Machine Learning Assisted Pb 2+ Ions Detection Using AlGaN/GaN HEMT Sensors

In this work, a novel machine learning-integrated AlGaN/GaN High Electron Mobility Transistor (HEMT) based sensor is reported for the detection of Pb2+ion contamination in aqueous media. Here, the gate region of the HEMT was functionalized with 2,5-dimercapto-1,3,4-thiadiazole (DMTD) to provide high affinity and selectivity toward Pb²⁺ ions. The DMTD functionalization enabled the device to exhibit excellent sensitivity at pH 5.5. Moreover, the lifetime stability analysis indicated consistent sensing performance over a period of 30 days. Furthermore, machine learning-based methodology was applied to detect the presence of Pb2+ions contamination in solution by analyzing the sensing response of AlGaN/GaN HEMT sensor. Here, a Random Forest classifier was trained on the obtained sensing response. The dataset was pre-processed, balanced using Synthetic Minority Oversampling Technique (SMOTE) to counteract class imbalance, and split into training (80%) and testing (20%) sets. The classifier achieved an accuracy of 90.50%, precision of 1.0 recall of 81.03%, and F1-score of 89.52%. Visualization through confusion matrix, Receiver Operating Characteristic (ROC) curves, and learning curves also demonstrates the robustness and reliability of the proposed model. The integration of machine learning with DMTD functionalized AlGaN/GaN HEMT sensing platforms establishes a promising pathway for rapid, selective and stable Pb2+ion detection in water quality monitoring applications.