Data-driven spectral response prediction of composite terahertz metamaterial absorbers using a VAE-transformer-LSTM network
Optical Materials Express · 01.12.2026 · 16 (12) · 4049
DOI: 10.1364/ome.608474 ↗Kurz gefasst
Fachartikel in der Zeitschrift Optical Materials Express, online erschienen am 05.11.2026. Verfasst von Yuying Jiang und 7 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Henan University of Technology, Henan Science and Technology Department.
Abstract (Originalsprache)
Terahertz (THz) metamaterial sensors have attracted significant attention due to their high sensitivity and label-free detection capability. However, the inverse design of metamaterial structures with desired spectral responses remains a challenging task because of the high-dimensional parameter space and nonlinear electromagnetic interactions. In this work, we propose a hybrid inverse design framework that integrates transfer learning with deep reinforcement learning to efficiently optimize THz metamaterial structures. First, a forward prediction model is constructed to establish the mapping between structural parameters and spectral responses, which is further enhanced via transfer learning to improve generalization across different structural configurations. Then, a DRL-based optimization strategy is employed to iteratively search for optimal structural parameters under target spectral constraints. Compared with conventional optimization methods, such as genetic algorithms and random search, the proposed approach demonstrates faster convergence and improved design efficiency. Numerical results show that the designed metamaterial structures achieve high consistency with target spectra, with reduced computational cost. In addition, the framework exhibits good adaptability to different design objectives, indicating its potential for generalized inverse design tasks. Our work provides an efficient and scalable solution for THz metamaterial inverse design and a promising way toward intelligent design of functional photonic devices.
Förderung
- National Natural Science Foundation of China
- Henan University of Technology
- Henan Science and Technology Department
- Education Department of Henan Province
- Ministry of Science and Technology of the People's Republic of China
- Horizontal Research Project of Henan University of Technology
- Supported by CultivationProject of Tuoxin Team in Henan University of Technology
Zitieren
APA
Jiang, Y., Wang, Q., Ge, H., Bu, Y., Lu, X., He, R., Wang, Z., & Ju, C. (2026). Data-driven spectral response prediction of composite terahertz metamaterial absorbers using a VAE-transformer-LSTM network. Optical Materials Express, 16(12), 4049. https://doi.org/10.1364/ome.608474
Vancouver
Jiang Y, Wang Q, Ge H, Bu Y, Lu X, He R, et al. Data-driven spectral response prediction of composite terahertz metamaterial absorbers using a VAE-transformer-LSTM network. Optical Materials Express. 2026;16(12):4049. doi:10.1364/ome.608474
BibTeX
@article{Jiang2026,
author = {Jiang, Yuying and Wang, Quanshu and Ge, Hongyi and Bu, Yuwei and Lu, Xuejing and He, Rong and Wang, Zhigao and Ju, Chenghui},
title = {Data-driven spectral response prediction of composite terahertz metamaterial absorbers using a VAE-transformer-LSTM network},
journal = {Optical Materials Express},
year = {2026},
volume = {16},
number = {12},
pages = {4049},
doi = {10.1364/ome.608474}
}
EUVIMED