Molecular Hessian matrices from a machine learning random forest regression algorithm

TL;DR

This paper introduces a machine learning approach using a random forest to rapidly and accurately predict molecular Hessian matrices, enabling efficient estimation of vibrational properties across different molecules.

Contribution

The novel method applies a random forest model to predict molecular Hessians directly from internal coordinates, ensuring invariance and broad applicability beyond the training set.

Findings

Accurate Hessian predictions for molecules in QM7 and QM9 datasets.

Estimates of vibrational frequencies, normal modes, and zero point energies.

Model demonstrates transferability to larger molecules.

Abstract

In this article we present a machine learning model to obtain fast and accurate estimates of the molecular Hessian matrix. In this model, based on a random forest, the second derivatives of the energy with respect to redundant internal coordinates are learned individually. The internal coordinates together with their specific representation guarantee rotational and translational invariance. The model is trained on a subset of the QM7 data set, but is shown to be applicable to larger molecules picked from the QM9 data set. From the predicted Hessian it is also possible to obtain reasonable estimates of the vibrational frequencies, normal modes and zero point energies of the molecules.