Benchmarking\nReconstructive Spectrometer with Multiresonant\nCavities

Recent years have seen the rapid development of miniaturized\nreconstructive\nspectrometers (RSs), yet they still confront a range of technical\nchallenges, such as bandwidth/resolution ratio, sensing speed, and/or\npower efficiency. Reported RS designs often suffer from insufficient\ndecorrelation between sampling channels, which, in essence, is due\nto inadequate engineering of sampling responses. This in turn results\nin poor spectral-pixel-to-channel ratios (SPCRs), typically restricted\nat single digits. So far, there lacks a general guideline for manipulating\nRS sampling responses for the effectiveness of spectral information\nacquisition. In this study, we shed light on a fundamental parameter\nfrom the compressive sensing (CS) theorythe average mutual\ncorrelation coefficient <i>ν</i>and provide\ninsight into how it serves as a critical benchmark in RS design. To\nthis end, we propose a novel RS design with multiresonant cavities,\nconsisting of a series of partial reflective interfaces. Such multicavity\nconfiguration allows the superlative optimization of sampling matrices\nto achieve minimized <i>ν</i>. Experimentally, we\nimplement a single-shot, dual-band RS on a SiN platform, realizing\nan overall operation bandwidth of 270 nm and a <0.5 nm resolution\nwith only 15 sampling channels per band. This leads to a record high\nSPCR of 18.0. Moreover, the proposed multicavity design can be readily\nadapted to various photonic platforms, ranging from optical fibers\nto free-space optics. For instance, we showcase that by employing\nmultilayer coatings, an ultrabroadband RS can be optimized to exhibit\na 700 nm bandwidth with an SPCR of over 100.