Nanosheet-type tin oxide gas sensor array for mental stress monitoring

Abstract

Mental stress management has become significantly important because excessive and sustained mental stress can damage human health. In recent years, various biomarkers associated with mental stress have been identified. One such biomarker is allyl mercaptan. A nanosheet-type tin oxide exhibited high gas selectivity for allyl mercaptan; thus, in this study, a sensor array comprising nanosheet-type tin oxide gas sensors was fabricated to detecting allyl mercaptan. Supervised learning algorithms were use to build gas classification models based on the principal component analysis of the sensor signal responses from the sensor array. The comprehensive data provided by the classification models can be used to forecast allyl mercaptan with high accuracy.

Figure 1

Allyl mercaptan conversion percentage (powder: 0.01 g, flow: 100 cm³/min, gas concentration: 20 ppm, tube diameter: 10 mm).

Figure 2

Transmission electron microscopy (TEM) images of (a) nanosheet-type tin oxide and (b) commercial tin oxide nanoparticle with corresponding fast Fourier transform (FFT) (insert).

Figure 3

Field emission scanning electron microscopy (FE-SEM) images of the sensor chip (blue area is electrode); (a) Pt electrodes on Al₂O₃ substrate, (b) between Pt electrodes, and nanosheet-type of tin oxide after synthesis for (c) 0.5 h, (d) 1.0 h, (e) 3.0 h, and (f) 6.0 h.

Figure 4

(a) Electrical resistance variations under air flow. (b) Sensor signal response for 54 ppm allyl mercaptan gas. (c) Electrical resistance variation at 300 °C for 11-ppm allyl mercaptan. (Line is the mean value, with less than 1% change in the range.).

Figure 5

(a) Sensor signal response to various gases. (b) Scores plot of principal component analysis for various gases (Eigenvector values are Table S3). (c) Coordinates plot of air and allyl mercaptan gas on principal component analysis result.

Figure 6

Gas classification models by (a) Gaussian naïve Bayes, (b) linear discriminant analysis, (c) k-nearest neighbor, (d) Random forest, (e) Linear support vector classification (SVC), (f) SVC with linear kernel, (g) SVC with polynomial kernel, and (h) SVC with radial basis function kernel.