Development of an In₂O₃/g-C₃N₄ Chemiresistive Sensor for Machine Learning-Assisted Discrimination of Acetone and Ethanol Vapors

Abstract

Differentiating volatile organic vapors, particularly acetone and ethanol remains a persistent limitation for metal oxide chemiresistive sensing devices. This investigation addresses the challenge by integrating the gas-sensing characteristics of In₂O₃/g-C₃N₄ (Indium Oxide/Graphitic Carbon Nitride) with advanced machine learning (ML) methods. The study encompasses the fabrication, physicochemical characterization, and sensing assessment of In₂O₃/g-C₃N₄ nanostructures toward ethanol and acetone vapors. Gas-sensing measurements performed over concentrations of 250–1000 ppm and operating temperatures between 210 and 300 °C revealed only marginal distinction between the two analytes. To improve their differentiation, the recorded transient sensing profiles were interpreted through the Langmuir–Hinshelwood kinetic model, providing microscopic insight into the surface reaction mechanism. The calculated activation energies were 0.28 eV for ethanol and 0.37 eV for acetone, indicating that ethanol undergoes surface interaction more readily, resulting in a comparatively stronger sensing signal. In addition, Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Artificial Neural Network (ANN), Random Forest (RF), and Light Gradient Boosting Machine (LightGBM) models were applied to the experimental sensing dataset to enable dependable identification and separation of ethanol and acetone vapors, with the ANN model demonstrating the highest classification performance, followed by RF, LightGBM, XGBoost, and SVM.