Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
FM: Fachverband Funktionsmaterialien
FM 12: German-French Focus Session: Materials Research in Polar Oxides – From Domain Engineering to Photonic and Electronic Devices I
FM 12.7: Talk
Wednesday, March 11, 2026, 11:30–11:45, BEY/0138
Machine learned potential for ferroelectric heterostructure BaTiO3/SrTiO3 — •Lan-Tien Hsu1, Jonathan Schmidt2, Aaron Iten2, Nicola Spaldin2, and Anna Grünebohm1 — 1Interdisciplinary Centre for Advanced Materials Simulation (ICAMS), Faculty of Physics and Astronomy, and Center for Interface-Dominated High Performance Materials (ZGH), Ruhr-University Bochum, Germany — 2Department of Materials, ETH Zürich, Zürich, CH-8093, Switzerland
Ferroelectric heterostructures offer a platform to realize topological patterns with functionalities relevant to future electronic devices.[1] While short-range machine-learned interatomic potentials can capture ferroelectric instabilities in pure materials, long-range dipole interactions remain essential for heterostructures, where depolarizing field plays a role.[2] We address this challenge by combining MACE message-passing networks with a latent Ewald-summation scheme[3] capable of learning long-range interactions and inferring Born effective charges without explicit training on response properties. The model predicts transition temperature and spontaneous polarization of BaTiO3 in close agreement with experiments. It further captures the multidomain formed in heterostructures, consistent with coarse-grained effective-Hamiltonian results, and generalizes well to untrained configurations, including those containing oxygen vacancies.
[1] Das et al, Nature 568, 368-372 (2019)
[2] Yu et al, Phys. Rev. B 112, 104324 (2025)
[3] Zhong et al, 10.48550/arXiv.2504.05169
Keywords: ferroelectric; heterogeneous interfaces; machine learning; depolarizing field; domain structure