Neural Network‐Assisted End‐to‐End Inverse Design for Polarimetric Microspectrometer DOI
Ting Ma, Xianjin Liu, Qiwen Bao

и другие.

Advanced Materials Technologies, Год журнала: 2025, Номер unknown

Опубликована: Апрель 15, 2025

Abstract Metasurfaces represent a cutting‐edge platform for the miniaturization of spectral and polarimetric sensing applications. However, achieving multi‐channel control has been significant challenge conventional design methodologies. This work introduces an end‐to‐end inverse framework capable directly optimizing metasurface geometry precise optical field manipulation. A cascaded system can be conceptualized as diffractive neural network. By developing deep network that correlates response meta‐atoms with their structural properties, differentiable pipeline is established to drive designs toward multi‐functional requirements. Leveraging this approach, polarization‐independent microspectrometer designed, which unaffected by polarization state incident light theoretically resolution down 2 nanometers across visible spectrum. Furthermore, polarization‐dependent demonstrated simultaneously capture both information from single exposure. device maintains same high also provides determination angle. Such polarization‐sensitive potential applications in detection chiral substances. The results will accelerate progress meta‐optics, implications spectro‐polarimetric detection, multi‐target holographic displays, parallel processing, computing.

Язык: Английский

Efficient Gradient-Based Metasurface Optimization toward the Limits of Wavelength-Polarization Multiplexing DOI
Yanjun Bao, Hongsheng Shi,

Wei Rui

и другие.

Nano Letters, Год журнала: 2025, Номер unknown

Опубликована: Апрель 3, 2025

Polarization and wavelength multiplexing are the two widely employed techniques to improve capacity in metasurfaces. While previous studies have pushed channel numbers of each technique its individual limits, achieving simultaneous limits both still presents challenges. Furthermore, current methods often suffer from computational inefficiencies, hindering their applicability computationally intensive tasks. In this work, we introduce experimentally validate a gradient-based optimization algorithm using deep neural network (DNN) achieve polarization with high efficiency. By leveraging efficiency DNN-based method, further implement nine multiplexed channels (three wavelengths × three polarizations) for large-scale image recognition tasks total 36 classes single-layer metasurface. The classification accuracy reaches 96% simulations 91.5% experiments. Our work sets new benchmark high-capacity inverse design advanced optical elements.

Язык: Английский

Процитировано

0

36‐Channel Spin and Wavelength Co‐Multiplexed Metasurface Holography by Phase‐Gradient Inverse Design DOI Creative Commons

C. G. Park,

Youngsun Jeon,

Junsuk Rho

и другие.

Advanced Science, Год журнала: 2025, Номер unknown

Опубликована: Май 8, 2025

Abstract Metasurface holography has emerged as a versatile tool for manipulating light at subwavelength scales, offering enhanced capabilities in multiplexing high‐resolution holographic images. However, the scalability of channel remains significant challenge. In this paper, high‐capacity single‐cell metasurface is presented capable maximizing channels by images across both spin and wavelength using single‐phase map. The achievement simultaneous left‐ right‐circular polarization states detailed broad spectral range, from visible to near‐infrared wavelengths, metasurface, optimized through an inverse design minimize loss between target output automatic differentiation. phase profile encode multiple without requiring complex meta‐atoms, thereby reducing fabrication complexity while maintaining high performance. Using method, two implementations are demonstrated, 8‐channel hologram covering regions 36‐channel operating full‐visible spectrum 18 wavelengths separated 20‐nm intervals. Furthermore, noise‐related functions incorporated into optimization process suppress background noise inter‐channel crosstalk, resulting significantly improved image quality fidelity. This approach offers reliable solution further photonic applications such displays, optical data storage, information encryption.

Язык: Английский

Процитировано

0

Advancements in Metasurfaces for Polarization Control: A Comprehensive Survey DOI Creative Commons
Humayun Zubair Khan, Junaid Zafar, Abdul Jabbar

и другие.

Next research., Год журнала: 2025, Номер unknown, С. 100407 - 100407

Опубликована: Май 1, 2025

Язык: Английский

Процитировано

0

Neural Network‐Assisted End‐to‐End Inverse Design for Polarimetric Microspectrometer DOI
Ting Ma, Xianjin Liu, Qiwen Bao

и другие.

Advanced Materials Technologies, Год журнала: 2025, Номер unknown

Опубликована: Апрель 15, 2025

Abstract Metasurfaces represent a cutting‐edge platform for the miniaturization of spectral and polarimetric sensing applications. However, achieving multi‐channel control has been significant challenge conventional design methodologies. This work introduces an end‐to‐end inverse framework capable directly optimizing metasurface geometry precise optical field manipulation. A cascaded system can be conceptualized as diffractive neural network. By developing deep network that correlates response meta‐atoms with their structural properties, differentiable pipeline is established to drive designs toward multi‐functional requirements. Leveraging this approach, polarization‐independent microspectrometer designed, which unaffected by polarization state incident light theoretically resolution down 2 nanometers across visible spectrum. Furthermore, polarization‐dependent demonstrated simultaneously capture both information from single exposure. device maintains same high also provides determination angle. Such polarization‐sensitive potential applications in detection chiral substances. The results will accelerate progress meta‐optics, implications spectro‐polarimetric detection, multi‐target holographic displays, parallel processing, computing.

Язык: Английский

Процитировано

0