Exposure to Volatile Organic Compounds (VOCs) is a significant public health and workplace safety issue. While several gold-standard analytical techniques exist for VOC detection, most of them work ex-situ and often lack sufficient sensitivity and molecular chemical specificity. 

Recently we developed an ultrasensitive approach for detecting and analyzing toxic VOCs in workplaces by combining Infrared (IR) spectroscopy with Machine Learning (ML) models. Focusing on six aromatic VOCs, we produce a fully comprehensive gas-phase IR spectral database further introducing, for the first time, universal IR calibration curves. Through these, ML-based system enables automatic, real-time detection of VOCs at concentrations below 1 ppm, supporting long-term exposure and multiple gases simultaneously monitoring. This research was carried out by the SapienzaTerahertz group (T. Mancini, A. D’Arco and S. Lupi) in collaboration with INFN, INAIL, and University of Roma Tre, and published on JECE.