A weighted one-level density of families of \(L\)-functions (Q6071805)

\textit{N. M. Katz} and \textit{P. Sarnak} [Bull. Am. Math. Soc., New Ser. 36, No. 1, 1--26 (1999; Zbl 0921.11047); Random matrices, Frobenius eigenvalues, and monodromy. Providence, RI: American Mathematical Society (1999; Zbl 0958.11004)] describe the \textit{Density Conjecture}, that the distributions of zeros in many natural families of \(L\)-functions seem to coincide with eigenvalue distributions of certain groups of random matrices. One aspect concerns ``one-level density'' for a family \(\mathcal{F}\) of \(L\)-functions, where the Density Conjecture would predict that \[ \lim_{X \to \infty} \frac{1}{\# \mathcal{F}_X} \sum_{L \in \mathcal{F}_X} \sum_{\gamma_L} f(c_L \gamma_L) \stackrel{?}{=} \int_{\mathbb{R}} f(x) W_{\mathcal{F}}(x) dx. \] Here, the inner sum is over imaginary parts of nontrivial zeros of \(L\) (denoted \(\gamma_L\)); the outer sum is over \(L\)-functions \(L\) in the family \(\mathcal{F}\) with log-conductor at \(1/2\) bounded by \(X\) (denoted by \(L \in \mathcal{F}_X\)); the constants \(c_L\) act to normalize the spacings of zeros; the function \(f\) is an even Schwartz function of rapid decay; and \(W_{\mathcal{F}}\) on the right is a one-level density function on an appropriate matrix group. This can be stated more succinctly as stating that the distribution of low-lying zeros (conjecturally) agrees with the distribution of low eigenvalues of random matrices in appropriate families. In this paper, the author studies a weighted version of this one-level density. The author weights (or \textit{tilts}, as stated in the paper) by certain functions \(V\) of the central value \(L(1/2)\), specifically \[ \mathcal{D}_k^{\mathcal{F}}(f, X) := \frac{1}{ \sum_{L \in \mathcal{F}_X} V\big( L(\tfrac{1}{2}) \big)^k } \sum_{L \in \mathcal{F}_X} \sum_{\gamma_L} f(c_L \gamma_L) V\big(L(\tfrac{1}{2})\big)^k. \] The weighting function \(V\) depends on the symmetry type of the family, but in all cases it has the effect of emphasizing \(L\)-functions where \(L(1/2)\) is large, where low-lying zeros should generically be less common. Broadly, this paper extends the Density Conjecture to a \textit{Weighted Density Conjecture}, asserting that \(\mathcal{D}_k^{\mathcal{F}}(f)\) should also agree with a \textit{weighted} average over random matrices. In support of this new conjecture, the author investigates three families of \(L\)-functions: the unitary family \(\{ \zeta(\tfrac{1}{2} + it) : t \in \mathbb{R} \}\), the symplectic family of quadratic Dirichlet \(L\)-functions, and the orthogonal family of quadratic twists of \(L(s, \Delta)\), the \(L\)-function associated to the discriminant modular form. Assuming the associated Riemann hypotheses and the Ratios Conjectures (of \textit{B. Conrey} et al. [Commun. Number Theory Phys. 2, No. 3, 593--636 (2008; Zbl 1178.11056)]), this paper proves a weighted form of the Density Conjecture for \(\mathcal{D}_k^{\mathcal{F}}\) for weight exponents \(k \leq 4\). The author also includes unconditional proofs of random matrix analogs. Finally, in sections 2A and 2B, the author gives conjectured shapes of the analogous results for all \(k \in \mathbb{N}\) and describes the matrix weights in terms of hypergeometric functions near \(0\). The structure of the proofs build on ideas introduced in the work of Conrey, Farmer, and Zirnbauer, as well as on strategies using the Ratios Conjecture described by \textit{J. B. Conrey} and \textit{N. C. Snaith} [Proc. Lond. Math. Soc. (3) 94, No. 3, 594--646 (2007; Zbl 1183.11050)].