Machine-intelligent inkjet-printed α-Fe₂O₃/rGO towards NO₂ quantification in ambient humidity

Metal oxides (MOx) represent one of the most investigated chemiresistive gas sensing platforms in spite of the challenges in selectivity to analytes and interference from humidity (RH). While selectivity is traditionally improved by cross-referencing sensor arrays, interferences from humidity (RH) in ambient environment, to which the majority of the MOx materials are susceptible, cannot be inherently quantified. For standalone MOx sensors, it is therefore difficult to discriminate responses from analytes and humidity. We develop a framework which employs temperature modulation (TM) algorithms and machine learning (ML) approaches using principal component analysis (PCA) and cluster analysis of transient features, to quantify NO₂ concentrations under specific RH conditions. With a single inkjet-printed MOx/reduced graphene oxide (rGO) complementary metal-oxide-semiconductor (CMOS)-integrated sensor, we achieve an overall discrimination accuracy of 97.3%. Our approach may enable the development of predictive systems for humidity sensitive sensors under ambient moisture conditions, towards the realisation of low-power, miniaturised adaptive air quality monitoring.