A unitary test of the Ratios Conjecture (Q708261)

The goal of the paper under review is to show agreement between predictions made using the \(L\)-functions Ratios Conjecture and theoretical predictions, which are rigorous. The paper brings new evidence of the validity of the Ratios Conjecture, which is a very active topic of research [\textit{J. B. Conrey} and \textit{N. C. Snaith}, Proc. Lond. Math. Soc. (3) 94, No. 3, 594--646 (2007; Zbl 1183.11050)], [\textit{F. W. K. Firk} and \textit{S. J. Miller}, Nuclei, primes and the random matrix connection. Symmetry 1, 64--105 (2009); doi:10.3390/sym1010064], [\textit{D. K. Huynh, J. P. Keating} and \textit{N. C. Snaith}, J. Number Theory 129, No. 12, 2883--2902 (2009; Zbl 1205.11077)], [\textit{S. J. Miller}, Int. Math. Res. Not. 2008, Article ID rnm146, 36 p. (2008; Zbl 1225.11104), Proc. Lond. Math. Soc. (3) 99, No. 2, 484--520 (2009; Zbl 1170.11027)], [\textit{S. J. Miller} and \textit{D. Montague}, Acta Arith. 146, No. 1, 53--90 (2011; Zbl 1233.11054)]. The Ratios Conjecture was formulated by \textit{B. Conrey, D.W. Farmer} and \textit{M. R. Zirnbauer} [Commun. Number Theory Phys. 2, No. 3, 593--636 (2008; Zbl 1178.11056)] in an attempt to understand the arithmetic arising in the leading term of the moment calculations of \(L\)-functions. While not being the only way to tackle arithmetic statistics questions, it has the advantage of being both very versatile and precise. Indeed, while originally intended for studying moments of \(L\)-functions, the conjecture can also be used to make predictions for many zero statistics such as low-lying zeros, \(n\)-level correlation, mollified moments and moments of \(|\zeta'(\rho)|\) or \(|\zeta(\rho+a)|\) (see [Conrey and Snaith (loc. cit.)]). Moreover, it is believed that in most cases its predictions are accurate up to square-root cancellation. The present paper focuses on the family \(\mathcal F_q\) of Dirichlet \(L\)-functions \(L(s,\chi)\) with \(\chi\) nonprincipal modulo \(q\), with \(q\) a prime. The statistic of interest in the current paper is the one-level density of low-lying zeros of the \(L\)-functions in the above family. More precisely, the authors find an estimate for the quantity \[ D_{1,\mathcal F_q}(g):= \frac 1{|\mathcal F_q|}\sum_{\chi \in \mathcal F_q} \sum_{\gamma_{\chi} : L(\frac 12+i\gamma_{\chi},\chi)=0} g \left( \frac{\gamma_{\chi} \log(q/\pi)}{2\pi}\right), \] where \(\mathcal F_q(X):= \{ \chi \bmod q \text{ nonprincipal}\}\) and \(g\) is an even Schwartz function whose Fourier transform is compactly supported in \((-2,2)\). Under these conditions, they show that \[ D_{1,\mathcal F_q}(g) = \hat{g}(0) + \frac 1{(q-2) \log \frac q{\pi}} \int_{\mathbb R} g(t) \sum_{\chi \in \mathcal F_q} \left[ \frac{\Gamma'}{\Gamma}\left(\frac 14+ \frac{a(\chi)}2+ \frac{\pi i \tau}{\log \frac q{\pi}} \right)\right] d\tau + O_{\epsilon}(q^{\frac{\sigma}2 -1 +\epsilon}), \] where \(\sigma\) is the supremum of the support of \(\hat g\). The authors also show how their result agree with the Ratios Conjecture's prediction, which as they show implies that whatever the support of \(\hat g\) is, we have \[ D_{1,\mathcal F_q}(g) = \hat{g}(0) + \frac 1{(q-2) \log \frac q{\pi}} \int_{\mathbb R} g(t) \sum_{\chi \in \mathcal F_q} \left[ \frac{\Gamma'}{\Gamma}\left(\frac 14+ \frac{a(\chi)}2+ \frac{\pi i \tau}{\log \frac q{\pi}} \right)\right] d\tau + O_{\epsilon}(q^{-\frac 12+\epsilon}). \] It is remarkable that when the support of \(\hat g\) is contained in the interval \((-1,1)\), their unconditional result is more precise than the Ratios Conjecture's prediction.