Machine Learning-Assisted ZnIn ₂ S ₄ /TiO ₂ Heterostructure Gas Sensor for Sensitive, Stable, and Reliable Triethylamine Detection in Complex Environments

The development of high-performance gas sensors for real-time detection of triethylamine (TEA) remains challenging due to sensitivity limitations and interference from coexisting gases. To address these challenges, a ZnIn2S4/TiO2 heterostructure was fabricated in this work via a facile hydrothermal strategy to harness the synergistic n-n heterojunction effects. Structural characterizations via SEM, TEM, and XPS verified that TiO2 nanoparticles are uniformly dispersed on the surface of ZnIn2S4 microflowers, forming a well-defined hierarchical heterostructure with intimate interfacial contact─features that lay a fundamental structural basis for efficient charge transfer and abundant active sites, thereby contributing to the enhanced gas-sensing performance. The sensing tests revealed the optimized sensor (ZnIn2S4/TiO2-20) to exhibit outstanding performances at 150 °C, with a response toward 50 ppm TEA reaching 120.2, coupled with rapid response/recovery times of 11 s/12 s, a detection limit of 1 ppm. Notably, the sensor exhibits no significant performance degradation even after 45 cycles of repeatability tests and a 90-day durability test. The mechanistic studies revealed the heterojunction to promote carrier transport and surface adsorption, while TiO2-induced oxygen vacancies enhance the reactive site densities. To mitigate false alarms in complex environments, a KNN+PCA machine learning model was integrated to achieve 100% binary classification accuracy for TEA/non-TEA gases. Overall, the combination of ZnIn2S4/TiO2 heterostructures with machine learning has potential for reliable TEA monitoring in industrial real complex environments.