Steady state entropy production rate for scalar Langevin field theories

TL;DR

This paper develops a method to compute the entropy production rate in scalar Langevin field theories, including active models, providing a way to quantify irreversibility at the field level.

Contribution

It introduces a general framework for calculating the field-level EPR in scalar Langevin theories, including a small-noise expansion and analysis of different observable choices.

Findings

Derived a non-negative quadratic expression for EPR in scalar Langevin fields.

Applied the method to Model AB and obtained explicit EPR calculations.

Showed that EPR depends on the choice of coarse-grained observables.

Abstract

The entropy production rate (EPR) offers a quantitative measure of time reversal symmetry breaking in non-equilibrium systems. It can be defined either at particle level or at the level of coarse-grained fields such as density; the EPR for the latter quantifies the extent to which these coarse-grained fields behave irreversibly. In this work, we first develop a general method to compute the EPR of scalar Langevin field theories with additive noise. This large class of theories includes active versions of Model A (non-conserved density dynamics) and Model B (conserved) and also models where both types of dynamics are simultaneously present (such as Model AB). Treating the scalar field $\phi$ (and its time derivative $\dot\phi$) as the sole observable(s), we arrive at an expression for the EPR that is non-negative for every field configuration and is quadratic in the time-antisymmetric component of the dynamics. Our general expression is a function of the quasipotential, which determines the full probability distribution for configurations, and is not generally calculable. To alleviate this difficulty, we present a small-noise expansion of the EPR, which only requires knowledge of the deterministic (mean-field) solution for the scalar field in steady state, which generally is calculable, at least numerically. We demonstrate this calculation for the case of Model AB. We then present a similar EPR calculation for Model AB with the conservative and non-conservative contributions to $\dot\phi = \dot\phi_{\rm A} + \dot\phi_{\rm B}$ viewed as separately observable quantities. The results are qualitatively different, confirming that the field-level EPR depends on the choice of coarse-grained information retained within the dynamical description.