Plasmonic sensor using generative adversarial networks integration

Area of Science:

• Photonics
• Sensing Technologies
• Machine Learning

Background:

• Machine learning (ML) is revolutionizing sensing technologies across diverse fields.
• Photonic crystal fibers (PCFs) offer unique properties for advanced sensor development.
• Plasmonic sensing enhances sensitivity by utilizing light-matter interactions at metal surfaces.

Purpose of the Study:

• To integrate ML techniques with PCF-based plasmonic sensing.
• To enhance sensor performance, including sensitivity, resolution, and prediction accuracy.
• To explore the use of generative adversarial networks (GANs) for improving artificial neural network (ANN) models in sensing.

Main Methods:

• Utilizing PCFs with open channels for augmented mode coupling.
• Employing a thin gold layer for efficient plasmon generation.
• Integrating ML models, including ANNs enhanced by GAN-generated data, for performance prediction.

Main Results:

• Achieved maximum wavelength sensitivity of 9000 nm/RIU and amplitude sensitivity of 490.41 RIU^-1.
• Demonstrated a high resolution of 1.11 × 10^-5 RIU and a maximum FOM of 138.04 RIU^-1.
• GAN-enhanced ANN models significantly improved confinement loss prediction accuracy.

Conclusions:

• The integrated ML and PCF plasmonic sensing approach offers superior performance.
• The sensor is versatile for applications in chemical sensing and medical diagnostics.
• GANs provide a viable method for augmenting training data to improve ML model accuracy in sensor applications.