This paper is an extension of a previous paper where the natural and purified clay in the homoionic Na form were physico-chemically characterized (doi:10.1016/j.clay.2008.04.014). In this study, the adsorption behavior of U (VI) on a purified Na-smectite suspension is studied using batch adsorption experiments and surface complexation modeling (double layer model). The sorption of uranium was investigated as a function of pH, uranium concentration, solid to liquid ratio, effect of natural organic matter (NOM) and NaNO3 background electrolyte concentration. Using the MINTEQA2 program, the speciation of uranium was calculated as a function of pH and uranium concentration. Model predicted U (VI) aqueous speciation suggests that important aqueous species in the [U (VI)] = 1 mg/L and pH range 3-7 including UO22+, UO2OH+, and (UO2)(3)(OH)(5+). The concentration of UO22+ decreased and that of (UO2)(3)(OH)(5+) increased with increasing pH. The potentiometric titration values and uptake of uranium in the sodium smectite suspension were simulated by FITEQL 4.0 program using a two sites model, which is composed of silicate and aluminum reaction sites. We compare the acidity constants values obtained by potentiometric titration from the purified sodium smectite with those obtained from single oxides (quartz and a-alumina), taking into account the surface heterogeneity and the complex nature of natural colloids. We investigate the uranium sorption onto purified Na-smectite assuming low, intermediate and high edge site surfaces which are estimated from specific surface area percentage. The sorption data is interpreted and modeled as a function of edge site surfaces. A relationship between uranium sorption and total site concentration was confirmed and explained through variation in estimated edge site surface value. The modeling study shows that, the convergence during DLM modeling is related to the best estimation of the edge site surface from the N-2-BET specific surface area, SSA(BET) (thus, total edge site concentrations). The specific surface area should be at least 80-100 m(2)/g for smectite clays in order to reach convergence during the modeling. The range of 10-20% SSA(BET) was used to estimate the values of edge site surfaces that led to the convergence during modeling. An agreement between the experimental data and model predictions is found reasonable when 15% SSA(BET) was used as edge site surface. However, the predicted U (VI) adsorption underestimated and overestimated the experimental observations at the 10 and 20% of the measured SSA(BET), respectively. The dependence of uranium sorption modeling results on specific surface area and edge site surface is useful to describe and predict U (VI) retardation as a function of chemical conditions in the field-scale reactive transport simulations. Therefore this approach can be used in the environmental quality assessment. (C) 2009 Elsevier B.V. All rights reserved.