Quantitative Decoding of Complex Gas Mixtures Using Mixed-Potential Sensor Arrays

Abstract

Advances in sensor technologies are required to go beyond the limitations of current commercial sensors and enable advanced combustion and emission control for fuel-saving lean-burn and low temperature combustion (LTC) engines. Oxygen l-sensors are an integral part of on-board diagnostics (OBD) of today’s spark-ignition vehicle emission systems, but a suitable analog is not available for NOx, NH3 and NMHC detection to control after-treatment systems and perform OBD functions in lean-burn engines. Mixed-potential sensors fabricated via well-established commercial manufacturing methods present a promising avenue to enable the widespread utilization of NOx, NH3 and NMHC sensing technology. Los Alamos National Lab (LANL) has developed patented pre-commercial prototype mixed-potential sensors exhibiting strong preferential selectivity and sensitivity to these target emission constituents. However, a single device with absolute selectivity remains elusive (1-3). The work herein presents an alternative strategy to absolute selectivity, in which the cross-sensitivity of a set of sensor elements run at varying operating conditions is exploited through the use of Bayesian inference techniques based on physical models of sensor-analyte interactions. We have previously shown that the individual concentration of each analyte species in NH3/C3H6 mixtures can be uniquely determined from the sensor voltage response of a single sensor operating at four different current bias points by employing a Bayesian interference model (4). This model was recently extended to de-convolute the gas speciation of NO/C3H8 mixtures utilizing the voltage response from two sensors of different electrode composition, LSCr|YSZ|Pt and Au|YSZ|Pt. In this work, we seek to further explore the mixed-potential array concept by employing multiple sensors in order to uniquely identify the concentration of NO, NO2, NH3, and C3H8 in complex mixtures relevant to diesel exhaust. Sensor voltage response data collected on a wide range of mixing ratios are used to train the Bayesian model. Gas mixtures not used in the training set are then used to test the accuracy of the model. This research represents the first stages of developing miniaturized mixed-potential electrochemical sensor (MPES) arrays as an all-in-one NOx/NH3/NMHC sensor package that will provide quantitative emissions monitoring and OBD for all lean-burn engine applications. Acknowledgements The research was funded by the Los Alamos National Laboratory Directed Research Development Exploratory Research (LDRD-ER) program.