Low-lying zeros of quadratic Dirichlet L-functions, hyper-elliptic curves and Random Matrix Theory

TL;DR

This paper proves that the distribution of low-lying zeros of quadratic Dirichlet L-functions matches predictions from Random Matrix Theory by using a function field analogue and taking limits over finite fields and genus.

Contribution

It establishes the equivalence between the low-lying zeros distribution and RMT predictions for all n by employing hyper-elliptic curve analogues and limit processes, extending previous partial results.

Findings

Distribution of zeros matches RMT predictions for all n

Function field analogue confirms combinatorial factors

Large genus limit aligns with Random Matrix Theory

Abstract

The statistics of low-lying zeros of quadratic Dirichlet L-functions were conjectured by Katz and Sarnak to be given by the scaling limit of eigenvalues from the unitary symplectic ensemble. The n-level densities were found to be in agreement with this in a certain neighborhood of the origin in the Fourier domain by Rubinstein in his Ph.D. thesis in 1998. An attempt to extend the neighborhood was made in the Ph.D. thesis of Peng Gao (2005), who under GRH gave the density as a complicated combinatorial factor, but it remained open whether it coincides with the Random Matrix Theory factor. For n at most 7 this was recently confirmed by Levinson and Miller. We resolve this problem for all n, not by directly doing the combinatorics, but by passing to a function field analogue, of L-functions associated to hyper-elliptic curves of given genus g over a field of q elements. We show that the answer in this case coincides with Gao's combinatorial factor up to a controlled error. We then take the limit of large finite field size q to infinity and use the Katz-Sarnak equidistribution theorem, which identifies the monodromy of the Frobenius conjugacy classes for the hyperelliptic ensemble with the group USp(2g). Further taking the limit of large genus g to infinity allows us to identify Gao's combinatorial factor with the RMT answer.