Machine Learning Band Gap Predictions: Linking Quasiparticle Self-Consistent GW and LDA-Derived Partial Density of States

TL;DR

This paper combines quasiparticle self-consistent GW calculations with machine learning to accurately and efficiently predict band gaps of materials based on LDA-derived partial density of states, significantly improving over traditional linear models.

Contribution

It introduces a novel approach integrating QSGW data with machine learning to predict band gaps from LDA PDOS, reducing computational costs and enhancing accuracy.

Findings

Model outperforms linear regression methods

Significantly improves band gap prediction accuracy

Demonstrates effectiveness on 1,516 materials from the Materials Project

Abstract

Accurately calculating band gaps for given crystal structures is highly desirable. However, conventional first-principles calculations based on density functional theory (DFT) within the local density approximation (LDA) fail to predict band gaps accurately. To address this issue, the quasi-particle self-consistent GW (QSGW) method is often employed as it is one of the most reliable theoretical approaches for predicting band gaps. Despite its accuracy, QSGW requires significant computational resources. To overcome this limitation, we propose combining QSGW with machine learning. In this study, we applied QSGW to 1,516 materials from the Materials Project [https://materialsproject.org/] and used machine learning to predict QSGW band gaps as a function of the partial density of states (PDOS) in LDA. Our results demonstrate that the proposed model significantly outperforms linear regression approaches with linearly-independent descriptor generation [https://github.com/Hitoshi-FUJII/LIDG]. This model is a prototype for predicting material properties based on PDOS.