EUROPEAN CONGRESS AND EXHIBITION ON ADVANCED MATERIALS AND PROCESSES - EUROMAT 2021

Fast and accurate interatomic potentials are critical for atomistic modeling in materials science, physics and chemistry. Driven by machine learning (ML) recent years have seen rapid progress in the field. In contrast to most ML models, the atomic cluster expansion (ACE) provides a complete representation of the atomic energy, expressed in polynomial multi-atom basis functions. ACE is amenable to physical and chemical interpretation, while its accuracy and computational efficiency was shown to be as good or better than that of leading ML models.

I will briefly summarize the derivation of ACE and discuss its application to metals and semiconductors. The extension of ACE to further variables, such as atomic charges and magnetic moments, as well as vectorial and tensorial properties will be introduced. I will then show applications of ACE including magnetism and the computation of phase diagrams.