Multianalyte\nDetection with Metasurface-Based Midinfrared\nMicrospectrometer

Midinfrared (2.5–25 μm) spectroscopy is\nan ideal tool\nfor identifying chemicals in a nondestructive manner. The traditional\nplatform is a Fourier transform infrared (FTIR) spectrometer, but\nthis is too bulky, expensive, and power-hungry for many applications.\nThere is therefore a growing demand for small, lightweight, and cost-effective\nmicrospectrometers for use in the field. One emerging platform is\nthe filter-array detector-array microspectrometer. It pairs a broadband\ndetector array with a thin and rigid array of spectral filters to\noffer a robust, compact platform for real-time in situ sensing. However,\nmost demonstrations have only focused on identifying a single chemical\nagainst a null sample, even though many applications would involve\nmultianalyte detection. In this work, we show a rare attempt at simultaneously\ntracking multiple analytes with a metasurface filter-array microspectrometer.\nThe metasurface consists of periodic lattices of subwavelength circular\napertures in an aluminum layer to create an array of bandpass filters.\nThe filter array is imaged with an off-the-shelf microbolometer via\na reverse-lens imaging setup to simultaneously monitor the concentration\nof ethanol and methanol in gasoline. This represents an important\napplication of fuel quality monitoring. Chemometric models (PLS and\nSVR) are trained and tested on gasoline blends with ethanol and methanol\ncontents, both ranging from 0% to 20% v/v. A support vector machine\nregression (SVR) model with a cubic kernel was found to have the lowest\ncombined prediction errors. The root-mean-square-error of prediction\n(RMSEP) for ethanol and methanol are 1.23% and 1.84% v/v; the corresponding\npseudounivariate limit of detection is found to be 4.22% and 6.86%\nv/v, respectively. This work takes the emerging field of metasurface-based\nmid-infrared spectrometers from single- to multianalyte detection,\nthereby considerably expanding their range of potential applications.