Journal of Membrane Science 382 (2011) 328–338

Contents lists available at SciVerse ScienceDirect

Journal of Membrane Science

journal homepage: www.elsevier.com/locate/memsci

Water reclamation using reverse osmosis: Analysis of fouling propagation given

tertiary membrane filtration and MBR pretreatments
Fraser C. Kent a,∗ , Khosrow Farahbakhsh a , Basuvaraj Mahendran b , Magdalena Jaklewicz c ,
Steven N. Liss b,d , Hongde Zhou a
a
School of Engineering, University of Guelph, 50 Stone Road East, Canada N1G 2W1
b
School of Environmental Sciences, University of Guelph, 50 Stone Road East, Canada N1G 2W1
c
GE Water & Process Technologies, 5316 John Lucas Dr., Burlington, ON, Canada L7L 6A6
d
School of Environmental Studies and Department of Chemical Engineering, Queen’s University, Kingston, ON, Canada K7L 3N6

a r t i c l e i n f o a b s t r a c t

Article history: A series of reverse osmosis (RO) bench-scale experiments were conducted using cross-flow cells at a
Received 17 January 2011 municipal wastewater treatment plant. The experiments were used to observe the accumulation of
Received in revised form 4 May 2011 effluent organic matter (EfOM) on the membrane surfaces over time and to compare two pretreatment
Accepted 12 August 2011
options including a membrane bioreactor (MBR) and conventional activated sludge with tertiary mem-
Available online 22 August 2011
brane filtration (CAS-TMF). The membrane surfaces after filtration from 3 to 36 days were characterized
by quantifying organic species, especially proteins and polysaccharides using Fourier-transform infrared
Keywords:
reflectometry (FTIR), correlative microscopy using confocal laser scanning microscopy (CLSM) and scan-
Water reclamation
Reverse osmosis
ning electron microscopy (SEM). The results suggest that the MBR resulted in less RO fouling. It was also
Fouling found that observations of fouling within the first few days of operation were distinctive from obser-
Membrane bioreactor vations following operations after a month or more, suggesting that short-term studies are unable to
Membrane filtration predict long-term fouling trends. It was also found that proteins were the predominant foulant found on
the surface within the first 2 weeks, but polysaccharide deposition became much higher than proteins
after 4 weeks of operation. An examination into the underlying fouling mechanisms is provided.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction matter and microorganisms continue to cause significant RO foul-

ing through deposition of organic macromolecules and biofilm
Water reclamation using reverse osmosis (RO) is increasingly development on membrane surfaces. An understanding of the
applied around the world to support growing populations. Fresh- mechanisms involved in both organic fouling and biofouling are
water sources are becoming scarce and it has been demonstrated the focus of many recent research projects however more research
that water reclamation has advantages over desalination of seawa- is required in order to provide better strategies for treatment or
ter in regions which desalinate [1]. Reclamation has the two-fold fouling control.
benefit of reducing the impact of withdrawal from source waters RO fouling has been divided into different categories based on
while also reducing the volume of wastewater released to the envi- the types of foulants involved. Particulate fouling, scaling, organic
ronment. Membrane fouling, however, is a problem with RO that fouling and biofouling are the four types that are often referenced
is prevalent in water reclamation applications. It has been found [5,6]. In particular, organic fouling and biofouling need further
that the use of low-pressure pretreatment membranes is useful clarification in the context of water reclamation. Organic fouling
for minimizing RO fouling [2,3] and as a result, integrated mem- refers to a permeability reduction caused by the accumulation of
brane processes are becoming more popular. A recent survey of RO dissolved or colloidal organic species onto a membrane surface.
plants treating wastewater effluent indicated that seventy-eight Biofouling involves attachment and growth of microorganisms on
(78%) of the facilities in the survey use membrane pretreatment the surface of a membrane [7]. Biofilms are composed of bacteria
[4]. Although low-pressure membranes provide high quality feed and extracellular polymeric substances (EPS) that are produced by
water for RO systems, the presence of low concentrations of organic the bacteria as well as arising from cell lysis. As with organic foul-
ing, biofouling is particularly prevalent within water reclamation
applications, since upstream treatment processes generally utilize
∗ Corresponding author. Present address: Box 660, Bracebridge, ON, Canada P1L bacteria and biofilms in engineered systems.
1T9. Tel.: +1 223 71 27 97 22. The organic fouling problems that have been reported for
E-mail address: fraskent@gmail.com (F.C. Kent). RO water reclamation applications have resulted in a number of

0376-7388/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2011.08.028
 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338 329

studies on this topic [8,10–13]. Dissolved organic matter con- Table 1

Summary of water quality from pretreatments.
tained in treated wastewater effluent, commonly known as effluent
organic matter (EfOM), has been found to be mostly made up of sol- Analyte Units TMF MBR
uble microbial products (SMP) [8] and has been associated with pH 7.3 7.6
significant RO fouling. Included in SMP are a number of differ- SUVA L/mg DOC m 1.56 2.00
ent constituents; however proteins and polysaccharides, which Colour PCU 20.6 30.8
are ubiquitous in wastewater effluent, have been identified as Alkalinity mg/L as CaCO3 242 294
COD mg/L 20.7 14.1
key fouling contributors [9]. Li et al. [10] suggested that proteins
TOC mg/L 7.7 5.6
and polysaccharides are key foulants for RO, with polysaccharides, Total nitrogen mg/L 28 26
in particular, causing relatively high losses of permeability [11]. Fecal Coliforms cfu/100 mL 0 0.1
Lee et al. [12] found that fouling with alginate, a model polysac-
charide, was exacerbated in the presence of calcium ions due to
inter-particle bridging. Other researchers have reported strong cor- confocal laser scanning microscopy (CLSM) and scanning electron
relations between intermolecular adhesion forces, measured using microscopy (SEM) to be employed. A series of 15 experimental
atomic force microscopy (AFM), and RO fouling rates in controlled runs were conducted using filtration durations between 3 and 36
experiments [13]. However, Li et al. [10] and Ang and Elimelech days. Each run involved parallel operation of two cross-flow cells,
[14] identified that there are synergistic effects when multiple one with MBR pretreatment and one with CAS-TMF pretreatment.
organic species are present in RO feed and it has been acknowl- Upon termination of each run, the surface foulants were anal-
edged that performing studies with individual organic species may ysed.
significantly underestimate membrane fouling. This suggests that Based on preliminary investigations that showed a statistically
conducting studies with real wastewater effluent is important. significantly higher total organic carbon (TOC) concentration in the
Research on RO fouling has also focused on biofouling. Flem- TMF effluent it was hypothesized that rate of foulant accumulation
ming et al. [15] suggested that it is the most difficult fouling and fouling on the RO membranes with TMF pretreatment would
mode to prevent since biofilms can develop even if only a few be more rapid. An attempt was also made to relate the findings to
cells are present, whereas all other fouling modes can be sig- more controlled experiments that have been conducted by other
nificantly reduced through removing a high percentage of the researchers. An effort was also made to compare the fouled surfaces
constituents during pretreatment. The research in this area has of RO membranes given TMF and MBR pretreatment in an attempt
focused on understanding how biofilms develop, what impacts to explain the difference in RO fouling rates that were observed in
their growth and how they negatively impact the filtration process. the pilot-scale testing that was conducted as a separate study.
Biofilm development involves four distinct stages: deposition of
and adhesion of microbial cells, multiplication into microcolonies, 2. Experimental
maturation and detachment [16]. Herzberg and Elimelech [17] con-
ducted controlled studies using dead and living bacteria and were 2.1. Experimental setup
able to show that bacterial cells alone do not cause a reduction
of permeability; rather the EPS matrix that they form is responsi- Pretreatments of wastewater were provided by ultrafiltration
ble for fouling. However, bacterial cells prevent diffusion of salts (UF) membranes for both TMF and MBR systems. Table 1 provides
away from the membrane surface causing an increase in concen- a summary of the permeate water quality characteristics for the
tration polarization and an increase in salt passage through the two pretreatment systems.
membrane. All testing was conducting using the six cross-flow cells that
The objective of the present study was to compare two different were set up in parallel to pilot-scale treatment systems. The SEPA
pretreatment methods in terms of their ability to prevent RO foul- cell (GE Water, Minnetonka, USA) is a commonly employed cross-
ing. Both pretreatment methods involved the use of low-pressure flow cell design that was designated by the USEPA as the standard
membranes to eliminate particulate fouling of the downstream RO. cell for evaluation of membranes [18]. It was used for all testing,
In one case a membrane bioreactor (MBR) was used and in the incorporating standard concentrate spacers and permeate carriers
other conventional activated sludge was used with tertiary mem- with continuous cross-flow rates to simulate the velocity profiles
brane filtration for polishing (CAS-TMF). Both of these pretreatment in a spiral-wound module. All cells were equipped with pressure
methods were operated given the same municipal wastewater. transducers to monitor trans-membrane pressure (TMP), rotame-
Prior to the bench-scale testing, a separate study was conducted ters for measuring concentrate flow rate and a custom permeate
using two identical pilot-scale RO units. The results of this sep- flow rate measurement system.
arate piloting study clearly showed that, for the conditions of All experimental runs were commenced with fresh RO mem-
the study, pretreatment with a MBR resulted in a significantly brane coupons. These coupons were cut from sheets of membrane
lower RO fouling rate (which was measured as a rate of change provided by the manufacturer (GE Water, Minnetonka, USA) that
of permeability). It was also found that MBR pretreatment con- were the same as those for the RO modules used in the pilot equip-
sistently provided an effluent with a significantly lower organics ment. The details of the pretreatment membranes, RO experimental
concentration. Given the nature of the pilot setup, it was not possi- conditions and RO membrane properties are given in Table 2.
ble to examine the RO membrane surfaces until each study was The cross-flow cells were set up onsite at a full-scale wastewater
completed, since any surface analysis would have involved the treatment plant in Guelph, Canada to allow the use of real munici-
destruction of the membrane modules. In order to investigate the pal wastewater and pretreatment systems. Fig. 1 shows how the
propagation of RO fouling, six bench-scale systems were used in cross-flow cells were incorporated into the experimental setup.
the present work. Three were fed with MBR permeate and the They were fed with concentrate from the associated pilot-scale RO
other three with CAS-TMF permeate. These cross-flow cells were systems. The concentrate pressure from the RO pilots was used
the focus of a series of experiments aimed at providing a better to drive permeation in the cross-flow cells, eliminating the need
understanding of how fouling occurs over time using a variety of for high-pressure pumps for the bench-scale testing. All cross-flow
surface analysis techniques applied after different filtration dura- cells were equipped with a needle valve on the feed and concentrate
tions. Use of the small RO membrane coupons allowed analytical lines to facilitate control over both the TMP and the concentrate
techniques such as Fourier transform infrared spectroscopy (FTIR), flow rate. Since the RO pilot systems operated at 15% recovery,
330 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338

Guelph Wastewater Treatment Plant

Primary Secondary
Chlorination
Screening Treatment Treatment
Rotating
Raw Media
Biological
Wastewater Filter
Contactor
De-gritting De-chlorination

Water Reclamation Research Pilot Plant

Tertiary
Membrane
Reverse
Filtration
Osmosis
RO1 Permeate
Fine (0.5mm) TMF RO1
Screening

RO1 Cross-flow
Test Cells
Membrane
(MBR)
Bioreactor
Reverse
Osmosis
Membrane RO2 Permeate
Bioreactor RO2
Filter

RO2 Cross-flow
Test Cells

Fig. 1. Process flow diagram of Guelph wastewater treatment plant and experimental setup.

this configuration allowed the introduction of slightly concen- in lower pilot-scale RO fouling compared with TMF pretreat-
trated feed to the cells. As a result, the cross-flow cells could be ment. To facilitate this, the cross-flow cells were operated in
considered to represent the end of a first RO element or the begin- pairs. Each experimental run with a cross-flow cell with MBR
ning of a second element in a typical RO full-scale treatment facility pretreatment was complimented with an associated cross-flow
for water reclamation. cell that had TMF pretreatment. Careful attention was given to
ensuring these pairs had approximately the same average flux.
2.2. Experimental procedure Each cross-flow cell pair had the same permeate volume upon
termination of the experimental run and harvesting of the RO mem-
To facilitate the investigation of fouling development on the brane.
RO membranes, the duration of experimental runs was varied The filtration runs can be divided into two phases based on the
from fairly short runs (3 days) to longer runs (36 days). Sur- analytical tests that were conducted on the membranes once the
face analyses were conducted for the different experimental runs runs were completed. In the first phase of experiments, 9 runs were
and results were plotted as a function of run length, or volume conducted. These filtration experiments were devoted to FTIR and
permeated per unit area. Ideally, this would give the effect of a SEM analyses. The membranes were analysed after the collection of
single RO membrane in operation that was analysed repeatedly 10, 20, 30, 40, 47, 65, 72, 84 and 95 L of permeate, respectively. The
over the course of the experiment. One of the major objectives second phase of experiments involved three runs which were anal-
of this work was to understand why MBR pretreatment resulted ysed using CLSM to determine the relative density of proteins and
polysaccharides, as well as the microbial cell density on the surface.
These analyses took place after 56, 90 and 120 L of permeate were
Table 2
Membrane properties and operating conditions. processed. Although the timing for analysis was determined based
on the volume processed, in the results this is expressed as filtered
TMF membrane pore size 0.02 mm
volume per unit area (L/m2 ) to allow the data to be compared with
TMF membrane chemistry Polyvinylidine fluoride (PVDF)
TMF membrane type Unsupported hollow-fibre
the pilot-scale results and with the results from other work.
MBR membrane pore size 0.04 mm Upon termination of a run, the membranes were carefully
MBR membrane chemistry Polyvinylidine fluoride (PVDF) removed from the cross-flow cells and cut into square pieces of
MBR membrane type Supported hollow-fibre approximately 2.5 cm × 2 cm and placed into labeled Petri dishes.
RO membrane chemistry Polyamide
For the FTIR and SEM analyses the samples were allowed to
RO membrane type Thin-film composite
RO membrane surface area 0.014 m2 dry, however, for the CLSM analyses, the samples were put in a
RO cross-flow rate 400 mL/min humidifier to prevent drying and immediately transported to the
RO flux range 12–16 Lmh laboratory to undergo sample preparation for CLSM analysis as
RO TMP range 10–20 bar
described below. All CLSM analyses took place within a few hours
RO average NaCl rejection 99%
of sampling of the membranes from the cross-flow cells.
 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338 331

2.3. Laboratory methods 1.0

MBR feed

2.3.1. Fourier transform infrared spectroscopy 0.9

TMF Feed

(Relative Permeability (P/Po)

Attenuated total reflectance Fourier transform infrared spec-
troscopy (FTIR) was employed to obtain finger printing FTIR spectra.
0.8
The analysis was carried out using a spectrum one Fourier trans-
form infrared spectrometer (PerkinElmer, USA) equipped with a
horizontal attenuated total reflectance accessory, with a diamond 0.7

internal reflectance element. Air-dried specimens were tested in

groups. Multiple spectra were collected. The most conservative 0.6
cases were compared.
0.5
2.3.2. Field emission scanning electron microscopy
The membrane surfaces were examined employing the Spectra 0.4
40 field emission scanning electron microscope (Zeiss, Germany). 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
Selected images were captured digitally as appropriate. Permeate volume per unit area (L/m )
2

The dry specimens were mounted on Cambridge stubs with

areas of interest exposed. To avoid excessive charging the surfaces Fig. 2. Relative permeability as a function of permeate volume per unit area for all
bench-scale runs conducted.
were coated with a thin layer of Iridium.

2.3.3. CLSM microscopy and image analysis intensity of surface area covered by cells on scanned membrane
2.3.3.1. Microscopy. The microscopy images of the membrane sam- surfaces using LCS-2.61.
ples were captured using an upright confocal laser scanning
microscopy (CLSM) system (Leica DM RE microscope connected to
3. Results and discussion
a Leica TCS SP2) (Leica, Germany) with three different visible light
lasers (argon, green helium), covering 6 excitation wavelengths.
3.1. Fouling results
Different magnification objectives (water-immersible 63× and also
40×, 20×, 10× Lenses (Leica)), and filter free spectral detector filter
The permeability was measured for each of the experimental
sets for monitoring the fluorescent staining were applied. Detection
runs to compare the fouling as a function of volume permeated
of polysaccharides was based on targeting glyconjugates specific to
for the two different pretreatments. Fig. 2 shows the relative per-
the lectin stain Concanavalin A and SYPRO orange which were used
meability (i.e. permeability divided by the initial permeability) as
to target all the proteins as described by Lin et al. [19]. The images
a function of volume filtered per unit area for both the MBR and
were processed using the confocal assistant software supplied by
TMF pretreatments. Note that all permeability values were tem-
the manufacturer (Leica Confocal Software (LCS, version 2.61)).
perature corrected to 20 ◦ C. Fouling occurred during all runs with
final permeability values varying from approximately 75% of the
2.3.3.2. EPS analysis. To observe EPS in the foulant layer, two dif- initial permeability for the shortest runs conducted (750 L/m2 ) to
ferent probes were collectively applied as described by Lin et al. approximately 60% of the initial permeability for the longest runs
[19]. Briefly, Concanavalin A, Alexa Flour 633 conjugate (5 mg/L, (8700 L/m2 ). It was interesting to note that three quarters of the
Invitrogen) to target carbohydrates (˛-Mannose, ˛-Glucose) and maximum permeability loss occurred within the first 750 L/m2 of
SYPRO orange (5 mg/L, Invitrogen) to target proteins were used. volume permeated. It is likely that compaction of the membranes
The fouled membrane samples were placed and stained in 7 cm dia. contributed to this initial permeability loss. However, the high
Petri-plates, incubated in the dark at room temperature (22 ± 2 ◦ C) initial fouling rate lasted for more than 24 h. This is a duration
for 30 min, and then washed gently three times with a phosphate beyond what other researchers have used to compact RO mem-
buffer (pH 7.0) to remove unbound probes. After washing, the branes [20,21]; although it should be noted that compaction runs
stained samples were immediately imaged by CLSM. Both 10× were conducted at 25–50% higher pressures in these two studies.
and 20× objectives, and a 63× water immersion objective, were Herzberg et al. [21] found a similar trend when conducting RO
used to view samples. Signals were recorded in the green channel filtration with secondary effluent with low-pressure membrane
(excitation 488 nm, emission 570 nm) for proteins and red channel polishing. These authors attributed this high initial flux decline
(excitation 633 nm, emission 647 nm) for polysaccharides. Z-stack to the accumulation of organic matter. Tansel et al. [22] reported
images were obtained beginning at the top of the fouling layer with changes in RO membrane morphology within 24 h of filtration with
optical slices taken every 0.4 ␮m through the foulant to determine a similar feed. They took images using AFM that identified a layer
the spatial distribution of EPS components while viewing samples of soft deposits forming on top of a strong sub-layer that firmly and
with a 63× water-immersible objective. The series of CLSM images rapidly covered the membrane surface. These authors suggested
were taken from different random locations on the RO membrane that the initial organic layer that deposited on the membrane sur-
samples and from replicate stained samples to acquire a number face is different from the subsequent layers. Brant and Childress
of images. The images were processed to measure the intensity [23] also suggested that foulant-membrane interactions only apply
of surface area covered by proteins and polysaccharides using the to the initial monolayer of fouling and that all subsequent fouling is
confocal assistant software (LCS, version 2.61). governed by foulant–foulant interactions. The large initial perme-
ability loss found in the present study may be associated with the
2.3.3.3. Microbial cell analysis. Acridine Orange (5 mg/L, Invitro- foulants that contributed to this initial sub-layer. Ang and Elimelech
gen) a nucleic acid selective fluorescent dye was used to target [20] conducted a fouling study using octanoic acid. They also found
microbial cells in the foulant layer. Samples staining and washing fouling trends similar to those shown in Fig. 2 and suggested that
procedures were similar to those mentioned above. For observa- the early significant loss of permeability was caused by an immedi-
tion, 10×, 20× and 63× water immersion lenses were used. Images ate absorption of octanoic acid molecules. The concept of an initial
were obtained and the images were processed to measure the fouling stage related to the deposition of a monolayer of organics,
332 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338

0.8
Δ Permeability (Lmh/bar)

0.6

0.4

0.2

0.0

3 data points associated

with pilot-scale runs
-0.2

-0.4
0 5000 10000 15000 20000 25000
2
Volume per unit area (L/m )

Fig. 3. Difference between TMF and MBR fouling as a function of permeate volume
per unit area for all bench-scale runs and pilot-scale runs.

that is different from subsequent organics deposition is discussed

further below.
To help compare the fouling rate for the cross-flow cells with
MBR pretreatment and the cross-flow cells with TMF pretreatment,
the total loss of permeability for each run with MBR pretreatment
was subtracted from the total loss of permeability for the paral-
lel run operated with TMF pretreatment. A positive value indicates
that TMF pretreatment resulted in more fouling than MBR pretreat-
ment. Fig. 3 shows the results in terms of difference in permeability
lost (or difference in fouling) as a function of permeated volume
per unit area. This analysis was done for all of the bench-scale runs
as well as the three pilot-scale runs conducted in the pilot-scale
study. As shown in the figure, there seems to be a positive trend
suggesting that, on average, the amount of fouling given TMF pre- Fig. 4. FTIR absorbance peaks for RO membrane surfaces associated with (a)
polysaccharides (1050 and 2940 cm−1 ) and (b) proteins (1665 and 1550 cm−1 ) for
treatment, relative to the fouling given MBR pretreatment becomes both MBR and TMF pretreatments as a function of permeate volume per unit area.
larger as the runs become longer. This is in agreement with the con-
clusions of the parallel pilot-scale study that was conducted where
it was noted that the difference in permeability for the two different
pretreatments became larger as the runs progressed. SMP [7,8,15,24,25]. In FTIR analyses, the detection of different
The results in Fig. 3 reveal an interesting phenomenon con- organic species can be achieved by associating the species with
cerning the short- and long-term fouling studies. This comparison their characteristic absorption responses at different wavelengths.
of two different RO pretreatments shows very different results in The responses are based on the functional groups present in the
short term experiments compared with longer term experiments. molecule and the associated wavelength that these functional
When considering all the tests that were conducted with less than groups absorb. According to Jarusutthirak and Amy [8], there are
5000 L/m2 , in 78% of these cases the TMF pretreatment exhibits characteristic peaks at approximately 1050 and 2940 cm−1 wave-
a lower fouling rate ( Permeability is negative). However, when lengths that are associated with the presence of polysaccharide-like
considering all the runs with greater than 5000 L/m2 of filtration, materials. There are also characteristic peaks at approximately
every result shows a higher fouling rate for TMF pretreatment ( 1550 and 1665 cm−1 wavelengths that are associated with the pres-
Permeability is positive). This suggests the potential for short-term ence of protein-like materials. In FTIR, quantification of organic
fouling experiments to provide inaccurate information on RO foul- species for different samples can also be achieved [26], however
ing given the long-term nature of RO operation in full-scale settings. a quantitative comparison can only be made for the same wave-
After approximately 5000 L/m2 of permeation, the higher fouling length through this technique. Therefore, in this study, responses
rate given TMF permeate as feed becomes clear. This corresponds to at a particular wavelength could be compared for different filtration
3–4 weeks of operation for a pilot RO system. A transition to a higher volumes and between samples with TMF and MBR pretreatments.
fouling rate for the RO membranes with TMF pretreatment after this However a comparison of quantities of constituents associated with
amount of time could be related to a transition from organic fouling different wavelengths (e.g. proteins and polysaccharides) could not
to biofouling. The higher concentration of organic matter in the TMF be achieved and the CLSM analyses were used to provide informa-
permeate may have provided more food for biofilm development tion in regard to relative quantities of proteins and polysaccharides.
causing an increased fouling rate for the RO with TMF pretreatment. For the FTIR and SEM analyses, 18 different cross-flow cells runs
This will be discussed further after more data are presented. were conducted (9 with MBR pretreatment and 9 with CAS-TMF
pretreatment) and harvested once a target cumulative volume was
3.2. FTIR results permeated. In Fig. 4 the absorbance peaks are shown for the 4 dif-
ferent wavelengths being considered within the FTIR spectrum for
Several researchers have used FTIR data to detect proteins the different runs.
and polysaccharides on the surfaces of membranes fouled by Three major observations can be made for these data:
 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338 333

0.60 0.20
TMF feed 2940 cm^-1

MBR feed 2940 cm^-1

Absorbance units
Absorbance units
0.15

0.50

0.10

TMF feed 1050 cm^-1

MBR feed 1050 cm^-1

0.40 0.05
0 3000 6000 0 3000 6000
Permeate volume per unit area (L/m2) Permeate volume per unit area (L/m2)

0.30 0.30

Absorbance units
Absorbance units

0.25

0.20

0.20
TMF feed 1665 cm^-1
TMF feed 1550 cm^-1
MBR feed 1550 cm^-1 MBR feed 1665 cm^-
1
0.15 0.10
0 3000 6000 0 3000 6000

Permeate volume per unit area (L/m2) Permeate volume per unit area (L/m2)

Fig. 5. FTIR absorbance data for samples with <6000 L/m2 of filtrations. Wavelengths of 1050 cm−1 (top left), 2940 cm−1 (top right), 1550 cm−1 (bottom left) and 1665 cm−1
(bottom right). Hatched lines are regressions for samples with TMF pretreatment; solid lines are for samples with MBR pretreatment.

1. For all wavelengths, responses were observed after 750 L/m2 of much higher absorbance values (after 5000 L/m2 ) did not exhibit a
filtration ranging from 0.1 to 0.5 absorbance units. This indicates relatively high degree of fouling. In general, it seems that as the foul-
the presence of organic material on the surface for the shortest ing layers grew, the change in permeability per unit of deposited
runs. foulant was small relative to the initial loss of permeability that
2. Between 750 and 5000 L/m2 of filtration, relatively small but occurred within the first 750 L/m2 of filtration.
consistent increases in absorbance were observed for all wave- The disproportionately large loss of permeability given the small
lengths as a function of permeate volume. This suggests that the absorption values observed for all wavelengths and samples after
sample surfaces incurred steady organic loading for this range of only 750 L/m2 of permeation could be associated with the forma-
filtrations volumes. The steady increase is difficult to observe in tion of an initial sub-layer as described by Tansel et al. [22]. The
Fig. 4 due to the scale and is illustrated in Fig. 5 where the FTIR sudden increase in the quantity of organics on the surface from
data for samples with filtration volumes less than 6000 L/m2 are 5000 to 7000 L/m2 could indicate a change in the way the foul-
presented. ing layer develops. There is no indication that the organic matter
3. Between 5000 and 6000 L/m2 a relatively large increase in deposition rate was higher for these longer runs, suggesting that
responses was observed for the samples with TMF pretreat- the change in the magnitude of peaks was associated not with
ment, however very little change was observed for samples with organics deposition, but a transition in the nature of foulant prop-
MBR pretreatment. Between 6000 and 7000 L/m2 of filtration the agation on the surface, perhaps due to biological activity. The fact
absorbance values for the MBR-RO samples did undergo a rela- that the magnitude of absorbance for TMF-RO samples with more
tively large change, however, the magnitude of responses at all than 5000 L/m2 of filtered water was higher than the magnitude
wavelengths remained well below the levels observed for the for MBR-RO samples suggests that the TMF permeate water qual-
samples with TMF pretreatment. This indicates that there was ity characteristics caused a more pronounced effect. A previous
a change in the nature of surface foulants after approximately study by Subramani and Hoek [27] demonstrated that bacterial
5000 L/m2 and that the change occurred earlier and was more adhesion can occur rapidly (within the first few hours of filtration)
substantial for the samples with TMF pretreatment. within cross-flow membrane filtration. However, it is likely that
the vast majority of microbes were removed by the membrane
In general the FTIR results suggest that there were two stages pretreatment in this study prior to RO filtration. It is interesting
in the propagation of organics deposition. First there was a surface that the data presented in Fig. 4 show a drastic change in foulant
loading of organic material that seemed to be relative to the amount propagation after weeks of filtration. The drastic change in foul-
of filtration volume. After this period, relatively large changes in ing may have been associated with significant biofilm development
absorbance values for all wavelengths suggest that a change in the and the relatively long lag period that occurred before the drastic
nature of fouling occurred and that it was no longer proportional change may have been a result of the robust pretreatment using
to the volumes filtered. low-pressure membranes, since they limit the concentration of
Although one might expect that the absorbance would be some- bacteria in the feed to the RO membranes. This explanation is sup-
what proportional to the degree of fouling, the samples that had ported by Durham et al. [28] who comment on the benefits of using
334 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338

Fig. 6. SEM images of RO membranes given TMF (left) and MBR (right) pretreatment for different permeated volumes per unit area. Volume filtered is given in text boxes
for each image.
 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338 335

low-pressure membranes as pretreatment for RO. It is also sup- 80

Protein Density MBR pretreatment
ported by the SEM data presented in the following section. Polysaccharide Density MBR pretreatment
Protein Density TMF pretreatment
70
Polysaccharide Density TMF pretreatment

3.3. SEM results 60

Scanning electron micrographs were generated for samples 50

Intensity
from the RO membranes to give a visual perspective of the sur-
faces. This allowed the SEM images of membranes with different 40

volumes of filtration to be observed. The intention was to pro-

30
duce the equivalent of a series of time lapse photographic images.
Fig. 6 shows the SEM images for a virgin RO membrane as well as 20
those with MBR and TMF pretreatment for a series of different fil-
tration volumes. Foulant can be seen accumulated on the surface 10
of the membranes and this accumulation increased as the filtra-
tion volume increased. In the SEM images shown in Fig. 6, it is 0
3000 4000 5000 6000 7000 8000 9000
clear that the entire surface is eventually coated with a layer of 2
Volume permeated per unit area (L/m )
foulants. When comparing the TMF and MBR pretreatments, there
seems to be more accumulation of material after 6800 L/m2 on the Fig. 7. CLSM results. Deposition intensity of protein (squares) and polysaccharide
membrane with TMF pretreatment. In some of the images, micro- (triangles) on membrane surfaces as a function of volume permeated. Results are
bial cells can be identified, with a matrix of EPS surrounding them. shown for pretreatment with CAS-TMF (open shapes) and MBR (solid shapes). Val-
Similar to the observations for the FTIR data, a slow accumula- ues were derived from 3 measurements from different samples and the standard
deviation was less than 5%.
tion of foulant material is observed at first, followed by a sudden
increase. The rate of build-up of foulant increases quickly between
2850 and 5175 L/m2 , whereas before 2850 L/m2 there is a very slow polysaccharides is shown as a red colour. Proteins clearly dominate
rate of increase in deposition on the membrane surface. Again, this the material deposited for the first sample, yet polysaccharides are
could indicate a transition from fouling governed by deposition of more pronounced in the final sample.
organic macromolecules, to fouling governed by the development The initial higher protein deposition may be due to early absorp-
of a biofilm. However, it is important to consider that the foul- tion which has been observed in other studies. Mo et al. [29]
ing rate for the membranes with less water filtered (between 750 observed that BSA absorption occurred for all surfaces in their
and 2850 L/m2 ) was higher than the fouling rate for the samples testing, regardless of pH and ionic strength. Although researchers
with more water filtered (between 2850 and 6800 L/m2 ), for both [10] have suggested that polysaccharides cause greater fouling
TMF and MBR pretreatments. Despite the greater accumulation of than proteins in controlled tests, it has been found that proteins
foulant material that was observed between 2850 and 6800 L/m2 , and polysaccharides as co-foulants cause more fouling than either
there was less of a loss of permeability. As with the FTIR data, the individual species due to a synergistic effect [14]. The deposition
amount of material measured on the membrane surface did not of proteins during the initial stage of membrane fouling has also
correspond to the degree of fouling when considering the overall been reported by other researchers [30]. Arabi and Nakhla [9] have
trends of the data sets for both pretreatments. There was, however, related higher protein to polysaccharide ratio in MBR studies to
a relationship between the accumulation of foulant material on the higher fouling propensity. Masse et al. [31] suggested that protein
membrane surface and fouling rate, when comparing the TMF and to polysaccharide ratios of bound EPS can be used as an indication
MBR pretreatments. The TMF pretreatment led to a higher fouling of the stickiness of sludge. It is interesting that the initial shorter
rate as a function of filtration volume (as shown in Fig. 3) and the filtration runs showed the highest amount of proteins and also the
TMF pretreatment also resulted in a greater observed accumula- highest loss of permeability.
tion of material on the membrane surface, compared with the MBR Similar to the changes observed for the FTIR data and the SEM
pretreatment (as shown in Fig. 6). images, in the CLSM results, there is a marked increase in deposi-
tion of polysaccharides between the first samples (4000 L/m2 ) and
3.4. CLSM results the second samples (6500 L/m2 ). The accumulation of polysaccha-
rides is expected as biofilm development continues and bacteria
For the CLSM testing, three pairs of cross-flow cells were oper- excrete polysaccharides in the production of biofilms. It is possi-
ated for 14, 23 and 30 days producing 4000, 6500 and 8700 L/m2 ble that the biofilms for both membranes reached the final stage of
of permeate, respectively. Upon completion of the run, membrane biofilm development: maturation and detachment, since the third
samples were analysed for protein and polysaccharide deposition samples had a similar amount of both polysaccharides and proteins
intensities using CLSM as described in Section 2.3. The results are compared with the second samples. It is also noted that for all three
shown in Fig. 7. sampling dates and for both polysaccharides and proteins, the RO
For both pretreatment methods the initial polysaccharide depo- membrane sample with TMF pretreatment had more deposition
sition levels are low compared to the protein deposition which than the sample with MBR pretreatment, which is consistent with
dominated the surface (protein to polysaccharide ratios were 4.1 the other results.
and 5.8 in TMF and MBR pretreatment samples, respectively). There
is a large increase in the deposition of polysaccharides as filtration 3.5. Microbial cell deposition
volume increases (protein to polysaccharide ratios were 0.5 and
0.4 in TMF and MBR pretreatment samples, respectively). It is also The deposition of microbial cells on the membrane surface was
noted that for all three sampling dates and for both polysaccharides also determined for the cross-flow cell testing. Fig. 9 shows the
and proteins, the RO membrane sample with TMF pretreatment had microbial cell deposition patterns for both TMF-RO and MBR-RO
more deposition than the sample with MBR pretreatment, which is samples. The accumulation of microbial cells follows a somewhat
consistent with the other results. Fig. 8 shows the images that cor- similar pattern to the results for polysaccharides. As with the pro-
respond to the CLSM analyses for these six membrane samples. The tein and polysaccharide data, all values for the RO membrane
presence of proteins is shown as a green colour and the presence of with TMF pretreatment showed higher microbial cells than the RO
336 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338

Fig. 8. Proteins (green) and polysaccharides (red) on the membrane surface for (a) 4000, (b) 6400 and (c) 8600 L/m2 shown for MBR pretreatment (above) and TMF
pretreatment (below). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of the article.)

membrane with MBR pretreatment. In addition, the microbial cell observation may be explained by the formation of an initial foulant
deposition for both membranes seemed to level off between 6500 layer that is different than the subsequent foulant accumulation
and 8700 L/m2 . Once again, this could correspond to a detachment as described by Tansel et al. [22]. Herzberg et al. [21] also identi-
stage of biofilm development that has been referred to in biofilm fied this first stage of fouling for membrane filtration of polished
studies [16]. secondary effluent. The degree of fouling for this period was dispro-
portionate to the accumulation of organic matter observed in the
3.6. Observation of high initial fouling rate FTIR data in Fig. 4, where the absorbance values for the different
wavelengths were relatively low after only 750 L/m2 of permeate.
It was observed that the vast proportion of permeability loss The magnitude of absorption values for both pretreatments exhib-
occurred within the first 750 L/m2 of filtration. As discussed, this ited relatively slow but steady increases for the samples analysed
as filtration volumes increased from 750 L/m2 up to approximately
5000 L/m2 as shown in Fig. 5. This high initial fouling rate, relative
100 to the quantity of organics deposition, could be associated with the
90 initial deposition of proteins according to the CLSM results in Fig. 7;
MBR pretreatment
TMF pretreatment
although a CLSM analysis of the membranes after 750 L/m2 would
80
be needed to show this. Mo et al. [29] found that BSA (a protein)
70 was absorbed onto surfaces despite electrostatic repulsion forces.
Research by Tsuneda et al. [32] identified that electrostatic repul-
Intensity

60
sion prevents microbial cell adhesion onto solid surfaces. This initial
50 layer of proteins may have provided surface-altering conditions to
40 facilitate microbial attachment that was observed between 4000
and 6500 L/m2 in Fig. 9.
30

20

10
3.7. Transition to biologically controlled fouling

0 A change in the foulant layer was detected in all of the

3000 4000 5000 6000 7000 8000 9000
2 surface analyses after approximately 5000 L/m2 . This may be
Volume permeated per unit area (L/m )
attributed to a transition to biologically governed fouling. For
Fig. 9. Microbial cell deposition intensity as a function of volume permeated for the CLSM results, proteins dominated the surface for the sam-
CAS-TMF (open diamonds) and MBR (solid diamonds). ples with less than 5000 L/m2 . However, for the samples with
 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338 337

more than 5000 L/m2 , polysaccharide deposition was found to be References

much higher than that of the proteins. Similarly, the microbial
cell densities became higher at this point, suggesting that biofilm [1] P. Cote, S. Siverns, S. Monti, Comparison of membrane-based solutions for water
reclamation and desalination, Desalination 182 (2005) 251–257.
formation was underway. The FTIR results also showed a sub- [2] T. Wintgens, T. Melin, A. Schafer, S. Khan, M. Muston, D. Bixio, The role of mem-
stantial increase in organics deposition levels for samples with brane processes in municipal wastewater reclamation and reuse, Desalination
more than 5000 L/m2 of filtration, even though all FTIR samples 178 (2005) 1–11.
[3] S. Adham, P. Gagliardo, L. Boulos, J. Oppenheimer, R. Trussell, Feasibility of the
between 750 and 5000 L/m2 showed only marginal increases in membrane bioreactor process for water reclamation, Water Sci. Technol. 43
response. A similar trend was also found within the SEM results, (2001) 203–209.
where much of the virgin membrane surface could be identified [4] A.A. Burbano, S.S. Adham, W.R. Pearce, The state of full-scale RO/NF
desalination—results from a worldwide survey, J. Am. Water Works Assoc. 99
for samples with less than 5000 L/m2 of filtration. The samples (2007) 116–127.
with 5175 L/m2 of permeation showed a large increase in cover- [5] S.B.S. Ghayeni, P.J. Beatson, R.P. Schneider, A.G. Fane, Water reclamation from
age and after 6800 L/m2 bacteria and biofilm material was clearly municipal wastewater using combined microfiltration-reverse osmosis (ME-
RO): preliminary performance data and microbiological aspects of system
observed. operation, Desalination 116 (1998) 65–80.
[6] J.S. Vrouwenvelder, J.A.M. van Paassen, L.P. Wessels, A.F. van Dam, S.M. Bakker,
The membrane fouling simulator: a practical tool for fouling prediction and
3.8. Higher observed fouling with TMF pretreatment control, J. Membr. Sci. 281 (2006) 316–324.
[7] H. Ivnitsky, I. Katz, D. Minz, E. Shimoni, Y. Chen, J. Tarchitzky, Characterization
of membrane biofouling in nanofiltration processes of wastewater treatment,
The fouling rate results in Fig. 3 indicate that the difference Desalination 185 (2005) 255–268.
in fouling rate between the RO membranes with TMF and MBR [8] C. Jarusutthirak, G. Amy, Role of soluble microbial products (SMP) in membrane
pretreatment were not clear for the shorter runs, but that they fouling and flux decline, Environ. Sci. Technol. 40 (2006) 969–974.
[9] S. Arabi, G. Nakhla, Impact of protein/carbohydrate ratio in the feed wastewater
became more evident with increased filtration time. This is sup-
on the membrane fouling in membrane bioreactors, J. Membr. Sci. 324 (2008)
ported by the analysis of accumulation of foulant material shown 142–150.
in the FTIR data in Fig. 4 and the SEM data in Fig. 6. The RO [10] Q. Li, Z. Xu, I. Pinnau, Fouling of reverse osmosis membranes by biopolymers
feed water with TMF pretreatment had a significantly higher in wastewater secondary effluent: Role of membrane surface properties and
initial permeate flux, J. Membr. Sci. 290 (2007) 173–181.
organics concentration compared with the feed water with MBR [11] M. Herzberg, S. Kang, M. Elimelech, Role of extracellular polymeric substances
pretreatment. This was demonstrated in a separate pilot study (EPS) in biofouling of reverse osmosis membranes, Environ. Sci. Technol. 43
where an extensive water quality comparison was conducted (2009) 4393–4398.
[12] S. Lee, W.S. Ang, M. Elimelech, Fouling of reverse osmosis membranes by
[33]. The higher concentration of organic material in the TMF- hydrophilic organic matter: implications for water reuse, Desalination 187
RO feed water may have led to an increase in the amount of (2006) 313–321.
fouling related to organics deposition and/or biofilm formation. [13] S. Lee, M. Elimelech, Relating organic fouling of reverse osmosis membranes
to intermolecular adhesion forces, Environ. Sci. Technol. 40 (2006) 980–
The higher organics concentration may have provided more food 987.
for attached microbial cells and led to a faster biofilm devel- [14] W.S. Ang, M. Elimelech, Protein (BSA) fouling of reverse osmosis membranes:
opment that was observed through the sudden increase in FTIR implications for wastewater reclamation, J. Membr. Sci. 296 (2007) 83–92.
[15] H.-C. Flemming, G. Schaule, T. Griebe, J. Schmitt, A. Tamachkiarowa,
absorbance values that were much higher than those for the MBR Biofouling—the Achilles heel of membrane processes, Desalination 113 (1997)
pretreatment. This could also account for the more highly devel- 215–225.
oped biofilm observed in the SEM images of the RO membrane [16] M. Herzberg, M. Elimelech, Physiology and genetic traits of reverse osmosis
membrane biofilms: a case study with Pseudomonas aeruginosa, ISME J. 2 (2008)
samples with TMF pretreatment after 6800 L/m2 of permeation.
180–194.
If this were the case, the difference in fouling between the two [17] M. Herzberg, M. Elimelech, Biofouling of reverse osmosis membranes: role of
pretreatments would have become larger if the study had been biofilm-enhanced osmotic pressure, J. Membr. Sci. 295 (2007) 11–20.
extended. [18] D.B. Mosqueda, R.M. Narbaitz, T. Matsuura, Membrane fouling test: apparatus
evaluation, J. Environ. Eng. 130 (2004) 90–99.
[19] H.J. Lin, K. Xie, B. Mahendran, D.M. Bagely, K.T. Leung, S.N. Liss, B.Q. Liao, Sludge
properties and their effects on membrane fouling in submerged anaerobic
4. Conclusions membrane bioreactors (SAnMBRs), Water Res. 43 (2009) 3827–3837.
[20] W.S. Ang, M. Elimelech, Fatty acid fouling of reverse osmosis membranes:
implications for wastewater reclamation, Water Res. 42 (2008) 4393–4403.
The analyses of RO membrane surfaces with a range of fil-
[21] M. Herzberg, D. Berry, L. Raskin, Impact of microfiltration treatment of sec-
tration volumes and with feed from two different pretreatment ondary wastewater effluent on biofouling of reverse osmosis membranes,
options provided a number of interesting observations. One of the Water Res. 44 (2010) 167–176.
most consistent observations in this series of experiments was [22] B. Tansel, J. Sager, J. Garland, S. Xu, L. Levine, P. Bisbee, Deposition of extracellu-
lar polymeric substances (EPS) and microtopographical changes on membrane
that foulant layers seemed to increase dramatically for samples surfaces during intermittent filtration conditions, J. Membr. Sci. 285 (2006)
with more than 5000 L/m2 of filtration. This included the amount 225–231.
of deposition of organic foulants found using FTIR as well as the [23] J.A. Brant, A.E. Childress, Assessing short-range membrane-colloid interactions
using surface energetics, J. Membr. Sci. 203 (2002) 257–273.
deposition density of polysaccharides and microbial cells as mea- [24] C. Laabs, G. Amy, M. Jekel, H. Buisson, Fouling of low-pressure (MF and UF)
sured by CLSM. The CLSM results suggest that proteins made up a membranes by wastewater effluent organic matter (EfOM): characterization
higher proportion of fouling layers initially (in the first two weeks) of EfOM foulants in relation to membrane properties, Water Sci. Technol. 3
(2003) 229–235.
after which the polysaccharide deposition densities increased dra- [25] C. Jarustthirak, G. Amy, J.P. Croue, Fouling characteristics of wastewater effluent
matically and became much higher than the protein deposition organic matter (EfOM) isolates on NF and UF membranes, Desalination 145
densities for both pretreatment methods. The RO with TMF pre- (2002) 247–255.
[26] M. Rabiller-Baudry, L. Begoin, D. Delaunay, L. Paugam, B. Chaufer, A dual
treatment had higher organics deposition and more rapid biofilm
approach of membrane cleaning based on physico-chemistry and hydrody-
development compared with the RO membrane with MBR pretreat- namics; application to PES membrane of dairy industry, Chem. Eng. Process. 47
ment. (2008) 267–275.
[27] A. Subramani, E.M.V. Hoek, Direct observation of initial microbial deposition
onto reverse osmosis and nanofiltration membranes, J. Membr. Sci. 319 (2008)
111–125.
Acknowledgement [28] B. Durham, M.M. Bourbigot, T. Pankratz, Membranes as pretreatment to
desalination in wastewater reuse: operating experience in the municipal and
The funding for this work was provided by the National Sci- industrial sectors, Desalination 138 (2001) 83–90.
[29] H. Mo, K.G. Tay, H.Y. Ng, Fouling of reverse osmosis membranes by protein
ence and Engineering Research Council of Canada (NSERC) and the
(BSA): effects of pH, calcium, magnesium, ionic strength and temperature, J.
Canadian Water Network (CWN). Membr. Sci. 315 (2008) 28–35.
338 F.C. Kent et al. / Journal of Membrane Science 382 (2011) 328–338

[30] H. Zhou, F. Fan, L. LaFleur, J. Peeters, H. Liu, Causes and measurements of mem- [32] S. Tsuneda, H. Aikawa, H. Hayashi, A. Yuasa, A. Hirata, Extracellular polymeric
brane fouling in MBR systems for wastewater treatment, in: Proceedings of the substances responsible for bacterial adhesion onto solid surfaces, FEMS Micro-
IWA North American Membrane Conference, Amherst, USA, 2008. biol. Lett. 223 (2003) 287–292.
[31] A. Masse, M. Sperandio, C. Cabassud, Comparison of sludge characteristics and [33] F.C. Kent, J. Citulski, K. Farahbakhsh, Water reclamation using membranes:
performance of a submerged membrane bioreactor and an activated sludge permeate water quality comparison of MBR and tertiary membrane filtration,
process at high solids retention time, Water Res. 40 (2006) 2405–2415. Desalination 274 (2011) 237–245.