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 behavior. Analytical models are often limited to a particular type of compound, 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 1364 dielectric constants. Analysis of Shapley additive explanations 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.
Bibliographical noteFunding Information:
The authors thank financial support from the Yoshida Scholarship Foundation, the Japan Student Services Organization, and the Centre for Doctoral Training on Theory and Simulation of Materials at Imperial College London. This research was also supported by the Creative Materials Discovery Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (Grant No. 2018M3D1A1058536). Through our membership of the UK’s HEC Materials Chemistry Consortium, which is funded by the EPSRC (Grant No. EP/L000202), this work used the ARCHER UK National Supercomputing Service (http://www.archer.ac.uk).
© 2020 Author(s).
All Science Journal Classification (ASJC) codes
- Physics and Astronomy(all)
- Physical and Theoretical Chemistry