Study On The Key Technologies Of Terahertz Spectroscopy In The Detection Of Pesticide Residues In Agricultural Products

In recent years,the quality and safety problems of agricultural products caused by pesticide residues are severe.Therefore,it is of great significance to study the rapid and accurate detection technology of pesticide residues to prevent pesticide abuse.As one of the major emerging science and technology in the 21st century,terahertz(THz)has many unique advantages,such as fingerprint,penetration and coherence,which makes it have considerable potential for development and application in the field of pesticide residue detection.In this paper,the key technologies for fingerprint peak analysis of 18 kinds of pesticides,quantitative detection of single-component residue,qualitative identification of mulit-component residues,as well as quantitative and qualitative detection of trace agricultural residues,were studied by using terahertz spectroscopy and imaging.It is of great significance to ensure the safety of agricultural products and human health.The main research contents and results are as follows:(1)Molecular dynamics simulation and accurate THz fingerprint peaks analysis of pesticides were realized based on density functional theory(DFT)and THz spectral pre-processing algorithm.a)The wavelet threshold de-noising and spectral baseline correction methods were used to optimized the absorption spectra(0.1?3.5 THz)of pesticides.b)The theoretical spectra of pesticide molecules were calculated by DFT simulation.c)The theoretical analysis of peak vibration modes of pesticide was completed by matching the THz experimental spectra with the DFT theoretical spectra.The characterized fingerprint peaks of pesticides include:chlorpyrifos(1.47,1.93 and 2.73 THz),fipronil(0.76,1.23 and 2.31 THz),carbendazim(2.72 and3.06 THz),dimethoate(1.05,1.89 and 2.92 THz),methomyl(1.01,1.65,1.91,2.72 and 3.20 THz),thidiazuron(0.99,1.57,2.17 and 2.66 THz),deltamethrin(0.90,1.49 and 2.32THz),fenvalerate(1.13,1.43,1.61,1.98 and 2.58 Hz),Beta cypermethrin(1.27,1.84,2.12 and 2.92 THz),6-benzylaminopurine(2.08 and 3.00 THz),paclobutrazol(0.71,1.30,1.88 and 2.67THz),and maleic hydrazide(2.34 THz).(2)Detection accuracy of 2,4-dichlorophenoxyacetic acid(2,4-D)pesticide in food matrices was improved by using spectral correction algorithms for eliminating the baseline drift of THz absorption spectra.a)The THz fingerprint peaks of 2,4-D were simulated and analyzed by using B3LYP/6-31G(d,p)basis set in DFT(1.35,1.60,2.37 and 3.00 THz).b)The effect of THz scattering baseline on the detection ability of 2,4-D in food matrices(Zizania latifolia,rice and corn)was analyzed.c)The asymmetric least square smoothing method(As LS),adaptive reweighting penalty least square(Air PLS),background correction method(Backcor)and sparse baseline estimation denoising(BEADS)were used for THz spectral baseline correction.d)The quantitative regression curve of 2,4-D was established based on the intensity of the absorption peak at 1.35 THz.Results showed that the detection limits of 2,4-D in Zizania latifolia,rice and corn were improved from 7%,5%and 7%to 1%,respectively.The quantitative detection results were R~2?0.9706(in Zizania latifolia),R~2?0.9671(in rice),and R~2?0.9277(in corn).e)The validity and reliability of THz absorption spectral baseline correction in improving the detection accuracy of 2,4-D residues in agricultural products were verified according to the results of recovery rate and prediction error.(3)High precision identification of trace multi-component pesticide residues on the surface of fresh leaves,as well as the image visualization of pesticide species and distribution were realized by strategy of THz imaging and deep learning.a)The fingerprint peaks of benomyl(0.70,1.07 and 2.20THz),carbendazim(1.16,1.35 and 2.32THz)and thiabendazole(0.92,1.24,1.24,1.96,1.95 and 2.58THz)were analyzed by the B3LYP/6-311G basis set in DFT.b)Different types of pesticide solutions(concentration of 10 mg/L)were dripped onto the leaf surface of T.sinensis to obtain their THz image and extract the spectra in range of 0.2?2.2 THz from the images.c)The fuzzy mapping cluster analysis was carried out to investigate the separability of pesticide residue types.d)The pesticide residue types on T.sinensis leaves were identified qualitatively by using deep convolution neural network(DCNN)and four back propagation neural network(BPNN)models(based on Train CGB,Train CGP,Train CGF and Train RP).Results showed that DCNN obtained the optimal training and prediction recognition accuracy(97.27%and 96.74%,respectively).e)The image visualization of pesticide types and their distribution on leaves was realized by the prediction class labels of DCNN model.(4)Signal enhancement sensing and quantitative and qualitative detection of trace pesticides in pepper extract were realized by using a metamaterial absorber with structure of split-ring resonator.a)The fingerprint peaks of indole-3-acetic acid(IAA,2.50THz)and tricyclozole(0.85,1.11 and 2.17THz)pesticides were analyzed by B3LYP/6-311G and B3LYP/6-31G+dp in DFT,respectively.b)According to the dual band perfect absorption of absorber under TM polarization(the absorptivity reached 90.05%at 0.918 THz and 94.68%at 1.575 THz),the standard curves for quantitative detection of trace pesticides were established based on the high sensitivity response of peak amplitude and frequency shift of metamaterial absorber to pesticide concentrations.Results showed that the sensing response of the absorption peak amplitude at1.575 THz was better.The quantitative results for IAA and tricyclozole were R~2=0.9544 and R~2=0.7837 respectively,and the detection limits were reached 10 ng/L.c)The qualitatively identify IAA and tricyclozole based on the absorber was realized by using partial least squares linear discriminant analysis(PLS-LDA)algorithm.The above research results proved the great advantages of terahertz spectroscopy in the rapid and accurate detection of pesticide residues in agricultural products.The key technologies discussed in this study,including the terahertz fingerprints analysis of pesticides,the improvement of quantitative detection accuracy of single component residue,the qualitative identification and visualization of trace multi-component residues,and the quantitative and qualitative sensing enhancement detection of trace residues,which provide a complete theoretical basis and guidance for food safety detection of agricultural products based on terahertz technology.