On the probability of generating prosoluble groups. (Q2472728)

Every compact topological group \(G\) admits a normalised Haar measure so that for every \(k\in\mathbb{N}\) the product space \(G^k\) carries a natural probability measure. Thus one can consider the probability \(P(G,k)\) that a random \(k\)-tuple generates \(G\) and, more specifically, the invariant \[ d_{\text{P}}(G):=\inf\{k\in\mathbb{N}\mid P(G,k)>0\}. \] \textit{A.~Mann} [Forum Math. 8, No. 4, 429-459 (1996; Zbl 0852.20019)] proved that every finitely generated prosoluble group \(G\) is positively finitely generated, i.e. satisfies \(d_{\text{P}}(G)<\infty\). The aim of the article under review is to describe \(d_{\text{P}}(F)\) for the free prosoluble group \(F\) of finite rank \(d\). The starting point is an exact formula for \(P(G,k)\), valid for any finite soluble group \(G\), in terms of invariants related to the \(G\)-module structure of the chief factors of \(G\). This formula is readily obtained from results of \textit{W.~Gaschütz} [Ill. J. Math. 3, 469-476 (1959; Zbl 0093.25002)]. The main results, Theorems~1 and 2, provide lower and upper bounds for \(d_{\text{P}}(F)\) in terms of constants \(c_2=3.056\dots\) and \(c_3=3.243\dots\); the number \(c_3-1\) is known as the Pálfy-Wolf constant. The numbers \(c_2\) and \(c_3\) constitute the beginning of a sequence of constants \(c_p\), indexed by the primes \(p\), which appeared in the work of \textit{P. P.~Pálfy} [J. Algebra 77, 127-137 (1982; Zbl 0489.20004)]. One has \(2<c_p<c_2\) for \(p\geq 5\). The constants \(c_p\) come into play, because Pálfy's results imply, for instance, that for every prime \(p\) and every \(n\in\mathbb{N}\) the orders of the maximal irreducible soluble subgroups of \(\text{GL}(n,p)\) are bounded by \(24^{-1/3}p^{(c_p-1)n}\). Theorem~1 states that \(d_{\text{P}}(F)\geq c_3(d-1)+1\), provided that \(d\geq 3\). The proof is based on Proposition~2.1 (wrongly labelled~1.1) which guarantees the existence of a family \(G_i\), \(i\geq 3\), of \(2\)-generated finite soluble groups with the following properties: (i) \(G_i\) admits an absolutely irreducible module \(\overline V\) of size \(n_i:=9^{4^i}\), (ii) \(G_i\) has no complemented chief factors of order \(n_i\), (iii) \(\lim_{i\to\infty} P(G_i,k)>0\) for every \(k\geq 3\), (iv) \(|G_i|\) is roughly of size \(n_i^{c_3-1}\), (v) \(|\Aut(G_i)|\) is roughly of size \(n_i^{c_3-1}\). The groups \(G_i\) are manufactured as subgroups of \(\text{GL}(2,3)\text{\,wr\,}\text{Sym}(4)\text{\,wr\,}\cdots\text{\,wr\,}\text{Sym}(4)\), with the factor \(\text{Sym}(4)\) appearing \(i\) times. Roughly half of the article is devoted to this construction, which is appropriately described as ``technical''. As the authors explain, their Theorem~1 provides a correction of Theorem~24(a) in [\textit{A.~Mann}, loc. cit.]; contrary to what Mann states, the groups \(K_n\) considered in his proof require more and more generators as \(n\) tends to infinity. Theorem~2 asserts that \(d_{\text{P}}(F)\leq\lceil\max\{c_3(d-1)+1,c_2d\}\rceil\). In order to prove this upper bound, Pálfy's work on primitive soluble groups is used to bound, for finite soluble groups \(G\), the number of irreducible \(G\)-modules of order \(p^n\). Here the case \(p=3\) requires a separate, more detailed treatment. For \(d\geq 10\) one has \(\lceil c_3(d-1)+1\rceil\geq\lceil c_2d\rceil\) and the estimates in Theorem~1 and 2 combine nicely to give Theorem~3: \[ d_{\text{P}}(F)=\lceil c_3(d-1)+1\rceil\qquad\text{for \(d\geq 10\)}. \] According to \textit{A.~Mann and A.~Shalev} [Isr. J. Math. 96, Pt. B, 449-468 (1996; Zbl 0877.20017)], a profinite group is positively finitely generated if and only if it has polynomial maximal subgroup growth. In the paper under review, the degree of maximal subgroup growth of such a group \(G\) is defined as \[ s(G):=\inf\{s\mid\exists C\in\mathbb{R}\,\forall n\in\mathbb{N}: m_n(G)\leq Cn^s\} \] where \(m_n(G)\) denotes the number of maximal subgroups of index precisely \(n\) in \(G\). Theorem~4 states that \[ c_3 (d-1)+1\leq s(F)\leq\max\{c_3(d-1)+1,c_2d\}\qquad\text{for }d\geq 3. \] The proof, which is outlined in the final section of the article, combines ideas from the main part of the paper with arguments from the work of \textit{A.~Mann} [loc. cit.]. Indeed, in order to establish these bounds one needs to control, for every \(n\in\mathbb{N}\), the number of homomorphisms \(\varphi\colon F\to\text{Sym}(n)\) whose image is a primitive soluble group. The lower bound can be proved by a construction of sufficiently many such homomorphisms, based on Proposition~2.1 described above. The upper bound follows from the proof of Theorem~2 in light of the work of Mann (cf.\ proofs of Theorem~23(a) and Proposition~19 in his article). In subsequent work, \textit{M.~Morigi} [Isr. J. Math. 155, 117-123 (2006; Zbl 1139.20025)] has extended the results of the article under review to cover also free prosoluble groups of ranks less than \(10\).