- Email: [email protected]
Autopsy of high-pressure membranes to compare effectiveness of MF and UF pretreatment in water reclamation
ARTICLE IN PRESS WAT E R R E S E A R C H
42 (2008) 697 – 706
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Autopsy of high-pressure membranes to compare effectiveness of MF and UF pretreatment in water reclamation Jeonghwan Kima,, Francis A. DiGianob, Roderick D. Reardonc a
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48823, USA Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, CB 7431, Chapel Hill, NC 27599-7431, USA c Carollo Engineers, 1800 Pembrook Drive, Suite 300 Orlando, FL 32810, USA b
art i cle info
ab st rac t
Article history:
A pilot-plant study was designed to compare the effectiveness of microfiltration (MF) and
Received 31 December 2006
ultrafiltration (UF) as pretreatment for high-pressure membranes in reclamation of
Received in revised form
biologically treated wastewater effluent. Granular media, filtered secondary effluent from
14 August 2007
a full-scale wastewater treatment plant, was fed to MF and UF units that operated in
Accepted 22 August 2007
parallel. Each of these filtrates served as the feedwater to two reverse osmosis (RO) units
Available online 21 September 2007
and one nanofiltration (NF) unit that operated in parallel. The decline in specific flux was
Keywords: Secondary effluent Pretreatment Membrane autopsy Specific flux Polysaccharides Microfiltration Ultrafiltration Reverse osmosis
substantially lower for high-pressure membranes receiving UF than MF pretreatment over the course of each of four pilot plant runs that lasted from 1 to 7 weeks. The removal of organic matter as measured by dissolved organic carbon (DOC) was somewhat higher by UF than MF pretreatment (about 15% by UF compared with 11% by MF). Addition of ferric chloride ahead of the UF unit, but not ahead of the MF unit, may account for this additional removal of organic matter. However, the additional DOC removal appeared insufficient to explain the differential in foulant accumulation between high-pressure membranes receiving UF and MF pretreatment. Extensive autopsy analyses of these high-pressure membranes showed from 35% to 56% less organic carbon on those receiving UF rather than MF pretreatment. A more specific indicator of a differential in organic fouling was the accumulation of polysaccharides and this showed from 27% to 38% less on UF- than on MFpretreated membranes. Yet another possible source of foulants is inorganic material given that the inorganic and organic weight percentages were nearly equal (56% vs. 44%) on the membrane surface. One specific source was aluminum added for phosphorus removal. Less fouling of high-pressure membranes pretreated by UF than MF could be due to the following: (1) a small, but very important, colloidal fouling fraction may have passed through MF but was rejected by UF pretreatment; (2) organic fouling was not related to organics in either the MF or UF filtrates but rather to organics that are generated in situ by microbial activity on the membrane surface; and/or (3) less passage of colloidal Al–P that carried over from secondary wastewater treatment. & 2007 Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +1 517 355 5155; fax: +1 517 355 0250.
E-mail address: [email protected] (J. Kim). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.08.042
ARTICLE IN PRESS 698
1.
WA T E R R E S E A R C H
Introduction
While membrane technology offers the potential to increase the value of wastewater reclamation, the occurrence of membrane fouling continues to limit the membrane operation efficiency (Jarusutthirak and Amy, 2002; Decarolis et al., 2001). High-pressure membranes such as nanofiltration (NF) and reverse osmosis (RO) membranes can be fouled easily by colloidal materials and organic matters present at high levels in secondary wastewater effluents (Her et al., 2003; Jarusutthirak et al., 2002; Parameshwaran et al., 2001). Foulants in wastewaters may fall into broad categories of particles, colloids, macromolecules, inorganics and even low molecular weight dissolved organics (Neis and Tiehm, 1997; Adin and Elimelech, 1989; Jarusutthirak and Amy, 2002; Van der Bruggen and Vandecasteele, 2001). The accumulation or adsorption of foulants on the surface or into the membrane matrix results in the loss of membrane performance over time, which results in increasing both capital and operational costs (Jarusutthirak et al., 2002). The extent of fouling of high-pressure membranes in water reclamation may depend upon the pretreatment processes as well as the chemical formulation of the highpressure membranes. In the past decade, pretreatment has shifted away from chemical coagulation–sedimentation– granular media filtration and towards low-pressure membrane filtration (Ghayeni et al., 1996; Reardon et al., 2005). The issue that remains unresolved is the choice between microfiltration (MF) or ultrafiltration (UF). Biomass flocs, individual bacterial cells and other particles carried over from the secondary clarifier should be removed by MF. Colloids, high molecular weight soluble microbial products and extracellular polymeric substances (EPS) generated by microbial activity (Jarusutthirak et al., 2002; Fonseca et al., 2003; Uhl et al., 2003) can be removed by UF, but the extent depends upon the molecular weight cut-off (MWCO) of the membrane. A direct comparison was made of the effectiveness of MF and UF as pretreatment for high-pressure membranes in a recent pilot-plant study of water reclamation (Reardon et al., 2007). The feedwater to the pilot plant was granular media, filtered secondary effluent from the North Buffalo Wastewater Reclamation Facility in Greensboro, NC. The decline in specific flux (flux/pressure) of three commercially available, high-pressure membranes that followed each pretreatment was the indicator of fouling. Statistical analysis showed that the decline in specific flux was lower for each of the three high-pressure membranes that received UF than MF pretreatment. For either type of pretreatment, the differences in rejection and fouling rate among the three high-pressure membranes were relatively small. The objective in this study was to use data collected from membrane autopsies at the end of the pilot-plant study at the North Buffalo Wastewater Reclamation Facility to explain less fouling of high-pressure membranes that received UF pretreatment. Autopsy measurements included microbial population, specific inorganic ions and several surrogates used to measure organic foulants (organic carbon, polysaccharides and molecular weight fraction).
42 (2008) 697– 706
2.
Methods
2.1.
Membrane pilot plant for wastewater reclamation
The process flow diagram for the membrane pilot plant is shown in Fig. 1. Granular media, filtered secondary effluent from the North Buffalo Wastewater Reclamation Facility (Greensboro, NC, USA), was pumped at a rate of about 100 gallon per minute (gpm) into a common tank. Feedwater was pumped from this tank to both MF and UF units. Filtrate from the MF (CMF 6M10C, US Filter MEMCOR) and UF (HYDRACap 60, Hydranautics) modules passed into the break tank from which the filtrate was pumped into high-pressure membranes as feedwater. Each type of high-pressure membrane was housed in two pressure vessels that operated in series with each pressure vessel containing three membrane elements. The total water recovery from these elements was 50%. Trains of low- and high-pressure membrane units were operated as MF–RO and UF–RO, although one of three highpressure membranes was classified as NF. Results from membrane pilot plant operation between April and August in 2005 are discussed in this paper. General characteristics of high-pressure membranes used in this pilot-plant study are shown in Table 1. All three highpressure membranes are spiral wound elements of thin film composite materials. The Hydranautics ESPA2 and TriSep X20 are RO membranes whereas the Dow/Film Tech NF90 is an NF membrane. The MWCO ranged from 50 to 500 Da, with the highest for the NF membrane. All three membranes are formed from polyamide that imparts a negative surface charge. However, the residual amino group in the polyamide-urea structure of the Trisep X-20 membrane, in contrast to the carboxylic acid group in the polyamide structure of the Dow/Film Tec NF-90 and the ESPA2 membranes, may suppress the negative charge. Differences in rate of decline in specific flux among the three high-pressure membranes during the pilot plant study, when normalized by the initial specific flux (lower for Trisep X20 membrane than either the ESPA2 or NF90 membranes), were relatively small and difficult to quantify (Reardon et al. 2007). Differences in foulant accumulation among the three high-pressure membranes were also difficult to discern from the autopsy results. For this reason, the discussion will be restricted to differences in foulant accumulation resulting from the pretreatment selection (MF vs. UF) for each highpressure membrane.
2.2.
Design and operation of low-pressure modules
Important characteristics of the MF and UF modules are given in Table 2. As stated by the manufacturers, the nominal pore size of the MF hollow fibers was 0.2 mm and the MWCO of the UF hollow fibers was 150 kDa. Using an available estimating method, the MWCO of the UF membrane corresponds to a nominal pore size of about 0.01 mm (Howe and Clark., 2002). Both the MF and UF modules operated in the dead-end mode. However, the flow pattern for MF was outside-in while that for the UF module inside-out. Another important difference is that 4 mg/L of FeCl3 is added to the feedwater of the UF
ARTICLE IN PRESS WAT E R R E S E A R C H
699
42 (2008) 697 – 706
Fig. 1 – Process flow diagram of membrane pilot plant to treat granular media, filtered, secondary effluent. Table 1 – Characteristics of high-pressure membranes applied in pilot-plant study Characteristic
Hydranautics ESPA2
Dow/FilmTec NF-90
TriSep X-20
Membrane type
Polyamide thin-film composite Est. 250–500 Negative (20 mV)
Polyamide thin-film composite 300 Slightly negative
Aromatic polyamide-urea thin-film composite Est. 50–100 Neutral to slightly negative
MWCO (Da) Membrane surface charge
Table 2 – Characteristics of MF and UF membranes as pretreatment for high-pressure membranes applied in pilot-plant study Characteristic
Membrane material Nominal pore size (mm) Affinity to water Ferric chloride dose (mg/L) Maximum TMP (psi) Backwash frequency (min) Backwash duration (min) Membrane surface area (ft2)
MF
UF
USFilter Memcor CMF 6M10C Polypropylene
Hydranautics HYDRAcap 60 Polyethersulfone
0.20 Hydrophilic 0
0.017 (MWCO ¼ 150 kDa) Hydrophilic 4 (as FeCl3)
12
20
20
25
2.5
0.5–1
2170 (outside)
1500 (inside)
module to coagulate particles that presumably enhance the efficiency of their removal during the backwash step. Backwashing was performed every 20 min in the MF module by
introducing high-pressure air to the inside of the hollow fiber while passing permeate over the outside surface at low pressure to flush away the foulants dislodged by air flow. Backwashing of the UF module was performed every 25 min by passing permeate from the outside to the inside of the hollow fiber (air is not used). According to the manufacturer’s recommendations, sodium hypochlorite (NaOCl) solution (6%) was added into backwash water (UF permeate) every fourth backwash cycle to reduce biological activity.
2.3. Measurements of inorganic fractions in foulant materials At completion of membrane pilot plant operation, the highpressure membrane elements were sent to Membrane Forensics (San Diego, CA, USA) to be unrolled and inspected to characterize inorganic foulant material. Approximately 10 g of foulant was removed with small flexible rubber spatula from one side of single membrane leaf. Loss on ignition (LOI) was determined by drying and weighing isolated foulant material at 110 1C. The samples were then ignited at 550 1C and weighed again to identify inorganic and organic portions of foulant materials. The loss on weight after ignition was taken as the organic portion of the foulant and the residual as the inorganic portion of the foulant.
ARTICLE IN PRESS 700
WA T E R R E S E A R C H
The foulant material was placed on a sample mount and examined with a scanning electron microscope (ETEC Autoscan SEM). Secondary electron images and backscatter electron images of the same sample area were taken at a 1500 magnification. Backscatter imaging distinguishes particles based on atomic weight. Targeted energy dispersive X-ray analysis (T-EDXA) uses backscatter images to determine inorganic elemental composition of foulant material. The wavelengths of X-ray were used to identify the presence and relative amounts of the chemical elements as the inorganic fraction in the foulant sample.
2.4. Measurements of organic fractions in foulant materials The organic fraction of the foulant material was quantified at the University of North Carolina (Chapel Hill, NC, USA). The foulant material was gently scraped from small segments of membrane surface (surface area is 100 cm2), added to 100 mL of ultrapure laboratory grade water and sonicated for 20 min (Croue et al., 2003). The sample was then filtered through a 0.45 mm Durapore hydrophilic PVDF membrane. The filtrate was analyzed by a Shimadzu Total Organic Carbon Analyzer model TOC-5000 (UV-persulfate digestion method). With the use of a 0.45 mm filter, the term ‘‘dissolved organic carbon’’ (DOC) is used here to describe the measurement technique, although this material would have originally been present in colloidal or particulate form on the membrane surface. Assuming that the scraping and sonication steps were efficient prior to passage of the sample through the 0.45 mm filter, DOC represents almost all of the organic carbon that had accumulated on the membrane surface. Polysaccharides in the fouling layer of all three unrolled high-pressure membranes were quantified by the phenol– sulfuric colorimetric method (Cho and Fane, 2002). This method was developed to measure total carbohydrates but has been adopted by various research groups to measure EPS (Cho and Fane, 2002; Kim and DiGiano, 2006). Polysaccharides are assumed to be a major contributor to EPS and hydrolyzed by sulfuric acid to products that react with phenol to produce a stable yellow color that can be measured spectrophotometrically (Hanson and Philips, 1981). The membrane specimen was cut into smaller segments and dissolved directly in 2 mL of 2.5% phenol solution after which 10 mL of 98% sulfuric acid was added for a 10-min reaction period at ambient temperature and 15 min at 25 1C. The absorbance at 488 nm was then measured by a UV–Vis spectrophotometer (U-2000, Hitachi) and compared against a calibration curve developed with glucose standard solutions that were prepared in the same way. High-pressure, size exclusion chromatography (HPSEC) was used to characterize the organic fraction of the foulant material that was removed from each of the unrolled highpressure membrane elements. The segments cut from each membrane were about 100 cm2 in size. The foulant samples for HPSEC analysis were prepared in the same manner as those for the organic carbon measurement described above. Filtrates from the 0.45 mm filtration step were sent to the University of Colorado (Boulder, CO, USA) for HPSEC analysis.
42 (2008) 697– 706
HPSEC separates molecules on a chromatographic column according to their molecular weight. The traditional detector is fixed wavelength (254 nm) UV. This detector is sensitive to humic natural organic matter (NOM) that has aromatic structural features but is less sensitive to non-humic NOM that does not have aromatic features such as polysaccharides. An on-line DOC detector that allows recognition of NOM components with low (e.g., proteins) or no (e.g., polysaccharides) UV absorption was implemented. A variable wavelength UV detector is used to distinguish NOM constituents with different UV spectra. The high-pressure liquid chromatography (HPLC) column was packed with Toyopearl (HW-50S) resin ˚. beads with a diameter of 20–40 mm and a pore size of 125 A The maximum detectable molecular weight was about 80 kDa.
2.5.
Measurement of microbial foulants
Membrane samples were cut from each unrolled membrane element, mounted with double-sided tape on a aluminum stubs, and sputter coated (Hummer X sputter coater, Anatech Ltd.) with a thin layer of approximately 10–20 nm of a conductive metal (60% gold and 40% palladium alloy). They were stored in a dry, dust-free environment until SEM (SEM, Stereoscan 200, Cambridge) observations of microbial foulants. The accelerating voltage was 15 kV and the working distance was about 20 mm. A 200 magnification was used to obtain appropriate resolution of the samples.
3.
Results and discussion
3.1.
Microscopic observations of foulant layers
SEM images taken from three high-pressure membrane types (X20, ESPA2 and NF90) are presented in Fig. 2. General conclusions cannot be reached about the effects of element position on the extent of fouling because the original positions of the membrane elements (first and sixth) were reversed during May 2005 due to the occurrence of irreversible fouling and membrane maintenance. These SEM images, however, show qualitatively that the fouling layer appeared much less dense for high-pressure membranes receiving UF filtrate than those receiving MF filtrate. While these measurements are only qualitative, microbial foulants (elliptical- or spherical-shaped cells) and perhaps abiotic foulants were present. A lower density of microbes on high-pressure membranes pretreated by UF may be due to more effective removal of microbes and/or removal of small but important amounts of substrate that results in the suppression of the microbial activity by pretreatment with UF membranes.
3.2.
Inorganic and organic fractions of foulant layer
The percent contribution of inorganic and organic foulants was estimated from the LOI test performed by Membrane Forensics on foulant materials isolated from the high-pressure membrane elements as described above. As listed in Table 3, the foulant layer on the MF pretreated,
ARTICLE IN PRESS WAT E R R E S E A R C H
701
42 (2008) 697 – 706
Fig. 2 – SEM images of foulant materials deposited on X20 (A), ESPA2 (B) and NF90 (C) membranes pretreated by MF membrane and X20 (D), ESPA2 (E) and NF90 (F) pretreated by UF membrane (element position #1). Table 3 – Percentage of inorganic and organic materials in foulant by loss of ignition analysis (membrane forensics) MF pretreatment
Organic (%) Inorganic (%)
UF pretreatment
ESPA2
X20
NF90
ESPA2
X20
NF90
49.6 50.4
54.7 45.3
43.9 56.1
43.9 56.1
44.4 55.6
44.4 55.6
high-pressure membranes was approximately equally divided on a percentage by weight basis between organic and inorganic material. The contributions of organic and inorganic foulants were 45.5% and. 54.5%, respectively, for the high-pressure membranes receiving UF pretreatment. Thus, the choice of pretreatment had little effect on the inorganic–organic composition of the foulant layer. Based on the large inorganic fraction of foulants, the presence of inorganic chemicals in the granular media, filtered, secondary effluent, can cause significant scale formation once their solubility products are exceeded in the concentrate stream. The membrane element selected for the LOI analyses was the fourth in a series of six for each membrane train. Given that the recovery of each element was about 11%, the feed stream would have been concentrated by a factor of 37% by the end of the fourth element. The extent of precipitation of inorganic compounds could be estimated from measurements of the concentrations of the inorganics in the feed stream and the possible solubility equilibrium; however, such calculations are beyond the scope of this study.
3.3.
Specific inorganic chemicals in foulant material
Results of T-EDXA of the foulant materials are shown in Fig. 3. The order of decreasing prevalence of inorganic ions for the MF pretreated, high-pressure membranes was P (30%), Al (25%) and Ca (15%). Similar results were also obtained with the NF90 high-pressure membrane. P-containing precipitates are very likely present on the membrane because of ineffective precipitation of P within the North Buffalo Wastewater Treatment Facility by addition of 1–27 mg/L of sodium aluminate (as Al) prior to the aeration basins (the average addition was about 5 mg/L). P concentrations in the secondary effluent were always significantly higher than the threshold concentrations considered necessary for balanced microbial growth (40.2 mg/L). As also shown in Fig. 3, the order of prevalence of inorganic ions shifted for UF-pretreated membranes. Instead of P, the most abundant ion was Fe (65%) followed by P (20%) and Ca (10%). Thus, the addition of 4 mg/L of FeCl3 to the UF feed stream, as recommended by the manufacturer to reduce pore fouling by formation of Fe-containing floc, was not
ARTICLE IN PRESS 702
WA T E R R E S E A R C H
42 (2008) 697– 706
70
35
34.6
Aluminum Manganese
30
Iron
Phosphorous
27.2
Calcium Iron
25
Titanium
20
Sulfur
15.4
15
Zinc
12.9
Silicon
10
Magnesium Potassium
4.2
5
Percentage of Inorganic Foulant (%)
Percentage of Inorganic Foulant (%)
40
Chloride
Calcium
22.9 22.6
Iron
20
Silicon Zinc
15
Sulfur
12.6
Titanium
10
Chloride Potassium
3.8
5
2.5 1.8 1.5 1.4
Percentage of Inorganic Foulant (%)
Aluminum Manganese
Titanium
20
Magnesium
9.2
10
Potassium
3.7
1.7 1.6 1.4 0.9 0.6 0.4 0.3
64.3
Iron
60
Phosphorous Calcium
50 Aluminum
40
Sulfur Silicon
30 Titanium
19.4 20
Chloride Potassium
7.6
10
3.2
1.5 1.4
1.4
0.7 0.5
0
45
70
40
Aluminum
32.5
Calcium Magnesium
25.8
Iron
25
Sulfur
20
Tin
15.1 14.6
Titanium
15
Silicon
10 Zinc
4.1 2.8 2.2 1.8
Percentage of Inorganic Foulant (%)
Phosphorous
Percentage of Inorganic Foulant (%)
Sodium
23.8
0.2 0.1
0
5
Silicon Sulfur
30
70
30
30
Aluminum
40
Phosphorous
30.6 Percentage of Inorganic Foulant (%)
Calcium
50
0
35
35
Phosphorous
56.4
1.8 1.6 1.2 0.9 0.1 0.1
0
25
60
Iron
60
Phosphorous Calcium
50
Aluminum
40
Sulfur Chloride
30 Silicon
20 20
Potassium
9.7
Magnesium
10 2.7
0.8 0.3
0
64.8
1.5
0.6
0.3 0.3
0.1
0
Fig. 3 – Distribution of inorganic foulants as measured by T-EDXA on ESPA2 (A). NF90 (C) and X20 membrane (E) pretreated by MF membrane and ESPA2 (B), NF90 (D) and X20 (F) membrane pretreated by UF membrane. (A) MF-Hydranautics ESPA2, (B) UFHydranautics ESPA2, (C) MF-Filmtec Nf90, (D) UF-Filmtec Nf90, (E) MF-Trisep X20, (F) UF-Trisep X20.
effectively bound into precipitates that could be rejected by UF. Also indicated in Fig. 3 is a much higher fraction of Al in the foulant layer on high-pressure membranes pretreated by MF than by UF (about 20% vs. 3%). This could be due to more effective rejection by UF of poorly settling Al precipitates that had carried over from the addition of sodium aluminate in
the activated sludge process. A much lower P in the UF than MF filtrate is possibly related. The average of eight measurements of orthophosphate over the pilot-plant study was 0.42 mg/L for UF compared with 0.85 mg/L for MF. The alternative explanation for lower P in the UF filtrate is additional precipitation by Fe. In either case, it is possible that the presence of less Al–P colloidal material on high-pressure
ARTICLE IN PRESS WAT E R R E S E A R C H
membranes pretreated by UF was an important explanation for less fouling. Although difficult to prove, the dominance of Fe in the foulant layer on the high-pressure membranes pretreated by UF as shown in Fig. 3 may also be an explanation for lower fouling of high-pressure membranes receiving UF than MF pretreatment. FeCl3 may not only alter the nature of the organic foulants that accumulated on the membrane surface but also increase the porosity of the foulant layer, which in turn produced less resistance to filtration.
3.4.
Accumulation of DOC in foulant material
The effect of pretreatment on DOC accumulation of highpressure membranes (X20, ESPA2 and X20) is presented in Fig. 4. The DOC on the membrane surface is expressed as mg of DOC per m2 of unit filtration area. The average DOC for the first and sixth elements will be discussed because these element positions were reversed during the course of the pilot-plant study such that longitudinal variation in foulant accumulation could not be analyzed. As shown in Fig. 4, the DOC on each of the three highpressure membrane types that followed UF pretreatment was from 35% to 59% of the DOC measured on these same membranes following MF pretreatment. Less DOC accumulation is consistent with less decline in specific flux during the pilot-plant study for UF pretreatment. While the DOC accumulations caused by UF and MF pretreatment were very different, DOC removal by UF was only 4% higher (15% vs. 11%) than by MF (Reardon et al., 2007). In fact, the difference in DOC removal between UF and MF was not significant statistically (p40.05), albeit the sample size was relatively small to make reliable statistical inferences. The difference in DOC removal should be small because of the high MWCO (150 kDa) of the UF membrane and the fact that the presence of the low molecular weight material is reasonable in granular media, filtered secondary effluent.
Although a small fraction, the differential in DOC between UF and MF filtrates represents colloidal material in the approximate size range of 0.01–0.2 mm that could produce a disproportionately large fraction of the foulant on the highpressure membranes (Howe and Clark, 2002). This additional colloidal fouling fraction that was removed by UF but not by MF pretreatment may have higher affinity for high-pressure membranes. Accordingly, it could explain why DOC accumulation on the high-pressure membranes receiving UF pretreatment was 35–59% of that for membranes receiving MF pretreatment. A difference in in-situ generation of organic matter by microbial growth (biofouling) is yet another possible explanation for far less DOC accumulation on membranes receiving UF pretreatment. Although quantitative evidence is lacking, the biofouling potential could be lower for UF than MF filtrate because of higher removal of seeding microbes by UF than MF given that pore sizes are about ten times smaller; higher removal of biodegradable substrates caused by FeCl3 addition to the UF feed stream; and/or inactivation of microbes by addition of NaOCl on every fourth backwash for UF unit.
3.5.
450 MF pretreatment
400
UF pretreatment
mg of Polysaccharides/m2
80 70 mg of DOC/m2
Accumulation of polysaccharides in foulant material
Fig. 5 shows the average value of the polysaccharide accumulation between the first and sixth element for each high-pressure membrane. Consistent with the results of DOC accumulation, there was far less accumulation of polysaccharide (27–38%) on membranes receiving UF pretreatment than MF pretreatment. Polysaccharides are often the main component of biopolymers, or EPS, and are typically in the colloidal size range which is defined as organic matter greater than a molecular weight of 3500 Da (Jarusutthirak et al., 2002). The maximum size of foulant material that contains polysaccharides, however, could be much higher than 3500 Da. For example, in a separated bench-scale study using this same UF hollow fiber and the same feedwater, polysaccharides
100 90
703
42 (2008) 697 – 706
60 50 40 30 20
MF pretreatment UF pretreatment
350 300 250 200 150 100 50
10
0
0 ESPA2
NF90 High pressure membranes
X20
Fig. 4 – Comparison of DOC accumulation in foulant material per unit area of each high-pressure membrane for MF and UF pretreatment.
ESPA2
NF90 X20 High pressure membranes
Fig. 5 – Comparison of polysaccharide accumulation in foulant material per unit area of each high-pressure membrane for MF and UF pretreatment.
ARTICLE IN PRESS 704
WA T E R R E S E A R C H
42 (2008) 697– 706
polysaccharides have the same molecular formula as glucose, the mass fraction of DOC should have been 0.4 rather than 0.2. The difference in this ratio could have resulted from the different techniques used to obtain samples for DOC and polysaccharides. The sample for DOC was scraped off of the membrane and sonicated by 20 min before passing through the 0.45 mm filter, while polysaccharides were measured directly after chemical agents came into contact with the membrane specimen. Thus, DOC analysis may not have captured all of the organic carbon due to lack of complete removal from the surface and/or inability to solubilize all organic material by sonication. Both of these effects would cause DOC to be lower than predicted from the molecular formula of polysaccharides.
accumulated on the membrane surface over the course of many backwashing cycles (Kim and DiGiano, 2006). Thus, a small fraction of polysaccharides in granular media filtered, secondary effluent is apparently larger than the stated MWCO of 150 kDa. These polysaccharides could either be generated from biological treatment or they have been present in the raw wastewater. The former source seems more likely based on other studies (Jarusutthirak et al., 2002; Cho and Fane, 2002). Less accumulation of polysaccharides on the highpressure membranes that were pretreated by UF could be consistent with the removal of a fraction of very large polysaccharide-containing colloids that were not rejected by MF. An alternative explanation for observing less polysaccharide accumulation on membranes receiving UF pretreatment is the notion of a lower biofouling potential in UF than in MF filtrate that produced less in-situ generation of organic matter by microbial growth as explained earlier. Although measured by two different techniques, the mass of DOC (Fig. 4) was compared to the mass of polysaccharides (Fig. 5). The ratio of DOC to polysaccharides, as expressed by its glucose equivalent (C6H12O6), was about 0.2. If the
3.6.
HPSEC results on foulant materials
The graphical representation of HPSEC results in Fig. 6 is in order of decreasing molecular weight because this corresponds to the order of appearance in the column eluant. The correspondence between peaks in the column elutriates and
0.5
0.5 MF-X20-1
MF-ESPA2-1
MF-X20-6
0.4
MF-ESPA2-6
0.4
UF-X20-1
UF-ESPA2-1
UF-X20-6
Responses
Responses
UF-ESPA2-6
0.3
0.2
0.1
DOC
UVA 0.0 1000000 100000
0.3
0.2
0.1
10000
1000
100
10
1
DOC
UVA 0 1000000 100000
Molecular Weight (dalton)
10000
1000
100
10
Molecular Weight (Da)
0.5 MF-NF90-1 MF-NF90-6
0.4
UF-NF90-1
Responses
UF-NF90-6
0.3
0.2
0.1
DOC
UVA 0 1000000 100000
10000 1000 100 Molecular Weight (Da)
10
1
Fig. 6 – HPSEC analysis with DOC and UV detectors of foulant sample removed from X20 (A), ESPA2 (B) and NF90 (C).
1
ARTICLE IN PRESS WAT E R R E S E A R C H
molecular weight needs to be cautiously interpreted. Firstly, a homologous series of an organic compound is used to obtain a correspondence between time of elutriation and molecular weight of the reference compound (largest molecular weight appears first). That is, the elutriation time of non-reference organic compounds with different structural features but the same molecular weight can be different. Secondly, the maximum detectable molecular weight is based on the pore size of the column packing materials (80 kDa in this study). Finally, the first peak to appear in the elutriation from injection of unknown mixture of organic compound is a collective response from all organic compounds of the reference molecular size at that elution time and larger (up to maximum detectable of 80 kDa). Thus, the first peak shown in Fig. 6 represents all molecular weights greater than about 20 and less than 80 kDa. The DOC peak responses in Fig. 6 show clearly that the largest is at a molecular weight of about 20 kDa regardless of the pretreatment. This eluted organic fraction comprises nonhumic, low aromaticity structures such as protein, polysaccharides and/or amino sugar that have been noted in secondary effluent (Jarusutthirak et al., 2002). The smaller DOC peaks at low molecular weights (from 100 to 1000 Da) should represent humic substances and organic acids and these are also similar in molecular weights for all membranes. However, the peak heights are lower for UF than MF pretreatment. This is consistent with observations of less accumulation of DOC and polysaccharides (Figs. 4 and 5) on membranes receiving UF pretreatment. Although HPSEC results are reported for the first and sixth elements, the effect of element position on the results cannot be determined because the original first and sixth elements were reversed part way through the pilot-plant study as was mentioned in discussion of the SEM results. Evidence to support the presence of biologically derived foulant material was provided by Fourier transform infrared (FTIR) spectra of foulant samples (Reardon et al., 2007). The FTIR spectra showed strong absorption peaks at 3397–3431, 1644–1656 and 1058–1086 cm1, and minor peaks at 2917–2928 and 545–567 cm1 for high-pressure membranes receiving either MF or UF pretreatment. Previous studies (Croue et al., 2003; Jarusutthirak et al., 2002) indicate that absorption bands at 1650, 1600, 1550, 1410 and 1060 cm1 are characteristic of proteins, amino sugars and carbohydrates that compose bacterial cell walls. Therefore, while the presence of polysaccharides in the raw wastewater cannot be ruled out, the more likely source is production in the biological treatment process. The major DOC peak of each HPSEC for both UF and MF pretreatment raises questions as to the origin of this nonhumic fraction of foulant materials. Its molecular weight (20 kDa) is far less than the MWCO of the UF membrane (150 kDa). If this 20 kDa material had originated directly from granular media, filtered secondary effluent, it would have passed through both the MF and UF. Therefore, the DOC peak heights should have been the same for membranes receiving UF and MF pretreatment. Yet, the peak height was lower for the membranes receiving UF than MF pretreatment (Fig. 6). One possible explanation for less 20 kDa material is that the addition of FeCl3 caused preferential removal of this small
42 (2008) 697 – 706
705
molecular weight colloidal fraction or it increased in size by the coagulating action; however, supporting data are not available for either mechanism. Another possibility is in-situ generation of an organic foulant by biofouling that produced a low molecular fraction of DOC directly upon the highpressure membranes. That is, the observation of a lower DOC peak height for organic material removed from membranes receiving UF pretreatment could be due to a lower biofouling potential of UF filtrate for the reasons noted earlier.
4.
Conclusions
Membrane autopsies showed less foulant accumulation by UF than MF pretreatment that suggested possible explanations for less flux decline of the high-pressure membranes receiving UF pretreatment. However, less accumulation of a specific foulant that might be responsible for less flux decline could not be identified. The weight fractions of inorganic and organic foulants were approximately equal, suggesting that either or both could contribute to a decline in specific flux with time but a differential in weight factions caused by UF compared with MF pretreatment was not apparent. Nevertheless, there was less Al–P colloidal material on the UFpretreated, high-pressure membranes and this may be an important explanation for less flux decline. Alternatively, the differential in DOC and polysaccharide accumulation between membranes receiving UF and MF pretreatment could also be consistent with less extent of flux decline for membranes receiving UF treatment. The UF filtrate would be expected to have less foulant material owing to somewhat greater removal of DOC (15% vs. 11%) perhaps caused by the addition of FeCl3. Moreover, FeCl3 could not only reduce the fouling of the UF membrane but also reduce the fouling potential of organic matter in the UF permeate and thus in the feed to the high-pressure membranes. Unfortunately, there were no direct data to support these possible mechanisms involving FeCl3 and evidence from the literature (Howe and Clark, 2002) is restricted to studies of natural waters. Although difficult to prove, less in-situ generation of organic foulants may be another possible explanation for less foulant accumulation on UF- than MF-pretreated, high-pressure membranes. The evidence is less accumulation of polysaccharides of a molecular weight fraction that should have passed through either MF or UF pretreatment. Equal amounts of this material should have been present on high-pressure membranes receiving either UF or MF pretreatment if it was not generated in situ. A lower biofouling potential after UF pretreatment could also be linked to autopsy results showing fewer microbes on high-pressure membranes receiving UF than MF pretreatment, albeit the differences are very difficult to quantify. Fewer microbes could be due to more efficient rejection by UF than MF. Alternatively, the addition of NaOCl on every fourth backwash of UF could also have inactivated microbes. This would lead to the suggestion that the disinfection strategy rather than membrane properties was responsible for less observed fouling after UF pretreatment. In summary, the autopsy results did not provide a clearcut, fundamental explanation of less flux decline for high-pressure membranes that received UF pretreatment.
ARTICLE IN PRESS 706
WA T E R R E S E A R C H
Nonetheless, valuable insights were gained into the possible roles of polysaccharides and inorganic colloids that need to be explored more fully in carefully controlled experiments that were not possible to achieve in these pilot-plant studies.
Acknowledgments We would like to thank the Water Environment Research Foundation and the North Carolina Water Resources Research Institute for their financial support. We also thank Membrane Forensics with the help of membrane autopsy works and Dr. Gary Amy with HPSEC analysis. R E F E R E N C E S
Adin, A., Elimelech, M., 1989. Particle filtration in wastewater irrigation. J. Irrig. Drainage Eng. (ASCE) 115, 474–487. Cho, B., Fane, A.G., 2002. Fouling transient in normally sub-critical flux operation of a membrane bioreactor. J. Membr. Sci. 209, 391–403. Croue, J.P., Grasset, L., Bacle, S., Jacquemet, V., 2003. Nanofiltration membrane autopsy of a full-scale unit: characterization of the organic and inorganic constituents of the foulants. In: Proceedings of the American Water Works Association Membrane Technology Conference, American Water Works Association, Denver, CO. Decarolis, J., Hong, S., Taylor, J., 2001. Fouling behavior of a pilot scale inside-out hollow fiber UF membrane during dead-end filtration of tertiary wastewater. J. Membr. Sci. 191, 165–178. Fonseca, A.C., Summers, R.S., Greenberg, A.R., Hernandez, M.T., 2003. Isolating and modeling critical factors governing biofouling of nanofiltration membranes. In: Proceedings of the American Water Works Association, Denver, CO. Ghayeni, S., Madaeni, S., Fane, A.G., Schneider, R.P., 1996. Aspects of microfiltration and reverse osmosis in municipal wastewater reuse. Desalination 117, 49–60. Hanson, R.S., Philips, J.A., 1981. Chemical composition. In: Gerhardt, P., et al. (Eds.), Manual of Methods for General
42 (2008) 697– 706
Bacteriology. American Society for Microbiology, Washington, DC, pp. 328–364. Her, N., Amy, G., Pecheux, A., Gatel, D., Cavard, J., 2003. Nanofiltration (NF) membrane foulant identification and fouling mechanisms. In: Proceedings of the American Water Works Association Membrane Technology Conference. Howe, K.J., Clark, M.M., 2002. Fouling of microfiltration and ultrafiltration membranes by natural waters. Environ. Sci. Technol. 36 (16), 3571–3576. Jarusutthirak, C., Amy, G., 2002. Membrane filtration of wastewater effluents for reuse: effluent organic matter rejection and fouling. Water Sci. Technol. 43 (10), 225–232. Jarusutthirak, C., Amy, G., Croue, J.P., 2002. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 145, 247–255. Kim, J., DiGiano, F.A., 2006. A two-fiber, bench-scale test of ultrafiltration (UF) for investigation of fouling rate and characteristics. J. Membr. Sci. 271, 196–204. Neis, U., Tiehm, A., 1997. Particle size analysis in primary and secondary wastewater effluents. Water Sci. Technol. 36 (4), 151–158. Parameshwaran, K., Fane, A.G., Cho, B.D., Kim, K.J., 2001. Analysis of microfiltration performance with constant flux processing of secondary effluent. Water Res. 35 (18), 4349–4358. Reardon, R.D., Paranjape, S.V., Foussereau, X.J., DiGiano, F.A., Aitken, M.D., Kim, J.H., Chang, S., Cramer, R., 2005. Membrane treatment of secondary effluent for subsequent reuse. Water Environment Research Foundation Report (01-CTS-6). IWA Publishing, London, UK. Reardon, R.D., DiGiano, F.A., Hobbs, C.M., Treadway, J.M., Kim, J.H., 2007. Membrane treatment of secondary effluent for subsequent use: phase 2—pilot plant comparisons of MF and UF for pretreatment of high-pressure membranes. Water Environment Research Foundation Report (01-CTS-6a). IWA Publishing, London, UK. Uhl, W., Hagmeyer, G., Huck, P.M., 2003. Biofouling of NF and RO membranes in drinking water production from surface waters; theoretical consideration of prerequisites and possible prevention strategies. In: Proceedings of the American Water Works Association, Denver, CO. Van der Bruggen, B., Vandecasteele, C., 2001. Flux decline during nanofiltration of organic components in aqueous solution. Environ. Sci. Technol. 35, 3535–3540.
42 (2008) 697 – 706
Available at www.sciencedirect.com
journal homepage: www.elsevier.com/locate/watres
Autopsy of high-pressure membranes to compare effectiveness of MF and UF pretreatment in water reclamation Jeonghwan Kima,, Francis A. DiGianob, Roderick D. Reardonc a
Department of Civil and Environmental Engineering, Michigan State University, East Lansing, MI 48823, USA Department of Environmental Sciences and Engineering, University of North Carolina at Chapel Hill, CB 7431, Chapel Hill, NC 27599-7431, USA c Carollo Engineers, 1800 Pembrook Drive, Suite 300 Orlando, FL 32810, USA b
art i cle info
ab st rac t
Article history:
A pilot-plant study was designed to compare the effectiveness of microfiltration (MF) and
Received 31 December 2006
ultrafiltration (UF) as pretreatment for high-pressure membranes in reclamation of
Received in revised form
biologically treated wastewater effluent. Granular media, filtered secondary effluent from
14 August 2007
a full-scale wastewater treatment plant, was fed to MF and UF units that operated in
Accepted 22 August 2007
parallel. Each of these filtrates served as the feedwater to two reverse osmosis (RO) units
Available online 21 September 2007
and one nanofiltration (NF) unit that operated in parallel. The decline in specific flux was
Keywords: Secondary effluent Pretreatment Membrane autopsy Specific flux Polysaccharides Microfiltration Ultrafiltration Reverse osmosis
substantially lower for high-pressure membranes receiving UF than MF pretreatment over the course of each of four pilot plant runs that lasted from 1 to 7 weeks. The removal of organic matter as measured by dissolved organic carbon (DOC) was somewhat higher by UF than MF pretreatment (about 15% by UF compared with 11% by MF). Addition of ferric chloride ahead of the UF unit, but not ahead of the MF unit, may account for this additional removal of organic matter. However, the additional DOC removal appeared insufficient to explain the differential in foulant accumulation between high-pressure membranes receiving UF and MF pretreatment. Extensive autopsy analyses of these high-pressure membranes showed from 35% to 56% less organic carbon on those receiving UF rather than MF pretreatment. A more specific indicator of a differential in organic fouling was the accumulation of polysaccharides and this showed from 27% to 38% less on UF- than on MFpretreated membranes. Yet another possible source of foulants is inorganic material given that the inorganic and organic weight percentages were nearly equal (56% vs. 44%) on the membrane surface. One specific source was aluminum added for phosphorus removal. Less fouling of high-pressure membranes pretreated by UF than MF could be due to the following: (1) a small, but very important, colloidal fouling fraction may have passed through MF but was rejected by UF pretreatment; (2) organic fouling was not related to organics in either the MF or UF filtrates but rather to organics that are generated in situ by microbial activity on the membrane surface; and/or (3) less passage of colloidal Al–P that carried over from secondary wastewater treatment. & 2007 Elsevier Ltd. All rights reserved.
Corresponding author. Tel.: +1 517 355 5155; fax: +1 517 355 0250.
E-mail address: [email protected] (J. Kim). 0043-1354/$ - see front matter & 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.watres.2007.08.042
ARTICLE IN PRESS 698
1.
WA T E R R E S E A R C H
Introduction
While membrane technology offers the potential to increase the value of wastewater reclamation, the occurrence of membrane fouling continues to limit the membrane operation efficiency (Jarusutthirak and Amy, 2002; Decarolis et al., 2001). High-pressure membranes such as nanofiltration (NF) and reverse osmosis (RO) membranes can be fouled easily by colloidal materials and organic matters present at high levels in secondary wastewater effluents (Her et al., 2003; Jarusutthirak et al., 2002; Parameshwaran et al., 2001). Foulants in wastewaters may fall into broad categories of particles, colloids, macromolecules, inorganics and even low molecular weight dissolved organics (Neis and Tiehm, 1997; Adin and Elimelech, 1989; Jarusutthirak and Amy, 2002; Van der Bruggen and Vandecasteele, 2001). The accumulation or adsorption of foulants on the surface or into the membrane matrix results in the loss of membrane performance over time, which results in increasing both capital and operational costs (Jarusutthirak et al., 2002). The extent of fouling of high-pressure membranes in water reclamation may depend upon the pretreatment processes as well as the chemical formulation of the highpressure membranes. In the past decade, pretreatment has shifted away from chemical coagulation–sedimentation– granular media filtration and towards low-pressure membrane filtration (Ghayeni et al., 1996; Reardon et al., 2005). The issue that remains unresolved is the choice between microfiltration (MF) or ultrafiltration (UF). Biomass flocs, individual bacterial cells and other particles carried over from the secondary clarifier should be removed by MF. Colloids, high molecular weight soluble microbial products and extracellular polymeric substances (EPS) generated by microbial activity (Jarusutthirak et al., 2002; Fonseca et al., 2003; Uhl et al., 2003) can be removed by UF, but the extent depends upon the molecular weight cut-off (MWCO) of the membrane. A direct comparison was made of the effectiveness of MF and UF as pretreatment for high-pressure membranes in a recent pilot-plant study of water reclamation (Reardon et al., 2007). The feedwater to the pilot plant was granular media, filtered secondary effluent from the North Buffalo Wastewater Reclamation Facility in Greensboro, NC. The decline in specific flux (flux/pressure) of three commercially available, high-pressure membranes that followed each pretreatment was the indicator of fouling. Statistical analysis showed that the decline in specific flux was lower for each of the three high-pressure membranes that received UF than MF pretreatment. For either type of pretreatment, the differences in rejection and fouling rate among the three high-pressure membranes were relatively small. The objective in this study was to use data collected from membrane autopsies at the end of the pilot-plant study at the North Buffalo Wastewater Reclamation Facility to explain less fouling of high-pressure membranes that received UF pretreatment. Autopsy measurements included microbial population, specific inorganic ions and several surrogates used to measure organic foulants (organic carbon, polysaccharides and molecular weight fraction).
42 (2008) 697– 706
2.
Methods
2.1.
Membrane pilot plant for wastewater reclamation
The process flow diagram for the membrane pilot plant is shown in Fig. 1. Granular media, filtered secondary effluent from the North Buffalo Wastewater Reclamation Facility (Greensboro, NC, USA), was pumped at a rate of about 100 gallon per minute (gpm) into a common tank. Feedwater was pumped from this tank to both MF and UF units. Filtrate from the MF (CMF 6M10C, US Filter MEMCOR) and UF (HYDRACap 60, Hydranautics) modules passed into the break tank from which the filtrate was pumped into high-pressure membranes as feedwater. Each type of high-pressure membrane was housed in two pressure vessels that operated in series with each pressure vessel containing three membrane elements. The total water recovery from these elements was 50%. Trains of low- and high-pressure membrane units were operated as MF–RO and UF–RO, although one of three highpressure membranes was classified as NF. Results from membrane pilot plant operation between April and August in 2005 are discussed in this paper. General characteristics of high-pressure membranes used in this pilot-plant study are shown in Table 1. All three highpressure membranes are spiral wound elements of thin film composite materials. The Hydranautics ESPA2 and TriSep X20 are RO membranes whereas the Dow/Film Tech NF90 is an NF membrane. The MWCO ranged from 50 to 500 Da, with the highest for the NF membrane. All three membranes are formed from polyamide that imparts a negative surface charge. However, the residual amino group in the polyamide-urea structure of the Trisep X-20 membrane, in contrast to the carboxylic acid group in the polyamide structure of the Dow/Film Tec NF-90 and the ESPA2 membranes, may suppress the negative charge. Differences in rate of decline in specific flux among the three high-pressure membranes during the pilot plant study, when normalized by the initial specific flux (lower for Trisep X20 membrane than either the ESPA2 or NF90 membranes), were relatively small and difficult to quantify (Reardon et al. 2007). Differences in foulant accumulation among the three high-pressure membranes were also difficult to discern from the autopsy results. For this reason, the discussion will be restricted to differences in foulant accumulation resulting from the pretreatment selection (MF vs. UF) for each highpressure membrane.
2.2.
Design and operation of low-pressure modules
Important characteristics of the MF and UF modules are given in Table 2. As stated by the manufacturers, the nominal pore size of the MF hollow fibers was 0.2 mm and the MWCO of the UF hollow fibers was 150 kDa. Using an available estimating method, the MWCO of the UF membrane corresponds to a nominal pore size of about 0.01 mm (Howe and Clark., 2002). Both the MF and UF modules operated in the dead-end mode. However, the flow pattern for MF was outside-in while that for the UF module inside-out. Another important difference is that 4 mg/L of FeCl3 is added to the feedwater of the UF
ARTICLE IN PRESS WAT E R R E S E A R C H
699
42 (2008) 697 – 706
Fig. 1 – Process flow diagram of membrane pilot plant to treat granular media, filtered, secondary effluent. Table 1 – Characteristics of high-pressure membranes applied in pilot-plant study Characteristic
Hydranautics ESPA2
Dow/FilmTec NF-90
TriSep X-20
Membrane type
Polyamide thin-film composite Est. 250–500 Negative (20 mV)
Polyamide thin-film composite 300 Slightly negative
Aromatic polyamide-urea thin-film composite Est. 50–100 Neutral to slightly negative
MWCO (Da) Membrane surface charge
Table 2 – Characteristics of MF and UF membranes as pretreatment for high-pressure membranes applied in pilot-plant study Characteristic
Membrane material Nominal pore size (mm) Affinity to water Ferric chloride dose (mg/L) Maximum TMP (psi) Backwash frequency (min) Backwash duration (min) Membrane surface area (ft2)
MF
UF
USFilter Memcor CMF 6M10C Polypropylene
Hydranautics HYDRAcap 60 Polyethersulfone
0.20 Hydrophilic 0
0.017 (MWCO ¼ 150 kDa) Hydrophilic 4 (as FeCl3)
12
20
20
25
2.5
0.5–1
2170 (outside)
1500 (inside)
module to coagulate particles that presumably enhance the efficiency of their removal during the backwash step. Backwashing was performed every 20 min in the MF module by
introducing high-pressure air to the inside of the hollow fiber while passing permeate over the outside surface at low pressure to flush away the foulants dislodged by air flow. Backwashing of the UF module was performed every 25 min by passing permeate from the outside to the inside of the hollow fiber (air is not used). According to the manufacturer’s recommendations, sodium hypochlorite (NaOCl) solution (6%) was added into backwash water (UF permeate) every fourth backwash cycle to reduce biological activity.
2.3. Measurements of inorganic fractions in foulant materials At completion of membrane pilot plant operation, the highpressure membrane elements were sent to Membrane Forensics (San Diego, CA, USA) to be unrolled and inspected to characterize inorganic foulant material. Approximately 10 g of foulant was removed with small flexible rubber spatula from one side of single membrane leaf. Loss on ignition (LOI) was determined by drying and weighing isolated foulant material at 110 1C. The samples were then ignited at 550 1C and weighed again to identify inorganic and organic portions of foulant materials. The loss on weight after ignition was taken as the organic portion of the foulant and the residual as the inorganic portion of the foulant.
ARTICLE IN PRESS 700
WA T E R R E S E A R C H
The foulant material was placed on a sample mount and examined with a scanning electron microscope (ETEC Autoscan SEM). Secondary electron images and backscatter electron images of the same sample area were taken at a 1500 magnification. Backscatter imaging distinguishes particles based on atomic weight. Targeted energy dispersive X-ray analysis (T-EDXA) uses backscatter images to determine inorganic elemental composition of foulant material. The wavelengths of X-ray were used to identify the presence and relative amounts of the chemical elements as the inorganic fraction in the foulant sample.
2.4. Measurements of organic fractions in foulant materials The organic fraction of the foulant material was quantified at the University of North Carolina (Chapel Hill, NC, USA). The foulant material was gently scraped from small segments of membrane surface (surface area is 100 cm2), added to 100 mL of ultrapure laboratory grade water and sonicated for 20 min (Croue et al., 2003). The sample was then filtered through a 0.45 mm Durapore hydrophilic PVDF membrane. The filtrate was analyzed by a Shimadzu Total Organic Carbon Analyzer model TOC-5000 (UV-persulfate digestion method). With the use of a 0.45 mm filter, the term ‘‘dissolved organic carbon’’ (DOC) is used here to describe the measurement technique, although this material would have originally been present in colloidal or particulate form on the membrane surface. Assuming that the scraping and sonication steps were efficient prior to passage of the sample through the 0.45 mm filter, DOC represents almost all of the organic carbon that had accumulated on the membrane surface. Polysaccharides in the fouling layer of all three unrolled high-pressure membranes were quantified by the phenol– sulfuric colorimetric method (Cho and Fane, 2002). This method was developed to measure total carbohydrates but has been adopted by various research groups to measure EPS (Cho and Fane, 2002; Kim and DiGiano, 2006). Polysaccharides are assumed to be a major contributor to EPS and hydrolyzed by sulfuric acid to products that react with phenol to produce a stable yellow color that can be measured spectrophotometrically (Hanson and Philips, 1981). The membrane specimen was cut into smaller segments and dissolved directly in 2 mL of 2.5% phenol solution after which 10 mL of 98% sulfuric acid was added for a 10-min reaction period at ambient temperature and 15 min at 25 1C. The absorbance at 488 nm was then measured by a UV–Vis spectrophotometer (U-2000, Hitachi) and compared against a calibration curve developed with glucose standard solutions that were prepared in the same way. High-pressure, size exclusion chromatography (HPSEC) was used to characterize the organic fraction of the foulant material that was removed from each of the unrolled highpressure membrane elements. The segments cut from each membrane were about 100 cm2 in size. The foulant samples for HPSEC analysis were prepared in the same manner as those for the organic carbon measurement described above. Filtrates from the 0.45 mm filtration step were sent to the University of Colorado (Boulder, CO, USA) for HPSEC analysis.
42 (2008) 697– 706
HPSEC separates molecules on a chromatographic column according to their molecular weight. The traditional detector is fixed wavelength (254 nm) UV. This detector is sensitive to humic natural organic matter (NOM) that has aromatic structural features but is less sensitive to non-humic NOM that does not have aromatic features such as polysaccharides. An on-line DOC detector that allows recognition of NOM components with low (e.g., proteins) or no (e.g., polysaccharides) UV absorption was implemented. A variable wavelength UV detector is used to distinguish NOM constituents with different UV spectra. The high-pressure liquid chromatography (HPLC) column was packed with Toyopearl (HW-50S) resin ˚. beads with a diameter of 20–40 mm and a pore size of 125 A The maximum detectable molecular weight was about 80 kDa.
2.5.
Measurement of microbial foulants
Membrane samples were cut from each unrolled membrane element, mounted with double-sided tape on a aluminum stubs, and sputter coated (Hummer X sputter coater, Anatech Ltd.) with a thin layer of approximately 10–20 nm of a conductive metal (60% gold and 40% palladium alloy). They were stored in a dry, dust-free environment until SEM (SEM, Stereoscan 200, Cambridge) observations of microbial foulants. The accelerating voltage was 15 kV and the working distance was about 20 mm. A 200 magnification was used to obtain appropriate resolution of the samples.
3.
Results and discussion
3.1.
Microscopic observations of foulant layers
SEM images taken from three high-pressure membrane types (X20, ESPA2 and NF90) are presented in Fig. 2. General conclusions cannot be reached about the effects of element position on the extent of fouling because the original positions of the membrane elements (first and sixth) were reversed during May 2005 due to the occurrence of irreversible fouling and membrane maintenance. These SEM images, however, show qualitatively that the fouling layer appeared much less dense for high-pressure membranes receiving UF filtrate than those receiving MF filtrate. While these measurements are only qualitative, microbial foulants (elliptical- or spherical-shaped cells) and perhaps abiotic foulants were present. A lower density of microbes on high-pressure membranes pretreated by UF may be due to more effective removal of microbes and/or removal of small but important amounts of substrate that results in the suppression of the microbial activity by pretreatment with UF membranes.
3.2.
Inorganic and organic fractions of foulant layer
The percent contribution of inorganic and organic foulants was estimated from the LOI test performed by Membrane Forensics on foulant materials isolated from the high-pressure membrane elements as described above. As listed in Table 3, the foulant layer on the MF pretreated,
ARTICLE IN PRESS WAT E R R E S E A R C H
701
42 (2008) 697 – 706
Fig. 2 – SEM images of foulant materials deposited on X20 (A), ESPA2 (B) and NF90 (C) membranes pretreated by MF membrane and X20 (D), ESPA2 (E) and NF90 (F) pretreated by UF membrane (element position #1). Table 3 – Percentage of inorganic and organic materials in foulant by loss of ignition analysis (membrane forensics) MF pretreatment
Organic (%) Inorganic (%)
UF pretreatment
ESPA2
X20
NF90
ESPA2
X20
NF90
49.6 50.4
54.7 45.3
43.9 56.1
43.9 56.1
44.4 55.6
44.4 55.6
high-pressure membranes was approximately equally divided on a percentage by weight basis between organic and inorganic material. The contributions of organic and inorganic foulants were 45.5% and. 54.5%, respectively, for the high-pressure membranes receiving UF pretreatment. Thus, the choice of pretreatment had little effect on the inorganic–organic composition of the foulant layer. Based on the large inorganic fraction of foulants, the presence of inorganic chemicals in the granular media, filtered, secondary effluent, can cause significant scale formation once their solubility products are exceeded in the concentrate stream. The membrane element selected for the LOI analyses was the fourth in a series of six for each membrane train. Given that the recovery of each element was about 11%, the feed stream would have been concentrated by a factor of 37% by the end of the fourth element. The extent of precipitation of inorganic compounds could be estimated from measurements of the concentrations of the inorganics in the feed stream and the possible solubility equilibrium; however, such calculations are beyond the scope of this study.
3.3.
Specific inorganic chemicals in foulant material
Results of T-EDXA of the foulant materials are shown in Fig. 3. The order of decreasing prevalence of inorganic ions for the MF pretreated, high-pressure membranes was P (30%), Al (25%) and Ca (15%). Similar results were also obtained with the NF90 high-pressure membrane. P-containing precipitates are very likely present on the membrane because of ineffective precipitation of P within the North Buffalo Wastewater Treatment Facility by addition of 1–27 mg/L of sodium aluminate (as Al) prior to the aeration basins (the average addition was about 5 mg/L). P concentrations in the secondary effluent were always significantly higher than the threshold concentrations considered necessary for balanced microbial growth (40.2 mg/L). As also shown in Fig. 3, the order of prevalence of inorganic ions shifted for UF-pretreated membranes. Instead of P, the most abundant ion was Fe (65%) followed by P (20%) and Ca (10%). Thus, the addition of 4 mg/L of FeCl3 to the UF feed stream, as recommended by the manufacturer to reduce pore fouling by formation of Fe-containing floc, was not
ARTICLE IN PRESS 702
WA T E R R E S E A R C H
42 (2008) 697– 706
70
35
34.6
Aluminum Manganese
30
Iron
Phosphorous
27.2
Calcium Iron
25
Titanium
20
Sulfur
15.4
15
Zinc
12.9
Silicon
10
Magnesium Potassium
4.2
5
Percentage of Inorganic Foulant (%)
Percentage of Inorganic Foulant (%)
40
Chloride
Calcium
22.9 22.6
Iron
20
Silicon Zinc
15
Sulfur
12.6
Titanium
10
Chloride Potassium
3.8
5
2.5 1.8 1.5 1.4
Percentage of Inorganic Foulant (%)
Aluminum Manganese
Titanium
20
Magnesium
9.2
10
Potassium
3.7
1.7 1.6 1.4 0.9 0.6 0.4 0.3
64.3
Iron
60
Phosphorous Calcium
50 Aluminum
40
Sulfur Silicon
30 Titanium
19.4 20
Chloride Potassium
7.6
10
3.2
1.5 1.4
1.4
0.7 0.5
0
45
70
40
Aluminum
32.5
Calcium Magnesium
25.8
Iron
25
Sulfur
20
Tin
15.1 14.6
Titanium
15
Silicon
10 Zinc
4.1 2.8 2.2 1.8
Percentage of Inorganic Foulant (%)
Phosphorous
Percentage of Inorganic Foulant (%)
Sodium
23.8
0.2 0.1
0
5
Silicon Sulfur
30
70
30
30
Aluminum
40
Phosphorous
30.6 Percentage of Inorganic Foulant (%)
Calcium
50
0
35
35
Phosphorous
56.4
1.8 1.6 1.2 0.9 0.1 0.1
0
25
60
Iron
60
Phosphorous Calcium
50
Aluminum
40
Sulfur Chloride
30 Silicon
20 20
Potassium
9.7
Magnesium
10 2.7
0.8 0.3
0
64.8
1.5
0.6
0.3 0.3
0.1
0
Fig. 3 – Distribution of inorganic foulants as measured by T-EDXA on ESPA2 (A). NF90 (C) and X20 membrane (E) pretreated by MF membrane and ESPA2 (B), NF90 (D) and X20 (F) membrane pretreated by UF membrane. (A) MF-Hydranautics ESPA2, (B) UFHydranautics ESPA2, (C) MF-Filmtec Nf90, (D) UF-Filmtec Nf90, (E) MF-Trisep X20, (F) UF-Trisep X20.
effectively bound into precipitates that could be rejected by UF. Also indicated in Fig. 3 is a much higher fraction of Al in the foulant layer on high-pressure membranes pretreated by MF than by UF (about 20% vs. 3%). This could be due to more effective rejection by UF of poorly settling Al precipitates that had carried over from the addition of sodium aluminate in
the activated sludge process. A much lower P in the UF than MF filtrate is possibly related. The average of eight measurements of orthophosphate over the pilot-plant study was 0.42 mg/L for UF compared with 0.85 mg/L for MF. The alternative explanation for lower P in the UF filtrate is additional precipitation by Fe. In either case, it is possible that the presence of less Al–P colloidal material on high-pressure
ARTICLE IN PRESS WAT E R R E S E A R C H
membranes pretreated by UF was an important explanation for less fouling. Although difficult to prove, the dominance of Fe in the foulant layer on the high-pressure membranes pretreated by UF as shown in Fig. 3 may also be an explanation for lower fouling of high-pressure membranes receiving UF than MF pretreatment. FeCl3 may not only alter the nature of the organic foulants that accumulated on the membrane surface but also increase the porosity of the foulant layer, which in turn produced less resistance to filtration.
3.4.
Accumulation of DOC in foulant material
The effect of pretreatment on DOC accumulation of highpressure membranes (X20, ESPA2 and X20) is presented in Fig. 4. The DOC on the membrane surface is expressed as mg of DOC per m2 of unit filtration area. The average DOC for the first and sixth elements will be discussed because these element positions were reversed during the course of the pilot-plant study such that longitudinal variation in foulant accumulation could not be analyzed. As shown in Fig. 4, the DOC on each of the three highpressure membrane types that followed UF pretreatment was from 35% to 59% of the DOC measured on these same membranes following MF pretreatment. Less DOC accumulation is consistent with less decline in specific flux during the pilot-plant study for UF pretreatment. While the DOC accumulations caused by UF and MF pretreatment were very different, DOC removal by UF was only 4% higher (15% vs. 11%) than by MF (Reardon et al., 2007). In fact, the difference in DOC removal between UF and MF was not significant statistically (p40.05), albeit the sample size was relatively small to make reliable statistical inferences. The difference in DOC removal should be small because of the high MWCO (150 kDa) of the UF membrane and the fact that the presence of the low molecular weight material is reasonable in granular media, filtered secondary effluent.
Although a small fraction, the differential in DOC between UF and MF filtrates represents colloidal material in the approximate size range of 0.01–0.2 mm that could produce a disproportionately large fraction of the foulant on the highpressure membranes (Howe and Clark, 2002). This additional colloidal fouling fraction that was removed by UF but not by MF pretreatment may have higher affinity for high-pressure membranes. Accordingly, it could explain why DOC accumulation on the high-pressure membranes receiving UF pretreatment was 35–59% of that for membranes receiving MF pretreatment. A difference in in-situ generation of organic matter by microbial growth (biofouling) is yet another possible explanation for far less DOC accumulation on membranes receiving UF pretreatment. Although quantitative evidence is lacking, the biofouling potential could be lower for UF than MF filtrate because of higher removal of seeding microbes by UF than MF given that pore sizes are about ten times smaller; higher removal of biodegradable substrates caused by FeCl3 addition to the UF feed stream; and/or inactivation of microbes by addition of NaOCl on every fourth backwash for UF unit.
3.5.
450 MF pretreatment
400
UF pretreatment
mg of Polysaccharides/m2
80 70 mg of DOC/m2
Accumulation of polysaccharides in foulant material
Fig. 5 shows the average value of the polysaccharide accumulation between the first and sixth element for each high-pressure membrane. Consistent with the results of DOC accumulation, there was far less accumulation of polysaccharide (27–38%) on membranes receiving UF pretreatment than MF pretreatment. Polysaccharides are often the main component of biopolymers, or EPS, and are typically in the colloidal size range which is defined as organic matter greater than a molecular weight of 3500 Da (Jarusutthirak et al., 2002). The maximum size of foulant material that contains polysaccharides, however, could be much higher than 3500 Da. For example, in a separated bench-scale study using this same UF hollow fiber and the same feedwater, polysaccharides
100 90
703
42 (2008) 697 – 706
60 50 40 30 20
MF pretreatment UF pretreatment
350 300 250 200 150 100 50
10
0
0 ESPA2
NF90 High pressure membranes
X20
Fig. 4 – Comparison of DOC accumulation in foulant material per unit area of each high-pressure membrane for MF and UF pretreatment.
ESPA2
NF90 X20 High pressure membranes
Fig. 5 – Comparison of polysaccharide accumulation in foulant material per unit area of each high-pressure membrane for MF and UF pretreatment.
ARTICLE IN PRESS 704
WA T E R R E S E A R C H
42 (2008) 697– 706
polysaccharides have the same molecular formula as glucose, the mass fraction of DOC should have been 0.4 rather than 0.2. The difference in this ratio could have resulted from the different techniques used to obtain samples for DOC and polysaccharides. The sample for DOC was scraped off of the membrane and sonicated by 20 min before passing through the 0.45 mm filter, while polysaccharides were measured directly after chemical agents came into contact with the membrane specimen. Thus, DOC analysis may not have captured all of the organic carbon due to lack of complete removal from the surface and/or inability to solubilize all organic material by sonication. Both of these effects would cause DOC to be lower than predicted from the molecular formula of polysaccharides.
accumulated on the membrane surface over the course of many backwashing cycles (Kim and DiGiano, 2006). Thus, a small fraction of polysaccharides in granular media filtered, secondary effluent is apparently larger than the stated MWCO of 150 kDa. These polysaccharides could either be generated from biological treatment or they have been present in the raw wastewater. The former source seems more likely based on other studies (Jarusutthirak et al., 2002; Cho and Fane, 2002). Less accumulation of polysaccharides on the highpressure membranes that were pretreated by UF could be consistent with the removal of a fraction of very large polysaccharide-containing colloids that were not rejected by MF. An alternative explanation for observing less polysaccharide accumulation on membranes receiving UF pretreatment is the notion of a lower biofouling potential in UF than in MF filtrate that produced less in-situ generation of organic matter by microbial growth as explained earlier. Although measured by two different techniques, the mass of DOC (Fig. 4) was compared to the mass of polysaccharides (Fig. 5). The ratio of DOC to polysaccharides, as expressed by its glucose equivalent (C6H12O6), was about 0.2. If the
3.6.
HPSEC results on foulant materials
The graphical representation of HPSEC results in Fig. 6 is in order of decreasing molecular weight because this corresponds to the order of appearance in the column eluant. The correspondence between peaks in the column elutriates and
0.5
0.5 MF-X20-1
MF-ESPA2-1
MF-X20-6
0.4
MF-ESPA2-6
0.4
UF-X20-1
UF-ESPA2-1
UF-X20-6
Responses
Responses
UF-ESPA2-6
0.3
0.2
0.1
DOC
UVA 0.0 1000000 100000
0.3
0.2
0.1
10000
1000
100
10
1
DOC
UVA 0 1000000 100000
Molecular Weight (dalton)
10000
1000
100
10
Molecular Weight (Da)
0.5 MF-NF90-1 MF-NF90-6
0.4
UF-NF90-1
Responses
UF-NF90-6
0.3
0.2
0.1
DOC
UVA 0 1000000 100000
10000 1000 100 Molecular Weight (Da)
10
1
Fig. 6 – HPSEC analysis with DOC and UV detectors of foulant sample removed from X20 (A), ESPA2 (B) and NF90 (C).
1
ARTICLE IN PRESS WAT E R R E S E A R C H
molecular weight needs to be cautiously interpreted. Firstly, a homologous series of an organic compound is used to obtain a correspondence between time of elutriation and molecular weight of the reference compound (largest molecular weight appears first). That is, the elutriation time of non-reference organic compounds with different structural features but the same molecular weight can be different. Secondly, the maximum detectable molecular weight is based on the pore size of the column packing materials (80 kDa in this study). Finally, the first peak to appear in the elutriation from injection of unknown mixture of organic compound is a collective response from all organic compounds of the reference molecular size at that elution time and larger (up to maximum detectable of 80 kDa). Thus, the first peak shown in Fig. 6 represents all molecular weights greater than about 20 and less than 80 kDa. The DOC peak responses in Fig. 6 show clearly that the largest is at a molecular weight of about 20 kDa regardless of the pretreatment. This eluted organic fraction comprises nonhumic, low aromaticity structures such as protein, polysaccharides and/or amino sugar that have been noted in secondary effluent (Jarusutthirak et al., 2002). The smaller DOC peaks at low molecular weights (from 100 to 1000 Da) should represent humic substances and organic acids and these are also similar in molecular weights for all membranes. However, the peak heights are lower for UF than MF pretreatment. This is consistent with observations of less accumulation of DOC and polysaccharides (Figs. 4 and 5) on membranes receiving UF pretreatment. Although HPSEC results are reported for the first and sixth elements, the effect of element position on the results cannot be determined because the original first and sixth elements were reversed part way through the pilot-plant study as was mentioned in discussion of the SEM results. Evidence to support the presence of biologically derived foulant material was provided by Fourier transform infrared (FTIR) spectra of foulant samples (Reardon et al., 2007). The FTIR spectra showed strong absorption peaks at 3397–3431, 1644–1656 and 1058–1086 cm1, and minor peaks at 2917–2928 and 545–567 cm1 for high-pressure membranes receiving either MF or UF pretreatment. Previous studies (Croue et al., 2003; Jarusutthirak et al., 2002) indicate that absorption bands at 1650, 1600, 1550, 1410 and 1060 cm1 are characteristic of proteins, amino sugars and carbohydrates that compose bacterial cell walls. Therefore, while the presence of polysaccharides in the raw wastewater cannot be ruled out, the more likely source is production in the biological treatment process. The major DOC peak of each HPSEC for both UF and MF pretreatment raises questions as to the origin of this nonhumic fraction of foulant materials. Its molecular weight (20 kDa) is far less than the MWCO of the UF membrane (150 kDa). If this 20 kDa material had originated directly from granular media, filtered secondary effluent, it would have passed through both the MF and UF. Therefore, the DOC peak heights should have been the same for membranes receiving UF and MF pretreatment. Yet, the peak height was lower for the membranes receiving UF than MF pretreatment (Fig. 6). One possible explanation for less 20 kDa material is that the addition of FeCl3 caused preferential removal of this small
42 (2008) 697 – 706
705
molecular weight colloidal fraction or it increased in size by the coagulating action; however, supporting data are not available for either mechanism. Another possibility is in-situ generation of an organic foulant by biofouling that produced a low molecular fraction of DOC directly upon the highpressure membranes. That is, the observation of a lower DOC peak height for organic material removed from membranes receiving UF pretreatment could be due to a lower biofouling potential of UF filtrate for the reasons noted earlier.
4.
Conclusions
Membrane autopsies showed less foulant accumulation by UF than MF pretreatment that suggested possible explanations for less flux decline of the high-pressure membranes receiving UF pretreatment. However, less accumulation of a specific foulant that might be responsible for less flux decline could not be identified. The weight fractions of inorganic and organic foulants were approximately equal, suggesting that either or both could contribute to a decline in specific flux with time but a differential in weight factions caused by UF compared with MF pretreatment was not apparent. Nevertheless, there was less Al–P colloidal material on the UFpretreated, high-pressure membranes and this may be an important explanation for less flux decline. Alternatively, the differential in DOC and polysaccharide accumulation between membranes receiving UF and MF pretreatment could also be consistent with less extent of flux decline for membranes receiving UF treatment. The UF filtrate would be expected to have less foulant material owing to somewhat greater removal of DOC (15% vs. 11%) perhaps caused by the addition of FeCl3. Moreover, FeCl3 could not only reduce the fouling of the UF membrane but also reduce the fouling potential of organic matter in the UF permeate and thus in the feed to the high-pressure membranes. Unfortunately, there were no direct data to support these possible mechanisms involving FeCl3 and evidence from the literature (Howe and Clark, 2002) is restricted to studies of natural waters. Although difficult to prove, less in-situ generation of organic foulants may be another possible explanation for less foulant accumulation on UF- than MF-pretreated, high-pressure membranes. The evidence is less accumulation of polysaccharides of a molecular weight fraction that should have passed through either MF or UF pretreatment. Equal amounts of this material should have been present on high-pressure membranes receiving either UF or MF pretreatment if it was not generated in situ. A lower biofouling potential after UF pretreatment could also be linked to autopsy results showing fewer microbes on high-pressure membranes receiving UF than MF pretreatment, albeit the differences are very difficult to quantify. Fewer microbes could be due to more efficient rejection by UF than MF. Alternatively, the addition of NaOCl on every fourth backwash of UF could also have inactivated microbes. This would lead to the suggestion that the disinfection strategy rather than membrane properties was responsible for less observed fouling after UF pretreatment. In summary, the autopsy results did not provide a clearcut, fundamental explanation of less flux decline for high-pressure membranes that received UF pretreatment.
ARTICLE IN PRESS 706
WA T E R R E S E A R C H
Nonetheless, valuable insights were gained into the possible roles of polysaccharides and inorganic colloids that need to be explored more fully in carefully controlled experiments that were not possible to achieve in these pilot-plant studies.
Acknowledgments We would like to thank the Water Environment Research Foundation and the North Carolina Water Resources Research Institute for their financial support. We also thank Membrane Forensics with the help of membrane autopsy works and Dr. Gary Amy with HPSEC analysis. R E F E R E N C E S
Adin, A., Elimelech, M., 1989. Particle filtration in wastewater irrigation. J. Irrig. Drainage Eng. (ASCE) 115, 474–487. Cho, B., Fane, A.G., 2002. Fouling transient in normally sub-critical flux operation of a membrane bioreactor. J. Membr. Sci. 209, 391–403. Croue, J.P., Grasset, L., Bacle, S., Jacquemet, V., 2003. Nanofiltration membrane autopsy of a full-scale unit: characterization of the organic and inorganic constituents of the foulants. In: Proceedings of the American Water Works Association Membrane Technology Conference, American Water Works Association, Denver, CO. Decarolis, J., Hong, S., Taylor, J., 2001. Fouling behavior of a pilot scale inside-out hollow fiber UF membrane during dead-end filtration of tertiary wastewater. J. Membr. Sci. 191, 165–178. Fonseca, A.C., Summers, R.S., Greenberg, A.R., Hernandez, M.T., 2003. Isolating and modeling critical factors governing biofouling of nanofiltration membranes. In: Proceedings of the American Water Works Association, Denver, CO. Ghayeni, S., Madaeni, S., Fane, A.G., Schneider, R.P., 1996. Aspects of microfiltration and reverse osmosis in municipal wastewater reuse. Desalination 117, 49–60. Hanson, R.S., Philips, J.A., 1981. Chemical composition. In: Gerhardt, P., et al. (Eds.), Manual of Methods for General
42 (2008) 697– 706
Bacteriology. American Society for Microbiology, Washington, DC, pp. 328–364. Her, N., Amy, G., Pecheux, A., Gatel, D., Cavard, J., 2003. Nanofiltration (NF) membrane foulant identification and fouling mechanisms. In: Proceedings of the American Water Works Association Membrane Technology Conference. Howe, K.J., Clark, M.M., 2002. Fouling of microfiltration and ultrafiltration membranes by natural waters. Environ. Sci. Technol. 36 (16), 3571–3576. Jarusutthirak, C., Amy, G., 2002. Membrane filtration of wastewater effluents for reuse: effluent organic matter rejection and fouling. Water Sci. Technol. 43 (10), 225–232. Jarusutthirak, C., Amy, G., Croue, J.P., 2002. Fouling characteristics of wastewater effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination 145, 247–255. Kim, J., DiGiano, F.A., 2006. A two-fiber, bench-scale test of ultrafiltration (UF) for investigation of fouling rate and characteristics. J. Membr. Sci. 271, 196–204. Neis, U., Tiehm, A., 1997. Particle size analysis in primary and secondary wastewater effluents. Water Sci. Technol. 36 (4), 151–158. Parameshwaran, K., Fane, A.G., Cho, B.D., Kim, K.J., 2001. Analysis of microfiltration performance with constant flux processing of secondary effluent. Water Res. 35 (18), 4349–4358. Reardon, R.D., Paranjape, S.V., Foussereau, X.J., DiGiano, F.A., Aitken, M.D., Kim, J.H., Chang, S., Cramer, R., 2005. Membrane treatment of secondary effluent for subsequent reuse. Water Environment Research Foundation Report (01-CTS-6). IWA Publishing, London, UK. Reardon, R.D., DiGiano, F.A., Hobbs, C.M., Treadway, J.M., Kim, J.H., 2007. Membrane treatment of secondary effluent for subsequent use: phase 2—pilot plant comparisons of MF and UF for pretreatment of high-pressure membranes. Water Environment Research Foundation Report (01-CTS-6a). IWA Publishing, London, UK. Uhl, W., Hagmeyer, G., Huck, P.M., 2003. Biofouling of NF and RO membranes in drinking water production from surface waters; theoretical consideration of prerequisites and possible prevention strategies. In: Proceedings of the American Water Works Association, Denver, CO. Van der Bruggen, B., Vandecasteele, C., 2001. Flux decline during nanofiltration of organic components in aqueous solution. Environ. Sci. Technol. 35, 3535–3540.