Augmentation of Universal Potentials for Broad Applications

TL;DR

This paper explores enhancing universal potentials like CHGNet for broader applications, demonstrating that fine-tuning or $ riangle$-learning improves predictions for specific systems and helps explain surface defect phenomena.

Contribution

It introduces augmentation methods such as fine-tuning and $ riangle$-learning to improve universal potentials' accuracy on specific systems.

Findings

Augmentation improves prediction accuracy for cluster and surface systems.

Fine-tuning and $ riangle$-learning enhance the performance of CHGNet.

The approach explains experimentally observed surface defects.

Abstract

Universal potentials open the door for DFT level calculations at a fraction of their cost. We find that for application to systems outside the scope of its training data, CHGNet\cite{deng2023chgnet} has the potential to succeed out of the box, but can also fail significantly in predicting the ground state configuration. We demonstrate that via fine-tuning or a $\Delta$-learning approach it is possible to augment the overall performance of universal potentials for specific cluster and surface systems. We utilize this to investigate and explain experimentally observed defects in the Ag(111)-O surface reconstruction and explain the mechanics behind its formation.