On a conjecture of Erdős and certain Dirichlet series (Q2346691)

Erdős conjectured (reported by \textit{A. E. Livingston} [Can. Math. Bull. 8, 413--432 (1965; Zbl 0129.02801)]) that for a periodic arithmetic function \(f\) with period \(q \geq 2\) such that \(f(n) =\pm 1\) if \(q\neq n\) and \(f(q) =0\), the sum \[ \sum_{n=1}^{\infty} \frac{f(n)}{n} \neq 0. \] There have been several attempts to resolve this conjecture, but till date it is open. Since \(\sum_{n=1}^{\infty} \frac{f(n)}{n}\) converges if and only if \(\sum_{a=1}^{q} f(a) = 0\), one sees that Erdös conjecture is true when \(q\) is even. When \(q\) is prime, this conjecture is a consequence of a work of \textit{A. Baker} et al. [J. Number Theory 5, 224--236 (1973; Zbl 0267.10065)]. Later, this conjecture was proved for \(q\)'s having exactly two distinct odd prime factors by \textit{T. Okada} [Acta Arith. 40, 143--153 (1982; Zbl 0402.10035)], for completely multiplicative periodic functions by \textit{R. Tijdeman} [in: Number theory for the millennium III. Proceedings of the millennial conference on number theory, Urbana-Champaign, USA, 2000. Natick, MA: A K Peters. 261--284 (2002; Zbl 1045.11022)] and for odd \(q \equiv 3\pmod 4\) by the second author and \textit{N. Saradha} [J. Number Theory 130, No. 12, 2671--2682 (2010; Zbl 1204.11114)]. In this nice article, the authors use a criterion of Okada as modified by \textit{N. Saradha} and \textit{R. Tijdeman} [J. Lond. Math. Soc., II. Ser. 67, No. 3, 580--592 (2003; Zbl 1045.11051)] to show that the falsity of Erdös' conjecture for an odd modulus \(q\) implies that \[ \sum_{ d\mid q \atop d \geq 3} \frac{1}{\varphi(d)} \geq 1. \] Thereafter, using power moment calculations of \(\sum_{ d\mid q \atop d \geq 3} \frac{1}{\varphi(d)} \geq 1\), they deduce that Erdös conjecture is true for 82\% of the integers \(q \equiv 1 \pmod 4\).