Machine Learning Approaches to Point Defects in Non-Metallic Materials: A Review of Methods

TL;DR

This review discusses recent machine learning methods for predicting point defect properties in non-metallic materials, highlighting achievements, challenges, and future research directions.

Contribution

It categorizes existing ML approaches, emphasizes dataset quality, and identifies key challenges in modeling charged-defect formation energies.

Findings

ML models predict defect energetics from local structures

MLPs approximate defect potential energy surfaces

Dataset quality impacts model performance

Abstract

We review recent machine-learning (ML) approaches for point defects in non-metallic materials, with an emphasis on defect formation energies. Existing studies largely fall into two categories: direct ML models that predict defect energetics from local structural representations, and machine-learning potentials (MLPs) that approximate the defect-containing potential energy surface. We summarize key achievements as well as persistent bottlenecks, emphasizing that dataset quality often dominates practical model performance. We further identify charged-defect formation energies as a central frontier, where Fermi-level alignment, finite-size corrections, and long-range electrostatics must be handled carefully and consistently to enable meaningful comparisons and transferable predictions across different materials.