Separating forward and backward pathways in nonequilibrium umbrella sampling.

Umbrella sampling enforces uniform sampling of steady-state distributions that are functions of arbitrary numbers of order parameters. The key to applying such methods to nonequilibrium processes is the accumulation of fluxes between regions. A significant difference between microscopically reversible and irreversible systems is that, in the latter case, the transition path ensemble for a reaction can be significantly different for "forward" and "backward" trajectories. Here, we show how to separately treat forward and backward pathways in nonequilibrium umbrella sampling simulations by working in an extended space. In this extended space, the exact rate (for equilibrium or nonequilibrium processes) can be calculated "for free" as a flux in phase space. We compare the efficiency of this rate calculation with forward flux sampling for a two-dimensional potential and show that nonequilibrium umbrella sampling is more efficient when an intermediate is present. We show that this technique can also be used to describe steady-state limit cycles by examining a simulation of circadian oscillations. We obtain the path of the limit cycle in a space of 22 order parameters, as well as the oscillation period. The relation of our method to others is discussed.