Borel hierarchies in infinite products of Polish spaces (Q2384694): Difference between revisions

Let \((X,\tau)\) be an uncountable Polish space. For \(n \in \mathbb N\) define \(\text{Pr}_{n}:X^{\mathbb N} \rightarrow X^n\) by setting, for \(\underline{x} = (x_{0}, x_{1}, \dots)\), \(\text{Pr}_{n}(\underline{x}) = (x_{0}, x_{1}, \dots, x_{n-1})\), and for \(A \subseteq X^n\) let \(\text{cyl}(A) = \{\underline{x} \in X^{\mathbb N}:\text{Pr}_{n}(\underline{x}) \in A\}\). In the Polish space \((X,\tau)^{\mathbb N}\) set \[ \overline{\Sigma}_0=\overline{\Pi}_0=\{\text{cyl}(A):A \subseteq X^n \text{ is Borel}, n \in \mathbb N\}, \] and for \(0 < \xi < \omega_{1}\) let \(\overline{\Sigma}_\xi=(\bigcup_{\eta <\xi}\overline{\Pi}_\eta)_\sigma\), \(\overline{\Pi}_\xi =\neg \overline{\Sigma}_\xi\). We define \({\mathcal B}=\bigcup_{\xi <\omega_1}\overline{\Sigma}_\xi=\bigcup_{\xi <\omega_1}\overline{\Pi}_\xi\). Note that the definition for the hierarchies \(\overline{\Sigma}_\xi\), \(\overline{\Pi}_\xi\) \((\xi < \omega_{1})\) in the paper under review differs from what we have given here. The original definition is incompatible with the results of the paper. Now endow \(X\) with the discrete topology \(\tau_{d}\) and consider the Borel hierarchy in \((X,\tau_{d})^{\mathbb N}\) as follows. Set \[ \Sigma_{0} = \Pi_{0} = \{\text{cyl}(A):A \subseteq X^n,~n \in \mathbb N\} \] and for \(0 < \xi < \omega_{1}\), \(\Sigma_{\xi} =(\bigcup_{\eta < \xi} \Pi_{\eta} )_{\sigma}\), \(\Pi_{\xi} = \neg \Sigma_{\xi}\). Assume \(X\) is recursively presented. Following the notation of [\textit{A.\ Louveau}, ``A separation theorem for \(\Sigma^ 1_ 1\) sets'', Trans. Am. Math. Soc. 260, 363--378 (1980; Zbl 0455.03021)], we define \(\Sigma^{\star}_{0} = \Pi^{\star}_{0} = \{\text{cyl}(A) :A \subseteq X^n\) is \(\Delta^{1}_{1}\), \(n\in \mathbb N\}\), and, for \(1 \leq \xi < \omega^{\text{CK}}_{1}\), \(\Sigma^{\star}_{\xi} = \bigcup^{1}_{1}(\bigcup_{\eta < \xi} \Pi^{\star}_{\eta})\), \(\Pi^{\star}_{\xi} = \neg \Sigma^{\star}_{\xi}\). Based on ideas of A.\ Louveau in the paper above and on [\textit{R.\ Barua}, ``On the Borel hierarchies of countable products of Polish spaces'', Real Anal. Exch. 16, No. 1, 60--66 (1991; Zbl 0770.54036)], the authors state that if every subset of \(\omega^{\omega}\) has the Baire property in the Gandy-Harrington topology then for every \(1 \leq \xi < \omega^{\text{CK}}_{1}\), if, for two \(\Sigma^{1}_{1}\) subsets \(A\), \(B\) of \((X,\tau)^{\mathbb N}\), \(A\) can be separated from \(B\) by a \(\Sigma_{\xi}\) set, then \(A\) can be separated from \(B\) by a \(\Sigma^{\star}_{\xi}\) set as well. This immediately implies \(\Sigma^{\star}_{\xi} = \Delta^{1}_{1} \cap \Sigma_{\xi}\) for every \(1 \leq \xi < \omega^{\text{CK}}_{1}\). Moreover, by relativization, if for every \(\alpha \in \omega^{\omega}\) every subset of \(\omega^{\omega}\) has the Baire property in the topology generated by the \(\Sigma^{1}_{1}(\alpha)\) sets then, for every \(1 \leq \xi < \omega_{1}\), \(\overline{\Sigma}_{\xi} = \Sigma_{\xi} \cap {\mathcal B}\). The set-theoretic assumptions fail in ZFC but hold in the Levy-Solovay model of ZF+DC, or in ZF+AD. Note that it remains open whether all these results hold in ZFC or not. E.g., the authors prove the separation theorem for \(\xi = 1\) in ZFC, without any additional assumptions.