Article pubs.acs.org/est
Mechanisms of Membrane Fouling Control by Integrated Magnetic Ion Exchange and Coagulation Haiou Huang,*,† Hyun-Hee Cho,‡ Joseph G. Jacangelo,†,§ and Kellogg J. Schwab† †
Johns Hopkins University, Bloomberg School of Public Health, 615 North Wolfe Street, Baltimore, Maryland 21205, United States Department of Civil and Environmental Engineering, Hanyang University, 17 Haengdang-dong, Seongdong-gu, Seoul 133-791, Republic of Korea § MWH Global, 40814 Stoneburner Mill Lane, Lovettsville, Virginia 20180, United States ‡
S Supporting Information *
ABSTRACT: Colloidal natural organic matter (NOM) is an important foulant to low-pressure membranes (LPMs) employed in drinking water treatment. Removal of colloidal NOM by magnetic ion exchange (MIEX), coagulation, and integrated MIEX and coagulation was investigated in this study to determine the relationship between colloidal NOM removal and membrane fouling reduction. The results showed that coagulation did not selectively remove colloidal NOM and the optimal coagulant dose was primarily determined by the concentration of humic substances. Comparatively, MIEX pretreatment preferentially removed humic substances and reduced the coagulant dose needed for colloidal NOM removal as a result of coagulation stoichiometry. A matched-pair analysis showed that integrated MIEX and coagulation pretreatment at much lower coagulant doses was as effective as coagulation in reducing membrane fouling. It is concluded that integrated MIEX and coagulation is potentially a viable pretreatment approach to reduce membrane fouling and in general removal of colloidal NOM in feedwater is an effective approach for membrane fouling control and should be considered in the research, development, and application of novel LPMbased treatment processes.
■
INTRODUCTION Colloidal natural organic matter (NOM) is operationally defined in this paper as NOM having high molecular weight based on size fractionation using size-exclusion chromatography with an organic carbon detector (SEC-OCD). As found by Huber et al.,1 colloidal NOM consists of primarily polysaccharides with some contribution from nitrogen-containing biopolymers. Characterization of organic NOM using nuclear magnetic resonance,2 transmission electron and atomic force microscopy,3 and environmental scanning electron microscopy4 revealed the predominance of polysaccharide material and presence of humic aggregates in the composition of colloidal NOM. Besides their large sizes, the polysaccharide fraction of colloidal NOM seems to be uncharged1 while the humic fraction appears to contain negatively charged functional groups that are reactive with divalent cations in water.4 Previous studies have found that organic fouling of lowpressure membranes (LPMs) employed in drinking water treatment is strongly associated with colloidal NOM in natural surface water,4−6 instead of organics with UV absorbance.7 Due to similarity between the size of colloidal NOM and membrane pore size, colloidal NOM can decrease the permeability of LPMs through pore blockage or cake layer formation.4,8 As a result, LPMs will gradually lose their water permeability as © 2012 American Chemical Society
colloidal NOM accumulate on membrane surfaces or plug into membrane pores. These fouled membranes need to be cleaned either hydraulically or chemically to restore their permeability, which adversely impacts the sustainability and productivity of LPMs for drinking water treatment. Therefore, removal of colloidal NOM from feedwater is potentially an important approach for the prevention/reduction of LPM fouling. Currently, there is no pretreatment technology specially designed for the removal of colloidal NOM from feedwater. Since colloidal NOM comprises a very small fraction of NOM,1 chemical pretreatments for LPM filtration are often designed for the removal of major NOM such as humic substances that can be the precursors of disinfection byproduct or trace organics of health concerns but may have limited efficiency in removing colloidal NOM. In our previous review,9 coagulation was found to be the most widely used pretreatment for membrane fouling control. However, the hydrolytic products of trivalent coagulants are known to react with a variety of negatively charged NOM moieties, including humic subReceived: Revised: Accepted: Published: 10711
March 14, 2012 June 29, 2012 August 27, 2012 August 27, 2012 dx.doi.org/10.1021/es300995x | Environ. Sci. Technol. 2012, 46, 10711−10717
Environmental Science & Technology
Article
stances.10 Presence of these competitors is expected to interfere with the removal of colloidal NOM by coagulation. There is not systematic understanding on the effects of competitive NOM components on the removal of colloidal NOM during coagulation as well as its consequences on LPM fouling. Information in this regard is important to the design and operation of integrated LPM filtration systems for sustainable water treatment. Therefore, this study was undertaken with two major objectives: (1) to evaluate the removal of colloidal NOM by coagulation before and after the removal of competitive NOM component; and (2) to determine the resulting effects of single-stage and integrated pretreatment on membrane fouling reduction. Magnetic ion exchange (MIEX) was chosen in this study as an effective pretreatment to selectively remove humic substances and decrease their competition with colloidal NOM during coagulation treatment. As an anion-exchange resin, MIEX can remove inorganic and organic anions in natural water by exchanging chloride or biocarbonate ions bound to its cationic backbones and subsequently separate from the treated water by magnetic-induced aggregation and sedimentation. Boyer et al. found that MIEX preferentially remove humic substances over other NOM constituents.11 This selective removal is probably due to the hydrophobicity and acidity of humic substances.12 In comparison, colloidal NOM is primarily composed of biopolymers that have neutral charges.1 Therefore, as confirmed in this study, most colloidal NOM are not expected to be removed by MIEX13,14 or other ion-exchange resins.1,14 As a result, MIEX treatment alone did not significantly affect the fouling of LPMs.13
Pretreatment Chemicals. Virgin MIEX resin sample was received from Orica Watercare as a concentrated suspension. During this study, aliquots of this suspension were transferred into 50 mL polypropylene centrifugal tubes; ultrapure water was added to the tubes to prepare stock suspensions with a MIEX volume fraction of 50%. The volume fraction of MIEX was measured by allowing MIEX to settle for 30 min and measuring the volume of settled MIEX resin in a total of 50 mL of suspension using the volume grade on the centrifugal tubes. Reagent-grade potassium alum, or aluminum potassium sulfate (KAl(SO4)2·12H2O, formula weight = 474.39 Da), was dissolved in ultrapure water to prepare a stock solution of 50 g/L. Jar Tests and MIEX Pretreatment. Different aliquots of coagulant stock solution were added to a series of 250 mL glass beakers, each containing 200 mL of natural water or water pretreated with 2.5 mL/L of virgin MIEX. These beakers were then placed on the jar tester (Phipps & Birds programmable jar tester, model PB-900). During each jar test, the alum solution was dosed while the water in the beakers was stirred at 100 rpm. After dosing the alum, the water in the beakers was mixed rapidly at 100 rpm for 2 min, followed by slow mixing at 25 rpm for 30 min. The treated water was allowed to settle for 60 min. Finally, 100 mL of the supernatant was carefully decanted from each beaker and used for water quality analyses. The optimal alum dose was determined as the minimum dose required to effectively remove the turbidity in the natural water, which shows as the starting point of a plateau region in the settled turbidity versus coagulation dose curve (Figure S1, Supporting Information). For MIEX pretreatment, virgin MIEX resin was added to the natural water at 2.5 mL/L, and the suspensions were stirred at 150 rpm for 15 min on the jar tester and settled for 10 min. After settling, the supernatant was poured out for subsequent coagulation experiment. According to the manufacturer, a MIEX dose of 2.5 mL/L was considered to be at the upper range of practical doses for full-scale water treatment plants. In comparison, a MIEX dose of 5 mL/L was found in preliminary experiments to be cost effective for dissolved organic carbon (DOC) removal (data not shown). Low-Pressure Membranes and Filtration System. Three commercially available, low-pressure, hollow-fiber membranes were tested for the impact of coagulation pretreatment. Properties of these membranes are summarized in Table 2. For the purpose of bench-scale filtration testing, mini-modules were prepared with these hollow-fiber membranes using an established method.15 The effective surface areas of membrane A, membrane B, and membrane C modules were 0.00548, 0.00503, and 0.00302 m2, respectively. Cleaned membrane modules were installed on a bench-scale membrane testing unit. This testing unit was configured in a submerged outside-in mode, pressurized outside-in, or pressurized insideout mode for the testing of membrane A, membrane B, and membrane C, respectively. A schematic diagram of the three
■
MATERIALS AND METHODS Natural Surface Water. Four batches of natural surface water were sampled from the source water of a drinking water treatment plant in California, USA, and transported to the lab via overnight shipping. Upon arrival, all water samples were immediately prefiltered using 1.2 μm glass fiber filters (Whatman GF/C) to remove coarse materials and then stored at 4 °C in the dark. Characteristics of the prefiltered water samples are presented in Table 1. Table 1. Characteristics of the Natural Surface Waters water sample parameter pH DOC UV254 absorbance specific UV absorbance alkalinity conductivity magnesium calcium
unit
batch 1 batch 2 batch 3 batch 4
mg/L cm−1 L/mg·m
8.04 5.5 0.189 3.4
8.28 3.6 0.118 3.3
8.27 8.7 0.288 3.3
8.40 16.8 0.551 3.3
mg CaCO3/L μS/cm mg/L mg/L
130 447 21.9 22.9
85 426 12.5 14.6
141 355 4.09 21.7
144 257 1.94 11.5
Table 2. Properties of the LPMs Employed in the Study membrane
material
membrane type
OD (mm)
ID (mm)
pure water permeability (L/m2·h/bar, 23 ± 1 °C)
nominal pore size (μm)
flow pattern
A B C
PVDF PVDF PES/PVP
MF UF UF
0.8 0.8 1.1
0.5 0.5 0.8
385.9 ± 25.2 208.6 ± 9.2 1488 ± 160
0.1 0.02 0.03
submerged, outside-in pressurized, outside-in pressurized, inside-out
10712
dx.doi.org/10.1021/es300995x | Environ. Sci. Technol. 2012, 46, 10711−10717
Environmental Science & Technology
■
RESULTS AND DISCUSSION NOM Composition in the Studied Water. NOM in the natural surface water showed multiple components based on the OCD chromatograms. As shown in Figure 1A and B, there
configurations is shown in Figure S2 of the Supporting Information. Bench-Scale Low-Pressure Membrane Filtration. LPM filtration experiments were conducted in four sequential steps: (1) 30 min filtration of ultrapure water to determine the baseline membrane specific flux or Js0; (2) 100 min constant flux filtration of the natural water pretreated by MIEX and/or coagulation; (3) a 1 min hydraulic backwash at the end of step 2, followed by 15 min filtration of the pretreated natural water; and (4) a chlorine cleaning at the end of step 3, followed by another 15 min filtration of the pretreated natural water. The waters were filtered at a constant permeate flux of approximately 82 ± 2 L/m2·h (LMH) in all experiments. The hydraulic backwash fluxes were 82 LMH for membrane A and membrane B, and 246 LMH for membrane C. Chlorine cleaning was conducted by soaking fouled membranes in sodium hypochlorite solution containing 500 mg/L of free chlorine (pH ∼ 10.3) for 30 min. The membrane fouling results were quantified using the Unified Membrane Fouling Index (UMFI), which is generally defined with the following expression:16 Js0
UMFI =
Js
Article
−1 Vs
Figure 1. Comparison of the effects of MIEX pretreatment, precoagulation, and integrated MIEX and coagulation pretreatment on the size fractionation of (A) DOC, (B) HMW DOC, and (C) DOC with UV254 absorbance in batch 4 of the natural surface water. MIEX dose = 2.5 mL/L; coagulant dose = 7.39 mg Al/L (coagulation only) or 3.70 mg Al/L (combined with 2.5 mL/L MIEX).
(1)
where Js0/Js and Vs are the inverse of the normalized specific flux (dimensionless) and the unit permeate throughput [L/m2], respectively, and Js or Js0 is specific permeate flux that is defined as the permeate flux normalized to transmembrane pressure [L/m2/bar]. UMFIT (UMFI for total fouling) was calculated with linear fitting of the Js0/Js versus Vs results obtained in step 2 of the filtration experiment; UMFIR (UMFI for hydraulically irreversible fouling) was computed by inserting the experimental values of Js0/Js and Vs obtained at the onset of step 3 into eq 1; UMFIC (UMFI for chemically irreversible fouling) was obtained by inserting the values of Js0/Js and Vs at the onset of step 4 into eq 1. Methods for the calculation of UMFI values have been described previously.16 Size-Exclusion Chromatography. Size-exclusion chromatography (SEC) with UV and organic carbon detectors (SECUV/OCD) was used in this study to determine the size fractionation of NOM in natural and treated water samples following the work of Lee et al.17 and Huber et al.1 The SEC column was a TSK-GEL G4000PWxl column with an effective separation range of 2000−300 000 for polyethylene glycol and polyethylene oxide (PEG and PEO) standards. The OCD was a modified GE Sievers 900 portable TOC analyzer. Phosphate buffer solution consisting of 1.2 g/L NaH2HPO4 and 2.5 g/L KH2PO4 (pH = 6.85) was employed as the mobile phase, and the flow rate was controlled at 1.00 mL/min. Water samples were prefiltered through 0.2 μm PES membrane and injected to the column at a volume of 500 μL. A chromatograph data system (BaseLine Chromtech) was used to acquire signals from the UV detector and the OCD and to calculate the individual and total areas of NOM peaks. The corresponding DOC concentration for each NOM peak was then calculated based on the percentage of its peak area and the DOC concentration of the water sample (measured separately using the TOC analyzer with grab samples).
are a major peak at an elution time of 9.5 min, a small peak at approximately 5 min (∼the exclusion limit of the column), and a small shoulder peak at approximately 11.5 min. According to Huber et al.,1 the first two OCD peaks are ascribed to colloidal NOM and humic substances, respectively, and the small shoulder peak is contributed by other small NOM constituents, such as low MW acids or neutrals. According to the manufacturer, the average pore size of the particulate medium in the SEC column is 50 nm, which suggests that colloidal NOM in the studied water probably has diameters equal or above 50 nm. This estimated size for colloidal NOM is close to an average diameter of approximately 80 nm found in our previous study using environmental scanning electron microscopy4 and in the size range of 10−100 nm estimated by Howe and Clark based on ultrafiltration separation.5 Compared to the OCD chromatogram, there were two overlapping peaks on the UV254 absorbance chromatogram (Figure 1C). In addition to the humic substance peak that is also shown in the OCD chromatogram, another peak was detected at approximately 10 min. This peak partially overlapped with the humic substance peak and is probably ascribed to low MW humic substances.1 Due to the low resolution of the SEC column employed in this study for small solutes, the low MW humic peak was completely overlapped with the dominating humic substance peak and is not distinguishable on the OCD chromatogram (Figure 1A). Colloidal NOM Removal by Ion Exchange or Coagulation. As an anion-exchange resin, MIEX preferentially remove negatively charged constituents in natural water, particularly humic acids and fulvic acids.11,12 For the studied natural surface water, MIEX treatment at 2.5 mL/L significantly 10713
dx.doi.org/10.1021/es300995x | Environ. Sci. Technol. 2012, 46, 10711−10717
Environmental Science & Technology
Article
Table 3. Concentrations of Different NOM Fractions in Natural Water after Different Pretreatmentsa DOC (mg/L) all fractions
a
colloidal NOM
humic substances
small NOM
pretreatment type
B1
B3
B4
B1
B3
B4
B1
B3
B4
B1
B3
B4
no pretreatment MIEX (5 mL/L) coagulation MIEX (2.5 mL/L) plus coagulation
5.1 1.9 3.6 2.0
8.8 2.4 5.2 2.2
16.8 6.8 6.0 3.6
0.15 0.13 0.07 0.06
0.25 0.20 0.12 0.07
1.2 0.65 0.13 0.13
4.5 1.4 3.2 1.6
7.9 1.9 4.7 1.9
14.7 5.4 5.2 2.9
0.47 0.39 0.32 0.31
0.74 0.34 0.38 0.29
1.0 0.81 0.73 0.55
B1, B3, and B4 refer to batch 1, batch 3, and batch 4 of the natural surface water, respectively.
coagulant dose, since approximately 64% of humic substances had been removed by upstream MIEX treatment. Similar trends were also observed for small NOM. Despite the low coagulant dose, both the small NOM peak on the SEC-OCD chromatograms (Figure 1A) and the small humic substance peak on the SEC-UV chromatograms (Figure 1C) were reduced by integrated MIEX and coagulation as compared to coagulation alone. Consequently, the integrated MIEX and coagulation resulted in the most efficient NOM removal. Practically, preferential removal of humic substances by MIEX may enhance the removal of nonhumic NOM constituents for LPM-based drinking water treatment integrated with other chemical pretreatments. A stoichiometric relationship was found between optimal coagulant dose and initial DOC of different batches of the water (Figure 2). A linear fitting of the experimental data using
decreased the dominating OCD peak for humic substances (Figure 1A), while the colloidal NOM peak was decreased to a lesser extent (Figure 1B). These trends were consistent with the finding of Drikas et al.12 On the other hand, the two UV254absorbing peaks corresponding to large and small humic substances showed similar decreases after MIEX treatment (Figure 1C), suggesting that both NOM fractions consist of anionic components that can be removed by MIEX. Overall, the SEC results indicate that MIEX treatment at 2.5 mL/L preferentially removed the major NOM component, that is, humic substances, in the studied water. At a dose of 5 mL/L, MIEX removed approximately 63.7% of humic substances but only 38.3% of colloidal NOM and 19.8% of small NOM in the batch 4 sample (Table 3). As a result, the composition of NOM in the water changed after MIEX treatment. Colloidal NOM content increased from 7% to 10% while the humic content decreased from 88% to 80%. Coagulation of the untreated batch 4 water at 7.39 mg Al/L simultaneously decreased the peaks for colloidal NOM and humic substances (Figure 1). At a dose of 5 mL/L, the colloidal NOM was removed by 87.6% as compared to 38.3% by MIEX whereas humic substances were removed to similar extents in two pretreatments (Table 3). Correspondingly, the content of colloidal NOM decreased from 7% to 2% and that of humic substances remained unchanged. However, the peak for large humic substances was almost completely eliminated after coagulation while the peak for small humic substances only decreased approximately by half (Figure 1C); this suggests that coagulation is more effective in removing large humic substances than smaller ones. Similar trends were observed during the treatment of batches 1 and 3 of the natural surface water. As shown in Table 3, coagulation removed approximately 52% and 56% of colloidal NOM in batches 1 and 3 of the natural water, respectively, as compared to 15% and 21% by MIEX. After MIEX pretreatment, colloidal NOM accounted for approximately 7% and 8% of the total DOC in batches 1 and 3 of the water, respectively, greater than 3% measured in the untreated waters. Effect of Combined MIEX and Coagulation on NOM Removal. Combination of coagulation and MIEX pretreatments had synergistic effects on the removal of colloidal NOM and other NOM constituents. After batch 4 of the natural water was pretreated by 2.5 mL/L of MIEX resin, the optimal coagulant dose decreased to 3.70 mg Al/L, which was approximately half of what was required for untreated water. However, this decrease in coagulant dose did not affect the removal of colloidal NOM (Table 3), suggesting that the optimal coagulant dose is governed by the concentration of humic substances (rather than colloidal NOM or small NOM constituents) in the natural water. Therefore, effective removal of colloidal NOM in batch 4 water was achieved at a low
Figure 2. Relationship between initial water DOC and the optimal coagulant dose determined in the coagulation of four batches of the natural surface water with and without MIEX pretreatment. The solid line and the dashed lines represent the least-squares linear regression line (n = 7) and the 95% confidence intervals, respectively. Optimal coagulant dose for batch 2 of the natural water was unavailable. The jar test results used in the determination of the optimal doses are presented in Figure S1 of the Supporting Information.
a least-squares approach yielded a correlation ratio of approximately 0.43 mg Al/mg DOC. In contrast, Shin et al. reported a ratio of 0.47 mg Al/mg DOC for the coagulation of synthetic waters containing Great Dismal Swamp Water NOM at pH 7.18 MIEX pretreatment did not affect the stoichiometry between alum demand and initial DOC of the water. This is attributable to the dominance of humic substances in NOM composition and their compatibility with both MIEX and coagulation removal. As shown in Table 3, humic substances accounted for approximately 88%, 89%, and 88% of the NOM in batches 1, 2, and 4 of the natural water, respectively. In the case of MIEX treatment, the amount of DOC removed by MIEX pretreatment resulted primarily from a decrease in the humic fraction (3.1 mg/L of 3.2 mg/L for batch 1, 6.0 mg/L of 6.4 mg/L for batch 3, and 9.3 mg/L of 10.0 mg/L for batch 4). 10714
dx.doi.org/10.1021/es300995x | Environ. Sci. Technol. 2012, 46, 10711−10717
Environmental Science & Technology
Article
backwash) by 50% to 80% for the studied membranes. In general, a greater UMFIT value indicates faster total membrane fouling during filtration and a greater UMFIR or UMFIC value indicates faster accumulation of irreversible fouling. The value of UMFIR for the three membranes, a measure of the residual fouling after hydraulic backwash, also decreased after coagulation or integrated MIEX and coagulation treatment (Figure 3). However, the decrease in UMFIC values (residual fouling after chlorine cleaning) was distinctive for membrane A and membrane C but not for membrane B. This probably resulted from the low level of chemically irreversible fouling (UMFIC < 0.001 m2/L) that was difficult to be quantified by short-term filtration experiments. Compared to coagulation and integrated MIEX and coagulation, MIEX reduced the total and hydraulically irreversible fouling to much less extent, consistent with low removal of colloidal NOM (Table 3). Similar trends were observed for batch 4 water (Figure S3, Supporting Information). The decrease in coagulant dose after MIEX pretreatment did not affect the effectiveness of coagulation in membrane fouling control. A matched pair analysis of the UMFI values for coagulation alone and integrated MIEX and coagulation pretreatment was conducted for three membranes and two waters. As shown in Figure 4, the line corresponding to y = 0
In comparison, DOC removal by coagulation was also primarily contributed by the removal of the humic fraction (1.3 mg/L of 1.4 mg/L for batch 1, 3.3 mg/L of 3.6 mg/L for batch 3, and 9.5 mg/L of 10.0 mg/L for batch 4). Therefore, removal of humic substances by MIEX pretreatment would reduce the amount to react with hydrolytic aluminum species thereby decreasing coagulant demand of the pretreated water. It is noteworthy that natural water samples used in the coagulation experiments were prefiltered by 1.2 μm glass fiber filters to remove coarse materials; this resulted in low initial turbidities (