Machine learning-driven grating-like SERS Platform toward ultra-sensitive detection of forsythin

Abstract

Due to the complexity of existing substrate preparation processes and the poor reproducibility of substrates prepared from a single material, a highly reliable and low-cost Ag20/V–Ti20/DVD multi-stage composite SERS substrate was constructed in this study. The grating-like SERS substrate has an enhancement factor of 8.44×106 for R6G and a detection limit as low as 1×10−10 M. Results indicated that the highly regular grating-like SERS substrate exhibited excellent reproducibility with the relative standard deviation (RSD) values less than 10%. The three-dimensional finite-difference time-domain (3D-FDTD) method simulated the spatial distribution of the electric field intensity of the substrate, and a large number of “hot spot” regions were distributed in the Ag nanogap in the grating structure nanostructure. The optimized substrate was subsequently utilized for the discrimination of R6G with different concentrations using SERS and machine learning methods. These outstanding SERS properties of the Ag20/V–Ti20/DVD multi-stage SERS sensing platform suggested its potential application in the identification of pharmacodynamic substance in forsythin when the concentration as low as 1×10−7 g/l. At the same time, the highly ordered grating structure provided a new idea for the preparation of regular nanoarray substrates, and the combination of SERS and machine learning paved a new way toward the ultra-sensitive detection of other pharmacodynamic substances in traditional Chinese medicine (TCM).