On-demand design of high-Q metasurfaces via physics-conditioned closed-loop learning
Optics Express · 10.08.2026 · 34 (16) · 29478
DOI: 10.1364/oe.603271 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optics Express, online erschienen am 30.07.2026. Verfasst von Yulu Qin und 8 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Zhejiang Provincial Natural Science Foundation.
Abstract (Originalsprache)
Accurate modeling of spectral evolution across broad geometric parameter spaces is critical for designing high-Q metasurfaces based on bound states in the continuum (BICs). However, conventional full-wave simulations are too slow for practical parameter exploration, and existing deep-learning models cannot capture both the ultra-sharp resonances and the dynamic mode-coupling evolution from bright resonances to hybridized states and quasi-BICs. Here, we present a physics-conditioned neural network (PCNN) for high-Q metasurface design, demonstrated on hexagonal boron nitride metasurfaces that support BICs. The architecture integrates a physics-constrained front-end with a feature-wise linear modulation-conditioned U-Net, enabling data-efficient and high-fidelity prediction of sharp resonances. An inverse network retrieves structural parameters from target spectra, which are validated through the forward PCNN, closing the computational design loop. We demonstrate accurate modeling and on-demand design of hybridized eigenstates, toroidal dipole modes, and quasi-BICs across a wide wavelength range (R 2 > 0.96). This work provides a unified platform for studying BIC physics and realizing customizable metasurfaces for sensing, nonlinear, and topological photonics.
Förderung
- National Natural Science Foundation of China
- Zhejiang Provincial Natural Science Foundation
Zitieren
APA
Qin, Y., Chen, Y., Wang, B., Tang, H., Cheng, H., Zhang, F., Hu, H., Zheng, H., & Zhou, X. (2026). On-demand design of high-Q metasurfaces via physics-conditioned closed-loop learning. Optics Express, 34(16), 29478. https://doi.org/10.1364/oe.603271
Vancouver
Qin Y, Chen Y, Wang B, Tang H, Cheng H, Zhang F, et al. On-demand design of high-Q metasurfaces via physics-conditioned closed-loop learning. Optics Express. 2026;34(16):29478. doi:10.1364/oe.603271
BibTeX
@article{Qin2026,
author = {Qin, Yulu and Chen, Yuyi and Wang, Bao and Tang, Huilin and Cheng, Haoyang and Zhang, Feng and Hu, Hanmin and Zheng, Haixia and Zhou, Xiaolong},
title = {On-demand design of high-Q metasurfaces via physics-conditioned closed-loop learning},
journal = {Optics Express},
year = {2026},
volume = {34},
number = {16},
pages = {29478},
doi = {10.1364/oe.603271}
}
EUVIMED