Deep Neural Network for Phonon-Assisted Optical Spectra in Semiconductors

TL;DR

This paper introduces a deep learning-based approach combining molecular dynamics and tight-binding models to efficiently simulate temperature-dependent phonon-assisted optical spectra in semiconductors with high accuracy, matching experimental results.

Contribution

The authors develop a novel method that integrates deep learning with ab initio fidelity to enable large-scale, high-accuracy simulations of phonon effects in semiconductors at finite temperatures.

Findings

Accurately models phonon-induced bandgap renormalization.

Replicates experimental optical spectra over five orders of magnitude.

Applicable to complex materials with computational efficiency.

Abstract

Ab initio based accurate simulation of phonon-assisted optical spectra of semiconductors at finite temperatures remains a formidable challenge, as it requires large supercells for phonon sampling and computationally expensive high-accuracy exchange-correlation (XC) functionals. In this work, we present an efficient approach that combines deep learning tight-binding and potential models to address this challenge with ab initio fidelity. By leveraging molecular dynamics for atomic configuration sampling and deep learning-enabled rapid Hamiltonian evaluation, our approach enables large-scale simulations of temperature-dependent optical properties using advanced XC functionals (HSE, SCAN). Demonstrated on silicon and gallium arsenide across temperature 100-400 K, the method accurately captures phonon-induced bandgap renormalization and indirect/direct absorption processes which are in excellent agreement with experimental findings over five orders of magnitude. This work establishes a pathway for high-throughput investigation of electron-phonon coupled phenomena in complex materials, overcoming traditional computational limitations arising from large supercell used with computationally expensive XC-functionals.