Special subsets of \(^{\text{cf}(\mu)}\mu\), Boolean algebras and Maharam measure algebras (Q1961228): Difference between revisions

Let \(\overline I=\langle I_i:i<\delta\rangle\) be a sequence of ideals on \(\text{Dom}(I_i)\), let \(\overline\eta= \langle\eta_\alpha:\alpha<\lambda\rangle\) be a sequence of members of \(\prod_{i<\delta}\text{Dom}(I_i)\), and let \(I\), \(J\) be ideals on \(\lambda\) and \(\delta\), respectively. It is said that \(\overline\eta\) is a \((\lambda,I,J)\)-sequence for \(\overline I\) if \(\{i<\delta: \{\eta_\alpha(i):\alpha\in X\}\in I_i\}\in J\) for every \(X\subseteq\lambda\), \(X\notin I\). If \(I_i\) is \(\kappa_i\)-complete, \(\kappa_i>\sum_{j<i}\kappa_j\), \(\mu=\sum_{i<\delta}\kappa_i\) is strong limit, \(|\text{Dom}(I_i)|<\mu\) and \(2^\mu=\mu^+=\lambda\), then there is such a sequence for ideals \(I\), \(J\) of bounded subsets of \(\lambda\) and \(\delta\), respectively. The main point of this paper is to build a \((\lambda,I,J)\)-sequence without using \(2^{\mu^+}=\mu^+\). The author handles these cases: 1. \(I_i=J^{\text{bd}}_{\lambda_i}\) (the ideal of bounded subsets of \(\lambda_i\)), \(\lambda=\text{cf}(\prod_{i<\delta}\lambda_i/J)\); 2. \(I_i=\prod_{l<n_i} J^{\text{bd}}_{\lambda_{l,i}}\) where \(\lambda_{l,i}\) are regular increasing with \(l\) and \(i\), and \(J\) is an ideal on \(\{(i,l):i<\delta\), \(l<n_i\}\); 3. \(I_i=\prod_{l<n_i} J^{\text{bd}}_{\lambda_{l,i}}\) where \(\lambda_{l,i}\) are regular decreasing with \(l\); and 4. two modifications of case 3 with nonstationary ideals or club guessing ideals instead of \(J^{\text{bd}}_{\lambda_{l,i}}\). The ideals \(I\) and \(J\) are often chosen to be \(J^{\text{bd}}_{\lambda}\) and \(J^{\text{bd}}_{\delta}\), respectively. Some results are obtained directly using pcf assumptions while, on the other hand, it is shown also that these assumptions are very reasonable. The existence of \((\lambda,I,J)\)-sequences is one of the main tools throughout the paper. The second theme of the paper is the question when a Boolean algebra \(\mathbf B\) has a free caliber \(\lambda\) (i.e., when for every set \(X\subseteq\mathbf B\) with \(|X|=\lambda\), there is \(Y\subseteq X\) with \(|Y|=\lambda\) such that \(Y\) is independent). Some results in this direction are obtained for Maharam measure algebras, ``small'' products of free Boolean algebras and \(\kappa\)-c.c.\ Boolean algebras. Further, it is proved that in Maharam measure algebra \(\mathbf B(\mu)\) (with \(\lambda=\mu^+=2^\mu\) where \(\mu\) is strong limit singular with countable cofinality) there is a strongly \(\lambda\)-anti-independent set of intervals (i.e., a system of intervals \(\{(a_\alpha,b_\alpha):\alpha<\alpha^*\}\) such that if \(\mathbf B'\supseteq\mathbf B(\mu)\), \(X\in[\alpha^*]^\lambda\) and \(a_\alpha\leq x_\alpha\leq b_\alpha\) in \(\mathbf B'\) for \(\alpha\in X\) then there are \(n<\omega\), \(\alpha_0<\alpha_1<\dots<\alpha_{n-1}\) from \(X\) such that some nontrivial Boolean combination of \(\langle x_{\alpha_l}:l<n\rangle\) is zero). It follows that \(\mathbf B(\mu)\) has a subalgebra of cardinality and dimension \(\lambda\) with no independent set of cardinality \(\lambda\) and there is a Hausdorff compact zero-dimensional space with measure on the family of Borel sets and having dimension \(\lambda\) so that as a measure space it is isomorphic to the Maharam space \(\mathcal B(\lambda)\) but there is no homeomorphism from \(X\) onto \(^\lambda 2\). It was a question of Monk whether \((\beth_\omega)^+\) is a free caliber of the product of the free Boolean algebras \(\text{FBA}(\beth_n)\) with \(\beth_n\) generators for \(n<\omega\). It is shown using some \((\lambda,I,J)\)-sequences that this is consistently not true. On the other hand \(\prod_{i<\omega_1}\text{FBA}(\beth_i)\) has free caliber \((\beth_{\omega_1})^+\). The proof involves some pcf considerations. Solving another problem of D. Monk it is proved that if \(\kappa\) is a weakly inaccessible cardinal such that \(\langle 2^\mu:\mu<\kappa\rangle\) is not eventually constant then there is a Boolean algebra of cardinality \(2^{<\kappa}\) with no independent subsets of cardinality \(\kappa^+\). Deducing from \textit{M. Gitik} and \textit{S. Shelah} [``On densities of box products'', Topology Appl. 88, No. 3, 219-237 (1998; Zbl 0926.03060)] some complementary consistency results are obtained (e.g., the first mentioned question of Monk is not answerable in ZFC) and also several other problems of \textit{J. D. Monk} [Cardinal invariants of Boolean algebras (Progress in Mathematics, Vol. 142, Birkhäuser, Basel) (1996; 849.03038)] are answered. Another question of Monk says: Assume \(\text{cf}(\mu)\leq\kappa<\mu< \lambda\leq\mu^{\text{cf}(\mu)}\). Is there a Boolean algebra of cardinality \(\lambda\) satisfying \(\kappa\)-c.c.\ with no independent set of cardinality \(\lambda\)? The consistency of the failure of this condition is proved although it is proved that ``almost always'' there is such a Boolean algebra. Finally, for several invariants \(\text{inv}\) (\(\text{inv}=\text{irr}\), \(s\), \(\text{hL}\), \(\text{hd}\), \(\text{Length}\)) of Boolean algebras consistency of the existence of Boolean algebras \(B_n\) and a nonprincipal ultrafilter \(\mathcal D\) such that \(\text{inv}(\prod_{n<\omega}B_n/\mathcal D)> (\prod_{n<\omega}\text{inv}(B_n))/\mathcal D\) is proved.