Hilbert spaces and low-lying zeros of \(L\)-functions (Q2099089)

This paper gives a general framework for estimating the frequency of low-lying zeros of \(L\)-functions, under GRH. This question is inspired by the pioneering work of \textit{H. Iwaniec} et al. [Publ. Math., Inst. Hautes Étud. Sci. 91, 55--131 (2000; Zbl 1012.11041)]. As a sample result, consider Dirichlet \(L\)-functions to prime modulus \(q\). For fixed \(t>0\) it is shown that \[ \frac{1}{q-2}\sum_{\substack{\chi(\mathrm{mod}\; q)\\ \chi\not=\chi_0}} \mathrm{ord}_{s=\tfrac12} L\left(s+\frac{2\pi it}{\log q},\,\chi\right) \le\frac12\left(1+\frac{|\sin 4\pi t|}{4\pi t}\right)^{-1}+ O_t\left(\frac{1}{\log q}\right). \] Thus for any fixed \(t\) at most a proportion \(\tfrac12+o(1)\) of the functions vanish, and for \(t\to 0^+\) the proportion is at most \(\tfrac14+o(1)\). (For \(t=0\) the method only gives a proportion \(\le\tfrac12+o(1)\), reproducing a result of \textit{V. A. Lyubetskij} [Algebra Logic 30, No. 6, 427--439 (1991; Zbl 0803.03034); translation from Algebra Logika 30, No. 6, 652--670 (1991)].) The paper handles general families of \(L\)-functions, and tackles the corresponding question of low-lying zeros via estimates for the 1-level density. The shape of the final result therefore depends on the symmetry type of the family. The available information about 1-level densities involves a weight function whose support has to lie in a restricted range, and one wishes to chose this weight function optimally. This lead to the central idea of the paper, which is to solve this optimization problem using reproducing kernels for appropriate Hilbert spaces. The authors find the reproducing kernels (for weight functions with the relevant support) for each of the various symmetry types.