SearcharxivSearch

arXiv · 2609.21195

The large sieve for square moduli under Hooley's hypothesis $R^*$

Abstract

Let $S(Q,M,N,(a_n)):=\sum_{q\le Q}\sum_{(a,q)=1}|\sum_{M<n\le M+N}a_ne(an/q^2)|^2$ be Zhao's large sieve sum with square moduli. At the critical point $N=Q^3$ the best known unconditional bound, due to Baier and Zhao (2008), is $S\ll Q^{1/2+\varepsilon}N\sum |a_n|^2$, against the conjectured $Q^{\varepsilon}N\sum|a_n|^2$, and the exponent $\tfrac12$ has not been lowered since. We prove that, under Hooley's Hypothesis $R^*$ for short Salié sums -- square-root cancellation for $\sum_{x_1<n\le x_2}\big(\tfrac nc\big)e_c(a\bar n+bn)$ over arbitrary subintervals of a period -- one has $S\ll Q^{1/2-1/134+\varepsilon}N\sum|a_n|^2$ at $N=Q^3$. The key estimate is a bound for the number $P(α)$ of fractions $a/q^2$, $q\le Q$, within $Q^{-3}$ of a point $α$ near $b/r$: we show $P(b/r+z)\ll(Q^{2/3}r^{-1/3}+Q^{1/4})Q^\varepsilon$ for every modulus $Q^{1/2+\varepsilon}\le r\le Q^{3/2}$, improving the bound $Q^{9/16}r^{-1/8}$ obtained by Baier (2026) for $r=p,p^2$ only, and reaching every modulus. The proof rests on a single observation: a sum of modular square roots $\sum_{n\in J}e_r(a\sqrt{jn})$ over an interval $J$ is, after completion and an exact evaluation of quadratic Gauss sums at every modulus, $r^{-1/2}$ times a Salié sum of length $r/|J|$. Hypothesis $R^*$ therefore yields square-root cancellation for these sums directly, at every modulus, without Weyl differencing; the saving over the trivial bound is the square of what the Weyl-differencing route gives. The Gauss-sum evaluations, including even moduli and coefficients sharing a factor with the modulus, are proved in full. The paper was prepared in collaboration with Claude (Anthropic); Section 1.9 sets out what each of us contributed.

Explore related subjects

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Stephan Baier. 2026-09-18. The large sieve for square moduli under Hooley's hypothesis $R^*$. https://arxiv.org/abs/2609.21195

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

On single-variable Witten zeta functions of rank two and three

By introducing a novel integration kernel for the Mellin transform, we uncover many previously unknown and intriguing properties of the Witten zeta functions of rank two and three. Detailed results concerning their pole locations, residues, and special values are obtained. We propose a non-trivial conjecture regarding their derivatives at the origin, which seems to encode deep information about the root system. We also discuss their behavior at negative integers, highlighting a connection with Eisenstein series and a $p$-adic observation.

math.NT

On graded Lie algebras associated to once-punctured elliptic curves with complex multiplication

We study a graded Lie algebra arising from the Galois action on the pro-$p$ fundamental group of a once-punctured elliptic curve with complex multiplication. Among other things, we provide a minimal generating set of the rationalized Lie algebra under suitable assumptions. The proof is based on a slight variant of the theory of weighted completion of profinite groups developed by Hain and Matsumoto.

math.NT

Burgess-type volume dependent bounds for character sums over $\mathbb{F}_{p^n}$

We establish a Burgess-type bound for short multiplicative character sums over finite fields $\mathbb{F}_{p^n}$. Let \[ B=\left\{\sum_{i=1}^{n}x_iω_i: N_i+1\le x_i\le N_i+H_i,1\le i\le n\right\}\subseteq\mathbb{F}_{p^n}, \] where $1\le H_i\le p$ for all $1\le i\le n$, and the side lengths satisfy $H_1\le H_2\le\cdots\le H_n.$ We prove that if the side lengths satisfy certain lower bounds in terms of the two largest side lengths, then a nontrivial cancellation occurs in the character sum over the boxes. This generalizes the work of Gabdullin \cite{GB} in dimensions $n=2,3$ to arbitrary dimension. This also generalizes the character sum estimate of Konyagin \cite{Kon} where each of the side lengths of the boxes are greater than $p^{1/4}$. The proof combines techniques from the geometry of numbers, multiplicative energy estimates, and Katz's bounds for multiplicative character sums.

math.NT