Fouling of Microfiltration and Ultrafiltration ... - ACS Publications

Jul 17, 2002 - Membrane filtration (microfiltration and ultrafiltration) has become an accepted process for drinking water treatment, but membrane fou...
6 downloads 0 Views 97KB Size
Environ. Sci. Technol. 2002, 36, 3571-3576

Fouling of Microfiltration and Ultrafiltration Membranes by Natural Waters KERRY J. HOWE* AND MARK M. CLARK Department of Civil and Environmental Engineering, MC-250, University of Illinois, Urbana, Illinois 61801-2352

Membrane filtration (microfiltration and ultrafiltration) has become an accepted process for drinking water treatment, but membrane fouling remains a significant problem. The objective of this study was to systematically investigate the mechanisms and components in natural waters that contribute to fouling. Natural waters from five sources were filtered in a benchtop filtration system. A sequential filtration process was used in most experiments. The first filtration steps removed specific components from the water, and the latter filtration steps investigated membrane fouling by the remaining components. Particulate matter (larger than 0.45 µm) was relatively unimportant in fouling as compared to dissolved matter. Very small colloids, ranging from about 3-20 nm in diameter, appeared to be important membrane foulants based on this experimental protocol. The colloidal foulants included both inorganic and organic matter, but the greatest fraction of material was organic. When the colloidal fraction of material was removed, the remaining dissolved organic matter (DOM), which was smaller than about 3 nm and included about 85-90% of the total DOM, caused very little fouling. Thus, although other studies have identified DOM as a major foulant during filtration of natural waters, this work shows that a small fraction of DOM may be responsible for fouling. Adsorption was demonstrated to be an important mechanism for fouling by colloids.

Introduction Membrane filtration has become an accepted process in water treatment. Many full-scale facilities are operating with either microfiltration (MF) or ultrafiltration (UF) membranes. One of the most significant issues affecting the development of membrane filtration is fouling (1). Interactions between the membrane and components in the raw water cause a rapid and often irreversible loss of flux through the membrane. Typical microfiltration plants operate with a flux of around 100 L/(m2‚h), less than 5% of the hydraulic permeability observed in laboratory filtration of deionized organic-free water through similar membranes. Many studies suggest that natural organic matter (NOM) is the most important foulant. With respect to fouling, researchers have studied the importance of membrane properties such as hydrophobicity, charge, and morphology (2-4); solution properties such as pH, ionic strength, and calcium concentration (3, 5, 6); and organic matter properties such as hydrophobicity, molecular * Corresponding author present address: Dept. of Civil Engineering, Tapy Hall, University of New Mexico, Albuquerque, NM 87131-1351; phone: (505)277-2722; fax: (505)277-1988; e-mail: [email protected]. 10.1021/es025587r CCC: $22.00 Published on Web 07/17/2002

 2002 American Chemical Society

weight, and charge density (3, 7-15). Some researchers have noted that both inorganic and organic compounds can contribute to fouling (15). Molecular size, which is indirectly related to molecular weight (MW), affects solubility and hydrophobicity; large molecules tend to be less soluble than small ones with similar functionality (16). Membrane surfaces would be expected, therefore, to preferentially adsorb higher-MW organics from solution. Lin and co-workers (7, 8) fractionated humic substances according to MW with gel permeation chromatography and hydrophobicity with resins. They found that the highest-MW components for both hydrophobic and hydrophilic fractions caused the greatest flux decline. Similarly, Yuan and Zydney (9, 10) filtered soil-derived humic acid solutions through 0.16-µm and 100 000-Da membranes and found that the water fractionated through the smallerpore membrane caused less fouling of a subsequent membrane. A similar result, less fouling after fractionation through small-pore membranes, was observed by Fan et al. (11) using natural surface water and membranes ranging from 0.45 µm to 10 000 Da. They proposed a mechanism for membrane fouling which involved a combination of adsorption of small molecules on the membrane pore wall and pore blockage by colloidal (>30 000 Da) organics. Habarou et al. (12) separated NOM by molecular weight using a 3500-Da dialysis bag. The material retained by the dialysis bag caused greater fouling than the material that passed through. The authors attributed the fouling to bacterial peptidoglycan residues. Each of these studies showed that high-MW components caused more fouling than smaller materials but did not identify an upper limit to the size of important membrane foulants. An opposing view was provided by Carroll et al. (13), who concluded that the greatest degree of fouling was by smallerMW molecules. They found greater fouling by the neutral hydrophilic fraction of dissolved organic matter (DOM) and characterized each fraction of DOM by size exclusion chromatography. The neutral hydrophilic fraction had the smallest MW distribution. The new research reported here elucidates significant aspects of membrane fouling by natural waters not previously demonstrated. First, this research successfully separated fouling components from nonfouling components in several natural waters. The importance of colloidal material as a membrane foulant is demonstrated. The most significant membrane fouling occurs in the presence of small colloidal matter, ranging from about 3-20 nm in diameter. This research demonstrated that the majority of DOM, by itself, does not cause membrane fouling; the actual foulant is a relatively small fraction of bulk DOM. Finally, it is shown that membrane foulants can include both inorganic and organic components, both of which must be considered in an overall understanding of fouling.

Experimental Methods and Materials Natural surface waters were filtered with laboratory-scale equipment. Bulk water samples were collected from three rivers and two lakes, which were selected to represent a geographic distribution across the United States. The waters had moderate levels of turbidity (10-35 NTU) and dissolved organic carbon (DOC) (2-4 mg/L) but varied significantly in hardness and alkalinity. The water was shipped to the University of Illinois and stored in a dark cold (4 °C) room before the experiments. Most of the experiments presented in this paper involved several filtration steps in series. The first filtration steps removed specific components from the water, and the latter VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3571

filtration steps investigated membrane fouling by the remaining components. Membrane filtration does not physically or chemically alter components that permeate the membrane. Thus, fouling by specific components after prefiltration is consistent with fouling by those components in the natural water, assuming that there are no interactions between permeating and retained components. Membrane filtration was conducted in unstirred deadend filtration cells (Amicon Corp., Beverly, MA) using flat sheet membrane material. New membranes were used for each test. Constant-pressure filtration was maintained by gas pressure regulated from a nitrogen cylinder. Pressure was measured with a digital pressure gage that had a resolution of 0.007 bar (0.1 psi). Membrane flux was determined by weighing the permeate on a top-loading balance at timed intervals with computerized data acquisition. All water samples were allowed to reach room temperature (22-24 °C) prior to filtration, and water temperature was measured to the nearest 0.1 °C at the beginning of each test. All flux values were corrected to a standard temperature (25 °C) and pressure using the following equation, adapted from American Society for Testing and Materials (ASTM) (17):

JS ) JM

( )

∆PS × 1.024(TS-TM) ∆PM

(1)

where J ) flux, ∆P ) transmembrane pressure, T ) temperature, and the subscripts M and S refer to measured and standard conditions, respectively. Clean-membrane permeability was determined by filtering reagent-grade water (resistivity > 10 MΩ‚cm, DOC < 0.2 mg/L) until the flux was constant (30 min, minimum). Flux decline was monitored during filtration of a predetermined volume of sample (typically, 400-900 mL, depending on experimental conditions). An external reservoir was connected between the pressure source and the filtration cell to increase sample volume. Separate reservoirs were used for samples and reagent-grade water to avoid contamination of the reagentgrade reservoir. Flux was normalized by dividing instantaneous measurements by the clean-water flux. Normalized flux is plotted as a function of the specific filtration volume (volume of permeate per unit of membrane area). A variety of membranes was used, depending on the objectives of specific experiments. The membranes included 0.2-0.6-µm glass-fiber (AP15, Millipore Corporation, Bedford, MA); 0.2-µm polypropylene, 0.2-µm polyethersulfone, 20 000Da polyethersulfone (Osmonics, Vista, CA); 100 000-Da cellulose acetate (Ondeo, Paris, France); and 3000-, 10 000-, 30 000-, and 100 000-Da regenerated cellulose (YM series, Amicon Corp., Beverly, MA) membranes. A number of experiments used a three-step filtration process, where the first step was prefiltration using a glassfiber filter to remove particulate matter. Glass-fiber prefilters were used in this research because the greater depth-filtration capabilities allow for a greater volume of water to be filtered without clogging that could change the size retention characteristics, and because glass-fiber exhibits low adsorbing and leaching characteristics. Quality control tests were conducted to verify that this prefiltration step did not have a measurable impact on the characterization of DOM, including the measurement of DOC concentration, UV254 adsorption, specific UV absorbance (SUVA), molecular weight distribution by high-performance size exclusion chromatography, and hydrophobic fraction by XAD resin adsorption. X-ray photoelectron spectrometry (XPS) was used to determine the elemental composition of clean and fouled membrane surfaces. XPS was performed with a Kratos Axis ULTRA imaging X-ray photoelectron spectrometer (Kratos Analytical Inc., Chestnut Ridge, NY). Small samples of fouled 3572

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002

FIGURE 1. Permeate flux of Lake Decatur water through 20 000-Da polyethersulfone membranes with and without prefiltration. Prefiltered water shows fouling by dissolved matter; raw water shows fouling by dissolved plus particulate matter (T ) 25 °C, P ) 1 bar). membranes (about 5 × 5 mm) were cut and placed in the sample chamber. The binding energies were used to determine the elemental composition of the materials. Standard methods (18) were used for all chemical analyses, when available. The concentration of DOM was measured as DOC using a Dorhmann Phoenix 8000 total organic carbon analyzer after filtration through a glass-fiber filter. Additional details of the experimental methods are provided elsewhere (19).

Results and Discussion Fouling Contribution by Particulate versus Dissolved Components. Material in natural waters is routinely resolved into particulate and dissolved components, where dissolved material is operationally defined as that which passes through a 0.45-µm filter (the dissolved fraction includes both colloidal and truly dissolved components). Simply speaking, fouling by particulate and dissolved matter can be considered to occur by different mechanisms. Particulate matter, which is clearly larger than the pores in commercial MF and UF membranes, forms a cake at the membrane surface. Dissolved matter, some of which can penetrate pores, causes fouling by a variety of mechanisms, such as forming a surface cake, penetrating and clogging pores, or adsorbing within pores to reduce the pore diameter. The relative contribution to membrane fouling by particulate and dissolved matter was assessed by conducting sets of parallel filtration experiments. The raw water was filtered directly in one test and was prefiltered through a glass-fiber filter in a parallel test. Particulate matter was removed during the prefiltration step, so a comparison of the flux decline in the two tests identifies the relative contribution of fouling by dissolved and particulate matter. An example of the results from these experiments is shown in Figure 1. Raw and prefiltered solutions both fouled 20 000Da polyethersulfone UF membranes rapidly, but the extent of fouling was slightly worse for the solution containing particulate matter. The flux decline during each experiment is calculated from initial and final flux values

φ)1-

JF JO

(2)

where φ is the flux decline and the subscripts O and F refer to initial and final values, respectively. After filtration of 150 L/m2 of water, the water without particulate matter caused a 66% flux decline, and the water with particulate matter

TABLE 1. Contribution of Particulate Matter to Flux Decline by Natural Waters source water Lake Decatur Medina River Ohio River Beaver Lake

membrane materiala

flux decline caused by raw water (%)

flux decline caused by prefiltered water (%)

contribution to flux decline from particulate matter (%)

CA PES PP CA PES PP CA PES PP CA PES PP

59 70 97 39 42 68 47 54 85 64 49 92

56 66 94 38 49 53 38 40 72 41 48 90

5 6 3 1 -22 19 27 15 36 2 2 13

average a

CA ) 100 000-Da cellulose acetate; PES ) 20 000-Da polyethersulfone; PP ) 0.2-µm polypropylene.

caused a 70% flux decline. The difference between these two is the additional fouling due to the presence of particulate matter. When expressed as percent of the total fouling that occurred, 94% of the flux decline was due to dissolved matter and 6% was due to particulate matter. Similar experiments conducted with other source waters and membrane materials, using a matrix of four natural waters and three membrane materials, are summarized in Table 1. In each experiment, the majority of the flux decline was due to dissolved matter. The contribution of particulate matter to the flux decline was between 1% and 36%; the average contribution was 13%. Variations may be due to differences in the amount and characteristics of the particulate matter in the various waters as well as differences in interactions between particulate matter and membrane materials. The evidence that particulate matter causes relatively little fouling has implications for full-scale installations treating water with similar quality. As noted earlier, the mechanism for fouling by particulate matter is cake formation. Full-scale membrane filters are vigorously backwashed every 30-90 min, and this backwashing is generally effective for removing surface cakes but less effective at removing material adsorbed to the membrane or lodged in pores. Thus, particulate fouling accumulates during the filtration cycle and is removed during the backwash cycle; there is little long-term accumulation of particulate matter. Fouling by dissolved matter, on the other hand, may be caused by mechanisms that continue to accumulate over time. These experiments utilized specific filtration volumes of 150-300 L/m2, as compared to typical volumes of about 100 L/m2 (between backwashes) in fullscale installations. As a result, the extent of particulate fouling in full-scale filtration may be even less than in these experiments. Accordingly, the rest of the experiments presented in this manuscript focus on fouling by dissolved components. Fouling by Colloidal versus Molecular Materials. Experiments quantified the size of dissolved matter causing fouling. Water from Beaver Lake, AR, was subjected to a threestep sequential filtration process (prefiltration, fractionation, fouling determination). First, raw water was prefiltered through glass-fiber filters to remove particulate matter. Next, aliquots of glass-fiber filter permeate were fractionated through regenerated cellulose (RC) membranes with nominal molecular weight cutoffs (MWCO) of 3000, 10 000, 30 000, and 100 000 Da, while one aliquot remained unfractionated. Regenerated cellulose exhibits low protein binding characteristics and is unlikely to retain material by adsorption or mechanisms other than size exclusion (20). Surface cake formation was minimized by removing particulate matter in the prefiltration step.

FIGURE 2. Permeate flux of Beaver Lake water through 0.2-µm polypropylene membranes after fractionation through regenerated cellulose membranes with various pore sizes (T ) 25 °C, P ) 0.69 bar). Next, the permeate from each RC membrane was filtered through an unused 0.2-µm polypropylene membrane, and the flux was monitored to determine the fouling caused by each aliquot. A new membrane was used with each aliquot. The results of these experiments are shown in Figure 2. The pore size of the RC membranes had a dramatic effect on the fouling of the subsequent polypropylene fouling. Unfractionated feedwater caused a 90% flux decline, but the flux only declined 15% when the feedwater was fractionated through the 3000-Da RC membrane. Using an analysis similar to that for Figure 1, the contribution of various components to membrane fouling can be determined. The material that passed through the 3000-Da membrane contributed 15% to the fouling, but the material that passed through the 100 000-Da membrane contributed to 65% of the fouling. This means that most of the foulants were able to pass through a 100 000-Da membrane but were not able to pass through a 3000-Da membrane. Less than 35% of the fouling was caused by VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3573

TABLE 2. Correlation between MWCO and Pore Size Based on Polyethylene Glycol

a

FIGURE 3. Permeate flux of Medina River water through 20 000-Da polyethersulfone membranes with and without fractionation through a 100 000-Da CA membrane (T ) 25 °C, P ) 1 bar). material retained by the 100 000-Da membrane, and less than 15% of the fouling was caused by material that was able to pass through the 3000-Da membrane. A similar three-step filtration experiment was conducted with water from the Medina River, TX. Water was prefiltered through a glass-fiber filter to remove particulate matter, one aliquot of the permeate was fractionated through a 100 000Da cellulose acetate membrane while a second aliquot remained unfractionated, and fouling was evaluated by filtration through 20 000-Da polyethersulfone membranes. The results shown in Figure 3 corroborate an important result from Beaver Lake water: a significant fraction of the foulants passed through the 100 000-Da membranes. Over 75% of the fouling was due to material smaller than 100 000 Da in the Medina River water. The identification of the MW of membrane foulants can be used to estimate their size. Molecular weight varies linearly with volume (assuming constant density), so diameter is a function of MW raised to a power

d ) β(MW)n

(3)

where d is the hydrodynamic diameter of a molecule (or particle), MW is the molecular weight, and β is a proportionality constant. The exponent n varies from 0.33 for spherical particles to nearly 1.0 for one-dimensional objects (such as linear long-chain polymers). Membrane MWCO values are determined by filtration of polymers such as polyethylene glycol (PEG). Lentsch et al. (21) determined the following relationship relating the hydrodynamic diameter and molecular weight of PEG:

d ) 0.09(MW)0.44

(4)

where d is in nanometers and MW is in daltons. The relationship between MWCO and pore size varies with the chemical selected for filtration because of steric, electrostatic, and chemical differences. Other methods for determining pore size also produce varying results (22). Nevertheless, pore size can be semiquantitatively estimated from MWCO. The approximate pore size for the RC membranes calculated from eq 4 is shown in Table 2. The pore sizes range from 3 nm for the 3000-Da membrane to 14 nm for the 100 000-Da membrane. Other research suggests these estimates are realistic. Kim et al. (22) measured pore sizes ranging from 4.4 to 13.4 nm for the YM30 membrane using three different analytical methods and measured a pore size of 13.3 nm for the YM100 membrane. The manufacturer of the cellulose acetate membrane used in Figure 3 estimates the pore size 3574

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002

membrane

MWCO (Da)

pore sizea (nm)

YM3 YM10 YM30 YM100

3000 10 000 30 000 100 000

3 5 8 14

Calculated from eq 4.

to be 10 nm. These pore sizes compare favorably with estimates for other membrane materials (20). The upper and lower boundaries of 100 000 and 3000 Da for a significant fraction of the foulants suggest that most of the foulants exist in a narrow size distribution. The upper size limit of material passing through a 100 000-Da membrane is probably less than 20 nm, allowing for some ambiguity between membrane MWCO and pore size. This upper limit applies to 65-75% of membrane foulants in the experiments shown in Figures 2 and 3. Similarly, the lower size limit for the majority of membrane foulants may be around 3 nm. Material ranging from about 3 nm to 20 nm in diameter can be described as very small colloids, whereas smaller material can be considered to be truly dissolved (23, 24). Material in the colloidal size range has been implicated in membrane fouling in other applications. Membranes are used in the food processing and pharmaceutical industries for separation or purification of solutions containing macromolecules such as proteins. Research in these application areas has demonstrated that proteins can cause rapid flux decline of MF membranes, even though proteins can be an order of magnitude smaller than the pore size of the membrane (20, 25). A common protein used in research, bovine serum albumin, has a MW of about 70 000 Da and is therefore in the same size range as the colloidal material that caused the greatest fouling in this research. Overall, this research suggests some similarities between membrane fouling by natural waters and solutions containing protein. Fouling by Dissolved Organic Matter. DOC concentration was monitored during the experiments shown in Figure 2. After prefiltration through the glass-fiber filter, the water contained 2.13 mg/L of DOC. Rejection of DOM through the RC membranes was calculated as

R)1-

CF CO

(5)

where R is rejection and C is concentration. Rejection of DOM by the RC membranes was low: 6% for the YM100 membrane, 15% for the YM10 membrane, and 9% for the YM3 membrane. The small difference in rejection between the YM100 and YM3 membranes suggests that very little DOM is in the colloidal size range. A total of 85-90% of the DOM in this water remained in the permeate from the YM3 and YM10 membranes, yet caused very little fouling after the colloidal matter was removed. These results demonstrate that the majority (85-90%) of DOM does not, by itself, cause fouling; the greatest fouling occurs when the remaining 1015% is present. Previous literature has demonstrated that colloidal DOM is a small fraction of total DOM. Schnoor et al. (26) found that 90% of DOM in a sample of Iowa River water was less than 3000 Da. Nilson and DiGiano (27) found that 80% of NOM in a sample of Tar River, NC, water was less than 3000 Da. Thurman (23) summarized MW data for humic and fulvic acids (the main components of DOM) from a large number of studies, which used four different analytical methods, and concluded that aquatic fulvic acid has a MW distribution between 500 and 2000 Da and that aquatic humic

acid has an average MW distribution between 2000 and 5000 Da. Thus, it appears that a relatively small fraction of DOM is in the size range that is most responsible for fouling. Adsorption as a Fouling Mechanism. These experiments were able to demonstrate the importance of adsorption in the fouling of membranes by small colloidal material. In the experiments shown in Figure 2, the membrane used in the third filtration step was a 0.2-µm polypropylene membrane. The nominal pore size of these membranes is at least 10 times larger than the colloids that cause fouling. Sieving or cake formation is not likely under these conditions, but adsorption appears to be a physically realistic explanation. To verify the importance of adsorption, additional experiments were conducted. Beaver Lake and Medina River water were subjected to a three-step sequential filtration process (prefiltration, fractionation, fouling determination). For each source water, three aliquots of prefiltered water were used in the fractionation step. One aliquot was fractionated through a 0.2-µm polypropylene membrane, the second was fractionated through a 0.2-µm polyethersulfone membrane, and the third was left unfractionated. Because the two fractionation membranes had the same pore size, material retained only because of its size should permeate both membranes equally. Any difference in fouling during the final filtration step would imply a difference in removal by other mechanisms during the fractionation step. The results are shown in Figure 4. For both source waters, the membranes that were fed water fractionated through the 0.2-µm polyethersulfone membrane exhibited flux decline curves almost identical to the unfractionated water. Thus, the 0.2-µm polyethersulfone membrane retained little or no foulants. However, little fouling occurred on the membrane that received permeate from the 0.2-µm polypropylene membrane; passage through the polypropylene membrane was sufficient to remove the foulants. The only variable in these experiments was the fractionation membrane, so differences in retention are due to differences in membrane material chemistry or morphology. Membrane morphology includes parameters such as physical structure, porosity, roughness, tortuosity, and thickness. The effect of physical structure on fouling is difficult to assess quantitatively. The morphology of the fractionation membranes was qualitatively assessed with scanning electron microscope imagery of the membrane surface and crosssection; in those images, the physical structures of the polypropylene and polyethersulfone membranes appear similar. An important factor in material chemistry is hydrophobicity. Polyethersulfone is more hydrophilic than polypropylene, when characterized by contact angle measurements. Previous research has shown that protein adsorption is greater on hydrophobic membranes (20, 25, 28). As a result, it can be inferred that adsorption had an effect on the retention of foulants by these membranes and that the greater retention by the polypropylene membrane was due to the greater hydrophobicity of the material. The results are similar regardless of the source of the feedwater or the pore size of the final membrane. The bottom panel in Figure 4 is dramatic because Medina River water exhibited virtually no fouling after being fractionated through the polypropylene membrane; complete removal of the fouling components was achieved. Complete removal of foulants in natural waters has not been reported in previous literature. Complete removal of foulants was similarly observed when Salt River water and Ohio River water were filtered in sequential filtration experiments. The complete removal of foulants in several different experiments demonstrates the importance of adsorption of very small colloidal matter as a fouling mechanism. DOC measurements in these tests indicated that the nonfouling feedwaters contained about 95% of the DOM in the raw water, cor-

FIGURE 4. Permeate flux after fractionation with 0.2-µm membranes. (A) Flux of Beaver Lake water through 0.2-µm PP membranes at P ) 0.69 bar. (B) Flux of Medina River water through 20 000-Da PES membranes at P ) 1 bar (T ) 25 °C). roborating the earlier evidence that the majority of DOM does not cause fouling by itself. Fouling by Inorganic versus Organic Matter. The previous experiments indicated that a small fraction of DOM was implicated in fouling. The experiments, however, only fractionated membrane foulants by size; there was actually no differentiation between organic and inorganic foulants. Some previous studies have attributed fouling to organic compounds (11-13), while others have addressed the significance of both DOM and inorganic colloids in membrane fouling (15). In the current research, XPS analyses were used to evaluate the elemental composition of membranes fouled with the Medina River and Beaver Lake waters. The elemental composition (excluding hydrogen) of the fouled membranes is shown in Table 3. The membrane fouled with Medina River water contains mostly carbon, oxygen, and nitrogen with a small amount of aluminum and silicon. The membrane appears to be fouled primarily with organic matter and a small amount of aluminum silicate. Aluminum silicate is a primary component of colloidal clay. In contrast, the membrane fouled with Beaver Lake water contains a significant amount of silica and aluminums30% of the total mass at the surface of the sample. Both membranes were VOL. 36, NO. 16, 2002 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3575

TABLE 3. Elemental Composition of Polypropylene Membranes Fouled with Medina River and Beaver Lake Water element

Medina River (% mass of element)

Beaver Lake (% mass of element)

carbon oxygen nitrogen silica aluminum calcium

68 19 6 4 2 1

26 39 5 18 12 1

also analyzed by attenuated total reflectance Fourier transform infrared spectrometry and absorbance that could be associated with aluminum silicate was observed (19). Thus, the foulants in both water sources appear to consist of both organic and inorganic matter but in different proportions. The fate of all components in natural water must be considered during any type of fractionation study. Several recent studies have fractionated natural waters through various nonionic and ionic resin columns (11, 13). These studies attributed membrane fouling to the neutral hydrophilic fraction of DOM, where the hydrophobicity and charge of DOM is inferred from its ability to adsorb to the resins. The neutral hydrophilic fraction was the fraction that passed through three resin columns without adsorbing and was present in the effluent from the final column. Fan et al. (11) noted that the neutral hydrophilic fraction had a greater concentration of colloidal DOM (defined as >30 000 Da in their study) and postulated that the accumulation of colloidal DOM within the membrane pore structure may have been a significant factor in flux decline, which is consistent with the results of this research. As an extension to their findings, it should be noted that the natural waters used might have contained inorganic as well as organic colloids. It is possible that the effluent from the final column in the fractionation procedure contained inorganic colloids and that both inorganic and organic colloids may have contributed to fouling.

Acknowledgments This research was funded by the American Water Works Association Research Foundation. The comments and views detailed herein may not necessarily reflect the views of AWWARF, its officers, directors, affiliates, or agents.

Literature Cited (1) AWWA Membrane Technology Research Committee. J. AWWA 1998, 90, 91-105. (2) Laıˆne´, J.-M.; Hagstrom, J. P.; Clark, M. M.; Mallevialle, J. J. AWWA 1989, 81, 61-67.

3576

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 36, NO. 16, 2002

(3) Jucker, C.; Clark, M. M. J. Membr. Sci. 1994, 97, 37-52. (4) Elimelech, M.; Zhu, X.; Childress, A. E.; Hong S. J. Membr. Sci. 1997, 127, 101-109. (5) Braghetta, A.; DiGiano, F. A.; Ball, W. P. J. Environ. Eng.sASCE 1998, 124, 1087-1097. (6) Kulovaara, M.; Metsa¨muuronen, S.; Nystro¨m, M. Chemosphere 1999, 38, 3485-3496. (7) Lin, C.-F.; Huang, Y.-J.; Hao, O. J. Water Res. 1999, 33, 12521264. (8) Lin, C.-F.; Lin, T.-Y.; Hao, O. J. Water Res. 2000, 34, 1097-1106. (9) Yuan, W.; Zydney, A. L. J. Membr. Sci. 1999, 157, 1-12. (10) Yuan, W.; Zydney, A. L. Environ. Sci. Technol. 2000, 34, 50435050. (11) Fan, L.; Harris, J. L.; Roddick, F. A.; Booker, N. A. Water Res. 2001, 35, 4455-4463. (12) Habarou, H.; Makdissy, G.; Croue, J.-P.; Amy, G.; Buisson, H.; Machinal, C. Proceedings of the AWWA Membrane Technology Conference 2001; AWWA: Denver, CO, 2001. (13) Carroll, T.; King, S.; Gray, S. R.; Bolto, B. A.; Booker, N. A. Water Res. 2000, 34, 2861-2868. (14) Cho, J.; Amy, G.; Pellegrino, J. J. Membr. Sci. 2000, 164, 89-110 (15) Scha¨fer, A. I.; Schwicker, U.; Fischer, M. M.; Fane, A. G.; Waite, T. D. J. Membr. Sci. 2000, 171, 151-172. (16) Clark, M. M. Transport Modeling for Environmental Engineers and Scientists; John Wiley and Sons: New York, 1996. (17) American Society for Testing and Materials (ASTM). D509090sStandard Practice for Standardizing Ultrafiltration Permeate Flow Performance Data. Annual Book of Standards; ASTM: Philadelphia, PA, 1999; Vol. 11.01. (18) American Public Health Association, American Water Works Association, and Water Environment Federation (APHA, AWWA, and WEF). Standard Methods for the Examination of Water and Wastewater, 18th ed.; APHA: Washington, DC, 1998. (19) Howe, K. J. Effect of Coagulation on Membrane Filtration Performance, Ph.D. Thesis, University of Illinois, Urbana, IL, 2001. (20) Cheryan, M. Ultrafiltration and Microfiltration Handbook; Technomic: Lancaster, PA, 1998. (21) Lentsch, S.; Aimar, P.; Orozco, J. L. Biotechnol. Bioeng. 1993, 41, 1039-1046. (22) Kim, K. J.; Fane, A. G.; Ben Aim, R.; Liu, M. G.; Jonsson, G.; Tessaro, I. C.; Broek, A. P.; Bargeman, D. J. Membr. Sci. 1994, 87, 35-46. (23) Thurman, E. M. Organic Geochemistry of Natural Waters; Martinus Nijhoff/Dr W. Junk Publishers: Boston, MA, 1985. (24) Evans, D. F.; Wennerstro¨m, H. The Colloidal Domain: Where Physics, Chemistry, and Biology Meet; Wiley-VCH: New York, 1999. (25) Belfort, G.; Davis, R. H.; Zydney, A. L. J. Membr. Sci. 1994, 96, 1-58. (26) Schnoor, J. L.; Nitzschke, J. L.; Lucas, R. D.; Veenstra, J. N. Environ. Sci. Technol. 1979, 13, 1134-1138. (27) Nilson, J. A.; DiGiano, F. A. J. AWWA 1996, 88, 53-66. (28) Fane, A. G.; Fell, C. J. D. Desalination 1987, 62, 117-136.

Received for review February 14, 2002. Revised manuscript received June 6, 2002. Accepted June 13, 2002. ES025587R