Gypsum Scaling and Cleaning in Forward Osmosis: Measurements

Feb 12, 2010 - Andrew S. Colburn , Noah Meeks , Steven T. Weinman , and Dibakar .... Jaichander Swaminathan , Gregory P. Thiel , Syed M. Zubair , Alla...
0 downloads 0 Views 4MB Size
Download PDF
Environ. Sci. Technol. 2010, 44, 2022–2028

Gypsum Scaling and Cleaning in Forward Osmosis: Measurements and Mechanisms B A O X I A M I * ,† A N D MENACHEM ELIMELECH‡ Department of Civil and Environmental Engineering, The George Washington University, 641 Academic Center, 801 22nd Street, NW, Washington, D.C. 20052, and Department of Chemical Engineering, Environmental Engineering Program, P.O. Box 208286, Yale University, New Haven, Connecticut 06520-8286

Received November 30, 2009. Revised manuscript received January 15, 2010. Accepted January 15, 2010.

This study investigates gypsum scaling and cleaning behavior in forward osmosis (FO). The results show that gypsum scaling in FO is almost fully reversible, with more than 96% recovery of permeate water flux following a water rinse without addition of chemical cleaning reagents. Parallel comparisons of fouling and cleaning were made between FO (without hydraulic pressure) and RO (under high hydraulic pressure) modes. The shape of the water flux decline curves during gypsum scaling is similar in the two modes, but the flux recovery in FO mode is higher than that in RO mode by about 10%. This behavior suggests that operating in FO mode may reduce the need for chemical cleaning. The role of membrane materials in controlling gypsum scaling and cleaning was investigated using cellulose acetate (CA) and polyamide (PA) membranes. GypsumscalingofPAmembranescausesmoreseverefluxdecline and is harder to clean than that of CA membranes. AFM force measurements were performed between a gypsum particle probe and the membrane surfaces to elucidate gypsum scaling mechanisms. Analysis of adhesion force data indicates that gypsum scaling of the PA membrane is dominated by heterogeneous/surface crystallization, while gypsum scaling of the CA membrane is dominated by bulk crystallization and subsequent particle deposition. This finding implies that membrane surface modification and new material development can be an effective strategy to mitigate membrane scaling.

Introduction In recent years, worldwide water scarcity has stimulated interest in using reverse osmosis (RO) and nanofiltration (NF) membrane desalination processes to produce clean water from saline waters, such as brackish ground/surface water, agricultural drainage water, and wastewater (1-4). However, the water recovery of these membrane processes is greatly limited by mineral salt scaling. Scaling occurs when the concentration of sparingly soluble saltsssuch as calcium sulfate, barium sulfate, and calcium carbonatesin feedwater becomes supersaturated at high product water recovery. As a result, precipitation of these salts may occur near or on the * Corresponding author phone: +1-202-994-5691; fax: +1-202994-0127; e-mail: [email protected]. † The George Washington University. ‡ Yale University. 2022

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 6, 2010

membrane surface, leading to severe membrane flux decline. Among the various salts that cause membrane scaling, calcium sulfate dihydrate (gypsum) is one of the most common scaling sources in desalination of brackish water (4, 5). In addition, gypsum scaling is often the most problematic as it cannot be controlled by adjusting pH. Membrane scaling can significantly shorten membrane life and increase the maintenance cost and energy consumption of membrane operation. Therefore, numerous research efforts have been made to seek strategies for scaling control (6-12). Scale mitigation strategies include addition of antiscalants (6-9), adjustment of operating conditions and membrane module configuration (10, 11), and pretreatment to lower the scaling potential of feedwater (12, 13). Although these strategies can mitigate membrane scaling to some extent, scaling still remains a limiting factor for achieving high product water recovery in membrane desalination. In order to develop more effective scaling mitigation strategies, it is imperative to understand gypsum scaling mechanisms (8-10, 14-16). Studies have demonstrated that gypsum scaling of membranes is controlled by both surface/ heterogeneous crystallization (i.e., crystallization takes place directly on the membrane surface) and deposition of bulk crystals (i.e., crystallization takes place in the bulk followed by deposition on the membrane surface) (10, 15). Shih et al. (14) studied the surface crystallization of gypsum on several polyamide (PA) RO membranes, and observed significant differences in the extent of surface scale coverage and surface crystal size among the selected membranes. Although this study (14) demonstrated a significant impact of membrane surface topology and chemistry on surface crystallization, no further studies, to the best of our knowledge, have investigated or explained the effects of membrane materials on gypsum scaling. Gaining a mechanistic understanding of the role of membrane materials in surface crystallization and scaling is essential to effectively modify membrane surfaces to minimize scaling potential. The recently emerging forward osmosis (FO) membrane process is a potential, sustainable alternative to conventional membrane processes in sea/brackish water desalination and wastewater reclamation (17-26). Instead of using hydraulic pressure, as in pressure-driven membrane processes, FO uses a concentrated draw solution to generate high osmotic pressure, which pulls water across a semipermeable membrane from the feed solution. The draw solutes are then separated from the diluted draw solution to recycle the solutes and to produce clean product water. The lack of hydraulic pressure in FO membrane operation has been demonstrated to be advantageous in regards to fouling and cleaning (18, 22, 26-28). Mi and Elimelech (27) demonstrated that organic (alginate) fouling in FO membranes can be fully reversed by rinsing with pure water without using chemical cleaning reagents. Holloway et al. (28) compared the fouling behavior of FO and RO for wastewater centrate treatment, demonstrating a slower flux decline rate in FO than in RO. Achilli et al. (18) and Cornelissen et al. (22) observed good fouling resistance for FO membranes in osmotic membrane bioreactors. Overall, the above studies with compressible organic foulants observed greater fouling resistance and/or higher cleaning efficiency for FO than for pressure-driven membranes. However, to date, the effect of hydraulic pressure on mineral salt scaling has not been studied. In addition, FO has the potential to achieve higher product water recovery, where scaling control is particularly 10.1021/es903623r

 2010 American Chemical Society

Published on Web 02/12/2010

FIGURE 1. Comparison of gypsum scaling and cleaning in FO and RO modes: (a) scaling of the CA membrane in FO and RO modes, (b) flux recovery after cleaning, (c) SEM image of gypsum scale in FO mode, and (d) SEM image of gypsum scale in RO mode. The scaling solution contains 35 mM CaCl2, 20 mM Na2SO4, and 19 mM NaCl, with a gypsum saturation index (SI) of 1.3. A 4 M NaCl draw solution is used in FO and 31 bar (450 psi) hydraulic pressure in RO. Other test conditions are the same for FO and RO: crossflow velocity of 8.5 cm/s, ambient pH (pH 6.8), and temperature of 20 ( 1 °C. For cleaning, DI water is used in both the FO and RO systems used DI water to rinse the membrane for 15 min at a crossflow velocity of 21 cm/s and a temperature of 20 ( 1 °C. The membrane water flux was tested with DI as the feed solution after the cleaning experiment. Note that the flux for the fouled membrane is normalized by the initial flux in the fouling experiments, while the recovered flux after cleaning is normalized by the pure water flux. important (26). Therefore, there is a pressing need to study the scaling and cleaning behavior of FO. This study investigates the gypsum scaling and cleaning behavior in FO membrane processes. Parallel scaling and cleaning experiments are performed in the RO mode for comparison of fouling/cleaning behaviors and elucidating the relevant mechanisms in FO. The impact of membrane materials on gypsum scaling and cleaning is also investigated. Atomic force microscopy (AFM) force measurements are used to elucidate the scaling mechanisms at the nanoscale and to explain the impact of membrane materials on gypsum scaling.

Materials and Methods Selected Membranes. Two membranes were used in this study. The first is a cellulose triacetate (designated as CA) membrane from Hydration Technologies, Inc. (Albany, OR). It has an asymmetric structure with an embedded woven mesh to enhance the mechanical strength of the membrane (24). The other membrane used is a modified thin-film composite polyamide (designated as PA) RO membrane from Dow Chemical Company (Midland, MI). The proprietary membrane is likely modified from a commercial seawater RO membrane (Filmtec SW30XLE-400i) by reducing the thickness of the membrane support layer. The two membranes have similar pure water permeabilities determined in RO mode: 3.6 × 10-12 m/(s Pa) for CA and 3.8 × 10-12 m/(s Pa) for PA. However, as we show later, in FO mode, the flux of the CA membrane is much higher than that of the PA membrane under the same testing conditions. The difference is attributed to the effect of internal concentration polarization, which is much more severe for the PA membrane because of the thick support layers (29). FO and RO Membrane Systems. One FO and one RO membrane system were used to conduct the scaling and cleaning experiments. Both systems contain custom-built

crossflow membrane cells. The feed and draw solution channels in FO and the feed channel in RO have similar dimensions: 77 mm in length, 26 mm in width, and 3 mm in depth. A schematic diagram and detailed description of the FO system are given in our recent publication (30). In the RO system, a Hydra-cell pump (Wanner Engineering, Inc., Minneapolis, MN) was used to generate cross-flow, and a bypass valve and backpressure regulator were used to control the crossflow velocity and hydraulic pressure. The permeate flux was monitored continuously by a digital flow meter (Optiflow 1000, Humonics, CA) interfaced with a computer. A constant temperature of 20 ( 1 °C was maintained by a water bath (Neslab, Newington, NH) for both the FO and RO experiments. Gypsum Scaling and Cleaning Experiments. The protocol for all scaling experiments comprised the following steps. First, a new membrane coupon, with the active layer facing the feed channel, was placed in the unit before each experiment and stabilized to obtain a constant flux. The stabilization process takes about 1 h for FO and 5 h for RO. The membrane in the FO mode was stabilized with DI as the feed and 4 M NaCl as the draw solution. The membrane in the RO mode was stabilized/compacted under 31 bar (450 psi) hydraulic pressure with DI as the feed. Next, the pure water flux of the stabilized membrane was obtained before the scaling experiment. The flux in the FO mode was obtained with DI water as the feed solution and 4 M NaCl as the draw solution. The flux in the RO mode was obtained with DI water as the feed solution and 31 bar (450 psi) feed pressure. Then, the gypsum scaling experiment was performed for about 24 h. Both the FO and RO experiments use the same scaling solution: 35 mM CaCl2, 20 mM Na2SO4, and 19 mM NaCl, with a gypsum (CaSO4 · 2H2O) saturation index (SI) of 1.3. To generate the driving force for water flux for the gypsum scaling experiments, 4 M NaCl draw solution is used in FO and 31 bar hydraulic pressure in RO. Other test conditions are the same for FO and RO: crossflow velocity of 8.5 cm/s, VOL. 44, NO. 6, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2023

FIGURE 2. Impact of membrane materials on gypsum scaling and cleaning in both the FO and RO modes: (a) scaling in FO mode, (b) scaling in RO mode, (c) SEM image of gypsum scale on the CA membrane, (d) SEM image of gypsum scale on the PA membrane, and (e) flux recovery after cleaning. The membranes used are cellulose acetate (CA) and polyamide (PA) membranes. In order to achieve the same initial flux in FO, the NaCl draw solution concentrations for the CA and PA membranes are 1.5 and 4 M, respectively. The pressure in RO is 31 bar (450 psi) for both the CA and PA membranes. Other scaling and cleaning conditions are the same as stated in Figure 1. ambient pH (pH 6.8), and temperature of 20 ( 1 °C. The gypsum scaling experiments in both FO and RO were performed in a nonrecycling mode—that is, the permeate was not recycled back to the feed solution such that the gypsum saturation index kept increasing over time. Water flux was continuously monitored throughout the fouling experiments by a PC. Membrane cleaning was performed immediately after the fouling experiments. DI water was used for both the FO and RO systems to rinse the membrane for 15 min at a crossflow velocity of 21 cm/s and a temperature of 20 ( 1 °C. The membrane water flux was tested with DI as the feed solution after the cleaning experiment. Note that the flux for the scaled membrane is normalized by the initial flux in the scaling experiments, while the recovered flux after cleaning is normalized by the pure water flux. AFM Force Measurements. The force measurements were performed with a Multimode AFM (Veeco Metrology Group, Santa Barbara, CA) using a gypsum particle probe. The probe was made by attaching a gypsum particle, measuring 8.0 µm in diameter, to a commercial tipless SiN AFM cantilever (Veeco Metrology Group, Santa Barbara, CA) with Norland Optical adhesive (Norland Products, Inc., Cranbury, NJ) and curing it under UV light for 20 min. To obtain a gypsum particle in the micrometer size range, large gypsum particles were ground and suspended in saturated CaSO4 solutions. The mixture was set for 15 min to allow some large particles to settle down. Then the supernatant was spread on a mica surface and dried. Several rounds of dilution were often needed in order to have single particles on the mica surface. Finally, the mica surface was placed under a microscope and one single gypsum particle was picked up by the AFM probe. An SEM image of the gypsum probe is shown in Supporting Information Figure S1. The AFM adhesion force measurements were performed in a fluid cell filled with the test solution. For each set of force 2024

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 6, 2010

measurements, a piece of fresh membrane was set up in the fluid cell and rinsed with DI water. A saturated gypsum solution was used as the first testing solution. Adhesion force curves were collected after allowing the system to stabilize for 30 min. In order to avoid dissolution of the gypsum probe or particle precipitation from the testing solution during the force measurements, the saturated solution was prepared by taking the supernatant from a supersaturated gypsum solution, which was allowed to sit for a few days to reach equilibrium. The supersaturated solution contains 70 mM CaCl2, 40 mM Na2SO4, and 38 mM NaCl, with a gypsum SI of 3.1. Then, a freshly prepared supersaturated gypsum solution, containing the same chemical composition as described earlier, was injected into the fluid cell. The test solution was left to equilibrate with the FO membrane for 30 min to allow the system to stabilize. The adhesion force curves were then collected at 30, 50, 70, 80, 90, and 100 min. The force measurements were conducted at five different locations on the membrane. Thirty force measurements were taken at each location to minimize inherent variability in the force data, which is mainly attributed to the heterogeneity of the membrane surface. Only the retracting (pull-off) force data were processed and converted to force versus distance curves.

Results and Discussion Comparison of Gypsum Fouling and Cleaning in FO and RO Modes. FO membrane processes use osmotic pressure as the driving force for water flux through the membrane, whereas RO membrane processes use hydraulic pressure as the driving force. In order to study whether the difference in driving force affects gypsum scaling and subsequent cleaning behavior, we tested the CA membrane in both the FO and RO modes. The corresponding flux decline curves are shown in Figure 1a. Note that the water flux for FO was corrected to account for the dilution of the draw solution

and the reduction of the osmotic driving force during the fouling run. Details on the correction protocol can be found in our previous publication (30). After the gypsum scaling experiments, cleaning experiments were immediately carried out. The flux recovery is shown in Figure 1b. Both the fouling and cleaning experiments were very reproducible, as shown in Supporting Information Figure S2. Therefore, not all the experiments were duplicated and no error bars are provided for the flux recovery data. As Figure 1 shows, the flux decline rates in the scaling experiments are practically the same for the FO and RO processes. However, the flux recoveries of the two processes are different. The membrane permeate water flux is almost 100% recovered in the FO mode, while 90% is recovered in the RO mode. This observation indicates that although pressure does not appear to have a significant effect on the flux decline curves, it may cause some differences in the gypsum scale structure and subsequent cleaning efficiency. However, the difference in gypsum scale structure is not large enough to discern in the SEM images of the gypsum scale formed in FO (Figure 1c) and RO (Figure 1d). Nevertheless, it is likely that the scale structure in the FO mode is less compact due to the lack of hydraulic pressure or the back diffusion of ions from the draw solution to the membrane surface on the feed side. As a result, the cleaning efficiency in the FO mode is slightly higher than that in RO mode. Note that our previous study demonstrated that the difference in fouling layer structure and cleaning efficiency is much more significant for a compressible foulant (alginate) (27). Effect of Membrane Materials on Gypsum Scaling and Cleaning. The FO membrane used in this study (denoted as CA) is made of cellulose acetate, which is known to have relatively low fouling potential. In order to study the possible effect of membrane materials on scaling, we chose a thinfilm composite polyamide (PA) membrane, the most widely used type in the RO membrane industry, to compare with the CA membrane. The impact of membrane materials on gypsum scaling was studied in both the FO and RO modes. Note that even though the manufacturer has slightly modified the PA membrane to suit FO processes, it still cannot generate water flux as high as the CA membrane in FO mode. Therefore, the scaling experiments in the FO mode have to be conducted at a much lower initial flux (7 L/m2-h) than those in the RO mode (28 L/m2-h). The flux decline curves in Figure 2a and b show that scaling of the PA membrane causes more severe flux decline than the CA membrane in both the FO and RO modes. In FO mode, the flux decline rate for the two membranes appears to be the same at the initial and intermediate stages. The relatively slow flux decline rate in FO is likely attributed to the relatively low initial flux. However, after about two days of the scaling experiment in FO, a drastic flux decline was observed for the PA membrane, with the flux decreasing to almost zero within a very short period (Figure 2a). In contrast, the flux decline rate of the CA membrane remained unchanged for the entire duration of the experiment. The faster flux decline for the PA membrane indicates much more severe gypsum scaling compared to the CA membrane. In order to determine whether membrane materials affect the structure of gypsum scale on the membrane surface, we took the membrane out of the FO membrane cell immediately after the scaling experiments to examine under SEM. Figure 2c and d present the SEM images of gypsum scale formed on the CA and PA membranes, respectively. Comparison of the two images clearly demonstrates that the gypsum scale formed on the CA membrane has a much looser structure with larger crystals than that on the PA membrane. We then performed cleaning experiments for both membranes. The flux recovery results of the two membranes in

FIGURE 3. AFM force measurements using a gypsum particle probe in a supersaturated gypsum solution. (a) Representative adhesion (pull-off) force curves for the CA membrane. (b) Representative adhesion force curves for the PA membrane. The adhesion force measurements are conducted in a supersaturated gypsum solution, with a saturation index of 3.1, to simulate the forces experienced during gypsum scaling. The forces at time zero are measured with a saturated gypsum solution without gypsum precipitation. The forces at 50 and 100 min are collected with supersaturated gypsum solution when precipitation is taking place. both the FO and RO modes are shown in Figure 2e. The recovered water fluxes of the CA membrane in both the FO and RO modes are above 90% of the initial flux, whereas the recovered fluxes of the PA membrane are only 75 and 60% for FO and RO, respectively. Because the membrane flux before cleaning is significantly different for the two membranes, we cannot directly compare the absolute cleaning efficiencies of the CA and PA membranes. However, the results demonstrate that a significant percentage of the water flux of the PA membrane cannot be recovered by pure water rinsing, consistent with the more compact gypsum scale observed on the surface of the PA membrane. The scaling and cleaning experiments demonstrated that membrane materials significantly affect gypsum scale formation on membrane surfaces. Since the structure of gypsum scale formed on the PA membrane is more compact (based on SEM images), it is hypothesized that the surface chemistry and/or topology of the PA membrane induce more severe heterogeneous crystallization than the CA membrane. To better understand the impact of membrane materials, we performed AFM force measurements to characterize interfacial phenomena on the nanoscale during gypsum scale formation. Material Specific Mechanisms Controlling Gypsum Scaling. The force measurements were performed with either saturated or supersaturated gypsum solutions to prevent dissolution of the gypsum particle probe. The supersaturated VOL. 44, NO. 6, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2025

FIGURE 4. Schematic description of the proposed mechanisms controlling gypsum scaling on the (a) CA and (b) PA membranes. solution is used to obtain the interfacial forces during the dynamic process of gypsum scaling. We used a higher SI in the force measurements (SI ) 3.1) than the scaling experiments (SI ) 1.3), because the solution with SI of 1.3 has 4-5 h induction time for nucleation, which is too long for the force measurements. The induction time of the solution with SI of 3.1 is around 40-50 min, making it suitable for the force measurements. The interfacial forces between the gypsum particle probe and the membrane surface are first measured with saturated gypsum solution, where no precipitation or dissolution takes place. Representative interfacial force curves for the CA and PA membranes are shown in Figure 3a and b, respectively. In this case, since there is no gypsum precipitation, we denoted a time 0 for the force measurements. The results in Figure 3 show that the adhesion force at time 0 for the CA membrane is roughly the same as the force for the PA membrane. After collecting the force curves at time 0 (without gypsum precipitation), we injected a freshly prepared supersaturated gypsum solution into the fluid cell, and collected force curves at 30, 50, 70, 80, 90, and 100 min. Two representative force curves at 50 and 100 min are shown for the CA and PA membranes in Figure 3a and b, respectively. Note that during these measurements, gypsum has already started precipitating in the solution and/ or on the membrane surface. Figure 3a demonstrates that the adhesion forces for the CA membrane at 50 and 100 min are not much different from the force without precipitation (0 min). This observation indicates that there is no or little heterogeneous (surface) crystallization on the CA membrane surface, as surface crystallization is expected to impact the adhesion forces. However, for the PA membrane, once crystallization takes place, the adhesion force curves change dramatically, with the rupture distance increasing from 80 nm at 0 min to 450 nm at 100 min. The force curves with the extended rupture distance clearly indicate a stepwise disruption of the interaction forces between the gypsum probe and the membrane surface. It appears that gypsum precipitation increases the number of interaction sites between the gypsum probe and the PA membrane surface, but does not change the 2026

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 6, 2010

FIGURE 5. The change of (a) force, (b) rupture distance, and (c) adhesion energy over time in the process of gypsum precipitation. (a), (b), and (c) share the same horizontal axis. The adhesion energy is obtained from integrating the force curve obtained at each point of time. This energy represents the total work needed to pull the gypsum particle probe off the membrane surface.

magnitude of adhesion force. We hypothesize that this phenomenon is due to heterogeneous nucleation taking place on the PA membrane. Recent studies (31, 32) on calcium carbonate crystal formation demonstrate that crystallization starts with the formation of prenucleation clusters that aggregate to form amorphous nanoparticles in the size range of 20-70 nm. These nanoparticles then assemble to form polycrystals in the size range of 100-500 nm, which develop crystalline domains after reaching a critical size. Note that this process is different from the classical nucleation theory, which considers that crystal formation occurs from a critical nucleus directly formed by the assembly of ions from solution (33). Although the above studies used calcium carbonate, it is indicated that the path of “prenucleation cluster - amorphous nanoparticle - polycrystal” is not restricted to calcium carbonate and is observed in other surface crystal formation (32).

FIGURE 6. Mechanisms controlling gypsum deposition on the surfaces of the CA and PA membranes. (a) and (b) are SEM images of the clean CA and PA membranes. (c) and (d) are SEM images of gypsum deposition on the CA and PA membranes after soaking the membranes in supersaturated gypsum solutions. The CA and PA membrane coupons are soaked for 24 h in 70 mM CaCl2, 40 mM Na2SO4, and 38 mM NaCl, with a gypsum saturation index (SI) of 3.1.

The new crystal formation theory can help to explain the stepwise disruption force curves observed in the force measurements for the PA membrane during crystal precipitation. It is known that the PA membrane is negatively charged due to carboxyl functional groups (34), which can form complexes with Ca2+ ions. These specific interactions between Ca2+ and the membrane surface result in increased Ca2+ concentration on the membrane surface, thereby initiating the formation of gypsum prenucleation clusters, and subsequently amorphous gypsum nanoparticles and polycrystals on the PA membrane surface (Figure 4b). The gypsum nanoparticles formed on the PA membrane surface can interfere with the AFM gypsum particle probe when the probe is pulled away from the membrane surface, causing the stepwise disruption feature observed in the adhesion force curves. In contrast, the predominant functional groups on the CA membrane surface are hydroxyls, which are neutral and do not have specific interactions with either Ca2+ or SO42- ions. Therefore, there is much lower probability for gypsum to form prenucleation clusters and amorphous nanoparticles directly on the CA membrane surfaces. The gypsum precipitation in this case is dominated by crystallization taking place in the bulk (illustrated in Figure 4a), which has negligible effect on the interaction forces between the gypsum probe and membrane surface. To further explain the effect of gypsum precipitation on the interaction forces between the gypsum particle probe and the membrane surface, we plotted the adhesion force and rupture distance as a function of time in Figures 5a and 5b, respectively. The magnitude of the adhesion force for both the CA and PA membranes varied within a relatively narrow range. However, the change of the rupture distance with time for both membranes differed markedly. While the rupture distance for the CA membrane remained relatively constant with time, the rupture distance for the PA membrane increased significantly after the onset of gypsum precipitation. We also integrated the area within the adhesion portion of the force-distance curve to determine the adhesion energy (or work of adhesion), which represents the overall work

required to pull the gypsum probe off the membrane surface. As shown in Figure 5c, the changes in the adhesion energy during gypsum precipitation are clearly different for the two membranes. Only small changes are observed for the CA membrane, indicating there is negligible surface crystallization. However, a drastic increase in the adhesion energy is observed for the PA membrane, pointing to a significant degree of surface crystallization. Surface Crystallization in the Absence of Membrane Permeate Flux. In order to further demonstrate that the gypsum scaling behaviors of the CA and PA membranes are dominated by different scaling mechanisms, we performed a gypsum scaling experiment in a batch system. CA and PA membrane coupons were soaked in supersaturated CaSO4 solution with an SI of 3.1. After soaking for 24 h, the two membranes were taken out and examined under SEM. Figure 6a and b illustrate the surfaces of the clean membranes, whereas Figure 6c and d are the images of the membrane surfaces after soaking in a supersaturated gypsum solution. Only a few scattered gypsum crystals are found on the CA membrane surface (Figure 6c). However, a large portion of the PA membrane surface is covered by large pieces of gypsum crystal clusters (Figure 6d). Since there is no permeate flux through the membrane in the batch scaling tests, the gypsum scale deposited on the PA membrane surface can only be a result of heterogeneous surface crystallization. The results again prove that the PA membrane has higher surface scaling potential than the CA membrane.

Acknowledgments We are grateful for the financial support received from the California Department of Water Resources under Award Number 4600007446.

Supporting Information Available SEM image of gypsum particle probe for AFM measurements (Figure S1). Duplicate runs of gypsum scaling and cleaning experiments demonstrating the reproducibility of our experiments (Figure S2). This material is available free of charge via the Internet at http://pubs.acs.org. VOL. 44, NO. 6, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2027

Literature Cited (1) Arras, W.; Ghaffour, N.; Hamou, A. Performance evaluation of BWRO desalination plantsA case study. Desalination 2009, 235 (1-3), 170–178. (2) Rieger, A.; Steinberger, P.; Pelz, W.; Haseneder, R.; Hartel, G. Mine water treatment by membrane filtration processes Experimental investigations on applicability. Desalination Water Treat.: Sci. and Eng. 2009, 6 (1-3), 54–60. (3) Yang, H.; Huang, C.; Pan, J. Characteristics of RO foulants in a brackish water desalination plant. Desalination 2008, 220 (13), 353–358. (4) Le Gouellec, Y.; Elimelech, M. Calcium sulfate (gypsum) scaling in nanofiltration of agricultural drainage water. J. Membr. Sci. 2002, 205 (1-2), 279–291. (5) Dydo, P.; Turek, M.; Ciba, J. Scaling analysis of nanofiltration systems fed with saturated calcium sulfate solutions in the presence of carbonate ions. Desalination 2003, 159 (3), 245– 251. (6) Le Gouellec, Y. A.; Elimelech, M. Control of calcium sulfate (gypsum) scale in nanofiltration of saline agricultural drainage water. Environ. Eng. Sci. 2002, 19 (6), 387–397. (7) Shih, W.; Gao, J.; Rahardianto, A.; Glater, J.; Cohen, Y.; Gabelich, C. Ranking of antiscalant performance for gypsum scale suppression in the presence of residual aluminum. Desalination 2006, 196 (1-3), 280–292. (8) Rahardianto, A.; Mccool, B.; Cohen, Y. Reverse osmosis desalting of inland brackish water of high gypsum scaling propensity: Kinetics and mitigation of membrane mineral scaling. Environ. Sci. Technol. 2008, 42 (12), 4292–4297. (9) Rahardianto, A.; Shih, W.; Lee, R.; Cohen, Y. Diagnostic characterization of gypsum scale formation and control in RO membrane desalination of brackish water. J. Membr. Sci. 2006, 279 (1-2), 655–668. (10) Lee, S.; Lee, C. Effect of operating conditions on CaSO4 scale formation mechanism in nanofiltration for water softening. Water Res. 2000, 34 (15), 3854–3866. (11) Jawor, A.; Hoek, E. Effects of feed water temperature on inorganic fouling of brackish water RO membranes. Desalination 2009, 235 (1-3), 44–57. (12) Rahardianto, A.; Gao, J.; Gabelich, C.; Williams, M.; Cohen, Y. High recovery membrane desalting of low-salinity brackish water: Integration of accelerated precipitation softening with membrane RO. J. Membr. Sci. 2007, 289 (1-2), 123– 137. (13) Gabelich, C.; Williams, M.; Rahardianto, A.; Franklin, J.; Cohen, Y. High-recovery reverse osmosis desalination using intermediate chemical demineralization. J. Membr. Sci. 2007, 301 (1-2), 131–141. (14) Shih, W.; Rahardianto, A.; Lee, R.; Cohen, Y. Morphometric characterization of calcium sulfate dihydrate (gypsum) scale on reverse osmosis membranes. J. Membr. Sci. 2005, 252 (1-2), 253–263. (15) Shirazi, S.; Lin, C.; Doshi, S.; Agarwal, S.; Rao, P. Comparison of fouling mechanism by CaSO4 and CaHPO4 on nanofiltration membranes. Sep. Sci. Technol. 2006, 41 (13), 2861– 2882. (16) Uchymiak, M.; Lyster, E.; Glater, J.; Cohen, Y. Kinetics of gypsum crystal growth on a reverse osmosis membrane. J. Membr. Sci. 2008, 314 (1-2), 163–172.

2028

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 6, 2010

(17) Cath, T.; Childress, A.; Elimelech, M. Forward osmosis: Principles, applications, and recent developments. J. Membr. Sci. 2006, 281 (1-2), 70–87. (18) Achilli, A.; Cath, T. Y.; Marchand, E. A.; Childress, A. E. The forward osmosis membrane bioreactor: a low fouling alternative to MBR processes. Desalination 2009, 239 (1-3), 10–21. (19) Cartinella, J. L.; Cath, T. Y.; Flynn, M. T.; Miller, G. C.; Hunter, K. W.; Childress, A. E. Removal of natural steroid hormones from wastewater using membrane contactor processes. Environ. Sci. Technol. 2006, 40 (23), 7381–7386. (20) Cath, T.; Gormly, S.; Beaudry, E.; Flynn, M.; Adams, V.; Childress, A. Membrane contactor processes for wastewater reclamation in space Part I. Direct osmotic concentration as pretreatment for reverse osmosis. J. Membr. Sci. 2005, 257 (1-2), 85–98. (21) Cath, T. Y.; Adams, D.; Childress, A. E. Membrane contactor processes for wastewater reclamation in space II. Combined direct osmosis, osmotic distillation, and membrane distillation for treatment of metabolic wastewater. J. Membr. Sci. 2005, 257 (1-2), 111–119. (22) Cornelissen, E.; Harmsen, D.; de Korte, K.; Ruiken, C.; Qin, J.; Oo, H.; Wessels, L. Membrane fouling and process performance of forward osmosis membranes on activated sludge. J. Membr. Sci. 2008, 319 (1-2), 158–168. (23) McCutcheon, J. R.; McGinnis, R. L.; Elimelech, M. A novel ammonia-carbon dioxide forward (direct) osmosis desalination process. Desalination 2005, 174 (1), 1–11. (24) McCutcheon, J. R.; McGinnis, R. L.; Elimelech, M. Desalination by ammonia-carbon dioxide forward osmosis: Influence of draw and feed solution concentrations on process performance. J. Membr. Sci. 2006, 278 (1-2), 114–123. (25) McGinnis, R. L.; Elimelech, M. Energy requirements of ammoniacarbon dioxide forward osmosis desalination. Desalination 2007, 207 (1-3), 370–382. (26) Martinetti, C.; Childress, A.; Cath, T. High recovery of concentrated RO brines using forward osmosis and membrane distillation. J. Membr. Sci. 2009, 331 (1-2), 31–39. (27) Mi, B.; Elimelech, M. Organic fouling of forward osmosis membranes: fouling reversibility and cleaning without chemical reagents. J. Membr. Sci. 2010, 348 (1-2), 337–345. (28) Holloway, R. W.; Childress, A. E.; Dennett, K. E.; Cath, T. Y. Forward osmosis for concentration of anaerobic digester centrate. Water Res. 2007, 41, 4005–4014. (29) McCutcheon, J. R.; Elimelech, M. Influence of concentrative and dilutive internal concentration polarization on flux behavior in forward osmosis. J. Membr. Sci. 2006, 284 (1-2), 237–247. (30) Mi, B.; Elimelech, M. Chemical and physical aspects of organic fouling of forward osmosis membranes. J. Membr. Sci. 2008, 320 (1-2), 292–302. (31) Gebauer, D.; Volkel, A.; Colfen, H. Stable prenucleation calcium carbonate clusters. Sci. 2008, 322 (5909), 1819–1822. (32) Pouget, E.; Bomans, P.; Goos, J.; Frederik, P.; de With, G.; Sommerdijk, N. The initial stages of template-controlled CaCO3 formation revealed by cryo-TEM. Sci. 2009, 323 (5920), 1555–1458. (33) Dydo, P.; Turek, M.; Ciba, J.; Wandachowicz, K.; Misztal, J. The nucleation kinetic aspects of gypsum nanofiltration membrane scaling. Desalination 2004, 164 (1), 41–52. (34) Childress, A.; Elimelech, M. Effect of solution chemistry on the surface charge of polymeric reverse osmosis and nanofiltration membranes. J. Membr. Sci. 1996, 119 (2), 253–268.

ES903623R