Multi-Dimensional Driven Verification Platform of Chemical Vapor Compositions

Since gas molecules do not exist in a single state in nature, future gas sensing technology will have to move toward predicting and determining the state of the gas mixture itself, rather than being limited to a specific single gas in the air. In this study, we propose a platform that detects gas in a mixed gas state under humidity conditions like human exhaled gas and predicts the type and mixture state of the gas. Our subminiature sensor is a multi-dimensional nano-electrode sensor. For each channel, six different ssDNA functionalized graphene are arranged, enabling multi-dimensional sensing at the same time. Our sensor confirmed the reactivity to the chemical vapor composition under the mixing conditions of various ratios of NO, NO₂, NH₃, and H₂S gases under low and high humidity conditions, and showed high reactivity even under high humidity conditions. Chemical vapor in a mixed state is simultaneously detected in multiple dimensions by our sensors, and the data is input in the form of electrical signals to our artificial intelligence model. Our convolutional neural network-based deep learning model is a very fast and efficient model that can process multi-dimensional data simultaneously and is trained and optimized through experimental data. Our artificial intelligence model can recognize the detected electrical signal, determine the type and ratio of gas included in the chemical vapor composition at the time of detection, and determine the status of harmful gases in the air. Our chemical vapor composition determination platform has verified its performance by achieving a recognition rate of over 99% in both low and high humidity conditions. Our research results are not limited to human expiratory gas diagnosis but are expected to serve as an excellent research background in various fields such as manufacturing, aerospace, and personal air management systems that require detection of harmful gases in real environments.