Physics-Driven Transferability: TrIP2's Framework for Interatomic Potentials

TrIP2 is an advanced version of the Transformer Interatomic Potential (TrIP) trained on the expanded ANI-2x dataset, including more diverse molecular configurations with sulfur, fluorine, and chlorine. It leverages the equivariant SE(3)-Transformer architecture, incorporating physical biases and continuous atomic representations. TrIP was introduced as a highly promising transferable interatomic potential, which we show here to generalize to new atom types with no alterations to the underlying model design. Benchmarking on COMP6 energy and force calculations, structure minimization tasks, and QC torsion drive energy profiles, as well as applications to molecules with unexpected conformational energy minima, demonstrates TrIP2's high accuracy and transferability with competitive performance against ANI-2x, the original TrIP, AIMNet2, and MACE-OFF23. Notably, TrIP2 achieves state-of-the-art force prediction performance on the COMP6 benchmarks and outperforms all reference models in structure minimization tasks. Without requiring any architectural modifications, TrIP2 successfully capitalizes on additional training data to deliver enhanced generalizability and precision, establishing itself as a robust and scalable framework capable of accommodating future expansions or applications to new domains with minimal retraining.