Conglomerated filters and statistical measures (Q2069797)

The paper under review is an interdisciplinary paper which deals with measure-theoretical problems connected to statistical convergence, from the point of view of analysis, and with filters and ultrafilters, from a set-theoretical point of view. Ultrafilters have been introduced by \textit{F.~Riesz} [``Stetigkeitsbegriff und abstrakte Mengenlehre'', Atti del IV Congresso Intern. Matem., Roma 1908, II, Roma, 18--24 (1909; JFM 40.0098.07)], see [\textit{H. L. Bentley} et al., in: Handbook of the history of general topology. Volume 2. Dordrecht: Kluwer Academic Publishers. 577--629 (1998; Zbl 0936.54028)], but have not received due attention until they were rediscovered by the French school. In the terminology of the paper under review, a \emph{statistical measure} is a non-negative finitely additive measure \(\mu \) defined on the collection of all subsets of $\mathbb N$ and such that \( \mu (\mathbb N) = 1\) and \( \mu (\{ k \} ) = 0\), for all \(k \in \mathbb N\). The name is motivated by the notion of statistical convergence, which plays an important role in mathematical analysis, measure theory and functional analysis. Notice that a \(\{ 0,1\}\)-valued statistical measure corresponds to a free ultrafilter over \(\mathbb N\). As pointed out by the authors, statistical measures, without using this name, have been considered earlier by other scholars working in axiomatic set theory, model theory and descriptive set theory. The \emph{filter generated by a statistical measure} \(\mu \) is the collection of all subsets \(A \subseteq \mathbb N\) such that \(\mu (A)=1\). With the aim of characterizing such filters, the authors introduce the notions of a \emph{poor} and of a \emph{conglomerated} filter. They show that every filter generated by a statistical measure is poor, that every conglomerated filter is not poor, and that there is a filter which is neither poor nor conglomerated. A filter with the Baire property is conglomerated, hence it is not generated by a statistical measure. Many examples of statistical measures are known such that the corresponding filter cannot be represented as a countable intersection of ultrafilters. In Section~3, the authors study intersections of families of ultrafilters. If some filter can be represented as a finite intersection of ultrafilters, the representation is unique; this is not the case for infinite intersections. In Section 4, among other things, an extract from a letter by Piotr Koszmider is reproduced, showing that there is a poor filter which is not generated by a statistical measure. The argument relies on Boolean algebra techniques. The authors credit valuable comments also to István Juhász. This well-written paper will surely foster further the collaboration among analysts and set theorists.