Application And Research Of Terahertz Spectroscopy In Field Of Food Additivity Safety

Illegal use or abuse of food additives can cause food safety issues.Terahertz spectroscopy was widely used in food additivity safety field with fast,accurate and non-destructive etc.advantages.In recent years,research on terahertz fingerprint spectra of illegal additives and food additives has been successively reported.The theoretical simulations were mostly performed using density functional theory(DFT).However,traditional algorithm lacked in-depth comparative analysis and improve according to the crystal structure characteristics.Moreover,previous studies usually performed qualitative or quantitative analysis on a single food additive.Considering that multiple food additives usually exist in practical food,distinguish and analysis of multiple food additive mixtures had important practical significance,the main contents and results were introduced as follows:(1)Based on B3 LYP and PBE functionals,the terahertz spectrum of the illegal additive melamine was studied in depth.The periodic crystal structure of melamine was simulated by CRYSTAL 14 software.In view of the fact that melamine is a hydrogen-bonded system,the London dispersion force cannot be ignored.The dispersion correction term was added to augment the PBE and B3 LYP functional and constructed the PBE-D and B3LYP-D models.Results showed that the dispersion correction models can obtain higher quality simulated spectra.The B3LYP-D model had higher accuracy with much lower simulation speed,while the PBE-D model provided slightly lower accuracy,but the time-consuming was only half of B3LYP-D model.(2)Three kinds of hydrogen-rich food additives,i.e.,benzoic acid,sorbic acid and xylitol were selected to explore the applicability of the above dispersion correction model in hydrogen-bonded systems.The B3 LYP and PBE functionals and two dispersion correction functionals(B3LYP-D and PBE-D)were used to calculate the three food additives.Results showed that the values of average absolute error based on dispersion correction models were decreased by 0.03?0.1 compared with primitive functionals,of which the B3LYP-D model has the highest accuracy.The universality of dispersion correction model for simulating hydrogen-bonded systems was verified to some extent.(3)Qualitative identification and quantitative analysis of food additives mixture.The sorbic acid,benzoic acid,xylitol and polyethylene were mixed with different proportion.An identification model based on chemometrics consisting of principal component analysis(PCA)and fuzzy pattern recognition was established to qualitative identify unknown food additives mixture.The proposed approach provided a new effective method for multi-component mixture identification.The partial least square(PLS),back propagation artificial neural network(BP-ANN),and least square support vector machine(LS-SVM)models were used for quantitative analysis.It was found that the LS-SVM model had the highest prediction accuracy,and the LSSVM model presented more obvious performance advantage with the complexity of system increased.Furthermore,an approach to reduce the sample preparation while ensuring predicted accuracy was proposed.