Orb-v3: atomistic simulation at scale

TL;DR

Orb-v3 introduces a highly efficient atomistic simulation model that balances speed, memory, and accuracy, enabling large-scale simulations with significant performance improvements over previous models.

Contribution

It presents the Orb-v3 interatomic potential family, achieving near state-of-the-art performance with reduced latency and memory, and challenges the necessity of equivariance and conservatism in modeling physical properties.

Findings

Non-equivariant, non-conservative models can accurately predict physical properties.

Orb-v3 achieves >10x reduction in latency and >8x reduction in memory.

The model advances the performance-speed-memory trade-off frontier.

Abstract

We introduce Orb-v3, the next generation of the Orb family of universal interatomic potentials. Models in this family expand the performance-speed-memory Pareto frontier, offering near SoTA performance across a range of evaluations with a >10x reduction in latency and > 8x reduction in memory. Our experiments systematically traverse this frontier, charting the trade-off induced by roto-equivariance, conservatism and graph sparsity. Contrary to recent literature, we find that non-equivariant, non-conservative architectures can accurately model physical properties, including those which require higher-order derivatives of the potential energy surface. This model release is guided by the principle that the most valuable foundation models for atomic simulation will excel on all fronts: accuracy, latency and system size scalability. The reward for doing so is a new era of computational chemistry driven by high-throughput and mesoscale all-atom simulations.