On practical partitions (Q1911745)

Let \({\mathcal A}= \{a_1= 1< a_2< \dots< a_k< \dots\}\) be an infinite subset of \(\mathbb{N}\). A partition of \(n\) with parts in \({\mathcal A}\) is a way of writing \(n= a_{i_1}+ a_{i_2}+ \dots+ a_{i_j}\) with \(1\leq i_1\leq i_2\leq \dots \leq i_j\). An integer \(a\) is said to be represented by the above partition, if it can be written \(a= \sum^j_{r=1} \varepsilon_r a_{i_r}\) with \(\varepsilon_r =0\) or 1. A partition will be called practical if all \(a\)'s, \(1\leq a\leq n\), can be represented. When \({\mathcal A}= \mathbb{N}\), it has been proved by P. Erdős and M. Szalay that almost all paritions are practical. In this paper, a similar result is proved, first when \(a_k= 2^k\), secondly when \(a_k\geq ka_{k-1}\). Finally an example due to D. Hickerson gives a set \({\mathcal A}\) and integers \(n\) for which a lot of non practical partitions do exist.