Ultrahigh-sensitive and real-time detection of BTXs for occupational safety via infrared spectroscopy coupled with machine learning technique

Exposure to Volatile Organic Compounds (VOCs) is one of the major human and occupational safety concern, as possible human carcinogens. Several gold standard methods are used for their detection in the atmosphere; however, most of them operate ex-situ or do not provide easy discrimination between different molecules with suitable sensitivity. Here, we introduce an ultrasensitive method based on Fourier Transform Infrared (FTIR) spectroscopy coupled with Machine Learning (ML) algorithms to analyse toxic gaseous substances in working sites down to a concentration of less than 1 ppm. We investigate six selected aromatic compounds (BTXs gases and styrene), building an accurate IR gas-phase database and providing, for the first time at the best of our knowledge, universal IR calibration curves still lacking in literature. Starting from this IR gas-phase database, we design and train a ML automatic and rapid recognition method. This advantageous combination between IR spectroscopy, including the estimated IR calibration curves, and ML algorithms demonstrates the strong ability of our strategy in discriminating between different gaseous VOCs indoor, with high accuracy and rapidity even when more compounds are present at the same time. The proposed approach responds to the fundamental needs (i) to evaluate low VOCs concentrations up to values less than 1 ppm (under the legislative levels), (ii) to monitor the VOCs presence in real-time for the accurate estimation of long-exposure levels and (iii) to discriminate the co-exposure at various compounds.