Recent years have seen an explosion of work related to machine-learning (ML) models trained on electronic structure data. These models typically make a locality assumption, which amounts to a neglect of long-range interactions (in particular of electrostatics). While this approach has been undeniably successful, it may become problematic when studying complex polar materials such as oxide catalysts or battery materials. To address this, we have explored how ML can be directly incorporated into electronic structure calculations. This dispenses of the<br/>separation between electronic structure theory as a mere data generation device on one hand, and ML as a mathematical fitting framework on the other. Instead, the ML model is used to predict more fundamental physical relationships. This has the advantage of making the ML models more transferable and data-efficient, since the mathematical framework of the electronic structure calculation is retained. Importantly, it also allows the natural incorporation of long-range physics.<br/><br/>In this talk, a recently reported size-extensive ML representation of the electron density will be discussed.[1] This was used to develop so-called Kernel Density Functional Approximations (KDFAs), trained on coupled cluster correlation energies. Importantly, KDFAs are pure functionals of the electron density that are nonetheless able to capture the non-local nature of the correlation energy. Currently, we are further developing this concept to obtain a self-consistent version of KDFA. This enables the prediction of electronic properties like the dipole moment and the density on the same footing as the energy and thus offers a route to CC-quality electron densities for systems of unprecedented size. A computationally efficient charge equilibration approach based on the same idea will also be discussed.

Describing complex polar materials with physics-enhanced machine learning

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.