Bounded Exponential Sums with Multiplicative Coefficients

Ihor Pylaiev, Fred Tyrrell · 2025

Read Bounded Exponential Sums with Multiplicative Coefficients by Ihor Pylaiev, Fred Tyrrell (2025) 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-2506-12845.txt.

Introduction

Pierre-Alexandre Bazin and co-authors ask when wave sums with number weights stay bounded in a 2025 paper. The sums add a multiplicative weight at each count, spun by a fixed twist. Most such sums grow without bound as the count climbs.

Bound sums turn out to be rare and tightly shaped. The weight must sit very close to a known twisted character form. For weights with just a few values, the shape list sharpens more. Full multiplicative weights with a wild twist get a full sort. Bound for a fat set of twists pins the weight down yet more.

Catalog record

CREATORIhor Pylaiev, Fred Tyrrell
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE318 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2506-12845

Browse the full library · About the Tower of Babel

Full text

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

Bounded Exponential Sums with Multiplicative Coefficients Ihor Pylaiev, Fred Tyrrell Abstract . We investigate when the exponential sum ∑ n ≤ x f ⁡ ( n ) ​ e 2 ​ UNKNOWN ​ i ​ n ​ UNKNOWN \mathchar 4944\displaylimits_{\mathchar 29038\mathchar 12820\mathchar 29048}\mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785\mathchar 29029^{\mathchar 28722\mathchar 28953\mathchar 29033\mathchar 29038\mathchar 28939} is bounded, for a multiplicative function f \mathchar 29030 and α ∈ ℝ \mathchar 28939\mathchar 12850\mathbb{\mathchar 29010} . We show that under certain natural assumptions, this exponential sum is bounded only when f \mathchar 29030 closely resembles a periodic function. For completely multiplicative functions or those taking only finitely many values, we obtain sharper classification results, including a full classification (up to a measure-zero set of α \mathchar 28939 ) in the completely multiplicative case. 1. Introduction 1.1. Background and motivation Exponential sums involving multiplicative functions 1 1 1 A function \mathchar 29030\mathchar 12346\mathbb{\mathchar 29006}\mathchar 12833\mathbb{\mathchar 28995} is multiplicative if \mathchar 29030\delimiter 67273472\mathchar 29037\mathchar 29038\delimiter 84054785\mathchar 12349\mathchar 29030\delimiter 67273472\mathchar 29037\delimiter 84054785\mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785 for all coprime \mathchar 29037\mathchar 24891\mathchar 29038\mathchar 12850\mathbb{\mathchar 29006} . If \mathchar 29030\delimiter 67273472\mathchar 29037\mathchar 29038\delimiter 84054785\mathchar 12349\mathchar 29030\delimiter 67273472\mathchar 29037\delimiter 84054785\mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785 for all \mathchar 29037\mathchar 24891\mathchar 29038\mathchar 12850\mathbb{\mathchar 29006} , we say \mathchar 29030 is completely multiplicative. form a central object of study in analytic number theory, bridging additive and multiplicative structures in the integers. For a multiplicative function \mathchar 29030\mathchar 12346\mathbb{\mathchar 29006}\mathchar 12833\mathbb{\mathchar 28995} , we consider sums of the form \mathchar 29011_{\mathchar 29030}\delimiter 67273472\mathchar 29048\mathchar 24635\mathchar 28939\delimiter 84054785\mathchar 12346\mathchar 12349\mathchar 4944\displaylimits_{\mathchar 29038\mathchar 12820\mathchar 29048}\mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785\mathchar 29029\delimiter 67273472\mathchar 29038\mathchar 28939\delimiter 84054785\mathchar 24891 where \mathchar 29029\delimiter 67273472\mathchar 28939\delimiter 84054785\mathchar 12349\mathchar 29029^{\mathchar 28722\mathchar 28953\mathchar 29033\mathchar 28939} and \mathchar 28939\mathchar 12850\mathbb{\mathchar 29010} is fixed. Understanding the size and behaviour of exponential sums with multiplicative coefficients sheds light on how multiplicative functions interact with additive characters, with applications to questions of equidistribution modulo 1, the pseudorandomness of arithmetic functions, and the classification of multiplicative functions according to their additive correlations. A classical result of Daboussi [ Dab75 ] shows that if \mathchar 29030 is bounded and satisfies certain orthogonality conditions, then \mathchar 29011_{\mathchar 29030}\delimiter 67273472\mathchar 29048\mathchar 24635\mathchar 28939\delimiter 84054785\mathchar 12349\mathchar 29039\delimiter 67273472\mathchar 29048\delimiter 84054785 for all irrational \mathchar 28939 . Montgomery and Vaughan [ MV77 ] further developed general bounds for such sums, achieving a close to optimal upper bound for these sums on major arcs. Several generalisations and extensions of these results have since appeared, and exponential sums involving multiplicative functions remains an active area of research within analytic number theory; see, for example, the work of Daboussi-Delange [ DD82 ] , Kátai [ Kát86 ] , Bachmann [ Bac99 ] , and Matomäki-Radziwiłł-Tao-Teräväinen-Ziegler [ MRTTZ23 ] . More recently, Granville and Soundararajan [ GS14 ] formalised the orthogonality between additive characters and multiplicative functions via the pretentious approach , introducing a quantitative framework to measure the extent to which a multiplicative function mimics a structured object such as \mathchar 29038^{\mathchar 29033\mathchar 29044}\mathchar 28959\delimiter 67273472\mathchar 29038\delimiter 84054785 . This perspective unifies and generalises many earlier results by showing that cancellation in exponential sums is intimately tied to the ‘distance’ between \mathchar 29030 and such structured functions, and was recently used by de la Bretèche and Granville [ BG22 ] to continue the study of these exponential sums over minor arcs. This paper investigates the most extreme version of this cancellation phenomenon, namely when \mathchar 29011_{\mathchar 29030}\left\delimiter 67273472\mathchar 28939\mathchar 24891\mathchar 29048\right\delimiter 84054785 is bounded. This is motivated by work on discrepancy problems, in particular Tao’s solution to the Erdős Discrepancy Problem [ Tao16 ] . The Erdős Discrepancy Problem was a long-standing conjecture of Erdős [ Erd57 ] , which states that for any \mathchar 29030\mathchar 12346\mathbb{\mathchar 29006}\mathchar 12833\{\mathchar 8704\mathchar 28721\mathchar 24891\mathchar 8235\mathchar 28721\} , the discrepancy \sup_{\mathchar 29028\mathchar 24891\mathchar 29038}\left\delimiter 69640972\mathchar 4944\displaylimits_{\mathchar 29034\mathchar 12349\mathchar 28721}^{\mathchar 29038}\mathchar 29030\delimiter 67273472\mathchar 29028\mathchar 29034\delimiter 84054785\right\delimiter 69640972 is unbounded. One of the key ingredients of Tao’s solution was the breakthrough of Matomäki-Radziwiłł [ MR16 ] and a result from the first Polymath project, which showed that solving the problem in the case where \mathchar 29030 is completely multiplicative was sufficient to prove the full conjecture. Numerous investigations into the discrepancy of various arithmetic functions have since appeared in the literature, including [ Klu17 ] , [ KM18 ] , [ KMPT21 ] , [ KMT24 ] , [ Aym22 ] , [ Aym22a ] and [ Fra19 ] . For further background on Tao’s solution to the Erdős Discrepancy Problem, see the survey article of Soundararajan [ Sou18 ] . Particularly relevant to the work in this paper is the work of Frantzikinakis [ Fra19 ] on the weighted Erdős discrepancy problem, which extends Tao’s ideas by asking for which weight functions \mathchar 29047\mathchar 12346\mathbb{\mathchar 29006}\mathchar 12833\mathbb{\mathchar 28995} it is possible to find sequences \mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785\mathchar 12850\{\mathchar 8704\mathchar 28721\mathchar 24891\mathchar 8235\mathchar 28721\} such that the weighted discrepancy \sup_{\mathchar 29028\mathchar 24891\mathchar 29038}\left\delimiter 69640972\mathchar 4944\displaylimits_{\mathchar 29034\mathchar 12349\mathchar 28721}^{\mathchar 29038}\mathchar 29030\delimiter 67273472\mathchar 29028\mathchar 29034\delimiter 84054785\mathchar 29047\delimiter 67273472\mathchar 29034\delimiter 84054785\right\delimiter 69640972 remains bounded. In particular, Frantzikinakis demonstrated that the weighted discrepancy is unbounded when the weight function is of the form \mathchar 29047\delimiter 67273472\mathchar 29038\delimiter 84054785\mathchar 12349\mathchar 29029\delimiter 67273472\mathchar 29038^{\mathchar 29035}\mathchar 28939\delimiter 84054785 for \mathchar 29035\mathchar 12821\mathchar 28722 , using ergodic methods. The story when \mathchar 29035\mathchar 12349\mathchar 28721 , however, is rather different - indeed simply letting \mathchar 29030\mathchar 12349\mathchar 28721 and \mathchar 28939\mathrel{\vtop{\halign{#\cr$\hfil\textstyle\mkern 1.0mu\delimiter 1299214\hfil$\crcr$\textstyle\mathchar 562$\crcr}}}\mathbb{\mathchar 29018} , we have \left\delimiter 0{\mathchar 4944\displaylimits\limits_{\mathchar 29038\mathchar 12820\mathchar 29048}\mathchar 29029\left\delimiter 67273472\mathchar 29038\mathchar 28939\right\delimiter 84054785}\right\delimiter 0\left\delimiter 0{\mathchar 29029\left\delimiter 67273472\mathchar 28939\right\delimiter 84054785{{\mathchar 28721\mathchar 8704\mathchar 29029\left\delimiter 67273472\left\delimiter 0\mathchar 29048\right\delimiter 0\mathchar 28939\right\delimiter 84054785\over\mathchar 28721\mathchar 8704\mathchar 29029\left\delimiter 67273472\mathchar 28939\right\delimiter 84054785}}}\right\delimiter 0\mathchar 12820{{\mathchar 28722\over\left\delimiter 0{\mathchar 28721\mathchar 8704\mathchar 29029\left\delimiter 67273472\mathchar 28939\right\delimiter 84054785}\right\delimiter 0}}\mathchar 24891 so the sum can be bounded in this case. 1.2. Main results In this paper, we investigate when exponential sums \mathchar 29011_{\mathchar 29030}\delimiter 67273472\mathchar 29048\mathchar 24635\mathchar 28939\delimiter 84054785 can remain bounded as \mathchar 29048\mathchar 12833\mathchar 561 , for a fixed value of \mathchar 28939 . Our primary aim is to classify the multiplicative functions \mathchar 29030 for which \delimiter 69640972\mathchar 29011_{\mathchar 29030}\delimiter 67273472\mathchar 29048\mathchar 24635\mathchar 28939\delimiter 84054785\delimiter 69640972\mathchar 12828\mathchar 28721\mathchar 314 Our first main result provides a general classification of multiplicative functions with bounded exponential sums. 2 2 2 Here, and throughout, \mathbb{\mathchar 29013}\mathchar 12349\{\mathchar 29050\mathchar 12850\mathbb{\mathchar 28995}\mathchar 12346\left\delimiter 0{\mathchar 29050}\right\delimiter 0\} is the unit circle in \mathbb{\mathchar 28995} . We will also use \mathbb{\mathchar 29012}\mathchar 12349\mathbb{\mathchar 29010}\delimiter 68408078\mathbb{\mathchar 29018} to denote the torus. Theorem 1.1 . Let \mathchar 29030\mathchar 12346\mathbb{\mathchar 29006}\mathchar 12833\mathbb{\mathchar 29013}\mathchar 8795\{\mathchar 28720\} be a multiplicative function, and assume that \mathchar 4944\displaylimits_{\mathchar 29040\mathchar 12850\mathchar 29012}{{\mathchar 28721\over\mathchar 29040}}\mathchar 12604\mathchar 561 , where \mathchar 29012\mathchar 12349\{\mathchar 29040\mathchar 12346\mathchar 29030\left\delimiter 67273472\mathchar 29040^{\mathchar 29035}\right\delimiter 84054785\mathchar 12349\mathchar 28720\text{ for some }\mathchar 29035\} . If, for some \mathchar 28939\mathchar 12850\mathbb{\mathchar 29010} , \limsup\limits_{\mathchar 29048\mathchar 12833\mathchar 561}\left\delimiter 0{\mathchar 4944\displaylimits\limits_{\mathchar 29038\mathchar 12820\mathchar 29048}\mathchar 29030\delimiter 67273472\mathchar 29038\delimiter 84054785\mathchar 29029\delimiter 67273472\mathchar 29038\mathchar 28939\delimiter 84054785}\right\delimiter 0\mathchar 12604\mathchar 561\mathchar 24891 then there is a Dirichlet character \mathchar 28959 and a real number \mathchar 29044\mathchar 12850\mathbb{\mathchar 29010} such that \lim_{\mathchar 29040\mathchar 12833\mathchar 561}\max_{\mathchar 29035}\left\delimiter 0{\mathchar 29030\delimiter 67273472\mathchar 29040^{\mathchar 29035}\delimiter 84054785\mathchar 8704\mathchar 28959\delimiter 67273472\mathchar 29040^{\mathchar 29035}\delimiter 84054785\mathchar 29040^{\mathchar 29033\mathchar 29035\mathchar 29044}}\right\delimiter 0\mathchar 314 We also have the

Read the complete text