Complex chemical processes, such as the decomposition of energetic materials and the chemistry of planetary interiors, are typically studied using large-scale molecular dynamics simulations that can run for weeks on high performance parallel machines. These computations may involve thousands of atoms forming hundreds of molecular species and undergoing thousands of reactions. It is natural to wonder whether this wealth of data can be utilized to build more efficient, interpretable, and predictive models. In this talk, we will use techniques from statistical learning to develop a framework for constructing kinetic Monte Carlo (kMC) models from molecular dynamics data. We demonstrate our framework throughout on a system of high-temperature high-pressure liquid methane, thought to be a major component of gas giant planetary interiors. We will show that 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 with a high degree of fidelity.<br/><br/>Our method first uses bond length and duration criteria to coarse grain from atomistic molecular dynamics data in phase space to molecules and elementary reactions. We then use maximum likelihood estimation to fit this coarse grained data to a kMC model that solves a corresponding chemical master equation. We show that the bond duration should be chosen systematically to minimize the error between kinetic Monte Carlo simulations and the corresponding molecular dynamics. We then apply three different model reduction methods to our kMC model: removing rare reactions, solving an integer program, and a new and efficient data-driven algorithm using L1-regularization. For our liquid methane system, less than 15% of the approximately 2000 reactions observed from molecular dynamics are necessary to reproduce the molecular concentrations over time of methane. This reduced set of reactions enables analysis of the important chemistry in this complex system.

Statistical learning of kinetic Monte Carlo models of high temperature chemistry from molecular dynamics

254th National Meeting (2017)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.