Some remarks on number theory. II (Q2524590)

[Part I, cf. Isr. J. Math. 3, 6--12 (1965; Zbl 0131.03902] According to the author's summary, this paper contains several disconnected remarks on number theory. The main results are: Theorem 1. Let \(f(k)\) be a real-valued arithmetical function, with \[ \lim_{n \to \infty} n^{-1} \sum_{k=1}^n f(k) = \alpha(\neq \pm\infty). \] Assume that for every \(\eta >0\), there is a \(g(\eta)\) so that for \(l > g(\eta)\) and \(n > 0\), \(l^{-1} \sum_{k=0}^{l-1} f(n+k) < \alpha+\eta\); then to every \(\varepsilon > 0\), \(\delta >0 \), there is an \(h(\varepsilon,\delta)\) so that for all but \(\varepsilon x\) integers \(n<x\), we have for every \(l > h(\varepsilon,\delta)\) that \[ \left|l^{-1} \sum_{k=0}^{l-1} f(n+k)-\alpha \right| < \delta. \] This generalizes (a strengthened form of) a result of \textit{R.Bellman} and \textit{H. N. Shapiro} [Duke Math. J. 21, 629--637 (1954; Zbl 0057.28602)]. Theorem 2. To every \(c_1\), there is a \(c_2 (c_1)\), so that if \(a_1 < a_2 < \cdots < a_k \leq n\) are integers, \(k > c_1n\), \(A = a_1a_2...a_n\), then \(\sum_{d \mid A} d^{-1}>c_2\log n\). The proof uses Brun's method. Also the following result (not stated as a formal theorem) is proved: Let \(a_1 < a_2 < \cdots < a_k \leq x\) be \(k\) integers such that no two of them are relatively prime, but every three are. If, for given \(x\), one sets \(\max k=f(x)\), then \(f(x)=({1 \over 2}+o(1))(\log x)/(\log\log x)\).