Related Experiment Video
Updated: Jul 19, 2026
09:33
Demonstration of Equal-Intensity Beam Generation by Dielectric Metasurfaces
Published on: June 7, 2019
6.6K
Supercell-based metasurfaces for arbitrary polarization beam splitting: physics-informed U-Net design with high
Optics Express
|November 11, 2025
Summary
A novel deep learning method designs metasurface beam splitters for precise polarization control. This approach enables independent manipulation of light polarization states, advancing integrated photonics and optical communications.
Area of Science:
- Photonics and Metasurface Technology
- Artificial Intelligence in Optics
Background:
- Growing demand for advanced polarization control in photonic systems.
- Metasurfaces offer unique capabilities for manipulating light polarization.
Purpose of the Study:
- To develop a deep learning approach for designing polarization-multiplexed metasurface beam splitters.
- To achieve arbitrary polarization control using metasurface designs.
Main Methods:
- A physical information-inspired deep learning approach using a modified U-Net architecture.
- Decomposition of target far-field patterns into orthogonal circular polarization components.
- Embedding wave propagation physics into the deep learning model.
Main Results:
- Efficient recovery of phase distributions with a Mean Squared Error (MSE) of 4.3 × 10-3.
- Silicon nanopillar supercell design achieved Jones matrix decoupling.
- Demonstrated 34.11 dB polarization extinction ratio and 63.91% transmission efficiency via FDTD simulations.
Conclusions:
- The proposed method provides a compact and efficient design framework for polarization-control devices.
- Enables independent control of orthogonal polarization states through geometric parameters and PB phase modulation.
- Applicable to integrated photonics, optical communications, and quantum computing.
