EUVIMEDEUVIMED entdecken
Crossref

Deep reinforcement learning-optimized inverse design of nanophotonic sensor

Md. Ikrama Hossain, Tasnia Nafs, Sakif Yeaser, A. K. M. Rakib, Rakibul Hasan Sagor

Optics Express · 24.08.2026 · 34 (17) · 31027

DOI: 10.1364/oe.604282 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Optics Express, online erschienen am 10.08.2026. Verfasst von Md. Ikrama Hossain und 4 weiteren Autor:innen. Die Arbeit nennt 91 Literaturangaben.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

A deep reinforcement learning (deep RL) based inverse design framework for geometric optimization of metal–insulator–metal (MIM) based plasmonic refractive index (RI) sensors is proposed in this paper. Traditional design methodologies rely extensively on manual trial-and-error and exhaustive parametric sweeps to navigate the nonlinear trade-offs between sensitivity and spectral resolution. In contrast, our approach utilizes a deep Q-network (DQN) integrated with full-wave finite-difference time-domain (FDTD) simulations to establish an automated optimization loop that refines geometric parameters based on spectral feedback from Lumerical. By systematically optimizing the structural dimensions of a nanorod-enhanced resonator, the framework reached an optimized state, achieving a peak sensitivity of 3291.84 nm/RIU and maximizing the figure of merit (FOM) of 37.94 RIU −1 . The final deep RL-optimized sensor was further simulated for detecting transformer oil aging. The interaction between Lumerical FDTD and deep RL, and how each geometric parameter can impact the performance of the RI sensor is discussed comprehensively in this study. To eliminate the dependence on human intuition and pre-labeled datasets, the proposed framework provides a data-efficient paradigm for maximizing light-matter interaction and can be implemented with high scalability for various structures of nanophotonic devices.

Zitieren

APA

Hossain, M. I., Nafs, T., Yeaser, S., Rakib, A. K. M., & Sagor, R. H. (2026). Deep reinforcement learning-optimized inverse design of nanophotonic sensor. Optics Express, 34(17), 31027. https://doi.org/10.1364/oe.604282

Vancouver

Hossain MI, Nafs T, Yeaser S, Rakib AKM, Sagor RH. Deep reinforcement learning-optimized inverse design of nanophotonic sensor. Optics Express. 2026;34(17):31027. doi:10.1364/oe.604282

BibTeX
@article{Hossain2026,
  author = {Hossain, Md. Ikrama and Nafs, Tasnia and Yeaser, Sakif and Rakib, A. K. M. and Sagor, Rakibul Hasan},
  title = {Deep reinforcement learning-optimized inverse design of nanophotonic sensor},
  journal = {Optics Express},
  year = {2026},
  volume = {34},
  number = {17},
  pages = {31027},
  doi = {10.1364/oe.604282}
}
Kostenlos starten →