Mid-infrared computational spectroscopy scheme via metasurface filter arrays and physics-data dual-driven spectral reconstruction

This study proposes a mid-infrared (MIR) spectral sensing scheme based on a metasurface filter array integrated with a physics-data dual-driven algorithm. Two types of metasurface structures, namely the silicon-based gold micro-hole array (Si-Au-MHA) and the silicon-based gold micro-pillar array (Si-Au-MPA), were designed through simulation to determine their structural parameters for fabrication, aiming to achieve spectral control across the 3-12 μm band. Samples were fabricated, and the transmission spectra were measured and validated. Furthermore, a physics-data dual-driven spectral reconstruction method combining regularized physical constraint with a neural network was introduced, demonstrating superior performance in both amplitude and spectral profile compared to purely physics-based or data-driven approaches. This work provides a viable solution for portable MIR spectral sensing applications.