Triple-Band Terahertz Metamaterial Absorber for Label-Free Detection of Industrial Contaminants  with Machine Learning Driven Technique

In this study, we present a triple-band terahertz metamaterial absorber optimized for label-free detection of industrial contaminants and various analytes using a machine learning (ML) optimization technique. The proposed structure consists of asymmetric copper resonators on an FR-4 substrate with a continuous copper ground plane, achieving near-unity absorption at 0.96 THz, 2.112 THz, and 3.94 THz. Detailed parametric studies and near-field analyses demonstrate high refractive index sensitivity, with maximum sensitivity values of 771.3 GHz/RIU and high Figure of Merit (FoM) values of 2.541, 2.376, and 3.361. The sensor exhibits strong polarization insensitivity and angular stability, making it highly robust for real-world applications. The integration of an ML accelerated framework for absorptivity prediction in circular-shaped terahertz metamaterial absorbers achieved superior performance, with the proposed Random Forest model attaining an unprecedented R² = 0.999156 (99.9156%) using an 80–20 train test split, surpassing Extra Trees and KNN, while reducing computational time by over 85% and effectively predicting missing parameter values. Extensive performance evaluations confirm the capability of sensors to detect contaminants in food, fuel, chemicals, pesticides, and biological samples, underscoring its potential for high-precision industrial sensing and real-time quality monitoring.