Variationally Enhanced Sampling

Abstract

Atomistic simulations can give a microscopic understanding of materials. However, their use is generally limited by the time scales that can be accessed as normally this falls short of what is needed to properly sample the complex free energy landscapes that characterize most material systems. Advanced methods are thus needed to enhance the sampling and overcome this time scale problem. Variationally enhanced sampling is one such approach based on identifying the important slow degrees of freedom and enhancing their fluctuations through the introduction of an external bias potential. The method is based on a variational principle that shows how the bias potential can be constructed by minimizing a convex functional, which brings a lot of flexibility. We introduce the theory and methodology of variationally enhanced sampling and discuss various novel and innovative applications and extensions.