Two-step asymptotics of scaled Dunkl processes

Abstract

Dunkl processes are generalizations of Brownian motion obtained by using the differential-difference operators known as Dunkl operators as a replacement of spatial partial derivatives in the heat equation. Special cases of these processes include Dyson’s Brownian motion model and the Wishart-Laguerre eigenvalue processes, which are well-known in random matrix theory. It is known that the dynamics of Dunkl processes is obtained by transforming the heat kernel using Dunkl’s intertwining operator. It is also known that, under an appropriate scaling, their distribution function converges to a steady-state distribution which depends only on the coupling parameter β as the process time t tends to infinity. We study scaled Dunkl processes starting from an arbitrary initial distribution, and we derive expressions for the intertwining operator in order to calculate the asymptotics of the distribution function in two limiting situations. In the first one, β is fixed and t tends to infinity (approach to the steady state), and in the second one, t is fixed and β tends to infinity (strong-coupling limit). We obtain the deviations from the limiting distributions in both of the above situations, and we find that they are caused by the two different mechanisms which drive the process, namely, the drift and exchange mechanisms. We find that the deviation due to the drift mechanism decays as t−1, while the deviation due to the exchange mechanism decays as t−1/2.