PVA/AgNPs hydrogel SERS substrate combined with machine learning for highly sensitive detection of organic selenium species

Abstract

Accurate detection of organic selenium speciation in soil and water was essential for preserving ecological safety and protecting human health. The performance of a surface-enhanced Raman scattering (SERS) sensor was critically dependent on the structural characteristics of the substrate material. In this work, a simple, low-cost and recyclable 3D porous polyvinyl alcohol/silver nanoparticles (PVA/AgNPs) hydrogel SERS substrate was prepared by the cyclic freeze-thaw method.This substrate exhibited exceptional SERS performance for both L-selenomethionine (SeMet) and Se-(methyl) seleno-L-cysteine (SeMCys). The enhancement factor (EF) reached values as high as 5.91×10 8 , and the limit of detection (LOD) was as low as 3.16 μg/L. Density functional theory (DFT) calculations indicated that charge transfer from the organic selenium molecules to the substrate was the primary mechanism for signal enhancement. Machine learning assisted analysis of SERS spectra effectively mitigated interferences from complex sample matrices. The support vector machine (SVM) model achieved a classification accuracy of 93% on the test set. The support vector regression (SVR) model achieved a coefficient of determination (R 2 ) greater than 0.991 on the prediction set, with root mean square error (RMSE) and mean absolute error (MAE) values of 35.884 μg/L and 8.663 μg/L, respectively. The proposed method enabled accurate identification and precise quantification of organic selenium species in authentic water and soil samples, demonstrated potential for practical application.