Automatic continuity of abstract homomorphisms between locally compact and Polish groups (Q2175518): Difference between revisions

The authors introduce the following class \(\mathcal{K}\) of \textit{almost Polish spaces}: \noindent Every \(X\in\mathcal{K}\) is a Hausdorff topological space, \(\mathcal{K}\) is closed under the passage to closed subspaces and closed under finite products, and the following properties are satisfied: (1) Every open covering of \(X\) has a countable subcovering. (2) The space \(X\) is not a countable union of nowhere dense subsets. (3) For each continuous image \(A\subseteq X\) of some \(\mathcal{K}\)-member there is an open set \(U\subseteq X\) such that the symmetric difference \((A\setminus U)\cap (U\setminus A)\) is a countable union of nowhere dense subsets of \(X\). The class \(\mathcal{P}\) of Polish spaces, the class \(\mathcal{L}^\sigma\) of locally compact \(\sigma\)-spaces and the class \(\mathcal{C}\) of compact spaces are almost Polish. If \(\mathcal{K}\) is an almost Polish class and \(X\in\mathcal{K}\), then \(A\subseteq X\) is called a \textit{\(\mathcal{K}\)-analytic set} if \(A=\psi(Z)\) holds for some \(Z\in\mathcal{K}\) and some continuous map \(\psi:Z\to X\). If \(G\) is a topological group with \(G\in\mathcal{K}\), then \(G\) is called \textit{\(\mathcal{K}\)-rigid} if the following holds: \noindent For every short exact sequence of groups \[1\to N\hookrightarrow K\overset{\varphi} \to G\to 1,\] where \(\varphi\) is an abstract group homomorphism, where \(G,K\in\mathcal{K}\), and where the kernel \(N\) of \(\varphi\) is \(\mathcal{K}\)-analytic, the homomorphism \(\varphi\) is automatically continuous and open. The authors show that rigidity fails in the following cases: \noindent (a) Abelian groups. (b) Groups which are not locally compact or \(\sigma\)-compact. (c) Infinite products of compact Lie groups if the kernel is not restricted. (d) \(\mathrm{SL}(n,\mathbb{C})\) and all infinite complex linear algebraic groups. (e) Certain connected perfect real algebraic groups. The first main result is Theorem 2.9. Let \(\mathcal{K}\) be a class of almost Polish spaces, and let \(K,G\in\mathcal{K}\) be topological groups. If \(\varphi:K\to G\) is an abstract group homomorphism such that for every identity neighborhood \(U\subseteq G\) there exists an identity neighborhood \(V\subseteq U\) such that \(\varphi^{-1}(V)\) is almost open (which is the case if \(\varphi^{-1}(V)\) is \(\mathcal{K}\)-analytic), then \(\varphi\) is continuous. If \(\varphi\) is in addition surjective, then it is open. \noindent In this situation, let \(G\) be a Lie group whose Lie algebra is perfect. If \((K,\ker\varphi)\in\mathcal{K}_a\) and if there exists a compact spacious subset \(C\subseteq G\) such that \(\varphi^{-1}(C)\) is \(\mathcal{K}\)-analytic, then \(\varphi\) is continuous and open. (Let \(\mathcal{K}_a\) denote the class of all pairs \((X,A)\), where \(X\in\mathcal{K}\) and \(A\subseteq X\) is \(\mathcal{K}\)-analytic. \(C\) is called \textit{spacious} if some product of finitely many translates of \(CC^{-1}\) has nonempty interior.) \noindent By setting \(C=G\) this result implies: Let \(G\) be a compact Lie group whose Lie algebra is semisimple. Then \(G\) is rigid within every almost Polish class \(\mathcal{K}\) that contains \(G\). \noindent This assertion contains results of \textit{R.~R. Kallman} [Adv. Math. 12, 416--417 (1974; Zbl. 0273.22009)] and \textit{P. Gartside} and \textit{B. Pejić} [Topology Appl. 155, 992--999 (2008; Zbl. 1151.54029)]. The second main result is Theorem 4.6: Let \(G\) be a Lie group such that the center of the connected component is finite and the Lie algebra of \(G\) is a direct sum of absolutely simple ideals. If \(\mathcal{K}\) is an almost Polish class containing \(G\), then \(G\) is rigid within \(\mathcal{K}\). Then the authors consider semiproducts of Lie groups. In Theorem 5.6 a list of such groups is given which are rigid in every Polish class \(\mathcal{K}\) containing them. In Section 6 the rigidity of topologically finitely generated profinite groups is studied. The main result is Theorem 6.3: Let \(G\) be a topologically finitely generated profinite group, and let \(\mathcal{K}\) be an almost Polish class. If \(G\) is contained in \(\mathcal{K}\), then \(G\) is rigid within \(\mathcal{K}\). \noindent This Theorem generalizes a result of Gartside and Pejić [loc. cit.]. Theorem 6.3 remains valid if \(G\) is a compact quasi-semisimple group (see Theorem 7.7). (A nontrivial compact group \(S\) with center \(C(S)\) is called \textit{quasi-semisimple} if the commutator subgroup is dense in \(S\) and if \(S/C(S)\) is topologically simple.) The very interesting paper ends with Theorem 8.1: Let \(G\) be a Lie group whose Lie algebra is perfect and let \(H\) be a topological group. Let \(\psi:G\to H\) be an abstract homomorphism such that there exists a compact spacious set \(C\subseteq G\) with compact \(\overline{\psi(C)}\), then \(\psi\) is continuous.