Multiphase phosgene detection on machine learning-enhanced fluorescent sensor arrays

Phosgene is a highly toxic, volatile asphyxiant and a notorious chemical warfare agent, posing severe threats to public safety, environmental security, and industrial operations. The capability for rapid, selective, and ultrasensitive on-site detection of phosgene is therefore critically important. Nevertheless, the coexistence of phosgene in liquid, gaseous, and solid-phase environments imposes substantially different sensing requirements, making universal detection across multiple phases a longstanding challenge. Herein, we report the rational design and synthesis of three donor-acceptor (D-A) fluorescent probes and the construction of a conjugated donor-regulated fluorescent sensor array for comprehensive phosgene sensing. The developed platform enables highly sensitive detection and selective of phosgene across liquid, gas, and solid phases, demonstrating excellent adaptability under diverse detection conditions. The detection limit of BTTANT in the liquid phase was 0.48 μM, in the gas phase 0.88 ppm, and the fluorescence quenching rate in the solid phase was 88.28%. By integrating RGB colorimetric analysis with a smartphone, we achieved portable on-site quantification. Leveraging machine learning to enhance recognition strategies, we performed in-depth analysis of multidimensional response signals to distinguish phosgene from various structurally similar interferents with high precision. This study established an integrated fluorescence sensing-intelligent recognition detection platform, providing a significant impetus for the rapid, accurate, and in-situ monitoring of phosgene in complex environments.