Sampling-accelerated First-principles Prediction of Phonon Scattering Rates for Converged Thermal Conductivity and Radiative Properties

TL;DR

This paper introduces a sampling-based method to efficiently estimate phonon scattering rates, significantly reducing computational costs and enabling accurate, converged predictions of thermal conductivity and radiative properties for materials like silicon.

Contribution

A novel sampling and maximum likelihood estimation approach that accelerates first-principles phonon scattering calculations by over 99%, achieving convergence with larger q-meshes.

Findings

Achieved converged thermal conductivity for silicon that matches experimental data.

Reduced computational cost of phonon scattering calculations by over 99%.

Enabled high-resolution predictions for thermal and radiative properties.

Abstract

First-principles prediction of thermal conductivity and radiative properties is crucial. However, computing phonon scattering, especially for four-phonon scattering, could be prohibitively expensive, and the thermal conductivity even for silicon was still under-predicted and not converged in the literature. Here we propose a method to estimate scattering rates from a small sample of scattering processes using maximum likelihood estimation. The computational cost of estimating scattering rates and associated thermal conductivity and radiative properties is dramatically reduced by over 99%. This allows us to use an unprecedented q-mesh of 32*32*32 for silicon and achieve a converged thermal conductivity value that agrees much better with experiments. The accuracy and efficiency of our approach make it ideal for the high-throughput screening of materials for thermal and optical applications.