Study Of Transport Mechanism Of Nanofiltration In Organic Phase

Nanofiltration (NF) is a relatively new membrane technique located between ultrafiltration (UF) and reverse osmosis (RO), and its application has been gaining popularity in water treatment, food, pharmaceutical and chemical industries. Solvent resistant nanofiltration (SRNF) may be considered as an extension to the NF process applied in organic solvent system. The advantages of SRNF are numerous, for example, the process is free of phase transition and the use of additives. Thermal degradation and deactivation can be minimized during the separation process due to the mild operation temperature. Organic solvents can be recycled at low pressure so as to decrease their emission to the environment, and reduce the energy consumption as compared with alternative unit operations like distillation and crystallization. Despite these advantages, the running large-scale SRNF processes are quite limited. Compared to the traditional NF process, the SRNF is less predictable due to the complicated solvent-solute-membrane interactions and the limited knowledge on the membrane transport mechanism, that is, whether transport (both solvent and solute) through SRNF membrane in organic solvent is dominated by viscous flow or diffusion. Present transport models have various limitations and a generalized model is not available, which hindered the practical applications of organic solvent nanofiltration. Therefore, investigation into the transport mechanisms of nanofiltration in organic solvent system is a frontier research area in the world.Firstly, the permeation of eight organic solvents through hydrophobic SRNF membrane (STARMEMTM122 membrane) has been measured in a Sterlitech HP4750 dead-end filtration cell. Considering the dependence of solvent flux on both solvent properties (such as molecular weight, viscosity, mole volume, and dielectric constant) and membrane-solvent interaction (such as membrane-solvent solubility parameter difference and membrane-solvent surface tension difference), correlation analysis is performed between the flux of 23 solvents and various solvent property parameters and membrane-solvent interaction parameters. Higher correlation coefficient is found between solvent flux and four parameters (solvent viscosity, dielectric constant, membrane-solvent solubility parameter difference and membrane-solvent surface tension difference). For the SRNF membrane systems studied, therefore, the four parameters may be confirmed as the main factors influencing the membrane flux.Then, taking the effect of the four parameters on solvent flux into account, a new semi-empirical solvent transport model based on the solution-diffusion with imperfection model has been developed, and its universal applicability is demonstrated by describing solvent transport behaviour in a wide range of pure solvent and binary solvent mixtures. The predictive ability of the new model was compared to five existing models by using the experimental results of solvent flux combined with some reference data. For the whole solvents studied, the new model shows a high correlation and minor standard errors between experimental and predicted solvent permeation, implying that the model may well describe the transport behaviour of pure solvent and binary solvent mixtures through hydrophobic SRNF membranes. The results demonstrate the suitability of the new model for the different hydrophobic membrane systems, and show that the viscous-diffusive transport mechanism may well describe the solvent permeation through hydrophobic solvent resistant nanofiltration membrane.In the NF membrane filtration process, the size and shape of a solute molecule significantly affects its transport and retention characteristics. When estimating the retention of a solute with a predictive model (either solution-diffusion model or pore flow model), some parameters associated with the effective solute size must be known. Present size descriptors could be used to describe solute retention for some membranes, however, they would often not apply equally to a wide range of membranes. In this study, according to the molecular geometry optimization, the influence of solvent on the molecular geometry was considered explicitly for the determination of a new molecular geometric size parameter ("calculated mean size"). The suitability of "calculated mean size" for describing solute retention behaviour is evaluated based on large amount of reference data and statistically compared with those previously reported geometric size descriptors.Finally, retention data for a series of small and neutral solutes in methanol are measured through three different SRNF membranes. The neutral solutes include linear alkyl-acetates, branched alkyl-acetates, cyclohexyl-acetates and aromatic acetates with molecular weight ranging from 116 to 228 g mol-1. For those solutes with approximately the same MW but different molecular shape, all three membranes show preferential transport of straight-chained solutes over cyclic solutes, while the branched solutes revealed the lowest permeability. In order to correlate the retention with the molecule size and shape of the solutes, eight traditional molecular size descriptors (MW, the Stokes diameter, the equivalent molar diameter, the empirical effective diameter, radius of gyration, "calculated molecular diameter", "molecular width", and "molecular length"), were compared to "calculated mean size". A strong correlation between molecular size and solute retention across a range of membranes in aqueous and non-aqueous systems was only found with "calculated mean size". In addition, correlation and regression analyses for various NF membranes in aqueous and non-aqueous systems were performed. The results justified "calculated mean size" as the most characteristic size parameter that can suitably discriminate the solutes molecules, and as a universal model parameter of the solution-diffusion model for well describing solute retention behaviour in aqueous and non-aqueous systems.