Non- Archimedean L- Functions of Siegel and Hilbert Modular Forms: Associated with Siegel and Hilbert Modular Forms

Lecture Notes in Mathematics Editors: J.--M. Morel, Cachan F. Takens, Groningen B. Teissier, Paris

1471

3 Berlin Heidelberg New York Hong Kong London Milan Paris Tokyo

Michel Courtieu Alexei Panchishkin

Non-Archimedean L-Functions and Arithmetical Siegel Modular Forms Second, Augmented Edition

13

Authors Michel Courtieu Lyc´ee Anna de Noailles 2, Avenue Anna de Noailles 74500 Evian les Bains, France e-mail: [email protected]

Alexei A. Panchishkin Institut Fourier Universit´e Grenoble I BP 74 38402 Saint-Martin d’H`eres, France e-mail: [email protected]

Cataloging-in-Publication Data applied for Bibliographic information published by Die Deutsche Bibliothek Die Deutsche Bibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data is available in the Internet at http://dnb.ddb.de

Mathematics Subject Classification (2000): primary: 11F, 11R, 11S secondary: 19K, 46F, 46G ISSN 0075-8434 ISBN 3-540-40729-4 Springer-Verlag Berlin Heidelberg New York ISBN 3-540-54137-3 1st edition Springer-Verlag Berlin Heidelberg New York This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specif ically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microf ilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer-Verlag. Violations are liable for prosecution under the German Copyright Law. Springer-Verlag is a part of Springer Science+Business Media springeronline.com c Springer-Verlag Berlin Heidelberg 1991, 2004
Printed in Germany The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specif ic statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Typesetting: Camera-ready TEX output by the author SPIN: 10951675

41/3142/du - 543210 - Printed on acid-free paper

Preface

The present book is a updated version of the LNM 1471 ”Non-Archimedean L-Functions of Hilbert and Siegel Modular Forms” by Alexei Panchishkin, appeared in 1991. A part of this new book uses the results of the PhD Thesis of Michel Courtieu (Grenoble, Institut Fourier, 2000). The main subject of the book is the p-adic theory of L-functions of Siegel modular forms. In the case of the Riemann zeta function„and of the Dirichlet L-functions, this theory goes back to the classical Kummer congruences for Bernoulli numbers, and to their p-adic interpretation given by Kubota and Leopoldt, and by Mazur. Using the techniques of the p-adic integration, and of the Rankin-Selberg convolution method, we construct a p-adic analytic continuation of the standard L-functions of Siegel modular forms in a very general case, that is, for any non-zero Satake p-parameter. This second version has three basic new features, found recently by Alexei Panchishkin: 1. The use of arithmetical nearly holomorphic Siegel modular forms, viewed as certain formal expansions of many variables with algebraic coefficients. 2. The use of arithmetical differential operators acting on nearly holomorphic Siegel modular forms. 3. The method of canonical projection allowing to deduce congruences between special L-values from congruences between arithmetical nearly holomorphic Siegel modular forms, via a systematic use of Atkin’s Up -operator (the Frobenius operator Π + (p) of A.N.Andrianov). This new method gives a conceptual explanation for the formulas of Manin type in the Siegel modular case, obtained in LNM 1471. In fact, the previously used distributions admit a canonical lift to distributions with values in an appropriate subspace of arithmetical Siegel modular forms. This lift depends on a choice of a non-zero Satake parameter. The book is intended for researches, to postgraduate students, and to professors, interested in representation theory, functional analysis and arithmetic algebraic geometry. It contains, together with new results, much background

VI

Preface

information about p-adic measures, their Mellin transforms, Siegel modular forms, Hecke operators acting on them, Euler products etc. It seems that the general methods developed in the book may have a number of other applications, in particular, to families of Siegel modular forms and L-functions. Saint-Hilaire du Touvet, August 2003

Contents

1

Non-Archimedean analytic functions . . . . . . . . . . . . . . . . . . . . . . . 1.1 p-adic numbers and the Tate field . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Continuous and analytic functions . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Distributions, measures, and the abstract Kummer congruences 1.4 Iwasawa algebra and the non-Archimedean Mellin transform . . 1.5 Admissible measures and their Mellin transform . . . . . . . . . . . . . 1.6 Complex valued distributions, associated with Euler products .

2

Siegel modular forms and the holomorphic projection operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1 Siegel modular forms and Hecke operators . . . . . . . . . . . . . . . . . . 2.2 Theta series, Siegel-Eisenstein series and the Rankin zeta function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.3 Formulas for Fourier coefficients of the Siegel-Eisenstein series . 2.4 Holomorphic projection and Maass operator . . . . . . . . . . . . . . . . 2.5 Explicit description of differential operators . . . . . . . . . . . . . . . . .

13 13 17 21 28 35 38

45 46 62 70 83 91

3

Arithmetical differential operators . . . . . . . . . . . . . . . . . . . . . . . . . 95 3.1 Description of the Shimura differential operators . . . . . . . . . . . . 95 3.2 Nearly holomorphic Siegel modular forms . . . . . . . . . . . . . . . . . . . 99 3.3 Algebraic differential operators of Maass and Shimura . . . . . . . . 106 3.4 Arithmetical variables of nearly holomorphic forms . . . . . . . . . . 117

4

Admissible measures for standard L–functions . . . . . . . . . . . . . 127 4.1 Congruences between modular forms and p-adic integration . . . 127 4.2 Algebraic differential operators and Siegel-Eisenstein distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 4.3 A general result on admissible measures . . . . . . . . . . . . . . . . . . . . 144 4.4 The standard L-function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 4.5 Algebraic linear forms on modular forms . . . . . . . . . . . . . . . . . . . 165 4.6 Congruences and proof of the Main theorem . . . . . . . . . . . . . . . . 174

VIII

Contents

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187 Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

Introduction

Kummer Congruences and Bernoulli Numbers The starting point in the theory of zeta functions is the expansion of the Riemann zeta-function ζ(s) into the Euler product: ζ(s) =

∞ 
(1 − p−s )−1 = n−s p

(Re(s) > 1).

n=1

The set of arguments s for which ζ(s) is defined can be extended to all s ∈ C, s = 1, and we may view C as the group of all continuous quasicharacters × C = Homcont (R× + , C ),

y −→ y s

of R× + . The special values ζ(1 − k) at negative integers are rational numbers: ζ(1 − k) = −

Bk , k

where Bk are Bernoulli numbers, which are defined by the formal power series equality ∞  tet Bn tn Bt = t , e = n! e −1 n=0 and we know (by the Sylvester-Lipschitz theorem, see [Mi-St], [Kat3]) that c ∈ Z implies ck (ck − 1)

Bk ∈ Z. k

The theory of non-Archimedean zeta-functions originates in the work of Kubota and Leopoldt [Ku-Le] containing p-adic interpolation of these special values. Their construction turns out to be equivalent to classical Kummer congruences for the Bernoulli numbers, which we recall here in the following form (see [Kat3]):

M. Courtieu and A. Panchishkin: LNM 1471, pp. 1–12, 2004. c Springer-Verlag Berlin Heidelberg 2004


2

Introduction

Let p be a fixed prime number, c > 1 an integer prime to p. Put (c)

ζ(p) (−k) = (1 − pk )(1 − ck+1 )ζ(−k) n and let h(x) = i=0 αi xi ∈ Zp [x] be a polynomial over the ring Zp of p-adic integers such that x ∈ Zp =⇒ h(x) ∈ pm Zp . Then we have that

n 

(c)

αi ζ(p) (−i) ∈ pm Zp .

i=0 (c)

This property expresses the fact that the numbers ζ(p) (−k) depend continuously on k in the p-adic sense. This can be deduced from the known formula for the sum of k-th powers: Sk (N ) =

N −1  n=1

nk =

1 [Bk+1 (N ) − Bk+1 ] k+1

k   in which Bk (x) = (x + B)k = i=0 ki Bi xk−i denotes the Bernoulli polynomial. Indeed, all summands in Sk (N ) depend p-adic analytically on k, if we restrict ourselves to numbers n, prime to p, so that the desired congruence (c) follows if we express the numbers ζ(p) (−k) in terms of Bernoulli numbers (see [Kat3]). p–adic L–Functions and Mellin Transforms The domain of definition of p-adic zeta functions is the p-adic analytic Lie group × Xp = Homcont (Z× p , Cp )  of all continuous p-adic characters of the profinite group Z× p , where Cp = Qp denotes the Tate field (completion of an algebraic closure of the p-adic field Qp ), so that all integers k can be viewed as the characters xkp : y → y k . The construction of Kubota and Leopoldt is equivalent to the existence a p-adic analytic function ζp : Xp → Cp with a single pole at the point x = x−1 p , which becomes a bounded holomorphic function on Xp after multiplication by the elementary factor (xp x − 1), (x ∈ Xp ), and is uniquely determined by the condition ζp (xkp ) = (1 − pk )ζ(−k) (k ≥ 1). This result has a very natural interpretation in framework of the theory of non-Archimedean integration (due to B. Mazur): there exists a p-adic measure  (c) xkp µ(c) = ζ(p) (−k). Indeed, if we µ(c) on Z× p with values in Zp such that Z× p integrate h(x) over Z× p we exactly get the above congruence. On the other

Introduction

3

hand, in order to define a measure µ(c) satisfying the above condition  it suffices for any continuous function φ : Z× → Z to define its integral φ(x)µ(c) . p p Z× p For this purpose we approximate φ(x) by a polynomial (for which the integral is already defined), and then pass to the limit. The important feature of the construction is that it equally works for primitive Dirichlet characters χ modulo a power of p. Throughout the book we fix an embedding ip : Q → Cp

(0.1)

and we shall identify Q with a subfield of C and of Cp . Then a Dirichlet × character χ : (Z/ZpN )× → Q becomes an element of the torsion subgroup × Xptors ⊂ Xp = Homcont (Z× p , Cp )

and the above equality also holds for the special values L(−k, χ) of the Dirichlet L-series ∞
 L(s, χ) = χ(n)n−s = (1 − χ(p)p−s )−1 , n=1

p

so that we have  ζp (χxkp ) = ip (1 − χ(p)pk )L(−k, χ) (k ≥ 1, k ∈ Z, χ ∈ Xptors ). (0.2) The original construction of T. Kubota and H.W. Leopoldt [Ku-Le] was successesfully used by K. Iwasawa [Iw] for the description of the class groups of cyclotomic fields. Since then the class of functions admitting p-adic analogues has gradually extended. Zeta-functions (of complex variable) can be attached as certain Euler products to various objects such as diophantine equations, representations of Galois groups, modular forms etc..., and they play a crucial role in modern number theory. Deep interrelations between these objects discovered in last decades are based on identities for the corresponding zeta functions which presumably all fit into a general concept of Langlands of L-functions associated with automorphic representations of a reductive group G over a number field K. From this point of view the study of arithmetic properties of these zeta functions is becoming especially important. The theory of modular symbols (due to B. Mazur and Y.I. Manin, see [Man1], [Man3], [Maz-SD]) provided a non-Archimedean construction of functions, which correspond to the case of the group G = GL2 over K = Q. Several authors (including P. Deligne, K.A. Ribet, N.M.Katz, P.F. Kurˇcanov and others, see [De-Ri], [Kat1], [Kat2], [Kurˇc1], [Kurˇc2], [Sho], [Vi1], [Vi2]) investigated this problem for the case G = GL1 and GL2 over totally real fields and fields of CM-type (i.e. totally imaginary quadratic extentions of totally real fields). But the case of more general reductive groups remained unclear until the mid-eighties although important complex analytic properties of the

4

Introduction

Langlands L-functions had been proved. In last decades a general approach to construction of non-Archimedean L-functions associated with various classes of automorphic forms was developed, in particular, for the case of symplectic groups of even degree over K = Q and the group G = GL2 × GL2 over a totally real field K. The main tool of the appearing theory is the systematic use of the RankinSelberg method for obtaining both complex-valued and p-adic distributions as certain integrals involving cusp forms and Eisenstein series. By this method we constructed non-Archimedean analogues of the standard zeta functions attached to Siegel cusp forms of even degree and of sufficiently large weight. p–adic L–Functions of Siegel Modular Forms. For a Siegel modular form f (z) of degree m and weight k, which is an eigenfunction of the Hecke algebra, and for each prime number p one can define the Satake p-parameters of f denoted by αi (p) with i = 0, 1, · · · , m. In this introduction we assume for simplicity that f is a modular form with respect to the whole Siegel modular group Γ m = Spm (Z), but in Chapter 3 we treat the greneral case of forms of level C with a Dirichlet character ψmodC. The standard zeta function of f is defined by means of the Satake p-parameters as the following Euler product:

m   −1
 χ(p)
χ(p)αi (p) χ(p)αi (p)−1 , 1− s D(s, f, χ) = 1− 1− p ps ps p i=1 where χ is an arbitrary Dirichlet character. According to A.N.Andrianov and V.L.Kalinin [An-Ka], this function can be represented in a form of an integral convolution of f and a theta series with a Siegel-Eisenstein series as a kernel. The construction of its p-adic analytic continuation is based on explicit formulas for the special values of the standard zeta function, and this result is equivalent to the existence of some generalized Kummer congruences for these values. Before giving the precise statements of our results we introduce the following normalized zeta functions   m
Γ (s + k − j) D(s, f, χ) D
(s, f, χ) = (2π)−m(s+k−(m+1)/2) Γ ((s + δ)/2)  j=1

D− (s, f, χ) = Γ ((s + δ)/2)−1 D
(s, f, χ) iδ π 1/2−s D
(s, f, χ) D− (s, f, χ) = Γ ((1 − s + δ)/2) where δ = 0 or 1 according as χ(−1) = 1 or χ(−1) = −1, and let  k a(ξ)em (ξz) ∈ Sm f (z) = ξ>0

Introduction

5

be the Fourier expansion of the Siegel cusp form f (z) of weight k, the sum is extended over all positive definite half integral m × m-matrices, z ∈ Hm , Hm = {z ∈ GLm (C) | tz = z,

Im(z) is positive definite}

is the Siegel upper half plane of degree m and em (z) = exp(tr(2πiz)). Assume that k > 2m + 2 and that m is even. Theorem A (Algebraic properties of the special values of the standard zeta functions). (a) For all integers s with 1 ≤ s ≤ k − δ − m and Dirichlet characters χ such that χ2 is non-trivial for s = 1, we have that f, f −1 D+ (s, f, χ) ∈ K = Q(f, Λf , χ), where K = Q(f, Λf , χ) denotes the field generated by Fourier coefficients of f , by the eigenvalues Λf (X) of Hecke operators X on f , and by the values of the character χ. (b) For all integers s with 1 − k + δ + m ≤ s ≤ 0 we have that f, f −1 D− (s, f, χ) ∈ K. We may suppose without loss of generality that a(ξ) ∈ Q for all ξ > 0, and let us assume also that a(ξ0 ) = 1 for some ξ0 > 0 with det(2ξ0 ) = 1. Moreover, we make the essential assumption that a0 (ξ0 ) = 0 for the Fourier coefficient a0 (ξ0 ) of the p-auxiliary form  f0 (z) = a0 (ξ)em (ξz) ξ>0

defined in the first section of chapter 2 by the equality(2.57). On the contrary, it is possible that α0 (p) is not a p-adic unit (see Theorem B below). We call the case |α0 (p)|p < 1 supersingular (i.e. non-p-ordinary). Note that in the previous book ([Pa6], p 4), only the p-ordinary case was treated. According to its definition, this form satisfies the following multiplicativity property a0 (pN ξ) = α0 (p)N a0 (ξ). The importance of the above non-vanishing condition of a0 (ξ0 ) was pointed out to the second author by S. B¨ ocherer. Theorem B (non-Archimedean interpolation of the standard zeta functions). Under the assumptions as above for each integer c > 1 prime to p there exist two Cp -analytic functions Dc+ (x, f ), Dc− (x, f ) : Xp → Cp , with the following properties:

6

Introduction

(i) for all pairs (s, χ) such that χ ∈ Xptors is a non-trivial Dirichlet character, s ∈ Z with 1 ≤ s ≤ k − δ − m, s ≡ δ (mod 2), and for s = 1 the character χ2 is non-trivial, the following equality holds   m(s+k−1−m) + Cχs Gm (χ)Cχ ¯ c+ s 2 −2s D (s, f, χ) (1 − χ ¯ (c)c ) D (χxp , f ) = ip α0 (Cχ )2 G(χ) ¯ f, f

(ii) for all pairs (s, χ) such that χ ∈ Xptors is a non-trivial Dirichlet character and s ∈ Z with 1 − k + δ + m ≤ s ≤ 0, s ≡ δ (mod 2), the following equality holds   m(s+k−1−m) − ¯ Gm (χ)Cχ c− s 2 2s−2 D (s, f, χ) (1 − χ (c)c ) D (χxp , f ) = ip α0 (Cχ )2 f, f

where Gm (χ) =



χ(det(h))em (h/Cχ )

h∈Mm (Z)modCχ

denotes the Gauss sum of degree m of the primitive Dirichlet character χ mod Cχ , Cχ = pNχ , α0 (Cχ ) = α0 (p)Nχ , G(χ) = G1 (χ). (iii) if ordp α0 (p) = 0 then the holomorphic functions in (i), (ii) are bounded Cp -analytic functions; (iv) in the general case (but assuming that α0 (p) = 0) the holomorphic functions in (i), (ii) belong to the type o(log(xhp )) with h = [2ordp α0 (p)] + 1 and they can be represented as the Mellin transform of h-admissible measures; (v) if h ≤ k − m then the functions Dc± are uniquely determined by the above conditions (i) and (ii). (see Main Theorem in §4.1.5 and Theorem 4.23). Distributions with Values in Siegel Modular Forms. One of the purposes of this book is to give a new conceptual construction of admissible measures (in the sense of Amice-V´elu) attached to a standard L-function of a Siegel cusp eigenform. For this purpose we use the theory of p-adic integration with values in spaces of arithmetical nearly holomorphic Siegel modular forms (in the sense of Shimura, see [Sh9]) defined over an Oalgebra A where O is the ring of integers in a finite extension K of Qp . Often we simply assume that A = Cp . Arithmetical nearly holomorphic Siegel modular forms admit two different descriptions:  a(ξ, Ri,j )q ξ ∈ Q[[q Bm ]][Ri,j ] such that • as formal power series g = ξ∈Bm

z − z))−1 and z ∈ Hm in for all R = (Ri,j )i,j=1,··· ,m = (4πIm(z))−1 = (2πi(¯ the Siegel upper half plane of degree m the series converges to a C∞ -Siegel modular form of a given weight k and character ψ;

Introduction

7

• as certain C∞ -Siegel modular forms g taking values in Q at all CM-points (up to a factor independnt of a concrete form). In fact, we need to fix throughout the book two embeddings i∞ : Q → C,

i p : Q → Cp ,

and we shall often view the field Q as a subfield of both C and Cp (the complex and the p–adic numbers) via these embeddings, omitting the symbols i∞ and  is the Tate field (the completion of a fixed algebraic closure i . Here C = Q p

p

p

Qp of Qp ), which is endowed with a unique norm | · |p such that | p |p = p−1 . We describe the action of certain arithmetical differential operators on algebraically defined nearly holomorphic Siegel modular forms (in the sense of Shimura, see [Sh9] over an O-algebra A where O is the ring of integers in a finite extension K of Qp . Often we simply assume that A = Cp . The important property of these arithmetical differential operators is their commutation with the Hecke operators (under an appropriate normalization, see Theorem 3.13). The reason of these nice arithmetical properties is that these (non-holomorphic) operators could be algebraically interpreted in terms of the Gauss-Manin connection acting on the de Rham cohomology sheaves of the arithmetical compactification of Siegel modular varieties (see [Fa-Ch90], [Ha2], [Ga-Ha]). Denote by M = M(A) the A-module (or simply the Cp -vector space) of nearly holomorphic Siegel modular forms . We show in Chapter 4 that the action of these arithmetical differential operators produces natural families of distributions on a profinite group Y = lim Yi with values in these A-modules ←− of nearly holomorphic Siegel modular forms inside a formal q-expansion ring like A[[q B ]][Ri,j ] where B = Bm = {ξ = t ξ ∈ Mm (Q)|ξ ≥ 0, ξ half-integral} is the semi-group, important for the theory of Siegel modular forms), and the nearly holomorphic parameters (Ri,j ) = R correspond to the matrix R = (4πIm(z))−1 in the Siegel modular case. We apply to these distributions the canonical projector πα : M → Mα onto the primary (characteristic) A-submodule associated to a non-zero eigenvalue α ∈ A× of the Frobenius operator U = Π + (p) = Up (the Atkin-Lehner operator in the Siegel modular case; Mα is the maximal A-submodule on which U − αI is nilpotent). This operator acts similar to the trace operator lowering the level of modular forms. On the other hand, U is invertible on Mα if α ∈ A× so that one can glue its action on forms of various levels. In this way one obtains the desired distributions with values in a finite dimensional vector space starting from naturally defined distributions with values in spaces of modular forms (like Siegel-Eisenstein distributions, theta distributions etc.). In order to obtain from them numerically valued distributions interpolating critical values attached to standard L-functions of Siegel modular forms one applies a suitable linear form coming from the Petersson scalar product (using the Andrianov identity of Rankin type for the standard L-function).

8

Introduction

In previous works [Pa7], [Pa6] a non–Archimedean interpolation of these special values was constructed fixing a non zero Satake p–parameter α0 (p) of the cusp eigenform f . It follows that the normalized critical values D(s, f, χ) of the standard L-function can be explicitely rewritten in terms of certain Cp –valued integrals of admissible p–adic measures (over a profinite group of the type Z× p of p–adic units) provided the character χ is non–trivial. More explicit description of these special values was given by M.Courtieu in his PhD thesis ([Cour], Institut Fourier, 2000) using detailed study of the action of the Shimura differential operator on Siegel modular forms. In the present book we give a conceptual explanation of the p–adic properties satisfyied by the special values of the standard L-function D(s, f, χ), where f is a Siegel cusp form of an even degree m and of weight k > 2m + 2, χ is a varying Dirichlet character. We show that these admissible measures can be lifted to arithmetical nearly holomorphic Siegel modular forms studied by G.Shimura [Sh9]. This lifting is given by a universal sequence Φ± s (χ) of distributions with values in arithmetical nearly holomorphic Siegel modular forms (for critical pairs (s, χ), see Proposition 4.21). It would be interesting to extend these lifting results to Siegel cusp eigenforms of odd degree, using the method of B¨ ocherer-Schmidt [B¨ oSch]. Families of Siegel Modular Forms. Note that the problem of construction of families of modular forms is closely related to the context of Wiles’ proof [Wi] which is based on a Galois cohomological construction of p-adic families of classical elliptc modular forms. It seems that a natural thing would be to try to extend constructions of such families to other classes of modular forms; the present book gives an example: a canonical lift of previously known p–adic distributions to distributions with values in an appropriate subspace of arithmetical Siegel modular forms. This lift depends on a choice of a non-zero Satake parameter, and it produces families by integration of arithmetical characters. A construction of rigid– analytic families of modular distributions is given recently in [PaInv]. Applications of this theory to constructions of certain new p-adic families of modular forms (families of Siegel-Eisenstein series, families of theta-series with spherical polynomials. . .) is based on the following main sources: • • • •

Serre’s theory of p-adic modular forms as certain formal q-expansions, [Se2]. Shimura’s theory of arithmeticity for nearly holomorphic forms, [Sh9]. Hida’s theory of p-adic modular forms and p-adic Hecke algebras, [Hi5]. Constructions of p-adic Siegel-Eisenstein series, and of p-adic Klingen– Eisenstein series by the author, [Pa8].

Note that the eigenspaces M(α) of U are contained in the primary subspaces Mα , and they where used by D. Kazhdan, B. Mazur, C.-G. Schmidt,

Introduction

9

see [KMS2000], in the p-ordinary case via a p–adic limit procedure. Notice that we do not need a p–adic limit procedure, and we treat the general case of any positive slope. Remarks on modular forms of positive slope According to R.Coleman, F. Gouvˆea and B. Mazur, the structure of modular forms of a given positive slope is more complicated than in the ordinary case, even for elliptic modular forms (see the theory of ”ferns” in [Gou-Ma], and [Co-Ma]). Our results mean that the p-adic behaviour of the standard p-adic L-functions depends not only on the slope of a Siegel modular form but also the eigenvalue itself, see also [PaInv]. This would be important for constructions of rigid–analytic families of Coleman type [CoPB] in the Siegel modular case, and the corresponding families of p-adic L-functions. Notice that the structure of weights is more complicated in the Siegel modular case due to vector-valued modular forms. It seems that besides fixing just a positive slope one needs some additional and more subtle parameters, given probably by an analogue of ”ferns”, which could provide a good understanding of overconvergency in the Siegel modular case. In the future, it would be interesting to combine our approach with geometric methods of Faltings–Chai [Fa-Ch90] and of Coleman–Mazur [Co-Ma]. For a fixed positive integer N ∈ N consider the profinite group Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

There is a natural projection yp : Y → Z× p. Definition 0.1. a) For h ∈ N, h ≥ 1 let C h (Y, A) denote the A-module of × locally polynomial functions of degree < h of the variable yp : Y → Z× p → A ; in particular, C 1 (Y, A) ⊂ C loc−const (Y, A) (the A-submodule of locally constant functions). We adopt the notation Φ(U) := Φ(χU ) for the characteristic function χU of an open subset U ⊂ Y . Let also denote C loc−an (Y, A) the A-module of locally analytic functions and C(Y, A) the A-module of continuous functions so that C 1 (Y, A) ⊂ C h (Y, A) ⊂ C loc−an (Y, A) ⊂ C(Y, A). b) For a given positive integer h an h-admissible measure on Y with values in M is an A-module homomorphism Φ˜ : C h (Y, A) → M such that for all a ∈ Y and for v → ∞

10

Introduction

      (yp − ap )j dΦ˜    a+(N pv )

= o(p−v(j−h) ) for all

j = 0, 1, · · · , h − 1,

p,M

where ap = yp (a). We adopt the notation (a)v = a + (N pv ) for both an element of Yv and the corresponding open compact subset of Y . U (p)–Operator and Method of Canonical projection. We explain in Section 4.3 of Chapter 4, how to construct an h-admissible α : C h (Y, A) → M(A) out of a sequence of distributions measure Φ Φj : C 1 (Y, A) → M with values in an A-module M = M(A) of nearly holomorphic modular forms over A (for all j ∈ N with j ≤ h − 1, where A is an O-algebra, and α ∈ A× is a fixed non-zero eigenvalue of the Frobenius operator U = U (p) = Π + (p) acting on  g= a(ξ, n)Rn q ξ ∈ A[[q Bm ]][Ri,j ], ξ∈Bm ,n≥0

by



g|U (p) =

a(pξ, n)(pR)n q ξ

ξ∈Bm ,n≥0

(over the complex numbers R = (4π)−1 Im(z)−1 and this notation corresponds to   1 u f |k m f |U (p)(z) = = 0m p1m t u=u∈Mm (Z) mod p



p−κm

tu=u∈M

m (Z)

f ((z + u)/p), mod p

Im ((z + u)/p) = Im(z/p) = Im(z)/p = (4π)−1 (pR)−1 (we use the Petersson-Andrianov notation for the action of matrices, Chapter 2). Then we consider (for an α ∈ A× ) the canonical projection operator πα : M(A) → M(A)α . We define an A-linear map α : C h (Y, A) → M Φ on local monomials ypj by 

α = πα (Φj ((a)v )) ypj dΦ

(a)v

where Φj : C 1 (Y, A) → M(A) are certain M(A)-valued distributions on Y for all j = 0, 1, . . . , h − 1.

Introduction

11

Automorphic L–Functions The standard zeta function D(s, f, χ) provides an example for the general definition of the Langlands L-functions. For a reductive group G over a number field K this definition is based on the notion of the Langlands L-group L G; this group is a complex analytic reductive group such that the lattice of characters of its maximal torus and the lattice of its one-parameter subgroups (cocharacters) are obtained from the analogous objects of the group G by inversion. The important fact in the representation theory of reductive groups over local fields is that semisimple conjugacy classes hv of L G for a place v of K correspond to certain infinite dimensional representations πv of the group G(Kv ) over the local field Kv (the completion of K at v). It is known that for groups of the type An and Dn this construction preserves their types, and interchanges the types Bn and Cn , so that if G = GLn then L G = GLn (C), and if G = GSpm then L G = Spin2m+1 (C), the universal covering of the orthogonal group SO2m+1 (C). For example, if G = GL 2 and v a non-Archimedean place α 0 1 then L G = GL2 (C) and for hv  the corresponding representation 0 α2 πv is the representation IndG T (µ1 ⊗ µ2 ), which is induced from the maximal torus T = GL1 × GL1 , µ1 , µ2 : Kv× → C× being unramified quasicharacters of Kv× with µi (πv ) = αi , i = 1, 2. Let π be an automorphic representation of the group G (which is an irreducible subrepresentation of the smooth regular representation of the adelic group G(AK )). Then there is the decomposition of π into the infinite tensor product: π = ⊗v πv where πv is a representation of G(Kv ) which correspond to certain classes hv from L G for almost all v (i.e for v ∈ S where S is a finite set of places of K). For a finite dimensional representation r : L G → GLt (C) of the L-group we define automorphic L-functions
L(s, π, r) = LS (s, π, r) = det(1t − (N v)−s r(hv ))−1 v∈S

where N v is the number of elements of the residue class field of v (which is a power of its characteristic), and the product is taken over all non-Archimedean places v, v ∈ S. In the Siegel modular case we consider, associated with f , the automorphic representation πf , which is generated by a function on GSpm (A) inflated from the cusp form f on Hm (as a subrepresentation of the regular representation of G(AQ ) = GSpm (A)). The irreducibility of πf is equivalent to the fact that f is an eigenfunction for the Hecke algebra Hm = ⊗p Hpm of the Siegel modular group Γm of degree m. In this case the corresponding character of Hp on f is completely determined by its Satake p-parameters, and for the universal covering r : Spin2m+1 (C) → SO2m+1 (C) with Spin2m+1 (C) ⊂ GL2m (C) we have that the classes hv and r(hv ) are represented by the matrices

12

Introduction

hv = Sp(hv )  diag{α0 (p)αi1 (p) · · · αir (p) | 0 ≤ r ≤ m, 1 ≤ i1 < · · · < ir ≤ m} r(hv ) = St(hv )  diag{1, α1 (p), · · · , αm (p), α1 (p)−1 , · · · , αm (p)−1 } where Sp and St are called, respectively, spinor and standard representations of the Langlands group L G. Therefore the standard zeta function D(s, f, χ) coincides with the L-function L(s, πf , St). The function 
 L(s, πf , Sp) =  p




−1  (1 − χ(p)α0 (p)αi1 (p) · · · αir (p)p−s )

0≤r≤m

1≤i1 0 there exists a polynomial h(x) ∈ Cp [x] such that |f (x) − h(x)|p < ε for all x ∈ W . If f (W ) ⊂ L for a closed subfield L of Cp then h(x) can be chosen so that h(x) ∈ L[x] (see [Kob1], [Wa]). Interesting examples of continuous p-adic functions are provided by interpolation of functions, defined on certain subsets, such as W = Z or N with K = Qp . Let f be any function on non-negative integers with values in Qp or in some (complete) Qp -Banach space. In order to extend f (x) to all x ∈ Zp we can use the interpolation polynomials  x x(x − 1) · · · (x − n + 1) . = n n!   Then we have that nx is a polynomial of degree n of x, which for x ∈ Z, x ≥ 0 gives the binomial coefficient. If x ∈ Zp then  x is close (in the p-adic topology) x to a positive integer, hence the value of n is also close to an integer, therefore  x . ∈ Z p n The classical Mahler’s interpolation theorem says that any continuous function f : Zp → Qp can be written in the form (see [La2], [Wa]):  x f (x) = an , n n=0 ∞ 

(1.15)

18

1 Non-Archimedean analytic functions

with an → 0 (p-adically) for n → ∞. For a function f (x) defined for x ∈ Z, x ≥ 0 one can write formally f (x) =

∞ 

an

n=0

 x , n

where the coefficients can be founded from the system of linear equations f (n) =

n 

am

m=0

that is am =

m  j=0

(−1)m−j

 n , m

 m f (j). j

The series for f (x) is always reduced to a finite sum for each x ∈ Z, x ≥ 0. If an → 0 then this series is convergent for all x ∈ Zp . As was noticed above, the inverse statement is also valid (“Mahler’s criterion”). If convergence of an to zero is so fast that the series defining the coefficients of the x-expansion of f (x) also converge, then f (x) can be extended to an analytic function, see 1.2.2 below. Unfortunately, for an arbitrary sequence an with an → 0 the attempt to use (1.15) for continuation of f (x) out of the subset Zp in Cp may fail. However, in the sequel we mostly consider anlytic functions, that are defined as sums of power series. 1.2.2 Analytic functions and power series of convergence of a series ∞(see [Kob1], p 13). The well known criterion  a is that the following partial sums n N ≤n≤M an are small for large n=0 N , M with M > N . In view of the non-Archimedean property (1.2) in Cp this occurs if and only if |an |p → 0 or  ordp (an ) → ∞ for n → ∞. Therefore the convergence of the power series n≥0 an xn depends only on |x|p but not on the precise value of x, hencethere is no “conditional convergence” in this case. Thus, for any power series n≥0 an xn we can define its radius of convergence r such that only one of the following holds: ∞  n=0 ∞ 

an xn converges ⇐⇒

x ∈ D0 (r− ),

(1.16)

an xn converges ⇐⇒

x ∈ D0 (r).

(1.17)

n=0

 An example of the first alternative is n≥0 xn , where (1.16) is satisfied with  n r = 1, and an example for the second is n≥0 pn xp −1 , where (1.17) is satisfied also with r = 1.

1.2 Continuous and analytic functions

19

The important examples of analytic functions are exp(x) and log(x), which are given as the power series exp(x) =

∞  xn , n! n=0

log(1 + x) =

∞  (−1)n+1 n x , n n=1

(1.18)

r = p−1/(p−1) ,

(1.19)

and we have that exp(x) converges on D0 (p−1/(p−1)− ), and

log(1 + x) converges on D0 (1− ),

r = 1,

(1.20)

ˇ (see [Kob1], [Bo-Sa]), so that exp(x) converges in a disc smaller than the unit disc, and log(1 + x) has better convergence that exp(x). Since the identity log(xy) = log(x) + log(y) (1.21) holds as a formal power series identity, it follows that (1.21) holds in Cp as long as |x − 1|p < 1 and |y − 1|p < 1. In particular, since |ζ − 1|p < 1 for ζ any pn -root of unity, we can obviously apply (1.21) to conclude that log(ζ) = 0. Also we have that for all x ∈ D0 (p−1/(p−1)− ) the following identities hold exp(log(1 + x)) = 1 + x, and log(exp(x)) = x, which are deduced from the corresponding properties of the formal series and can be used for establishing isomorphisms between certain additive and ν− multiplicative subgroups in Cp and C× ; Qp ) (with ν as in p ; for U = D1 (p 1.1.3) there are the isomorphisms ∼

n

exp : pν+n Zp −→ U p

(with n ≥ 0).

(1.22)

Theorem 1.1 (On analyticity of interpolation series). Let r < p−1/(p−1) < 1, and let  ∞  x an f (x) = , n n=0 be a series with the condition |an |p ≤ M rn for some M > 0. Then f (x) is expresible as a certain power series whose radius of convergence is not less than R = (rp1/(p−1) )−1 > 1. Proof. (see [Wa], p. 53) As an example let us consider the function a x which is defined for a ∈ Z× p by means of the decomposition a = ω(a) a where ω(a) is the Teichm¨ uller representative of a for p > 2 and ω(a) = ±1 for p = 2 with ω(a) ≡ a (mod 4), and the exponentiation is given by the binomial formula

20

1 Non-Archimedean analytic functions ∞   x a = (1 + a − 1) = ( a − 1)n . n n=0 x

x

(1.23)

Since | a − 1|p ≤ pν we may put in the above theorem r = p−ν and get that the function a x is a power series in x with the radius of convergence not less than pν−1/(p−1) > 1 and the following equality holds: ∞   x (1.24) exp(x log( a )) = ( a − 1)n . n n=0 in which both parts are analytic in x and coincide for x ∈ N. 1.2.3 Newton polygons (see [Kob1], p 21). The Newton polygon Mf for a power series f (x) =

∞ 

an xn ∈ Cp [[x]]

n=1

is defined as the convex hull of the points (n, ordp (an )) (where we agree to take ordp (0) = ∞). It is not hard to prove the following. Proposition 1.2. If a segment of Mf has slope λ and horizontal length N (i.e. it extends from (n, ordp (an )) to (n + N, λN + ordp (an )) then f has precisely N roots rn with ordp (rn ) = −λ (counting multiplicity). The following theorem is the p-adic analog of the Weierstrass Preparation theorem (see [Kob1], p. 21). Theorem 1.3. Let f (x) = am xm + am+1 xm+1 + · · · ∈ Cp [[x]], am = 0 be a power series which converges on D0 (pλ ; Cp ). Let (N, ordp (aN )) be the right endpoint of the last segment of Mf with slope ≤ λ, if this N is finite. Otherwise, there will be a last infinitely long segment of slope λ and only finitely many points (n, ordp (an )) on that segment. In that case let N be the last such n. Then there exists a unique polynomial h(x) of the form bm xm + bm+1 xm+1 + · · · + bN xN with bm = am and a unique power series g(x) which converges and does not vanish on D0 (pλ ; Cp ) such that f (x) =

h(x) on D0 (pλ ; Cp ). g(x)

In addition Mh coincide with Mf as far as the point (N, ordp (aN )).

1.3 Distributions, measures, and the abstract Kummer congruences

21

Corollary 1.4. A power series which converges everywhere and has no zeroes is a constant. A simple proof of the Weierstrass Preparation Theorem for power series of the type ∞  f (x) = an xn ∈ Op [[x]] n=1

is based on a generalization of the Euclid algorithm (see [Man1]). There exists another definition (dual) of the Newton polygon (see [Kob1], [Vi1]) of a series ∞  an xn ∈ Cp [[x]]. f (x) = n=1

Instead of the points (n, ordp (an )) let us look at the lines ln : y = nx+ordp (an ) f is defined as the graph of the function minn ln (x). The with slope n. Then M f give x-coordinate of the points of intersection of the ln which appear in M ordp of the zeroes, and the difference between the slopes n of the successive ln f give the number of zeroes with given ordp . This definition which appear in M f coincides with the graph of the function is explained by the fact that M    Mf (t) = log sup |f (x)|p p

|x|p 1 and admits an analytic continuation over all s ∈ C. For this series we have that

24

1 Non-Archimedean analytic functions

Bk,f . (1.29) k For example, if f ≡ 1 is the constant function with the period M = 1 then we have that ∞  t Bk k Bk t = t , ζ(1 − k) = − , k k! e −1 L(1 − k, f ) = −

k=0

Bk being the Bernoulli number. The formula (1.29) is established by means of the contour integral discovered by Riemann (see [La1]). This formula apparently implies the desired independence of Bk,f on the choice of M . We note also that if K ⊂ C is an arbitrary subfield, and f (Y ) ⊂ K then we have from the formula (1.27) that Bk,f ∈ K hence the distribution Ek is a K-valued distribution on Y . 1.3.2 Measures Let R be a topological ring, and C(Y, R) be the topological module of all R-valued functions on a profinite set Y . Definition 1.6. A measure on Y with values in the topological R-module A is a continuous homomorphism of R-modules µ : C(Y, R) −→ A. The restriction of µ to the R-submodule Step(Y, R) ⊂ C(Y, R) defines a distribution which we denote by the same letter µ, and the measure µ is uniquely determined by the corresponding distribution since the R-submodule Step(Y, R) is dense in C(Y, R). The last statement expresses the well known fact about the uniform continuity of a continuous function over a compact topological space. Now we consider any closed subring R of the Tate field Cp , R ⊂ Cp , and let A be a complete R-module with topology given by a norm | · |A on A compatible with the norm | · |p on Cp so that the following conditions are satisfie: • for x ∈ A the equality |x|A = 0 is equivalent to x = 0, • for a ∈ R, x ∈ A: |ax|A = |a|p |x|A , • for all x, y ∈ A: |x + y|A < max(|x|A , |y|A ). Then the fact that a distribution (a system of functions µ(i) : Yi → A) gives rise to a A-valued measure on Y is equivalent to the condition that the system µ(i) is bounded, i.e. for some constant B > 0 and for all i ∈ I, x ∈ Yi the following uniform estimate holds: |µ(i) (x)|A < B.

(1.30)

This criterion is an easy consequence of the non-Archimedean property |x + y|A ≤ max(|x|A , |y|A )

1.3 Distributions, measures, and the abstract Kummer congruences

25

of the norm | · |A (see [Man2], [Vi1]). In particular if A = R = Op = {x ∈ Cp | |x|p ≤ 1} is the subring of integers in the Tate field Cp then the set of Op -valued distributions on Y coincides with Op -valued measures (in fact, both sets are R-algebras with multiplication defined by convolution, see section 1.4). Below we give some examples of measures based on the following important criterion of existence of a measure with given properties. Proposition 1.7 (The abstract Kummer congruences). (see [Kat3]). Let {fi } be a system of continuous functions fi ∈ C(Y, Op ) in the ring C(Y, Op ) of all continuous functions on the compact totally disconnected group Y with values in the ring of integers Op of Cp such that Cp -linear span of {fi } is dense in C(Y, Cp ). Let also {ai } be any system of elements ai ∈ Op . Then the existence of an Op -valued measure µ on Y with the property  fi dµ = ai Y

is equivalent to the following congruences, for an arbitrary choice of elements bi ∈ Cp almost all of which vanish   bi fi (y) ∈ pn Op for all y ∈ Y implies bi ai ∈ pn Op . (1.31) i

i

Remark 1.8. Since Cp -measures are characterised as bounded Cp -valued distributions, every Cp -measures on Y becomes a Op -valued measure after multiplication by some non-zero constant. Proof of proposition 1.7. The nessecity is obvious since   b i ai = (pn Op − valued function)dµ = Y i  = pn (Op − valued function)dµ ∈ pn Op . Y

In order to prove the sufficiency we need to construct a measure µ from the numbers ai . For a function f ∈ C( Y, Op ) and a positive integer n there exist elements bi ∈ Cp such that only a finite number of bi does not vanish, and  f− bi fi ∈ pn C(Y, Op ), i

according to the  density of the Cp -span of {fi } in C(Y, Cp ). By the assumption (1.31) the value i ai bi belongs to Op and is well defined modulo pn (i.e. does not depend on the choice of bi ). Following N.M. Katz ([Kat3]), we denote this value by “ Y f dµ mod pn ”. Then we have that the limit procedure   f dµ = lim “ f dµ mod pn ” ∈ lim Op /pn Op = Op , ←− Y

gives the measure µ.

n→∞

Y

n

26

1 Non-Archimedean analytic functions

1.3.3 The S-adic Mazur measure Let c > 1 be a positive integer coprime to
M0 = q q∈S

with S being a fixed set of prime numbers. Using the criterion of the proposition 1.7 we show that the Q -valued distribution defined by the formula Ekc (f ) = Ek (f ) − ck Ek (fc ),

fc (x) = f (cx),

(1.32)

turns out to be a measure where Ek (f ) are defined in 1.3.1, f ∈ Step(Y, Qp ) and the field Q is regarded as subfield of Cp . Define the generelized Bernoulli (M) polynomials Bk,f (X) as ∞ 

(M)

Bk,f (X)

k=0

M−1  tk te(a+X)t = , f (a) Mt k! e −1 a=0

(1.33)

and the generalized sums of powers Sk,f (M ) =

M−1 

f (a)ak .

a=0

Then the definition (1.33) formally implies that 1 (M) (M) [B (M ) − Bk,f (0)] = Sk−1,f (M ), k k,f

(1.34)

and also we see that k   k (M) Bk,f (X) = Bi,f X k−i = Bk,f + kBk−1,f X + · · · + B0,f X k . (1.35) i i=0 The last identity can be rewritten symbolically as Bk,f (X) = (Bf + X)k . The equality (1.34) enables us to calculate the (generalized) sums of powers in terms of the (generalized) Bernoulli numbers. In particular this equality implies that the Bernoulli numbers Bk,f can be obtained by the following p-adic limit procedure (see [La1]): Bk,f = lim

n→∞

1 Sk,f (M pn ) M pn

(a p-adic limit),

(1.36)

where f is a Cp -valued function on Y = ZS . Indeed, if we replace M in (1.34) by M pn with growing n and let D be the common denominator of all (M) coefficients of the polynomial Bk,f (X). Then we have from (1.35) that

1.3 Distributions, measures, and the abstract Kummer congruences

! 1 1 2 (Mpn ) (M) Bk,f (M ) − Bk,f (0) ≡ Bk−1,f (M pn ) (mod p n). k kD

27

(1.37)

The proof of (1.36) is accomplished by division of (1.37) by M pn and by application of the formula (1.34). Now we can directly show that the distribution Ekc defined by (1.32) are in fact bounded measures. If we use (1.31) and take the functions {fi } to be all of the functions in Step(Y, Op ). Let {bi } be a system of elements bi ∈ Cp such that for all y ∈ Y the congruence  bi fi (y) ≡ 0 (mod pn ) (1.38) i



holds. Set f = i bi fi and assume (without loss of generality) that the number n is large enough so that for all i with bi = 0 the congruence 1 Sk,fi (M pn ) (mod pn ) M pn

Bk,fi ≡

(1.39)

is valid in accordance with (1.36). Then we see that  Mp −1 n

n −1

Bk,f ≡ (M p )

i

bi fi (a)ak

(mod pn ),

(1.40)

a=0

hence we get by definition (1.32): Ekc (f ) = Bk,f − ck Bk,fc ≡ (M pn )−1

(1.41)

n  Mp −1

i

 bi fi (a)ak − fi (ac)(ac)k

(mod pn ).

a=0

Let ac ∈ {0, 1, · · · , M pn − 1}, such that ac ≡ ac (mod M pn ), then the map a −→ ac is well defined and acts as a permutation of the set {0, 1, · · · , M pn − 1}, hence (1.41) is equivalent to the congruence Ekc (f )

= Bk,f −c Bk,fc ≡ k

 ak − (ac)k c

i

M pn

n Mp −1

bi fi (a)ak

(mod pn ). (1.42)

a=0

Now the assumption (1.37) formally inplies that Ekc (f ) ≡ 0 (mod pn ), completing the proof of the abstact congruences and the construction of measures Ekc . Remark 1.9. The formula (1.41) also implies that for all f ∈ C(Y, Cp ) the following holds f) (1.43) Ekc (f ) = kE1c (xk−1 p where xp : Y −→ Cp ∈ C(Y, Cp ) is the composition of the projection Y −→ Zp and the embedding Zp → Cp .

28

1 Non-Archimedean analytic functions

Indeed if we put ac = ac + M pn t for some t ∈ Z then we see that akc − (ac)k = (ac + M pn t)k − (ac)k ≡ kM pn t(ac)k−1

(mod (M pn )2 ),

and we get that in (1.42): akc − (ac)k ac − ac ≡ k(ac)k−1 n Mp M pn

(mod M pn ).

The last congruence is equivalent to saying that the abstract Kummer congruences (1.31) are satisfied by all functions of the type xk−1 fi for the measure p E1c with fi ∈ Step(Y, Cp ) establishing the identity (1.43).

1.4 Iwasawa algebra and the non-Archimedean Mellin transform 1.4.1 The domain of definition of the non-Archimedean zeta functions In the classical case the set on which zeta functions are defined is the set of complex numbers C which may be viewed equally as the set of all continuous characters (more precisely, quasicharacters) via the following isomorphism: ∼

× C −→ Homcont (R× +, C ) s s −→ (y −→ y )

(1.44)

The construction which associates to a function h(y) on R× + (with certain growth conditions as y → ∞ and y → 0) the following integral  dy Lh (s) = h(y)y s × y R+ (which converges probably not  for all values of s) is called the Mellin transform. For example, if ζ(s) = n≥1 n−s is the Riemann zeta function, then the function ζ(s)Γ (s) is the Mellin transform of the function h(y) = 1/(1 − e−y ): ζ(s)Γ (s) =

∞  0

1 dy ys , −y 1−e y

(1.45)

so that the integral and the series are absolutely convergent for Re(s) > 1. For an arbitrary function of type f (z) =

∞  n=1

a(n)e2iπnz

1.4 Iwasawa algebra and the non-Archimedean Mellin transform

29

with z = x + iy ∈ H in the upper half plane H and with the growth condition a(n) = O(nc ) (c > 0) on its Fourier coefficients, we see that the zeta function L(s, f ) =

∞ 

a(n)n−s ,

n=1

essentially coincides with the Mellin transform of f (z), that is  ∞ Γ (s) dy L(s, f ) = f (iy)y s . (2π)s y 0

(1.46)

Both sides of the equality (1.46) converge absolutely for Re(s) > 1 + c. The identities (1.45) and (1.46) are immediately deduced from the well known integral representation for the gamma-function  ∞ dy Γ (s) = (1.47) e−y y s , y 0 × where dy y is a measure on the group R+ which is invariant under the group translations (Haar measure). The integral (1.47) is absolutely convergent for Re(s) > 0 and it can be interpreted as the integral of the product of an additive character y → e−y of the group R(+) restricted to R× + , and of the multiplicative character y → y s , integration is taken with respect to the Haar measure dy/y on the group R× +. In the theory of the non-Archimedean integration one considers the group Z× S (the group of units of the S-adic completion of the ring of integers Z) ˆ (the completion of instead of the group R× , and the Tate field C = Q p

+

p

an algebraic closure of Qp ) instead of the complex field C. The domain of definition of the p-adic zeta functions is the p-adic analytic group × × XS = Homcont (Z× S , Cp ) = X(ZS ),

where: and the symbol

(1.48)

× ∼ Z× S = ⊕q∈S Zq ,

X(G) = Homcont (G, C× p)

(1.49)

denotes the functor of all p-adic characters of a topological group G (see [Vi1]). 1.4.2 The analytic structure of XS Let us consider in detail the structure of the topological group XS . Define Up = {x ∈ Z× p

| x ≡ 1 (mod pν )},

where ν = 1 or ν = 2 according as p > 2 or p = 2. Then we have the natural decomposition

30

1 Non-Archimedean analytic functions

 XS = X (Z/pν Z)× ×




  × X(Up ). Z× q

(1.50)

q=p

The analytic dstructure on X(Up ) is defined by the following isomorphism (which is equivalent to a special choice of a local parameter): ∼

ϕ : X(Up ) −→ T = {z ∈ C× | |z − 1|p < 1}, p where ϕ(x) = x(1 + pν ), 1 + pν being a topoplogical generator of the multiplicative group Up ∼ = Zp . An arbitrary character χ ∈ XS can be uniquely represented in the form χ = χ0 χ1 where χ0 is trivial on the component Up , and χ1 is trivial on the other component
(Z/pν Z)× × Z× q . q=p

The character χ0 is called the tame component, and χ1 the wild component of the character χ. We denote by the symbol χ(t) the (wild) character which is uniquely determined by the condition χ(t) (1 + pν ) = t with t ∈ Cp , |t|p < 1. In some cases it is convenient to use another local coordinate s which is analogous to the classical argument s of the Dirichlet series: Op −→ XS s −→ χ(s) , where χ(s) is given by χ(s) ((1 + pν )α ) = (1 + pν )αs = exp(αs log(1 + pν )). The character χ(s) is defined only for such s for which the series exp is padically convergent (i.e. for |s|p < pν−1/(p−1) ). In this domain of values of the argument we have that t = (1 + pν )s − 1. But, for example, for (1 + n t)p = 1 there is certainly no such value of s (because t = 1), so that the scoordonate parametrizes a smaller neighborhood of the trivial character than the t-coordinate (which parametrizes all wild characters) (see [Man2], [Man3]). | |z − 1|p < Recall that an analytic function F : T −→ Cp (T = {z ∈ C× p  1}), is defined as the sum of a series of the type i≥0 ai (t − 1)i (ai ∈ Cp ), which is assumed to be absolutely convergent for all t ∈ T . The notion of an analytic function is then obviously extended to the whole group XS by shifts. The function ∞  F (t) = ai (t − 1)i i=0

is bounded on T iff all its coefficients ai are universally bounded. This last fact can be easily deduced for example from the basic properties of the Newton

1.4 Iwasawa algebra and the non-Archimedean Mellin transform

31

polygon of the series F (t) (see [Kob1], [Vi1], [Vi2]). If we apply to these series the Weierstrass preparation theorem (see [Kob2], [La1], [Man1] and theorem 1.3) we see that in this case the function F has only a finite number of zeroes on T (if it is not identically zero). In particular, consider the torsion subgroup XStors ⊂ XS . This subgroup is discrete in XS and its elements χ ∈ XStors can be obviously identified with primitive Dirichlet characters χmodM such that the support S(χ) = S(M ) of the conductor of χ is containded in S. This identification is provided by a fixed embedding denoted ×

ip : Q → C× p if we note that each character χ ∈ XStors can be factored through some finite factor group (Z/M Z)× : × ip

× × χ : Z× S → (Z/M Z) → Q → Cp ,

and the smallest number M with the above condition is the conductor of χ ∈ XStors . The symbol xp will denote the composition of the natural projection Z× S → × × Zp and of the natural embedding Z× p → Cp , so that xp ∈ XS and all integers k can be considered as the characters xkp : y −→ y k . Let us consider a bounded Cp -analytic function F on XS . The above statement about zeroes of bounded Cp -analytic functions implies now that the function F is uniquely determined by its values F (χ0 χ), where χ0 is a fixed character and χ runs through all elements χ ∈ XStors with possible exclusion of a finite number of characters in each analyticity component of the decomposition (1.50). This condition is satisfied, for example, by the set of characters χ ∈ XStors with the S-complete conductor (i.e. such that S(χ) = S), and even for a smaller set of characters, for example for the set obtained by imposing the additional assumption that the character χ2 is not trivial (see [Man2], [Man3], [Vi1]). 1.4.3 The non-Archimedean Mellin transform Let µ be a (bounded) Cp -valued measure on Z× S . Then the non-Archimedean Mellin transform of the measure µ is defined by  Lµ (x) = µ(x) = xdµ, (x ∈ XS ), (1.51) Z× S

which represents a bounded Cp -analytic function Lµ : XS −→ Cp .

(1.52)

Indeed, the boundedness of the function Lµ is obvious since all characters x ∈ XS take values in Op and µ also is bounded. The analyticity of this function

32

1 Non-Archimedean analytic functions

expresses a general property of the integral (1.51), namely that it depends analytically on the parameter x ∈ XS . However, we give below a pure algebraic proof of this fact which is based on a description of the Iwasawa algebra. This description will also imply that every bounded Cp -analytic function on XS is the Mellin transform of a certain measure µ. 1.4.4 The Iwasawa algebra (see [La1]). Let O be a closed subring in Op = {z ∈ Cp | |z|p ≤ 1}, (i ∈ I),

G = lim Gi , ←− i

πij

a profinite group. Then the canonical homomorphism Gi ←− Gj induces a homomorphism of the corresponding group rings O[Gi ] ←− O[Gj ]. Then the completed group ring O[[G]] is defined as the projective limit O[[G]] = lim O[[Gi ]], ←−

(i ∈ I).

i

Let us consider also the set Dist(G, O) of all O-valued distributions on G which itself is an O-module and a ring with respect to multiplication given by the convolution of distributions, which is defined in terms of families of functions (i) (i) µ1 , µ2 : Gi −→ O, (see the previous section) as follows:  (i) (i) µ1 (y1 )µ2 (y2 ), (µ1  µ2 )(i) (y) =

(y1 , y2 ∈ Gi )

(1.53)

y=y1 y2

Recall also that the O-valued distributions are identified with O-valued measures. Now we describe an isomorphism of O-algebras O[[G]] and Dist(G, O). In this case when G = Zp the algebra O[[G]] is called the Iwasawa algebra. Theorem 1.10. (a) Under the same notation as above there is the canonical isomorphism of O-algebras ∼

Dist(G, O) −→ O[[G]].

(1.54)

(b) If G = Zp then there is an isomorphism ∼

O[[G]] −→ O[[X]],

(1.55)

where O[[X]] is the ring of formal power series in X over O. The isomorphism (1.55) depends on a choice of the topological generator of the group G = Zp .

1.4 Iwasawa algebra and the non-Archimedean Mellin transform

33

1.4.5 Formulas for coefficients of power series We noticed above that the theorem 1.10 would imply a description of Cp -analytic bounded functions on XS in terms of measures. Indeed, these functions are defined on analyticity components of the decomposition (1.50) as certain power series with p-adically bounded coefficients, that is, power series, whose coefficients belong to Op after multiplication by some constant from C× p . Formulas for coefficients of these series can be also deduced from the proof of the theorem. However, we give a more direct computation of these coefficients in terms of the corresponding measures. Let us consider the component aUp of the set Z× S where
a ∈ (Z/pν Z)× × Z× q , q=

and let µa (x) = µ(ax) be the corresponding measure on Up defined by restric∼ tion of µ to the subset aUp ⊂ Z× S . Consider the isomorphism Up = Zp given by: y = γ x (x ∈ Zp , y ∈ Up ), with some choice of the generator γ of Up (for example, we can take γ = 1+pν ). Let µ a be the corresponding measure on Zp . Then this measure is uniquely determined by values of the integrals   x dµ a (x) = ai , (1.56) i Zp   with the interpolation polynomials xi , since the Cp -span of the family " # x (i ∈ Z, i ≥ 0) i is dense in C(Zp , Op ) according to the Mahler’s interpolation theorem for continuous functions on Zp (see 1.2.1 and [Mah]). Indeed, from the basic properties of the interpolation polynomials it follows that  x bi ≡ 0 (mod pn ) (for all x ∈ Zp ) =⇒ bi ≡ 0 (mod pn ). i i We can now apply the abstract Kummer congruences (see proposition 1.7), which imply that for arbitrary choice of numbers ai ∈ Op there exists a measure with the property (1.56). On the other hand we state that the Mellin transform Lµa of the measure µa is given by the power series Fa (t) with coefficients as in (1.56), that is     ∞  x χ(t) (y)dµ(ay) = (1.57) dµa (x) (t − 1)i i Up Z p i=0

34

1 Non-Archimedean analytic functions

for all wild characters of the form χ(t) , χ(t) (γ) = t, |t − 1|p < 1. It suffices to show that (1.57) is valid for all characters of the type y −→ y m , where m is a positive integer. In order to do this we use the binomial expansion x

γ mx = (1 + (γ m − 1)) =

∞   x (γ m − 1)i , i i=0

which implies that 

 m

y dµ(ay) =

γ Zp

up

mx

dµ a (x)

=

  x dµa (x) (γ m − 1)i , i Zp

 ∞  i=0

establishing (1.57). 1.4.6 Example. The S-adic Mazur measure and the non-Archimedean Kubota-Leopoldt zeta function (see [La1], [Ku-Le], [Le2], [Wa]). Let us first consider a positive integer c ∈ Z× S ∩ Z, c > 1 coprime to all primes in S. Then for each complex number s ∈ C there exists a complex distribution µcs on GS = Z× S which is uniquely determined by the following condition µcs (χ) = (1 − χ−1 (c)c−1−s )LM0 (−s, χ),

(1.58)

$ where M0 = q∈S q (see 1.3.1). Moreover, the right hand side of (1.58) is holomorphic for all s ∈ C including s = −1. If s is an integer and s ≥ 0 then according to criterion of proposition 1.7 the right hand side of (1.58) belongs to the field Q(χ) ⊂ Qab ⊂ Q generated by values of the character χ, and we get a distribution with values in Qab . If we now apply to (1.58) the fixed embedding ip : Q → Cp we get a Cp -valued distribution µ(c) = ip (µc0 ) which turns out to be an Op -measure in view of proposition 1.7, and the following equality holds µ(c) (χxrp ) = ip (µcr (χ)). This identity relates the special values of the Dirichlet L-functions at different non-positive points. The function  −1 L(x) = 1 − c−1 x(c)−1 Lµ(c) (x)

(x ∈ XS )

(1.59)

is well defined and it is holomorphic on XS with the exception of a simple pole at the point x = xp ∈ XS . This function is called the non-Archimedean zetafunction of Kubota-Leopoldt. The corresponding measure µ(c) will be called the S-adic Mazur measure.

1.5 Admissible measures and their Mellin transform

35

1.4.7 Measures associated with Dirichlet characters Let ω mod$M be a fixed primitive Dirichlet character such that (M, M0 ) = 1 with M0 = q∈S q. This section gives a construction of the direct image of × ¯ the Mazur measure under the natural map Z× ¯ → ZS where S = S ∪ S(M ), S $ ¯0 = M ¯ q. This construction is used in Chapter III. We show that for any q∈S ¯ 0 ) = 1, c > 1, there exist Cp -measures µ+ (c, ω), positive integer c with (c, M × − µ (c, ω) on ZS which are determined by the following conditions, for s ∈ Z, s > 0:   Cωχ¯ −1 s + −s · (1.60) χxp dµ (c, ω) = (1 − χω(c)c ¯ ) ip × G(ω χ) ¯ ZS
 1 − χ¯ ω (q)q s−1 · ¯ L+ M0 (s, χω), −s 1 − χω(q)q ¯ q∈S\S(χ)

and for s ∈ Z, s ≤ 0,  i−1 p

Z× S



χxsp dµ− (c, ω)

= (1 − χ¯ ω (c)cs−1 )L− ¯ M0 (s, χω),

(1.61)

where δ ¯ = LM¯ (s, χω)2i ¯ L+ M0 (s, χω)

Γ (s) cos(π(s − δ)/2) (2π)s

¯ = LM¯ (s, χω) ¯ L− M0 (s, χω)

(1.62) (1.63)

are the normalized Dirichlet L-functions with δ ∈ {0, 1}, and χω(−1) ¯ = (−1)δ . The function G(ω χ) ¯ denote the Gauss sum of the Dirichlet character ω χ. ¯ The functions (1.60), (1.61) satisfy the functional equation
 1 − χ¯ ω (q)q s−1 (1 − s, χω) ¯ = ¯ L− L+ M0 M0 (s, χω). −s 1 − χω(q)q ¯ q∈S\S(χ)

¯ Indeed, by the definition of the S-adic Mazur measure µc on Z× S , (1.60) and (1.61) are given by  def  −1 dµ+ (c, ω) = Z× xx−1 dµc , p ω Z× ¯ S S  def  xdµ− (c, ω) = Z× x−1 ωdµc , Z× S

¯ S

where x ∈ XS and XS¯ is viewed as a subgroup of XS .

1.5 Admissible measures and their Mellin transform 1.5.1 Non-Archimedean integration This construction was nicely explained by J. Coates and B. Perrin-Riou [Co-PeRi] using the Fourier transform of distributions. Let S be a finite set

36

1 Non-Archimedean analytic functions

of primes containing p. The set on which our non-Archimedean zeta functions are defined is the Cp -adic analytic Lie group XS = Homcont (GalS , C× p ),

(1.64)

where GalS is the Galois group of the maximal abelian extension of Q unramified outside S and infinity. Now we recall the notion of h-admissible measures on GalS and properties of their Mellin transform. These Mellin transforms are certain p-adic analytic functions on the Cp -analytic group XS . Recall that by class field theory the group GalS is described as the projective limit (Z/M Z)× = Z× GalS ∼ = lim S, ←−

(1.65)

M

where M runs over integers with support in the set of primes S (i.e. S(M ) ⊂ S). The canonical Cp -analytic structure on XS is obtained by shifts from the × obvious Cp -analytic structure on the subgroup Homcont (Z× p , Cp ) ⊂ XS . We tors regard the elements of finite order χ ∈ XS as Dirichlet characters whose conductor c(χ) may contain only primes in S, by means of the decomposition × i

∞ × × × χ : A× Q /Q −→ ZS −→ Q −→ C ,

(1.66)

where i∞ is a fixed embedding. The character χ ∈ XStors form a discrete subgroup XStors ⊂ XS . We shall need also the natural homomorphism × × xp : Z× S −→ Zp −→ Cp ,

xp ∈ XS ,

(1.67)

so that all integers k ∈ Z can be regarded as characters of the type xkp : y −→ yk . Recall that a p-adic measure on Z× S may be regarded as a bounded Cp linear form µ on the space C(Z× , C ) of all continuous Cp -valued functions p S C(Z× S , Cp ) −→ Cp



ϕ −→ µ(ϕ) =

Z× S

ϕdµ

which is uniquely determined by its restriction to the subspace Step(Z× S , Cp ) of locally constant functions. We denote by µ(a + (M )) the value of µ on the characteristic function of the set a + (M ) = {x ∈ Z× S

| x ≡ a (mod M )} ⊂ Z× S.

The Mellin transform Lµ of µ is a bounded analytic function Lµ : XS −→ Cp χ −→ Lµ (χ) =

 Z× S

χdµ

on XS , which is uniquely determined by its values Lµ (χ) for the characters χ ∈ XStors .

1.5 Admissible measures and their Mellin transform

37

1.5.2 h-admissible measure A more delicate notion of an h-admissible measure was introduced by Y. Amice, J. V´elu and M.M. Viˇsik (see [Am-Ve], [Vi1]). Let C(Z× S , Cp ) denote the space of Cp -valued functions that can be locally represented by polynomials of degree less than a natural number h of the variable xp ∈ XS introduced above. Definition 1.11. A Cp -linear form µ : C h (Z× S , Cp ) −→ Cp is called an hadmissible measure if for all r = 0, 1, · · · , h − 1 the following growth condition is satisfied          . (xp − ap )r dµ = o |M |r−h  sup p  a∈Z× a+(M) S

p

Note that the notion of a bounded measure is covered by the case h = 1, but the set of 1-admissible measures is bigger; it consists of the so called measures of bounded growth (see [Man2], [Vi1]), which are characterized by the property that they grow on open compact sets a + (M ) ⊂ Z× S slower than o(|M |−1 elu and M.M. Viˇsik) p ). We know (essentially due to Y. Amice, J. V´ that each h-admissible measure can be uniquely extended to a linear form on the Cp -space of all locally analytic functions so that one can associate to its Mellin transform Lµ : XS −→ Cp  χ −→ Lµ (χ) = χdµ Z× S

which is a Cp -analytic function on XS of the type o(log(xhp )). Moreover, the measure µ is uniquely determined by the special values of the type Lµ (χxrp ) with χ ∈ XStors and r = 0, 1, · · · , h−1. First example of h-admissible measures and their L-functions concerned the L-function Lf (s, χ) =

∞ 

χ(n)an n−s

n=1

of an elliptic normalized Hecke cusp eigenform f (z) =

∞ 

an exp(2iπz)

(Im(z) > 0)

n=1

with ap divisible by p (the supersingular case). This notion is used in Chapter 4. We can consider another example, provided by the analytic function log(x). For anys ∈ N let us define a distribution µs by µs (χ) = µ(χxsp ) = s log(1 + pν ),

for all χ ∈ XStors

and s ∈ N.

38

1 Non-Archimedean analytic functions

This sequence of distributions turns out to be a 2-admissible measure. Indeed, for r = 0, 1    (xp − ap )r dµ = χmodM χ−1 (a) Z× χ(x)(xp − ap )r dµ S a+(M) " 0, if r = 0,  = log(1 + pν ) χmodM χ−1 (a), if r = 1  But, in the case r = 1 the last sum χmodM χ−1 (a) is equal to 0 if a = 1, and if a = 1 then this sum is equal to ϕ(M ). So we see that the conditions of the definition 1.11 are satisfied, so that we obtain a 2-admissible measure. If we look now at its Mellin transform  xdµ, Lµ (x) = Z× S

we see immediately that Lµ (χxp ) = log(xp ) which is a Cp -analytic function on XS× of the type o(log(xp )2 ).

1.6 Complex valued distributions, associated with Euler products In this section we give a general construction of distributions, attached to rather arbitrary Euler products. This construction provides a generalization of measures, which were first introduced by Y.I. Manin, B. Mazur and H.P.F. Swinnerton-Dyer (see [Man2], [Maz-SD]). Our construction ([Pa2], [Pa3]) was already successfully used in several problems concerning the p-adic analytic interpolation of Dirichlet series (see [Ar], [Co-Schm], [Sc]). 1.6.1 Dirichlet series Let S be a fixed finite set of primes and D(s) =

∞ 

an n−s

(s, an ∈ C)

(1.68)

n=1

be a Dirichlet series with the following multiplicativity property of its coefficients an :
 Fq (q −s )−1 an n−s , (1.69) D(s) = q∈S

n=1

(n,S)=1

where the condition (n, S) = 1 means that n is not divisible by any prime in S, and Fq (X) are polynomials with the constant term equal to 1:

1.6 Complex valued distributions, associated with Euler products mq 

Fq (X) = 1 +

Aq,i X i .

39

(1.70)

i=1

We assume also that the series (1.68) is absolutely convergent in some right half plane Re(s) > 1 + c (c ∈ R). This assumption is satisfied in most cases, for example, when the coefficients an satisfy the estimate |an | = O(nc ). For a Dirichlet character χ : (Z/M Z)× → C× modulo M ≥ 1 the twisted Dirichlet series is defined by ∞  D(s, χ) = χ(n)an n−s . (1.71) n=1

For all s ∈ C such that the series (1.68) is absolutely convergent, let us define the function Ps : Q → C by the equality ∞ 

Ps (x) =

e(nx)an n−s

(e(x) = exp(2iπx)).

(1.72)

n=1

Using the functions (1.72) we construct C-valued distributions µs on the compact group Z× S such that for every primitive Dirichlet character χ viewed × the value of the series D(s, χ) ¯ at s with as a homomorphism χ : Z× S → C Re(s) > 1 + c is canonically expressed in terms of the integral   def χdµs = χ(a)µs (a + (M )). Z× S

amodM

(a,S)=1

Let for every q ∈ S, α(q) denote a fixed root of the inverse polynomial X

mq

Fq (X

−1

)=X

mq

+

mq 

Aq,i X mq −i

i=1

(that is, an inverse root of Fq (X)). We suppose that α(q) = 0 for every q ∈ S, and let us extend the definition of numbers α(n) to all positive integers whose support is contained in S (by multiplicativity):
α(q)ordq (n) (S(q) ⊂ S). α(n) = q∈S

Let us define an auxiliary polynomial mq −1

Hq (X) = 1 +



Bq,i X i

(1.73)

i=1

by means of the relation Fq (X) = (1 − α(q)X)Hq (X),

(1.74)

40

1 Non-Archimedean analytic functions

which implies the identities Bq,i = −

mq 

(i = 1, · · · , mq − 1)

Aq,j α(q)i−j

(1.75)

j=I+1

for the coefficients of the polynomial (1.73). Let us also introduced the following finite Euler product
 B S (n)n−s = Hq (q −s ), (1.76) q∈S

S(n)⊂S

in which the coefficients B S (n) are given by means of (1.75), namely,
B S (n) = B S (q ordq (n) ) (S(n) ⊂ S), q∈S

"

with

Bq,i , for i < mq 0, otherwise. Now we state the main result of the section. S

i

B (q ) =

(1.77)

Theorem 1.12. (a) For any choice of the inverse roots α(q) = 0 (q ∈ S), and for any s from the convergency region of the series (1.68) there exists a distribution µs = µs,α on Z× S whose values on open compact subsets of the are given by the following type a + (M ) ⊂ Z× S a  M s−1  n µs (a + (M )) = n−s , B S (n)Ps (1.78) α(M ) M S(n)⊂S

so that the sum in (1.78) is finite and the numbers B S (n) are defined by (1.77). (b) For any primitive Dirichlet character χ viewed as a function χ : Z× S → × C the following equality holds 
    Cχs−1 −s G(χ)D(s, χ), ¯ 1 − χ(q)α(q)−1 q −s Hq χ(q)q ¯ χdµs = α(Cχ ) Z× S q∈S\S(χ)

(1.79) with G(χ) =



 χ(a)e

amodCχ

a Cχ



being the Gauss sum, Cχ the conductor, and S(χ) the support of the conductor of χ. Remark 1.13. The distribution µs can be obtained as the Fourier transform of the standard zeta-distribution attached to the Diriclet series
 D(s, χ) = Hq (χ(q)q −s ) = χ(n)a0 (n)n−s , q∈S

n≥1

where a0 (qn) = α(q)a0 (n) for each n ∈ N and q ∈ S (see [Co-PeRi]).

1.6 Complex valued distributions, associated with Euler products

41

Proof of theorem 1.12 The following proof of this theorem differs from that given in [Pa2] and is based on the compatibility criterion (proposition 1.7). We check that the sum  χ(a)µs (a + (M )), (1.80) amodM

(a,S)=1

does not depend on the choice of a positive integer M with the condition Cχ |M , S(M ) = S. This will be provided by a calculation which also implies that (1.80) coincides with the right hand side of (1.79) (and therefore is independent of M ). Lemma 1.14. For an arbitrary positive integer n and Cχ |M put  χ(a)e(an/M ). Gn,M = amodM

(a,S)=1

Then the following holds 

M Gn,M (χ) = G(χ) Cχ

−1

µ(d)d

 χ(d)δ

d|(M/Cχ )

dn (M/Cχ )

 χ ¯

dn (M/Cχ )

,

in which µ denotes the M¨ obius function, δ(x) = 1 or 0 according as x ∈ Z or not, and we assume that the character χ is primitive modulo Cχ . Proof. The proof of this lemma is deduced from the well known property of the M¨ obius function: "  1, if n = 1; µ(d) = 0, if n > 1. d|n

Consequently, Gn,M takes the following form:  µ(d)χ(a)e(an/M ) = d|(a,M )

amodM

=

 

d|M

=



µ(d) µ(d)d

χ(d)

d|M

=



d|(M/Cχ )

χ(da1 )e(da1 n/M )

a1 modM/d −1

−1

µ(d)d



a1 modM 

χ(d)δ

χ(a1 )e(da1 n/M ) dn (M/Cχ )

since χ(a1 ) depend only on a1 modCχ , and



 a1 modM

χ(a1 )e(da1 n/M ),

42

1 Non-Archimedean analytic functions

e(a1 dn/M ) = e((a1 /Cχ )(dn/(M/Cχ ))). In the above equality we changed the order of summation, then we replaced the index of summation a by da1 and extended a system of residue classes a1 modM/d to a system a1 modM . Now we transform the summation into that one modulo Cχ . It remains to use the well known property of Gauss sums (see, for example, [Sh1], lemma 3.63): Gn,Cχ (χ) = χ(n)G(χ), ¯ establishing the lemma. In order to deduce the theorem, we now transform (1.80), taking into account the definition (1.78) and lemma 1.14: M s−1 α(M )



χ(a)



B S (n)n−s

n

amodM



an1 e

 ann 

n1

(a,M)=1

1

M

n−s 1 =

(1.81)

M s−1   S B (n)n−s an1 n−s 1 Gnn1 ,M (χ) α(M ) n n 1  Ms = G(χ) B S (n)n−s an1 n−s 1 · α(M )Cχ n n1   µ(d) dnn1 dnn1 χ(d)δ · χ ¯ . d (M/Cχ ) (M/Cχ ) =

d|(M/Cχ )

From the last formula we see that non-vanishing terms in the sum over n and n1 must satisfy the condition (M/Cχ d)|nn1 . Let us now split n1 into two factors n1 = n 1 · n 1 so that S(n 1 ) ⊂ S, and (n 1 , S) = 1. Then   nn 1 dnn1 ¯ χ ¯ = χ(n ¯ 1 )χ (1.82) (M/Cχ ) (M/Cχ d) 

and

dnn1 (M/Cχ )





(1.83)

since (n 1 , M ) = 1. According to (1.69) one has
 an1 n 1 −s = Fq (q −s )−1 .

(1.84)

S(n1 )⊂S

=δ

nn 1 (M/Cχ d)

,

δ

q∈S

Now we use the definition of the finite Euler product (1.76) and of the polynomials Hq (X) which we rewrite here in the form

 B S (n)n−s Fq (q −s )−1 = (1 − α(q)q −s )−1 . n

q∈S

q∈S

1.6 Complex valued distributions, associated with Euler products

43

Consequently, 

B S (n)n−s

 S(n1 )⊂S

n



an1 n 1 −s =

an2 n−s 2

S(n2 )⊂S

and for S(n2 ) ⊂ S we have that 

α(n2 ) =

n2 =nn1

B S (n)an1 .

(1.85)

Keeping in mind (1.82) and (1.83) we transform (1.81) to the following  µ(d) χ(d) χ(n ¯ 1 )an1 n 1 −s · (1.86) d  d|(M/Cχ ) (n1 ,S)=1    nn1 nn 1 · B S (n)an1 (nn 1 )−s δ χ ¯ . (M/Cχ d) (M/Cχ d)   Ms G(χ) α(M )Cχ n



n,n1

Now we transform (1.86) with the help of the relation (1.85), taking into account that non-zero summands can only occur for such n2 = nn 1 which are divisible by M/(Cχ d), (i.e. we put n2 = (M/Cχ d)n3 , S(n3 ) ⊂ S). We also note that by the definition of our Dirichlet series we have
 −s B S (n)an1 (nn 1 )−s = D(s, χ) ¯ Fq (χ(q)q ¯ ). n,n1

q∈S\S(χ)

Therefore (1.86) transforms to the following Ms G(χ) α(M )Cχ






−s Fq (χ(q)q ¯ )

q∈S\S(χ)

d|(M/C )



µ(d) χ(d)· d

χ  −s n3 M n3 M · χ(n ¯ 3 )α = Cχ d cχ d S(n3 )⊂S
 Cχs−1 −s G(χ)D(s, χ) ¯ = Fq (χ(q)q ¯ ) µ(d)ds−1 χ(d)α(d)−1 · α(Cχ ) q∈S\(χ) d|(M/Cχ )  χ(n ¯ 3 )α(n3 )n−s . · 3



S(n3 )⊂S

The proof of the theorem is accomplished by noting that 
µ(d)ds−1 χ(d)α(d)−1 = (1 − χ(q)α(q)−1 q s−1 ), 

d|(M/Cχ )




q∈S\S(χ)

χ(n ¯ 3 )α(n3 )n−s 3

=

S(n3 )⊂S




−s −1 (1 − χ(q)α(q)q ¯ )

q∈S\S(χ)

=

q∈S\S(χ)

−s −1 −s Fq (χ(q)q ¯ ) Hq (χ(q)q ¯ ).

44

1 Non-Archimedean analytic functions

1.6.2 Concluding remarks This construction admits a generalization [Pa4] to the case of rather general Euler products over prime ideals in algebraic number fields. These Euler products have the form
 an N (n)−s = Fp (N (p)−s )−1 , D(s) = n

p

where n runs over the set of integrals ideals, and p over the set of prime ideals of integers OK of a number field K, with N (n) denoting the absolute norm of an ideal n, and Fp ∈ C[X] being polynomials with the condition Fp (0) = 1. In [Pa4] we constructed certain distributions, which provide integral representations for special values of Dirichlet series of the type  χ(n)an N (n)−s , D(s, χ) = n

where χ denote a Hecke character of finite order, whose conductor consists only of prime ideals belonging to a fixed finite set S of non-Archimedean places of K. The main result of [Pa4] provides a generalization of theorem 4.2 of the earlier work of Yu.I. Manin [Man3]. In the construction of non-Archimedean convolutions of Hilbert modular forms given in chapter 4, we give another approach to local distributions, which differs from the given above and is applicable only to certain Dirichlet series (namely, to convolutions of Rankin type). However, there is an interesting link between these two approaches, which is based on a general construction of p-adic distributions attached to motives. It turns out that both types of distributions can be obtained using the Fourier transform in the distribution space (see [Co-PeRi], [Co-Schn]).

2 Siegel modular forms and the holomorphic projection operator

This chapter contains some preparatory facts which we shall use in the construction of non-Archimedean standard zeta-functions in the next chapter. We recall main properties of Siegel modular forms and of the action of the Hecke algebra on them, as well as the definitions of spinor zeta-functions and standard zeta-functions, see also [An1], [An3]. Then in the second section, we present some standard results on theta series with Dirichlet character (see [An-Ma1], [An-Ma2], [St2]), and recal the definitions of Siegel-Eisenstein series, and of Rankin type convolutions of Siegel modular forms. Also, we recall their relation with the standard zeta functions (Andrianov’s identity). In the third section, we give an exposition of some results of G. Shimura and P. Feit on real analytic Siegel-Eisenstein series and their analytic continuation in terms of confluent hypergeometric functions (see [Fe], [Sh5], [Sh7]). These results extend previous results of V.L. Kalinin [Kal] and R. Langlands [Ll]. In the final section, a detailed study of holomorphic projection operator and its basic properties is given. Note that the most complete and elegant exposition of properties of the holomorphic projection operator on symplectic and unitary groups is contained in [Sh9]. The formula of theorem 2.18 provides an explicit formula for computing the holomorphic projection onto the space of holomorphic (not necessary cusp) modular forms for functions belonging to a wide class of C ∞ -Siegel modular forms. Previously the holomorphic projection operator onto the space of cusp form was studied by J. Sturm [St1], [St2], B. Gross and D. Zagier [Gr-Za] under some assumptions on the growth of modular forms. Theorem 2.16 gives an explicit description of the action of this operator in terms of Fourier expansions. Here we also give a very explicit formula (2.121) for the special (critical) values of the confluent hypergeometric function. Notations Let A be a commutative ring with identity, then Mr,s (A) denote the set of all (r × s)-matrices with coefficients in A. For z ∈ Mr (C) put er (z) = e(tr(z)) with e(u) = exp(2iπu) for u ∈ C. We denote by tz ∈ Mr,s (A) the matrix, which

M. Courtieu and A. Panchishkin: LNM 1471, pp. 45–93, 2004. c Springer-Verlag Berlin Heidelberg 2004


46

2 Siegel modular forms and the holomorphic projection operator

is transpose to z ∈ Mr,s (A), and write ξ[η] for tηξη. For a non-degenerate square matrix ξ we put ξ
= tξ −1 . If ξ is a hermitian matrix then we write ξ ≥ 0 or ξ > 0 according as ξ is non negative or positive definite. Let Hm denote the Siegel upper half plane on the degree m, Hm = {z ∈ Mm (C) |

t

z = z = x + iy, y > 0},

so that Hm is a complex analytic variety whose dimension is denoted by m = m(m + 1)/2. Let the symbol Am denote the lattice of all half integral symetric matrices in the vector space V = {y ∈ Mm (R) | ty = y}. This lattice is dual to the latice L = Mm (Z) ∩ V with respect to the pairing given by (u, v) −→ em (uv). For a function f : Hm −→ C of the form  c(ξ)em (ξz) (z ∈ Hm ) f= ξ∈Am

and for a positive integer N we use the notations  c(ξ)em (N ξz), f |V (N )(z) = f (N z) = f |U (N )(z) = fρ =

 ξ∈A m

ξ∈Am

c(N ξ)em (ξz), c(ξ)em (ξz),

ξ∈Am

as well as the notation of A.N. Andrianov for the action of the Frobenius elements Π + (q), Π − (q) given in 2.1.8. Moreover, for N ≥ 1 and an integer k we put √ f |W (N )(z) = f |k W (N )(z) = det( Nz)k f (−(N z)−1 ), so that (f |W (N ))|W (N ) = (−1)mk f.

2.1 Siegel modular forms and Hecke operators 2.1.1 Symplectic group and Siegel upper half plane (see [An1], [An5], [Sh3], [Si2], [Fr], [Maa]). Let G = GSpm be the algebraic subgroup of GL2m defined by GA = {γ ∈ GL2m (A) |

γJm γ = ν(γ)Jm , ν(γ) ∈ A× },

t

for any commutative ring A, where  0m −1m Jm = . 1m 0m

(2.1)

2.1 Siegel modular forms and Hecke operators

47

The elements of GA are characterized by the conditions bta − atb = dtc − ctd = 0m , 

and if γ=

ab cd



dta − ctb = 1m ,

∈ GA then γ −1 = ν(γ)−1



d −tb −tc ta t

(2.2) .

The multiplier ν defines a homomorphism ν : GA → A× so that ν(γ)2m = det(γ)2 and ker(ν) is denoted by Spm (A). We also put G∞ = GR ,

G+ ∞ = {γ ∈ G∞ | ν(γ) > 0},

+ G+ Q = G∞ ∩ GQ .

(2.3)

The group G+ ∞ acts transitively on the upper half plane Hm by the rule   ab −1 + z −→ γ(z) = (az + b)(cz + d) γ= ∈ G∞ , z ∈ Hm cd so that the scalar matrices acts trivially, and Hm can be identified with a homogeneous space of the group Spm (R). Let Km denote the stabilizer of the point i1m ∈ Hm in the group Spm (R), Km = {γ ∈ Spm (R) | γ(i1m ) = i1m }, then there is a bijection Spm (R)/Km  Hm and Km = Spm (R) ∩ SO2m . The group Km is a maximal compact subgroup of the Lie group Spm (R) and it can be identified  with the group U(m) of all unitary m × m-matrices via the ab map γ = −→ a + ib. We adopt also the notations cd

dxij , dy = , dz = dxdy, (2.4) dx = i≤j

d× y = det(y)−κ dy,

i≤j

d× z = det(y)−κ dz,

where z = x + iy, x = (xij ) = tx, y = (yij ) = ty > 0. Then d× z is a differential × on Hm invariant under the action of the group G+ ∞ , and the measure d y is invariant under the action of elements a ∈ GLm (R) on Y = {y ∈ Mm (R) |

t

y = y > 0}

defined by the rule y −→ taya. 2.1.2 Siegel modular forms Let us consider the Siegel modular group denoted Γ m = Spm (Z) and let Γ ⊂ G+ Q be an arbitrary congruence subgroup. This means that Γ is commeasurable with Γ m in G+ Q modulo its center (i.e. as a group of transformations of Hm ), and Γ ⊃ Γ m (N ) for some N ∈ N, where

48

2 Siegel modular forms and the holomorphic projection operator

Γ m (N ) =

"  ab γ= ∈ Γ m | γ ≡ 12m cd

# (mod N )

is the main congruence subgroup of level N in Γ m . In order to give the general definition of modular forms we consider ρ : GLm (C) →  GLr (C) a rational ab representation which will also be denoted by ρ. For γ = ∈ G+ ∞ and cd for any complex valued function f : Hm → Cr we use the notation f |ρ γ(z) = ρ(cz + d)−1 f (γ(z)).

(2.5)

Definition 2.1. A function f : Hm → Cr is called a holomorphic modular form of weight ρ on Γ if the following conditions (2.6) - (2.8) are satisfied: f |ρ = f, f is holomorphic on Hm ,

(2.6) (2.7)

if m = 1 then f is holomorphic at cusps of Γ.

(2.8)

Let Mρ (Γ ) be the complex vector space of functions satisfying the above conditions. For each f ∈ Mρ (Γ ) there is the following Fourier expansion  f (z) = c(ξ)em (ξz), ξ

where c(ξ) ∈ Cr , ξ run over all ξ = tξ ∈ Mm (Q), ξ > 0 (for m > 1 the last condition automatically follows by the Koecher principle). More precisely, let M be the smallest integer such that # " 1m M u t Γ ⊃ | u ∈ Mm (Z), u = u 0m 1m and we put A = Am = {ξ = (ξij ) ∈ Mm (R) | ξ = tξ, ξij , 2ξii ∈ Z} , B = Bm = {ξ ∈ A | ξ ≥ 0} , C = Cm = {ξ ∈ A | ξ > 0} . Then Am is a lattice of half-integral matrices in the R-vector space of | tx = x} dual to the lattice symmetric matrices V = {x ∈ Mm (R) L = Mm (Z) ∩ V with respect to the action (ξ, x) −→ em (ξx) and for each f ∈ Mρ (Γ ) there is the following Fourier expansion  f (z) = c(ξ)em (ξz), (2.9) ξ∈M −1 B

Moreover, for each γ ∈ G+ Q we have that f |ρ γ ∈ Mρ (Γ (γ)), where Γ (γ) is a congruence subgroup,

2.1 Siegel modular forms and Hecke operators



(f |ρ γ)(z) =

cγ (ξ)em (ξz),

49

(2.10)

ξ∈Mγ−1 B

with cγ (ξ) ∈ Cr , Mγ ∈ N. A form f is called a cusp form if for all ξ with det(ξ) = 0 in expansion (2.10) one has cγ (ξ) = 0 for all γ ∈ G+ Q that is 

(f |ρ γ)(z) =

cγ (ξ)em (ξz).

ξ∈Mγ−1 C

We denote by Sρ (Γ ) ⊂ Mρ (Γ ) the subspace of cusp forms. Definition of the vector spaces M(N, ψ). Let us consider congruence subgroups Γ1m (N ) ⊂ Γ0m (N ) ⊂ Γ m (N ) = Spm (Z), defined by "  # ab m Γ0 (N ) = γ = ∈ Spm (Z) | c ≡ 0m (mod N ) , "  c d ab Γ1m (N ) = γ = ∈ Spm (Z) | c ≡ 0m (mod N ), cd # det(a) ≡ 1 (mod N ) , and let r = 1, ρ(x) = ρk (x) = det(x)k ( k ∈ N ). Then the vectore space M(Γ1m (N )) has already been defined, and we set " Mkm (N, ψ) = f ∈ Mρ (Γ1m (N )) | f |ρ γ = ψ(det(a))f (2.11) #  ab for all γ = ∈ Γ0m (N ) , cd where ψ is a Dirichlet character modulo N . Elements f ∈ Mkm (N, ψ) admits a Fourier expansion of the form  f (z) = c(ξ)em (ξz), (2.12) ξ∈Bm

with z ∈ Hm , c(ξ) ∈ C, and the condition c(uξ tu) = ψ(det(u))det(u)k c(ξ)

(ξ ∈ Bm , u ∈ GLm (Z)).

(2.13)

Put k (N, ψ) = Mkm (N, ψ) ∩ Sρk (Γ1k (N )). Sm k (N, ψ) and h ∈ Mkm (N, ψ) The Petersson scalar product. For f ∈ Sm the Petersson scalar product is defined by  f, h N = f (z)h(z)det(y)k d× z, (2.14) Φ0 (N )

where Φ0 (N ) = Γ0m (N )\Hm is a fundamental domain for the group Γ0m (N ).

50

2 Siegel modular forms and the holomorphic projection operator

The Siegel operator connects the vector spaces Mkm (N, ψ) for different values of m. If f ∈ Mkm (N, ψ), z ∈ Hm−1 and λ > 0 then we have that  z 0 ∈ Hm , 0 iλ and it follows from (2.12) that there exists the limit    z 0 ξ 0 (Φf )(z ) = lim f c = em−1 (ξ z ), 0 iλ 0 0 λ→∞

(2.15)

ξ∈Bm−1

where c(ξ) are the Fourier coefficients of f . Then Φf ∈ Mkm−1 (N, ψ), (we put Mk0 (N, ψ) = 0. We then have that % & k Sm (N, ψ) ⊆ ker Φ = f ∈ Mkm (N, ψ) | Φf = 0 , (for N = 1 both sets coincide, see [Maa]). k (N, ψ) then there is the Estimates for Fourier coefficients. If f ∈ Sm following upper estimate   (z = x + iy ∈ Hm ), |f (z)| = O det(y)−k/2 (2.16) which provides us also with the estimate   |c(ξ)| = O det(ξ)k/2 .

(2.17)

For modular (not necessary cusp) forms  f (z) = c(ξ)em (ξz) ∈ Mkm (N, ψ) ξ∈Bm

there is the upper estimate of their growth: |c(ξ)| ≤ c1

m


(1 + λkj ),

(2.18)

j=1

with λ1 , · · · , λm being eigenvalues of the matrix y, z = x + iy (see [St2], p 335). In this situation one has also the following estimate |c(ξ)| ≤ c2 det(ξ )k

(2.19) 

in which c2 is a positive constant depending only on f , ξ = tu

ξ 0 0 0

u, where

u ∈ SLm (Z), ξ ∈ Br , det(ξ ) > 0, r < m. We refer the reader to [Fo], [Ki], [Ra1], [Ra3] for a more detailed discussion of various estimates for Fourier coefficients and for growth of Siegel modular forms, as well as for some intersting applications to quadratic forms.

2.1 Siegel modular forms and Hecke operators

51

2.1.3 The Hecke algebra (see [An5], [B¨o2], [Sat]). Let q be a prime, q
N ,   ' ab + −1  ∆ = ∆m (N ) = γ = ∈ G ∩ GL (Z[q ])  2m q Q cd ( ν(γ)± ∈ Z[q −1 ], c ≡ 0m (mod N ) m be a subgroup in G+ Q containing Γ = Γ0 (N ). The following Hecke algebra m over Q: L = Lq (N ) = DQ (Γ, ∆) is then defined as a Q-linear space generated by the double cosets (g) = (Γ gΓ ), g ∈ ∆ of the group ∆, with respect to the subgroup Γ for which multiplication is defined by the standard rule (see [An5], [Sh1] and in 2.1.7 below). We recall the description of the structure of L = Lm q (N ), (q
N ); for each j, 1 ≤ j ≤ m let us denote by ωj an ±1 ±1 automorphism of the algebra Q[x±1 0 , x1 , · · · , xm ] defined on its generators by the rule:

x0 −→ x0 xj , xj −→ x−1 j , xi −→ xi

(where 1 ≤ i ≤ m, i = j).

Then the automorphisms ωj and the permutation group Sm of the variables xi (1 ≤ i ≤ m) generate together the Weyl group W = Wm , and there is the Satake isomorphism : ∼

±1 ±1 Wm Sat : L −→ Q[x±1 0 , x1 , · · · , xm ]

(2.20)

where Wm indicates the subalgebra of elements fixed by Wm . For any commutative Q-algebra A the group Wm acts on the set (A× )m+1 , therefore any homomorphism of Q-algebras λ : L → A can be identified with some element (α0 , α1 , · · · , αm ) ∈ [(A× )m+1 ]Wm .

(2.21)

An explicit description of the Satake isomorphism is given below in 2.1.7. 2.1.4 Hecke operators Any double coset (g) = (Γ gΓ ) (g ∈ ∆ = ∆m q (N )) can be represented as a disjoint union of left cosets: )

t(g)

(g) =

Γ gi ,

i=1

therefore any element X ∈ L of the Hecke algebra L takes the form of a finite linear combination t(X)  X= µi (Γ gi ), i=1

52

2 Siegel modular forms and the holomorphic projection operator

withµi a g= c

∈ Q, gi ∈ ∆. In order to define the Hecke operators we put for any b ∈∆ d (f |k,ψ g)(z) = det(g)k−κ ψ(det(a))det(cz + d)−k f (g(z))

(2.22)

(this convenient notation, suggested by Petersson and Andrianov, is especially useful when dealing with Hecke operators in their normalized form, compare with (2.5)). With this notation the automorphy condition can be rewritten as follows (2.23) (f |k,ψ γ)(z) = f for all γ ∈ Γ = Γ0m (N ). In this case for any 

t(X)

X=

µi (Γ gi ) ∈ L

i=1

we have that the expression 

t(X)

f |X =

µi f |k,ψ gi ,

(2.24)

i=1

is well defined and f |X ∈ Mkm (N, ψ) so that the formula (2.24) gives a representation of the Hecke algebra L = Lm q (N ) on the complex vector space k Mm (N, ψ) (q
N ). 2.1.5 Hecke polynomials ˜ Following A.N.Andrianov (see [An5]), let us consider polynomials Q(z) ∈ ±1 ˜ ∈ Q[x±1 , · · · , x ][z] : Q[x0 , · · · , xm ][z] and R(z) m 0 ˜ ˜ 0 , x1 , · · · , xm ; z) Q(z) = Q(x m

= (1 − x0 z)

(2.25) (1 − x0 xi1 · · · xir z),

r=1 1≤i1 1, then (a) If χ2 = 1 then the function G
(z, s) is an entire function of the variable s; (b) Suppose that χ2 is trivial, then we have • (b1 ) if either 2k ≥ m and m odd or 2k ≥ m and m is even, but (m/2) + k is odd (i.e. µ = 1), then the function G
(z, s) is an entire function of the variable s; • (b2 ) if 2k ≥ m and both numbers m and (m/2)+k are even (i.e. µ = ε(m) = 0) then the function G
(z, s) is an entire function of the variable s with the exclusion of a possible simple pole at the point s = (m + 2 − 2k)/4; • (b3 ) if m > 2k ≥ 0 then the function G
(z, s) is an entire function of the variable s with possible exclusion of simple poles at those points s for which 2s is an integer and [(m − 2k + 3)/2] ≤ 2s ≤ (m + 1 − 2k)/2; • (b4 ) if k = 0 then the function G
(z, s) has a simple pole at the point s = (m + 1)/2 iff χ is trivial, and in this case we have that the function Ress=(m+1)/2 G
(z, s; 0, 1, N ) of the variable z is a non-zero constant.

2.3 Formulas for Fourier coefficients of the Siegel-Eisenstein series

81

Theorem 2.14 (On positivity properties of Fourier expansions of the normalized Siegel-Eisenstein series). Assume that 2k > m and define the numbers A(χ), B(χ), C(χ, k) as follows: (a) If χ2 is not trivial, m is even and µ = ε(k + (m/2)) then put C(χ, k) = (m − 2k + 2 − 2µ)/4;

A(χ) = B(χ) = 1 + (m/2),

(b) If χ2 is trivial, m is even and µ = ε(k + (m/2)) then put A(χ) = B(χ) = (m/2),

C(χ, k) = (m − 2k + 2 − 2µ)/4;

(c) If χ2 is not trivial, m is odd,then put A(χ) = B(χ) = (m + 3)/2,

C(χ, k) = [(1 + m − 2k)/4];

(d) If χ2 is trivial, m is odd, then put A(χ) = (m + 5)/2,

B(χ) = (m + 1)/2,

C(χ, k) = [(3 + m − 2k)/4];

under these notation and assumptions there are the following positivity properties of matrices h ∈ Am , by which the Fourier coefficients of the series G
(z, s) are indexed: (1) if s ≤ 0, s ∈ Z and k + 2s ≥ A(χ) then  b
(h, y, s)em (hz), (2.138) G
(z, s) = Am h>0

(2) if k + s − κ ∈ Z, k + s − κ ≥ 0, (κ = (m + 1)/2), s ≤ C(χ, k) then  b
(h, y, s)em (hz), (2.139) G
(z, s) = Am h≥0

Proof. The proof of this theorem is contained in the book of P.Feit [Fe], theorems 14.1.A – 14.1.F; it is based on a detailed study of poles and residues of the Γ -factor Γ
(h, s) and of the Dirichlet L-function L
(h, χ, k + 2s), which was carried out in [Fe] in terms of the functions f (n, s) = Γ (n + s)/Γ (s); for positive integers n these functions turn out to be polynomials with zeros given by s ∈ Z, s ≤ 0, s + n > 0. It was shown in [Fe] that the factor Γ
(g, s) = Γ (h, s)/Γ (1m , s) is equivalent (up to multiplication by an invertible entire function) to a certain explicitly given polynomial in C[s] (see [Fe], §11). It follows also from this calculation that (a) if χ2 is trivial, m is odd and s = s0 = (m + 2 − 2k)/4 then the function
G (z, s) has a pole at the point s = s0 such that the residue Ress=s0 G
(z, s) has a non negative Fourier expansion;

(2.140)

([Fe], theorem 14.1.C); (b) if χ2 is trivial and m is odd then the function G
(z, s) is finite at the point s = s0 and has a non negative Fourier expansion.

82

2 Siegel modular forms and the holomorphic projection operator

It is essential for our purposes to reformulate the corresponding statements for the series G+ (z, s) and G− (z, s) (see (2.135), (2.136)), which are obtained from G
(z, s) by an additional normalization. The following theorem is an immediate consequence of the theorem 2.13 on holomorphy and the properties (2.138)-(2.140). Theorem 2.15 (On Fourier coefficients with positive matrix numbers). Let m be an even integer such that 2k > m, then: (a) For 2s an integer, s ≤ 0, k + 2s ≥ 1 + m/2 there is the following Fourier expansion  b+ (h, y, s)em (hz), (2.141) G+ (z, s) = Am h>0

where for s > (m + 2 − 2k)/4 in (2.141) non-zero terms only occur for positive definite h > 0, and for all s from (a) with h > 0, h ∈ Am the following identity holds b+ (h, y, s) = W
(y, h, s)L+ N (k + 2s − (m/2), χχh )M (h, χ, k + 2s), with L+ N (s, χ) =

2iδ Γ (s) cos(π(s − δ)/2) LN (s, χ) (2π)s

is the normalized Dirichlet L-function, δ = 0 or 1 according to χ(−1) = (−1)δ , the factor M (h, χ, k + 2s) is defined by (2.132), W
(y, h, s) = 2−mκ det(h)k+2s−κ det(4πy)s R(4πhy; −s; κ − k − s), provided s is an integer, where R(y; n, β) is defined by (2.120), and b+ (h, y, s) = 0 otherwise (if s ∈ Z). (b)If 2s is an integer with k + 2s ≤ m/2, k + s ≥ κ then there is the following Fourier expansion  b− (h, y, s)em (hz), (2.142) G− (z, s) = Am h≥0

and for all s from (b) with h > 0, h ∈ Am the following identity holds b− (h, y, s) = W
(y, h, s)L− N (k + 2s − (m/2), χχh )M (h, χ, k + 2s), where L− N (s, χ) = LN (s, χ), W
(y, h, s) = 2−mκ det(4πy)κ−k−s R(4πhy; k + s − κ, s), provided s + k − κ is an integer, and b− (h, y, s) = 0 otherwise.

2.4 Holomorphic projection and Maass operator

83

Proof. The proof is deduced from the expansions (2.137) if we remember the definition of the normalizing factors and the positivity property in the theorem 2.14. We also note that by (2.129) W
(y, h, s) = em (−ihy)ω(2πy, h; k + s, s)det(y)κ−k−s · ·δ+ (hy)k+s−κ+q/4 δ− (hy)s−κ+p/4 , and then take into account the formula (2.126) for the critical values of the function ω. In case of the odd parity 2s ∈ Z we get vanishing of the Fourier coefficients because of the Γ -factors in (2.135), (2.136).

2.4 Holomorphic projection operator and the Maass differential operator 2.4.1 Holomorphic projection operator We start with describing a vector space on which this operator acts. A function F : Hm → C, F ∈ C ∞ (Hm ) is called a C ∞ -modular form of weight k on the group Γ0m (N ) with a Dirichlet character ψmodN if the following automorphy condition is satisfied: F ((az + b)(cz + d)−1 ) = ψ(det(d))det(cz + d)k F (z) 

for all γ=

ab cd

∈ Γ0m (N )

(compare with (2.11)). The space of functions F with the above condition is ˜ k (N, ψ) there is the following Fourier ˜ k (N, ψ). For all F ∈ M denoted by M m m expansion  F (z) = A(y, h)em (hx), (2.143) h∈Am

where A(y, h) are some C ∞ -functions on Y . The Petersson inner prodk (N, ψ) and uct is defined for an arbitrary holomorphic cusp form f ∈ Sm k ˜ F ∈ Mm (N, ψ) by  f, F N = f (z)F (z)det(y)k−m−1 dxdy, Φ0 (N )

where Φ0 (N ) = Γ0m (N )\Hm is a fundamental domain for the group Γ0m (N ). ˜ k (N, ψ) a function of a bounded growth if for We call a function F ∈ M m each ε > 0 the following integral converges:   |F (z)|det(y)k−1−m e−εtr(y) dydx < ∞ (2.144) X

Y

84

2 Siegel modular forms and the holomorphic projection operator

where

X = {x ∈ Mm (R) | Y = {y ∈ Mm (R) |

x = x, |xij | ≤ 1/2 for all i, j}, y = y > 0}.

t t

˜ k (N, ψ) a function of a moderate Respectively, we call a function F ∈ M m growth if for all z ∈ Hm and for all sufficiently large values of Re(s)  0 the integral  F (w)det(w ¯ − z)−k−|2s| det(Im(w))k+s d× w (2.145) Hm

is absolutely convergent and admits an analytic continuation over s to the point s = 0. The last definition may differ from a traditional one; its meaning is clarified by the following result (Theorem 2.16), which provides a refinement of theorem 1 of the Sturm’s paper [St2]. It will follow from the proof that all functions of bownded growth automatically turn out to be of a moderate growth in the sense of definitions (2.144), (2.145) given above. ˜ k (N, ψ) and k > 2m. Put for h > 0, h ∈ Am Theorem 2.16. Let F ∈ M m  a(h) = c(k, m)−1 det(4h)k−(m+1)/2 A(y, h)em (ihy)det(y)k−1−m dy, Y

(2.146) with

c(t, m) = Γm (t − (m + 1)/2)π −m(t−(m+1)/2) ,

and A(y, h) being coefficients of the expansion (2.143) and suppose that the integral (2.146) is absolutely convergent. Define the function  a(h)em (hz). (2.147) Hol (F )(z) = Am h>0

Then ˜ k (N, ψ) is of a bounded growth then Hol (F )(z) ∈ (a) if the function F ∈ M m k Sm (N, ψ). ˜ k (N, ψ) is of a moderate growth and the ex(b) If the function F ∈ M m pansion (2.143) contains only terms with positive definite matrices h ∈ Am , k (N, ψ) the following then Hol (F )(z) ∈ Mkm (N, ψ). In both cases for all g ∈ Sm equality holds: g, F N = g, Hol (F ) N . (2.148) Remark 2.17. The cusp form Hol (F ) is uniquely determined by (2.148) under the assumptions of (a), but in (b) this equality is not sufficient to identify the modular form Hol (F ). For example, (2.148) does not change if we replace this modular form by adding to it an Eisenstein series (of Siegel or of Klingen type). Part (a) of the theorem 2.16 was established by Sturm [St2].

2.4 Holomorphic projection and Maass operator

85

2.4.2 Poincar´ e series of two variables (of exponential type) of higher level In order to prove the theorem 2.16 we use this kind of Poincar´e serie introduced by H. Klingen [Kl1] and used by S. B¨ ocherer in [B¨ o2] instead of Poincar´e series of one variable from [St2], [Gr-Za]. We consider an element (Γ gΓ ) of the Hecke algebra Lm (N ) with Γ = Γ0m (N ), g ∈ ∆ = ∆m q with   ' ab  ∩ GL2m (Z[q −1 ])  ∆= γ= ∈ G+ Q cd ( ν(γ)± ∈ Z[q ± ], c ≡ 0m (mod N ) , where tgJm g = ν(g)Jm , ν(g) > 0. Put  k Pm (z, w, g, s) = ψ(det(a))j(γ, z)−k−|2s| det(γ(z) + w)k−|2s| ,

(2.149)

γ∈Γ gΓ



ab def ∈ Γ gΓ and uk−|2s| = u−k |u|−2s for u ∈ C× , s ∈ C. The cd series in (2.149) converges absolutely and uniformly on products of the type Vm (d) × Vm (d) for k + Re(2s) > 2m + 1, d > 0, % & Vm (d) = z = x + iy ∈ Hm | y ≥ d1m , tr(txx) ≤ 1/d . γ =

We also put k k (z, w, g, s) = det(Im(z))s det(Im(w))s Pm (z, w, g, s). Pm

(2.150)

The following properties of these series were established by B¨ocherer in [B¨o2] (a) symmetry k k (z, w, g, s) = Pm (w, z, g, s); (2.151) Pm (b) automorphy with respect to both arguments: k Pm (γ(z), γ(w), g, s) k (z, w, g, s), = ψ(det(d1 ))ψ(det(d2 ))j(γ1 , z)k j(γ2 , w)k Pm  ai b i ∈ Γ , i = 1, 2; where γi = ci di (c) action of Hecke operators:

(2.152)

k k Pm (z, w, 1m , s)|k,ψ (Γ gΓ )z = Pm (z, w, g, s) k = Pm (z, w, 1m , s)|k,ψ (Γ gΓ )w ,

(2.153)

where the subscript indicates to which of the variables the Hecke operator is being applied with the action defined by (2.23).

86

2 Siegel modular forms and the holomorphic projection operator k (d) the integral representation: for all f ∈ Sm (N, ψ) we have that k (−¯ z , w, g, s), f (w) N,w = µ(m, k, s)f |k,ψ (Γ gΓ )(z), Pm

(2.154)

with µ(m, k, s) = 2m+m(m+1)/2−2ms+1 i−mk π m(m+1)/2

Γm (k + s − (m + 1)/2) . Γm (k + s)

The proof of the properties (2.151), (2.153) is easily deduced from the symmetry relation j(γ, z)det(γ(z) + w) = j(˜ γ , w)det(˜ γ (w) + z),

(2.155)

which is valid for all γ ∈ Spm (R) with 1 2 1m 0m γ˜ = γ 0m −1m and Γ gΓ = Γ g˜Γ for g ∈ ∆. Then (2.152) is immediately deduced from the definition (2.149). The proof of the integral formula (2.154) is carried out similarly to that in Klingen’s article [Kl2]; for this purpose we may assume that g = 1m . The integration in the left hand side of (2.154) can be reduced by the standard unfolding procedure to that over the whole Siegel upper half plane: notice that Hm = ∪γ γ(Φ0 (N )),

γ ∈ Γ = Γ0m (N ),

where Φ0 (N ) = Γ0m (N )\Hm is a fundamental domain for the group Γ0m (N ). The required property follows then from the integral representation  f (w)det(w¯ − z)−k−|2s| det(Im(w))k+s d× w = (2.156) Hm

= imk 2m(m+1)−2ms−mk Im (k + s − m − 1)f (z)det(Im(z)), where

 det(1m − ww) ¯ s dudv

Im (s) =

(2.157)

Em

=

 π mκ Γ (s + 1 + (m − 1)/2) m 2 Γm (s + m + 1)

denotes the integral studied by Hua Lo-Ken [HuLK]. This integral is absok (N, ψ), the integration in (2.156) is lutely convergent for Re(s) > −1, f ∈ Sm extended over the generalized unit disc Em = {w = u + iv ∈ Mm (C) |

w = w, 1m − ww ¯ > 0}

t

of the degree m, the image of Hm via the Cayley transform

2.4 Holomorphic projection and Maass operator

w → (w − i1m )(w + i1m )−1

87

(w ∈ Hm ).

k (N, ψ) then there is the In order to prove (2.156) we note that if f ∈ Sm following upper estimate

f (z) ≤ cdet(Im(w))−k/2 , and the integral in (2.156) is majorated by  |det(w ¯ − z)|−k−2Re(s) det(Im(w))(k/2)+Re(s) d× w, Hm

which provides the absolute convergence of (2.156) in this domain for (k/2) + Re(s) > m. Next we rewrite the integrand in (2.156) in the form g(w)det(w ¯ − z)−k−s det(Im(w))(k+s) with the holomorphic function g(w) = f (w)det(w − z¯)−s , which is integrated using the Cayley transform, so that (2.156) follows. 2.4.3 Reduction of theorem 2.16 to properties of Poincar´ e series We restrict ourselves to the case of functions F satisfying assumptions of (b). Set in formulas (2.152), (2.154) g = 1m and define a function of two k variables Km (z, w, s) by the equality k k (z, w, s) = µ(m, k, s)−1 Pm (−¯ z , w, 12m , s). Km

(2.158)

We show that the function k Hol (F )(z, s)|s=0 = Km (z, w, s), F (w) N,w |s=0

(2.159)

obtained by analytic continuation of the right hand side to the point s = 0 satisfies all conditions of the theorem, i.e. it coincides with the function defined by (2.146), (2.147): Hol (F )(z, s)|s=0 ∈ Mkm (N, ψ) and the equality g(z), F (z) N = g(z), Hol (F )(z, s)|s=0 N holds for all g ∈ Skm (N, ψ). For this purpose we note that for sufficiently large value of Re(s) the right hand side of (2.159) can be rewritten in the form of an integral over the whole Siegel upper half plane Hm of degree m: k Km (z, w, s), F (w) N,w = µ(m, k, s)−1 det(Im(z))s ·  F (w)det(w ¯ − z)−k−|2s| det(Im(w))k+s d× w, Hm

(2.160)

88

2 Siegel modular forms and the holomorphic projection operator

(due to the assumption on the growth of F the integral (2.160) is absolutely convergent for all Re(s)  0). Next let us consider the subgroup # "  1m b 0 Γ∞ = γ = ± | γ ∈ Γ ⊂ Γ = Γ0m (N ). 0m 1m Then the set {w = u + iv ∈ Hm | u ∈ X, v ∈ Y } 0 is a fundamental domain for the action of Γ∞ on Hm , and we see that for Re(s)  0 the right hand side of (2.160) takes the form    −1 s 2µ(m, k, s) det(Im(z)) F (w) det(w ¯ − z + b)−k−|2s| · X

Y

b∈L

·det(Im(w))k+s d× w = (2.161)   = 2µ(m, k, s)−1 det(Im(z))s F (w)S(w¯ − z, L; k + s, s)· X

Y

·det(Im(w))k+s d× w, where L = Mm (Z) ∩ V is a lattice in V = {x ∈ Mm (R) | function  det(z + b)−k−|2s| S(z, L; k + s, s) =

t

x = x}, and the

b∈L

admits an analytic continuation to all s ∈ C by means of the Fourier expansion of (2.156). Moreover, for k > m we have that S(w ¯ − z, L; k + s, s)|s=0 =  (−2πi)mk Γm (k)−1

(2.162) k−κ

det(h)

em (f (z − w)). ¯

Am h>0

Under the assumption on growth of F the integral admits an analytic continuation to the point s = 0. This analytic continuation can be explicitly given in the form of a Fourier expansion by means of (2.161), (2.162) using the positivity of h. As a result the function Hol (F )(z, s)|s=0 takes the form (4π)m(k−(m+1)/2) Γm (k − (m + 1)/2)−1 Hol (F )(z, s)|s=0 =   k × det(h)k−κ · F (z)em (h(z − w))det(Im(w)) ¯ d w. Am h>0

X

Y

Then the formula (2.146) follows from the obvious equality:   k × F (z)em (h(z − w))det(Im(w)) ¯ d w= X Y  em (hz) Y A(v, h)em (ihv)det(v)k−1−m dv. In order to prove the remaining statements we note that the function

2.4 Holomorphic projection and Maass operator

89

Hol (F )(z, s)|s=0 defined by (2.159) is holomorphic and it satisfies the automorphic properties with respect to Γ0m (N ) of weight k with the Dirichlet character ψ (mod N ). k (z, w, s) and conseIndeed, these properties are satisfied by the function Km quently by (2.160) for Re(s)  0. But the identity expressing the automorphy condition (2.11) does not change by analytic continuation, and we get Hol F (z, s)|s=0 ∈ Mm k (N, ψ) (note for m > 1 also the Koecher principle is applicable). The equality (2.148) is then deduced from definitions (2.158) and (2.159), and from the automorphy property (2.154) of the Poincar´e series. For Re(s)  0 we get k g, Km (z, w, s), F (w) N,w N,z = g, F N

=

(2.163)

k (z, w, s), g(z)

Km N,z , F (z) N,w .

These equalities remain valid by the analytic continuation and we get (2.148). In the equality (2.163) the property k (z, w, s) = K k (z, w, s Km ¯). m

was taken into account. 2.4.4 Fourier expansion of the holomorphic projection of special modular forms When applying the formula (2.146), it is convenient to use the integral representation (2.95) for the Γ -function of degree m, Γm (s) = π m(m−1)/4

m−1


Γ (s − (j/2)),

j=0

This integral representation can be rewritten in the equivalent form  Γm (ν − κ)det(u)κ−ν = det(y)ν−m−1 e−tr(uy) dy (2.164) Y  = det(y)ν−κ em (i(2π)−1 uy)d× y. Y

Moreover, if R(y) ∈ C[yij ] is a polynomial of y = (yij ), i ≤ j then for all ν ∈ Z, ν > m we have that   ! R(y)det(y)ν−κ e−tr(uy) d× y = R(−∂/∂u)e−tr(uy) det(y)ν−κ d× y Y Y  (2.165) = R(−∂/∂u) Γm (ν − κ)det(u)κ−ν , where ∂/∂u = ∂ij , ∂ij = 2−1 (1 + δij ∂/∂uij ). Indeed, it suffices to verify the statement (2.165) for monomials of the form

90

2 Siegel modular forms and the holomorphic projection operator

R(y) =




a(i,j)

yij

,

a(i, j) ∈ Z, a(i, j) ≥ 0.

i≤j

In this particular case this is done by application of the differential operator
a(i,j) ∂/∂uij i≤j

to both sides of the equality (2.164). We shall need formulas (2.146) and (2.165) in a special situation, which is described in the next theorem below. m Theorem 2.18. Let F ∈ Mm k (N, ψ) be a C-modular form F ∈ Mk (N, ψ) which has the form of a product of the type F (z) = g(z)G(z), where  B(h)em (hz), g(z) =

G(z) =

Am h>0

C(h)det(4πy)−n R(4πhy; n, β)em (hz),

Am h≥0

F (z) satisfies one of the two assumptions (a) or (b) of the theorem 2.16 and R(z; n, β) is the polynomial (2.120) defined for any integer n ≥ 0, β ∈ C and z = tz ∈ Mn (C) by ! R(z; n, β) = (−1)mn etr(z) det(z)n+β ∆nm etr(z) det(z)−β , where ∆m = det(∂ij ),

(∂ij = 2−1 (1 + δij )∂/∂ij z, i ≤ j)

is the Maass differential operator. Then the following equality holds  Hol F (z) = B(h1 )C(h2 )P (h2 , h; n, β)em (hz),

(2.166)

Am h=h1 +h2 >0

where P (v, u) = P (v, u; n, β) denotes a polynomial of u = tu = (uij ) and v = tv = (vij ) with the property P (v, u; n, β) ≡ (−1)mn det(v)n

(mod uij ),

(2.167)

and P (v, u; n, β) ∈ Q[u, v] for any β ∈ Q. Proof ( of 2.18). The proof of 2.18 is carried out by a straightforward application of the integral formula (2.146) for the action of Hol on each of the Fourier coefficients of the function F (z):  B(h1 )C(h2 )det(4πy)−n R(4πh2 y; n, β)em (ihz). A(y, h) = Am h=h1 +h2 >0

As a result we get

2.5 Explicit description of differential operators

A(h) =



91

B(h1 )C(h2 )P (h2 , h; n, β),

Am h=h1 +h2 >0

where P (v, u) (2.168)  det(4πu)k−κ = R(4πvy; n, β)det(4πy)−n det(y)k−κ em (2iuy)d× y Γm (k − κ) Y  det(4πu)k−κ = R(4πvy; n, β)det(4πy)−n+k−κ em (2iuy)d× y Γm (k − κ) Y  det(4πu)k−κ R(vy; n, β)det(y)−n+k−κ e−tr(uy) d× y = Γm (k − κ) Y  Γm (k − n − κ) det(u)k−κ R(v · ∂/∂u; n, β) det(u)κ−k+n , = Γm (k − κ) with n ∈ Z, n ≥ 0, β ∈ C. In order to accomplish the proof it suffices to show that the function P (v, u) = P (v, u; n, β) is a polynomial with the desired properties (2.167). This last fact is deduced from the last of the equalities (2.168) and some general properties of the differential operator ∂/∂u which are given below (see also [Kl2]).

2.5 Explicit description of differential operators 2.5.1 The polynomial R(z; r, β) Here we want to describe explicitely the polynomial R(z; r, β), introduced in the section 2.3 (see (2.120)). We show that this is a polynomial invariant for the action of conjugation on the space of symetric matrices y ∈ V = {h ∈ Mm (R) | th = h}. Let us consider the natural representation (0 ≤ r ≤ m) ρr : GLm (C) −→ GL(Λr Cm ) z −→ ρr (z) of the group GLm (C) on the vector space Λr Cm with respect to  the basis {ei1 ∧ · · · ∧ eir : i1 < · · · < ir }. Thus ρr (z) is a matrix of size m r composed of the subdeterminants of z of degree r. Put ρ
r (z) = det(z)ρm−r (tz)−1

(r = 0, 1, · · · , m),

in other words, ρ
m−r (z) is the matrix representing the action of z on Λm−r Cm with respect to the basis dual to the above basis of Λr Cm . Then the representations ρr and ρ
r turn out to be polynomial representations so that for each z ∈ Mm (C) the linear operators ρr (z), ρ
r (z) are well defined. We consider

92

2 Siegel modular forms and the holomorphic projection operator

the differential operators ρr (∂/∂z) and ρ
r (∂/∂z) which associate to each C valued function on Hm a certain Mt (C) valued function on Hm with t = m r . In particular, the Maass differential operator is done by ∆ = ρm (∂/∂z) = ρ
m (∂/∂z) = det(2−1 (1 + δij )∂/∂zij ). We define the invariant polynomials λr (z) = tr(ρr (z)). They are the coefficients of the caracteristic polynomial of z det(t1m − z) =

m 

(−1)j λj (z)tm−j .

j=0

The following differentiation rules are valid (see [Sh8], lemma 9.1): ∆(f g) =

m 

  tr tρr (∂/∂z)f · ρ
m−r (∂/∂z)g ,

(2.169)

r=0

ρr (∂/∂z)det(z)α = cr (α)det(z)α−1 ρ
m−r (z), ρ
r (∂/∂z)det(z)α = cr (α)det(z)α−1 ρm−r (z), ρr (∂/∂z)etr(uz) = ρr (u)etr(uz) , m    ∆ etr(uz) det(z)α = etr(uz) det(z)α−1 cm−r (α)λr (uz),

(2.170) (2.171) (2.172) (2.173)

r=0

for α ∈ C with cr (α) =

r−1


(α + j/2) =

j=0

Γr (α + (r + 1)/2) Γr (α + (r − 1)/2)

(2.174)

Now we generalized formula (2.169) to the application of the Maass operator ∆ on n functions f1 , · · · , fn .

2.5 Explicit description of differential operators

93

Lemma 2.19. Let f1 , · · · , fn : Hm → C be n C ∞ -functions, then we get ∆(f1 · · · fn ) = (2.175)   L L t (−1)ε(σK,L ) ρt1 (∂/∂z)Kt1 (f1 ) · · · ρtn (∂/∂z)Kttnn (fn ), 1

|T |=m K,L

where T goes through the multi-indices T = {0 ≤ t1 , · · · , tn ≤ m} which satisfy |T | = t1 + · · · + tn = m. The second sum is made on the multiindices K and L of multi-indices strictely ordonnered Ktj and Ltj such that card(Ktj ) = card(Ltj ) and ∪nj=1 Ktj = ∪nj=1 Ltj = m = {0, · · · , m} are two partitions of m. The permutation σK,L of m is determined by: for any h ∈ m, there is tj such that h = hi ∈ Ktj with i the rank of h in Ktj with respect to the natural order, then σK,L (hi ) = h i where h i is the element of rank i of Ltj . We note ε(σK,L ), the sign of the permutation σK,L . Proof. We prove this lemma by induction on the number n of functions. • If n = 2, this is the formula (2.169). • We assume that (2.175) is true for (n − 1). By hypothesis we have ∆(f1 , · · · , fn ) = (2.176)   Ltn−1 Lt1 ε(σK,L ) (−1) ρt1 (∂z )Kt (f1 ) · · · ρtn−1 (∂z )Kt (fn−1 fn ). 1

|T |=m K,L

n−1

Lt

But the operator ρtn−1 (∂z )Ktn−1 coincides with the operator ∆tn−1 (z1 ) for the n−1 variable z1 = (zij )i∈Ktn−1 ,j∈Ltn−1 . Thus formula (2.176) give Lt

ρtn−1 (∂z )Ktn−1 (fn−1 fn ) = n−1





Lt

(−1)ε(σK,L ) ρtn (∂z )Ktn (fn−1 )·

tn−1 +tn =tn−1 Lt ·ρtn (∂z )Ktn (fn ),

n

n

(see also Section 3.3 of Chapter 3 where a more detailed proof of the Lemma 2.19 is given); compare with the proof of Lemma 3.6)

3 Arithmetical differential operators on nearly holomorphic Siegel modular forms

3.1 Description of the Shimura differential operators Let p be a prime number (we often assume p ≥ 5). The purpose of this chapter is to describe the action of certain arithmetical differential operators on algebraically defined nearly holomorphic Siegel modular forms (in the sense of Shimura, see [Sh9] over an O-algebra A where O is the ring of integers in a finite extension K of Qp . Often we simply assume that A = Cp . We view such modular forms as certain formal expansions. The important property of these arithmetical differential operators is their commutation with Hecke operators (under an appropriate normalization, see Theorem 3.13). Their action produces natural families of distributions on a profinite group Y = lim Yi with values in these A-modules of nearly holomorphic Siegel mod←− ular forms inside a formal q-expansion ring like A[[q B ]][Ri,j ] where B is an additive semi-group, q B = {q ξ |ξ ∈ B} the corresponding formally written multiplicative semi-group (for example B = Bm = {ξ = tξ ∈ Mm (Q)|ξ ≥ 0, ξ half-integral} is the semi-group, important for the theory of Siegel modular forms), and the nearly holomorphic parameters (Ri,j ) = R correspond to the matrix R = (4πIm(z))−1 in the Siegel modular case. There exist nice applications of this theory to construction of certain new p-adic families of modular forms (families of Siegel-Eisenstein series, families of theta-series wth spherical polynomials. . .). Main sources of this theory are: Serre’s theory of p-adic modular forms as certain formal q-expansions (J.-P. Serre, Formes modulaires et fonctions zˆeta p-adiques, LNM 350 (1973) 191268) [Se2]. Shimura’s theory of arithmeticity for nearly holomorphic forms (Shimura, G., Arithmeticity in the theory of automorphic forms. AMS, 2000, [Sh9]).

M. Courtieu and A. Panchishkin: LNM 1471, pp. 95–125, 2004. c Springer-Verlag Berlin Heidelberg 2004


96

3 Arithmetical differential operators

Hida’s theory of p-adic modular forms and p-adic Hecke algebras (H. Hida, Elementary theory of L-functions and Eisenstein series, Cambridge University Press, 1993 [Hi5]). Construction of p-adic Siegel-Eisenstein series by the author, appeared in Israel Journal of Mathematics in 2000, [Pa8]. Nerarly holomorphic Siegel modular forms provide additional arithmetical parameters. Moreover they admit an algebraic interpretation in terms of sections of certain de Rham cohomology sheaves on Siegel modular varieties (see [Ha3]), and a typical example of the arithmetical differential operators is given by the Gauss-Manin connection. On the other hand, differential operators preserving rational structures naturally come from the action of universal envelopping algebras of corresponding Lie algebras on automorphic forms. subsection Arithmetical nearly holomorphic Siegel modular forms They were studied by G.Shimura [Sh9] and descriptions:  they admitξtwo different a(ξ, Ri,j )q ∈ Q[[q Bm ]][Ri,j ] such that • as formal power series g = ξ∈Bm

for all R = (Ri,j )i,j=1,·...·,m = (4πIm(z))−1 = (2πi(¯ z − z))−1 and z ∈ Hm in the Siegel upper half plane of degree m the series converges to a C∞ -Siegel modular form of a given weight k and character ψ; • as certain C∞ -Siegel modular forms g taking values in Q at all CMpoints (up to a factor independnt of a concrete form) and satisfying certain reciprocity laws at these points. The first description motivates our choice of arithmetical variables: q ξ , (r) R = (Ri,j ) = (4πIm(z))−1 = 2πi(¯ z − z)−1 . We describe the action of δk in terms of these arithmetical variables. Consider the differential operator ∆m (the Maass differential operator) of degree m, acting on complex C ∞ -functions on Hm , defined by the equality: ∆m = det(∂˜ij ),

∂˜ij = 2−1 − (1 + δij )∂/∂ij

For any f ∈ C∞ (Hm , C), one defines also the operators  Mk f (z) = det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) , δk f (z) =

 m 1 − det(y)−1 Mk f (z). 4π

Due to H. Maass [Maa], for an integer k and a Dirichlet character ψ modulo N , the differential operator δk acts on C∞ -Siegel modular forms, δk : ˜ m (N, ψ) → M ˜ m (N, ψ). One defines then the Shimura differential operator M k k+2 as the composition (r)

˜ m (N, ψ) → M ˜ m (N, ψ). δk = δk+2r−2 ◦ · · · ◦ δk : M k k+2r In order to describe explicitely the action of δk on Fourier expansions consider, for an integer r ≥ 0 and for a complex number β, the polynomial

3.1 Description of the Shimura differential operators

97

! Rm (z; r, β) = (−1)mr etr(z) det(z)r+β ∆rm e−tr(z) det(z)−β . According to its definition the degree of the polynomial Rm (z; r, β) is equal to mr and the term of the highest degree coincides with det(z)r . We have also that for β ∈ Q the polynomial Rm (z; r, β) has rational coefficients. Notice also that for m = 1 one has r   r r−t R1 (z, r, β) = z β(β + 1) · · · (β + t − 1). t t=0 M.Courtieu has established in his PhD thesis the following more explicit expression for the function Rm (z; r, β) for arbitrary m (see [Cour], Chapter 3 and Theorem 3.12): Rm (z; r, β) =

r   r det(z)r−t t t=0



RL (β)λl1 (z) · . . . · λlt (z),

|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 +· · ·+lt ≤ mt−t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt−|L|) and with coefficients in the ring Z[1/2]. The polynomials λi (z) are defined by det(t1m + z) =

m 

λi (z)tm−i

i=0

(λi (z) is the sum of all diagonal minors of size i × i of the matrix z). Let f ∈ Mm k (N,  ψ) be a Siegel modular form whose Fourier expansion is given by f (z) = ξ∈Bm c(ξ)em (ξz). Let us consider the natural representation (0 ≤ r ≤ m) ρr : GLm (C) −→ GL(∧r Cm ) z −→ ρr (z) of the group GLm (C) on the vector space Λr Cm with respect to the basis {ei1 ∧ · · · ∧ eir : i1 < · · · < ir }.   m Thus ρr (z) is a matrix of size m r × r composed of the subdeterminants of z of degree r. Put ρ
r (z) = det(z)ρm−r (tz)−1

(r = 0, 1, · · · , m),

in other words, ρ
m−r (z) is the matrix representing the action of z on Λm−r Cm with respect to the basis dual to the above basis of Λr Cm . Then the representations ρr and ρ
r turn out to be polynomial representations so that for each z ∈ Mm (C) the linear operators ρr (z), ρ
r (z) are well defined.

98

3 Arithmetical differential operators

We prove in this chapter (Theorem 3.14) then that the nearly holomorphic Siegel modular form (r) δk f (z) is given by the following formal power series expansion r    r (r) det(ξ)r−t c(ξ) δk f (z) = t t=0 ξ∈Bm



=

    RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )q ξ

|L|≤mt−t

 ξ∈Bm



=

r   r c(ξ) det(ξ)r−t t t=0

c(ξ)Q(R, ξ)q ∈ Q(f )[[q ξ



RL (κ − k − r)QL (R, ξ)q ξ

|L|≤mt−t

Bm

]][Ri,j ]i,j=1,··· ,m ,

ξ∈Bm

where Q(f ) is the subfield of C generated by the Fourier coefficients of f , L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m as above, r    r Q(R, ξ) = Q(R, ξ; k, r) = RL (κ − k − r)QL (R, ξ)), det(ξ)r−t t t=0 |L|≤mt−t

and

    QL (R, ξ) = tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ).

For r = 1 this gives: δk f (z) =



c(ξ)

m 

ξ∈Bm

  (−1)m−l cm−l (k + 1 − κ)tr tρm−l (R) · ρ
l (ξ) q ξ ,

l=0

where cm−l (k + 1 − κ) =

m−l−1


(k + 1 − κ + j/2) =

j=0

Γm−l (k + 1/2) Γm−l (k − 1)/2)

(l ≤ m − 1),

and in the elliptic modular case m = 1, κ = 1, the only possibilities are t = 0, 1 when c0 (k) = 1, c1 (k) = k, and one obtains again the classical formula     k  δk ( an q n ) = nan q n − an q n = nan q n − kR an q n 4πy n≥0

n≥1

in which R=

1 , 4πy

 n≥1

n≥0

 nan q n =

1 ∂  2πi ∂z

 n≥0

n≥1

 an q n  = q

n≥0

 d  dq



 an q n  .

n≥0

Note that the operators δk act on nearly holomorphic forms and they are very different from operators studied in [Ch-Eh] and [Eh-Ib] although some polynomials which come up in their action on Fourier expansions have similar properties.

3.2 Nearly holomorphic Siegel modular forms 3.2.1 Algebraic nearly holomorphic Siegel modular forms Let us define algebraic nearly holomorphic Siegel modular forms over an O-algebra A where O is the ring of integers in a finite extension K of Qp . Recall that we fix throughout the book two embeddings i∞ : Q → C,

i p : Q → Cp ,

and we often view the field Q as a subfield of both C and Cp (the complex and the p–adic numbers) via these embeddings, omitting the symbols i∞ and ip . We assume that A is a closed subring of Cp containing ip (Qp ). Recall some notations and definitions concerning Siegel modular forms. Let A be a commutative ring with identity, then Mr,s (A) denote the set of all (r × s)-matrices with coefficients in A. For z ∈ Mr (C) put er (z) = e(tr(z)) with e(u) = exp(2iπu) for u ∈ C. We denote by tz ∈ Mr,s (A) the matrix, which is transpose to z ∈ Mr,s (A), and write ξ[η] for tηξη. For an invertible matrix ξ we put ξ
= tξ −1 . If ξ is a hermitian matrix then we write ξ ≥ 0 or ξ > 0 according as ξ is non negative or positive definite. Let Hm denote the Siegel upper half plane on the degree m, Hm = {z ∈ Mm (C) |

t

z = z = x + iy, y > 0},

so that Hm is a complex analytic variety whose dimension is denoted by m = m(m + 1)/2. Let the symbol Am denote the lattice of all half integral symetric matrices in the vector space V = Vm = {y ∈ Mm (R) | ty = y}. This lattice is dual to the latice L = Mm (Z) ∩ V with respect to the pairing given by (u, v) −→ em (uv). Let G = GSpm be the algebraic subgroup of GLm defined by GA = {γ ∈ GLm (A) |

γJm γ = ν(γ)Jm , ν(γ) ∈ A× },

t

for any commutative ring A, where  0m −1m Jm = . 1m 0m The elements of GA are characterized by the conditions bta − atb = dtc − ctd = 0m , dta − ctb = 1m . The multiplier ν defines a homomorphism ν : GA → A× so that ν(γ)2m = det(γ)2 and ker(ν) is denoted by Spm (A). We also put + + G∞ = GR , G+ ∞ = {γ ∈ G∞ | ν(γ) > 0}, GQ = G∞ ∩ GQ .

The group G+ ∞ acts transitively on the upper half plane Hm by the rule

100

3 Arithmetical differential operators

  ab z −→ γ(z) = (az + b)(cz + d)−1 γ = , z ∈ H ∈ G+ m ∞ cd so that the scalar matrices acts trivially, and Hm can be identified with a homogeneous space of the group Spm (R). Let Km denote the stabilizer of the point i1m ∈ Hm in the group Spm (R), Km = {γ ∈ Spm (R) | γ(i1m ) = i1m }, then there is a bijection Spm (R)/Km  Hm and Km = Spm (R) ∩ SO2m . The group Km is a maximal compact subgroup of the Lie group Spm (R) and it can be identified  with the group U(m) of all unitary m × m-matrices via the ab map γ = −→ a + ib. cd For a function f : Hm −→ C of the form  f= c(ξ)em (ξz) ξ∈Am

and for an integer k we use the notation of Petersson for the action of γ ∈ GSpm on f : (f |k γ)(z) = det(γ)k−κ det(cz + d)−k f (γ(z)) (where

κ = (m + 1)/2),

we put also for N ∈ N: f |U (N )(z) =

 ξ∈Am

f |V (N )(z) =



c(N ξ)em (ξz) =



 tu=u∈M

m (Z)

f |k mod N

1m u 0m N 1m 

c(ξ)em (N ξz) = f (N z) = N

m(κ−k)

ξ∈Am

f |k



N 1m 0m 0m 1m

√ f |W (N )(z) = det( N z)−k f (−(N z)−1 ) = N m(κ−k/2) f |k W (N ),  0m −1m W (N ) = . N 1m 0m There are the following relations for these operators: (f |W (N ))|W (N ) = (−1)mk f, f |W (N N ) = N mk/2 f |W (N )V (N ). We adopt also the notations  c(ξ)em (ξz), Y = {y ∈ Mm (R) | fρ =

t

y = y > 0},

ξ∈Am

Bm = {h ∈ Am | h ≥ 0}, Cm = {h ∈ Am | h > 0}.   The bloc notation g = ac db ∈ Spm (Z) defines the congruence subgroup

, ,

3.2 Nearly holomorphic Siegel modular forms

Γ0m (N ) =

101

"  # ab g= ∈ Spm (Z)|c ≡ 0 mod N cd

of the Siegel modular group Γ m = Spm (Z), and its normal subgroup # "  ab  m Γ1 (N ) = ∈ Spm (Z)det(d) ≡ 1 mod N, c ≡ 0 mod N ⊂ Γ0m (N ). cd A classical Siegel modular form f ∈ Mk (N, ψ) of weight k and character ψ for Γ0m (N ) is a holomorphic function f : Hm → C such that for every ab γ = c d ∈ Γ0m (N ) one has f (γ(z)) = ψ(det d)det(cz + d)k f (z) (for m > 1 the regularity at ∞ is automatically satisfied by Koecher). The Fourier expansion of such f uses the semi-group Bm :  f (z) = a(ξ)q ξ ∈ C[[q Bm ]] ξ∈Bm

where the symbols q ξ = exp(2πitr(ξz)) =

m
i=1

ξii qii




−1 qij ij ⊂ C[[q11 , . . . , qmm ]][qij , qij ]i,j=1,··· ,m 2ξ

i<j

√ are used (with qij = exp(2π( −1zi,j ))); they form a multiplicative semi-group so that one may consider f as a formal q-expansion and one can introduce ˜ k (N, ψ) Siegel modular forms over A as certain elements of A[[q Bm ]]. Let M denote the complex vector space of C∞ -Siegel modular forms of weight k and character ψ for Γ0m (N ). For a non negative integer r and variables q, R = (Ri,j )i,j=1,...,m consider the ring A[[q Bm ]][Ri,j ] (over the complex numbers this notation corresponds to q ξ = exp(2πitr(ξz)), R = (4πIm(z))−1 ). Consider the following A-submodules  '   Pr (A) = g = a(ξ, Ri,j )q ξ ∈ A[[q Bm ]][Ri,j ] ξ∈Bm

( a(ξ, Ri,j ) ∈ A[Ri,j ], degR a(ξ, Ri,j ) ≤ r ⊂ A[[q Bm ]][Ri,j ] Consider also for any natural number a the A-modules ( '  −1 Qr,a = g = a(ξ, Ri,j )q a ξ , ξ∈Bm

and put Qr =

3 a≥1

Qr,a .

102

3 Arithmetical differential operators

Definition 3.1. Let Mm (N, ψ, A) denote the A-submodule of A[[q Bm ]][Ri,j ]  r,k generated by all F = a(ξ, R)q ξ such that all a(ξ, R) ∈ A[Ri,j ]i,j=1,...,m ξ∈Bm ≤r

are polynomials of degree

over A ⊂ Cp which belong to

ip (Q)[Ri,j ]i,j=1,...,m and F =

 ξ∈Bm

  i−1 (a(ξ, R))  p

R=(4πY )−1

m q ξ ∈ Mm r,k (N ), ψ, Q) ⊂ Mr,k (N ), ψ, C)

is a C∞ -complex function on Hm satisfying the following two conditions  ab ∀γ = ∈ Γ0m (N ) F |k γ = ψ(det(d))F (3.1) cd ∀γ ∈ Spm (Z) F |k γ ∈ Qr

(3.2)

(a complex nearly holomorphic modular form of type r and weight k over Q in the sense of Shimura, the second condition is necessary only for m = 1). Put Mm r,k (A) =

)

v Mm r,k (N p , ψ; A)

(3.3)

v≥1

(an A-module of nearly holomorphic Siegel modular forms of wight k, Dirichlet character ψ mod N , type r and level N pv (v ≥ 0) with coefficients in A; we view ψ mod N ). 3.2.2 Formal expansions of nearly holomorphic forms. We may view a given cusp form  a(ξ; f )q ξ ∈ A[[q Bm ]], f= ξ∈Bm Bm ]][Ri,j ] in the formal q-expansion ring as an element of Mm r,k (A) ⊂ A[[q

A[[q Bm ]][Ri,j ]). 3.2.3 Action of the U -operator. Define the action of the operator U = U (p) on  g= a(ξ, n)Rn q ξ ∈ A[[q Bm ]][Ri,j ], ξ∈Bm ,n≥0

3.2 Nearly holomorphic Siegel modular forms

by



g|U (p) =

103

a(pξ, n)(pR)n q ξ

ξ∈Bm ,n≥0

(over the complex numbers R = (4π)−1 Im(z)−1 and this notation corresponds to   1m u f |U (p)(z) = f |k 0m p1m t u=u∈Mm (Z) mod p



p−κm

tu=u∈M

m (Z)

f ((z + u)/p), mod p

Im ((z + u)/p) = Im(z/p) = Im(z)/p = (4π)−1 (pR)−1  Let f0 = a(ξ, n; f0 )Rn q ξ ∈ A[[q Bm ]][Rij ] denote an eigenfunction ξ∈Bm ,n≥0

of U with the eigenvalue α ∈ A, (U (f0 ) = αf0 ) so that for any fixed ξ = pv · ξ0 we have pv|n| a(pv ξ0 , n, ; f0 ) = αv a(ξ0 , n; f0 ) ∈ A,  with |n| = ij nij , (pR)vn = pv|n| Rvn . An element α ∈ A is called a U - characteristic value on M = Mm r,k (A) attached to a cusp form if there exists a cusp form f0 which is an eigenfunction of U with the eigenvalue α ∈ A (U (f0 ) = αf0 ). Definition 3.2. Let α ∈ A. a) The α-proper submodule M(α) ⊂ M of the operator U is the following A-submodule M(α) = M(α) (A) = Ker(αI − U ) ⊂ M = M(A). b) The α-characteristic (primary) submodule of Mα = Mα (A) ⊂ M of the operator U is the maximal A-submodule on which U − αI is nilpotent. ) Ker(U − αI)r ⊂ M(A). Mα (A) = r≥1

c) For each fixed level N pv put M(α) (N pv ) = M(α) (N pv , ψ; A) = M(α) ∩ M(N pv ), Mα (N pv ) = Mα (N pv , ψ; A) = Mα ∩ M(N pv ). Proposition 3.3. Suppose that α ∈ A× is an invertible element of the algebra A then a) the A-linear map U v takes M(N pv+1 ) to M(N p); b) assume that Mα (N pv ; A) is a locally free A-module of finite rank then for all v ≥ 1 the A-linear map U is invertible on Mα (N pv ) ⊂ M(N pv ); c) assume that Mα (N pv ; A) is a locally free A-module of finite rank then for all v ≥ 1 Mα (N pv ; A) = Mα (N p; A) is independent of v;

104

3 Arithmetical differential operators

Proof The statement a) is verified using a known relation of U v with the trace operator on modular forms of weight k:  pv+1 = g|k γ, g|TrN Np γ∈Γ0 (N pv+1 )\Γ0 (N p)

then

v+1

vm(k−m−1)

p 2 g|U v = p g|WN pv+1 TrN WN p , Np  0 −1 is the standard involution. where WM = M 0 v+1 p ˜ m (N pv+1 , ψ) by acts on the space M This operator TrN k Np   1m 0m N pv+1 g|k TrN p (g) = N pu 1m t v u=u∈Mm (Z)

(mod p )

One can write down the action of this operator on Fourier expansions using the operators U (pv ) acting by  m(m+1) F |U (pv )(z) = (pv )− 2 F ((z + u)/pv ). tu=u∈M

m (Z)

(mod pv )

One uses the following matrix identity:  1m 0m N pu 1m    −1m 0m 1m −u 0m −1m v+1 −1 = (−N p ) = , 0 m pv 1 m N pv+1 1m 0m N p1m 0m which implies the relation: v+1

p TrN Np

(g) = p−

vm(k−m−1) 2

g|W (N pv+1 )U (pv )W (N p).

In order to prove b) notice that the operator U acts on the A-module of finite rank Mα (N pv ; A) ⊂ M(N pv ; A) and its determinant is in A× hence U is invertible on Mα (N pv ; A). The statement c) then directly follows from a) and b). Definition 3.4. Suppose that α ∈ A× is an invertible element of the algebra A. a) Define the α-characteristic projection πα,v : M(N pv ) → Mα (N pv )

(v ≥ 1)

as the canonical projector to the α-characteristic submodule ) Ker(U − αI)r Mα (N pv ) = of the U -operator with the kernel

4

r≥1

r≥1

Im(U − αI)r ;

3.2 Nearly holomorphic Siegel modular forms

105

Proposition 3.5. a) For each fixed level N pv the following diagram is commutativ M(N pv+1 ) −→ Mα (N pv+1 ) πα,v+1    v v U 6 6 U M(N p) −→ Mα (N p)

(3.4)

πα (g) = U −v [πα,1 U v (g)]

(3.5)

πα,1

b) For any g ∈ M put

(note that πα (g) is well defined if v is sufficiently large so that g ∈ M(N pv+1 ), because of the commutative diagram (3.4)).

3.3 Algebraic differential operators of Maass and Shimura 3.3.1 Differential operators on Hm . Let us consider the differential operator ∆m (the Maass differential operator) of degree m, acting on complex C ∞ -functions on Hm , defined by the equality: ∆m = det(∂˜ij ),

∂˜ij = 2−1 − (1 + δij )∂/∂ij

(3.6)

For any f ∈ C∞ (Hm , C), one defines also the operators  Mk f (z) = det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) ,  m 1 δk f (z) = − det(y)−1 Mk f (z). 4π Due to H. Maass [Maa], for an integer k and a Dirichlet character ψ modulo N , the differential operator δk acts on C∞ -Siegel modular forms, δk : ˜ m (N, ψ). One defines then the Shimura differential operator ˜ m (N, ψ) → M M k k+2 as the composition (r)

˜ m (N, ψ) → M ˜ m (N, ψ). δk = δk+2r−2 ◦ · · · ◦ δk : M k k+2r In order to describe explicitely the action of δk on Fourier expansions consider, for an integer r ≥ 0 and for a complex number β, the polynomial ! (3.7) Rm (z; r, β) = (−1)mr etr(z) det(z)r+β ∆rm e−tr(z) det(z)−β , with z ∈ V ⊗ C, where the exponentiation is well defined by det(y)β = exp (β log[det(y)]) , for det(y) > 0, y ∈ Y ⊗ C. According to definition (3.7) the degree of the polynomial Rm (z; r, β) is equal to mr and the term of the highest degree coincides with det(z)n . We have also that for β ∈ Q the polynomial Rm (z; r, β) has rational coefficients. Notice also that for m = 1 one has r   r R1 (z, r, β) = β(β + 1) · . . . · (β + t − 1)z r−t . t t=0 M.Courtieu has established in his PhD thesis in 2000 the following more explicit expression for the function Rm (z; r, β) for arbitrary m (see [Cour], Chapter 3)

3.3 Algebraic differential operators of Maass and Shimura r   r Rm (z; r, β) = det(z)r−t t t=0



107

RL (β)λl1 (z) · . . . · λlt (z), (3.8)

|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 +· · ·+lt ≤ mt−t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt−|L|) and with coefficients in the ring Z[1/2]. The polynomials λi (z) are defined by det(t1m + z) =

m 

λi (z)tm−i ,

(3.9)

i=0

In other words λi (z) is the sum of all diagonal minors of size i×i of the matrix z. 3.3.2 The polynomial Rm (z; r, β). We now describe explicitely the polynomial Rm (z; r, β) defined by (3.7). We show that this is a polynomial invariant for the action of conjugation on the space of symetric matrices y ∈ V = {ξ ∈ Mm (R) | tξ = ξ}. Let us consider the natural representation (0 ≤ r ≤ m) ρr : GLm (C) −→ GL(∧r Cm ) z −→ ρr (z) of the group GLm (C) on the vector space Λr Cm with respect to the basis {ei1 ∧ · · · ∧ eir : i1 < · · · < ir }.   m Thus ρr (z) is a matrix of size m r × r composed of the subdeterminants of z of degree r. Put ρ
r (z) = det(z)ρm−r (tz)−1

(r = 0, 1, · · · , m),

in other words, ρ
m−r (z) is the matrix representing the action of z on Λm−r Cm with respect to the basis dual to the above basis of Λr Cm . Then the representations ρr and ρ
r turn out to be polynomial representations so that for each z ∈ Mm (C) the linear operators ρr (z), ρ
r (z) are well defined. We consider ˜ ˜ the differential operators ρr (∂/∂z) and ρ
r (∂/∂z) which associate to each C valued function on Hm a certain Mt (C) valued function on Hm with t = m r . In particular, the Maass differential operator is given by ˜ ˜ ˜ ∆ = ρm (∂/∂z) = ρ
m (∂/∂z) = det(2−1 (1 + δij )∂/∂z ij ). We define the invariant polynomials λr (z) = tr(ρr (z)). They are the coefficients of the characteristic polynomial of z

108

3 Arithmetical differential operators

det(t1m − z) =

m 

(−1)j λj (z)tm−j .

j=0

The following differentiation rules are valid (see [Sh8], lemma 9.1):   ˜ ˜ tr tρr (∂/∂z)f · ρ
m−r (∂/∂z)g ,

(3.10)

α ˜ ρr (∂/∂z)det(z) = cr (α)det(z)α−1 ρ
m−r (z),

(3.11)

α ˜ ρ
r (∂/∂z)det(z) = cr (α)det(z)α−1 ρm−r (z),

(3.12)

∆(f g) =

m  r=0

tr(uz) ˜ ρr (∂/∂z)e = ρr (u)etr(uz) , m    ∆ etr(uz) det(z)α = etr(uz) det(z)α−1 cm−t (α)λt (uz),

(3.13) (3.14)

t=0

for α ∈ C with c0 (α) = 1,

cr (α) =

r−1


(α + j/2) =

j=0

Γr (α + (r + 1)/2) (r ≥ 1). Γr (α + (r − 1)/2)

(3.15)

One can extend the formula (3.10) to the case of the action of ∆ on n functions f1 , · · · , fn : Lemma 3.6. Let f1 , · · · , fn : Hm → C be n C∞ -functions, then we get ∆(f1 · · · fn ) =   Lt1 Ltn ˜ ˜ (−1)ε(σK,L ) ρt1 (∂/∂z) Kt (f1 ) · · · ρtn (∂/∂z)Ktn (fn ),

|T |=m K,L

(3.16)

1

where T runs over the sets of multiplets T = {0 ≤ t1 , · · · , tn ≤ m} which satisfy |T | = t1 + · · · + tn = m. The second sum is extended over the sets of all multi-indices K and L consisting of strictely ordered subsets of multi-indices Ktj and Ltj such that card(Ktj ) = card(Ltj ) and ∪nj=1 Ktj = ∪nj=1 Ltj = m = {0, · · · , m} are two partitions of m. The permutation σK,L of m is determined by the following rule: for any h ∈ m, take tj such that h = hi ∈ Ktj where i the rank of h in Ktj (with respect to the natural order), then σK,L (hi ) = h i where h i is the element of rank i of Ltj , and denote by ε(σK,L ) the sign of the permutation σK,L . Proof of Lemma 3.6. One proves this lemma by induction on the number n of functions. • If n = 2, this is the formula (3.10) • We assume that (3.16) is true for (n − 1). By hypothesis we have

3.3 Algebraic differential operators of Maass and Shimura

109

∆(f1 , · · · , fn ) = (3.17)   Ltn−1 Lt1 ε(σK,L ) (−1) ρt1 (∂z )Kt (f1 ) · · · ρtn−1 (∂˜z )Kt (fn−1 fn ). 1

|T |=m K,L

n−1

Lt

But the operator ρtn−1 (∂˜z )Ktn−1 coincides with the operator ∆tn−1 (z1 ) (with n−1 respect to the variable z1 = (zij )i∈Ktn−1 ,j∈Ltn−1 ). Thus formula (3.10) gives Lt (3.18) ρtn−1 (∂˜z )Ktn−1 (fn−1 fn ) n−1  L  Lt  t = (−1)ε(σK,L ) ρtn−1 (∂˜z )K n−1 (fn−1 ) · · · ρtn (∂˜z )Ktn (fn ),  t n−1

tn−1 +tn =tn−1

n

with card(Ktn−1 ) = card(Ltn−1 ) = t n−1 , card(Ktn ) = card(Ltn ) = t n and one has Ktn−1 ∪ Ktn = Ktn−1 , Ltn−1 ∪ Ltn = Ltn−1 and (3.17) follows. The following more delicate differentiation rule holds: Lemma 3.7. Let u = tu be a real symmetric m × m-matrix, α a complex number, r, j two integers such that 0 ≤ r ≤ j ≤ m and z ∈ Hm then one has the following equality: ˜ (λj (z)α ) = cr (α)λj (z)α−1 µ
(3.19) ρr (∂) r,j−r (z).     m Here µ
r,j−r (z) is a matrix of size m r × r with the coefficients given by µ
r,j−r (z)L K =

 T ⊃K,L

T \L , T \K

ρ
j−r (z)

the sum is being taken over all the multi-indices T : 1 ≤ t1 < · · · < tj ≤ m containing the multiindices K : 1 ≤ k1 < · · · < kr ≤ m et L : 1 ≤ l1 < ¯ is used for the the complementary multi-index · · · < lr ≤ m. The notation L in m = {0, · · · , m} of the multi-index L. Recall that the matrix coefficient ρ
m−r (z)L K is equal (up to a sign) to the determinant of the complementary matrix in z for the determinant ρr (z)L K more precisely one has ¯

ε(σK,L ) ρm−r (z)L ρ
m−r (z)L ¯, K = (−1) K

where σK,L is the permutation of Sm given by the following rule: for every ¯ σK,L (k¯i ) = ¯li with l¯i ∈ L ¯ and i ki ∈ K, σK,L (ki ) = li and for every k¯i ∈ K, ¯ ¯ denoting the rank of the index ki (respectively li , ki , li ) with respect to the ¯ L) ¯ (recall that ε(σK,L ) denotes the natural order in K (respectively L, K, signature of the permutation σK,L ). 3.3.3 Proof of Lemma 3.7  is deduced from (3.13) as follows. Notice that λj (z) = T ρj (z)TT is the sum ˜ L , one gets non zero of all diagonal minors of size j in z. If one applies ρr (∂) K

110

3 Arithmetical differential operators

terms only for those diagonal minors ρj (z)TT , when the multi-index T contains the multi-indices K et L. The corresponding minor is then the determinant of a submatrix z, so that the computation is similar to that in (3.13) but for a matrix of a lower size. This gives us then ˜ L (λj (z)α ) = cr (α)λj (z)α−1 ρr (∂) K



T \L , T \K

ρ
j−r (z)

T ⊃K,L

and (3.18) follows. 3.3.4 Action of the Shimura differential operator on formal Fourier expansions We now apply the above results on Rm (z; r, β) in order to obtain an explicit dscripton of the action of the Shimura differential operator on formal Fourier expansions of nearly holomorphic Siegel modular forms. First of all we recall that the Shimura differential operator δk in the Siegel modular case is given as the twisted Maass operator ∆ [Maa], Chapter 19. Definition 3.8. For any f ∈ C∞ (Hm , C), define the operators  Mk f (z) = det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) , δk f (z) =

 m 1 − det(y)−1 Mk f (z). 4π

(3.20) (3.21)

Due to H. Maass [Maa], for an integer k and a Dirichlet character ψ modulo N , the Shimura differential operator acts on C∞ -Siegel modular forms δk : ˜ m (N, ψ) → M ˜ m (N, ψ). One defines then δ (r) = δk+2r−2 ◦ · · · ◦ δk . M k k+2 k  (r) We show now that the Fourier expansion of δk f , where f = Bm c(ξ)em (ξz) ∈ Mm k (N, ψ), is given by the formula  (r) c(ξ)det(4πy)−r Rm (4πξy; r, β)em (ξz). δk f = ξ∈Bm (r)

In order to describe the action of δk f one establishes first the following Lemma 3.9. Let α be a complex number, r a positive integer, u a real symmetric matrix and z ∈ Hm then one has the following relation      det(z)−r ∆ det(z)r+1 ∆r etr(uz) det(z)α = ∆r+1 etr(uz) det(z)α+1 .

3.3 Algebraic differential operators of Maass and Shimura

111

Proof of Lemma 3.9 Let us use the formula (3.14) giving   ∆ etr(uz) det(z)α   in order to compare it with the action of the operator det(z)−r ∆ det(z)r+1 · one has  det(z)−r ∆ etr(uz) det(z)α+1 · m 

·

 cm−l1 (α) · . . . · cm−lr (α − r + 1)Pl1 ···lr (u, z) =

l1 ,··· ,lr =0

etr(uz) det(z)α−r · ·

m 

cm−l1 (α) · . . . · cm−lr (α − r + 1)cm−lr+1 (α + 1)Pl1 ···lr lr+1 (u, z).

l1 ,··· ,lr+1 =0

The polynomials Pl1 ···lr are defined recursively in such a way that   ∆r etr(uz) det(z)α = e

tr(uz)

α−r

det(z)

m 

cm−l1 (α) · . . . · cm−lr (α − r + 1)Pl1 ···lr (u, z),

l1 ,··· ,lr =0

hence one has the following equality satisfied by Pl1 ···lr (u, z) : Pl1 ···lr (u, z) = 



L

L

L

˜ i2 (Pl ···l (u, z)), (−1)ε(σK,L ) ρi1 (u)Kii1 ρ
m−lr (z)Kllrr ρi2 (∂) 1 r−1 Ki 1

2

i1 +i2 =m−lr K,L

(a polynomial in matrix variables u and z independent of α and symmetric in lj ). The independence of Pl1 ···lr (u, z) on α results from the fact that the terms depending on α factorise out using (3.11). The symmetry in lj results from the theorem of Schwarz on the independence of the mixed partial derivatives of C∞ -functions on the order of differentiaion. Here the choice of l1 , · · · , lr corresponds to another order of derivation applied to the factor det(z)
. Replacing u by −1m and α by −β, and using then the formula defining R(z; r, β), one finds that this is indeed a polynomial as the factors etr(z) and e−tr(z) cancel, as well as the factors det(z)β and det(z)−β . One obtains then: R(z; r, β) =

m  l1 ,··· ,lr =0

cm−l1 (−β) · . . . · cm−lr (−β − r + 1)Pl1 ···lr (−1m , z).

112

3 Arithmetical differential operators

In particular the polynomials Pl1 ···lr are symmetric in lj , and one can replace the index of summation l1 = lr+1 , l2 = l1 , · · · , lr+1 = lr so that one ob tr(uz)  r+1 α+1 e . This gives exactly the announced tains the expression ∆ det(z) (r)

property. In order to compute the Fourier expansion of δk f , one proceeds by induction on r. One computes the action of the Shimura differential operator on the test functions em (ξz), more precisely one shows that δk (em (ξz)) = det(4πy)−r Rm (4πξy; r, κ − k − r)em (ξz).  c(ξ)em (ξz), one obtains then by linearity the In fact if one takes f (z) = (r)

ξ∈Bm (r) δk f ,

resulting Fourier expansion of  (r) c(ξ)det(4πy)−r Rm (4πξy; r, κ − k − r)em (ξz), δk f (z) = ξ∈Bm

It remains to prove the following Lemma 3.10. The action of the Shimura differential operator on the test functions em (ξz) (where ξ ∈ Bm and z ∈ Hm ) is given by the formula δk (em (ξz)) = det(4πy)−r Rm (4πξy; r, κ − k − r)em (ξz). (r)

(3.22)

Proof of Lemma 3.10 : One proves this lemma by induction on r : • If r = 0, it is obvious. • Suppose next that it holds for some r then if one applies the Shimura operator δk+2r one obtains (r+1)

δk

(em (ξz)) = (−1)m det(4πy)−1 det(z − z¯)κ−k−2r ·  ·∆ det(z − z¯)k+2r+1−κ det(4πy)−r em (ξz) · · Rm (4πξy; r, κ − k − r)] .

Notice that 4πξy = −2iπξ(z − z¯). On the other hand z¯ is aniholomorphic hence it behaves as a constant under the action of ∆, and one can write, by putting u = −2iπξ, β = κ − k − r:   z); r, β) = (−1)mr etr(u(z−¯z)) det(z−¯ z )r+β ∆r e−tr(u(z−¯z)) det(z − z¯)−β . Rm (u(z−¯ One replaces this expression in the first formula pulling out the factor em (ξ z¯) = etr(−u¯z) which is an antiholomorphic function. One obtains then (r+1)

δk

(em (ξz)) = det(4πy)−(r+1) em (ξz)(−1)m(r+1) etr(u(z−¯z )) det(z − z¯)β ·

3.3 Algebraic differential operators of Maass and Shimura

113

 ! ·∆ det(z − z¯)r+1 ∆r e−tr(u(z−¯z)) det(z − z¯)−β . Let us now apply lemma 3.9 which gives (r+1)

δk

(em (ξz)) = det(4πy)−(r+1) em (ξz)(−1)m(r+1) etr(u(z−¯z)) ·   ·det(z − z¯)r+β ∆r+1 e−tr(u(z−¯z)) det(z − z¯)−β

= det(4πy)−(r+1) Rm (4πξy; r + 1, κ − k − (r + 1))em (ξz), finishing the proof by induction. Now we can write down the action of the Shimura operator on the Fourier expansions of Siegel modular forms in terms of the polynomial Rm (z; r, β) as follows: Proposition 3.11. Let f ∈ Mm k (N,  ψ) be a Siegel modular form whose Fourier expansion is given by f (z) = ξ∈Bm c(ξ)em (ξz), then  (r) c(ξ)det(4πy)−r Rm (4πξy; r, κ − k − r)em (ξz). δk f (z) = ξ∈Bm

Let us now use the explicite expressions (3.8) of M.Courtieu for the func(r) tion Rm (z; r, β) for arbitrary m in order to describe the action of δk on Fourier expansions. Theorem 3.12. a) Let f ∈ Mm k (N,  ψ) be a Siegel modular form whose Fourier expansion is given by f (z) = ξ∈Bm c(ξ)em (ξz), then r    r (r) δk f (z) = c(ξ)det(4πy)−r (3.23) det(2πiξ(¯ z − z))r−t t t=0 

ξ∈Bm

RL (κ − k − r)λl1 (2πiξ(¯ z − z)) · . . . · λlt (2πiξ(¯ z − z))em (ξz)

|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 + · · · + lt ≤ mt − t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt − |L|) and with coefficients in thering Z[1/2] (recall that m−i . the polynomials λi (z) are defined by det(t1m + z) = m i=0 λi (z)t b) In particular, if r = 1, δk f (z) = (−1)m



c(ξ)det(2πi(¯ z − z))−1

ξ∈Bm m 

cm−t (k + 1 − κ)λt (−2πiξ(¯ z − z))em (ξz)

t=0

where c0 (α) = 1,

cr (α) =

r−1


(α + j/2) =

j=0

Γr (α + (r + 1)/2) . Γr (α + (r − 1)/2)

(3.24)

114

3 Arithmetical differential operators

Proof of Theorem 3.12 Let us write by (3.22)  (r) δk f (z) = c(ξ)det(4πy)−r Rm (4πξy; r, κ − k − r)em (ξz) ξ∈Bm

and substitute in it r   r Rm (z; r, β) = det(z)r−t t t=0



RL (β)λl1 (z) · . . . · λlt (z),

|L|≤mt−t

with 4πξy = 2πiξ(¯ z − z), β = κ − k − r: (r) δk f (z)

=



−r

c(ξ)det(4πy)

t=0

ξ∈Bm



r   r

t

det(2πiξ(¯ z − z))r−t

RL (κ − k − r)λl1 (2πiξ(¯ z − z)) · . . . · λlt (2πiξ(¯ z − z))em (ξz)

|L|≤mt−t

Note also that by (3.14) m    ∆ etr(uz) det(z)α = etr(uz) det(z)α−1 cm−t (α)λt (uz), t=0

for α ∈ C with c0 (α) = 1, cr (α) =

r−1


(α + j/2) =

j=0

Γr (α + (r + 1)/2) Γr (α + (r − 1)/2)

and by definition (3.7) ! Rm (z; n, β) = (−1)mn etr(z) det(z)n+β ∆nm e−tr(z) det(z)−β , hence

! Rm (z; 1, β) = (−1)m etr(z) det(z)1+β ∆m e−tr(z) det(z)−β = (−1)m etr(z) det(z)1+β e−tr(z) det(z)−β−1

m 

cm−t (−β)λt (−z)

t=0

giving δk f (z) =

 ξ∈Bm

c(ξ)det(4πy)−1 Rm (4πξy; 1, κ − k − r)em (ξz) =

3.3 Algebraic differential operators of Maass and Shimura



115

c(ξ)det(2πi(¯ z − z))−1 (−1)m etr(2πiξ(¯z −z)) det(2πiξ(¯ z − z))1+β ·

ξ∈Bm

·e−tr(2πiξ(¯z−z)) det(2πiξ(¯ z −z))−β−1

m 

cm−t (k+1−κ)λt (−2πiξ(¯ z −z))em (ξz) =

t=0

(−1)m



c(ξ)det(2πi(¯ z − z))−1

m 

cm−t (k + 1 − κ)λt (−2πiξ(¯ z − z))em (ξz).

t=0

ξ∈Bm

3.3.5 Commutation of the Shimura operator with Hecke operators We conclude this section with a general result on commutation of the Shimura differential operator with Hecke operators (up to a normalization). Recall some basic fact about the action of Hecke algebras (see [An5]). Let q be a prime, q  | N , ∆ = ∆m q (N ) = "   ab  −1 ∩ GL (Z[q ]) γ= ∈ G+  ν(γ)± ∈ Z[q −1 ], c ≡ 0m 2m Q cd

# (mod N )

m be a subgroup in G+ Q containing Γ = Γ0 (N ). The Hecke algebra

L = Lm q (N ) = DQ (Γ, ∆) over Q is then defined as a Q-linear space generated by the double cosets (g) = (Γ gΓ ), g ∈ ∆ of the semigroup ∆ with respect to the subgroup Γ , for which multiplication is defined by the standard rule (see [An5], [Sh1]). We recall the description of the structure of L = Lm q (N ), (q
N ): for each j, 1 ≤ j ≤ m let us denote by wj an automorphism of the algebra ±1 ±1 Q[x±1 0 , x1 , · · · , xm ]

defined on its generators by the rule: x0 → x0 xj , xj → x−1 j , xi → xi

(1 ≤ i ≤ m, i = j).

Then the automorphisms wj and the permutation group Σm of the variables xi (1 ≤ i ≤ m) generate together the Weyl group W = Wm , and there is the Satake isomorphism (see (2.1.3) and 2.1.7): ∼

±1 ±1 Wm . Sat : L → Q[x±1 0 , x1 , · · · , xm ]

Recall that we define Hecke operators  using the notation of Petersson and ab Andrianov (2.22): for any g = ∈∆ cd (f |k,ψ g)(z) = (detg)k−(m+1)/2 ψ(deta)det(cz + d)−k f (g(z)).

116

3 Arithmetical differential operators

Theorem 3.13. Let  f (z) = c(ξ)em (ξz) ∈ Mm k (N, ψ)

(ξ ∈ Bm ),

ξ∈Bm

and let

(r)

˜ m (N, ψ) → M ˜ m (N, ψ) δk = δk+2r−2 ◦ · · · ◦ δk : M k k+2r be the Shimura differential operator defined by (3.21). Consider for a fixed prime q an arbitrary element 

t(X)

X=

vi (Γ0 gi ) ∈ L0 (Γ0 = Γ0m , L0 = Lm 0,q )

i=1

of the extended Hecke algebra L0 with  ν ∗ q i di bi gi = ∈ ∆m 0,q , 0 di and assume X homogeneous (ν1 = ν2 = · · · νt = ν). Then (r) (r) (δk f )|k+2r X = q νrm δk (f |k X).

(3.25)

In particular (r)

(r)

(r)

(r)

m m (δk f )|k+2r Π+ (q) = q rm δk (f |k Π+ (q)),

(3.26)

m m (δk f )|k+2r Π− (q) = q rm δk (f |k Π− (q))

Proof of Theorem 3.13 : We use the known property δk (f |k γ) = (δk f )|k+2 γ for all γ ∈ Sp2m (R) and k ∈ N. Taking into account (δk f )|k+2 gi = det(gi )δk (f |k gi ), one immediately obtains (3.25) from the relation detgi = q νi rm taking into account the homogeneousity of X.

3.4 Arithmetical variables of nearly holomorphic Siegel modular forms and differerntial operators. 3.4.1 Arithmetical nearly holomorphic Siegel modular forms where studied by G.Shimura [Sh9] two different descriptions: and they admit a(ξ, Ri,j )q ξ ∈ Q[[q Bm ]][Ri,j ] such that for • as formal power series g = ξ∈Bm

all R = (4πIm(z))−1 and z ∈ Hm the series converges to a C∞ -Siegel modular form of a given weight k and character ψ; • as certain C∞ -Siegel modular forms g taking values in Q at all CMpoints (up to a factor independent of a concrete form) and satisfying certain reciprocity laws at these points. The first description motivates our choice of arithmetical variables: q ξ , z − z)−1 . Our task in this Section is to R = (Ri,j ) = (4πIm(z))−1 = 2πi(¯ (r) expliciteltly describe the action of δk in terms of these arithmetical variables. Theorem 3.14. Let f ∈ Mm k (N, ψ) be a Siegel modular form whose Fourier expansion is given by f (z) = ξ∈Bm c(ξ)q ξ , then (r)

δk f (z) =



c(ξ)

ξ∈Bm



    RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )q ξ

|L|≤mt−t

=

r   r det(ξ)r−t t t=0

 ξ∈Bm

r   r c(ξ) det(ξ)r−t t t=0



=



RL (κ − k − r)QL (R, ξ)q ξ

|L|≤mt−t

c(ξ)Q(R, ξ)q ξ ∈ Q(f )[[q Bm ]][Ri,j ]i,j=1,··· ,m ,

ξ∈Bm

where Q(f ) is the subfield of C generated by the Fourier coefficients of f , L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m as above, r    r Q(R, ξ) = Q(R, ξ; k, r) = RL (κ − k − r)QL (R, ξ)), det(ξ)r−t t t=0 |L|≤mt−t

    QL (R, ξ) = tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ). If r = 1 one has δk f (z) =

 ξ∈Bm

where

c(ξ)

m  l=0

  (−1)m−l cm−l (k + 1 − κ)tr tρm−l (R) · ρ
l (ξ) q ξ

118

3 Arithmetical differential operators

cm−l (k + 1 − κ) =

m−l−1


(k + 1 − κ + j/2) =

j=0

Γm−l (k + 1/2) Γm−l (k − 1)/2)

(l ≤ m − 1).

Remark 3.15. In the elliptic modular case m = 1, κ = 1, the only possibilities are t = 0, 1 when c0 (k) = 1, c1 (k) = k, and one obtains again the classical formula δk (





an q n ) =

n≥0

nan q n −

n≥1

  k  an q n = nan q n − kR an q n 4πy n≥0

n≥1

n≥0

in which

       ∂ 1 1 d  , R= nan q n = an q n  = q  an q n  . 4πy 2πi ∂z dq n≥1

n≥0

n≥0

Proof of Theorem 3.14 Let us change the variable by putting R equal (4πy)−1 = −2πiz − z¯)−1 and let us use the equality (3.14) with u = −2πiξ and the polynomial (3.7) with n = r, β = κ − k − r  m  1 det(y)−1 det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) δk f (z) = − 4π  = det(2πi)−m det(z − z¯)κ−k−1 ∆ det(z − z¯)k−κ+1 f (z) . δk (em (ξz)) = det(4πy)−r em (ξz)R(4πξy; r, κ − k − r) (r)

z − z); r, κ − k − r) = det(2πi(¯ z − z))−r em (ξz)R(2πiξ(¯ We have by Theorem 3.11 (r) δk f (z)

=



−r

c(ξ)det(4πy)

ξ∈Bm



r   r det(2πiξ(¯ z − z))r−t · t t=0

(3.27)

RL (κ −k −r)λl1 (2πiξ(¯ z −z))·. . .·λlt (2πiξ(¯ z −z))em (ξz)

|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 + · · · + lt ≤ mt − t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt − |L|) and with coefficients in the  ring Z[1/2] (recall that m the polynomials λj (z) are defined by det(t1m + z) = j=0 λj (z)tm−j . We now express all terms of type λj (2πiξ(¯ z − z)) in terms of our arithz − z)) = λj (ξR−1 )). metical variable R: 2πi(¯ z − z) = R−1 and λj (2πiξ(¯ Viewing symmetric matrices R and ξ as independent variables let us write the definition

3.4 Arithmetical variables of nearly holomorphic forms

119

det(tR + ξ) = det(R)det(t1m + ξR−1 ) = det(R)det(t1m + ξR−1 ) =

m 

det(R)λj (ξR−1 )tm−j = tm det(ξ)det(Rξ −1 + t−1 1m ) =

j=0

tm det(ξ)

m 

λj (Rξ −1 )t−m+j =

(3.28)

j=0 m 

det(ξ)λj (Rξ −1 )tj

j=0

Let us show that the coefficients of the polynomial det(tR + ξ) admit the following very explicite expression det(tR + ξ) =

m 

  tr tr tρr (R) · ρ
m−r (ξ))

(3.29)

r=0

In order to prove 3.29 we use the differentiation rules (3.10 ) and (3.13 ) ∆(f g) =

m 

  ˜ ˜ tr tρr (∂/∂z)f · ρ
m−r (∂/∂z)g ,

r=0 tr(uz) ˜ ρr (∂/∂z)e = ρr (u)etr(uz) ,

applied to the functions f = etr(tRz) and g = etr(ξz) : ∆(etr(tR+ξ)z ) = ∆(etr(tRz) · etr(ξz) ) = m 

(3.30)

  tr(tRz) tr(ξz) ˜ ˜ = tr tρr (∂/∂z)e · ρ
m−r (∂/∂z)e

r=0 m 

  tr tρr (tRz)etr(tRz) · ρ
m−r (ξz)etr(ξz) ) =

r=0 m 

  tr tρr (tRz) · ρ
m−r (ξz) etr((tR+ξ)z) ,

r=0

on the other hand tr((tR+ξ)z) ˜ ∆(e(tr(tR+ξ)z) ) = ρm (∂/∂z)e =

ρm ((tR + ξ)z)e

tr((tR+ξ)z)

= det((tR + ξ)z)e

(3.31) tr((tR+ξ)z)

,

and the substutution of z = 1m to (3.28) and (3.31) implies (3.29). It follows from (3.29) and (3.28) that   det(R)λj (ξR−1 ) = tr tρm−j (R) · ρ
j (ξ)

(3.32)

120

3 Arithmetical differential operators

Let us substitute (3.32) to (3.27) writng 4πy = 2πi(¯ z − z) = R−1 ,   λj (2πiξ(¯ z − z)) = λj (ξR−1 ) = det(R)−1 tr tρm−j (R) · ρ
j (ξ)

(r) δk f (z)



=

c(ξ)det(R

−1 −r

ξ∈Bm



)

r   r det(ξR−1 )r−t t t=0

RL (κ − k − r)λl1 (ξR−1 ) · . . . · λlt (ξR−1 )em (ξz)

|L|≤mt−t

=

 ξ∈Bm



r   r c(ξ) det(ξ)r−t det(R)t t t=0

RL (κ − k − r)λl1 (ξR−1 ) · . . . · λlt (ξR−1 )em (ξz)

|L|≤mt−t

r   r = c(ξ) (3.33) det(ξ)r−t t t=0 ξ∈Bm      RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )em (ξz)



|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 + · · · + lt ≤ mt − t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt − |L|) and with coefficients in the ring Z[1/2] and the first statement of Theorem 3.14 follows. In order to prove the second statement it suffices to specialize these formulae to the case of r = 1 using (3.24). 3.4.2 Action on Fourier expansion (r)

Note from (3.33) that the action of δk on the exponential q ξ = em (ξz) written in variables q ξ , R = (Ri,j ) is given by (r) δk (q ξ )

r   r = det(ξ)r−t t t=0



(3.34)

    RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )q ξ ,

|L|≤mt−t

and one may view ξ also as an independent variable for which the notation ∂ξ = (∂ξ,i,j ), ∂˜ξ = (∂˜ξ,i,j ), ∂˜ξ,i,j = 2−1 (1 + δi,j )∂˜ξ,i,j will be used.

(3.35)

3.4 Arithmetical variables of nearly holomorphic forms

121

3.4.3 Differentiation on monomials Our next task is to write down the formal action (r)

δk (Ria11,j1 · · · Riann,jn q ξ )

(3.36)

(r)

of δk using the differentiaion over the formal parameter ξ of the equality (3.34). Let us show first that Ria11,j1 · · · Riann,jn q ξ = det(R)a1 +···+an etr(2πiξ(−¯z )) · Dξ (etr(2πiξ(¯z −z)) )

(3.37)

for an appropriate differential polynomial on ∂˜ξ,i,j which we now specify. Recall that (Ri,j ) = (2πi(¯ z −z))−1 = ρ1 ((2πi(¯ z −z))−1 ) = det(2πi(¯ z −z))−1 ·tρ
m−1 ((2πi(¯ z −z))) on the other hand t
ρm−1 ((2πi(¯ z−

z)))etr(2πiξ(¯z −z)) = tρ
m−1 ((∂˜ξ ))etr(2πiξ(¯z −z))

hence a1 an Dξ = Dξ,i · · · Dξ,i 1 ,j1 n ,jn

(3.38)

where (Dξ,i,j ) = tρ
m−1 ((∂˜ξ )). (we used the commutation of ∂˜z,i,j with ∂˜ξ,i,j with respect to the independent formal variables z and ξ). Let us compute   (r) (r) δk (Ria11,j1 · · · Riann,jn q ξ ) = δk det(R)a1 +···+an etr(2πiξ(−¯z )) · Dξ (etr(2πiξ(¯z −z)) ) using (3.37), (3.7), Lemma 3.9 and (3.34): r   r (r) (3.39) det(ξ)r−t δk (q ξ ) = t t=0      RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )q ξ , |L|≤mt−t (r)



δk (Ria11,j1 · · · Riann,jn q ξ )

 det(R)a1 +···+an etr(2πiξ¯z ) · Dξ (etr(2πiξ(z−¯z )) )   (r) = etr(2πi(ξ¯z )) δk det(R)a1 +···+an · Dξ (etr(2πiξ(z−¯z )) )    (r) = etr(2πiξ¯z )) Dξ δk det(R)a1 +···+an · etr(2πiξ(z−¯z )) (r)

= δk

and our computation of (3.39) is reduced to computation of     (r) (r) z)−a ·etr(2πi(ξz)) δk det(R)a · etr(2πiξ(z−¯z )) = (−2πi)−am etr(−2πi(ξ¯z )) δk det(z−¯ where a = a1 + · · · + an .

122

3 Arithmetical differential operators

Lemma 3.16. Let a ≥ 0 be a natural number then the action of the Shimura differential operator on the functions det(¯ z − z)−a em (ξz) (where ξ ∈ Bm and z ∈ Hm ) is given by the formula (r)

δk

  det(z − z¯)−a em (ξz)

= (−2πi)−rm det(z − z¯)−r−a em (ξz)Rm (2πiξ(¯ z − z); r, κ − k − r + a)   = (2πi)−rm etr(2πiξ(¯z )) det(z − z¯)κ−k−r ∆r etr(2πiξ(z−¯z )) det(z − z¯)−κ+k+r−a (r)

and δk = δk+2r−2 ◦ · · · ◦ δk , δk f (z) =  m  1 det(y)−1 Mk f (z) = (2πi)−m det(z − z¯)κ−k−1 ∆ det(z − z¯)−κ+k+1 f − 4π  Mk f (z) = det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) , ! Rm (z; r, β) = (−1)mr etr(z) det(z)r+β ∆rm e−tr(z) det(z)−β . Proof of Lemma 3.16 It is very similar to that of Lemma 3.10. One uses induction on r : • If r = 0, it is obviously true. (r+1) (r) • Suppose next that it holds for some r then δk = δk+2r ◦ δk , and if one applies the Shimura operator δk+2r by induction one obtains   (r+1) det(z − z¯)−a em (ξz) = (−2πi)−m det(z − z¯)κ−k−2r−1 (3.40) δk ·∆ det(z − z¯)k+2r+1−κ (2πi)−rm det(z − z¯)−a−r em (ξz) · ! ·Rm (2πiξ(¯ z − z); r, κ − k − r + a) (recall that 4πξy = −2iπξ(z − z¯)). On the other hand z¯ is aniholomorphic hence it behaves as a constant under the action of ∆, and it makes it possible to write, by putting u = −2iπξ, β = κ − k − r + a : Rm (u(z − z¯); r, β) =



(−1)mr etr(u(z−¯z)) det(z − z¯)r+β ∆r e−tr(u(z−¯z)) det(z − z¯)−β



(3.41)

(compare with the proof of 3.10). One substituts (3.41) into (3.40) , then the factor em (ξ z¯) = etr(−u¯z) which is an antiholomorphic function can be pulled out. One obtains then

3.4 Arithmetical variables of nearly holomorphic forms (r+1)



δk

 det(z − z¯)−a em (ξz)

123

(3.42)

= (−2πi)−m det(z − z¯)κ−k−2r−1 · ·∆ det(z − z¯)k+2r+1−κ (2πi)−rm det(z − z¯)−a−r em (ξz) ·  ! ·(−1)mr etr(u(z−¯z)) det(z − z¯)r+β ∆r e−tr(u(z−¯z)) det(z − z¯)−β = (2πi)−(r+1)m (−1)m(r+1) det(z − z¯)κ−k−2r−1 em (ξ z¯)) ·  ! ·∆ det(z − z¯)k+2r+1−κ−a−r+r+β ∆r e−tr(u(z−¯z )) det(z − z¯)−β = (−2πi)−(r+1)m det(z − z¯)κ−k−r−1 em (ξ z¯)) ·  ! det(z − z¯)−r · ∆ det(z − z¯)r+1 ∆r e−tr(u(z−¯z)) det(z − z¯)−β Let us now apply lemma 3.8 which gives  ! det(z − z¯)−r · ∆ det(z − z¯)r+1 ∆r e−tr(u(z−¯z)) det(z − z¯)−β+1   = ∆r+1 e−tr(u(z−¯z)) det(z − z¯)−β+1 and (3.42) transforms to   (r+1) δk det(z − z¯)−a em (ξz) =

(3.43)   (−2πi)−(r+1)m det(z − z¯)κ−k−r−1 em (ξ z¯)∆r+1 e−tr(u(z−¯z)) det(z − z¯)−β+1 = (−2πi)−(r+1)m det(z − z¯)κ−k−(r+1) em (ξ z¯) ·   ·∆r+1 e−tr(u(z−¯z )) det(z − z¯)−κ+k+r+1−a

finishing the proof by induction. Corollary 3.17. Let a ≥ 0 be a natural number then the action of the Shimura differential operator on the functions det(2πi(¯ z −z))−a em (ξz) (where ξ ∈ Bm and z ∈ Hm ) is given by the formula (r) 

δk

 det(R)a q ξ = det(R)r+a Rm (ξR−1 ; r, κ − k − r + a)q ξ a ξ

= det(R) q

r   r t=0



t

det(ξ)r−t

(3.44)

    RL (κ − k − r + a)tr tρm−l1 (R) · ρ
l1 (ξ) · . . . · tr tρm−lt (R) · ρ
lt (ξ)

|L|≤mt−t (r)

where δk = δk+2r−2 ◦ · · · ◦ δk , and the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt − |L|) and with coefficients in the ring Z[1/2].

124

3 Arithmetical differential operators

Proof of Corollary 3.17 Let us substitute into (3.8) 4πξy = 2πiξ(¯ z − z) = ξR−1 as z, β = κ − k − r + a: r    r RL (β)λl1 (z) · . . . · λlt (z) Rm (z; r, β) = det(z)r−t t t=0 |L|≤mt−t

giving Rm (ξR−1 ; r, κ − k − r + a) = r   r det(ξR−1 )r−t t t=0



RL (κ − k − r + a)λl1 (ξR−1 ) · . . . · λlt (ξR−1 ) =

|L|≤mt−t

r   t=0

r det(ξ)r−t det(R)−r t



    RL (κ − k − r + a)tr tρm−l1 (R) · ρ
l1 (ξ) · . . . · tr tρm−lt (R) · ρ
lt (ξ)

|L|≤mt−t

using again (3.32):   λj (ξR−1 ) = det(R)−1 tr tρm−j (R) · ρ
j (ξ) , and (3.44) follows. In a more general case let us apply Lemma 3.31 to (3.39): (r)

(3.45) δk (Ria11,j1 · · · Riann,jn q ξ )    (r) = etr(2πiξ¯z )) Dξ δk det(R)a1 +···+an · etr(2πiξ(z−¯z ))    (r) = etr(2πiξ¯z )) Dξ (−2πi)−am etr(−2πi(ξ¯z )) δk det(z − z¯)−a · etr(2πi(ξz))    (r) = (−2πi)−am Dξ δk det(z − z¯)−a · etr(2πi(ξz))   = (−2πi)−am−rm Dξ det(z − z¯)−r−a em (ξz)Rm (2πiξ(¯ z − z); r, κ − k − r + a) where a = a1 + · · · + an . Let us use again the explicit expression (3.8) for the polynomial Rm (z; r, β) =

r   r det(z)r−t t t=0



RL (β)λl1 (z) · . . . · λlt (z),

|L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 + · · · + lt ≤ mt − t. Let us substitute (3.32) to (3.45) writing 4πy = 2πi(¯ z − z) = R−1 ,

3.4 Arithmetical variables of nearly holomorphic forms

125

  λj (2πiξ(¯ z − z)) = λj (ξR−1 ) = det(R)−1 tr tρm−j (R) · ρ
j (ξ) implying directly: (r)

δk (Ria11,j1 · · · Riann,jn q ξ ) (3.46)   = (−2πi)−am−rm Dξ det(z − z¯)−r−a q ξ Rm (2πiξ(¯ z − z); r, κ − k − r + a) r    r −am−rm −r−a ξ det(2πiξ(¯ z − z))r−t Dξ det(z − z¯) q = (−2πi) t t=0   RL (β)λl1 (2πiξ(¯ z − z)) · . . . · λlt (2πiξ(¯ z − z)) |L|≤mt−t

r   r z − z)) q = Dξ det((2πi)(¯ det(2πiξ(¯ z − z))r−t t t=0   RL (β)λl1 (2πiξ(¯ z − z)) · . . . · λlt (2πiξ(¯ z − z))



−r−a ξ

|L|≤mt−t

r   r = det(R) Dξ q det(ξ)r−t t t=0  t    RL (β)tr ρm−l1 (R) · ρ
l1 (ξ) · . . . · tr tρm−lt (R) · ρ
lt (ξ) . a

|L|≤mt−t



ξ

4 Admissible measures for standard L–functions and nearly holomorphic Siegel modular forms

4.1 Congruences between modular forms and p-adic integration 4.1.1 Integration in nearly holomorphic Siegel modular forms Let p be a prime number (we often assume p ≥ 5). The purpose of this chapter is to give a new conceptual construction of admissible measures (in the sense of Amice-V´elu) attached to a standard Lfunction of a Siegel cusp eigenform. For this purpose we use the theory of p-adic integration in spaces of nearly holomorphic Siegel modular forms (in the sense of Shimura, see [Sh9]) over an O-algebra A where O is the ring of integers in a finite extension K of Qp . Often we simply assume that A = Cp . We view such modular forms as certain formal expansions and we study the action of certain differential operators on Siegel-Eisenstein distributions and on theta distributions. The important property of these arithmetical differential operators is their commutation with Hecke operators (under an appropriate normalization). Their action produces natural families of distributions with values in spaces of nearly holomorphic Siegel modular forms. Denote by M = M(A) an A-module (or simply a Cp -vector space) of nearly holomorphic Siegel modular forms. In the present paper we apply to these distributions the canonical projector πα : M → Mα onto the primary (characteristic) A-submodule associated to a non-zero eigenvalue α ∈ A× of the Frobenius operator U = Π + (p) = Up (the Atkin-Lehner operator in the Siegel modular case; Mα is the maximal A-submodule on which U − αI is nilpotent). This operator acts similar to the trace operator lowering the level of modular forms. On the other hand, U is invertible on Mα if α ∈ A× so that one can glue its action on forms of various levels. In this way one obtains the desired distributions with values in a finite dimensional vector space starting from naturally defined distributions with values in spaces of modular forms (like Siegel-Eisenstein distributions, theta distributions etc.). In order to obtain from them numerically valued distributions interpolating

M. Courtieu and A. Panchishkin: LNM 1471, pp. 127–186, 2004. c Springer-Verlag Berlin Heidelberg 2004


128

4 Admissible measures for standard L–functions

critical values attached to standard L-functions of Siegel modular forms one applies a suitable linear form coming from the Petersson scalar product (using the Andrianov identity of Rankin type for the standard L-function). In previous works [Pa7], [Pa6] a non–Archimedean interpolation of these special values was constructed fixing a non zero Satake p–parameter α0 (p) of the cusp eigenform f . It follows that the normalized critical values D(s, f, χ) of the standard L-function can be explicitely rewritten in terms of certain Cp –valued integrals of admissible p–adic measures (over a profinite group of the type Z× p of p–adic units) provided the character χ is non–trivial. More explicit description of these special values was given by M.Courtieu in his PhD thesis ([Cour], Institut Fourier, 2000) using detailed study of the action of the Shimura differential operator on Siegel modular forms. In the present chapter we give a conceptual explanation of these p– adic properties satisfyied by the special values of the standard L-function D(s, f, χ), where f is a Siegel cusp form of an even degree m and of weight k > 2m + 2, χ is a varying Dirichlet character. We show that these admissible measures can be lifted to arithmetical nearly holomorphic Siegel modular forms studied by G.Shimura [Sh9]. This lifting is given by a universal sequence Φ± s (χ) of distributions with values in arithmetical nearly holomorphic Siegel modular forms (for critical pairs (s, χ), see Proposition 5.4). It would be interesting to extend these lifting results to Siegel cusp eigenforms of odd degree, using the method of B¨ ocherer-Schmidt [B¨ oSch]. 4.1.2 Arithmetical nearly holomorphic Siegel modular forms They admit two different descriptions:  a(ξ, Ri,j )q ξ ∈ Q[[q Bm ]][Ri,j ] such that • as formal power series g = ξ∈Bm

for all R = (Ri,j )i,j=1,··· ,m = (4πIm(z))−1 = (2πi(¯ z − z))−1 and z ∈ Hm in the Siegel upper half plane of degree m the series converges to a C∞ -Siegel modular form of a given weight k and character ψ; • as certain C∞ -Siegel modular forms g taking values in Q at all CM-points (up to a factor independnt of a concrete form). Rcall that throughout the paper we fix embeddings i∞ : Q → C,

i p : Q → Cp ,

and we shall often view algebraic numbers (via these embeddings) as both complex and p–adic numbers. 4.1.3 The group For a fixed positive integer N ∈ N consider the profinite group Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

There is a natural projection yp : Y → Z× p.

4.1 Congruences between modular forms and p-adic integration

129

Definition 4.1. a) For h ∈ N, h ≥ 1 let C h (Y, A) denote the A-module of × locally polynomial functions of degree < h of the variable yp : Y → Z× p → A ; in particular, C 1 (Y, A) ⊂ C loc−const (Y, A) (the A-submodule of locally constant functions). We adopt the notation Φ(U) := Φ(χU ) for the characteristic function χU of an open subset U ⊂ Y . Let also denote C loc−an (Y, A) the A-module of locally analytic functions and C(Y, A) the A-module of continous functions so that C 1 (Y, A) ⊂ C h (Y, A) ⊂ C loc−an (Y, A) ⊂ C(Y, A). b) For a given positive integer h an h-admissible measure on Y with values in M is an A-module homomorphism Φ˜ : C h (Y, A) → M such that for all a ∈ Y and for v → ∞       (yp − ap )j dΦ˜ = o(p−v(j−h) ) for all    a+(N pv )

j = 0, 1, · · · , h − 1,

p,M

where ap = yp (a). We adopt the notation (a)v = a+(N pv ) for both an element of Yv and the corresponding open compact subset of Y . 4.1.4 Canonical projection α : We explain in Section 4.3, how to construct an h-admissible measure Φ C h (Y, A) → M(A) out of a sequence of distributions Φj : C 1 (Y, A) → M with values in an A-module M = M(A) of nearly holomorphic modular forms over A (for all j ∈ N with j ≤ h − 1, where A is an O-algebra, and α ∈ A× is a fixed non-zero eigenvalue of the Frobenius operator U = U (p) = Π + (p) acting on  g= a(ξ, n)Rn q ξ ∈ A[[q Bm ]][Ri,j ], ξ∈Bm ,n≥0

by g|U (p) =



a(pξ, n)(pR)n q ξ

ξ∈Bm ,n≥0

(over the complex numbers R = (4π)−1 Im(z)−1 and this notation corresponds to   u 1 f |k m f |U (p)(z) = = 0m p1m t u=u∈Mm (Z) mod p

130

4 Admissible measures for standard L–functions



p−κm

tu=u∈M

m (Z)

f ((z + u)/p), mod p

Im ((z + u)/p) = Im(z/p) = Im(z)/p = (4π)−1 (pR)−1 (we use the Petersson-Andrianov notation for the action of matrices). Then we consider (for an α ∈ A× ) the canonical projection operator πα : M(A) → M(A)α . We define an A-linear map α : C h (Y, A) → M Φ on local monomials ypj by 

α = πα (Φj ((a)v )) ypj dΦ

(a)v

where Φj : C 1 (Y, A) → M(A) are certain M(A)-valued distributions on Y for all j = 0, 1, · · · , h − 1. 4.1.5 The standard zeta function of a Siegel cusp eigenform In order to describe the main result of the paper let  f= a(ξ)em (ξz) ξ>0

be a Siegel cusp form of the even degree m of weight k on the congruence subgroup " # ab ∈ Spm (Z) | c ≡ 0 mod C Γ0 (C) = cd with a Dirichlet character ψ mod C. Here z belongs to the Siegel upper half plane % & Hm = z ∈ GLm (C)|z = tz, Im(z) > 0 and we adopt the standard notation em (z) = exp(2πitr(z)). Suppose that f is an eigenfunction of the global Hecke algebra (m)

L(m) (C) = ⊗q
C L(q) (C) with the eigenvalue given by a homomorphism λf : L(m) (C) → C, i.e. (m)

f | X = λf (X)f for all X ∈ L(q) (C). Let α0 (q), α1 (q), · · · , αm (q)

4.1 Congruences between modular forms and p-adic integration

131

be the (m + 1)–tuple of the Satake q–parameters of λf , which uniquely determine λf . It is known [An5] that the relation α20 (q)α1 (q) . . . αm (q) = ψ(q)m q m(k−(m+1)/2) holds. Recall that the standard L-function of f with a Dirichlet character χ mod N is defined as the Euler product
D(s, f, χ) = D(q) (s, f, χ), (4.1) q
C

with D(q) (s, f, χ)−1 =  m   χ(q)ψ(q
χ(q)ψ(q)αi (q) χ(q)ψ(q)α−1 i (q) 1− 1− 1− , qs qs qs i=1 (the product is absolutely convergent for Re(s) > 1 + m). Recall the three types of the normalized zeta function: for κ = (m + 1)/2, s ∈ C D∗ (s, f, χ) = (2π)−m(s+k−κ) Γ ((s + δ)/2)

m


Γ (s + k − j)D(s, f, χ),

(4.2)

j=1

D− (s, f, χ) = Γ ((s + δ)/2)−1 D∗ (s, f, χ) = (2π)−m(s+k−κ)

m


(4.3)

Γ (s + k − j)D(s, f, χ),

j=1

D+ (s, f, χ) =

2 iδ Γ (s) cos(π(s − δ)/2) − D (s, f, χ) = (2πi)s

(4.4)

iδ π (1−2s)/2 D∗ (s, f, χ), Γ ((1 + ∆ − s)/2) where δ = 0 or 1 according as ψχ(−1) = (−1)δ with f ∈ Sm k (C, ψ) be an eigenfunction of the global Hecke algebra Lm (C) = ⊗q
C Lm (C). Define q h = [2ordp α0 (p)] + 1, Assume that a(ξ0 ) = 0 for some ξ0 > 0 with det(2ξ0 ) = 1. Moreover, we make the essential assumption (*) that a0 (ξ0 ) = 0

(∗) 

for a Fourier coefficient a0 (ξ0 ) of the U -eigenform f0 (z) = ξ>0 a0 (ξ)em (ξz) defined in Section 4.4 and in §1, (1.47), chapter 2 of [Pa6]. The importance of the above non–vanishing condition was pointed out to us by S. B¨ ocherer.

4 Admissible measures for standard L–functions

132

It turns out that it actually may not be satisfied by certain Eisenstein series (note that we deal here with cusp forms). However it is not known in general, whether there exist cusp forms violating the non–vanishing condition. Main Theorem. Let f ∈ Skm (N, ψ) be a Siegel cusp eigenform of even degree m and of weight k > 2m + 2 satisfying the above non-vanishing condition (*). Define h = [2ordp (α0 (p))] + 1 then for any positive integer c > 1, (c, N p) = 1 there exist two Cp -valued h-admissible measures Dc± on Y such that: (i) For all pairs (s, χ) with s ∈ Z, 0 ≤ s ≤ k − m − 2 and a primitive non-trivial Dirichlet character χ ∈ XStors of conductor Cχ = pv one has 

m(2s+2k−m)/4

•

χxsp dDc+ = Y

·

GCχ (1m , χ)Cχ (iCχ )m2 /2 α0 (Cχ2 )
q|N

 •

G(χψ) ¯

(1 − (χψ) ¯ 2 (c)c−2s−2 )·

¯ 0 (q)q s ) D+ (s + 1, f, χ) (1 − (χψ) ¯ . (1 − (χψ) ¯ 0 (q)q −s−1 ) f, f N m(2−2s+2k−m)/4

χxsp dDc− = Y

s+1 Cχψ ¯

¯ χ GCχ (1m , χ)C (iCχ )m2 /2 α0 (Cχ2 ) ·

¯ 2 (c)c−2s−2 )· (1 − (χ ¯ψ)

D− (−s, f, χ) . f, f N

¯ viewed as elements of the Tate field These integrals are algebraic numbers Q ¯ Cp (via the fixed embedding ip : Q → Cp ). (ii) If ordp (α0 (p)) = 0 (i.e. f is p-ordinary), then the measures Dc± are bounded Cp -analytic functions where  χ(detξ )em (tξξ /Cχ ), GCχ (1m , χ) = ξ  ∈Mm (Z)

(mod Cχ )

denotes the Gauss sum of degree m of χ.

4.2 Algebraic differential operators and Siegel-Eisenstein distributions 4.2.1 Operatots of Maass and Shimura Consider again the differential operator ∆m (the Maass differential operator) of degree m, acting on complex C ∞ -functions on Hm , defined by the equality: ∂˜ij = 2−1 − (1 + δij )∂/∂ij (4.5) ∆m = det(∂˜ij ), For any f ∈ C∞ (Hm , C), one defines also the operators  Mk f (z) = det(z − z¯)κ−k ∆ det(z − z¯)k−κ+1 f (z) ,  m 1 det(y)−1 Mk f (z). δk f (z) = − 4π Due to H. Maass [Maa], for an integer k and a Dirichlet character ψ modulo N , the differential operator δk acts on C∞ -Siegel modular forms, δk : ˜ m (N, ψ). Recall also that the Shimura differential operator ˜ m (N, ψ) → M M k k+2 is the composition (r) δk = δk+2r−2 ◦ · · · ◦ δk . (r)

In order to describe explicitely the action of δk on Fourier expansions consider, for an integer r ≥ 0 and for a complex number β, the polynomial ! Rm (z; r, β) = (−1)mr etr(z) det(z)r+β ∆rm e−tr(z) det(z)−β . (4.6) According to its definition the degree of the polynomial Rm (z; r, β) is equal to mr and the term of the highest degree coincides with det(z)r . We have also that for β ∈ Q the polynomial Rm (z; r, β) has rational coefficients. Notice also that for m = 1 one has r   r r−t R1 (z, r, β) = z β(β + 1) · . . . · (β + t − 1). t t=0 M.Courtieu has established in his PhD thesis the following more explicit expression for the function Rm (z; r, β) for arbitrary m (see [Cour], Chapter 3 and Theorem 3.12: r    r Rm (z; r, β) = RL (β)λl1 (z) · . . . · λlt (z), (4.7) det(z)r−t t t=0 |L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 +· · ·+lt ≤ mt−t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt−|L|) and with coefficients in the ring Z[1/2]. The polynomials λi (z) are defined by det(t1m + z) =

m 

λi (z)tm−i

i=0

(λi (z) is the sum of all diagonal minors of size i × i of the matrix z).

4 Admissible measures for standard L–functions

134

4.2.2 Formulas for Fourier expansions Let f ∈ Mm k (N,  ψ) be a Siegel modular form whose Fourier expansion is given by f (z) = ξ∈Bm c(ξ)em (ξz), then we proved in Chapter 3, (Theorem (r)

3.14) that the nearly holomorphic Siegel modular form δk f (z) is given then by the following formal power series expansion (r) δk f (z)

=

r   r c(ξ) det(ξ)r−t t t=0

 ξ∈Bm



(4.8)

    RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )q ξ =

|L|≤mt−t

 ξ∈Bm

r   r c(ξ) det(ξ)r−t t t=0





RL (κ − k − r)QL (R, ξ)q ξ =

|L|≤mt−t

c(ξ)Q(R, ξ)q ξ ∈ Q(f )[[q Bm ]][Ri,j ]i,j=1,··· ,m ,

ξ∈Bm

where Q(f ) is the subfield of C generated by the Fourier coefficients of f , L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m as above, r    r Q(R, ξ) = Q(R, ξ; k, r) = RL (κ − k − r)QL (R, ξ)), det(ξ)r−t t t=0 |L|≤mt−t

    QL (R, ξ) = tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ). For r = 1 this gives: δk f (z) =



c(ξ)

ξ∈Bm

m 

  (−1)m−l cm−l (k + 1 − κ)tr tρm−l (R) · ρ
l (ξ) q ξ ,

l=0

where cm−l (k + 1 − κ) =

m−l−1


(k + 1 − κ + j/2) =

j=0

Γm−l (k + 1/2) Γm−l (k − 1)/2)

(l ≤ m − 1),

4.2.3 Siegel-Eisenstein series. (See also Section 2.2 of Chapter 2). Let us recall the definition of the SiegelEisenstein series Ek (z, s, χ) which come up in the integral representation of Rankin type for the standard L-function. We introduce some additional notation. Let us fix a positive integer m then a pair of integral matrices c, d ∈ Mm (Z) is called primitive if any matrix with rational coefficients γ ∈ Mm (Q) such that γc, γd ∈ Mm (Z) is integral: γ ∈ Mm (Z). The pair (c, d)

4.2 Algebraic differential operators and Siegel-Eisenstein distributions

135

is said symmetric if c·td = d·tc. Two primitive symmetric pairs (c1 , d1 ) and (c2 , d2 ) are said equivalent if (c1 , d1 ) = (uc2 , ud2 ) for a matrix u ∈ SLm (Z). Let Dm denote the set of equivalence classes of symmetric couples of coprime matrices. Now let k, N be positive integers, s a complex number and χ a Dirichlet character modulo N such that χ(−1) = (−1)k . For z ∈ Hm the Siegel-Eisenstein series is defined by  χ(det(d))det(cz + d)−k−|2s| , (4.9) E(z, s; k, χ, N ) = E(z, s) = det(y)s where the summation is taken over all (c, d) ∈ Dm with the condition c ≡ 0 (mod N ) and we adopt the convenient notation by Deligne and Ribet [De-Ri]: z −k−|2s| = z −k |z|−2s for z ∈ C∗ def

The series (4.2) is absolutely convergent for k + 2Re(s) > m + 1 and it admits a meromorphic analytic continuation  over the whole complex s-plane. Put ab j(α, z) = det(cz + d) for α = and z ∈ Hm , then it follows from a cd known description of Dm that  χ(det(d)j(α, z)−k−|2s| , (4.10) E(z, s; k, χ, N ) = det(y)s 

α∈P ∩Γ \Γ

ab , and P denotes the subgroup of P ⊂ G∞+ , where Γ = Γ0m (N ), α = cd consisting of elements α with the condition c = 0. The series (4.10) define ˜ k (N, χ) ¯ and if s = 0 and k > m + 1, C∞ -modular forms Ek (z, s, χ) ∈ M m ¯ This series has a Fourier these forms are holomorphic: Ek (z, χ) ∈ Mm k (N, χ). expansion of the type  bk (ξ)em (ξz). Ek (z, χ) = ξ∈Bm

Following the works of H. Maass (see [Maa]], chap. 19), the series Ek (z, s, χ) are related to each other for different values of k and s via the Shimura differential operator: Ek+2 (z, s − 1, χ) = (−4π)m

Γm (k + s) δk Ek (z, s, χ). Γm (k + s + 1)

(4.11)

One obtains by iteration of (4.11) r times Ek+2r (z, s − r, χ) = (−4π)mr

Γm (k + s) (r) δ Ek (z, s, χ) Γm (k + s + r) k

(r)

where δk = δk ◦· · ·◦δk+2r−2 . It follows from (4.7) that the series Ek+2r (z, −r, χ) is a nearly holomorphic Siegel modular form with the following Fourier expansion:  Γm (k) Ek+2r (z, −r, χ) = det(−y)−r bk (ξ)R(4πξy; r, κ−k −r)em (ξz). Γm (k + r) ξ∈Bm

136

4 Admissible measures for standard L–functions

4.2.4 Normaized Siegel-Eisenstein series In order to obtain arithmetical nearly holomorphic forms whose Fourier coefficients depend nicely on the parameter s = −r ≤ 0 we introduce normalized Siegel-Eisenstein series starting from the series (4.9). Such series where studied in detail by Shimura (voir [Sh8]. Keeping the same notation as above define (4.12) Ek
(z, s, χ) = Ek (−z −1 , s, χ)det(z)−k . (the involuted Siegel-Eisenstein series): Ek
(z, s, χ) = Ek (z, s, χ)|k Jm . The series (4.12) admit the following Fourier expansion  b(ξ, y, s, χ)em (ξz), (4.13) Ek
(z, s, χ) = ξ∈N −1 Am

where the coefficients b(ξ, y, s, χ) are explicitely given in [Sh8] (see also [Pa6] (chap. 2, prop. 3.4.)) in terms of the confluent hypergeometric function. We introduce here other three types of normalized Siegel-Eisenstein series in order to give a precise statement on their holomorphy properties with respect to the variable s, the properties of positivity of matrices ξ by which their Fourier coefficients are indexed, and also algebraic properties of these Fourier coefficients: G∗ (z, s) = G∗ (z, s; k, χ, N ) = N

m(k+s)

Γ(k, s)LN (k + 2s, χ)




(4.14)

[m/2]

LN (2k + 4s − 2i, χ2 )E ∗ (N z, s),

i=1

with E ∗ (N z, s) = E(−(N z)−1 , s)det(N z)−k = N −km/2 E|W (N ),

imk 2−m(k+1) π −m(s+k) ×

"

(4.15)

Γ(k, s) = Γm (k + s)Γ (s + (k − Γm (k + s),

and LN (s, χ) =




m 2

+ µ)/2), if m is even; otherwise.

(1 − χ(q)q −s )−1

q |N

is the Dirichlet series associated to the character χ and µ = ε(m/2 + k), with ε(r) ∈ {0, 1} such that ε(r) ≡ r (mod 2) for any integer r ∈ N. If the character χ2 is non trivial, the function G
k (z, s, χ) is an entire function of the complex variable s (see [Fe], Th. 9.1, p. 49, and [Pa6], Chap. 3, Th. 3.6).

4.2 Algebraic differential operators and Siegel-Eisenstein distributions

137

Proposition 4.2. Let N > 1 a positive integer such that 2|N then (a) If χ2 = 1 the series G
k (z, s, χ) is an entire function of the variable s. (b) If χ2 is trivial then (b1 ) if either 2k ≥ m and m odd or 2k ≥ m and m is even, but (m/2) + k is odd (i.e. µ = ε((m/2) + k) = 1 ), then the function G∗ (z, s) is an entire function of the variable s; (b2 ) if 2k ≥ m and both numbers m and (m/2) + k are even (i.e. µ = ε(m) = 0) then the function G∗ (z, s) is an entire function of the variable s with the exclusion of a possible simple pole at the point s = (m + 2 − 2k)/4; (b3 ) if m > 2k ≥ 0 then the function G∗ (z, s) is an entire function of the variable s with possible exclusion of simple poles at those points s for which 2s is an integer and [(m − 2k + 3)/2] ≤ 2s ≤ (m + 1 − 2k)/2 ; (b4 ) if k = 0 then the function G∗ (z, s) has a simple pole at the point s = (m + 1)/2 iff χ is trivial, and in this case we have that the function Ress=(m+1)/2 G∗ (z, s; 0, 1, N ) of the variable z is a non-zero constant. Proof of Proposition 4.2 (See [Fe], Th. 9.1, p. 49). If m is odd then we put G+ (z, s, χ) = G− (z, s, χ) = G∗ (z, s, χ). If m is even then we define G− (z, s) = Γ ((k + 2s − (m/2) + µ)/2)−1 G∗ (z, s), G+ (z, s) =

iµ π κ−k−2s G∗ (z, s) = Γ ((1 − k − 2s + (m/2) + µ)/2)

(4.16) (4.17)

2iµ Γ (k + 2s − (m/2)) cos(π(k + 2s − (m/2) − µ)/2) − G (z, s), (2π)k+2s−(m/2) where µ = ε((m/2) + k), κ = (m + 1)/2. The Fourier expansions of G
k (z, s, χ) and G± k (z, s, χ) are given by G
k (z, s, χ) =



b
(ξ, y, s, χ)em (ξz),

ξ∈Am

G± k (z, s, χ) =



b± (ξ, y, s, χ)em (ξz).

ξ∈Am

The Fourier coefficients b (ξ, y, s, χ), b± (ξ, y, s, χ) are explicitely given in [Pa6], (Chap. 2, §3.5) and in Theorem 4.3 below, and we show that for all crtical pairs (s, χ) these series define arithmetical nearly holomorphic Siegel modular forms.


138

4 Admissible measures for standard L–functions

Next we define an additional quadratic Dirichlet character χξ depending on ξ ∈ Am and defined only for even rank r. Namely, for ξ = 0 let χξ = χ0 be trivial; for ξ = 0 we know that for some matrix u ∈ GLm (Q)  ξ1 0 t uξu = (4.18) with det ξ1 = 0, 0 0 then let χξ denote the quadratic character attached to the quadratic field 0 Q( detξ1 )/Q (this definition does not depend on choice of a matrix u). Theorem 4.3. Let m is even, 2k > m. Then: (a) If 2s is an integer, s ≤ 0, k + 2s ≥ 1 + (m/2) there is the following Fourier expansion  b+ (ξ, y, s, χ)em (ξz), (4.19) G+ (z, s) = Am ξ>0

where for s > (m + 2 − 2k)/4 in (4.19) non-zero terms only occur for positive definite ξ > 0, and for all s from (a) with ξ > 0, ξ ∈ Am the following identity holds b+ (ξ, y, s, χ) = W + (y, ξ, s)L+ N (k + 2s − (m/2), χχξ )M (ξ, χ, k + 2s), with L+ (s, χ) =

2iδ Γ (s) cos(π(s − δ)/2) LN (s, χ) (2π)s

is the normalized Dirichlet L-function, δ = 0 or 1 according to χ(−1) = (−1)δ , the factor M (ξ, χ, k + 2s) is a finite Euler product
Mq (ξ, χ(q)q −k−2s ) M (ξ, χ, k + 2s) = q∈P (ξ)

with Mq (ξ, t) ∈ Z[t], extended over primes q in the set P (ξ) of prime divisors of the number N and of all elementary divisors of the matrix ξ,and the character χξ is given by (4.18). W + (y, ξ, s) = 2−mκ detξ k+2s−κ det(4πy)s R(4πξy; −s; κ − k + s), provided s is an integer, where R(y; n, β) is defined by (4.2.2), and b+ (ξ, y, s, χ) = 0 otherwise (if s ∈ Z). (b)If 2s is an integer with k + 2s ≤ m/2, k + s ≥ κ then there is the following Fourier expansion  b− (ξ, y, s, χ)em (ξz), (4.20) G− (z, s) = Am ξ≥0

4.2 Algebraic differential operators and Siegel-Eisenstein distributions

139

and for all s from (b) with ξ > 0, ξ ∈ Am the following identity holds b− (ξ, y, s, χ) = W − (y, ξ, s)L− N (k + 2s − (m/2), χχξ )M (ξ, χ, k + 2s), where L− (s, χ) = LN (s, χ), W − (y, ξ, s) = 2−mκ det(4πy)κ−k−s R(4πξy; k+s−κ, −s), provided s + k − κ is an integer, and b− (ξ, y, s, χ) = 0 otherwise. Proof of Theorem 4.3 (See in [Pa6], Chap. 2, Th. 3.8). Let us now write down the action of the Shimura differential operator on these normailzed Siegel-Eisenstein series. Lemma 4.4. The action of the Shimura differential operator on Ek
(z, s, χ) is given by: δk Ek
(z, s, χ) = δk Ek
(N z, s, χ) =

Γm (k + s + 1)
1 Ek+2 (z, s − 1, χ), m Γm (k + s) (−4π)

(4.21)

N m Γm (k + s + 1)
Ek+2 (N z, s − 1, χ). (−4π)m Γm (k + s)

(4.22)

Proof of Lemma 4.4 We use the known property δk (f |k γ) = (δk f )|k+2 γ for all γ ∈ Sp2m (R) and k ∈ N. Taking into account Ek
(z, s, χ) = Ek (z, s, χ)|k Jm , one immediately obtains (4.21) from the relation (4.11) satisfied by Ek (z, s, χ). In order to prove (4.22), one changes the variable z = N z. One obtains then δk (z)Ek
(N z, s, χ) = det(N 1m )δk (z )Ek
(z , s, χ), and (4.22) follows. Corollary 4.5. One has the following relations : δk (G
k (z, s, χ)) = G
k+2 (z, s − 1, χ), + δk (G+ k (z, s, χ)) = Gk+2 (z, s − 1, χ), − δk (G− k (z, s, χ)) = Gk+2 (z, s − 1, χ).

(4.23)

140

4 Admissible measures for standard L–functions

Proof of Corollary 4.5 One shows all the three equalities in the same way, so let us consider only the first one. One applies the Shimura differential operator to G
k (z, s, χ). The only factors in the formula (4.14) for G
k (z, s, χ) which are not invariant with respect to the change of variables k → k + 2, s → s − 1, are the following: N m(s+k) imk 2−m(k+1) π −m(s+k) Γm (k + s). These factors come up essentially in the factor Γ(k, s) (see (4.15)). One applies then the Shimura differential operator to G
k (z, s, χ) using (4.22). A simplification of the powers of N , π, 2, and i, and of the factor Γm (k + s) gives δk (G
k (z, s, χ)) = G
k+2 (z, s − 1, χ). 4.2.5 Distributions with values in nearly holomorphic Siegel modular forms. Consider a profinite group Y = lim Yv for example ← v

Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

There is a natural projection yp : Y → Z× p . Let A be a commutative ring, and let M be an A-module. Let Cloc−const (Y, A) denote the A-module of locally constant functions. A distribution on Y with values in M is an A-module homomorphism Φ : Cloc−const (Y, A) → M. We adopt the notations Φ(U) := Φ(χU ) for the characteristic function χU of an open subset U ⊂ Y , and (a)v = a + (N pv ) for both an element of Yv and the corresponding open compact subset of Y . Recall an algebraic version of nearly holomorphic Siegel modular forms over an O-algebra A where O is the ring of integers in a finite extension K of Qp . We assume that A is equipped with an augmentation eA : A → O, and fix embeddings i∞ : Q → C, ip : Q → Cp .  Let us fix another posivite integer N and let denote by g = a(ξ, R)q ξ ∈ ξ∈Bm

Mm r,k (A) an element of m Mm r,k (A) = Mr,k (A; N ) =

)

v Mm r,k (N p ; A)

v≥1

(A-module of nearly holomorphic Siegel modular forms of type r and level N pv (v ≥ 0) with coefficients in A). Let M = Mm r,k (A) and let Φ : loc−const C (Y, A) → M be a distribution on Y with values in M . Assume that

4.2 Algebraic differential operators and Siegel-Eisenstein distributions

141

A contains (ϕ(N pv ))−1 and all the values of Dirichlet characters χ mod N pv viewed as homomorphisms Y → A× then obviously  χ(a)Φ((a)v ), (4.24) Φ(χ) = a mod N pv

1 ϕ(N pv )

Φ((a)v ) =



χ(a)−1 Φ((a)v ).

χ mod N pv

Proposition 4.6 (Normaized Siegel-Eisenstein distributions). a) Let N > 1 a positive integer such that 2|N and let k, s be integers such that − 2k > m and κ − k ≤ s ≤ 0 then there exist distributions G
k,s , G+ k,s , Gk,s on v Y with values in Mm r,k (A) such that for any character χ mod N p G
k,s (χ) = G
k (z, s, χ), G+ k,s (χ)

=

− G+ k (z, s, χ), Gk,s (χ)

(4.25) =

G− k (z, s, χ),

where r = m · max(−s, k − κ + s). b) Suppose that κ − k + 1 ≤ s ≤ 0 then the action of the Shimura differential − operators on G
k,s , G+ k,s , Gk,s is given by δk (G
k,s (χ)) = G
k+2,s−1 (χ), + δk (G+ k,s (χ)) = Gk+2,s−1 (χ), − δk (G− k,s (χ)) = Gk+2,s−1 (χ).

(4.26)

c) Moreover if t ≥ 0 is an integer such that κ − k + t ≤ s ≤ 0 then the action (t) − of δk on G
k,s , G+ k,s , Gk,s is given by (t)

δk (G
k,s (χ)) = G
k+2t,s−t (χ),

(4.27)

(t)

+ δk (G+ k,s (χ)) = Gk+2t,s−t (χ), − δk (G− k,s (χ)) = Gk+2t,s−t (χ). (t)

Proof of Proposition 4.6 One deduces a) from (4.24) taking into account that for all integrs k, s such that 2k > m and κ − k ≤ s ≤ 0 the Fourier coefficients b
(ξ, y, s, χ), b± (ξ, y, s, χ)) are polynomials of the variables Ri,j (, j = 1, . . . , m), where R = (4πIm(z))−1 , of degree ≤ r = m · max(−s, k − κ + s) which belong to Q[Ri,j ]i,j=1,...,m . They are explicitely given in Theorem 4.3 above (4.19), (4.20), and in [Pa6] (Chap. 2, §3.5) . Let us consider the natural representation (0 ≤ r ≤ m) ρr : GLm (C) −→ GL(∧r Cm ) z −→ ρr (z)

142

4 Admissible measures for standard L–functions

of the group GLm (C) on the vector space Λr Cm with respect to the basis {ei1 ∧ · · · ∧ eir : i1 < · · · < ir }.   m Thus ρr (z) is a matrix of size m r × r composed of the subdeterminants of z of degree r. Put ρ
r (z) = det(z)ρm−r (tz)−1

(r = 0, 1, · · · , m),

in other words, ρ
m−r (z) is the matrix representing the action of z on Λm−r Cm with respect to the basis dual to the above basis of Λr Cm . Then the representations ρr and ρ
r turn out to be polynomial representations so that for each z ∈ Mm (C) the linear operators ρr (z), ρ
r (z) are well defined. Then by Lemma 3.10 and (3.34) we have that if r and k − κ − r ≥ 0, then r   r −r det(4πy) Rm (4πξy; r, κ − k − r) = (4.28) det(ξ)r−t t t=0      RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ), |L|≤mt−t

where L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m, such that |L| = l1 + · · · + lt ≤ mt − t, the coefficients RL (β) ∈ Z[1/2][β] are polynomials in β of degree (mt − |L|) and with coefficients in the ring Z[1/2], and the polynomials λi (z) are defined by det(t1m + z) =

m 

λi (z)tm−i

i=0

(λi (z) is the sum of all diagonal minors of size i × i of the matrix z). It follows that in (4.19) one has for an integer r = −s ≥ 0: r   r det(4πy)s R(4πξy; −s; κ − k + s) = (4.29) det(ξ)r−t t t=0      RL (κ − k − r)tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ), |L|≤mt−t

and in (4.20) one has for an integer r = k − κ + s ≥ 0: r    r κ−k−s R(4πξy; k + s − κ, −s) = (4.30) det(4πy) det(ξ)r −t t t=0      RL (κ − k − r )tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) ). |L|≤mt−t

These are polynomials of degree ≤ m · max(−s, k − κ + s) with rational coefficients , hence the Fourier coefficients b
(ξ, y, s, χ), b± (ξ, y, s, χ)) define for each ξ distributions with values in Q[Ri,j ]i,j=1,...,m (evaluated at χ). In order to deduce b) and c) (by iteration) it suffices to use then Corollary 4.5.

4.2 Algebraic differential operators and Siegel-Eisenstein distributions

143

4.2.6 Convolutions of distributions with values in nearly holomorphic Siegel modular forms. Consider again the profinite group Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

loc−const Let M = Mm (Y, A) → M be two distributions r,k (A) and let Φi : C m on Y with values in Mi = Mri ,ki (A), (i = 1, 2). Assume that A contains (ϕ(N pv ))−1 and all the values of Dirichlet characters χ mod N pv viewed as homomorphisms Y → A× then obviously  χ(a)Φi ((a)v ), (4.31) Φi (χ) = a mod N pv

1 ϕ(N pv )

Φi ((a)v ) =

 χ mod

χ(a)−1 Φi (χ).

N pv

Let us fix a Dirichlet character ψ on Y with values in A then there exists a unique distribution Φ = Φ1 ∗ψ Φ2 on Y with values in M = Mm r1 +r2 ,k1 +k2 (A) such that for all Dirichlet characters χ on Y with values in A one has Φ(χ) = Φ1 (χ) · Φ2 (ψχ−1 ). One defines Φ on basic open subsets by  Φ((a)v ) = ψ(y)Φ1 (ya)Φ2 (y). (4.32) y∈Yv

for any v such that ψ is defined modN pv . Indeed, Φ((a)v ) =



ψ(y)Φ1 (ya)Φ2 (y) =

y∈Yv

 1 ψ(y) v 2 (ϕ(N p )) y∈Yv

 χ

mod



χ (ya)−1 Φ1 (χ ) χ

N pv

mod

χ (y)−1 Φ2 (χ )

N pv

hence Φ(χ) =



χ(a)

a∈Yv

 χ

mod



ψ(y)Φ1 (ya)Φ2 (y) =

y∈Yv

a∈Yv



χ (ya)−1 Φ1 (χ ) χ

N pv

  1 χ(a) ψ(y) v 2 (ϕ(N p ))

mod

χ (y)−1 Φ2 (χ ) = Φ1 (χ) · Φ2 (ψχ−1 ) =

N pv

  1 (χχ −1 )(a) (ψχ −1 χ −1 )(y) v 2 (ϕ(N p )) a∈Yv

y∈Yv

y∈Yv

= Φ1 (χ) · Φ2 (ψχ

−1

 χ mod N pv

Φ1 (χ )

 χ mod N pv

)

because of the orthogonality relations which give χ = χ, χ = ψχ−1 .

Φ2 (χ )

4.3 A general result on admissible measures with values in nearly holomorphic Siegel modular forms. 4.3.1 Profinite group Yv for example Consider again the profinite group Y = lim ← v

Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

There is a natural projection yp : Y → Z× p . Let A be a normed topological ring over Zp with the norm | · | = | · |p,A , and M be a normed A-module with the norm | · |p,M . Definition 4.7. a) For h ∈ N, h ≥ 1 let Ch (Y, A) denote the A-module of × locally polynomial functions of degree < h of the variable yp : Y → Z× p → A ; in particular, C1 (Y, A) ⊂ Cloc−const (Y, A) (the A-submodule of locally constant functions). We adopt the notation Φ(U) := Φ(χU ) for the characteristic function χU of an open subset U ⊂ Y . Let also denote Cloc−an (Y, A) the A-module of locally analytic functions and C(Y, A) the A-module of continous functions so that C1 (Y, A) ⊂ Ch (Y, A) ⊂ Cloc−an (Y, A) ⊂ C(Y, A). b) For a given positive integer h an h-admissible measure on Y with values in M is an A-module homomorphism  : Ch (Y, A) → M Φ such that for fixed a ∈ Y and for v → ∞       (yp − ap )j dΦ = o(p−v(j−h) ) for all    a+(N pv )

j = 0, 1, . . . , h − 1,

p,M

where ap = yp (a). We adopt the notation (a)v = a+(N pv ) for both an element of Yv and the corresponding open compact subset of Y . 4.3.2 Measures and sequences of distributions α : Ch (Y, A) → We wish now to construct an h-admissible measure Φ M (A) out of a sequence of distributions r∗

1 Φα j : C (Y, A) → Mr ∗

with values in an A-module M = Mr∗ (A) of nearly holomorphic Siegel modular forms over A of type r∗ (for all j ∈ N with j ≤ h − 1 where r∗ ∈ N be a

4.3 A general result on admissible measures

145

natural number, and A an algebra over O as in Section 1). For this purpose we define α : Ch (Y, A) → Mr∗ Φ by



α = Φα ((a)v ) = πα (Φj ((a)v ) ypj dΦ j (a)v

where Φj : C (Y, A) → Mr∗ (A) are certain Mr∗ (A)-valued distributions on Y for all j = 0, 1, . . . , h − 1 , and Φα j ((a)v ) their α-characteristic projections given by !  −v  πα,1 U v Φj ((a)v ) = πα,v +1 (Φj ((a)v ) Φα j ((a)v ) = U 1

for any v ≥ κv. Note first of all that the definition  α = Φα ((a)v ) = U −κv [πα,1 U κv Φj ((a)v )] ypj dΦ j (a)v

α : Ch (Y, A) → Mr∗ (A) is undependent on the choice of of the linear form Φ the level: for any v ≥ κv we have by Proposition 3.5 the following comutative diagram πα,v +1  v  +1 Mr∗ (N pv +1 ; A) −→ Mα ; A) r ∗ (N p    6 U v

  Uv 6 πα,1

Mr∗ (N p; A) −→

Mα r ∗ (N p; A)

in which the right vertical arrow is an A-isomorphism by Proposition 1.6 b), and by linear algebra the A-linear endomorphism U commutes with the characteristic projectors πα,v +1 , πα,1 hence !   U −κv [πα,1 U κv Φj ((a)v )] = U −v πα,1 U v Φj ((a)v ) = πα (Φj ((a)v )). Theorem 4.8. Let α ∈ A be an element of A whose absolute value |α|p satisfies 0 < |α|p < 1 and put h = [vp (α)] + 1. Suppose that there exists a positive integer κ such that for any (a)v ⊂ Y the following two conditions are satisfied:   (level) Φj (a)v ∈ Mr∗ (N pκv ),     j   κv     j 0 j−j  U  ≤ Cp−vj  (a) ) Φ ) (growth) (−a j v p   j    j =0

p

for all j = 0, 1, . . . , κh − 1. α : Ch (Y, A) → Mr∗ with Then there exists an h∗ -admissible measure Φ ∗ h = hκ such that for any ((a)v ) ∈ Y

4 Admissible measures for standard L–functions

146



α = Φα ((a)v ) ypj dΦ j

(a)v

for all j = 0, 1, . . . , h∗ − 1 where −κv Φα [πα,1 U κv Φj ((a)v )] . j ((a)v ) = U

and v πα,v : Mr∗ (N pv ) → Mα r ∗ (N p )

(v ≥ 1)

denotes the α-characteristic projector (canonical projector πα to the α-characteristic submodule of the U -operator ) v Ker(U − αI)n Mα r ∗ (N p ) = n≥1

(with the kernel

7

Im(U − αI)n , see definition 1.5) (where

n≥1 α U κv Φj ((a)v ) ∈ Mα r ∗ (N p) = Mr ∗ (N p; A)

because of the inclusion U κv−1 (Mr∗ (N pκv ; A)) ⊂ Mr∗ (N p; A) (for all v ≥ 1)). (see 3.3 a)) Proof of Theorem 4.8 We need to check the h∗ -growth condition for the linear form α : Ch∗ (Y, A) → Mr∗ Φ (given by the condition of the theorem 4.8), which says that for fixed a, b ∈ Y and for v → ∞      α  (yp − ap )j dΦ = o(p−v(j−hκ) )    (a)v p,Mr∗

for all j = 0, 1, · · · , hκ − 1 where ap = yp (a). α using the binomial formula: Let us develop the definition of Φ  α = (yp − ap )j dΦ (a)v j    j (−ap )j−j Φα j  ((a)v ) = j 

j =0

4.3 A general result on admissible measures

147



α−vκ · αvκ

 j       j · U −vκ πα,1 U κv (−ap )j−j Φj  (a)v  j  j =0

First we notice that all the operators vκ

α

·U

−vκ



−1

= α

U

−vκ

 −vκ   −vκ n−1 t −1 α−1 Z = I +α Z = t t=0

are uniformely bounded for v → ∞ by a positive constant C1 (where U = (N p; A)) because n does not depend on v αI + Z and Z n = 0 for n = rkA Mr∗ −vκ and the binomial coefficients are all Zp -integers. t On the other hand by the condition (growth) of the theorem (for the distributions Φj )     j   κv    j j−j U ( Φj  ((a)v )) ≤ Cp−vj (−ap )  j    j =0 p,Mr∗

for all j = 0, 1, . . . , κh − 1. If we apply to this estimate the previous bounded operators πα,1 U κv we get the following inequality      j α  (yp − ap ) dΦ  ≤ C · C1 |α−vκ |p · p−vj = o(p−v(j−hκ) )    (a)v p,Mr∗

because of the estimate  vκ |α−vκ |p = pvp (α) = o(pvhκ )

4.4 The standard L-function of a Siegel cusp eigenform and its critical values. 4.4.1 The standard L function Let f ∈ Mm k (C, ψ) be an eigenfunction of all Hecke operators and let (α0 , α1 , · · · αm ) = (α0,f (q), α1,f (q), · · · , αm,f (q)) ∈ [(C× )m+1 ]Wm

(4.33)

be the Satake q-parameters of the modular form f (for all primes q not dividing the level C). Recall that the standard L-function D(s, f, χ) of f ∈ Mm k (C, ψ) is the $ product D(s, f, χ) = q
N D(q) (s, f, χ) with D(q) (s, f, χ) = (1 − χ(q)ψ(q)q −s )−1 Rf,q (χ(q)ψ(q)q −s )−1 , that is D(s, f, χ) = (4.34)

−1 "   m
χ(q)ψ(q)
χ(q)ψ(q)αi (q)−1 χ(q)ψ(q)αi (q) , 1− 1− 1− qs qs qs q i=1 where χ is a Dirichlet character mod M . Analytic properties of the standard zeta functions were studied by A. N. Andrianov and V. L. Kalinin [An-Ka] in the case of even degree m, and S. B¨ ocherer [B¨o2] extended these results to the case of arbitrary degree using a different approach (see also [B¨ oSch] where a p-adic version of his construction was developped). For m = 1 and a ∞ normalized cusp eigenform f (z) = n=0 a(n)e(nz) ∈ M1k (C, ψ) we have that D(s, f, χ) = L2,f (s + k − 1, χ), where L2,f (s, χ) = LCM (2s − 2k + 2, χ2 ψ 2 )

∞ 

χ(n)a(n2 )n−s

n=1

is the symmetric square of the modular form f . Let us recall some basic equalities from 2.2 concerning the integral representation of the standard L-function 4.4.2 Theta series (See in section 2.2.1 (Chapter 2), and in [An2], [An-Ma1], [An-Ma2], [St2]). Let F ∈ 2Cm be an even symmetric positive definite matrix, and q0 its level (i.e. the smallest positive integer such that q0 F −1 ∈ 2Cm , and χ a Dirichlet character modulo Q (not necesarily primitive). Put ν = 1 or 0 and the theta function was defined by (2.60):  (ν) χ(det(ξ))det(ξ)ν em (zF [ξ]/2). (4.35) θ(χ) = θF (z, χ) = ξ∈Mm (Z)

4.4 The standard L-function

149

Proposition 4.9. (a) If  γ=

ab cd

∈ Γ0m (q0 Q2 ),

then the following transformation formula (2.61) holds (ν)

(m)

(ν)

θF (γ(z), χ) = χ(det(d))χQ2 F (γ)det(cz + d)(m/2)+ν θF (z, χ),

(4.36)

(m)

where χF (γ) is a root of unity of the eighth degree, and if m is even, then  (−1)m/2 det(F ) (m) χF (γ) = . (4.37) det(d)  0m −1m . If χ is primitive (b) Let J(M ) denote for M > 0 the matrix M 1m 0m modulo Q then the action of the involution J(Q2 q0 ) on (2.60) is given by (2.63) (ν) θF (J(Q2 q0 )z, χ) = χ(−1)m (iQ)mν GQ (1m , χ)det(Fˆ )m/2+ν ·

(4.38)

(ν) ·det(−iz)m/2+ν θFˆ (z, χ) ¯

with Fˆ = q0 F −1 and 

GQ (ξ, χ) =

χ(det(h))em (tξh/Q)

h∈Mm (Z)modQ

is the Gauss sum of degree m of the character χ.

4.4.3 The Rankin zeta function Let f and g be two holomorphic modular forms of weights k and l on the congruence subgroup Γ0m (N ) ⊂ Γ m , and with Dirichlet characters ψ and ω. More precisely we suppose that  k a(ξ)em (ξz) ∈ Sm (N, ψ), (4.39) f (z) = ξ∈Cm

g(z) =



b(ξ)em (ξz) ∈ Mlm (N, ω),

(4.40)

ξ∈Bm

with Bm being the set of half integral non negative matrices of size m × m, and Cm ⊂ Bm the subset of all positive definite matrices (see section 2.1). Define an equivalence relation on Bm by ξ1 ∼ ξ2 iff ξ1 = tuξ2 u for some matrix u ∈ SLm (Z). Then the Rankin zeta function (convolution of two Siegel modular forms) was defined in (2.80) as the series [St2]:

150

4 Admissible measures for standard L–functions

L(s, f, g) =



ε(ξ)−1 a(ξ)b(ξ)det(ξ)−s

(s ∈ C),

(4.41)

ξ∈Cm /∼

where ε(ξ) = |Aut(ξ)| is the order of the unit group ' ( Aut(ξ) = η ∈ SLm (Z) | tηξη = ξ[η] = ξ . This series is well defined due to basic properties of Fourier coefficients of Siegel modular forms; Cm / ∼ denotes the orbit space with respect to ∼ and Re(s) is supposed to be large enough: Re(s) ≥ 0. Proposition 4.10 (Integral representation for the Rankin zeta function). (See Proposition 2.8) and [St2], proposition 6). For s with Re(s)  0 there is the following integral representation. (4π)−ms Γm (s)L(s, f, g) = . / f ρ , gE(z, s − k + (m + 1)/2; k − l, ψω, N ) ,

(4.42)

N

where the inner product is defined by (2.14),  k ¯ a(ξ)em (ξz) ∈ Sm (N, ψ), f ρ (z) = ξ∈Bm

and Γm (s) denotes the Γ -function of degree m, i.e. Γm (s) = π m(m−1)/4

m−1


Γ (s − (j/2)),

j=0

(see also (2.94)). 4.4.4 The standard zeta function D(s, f, χ) as the Rankin convolution Now let f (z) =



a(ξ)em (ξz) ∈ Skm (N, ψ),

ξ∈Cm

be a cusp form of even degree m which is an eigenfunction of the Hecke algebra Lm q (N ) for q not dividing N , and let (α0 , α1 , · · · , αm ) = (α0 (q), α1 (q), · · · , αm (q))) ∈ [(A× )m+1 ]Wm be the corresponding (m+1)-tuple of the Satake q-parameters of f (see (2.21)). Fix a matrix ξ0 ∈ Cm such that a(ξ0 ) = 0 and consider the primitive quadratic Dirichlet character χξ0 definied for positive integers d with (d, 2det(2ξ0 )) = 1 by the formula

4.4 The standard L-function

 χξ0 (d) =

(−1)m/2 det(2ξ0 ) d

151

,

and when det(2ξ0 ) is odd, then by definition χξ0 (2) = 1 or −1 according as to which of the two following quadratic forms x1 x2 + x3 x4 + . . . xm−1 xm x1 x2 + · · · + xm−3 xm−2 + x2m−1 + xm−1 xm + x2m our quadratic form ξ0 is equivalent over the field F2 of two elements. Assume also that χ mod M is chosen so that (−1)ν = χ(−1). The main result of the A.N.Andrianov’s work [An3] can be stated as a certain identity expressing the standard zeta function D(s, f, χ) as a Rankin zeta function, namely the convolution of the given form f and a theta function with the Dirichlet character χ mod M . The precise statement of the result is given in the following proposition. Proposition 4.11. Under the notation and assumptions as above for the sufficiently large values of Re(s) the following identity holds a(ξ0 )R(s, f, χ) = 2

−1

(4.43) (s+k−1+ν)/2

det(ξ0 )

L((s + k − 1 +

(ν) ν)/2, f, θ2ξ0 (z; χ)),

where the function R(s, f, χ) is defined by the following equality D(s, f, χ) =






m/2−1

L(s + (m/2), χψχξ0 ) 



(4.44)

L(2s + 2i, χ2 ψ 2 ) R(s, f, χ),

i=0

and it is assumed that the modulus M of the character χ is divisible by all (ν) prime divisors of the number N det(2ξ0 ), where θ2ξ0 (z, χ) is the theta function as in 2.2.1. More explicitly, the right hand side of (2.83) can be represented as the series  χ(det(ξ))det(ξ)ν a(ξ0 [ξ])det(ξ0 [ξ])−(s+k−1+ν)/2 = det(ξ0 )(s+k−1+ν)/2 ξ

=



χ(det(ξ))a(ξ0 [ξ])det(ξ)−(s+k−1+ν)/2 ,

(4.45)

ξ

where the summation is taken over the set of equivalence classes ξ ∈ SLm (Z)\M+ m (Z) of the form SLm (Z)ξ. Now we set q0 to be equal to the level of the quadratic form with the matrix 2ξ0 (see 2.2.1). In order to get the integral representation of the standard zeta function we apply to it the result of proposition 2.8. For this purpose we put

152

4 Admissible measures for standard L–functions

in the notation of proposition 2.8 l = (m/2) + ν, ω = χχξ0 with χ being a Dirichlet character modulo M , and take the number N q0 M 2 as N . We note also that both parts of (2.83) and (2.84) converge absolutely for Re(s) > m (see also [An5], p. 133). Proposition 4.12 (integral representation of the standard zeta function). For the notations and assumptions as above about Re(s), the following equality holds: 2a(ξ0 )det(ξ0 )−(s+k−1+ν)/2 (4π)−ms Γm ((s + k − 1 + ν)/2)R(s, f, χ) = . / (ν) = f ρ , θ2ξ0 (z, χ)E(z, (s − k + m + ν)/2) , (4.46) N M 2 q0

where the Siegel-Eisenstein series in the right hand side E(z, s) = E(z, s; k − ν − (m/2), χψχξ0 , N M 2 q0 ) is defined by (2.76). 4.4.5 Algebraic properties of the special values of normalized distributions. Consider only the case of even m. In order to give a precise statements on algebraic properties of the standard zeta functions and of the corresponding distributions, it is convenient to make some additional normalization of these values because these properties look different for the integral points to the left and to the right of the critical line Re(s) = 12 (the same effect happens in the case of the Riemann zeta function). Recall the following three types of the normalized zeta function: for κ = m+1 2 , s∈C D∗ (s, f, χ) = (2π)−m(s+k−κ) Γ ((s + δ)/2)

m


Γ (s + k − j)D(s, f, χ), (4.47)

j=1

D− (s, f, χ) = Γ ((s + δ)/2)−1 D∗ (s, f, χ) = (2π)−m(s+k−κ)

m


Γ (s + k − j)D(s, f, χ),

(4.48)

j=1

D+ (s, f, χ) =

2 iδ Γ (s) cos(π(s − δ)/2) − D (s, f, χ) = (2πi)s

(4.49)

iδ π (1−2s)/2 D∗ (s, f, χ), Γ ((1 + δ − s)/2) where δ = 0 or 1 according as ψχ(−1) = (−1)δ with f ∈ Sm k (C, ψ) be an eigenfunction of the global Hecke algebra Lm (C) = ⊗q |C Lm q (C) with the eigenvalue given as a homomorphism λf : Lm (C) → C.

4.4 The standard L-function

153

Now we turn to the normalizations (4.48) and (4.49). Their convenience is illustrated by the following result on algebraicity of the special values of (4.48) and (4.49). Theorem 4.13 (Algebraic properties of the special values of the standard zeta function). Assume that the cusp eigenform f ∈ Sm k (C, ψ) satisfies the condition a(ξ0 ) = 1 for some ξ0 ∈ Am , ξ0 > 0. Then a) For all integers s satisfying 1 ≤ s ≤ k − δ − m and s = 1 if the character χ2 ψ 2 is trivial we have that f, f −1 D+ (s, f, χ) ∈ K = Q(f, λf , ψ, χ),

(4.50)

where K = Q(f, λf , ψ, χ) denote the field generated by Fourier coefficients of f , by the eigenvalues λf (X) of Hecke operators X on f , and by the values of the characters χ and ψ. Moreover, D+ (s, f, χ) = 0 for s ≡ δ(mod2). b) For all integers s satisfying 1 − k + δ + m ≤ s ≤ 0 we have that f, f −1 D− (s, f, χ) ∈ K, and

(4.51)

D− (s, f, χ) = 0 for s ≡ δ(mod2), s ∈ Z, s ≤ 0.

Proof of Theorem 4.13 (See also [Pa6], Chapter 3). One uses the identity (4.46), which says that for all non-trivial Dirichlet characters χ mod M : 2a(ξ0 )((4π)m det(ξ0 ))−(s+k−1+ν)/2 Γm ((s + k − 1 + ν)/2) D(s, f, χ) = LN (s +

m , ψχξ0 χ) 2




(m/2)−1

LN (2s + 2i, ψ 2 χ2 )) ×

(4.52)

i=0

(ν)

× f ρ (z), θ2ξ0 (z; χ)E(z, (s − k + m + ν)/2) N , E(z, s) = E(z, s; k − (m/2) − ν, ψχξ0 χ, N ) is the Siegel – Eisenstein series of the weight k − (m/2) − ν, the level N = 4q0 CM 2 , and the Dirichlet character ψχξ0 χ is viewed modN . For the convenience of the reader we give here more details on the proof of the algebraicity results (taken from [Pa6], p. 101-110). Proposition 4.14. Le f ∈ Sm k (C, ψ) be a cusp form of weight k ≥ m + 1 where m is even, χ be a Dirichlet character modulo M ≥ 1. Put N = 4qo M 2 C where q0 is the level of a quadratic form with the matrix 2ξ0 such that a(ξ0 ) = 0. Then we have that

4 Admissible measures for standard L–functions

154

∗ a(ξ0 )DN (s, f, χ) = f ρ , K ∗ (z, s; ξ0 , χ) N ,

where K ∗ (z, s; ξ0 , χ) = N −m(2s+m)/4 2m(2k−m+2−κ) i−m(k−(m/2)−ν) × (4.53) (s+k−1+ν)/2 (ν) θ2ξ0 (χ)G∗ (z, (s

×detξ0

− k + ν + m)/2)|W (N ),

∗ (s, f, χ) indicates that all Euler where the subscript N in the notation DN factors corresponding to q, q|N are removed from the Euler product, and the series G∗ (z, s) = G∗ (z, s; k − (m/2) − ν, χχξ0 ψ, N )

being defined by the equality (2.133) of chapter 2. The proof is deduced from the integral representation ((4.46)) rewritten in the form 2a(ξ0 )((4π)m detξ0 )−(s+k−1+ν)/2 Γm ((s + k − 1 + ν)/2) D(s, f, χ) = ˜ s; ξ0 , χ) N , (4.54) f ρ (z), K(z, where ˜ s; ξ0 , χ) = K(z, LN (s +

m , ψχξ0 χ) 2

(ν) ×θ2ξ0 (z; χ)[E(z, (s

(4.55)


(m/2)−1

LN (2s + 2i, ψ 2 χ2 )) ×

i=0

− k + m + ν)/2)|k−(m/2)−ν W (N )]|k−(m/2)−ν W (N ).

In the above equality we used definition (4.47) of the normalized zeta functions, the definition of the series G∗ (z, s) and the relation (2.137 ) for the Γ -factors. Now Theorem 4.13 is implied by Proposition 4.14 and Theorem 2.13 of chapter 2 in which we take k to be equal to k − (m/2) − ν. ∗ (s, f, χ) is obtained from the function D∗ (s, f, χ) Note that the function DN if we multiply it by an elementary holomorphic factor; however we do not know in general how to obtain holomorphy properties of the function D∗ (s, f, χ) itself from the theorem, and this interesting question needs a further study. However, under the assumptions of Theorem 4.13 we have that ∗ (s, f, χ), D∗ (s, f, χ) = DN

hence the proof is completed. 4.4.6 Integral representation for the functions D ± (s, f, χ) In order to prove Theorem 4.13 on algebraic properties we need an integral representation for the functions D± (s, f, χ) analogous to that of Proposition 4.14.

4.4 The standard L-function

155

Let χ be a Dirichlet character modulo M and assume that all conditions 2 of Theorem 4.13 are satisfied. Put N = 4qM C, then we have the following integral representation ± (s, f, χ) = f ρ , K ± (z, s; ξ0 , χ) N , a(ξ0 )DN

(4.56)

with K ± (z, s; ξ0 , χ) = N −m(2s+m)/4 2m(2k−2−m−κ) i−m(k−(m/2)−ν) × (s+k−1+ν)/2

×detξ0

Hol [θ2ξ0 (χ)|W (N )G± (z, (s − k + ν + m)/2)]|W (N ), (ν)

in which K ± (z, s; ξ0 , χ) ∈ Mm k (N, ψ), the series

G± (z, s) = G± (z, s; k − (m/2) − ν, χχξ0 ψ, N )

are defined by (2.135), (2.136) of Chapter 2 and the symbol Hol denotes the holomorphic projection operator from Theorem 2.16 of chapter 2. The proof of (4.56) is carried out in exactly the same way as that of the proposition 3.9 if we take into account definitions (4.2) and (4.2) of the functions D± (s, f, χ), the definition of the series G± (z, χ) and use the relation (3.12) for the Γ -factors. The possibility of applying Hol to the function θ2ξ0 (χ)|W (N )G± (z, (s − k + ν + m)/2) (ν)

(4.57)

by formulas of Theorem 2.147 of chapter 2 is justified as in the end of Section 2.4 in chapter 2 bearing in mind positivity properties of Fourier expansions of the series G± (z, s) in Theorem 2.147 of chapter 2 and the growth estimates mentioned above. From these estimates it follows that the function (4.57) satisfy the bounded growth condition, and its Fourier expansion contains only terms with positive definite matrix indices. Remark 4.15. If k > 2m + 2 then for s = k − ν − m the series defining the function G± (z, 0) = G∗ (z, 0; k − (m/2) − ν, χχξ0 ψ, N ) is absolutely convergent so that this function is holomorphic, and we can omit the symbol Hol in the integral representation (4.56): a(ξ0 )Γ + (k − ν − m)DN (k − ν − m, f, χ) = f ρ , K ± (z, k − ν − m; ξ0 , χ) N , with

(4.58)

156

4 Admissible measures for standard L–functions

K ± (z, k − ν − m; ξ0 , χ) =

(4.59)

−m(2k−2ν−m)/4 m(2k−2−m−κ)+m(2k−2ν−m)/2 −m(k−(m/2)−ν)

N 2 (ν) ×θ2ξ0 (χ)[G± (z, 0)|W (N )],

i

×

in which Γ + (s) = (2π)−m(s+k−κ)

m 2iδ Γ (s) cos(π(s − δ)/2)
Γ (s + k − j) (2π)s j=1

(4.60)

is the gamma-factor. With these values of k and s the identity of A.N.Andrianov (Proposition 2.84 of Chapter 2 and the equality (4.46) takes the form: a(ξ0 )D+ (s, f, χN ) = a(ξ0 )D+ (s, f, χ) =


(4.61)

(m/2)−1

Γ + (s)LN (s + (m/2), χ χξ0 ψ) ×

LN (2s + 2i, χ2 ψ 2 ) ×

i=0



χN (deth)a(ξ0 [h])deth−(s+k−1) ,

h∈SLm (Z)\M+ m (Z)

where χ be a Dirichlet character modulo N defined by χN (d) = χ(d) for det2ξ0 |N . The series in (4.61 ) is absolutely convergent for Re(s) > 1 + m due to the estimate |a(h)| = O(deth(k/2) ) for the Fourier coefficients. 4.4.7 Action of the group Aut(C) on scalar products of modular forms. Let us fix a positive integer N1 . Recall that the group Aut(C) acts on modular forms  a(ξ)em (ξz) ∈ Mm f= k (N1 , ψ) Am ξ≥0

by the following rule:  σ fσ = a(ξ)σ em (ξz) ∈ Mm k (N1 , ψ ) (σ ∈ Aut(C)), Am ξ≥0

and this action commutes with the action of the Hecke algebra, see [Sh7], [St2]. Consider the global Hecke algebra L(N1 ) = ⊗q
N1 Lm q (N1 ) and suppose that f ∈ M(N1 , ψ) is an eigenfunction of the Hecke algebra L(N1 ) with the eigenvalue given by a homomorphism Λ : L(N1 ) → C, i.e.

4.4 The standard L-function

157

f |X = Λ(X)f for all X ∈ Lm q (N1 ) and all q
N1 . Let N1 |N . We define a Λ-packet of modular forms as the following subspace of Sm k (N, ψ): % & m Hkm (Λ, N, ψ) = f ∈ Sm k (N, ψ)|f |X = Λ(X)f, X ∈ Lq (N1 ), q
N , and put Hkm (Λ, ψ) = Sm k (ψ)

=

3 N ≡0( mod N1 )

3

Hkm (Λ, N, ψ),

(4.62)

m N ≡0( mod N1 ) Sk (N, ψ).

We know due to Shimura that the action of Aut(C) commutes with the action of the Hecke algebra [Sh7], therefore for each σ ∈ Aut(C) f ∈ Hkm (Λ, ψ) ⇐⇒ f σ ∈ Hkm (Λσ , ψ σ ),

(4.63)

where Λσ (X) = Λ(X)σ . On the other hand, if we use the normality property of the Hecke operators with respect to the Petersson scalar product and commutativity of the Hecke algebra Λ(N1 ) we see (as in the classical case) that for a certain set of homomorphisms Λ = Λ1 , · · · , Λt there is the decomposition of S(N, ψ) into the orthogonal direct sum of the corresponding Λ-packets: S(N, ψ) =

t 8

Hkm (Λi , N, ψ).

(4.64)

i=1

The following proposition was established by J.Sturm ([St2], Theorem 3). We statehere this result in a form, which is more suitable for our applications. Proposition 4.16. Let m be even then for any integer k with k > 2m + 2, a Dirichlet character ψ mod N , and a homomorphism Λ : L(N ) → C there exists a non zero constant µ(Λ, k, ψ) ∈ C× depending only on Λ, k, ψ such that 2σ 1 ρ f σρ , g σ N f , g N (4.65) = µ(Λ, k, ψ) µ(Λσ , k, ψ σ ) for all f ∈ Hkm (Λ, N, ψ), g ∈ Mm k (N, ψ), σ ∈ Aut(C). Remark 4.17. If we take in equality (4.65) g equal to f ρ then proposition 4.16 implies that f, f N µ(Λ, k, ψ)−1 ∈ Q(Λ, f ), with Q(Λ, f ) being the subfield of C generated by the values of the homomorphism Λ and the Fourier coefficients of f . We give here a proof based on the Andrianov’s identity (4.102) in which the right-hand side has the form:
D+ (s, f, χN ) = Γ + (s) R(Λ, q, k)(χ(q)ψ(q)q −s ) (4.66) q|N

158

4 Admissible measures for standard L–functions

where R(Λ, q, k)(t) ∈ Q[Λ(X), X ∈ Lm q (N )] [t] are polynomials satisfying R(Λ, q, k)(0) = 1 which depend only on the Λpacket of the form f and on the numbers q and k. The product in (4.66) converges absolutely for Re(s) > 1 + m. Put s = k − ν − m where ν = 0, 1, k ≡ ν mod 2, take as χN the trivial character modulo N and define
R(Λ, q, k)(ψ(q)q −(k−ν−m) ), (4.67) µ(Λ, k, ψ) = G(ψ)m−1 Γ + (k − ν − m) q
2B

with the Γ -factor being defined by (4.93), B = B(Λ, k, ψ) being a positive integer such that the product in (4.67) does not vanish; B exists due to absolute convergence of the product (see the Remark 4.15), so that we can and we will assume that B(Λ, k, ψ)σ = B(Λσ , k, ψ σ ). Now, with the number µ(Λ, k, ψ) already been defined, we prove first the proposition 4.16 for the special modular form g = G(ψ)m−1 K + (ξ0 , ψ), where we adopt the notation K + (ξ0 , ψ) = K + (z, k −ν −m; ξ0 , χN ) in order to stress, that the function K + (z, k − ν − m; ξ0 , χN ) depends on ξ0 and on χN (recall that we took χN to be the trivial character modulo N ). Then the following identity holds: [G(ψ)m−1 K + (ξ0 , ψ)]σ = G(ψ σ )m−1 K + (ξ0σ , ψ σ ).

(4.68)

This important fact expresses in a more precise form the result of proposition 2.15 of Chapter 2 on the cyclotomy of the Fourier coefficients of this Siegel modular form. The identity (4.68) will be proved later in Proposition 4.19, and now we deduce from it Proposition 4.16. According to the equalities (4.66) and (4.58), there is the following relation: G(ψ)m−1 f ρ , K + (ξ0 , ψ) N µ(Λ, k, ψ)
= a(ξ0 ) R(Λ, q, k)(ψ(q)q −(k−ν−m) ),

(4.69)

q|2B q |det2ξ0

with the finite Euler product in the right hand side on which the automorphisms σ ∈ Aut(C) act term-by-term. Therefore, from (4.68) and (4.69) it follows that for the functions of the type g = G(ψ)m−1 K + (ξ0 , ψ) satisfy the relation (4.65). In order to deal with the general case we vary ξ0 ∈ Am , ξ0 so that the number N = N (ξ0 ) will now depend on ξ0 , and consider the trace operator Tr(N2 , N1 , ψ) =  where elements g(i) = of the right cosets:

a(i) b(i) c(i) d(i)



d 

ψ(a(i))F |k g(i), )

i=1

form a complete system of representatives

4.4 The standard L-function

Γ0m (N1 ) =

d )

159

Γ0m (N2 )g(i).

i=1

The important property of this action is that it commutes with the trace operator. This fact is stated more precisely in the following proposition. Proposition 4.18. Let F ∈ Mm k (N2 , ψ) be a Siegel modular form with cyclotomic Fourier coefficients,  F (z) = A(ξ)em (ξz), A(ξ) ∈ Qab . ξ≥0

Then for all σ ∈ Aut(C) the following equality holds [f |Tr(N2 , N1 , ψ)]σ = [f σ |Tr(N2 , N1 , ψ σ )]. Proof is given in [St2], lemma 11, and it is based on properties of the action of the restricted adele group GA+ on the graded ring of automorphic forms studied by Shimura [Sh8], where GA+ denotes the subgroup of the adelization % & GA = α ∈ GL2m |t αJm α = ν(α)Jm , ν(α) ∈ GL1 (A) consisting of all elements α ∈ GA for which Archimedean component of ∈ Mm the ν(α) is positive. For x ∈ GA+ and a modular form k = 3 idele $F + m m x M (Γ (M )) the action of x on F is denoted by F . If t ∈ Z an idele k M q q ab whose action on Q by class field theory coincides with that of σ ∈ Aut(C), then we have that  1m 0m σ x(t−1 ) F =F with x(t) = ∈ GA+ 0m t1m for the above action on modular forms. Now the proposition is easily deduced from this equality. In view of the strong approximation theorem for the group Sp2m (A) we can choose for each representative g(i) in (4.4.7) elements u(i) ∈ Sp2m (A), h(i) ∈ Sp2m (Z) such that u(i)q ≡ 12m (modN2 ) for all primes q, q|N2 and   1m 0m 1m 0m g(i) = u(i) h(i). 0m t−1 1m 0m t1m −1

Let us take into account that F σ = F x(t 

(F σ )|Tr(N2 , N1 , ψ σ

σ−1

)

then we get the equality  σ−1 = ψ(a(i))σ F σ |g(i) i

and from the choice of h(i) it follows that

160

4 Admissible measures for standard L–functions

Γ0m (N1 ) =

d )

Γ0m (N2 )h(i),

i=1









(i) b (i) and a(i) ≡ a(i) mod N2 , so that the proposition with h(i) = ac (i) d (i) follows. Now we are able to accomplish the proof of the proposition 4.16. We let the element ξ0 in the equality (4.69) vary, and put N2 = B 2 N det2 2ξ0 , N1 = N. Then f ρ , K + (ξ0 , ψ) N2 = f ρ , K + (ξ0 , ψ)|Tr(N2 , N1 , ψ) N1 ,

and from proposition 4.18 it follows that the equality (4.65) holds for all modular forms g from the set  ( '  V = G(ψ)m−1 K + (ξ0 , ψ)|Tr(N2 , N1 , ψ)Am  ξ0 > 0, N2 = B 2 det2 2ξ0 , N1 = N . Let

 ' (  V1 = g1 ∈ Hkm (Λ, N, ψ)g − g1 is orthogonal to some g ∈ V .

In other words, the set V1 consists of those elements in Hkm (Λ, N, ψ) which are orthogonal projections of the special elements g ∈ V as above. One checks as in [St2] that V1 generates the whole Λ-packet Hkm (Λ, N, ψ). Indeed, if  a1 (ξ)em (ξz) ∈ Hkm (Λ, N, ψ) f1 = ξ>0

satisfies f1 , g1 N = 0 for all g1 ∈ V1 then (3.39) implies that a(ξ0 ) = 0 for all ξ0 ∈ Am , ξ0 > 0 hence f1 ≡ 0. Also, it follows that proposition 4.16 is valid for all g1 ∈ Hkm (Λ, N, ψ). For an arbitrary Siegel modular form g ∈ Mm k (N, ψ) let us write g = g1 + h with g1 ∈ Hkm (Λ, N, ψ) and g1 , h1 N = 0. Now if we combine (4.63) and (4.64) and get the following: σ

g σ = g1σ + hσ1 , g1σ ∈ Hkm (Λσ , N, ψ ), g1σ , hσ1 N = 0.

(4.70)

In order to obtain the above equality we used the important fact about the invariance of the subspace of all Eisenstein series in Mm k (N, ψ) under the action of σ ∈ Aut(C). In turn, this property follows from the general decomposition theorem [Kl3], describing the subspace of all Eisenstein series ( the orthogonal complement to the subspace of cusp forms) in terms of the Klingen – Eisenstein series, and from the invariance properties of these series under the action of complex automorphisms σ ∈ Aut(C), established by M.Harris [Ha2], [Ha3] (see also [B¨o2], [Ku-Mi], [Miz1], [Miz2]). This last fact (stated in a more precise form) comprise the content of the Garrett’s conjecture, which was proved in [Ha3]. However, we do not use this fact any more and therefore will not go into detail of this interesting research. Returning to the equality (4.70) we get

4.4 The standard L-function

161

[ f ρ , g N µ(Λ, k, ψ)−1 ]σ = [ f ρ , g1 N µ(Λ, k, ψ)−1 ]σ = f σρ , g1σ N µ(Λσ , k, ψ σ )−1 = f σρ , g σ N µ(Λσ , k, ψ σ )−1 , To accomplish the proof of proposition 4.16 we need only to check the property (4.68) which will be now stated in the following more general form. Proposition 4.19. Let χ mod N be an arbitrary Dirichlet character, and K + (ξ0 , ψ, χ) = K + (z, k − ν − m; ξ0 , ψ, χ) ∈ Mm k (N, ψ) denote a modular form in the integral representation (3.30). Then for all σ ∈ Aut(C) there is the following relation [G(ψχ)m−1 K + (ξ0 , ψ, χ)]σ = G(ψ σ χσ )m−1 K + (ξ0 , ψ σ , χσ )

(4.71)

Proof From the definition of the theta series we easily see that (ν)

(ν)

θ2ξ0 (z, χ)σ = θ2ξ0 (z, χσ ). Therefore it suffices to check the following property: 1/2

[G(ψχ)m−1 detξ0 G+ (z, 0; χχξ0 ψ, k − (m/2) − ν, N )|W (N )]σ = σ

σ m−1

G(ψ χ )

) ((detξ0 1/2)σ G+ (z, 0; (χχξ0 ψ)σ , k

(4.72)

− (m/2) − ν, N )|W (N ),

(in view of the equality k−(m+1)/2

detξ0

k−(m/2)−1

= detξ0

, m/2 ∈ Z.)

According to formulas in §2.3.1, ( 2.141) of chapter 2, the Fourier coefficient of the series G+ (z, 0; χχξ0 ψ, k − (m/2) − ν, N ) by em (hz) with h ∈ Am , h > 0 has the form (det h)k−ν−m−(1/2) L+ (k − ν − m, χχh χξ0 ψ)M (h, χχξ0 ψ, s), where M (h, χχξ0 ψ, s) =




(4.73)

Mq (h, χχξ0 ψ(q), q −s )

q∈P (h)

is the finite Euler product (2.132) of Chapter 2, where the product is extended to all primes q in the set P (h) of prime divisors of N and the elementatry divisors of the matrix h, with the property Mq (h, t) ∈ Z[t]. Therefore M (h, χχξ0 ψ, k − ν − (m/2))σ = M (h, (χψ)σ χξ0 , k − ν − (m/2)). Now let us consider the factor L+ (k−ν−m, χχh χξ0 ψ) and recall an elementary result on the special values of the Dirichlet L-functions [La1], [Sh3], [Wa],

162

4 Admissible measures for standard L–functions

[Le2]. Let χ mod N be a Dirichlet character of conductor N0 , and χ0 mod N0 be the corresponding primitive character, G(χ) = G(χ0 ) =

N0 

χ0 (x)e(x/N0 )

x=1

be its Gauss sum. We set, for a positive integer r P (r, χ) = G(χ)−1 (2πi)−r L(r, χ). Then for all σ ∈ Aut(C) and χ(−1) = (−1)r we have that P (r, χ)σ = P (r, χσ )

(4.74)

If we apply this property to normalized Dirichlet L-series we see that [G(χ)−1 L+ (r, χ)]σ = G(χσ )−1 L+ (r, χσ ) (r ∈ Z, r > 0)

(4.75)

L− (r, χ)σ = L− (r, χσ ) (r ∈ Z, r ≤ 0)

(4.76)

so that for the values of the “wrong parity” the corresponding values vanish. From the basic property of Gauss sums it follows that G(χ)σ = χσ (v)−1 G(χσ ) f or v ∈ (Z/N Z)× , such that e(1/N )σ = e(v/N ). The last property implies the useful relation: G(ψ)σ G(χ)σ G(ψχ)σ = G(ψ σ χσ ) G(ψ σ )G(χσ )

(4.77)

Let us now apply (4.75) and (4.77) to the coefficients (4.73), then we get [G(ψχ)−1 (detξ0 h)1/2 L+ (k − ν − m, χχh χξ0 ψ]σ = σ

σ −1

G(ψ χ )

((detξ0 h)

(4.78)

) L (k − ν − m, (χψ) χξ0 χh ).

1/2 σ

+

σ

In the equality (4.79) we used the following elementary property of Gauss sums: G(χξ0 χh ) (det(ξ0 h))1/2 ∈ Q ˇ which is due to the fact that χξ0 χh is an even quadratic character (see [Bo-Sa]). To complete the proof of (4.72) and (4.73) we need the following general compatibility property of the action of Aut(C) and of the involution W (N ) (see [St2], lemma 5, and [Sh8]): for a modular form f ∈ Mm k (N, ψ) with cyclotomic Fourier coefficients and for all σ ∈ Aut(C) we have that f σ |W (N ) = ψ(v m )σ (f |W (N ))σ ,

(4.79)

where v ∈ (Z/N Z)× is chosen so that e(1/N )σ = e(v/N ). Now the proof of Proposition 4.16 is finished.

4.4 The standard L-function

163

4.4.8 Algebraicity properties and Fourier coefficients Now in order to deduce theorem 3.2 on algebraicity of the normalized standard zeta function and the algebraicity property of the normalized distributions we use the already proved algebraicity properties of Fourier coefficients of the functions K ± (z, s; ξ0 , χ)|W (N ) in the corresponding integral representation (3.28) and of the functions F ± (z, χ) in (3.13) and (3.14). We apply these properties in the form given below: let k > 2m + 2, f ∈ Hkm (Λ, N, ψ) ⊂ m Sm k (N, ψ) be a cusp form, an eigenfunction of the Hecke algebra L (N ) with m an eigenvalue given as the homomorphism Λ : L (N ) → C. Let us consider the following linear form Lf : g →

f ρ , g|W (N ) N f, f N

(4.80)

on the vector space Mm k (N, ψ) with  g= b(h)em (hz) ∈ Mm k (N, ψ) h≥0

being its arbitrary element. Then there exist positive matrices h1 , h2 , · · · , ht ∈ Am and algebraic numbers α1 , α2 , · · · , αt ∈ Q(f, Λ, ψ) from the field Q(f, Λ, ψ) generated by the Fourier coefficients of f and values of the homomorphism Λ and the character ψ such that for all g ∈ Mm k (N, ψ) the linear form is explicitly given by  αi b(hi ). (4.81) Lf (g) = i

Indeed, we notice that every Siegel modular form of weight k > 2m is uniquely determined by its Fourier coefficients with positive matrix indices h ∈ Am . This fact is equivalent to saying that for such a weight k there are no singular modular forms (i.e. having only Fourier coefficients with degenerate h ∈ Am , det h = 0), which was established by G.L.Resnikov ([Res],[Ra3]). Then Proposition 4.16 implies that the number f ρ , g|W (N ) N ∈ Q(f, g, Λ, ψ) µ(Λ, k, ψ)−1 belongs to the field Q(f, g, Λ, ψ) generated by the Fourier coefficients of the forms f and g, by the values of Λ and ψ. Moreover we have that f, f N µ(Λ, k, ψ)−1 ∈ Q(f, Λ, ψ)× (see the remark after proposition (4.16), and (4.81) follows. In order to prove Theorem 4.13 we use Proposition 4.16 and substitute in it the modular forms K ± (z, s; ξ0 , χ)|W (N ) ∈ Mm k (N ; ψ),

164

4 Admissible measures for standard L–functions

for g. This modular for has cyclotomic Fourier coefficients vanishing for degenerate matrix indices h ∈ Am such that the action of σ ∈ Aut(C) on them is described as in Proposition 4.19 As a consequence, we obtain also the following more explicit description of the action of σ ∈ Aut(C) on the special values in question. Put ˜ + (s, f, χ) = G(ψχ)m−1 D+ (s, f, χ)µ(Λ, k, ψ)−1 , D ˜ − (s, f, χ) = G(ψχ)m D− (s, f, χ)µ(Λ, k, ψ)−1 . D Then under the assumption of Theorem 4.13 for every σ ∈ Aut(C) we have that ˜ ± (s, f, χ)σ = D ˜ ± (s, f σ , χσ ). D

(4.82)

In particular, this relation implies the algebraicity result in Theorem 4.13. Notice that in this situation the constant µ(Λ, k, ψ) depends only on the Λpacket of f and it coincides with that for the original cusp form f as above.

4.5 Algebraic linear forms on arithmetical nearly holomorphic Siegel modular forms 4.5.1 Convolutions of theta distributions and Eisenstein distributions with values in nearly holomorphic Siegel modular forms. Consider again the profinite group Y = lim Yv , where Yv = (Z/N pv Z)× . ← v

loc−const (Y, A) → M be two distributions Let M = Mm r,k (A) and let Φi : C m on Y with values in Mi = Mri ,ki (A), (i = 1, 2). Assume that A contains (ϕ(N pv ))−1 and all the values of Dirichlet characters χ mod N pv viewed as homomorphisms Y → A× then obviously

Φi (χ) =



χ(a)Φi ((a)v ), Φi ((a)v ) =

a mod N pv

1 ϕ(N pv )



χ(a)−1 Φi (χ).

χ mod N pv

Let us fix a Dirichlet character ψ on Y with values in A then there exists a unique distribution Φ = Φ1 ∗ψ Φ2 on Y with values in M = Mm r1 +r2 ,k1 +k2 (A) such that for all Dirichlet characters χ on Y with values in A one has Φ(χ) = Φ1 (χ) · Φ2 (ψχ−1 ) (Φ1 ∗ψ Φ2 is called the twisted convolution of Φ1 and Φ2 . Let us put M = pv , N = 4q0 C and take as Φ1 the theta distribution on Y defined on Dirichlet characters χ mod M by (ν)

Φ1 (χ) = θ2ξˆ (χ)|V (C)

(4.83)

0

and let us take as Φ± 2,s the normalized Siegel-Eisenstein distribution (4.25) defined on Dirichlet characters ω by ± Φ± 2,s (ω) = G (z, (s − k + m + ν)/2); k − (m/2) − ν, ω, N M ) =

(4.84)

±

(r ) ± G± Ψ2,s (ω) (s−k+m+ν)/2),k−(m/2)−ν (ω), N M ) = δ

(we use it for the Dirichlet character ω = ψχξ0 χ ¯ ) where M = M 2 , + Ψ2,s (ω) = G+ m/2+s,0 (ω), − (ω) = G− Ψ2,s m/2+1−s,s−(1/2) (ω)

are two holomorphic Siegel-Eisenstein series. ± Next define Φ± s = Φ1 ∗ψχξ0 Φ2,s (on basic open subsets one has Φ± s ((a)v ) =

 y∈Yv

ψχξ0 (y)Φ1 (ya)Φ± 2,s (y).

(4.85)

166

4 Admissible measures for standard L–functions

for any v such that ψχξ0 is defined modN pv ). We show that for all critical pairs (s, χ) as Theorem 4.13 the function Φ± s (χ) is an arithmetical nearly holomorphic Siegel modular form in m v m v (Γ Mm 1 (N p ), Q), and for an appropriate linear form Lf,α on Mr,k (Γ1 (N p ), Q) r,k ± ¯ the value Lf,α (Φ (χ)) coincides, up to a multiplicative con(defined over Q) s × ± stant in Q , with the corresponding value Ds,χ of the normalized distribution constructed in [Pa6], Chapter 3 (which produces the critical value ¯ D± (s, f, χ))). Let us consider (as in Proposition 3.3) the canonical projection operator m m α πα : Mm r,k (C) → Mr,k (Γ0 (N p), ψ; C)

attached to a non-zero eigenvalue α of U = Π + (p) acting on ) m v Mm Mm r,k (N, p; C) = r,k (Γ0 (N p ); ψ, C) : v≥1

for any g ∈ M = Mm r,k (C) put πα (g) = U −v [πα,1 U v (g)] (πα (g) is well defined if v is sufficiently large so that g ∈ Mm (N pv+1 )). Proposition 4.20. Let α ∈ Q ⊂ C be a non-zero eigenvalue of U = m Π + (p) on Mm r,k (C) associated with a cusp eigenform f ∈ Sk (Γ0 (N ), ψ) and let f0 = f0,α be a corresponding eigenfunction (f0 |U = αf0 ), let us define  f 0 = f0ρ |WN0 , f0ρ = aξ (f0 )q ξ . ξ

Then a) U ∗ = WN−1 U WN0 , the adjoint operator of U in the hermitian vector 0 space Sr,k (Γ1m (N p), ψ; C) with respect to the Petersson scalar product. b) One has f 0 |U ∗ = α ¯ f 0 , and for all ”good” Hecke operators T ∈ L(N pv ) 0 0 one has T f = λf (T )f . m c) The linear form f,α : g → f 0 , g on Mm r,k (Γ0 (N p), ψ; C) vanishes on Kerπα,1 , where m m m α πα,1 : Mm r,k (Γ0 (N p), ψC) → Mr,k (Γ0 (N p), C)

(the projector onto the α-primary subspace with the kernel Kerπα,0 = Im(U − αI)n ) hence f 0 , g = f 0 , πα,1 (g) = f 0 , πα (g) .

4.5 Algebraic linear forms on modular forms

167

d) If g ∈ Mm (N pv+1 , Q) = Mm (N ppv , Q) and α = 0, one has f 0 , g α = f 0 , πα (g) = α−v f 0 , g | U v

where

g α = πα (g) = (U α )−v πα,1 (g | U v ) ∈ Mm (N p)α

is the α-primary part of g (the image of g under the canonical projection πα ): e) Let us define Lf,α (g) =

f 0 , α−v g|U v N p f 0 , πα (g) N p = , 0 f , f0 N p f 0 , f0 N p

m hence Lf,α : Mm r,k (N, p; Q) → Q is the linear form Lf,α on Mr,k (N, p; C) which is defined over Q and factorizes through the finite dimensional projection m α m m πα : Mm r,k (N, p; C) → Mr,k (N, p; C) ⊂ Mr,k (Γ0 (N p); C);

There exists a unique Cp -linear form

f,α

α ∗ ∈ (Mm r,k (N0 , Cp ) ) such that

 f,α (ip (πα (g)))

= ip (Lf,α (g)) = ip

f 0 , α−v U v (g) N0 f 0 , f0 N0



(g ∈ Mm r,k (N, p; Q) with coefficients in Q). Proof of Proposition 4.20 a) See [Miy], Th. 4.5.5 for m = 1 (the general case is very similar). ¯ f0ρ | WN p = α ¯f 0. b) By definition, f 0 |U ∗ = f0ρ | WN p WN−1p U WN p = α c) For any function g1 = (U − αI)n g ∈ Kerπα,1 = Im(U − αI)n one has f 0 , g1 = f 0 , (U − αI)n g = (U ∗ − α ¯ I)f 0 , (U − αI)n−1 g = 0 hence for g1 = g − πα,1 (g) we get f 0 , g = f 0 , πα,1 (g) + (g − g α ) = f 0 , πα,1 (g) + f 0 , g1 = f 0 , πα,0 (g)

d) Let us use directly the equality (U ∗ )v f 0 = α ¯ v f 0 of b): αv · f 0 , g α = (U ∗ )v f 0 , U −v πα,1 (g | U v ) = f 0 , πα,1 (g | U v ) = f 0 , g | U v

by c) because g | U v ∈ Mm (N p). e) Note that Lf,α (f0 ) = 1, f0 ∈ Mm (N p; Q); consider the complex vector space

168

4 Admissible measures for standard L–functions

KerLf,α = f 0 ⊥ = {g ∈ Mm (N p; C) | f 0 , g = 0} which admits a Q-rational basis because it is stable by the action of for all ”good” Hecke operators T ∈ L(N pv ): f 0 , g = 0 =⇒ f 0 , g|T = f 0 |T ∗ , g = 0 and one obtains such a basis by the diagonalisation of the action of all the T (a commutative family of normal operators) and e) follows (see also Section 4.4.7 4.5.2 Evaluation of algebraic linear forms Our next task is to compute the value ± −1 Lf,α (Φ± )) = s (χ)) = Lf,α (Φ1 (χ) · Φ2,s (ψχξ0 χ

(4.86)

Lf,α (θ2ξˆ (χ)|V (C)G± (z, (s − k + m + ν)/2; k − (m/2) − ν, ψχξ0 χ, N M )) = (ν)

0

−1 ))|U 2v N p α−2v f 0 , (Φ1 (χ) · Φ± 2,s (ψχξ0 χ f 0 , f0 N p

α−2v f 0 , (θ2ξˆ (χ)|V (C)G± (z, s−k+m+ν ;k − 2 (ν)

0

m 2

− ν, ψχξ0 χ, N M ))|U 2v N p

f 0 , f0 N p

α−2v f0ρ |WN p , (θ2ξˆ (χ)|V (C)G± (z, s−k+m+ν ;k − 2 (ν)

0

m 2

− ν, ψχξ0 χ, N M ))|U 2v N p

f 0 , f0 N p

where χ mod N0 pv is a Dirichlet character of conductor Cχ = pv . Let us unfold the integral (4.21) using the relation (given in Proposition 3.3) of U 2v with the trace operator 2v+1

p g|U 2v = pvm(k−m−1) g|WN p2v+1 TrN Np

Lf,α (Φ± s (χ)) =

WN p :

(4.87)

mk vm(k−m−1)

(−1)

p α2v

2v+1

p −1 f0ρ , (Φ1 (χ) · Φ± ))|WN p2v+1 TrN 2,s (ψχ Np

f 0 , f

N p

0 N p

mk vm(k−m−1)

p α2v (ν) ρ f0 , (θ2ξˆ (χ)|V (C)G± (z, s−k+m+ν ;k − 2

=

(−1)

0

=

m 2

− ν, ψχξ0 χ, N p2v+1 ))|WN p2v+1 N p2v+1

f 0 , f0 N p

(−1)mk pvm(k−m−1) −1 · f0ρ , Φ1 (χ)|WN p2v+1 Φ± )|WN p2v+1 N p2v+1 2,s (ψχξ0 χ α2v f 0 , f0 N p

4.5 Algebraic linear forms on modular forms

169

Let us compute separately (ν)

Φ1 (χ)|WN p2v+1 = θ2ξˆ (χ)|V (C)WN p2v+1

(4.88)

−1 )|WN p2v+1 = Φ± 2,s (ψχξ0 χ

(4.89)

0

and

m s−k+m+ν ;k − − ν, ψχξ0 χ, N M )|WN p2v+1 2 2 using the action of the main involution on theta series, and using the definitions (4.16) and (4.17) of the normalized Siegel-Eisenstein series. ˜ M Cχ−2 so that the Now let us put N0 M = 4q0 Cχ2 N with N = N M0m−1 following transformation formula holds: G± (z,

θ2ξˆ (χ)|W (N0 M ) = N m(m+2ν)/4 θ2ξˆ (χ)|W (4q0 Cχ2 )V (N ) = (ν)

(ν)

0

0

GCχ (1m , χ)

˜ (N M0m−1 M )m(m+2ν)/4 −1/2

·det(q0

−m(m+2ν)/2 (iCχ )m2 /2 Cχ

(4.90)

·

ξ0 )m/2+ν θ2ξ0 (χ)|V ¯ (N ) (ν)

Recall the definition of the normalized Siegel-Eisenstein series of the level N0 M = N p2v+1 : G− (z, s) = Γ ((k + 2s − (m/2) + µ)/2)−1 G∗ (z, s), iµ π κ−k−2s G∗ (z, s) = Γ ((1 − k − 2s + (m/2) + µ)/2)

G+ (z, s) =

2iµ Γ (k + 2s − (m/2)) cos(π(k + 2s − (m/2) − µ)/2) − G (z, s). (2π)k+2s−(m/2) where µ = ε((m/2) + k), and G∗ (z, s) = G∗ (z, s; k, χ, N0 M ) = (N0 M )m(k+s) Γ(k, s)LN0 (k + 2s, χ)




(4.91)

[m/2]

LN0 (2k + 4s − 2i, χ2 )E ∗ (N0 M z, s),

i=1

with E ∗ (N0 M z, s) =

−1

E(−(N0 M z)

i

2

π

−k

, s)det(N0 M z)

Γ(k, s) = mk −m(k+1) −m(s+k)

(4.92)

" ×

−km/2

= (N0 M )

Γm (k + s)Γ (s + (k − Γm (k + s),

m 2



E|W (N0 M ),

+ µ)/2), if m is even; otherwise.

4 Admissible measures for standard L–functions

170

and LN (s, χ) =




(1 − χ(q)q −s )−1

q
N

is the Dirichlet series associated to the character χ and µ = ε(m/2 + k), with ε(r) ∈ {0, 1} such that ε(r) ≡ r (mod 2) for any integer r ∈ N. Let us introduce the Gamma-factors Γ ± (k, s) by the equality G± (z, s) = G± (z, s; k, χ, N0 M )

(4.93)

= (N0 M )m(k+s) Γ ± (k, s)


[m/2]

LN0 (k + 2s, χ)

LN0 (2k + 4s − 2i, χ2 )E ∗ (N0 M z, s)

i=1

= (N0 M )m((k/2)+s) Γ ± (k, s)


[m/2]

LN0 (k + 2s, χ)

LN0 (2k + 4s − 2i, χ2 )E|W (N0 M ),

i=1

so that the change of variables s := (s − k + m + ν)/2, k := k − (m/2) − ν in (4.93) gives k + 2s := s + (m/2) = (s + 2m)/2 and  m s−k+m+ν ± G ;k − − ν, ψχξ0 χ, N0 M |W (N0 M ) z, 2 2 s−k+m+ν m − ν, ) 2 2

= (N0 M )m((s+2m)/4 Γ ± (k −


[m/2]

LN0 ((2s + m)/2, ψχξ0 χ)

LN0 (2s + m − 2i, χ2 ψ 2 )

(4.94)

i=1

E(z, (s − k + m + ν)/2; k − (m/2) − ν, ψχξ0 χ, N0 M ). Let us substitute it to (4.46) with χ replaced by the Dirichlet character χ 2a(ξ0 )(detξ0 )−(s+k−1+ν)/2 (4π)−ms Γm ((s + k − 1 + ν)/2) R(s, f, χ)

(4.95)

(ν)

= f0ρ , θ2ξ0 (z; χ)E(z, (s − k + m + ν)/2; k − (m/2) − ν, ψχξ0 χ) N M 2 q0 , where a(ξ0 )R(s, f, χ) = 2−1 det ξ0

(s+k−1+ν)/2

(4.96) (ν)

L((s + k − 1 + ν)/2, f0 , θ2ξ0 (z; χ)).

The function R(s, f, χ) is related to D(s, f, χ) by the following equality


(m/2)−1

D(s, f, χ) = L(s + (m/2), ψχξ0 χ)

i=0

L(2s + 2i, ψ 2χ2 ψ 2 )R(s, f, χ) (4.97)

4.5 Algebraic linear forms on modular forms

171

therefore (4.95) takes the form 2a(ξ0 )(detξ0 )−(s+k−1+ν)/2 (4π)−ms × Γm ((s+ k − 1 + ν)/2)D(s, f, χ) = (4.98)


(m/2)−1

L(s + (m/2), ψχξ0 χ)

L(2s + 2i, ψ 2 χ2 ψ 2 )

i=0 (ν)

f0ρ , θ2ξ0 (z; χ)E(z, (s − k + m + ν)/2; k − (m/2) − ν, ψχξ0 χ) N M 2 q0 , We see that the integral in (4.87) transforms to the following −1 )|WN p2v+1 N p2v+1 = f0ρ , Φ1 (χ)|WN p2v+1 Φ± 2,s (ψχ

(4.99)

(ν)

f0ρ , θ2ξˆ (χ)|V (C)WN0 M  0

m s−k+m+ν ;k − − ν, ψχξ0 χ, N0 M )|WN0 M  N0 M  2 2 and (4.99) becomes G± (z,

−1 f0ρ , Φ1 (χ)|WN p2v+1 Φ± )|WN p2v+1 N p2v+1 = 2,s (ψχξ0 χ ˜ M )m(m+2ν)/4 (N M0m−1

GCχ (1m , χ) −m(m+2ν)/2 (iCχ )m2 /2 Cχ

(N0 M )m((s+2m)/4 Γ ± (k −

−1/2

det(q0

(4.100) ξ0 )m/2+ν

s−k+m+ν m − ν, ) 2 2


[m/2]

LN0 ((2s + m)/2, ψχξ0 χ)

LN0 (2s + m − 2i, χ2 ψ 2 )

i=1

m s−k+m+ν ;k − − ν, ψχξ0 χ, N0 M ) N0 M  2 2 Let us now compare the value of the linear form (4.87) expliciely given by (4.100), with the integral representation (4.98) for the standard L-function: f0ρ , θ2ξ0 (χ)|V ¯ (N )E(z, (ν)

Lf,α (Φ± s (χ)) = mk vm(k−m−1)

(4.101)

=

(−1) p α2v f 0 , f0 N p

=

(−1)mk pvm(k−m−1) × α2v f 0 , f0 N p

−1 · f0ρ , Φ1 (χ)|WN p2v+1 Φ± )|WN p2v+1 N p2v+1 2,s (ψχξ0 χ

˜ ×(N M0m−1 M )m(m+2ν)/4

GCχ (1m , χ)

−1/2

det(q0 −m(m+2ν)/2 (iCχ )m2 /2 Cχ s−k+m+ν m − ν, )× ×(N0 M )m((s+2m)/4 Γ ± (k − 2 2

ξ0 )m/2+ν ×

4 Admissible measures for standard L–functions

172




[m/2]

×LN0 ((2s + m)/2, ψχξ0 χ)

LN0 (2s + m − 2i, χ2 ψ 2 )

i=1

m s−k+m+ν ;k − − ν, ψχξ0 χ, N0 M ) N0 M  2 2 Let us remove from the Euler products D(s, f, χ) and L(s, χ) all the factors corresponding to the prime divisors of N0 : f0ρ , θ2ξ0 (χ)|V ¯ (N )E(z, (ν)

2a(ξ0 )(detξ0 )−(s+k−1+ν)/2 (4π)−ms × Γm ((s + k − 1 + ν)/2)DN0 (s, f, χ) =


(m/2)−1

LN0 (s + (m/2), ψχξ0 χ)

LN0 (2s + 2i, ψ 2 χ2 ψ 2 )×

(4.102)

i=0 (ν)

× f0ρ , θ2ξ0 (z; χ)E(z, (s − k + m + ν)/2; k − (m/2) − ν, ψχξ0 χ) N M 2 q0 . Let χ be a non-trivial Dirichlet character modulo N pv of conductor Cχ such that Cχ |pv ). Consider fhe complex valued distributions Lf,α (Φ± s (χ)) defined for the critical pairs (s, χ) as in Theorem 4.13 by ± −1 Lf,α (Φ± )). s (χ)) = Lf,α (Φ1 (χ) · Φ2,s (ψχξ0 χ

Proposition 4.21. Assume that the cusp eigenform f ∈ Sm k (C, ψ) satisfies the condition a(ξ0 ) = 1 for some ξ0 ∈ Am , ξ0 > 0. Then a) For all integers s satisfying 1 ≤ s ≤ k − δ − m and s = 1 if the character χ2 ψ 2 is trivial we have that the value Lf,α (Φ+ s (χ)) ∈ K = Q(f, λf , ψ, χ), where K = Q(f, λf , ψ, χ) denotes the field generated by Fourier coefficients of f , by the eigenvalues λf (X) of Hecke operators X on f , and by the values of the characters χ and ψ. Moreover, Lf,α (Φ+ s (χ)) = 0 for s ≡ δ(mod2). b) For all integers s satisfying 1 − k + δ + m ≤ s ≤ 0 we have that Lf,α (Φ− s (χ)) ∈ K, and

Lf,α (Φ− s (χ)) = 0 for s ≡ δ(mod2), s ∈ Z, s ≤ 0.

c) Suppose moreover that N is divisible by all of the prime divisors of 2det(2ξ0 ) then there is the following equality Lf,α (Φ± s (χ)) = t±

GCχ (1m , χ) Cχm(2s+2k−2−m)/4 ± D (s, f, χ), ¯ α0 (Cχ )2 (iCχ )m2 /2

4.5 Algebraic linear forms on modular forms

173

×

where t± is a multiplicative constant in Q , independent of the critical pair (s, χ), ¯ denotes the normalized standard L-function as in Theorem 4.13, D± (s, f, χ)  χ(detξ )em (tξξ /Cχ ), GCχ (1m , χ) = ξ  ∈Mm (Z)

(mod Cχ )

is the Gauss sum of degree m of χ. Proof of Proposition 4.21 It is directly implied by the equalities (4.101) and (4.102) taking into account Theorem 4.13 on the critical calues of the standard L-function.

4.6 Congruences between nearly holomorphic modular forms and proof of the Main theorem. 4.6.1 Regularized distributions in Siegel modular forms. We now outline the method of the proof of Main Theorem and show the existence of the h–admissible measures attached to the standard zeta functions D(s, f, χ). This proof is based on congruences for the Fourier coefficients of nearly holomorphic Siegel modular forms. Consider again the profinite group Yv , where Yv = (Z/N pv Z)× . Y = lim ← v

and consider fhe complex valued distributions Lf,α (Φ± s (χ)) defined for the critical pairs (s, χ) as in Theorem 4.13 and Proposition 4.21 by ± −1 )). Lf,α (Φ± s (χ)) = Lf,α (Φ1 (χ) · Φ2,s (ψχξ0 χ

(4.103)

The proof is based on a regularization of the distributions Lf,α (Φ+ s (χ)) (for 1 ≤ s ≤ k − ν − m) and Lf,α (Φ− (χ)) (for 1 − k + ν + m ≤ s ≤ 0) (with values s in Q). Recall that by Proposition 4.20 the linear form Lf,α factorizes through the canonical projection πα : ± Lf,α (Φ± s (χ)) = Lf,α (πα (Φs (χ))). ˜ . Consider a positive integer c such that (c, N0 ) = 1 and N0 = 4q0 N M0m−1 One supposes always that m is even and that k > 2m + 2. Hence there exist c− regularized distributions Φc+ s , Φs on Y with values in the module ) m v Mm Mm r,k (Q) = r,k (Γ0 (N p ), ψ; Q) : v≥1

of arithmetical nearly holomorphic Siegel modular forms. These regularized distributions are uniquely determined by their values as follows:   s Φc+ ¯ −1 1 − (χψ) (4.104) ¯ 2 (c)c−2s Φ+ s (χ) = Cχψ ¯ G(χψ) s (χ)
 ¯ 0 (q)q s−1 )(1 − (χψ) (1 − (χψ) ¯ 0 (q)q −s )−1 , q|N0

  ¯ 2 (c)c2(s−1) Φ− (χ). (χ) = 1 − ( χ ¯ ψ) Φc− s s ¯

Next consider the Cp -linear forms Dc± : Ck−m (Y ; Cp ) → Cp defined on the local monomials xjp for j = 0, 1, · · · , k − m − 1 by:    xjp Dc+ = Lf,α (Φc+ ) = Lf,α (πα (Φc+ (4.105) j+1 j+1 )), a+(M)

a+(M)

a+(M)

4.6 Congruences and proof of the Main theorem



 xjp Dc− = a+(M)

175



a+(M)

Lf,α (Φc− −j ) =

a+(M)

Lf,α (πα (Φc− −j )).

k−m c± It follows that there exist the Cp -linear maps Φ (Y ; Cp ) → Mm α :C r,k (Cp ) defined on the local monomials xjp for j = 0, 1, · · · , k − m − 1 by:   j c± xp Φα = πα (Φc+ (4.106) j+1 )), a+(M)



a+(M)

c± xjp Φ α =

a+(M)



a+(M)

πα (Φc− −j ).

In order to prove that Cp -linear forms Dc± : Ck−m (Y ; Cp ) → Cp define in fact certain admissible measures we observe that c± ) Dc± = Lf,α (Φ α c± where the linear forms Φ α depend on the choice of α (but not on the choice of f ), and in fact they take values in a finite dimensional Cp -vector space (the image of the finite dimensional projection: m α m m πα : Mm r,k (N, p; C) → Mr,k (N, p; C) ⊂ Mr,k (Γ1 (N p); C),

see Proposition 4.20 (d)) hence it suffuces to prove the admissibility of the c± linear maps Φ α with values in spaces of arithmetical nearly holomorphic Siegel modular forms. For this purpose we use a general result on admissible measures with values in nearly holomorphic Siegel modular forms given in Theorem 4.8. We have for the positive critical values   r   r c+ (x − a)r dDc+ = dDj+1 (4.107) (−a)r−j j a+(M) a+(M) j=0 r    r 1 c+ (−a)r−j χ−1 (a)Dj+1 (χ) j ϕ(M ) j=0 χmodM r    r 1 = χ−1 (a)Lf,α (Φc+ (−a)r−j j+1 (χ)). j ϕ(M ) j=0

=

χmodM

similarly for D 

c−

we have

(x − a)r dDc− = a+(M)

=

r   r j=0

j

 r   r c− dD−j (−a)r−j j a+(M) j=0

(−a)r−j

 1 c− χ−1 (a)D−j (χ) ϕ(M ) χmodM

r    r 1 = χ−1 (a)Lf,α (Φc− (−a)r−j −j (χ)). j ϕ(M ) j=0 χmodM

(4.108)

4 Admissible measures for standard L–functions

176

4.6.2 Sufficient conditions for admissibility of measures with values in nearly holomorphic Siegel modular forms. In order to prove the admissibility property Dc± of we need to verify the following growth condition for Dc :        r c±  p α0 (p) , (4.109) sup  (x − ap ) dD  = o |M |r−2ord p  a∈Y  a+(m) p

for r = 0, 1, · · · , k − m − 1, which will be implied by         r−2ordp α0 (p) c± , sup  (x − ap )r dΦ α  = o |M |p  a∈Y  a+(m)

(4.110)

p

Let us use a general result giving a sufficient condition for the admissibility of measures with values in nearly holomorphic Siegel modular forms (given in Theorem 4.8): with κ = 2, h = [2ordp α0 (p)] + 1. Then we need to prove that    j    2v      j  ≤ Cp−vj U (4.111) (−a0p )j−j Φc+ 1+j  (a)v )   j   j  =0 p

and

    j    2v      j 0 j−j  c− U  ≤ Cp−vj (a) ) Φ ) (−a v p −j    j   j  =0

(4.112)

p

for all j = 0, 1, · · · , k − m − 1. 4.6.3 Fourier expansions of distributions with values in nearly holomorphic Siegel modular forms. Let us use that fact that (r + )

δm/2+s G+ ¯ h0 ) = G+ ¯ h0 ), k (z, s , χψχ m/2+s (z, 0, χψχ

(4.113)

(r − )

δm/2+1−s G− ¯ h0 ) = G− ¯ h0 ), k (z, s , χψχ m/2+1−s (z, s − 1/2, χψχ

where k = k−m/2−ν, s = (s−k+m+ν)/2, r+ = r− = (k−m−ν−j−1)/2 ∈ N and if j + 1 ≡ δ (mod 2) (otherwise the series vanishes). The Fourier expansions of these series are given by  ¯ ξ0 ) = b+ (ξ)em (ξz), (4.114) G+ m/2+s (z, 0, χψχ ξ∈Cm

4.6 Congruences and proof of the Main theorem

G− ¯ψχξ0 ) = m/2+1−s (z, s − 1/2, χ



177

b− (ξ)em (ξz),

ξ∈Cm

where b+ (ξ) = 2−mκ det(ξ)m/2+s−κ L+ N1 (s, ωχξ )M (ξ, ω, m/2 + s),

(4.115)

b− (ξ) = 2−mκ L− N1 (s, ωχξ )M (ξ, ω, m/2 + s), ¯ is a Dirichlet character, ω = ψχξ0 χ M (ξ, ω, m/2 + s) =




Mq (ξ, ω(q)q −m/2−s ),

q∈P (ξ)

is a finite Euler product such that Mq (ξ, t) ∈ Z[t]. The normalized Dirichlet L-functions are given by δ (1−2s)/2 L+ N1 (s, ωχξ ) = i π

Γ (s + δ)/2) LN (s, ωχξ ), Γ ((1 − s + δ)/2 1

(4.116)

L− N1 (s, ωχξ ) = LN1 (s, ωχξ ). Let us use these Fourier expansions in order to compute the following nearly holmorphic Fourier expansion ± Φ± ¯ s (χ) = Φ1 (χ)Φ2,s (ψχξ0 χ)  ± ± (ω) = v ± (R, ξ, s, χ)em (ξz), Φ1 (χ)δ (r ) Ψ2,s



Φ± s (χ)|U (M ) =

=



ξ∈Cm

v ± (M R, M ξ, s, χ)em (ξz)

(4.117)

ξ∈Cm



d± (s, M R, ξ1 , ξ2 )em (ξz)

ξ∈Cm N ξˆ0 [ξ1 ]+ξ2 =M  ξ

where M = p2v , d+ (s, M R, ξ1 , ξ2 ) = (q0 N )−m(s−1+k−m−ν)/2 det(ξ0 )(s−1+k−m−ν)/2 (4.118) χ(det(ξ1 ))det(ξ1 )ν det(ξ2 )s−1/2 L+ N1 (s, ωχξ2 ) M (ξ2 , ω, s + m/2)Q(M R, ξ2 ; r+ , β + ), under the condition s ≡ δ (mod 2) (otherwise d+ (s, M Rξ 1 , ξ2 ) = 0). Recall the notation ω = χψχ ¯ ξ0 , β + = (1−s−k+ν +m)/2 and r+ = (k−ν −m−s)/2. − The coefficients d (s, M R, ξ1 , ξ2 ) are given by d− (s, M R, ξ1 , ξ2 ) = (q0 N )−m(s−1+k−m−ν)/2 det(ξ0 )(s−1+k−m−ν)/2 (4.119) χM (det(ξ1 ))det(ξ1 )ν L− N1 (s, ωχξ2 ) M (ξ2 , ω, s + m/2)Q(M R, ξ2 ; r− , β − ), under the condition s ≡ δ (mod 2) (otherwise d− (s, ξ1 , ξ2 ) = 0. Recall also that β − = (s − k + ν + m)/2 and that r− = (k − ν − m + s − 1)/2.

4 Admissible measures for standard L–functions

178

4.6.4 Fourier expansions of regularized distributions. Let us apply the above formulas for the normalized distributions:  Φc± v c± (M R, M ξ, s, χ)em (ξz) (4.120) s (χ)|U (M ) = ξ∈Cm





=

dc± (s, M R, ξ1 , ξ2 )em (ξz),

ξ∈Cm N ξˆ0 [ξ1 ]+ξ2 =M  ξ

2v

where M = p , dc+ (s, M R, ξ1 , ξ2 ) = (q0 N )−m(s−1+k−m−ν)/2 det(ξ0 )(s−1+k−m−ν)/2 (4.121) χ(det(ξ1 ))det(ξ1 )ν det(ξ2 )s−1/2 Q(M R, ξ2 ; r+ , β + ) M (ξ2 , χψχ ¯ ξ0 , s + m/2)L+ ¯ ξ0 χξ2 ) N1 (s, χψχ   s s−1
¯ Cχψ (1 − (χψ)0 (q)q )  ¯ (1 − (χψ) ¯ 2 (c)c−2s )  . G(χψ) ¯ (1 − (χψ) ¯ 0 (q)q −s ) q|N0

c−

The coefficients d c−

d



(s, M R, ξ1 , ξ2 ) are given by



(s, M R, ξ1 , ξ2 ) = (q0 N )−m(s−1+k−m−ν)/2 det(ξ0 )(s−1+k−m−ν)/2 (4.122) ¯ 2 (c)c−2s )L− χM (det(ξ1 ))det(ξ1 )ν (1 − (χψ) N1 (s, ωχξ2 ) M (ξ2 , ω, s + m/2)Q(M R, ξ2 ; r− , β − )

under the condition s ≡ δ (mod 2) (otherwise d+ (s, M Rξ 1 , ξ2 ) = 0). Recall the notation ω = χψχ ¯ ξ0 , β + = (1−s−k+ν +m)/2 and r+ = (k−ν −m−s)/2, − β = (s − k + ν + m)/2 and that r− = (k − ν − m + s − 1)/2. 4.6.5 Main congruences for the Fourier expansions of regularized distributions. Let us use the orthogonality relations for Dirichlet characters in order to prove (4.111) and (4.112) using the Fourier expansions (4.120):     j     j  1 0 j−j −1 c+  χ (a)v (M R, M ξ, 1 + j , χ) (−ap )  j ϕ(M )   j =0

≤ Cp and

χmodM

−vj

p

(4.123)

        j  j 1 0 j−j −1 c−  χ (a)v (M R, M ξ, −j , χ) (−ap )  ϕ(M )  j  =0 j χmodM ≤ Cp−vj

for all j = 0, 1, · · · , k − m − 1, M = pv , M = p2v .

p

(4.124)

4.6 Congruences and proof of the Main theorem

179

4.6.6 Kummer congruences and Mazur’s measure. Let S be a finite set of primes containing p and all the prime divisors of N . Consider the Cp –adic analytic Lie groups × XS = Homcontin (Z× S , Cp ),

where

X = XN = Homcontin (Y, C× p)

(Z/QZ)× Z× S = lim ←− Q

Z× S

The projection homomorphism → Y induces the analytic embedding X → XS . Recall the notion of the h–admissible measures on Z× S and properties of their Mellin transform. These Mellin transform are certain p–adic analytic functions on the Cp –analytic Lie group XS . The canonical Cp –analytic structure on XS is obtained by shifts from the obvious Cp –analytic structure on the subgroup × Homcontin (Z× p , Cp ) ⊂ XS .

We view the elements of finite order χ ∈ XStors as Dirichlet characters whose conductor C(χ) contain only primes in S, by means of the decomposition × i

∞ × χ : Z× S → Q −→ C ,

where i∞ is the fixed embedding. The characters χ ∈ XStors form a discrete subgroup XStors ⊂ XS . We use also the following natural homomorphism × × xp : Z× S → Zp → Cp ,

xp ∈ XS .

so that all integers k ∈ Z may be viewed as characters of the type xkp : y → y k . Recall that a p–adic measure on Z× S may be viewed as a bounded Cp –linear form µ on the space C(Z× ) of all continuous Cp –valued functions S  ϕ → µ(ϕ) = dµ ∈ Cp , ϕ ∈ C(Z× S ), Z× S

which is uniquely determined by its restriction to the subspace C1 (Z× S ) of locally constant functions. We denote by µ(a + (Q)) the value of µ on the characteristic function of the set & % × a + (Q) = x ∈ Z× S | x ≡ a mod Q ⊂ ZS . The Mellin transform Lµ of µ is a bounded analytic function  Lµ : XS → Cp , Lµ (χ) = dµ ∈ Cp , χ ∈ XS , Z× S

on XS , which is uniquely determined by its values Lµ (χ) for the characters χ ∈ XStors .

4 Admissible measures for standard L–functions

180

Non–Archimedean integration and admissible measures A more delicate notion of an h–admissible measure was introduced by Amice–Velu and Viˇsik (see [Am-Ve], [Vi1]). Let Ch (Z× S ) denote the space of Cp –valued functions which can be locally represented by polynomials of degree less than a natural number h of the variable xp ∈ XS introduced above. Definition 4.22. A Cp –linear form µ : Ch (Z× S ) → Cp is called h– admissible measure if for all a ∈ Z× S and for all r = 0, 1, · · · , h − 1 the following growth condition is satisfied  | sup (xp − ap )r dµ |= o(| Q |r−h ) p a∈Z× S

a+(Q)

We know (essentially due to Amice–Velu and Viˇsik) that each h–admissible measure can be uniquely extended to a linear form on the Cp –space of all locally analytic functions so that one can associate to its Mellin transform  Lµ : XS → Cp , Lµ (χ) = χdµ ∈ Cp , χ ∈ XS , Z× S

which is a Cp –analytic function on XS of the type o(log xhp ). Moreover, the measure µ is uniquely determined by the special values of the type Lµ (χxrp ) (χ ∈ XStors , r = 0, · · · , h − 1). Let ω mod such that (A, M0 ) = 1 $ A be a fixed primitive Dirichlet character $ with M0 = q∈S q. Put S = S ∪ S(A), M = q∈S q. Then for any positive integer c with (c, M ) = 1, c > 1 there exist Cp -measures µ+ (c, ω), µ− (c, ω) on Z× S which are uniquely determined by the following conditions: for s ∈ Z, s > 0   Cωχ¯ −1 s + × χ xp dµ (c, ω) = (1 − χω(c)c−s ) ip G(ω χ) ¯ Z× S
% & (1 − χω(q)q s−1 )/(1 − χω(q)q −s ) L+ × (4.125) M0 (s, χω), q∈S\S(χ)

and for s ∈ Z, s ≤ 0  i−1 p

Z× S

χ xsp

 −

dµ (c, ω)

= (1 − χω(c)cs−1 )L− M0 (s, χ, ω),

(4.126)

2iδ Γ (s) cos(π(s − δ)/2) , (2π)s

(4.127)

where ¯ L+ M0 (s, χω) = LM (s, χω)

4.6 Congruences and proof of the Main theorem

181

L+ ¯ M0 (s, χω) = LM (s, χω) are the normalized Dirichlet L-functions with δ = 0, 1, (−1)δ = χω(−1). The functions (4.125) and (4.126) satisfy the following functional equation LM0 (1 − s, χω) =
% & (1 − χω(q)q s−1 )/(1 − χω(q)q −s ) L+ M0 (s, χω).

(4.128)

q∈S\S(χ)

the Recall that by the definition of the S-adic Mazur measure µc on Z× S distributions (4.125) and (4.126) are given by   − x dµ (c, ω) = x−1 ω dµc , Z× S



Z× S

 Z× S

dµ+ (c, ω) =

Z×

−1 x x−1 dµc , p ω

S

where x ∈ XS , and XS is regarded as a subgroup of XS . 4.6.7 Reduction of the Main congruence to congruences for partial sums. Recall from (4.2.4) that  (r) c(ξ)Q(R, ξ)q ξ ∈ Q(f )[[q Bm ]][Ri,j ]i,j=1,··· ,m , δk f (z) = ξ∈Bm

where Q(f ) is the subfield of C generated by the Fourier coefficients of f , L runs over all the multi-indices 0 ≤ l1 ≤ · · · ≤ lt ≤ m as above, r    r Q(R, ξ) = Q(R, ξ; k, r) = RL (κ − k − r)QL (R, ξ), det(ξ)r−t t t=0 |L|≤mt−t

    QL (R, ξ) = tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) . Hence combining all the above remarks we deduce that the coefficients ip (v c± (M R, M ξ, s, χM )) are linear combinations with coefficients independent of M , M and s of the following expressions: B+ = χM (det(ξ1 ))det(ξ1 )ν det(ξ2 )s−1/2 det(ξ2 )r

(4.129) +

r+  +  t=0

r t

 |L|≥t

cL (β + )

182

4 Admissible measures for standard L–functions

 r+ +s−1/2 χψ(τ ¯ ) QL (M R, ξ2 ) (q0 N )−m det(ξ0 ) s−1/2 τ

 Z× S

r  −   r

χxsp dµc+ (ω)

−

B− = χ ¯M (det(ξ1 ))det(ξ1 )ν det(ξ2 )r

−

t

t=0

 r− QL (M R, ξ2 ) (q0 N )−m det(ξ0 )

cL (β − )

(4.130)

|L|≥t



χx ¯ sp dµc− (ω),

Z× S

with ω = ψχξ0 χξ2 and det(2ξ2 )det(2ξ0 ) = τ a2 for a, τ √∈ Z such that τ is square free. Hence Cχω = τ Cχψ and G(χω) ¯ = χψ(τ ¯ ) τ G(χψ) ¯ and us¯ ¯ ing N ξˆ0 [ξ1 ] + ξ2 = M ξ we show the following congruence det(ξ2 )det(ξ0 ) ≡ (q0 N )m det(ξ1 )2 (mod M ). Moreover m is even, and one deduces then that det(ξ2 )det(ξ0 ) is a square modulo q for all q|N , then we simply obtain (6.28), (6.29) from (6.23) and (6.24). Concider M and M large enough such that M 2 |M and q0m |M and recall that N ∈ Z× S hence on can find det(ξ2 )(q0 N )−m det(ξ0 ) ≡ det(ξ1 )2

(mod M ).

Thus the coefficients χM (det(ξ1 )) do not vanish only if det(ξ1 ) ∈ Z× S . We see then that det(ξ2 )(q0 N )−m det(ξ0 ) is a square in Z× and we put S 0 λ = λ+ = τ −1 det(ξ2 )(q0 N )−m det(ξ0 ) ∈ Z× S in the case +, 0 λ = λ− = det(ξ2 )(q0 N )−m det(ξ0 ) ∈ Z× S in the case −. It follows that the expressions B + and B − become B + = ψ(τ )τ k−m−ν−1/2 det(ξ1 )ν χ(λ)λs−1+k−m−ν 

cL (β + )QL (M R, ξ2 )

ν

s−1+k−m−ν

B = det(ξ1 ) χ(λ)λ ¯

QL (M R, ξ2 )

Z× S



Z× S

(4.131)

cL (β − )

(4.132)

χxsp dµc+ (ω),

[(k−m−1)/2]  −  t=0

r t

t=0



|L|≥t −

[(k−m−1)/2]  + 

r t

 |L|≥t

χx ¯ sp dµc− (ω).

In this way the number A± is a linear combination with coefficients independent on M , M and j of type A+ =

(4.133) 



t=0

|L|≥t

[(k−m−1)/2]

det(ξ1 )ν λk−m−ν−1−l

QL (M R, ξ2 )

l  ±  rj j=0

t

4.6 Congruences and proof of the Main theorem

cL (βj± )lj(−aλ−1 )l−j

 Z× S

183

 1 χ(xa−1 )xj+1 dµc± (ω). ϕ(M ) χmodM

Note that cL (βj+ ) is a polynomial of degree |L| in βj+ hence of degree |L| in j  + because βj+ = (−j − k + ν + m)/2 and rtj is a polynomial of degree t in rj+ thus of degree tin j in view of rj+ = (k − m − ν − j − 1)/2. One can therefore write

 + t+|L|  (j + n + 1)! rj . µn cL (βj+ ) = t (j + 1)! n=0

Here the coefficients µn are certain fixed rational numbers. Using again the orthogonality relations we see that the sum over j, denoted by C + , takes the form 



t+|L|

C+ =

x≡aλ−1 modM n=0

µn

l   l (j + n + 1)! j+1 + x dµ (c, .) (−aλ−1 )l−j j (j + 1)! j=0 9 :; < n   −n ∂ n+1 −1 l (x − aλ ) x x ∂xn

Therefore (x − aλ−1 )l ≡ 0 (mod M l ) giving the congruence 2νm,k C + ≡ 0 (mod M l−n ) ≡ 0

(mod M l−t−|L| ).

(4.134)

4.6.8 Proof of the Main congruence. Now the expression (6.31) transforms to 



t=0

|L|≥t

[(k−m−1)/2]

A+ = det(ξ1 )ν λk−m−ν−1+l

QL (M R, ξ2 )C + ,

(4.135)

where QL (M R, ξ2 ) is a homogeneuos polynomial of degree |L| in the variables M Rij implying the congruence QL (M R, ξ2 ) ≡ 0 We deduce that

(mod M |L| ) ≡ 0 (mod M 2|L| ).

  . A+ = o |M |2|L|+l−t−|L| p

(4.136) (4.137)

On the other hand we know from the description of the polynomial r    r Q(R, ξ) = Q(R, ξ; k, r) = RL (κ − k − r)QL (R, ξ)), det(ξ)r−t t t=0 |L|≤mt−t

    QL (R, ξ) = tr tρm−l1 (R)ρ
l1 (ξ) · . . . · tr tρm−lt (R)ρ
lt (ξ) )

4 Admissible measures for standard L–functions

184

that |L| ≥ t so we obtain finally the congruence   A+ = o |M |lp .

(4.138)

This shows that Dc+ is an h-admissible measure by the sufficient condition (4.111) (implying (4.109)). ¯ Thus In the case Dc− the proof is similar but one needs to replace χ by χ. − A is also a linear combination with coefficients independent of M , M and j of type 

[(k−m−1)/2] −

ν k−m−ν−1

A = det(ξ1 ) λ

t=0

  l (−aλ)l−j cL (βj− ) j Z× S(N

0)





QL (M R, ξ2 )

|L|≥t

l  −  r j

j=0

t

(4.139)

 1 −j χ(aλx)x ¯ dµ− (c). ϕ(M ) χmodM

Let us use again the orthogonality for the Dirichlet characters and the fact  − that rtj cL (βj− ) is a polynomial of degree t + |L| en j. One represents this polynomial in the form  − t+|L|  rj (j + 1 + n)! . µn cL (βj− ) = t (j + 1)! n=0

(4.140)

One obtains the following expression for the sum C − over j in the formula giving A− : l   l (j + 1 + n)! −j − x dµ (c, .), C = µn (−aλ)l−j j (j + 1)! x−1 ≡aλmodM n=0 j=0 9 :; < n   ∂ (aλ − x1 )l xn1 n xn+1 ∂x1 1 (4.141) where x1 = x−1 . Then we use the same argument as for A+ in (4.138) in order to prove that     A− = o |M |lp = o |M |l−h . (4.142) p −





t+|L|

It follows from (4.142) that the linear form Dc− is an admissible measure. Theorem 4.23. Let f ∈ Sm k (N, ψ) be a Siegel cusp eigenform of even degree m and of weight k > 2m + 2 satisfying the nonvanishing condition (*) of Main Theorem. For a positive integer c > 1, (c, N p) = 1 there exist two Cp -analytic functions Lc± : XS → Cp such that: (i) For all pairs (s, χ) with s ∈ Z, 0 ≤ s ≤ k − m − 2 and a primitive non-trivial Dirichlet character χ ∈ XStors of conductor Cχ = pv one has

4.6 Congruences and proof of the Main theorem m(2s+2k−m)/4

• Lc+ (χxsp ) =

GCχ (1m , χ)Cχ (iCχ )m2 /2 α0 (Cχ2 )
q|N

s+1 Cχψ ¯

G(χψ) ¯

185

(1 − (χψ) ¯ 2 (c)c−2s−2 )

¯ 0 (q)q s ) D+ (s + 1, f, χ) (1 − (χψ) ¯ . −s−1 (1 − (χψ) ¯ 0 (q)q ) f, f N m(2−2s+2k−m)/4

• Lc− (χxsp ) =

¯ χ GCχ (1m , χ)C (iCχ )m2 /2 α0 (Cχ2 )

¯ 2 (c)c−2s−2 ) (1 − (χ ¯ψ)

D− (−s, f, χ) . f, f N ¯ viewed as elements of the Tate field Cp These values are algebraic numbers Q ¯ → Cp ). On the other hand the Cp -analytic (via the fixed embedding ip : Q functions vanish at the following points Lc+ (χxsp ) = Lc− (χxsp ) = 0 if s ≡ 0 (mod 2). (ii) If ordp (α0 (p)) = 0 (i.e. f is p-ordinary), then the holomorphic functions Lc+ et Lc− are bounded Cp -analytic functions. (iii) In the general case the functions Lc+ and Lc− are of type o(log(xhp )) where h = [2ordp (α0 (p))] + 1 and they are Mellin transforms of certain hadmissible measures. (iv) If h ≤ k − m − 1, then the functions Lc± are uniquely determined by (i). 4.6.9 Proof of Theorem 4.23 The assertion (ii) (i.e. the case ordp (α0 (p)) = 0) which was proved in [Pa6], also follows easily from the Main congruence (4.111) and (4.112) taking into account the description of the canonical projection operator in Proposition c± ) let us develop the definition of Φ α using the 3.3: for α = α0 (p), Φα = πα (Φ binomial formula: 

α = (yp − ap )j dΦ

(a)v

j    j −vκ · (−ap )j−j Φα j  ((a)v ) = α j 

j =0



αvκ

 j       j · U −vκ πα,1 U κv (−ap )j−j Φj  (a)v  j  j =0

First we note that all the operators vκ

α

·U

−vκ



−1

= α

U

−vκ

 −vκ   −vκ n−1 t −1 = I +α Z = α−1 Z t t=0

186

4 Admissible measures for standard L–functions

are uniformely bounded for v → ∞ by a positive constant C1 (where U = αI + Z and Z n = 0 for n = rkA Mr∗ (N p; A)) because n does not depend on v −vκ and the binomial coefficients are all Zp -integers. t On the other hand by the Main congruences (4.111) and (4.112) with κ = 2 c− (for the distributions Φj = Φc+ j+1 or Φj = Φ−j )     j    κv   j j−j U ( Φj  ((a)v )) (−ap )  j   j  =0

≤ Cp−vj p,Mr∗

for all j = 0, 1, . . . , k−m−1. If we apply to this estimate the previous bounded operators πα,1 U κv we get the following inequality      j α  (yp − ap ) dΦ  ≤ C · C1 |α−vκ |p · p−vj = CC1 p−vj    (a)v p,Mr∗

α = πα (Φ0 ) is a because |α−vκ |p = 1. In particular, for j = 0 it implies that Φ bounded measure, and for all j = 0, 1, . . . , k − m − 1 we have moreover that    j α dπα (Φj ) = yp dπα (Φ0 ) = ypj dΦ (a)v

(a)v

(a)v

Remark 4.24. We see in this case that all distributions πα (Φj ) are bounded but we do not claim that the distributions Φj are bounded themselves. In the general case h > k − m − 1 the integer j run over {0, · · · , h − 1} and one can extend the values of our functions Lc± by the equality Lc± (χxsp ) = 0 (for all χ ∈ XStors of a conductor divisible by all the prime divisors of N p) for j > k − m − 1 − ν (but keeping the same values for 0 ≤ s ≤ k − m − 1 − ν).  ± These are the factors rt which vanish in the above formulas (4.131) for the distributions B ± and the verification of the h-admissibility goes without change in this situation. Also, one obtains again the h-admissible measures Dc± with h = [2ordp (α0 (p))] + 1. The functions Lc± coincide therefore with the Mellin transforms of these h-admissible measures. Finally, if h ≤ k − m − 1 then the conditions in (i) uniquely determine the analytic functions Lc± of type o(log(xhp )) by their values following a general property of admissible measures (see [Am-Ve], [Vi1]). In the case h > k − m − 1, there exist many analytic functions Lc± verifying the conditions in (i) which depend on a choice of analytic continuation (interpolation) for the values Lc± (χxjp ) if j > k − m− 1 − ν but one shows in the theorem 4.23 that there exists at least one such continuation (for example the one which was described in the proof of (iii)).

References

Y. Amice, J. Velu, Distributions p-Adiques associ´ees aux S´eries de Hecke, Journ´ees Arithm´etiques de Bordeaux (Conf. Univ. Bordeaux (1974)), Ast´erisque N◦ 24/25, Soc. Math. France, Paris (1975), p 119131. [An1] A.N. Andrianov, Euler products attached to Siegel Modular Forms of Degree 2, Uspekhi Mat. Nauk 29 (1974), p 44-109 (in Russian). [An2] A.N. Andrianov, Symmetric Squares of Zeta Functions of Siegel Modular Forms of degree 2, Trudy Steklov Math. Institut USSR. 142 (1976), p 2245 (in Russian). [An3] A.N. Andrianov, Euler Expansions of the Theta Transformations of Siegel Modular Forms of Degree n, Mat. Sbornik 104 (1977), p 390-427. [An4] A.N. Andrianov, On the Decomposition of Hecke Polynomials for the Symplectic Group of Degree n, Mat. Sbornic 104 (1978), p. 291-341 (in Russian). [An5] A.N. Andrianov, Multiplicative Arithmetic of Siegel Modular Forms, Uspekhi Mat. Nauk 34 (1979), p 67-135 (in Russian). [An6] A.N. Andrianov, Modular Descent and the Saito-Kurokawa Conjecture, Inv. Math. 53 (1979), p 267-280. [An-Ka] A.N. Andrianov, V.L. Kalinin, On Analytic Properties of Standard Zeta Functions of Siegel Modular Forms, Mat. Sbornik 106 (1978), p 323-339 (in Russian). [An-Ma1] A.N. Andrianov, G.N. Maloletkin, Behaviour of the Theta Series of Degree n under Modular Susbstitutions, Izvestija Akad. Nauk. Ser. Matem. 39 (1975), p 243-258 (in Russian). [An-Ma2] A.N. Andrianov, G.N. Maloletkin, Behaviour of the Theta Series of Degree n of indefinite Quadratic Forms under Modular Substitutions, Trudy Steklov Mat. Institut USSR 148 (1978), p 5-15 (in Russian). [An-Pa] A.N. Andrianov, A.A.Panchishkin, Singular Frobenius operators on Siegel modular forms with characters and zeta functions, St.-Petersbourg Math.J., vol. 12, no. 2 (2001), pp.233-257. [Ar] B. Arnaud, Interpolation p-Adique d’un Produit de Rankin, CRAS Paris Ser.1, 299 (1984), p 527-530. [Ba] W.L. Baily, Eisenstein Series on tube domains, In: Problems in analysis. Princeton University Press (1970). [Am-Ve]

M. Courtieu and A. Panchishkin: LNM 1471, pp. 187–193, 2004. c Springer-Verlag Berlin Heidelberg 2004


188

References

¨ ¨ cherer, Uber S. Bo die Fourier-Jacobi-Entwicklung Siegelscher Eisensteinreihen, Math. Z. 183 (1983), p. 21-46. ¨ ¨ cherer, Uber [B¨ o2] S. Bo die Funktionalgleichung Automorpher L-Funktionnen zur Siegelschen Modulgruppe, J. reine angew. Math. 362 (1985), p 146-168. ¨ ¨ cherer, Uber [B¨ o3] S. Bo die Fourier-Jacobi-Entwickelung Siegelscher Eisensteinreihen. II., Math. Z. 189 (1985) , p. 81-100. ¨ cherer, C.-G. Schmidt, p-adic measures attached to Siegel modu[B¨ oSch] S. Bo lar forms, Ann. Inst. Fourier 50 (2000), no. 5, 1375–1443. ˇ ˇ ˇ, Number theory, 3d ed. Nauka : Moscow [Bo-Sa] S.I. Borevic, I.R. Safarevi c (1985) (in Russian). ¨ lich, Algebraic Number Theory, Acad. Press. [Ca-Fr] J.W.S. Cassels, A. Fro London, N.Y., San Fransisco (1967), 366 p. [Ch-Eh] Young Ju Choie, W. Eholzer, Rankin-Cohen operators for Jacobi and Siegel forms, J. Number Theory 68 (1998), no. 2, 160–177. [Co-PeRi] J. Coates, B. Perrin-Riou, On p-adic L-functions attached to Motives over Q, Advanced Studies in Pure Math. 17 (1989), 23-54 [Co-Schm] J. Coates, C.G. Schmidt, Iwasawa theory for the Symetric Square of an Elliptic Curve, J. reine angew. Math. 375/376 (1987), p 104-156. [Co-Schn] J. Coates, P. Schneider, A note on p-Adic Distribution, Preprint, Heidelberg (1985). [Colm98] P. Colmez, Fonctions L p-adiques, S´eminaire Bourbaki, 51 `eme ann´ee, 1998–99, n◦ 851. Novembre 1998. [CoPB] R. Coleman, p-adic Banach spaces and families of modular forms. Invent. Math. 127, No.3, 417-479 (1997) [Co-Ma] R. Coleman, B. Mazur, The eigencurve. Galois representations in arithmetic algebraic geometry (Durham, 1996), 1–113, London Math. Soc. Lecture Note Ser., 254, Cambridge Univ. Press, Cambridge, 1998. [Cour] M. Courtieu, Familles d’op´erateurs sur les formes modulaires de Siegel et fonctions L p-adiques, http://www-fourier.ujfgrenoble.fr/THESE/ps/t101.ps.gz, Th`ese de Doctorat, Institut Fourier, 2000, 1–122. [Eh-Ib] W. Eholzer, T. Ibukiyama Rankin-Cohen type differential operators for Siegel modular forms. , Internat. J. Math , 9 (1998) no. 4, 443–463. [De1] P. Deligne, Formes Modulaires et Repr´esentations l-Adiques, S´em. Bourb. 1968/69, exp. N◦ 335. Springer-Verlag, Lect. Notes in Math. N◦ 179 (1971), p 139-172. [De2] P. Deligne, La Conjecture de Weil. I, Publ. math. IHES 43 (1974), p 273-307. [De-Ri] P. Deligne, K.A. Ribet, Values of Abelian L-Functions at Negative Integers over Totally Real Fields, Invent. Math. 59 (1980), p 227-286. [Fa-Ch90] G. Faltings, Ching-Li Chai, Degeneration of abelian varieties. Ergebnisse der Mathematik und ihrer Grenzgebiete. 3. Folge, 22. Berlin etc.: Springer-Verlag. xii, 316 p. (1990). [Fe] P. Feit, Poles and Residues of Eisenstein Series for Symplectic and Unitary Groups, Memoir AMS 61 N◦ 346 (1986), 89 p. [Fo] O.M. Fomenko, Applications of the Theory of Modular Forms to Number Theory, In the series “Algebra. Topology. Geometry.” Vol. 15. VINITI Publ. Moscow (1977), p 5-91 (in Russian). [Fr] E.Freitag, Siegelsche Modulfunktionnen, Springer-Verlag : Berlin e.a. (1983), x+341 p. [B¨ o1]

References

189

P.B.Garrett, M.Harris, Special values of triple product L–functions, Amer. J. Math 115 (1993) 159–238. [Gou-Ma] F.Gouvˆ ea, B.Mazur, On the density of modular representations, in: Computational perspectives in Number Theory, ed. by D.A. Buell, J.T. Teitelbaum, Amer. Math. Soc., 127–142 (1998). [Gr-Za] B. Gross, D. Zagier, Heegner Points and Derivatives of L-Series, Invent. Math. 84 (1986), p 225-320. [Ha1] M. Harris, The Rationality of Holomorphic Eisenstein Series, Invent. Math. 63 (1981), p 305-310. [Ha2] M. Harris, Special Values of Zeta-Functions attached to Siegel Modular Forms, Ann. Sci. Ec. Norm. Sup. 14 (1981), p. 77-120. [Ha3] M. Harris, Eisenstein Series on Shimura Varieties, Ann. of Math. 119 (1984), p. 59-94. [He] E. Hecke, Theorie der Eisensteinscher Reihen und ihre Anwebdung auf Fonktionnentheorie und Arithmetik, Abh. Math. Sem. Hamburg 5 (1927), p. 199-224. [Hi1] H. Hida, A p-Adic Measure attached to the Zeta Functions associated with two Elliptic Cusp Forms. I, Invent. Math. 79 (1985), p 159-195. [Hi2] H. Hida, Galois representations into GL2 (Zp [[X]]) attached to ordinary cusp forms, Invent. Math. 85 (1986) 545–613. [Hi3] H. Hida, Le produit de Petersson et de Rankin p-adique, S´eminaire de Th´eorie des Nombres, Paris 1988–1989, 87–102, Progr. Math., 91, Birkh¨ auser Boston, Boston, MA, 1990. [Hi4] H. Hida, On p-adic L-functions of GL(2) × GL(2) over totally real fields, Ann. Inst. Fourier, (Grenoble) 40, no.2 (1991) 311–391. [Hi5] H. Hida, Elementary theory of L-functions and Eisenstein series, London Mathematical Society Student Texts. 26 Cambridge University Press [Hi6] H. Hida, p-adic ordinary Hecke algebras for GL(2), Ann. Inst. Fourier (Grenoble) 44 (1994), no. 5, 1289–1322. [Ho-PSh1] R. Howe, I.I. Piatetski-Shapiro, A Couterexample to the “Generalized Ramanujan Conjecture” for (quasi-) Split Groups, Automrphic Forms, Representations and L-Functions. Proc. Pure Math. Vol. 33 Part. 1. Providence : R.I. (1979), p 315-322. [Ho-PSh2] R. Howe, I.I. Piatetski-Shapiro, Some Examples of Automorphic Forms on SP4 , Duke Math. J. 50 (1983), p 55-106. [HuLK] Hua Lo-Ken, Harmonic analysis of functions of several complex variables in the classical domains, Providence: R.I. (1963). [Iw] K. Iwasawa, Lectures on p-Adic L-Functions, Ann. of Math. Studies, N◦ 74. Princeton Univ. Press (1972). [Ja] H. Jacquet, Automorphic Forms on GL(2). Part II, Springer-Verlag, Lect. Notes in Math. N◦ 114 (1972), 548 p. [Kal] V.L. Kalinin, Eisenstein Series on the Symplectic Group, Mat. Sbornik 103 (1977), p 519-549 (in Russian). [Kat1] N.M. Katz, p-Adic Interpolation of Real Analytic Eisenstein Series, Ann. of Math. 104 (1976), p 459-571. [Kat2] N.M. Katz, p-Adic L-Functions for CM-Fields, Invent. Math. 48 (1978), p 199-297. [Kat3] N.M. Katz, p-Adic L-Functions for CM -Fields, Invent. Math. 48 (1978), p 199-297. [Ga-Ha]

190

References

[KMS2000] D. Kazhdan, B. Mazur, C.-G. Schmidt, Relative modular symbols and Rankin-Selberg convolutions. J. Reine Angew. Math. 519, 97-141 (2000). [Ki] Y. Kitaoka, Lectures on Siegel Modular Forms and Representations by Quadratic Forms, Springer-Verlag, Berlin e.a. (1986). [Kl1] H. Klingen, Zum Darstellungssatz f¨ u Siegelsche Modulformen, Math. Z. 102 (1967), p. 30-43. ¨ [Kl2] H. Klingen, Uber Poincar´esche Reihen vom Exponentialtyp, Math. Ann. 234 (1978), p. 145-170. [Kl3] H. Klingen, Zum Darstellungssatz f¨ ur Siegelsche Modulformen, Math. Z. 102 (1967) 30–43 [Kob1] N. Koblitz, p-Adic Analysis : A Short Course on Recent Work, London Math. Soc. Lect. Notes Series, N◦ 46. Cambridge University Press. London, Cambridge (1980). [Kob2] N. Koblitz, p-Adic Numbers, p-Adic Analysis and Zeta Functions, 2nd ed. Springer-Verlag (1984). [Koj] H. Kojima, On Construction of Siegel Modular Forms of degree two, Math. Soc. Japan 34 (1982), p 393-412. [Ku-Le] T. Kubota, H.W. Leopoldt, Eine p-Adische Theorie der Zetawerte, J. Reine Angew. Math. 214/215 (1964), p 328-339. ˇanov, Zeta Functions of Elliptic Curves over certain Abelian [Kurˇc1] P.F. Kurc Extension of Imaginary Quadratic Fields, Mat. Sbornik 108 (1977), p 56-70 (in Russian). ˇanov, Local Measures connected with Cusp Forms of Jacquet[Kurˇc2] P.F. Kurc Langlands over CM-Fields, Izvestija Akad. Nauk. Ser. Matem. 42 (1978), p 588-601 (in Russian). [Kuro] N. Kurokawa, Examples of Eigenvalues of Hecke Operators on Siegel Cusp Forms of degree two, Invent. Math. 49 (1978), p 149-165. [Ku-Mi] N. Kurokawa, S. Mizumoto, On Eisenstein Series of Degree two, Proc. Jap. Acad. A57 (1981) , p. 134-139. [La1] S. Lang, Introduction to Modular Forms, Springer-Verlag (1976). [La2] S. Lang, Cyclotomic Fields, (Graduate Texts in Math.) Springer-Verlag, N.Y. (1980). [Ll] R. Langlands, On the Functional Equations satisfied by Eisenstein Series, Springer-Vrelag, Lect. Notes in Math. N◦ 544 (1976), 337 p. [LBP] Y.-H. Le Bras, A.A. Panchishkin, Sur les produits triples Λ-adiques, Communications in Algebra, 29 (2001) No. 9, 3727-3740. [Le1] H.W. Leopoldt, Eine Verallgemeinerung der Bernoullischen Zahlen, Abh. Math. Sem. Univ. Hambourg 22 (1958), p 131-140. [Le2] H.W. Leopoldt, Eine p-Adische Theorie der Zetawerte. II, J. Reine Angew. Math. 274/275 (1975), p 224-239. [Maa] H. Maass, Siegel’s Modular Forms and Dirichlet Series, Springer-Verlag, Lect. Notes in Math. N◦ 216 (1971). [Mah] K. Mahler, An Interpolation Series for a Continuous Function of p-Adic Variable, J. Reine Angew. Math. 199 (1958), p 23-34. [Man1] Y.I. Manin, Cyclotomic Fields and Modular Curves, Uspekhi Mat. Nauk 26 (1971), p 7-78 (in Russian). [Man2] Y.I. Manin, Periods of Cusp Forms and p-Adic Hecke Series, Mat. Sbornik 92 (1973), p 378-401 (in Russian). [Man3] Y.I. Manin, Non-Archimedean Integration and p-Adic L-Functions of Jacquet-Langlands, Uspekhi Mat. Nauk 31 (1976), p 5-54 (in Russian).

References

191

[Maz-SD] B. Mazur, H.P.F. Swinnerton-Dyer, Arithmetic of Weil Curves, Invent. Math. 25 (1974), p 1-61. [Mi-St] J. Milnor, J. Stasheff, Characteristic Classes, Ann. of Math. Studies N◦ 76, Princeton Univ. Press. (1974), p 231-264. [Miy] Toshitsune. Miyake, Modular forms. Transl. from the Japanese by Yoshitaka Maeda., Berlin etc.: Springer-Verlag. viii, 335 p. (1989). [Miz1] S. Mizumoto, Fourier Coefficients of Generalized Eisenstein Series of Degree two, Proc. Jap. Acad. A59 (1983), p. 231-264. [Miz2] S. Mizumoto, Fourier Coefficients of Generalized Eisenstein Series od Degree two, Kodai Math. J. 7 (1984), p. 86-110. [MyVQ] My Vinh Quang, Non-Archimedean Rankin Convolution of Unbounded growth, Math. USSR Sbornik 72 (1992), p 151-161. [Nag1] S. Nagaoka, A remark on Serre’s example of p-adic Eisenstein series, Math. Z. 235 (2000) 227-250. [Nag2] S. Nagaoka, Note on mod p Siegel modular forms, Math. Z. 235 (2000) 405-420. [Pa1] A.A. Panchishkin, Symmetric Squares of Hecke Series and their Values at Integral Points, Mat. Sbornik 108 (1979), p 393-417 (in Russian). [Pa2] A.A. Panchishkin, Complex valued Measures attached to Euler Products, Trudy Sem. Petrovskogo 7 (1981), p 239-244 (in Russian). [Pa3] A.A. Panchishkin, Le prolongement p-Adique Analytique de Fonctions L de Rankin, CRAS Paris 294 (1982) 51-53, p 227-230. [Pa4] A.A. Panchishkin, Local Measures attached to Euler Products in Number Fielfds, In “Algebra” (Ed. by A.I. Kostrikin), Moscow Univ. Press. (1982), p 119-138 (in Russian). [Pa5] A.A. Panchishkin, A Functional Equation of the non-Archimedean Rankin Convolution Duke Math. J. 54 (1987), p 77-89. [Pa6] A.A. Panchishkin, Non-Archimedean L-Functions of Siegel and Hilbert Modular Forms, Lecture Notes in Math. 1471, Springer-Verlag (1991), 166 p. [Pa7] A.A. Panchishkin, Admissible Non-Archimedean Standard Zeta Functions of Siegel Modular Forms, Proceedings of the joint AMS Summer Conference on Motives, Seatle, July 20 - August 2 1991, Seatle, Providence, R.I. vol. 2 (1994), p 251-292. [Pa8] A.A. Panchishkin, On the Siegel-Eisenstein measure and its applications, Israel Journal of Mathemetics, 120, Part B (2000) 467-509. [Pa9] A.A. Panchishkin, On p-adic integration in spaces of modular forms and its applications, Preprint of IAS, January (2000) 1-31. [Pa10] A.A. Panchishkin, A new method of constructing p-adic L-functions associated with modular forms, Moscow Mathematical Journal, 2 (2002), Number 2, 1-16 [PaInv] A.A. Panchishkin, Two variable p-adic L functions attached to eigenfamilies of positive slope, Invent. Math. Published Online on July 1 (2003). [Pa12] A.A. Panchishkin, Sur une condition suffisante pour l’existence des mesures p-adiques admissibles, Preprint No.560 (2002) of the Institut Fourier (accepted in Journal de Th´eorie des Nombres de Bordeaux, 2003) [Ra1] S. Raghavan, Modular Forms of Degree n and Representations by Quadratic Forms, Ann. of Math. 70 (1959), p 446-477.

192 [Ra2] [Ra3]

[Res] [Res-S] ˇ [Saf] [Sat] [Sc] [Se1] [Se2]

[Sh1] [Sh2] [Sh3] [Sh4] [Sh5] [Sh6] [Sh7] [Sh8] [Sh9]

[Sho] [Si1] [Si2] [St1] [St2]

References S. Raghavan, Singular Modular Forms of Degree s, Papers Colloq. S. Ramanujan. A Tribute. Berlin e.a. (1978), p. 263-272. S. Raghavan, Estimates of Coefficients of Modular Forms and generalized Modular Relations, Automorphic Forms, Representation Theory and Arithmetic. Papers Colloq., Bombay(1979), p. 8-15, Berlin e.a. (1981), p 247-254. H.L. Resnikoff , Automorphic forms of singular weight are singular forms, Math. Ann. 215 (1975) 173–183. H.L. Resnikoff , R.L.Saldana Some algebraic properties of Fourier coefficients of Eisenstein series of degree two, J.reine angew. Math. (1974) 90–109. ˇ ˇ, Zeta Function, Moscow University Press (1969). I.R. Safarevi c I. Satake, Theory of Spherical Functions on Reductive Algebraic Groups over p-Adic Fields, Publ. Math. IHES N◦ 18 (1963). C-.G. Schmidt, The p-Adic L-Functions attached to Rankin Convolutions of Modular Forms, J. reine angew. Math. 368 (1986), p 201-220. J.-P. Serre, Cours d’Arithm´ etique, Press. Univ. France : Paris, 1970; 2nd ed. (1977), 188 p. J.-P. Serre, Formes modulaires et fonctions zˆeta p-adiques, Modular functions of one variable, III (Proc. Internat. Summer School, Univ. Antwerp, 1972) 191–268, Lecture Notes in Math., Vol. 350, Springer, Berlin, 1973. G. Shimura, Introduction to the Arithmetic Theory of Automorphic Functions, Princeton Univ. Press. (1971). G. Shimura, On the Holomorphy of certain Dirichlet Series, Proc. Lond. Math. Soc. 31 (1975), p 79-98. G. Shimura, On the Periods of Modular Forms, Math. Annalen 229 (1977), p 211-221. G. Shimura, The Special Values of Zeta Functions associated with Hilbert Modular Forms Duke Math. J. 48 (1978), p 637-679. G. Shimura, Arithmetic of Differential Operators on Symetric Domains, Duke Math. J. 48 (1981), p 813-843. G. Shimura, Confluent Hypergeometric Functions on Tube Domains, Math. Ann. 260 (1982), p. 269-302. G. Shimura, Algebraic Relations between Critical Values of Zeta Functions and inner Products, Amer. J. Math. 105 (1983), p 253-285. G. Shimura, On Eisentsein Series, Duke Math. J; 50 (1983), p. 417-476. G. Shimura, Arithmeticity in the theory of automorphic forms, Mathematical surveys and monographs, 82 AMS, Providence, RI 2000, 302 p. V.V. Shokurov, The Integrals of Shimura of Cusp Forms, Izvestija Akad. Nauk. Ser. Mat. 44 (1980), p 670-718. ¨ C.L. Siegel, Uber die Zetafunktionnen Indefiniter Quadratischer Formen, Math. Z. 43 (1938), p. 682-708. C.L. Siegel, Ein¨ uhrung in die Theorie des Modulfunktionnen n-ten Grades, Math. Ann. 116 (1939), p 617-657. J. Sturm, Special Values of Zeta Functions and Eisenstein Series of half integral weight, Amer. J. Math. 102 (1980), p 219-240. J. Sturm, The Critical Values of Zeta-Functions associated to the Symplectic group, Duke Math. J. 48 (1981), p 327-350.

References [Til-U]

[Vi1] [Vi2]

[Wa] [Weil] [Weis1] [Weis2] [Weis3] [Wi] [Za]

193

J. Tilouine, E. Urban, Several variable p-adic families of Siegel-Hilbert ´ Norm. cusp eigenforms and their Galois representations, Ann. scient. Ec. Sup. 4e s´erie, 32 (1999) 499–574. M.M. Viˇsik, Non-Archimedean Measures associated with Dirichlet series, Mat. Sbornik 99 (1976), p 248-260. M.M. Viˇsik, Non-Archimedean Spectral Theory, In the series : “Modern Problems of Mathematics”, Vol. 25, Moscow : VINITI Publ. (1984), p 51-114. L. Washington, Introduction to Cyclotomic Fields, Springer-Verlag : N.Y. e.a. (1982). A. Weil, Basic Number Theory, Springer-Verlag, Hdlbrg-N.Y. (1973), 312 p. ¨ R. Weissauer, Uber die Konstruktion von Spitzenformen mit Hilfe von Thetareihen, Abh. Math. Sem. Hamburg 56 (1986), p. 133-151. R. Weissauer, Stabile Modulformen und Eisensteinreihen, SpringerVerlag, Lect. Notes in Math. N◦ 1219 (1987). R. Weissauer, Divisors on the Siegel Modular Variety, Springer-Verlag, Lect. Notes in Math. N◦ 1240 (1987). A.Wiles, Modular elliptic curves and Fermat’s Last Theorem. Ann. Math., II. Ser. 141, No.3, 443-551 (1995) D.B. Zagier, Modular Forms whose Fourier Coefficients involve ZetaFunctions of Quadratic Fields, In : Modular Functions. V, SpringerVerlag, Lect. Notes in Math. N◦ 627 (1977), p 106-168.

Index

S-adic Mazur measure, 34 C ∞ -modular form, 83 h-admissible measure, 9, 37, 129, 132, 144, 185 analytic function, 18 arithmetical Siegel modular forms, 174 Atkin-Lehner operator, 7 canonical projection operator, 7, 130, 146 compatibility criterion, 22 completed group ring, 32 confluent hypergeometric function, 45, 74, 77 convolution of distributions, 32 critical values, 75 equivalent symmetric couples, 65 Fourier expansions, 106, 110, 136 Frobenius element, 58 Frobenius operator, 7 function of bounded growth, 83 function of moderate growth, 84 Gauss sum of degree m, 6, 132, 149 Gauss-Manin connection, 7, 96 growth condition, 37, 146 Hecke Hecke Hecke Hecke

algebra, 53 couple, 56 operators, 7, 53, 95, 127 polynomial, 53, 54

holomorphic projection operator, 45 inertial degree, 14 involuted Siegel-Eisenstein series, 136 Iwasawa algebra, 32 Klingen-Eisenstein series, 71 locally analytic functions, 129, 144 locally constant functions, 9, 129 locally polynomial functions, 9, 129, 144 Maass differential operator, 75, 96, 106, 133 Mahler’s criterion, 18 measures of bounded growth, 37 Mellin transform, 6, 28, 37, 54 nearly holomorphic Siegel modular form, 7, 95, 96, 99, 102, 110, 127, 135, 144 Newton polygon, 20 non–vanishing condition, 131, 132 non-Archimedean Mellin transform, 31 non-Archimedean zeta function of Kubota-Leopoldt, 34 nonvanishing condition, 184 normaized Siegel-Eisenstein distributions, 141 normalized Siegel-Eisenstein distribution, 165 normalized Siegel-Eisenstein series, 136, 169 normalized standard L-function, 173 notation of Petersson, 100, 115, 130

196

Index

Ramanujan-Petersson conjecture, 54 ramification index, 14 Rankin zeta function, 67, 149 Saito-Kurokawa lifting, 54 Satake q-parameters, 53, 131, 148 Satake isomorphism, 51, 52, 57, 115 sequence of distributions, 144 Shimura differential operator, 140 Shimura differential operator, 106, 110, 115, 116, 128, 133, 140, 141 Siegel cusp eigenform, 132 Siegel modular forms, 45 Siegel-Eisenstein series, 66, 134, 135 spinor representation, 12 spinor zeta function, 12, 53, 60

standard L-function, 131, 148 standard representation, 12 standard zeta function, 54 sum of eigenvalues, 74 supersingular case, 37 symmetric couple, 65 tame component, 30 Teichm¨ uller representative, 14 totally ramified extension, 15 trace operator, 7 twisted convolution, 165 unramified extension, 14 wild component, 30