Chemometric Analysis of Substituent Effects. I. Substituent Effects on the Dissociation of Monosubstituted Benzoic Acids

Abstract

Forty-six representative sets of data relating to the dissociation of monosubstituted benzoic acids in various solvents were extracted from the literature. The set of substituents included 25 common substituents in the meta position and the same number of substituents in the para position. Hydrogen served as the reference standard. The sets were subjected to regression analysis using conventional empirical models. The Hammett model was found to be valid within the limits of experimental error. The Taft model with the σR0, σR+ and σR- parameters is the best model to account for the substituent effects from the meta or para position solely. The inductive and mesomeric effects of the substituent are also best separated on this parametric scale. By applying the method of conjugated deviations (analysis of latent variables), a single latent variable was found to be sufficient to describe the data variability in all the three data sets analyzed (meta + para in the Hammett model and sets of meta and para substituted derivatives separately). The relationship between the first latent variables from the meta and para positions is isoparametric, the substituents lie on three straight lines intersecting in one point. The first straight line corresponds to substituents with the I effect (CH3, C2H5, tert-C4H9, C6H5, SO2NH2, CN, NO2 and hydrogen as the reference standard), the second straight line corresponds to substituents with I and +M effects (NH2, N(CH3)2, NHCOCH3, CH3O, SH, F, Cl, Br), and the third straight line corresponds to substituents with I and -M effects (CHO, CH3CO, COOR, SO2CH3, CF3). The +M mesomeric effect is twice as strong as the -M effect. These facts were used to propose a new empirical model for the description of substituent effects by means of one substituent constant and one (meta) or two (para) reaction constants. The PLS method revealed that the additional effects contribute about 8% to the data variability in the interpretation of the para substitution through meta substitution.