Enabling Weather-Independent Gas Detection through Deep Learning on Light-Activated Sensors

Area of Science:

• Materials Science
• Chemical Sensing
• Nanotechnology

Background:

• Light-activated gas sensors offer low-temperature, low-power detection.
• Integrating materials onto micro light-emitting diode (μLED) platforms enhances sensor performance.

Purpose of the Study:

• To develop a high-performance sensor for simultaneous NO2 and H2O detection.
• To leverage μLEDs and deep learning for enhanced, weather-independent gas sensing.

Main Methods:

• Direct integration of Bismuth (Bi)-doped Indium Oxide (In2O3) nanofibers onto μLED platforms.
• Utilizing blue illumination for sensor activation and a convolutional neural network (CNN) for signal analysis.

Main Results:

• Achieved high NO2 sensitivity (response value of 264.9 at 1 ppm) with fast response/recovery times (<30 s).
• Demonstrated accurate prediction of NO2 and H2O concentrations with 99% classification accuracy and 10% regression error.
• Enabled weather-independent sensing under variable outdoor conditions.

Conclusions:

• The Bi-doped In2O3/μLED system with deep learning provides effective real-time environmental monitoring.
• This approach enhances light activation efficiency for superior NO2 sensing.
• The sensor system shows significant potential for practical applications in environmental monitoring.