Tracing Molecular Reactions and Decomposition Using Machine Learning Force Fields and Active Learning

Molecular reactivity spans extended length and time scales, making them costly to simulate using ab initio approaches or limited in chemical transferability/scope using reactive force fields. In this work, we describe how simulations of liquid structure and dynamics of organic molecules undergoing thermal decomposition reactions can be achieved using hybrid density functional theory (DFT), active learning [1], and machine learning force fields (MLFF) [2]. Active learning is essential for several reasons: 1) The overall computational cost of hybrid DFT is reduced dramatically as the uncertainty-aware force field collects only sufficiently-uncorrelated DFT frames; 2) High-temperature configurations with radicals or free gases, which are potential decomposition products, can also be collected on-the-fly; 3) The approach is useful in situations where classical force fields are either unavailable or lacking in expressivity.<br/><br/>When the training data from active learning are fed into a data-efficient equivariant neural network, molecular decomposition and reaction pathways can be traced with first-principles, all-atom resolution by identifying reaction pathways to product formation. We apply the approach to study the thermal decomposition of a “green” solvent used in battery recycling and validate our results against experimental characterization.<br/><br/>[1] Vandermause, J., Xie, Y., Lim, J.S. et al. Nat Commun 13, 5183 (2022).<br/>[2] Musaelian, A., Batzner, S., Johansson, A. et al. Learning local equivariant representations for large-scale atomistic dynamics. Nat Commun 14, 579 (2023).