Machine learning force fields: Construction, validation, and outlook

TL;DR

This paper discusses the development, validation, and future outlook of machine learning-based force fields that leverage quantum mechanics, emphasizing their workflow, validation through complex simulations, and uncertainty estimation for improved versatility.

Contribution

It introduces a comprehensive workflow for constructing ML force fields, validates them with complex phenomena, and proposes uncertainty estimation to enhance their adaptability.

Findings

Validated force fields can simulate complex phenomena beyond ab initio limits.

Workflow includes environment generation, representation, training set selection, and learning.

Uncertainty estimation helps identify and improve areas of poor performance.

Abstract

Force fields developed with machine learning methods in tandem with quantum mechanics are beginning to find merit, given their (i) low cost, (ii) accuracy, and (iii) versatility. Recently, we proposed one such approach, wherein, the vectorial force on an atom is computed directly from its environment. Here, we discuss the multi-step workflow required for their construction, which begins with generating diverse reference atomic environments and force data, choosing a numerical representation for the atomic environments, down selecting a representative training set, and lastly the learning method itself, for the case of Al. The constructed force field is then validated by simulating complex materials phenomena such as surface melting and stress-strain behavior - that truly go beyond the realm of $ab\ initio$ methods both in length and time scales. To make such force fields truly versatile an attempt to estimate the uncertainty in force predictions is put forth, allowing one to identify areas of poor performance and paving the way for their continual improvement.