Snapshot full-Stokes imaging spectropolarimeter based on polarization-resolved metasurface with AI-assisted inverse design

• Applicant: Dr.-Ing. Chia-Wei Chen
• Subject Area: Measurement Systems ; Coating and Surface Technology ; Geodesy, Photogrammetry, Remote Sensing, Geoinformatics, Cartography ; Optics, Quantum Optics and Physics of Atoms, Molecules and Plasmas
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 565580671

This DFG-NSTC joint proposal will contribute to the design, fabrication, and assembly of a novel, compact, and fast sensor system for acquiring quantitative spatial, spectral, and polarization images with a multifunctional metasurface grating. The polarimeter can measure spectroscopic imaging and obtain full Stokes vectors without wavelength scanning. The compactness of metasurfaces allows a miniature size of the imaging spectropolarimeter. In addition, the system does not need mechanical components (e.g., rotary stages for waveplates) to analyze full Stokes vectors. The main idea to achieve this is to use a metasurface grating, which splits the light based on its polarization state. The proposed grating can generate four spectroscopic images with different polarization states by +1 and -1 diffraction orders on two different Cartesian coordinates. The four images provide sufficient information to calculate the Stokes parameters. The polarization-resolved hyperspectral imaging is capable of providing more spectral, spatial and polarization information for researchers and industrial applications, e.g., three-dimensional imaging for geometry and surface reconstruction, measurements of physical properties for birefringence induced by stress and refractive index, object detection and quality inspection, and remote sensing for pollutants for air quality. The design process of metasurfaces usually takes numerous trial and error, and the time requirement depends on the scale and range of design parameters. In order to accelerate and simplify the design process in metasurface for polarization imaging, an inverse design framework based on a generative model (GPT-based) is proposed for metasurface structures (materials and shapes). Structure tokens and structure serialization will be used to convert the metasurface structure into a sequence. The proposed framework can generate suitable metagratings based on the required polarization states and spectra. This foundation model can enhance the diversity and flexibility of the inverse design process. After the structure generation, the forward model (training or physical-based) will evaluate the spectra and polarization states for the generated structures. Then, the selected structures will be further optimized (fine-tuning) by the physical simulation for the final design. This framework will significantly reduce the time requirement compared to the conventional design process. In addition, the model can be applied to any polarization application, e.g., polarimetry, ellipsometry, polarization holography and bioimaging, not limited to the Stokes imaging.

• DFG Programme: Research Grants
• International Connection: Taiwan
• Partner Organisation: National Science and Technology Council (NSTC)
• Cooperation Partner: Professor Dr. Chih-Ming Wang