A remark on a summation formula related with the sum of digits (Q2494359)

Let \(g\) be an integer and let \(s(n)\) be the sum of digits of \(n\) to the base of \(g\). Let \(f\) be a complex valued function. The main question of this paper is: Under what conditions on \(f\) does \(\sum_{n\leq x} f(n)=o(x)\) imply that \(F_g(x):=\sum_{0\leq n<x}f(s(n))=o(x)\)? The author gives sufficient conditions on \(f\). This involves oscillations of \(f\) on every segment of the form \([Y,Y+Z]\) for \(Z\) slightly (refer to the original paper for a precise statement) larger than \(\sqrt{Y}\). As an application, it is proved that for a positive real non-integer number \(\alpha\), the sequence \((s(n))^\alpha\) is uniformly distributed modulo \(1\) if and only if \(\alpha>1/2\). Further applications include the characteristic function of integers which are not divisible by any \(k\)-th power, the number of representations of an integer as sum of two squares, and the number of divisors of \(n\). Generalisations for strictly \(g\)-additive functions \(h(n)\) (instead of \(s(n)\)) are indicated.