Transition metal single-atom doped MoS2 for gas adsorption: A combined density functional theory and machine learning study

Indoor air quality is intrinsically linked to human health, making accurate detection essential, in which gas sensors play a crucial role. Recent advancements in two-dimensional materials and doping technologies have led to the development of gas sensors with enhanced performance. Due to the limitations of traditional experimental methods, the combination of first-principles and machine learning method has emerged as a novel approach for material screening. In this study, we present a comprehensive investigation utilizing first-principles and machine learning methods to analyze the adsorption behaviors of indoor air gases on transition metal single-atom doped MoS2 (TM-MoS2) materials. The stable TM-MoS2 structures were identified through density functional theory, with formaldehyde (HCHO), ammonia (NH3), and hydrogen sulfide (H2S) selected as target gases. Six machine learning models were employed to establish the relationship between the adsorption energies of these target gases and the structural features of TM-MoS2. The SHapley Additive exPlanations approach was used to elucidate the importance of various features for the best predictive model. Our work provides a straightforward method for screening novel MoS2-based gas-sensitive materials.