Molecular-Level Recognition Surfaces Drive Machine-Learning Discrimination of Homologous Gases

Area of Science:

• Materials Science
• Chemical Sensing
• Machine Learning

Background:

• Chemoresistive sensors struggle to differentiate homologous gases due to similar electronic responses.
• Machine learning (ML) shows promise for gas discrimination but requires distinct signals.

Purpose of the Study:

• To develop molecular-level recognition surfaces for enhanced ML-assisted gas sensing.
• To improve the discrimination and quantification of chemically homologous gases, particularly ammonia and amines.

Main Methods:

• Conformal integration of amorphous phosphotungstic acid (∼2 nm) onto WO3 nanowires.
• Creation of diverse Lewis and Brønsted acid sites to modulate gas adsorption.
• Utilizing temperature-dependent response and transient kinetics for characteristic fingerprint spectra.

Main Results:

• Achieved 91.0% recognition accuracy for four ammonia/amine gases with minimal ML data.
• Successfully quantified ethylamine/dimethylamine isomer mixtures with high precision (R² > 0.99, error < 4%).
• Demonstrated excellent generalization capabilities for the sensing system.

Conclusions:

• The near molecular-level amorphization strategy effectively modulates gas recognition.
• This approach provides a powerful platform for ML-driven discrimination and quantification of homologous gases.
• The developed sensing system overcomes previous limitations in resolving complex gas mixtures.