Machine Learning‐Enabled Smart Gas Sensing Platform for Identification of Industrial Gases

Abstract

Both ammonia and phosphine are widely used in industrial processes, and yet they are noxious and exhibit detrimental effects on human health. Despite the remarkable progress on sensors development, there are still some limitations, for instance, the requirement of high operating temperatures, and that most sensors are solely dedicated to individual gas monitoring. Herein, an ultrasensitive, highly discriminative platform is demonstrated for the detection and identification of ammonia and phosphine at room temperature using a graphene nanosensor. Graphene is exfoliated and successfully functionalized by copper phthalocyanine derivate. In combination with highly efficient machine learning techniques, the developed graphene nanosensor demonstrates an excellent gas identification performance even at ultralow concentrations: 100 ppb NH3 (accuracy—100.0%, sensitivity—100.0%, specificity—100.0%) and 100 ppb PH3 (accuracy—77.8%, sensitivity—75.0%, and specificity—78.6%). Molecular dynamics simulation results reveal that the copper phthalocyanine derivate molecules attached to the graphene surface facilitate the adsorption of ammonia molecules owing to hydrogen bonding interactions. The developed smart gas sensing platform paves a path to design a highly selective, highly sensitive, miniaturized, low‐power consumption, nondedicated, smart gas sensing system toward a wide spectrum of gases.

Keywords

• ammonia and phosphine
• gas detection
• gas identification
• machine learning techniques
• molecular dynamics simulations
• smart gas sensing