Although nanofiltration (NF) has attracted increasing attention over the recent years, its current industrial applications are still limited compared to the great potentialities of this separation process. The major reason is that the physical phenomena involved in the separation process by nanoporous membranes like those used in NF are still poorly understood. Indeed, most current NF models are essentially empirical tools the success of which can be attributed in large part to the fact that key properties of membranes (like their surface charge density) and confined fluids (like the dielectric constant of the pore-filling solution) are most often fitting parameters that have only limited correspondence to reality. An optimal development of the NF process therefore requires building relevant modeling tools that relate the properties of the membrane material and the fluid confined inside the membrane pores to the separation efficiency. A theoretical work that points out the links between the membrane structure and the transfer properties is then essential but the complexity of the transport phenomena inside nanometric paths makes this task cumbersome. This program aims at providing a better understanding of ion-transfer mechanisms through NF membranes. In order to catch all processes involved in a separation at the nanometer scale, a multi-physic modeling coupling the hydrodynamics of the system with fluid / material interactions (steric hindrance, electrostatic interactions, dielectric phenomena…) will be developed and combined with molecular dynamics simulations that will account for electronic polarizability of ions by means of a core-shell model. Indeed, molecular simulations of fluids confined in nanopores offer unique possibilities to connect some macroscopic properties to a microscopic description of the physical phenomena at play in nanoconfined phases. This original approach will be first validated with model silica nanopores and it will be further applied to polyamide membranes which are currently the main class of NF membranes available on the market. Experimental measurements of both the surface charge density (inferred from tangential streaming current experiments) and rejection rate of commercial polyamide membranes will be carried out with aqueous solutions of single salts and electrolyte mixtures so as to test the predictive abilities of the multi-scale approach developed within the scope of this program.