Modes of Natural Organic Matter Fouling during Ultrafiltration

GEORGES BELFORT* , †. Howard P. Isermann Department of Chemical Engineering and Department of Civil and Enviromental Engineering,. Rensselaer ...
0 downloads 0 Views 186KB Size
Download PDF
Environ. Sci. Technol. 2003, 37, 1676-1683

Modes of Natural Organic Matter Fouling during Ultrafiltration M A S A H I D E T A N I G U C H I , †,§ J A M E S E . K I L D U F F , * ,‡ A N D G E O R G E S B E L F O R T * ,† Howard P. Isermann Department of Chemical Engineering and Department of Civil and Enviromental Engineering, Rensselaer Polytechnic Institute, 110 8th Street, Troy, New York 12180-3190

The fouling of ultrafiltration membranes by natural organic matter (NOM), isolated from a potable surface water source, was studied with an emphasis on elucidating fouling modes and the role of aggregates. NOM size was related to membrane pore sizes using parallel membrane fractionation and size exclusion chromatography; such analyses confirmed the predominance of low MW species and identified the presence of aggregates in concentrated NOM solutions. Cake formation was the dominant mode of fouling by the unfiltered feed, which contained aggregates. This was identified by a constant rate of increase in membrane resistance with permeate throughput and was independent of pore size over a 10-1000 kDa molecular weight cutoff (MWCO) range. Prefiltration (to remove aggregates) and dilution (to reduce aggregate concentration) reduced the rate of increase in membrane resistance for the low MWCO membranes but did not change the fouling mode. In contrast, such pretreatment prevented cake formation on the larger MWCO membranes and shifted the mode of fouling to pore blockage. The data lend support for the idea that an initial fouling layer of large aggregates can catalyze the fouling by lower MW species. The fouling layer could be removed from the large MWCO membranes by backwashing, but the lower MWCO membranes exhibited some irreversible fouling, suggesting that low MW species penetrated into the pore structure. A combined pore blockage-cake formation model described the data well and provided insight into how fouling modes evolve during filtration.

Introduction While membrane separation processes have been widely used to recover valuable products from complex mixtures, only recently have membrane technologies emerged as viable for drinking water production (1-5) and for wastewater treatment (6-8). In many applications, process economics is governed by membrane fouling, which can significantly reduce membrane performance. Understanding fouling phenomena and developing ways to control them is critical for the development of water treatment technologies. Each * Corresponding authors phone: (518)276-2042; fax: (518)2763055; e-mail: [email protected] (J.E.K.) and phone: (518)276-6948; fax: (518)276-4030; e-mail: [email protected] (G.B.). † Howard P. Isermann Department of Chemical Engineering. ‡ Department of Civil and Enviromental Engineering. § Present address: Global Environment Research Labs, Toray Industries, Inc., Shiga, Japan. 1676

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 8, 2003

potential drinking water source (e.g., seawater, surface water, and groundwater) has characteristic properties that impact membrane filtration. For surface waters and some groundwaters, NOM is thought to be a primary contributor to membrane fouling, either acting alone or in concert with bound metals and particles (9-11). Factors potentially affecting membrane fouling by NOM include properties of the feed constituents such as size, hydrophobicity, charge density, and isoelectric point (1218); properties of the membrane including hydrophobicity, charge density, surface roughness, and porosity (19, 20); properties of the solution phase such as pH, ionic strength, and concentration of metals (13, 14, 16, 19, 21-25); and the hydrodynamics of the membrane system, characterized by solution flux and surface shear (14, 18, 22). The role of solution composition appears to be well established, with fouling exacerbated by conditions of low pH (13, 14, 24, 25), high ionic strength (14, 25, 26), and high divalent cation concentration (i.e., hardness ions) (14, 23, 26). Other factors, including the role of NOM polarity, are less well understood. For example, Carroll et al. (17) fractionated NOM by charge and polarity and found that the major contribution to fouling was from neutral, hydrophilic NOM components; however, others report that the most extensive membrane fouling was caused by hydrophobic NOM components (12). In some studies, fouling by NOM was found to be reversible (11) while in others fouling was irreversible (9). The lack of consistent findings is likely due to the complexity of NOM and its interaction with membrane surfaces. Flux decline caused by organic macromolecules may result from concentration polarization, from pore blockage by solute adsorbed on the membrane surface or within pores, and from the formation of a cake layer on the membrane surface, which presents a resistance to flow in addition to the membrane itself. Concentration polarization effects have been shown to be small in comparison with pore blockage and surface deposition (18, 26). Based on the complexity of NOM, including the wide distribution of molecular sizes, it is reasonable to expect fouling by both pore blockage and surface deposition (26). Of particular importance in this regard is the size distribution of feed solution components relative to the pore size distribution of the membrane. Components smaller than membrane pores can adsorb to surfaces and reduce the cross sectional area for flow, while larger components can block pore entrances and contribute to cake or gel formation. Recent research on humic acid fouling of microfiltration membranes has suggested that the convective transport and deposition of large aggregates on the membrane surface dominates the fouling by humic substances (24, 27). It was shown that initial fouling occurred as the result of aggregate deposition, which could catalyze subsequent fouling, even by feed solutions that had been prefiltered to remove aggregates. However, the role of large aggregates was shown to be less significant for smaller molecular-weight-cutoff (MWCO) UF membranes (26). Much of this work was done using Aldrich humic acid, which is known to be much larger, have a higher aromatic content, have a higher metal content, and have a lower acidic functional group content than aquatic humic substances, which are dominated by fulvic acids (28). While several studies have interpreted NOM fouling behavior in terms of the resistance-in-series model (e.g., 1618, 22), fewer studies have examined how different fouling modes evolve during filtration. Huang and Morrissey (28) found that fouling of microfiltration membranes by a complex mixture of both soluble and suspended protein solids was 10.1021/es020555p CCC: $25.00

 2003 American Chemical Society Published on Web 03/07/2003

dominated initially by pore blocking, followed by cake formation. This is consistent with other studies that have shown fouling to shift to a cake formation mode after pore blocking processes proceed to completion (29-31). The focus of the work described in this paper was to evaluate the relative importance of pore blockage and cake formation modes of fouling by NOM from a potable water source (Tomhannock Reservoir, Troy, NY), with attention to the role of aggregates. We have employed concentrated NOM solutions to simulate conditions that might exist in the retentate stream in latter stages of a membrane filtration plant, where the most severe fouling problems are anticipated. The role of aggregates was studied by comparing the filtration performance of unfiltered with that of prefiltered RO isolates (to remove aggregates) as was done by Yuan and Zydney (24, 26, 27). In addition, by employing diluted, unfiltered RO isolate, we examine the effect of aggregate concentration. The existence of aggregates was demonstrated by measuring the NOM molecular size distribution. These data were also used to support interpretations of fouling modes. Two different techniques were usedssize exclusion chromatography (SEC) and a membrane fractionation technique originally developed by Aiken (32). Ultrafiltration performance was measured in terms of solution flux, flux recovery after backwashing, and NOM rejection. Fouling was evaluated using a combined pore blockage-cake formation model and the fouling models of Hermans and Bredee (33).

Experimental Materials and Methods Membranes. Poly(ether sulfone) (PES) membranes were obtained from Pall Filtron Corp. (OMEGA lots 7099B, 8220B, 9140E, 7309A, 7265G, 9336J, 0073F, 0083B; respectively, East Hills, NY). A wide range of molecular weight cutoffs (MWCOs) was employed (10, 30, 50, 70, 100, 300, 500, and 1000 kDa). These OMEGA series membranes were weakly hydrophilized during manufacture by an undisclosed process. The membranes were used for both the NOM size distribution measurements and the fouling experiments. In addition, other PES membranes, with 0.2 and 0.45 µm nominal pore sizes, were purchased from Gelman Science Corp. (Supor lots 8760, 81739; respectively, Ann Arbor, MI), and a glassfiber filter with 1.3 µm nominal pore size was purchased from Whatman Inc. (GF/B Lot. 622497; Clifton, NJ). These were used for both the NOM size distribution tests and for NOM prefiltration. All the membranes were soaked in deionized (DI) water overnight to remove the wetting agent, washed several times with DI water, and soaked again in DI water overnight. Natural Organic Matter. Natural organic matter was isolated from the Tomhannock Reservoir, the drinking water supply for the City of Troy, NY, using a field reverse osmosis system designed after Serkiz and Perdue (34). The treatment and recovery scheme is shown in Figure 1. This source water had a dissolved organic carbon content of about 3 mg/L. Pretreatment with cartridge- and microfilters (1.0 µm and 0.2 µm) was applied to reduce turbidity, and a cation exchange softener was used to remove calcium and magnesium. Then, the NOM in the pretreated water was concentrated to slightly greater than 100 mg/L, using a reverse osmosis membrane (Fastek TLC S4040, Osmonics Inc., Minnetonka, MN) with a feed flow rate of 14 L/min, a retentate to permeate flow ratio of 7, and an applied pressure of 550 kPa (80 psig). This low pressure was used to maximize NOM rejection (>99%) and mass recovery (>95%). Prior to using RO isolate as a feed solution for the UF experiments, the pH was adjusted to 7.0 using hydrochloric acid, and the conductivity was adjusted to 5.3 mS/cm using sodium chloride. Some of the NOM solution was prefiltered with 0.45 µm PES membranes using a stirred filtration cell (Model 8050, Millipore Corp., Bedford, MA) under a pressure of 34 kPa (5 psig), which

FIGURE 1. Flow diagram of the pretreatment, recovery, and isolation of NOM from the Tomhannock Reservoir. removed about 5% of the initial organic carbon. The total organic carbon concentration of the feed solution was measured using a carbon analyzer (Oceanographic International, College Station, TX) calibrated externally with potassium hydrogen phthalate. Total organic carbon was correlated with UV absorbance at 254 nm, and routine determinations of NOM concentration were made spectroscopically (Hitachi U 2000 Double-Beam UV/vis spectrophotometer, Hitachi Instruments, Inc., Danbury, CT). Water. The DI water was produced from Troy City tap water using an in-house deionized water system consisting of (in order) reverse osmosis membranes (FT-30, FilmTech, MN), UV (254 nm) irradiation, and a Teflon microfilter. Chemicals. A mixture of dextrans was used to characterize the MWCO of the UF membranes (described below). The mixture included dextran 4 (Serva Biochemicals, Paramus, NJ) and T10, T40, T70, T500, and T2000 (Amersham Pharmacia, Piscataway, NJ). The size exclusion chromatography column was calibrated using dextran molecular weight standards (DXT1750k, DXT750k, DXT550k, DXT165k, DXT97k, DXT47k, DXT25k, DXT11k, and DXT3k, American Polymer Standards Co., Mentor, OH), maltodextrin (MW ) 1153 Da, Aldrich Chemical Co., Milwaukee, WI), and glucose (MW ) 180 Da, Aldrich Chemical Co., Milwaukee, WI). Size Exclusion Chromatography. Size exclusion chromatography (SEC) (Shodex Ohpak KB-806M polyhydroxymethacrylate gel, exclusion limit 0.50) was independent of the membrane pore size. This is consistent with the findings of Yuan and Zydney (26) who showed that the hydraulic resistance of deposited humic acids did not depend on UF membrane MWCO. The slower initial increase in resistance of the 300 and 500 kDa MWCO membranes appears to correlate with the small fraction of NOM components in this size range, as demonstrated in Figure 3. Examination of the data in Table 1 reveals that dilution of the feed solution by a factor of 2 reduced the rate of fouling by cake formation proportionally (dRT/dW ) 0.07 ( 0.01 × 106 m4‚Pa‚s/kg2) for the lower MWCO membranes (up to 100 kDa) but did not change the fouling mode. In contrast, dRT/ dW was not constant for the larger MWCO membranes, suggesting a different mode of fouling, which is addressed below. Prefiltration of the RO isolate, which removed large NOM aggregates (as confirmed by both size exclusion chromatography and parallel membrane fractionation), had no effect on the mode of fouling, and only a limited effect on its rate of development for the smaller MWCO membranes (again up to 100 kDa). Therefore, small MWCO membranes fouled 1680

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 8, 2003

FIGURE 5. The effect of membrane MWCO on the development of total resistance as a function of reduced time, t/tmax;tmax is the time to filter 27 kg/m2 NOM feed solution. Feed solutions: (a) 100 mg/L of RO isolate; (b) 50 mg/L of RO isolate; (c) 100 mg/L of prefiltered (0.45 µm) RO isolate. Membrane MWCOs: 10 kDa (9); 30 kDa ([); 100 kDa (2); 300 kDa (b); 500 kDa (0); 1000 kDa (]). Lines are fits of the combined pore blockage-cake formation model (eqs 4 and 5); parameters are reported in the Supporting Information. in a cake formation mode in the absence of aggregates. Inspection of Figure 4 reveals that even in the absence of aggregates, a significant fraction of the NOM solution contains species large enough to be rejected by these smaller MWCO membranes. This suggests that cake formation will be the predominant fouling mode in the presence of species significantly larger than the membrane pore size. Prefiltration of the RO isolate significantly reduced the rate of fouling of the 300 kDa membrane, and the fouling was not dominated by cake formation (Figure 5c). Especially

TABLE 1. Average Change in Differential Resistance per Mass of Permeate Filtered at Steady State

MWCO [kDa]

dRT/dW (untreated NOM) ×10-6 [m4‚ Pa‚s kg-2]

dRT/dW (diluted NOM) ×10-6 [m4‚ Pa‚s kg-2]

dRT/dW (prefiltered NOM) ×10-6 [m4‚ Pa‚s kg-2]

10 30 50 70 100 300 500 1000 mean 〈dRT/dW〉

0.16 ( 0.02 0.17 ( 0.01 0.16 ( 0.02 0.15 ( 0.02 0.16 ( 0.02 0.16 ( 0.02 0.16 ( 0.01 0.16 ( 0.01 0.16 ( 0.02

0.07 ( 0.01 0.07 ( 0.00 0.08 ( 0.01 0.07 ( 0.01 0.07 ( 0.01

0.11 ( 0.02 0.11 ( 0.03 0.10 ( 0.02 0.10 ( 0.01 0.11 ( 0.02

0.07 ( 0.01

0.11 ( 0.02

FIGURE 7. Progression of the filtration constant, n, during filtration of the 100 mg/L of RO isolate. Values plotted are the log slope of curves shown in Figure 6, calculated from the fitted combined pore blockage-cake formation model (eqs 4 and 5); parameters are reported in the Supporting Information). Values of dt/dW between the two data points shown on each curve represent the range of experimental data. Membrane MWCOs: 10 kDa (9); 30 kDa ([); 300 kDa (b); 500 kDa (0).

FIGURE 6. Representative filtration data for 100 mg/L of RO isolate, plotted according to eq 3. Lines are derivatives calculated from the fitted combined pore blockage-cake formation model (eqs 4 and 5); parameters are reported in the Supporting Information). The log slope of the plot yields the filtration constant, n. Membrane MWCOs: 30 kDa ([); 500 kDa (0). dramatic was the performance of the larger MWCO membranes (500 and 1000 kDa), for which dRT/dW was very small. These findings are consistent with the data presented by Yuan and Zydney, who showed that prefiltration (0.16 µm) of an Aldrich humic acid solution did little to mitigate fouling of a 30 kDa UF membrane, while such prefiltration significantly reduced the flux reduction exhibited by a 300 kDa membrane. Additional insights into modes of fouling can be gained by plotting data on the coordinates suggested by eq 3. The slope of d2t/dW2 versus dt/dW on log-log coordinates yields the filtration constant, n. While derivatives calculated from experimental data are subject to significant variability, such derivatives were readily computed from fits of the combined pore blockage-cake formation model using finite differences over one-second time intervals. Such a plot is shown in Figure 6 for filtration of the 100 mg/L unfiltered RO isolate through the 30 and 500 kDa MWCO membranes. To clarify the presentation, arithmetic coordinates were chosen and two representative data sets were plotted. The rate of change of dt/dW plotted on the y-axis increases steeply early in the filtration run, passes through a maximum, and finally reaches a constant value. The progression of the filtration constant, n, obtained as the log slope of the model fits shown in Figure 6, is shown in Figure 7 for several membranes. Data were plotted versus dt/dW to facilitate comparisons between Figures 6 and 7. Values of dt/dW between the two data points shown on each curve represent the range of experimental data. It is clear that pore blockage (n ) 2) dominates early in the filtration run, while cake formation (n ) 0) controls later in the run, consistent with the plateau in Figure 6. The maximum in Figure 6, corresponding to the minimum in

FIGURE 8. Representative filtration data for 50 mg/L of RO isolate plotted on the coordinates of eq 3. Lines are derivatives calculated from the fitted combined pore blockage-cake formation model (eqs 4 and 5); parameters are reported in the Supporting Information). The log slope of the plot yields the filtration constant, n. Membrane MWCOs: 30 kDa ([); 100 kDa (2); 1000 kDa (]). Figure 7, marks the transition from pore blockage to cake formation (30). The mode of fouling shifted to cake formation for membranes of all MWCO values during filtration of the 100 mg/L unfiltered RO isolate. Membranes having a MWCO >100 kDa exhibited a distinct pore blockage regime (data for the 500 kDa MWCO membrane shown in Figure 6 provides an example) prior to cake formation, while the smaller pore size membranes exhibited a rapid transition to cake formation; i.e., no experimental data were collected before the transition. The rapid shift to cake formation is exemplified by the data for the 30 kDa MWCO membrane shown in Figure 6 and the steep initial slope of the n vs dt/dW curve for the 10 and 30 kDa MWCO membranes in Figure 7. As shown in Figure 8, the mode of fouling did not shift fully to cake formation for membranes having a MWCO >100 kDa during filtration of the 50 mg/L unfiltered RO isolate, in contrast to the 100 mg/L unfiltered RO isolate. However, the transition regime between pore blockage and cake formation was reached, yielding negative values of the filtration constant, n, at the end of the filtration run. The final n-values reached after filtration of ∼27 kg/m2 of NOM solution are plotted in Figure 9. The mode of fouling for membranes having smaller MWCO values shifted from pore blockage to VOL. 37, NO. 8, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1681

aggregates could effectively “catalyze” fouling by smaller molecular weight species. In the absence of aggregates, the smaller NOM species penetrated into the membrane pore structure, causing pore blockage instead of forming a cake layer on the surface. The observation that dilution of the NOM feed (and hence dilution of aggregates) prevented a full transition to cake formation suggests two possibilities. First, the extent of aggregation may have been lowered as a result of dilution, either by reversing aggregate formation or slowing the rate of formation. Second, the ability of aggregates to catalyze fouling may depend on their surface density, which was likely lower by at least a factor of 2 for the diluted feed.

FIGURE 9. The final value of the filtration constant, n, reached after filtration of ∼27 kg/m2 of NOM solution. A value of zero indicates cake formation; positive values indicate pore blockage; negative values indicate a transition from pore blockage to cake formation. NOM solutions: 100 mg/L of RO isolate (b); 50 mg/L of RO isolate ([); 100 mg/L of prefiltered (0.45 µm) RO isolate (9). cake formation, as for the 100 mg/L unfiltered feed; however, the transition region was now observed experimentally. As shown in Figure 5c, prefiltration of the 100 mg/L RO isolate had a dramatic effect on the development of resistance and the modes of fouling for membranes having a MWCO >100 kDa. A shown in Figure 9, the value of the filtration constant, n, reached after filtration of ∼27 kg/m2 of NOM solution was greater than 1.4 for these membranes, signifying fouling by a pore blockage mode onlysno transition to cake formation was observed. The shift in fouling mode from cake formation in the presence of NOM aggregates to pore blockage upon either dilution or prefiltration (and subsequent removal of aggregates) suggests that even though such aggregates comprised a small proportion of the NOM mass, they were responsible for the cake formation on the larger membranes. This supports the findings of Yuan and Zydney, who proposed that an initial fouling layer consisting of large

The reversibility of NOM fouling (by backwashing) is shown in Figure 10. The reversible resistance (removed by physical cleaning) after filtration of the 100 mg/L unfiltered RO isolate was independent of membrane MWCO. This supports an interpretation of the reversible resistance as that contributed by the formation of a cake layer on the membrane surface and is consistent with the ability of aggregates to catalyze fouling by cake formation. The reversible resistance after filtration of the 50 mg/L unfiltered RO isolate was essentially the same for membranes having MWCOs less than or equal to 100 kDa but decreased significantly for the larger pore size membranes. This correlates with the mode of fouling, which did not shift fully to cake formation for the larger membranes. Similar trends were observed for the reversible component of resistance after filtration of the prefiltered 100 mg/L RO isolatesremoval of aggregates prevented cake formation and either significantly reduced (300 kDa membrane) or essentially eliminated (500 and 1000 kDa) reversible fouling. Irreversible fouling was observed with smaller MWCO membranes (MWCO up to 100 kDa), a finding consistent with that of Cho et al. (18) for filtration of prefiltered surface waters with a 10 kDa PES UF membrane. However, irreversible fouling was virtually absent for the larger MWCO membranes. This is consistent with the measured size distribution of NOMsmost of the NOM was between 5 and 100 kDa (as protein) with a small amount present as large aggregates and few components in the 300-500 kDa size

FIGURE 10. Comparison of reversible (removed by backwashing) and irreversible resistances after filtration: 100 mg/L of RO isolate: downward dash (reversible), upward dash (irreversible); 50 mg/L of RO isolate: downward diagonal (reversible), upward diagonal (irreversible); and 100 mg/L of prefiltered (0.45 µm) RO isolate: solid bar (reversible), unfilled bar (irreversible). 1682

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 8, 2003

range. Smaller NOM components were either able to pass through the larger membranes (300-1000 kDa) or were easily removed by backwashing. Where present, the aggregates were rejected and contributed to reversible cake formation but did not cause irreversible fouling. In contrast, the NOM solution contained some components that were about the same size as the pore size of the 10 through 100 kDa membranes. These NOM components contributed to pore blockage and were difficult to remove, perhaps due to steric hindrance. However, pore blockage was not a dominant fouling mode, and the irreversible resistance was a relatively small fraction of the total resistance. Irreversible fouling was reduced by both dilution and prefiltration of the feed. Apparently, prefiltration to remove aggregates also removed some of the more reactive species that contributed to irreversible fouling.

Acknowledgments We appreciate John Pieracci’s help with the SEC analysis. Jonathan Dordick and Steven Cramer are thanked for allowing us to use their respective HPLC systems. Andrew Wigton and Supatpong Mattaraj isolated and purified the NOM. Pall Filtron Corporation donated the PES membranes, and Millipore Corporation donated the SEC column. We also acknowledge the support of Toray Industries Inc., Shiga, Japan, and the Eastman Kodak Company. Grants to G. Belfort from the U.S. Department of Energy (Grant No. DE-FG0290ER14114) and the National Science Foundation (Grant No. CTS-94-00610) and to J. Kilduff from the National Science Foundation (Grants BES-9871241 and BES-99-84709) are gratefully acknowledged.

Supporting Information Available Molecular size evaluation of NOM, combined pore blockagecake formation model, and model parameters and sensitivity (Table S1). This material is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) AWWA Membrane Technology Research Committee J. Am. Water Works Assoc. 1992, 84 (Jan), 59-67. (2) Adham, S. S.; Jacangelo, J. G.; Laine, J. M. J. Am. Water Works Assoc. 1996, 88 (May) 22-31. (3) Glucina, K.; Alvalez, A.; Laıˆne´, J. M. Desalination 2000, 132, 73-82. (4) Wright, A. G. Eng. News-Rec. 1998, 241(11), 34. (5) Sethi, S.; Juby, G.; Schuler, P.; Holmes, L. Proc. AWWA Annual Conference; Washington, DC, June 2001.

(6) Stephenson, T.; Brindle, K.; Judd, S.; Jefferson, B. Membrane Bioreactors for Wastewater Treatment; IWA Publishing: London, UK, 2000. (7) Coˆte´, P.; Buisson, H.; Praderie, M. Water Sci. Technol. 1998, 38(4-5), 437-442. (8) Churchouse, S.; Wildgoose, D. Membr. Technol. 1999, 111, 4-8. (9) Lahoussine-Turcaud, V.; Wiesner, M. R.; Bottero, J. Y. J. Membr. Sci. 1990, 52, 173-190. (10) Kaiya, Y.; Itoh, Y.; Fujita, K.; Takizawa, S. Desalination 1996, 106, 71-77. (11) Nystro¨m, M.; Ruohomaki, K.; Kaipia, L. Desalination 1996, 106, 79-87. (12) Nilson, J. A.; DiGiano, F. A. J. Am. Water Works Assoc. 1996, 88 (May), 53-66. (13) Jucker, C.; Clark, M. M. J. Membr. Sci. 1994, 97, 37-52. (14) Hong, S.; Elimelech, M. J. Membr. Sci. 1997, 132, 159-181. (15) Lin, C.; Lin, T.; Hao, O. J. Water Res. 2000, 34, 1097-1106. (16) Cho, J.; Amy, G.; Pellegrino, J. J. Membr. Sci. 2000, 164, 89-110. (17) Carroll, T.; King, S.; Gray, S. R.; Bolto, B. A.; Booker, N. A. Water Res. 2000, 34, 2861-2868. (18) Cho, J.; Amy, G.; Pellegrino, J. Water Res. 1999, 33, 2517-2526. (19) Combe, C.; Molis, E.; Lucas, P.; Riley, R. and Clark, M. M. J. Membr. Sci. 1999, 154, 73-87. (20) Kilduff, J. E.; Mattaraj, S.; Sensibaugh, J.; Pieracci, J. P.; Belfort, G. Desalination 2000, 132, 133-142. (21) Braghetta, A.; DiGiano, F. A.; Ball, W. P. J. Environ. Eng. 1997, 123, 628-641. (22) Braghetta, A.; DiGiano, F. A.; Ball, W. P. J. Environ. Eng. 1998, 124, 1087-1098. (23) Yoon, S.; Lee, C.; Kim, K.; Fane, A. G. Water Res. 1998, 32, 21802186. (24) Yuan, W.; Zydney, A. L. Desalination 1999, 122, 63-76. (25) Jones, K. L.; O′Melia, C. R. J. Membr. Sci. 2000, 165, 31-46. (26) Yuan W.; Zydney, A. W. Environ. Sci. Technol. 2000, 34, 50435050. (27) Yuan, W. Zydney, A. L. J. Membr. Sci. 1999, 157, 1-12. (28) Huang, L.; Morrissey, M. T. J. Membr. Sci. 1998, 144, 113-123. (29) Bowen, W. R.; Calvo, J. I.; Hernandez, A. J. Membr. Sci. 1995, 101, 153-165. (30) Ho, C. C.; Zydney, A. L. J. Colloid Interface Sci. 2000, 232, 389399. (31) Lu, W. M.; Tung, K. L.; Hwang, K. J. Chem. Eng. Sci. 1997, 52, 1743-1756. (32) Aiken, G. R. Environ. Sci. Technol. 1984, 18, 978-981. (33) Hermans, P. H.; Bredee, H. L. J. Soc. Chem. Ind. 1936, 55, 1T-4T. (34) Serkiz, S. M.; Perdue, E. M. Water Res. 1990, 24, 911-916. (35) Tkacik, G.; Michaels, S. Bio/Technology 1991, 9, 941-946.

Received for review January 25, 2002. Revised manuscript received December 18, 2002. Accepted January 22, 2003. ES020555P

VOL. 37, NO. 8, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

1683