On the number of generators needed for free profinite products of finite groups. (Q926399)

Let \(G_1,G_2,\dots,G_n\) be finite groups and let \(\Gamma=G_1*G_2\cdots*G_n\) be the free product of the \(G_i\). It is an interesting and difficult problem to compare the minimal number of generators \(d(\Gamma)\) of \(\Gamma\) with the minimal number of topological generators \(d(\widehat\Gamma)\) of the profinite completion \(\widehat\Gamma\) of \(\Gamma\). The so called Grushko-Neumann theorem says that \(d(\Gamma)=\sum_id(G_i)\). Trivially \(d(\widehat\Gamma)\leq d(\Gamma)\). It was proved by the reviewer [J. Algebra 245, No. 2, 552-561 (2001; Zbl 1001.20021)] that \(d(\widehat\Gamma)\) can be smaller than \(d(\Gamma)\) but it remains to understand how large the difference \(d(\Gamma)-d(\widehat\Gamma)\) can be. It was conjectured that if the \(G_i\) have pairwise coprime orders then in fact \(d(\widehat\Gamma)=n+s-1\), with \(s=\max_id(G_i)\). In this paper some interesting lower bounds for \(d(\widehat\Gamma)\) are obtained. For example it is proved that \(d(\widehat\Gamma)\geq n\) (hence the above conjecture is true when \(s=1\)) and that \(d(\widehat\Gamma)\geq n+s'-1\), with \(s'=\max_id(G_i/G_i')\). In particular, if all the \(G_i\) are nilpotent then \(d(\widehat\Gamma)\geq n+s-1\) which sets the conjectured lower bound. Moreover, if the \(G_i\) are \(p_i\)-groups for distinct primes \(p_i\), then this result combined with an upper bound proved by the reviewer implies \(n+s-1\leq d(\widehat\Gamma)\leq n+s\). For groups of pairwise coprime order where the minimal number of generators is not witnessed by the Abelianization, it seems a hard problem to set the conjectured lower bound in general, but in the case \(s=2\) the authors show that this is the best possible one can hope for. Indeed they prove that for any positive integer \(n\) there exist solvable groups \(G_1,G_2,\dots,G_n\) of pairwise coprime order such that \(d(G_i)=2\), \(G_i/G_i'\) is cyclic and \(d(\widehat\Gamma)\geq n+1\).