Exact first-passage time distributions for three random diffusivity models

TL;DR

This paper derives exact first-passage time distributions for a stochastic process with variable diffusivity modeled by three different stochastic processes, revealing surprising and anomalous diffusion behaviors.

Contribution

It provides the first exact formulas for FPT PDFs in three distinct stochastic diffusivity models, including a surprising Levy-Smirnov form for one model.

Findings

FPT PDF for Model II matches Levy-Smirnov distribution.

Models I and III show non-standard tail behaviors.

All models have broad PDFs with no finite first moment.

Abstract

We study the extremal properties of a stochastic process $x_t$ defined by a Langevin equation $\dot{x}_t=\sqrt{2 D_0 V(B_t)}\,\xi_t$, where $\xi_t$ is a Gaussian white noise with zero mean, $D_0$ is a constant scale factor, and $V(B_t)$ is a stochastic "diffusivity" (noise strength), which itself is a functional of independent Brownian motion $B_t$. We derive exact, compact expressions for the probability density functions (PDFs) of the first passage time (FPT) $t$ from a fixed location $x_0$ to the origin for three different realisations of the stochastic diffusivity: a cut-off case $V(B_t) =\Theta(B_t)$ (Model I), where $\Theta(x)$ is the Heaviside theta function; a Geometric Brownian Motion $V(B_t)=\exp(B_t)$ (Model II); and a case with $V(B_t)=B_t^2$ (Model III). We realise that, rather surprisingly, the FPT PDF has exactly the L\'evy-Smirnov form (specific for standard Brownian motion) for Model II, which concurrently exhibits a strongly anomalous diffusion. For Models I and III either the left or right tails (or both) have a different functional dependence on time as compared to the L\'evy-Smirnov density. In all cases, the PDFs are broad such that already the first moment does not exist. Similar results are obtained in three dimensions for the FPT PDF to an absorbing spherical target.