Removal of organic foulants from feed waters by dynamic membranes

Seeking economic solutions for the pre-treatment of polluted waters is at the forefront of current reverse osmosis (RO) desalination research. The modern trend is to remove the polluting species by various UF/MF membrane pretreatment schemes. This paper describes progress in an ongoing project aiming to improve membrane pre-treatment processes by placing dynamic membranes in front of the UF/MF unit so as to relieve the load on the downstream processing. Dynamic membranes (DM), a class of membranes that has not yet found general industrial application, are formed in situ by colloid deposition on a porous support. They are an attractive method to relieve the load on the downstream pre-treatment membranes because, once fouled, they can be removed and reformed in place. This research has shown that DM formation conditions for organic foulant retention are not necessarily those suggested in the literature for other applications. The membranes tested were formed by dead end filtration of a hydrous zirconium oxide colloidal suspension on inexpensive non-woven flat sheet supports and post-treated by a poly(acrylic acid) solution. The quality of a DM membrane was characterized by measurements of permeate flux and retention of ovalbumin, used as a model contaminant. The properties of dynamic membranes are known to depend on a large number of variables. Results are described showing the effect of parameters found to exert a significant influence on membrane properties. They include the zirconium oxide colloid concentration, the tightness of the support fabric, the pH level of the colloidal suspension and the poly(acrylic acid) post-treatment. The best dynamic membrane achieved exhibited 85% ovalbumin retention at a concentration of 1000 ppm and over 95% retention at a concentration of 50 ppm. Permeation rates were in the range of UF fluxes (10–50 l/m2 h bar). The work carried out to date has yielded encouraging results that are being currently extended.

Dynamic membranes, with reasonably good separation characteristics and permeabilities similar to those of commercial UF membranes, can be constructed. Flat sheet membranes were formed in the dead-end configuration and operated at pressures of about 5 bar. The rate of hydrous zirconium oxide deposition obeyed conventional filtration laws. The colloid layer was modified with PAA. The modification was vital for producing a viable membrane; without PAA post-treatment almost no retention was observed. The porosity of the support fabric played an important role in membrane formation. The poreclogging stage of a tighter membrane was much shorter than for a looser fabric, leading to quicker membrane formation. More importantly, the membrane formed on the tighter fabric exhibited superior separation properties, affording protein rejection 40% higher than on the looser fabric, albeit at lower permeabilities. The membranes formed from particles in contact with an aqueous solution at pH 6 had higher retention rates those formed from more acidic solutions. Furthermore, the membranes formed at the higher pH also yielded higher permeabilities. The effect of increasing the model foulant concentration from 50 ppm to 1000 ppm was consistent with ultrafiltration membrane performance. At the higher protein concentrations the retention and permeability decreased. The permeate fluxes achieved were in the range of ultrafiltration fluxes (over 40 l/m2 h bar) after 4 h of continuous operation with a contaminant solution of 50 ppm protein. Ovalbumin retention in the various tests ranged from less than 40% for mediocre membranes and over 95% for the best membranes. Experiments that produced membranes that lacked integrity showed much higher nominal fluxes with almost no retention. Of the combinations tested, the best dynamic membranes in terms of retention were formed on the thicker Viledon® FO2430 support with a concentration of 0.252 mM ZrCl4 at pH 6. For a feed of 1000 ppm ovalbumin, retention of about 85% was achieved over a period of 4 h at a transmembrane pressure of 5 bar. The permeate turbidity throughout the experiment was about 1 NTU. Membrane fouling, however, continued throughout the experiment, reaching 10 l/m2 h bar after 4 h. At a lower feed concentration of 50 ppm, the same membrane afforded retention of about 95% with permeabilities of over 40 l/m2 h bar. The work carried out to date has yielded encouraging results that are being currently extended.