Simultaneous Determination of Colorant Mixtures Used in Cosmetics by Partial Least-Squares Multivariate Calibration Spectrophotometry

Abstract

Quinoline Yellow (C.I. 47005), Sunset Yellow (C.I. 15958), Tartrazine (C.I. 19140) and Brilliant Blue FCF (C.I. 42090) are four synthetic colorant matters widely used as additives in cosmetic products. Quinoline Yellow (QY) E-104 is a colorant approved by the European Union for use in cosmetics, except in the area around the eyes.1 Sunset Yellow (SY) E-110 is an azo dye to which some people have an allergic reaction.2 Tartrazine (TT) E-102 is a synthetic product that can produce urticaria, asthma, a running nose and other discomfort in susceptible individuals.3 Finally, Brilliant Blue FCF (BB) E-133 is a synthetic dye under consideration by the E.U. for an E prefix. Due to such difficulties in using these additives, only small amounts can be used in the manufacturing of cosmetic products4 and analyses are needed. The analytical techniques that have been used for the analysis of these chemicals are mainly the following: differential pulse voltammetry with a solid electrode or with a dropping mercury electrodes-', liquid chromatography8, high-performance liquid chromatography9 and UV-spectrophotometry.10 11 The main obstacle in the spectrophotometric determination of mixtures of these colorants is the overlapping of their spectra. However, in recent years complex mixtures of chemicals with similar spectral characteristics have been resolved by applying multivariate analysis methods of partial least-squares (PLS), originally proposed by Wold.12 PLS methods have been applied to the determination of mixtures of flavour enhancers in foods13 or aromatic aldehydes in biological samples14 among other applications. Since the four colorants studied here show similar chemical structures, they present a high spectral overlapping, and their simultaneous determination is hard when conventional methods are used. Hence, we propose the utilization of the PLS statistical method for treating of the analytical signal and absorbance, and to resolve quaternary, ternary, binary mixtures and single determinations of the above-cited colorants. In general, a multivariate calibration model is constructed from instrumental response data collected for a set of multicomponent samples of known concentrations with respect to the analytes of interest. In these cases we used partial least-squares regression, PLS. This method is based on the concentration of the total information of a response matrix in the fewest number of new variables, called principal components or factors, which can describe the significant information contained in multivariate data, such as the spectra, kinetics and chromatographic peaks. Its main difference with other multivariate calibration methods, as a principalcomponent regression (PCR), is that PCR uses only information of the response matrix and PLS uses the response matrix and the concentration matrix for calibration in order to define which dominant factors in the response data are most relevant to the concentration of the components. The orthogonality of the principal components is lost in the PLS model. Two forms of the algorithm exist, namely, PLS-1 and PLS-2. The former calibrates each analytes individually, while the latter can model a number of components simultaneously. In this work, since PLS-1 provided the most accurate predictions, it was applied in the proposed method for determining these colorants in commercial cosmetic products as colognes, facial tonics, deodorants, after shave lotions, bath gels, bath salts and shampoos. In all cases the obtained results are satisfactory.