Girsanov reweighting for simulations of underdamped Langevin dynamics. Theory

Abstract

The critical step in a molecular process is often a rare-event and has to be simulated by an enhanced sampling protocol. Recovering accurate dynamical estimates from such biased simulation is challenging. Girsanov reweighting is a method to reweight dynamic properties formulated as path expected values. The path probability is calculated at the time-step resolution of the molecular-dynamics integrator. But the theory is largely limited to overdamped Langevin dynamics. For underdamped Langevin dynamics, the absolute continuity of the path probability ratio for the biased and unbiased potential is not guaranteed, but it depends on the Langevin integrator. We develop a general approach to derive the path probability ratio for Langevin integrators and to analyze whether absolute continuity is fulfilled. We demonstrate our approach on symmetric splitting methods for underdamped Langevin dynamics. For methods that obey absolute continuity, and thus can be used for Girsanov reweighting, we provide an expression for the relative path probability. %