Probabilistic nilpotence in infinite groups (Q2055282): Difference between revisions

While there is a large literature on the probability of commuting elements in topological group theory, or in abstract group theory, or in finite group theory, there is much less in connection with geometric group theory. The present contribution is placed exactly along this line of research, continuing some recent investigations in [\textit{Y. Antolín} et al., Proc. Am. Math. Soc. 145, No. 2, 479--485 (2017; Zbl 1380.20064)] and [\textit{M. C. H. Tointon}, J. Lond. Math. Soc., II. Ser. 101, No. 3, 1280--1297 (2020; Zbl 1504.20074)]. The best way to review the present contribution is probably with the same words of the authors in the abstract: ``The `degree of \(k\)-step nilpotence' of a finite group \(G\) is the proportion of the tuples \((x_1 , \ldots , x_{k+1}) \in G^{k+1}\) for which the simple commutator \([x_1 , \ldots , x_{k+1}]\) is equal to the identity. In this paper we study versions of this for an infinite group \(G\), with the degree of nilpotence defined by sampling \(G\) in various natural ways, such as with a random walk, or with a Følner sequence if \(G\) is amenable. In our first main result we show that if \(G\) is finitely generated, then the degree of \(k\)-step nilpotence is positive if and only if \(G\) is virtually \(k\)-step nilpotent (Theorem 1.5). This generalises both an earlier result of the second author treating the case \(k=1\) and a result of Shalev for finite groups, and uses techniques from both of these earlier results. We also show, using the notion of polynomial mappings of groups developed by Leibman and others, that to a large extent the degree of nilpotence does not depend on the method of sampling (Theorem 1.12). As part of our argument we generalise a result of Leibman by showing that if \(\varphi\) is a polynomial mapping into a torsion-free nilpotent group, then the set of roots of \(\varphi\) is sparse in a certain sense (Theorem 5.1). In our second main result we consider the case where \(G\) is residually finite but not necessarily finitely generated. Here we show that if the degree of \(k\)-step nilpotence of the finite quotients of \(G\) is uniformly bounded from below, then \(G\) is virtually \(k\)-step nilpotent (Theorem 1.19), answering a question of Shalev. As part of our proof we show that degree of nilpotence of finite groups is sub-multiplicative with respect to quotients (Theorem 1.21), generalising a result of Gallagher.''