THE WEIGHT TWO AND OPPOSITE SIGN CASES FOR THE FOURIER RELATIVE TRACE FORMULAS

Matteo Di Scipio · 2026

Read THE WEIGHT TWO AND OPPOSITE SIGN CASES FOR THE FOURIER RELATIVE TRACE FORMULAS by Matteo Di Scipio (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-13062.txt.

Introduction

Matteo Di Scipio gives an adelic relative trace formula proof of the Petersson and Bruggeman-Kuznetsov formulas in two hard cases. One case is the holomorphic setting with weight two. The other is the non-holomorphic setting where the product of the two parameters is negative. Two sets of hypotheses on the non-archimedean test function frame the proof. They are named the geometric and spectral assumptions. Under them, the approach yields refined versions of both formulas.

Catalog record

CREATORMatteo Di Scipio
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE81505
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.13062

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

THE WEIGHT TWO AND OPPOSITE SIGN CASES FOR THE FOURIER RELATIVE TRACE FORMULAS Matteo Di Scipio Abstract We provide an adelic relative trace formula proof to the Petersson/Bruggeman-Kuznetsov (PBK) formulas, specifically in the holomorphic case for κ = 2 \kappa=2 and the non-holomorphic case for m 1 ​ m 2 < 0 m_{1}m_{2}<0 . Given two sets of hypothesis on the non archimedean test function f f , called the geometric and spectral assumptions, this approach allows us to obtain refined PBK formulas. 1 Introduction 1.1 Motivation The Petersson and the Bruggeman-Kuznetsov formulas are some of the most important tools in analytic number theory. Applications abound in literature, see for instance [ 4 ] , [ 1 ] , [ 27 ] , [ 12 ] for moments and subconvexity, [ 5 ] for large sieve inequalities, [ 8 ] and [ 15 ] for non-vanishing of L-functions at central values, [ 2 ] and [ 23 ] for primes in arithmetic progression, [ 14 ] for low lying zeroes, and [ 25 ] for a survey on Kloostermania and its applications. There are two main ways of obtaining these formulas: a more classical approach involving Poincare series (see e.g [ 26 ] and [ 21 ] ) and the pre-trace formula approach developed by Jacquet [ 17 ] and Zagier [ 29 ] , as exposited by Knigthly and Li in [ 18 ] and [ 20 ] for the Petersson formula of weight \kappa>2 and the Bruggeman-Kuznetsov formula for pairs of coefficient with the same sign. One of the advantages of working with the pre-trace formula is the greater flexibility in selecting families of local non-archimedean representations underlying automorphic forms, obtaining generalised formulas as in [ 13 ] . The aim of this paper is to fill a gap in the literature, developing similar formulas for Maass forms in the opposite sign case, as well as in the \kappa=2 holomorphic case via the pre-trace approach. It is worth highligthing that the holomorphic case is particularly interesting for applications given the connections with elliptic curves over \mathbb{Q} . We first prove a version of the opposite sign formula under minimal assumptions; the geometric side will be in terms of generalised Kloosterman sums, as in [ 13 ] . Following the same paper, we outline how additional constraints of the test function, namely the spectral and geometric assumptions , allow to obtain refined formulas. We prove this for a large class of archimedean test functions h_{\infty} , namely those that satisfy \begin{cases}h_{\infty}(t)\text{ is even}\\ h_{\infty}(t)\ \text{is holomorphic in }|\text{Im}(t)|\leq A\\ h_{\infty}(t)\ll(1+|t|)^{-B}\ \text{in }|\text{Im}(t)|\leq A\end{cases} (1.1) for any A>13 and B>2 as in [ 20 ] (see Theorem 8.1 and its proof loc. cit.). The derivation of this formula runs very similar to the same sign case, with the only major change arising on the archimedean component of the off-diagonal terms on the geometric side, for which we introduce a counterpart to the Zagier transform given in [ 20 , §7.5.2] . As a simple corollary, we prove an equidistribution result for the parities of automorphic representations that are unramified at infinity with specified non-archimedean components. We also prove the refined weight two Petersson formula by a limiting argument, taking a smooth truncation of the matrix coefficient associated with the newform of the discrete series \mathcal{D}(2) . We approach the problem in this way because, unlike in the case \kappa\geq 4 described in [ 18 ] , the matrix coefficient is not in L^{1}([G]) , and the geometric kernel attached to it need not be absolutely and uniformly convergent on compacta. We mention in passing that we chose to work over \mathbb{Q} for sake of presentation, but the results of this paper should generalise to arbitrary number fields. 1.2 Statement of main results For a choice of a given test function f_{\mathbb{A}}=f_{\infty}\otimes f , the trace formulas we shall discuss provide an equality between two series. One side, commonly referred to as the spectral side, involves a product of pairs of Fourier coefficients, averaged over a family of automorphic forms cut by f , whereas the other involved generalised Kloosterman sums. Under mild conditions on the support of f_{p} , one may obtain the following Bruggeman-Kuznetsov formula. Theorem 1.2.1 (Unrefined BK formula) . Suppose f=\bigotimes_{p}f_{p}\in\mathcal{H} is non-zero, and that for each p that f_{p} is supported inside the subgroup of matrices g\in G(\mathbb{Q}_{p}) with v_{p}(\det g)\in 2\mathbb{Z} . For m_{1},m_{2}\in\tfrac{1}{N}\mathbb{Z} with m_{1}m_{2}<0 we have that \displaystyle\sum_{\pi\in\mathcal{F}_{0}(f)} \displaystyle h_{\infty}(t_{\pi})\sum_{\varphi\in\mathcal{B}(\pi)}a_{u_{\pi(f)\varphi}}(m_{1})\overline{a_{u_{\varphi}}(m_{2})} \displaystyle+\frac{1}{4\pi}\sum_{\chi\in\mathcal{F}_{E}(f)}\sum_{\phi\in\mathcal{B}(\chi,\chi^{-1})}\int_{-\infty}^{\infty}h_{\infty}(t)a_{u_{E(\pi_{it}(\chi)(f)\phi_{it})}}(m_{1})\overline{a_{u_{E(\phi_{it})}}(m_{2})}dt (1.2) \displaystyle=\sum_{c\in\mathcal{C}(f)}\frac{H(m_{1},m_{2},c)}{c}H_{\infty}^{-}\left(\frac{4\pi\sqrt{|m_{1}m_{2}|}}{c}\right) as absolute convergent sums/integrals. Here • N is the level of f , • \mathcal{F}_{0}(f) and \mathcal{F}_{E}(f) are the global families cut by f as in Section 1.4.5, • H(m,n,c) and \mathcal{C}(f) are the generalised Kloosterman sums and moduli space associated to f given in Section 2.5 , • \mathcal{B}(\pi) (resp. \mathcal{B}(\chi,\chi^{-1}) ) is an orthonormal basis consisting of pure tensors for \pi^{K_{\infty}\times K(N)} (resp. \pi_{\chi,\chi^{-1}}^{K_{\infty}\times K(N)} ), • u_{\varphi} (resp u_{E(\phi_{it})} ) are the classical \Gamma(N) -Maass forms (resp. Eisenstein series) with Fourier coefficients a_{u_{\varphi}}(m) (resp a_{u_{E(\phi_{it})}}(m) ), as defined in ( 2.9 ), • h_{\infty}(t)\in PW(\mathbb{C}) where PW(\mathbb{C}) is the Paley-Wiener space as in [ 20 , §3.3] , whereas H_{\infty}^{-}(x) is the transform of h_{\infty} given by H_{\infty}^{-}(x)=\frac{1}{\pi}\int_{-\infty}^{\infty}K_{2it}(x)\sinh(\pi t)th_{\infty}(t)dt (1.3) Theorem 1.2.1 is often too general for applications, and we often desire better control on both the geometric and spectral sides. With this purpose in mind, we introduce two sets of conditions, called the Geometric and Spectral assumptions. Geometric assumption : There exists y\in\mathbb{Q}_{+} such that \text{supp}(f)\subset a(y)^{-1}K_{p}a(y) . For any y\in\mathbb{Q}_{+} such that the Geometric assumption holds, we say that y controls the support of f . Spectral assumption : The function f\in\mathcal{H} is non-zero and admits a representation f=\bigotimes_{p}f_{p} where f_{p} is either a newform projector, or there exists c\in\mathbb{Z}_{\geq 0} such that f_{p}=\nu(p^{c})1_{K_{0}(p^{c})} , where \nu(n)=n\prod_{p|n}(1+p^{-1}) . Given a cuspidal representation \pi of G(\mathbb{A}) that factors as \pi\simeq\bigotimes^{\prime}_{v}\pi_{v} , if f satisfies the spectral assumptions, the operator \pi(f) is an orthogonal projection onto the subspace V_{f}=\pi_{\infty}\otimes\bigotimes_{p:f_{p}\ \text{newform proj}}\mathbb{C}v_{0,p}\otimes\bigotimes_{p:f_{p}=v(p^{c})1_{ZK_{0}(p^{c})}}\pi_{p}^{K_{0}(p^{c})} (1.4) where v_{0,p} is the unique newform of \pi_{p} , up to scalars. Following [ 28 , §7] , if \pi\in\mathcal{F}_{0}(f) , there exists an orthonormal basis \mathcal{B}_{f}(\pi) of the subspace V_{f}^{K_{\infty}} , and weights w(\pi,f)\in\mathbb{C} such that for any m_{1},m_{2}\in\mathbb{N} with (m_{1}m_{2},N)=1 (here N is the level of f ) \sum_{\varphi\in\mathcal{B}_{f}(\pi)}a_{u_{\varphi}}(m_{1})\overline{a_{u_{\varphi}}(m_{2})}=w(\pi,f)\lambda_{\pi}(m_{1})\overline{\lambda_{\pi}(m_{2})} (1.5) thus recovering the Hecke eigenvalues, normalised so that the Ramanujan-Petersson conjecture predicts |\lambda_{\pi}(q)|\leq 2 for q prime coprime to N . The same results holds for \pi\in\mathcal{F}_{\kappa}(f) and the right-isotypic space for k_{\theta}\to e(\kappa\theta) . In an analogous fashion, for a Hecke character \chi , letting \lambda_{\pi_{it}(\chi)}(m)=\sum_{ab=m}\chi(a)\chi^{-1}(b)(b/a)^{it} (1.6) we have \sum_{\phi\in\mathcal{B}_{f}(\chi,\chi^{-1})}a_{u_{E(\phi_{it})}}(m_{1})\overline{a_{u_{E(\phi_{it})}}(m_{2})}=w(\pi_{it}(\chi),f)\lambda_{\pi_{it}(\chi)}(m_{1})\overline{\lambda_{\pi_{it}(\chi)}(m_{2})} (1.7) where \mathcal{B}_{f}(\chi,\chi^{-1}) is a basis for V_{f}^{K_{\infty}} . With these further assumptions at hand, one may derive from Theorem 1.2.1 the following formula Theorem 1.2.2 (Refined BK formula) . Let f\in\mathcal{H} satisfy both the spectral and geometric assumptions. For m_{1},m_{2}\in\mathbb{Z} with m_{1}m_{2}<0 and (m_{1}m_{2},N)=1 we have that \displaystyle\sum_{\pi\in\mathcal{F}_{0}(f)}p(\pi)h_{\infty}(t_{\pi})w(\pi,f)\lambda_{\pi}(|m_{1}|)\overline{\lambda_{\pi}(|m_{2}|)} \displaystyle+\frac{1}{4\pi}\sum_{\chi\in\mathcal{F}_{E}(f)}p(\pi_{\chi,\chi^{-1}})\int_{-\infty}^{\infty}h_{\infty}(t) \displaystyle w(\pi_{it}(\chi),f)\lambda_{\pi_{it}(\chi)}(|m_{1}|)\overline{\lambda_{\pi_{it}(\chi)}(|m_{2}|)}dt (1.8) \displaystyle=\sum_{c\in\mathcal{C}(f)}\frac{H(m_{1},m_{2},c)}{c}H_{\infty}^{-}\left(\frac{4\pi\sqrt{|m_{1}m_{2}|}}{c}\right) as absolutely convergent sums/integrals. Here • N , \mathcal{F}_{0}(f),\mathcal{F}_{E}(f) , \mathcal{C}(f) , H(m,n,c) are as in Theorem 1.2.1 , • p(\pi) is the parity of \pi , as given in ( 2.5 ), • h_{\infty} is now a

Read the complete text