Journal of Membrane Science 272 (2006) 198–210

Chemical and physical aspects of cleaning of

organic-fouled reverse osmosis membranes
Wui Seng Ang, Sangyoup Lee, Menachem Elimelech ∗
Environmental Engineering Program, Department of Chemical Engineering, P.O. Box 208286, Yale University, New Haven, CT 06520-8286, USA

Received 19 May 2005; received in revised form 13 July 2005; accepted 20 July 2005
Available online 6 September 2005

Abstract

The role of chemical and physical interactions in cleaning of organic-fouled reverse osmosis (RO) membranes is systematically investigated.
Fouling and cleaning experiments were performed with organic foulants (alginate and natural organic matter) that simulate effluent organic
matter (EfOM) and selected cleaning agents using a laboratory-scale crossflow test unit. Following accelerated organic fouling runs with
divalent calcium ions, cleaning experiments were performed at various chemical and physical conditions. The results showed that a metal
chelating agent (EDTA) and an anionic surfactant (SDS) were able to clean the fouled RO membrane effectively by optimizing chemical
(dose and pH) and physical (time, crossflow velocity, and temperature) conditions during cleaning. The permeate flux was, however, poorly
recovered when cleaning was performed with NaOH (pH 11). Interfacial force measurements (by AFM) further confirmed the cleaning results,
demonstrating the influence of cleaning agents on reducing the foulant–foulant adhesion force. The results showed that the adhesion force
caused by calcium-induced foulant–foulant interaction was reduced significantly with EDTA and SDS cleaning, while substantial adhesion
force still remained following NaOH cleaning. It is proposed that an efficient cleaning agent and favorable cleaning conditions could be
selected by considering two important mechanisms: chemical reaction between cleaning agents and foulants in the fouling layer, and mass
transfer of chemical agents (from the bulk phase to the fouling layer) and foulants (from the fouling layer to the bulk phase).
© 2005 Elsevier B.V. All rights reserved.

Keywords: Membrane cleaning; Chemical cleaning; Organic fouling; RO membrane; Intermolecular adhesion force; Foulant–foulant interaction

1. Introduction the RO membrane processes as feed water, the presence of

EfOM contributes to organic fouling [3]. EfOM represents a
Reverse osmosis (RO) membranes are widely used in range of organic compounds, such as polysaccharides, pro-
drinking water, wastewater, and industrial applications. The teins, aminosugars, nucleic acids, humic and fulvic acids,
use of RO membranes in advanced wastewater reclamation organic acids, and cell components [3]. These complex and
using secondary treated wastewater effluent to produce water heterogeneous compounds can generally be classified into
for indirect potable use has also increased over the past few two groups according to their origin: (i) soluble microbial
years [1,2]. However, a major impediment in the application products (SMPs) derived during substrate metabolism in
of RO membrane technology for desalination and wastewater the biological wastewater treatment process and (ii) natural
reclamation is membrane fouling. organic matter (NOM) originating from the drinking water
In advanced water reclamation, secondary effluent from source [3,4]. It has been reported that the majority of EfOM
wastewater treatment plants contains dissolved organic mat- in secondary wastewater effluent is made up of SMPs [5,6].
ter, commonly known as effluent organic matter (EfOM). Polysaccharides, one of the major constituents of SMPs [7],
When the secondary wastewater effluent is introduced to have been found to play an important role in the fouling of
nanofiltration (NF) and ultrafiltration (UF) membranes [8].
∗ Corresponding author. Tel.: +1 203 432 2789; fax: +1 203 432 2881. Developing strategies for fouling control has always been
E-mail address: menachem.elimelech@yale.edu (M. Elimelech). a major challenge in membrane research. However, despite

0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2005.07.035
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 199

the many preventive strategies, fouling is inevitable. The profiles from AFM force measurements were used to explain
long-term solution would be to remove the foulant deposited the cleaning mechanisms involved.
on membrane surfaces via chemical cleaning [9–14]. Mem-
brane cleaning is performed when there is a significant drop
in permeate flux or salt rejection, or when there is a need to 2. Experimental
increase the transmembrane pressure significantly to main-
tain the desired water flux [15]. There are five categories of 2.1. Organic foulants
cleaning agents — alkaline solutions, acids, metal chelating
agents, surfactants, and enzymes [9,16]. Commercial clean- The organic foulants chosen as representative hydrophilic
ing products are often mixtures of these compounds, but in and hydrophobic EfOM were sodium alginate (Sigma-
most cases the actual composition is unknown. Therefore, Aldrich, St. Louis, MO) and Suwannee River natural organic
many past studies on chemical cleaning are not able to provide matter (SRNOM) (International Humic Substances Society,
useful information to illuminate the mechanisms of chemical St. Paul, MN), respectively. Sodium alginate was extracted
cleaning. from brown seaweed. Based on the manufacturer, the molec-
In this study, alkaline solution, metal chelating agent, ular weight of the sodium alginate ranges from 12 to 80 kDa.
and surfactant are used to clean the organic-fouled mem- Other characteristics of SRNOM, including molecular weight
branes. Alkaline solutions clean organic-fouled membranes and mass fraction of hydrophobic NOM, can be found else-
by hydrolysis and solubilization. Alkaline solutions increase where [26,27]. Both organic foulants were received in a
the solution pH, and therefore increase the negative charge powder form, and stock solutions (2 g/l) were prepared by
and solubility of the organic foulant. For example, when car- dissolving each of the foulants in deionized (DI) water. Mix-
boxylic functional groups of the organic foulant are deproto- ing of the stock solutions was performed for over 24 h to
nated at alkaline conditions, solubility increases nearly three ensure complete dissolution of the foulants, followed by fil-
orders of magnitude [17]. In the presence of divalent cations, tration with a 0.45 ␮m filter (Durapore, Millipore, Billerica,
such as Ca2+ , alginate, which is selected in this study to rep- MA). The filtered stock solutions were stored in sterilized
resent the hydrophilic (polysaccharide) fraction of EfOM, glass bottles at 4 ◦ C.
forms complexes with calcium ions, resulting in a highly
compacted gel network of fouling layer [18,19]. Metal chelat- 2.2. Chemical cleaning agents
ing agents, such as EDTA, remove divalent cations from the
complexed organic molecules and improve the cleaning of The chemical cleaning agents used were NaOH (pH 11.0)
the fouled membrane [20]. Surfactants are compounds that as an alkaline solution, certified grade disodium ethylenedi-
have both hydrophilic and hydrophobic groups, and are semi- aminetetraacetate (Na2 -EDTA) as a metal chelating agent,
soluble in both organic and aqueous solvents. Surfactants can and certified grade sodium dodecyl sulfate (SDS) as an
solubilize macromolecules by forming micelles around them anionic surfactant. These cleaning agents are the most
[21], and help to remove the foulant from the membrane sur- common compounds in commercial cleaning products for
face. organic-fouled membranes [9,16]. The NaOH, EDTA, and
Atomic force microscopy (AFM) has recently been used SDS were purchased from Fisher Scientific (Pittsburgh, PA)
to quantify intermolecular forces in membrane research and used with no further purification. The stock chemical
[22–24], as it is able to measure interaction forces in liquids solutions were prepared freshly by dissolving each chemical
at pico- or nano-Newton level. Li and Elimelech [25] have in deionized (DI) water. Cleaning with DI water served as a
recently demonstrated that foulant–foulant interactions could baseline. The pH of the EDTA and SDS cleaning solutions
be determined by performing force measurements using a was adjusted with 1.0 M NaOH as needed.
carboxylate-modified latex colloid probe in an AFM fluid
cell. The technique has been used to quantify the fouling 2.3. RO membrane
behavior of a nanofiltration membrane by humic acid as
well as the cleaning efficiency by several chemical clean- The relatively well characterized thin-film composite
ing agents. The study showed a good correlation between LFC-1 (Hydranautics, Oceanside, CA) was used as a model
the measured adhesion force and the fouling and cleaning RO membrane. The average hydraulic resistance was deter-
behaviors. mined to be 9.16 (±0.11) × 1013 m−1 . The observed salt
The objective of this paper is to systematically study rejection was 98.7–99.3% — determined with a 10 mM
the physical and chemical aspects of cleaning of organic- (584 mg/l) NaCl feed solution at an applied pressure of
fouled RO membranes, with emphasis on cleaning of an 300 psi (2068.5 kPa) and a crossflow velocity of 8.1 cm/s.
RO membrane fouled by acidic polysaccharides. AFM was Membrane samples were received as dry large sheets, and
used to study the interaction forces between the foulant in were cut and stored in DI water at 4 ◦ C. The membrane has
the bulk solution and the foulant accumulated on the mem- been reported to be negatively charged at solution chemistries
brane surfaces. The results from the bench-scale crossflow typical to natural and waste waters, with an isoelectric point
fouling/cleaning experiments and the relevant adhesion force at about pH 4 [28].
200 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

2.4. Crossflow membrane test unit reservoir was disposed and cleaning of the fouled membrane
was performed by adding the chemical cleaning solution to
Fouling and subsequent cleaning experiments were per- the feed reservoir. At the end of the cleaning stage, the chem-
formed with a laboratory-scale crossflow test unit. The mem- ical cleaning solution in the reservoir was emptied, and the
brane test unit consists of a membrane cell, pump, feed reser- reservoir and the membrane cell were rinsed with DI water
voir, temperature control system, and data acquisition system. to flush out the residual chemical cleaning solution. Finally,
The membrane cell was a rectangular plate-and-frame unit, the cleaned RO membrane was subjected to the second base-
containing a flat membrane sheet placed in a rectangular line performance with DI water for 2 h to determine the pure
channel of dimensions 7.7 cm long, 2.6 cm wide, and 0.3 cm water flux again. The operating conditions (i.e., initial flux,
high. Both permeate and retentate were recirculated back crossflow velocity, and temperature) at this stage were iden-
to the feed reservoir. Permeate flux was registered continu- tical to those applied during the initial baseline performance,
ously by a digital flow meter (Optiflow 1000, Humonics, CA) so as to determine the cleaning efficiency by comparing the
interfaced with a computer. A floating disc rotameter (King pure water fluxes determined before fouling and after clean-
Instrument, Fresno, CA) was used to monitor the retentate ing. To confirm the reproducibility of determined cleaning
flow rate. The crossflow velocity and the operating pressure efficiency, selected fouling/cleaning runs were duplicated.
were adjusted using a bypass valve (Swagelok, Solon, OH) Results showed that fouling rate and cleaning efficiency
in conjunction with a back-pressure regulator (US Paraplate, obtained from the duplicate runs were always within less than
Auburn, CA). Temperature was controlled by a recirculating 5% difference. Throughout all the fouling/cleaning stages,
chiller/heater (Model 633, Polysciences) with a stainless- the feed water in the reservoir located on top of a magnetic
steel coil submerged in the feed water reservoir. stirrer was mixed vigorously to ensure complete mixing of
the feed water and cleaning solution.
2.5. Fouling and cleaning experiments Cleaning efficiency varies with respect to the conditions
applied during cleaning, namely, type of cleaning agent,
The experimental protocol developed for fouling/cleaning cleaning solution pH, cleaning agent dose, cleaning time,
is shown in Fig. 1. The membrane was first compacted with crossflow velocity during cleaning, and cleaning solution
DI water for 12 h until the permeate flux became constant, fol- temperature. Furthermore, cleaning efficiency, even at fixed
lowed by the initial baseline performance for 2 h. Following cleaning conditions, is also influenced by the conditions
the initial baseline performance, the membrane was stabi- applied during fouling. For a systematic investigation on
lized and equilibrated with a foulant-free electrolyte solution the role of chemical and physical interactions in cleaning of
for 4 h. After attaining a stable flux, the initial flux and the fouled membranes, it is necessary to isolate cleaning condi-
crossflow velocity were adjusted to the values used for the tions from the influence of fouling conditions, thus focusing
subsequent fouling runs. The chemistry of feed solution and solely on the influence of cleaning conditions on cleaning effi-
operating conditions adjusted in this stage were similar to ciency. Therefore, all the fouling runs subjected to cleaning
those used for the ensuing fouling runs. Fouling was then experiments were performed at the same fouling conditions
initiated by adding organic foulants (i.e., alginate, SRNOM, in terms of feed solution chemistry (i.e., pH 6.0, 0.5 mM
or both alginate and SRNOM) to the feed water to achieve calcium, and 10 mM total ionic strength adjusted by varying
the desired organic concentration. Fouling runs were carried NaCl concentration) and operating conditions (i.e., initial flux
out for 17 h. At the end of fouling, the solution in the feed of 20 ␮m/s, crossflow velocity of 8.56 cm/s, and temperature
of 20 ◦ C); these conditions yielded similar flux decline pro-
files. This implies that each cleaning is applied to an identical
organic fouling layer since the same organic fouling layer
produces identical resistance to permeate flow and, hence,
the same flux profiles. A few fouling runs were performed
with feed waters containing different organic foulants (i.e.,
SRNOM or mixture of alginate and SRNOM); however, the
fouling and cleaning conditions were the same as above, thus
allowing to investigate the influence of foulant composition
on cleaning efficiency.
During the fouling experiments, calcium was used for two
reasons. First, calcium was used to accelerate fouling of the
charged organic foulant, leading to a substantial flux decline,
which was needed to differentiate the cleaning efficiency
determined at various conditions. Based on our previous find-
ings [20,25], it was shown that calcium ions specifically bind
with carboxylic functional groups of organic foulants (i.e.,
Fig. 1. Experimental protocol for the fouling/cleaning runs. NOM and acidic polysaccharides) and form bridges between
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 201

adjacent foulants. This results in the formation of a cross- at each location. Because the focus of this study was on
linked fouling layer at the membrane surface, thus producing the foulant–foulant interaction (adhesion), only the raw data
substantial resistance to permeate flow. In addition, calcium- obtained from the retracting force curves were processed and
induced organic fouling is one of the major causes for severe converted to obtain the force versus surface-to-surface sepa-
flux decline and known to be most recalcitrant since natural ration curves. The force curves presented were the averages
and waste waters contain substantial amounts of calcium. of all the representative force curves obtained at the different
locations. The detailed procedures of using AFM in con-
2.6. AFM adhesion force measurements junction with a colloid probe to determine the intermolecular
adhesion forces in membrane fouling/cleaning can be found
Atomic force microscopy (AFM) was used to measure the in our recent work [25].
interfacial force between the foulant in the bulk solution and
the foulant in the fouling layer on the membrane. The force
measurements were performed with a colloid probe, modified 3. Results and discussion
from a commercialized SiN AFM probe (Veeco Metrology
Group, Santa Barbara, CA). A carboxylate-modified latex 3.1. Chemical aspects of cleaning
(CML) particle (Interfacial Dynamics Corp., Portland, OR)
was used as a surrogate for the organic foulants, because The chemical aspects of cleaning – cleaning solution type,
organic foulants (alginate and SRNOM) carry predominantly cleaning solution pH, and cleaning agent dose – have been
carboxylic functional groups. To make a colloid probe, a systematically investigated to find out the favorable cleaning
CML particle of diameter 4.0 ␮m was attached by Norland conditions in terms of chemical reaction between the cleaning
Optical adhesive (Norland Products Inc., Cranbury, NJ) to a agent and foulants. In addition, the influence of fouling layer
tipless SiN cantilever. The colloid probe was cured under UV composition on cleaning efficiency is examined with feed
light for 20 min. waters containing both alginate and SRNOM with different
The AFM adhesion force measurements were performed mass ratios. Cleaning experiments presented in this section
in a fluid cell using a closed inlet/outlet loop. The solution were performed at fixed physical conditions so that only the
chemistries of the test solutions in the fluid cell were the chemical aspects of cleaning could be investigated.
same as those used in the bench-scale fouling/cleaning exper-
iments. The fluid cell was first rinsed with 6 ml of DI water 3.1.1. Effect of cleaning solution type
and 6 ml of the test solution, before injecting another 6 ml The cleaning efficiency with different cleaning solutions
of test solution. Once all the air bubbles had been flushed – NaOH (pH 11), SDS, and EDTA – is compared in Fig. 2.
out of the fluid cell, the injection would stop and the outlet DI water was also used to serve as a baseline for compari-
was closed. The membrane was equilibrated with the test son with the cleaning solutions. Because fouling runs prior
solution for 60 min before force measurements were per- to each cleaning experiment were performed under identical
formed. The force measurements were conducted at three conditions, the flux behavior of each run was almost identical,
to five different locations, and 15 measurements were taken except the pure water flux obtained after cleaning. Therefore,

Fig. 2. Influence of cleaning solution type on cleaning of organic-fouled RO membranes: (a) flux behavior during the fouling/cleaning experiments and (b)
variation in cleaning efficiency with respect to cleaning solution type. Conditions for fouling experiments: alginate concentration = 20 mg/l, calcium concen-
tration = 0.5 mM, total ionic strength = 10 mM (i.e., 0.5 mM Ca2+ plus 8.5 mM NaCl), pH 5.90 ± 0.05, initial flux = 20 ␮m/s, crossflow velocity = 8.56 cm/s,
and temperature = 20.0 ± 0.2 ◦ C. Conditions for cleaning experiments: time = 15 min, temperature = 20.0 ± 0.2 ◦ C, crossflow velocity = 42.8 cm/s. Cleaning
solutions were used without pH adjustment (ambient pH) except for the NaOH cleaning. Note that the last data points in Fig. 2a are for flux measured with DI
water after the corresponding cleaning.
202 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

the flux profiles for subsequent fouling/cleaning experiments increase in cleaning efficiency of an NOM-fouled NF mem-
will not be shown, unless different concentrations and types brane was observed when SDS concentration was higher than
of organic foulants were used, which would result in flux the CMC. The influence of SDS concentration on cleaning
decline profiles different from those shown in Fig. 2. efficiency will be discussed later.
Fig. 2a clearly shows that permeate flux during fouling In the case of EDTA cleaning, cleaning efficiency is quite
runs decreased dramatically since alginate fouling of the sensitive to solution pH. The chelating ability of EDTA to
RO membrane was accelerated in the presence of calcium remove complexed calcium ions from alginate–calcium com-
ions (0.5 mM). As described earlier, calcium ions bind with plexes through a ligand-exchange reaction increases at higher
carboxylic groups of alginate and bridge between adjacent pH where more carboxylic groups of EDTA are deprotonated.
alginate molecules, leading to the formation of a cross-linked This will be confirmed when the effect of cleaning solution
alginate fouling layer on the membrane surface. This algi- pH on cleaning efficiency is reported.
nate fouling layer can also be explained by the gel forma- For the above runs, visual inspection of the membranes
tion of alginate in presence of calcium ions as commonly after cleaning by all tested solutions revealed that there was
described by the ‘egg-box’ model [18,19]. Based on this still substantial alginate gel layer on the membrane surface.
model, divalent cations (such as calcium ions) induce gela- This observation is not surprising since the cleaning effi-
tion of alginate through binding with and bridging between ciency was less than 25% for all cases shown in Fig. 2b.
alginate molecules in a highly cooperative manner, leading to
the ‘egg-box’ shaped gel network. In fact, visual inspection 3.1.2. Effect of cleaning solution pH
of the membrane surface at the end of the fouling runs con- The influence of solution pH on the cleaning efficiency of
firmed the presence of a thick alginate gel layer. Therefore, SDS and EDTA is illustrated in Fig. 3. It is shown that clean-
the efficiency of a cleaning agent depends on the ability of ing efficiency increased noticeably (i.e., from 25 to 44%) with
the cleaning agent to break down the alginate gel network in increasing pH from 4.9 to 11.0 for EDTA, while there was
the fouling layer through chemical reaction of the cleaning only a slight increase for SDS at higher pH. At pH 11.0, all
agent with the fouling layer. the carboxylic functional groups of EDTA are deprotonated
Fig. 2b presents the efficiency of each cleaning solution, (pKa values are 1.99, 2.67, 6.16, and 10.26 [30,31]). The
determined by comparing the pure water fluxes before fouling increase in the chelating ability of EDTA at pH 11.0 resulted
and after cleaning (the initial and final water fluxes are shown in a more effective ligand-exchange reaction between EDTA
in Fig. 2a). The cleaning efficiencies with NaOH (pH 11) and and alginate–calcium complexes within the alginate gel layer.
2 mM SDS were 14 and 15%, respectively. This points out Consequently, the alginate gel layer was broken down rela-
that NaOH and SDS cleaning performed at these experimen- tively more easily compared to lower pH and, thus, resulted in
tal conditions was ineffective since the cleaning efficiency a higher cleaning efficiency. On the other hand, solution pH
with DI water was already 12%. Cleaning with 0.5 mM EDTA showed very little effect on SDS cleaning, because the pKa of
was relatively more effective compared to NaOH and SDS, the sulfate functional group of SDS is 2.12 [30,31], implying
and the cleaning efficiency (25%) was two times higher that SDS is in its ionic form at both pH values examined. The
than that of DI water. This is attributed to the preferential
EDTA–calcium complexation compared to alginate–calcium
complexation.
In the case of NaOH cleaning, it is inferred that chemi-
cal reaction between NaOH and alginate–calcium complexes
in the alginate gel layer was limited and, thus, resulting in
a poor cleaning efficiency. There are two possible ways to
increase NaOH cleaning efficiency: an increase in NaOH
concentration (i.e., higher pH) and/or use of more favorable
physical conditions. The former, however, is not applicable in
real world applications since most polymeric RO membranes
tolerate a limited pH range (i.e., pH 3–12). Therefore, the lat-
ter possibility has been investigated, and the results will be
shown later in this paper.
The ineffective SDS cleaning is attributed to the low SDS
concentration applied during cleaning. The 2 mM concentra-
tion was much lower than the critical micelle concentration
(CMC) of SDS, reported as 8.36 mM in DI water [29]. At this
Fig. 3. Variation in SDS and EDTA cleaning efficiencies as a function of
low SDS concentration, the interaction of the adsorbed SDS
cleaning solution pH. Cleaning experiments were performed following foul-
molecules with alginate–calcium complexes was not strong ing runs carried out at the same conditions as those in Fig. 2a. Conditions
enough to break the intermolecular bridging formed with cal- for cleaning experiments were identical to those in Fig. 2b except for the
cium ions. Li and Elimelech [25] demonstrated that a marked cleaning solution pH as indicated in this figure.
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 203

results imply that cleaning solution pH is a governing factor

affecting chemical reaction between EDTA and deposited
foulants, while chemical reaction between SDS and foulants
is less influenced by cleaning solution pH.
It is noteworthy that the efficiency of the 0.5 mM EDTA
cleaning solution at pH 11 was relatively low (less than 45%)
compared to the results of other studies on EDTA clean-
ing of NOM-fouled NF membranes [20,25]. These studies
showed that EDTA treatment was effective (almost 100%
cleaning efficiency) in cleaning NOM-fouled NF membranes
in the presence of calcium ions when the molar EDTA con-
centration (cleaning solution) was the same as the calcium
concentration used during the fouling runs. Note that in our
study, both EDTA and calcium concentrations during the
cleaning and fouling experiments were 0.5 mM. The reason
for this discrepancy is attributed to the structural difference
in the fouling layers as well as the different extent of foulant Fig. 5. Variation in SDS cleaning efficiency with SDS concentration. SDS
cleaning experiments were performed at pH 11 following fouling runs carried
accumulation on the membrane surface for these two cases.
out at the same conditions as those in Fig. 2a. Other cleaning conditions were
The fouling layer of the other studies comprised hydropho- identical to those in Fig. 2b.
bic NOM with smaller molecular weight, whereas the fouling
layer in this study comprised hydrophilic alginate with larger
molecular weight. This argument will be discussed later in notable increase in cleaning efficiency when SDS concen-
this paper (Section 3.1.4). tration increased from 5 to 8 mM. We note that an 8 mM
SDS concentration is close to the CMC of SDS, reported as
8.36 mM in DI water [29]. Therefore, it is concluded that
3.1.3. Effect of cleaning chemical dose cleaning with SDS above the CMC is the critical factor for
Cleaning experiments were also performed with clean- efficient SDS cleaning, allowing adequate chemical reaction
ing solutions containing different doses of EDTA or SDS between SDS and the foulants to break down the alginate gel
(both at the pH 11), as presented in Figs. 4 and 5, respec- network. A similar mechanism of SDS cleaning of fouled NF
tively. The results clearly show that cleaning efficiency with membranes by humic acid in presence of calcium ions, has
EDTA and SDS cleaning increased with increasing cleaning been discussed in our recent work [25].
agent dose. For EDTA, cleaning efficiency increased propor-
tionally with EDTA concentration and reached near 100%
efficiency at 2.0 mM EDTA. Note that this molar EDTA con- 3.1.4. Effect of organic foulant composition
centration is four times higher than the calcium concentration As discussed in previous sections, the fouling layer struc-
employed during fouling. For SDS cleaning, there was a ture as well as the extent of foulant accumulation on the
membrane surface is expected to affect cleaning efficiency
for a given cleaning agent under the same cleaning condi-
tions. To investigate the influence of structural difference of
the organic fouling layer on cleaning efficiency, fouling runs
were performed with feed waters containing different mass
ratios of alginate to SRNOM. Several feed water foulant com-
positions were examined with the total foulant concentration
maintained at 20 mg/l: (i) 20 mg/l alginate and no SRNOM,
(ii) 14 mg/l alginate plus 6 mg/l SRNOM, (iii) 6 mg/l algi-
nate plus 14 mg/l SRNOM, and (iv) 20 mg/l SRNOM and
no alginate. It is expected that the difference in feed water
organic composition will result in the formation of organic
fouling layer, which is structurally different from one another,
thus resulting in different cleaning efficiencies with the same
cleaning solution. Following the fouling runs, each cleaning
experiment was performed with 0.5 mM EDTA (pH 11). The
flux profiles of these fouling runs and the cleaning efficien-
cies determined from the subsequent cleaning experiments
Fig. 4. Variation of EDTA cleaning efficiency with EDTA concentration.
EDTA cleaning experiments were performed at pH 11.0 following fouling are presented in Fig. 6.
runs carried out at the same conditions as those in Fig. 2a. Other cleaning The flux behaviors depicted in Fig. 6a varied, indicat-
conditions were identical to those in Fig. 2b. ing that the structure and/or amount of fouling layer were
204 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

Fig. 6. Influence of feed water organic composition on cleaning of organic-fouled RO membranes: (a) flux behavior during the fouling/cleaning experiments
and (b) variation in cleaning efficiency with respect to feed water organic composition. Fouling experiments were performed with feed waters containing:
alginate = 20 mg/l (i.e., alginate only); alginate = 14 mg/l and SRNOM = 6 mg/l (i.e., alginate:SRNOM = 7:3); alginate = 6 mg/l and SRNOM = 14 mg/l (i.e.,
alginate:SRNOM = 3:7); or SRNOM = 20 mg/l (i.e., SRNOM only). Other fouling conditions were identical to those in Fig. 2a. Cleaning experiments were
performed with 0.5 mM EDTA cleaning solutions at pH 11.0. Other cleaning conditions were identical to those in Fig. 2b. Note that the last data points in
Fig. 6a are for flux measured with DI water after the corresponding cleaning.

different for each foulant composition. A more severe flux that for SRNOM fouling (i.e., 20 mg/l SRNOM and absence
decline was observed with increasing alginate concentration, of alginate). This suggests that a small amount of alginate
and there was a significant difference in the flux profiles for (6 mg/l alginate and 14 mg/l SRNOM) was enough to form
alginate (i.e., absence of SRNOM) and SRNOM (i.e., absence alginate gel layer on the membrane surface, and the alginate
of alginate) fouling runs. This implies that the alginate foul- molecules more easily accumulated on the membrane surface
ing layer produced greater hydraulic resistance to permeate compared to SRNOM molecules.
flow than the SRNOM fouling layer. This is attributed to the Fig. 6b clearly shows that the cleaning efficiency with
formation of gel-type alginate fouling layer as described ear- 0.5 mM EDTA (pH 11) increased when the fouling runs were
lier based on the ‘egg-box’ model. The gelation of charged performed with feed waters containing a higher proportion
macromolecules in the presence of calcium ions is predomi- SRNOM to alginate. The SRNOM-fouled membrane (i.e.,
nant for hydrophilic organic macromolecules (such as acidic absence of alginate) achieved the highest cleaning efficiency
polysaccharides) compared to hydrophobic organic macro- (108%), while the alginate-fouled membrane (i.e., absence
molecules (such as humic acids) [18,19]. However, it should of SRNOM) had the lowest efficiency (44%). The combined
be noted that NOM fouling in presence of calcium ions also organic fouling with alginate to SRNOM ratios of 3:7 and
results in severe flux decline during nanofiltration and tight- 7:3 had cleaning efficiencies of 102 and 64%, respectively.
ultrafiltration [20,32,33]. The reason for the less pronounced These values were expected as they were within the range
flux decline with SRNOM observed in this study is that the of the cleaning efficiencies for the alginate- and SRNOM-
additional hydraulic resistance of the SRNOM fouling layer fouled membranes. The high cleaning efficiency of SRNOM-
to permeate flow is relatively much smaller than that of the fouled membrane was in agreement with our recent study
RO membrane used in this study. on EDTA cleaning of NOM-fouled NF membrane [20]. A
It is also noteworthy that, in the case of alginate fouling, cleaning efficiency of over 100% suggests that the recovered
the flux declined rapidly in the initial stage and, then, declined flux after EDTA cleaning was higher than the flux of the
gradually. On the other hand, for SRNOM fouling, the fouling original clean membrane, probably due to the presence of
rate increased as fouling progressed (i.e., flux declined more a small amount of EDTA and/or NOM on the membrane
rapidly with filtration time). This implies that the alginate surface, making the membrane more hydrophilic, and thus,
gel layer was denser and more compact than the SRNOM enhancing the partitioning and passage of water molecules
fouling layer, leading to a substantial hydraulic resistance [20].
upon gel formation. However, this alginate fouling layer The difference in the EDTA cleaning efficiency for the
was likely less compressible than the SRNOM fouling layer alginate- and SRNOM-fouled membranes can be attributed
that became more compact during filtration, resulting in a to one or more of the following: (i) the stronger bind-
higher fouling rate as fouling progressed. In addition, the flux ing and bridging of calcium to alginate than SRNOM, (ii)
decline behavior with the higher SRNOM proportion (i.e., the gel-forming nature of alginate which yields a denser
alginate:SRNOM = 3:7) was more similar to that for alginate and more compact fouling layer compared to SRNOM,
fouling (i.e., 20 mg/l alginate and absence of SRNOM) than and (iii) the larger mass of alginate accumulating on the
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 205

membrane surface compared to SRNOM. The first expla- 3.2. Physical aspects of cleaning
nation is not likely since a strong chelating agent, such
as EDTA, can easily remove the complexed calcium from In this section, we investigate the physical aspects of clean-
both alginate–calcium and SRNOM–calcium complexes via ing – crossflow velocity, cleaning time, and temperature – to
a ligand-exchange reaction. The second explanation is based determine the favorable conditions for effective mass transfer.
on the observation that, in the presence of divalent cations, During cleaning, mass transfer of foulants from the fouling
the alginate gel layer produced more resistance to permeate layer to the bulk solution takes place after the chemical reac-
flow than the SRNOM fouling layer. This behavior sug- tion between the cleaning agent and deposited foulants has
gests that the alginate fouling layer is much denser and weakened the structural integrity of the fouling layer. Cross-
more compact than the SRNOM fouling layer due to its gel- flow velocity in the membrane cell mainly controls the mass
like structure. Consequently, during EDTA cleaning, EDTA transfer, and cleaning time and temperature affect the overall
molecules diffused more easily into the looser SRNOM foul- mass transfer as the rate and extent of solute transport are
ing layer. Thus, a more rigorous ligand-exchange reaction functions of time and temperature. Cleaning experiments –
is likely to occur within the SRNOM fouling layer than following fouling runs carried out under identical conditions
within the alginate fouling layer and, consequently, this has (Section 2.5) – were performed at fixed chemical conditions,
more impact on breaking up the intermolecular bridging of thus focusing on the physical aspects of cleaning. However,
SRNOM induced by calcium ions than that of alginate. A it is important to note that cleaning time and temperature also
third possible explanation for the difference in cleaning of affect chemical reaction between cleaning agent and foulants
the alginate- and SRNOM-fouled membranes is the variation since contact time (between cleaning agent and fouling layer)
in the amount of accumulated foulant on the membrane sur- and solution temperature influence the extent and rate of reac-
face. At the given hydrodynamic conditions (i.e., initial flux tion. While investigating the influence of a certain physical
and crossflow velocity), the larger alginate macromolecules factor on the efficiency of a cleaning agent, all other physical
may accumulate more effectively on the RO membrane sur- factors are kept constant.
face compared to the smaller SRNOM molecules because
of the lower back-diffusion experienced by larger macro- 3.2.1. Effect of cleaning time
molecules. In the previous results (Fig. 3), an almost 100% The influence of cleaning time on the efficiency of each
cleaning efficiency was demonstrated for the alginate-fouled cleaning agent was investigated. The flux behaviors during
RO membrane when 2.0 mM EDTA was used. This sug- the fouling/cleaning runs and the resulting cleaning efficien-
gests that 0.5 mM EDTA was not enough to react with all cies with NaOH (pH 11) and 0.5 mM EDTA (pH 11) are
the alginate–calcium complexes in the gel layer (within the shown in Fig. 7. The influence of cleaning time on cleaning
15 min cleaning time), while this dose was sufficient to react efficiency is also compared for two different SDS concentra-
with all the SRNOM–calcium complexes in the fouling layer tions (2 and 10 mM; both at pH 11) as shown in Fig. 8. In all
(over 100% cleaning efficiency). Thus, to better understand cases, the results are compared for cleaning performed for 15
the cleaning mechanisms, the “stoichiometry” between the and 60 min.
cleaning agent dose and amount of foulant on the membrane For NaOH (pH 11) cleaning, an increase in cleaning time
surface must be considered. from 15 to 60 min resulted in negligible difference in cleaning

Fig. 7. Variation in NaOH and EDTA cleaning efficiencies with respect to cleaning time: (a) flux behavior during fouling/cleaning experiments and (b) cleaning
efficiency. Cleaning experiments were performed at pH 11.0 following fouling runs carried out at the same conditions as those in Fig. 2a. Other cleaning
conditions were identical to those in Fig. 2b except the cleaning time as indicated in the figure. Note that the last data points in Fig. 7a are for flux measured
with DI water after the corresponding cleaning.
206 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

Fig. 8. Variation in SDS cleaning efficiency with respect to cleaning time Fig. 9. Variation in EDTA cleaning efficiency with the crossflow velocity
at different SDS concentrations. Cleaning experiments were performed at during cleaning with different EDTA concentrations. Cleaning experiments
pH 11 following fouling runs carried out at the same conditions as those in were performed at pH 11 following fouling runs carried out at the same
Fig. 2a. Other cleaning conditions were identical to those in Fig. 2b except conditions as those in Fig. 2a. Other cleaning conditions were identical to
the cleaning time as indicated in the figure. those in Fig. 2b except the crossflow velocity as indicated in the figure.

of cleaning time should be taken into account since increas-

efficiency (Fig. 7). This is attributed to the limited chemical ing cleaning time does not always result in a corresponding
reaction between NaOH and foulants in the alginate gel layer. increase in cleaning efficiency.
In other words, a longer contact time did not contribute to
enhancing cleaning efficiency, unless there was a favorable 3.2.2. Effect of crossflow velocity
chemical reaction between the cleaning agent and foulants The influence of crossflow velocity (during cleaning) on
in the fouling layer to lessen foulant–foulant interactions. cleaning efficiency was investigated with solutions contain-
Therefore, we conclude that NaOH is not a suitable cleaning ing different doses of EDTA (0.5 and 2.0 mM; both at pH 11),
agent for membranes fouled by organic matter in presence as shown in Fig. 9. It is shown that cleaning efficiency with
of calcium ions. Besides, as mentioned earlier, increasing 0.5 mM EDTA did not improve significantly with increasing
the NaOH concentration and, hence, resulting in a pH higher crossflow velocity, while there was a discernible increase in
than 12, is not applicable for most polymeric RO membranes. cleaning efficiency with 2.0 mM EDTA as crossflow velocity
In the case of EDTA cleaning, there was a twofold increase increased. We have shown earlier that, for a cleaning agent
in cleaning efficiency when the cleaning time was increased to be effective, the chemical reaction with the foulants has
from 15 to 60 min. Contrary to NaOH, EDTA is a favorable to be favorable before physical interaction (due to crossflow
cleaning agent since EDTA reacts specifically within the foul- velocity) takes place. With 0.5 mM EDTA, the concentra-
ing layer through a ligand-exchange reaction. tion of cleaning agent was not enough to produce a favorable
Previous results (Fig. 5) demonstrated that the efficiency chemical reaction with the foulants in the fouling layer within
of SDS cleaning is significantly influenced by SDS concen- the 15 min cleaning time. An increase in crossflow velocity
tration with respect to its CMC. As shown in Fig. 8, increasing which resulted in an increase in the shear rate did not enhance
cleaning time from 15 to 60 min resulted in nearly 100% the mass transfer of foulants in the fouling layer to the bulk
cleaning efficiency when 10 mM SDS was used. However, solution because the EDTA molecules did not react strongly
cleaning with 2 mM SDS for 60 min was still ineffective (less with the foulants, and the structural integrity was still very
than 30% cleaning efficiency). This observation is closely much intact. However, with 2.0 mM EDTA, an increase in
related to the previous results for NaOH and EDTA clean- shear rate enhanced the mass transfer of the foulants from the
ing performed for different cleaning times. Similar to NaOH, fouling layer to the bulk solution, as 2.0 mM EDTA is favor-
cleaning with 2 mM SDS was, to some extent, limited in terms able for a rigorous chemical reaction. This led to an increase
of chemical reaction and, thus, increasing cleaning time did in cleaning efficiency with increasing crossflow velocity. We
not significantly enhance the cleaning efficiency. However, conclude that the chemical reaction of the cleaning agent with
increasing cleaning time for the 10 mM SDS solution led to deposited foulants and the associated mass transfer phenom-
an almost complete cleaning of the fouled membrane since ena are quite important in membrane cleaning.
cleaning with 10 mM SDS was favorable in terms of chemi-
cal reaction between SDS and the foulants within the fouling 3.2.3. Effect of temperature
layer. Therefore, while considering the chemical aspects of The results for EDTA cleaning performed at different tem-
cleaning (e.g. deciding the cleaning agent dose), optimization peratures are presented in Fig. 10. We note that, in both
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 207

the fouling layer plays a critical role in determining the overall

cleaning efficiency. Determining the intermolecular adhesion
forces between foulants in the fouling layer in the presence, or
absence, of cleaning solution enables us to predict the capa-
bility of chemical reaction between cleaning agent and the
foulants. A more favorable (or reactive) cleaning agent will
result in less intermolecular adhesion force between foulants
in the fouling layer.
The intermolecular adhesion force was first determined in
the absence of cleaning solution to serve as a baseline and,
then, in the presence of the cleaning agent. In the absence of
cleaning solution, the interfacial adhesion force was deter-
mined with the same solution used as feed water for the
fouling runs (i.e., 20 mg/l alginate, 0.5 mM Ca2+ , 10 mM total
ionic strength, and pH 6.0). The interfacial adhesion forces
determined in the presence of cleaning solutions (same solu-
Fig. 10. Variation in EDTA cleaning efficiency with respect to cleaning tion chemistries as those used in the cleaning experiments)
solution temperature. Cleaning experiments were performed with 0.5 mM were compared with the cleaning efficiency determined from
EDTA solution at pH 11 following fouling runs carried out at the same
fouling/cleaning experiments, as shown in Fig. 11.
conditions as those in Fig. 2a. Other cleaning conditions were identical to
those in Fig. 2b except the temperature as indicated in the figure. Fig. 11a shows the intermolecular adhesion forces deter-
mined in the presence of different cleaning solutions (i.e.,
cases, the pure water flux after cleaning was determined at NaOH, 0.5 mM EDTA, and 10 mM SDS, all at pH 11) as
20 ◦ C. The cleaning efficiency with 0.5 mM EDTA (pH 11) well as in the absence of cleaning solution. The maximum
increased dramatically with increasing cleaning temperature adhesion force (corresponds to the minimum in the adhesion
from 20 to 40 ◦ C. Both the rate of chemical reaction of EDTA force curve) in the absence of cleaning solution was about
with the deposited foulants and the diffusive transport of 1.5 mN/m. The addition of SDS to the test solution was the
foulants from the fouling layer to the bulk solution increased most effective, almost eliminating the adhesion force. The
as temperature was increased. At a higher temperature, the addition of EDTA to the test solution decreased substantially
swelling of the alginate gel layer might have also contributed the adhesion force, reducing the maximum adhesion force
to weakening its structural stability [34,35]. to 0.4 mN/m. In the presence of NaOH (pH 11), however,
a significant adhesion force (1.3 mN/m) was still present. It
3.3. Relating cleaning efficiency to measured adhesion is also noted that, with the reduction in adhesion force in
forces presence of the cleaning solution, the distance from the foul-
ing layer on the membrane surface to the point where no
Based on the preceding results, it has been shown that adhesion force was observed was reduced correspondingly.
chemical reaction of the cleaning agent with the foulants in The reduction in intermolecular adhesion force with each

Fig. 11. Relationship between intermolecular adhesion forces and cleaning efficiency: (a) variation in foulant–foulant adhesion forces as a function of cleaning
solution type (the curve indicated as ‘no cleaning agent’ was determined in the absence of cleaning solution) and (b) variation in cleaning efficiency as a function
of cleaning solution type. The test solutions used in the interfacial force measurements were the same as those used during fouling runs and also contained
cleaning solutions as indicated in the figure. Cleaning experiments were carried out for 60 min following fouling runs performed at the same conditions as those
in Fig. 2a. Other cleaning conditions were identical to those in Fig. 2b.
208 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

cleaning solution was directly related to the cleaning effi- governed by the chemical reaction of the cleaning agent
ciency as shown in Fig. 11b. These cleaning efficiencies were with the foulants in the fouling layer. A more favorable
determined from experiments with each cleaning solution chemical reaction would lead to a lower foulant–foulant
at fixed physical cleaning conditions (60 min cleaning time, adhesion force and hence higher cleaning efficiency. This
crossflow velocity of 42.8 cm/s, and 20 ◦ C). general relationship between intermolecular adhesion force
The results in Fig. 11 demonstrate a strong relationship and cleaning efficiency has also been demonstrated in our
between the efficiency of a cleaning solution and the resid- recent study with humic acid-fouled NF membranes [25].
ual intermolecular adhesion force in the presence of the Therefore, the selection of a favorable cleaning solution in
same cleaning solution. This is because the efficiency of a terms of chemical reaction between the cleaning agent and the
cleaning solution at fixed physical conditions was mainly foulants in the fouling layer can be predicted by measuring the

Fig. 12. Schematic representation for effective cleaning of organic-fouled RO membranes. A cross-linked fouling layer is formed on the membrane surface
in presence of calcium ions, which bind to organic foulants and form bridges between adjacent foulant molecules. During cleaning, the cleaning agent reacts
with the foulants in the fouling layer yielding loosened foulants. These reaction products are removed from the fouling layer to the bulk solution through the
hydrodynamics/mass transfer. Thus, efficient cleaning can be achieved through the coupling between the chemical reaction and mass transfer along with the
optimization of cleaning conditions responsible for the favorable chemical reaction and mass transfer.
 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210 209

intermolecular adhesion force prior to the fouling/cleaning of loosened foulants from the fouling layer to the bulk solu-
experiments. Following this prediction, the selected favor- tion. The chemical reaction is greatly influenced by the type
able cleaning solution can be used for cleaning experiments to and dose of cleaning agent as well as cleaning solution pH.
determine the optimal physical cleaning conditions in terms In addition, the fouling layer composition also influences the
of both cleaning efficiency and operating cost. reactivity of a cleaning agent with foulants in the fouling
layer.
3.4. Effective cleaning requires favorable chemical and EDTA and SDS were quite effective in reacting with
physical interactions organic foulants in the fouling layer formed in the presence of
calcium ions, while NaOH cleaning resulted in poor cleaning
Membrane cleaning involves both chemical and physical efficiency due to its limited reactivity with deposited foulants.
interactions. These interactions include (i) chemical reaction Cleaning efficiency with EDTA and SDS was improved by
between the cleaning agent and the foulants in the fouling optimizing cleaning agent dose and solution pH. However,
layer and (ii) mass transfer of cleaning agent from the bulk these chemical factors hardly contributed to improving the
solution to the fouling layer and the foulants from the fouling efficiency of NaOH cleaning. For favorably reactive clean-
layer back to the bulk solution. A schematic representation ing agents, cleaning efficiency can be further improved by
of the roles of chemical and physical interactions for effec- enhancing the mass transfer of the reaction products from
tive cleaning of organic-fouled RO membranes is shown in the fouling layer to the bulk solution. Mass transfer of the
Fig. 12. Effective cleaning can be achieved only when both foulants from the fouling layer to the bulk solution was mainly
the chemical and physical interactions are favorable as dis- controlled by the crossflow velocity, whereas cleaning time
cussed below. and temperature affected both the mass transfer and chemical
In the presence of a cleaning solution, chemical reaction reaction.
will occur between the cleaning agent and the foulants in the The AFM force measurements show that the chemical
fouling layer. In this study, we have seen that the effective- reaction between cleaning agents and deposited foulants
ness of cleaning in terms of chemical reactivity depends on plays an important role in reducing the intermolecular adhe-
the type of cleaning solution, cleaning solution pH, cleaning sion force between foulants in the fouling layer. A remarkable
chemical dose, and the foulant chemical composition. When a correlation is noted between the cleaning efficiencies and the
cleaning agent has a favorable chemical reactivity, the clean- magnitudes of the adhesion force measured in the presence
ing agent, upon contact with the foulants, will be able to of the corresponding cleaning solutions. The cleaning effi-
react with the foulants and weaken the structural integrity of ciency was inversely related to the residual foulant–foulant
the fouling layer. The physical (hydrodynamic) conditions, adhesion force, implying that favorable cleaning agents are
which are mainly responsible for the mass transfer of the those effective in reducing the intermolecular adhesion force.
reaction products, then play an important role in removing It is therefore suggested that AFM force measurements can
the foulants from the fouling layer. be used as a tool to predict the effectiveness of the chemical
As we have shown earlier, some physical conditions (i.e., reactivity of a cleaning agent. The selection of reactive clean-
cleaning time and temperature) affect the mass transfer and ing agents is of paramount importance in determining the
chemical reaction concurrently, unless the cleaning agent is overall cleaning efficiency since the optimization of physical
limited in terms of chemical reactivity. When a cleaning agent conditions could improve the cleaning efficiency only when
has a favorable chemical reaction with the foulants in the foul- the chemical reaction of the cleaning agent with deposited
ing layer, an optimization of the physical conditions improves foulants is favorable.
its cleaning efficiency significantly, as illustrated for 0.5 mM
EDTA and 10 mM SDS solutions when the cleaning time
was increased from 15 to 60 min. However, optimizing the Acknowledgements
physical conditions will not enhance the cleaning efficiency
of a cleaning agent which does not have a favorable chemi- This study was supported by the US Department of Inte-
cal reaction with the foulants, as shown for NaOH solution rior, Bureau of Reclamation, Desalination and Water Purifi-
when the cleaning time was increased from 15 to 60 min. As cation Research and Development Program (Award No. 03-
we emphasized earlier, effective cleaning of organic-fouled FC-81-0846), by the Public Utilities Board of Singapore
RO membranes requires both the physical and chemical inter- (PUB) for a doctoral fellowship to W.S. Ang, and by the
actions to be favorable. post-doctoral fellowship program of Korea Science and Engi-
neering Foundation (KOSEF) for S. Lee.

4. Conclusion
References
The cleaning of organic-fouled RO membranes is accom- [1] J.E. Drewes, M. Reinhard, P. Fox, Comparing microfiltration-reverse
plished by the chemical reaction between cleaning agents and osmosis and soil-aquifer treatment for indirect potable reuse of water,
foulants in the fouling layer, followed by the mass transfer Water Res. 37 (2003) 3612–3621.
210 W.S. Ang et al. / Journal of Membrane Science 272 (2006) 198–210

[2] M.P. del Pino, B. Durham, Wastewater reuse through dual-membrane [21] M.J. Rosen, Surfactant and Interfacial Phenomena, 2nd ed., Wiley,
processes: opportunities for sustainable water resources, Desalination New York, 1989.
124 (1999) 271–277. [22] N. Hilal, L. Al-Khatib, B.P. Atkin, V. Kochkodan, N. Potapchenko,
[3] D.J. Barker, S.M.L. Salvi, A.A.M. Langenhoff, D.C. Stuckey, Solu- Photochemical modification of membrane surfaces for (bio)fouling
ble microbial products in ABR treating low-strength wastewater, J. reduction: a nano-scale study using AFM, Desalination 158 (2003)
Environ. Eng. 126 (2000) 239–249. 65–72.
[4] J.E. Drewes, P. Fox, Fate of natural organic matter (NOM) during [23] N. Hilal, A.W. Mohammad, B. Atkin, N.A. Darwish, Using atomic
groundwater recharge using reclaimed water, Water Sci. Technol. 40 force microscopy towards improvement in nanofiltration membranes
(1999) 241–248. properties for desalination pre-treatment: a review, Desalination 157
[5] P. Sciener, S. Nachaiyasit, D.C. Stuckey, Production of soluble (2003) 137–144.
microbial products (SMP) in an anaerobic baffled reactor: composi- [24] N. Hilal, W.R. Bowen, Atomic force microscope study of the
tion, biodegradability and the effect of process parameters, Environ. rejection of colloids by membrane pores, Desalination 150 (2002)
Technol. 19 (1998) 391–400. 289–295.
[6] B.E. Rittmann, W. Bae, E. Namkung, C.J. Lu, A critical evaluation [25] Q. Li, M. Elimelech, Organic fouling and chemical cleaning if
of microbial product formation in biological processes, Water Sci. nanofiltration membranes: measurements and mechanisms, Environ.
Technol. 19 (1987) 517–528. Sci. Technol. 38 (2004) 4683–4693.
[7] J. Manka, M. Rebhun, A. Mandelbaum, A. Bortinger, Characteri- [26] S. Lee, B. Kwon, M. Sun, J. Cho, Characterization of NOM
zation of organics in secondary effluents, Environ. Sci. Technol. 8 included in NF and UF membrane permeates, Desalination 173
(1974) 1017–1020. (2005) 131–142.
[8] C. Jarusutthirak, G. Amy, J.-P. Croué, Fouling characteristics of [27] S. Lee, G. Amy, J. Cho, Applicability of Sherwood correlations
wastewater effluent organic matter isolates on NF and UF mem- for natural organic matter (NOM) transport in nanofiltration (NF)
branes, Desalination 145 (2002) 247–255. membranes, J. Membr. Sci. 240 (2004) 49–65.
[9] T. Mohammadi, S.S. Madaeni, M.K. Moghadam, Investigation of [28] E.M. Vrijenhoek, S.K. Hong, M. Elimelech, Influence of mem-
membrane fouling, Desalination 153 (2002) 155–160. brane surface properties on initial fouling rate of colloidal fouling of
[10] R. Liikanen, J. Yli-Kuivila, R. Laukkanen, Efficiency of various reverse osmosis and nanofiltration membranes, J. Membr. Sci. 188
chemical cleanings for nanofiltration membranes fouled by conven- (2001) 115–128.
tionally treated surface water, J. Membr. Sci. 195 (2002) 265–276. [29] P. Mukerjee, K.J. Mysels, Critical Micelle Concentrations of Aque-
[11] S.S. Madaeni, T. Mohammadi, M.K. Moghadam, Chemical cleaning ous Surfactant Solutions, Nat. Stand. Ref. Data Ser., U.S. Depart-
of reverse osmosis membranes, Desalination 134 (2001) 77–82. ment of Commerce, National Bureau of Standards, Washington, DC,
[12] J.J. Sadhwani, J.M. Vesa, Cleaning tests for seawater reverse osmosis NSRDS–NBS 36, 1970.
membranes, Desalination 139 (2001) 177–182. [30] The Merck Index, S. Budavari, (Eds.), An Encyclopaedia of Chem-
[13] S. Ebrahim, Cleaning and regeneration of membranes in desalination icals, Drugs, and Biologicals, Merck and Company Inc. Publishers,
and wastewater applications: state-of-the-art, Desalination 96 (1994) Rahway, New Jersey, 1989.
225–238. [31] J.A. Dean (Ed.), Handbook of Organic Chemistry, McGraw-Hill
[14] R. Deqian, Cleaning and regeneration of membranes, Desalination Book Company, New York, 1987.
62 (1987) 363–371. [32] A. Seidel, M. Elimelech, Coupling between chemical and physical
[15] T.K. Osta, L.M. Bakheet, Pretreatment system in reverse osmosis interactions in natural organic matter (NOM) fouling of nanofiltration
plants, Desalination 63 (1987) 71–80. membranes: implications for fouling control, J. Membr. Sci. 203
[16] G. Trägårdh, Membrane cleaning, Desalination 71 (1989) 325–335. (2002) 245–255.
[17] E.M. Thurman, Organic Geochemistry of Natural Waters, Martinus [33] J. Cho, G. Amy, J. Pellegrino, Membrane filtration of natural organic
Nijhoff/Dr. W. Junk Publishers, Boston, MA, 1985. matter: factors and mechanisms affecting rejection and flux decline
[18] D.A. Rees, Polysaccharide shapes and their interactions — some with charged ultrafiltration (UF) membrane, J. Membr. Sci. 164
recent advances, Pure Appl. Chem. 53 (1984) 1–14. (2000) 89–110.
[19] G.T. Grant, E.R. Morris, D.A. Rees, J.C. Smith, D. Thom, Biological [34] S.T. Moe, K.I. Draget, G. Skjåk-Bræk, O. Smidsrød, Tempera-
interaction between polysaccharides and divalent cations: the egg- ture dependence of the elastic modulus of alginate gels, Carbohyd.
box model, FEBS Lett. 32 (1973) 195–198. Polym. 19 (1992) 279–284.
[20] S. Hong, M. Elimelech, Chemical and physical aspects of natu- [35] A.H. Clark, S.B. Ross-Murphy, Structural and mechanical properties
ral organic matter (NOM) fouling of nanofiltration membranes, J. of biopolymer gels, Adv. Polym. Sci. 83 (1987) 57–192.
Membr. Sci. 132 (1997) 159–181.