Machine Learning-Based Optimization of a High-Performance Dual-Band Graphene THz Metamaterial Absorber
IEEE Transactions on Plasma Science · 2026-06 · 54 (6) · 3292-3306
DOI: 10.1109/tps.2026.3687996 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Plasma Science, 2026-06. Verfasst von Youssef Amraoui und 4 weiteren Autor:innen. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 39 Literaturangaben.
Abstract (Originalsprache)
This article presents the design and analysis of a dual-band graphene-based terahertz (THz) metamaterial absorber with enhanced absorption performance and machine learning-assisted optimization. The recommended structure resonates at 3.15 and 4.12 THz, achieving high absorptivities of 99.73% and 99.99%, with quality factors of 6.35 and 5.63, respectively. Owing to the unique tunable surface conductivity of graphene, the absorber exhibits dynamic reconfigurability, enabling efficient control of its electromagnetic response. The absorption mechanism is thoroughly investigated through impedance matching conditions, magnetic and electric field distributions, and parametric analysis, confirming efficient electromagnetic energy confinement. The metamaterial design described can be applied as a refractive index (RI) sensor with a high sensitivity of 0.490 THz/RIU of RI over an RI range of 1–2, for an analyte thickness of $0.5~\mu $ m. To further improve design efficiency, several machine learning (ML) regression models are employed to predict the absorber’s performance based on key physical parameters. A comparative analysis of random forest (RF), K-nearest neighbors (KNNs), decision tree (DT), and extra trees (ET) models is conducted. Among these, the ET regressor demonstrates superior predictive capability, achieving low error metrics and high accuracy. The ET model achieved a root mean squared error of 0.003, a mean absolute error (MAE) of 0.001, a mean squared error of 0.00001, a variance of 98.63%, and an R2 of 98.6%. The integration of ML considerably reduces computational cost and allows a rapid prediction of absorption characteristics. The proposed framework proposes an effective approach for the optimization and design of THz metamaterial absorbers and shows strong potential for applications in sensing, material characterization, and communication systems.
Zitieren
APA
Amraoui, Y., Halkhams, I., Alami, R. E., Jamil, M. O., & Qjidaa, H. (2026). Machine Learning-Based Optimization of a High-Performance Dual-Band Graphene THz Metamaterial Absorber. IEEE Transactions on Plasma Science, 54(6), 3292-3306. https://doi.org/10.1109/tps.2026.3687996
Vancouver
Amraoui Y, Halkhams I, Alami RE, Jamil MO, Qjidaa H. Machine Learning-Based Optimization of a High-Performance Dual-Band Graphene THz Metamaterial Absorber. IEEE Transactions on Plasma Science. 2026;54(6):3292-3306. doi:10.1109/tps.2026.3687996
BibTeX
@article{Amraoui2026,
author = {Amraoui, Youssef and Halkhams, Imane and Alami, Rachid El and Jamil, Mohammed Ouazzani and Qjidaa, Hassan},
title = {Machine Learning-Based Optimization of a High-Performance Dual-Band Graphene THz Metamaterial Absorber},
journal = {IEEE Transactions on Plasma Science},
year = {2026},
volume = {54},
number = {6},
pages = {3292--3306},
doi = {10.1109/tps.2026.3687996}
}
EUVIMED