Ting Ma / Harbin Institute of Technology (Shenzhen)
Xianjin Liu / Harbin Institute of Technology (Shenzhen)
Qiwen Bao / Harbin Institute of Technology (Shenzhen)
Jun-Jun Xiao / Harbin Institute of Technology (Shenzhen)
The continuous advancement of science and technology has led to an increasing demand for the miniaturization and portability of spectrometers. As a novel type of spectral analysis instrument, the miniaturized polarization spectrometer, with its significant advantages of small size, lightweight and low cost, shows broad application prospects in the fields of material science, biomedicine, environmental monitoring and consumer electronics. In this study, we introduce a design methodology for a polarization-dependent microspectrometer based on a differentiable diffraction model. This model links the optical responses of meta-atoms to their structural parameters using deep learning techniques. The differentiable surrogate model is then embedded in an inverse design pipeline, enabling the integration of local transmission coefficient engineering and global engineering of the scatterer in a unified framework, which enables end-to-end optimization of the design target to the meta-atom’s structure parameters. This innovative design modulates the incident light through only a single metasurface and accurately determines the spectral and polarization information of the illumination from a single image captured by a monochrome camera. The system expands the horizons of portable spectroscopy and promises to be integrated with smartphones and lab-on-a-chip platforms, playing an important role in on-site, on-time, in-situ polarization spectroscopic analysis.