Classifying Erdős type spaces of higher descriptive complexity (Q375688)

Let \(\mu\) be an infinite cardinal and \(\ell_\mu^p=\{(x_\alpha)\in \mathbb{R}^\mu: \sum_{\alpha\in\mu}|x_\alpha|^p<\infty\}\). Given a fixed collection \((E_\alpha)_{\alpha\in\mu}\) of subsets in \(\mathbb{R}\), the Erdös type space \(\mathcal E_\mu\) is defined as \(\mathcal E_\mu=\{x\in \ell_\mu^p: x_\alpha\in E_\alpha, \alpha\in\mu\}\). For \(\mu=\omega\), \(p=2\) and \(E_n=\mathbb{Q}\), \(n\in\omega\), the Erdös space \(\mathfrak{C}=\{x\in\ell^2: x_n\in\mathbb{Q}, n\in\mathbb{N}\}\) is obtained. For \(T=\{0\}\cup\{1/m: m\in\mathbb{N}\}\), \(\mu=\omega\), \(p=2\) and \(E_n=T\) for each \(n\in\omega\), the complete Erdös space \(\mathfrak C_c\) is obtained. Both spaces were introduced in [\textit{P. Erdös}, Ann. Math. (2) 41, 734--736 (1940; JFM 66.0537.01)]. The paper presents a theory for Erdös type spaces constructed with zero-dimensional sets of higher Borel complexities. To formulate the results, we recall that \(w(X)\) denotes the weight of a topological space \(X\) and \(lw(X)\) is the local weight of \(X\), i.e., \(lw(X)=\min\{w(U): U\subset X\text{ nonempty open}\}\). For a set \(X\), \(X_D\) denotes this set equipped with the discrete topology. For \(\omega\leq \lambda\leq \kappa\), let \[ \mathfrak{C}_c(\lambda,\kappa)=\{x\in \ell_\kappa^1: x_\alpha\in T\text{ for }\alpha\in\lambda, x_{\alpha}\in\{0,1\}\text{ for }\alpha\in \kappa\setminus \lambda\} \] and \[ \mathfrak{C}(\lambda,\kappa)=\{x\in \ell_\kappa^1: x_\alpha\in \mathbb{Q}\text{ for }\alpha\in\lambda, x_{\alpha}\in\{0,1\}\text{ for }\alpha\in \kappa\setminus \lambda\}. \] Let \(\Pi_\gamma^0\) (respectively \(\Sigma_\gamma^0\)) for \(2\leq \gamma<\omega_1\) denote the class of all zero-dimensional separable metrizable spaces in the multiplicative (respectively additive) absolute Borel class of rank \(\gamma\). A space \(Z\) is said to absorb a class of topological spaces \(\mathcal{A}\) if \(Z\times X\approx Z\) for every \(X\in \mathcal{A}\setminus\{\emptyset\}\). If moreover \(Z\in\mathcal{A}\), then \(Z\) is called the stable space of \(\mathcal{A}\) (this definition is meaningful since stable spaces are topologically unique). By \textit{A. J. M. van Engelen} [Homogeneous zero-dimensional absolute Borel sets. CWI Tracts, 27. Centrum voor Wiskunde en Informatica. Amsterdam: Mathematisch Centrum. (1986; Zbl 0599.54044)], both \(\Pi_\gamma^0\) and \(\Sigma_\gamma^0\) contain stable spaces, which are denoted by \(P_\gamma\) respectively \(S_\gamma\). The spaces \(P_\gamma\) and \(S_\gamma\) are represented by subsets of \(\mathbb{R}\) that contain \(0\). For \(\omega\leq \lambda\leq \kappa\) and \(2\leq \gamma<\omega_1\), the Borel-Erdös space for \(\lambda,\kappa\) and \(\gamma\) are defined by \[ \mathfrak{C}_\gamma(\lambda,\kappa)=\{x\in\ell_\kappa^1: x_\alpha\in P_\gamma\text{ for }\alpha\in\lambda, x_\alpha\in\{0,1\}\text{ for }\alpha\in\kappa\setminus \lambda\}. \] Let \(\mathfrak{C}_\gamma=\mathfrak{C}_\gamma(\omega,\omega)\). It follows from the paper \textit{J. J. Dijkstra} et al. [J. Math. Soc. Japan 60, No. 3, 793--818 (2008; Zbl 1154.54021)], that \[ \mathfrak{C}_2(\lambda,\kappa)\approx P_2\times \mathfrak{C}_c(\lambda,\kappa)\approx P_2\times \mathfrak{C}_c\times (\lambda_D)^\omega\times \kappa_D. \] Further, the paper \textit{J. J. Dijkstra} and \textit{K. I. S. Valkenburg} [Topology Appl. 159, No. 7, 1875--1886 (2012; Zbl 1257.54037)], yields \[ \mathfrak{C}_3(\lambda,\kappa)\approx P_3\times \mathfrak{C}_c(\lambda,\kappa)\approx P_3\times\mathfrak{C}_c\times (\lambda_D)^\omega\times \kappa_D\quad\text{and}\quad \mathfrak{C}_3\approx\mathfrak{C}\approx P_3\times \mathfrak{C}_c. \] The following generalization of these results is proved in the paper. Let \(2\leq \gamma<\omega_1\) and \(\omega\leq \lambda \leq \kappa\), then \(\mathfrak{C}_\gamma(\lambda,\kappa)\approx P_\gamma\times \mathfrak{C}_c(\lambda,\kappa)\approx P_\gamma\times \mathfrak{C}_c\times (\lambda_D)^\omega\times \kappa_D\). In particular, \(\mathfrak{C}_\gamma\approx P_\gamma\times \mathfrak{C}_c\). For Erdös type spaces with all sets \(E_\alpha\) in the same multiplicative Borel class the authors prove the following theorem. Suppose \(\mathcal{E}_\mu\) is such that \(lw(\mathcal{E}_\mu)=\lambda\), \(w(\mathcal{E}_\mu)=\kappa\) and let \(2\leq \gamma<\omega_1\). Assume that there exists an \(\alpha\in\mu\) such that \(E_\alpha\approx P_\gamma\). Then \(\mathcal{E}_\mu\approx \mathfrak{C}_\gamma(\lambda,\kappa)\) if and only if every \(E_\alpha\in \Pi_\gamma^0\) and \(\text{ind}\mathcal{E}_\mu>0\). Further they show the following result. Suppose \(\mathcal{E}_\mu\) is such that \(lw(\mathcal{E}_\mu)=\lambda\), \(w(\mathcal{E}_\mu)=\kappa\) and let \(2\leq \gamma<\omega_1\). Assume that there exist infinitely many \(\alpha\in\mu\) such that \(E_\alpha\approx S_\gamma\). Then \(\mathcal{E}_\mu\approx \mathcal{C}_{\gamma+1}(\lambda,\kappa)\) if and only if every \(E_\alpha\in \Pi_{\gamma+1}^0\) and \(\text{ind}\mathcal{E}_\mu>0\). The last part of the paper studies the case where every \(E_\alpha\) is coanalytic and a link between these spaces and homeomorphism groups of manifolds that leave the zero-dimensional pseudoboundary invariant is established.