Some consequences of Rado's selection lemma (Q1938400)

The paper under review is rich in very interesting results on certain consequences of Rado's selection lemma (\(\mathbf{RL}\)), which is the following combinatorial principle: Let \(\mathcal{F}\) be a family of finite sets and suppose that to every finite subset \(F\) of \(\mathcal{F}\) there corresponds a choice function \(\phi_F\) whose domain is \(F\) such that \(\phi_{F}(T)\in T\) for each \(T\in F\). Then there is a choice function \(f\) whose domain is \(\mathcal{F}\) with the property that for every finite subset \(F\) of \(\mathcal{F}\), there is a finite subset \(F'\) of \(\mathcal{F}\) such that \(F\subseteq F'\) and \(f(T)=\phi_{F'}(T)\) for all \(T\in F\), and which was introduced by \textit{R. Rado} [Can. J. Math. 1, 337--343 (1949; Zbl 0033.25302)]. \(\mathbf{RL}\) is known to be a consequence of the Boolean Prime Ideal Theorem (\(\mathbf{BPI}\)) (``Every (non-trivial) Boolean algebra has a prime ideal''), the latter being equivalent to the following variant of Tychonoff's compactness theorem: ``The Tychonoff product of compact Hausdorff topological spaces is compact''. (For the above implication, see [\textit{Y. Rav}, Math. Nachr. 79, 145--165 (1977; Zbl 0359.02066); \textit{W. H. Gottschalk}, Proc. Am. Math. Soc. 2, 172 (1951; Zbl 0043.05403)].) Moreover, \(\mathbf{RL}\) is equivalent to \(\mathbf{BPI}\) + \(\mathbf{AC^{fin}}\) (where \(\mathbf{AC^{fin}}\) is the Axiom of Choice (\(\mathbf{AC}\)) restricted to families of non-empty finite sets); see [Rav, loc. cit.]. On the other hand, \(\mathbf{RL}\) does not imply \(\mathbf{BPI}\) in \(\mathbf{ZFA}\) set theory (Zermelo-Fraenkel set theory with the Axiom of Extensionality weakened to allow the existence of atoms); see [\textit{P. E. Howard}, Z. Math. Logik Grundlagen Math. 30, 129--132 (1984; Zbl 0519.03026)]. It is an open problem whether \(\mathbf{RL}\) implies \(\mathbf{BPI}\) in \(\mathbf{ZF}\) set theory (Zermelo-Fraenkel set theory without \(\mathbf{AC}\)). This problem is also mentioned in the present paper. The paper under review basically consists of four parts: the relationship of \(\mathbf{RL}\) with: (1) the Hahn-Banach theorem, (2) the Axiom of Multiple Choice, (3) variants of Tychonoff's compactness theorem, and (4) the extendability of filters, of linearly orderable Boolean algebras, or on well-orderable sets, or on \(\omega\) in particular, to ultrafilters. The main results of the paper are as follows: {\parindent=5mm \begin{itemize}\item[1.)] \(\mathbf{RL}\) implies the Hahn-Banach theorem. In particular, it is shown that \(\mathbf{RL}\) implies the following equivalent form of the Hahn-Banach theorem: ``For every non-trivial Boolean algebra \((B,\oplus,\cdot)\), there exists a unitary measure \(m:B\rightarrow [0,1]\)'' (that is, \(m\) is finitely additive -- for all \(x,y\in B\) satisfying \(x\cdot y=0_B\), \(m(x\oplus y)=m(x)+m(y)\) -- and \(m(1_B)=1\)). \item[2.)] The Axiom of Multiple Choice (\(\mathbf{MC}\)) (``For every family \(\mathcal A\) of non-empty sets there exists a function \(f\) with domain \(\mathcal A\) such that, for all \(A\in\mathcal A\), \(f(A)\) is a non-empty finite subset of \(A\)'') implies, in \(\mathbf{ZFA}\), \(\mathbf{RL}\). This also yields that \(\mathbf{RL}\) does not imply \(\mathbf{BPI}\) in \(\mathbf{ZFA}\). (\(\mathbf{MC}\) is true in the second Fraenkel permutation model, whereas \(\mathbf{BPI}\) is false in that model). \item[3.)] \(\mathbf{RL}\) is equivalent to each one of the following statements: {\parindent=8mm \begin{itemize}\item[{\(\bullet\)}] Given a set \(I\) and a non-empty filter base \(\mathcal F\subseteq\mathcal{L}_{2^I}\setminus\{\emptyset\}\) (where \(\mathcal{L}_{2^I}\) is the lattice generated by closed subsets of the Tychonoff product \(2^I\) of the form \(\pi_i^{-1}(F)\), where \(i\in I\) and \(F\) is a subset of the discrete space \(2=\{0,1\}\)), if \(\mathcal F\) has a choice function, then \(\bigcap\mathcal F\) is non-empty. \item[{\(\bullet\)}] For every set \(I\) such that the family of all non-empty finite subsets of \(I\) has a choice function, the Tychonoff product \(2^I\) (where \(2=\{0,1\}\) has the discrete topology) is compact. \item[{\(\bullet\)}] Given a Boolean algebra \(B\) such that the family of all non-empty finite subsets of \(B\) has a choice function, every proper filter of \(B\) is included in an ultrafilter. \end{itemize}} \item[4.)] \(\mathbf{RL}\) implies ``Every proper filter of a linearly orderable Boolean algebra is included in an ultrafilter'', which in turn implies ``Every filter on a well orderable set is included in an ultrafilter'' (hence, every filter on \(\omega\) is included in an ultrafilter, which in turn implies ``there is a free ultrafilter on \(\omega\)''). \end{itemize}} The paper also contains several interesting open problems: {\parindent=5mm \begin{itemize}\item[{\(\bullet\)}] Is the (stronger than \(\mathbf{RL}\) and weaker than \(\mathbf{BPI}\)) statement ``Given a set \(I\) and a non-empty set \(\mathcal F\subseteq\mathcal L_{2^I}\setminus\{\emptyset\}\) satisfying the finite intersection property, if \(\mathcal F\) has a choice function, then \(\bigcap\mathcal F\) is non-empty'' equivalent to \(\mathbf{BPI}\) or to \(\mathbf{RL}\) in \(\mathbf{ZF}\)? \item[{\(\bullet\)}] Does \(\mathbf{RL}\) imply \(\mathbf{AC}^2\), i.e., \(\mathbf{AC}\) restricted to families of pairs, in \(\mathbf{ZF}\)? \item[{\(\bullet\)}] Does ``Every proper filter of a linearly orderable Boolean algebra is included in an ultrafilter'' or ``Every filter on a well orderable set is included in an ultrafilter'' imply \(\mathbf{RL}\) in \(\mathbf{ZF}\)? \item[{\(\bullet\)}] Let \(\mathbf{H}\) -- following the notation in Morillon's paper -- denote M. Hall's theorem: ``Let \(I\) be an infinite set of finite sets such that for each finite subset \(F\) of \(I\), there is an injective choice function on \(F\). Then there is an injective choice function on \(I\)''. Does (Hahn-Banach theorem) + \(\mathbf{H}\) imply \(\mathbf{BPI}\) in \(\mathbf{ZF}\)? Does (Hahn-Banach theorem) + \(\mathbf{H}\) + ``every filter on a well orderable set is included in an ultrafilter'' imply \(\mathbf{BPI}\) in \(\mathbf{ZF}\)? \end{itemize}} A diagram of implications/non-implications between the principles discussed in the paper and a list of 10 bibliographical items conclude the paper.