Coarse-graining in time with the Functional Renormalisation Group: Relaxation in Brownian Motion

TL;DR

This paper applies the functional Renormalisation Group to analyze relaxation in a stochastic Langevin system, demonstrating how coarse-graining in time improves understanding of dynamics and equilibrium properties across different potentials and temperatures.

Contribution

It introduces a novel application of the fRG to time-coarse-grain stochastic processes, connecting supersymmetric quantum mechanics with relaxation dynamics.

Findings

Effective potential captures fluctuations at all timescales.

Gradient expansion accuracy improves at higher temperatures.

Method handles complex potentials with multiple wells.

Abstract

We apply the functional Renormalisation Group (fRG) to study relaxation in a stochastic process governed by an overdamped Langevin equation with one degree of freedom, exploiting the connection with supersymmetric quantum mechanics in imaginary time. After reviewing the functional integral formulation of the system and its underlying symmetries, including the resulting Ward-Takahashi identities for arbitrary initial conditions, we compute the effective action $\Gamma$ from the fRG, approximated in terms of the leading and subleading terms in the gradient expansion: the Local Potential Approximation and Wavefunction Renormalisation respectively. This is achieved by coarse-graining the thermal fluctuations in time resulting in e.g. an effective potential incorporating fluctuations at all timescales. We then use the resulting effective equations of motion to describe the decay of the covariance, and the relaxation of the average position and variance towards their equilibrium values at different temperatures. We use as examples a simple polynomial potential, an unequal Lennard-Jones type potential and a more complex potential with multiple trapping wells and barriers. We find that these are all handled well, with the accuracy of the approximations improving as the relaxation's spectral representation shifts to lower eigenvalues, in line with expectations about the validity of the gradient expansion. The spectral representation's range also correlates with temperature, leading to the conclusion that the gradient expansion works better for higher temperatures than lower ones. This work demonstrates the ability of the fRG to expedite the computation of statistical objects in otherwise long-timescale simulations, acting as a first step to more complicated systems.