On the Fluctuation Relation for Nose-Hoover Boundary Thermostated Systems

TL;DR

This paper investigates the fluctuation relations in a boundary thermostated Lorentz gas with Nosé-Hoover thermostats, confirming the transient relation's dependence on initial conditions and analyzing the relaxation to steady state for phase space contraction and dissipation.

Contribution

It demonstrates how transient fluctuation relations depend on initial ensembles and verifies their asymptotic convergence to steady state relations in a Nosé-Hoover thermostated system.

Findings

Transient fluctuation relation holds for specific initial ensembles.

Steady state fluctuation relation is verified after mesoscopic times.

The dissipation function satisfies the fluctuation relation more accurately than phase space contraction rate.

Abstract

We discuss the transient and steady state fluctuation relation for a mechanical system in contact with two deterministic thermostats at different temperatures. The system is a modified Lorentz gas in which the fixed scatterers exchange energy with the gas of particles, and the thermostats are modelled by two Nos\'e-Hoover thermostats applied at the boundaries of the system. The transient fluctuation relation, which holds only for a precise choice of the initial ensemble, is verified at all times, as expected. Times longer than the mesoscopic scale, needed for local equilibrium to be settled, are required if a different initial ensemble is considered. This shows how the transient fluctuation relation asymptotically leads to the steady state relation when, as explicitly checked in our systems, the condition found in [D.J. Searles, {\em et al.}, J. Stat. Phys. 128, 1337 (2007)], for the validity of the steady state fluctuation relation, is verified. For the steady state fluctuations of the phase space contraction rate $\zL$ and of the dissipation function $\zW$, a similar relaxation regime at shorter averaging times is found. The quantity $\zW$ satisfies with good accuracy the fluctuation relation for times larger than the mesoscopic time scale; the quantity $\zL$ appears to begin a monotonic convergence after such times. This is consistent with the fact that $\zW$ and $\zL$ differ by a total time derivative, and that the tails of the probability distribution function of $\zL$ are Gaussian.