skip to content

Department of Applied Mathematics and Theoretical Physics

A random subset X of Z^d is ASL_d(Z)-invariant if its law is invariant under the affine special linear group ASL_d(Z) = SL_d(Z) x Z^d. Natural examples fall into two essentially different families: Bernoulli random subsets (each point is kept independently with a fixed probability), which are chaotic, and random translates of nZ^d, which are highly structured. Motivated by questions from mathematical physics, we are interested in classifying all of them.

In this talk, I will explain that the structured family is richer than it first appears but that the classification shows a surprising rigidity: for d \geq 3, every invariant random subset is essentially built from random polynomials and Bernoulli thinnings.

I will focus on a key (yet elementary) step that frames our approach, namely how unipotent invariance relates the marginal laws of X to non-conventional ergodic averages and, through them, to Gowers norms. This brings together three a priori unrelated areas: the theory of point processes, the dynamics of simple Lie groups and their lattices, and higher-order Fourier analysis. The talk will, however, be devoted entirely to this elementary reframing, and no knowledge of ergodic theory or Lie groups will be assumed.

Based on joint works with Mikolaj Fraczyk and with Asgar Jamneshan.

Further information

Time:

18Nov
Nov 18th 2026
13:30 to 14:30

Venue:

MR4, CMS

Series:

Discrete Analysis Seminar