Flux Decline during Nanofiltration of Organic Components in Aqueous

Jul 18, 2001 - This paper studies different mechanisms of flux decline for the nanofiltration of aqueous solutions containing organic compounds. The r...
13 downloads 18 Views 71KB Size
Environ. Sci. Technol. 2001, 35, 3535-3540

Flux Decline during Nanofiltration of Organic Components in Aqueous Solution BART VAN DER BRUGGEN* AND CARLO VANDECASTEELE Department of Chemical Engineering, University of Leuven, W. de Croylaan 46, B-3001 Heverlee, Belgium

Flux decline due to interaction of the membrane with the feed solution is a major drawback for the use of nanofiltration in environmental applications. This paper studies different mechanisms of flux decline for the nanofiltration of aqueous solutions containing organic compounds. The resistance model for flux decline is used: different mechanisms contribute through an increase of the resistance of the membrane against mass transport. The focus in this research is on pore blocking and adsorption inside the membrane pores. Osmotic pressure is also taken into account as it decreases the driving force. The nanofiltration membranes used were NF70 (Dow), UTC-20 and UTC-60 (Toray Ind.), and NTR 7450 (Nitto-Denko). Experiments with different organic components in aqueous solution showed that adsorption resulted in a strong decrease of the water flux. The results of the flux decline as a function of the concentration could well be fitted with the Freundlich equation for adsorption. The components that showed the largest effect had the highest polarity (permanent dipole moment or polarizability), which indicates that adsorption is favored by the polarity of the components in solution. Moreover, the molecules with a size similar to the pore size had a stronger effect on the water flux than other molecules. This can be explained by blocking of the pores by adsorbed compounds.

Introduction

Materials and Methods

Pressure-driven membrane processes such as ultrafiltration and reverse osmosis, and especially nanofiltration, are often necessary to meet the increasingly stringent requirements for process waters and wastewaters. With nanofiltration, a wide diversity of compounds can be removed in one step: organic molecules above a given minimum size and inorganic compounds. Energy consumption is much lower than for reverse osmosis, which makes it an economically feasible unit operation (1). These advantages lead to a growing list of possible applications: removal of hardness, pesticides, and natural organic matter in the drinking water industry (2); water reclamation in the textile industry (3); recovery of valuable pharmaceutical products (4); and removal of heavy metals (5). In nanofiltration, retention properties are very important: the possibility to retain relatively small organic molecules and multivalent ions from aqueous solution at moderate pressures is crucial for most applications. There* Corresponding author telephone: +32 16 32.23.40; fax: +32 16 32.29.91; e-mail: [email protected]. 10.1021/es0100064 CCC: $20.00 Published on Web 07/18/2001

fore, most research has focused on retention properties (6-9). In addition to the retention of the components to be removed from the solution, the water flux through the membrane is an important parameter for the design of a nanofiltration unit. The membrane surface necessary to process a given water stream is determined by the amount of water passing through the membrane per unit of pressure and of surface. Moreover, the retention of a dissolved component is influenced by the water flux: the larger the flux, the less transport through the membrane by diffusion. Of course, a high water flux is desirable to keep the investment cost for the membranes as low as possible (1). Two phenomena are important: the immediate decline of the water flux when applying the solution to be filtered to the unit and the stability of the water flux as a function of time. The first phenomenon, flux decline due to the composition of the feed solution, is reflected by a difference in water flux between feed solution and pure water. The second phenomenon is also crucial: the water flux should be constant as a function of time; a gradual loss of flux due to membrane fouling is to be avoided. A successful implementation of nanofiltration requires a limited immediate flux decline due to the feed solution and a constant water flux as a function of time. The purpose of this research is to study the mechanisms of flux decline for aqueous solutions containing organic components. Different phenomena will be distinguished: adsorption at the membrane surface or inside the pores, pore blocking, osmotic pressure, concentration polarization, formation of a gel layer, and deposition of suspended solids on the membrane. The two latter effects are very specific for solutions containing high concentrations of macromolecules or suspended solids and will be avoided in the experiments. Concentration polarization is a phenomenon that is wellknown and adequately described (10); it will not be discussed in detail here. It will be shown that for aqueous solutions containing organic components, pore blocking and adsorption on the membrane surface are important mechanisms of flux decline, whereas the effects of the osmotic pressure are less important. Flux decline or fouling caused by inorganic compounds (e.g., CaCO3, CaSO4, or BaSO4 scaling) will not be studied in this paper.

 2001 American Chemical Society

Equipment. For the filtration experiments, a commercial nanofiltration unit on laboratory scale was used (Amafilter, Test Rig PSS1TZ). Samples of permeate and retentate were taken for each measurement; sampling was done in steadystate conditions (10 min after the last change of the parameter settings). In all experiments, a feed velocity of 6 m/s was maintained. The feed temperature was set at 25 ( 0.1 °C. Flat sheet membranes were used with a diameter of 0.09 m. The membranes used are mostly commercially available nanofiltration membranes. They were selected because they cover the whole range of operation of nanofiltration (from reverse osmosis to ultrafiltration). Table 1 summarizes the most important characteristics of these membranes. All membranes are thin film composite. NF70 has a cross-linked aromatic polyamide top layer; NF45, UTC-20, and UTC-60 have a polypiperazine amide top layer; and NTR 7450 has a sulfonated polyether sulfone top layer. All the membranes used have a surface charge. VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3535

TABLE 1. Properties of Selected Membranes membrane type

NF70

NF-45

UTC-20

UTC-60

NTR 7450

manufacturer MWCa (g/mol) permeability (L m-2 h-1 bar-1) max temp (°C) max pressure (bar) pH range charge

Dow-FilmTec 200 11 35 17 3-9 -

Dow-FilmTec

Toray Ind. Inc. 180 15 35 15 3-10 -

Toray Ind. Inc.

Nitto-Denko 600-800 12 40 30 2-11 -

7 35 41 2-11 -

10 35 15 3-9 +

a MWC is the molecular weight of a component that is retained for 90% and is considered to be the lower limit for the size of a molecule that is still retained by the membrane.

TABLE 2. Components Used in Experiments Together with Analysis Method Used for Their Determination effective dipole diameter moment (nm) (D) methanol ethanol 2-propanol methyl ethyl ketone (MEK) ethyl acetate toluene aniline phenol cyclohexanone methyl metacrylate isobutyl methyl ketone (BMK) benzonitrile benzyl alcohol caprolactam benzoic acid nitrobenzene xylose galactose maltose raffinose

determination

0.27 0.34 0.39 0.42

1.6 1.7 1.8 2.8

gas chromatography gas chromatography gas chromatography gas chromatography

0.48 0.50 0.49 0.49 0.45 0.52 0.52

1.7 0.4 1.5 1.7 2.8 2.0 2.7

gas chromatography gas chromatography gas chromatography UV spectrophotometry gas chromatography gas chromatography gas chromatography

0.51 0.54 0.50 0.52 0.52 0.55 0.66 0.84 0.94

3.9 1.7 3.9 4.2 3.8 1.0

UV spectrophotometry UV spectrophotometry gas chromatography UV spectrophotometry gas chromatography UV spectrophotometry UV spectrophotometry UV spectrophotometry UV spectrophotometry

Experimental. A series of organic molecules was selected for nanofiltration experiments in deionized water. Most of these molecules are uncharged at the pH values used in the experiments (pH 6-8). The series includes aromatic as well as aliphatic compounds, alcohols, ketones, esters, and sugars. They were not selected because of environmental significance but rather as model components because their size corresponds to the pore sizes of the different membranes. Table 2 summarizes the used compounds together with the analysis method applied for their determination. Since concentration polarization was not a main research topic in this study, it has been minimized by applying a high cross-flow velocity in all experiments (usually 6 m/s). The Reynolds number that was obtained this way was as high as 25 000, which justifies neglecting concentration polarization for the experiments. In the first series of experiments that was carried out, solutions containing one single component in increasing concentrations was used, starting with filtration of pure water. The components studied were benzonitrile, benzyl alcohol, phenol, cyclohexanone, caprolactam, toluene, and nitrobenzene. These were selected because a significant flux decline was observed with these components in preliminary experiments. The NF70 membrane was used for this research. The water flux through the membrane was measured as a function of concentration for all components. The pressure in these experiments was 10 bar, so that the water flux is well below the critical flux. Hence, operation above the critical flux might as well have an influence on the measurement of flux decline. In a second series of experiments, the influence of the 3536

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001

FIGURE 1. Influence of the concentration on the water flux for benzonitrile (membrane, NF70). structure and size of the solute molecules on the water flux through the membrane was examined by means of the effective diameter (9) and dipole moment of the molecules. Solutions containing a single component from Table 2 in a concentration of 250 mg/L were nanofiltrated. Of course, this might overestimate the influence of molecules with a lower molecular weight such as methanol or ethanol because their molar concentration is relatively higher. This effect is neglected because molecular weight differences in the range where flux decline occurs are small. The experiments can be divided into three steps: (a) measurement of the pure water flux at pressures of 5, 8, 10, 15, and 20 bar; (b) measurement of the water flux at the same pressures with the selected solution; and (c) second measurement of the pure water flux at these pressures. The comparison with the initial water flux is necessary as a reference value to measure the flux decline. Afterward, another solution was measured. The last pure water flux measured served as the initial pure water flux for this next solution. In this way, a quick scan of all the components was possible. A disadvantage of this approach is that preceding components can influence the results for the following ones. However, when the largest fraction of the flux decline is reversible, the results will still be reliable. Additionally, the “quick-scan” approach has the advantage that water fluxes can be numerically compared because the same membrane sheet is always used. Use of different membrane sheets that may have a slightly different water flux makes such a comparison impossible for experiments where membrane changes are necessary; lower water fluxes can be related to slight changes in the membrane properties because a different sheet is used as well as related to flux decline due to the component that was used.

Results and Discussion Mechanisms of Flux Decline. A gradual decrease of the water flux as a function of the concentration was found for each component in the first series of experiments; a typical curve that was obtained this way is shown in Figure 1 for benzonitrile (membrane, NF70).

A model that is commonly used to describe phenomena of flux decline is the resistance model (10). The water flux is written as

flux )

driving force viscosity × total resistance

which, in case of nanofiltration, becomes

J)

∆P ηRtot

The flux decline that was found in the experiments should be explained by an increase in the total resistance against mass transport. The expression for the water flux can be compared to the Hagen-Poiseuille equation (ideal conditions):

J)

r 2 ∆P 8ητ ∆x

Thus, the membrane resistance depends on the porosity (), the tortuosity (τ), the pore radius (r), and the membrane thickness (∆x). Other equations such as the Kozeny-Carman equation, used for membrane structures consisting of packed spheres, lead to a comparable dependency of the water flux on membrane properties. The Hagen-Poiseuille equation is valid when pure water is applied to the membrane. When solutions of organic molecules in water are applied, the water flux will often be lower. Different mechanisms of flux decline can be distinguished (10, 11). Adsorption inside the pores or at the membrane surface narrows the pores. When the molecules have a similar size as the pores, permeation can lead to pore blocking, a phenomenon that can be enhanced or caused by adsorption. Pore blocking has been observed for ultrafiltration, where macromolecules are filtered (12). For the filtration of nonmacromolecular components with nanofiltration, this phenomenon has not yet been described. Concentration polarization results in an increased concentration of solutes at the membrane surface; this high concentration of dissolved molecules provides an extra barrier for mass transport. Eventually, the concentration at the membrane can become so high that a gel layer is formed, preventing mass transport. This effect is somewhat comparable to the deposition of suspended solids on the membrane surface. Each of these mechanisms corresponds to an increase in the total resistance for mass transport. The total resistance is the sum of the different individual resistances, i.e., Rtot ) Rp + Ra + Rm + Rg + Rcp + Ri + Rd (Rp, resistance due to pore blocking; Ra, resistance due to adsorption inside the pores; Rm, membrane resistance (intrinsic); Rg, resistance caused by the formation of a gel layer; Rcp, concentration polarization resistance; Ri, resistance caused by specific interactions; Rd, resistance from deposits on the membrane). In the ideal case, e.g., filtration of pure water, the membrane resistance (Rm) is the only resistance involved. This is an intrinsic membrane characteristic that corresponds to the resistance calculated from, for example, the HagenPoiseuille equation and does not change during filtration or by changing the feed solution. It reflects the minimal resistance of the system against mass transport and thus determines the maximal water flux at a given pressure. The other phenomena can only make pores narrower (or the membrane thicker), resulting in an increase of the total resistance or the addition of an extra resistance term to the intrinsic membrane resistance. The gel layer resistance, the adsorption resistance, the pore blocking resistance, the deposition resistance, and the

concentration polarization resistance depend strongly on the type of feed solution that is used. In this case, the gel layer resistance is not present, as the formation of a gel layer is related to macromolecules, which are not present here. In all experiments, synthetic solutions of organic components in pure water were used, so that suspended solids do not occur; this contribution to resistance against mass transport is thus avoided. Concentration polarization can be neglected because of the experimental conditions. The only factors that remain are the membrane resistance, pore blocking, and adsorption. Before considering these resistance factors, the osmotic pressure should be taken into account. The retention of (multivalent) ions and small organic molecules in nanofiltration causes buildup of an osmotic pressure. This pressure has to be counterbalanced by the applied transmembrane pressure. Therefore, the pressure needed to obtain a given water flux will be higher, or the water flux at a given transmembrane pressure will be lower. Thus, the osmotic pressure causes flux decline, but this is due to a decrease of the driving force instead of an increase of the resistance against mass transport. This can be expressed by the phenomenologic equation for the water flux (13), originally introduced by Kedem and Katchalsky:

Jv ) Lp(∆P - σ∆Π) If the reflection coefficient (σ), the maximal retention of the component at an “infinite” pressure, can be assumed to be equal to 1, the water flux would be 0 when the applied pressure equals the osmotic pressure. The extent to which the osmotic pressure will play a role is determined by (a) the retention of the components in the solution, (b) their concentration, and (c) their molecular mass. Nanofiltration is a process that works on the interface of reverse osmosis (where low molecular weight components are retained; osmotic pressures are generally high) and ultrafiltration (where high molecular weight components are retained; osmotic pressures are generally negligible). Therefore, specific membrane properties determine the effect of the osmotic pressure. Because retention of uncharged molecules is related to pore sizes, it is obvious that membranes with the smallest pore sizes will show the largest osmotic pressures. The osmotic pressure should thus be deducted from the total pressure or, the water flux at the chosen pressure of 10 bar should be corrected by a factor ∆P/(∆P - ∆π). Because the osmotic pressures are small for these solutions, the flux decline due to the osmotic pressure is small. The osmotic pressure was calculated with the van’t Hoff equation for the concentrations used in the experiments. The influence of the osmotic pressure is only minor. For example, a concentration of 1 g/L benzonitrile would result in a flux decline due to osmotic pressure of 1.64 L m-2 h-1 at 10 bar, which is less than 5% of the total flux decline that was found experimentally. This conclusion can be extended to all experiments carried out in this framework because osmotic pressures would be comparable. The remaining flux decline can be explained by adsorption inside the membrane pores or at the membrane surface, possibly enhanced by pore blocking. The experimental flux decline for all components (corrected for the osmotic pressure) is given in Figure 2. A commonly used equation to model adsorption phenomena is the Freundlich equation:

q ) Kfcn where c is the concentration of the component to be adsorbed at equilibrium and q is the amount of the component that VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3537

TABLE 4. Flux Decline (%) between Initial Water Flux and Water Flux after Filtration with the Last Component 5 bar 8 bar 10 bar 15 bar 20 bar

FIGURE 2. Flux decline for the used set of molecules (membrane, NF70).

FIGURE 3. Flux decline as a function of concentration (extrapolated) for benzonitrile (membrane, NF70).

TABLE 3. Modeling of Flux Decline with the Freundlich Equation molecule

Kf

n

molecule

Kf

n

benzyl alcohol benzonitrile benzoic acid cyclohexanone

72.0 47.1 46.1 33.3

0.47 0.48 0.80 0.42

phenol caprolactam toluene nitrobenzene

31.8 18.7 27.9 20.7

0.35 0.19 0.75 0.66

is adsorbed on the material, divided by the amount of material. Kf and n are empirical constants. If it is assumed that adsorption and flux decline are proportional, q in the Freundlich equation can be replaced by the flux decline ∆J:

∆J ) Kfcn This equation was fitted to the experimental results from Figure 2. Kf and n values were determined for each of the molecules (Table 3). The correlation between experimental and calculated data was very good. Correlation coefficients were always above 0.98, except for toluene (0.96) and nitrobenzene (0.93). Figure 3 represents the curve for benzonitrile (correlation coefficient 0.998). The water flux would drop back to zero at about 6.5 g/L. At this point, all flux is obstructed by adsorption of benzonitrile on the membrane or inside the pores. For charged molecules, electrostatic interactions between solute and membrane are to be expected. These effects might be a major cause for flux decline or fouling. For uncharged organic compounds, adsorption is the process that is most likely to occur (14, 15). Molecules can get attached to the membrane pores or to the membrane surface by adsorption or chemisorption. Inside the pores, they narrow the free pathway for the water flow, hence decreasing the net pore opening. From the Hagen-Poiseuille equation, it can be seen that this should lead to a flux decline. When adsorption has a strong effect, it could even lead to pore blocking when the whole cross section of the pore is filled. 3538

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001

NF70

UTC-60

UTC-20

NTR 7450

5.4 -1.1 4.2 6.3 -2.8

32.1 29.7 32.3 29.4 31.2

4.5 2.3 0.8 -7.7 -16.1

59.4 61.0 61.1 64.6 63.4

Membranes can become blocked when molecules with approximately the same size as the pore opening fill the membrane pores and are prevented from further permeation, e.g., when the pore becomes narrower toward the end (16). Because it has been shown that the range of pore sizes can differ significantly between different membrane types, this effect should be different for each membrane (9). Not only the average pore size differs, but the distribution of pore sizes can also be different. A wide pore size distribution is expected to lead to less flux decline due to pore blocking when a single component solution is used, because only a fraction of the pores would be blocked. When fouling is caused by irreversible adsorption or chemisorption, a stable flux is expected once all adsorption sites in the membrane pores or at the surface are occupied and an adsorption equilibrium is established. Reversibility by mechanical or chemical cleaning would involve desorption of the organic material, which can only be obtained with organic solvents that have more affinity for the molecules. These organic solvents might possibly damage the polymeric membranes. Pore blocking can be reversible or irreversible. If the blocked molecules are not removed by cleaning with pure water (this is irreversible flux decline or fouling), the effect would anyhow be limited because once the pores with the appropriate size are filled, the pore blocking will stop at this level of flux decline. Adsorption already occurs before pressure has been applied and the membrane process has been started (17). As soon as the top surface of the membrane is in contact with the solution, solute molecules will adsorb at the membrane surface due to physicochemical interactions, e.g., hydrophobic interactions (dispersion forces), polar interactions (dipole-dipole and dipole-induced-dipole forces), and charge transfer (hydrogen bonding). The nature of the membrane material, the type of solute, the solute concentration, and the pH are parameters that can determine the extent of adsorption. For real samples of wastewater or process water, other processes could also cause flux decline. For solutions with high concentrations of low molecular compounds (e.g., salt solutions), the osmotic pressure is the most important effect. Specific feed-related problems such as bacterial growth or deposition of suspended solids can usually be solved with an appropriate pretreatment (flocculation or ultrafiltration, desinfection when needed). After a constant water flux is reached, additional flux decline on a longer time scale can still occur. This can be caused by bacterial growth (18), accidental presence of pollutants such as colloids or suspended solids, or formation of a thin layer of polluting compounds on the surface of the membrane (19). Influence of Molecular Size and Dipole Moment. For the second series of experiments, the comparison for the different membrane types between the initial water flux and the water flux after filtration of all 20 organic components is very important. In Table 4, the difference between NTR 7450 and UTC-60 on one hand and NF70 and UTC-20 on the other hand is obvious. The irreversible flux decline (without cleaning) is small for NF70 and UTC-20 but quite considerable

TABLE 5. Molecules That Cause Flux Decline with NF70 and UTC-20a NF70 UTC-20 toluene aniline phenol cyclohexanone benzonitrile

X X X XX XXX

X X XX

NF70 UTC-20 benzyl alcohol XXX caprolactam XX benzoic acid nitrobenzene XXX

XXX X XX X

a X refers to average flux decline between 5 and 10%. XX refers to average flux decline between 10 and 15%. XXX refers to flux decline over 15%.

FIGURE 4. Flux decline (%) as a function of effective diameter for (a) NF70 and (b) UTC-20. for UTC-60 and NTR 7450. Indeed, for NF70 and UTC-20 the water flux in each new experiment is approximately the same as in the preceding experiment. Even negative values for the flux decline were found, which corresponds to a flux increase. However, these negative values are rather caused by the repeated effect of flux decline followed by a flux increase during the rinsing step; when both effects do not counterbalance each other completely, small negative values can be found. For NTR 7450 and UTC-60, the water flux is gradually lower after each experiment. Therefore, no strict conclusions can be made from this experiment for NTR 7450 and UTC60. Obviously, fouling occurs for these two membranes. Electrostatic interactions could explain the fouling with the UTC-60 membrane, which is positively charged. NTR 7450 is negatively charged, but hydrophobic interactions are a possible reason for the fouling with this membrane. The fouling effect seemed to be the most explicit for the experiments with toluene, aniline, phenol, cyclohexanone, benzonitrile, isobutylmethyl ketone, benzyl alcohol, caprolactam, and benzoic acid. Although some reserve is necessary because of the irreversible effects in the experiment for NTR 7450 and UTC-60, it can be seen from Table 2 that these are the components with the highest dipole moment or polarizability. For NF70 and UTC-20, the flux decline is presented as a function of molecular size. In Figure 4, the effective diameter of the tested molecules is used as the size parameter (9); flux decline is calculated as the percent decrease between the previous pure water flux and the feed solution that contained the particular component. Both membranes show a peak in flux decline around 0.5 nm, the effect for NF70 being somewhat sharper than for UTC-20. The average flux decline for NF70 is larger: for molecules between 0.45 and 0.55 nm, the average flux decline is 3.4% for UTC-20 but 11.0% for NF70. If a flux decline of (5% is taken as a minimum for a significant effect, then the flux decline is limited to the narrow range between 0.45 and 0.55 nm. This corresponds to the pore size and pore size distributions that were obtained with the log-normal model (9). The logarithm of the average pore size for UTC-20 is -0.27 (corresponding to a pore size of 0.54 nm) with a standard deviation of 0.17; the logarithm of the average pore size for NF70 is -0.46 (corresponding to a pore size of 0.34 nm) with a standard deviation of 0.54 nm. NF70 has smaller pores but a wider range of pore sizes; the wide range of pore sizes for NF70 causes a more severe effect of decline because a larger

variation of pore sizes is “available” for the molecules to fit in. This shows that molecular size is an important factor for flux decline because the flux decline can only occur when the molecule can enter the membrane structure and has the right size to prevent further water passage. Molecules that are too small can freely pass the membrane or do not hinder water transport; molecules that are too large cannot enter the membrane structure and can therefore not cause flux decline or fouling. Pore size alone cannot explain flux decline for all components. For example, a solution containing 250 mg/L ethyl acetate (effective diameter 0.48 nm) caused a flux decline of 2.6% with NF70, but a solution containing 250 mg/L phenol (effective diameter 0.49 nm) caused a much larger flux decline: 13.9%. In Table 5, the different molecules that were found to cause flux decline for NF70 and UTC-20 are summarized. Both membranes show flux decline with approximately the same components, with a shift to somewhat larger molecules for UTC-20 due to the slightly larger pores in this membrane. A noticeable conclusion is that flux decline occurs for largely comparable molecules. The four molecules with a dipole moment above 3 D (benzonitrile, caprolactam, nitrobenzene, and benzoic acid) are all included in Table 5. Other molecules such as cyclohexanone are also fairly polar (cyclohexanone has a dipole moment of 2.8 D). All other molecules are cyclic (apart from cyclohexanone, all are aromatic) and have retentions with the two membranes that are not near 0% or 100%. In other words, the molecules where flux decline occurs are those that have a high dipole moment or that have a high polarizability. The influence of polarity and polarizability is not surprising. Indeed, in the first series of experiments it has been found that adsorption on the membrane surface is a major cause of flux decline. Because the nanofiltration membranes are charged, adsorption is expected to be enhanced by a high polarity or polarizability; thus, solutions containing polar molecules are subject to flux decline. The larger the polarity or polarizability of the molecule, the more effect adsorption would have, and, thus, the more flux decline would occur.

Literature Cited (1) Owen, J.; Bandi, M.; Howell, J. A. Churchhouse, S. J. J. Membr. Sci. 1995, 102, 77. (2) Rachwal, A. J.; Khow, J.; Colbourne, J. S.; O’Donnell, J. Desalination 1994, 97, 427. (3) Van der Bruggen, B.; Daems, B.; Wilms, D.; Vandecasteele, C. Sep. Purif. Technol. 2001, 22-23, 519. (4) Raman, L. P.; Cheryan, M.; Rajagopalan, N. Chem. Eng. Prog. 1994, 68. (5) Alami-Younssi, S.; Larbot, A.; Persin, M.; Sarrazin, J.; Cot, L. J. Membr. Sci. 1994, 91, 87. (6) Bowen, W. R.; Mukhtar, H. J. Membr. Sci. 1996, 112, 263. (7) Long, T. D.; Jacobs, P. L.; Anderson, J. L. J. Membr. Sci. 1981, 9, 13. (8) Mitchell, B. D.; Deen, W. N. J. Colloid Interface Sci. 1986, 113, 132. (9) Van der Bruggen, B.; Schaep, J.; Wilms, D.; Vandecasteele, C. Sep. Sci. Technol. 2000, 35 (2), 169. VOL. 35, NO. 17, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

3539

(10) Mulder, M. Basic Principles of Membrane Technology, 2nd ed.; Kluwer Academic: Dordrecht, The Netherlands, 1996. (11) Eykamp, W. In Membrane Separations Technology, Principles and Applications; Noble, R. D., Stern, S. A., Eds.; Elsevier: Amsterdam, 1995; pp 1-43. (12) Doyen, W. Latest Developments in Micro- and Ultrafiltration Membranes for Large Scale Drinking Water Applications; Workshop on Membranes in Drinking Water Production, l’Aquila, Italy, June 1-4, 1997. (13) Wang, X. L.; Tsuru, T.; Togo, M.; Nakao, S. I.; Kimura, S. J. Chem. Eng. Jpn. 1995, 28 (2), 186. (14) Kiso, Y.; Kitao, T.; Nishimura, K. J. Appl. Polym. Sci. 1999, 74, 1037. (15) Kiso, Y.; Kitao, T.; Nishimura, K. J. Appl. Polym. Sci. 1999, 71, 1657. (16) Boerlage, S. F. E.; Kennedy, M. D.; Bonne, P. A. C.; Galjaard, G.; Schippers, J. C. Prediction of Flux Decline in Membrane Systems

3540

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 17, 2001

due to Particulate Fouling; Workshop on Membranes in Drinking Water ProductionsTechnical Innovations and Health Aspects, l’Aquila, Italy, June 1-4, 1997. (17) Mulder, M. In Membrane Separations Technology, Principles and Applications; Noble, R. D., Stern, S. A., Eds.; Elsevier: Amsterdam, 1995; pp 45-84. (18) Coker, S.; Sehn, P. Proceedings of the Conference on Membranes in Drinking and Industrial Water Production; l’Aquila, Italy, October 2000; Vol. 2, pp 447-451. (19) Zhu, X.; Elimelech, M. Environ. Sci. Technol. 1997, 31, 3654.

Received for review January 3, 2001. Revised manuscript received May 16, 2001. Accepted May 31, 2001. ES0100064