Finitely additive measures on \(\mathbb{N}\) (Q5905647)

This long and carefully written paper presents a detailed study of finitely additive probability measures \(\mu\) on the power set of \(\mathbb{N}\), the set of positive integers. Numerous properties of such measures are considered (such as scale-invariant, stretchable, thinnable, extending density, etc.) and there is the usual interminable series of lemmas and examples serving to illustrate and relate the properties. Among these, the most interesting is the notion of an elastic measure, which is defined by the following two properties: (1) For all \(K\subseteq \mathbb{N}\), for all \(n\in\mathbb{N}\), we have \(\mu(nK)=n^{-1}\mu(K)\), where \(nK=\{nx: x\in K\}\). (2) For all functions \(f,g: \mathbb{N}\to\mathbb{N}\) such that \(f(n)\leq g(n)\) for all \(n\in\mathbb{N}\), we have \(\mu(f(\mathbb{N}))\geq\mu(g(\mathbb{N}))\). The author shows that the set \({\mathcal E}\) of all elastic measures on \(\mathbb{N}\) is nonempty, and scrutinizes the need for the axiom of choice in the proof. It is widely accepted that the density \(d(K)\) of a set \(K\subseteq\mathbb{N}\), defined by \(d(K)=\lim_{n\to\infty}[| K\cap[1,n]|/n]\), where \(| S|\) denotes the number of elements in a set \(S\), provides the best ``intuitively correct'' measure of the ``size'' of \(K\). But what is the ``size'' of a set \(K\) for which the above limit does not exist? If \(\mu\) is elastic, then \(\mu(K)=d(K)\) whenever \(d(K)\) is defined. The author argues that \(\mu(K)\) may be the best ``intuitively correct size'' of \(K\) when \(K\) is such that \(\mu(K)=\nu(K)\) for all \(\mu,\nu\in{\mathcal E}\).