SERS技术结合PLS模型定量检测牛奶中二元污染物

Quantitative detection of multiple contaminants in milk by SERS combining partial least squares regression

为实现牛奶等复杂体系下多元污染物的快速定量检测,提出了表面增强拉曼散射(SERS)光谱技术结合偏最小二乘(PLS)回归模型的定量分析方案。SERS的高灵敏性和丰富的指纹信息有利于多元分析物的检测和识别,但多元分析物在金属纳米颗粒上的竞争吸附导致的光谱强度改变使准确定量成为困难。以存在竞争吸附条件下获得的多组混合光谱数据作为训练集建立PLS模型,极大地提高了真实体系下多元分析物浓度的预测准确度。基于上述方法,进行了牛奶中噻菌灵(TBZ)和环丙沙星(CIP)残留的检测,预测结果中TBZ和CIP浓度的相关系数(R^(2))分别为0.9722和0.9719,测试集均方误差(RMSEP)分别为4.0592 mg/L和0.4032 mg/L。实验结果表明,该方法在复杂体系多元检测应用中具有可行性。结合便携式拉曼光谱仪,该方案有望在食品安全等领域现场快速检测中得到实际应用。

A quantitative analysis scheme based on surface-enhanced Raman scattering(SERS)spectroscopy combined with partial least squares(PLS)regression is proposed in this paper,to realize the rapid quantitative detection of multiple pollutants in complex systems,such as milk.The high sensitivity and rich fingerprint information of SERS are beneficial to the detection and identification of various pollutants,but the changes of spectral intensity caused by the competitive absorption of multivariates on metal nanoparticles make the accurate quantification difficult.To solve this problem,a multivariate quantitative analysis model was established by using acquired mixed spectra of multivariatess under competitive absorptioncondition as the training set,which greatly improved the prediction accuracy of multivariate analyte concentration in real systems.Thiabendazole(TBZ)and ciprofloxacin(CIP)residues in milk were detected and quantified with the proposed method.The predicted correlation coefficients R^(2) of TBZ concentration and CIP concentration reached 0.9722 and 0.9719,respectively,and root mean square error of prediction(RMSEP)were 4.0592 mg/L and 0.4032 mg/L,respectively.It is demonstrated that the proposed method can implement fast quantification of multivariate pollutants in complex systems.Combined with portable Raman spectrometer,it is expected to be practically applied to the on-site rapid detection in food safety and other fields.