Nonequilibrium Steady State of a Weakly-Driven Kardar-Parisi-Zhang Equation

TL;DR

This paper investigates the non-equilibrium steady-state distribution of interface heights in a high-dimensional KPZ equation with weak noise, revealing Gaussian small fluctuations and asymmetric non-Gaussian tails with distinct scaling behaviors.

Contribution

It provides an analytical and numerical study of the steady-state height distribution in the weak-noise, high-dimensional KPZ equation, highlighting the nature of fluctuations and tail behaviors.

Findings

Steady state distribution approaches a non-equilibrium form in $d>2$.

Small fluctuations are Gaussian with variance from equilibrium free energy.

Tail behaviors are non-Gaussian; slow tail scales as |H|, fast tail as |H|^3.

Abstract

We consider an infinite interface in $d>2$ dimensions, governed by the Kardar-Parisi-Zhang (KPZ) equation with a weak Gaussian noise which is delta-correlated in time and has short-range spatial correlations. We study the probability distribution of the interface height $H$ at a point of the substrate, when the interface is initially flat. We show that, in a stark contrast with the KPZ equation in $d<2$, this distribution approaches a non-equilibrium steady state. The time of relaxation toward this state scales as the diffusion time over the correlation length of the noise. We study the steady-state distribution $\mathcal{P}(H)$ using the optimal-fluctuation method. The typical, small fluctuations of height are Gaussian. For these fluctuations the activation path of the system coincides with the time-reversed relaxation path, and the variance of $\mathcal{P}(H)$ can be found from a minimization of the (nonlocal) equilibrium free energy of the interface. In contrast, the tails of $\mathcal{P}(H)$ are nonequilibrium, non-Gaussian and strongly asymmetric. To determine them we calculate, analytically and numerically, the activation paths of the system, which are different from the time-reversed relaxation paths. We show that the slower-decaying tail of $\mathcal{P}(H)$ scales as $-\ln \mathcal{P}(H) \propto |H|$, while the faster-decaying tail scales as $-\ln \mathcal{P}(H) \propto |H|^3$. The slower-decaying tail has important implications for the statistics of directed polymers in random potential.