Elliptic Bessel processes and elliptic Dyson models realized as temporally inhomogeneous processes

TL;DR

This paper introduces elliptic Bessel and Dyson processes on a circle, extending classical models with elliptic functions, and demonstrates their construction via Girsanov transformations, with special cases exhibiting determinantal structures.

Contribution

The paper constructs elliptic Bessel and Dyson processes using elliptic functions and Girsanov transformations, extending classical models to inhomogeneous, finite-time processes on a circle.

Findings

Elliptic processes are realized as Girsanov transforms of Brownian motions.

Special cases exhibit determinantal martingale representations.

Elliptic Dyson model is determinantal for all observables without multiple points.

Abstract

The Bessel process with parameter $D>1$ and the Dyson model of interacting Brownian motions with coupling constant $\beta >0$ are extended to the processes in which the drift term and the interaction terms are given by the logarithmic derivatives of Jacobi's theta functions. They are called the elliptic Bessel process, eBES$^{(D)}$, and the elliptic Dyson model, eDYS$^{(\beta)}$, respectively. Both are realized on the circumference of a circle $[0, 2 \pi r)$ with radius $r >0$ as temporally inhomogeneous processes defined in a finite time interval $[0, t_*), t_* < \infty$. Transformations of them to Schr\"odinger-type equations with time-dependent potentials lead us to proving that eBES$^{(D)}$ and eDYS$^{(\beta)}$ can be constructed as the time-dependent Girsanov transformations of Brownian motions. In the special cases where $D=3$ and $\beta=2$, observables of the processes are defined and the processes are represented for them using the Brownian paths winding round a circle and pinned at time $t_*$. We show that eDYS$^{(2)}$ has the determinantal martingale representation for any observable. Then it is proved that eDYS$^{(2)}$ is determinantal for all observables for any finite initial configuration without multiple points. Determinantal processes are stochastic integrable systems in the sense that all spatio-temporal correlation functions are given by determinants controlled by a single continuous function called the spatio-temporal correlation kernel.