Study of the effects of defects in ultrafiltration membranes on the water flux and the molecular weight cut-off

The aim of this study was to evaluate how defects in a membrane surface can affect the molecular weight cut-off (MWCO) and the membrane permeability as they are traditionally defined. These two parameters are important criteria for the selection of ultrafiltration membranes with regard to efficiency and rejection of dissolved and particulate substances and especially disinfective retention. Tests were conducted with a flat-sheet membrane with an effective filtration area of 1.45×10!3 m!2. Dead-end filtration experiments were performed with various feed solutions: pure water to measure the hydraulic permeability, a mixture of dextrans to determine the apparent MWCO and protein solutions. This was done before and after the membrane integrity was altered by perforating its surface with a sharp tip.

The first results reported in this paper seem to show that intrinsic characteristics of the membrane — hydraulic permeability and apparent MWCO, as they are usually determined — are not sufficiently sensitive to the presence of defects in the surface, whose presence could be responsible for leakage of viruses or microorganisms if they were present in a plant in operation. Even if hydraulic permeability increases in presence of defect on membrane surface, the sensitivity of this parameter is not sufficient to reveal the presence of imperfections smaller than 50 µm in diameter. As for apparent MWCO values determined on perforated membranes, no 8.1 change was observed in presence of small defects (50 µm) even when this parameter decreased when big holes (150 µm) were created on the membrane surface. This last result, which was opposed to the evolution we should expect, could be due to an healing effect by dextran molecules. Lastly, protein retention measurements have shown the difficulty in using the characteristic retention curve obtained with a mixture of dextran molecules in predicting rejection of other isolated solutes. This result suggests that a characterisation method using monodisperse tracer molecules instead of polydisperse ones would be more adequate to predict rejection based on size exclusion.