Machine learning-driven 3D plasmonic cellulose sensor for in situ rapid SERS detection of bisphenol compounds in water sample.

Rapid component separation and accurate identification of bisphenols compounds (BPs) in real water sample remain an attractive challenge due to the trace amounts and structural similarities of BPs, and complexity of real samples. Here, we designed and synthesized chemically modified cellulose p-toluenesulfonate (CTSA) to encapsulate octadecylamine-modified gold nanoparticles (Au-ODA), obtaining 3D plasmonic cellulose (Au@CTSA). Simultaneously, by virtue of the high surface area in the 3D network of CTSA and the solvent volatile deposition, BPs in water were in situ extracted and concentrated in Au@CTSA microspheres. Since the 3D network of Au@CTSA supports the formation of "hotspots", the number of "hotspots" available is greatly improved, enabling excellent SERS detection of BPs. Based on the collected SERS spectra, machine learning was utilized to analyze the overall profile of BPs, which eliminated the subjective judgment of the concentration by the Au@CTSA sensor using a single characteristic peak. In this way, the accuracy of identification of BPs was significantly improved. The machine learning-driven Au@CTSA sensor realized the detection of traces bisphenol A (BPA), bisphenol S (BPS), and bisphenol F (BPF) in water sample, pushing quantitative detection of different concentrations of BPs and contributing facile indicators for water quality monitoring.