On refinements of \(\omega \)-bounded group topologies (Q714739)

According to \textit{N. T. Varopoulos} [Proc. Camb. Philos. Soc. 60, 465--516 (1964; Zbl 0161.11103)], two topological abelian groups \(G\) and \(G^{\prime }\) are said to be in duality (denoted by \(\langle G, G'\rangle\)) if and only if there is a function \(\langle .\, ,.\rangle :G\times G^{\prime }\longrightarrow \mathbb T\) (\(\mathbb{T}\) denotes the circle group) such that \(\langle g_{1}g_{2},g^{\prime }\rangle =\langle g_{1},g^{\prime }\rangle \cdot \langle g_{2},g^{\prime }\rangle\); \(\langle g,g_{1}^{\prime }g_{2}^{\prime }\rangle =\langle g,g_{1}^{\prime }\rangle \cdot \langle g,g_{2}^{\prime }\rangle\) for all \(g, g_{1}, g_{2}\in G\), \(g^{\prime }, g_{1}^{\prime }, g_{2}^{\prime }\in G^{\prime }\); and it holds that (i) if \(g\neq 0_{G}\), the neutral element of \(G\), then there exists \(g^{\prime }\in G^{\prime }\) such that \(\langle g,g^{\prime }\rangle \neq 1\); and (ii) if \(g^{\prime }\neq 0_{G^{\prime }}\) , the neutral element of \(G^{\prime }\), there exists \(g\in G\) such that \( \langle g,g^{\prime }\rangle \neq~1\). Given a duality \(\langle G,G^{\prime }\rangle \), we can associate the following two canonical topologies on \(G\), respectively, \(G^{\prime }\), the topology \(\sigma(G,G^{\prime })\) on \(G\) which is the weak topology generated by all the elements in \(G^{\prime }\) considered as continuous homomorphisms into \(\mathbb T\) and the topology \(\sigma(G^{\prime },G)\) defined similarly. Both topologies are precompact and compatible with the given duality, in the sense that the Pontryagin dual of \((G,\sigma(G,G'))\) coincides with \(G'\). Therefore, their completions are compact groups denoted by \(b(G,G^{\prime })\) and \(b(G^{\prime },G)\), respectively [Varopoulos, loc. cit.]. The following question has been studied recently by several authors: for a duality \(\langle G,G^{\prime }\rangle\), consider pairs of properties \((P,Q)\) which are in duality in the sense that \(G\) satisfies \(P\) if and only if \(G'\) satisfies \(Q\). In this paper, the author sheds fresh light on these matters looking at the properties of \(\omega\)-bounded group and \(P\)-group. A topological group \(G\) is \(\omega\)-bounded if the closure of any countable subset in \(G\) is compact; and a topological group \(G\) is a \(P\)-group if every \(G_\delta\)-set is open. Each topological group \((G,\tau)\) has associated a \(P\)-group (its \(P\)-modification \((G,P\tau)\)), whose topology \(P\tau\) has the family of \(G_\delta\)-sets of \(G\) as an open base. Among others, the following results are proved. (1) (Theorem 2.4) Given a duality \(\langle G,G^{\prime }\rangle\), then \((G',\sigma(G',G))\) is \(\omega\)-bounded if and only if the topology \(P\sigma(G,G')\) is compatible with the dual pair. (2) (Theorem 3.5) The following conditions are equivalent for a precompact abelian group \(G\): (a) \(G\) is realcompact; (b) every \(\sigma(\widehat G, G)\)-countably continuous homomorphism \(f\in b(G,G')\) is \(\sigma(\widehat G, G)\)-continuous; (c) \(PG\) is complete. (3) (Theorem 4.1) For an \(\omega\)-bounded group \(G\), the following assertions are equivalent: (a) there exists an \(\omega\)-bounded group topology on \(G\) strictly finer than the original; (b) \((\widehat G,\sigma(\widehat G,G))\) is not realcompact; (c) \((\widehat G,P\sigma(\widehat G,G))\) is not complete. The latter result is related to a recent one by \textit{W. W. Comfort} and \textit{J. van Mill} [Proc. Am. Math. Soc. 135, No. 12, 4039--4044 (2007; Zbl 1138.22002)], where it was proved that, for every pseudocompact abelian group of uncountable weight \((G,\tau)\), there exists a pseudocompact group topology strictly finer than \(\tau\).