$p$-adic $L$-functions for unitary Shimura varieties. I: Construction of the Eisenstein measure
Michael Harris,Jian-Shu Li,Christopher M. Skinner
DOI: https://doi.org/10.4171/dms/4/12
2006-01-01
Abstract:We construct the Eisenstein measure in several variables on a quasi-split unitary group, as a first step towards the construction of p-adic L-functions of families of ordinary holomorphic modular forms on unitary groups. The construction is a direct generalization of Katz’ construction of p-adic L-functions for CM fields, and is based on the theory of p-adic modular forms on unitary Shimura varieties developed by Hida, and on the explicit calculation of non-degenerate Fourier coefficients of Eisenstein series. 2000 Mathematics Subject Classification: Primary 11F33, 11R23; Secondary 14G35 Introduction This is the first of a projected series of papers devoted to studying the relations between p-adic L-functions for GL(n) (and unitary groups), congruences between stable and endoscopic automorphic forms on unitary groups, and Selmer groups for p-adic representations. The goals of these papers are outlined in the survey article [HLS]. The purpose of the present installment is to prepare the ground for the construction of p-adic L-functions in sufficient generality for the purposes of subsequent applications to congruences and Selmer groups. Institut de Mathématiques de Jussieu, U.M.R. 7586 du CNRS; Member, Institut Universitaire de France Partially supported by RGC-CERG grants HKUST6115/02P, HKUST601805 and a grant from the Chinese Academy of Science Partially supported by a grant from the National Science Foundation and a fellowship from the David and Lucile Packard Foundation Documenta Mathematica · Extra Volume Coates (2006) 393–464 394 Michael Harris, Jian-Shu Li, Christopher M. Skinner The first general conjectures on the construction of p-adic L-functions for ordinary motives were elaborated by Coates in [Co]. The conjectured p-adic analytic functions of [Co] interpolate the quotients of normalized values of Lfunctions at critical points, in the sense of Deligne. The normalization proceeds in two steps. The critical values are first rendered algebraic, by dividing by their Deligne periods. Next, they are p-stabilized: the Euler factors at p and ∞ are modified according to a complicated but explicit recipe. Coates’ conjecture is that the resulting values are p-adically interpolated by a p-adic analytic function of Iwasawa type, associated to a p-adic measure. In our setting, the Deligne period is generally replaced by a certain Petersson norm or an algebraic multiple thereof; the relation of this Petersson norm to the Deligne period is discussed at length in [H3]. In [Pa], Panchishkin points out that Coates’ recipe can be adapted unchanged for motives satisfying a condition weaker than ordinarity, which he calls admissibility and which Perrin-Riou and Greenberg have called the Panchishkin condition. Although this is somewhat obscured by the automorphic normalization, we work in the generality of Panchishkin’s admissibility condition. Panchishkin also conjectures the existence of more general p-adic L-functions in the absence of admissibility; we do not address this question. We work with automorphic forms on the unitary groups of hermitian vector spaces over a CM field K, with maximal totally real subfield E. We assume every prime of E dividing p splits in K; we also impose a hypothesis (1.1.2) linking primes above p to signatures of the unitary group at real places of E. Unitary groups, unlike GL(n), are directly related to Shimura varieties. We show that the special values of L-functions of automorphic forms on unitary groups satisfy the congruences needed for the construction of p-adic L-functions by appealing to the fact that the corresponding Shimura varieties are moduli spaces for abelian varieties of PEL type. In this our approach is directly modeled on Katz’s construction [K] of p-adic L-functions for Hecke characters of CM fields; indeed, for groups of type U(1) our results reduce to those of Katz. The starting point of Katz’s construction is Damarell’s formula and its generalizations due to Shimura, which relate the values of arithmetic Eisenstein series at CM points to special values of L-functions of arithmetic Hecke characters. A generalization of Damerell’s formula in higher dimensions is the construction of standard L-functions of unitary groups by the doubling method. This was first developed systematically in the article [PSR] of Piatetski-Shapiro and Rallis, though special cases had been discovered independently by Garrett, and a more thorough development in classical language is contained in the books [S97, S00] of Shimura. The local theory for unitary groups was ignored in [PSR] but was worked out in [L2] and [HKS]. Our p-adic L-functions are actually attached to Hida families of nearly ordinary modular forms on a unitary group G = U(V ). As in [K], the main step is the construction of an Eisenstein measure on a large unitary group H, attached to the sum of two copies V ⊕ (−V ) of V . The hermitian form on −V has been multiplied by −1, so thatH is quasi-split and its associated Shimura variety has Documenta Mathematica · Extra Volume Coates (2006) 393–464 p-adic L-functions for unitary Shimura varieties, I 395 a point boundary component, stabilized by a maximal parabolic subgroup, the Siegel parabolic. The Eisenstein series attached to the Siegel parabolic are the direct generalizations of the classical Eisenstein series onGL(2). The Eisenstein measure is a p-adic measure on a product T of copies of Zp with values in the algebra of p-adic modular forms on H interpolating such Eisenstein series. The theory of p-adic modular forms on H was developed by Hida in [Hi04, Hi05]. As in [K], these forms belong to the algebra of functions on the Igusa tower, which is a rigid analytic étale covering of the ordinary locus of the Shimura variety attached to H. The existence of the Eisenstein measure relies crucially on the irreducibility of the Igusa tower; this was established in some generality by Hida, though easier arguments due to Chai and Hida himself suffice for the case at hand (cf. [Ch, Hi06]). The Eisenstein measure associates, by integrating over T with respect to this measure, p-adic modular forms to continuous functions on T . The integrals of characters of T of finite order, which determine the measure, are classical holomorphic (Siegel) Eisenstein series on H and as such are associated to explicit functions (“sections”) belonging to degenerate principal series induced from characters of the Siegel parabolic. These sections factor as tensor products of local sections over the primes of E. At almost all finite primes the local sections are unramified and present no difficulty, and we simplify the theory by choosing local sections at ramified primes, other than those dividing p, that are insensitive to p-adic variation of the character of T . With our choice of data, the Fourier coefficients of the Eisenstein series at a chosen point boundary component also factor over primes. All the work in constructing the Eisenstein measure then comes down to choosing local data at primes above p such that the corresponding local coefficients satisfy the necessary Kummer congruences. Our strategy for choosing local data follows [K] in making use of a partial Fourier transform. Unlike in [K], our construction is systematically adelic and isolates the local considerations at p. The Eisenstein measure is designed to pair with Hida families – on G × G, not on G itself – and thus depends on several variables, considerably complicating the calculations. The doubling method was used by Böcherer and Schmidt in [BS] to construct standard p-adic L-functions for Siegel modular forms. They do not use p-adic modular forms; their approach is to construct the p-adic measure directly in terms of normalized special values of complex L-functions. Their approach applies to all critical values, unlike the present paper, which avoids reference to non-holomorphic differential operators (and their p-adic analogues). Presumably their techniques work for quasi-split unitary groups as well. We have not attempted to compare our results where they can be compared, namely in the local analysis at the prime p, since our group is locally isomorphic to GL(2n), in principle much simpler than a symplectic group. As predicted by Coates, the shape of the modified Euler factor at a prime v dividing p depends on the p-adic valuations of the eigenvalues of Frobenius at v. On the other hand, as in [H3], the fact that a critical value of the standard L-function is an algebraic multiple of a period of an arithmetic modular form on Documenta Mathematica · Extra Volume Coates (2006) 393–464 396 Michael Harris, Jian-Shu Li, Christopher M. Skinner the doubled group G×G – in other words, the Petersson norm of an arithmetic modular form on G – can be expressed in terms of Hodge numbers. Then the Panchishkin condition, applied to the standard L-function for GL(n)K, roughly states that, for each v dividing p, the modified Euler factor at v is given by a natural partition of the Frobenius eigenvalues at v that corresponds to the signature of the unitary group at real places assigned to v by Hypothesis (1.1.2). The form of the modified Euler factor at p is thus linked to the real form of G. This is reflected in the fact that the natural embedding of the Shimura variety attached to G × G in that attached to H in general does not define a map of Igusa towers. In order to pair p-adic modular forms on H with p-adic modular forms on G × G, the natural embedding has to be replaced by a padic translation (cf. (2.1.11)), which is exactly what is needed to provide the expected modification of the Euler factor. The main innovation of our construction concerns the zeta integral at p. As in [K], the use of a partial Fourier transform to define local data at p with the appropriate congruence properties to construct the Eisenstein measure is precisely what is needed to obtain the modified Euler factor at p directly as a local zeta integral, up to some volume factors.