Inverse multivariate graduations as a means of separate determination of similar analytes from the spectrum of a non-additive light absorption mixture

Multivariate graduations are used in spectrophotometric analysis for the determination of a number of analytes in multicomponent solutions. These graduations relate generalized signals measured at multiple wavelengths to analyte concentrations. The purpose of this work is to test the applicability of inverse multivariate graduations (OMG) for the separate determination of analytes of the same type when their light absorption is non-additive. The objects of analysis were model aqueous solutions simultaneously containing Cu(II), Co(II), Ni(II), Zn(II), Pb(II) and an excess of the photometric reagent 4-(2-pyridylazo) resorcinol. In such solutions, statistically significant deviations from light absorption additivity were observed, probably caused by a shift in the complexation equilibrium. The initial data for the construction of OMGs were spectra of model mixtures from the training sample. The number of analytical wavelengths (m) and the number of mixtures in the training sample (n) were varied during the experiment. Metal concentrations in the mixtures from the test sample were calculated separately by multiple linear regression using different spectral intervals and different OMGs. The best results were obtained at m = 16 and n = 30. The errors of determination of Co, Ni and Zn in single mixtures do not exceed 25 oz.% (modulo), and the generalized errors (RMSEP) were 10-15 oz.%. The errors of copper and lead determination were characterized by much higher values. The experiment showed that OMG can be used to separately determine components of mixtures with similar but not additive spectra. However, the volume of initial data in this case should be much larger than in the assessment of the total content of the same analytes, the accuracy of the results will be lower, and the possibility of correct determination of all analytes is not guaranteed.