a lattice structure of P( n ).
We shall not stress this fact, but it should be mentioned that this lattice structure is the combinatorial background of a large part of the ordinary representation theory of Sn' as will soon become apparent. We call this lattice (P( n), /\, V) the diagram lattice of order n, since the name "partition lattice" might be misleading: it is already the standard name for a different lattice structure. In order to apply the structure of the diagram lattice P( n), we need to characterize the dominance a ~ /3 of partitions a and /3 of n in terms related to representations of symmetric groups. And there is in fact a classical characterization of a ~ /3 in terms of the existence of certain 0-1 matrices which we can easily apply to representations of symmetric groups by way of 1.3.13. This characterization of a Q/3 which we have in mind is the GaleRyser theorem, one of the most important existence theorems in combinatorics, where it is used mainly to prove the existence of incidence structures, by way of demonstrating the existence of corresponding incidence matrices. It reads as follows: 1.4.17 THEOREM OF GALE AND RYSER. If a=(a l , a2'. 00) and /3=(/31' /32' 00.) are partitions of n, then there exist 0-1 matrices with row sums at, a2' ... and column sums /3;, /3~, . .. if and only if a ~ /3.
We do not intend to give a proof of this theorem here, since the standard one is purely combinatorial. (That the existence of such a 0-1 matrix implies a ~ /3 is trivial; the other direction is proved by exhibiting an algorithm which describes the construction of such a 0-1 matrix. The interested reader is referred to the book of H. J. Ryser [1963].) Applying 1.3.7 and 1.3.13, we obtain a translation of this characterization of a~/3 into representation-theoretical terms: 1.4.18 THEOREM OF RUCH AND SCHONHOFER. If a and /3 are partitions of n, and S'" and SfJ' are Young subgroups ofSn which correspond to a and /3', and if the groundfield F has charF;io2, then i(IS", tSn,ASwtSn)*O if and only if a~/3.
28
Symmetric Groups and Their Young Subgroups 1.4
The proof of this theorem given by Ruch and Schonhofer does not use the Gale-Ryser theorem but another characterization of a ~/3 in terms of so-called Young tableaux. A Young-tableau t a with Young diagram [a] (sometimes called an a-tableau for short) arises from [a] by replacing the nodes X of [a] by the points i of n= p, ... , n}. for example, here are two of the 7! Young tableaux with diagram [Y, 1]: 1 2 3
4 5
6 7
516 243 7
The tableau obtained by replacing the nodes by 1,2,... in order down successive columns is denoted by tf: 1 1.4.19
tao 1'-
2
a; + 1 a; +2
a', a; + a; 1.4.20 LEMMA. If a, /3~n and t a is an a-tableau, then a~/3 if and only if there exists a /3-tableau t f3 such that any two points which occur in t a in the same row occur in t f3 in different columns.
Proof Assume the existence of tao t f1 with the given property. Then a, ~ /3,. since the points in the first row of t a occur in different columns of f1 t . As this also holds for the points in the second row of t a , there must be a 2 f3 places in different columns of t which have not yet been accounted for. Thus a 2 ~/31 +/3 2 -a,. Continuing in this way, we get a~/3. f1 Conversely, if a~/3 then t can be constructed algorithmically from an a arbitrary t (see Dress [1979]. Thiirlings [1977]). One can proceed recursively by taking the last point of the last row (row h, say) of t a , and putting f3 it at the end of row j of t , where): =max{i I /3i ;;;;:a h }· • The last lemma and the results of the preceding section give the following characterizations of the dominance order in terms of group theory, representation theory, and combinatorics:
1.4.21 THEOREM. If a and /3 are partitions of n, then the following properties of a and /3 are equivalent when char F=I=2: (i) a ~/3, (ii) for any two Young subgroups Sa and Sf3' corresponding to a and /3'.
there exist double cosets Sa7TSf3' with the trivial intersection property Sa n 7TSf1,7T - I = P},
1.5 Young Subgroups as Horizontal and Vertical Groups of Young Tableaux
29
(iii) for any two Young subgroups Sa and Sf3' corresponding to a and {3' and their representations ISa and ASf3' over F, we have i(/Sa iSn, ASf3 , iSn)*O, (iv) there exist 0-1 matrices with a as vector of row sums and {3' as vector of column sums, f3 (v) for each a-tableau t a there exists a {3-tableau t where any two points f3 which occur in t a in the same row occur in t in different columns, (vi) the coefficient of xayf3' in IIi,k(l + x;yd is *0. We have seen already that (ii), (iii), (iv), and (vi) are equivalent and that (i) and (v) are equivalent. The rest of the statement is obtained from the Gale-Ryser theorem or from the Ruch-Schonhofer theorem, neither of which we have proved yet. Since we are going to apply some of these results, it should be pointed out that we shall use only the trivial part of the Gale-Ryser theorem, which gives a ~ {3 from the existence of the relevant 0-1 matrix. A proof of the Ruch-Schonhofer theorem will be given later so that in fact the story will be complete.
1.5
Young Subgroups as Horizontal and Vertical Groups of Young Tableaux
In the last section we introduced the notion of a Young tableau t a with Young diagram [a]. If we are given such a Young tableau t a , then its rows and its columns define two dissections of n: the first one is a dissection of n into pairwise disjoint subsets of orders a" the second one is a dissection of n into pairwise disjoint subsets of orders For example
a;.
I 4
2 5
3
defines by its rows the dissection of 5 into subsets {l, 2, 3} and {4, 5}, while it defines by its columns the dissection of 5 into subsets {l.4}, {2,5}, and {3}. Correspondingly we obtain from the rows of t a a Young subgroup Sa' which is called the horizontal group of t a, and which will be denoted by or, if necessary, by H(t a ). The columns of t a define a Young subgroup Sa" which is called the vertical group of t a, and which will be denoted by
or, if necessary, by V(t a ). These specific Young subgroups Ha and va obviously have the property
1.5.1
30
Symmetric Groups and Their YDung Subgroups 1.5
Therefore a Young tableau t a displays in a certain sense the double coset SaTTSa' with the trivial intersection property Sa n TTSa,TT - I = {I}. For if we are given Sa and Sa" then we need only consider a tableau ta, which has Sa as its horizontal group, and take a TT such that TTSa,TT -1 is the vertical group of tao It is the aim of this section to place several properties of horizontal and vertical groups of tableaux at the reader's disposal. For a while we shall keep the diagram [0:] fixed, so that we need write only t for the tableau in question and H and V for its horizontal and vertical groups. If now TTESn , then we define TTt to be the tableau which arises from t by an application of TT to the points in the tableau t: 1.5.2
If
t= ... i . . .
, then TTt·-
.TT(i).
It is clear that the following holds:
In Chapter 3 it will emerge that elements of the following form are essentially (i.e. up to a scalar factor) idempotent elements of the group algebra FSn of Sn over the groundfield F, char Fln!, which generate minimal left ideals: 1.5.4
e: = ~
~ sgnTT·TTp.
7TEV pEH
Thus it will be useful to have a few results to hand on permutations of the form TTp, TTE V, pEH. A trivial but useful remark is that for every TTE V, pEH, 1.5,5
t ' : = TTpt=p'TTt,
where
p':= TTpTT- 1 EH(TTt).
It means that instead of first carrying out p, which leaves the points of t in their rows, and then TT, which usually does not leave the points of pt in their columns, we may first apply TT, which leaves the points of t in their columns, and then apply pi, which leaves the points of TTt in their rows. We use this remark in order to prove that for p E H( t), TT E V( t), we have: 1.5.6 LEMMA. Any two points which occur in the same column of t occur in t' : = TTpt in different rows.
Proof This follows immediately from tl=P'TTt. For two points in the same column of t are in the same column of TTt and hence in different rows, and they do not leave their row when we shift from TTt to P'TTt, since p'EH(TTt).
•
; 5 Young Subgroups as Horizontal and Vertical Groups of Young Tableaux
31
But the converse is also true: 1.5.7 LEMMA. If every two points which occur in the same column of t, occur different rows of t' : = at, then a is of the form 7Tp for suitable 7T E V( t ),
In
pEH(t). Proof By assumption, all the points of the first row of t' occur in different columns of t, so that a vertical permutation applied to t moves them into the first row. Leaving these points fixed, another element of V( t) moves the points of the second row of t' into the second row, and so on. Hence there exist 7TE V(t) such that 7Tt contains each point in the same row as does t'. A final application of a suitable p' EH( 7Tt)=7TH(t)7T - I yields t'=p'7Tt=7Tpt, for p:= 7T- I p'7T. • 1.5.8 LEMMA. If a f£. VH, then there exist transpositions
7T E
V, P E H such that
7Tap = a. Proof Since af£. VH, we obtain from 1.5.7 that there exist two points, say i and k, which occur in t in the same column, and which appear in t': = at in the same row. Let 7: = (ik) denote their transposition. Then 7 E V( t) and T EH( t') = aH( t)a -1, so that a -IT(] E H( t). Therefore 7T: = T and p: = a -1 7a fulfill the statement. •
With these results in mind, we study the following elements of the group algebra FSn (where F is an arbitrary field): 1.5.9
'If: = ~
Sgn7T'7T
and
X: =
1TEV
~ p, pEH
together with their product 1.5.10
e : = 'lfX= ~
~ sgn 7T' 7Tp.
1TEV pEH
We notice that it suffices to take the sum over all those a E Sn which are of the form 7Tp, 7TE V, pEH:
1.5.11
e= ~ sgn 7T '7Tp, 1Tp
for if a=7T I PI =7T2P2, then 7T2- 17T 1 = p2p11 EHn V= {l}, so that both 7T 1 =7T2 and PI =P2' A very important property of e is
1.5.12
"hE V, pEH
(7Tep=sgn7T·e).
Symmetric Groups and Their Young Subgroups 1.5
32
Besides this, we need to know which products ee' are zero. Since ee' =
'YX'Y'X', it will be useful to consider expressions of the form X"('. 1.5.13 LEMMA. Suppose that t and t' are tableaux with the same diagram and :}c: = X (t), "'(': = "'(( n. Then :ley' = if and ollly if there exist two points which occur in t in the same row and in t' in the same column.
°
Proof
(i) If there exist two such points, say i and k, then there exist subgroups H and V' in H and V' such that
(ii) Conversely, if no such pair of points exists, then by 1.5.7
IT'EV',
p'EH'.
This yields
X 'I" =X( t )'I.r( t') =X( IT'p't')'Y(t') =IT'p'X( t')p' -117" -I,y( t')
=IT'X (t')( sgn 17" =sgn 17" 'IT'X''Y'
I
)'\1'( t')
*0.
•
The second part of the proof can be rephrased as follows. Suppose that t and t' have the property that every pair of points which occur in t in the same row occur in different columns of t'. Then we may construct a new tableau t" by following the mechanical procedure sketched below:
1.5.14
1 1 1
r--I---, I
.
- --t------1
/
I
/
I
//
I
It"
~_./
/
//
)
33
Exercises
For the new tableau t" we have 1.5.15. X"=X, CV"=CV', and therefore XCV' = X"'Y" *0. If we write t" =a't'= at, then a' E V' and a EH, so that a' - Ia E V' H. This proves:
1.5.16 COROLLARY. Suppose that t and t' are tableaux with the same diagram, and t'=rt. Then XCV'*O if and only if r=7T'p forsome 7T' E V', p EH.
The argument used in the first part of the proof of 1.5.13 does not use the fact that t and t' are of the same diagram. Hence our next remark is 1.5.17 LEMMA. If a'¢/3 and t a , t f3 are tableaux with diagrams [a],[f3], then there exist two points which occur in t a in the same row while they occur in t f3 in the same column. Also X a 'Y f3 =0, and hence e ae f3 =0. Proof The existence of two such points is immediate from 1.4.20, and now part (i) of the proof of 1.5.13 completes the proof. •
In Chapter 3 we shall see that the elements e a are in fact essentially primitive idempotents, so that the results we have obtained can be interpreted in terms of irreducible representations of Sn'
Exercises 1.1 Prove that As is simple in the following two ways: (a) by considering the orders of the conjugacy classes of As, (b) by showing that the assumption {l}4. 1.3 (a) Show that, for n*6, each automorphism
*
CHAPTER 2
Ordinary Irreducible Representations and Characters of Symmetric and Alternating Groups The present chapter contains a characterization of the ordinary irreducible representations of symmetric groups as common constituents of monomial representations induced from Young subgroups. Besides this, the ordinary irreducible characters are given as .l-linear combinations of permutation characters induced from Young subgroups, and an important recursion formula is derived. As an application, the ordinary irreducible characters and representations of An are given. It is shown furthermore, that Sn is characterized by its character table. An examination of the values of the irreducible characters leads to the notions of cores and quotients of partitions which later on are of great importance for modular purposes. Finally, the Littlewood-Richardson rule, which is one of the most useful results of the whole theory, is derived.
2. \ The Ordinary Irreducible Representations of Sn
35
We know from general representation theory of finite groups that Sn has the same number of ordinary irreducible representations as conjugacy classes. 1.2.6 has shown that a conjugacy class of Sn is characterized by the cycle partition y~n of its elements. For each such partition y of n we constructed Young subgroups Sy and considered their trivial one-dimensional representations ISy ' the identity representation of Sy' and ASy ' the alternating representation of Sy' In Section 1.3 we expressed the intertwining number
a, f3~n, of the induced representations in terms of numbers of double cosets
and in terms of numbers of 0-1 matrices with prescribed row and column sums. The particular intertwining number
for example, is equal to the number of 0-1 matrices with row sums a, and column sums a~. But it is not difficult to see that there is exactly one 0-1 matrix which satisfies these conditions on its row and column sums, namely the matrix 2.1.1
(
~
1
: :.: : : : : : .1 . 10
~ : : : : : :~ ) 0
where the 1's are placed in each row as far to the left as possible. This yields 2.1.2
A careful interpretation of these equations yields several important results. First of all we remember that these equations hold whenever the groundfield F is of characteristic,,", 2 (cf. 1.3.7). Therefore let us interpret 2.1.2 first as a result on representations of Sn over the field a. Since the intertwining number i( " . ) is "bilinear" and has all its values in No, 2.1.2 implies that there is exactly one irreducible a-representation of Sn' say D, which is contained in both ISa t Sn and ASa. t Sn' Furthermore i(D, D)= 1, and D is contained in both ISa tSn and A Sa' tSn with multiplicity 1. Since i(D, D)= 1, D must be absolutely irreducible, so that it is also an irreducible C-representation of Sn' Altogether this yields the following basic result: 2.1.3 THEOREM. If a is a partition of n and if Sa and Sa' are Young subgroups of Sn' which correspond to a and ai, then the induced representations ISa tSn and A Sa' t Sn have exactly one ordinary irreducible constituent in common.
Ordinary Irreducible Representations and Characters 2.1
36
Furthermore this irreducible constituent can be realized over 0, and it is contained with multiplicity 1 in both IS" i Sn and AS". is,,.
As any two Young subgroups of SrI which correspond to the same partition a' ~ n are conjugate subgroups, the characters of IS" is,, and AS", i Sn depend only on the partition a of the integer n and not on the dissections of the set n which yield S" and S"" Hence the same must be true for the common irreducible constituent mentioned in Theorem 2.1.3; it depends only on a. We therefore denote this uniquely determined constituent, and also its equivalence class of representations, by [a], since there is no danger of confusing this with the corresponding Young diagram. Abusing the symbol n, we have 2.104
Two special cases are easy to identify, namely the identity representation ISn of Sn and the alternating representation ASn of Sn' We denoted by (I") the partition (I, ... , 1) of n. Since the only c.orresponding Young subgroup is S(I") = { 1}, both I S(I") i Sn and A SO") i SrI are equal to the regular representation RSn of Sn: 2.1.5
Furthermore from
S(,,)
=Sn it follows that
2.1.6
Thus 2.1.4 implies 2.1.7
It is important to perceive the close connection between [a] and [a ' ]. Since
and
(use the standard argument for characters J.L of G and v of H:;;;;; G that J.L0v i G=(J.L t H0v)i G), we have for each a~n: 2.1.8
This means that the ordinary irreducible representation [a' ] differs from [a] only on the odd permutations, and there just by the sign.
37
2. I The Ordinary Irreducible Representations of Sn
The set {[a]la~n}
2.1.9
is a subset of the set of equivalence classes of ordinary irreducible representations of Sn' We would like to show that 2.1.9 is already the complete set of these equivalence classes. A preliminary consideration shows 2.1.10 LEMMA. The multiplicity i(IS" iSn ,[,8]) of [,8] in IS" iSn is nonzero only if a~,8.
Proof 2.104 shows that i(ASw i Sn' [,8]) = 1. The assumption that i(IS" i SII' [,8])*0 yields therefore i(IS" i SII' ASf3 , 1SII )*0. Hence there exist 0-1 matrices with row sums a I and column sums p} f)I, which implies a ~,8. • Notice that in this proof we did only use the trivial part of the Gale-Ryser theorem, as we promised at the end of Section 104. We are now in a position to prove the main theorem of this section. 2.1.11 THEOREM. {[a]la~n} is the complete set of equivalence classes of ordinary irreducible representations of SII'
Proof We need only show that [a] = [,8] implies a =,8, for then the cardinality of {[a]la~n} is equal to the number of conjugacy classes of SII' so that this system must be complete. But if [a]=[,8], we can argue in the following way:
Thus by 2.1.10 we have both done.
a~,8
and
,8~a,
which imply a=,8, and we are •
Hence each ordinary irreducible representation of SII is realizable over C, so that C is a splitting field for Sn' This yields (apply (83.7) in Curtis and Reiner [1962]): 2.1.12
THEOREM.
Each field is a splitting field for Sn'
This is not true for alternating groups; A 3 provides a counterexample. Let us conclude this section with a numerical example. By 2.1.11, we know that [3], [2,1], and [13] are all the equivalence classes of ordinary irreducible representations of S3' 2.1.7 yields
Ordinary Irreducible Representations and Characters 2.2
38
Since the squares of the dimensions add up to 6, [2,1] has dimension 2. Since 0:: = (2,1) is self-associated, 2.1.8 yields that the character of [2,1] vanishes on the class of transpositions. This gives the character table of S3 up to one remaining entry: (2,1 )
-1
(3 )
a
*
1)
[I' J [2, I]
1
1
[3 J
The remaining entry turns out to be - 1 by the orthogonality relations, say. Hence the complete character table of S3 is (I')
(2. I)
-1
(3)
1)
a
-1
1
1
[! 'I [2,11
[3]
But we are far from being able to evaluate the character table for every n, so that we need to examine more closely the permutation characters induced by Young subgroups. Before we start doing this we should not forget to mention a result which follows from 1.1.26, 2.1.7, and the well known form of the composition series of symmetric groups: 2.1.13 THEOREM. All the ordinary irreducible representations of Sn are faithful except [n] (for n;;;'2), W] (for n;;;.3), and [2 2 ] (for n=4).
2.2
The Permutation Characters Induced by Young Subgroups
In the preceding section we constructed with the aid of the representations ISa i Sn a complete system of ordinary irreducible representations of Sn' The former are transitive permutation representations, and we would like to examine them more thoroughly. In order to do this, we put the p( n): = 1P( n)1 proper partitions o:i of n in lexicographic order: 2.2.1 and define a matrix 2.2.2 with the foHowing multiplicities as entries: 2.2.3
39
:.2 The Permutation Characters Induced by Young Subgroups
°
2.1.3 together with 1.4.8 and 2.1.1 shows that M n is an upper triangular matrix with l's along the main diagonal:
2.2.4
*
M= n
o
This has several important consequences. We denote by 2.2.5
the value of the character cycle partition f3, and by
~a
of [a] on the conjugacy class of elements with
2.2.6 we denote the value of the character
~a
of ISa i Sn on this same class. Then
2.2.3 yields
2.2.7 If we put the values
~;
and
~p
into matrices, say l~i,k~p(n),
2.2.8
then Zn is the character table of Sn' and En satisfies (cf. 2.2.7)
so that we obtain from det Mn *0 (cf. 2.2.4): 2.2.9 M n is a matrix over No and has determinant l; hence M n-
1
is a matrix over Z, so that by 2.2.9 each ordinary irreducible character ~a of Sn is a Z-linear combination of the permutation characters As there are exactly p( n) such permutation characters, this Z-linear combination is uniquely determined. Thus the following holds:
e.
2.2.10 THEOREM. The ring char(Sn)= EBfin/z~a' of generalized ordinary characters of Sn possesses besides its Z-basis {~alaf-n} the Z-basis aa/af-n}, which consists of characters of transitive permutation representations of Sn'
40
Ordinary Irreducible Representations and Characters 2.2
If we want to express fa in terms of this basis of permutation characters, we need to know M n- 1 (or M n ). It should be mentioned that M n can be evaluated using enumerations of double cosets. For M n is triangular and has l's along its main diagonal, while the scalar product of the ith and the jth row of M n satisfies (apply 1.3.5)
2. m ik m jk =i(ISa
2.2.11
i
iSn , ISa) iSn )
k
=
2.
1.
This number of double cosets can be found as a certain coefficient in II(1-x i y)-1 (d. 1.3.14). A numerical example illustrates this: M 3 is of the
form (notice (13)«2, 1)«3»
* 1
o
* *
and we know
2.
(IS(3)iS 3,IS(2,I)iS3 )=
1=1,
53,,5(2.1)
2.
( I S(3) i S3' I S(l J) i S3 ) =
1 = 1,
53 "5(13)
so that
*
1
o
)
Furthermore, (IS(2, I) i S3 ,IS(I) i S3) =
2.
1= IS3: S(2, \)1 = 3,
512 ,11"5(1'1
so that finally
M3 =
(6
2 1
o o
)
41
2.2 The Permutation Characters Induced by Young Subgroups
We have obtained from 2.2.11 and 1.3.14 the following result: 2.2.12 THEOREM. The matrix M n can be evaluated from the coefficients of the generating function IIi. k( 1- Xi Yk ) - I.
Since 1S(I") i Sn = RSn (d. 2.1.5), M n contains mensions
In
its first row the di-
2.2.13 of the ordinary irreducible representations [a] of Sn' And as i(IS(n) i Sn, 1Sa iSn)=i(ISn' 1Sa iSn)= 1, it contains in its last column only l's, so that we have
2.2.14
M= n
*
o
Furthermore the scalar product of the first row of M n with its ith row is just the index ISn: Sa'l, l";;;;i";;;;p(n). But the knowledge of M n alone does not suffice to evaluate Zn' We also need :::n' the matrix of the permutation characters. Since 1Sa i Sn is the permutation representation of Sn on the left cosets of Sa it is in principle possible to evaluate ~a( 17) by checking which left cosets of Sa remain fixed under left multiplication by 17. If we want to do this, it makes life easier to visualize the left cosets by so-called tabloids, which may be introduced as equivalence classes of tableaux. We call two tableaux t and t' (with the same diagram) row equivalent if and only if t' arises from t by a horizontal permutation: 2.2.15
t~t'
:=
317EH(t) (t'=17t).
The equivalence class of t is denoted by 2.2.16
{t} .
In other words: {t} arises from t by neglecting the order of points in the rows. We shall indicate this by drawing lines between the rows of t. It is obvious that the tabloids of shape [a] are in one-to-one correspondence with the left cosets of Sa and admit an action of Sn which is equivalent to the left
42
Ordinary Irreducible Representations and Characters 2.2
multiplication of Sn on the cosets of Sa' For example the (2, I)-tableaux are 1 3
2
1 2
2 3
3
2
3
3 2
1
3
2
1
while the tabloids with diagram [2, 1] are 1 2
3
3
2
3
2
I s3 leaves each one of them fixed, (12) leaves just the first one fixed, and (123) has no fixed tabloid. This enables us to evaluate ~(z, I): ~(z, 1)(1) = 3, ~(Z,I)«(12»= 1, ~(Z,I)«(123»=0. Thus we obtain for :::3' since
IS(13)
IS3 =RS3
and IS(3) IS3 =IS3, '=' _
-3-
(63 o1 1
1
We now obtain 2 3 in the following way:
-2
-1
o
1
o
1
-: )
A further result on M n can be derived from 2.1.10. For this lemma says that m ik ~O only if exi~exk. This means for the characters
2.2.17
~a =
r
a
+
~ (~a,
[P ] W~
f3r>a
and yields a lot of zeros in the matrix M n , n;,.6. We would like to derive a result, which in a sense reverses 2.2.17, and which will turn out to allow a representation theoretical proof of the Ruch-Schonhofer theorem. We have already seen that i(ISa I Sn' ISp I Sn) is equal to the number of matrices over No with ex as vector of row sums and p as vector of column sums. We can apply this if we happen to know the number of such matrices. The following lemma gives the number of matrices in a particular case:
2.2.18 LEMMA. If r l , rz , c I , and C z are nonnegative integers with the property r l +rz =c I +c z ' then the number of2X2 matrices over No with row sums r l , rz and column sums c l ' Cz is equal to 1+min{r p rz , c l ' cz}.
43
2.2 The Permutation Characters Induced by Young Subgroups
Proof If for example r\ =min{r" r2 , c" c2 }, then we have the following I + r I choices for the entries of the first row of such a 2 X 2 matrix:
(:
r\ -r) *
'
Each of these choices yields exactly one 2 X 2 matrix with row sums r, and column sums c" since r] was assumed to be the minimum, so that a suitable second row can be found and is uniquely determined. In the case when r2 , c" or c2 is the minimum, an analogous argument yields the statement. • This helps in the proof of 2.2.19 LEMMA. If a=(a" a 2 )t-n, a2 >0, and a*: =(a]
+ 1, a 2 -1),
then
(i) [aj+IS,," iSn = IS" iSn, (ii) the dimension of[ a j satisfies
r
"2
(iii) IS" iSn = ~ [n-p, pj. • =0
~"
Proof (i): 2.2.18 yields for the inner product of the generalized character - ~,," with itself: i(IS" i Sn' IS" i SJ + i(IS,," i Sn, IS,," i Sn) - 2i(IS" i Sn' IS,," i Sn) = (a 2 + 1) + (a 2 -1 + 1) - 2( a 2 -1 + 1) = 1.
e-
Hence ~,," is ± 1 times an irreducible character of Sn' Since (IS" iSn,[a])= 1, but (IS,," iSn,[a])=O (by 2.1.10), we obtain ~,,-~,," =K", as stated. (ii): The statement follows from the dimensions
by an application of (i). (iii): This part follows immediately from part (i), by induction on a2'
•
44
Ordinary Irreducible Representations and Characters 2.2
2.2.20 THEOREM. If a, {3~n. then there is a representation Df3a of Sn with ISf3 iSIl + Df3a =ISa iSIl if and only if a
Proof (i) If there exists Df3a with ISf3 i SIl + Df3a = ISa i SII' then (ISf3 i SIl' [{3]) = I shows that (ISa i SIl' [{3]);;;. I; thus a Q{3 by 2.1.1 O. Also a =/= {3. (ii) If a
This shows that it suffices to prove the existence of a Df3a of the required form only for the case a
for suitable i and j (cf. 1.4.10). If we denote by # the outer tensor-product multiplication of representations, then ISa iSn=([a,]#" '#[all])iSn = ([ a i ] # [a)] # (
#
[a v ] )
)
t Sn
V*l.)
which by 2.2.19(i) equals ( ([ a, . a)] + ([ a i + I] # [ a) -
I] ) i Sa, +,,)
)#
(
= ([ a,. a)] # ( #
vt.) [a v]))i SII [a,,] ) ) i Sn + ISfJ i Sn
"*1.)
where
• Expressed in terms of the matrix Mil this reads as follows: 2.2.21 COROLLARY. For given i and j. each entry m ik of the ith row of M n is greater than or equal to the corresponding entry m)k of the jth row of M n if and only if a ' Qa).
Furthermore we obtain from 2.2.20 and 2.1.10: 2.2.22
COROLLARY.
(IS" is,,.[{3])>O=aQ{3.
45
:3 Irreducible Characters as Combinations of Permutation Characters
Finally we obtain the Ruch-Schonhofer theorem:
For 2.2.22 yields the = part, while the other direction of the statement follows from the trivial part of the Gale-Ryser theorem. This completes the story and gives at the same time a representation theoretical proof of both the Gale-Ryser theorem and the Ruch-SchOnhofer theorem.
2.3
The Ordinary Irreducible Characters as Z-Iinear Combinations of Permutation Characters
An important result of the last section was Theorem 2.2.10, which says that each ordinary irreducible character ~a of Sn is a uniquely determined I-linear combination of the permutation characters It is the aim of the present section to derive a direct method for determining these linear combinations. The expression for ~a in terms of the is of a determinantal form. so we first introduce a multiplication of representations [a] of symmetric groups. If m and n are natural numbers, then Sm X Sn can be embedded into Sm+n in a natural way:
e.
e
2.3.1 since Sn ~Sm+n"m (where m+n: = p, ... , m+n}). Hence if af-m and f3f-n, then the (ordinary irreducible) representation [a]#[f3] of Sm XSn defines a representation of a subgroup of Sm+n which is isomorphic to Sm X Sn' and we can induce into Sm+n' The resulting representation is denoted by [a] [131 and called the outer product of raj and [f3J:
[a] [13] : = [ a] # [ 13 h Sm + n .
2.3.2
It is easy to see that this multiplication is associative (apply (43.3) in Curtis and Reiner [1962]), and it is commutative. Hence for example
2.3.3 With the aid of this multiplication we now define a determinant which corresponds to a=(a l ,a2, ... )f-n as follows: 2.3.4
l[ai+j-i]l=
[all
[at+l] [a l +2] [al+h-I] [ a 2] [ a 2 + I] [ a 2 + h - 2] ............................................ [ a 2 - 1]
[ah-h+l]
[a h -h+2]
[a h -h-1:"3]
[a h ]
Ordinary Irreducible Representations and Characters 2.3
46
when
2.3.5
ah+ 1 =
0, subject to the definitions
[r]:={~
if r=O (which is consistent with 2.3.3), if r
(We put I'[r]: =[r] and O·[r]: =0.) Note that 2.3.5 ensures that the determinant's definition is independent of the choice of h (subject only to ah+ I = 0). For example if a: = (3,1 2 )f-5, we get
[3]
[4]
I
[I]
0
I
[5] [2] =[3][1][1]-[3][2]-[4][1]+[5] [I]
[3]
[4] [1]
[5]
I
[2]
[6] [3]
0
I
[ I]
[2]
0
0
0
I
Since minus signs appear, the determinant 2.3.4 has to be interpreted as a generalized "representation". The character of our example is
If we had M s at hand, then we would see that this character is equal to ~(3.12l, so that the above determinant which corresponds to a=(3, 12 ) yields
the desired expression for ~a as a I-linear combination of the permutation characters And it is in fact true that for any partition a, the determinant 2.3.4 gives this expression, for short:
e.
[a] = I[a i +j - i ]1· It is our aim to provide a proof of this. In order to do this, we first generalize our definition of permutation characters ~a slightly. We defined an improper partition A of n to be a sequence of nonnegative integers whose sum is n. Thus, A may be regarded as an element of IN, the set of mappings from N into I. Now assume only that AE I Nand '2% I Ai = n (so that A still has finite support, but we no longer assume that its parts are nonnegative). Then A will be called a composition of n. (The reader will be encouraged to hear that we do not intend to introduce any more terms to describe elements of IN.) He who likes his definitions to be as general as possible may want occasionally to drop the assumption that '2~= lA, =n (e.g. in 2.3.6).
=3 Irreducible Characters as Combinations of Permutation Characters
47
Sometimes, when using compositions of n, we shall impose the restriction that Ai =0 for i>n (this works automatically for proper partitions)-a restriction which could, in fact, be avoided in all places where we make it, but this would force the introduction of some really unwieldy notation. To continue with the story, let A be a composition of n, and put 2.3.6
~
>... _
.-
{the character of IS>.. i Sn the function mapping Sn to 0
if AF n, otherwise.
Although it is unusual to consider "the zero character", as in the case where A is not an improper partition of n, these zero characters behave in the obvious way. The properties we need are these: (i) (the zero character of Sm) # (a character of Sd=the zero character of
Sm+k'
(ii) Inducing or restricting a zero character gives a zero character. These results follow by inspecting the processes involved; alternatively, they may be taken as the definitions of the restriction of a zero character, etc. Now regard Sn as the subgroup of SN consisting of the permutations fixing each point i>n. Then, by definition, every 7T which belongs to S", belongs to some Sn' (Remember that we chose to define a permutation to be a bijection fixing all but finitely many points.) Hence we may extend the function sgn to S", in a well-defined manner. For every composition A of n, and every 7T E S"" A0 7T is a composition of n, and
We define addition on .l N pointwise. Then 2.3.7
A-id+7T= (AI -I +7T(I), A2 -2+7T(2), ... )
is a composition of n. Using this notation, and having 2.3.6 in mind, we put for compositions A of n 2.3.8
x>..:=
~
sgn7T·~>"-id+".
"ESN
Since ~~= I AI = n, there exists hE'" 0 such that Ai = 0 for all i > h; then for all7TES N \Sh' A-id+7T has a negative part. Therefore, the sum in 2.3.8 is finite and 'Vi~h,
X>" = ~ sgn7T·g>..-id+". "ES,
48
Ordinary Irreducible Representations and Characters 2.3
Thus X A is a generalized character of Sn' In particular, if A: = (X~n, then X a is clearly the character of the determinant I[ (Xi +j - ill, so that it is our aim to prove Xa = ~a. We divide the proof of this basic result into several steps. A preliminary lemma gives an important property of X A:
2.3.9
LEMMA.
If A is a composition of n, and
Proof We put 7": =(i, i+ I)ESN . Then we have
if j"", i, i + I, if j=i, if j=i+ 1. This implies
2.3.10
~1'-id+?TT =e-id+?T,
LEMMA.
so that
Suppose that A is a composition of n=m+k. Then
(i) ~AiSmXSk=~I'Fke-I'#~I', (ii) X A iSm X Sk = ~I'F kXA-I'#~I'.
Proof (i): Both sides of the equation equal zero unless A is an improper partition of m+k. Assume. therefore, that AFm+k. (The sum over Jl has only finitely many nonzero terms, since ~A-I'#~I' =0 unless A- JlF m.) Mackey's subgroup theorem then yields
~A iSm X Sk =
~
I( (Sm XS k ) n7TSA7T -j) iSm XS k .
Sm XSk?TSx
By 1.3.10 the double-coset Sm X Sk7TSA is characterized by the 2 X n matrix
It is therefore uniquely determined by
2.3 Irreducible Characters as Combinations of Permutation Characters
49
The character of this representation is
and so (i) is proved. (ii): The definition 2.3.8 of XA yields
and by (i), this equals ~ ~ sgn'IT·~(A-I')-idh#~1'= ~ XA-I'#~I'.
I'F k 7T
I'F k
•
If A is a composition of n, and p. is a composition of k, we define sgn'IT·e- id -(I'-id)07T.
~
XVI':=
2.3.11
7TESN We claim that the following is true:
2.3.12 LEMMA. If A is a composition of m + k, then X A t Sm X Sk = " /3' k XA//3#X/3 • .:. Proof The proof of 2.3.1O(ii) has already shown that
xAt Sm X Sk =
~ sgn 'IT ~ e-id+7T-I'#~I'.
7TESN Replacing p. by
P.O'IT
we get (as
xAlSm XS k =
J1-Fk
~I'07T =~I')
~
sgn'IT ~ e-id-(I'-id)07T#~I'.
7T
~k
Now, for each p., P.i is eventually zero, and thus there exist uniquely determined 13 F k and a E S N such that
(13- id) and
0
a= p.- id
50
Ordinary Irreducible Representations and Characters 2.3
This enables us to proceed as follows (recall that all the sums are finite, so there is no problem about rearranging): x>'J.SmXS k =
sgn'7T
~ "ES",
=~
~ oES",
~~>'-id-(P-id)ooo"#~(P-id)oo+id
P
sgn(a'7T)sgna~e-id-(.B--idJoo"#e-id+o-1
".0
=~ P
P ~
sgn pe-id-(p-id)op# ~ sgnae- id +o- I o,=S",
pES",
= ~ X>'/fJ#X fJ P
= ~ X>'/fJ#X fJ . P,k
The last equation holds because xP =0 if Pi -i=Pi + 1 -(i+ 1) (by 2.3.9). • The lemma shows the importance of the generalized characters X>'/fJ of Sn-k'
2.3.13 THEOREM. If af-n, k";;;;n, and pf-k, then (i) xo./ fJ "",0 only if a; ~Pi for all i,
(ii) (xa/P,~(n-k»= {l ifal~PI~a2~P2~a3~"', otherwlse.
°
Proof (i): We consider the determinant by which X a / P is defined (cf. 2.3.11). It is (cf. 2.3.4)
As the sequences a-id and P-id are strictly decreasing, an entry [ai-i(Pj-j)]=O, i.e. a;-i-(p;-j)a; - i - (P, - i) = a; - Pi' Hence xa / P "'" implies P; .,;;;;a;. (ii): For each composition A of n-k we have (cf. 2.2.14)
°
(~\~(n-k»)=
{I
if A'F~-k, ° otherwIse.
Thus, if ~ denotes the Kronecker symbol,
TIsa, -;-(fJ.",-"U)).'"'O
~ ( X alP ,'i>/::.(n-k»)- "" sgn1T . "
I
2.3 Irreducible Characters as Combinations of Permutation Characters
51
If al ~f3l ~a2 ~f32 ~a3 ~ "', then this determinant has ]'s along its main diagonal and O's below, and hence it is equal to ]. Otherwise it is not of this triangular form, so that by the monotonicity of the sequences a - id and 13 - id two columns are equal, and hence the determinant must vanish in this case. •
This result turns out to be crucial in the proof of 2.3.14 YOUNG'S RULE (FIRST VERSION). For each AF n such that Ai ==0 when i>n and every partition a of n, we have that the inner product (x a , is equal to the number of (n - ])-tuples (13(1), .. . , f3( n- I» such that
e)
(i) 'VI ,,;;;;i";;;;n-] (f3(i)~2:~= ,A), (ii) 'V] ,,;;;;j";;;;n (f3/') ,,;;;;f3j2),,;;;; ... ,,;;;;f3?-IJ ";;;;a), (iii) 'Vj>], i~ I (f3p> ";;;;f3}!..-i », if we set 13(0): =(0) and f3(n): =a. '
Proof Applying 2.3.12, we have
which is by 2.3.]3
if the sum is taken over all the f3(n-l) which satisfy n-I f3(n-I)~ ~
Aj
and
al;;;"f3ln-I)~a2~f3in-I)~ ....
i=l
Another application of 2.3.]2 and 2.3.13 yields
where the sums have to be taken over f3(n-l) and f3(n-2) subject to the
52
Ordinary Irreducible Representations and Characters 2.3
following conditions: (i) ,8(n-I)~~7-IAj' ,8(n-2)~~;1-2Aj'
(ii) CX I ;;:,,8\n-l) ;;:'cx 2 ;;:,,Bin-l);;:, "', (iii) ,8\n-I);;:, ,8\n-2);;:, ,81 n- I);;:, ,81 n- 2);;:, .... Further iterations yield
(x",e)=
2: 2: ... 2:
p,n-I) p,n-2)
=
(XI1''',~(AI))
p(l)
2: .. ·2:1 p'n-I)
p(l)
2:
1,
(p'l), .... p'n-l))
where the sum is taken over all (n - 1)-tuples (,8(1), ... , ,8( n conditions described in the statement.
I)
subject to the •
We are now in a position to prove the main result of this section: 2.3.15
THEOREM.
For each
cx~n
we have that
the character of the ordinary irreducible representation [cx] of Sn' Thus, in particular, the set
is the complete set of ordinary irreducible characters of Sn' Each expressed in the determinantal form ~"=
2:
~"
can be
sgn7T·~"-id+"
"ESn
as a linear combination of permutation characters
e, AF n, with coefficients
0, ± 1. We can therefore express the representation [cx] itself as a generalized "representation" in the following deterrninantal form:
subject to the conventions [r]: = I if r=O, and [r]: =0 if r
2.3 Irreducible Characters as Combinations of Permutation Characters
53
Proof (i) We prove first that for any AF n (with A; =0 for i>n), (X", f')*0 implies a~A: This follows from Young's rule, for condition 2.3.14(iii) yields /3i l ) =/3j2) = ... =f3,~n~ll =0. Hence /3(I)f-A 1 together with a l ~/3\I) gives a l ~AI (by f3i l ) =0); f3( 2l f-A I +A 2 together with a l ~f3\2), a 2 ~f3f) yields a l + a 2 ~ 13\2) + /3i 2l ~A I + A2 (by f3j2l = 0), and so on. (ii) (X", ~")= I: By Young's rule (X", ~") is equal to the number of (n-l)-tuples (/3(1), ... ,f3(n~I) which satisfy (see (i»
Thus (( a l ),( ai' ( 2 ), ... ,( a l , ... , an~ 1» is the only possible (n -I)-tuple. (iii) (X", X") = 1: If (X", ~"-id+")*O, then a - id +1TF n anda~a - id+'17, by (i). But a-id +'17~a, as it is very easy to see, and so a=a-id +'17 and '17= 1. Together with part (ii), this shows that ~" is the unique summand f' of X" for which (X", f')*0. Therefore, (X", X")=(X", ~")= 1. (iv) In order finally to identify X" with ~", we consider the determinantal form I[a; -i+jJI which shows that [aJ is contained (use 2.2.22) in the main diagonal term only. Thus (X", ~")= 1. But (iii) shows that X" is ± an irreducible character, and so X" 0::: ~". •
A numerical example is
But
has the character
for '17 E S3 induces on the left cosets of S2 in S3 the same permutation as on the tabloids I 3 Thus
2
13 2
T3 I
Ordinary Irreducible Representations and Characters 2.3
54
This yields again the character table of S3:
More generally we have by 2.3.15 for each
n~2:
(i) [n-l,l)::::[n-l][l)-[n],
2.3.16
(ii)
~(n - I, I) ( 7T ) :::: a I ( 7T ) -
1.
(This of course also follows from 2.2.l9(iii).) The representation [n-l][l], which has the character
is sometimes called the natural representation of Sn' Another useful corollary of the deterrninantal formula is the following result on the values of the ordinary irreducible characters of Sn on the class of elements which consist of a single n-cycle:
2.3.17 f"((l ... n»::::{~-lr
if a::::(n-r,lr),
O:;;;;r:;;;;n-l,
otherwise.
Proof The only Young subgroup which contains n-cycles is the Young subgroup Sp: ::::Sn' Thus 'n))"",O only if f3::::(n). The deterrninantal form therefore shows ~a«(l ·n))""'O only if a summand ±~(n) occurs in
e«(l .. "
But this is the case only if I[a;+j-ill is of the form
I[a, +j-i)l::::
+ h - 1:::: n (d. 2.3.4), i.e. h::::n-a 1 + 1, or equivalently a is of the form a::::(n-r, 1') for a suitable r:;;;;n-l.
(h minimal such that ah+ I :::: 0), so that we must have a l
2.3 Irreducible Characters as Combinations of Permutation Characters
55
If on the other hand a=(n-r, 1'), then ~a((l ... n))=(_l)h-l~(n)((l. .. n))=(_l)h-l
=(-1)',
•
as stated. This shows that
~a((l
.. ·n»*O if and only if [a] is a f-shaped diagram:
x
X
X
X
X
[n-r,l']= : X X
Such diagrams are called hooks and play an important part in the representation theory of Sn. We would like to conclude this section with a proof of a formula which expresses the dimension of [a] in terms of hooks. It was T. Nakayama who first introduced the concept of hooks, by which we mean specific f -shaped subsets of diagrams [a]. In order to make this precise, we call the node of [a] which lies in the i th row and j th column the (i, j)-node of [a]. Then we denote by
r
2.3.18 the (i, j )-hook of [a], which consists of the (i, j )-node, called the corner of the hook, and all the nodes to the right of it in the same row together with all the nodes lower down and in the same column as the comer: j
The (i, k)-nodes, k>j, form the arm of the hook, while the (I, j)-nodes, l>i, form the leg of the hook. The (i, ai)-node is called the hand of the hook, while the (a), j)-node is called the foot: comer
hand
I!' I!' X X···········X a Hi} -
foot?
X] . :
X
arm leg
56
Ordinary Irreducible Representations and Characters 2.3
The number
h~j
of nodes of Hi), i.e.
2.3.19 is called the length of Hi). The number I':) of nodes in the leg of Hi), i.e.
2.3 .20 is called the leg length of Hij. To Hi~ there corresponds a part of the rim of [a] which is of the same length. It consists of the nodes on the rim between the hand and the foot of 2 Hi), including the hand and the foot node. To H,Oj2.1 j for example there correspond the encircled nodes of [3,2,1 2 ] as follows:
x
0 0
o o o
0
This associated part of the rim will be denoted by
It is important to notice that the result
of removing
R~j
from [a] is again a Young diagram: e.g.
x [ 3 2 12 ]\R(3.2.1 "
1.1
2
)
=
X
X X
=X=[l].
X X
We intend to show that the dimension product of all the hook lengths: 2.3.21
THEOREM.
r
r
of [a] is just n! divided by the
=n!/IIih~j.' .j
Proof Notice that for each composition A= (A I' A2'" .) of n we have A
_
n!
g (l) - AI''A 2', ... ' if we put (cf. 2.3.5) 1/r!=0 when r
2.3 Irreducible Characters as Combinations of Permutation Characters
57
h nonzero parts) the determinantal formula gives
an h X h determinant. Now
Elementary transformations of the determinant on the right hand side of this equation yield the Vandermonde determinant
Therefore 2.3.22
r
This equation shows that can be expressed in terms of the hooks which have their corner in the first column. To complete the proof we show that for 1 ,,;;;,i,,;;;,h we have h (h"il -h")1 )II'""=1 h"i" =h"iI", II )=i+l
Hi~
(1)
On each side of (1) there are hfl factors. The factors hfl -hjl are strictly increasing as j increases, while the hf" are strictly decreasing as P increases. Furthermore the factors on the left-hand side are all ,,;;;,hfl' It therefore suffices to prove that the factors on the left-hand side are pairwise different. This can be done by a verification of (2)
for a suitable j, which depends on Thus
a)+ 1
P.
We put j: =
a~,
so that aJ ~ P, while
and
This proves (2), which implies (1), which then implies the statement.
•
58
Ordinary Irreducible Representations and Characters 2.4
r,
This formula relates the prime divisors of n!, and the hook lengths so it is helpful when we come to consider modular representations.
2.4
h~J'
A Recursion Formula for the Irreducible Characters
In the preceding section we saw how ~o: can be expressed as a I-linear combination of permutation characters so that we are now able to of the characters calculate the character table Z; of S, from the table :::n by suitable applications of the determinantal form of [a]. But this is still unsatisfactory; we would prefer to evaluate Z; either directly or recursively. To obtain a recursion formula for ~o: we need to know what happens when [a], a a partition of n, is restricted to subgroups Sm, m..is a composition of n, we put
e,
e
e
2.4.1
+" ...),
X- :=(A"
, Ai-I' Ai -I,
N+: = (A"
, A;_I' A; + I, AI+"''')'
Ai
e,'-
Then 2.3.10 yields (since =0 if Ai - is not an improper partition of n - I; in particular, =0 for all sufficiently large i):
e'-
2,4.2 LEMMA. 00
e~Sn-I=2:e-. i=1
Applying this to the determinantal form of
~o:,
we obtain
~O:~Sn_'= ~ sgn7T(gO:-idh~Sn_l) ?TES
n
= 2:sgn 7T2:g
/
?T
By 2.3.9 we have that X a'-7"=O only if
(\:j-I;;;'a
j
+,. Therefore only sum-
59
2.4 A Recursion Formula for the Irreducible Characters
mands for which cx;>cx;+I' and hence cx'-~n-I contribute to KatSn~I' This proves the first half of the following theorem, the second half of which follows by Frobenius's reciprocity law:
2.4.3 THE BRANCHING THEOREM. If cx=(cxl'cx 2 , ... ) is a partition of n , then we have for the restriction of [cx1to the stabilizer Sn-l of the point n
; ll:'j>a'+l
On the other hand, if S, denotes the stabilizer of the point n + I in Sn+ I' then we have for the induced representation
(ii) If a is p-regular, >[ a] is a constituent of [a] with multiplicity 1. The other constituents of [a] have the form >[ y] with y p-regular and [y] c> [a]. More generally, (iii) If f3 is any partition of n, then [f3] contains >[f3]R with multiplicity 1, and the other constituents of [f3] have the form >[y] with y p-regular and [y]c>[f3]R. In terms of individual blocks, Theorem 6.3.50 says: 6.3.60 THEOREM. Within each block arrange the row labels so that all p-regular diagrams (in lexicographic order) come before all the p-singular diagrams (in lexicographic order). Then the block of the decomposition matrix looks like
o * The columns will be labeled > [a], where a is p-regular, and automatically these labels will be in lexicographic order. Further zeros can be inserted using the fact that if [a] is p-regular and [a] ~ [f3]R, then there is a zero at the intersection of the row labeled [f3] and the column labeled >[ a]. Theorem 6.3.50(iii), may also imply some zeros. 6.3.61 EXAMPLES. p=2, n= 13. >[7,5,1] is not a constituent of [8,4,1], because [7,5, 1] ~ [8,4,1]; nor is it a constituent of [7,1 6 ], because [7,5, 1] ~ [7, 16 ]R =[7,6]. By Theorem 6.3.50(iii), >[7,6] is not a constituent of [5,4 2 ], as is seen in the pictures below:
[7,6] =
~ o1 o1
1
o
0
1
1
%
/l'
1
'
o 1
o 1
o
1
o
o
1
It is clear that we have not got all the mileage we could out of theorem 6.3.50(iii), since it was not used in the induction hypothesis. An improvement of this result and other r-inducing information could be given, but it is hardly worthwhile formulating the statements, since they seem to be of no general value, and the results would become apparent in numerical calculations. Several other corollaries of Theorem 6.3.50 can be listed.
6.3 Techniques for Finding Decomposition Matrices
287
6.3.62 COROLLARY. The first ordinary character in lexicographic order within each block is modularly irreducible.
Because of 6.3.51(i), the decomposition matrix of Sn will be "wedgeshaped" (6.3.1) if the p-regular ordinary characters are put at the beginning, in any order which agrees with the dominance order. It is clear that the first p-regular diagram we must take in each block is unique (add nodes to the first row of the core). Less clear is that there is always a unique p-regular diagram [a] satisfying: 6.3.63. [f3] is ap-regular diagram in the same block as [a] =[a]¢>[,B]. But such a diagram obviously exists, and we can say of it: 6.3.64 COROLLARY. The column of the decomposition matrix corresponding to a p-regular diagram [a] satisfying 6.3.63 can be written down at once, since there is a 1 opposite [f3] if[f31R =[a], and 0 otherwise. EXAMPLE. The 3-regular diagram in question for the principal 3-block of S6 is [3,2, 1]. The diagrams [f3] satisfying [f3]R =[3,2, 1] are [3,2, 1], [3,1 3 ], [2 3 ], and [2,1 4 ]. (See Cs in 6.3.25.)
6.3.65
A similar result is 6.3.66 COROLLARY. r-inducing provides an algorithm for wrltmg down a complete set of linearly independent projective characters of Sn'
Proof It is not necessary to determine the decomposition matrix of Sn before inducing further the projective characters found by r-inducing. •
The matrices obtained by following Corollary 6.3.66 are presented, incorrectly, as decomposition matrices in Robinson's book. Compare 6.3.11 with the next corollary of Theorem 6.3.50. 6.3.67 COROLLARY. Each p-singular character can be written as a linear combination, with integer coefficients, of the p-regular characters on p-regular classes.
When n;;;'5, we find that [n]+[n-2,2]=[n-2,1 2 ] on 2-regular classes. Such relations also follow from the Murnaghan-Nakayama formula (d. Littlewood's paper [1951]).
6.3.68
EXAMPLE.
It is clear that a similar result to Theorem 6.3.50 could be proved by interchanging the words "row" and "column" (referring to diagrams). In particular. the modular irreducible characters can be labeled not only by
288
Modular Representations 6.3
p-regular diagrams, but also by diagrams which have no p columns of the same length; that is, by diagrams [13] whose associated diagram [13'] is p-regular. We now return to the problem of how to proceed when inducing projective indecomposable and modular irreducible characters from Sn-l to Sn does not provide the whole decomposition matrix. The short answer is to try the same tricks on a different subgroup. When most of the decomposition matrix of a group is known, the plan of working entirely with modular irreducible characters (6.3.27 and 6.3.33) really comes into its own. As we have explained, there are equivalent methods using projective characters, but these do take longer. This is a most frustrating stage of the game. Suppose we are trying to determine the multiplicity of a known modular irreducible ep in a character X, and we know that this multiplicity is 0 or 1. The two methods to try are: 6.3.69. Prove (X -
ep ) ~ H is not a modular character of H.
6.3.70. Find a modular character ~ of H such that ~ i G cannot be a modular character of G unless ep c:; X.
These two methods are virtually independent, and there is a fifty-fifty chance that at any given time we are trying the wrong one. Two notes of warning: 6.3.71. If a subgroup H fails to give the answer, it is no good trying 6.3.69 and 6.3.70 on a subgroup contained in H, because of the transitivity of the process involved. 6.3.72. The method of 6.3.70 will not work if ~ agrees with an ordinary character of H, because inducing an ordinary character gives an ordinary character.
I
We shall illustrate with Sn' For n";;; 10, there are only four problem cases not solved by inducing from Sn _ I' namely
ep [8 ] in ep[5] in The multiplicity of ep[7] in
I
[ 5 ,3] mod 3 [3,2] mod 2 [5,2] mod 2
ep[lO] in [6,3,1] mod 2
.
It
IS
0 or 1 } 0 or 1 0, 1, or 2 . 1,2, or 3.
1
is undecided.
6.3 Techniques for Finding Decomposition Matrices
289
6.3.73 EXAMPLE. The multiplicity of <1>[7] in [5,2] mod 2 is O. We must choose a subgroup of S7 not contained in S6' In passing, it must be noted that, in general, the decomposition matrix of An is as hard to find as that of SIl' so we assume no knowledge about A 7. Since we do not want to become too involved in calculating the decomposition matrix of our subgroup, we choose one which has a simple decomposition matrix. A subgroup of odd order would be good, because 6.3.74. If pil H I, the ordinary characters of H are modularly irreducible.
The most obvious subgroup of S7 to try is that generated by a 7-cycle. This does not work, because ([5,2]t H, I)H =2. Try, therefore, a slightly bigger subgroup. The group N generated by (1234567) and (235)(476) contains Is" 6 7-cycles. and 14 elements whose cycle partition is (3 2 , I). Part of the character table of S7 is [5,2]
(17)
(3 2 1)
(7)
14
-I
a
Therefore [5, 2]t N does not contain the identity character. Since ([5,2][7]) ~ N is not a modular character of N, we are finished. 6.3.75 EXAMPLE: The multiplicity of [8J in [5, 3J mod 3 is O. Bearing in mind the method of the last example, and looking at the character table of Sg, we soon come around to trying the group H generated by (1234) and (15)(26)(37)(48) (H c:::=. C4 wrC 2 , and 3fi H 1= 32). Sure enough, [5, 3]t H does not contain the identity character, and the problem is resolved. The above examples are just the start of many troubles. If pU > 1, and the power of p dividing n-pu + 1 is exactly a, then there is difficulty in deciding the multiplicity of the identity character in [n - pU, pU J. This multiplicity is a (see 7.3.23), but apparently could be >0 if we use only Sn-l' The first problem like this for p =: 5 is in [9,5J, and no elementary character-theoretic method of finding the multiplicity of the identity character here is known. 6.3.76 EXAMPLE: The multiplicity of (5J in (3,2] mod 2 is I. (Cf. Example 6.3.15.) This is an interesting problem. As the alternative multiplicity is 0, we should try 6.3.70 this time. But since S4 fails, 6.3.71 and 6.3.72 imply that we have to use As. It is lucky that a trick works, becau:-e we certainly do not want to assume knowledge about A n in general. The idea is to see that (3,1 2 ] splits when restricted to As. Since <1>(3,2] is the only constituent of (3,1 2 ] besides the trivial character, (3,2J must split into two modular characters of equal degrees. Thus <1>[3,2] has even degree. But [3,2]=a[5J+[3,2J with a =: a or 1, and deg(3,2] is odd. Therefore, a = I.
Modular Representations 6.3
290
We shall use this last example to show that the most elementary properties of representation modules are sometimes more powerful than mere character-theoretic results (6.3.82 and 6.3.88). It seems to be very hard to determine the multiplicity mod 2 of
6.3.77 EXAMPLE. When p=2,
- ([11] + [10, 1] + [9,2] + [8,3] + [7,4] + [6,5]) is a 2-modular character of SII' Taking only those characters having 2-core [2, I] gives [6,3,2]-[10,1]-[6,5], and the result follows. Before moving on to methods which involve more than just character theory, we mention two elementary methods which may be of use. 6.3.78. Restricting projective indecomposables and modular irreducibles from a group containing G gives projective and modular characters of G.
6.3.79 EXAMPLE. [3,2,1] is a projective indecomposable of S6 modulo 2 (it is in a block of defect 0). Hence [3,2] + [3,1 2] + [2 2.1] is a projective character of Ss' 6.3.80. Tensoring two modular characters gives a modular character.
This can be useful in one of two ways. We may have to subtract constituents from known modular characters in order to write the product as the sum of modular characters. Alternatively, it may be possible to prove
6.3 Techniques for Finding Decomposition Matrices
291
modules. It is encouraging to see that most of Theorem 6.3.50 (which has relied heavily on the Nak~yama conjecture) can be proved quite simply by examining composition factors of the representation modules. That some results can be found very easily by considering representation modules is illustrated by
6.3.81 THEOREM. Suppose M is a permutation representation of G on n points, and p divides n. Then the identity p-modular representation is a composition factor of M at least twice. Proof Let v I' v 2 , ••• , vn be the n points on which G acts. Then we may regard M as the vector space over a field of characteristic p whose basis elements are v I' v 2 , ••• , v n. The following are submodules of M: VI = {a\v\
+ ... +anvnla\ + ... +a n =O},
V2 =(V\+V 2 +' "+Vn)F'
.
M / V\ and V2 are both the identity module. Since pin, VI J V2 , and we are ~~
Of course, the hypothesis of this theorem shows that there is a homomorphic image of Gin Sn' and the theorem is essentially about the group Sn'
6.3.82 EXAMPLE. [5]+[4,1]+[3,2] is the permutation character of Ss on the subgroup S3 X S2' of index 10. Therefore the identity p-modular character [5] is a constituent of [4, 1]+[3,2] if p=2 or 5. By Nakayama's conjecture, [5] <;;;[3,2] if p=2, and [5] <;;;[4, I] if p=5. Very occasionally, a special piece of information is available which helps to find modular characters.
6.3.83 EXAMPLE. As '="'SLi4), the group of 2X2 matrices with determinant lover the field of order 4. Hence A s has a 2-dimensional representation over the field of order 4. Applying the field automorphism, an inequivalent representation is obtained. These are the two parts into which [3,2] splits (see Example 6.3.76). In conclusion, we give two results which sometimes solve problems which appear not to yield to purely character-theoretic arguments. The first is an easy, and very weak, result analogous to the Frobenius reciprocity theorem. It is useful only when we have a representation which is "nearly" irreducible.
6.3.84 THEOREM. Suppose M is an irreducible G-module, and H is a subgroup of G. If L is an H-submodule of M, then M is a top composition factor of L G.
292
Modular Representations 6.3
Proof L G =L0g 1 + ... + L0g". where the g;'s are coset representatives of H in G. (tEL) is a nonzero G-module homomorphism from L G into M. Since M is irreducible, we must have L G/Ker () ~ M. • 6.3.85 EXAMPLE. For p=2, see the decomposition matrix of SII' [4 3]tSII = [4 2,3]=2<[>[11]+<[>[9,2]+<[>[5,4,2]. Now, if [4 3] is irreducible, either its representation module or the dual of this has an S II submodule admitting <[>[ II] or <[>[9,2]. But <[>[II]TSI2 ~[43], <[>[9, 2]T SI2 ~[43], contradicting Theorem 6.3.84. Therefore, [4 3] is reducible. The same argument works for any factor of [4 3] which restricts to Sll to contain <[>[5.4.2] and another factor. Hence [4 3] contains a factor (which must be irreducible) which restricts to SII to be <[>[5,4,2]. 6.3.86 THEOREM. If <[> is any irreducible 2-modular character of a finite group G, then either <[> is the identity character. or <[> is not real. or the degree of <[> is even.
Proof Suppose that F is a field of characteristic 2. and G is a finite absolutely irreducible subgroup of GL(n, F) with n> 1. Suppose too, that for every g in G, tr( g) = tr( g' - I), where g' is the transpose of g. We must show that n is even. There is a nonsingular matrix a such that for all g E G g= ag' -J a -I. By Schur's lemma. a'=p,a where p,EF. Therefore, a=a"=p,a'=p,2a. Thus p,= I and a=a'. Define a bilinear form «1>( , ) on F" by letting «I>(v,w)=vaw'. By construction, this is a G-invariant symmetric bilinear form. V: = {v 1«1>( v, v) = O} is an invariant subspace of F" of codimension at most 1. Therefore V=F", and the bilinear form is alternating. This shows that n is ~w.
•
6.3.87 COROLLARY. All the irreducible 2-modular characters of S", except the identity one. have even degree. If <[> is a 2-modular character of S'P then the identity character has even multiplicity in <[> if deg <[> is even, or odd multiplicity if deg <[> is odd. 6.3.88 EXAMPLE. deg[3,2] is odd, so it contains the identity 2-modular character.
Exercises 6.1 Let a be a p-core. Using the technique employed for 6.1.9, prove that the number of p-regular partitions whose p-core is Ii equals the number of partitions whose associated partition is p-regular and whose p-core is Ii.
6.3 Techniques for Finding Decomposition Matrices
293
6.2 Provide an alternative proof of the p-block structure of An (Theorem 6.1.46), by considering the central character Wi, whose definition is
d(g)= ICG(g)I~I(g).
I' 6.3 Prove that the Brauer tree for the principal p-block of Sp is given by 6.3.9. 6.4 Write n=x(p-1)+r, where O~r
[In]=[(x+ 1)', x p - 1 - r ]. 6.5 Use the Littlewood-Richardson rule, and the fact that [ll]=[y] on 2-regular classes, to prove that r
[x+I,I'-']+[x,I']=
2:
[x+y-i,i]
;=0
on 2-regular classes. Deduce necessary and sufficient conditions for [x, 1'] to be 2-modularly irreducible.
CHAPTER 7
Representation Theory of Sn over an Arbitrary Field
Although much information about the p-modular representations of Sn can be obtained in terms of characters, difficulties remain which apparently cannot be overcome by character-theory arguments alone. More results are found by examining the representation modules of Sn' It is useful to have a module for each ordinary irreducible representation. Such modules have already been constructed-the left ideals of the group algebra in Section 3.1- but it is easier to work in terms of Specht modules, which will be defined in the first section of this chapter. Each left ideal of Section 3.1 is isomorphic to some Specht module, and so every result could be interpreted in terms of the group algebra. Essentially, the advantage enjoyed by the method of examining Specht modules, modulo p, over that of looking at the p-modular components of the ordinary characters is that the order of the factors in a composition series can be noted.
Let F be an arbitrary field, and consider the F vector space MfJ spanned by f3-tabloids. Recall from 2.2.16 that a f3-tab10id is a f3-tableau with unordered row entries; formally, the tabloid {t) containing t is the equivalence class of t under the equivalence relation
As before, in examples we shall draw lines between the rows of t to denote the tabloid {t}. If we let Sn act on the set of f3-tabloids by
7T{ t} : = {7Tt } then this action is well defined, and extending the action to be F-linear turns MfJ into an FS n -module. Clearly Sn acts transitively on the set of f3-tabloids, and the stabilizer of a given f3-tabloid is a Young subgroup Sp. Therefore
7.1.1. MfJ: = the FSn -module spanned by f3-tabloids is isomorphic to the permutation module of Sn on a Young subgroup Sp. MP is a cyclic module, generated by anyone f3-tabloid. A f3-polytabloid is an element of MP of the form
where, as usual, Clfl is the signed column sum for t.
7.1.2
EXAMPLE.
If
t= 1 4
2 3
5
then Ci{1
= (1- (14))(1- (23))
and
e =_1_2__ 5 I 4 3
4
2
3
5
1 4
3 2
5 +_4_3__5 1
2
It is important to note that the polytabloid e l depends on the tableau t, not just the tabloid {t}.
296
Representation Theory of S" over an Arbitrarv Field 7.1
The Specht module SfJ associated with the partition 13 is defined to be the subspace of MfJ spanned by polytabloids. Now, w'lr l ='Y"'IW, so weI =e."t' and SfJ is an FSn-submodule of MfJ. Further, 7.1.3. SfJ: = the FSn-module spanned by f3-po(vtabloids is a cyclic module, generated by anyone f3-polytabloid.
Notice that when f3=(n), SfJ =MfJ =the identity FSn-module. When MfJ is isomorphic to the regular representation module, and SfJ affords the alternating representation.
13 = (I n),
7.1.4 LEMMA. Each left ideal FSnC"I{I,:}C (t a f3-tableau) of the group algebra defined in Section 3.1 is isomorphic to the Specht module SfJ.
Proof The FSIl-homomorphism from FSIl 5C to MfJ given by 5C --+ {t} is clearly a well-defined isomorphism, since w fixes 5C <=> wE H {<=> w fixes {t}. Restricting this isomorphism to the subideal FS,,'1(15C proves the lemma. •
This lemma illuminates the advantages of working with Specht modules, since the ideal FS,,'YI']C depends upon the tableau chosen, whereas SfJ depends only on the partition 13. We have also replaced 5C, which is a long sum of group elements, by the single object {t}. An independent proof that Specht modules can be used to describe all the ordinary irreducible representations of Sn will be given shortly. A few preliminary results are needed. 7.1.5
LEMMA.
Suppose that a, f3~n. Then
(i) 'YIMfJ =0 if t is an a-tableau and a~f3: (ii) 'YIMfJ =Fe l if t is a f3-tableau.
Proof Let t* be a f3-tableau, and a and b be two numbers in the same row of t*. Then ( 1- ( ab )) {t *} = {t *} - {( ab )t *} = 0.
'1r l{t *}*0, then a and b must lie in different columns of t; for, otherwise, we could select signed coset representatives a I' ... ,ok for the subgroup of V I consisting of I and (ab), and obtain '1(1{t*} =(0, + ···+ok)(l-(ab)){t*}=O. If l is an a-tableau, ,\il{t*}*O therefore implies ar;:;. 13, by Lemma 1.4.20, and this proves the first result. If t is a f3-tableau, and 'l,rl{ t*} * 0, then the fact that the numbers from each row of t* must belong to different columns of t shows that {t*} = w{t} for some wE VI. In this case, 'l,rl{t*} ='l,rlw{t} = ±'YI{t} = ±et' and the second result follows. •
If t is any other tableau, and
297
7.1 Specht Modules
Before going further, a bilinear form <1>( , ) is introduced on MfJ. 7.1.6. Let
"'({f,).(f,})~ {~
if
{ t d = {t 2 },
if
{t d
*{t
2},
and extend to be bilinear on MfJ. This is a natural symmetric, Sn-invariant, bilinear form which is nonsingular on MfJ. We shall often use tricks like the following one: For u, vEMfJ,
(CV'u. v)= ~ ((sgn'1T )'1TU, v) 7IEV'
=
~ (u,(sgn'1T)'1T- 1v) 7IEV'
=
~ ( u, (sgn '1T ) '1TV
)
7IEV'
= <1>( u, 'Y'v ).
With this preparation behind us, we give the crucial theorem: 7.1.7
THEOREM.
If A is an FSn-submodule of MfJ, then either Ad SfJ or
A CSfJ.l.
Proof Suppose that a EA and t is a ,B-tableau. By 7.1.5(ii) CV 'a is a multiple of e,. If we can choose a and t so that this multiple is non-zero, then e, EA. Since SfJ is generated bye" we have A d SfJ. If for every a and t, CV'a=O, then for all a and t 0= ( 'Y 'a, {t } ) = ( a, 'y I {t }) = ( a, e, ). That is, A C SfJ.l .
•
7.1.8 THEOREM. If F is any field (in particular if F=Q or Z.p), then SfJI(SfJ nsfJ.l) is an absolutely irreducible FSn-module, or zero.
Proof By Theorem 7.1.7, any submodule of SfJ is either SfJ itself or is contained in SfJ n SfJ.l. Therefore, SfJI( SfJ n SfJ -L ) is irreducible or zero. Choose a basis e 1, ... , ek for SfJ where each e; is a polytabloid. Then dime sfJI( sfJ n sfJ.l » is the rank of the Gram matrix (i.e. the matrix whose 0, j)th entry is (e;, e» with respect to this basis. But the entries in the
298
Representation Theory of Sn over an Arbitrary Field 7.1
Gram matrix belong to the prime subfield of F, since polytabloids involve tabloids with coefficients ± 1. Therefore, the rank of the Gram matrix is the same over F as over the prime subfield, and so Sf3 n Sf3 -L does not increase in dimension if we extend F. Since Sf3/( Sf3 n Sf3-L) is always irreducible when it is nonzero, it follows that it is absolutely irreducible. • When F=Q, Sf3 nsf3-L =0, and no two Specht modules for different partitions are isomorphic (this is proved in 7.1.14 below, but can readily be deduced in this special case using 7.1.5(i». Since the number of partitions of n equals the number of conjugacy classes of Sn' we have:
7.1.9 THEOREM. When F=Q, the various Specht modules Sf3 (j3f-n) give all the ordinary irreducible representations of Sn' Returning to the case where F is arbitrary, we have to decide when Sf3 is contained in Sf3 -L.
7.1.10 Then
LEMMA.
Suppose that the partition j3 has exactly
Zj
parts equal to j.
(i) for every pair of j3-polytabloids e/ and e/*, H7'= I z/ divides <1>( el' e/*), and (ii) for every j3-polytabloid el' there is a j3-polytabloid e/* such that <1>( el' e/*) = H~= I( Zj!)i.
Proof Since O! = 1, there is no problem about taking infinite products. Define an equivalence relation ~ on the set of j3-tabloids by {td~{tz} if and only if, for all i and j, i and j belong to the same row of {t d when i and j belong to the same row of {tz}. Informally, this simply says that we can go from {t d to {t z} by shuffling rows. The equivalence classes have size IIj'=l z/, N ow, if {t d is involved in e/ and {t d ~ {t z}, then the definition of a polytabloid shows that {tz} is involved in e/, and whether the coefficients (which are ± 1) are the same or have opposite signs depends only on {t,} and {tz}. Therefore, any two j3-polytabloids have a multiple of IIj=lzJ! tabloids in common, and part (i) follows. Next, let t be any j3-tableau, and obtain t* from t by reversing the order of the numbers in each row of t. For example,
if
1
2
t= 5
6 9
8 11
3 7 10
4
4
then
t*= 7
10 11
3 6 9
2 5 8
Let wE V' have the property that for every i, i and wi belong to rows of t
299
7.1 Specht Modules
which have the same length. (In the example, 7T can be any permutation in the group generated by (5,8), (6,9), (7,10).) Then {7Tt} is involved in e/ and in e/*, with the same coefficient in each. It is easy to see that all tabloids common to e/ and e/* have this form. Therefore, lI>(e/.e/*)=Ilj=l(zj!»)' which proves part (ii). • 7.1.11 COROLLARY. Suppose that SfJ is defined over a field of characteristic p. Then SfJI(SfJ n SfJ 1- ) is nonzero if and only if 13 is p-regular.
Proof By definition, 13 is p-regular if and only if for every j~ 1. Zj (el' e/*)'1"'O for the f3-polytabloids e/ and e t* in part (ii) of the lemma. Therefore, SfJ g; SfJ 1- . If 13 is p-singular, lI>( el' e t*) = for every pair of f3-polytabloids, by part (i) • of the lemma. Therefore, SfJ ~ SfJ 1- .
°
7.1.12 COROLLARY. Suppose that SfJ is defined over a field of characteristic p, and 13 is p-regular. Then, for every f3-tableau t, there is a 7TESn such that Cif/7Te t is a nonzero multiple of el'
Proof By part (ii) of the lemma, given t, we can choose 7T such that
lI>(el'7Te/)=nj;,I(Z)!)l. But Cif t7Te t is a multiple of el' by 7.1.5(ii); say
Cif/7Te/ =he/. Now, h = h lI> ( { t }, e/ ) = lI> ( { t }, he/ ) = lI> ( { t }, Cif t7Te/ ) 00
=lI>(Cift{t},7Te/)=lI>(el'7Te t )= Since 13 is p- regular, h
II (z)!Y· )=1
* 0.
•
We now give SfJI(SfJ n SfJ 1-) a name: 7.1.13. Assume that char F= p (prime or DfJ: =SfJI(SfJ ns fJ 1-).
= (0) and
that 13 is p-regular. Let
The main theorem of this section is: 7.1.14 THEOREM. Suppose that F is an arbitrary field, and char F= p (prime or = 00). Then (i) As 13 varies over p-regular partitions of n, DfJ varies over a complete set
of ineqUivalent irreducible FSn -modules. (ii) DfJ is self-dual and absolutely irreducible. (iii) All the composition factors of sfJ n sfJ 1- and of MfJI sfJ are isomorphic to modules DO. with aC> 13.
300
Proof
Representation Theory of Sn over an Arbitrary Field 7.1
Suppose that some composition factor of Mf3 is isomorphic to D a .
D ais a homomorphic image of sa, so there is a nonzero FSn -homomorphism 8 from sa into a quotient of Mf3. By Corollary 7.1.12, there exist a-tableaux t
and t* such that CV·/e/. =hel' where h=l=O. But
7.1.15 Since 8 =1= 0, 7.1.5(i) gives a~ /3. N ow, if D a c::=. Df3, then Mf3 has a composition factor isomorphic to D a . Therefore, a~/3. Similarly, /3~a, so a=/3. But 6.1.2 shows that the number of inequivalent modules Df3 equals the number of absolutely irreducible FSn-modules. We have already proved that Df3 is absolutely irreducible, so the proof of part (i) is complete. Now, Df3 = Sf3j( Sf3 n Sf3 1. ) c::=. (Sf3 + Sf3 1. )1 Sf3 1., by the second isomorphism theorem, which in turn c::=. dual of Sf3I( Sf3 n Sf3 1.), since always A I B c::=. dual of B 1. I A 1.. Therefore, Df3 is self-duaL which finishes part (ii). Next, 7.1.15 shows that the only possible composition factors of Mf3 have the form D a with a~/3. Furthermore, if A is an FSn-submodule of Mf3 and Df3 c::=. Mf3I A then there is a homomorphism from Sf3 to Mf3j A:
and 7.I.I5 and 7.1.5(ii) prove that the image of this homomorphism is (Sf3 +A)IA. Therefore no composition factor of Mf3jS/3 or of S/3 ns f3 1. is isomorphic to Df3. and this completes the proof of part (iii). • Compare the ease with which we have the next result, with the struggle to get 6.3.50: 7.1.16
COROLLARY. The matrix recording the composition factors of Specht modules over a field of characteristic p has the form D#
(f3 p·regular )
0
{
S#
(/3 p·regular)
S#
(/3 p'singular ) {
*
when the p-regular partitions are placed in lexicographic order before all the p-singular partitions. Proof
Recall that Df3: = S/3I( S/3
n Sf3 1. ), and use part (iii) of the theorem.
•
301
7.2 The Standard Basis of the Specht Module
We have carefully avoided saying that the above matrix is the decomposition matrix of Sn' although we shall see in the next section that this is indeed the case. The problem is that we are not yet certain that Sf3 defined over the field of p elements is the reduction modulo p of Sf3 defined over the rational field.
7.2
The Standard Basis of the Specht Module
Our objective in this section is to prove that
{e( It is a standard .a-tableau} is a basis for Sf3, whatever field we work over. We start by finding elements in the group algebra FSn which annihilate a given polytabloid e(. There are some elements of FSn which obviously annihilate e p namely 1- (sgn 7T)7T
7.2.1
annihilates e(
Proof 'Iff =(sgn7T)7T'If(, so (l-(sgn7T)7T)'Y({t}=(I-(sgn7T)7T)e r =0.
•
Now let X be a subset of the ith column, and Y be a subset of the jth column of t, with i
Gx ,y: = ~ (sgnae )oe' e= 1
G x, y is called a Garnir element for XU Y. The permutations a I' . , . , ak are, of course, not unique, but for practical purposes note that we may take aI' ... , ak so that a It, . ,. ,ok t are all the tableaux which agree with t except in the positions occupied by XU Y, and whose entries increase vertically downwards in the positions occupied by XU Y. 7.2.2
EXAMPLE.
If I
t= 4 5 then we may take
1 4 5
2 3,
X= {4,
al!' ... ,ak !
2 3,
I 3 5
2 4,
5}
and
Y= {2,
3} ,
to be
1 3 4
2 5,
1 2 5
3 4,
1 2 4
3 5,
1 2 3
4 5
302
Representation Theory of Sn over an Arbitrary Field 7.2
With this choice of coset representatives, we have G x. y = 1- (34) + (354) + (234) - (2354) + (24 )(35). 7.2.3
THEOREM. Let t be a f3-tableau, X a subset of the i th column of t, and Y a subset of the j th column of t. with i f3: (where f3' is the partitio~ associated with f3), then
Gx,yet=O. Proof Write
s;s;
for
~{(sgn(J)(JI(JESxXSy},
for
~{(sgn(J)(JI(JESXuy}·
and SXUY
Since xu Y!>f3:, for every 'TE V t some pair of numbers in same row of 'Tt. Hence 1
xu Yare in the
Therefore,
Now, S; S; is a factor of 'Y t and
Therefore,
O=Sxu y''Ij't{t} =
1
XI!I YI!G x . y'Yt{t}.
Thus, Gx y'Yt{t} =0 when the field is the field of rational numbers. But all the tabloi'd coefficients in G x. yC'\f t{t} are integers in this special case, and so G x. y'Y t{ t} = 0, whatever the field. • In fact, the left ideal of FSn annihilating e t is generated by the elements of the group algebra described in 7.2.1, together with all Garnir elements for XU Y, with X, Y ranging over subsets of adjacent pairs of columns of t (say X a subset of the i th and Ya subset of the (i + 1)th) with the property that 1xu Y 1= f3: + 1; but we do not require this result.
303
7.2 The Standard Basis of the Specht Module
7.2.4 EXAMPLE. If 1 t= 4 5
and t I'
... , t 6
2
3
are the six tableaux listed in Example 7.2.2, then
so
In this way, we have written e( (where t is not standard) as a linear combination of polytabloids e( for which t i is standard. Our plan is to do this in general. ' Remembering that a tableau t is standard if the numbers increase along the rows and down the columns of t, we define e( to be a standard polytabloid if t is standard. Now totally order the set of j1-tabloids by saying 7.2.5. {tl}<{tz} if for some i
(i) when»i,) is in the same row of {td and {tz}, and (ii) i is in a higher row of {tl} than {tz}. For example, 3 I
4 2
5<2 I
4 3
5< I 2
4 3
5<2 1
3 4
5< I 2
3 4
5
<1 3
2 4
5<2 1
3 5
4<1 2
3 5
4<1 3
2 5
4<1 4
2 5
3
It is clear (d. Example 7.1.2) that 7.2.6. If the numbers increase down the columns of t, then {t} is the last tabloid involved in e r
The basis of Sf3 described in the next theorem is called the standard basis of Sf3. 7.2.7 THEOREM. {e( It is a standard j1-tableau} is a basis for sf3.
304
Representation Theory of Sn over an Arbitrary Field 7.2
Proof By 7.2.6, the standard polytabloids are linearly independent. It is therefore sufficient to prove that any polytabloid can be written as a linear combination of standard polytabloids. Let [t] denote the column equivalence class of the tableau t; that is, [t]={t,lt j =m for some ?TEV t }. Totally order the set of column equivalence classes in a way similar to the total order on the set of tabloids. Suppose that t is not standard. By induction, we may assume that e( can be written as a linear combination of standard polytabloids when [t']<[t]. We want to prove the same result for et • (There is no need to "get the induction started", since we are assuming nothing when there is no t' with [t']<[t].) Since ?Te t = (sgn?T)e t when?TE vt, we may suppose that the entries in t are in increasing order down columns. Since t is not standard, some adjacent pair of columns, say the i th and (i + I)th columns, have entries a, bq for some q:
Thus, if D is an ordinary representation of a group G, and we take W(I) to be a G-module affording D, then 8.3.6. W li is a G-module affording the symmetrized representation D c:::J (al.
Returning to the case where F is arbitrary, note that 8.3.7 THEOREM. The dimension of wli/(W a n W d ) equals the rank of the Gram matrix with respect to the semistandard basis of w a. 8.3.3, combined with the technique used in the proof of 7.1.8, now gives:
8.3.8 THEOREM. Suppose that a~ n has at most m nonzero parts (so that W a -=1= 0). W a/ (W a n wa 1- ) is an absolutely irreducible UF-module. If IF I;;;. n + 2, then Wli/( W li n W d ) is an absolutely irreducible GL( m, F)-module.
337
8.3 Irreducible GL( m, F)-Modules over F
In Exercise 8.4, we outline how to construct all the irreducible FGL(m, F) modules when F is finite. 8.3.9
COROLLARY.
Won) is an absolutely irreducible UF-module (if n";;;;m).
Proof The Gram matrix with respect to the semistandard basis of
the identity matrix. w"/(W"
n W"-l.) need not be an irreducible
n
W(t
)
is •
GL(m, F)-module when F
is small. 8.3.10 EXAMPLE. Let m = 2. The semistandard basis of W(3) is {w 1w twl' W 2 W 2W 2 ' x, y}, where
as in Example 8.2.15. The Gram matrix for this basis is
3
J
When char F= 2, W(3) is therefore an irreducible UF-module, but we have seen that it is not an irreducible G L(2, F)-module when IF I = 2. The Gram matrix proves that the dimension of W(3)/(W(3) n W(3)-l.) is 2 if char F= 3 and m = 2. We recommend that the reader verifies directly that W(3) n W(3)-l., the space spanned by x and y, i~.PF-invariant in this case. We now prove that no two of the irreducible Urmodules we have found are equivalent. To do this, the elements (~i) of UF are brought into play. An elementary property of binomial coefficients is 8.3.11 This works even for a> x if we let (
Now, for
a;;;'
x-a). = (x-a )(x-a-l) ... (x-a+ I-b) b . b!
0, define
338
Representations of General Linear Groups 8.3
as follows. Let
Assuming inductively that ( hib-a) has been defined for O,,;;,b";;'c-I, let
Here, of course,
( c~
b)
means
I+ I+
... + I
(( c ~
b) times)
E F.
8.2.5 and 8.3.11 give: 8.3.12. If N; of the subscripts ii' i 2' ... , in equal i, then (N-a) wl8l···l8lw. ,c wl8l···l8lw=' ( h-a) I] III C I] 1"
8.3.13 LEMMA.
-a) w 181 ... 181
h ( ,C
II
w =
'n
0 for all c ~ 1
if and only if precisely a of the subscripts it, i 2' ... , in are equal to i. Proof Let N, of the subscripts equal i. Then h-a) wl8l···l8lw=' (N-a) wl8l···l8lw. ( ,c I" C 1] til I]
If N; =a, then forall
c~l.
If Ni >a, then for
c=N, -a.
8.3 Irreducible GL( m, F)-Modules over F
339
When c= 1, this equals N -a, which is nonzero if char F=O. On the other I
hand if char F=p and p divides (a- N; :c-l ) for every c with 1 ~c~aN;, then p divides
8.3.14
(:= ~ ),
a contradiction. (Consider Pascal's triangle.)
•
Let af-n', {3f-n. Suppose that A is a UF-submodule of Lfl and that there exists a nonzero Urhomomorphism 0 from W" into Lfl/ A. Then n=n' and {3~a. Furthermore, ifa={3, then ImO=(W"+A)/A. LEMMA,
Proof Choose a basis 1"/ 2 ,,,, of Lfl in such a way that each Ij has the form 'yflwT; for some {3-tableau ~. Let O('I["w") =bJl +
... +bJ, + A
We may assume that the coset representative bt/ t + ... +bJr is such that no element of A has the form xIII + ... + xJ, with x; EF, x r *0. If the content of T, is not a, then there exists an i such that the tensors W i ,Q9 ••• Q9 Win involved in I, do not have precisely a; subscripts equal to i. By Lemma 8.3.13, for some c;;;'l, the element
-a) -
u:= ( h ; c ; EUF
has the property that ulr =Y,lr with O*Y, EF. But 8.3.12 shows that ul; =Jjlj (Y; EF) for each), and by Lemma 8.3.13 again, uO(V"w")=O. Therefore, when we apply u to our expression for O('\("w"), we deduce that b,y,/, + ... + b,yJrEA, in contradiction to our choice of coset representative. Thus, the content of T, is a. But unless n = n' and {3~a, there is no {3-tableau T, of content a such that ('I,(fl WT *0 (see 8.1.10). This proves the first part of the lemma. When a = {3, the ~nly nonzero vectors of the form '\("w T where T is a {3-tableau of content a are ± 'Y"w". Therefore, in this case, r= 1 and O('Y"w")=bl'Y"w" + A, and from this the last statement of the lemma follows.
•
If a has at most m nonzero parts, define
F": = W"/( W" n W«-.L ).
340
Representations of General Linear Groups 8.3
This is both a UF-module and a GL(m, F)-module; it is an irreducible UF-module. 8.3.15 THEOREM. If F" is a composition factor of LfJ as a UF-module, then 0: and {3 are partitions of the same integer and {3r;:;;.0:. Furthermore, the multiplicity of F" as a composition factor of L" is one.
Proof Since F" is a homomorphic image of W", F" is a composition factor of LfJ if and only if there exists a nonzero UF-homomorphism from W" into a quotient of Lf3. Now apply the last lemma. • 8.3.16
COROLLARY.
(i) If F" and Ff3 are isomorphic as UF-modules, then 0: = {3. (ii) If F is infinite, and F" and Ff3 are isomorphic as GL(m, F)-modules, then 0:={3.
In particular, 8.3.17 COROLLARY. If charF=O, and W""'" WfJ (and not both are zero), then 0:={3.
When F is algebraically closed of characteristic zero, it can be proved that the various nonzero Weyl modules give all the irreducible integral representations of GL(m, F); to obtain all the irreducible rational representations, simply tensor Weyl modules with ~ -d, a negative integral power of the determinant module. The reader interested in constructing irreducible modules for SL(m, F) over F (where F is infinite) should note that an almost identical proof to that given shows that F" is an irreducible SL(m, F)-module. Our proof of inequivalence, though, used the elements
(h a) i;
of UF' which are not
available in the special-linear-group case. However, the reader should have no difficulty (except when o:f- n', {3r n, and n'
8.4 Further Connections between Specht and Weyl Modules
341
8.3.18 THEOREM. The matrix recording the UF composition factors of Weyl modules for partitions of n has the form
o
W"
*
if the partitions are written down in reverse lexicographic order. 8.3.19 EXAMPLE. If char F=3, the UF composition factors of Weyl modules for partitions of 3 are given by F1 1')
w(\J)
w(2,1) w(3)
[:
FI2,1)
F(3)
J
It is known [James, 1980] that the matrix appearing in Theorem 8.3.18 contains the decomposition matrix of Sn as a submatrix. To be explicit, let SfJ be the dual of SfJ, and when the partition {3' associated with {3 is p-regular, let jjfJ be the quotient of SfJ by its unique maximal FSn-submodule (see Exercise 7.1). Then when /3' is p-regular, the multiplicity of jjfJ as a composition factor of equals the multiplicity of FfJ as a composition factor of W a . Compare the matrix in Example 8.3.19 with the decomposition matrix of S3 for p = 3, written in terms of jja and Sa:
sa
8.4
Further Connections between Specht and Weyl Modules
We have already proved many results for Weyl modules which are analogous to those for Specht modules, and in this section we push the connection between these modules further, by examining the properties of the copy, FSnCVaw a, of the Specht module sa which lies in L(n).
342
Representations of General Linear Groups 8.4
Recall that M a is the space spanned by those wii~ ... Q9win where, for each i, precisely O:i of the subscripts are i. By 8.3.13, M a = {WlwELtnland
Proof Let A denote the right-hand side. We have already proved that sa \:;A.
Under the isomorphism 8.1.1, the total order 7.2.5 on the set of a-tabloids induces a total order on {w,,@ ... @wi,EM a }. Although our field is arbitrary, we may define a bilinear form cP on Ltnl as in 8.2.7 (in fact, the restriction of cP to M a is just the bilinear form used in Chapter 7). Now, suppose that W is a nonzero element of A. We claim that the last wi,@"'@wi,involvedin W is such that i"i 2 , ... ,i n is a lattice permutation. Suppose not. Then for some x and z, i z =x+ 1, and the number of subscripts before i z which equal x + 1 is the same as the number of subscripts before i z which equal x. Let i u,' i u,'"'' i u, be the subscripts after i z (excluding i z ) which equal x+ 1, and let i u ,+,,'''' i u , be the subscripts before i z which equal x. Theny=ax+1-1. Since e~.x+l/a! annihilates w, W is orthogonal (under CP, defined above) to all vectors of the form (e~+l,x/a!)w' with w'ELtnl. In particular, by Lemma 8.4.1, w is orthogonal to the sum v of all tensors wi ,@ ... @wi , in M a such that (i))k =i k for i k *x, x+ 1 and (ii) )u, = i u, for 1.;; r';;y. But wi,@ ... @wi, is the first tensor satisfying these conditions, and by hypothesis wi ,@ ... @wi , is the last tensor in w. This contradicts CP( v, w)=O. We have now proved that the last tensor involved in every nonzero element of A has its subscripts in a lattice permutation. Therefore, A = sa, by 8.1.2 and 7.2.10. • Compare the next corollary with the definition of the Weyl module. 8.4.3
COROLLARY.
344
Representations of General Linear Groups 8.4
Proof Since the action of Sn on L(n) commutes with that of UF, and'l["w" generates S" as an FS,cmodule, every element of S" is certainly annihilated by those elements in UF which annihilate Y"w". This gives one inclusion, and the other follows from the descriptions of M" and S" we have given above. •
Surprisingly, not only does tensor space contain Weyl modules and Specht modules, but it also contains copies of many vector spaces of module homomorphisms: 8.4.4 THEOREM. Suppose that a and there exist injective linear maps
(i) (ii) (iii) (iv)
from from from from
f3 have at most m nonzero parts. Then
W" ns fJ into HomFS
The third and fourth maps are isomorphisms, and so are the first and second unless char F= 2 and IX is 2-singular. Proof (i): Since (Y"w" generates W" as a UF-module, every element of W" n SfJ has the form uCV"w" for some u E UF. Let
Then 0 is well defined, since s,CV"w" =S2(1("W" implies that SI -S2 annihilates CV"w", and so annihilates uCV"w". Thus, we have associated a unique element of HomFS(S", SfJ) with each element of W" ns fJ . If
EXAMPLE.
Suppose char F= 2. Then
8.4 Fu~ther Connections b~tween Specht and Weyl Modules
345
spanned by w,0wZ -wZ0w], so w(l')ns(2)=o. Since M(Z)=S(2), this example simultaneously shows that the maps in parts (i) and (ii) of the theorem need not be isomorphisms when char F= 2 and a is 2-singular. Theorems 8.4.2 and 8.4.4 are very powerful when it comes to finding composition factors of Specht modules. To decide whether or not HomFdS a, SfJ) is zero, all we have to do (at least in the case char F~2 or a 2-regu!
(b)