The probability of generating a finite linear group. (Q1764451): Difference between revisions

For any finite group \(G\) and any nonnegative integer \(t\), let \(P_G(t)\) denote the probability that a random \(t\)-tuple of elements from \(G\) generates \(G\). Let \({\mathcal X}_n\) denote the class of all finite groups which have a completely reducible linear representation of degree \(n\) over some field. Then \textit{L. G. Kovács} and \textit{G. R. Robinson} [Proc. Am. Math. Soc. 112, No. 2, 357-364 (1991; Zbl 0725.20031)] have shown that \(P_G(\lfloor\tfrac 32n\rfloor)>0\) for all \(G\in{\mathcal X}_n\) and this result was refined for suitable fields by \textit{E. Ditomi, A. Lucchini} and \textit{F. Morini} [Isr. J. Math. 132, 29-44 (2002; Zbl 1042.20046)]. Define \(\overline G:=G/O^2(G)\) as the largest \(2\)-power factor group of \(G\). In the present paper the authors show that for any sequence of groups \(G_n\in{\mathcal X}_n\) we have \(P_{\overline G_n}(\lfloor\beta n\rfloor)-P_{G_n}(\lfloor\beta n\rfloor)\to 0\) as \(n\to\infty\) whenever \(\beta\geq\tfrac 32\). As a consequence, under the same hypothesis, \(P_G(\lfloor\beta n\rfloor)\to 1\) as \(n\to\infty\) whenever \(\beta>\tfrac 32\) (the behaviour is different at \(\beta=\tfrac 32\)).