Multi-point distribution for Gaussian non-equilibrium non-Markovian observables

Abstract

When analyzing experimental or simulation time-series data, the question arises whether it is possible to tell from a one-dimensional time-dependent trajectory whether the system is in equilibrium or not. We here consider the non-equilibrium version of the generalized Langevin equation for a Gaussian observable and show that i) the multi-point joint distribution solely depends on the two-point correlation function and that ii) the two-point correlation function for a non-equilibrium process is identical to an equilibrium process with uniquely determined parameters. Since the multi-point joint distribution completely characterizes the dynamics of an observable, this means that the non-equilibrium character of a system, in contrast to its non-Markovianity, cannot be read off from a one-dimensional trajectory.