Learning the Electrostatic Response of the Electron Density through a Symmetry-Adapted Vector Field Model

Area of Science:

• Computational Chemistry
• Materials Science
• Machine Learning

Background:

• Predicting electron density response to electric fields is a key challenge in atomistic machine learning.
• Existing methods often lack efficiency and scalability for complex molecular and material systems.

Purpose of the Study:

• To develop an efficient and accurate method for predicting electron density response under electric fields.
• To leverage rotational symmetry for improved performance in atomistic machine learning models.

Main Methods:

• Developed symmetry-adapted kernel functions that explicitly incorporate rotational symmetry of vector fields.
• Demonstrated equivariance using rotated water molecules.
• Evaluated efficiency on liquid water, naphthalene crystals, and gold nanoparticles.

Main Results:

• The method shows high efficiency concerning training configurations and features.
• Accurately predicts electronic polarizability scaling with size for gold nanoparticles up to 2000 atoms.
• Provides an accurate and inexpensive strategy for predicting electrostatic response.

Conclusions:

• Symmetry-adapted kernel functions offer a powerful approach for atomistic machine learning.
• The method extends equivariant learning models for electron density prediction.
• Enables accurate and cost-effective electrostatic response predictions for diverse systems.