Modelling the dielectric constants of crystals using machine learning

Abstract

The relative permittivity of a crystal is a fundamental property that links microscopic chemical bonding to macroscopic electromagnetic response. Multiple models, including analytical, numerical and statistical descriptions, have been made to understand and predict dielectric behaviour. Analytical models are often limited to a particular type of compounds, whereas machine learning (ML) models often lack interpretability. Here, we combine supervised ML, density functional perturbation theory, and analysis based on game theory to predict and explain the physical trends in optical dielectric constants of crystals. Two ML models, support vector regression and deep neural networks, were trained on a dataset of 1,364 dielectric constants. Shapley additive explanations (SHAP) analysis of the ML models reveals that they recover correlations described by textbook Clausius-Mossotti and Penn models, which gives confidence in their ability to describe physical behavior, while providing superior predictive power.

14 pages, 4 figures, 4 tables

References in corpus (7)

• Scikit-learn: Machine Learning in Python
• Phonons and related properties of extended systems from density-functional perturbation theory
• A Unified Approach to Interpreting Model Predictions
• Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals
• Universal Fragment Descriptors for Predicting Electronic Properties of Inorganic Crystals
• Benchmarking Materials Property Prediction Methods: The Matbench Test Set and Automatminer Reference Algorithm
• Searching for Materials with High Refractive Index and Wide Band Gap: A First-Principles High-Throughput Study

Cited by in corpus (7)

• Interpretable and Explainable Machine Learning for Materials Science and Chemistry
• Perovskite-Inspired Materials for Photovoltaics -- From Design to Devices
• Deep learning of spectra: Predicting the dielectric function of semiconductors
• Identification of High-Dielectric Constant Compounds from Statistical Design
• Bridging Fidelities to Predict Nanoindentation Tip Radii Using Interpretable Deep Learning Models
• Entropy-based Active Learning of Graph Neural Network Surrogate Models for Materials Properties
• General Learning of the Electric Response of Inorganic Materials