Fourier optimization, the least quadratic non-residue, and the least prime in an arithmetic progression

Emanuel Carneiro,Micah B. Milinovich,Emily Quesada-Herrera,Antonio Pedro Ramos

2024-04-12

Abstract:By means of a Fourier optimization framework, we improve the current asymptotic bounds under GRH for two classical problems in number theory: the problem of estimating the least quadratic non-residue modulo a prime, and the problem of estimating the least prime in an arithmetic progression.

Number Theory,Classical Analysis and ODEs

What problem does this paper attempt to address?

This paper attempts to solve the asymptotic bound estimation in two classical number theory problems, as follows:

1. **The least quadratic non - residue problem**:
- Problem description: For an odd prime \(p\), find the least quadratic non - residue \(n_p\) modulo \(p\). That is, find the smallest positive integer \(n_p\) such that for all \(a\in\mathbb{Z}\), \(a^{2}\not\equiv n_p\pmod{p}\).
- Currently known results: Under the Generalized Riemann Hypothesis (GRH), the best asymptotic bound currently is \(n_p\leq(0.794 + o(1))\log^{2}p\).

2. **The least prime in an arithmetic progression problem**:
- Problem description: Given natural numbers \(q\) and integers \(a\) satisfying \(\gcd(a, q)=1\), find the least prime \(P(a, q)\) of the form \(a\pmod{q}\).
- Currently known results: Under the GRH, the best asymptotic bound currently is \(P(a, q)\leq(1 + o(1))(\varphi(q)\log q)^{2}\), where \(\varphi\) is the Euler function.

The main contribution of the paper is to improve the asymptotic bound estimations of the above two problems under the GRH by introducing the Fourier optimization framework. Specifically:

- For the least quadratic non - residue problem, the author proves that under the GRH, there exists a constant \(C(1)\) such that
\[\limsup_{p\rightarrow\infty}\frac{n_p}{\log^{2}p}\leq C(1)^{-2}< 0.7615.\]

- For the least prime in an arithmetic progression problem, the author proves that under the GRH, there exists a constant \(C(3)\) such that
\[\limsup_{q\rightarrow\infty}\frac{P(a, q)}{(\varphi(q)\log q)^{2}}\leq C(3)^{-2}<\frac{8}{9}.\]

These results show that under the GRH assumption, the author's method can provide tighter asymptotic bound estimations than existing methods. In addition, the paper also discusses related Fourier optimization problems and gives specific numerical bounds, further verifying the validity of these results.