Journal of Membrane Science 365 (2010) 34–39 Contents lists available at ScienceDirect Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci Comparison of fouling behavior in forward osmosis (FO) and reverse osmosis (RO) Sangyoup Lee a , Chanhee Boo a , Menachem Elimelech b , Seungkwan Hong a,∗ a b Department of Civil, Environmental & Architectural Engineering, Korea University, 1-5 Ga, Anam-Dong, Sungbuk-Gu, Seoul 136-713, Republic of Korea Department of Chemical Engineering, Environmental Engineering Program, Yale University, New Haven, CT 06520-8286, USA a r t i c l e i n f o Article history: Received 7 June 2010 Received in revised form 17 August 2010 Accepted 22 August 2010 Keywords: Forward osmosis Reverse osmosis Cake-enhanced osmotic pressure (CEOP) Fouling reversibility a b s t r a c t Fouling behaviors during forward osmosis (FO) and reverse osmosis (RO) are compared. Alginate, humic acid, and bovine serum albumin (BSA) are used as model organic foulants, and two suspensions of silica colloids of different sizes are chosen as model particulate foulants. To allow meaningful comparison of fouling behavior, identical hydrodynamic operating conditions (i.e., initial permeate flux and cross-flow velocity) and feed water chemistries (i.e., pH, ionic strength, and calcium concentration) are employed during FO and RO fouling runs. The observed flux-decline behavior in FO changed dramatically with the type of organic foulant, size of colloidal foulant, and the type of the draw solution employed to generate the osmotic driving force. Based on these experimental data and the systematic comparisons of fouling behaviors of FO and RO, we provide new insights into the mechanisms governing FO fouling. In FO, reverse diffusion of salt from the draw solution to the feed side exacerbates the cake-enhanced osmotic pressure within the fouling layer. The elevated osmotic pressure near the membrane surface on the feed side leads to a substantial drop in the net osmotic driving force and, thus, significant decline of permeate flux. Our results further suggest that the structure (i.e., thickness and compactness) of the fouling layers of FO and RO is quite different. By varying the cross-flow velocity during the organic fouling runs, we were able to examine the fouling reversibility in FO and RO. The permeate flux during organic fouling in FO recovered almost completely with increasing cross-flow velocity, while no noticeable change was observed for the RO system. Our results suggest that organic fouling in FO could be controlled effectively by optimizing the hydrodynamics in the feed stream without employing chemical cleaning. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Membrane-based seawater desalination and wastewater reuse are widely considered as promising solutions to augment water supply and alleviate water scarcity. At present, reverse osmosis (RO) is one of the most effective and robust technologies for seawater desalination and wastewater reuse. However, RO uses an average of 4 kW prime (electric) energy to produce one cubic meter of product water [1,2], which results in emission of 1.8 kg CO2 per cubic meter of product water [3]. In addition, fouling is inevitable in RO systems, thereby requiring the use of chemical cleaning agents and increasing the cost of water production by RO technology [4–6]. In recent years, there has been a growing interest in forward osmosis (FO). Recent studies have shown that FO could be of strategic importance in several applications where RO has dominated for several decades as well as in other applications such as liquid food processing and material recycling [7–10]. The distinguishing feature of FO compared to RO is the use of osmotic pressure gradient ∗ Corresponding author. Tel.: +82 2 3290 3322; fax: +82 2 928 7656. E-mail address: skhong21@korea.ac.kr (S. Hong). 0376-7388/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2010.08.036 as the driving force for water permeation across a semi-permeable membrane. The main advantages of using FO are that it operates at low or no hydraulic pressure, it can achieve high rejection of a wide range of contaminants, and it may have lower irreversible fouling than pressure-driven membrane processes because of the lack of applied hydraulic pressure. Among these advantages, the latter is quite attractive, especially in the area of water and wastewater treatment as well as sea and brackish water desalination. However, to date, no work has been done to directly compare FO and RO fouling behaviors in fouling runs performed under identical physicochemical conditions. In addition, studies investigating the mechanisms of fouling in FO are rather scarce [11,12]. Hence, delineating the differences between FO and RO fouling behavior and elucidating the mechanisms and factors governing FO fouling are of paramount importance. The objectives of this study are to systematically compare the fouling behaviors in FO and RO and to further elucidate the fouling mechanisms in FO. To do so we employed identical hydrodynamic operating conditions and feed water chemistries as well as plateand-frame cells with identical channel dimensions. Various organic macromolecules and colloidal particles of different sizes were used as model foulants. We also examined the reversibility of fouling in FO by altering the hydrodynamic conditions during fouling. Our S. Lee et al. / Journal of Membrane Science 365 (2010) 34–39 observations provide new insights into the fouling mechanisms in FO and strategies for fouling control. 2. Materials and methods 2.1. Organic foulant Alginate, Suwannee River humic acid (SRHA), and BSA were used as model organic foulants to represent common polysaccharides, natural organic matter, and proteins, respectively. These organic macromolecules have been reported to be the major components of organic fouling during membrane filtration of surface water, seawater, and wastewater effluent [13,14]. Alginate (Sigma–Aldrich, St. Louis, MO), SRHA (International Humic Substances Society, St. Paul, MN), and BSA (Sigma–Aldrich, St. Louis, MO) were received in a powder form. Stock solutions (2 g/L) were prepared by dissolving the organic foulant in deionized (DI) water, followed by filtration with a 0.45 ␮m filter (Millipore, Billerica, CA). Each stock solution was stored in sterilized glass bottle at 4 ◦ C. The molecular weights of alginate, SRHA, and BSA are about 12–80 kDa, 1–5 kDa, and 60–70 kDa, respectively as outlined in our previous studies [15]. 2.2. Colloidal particle foulants The model colloidal foulants used are two types of silica (SiO2 ) colloids with different particle sizes (ST-50 and MP-3040, Nissan Chemical Industries, Ltd., NY). Both colloidal suspensions are spherical and monodisperse with average diameters of 20–30 nm for ST-50 and 300 nm for MP-3040. The particles were supplied as a stable concentrated aqueous suspension at an alkaline pH (8.5–9.5). Gravimetric analysis revealed the density of the 20 and 300 nm particles to be 1.4 and 2.3 g/cm3 , respectively [16]. Concentrated stock suspensions of silica were stored at 4 ◦ C. Prior to use, the stock suspension was hand shaken and then sonicated for at least 20 min to ensure good dispersion of the colloidal particles. The concentration of the stock suspension was routinely monitored following the same dispersing procedure. Detailed characteristics of these particles can be found elsewhere [16]. 2.3. Membrane The membrane used in this study was provided by Hydration Technologies (Albany, OR). The membrane chemistry is proprietary, though it is believed to consist of cellulose-based polymers. The membrane has an embedded polyester mesh to provide mechanical support. The total thickness of the membrane is approximately 50 ␮m. Detailed description of the structure and properties of the membrane is available in our previous studies [17]. 2.4. FO and RO systems Schematic diagrams and detailed description of the cross-flow bench-scale FO and RO systems used in our study are given elsewhere [11,18]. The membrane cells in both systems have the same geometry, except that the FO cell has two symmetric channels on both sides of the membrane for co-current flows of the feed and draw solutions. The cross-flow FO unit is custom built with channel dimensions of 77 mm long by 26 mm wide by 3 mm deep. Variable speed gear pumps (Micropump, Vancouver, WA) were used to pump the liquids in a close loop. A constant temperature water bath was used to maintain both the feed and draw solution temperature at 20 ◦ C. The draw solution tank rested on a digital scale (Denver Instruments, Denver, CO) and weight changes were measured over time to determine the permeate water flux by a computer. 35 2.5. Fouling tests The hydrodynamic operating conditions (i.e., initial permeate flux and cross-flow velocity) and feed water compositions (i.e., foulant concentration, pH, ionic strength, and calcium concentration) during the FO and RO fouling tests were identical. Prior to fouling tests, the membranes in both systems were equilibrated with foulant-free test solutions for several hours, and during this stage, the initial flux was adjusted to the same value for both systems. For the FO tests, NaCl or dextrose was used as draw solutions to produce the osmotic pressure driving force. Fouling tests were initiated by adding proper amounts of foulant stock solution to the feed water reservoir. Ionic strength of feed solution was adjusted with NaCl and calcium ion concentration was kept constant at a concentration of 1 mM. Permeate flux was continuously monitored using a digital balance and recorded in real time on a laboratory computer. Throughout the fouling tests, the temperatures of the feed (both FO and RO) and draw (only FO) solutions were maintained at 20 ◦ C. In the FO runs, the permeate flows from the feed tank to the draw solution tank; therefore, the feed solution is gradually concentrated while the draw solution is gradually diluted. To circumvent these phenomena in the FO runs, a baseline test was performed with foulant-free feed solution prior to each fouling run. During this baseline test, feed solution conductivity was measured to determine reverse salt diffusion from draw solution to feed solution. The flux curves obtained from the baseline tests were used to correct the flux curves obtained during fouling runs. In addition, the permeate from the RO fouling runs was discarded rather than recycled back to the feed tank to simulate the gradual concentration of foulants during FO fouling runs. All other techniques and procedures involved in the operation of the FO and RO systems are available in our recent studies [11,15]. 3. Results and discussion 3.1. Organic fouling in FO and RO The flux-decline curves obtained during the FO and RO organic fouling runs are depicted in Fig. 1. The same initial flux of 7.0 ␮m/s (25.2 L/m2 h) was applied to all tests, using 5.0 M of NaCl draw solution for FO and a hydraulic pressure of 3102.6 kPa (450 psi) for RO. As shown in Fig. 1, permeate flux declines more significantly in FO for alginate and humic acid, while a much lower flux decline rate is observed in FO for BSA. In the case of organic fouling of salt rejecting membranes, the decline in permeate flux is generally attributed to the increase in the total hydraulic resistance contributed by the organic fouling layer [19]. Therefore, the structure (compactness) and thickness of the fouling layer are the main factors controlling the flux decline behavior during organic fouling. The structure of the organic fouling layer is affected by both chemical (i.e., pH, ionic strength, and divalent cations) and physical (i.e., initial permeate flux, cross-flow velocity, and applied pressure) conditions [20,21]. Among these factors, it is known that divalent calcium cations play a predominant role in organic fouling since the cross-linking of organic macromolecules by divalent cations forms a compact, dense and thick fouling layer [19]. The structure of the organic fouling layer is also determined by the applied hydraulic pressure. Organic foulants such as those used in this study form deformable gel-like fouling layers that should be significantly affected by the applied pressure. As previously discussed, the driving force for separation in FO and RO is different. Because FO uses an osmotic pressure gradient, the organic foulants loosely accumulate on the membrane surface and form a sparse and thick fouling layer. On the other hand, RO uses high hydraulic 36 S. Lee et al. / Journal of Membrane Science 365 (2010) 34–39 brane. The large NaCl gradient induces reverse diffusion of NaCl from the draw solution side of the membrane to the feed side of the membrane. Since the organic foulants form a thick gel layer in the presence of calcium ions, the salt that passed from the draw solution side of the membrane is trapped by the organic fouling layer leading to significant cake-enhanced osmotic pressure and higher flux decline. Although RO is also subject to CEOP by the rejected salt, the effect is much less pronounced. Accelerated CEOP in FO is mostly due to the reverse salt diffusion from the draw solution and the much thicker cake layers compared to RO. The reason that only BSA protein exhibits lower flux decline in FO is attributed to the salting in and/or salting out effects (i.e., the Hofmeister effects) for proteins [26]. It is known that salts destabilize (i.e., salting in) or stabilize (salting out) many proteins when added to their solutions. These Hofmeister effects are manifested via salt-induced changes of the hydrophobic/hydrophilic properties of protein–water interfaces [27]. In FO, the BSA fouling layer is exposed to a salinity-rich environment due to the salts reversely diffused from the draw solution to fouling layer on the feed side. Therefore, it can be assumed that the protein fouling layer undergoes structural deformation by salting in and/or out effects. This allows the protein fouling layer to be easily removed from the membrane surface by the hydrodynamic shear generated by the flow, thereby leading to less accumulation of foulants on the membrane surface. Reciprocally, the role of accelerated CEOP in flux decline during alginate and humic acid fouling is also noticeably diminished in BSA fouling. 3.2. Colloidal particle fouling in FO and RO Fig. 1. Flux-decline curves obtained during the FO and RO organic fouling runs with (a) alginate, (b) humic acid, and (c) bovine serum albumin (BSA). All fouling runs were performed with identical solutions (i.e., 200 mg/L total organic foulant concentration, 1 mM calcium, 50 mM total ionic strength, and pH 6–7). A 5.0 M NaCl draw solution is used in FO and a hydraulic pressure of 3102.6 kPa (450 psi) is applied in RO. Experimental conditions during the fouling runs with both FO and RO: initial flux of 7.0 ␮m/s (25.2 L/m2 h), cross-flow velocity of 8.5 cm/s, and temperature of 21.0 ± 1.0 ◦ C. pressure, therefore resulting in a more compact, dense, and thin fouling layer under the action of the applied pressure. For salt rejecting membranes, like NF and RO, cake-enhanced osmotic pressure (CEOP) has been recognized as an important mechanism for flux decline, particularly for colloidal particles and bacterial cells [22–25]. In this mechanism, the cake layer hinders the back diffusion of salt thereby resulting in elevated osmotic pressure near the membrane surface. In this study, 5 M NaCl is used as the draw solution, forming a large gradient of salt across the mem- To compare the flux decline behaviors during the colloidal fouling of FO and RO, we used two colloidal suspensions of different size as model particulate foulants. The cake-enhanced osmotic pressure mechanism (CEOP) has been shown previously as a major cause of flux decline for salt rejecting membranes [28]. According to this mechanism, permeate flux decline is not caused by the cake layer resistance, but rather due to the enhanced concentration polarization and hence osmotic pressure within the particle cake layer near the membrane surface. The data in Fig. 2 supports this mechanism of flux decline. The results further show that the flux-decline behavior in FO and RO changes dramatically according to the particle size. Similar flux decline rates were observed with the 20 nm silica particles (Fig. 2a). This result can be explained by both the fouling layer characteristics and the dependence of the CEOP mechanism on particle size. The small colloidal particles have large back diffusion compared to the convective permeate flow, thereby resulting in a thin cake layer. CEOP is insignificant for the thin colloid cake layer because of the short pathway for back diffusion. The thin cake layer also provides a small hydraulic resistance to permeate flow in both FO and RO, yielding very small flux decline rate for FO and RO. Because colloidal silica particles are not deformable, the cake layer in RO could not be further compressed above the maximum packing density of hard, rigid spheres. Interestingly, at the end of the fouling runs, a slightly greater flux decline was observed in FO due to the continual intrusion of salt from draw solution. We verified the relationship between particle size and flux decline mechanisms by conducting fouling runs with larger silica particles (300 nm) (Fig. 2b). Based on Carman–Kozeny equation, the cake layer formed with these particles has negligible resistance compared to the intrinsic membrane resistance [29–31]. The 300 nm colloidal particles develop a thicker porous cake layer which hinders back diffusion of salt. Notably, in FO, there is a greater salt build-up near the membrane surface due to salt intrusion from the draw solution. This results in a substantial drop in the net osmotic pressure difference and hence severe flux decline. S. Lee et al. / Journal of Membrane Science 365 (2010) 34–39 37 Fig. 3. Effect of salt reverse diffusion from the draw solution to the feed side on FO alginate fouling by cake-enhanced osmotic pressure (CEOP). Dextrose was used as a draw solution in addition to NaCl. In order to achieve the same initial flux in FO, 3.0 M dextrose and 0.6 M NaCl were used in FO and a hydraulic pressure of 1241.1 kPa (180 psi) is applied in RO. Other experimental conditions were identical to those in Fig. 1. reverse diffusion on the development of the CEOP for two different draw solutions. As shown, the reverse diffusion of dextrose is negligible compared to NaCl, therefore not accelerating the buildup of cake-enhanced concentration polarization. This explains the much milder flux decline in FO with dextrose as a draw solution compared to NaCl. 3.4. Fouling in FO is reversible Fig. 2. Flux-decline curves obtained during the FO and RO colloidal fouling runs with silica colloids: (a) 5 g/L of 20 nm silica colloids and (b) 400 mg/L of 300 nm silica colloids. A 3.0 M NaCl draw solution is used in FO and a hydraulic pressure of 2068.4 kPa (300 psi) is applied in RO. Other experimental conditions during the fouling runs were identical to those in Fig. 1. Our results in Fig. 2b demonstrate a remarkable difference between flux decline rates of two systems due to the predominance of cakeenhanced osmotic pressure in FO. 3.3. Cake-enhanced osmotic pressure is an important fouling mechanism in FO To verify that the reverse diffusion of salt from the draw solution to feed side accelerates CEOP in FO, dextrose was used as another draw solution. In Fig. 3, three flux-decline curves are presented, where alginate was used as organic foulant, similar to Fig. 1b. Comparing the flux-decline curves for ‘RO (1241.1 kPa (180 psi))’ and ‘FO (0.6 M NaCl)’ shows that the flux decline for FO is much severer than RO. This result is quite similar to the previous result shown in Fig. 1b, where reverse diffusion of NaCl from the draw to feed accelerated CEOP, resulting in more flux decline in FO than RO. Interestingly, however, almost identical flux-decline behavior was observed when 3.0 M of dextrose was used as draw solution instead of 0.6 M NaCl. This result clearly demonstrates that the reverse diffusion of salt from the draw solution to the fouling layer on the feed side accelerates CEOP and consequently reduces the net driving force for water permeation in FO. Dextrose is effectively retained by the membrane because of its much larger hydrated radius than that of NaCl, leading to negligible reverse diffusion of dextrose from the draw to the feed side. Fig. 4 illustrates the impact of draw solute Organic foulants loosely accumulate on the membrane surface and form a sparse and thick fouling layer in FO due to the absence of mechanical hydraulic pressure. With this loose and sparse fouling layer, there might be no need to apply harsh chemical cleaning, since the fouling could be reversible by simple physical cleaning such as hydraulic flushing. This is not likely the case for organic fouling in RO as the fouling layer is compact, dense, and cross-linked and, hence, mostly irreversible when subject to physical cleaning. At present, in RO systems, chemical cleaning is inevitable. To verify these structural differences between FO and RO fouling layers, the cross-flow velocity was increased during the course of FO and RO fouling runs (Fig. 5). The cross-flow velocity was increased to a level three times higher than the initial cross-flow velocity at about 12 h after the initiation of fouling run. Prior to increasing the cross-flow velocity, FO suffered from flux decline more severely than RO. When increasing the cross-flow velocity, however, the reduced flux during FO fouling run recovered almost completely and reached to the initial value, while there was no noticeable flux recovery in RO. This demonstrates that the structure of the FO fouling layer is loose and sparse enough to be disrupted and removed by the hydraulic shear generated by the increased cross-flow. On the other hand, the same shear rate is not enough to disrupt the compact and cohesive fouling layer formed during the RO fouling run and, thus no visible change in flux behavior was observed. In addition, it is interesting to note that the flux-decline behavior in FO after increasing the cross-flow velocity is quite different from the initial flux-decline behavior. At the higher cross-flow velocity, the flux decline rate is more moderate than the case before increasing the cross-flow. This implies that foulant accumulation on the membrane surface is much less at the higher cross-flow velocity, thereby leading to the formation of a much thinner and looser fouling layer. This thin fouling layer contributes much less to CEOP acceleration by the salt intruded from the draw solution as concentration polar- 38 S. Lee et al. / Journal of Membrane Science 365 (2010) 34–39 Fig. 4. A conceptual illustration of the effect of draw solute reverse diffusion on cake-enhanced osmotic pressure (CEOP) in FO for different draw solutions: (a) NaCl and (b) dextrose. CEOP is greatly enhanced due to the reverse diffusion of NaCl (a) compared to the case with negligible reverse diffusion (i.e., with dextrose) (b). ization within the fouling layer is highly dependent of the fouling layer thickness. Compressible and deformable organic foulants, like alginate in our case, would compress significantly under the action of hydraulic pressure, thereby yielding a compact and cohesive fouling layer. This is not the case under the action of osmotic pressure in FO, which results in a less compact and cohesive fouling layer that can easily be removed by simple physical cleaning. On the other hand, hard colloidal particles, like the silica colloids used in this study, are not compressible and the cake (fouling) layer can attain a maximum packing density of 0.64 (i.e., maximum random packing for rigid spherical particles). In this case, the drag force depends solely on the fluid velocity which is proportional to the pressure gradient, not the pressure itself. So, in this case, cake structures in FO and RO may be similar. In many practical cases, however, We have shown in Fig. 5 that the reduced water flux was easily recovered by increasing the cross-flow velocity during the FO fouling run, and that the ensuing flux-decline rate was lower after increasing the cross-flow velocity. In Fig. 6, three different initial cross-flow velocities (i.e., 17.1, 25.6, and 34.2 cm/s) were employed in the FO fouling runs, and the flux-decline behaviors were com- Fig. 5. Comparison of fouling layer cohesion/compactness in FO and RO during organic (alginate) fouling. Upon increasing cross-flow velocity from 8.54 to 25.6 cm/s, the water flux in FO increased due to removal of the loose cake layer, but no change was observed in RO because of the compacted, cohesive fouling layer. Experimental conditions for fouling with alginate are similar to those described in Fig. 1. Fig. 6. Mitigation of flux decline in FO at high cross-flow velocity. Three different initial cross-flow velocities (17.1, 25.6, and 34.2 cm/s) were employed during the FO fouling runs, and the flux-decline behaviors are compared. The flux curve for the lowest cross-flow velocity of 8.54 cm/s was adopted from FO data in Fig. 5. fouling layers are most compressible as both compressible and incompressible foulants are combined in the fouling layer. Moreover, even though incompressible colloidal particles may dominate in certain cases, there is still a possibility of cake layer compression when different sizes of particles are present (i.e., polydisperse suspension). 3.5. Fouling in FO is mitigated at higher cross-flow velocities S. Lee et al. / Journal of Membrane Science 365 (2010) 34–39 pared with each other. The flux curve for the lowest cross-flow velocity of 8.54 (from data in Fig. 5) is also included for comparison. It is clearly shown that flux decline (or fouling) is effectively mitigated with increasing cross-flow velocity (i.e., from 8.54 to 34.2 cm/s). Interestingly, at the highest cross-flow velocity (i.e., 34.2 cm/s), almost no flux decline was observed. It should be noted that this effective fouling control by increasing cross-flow velocity is not possible with RO fouling. As we have shown earlier in Fig. 5, there was no flux recovery even after increasing the crossflow velocity by three times (i.e., from 8.54 to 25.6 cm/s). These results imply that FO fouling can be effectively controlled by optimizing hydrodynamic operating conditions and cleaned easily by simple physical cleaning methods without harsh chemical cleaning. 4. Conclusions In this study, we systematically compared the fouling behaviors in FO and RO and further elucidated the fouling mechanisms in FO. It has been found that the key mechanism of flux decline in FO is rather accelerated cake-enhanced osmotic pressure (CEOP) due to reverse salt diffusion from the draw to feed than the increase in fouling layer resistance. This implies that selecting the proper draw solution (i.e., less back diffusion) and/or improving the membrane property (i.e., higher selectivity) are of paramount importance in efficient operation of FO. Interestingly, fouling in FO is almost reversible while irreversible in RO. Deformable organic foulant without hydraulic pressure (i.e., FO) makes loose and sparse fouling layer that can easily be removed by simple physical cleaning. The fact that accelerated cake-enhanced osmotic pressure (CEOP) due to reverse salt diffusion from the draw solution is a key mechanism of FO fouling implies that FO fouling could be minimized by selecting proper draw solution and/or improving reverse salt rejection by membrane. Fouling control and membrane cleaning in FO are much more feasible than RO since FO fouling is reversible to simple physical cleaning. For example, fouling in membrane bioreactors (MBR) in advanced wastewater treatment may be reduced significantly when FO-hybrid MBR system is used as the continuous bubble aeration in MBR could efficiently prevent foulant accumulation on the membrane surface. Acknowledgements This research was supported by World Class University (WCU) program (Case III) through the National Research Foundation of Korea funded by the Ministry of Education, Science and Technology (R33-10046) and partly by Seawater Engineering & Architecture of High Efficiency Reverse Osmosis (SEAHERO) program supported by the Ministry of Land, Transport and Maritime Affairs (MLTM). References [1] R. Semiat, Energy issues in desalination processes, Environ. Eng. Sci. 42 (2008) 8193–8201. [2] S. Liang, C. Liu, L.F. Song, Two-step optimization of pressure and recovery of reverse osmosis desalination process, Environ. Eng. Sci. 43 (2009) 3272– 3277. [3] G. Raluy, L. Serra, J. Uche, Life cycle assessment of MSF, MED and RO desalination technologies, Energy 31 (2006) 2361–2372. 39 [4] S.M. Mylon, K.L. Chen, M. Elimelech, Influence of natural organic matter and ionic composition on the kinetics and structure of hematite colloid aggregation: implications to iron depletion in estuaries, Langmuir 20 (2004) 9000–9006. [5] V.T. Lahoussine, M.R. Wiesner, J.Y. Bottero, Fouling in tangential-flow ultrafiltration: the effect of colloid size and coagulation pretreatment, J. Membr. Sci. 52 (1990) 173–190. [6] M. Herzberg, S. Kang, M. Elimelech, Role of extracellular polymeric substances (EPS) in biofouling of reverse osmosis membranes, Environ. Sci. Technol. 43 (2009) 4393–4398. [7] T.Y. Cath, A.E. Childress, M. Elimelech, Forward osmosis: principles, applications, and recent developments, J. Membr. Sci. 281 (2006) 70–87. [8] M.I. Dova, K.B. Petrotos, H.N. Lazarides, On the direct osmotic concentration of liquid foods. Part I. Impact of process parameters on process performance, J. Food Eng. 78 (2007) 422–430. [9] B. Jiao, A. Cassano, E. Drioli, Recent advances on membrane processes for the concentration of fruit juices: a review, J. Food Eng. 63 (2004) 303–324. [10] M.A. Shannon, P.W. Bohn, M. Elimelech, J.G. Georgiadis, B.J. Marinas, A.M. Mayes, Science and technology for water purification in the coming decades, Nature 452 (2008) 301–310. [11] B. Mi, M. Elimelech, Chemical and physical aspects of organic fouling of forward osmosis membranes, J. Membr. Sci. 320 (2008) 292–302. [12] B. Mi, M. Elimelech, Organic fouling of forward osmosis membranes: fouling reversibility and cleaning without chemical reagents, J. Membr. Sci. 348 (2010) 337–345. [13] G.T. Grant, E.R. Morris, D.A. Rees, J.C. Smith, D. Thom, Biological interaction between polysaccharides and divalent cations: the egg-box model, FEBS Lett. 32 (1973) 195–198. [14] H. Ma, H.E. Allen, Y. Yin, Characterization of isolated fractions of dissolved organic matter from natural waters and a wastewater effluent, Water Res. 35 (2001) 985–996. [15] S. Lee, M. Elimelech, Relating organic fouling of reverse osmosis membranes to intermolecular adhesion forces, Environ. Sci. Technol. 40 (2006) 980–987. [16] S. Lee, J. Cho, M. Elimelech, Combined influence of natural organic matter (NOM) and colloidal particles on nanofiltration membrane fouling, J. Membr. Sci. 262 (2005) 27–41. [17] J.R. McCutcheon, M. Elimelech, Influence of membrane support layer hydrophobicity on water flux in osmotically driven membrane processes, J. Membr. Sci. 318 (2008) 458–466. [18] W.S. Ang, S. Lee, M. Elimelech, Chemical and physical aspects of cleaning of organic-fouled reverse osmosis membranes, J. Membr. Sci. 272 (2006) 198–210. [19] S. Hong, M. Elimelech, Chemical and physical aspects of natural organic matter (NOM) fouling of nanofiltration membranes, J. Membr. Sci. 132 (1997) 159–181. [20] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic matter: factors and mechanisms affecting rejection and flux decline with charged ultrafiltration (UF) membrane, J. Membr. Sci. 164 (2000) 89–110. [21] A. Braghetta, F.A. DiGiano, W.P. Ball, Nanofiltration of natural organic matter: pH and ionic strength effects, J. Environ. Eng. ASCE 132 (1997) 628–641. [22] E.M.V. Hoek, A.S. Kim, M. Elimelech, Influence of crossflow membrane filter geometry and shear rate on colloidal fouling in reverse osmosis and nanofiltration separations, Environ. Eng. Sci. 19 (2002) 357–372. [23] E.M.V. Hoek, M. Elimelech, Cake-enhanced concentration polarization: a new fouling mechanism for salt-rejecting membranes, Environ. Sci. Technol. 37 (2003) 5581–5588. [24] S. Lee, J. Cho, M. Elimelech, Influence of colloidal fouling and feed water recovery on salt rejection of RO and NF membranes, Desalination 160 (2004) 1–12. [25] M. Herzberg, M. Elimelech, Biofouling of reverse osmosis membranes: role of biofilm-enhanced osmotic pressure, J. Membr. Sci. 295 (2007) 11–20. [26] S. Wang, N. Amornwittawat, J. Banatlao, M. Chung, Yu. Kao, X. Wen, Hofmeister effects of common monovalent slats on the beetle antifreeze protein activity, J. Phys. Chem. B 113 (2009) 13891–13894. [27] A. Dér, L. Kelemen, L. Fábián, S.G. Taneva, E. Fodor, T. Páli, A. Cupane, M.G. Cacace, J.J. Ramsden, Interfacial water structure controls protein conformation, J. Phys. Chem. B 111 (2007) 5344–5350. [28] S. Lee, J. Cho, M. Elimelech, A novel method for investigating the influence of feed water recovery on colloidal and NOM fouling of RO and NF membranes, Environ. Eng. Sci. 22 (2005) 496–509. [29] R.S. Faibish, M. Elimelech, Y. Cohen, Effect of interparticle electrostatic double layer interactions on permeate flux decline in crossflow membrane filtration of colloidal suspensions: an experimental investigation, J. Colloid Interface Sci. 204 (1998) 77–86. [30] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publishers, Dordrecht/Boston/London, 1996. [31] P.C. Carmen, Fluid flow through a granular bed, Trans. Inst. Chem. Eng. 15 (1937) 150–156.