Discrimination of trace nitroaromatics using linear discriminant analysis on aerosol jet printed fluorescent sensor arrays

In this work, we report on fluorescent sensor arrays fabricated by aerosol jet printing on glass substrates to detect explosives-related nitroaromatic species. The printed sensor arrays consist of six different fluorescent polymers responding to nitroaromatic vapors through a photo-induced electron transfer. This results in a quenched fluorescence proportional to the vapor concentration. Distinct fluorescence quenching patterns are detected for nitroaromatic species including nitrobenzene, 1,3-dinitrobenzene and 2,4-dinitrotoluene. The detected fingerprints are evaluated at low concentrations of only 1, 3 and 10 parts-per-billion in air. Linear discriminant analysis is used to train each sensor array enabling the discrimination of the target analyte vapors. To investigate the reproducibility of multiple sensor arrays on a single substrate, the measured fluorescence quenching patterns are used to benchmark the linear discriminant models. For this purpose, the target analytes and vapor concentrations are predicted for each sensor array. On average, we report low and reproducible misclassification rates of about 4 % indicating excellent discriminatory abilities at low concentrations close to the detection limits. We conclude that digital printing of fluorescent polymers offers the potential to realize low-cost sensor arrays for a reliable detection of trace explosives.