Efficient Interfacial Electron Modulation Coupled with Machine Learning for Intelligent Microfluidic Photoelectrochemical-Electrochromic Sensing of Tetracycline

Abstract

Herein, an innovative microfluidic photoelectrochemical-electrochromic (PEC-EC) dual-mode sensing platform integrated with a machine learning (ML) quantitative strategy was constructed for the sensitive detection of tetracycline (TC). First, a brand-new Z-scheme WO3:Yb,Tm/Bi2S3 composite was utilized as a photoanode matrix to supply stable electrons to the PB@N-CDs@PB electrochromic material. Density functional theory (DFT) calculations including work function and differential charge density verified the successful construction of the Z-scheme heterostructure, which accelerated interfacial charge separation and greatly boosted photocurrent output. Second, to achieve effective signal amplification, a PCN-224-Ag-hyaluronic acid (HA) composite was designed for the controlled release of Ag+, triumphantly triggering in situ formation of photoactive AgBiS2 on the photoanode. Meanwhile, both PCN-224 and the photoelectrode could synergistically generate reactive oxygen species (ROS), achieving a highly efficient photocatalytic degradation of TC. Third, the three-channel microfluidic chip with favorable portability and high throughput exhibited ultralow detection limits of 0.25 pmol/L (PEC channel) and 3.1 pmol/L (EC channel). More importantly, a machine learning (ML) strategy was employed using the random forest (RF) algorithm to extract and fuse multimodal RGB features from electrochromic images. This intelligent calibration technology effectively improved the quantitative accuracy of the TC detection. This work provides a new design strategy for modulating interfacial electron transport and demonstrates machine learning as a powerful tool for accurate antibiotic monitoring in food and environmental samples.