Average sizes of mixed character sums

Victor Y. Wang,Max Wenqiang Xu
2024-11-21
Abstract:We prove that the average size of a mixed character sum $$\sum_{1\le n \le x} \chi(n) e(n\theta) w(n/x)$$ (for a suitable smooth function $w$) is on the order of $\sqrt{x}$ for all irrational real $\theta$ satisfying a weak Diophantine condition, where $\chi$ is drawn from the family of Dirichlet characters modulo a large prime $r$ and where $x\le r$. In contrast, it was proved by Harper that the average size is $o(\sqrt{x})$ for rational $\theta$. Certain quadratic Diophantine equations play a key role in the present paper.
Number Theory,Probability
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the average size of mixed characters and (mixed character sums). Specifically, the author studies the mixed character sums of the following form: \[ \sum_{1 \leqslant n \leqslant x} \chi(n) e(n\theta) w\left(\frac{n}{x}\right) \] where: - $\chi$ is a Dirichlet character modulo a large prime $r$, - $e(n\theta) = e^{2\pi i n \theta}$ is an additive character, - $w$ is a suitable smooth function. ### Main problem The main problem of the paper is to determine the average size of the above - mentioned mixed character sums under different conditions. Specifically, the author hopes to prove that for an irrational number $\theta$ satisfying the weak Diophantine condition, when $x \leqslant r$, the average size of the mixed character sums is of the order of $\sqrt{x}$. This is in contrast to Harper's previous result, that is, for a rational number $\theta$, the average size is $o(\sqrt{x})$. ### Diophantine condition The weak Diophantine condition mentioned in the paper means that there exists a positive constant $C = C(\theta)$ such that for all natural numbers $q$, we have: \[ \| q\theta \| := \min_{n \in \mathbb{Z}} |q\theta - n| \geqslant C \exp\left(-q^{\frac{1}{4}}\right) \] This condition ensures that $\theta$ is not "too close" to a rational number. ### Main conclusion The main conclusion of the paper (Theorem 1.1) shows that for an irrational number $\theta$ satisfying the above - mentioned weak Diophantine condition, if $x$ is sufficiently large, then for all $x \leqslant r$, we have: \[ \sqrt{x} \ll \frac{1}{r - 1} \sum_{\chi \mod r} \left| \sum_{1 \leqslant n \leqslant x} \chi(n) e(n\theta) w\left(\frac{n}{x}\right) \right| \ll \sqrt{x} \] This means that the average size of the mixed character sums is approximately $\sqrt{x}$, and this result holds for all $x \leqslant r$. ### Research methods To prove this conclusion, the author uses the Poisson summation formula, exponential sum estimates, and some methods in combinatorial number theory. In particular, when dealing with the case of $x \geqslant \sqrt{r}$, by transforming the problem into a counting problem of quadratic Diophantine equations and using the pigeonhole principle to control the number of solutions. In summary, the main contribution of this paper is to reveal that the mixed character sums have different asymptotic behaviors in the case of irrational numbers compared to rational numbers, and to provide a rigorous mathematical proof.