On connected groups and related topics (Q1193041)

The paper deals with the structure of locally compact connected Lie groups, apparently an inexhaustible source of problems. I venture to say that there are two main tools for dealing with them: One, Lie group theory, two the character theory of compact (connected) abelian groups, and that's it. Almost. -- By a theorem of Iwasawa's and after the results of Montgomery, Gleason, and Yamabe we know that a locally compact connected group \(G\) is homeomorphic to \(E\times K\) where \(K\) is a maximal compact subgroup (which is connected) and a so-called manifold factor \(E\) concoted from \(n\) one-parameter subgroups of \(G\) such that \(E\) is diffeomorphic to \(\mathbb{R}^ n\). The structure of compact connected groups is fairly well understood. The commutator subgroup \(K'\) is closed, arcwise and locally arcwise connected and is very close to a direct product of connected compact simple Lie groups. Recall that \(K\) contains a compact connected abelian subgroup \(A\) such that \(K\) is the semidirect product \(K'\rtimes A\) of \(K'\) by \(A\cong K/K'\). Thus \(G\) is homeomorphic to \(E\times K'\times A\). It therefore follows that the topological properties of \(G\) are known if those of \(K'\) and \(A\cong K/K'\) are known. In particular, \(G\) is locally connected iff \(A\) is locally connected, it is arcwise connected iff \(A\) is arcwise connected. The highly developed duality theory of compact connected abelian groups applies to \(A\) and lets us determine \textit{precisely} when \(A\) is locally connected, or arcwise connected. There are compact connected and locally connected abelian groups which are not arcwise connected, and they are big. The issue of arc connectivity impinges on logic. Indeed \(A\) is arcwise connected iff \(\text{Ext}(\widehat A,\mathbb{Z})=0\). The statement that this condition implies that \(\widehat A\) is free, i.e., that \(A\) is a torus, i.e. a product of circles, is independent from Zermelo-Fraenkel-Bernays set theory and the Axiom of Choice. However, if \(\widehat A\) is countable, the statement is true, regardless. Thus if \(A=K/K'\) is metrizable, then \(A\) is arcwise connected iff it is a torus (iff it is locally connected). Thus all topological complications relating to connectivity of a locally compact connected group \(G\) are located in the corresponding complications of the group \(A\cong K/K'\) for which the answers are there. The arc component \(G_ a\) of the identity of \(G\) corresponds to \(E\times K'\times A_ a\) under the homeomorphism \(E\times K'\times A\). As Dixmier observed, \(A_ a\) is the set of points reached by one-parameter groups. With the aid of Iwasawa's local product decomposition theorem one shows that \(G_ a\) is generated by one parameter subgroups. In fact \(G_ a\) is the product of the injective homeomorphic image of a Lie group (normal in \(G\)) and \(K'\) and \(A_ a\). We note \(G=G_ aA\). There are many variations to that theme, and a number of them is discussed in the paper under review. Some of them rectify statements floating in the literature for some 30 years.