Fluctuation-Enhanced Sensing of Organic Solvent Vapors Mixture by Machine Learning

Graphene back-gated FETs operate as gas sensors and can detect different gases by the fluctuation-enhanced gas sensing (FES) method, utilizing low-frequency noise measurements. Individual Lorentzian components are observed in low-frequency noise and identified by their corner frequency <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">f</i><sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">c</sub>, which is characteristic of the adsorbed-desorbed gas molecules. The sensor was irradiated by UV light (275 nm) to enhance its sensing properties and reduce humidity impact. The estimated resistance noise power spectral density can evaluate gas mixture components by applying a detection algorithm. We proposed an efficient method of power spectra processing to realize this task. It utilizes a support vector machine (SVM) detection algorithm implemented in a MATLAB environment. The power spectrum is an input data vector comprising flicker noise and additive Lorentzians having intensities proportional to the gas concentration responsible for its generation. The results confirmed that a single graphene back-gated FET can determine the components of the studied gas mixture (chloroform, acetone). The accuracy of gas detection depends on the intensity of the generated noise components related to gas molecules’ activity. The selected algorithm was the most efficient among other machine learning algorithms available in MATLAB software. We underline that the proposed method is characterized by a low limit of gas detection (LOD) and robust against gas sensor drift in time because it only considers noise components related to the selected gases.