Unlike the popular perception, it has been observed that pore size is not an important factor affecting permeate quality in MF and UF with possible exceptions with very tight membranes, e.g. molecular weight cut off (MWCO) < 100kDa. When MF and UF membranes with various pore sizes were compared in anaerobic digester broth filtration, there was no noticeable difference in permeate quality (Imasaka, 1989). Similar observations were made in alcohol distillery wastewater filtration (Yoon, 1994), virus removal from effluent (Hirani, 2010), etc.
It is because the cake layer formed in the beginning of the filtration cycle acts as a new membrane surface of which effective pore size is decided by the particles in feed water as shown in Fig. 1. Since particles in feed water see only the cake layer, initial membrane pores are rarely influential in permeate quality. The cake layer formed on membrane surface is called “dynamic membrane” that can be removed during membrane cleaning. It can typically reject most colloids such as single cell bacterium and virus, but it may not efficiently reject relatively small macromolecules since the particles in the cake layer form large enough spaces to pass them.
One implication of dynamic membrane formation is that when membrane pore size is measured using surrogate particles such as latex particles, concentration of latex solution must be kept low and liquid velocity on membrane surface must be kept high in order to prevent cake layer formation. If cake layer is formed during the membrane characterization, smaller particles than membrane pore can be rejected by the cake layer and the pore size can be underestimated.
Fig. 1. Dynamic membrane formation by particles existing in retentate.
Fig. 2 illustrates the effect of dynamic membrane formation on permeate quality. In the experiment, 0.45 micron flat sheet immersed membranes (Yuasa, Japan) were used to filter 1.0 g/L silica particles with nominal size of 0.05 micron. In the beginning, permeate turbidity was measured at around 100 NTU, but it gradually decreased to below 10 within 6 hours. Meanwhile, TMP increased from <5 kPa to >55 kPa at a constant flux of 25 LMH. These experimental results clearly indicate that the cake layer formed on membrane surface enables the membranes with 0.45 micron pores to reject 0.05 micron particles at the expense of higher filtration pressure.
Fig. 2. Time curve of permeate turbidity when 0.05 micron silica particles (1.0 g/L) are filtered by a flat sheet immersed membrane with 0.45 micron pores (Unpublished data, Yoon, 2002)
Dynamic membrane can also reduce membrane fouling by forming a cake layer with relatively high permeability in the beginning of a filtration cycle, which not only prevents small particles from passing through the membrane, but also makes membrane cleaning easier. When bovin serum albumin (BAS) was filtered in the presence of yeast cells by a MF membrane under crossflow condition, initially formed yeast cell layer not only reduced flux decline but also made the membrane cleaning easier (Güell, 1999). Similar flux enhancing effect was found when bovine serum albumin (BSA) was filtered by the membranes pre-coated with yeast cells (Arora, 1994).
When a particle containing water is filtered by ceramic membranes in crossflow condition, higher flux can be obtained at a lower cross flow velocity in some conditions. According to Imasaka (1993)‘s explanation, relatively large particles form a cake layer at a low flow condition and the cake layer reduces further membrane fouling by smaller particles. At a high flow condition, in contrast, relatively small particles are dominant in cake layer since larger particles are scoured. Thus the resistance of the cake layer can be higher at the high flow condition although the cake layer is thinner. Similar observation was made when a mixture of two different particles were filtered, where flux was higher with lower crossflow velocity (Foley, 1995). However, these observations can be considered exceptional, where initially formed cake layer interfere with further particle deposition. In general, higher crossflow velocities result in higher flux in most cases.
Under the condition where cake (or gel) layer does not exist, high particle concentration near membrane surface can cause a high particle loss through membrane due to the higher chances of particles passing through membrane. When a model bacteriophase Qβ, suspended in clean water was filtered by a 0.1 µm ceramic membrane, rejection efficiency increased with increasing crossflow velocity as shown in Fig. 3 (Urase, 1993). In was attributed to the low particle concentration in the concentration polarization layer at high crossflow velocity on membrane surface. It must be emphasized that this phenomenon is not commonly observed in practical situation, where some extent of cake layer always exits on membrane surface.
Fig. 3. Bacteriophase Qβ rejection as a function of crossflow velocity at constant flux levels (Urase, 1993).
© Seong Hoon Yoon