Effects of data scatter and inconsistency in sorption modeling

Surface Complexation Models (SCM) can describe sorption processes at the mineral-fluid interface on a scientific, quasi-thermodynamic basis. The effect of parameter uncertainty in blind predictions is illustrated by the example of Np(V) sorption onto hematite.
The Diffuse Double Layer (DDL) model was chosen as SCM submodel to keep the number of parameters at a minimum. From the surface species reported so far, =FeOH₂ ⁺, =FeO^-, =FeONpO₂, =FeONpO₂(HCO₃)₂ ^2-, =FeOHCO₂, and =FeOCO₂ ^- were selected with their normalized DDL reaction constants log K_n, taken from the RES³T database (Brendler et al., 2003).
The modeling was performed with the FITEQL code, version 3.2. The predictions were evaluated by comparison with the experimental results by Kohler et al. (1999), comprising of 52 data points for the Np(V) sorption at varied pH, ionic strength, solid/liquid ratio, Np content, and air P_CO2. When focusing on the conventional distribution coefficient K_D, the maximum deviation for all data subsets is less than one order of magnitude, a spreading considered to be reasonable in performance assessment applications.
Because the protolysis constants can not be varied independently, the point of zero charge (PZC) and delta pK were calculated based on experimental pK₁ and pK₂. Consequential we generated 20 independent pK data sets with the aid of Gaussian-distributed random numbers. None of these parameter sets did deliver unacceptable blind predictions for the distribution coefficients.
The formally large scatter of pK values as extracted from literature is actually not critical. Even the use of inconsistent values from other electrostatic models will yield similar results. For well-defined mineral systems, provided a suitable database is accessible, the SCM approach seems to be promising.
(References: Brendler V. et al. (2003), J. Cont. Hydrol., 61, 281-291.; Kohler M. et al. (1999), Radiochim. Acta, 85, 33-48.)