📝 SummarXiv

Abstract

The macroscopic fluctuation theory is a powerful tool to characterise the large scale dynamical properties of diffusive systems, both in- and out-of-equilibrium. It relies on an action formalism in which, at large scales, the dynamics is fully determined by the minimum of the action. Within this formalism, the analysis of the statistical properties of a given observable reduces to solving the Euler-Lagrange equations with the appropriate boundary conditions. One must then compute the action at its minimum to deduce the cumulant generating function of the observable. This typically involves computing multiple integrals of cumbersome expressions. Recently, a simple formula has been conjectured to shortcut this last step, and compute the cumulant generating function of different observables (integrated current or position of a tracer) without the need to compute any integral. In this work, we prove this simple formula, and extend it to more general observables. We then illustrate the efficiency of this approach by applying it to compute the variance of a generalised current in the semi-infinite symmetric exclusion process and the joint properties of two occupation times in any diffusive system. In the case of the integrated current, our formula can be interpreted as a generalisation of Fick's law to obtain all the cumulants of the current beyond the average value.