J Colloid Interface Sci. 2026 Jul 5;724(Pt 1):141071. doi: 10.1016/j.jcis.2026.141071. Online ahead of print.

ABSTRACT

Gas sensing plays a critical role across diverse fields, including medical diagnostics, industrial safety, and environmental monitoring. However, conventional gas sensors often suffer from poor selectivity and limited sensitivity, especially at low concentrations. Herein, we present a light-regulated gas-sensing system based on MXene-hydrogel composites, integrated with machine learning, for a low detection limit (5 ppb) and high-accuracy classification. PNIPAM hydrogel provides a reconfigurable and adsorptive surface, while MXene offers excellent electrical conductivity and photothermal conversion. Near-infrared light modulation further enhanced selectivity and reduced response/recovery times. When integrated with machine learning classification algorithms, the sensing system enabled robust classification of ten gas molecules with an accuracy of 98.64%. In a pilot breath-sample discrimination task, the system further distinguished cancer patients from healthy controls with a binary classification accuracy of 97.3%. These findings highlight the potential of combining light-regulated sensing materials with machine learning analysis for compact gas identification and exploratory breath-based screening.

PMID:42424889 | DOI:10.1016/j.jcis.2026.141071