Abstract :
A critical comparison of the various PCA methods on the absorbance matrix data concerning the complexation equilibria between SNAZOXS and Cd^2+, Co^2+, Cu^2+, Ni2+, Pb^2+ and Zn^2+ or Naphtylazoxine 6S and Cd2+, Cu^2+, Ni^2+ and Zn2+ at 25 °C is performed. The number of complex species in a complexforming equilibria mixture is the first important step for further qualitative and quantitative analysis in all forms of spectral data treatment. Therefore, the accuracy of the nine selected index functions for the prediction of the number of light-absorbing components that contribute to a set of spectra is critically tested using the principal component PCA algorithm INDICES in S-Plus software. Four precise methods based upon a knowledge of the experimental error of the absorbance data and five approximate methods requiring no such knowledge are discussed. Precise methods always predict the correct number of components even a presence of the minor species in mixture. Due to the large variations in the index values and even at logarithmic scale they do not reach an obvious point where the slope changes. An improved identification with the second or third derivative and derivative ratio function for some indices is preferred. Behind the number of various complexes formed the stability constants of species ML, ML2, (and ML3, respectively) type log (beta)11, log (beta)12, (and log (beta)13, respectively) for the system of SNAZOXS (ligand L) with six metals (the standard deviation s(log (beta)pq) of the last valid digits are in brackets) Cd^2+ (4.50(3), 8.36(7)), Co^2+ (5.75(6), 9.79(9), 13.05(2)), Cu^2+ (6.69(6), 11.40(7)), Ni^2+ (6.44(8), 10.91(11), 15.07(10)), Pb^2+ (5.63(5), 9.97(9)) and Zn^2+ (5.11(3), 8.84(5)) and for system of Naphtylazoxine 6S with Cd^2+ (6.08(4), 11.44(7), 16.06(11)), Cu^2+ (7.80(8), 13.41(14)), Ni^2+ (6.35(12), 11.43(19), 16.68(24)) and Zn^2+ (7.01(8), 12.65(15)) at 25 °C are estimated with SQUAD(84) nonlinear regression of the moleratio spectrophotometric data. The proposed strategy of an efficient experimentation in a stability constants determination, followed by a computational strategy, is presented with goodness-of-fit tests and various regression diagnostics able to prove the reliability of the chemical model proposed.
