Realizing weakly supervised learning by physical information for rapidly and accurately predicting three-dimensional weak electromagnetic field of metallic metasurfaces and zero-shot inverse design

T Tianyou Zeng (School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,) M Manqi Liu (School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,) W Weitao Yang (Department of Radiology, Tongji Hospital, Shanghai Frontiers Science Center of Nanocatalytic Medicine, The Institute for Biomedical Engineering & Nano Science, School of Medicine) K Kunyuan Xu (School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,) S Shuqin Wang J Jianping Guo

Abstract

This work introduces a separable, complex-valued 3D Maxwell loss function to physically constrain a 3D U-Net. This framework enables the prediction of complete vectorial electromagnetic fields for metallic metasurfaces across a broad spectral range, achieving a computational speedup exceeding two orders of magnitude (>100×) relative to the FDTD. A recognized limitation of purely data-driven networks is their inability to accurately predict weak field regions, which can even produce negative spectrum correlations. Incorporating Maxwell loss ensures accurate weak field predictions for absorptive materials. Additionally, this constraint allows effective weak supervision, reducing labeled points needed to just 12.5% of the total field points. More importantly, we have achieved zero-shot inverse design of meta-material structures for arbitrary materials and arbitrary geometries in three dimensions.

Article Details

Volume / Issue Vol. 127, Issue 6
Published August 11, 2025
ISSN 0003-6951
Publisher American Institute of Physics

Journal Info

Applied Physics Letters

American Institute of Physics

ISSN: 0003-6951 Physical Sciences

Authors (6)

T

Tianyou Zeng

School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,

M

Manqi Liu

School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,

W

Weitao Yang

Department of Radiology, Tongji Hospital, Shanghai Frontiers Science Center of Nanocatalytic Medicine, The Institute for Biomedical Engineering & Nano Science, School of Medicine

K

Kunyuan Xu

School of Physics, South China Normal University 1 Key Laboratory of Atomic and Subatomic Structure and Quantum Control (Ministry of Education), Guangdong Basic Research Center of Excellence for Structure and Fundamental Interactions of Matter, , Guangzhou, Guangdong 510006,

S

Shuqin Wang

J

Jianping Guo