Groups of quasi-invariance and the Pontryagin duality (Q616919)

For a probability measure on a Polish group \(X\) and \(g\in X\) let \(\mu_g\) denote the measure defined by \(\mu_g(E)=\mu(g^{-1}\cdot E)\) for all Borel sets \(E\). The set \(E(\mu)=\{g\in X:\mu_g\sim\mu\}\) is a~subgroup of \(X\) and admits a Polish group topology. This article was inspired by an old problem in abstract harmonic analysis and topological algebra: find the ``right'' generalization of the class of locally compact groups. The author proposes this generalization. A Polish group \(G\) is called a QI-group if there is a locally compact Polish group \(X\) with a probability measure on \(X\) and a~continuous bijection from \(G\) onto \(E(\mu)\) such that \(E(\mu)\) is dense in \(X\) and \(\mu_g\perp\mu\) for all \(g\notin E(\mu)\). The author proves that if \(G\) is a QI-group represented in~\(X\) by \(\mu\), then its image \(E(\mu)\) in \(X\) is \(\sigma\)-compact. The author considers some general problems of the Pontryagin duality theory for QI-groups. He shows that there exists a Polish non-locally quasi-convex QI-group such that its bidual is not a QI-group and he proves that the bidual of a QI-group may be not saturated. He also defines a reflexive non-discrete group topology on the set of integers.