Mutual Influences between Natural Organic Matter and Inorganic

Fouling is one of the most critical aspects of membrane technology and is strongly influenced by natural water characteristics. This study focuses on ...
0 downloads 0 Views 2MB Size
Download PDF
Environ. Sci. Technol. 2008, 42, 9129–9136

Mutual Influences between Natural Organic Matter and Inorganic Particles and Their Combined Effect on Ultrafiltration Membrane Fouling DORIS JERMANN, WOUTER PRONK,* AND MARKUS BOLLER Eawag, Swiss Federal Institute of Aquatic Science and Technology, P.O. Box 611, 8600 Du ¨ bendorf, Switzerland

Received March 5, 2008. Revised manuscript received September 10, 2008. Accepted September 15, 2008.

Fouling is one of the most critical aspects of membrane technology and is strongly influenced by natural water characteristics.Thisstudyfocusesonamechanisticunderstanding of the impact of interactions between natural organic matter (NOM) and particles on fouling. The model substances used were humic acid, alginate (polysaccharide), and kaolinite. NOM-kaolinite adsorption experiments, particle characterization, and dead-end ultrafiltration (UF) batch experiments were performed. The adsorption experiments indicated particle stabilization at low NOM equilibrium concentrations, whereas calcium induced significant aggregation, especially with alginate. UF experiments implicated a synergistic fouling effect of particle-NOM combinations, which was greatly reduced bycalcium.Moreover,irreversibleNOMfoulingwasonlyprevented by particles in the presence of calcium. On the basis of our results,wepresentamechanisticmodelsuggestingthatsynergistic fouling effects occur due to particle stabilization by NOM adsorption, especially shown for HA, and antagonistic effects due to particle destabilization by calcium. However, synergistic fouling can also be based on sterical interferences between larger NOM in the form of polysaccharides and particles during simultaneous pore blocking and cake formation. A heterogeneous NOM-particle fouling layer is ultimately formed with membrane associations dominated by NOM. The combined fouling is conclusively determined by the type of NOM, its specific fouling mechanisms, and its particle interactions prior to and during the filtration process.

Introduction During the past decade, ultrafiltration (UF) has emerged as one of the most reliable, cost-effective, and sustainable unit processes for the production of drinking water. This has stimulated research aiming to gain an understanding of fouling, which is currently the major problem in membrane applications. Natural organic matter (NOM) is generally accepted to represent the main foulant, especially for surface waters. Within NOM, humic substances and polysaccharides were shown to be critical for UF due to membrane adsorption and pore blocking (1-4). Besides NOM, particles have also been recognized as critical membrane foulants. Particulate cake fouling is * Corresponding author e-mail: [email protected]; phone: +41 44 823 5381; fax: +41 44 823 53 89. 10.1021/es800654p CCC: $40.75

Published on Web 11/14/2008

 2008 American Chemical Society

characterized by the layer morphology that governs the flux decline and by the layer-membrane associations that determine the fouling reversibility. Particle characteristics such as size, charge, and surface chemical stability crucially affect the cake layer morphology (5, 6). Earlier research in aquatic science suggested that NOM would stabilize inorganic particles, as observed principally with humic substances, and that divalent cations, in particular calcium, would destabilize them (7, 8). More recently, however, particle destabilization has also been shown to occur with specific fractions of NOM, namely, polysaccharides (9, 10). Empirical studies and membrane fouling layer analysis suggest that the presence of NOM affects particulate membrane fouling by impairing the fouling layer structure. Moreover, NOM increases irreversible fouling by membrane adsorption and gel layer formation (2, 11-13). Practically all mechanistic studies have aimed at separately determining the effects caused by either NOM or particle fouling. Research on combined NOM-particle filtration has been published only recently. Particle aggregation/stabilization by NOM was found to be a crucial effect during combined filtration. Nevertheless, combined NOM-particle fouling is still poorly understood. However, as NOM and particles are ubiquitous in natural surface waters, knowledge of such combined fouling could well be vital to understand membrane fouling under practical conditions. The present study aims at acquiring a further understanding of the fouling mechanisms of NOM in combination with inorganic particles. The focus is on NOM-particle interactions in the presence and absence of calcium. The impact of the NOM type and concentration and the presence of calcium on NOM-particle interactions and the consequences for combined UF fouling are studied in detail. To investigate these parameters systematically and in detail, only one type of particle (kaolinite) and only one particle size were used for reasons of simplification.

Experimental Section Model Compounds and Solution. For the experiments, humic acid (HA; IHSS, 1-5 kDa) and the polysaccharide alginate (alginic acid, Sigma-Aldrich, Switzerland, 12-80 kDa) were selected as representatives of NOM substances, and kaolinite (Clay Society, particle size 450 nm, specific surface area (BET) 22.5 m2/g (14)) was selected to represent the inorganic particles. The preparation details for the NOM solutions can be found elsewhere (15). Fresh kaolinite solutions were prepared and placed in an ultrasonic bath for 10 min directly prior to usage. Details on the physical and chemical characteristics of the model substances as well as the selection criteria to use these substances as model compounds are given in Supporting Information Table 1. The background solutions consisted of deionized (DI) water, NaCl, NaHCO3 (1 mM), and calcium (0 or 1.25 mM). The pH was adjusted to 7.5 with HCl (0.2 M)/NaOH (0.2 M). The ionic strength was 20 mM in all experiments. UF Experiments. UF experiments were performed to study fouling during individual and combined NOM-particle filtration. In the latter case, this was done after NOM-kaolinite interactions had taken place in solution. The concentrations used were 0.02, 0.2, and 2 mg of C/L HA, 0.02, 0.2, and 1.0 mg of C/L alginate, and 10 and 100 mg/L particles (dry solid mass). In the NOM-particle combinations, the organics were always added to the particles in glassware and the solutions were stirred overnight. VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9129

For all experiments, we used a fresh flat sheet of poly(ether sulfone) (PES) UF membrane (100 kDa, PBHK, Biomax). Prior to the experiments, the membranes were washed by a leaching procedure described elsewhere (15). The UF test unit was a stirring cell with the stirrer removed (400 mL, Amicon). Its setup is given in detail elsewhere (15). The cell pressure was regulated at 0.5 ( 0.01 bar by a pressure control device. Prior to every experiment, the clean membrane flux was assessed with DI water. The experiments comprised three cycles. Each cycle consisted of solution filtration (1 L), backwash (DI water, 50 mL, 0.5 ( 0.01 bar), and filtration with DI water (200 mL) to assess the reversibility of the flux decline. The following permeate fractions were sampled (mL): first cycle, 0-100, 100-900, and 900-1000; second cycle, 0-100; third cycle, 0-100 and 900-1000. The first 10 mL of each cycle was discarded. NOM-Particle Adsorption Experiments. NOM-kaolinite adsorption experiments were performed to characterize the NOM-kaolinite interactions in solution. The concentration of the particles was 1 g/L, while that of the organics was increased from 0 to 80 mg of C/L. The solutions were mixed by adding the organics to the particles and shaken for 24 h. They were finally ultracentrifuged (32 000 rpm, 60 min), and the organics in the supernatant were measured. NOM and Particle Analytics. The HA concentration was quantified by UV spectrometry (Varian, CARY.100scan, 254 nm). The particle size was measured by photon correlation spectroscopy (PCS) (Zetasizer NS, Malvern, United Kingdom). The measurable particle size range given by the supplier is 0.6 nm to 10 µm. A Zetamaster (Malvern) was used to analyze the zeta potential of the particles.

Theoretical Basis Adsorption Isotherms. The adsorption of organics onto particles is often described by the Langmuir or Freundlich isotherm. Langmuir adsorption describes a monolayer process, whereas the empirical Freundlich isotherm describes any adsorption and is often used for adsorption processes of greater complexity than the monolayer type. Details of and formulas for the adsorption isotherms can be found in ref 16. Membrane Filtration Cake Resistance. Particles that are larger than the membrane are assumed to cause fouling by forming a cake on the membrane surface. According to filtration theory, the resistance of a cake layer RC can be expressed using Darcy’s law and the Carman-Kozeny equation (eq 1) (17). RC ) (180(1 - ε)/(d2ε3Fp))(mp/Am)

(1)

This equation shows that the cake layer resistance is determined by particle characteristics such as the diameter d, the density Fp, the mass mp, the membrane surface area Am, and the cake porosity ε. The hydraulic resistance of a fouling cake depends strongly on the cake porosity. The Carman-Kozeny equation is suited to describe the resistance of relatively dense cakes (porosity e0.5). Aggregates are known to form rather porous cakes, as they have their own fractal porosity (6, 18).

Results and Discussion Adsorption Isotherms. The measured relationships of the NOM-kaolinite adsorption were modeled by the Langmuir and Freundlich adsorption isotherms with the Excel Solver programmed to minimize the sum of squared errors as shown in Figure 1. Both NOM substances show a relatively large adsorption at low NOM concentrations-up to 50% and 30% 9130

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

FIGURE 1. Measured and modeled adsorption of HA and alginate onto kaolinite with 0 and 1.25 mM calcium: (a) 0-80 mg of C/L and (b) enlarged for the 0-10 mg of C/L range. of the initial HA and alginate, respectively, in solution at a C/L equilibrium concentration of 2 mg. The data for HA correspond to those from other studies (19), while no literature values were found for alginate-kaolinite adsorption. In the absence of calcium, HA adsorbs better onto kaolinite than alginate, and this behavior is best described by the Freundlich isotherm. Multilayer adsorption can be assumed to occur by intermolecular hydrophobic HA-HA attraction. However, HA adsorption onto kaolinite is assumed to take place mainly by interactions between negative HA groups and amphoteric kaolinite edges (∼20% of the total surface area), which are positively charged at a neutral pH (8, 20). The adsorption of alginate onto kaolinite is slightly better represented by a Langmuir isotherm than a Freundlich isotherm (sum of least squares n ) 1.55 compared to n ) 1.99). It is suggested that alginate, a polyelectrolyte, adsorbs onto positively charged kaolinite edges via electrostatic attraction (9, 10). Monolayer adsorption is likely to occur with strong intermolecular alginate charge rejection and the absence of hydrophobic interactions. NOM adsorption onto kaolinite is increased in the presence of calcium, particularly for alginate, and is well described by a Freundlich isotherm. In the case of HA, however, the presence of calcium is observed to lead to higher adsorption only in NOM equilibrium concentration ranges above 8 mg of C/L. It was reported earlier that calcium can increase the adsorption of HA onto clay particles (21). However, that particular study used higher HA/clay ratios than this one. Particle Size Measurements. Parts a and b of Figure 2 present the average size of the kaolinite particles after NOM adsorption. In the absence of NOM, an average particle size above 1.5 µm indicates aggregation of particles independently of the presence of calcium (individual particle size 450 nm). In the absence of calcium, the particle size decreases with increasing NOM concentration until particle stabilization is reached. In the case of alginate, the presence of calcium resulted in a pronounced increase of particle size at around 0.5 mg of C/L (equilibrium), showing aggregates of up to 5.5

FIGURE 2. Particle size and zeta potential in NOM-kaolinite suspensions in the presence of 0 and 1.25 mM calcium: (a) particle size as a function of the NOM equilibrium concentration, (b) particle size as a function of the amount of NOM adsorbed onto kaolinite, (c) zeta potential as a function of the NOM equilibrium concentration, (d) zeta potential as a function of the amount of NOM adsorbed onto kaolinite. µm. For HA, particle stabilization was also observed at larger NOM equilibrium concentrations when calcium was present (Figure 2b). Zeta Potential Measurements. The kaolinite zeta potential is presented in Figure 2c,d. It shows that NOM adsorption causes a significant negative shift of the zeta potential at low equilibrium concentrations and stable values at higher concentrations above 1 mg of C/L in solutions without calcium. Calcium increased the zeta potential of all samples by about 20 mV. The values stabilized only above an HA and alginate equilibrium concentration of 8 mg of C/L. Particle size measurements (Figure 2) indicate that particle stabilization occurs between 0 and 0.5 mg of C/L NOM equilibrium concentrations in the absence of calcium. In the same concentration range, large amounts of NOM adsorb onto kaolinite (Figure 1). Aggregates are present below 0.5 mg of C/L. Particles are stabilized when 0.5 mg of HA or 0.5 mg of alginate is adsorbed onto 1 g of kaolinite (Figure 2b). This number is in line with a theoretical calculation of the amount of HA used for monolayer surface coverage. Such a calculation shows that 0.4 mg of HA (C) is used per gram of kaolinitesassuming that one HA molecule occupies a kaolinite surface area of 44 nm2 and that only the edges are occupied by HA (22). Particle destabilization is favored in the presence of calcium, particularly in the case of alginate. This is explained by calcium interactions with NOM leading to a lower negative charge of the particles with an adsorbed NOM layer (Figure 2c,d), which can increase NOM adsorption. This ultimately leads to particle bridging by NOM-NOM and probably also to NOM-kaolinite associations in low organic concentration ranges (7). Strong associations between NOM carboxyl groups and calcium were found previously in molecular modeling studies (23, 24). The particular effect in the presence of alginate can be explained by preferential calcium complexation. It has been reported in the literature and is known as the egg-box model (23). The strong associations between

alginate and calcium can also explain that aggregation of alginate-coated particles is more pronounced than in the case of humic acid-coated particles, although the negative charge of the alginate-coated particles is larger. The alginate-Ca association can lead to relatively strong shielding of the positive charge of calcium by alginate. For alginate, the particles are stabilized when 4 mg of C is adsorbed per gram of kaolinite (Figure 2b). However, particle rejection rises with increasing adsorption of negative NOM, which induces particle stabilization (7, 8). Finally, we observed in the case of HA that the favored particle destabilization in the presence of calcium correlated with a lower adsorption of HA onto kaolinite. This suggests that calcium mitigates HA-kaolinite monolayer adsorption by lowering the negative particle charge. However, calcium increases the multilayer adsorption of HA, which occurs at higher concentration ranges due to HA-HA interactions. Particles were stable when 1 mg of C from HA was adsorbed (Figure 2b). However, the zeta potential further decreased to more negative values up to a particle load of 2 mg of C/g of kaolinite. The characteristics of the particles used for the UF experiments could be assessed on the basis of the above results. First, the adsorption isotherms and their model parameters allowed the NOM load on the kaolinite mass used (mkaolinite, 100 mg/L) to be calculated (eq 2). KFcn ) q ) (c0 - c)/mkaoliniteV

(2)

In this mass balance equation, combined with the Freundlich isotherm, KF is the Freundlich sorption coefficient and n the Freundlich exponent. The volume used is V, c is the equilibrium, and c0 is the initial NOM concentration. With the calculated NOM load of the particles, the particle size and zeta potential were determined from the relationship between the NOM load and the particle size (Figure 2b) and the NOM load and the zeta potential, respectively (Figure 2d). The presence of the particles could then be classified VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9131

TABLE 1. DOC Adsorbed on Kaolinite and Particle State in the Suspensions Used for the UF Experiments† model substance

mass adsorbed (mg of DOC)

specific load (mg of DOC/g)

particle state

0 mM Calcium 0.000 0.019 0.078 0.180 0.01 0.02

0.00 0.19 0.78 1.80 0.50 2.00

aggregated aggregated stable stable aggregated/stable stable

1.25 mM Calcium 0.09

0.90

aggregated

0 mM Calcium 0.000 0.02 0.06 0.08 0.005 0.01

0.00 0.20 0.60 0.77 0.50 0.9

aggregated aggregated stable stable aggregated/stable stable

1.25 mM Calcium 0.13

1.3

aggregated

(a) Humic Acid kaolinite, kaolinite, kaolinite, kaolinite, kaolinite, kaolinite,

100 mg/L-0 mg of C/L 100 mg/L-HA, 0.02 mg of C/L 100 mg/L-HA, 0.2 mg of C/L 100 mg/L-HA, 2 mg of C/L 10 mg/L-HA, 0.02 mg of C/L 10 mg/L-HA, 2 mg of C/L

kaolinite, 100 mg/L-HA, 2 mg of C/L

(b) Alginate kaolinite, kaolinite, kaolinite, kaolinite, kaolinite, kaolinite,

100 mg/L-0 mg of C/L 100 mg/L-alginate, 0.02 mg of C/L 100 mg/L-alginate, 0.2 mg of C/L 100 mg/L-alginate, 1.0 mg of C/L 10 mg/L-alginate, 0.02 mg of C/L 10 mg/L-alginate, 0.2 mg of C/L

kaolinite 100 mg/L-alginate, 0.2 mg of C/L

Adsorption parameters: HA-kaolinite, 0 mM Ca, KF ) 1.41, n ) 0.29; 1.25 mM Ca, KF ) 0.56, n ) 0.75; alginate-kaolinite, 0 mM Ca, KF ) 0.82, n ) 0.11; 1.25 mM Ca, KF ) 3.88, n ) 0.43. †

into two groups, stable and aggregated (Table 1). In this paper, this kind of particle characteristic is named the particle state. The particles were regarded as stable if a further increase of NOM load did not significantly change the particle zeta potential or size. “Aggregated/stable” indicates that the particles are still aggregated but are partly stabilized as indicated by lower measured particle sizes compared to that of the aggregated kaolinite in the absence of NOM (below half the diameter of the aggregated kaolinite in the absence of NOM). As shown in Table 1, in the case of NOM and kaolinite without calcium the particles were aggregated in combination with the lowest NOM concentration (0.02 mg of C/L), but stable in combination with the higher NOM concentrations (0.2 and 2 mg of C/L). In the presence of calcium, the particles were aggregated when the NOM concentrations were 2 mg of C/L HA and 0.2 mg of C/L alginate. Flux Decline. Figure 3 shows the flux decline of UF of solutions with individual and premixed compounds. Kaolinite alone causes only a minor flux decline, whereas both NOM compounds lead to substantial fouling. The flux decline of alginate at a concentration of 0.2 mg of C/L was comparable to that with 2 mg of C/L HA. It can be seen that in the mixtures of kaolinite (100 mg/L) with HA (2 mg of C/L) and alginate (0.2 mg of C/L) the total flux decrease was greater than the sum of the individually filtered substances. This indicates strong synergistic effects. An increase of the organic concentration from 0.02 to 0.2 mg of C/L resulted in a significant flux decrease in the combined NOM-kaolinite UF (Figure 3a,c). This indicates a large effect of the organics on the combined fouling at low NOM concentrations. Table 1 shows that the increase of organics from 0.02 to 0.2 mg C/L correlates with a change in the particle state from aggregated to stable. In the case of HA, a further concentration increase (from 0.2 to 2 mg of C/L) resulted in a much smaller effect. This correlates with the fact that the kaolinite particles are stable at both HA concentrations (0.2 and 2 mg of C/L). These results indicate that the synergistic effect of HA and kaolinite on fouling already occurs with small amounts of HA. They additionally reveal that a change in particle state from aggregated to stable by NOM plays a crucial role in the flux decline. 9132

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

A further increase in the alginate concentration (from 0.2 to 1 mg C/L) led to another significant decrease in flux. This indicates a successive buildup of a relatively dense fouling layer induced by the presence of alginate and demonstrates the dominant fouling by alginate already observed in our previous study (15). A 10-fold decrease of the particle concentration resulted in a lower flux loss with both NOM substances, which is explained by a lower deposited total mass independent of the fouling layer structure. The particle concentration has a greater impact in combination with HA than with alginate, which gives further proof of the successive dominance of the already described fouling by alginate alone. Calcium has a reducing effect on fouling in combined NOM-particle filtration. Table 1 shows that the particles are aggregated with both types of NOM in the UF solutions with calcium. This indicates that the aggregation of the NOMkaolinite particles by calcium induced a lower flux decline than in the case of the stabilized NOM-kaolinite particles in the absence of calcium. In the case of NOM only, calcium had a minor effect on flux loss by HA, but raised the flux loss detrimentally with alginate. This is explained by increased intermolecular alginate interactions and alginate-gel formation by calcium (15). Flux Decline Reversibility. Reversible fouling refers to fouling reversed by membrane cleaning and can be divided into fouling reversible by hydraulic backwashing and fouling reversible by chemical backwashing (25). In this study the hydraulic reversible fouling was assessed by the flux decline reversibility, more specifically by the flux after the hydraulic backwash. Parts b and d of Figure 3 show that in the absence of calcium the reversibility decreases with increasing NOM concentration independent of the particle concentration. This indicates that the initial NOM concentration determines the irreversible fouling by NOM-membrane interactions that cannot be prevented by kaolinite. The irreversible fouling is stronger with HA than with alginate, a fact attributed to hydrophobic HA-membrane interactions. Calcium increases the irreversible NOM fouling without particles, especially in the case of alginate. This can be explained by ion bridging between NOM and the membrane,

FIGURE 3. Relative flux decline at the end of a filtration cycle before backwash (Jbb) for different concentration combinations of (a) HA and kaolinite and (b) alginate and kaolinite and after hydraulic backwash (Jab) for different concentration combinations of (c) HA and kaolinite and (d) alginate and kaolinite. Key: bold, experiments using the same NOM and kaolinite concentrations as those with NOM and kaolinite alone; shaded box, experiments performed with calcium (1.25 mM). J0 ) initial membrane water (DI) flux.

FIGURE 4. Transmission of HA with 0 and 1.25 mM calcium. as both are negatively charged, and/or reduction of electrostatic rejection between NOM and the membrane due to charge shielding by calcium. However, the particles reduced the irreversible NOM/calcium fouling. This is attributed to calcium associated with kaolinite and the fact that less free calcium is available for membrane-NOM bridging, as observed in another study (26). DOC Transmission through the Membrane. The DOC transmission is defined as the measured permeate concentration divided by the measured feed concentration and is shown in Figure 4. The feed concentration values refer to the NOM equilibrium concentration after adsorption of NOM onto kaolinite.

The transmission of HA in combination with kaolinite is only a little lower than when HA is filtered alone. There is consequently no additional retention of HA in the combined system apart from the HA adsorbed onto kaolinite. Moreover, HA transmission increases over the filtrated volume, which indicates initial adsorption onto the membrane. The similar transmission of HA in combination with kaolinite and of HA alone indicates that the same amount of HA is adsorbed onto the membrane independently of the amount adsorbed onto the kaolinite. These observations support the conclusion that kaolinite did not affect the irreversible HA-membrane interactions and hence irreversible fouling in our experiments. However, it has to be considered that the HA adsorption onto kaolinite was relatively low (less than 10%) in the experiment with kaolinite and 2 mg of C/L HA. It therefore had only an insignificant impact on the HA adsorption onto kaolinite. At lower initial HA concentrations, the relative adsorption is larger (Figure 1), and it is likely that the HA adsorption onto kaolinite can reduce the HA in solution to such low values that it affects the HA membrane adsorption significantly. Calcium reduced the HA transmission. We showed in previous studies that the alginate transmission initially decreases before stable values are reached. This suggests that initial membrane-pore blockage followed by cake formation is the major fouling mechanism (15, 27). Furthermore, it was shown that alginate transmission is significantly lower in the presence of kaolinite than when alginate is filtered alone. This indicated that another mechanism apart from kaolinite adsorption is responsible for the retention. Kaolinite also deposits on the membrane to form VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9133

FIGURE 5. Schematic fouling model of (a) kaolinite, (b) HA and kaolinite in the absence of calcium, (c) alginate and kaolinite in the absence of calcium, (d) HA and kaolinite in the presence of calcium, and (e) alginate and kaolinite in the presence of calcium. a cake. The presence of kaolinite can consequently interfere with individual alginate fouling due to its steric effects, such as size exclusion by the kaolinite cake or alginate trapping during deposition. Fouling Model of Combined NOM-Kaolinite Ultrafiltration. On the basis of our experimental results, a model for combined NOM-particle fouling is proposed and shown schematically in Figure 5. The model explains the differences between the fouling layer structures built during individual and combined NOM and kaolinite UF in a qualitative way. To simplify the model, the natural kaolinite particles rather resembling sheets are represented as spheres. A comparison between the characteristics of the particles in suspension (Table 1) and the corresponding flux decline (Figure 3) shows that for a certain particle concentration the (reversible) flux decline is significantly bigger in the case of stable particles than of aggregated ones. This indicates that the particle state is a crucial factor for the resistance caused by a fouling layer. Kaolinite loosely aggregates in solution at neutral pH values, although its zeta potential is relatively strongly negative (-47 mV). However, kaolinite aggregation is explained by “face-edge” charge attraction of its negatively charged silica surface and its conditional positively charged edges (8, 20). Such aggregation leads to a fouling cake of higher porosity compared to that formed by stabilized particles (Figure 5a). High porosity results in a relatively low resistance toward flux (eq 1) as confirmed by the low flux decline of the kaolinite suspension (Figure 3). In contrast, stabilized particles lead to a denser cake with a higher 9134

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

resistance than aggregated particles (Figure 5b). It was shown that in the absence of calcium the particles are stabilized at NOM loads of 0.5 mg of C/g of kaolinite (Table 1). Such particle stabilization explains the synergistic fouling effect of NOM and kaolinite, especially in the case of HA (Figure 3a). This is supported by our previous finding that the premixing of HA and kaolinite, during which HA has time to adsorb onto kaolinite, is essential for their synergistic fouling effect (27). Calcium can enhance the aggregation of particles with NOM adsorbed on their surface (Figure 2). In the case of both solution mixtures with calcium, namely, kaolinite-HA and kaolinite-alginate, the particles were aggregated (Table 1). This led to the formation of a fouling layer of relatively high porosity (Figure 5d,e) and of a low hydraulic resistance as demonstrated in our combined substance experiments in the presence of calcium (Figure 3). Similar relationships between particle aggregation and hydraulic resistance have been shown in earlier flocculation studies using flocculation chemicals (28, 29). However, this study showed that the aggregation of naturally occurring particles can also influence membrane filtration without pretreatment/flocculation. In the case of alginate, it was previously shown that synergistic fouling effects of alginate and kaolinite occur with premixed compounds as well as with subsequent filtration of kaolinite and alginate (27). This indicates that kaolinite stabilization is not the only synergy mechanism for alginatekaolinite UF. Rather, due to the high molecular weight of alginate and its longitudinal shape, it is likely to be effectively retained by the membrane (Figure 4b) and can fill up the

interstices between the kaolinite particles (see Figure 5c). This assumption is further supported by the increased alginate retention in the presence of kaolinite (Figure 4b). Moreover, we showed that irreversible NOM membrane fouling was not affected significantly by particles (without calcium). The HA-kaolinite adsorption as such cannot prevent irreversible HA fouling appreciably since the amount of HA left in solution is still high (>90%). In the case of alginate, the irreversibility was relatively low with and without particles. In the presence of calcium, large irreversible fouling by alginate was observed. In a previous study, calcium was found to have an insignificant effect on irreversible fouling by alginate, although a larger amount of alginate was used. It seems that the calcium/ alginate ratio is critical for the effect of calcium on irreversible fouling by alginate, which is probably due to the large affinity of calcium toward alginate. However, particles reduced the irreversible flux decline by alginate. This is explained by the competition between NOM and particles for calcium (25). In a previous study we showed that combination of the three model compounds (humic acid, alginate, kaolinite) led to a flux decrease more detrimental compared to that of the experiments shown here with kaolinite and humic acid or alginate (27). It can be assumed on the basis of our model that the synergistic effect of these substances is due to the interplay of their individual fouling mechanisms. This is supported by studies reported in the literature on the synergistic effects of organic and colloidal fouling as well as of fouling by various organic substances (15, 26). It is essential to understand fouling mechanisms to optimize membrane filtration, and this also applies to pretreatment processes. Our model shows that the NOM/ particle ratio and the calcium concentration are crucial factors for fouling. Knowledge of these two factors for specific feedwater can improve the prediction of its membrane fouling characteristics and thus operation strategies. In practice, the NOM/particle ratio can be influenced by pretreatment. For example, activated carbon can adsorb NOM, thus reducing the NOM/particle ratio. According to our model, this favors particle destabilization and the formation of fouling cakes of lower resistance. The removal of polysaccharides is a critical factor, since they were shown to be removed only to a small extent during pretreatment steps such as ozone treatment and activated carbon adsorption (30, 31). Our results show that an appreciation of particle-NOM interactions in natural waters and during UF is crucial for understanding fouling mechanisms. This is of particular importance because the raw water characteristics may vary greatly with respect to the NOM and particle concentrations occurring in natural waters. Our study results may improve predictions of UF membrane fouling by natural waters and may be used to optimize UF processes by the application of different pretreatment steps. Further studies on NOM-particle interactions using particles of various size classes and various kinds of particles (e.g., goethite, organic detritus) may be conducted to extend the knowledge on the role of NOM-particle interactions in UF fouling.

Acknowledgments We acknowledge WVZ (Wasserversorgung Zu ¨ rich) and WABAG AG, Winterthur, Switzerland, for their support and successful collaboration within the project WAVE 21. Se´bastien Meylan and Jacqueline Traber are acknowledged for their support in the analyses and experimental work.

Supporting Information Available More detailed information on the characteristics of the model substances kaolinite, alginate, and humic acid and on the

membrane characteristics. This information is available free of charge via the Internet at http://pubs.acs.org.

Literature Cited (1) Clark, M. M.; Lucas, P. Diffusion and partitioning of humic acid in a porous ultrafiltration membrane. J. Membr. Sci. 1998, 143, 13–25. (2) Jarusutthirak, C.; Amy, G.; Croue, J.-Ph. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 2002, 145, 247–255. (3) Kimura, K.; Hane, Y.; Watanabe, Y.; Amy, G.; Ohkuma, N. Irreversible membrane fouling during ultrafiltration of surface water. Water Res. 2004, 38, 3431–3441. (4) Yamamura, H.; Chae, S.; Kimura, K.; Watanabe, Y. Transition in fouling mechanism in microfiltration of a surface water. Water Res. 2007, 41, 3812–3822. (5) Faibish, R. S.; Elimelech, M.; Cohen, Y. Effect of interparticle electrostatic double layer interactions on permeate flux decline in crossflow membrane filtration of colloidal suspensions: An experimental investigation. J. Colloid Interface Sci. 1998, 204, 77–86. (6) Wiesner, M. R. Morphology of particle deposits. J. Environ. Eng. 1999, 125, 1124–1132. (7) O’Melia, C. R.; Tiller, C. L. Physicochemical aggregation and deposition in aquatic environments. Environmental Particles; Lewis Publishers: Boca Raton, FL, 1993; Vol. 2. (8) Stumm, W.; Morgan, J. Aquatic Chemistry; John Wiley & Sons, Inc.: New York, 1995. (9) Buffle, J.; Wilkinson, K. J.; Stoll, S.; Filella, M.; Zhang, J. W. A generalized description of aquatic colloidal interactions: The three-colloidal component approach. Environ. Sci. Technol. 1998, 32, 2887–2899. (10) Wilkinson, K. J.; Balnois, E.; Leppard, G. G.; Buffle, J. Characteristic features of the major components of freshwater colloidal organic matter revealed by transmission electron and atomic force microscopy. Colloids Surf., A 1999, 155, 287–310. (11) Tarabara, V. V.; Pierrisnard, F.; Parron, C.; Bottero, J.-Y.; Wiesner, M. R. Morphology of deposits formed from chemically heterogeneous suspensions: Application to membrane filtration. J. Colloid Interface Sci. 2002, 256, 367–377. (12) te Poele, S.; Menkveld, W.; Boom, J.; van Bragt, W. Effluent treatment by multi-media filtration, microfiltration and ultrafiltration: Results of a pilot investigation at WWTP Hoek van Holland. Water Sci. Technol. 2005, 52, 99–105. (13) Waite, T. D.; Scha¨fer, A. I.; Fane, A. G.; Heuer, A. Colloidal fouling of ultrafiltration membranes: Impact of aggregate structure and size. J. Colloid Interface Sci. 1999, 212, 264–274. (14) Bereznitski, Y.; Jaroniec, M.; Maurice, P. Adsorption characterization of two Clay Minerals Society standard kaolinites. J. Colloid Interface Sci. 1998, 205, 528–530. (15) Jermann, D.; Pronk, W.; Meylan, S.; Boller, M. Interplay of different NOM fouling mechanisms during ultrafiltration for drinking water production. Water Res. 2007, 41, 1713–1722. (16) Sontheimer, H., Frick, B., Fettig, J., Ho¨rner, G., Hubele, C., Zimmer, G. Adsorptionsverfahren zur Wasserreinigung; EnglerBunte-Institut der Universita¨t Karlsruhe: Karlsruhe, Germany, 1985. (17) Mulder, M. Basic Principles of Membrane Technology; Kluwer Academic Publisher: Dordrecht, The Netherlands, 2000. (18) Jiang, Q.; Logan, B. E. Fractal dimensions of aggregates determined from steady-state size distributions. Environ. Sci. Technol. 1991, 25, 2031–2038. (19) Tombacz, E.; Filipcsei, G.; Szekeres, M.; Gingl, Z. Particle aggregation in complex aquatic systems. Colloids Surf., A 1999, 151, 233–244. (20) Tombacz, E.; Szekeres, M. Surface charge heterogeneity of kaolinite in aqueous suspension in comparison with montmorillonite. Appl. Clay Sci. 2006, 34, 105–124. (21) Feng, X.; Simpson, A. J.; Simpson, M. J. Chemical and mineralogical controls on humic acid sorption to clay mineral surfaces. Org. Geochem. 2005, 36, 1553–1566. (22) Tiller, C. L.; O’Melia, C. R. Natural organic matter and colloidal stability: Models and measurements. Colloids Surf., A 1993, 73, 89–102. (23) Ahn, W.-Y.; Kalinichev, A. G.; Clark, M. M. Effects of background cations of the fouling of polyethersulfone membranes by natural organic matter: Experimental and molecular modeling study. J. Membr. Sci. 2008, 309, 128–140. (24) Perry, T. D.; Cygan, R. T.; Mitchell, R. Molecular modeling of alginic acid: Interactions with calcium ions and calcite surfaces. Geochim. Cosmochim. Acta 2006, 70, 3508–3532. VOL. 42, NO. 24, 2008 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

9135

(25) Kulkarni, S. K., Funk, E. W., Li, N. N. Introduction and definitions. In Membrane Handbook; Chapman and Hall: New York, 1992. (26) Li, Q.; Elimelech, M. Synergistic effects in combined fouling of a loose nanofiltration membrane by colloidal materials and natural organic matter. J. Membr. Sci. 2006, 278, 72–82. (27) Jermann, D.; Pronk, W.; Ka¨gi, R.; Halbeisen, M.; Boller, M. Influence of interactions between NOM and particles on UF fouling mechanisms. Water Res. 2008, 42, 3870–3878. (28) Choi, K.; Young-june, D.; Brian, A. In-line coagulation with low-pressure membrane filtration. Water Res. 2004, 38, 4271– 4281.

9136

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 42, NO. 24, 2008

(29) Pikkarainen, A. T.; Judd, S. J.; Jokela, J.; Gillberg, L. Precoagulation for microfiltration of an upland surface water. Water Res. 2004, 38, 455–465. (30) Velten, S.; Traber, J.; Knappe, D. R.; von Gunten, U.; Boller, M.; Meylan, S. Evolution of natural organic matter adsorption on granular activated carbon in a packed bed reactor. Manuscript in preparation. (31) Widrig, L. D.; Gray, K. A.; McAuliffe, K. S. Removal of algalderived organic material by preozonation and coagulation: monitoring changes in organic quality by pyrolysis-GC-MS. Water Resour. 1996, 30, 2621–2632.

ES800654P