Statistical analysis employing multiparametric regression analysis was used to correlate the pKa values of 16 substituted Schiff bases (imines) derived from heterogeneous carbonyl compounds with various descriptors as parameters. These parameters are based on quantum mechanical methods. Four sets of parameters are used for these treatments. Two of them were derived depending on the semiempirical calculation represented by Austin method1 (AM1) and parameterized method3 (PM3) ..
The other two sets were chosen as Ab initio models, expressed by Hartree fock (HF) and MP2 calculation models performed at 6.311G (d,p) level of theory. The descriptors employed to test their impact on the variation of the experimental pKa(s) are of two kinds, electronic and geometrical including atomic charges (Muliken and lowding), length of the imine (C=N) and connected bonds. Other molecular characteristic are also used such as energies of the HOMO and LUMO orbitals, total energies (TE), kinetic energy (KE), torsion, Vender Waals interactions of the 1,4 type (VDW 1,4) and non 1,4 type (Non 1,4 VDW).
The correlation between the pKa values of the studied compounds and the selected parameters are investigated. Depending on the obtained results, several sets of parameters were constructed. The best set is indicated by the values of correlation coefficient (R‌‌‌‌‌2) and standard error (SE). Comparison are carried out between the experimental and calculated pKa obtained from AM1, PM3, MP2, and HF methods. Models of HF showed better consistency among the experimental and the calculated pKa(s) of the studied compounds.
Hypothetical compounds of similar structures but with different substituents not included in the regression analysis are used to test the success of the derived equations. Good results are obtained.