|
We demonstrate a novel, fully-integrated approach to spectral sensing in the near-infrared range suitable for analyzing the chemical composition of organic materials. The sensor consists of 16 detector pixels, each forming a resonant-cavity enhanced photodetector consisting of an InGaAs/InP photodiode and a tuning layer enclosed in a planar cavity formed by two metal mirrors. For wavelengths meeting the resonance condition of the optical cavity, the absorption in the photodiode is enhanced, which leads to a wavelength-specific response of the photodetector. As the thickness of the tuning layer is varied throughout the pixels, each of the 16 photodetectors features an individual complex spectral response with several peaks of about 50 nm linewidth and responsivity above 0.1 A/W. All pixels together cover the whole wavelength range from 900 nm to 1700 nm, allowing for the analysis of broad spectral features typical for diffuse reflectance spectra of organic materials in the near-infrared range. The photocurrents read-out from the spectral sensors can be combined with chemometric analysis methods to determine the material composition. We demonstrate the performance of the spectral sensor for the determinate of moisture in rice grains, leading to a coefficient of determination of R² = 0.97. Other demonstrated applications include the quantification of the sugar content in tomatoes, fat and protein content in raw cow milk and the classification of different types of plastic. With a size of 1.5 mm by 1.5mm and a fabrication scheme based on optical lithography, this on-chip spectral sensor yields potential for large-scale production. Together with the mechanical stability of the sensor, this approach is an important step towards portable, low-cost spectral sensing solutions.
|
