Invariable generation of infinite groups. (Q468705)

A group \(G\) is said to be invariably generated by a set \(S\) if \(G=\langle s^{g(s)}\mid s\in S\rangle\) for each choice of \(g(s)\in G\), \(s\in S\). The notion of invariable generation occurs naturally for Galois groups, where elements are only given up to conjugacy [see \textit{J. D. Dixon}, Discrete Math. 105, No. 1-3, 25-39 (1992; Zbl 0756.60010)]. A group \(G\) is called IG if it is invariably generated by some subset \(S\) of \(G\) and it is called FIG if it is invariably generated by some finite subset. In a previous paper the authors proved that a finite group \(G\) is invariably generated by at most \(\log_2|G|\) elements [J. Algebra 348, No. 1, 302-314 (2011; Zbl 1248.20036)]. In this paper they consider infinite groups. Which of them are FIG? The main result solves this problem for linear groups: a linear group is FIG if and only if it is finitely generated and virtually solvable. The linearity assumption cannot be dropped: for example the Grigorchuk group is FIG but not virtually solvable. The authors notice that many linear groups are not IG (an example is given by \(\text{SL}_n(\mathbb C)\)). In particular they prove that a linear algebraic group over an algebraically closed field is IG if and only if it is virtually solvable. In the second part of the paper it is investigated the analogous question for profinite groups. They are always IG, but they are not necessarily FIG (even if they are finitely generated). It is intriguing that, while arithmetic groups are not FIG, their profinite completions are often FIG. The authors prove for example that the profinite completion of an arithmetic group having the congruence subgroup property is invariably generated by a finite set of elements.