Machine Learning-Assisted Sensor Based on CsPbBr3@ZnO Nanocrystals for Identifying Methanol in Mixed Environments

Methanol is a respiratory biomarker for pulmonary diseases, including COVID-19, and is a common chemical that may harm people if they are accidentally exposed to it. It is significant to effectively identify methanol in complex environments, yet few sensors can do so. In this work, the strategy of coating perovskites with metal oxides is proposed to synthesize core-shell CsPbBr3@ZnO nanocrystals. The CsPbBr3@ZnO sensor displays a response/ recovery time of 3.27/3.11 s to 10 ppm methanol at room temperature, with a detection limit of 1 ppm. Using machine learning algorithms, the sensor can effectively identify methanol from an unknown gas mixture with 94% accuracy. Meanwhile, density functional theory is used to reveal the formation process of the core-shell structure and the target gas identification mechanism. The strong adsorption between CsPbBr3 and the ligand zinc acetylacetonate lays the foundation for the formation of the core-shell structure. The crystal structure, density of states, and band structure were influenced by different gases, which results in different response/recovery behaviors and makes it possible to identify methanol from mixed environments. Furthermore, due to the formation of type II band alignment, the gas response performance of the sensor is further improved under UV light irradiation.