Sb-doped SnO2 gas sensor array with machine learning: Enhanced selectivity at room temperature
Journal
Journal of Alloys and Compounds
Journal Volume
1047
Start Page
182784
ISSN
09258388
Date Issued
2025-12-05
Author(s)
Lin, Yan-Fong
Chi, Yu-Chen
Chen, Shih-Hong
Lin, Zhan-Yu
Hsieh, Hung-Yu
Dong, Bo-Chang
Huang, Chun-Ying
Abstract
We report a single-material, room-temperature gas sensor array strategy that utilizes Sb-doped SnO2 films—fabricated via atmospheric pressure chemical vapor deposition (APCVD) for selective detection of CO, NH3, H2, and NO2. By systematically varying Sb doping levels (0–4 wt%), we create distinct sensor “fingerprints” within the same base material, thereby eliminating the need to integrate multiple materials. Different Sb concentrations affect electron concentration, oxygen vacancy formation, and surface morphology in unique ways, giving rise to varied response patterns that serve as recognizable signatures for machine learning algorithms. We evaluated these arrays using four supervised models, including Random Forest (RF), Naive Bayes (NB), Support Vector Machine (SVM), and K-Nearest Neighbors (KNN), relying solely on a single feature (i.e., gas response). Using only three Sb doping levels can achieve 100 % classification accuracy under certain conditions, highlighting the potential of Sb doping for cost-effective and highly selective sensor arrays. This approach paves the way for next-generation single-material, multi-gas identification at room temperature.
Subjects
Gas sensor array
Machine learning
Selectivity
SnO2
SDGs
Publisher
Elsevier Ltd
Type
journal article