An integrated microfluidic, metal oxide semiconductor gas sensor combined with machine learning optimization for multiplexed greenhouse gas detection

点击翻译按钮获取中文摘要

Significance: Anthropogenic greenhouse gas (GHG) emissions in the earth's atmosphere have resulted in the gradual heating of the atmosphere, causing severe damage to human health, and negatively impacting the economy. The continuous atmospheric monitoring of the GHGs, such as carbon dioxide (CO2) and methane (CH4), the two most abundant GHGs, is thus necessary to prevent environmental deterioration. Gas chromatography and mass spectrometry are the most commonly used techniques for analyzing GHGs. However, these methods are expensive, time-consuming, and require bulky devices and highly skilled personnel. Recent research and industrial efforts have focused on replacing the conventional analysis techniques with the lab-on-chip (LOC)-based sensors for detecting GHGs. LOC-based sensors offer alternative miniaturized analysis tools, with advantages, such as enhancing accuracy and reducing processing time and cost. In addition, a complete analysis can be performed on a single system, enhancing portability.Innovation & Objectives: Our long-term goal is to develop a "smart gas sensing" technique that enables portable, low-cost, highly selective, and efficient detection of GHGs. In collaboration with the industry partner, Pro-Flange, we will develop an ultrasensitive microfluidic, metal oxide semiconductor sensor that can detect CO2 and CH4. The proposed sensor will be characterized in terms of microchannel geometry, sensor recovery time, and selective coating materials for CH4 and CO2 to detect small concentrations (ppm level) in real-time. Enhanced selectivity will be achieved by implementing novel coatings on the microchannel surface and optimizing the microfluidic channel geometry. The sensitivity will be improved by electrodepositing nanostructures onto the MOS to increase the surface area. In addition, we will develop a machine learning (ML) algorithm employing pattern recognition methods by collecting gas data from the literature and our sensor to optimize the sensor design.