Obtaining accurate optical spectra is equally necessary for a fundamental physical understanding of semiconductors on the one hand and their technological and economic utilization on the other hand, as well as for many research questions in between, but it remains a challenge both experimentally and theoretically. Machine learning (ML) models present a promising alternative. Recently, the first working ML models became available; however, each one comes with its own significant limitations. In this study, we present OptiMate3B, a ML model that incorporates the advances made separately by the first wave of models. OptiMate3B, a graph attention network utilizing three-body information, is able to predict the frequency-dependent complex dielectric function under the independent particle approximation and the random-phase approximation to quantitative accuracy for a wide range of crystalline semiconductors and insulators within milliseconds. It works particularly well for compounds containing elements that are commonly used for optical and optoelectronic applications. An easy-to-use interface for OptiMate3B significantly lowers the barrier to entry for the broad community.
Grunert et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: