Ion Gel Modulated Channel Interface Engineering: Multidimensional Recognition of Volatile Organic Compounds in a Single CNTFET

Field-effect transistor (FET)-type sensors have attracted extensive attention for detecting trace hazardous gases by leveraging their multi-parameter capabilities. However, their selectivity is often constrained when identifying volatile organic compounds (VOCs) across different chemical classes due to a reliance on the singular sensing modality of surface charge transfer. In this study, we address this limitation by designing an ionic gel (Ion Gel)-carbon nanotube (CNT) FET (IG-CNTFET). In our design, a meticulously engineered Ion Gel gate dielectric is coupled to the CNT channel, introducing a potent electrical double-layer (EDL) capacitive effect at the interface. The resulting architecture enables a composite-action sensing mechanism, in which gas molecules not only interact directly with the CNT channel but also induce a concurrent redistribution of ions within the EDL. This synergistic modulation of the device's electronic properties fundamentally enriches the sensing modality beyond conventional charge-transfer limits. Four representative VOCs, including dimethyl methylphosphonate (DMMP), triethylamine (TEA), 1,2-dichloroethane (DCE), and aniline, exhibit distinct adsorption behaviors that uniquely modulate carrier transport and shift key electrical parameters like carrier mobility and threshold voltage. We established a multi-dimensional sensing dataset by extracting electrical parameters from the device's transfer curves, which we visualized in radar charts to reveal analyte-specific patterns. For classification, we employed Principal Component Analysis (PCA) for dimensionality reduction, followed by a Multi-Layer Perceptron (MLP) algorithm, thereby enabling the accurate differentiation of the four structurally distinct VOCs. The work pioneers a single-device approach for recognizing multi-VOCs, marking a significant leap toward intelligent, multi-dimensional gas sensing with high selectivity and data richness.