No label defined (Q4798615)

The Lerch zeta-function with real parameters \(\lambda,\alpha>0\) is given by \[ L(\lambda,\alpha,s)=\sum_{m=0}^\infty {\exp(2\pi i\lambda m)\over (m+\alpha)^s}. \] In the present paper the mean square with respect to the parameter \(\alpha\) is studied. Define \[ I(s,\lambda)=\int_0^1|L(\lambda,\alpha,s)-\alpha^{-s}|^2 d\alpha \] and put \(\tilde{\zeta}(s,\lambda)=\exp(2\pi i\lambda)L(\lambda,1,s)\). The author proves for sufficiently large \(t\) \[ \begin{matrix} I(1/2+it,\lambda)&=&\gamma+2\Re\left\{\tilde{\zeta}'(\lambda,0)- \tilde{\zeta}(\lambda,0){\Gamma'\over \Gamma}(1/2+it)\right\}\\ &&-2\Re{\exp(-2\pi i\lambda)\tilde{\zeta}(\lambda,1/2+it)-1\over 1/2+it}+O(1/t), \end{matrix} \] and a similar formula for \(s=1+it\); the derivative is taken with respect to \(s\). These formulae generalize a nice result for the Hurwitz zeta-function (integral \(\lambda\)) due to \textit{J. Andersson} [Math. Scand. 71, 295-300 (1992; Zbl 0815.11042)], and its improvement by \textit{M. Katsurada} and \textit{K. Matsumoto} [Proc. Japan Acad., Ser. A 69, 303-307 (1993; Zbl 0799.11027)]. The method of proof relies on a representation of \(I(s,\lambda)\) by hypergeometric functions.