A connection between covers of the integers and unit fractions (Q2643870)

For integers \(a\) and \(n>0\), let \(a(n)\) denote the residue class \(\{x\in \mathbb Z: x=a \pmod n\}\). Let \(A\) be a collection \(\{a_s(n_s)\}_{s=1}^k\) of finitely many residue classes such that \(A\) covers all the integers at least \(m\) times but \(\{a_s(n_s)\}_{s=1}^{k-1}\) does not. Then the author proves that if \(n_k\) is a period of the covering function \(w_A(x)=|\{1\leq s\leq k: x\in a_s(n_s)\}|\) then for any \(r=0,\dots,n_k-1\) we have \[ \left| \left\{\left\lfloor \sum_{s \in I}\frac{1}{n^s}\right\rfloor: I\subseteq [1,k-1]\;\text{and}\;\left\{\sum_{s \in I}\frac{1}{n^s}\right\}=\frac{r}{n_k}\right\}\right| \geq m, \] where \(\lfloor \alpha \rfloor\) and \(\{ \alpha \}\) denote, respectively, the integral part and the fractional part of \(\alpha\). Several remarks are given: \(n_k\) need not be the largest modulus among \(n_1,\dots,n_k\); for \(m=1\), \(n_k=N_A\) (\(N_A\) denotes the least common multiple of the moduli \(n_1,\dots,n_k\)) the theorem is an easy consequence of a previous paper of the author [Acta Arith. 72, No. 2, 109--129 (1995; Zbl 0841.11011)] (as observed by the author's twin brother Z. H. Sun) and finally a stronger result can be proved when \(w_A(x)=m\) for all \(x\in \mathbb Z\).