Optical fingerprint for gas identification at room temperature using light-activated a-IGZO thin films and machine learning

Abstract

A single metal oxide semiconductor (MOS) based gas sensor with distinct thermal fingerprints can replace gas sensor arrays using machine learning (ML) techniques. However, the gradual adjustment of operating temperatures hinders real-time monitoring capabilities and accelerates material degradation. In this study, we utilize a light-activated a-IGZO gas sensor and vary light intensities to generate optical fingerprints, replacing thermal fingerprints for gas identification. To evaluate the ability of the sensor to discriminate between O₃, NO₂, acetone, and toluene, we conducted Principal Component Analysis (PCA) based on their response features. The identification accuracy of the sensor is evaluated using four ML methods: Support Vector Machines (SVM), Naive Bayes (NB), K-nearest neighbors (KNN), and Random Forest (RF). Furthermore, the concentration prediction for each target gas is performed using linear regression and SVM regression models. These findings illustrate that employing a single light-activated a-IGZO sensor, combined with a PCA-driven ML algorithm, achieves an accuracy of over 90 % in discriminating between gas types and concentrations. This strategy paves the way for using a single MOS-based sensor for gas identification at room temperature.