Inverse-designed integrated computational spectrometer with sub-ångström resolution
Abstract
Computational spectrometers have garnered significant interest owing to their robust tolerance to fabrication imperfections and potential for compact, cost-effective spectral sensing. However, most existing schemes rely on stochastic design strategies to produce random spectral responses, which often underutilize the device’s inherent performance limits. In this work, we propose an inverse design methodology for computational spectrometers leveraging a modified particle swarm optimization (PSO) algorithm. Using cascaded add-drop microring resonators (MRRs) as fundamental filtering units, we develop a global optimization framework to tailor the spectral responses of the complete MRR filter bank and suppress both auto-correlation and inter-channel cross-correlation. Compared with randomly generated designs, the proposed approach reduces the average auto-correlation HWHM from 0.32 nm to 0.08 nm, demonstrating enhanced spectral encoding capability. Fabricated on an 8-inch silicon nitride (SiN) platform via 180-nm photolithography, the spectrometer achieves a spectral resolution of 80 pm across the C-band, validating the efficacy of the proposed design framework.