Averages of Character Sums

From MaRDI portal





Let \(\chi\) be a primitive character, mod \(k\), \(k>0\). Let \[ S(m)=\sum_{0 \leq l \leq m} \chi(l)\quad\text{and } T(n)=\sum_{0 \leq m<n} S(m). \] Hua proved that \(|T(n)| \leq \frac{n}{2}\left(k^{\frac{1}{2}}-n k^{-\frac{1}{2}}\right)\) if \(\chi(-1)=1\) and \(0 \leq n \leq k\). By means of this result, the authors give a very simple proof of a result due to Davenport, namely, for \( 0<\sigma = \Re(s)<1\), then \[ |L(s, \chi)| \leq \frac{|s(s+1)|}{\sigma(1-\sigma)} k^{\frac12 (1-\sigma)} \quad\text{for }\chi(-1)=1. \] In case \(\chi(-1)=-1\), instead of Hua's result they obtain the inequality \[ \left|T(n) - \frac{\tau(\chi)}{\pi i} L(1, \chi) n\right| < n \left(2.1+\frac{\pi}{4} \log \frac{1}{\alpha}\right) k^{\frac{1}{2}}, \] where \(\alpha k \le n \le k\). But it is not sharp enough to give a new proof of Davenport's theorem for the case \(\chi(-1)=-1\). By means of a result due to Chowla, i. e. for \( \chi(-1)=-1\) \[ \varlimsup_{k \rightarrow \infty} \frac{L(1, \chi)}{\log \log k} \ge e^{\gamma}, \] they deduce also \[ \varlimsup_{k \rightarrow \infty} \frac{\tau(k)}{k^{\frac12} \log \log k} \ge \frac{e^{\gamma}}{\pi}, \] which is an improvement of a theorem due to Paley.











This page was built for publication: Averages of Character Sums

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5800861)