Machine Learning-Driven Surface Plasmon-Enhanced Dual Spectroscopies Improve Recognition and Real-Time Monitoring of Hazardous Chemicals

To address the challenges of precise identification and real-time monitoring of hazardous chemicals, this work proposes and develops surface-plasmon-enhanced dual spectroscopies (SPEDS). This technique combines highly recognizable surface-enhanced Raman spectroscopy (SERS) with real-time plasmon-mediated differential ultraviolet-visible spectroscopy (P-DUS). The feasibility of this technique is demonstrated by successfully acquiring SPEDS signals of thiourea with a plasmonic gold colloidal system. By combining SPEDS with machine learning algorithms, we achieve accurate identification and precise quantification of chemicals, with accuracies of 98.2 and 98.6%, respectively, significantly outperforming single P-DUS (63.2 and 95.1%) and SERS (80.3 and 86.5%). Additionally, we demonstrate the universality and expandability of SPEDS through other plasmonic nanostructures of various shapes and surface modifications. Using a CuS-coated Au nanoarray, we demonstrate multiple 8-h monitoring sessions of Hg2+ with good anti-interference and robust quantification, thereby highlighting the practical potential of SPEDS in real-world applications. These results position SPEDS as a cutting-edge and multifunctional chemical sensing platform, unlocking transformative possibilities for advancing environmental monitoring, industrial safety, and public health.