Predictive modelling of nanofiltration: membrane specification and process optimisation

An important challenge for nanofiltration processes is the development of predictive models that convey a fundamental understanding and simple quantification of the governing phenomena in a way that has the potential for industrial application. The paper reviews one such approach, including: the formulation of a mathematically consistent description of rejection and flux at nanofiltration pores, the inclusion in calculations of pore size distributions with pore size dependent physical properties, the linearisation of these models to facilitate rapid calculations, and the use of such calculations to specify membranes for technically demanding separations. The present achievements and future challenges for predictive modelling of nanofiltration processes are assessed.

The paper has described the development of a continuum approach to NF processes that has the potential for industrial application. Within such a framework useful calculations can be made. Indeed, given the many known limitations of such models, they are surprisingly successful. This is in part due to the use of experimental parameters in the calculations which may implicitly compensate for the shortcomings. Model calculations are already valuable in predicting the separation of uncharged solutes, though it is essential that a pore size distribution is included if fractionation is to be quantitatively assessed. A significant weakness of the present approach for ions and other charged solutes is the requirement to experimentally determine the effective membrane charge density from separation data. Isotherms can be used to describe ion adsorption, though they are not always successful for multicomponent systems. Independent measurements quantifying membrane charge density or potential are difficult to apply successfully to process predictions. The inclusion of a dielectric exclusion term is essential for ions and other charged solutes, but further work is needed to determine the most appropriate means for its calculation. There is a need to learn more about the properties of solvents and solutes in confined volumes. Molecular dynamic calculations of properties such as dielectric constant and ion transport are useful. However, they need to be verified experimentally if they are to be used with confidence. It is likely that new theoretical and experimental approaches will need to be developed if substantial progress is to be made. In this context, membrane technologists need to be aware of advances being made in the biosciences, such as those describing solvent and ion transport in protein transport channels. Finally, there is the possibility that advances in understanding the physics of solutions and increases in computing power will allow the application of molecular level descriptions of NF — but this is likely to be many years in the future.