A super-class walk on upper-triangular matrices (Q1882997)

Let \(G\) be the group of \(n\times n\), upper triangular matrices over the finite field \(\mathbb{F}_p\) with ones on the diagonal. The authors investigate the rate of convergence of certain conjugation-invariant random walks on \(G\) to the uniform distribution on \(G\) w.r.t. the total variation norm. It turns out that after \(k\) steps the error can be estimated by \(c_1 e^{-c_2k/(p^2 n\ln n)}\) for universal constants \(c_1,c_2> 0\). As such error estimates usually rely on character theory of \(G\) and as the character theory of \(G\) seems to be too complicated up to now, the authors restrict their attention to certain random walks with laws which are constant on particular unions of conjugacy classes, so-called super-classes, for which recently a complete character theory was developed by André, Carter, and Yan. This class of random walks on \(G\) is in particular closely related to those random walks, where in each step from a random row a random multiple of the next row is subtracted.