Machine learning–based six-channel dual-peak photonic nose for identifying real organophosphorus nerve agents and their simulants

Abstract

The timely identification and characterization of chemical agents present in battlefield environments are crucial objectives for ensuring national public safety. In this study, we synthesized two polymers with distinct solubility parameters by encapsulating two styrene materials, each featuring different functional groups. The synthesized polymers were utilized as low-refractive-index layers and were alternately combined with three metal–organic frameworks that possessed varying refractive indices to construct multi-species photonic units. Based on the unique interactions of each photonic unit with different types of gases, we developed a six-channel photonic nose with dual-peak and dual-pattern recognition capabilities for distinguishing nine types of organophosphorus chemical agent gases. Moreover, for the first time, the photonic nose has been integrated with a machine learning algorithm to identify nerve agents. The developed photonic nose successfully detected four organophosphorus agents, including sarin, soman, VX and tabun, within 40 s, achieving an impressive accuracy rate of 100 %. Notably, by utilizing the algorithm, the developed photonic nose could effectively distinguish the organophosphorus nerve agents and their five common simulants. Additionally, the photonic nose was quantitatively evaluated using dimethyl methylphosphonate as a representative agent, yielding results that also demonstrated 100 % accuracy. Furthermore, the photonic nose sensor successfully identified mixed agents, highlighting its potential for monitoring public safety concerning organophosphorus nerve agents.