Application of inverted multivariate calibrations to determination of the total content of phenols

Abstract

The total content ( c Σ ) of toxic phenols in water is usually determined using the procedures which include the introduction of a group reagent, measurement of the generalized signal ( A Σ ) at a selected wavelength, and assessment of c Σ in terms of C 6 H 5 OH. The use of diazotized sulfanilic acid (DSA) as a group reagent leads to the errors of c Σ determination which do not exceed 30% thus ensuring more adequate estimates of c Σ than measuring phenolic index. We suppose that further reduction of errors in group analysis would be possible with multiwavelength measurements and multivariate calibrations but these promising techniques have not been used yet for the determination of total phenol. To check up this assumption, model multicomponent colored solutions of the known composition were prepared and analyzed. These model mixtures contained simultaneously up to 5 different phenols with a total concentration ranged from 15 to 70 μmol/L. After converting all phenols to corresponding azo-dyes their generalized signals were measured at m wavelengths in the UV region of the spectrum 10 minutes after mixing the solutions. The results of group analysis were calculated with the multivariate calibrations; the recalculation of A Σ values to standard substance concentration we used only for comparison. The inverted multivariate calibrations were calculated with A Σ values of n model mixtures which formed a training set. For optimized conditions ( m = 7, n = 10) the systematic error of c Σ prediction is less than 13 % rel., that is half of errors for total index calculation. Therefore, inverted multivariate calibrations can be rather useful to control the total content of phenolics in natural and waste waters (instead of the total indices). However, systematic errors raised sharply when the analyzed sample contained some individual phenols which were absent in samples of the training set; in such cases the errors can increase up to 80 % rel. To maximize the correctness of corresponding techniques, it is desirable to elucidate beforehand the qualitative composition of phenol mixtures in water samples under study and take it into account when the multivariate calibration is used.