Ultrasensitive SERS Detection and AI-Assisted Differentiation of Structurally Similar Pollutants on Flexible Aluminum...

Surface-enhanced Raman spectroscopy (SERS) achieves ultrasensitive molecular detection; however, reliable discrimination of structurally similar analytes remains a critical challenge due to spectral overlap. Here, we report a flexible aluminum (Al)-based SERS substrate that integrates efficient plasmonic hotspot generation with machine learning-assisted spectral analysis for the detection and differentiation of structurally similar organic dyes, including malachite green, crystal violet, and methylene blue. The as-prepared Al substrate underpins dense and reproducible hotspot formation, exhibiting a high enhancement factor (EF) of 2.11 × 1011 and enabling ultrasensitive detection down to a limit of detection (LOD) of 10-15 M. To address spectral complexity, a data-driven approach combining principal component analysis (PCA) and support vector machine (SVM) classification is employed, achieving a high classification accuracy of 96.7% for single analytes as well as binary and ternary mixtures across multiple concentration levels. The proposed approach offers strong potential for multiplex detection and practical applications in environmental monitoring, food safety, and on-site analytical sensing.