Solute Fe(II) - montmorillonite interactions are studied in anoxic conditions and at room temperature for reaction times from hour to week. Fe2+ is shown to be sorbed on cation exchange site with the same affinity than Ca2+. In chloride anionic medium, Fe(II) form ionic pairs - FeCl+ - which is sorbed with almost the same affinity than CaCl+ and MgCl+ are. The exchange thermodynamics constants derived from this study are used to simulate the change in the exchanger composition as clay river particles enter seawater. In high concentration chloride medium, as seawater, monovalent ions (Na+ and CaCl+, MgCl+ ionic pairs) are shown to be the major species of the exchanger. Fe2+ is sorbed specifically on the montmorillonite edge surfaces with a very high affinity. Simple complexation model are able to model the sorption data and show that the Fe2+ affinity for clay edge surfaces is ~1000 times higher than the Zn2+ one. Mössbauer experiments combined to sorption, titration and dissolution experiments show that the Fe2+ sorption is due to several different reactions: effective competitive sorption with replacement of previously sorbed or structural cations (Zn2+, Mg2+); cooperative sorption together with H4SiO4, in agreement with a possible surface precipitation of a Fe - Si phase; a sorption mechanism followed by an oxidation reaction, with a release of two H+ in solution per Fe(II) sorbed, and a product (Fe(III)) fitting better octahedral surface "sites". All these phenomena can not be taken into account in a classical surface complexation model. Hence, an innovative model is developed to model clay - solute interactions, based on a morphological and structural approach. Montmorillonite edge surface area was determined using two independent methods, AFM measurement and low-pressure gas adsorption, that give the same value for this area, i.e. 8.5 m2 g-1. The clay - solute interface was found to be constituted by a mix of, at least, 27 reactive sites. Their reactivity towards H+ was modeled with the MUSIC model. An excellent agreement is found between potentiometric measurement data and predicted curves. This model should be further developed to predict the clay surfaces - metallic cations interactions.
