Simultaneous UV-Vis spectrophotometric quantification of Auramine O and Acridine orange dyes by partial least squares

One of the main difficulties in quantification of dyes in industrial wastewaters is the fact that dyes are usually in complex mixtures rather than being pure. Auramine O (AU) and Acridine orange (AO) in aqueous solutions were selected as two model dyes whose UV-Vis absorption spectra highly overlap each other. The spectral overlap between AU and AO dyes is extreme and this will retard their direct quantification by simple spectrophotometry without separation. Therefore, univaraite calibration is not applicable for the current sorption system and an analysis must be performed by chromatographic procedures or multivariate calibration methods in order to overcome such degree of overlapping [1,2]. Partial least squares (PLS-1) is factor-based chemometric method which can analyze highly collinear noisy data [3]. We developed one rapid and powerful method for spectral resolution of a highly overlapping binary dye system in the presence of interferences. The calibration samples were built by random number generation method with dye concentrations ranging from 3–32 mg/L. Once the calibration/training has been done, the final PLS-1 model may be further used to predict the concentrations of new samples. The spectra of an independent test dataset consisting of 10 samples were applied to the PLS-1 model. High accuracies in prediction of AU and AO dyes were obtained and all predicted values being close to the reference ones. A number of important statistical parameters such as correlation coefficient (R2 pred), root mean square error of prediction (RMSEP) and relative error of prediction (REP) were calculated. In the case of a river water sample, taking into account the complexity of the matrices available in natural waters, the prediction power of PLS-1 model for these dyes can be considered acceptable with recoveries between 92-98%.