Résumé
Ion separation by liquid-liquid extraction (LLE) is a process based on the difference in affinity of a solute between two immiscible phases e.g. water/oil. In order to increase the affinity of the solute to be extracted, an amphiphilic transfer agent called ligand or extractant molecule is usually added in the organic phase, which is often the case for metal ion separation. However, when the complexation "ligand/metal ion" is not so strong and the solubilization of the complex in the oil phase depends on the extractant aggregation in this same phase, then the structural organization of these amphiphilic molecules at the water/oil interfaces can be determinant for the ion transfer kinetics. This is the case for the reprocessing of the nuclear waste and more specifically for actinide/lanthanide separation through the DIAMEX process. Thus, a detailed knowledge of the supramolecular structuring of these interfaces containing extractive molecules is essential for understanding and simulate the ion transfer phenomena. Unfortunately, experimental data to access structural information of these buried interfaces at the nanometer scale are almost inexistent. Recently, the combination experiments involving x-ray and neutron reflectivity has made the access to these information possible. The analysis of these measurements shows that trivalent cations can be repelled or attracted by an interface enriched with extractant depending on the nature of the ligand molecule. Since the observed experimental structure was the result of a complex processing of many reflectivity data; it was interesting to see whether molecular modelling (MD) could yield comparable results without preliminary hypothesis on the ligand molecule. MD simulations of real solutions were performed explicitly taking into account the polarization effects of all the atoms using the AMBER software to simulate a water/oil interface containing ligand molecules distribution at equilibrium between the interface and the organic volume. The comparison of measurement and MD (Fig.) results supports not only the analysis of data from large instruments, but also the molecular interaction parameters used for MD simulations.