Traveling without dwelling: Extending the timescale accessible to molecular dynamics simulation

Abstract

Molecular dynamics (MD) simulations have been widely recognized as a promising tool for investigating dynamical processes with an atomic resolution. The time span accessible to MD simulations is, however, extremely limited, which makes it difficult to simulate dynamical events on a timescale of milliseconds or longer while maintaining the microscopic-level resolution. Here, we present an approach that allows us to accelerate the dynamical processes (infrequent events) by introducing an additional bias potential and to describe these events on the correct timescale. To this end, we utilize a reparametrization of time (generalized Poincaré time transformation), which leads to the accelerated dynamics and the biased sampling in phase space. The direct link between the timescale and biased sampling demonstrated in this study makes it possible to extend the timescale in MD simulations without invoking the approximation of the transition state theory. A general form of the bias potential is also introduced, which is easily constructed and straightforwardly applied to various systems. The present approach, Poincaré boost dynamics, is applied to the diffusion process of an impurity atom in an amorphous matrix and to conformational changes of a model protein for demonstrating its validity and versatility.