Dual electromagnetic enhancement engineering of hybrid Au film–poly(ionic liquid) nanohole–SiO2 nanosphere–Au nanosphere substrates integrated with machine learning algorithm for ultrasensitive SERS detection

An Au film–PIL nanohole–SiO 2 nanosphere–Au nanosphere SERS substrate integrated with machine learning was fabricated. • Dual electromagnetic enhancement was realized by waveguide modes of hybrid structure and LSPR of Au nanospheres. • PIL nanoholes around Au nanospheres promoted PAHs adsorption through hydrophobic effect and π-π stacking interaction. • The PCA-SVM model achieved over 95% prediction accuracy for PAHs of various concentrations and species in river water. Surface-enhanced Raman spectroscopy (SERS) is a promising trace detection technique, yet its practical application in environmental pollutant analysis remains severely restricted by insufficient electromagnetic enhancement and weak adsorption capacity toward target analytes. Herein, a novel hybrid SERS substrate with dual electromagnetic enhancement and highly efficient molecular enrichment was facilely fabricated, featuring a hierarchical structure of Au film–poly(ionic liquid) (PIL) nanohole–SiO 2 nanosphere–Au nanosphere. In this system, the waveguide optical modes of the Au film–PIL nanohole–SiO 2 nanosphere structure induce the primary electromagnetic enhancement, while Au nanospheres are precisely positioned in the primary enhancement regions to deliver a secondary electromagnetic enhancement through localized surface plasmon resonance (LSPR), thereby realizing a high enhancement factor of 2.35 × 10 7 for pyrene. Meanwhile, PIL nanoholes facilitate specific adsorption and enrichment of polycyclic aromatic hydrocarbons (PAHs) within hotspots via the hydrophobic effect and π-π stacking interactions, significantly amplifying the SERS response of analytes. The hybrid substrate exhibits excellent analytical performance, with limits of detection (LODs) as low as 5.33 × 10 −10  mol/L for pyrene and 3.97 × 10 −10  mol/L for anthracene, respectively. Furthermore, a combined principal component analysis (PCA) and support vector machine (SVM) algorithm was employed to efficiently extract spectral features and reduce data dimensionality, achieving a prediction accuracy exceeding 94% for varying concentrations and species of PAHs in actual river water samples. Taken together, the combination of this hybrid SERS substrate and the PCA-SVM algorithm provides a novel sensing platform, enabling ultrasensitive, accurate, and intelligent trace-level detection of environmental pollutants in complex matrices. Download: Download high-res image (185KB) Download: Download full-size image