Md Koushik Alam, Massoumeh Nazari, Sumit Goswami, Binbin Weng, University of Oklahoma, Norman, OK
Mid-infrared (MWIR) photodetectors are widely used in applications such as chemical sensing, thermal imaging, and free-space communications, where maximizing light absorption and minimizing noise are essential for optimal performance. Achieving this requires precise control over light–matter interactions, which can be realized using metasurface-based designs. In this study, we demonstrate a Mie metasurface engineering approach to enhance broadband light absorption in narrowband PbSe semiconductors. By tuning the radius and height of disk-shaped PbSe nanostructures, the resonances along the vertical (thickness) and lateral (in-plane) directions can be made to partially overlap across a wavelength range. This overlap traps and concentrates the incoming light within the nanostructures, enhancing the interaction between light and the material and resulting in strong broadband absorption. We find that disks in air achieve over 99% absorption at 3.65 µm, while placing the disks on a SiO_₂ substrate shifts the absorption peak to 3.7 µm due to the higher refractive index increasing the effective optical path, producing a redshift in the resonance. Traditionally, designing such structures requires performing full-wave numerical simulations for each geometric configuration, which is both time-consuming and computationally intensive. To overcome this limitation, we developed a deep neural network (DNN) model capable of predicting absorption spectra directly from the geometric parameters of the PbSe disks. By rapidly exploring a large design space, the model identifies promising configurations that are then verified with numerical simulations, ensuring both accuracy and physical insight. This combined approach of metasurface engineering and neural network assisted design offers a powerful and efficient route to realize high-performance broadband MWIR photodetectors with tailored absorption characteristics, paving the way for next-generation infrared sensing technologies.