Thermodynamic Sorption Models: Derivation of Recommended Data Sets and Application to Performance Assessment

This paper integrally addresses three topics: the general predictive capabilities of Surface Complexation Models (SCM) for radionuclide migration modeling, their current state of parametrization based on a respective database, and the application for performance assessment (PA).

The SCM concept is mineral-specific and can therefore also be used for additive models of more complex solid phases such as rocks or soils. It properly takes into account the physico-chemical phenomena governing the contamination source term development in time and space. One major aspect there is the substitution of conventional distribution coefficients (KD-values) for the empirical description of sorption processes, most often applied in risk assessment studies so far. Thus, the framework of a "Smart KD" is developed for complex scenarios with a detailed explanation of the underlying assumptions and theories. It helps to identify essential processes and the associated most critical parameters, easing further experimental refinement studies.

As to the knowledge of the authors, so far there is no digital thermodynamic database for surface complexation equilibria existent world-wide, despite the vast amount of available data. Therefore, the "Rossendorf Expert System for Surface and Sorption Thermodynamics" (RES³T) was developed [1]. Originally implemented as a stand-alone relational database under MS Access on a PC, now RES³T is also connected to the WWW and accessible through commonly used web browser interfaces. An integrated user interface allows to easily access selected mineral and sorption data, to convert parameter units, to extract internally consistent data sets for sorption modeling, and to export them into formats suitable for other modeling software. Data records comprise of mineral properties, specific surface area values, characteristics of surface binding sites and their protolysis, sorption ligand information, and surface complexation reactions. All relevant submodels are already implemented, extensions to other variants are straightforward. An extensive bibliography is attached, providing links not only to the above listed data items, but also to background information concerning surface complexation model theories, related software for data processing and modeling, sorption experiment techniques, and independent spectroscopic evidence of surface species. RES³T assists the identification of critical data gaps, the evaluation of existing parameter sets, consistency tests and the establishment of selected reference data sets.

High-quality experimemtal sorption data sets as provided by Phase II of the NEA Sorption Project for its fitting assessment efforts were used. The systems covered are Np(V) sorption onto hematite, U(VI) sorption onto quartz and Se sorption onto goethite. To keep the number of parameters at a minimum, the Diffuse Double Layer model was selected to account for electrostatics. All calculations were performed with the FITEQL code, version 3.2 [2]. Based on the information in the sorption database RES³T for the above minerals and chemically similar phases, first a set of relevant species was formed. Then respective surface complexation parameters were taken from RES³T: the binding site density for the minerals, the surface protolysis constants, and the brutto stability constants for all relevant surface complexes. To be able to compare and average thermodynamic constants originating from different sources, the normalization concept as introduced by Kulik [3] was applied. Lacking data was substituted by estimates exploiting chemical analogy. The only system-specific parameters directly going into the computations were the solid-liquid ratio and the specific surface area. The model prediction almost always represented the experimental values for the sorbed amount of Np, U and Se, expressed as conventional distribution coefficients KD as required by PA software, within one order of magnitude or better, provid...