William Fulton Joe Harris Representation Theory A First Course With 144 Illustrations Springer-Verlag New York Berlin Heidelberg London Paris Tokyo Hong Kong Barcelona BudapestWilliam Fulton Joe Harris Department of Mathematics Department of Mathematics University of Chicago Harvard University Chicago, IL 60637 Cambridge, MA 02138 USA USA Editorial Board JH. Ewing F.W. Gehring P.R. Halmos Department of Department of Department of Mathematics Mathematics Mathematics Indiana University University of Michigan University of Santa Clara Bloomington, Ann Arbor, Santa Clara, IN 47405 USA MI 48109 USA CA 95053 USA Mathematics Subject Classification: 20GOS, 17B10, 17B20, 22646 Library of Congress Cataloging-in-Publication Data Fulton, William, 1939- Representation theory: a first course / William Fulton and Joe Harris. p. cm—(Graduate texts in mathematics) Includes bibliographical references and index. 1. Representationsof groups. 2. Representations of algebras. 3. Lie groups. 4. Lie algebras. I. Harris, Joc. Il. Title IML, Series. QAITL.F85 199) 512.2—de20 90.24926 Printed on acid-free paper © 1991 Springer-Verlag New York Inc. All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer-Verlag New York, Inc, 175 Fifth Avenue, New York, NY 10010, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar of dissimilar methodology now known or hereafter developed is forbidden. The use of general descriptive names, trade names, trademarks, etc, in this pobifcafion, even if the former are not especially identified, is not to be taken asa sign that such names; as understood by the Trade Marks and Merchandise Marks Act, may accordingly be used freely by anyone. ‘Typeset by Asco Trade Typesetting, Ltd., Hong Kong. Printed and bound by R.R. Donnelley & Sons Co., Harrisonburg, VA. Printed in the United States of America. 987654325 ISBN 0-387-97495-4 Springer-Verlag New York Berlin Heidelberg (softeover) ISBN 3-540-97495-4 Springer-Verlag Berlin Heidelberg New York ISBN 0-387-97527-6 Springer-Verlag New York Berlin Heidelberg (hardcover) ISBN 3-540-97527-6 Springer-Verlag Berlin Heidelberg New YorkPreface The primary goal of these lectures is to introduce a beginner to the finite- dimensional representations of Lie groups and Lie algebras, Since this goal is shared by quite a few other books, we should explain in this Preface how our approach differs, although the potential reader can probably see this better by a quick browse through the book. Representation theory is simple to define: it is the study of the ways in which a given group may act on vector spaces. It is almost certainly unique, however, among such clearly delineated subjects, in the breadth of its interest to mathematicians. This is not surprising: group actions are ubiquitous in 20th century mathematics, and where the object on which a group acts is not a vector space, we have learned to replace it by one that is (¢.g., a cohomology group, tangent space, etc,). As a consequence, many mathematicians other than specialists in the field (or even those who think they might want to be) come in contact with the subject in various ways. It is for such people that this text is designed. To put it another way, we intend this as a book for beginners to learn from and not as a reference. This idea essentially determines the choice of material covered here. As simple as is the definition of representation theory given above, it fragments considerably when we try to get more specific. For a start, what kind of group G are we dealing with—a finite group like the symmetric group S, or the general linear group over a finite field GL,(F,), an infinite discrete group like SL,(Z), a Lie group like SL,C, or possibly a Lie group over a local field? Needless to say, each of these settings requires a substantially different approach to its representation theory. Likewise, what sort of vector space is G acting on: is it over C, R, Q, or possibly a field of finite characteristic? Is it finite dimensional or infinite dimensional, and if the latter, what additional structure(such as norm, or inner product) does it carry? Various combinationsvi Preface of answers to these questions lead to areas of intense research activity in representation theory, and it is natural for a text intended to prepare students for a career in the subject to lead up to one or more of these areas. As a corollary, such a book tends to get through the elementary material as quickly as possible: if one has a semester to get up to and through Harish—Chandra modules, there is little time to dawdle over the representations of G, and SL,C. By contrast, the present book focuses exactly on the simplest cases: repre- sentations of finite groups and Lie groups on finite-dimensional real and complex vector spaces, This is in some sense the common ground of the subject, the area that is the object of most of the interest in representation theory coming from outside. The intent of this book to serve nonspecialists likewise dictates to some degree our approach to the material we do cover. Probably the main feature of our presentation is that we concentrate on examples, developing the general theory sparingly, and then mainly as a useful and unifying language to describe phenomena already encountered in concrete cases. By the same token, we for the most part introduce theoretical notions when and where they ate useful for analyzing concrete situations, postponing as long as possible those notions that are used mainly for proving general theorems, Finally, our goal of making the book accessible to outsiders accounts in part for the style of the writing. These lectures have grown from courses of the second author in 1984 and 1987, and we have attempted to keep the informal styte of these lectures. Thus there is almost no attempt at efficiency: where it seems to make sense from a didactic point of view, we work out many special cases of an idea by hand before proving the genera! case; and we cheerfully give several proofs of one fact if we think they are illuminating. Similarly, while it is common to develop the whole semisimple story from one point of view, say that of compact groups, or Lie algebras, or algebraic groups, we have avoided this, as efficient as it may be. It is of course not a strikingly original notion that beginners can best learn about a subject by working through examples, with general machinery only introduced slowly and as the need arises, but it seems particularly appropriate here. In most subjects such an approach means one has a few: out of an unknown infinity of examples which are useful to illuminate fhe general situation. When the subject is the representation theory of complex semisimple Lie groups and algebras, however, something special happens: once one has worked through all the examples readily at hand— the “classical” cases of the special linear, orthogonal, and symplectic groups—one has not just a few useful examples, one has all but five “exceptional” cases. This is essentially what we do here. We start with a quick tour through representation theory of finite groups, with emphasis determined by what is useful for Lie groups. In this regard, we include more on the symmetric groups than is usual. Then we turn to Lie groups and Lie algebras. After some preliminaries and a look at low-dimensional examples, and one lecture withPreface vii some general notions about semisimplicity, we get to the heart of the course: working out the finite-dimensional representations of the classical groups. For each series of classical Lie algebras we prove the fundamental existence theorem for representations of given highest weight by explicit construction. Our object, however, is not just existence, but to see the representations in action, to see geometric implications of decompositions of naturally occurring representations, and to see the relations among them caused by coincidences between the Lie algebras. The goat of the fast six lectures is to make a bridge between the example- oriented approach of the earlier parts and the general theory. Here we make an attempt to interpret what has gone before in abstract terms, trying to make connections with modern terminology. We develop the general theory enough to see that we have studied all the simple complex Lie algebras with five exceptions. Since these are encountered less frequently than the classical series, it is probably not reasonable in a first course to work out their representations as explicitly, although we do carry this out for one of them, We also prove the general Weyl character formula, which can be used to verily and extend many of the results we worked out by hand earlier in the book. Of course, the point we reach hardly touches the current state of affairs in Lie theory, but we hope it is enough to keep the reader’s eyes from glazing over when confronted with a lecture that begins: “Let G be a semisimple Lie group, P a parabolic subgroup, ...” We might also hope that working through this book would prepare some readers to appreciate the elegance (and efficiency) of the abstract approach. In spirit this book is probably closer to Weyl's classic [Wet] than to others written today. Indeed, a secondary goal of our book is to present many of the results of Weyl and his predecessors in a form more accessible to modern readers. In particular, we include Weyl’s constructions of the representations of the general and special linear groups by using Young's syminetrizers; and we invoke a little invariant theory to do the corresponding result for the orthogonal and symplectic groups. We also include Weyl's formulas for the characters of these representations in terms of the elementary characters of symmetric powers of the standard representations. (Interestingly, Weyl only gave the corresponding formulas in terms of the exterior powers for the general linear group. The corresponding formulas for the orthogonal and symplectic groups were only given recently by Koike and Terada. We include a simple new proof of these determinantal formulas.) More about individual sections can be found in the introductions to other parts of the book. Needless to say, a price is paid for the inefficiency and restricted focus of these notes. The most obvious is a lot of omitted material: for example, we include little on the basic topological, differentiable, or analytic properties of Lie groups, as this plays a small role in our story and is well covered in dozens of other sources, including many graduate texts on manifolds. Moreover, there are no infinite-dimensional representations, no Harish-Chandra or Vermaviii Preface modules, no Steifel diagrams, no Lie algebra cohomology, no analysis on symmetric spaces or groups, no arithmetic groups or automorphic forms, and nothing about representations in characteristic p > 0. There is no consistent attempt to indicate which of our results on Lie groups apply more generally to algebraic groups over fields other than R or C (e.g,, local fields). And there is only passing mention of other standard topics, such as universal enveloping algebras or Bruhat decompositions, which have become standard tools of representation theory. (Experts who saw drafts of this book agreed that some topic we omitted must not be left out of a modern book on representation theory—but no two experts suggested the same topic.) We have not tried to trace the history of the subjects treated, or assign credit, or to attribute ideas to original sources—this is far beyond our knowl- edge. When we give references, we have simply tried to send the reader to sources that are as readable as possible for one knowing what is written here. A good systematic reference for the finite-group material, including proofs of the results we leave out, is Serre [Se2]. For Lie groups and Lie algebras, Serre [Se3], Adams [Ad], Humphreys [Hut], and Bourbaki [Bour] are recommended references, as are the classics Weyl [Wel] and Littlewood [Litt]. We would like to thank the many people who have contributed ideas and suggestions for this manuscript, among them J-F. Burnol, R. Bryant, J. Carrell, B. Conrad, P. Diaconis, D. Eisenbud, D. Goldstein, M. Green, P. Griffiths, B. Gross, M. Hildebrand, R. Howe, H. Kraft, A. Landman, B. Mazur, N. Chriss, D. Petersen, G. Schwartz, J. Towber, and L. Tu. In particular, we would like to thank David Mumford, from whom we learned much of what we know about the subject, and whose ideas are very much in evidence in this book. Had this book been written 10 years ago, we would at this point thank the people who typed it. That being no longer applicable, perhaps we should thank instead the National Science Foundation, the University of Chicago, and Harvard University for generously providing the various Macintoshes on which this manuscript was produced. Finally, we thank Chan Fulton for making the drawings. Bill Fulton and Joe HarrisUsing This Book A few words are in order about the practical use of this book. To begin with, prerequisites are minimal: we assume only a basic knowledge of standard first-year graduate material in algebra and topology, including basic notions about manifolds. A good undergraduate background should be more than enough for most of the text; some examples and exercises, and some of the discussion in Part IV may refer to more advanced topics, but these can readily be skipped. Probably the main practical requirement is a good working knowledge of multilinear algebra, including tensor, exterior, and symmetric products of finite dimensional vector spaces, for which Appendix B may help. We have indicated, in introductory remarks to each lecture, when any back- ground beyond this is assumed and how essential it is. For a course, this book could be used in two ways. First, there are a number of topics that are not logically essential to the rest of the book and that can be skimmed or skipped entirely. For example, in a minimal reading one could skip §§4, 5, 6, 11.3, 13.4, 15.3-15.5, 17.3, 19.5, 20, 22.1, 22.3, 23.3-23.4, 25.3, and 26.2; this might be suitable for a basic one-semester course. On the other hand, in a year-long course it should be possible to work through as much of the material as background and/or interest suggested. Most of the material in the Appendices is relevant only to such a long course. Again, we have tried to indicate, in the introductory remarks in each lecture, which topics are inessential and may be omitted. Another aspect of the book that readers may want to approach in different ways is the profusion of examples. These are put in largely for didactic reasons: we feel that this is the sort of material that can best be understood by gaining some direct hands-on experience with the objects involved. For the most part, however, they do not actually develop new ideas, the reader whose tastes run more to the abstract and general than the concrete and special may skip manyx Using This Book of them without logical consequence. (Of course, such a reader will probably wind up burning this book anyway.) We include hundreds of exercises, of wildly different purposes and difficulties. Some are the usual sorts of variations of the examples in the text or are straightforward verifications of facts needed; a student will probably want to attempt most of these. Sometimes an exercise is inserted whose solution is a special case of something we do in the text later, if we think working on it will be useful motivation (again, there is no attempt at “efficiency,” and readers are encouraged to go back to old exercises from time to time). Many exercises are included that indicate some further directions or new topics (or standard topics we have omitted); a beginner may best be advised to skim these for general information, perhaps working out a few simple cases. In exercises, we tried to include topics that may be hard for nonexperts to extract from the literature, especially the older literature. In general, much of the theory is in the exercises—and most of the examples in the text. We have resisted the idea of grading the exercises by (expected) difficulty, although a “problem” is probably harder than an “exercise.” Many exercises are starred: the + is not an indication of difficulty, but means that the reader can find some information about it in the section “Hints, Answers, and References” at the back of the book. This may be a hint, a statement of the answer, a complete solution, a reference to where more can be found, or a combination of any of these. We hope these miscellaneous remarks, as haphazard and uneven as they are, will be of some use.Contents Preface Using This Book Part I: Finite Groups 1. Representations of Finite Groups §1.1: Definitions §1.2: Complete Reducibility; Schur's Lemma §1.3: Examples: Abelian Groups; S5 2. Characters §2.1: Characters §2.2: The First Projection Formula and ts Consequences §2.3: Examples: S, and &, §2.4: More Projection Formulas; More Consequences 3. Examples; Induced Representations; Group Algebras; Real Representations : Examples: S, and 9%, Exterior Powers of the Standard Representation of S, Induced Representations The Group Algebra : Real Representations and Representations over Subfields of C one Ww 12 15 18 21 26 26 3t 32. 36 39xii . Representations of S,: Young Diagrams and Frobenius’s Character Formula §4.1: Statements of the Results §4.2: Irreducible Representations of S, §4.3: Proof of Frobenius’s Formula . Representations of Mt, and GL,(F,) §5.1: Representations of Wf, §5.2: Representations of GL,(F,) and SL2(F,) . Weyl’s Construction §6.1: Schur Functors and Their Characters §6.2: The Proofs Part 1: Lie Groups and Lie Algebras |. Lie Groups §7.1: Lie Groups: Definitions §7.2: Examples of Lie Groups §7.3: Two Constructions . Lie Algebras and Lie Groups §8.1: Lie Algebras: Motivation and Definition §8.2: Examples of Lie Algebras §8.3: The Exponential Map . Initial Classification of Lie Algebras $9.1: Rough Classification of Lie Algebras : Engel's Theorem and Lie's Theorem Semisimple Lie Algebras |: Simple Lie Algebras ). Lie Algebras in Dimensions One, Two, and Three $10.1: Dimensions One and Two §10.2: Dimension Three, Rank | §10.3: Dimension Three, Rank 2 §10.4: Dimension Three, Rank 3 . Representations of s1,C §11.1: The Irreducible Representations §11.2: A Little Plethysm $11.3: A Little Geometric Plethysm Contents 63 67 15 15 93 93 101 104 i 14 121 121 125 128 138 133 133 136 139 141 146 146 151 153Contents 12. Representations of s{,€, Part I 13, 18. Representations of s[,C, Part Il: Mainly Lots of Examples §13.t: Examples §13.2: Description of the irreducible Representations §13.3: A Little More Plethysm §13.4: A Little More Geometric Plethysm Part Ill: The Classical Lie Algebras and Their Representations . The General Set-up: Analyzing the Structure and Representations of an Arbitrary Semisimple Lie Algebra §14.f: Analyzing Simple Lie Algebras in General §14.2: About the Killing Form . sl, C and sf,C §15.1: Analyzing s1,C §15.2: Representations of sIgC and s1,C §15.3: Weyl's Construction and Tensor Products §15.4: Some More Geometry §15.5: Representations of GL,C . Symplectic Lie Algebras §16.1: The Structure of Spz,C and sp2,C §16.2: Representations of sp,C . spe€ and sp, §17.1: Representations of sp,C §17.2: Representations of sp,,€ in General §17.3: Wey!’s Construction for Symplectic Groups Orthogonal Lie Algebras §18.1: SO,C and £0, $18.2: Representations of so,C, s0,€, and s0,C . 606C, 80,C, and so,,C §19.1: Representations of s0,C §19.2: Representations of the Even Orthogonal Algebras §19.3: Representations of 0,C §19.4: Representations of the Odd Orthogonal Algebras §19.5: Weyl's Construction for Orthogonal Groups xiii t6l 175 11S 182 185 189 195 197 197 206 211 2in 217 222 227 231 238 238 244 253 253 259 262 267 267 273 282 282 286 292 294 296xiv Contents 20. Spin Representations of so,,C 299 §20.1: Clifford Algebras and Spin Representations of s0,,C 299 §20.2: The Spin Groups Spin,,C and Spin,,R 307 §20.3: SpingC and Triality 312 Part IV: Lie Theory 317 21. The Classification of Complex Simple Lie Algebras 319 §21.1: Dynkin Diagrams Associated to Semisimple Lie Algebras 319 §21.2: Classifying Dynkin Diagrams 325 §21.3: Recovering a Lie Algebra from Its Dynkin Diagram 330 22. g and Other Exceptional Lie Algebras 339 §22.t: Construction of g, from Its Dynkin Diagram 339 §22.2: Verifying That g, is a Lie Algebra 346 §22.3: Representations of g 350 §22.4: Algebraic Constructions of the Exceptional Lie Algebras 359 23, Complex Lie Groups; Characters 366 §23.1: Representations of Complex Simple Groups 366 §23.2: Representation Rings and Characters 375 §23.3: Homogeneous Spaces 382 §23.4: Bruhat Decompositions 395 24. Weyl Character Formula 399 §24.1: The Weyl Character Formula 399 §24.2: Applications to Classical Lie Algebras and Groups 403 25. More Character Formulas 41s §25,.1: Freudenthal's Multiplicity Formula 415s §25.2: Proof of (WCF); the Kostant Multiplicity Formula 419 §25.3: Tensor Products and Restrictions to Subgroups 424 26, Real Lie Algebras and Lie Groups 430 §26.1: Classification of Real Simple Lie Algebras and Groups 430 §26.2: Second Proof of Weyl'’s Character Formula 440 §26.3: Real, Complex, and Quaternionic Representations 444 Appendices 45h A. On Symmetric Functions 453 §A.1: Basic Symmetric Polynomials and Relations among Them. 453 §A.2: Proofs of the Determinantal Identities 462 §A.3: Other Determinantal Edentities 465Contents B. On Multilinear Algebra e t: Tensor Products Exterior and Symmetric Powers o 3: Duals and Contractions C. On Semisimplicity §C.1: The Killing Form and Cartan’s Criterion §C.2: Complete Reducibility and the Jordan Decomposition §C.3: On Derivations D. Cartan Subalgebras §D.1: The Existence of Cartan Subalgebras §D.2: On the Structure of Semisimple Lie Algebras §D.3: The Conjugacy of Cartan Subalgebras §D.4: On the Weyl Group E, Ado's and Levi's Theorems §E.1: Levi's Theorem §E.2: Ado’s Theorem F. Invariant Theory for the Classical Groups §F.1: The Polynomial Invariants §F.2: Applications to Symplectic and Orthogonal Groups §F.3; Proof of Capelli Identity Hints, Answers, and References Bibliography Index of Symbols Index xv 4 4 472 475 478 478 481 483 487 487 489 491 493 499 499 Sul 5i4 516 536 543 547PART I FINITE GROUPS Given that over three-quarters of this book is devoted to the representation theory of Lie groups and Lie algebras, why have a discussion of the represen- tations of finite groups at all? There are certainly valid reasons from a logical point of view: many of the ideas, concepts, and constructions we will introduce here will be applied in the study of Lie groups and algebras. The real reason for us, however, is didactic, as we will now try to explain. Representation theory is very much a 20th-century subject, in the following sense. In the 19th century, when groups were dealt with they were generally understood to be subsets of the permutations of a set, or of the automor- phisms GL(V) of a vector space V, closed under composition and inverse. Only in the 20th century was the notion of an abstract group given, making it possible to make a distinction between properties of the abstract group and properties of the particular realization as a subgroup of a permutation group or GL(V). To give an analogy, in the 19th century a manifold was always a subset of R"; only in the 20th century did the notion ofan abstract Riemannian manifold become common. In both cases, the introduction of the abstract object made a fundamental difference to the subject. In differential geometry, one could make a crucial distinction between the intrinsic and extrinsic geometry of the manifold: which properties were invariants of the metric on the manifold and which were properties of the particular embedding in R". Questions of existence or non- existence, for example, could be broken up into two parts: did the abstract manifold exist, and could it be embedded, Similarly, what would have been called in the 19th century simply “group theory” is now factored into two parts. First, there is the study of the structure of abstract groups (eg., the classification of simple groups). Second is the companion question: given a group G, how can we describe all the ways in which G may be embedded in2 I. Finite Groups (or mapped to) a linear group GL(V)?. This, of course, is the subject matter of representation theory. Given this point of view, it makes sense when first introducing representa- tion theory to do so in a context where the nature of the groups G in question is itself simple, and relatively well understood. It is largely for this reason that we are starting off with the representation theory of finite groups: for those readers who are not already familiar with the motivations and goals of representation theory, it seemed better to establish those first in a setting where the structure of the groups was not itself an issue. When we analyze, for example, the representations of the symmetric and alternating groups on 3, 4, and 5 letters, it can be expected that the reader is already familiar with the groups and can focus on the basic concepts of representation theory being introduced. We will spend the first six lectures on the case of finite groups. Many of the techniques developed for finite groups will carry over to Lie groups; indeed, our choice of topics is in part guided by this. For example, we spend quite a bit of time on the symmetric group; this is partly for its own interest, but also partly because what we learn here gives one way to study representations of the general linear group and its subgroups. There are other topics, such as the alternating group %,, and the groups SL,(F,) and GL,(F,) that are studied purely for their own interest and do not appear later. (In general, for those readers primarily concerned with Lie theory, we have tried to indicate in the introductory notes to each lecture which ideas will be useful in the succeeding parts of this book.) Nonetheless, this is by no means a comprehensive treat- ment of the representation theory of finite groups; many important topics, such as the Artin and Brauer theorems and the whole subject of modular representations, are omitted.LECTURE 1 Representations of Finite Groups In this lecture we give the basic definitions of representation theory, and prove two of the basic results, showing that every representation is a (unique) direct sum of itreduc- ible ones, We work out as examples the case of abelian groups, and the simplest nonabelian group, the symmetric group on 3 letters, In the latter case we give an analysis that will turn out not to be useful for the study of finite groups, but whose main idea is central to the study of the representations of Lie groups. §1.1: Definitions §1.2: Complete reducibility; Schur’s lemma §1.3: Examples: Abelian groups; S3 §1.1. Definitions A representation of a fi group G on a finite-dimensional complex vector space V is a homomorphism p: G + GL(V) of G to the group of automor- phisms of V; we say that such a map gives V the structure of a G-module. When there is little ambiguity about the map p (and, we're afraid, even sometimes when there is) we sometimes call V itself a representation of G; in this vein we will often suppress the symbol p and write gv or gu for p(g)(v). The dimension of V is sometimes called the degree of p. A map ¢ between two representations V and W of G is a vector space map g: V > W such that = fy set < —4 1, Representations of Finite Groups commutes for every g € G. (We will call this a G-linear map when we want to distinguish it from an arbitrary linear map between the vector spaces V and W,) We can then define Ker g, Im g, and Coker g, which are also G-modules, A subrepresentation of a representation V is a vector subspace W of V which is invariant under G. A representation V is called irreducible if there is no proper nonzero invariant subspace W of V. If Vand Ware representations, the direct sum V © W and the tensor product V @ W are also representations, the latter via a(v @ w) = gu @gw. For a representation V, the nth tensor power V®" is again a representation of G by this rule, and the exterior powers A"(V) and symmetric powers Sym"(V) are subrepresentations! of it. The dual V* = Hom(V, C) of V is also a repre- sentation, though not in the most obvious way: we want the two representa- tions of G to respect the natural pairing (denoted ¢ , >) between V* and V, so that if g: G+ GL(V) is a representation and p*: G + GL(V*) is the dual, we should have <*(g)(0*), a(g)(0)) = Aut(X) is a homomorphism to the permutation group of X, there is an associated per- mutation representation: let V be the vector space with basis {e,: x eX}, and let G act on V by GY Als = Ye gx The regular representation, denoted Rg or R, corresponds to the left action of G on itself. Alternatively, R is the space of complex-valued functions on G, where an element g € G acts on a function a by (ga)(h) = a(g7'h). Exercise 1.4*. (a) Verify that these two descriptions of R agree, by identifying the element e, with the characteristic function which takes the value ! on x, 0 on other elements of G. (b) The space of functions on G can also be made into a G-module by the rule (gx)(h) = a(hg). Show that this is an isomorphic representation. §1.2. Complete Reducibility; Schur’s Lemma As in any study, before we begin our attempt to classify the representations of a finite group G in earnest we should try to simplify life by restricting our search somewhat. Specifically, we have seen that representations of G can be6 1. Representations of Finite Groups built up out of other representations by linear algebraic operations, most simply by taking the direct sum. We should focus, then, on representations that are “atomic” with respect to this operation, i.c., that cannot be expressed as a direct sum of others; the usual term for such a representation is Inde- composable. Happily, the situation is as nice as it could possibly be: a repre- sentation is atomic in this sense if and only if it is irreducible (ie. contains no Proper subrepresentations); and every representation is the direct sum of irreducibles, in a suitable sense uniquely so. The key to all this is Proposition 1.5. If W is a subrepresentation of a representation V of a finite group G, then there is a complementary invariant subspace W' of V, so that v=wew'. Proor. There are two ways of doing this, One can introduce a (positive definite) Hermitian inner product H on V which is preserved by each g eG (ie. such that H(ge, gw) = H(v, w) for all v, w € V and g € G). Indeed, if Ho is any Hermitian product on ¥, one gets such an H by averaging over G: H(v, w) = x Ho(g», gw). se Then the perpendicular subspace W* is complementary to W in V. Alterna- tively (but similarly), we can simply choose an arbitrary subspace U comple- mentary to W, let 19: V-» W be the projection given by the direct sum decomposition V = W @ U, and average the map % over G: that is, take (0) = Y) g(to(g"'»)). ae This will then be a G-linear map from V onto W, which is multiplication by {G| on W; its kernel will, therefore, be a subspace of V invariant under G and complementary to W. oO Corollary 1.6. Any representation is a direct sum of irreducible representations. This property is called complete reducibility, or semisimplicity. We will see that, for continuous representations, the circle S', or any compact group, has this property; integration over the group (with respect to an invariant measure on the group) plays the role of averaging in the above proof. The (additive) group R does not have this property: the representation «(5 ‘) leaves the x axis fixed, but there is no complementary subspace. We will see other Lie groups such as SL,(C) that are semisimple in this sense. Note also that this argument would fail if the vector space V was over a field of finite characteristicsince it might then be the case that x(v) = O for ve W. The failure§1.2, Complete Reducibility; Schur’s Lemma 7 of complete reducibility is one of the things that makes the subject of modular representations, or representations on vector spaces over finite fields, so tricky. The extent to which the decomposition of an arbitrary representation into a direct sum of irreducible ones is unique is one of the consequences of the following: Schur’s Lemma 1.7. If V and W are irreducible representations of G and gp: V > W is a G-module homomorphism, then (1) Either ¢ is an isomorphism, or p = 0. (2) If V = W, then p =A-I for some Le C, I the identity. Proor. The first claim follows from the fact that Ker g and Im ¢ are invariant subspaces. For the second, since C is algebraically closed, ¢ must have an eigenvalue A, ie., for some A eC, y — AI has a nonzero kernel. By (1), then, we must have — al = 0, s0 9 = A. Qo We can summarize what we have shown so far in Proposition 1.8. For any representation V of a finite group G, there is a decomposition V = VO" @--@ 0%, where the V, are distinct irreducible representations. The decomposition of V. into a direct sum of the k factors is unique, as are the V, that occur and their multiplicities a,. Proor. It follows from Schur’s lemma that if W is another representation of G, with a decomposition W = ® W,%", and @: V > W is a map of represen- tations, then @ must map the factor V, into that factor Wj°™ for which W, = V,;; when applied to the identity map of V to V, the stated uniqueness follows. o In the next lecture we will give a formula for the projection of V onto HW, The decomposition of the ith summand into a direct sum of a; copies of V, is not unique if a, > 1, however. Occasionally the decomposition is written V=aV, @-Oah =a +--+ 4h, (1.9) especially when one is concerned only about the isomorphistn classes and multiplicities of the yj. One more fact that will be established in the following lecture is that a finite group G admits only finitely many irreducible representations 1, up to iso- morphism (in fact, we will say how many). This, then, is the framework of the classification of all representations of G: by the above, once we have described8 1. Representations of Finite Groups the irreducible representations of G, we will be able to describe an arbitrary representation as a linear combination of these. Our first goal, in analyzing the representations of any group, will therefore be: (i) Describe all the irreducible representations of G. Once we have done this, there remains the problem of carrying out in practice the description of a given representation in these terms. Thus, our second goal will be: (ii) Find techniques for giving the direct sum decomposition (1.9), and in particular determining the multiplicities a, of an arbitrary representation V. Finally, it is the case that the representations we will most often be concerned with are those arising from simpler ones by the sort of linear- or multilinear- algebraic operations described above. We would like, therefore, to be able to describe, in the terms above, the representation we get when we perform these operations on a known representation. This is known generally as (iii) Plethysm: Describe the decompositions, with multiplicities, of represen- tations derived from a given representation V, such as V@V, V*, N(V), Sym*{V), and A(A'V). Note that if V decomposes into a sum of two represen- tations, these representations decompose accordingly; e.g., if V = U ® W, then ke i AV DP rvenm so it is enough to work out this plethysm for irreducible representations. Similarly, if V and W are two irreducible representations, we want to decom- pose V ® W; this is usually known as the Clebsch-Gordon problem. §1.3. Examples: Abelian Groups; S3 One obvious place to look for examples is with abelian groups. It does not take long, however, to deal with this case. Basically, we may observe in general that if V is a representation of the finite group G, abelian or not, each ge G gives a map p(g): V > V; but this map is not generally a G-modile homomor- phism: for general h € G we will have : g(h(o)) # h(g(o)). Indeed, p(g): V + V will be G-linear for every p if (and only if) g is in the center Z(G) of G. In particular if G is abelian, and V is an irreducible representation, then by Schur’s lemma every element g € G acts on V by a scalar multiple of the identity. Every subspace of V is thus invariant; so that V must be one dimensional. The irreducible representations of an abelian group G are thus simply elements of the dual group, that is, homomorphisms p:Gac%§1.3. Examples: Abelian Groups; G5, 9 We consider next the simplest nonabelian group, G = G3. To begin with, we have (as with any symmetric group) two one-dimensional representations; we have the trivial representation, which we will denote U, and the alternating representation U', defined by setting gv = sgn(g)v for g€ G, ve C. Next, since G comes to us as a permutation group, we have a natural permutation representation, in which G acts on C* by permuting the coordinates. Explicitly, if {e,, e2, ¢3} is the standard basis, then g-e, = egy, or, equivalently, . 9 (24s 2ay 2a) = (y-1¢1y9 29-1 ay 29° y)- This representation, like any permutation representation, is not irreducible: the line spanned by the sum (1, 1, !) of the basis vectors is invariant, with complementary subspace V = {(2), 22, 25) € C7: 2, + 22 + 23 = O}. This two-dimensional representation V is easily seen to be irreducible; we call it the standard representation of S,. Let us now turn to the problem of describing an arbitrary representation of G,. We will see in the next lecture a wonderful tool for doing this, called character theory; but, as inefficient as this may be, we would like here to adopt a more ad hoc approach. This has some virtues as a didactic technique in the present context (admittedly dubious ones, consisting mainly of making the point that there are other and far worse ways of doing things than character theory). The real reason we are doing it is that it will serve to introduce an idea that, while superfluous for analyzing the representations of finite groups in general, will prove to be the key to understanding representations of Lie groups. The idea is a very simple one: since we have just seen that the representation theory of a finite abelian group is virtually trivial, we will start our analysis of an arbitrary representation W of ©, by looking just at the action of the abelian subgroup 2%, = Z/3 c S, on W. This yields a very simple decom- Position: if we take t to be any generator of 2, (that is, any three-cycle), the space W is spanned by eigenvectors v, for the action of t, whose eigenvalues are of course all powers of a cube root of unity w = e7""3. Thus, w= Oh, where V=Cv, and ry, = wv, Next, we ask how the remaining elements of G, act on I in terms of this decomposition. To see how this goes, let o be any transposition, so that t and o together generate S3, with the relation oto = 17, We want to know where o sends an eigenvector v for the action of t, say with eigenvalue w'; to answer10 1. Representations of Finite Groups this, we look at how t acts on o(v). We use the basic relation above to write t(o(v)) = 0(2?(0)) = 0(w"' 0) = w'-o(v). The conclusion, then, is that if v is an eigenvector for t with eigenvalue w', then o(v) is again an eigenvector for t, with eigenvalue w”', Exercise 1.10. Verify that with ¢ = (12), r = (123), the standard representation has a basis « = (w, 1, ”), 8 = (1, w, w”), with ta=or, th=o"f, on=f, of=a Suppose now that we start with such an eigenvector v for t. If the eigenvalue of vis w! ¥ 1, then o(v) is an eigenvector with eigenvalue w” ¥ w', and so is independent of v; and vy and o(v) together span a two-dimensional subspace V’ of W invariant under G3. In fact, V’ is isomorphic to the standard repre- sentation, which follows from Exercise 1.10. If, on the other hand, the eigen- value of v is 1, then o(v) may or may not be independent of v. If it is not, then v spans a one-dimensional subrepresentation of W, isomorphic to the trivial representation if o(v) = v and to the alternating representation if o(v) = —v. Ifo(v) and v are independent, then v + o(v) and v — o(v) span one-dimensional representations of W isomorphic to the trivial and alternating representations, respectively. We have thus accomplished the first two of the goals we have set for ourselves above in the case of the group G = S,. First, we see from the above that the only three irreducible representations of S, are the trivial, alternating, and standard representations U, U' and V. Moreover, for an arbitrary repre- sentation W of G, we can write W =U QU Ove; and we have a way to determine the multiplicities a, b, and c: c, for example, is the number of independent eigenvectors for t with eigenvalue w, whereas a + cis the multiplicity of 1 as an eigenvalue of g, and b + c is the multiplicity of —1 as an eigenvalue of o. ‘ In fact, this approach gives us as well the answer to our third problem, finding the decomposition of the symmetric, alternating, or tensor powers of a given representation W, since if we know the eigenvalues of t on such a representation, we know the eigenvalues of t on the various tensor powers of W. For example, we can use this method to decompose V @ V, where V is the standard two-dimensional representation. For V @ V is spanned by the vectors @a, c@f, P@a, and P@ f; these are eigenvectors for t with eigenvalues ?, 1, 1, and , respectively, and o interchanges «@a with B® p, and « @ f with f @ a. Thus « @ a and fp @ f span a subrepresentation$1.3. Examples: Abelian Groups; S it isomorphic to V, a@ f+ #@« spans a trivial representation U, and %@ fp — 8 @a spans U’,so V@V=UGU'OV. Exercise 1.11. Use this approach to find the decomposition of the represen- tations Sym?V and Sym3V. Exercise 1.12. (a) Decompose the regular representation R of G,. (b) Show that Sym***V is isomorphic to Sym*V¥ @R, and compute Sym*V for all k. Exercise 1.13%. Show that Sym?(Sym?V) = Sym*(Sym?V). Is Sym™(Sym"V) isomorphic to Sym"(Sym"V)? As we have indicated, the idea of studying a representation V of a group G - by first restricting the action to an abelian subgroup, getting a decomposition of V into one-dimensional invariant subspaces, and then asking how the remaining generators of the group act on these subspaces, does not work well for finite G in general; for one thing, there will not in general be a convenient abelian subgroup to use. This idea will turn out, however, to be the key to understanding the representations of Lie groups, with a torus subgroup playing the role of the cyclic subgroup in this example. Exercise 1.14*. Let V be an irreducible representation of the finite group G. Show that, up to scalars, there is a unique Hermitian inner product on V preserved by G.LECTURE 2 Characters This lecture contains the heart of our treatment of the representation theory of finite groups: the definition in §2.L of the character of a representation, and the main theorem (proved in two steps in §2.2 and §2.4) that the characters of the irreducible representa- tions form an orthonormat basis for the space of class functions on G. There will be more exaniples and more constructions in the following lectures, but this is what you need to know. : Characters The first projection formula and its consequences Examples: G, and Wy : More projection formulas; more consequences §2.1. Characters As we indicated in the preceding section, there is a remarkably effective tool for understanding the representations of a finite group G, called character theory. This is in some ways motivated by the example worked out in the last section where we saw that a representation of S,'was determined by knowing the eigenvalues of the action of the elements t and ¢ € G3. Fora general group G, it is not clear what subgroups and/or elements should play the role of %,, t, and o; but the example certainly suggests that knowing all the eigenvalues of each element of G should suffice to describe the representation. Of course, specifying all the eigenvalues of the action of each element of G is somewhat unwieldy; but fortunately it is redundant as well. For example, if we know the eigenvalues {2,} of an element g € G, then of course we know the eigenvalues {Af} of g* for each k as well. We can thus use this redundancy§2.1. Characters 13 to simplify the data we have to specify. The key observation here is it is enough to give, for example, just the sum of the eigenvalues of each element of G, since knowing the sums ) Af of the kth powers of the eigenvalues of a given element g € G is equivalent to knowing the eigenvalues {A,} of g themselves. This then suggests the following: Definition. If V is a representation of G, its character yy is the complex-valued function on the group defined by xv(9) = Tr(glv), the trace of g on V. In particular, we have ayligh™) = x(a so that y, is constant on the conjugacy classes of G; such a function is called a class function. Note that y,(1) = dim V. Proposition 2.1. Let V and W be representations of G. Then Xvow = Xv + kw, Xvew = Xv hws w= Ky and xpry(9) = 2Lav(9)? — xv(97)}- Proor. We compute the values of these characters on a fixed element g € G. For the action of g, V has eigenvalues {2,} and W has eigenvalues {,1,}. Then {a + 4g} and {A,- uj} are eigenvalues for V ® W and V @ W, from which the first two formulas follow. Similarly {A;' = /,} are the eigenvalues for gon V*, since all eigenvalues are nth roots of unity, with the order of g. Finally, {A, Ali < j} are the eigenvalues for g on A?¥, and (AY - Ya, Ban ee and since g? has eigenvalues {47}, the last formula follows. oa Exercise 2.2. For Sym?V, verify that Asymav(9) = $0xv(g)? + xv(9)}- Note that this is compatible with the decomposition V@V =Sym’V @/NV. Exercise 2.3*. Compute the characters of Sym*V and A‘V. Exercise 2.4*. Show that if we know the character x, of a representation V, then we know the eigenvalues of each clement g of G, in the sense that we14 2. Characters know the coefficients of the characteristic polynomial of g: V — V. Carry this out explicitly for elements g € G of orders 2, 3, and 4, and for a representation of G ona vector space of dimension 2, 3, or 4. Exercise 2.5. (The original fixed-point formula). If V is the permutation repre- sentation associated to the action of a group G on a finite set X, show that X0(g) is the number of elements of X fixed by g. As we have said, the character of a representation of a group G is really a function on the set of conjugacy classes in G. This suggests expressing the basic information about the irreducible representations of a group G in the form of a character table. This is a table with the conjugacy classes [g] of G listed across the top, usually given by a representative g, with (for reasons that will become apparent later) the number of elements in each conjugacy class over it; the irreducible representations V of G listed on the left; and, in the appro- priate box, the value of the character x, on the conjugacy class [g]. Example 2.6. We compute the character table of G3. This is easy: to begin with, the trivial representation takes the values (1, 1, 1) on the three conjugacy classes [1], [(12)], and [(123)], whereas the alternating representation has values (1, —1, 1). To see the character of the standard representation, note that the permutation representation decomposes: C? = U @ V; since the character of the permutation representation has, by Exercise 2.5, the values (3, 1,0), we have xy = x¢9 — xy = (3, 1, 0) — (1, 1, 1) = (2, 0, —1). In sum, then, the character table of S, is I 3 2 (12) (123) trivial U alternating U’ standard V This gives us another solution of the basic problem posed in Lecture 1: if W is any representation of S, and we decompose W into irreducible repre- sentations W = U®"@ U'® @ V®, then Zw = aXy + bxy- + cxy- In particu- lar, since the functions yy, yy: and yy are independent, we see that W is determined up to isomorphism by its character Yy. Consider, for example, V @ V. Its character is (y,)*, which has values 4, 0, and 1 on the three conjugacy classes. Since V ® U @ U’ has the same char- acter, this implies that V @ V decomposes into V ® U @ U’, as we have seen directly. Similarly, V @ U’ has values 2,0, and —1,s0 V@U’ = ¥V.§2.2. The First Projection Formula and Its Consequences 15 Exercise 2.7*. Find the decomposition of the representation V®" using char- acter theory. Characters will be similarly useful for larger groups, although it is rare to find simple closed formulas for decomposing tensor products. §2.2. The First Projection Formula and Its Consequences In the last lecture, we asked (among other things) for a way of locating explicitly the direct sum factors in the decomposition of a representation into irreducible ones. In this section we will start by giving an explicit formula for the projection ofan irreducible representation onto the direct sum of the trivial factors in this decomposition; as it will turn out, this formula alone has tremendous consequences. To start, for any representation V of a group G, we set VS ={veV:gu=0 VgeG}. We ask for a way of finding V° explicitly. The idea behind our solution to this is already implicit in the previous lecture. We observed there that for any representation V of G and any geG, the endomorphism g: V — V is, in general, not a G-module homomorphism. On the other hand, if we take the average of all these endomorphisms, that is, we set 1 == € End(V) o> igi 06 Ea” then the endomorphism ¢ will be G-linear since )g = Yhgh™'. In fact, we have Proposition 2.8. The map ¢ is a projection of V onto V®. PRrooF. First, suppose v = g(w) = (1/|G|) ). gw. Then, for any he G, 1 bo = Tq L how = so the image of @ is contained in V°. Conversely, if ve VS, then p(v) = (IG) Sv = v, 80 V9 & Im(o}; and g og =o. a We thus have a way of finding explicitly the direct sum of the trivial subrepresentations of a given representation, although the formula can be hard to use if it does not simplify. If we just want to know the number m of copies of the trivial representation appearing in the decomposition of V, we can do this numerically, since this number will be just the trace of the16 2. Characters projection p. We have m= dim VS = Tae ad 2, Trace(q) = ae Y tv(o. (29) In particular, we observe that for an irreducible representation V other than the trivial one, the sum over all g € G of the values of the character yy is zero. We can do much more with this idea, however. The key is to use Exercise 1.2: if V and W are representations of G, then with Hom(V, W), the representa- tion defined in Lecture I, we have Hoin(V, W)$ = {G-module homomorphisms from V to 1}. If V is irreducible then by Schur’s lemma dim Hom(V, W)° is the multiplicity of Vin W; similarly, if W is irreducible, dim Hom(V, W)° is the multiplicity of W in V, and in the case where both V and W are irreducible, we have a 1 ifvVew dim Hom,(V, W) = { tveW. But now the character Yyjonqv, sv) Of the representation Hom(V, W) = V* @ W is given by Anomey 199) = XrLG)- Xow 9)- We can now apply formula (2.9) in this case to obtain the striking l itvew iis XW(@xw(9) = {i avew. (2.10) To express this, let Coes(G) = {class functions on G} and define an Hermitian inner Prowse on C iass(G) by (a, B) = x a(g) f(g). (2.11) ii lee Formula (2.10) then amounts to i Theorem 2.12, In terms of this inner product, the characters of the irreductble representations of G are orthonormal. For example, the orthonormality of the three irreducible representations of G, can be read from its character table in Example 2.6. The numbers over each conjugacy class tell how many times to count entries in that column. Corollary 2.13. The number of irreducible representations of G is less than or equal to the number of conjugacy classes.§2.2. The First Projection Formula and Its Consequences 17 We will soon show that there are no nonzero class functions orthogonal to the characters, so that equality holds in Corollary 2.13. Corollary 2.14. Any representation is determined by its character. Indeed if V = V2" @ --- @ V2%, with the V; distinct irreducible characters, then xy = YqXy,, and the yy, are linearly independent. Corollary 2.15, A representation V is irreducible if and only if (x,. xy) = 1. In fact, if V = Vi" +: @® HE* as above, then (xy, Xv) =), a7. The multiplicities a, can be calculated via Corollary 2.16. The multiplicity a, of V; in V is the inner product of xy with xy,, ie, a) = (Xv Xv,)- We obtain some further corollaries by applying all this to the regular representation R of G. First, by Exercise 2.5 we know the character of R; it is simply 0 ifg#e XR(g) = 1a itgee. Thus, we see first of all that R is not irreducible if G # {e}. In fact, if we set R= @vVe, with V, distinct irreducibles, then 1 . 4 = (Xv Xn) = jaz Il = dim h. (2.17) Corollary 2.18. Any irreducible representation V of G appears in the regular representation dim V times. In particular, this proves again that there are only finitely many irreducible representations. As a numerical consequence of this we have the formula 1G| = dim(R) = ¥ dim(v,). {2.19) 7 Also, applying this to the value of the character of the regular representation on an element g € G other than the identity, we have 0=Lidim Y)-x,(9) ifg #e. (2.20) These two formulas amount to the Fourier inversion formula for finite groups, cf. Example 3,32. For example, if all but one of the characters is known, they give a formula for the unknown character. Exercise 2.21. The orthogonality of the rows of the character table is equiv- alent to an orthogonality for the columns (assuming the fact that there are as18 2. Characters many rows as columns). Written out, this says: (@) For geG, where the sum is over all irreducible characters, and c(g) is the number of elements in the conjugacy class of g. (ii) If g and ht are elements of G that are not conjugate, then D xox) = 0. x Note that for g = e these reduce to (2.19) and (2.20). §2.3. Examples: G, and 2, To see how the analysis of the characters of a group actually goes in practice, we now work out the character table of S,. To start, we list the conjugacy classes in G, and the number of elements of S, in each. As with any symmetric group G,, the conjugacy classes correspond naturally to the partitions of d, that is, expressions of d as a sum of positive integers a), a2, ..., 4, Where the correspondence associates to such a partition the conjugacy class of a permutation consisting of disjoint cycles of length a,, a2, ..., a,- Thus, in Sy we have the classes of the identity element 1 (4=1+ 1+ 1+ 1), a trans- position such as (12), corresponding to the partition 4 = 2 + 1 + 1; a three- cycle (123) corresponding to 4 = 3 + 1; a four-cycle (1234) (4 = 4); and the product of two disjoint transpositions (12)(34) (4 = 2 + 2). Exercise 2,22. Show that the number of elements in each of these conjugacy classes is, respectively, 1, 6, 8, 6, and 3. As for the irreducible representations of G,, we start with the same ones that we had in the case of S,: the trivial U, the alternating U', and the standard representation V, ic., the quotient of the permutation representation associated to the standard action of S, on a set of four elements by the trivial subrepresentation. The character of the trivial representation on the five conjugacy classes is of course (1, 1, 1, 1, 1), and that of the alternating representation is (1, —1, 1, —1, 1). To find the character of the standard representation, we observe that by Exercise 2.5 the character of the permuta- tion representation on C* is xc+ = (4, 2, 1, 0, 0) and, correspondingly, kv = Xe — ty = GB, 1,0, —1, — 1). Note that |zy| = |, so V is irreducible. The character table so far looks like§2.3, Examples: G, and %, 19 1 6 8 6 3 SG, | t (12) (123) (1234) (12)(34) trivial U 1 1 1 I 1 alternating U’ t -t 1 -t ' standard V | 3 1 0 -1 -1 Clearly, we are not done yet: since the sum of the squares of the dimensions of these three representations is 1 + 1+9 = 11, by (2.19) there must be additional irreducible representations of G,, the squares of whose dimensions add up to 24 — 11 = 13, Since there are by Corollary 2.13 at most two of them, there must be exactly two, of dimensions 2 and 3. The latter of these is easy to locate: if we just tensor the standard representation V with the alternating one U', we arrive at a representation V' with character xy- = Yy"Xv- = (3, —1, 0, 1, — 1). We can see that this is irreducible either from its character {since |yy-| = 1) or from the fact that it is the tensor product of an irreducible representation with a one-dimensional one; since its character is not equal to that of any of the first three, this must be one of the two missing ones. As for the remaining representation of degree two, we will for now simply call it W; we can determine its character from the orthogonality relations (2.10). We obtain then the complete character table for S,: 1 6 8 6 3 (12) (123) (1234) (12)(34) 1 trivial U 1 ' alternating U" I -t 1 -! 1 standard V 3 1 0 -t -t Veveu | 3 -t 0 1 -1 Another W 2 0 -t 0 2 Exercise 2.23. Verify the last row of this table from (2.10) or (2-20). We now get a dividend: we can take the character of the mystery represen- tation W, which we have obtained from general character theory alone, and use it to describe the representation W explicitly! The key is the 2 in the last column for yy: this says that the action of (12)(34) on the two-dimensional vector space W is an involution of trace 2, and so must be the identity. Thus, W is really a representation of the quotient group! "ICN isa normal subgroup of a group G, a representation p: G + GL(P) is trivial on N if and only if it factors through the quotient G— G/N + GLY). Representations of G/N can be identified with representations of G that are trivial on N.20 2. Characters 4/1, (12)(34), (13)(24), (14)(23)} = So. [One may see this isomorphism by letting S, act on the elements of the conjugacy class of (12)(34);, equivalently, if we realize G, as the group of rigid motions of a cube (see below), by looking at the action of S, on pairs of opposite faces.] W must then be just the standard representation of G3 pulled back to S, via this quotient. Example 2.24. As we said above, the group of rigid motions of a cube is the symmetric group on four letters; S, acts on the cube via its action on the four long diagonals. It follows, of course, that G, acts as well on the set of faces, of edges, of vertices, etc.; and to each of these is associated a permutation representation of S,. We may thus ask how these representations decompose; we will do here the case of the faces and leave the others as exercises. We start, of course, by describing the character x of the permutation representation associated to the faces of the cube. Rotation by 180° about a line joining the midpoints of two opposite edges is a transposition in G, and fixes no faces, so y(12) = 0. Rotation by 120° about a long diagonal shows x(123) = 0. Rotation by 90° about a line joining the midpoints of two opposite faces shows (1234) = 2, and rotation by 180° gives x((12)(34)) = 2. Now (x% x) = 3, so x is the sum of three distinct irreducible representations, From the table, (x, xv) = (% %v-) = 0 Xw) = 1, and the inner products with the others are zero, so this representation is U @ V’ @ W. In fact, the sums of opposite faces span a three-dimensional subrepresentation which contains U (spanned by the sum of all faces), so this representation is U @ W. The differences of opposite faces therefore span V'. Exercise 2.25*. Decompose the permutation representation of S, on (i) the vertices and (ii) the edges of the cube. Exercise 2.26. The alternating group , has four conjugacy classes. Three fepresentations U, U', and U" come from the representations of W/L, (12}(34), (13)(24), (14)(23)} & 2/3, so there is one more irreducible representation V of dimension 3. Compute the character table, with wo = e?*: t 4 4 3 (123) (132) (1234)§2.4. More Projection Formulas; More Consequences 21 Exercise 2.27. Consider the representations of S, and their restrictions to %,. Which are still irreducible when restricted, and which decompose? Which pairs of nonisomorphic representations of S, become isomorphic when restricted? Which representations of 9, arise as restrictions from ©,? §2.4. More Projection Formulas; More Consequences In this section, we complete the analysis of the characters of the irreducible Tepresentations of a general finite group begun in §2.2 and give a more general formula for the projection of a general representation V onto the direct sum of the factors in V isomorphic to a given irreducible representation W. The main idea for both is a generalization of the “averaging” of the endomorphisms g: V + V used in §2.2, the point being that instead of simply averaging all the g we can ask the question: what linear combinations of the endomorphisms g: V > V are G-linear endomorphisms? The answer is given by Proposition 2.28. Let «: G+ C be any function on the group G, and for any representation V of G set Gay = Lag) gi VV. Then q,y is a homomorphism of G-modules for all V if and only if is a class Junction. Proor, We simply write out the condition that ¢,,y be G-linear, and the result falls out: we have a.v{hv) = a(g)- (hv) = Ya(hgh"!): hgh-*(he) (substituting hgh"! for g) = WY a(hgh “*)-g(o)) = MX a(a)-ote) (if @ is a class function) = h(g,.v(0)). Exercise 2.29*. Complete this proof by showing that conversely if « is not a class function, then there exists a representation V of G for which @, , fails to be G-linear. Qa As an immediate consequence of this proposition, we have22 2. Characters Proposition 2.30. The number of irreducible representations of G is equal to the number of conjugacy classes of G. Equivalently, their characters {xy} form an orthonormal basis for C .tas5(G). Proor. Suppose a: G + Cis a class function and (a, xy) = 0 for all irreducible representations V; we must show that a = 0. Consider the endomorphism Pav = ¥,Ug)-g: VV as defined above. By Schur’s lemma, ¢,,, = 4-Id; and ifn = dim V, then 1 d= a trace(,,) 1 = Lala G = ud (a, xv*) =0. Thus, 9,,) = 0, or Y.a(g)g = 0 on any representation V of G; in particular, this will be true for the regular representation V = R. But in R the elements {gq € G}, thought of as elements of End(R), are linearly independent. For example, the elements {g(e)} are all independent. Thus «(g) = 0 for all g, as required. o This proposition completes the description of the characters of a finite group in general. We will see in more examples below how we can use this information to build up the character table of a given group. For now, we mention another way of expressing this proposition, via the representation ring of the group G. The representation ring R(G) of a group G is easy to define. First, as a group we just take R(G) to be the free abelian group generated by all (isomorphism classes of) representations of G, and mod out by the subgroup generated by elements of the form V + W — (V @® W). Equivalently, given the statement of complete reducibility, we can just take all integral linear combinations ) a, V, of the irreducible representations V, of G; elements of R(G) are correspondingly called virtual representations. The ring structure is then given simply by tensor product, defined on the generators of R(G) and extended by linearity. We can express most of what we have learned so far about representations of a finite group G in these terms. To begin, the character defines a map 1 RG) > CorassG) from R(G) to the ring of complex-valued functions on G; by the basic formulas of Proposition 2.1, this map is in fact a ring homomorphism. The statement that a representationin determined by its character then says that x is injective;§2.4. More Projection Formulas; More Consequences 23 the images of y are called virtual characters and correspond thereby to virtual representations. Finally, our last proposition amounts to the statement that x induces an isomorphism Xe: RIG) @C — Cotass(G). The virtual characters of G form a lattice A = Z° in Cyy,44(G), in which the actual characters sit as a cone Ag = N‘c Z*. We can thus think of the problem of describing the characters of G as having two parts: first, we have to find A, and then the cone Ag < A (once we know Ag, the characters of the irreducible representations will be determined). In the following lecture we will state theorems of Artin and Brauer characterizing A @ Q and A. The argument for Proposition 2.30 also suggests how to obtain a more general projection formula. Explicitly, if W is a fixed irreducible representation, then for any representation V, look at the weighted sum a = : V). v ial oe xw(g)-9 € End(V). By Proposition 2.28, p is a G-module homomorphism. Hence, if V is irreduc- ible, we have y = A-Id, and 1 = dmv Trace p mV a Lxwl)- x9) 1 -Jdmy *7=" 0 itvew. For arbitrary V, . 1 — by = dim We YS twlg) gi VV (2.31) IG| sec is the projection of V onto the factor consisting of the sum of all copies of W appearing in V. In other words, if V = om then 1, = dim Y- ia. x XvAd-9 (2.32) is the projection of V onto V;°", Exercise 2,33*, (a) In terms of representations V and W in R(G), the inner product on C.,,55(G) takes the simple form (V, W) = dim Homg(V, W).24 2. Characters (b) Ify € C,,,,,(G) is a virtual character, and (x, x) = 1, then either x or —y is the character of an irreducible representation, the plus sign occurring when x1) > 0. If (x, x) = 2, and x(1) > 0, then x is either the sum or the difference of two irreducible characters. () If U, V, and W are irreducible representations, show that U appears in V @ W if and only if W occurs in V* @ U. Deduce that this cannot occur unless dim U > dim W/dim V. We conclude this lecture with some exercises that use characters to work out some standard facts about representations. Exercise 2.34*. Let V and W be irreducible representations of G, and Lo: V > W any linear mapping. Define L: V-> W by t Livy) = — Lo(g-o). () = gb, 9 Lolo») Show that L = 0 if V and W are not isomorphic, and that L is multiplication by trace(Lg)/dim(V) if V = W. Exercise 2.35*. Show that, if the irreducible representations of G are represented by unitary matrices [cf. Exercise 1.14], the matrix entries of these representa- tions form an orthogonal basis for the space of all functions on G [with inner product given by (2.11)J- Exercise 2.36*. If G, and G, are groups, and V, and V, are representations of G, and G,, then the tensor product V, @ V, is a representation of G, x G2, by (91 * go)-(¥; @ 02) = gy" 0, @ Gz" v. To distinguish this “external” tensor product from the internal tensor product—when G, = G,—this external tensor product is sometimes denoted V, @ V,. If y, is the character of V,, then the value of the character y of V, @ V, is given by the product: 9s % G2) = x1(91)x2(g2)- If V, and V, are irreducible, show that V, @ V; is also irreducible and show that every irreducible representation of G, x G, arises this way. In terms of representation rings, RG, x G2) = RG.) ® RG,). In these lectures we will often be given a subgroup G of a general linear group GL(V), and we will look for other representations inside tensor powers of V. The following problem, which is a theorem of Burnside arid Molien, shows that for a finite group G, all irreducible representations can be found this way.§2.4. More Projection Formulas; More Consequences 25 Problem 2.37*. Show that if V is a faithful representation of G, ie. p: G > GL(V) is injective, then any irreducible representation of G is contained in some tensor power V®" of ¥, Problem 2.38*. Show that the dimension of an irreducible representation of G divides the order of G. Another challenge: Problem 2.39*. Show that the character of any irreducible representation of dimension greater than 1 assumes the value 0 on some conjugacy class of the group.LECTURE 3 Examples; Induced Representations; Group Algebras; Real Representations This lecture is something of a grabbag. We start in §3.1 with examples illustrating the use of the techniques of the preceding lecture. Section 3.2 is also by way of an example. We will sce quite a bit more about the representations of the symmetric groups in general later; §4 is devoted to this and will certainly subsume this discussion, but this should provide at least a sense of how we can go about analyzing representations of a class of groups, as opposed to individual groups. In §§3.3 and 3.4 we introduce two basic notions in representation theory, induced representations and the group algebra. Finally, in §3.5 we show how to classify representations of a finite group on a real vector space, given the answer to the corresponding question over C, and say a few words about the analogous question for subfields of C other than R. Everything in this lecture is elementary except Exercises 3.9 and 3.32, which involve the notions of Clifford algebras and the Fourier transform, respectively (both exercises, of course, can be skipped). §3.1: Examples: Sx and Ws §3.2: Exterior powers of the standard representation of S, §3.3: Induced representations §3.4: The group algebra §3.5: Real representations and representations over subfields of C / §3.1. Examples: G, and 2, We have found the representations of the symmetric and alternating groups for n < 4, Before turning to a more systematic study of symmetric and alter- nating groups, we will work out the next couple of cases.§3.1. Examples: , and %, un Representations of the Symmetric Group S, As before, we start by listing the conjugacy classes of S, and giving the number of elements of each: we have 10 transpositions, 20 three-cycles, 30 four-cycles and 24 five-cycles; in addition, we have 15 elements conjugate to (12)(34) and 10 elements conjugate to (12)(345). As for the irreducible representations, we have, of course, the trivial representation U, the alternating representation U’, and the standard representation V; also, as in the case of S, we can tensor the standard representation V with the alternating one to obtain another irreducible representation V’ with character xy. = xy “Xu. Exercise 3.1. Find the characters of the representations V and V’; deduce in particular that V and V’ are distinct irreducible representations. The first four rows of the character table are thus 1 10 20 30 24 15 20 Ss 1 (12) (123) (1234) (12345) (12)(34) (12)(345) U 1 1 1 1 1 1 1 u 1 -t 1 -t 1 1 -t Vv 4 2 1 0 -t 0 -1 v' 4 —2 1 0 -t 0 1 Clearly, we need three more irreducible representations. Where should we look for these? On the basis of our previous experience (and Problem 2.37), a natural place would be in the tensor products/powers of the irreducible representations we have found so far, in particular in V @ V (the other two possible products will yield nothing new: we have V'@ V = V@V @U' and V’'@ V' = V®@ PV). Of course, V@ V breaks up into A?V and Sym?V, so we look at these separately. To start with, by the formula Aawla) = Harta)? — xr(97)) we calculate the character of A?V: Any = (6, 0, 0, 0, 1, —2, 0); we see from this that it is indeed a fifth irreducible representation (and that NV @U' = AV, so we get nothing new that way). We can now find the remaining two representations in either of two ways. First, ifn, and n, are their dimensions, we have St = 120 = 17 4 17 447 4+ 4? + 6? +n} +n}, so n? + n3 = 50. There are no more one-dimensional representations, since these are trivial on normal subgroups whose quotient group is cyclic, and U,28 3. Examples; Induced Representations; Group Algebras; Real Representations is the only such subgroup. So the only possibility isn, =n = 5. Let W denote one of these five-dimensional representations, and set W’ = W@ U’. In the table, if the row giving the character of W is (5 ay a2 3 Oy Os 6), that of W'is(5 —a, a, —a@3 a, as ~—d). Using the orthogonality relations or (2.20), one sees that W’ 4 W; and with a little calculation, up to interchanging W and W’, the last two rows are as given: 1 10 20 30 24 15 20 Ss 1 (12) (123) (1234) (12345) (12)(34) (12)(345) U 1 1 1 1 1 1 1 u 1 -t J 1 1 4 -t vj4 2 1 0 =-1 0 1 ve 4 -2 1 0 -1 0 I Nv 6 0 0 0 1 -2 0 Ww 5 1 -t 1 0 I 1 w 5 -t -l 1 0 1 1 From the decomposition V ® U = C5, we have also A4V = ASC = U’, and V* = V. The perfect pairing! VxvoNVe=U, taking v x (v, A v2 A v3) tov A vy A 02 A v3 Shows that AV is isomorphic toV* QU =V. Another way to find the representations W and W’ would be to proceed with our original plan, and look at the representation Sym?V, We will leave this in the form of an exercise: Exercise 3.2. (i) Find the character of the representation Sym7V. (ii) Without using any knowledge of the character table of S,, use this to show that Sym?V is the direct sum of three distinct irreducible representations. (iii) Using our knowledge of the first five rows of the character table, show that Sym?V is the direct sum of the representations U, V, and a third irreduc- ible representation W. Complete the character table for S.. Exercise 3.3, Find the decomposition into irreducibles of the representations. AW, Sym?W, and V@ W. “IE V and W are n-dimensional vector spaces, and U is one dimensional, a perfect pairing is a bilinear map fi: V x WU such that no nonzero vector v in V has Ao, V7) = 0. Equivalently, the map V -+ Hom(W, U) = W* @ U, v++(w+s fo, 1), is an isomorphism.§3.1. Examples: Sq and W, 29 Representations of the Alternating Group 2, What happens to the conjugacy classes above if we replace S, by 2,? Obviously, all the odd conjugacy classes disappear; but at the same time, since conjugation by a transposition is now an outer, rather than inner, auto- morphism, some conjugacy classes may break into two. Exercise 3.4. Show that the conjugacy class in S, of permutations consisting of products of disjoint cycles of lengths b,, bz, ... will break up into the union of two conjugacy classes in %, if all the 4, are odd and distinct; if any b, are even or repeated, it remains a single conjugacy class in %,. (We consider a fixed point as a cycle of length 1.) In the case of U5, this means we have the conjugacy class of three-cycles (as before, 20 elements), and of products of two disjoint transpositions (15 elements); the conjugacy class of five-cycles, however, breaks up into the conjugacy classes of (12345) and (21345), each having 12 elements. As for the representations, the obvious first place to look is at restrictions to %. of the irreducible representations of S, found above. An irreducible representation of S, may become reducible when restricted to U,; or two distinct representations may become isomorphic, as will be the case with U and U’, V and V’, or W and W’. In fact, U, V, and W stay irreducible since their characters satisfy (y, x) = 1. But the character of A?V has values (6,0, —2, 1, 1) on the conjugacy classes listed above, so (x, x) = 2, and A?V is the sum of two irreducible representations, which we denote by Y and Z. Since the sums of the squares of all the dimensions is 60, (dim Y)* + (dim Z)? = 18, so each must be three dimensional. Exercise 3.5. Use the orthogonality relations to complete the character table of U,: t 20 15 12 12 (123) (12934) (12345) (21345) The representations ¥ and Z may in fact be familiar: &, can be realized as the group of motions of an icosahedron (or, equivalently, of a dodecahedron)30 3. Examples; Induced Representations; Group Algebras; Real Representations and ¥ is the corresponding representation. Note that the two representations ‘M, -+ GL3(R) corresponding to Y and Z have the same image, but (as you can see from the fact that their characters differ only on the conjugacy classes of (12345) and (21345)) differ by an outer automorphism of U,. Note also that A?V does not decompose over Q; we could see this directly from the fact that the vertices of a dodecahedron cannot all have rational coordinates, which follows from the analogous fact for a regular pentagon in the plane. Exercise 3.6. Find the decomposition of the permutation representation of U1, corresponding to the (i) vertices, (ii) faces, and (iii) edges of the icosahedron. Exercise 3.7. Consider the dihedral group D,,, defined to be the group of isometries ofa regular n-gon in the plane. Let F = Z/n c D,, be the subgroup of rotations. Use the methods of Lecture J (applied there to the case S; = Dg) to analyze the representations of D,,: that is, restrict an arbitrary representa- tion of D,, to I, break it up into eigenspaces for the action of F, and ask how the remaining generator of D,, acts of these eigenspaces. Exercise 3.8. Analyze the representations of the dihedral group D,, using the character theory developed in Lecture 2. Exercise 3.9. (a) Find the character table of the group of order 8 consisting of the quaternions {+1, +i, tj, +k} under multiplication, This is the case m = 3 of acollection of groups of order 2”, which we denote H,,. To describe them, let C,, denote the complex Clifford algebra generated by v,,..., v, with telations v7 = —{ and v,-v, = —9-0,, 80 C, has a basis 0, = ¥;,"...°0,, aS 1 = {i, < +++ =2/3 001 of H(Z/3)), and then by character theory. §3.2. Exterior Powers of the Standard Representation of S, How should we go about constructing representations of the symmetric groups in general? The answer to this is not immediate; it is a subject that will occupy most of the next lecture (where we will produce all the irreducible representations of S,). For now, as an example of the elementary techniques developed so far we will analyze directly one of the obvious candidates: Proposition 3.12. Each exterior power \*V of the standard representation V of G, is irreducible, O (sgn a)?(sgn b)(sgn c) “5 Ema — 2k +01 (3. sen ) Cz se 6) These last sums are zero unless k — | = 0 or 1. The case k = | gives 1 td ay ka- Kl = a(;)ee l= 1. Similarly, the terms with k — | = 1 also add up to 1, so (x, x) = 2, as required. Note by way of contrast that the symmetric powers of the standard repre- sentation of S, are almost never irreducible. For example, we already know that the representation Sym? V contains one copy of the trivial representation: this is just the statement that every irreducible real representation (such as V) admits an inner product (unique, up to scalars) invariant under the group action; nor is the quotient of Sym?V by this trivial subrepresentation neces- sarily irreducible, as witness the case of S,. §3.3. Induced Representations If H ¢ Gisa subgroup, any representation V of G restricts to a representation of H, denoted Res§ V or simple Res V. In this section, we describe an impor- tant construction which produces representations of G from representations of H. Suppose V is a representation of G, and W c V is a subspace which is H-invariant. For any g in G, the subspace g- W = {g-w: we W} depends only on the left coset gH of g modulo H, since gh: W = g-(h: W) = g: W; fora coset§3.3. Induced Representations 33 a in G/H, we write o- W for this subspace of V. We say that V is induced by W if every element in V can be written uniquely as a sum of elements in such translates of W, i.e., V= @®ow oeGiHt In this case we write V = Ind§W = Ind W. Example 3.13. The permutation representation associated to the left action of G on G/H is induced from the trivial one-dimensional representation W of H. Here V has basis {e,: ¢ € G/H}, and W = C-e,, with 1 the trivial coset. Example 3.14. The regular representation of G is induced from the regular representation of H. Here Y has basis {e,:g eG}, whereas W has basis {e,: h € H}. We claim that, given a representation W of H, such V exists and is unique up to isomorphism. Although we will later give several fancier ways to see this, it is not hard to do it by hand. Choose a representative g, € G for each coset o € G/H, with e representing the trivial coset H. To see the uniqueness, note that each element of V has a unique expression v = )'g,w,, for elements w, in W. Given g in G, write g-g, = g,"h for some t € G/H and he H. Then we must have 9'(9oWe) = (9° Go)e = (e' hw, = 9,(hw,)- This proves the uniqueness and tells us how to construct V = Ind(W) from W. Take a copy W’ of W for each left coset o e G/H; for we W, let g,w denote the element of W* corresponding to w in W. Let V= @ _W?, so every oeGit element of V has a unique expression v = )'g,w, for elements w, in W. Given g& G, define 9°(deW%) = 9elhwe) ifg-Go = gh. To show that this defines as action of G on V, we must verify that g’-(g-(g.,)) = (9'-g)-(go¥,) for another element g’ in G. Now if g'-g, = g,'H’, then 4° (G°(9.W_)) = 9°: (gelhw,)) = 9,(h'(hw,)). Since (9°: 9)- Go = 9'-(9° 9.) = 9° Gh = g,"h'-h, we have (9°: 9)" (GoWe) = Opl(Ht “h)w,) = gp{h’ -(Aw,)), as required. Example 3.15. If W = QW, then Ind W = @Ind W,. The existence of the induced representation follows from Examples 3.14 and 3.15 since any W is a direct sum of summands of the regular representation.34 3, Examples; Induced Representations; Group Algebras; Real Representations Exercise 3.16. (a) If U is a representation of G and W a representation of H, show that (with all tensor products over C) U @ Ind W = Ind(Res(U) ® W). In particular, Ind(Res(U)) = U @ P, where P is the permutation representa- tion of G on G/H. For a formula for Res(Ind(W)), for W a representation of H, see [Se2, p. 58]. (b) Like restriction, induction is transitive: if H ¢ K ¢ G are subgroups, show that Ind§(W) = Indg(Indk W). Note that Example 3.15 says that the map Ind gives a group homomor- phism between the representation rings R(H) and R(G), in the opposite direc- tion from the ring homomorphism Res: R(G)~ R(H) given by restriction; Exercise 3.16(a) says that this map satisfies a “push-pull” formula a: Ind(f) = Ind(Res(a)- B) with respect to the restriction map. Proposition 3.17. Let W be a representation of H, U a representation of G, and suppose V = Ind W. Then any H-module homomorphism go: W— U extends uniquely to a G-module homomorphism @: V + U. i.e., Hom,,(W, Res U) = Hom,(Ind W, U). In particular, this universal property determines Ind W up to canonical isomorphism. Proor. With V = @, eqn 0° W as before, define @ on a: W by o W W825 Us U, which is independent of the representative g, for o since g is H-linear. To compute the character of V = Ind W, note that g € G maps oW to goW, so the trace is calculated from those cosets o with go = 9, ie., s'gs € H for seo. Therefore, Xana w(9) aw(s'gs) (seo arbitrary). (3.18) Exercise 3.19. (a) If C is a conjugacy class of G, and Cn H decomposes into conjugacy classes D,,..., D, of H, (3.18) can be rewritten as: the value of the character of Ind W on C is IGl & [Dil na (C) = Ye tonal = iG D TG] (b) If W is the trivial representation of H, then G:H timate Xw(D)).§3.3. Induced Representations 35 Corollary 3.20 (Frobenius Reciprocity). If W is a representation of H,and Ua representation of G, then Chana Xu)e = (hrs Xresu)n- Proor. It suffices by linearity to prove this when W and U are irreducible. The left-hand side is the number of times U appears in Ind W, which is the dimension of Hom,(Ind W, U). The right-hand side is the dimension of Hom,,(W, Res U). These dimensions are equal by the proposition. o If W and U are irreducible, Frobenius reciprocity says: the number of times U appears in Ind W is the same as the number of times W appears in Res U. Frobenius reciprocity can be used to find characters of G if characters of H are known. Example 3.21. We compute Indj W, when H = S, c G = G3, W = J, (the standard representation) = U; (the alternating representation). We know the irreducible represenatations of G3: U;, U3, V3, which restrict to U,, U; = Vz, U, ® Uj, respectively. Thus, by Frobenius, Ind V, = U3 ® Vs. Example 3,22. Consider next H = S3 ¢ G = G4, W = Vy. Again we know the irreducible representations, and Res U, = U;, Res Uj = Uj, Res Vs = Us ® Vs [the vector (1, 1 1, = 3) © Va = {ens Xa5 X35 Xa): Dr = OF is fixed by H], Res Vj = Us ® V3, with V3 = V3, and Res W, = V3 (as one may see directly). Hence, Ind V, = V, ® 7; ® W,. (Note that the isomorphism Res W, = V, actually follows, since one W, is all that could be added to V, ® Vj to get Ind Vs.) Exercise 3.23, Determine the isomorphisin classes of the representations of S, induced by (i) the one-dimensional representation of the group generated by (1234) in which (1234)-v = iv, i = ./ — 1; (ii) the one-dimensional representa- tion of the group generated by (123) in which (123)-v = e?"p, Exercise 3.24, Let H = WM, < G = Ss. Show that Ind U = U@U’, Ind V= V@ V‘, and Ind W = W@ W’, whereas Ind Y = Ind Z = A?V. Exercise 3.25*. Which irreducible representations of S, remain irreducible when restricted to %,? Which are induced from %,? How much does this tell you about the irreducible representations of 2,7 Exercise 3.26*, There is a unique nonabelian group of order 21, which can be realized as the group of affine transformations x ++ ax + of the line over the field with seven elements, with « a cube root of unity in that field. Find the irreducible representations and character table for this group.36 3, Examples; Induced Representations; Group Algebras; Real Representations Now that we have introduced the notion of induced representation, we can state two important theorems describing the characters of representations of a finite group. In the preceding lecture we mentioned the notion of virtual character; this is just an element of the image A of the character map XRG) > Corus) from the representation ring R(G) of virtual representations. The following two theorems both state that in order to generate A ® Q (resp. A) it is enough to consider the simplest kind of induced representations, namely, those induced from cyclic (respective elementary) subgroups of G. For the proofs of these theorems we refer to [Se2, §9, 10]. We will not need them in these lectures, Artin’s Theorem 3,27. The characters of induced representations from cyclic subgroups of G generate a lattice of finite index in A. A subgroup H of G is p-elementary if H = A x B, with A cyclic of order prime to p and B a p-group. Brauer’s Theorem 3.28. The characters of induced representations from elemen- tary subgroups of G generate the lattice A. §3.4. The Group Algebra There is an important notion that we have already dealt with implicitly but not explicitly; this is the group algebra CG associated toa finite group G. This is an object that for all intents and purposes can completely replace the group G itself; any statement about the representations of G has an exact equivalent statement about the group algebra. Indeed, to a large extent the choice of language is a matter of taste. The underlying vector space of the group algebra of G is the vector space with basis {e,} corresponding to elements of the group G, that is, the under- lying vector space of the regular representation. We define the algebra struc- ture on this vector space simply by 5" &n = Sgn By a representation of the algebra CG on a vector space V we mean simply an algebra homomorphism CG -— End(V), so that a representation V of CG is the same thing as a left CG-module. Note that a representation p: G -» Aut(V) will extend by linearity toa map p: CG > End(V), so that representations of CG correspond exactly to representations of G; the left CG-module given by CG itself corresponds to the regular representation.§3.4. The Group Algebra 37 If {W,} are the irreducible representations of G, then we have seen that the regular representation R decomposes R= Dimeoanm, We can now refine this statement in terms of the group algebra: we have Proposition 3.29. As algebras, CG = @ End(w). Proor. As we have said, for any representation W of G, the map G ~ Aut(W) extends by linearity to a map CG —~ End(W); applying this to each of the irreducible representations W, gives us a canonical map e: CG + @ End(). This is injective since the representation on the regular representation is faithful. Since both have dimension ) (dim W/,)’, the map is an isomorphism. o A few remarks are in order about the isomorphism ¢ of the proposition. First, @ can be interpreted as the Fourier transform, cf. Exercise 3.32. Note also that Proposition 2.28 has a natural interpretation in terms of the group algebra: it says that the center of CG consists of those )’ «(g)e, for which a is a class function. Next, we can thinkof ¢ as the decomposition of the semisimple algebra CG into a product of matrix algebras. It implies that the matrix entries of the irreducible representations give a basis for the space of all functions on G, cf. Exercise 2.35. Note in particular that any irreducible representation is isomorphic to a (minimal) left ideal in CG. These left ideals are generated by idempotents. In fact, we can interpret the projection formulas of the last lecture in the language of the group algebra: the formulas say simply that the elements to. idi x Xw(9) e, € CG a dim W- are the idempotents in the group algebra corresponding to the direct sum factors in the decomposition of Proposition 3.29. To locate the irreducible representations W, of a group G [not just a direct sum of dim(W;) copies], we want to find other idempotents of CG. We will see this carried out for the symmetric groups in the following lecture. The group algebra also gives us another description of induced representa- tions: if W is a representation of a subgroup H of G, then the induced representation may be constructed simply by Ind W = CG @cn W,38 3. Examples; Induced Representations; Group Algebras; Real Representations where G acts on the first factor: g-(e,- ® w) = ég¢ © w. The isomorphism of the reciprocity theorem is then a special case ofa general formula for a change of rings CH + CG: Homeg(W, U) = Homeg(CG @cu W, UV). Exercise 3.30*. The induced representation Ind(W) can also be realized con- cretely as a space of W-valued functions on G, which can be useful to produce matrix realizations, or when trying to decompose Ind(W) into irreducible Pieces. Show that Ind(W) is isomorphic to Hom,(CG, W) = {f: G > W:f(hg) = hf(g), Vhe H,g eG}, where G acts by (g’- f)(g) = f(gq’). Exercise 3.31. If CG is identified with the space of functions on G, the function corresponding to J°,.¢(g)e,, show that the product in CG corresponds to the convolution * of functions: (e+ Wg) = a (hye ('g). (With integration replacing summation, this indicates how one may extend the notion of regular representation to compact groups.) Exercise 3.32*. If p: G + GL(V,) is a representation, and ¢ is a function on G, define the Fourier transform (p) in End(V,) by the formula 9(p) = oe (9): (9). A (a) Show that 9*H() = @l0) V0). (b) Prove the Fourier inversion formula 1 . = 0D = Gy Ldim(¥,): Trace(o(g™'): (9), the sum over the irreducible representations p of G. This formula is equivalent to formulas (2.19) and (2.20). (©) Prove the Plancherel formula for functions @ and ys on G: wa sir Teace(O(n)¥(0) EY o(g (9) = ae Our choice of left action of a group on a space has been perfectly arbitrary, and the entire story is the same if G acts on the right instead. Moreover, there is a standard way to change a right action into a left action, and vice versa: Given a right action of G on V, define the left action by gu=v(g'), geGvev.§3.5, Real Representations and Representations over Subfields of € 39 If A = CG is the group algebra, a right action of G on V makes V a right A-module, To turn right modules into left modules, we can use the anti- involution a++d of A defined by (a,¢,)* = )a,e,-1. A right A-module is then turned into a left A-module by setting a-v = 0-4. The following exercise will take you back to the origins of representation theory in the 19th century, when Frobenius found the characters by factoring this determinant. Exercise 3.33*. Given a finite group G of order n, take a variable x, for each element g in G, and order the elements of G arbitrarily. Let F be the deter- minant of the n x 1 matrix whose entry in the row labeled by g and column labeled by h is x,.4-1. This is a form of degree n in the n variables x,, which is independent of the ordering. Normalize the factors of F to take the value 1 when x, = 1 and x, =0 for g #e. Show that the irreducible factors of F correspond to the irreducible representations of G. Moreover, if F, is the factor corresponding to the representation p, show that the degree of F, is the degree d(p) of the representation p, and that each F, occurs in F d(p) times. If x, is the character of p, show that y,(g) is the coefficient of x, - x4"! in F,. §3.5. Real Representations and Representations over Subfields of C Ifa group G acts on a real vector space Vo, then we say the corresponding complex representation of V = ¥) @g C is real. To the extent that we are interested in the action of a group G on real rather than complex vector spaces, the problem we face is to say which of the complex representations of G we have studied are in fact real. Our first guess might be that a representation is real if and only if its character is real-valued. This turns out not to be the case: the character of a real representation is certainly real-valued, but the converse need not be true. To find an example, suppose G ¢ SU(2)isa finite, nonabelian subgroup. Then G acts on C? = V with a real-valued character since the trace of any matrix in SU(2) is real. If V were a real representation, however, then G would be a subgroup of SO(2) = S', which is abelian. To produce such a group, note that SU(2) can be identified with the unit quaternions. Set G = {+1, ti, +i, +k}. Then G/{ + 1} is abelian, so has four one-dimensional representations, which give four one-dimensional representations of G. Thus, G has one irreducible two-dimensional representation, whose character is real, but which is not real, Exercise 3.34%. Compute the character table for this quaternion group G, and compare it with the character table of the dihedral group of order 8.40 3. Examples; Induced Representations, Group Algebras; Real Representations A more successful approach is to note that if V is a real representation of G, coming from V, as above, then one can find a positive definite symmetric bilinear form on VY which is preserved by G. This gives a symmetric bilinear form on V which is preserved by G. Not every representation will have such a form since degeneracies may arise when one tries to construct one following the construction of Proposition 1.5, In fact, Lemma 3.35. An irreducible representation V of G is real if and only if there is a nondegenerate symmetric bilinear form B on V preserved by G. Proor. If we have such B, and an arbitrary nondegenerate Hermitian form H, also G-invariant, then vavesy gives a conjugate linear isomorphism @ from V to V: given x € V, there is a unique (x) € V with B(x, y) = H(@(x), y), and @ commutes with the action of G. Then 9? = go ¢ is a complex linear G-module homomorphism, so @? = A-Id. Moreover, H(@(x), ») = B(x, y) = Bly, ») = H(9(y), 0 = Hl 0), from which it follows that H(@?(x), y) = H(x, @?(y)), and therefore 2 is a Positive real number. Changing H by a scalar, we may assume 4 = I, so @ = Id. Thus, V is a sum of real eigenspaces V, and V_ for @ corresponding toeigenvalues 1 and — 1. Since ¢ commutes with G, V, and V_ are G-invariant subspaces. Finally, (ix) = —ie(x), so iV, = V.,and V=V, @C. 0 Note from the proof that a real representation is also characterized by the existence of a conjugate linear endomorphism of V whose square is the identity; if V = Vo @g C, it is given by conjugation: vp @ AH vy @ 1. A warning is in order here: an irreducible representation of G on a vector space over R may become reducible when we extend the group field to C. To give the simplest example, the representation of Z/n on R? given by 2nk al cos—— = —sin—- k n n pki . 2mk 2nk sin— cos — n n is irreducible over R for n > 2 (no line in R? is fixed by the action of Z/n), but will be reducible over C. Thus, classifying the irreducible representations of G over C that are real does not mean that we have classified all the irreducible real representations. However, we will see in Exercise 3.39 below how to finish the story once we have found the real representations of G that are irreducible over C.§3.5. Real Representations and Representations over Subfields of € 4 Suppose V is an irreducible representation of G with x, real. Then there is a G-equivariant isomorphism V = V*, ic. there is a G-equivariant (non- degenerate) bilinear form B on V; but, in general, B need not be symmetric. Regarding B in V*@V* =Sym’v*@NV*, and noting the uniqueness of B up to inultiplication by scalars, we see that B is either symmetric or skew-symmetric. If B is skew-symmetric, proceeding as above one can scale so p? = —Id. This makes V “quaternionic,” with @ becoming multiplication? by j: Definition 3.36. A quaternionic representation is a (complex) representation V which has a G-invariant homomorphism J: V > V that is conjugate linear, and satisfies J? = —Id, Thus, a skew-symmetric nondegenerate G-invariant B determines a quaternionic structure on V. Summarizing the preceding discussion we have the Theoem 3.37. An irreducible representation V is one and only one of the following: (1) Complex: yy is not real-valued; V does not have a G-invariant non- degenerate bilinear form. (2) Real: V = V, @C, a real representation; V has a G-invariant symmetric nondegenerate bilinear form. (3) Quaternionic: yy is real, but V is not real; V has a G-invariant skew- symmetric nondegeneate bilinear form. Exercise 3.38. Show that for V irreducible, 0 if V is complex x ¥Y wlg?)=4 1 if Vis real IG| ee —L_ if V is quaternionic. This verifies that the three cases in the theorem are mutually exclusive. It also implies that if the order of G is odd, all nontrivial representations must be complex. Exercise 3.39. Let Vy be a real vector space on which G acts irreducibly, V = ¥®C the corresponding real representation of G. Show that if V is not irreducible, then it has exactly two irreducible factors, and they are conjugate complex representations of G. 7 See §7.2 for more on quaternions and quaternonic representations.42 3. Examples; Induced Representations; Group Algebras; Real Representations Exercise 3.40. Classify the real representations of 9(,. Exercise 3.41*. The group algebra RG is a product of simple R-algebras corre- sponding to the irreducible representations over R. These simple algebras are matrix algebras over C, R, or the quaternions H according as the representa- tion is complex, real, or quaternionic. Exercise 3.42*. (a) Show that all characters of a group are real if and only if every element is conjugate to its inverse. (b) Show that an element o in a split conjugacy class of Mf, is conjugate to its inverse if and only if the number of cycles in ¢ whose length is congruent to 3 modulo 4 is even. (c) Show that the only d’s for which every character of 9, is real-valued are d=1,2,5,6, 10, and 14, Exercise 3.43*. Show that: (i) the tensor product of two real or two quater- nionic representations is real; (ii) for any V, V* @ Vis real; (iii) if V is real, so are all A'V; (iv) if V is quaternionic, A'V is real for k even, quaternionic for k odd. Representations over Subfields of C in General We consider next the generalization of the preceding problem to more general subfields of C. Unfortunately, our results will not be nearly as strong in general, but we can at least express the problem neatly in terms of the representation ring of G. To begin with, our terminology in this general setting is a little different. Let K < C be any subfield. We define a K-representation of G to be a vector space ¥, over K on which G acts; in this case we say that the complex representation V = Vo @ C is defined over K. One way to measure how many of the representations of G are defined over a field K is to introduce the representation ring Rx(G) of G over K. This is defined just like the ordinary representation ring; that is, it is just the group of formal linear combinations of K-representations of G modulo relations of the form V + W —(V® W), with multiplication given by tensor product. Exercise 3.44*. Describe the representation ring of G over R for some of the groups G whose complex representation we have analyzed above. In partic- ular, is the rank of Ry(G) always the same as the rank of R(G)? Exercise 3.45*. (a) Show that Rx(G) is the subring of the ring of class functions on G generated (as an additive group) by characters of representations defined over K,§3.5. Real Representations and Representations over Subfields of C 43 (b) Show that the characters of irreducible representations over K form an orthogonal basis for Rx(G). (©) Show that a complex representation of G can be defined over K if and only if its character belongs to Rx(G). For more on the relation between Rx(G) and R(G), see [Se2].LECTURE 4 Representations of S,: Young Diagrams and Frobenius’s Character Formula In this lecture we get to work. Specifically, we give in §4.1 a complete description of the irreducible representations of the symmetric group, that is, a construction of the representations (via Young symmetrizers) and a formula (Frobenius’ formula) for their characters. The proof that the representations constructed in §4.1 are indeed the irreducible representations of the symmetric group is given in §4.2; the proof of Frobenius’ formula, as well as a number of others, in §4.3. Apart from their intrinsic interest (and undeniable beauty), these results turn out to be of substantial interest in Lie theory: analogs of the Young symmetrizers will give a construction of the irreduc- ible representations of SL, C. At the same time, while the techniques of this lecture are completely elementary (we use only a few identities about symmetric polynomials, proved in Appendix A), the level of difficulty is clearly higher than in preceding lectures. The results in the latter half of §4.3 (from Corollary 4.39 on) in particular are quite difficult, and inasmuch as they are not used later in the text may be skipped hy readers who are not symmetric group enthusiasts. .1: Statements of the results .2: Irreducible representations of S, §4.3; Proof of Frobenius's formula §4.1. Statements of the Results The nuinber of irreducible representaton of S, is the number of conjugacy classes, which is the number p(d) of partitions’ of d:d=A, ++ +4,, Ay > 2A, > 1. We have "ILis sometimes conveniemt, and sometimes a nuisance, to have partitions that end in one or mnveniertt, we allow some of the A, on the end to be zero, Two sequences define the same partition, of course, if they differ only by zeros at the end.$4.1. Statements of the Results 45 7) K(LECF PH IEEE LE peje ¥ va = TI (; ro neh which converges exactly in {t| < 1. This partition number is an interesting arithmetic function, whose congruences and growth behavior as a function of dhave been much studied (cf. [Har], [And]). For example, p(d) is asymptoti- cally equal to (I/ad)e* v4, with a = 4,/3 and B = n./2/3. To a partition A = (A;,..., A,) is associated a Young diagram (sometimes called a Young frame or Ferrers diagram) with A, boxes in the ith row, the rows of boxes lined up on the left. The conjugate partition 1’ = (4,,,..., 44) to the partition A is defined by inter- changing rows and columns in the Young diagram, i.c,, reflecting the diagram in the 45° line. For example, the diagram above is that of the partition (3, 3,2, 1, 1), whose conjugate is (5, 3, 2). (Without reference to the diagram, the conjugate partition to can be defined by saying 2/ is the number of terms in the partition 4 that are greater than or equal toi.) Young diagrams can be used to describe projection operators for the regular representation, which will then give the irreducible representations of G,. For a given Young diagram, number the boxes, say consecutively as shown: [273] 5 7 oolalay— More generally, define a tableau on a given Young diagram to be a numbering of the boxes by the integers 1,...,d. Given a tableau, say the canonical one shown, define two subgroups? of the symmetric group 7 Ifatableau other than the canonical one were chosen, one would get different groups in place of P and Q, and different elements in the group ring, but the representations constructed this way will be isomorphic.46 4, Representations of S,: Young Diagrams and Frobenius’s Character Formula P = P, = (ge S,:g preserves each row} and Q = 0, = {9 € Sa: g preserves each column}. In the group algebra CG,, we introduce two elements corresponding to these subgroups: we set a,= Ye, and b= Y) sgn(g)-e,. (4.1) oP gO To see what a, and b, do, observe that if V is any vector space and G, acts on the dth tensor power V ® by permuting factors, the image of the element a, € CG, — End(V ®) is just the subspace Im(a,) = Sym" V @Sym?V @ + @Sym*V c V%, where the inclusion on the right is obtained by grouping the factors of V® according to the rows of the Young tableaux. Similarly, the image of b, on this tensor power is Im(h,) = NV @AVV@- @NV c Ve, where j: is the conjugate partition to A. Finally, we set eg = a,b, ECS, (4.2) this is called a Young symmetrizer. For example, when A = (d), ej) = Ga) = Yeees% and the image of cg on V% is Sym*¥. When A=(1,..., 1), Con ty = Pita) = Depo #,88M(g)ep» and the image of c,,_..1 00 Vis NY. We will eventually see that the image of the symmetrizers c, in V®? provide essentially all the finite-dimensional irreducible representations of GL(V). Here we state the corresponding fact for representations of S,: Theorem 4.3. Some scalar multiple of c, is idempotent, i.e., c? = n,c,, and the image of c, (by right multiplication on CG,) is an irreducible representation V, of G,. Every irreducible representation of. S, can be obtained in this way for a unique partition. We will prove this theorem in the next section, Note that, as a corollary, each irreducible representation of G, can be defined over the rational numbers since c, is in the rational group algebra QG,. Note also that the theorem gives a direct correspondence between conjugacy classes in ©, and irreducible representations of G,, something which has never been achieved for general groups. For example, for 1 = (d), Va = CG, (a ad Py 0§4.1. Statements of the Results 47 is the trivial representation U, and when A = (1, ..., 1), Vo. = cee ye sgn(g)e, = cx sgn(g)e, is the alternating representation U’. For A = (2, 1), Cea,ny = Cr + ay)" (C1 — Cc) = E+ e4ay ~ 13) — 132 in CG,, and Vj, 1,is spanned by cy, 4, and (13)- cy, 1), 80 Viz, 1, is the standard representation of G3. Exercise 4.4%. Set A = CG,, 80 Vy = Ac, = Aajh,. {a) Show that V, = Ad,a,. (b) Show that V, is the image of the map from Aa, to Ab, given by right multiplication by 6,. By (a), this is isomorphic to the image of Ab, > Aa, given by right multiplication by a,. {c) Using (a) and the description of V, in the theorem show that Vy = YOu, where J’ is the conjugate partition to A and U’ is the alternating representation. Examples 4.5. In earlier lectures we described the irreducible representations of G, ford < 5. From the construction of the representation corresponding to a Young diagram it is not hard to work out which representations come from which diagrams: Ss, trivial H altemating SG ] U tivial A U' altemating FP V_ standard Sy om vu H u fe v FF vo fy w48 4. Representations of S,: Young Diagrams and Frobenius’s Character Formula 8, com vu H ul fFFo iv [Py fe nv FH w AP ow Exercise 4.6*. Show that for general d, the standard representation V corre- sponds to the partition d = (d — t) + £. Asa challenge, you can try to prove that the exterior powers of the standard representation V are represented by a “hook”: foo £ nv Lo Note that this recovers our theorem that the A'V are irreducible. Next we turn to Frobenius's formula for the character x, of V,, which includes a formula for its dimension. Let C, denote the conjugacy class in S, determined by a sequence i= (i, f2,..., i) with Yai, = d: C, consists of those permutations that have i, 1-cycles, i, 2-cycles, ..., and i, d-cycles. Introduce independent variables x,,..., X,, with k at least as large as the number of rows in the Young diagram of 4. Define the power sums P,(x), 1 + > A, = O of d, set aAtk—-t b=Aytk—-2,..,h =a, (49) (47)§4.1. Statements of the Results 49 a strictly decreasing sequence of k non-negative integers. The character of V, evaluated on g € C, is given by the remarkable Frobenius Formula 4.10 (G) = [ac iF ye J Hl y 5 seo Ed For example, if d = 5, A = (3, 2), and C, is the conjugacy class of (12)(345), ie, iy =0,i, = 1,13 = 1, then Yea.) = [Oe — 2) OF + DOE + Daa = b Other entries in our character tables for $3, S,, and G, can be verified as easily, verifying the assertions of Examples 4.5. In terms of certain symmetric functions S, called Schur polynomials, Fro- benius’s formula can be expressed by THe = Dal, the sum over all partitions 1 of d in at most & parts (cf. Proposition 4.37 and (A.27)). Although we do not use Schur polynomials explicitly in this lecture, they play the central role in the algebraic background developed in Appendix A. Let us use the Frobenius formula to compute the dimension of V,. The conjugacy class of the identity corresponds to i = (d), so . dim Vy = zal) = (AG) (er 2 + 4)... ta Now A(x) is the Vandermonde determinant: Tog xg Do: = Y (ema)xgr tee segth-t, foxy xb © o8by The other term is ty beta = So ata a the sum over k-tuples (r,,...,%) that sum to d, To find the coefficient of i'+...°X,"in the product, we pair off corresponding terms in these two sums, dad Lsenio): G = 6) + OP G—o(f) + OP the sum over those o in G, such that ky, — (i) + 12 0 forall t 1), and a(I2... m) = Ak m 9X) = i hs ¥E)= 9 —m i (&-j+0. The sum in (4.18) can be realized as the coefficient of x7! in the Laurent expansion of /(x)/@(x) at x = 00. Define the rank r of a partition to be the length of the diagonal of its Young diagram, and let a; and 6, be the number of boxes below and to the right of the ith box of the diagonal, reading from lower right to upper left. Frobenius 14 0. A, called ( by ed areverse notation for the characteristics, writing (b,, ..., ),|a,,..., 4,) instead.) For the partition (10, 9, 9, 4, 4, 4, 1): J r=4 characteristics 2346 0967 9 ) the characteristics of the partition. (Many writers now use Algebraically, r and the characteristics a, <-- gq. >" > q, > 0 be the lengths of the symmetric hooks that form the diagram of A; thus, 4; = 2A, — 1, q, = 2A, — 3,.... Show that if g is a product of disjoint cycles of lengths 41, 42, ..-, 4, then wala) = (= DI”. §4.2. Irreducible Representations of S, We show next that the representations V, constructed in the first section are exactly the irreducible representations of Gj. This proof appears in many standard texts (e.g. [C-R], [Ja-Ke], [N-S], [Wet]), so we will be a litte concise. Let A = CG, be the group ring of S,. For a partition A of d, let P and Q be the corresponding subgroups preserving the rows and columns of a Young tableau T corresponding to 4, let a=a,, b= 6,, and let c=c, =ab be§4.2. Irreducible Representations of S, 53 the corresponding Young symmetrizer, so V, = Ac, is the corresponding representation. (These groups and elements should really be subscripted by T to denote dependence on the tableau chosen, but the assertions made depend only on the partition, so we usually omit reference to T.) Note that PQ = {1}, so an element of G, can be written in at most one way as a product p-q, p€ P, q € Q. Thus, c is the sum J) te,, the sum over all g that can be written as p- q, with coefficient + 1 being sgn(q); in particular, the coefficient of e, in c is 1. Lemma 4.21. (1) For pe P, p-a = a-p =a. (2) For q yp ifthe first nonvanishing 4, ~ 1; is positive. (4.22) Lemma 4.23.(1) If A > 1, then forall x € A,a,:x-b, = 0. In particular, if A > pt, then c,*¢, = 0. (2) For all x € A, ¢,°x*¢, = isa scalar multiple of c,. In particular, c,-¢, = nyc, for some n,é C. Proor. For (1), we may take x = g € G,. Since g:b,-g~! is the element con- structed from gT’, where T” is the tableau used to construct b,, it suffices to show that a,°b, = 0. One verifies that 1 > implies that there are two integers in the same row of T and the same column of 7’. If ¢ is the transposition of these integers, then a,:t = a, t°b, = —b,, 80 a," by = ay't- tb, = — ay by, as required. Part (2) follows from Lemma 4.21 (3). Exercise 4.24*. Show that ifA # p, thenc,-A-c, = 0; in particular, ¢,-cy54 4, Representations of Sj: Young Diagrams and Frobenius’s Character Formula Lemma 4.25. ({) Each V, is an irreducible representation of S,. (2) If. A# 1p, then V, and V, are not isomorphic. Proor. For (1) note that c,V¥,¢ Cc, by Lemma 4.23. If We V, is a subrepresentation, then c,W is either Cc, or 0. If the first is true, then V, = A-c,¢ W. Otherwise W- Wc A-c,W =0, but this implies W = 0. Indeed, a projection from A onto W is given by right multiplication by an element ge A with g = y?« W-W =0. This argument also shows that c,V, # 0, ic., that the number 1, of the previous lemma is nonzero. For (2), we may assume A > p. Thence, V, = Ce, # 0, bul, V, = ¢,° Acy = 0, so they cannot be isomorphic A-modules. oO Lemma 4.26. For any A, c,"¢4 =, C,, with n, = di/dim V,, Proor. Let F be right multiplication by c, on A. Since F is multiplication by n, on V,, and zero on Ker(c,), the trace of F is n, times the dimension of V,. But the coefficient of e, in e,-c, is 1, so trace(F) = |G,| = dl. im} Since there are as many irreducible representations V, as conjugacy classes of G,, these must form a complete set of isomorphism classes of irreducible representations, which completes the proof of Theorem 4.3, In the next section we will prove Frobenius’s formula for the character of V,, and, in a series of exercises, discuss a little of what else is known about them: how to decompose tensor products or induced or restricted representations, how to find a basis Tor V,, ete. §4.3. Proof of Frobenius’s Formula For any partition 1 of d, we have a subgroup, often called a Young subgroup, GS, = Gy Kx GG, (427) Let U, be the representation of G, induced from the trivial representation of G,. Equivalently, U, = A-a,, with a, as in the preceding section. Let Vi = Xu, = character of U,. (4.28) Key to this investigation is the relation between U, and ¥,, i.e., between p, and the character x, of V,. Note first that V, appears in U,, since there is a surjection U, = Aa, -» Vy = Aah, xH+x- by. (4.29) Alternatively, V, = Aa,h, = Abja, < Aa, = U,,§4.3. Proof of Frobenius’s Formula 55 by Exercise 4.4. For example, we have Yuet.) % Yar.) ® May which expresses the fact that the permutation representation C? of G, is the sum of the standard representation and the trivial representation. Eventually we will see that every U, contains V, with multiplicity one, and contains only other V, for p> A. The character of U, is easy to compute directly since U, is an induced representation, and we do this next. Fori = (i;,..., ig) d-tuple of non-negative integers with )\ai, = d, denote by Ges the conjugacy class consisting of elements made up of i, I-cycles, i, 2-cycles, +++, ig d-cycles. The number of elements in C, is easily counted to be d! =o a 4, lal 185, 12h, dsi,1 (4.30) By the formula for characters of induced representations (Exercise 3.19), Va(C) = mite STIG S,) ti! Il pa Pry, a hkl where the sum is over all collections {r,4:1 0. (4.38) n Consider this equation together with (4.35). We deduce first that each w, is a§4.3. Proof of Frobenius's Formula 57 virtual character: we can write O,= 2 Myke Maye Z But the @,, like the y,, are orthonormal by (4.36), so 1 = (0, 0) = my ¥ and hence w, is + x for some itreducible character y. (It follows from the hook length formula that the plus sign holds here, but we do not need to assume this.) Fix A, and assume inductively that y, = @, for all p > A, so by (4.35) Ya=o,+ zs Kuikw Comparing this with (4.38), and using the linear independence of characters, the only possibility is that w, = x4. a. Corollary 4.39 (Young’s rule). The integer K,,, is the multiplicity of the irreduc- ible representation V, in the induced representation U,: UAZnODKwhe ett D Kate na era Note that when A= (1,..., 1), U, is just the regular representation, so Kya,...1) = dim V,. This shows that the dimension of V, is the number of standard tableaux on A, i.e. the number of ways to fill the Young diagram of Awith the numbers from { to d, such that all rows and columns are increasing. The hook length formula gives another combinatorial formula for this dimen- sion. Frame, Robinson, and Thrall proved that these two numbers are equal. Fora short and purely combinatorial proof, see [G-N-W]. For another proof that the dimension of V, is the number of standard tableaux, see [Jam]. The latter leads to a canonical decomposition of the group ring A = CG, as the direct sum of left ideals Ae, summing over all standard tableaux, with er = (dim V,/d!)- cy, and cy the Young symmetrizer corresponding to T, cf. Exercises 4.47 and 4.50. This, in turn, leads to explicit calculation of matrices of the representations V, with integer coefficients. For another example of Young's rule, we have a decomposition Ua-aa) = Q Ya-t- In fact, the only 4 whose diagrams can be filled with d— |’s and a 2's, nondecreasing in rows and strictly increasing in columns, are those with at most two rows, with the second row no longer than a; and such a diagrain has only one such tableau, so there are no multiplicities. Exercise 4.40*. The characters y, of G, have been defined only when dis a pattition of d. Extend the definition to any k-tuple a = (a,,..., a,) of integers58 4. Representations of S,: Young Diagrams and Frobenius’s Character Formula that add up to d by setting y, = 0 if any of the a, are negative, and otherwise Wa = W,, where A is the reordering of a,,..., a, in descending order. In this case Wy, is the character of the representation induced from the trivial represen- tation by the inclusion of G,, x ++: x S,, in G,. Use (A.5) and (A.9) of Appendix A to prove the determinantal formula for the irreducible characters X, in terms of the induced characters p,: ae X SBO(C)Weay beg tpt dy (2) Zon dt) Ifone writes y, as a formal product ¥,,°W., can be written » the preceding formula Ya Wart Wayte = Wass Ort Mare Vane Me The formal product of the preceding exercise is the character version of an “outer product” of representations. Given any non-negative integers d,,... d,, and representations V, of G,,, denote by V, © --- © K the (isomorphism class of the) representation of S,,d = }.d,, induced from the tensor product repre- sentation V, @---BV, of S,, x --- x S,, by the inclusion of Gy, x" x Sa in S, (see Exercise 2.36). This product is commutative and associative. It will turn out to be useful to have a procedure for decomposing such a representa- tion into its irreducible pieces. For this it is enough to do the case of two factors, and with the individual representations V, irreducible. In this case, one has, for V, the representation of S, corresponding to the partition 4 of d and V, the representation of S,, corresponding to the partition yt of m, Kok =T Nah (441) the sum over all partitions v of d + m, with N,,, the coefficients given by the Littlewood- Richardson rule (A.8) of Appendix A. Indeed, by the exercise, the character of V, ¢ V, is the product of the corresponding determinants, and, by {A.8), that is the sum of the characters Nj,,%y- When m = | and p = (mn), F, is trivial; this gives Inde", = PV, (4.42) the sum over all v whose Young diagram is obtained from that of A by adding one box. This formula uses only a simpler form of the Littlewood—Richardson rule known as Pieri’s formula, which is proved in (A.7). Exercise 4.43*. Show that the Littlewood-Richardson number N,,, is the multiplicity of the irreducible representation V, 8 V, in the restriction of V, from G,,,, to Gx S,,. In particular, taking m = 1, p = (1), Pieri’s formula (A.7) gives Res8e'¥, = 3 Vs,§4.3. Proof of Frobenius’s Formula 59 the sum over all 1 obtained from v by removing one box. This is known as the “branching theorem,” and is useful for inductive proofs and constructions, particularly because the decomposition is multiplicity free. For example, you can use it to reprove the fact that the multiplicity of V, in U, is the number of semistandard tableaux on j of type J. It can also be used to prove the assertion made in Exercise 4.6 that the representations corresponding to hooks are exterior powers of the standard representation. Exercise 4.44* (Pieri’s rule), Regard S, as a subgroup of G,,,, as usual. Let 2 be a partition of d and v a partition of d + m. Use Exercise 4.40 to show that the multiplicity of V, in the induced representation Ind(V,) is zero unless the Young diagram of / is contained in that of v, and then it is the number of ways to number the skew diagram lying between them with the numbers from f to m, increasing in both row and column. By Frobenius reciprocity, this is the same as the multiplicity of V, in Res(V,). When applied to d = 0(or 1), this implies again that the dimension of V, is the number of standard tableaux on the Young diagram of v. For a sampling of the many applications of these rules, see [Dia §7, §8]. Problem 4.45*. The Murnaghan- Nakayama rule gives an efficient inductive method for computing character values: If J is a partition of d, and g € G, is written as a product of an m-cycle and a disjoint permutation h € S,_,,, then ral) = L(— x), where the sum is over all partitions jt of d — m that are obtained from 4 by removing a skew hook of length m, and r(j1) is the number of vertical steps in the skew hook, i.e., one less than the number of rows in the hook. A skew hook for A is a connected region of boundary boxes for its Young diagram such that removing them leaves a smaller Young diagram; there is a one-to-one correspondence between skew hooks and ordinary hooks of the same size, as indicated: A=(7,6,5,5,4,4, b 1) n= (7,4,4,3,3, 1, 1,1) hook length = 9, r= 4 For example, if 1 has no hooks of length m, then y,(g) = The Murnaghan -Nakayama rule may be written inductively as follows: If gisa written as a product of disjoint cycles of lengths m,, m2, ..., m1, ips With the lengths m, taken in any order, then x,(q) is the sum }(— iy, where the sum is over all ways s to decompose the Young diagram of 4 by successively60 4, Representations of Sj: Young Diagrams and Frobenius's Character Formula removing p skew hooks of lengths m,, ..., m,, and r(s) is the total number of vertical steps in the hooks of s. (a) Deduce the Murnaghan—Nakayama rule from (4.41) and Exercise 4.16, using the Littlewood—Richardson rule. Or: (b) With the notation of Exercise 4.40, show that Vote 2 Vol) = SW Wayatban 0 Woe Exercise 4.46*. Show that Corollary 4.39 implies the “Snapper conjecture”: the irreducible representation V, occurs in the induced representation U, if and only if Sasdm for all j > t. oa fi Problem 4.47*, There is a more intrinsic construction of the irreducible representation V,, called a Specht module, which does not involve of the choice of a tableau; it is also useful for studying representations of S, in positive characteristic. Define a tabloid {T} to be an equivalence class of tableaux (numberings by the integers 1 to d) on A, two being equivalent if the rows are the same up to order. Then , acts by permutations on the tabloids, and the corresponding representation, with basis the tabloids, is isomorphic to U,. For each tableau 7, define an element £; in this representation space, by Ey = by{T} = ¥ santa) {qT}, the sum over the q that preserve the columns of T. The span of all E;’s is isomorphic to V,, and the E;’s, where T varies over the standard tableaux, form a basis. Another construction of V, is to take the subspace of the polynomial ring C[x,,..., Xa] spanned by all polynomials F,, where F; = | (x; — x), the product over all pairs i < j which occur in the same column in the tableau T: Exercise 4.48*. Let U; be the representation A -b,, which is the representation of G, induced from the tensor product of the alternating representations on the subgroup S, = G,, x --- x G,,, where # = 2’ is the conjugate partition. Show that the decomposition of U; is Uh = E Kyte Deduce that V, is the only irreducible representation that occurs in both U, and Uj, and it occurs in each with multiplicity one. Note, however, that in general A-c, # A-a,7 A°h, since A-c, may not be contained in A-a,.§4.3. Proof of Frobenius's Formula 61 Exercise 4.49*. With notation as in (4.41), if U’ = Viy___1, is the alternating representation of G,,, show that V, 0 M1, decomposes into a direct sum ®P,, the sum over all x whose Young diagram can be obtained from that of A by adding m boxes, with no two in the same row. Exercise 4.50, We have seen that A = CG, is isomorphic to a direct sum of mi, copies of V, = Ac,, where m, = dim V, is the number of standard tableaux on A. This can be seen explicitly as follows. For each standard tableau T on each J, let cz be the element of CG, constructed from T. Then A = @A cr. Indeed, an argument like that in Lemma 4.23 shows that c;-c7- = 0 whenever T and 7" are tableaux on the same diagram and T > T’, i.e. the first entry (reading from left to right, then top to bottom) where the tableaux differ has the entry of T larger than that of 7’. From this it follows that the sum LA-c, is direct. A dimension count concludes the proof. (This also gives another proof that the dimension of V, is the number of standard tableaux on 4, provided one verifies that the sum of the squares of the latter numbers is dl, cf. [Boe] or [Ke].) Exercise 4.51*, There are several methods for decomposing a tensor product of two representations of ,, which amounts to finding the coefficients C,,, in the decomposition VYi@V, = LCs for A, n, and v partitions of d. Since one knows how to express V, in terms of the induced representations U,, it suffices to compute V,@ U,, which is isomorphic to Ind(Res(V,)), restricting and inducing from the subgroup G, = G,, x G,, x +++; this restriction and induction can be computed by the Littlewood-Richardson rule. For d < 5, you can work out these coefficients using only restriction to G,_, and Pieri’s formula. (a) Prove the following closed-form formula for the coefficients, which shows in particular that they are independent of the ordering of the subscripts Anand v: C= a,00,0)0,(0, T 2(i) the sum over all afi) = itd 1,82! (b) Show that ca ft ituaa c _ ft ifmed au) = 0 otherwise, Awttr 10 otherwise. iy, -.., ig) with Lai, = d, and with «,(i) = x,(C,) and igh", Exercise 4.52*, Let Ry = R(G,) denote the representation ring, and set R = @ixo Ry. The outer product of (4.41) determines maps Ry @ Rn > Rasms62. 4, Representations of S,; Young Diagrams and Frobenius’s Character Formula which makes R into a commutative, graded Z-algebra. Restriction determines maps Ryam = R(Sysm) > R(Sq X Sq) = Ry @ Rms which defines a co-product 5: R + R@ R. Together, these make R into a (graded) Hopf algebra. (This assertion implies many of the formulas we have proved in this lecture, as welt as some we have not.) (a) Show that, as an algebra, R=Z[M,...,He---], where H, is an indeterminate of degree d; H, corresponds to the trivial representation of G,. Show that the co-product 6 is determined by O(H,) = H, @ 1 + Hy; @H, ++ +1 @H,. If we set A = Z[Hj,..., Ha...) = Ag we can identify Ay with the symmetric polynomials of degree d in k = d variables. The basic symmetric polynomials in A, defined in Appendix A therefore correspond to virtual representations of G,. (b) Show that E, corresponds to the alternating representation U’, and HieU, SV, By Uz. (c) Show that the scalar product ¢ , ) defined on A, in (A.16) corresponds to the scalar product defined on class functions in (2.11). (d) Show that the involution 9 of Exercise A.32 corresponds to tensoring a representation with the alternating representation U’. (e) Show that the inverse map from R, to A, takes a representation W to i F ytw(u)P, where 2(i) = i, 11 -i,!2!-, + igldls, The (inner) tensor product of representations of S, gives a map Ry @ Ry > R, which corresponds to an “inner product” on symmetric functions, some- times denoted +. (f) Show that 0 forj#i i) p peeps = Py ifj=i. Since these P" form a basis for A, @ Q, this formuta determines the inner product.LECTURE 5 Representations of Y, and GL,(F,) In this lecture we analyze the representation of two more types of groups: the alternat- ing groups W, and the linear groups GL,(F,) and SL,(F,) over finite fields. In the former case, we prove some general results relating the representations of a group to the representations of a subgroup of index two, and use what we know about the symmetric group; this should be completely straightforward given just the basic ideas of the preceding lecture. In the latter case we start essentially from scratch. The two sections can be read (or not) independently; neither is fogically necessary for the remainder of the book. §5.1: Representations of Uy §5.2: Representations of GL,(F,) and SL,(F,) §5.1. Representations of 2, The alternating groups M,, d > 5, form one of the infinite families of simple groups. In this section, continuing the discussion of §3.1, we describe their irreducible representations, The basic method for analyzing representations of U, is by restricting the representations we know from ©,. In general when H is a subgroup of index two in a group G, there is a close relationship between their representations. We will see this phenomenon again in Lie theory for the subgroups SO, of the orthogonal groups O,. Let U and U’ denote the trivial and nontrivial representation of G obtained from the two representations of G/H. For any representation V of G, let V' =V@U'; the character of V' is the same as the character of V on elements of H, but takes opposite values on elements not in I, In particular, Res V' = Res§v.64 5. Representations of Wf, and GL,(F,) If W is any representation of H, there is a conjugate representation defined by conjugating by any element ¢ of G that is not in H; if y is the character of W, the character of the conjugate is hr (tht), Since ¢ is unique up to multiplication by an element of H, the conjugate representation is unique up to isomorphism. Proposition 5.1. Let V be an irreducible representation of G,and let W = ResGV be the restriction of V to H. Then exactly one of the following holds: (I) Visnot isomorphic to V’; W is irreducible and isomorphic to its conjugate; Indgw=V@V'. Q) VZV W = W@W", where W' and W" are irreducible and conjugate but not isomorphic; Ind§ W' = Ind§ W” = V. Each irreducible representation of H arises uniquely in this way, noting that in case (1) V’ and V determine the same representation. Proor. Let x be the character of V. We have G|=2Hl= ¥ Ixy? + oP. 1G] = 21h1 on (hy 3, ol Since the first sum is an integral multiple of |/|, this multiple must be I or 2, which are the two cases of the proposition. This shows that W is either irreducible or the sum of two distinct irreducible representations W’ and W". Note that the second case happens when x(t) = 0 for all t ¢ H, which is the case when V’ is isomorphic to V. In the second case, W’ and W” must be conjugate since W is self-conjugate, and if W‘ and W* were self-conjugate V would not be irreducible. The other assertions in (I) and (2) follow from the isomorphism Ind(Res V) = V @ (U ® U’) of Exercise 3.16. Similarly, for any representation W of H, Res(Ind W) is the direct sum of W and its conjugate— as follows say from Exercise 3.19—from which the last statement follows readily. oO Most of this discussion extends with little change to the case where H is a normal subgroup of arbitrary prime index in G, cf. [B-tD, pp. 293-296]. Clifford has extended much of this proposition to arbitrary normal subgroups of finite index, cf. [Dor, §14]. There are two types of conjugacy classes ¢ in H: those that are also conjugacy classes in G, and those such that c Uc’ is a conjugacy class in G, where c’ = tct”', t ¢ H; the latter are called split. When W is irreducible, its character assumes the same values—those of the character of the representa- tion V of G that restricts to W—on pairs of split conjugacy classes, whereas in the other case the characters of W' and W” agree on nonsplit classes, but they must disagree on some split classes. If yy-(c) = Yw-(c’) = x, and yy(c’) = Xw-(c) = y, we know the sum x + y, since it is the value of the character of the representation V that gives rise to W’ and W” on c Uc’, Often the exact values of x and y can be determined from orthogonality considerations.§5.1. Representations of Uy 65 Exercise 5.2*. Show that the number of split conjugacy classes is equal to the number of irreducible representations V of G that are isomorphic to V', or to the number of irreducible representations of H that are not isomorphic to their conjugates. Equivalently, the number of nonsplit classes in H is same as the number of conjugacy classes of G that are not in H. We apply these considerations to the alternating subgroup of the symmetric group. Consider restrictions of the representations V, from , to %,. Recall that if 1’ is the conjugate partition to A, then Vy = VY, @U', with U’ the alternating representation. The two cases of the proposition correspond to the cases (1) 1’ #4 and (2) A’ =A. If 4 £A, let W, be the restriction of V, to U,. If A’ = A, let Wy and W' be the two representations whose sum is the restriction of V,. We have Ind = VY@My, Res ¥y= Res Vy = W, when i # A, Ind Wij=Ind Wr =, Resly= W@W when i’ Note that # {self-conjugate representations of S, = # {symmetric Young diagrams} = # {split pairs of conjugacy classes in %,} = # {conjugacy classes in S, breaking into two classes in %,}. Now a conjugacy class of an element written as a product of disjoint cycles is split if and only if there is no odd permutation commuting with it, which is equivalent to all the cycles having odd length, and no two cycles having the same length. So the number of self-conjugate representations is the number of partitions of d as a sum of distinct odd numbers. In fact, there is a natural correspondence between these two sets: any such partition corresponds to a symmetric Young diagram, assembling hooks as indicated: If A is the partition, the lengths of the cycles in the corresponding split conjugacy classes are q, = 2A, — 1, q2 = 2A, — 3,43 = 243 —5,....66 5, Representations of %, and GL,(F,) For a self-conjugate partition A, let 7’, and yi denote the characters of Wy and Wy, and let c and c’ bea pair of split conjugacy classes, consisting of cycles of odd lengths q, >, >-*: > q,. The following proposition of Frobenius completes the description of the character table of %,. Proposition 5.3. (1) If c and c’ do not correspond to the partition A, then wal) = rile’) = XH) = Ale’) = daleve’). (2) If c and ¢’ correspond to A, then BO=UGE)=x% nC)=GO=), with x and y the two numbers Y(-I t Jaa as and m = H([]q,— = 4(d — 2. For example, if d = 4 and A = (2, 2), we have r = 2, q, = 3, q, = 1, and x and y are the cube roots of unity; the representations Wj and W7' are the representations labeled U' and U” in the table in §2.3. For d = 5,4 = (3, 1, 1), r= 1, q,=5, and we find the representations called Y and Z in §3.1. For d <7, there is at most one split pair, so the character table can be derived from orthogonality alone. Note that since only one pair of character values is not taken care of by the first case of Frobenius’s formula, the choice of which representation is Wi and which W/' is equivalent to choosing the plus and minus sign in (2). Note also that the integer m occurring in (2) is the number of squares above the diagonal in the Young diagram of 1. We outline a proof of the proposition as an exercise: Exercise 5.4*, Step 1. Let q = (q, > --: > q,) be a sequence of positive odd integers adding to d, and let c’ = c'(q) and c” = c"(q) be the corresponding conjugacy classes in 2. Let A be a seif-conjugate partition of d, and let zy; and x4 be the corresponding characters of Uy. Assume that 7’, and 73 take on the. same values on each element of 2, that is not in c’ or c’. Let u= y4(c') = alc") and v = x4(c") = x3(c’). (i) Show that u and v are real when m = $£(q, — Lis even, and # = v when mis odd, (ii) Let 3 = x4 — 4. Deduce from the equation (9, 9) = 2 that ju — v|? = Gr (iii) Show that A is the partition that corresponds to q and that u + v = (-1)", and deduce that « and v are the numbers specified in (2) of the proposition. Step 2. Prove the proposition by induction on d, and for fixed d, look at that q which has smallest q,, and for which some character has values on the classes c’(q) and c”(q) other than those prescribed by the proposition. uN§5.2. Representations of GL,(F,) and SL,(F,) 67 (i) Ifr = 1,s0q, = d = 2m + 1, the corresponding self-conjugate partition is 4 = (+ 1, 1,..., 1). By induction, Step 1 applies to ', and 7%. (ii) Ifr > 1, consider the imbedding H = %,, x W%y_,, © G = W,, and let X' and X” be the representations of G induced from the representations Wie W; and WW, where W; and W," are the representations of U,, corresponding to qj, i-e., to the self-conjugate partition (4(q, — 1), I, ..., 1) of 41; W; is one of the representations of U,_,, corresponding to (q, ..., g,); and denotes the external tensor product (see Exercise 2.36). Show that X’ and X" are conjugate representations of U,, and their characters y' and y” take equal values on each pair of split conjugacy classes, with the exception of c’(q) and c”(q), and compute the values of these characters on c’(q) and c"(q). (iii) Let 9 = x’ — x”, and show that (9, 9) = 2. Decomposing X’ and X” into their irreducible pieces, deduce that X’ = Y ® Wj and X” = Y @ WY for some self-conjugate representation Y and some self-conjugate partition A of d. (iv) Apply Step 1 to the characters 7, and 73, and conclude the proof. Exercise 5.5*. Show that if d > 6, the only irreducible representations of Q, of dimension less than d are the trivial representation and the (n — 1)- dimensional restriction of the standard representation of G,. Find the excep- tions for d < 6. We have worked out the character tables for all G, and U, for d < 5. With the formulas of Frobenius, an interested reader can construct the tables for a few more d—until the number of partitions of d becomes large. §5.2. Representations of GL,(F,) and SL,(F,) The groups GL,(F,) of invertible 2 x 2 matrices with entries in the finite field F, with q elements, where q is a prime power, form another important series of finite groups, as do their subgroups SL2(F,) consisting of matrices of determinant one. The quotient PGL,(F,) = GL,(F,)/F¢ is the automor- phism group of the finite projective line P'(F,). The quotients PSL,(F,} = SL,(F,)/{ +1} are simple groups if q # 2, 3 (Exercise 5.9). In this section we sketch the character theory of these groups. We begin with G = GL;(F,). There are several key subgroups: coef Mew (G Dh (This “Borel subgroup” B and the group of upper triangular unipotent matrices N will reappear when we look at Lie groups.) Since G acts transitively on the projective line P'(F,), with B the isotropy group of the point (1:0), we have 1G] = |BI-|P'()1 = q — 1)?4@ + 1).68 5. Representations of 2%, and GL,(F,) We will also need the diagonal subgroup o-{(0 )-r xe where we write F for F,. Let F° = F,2 be the extension of F of degree two, unique up to isomorphism. We can identify GL,(F,) as the group of all F-linear invertible endomorphisms of F’. This makes evident a large cyclic subgroup K = (F')* of G. At least if q is odd, we may make this isomorphism explicit by choosing a generator ¢ for the cyclic group F* and choosing a square root Je in F’, Then t and Je form a basis for F’ as a vector space over F, so we can make the identification: _S{* o\ sens x ev) . Kf 2)psen ¢ ) boxtays K is acyclic subgroup of G of order q? — 1. We often make this identification, leaving it as an exercise to make the necessary modifications in case qis even. The conjugacy classes in G are easily found: Representative No. Elements in Class No. Classes x 0 a=(5 °) 1 q-t n=(3 ) e-1 qt x 0 -Nq-2 con(5 pkey ata ae _{* ¥ 2 aq —') tas=( ).y#0 v-4 I ol Here c,,, and c, , are conjugate by ( -1 oF and d, ,and d, _, are conjugate a by any c “ . To count the number of elements in the conjugacy class of b,, look at the action of G on this class by conjugation; the isotropy group . f(a b . is (5 \t so the number of elements in the class is the index of this group in G, whichis q? — 1. Similarly the isotropy group for c,_, is D, and the isotropy group for d,,,is K. To see that the classes are disjoint, consider the eigenvalues and the Jordan canonical forms. Since they account for |G| elements, the list is complete. There are q? — 1 conjugacy classes, so we must find the same number of irreducible representations. Consider first the permutation representation of G on P'(F), which has dimension q + 1. It contains the trivial representation;§5.2. Representations of GLa (F,) and SLa(F,) 0 let V be the complementary q-dimensional representation. The values of the character y of V on the four types of conjugacy classes are x(a,) = q, x(b,) = 0, X(Ce,y) = 1, x(dy.y) = — 1, which we display as the table: Vio gq 0 1 =I Since (x, x) = L, V is irreducible. For each of the q—I characters «:F* + C* of F*, we have a one- dimensional representation U, of G defined by U,(g) = a(det(g)). We also have the representations V, = V @ U,. The values of the characters of these representations are Ug a(x)? a(x)? alxar(y) x(x? — ey?) Vas qae(x)? 0 a(xja(y) — —a(x? — ey?) Note that if we identify (; y ) with {=x + y/é in F, then x x? — ay? = det (; ”) = Normyye(0) = CC = CU". The next place to look for representations is at those that are induced from large subgroups. For each pair a, B of characters of F*, there is a character of the subgroup B: B- B/N = D = F* x F* 4 C* x C* 4 C4, b which takes (¢ ’) to «(a)B(d). Let W, y be the representation induced from Bto G by this representation; this is a representation of dimension [G: B] = q+ 1. By Exercise 3.19 its character values are found to be: Wao? (9+ NalX)BX) (DBR) a(x) + a(V)B(x) 0 We see from this that W, 4 Wh. that W,,, = U,@ V,, and that for a # B the representation is irreducible. This gives }(q — 1)(q — 2) more irreducible representations, of dimension q + 1. Comparing with the list of conjugacy classes, we see that there are }q(q — 1) irreducible characters left to be found. A natural way to find new characters is to induce characters from the cyclic subgroup K. For a representation go: K =(F)* > Ct, the character values of the induced representation of dimension [G: K] = q — tare Ind(g): g(q- De) 9 0 — oll) + ey Here again (=x + y./eeK =(F)*. Note that Ind(%) = Ind(g), so the representations Ind(g) for 9? # @ give 4q(q — |) different representations.10 5. Representations of U, and GL(F,) However, these represenations are not irreducible: the character y of Ind(@) satisfies (x, x)= q — | if op’ ¥ @, and otherwise (, x) = q. We will have to work a little harder to get irreducible representations from these Ind(g). Another attempt to find more representations is to look inside tensor products of representations we know. We have V, @ U, = V,,,and W, 4@ U, = Wy.» 80 there are no new ones to be found this way. But tensor products of the V,’s and W, 9's are more promising. For example, V @ W,,; has character values: VOM: aa t Nar) 0 a(x) +a(y) 0 We can calculate some inner products of these characters with each other to estimate how many irreducible representations each contains, and how many they have in common. For example, (trem. Xm,,) = 2 Cine Xn...) = 1 ele = a, (ivew.> vow.) =9 +3, (Lrem.y Xnae) = 4 if ple = % Comparing with the formula (Jina) inde) = 4 — 1, one deduces that V@W,,, and Ind(g) contain many of the same representations. With any luck, Ind(g) and W,, should both be contained in V @ W,,,. This guess is easily confirmed; the virtual character Ae = kvew., — Xm. — Xinatoy takes values (q — !)a(x), —a(x), 0, and —(@(C) + e(¢)*) on the four types of conjugacy classes. Therefore, (y,,7,) = 1, and x,(1)=q—1!>0, so x, is, in fact, the character of an irreducible subrepresentation of V@ Wz, of dimension q — 1. We denote this representation by X,. These 4q(q — 1) representations, for ¢ # 9%, and with X, = X,«, therefore complete the list of irreducible representations for GL,(F). The character table is ' v-t Pra v-4 x 0 xt x 0 x ey own ta-G) AG) er) a6 De U, a(x?) a(x?) ae(xy) att) % a(x?) o (xy) ~a(C") Wee | (a+ tax) exy a(x) f(x) a(x) B(y) + a(y)ACx) 0 Xe! @— Nex) —9(x) 0 —(0) + oC) Exercise 5.6, Find the multiplicity of each irreducible representation in the representations V @ W, , and Ind(g).§5.2, Representations of GL, (F,) and SL,(F,) nN Exercise 5.7. Find the character table of PGL,(F) = GL,(F)/F*. Note that its characters are just the characters of GL,(F) that take the same values on elements equivalent mod F*. We turn next to the subgroup SL,(F,) of 2 x 2 matrices of determinant one, with q odd. The conjugacy classes, together with the number of elements in each conjugacy class, and the number of conjugacy classes of each type, are Representative No, Elements in Class No. Classes » (69 : : a eh . nh) nr) 6 (32) © (4) 0 ( Saxe a a+!) ®) (x xe tl aq — 1) The verifications are very much as we did for GL,(F,). In (7), the classes of -1 ( °) and (5 °) are the same. In (8), the classes for (x, y) and (x, —y) x x are the same; as before, a better labeling is by the element in the cyclic group C= Ce (Fy ct =I}; the elements + 1 are not used, and the classes of and (~' are the same. The total number of conjugacy classes is q + 4, so we turn to the task of finding q + 4 irreducible representations. We first see what we get by restrict- ing representations from GL,(F,). Since we know the characters, there is no problem working this out, and we simply state the results; (1) The U, all restrict to the trivial representation U. Hence, if we restrict any representation, we will get the same for all tensor products by U,'s.n 5. Representations of W, and GL,(F,) (2) The restriction V of the V,’s is irreducible. (3) The restriction W, of W, , isirreducible if” 4 land W, = W, when f = « or fi = a7". These give 4(q — 3) irreducible representations of dimension +1. (3) Let t denote the character of F* with 1? = 1,1 # 1. The restriction of Wi, is the sum of two distinct irreducible representations, which we denote W' and W”. (4) The restriction of X, depends only on the restriction of g to the subgroup C,and g and g~! determine the same representation. The representation is irreducible if g? # 1. This gives }(q — 2) irreducible representations of dimension q — 1. (4’) If y denotes the character of C with ? = 1, y # 1, the restriction of Xy is the sum of two distinct irreducible representations, which we denote X' and X”. Altogether this list gives q + 4 distinct irreducible representations, and it is therefore the complete list. To finish the character table, the problem is to describe the four representations W’, W”, X’, and X”. Since we know the sum of the squares of the dimensions of all representations, we can deduce that the sum of the squares of these four representations is q? + i, which is only possible if the first two have dimension 3(q + 1) and the other two 4(q — 1). This is similar to what we saw happens for restrictions of representations to subgroups of index two. Although the index here is larger, we can use what we know about index two subgroups by finding a subgroup H of index two in GL,(F,) that contains SL,(F,), and analyzing the restrictions of these four Tepresentations to H. For H we take the matrices in GL,(F,) whose determinant is a square. The representatives of the conjugacy classes are the same as those for GL2(F,), including, of course, only those representatives whose determinant is a square, but we must add classes represented by the elements ( ‘). xe F*. These x, 0 ‘) in GL,(F,), but not in H. These are the q — | split conjugacy classes. The procedure of the preceding section can be used to work out all the representations of H, but we need onlya little of this. Note that the sign representation U’ from G/H is U,, so that W, = W,,@ U' and X, = X, @ U’; their restrictions to H split into sums of conju- gate irreducible representations of half their dimensions. This shows these representations stay irreducible on restriction from H to SL,(F,), so that W’ and W” are conjugate representations of dimension 4(q + t), and X’ and X” are conjugate representations of dimension 4(q — 1), In addition, we know that their character values on all nonsplit conjugacy classes are the same as half the characters of the representations W,,, and X,, respectively. This is all the information we need to finish the character table. Indeed, the only values not covered by this discussion are are conjugate to the elements (§5.2. Representations of GL,(F,) and SL,(E,) 3 Gi) Gi) Con) (04) w s t s v Ww" t s C s x’ u v a’ v x" v u v ul The first two rows are determined as follows. We know that s + t= 1 1) <1 tn addition, since (1 ')" = (1 ~")is conjugate t Ins o We! In } since | ¢ I Flo 4 is conjugate to 1 } ' ; 1) otherwise, and since x(g7') = xg) for any character, we conclude that s and ¢ are real if q = 1 mod(4), and s = Tif q = 3 mod(4). In addition, since —e acts as the identity or minus the identity for any irreducible representation (Schur’s lemma), x(—9) = x19) xia e) for any irreducible character x. This gives the relations s’ = t(—1)s and t= 1(—1)t. Finally, applying the equation (x, x) = 1 to the character of W” gives a formula for sf + (5. Solving these equations gives s, t= 4+ 4./ogq, where w = t(—1) is 1 or —t according as q = 1 or 3 mod(4). Similarly one computes that u and v are —4 + 4 Jog. This concludes the computations needed to write out the character table. if q is congruent to 1 modulo 4, and to Exercise 5.8. By considering the action of SL,(F,) on the set P'(F,), show that SL,(F,) © Ss, PSL2(F,) = Wg, and SL,(F,) = Ws. Exercise 5.9*. Use the character table for SL,(F,) to show that PSL,(F,) is a simple group if q is odd and greater than 3. Exercise 5.10. Compute the character table of PSL,(F,), either by regarding it as a quotient of SL,(F,), or as a subgroup of index two in PGL,(F,). Exercise 5.11*. Find the conjugacy classes of GL,(F,), and compute the char- acters of the permutation representations obtained by the action of GL;(F,) on (i) the projective plane P?(F,) and (ii) the “flag variety” consisting ofa point on a Sine in P(F,). Show that the first is irreducible and that the second is a sum of the trivial representation, two copies of the first representation, and an irreducible representation. Although the characters of the above groups were found by the early pioneers in representation theory, actually producing the representations in anatural way is more difficult. There has been a great deal of work extending4 5. Representations of %, and GL,(F,) this story to GL,(F,) and SL,(F,) for n > 2 (cf. [Gr]), and for corresponding groups, called finite Chevalley groups, related to other Lie groups. For some hints in this direction see [Hu3}], as well as [Ti2]. Since all but a finite number of finite simple groups are now known to arise this way (or are cyclic or alternating groups, whose characters we already know), such representations play a fundamental role in group theory. In recent work their Lie-theoretic origins have been exploited to produce their representations, but to tell this story would go far beyond the scope of these lecture(r)s.LECTURE 6 Weyl’s Construction In this lecture we introduce and study an important collection of functors generalizing the symmetric powers and exterior powers. These are defined siinply in terms of the Young symmetrizers c, introduced in §4: given a representation V ofan arbitrary group G, we consider the dth tensor power of V, on which both G and the symmetric group on d letters act. We then take the image of the action of c, on ¥®; this is again a representation of G, denoted S,(V). This gives us a way of generating new representa- tions, whose main application will be to Lie groups: for example, we will generate all representations of SLC by applying these to the standard representation C” of SL,C. While it may be easiest to read this material while the definitions of the Young symmetrizers are still fresh in the mind, the construction will not be used again until §15, so that this lecture can be deferred until then. §6.1: Schur functors and their characters §6.2: The proofs §6.1. Schur Functors and Their Characters For any finite-dimensional complex vector space V, we have the canonical decomposition V@V =Sym’V @/*V. The group GL(V) acts on V @ V, and this is, as we shall soon see, the decom- position of V @ V into a direct sum of irreducible GL(V)-representations. For the next tensor power, V@V@V =Sym*V ®A*V @ another space. We shall see that this other space is a sum of two copies of an irreducible6 6. Weyl’s Construction GL(¥)-representation. Just as Sym*V and A‘V are images of symmetrizing operators from V®4 = V@ V@:-@ V to itself, so are the other factors. The symmetric gtoup S, acts on V4, say on the right, by permuting the factors (01 B+ @ 4) = Vg @-"* @ Yotay- This action commutes with the left action of GL(V). For any partition A of d we have from the last Jecture a Young symmetrizer c, in CG,. We denote the image of c, on V® by S,V: S1V = Im(cilyea) which is again a representation of GL(V). We call the functor! V ~~ S,V the Schur functor or Wey! module, or simply Weyl’s construction, corresponding to A. It was Schur who made the correspondence between representations of symmetric groups and representations of general linear groups, and Weyl who made the construction we give here.” We will give other descriptions later, cf. Exercise 6.14 and §15.5. For example, the partition d = d corresponds to the functor V ~v Sym¢ V, and the partition d = 1 +--+ + I to the functor V~~'A‘V. We find something new for the partition 3 = 2 + 1. The corresponding syminetrizer c, is Ca.) = 1 + ey ~ Cay ~ C32» so the image of c; is the subspace of V ® spanned by all vectors Dy OV, OU; + 02 @ 0, @ v3 — v3 Ov, Ov, — 13 @V, @ P2.. If ?V @ V is embedded in V®? by mapping (04 A 03) @ vat 0, @ 0; @ 03 — 43 @V2WM, then the image of c, is the subspace of \?V @ V spanned by all vectors (21 A 03) @ v2 + (bz A ¥3)@v4- It is not hard to verify that these vectors span the kernel of the canonical map from A?V @ V to A*V, so we have Sa,nV = Ker(NV@ VAY). (This gives the missing factor in the decomposition of V®3.) Note that some of the S, V can be zero if V has small dimension. We will see that this is the case precisely when the number of rows in the Young diagram of / is greater than the dimension of V. ' The functorialily means simply that a linear map g: V-+ W of vector spaces determines a linear map S,(9): S, V+ S,1Y, with S,(9 © ¥) = S.(9) © S,(W) and S,(Id)) = Ids,y. ? The notion goes by a variety of names and notations in the literature, depending on the context, Constructions differ markedly when not over a field of characteristic zero; and many authors now Parametrize hem by the conjugate partitions. Our choice of notation is guided by the corre- spondence between these functors and Schur polynomials, which we will see are their characters.§6.1. Schur Functors and Their Characters. wie When G = GL(V), and for important subgroups G < GL(V), these S,V give many of the irreducible representations of G; we will come back to this later in the book. For now we can use our knowledge of symmetric group representations to prove a few facts about them—in particular, we show that they decompose the tensor powers V®, and that they are irreducible repre- sentations of GL(V). We will also compute their characters; this will eventually be seen to be a special case of the Weyl character formula. Any endomorphism g of V gives rise to an endomorphism of S, V. In order to tell what representations we get, we will need to compute the trace of this endomorphism on S,V; we denote this trace by y5,(g). For the computation, let x,,..., x, be the eigenvalues of g on V, k = dim V. Two cases are easy. For A= (a), SV =Sym*V, X5,q7(G) = Hal%14---1 Xs (6.1) where H,(x,,...,X,) is the complete symmetric polynomial of degree d. The definition of these symmetric polynomials is given in (A.t) of Appendix A. The truth of (6.1) is evident when g is a diagonal matrix, and its truth for the dense set of diagonalizable endomorphisms implies it for all endomorphisms; ‘or one can see it directly by using the Jordan canonical form of g. For A=(l,..., l), we have similarly Sana V= NV s.r) = Ears Meh (6.2) with E,(x,,...,,) the elementary symmetric polynomial [see (A.3)]. The polynomials H, and E, are special cases of the Schur polynomials, which we denote by S, = 5,(x,, ..., x4). As A varies over the partitions of d into at most k parts, these polynomials S, form a basis for the symmetric polynomials of degree d in these k variables. Schur polynomials are defined and discussed in Appendix A, especially (A.4)-(A.6). The above two formulas can be written X8v(9) = SxlXqy 005 Xe) for A= (d)and A= (1,..., 1). We will show that this equation is valid for all A: Theorem 6.3. (1) Let k=dim V. Then S,V is zero if As, #0. If A= (A, 201° > Ay 2 0), then dim §,V = $,(1,..., 1) = Ae 1si Sym**'V), and similarly for the conjugate partition, Sau..nl = Ker(NV®@ VAY), Exercise 6.10*. One can also derive the preceding decompositions of tensor products directly from corresponding decompositions of representations of symmetric groups. Show that, in fact, S, V @ S,V corresponds to the “inner product” representation V, o V, of S,,,, described in (4.41). Exercise 6.11%. (a) The Littlewood-Richardson rule also comes into the de- composition of a Schur functor of a direct sum of vector spaces V and W. This80 6. Weyl's Construction generalizes the well-known identities Sym"(V ® W) = ® (Sym*V @ Sym? W¥), athen NVOW)= OB AVON). an Prove the general decomposition over GL(V) x GL(W): SAV BW) = © NlSiV @ SW), the sum over all partitions A, 4 such that the sum of the numbers partitioned by A and yp is the number partitioned by v. (To be consistent with Exercise 2.36 one should use the notation @ for these “external” tensor products.) (b) Similarly prove the formula for the Schur functor of a tensor product: S(V® W) = BD CylS.V ® SW), where the coefficients C,,,, are defined in Exercise 4.51.In particular show that Sym4(V @ W) = O®S,V@S,W, the sum over all partitions A of d with at most dim V or dim W rows. Replacing W by W*, this gives the decomposition for the space of polynomial functions of degree d on the space Hom(V, W) over GL(V) x GL(W). For variations ‘on this theme, see [Ho3]. Similarly, NV QW) = OSV OS, the sum over partitions 4 of d with at most dim V rows and at most dim W columns, Exercise 6.12. Regarding GL, € =GL,€ x {1} < GL,€ x GL, Co Gla the preceding exercise shows how the restriction of a representation de- composes: Res(S,(C"*")) = Y (Naw dim S,(C"))S,(C"). In particular, for m = 1, Pieri’s formuta gives Res(S,(C""") = OSC"), the sum over all 2 obtained from v by removing any number of boxes from its Young diagram, with no two in any column. Exercise 6.13*. Show that for any partition w= (jt), ..., u,) of d, N'VO@NVQ-@NV = PKuSeh, where K,, is the Kostka number and 2’ the conjugate of 2.§6.1. Schur Functors and Their Characters 81 Exercise 6.14*. Let 1 = 2’ be the conjugate partition. Put the factors of the dth tensor power V ® in one-to-one correspondence with the squares of the Young diagram of A. Show that S, V is the image of this composite map: Q(N'V) + Q(B"V) > VE* > @, (@*V) > @, Sym*v), the first map being the tensor product of the obvious inclusions, the second grouping the factors of V®4 according to the columns of the Young diagram, the third grouping the factors according to the rows of the Young diagram, and the fourth the obvious quotient map. Alternatively, S, V is the image of a composite map @i(Syn*V) + Bs (@*V) > VE > BW, (@"V) > @, (NV). In particular, S, V can be realized as a subspace of tensors in V® that are invariant by automorphisms that preserve the rows of a Young tableau of A, or a subspace that is anti-invariant under those that preserve the columns, but not both, cf. Exercise 4.48. Problem 6.15*. The preceding exercise can be used to describe a basis for the space S,V. Let v,,.-., 0; be a basis for V. For each semistandard tableau 7° on A, one can use it to write down an element v7 in ()(N"V); 0; is a tensor product of wedge products of basis elements, the ith factor in A“V being the wedge product (in order) of those basis vectors whose indices occur in the ith column of 7: The fact to be proved is that the images of these elements vy under the first composite map of the preceding exercise form a basis for S,V. At the end of Lecture 15, using more representation theory than we have at the moment, we will work out a simple variation of the construction of S,V which will give quick proofs of refinements of the preceding exercise and problem. Exercise 6.16*. The Pieri formula gives a decomposition Sym'V @ Sym'V =D Syraan¥s the sum over 0 < a < d. The left-hand side decomposes into a direct sum of Sym?(Sym*V) and A?(Sym*V). Show that, in fact, Sym?(Sym*V) = Sy2u,0)V ® Syre-2,2)V ® Se2a-4,4V Os M Gym‘V) = Se4-1,y¥ ® See-a,3V ® Seae-s, VO" - Similarly using the dual form of Pieri to decompose A‘V @ A’V into the sum @S,V, the sum over all A = (2,..., 2, 1,..., B) consisting of d — a 2's and 2a I's, 0 Ady, given by right multiplication by a,,,, or the image of the map Aajy, > Aby, given by right multiplication by b,,,. The decomposition of Vajq into irreducible representations is ) Yan = E Nava Yor Similarly there are skew Schur functors S,,,, which take a vector space V to the image of c,,, on V ®; equivalently, S,,, V is the image of a natural map (generalizing that in the Exercise 6.14) (vi) (NV) + VOF > Dy (Syms V), or (vii) Qj Gym*-HV) > VE4s @, (AU-HY)., Given a basis v,, ..., for V and a standard tableau Ton 2/p, one can write down an element vz in @,(N'-4V); for example, corresponding to the dis- played tableau, vz = v, @ v, @ (v, A v5). A key fact, generalizing the result of Exercise 6.15, is that the images of these elements under the map (vi) form a basis for Sy),.V. The character of S,,V is given by the Schur function S,,,: if g is an endomorphism of V with eigenvalues x,, ..., X,, then (viii) L8aqv (9) = Sryp(% 14 +++ Xe) In terms of basic Schur functors, (ix) Say V = LN Seh Exercise 6,20*. (a) Show that if 2 = (p, 9), Sip, V is the kernel of the contrac- tion map Cp.q: Sym?V @ SymtV -» Sym?*!V @ Syme p. (b) If A = (p, 4, 1), show that Sq,,,,)V is the intersection of the kernels of two contraction maps c,,¢@ f,and {,, @ ¢,,-, where I, denotes the identity map on Sym! V. In general, for 2=(A,,..., Ay), SV c Sym" V @---@Sym*Y is the inter- section of the kernels of the k — 1 maps Ve = 1, O° Bly, @Ca.24,@ hy 8 Oly Usisk=. (c) For A =(p,1,...., 1), show that S,V is the kernel of the contraction map: Soot.) V = Ker(Sym’V @ MPV > Sym? V@AT PY). In general, for any choice of a between | and k — 1, the intersection of84 6. Weyl's Construction the kernels of all y, except y, is S,V@S,V, where ¢ =(A,,...,4,) and T= (Agsts «+. 44) 80 S,V is the kernel of a contraction map defined on S,V @ §,V. For example, ifa is k — 1,and weset r = A,, Pieri’s formula writes S,V@ Sym’ as a direct sum of S,V and other factors S,V; the general assertion in (b) is equivalent to the claim that S, V is the only factor that is in the kernel of the contraction, ie, S,V = Ker(Sa,,....4-V ® Sym'V > V 4-14) @ Sym! V), These results correspond to writing the representations V, < U, of the sym- metric group as the intersection of kernels of maps to Ug, ayttsayay—teeeos da Exercise 6.21. The functorial nature of Wey!’s construction has many conse- quences, which are not explored in this book. For example, if E, isa complex of vector spaces, the tensor product E%* is also a complex, and the symmetric group GS, acts on it; when factors in E, and E, are transposed past each other, the usual sign (— 1)°* is inserted. The image of the Young symmetrizer c, is a complex S,(E,), sometimes called a Schur complex. Show that if E, is the complex E_,; = V+ Eg = V, with the boundary map the identity map, and A =(d), then S,(F,) is the Koszul complex 05 N+ AT @S! 5AM? @ $2 ++ +N @ SH > S40, where A! = A'V, and S! = Sym/V. §6.2. The Proofs We need first a small piece of the general story about semisimple algebras, which we work out by hand, For the moment G can be any finite group, although our application is for the symmetric group, If U is a right module over A = CG, let B = Hom,(U, U) = {g: U > U: o(v-g) = e(v)-g, Wu € U,g € G}. Note that B acts on U on the left, commuting with the right action of A; B is called the commutator algebra. IfU = @ U2" is an irreducible decomposition with U, nonisomorphic irreducible right A-modules, then by Schur’s Lemma 17 B= Q Homg(", uP") = Bi Mn (C), where M,,(C) is the ring of n, x n, complex matrices. If W is any left A-module, the tensor product U @, W = U @c W/subspace generated by {va @ w — v@ aw} isa left B-module by acting on the first factor: b-(v @ w) = (b- 1) @w.§6.2. The Proofs 85 Lemma 6.22. Let U be a finite-dimensional right A-module. (i) For any c € A, the canonical map U @, Ac + Uc is an isomorphism of left B-modules. (ii) If W = Ac is an irreducible left A-module, then U @, W = Uc is an irreducible left B-module. (iii) If W, = Ac, are the distinct irreducible left A-modules, with m, the dimension of W,, then U = BU @ MIO" = GD Uey™ is the decomposition of U into irreducible left B-modules. Proor. Note first that Ac is a direct summand of A as a left A-module; this is a consequence of the semisimplicity of all representations of G (Proposition 1.5). To prove (i), consider the commutative diagram U@Q,A —*> UG Ac — UQA | U + Ue & U where the vertical maps are the maps v @ a+ v-a; since the left horizontal maps are surjective, the right ones injective, and the outside vertical maps are isomorphisms, the middle vertical map must be an isomorphism. For (ii), consider first the case where U is an irreducible A-module, so B=C. It suffices to show that dim U @, W < 1. For this we use Proposition 3.29.to identify A with a direct sum (Py-, M,,,€ of r matrix algebras. We can identify W with a minimal left ideal of A. Any minimal ideal in the sum of matrix algebras is isomorphic to one which consists of r-tuples of matrices which are zero except in one factor, and in this factor are all zero except for one column. Similarly, U can be identified with the right ideal of r-tuples which are zero except in one factor, and in that factor all are zero except in one row. Then U @, W will be zero unless the factor is the same for U and W, in which case U @, W can be identified with the matrices which are zero except in one row and column of that factor. This completes the proof when U is irreducible. For the general case of (ii), decompose U = @); U®" into a sum of irreducible right A-modules, so U @, W = (,(U, @, W)®" = C®™ for some k, which is visibly irreducible over B = @ M,(C). Part (iii) follows, since the isomorphism A = (R) W,%™ determines an iso- morphism UZU QA ZUR (DM) = BU @ M)™. Q To prove Theorem 6.3, we will apply Lemma 6.22 to the right CS,-module U = V®. That lemma shows how to decompose U as a B-module, where B86 6. Weyl’s Construction is the algebra of all endomorphisms of U that commute with all permuta- tions of the factors. The endomorphisms of U induced by endomorphisms of V are certainly in this algebra B. Although B is generally much larger than End(V), we have Lemma 6.23, The algebra B is spanned as a linear subspace of End(V®) by End(V). A subspace of V® is a sub-B-module if and only if it is invariant by GL(V). Proor. Note that if W is any finite-dimensional vector space, then Sym“W is the subspace of W® spanned by all w! = d!w@ --- @ w as w runs through W. Applying this to W = End(V) = V* @ V proves the first statement, since End(V®) =(V*)®! @ V® = W®4, with compatible actions of S,. The second follows from the fact that GL(V) is dense in End(V). Oo We turn now to the proof of Theorem 6.3. Note that S,V is Uc,, so parts (2) and (4) follow from Lemmas 6.22 and 6.23. We use the same methods to give a rather indirect but short proof of part (3); for a direct approach see Exercise 6.28. From Lemma 6.22 we have an isomorphism of GL(V)-modules: S\VxzVv@,y (6.24) with V, = A-c,. Similarly for U, = A-a,, and since the image of right multi- plication by a, on V® is the tensor product of symmetric powers, we have Sym**V @ Sym®V @---@ Symn*V = V%@, U,. (6.25) But we have an isomorphism U, = @, K,V, of A-modules by Young's rule (4.39), so we deduce an isomorphism of GL(V)-modules Sym* V @ Sym*?V @ ++ @ Sym*V = | K,,S,V. (6.26) x By what we saw before the statement of the theorem, the trace of g on the left-hand side of (6.26) is the product H,(x,, ..., X) of the complete symmetric polynomials Hj,(x1,..., x4). Let S,(g) denote the endomorphism of S,V determined by an endomorphism g of V. We therefore have Hylx1, +15 %4) = EK yy Trace(S,(9)). But these are precisely the relations between the functions H, and the Schur polynomials 5, [see formula (A.9)], and these relations are invertible, since the matrix (K,,,) of coefficients is triangular with 1’s on the diagonal. It follows that Trace(S ,(g)) = 5,(x,, .... x4), which proves part (3). Note that if A= (4,,-.., 44) with d > k and Ay4, # 0, this same argument shows that the trace is S,(x,,..., X4,0,-.., 0), which is zero, for example by (A.6). For g the identity, this shows that S,V = 0 in this case. From part (3) we also get§6.2. The Proofs 87 dim S,V = S,(,..., 0, (6.27) and computing S,(I,..., 1) via Exercise A.30(ii) yields part (1). o Exercise 6.28. If you have given an independent proof of Problem 6.15, part (3) of Theorem 6.3 can be seen directly. The basis elements v; for S, V specified in Problem 6.15 are eigenvectors for a diagonal matrix with entries x1, ..., X4, with eigenvalue X* = x{-...-xg*, where the tableau T has a, 1's, a3 2's,..., a, k’s. The trace is therefore )’K,,X", where K,, is the number of ways to number the boxes of the Young diagram of A with a, 1's, a, 2's, ..., a, k's. This is just the expression for 5, obtained in Exercise A.31(a). We conclude this lecture with a few of the standard elaborations of these ideas, in exercise form; they are not needed in these lectures. Exercise 6.29*. Show that, in the context of Lemma 6.22, if U is a faithful A-module, then A is the commutator of its commutator B: A = {h: U > U: (bv) = by(v), Yo € U, b © BY. If U is not faithful, the canonical map from A to its bicommutator is surjective. Conclude that, in Theorem 6.3, the algebra of endomorphisms of V® that commute with GL(V) is spanned by the permutations in S,. Exercise 6.30. Show that, in Lemma 6.22, there is a natural one-to-one cor- respondence between the irreducible right A-modules U, that occur in U and the irreducible left B-modules V;. Show that there isa canonical decomposition U=QU@c u) asa left B-module and as a right A-module. This shows again that the number of times V, occurs in U is the dimension of U,, and dually that the number of times U, occurs is the dimension of V,. Deduce the canonical decomposition V% = DSV Och the sum over partitions A of d into at most k = dim V parts; this decomposition is compatible with the actions of GL(V) and ,. In particular, the number of times V, occurs in the representation V® of G, is the dimension of S,V. Exercise 6.31. Let e be an idempotent in the group algebra A = CG, and let U =eA be the corresponding right A-module. Let E = eAe, a subalgebra of A. The algebra structure in A makes eA a left E-module. Show that this defines an isomorphism of C-algebras E = ede = Hom, (eA, eA) = Hom, (U, U) = B.