Analysis of interferents by means a D-optimal screening design and calibration using partial least squares regression in the spectrophotometric determination of Cr(VI)

Using a central composite design, the signal of the process for the spectrophotometric determination of hexavalent chromium (λ = 543 nm) is maximised and its variability minimised using as complexing agent 1,5-diphenylcarbazide in sufficiently acid medium. To analyse the interference of various analytes (Mo(VI), V(V), Fe(III) and Mn(VII)) on the Cr(VI) as a function of concentration of interferent, a factorial design was prepared at three levels of each (zero, medium and high concentration), which implies performing 81 determinations. However, a D-optimal design with just nine experiments is sufficiently good to estimate the model proposed.The interference of these metals makes it impossible to determine Cr(VI) when they are present in the sample. To avoid prior separation steps, a multivariate regression by partial least squares, PLS, is proposed to calibrate the Cr(VI) in the presence of these analytes varying the concentration of the Cr(VI) between 0.1 and 0.9 μg ml−1 and that of the interferents between 3 and 5 μg ml−1. The average errors obtained were 4.5% and 3.29% fitted and in prediction, respectively, with a standard error in prediction (RMSEP) of 0.016% presenting absence of both constant and proportional bias.The detection limit with the PLS regression in the presence of interferents is 0.1 μg ml−1 with a probability of false positive equal to 5% and less than 5% for false negative. The capability of detection is similar to that obtained with the univariate calibration (absorbance at 543 nm) in absence of interferents.With the PLS regression it is possible to discriminate 0.085 μg ml−1 of Cr(VI) in a sample with 0.5 μg ml−1 of Cr(VI) with probabilities of false compliance and false non-compliance equal to 0.05. For the univariate calibration without interferents, it was established at 0.0971 μg ml−1 of Cr(VI) for the same nominal concentration.In relation to interference of V(V), Fe(III) and Mn(VII), the PLS calibration could be an efficient alternative to the separation step for Cr(VI) spectrophotometric determination using 1,5-diphenylcarbazide.