Ultrasensitive detection and distinction of pollutants based on SERS assisted by machine learning algorithms

SERS as a promising sensing technique still faces challenges in the precise identification of trace-amount mol-ecules due to the limitation of sensitivity and cleanliness of SERS substrate. Here, we report a precise and ul-trasensitive identification of multiple pollutants via 3D clean cascade-enhanced nanosensor assisted by machine learning algorithms. This SERS substrate could achieve cascading electromagnetic energy with a remarkable enhancement factor as high as 8.35 x 109, which is attributed to the combination of micro-level polystyrene sphere (PS) porous array and nano-level Au-Ag clusters of this substrate. Benefitting from high cleanliness and ultra-sensitivity, multiple hazardous pollutants with similar geometry and Raman peaks at ultra-low concen-tration were successfully distinguished assisted by principal component analysis (PCA). As a result, this efficient and clean SERS substrate together with artificial intelligence could promote the application of SERS technology in the accurate identification of trace contaminants. Data Availability: Data will be made available on request.