Machine Learning for Quantum Mechanical Properties of Atoms in Molecules

TL;DR

This paper presents machine learning models that accurately predict quantum mechanical properties of atoms in molecules, achieving density functional theory accuracy with scalable computational costs across diverse molecular datasets.

Contribution

The authors develop local, atom-centered machine learning models that predict quantum observables with high accuracy and linear scaling, validated on a large set of small organic molecules.

Findings

Predictions match density functional theory accuracy.

Linear computational scaling demonstrated for large polymers.

Validated on 9,000 diverse organic molecules.

Abstract

We introduce machine learning models of quantum mechanical observables of atoms in molecules. Instant out-of-sample predictions for proton and carbon nuclear chemical shifts, atomic core level excitations, and forces on atoms reach accuracies on par with density functional theory reference. Locality is exploited within non-linear regression via local atom-centered coordinate systems. The approach is validated on a diverse set of 9k small organic molecules. Linear scaling of computational cost in system size is demonstrated for saturated polymers with up to sub-mesoscale lengths.