Abstract:We prove a result on the large deviations of the central values of even primitive Dirichlet $L$-functions with a given modulus. For $V\sim \alpha\log\log q$ with $0<\alpha<1$, we show that \begin{equation}\nonumber\frac{1}{\varphi(q)} \# \left\{\chi \text{ even, primitive mod }q: \log \left|L\left(\chi,\frac{1}{2}\right)\right| >V\right\}\ll \frac{e^{-\frac{V^2}{\log\log q}}}{\sqrt{\log\log q}}.\end{equation} This yields the sharp upper bound for the fractional moments of central values of Dirichlet $L$-functions proved by Gao, upon noting that the number of even, primitive characters with modulus $q$ is $\frac{\varphi(q)}{2}+O(1).$ The proof is an adaptation to the $q$-aspect of the recursive scheme developed by Arguin, Bourgade and Radziwill for the local maxima of the Riemann zeta function, and applied by Arguin and Bailey to the large deviations in the $t$-aspect. We go further and get bounds on the case where $V=o(\log\log q)$. These bounds are not expected to be sharp, but the discrepancy from the Central Limit Theorem estimate grows very slowly with $q$. The method involves a formula for the twisted mollified second moment of central values of Dirichlet $L$-functions, building on the work of Iwaniec and Sarnak.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is about the upper bound of large deviations in the q - aspect of the central values of Dirichlet L - functions. Specifically, the author studied the probability that the logarithm of the central value of even primitive Dirichlet L - functions with a given modulus \(q\) exceeds a certain threshold \(V\), and gave an upper bound estimate of this probability.
### Main Problem Description
For \(V\sim\alpha\log\log q\) (where \(0 < \alpha < 1\)), the author proved the following result:
\[
\frac{1}{\phi(q)}\#\left\{\chi\text{ even primitive mod }q:\log\left|L\left(\chi,\frac{1}{2}\right)\right| > V\right\}\ll e^{-\frac{V^{2}}{\log\log q}}\sqrt{\log\log q}
\]
This result shows that when \(V\) is in the range of \(\alpha\log\log q\), the large deviations of the central values of even primitive Dirichlet L - functions can be upper - bounded by the Gaussian tail.
### Research Background
Selberg's central limit theorem states that the normalized distribution of the logarithm of the Riemann zeta function on the critical line is asymptotically Gaussian. Similarly, many studies are devoted to generalizing Selberg's theorem to other types of L - functions and exploring their central value distributions.
### Main Contributions of the Paper
1. **Upper Bound of Large Deviations**: The author successfully proved the upper bound of large deviations in the q - aspect of Dirichlet L - functions through a recursive scheme and a twisted mollifier formula.
2. **Upper Bound of Fractional Moments**: As a corollary, the author also obtained the upper bound of the fractional moments of the central values of Dirichlet L - functions, which is consistent with Gao's result.
3. **Improved Bounds**: By using an improved method, the author can also obtain a reasonable upper bound estimate when \(V = o(\log\log q)\).
### Method Overview
The author used a recursive scheme similar to that developed by Arguin, Bourgade and Radziwiłł for the Riemann zeta function, combined with the twisted mollifier formula of Iwaniec and Sarnak. Specific steps include:
- Approximating the logarithmic L - function with Dirichlet polynomials.
- Analyzing the variances of these polynomials.
- Using mollifier techniques to control deviations.
- Estimating through Euler integral formulas and orthogonality relations.
These methods work together to enable the author to accurately estimate the large - deviation behavior of the central values of Dirichlet L - functions.
### Conclusion
This paper provides new insights into the large deviations in the q - aspect of Dirichlet L - functions through rigorous mathematical proofs and lays the foundation for future research. In particular, these results are helpful for understanding the non - zero properties of L - functions and other related problems.