Derivation of the non-equilibrium generalized Langevin equation from a generic time-dependent Hamiltonian

Abstract: It has been become standard practice to describe steady-state non-equilibrium phenomena by Langevin equations with colored noise and time-dependent friction kernels that do not obey the fluctuation-dissipation theorem, but since these Langevin equations are typically not derived from first-principle Hamiltonian dynamics it is not clear whether they correspond to physically realizable scenarios. By exact Mori projection in phase space we derive the non-equilibrium generalized Langevin equation (GLE) from a generic many-body Hamiltonian with a time-dependent force h(t) acting on an arbitrary phase-space dependent observable $A$. The GLE is obtained in explicit form to all orders in $h(t)$. For non-equilibrium observables that correspond to a Gaussian process, the resultant GLE has the same form as the equilibrium Mori GLE, in particular the memory kernel is proportional to the total force autocorrelation function. This means that the extraction and simulation methods developed for equilibrium GLEs can be used also for non-equilibrium Gaussian variables. This is a non-trivial and very useful result, as many observables that characterize non-equilibrium systems display Gaussian statistics. For non-Gaussian non-equilibrium variables correction terms appear in the GLE and in the relation between the complementary force autocorrelations and the memory kernels, which are explicitly given in term of cubic correlation functions of $A$. Interpreting the time-dependent force h(t) as a stochastic process, we derive non-equilibrium corrections to the fluctuation-dissipation theorem and methods to extract all GLE parameters from experimental or simulation data, thus making our non-equilibrium GLE a practical tool to study and model general non-equilibrium systems.