On large spacing of the zeros of the Riemann zeta-function (Q2444994): Difference between revisions

Let \(\zeta(s)\) denote the Riemann zeta-function, and let \(0<\gamma\leq\gamma'\) denote the ordinates of consecutive non-trivial zeros of \(\zeta(s)\). It is well-known that \(\gamma'-\gamma\sim 2\pi/\log \gamma\). Define \[ \lambda=\limsup_{\gamma}\frac{\gamma'-\gamma}{2\pi/\log \gamma}\quad\text{and}\quad \mu=\liminf_{\gamma}\frac{\gamma'-\gamma}{2\pi/\log \gamma}. \] It is widely believed that \(\lambda=\infty\) and \(\mu=0\). That is to say there exist arbitrarily large (normalised) gaps and arbitrarily small (normalised) gaps between consecutive zeros of the Riemann zeta-function. In this paper, the authors focus on the large gaps. In 1940's, Selberg remarked that he could prove \(\lambda>1\) unconditionally. Here the authors assume the Generalised Riemann Hypothesis and prove that \[ \lambda > 3.072. \] This improves the previous results of \textit{N. Ng} [J. Number Theory 128, No. 3, 509--556 (2008; Zbl 1182.11038)] and of the reviewer [J. Number Theory 131, No. 1, 67--95 (2011; Zbl 1269.11078)] under the same hypothesis. These papers all use the following idea of \textit{J. Mueller} [J. Number Theory 14, 327--331 (1982; Zbl 0483.10035)]. Let \(H:\mathbb{C}\rightarrow\mathbb{R}_{\geq 0}\) be continuous and let \[ M_1(H,T)=\int_{0}^{T}H(1/2+it)\,dt, \] \[ m(H,T;\alpha)=\sum_{0<\gamma\leq T}H\big(1/2+i(\gamma+\alpha)\big), \] \[ M_2(H,T;c)=\int_{-c/L}^{c/L}m(H,T;\alpha)\,d\alpha, \] where \(L=\log\frac{T}{2\pi}\). Then we have \(\lambda>c/\pi\) provided that \[ \frac{M_2(H,2T;c)-M_2(H,T;c)}{M_1(H,2T)-M_1(H,T)}<1. \] The authors obtain \(\lambda > 3.072\) by choosing a new function \(H(s)\). It is not intuitively clear at the first glance why this choice of \(H\) will give a better lower bound for \(\lambda\) though.