Functions with integer-valued divided differences (Q2071700)

It has been shown by \textit{R. R. Hall} [Mathematika 18, 71--77 (1971; Zbl 0226.10019)]. Cf. \textit{I. Ruzsa} [Mat. Lapok 22, 125--134 (1971)] that if a function \(s(n)\) is integer-valued, satisfies \[ |s(n)|\ll \theta^n\quad\quad (*) \] for some \(\theta<e-1\), and for all integers \(n,k>0\) one has \[ s(n+k)\equiv s(n)\pmod k, \] then \(s(n)\) is a polynomial. This is the case \(m=1\) of the following Theorem 1: If the \(m\)th divided difference \(\delta_ms\) of \(s(n)\) is integer-valued, satisfies (*) for some \(\theta < \exp(H_m)-1\) with \[ H_m=\sum_{k=1}^m\frac1k, \] then \(s(n)\) is a polynomial. The proof is based on local analysis and utilizes a generalization of the classical theorem of \textit{K. Mahler} [J. Reine Angew. Math. 199, 23--34 (1958; Zbl 0080.03504)] to bounded functions which are not continuous, established by the author (Theorem 2). Theorem 4 establishes the formula \[ \sum_{p\le n}\tau_{p.m}(n)\log p = nH_m + O\left(n\exp(-\alpha\sqrt{\log n})\right), \] where \(\tau_{p,m}(n)\) is the maximal power of a prime \(p\) dividing the product of \(m\) distinct positive integers not exceeding \(n\).