Discrete and continuous Muttalib--Borodin processes I: the hard edge (2010.15529v1)

Abstract: In this note we study a natural measure on plane partitions giving rise to a certain discrete-time Muttalib-Borodin process (MBP): each time-slice is a discrete version of a Muttalib-Borodin ensemble (MBE). The process is determinantal with explicit time-dependent correlation kernel. Moreover, in the $q \to 1$ limit, it converges to a continuous Jacobi-like MBP with Muttalib-Borodin marginals supported on the unit interval. This continuous process is also determinantal with explicit correlation kernel. We study its hard-edge scaling limit (around 0) to obtain a discrete-time-dependent generalization of the classical continuous Bessel kernel of random matrix theory (and, in fact, of the Meijer $G$-kernel as well). We lastly discuss two related applications: random sampling from such processes, and their interpretations as models of directed last passage percolation (LPP). In doing so, we introduce a corner growth model naturally associated to Jacobi processes, a version of which is the "usual" corner growth of Forrester-Rains in logarithmic coordinates. The aforementioned hard edge limits for our MBPs lead to interesting asymptotics for these LPP models. In particular, a special cases of our LPP asymptotics give rise (via the random matrix Bessel kernel and following Johansson's lead) to an extremal statistics distribution interpolating between the Tracy-Widom GUE and the Gumbel distributions.