Gao Zhi-Wei, Yu Yan, Chen Shi-Hua, Li Yong-Yu, Liu Zi-Hao, Yang Meng, Li Pei-Hua, Song Zong-Yin, Huang Xing-Jiu
Key Laboratory of Environmental Optics and Technology, And Environmental Materials and Pollution Control Laboratory, Institute of Solid State Physics, HFIPS, Chinese Academy of Sciences, Hefei 230031, China; Department of Materials Science and Engineering, University of Science and Technology of China, Hefei 230026, China; Institute of Environment, Hefei Comprehensive National Science Center, Hefei 230088, China.
Key Laboratory of Environmental Optics and Technology, And Environmental Materials and Pollution Control Laboratory, Institute of Solid State Physics, HFIPS, Chinese Academy of Sciences, Hefei 230031, China.
J Hazard Mater. 2025 Jul 5;491:138030. doi: 10.1016/j.jhazmat.2025.138030. Epub 2025 Mar 23.
Simultaneous quantification of multiple heavy metal ions remains a significant challenge in electrochemical methods, as complex high-throughput data from signal interference cannot be accurately analyzed through individual expertise and calibration curves. In this study, machine learning techniques were introduced to co-detect Cd(II) and Cu(II), with their electrochemical interference mechanisms explored on highly active CoP/CoP heterostructures. The random forest (RF) model initially identified key feature variables in response currents, which were subsequently input into the convolutional neural network (CNN) to uncover the relationship between electrochemical signals and ion concentrations, demonstrating excellent reliability with R values of 0.996 for both Cd(II) and Cu(II). The root mean square error (RMSE) values for Cd(II) and Cu(II) were 0.0177 and 0.0206 μM, respectively, indicating high predictive accuracy. The experiments and theory calculations revealed that Cu(II) preferentially bonded with P sites over Cd(II). Enhanced electron transfer from Co to P atoms and weakened Cu-P bonds facilitated Cu(II) reduction and desorption from CoP/CoP, thereby boosting electrochemical signals, while Cd(II) signals were inhibited due to active site loss. Herein, the integration of machine learning provides robust support for simultaneous detection of multiple analytes, accelerating the practical application of electrochemical methods in environmental monitoring.
在电化学方法中，同时对多种重金属离子进行定量分析仍然是一项重大挑战，因为信号干扰产生的复杂高通量数据无法通过个人专业知识和校准曲线进行准确分析。在本研究中，引入机器学习技术来同时检测Cd(II)和Cu(II)，并在高活性CoP/CoP异质结构上探索它们的电化学干扰机制。随机森林（RF）模型首先识别响应电流中的关键特征变量，随后将这些变量输入卷积神经网络（CNN）以揭示电化学信号与离子浓度之间的关系，Cd(II)和Cu(II)的R值均为0.996，显示出出色的可靠性。Cd(II)和Cu(II)的均方根误差（RMSE）值分别为0.0177和0.0206 μM，表明预测准确性高。实验和理论计算表明，Cu(II)比Cd(II)更倾向于与P位点结合。从Co到P原子的电子转移增强以及Cu-P键减弱促进了Cu(II)从CoP/CoP上的还原和解吸，从而增强了电化学信号，而Cd(II)信号由于活性位点的损失而受到抑制。在此，机器学习的整合为多种分析物的同时检测提供了有力支持，加速了电化学方法在环境监测中的实际应用。
