A note on the Hurwitz zeta function
Let, as usual, \[ \zeta(\nu,a) = \sum_{n=0}^\infty (n+a)^{-\nu} \quad (0 < a \leq 1, \Re\nu > 1), \quad \text{Li}_\nu(z) = \sum_{k=1}^\infty z^k k^{-\nu} (\Re \nu > 1) \] denote the Hurwitz zeta-function and the polylogarithm function, respectively. For \(t \in \mathbb N\) and \(\omega = \exp(2\pi i/t)\) the authors derive the elementary identities \[ \zeta\left(\nu,{s\over t}\right) = {1\over t}\sum_{r=1}^t t^\nu\text{Li}_\nu(\omega^r) \omega^{-rs}\qquad(s = 1,2,\ldots,t)\tag{1} \] and \[ \text{Li}_\nu(\omega^r) = {1\over t^\nu} \sum_{s=1}^t \zeta\left(\nu,{s\over t}\right) \omega^{rs}\qquad(r = 1,2,\ldots,t).\tag{2} \] From (1) and (2) several corollaries are deduced, which involve various sums and the Bernoulli numbers. To prove (1) and (2) it suffices, by analytic continuation, to consider only \(\Re \nu > 1\). The proof of (2) is immediate from \[ \text{Li}_\nu(\omega^r) = \sum_{m=0}^\infty \sum_{s=1}^t {\exp(2\pi \text{ir}(tm+s)/t)\over(tm+s)^\nu} = {1\over t^\nu}\sum_{s=1}^t\exp\left({2\pi \text{ir}s\over t}\right) \sum_{m=0}^\infty{1\over\left(m + {s\over t}\right)^\nu}. \] Substituting the expression (2) in the right-hand side of (1) we obtain (1), with the remark that this is not correctly given on p. 50, since the ``orthogonality relationship \[ \sum_{r=1}^t\omega^{rs}\omega^{-rs} = \begin{cases} t &\text{if }r=s,\\ 0&\text{otherwise},\end{cases} \] is obviously not true (the dummy variable \(s\) in (2) is not to be confused with \(s\) in (1)).
This page was built for publication: A note on the Hurwitz zeta function
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2724052)
