Machine Learning-Assisted Development of Sensitive Electrode Materials for Mixed Potential-Type NO₂ Gas Sensors

Area of Science:

• Materials Science
• Chemical Sensing
• Computational Chemistry

Background:

• Yttrium-stabilized zirconia (YSZ)-based sensors are crucial for automotive exhaust gas detection.
• Developing new sensitive electrode materials for mixed potential-type NO2 sensors is challenging, with limited progress in recent years.
• Traditional methods for discovering new sensing materials are slow and inefficient.

Purpose of the Study:

• To employ machine learning to accelerate the discovery of novel sensitive electrode materials for NO2 detection.
• To identify high-performance materials for YSZ-based mixed potential-type gas sensors.
• To overcome the limitations of traditional trial-and-error approaches in materials discovery.

Main Methods:

• Training four tree-based machine learning models on a dataset of approximately 8000 materials.
• Selecting over 400 potential candidate materials using the trained models.
• Synthesizing and testing 13 selected materials, including those with unexploited elements and a non-metal material (BPO4), for NO2 sensing performance.

Main Results:

• The machine learning models successfully predicted numerous promising electrode materials.
• All 13 experimentally tested materials demonstrated good gas-sensing performance for NO2.
• A notable finding was the identification of Boron Phosphate (BPO4), a non-metal material, as an effective NO2 sensor material.

Conclusions:

• Machine learning significantly accelerates the identification of new sensitive electrode materials for NO2 detection.
• The developed models can efficiently screen a large number of materials, reducing experimental effort.
• This approach opens new avenues for discovering advanced materials for gas sensing applications, including non-metal-based sensors.