Design and Optimization of Full-Stokes Hyperspectro-Polarimetric Encoding Metasurfaces Based on Conditional...

Area of Science:

• Metasurfaces and optical engineering.
• Advanced materials science.
• Computational electromagnetics.

Background:

• Metasurfaces offer advanced optical detection capabilities.
• Current metasurface design is inefficient, relying on trial-and-error.
• Existing methods limit broadband precision and design efficiency.

Purpose of the Study:

• To develop an efficient, end-to-end framework for designing metasurfaces.
• To integrate physical encoding constraints into the design process.
• To enable on-demand metasurface generation with controlled encoding independence.

Main Methods:

• Proposed an end-to-end conditional multitask learning framework.
• Incorporated correlation constraints as physical conditions during network training.
• Developed a metasurface array screening strategy using forward prediction and inverse design networks.

Main Results:

• Designed metasurface arrays (4x4 to 10x10) with significant Relevance reductions (up to 36.5%).
• Achieved superior performance compared to manual selection and existing ML methods.
• Experimentally validated a 4x4 array for full-Stokes spectro-polarimetric reconstruction (400-900 nm, 4 nm resolution).

Conclusions:

• The proposed framework enables efficient, on-demand metasurface design.
• Demonstrated high-performance spectro-polarimetric reconstruction capabilities.
• Highlights potential for high-performance, integrated multidimensional optical systems.