Invariance of compactness for the Bohr topology (Q5929008)

\textit{E. van Douwen}'s paper [Topology Appl. 34, 69-91 (1990; Zbl 0696.22003)] and, subsequently, \textit{Trigos-Arrieta}'s work on LCA groups [Ph. D. dissertation, Wesleyan University, Middletown CT, USA, 1991] relaunched the interest on topological properties of weakly topologized groups, particularly on the preservation of topological properties after passing from the Bohr (weak) topology to the strong topology. Recently, several authors (\textit{W. Banaszczyk} and \textit{E. Martín-Peinador} [J. Pure Appl. Algebra 138, 99-106 (1999; Zbl 0935.22004)], \textit{S. Hernández} and the reviewer [Fundam. Math. 159, 195-218 (1999; Zbl 0934.22008)], \textit{K. Kunen} and \textit{W. Rudin} [Math. Proc. Camb. Philos. Soc. 126, 117-137 (1999; Zbl 0915.43003)]) have chosen the existence of conveniently placed discrete subsets of a special nature as a way to study the Bohr topology of maximally almost periodic Abelian (MAPA, for short) groups. Together with discrete subsets in the Bohr topology which are either \(C\)-embedded in the group or \(C^\ast\)-embedded in the Bohr compactification, the present paper deals with subsets equivalent to the \(\ell^1\)-basis, a concept based on Bourgain's extension of the usual one in the Banach space setting (a subset of a Banach space is equivalent to the \(\ell^1\)-basis if it generates a subspace isomorphic to \(\ell^1\)). The duality theory of MAPA groups often takes the well-understood duality theory of locally convex vector spaces as a source of inspiration (this is natural, since every locally convex space is a MAPA group in its additive structure) but some difficulties usually arise in this approach (the class of MAPA groups is simply too wide) and some restriction to the theory has to be made. That kind of situation appears in the present paper when the authors need an analog of Grothendieck's characterization of the completion. The result turns out to fail for general MAPA groups and the authors are forced to find a good class of groups for which this fundamental result holds. The resulting groups are named \(g\)-groups and comprise LCA groups and additive groups of locally convex spaces. The main theorem of the paper states that the weak topology of a complete \(g\)-group must be a \(\mu\)-space (functionally bounded subsets must be relatively compact), giving thus the counterpart to an analogous result of \textit{M. Valdivia} for Banach spaces [J. Funct. Anal. 24, 1-10 (1977; Zbl 0344.46004)]. An immediate consequence is that, for complete \(g\)-groups, the preservation of several topological properties (compactness, countable compactness, pseudocompactness) is, in essence, the same as the preservation of functional boundedness and that this is equivalent to the existence of infinite, discrete and \(C\)-embedded subsets (with respect to the Bohr topology) in every non-totally bounded subset of a complete \(g\)-group. The last equivalence quoted above leads the authors to considering sequences equivalent to the \(\ell^1\)-basis and to prove some Rosenthal-type theorems which yield the existence of sequences equivalent to the \(\ell^1\)-basis contained in sequences without any Cauchy subsequence. With all these ingredients, the paper comes to its last theorem: in a metrizable and complete \(g\)-group \(G\) the Bohr and original topologies have the same compact subsets if and only if every subset \(A\) of \(G\) whose closure is not compact contains an infinite subset \(B\) which is (a) discrete, (b) \(C\)-embedded in \(G\) (with respect to the Bohr topology), (c) \(C^{\ast}\)-embedded in the Bohr compactification of \(G\) and (d) equivalent to the \(\ell^1\)-basis. As to the proof, it is first shown that the \(C\)-embedding property is equivalent to the preservation of the compactness property (this follows directly from previous results in the paper -- the Bohr topology is a \(\mu\)-space) while the other properties of \(B\) follow from the Rosenthal-type theorems quoted above.