Maximal almost disjoint families and pseudocompactness of hyperspaces (Q2052563)

The main object of study of this paper is the Mrówka space \(\Psi(\mathcal A)\) of a MAD family \(\mathcal A\), and the main question addressed is whether the Vietoris hyperspace \(\mathrm{exp}(\Psi(\mathcal A))=\{F\subseteq\Psi(\mathcal A)\mid \varnothing\neq F\text{ is closed}\}\) of such a space is pseudocompact (that is, every continuous function to \(\mathbb R\) is bounded). In case it is, one simply says that \(\mathcal A\) is pseudocompact, for short. Tackling a particular case (the one where the only spaces under consideration are those of the form \(\Psi(\mathcal A)\)) of a question of \textit{J. Ginsburg} [Can. J. Math. 27, 1392--1399 (1975; Zbl 0314.54007)], the authors of the paper under review determine various sets of conditions under which a given MAD family \(\mathcal A\) is pseudocompact. There are three main results, one per section (excluding the Introduction). The first one is a set of equivalences, namely that the statements \begin{itemize} \item[1.] \(\mathsf{MA}_{\mathfrak c}(\wp(\omega)/\mathrm{Fin})\), \item[2.] every MAD family is pseudocompact, \item[3.] \(\mathfrak h=\mathfrak c\) and every base tree has a cofinal branch, \item[4.] \(\mathfrak n(\omega^*)>\mathfrak c\), \end{itemize} are all equivalent. (Everything in those statements is basically self-explanatory, except perhaps for the notation \(\mathfrak n(X)\), used to denote the \textit{Baire number} of the topological space \(X\) -- the least number of open dense subsets of \(X\) that have empty intersection). In the same section, the authors furthermore prove that \(\mathfrak p<\mathfrak c\) is consistent with either of the previous statements (by forcing with a Suslin tree over a ground model (containing one such tree and) satisfying \(\mathfrak p=\mathfrak c=\omega_2\)). After that, in the next section, the question about generic existence of pseudocompact MAD families is studied. By using a matrix iteration of Mathias forcings, a model with \(\mathfrak a=\omega_1\) plus an ultrafilter that is \(\mathfrak c=\omega_2\)-closed (in the sense of pseudointersections, i.e. in particular this ultrafilter is a P-point) is constructed; earlier results in those sections show that these conditions on the model imply the generic existence of pseudocompact MAD families (that is, every AD family of size \(<\mathfrak c\) can be extended to a pseudocompact MAD family). Of course, generic existence is implied by every MAD family being pseudocompact; this model, however, shows that the implication is not reversible (because in particular \(\mathfrak h\leq\mathfrak a\) and so one of the conditions, from the previously mentioned result, equivalent to all MAD families being pseudocompact, does not hold). Finally, in Section 4, the authors prove that it is also consistent that the generic existence of pseudocompact MAD families fails; they do this by building a model with a MAD family that has cardinality \(\omega_2<\mathfrak c=\omega_3\) and is not pseudocompact (in particular, that very family provides an AD family of size \(<\mathfrak c\) that cannot be extended to a pseudocompact MAD family). This is done by means of a short iteration (a finite support iteration of length \(\omega_2\) over a ground model where \(\mathfrak c=\omega_3\)) on which \(\omega_1\) MAD families of size \(\omega_2\) are carefully added, in such a way that a MAD family of functions (also added step by step by the iteration) interacting nicely with the aforementioned MAD families turns out not to be pseudocompact.