Prof. Dr. Gu¨nter Harder Cohomology of Arithmetic Groups Zu erscheinen 2017 Erstellt mit LATEX2ε Preface Finallythisisnowthebookon”CohomologyofArithmeticGroups”whichwas announced in my ”Lectures on Algebraic Geometry I and II.” It starts with Chapter II because the material in what would be Chapter I is covered in the first four chapters of LAG I. In these four chapters we provide the basics of homological algebra which are needed in this volume. During the years 1980-2000 I gave various advanced courses on number the- ory, algebraic geometry and also on ”cohomology of arithmetic groups” at the university of Bonn. I prepared some notes for the lectures on ”cohomology of arithmetic groups”, because there was essentially no literature covering this subject. At some point I had the idea to use these notes as a basis for a book on this subject, a book that introduced into the subject but that also covered applications to number theory. It was clear that a self-contained exposition needs some preparation, we need homological algebra and later if we treat Shimura varieties, we need also a lot of algebraic geometry, especially the concept of moduli spaces. Since the cohomologygroupsofarithmeticgroupsaresheafcohomologygroups,andsince the theory of sheaves and sheaf cohomology is ubiquitous in algebraic geometry I had the idea to write a volume ”Lectures on Algebraic Geometry” where I discuss the impact of sheaf theory to algebraic geometry. This volume became the two volumes mentioned above and the writing of these volume is at last partly responsible for the delay. The applications to number theory concern the relationship between special values of L-functions and the integral structure of the cohomology as module under the Hecke algebra. On the one hand we can prove rationality statements for special values (Manin and Shimura) on the other these special values tell us something about the denominators of the Eisenstein classes. These connections wasalreadydiscussedintheoriginalnotesin1985forthespecialcaseofSl (Z). 2 and the precise results are stated at the end of Chapter II. In more general cases this relationship is conjectural and it was very impor- tantformethattheseconjecturesgotsomesupportbyexperimentalcalculations by G. van der Geer and C. Faber and others. Thistells usthat thewholesubjecthas interestingaspects from thecompu- tational side. In Chapter II we discuss a strategy to compute the cohomology and the Hecke endomorphisms explicitly so that we can verify the above con- jectures in explicit examples. For the group Sl (|Z) such explicit calculations 2 v vi have been done by my former student X.-D. Wang in his Bonn dissertation and are now resumed in Chapter II. I hope that this book will be a substantial contribution to a beautiful field in mathematics, it contains interesting results and it also points to challenging questions. Contents 0.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix 1 Basic Notions and Definitions 1 1.1 Affine algebraic groups over Q. . . . . . . . . . . . . . . . . . . . 1 1.1.1 Affine groups schemes . . . . . . . . . . . . . . . . . . . . 3 1.1.2 k-forms of algebraic groups . . . . . . . . . . . . . . . . . 5 1.1.3 The Lie-algebra . . . . . . . . . . . . . . . . . . . . . . . 8 1.1.4 StructureofsemisimplegroupsoverRandthesymmetric spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1.1.5 Special low dimensional cases . . . . . . . . . . . . . . . 15 1.2 Arithmetic groups . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.2.1 The locally symmetric spaces . . . . . . . . . . . . . . . . 19 1.2.2 Compactification of Γ\X . . . . . . . . . . . . . . . . . . 28 2 The Cohomology groups 47 2.1 Cohomology of arithmetic groups as cohomology of sheaves on Γ\X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 2.1.1 The relation between H•(Γ,M) and H•(Γ\X,M˜). . . . 48 2.1.2 How to compute the cohomology groups Hi(Γ\X,M˜)? . 51 2.1.3 The homology as singular homology . . . . . . . . . . . . 58 2.1.4 The fundamental exact sequence . . . . . . . . . . . . . . 60 2.1.5 How to compute the cohomology groups Hq(Γ\X,M˜). . 64 c 3 Hecke Operators 69 3.1 The construction of Hecke operators . . . . . . . . . . . . . . . . 69 3.1.1 Commuting relations . . . . . . . . . . . . . . . . . . . . . 72 3.1.2 Relations between Hecke operators . . . . . . . . . . . . . 75 3.2 Some results on semi-simple modules for algebras . . . . . . . . 78 3.3 Explicit formulas for the Hecke operators, a general strategy. . . 81 3.3.1 Hecke operators for Gl : . . . . . . . . . . . . . . . . . . 82 2 3.3.2 The special case Sl . . . . . . . . . . . . . . . . . . . . . 83 2 3.3.3 The boundary cohomology . . . . . . . . . . . . . . . . . 84 3.3.4 The explicit description of the cohomology . . . . . . . . 85 3.3.5 The map to the boundary cohomology . . . . . . . . . . . 86 3.3.6 The first interesting example . . . . . . . . . . . . . . . . 91 3.3.7 Computing mod p . . . . . . . . . . . . . . . . . . . . . 95 3.3.8 The denominator and the congruences . . . . . . . . . . . 96 3.3.9 The p-adic ζ-function . . . . . . . . . . . . . . . . . . . . 103 3.3.10 The Wieferich dilemma . . . . . . . . . . . . . . . . . . . 105 vii viii CONTENTS 4 Representation Theory, Eichler Shimura 107 4.1 Harish-Chandra modules with cohomology . . . . . . . . . . . . . 107 4.1.1 The principal series representations. . . . . . . . . . . . . 109 4.1.2 Reducibility and representations with non trivial coho- mology . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.1.3 The Eichler-Shimura Isomorphism . . . . . . . . . . . . . 123 4.1.4 The Periods . . . . . . . . . . . . . . . . . . . . . . . . . . 130 4.1.5 The Eisenstein cohomology class . . . . . . . . . . . . . . 132 5 Application to Number Theory 135 5.1 Modular symbols, L− values and denominators of Eisenstein classes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 5.1.1 Modular symbols attached to a torus in Gl . . . . . . . . 135 2 5.1.2 Evaluation of cuspidal classes on modular symbols . . . . 136 5.1.3 Evaluation of Eisenstein classes on capped modular sym- bols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 5.1.4 The capped modular symbol . . . . . . . . . . . . . . . . 141 6 Cohomology in the adelic language 155 6.1 The spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 6.1.1 The (generalized) symmetric spaces . . . . . . . . . . . . 155 6.1.2 The locally symmetric spaces . . . . . . . . . . . . . . . . 158 6.1.3 Thegroupofconnectedcomponents,thestructureofπ (SG ).161 0 Kf 6.1.4 The Borel-Serre compactification . . . . . . . . . . . . . . 162 6.1.5 The easiest but very important example . . . . . . . . . 164 6.2 The sheaves and their cohomology . . . . . . . . . . . . . . . . . 164 6.2.1 Basic data and simple properties . . . . . . . . . . . . . . 164 6.3 The action of the Hecke-algebra . . . . . . . . . . . . . . . . . . . 169 6.3.1 The action on rational cohomolgy . . . . . . . . . . . . . 169 6.3.2 The integral cohomology as a module under the Hecke algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171 6.3.3 Excursion: Finite dimensional H−modules and represen- tations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175 6.3.4 Back to cohomology . . . . . . . . . . . . . . . . . . . . . 185 6.3.5 Some questions and and some simple facts . . . . . . . . . 188 6.3.6 Poincar´e duality . . . . . . . . . . . . . . . . . . . . . . . 190 7 The fundamental question 195 7.0.7 The Langlands philosophy . . . . . . . . . . . . . . . . . 195 8 Analytic methods 213 8.1 The representation theoretic de-Rham complex . . . . . . . . . . 213 8.1.1 Rational representations . . . . . . . . . . . . . . . . . . 213 8.1.2 Harish-Chandra modules and (g,K )-cohomology. . . . 214 ∞ 8.1.3 Input from representation theory of real reductive groups. 217 8.1.4 Representation theoretic Hodge-theory. . . . . . . . . . . 218 8.1.5 Input from the theory of automorphic forms . . . . . . . . 220 8.1.6 Consequences. . . . . . . . . . . . . . . . . . . . . . . . 228 8.1.7 Franke’s Theorem . . . . . . . . . . . . . . . . . . . . . . 233 8.2 Modular symbols . . . . . . . . . . . . . . . . . . . . . . . . . . 233 0.1. INTRODUCTION ix 8.2.1 The general pattern . . . . . . . . . . . . . . . . . . . . . 233 8.2.2 Rationality and integrality results . . . . . . . . . . . . . 239 8.2.3 The special case Gl . . . . . . . . . . . . . . . . . . . . . 240 2 8.2.4 The L-functions . . . . . . . . . . . . . . . . . . . . . . . 256 8.2.5 The special values of L-functions . . . . . . . . . . . . . . 258 9 Eisenstein cohomology 261 9.1 The Borel-Serre compactification . . . . . . . . . . . . . . . . . . 261 9.1.1 The two spectral sequences . . . . . . . . . . . . . . . . . 263 9.1.2 Induction . . . . . . . . . . . . . . . . . . . . . . . . . . . 264 9.1.3 A review of Kostants theorem . . . . . . . . . . . . . . . 264 9.1.4 The inverse problem . . . . . . . . . . . . . . . . . . . . . 268 9.2 The goal of Eisenstein cohomology . . . . . . . . . . . . . . . . . 269 9.2.1 Inductionandthelocalintertwiningoperatoratfiniteplaces271 9.3 The Eisenstein intertwining operator . . . . . . . . . . . . . . . . 271 9.4 The special case Gl . . . . . . . . . . . . . . . . . . . . . . . . . 276 n 9.4.1 Resume and questions . . . . . . . . . . . . . . . . . . . . 281 9.5 Residual classes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 9.6 Detour: (g,K )− modules with cohomology for G=Gl . . . . 284 ∞ n 9.6.1 The tempered representation at infinity . . . . . . . . . . 285 9.6.2 The lowest K type in D . . . . . . . . . . . . . . . . . 292 ∞ λ 9.6.3 The unitary modules with cohomology, cohomological in- duction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294 9.6.4 The A (λ) as Langlands quotients . . . . . . . . . . . 299 qu,v 9.6.5 Congruences . . . . . . . . . . . . . . . . . . . . . . . . . 304 9.7 The example G=Sp /Z . . . . . . . . . . . . . . . . . . . . . . . 308 2 9.7.1 Some notations and structural data . . . . . . . . . . . . 308 9.7.2 The cuspidal cohomology of the Siegel-stratum . . . . . . 309 0.1 Introduction This book is meant to be an introduction into the cohomology of arithmetic groups. Thisiscertainlyasubjectofinterestinitsownright,butmymaingoal will be to illustrate the arithmetical applications of this theory. I will discuss the application to the theory of special values of L-functions and the theorem of Herbrand-Ribet (See Chap V, [Ri], Chap VI, Theorem II). Ourmainobjectsofinterestarethecohomologygroupsoflocallysymmetric spaces Γ\X with coefficients in sheaves M˜ which are obtained from a finitely generated Γ-module M, they will be denoted by H•(Γ\X,M˜). Ontheotherhandthesubjectisalsoofinterestfordifferentialgeometersand topologists, since the arithmetic groups provide so many interesting examples of Riemannian manifolds. My intention is to write an elementary introduction. The text should be readable by graduate students. This is not easy, since the subject requires a x CONTENTS considerable background: One has to know some homological algebra ( coho- mology and homology of groups, spectral sequences, cohomology of sheaves), the theory of Lie groups, the structure theory of semisimple algebraic groups, symmetric spaces, arithmetic groups, reduction theory for arithmetic groups. At some point the theory of automorphic forms enters the stage, we have to understand the theory of representations of semi-simple Lie groups and their cohomology. Finally when we apply all this to number theory (in Chap. V and VI) one has to know a certain amount of algebraic geometry ((cid:96)−adic cohomol- ogy, Shimura varieties (in the classical case of elliptic modular functions)) and some number theory( classfield theory, L−functions and their special values).I willtrytoexplainasmuchaspossibleofthegeneralbackground. Thisshouldbe possible, because already the simplest examples namely the Lie groups Sl (R) √ 2 and Sl (C) and their arithmetic subgroups Sl (Z) and Sl (Z[ −1]) are very 2 2 2 interesting and provide deep applications to number theory. For these special groups the results needed from the structure theory of semisimple groups, the theory of symmetric spaces and reduction theory are easy to explain. I will therefore always try to discuss a lot of things for our special examples and then to refer to the literature for the general case. I want to some words about the general framework. Arithmetic groups are subgroups of Lie groups. They are defined by arith- meticdata. TheclassicalexampleisthegroupSl (Z)sittingintherealLiegroup √ 2 Sl (R)orthegroupSl (Z[ −1])asasubgroupofSl (C),whichhastobeviewed 2 2 2 as real Lie group (See ..). Of course we may also consider Sl (Z) ⊂ Sl (R) as n n an arithmetic group. We get a slightly more sophisticated example, if we start from a quadratic form, say f(x ,x ,...,x )=−x2+x2+···+x2 1 2 n 1 2 n the orthogonal group O(f) is a linear algebraic group defined over the field Q ofrationalnumbers, thegroupofitsrealpointsisthegroupO(n,1)=O(f)(R) andthegroupofintegralmatricespreservingthisformisanarithmeticsubgroup Γ⊂O(f)(R) The starting point will be an arithmetic group Γ⊂G , where G is a real ∞ ∞ Lie group. This group is always the group of real points of an algebraic group over Q or a subgroup of finite index in it. To this group G one associates ∞ a symmetric space X = G /K , where K is a maximal compact subgroup ∞ ∞ ∞ of G , this space is diffeomorphic to Rd. The next datum we give ourselves ∞ is a Γ-module M from which we construct a sheaf M˜ on the quotient space Γ\X. This sheaf will be what topologists call a local coefficient system, if Γ acts without fixed points on X. We are interested in the cohomology groups H•(Γ\X,M˜). Under certain conditions we have an action of a big algebra of operators on these cohomology groups, this is the so called Hecke algebra H , it originates from the structure of the arithmetic group Γ (Γ has many subgroups of finite index, which allow the passage to coverings of Γ\X and we have maps going back and forth). It is the structure of the cohomology groups H•(Γ\X,M˜) as 0.1. INTRODUCTION xi a module under this algebra H, which we want to study, these modules contain relevant arithmetic information. Now I give an overview on the Chapters of the book. In chapter I we discuss some basic concepts from homological algebra, espe- cially we introduce to the homology and cohomology of groups, we recall some facts from the cohomology of sheaves and give a brief introduction into the theory of spectral sequences . Chapter II introduces to the theory of linear algebraic groups, to the theory of semi simple algebraic groups and the corresponding Lie groups of their real points. We give some examples and we say something about the associated symmetric spaces. We consider the action of arithmetic groups on these sym- metricspaces,anddiscusssomeclassicalexamplesindetail. Thisisthecontent of reduction theory. As a result of this we introduce the Borel-Serre compacti- fication Γ\X¯ of Γ\X, which will be discussed in detail for our examples. After this we take up the considerations of chapter I and define and discuss the co- homology groups of arithmetic groups with coefficients in some Γ-modules M. We shall see that these cohomology groups are related (and under some condi- tionsevenequal)tothecohomologygroupsofthesheavesM˜ onΓ\X. Another topic in this chapter is the discussion of the homology groups, their relation to the cohomology with compact supports and the Poincar´e duality. We will also explain the relations between the cohomology with compact supports the ordinary cohomology and the cohomology of the boundary of the Borel-Serre compactification. Finally we introduce the Hecke operators on the cohomology. Wediscusstheseoperatorsindetailforourspecialexamples,andweprovesome classically well known relations for them in our context. In these classical cases we also compute the cohomology of the boundary as a module over the Hecke algebra H At the end of this chapter we give some explicit procedures, which allow an explicit computation of these cohomology groups in some special cases. It may be of some interest to develop such computational techniques sinces this allows tocarryoutnumericalexperiments(See.. and... ). Weshallalsoindicatethat this apparently very explicit procedure for the computation of the cohomology does not give any insight into the structure of the cohomology as a module under the Hecke algebra. This chapter II is still very elementary. In Chapter III we develop the analytic tools for the computation of the co- homology. Here we have to assume that the Γ-module M is a C-vector space and is actually obtained from a rational representation of the underlying alge- braic group. In this case one may interprete the sheaf M˜ as the sheaf of locally constant sections in a flat bundle, and this implies that the cohomology is com- putable from the de-Rham-complex associated to this flat bundle. We could even go one step further and introduce a Laplace operator so that we get some kind of Hodge-theory and we can express the cohomology in terms of harmonic forms. Here we encounter serious difficulties since the quotient space Γ\X is not compact. But we will proceed in a different way. Instead of doing analysis on Γ\X we work on C (Γ\G ). This space is a module under the group G , ∞ ∞ ∞ xii CONTENTS which acts by right translations, but we rather consider it as a module under theLiealgebragofG onwhichalsothegroupK acts,itisa(g,K)-module. ∞ ∞ SinceourmoduleMcomesfromarationalrepresentationoftheunderlying group G, we may replace the de-Rham-complex by another complex H•(g,K ,C (Γ\G )⊗M, ∞ ∞ ∞ this complex computes the so called (g,K)-cohomology. The general principle will be to ”decompose” the (g,K)-module C into irreducible submodules and ∞ therefore to compute the cohomology as the sum of the contributions of the in- dividualsubmodules. Thisisagrouptheoreticversionoftheclassicalapproach by Hodge-theory. Here we have to overcome two difficulties. The first one is thatthequotientΓ\G isnotcompactandhencetheabovedecompositiondoes ∞ notmakesense,thesecondisthatwehavetounderstandtheirreducible(g,K)- modulesandtheircohomology. Thefirstproblemisofanalyticalnature,wewill give some indication how this can be solved by passing to certain subspaces of the cohomology the so called cuspidal and the discrete part of the cohomology. We shall state some general results, which are mainly due to A. Borel and H. Garland. We shall shall also state some general results concerning the second problem. The general result in this chapter is a partial generalization of the theorem of Eichler-Shimura, it describes the cuspidal part of the cohomology in terms of irreducible representations occurring in the space of cusp forms and containssomeinformationonthediscretecohomology,whichisslightlyweaker. We shall also give some indications how it can be proved. In the next chapter IV we resume the discussion of the previous chapter but we restrict our attention to the specific groups Sl (R) and Sl (C) and their 2 2 arithmetic subgroups. At first we give a rather detailed discussion of their representation-theory (i.e. the theory of representations of the corresponding (g,K)-modules) and we compute also the (g,K)-cohomology of the most im- portant (g,K)-modules, this is the second ingredient in the theorem of Eichler -Shimura. But in this special case we give also a complete solution for the an- alytical difficulties, so that in this case we get a very precise formulation of the Eichler-Shimura theorem, together with a rather complete proof. In the following chapter V we discuss the Eisenstein-cohomology. The the- orem of Eichler-Shimura describes only a certain part of the cohomology , the Eisenstein -cohomology is meant to fill the gap, it is complementary to the cuspidal cohomology. These Eisenstein classes are obtained by an infinite sum- mation process, which sometimes does not converge and is made convergent by analytic continuation. We shall discuss in detail the cases of the special groups Sl (R) and Sl (C) (the second case is not yet in the manuscript). Here we will 2 2 be able to explain an arithmetic application of our theory. Recall that we have to start from a rational representation of the underlying algebraic group G/Q and this representation is defined over Q or at least over some number field. Hence we actually get a Γ-module M which is a Q- vector space, and hence we may study the cohomology H•(Γ\X,M˜) which then is a Q-vector space. The Eisenstein classes are a priori defined by transcendental means, so they define a subspace in H•(Lieg,K,M˜)C . But we have still the action of the Hecke- algebra H, and this acts on the Q-vector space H•(Γ\X,M˜), and using the so called Manin-Drinfeld argument we can characterize the space of Eisenstein- classes as an isotypical piece in the cohomology, hence it is defined over Q. We 0.1. INTRODUCTION xiii shallindicatethatwecanevaluatethenowrationalEisenstein-classesoncertain homology-classes, which are also defined over Q, hence the result is a rational number. On the other hand we can-using the transcendental definition of the Eisenstein class-express the result of this evaluation in terms of special values ofL−functions. ThisyieldsrationalityresultsforspecialvaluesofL−functions (see [Ha] and [Ha -Sch]). This gives us the first arithmetic informations of our theory. In Chapter VI we discuss the arithmetic properties of the Eisenstein- classes. in the previous chapter we have seen, that the Eisenstein-classes are rational classes despite of the fact, that they are obtained by an infinite sum- mation. Now we will discuss the extremely special case where Γ = Sl (Z) and 2 our Γ-module is (cid:88) 1 M ={ a XνYn−ν|a ∈Z[ ]}. n ν ν 6 We also introduce the dual module 1 M∨ =Hom(M,Z[ ]). n 6 WethenaskwhethertheEisenstein-classisactuallyanintegralclass,thismeans whether it is contained in H1(Γ\X),M∨). The answer is no in general, the n Eisenstein-classhasadenominator,whichisapartfrompowersof2and3exactly the numerator of the number B ζ(1−(n+2))=± n+2. n+2 (See Chap. VI, Theorem I) This result is obtained by testing the Eisenstein- classes on certain homology classes, the so called modular symbols, which have beenintroducedinchapterII.ThisresultgeneralizesresultsofHaberland[Hab] andmystudent[Wg]. Iwillindicatethatthisresulthasarithmeticimplications in the direction of the theorem of Herbrand -Ribet. We cannot prove this theorem here since we need some other techniques from arithmetic algebraic geometrytocompletetheproof. Weshallalsodiscusssomecongruencerelations between Eisenstein classes of different weights, which arise from congruence relationsonthelevelofsheaves. Thesecongruencerelationsbetweenthesheaves have also been exploited by Hida and R. Taylor Finally I want to discuss some possible generalizations of all this and some open interesting problems. During the whole book I always tried to keep the door open for such generalizations. I presented the cohomology of arithmetic groups in such a way that we have the necessary tools to extend our results. Thismayhavehadtheeffect,thatthepresentationoftheresultsintheclassical case of Sl (Z) looks to complicated, but I hope it will pay later on. 2 Some of these generalisations are discussed in [HS].
Description: