Trace-Level Discrimination and Detection of Polycyclic Aromatic Hydrocarbons via hybrid Photonic-Plasmonic SERS Sensors Integrated with Machine Learning Algorithms

Polycyclic aromatic hydrocarbons (PAHs) pose severe environmental threats, yet their trace-level detection via surface-enhanced Raman spectroscopy (SERS) remains challenging, primarily due to insufficient electromagnetic field enhancement and weak substrate-analyte affinity. Herein, a hybrid photonic-plasmonic SERS platform integrated with machine learning algorithms was introduced to address these critical limitations and enhance detection accuracy as well as analysis efficiency. The engineered architecture, comprising Au film-poly(ionic liquid) (PIL) nanobowl-Au nanosphere, exhibited exceptional detection performance through the synergistic coupling of photonic nanocavities and plasmonic hotspots, which collectively generated high-intensity electromagnetic field regions. Concurrently, the PIL nanobowl structures enabled efficient PAH enrichment via hydrophobic interactions and π-π stacking. This integrated system achieved ultra-sensitive quantification of pyrene, anthracene, phenanthrene, and benzo[a]pyrene, with a limit of detection (LOD) ranging from 6.1–8.5×10^-10 mol/L. Notably, a strong linear relationship (R²=0.998) was established between principal component analysis (PCA)-derived Euclidean distances and molar ratios in binary mixtures. Furthermore, by leveraging PCA and support vector machine (SVM) algorithms, the sensing platform demonstrated robust discriminative capability for seven structurally analogous PAHs—including single-component analytes and binary/ternary mixtures—in real river water matrices. This study presents a data-driven intelligent sensing strategy integrating nanomaterial engineering and machine learning algorithms, thereby enabling rapid, low-cost detection of trace organic contaminants, especially those with similar chemical structures.