Abstract:We consider the moment of quadratic twists of the Möbius function of the form \[
S_k(X,Y) = \sum_{d\leq X} \left( \sum_{n\leq Y} \left(\frac{8d}{n}\right) \mu(n)\right)^k, \] where $\left(\frac{8d}{\cdot}\right)$ is the Kronecker symbol and $d$ runs over positive, odd and square-free integers. We give unconditional results for their asymptotic behaviors.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the asymptotic behavior of the moments of the Möbius function with quadratic twists. Specifically, the author studies the quantity of the following form:
\[
S_k(X,Y) = \sum_{d \leq X} \left( \sum_{n \leq Y} \left( \frac{8d}{n} \right) \mu(n) \right)^k,
\]
where $\left( \frac{8d}{n} \right)$ is the Kronecker symbol, $d$ is a positive, odd and square - free integer, and $\mu(n)$ is the Möbius function.
### Research Background and Motivation
1. **Mertens Function and Riemann Hypothesis**
- The Mertens function is defined as $M(x) := \sum_{n \leq x} \mu(n)$.
- Classical results show that the Mertens function is closely related to the Riemann Hypothesis (RH). In fact, RH is equivalent to $M(x) \ll x^{1/2+\varepsilon}$ for any $\varepsilon > 0$.
- Under GRH (Generalized Riemann Hypothesis), there have been some studies on the Mertens function and its generalizations, such as the work of Halupczok and Suger [8].
2. **Generalized Mertens Function**
- For a Dirichlet character $\chi$ modulo $q$, the corresponding generalized Mertens function is defined as $M(x,\chi) := \sum_{n \leq x} \mu(n)\chi(n)$.
- Similarly, it is related to the GRH of the corresponding Dirichlet L - function.
3. **Moments of the Möbius Function with Quadratic Twists**
- Gao and Zhao [6] studied the case when $k = 2$ and introduced smoothed summation.
- The goal of this paper is to give an asymptotic formula for $S_k(X,Y)$ without assuming GRH, especially under the condition $Y \ll X^{1/3-\varepsilon}$.
### Main Results
- **Theorem 1.2**: For large $X$ and $Y$, when $Y \ll X^{1/3-\varepsilon}$, we have
\[
S_2(X,Y) = \frac{4}{\pi^2} Z_2 XY+O(X Y^{7/12})+O(X^{1/2+\varepsilon} Y^{2+\varepsilon})+O(X^{1/2} Y^{5/2}(\log Y)^2).
\]
- **Theorem 1.3**: For large $X$ and $Y$, when $Y \ll X^{2/3k-\varepsilon}$, we have
\[
S_k(X,Y) = \frac{4}{\pi^2} Z_k XY^{k/2} Q(\log Y)+O(X Y^{k/2-\theta})+O(X^{1/2+\varepsilon} Y^{k+\varepsilon})+O(X^{1/2} Y^{5k/4}(\log Y)^k),
\]
where $Q\in\mathbb{R}[x]$ is a monic polynomial of exact degree $\binom{k}{2}-k$, $\theta > 0$, and $Z_k$ is given by a specific Euler product.
### Methods and Tools
- **Multivariate Tauberian Theorem**: It is used to deal with the asymptotic behavior of multivariate arithmetic functions.
- **Estimates of Rudnick and Soundararajan**: They are used to deal with the estimates of quadratic characters.
- **Perron Formula**: It is used for the specific calculations when $k = 2$.
Through these methods, the author is able to give the Möbius with quadratic twists without relying on GRH.