On the zeros of \(\zeta'(s)\) near the critical line (Q1847884)

This paper improves some results and brings some new results concerning the zeros of \(\zeta'(s)\). The standard notation is used: \(s=\sigma + it\), \(T\) is a large parameter, \(L=\log T\), \(\rho ' = \beta ' + i\gamma '\) denotes a zero of \(\zeta'(s)\), and \(N_{1}(T)\) denotes the number of zeros of \(\zeta'(s)\) in \(0<t\leq T\) counted according to multiplicity. Let for \(\nu \in \mathbb R\), \(m^{-}(\nu)\) and \(m^{+}(\nu)\) (resp.) be the \(\liminf\) and \(\limsup\) (resp.) as \(T\to\infty\) of \[ {1\over N_{1}(T)}\sum_{_{\substack{ 0 < \gamma ' \leq T \\ {\beta ' \leq {1\over 2} + {\nu\over L}}}}}1. \] An old result due to Speiser says that the Riemann Hypothesis (RH) is equivalent to \(\zeta'(s)\) having no zeros in \(0 < \sigma < {1\over 2}\). Other known facts about this matter are: \(m^{+}(\nu)\to 0\) as \(\nu\to -\infty\) (Levinson and Montgomery); \(m^{+}(\nu) < 1\) for all \(\nu\) (Conrey and Ghosh); RH implies \(m^{-}(\nu)> 0\) for \(\nu>0\) (Soundararajan). It was conjectured by Soundararajan that \(m^{-}(\nu) \equiv m^{+}(\nu) (=m(\nu))\), \(m(\nu)\) is continuous, \(m(\nu)>0\) for all \(\nu >0\), and \(m(\nu)\to 1\) as \(\nu\to\infty\). In this paper the author proves unconditionally that \(m^{-}(\nu)>0\) for sufficiently large \(\nu\), and if RH and a conjecture named SGZ are assumed then \(m^{-}(\nu)>0\) for any \(\nu>0\). The SGZ conjecture asserts that there is a positive proportion of gaps of size any fixed positive constant times the average spacing between the ordinates of the zeros of \(\zeta(s)\). Assuming RH, it is also proved that, qualitatively speaking, two zeros of \(\zeta(s)\) close to each other will induce a zero of \(\zeta'(s)\) to be located nearby. To describe this quantitatively, let \(\alpha_{1}\) and \(\alpha_{2}\) be positive constants, \(\alpha_{1} <2\pi, \alpha_{2}> \alpha_{1}(1-\sqrt{{\alpha_{1}\over 2\pi}})^{-1}\). If \({1\over 2}+i\gamma\) is a zero of \(\zeta(s)\) with \(\gamma\) sufficiently large, \({1\over 2}+i\gamma^{+}\) is the zero with the next larger ordinate, and if \(\gamma^{+}-\gamma < \alpha_{1}(\log \gamma)^{-1}\), then there exists \(\rho '\) such that \( \rho ' -\rho <\alpha_{2}(\log \gamma)^{-1}\). The proofs are based on the expression of the logarithmic derivative of \(\pi^{-{s\over 2}}\Gamma({s\over 2})\zeta'(s)\) in terms of the zeros of \(\zeta'(s)\) by means of Hadamard's theorem for entire functions, the use of the argument principle for counting zeros, a result of Selberg and of Fujii that there exists a positive proportion of gaps between the ordinates of consecutive zeros of \(\zeta(s)\) which are larger than the average gap. A well-known result of Levinson and Montgomery, stating that the number of zeros of \(\zeta(s)\) and of \(\zeta'(s)\) in \(0<\sigma< {1\over 2}, 0<t<T\) differ by at most \(O(\log T)\), is also invoked to take care of the possible presence of zeros to the left of the critical line.