Data-Driven Methods for Building Reduced Kinetic Monte Carlo Models of Complex Chemistry from Molecular Dynamics Simulations

Abstract

Complex chemical processes such as those found in combustion, the decomposition of energetic materials, and the chemistry of planetary interiors, are typically studied at the atomistic level using molecular dynamics (MD) simulations. A nascent but growing trend in many areas of science and technology is to consider a data-driven approach to studying complex processes, and molecular dynamics simulations, especially at high temperatures and pressures, are a prime example of an area ripe for disruption with this approach. MD simulations are expensive, but each simulation generates a wealth of data. In this chapter, we discuss a statistical learning framework for extracting information about the underlying chemical reactions observed in MD data, and using it to build a fast kinetic Monte Carlo (KMC) model of the corresponding chemical reaction network. We will show our KMC models can not only extrapolate the behavior of the chemical system by as much as an order of magnitude in time but can also be used to study the dynamics of entirely different chemical trajectories. We will also discuss a new and efficient data-driven algorithm for reducing our learned KMC models using L1-regularization. This allows us to reduce complex chemical reaction networks consisting of thousands of reactions in a matter of minutes.