Synthetic food dyes are usually added to processed foods to give and restore their desired esthetic quality. This aspect and the toxicological evidence justify quality control and the development of methodologies to quantify these food additives. In this work, the mathematical methodologies such as the additivity principle, derivative spectrophotometry and multivariate technique (Partial Least Square Regression - PLSR) were studied in the simultaneous determination of two food colorants: Tartrazine (TY) and Sunset Yellow (SY), extracted from natural wool. These methodologies were evaluated and compared according to their prevision capacities. The best PLSR model (spectral range of 305 to 645 nm with data transformation by first derivative and two principal components) presented the lower Root Mean Square Error Prediction value, RMSEP (TY = 0.191 and SY = 0.102). The concentration range was from 1.0 to 16.0 mg L-1 for SY and 2.0 to 22.0 mg L-1 for TY. The validation set was used and showed an average error of 1.97% for SY and 1.39% for TY. The application of this methodology in real samples showed concentration values in agreement with the limits established by the Brazilian law and regulations
PLS; first derivative; vierordt's method; synthetic colorants
