Essentials of Computational Chemistry

11.4 STRENGTHS AND WEAKNESSES OF CONTINUUM SOLVATION MODELS 413
One approach for reducing the errors associated with the prediction of pKa values is to
employ an isodesmic reaction. To illustrate with a specific example, it may be very hard to
correctly predict the free energy change for the aqueous reaction
+
NH3 F
NH 2
F
+
+
H
(11.26)
However, if the theoretical target is instead the free energy change for the isodesmic equation
+
NH3 F
+
NH 2
NH 2
F
+
+
NH3 (11.27)
one may well expect this to be computed far more accurately, since errors in levels of theory
should largely cancel from left to right. Provided experimental data are available for the
unsubstituted case
NH 2
+
+ H
+
NH3 (11.28)
then the free energy change for Eq. (11.26) may be estimated from the difference between
the computed value for Eq. (11.27) and the experimental value for Eq. (11.28). Chen and
MacKerell (2000) have provided a more detailed demonstration of the utility of this approach
for a series of substituted pyridines using a variety of different levels of theory for the gas
phase and computed solvation free energies.
An alternative approach for improving predicted pKa values has been suggested by Klicic
et al. (2002), who developed functional-group-specific linear regression corrections for pKa
values computed from a particular DFT SCRF PB formalism. Correction of the raw computed
pKas increases the model’s accuracy to about 0.5 pK units for those acidic functional groups
well represented in their parameterization set.
11.4.1.2 Redox potentials
Oxidation and reduction potentials in solution are also computed via reference to
particular thermodynamic cycles as illustrated in Figure 11.10. In this case, however, the