Direct evaluation of rare events in active matter from variational path sampling

Active matter is driven far-from-equilibrium due to autonomous self-propulsion at the scale of individual particles. Sampling rare trajectories and computing their rates in active systems is complicated by the breaking of detailed balance. Using tools from trajectory reweighting and reinforcement learning, we have developed a novel numerical algorithm called Variational Path Sampling(VPS) that uses optimal control forces to directly generate uncorrelated reactive trajectories and compute exact rate estimates in arbitrary out-of-equilibrium systems. VPS optimizes a control force to render rare fluctuations as typical, by using a variational principle arising from a ratio of path partition functions. We have applied VPS to sample rare conformational fluctuations and compute the rate of isomerization of a passive solute dimer in a dense active bath. We find significant rate enhancement with increasing self-propulsion of the active bath, resulting from driven collisions of motile bath particles with the cross-section of the passive particles.