Machine-learning-assisted discrimination of NO2 and NH3 using a highly responsive optoelectrically driven CsPbBr3 single-microcrystal gas sensor

Halide perovskite single crystals offer a promising platform for gas sensing application owing to their structural and optoelectrical properties, which enable high sensitivity toward various gaseous analytes. However, the discrimination of various gases within one device remains challenging. This study proposes a design of optoelectronic gas sensor based on CsPbBr3 perovskite single microcrystal for the detection of NO2 and NH3 in the mixture with air at room temperature. Tunable light intensity excitation affords fast response time of 2 s and high sensitivity with detection limit reaching 0.1 and 0.24 ppm for NO2 and NH3 in the mixture with air, respectively. In combination with machine-learning algorithms, the exact type of analyte is identified with 96% accuracy using a Random Forest Classifier. Additionally, using a regression model derived from the Langmuir adsorption equation, the exact concentration of gas was estimated with mean absolute percentage errors in the range of 20.1%–43.9%. The reported results establish single-crystal perovskite gas sensors as promising devices for real-time environmental monitoring.