Bond-Valence methods for pKa prediction:: critical reanalysis and a new approach

Bond-valence methods for the prediction of (hydr)oxide solution monomer and surface functional group acidity constants are examined in light of molecular structures calculated using ab initio methods. A new method is presented that is based on these calculated structures, and it is shown that previously published methods have neglected one or more of four essential features of a generalized model. First, if the apparent pK(a) values of solution monomers are to be used to predict intrinsic pK(a) values of surface functional groups, similar electrostatic corrections must be applied in both cases. In Surface complexation models, electrostatic corrections are applied by representing a charged surface as a uniform plane of charge density, and an analogous correction can be made to Solution monomers by treating them as charged spheres. Second, it must be remembered that real surfaces and real monomers are not homogeneous planes or spheres. Rather, charge density is distributed rather unevenly, and a further electrostatic correction (which is often quite large) must be made to account for the proximity of electron density to the point of proton attachment. Third, the unsaturated valence of oxygen atoms in oxyacids, hexaquo cations, and oxide surfaces is strongly correlated with acidity after electrostatic corrections are made. However, calculation Of unsaturated valence for oxyacids and oxide surfaces must be based on realistic Me-O bond lengths (taking into account bond relaxation), which can be obtained from ab initio structure optimizations. Finally, unsaturated valence must be divided between possible bonds (four for oxygen atoms) to reflect the fact that O-H bonds are localized to particular regions of the O atoms. Empirical models that take all these factors into account are presented for oxyacids and hexaquo cations. These models are applied to the gibbsite (100), (010), (001), and cristobalite (100) surfaces, and it is demonstrated that the model for oxyacids predicts reasonable intrinsic pK(a) values for oxide surfaces. However. the prediction of surface pK(a) values is complex, because the protonation state of one functional group affects the pK(a) values of neighboring groups. Therefore, calculations of larger periodic systems, progressively protonated and reoptimized, are needed. Copyright (C) 2004 Elsevier Ltd.