Inverted orbits of exclusion processes, diffuse-extensive-amenability and (non-?)amenability of the interval exchanges

Abstract

The recent breakthrough works [6,8,9] which established the amenability for new classes of groups, lead to the following question: is the action $W(\mathbb{Z}^d) \curvearrowright \mathbb{Z}^d$ extensively amenable? (Where $W(\mathbb{Z}^d)$ is the {\em wobbling group} of permutations $\sigma:\mathbb{Z}^d \to \mathbb{Z}^d$ with bounded range). This is equivalent to asking whether the action $(\mathbb{Z}/2\ \mathbb{Z})^{(\mathbb{Z}^d)} \rtimes W(\mathbb{Z}^d) \curvearrowright (\mathbb{Z}/2\mathbb{Z})^{(\mathbb{Z}^d)}$ is amenable. The $d=1$ and $d=2$ and have been settled respectively in [6,8]. By [9], a positive answer to this question would imply the amenability of the IET group. In this work, we give a partial answer to this question by introducing a natural strengthening of the notion of extensive-amenability which we call diffuse-extensive-amenability. Our main result is that for any bounded degree graph $X$, the action $W(X)\curvearrowright X$ is diffuse-extensively amenable if and only if $X$ is recurrent. Our proof is based on the construction of suitable stochastic processes $(\tau_t)_{t\geq 0}$ on $W(X)\, <\, \mathfrak{S}(X)$ whose {\em inverted orbits} $$ \bar O_t(x_0) = \{x\in X, \exists s\leq t,\, \tau_s(x)=x_0\} = \bigcup_{0\leq s \leq t} \tau_s^{-1}(\{x_0\}) $$ are exponentially unlikely to be sub-linear when $X$ is transient. This result leads us to conjecture that the action $W(\mathbb{Z}^d) \curvearrowright \mathbb{Z}^d$ is not extensively amenable when $d\geq 3$ and that a different route towards the (non-?)amenability of the IET group may be needed.