English

The frequency and the structure of large character sums

Number Theory 2020-06-29 v2

Abstract

Let M(χ)M(\chi) denote the maximum of nNχ(n)|\sum_{n\le N}\chi(n)| for a given non-principal Dirichlet character χ(modq)\chi \pmod q, and let NχN_\chi denote a point at which the maximum is attained. In this article we study the distribution of M(χ)/qM(\chi)/\sqrt{q} as one varies over characters (modq)\pmod q, where qq is prime, and investigate the location of NχN_\chi. We show that the distribution of M(χ)/qM(\chi)/\sqrt{q} converges weakly to a universal distribution Φ\Phi, uniformly throughout most of the possible range, and get (doubly exponential decay) estimates for Φ\Phi's tail. Almost all χ\chi for which M(χ)M(\chi) is large are odd characters that are 11-pretentious. Now, M(χ)nq/2χ(n)=2χ(2)πqL(1,χ)M(\chi)\ge |\sum_{n\le q/2}\chi(n)| = \frac{|2-\chi(2)|}\pi \sqrt{q} |L(1,\chi)|, and one knows how often the latter expression is large, which has been how earlier lower bounds on Φ\Phi were mostly proved. We show, though, that for most χ\chi with M(χ)M(\chi) large, NχN_\chi is bounded away from q/2q/2, and the value of M(χ)M(\chi) is little bit larger than qπL(1,χ)\frac{\sqrt{q}}{\pi} |L(1,\chi)|.

Keywords

Cite

@article{arxiv.1410.8189,
  title  = {The frequency and the structure of large character sums},
  author = {Jonathan Bober and Leo Goldmakher and Andrew Granville and Dimitris Koukoulopoulos},
  journal= {arXiv preprint arXiv:1410.8189},
  year   = {2020}
}

Comments

58 pages, 4 figures, 1 table. Minor changes, and small improvement of Theorem 2.2. Final version, to appear in J. Eur. Math. Soc