Ag@CDS SERS substrate coupled with lineshape correction algorithm and BP neural network to detect thiram in beverages.

Surface enhanced Raman scattering (SERS) has been proved an effective analytical technique due to its high sensitivity, however, how to identify and extract useful information from raw SERS spectra is still a problem that needs to be resolved. In this work, a composite SERS substrate was prepared by encapsulating Ag nanoparticles within dialdehyde starch (Ag@CDS) to obtain dense "hot spot", and then a novel spectral preprocessing algorithm namely lineshape correction algorithm (LCA) was developed to separate the characteristic peaks of analytes from the original SERS spectra. Based on Ag@CDS and LCA, thiram residues in different beverages were quantitatively detected using back propagation (BP) neural network regression model. It was found that LCA provided an easy-to-use method for improving prediction ability of BP model. The R p 2 of BP model was improved from 0.2384, 0.3647 and 0.5581 to 0.9327, 0.9127 and 0.9251 for the quantitative detection of thiram residue in apple juice, grape juice and milk, respectively, while LCA was used for SERS spectra preprocessing. The optimal model can accurately detect thiram residue with a low limit of detection at 1.0 × 10−7 M, which is far below the maximum residue limit of thiram (2.9 × 10−5 M) regulated by the US Environmental Protection Agency. This study demonstrated that the proposed LCA can be used as a simple and valid spectra-preprocessing method in SERS quantitative detection. This paper aims to propose a rapid chemometrics algorithm, namely lineshape correction algorithm, coupled with a composite SERS substrate to detect the thiram residues in apple juice, grape juice, and milk. [Display omitted] • A dialdehyde starch-base SERS substrate (Ag@CDS) was synthesized. • Proposed a new data preprocessing algorithm named lineshape correction algorithm (LCA). • The LCA model use Ag@CDS to quantitative detect the residue of thiram in apple juice, grape juice and milk. [ABSTRACT FROM AUTHOR]