Selectivity Optimization of GDC-Based Mixed-Potential Volatile Organic Compound Gas Sensors

Selectivity is an important criterion for evaluating the gas-sensing performance of a sensor. In this work, we focus on compounding with a second component into the sensitive material and combining the sensor with machine learning to enhance the selectivity of the sensor. NiFe₂O₄ and ZnFe₂O₄ are prepared by the sol-gel method, and ZnFe₂O₄ is mixed with NiFe₂O₄ at different mass ratios. The sensor fabricated from the mixed sensitive materials achieves a significant increase in response to five different volatile organic compound (VOC) gases. Among them, the sensor with a NiFe₂O₄: ZnFe₂O₄ ratio of 1:2 exhibits the best sensing performance to five VOC gases. This sensor demonstrates a detection range for Triethylamine (TEA) from 0.01 to 100 ppm, with a response value of -97.1 mV for 50 ppm TEA at 350 °C. In addition, the sensor with a NiFe₂O₄: ZnFe₂O₄ ratio of 1:3 has an extremely wide detection range for TEA from 2 ppb to 200 ppm. All devices exhibit good humidity resistance and reliable repeatability. To further enhance the selectivity of the device, five machine learning models are trained using eigenvalues of response values, response time, maximum reaction rate, and ratio coefficients. Of these, a recognition rate of 94% for the five gases is achieved using the Random Forest. This work explores the potential of machine learning in the field of sensors and provides an effective method to improve the selectivity of sensors.