Including physics-informed atomization constraints in neural networks for reactive chemistry

Machine learning interatomic potentials (MLIPs) have emerged as powerful tools for investigating atomistic systems with high accuracy and relatively low computational cost. However, a common and unaddressed challenge with many current neural network MLIP models forms is their limited ability to accurately predict the relative energies of systems containing isolated or nearly isolated atoms, which appear in various reactive processes. To address this limitation, we present a mathematical technique for modifying any existing atom-centered neural network architecture to account for the energies of isolated atoms. The result produces a consistent prediction of atomization energy (AE) of a system using minimal constraints on the model. Using this technique, we build a model architecture we call HIP-NN-AE, an AE-constrained version of the Hierarchically Interacting Particle Neural Network (HIP-NN), as well as ANI-AE, the AE-constrained version of the Accurate Neural Network Engine for Molecular Energies (ANI). Our results demonstrate AE consistency of AE-constrained models, which drastically improves the AE predictions for the models. We compare the AE-constrained approach to an unconstrained models as well as models from the literature in other scenarios, such as bond dissociation energies, bond dissociation pathways, and extensibility tests. These results show that the constraints improve the model performance in some of these tasks, and do not negatively affect the performance on any tasks. The AE constraint approach thus offers a robust solution to the challenges posed by isolated atoms in energy prediction tasks.