Remarks on the mean values of multiplicative functions (Q2467595)

Let \(f\) denote a real-valued multiplicative function and \(M(f)\) its mean value when this exists. Theorem 1.1 of this paper establishes that if \(M(f^2)\) exists then so does \(M(f)\), and moreover that \(M(f)= 0\) if the series \(\sum_p {(f(p)- 1)^2\over p}\) diverges. Under certain conditions, including that \(M(f^2)\neq 0\), an upper bound for \({1\over x}\sum_{n\leq x}f(n)\) is obtained in Theorem 1.2. The result of Theorem 1.1 was proved for positive valued multiplicative functions \(f\) by \textit{P. D. T. A. Elliott} in [J. Aust. Math. Soc., Ser. A 30, 461--468 (1980/81; Zbl 0469.10024)], and he gave another proof in Theorem 19.1 of his book [Duality in analytic number theory, Cambridge Tracts in Mathematics. 122. Cambridge: Cambridge University Press (1997; Zbl 0887.11002)]. The proof of the theorem in this paper is more direct and self-contained, and depends on several lemmas. The first of these establishes a new property on the local regularity of the function \(M(x;f)= \sum_{n\leq x}f(n)\). This is then used to obtain the key result that \(\sum_p \sum_{\nu\geq 2} {f(p^\nu)^2\over p^\nu}\) is finite when \(M(f^2)\) exists and \(\limsup_{x\to\infty}\, {M(x;f)\over x}\)is positive. Let \(\tau_0(n)= |\tau(n)|n^{-11/2}\), where \(\tau(n)\) denotes Ramanujan's function; this function has been studied by \textit{R. A. Rankin} who showed that \(M(\tau^2_0)\neq 0\), by \textit{P. D. T. A. Elliott} (loc. cit.) and by several other authors. As a corollary to the main topic, the present author improves the previously known upper bound for \(\sum_{n\leq x}\tau_0(n)\) from \(x(\log x)^{-1/18}\) to \(x(\log x)^{-s}\), where \(s\approx 0.11852\) and is precisely defined.