On the number of generators of finite images of free products of finite groups (Q5958875)

An open question known to specialists for a long time is settled in this paper; namely the question of the validity of the Grushko-Neumann Theorem in the category of profinite groups. The Grushko-Neumann Theorem states that the minimal number of generators \(d(G)\) of a free product \(G=G_1*G_2\) is equal to \(d(G_1)+d(G_2)\). In the profinite context one thinks in terms of topological generators. The author [J. Group Theory 4, No. 1, 53-58 (2001; Zbl 0973.20021)] proved that a profinite version of the Grushko-Neumann Theorem does not hold: namely he proved that if \(H_1,H_2\) are finite groups of coprime orders such that \(H_1\) is generated by \(r\) involutions and \(d(H_2)\leq r\) then the minimal number of generators \(d(G)\) of their free profinite product \(G=H_1\amalg H_2\) is less than or equal to \(2r\). It is proved in the paper that the same conclusion holds just assuming that \(H_1\) can be generated by \(r\) elements, not necessarily of order 2. In fact the author gives bounds in a more general situation. Theorem A. There exist two constants \(\alpha\) and \(c\), with \(\alpha<1\), such that the following holds. Suppose \(G\) is generated by \(s\) subgroups \(H_1,\dots,H_s\) of pairwise coprime orders and that each of these subgroups can be generated by \(r\) elements. Then \(\bullet\) \(d(G)\leq r(1+\alpha)+\tfrac{7c}3+1\) if \(s=2\); \(\bullet\) \(d(G)\leq(s-1)(\tfrac{4c}3+1)+2r+c\) if \(s>2\). In special cases the author obtains a stronger bound. If \(H_1,\dots,H_s\) are solvable then the bound is \(d(G)\leq s(\tfrac{4c}3+1)+r+c\). If they are \(p\)-groups, then \(d(G)\leq s+r\).