Cutoff for Almost All Random Walks on Abelian Groups
Abstract
Consider the random Cayley graph of a finite group with respect to generators chosen uniformly at random, with ; denote it . A conjecture of Aldous and Diaconis (1985) asserts, for , that the random walk on this graph exhibits cutoff. Further, the cutoff time should be a function only of and , to sub-leading order. This was verified for all Abelian groups in the '90s. We extend the conjecture to . We establish cutoff for all Abelian groups under the condition , where is the minimal size of a generating subset of , which is almost optimal. The cutoff time is described (abstractly) in terms of the entropy of random walk on . This abstract definition allows us to deduce that the cutoff time can be written as a function only of and when and ; this is not the case when . For certain regimes of , we find the limit profile of the convergence to equilibrium. Wilson (1997) conjectured that gives rise to the slowest mixing time for amongst all groups of size at most . We give a partial answer, verifying the conjecture for nilpotent groups. This is obtained via a comparison result of independent interest between the mixing times of nilpotent and a corresponding Abelian group , namely the direct sum of the Abelian quotients in the lower central series of . We use this to refine a celebrated result of Alon and Roichman (1994): we show for nilpotent that is an expander provided . As another consequence, we establish cutoff for nilpotent groups with relatively small commutators, including high-dimensional special groups, such as Heisenberg groups.
Cite
@article{arxiv.2102.02809,
title = {Cutoff for Almost All Random Walks on Abelian Groups},
author = {Jonathan Hermon and Sam Olesker-Taylor},
journal= {arXiv preprint arXiv:2102.02809},
year = {2025}
}
Comments
Accepted at Journal of European Mathematical Society (JEMS), Sept '25. This is part of a multi-paper project investigating properties of Cayley graphs with divergently many generators chosen uniformly at random. There is some textual overlap between the introductions of the different papers