Mixing of the averaging process and its discrete dual on finite-dimensional geometries (Q6103993)

The averaging process is a Markovian model of mass redistribution among nearest-neighboring sites of a graph and closely related to a large number of other models. In this paper a generalization of \(L^1\)-mixing of the averaging process and its discrete dual on finite dimensional geometries is given. The setting is that of large undirected graphs satisfying finite-dimensional Nash inequalities. A complete picture of the total variation mixing of a discrete dual of the averaging process, which here is called binomial splitting process, is obtained. A spectral gap identity for the binomial splitting process is obtained, showing that the \(k\)-particle system's spectral gap coincides with the spectral gap of the single-particle system on any graph. Multicolored averaging are introduced and a intertwining relation with binomial splitting process is proved. Sharp upper bounds for the averaging process from proprieties of a few-particle binomial splitting are derived and results on the many-particle binomial splitting are deduced.