Journal of Membrane Science 281 (2006) 139–148 Modelling of solute transport in non-aqueous nanofiltration J. Geens, K. Boussu, C. Vandecasteele, B. Van der Bruggen ∗ Laboratory for Applied Physical Chemistry and Environmental Technology, Department of Chemical Engineering, Katholieke Universiteit Leuven, W. de Croylaan, 46 B-3001 Belgium Received 15 November 2005; received in revised form 6 March 2006; accepted 15 March 2006 Available online 18 April 2006 Abstract This work presents a new approach for modelling of transport of small organic components in nanofiltration of organic solvents. The solute rejection in non-aqueous nanofiltration is affected by a number of membrane–solvent–component interactions, which must be incorporated in a generalised transport model. Since measuring or calculating membrane–solvent–component interactions is difficult, a simple empirical approach is adopted to model the data based on a pore-flow formalism. Previous work showed that solute transport in non-aqueous NF is mainly dominated by convection, but transport models for pore flow, available in the literature, only consider aqueous feed streams. In this article, solvent dependency was implemented in these models in terms of effective pore and solute diameters. This approach was applied on four literature models (the SHP-model, the Ferry model, the Verniory model and the lognormal model). Descriptive modelling of the experimental rejection data of different reference components in methanol, ethanol, acetone and ethyl acetate resulted in the identification of unknown membrane characteristics (pore size or pore size distribution), clearly confirming the solvent dependence of these parameters. This effect was observed for both polymeric and ceramic NF-membranes. The membrane parameters identified from the data fitting were used for predictive modelling of the rejection of raffinose in methanol and ethanol, with a hydrophilic and a hydrophobic NF membrane. A high correlation between predicted values and experimental data was found. © 2006 Elsevier B.V. All rights reserved. Keywords: Nanofiltration; Solute transport; Pore flow; Reflection curve; Predictive modelling 1. Introduction Due to the use of new materials with a high degree of solvent resistance and improved process controlling during membrane manufacturing, the performance of nanofiltration in organic sol- vents increased significantly. Over the last years, non-aqueous nanofiltration became an important alternative to traditional sep- aration units, such as distillation or evaporation, which are highly energy consuming. Different applications were reported in the literature, e.g. solvent recovery from lube oil filtrates [1], in the vegetable oil [2,3] and the pharmaceutical industry [4], for the separation and reuse of homogeneous catalysts [5,6] or for solvent exchange in the chemical industry [7]. Impor- tant economic and environmental benefits can be realised, and working conditions are often much more moderate, thus safer [8]. ∗ Corresponding author. Tel.: +32 16 32 23 40; fax: +32 16 32 29 91. E-mail address: bart.vanderbruggen@cit.kuleuven.be (B. Van der Bruggen). However, the scientific background on transport mechanisms in non-aqueous nanofiltration is still relatively limited. The per- formance of membranes was often characterised in aqueous media, but these data cannot be straightforwardly transferred to applications in organic solvents. The use of the molecular weight cut-of (MWCO: the molecular weight of a reference component that has a 90% rejection), specified in aqueous nanofiltration, makes no sense with respect to non-aqueous nanofiltration [9]. Different approaches for the modelling of solute transport in organic solvents have been presented, but these were often focused on dense materials [10]. Other attempts showed poor results, as the solvent effect on solute rejection could not accu- rately be incorporated in these models [11]. In this work, a new methodology is presented for the mod- elling of solute transport in non-aqueous nanofiltration. As pre- vious work indicated that solute transport is mainly dominated by convection [12], the new methodology is based on existing transport models for porous membranes. A first part of this work is focused on the descriptive modelling of experimental data, obtained with six reference 0376-7388/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2006.03.028