【Device Papers】Machine Learning-Assisted CdS/Ga₂O₃ Heterojunction PEC Sensor for Sensitive Tetracycline Detection

       Researchers from the Zhejiang A&F University have published a dissertation titled "Machine Learning-Assisted CdS/Ga₂O₃ Heterojunction PEC Sensor for Sensitive Tetracycline Detection" in Electrochimica Acta.

Abstract

       The PEC sensor constructed from wide-bandgap Ga₂O₃ and narrow-bandgap CdS for enhanced tetracycline (TC) detection, overcoming conventional limitations like low sensitivity and complex operation. CdS/Ga₂O₃ composites were synthesized via a one-step hydrothermal method, enabling efficient separation and transport of photogenerated carriers through band modulation. Compared with pure CdS or Ga₂O₃, the heterojunction exhibited a sixfold photocurrent enhancement, improved response time, and superior stability. The sensor demonstrated excellent repeatability (RSD = 0.359 %), selectivity, and sensitivity for TC concentrations from 1 pM to 50 μM, achieving a low detection limit of 0.045 μM. Tests on lake and river water confirmed high practical applicability, with RSD<2 %. To further improve detection accuracy, three machine learning algorithms—SVR, Random Forest, and XGBoost—were applied to photoelectric response data across TC concentrations, using cross-validation and feature selection to build predictive models. Random Forest outperformed others, offering the best error evaluation and fitting accuracy while reinforcing the linear correlation between TC concentration and photoelectric response. This work provides new insights into quantitative antibiotic detection and underscores the potential of heterojunction engineering for PEC sensor optimization.

DOI：

https://doi.org/10.1016/j.electacta.2025.147851