Stability, invariant random subgroups, and property testing

Given two permutations A and B which "almost" commute, are they "close" to permutations A' and B' which really commute? Arzhantseva and Paunescu (2015) formalized this question and answered affirmatively. This can be viewed as a property of the equation XY=YX, and turns out to be equivalent to the following property of the group Z^2 = < X,Y | XY=YX >: Every "almost action" of Z^2 on a finite set is close to a genuine action of Z^2. This leads to the notion of stable groups. Another point of view is that of property testing: The result of [AP] implies that for a pair of permutations, “being (close to) a solution for XY=YX” is a locally testable property, and one may ask which other equations, or systems of equations, are locally testable in this sense.

We will describe a relationship between stability, invariant random subgroups and sofic groups, giving, in particular, a characterization of stability among amenable groups. We will then show how to apply the above in concrete cases to prove and refute stability of some classes of groups. Finally, we will discuss stability of groups with Kazhdan's property (T), and some results on the quantitative aspect of stability.

Based on joint works with Alex Lubotzky, Andreas Thom and Jonathan Mosheiff.