Abstract:Noble metal-based particulate catalysts have traditionally been loaded or decorated onto the semiconductor metal oxide (SMO) surfaces to enhance the performance of gas sensors. Fabricating SMO-based sensors for selective detection of particular gases with high sensitivity remains a challenge. This study aims to enhance the efficacy of pure SMO and their heterostructure with gold (Au) decoration, along with an attempt to address the cross-sensitivity towards other gases through machine learning-assisted discriminative analysis. The sensitivity of sputtered ZnO thin films towards different VOCs such as acetone, ethanol, methanol, and toluene was improved by loading particulate Au. The highest sensitivity of 25 and low response time of ~6 s and ~4 s were achieved towards 10 ppm of ethanol and toluene, respectively. Further, Au-loaded ZnTiO 3 -TiO 2 (ZTO) layered heterostructure was also adopted to achieve selectivity towards toluene. ZTO thin film sensor was highly sensitive towards 10 ppm of toluene, with a sensitivity of 49 and a response time of ~6 s. Loading Au on ZnO thin film increases sensitivity towards multiple VOCs, whereas Au on ZTO heterostructure improves the selectivity towards toluene. To overcome the problem arising due to the cross-sensitivity, the idea of operating multiple sensors with different Au loading is proposed. The discriminative property of different sensors as an array, towards VOCs was analysed in detail using machine learning algorithms such as principal component analysis (PCA), linear discriminant analysis (LDA), logistic regression (LR), and artificial neural networks (ANN). The multilayer perceptron model implemented to discriminate multiple VOCs worked at a discrimination accuracy of 100%.

Discriminative analysis of volatile organic compounds using machine-learning assisted Au loaded ZnO and TiO 2 -based thin film sensors