Chemiresistive gas sensors play a critical role in environmental monitoring, industrial safety, and medical diagnostics, where high selectivity toward specific target analytes is of paramount importance but remains challenging. Among various sensing materials, SnO₂ is one of the most widely used materials in commercial gas-sensing platforms due to its high sensitivity, low cost, and technological maturity. Conventional strategies to improve selectivity in SnO₂-based sensors primarily rely on metal doping or loading, which modulates surface reactions and electronic structures. However, such approaches are largely guided by empirical trial-and-error, lacking predictive design principles. To address these challenges, this study aims to develop a machine learning–assisted approach to rationally design the selectivity of sensing materials toward target analytes. A dataset of metal-doped SnO₂ systems on different gas is first constructed using density functional theory calculations, from which adsorption energy and electronic descriptors for different metal dopants and gas molecules were extracted. These features are used to train supervised learning models capable of predicting target-specific selectivity in the presence of multiple interfering gases. The established framework is further extended to metal–organic frameworks (MOFs), demonstrating its cross-material applicability. This work provides a general data-driven strategy for the targeted discovery of selective gas-sensing materials and offers fundamental insights into the physicochemical origins of gas selectivity across different material platforms.
Project is currently funded by: State & Other Govt