Molecular electrostatic potentials from machine learning models for dipole and quadrupole predictions

TL;DR

This study demonstrates that machine learning models, especially those incorporating quadrupole moments, can accurately predict molecular electrostatic potentials, significantly enhancing the speed and accuracy of such predictions in organic molecules.

Contribution

The paper introduces ML models that effectively incorporate quadrupole moments to improve MEP predictions, highlighting the importance of quadrupoles in ML-based electrostatic modeling.

Findings

Including quadrupole moments improves MEP prediction accuracy.

ML models trained on QM9 and SPICE datasets perform well across diverse organic molecules.

Quadrupole moments are essential for rapid and accurate MEP estimation.

Abstract

The molecular electrostatic potential (MEP) is a key quantity for describing and predicting intermolecular and ion-molecule interactions. Here, we assess the ability of machine-learning (ML) models to infer the MEP, based on the equivariant graph-convolutional neural network architecture PiNet2 and trained on dipole and quadrupole moments. For the established QM9 dataset, we find that including the quadrupole contribution in the ML models substantially improves their ability to recover the MEP compared to dipole-only models. This trend is confirmed on the SPICE dataset, which spans a much broader region of organic chemical space. Together, this study underscores the central role of the quadrupole moment as a fitting target for ML models aiming at rapid access to the MEP.