InGaN-GaN-MQW-ZnO based e-nose sensors for nitrogen dioxide detection using advanced machine learning approaches

초록

Accurate detection of nitrogen dioxide (NO<inf>2</inf>) at low concentrations is essential for environmental and health monitoring. This study presents a high-performance NO<inf>2</inf> sensor based on a ZnO-coated InGaN-GaN multi-quantum well (Z-IG-G-MQW) structure. Among the different configurations, the Z-IG-G-MQW<inf>3</inf> sensor exhibits the best performance, achieving a detection limit of 3 ppb and a maximum response of 41.453 at 1.0 ppm NO<inf>2</inf>. Compared to bare GaN and other MQW sensors, it shows enhancements of 2.381-, 1.693-, and 1.377-times, respectively, due to improved gas adsorption, charge transport, and heterojunction efficiency. The sensor demonstrates excellent selectivity toward NO<inf>2</inf>, even in the presence of interfering gases, and maintains stable performance under 0–50 % relative humidity (RH), though higher humidity reduces sensitivity due to adsorption site competition. To improve predictive accuracy under variable conditions, a machine learning approach is implemented. In Trial 1, where full concentration data are used for training and testing, the neural network model achieves the highest accuracy (R² = 0.98). In Trial 2, designed to evaluate generalizability using boundary training data, the Poly2 model performs best (R² = 0.96). This work demonstrates the synergy between advanced heterostructure engineering and machine learning for real-time NO<inf>2</inf> sensing in environmental and healthcare applications. © 2025 Elsevier B.V.

키워드