Completely multiplicative functions with sum zero over generalised prime systems (Q2221687)

In 1903, Landau discovered another proof of the Prime Number Theorem which used relatively weak information about integers. Thirty years later, this gave Beurling the idea that the PNT could be established for a system of `generalized integers`, which are defined as follows: A \textit{generalized prime system}, also called \(g\)-prime system, is a non-decreasing sequence \(\mathcal{P}\) of real numbers \[1 \leqslant p_1 \leqslant p_2 \leqslant \dotsb \leqslant p_n \leqslant \dotsb\] such that \(p_n \to \infty\) as \(n \to \infty\). From these can be formed the system \(\mathcal{N}\) of \textit{Beurling generalized integers}, or \(g\)-integer system, defined as numbers of the shape \(p_1^{a_1} \dotsb p_n^{a_n}\), where \(a_k \in \mathbb{Z}_{\geqslant 1}\). In [Expo. Math. 35, 364--389 (2017; Zbl 1405.11050)], \textit{J.-P. Kahane} and \textit{E. Saias} studied the so-called CMO-functions, i.e., complex-valued completely multiplicative functions \(f\) satisfying \(\sum_{n=1}^\infty f(n) = 0\), and among other things considered the question of the rate of convergence to zero of the partial sums \(\sum_{n \leqslant x} f(n)\). In the paper under review, the author generalizes the work of Kahane and Saias by considering the CMO\(_{\mathcal{P}}\)-functions, i.e. completely multiplicative functions \(f : \mathcal{N} \to \mathbb{C}\) satisfying \(\sum_{n \in \mathcal{N}} f(n) = 0\), where \(\mathcal{N}\) is a \(g\)-integer system. The author investigates several properties of such functions, and also asks about the rate of convergence to zero of the partial sums. For instance, it is proved that, if \(\mathcal{P}\) is a \(g\)-prime system such that all \(g\)-primes are distinct and such that the series \(\sum_{p \in \mathcal{P}} \frac{1}{p \log p}\) converges, then, for all CMO\(_{\mathcal{P}}\)-functions \(f\), the estimate \[\sum_{\substack{n \leqslant x \\ n \in \mathcal{N}}} f(n) = \Omega \left( \frac{1}{x \log x}\right) \] holds. For specific CMO\(_{\mathcal{P}}\)-functions, this result can be refined. For instance, defining the Generalized Liouville function \(\lambda_{\mathcal{P}}\) in an obvious way, the author shows that, if \(\mathcal{P}\) is a \(g\)-prime system such that there exists \(\rho > 0\) such that \[\sum_{\substack{n \leqslant x \\ n \in \mathcal{N}}} 1 = \rho x + O \left( x^\beta \right)\] for some \(\beta < \frac{1}{2}\), then \[\sum_{\substack{n \leqslant x \\ n \in \mathcal{N}}} \frac{\lambda_{\mathcal{P}} (n)}{n} = \Omega \left( \frac{1}{x^{1/2}}\right).\] At the end of the paper, the author surmises that, if \(\mathcal{P}\) is a \(g\)-prime system with abscissa of convergence \(\sigma_c = 1\), then, for all CMO\(_{\mathcal{P}}\)-functions \(f\) \[\sum_{\substack{n \leqslant x \\ n \in \mathcal{N}}} f(n) = \Omega \left( \frac{1}{x^{1/2}}\right).\]