The Keisler-Shelah isomorphism theorem and the continuum hypothesis. II (Q6097051): Difference between revisions

Let \(\mathfrak{M}\), \(\mathfrak{N}\) be \(\mathcal{L}\)-structures of size \(\leq 2^{\aleph_0}\), where \(\mathcal{L}\) is a countable language. In [Nederl. Akad. Wet., Proc., Ser. A 64, 477--495 (1961; Zbl 0118.01501)], \textit{H. J. Keisler} showed that the statement \(\mathfrak{N}\equiv\mathcal{M}\) if and only if \(\mathfrak{N}^\omega/U\cong\mathcal{M}^\omega/U\), for some ultrafilter \(U\) in \(\omega\), follows from the CH. In [Isr. J. Math. 10, 224--233 (1971; Zbl 0224.02045)], the second author proved that if \(\mathcal{L}\) is a countable language, \(\mathfrak{N}\equiv\mathcal{M}\) if and only if there exists an ultrafilter \(U\) in \(2^\omega\) such that \(\mathfrak{N}^\omega/U\cong\mathcal{M}^\omega/U\), without the hypothesis of CH. Afterwards, in [Math. Sci. Res. Inst. Publ. 26, 357--405 (1992; Zbl 0789.03035)] he provided a model of ZFC, where \(2^{\aleph_0}=\aleph_1\) and there are countable graphs \(\mathfrak{H}, \mathfrak{G}\) which are elementarily equivalent but for any ultrafilter \(U\) on \(\omega\), \(\mathfrak{H}^{\omega}/U\ncong\mathfrak{G}^\omega/U\). So, CH is an unavoidable assumption in order to obtain the Keisler result, even for countable models. The authors of the paper under review proved in [Fundam. Math. 260, No. 1, 59--66 (2023; Zbl 07687299)] that if for any two elementarily equivalent structures \(\mathfrak{N}\) \(\mathcal{M}\) of size \(\leq 2^{\aleph_0}\) in a countable language, \(\mathfrak{N}^\omega/U\cong\mathcal{M}^\omega/U\) for some ultrafilter \(U\) in \(\omega\), then CH holds. In this sequel under review, they consider some extensions of Keisler isomorphism theorem. In order to correctly state the results, several definitions are necessary. A pseudo ultraproduct problem is a sequence \[ m=\langle(\mathbb{M}_n^{m,1},\mathbb{M}_n^{m,2},\tau_{m,n}): n< \omega\rangle \] where \begin{itemize} \item[1.] \(\langle\tau_{m,n}: n<\omega\rangle\) is a \(\subseteq\)-increasing sequence of finite languages, with \(\tau_{m,0}=\emptyset\). Set \(\tau_m=\bigcup_n\tau_{m,n}\), \item[2.] every \(\mathbb{M}_n^{m,l}\) is a \(\tau_m\)-model, \item[3.] \(\kappa(m)\leq\aleph_1\), where \(\kappa(m)=\sup\{|\mathbb{M}_n^{m,l}|: l=1,2, n<\omega\}\) and \(|\mathbb{M}_n^{m,l}|\) is the size of the universe of the model \(\mathbb{M}_n^{m,l}\). \end{itemize} Then, they introduce an Ehrenfeucht-Fraïsse game \(\Game_k^n(m)=\Game_{\tau_{m,k},k}\) \((\mathbb{M}_n^{m,1}\), \(\mathbb{M}_n^{m,2})\). They define an ultraproduct problem through the game \(\Game\). The authors show the consistency of the failure of the CH with the assertion that for any non-principal ultrafilter \(D\) on \(\omega\), if the ultra\-pro\-ducts \(\prod_D\mathbb{M}_n^{m,1}\) and \(\prod_D\mathbb{M}_n^{m,2}\) are elementarily equivalent, then they are isomorphic. In the last section, it is proved that the above result can not be extended to a model of \(\mathrm{MA}+2^{\aleph_0}>\aleph_1\). In fact, they show that the above conclusion fails if \(\mathfrak{b}>\aleph_1\), where \(\mathfrak{b}\) denotes the bounding number. The paper (as well as the first part) is very well written and not difficult to follow.