Charged SERS Probe Combined With Chemometric Method To Detect Pesticide Residues In Rice

Pesticides are substances that can control pests or regulate crop growth.Rice is one of the main food crops in my country.The use of pesticides in rice can protect rice production from pests and diseases,promote its growth,and increase the quantity and quality of rice.However,due to unreasonable application,the residues in the rice exceeded the standard,causing the quality of rice to decline and trade losses.The detection of pesticide residues among them is the key way to solve the above dilemma.Analysis methods such as chromatography,spectrophotometry,and immunoassay have the problems of complicated operation and long process.Recently,surface-enhanced Raman spectroscopy(SERS)has become an efficient method for pesticide residue detection due to fingerprint characteristics,high sensitivity,and simple pre-processing.Due to the charge attraction between the analyte and the nanoparticles,the charged SERS probe can further improve the specificity and responsiveness of detection.In this paper,a charged SERS probe and chemometric methods were used to study the rapid and sensitive detection of chlormequat chloride,thiram and cartap in rice,and further explored the liquid-liquid interface self-assembly to construct a SERS hotspot array.The main research work is as follows:(1)The charged SERS probe was studied to detect negatively charged chlormequat chloride and thiram residues in rice.First,lysine modified gold nanorods(GNRs)are positively charged,and the spectra of rice samples containing pesticide residues are collected.Then,the effects of full-band spectral data and feature extraction spectral data are discussed.Multivariate scattering correction(MSC),Savitzky-Golay(S-G)smoothing,standard normal variable(SNV)are used to remove or reduce background interference,and use Support Vector Machine(SVM),Decision Tree(DT),Partial Least Squares(PLS),Random Forest(RF),Linear Regression(LD)and K Nearest Neighbors(KNN),etc.,established concentration prediction regression models,further explored the prediction effect of Convolutional Neural Network(CNN).From the results,it can be seen that SNV combined with DT in the full-band spectrum of chlormequat chloride has a better prediction effect.The root mean square error of the prediction set is(RMSEP)=0.2628,and the coefficient of determination of the prediction set(R_p~2)=0.9996;The prediction effect of MSC combined with DT in the double full-band spectra of thiram is better,the result is RMSEP=0.1987,R_p~2=0.9998;the result of CNN's prediction set of chlormequat chloride is RMSEP=0.7334,R_p~2=0.9950,and the result of thiram is RMSEP=0.5514,R_p~2=0.9925.After PCA performs feature extraction on the full-band spectral data,the prediction effect of SNV combined with DT on chlormequat chloride is better than other methods,which is RMSEP=0.1683,R_p~2=0.9980;SNV combined with DT has a better prediction effect on thiram,which is RMSEP=0.2982,R_p~2=0.9980;Although the spectral data modeling results of chlormequat chloride and thiram after feature extraction are worse than the full-band spectral data modeling results,the processing speed is faster.(2)The negatively charged detection of positively charged pesticide residues in rice after modification of GNRs by alanine was studied.First,alanine-modified GNRs were used to collect the positively charged spectra of cartap in rice samples.Then,the influence of full-band spectral data and feature extraction spectral data is discussed.For the full-band spectrum,MSC,S-G smoothing and SNV are used to remove background interference.SVM,KNN,DT,PLS,LD and RF are used to build regression prediction models.The combination of S-G smoothing and DT in cartap has a good prediction effect,and the result is RMSEP=0.2411,R_p~2=0.9997;At the same time,the collected spectral data of cartap is used to establish a model for processing with CNN,and the result is RMSEP=0.7514,R_p~2=0.9869.Then,PCA is used for feature extraction,and the prediction performance of SNV combined with DT is better than other models.The result is RMSEP=0.6552,R_p~2=0.9979.(3)The self-assembly of the SERS hotspot array at the liquid-liquid interface was explored.First,synthesize GNRs by improving the seed growth method,and then replace the cetyltrimethylammonium bromide(CTAB)surfactant in GNRs with polyvinylpyrrolidone(PVP),which reduces the charge of GNRs,and also reduces repulsive force.Finally,add water,cyclohexane,the replaced substrate,the silicon wafer and the pesticide into the reaction tank to form a layered interface after self-assembly.The residues of parathion-methyl in rice are detected,which is a new way to detect pesticide residues in liquid environment.In view of the above research work,the charged SERS probe combined with chemometric methods can sensitively and rapidly detect pesticide residues in rice,and can efficiently analyze pesticide residues in complex backgrounds.At the same time,the liquid-liquid interface self-assembly method provides a way for SERS detection to integrate target sample separation and hotspot array assembly,and later combined with charged SERS probes,will construct a simple,specific and sensitive pesticide residue analysis system.