Design and Performance Evaluation of Machine Learning-Based Terahertz Metasurface Chemical Sensor

Area of Science:

• Terahertz (THz) sensing
• Metasurface technology
• Plasmonic materials

Background:

• Chemical sensing requires highly sensitive and accurate detection methods.
• Terahertz metasurfaces offer unique properties for sensing applications.
• Graphene and plasmonic materials are promising for enhancing sensor performance.

Purpose of the Study:

• To design and optimize a terahertz metasurface sensor for sensitive chemical detection.
• To explore the use of graphene and multiple resonator designs for enhanced sensitivity.
• To integrate machine learning for sensor design optimization and performance prediction.

Main Methods:

• Fabrication of a terahertz metasurface sensor with circular and square ring resonators.
• Incorporation of graphene and plasmonic materials.
• Application of Random Forest regression for design enhancement and performance prediction.
• Simulation and analysis of sensor sensitivity, detection limit, and quality factor.

Main Results:

• Achieved a high sensitivity of 417 GHz/RIU and a low detection limit of 0.264 RIU for ethanol and benzene.
• Demonstrated a significant reduction (approx. 90%) in simulation time and computational requirements using machine learning.
• Obtained a high-quality factor of 14.476.
• Showcased potential for 2-bit encoding applications via graphene chemical potential modulation.

Conclusions:

• The proposed terahertz metasurface sensor offers a high-performance solution for chemical sensing and environmental monitoring.
• The integration of machine learning significantly improves the efficiency of sensor design and development.
• This work represents a significant advancement in terahertz sensing technology through novel materials, structures, and computational methods.