Design of a Deep Learning-Based Metalens Color Router for RGB-NIR Sensing

Area of Science:

• Optics and Photonics
• Materials Science
• Artificial Intelligence

Background:

• Metalenses offer advanced light modulation capabilities.
• Traditional metalens design is computationally intensive and time-consuming.
• Existing color separation methods often lack high optical efficiency.

Purpose of the Study:

• To develop an efficient metalens-based color router.
• To overcome limitations of traditional metalens design methods.
• To achieve broadband spectral separation from visible to near-infrared light.

Main Methods:

• Proposed a deep learning network for metalens forward prediction.
• Combined deep learning with particle swarm optimization for efficient design.
• Simulated and validated the metalens performance across a broad wavelength range.

Main Results:

• The designed metalens color router effectively separates red, green, blue, and near-infrared light.
• Achieved precise focusing of separated spectra into designated areas.
• Attained a high optical efficiency of 40%, outperforming traditional color filters.

Conclusions:

• The integrated deep learning and optimization approach enables efficient metalens design.
• The developed color router demonstrates superior performance in spectral separation and focusing.
• This work paves the way for advanced optical components with enhanced efficiency.