Michael J. Curran
Abstract:Assuming the Riemann hypothesis, we investigate the shifted moments of the zeta function \[ M_{\alpha,{\beta}}(T) = \int_T^{2T} \prod_{k = 1}^m |\zeta(\tfrac{1}{2} + i (t + \alpha_k))|^{2 \beta_k} dt \] introduced by Chandee, where ${\alpha} = {\alpha}(T) = (\alpha_1, \ldots, \alpha_m)$ and ${\beta} = (\beta_1 \ldots , \beta_m)$ satisfy $|\alpha_k| \leq T/2$ and $\beta_k\geq 0$. We shall prove that \[ M_{{\alpha},{\beta}}(T) \ll_{\beta} T (\log T)^{\beta_1^2 + \cdots + \beta_m^2} \prod_{1\leq j < k \leq m} |\zeta(1 + i(\alpha_j - \alpha_k) + 1/ \log T )|^{2\beta_j \beta_k}. \] This improves upon the previous best known bounds due to Chandee and Ng, Shen, and Wong, particularly when the differences $|\alpha_j - \alpha_k|$ are unbounded as $T \rightarrow \infty$. The key insight is to combine work of Heap, Radziwiłł, and Soundararajan and work of the author with the work of Harper on the moments of the zeta function.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the estimation of the shifted moments of the Riemann zeta function. Specifically, under the assumption that the Riemann hypothesis holds, the author studies the following form of integral:
\[ M_{\alpha, \beta}(T) = \int_T^{2T} \prod_{k = 1}^m |\zeta(1/2 + i(t + \alpha_k))|^{2\beta_k} dt \]
where \(\alpha=\alpha(T)=(\alpha_1,\ldots,\alpha_m)\) and \(\beta = (\beta_1,\ldots,\beta_m)\) satisfy \(|\alpha_k|\leq T/2\) and \(\beta_k\geq0\).
### Research Background
1. **Chandee's Work**:
- Chandee [4] first studied the general form of such shifted moments and gave a lower bound when \(\beta_k\) are integers and \(\alpha_k = O(\log\log T)\).
- When \(\alpha_j-\alpha_k = O(1)\), Chandee also gave an upper bound under the assumption that the Riemann hypothesis holds. These upper bounds are optimal within a loss of \((\log T)^\varepsilon\).
2. **Ng, Shen, and Wong's Work**:
- Ng, Shen, and Wong [22] removed the loss of \((\log T)^\varepsilon\), especially in the case of \(\beta = (\beta,\beta)\), using the research results of Harper [11] on the moments of the zeta function.
- They also gave the bounds when \(|\alpha_1+\alpha_2|\leq T^{0.6}\).
### Main Contributions of the Paper
1. **Improved Upper Bound**:
- The author proves that, under the assumption that the Riemann hypothesis holds, for \(\alpha\) and \(\beta\) satisfying \(|\alpha_k|\leq T/2\) and \(\beta_k\geq0\), the following inequality holds:
\[
M_{\alpha, \beta}(T)\ll_\beta T (\log T)^{\beta_1^2+\cdots+\beta_m^2} \prod_{1\leq j < k\leq m} |\zeta(1 + i(\alpha_j - \alpha_k)+1/\log T)|^{2\beta_j\beta_k}
\]
- This result improves the previously best - known bounds of Chandee [4] and Ng, Shen, and Wong [22], especially when \(|\alpha_j - \alpha_k|\) is unbounded.
2. **Key Insight**:
- The author combines the work of Heap, Radziwiłł, and Soundararajan [12] and Harper [11], as well as his own previous work [7], and proposes new methods to deal with these shifted moments.
### Conclusion
The main goal of the paper is to extend the work of Ng, Shen, and Wong [22] to arbitrary \(\alpha\) and \(\beta\) under the assumption that the Riemann hypothesis holds, and to give stronger bounds, especially when \(|\alpha_j - \alpha_k|\) is unbounded. These results are not only theoretically significant but also provide new tools and methods for further studying the properties of the Riemann zeta function.