Rapid sampling of reactive Langevin trajectories via noise-space Monte Carlo

A noise-space Monte Carlo approach to sampling reactive Langevin trajectories is introduced and compared to a configuration based approach. The noise sampling is shown to overcome the slow relaxation of the configuration based method. Furthermore, the noise sampling is shown to sample multiple pathways with the correct probabilities without any additional work being required formally or algorithmically. The path sampling proceeds without any introduction of fictitious interactions and includes only the parameters appearing in Langevin's equation.