Time correlation functions of equilibrium and nonequilibrium Langevin dynamics: Derivations and numerics using random numbers

TL;DR

This paper derives and numerically verifies time correlation functions for equilibrium and nonequilibrium Langevin dynamics, including inertia effects, providing a comprehensive resource for understanding and simulating such stochastic systems.

Contribution

It presents the first full derivation of nonequilibrium Langevin correlation functions and compares multiple numerical methods, enhancing understanding and teaching of stochastic dynamics.

Findings

Analytical solutions for Langevin correlation functions in equilibrium and nonequilibrium.

Numerical verification of derived correlation functions.

Demonstration of how inertia effects diminish in the zero-mass limit.

Abstract

We study the time correlation functions of coupled linear Langevin dynamics without and with inertia effects, both analytically and numerically. The model equation represents the physical behavior of a harmonic oscillator in two or three dimensions in the presence of friction, noise, and an external field with both rotational and deformational components. This simple model plays pivotal roles in understanding more complicated processes. The presented analytical solution serves as a test of numerical integration schemes, its derivation is presented in a fashion that allows to be repeated directly in a classroom. While the results in the absence of fields (equilibrium) or confinement (free particle) are omnipresent in the literature, we write down, apparently for the first time, the full nonequilibrium results that may correspond, e.g., to a Hookean dumbbell embedded in a macroscopically homogeneous shear or mixed flow field. We demonstrate how the inertia results reduce to their noninertia counterparts in the nontrivial limit of vanishing mass. While the results are derived using basic integrations over Dirac delta distributions, we mention its relationship with alternative approaches involving (i) Fourier transforms, that seems advantageous only if the measured quantities also reside in Fourier space, and (ii) a Fokker--Planck equation and the moments of the probability distribution. The results, verified by numerical experiments, provide additional means of measuring the performance of numerical methods for such systems. It should be emphasized that this manuscript provides specific details regarding the derivations of the time correlation functions as well as the implementations of various numerical methods, so that it can serve as a standalone piece as part of education in the framework of stochastic differential equations and calculus.