We demonstrate a computational multispectral metasurface employing a 3 \(\times\) 3 photonic crystal array architecture that operates across the longwave infrared spectrum (8–11.5 \(\upmu\) m). The designed structure achieves remarkable optical performance with peak transmittance reaching 75.8% and broadband energy utilization efficiency of 41.37%. Notably, the inter-channel transmittance correlation coefficient of 0.17 indicates superior spectral discrimination compared to conventional grating-based systems. We also considered the angular dependence of the array on the incident light. Additionally, to evaluate the spectral reconstruction performance of the transmission spectra under different photonic crystals, a spectral reconstruction deep learning network was constructed with the mean squared error is 2.86 \(\times 10^{-3}\) . This architecture establishes a hardware-algorithm co-design framework for next-generation infrared multispectral systems, demonstrating the potential for integrated superlattice detectors with sub-100 \(\upmu\) m \(^{2}\) pixel pitch, which represents a critical advancement for portable spectroscopic applications.
