Degree Name

Doctor of Philosophy


School of Civil, Mining, and Environmental Engineering


Forward osmosis (FO) can substantially advance water and wastewater treatment, particularly in battling against the wide occurrence of emerging trace organic contaminants (TrOCs). This thesis aims to elucidate four key effects on TrOC rejection by FO, namely, membrane properties and draw solution, key operating conditions, membrane fouling, and development of an FO – membrane distillation (MD) hybrid system.

Two types of FO membranes were employed, an asymmetric cellulose triacetate (CTA) and a thin-film composite (TFC) polyamide FO membrane, for the rejection of a wide range of TrOCs. A number of principal membrane parameters were characterised to facilitate the understanding of TrOC transport behaviour in FO, including effective average pore radius (rp), selective barrier thickness over porosity parameter (l/ε), surface charge, support layer structural parameter (S), pure water permeability coefficient (A) and salt (NaCl) permeability coefficient (B).

Membrane properties governed TrOC rejection in FO. The TFC membrane exhibited higher rejection of neutral TrOCs with low molecular weight than the CTA membrane, although the estimated pore size of the TFC membrane (0.42 nm) was slightly larger than that of the CTA membrane (0.37 nm). This higher rejection of neutral TrOCs by the TFC membrane was attributed to its active layer properties, namely a more effective active layer structure that was indicated by a larger l/ε parameter, and pore hydration induced by the negative surface charge.

Reverse draw solute diffusion, which is a unique mass transfer phenomenon in FO, was found to retard forward feed solute diffusion. This “retarded forward diffusion” phenomenon was further elucidated by examining rejection behaviour of three hydrophobic TrOCs, bisphenol A, triclosan and diclofenac, using the CTA membrane in FO and reverse osmosis. The reverse NaCl flux hindered the pore diffusion and subsequent adsorption of the TrOCs within the membrane.

Key operating parameters, including feed pH, membrane orientation, feed and draw solution temperature, was found to exert significant impact on TrOC rejection by FO. Neutral carbamazepine rejection was generally pH independent in both membrane orientations. Carbamazepine rejection in pressure retarded osmosis mode was lower than that in FO mode due to the higher concentration gradient caused by concentrative internal concentration polarization in the porous supporting layer. Conversely, sulfamethoxazole rejection was significantly affected by the feed solution pH in both membrane orientations. Variation in the rejection of sulfamethoxazole could be attributed to the electrostatic repulsion between the negatively charged FO membrane surface and varying effective charge of the sulfamethoxazole molecule. Rejection of charged TrOCs was higher than that of neutral TrOCs and was insensitive to temperature variation. On the other hand, rejection of neutral TrOCs decreased significantly when the feed and draw solution temperatures were 40 and 20 °C, respectively, due to the increase in their diffusivity at an elevated temperature. By contrast, rejection of neutral TrOCs increased when the feed and draw solution temperatures were 20 and 40 °C, respectively. The reverse salt (NaCl) flux increased due to an increase in the draw solute diffusivity.

Membrane fouling affected membrane performance and subsequent TrOC rejection. Deposition of humic acid onto the membrane surface was promoted by the complexation with calcium ions in the feed solution and the increase in ionic strength at the membrane surface due to the reverse transport of NaCl draw solute. As the deposition of humic acid increased, the permeation of carbamazepine and sulfamethoxazole decreased, which correlated well with the decrease in the membrane salt (NaCl) permeability coefficient. It was hypothesized that the hydrated humic acid fouling layer hindered solute diffusion through the membrane pore and enhanced solute rejection by steric hindrance, but not the permeation of water molecules. The membrane water and salt (NaCl) permeability coefficients were fully restored by physical cleaning of the membrane, suggesting that humic acid did not penetrate into the membrane pores. Membrane fouling was also simulated using either humic acid or colloidal particles as model foulants at different initial permeate water fluxes. Water flux decline was insignificant at an initial permeate flux of 9 L/m2h and the fouling layer was fluid-like, spare, and loose. By contrast, the water flux decline was substantial at an initial permeate flux of 20 L/m2h, resulting in the formation of a compact and cohesive fouling layer. Water flux recovery after physical cleaning for both humic acid and colloidal particle fouled membranes was consistently higher at an initial permeate flux of 9 L/m2h compared to 20 L/m2h. This markedly different fouling behavior at low and high initial permeate fluxes suggests that the fouling layer structure varied from a fluid-like loose layer at low initial permeate flux to a more cohesive and compact layer at high initial permeate flux. We surmise that the fluid-like loose layer formed at low initial water flux contributed to pore blockage and thus enhanced steric hindrance, thereby leading to an increase in TrOC rejection. By contrast, the cohesive and compact fouling layer formed at high initial water flux exacerbated cake-enhanced concentration polarization and resulted in a decrease in TrOC rejection.

Major outcome from the study was the demonstration of the robustness and treatment capacity of an FO – MD hybrid system for small-scale decentralized sewer mining. A stable water flux was realized using a laboratory-scale FO – MD hybrid system operating continuously with raw sewage as the feed at water recovery up to 80%. The hybrid system also showed an excellent capacity for TrOC removal, with removal rates ranging from 91 to 98%. Concentrations of organic matter and TrOCs in the draw solution increased substantially as the water recovery increased. This accumulation of some contaminants in the draw solution was attributed to the difference in their rejection by the FO and MD systems. It was shown that granular activated carbon adsorption or ultraviolet oxidation could be used to prevent contaminant accumulation in the draw solution, resulting in near complete rejection (>99.5%) of TrOCs. This FO – MD hybrid system was also applied to the simultaneous extraction of phosphorus and clean water from digested sludge centrate. FO concentrates orthophosphate and ammonium for subsequent phosphorus recovery in the form of struvite (MgNH4PO4·6H2O), while MD was used to recover the draw solution and extract clean water from the digested sludge centrate. The FO process experienced water flux decline during operation, but fouling was largely reversible after a brief, simple membrane flushing using deionized water. The FO process also provided an effective pretreatment capacity to the subsequent MD process, which exhibited stable water flux. The use of MgCl2 as the draw solute for the FO process was another novel aspect of the system. The reverse salt flux of magnesium to the concentrated digested sludge across the FO membrane and the diffusion of protons away from the digested sludge create favorable conditions for the formation of struvite crystals. The precipitates obtained in the hybrid process were verified to be struvite crystals by examining crystal morphology, element composition, and crystal structure.

FoR codes (2008)




Unless otherwise indicated, the views expressed in this thesis are those of the author and do not necessarily represent the views of the University of Wollongong.