Sojourn probabilities in tubes and pathwise irreversibility for It\^o processes

TL;DR

This paper derives a new expression for the probability that an Itô process remains near a path, introduces a pathwise irreversibility measure based on tube exit rates, and explores its implications for stochastic dynamics and entropy production.

Contribution

It provides a general formula for sojourn probabilities in tubes of finite radius for multidimensional Itô processes, and introduces a pathwise irreversibility measure independent of diffusivity.

Findings

Derived explicit Lagrangian for 1D systems in terms of drift and diffusivity.

Showed convergence of path-reversal probability ratios, defining irreversibility.

Demonstrated sensitivity of most probable crossing paths to tube radius.

Abstract

The sojourn probability of an It\^o diffusion process, i.e. its probability to remain in the tubular neighborhood of a smooth path, is a central quantity in the study of path probabilities. For $N$-dimensional It\^o processes with state-dependent full-rank diffusion tensor, we derive a general expression for the sojourn probability in tubes whose radii are small but finite, and fixed by the metric of the ambient Euclidean space. The central quantity in our study is the exit rate at which trajectories leave the tube for the first time. This has an interpretation as a Lagrangian and can be measured directly in experiment, unlike previously defined sojourn probabilities which depend on prior knowledge of the state-dependent diffusivity. We find that while in the limit of vanishing tube radius the ratio of sojourn probabilities for a pair of distinct paths is in general divergent, the same for a path and its time-reversal is always convergent and finite. This provides a pathwise definition of irreversibility for It\^o processes that is agnostic to the state-dependence of the diffusivity. For one-dimensional systems we derive an explicit expression for our Lagrangian in terms of the drift and diffusivity, and find that our result differs from previously reported multiplicative-noise Lagrangians. We confirm our result by comparing to numerical simulations, and relate our theory to the Stratonovich Lagrangian for multiplicative noise. For one-dimensional systems, we discuss under which conditions the vanishing-radius limiting ratio of sojourn probabilities for a pair of forward and backward paths recovers the established pathwise entropy production. Finally, we demonstrate for our one-dimensional example system that the most probable tube for a barrier crossing depends sensitively on the tube radius, and hence on the tolerated amount of fluctuations around the smooth reference path.