On the mean square formula for the Riemann zeta-function on the critical line (Q1321531)

For \(T\geq 2\) \textit{J. L.Hafner} and the reviewer [J. Number Theory 32, 151-191 (1989; Zbl 0668.10045)] defined the function \(G(t)\) by \[ G(t) =\int_ 2^ T (E(t)- \pi) dt, \qquad E(T)= \int_ 0^ T \Biggl| \zeta \Biggl( {1\over 2}+ it \Biggr)\Biggr|^ 2 dt-T \log \Biggl( {T\over {2\pi}}\Biggr)- (2\gamma-1)T \] and established a mean square formula for it. The reviewer [Acta Arith. 56, 135-159 (1990; Zbl 0659.10053)] proved that \(u_{n+1}- u_ n \ll u_ n^{1/2}\) and that \(u_{n+1}- u_ n \gg u_ n^{1/2} (\log u_ n)^{-5}\) holds for infinitely many \(n\), where \(u_ n\) is the \(n\)-th zero of \(G(T)\). It was also pointed out that \[ u_{n+1}- u_ n \gg u_ n^{1/2}\tag{1} \] holds for infinitely many \(n\) if a sharp mean square formula for \(E(T)\) holds. The author establishes now the best possible bound (1) unconditionally by showing that \[ \int_ T^{2T} \biggl( G^ 2(t)- \max_{0\leq h\leq H} (G(t+h)-G(t))^ 2\biggr) dt>0, \qquad H=\delta T^{1/2} \] with suitable small \(\delta\).