The complementary beta distribution (Q1611781)

Let \((g_1,\dots, g_k)\) be a \(k\)-tuple of elements in a finite group \(G\) of order \(n\), let \(X_1\), \(X_2,\dots\) be i.i.d. \(G\)-valued random variables, and assume a random walk on \(G\) determined by the ``steps'' \(X_j\) is given in such a way that one multiplies the previous position of the walk by \(g_i^\gamma\), where \(i\) is uniform on \(\{1,\dots,k\}\) and \(\gamma\) is uniform on \(\{-1,0,1\}\), and \(i\) and \(\gamma\) are independent. Put \(P^{*m}(s)= \text{Pr} (X_1\dots X_m=s)\), \(s\in G\), and let \(d_P(m)\) denote the variation distance of \(P^{*m}\) from the uniform distribution, i.e. \(d_P(m)= {1\over 2}\sum_{s\in G} |P^{*m}(s)-1/ |G|\|\). The main new result is as follows. Suppose \(k=\log_2 n+f(n)\) with \(\lim_{n\to\infty} f(n)=\infty\), and let \(\varepsilon >0\). If \(m=m(n)> (1+\varepsilon) (9/8)(\log_2n) \ln(\log_2n)\), then \(\lim_{n\to \infty}E[d_P(m)] =0\), where the expectation is over a uniform choice of the \(n^k\) possible \(k\)-tuples \((g_1,\dots,g_k)\).