Lecture Notes ni Mathematics Edited by .A Dold and .B Eckmann 4031 odnanreF .Q Gouv6a citemhtirA fo p-adic raludoM smroF galreV-regnirpS Heidelberg Berlin NewYork London Paris oykoT Author Fernando Quadros Gouv~a Instituto de Matem&tica e Estat[stica Universidade de S~.o Paulo Caixa Postal 20570 (Ag. Iguatemi) 01498 S~.o Paulo, SP, Brazil Mathematics Subject Classification (1980, revised 1985): 11 F ISBN 3-540-18946-7 Springer-Verlag Berlin Heidelberg New York ISBN 0-387-18946-? Springer-Verlag New York Berlin Heidelberg This work is subjectto copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, re-use of iflustrations, recitation, broadcasting, reproduction on microfilms or in other ways, and storage in data banks. Duplication of this publication or parts thereof is only permitted under the provisions otfh eG erman Copyright Law of September 9, 1965, ni its version of June 24, 1985, and a copyright fee must alwaybse paid. Violations fall under the prosecution act of the German Copyright Law. © Springer-Vedag Berlin Heidelberg 1988 Printed in Germany Printing and binding: Druckhaus Beltz, Hemsbach/Bergstr. 2146/3140-543210 Introduction The theory of p-adic modular forms was initiated by Serre, Katz, and Dwork, who, in the early 1970's, attempted to define objects which would be recognizably modular forms but which would be truly p-adic, reflecting the p-adic topology in an essential way. Specifically, one wanted a theory where two modular forms with highly congruent q-expansion coefficients would be p-adically close, and where limits of modular forms (with respect to this topology) would exist. The initial motivation for this construction was the problem of p-adic interpolation of special values of L-functions. The first difficulty, of course, was to find the correct definitions. Serre's approach (see [Se73]) was the most elementary: modular forms were identified with their q-expansions and p-adic modular forms were considered as limits of such q-expansions. Serre showed that such limits have "weights" (in a suitable sense), and developed the theory suffi- ciently to be able to obtain p-adic L-functions by constructing suitable analytic families of p-adic modular forms. His theory of analytic families of modular forms will be briefly discussed in our third chapter. He was also the first to notice that modular forms of level Np (and appropriate nebentypus) are of level N when considered as p-adic modular forms, which is a crucial aspect of the p-adic theory. Dwork's approach was more analytic. In [Dw73], for example, he restricts himself to p-adic analytic functions on a modular curve (i.e., p-adic modular forms of weight zero); in the same article, he notes the importance of growth conditions and of the fact that the U operator is completely continuous (in the sense of Serre) in the case which he is considering. Katz's work brought these approaches together, showing how to define p-adic rnodu- lar forms in modular terms, generalizing the results of Dwork on the U operator and of Serre on congruences ([Ka731, [Ka75a], and [Ka76]). As defined by Katz, p-adic modu- lar forms are clearly modular objects, i.e., they are clearly related to classifying elliptic curves with some additional structure. This allows a much more conceptuM approach to the theory, but at the same time makes it less accessible. Using his results, Katz was able to obtain important results on higher congruences between q-expansions of modular forms and several interpolation theorems for L-functions and Eisenstein series. It is this approach that we follow on foundational matters, and it will be fully described in our first chapter. During the preparation of this book, the author received the financial support of the Funda~.o de Amparo Pesquisa ~ do Estado de o~S Paulo, Brazil, of the ,PPAC BUCBFOC-S project, and of the Institut des Hautes Scientifiques l~tudes in Bures-sur-Yvette, Prance. iv More recently, attention has once again been called to the subject of p-adic modular forms by the work of Hida, who has used the theory, especially through his construc- tion of the "ordinary part" of the space of p-adic modular functions, in several different contexts (see [Hi86b], [Hi86a], etc.). Most notably, from our point of view, he has shown how to use ordinary p-adic modular forms to construct analytic families of Ga- lois representations. This has been considered in more detail by Mazur and Wiles in [MW86], and has motivated Mazur's construction of a general theory of deformations of Galois representations in [Ma], in relation to which several interesting results have been obtained by Mazur and Boston. The aim of this book is twofold. In the first place, we have tried to put together a unified and coherent exposition of the foundations of the theory, pointing out, in particular, the connections between the various approaches of Serre, Katz, and Hida. In the course of so doing, we have filled in various gaps and obtained several new results. In particular, we have obtained quite a bit of information about the action of the U operator, especially as relates to the spaces of overconvergent forms. In the second place, we have considered the problem of constructing deformations of residual representations, and shown that many of these are attached (as in the classical case) to p-adic modular forms which are eigenforms under the action of the Hecke operators, provided the original residual representation is absolutely irreducible and attached to a classical modular form over a finite field. The chapters break down as follows: the first chapter is largely foundational in nature, and most of its contents are to be found, implicitly or explicitly, in the works of Serre and Katz. It is intended to give the reader ready access to the theory. We have tried to give precise references for all the quoted theorems, and have frequently given informal descriptions of their proofs. The second chapter introduces the Hecke algebra, and, in particular, the U operator, and considers the problem of obtaining eigenforms for the Hecke algebra by way of studying the spectral theory of the U operator. These results are extensions of those obtained by Katz in [Ka73] (he considers only the case of overeonvergent modular forms of weight zero). The main payoff here is the fact that there are "very few" overconvergent eigenforms for the U operator outside the kernel of U, in the precise sense that the space of overconvergent eigenforms for U with fixed weight k C Z and fixed valuation for the eigenvalue of U is finite-dimensional. In fact, if one requires that the eigenvalue of U be a unit (the "ordinary" case considered by Hida), one knows that any such eigenform of weight k _> 3 will necessarily be classical. In the non-ordinary case (i.e., when the eigenvalue is assumed to belong to the maximal ideal of B), one does not know any examples that are not classical modular forms. By contrast, if we do not require overconvergence, one can produce a large number of (non-ordinary) eigenforms by a simple construction. The third chapter deals with constructing Galois representations; for this, we first construct a good duality theory for the Hecke algebra (related to the one obtained by ttida), and then use it in a crucial way to construct a "universal modular deformation" Y of a given (modular) absolutely irreducible residual representation, which defines a sub- space of Mazur's "universal deformation space". We then study this "universal modular deformation space" with a view to the natural question of determining its dimension, and, in particular, of deciding how close it comes to filling the entire deformation space. In this direction, we show that the Krull dimension of the modular deformation ring is at least three, which implies that the dimension of the modular deformation space, as a formal scheme, is at least two. One suspects that all of the deformations are in fact modular, so that the two deformation spaces are the same; we have not been able to show this, even in the well-understood "neat $3" case considered by Mazur and Boston, but we have several partial results that tend to support the conjecture. Readers who are already familiar with the theory of p-adic modular forms as for- mulated by Katz may wish to skip all of the first chapter and the first sections of the second, beginning to read at the point where we begin to study the U operator and its spectral theory. Readers who are only interested in Galois representations can begin directly with Chapter 3, and refer back as necessary. Most of the work contained herein was done while at Harvard University and at the Institut des Hautes ]~tudes Scientifiques; I would like to thank both institutions for their hospitality, and in particular for having been able to use their computer facilities so as to give this text the appearance it has. My debt to the mathematical work of Barry Mazur, Nicholas M. Katz, J.-P. Serre, and Haruzo Hida, among others, will, I hope, become abundantly clear in the text. On a more personal level, many people have helped me on the way to the completion of this book, which is a expanded and corrected version of my Harvard Ph.D. thesis. Barry Mazur, who was my thesis advisor~ suggested the topic, expressed interest at every stage, and supplied crucial help at many points. David Roberts, Dinesh Thakur, Jeremy Teitelbaum, Alan Fekete, and others of my fellow graduate students at Harvard were always ready to discuss this and other topics, and contributed greatly to the intellectual excitement of doing mathematics. J. F. Mestre of the ]~cole Normale Superieure helped me with the complexities of incompatible computer equipment at a critical point in the preparation of the final version. My sons Heitor and Marcos reminded me every night that there was more to life than mathematics, and cheered "Daddy's big project" on from day to day. My parents helped in many ways, both financially and personal]y, and my wife contributed in more ways than one can count. I thank you all; may God bless you all greatly. Bures-sur-Yvette, January of 1988 Fernando Quadros Gouv~a ...r~ rose,re, 7r&vrce ¢~e 85~a~, 0co9 .er~e~or~ Contents p-adic Modular Forms 1 1.1 Level Structures and Trivializations . . . . . . . . . . . . . . . . . . . . 3 1.2 p-adic Modular Forms with Growth Conditions .............. 4 1.2.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 1.2.2 Basic Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 1.3 Generalized p-adic Modular Functions . . . . . . . . . . . . . . . . . . . 13 1.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.3.2 The q-expansion map . . . . . . . . . . . . . . . . . . . . . . . . 16 1.3.3 Diamond operators . . . . . . . . . . . . . . . . . . . . . . . . . . 17 1.3.4 Weight and nebentypus ........ , . . . . . . . . . . . . . . . 19 1.3.5 Modular forms and modular functions ............... 19 1.3.6 Divided congruences . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.3.7 Appendix: Modular forms of weight X . . . . . . . . . . . . . . . 26 II ttecke and U Operators 30 II.1 Hecke Operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 II.l.1 Direct definition . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 II.1.2 Hecke operators on divided congruences .............. 33 II.2 The Frobenius Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 II.3 The U Operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 II.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 II.3.2 U and overconvergence . . . . . . . . . . . . . . . . . . . . . . . . 43 II.3.3 U and Frobenius . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 II.3.4 Spectral theory: the overconvergent case .............. 54 II.3.5 Spectral theory: the ordinary case . . . . . . . . . . . . . . . . . 59 II.3.6 The characteristic power series . . . . . . . . . . . . . . . . . . . 61 II.3.7 Varying the weight . . . . . . . . . . . . . . . . . . . . . . . . . . 66 II.4 Appendix: Hida's theory of the ordinary part . . . . . . . . . . . . . . . 69 III Galois Representations 71 III.1 Duality Theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 III.l.1 Classical duality . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 III.1.2 Duality for parabolic p-adic modular functions .......... 73 Vll" 1" " Contents III.1.3 The non-parabolic case . . . . . . . . . . . . . . . . . . . . . . . 77 III.2 Families cf Modular Forms . . . . . . . . . . . . . . . . . . . . . . . . . 80 IH.3 Changing the Level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 IH.4 Deformations cf Residual Eigenforms . . . . . . . . . . . . . . . . . . . . 87 III.4.1 Universal deformations . . . . . . . . . . . . . . . . . . . . . . . . 88 III.4.2 Deformations ovtside Np . . . . . . . . . . . . . . . . . . . . . . 89 III.4.3 Some classical results . . . . . . . . . . . . . . . . . . . . . . . . 92 HI.5 Deformations of Galois Representations .................. 93 III.6 The modular deformation space . . . . . . . . . . . . . . . . . . . . . . . 99 III.6.1 Changing the weight . . . . . . . . . . . . . . . . . . . . . . . . . 99 III.6.2 Twisting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 III.6.3 Families of twists, and an estimate for the Krull dimension of the modular deformation ring . . . . . . . . . . . . . . . . . . . . . . 106 III.6.4 The ordinary case . . . . . . . . . . . . . . . . . . . . . . . . . . 109 HI.7 Further Questions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 Chapter I p-adic Modular Forms We begin by giving an overview of the basic theory of p-adic modular forms, following, for the most part, the approach of Katz. First, we explain Katz's "p-adic modular forms with growth conditions" and how they relate to Serre's version of the theory. Then, we go on to discuss "generalized p-adic modular functions" (which, as the name suggests, include all the objects defined previously). The main references for this chapter are the foundational papers of Serre and Katz, especially [Se73], [Ka73], [Ka76], and [Ka75b]. To understand the definition of p-adic modular forms as functions of elliptic curves with extra structure (which is what we mean by a "modular" definition), we should first recall how to interpret classical modular forms in these terms. Classical (meromorphic) modular forms can be interpreted as functions of triples (E/A, w, z), composed of an elliptic curve E over A, a non-vanishing invariant differential w on E and a level structure ,, obeying certain transformation laws. Equivalently, they can be thought of as global sections of certain invertible sheaves over the moduli space of elliptic curves with the given kind of level structure, which is of course a modular curve. If we restrict A to be an algebra over the complex numbers, then it is easy to see that this is equivalent to the classical theory, since the quotient of the upper hMf-plane by a congruence subgroup classifies elliptic curves with a level structure. The fact that allowing A to run over R-algebras produces the classical theory of "modular forms defined over R" (i.e., complex modular forms whose Fourier expansion coefficients belong to R) is known as the "q-expansion principle". An exposition of the classical theory in this spirit can be found in the first chapter of [Ka73]; the first appendix to the same paper explains how this formulation relates to the classical definitions. We wish to obtain a p-adic theory of modular forms, i.e., a theory which reflects the p-adic topology in an essential way. This cannot be done by simply mimicking the definition of classical modular forms as functions of elliptic curves with level structure and differential, because the space of "classical modular forms over Zp" thus obtained is simply the tensor product with Zp of the space of classical modular forms over Z. To obtain a properly p-adic theory, we should take into account the p-adic topology by allowing limits of classical forms. This can either be done directly in terms of Fourier expansions of modular forms, or one can try to obtain a "modular" definition. 2 Chapter I. p-adic Modular Forms The first approach, which is due to Serre (in [Se73]), is to identify a classical modular form with the set of its q-expansions, and then to consider limits by using the p-adic topology on Zp[[q]]. One is then able to show that whenever a sequence of classical forms tends to a limit, their weights tend to a "p-adic weight" X, which is just a character X : XZ ~ Z~. This produces an elementary theory with strong ties to the theory of congruences between classical modular forms, which turns out to be a special case of the "modular" theory developed by Katz in [Ka73]. To obtain a "modular" theory of p-adic modular forms, one should define them as functions on elliptic curves, or, equivalently, as sections of bundles on a modular curve. This is achieved by Katz's idea of considering the rigid analytic space obtained by deleting p-adic disks around the supersingular points in the (compactified) moduli space M(N) of elliptic curves with a r~(N)-structure over Zp. To do this, we recall that, for p >_ 5, the classical modular form Ep-1 is a p-adic lifting of the Hasse invariant (this is equivalent to the well-known fact that the q-expansion of Ep-1 is congruent to 1 modulo p, see [Ka73, Section 2.11), and consider regions on A4(N) where Ep_l is "not too near zero". Since we would like to remove as small a disk as possible, we will allow the meaning of "not too near zero" to vary in terms of a parameter r, which we call the "growth condition". Taking r = 1 amounts to restricting ourselves to ordinary curves (i.e., to deleting the supersingular disks completely), and the resulting theory is the same as Serre's. However, if r is not a p-adic unit, we get a smaller space of "overconvergent" forms, which can be evaluated at curves which are "not too supersingular". Many of the interesting questions of the theory turn on the relation between these spaces as one varies r. The idea of considering modular forms with growth conditions at the supersingular points seems to be originally due to Dwork (for example, in [Dw73]); it was first developed systematically by Katz in [Ka73]. The idea of deleting the supersingular curves (more precisely, the curves with super- singular reduction) may sound strange at first, but is in fact quite natural in the context of what is wanted. We would like congruences of q-expansions to reflect congruences of the modular forms themselves, so that, for example, Ep-l(q) -- 1 (mod p) in Zp[[q]] should imply that there exists a modular form f such that Ep_, -1 =p f; this, however, is certainly false if we allow f to be evaluated at (a lifting of) a supersin- gular curve since the value of Ep_~ at any such curve must be divisible by p (because it lifts the Hasse invariant). It turns out that omitting the supersingular disks (and pos- sibly restricting the choice of differential on the curve) does the trick, and congruence properties of q-expansions are then reflected in congruence properties of p-adic modular forms. It also turns out that it is interesting to vary the radius of the omitted disk, introducing growth conditions and making the theory richer. 1.1. Level Structures and Trivializations 3 The choice of Ep-1 as a lifting of the Hasse invariant automatically restricts us to primes p > 5. For p = 2 and p = 3, one must choose a different lifting of the Hasse invariant (of higher level, since there is no lifting of level 1); one knows that such liftings exist for p = 3 and N > 2 and for p = 2 and 3 < N < 11. This means that it is possible to construct a theory on the same lines for p = 3 and any level N > 2 (and then obtain a level N = 1 theory by taking the fixed points under the usual group action); for p = 2, however, one will only get a theory for levels divisible by some number between 3 and 11, and again try to use group actions to get the full theory. In any case, since we will later need to restrict our theory to the case p __> 7 (for the spectral theory of the U operator), we have preferred to avoid these questions entirely by stating our results only for primes p >_ 5. In [Ka73], Katz discusses these problems further, and takes the cases p = 2 and p = 3 into account in the statements of his theorems. 1.1 Level Structures and Trivializations In what follows, p will denote a fixed rational prime, p > 5, and N a fixed level with (N,p) = 1. To guarantee that the moduli problems under consideration are repre- sentable, we will often assume that N > 3 (especially when discussing forms with growth conditions). We let B denote a "p-adic ring", i. e., a Zp-algebra which is complete and separated in the p-adic topology. In most cases, B will be a p-adically complete discrete valuation ring or a quotient of such. Let E be an elliptic curve over a p-adic ring B. We will consider, following Katz, level structures on E of the following kind: Definition 1.1.1 Let E be an elliptic curve over B, and let E[Np ]~ denote the kernel o] multiplication by Np~on E, considered as a group scheme over B. An arithmetic level Np~structure, or Fl(Np')a'ith-structure, on E is an inclusion z : ttNp~ ~ E[Np ]~ of finite flat group schemes over B. It is clear that, if v > ,O the existence of such an inclusion implies that E is fiber- by-fiber ordinary, so that we are automatically restricting our theory to such curves whenever the level is divisible by p. We will denote by AA°(Np ~) the moduli space of elliptic curves with an arithmetic level Np ~ structure. When v > ,O this is an open subscheme of the moduli space of elliptic curves with a Fl(Np~)-structure as defined by Katz and Mazur in [KMSh], which we denote by AAI(Np~). This tast space may be compactified by adding "cusps" (see [KM85], [DeRa]); this produces a proper scheme which we will denote by MI(Np~), which contains as an (affine when v > 0) open subscheme the scheme obtained by adding