arXiv reader

On algebraic twists with composite moduli

Yongxiao Lin and Philippe Michel

arXiv
2304.08149v1 · PDF · 17 April 2023
Categories
math.NT
Licence
CC BY 4.0 The site shows the text.

Abstract

We study bounds for algebraic twists sums of automorphic coefficients by trace functions of composite moduli.

Items 16

Theorem

Theorem 1.1

N2OW

Let f(z) be a Hecke eigencuspform of level 1 either holomorphic of weight k\geqslant 2 or a Laplacian eigenform with spectral parameter t_f; let (\lambda_f(n))_{n\geqslant 1} be its Hecke eigenvalues. Let q=q_0q_1 be a product of two distinct primes and let

K_0:\mathbf{Z}/{q_0}\mathbf{Z}\mapsto \mathbf{C},\ K_1:\mathbf{Z}/{q_1}\mathbf{Z}\mapsto \mathbf{C}

be two complex valued functions which we identify with functions on \mathbf{Z} of period q_0 and q_1 respectively; we assume that K_0 is the trace function attached to an \ell-adic middle extension Fourier sheaf \mathcal{F} on \mathbf{A}^1_{\mathbf{F}_{q_0}}, geometrically irreducible, pure of weight 0 and such that

  • The automorphism group (see [FKMS, § 7.1]) of the Fourier transform sheaf \widehat\mathcal{F} is trivial.

  • (MO) (see [LMS, § 1]) There is no \lambda \in {\mathbf{F}^\times_q} such that the geometric monodromy group of \mathcal{F} has some quotient which is equal, as a representation of the geometric fundamental group \pi_1 into an algebraic group, to the geometric monodromy group of the Kloosterman sheaf [\times \lambda]^*\mathcal{K}\ell_2 modulo \pm 1.

Let K be the q-periodic function given by

K(\cdot)=K_0(\cdot)K_1(\cdot).

Let Z\geqslant 1 be some parameter and V\in \mathcal{C}^\infty_c(\mathbf{R}) be a smooth function compactly supported in the interval [1,2[ satisfying for all j\geqslant 0,

V^{(j)}(x)\ll_j Z^j.

For X\geqslant 1 we have the bound

\sum_{n=1}^{\infty}\lambda_f(n)K(n)V\left(\frac{n}{X}\right)\ll \|\widehat{K_1}\|_{\infty} X^{o(1)}\left( Z^{1/2}X^{1/2} q_0^{1/2}+ZX^{1/2}q^{1/2}{q_0}^{-1/4}+ Zq^{1/2}{q_0}^{1/4}\right),

where the implicit constant depends at most on f, on the conductor C_0=C(\mathcal{F}) of \mathcal{F} and on the implicit constants in (?).

Remark

Remark 1.2

8OCG

Examples of traces functions K_0 whose associated sheaf \mathcal{F} satisfy the conditions of Theorem 1.1 include the hyper-Kloosterman sums \operatorname{\mathrm{Kl}}_d(\cdot;q_0) for d\not=2.

Remark

Remark 1.3

NWJA

In particular, under the above assumptions on K, for Z=1 and

q_0= q^{2/3+o(1)}

from (?) we see that the bound (?) holds as long as

X\geqslant q^{2/3+\delta}

for any given \delta>0. The range (for d=2)

q^{2/3}=X_{\mathrm{cv}}^{2/3}>X_{\mathrm{ap}}=q^{1/2}

is sometimes called the Weyl range. It can be related to the work of Heath-Brown [H-B78] who obtained a Weyl-type subconvexity bound L(\chi,1/2)=O\big(q^{1/6+o(1)}\big) for the Dirichlet L-functions provided that q has a factor q_0 of size q_0\approx q^{2/3}.

Remark

Remark 1.4

VJF1

It is possible to improve this range further assuming that q_0 is squarefree and suitably factorable (under additional assumptions on K_0) by using the recent work of Wu, Xi, and Sawin [PW] which extends Heath-Brown’s q-van der Corput’s method to general trace functions.

Remark

Remark 1.5

0SGP

In [FKM1] the bound (?) was obtained for q prime in the range (?) under the sole assumption that \mathcal{F} is a Fourier sheaf (not geometrically isomorphic to the constant sheaf or any Artin–Schreier sheaf). Theorem 1.1 holds in this generality as well, but to simplify the exposition, we have chosen to make this extra assumption regarding the automorphism group of \widehat\mathcal{F}.

Theorem

Theorem 1.6

XC7A

Let f(z) be a Hecke eigencuspform of level 1 either holomorphic of weight k\geqslant 2 or a Laplacian eigenform with spectral parameter t_f; let (\lambda_f(n))_{n\geqslant 1} be its Hecke eigenvalues. Let \varphi be a Hecke–Maass cuspform for \operatorname{GL}_{3,\mathbf{Q}} of level 1 with Fourier coefficients (\lambda_\varphi(n,r))_{n\geqslant 1,r\not=0}.

Let q=q_0q_1 be a product of two distinct primes and let

K_0:\mathbf{Z}/{q_0}\mathbf{Z}\mapsto \mathbf{C},\ K_1:\mathbf{Z}/{q_1}\mathbf{Z}\mapsto \mathbf{C}

be two complex valued functions which we identify with functions on \mathbf{Z} of period q_0 and q_1 respectively; we assume moreover that K_0 is the trace function attached to an \ell-adic middle extension Fourier sheaf \mathcal{F} on \mathbf{A}^1_{\mathbf{F}_{q_0}}, geometrically irreducible and pure of weight 0, of conductor C_0 and which is good in the sense of [LMS, § 1]. Let K be the q-periodic function given by

K(\cdot)=K_0(\cdot)K_1(\cdot).

Let V\in \mathcal{C}^\infty_c(\mathbf{R}) be a smooth function satisfying (?). Let X\geqslant 1 be such that X\geqslant Z^4q^2q_0^{1/2}. Then

\begin{gathered} \sum_{n\geqslant 1,r\not=0}\lambda_\varphi(n,r)\lambda_f(n)K(nr^2)V\left(\frac{nr^2}{X}\right)\ll_{\varphi,f,\|K_1\|_\infty,C_0}X^{o(1)}Z^2\Bigg(X^{3/4}{q_0}^{3/4}\\ + X^{\frac{2-\theta_3}{3-2\theta_3}}(q^2q_0^{1/2})^{\frac{1-\theta_3}{3-2\theta_3}}+\frac{X}{{q_0}^{1/4}}+ \frac{X^{3/4}q}{q_0^{1/2}}\Bigg). \end{gathered}

Here \theta_3=5/14 is the best known bound towards the Ramanujan–Petersson conjecture on \operatorname{GL}_3.

Remark

Remark 1.7

8BWR

In particular for

q_0= q^{4/5+o(1)}

we obtain that

\sum_{n\geqslant 1,r\not=0}\lambda_\varphi(n,r)\lambda_f(n)K(nr^2)V\left(\frac{nr^2}{X}\right)\ll_{\varphi,f,\|K_1\|_\infty,C_0}X^{o(1)}Z^2\Bigg(X^{3/4}q^{3/5}+X^{\frac{2-\theta_3}{3-2\theta_3}}q^{\frac{12(1-\theta_3)}{5(3-2\theta_3)}}+\frac{X}{q^{1/5}}\Bigg)

and when Z=1 this bound is non-trivial (i.e., (?) holds) as soon as

X\geqslant q^{12/5+\delta}=q^{5/2-1/10+\delta}

for some \delta>0, in which case the second factor inside the parentheses can be removed.

Corollary

Corollary 1.8

EYPM

Notations be as above. Assume that

q_0= q^{4/5+o(1)}.

We have for any (a,q)=1,

\sum_{\stackrel{{\scriptstyle n,|r|\geqslant 1}} {{\scriptstyle nr^2\equiv a\bmod q}}}\lambda_\varphi(n,r)\lambda_f(n)V\left(\frac{nr^2}{X}\right)\ll_{\varphi,f} X^{o(1)}\left(X^{1/4}q^{8/5}+q^{23/10}\right).

In particular, given \eta>0, we have

\sum_{\stackrel{{\scriptstyle n,|r|\geqslant 1}} {{\scriptstyle nr^2\equiv a\bmod q}}}\lambda_\varphi(n,r)\lambda_f(n)V\left(\frac{nr^2}{X}\right)\ll_{\varphi,f,\eta} (\frac{X}{q})^{1-\delta}

for some \delta=\delta(\eta)>0 as long as

q\leqslant X^{2/7+1/364-\eta}=X^{\theta_6+1/364-\eta}.
Remark

Remark 1.9

8N4U

The exponent 1/364 matches the one presented in [Lin-Sun, Cor. 1.4]. In that work, Q. Sun and the first named author obtained the following error term for the “sharp cut" sum of the coefficients \sum_{mr^2=n}\lambda_\varphi(m,r)\lambda_f(m)

\sum_{nr^2\leqslant X}\lambda_\varphi(n,r)\lambda_f(n)=O\left(X^{\tau_6-1/364+o(1)}\right)

where

\tau_d=\frac{d-1}{d+1}

is the “standard" exponent of Landau (see [FICan, Prop. 1.1]) for sharp-cut sums of coefficients of automorphic L-functions of degree d.

Remark

Remark 2.1

XLWH

Choosing the parameter q_0 in place of q is one chief difference between this paper and [AHLS].

Notation

Notation

X42J

To lighten the expressions to come we write

A\lesssim B\hbox{ in place of }A\leqslant X^{o(1)}B.
Remark

Remark 2.2

7XBL

Estimating trivially (assuming the sum over \alpha\bmod q_0 in (?) is bounded above by O(q_0^{1/2})) and ignoring the error terms, we have

\begin{split} S(X)\ll& \frac{X^{2+o(1)}}{P^3{q_0}^2q^{1/2} }\frac{p^2{q_0}^2}{X}P\frac{Zpq}{X}q_0^{1/2} \ll X^{o(1)}ZPq^{1/2}q_0^{1/2}. \end{split}

In view of the constraint (?), this bound is insufficient to beat the trivial bound O(X) yet.

Remark

Remark 2.3

XIBK

As can be expected, if the modulus q=q_0q_1 of K(\cdot) admits further factorisation, then it is possible to apply a q-analogue of van der Corput’s argument to the (r_1,r_2)-sum to obtain further improvements of our final result, cf. see also Remark 1.4.

Remark

Remark 2.4

B1IH

(1) The parameter P does not play an essential role in the final bounds (cf. see also [AHLQ]).

(2) Similar results can also be proved using an amplified second moment approach as in [Bykovskii] and [FKM1].

Lemma

Lemma 4.1

2J8O

We have the following estimates

\mathrm{FT}(0;k)\ll \sqrt{k}rc_1c_2\mathop{\sum_{d|c_1c_2}\sum_{d'|c_1c_2}}_{(d,d')|(m -m')} (d ,d'),

and

\begin{split} \mathrm{FT}(n;k)\ll \sqrt{k}\sum_{d_1|c_1}d_1\sum_{d'_1|c_1}d'_1 \mathop{\sideset{}{^\star}\sum_{\stackrel{{\scriptstyle x_1(rc_1/n_1)}} {{\scriptstyle q^2_1 n_1x_1\equiv \mp m\bmod {d_1}}}}}\mathop{\sum\sum}\limits_{\stackrel{{\scriptstyle d_2|(c_2,q^2_1 n_1c'_2+nm)}} {{\scriptstyle d'_2|(c'_2,q^2_1 n_1c_2+nm')}}}d_2d'_2. \end{split}
Proposition

Proposition 4.2

HX1L

Let T_\mathcal{F}({\mathbf{F}_q}) be the subgroup of {\mathbf{F}^\times_q} defined by

T_\mathcal{F}({\mathbf{F}_q})=\{\lambda\in{\mathbf{F}^\times_q},\ [\times\lambda]^*\mathcal{F}\hbox{ is geometrically isomorphic to }\mathcal{F}\}.

Assuming that the sheaf \mathcal{F} is good, then for any \alpha,\beta,\alpha',\beta',\gamma,\gamma',\delta\in{\mathbf{F}^\times_q}, we have

\sum_{v}Z(v)\overline{Z'(v-\delta)}=O(q^{1/2}).

If \delta=0 the above bound holds unless

\alpha/\alpha'=\beta\gamma/\beta'\gamma'\in T_\mathcal{F}({\mathbf{F}_q})

in which case

\sum_{v}Z(v)\overline{Z'(v)}=c_\mathcal{F}(\alpha/\alpha')q+O(q^{1/2})

for c_\mathcal{F}(\alpha/\alpha') some complex number of modulus 1. Here the implicit constants depend only on C(\mathcal{F}).

Source and licence

The text on this page is from arXiv:2304.08149v1 by Yongxiao Lin and Philippe Michel, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2304/2304.08149.jsonl.