Dresden 2026 – scientific programme

Parts | Days | Selection | Search | Updates | Downloads | Help

MM: Fachverband Metall- und Materialphysik

MM 19: Poster Session

MM 19.49: Poster

Tuesday, March 10, 2026, 18:00–20:00, P5

Transferable Hamiltonian-learning model for large-scale finite temperature electronic-structure calculations — •Kaiwen Chen, Martin Schwade, and David Egger — Physics Department, TUM School of Natural Sciences, Technical University of Munich, 85748 Garching, Germany

Exploring the optoelectronic properties of large-scale systems across various temperatures and structures using conventional density functional theory (DFT) often encounters significant computational challenges. Recent advancements in machine learning (ML) Hamiltonians and the wide availability of DFT-databases have made it possible to train models, capable of predicting accurate Hamiltonians and electronic structure across the chemical space. However, creating a model that can determine temperature-dependent electronic structure with such transferability remains a difficult task. Building on our earlier work involving a physics-informed temperature-transferable Hamiltonian-learning model [1], we introduce an extension of this method that enables it to be trained on a wider field of compositions and thus be able to predict accurate effective Hamiltonians for different chemical compositions.

[1] M. Schwade, S. Zhang, F. Vonhoff, F. P. Delgado, D. A. Egger, *Physics-informed Hamiltonian learning for large-scale optoelectronic property prediction,* arXiv:2508.20536 (2025).

Keywords: Hamiltonian-learning; Machine learning; tight binding; electronic structure; Halide perovskites