On the Random Generation of Finite Simple Classical Groups (Q3158385)

Let \(r\leq s\) be primes with \(s>2\). The main result of this paper is that there exists a quadratic function \(f\) such that if \(G\) is a classical simple group over a finite field with the dimension of the natural module at least \(n\geq f(s)\), then the probability that \(G\) is generated by a random pair of elements of orders \(r\) and \(s\) tends to \(1\) as \(|G|\) tends to \(\infty\). A similar result was proved by \textit{M. W. Liebeck} and \textit{A. Shalev} [Bull. Lond. Math. Soc. 34, No. 2, 185-188 (2002; Zbl 1046.20046)]. The nature of the function was not described in that paper. The proof proceeds by estimating the number of elements of orders \(r\) and \(s\) and getting a similar estimate of all maximal subgroups. This is a special case of a conjecture of the reviewer: Let \(r\leq s\) be primes and assume that \((r,s)\neq (2,2)\) or \((2,3)\). Let \(X(r,s)\) be the family of simple groups whose orders are divisible by \(r\) and \(s\). Then the probability that a random pair of elements of orders \(r\) and \(s\) generate \(G\) tends to \(1\) as \(G\in X(r,s)\) has order tending to \(\infty\). If \(r=s=2\), then the group generated is dihedral. If \((r,s)=(2,3)\), there are counterexamples (but in fact by results of Liebeck and Shalev and the reviewer, Lübeck and Shalev, only \(4\)-dimensional symplectic groups fail). The remaining cases are Chevalley groups in bounded dimension where techniques of algebraic geometry may be used.