Cross sections and homeomorphism classes of Abelian groups equipped with the Bohr topology (Q5943949)

Let \(G^{\#}\) denote an Abelian group \(G\) equipped with its largest totally bounded group topology or, in other words, with the Bohr topology of the discrete group \(G\). This paper provides a substantially new approach to the general problem of studying how do algebraic properties of \(G\) affect the topological ones of \(G^{\#}\). The following two questions due to \textit{E. van Douwen} [Topology Appl. 34, 69-91 (1990; Zbl 0696.22003)] that are addressed in this paper constitute the foundations of this problem: (1) If two groups \(G_1\) and \(G_2\) have the same cardinality, must the spaces \(G_1^{\#}\) and \(G_2^{\#}\) be homeomorphic? and, (2) is every subgroup \(H\) of \(G^{\#}\) a retract of \(G^{\#}\)? It should be noted that question (2) remains unsolved and that question (1) has been solved in the negative by \textit{K. Kunen} [Topology Appl. 90, 91-107 (1998; Zbl 0974.54025)] and \textit{D. Dikranjan} and \textit{St. Watson} [J. Pure Appl. Algebra 163, 147-158 (2001; Zbl 0983.22001)] but still the problem of determining which countable groups have homeomorphic \(\#\)-topologies is far from being well understood. The paper presents a common approach to questions (1) and (2) based on the existence of continuous cross sections. If \(H\) is a subgroup of a topological group \(G\), a continuous cross section from \(G/H\) is a continuous mapping \(\Gamma:G/H \rightarrow G\) with the property \(\pi \circ \Gamma = id|_{G/H}\), where \(\pi:G\rightarrow G/H\) denotes the natural homomorphism. While cross sections always exist (it is just a matter of \textit{choosing} a representative of each coset \(g+H\)), continuous cross sections can be impossible to find and a subgroup \(H\) is called a ccs-subgroup when such a continuous cross section can be defined. One of the main results of the paper proves that \({\mathbb Z}^{\#}\) is a ccs-subgroup of \({\mathbb Q}^{\#}\). This has the following interesting consequences: \({\mathbb Z}^{\#}\) is a retract of \({\mathbb Q}^{\#}\) and \({\mathbb Q}^{\#}\) is homeomorphic to \(({\mathbb Q}/{\mathbb Z})^{\#} \times {\mathbb Z}^{\#}\). Within the general approach of the paper it is possible to extend these results from \({\mathbb Z}\) to finitely generated groups. The other main result goes in the opposite direction in presenting an example of a group \(G\) having a subgroup \(H\) that is not a ccs-subgroup for the \(\#\)-topology. The example is obtained by embedding an \(\omega\)-dimensional \({\mathbb Z}_q\)-module, \(M_q\) (i.e. a countable direct sum \(\oplus_{\omega} {\mathbb Z}_q\) of the cyclic group of order \(q\)) in a \({\mathbb Z}_{q^2}\)-module, \(M_{q^2}\). The construction is of a set-theoretic nature and is inspired by the counterexample of Kunen [op. cit.] for the homeomorphism problem. Even if the paper does not completely solve the main questions (1) and (2), it does present significant advances and sheds new light on both of them.