BIGDML: Towards Exact Machine Learning Force Fields for Materials

TL;DR

BIGDML introduces a highly data-efficient machine learning approach for accurate force fields applicable to diverse materials, achieving state-of-the-art precision with minimal training data and enabling advanced molecular dynamics simulations.

Contribution

The paper presents BIGDML, a novel symmetry-aware ML force field method that requires only 10-200 geometries for training, outperforming existing models in accuracy and efficiency.

Findings

Achieves energy errors below 1 meV per atom.

Successfully models dynamics of complex materials including 2D and 3D semiconductors.

Demonstrates nuclear quantum effects and diffusion behaviors in simulations.

Abstract

Machine-learning force fields (MLFF) should be accurate, computationally and data efficient, and applicable to molecules, materials, and interfaces thereof. Currently, MLFFs often introduce tradeoffs that restrict their practical applicability to small subsets of chemical space or require exhaustive datasets for training. Here, we introduce the Bravais-Inspired Gradient-Domain Machine Learning (BIGDML) approach and demonstrate its ability to construct reliable force fields using a training set with just 10-200 geometries for materials including pristine and defect-containing 2D and 3D semiconductors and metals, as well as chemisorbed and physisorbed atomic and molecular adsorbates on surfaces. The BIGDML model employs the full relevant symmetry group for a given material, does not assume artificial atom types or localization of atomic interactions and exhibits high data efficiency and state-of-the-art energy accuracies (errors substantially below 1 meV per atom) for an extended set of materials. Extensive path-integral molecular dynamics carried out with BIGDML models demonstrate the counterintuitive localization of benzene--graphene dynamics induced by nuclear quantum effects and allow to rationalize the Arrhenius behavior of hydrogen diffusion coefficient in a Pd crystal for a wide range of temperatures.