The thresholds for diameter 2 in random Cayley graphs

From MaRDI portal



Abstract: Given a group G, the model mathcalG(G,p) denotes the probability space of all Cayley graphs of G where each element of the generating set is chosen independently at random with probability p. In this article we show that for any epsilon>0 and any family of groups G_k of order n_k for which nkoinfty, a graph GammakinmathcalG(Gk,p) with high probability has diameter at most 2 if pgeqslantsqrt(2+epsilon)fraclognknk and with high probability has diameter greater than 2 if pleqslantsqrt(1/4+epsilon)fraclognknk. We also provide examples of families of graphs which show that both of these results are best possible. Of particular interest is that for some families of groups, the corresponding random Cayley graphs achieve diameter 2 significantly faster than the ErdH{o}s-Renyi random graphs.


Given a group \(G\), the model \(\mathcal{G}(G,p)\) denotes the probability space of all Cayley graphs of \(G\) with each generator chosen independently with probability \(p\) from \(G\). The authors investigate the properties of graphs \(\Gamma\in \mathcal{G}(G,p)\) for an arbitrary group \(G\), \(|G|=n\rightarrow \infty\). In particular they prove that with high probability the diameter of \(\Gamma\) is at most 2 when \(p\geq \sqrt{(2+\epsilon)\frac{\log{n}}{n}}\) and this result is best possible as with high probability the diameter of \(\Gamma\) is greater than 2 when \(\Gamma\in \mathcal{G}(\mathbb{Z}_2^n,p)\) and \(p\leq \sqrt{(2-\epsilon)\frac{\log{2^n}}{2^n}}\).NEWLINENEWLINENEWLINE They also prove that with high probability the diameter of \(\Gamma\) is greater than 2 when \(p\leq \sqrt{(\frac{1}{4}-\epsilon)\frac{\log{n}}{n}}\) in the case of a general group \(G\) and when \(p\leq \sqrt{(\frac{1}{2}-\epsilon)\frac{\log{n}}{n}}\) in the case of an abelian group \(G\). Also, these results are best possible. Namely, if \(G\) is a cyclic group and \(p\geq \sqrt{(\frac{1}{2}+\epsilon)\frac{\log{n}}{n}}\), then with high probability the diameter of \(\Gamma\) is at most 2. Similarly, if \(0<\epsilon<1/4\), \(G\) has at most \(O(n^{(1+\epsilon)/2})\) involutions, at most \(O(n^{(1+\epsilon)/2})\) elements \(x\) with \(\mathrm{cl}(x)\leq 1/\epsilon\), at most \(O(n^{(1+\epsilon)/4})\) involutions \(x\) with \(\mathrm{cl}(x)\leq 1/\epsilon\) and at most \(\epsilon^2 n/49\) conjugacy classes, then with high probability the diameter of \(\Gamma\) is at most 2, when \(p\geq \sqrt{(\frac{1}{4}+\epsilon)\frac{\log{n}}{n}}\).NEWLINENEWLINENEWLINE The authors consider also the model \(\mathcal{G}(L,p)\), where \(L\) is a Latin square. In this model the set of vertices is \([n]\) and two vertices are adjacent if and only if \(L_{ij}\in S\) or \(L_{ji}\in S\), where the elements of \(S\) are chosen from \([n]\) independently with probability \(p\). The authors prove that in this case with high probability the diameter of \(\Gamma\in \mathcal{G}(L,p)\) is at most 2 when \(p\geq \sqrt{(26+\epsilon)\frac{\log{n}}{n}}\) and with high probability is greater than 2 when \(p\leq \sqrt{(\frac{1}{4}-\epsilon)\frac{\log{n}}{n}}\).











This page was built for publication: The thresholds for diameter 2 in random Cayley graphs

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2925523)