Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

HL: Fachverband Halbleiterphysik

HL 1: Optical Properties I

HL 1.2: Vortrag

Montag, 9. März 2026, 09:45–10:00, POT/0006

OptiMate3B: Three-body-information-based prediction of optical spectra for semiconductors and insulators — •Malte Grunert, Max Großmann, and Erich Runge — Institute of Physics and Institute of Micro- and Nanotechnologies, Technische Universität Ilmenau, 98693 Ilmenau, Germany

Optical spectra of semiconductors and insulators are essential both for many technologically important applications as well as for the understanding of materials. However, their computation via adequate ab initio techniques remains prohibitively expensive for either large-scale studies or many realistic systems. Recently however, machine learning techniques have emerged as promising alternative. We present OptiMate-3B, a line graph attention network capable of predicting the optical spectra of semiconductors and insulators under both the independent particle approximation and the random phase approximation - i.e. incorporating local-field effects - within milliseconds. OptiMate-3B incorporates three-body information to capture angular information in crystalline materials, further enhancing its predictive accuracy. Having been trained on tens of thousands of high-fidelity first-principles spectra, the model generalizes across a broad chemical and structural space. Its open and user-friendly interface facilitates direct integration into high-throughput materials workflows, providing an efficient surrogate for ab initio optical spectroscopy.

Keywords: Machine Learning; Semiconductors; Optical Spectra; High-Throughput; DFT