Uploaded by onlyname13

A STUDY ON ORGANIC FOULING OF REVERSE

advertisement
View metadata, citation and similar papers at core.ac.uk
brought to you by
CORE
provided by ScholarBank@NUS
A STUDY ON ORGANIC FOULING OF REVERSE
OSMOSIS MEMBRANE
MO HUAJUAN
(B.&M.Eng., ECUST)
A THESIS SUBMITTED
FOR THE DEGREE OF PhD OF ENGINEERING
DEPARTMENT OF CIVIL ENGINEERING
NATIONAL UNIVERSITY OF SINGAPORE
2009
Acknowledgement
Acknowledgement
This is for me an enriching journey of challenges, opportunities, and excitement.
Without the many wonderful people to whom I owe millions of supports and help, it
could be impossible. My deepest appreciation goes to my supervisor Associate
Professor Ng How Yong and Professor Ong Say Leong for their constant
encouragement, invaluable guidance, patience and understanding in research and life
throughout the whole length of my PhD candidature. Special thanks also to all the
laboratory officers, friends and my family; as well as anyone who have helped me in
one way or another during my PhD study.
i
Table of contents
Table of Contents
Acknowledgement
i
Table of Contents
ii
Summary
vii
List of Tables
ix
List of Figures
x
Nomenclature
xv
Chapter 1 Introduction
1
1.1 Background
2
1.2 Problem statement
7
1.3 Research objectives
10
1.4 Organization of thesis
12
Chapter 2 Literature Review
2.1 Dissolved organic matters in water reclamation system
15
15
2.1.1 Source of dissolved organic matters
15
2.1.2 Dissolved organic matters removal by MF/UF
17
2.1.3 Model polysaccharides and proteins
18
2.2 Membrane and membrane process system
21
2.2.1 Membrane definition and process classification
21
2.2.2 Basic membrane transport theory for RO process
24
2.2.3 Concentration polarization
26
ii
Table of contents
2.2.4 Spiral wound membrane module and the permeate flux behavior
2.3 Membrane fouling
29
33
2.3.1 Definition and types of membrane fouling in RO process
33
2.3.2 Fouling mechanisms in RO process
35
2.3.3 Key issues in organic fouling
37
2.3.4 Membrane cleaning
46
Chapter 3 Materials and Methods
50
3.1 Chemical solution preparation
50
3.1.1 Model organic matters
50
3.1.2 Other chemicals
50
3.2 RO membranes
51
3.3 RO filtration setup
52
3.3.1 Small lab-scale crossflow membrane cell
52
3.3.2 Long channel crossflow RO membrane cell
54
3.4 Filtration and cleaning operation
57
3.4.1 Salt solution filtration tests
57
3.4.2 Organic fouling tests
57
3.4.3 Cleaning tests
58
3.5 Membrane hydraulic resistance measurement
59
3.6 Beaker tests of gel formation
60
3.7 Analytical techniques
60
3.7.1 Streaming potential analyzer
60
3.7.2 SEM-EDX
61
3.7.3 Polysaccharide and protein assay
62
iii
Table of contents
Chapter 4 Polysaccharide Fouling and Chemical Cleaning
4.1 Polysaccharide fouling
63
64
4.1.1 Effects of calcium concentration on alginate fouling
64
4.1.2 Effects of alginate concentration on alginate fouling
65
4.1.3 Gel formation in beaker tests
68
4.2 Chemical cleaning of membranes fouled by polysaccharide
70
4.2.1 Effects of calcium on membrane cleaning
70
4.2.2 Effects of pH and concentration of cleaning solution
72
4.2.3 Impact of multiple cleaning cycles
76
4.3 Summary
Chapter 5 Protein Fouling and Chemical Cleaning
5.1 Protein fouling
78
79
79
5.1.1 Adsorption of BSA on membrane
79
5.1.2 Effects of ionic strength
81
5.1.3 Effects of cations
84
5.1.4 Effects of temperature
89
5.2 Chemical cleaning of membranes fouled by protein
93
5.2.1 Effects of cleaning solution concentration
93
5.2.2 Effects of cleaning solution pH
95
5.2.3 Effects of cleaning time
96
5.3 Summary
98
iv
Table of contents
Chapter 6 Permeate Behavior and Concentration Polarization in a Long
RO Membrane Channel
100
6.1 Calculation of concentration polarization
101
6.2 Variation of permeate flux along the channel
104
6.3 Variation of rejection along the channel
109
6.4 CP and CF variation along the channel
112
6.5 Correlation between recovery and CP
115
6.6 Permeate performance in a spacer-filled channel
119
6.6 Summary
124
Chapter 7 Organic Fouling Development in a Long RO Membrane Channel 127
7.1 Organic fouling development along the channel
128
7.1.1 The permeate behavior in alginate and BSA fouling along the
RO membrane channel
128
7.1.2 Effects of operating conditions on organic fouling development
along the channel
7.1.3 Effects of feed spacer(s) on alginate fouling
133
138
7.2 Key factors in organic fouling development in a long membrane
channel and numerical study
139
7.2.1 Model development
140
7.2.2 Comparison between experimental work and numerical
simulation
7.3 Summary
145
149
v
Table of contents
Chapter 8 Conclusions and Recommendations
151
8.1 Conclusions
152
8.2 Recommendations
155
References
158
Appendix
170
vi
Summary
Summary
Reverse osmosis (RO) is a valuable membrane separation process and is increasingly
used in water reclamation because of its high product quality and low costs. The
efficiency of RO membrane is limited most notably by membrane fouling, which
refers to the accumulation of foulant present even in minute quantity in the RO feed.
An understanding of the feed solution, that is, the foulant composition, is the first step
towards formulating a fouling mitigation strategy. Within the commonly encountered
foulants in water reclamation, organic fouling is a major category which include
humic acids, polysaccharides, proteins, etc. A key issue in organic fouling is the
various interactions between organic foulants, inorganic components of the feed and
the RO membranes.
Typically, a small lab-scale RO membrane cell can be used to investigate the organic
fouling behavior, but it cannot completely represent what actually happens in a
membrane module. The full-scale RO membrane channel has been theoretically
shown to be of a heterogeneous system, which is characterized by variation of water
flow and mass concentration along the flow channel. These variable parameters will
inevitably affect the distribution of the deposited organic foulants. Hence, compared
with an average permeate flux, local permeate flux is more reliable to describe the
fouling development in RO membrane channel, which will add further to our
knowledge on organic fouling in a real plant.
vii
Summary
In this thesis, sodium alginate and bovine serum albumin (BSA) chosen as model
polysaccharide and protein, respectively, were used to study the polysaccharide and
protein fouling behavior in two lab-scale RO membrane cells of different dimensions.
The first test cell was 0.1 m long and was treated as a homogenous system while the
second one was a 1-m long cell which was designed to measure five local permeate
flux along the channel. The study began with an investigation of RO membrane
fouling by alginate. The presence of calcium in the feed solution intensively
magnified alginate fouling potential. Other chemical (pH, ionic strength, cation
species) and physical (temperature) parameters of feed water were investigated in the
study of RO membrane fouling by BSA. It is noted that the most severe BSA fouling
occurred at pH near the iso-electric point (IEP) of BSA. The study proceeded with an
investigation into the behavior of permeate flux in a long RO membrane channel. This
is the first report to experimentally show the heterogeneous distribution of flow and
mass due to exponential growth of salt concentration polarization in a long RO
membrane channel. Interestingly, in this long membrane channel, permeate flux was
observed to decline faster at one end than the other end of the channel when the
organic fouling progressed. In addition, modeling efforts simulated the alginate
fouling development in the 1-m long RO membrane channel by incorporating a
modified fouling potential and deposition ratio and predicted well the experimental
results.
viii
List of tables
List of Tables
Table 2.1
Some membrane processes and their driving forces (Mulder, 1996).
Table 2.2
Classification of pressure-driven membrane processes (Mulder,
1996).
Table 2.3
Empirical relations of the concentration dependence of osmotic
pressure for different salt (Lyster and Cohen, 2007).
Table 3.1
Surface characterization of LFC1 and ESPA2 RO membrane.
Table 4.1
Atomic weight percentage on the membrane surface by SEM-EDX.
Table 7.1
RO parameter values for simulation.
ix
List of figures
List of Figures
Figure 1.1
A schematic diagram of the research objectives and scope of this
study.
Figure 2.1
Alginate molecular structure: (a) alginate monomers (uronic acids:
M vs G. The carbon atoms C-2 and C-3 of the mannuronate units are
partially acetylated (R= -H or -COCH3), all C-5 carbon atoms carry a
carboxylate group that may be partially protonated); (b)
macromolecular conformation of the alginate polymer; (c) chain
sequences; block copolymer structure; (d) calcium induced gelation
of alginate: schematic representation in accordance with the “eggbox” structure (Davis et al., 2003).
Figure 2.2
A schematic of a spiral wound module showing the flow directions,
feed and permeate channels including spacers.
Figure 3.1
(a) Schematic diagram of the crossflow RO filtration setup. (b)
Picture of the small lab-scale RO setup.
Figure 3.2
(a) Schematic diagram of the long channel RO membrane cell. (b)
Picture of the long channel RO membrane cell (to be continued).
Figure 4.1
Effects of calcium concentration (0, 0.1, 0.3, and 1.0 mM) on
permeate flux with time over a period of 50 h. Ionic strength was
maintained at 10 mM by varying the sodium chloride concentration.
Sodium alginate concentration was 50 mg/L. pH was unjusted at
6.0±0.1. Initial permeate flux was 1.38×10-5 m/s. Crossflow velocity
was 0.0914 m/s.
Figure 4.2
Permeate flux over a period of 50 h at different alginate
concentrations (10, 20, and 50 mg/L). Calcium concentration was
1.0 mM and ionic strength was 10 mM. pH was unjusted at 6.0±0.1.
Initial permeate flux was 1.38×10-5 m/s. Crossflow velocity was
0.0914 m/s.
Figure 4.3
Carbon to calcium weight ratio on the fouling layer of the
membrane. The fouled membrane samples were obtained from six
runs with different filtration time. Sodium alginate concentration
was 10, 20, or 50 mg/L and calcium concentration was 1.0 mM for
all runs. pH was unjusted at 6.0±0.1. Initial permeate flux was
1.38×10-5 m/s. Crossflow velocity was 0.0914 m/s.
x
List of figures
Figure 4.4
Turbidity change with gel formation by sodium alginate and calcium
in beaker tests. Figure 4.4b is the extended range of Y axis for 10
mM calcium. The ovals in Figure 4.4a highlight the turning points.
Figure 4.5
Comparison of the extent of flux restoration with and without
calcium in the feed water using different types of chemical cleaning
agents. The fouling test conditions: sodium alginate of 50 mg/L;
calcium of 0 or 1.0 mM; ionic strength of 10 mM; pH unjusted
(6.0±0.1); initial permeate flux of 1.38×10-5 m/s; crossflow velocity
of 0.0914 m/s. Cleaning conditions are described in Chapter 3.
Figure 4.6
Flux restoration with EDTA cleaning at different (a) pH (the
concentration of EDTA solution was fixed at 1mM); and (b) EDTA
concentrations (the pH of EDTA solution was fixed at 4.5). Fouling
conditions: sodium alginate of 50 mg/L; calcium of 1.0 mM; ionic
strength of 10 mM; pH unjusted (6.0±0.1); initial permeate flux of
1.38×10-5 m/s; and crossflow velocity of 0.0914 m/s. Cleaning
conditions are described in Chapter 3.
Figure 4.7
Flux restoration after membrane cleaning with SDS at different (a)
pH (the concentration of SDS solution was fixed at 1mM); and (b)
SDS concentrations (the pH of SDS solution was fixed at 8). Fouling
conditions: sodium alginate of 50 mg/L; calcium of 1.0 mM; ionic
strength of 10mM; pH unjusted (6.0±0.1); initial permeate flux of
1.38×10-5 m/s; crossflow velocity of 0.0914 m/s. Cleaning
conditions are described in Chapter 3.
Figure 4.8
Normalized fluxes over three successive cycles of alginate fouling
and EDTA cleaning. Fouling conditions: alginate of 50 mg/L;
calcium of 1.0 mM; ionic strength 10 of mM, pH unjusted (6.0±0.1);
initial permeate flux of 1.38×10-5 m/s; and crossflow velocity 0.0914
m/s.
Figure 4.9
Comparison of flux restoration after each successive EDTA
cleaning. Cleaning conditions: EDTA concentration of 1 mM and pH
4.5. Other conditions are described in Chapter 3.
Figure 5.1
Zeta potential on RO membrane surface as a function of pH at
different ionic strength and in the presence/absence of BSA (50
mg/L BSA).
Figure 5.2
Normalized flux with time at three different pH under different ionic
strength of (a) 1 mM and (b) 10 mM.
Figure 5.3
Zeta potential on RO membrane surface as a function of pH with
different cation species. Ionic strength was maintained at 10 mM for
all cation species and BSA concentration was maintained at 50
mg/L.
xi
List of figures
Figure 5.4
Normalized flux decline with time at three different pH with
different cation species (a) 10 mM Na+; (b) 7 mM Na+ and 1 mM
Ca2+; (c) 7 mM Na+ and 1 mM Mg2+ ( to be continued).
Figure 5.5
Zeta potential on RO membrane surface as a function of temperature
for three different pH. BSA concentration was 50 mg/L and ionic
strength of solution was 10 mM (1 mM Ca2+ and 7 mM Na+).
Figure 5.6
Normalized flux with time at different temperatures for different pH:
(a) pH=3.9; (b) pH=4.9; (c) pH=7. BSA concentration was 50 mg/L
and ionic strength of solution was 10 mM (1 mM Ca2+ and 7 mM
Na+) (to be continued).
Figure 5.7
Variation of SDS, urea and EDTA cleaning efficiencies as a function
of cleaning solution concentration (Cleaning time of 60 min;
operating pressure of 15 psi; crossflow velocity 0.29 m/s; and pH 8.0
for SDS, 7.8 for urea and 5.2 for EDTA).
Figure 5.8
Variation of SDS, urea and EDTA cleaning efficiencies as a function
of cleaning solution pH (Cleaning conditions: cleaning time of 60
min; operating pressure of 15 psi; crossflow velocity of 0.29 m/s;
solution concentration of 1 mM for SDS and urea and 0.1 mM for
EDTA).
Figure 5.9
Variation of SDS and urea cleaning efficiencies as a function of
cleaning time (Cleaning conditions: operating pressure of 15 psi;
crossflow velocity of 0.29 m/s; cleaning solution pH of 8.0 for SDS
and 7.8 for urea).
Figure 6.1
Clean ESPA2 membrane hydrodynamic resistance at different points
along the channel at different applied pressure.
Figure 6.2
Permeate flux variation along the channel for DI water filtration
(void symbol) and salt solution filtration (feed concentration of 30
mM, solid symbol) at (a) different pressure (feed flow of 0.091 m/s)
and (b) different feed flow (applied pressure of 300 psi).
Figure 6.3
Net driving force on permeate (solid symbol) and trans-membrane
osmotic pressure (void symbol) variation along the channel for salt
solution filtration (feed concentration of 30 mM) at different (a)
applied pressure (feed flow of 0.091 m/s) and feed flow (applied
pressure of 300 psi).
Figure 6.4
True rejection for salt solution filtration (feed concentration of 30
mM) at (a) different applied pressure (feed flow of 0.0914 m/s) and
(b) feed flow (applied pressure of 300 psi).
xii
List of figures
Figure 6.5
Observed rejection variation along the channel for salt solution
filtration (feed concentration of 30 mM) at different (a) applied
pressure (feed flow of 0.091 m/s) and (b) feed flow (applied pressure
of 300 psi).
Figure 6.6
CP modulus (solid symbol) and CF (void symbol) variation along
the channel for salt solution filtration (feed concentration of 30 mM)
at different (a) applied pressure (feed flow of 0.091 m/s) and (b) feed
flow (applied pressure of 300 psi).
Figure 6.7
(a) Cumulative recovery and correlation between cumulative
recovery and CP modulus for salt solution filtration at different
applied pressure (feed concentration of 30 mM and feed flow of
0.091 m/s).
Figure 6.8
(a) Cumulative recovery and (b) correlation between cumulative
recovery and CP modulus for salt solution filtration at different feed
flow (feed concentration of 30 mM and applied pressure of 300 psi).
Figure 6.9
Permeate flux variation along the channel with or without a spacer
inserted (applied pressure of 300 psi and feed flow of 0.091 m/s).
Figure 6.10
CP modulus growth along the channel with or without a spacer
inserted (applied pressure of 300 psi and feed flow of 0.091 m/s).
Figure 6.11
CF variation along the channel with or without a spacer inserted
(applied pressure of 300 psi and feed flow of 0.091 m/s).
Figure 6.12
Observed rejection variation along the channel with or without a
spacer inserted (applied pressure of 300 psi and feed flow of 0.091
m/s).
Figure 7.1
Permeate flux (a) and normalized flux (b) evolution over 25 h of
alginate fouling at different locations along the channel (Fouling
conditions: alginate of 50 mg/L; ionic strength of 10 mM; calcium of
1.0 mM; unjusted pH at 6.0±0.1; temperature 25oC; initial flux of
2.00×10-5 m/s; crossflow velocity of 0.09 m/s).
Figure 7.2
Permeate flux (a) and normalized flux (b) evolution over 25 h of
protein fouling at different locations along the channel (Fouling
conditions: BSA of 50 mg/L; ionic strength of 10 mM; calcium of
1.0 mM; unjusted pH at 6.0±0.1; temperature of 25oC; initial flux of
2.00×10-5 m/s; crossflow velocity of 0.09 m/s).
Figure 7.3
Effects of the initial flux on alginate (a) and BSA (b) fouling along
the channel (Fouling conditions: alginate or BSA of 50 mg/L; ionic
strength of 10 mM; calcium of 1.0 mM; unjusted pH at 6.0±0.1;
temperature of 25oC; crossflow velocity of 0.09 m/s).
xiii
List of figures
Figure 7.4
Effects of crossflow velocity on (a) alginate and (b) BSA fouling
along the channel (Fouling conditions: alginate or BSA of 50 mg/L;
ionic strength of 10 mM; calcium of 1.0 mM; unjusted pH at
6.0±0.1; temperature of 25oC; initial flux of 2.00×10-5 m/s).
Figure 7.5
Effects of feed spacer on the alginate fouling development along the
spacer-free channel and the channel with one or two spacers inserted
(Fouling conditions: alginate of 50 mg/L; ionic strength of 10 mM;
calcium of 1.0 mM; unjusted pH at 6.0±0.1; temperature of 25oC;
initial flux of 2.00×10-5 m/s; crossflow velocity of 0.09 m/s).
Figure 7.6
An example of the modified fouling potential km evolution during
alginate fouling process.
Figure 7.7
Numerically simulated permeate flux and experiential data of
alginate fouling in a long RO membrane channel.
xiv
Nomenclature
Nomenclature
c
Solute concentration, mg/L
cf
Organic foulant concentration in the bulk solution, mg/L
cf0
Organic foulant concentration in the feed, mg/L
ci
Molar concentration of the solute, M
cm
Salt concentration at the membrane wall, mg/L
cp
Salt concentration in the permeate, mg/L
cs
Salt concentration in the bulk solution, mg/L
cs0
Salt concentration in the feed, mg/L
D
Diffusion coefficient, m2/s
ds
Diameter of the solute particles, m
H
Channel height, m
i
Serial number of permeate location
J
Permeat flux, m/s
J0
Initial flux, m/s
Js
Solute flux, m/s
Jw
Water flux, m/s
k
Mass transfer coefficient, m/s
kf
Fouling potential, Pa.s/m2
km
Feed water modified fouling potential, Pa.s/m2
ks
Membrane permeability coefficient for solute, m.L/s.mg
kw
Membrane permeability coefficient for water, m/s.Pa
M
Amount of foulant deposited on the membrane, mg/m2
xv
Nomenclature
n
The segment number
R
Gas constant, 8287.7 Pa.L/mole.K
Rc
Cake layer resistance, Pa.s/m
Rec
Cumulative recovery
Rm
Total membrane resistance, Pa.s/m
Rm0
Clean membrane resistance, Pa.s/m
robs
Observed rejection of salt
rs
Specific resistance of the fouling layer, Pa.s.m/mg
rtru
True rejection of salt
t
Time, s
T
Absolute temperature, K
u
Crossflow velocity, m/s
u0
Feed flow velocity, m/s
vp
Permeate flux, m/s
x
Distance from the channel inlet, m
Greek
α
Ratio of organic concentration in bulk solution over organic
concentration in the feed
β
Ratio of salt concentration at the membrane wall
concentration in the feed
γ
Ratio of crossflow velocity over the feed flow velocity
Δc
Solute concentration difference across the membrane, mg/L
Δp
Driving pressure, Pa
Δπ
Osmotic pressure difference across the membrane, Pa
ε
Porosity of the cake layer
θ
Ratio of organic matters fouling the membrane over the total organic
matters transferred to the membrane
over salt
xvi
Nomenclature
σi
Number of ions formed if the solute dissociates
φ
Osmotic pressure coefficient, 75.8 Pa.L/mg.
xvii
Introduction
Chapter 1
Introduction
The scarcity of fresh water has urged the need to seek alternative water sources apart
from traditional water sources such as river and ground water. On the other hand,
discharge of untreated or poorly treated wastewater from domestic and industrial
sources has degraded the water quality of traditional water sources. One of the ways
to resolve water scarcity is to reclaim used water for indirect portable purpose
(Vedavyasan, 2000; Wilf and Alt, 2000). Reverse osmosis (RO) has been widely
accepted as a preferred advanced treatment process to produce high quality water
from microfiltration (MF) or ultrafiltration (UF) pretreated secondary effluent for
water reclamation. However, successful operation of RO process for water
reclamation is accompanied by a number of challenging issues, one of which is
membrane fouling that reduces water productivity and quality, the lifespan of RO
membrane due to frequent chemical cleaning and increases operating cost. More
efforts are made on the study of organic fouling, which is predominant in water
reclamation using RO process.
In the following sections of Chapter 1, the worldwide RO-based water reclamation
industry showing the increasing usage of RO process in water reclamation is briefly
discussed. The emphasis is placed on specifying a few aspects of the problem of
organic fouling, including organic fouling caused by polysaccharide and protein and
chemical cleaning of the fouled membrane. More details on organic fouling are
1
Introduction
reviewed in Chapter 2. Finally, the objectives and scopes of this study are stated at the
end of this chapter.
1.1 Background
The idea of sustainable development has diverted the attention of engineers from the
end point of processes to the beginning point. Traditionally, secondary effluent of
treated wastewater is discharged into rivers, and hence residual contaminants in
secondary effluent are introduced into the environment. However, one of the current
efforts to reduce contamination of the environment is to make use of secondary
effluent from municipal wastewater treatment plants as a potential feed source for
water reclamation to produce potable water. After MF/UF pretreatment or membrane
bioreactor (MBR) treatment, the water is treated with RO for the removal of residual
colloid, organic matters, bacteria and salt. The permeate from RO process has been
used for different purposes, such as ground water recharge in Water Factory 21 in
USA (Wehner, 1992), and boiler feed water in Peterborough Power Station in UK
(Murrer and Latter, 2003). RO-based water reclamation is playing its part
inalleviating the diminishing freshwater supply and meeting the increasing water
demand throughout the world.
In Singapore, RO technology for water reclamation is playing a key role in producing
Newater (Newater is the name given to the reinvented high grade water from
municipal wastewater in Singapore) to secure the country’s water supply. Newater is
either supplied directly to industry as cooling water, boiler feed water, process water,
etc (i.e., direct non-potable use), or to reservoir where it is mixed with natural water
prior to traditional water treatment (i.e., indirect potable use). The first two Newater
2
Introduction
plants were opened in 2003, followed by two more Newater plants in 2004 and 2007.
Altogether, Newater produced by the four plants can meet 15% of Singapore’s water
needs. When the fifth is ready in 2010, Newater will meet 30% of Singapore current
water needs (PUB, 2008).
Compared to other water resources, the benefits of reuse of wastewater as water
source are commonly recognized as (Mujeirigo, 2000):
-
An additional contribution to water resources;
-
A reduction in the disposal of wastewater;
-
A reduction in the pollutant load to surface water;
-
A reduction, postponement, or cancellation of building new drinking water
treatment facilities, with the positive consequence on natural water courses
and water costs;
-
The beneficial use of nutrients (nitrogen and phosphorous) in reclaimed water,
when it is used for agricultural and landscape irrigation (eg., golf courses);
-
A considerably higher reliability and uniformity of the available water flows.
Despite the attractive contribution and the increasing acceptance of RO technology,
separation process via a semi-permeable membrane is plagued by a critical problem,
membrane fouling (Kimura et al., 2004; Lapointe et al., 2005; Mulder, 1996).
Membrane fouling is a result of contaminant or foulant rejected and accumulated on
the membrane surface in a pressure-driven separation process, which leads to the
reduction of water productivity, deterioration of water quality, and shortening of the
membrane lifespan. Since membrane fouling is inevitable in all membrane processes
3
Introduction
including RO process, cleaning becomes an integral part of membrane processes to
mitigate or minimize the fouling.
Effective pretreatment of the RO feed water before entering the RO process is
required to reduce the membrane fouling rate and fouling extent. Many preventive
strategies have been developed for this purpose.
In the early stages of RO
applications, technologies such as MF/UF pretreatment, coagulation and reduction of
alkalinity by pH adjustment were the main treatment steps to control fouling. More
recently, novel methods, such as Fenton process pretreatment (Chiu and James, 2006),
magnetic ion exchange resin (Zhang et al., 2006), ultrasonication (Chen et al., 2006),
and TiO2-UV (Wei et al., 2006) have been investigated and reported to have a good
control of organic fouling with sustained high permeate flux. However, the cost
associated with fouling control and membrane cleaning represents a significant
proportion of the total operating cost. It has been estimated that the pretreatment cost
in RO systems in the Middle East ranged from 10 to 25% of the total operating cost
(Shahalam et al., 2002). Madaeni et al. (2001) reported that the cost of membrane
cleaning represented about 5 to 20% of the operating cost of a RO process.
The success of fouling control is based on a deep understanding of the chemical
components in the RO feed water. Determination of the chemical composition of
fouling substances is indispensable to the development of proper measures and
methods for pretreatment or post treatment (cleaning). The chemical components in
the feed water for water reclamation vary widely; hence, the fouling phenomenon is a
complex issue. Common inorganic particles such as silicate clay and inorganic
colloids and organic matters such as polysaccharides, proteins, nucleic acids and
4
Introduction
humic acids that are present in the natural water or effluent will play important roles
in the fouling process. A significant amount of research has been invested on isolation
and fractionation of feed components (Barker and Stuckey, 1999; Imai et al., 2002.),
but further investigation of the feasibility and fouling potential of each feed
component using membrane filtration is needed (Hu et al., 2003; Jarusutthirak et al.,
2002).
Organic fouling is associated with natural organic matters that are present in surface
water. The widely occurring humic acids in surface water are one of the main organic
components causing organic fouling for RO process (Nystrom et al., 1996; Yoon et al.,
1997). With the advent of wastewater reclamation, the focus of organic fouling is
extended to the dissolved organic matters focused in the secondary effluent (Ang and
Elimelech, 2007; Lee et al., 2006). Soluble microbial products (SMP), especially
polysaccharide and protein, are not only the main cause of organic fouling in
membrane bioreactor (Ng et al., 2006), but also contribute significantly to organic
fouling in RO process (Schneider et al., 2005).
The success of fouling control also requires an understanding of the transfer and
distribution of foulants inside membrane modules, which is an integrated result of the
membrane module configuration and operating conditions. To reduce the overall
capital and operating costs, there is a trend towards operating RO process at a high
water recovery (Rautenbach et al., 2000; Wilf and Klinko, 2001), which resulted in
more particle remaining on the membrane surface and subsequently accelerated
fouling rate. With the advent of recent technology in producing highly permeable and
low fouling membranes, a new phenomenon known as hydraulic imbalance emerges
5
Introduction
(Tay, 2006). In a long pressure vessel containing several highly permeable RO
membrane modules connected in series, the permeate would be mainly produced in
the first few membrane modules, while the contribution from the last few membrane
modules is limited (Song et al., 2003; Wilf, 1997). The imbalance in permeate flux
also results in the heterogeneous distribution of foulants in a long feed side channel
(Chen et al., 2004; Hoek et al., 2008) with the membrane fouling starting from the
inlet end of the membrane channel and gradually proceeding to the rear part of the
channel.
In the RO industry, the silt density index (SDI) and modified fouling index (MFI) are
used for the evaluation of membrane fouling potential of dispersed particulate matters
(suspended and colloidal) in the feed water for RO process (Brauns et al., 2002).
These tests involve filtering feed water through a 0.45 μm microfiltration membrane
at a constant pressure in a dead-end filtration device. These indexes give a satisfaction
limit for feed water acceptable by a RO system, but they do not measure the variation
rate in terms of membrane resistance during the tests. Most fouling studies, if not all,
are conducted using a lab-scale crossflow membrane cell where membrane fouling is
characterized by the permeate flux decline rate (Hong and Elimelech, 1997; Lee et al.,
2005; Tang, et al., 2009). Lab-scale tests are relatively easy and economic in
operation. Therefore, it is feasible to investigate the interactions of various parameters
by a series of filtration tests in lab-scale setup. Tay and Song (2005) developed a
fouling potential indicator obtained in a lab-scale crossflow RO membrane module for
fouling characterization in a full-scale RO process. However, lab-scale tests represent
to some extent only the fouling tendency and fouling rate in full-scale RO process.
Pilot-scale tests are usually conducted for observation of fouling development in a
6
Introduction
full-scale RO process to generate the recommended design parameters for fouling
mitigation and control (Chen et al., 2004). In addition, pilot tests produce the spatial
fouling distribution along the long feed channel other than temporal fouling
development (Hoek et al., 2008; Schneider et al., 2005). However, pilot tests are
usually costly and time-consuming, and only limited operating scenarios can be tested
and evaluated. In both cases of the small lab-scale RO membrane cell or large pilot
RO membrane module, the average permeate flux cannot unveil the hydraulic or
fouling imbalance along the feed channel (Chen et al., 2007; Zhou et al., 2006).
Hence, the information from lab-scale or pilot-scale tests for predicting fouling
development in a full-scale RO process is quite limited and incomplete.
1.2 Problem statement
Fouling control is crucial to the success of RO processes for water treatment. A great
number of both conventional and novel materials, technologies and processes are
available for fouling control. The selection of these materials, technologies and
processes is difficult without the correct characterization of specific feed water
fouling tendency and fouling prediction in a specific RO process. The conventional
SDI and succeeding MFI are widely recognized with limitations to fouling prediction
and simplification of complex interactions (Brauns et al., 2002; Yiantsios et al., 2005).
The most notable limitation is that they are not capable of measuring the fouling
potential of the organic components, which are one of main category of foulants in
RO process. Tay and Song (2005) recently developed a fouling potential index, which
is an inclusive index and is capable of capturing all possible foulants in feed water to
RO process. However, its theory and experimental work were based on colloidal
fouling and are yet to be validated for application in the organic fouling. Therefore,
7
Introduction
extensive efforts are needed to investigate the organic fouling behavior of specific
organic components in order to establish a fouling index.
Although polysaccharide, protein, humic acid, and nucleic acid are the main organic
components in the secondary effluent, polysaccharide and protein, in contrast to
humic acid, are seldom studied in relation to fouling of RO process for water
reclamation. The macromolecules of polysaccharide and protein can be readily
removed by RO membranes, but their fouling behavior in RO process is still relatively
unexplored. Therefore, there is a need to study the roles of polysaccharide and protein
on organic fouling. The study of the fouling behavior of macromolecular biopolymers
is mainly reported with MF/UF applied in the food or pharmaceutical industries
(Maruyama et al., 2001; Simmons et al., 2006). Separate in-depth study of
polysaccharide and protein fouling of RO membranes is essential for the differences
between RO and MF/UF membranes in terms of membrane materials and pore
structure, and different fouling mechanisms that require different appropriate fouling
mitigation measures.
At present, the characterization of polysaccharide and protein fractionated directly
from the secondary effluent is still in its infancy stage. This is probably attributed to
the composition and concentration variability of secondary effluent even from the
same wastewater treatment plant and availability of the difficulty of access to highly
reliable isolation and fractionation technology as well. The fractions of
macromolecules in secondary effluent are therefore mainly limited to hydrophilic and
hydrophobic portions (Hu et al., 2003; Zhao et al., 2010;). Hence, before fractionation
of secondary effluent is fully developed, model polysaccharide and protein are
8
Introduction
currently used for studies even though synthetic solutions made of model components
are not ideal and cannot replace real feed water to RO process for fouling studies.
Synthetic solutions should be made to represent well the characteristics of real feed
water to RO process by including the organic matter properties and solution chemistry.
It is desirable to understand the different roles of organic matters and effects of
inorganic components, which is essential to assess the interactions involved in the
fouling process and to choose the proper fouling control technology.
In addition to the problem of determining the target organic matters, another issue is
the design of the membrane test setup. A commonly used membrane test setup for
fouling study is a small lab-scale crossflow plate-and-frame cell with a small filtration
area of 100 cm2 or less, and a short feed channel about 10 cm. The small dimensions
of the membrane test setup assume homogeneous membrane characteristics and flow
condition. The rate of change of the average permeate flux during the test reflects the
impacts of organic matter on fouling of the RO membrane. However, when the feed
channel becomes longer, like a full-scale RO process, an average permeate flux and
salt rejection cannot unveil the variations of hydraulic properties and parameters and
bulk solution properties along the long membrane module. To compensate for these
inadequacies using a small lab-scale membrane cell, it is necessary to design a long
feed channel to simulate a full-scale RO process to give a better understanding of
fouling development of organic matters.
A pilot-scale RO system is attractive for studying the permeate and fouling variation
along a long RO membrane channel over time. However, such a system does not
allow the RO process to be operated in a manner where inner parameters can be
9
Introduction
directly indentified or measured. As such, most current studies turn to theoretical
modeling to predict fouling behavior along a long channel. Therefore, it is vital to
design a novel RO membrane setup with a long channel for conducting experimental
work to study organic fouling behavior for the validation of theoretical results. Using
a long feed channel membrane setup, instead of a test setup that averages the flux over
the whole channel, varying local permeate flux due to the variations of local
parameters can be determined. With a knowledge of the fouling behavior of the feed
water along a long channel RO membrane, the experimental results obtained can
further improve the existing theoretical work on spatial and temporal development of
organic fouling
1.3 Research objectives
The overall objectives of this project are to systematically study the interactions
involved in the fouling behavior of two specific organic matters and to elucidate the
fouling development in the RO process using both small-scale and long channel RO
membrane setups. To achieve the intended objectives, this study is organized into two
phases as illustrated in Figure 1.1.
In phase I, the main goal is to understand the interactions involved in organic fouling
by model polysaccharide sodium alginate and model protein BSA. Series of
laboratory filtration tests were conducted in a small lab-scale crossflow plate-andframe RO membrane cell. The specific objectives in this phase are:
(1)
to study the fouling behavior of alginate focusing on the interactions
between alginate and calcium;
10
Introduction
(2)
to investigate cleaning efficiency of alginate fouled-membrane by EDTA,
SDS and NaOH;
(3)
to study the fouling behavior of BSA and the effect of pH, ionic strength,
divalent ions, and temperature;
(4)
to investigate the cleaning efficiency of BSA fouled–membrane by EDTA,
SDS and urea.
In phase II, the main goal is to study organic fouling development in a l-m long
channel crossflow RO membrane cell. The same model polysaccharide and protein
were used in this phase. The fouling tendency of feed water obtained in phase I will
be used to explain the experimental results in the long channel membrane cell and as a
basis for the modeling study of alginate fouling development in phase II. The specific
objectives in this phase are:
(1) to study the local permeate fluxes and concentration polarization along the
long RO feed channel;
(2) to study the fouling development of alginate and BSA in a long channel RO
membrane cell;
(3) to investigate the effect of crossflow velocity and initial flux on fouling
development;
(4) to modify previous fouling predictive model and simulate alginate fouling
development in a long channel membrane cell.
The ultimate goal of this study is to provide a preliminary quantitative description of
organic fouling development in a long channel membrane cell. Although basic
membrane transfer models and recently developed fouling models can describe fully
11
Introduction
the fouling mechanisms, the study on organic fouling development in a full-scale RO
process lacks inclusion of important effects such as complex specific and non-specific
interactions and concentration polarization. A more accurate description of organic
fouling development in a long channel membrane cell is meaningful for organic
fouling prediction in a RO process and design optimization.
1.4 Organization of thesis
The subsequent parts of this thesis are divided into the following chapters:
Chapter 2 – Literature review
This chapter presents a comprehensive review of published literature, covering the
organic components in the secondary effluent in water reclamation, membranes and
membrane process, and membrane fouling. Discussions will focus mainly on the key
issues involved in organic fouling.
Chapter 3 – Materials and methods
This chapter describes the small lab-scale RO membrane cell and a long channel RO
membrane cell and their operating conditions used in this study. It also describes in
detail various analytical methods employed in this study.
Chapter 4 – Polysaccharide fouling and chemical cleaning
Fouling tests using model polysaccharide alginate were conducted to study the
polysaccharide fouling behavior along with chemical cleaning efficiency of tests
using three types of cleaning agents. The effects of calcium ions on RO membrane
fouling and cleaning were presented and discussed.
12
Introduction
Chapter 5 – Protein fouling and chemical cleaning
Fouling tests using model protein BSA were carried out to study the protein fouling
behavior along with cleaning efficiency of tests using three types cleaning agents. The
effects of various physical and chemical aspects on the fouling and cleaning were
investigated.
Chapter 6 – Permeate behavior and concentration polarization in a long channel RO
membrane cell
The behavior of local permeate flux and salt rejection in a long channel RO
membrane cell was experimentally investigated using a laboratory-scale 1-m long RO
membrane channel. Concentration polarization modulus (CP) was calculated to
correlate the recovery and concentration polarization. The effect of spacers on
minimizing concentration polarization formation was also investigated.
Chapter 7 – Organic fouling development in a long channel RO membrane cell
Organic fouling tests were conducted to demonstrate the organic fouling development
in a 1-m long RO membrane channel. The effects of operating conditions on fouling
development were discussed. New factors were introduced into the previous fouling
model to predict the fouling development in a full-scale RO process. In addition, the
effect of spacers on the organic fouling was also investigated.
Chapter 8 – Conclusions and recommendations
This chapter summarizes the major conclusions derived from this study. Based on the
experimental findings obtained from this study, recommendations are also provided
for future studies.
13
Introduction
Phase I:
Organic fouling:
Feed properties
Polysaccharide fouling:
Effect of Ca2+
Protein fouling:
Effect of ionic strength, pH,
cationic ions, temperature
Chemical cleaning
(EDTA, SDS, NaOH):
Effect of Ca2+
Chemical cleaning
(EDTA, SDS, urea):
Effect of concentration, pH
and cleaning time
Alginate
BSA
Phase II:
Organic fouling:
Membrane module and operating conditions
Long RO
membrane
channel
Concentration polarization
Fouling development in long
channel:
Effect of crossflow velocity and
initial flux
Figure 1.1 A schematic diagram of the research objectives and scope of this study.
14
Literature review
Chapter 2
Literature Review
Prediction and control of membrane fouling is concerned with identification of the
feed characteristics, designing of the membrane module and process, and operation of
the membrane system. Efforts have been made to investigate the causes and effects of
organic fouling from the beginning to the end of the process. This chapter gives a
literature review in detail and a critical analysis on the topics in three areas:
identification of polysaccharide and protein in the secondary effluent, fundamental
knowledge of the RO membrane and membrane process, and the key issues involved
in organic fouling. The key issues covered include:

The organic contents and the solution chemistry of feed water;

The membrane properties and membrane module configuration;

The operating conditions.
2.1 Dissolved organic matters in water reclamation system
2.1.1 Source of dissolved organic matters
Dissolved organic matters (DOM) present in the secondary effluent are commonly
known as effluent organic matters (EfOM). Many studies have been conducted to
identify and characterize EfOM in the secondary effluent (Hu et al., 2003; Imai et al.,
2002; Rebhume and Manka, 1971). EfOM was reported as of microbial origin, which
mainly consists of a significant amount of soluble microbial products (SMP). SMP is
defined “as the pool of organic compounds that result from substrate metabolism
15
Literature review
(usually with biomass growth) and biomass decay during the complete mineralization
of simple substrates” (Norguera et al., 1994). Thus, some organic compounds such as
humic substance, low molecular weight (hydrophilic) acids, protein, carbohydric acid,
amino acid, and hydrocarbon are commonly found in secondary effluent. An
important review on SMP was published by Barker and Stuckey (1999). The typical
concentration of DOM in secondary effluent was about 10 mg/L in term of dissolved
organic carbon. It was noted that a greater amount of high molecular weight (MW)
compounds were found in secondary effluent than in the corresponding influent and
that SMP has a broad spectrum of MW (<0.5 to >50 kDa). The production of SMP is
affected by several process parameters such as feed strength, hydraulic retention time
(HRT) and sludge retention time (SRT). The complexity of DOM suggests that the
results on DOM during the water treatment process cannot be directly compared
unless their origin and evolution are well understood. The complexity of DOM
therefore made it difficult to understand organic fouling in the subsequent membrane
system.
As DOM components consist of a heterogeneous mixture of complex organic
materials, many techniques have been developed to isolate and fractionate DOM
present in the secondary effluent. These methods include vacuum evaporation,
chemical precipitation, adsorption on XAD resin, and membrane filtration
(Jarusutthirak et al., 2002; Ma et al., 2001; Painter, 1973; Schiener et al., 1998). The
composition and nature of DOM in the secondary effluent have been reported in the
literature. For example, Rebhun and Manka (1971) used ether extraction to classify
40-50% of the organics as humic substances (fulvic acid being the major fraction of
16
Literature review
this class). The remaining organic matters were ether extractables (8.3%), anionic
detergents (13.9%), carbohydrates (11.5%), protein (22.4%), and tannins (1.7%).
However, there are some limitations in the resin fractionation method when they are
applied to organic fouling studies. Firstly, adsorption and desorption occurring on the
resins affect the properties of organic matters. Secondly, the recovery of DOM is not
high and inevitably affects the accuracy of the fractionation results. Thirdly, XAD-8
used in the fractionation method is specially designed to isolate the humic fraction. As
a result, most of the past studies have been focused on fouling caused by humic
substances while very few studies have been conducted on other organic matters.
2.1.2 Dissolved organic matters removal by MF/UF
In water reclamation system, secondary effluent from the secondary clarifier usually
undergoes pretreatment by MF/UF systems before feeding into RO systems or the
MF/UF filtrate from MBR systems instead of the secondary effluent is fed directly
into the RO process. Unfortunately, the MF/UF process is not always effective for
complete removal of DOM that is present in raw water sources due to its large
membrane pore size (Jacangelo et al., 2006). Ognier et al. (2002) used model protein
β-lactoglobulin solution to study the adsorption efficiency of bacterial suspension in
MBR, where the protein molecular weight was 18 kDa while the UF membrane
molecular weight cutoff (MWCO) was 100 kDa, and this allowed the penetration of
protein molecules through the UF membrane and hence the occurrence of the model
protein in the RO feed water. Lin et al. (2001) also reported that the pore size of the
UF membrane affects its permeate quality. The dissolved organic carbon (DOC)
removal efficiencies ranging from 75 to 80% were observed from two small pore17
Literature review
sized membranes, whereas a large pore-sized UF membrane (100 kDa) could only
achieve about 20-30% DOC removal. The remaining DOM therefore will have to be
removed by RO process. Therefore, attention must be paid on the DOM, especially
macromolecules in the RO feed water and its organic fouling potential in RO process.
2.1.3 Model polysaccharides and proteins
Among the wide range of DOM, polysaccharides are some of the most ubiquitous
hydrophilic macromolecules in the secondary effluent (Lee et al., 2006). A major
source of polysaccharides is likely to be the bacterial cell wall, released during the
endogenous phase of microbial growth (Jarusutthirak et al., 2002). Its concentration
could be about 4 mg/L measured as DOC in the treated secondary effluent.
(Jarusutthirak et al., 2002). Alginate, a kind of acidic polysaccharide, is produced by
bacteria, microalgae, or macroalgae (Davis et al., 2003; Lattner et al., 2003; Nunez et
al., 2000). Alginate is typically made up of repeating α-L-guluronic(G) and β-Dmanuronic (M) acids as depicted in Figure 2.1. In one case of bacterial alginate
isolated from mucoid Pseudomonas aeruginosa, the polymeric chain structure
comprises varying proportions of alternating MG-blocks and homopolymer M-blocks,
but lacks mono-G-block structures (Lattner et al., 2003). Another example of alginate
extract from a suite of Sargassum brown algae displays unusual enrichment in
homopolymeric G blocks (Davis et al., 2003). Both of these two conformations of
mannuronate-guluronate MG pair or homopolymeric G blocks demonstrated the
enhanced selectivity for calcium relative to monovalent ions as shown in Figure 2.1d.
18
Literature review
Figure 2.1 Alginate molecular structure: (a) alginate monomers (uronic acids: M vs G.
The carbon atoms C-2 and C-3 of the mannuronate units are partially acetylated (R= H or -COCH3), all C-5 carbon atoms carry a carboxylate group that may be partially
protonated); (b) macromolecular conformation of the alginate polymer; (c) chain
sequences; block copolymer structure; (d) calcium induced gelation of alginate:
schematic representation in accordance with the “egg-box” structure (Davis et al.,
2003).
19
Literature review
In the field of wastewater treatment employing biological treatment processes,
numerous researchers have studied the physicochemical properties of alginate since it
plays an important role in bioflocculation, and thus governs the efficiency of
solid/liquid separation, settling, and dewatering (Bruus et al., 1992; Dignac et al.,
1998; Sainin and Vesilind, 1996). Ye and co-workers used alginate as a model
extracellular polymeric substance to evaluate the fouling contribution of the
polysaccharide component via its rejection, specific cake resistance, and membrane
morphology in a dead-end UF device (Ye et al., 2005; Ye et al., 2006). However, very
few studies have been conducted to address the role of alginate in the fouling of RO
membranes (Lee et al., 2006).
It is difficult to determine the protein component in the supernatant in the secondary
clarifier in biological treatment plant due to the variability of biological treatment
process and the complex nature of protein itself. Protein occupied 22.4 % of the COD
of the total dissolved organic matters in the secondary effluent, higher than the
polysaccharides’ fraction of 11.5% (Rebhum, 1971). Due to the wide usage of
membrane technology in the food and pharmaceutical industries adsorption of a wide
range of proteins such as bovine serum albumin (BSA), human serum albumin(HSA),
immunoglobulin G (IgG), α-lactoalbumin, lysozyme, and β-lactoglobulin(bLG) onto
membrane surface have been investigated on its adsorption on the membrane surface,).
In this study, bovine serum albumin (BSA) was used as the model protein. The choice
of this protein was based on two factors: (i) the BSA properties are perfectly defined
in the literature; and (ii) according to the molecular weigh cutoff of the UF membrane
(up to 100 kDa), its penetration through the membrane pores is possible.
20
Literature review
BSA is a single polypeptide chain consisting of about 583 amino acid residues and no
carbohydrates. BSA contains 35 polar cysteine residues, 34 of which are covalently
linked to form 17 intramolecular disulfide bonds (-S-S-), with the remaining cysteine
residue present as a free thiol (-SH) (Kelly and Zydney, 1994). BSA is a globular
protein and its backbone folds on itself to produce a more or less spherical shape. It is
water soluble and has a compact structure. The BSA structure is determined by a
variety of interactions. Other than covalent peptide bonds determining the primary
structure, other non-covalent stabilizing forces contribute to the most stable structure
of BSA, including hydrogen bonding, hydrophobic interaction, electrostatic attraction,
complexation with a single metal ion such as Ca2+ and K+, and disulfide bonds
(Compbell and Farrell, 2006).
Hydrophobic interaction is a major factor in the
folding of protein into specific three-dimensional structures. The hydrophobic side
groups contribute to the folding of BSA and leave the polar hydrophilic side chain lie
on the exterior of the molecular and accessible to the aqueous environment. The
folding or unfolding of protein causing compaction or expansion of the molecular
structure is affected by heat, high or low extremes of pH of aqueous environment,
exposure to detergents, and urea and guanidine hydrochloride (Bloomfield, 1966).
2.2 Membrane and membrane process system
2.2.1 Membrane definition and process classification
In the book titled Diffusion and Membrane Technology (Tuwiner, 1962), membrane
was defined as “a barrier, usually thin, which separate two fluids. May be intended as
a seal or formulated to be semi-permeable, i.e., permit transfer of some components
and none of others or, at least to possess transfer properties which are selective”.
Although there is no exact definition of a membrane at the microscopic level, the
21
Literature review
above description adequately defines the physical structure and macroscopic function
of a membrane, which is commonly recognized as a selective semi-permeable barrier
between two phases (Aptel and Buckley, 1996; Mulder, 1996).
Membrane can be classified according to different mechanisms of operation, physical
morphology and materials (Aptel and Buckley 1996). In water and waste water
treatment, organic polymeric membranes are the most common type of membranes
used. Among the organic materials, the two most important materials are cellulose
acetate (CA) and polyamide (PA) (AWWA, 1999; Byrne, 1995). CA membranes are
low in cost, have good resistance against chlorine and have very smooth surfaces. CA
membranes are considered an uncharged membrane and are less likely to attract
foulants to the membrane surface. The smooth skin layer of CA membranes also aids
in resisting fouling of CA membrane. However they can only operate within a small
pH range (4 ≤ pH ≤ 7) as they can be hydrolyzed easily. They have low upper
operational temperature limits and do not have good organics rejection properties. PA
membranes, on the other hand, are growing in popularity because they have higher
water flux with slightly better salt rejection and a higher range of operating
temperatures. They reject organics well and resist membrane compaction. Therefore,
they are widely used in RO process. PA membranes used in water purification
industries have a negative charge characteristic which increases the fouling rate of the
PA membrane. Another shortcoming of PA membrane is its sensitivity to chlorine.
Fortunately, a called thin-film structure of PA membranes in spiral wound modules is
cross-liked in its chemical structure, which gives it tolerance to attack by oxidizing
agents. Recently, the advancement of membrane material and surface modification
technology have produced novel RO membranes such as ESPA and LFC membrane
22
Literature review
produced by Hydranautics requiring lower energy and are more resistant to fouling
(Gerard, et al., 1998).
Transport through the membrane takes places when a driving force is applied to the
components of the feed. Table 2.1 lists some examples of driving forces in membrane
processes (Mulder, 1996). In most membrane processes the driving force is a pressure
difference across the membrane. The characteristics of these process is that the
solvent is continuous and driven through the membrane by applied pressure, while the
solutes at relatively low concentration are retained in various extent depending on the
particle size and chemical properties of the solutes and the membrane structure. The
separating mechanisms in the pressure-driven processes are mainly based on sieving
and solution-diffusion as shown in Table 2.2 (Metcalf & Eddy, Inc, 2004).
Table 2.1 Some membrane processes and their driving forces (Mulder, 1996).
Membrane process
Driving force
Microfiltration
Pressure
Ultrafiltration
Pressure
Nanofiltration
Pressure
Reverse Osmosis
Pressure
Pressure Retard Osmosis
Pressure
Gas Separation
Concentration
Pervaporation
Concentration
Dialysis
Concentration
Electrodialysis
Electrically
Membrane Distillation
Thermally
23
Literature review
Table 2.2 Classification of pressure-driven membrane processes (Mulder, 1996).
Membrane
Pressure range
Membrane
Separating
process
(bar)
structure
mechanism
Microfiltration
0.1-2.0
Macropores
Sieving
>50 nm
Mesopores
Ultrafiltration
1.0-5.0
Sieving
2-50 nm
Nanofiltration
Micropores
Sieving +
<2 nm
Solution-diffusion
5.0-20
Dense
Reverse
10-20
Osmosis
Solution-diffusion
<2 nm
2.2.2 Basic membrane transport theory for RO process
There are two general approaches in describing membrane transport models for RO
process (Mulder, 1996). One class is based on a phenomenological approach and nonequilibrium thermodynamics, which treat membrane process as a black-box and
provide no information on how separation actually occurs. The other class is
mechanistic models such as the pore model and solution-diffusion model. Mechanistic
models try to relate separation with structurally-related membrane parameters in an
attempt to describe mixtures. They are useful to know how separation actually occurs
and which factors are important.
Despite various objectives and approaches of different models, they end up with the
unified transport equations. It is commonly accepted that the driving forces for water
24
Literature review
and solute transports are different. Water is primarily driven through the RO
membrane by a hydraulic pressure difference, while solute transport across the
membrane is driven primarily by the concentration difference. The water flux and
solute flux passing through the RO membrane are given respectively by
J w  k w (p   )
J s  k s c
2.1
2.2
where Jw is the water flux (m/s), kw is the membrane permeability coefficient for
water (m/s.Pa), ∆p is the driving pressure (Pa), ∆π is the osmotic pressure difference
across the membrane (Pa), Js is the solute flux (mg/L), ks is the membrane
permeability coefficient for salt (m.L/s.mg), and ∆c is the solute concentration
difference across the membrane (mg/L).
Osmotic pressure can be calculated with van’t Hoff equation if the solution contains
only a single solute (Brandt et al., 1993):
   i ci RT
2.3
where ci is the molar concentration of the solute (mole), σi is number of ions formed if
the solute dissociates(-), R is the gas constant (8287.7 Pa.L/mole.K), and T is absolute
temperature (K). When the solution contains multiple salts, osmosis pressure is
generally estimated with the following empirical expression (Eq. 2.4) with equation
parameters listed in Table 2.3 (Lyster and Cohen, 2007)
25
Literature review
  A c  B c 2  C c 3
2.4
where c here denotes the salt concentration in mM units and Osmotic pressure here in
bar units.
Table 2.3 Empirical relations of the concentration dependence of osmotic pressure for
different salt (Lyster and Cohen, 2007).
Equation parameters
Salt
Aπ
Bπ
Cπ
NaCl
0.04572
-1.797×10-6
4.631×10-9
CaCl2
0.06235
1.314×10-5
8.993×10-9
2.2.3 Concentration polarization
Concentration polarization refers to the build up of a boundary layer of more highly
concentrated solute adjacent to the membrane surface than in the bulk liquid. This
occurs because membranes have the ability to transport one component more readily
than another. The more concentrated solute layer will diffuse back into the bulk
solution, but after a given period of time, steady-state conditions will be established. It
should be noted that concentration polarization is also considered in this study with
fouling being excluded. Compared to fouling, concentration polarization is considered
to be reversible (Sablani et al., 2001). If the concentrated solute contains sparingly
soluble salts, scaling will occur when their concentration in the boundary layer
exceeds their solubility limits. At such higher concentrations, colloidal materials may
agglomerate and foul the membrane surface (Brandt et al., 1993). Concentration
polarization increases the osmotic pressure due to concentration buildup at the
26
Literature review
membrane surface, which causes a reduction in water flux and an increase in salt
transport across the membrane.
Most analyses of concentration polarization have employed the simple stagnant film
model originally presented by Michaels (1968). The model assumed a stagnant film
where mass transfer reaches to a one-dimensional steady-state mass balance, that is,
the convective transport of solute (salt in this thesis) to the membrane is equal to the
sum of the permeate flow plus the diffusive back transport of the solute, i.e.,
v pc  D
dc
 vpcp
dx
2.5
where D is the diffusion coefficient for solute transport through the membrane (m2/s),
vp is the permeate flux (m/s), c is the concentration of membrane-retained salt (mg/L),
cp is the salt concentration in the permeate (mg/L), and dc/dx is the salt concentration
gradient along the distance away from the membrane surface (mg/L.m). The above
equation is integrated over the boundary layer thickness to yield:
v p  k ln(
cm  c p
cs  c p
)
2.6
where cm is the salt concentration at the membrane wall (mg/L), cs is the salt
concentration in the bulk solution (mg/L), and k is mass transfer coefficient (m/s).
This allowed the calculation of cm when k is known. If the rejection is 100%, cm/cs can
be obtained from Eq. 2.6
27
Literature review
vp
cm
 exp( )
cs
k
2.7
which is known as the concentration polarization modulus (CP) (Mulder, 1996).
Therefore, estimating mass transfer coefficient k in RO process is very important, but
very difficult because the theoretical expression k is borrowed from mass or heat
transfer from non-porous smooth duct flow and yet to be proven to be adopted in
membrane separation. Three approaches to estimate k are employed: direct
measurements using optical or microelectrode measurements, indirect measurements,
in which the true rejection is calculated by extrapolation to infinite feed circulation,
and indirect measurements, in which a concentration polarization combined with a
measurement transport model is used for mass transfer coefficient calculation (Murthy
and Gupta, 1997).
According to Eq. 2.7, two factors: permeate flux vp and mass transfer coefficient k are
responsible for concentration polarization. In order to reduce concentration
polarization, the feed channel spacer is designed as a turbulence promoter to enhance
mass transfer near the membrane surface and mitigate concentration polarization
(Ahmad et al., 2005; Geralds, 2002). Song and his research group demonstrated that
vertical concentration polarization effect on filtration performance was less critical
with inserted spacer than that of the salt accumulation along the membrane channel
(longitudinal concentration polarization or concentration factor) by comparing
experimental average permeate flux with numerical simulation of the same system
(Zhou et al., 2006). They suggested that the designed spacer must provide hydraulic
dispersion of at least 10 times of the molecular diffusion coefficient of sodium
28
Literature review
chloride to achieve complete depolarization in full-scale spiral wound modules. On
the other hand, horizontal pressure loss will be increased by inserting spacer along the
membrane channel, which will inevitably reduce water productivity (Dacosta, 1993;
Zhou et al., 2006). Hence, the effectiveness of feed spacers to reduce concentration
polarization in RO process remains in dispute and optimization in the use of feed
spacers is required.
2.2.4 Spiral wound membrane module and the permeate flux behavior
Larger membrane areas per unit volume is typically preferred for full-scale
application of membranes. The smallest unit into which a membrane area is packed is
called a module. The module is the central part of a membrane installation.
Commercially available modules include spiral wound, hollow fiber, tubular and
plate-and-frame modules. Amongst these, spiral wound modules are often preferred
for RO process in the application of water reclamation because they offer a good
balance between ease of operation, fouling control, permeation rate and packing
density (Byrne, 1995; Schwinge et al., 2004). However, major problems for a spiral
wound module are concentration polarization, fouling and high pressure loss
(Schwinge et al., 2004).
A schematic diagram of a spiral wound module is shown in Figure 2.2. The
components of a spiral wound module are the membrane, feed and permeate channel,
spacers which keep the membrane leaves apart, permeate carrier, permeate tube and
membrane housing. As the next logic step from a flat membrane, the spiral wound
module actually is a plate-and-frame system by rolling several sheets of flat RO
membrane around the central permeates tube. The feed solution flows in an axial
29
Literature review
direction parallel to the permeate tube through the feed channel. Water passes through
the membrane and is collected as permeate in the permeate tube. The concentrate
flows out from the other side of the feed channel in the membrane module. In a fullscale process, several spiral wound membrane modules are laid in series in a pressure
vessel to form a long membrane channel. A number of these pressure vessels are
connected or arranged in parallel or in series to form a single-stage or multi-stage
process according to the recovery requirement.
Permeate carrier
Figure 2.2 A schematic of a spiral wound module showing the flow directions, feed
and permeate channels including spacers.
Continuous permeation of water and rejection of salt by the RO membrane result in
the variation of process parameters along the feed channel in a full-scale process,
which include decreasing axial velocity, increasing bulk salt concentration and
increasing concentration polarization. Therefore, the identification of flow
hydrodynamic, i.e., local flow field, is a precondition for the prediction of the salt
30
Literature review
concentration variation and membrane performance, and optimization of spiral wound
modules, or, generally, a long membrane channel.
Most theoretical studies are based on the classic film theory model to deal with flow
inside a membrane channel (Avlonitis et al., 1993; Gupta, 1992; Zydney, 1997). It
simplified the membrane filtration channel as a homogenous unit, that is, the salt
concentration and the permeate flux remains invariant along the feed channel, which
underestimates the increase in the thickness of the boundary concentration buildup
layer in a long membrane channel (Song et al., 2003). A solution is to separate the
long membrane channel into finite elements and then apply the film theory to each
element. Numerical calculation of finite elements can present axial variation of
permeate flux and concentration accumulation in a full-scale system (Bhattacharyya et
al., 1990; Hoek et al., 2008; Lyster and Cohen, 2007). These studies confirmed that
the permeate flux decreased and solute concentration increased along the channel due
to the effect of concentration polarization. The problem of hydraulic imbalance arises
when high specific flux from the first few membrane modules increases the salt
concentration dramatically downstream along the membrane channel, which can lead
to high permeate flux being produced in the first few membrane modules, while little
or no flux is produced in the last few membrane modules in a pressure vessel (Wilf,
1997; Nemeth, 1998). Therefore, the channel length should be optimally controlled to
fully utilize the membrane system.
In addition, the behavior of concentration
accumulation is affected by the feed properties and operating conditions (de Pinho et
al., 2002).
Apart from permeate flux, rejection coefficient decreases along the
membrane channel as a result of salt concentration buildup in the boundary layer
(Bouchard et al., 1994; Geraldes et al., 2001).
31
Literature review
Although numerical simulation can be used to predict the RO process performance, it
is still subjected to the verification of experimental work. The work based on
experimental tests use physical models of spiral wound modules in the laboratory that
consist of rectangular thin channels-slits-filled with spacer (Geralds et al., 2002). The
measurements in this system have been restricted to average permeate quality and
quantity analysis over the whole channel and cannot demonstrate the heterogeneity
inside the feed channel of the spiral wound module. An insight on the local mass
transfer inside the feed channel of spiral wound module is important and therefore,
there is a need to develop new testing membrane setups that can monitor permeate
flux and quality along a long RO membrane channel.
Song and his research group adopted a pilot-scale RO setup consisting of four
membrane modules in series and showed that the global recovery was limited by
thermodynamics restriction due to the increasing osmotic pressure in the feed side of
the membrane channel (Song and Tay, 2006; Zhou et al., 2006). However, this test
failed to present the data of local variables with the channel length as calculated in the
modeling work. Chen et al. (2007) investigated the uneven axial distribution of
protein deposition on the membrane surface and claimed the feed spacer could abate
such heterogeneity. Nevertheless, their work only recorded the average permeate flux
for a 50-cm channel, unveiling the information of the permeate flux variation along
the channel. The results from another experimental study of 20-cm long nanofiltration
cell were in good agreement with their simulation work (Geraldes, 2001), whereas the
experimental data were only collected from the inlet channel of 6 cm long, still a big
gap away from a long channel of spiral wound membrane modules. Bu-Ali et al.
(2007) simulated a long channel up to 230 cm by a set of five membrane modules in
32
Literature review
series and collected permeate flux at the end of each membrane module. This work
made good contribution to experimental studies through monitoring the permeate flux
along the channel although the channel was not continuous. However, no
concentration polarization was considered in this study. Therefore, a new RO
membrane test setup characterized by a long membrane channel is needed to
investigate the variation of the local flow, concentration and pressure along the
membrane channel.
2.3 Membrane fouling
2.3.1 Definition and types of membrane fouling in RO process
In practice, a continuous decline of permeate flux can often be observed in RO
process. Flux decline is manifested by four major factors: physical compaction under
high hydraulic pressure, membrane structure deformation by specific chemicals,
concentration polarization, and membrane fouling (Hoek et al., 2008; Mulder, 1996)..
Owing to its capability to reject most of the particles, colloids, emulsion, suspension,
macromolecules, and salt, the RO membrane is susceptible to the deposition of these
rejected substances, which are named foulant in the fouling process. Generally,
membrane fouling refers to the attachment, accumulation, or adsorption of foulants
onto the membrane surface and/or within the membrane, resulting in an extra
resistance to flux and a decline of permeate flux (Zhu and Elimelech, 1995). In RO
process, membrane fouling is primarily ascribed to fouling layer formation on the
membrane surface and seldom to pore plugging due to the “dense” nature of RO
membranes. Therefore, in this study, membrane fouling in RO process is defined as
the decline in permeate flux caused by the formation of a fouling layer on the
33
Literature review
membrane surface that excludes the contribution of physical compaction, chemical
damage of membrane structure, and concentration polarization.
In RO process, the common types of fouling are colloidal (Vrijenhoek et al., 2001),
organic (Lee et al., 2006), biological (Tasaka et al., 1994; Vrouwenvelder et al., 2009)
and scaling (Le Gouellec and Elimelech, 2002). More often, membrane fouling
involves a combination of different fouling types depending on the constituents in the
feed water (Lee et al., 2005; Schneider et al., 2005).
Colloids are ubiquitous in natural and process waters. Examples include clays,
colloidal silica, iron oxyhydroxide, organic colloids and suspended matters, and
calcium carbonate precipitates (Elimelech et al., 1997). The size of these particles
ranges from a few nanometers to a few micrometers. These colloidal particles are
considered to be the principal cause of membrane fouling (Cohen and Probstein,
1986). The gel layer model describes colloids accumulating on the membrane surface,
which form a cake layer that increases the resistance to water flow through the
membrane, whereas according to the cake-enhanced osmosis pressure model, the
extra resistance of the cake layer is negligible and the enhanced osmosis pressure is
the main cause which reduces the net driving force on water.
Organic fouling is one of the main fouling in water treatment and reclamation plants.
As stated in Section 2.1, the main cause of organic fouling of RO membranes is traced
to the dissolved organic matters in secondary effluent or MBR effluent. Organic
fouling has a greater effect in membrane processes than colloidal fouling because
organic materials can form a more compact fouling layer at favorable solution
34
Literature review
chemistry (Hong and Elimelech, 1997). In addition, organic matters in the feed water
to RO process provide nutrients for microorganisms and promot biofouling (Herzberg,
2009).
Biological fouling or biofouling begins with the attachment of microorganisms to the
membrane surface mediated by the interaction between cells and membranes. The
attached microbes are capable of utilizing organic matters from the feed water or
organics rejected by membranes to reproduce cell and form an extracellular biological
film or biofilm that increases the membrane resistance (Schneider et al., 2005). It is
also noted that the biofilm would enhance the osmotic pressure, leading to less net
driving force across the membrane (Hoek and Elimelech, 2003). Biofouling is an
important potential factor that causes RO membrane performance decline and
accounts for over half of the performance decline factors for an ultrapure water
system (Tasaka et al., 1994).
Scaling refers to fouling caused by sparingly insoluble salts. As the salt concentration
in the feed water increases downstream due to the loss of water through permeation,
dissolved inorganic components such as Ca2+, Mg2+, CO32-, SO42-, PO43-, silica and
iron can precipitate as insoluble salts (or scales) on the membrane surface if the
solubility limits are exceeded.
2.3.2 Fouling mechanisms in RO process
The phenomenon of fouling is very complex and difficult to be described theoretically.
However, reliable values of flux decline are necessary for process design. A
resistances-in-series model can describe the flux behavior (Mulder, 1996). In this
35
Literature review
model, foulant is deposited on the membrane surface to form a cake layer and the
cake layer resistance is in series with the membrane resistance. The flux can be
described by
vp 
p
R m 0  Rc
2.8
where Rm0 is the clean membrane resistance (Pa.s/m), and Rc is the resistance of the
cake layer (Pa.s/m). The total cake layer resistance Rc is equal to the specific
resistance of the cake rs (Pa.s/m2) multiplied by the cake layer thickness (m). The
specific cake resistance rs is often expressed by the Kozeny-Carman relationship:
rs  180
(1   ) 2
(d s ) 2  3
2.9
where ds is the “diameter” of the solute particles (m), and ε is the porosity of the cake
layer(-).
The difficulty in applying resistance-in-series is to the estimation of the specific
resistance and thickness of the cake layer due to the complexity of organic matters.
Even for a given organic matter, compactness and thickness of the cake layer will
depend on many physical and chemical factors such as concentration, temperature, pH,
ionic strength and specific interactions. In a modeling study on fouling of a full-scale
RO process, an empirical parameter was successfully used for the first time to
describe the foulant sticking efficiency in membrane separation to calculate the cake
layer thickness (Hoek et al., 2008). This parameter characterized the stabilizing effect
36
Literature review
of electrostatic double layer interactions in particles aggregation, which is a very
important interaction for fouling in membrane separation. However, the specific
resistance rs was calculated based on the maximum packing of particles, which is
applicable for mono-disperse of spherical colloids but not sufficient for organic
fouling (Hong et al., 1997).
Despite the popularity of resistance-in-series model in fouling studies, another fouling
mechanism via a cake- and biofilm-enhanced osmotic pressure was reported to
contribute more to the flux decline in RO process than the increased hydraulic
resistance to permeate flux (Herzberg and Elimelech, 2007; Hoek and Elimelech,
2003). The hindered back diffusion of salt ions within the deposit layer results in
elevated salt concentration and, thus, enhanced trans-membrane osmotic pressure near
the membrane surface. The latter leads to a substantial decrease in the net driving
pressure across the membrane and, consequently, leading to a more severe flux
decline. However, no such enhanced osmotic pressure was reported in the case of
organic fouling (Lee et al., 2005). Therefore, it can be assumed that the formation of
organic foulant cake layer has insignificant influence on the trans-membrane osmotic
pressure.
2.3.3 Key issues in organic fouling
Despite a significant amount of research efforts made and numerous papers published
on membrane fouling, a comprehensive study of polysaccharide and/or protein
fouling in different RO testing cells is not available. However, significant progress
has been made over the years to identify the important parameters governing the
fouling behavior of organic matters. Theses parameters include (i) the properties of
37
Literature review
the feed water (foulant concentration and polarity, pH, ionic strength, ion type and
temperature) (Bellona et al., 2004; Kelly and Zedney, 1994; Tang et al., 2007); (ii)
membrane
properties
and
membrane
module
configuration
(morphology,
hydrophobicity, surface charge, channel dimension, and spacers) (Combe et al., 1999;
Davis, 1992; Koutsou et al., 2009; Li et al., 2007); and (iii) operating conditions
(applied pressure and crossflow velocity) (Tang et al., 2007).
2.3.3.1 The feed water
One nucleic element of organic fouling is the characteristics of the organic matters.
Macromolecular organic matters have a broad range of chemical groups which
interact with other organic molecules and membrane through van der Waals forces,
electrostatic interaction, hydrophobic interaction, and/or hydrogen bonding (Wilbert
et al., 1998). Therefore, the foulant size and structure intensively influence the ability
of foulant approaching other organic molecules and the membrane surface. A higher
foulant concentration is easily expected to pose a greater fouling problem due to the
presence of a larger pool of foulants. If more than one type of organic foulant exists in
the feed solution, fouling will become more complex and severe due to additional
interactions between the different organic molecules (Chen et al., 2007). In the case of
protein as a foulant, a particular interaction exists in protein fouling. Structurally soft
protein with low internal structure stability generally tends to adsorb on all surface
irrespective of electrostatic interactions, owing to a gain in conformational entropy
resulting from adsorption (Bos et al., 1994).
One physical aspect of the feed water which is also an important operating factor is
temperature which, surprisingly, is often not included in the study of organic fouling
38
Literature review
of RO membranes. Changes in temperature can strongly affect membrane fouling and
filtration performance of RO membranes, depending on the type of foulant, feed water
and operating conditions. In membrane filtration of protein solution, there are
conflicting requirements on temperature to enhance the filtration performance of
filtration process. It is noticed that a higher temperature was physicochemically
favorable to membrane filtration due to the reduction of feed viscosity and
concentration polarization arising from the increase in salt diffusivities and diffusion
coefficients of macromolecules, which ultimately led to the increase in permeate flux
(Campbel et al., 1993; Wang et al., 2005). However, on the other hand, higher
temperature could also trigger denaturation and expose protein’s inner hydrophobic
regions to the surrounding, which increased the tendency of aggregation or adsorption
to the membrane surface and increase the rate of membrane fouling (Kelly and
Zedney, 1994; Meireles et al., 1991; Simmons et al., 2007).
Among the various parameters of solution chemistry, the most important are ion types,
ionic strength, and pH. The effect of solution chemistry on organic fouling has been
investigated extensively. For example, a higher ionic strength can lead to a greater
organic fouling problem by altering the surface charge of organic matters and
membrane surface and reducing the repulsion interaction between the organic matters
and membrane surface (Hong and Elimelech 1997; Lee et al., 2006; Tang et al., 2007).
pH of the solution has different effects depending on the organic matters. Generally,
low pH enhances the accumulation of polysaccharides and humic acid on the
membrane surface, but higher fouling rate of protein was observed at pH near to the
iso-electric point of protein (Ang and Elimelech, 2007; Hong and Elimelech, 1997;
Lee et al., 2006). The presence of divalent cation ions will enhance the binding of
39
Literature review
organic matters leading to a compact fouling layer (Frank and Belfort, 2003; Grant et
al., 1973; Latter et al., 2003, Xue et al., 2009). These qualitative results indicated that
adjustment of the solution chemistry can control the interactions between organic
matters and membrane surface for better fouling control. However, it is not easy to
quantitatively describe the effect of feed solution chemistry, especially accompanied
by concentration polarization. If concentration polarization occurs severely, the feed
characteristics cannot represent the true condition near the membrane surface inside
the membrane module depending on the membrane module configuration and
operating conditions.
Song and co-workers developed a collective index kf termed as the fouling potential to
indicate fouling strength of feed water (Chen et al., 2004; Tay and Song, 2005). This
fouling potential kf is the product of the specific resistance rs of the fouling layer and
particles concentration in feed cf0.
k f  rs c f 0
2.10
Therefore, kf is an intrinsic indicator of the feed water and not dependent on the
operating system and conditions. A small piece of RO membrane and a single labscale RO membrane cell can determine kf experimentally without any prior
knowledge concerning the nature of foulants and the associated fouling mechanisms.
If the clean membrane resistance and the flux are known, kf can be calculated
according to Eq. 2.11:
40
Literature review
kf 
Rm (t )  Rm 0
t
v
p
2.11
(t )dt
0
where Rm(t) is the total membrane resistance at time t. It can be used to quantitatively
predict fouling development in a full-scale RO process. Compared to the previous
fouling indices SDI and MFI, this fouling potential is more accurate and effective to
quantitatively predict the fouling in RO process.
2.3.3.2 Membrane properties
Membrane is the nucleic element in membrane processes. Recent studies
characterized the RO membrane with key membrane surface properties (surface
morphology, charge, and hydrophobicity) and correlated these properties to the
fouling behavior (Fu et al., 2008; Li et al., 2007; Vrijenhoek et al., 2001). In general,
surfaces that are most likely to resist organic fouling are characterized as being
hydrophilic, H-bond acceptors, non-H-bond donors, and neutrally charged surface.
Rougher and more hydrophobic membrane has a higher tendency for fouling by
natural organic matters (Park et al., 2005). According to such interactions as
hydrophobic and electrostatic interactions between organic matters and membranes, a
variety of surface modification strategies to prevent fouling have been examined and
reported. They are i) increasing hydrophilicity of the membrane to provide more
opportunity for water, rather than organic matters, to chemically associate with the
membrane; ii) introducing of negative charges to increase the electrostatic repulsion
between the membrane and the negative-charged organic foulant; and iii) creating
steric hindrance to exclude the organic foulant from the immediate vicinity of the
membrane (Abu Tarboush et al., 2008; Wilbert et al., 1998).
41
Literature review
However, some conflicting organic fouling behavior has been noted. Usually protein
could cause more severe fouling on hydrophobic than hydrophilic membrane surface
due to hydrophobic effect (Matthiasson, 1983), but substantial protein fouling on
hydrophilic membrane was observed in another study (Mueller and Davis, 1996). This
common conflicting observation in organic fouling is possibly attributed to the
difficulty in determining the predominant interactions for different proteins,
membrane and operating conditions used by different researchers.
2.3.3.3 Membrane module configuration
Among the important variables in the membrane module configuration are height and
length of the feed channel. The channel height and feed flow rate, resulting in
different tangential crossflow velocities, determine the shear rate at the membrane
surface in the crossflow membrane module. A high shear rate at the membrane
surface reduces both concentration polarization and cake build-up (Davis, 1992;
Jaffrin, 2008). Another effect of increased channel height is the reduction of the total
membrane filtration area in a membrane module.
The foulant deposition does not only evolve with time, but also distribute spatially
along the feed channel. For example, heterogeneous development of foulant
deposition along the channel was observed for a two-stage RO train in a pilot water
reclamation plant (Schneider et al., 2005). The proportion of organic matters in the
fouling layer increased in the first stage from approximately 30% in the first two
elements to 55% in element 4–6, and remained constant for subsequent elements in
stage 2. On the contrary, the protein content in the organic fraction decreased linearly
from around 70% in element 1, stage 1 to about 40% in element 6, stage 1 and
42
Literature review
remained constant for subsequent elements in stage 2. The spatial distribution of the
fouling deposition may be ascribed to the local variation of flow, pressure, and
concentration along the feed channel throughout the whole system (Schwinge et al.,
2004; Song and Tay, 2006). A direct result of foulant deposition developing from the
inlet elements to the outlet elements is “flux leveling”, which refers to the permeate
flux shifting from the inlet towards the outlet to maintain a constant global water
production as fouling progresses (Chen et al., 2004; Hoek et al., 2008). Therefore, the
channel length is a critical factor to predict the temporal and spatial fouling
development.
However, it is noted that it is not always true that organic deposition developed from
the inlet towards the outlet. In the above mentioned two-stage RO process, total
organic deposition was more severe in the latter elements than the former elements
(Schneider et al., 2005), which suggests that the deposition of certain organic matters
was faster downstream than upstream along the feed channel, in opposition to the
protein fouling development. These phenomena were not taken into consideration in
recently published models of full-scale RO process (Chen, et al., 2004; Hoek et al.,
2008). The possible reason is that these modeling studies did not take into account the
organic matters, which involves more complex physicochemical interactions.
Therefore, due to the lack of experimental data evidence of the local permeate flux
and analysis of the organic foulant deposition in the controlled lab-scale setup and
pilot-scale plant, the nature of the organic fouling development along the feed channel
is not fully understood.
43
Literature review
Spacers are an integrated part of membrane module and have been widely
acknowledged that they can produce a favorable hydrodynamic condition to suppress
the formation of concentration polarization and foulant deposition on the membrane
surface (Ahmad et al., 2005; Chen et al., 2007; Geraldes, 2001). However, the shear
field is homogeneous and deposition may be more severe near the points where the
spacer contacts the membrane, yielding a deposit distribution pattern similar to the
spacer geometric pattern (Yiatsios et al., 2005).
Computational fluid dynamics (CFD) simulation and modeling are usually used to
study the flow profile in a spacer-filled channel. It is a challenge to incorporate the
permeation performance of an RO process in CFD simulation. Experimental work
also has been done to observe the effects of spacer on the RO filtration performance.
An experimental study of solution deposition conducted by Chen et al. (2007)
investigated the uneven axial distribution of deposition on the membrane surface and
claimed the feed spacer could abate such heterogeneity. Nevertheless, their work only
recorded the average permeate flux for a 50-cm channel, unveiling the information of
the local permeate flux variation along the channel.
Spacers also display some
negative effects on RO process due to reduction of effective driving force. It has been
shown that spacers can generate certain degree of pressure drop in the narrow
membrane channel due to friction (Zhou et al. 2006). On the other hand, inserting of
spacer between membranes will reduce the effective filtration area, causing less water
production, which has not been investigated in depth.
44
Literature review
2.3.3.4 Operating conditions
Foulants in the bulk solution in the RO membrane feed channel is susceptible to a
convective flow perpendicular to the membrane surface (permeate flux) and a
tangential shear flow parallel to the membrane surface (crossflow) (Hong and
Elimelech, 1997). Foulants are transported to the membrane surface by the permeate
flow and swept away from the membrane surface by the crossflow. As permeate is
primarily driven by the driving pressure, an increase in driving pressure increases the
accumulation of foulant on the membrane surface as well as the initial permeate flux
before fouling occurs. Tang and Leckie (2007) proposed a concept of a limiting flux –
a maximum pseudo-stable flux, beyond which further increase in applied pressure did
not increase the stable flux. If the initial flux was below the limiting value, little
foulant deposition was observed. The driving pressure also affects the porosity of the
fouling layer depending on the compressibility of the organic fouling layer. At a
higher driving pressure, a more compact fouling layer produces higher resistance and
reduces the permeate flux. Previous studies shows higher membrane flux not only
impose greater hydrodynamic drag on humic acid molecules towards membrane
surfaces, but also increase the humic acid and salt concentrations at the membrane
surface through concentration polarization (Goosen et al., 2004). Although flux
reduction directly due to concentration polarization of the foulant is not significant,
concentration polarization can contribute to flux reduction indirectly due to elevated
salt concentration near the membrane surface (Tang et al., 2007).
Increasing crossflow velocity will lessen the organic fouling because higher crossflow
velocity increases the shear rate, thus reducing the rate of extent of organic fouling
(Seidel and Elimelech, 2002). Higher crossflow velocity also reduces the salt
45
Literature review
concentration on the membrane surface (Tang et al., 2007), which could abate the flux
reduction. However, one study reported that the fouling rate didn’t depend on the
crossflow velocity in the initial stage of filtration (Lee et al., 2006). Therefore, for a
constant-pressure RO process, careful choice of operating pressure (or initial flux) and
crossflow velocity must be exercised to control fouling.
2.3.4 Membrane cleaning
Membrane cleaning is one of the methods adopted to alleviate membrane fouling and
is always employed in practice. Compared to the other methods such as pretreatment
and membrane modification, membrane cleaning is more direct and quick to restore
the permeability of RO membranes. The cleaning methods currently being used can
be distinguished as physical cleaning and chemical cleaning. In a full-scale RO
process where spiral-wound membrane modules are the predominant membrane
configuration, membrane cleaning is commonly done with chemicals (Madaeni et al.,
2001). Because organic fouling is strengthened and accelerated by the various
interactions, more effective counter measures are necessary to reduce the interactions
between foulants and membranes. Membrane chemical cleaning occurs through
chemical reactions between cleaning chemicals and membranes or organic foulants to
remove the fouling layer partially or completely. In general, chemical cleaning can be
broadly classified based on cleaning agents into six categories: acids (such as citric
acid, HCl), alkaline (such as NaOH), chelating agents (such as EDTA, polyacrylates),
surfactants (such as SDS), enzymes (such as proteases, amylases), and combination of
multi-agents. Acid and base can create a suitable pH environment (down to 2 and up
to 11.5) for organic foulant hydrolysis or affinity reduction (Kuzmenko et al., 2005).
Chelating agents can combine with metal ions which link organic matters in the
46
Literature review
fouling layer (Ang et al., 2006). As a result, the fouling layer becomes loose and is
removed. Enzymes play a vital role in scissoring specific point in the proteinaceous
foulants and surfactants can rapidly solubilize the loose foulant fragments (MuñozAguado et al., 1996).
However, the general chemical reaction mechanism cannot adequately explain the
cleaning efficiency observed. On one hand, the bindings between the foulants or
between membranes and foulants vary with the molecular weight, functional groups,
surface charge and the nature of membranes. It is obvious that different forces are
needed to break these bindings. On the other hand, cleaning efficiency is affected by
many chemical and physical parameters other than the type of cleaning agents, which
includes pH, concentration, time, temperature and cleaning protocol. The relationship
between cleaning efficiency and a single parameter is often non-linear as membrane
cleaning is a complex function of multiple parameters. It is not surprising to see that
many past studies demonstrated different chemical cleaning performance under the
same cleaning conditions (Farinas et al., 1987; Li et al., 2005; Madaeni and
Mansourpanah, 2004; Muñoz-Aguado et al., 1996). These observations indicated that
the study on membrane cleaning is still very much in its infancy stage. Thus, there is a
need to conduct further research on membrane fouling and to develop effective
cleaning protocol as well as to formulate effective cleaning agents.
As chemical cleaning could remove the foulant deposited on membrane, it could
induce undesirable effects such as the damage of membrane structure due to improper
selection of chemical agents and operating conditions. Kuzmenko et al. (2005)
reported the formation of holes in the UF membrane skin layer caused by forced
47
Literature review
cleaning conditions. Thus, a higher dosage of cleaning agents that resulted in a
complete restoration to the initial permeate flux may eventually lead to more severe
problems that in turn require a more frequent clean-in-place operation and a shortened
membrane life. Depending on the cleaning chemicals and membrane, the membrane
surface may be affected and the binding of organic matters with membranes in the
subsequent runs may change after chemical cleaning (Kuzmenko et al., 2005;
Lawrence et al., 2006). A possible consequence was that the subsequent flux was not
easily restored due to the enhanced binding between the foulants and membranes after
chemical cleaning. On the contrary, a study reported that if flux was not totally
restored by cleaning, the subsequent additional fouling layer was easier to remove
(Muñoz-Aguado et al., 1996).
Therefore, a successful cleaning operation is judged by not only the flux recovery but
also the performance of subsequent operation to facilitate having a long membrane
life-span and high operation efficiency of the membrane system. In order to achieve
the best cleaning performance, most of the cleaning studies were conducted under
extreme conditions, such as high pH (higher than what a membrane can typically bear)
(Ang et al., 2006). As a result, the findings of these cleaning studies may not be
useful for long-term operation of a membrane. Therefore, it is essential that a
membrane cleaning studies should: (i) choose a typical range of conditions (chemical
types, concentration, and pH) for evaluating the cleaning performance; and (ii) adjust
physical variables to relieve the stress on the membrane system induced by extreme
chemical conditions. The above discussion clearly indicate the needs for conducting
further research on membrane cleaning with the aim of obtaining a better
48
Literature review
understanding on cleaning mechanism and to formulate effective cleaning agent and
cleaning protocol.
49
Materials and methods
Chapter 3
Materials and Methods
3.1 Chemical solution preparation
3.1.1 Model organic matters
The model polysaccharide used in this study is a commercial high grade sodium
alginate (Product No. A2158, Sigma-Aldrich, Saint Louis, MO) that was delivered in
powder form. From the manufacturer’s specifications, alginate has molecular weight
ranging from 12 to 80 kDa, and a low viscosity of approximately 250 cP for a 2%
solution at 25oC. Stock solution containing 5 g/L of sodium alginate was prepared by
dissolving the sodium alginate powder in 18.2-MΩ deionized (DI) water (Milli-Q,
Millipore, Billerica, MA) 12 h before the tests.
Commercially available biopolymer bovine serum albumin (BSA) (Sigma-Aldrich,
Saint Louis, MO) was used as the model protein in this study. The molecular weight
of the commercial BSA was specified to be about 66 kDa by the manufacturer. BSA
was reported to have an isoelectric point (IEP) at pH 4.7-4.9 (Salgın et al., 2006).
Stock solution of BSA was freshly prepared in DI water to the required concentration
prior to each experiment.
3.1.2 Other chemicals
Reagent grade chloride salts of sodium, calcium and magnesium (Merck, Darmstadt,
Germany) and hydrochloric acid and sodium hydroxide were used to prepare different
50
Materials and methods
electrolyte solution with different concentration and pH. Sodium hydroxide,
ethylenediamine tetraacetic acid (EDTA, Goodrich Chemical Enterprise, Bangpoo,
Thailand), sodium dodecyl sulfate (SDS, Fluka, Saint Louis, MO), and urea (SigmaAldrich, Saint Louis, MO) were used for cleaning the RO membranes. Except for urea,
these chemicals are commonly used in commercial RO process using organic
membranes (Trägårdh, 1998). EDTA is a common metal chelating agent in many
chemical reactions while SDS is an anionic surfactant that is used in many household
products. Urea is not a typical cleaning agent in membrane processes, but has specific
reactions with protein and is relatively cheap. Chemical cleaning solutions were
freshly prepared with DI water just before the experiments.
3.2 RO membranes
Hydrophilic polyamide membrane LFC1 and ESPA2 from Hydranautics Inc.
(Oceanside, CA) were used in phase I and II, respectively. As required by the
membrane filtration cells, the flat sheet membrane was cut into coupons with an
effective surface area of 2.9×10-3 m2 in phase I or long strips with an effective surface
area of 0.03 m2 in phase II, stored in DI water at about 5oC, with DI water being
replaced regularly. Different membrane was used in Phase I and Phase II just because
LFC1 membrane in larger dimensions for the long channel RO cell in phase II was
not available. The intrinsic physical and chemical properties of LFC1 and ESPA2
membranes have been characterized extensively in previous studies (Gerard et al.,
1998; Vrijenhoek et al., 2001). Table 3.1 shows some physicochemical characteristics
of these two types of membrane with a structure of polyamide film (Norberg, 2007).
51
Materials and methods
Table 3.1 Surface characterization of LFC1 and ESPA2 RO membranes.
Surface roughness (nm)
Zeta potential (mV)
Contact angle (o)
LFC1
67.4
-5.5
51.8
EPA2
130.2
-7.7
69.0
3.3 RO filtration setup
Two RO filtration setups were used in this study. They consisted of high pressure
system, membrane filtration system and one data acquisition system. The different
part of these two setups was the membrane filtration cell. A small lab-scale crossflow
membrane cell was used in Phase I and a long channel RO membrane cell in Phase II.
3.3.1 Small lab-scale crossflow membrane cell
The small lab-scale membrane cell was made of stainless steal and had a channel
width of 2.9 cm, a length of 10 cm, and a height of 0.2 cm with an effective filtration
area of 2.9×10-3 m2. The feed water was stored in a high-density polyethylene (HDPE)
cylindrical tank of 20 L (Nalgene, Rochester, NY) and was magnetically stirred and
maintained at 25±0.5oC by a chiller (Model CWA-12PTS, Wexten Precise Industries
Co., Taiwan). The feed water was circulated into the membrane cell by a diaphragm
pump (Hydra-cell D-03, Wanner Engineering, Inc., Minneapolis, MN). Desired
pressure and feed water flow rate were achieved by adjusting the bypass needle valve
and back-pressure regulator. The applied pressure and concentrate flow rate were
monitored by a digital pressure gauge (PSI-Tronix, Inc., Tulane, CA) and a variable
area flow meter (Blue-White industries, Ltd., Huntington Beach, CA), respectively.
The permeate flow rate was measured by a digital flowmeter (Optiflow 1000, Agilent
52
Materials and methods
Technologies, Palo Alto, CA) that was linked to a personal computer for continuous
logging and monitoring. Permeate and concentrate were recycled back to the feed
tank.
Feed tank
Permeate
flowmeter
Chiller
Variable area
flowmeter
Bypass needle
valve
Pressure
gauge
Membrane cell
High pressure
pump
Pressure
regulator
(a) Schematic diagram
Permeate
Concentrate
Feed
(b) Picture
Figure 3.1 (a) Schematic diagram of the crossflow RO filtration setup. (b) Picture of
the small lab-scale RO setup.
53
Materials and methods
3.3.2 Long channel crossflow RO membrane cell
A schematic diagram of the longitudinal section of the RO membrane cell is shown in
Figure 3.2. The cell consisted of two long detachable parts made of stainless steel,
forming a longitudinal channel with an effective length of 100 cm and a width of 3 cm
in which a long strip of RO membrane was placed. The selective layer of the
membrane was placed faced down with a gap of 0.8 mm (channel height) towards the
impermeable wall of the bottom part of the module. Spacer was not inserted between
the membrane and the impermeable wall in most runs unless indicated otherwise. The
two ends of the bottom part of the membrane cell were provided with two 5-mm wide
inlet and outlet openings. Ten pressure transducers (Gauge pressure sensor A-105,
Honeywell, Columbus, Ohio) were installed at equal intervals to the bottom part to
monitor the pressure in the feed side along the flow channel. The supported layer of
the membrane was tightly faced against the permeable wall of the upper part of the
membrane cell during testing. The part was split into 10 porous segments with each
10 cm in length and 3 cm in width (that was the effective filtration area of each
segment) to channel permeate off into 10 small openings with diameter of 1.25 cm to
collector permeate and to connecte to digital flow meters in parallel along the channel.
The 10 filtration segments were each separated by a small nonporous zone of 0.2 cm
in length and 3 cm in width in order to prevent cross-mixing of permeate from
different permeate collectors. In this study, every adjacent permeate collectors were
connected to one digital flow meter to record the permeate flow from each two
segments (or one 20-cm permeate segment) from inlet to outlet. In the spacer-filled
channel, spacer was cut into the same dimensions with the channel of 3 cm in width
and 100 cm in length and was placed between the membrane and the impermeable
wall of the bottom part from the inlet to the outlet.
54
Feed from
pump
2
3
2
5
6
(a ) Schematic diagram
4
3
7
4
8
10
Concentrate
recycled to
feed tank
Porous stainless
steel wall
Non-porous zone
9
5
Figure 3.2 (a) Schematic diagram of the long channel RO membrane cell. (b)
Picture of the long channel RO membrane cell (to be continued).
Membrane
10 pressure sensors
to data acquisition
system
1
1
Each two segments measured by one permeate flow meter
Materials and methods
55
Concentrate
Figure 3.2 (Cont.)
(b) Picture
Permeate
Bottom part
Upper part
Feed
Materials and methods
56
Materials and methods
3.4 Filtration and cleaning operation
3.4.1 Salt solution filtration tests
The filtration of salt solution was conducted in the long channel RO membrane cell to
study the permeate behavior along the RO membrane channel. Two M NaCl stock
solution was added to the feed water tank to achieve a concentration of 30 mM (1752
mg/L) after compaction of the RO membrane with 14 L of DI water at 300 psi (2068
kPa) for 24 h to obtain a stable flux. Subsequently, the applied pressure was adjusted
at a value ranging from 414-2,068 kPa (60-300 psi) with an interval of 276 kPa (40
psi) and the feed water flow was adjusted (4.6, 9.1, and 22.6 cm/s) to start the series
of salt solution filtration. The corresponding permeate flow under different operating
conditions was recorded and permeate and concentrate samples were taken after 1 h
of equilibration for each adjustment. The salt concentration in feed water, permeate
and concentrate was tested via conductivity measurement (F-54 BW, Horiba, Japan).
All the experiments were duplicated to ensure repeatability. All the runs were
conducted at a temperature of 25oC and an unadjusted pH at 6.0±0.1. Between each
experimental runs, the feed water tank was emptied and refilled with DI water to flush
the membrane setup in recycle mode for about 1 h at low pressure and high crossflow
velocity. The washing process was repeated at least three times to ensure no salt
residual remained in the system.
3.4.2 Organic fouling tests
Organic fouling tests were conducted using the small lab-scale RO membrane cell and
the long channel RO membrane cell. After the RO membrane system was flushed
with DI water, fresh RO membrane was first compacted with DI water for 24 h at
2.07×106-2.41×106 Pa (300-350 psi). Stock solutions of salt were introduced into the
57
Materials and methods
feed water to achieve the ionic strength of 10 mM unless indicated otherwise and the
system operated for another 2 h at the same pressure and crossflow velocity of 0.091
m/s to acquire a low Reynolds number of 400 or other investigated values of
crossflow velocity as indicated in the result sections. The pressure was then adjusted
to achieve a predetermined initial permeate flux of 1.38×10-5 m/s or other investigated
value of pressure indicated in the result sections, which was maintained for 2 h to
observe a steady permeate flux. The organic stock solution was subsequently added
into the feed water to initiate the fouling process. Permeate flow was continuously
recorded until the test terminated when the permeate flux reached a predetermined
value.
In the organic fouling tests, the salt and organic concentrations were varied to
investigate the effect of salt and organic matters on the fouling. The concentrations
tested are indicated in the result sections. In order to enhance the influence of organic
fouling in this study organic concentration was set at a maximum value of 50 mg/L
which contributed to about 10 mg/L of total organic carbon (TOC), typical one in the
secondary effluent, but much higher than the concentration of polysaccharides and
protein in the secondary effluent. All fouling experiments were conducted at
temperature of 25oC with uncontrolled feed water pH of about 6.0, unless pH and
temperature were an investigated variable as indicated in the result sections.
3.4.3 Cleaning tests
Membrane cleaning was carried out after the permeate flux declined to about half of
the initial flux in the fouling test. The cleaning process started with the circulation of
selected cleaning solution for 30 min at 15 psi (1.03×105 Pa) and crossflow velocity
58
Materials and methods
of 0.29 m/s. After which, the pump was switched off and the RO membrane was
soaked in the cleaning solution for the next 30 min. The same cleaning solution was
again circulated throughout the system for another 30 min before being drained off.
The system was then flushed with DI water for 30 min to remove the residual
cleaning chemical. When the cleaning time was an investigated parameter in the
protein studies, the total cleaning time were 60 or 15 min. The restored flux using
organic-free test solution at identical operating conditions used in the fouling
experiment determined the efficiency of the membrane cleaning. All the operations
were carried out at a temperature of 25oC.
3.5 Membrane hydraulic resistance measurement
Clean water permeability of the membrane was determined by measuring the
stabilized permeate fluxes at various pressures (up to 350 psi) with DI water. After
membrane compaction using DI water and when the stable flux was achieved,
stabilized permeate flux (vp) was measured for a range of operating pressures with DI
water. The clean membrane resistance (Rm0) at different operating pressure (p) was
calculated based on Eq. 3.1.
Rm 0 
p
vp
3.1
LFC1 membrane resistance was determined to be about 8.83×1010 Pas/m on average.
In the light of the large length in the long channel membrane cell, ESPA2 membrane
resistance was plotted with the channel which will be discussed in the following result
sections.
59
Materials and methods
3.6 Beaker tests of gel formation
Beaker tests were used to observe the alginate-Ca2+ gel formation in a beaker to
understand the interaction between alginate and Ca2+. 50-mL organic-free salt solution
with variable Ca2+ (0, 1, 2, and 10 mM) concentration was prepared in different
beakers and sodium alginate stock solution at a concentration of 2 g/L was added into
beakers gradually. After the solution was stirred for 10 min after each alginate
introduction, the turbidity of the solution (2100N Turbidimeter, Hach) was recorded
to characterize the gel formation.
3.7 Analytical techniques
3.7.1 Streaming potential analyzer
Zeta potential of the membrane surface was measured using a streaming potential
analyzer (EKA, Anton Paar GmbH, Graz, Austria). The zeta potential of membrane
surface was calculated from the measured streaming potential using FairbrotherMastin approach (Fairbrother and Mastin, 1924). For each measurement of the
streaming potential, electrolyte composition and concentration in the test solutions
were identical with those of feed solutions in the fouling tests. BSA was also added
into the test solutions for streaming potential measurement at various electrolyte
compositions and concentrations to reflect the actual electrokinetic property of the
membrane during the fouling tests. Hydrochloric acid and sodium hydroxide were
added manually to adjust the solution pH. Before the beginning of each measurement
or after the pH adjustment of the test solution, the test solution was used to thoroughly
flush the EKA cell (about three times at each flow direction and 80 s each time) where
a fresh membrane was installed to remove any air bubbles. This was to achieve the
60
Materials and methods
desired test solution chemistry and to ensure correct results were obtained. Except for
measurements using temperature as a test variable (with a range from 13 – 38oC), all
streaming potential measurements were performed at temperature of 25oC, which was
monitored by the temperature probe in the EKA analyzer. Temperature was controlled
by placing the test beaker into the ice water or warm water bath. Before measurement
of streaming potential, the test cell and the bypass tubing of the EKA analyzer were
rinsed four times (each rinse lasted 80 s) by test solution. Each sample measurement
under one condition was repeated two times, and each time produced four streaming
potential value (measurement of each value took 120 s). Hence, the final resultes
shown in this dissertation was average value of eight data.
3.7.2 SEM-EDX
A scanning electron microscopy (SEM) (JSM-5600LV, JEOL Ltd, Tokyo Japan)
equipped with an energy dispersion of X-Ray spectroscopy (EDX) analysis system
(Link Isis, Oxford Instruments, England) was used to analyze sodium alginate powder,
the new membrane and the fouled membrane. The membrane samples were dried in
the liquid nitrogen following being dried in the air and then were stored in the
desiccators before measurement. The atomic analysis was expressed as weight ratio of
carbon to calcium (C/Ca) which was used as an indicator of membrane fouling layer
matrix. Apart from new membranes and fouled membranes, sodium alginate powder
was also analyzed. The value of sodium alginate powder composition obtained from
EDX analysis was consistent with its theoretical value calculated from the alginate
chemical formula (C6H7NaO6) n.
61
Materials and methods
3.7.3 Polysaccharide and protein assay
The polysaccharide and protein concentration in the feed, permeate and concentrate
were
monitored.
Polysaccharide
concentration
was
determined
using
the
phenol/sulphuric acid assay (Dubios, 1956) with glucose (Amresco, Anhydrous DGlucose) as the calibration standard. Briefly, 2 mL of sample was first mixed with 1
mL of Phenol Reagent (5% phenol solution) and followed by 5 mL of concentrated
sulphuric acid. The solutions were mixed immediately by vortex and allowed to stand
at room temperature for 30 min. Absorbance measurements were taken at 490 nm
with a spectrophotometer (DR/4000U, Hach, Loveland, Colorado). Standards were
prepared by diluting various amount of glucose stock solution (100 μg/mL) in
distilled water.
Protein concentration was determined using the modified Lowry’s Method (Lowry et
al., 1951) with bovine serum albumin (Sigma-Aldrich, B-4287, St. Louis, MO) as the
calibration standard. Briefly, 1 mL of sample was mixed with 5 mL of Assay Mix
using a vortex. Assay Mix was prepared with 5 mL alkaline reagent (0.1 M NaOH,
2% NaCO3, 0.5% Sodium tartrate and 0.5% Sodium dodecylsulfate) and 0.5 mL
copper reagent (1% CuSO4·5H2O). After 10 min of incubation at room temperature,
0.5 mL of diluted Folin-Ciocalteu reagent (1 N) was added to the solution and well
mixed immediately using a vortex. The solutions were allowed to stand at room
temperature for another 30 min. Absorbance measurements were taken at 650 nm
with a spectrophotometer (DR/4000U, Hach). Standards were prepared by diluting
various amount of BSA stock solution (1 μg/μL) in DI water.
62
Polysaccharide fouling and chemical cleaning
Chapter 4
Polysaccharide Fouling and Chemical Cleaning
A small laboratory-scale crossflow RO membrane cell was set up to investigate the
effect of polysaccharide-calcium interaction on membrane fouling and related
chemical cleaning of RO membranes. Sodium alginate was used to model
polysaccharide, a typical macromolecular biopolymer, in the secondary effluent.
Among the different solution chemistry parameters (i.e., pH, ion type, and ionic
strength) affecting the organic fouling, divalent cation ion calcium was chosen as the
variable in this chapter since the predominant alginate fouling of RO membrane is the
specific interaction between alginate and calcium. Apart from the behavior of the
permeate flux characterizing polysaccharide fouling, a beaker test was designed to
simulate the alginate-calcium gel formation in the fouling process. This chapter
discusses the respective roles of sodium alginate and calcium on the initial fouling
rate and long-term fouling extent. The alginate-calcium interaction enhanced-fouling
requires specific cleaning agents. Three types of cleaning agents EDTA, SDS and
NaOH were chosen to clean alginate-fouled membranes. The comparison of their
cleaning efficiency for different solution concentration and pH is also discussed in
this chapter.
63
Polysaccharide fouling and chemical cleaning
4.1 Polysaccharide fouling
4.1.1 Effects of calcium concentration on alginate fouling
Flux decline associated with sodium alginate fouling of RO membranes was studied
with different calcium concentrations (0, 0.1, 0.3, and 1.0 mM) while maintaining the
sodium alginate concentration at 50 mg/L and the results are shown in Figure 4.1. All
flux profiles were characterized by a rapid initial flux decline, followed by a gradual
flux decline, before the flux approached a stabilized value. The extent of flux decline
was found to increase with calcium concentration although the initial fouling rates for
all calcium concentrations were observed to be similar, even the scenario without
calcium in the feed solution. Such an effect of calcium on organic fouling was also
observed in the study of humic acid fouling of nanofiltration (NF) membranes (Hong
and Elimelech, 1997; Li and Elimelech, 2004). The severe decline of permeate flux
was attributed to the substantial increase in membrane hydrodynamic resistance due
to the development of alginate fouling layer on the membrane surface. This alginate
foulant layer was the product of alginate-calcium complexation, followed by calciumbridging of alginate macromolecules, which could be explained by the popular “eggbox” model (Grant et al., 1973). This model stated that the specificity of complexation
between cation ions and polysaccharide to lead to firm cohesion between
macromolecular chains. Hence, with higher calcium concentration, there would be
greater capacity to form the cross-linked alginate chains leading to the formation of a
thicker, denser, and more compact fouling layer.
64
Polysaccharide fouling and chemical cleaning
1.0
Normalized flux, J/J0
0.8
0.6
0.4
NaCl 10 mM, CaCl2 0 mM
NaCl 9.7 mM, CaCl2 0.1 mM
NaCl 9.1 mM, CaCl2 0.3 mM
NaCl 7.0 mM, CaCl2 1.0 mM
0.2
0.0
0
10
20
30
40
50
Time (h)
Figure 4.1 Effects of calcium concentration (0, 0.1, 0.3, and 1.0 mM) on permeate
flux with time over a period of 50 h. Ionic strength was maintained at 10 mM by
varying the sodium chloride concentration. Sodium alginate concentration was 50
mg/L. pH was unjusted at 6.0±0.1. Initial permeate flux was 1.38×10-5 m/s. Crossflow
velocity was 0.0914 m/s.
4.1.2 Effects of alginate concentration on alginate fouling
Alginate fouling was also investigated with different sodium alginate concentrations
(10, 20, and 50 mg/L) in the presence of fixed calcium concentration of 1.0 mM.
Figure 4.2 shows the changes in permeate flux due to alginate fouling at three
different sodium alginate concentrations over about 50 h of operation. Initial fouling
rates were observed to correspond with the sodium alginate concentration, i.e.,
increasing sodium alginate concentration resulting in higher fouling rate. This
observation contrasted greatly to the flux curves shown in Figure 4.1, which showed
similar initial fouling rate. It seemed that the alginate concentration, rather than the
calcium concentration, affected the initial fouling rate.
Near the end of the
65
Polysaccharide fouling and chemical cleaning
experimental runs, the stabilized normalized fluxes at three different sodium alginate
concentrations (Figure 4.2) were less different as compared to normalized fluxes for
varying calcium concentration but with constant sodium alginate concentration
(Figure 4.1). Hence, calcium concentration seemed to have greater influence on the
long term flux than the sodium alginate concentration.
1.0
10 mg/L
20 mg/L
50 mg/L
Normalized flux, J/J0
0.8
0.6
0.4
0.2
0.0
0
10
20
30
40
50
Time (h)
Figure 4.2 Permeate flux over a period of 50 h at different sodium alginate
concentrations (10, 20, and 50 mg/L). Ionic strength was 10 mM with sodium chloride
of 7.0 mM and calcium chloride of 1.0 mM. pH was unjusted at 6.0±0.1. Initial
permeate flux was 1.38×10-5 m/s. Crossflow velocity was 0.0914 m/s.
The element analysis by SEM-EDX showed the gel formation due to alginate –
calcium interaction during the fouling development. Figure 4.3 shows the weight ratio
C/Ca with operation time for six runs under varying the sodium alginate concentration
(10, 20 and 50 mg/L) and fixed calcium concentration (1.0 mM). At the beginning, no
66
Polysaccharide fouling and chemical cleaning
fouling occurred and the ratio at time of 0 was calculated according to the sodium
alginate and calcium concentration in the feed water. Other six data was obtained with
different fouled membranes from six runs stopping at the various times. The carbon
content in sodium alginate measured by SEM-EDX was about 38% (w/w). With
higher sodium alginate concentration, the C/Ca ratio increased faster at the initial
fouling stage, indicating alginate contributed much more to the fouling than calcium.
However, the final carbon and calcium ratio reached to a similar value for all cases,
which supported the alginate-calcium interaction and one component’s priority over
the other during the fouling development.
2.0
C/Ca, w/w
1.6
1.2
0.8
Sodium alginate:
10 mg/L
20 mg/L
50 mg/L
0.4
0.0
0
10
20
30
40
50
Time (h)
Figure 4.3 Carbon to calcium weight ratio on the fouling layer of the membrane. The
fouled membrane samples were obtained from six runs with different filtration time.
Sodium alginate concentration was 10, 20, or 50 mg/L and calcium concentration was
1.0 mM for all runs. pH was unjusted at 6.0±0.1. Initial permeate flux was 1.38×10-5
m/s. Crossflow velocity was 0.0914 m/s.
67
Polysaccharide fouling and chemical cleaning
4.1.3 Gel formation in beaker tests
Figure 4.4 shows a positive relationship between turbidity and sodium alginate at
different calcium concentrations (0, 1, 2 and 10 mM). The turbidity increased with
alginate concentration, which is due to more gel formed at higher alginate
concentration. In addition, the turbidity increasing rate was higher at high calcium
concentration, 0.02 NTU/alginate (mg/L) at calcium concentration of 10 mM while
0.0035 NUT/alginate (mg/L) at calcium concentration of 1mM.
On closer inspection, two important turning points of alginate concentration (334.5
mg/L and 666.7 mg/L) were observed at calcium concentration of 1 and 2 mM,
respectively, while such a phenomenon was not observed for zero calcium
concentration and a high calcium concentration of 10 mM (Figure 4.4b) within the
tested alginate concentration rage in this study. After the turning point, the turbidity
increasing rate at calcium concentration 1 and 2 mM became as low as calcium
concentration of 0 mM, 0.0017 NTU/alginate (mg/L), where no alginate-calcium gel
formation was observed. Hence no more gel was formed even though more sodium
alginate was continuously introduced into the solution beyond the turning point at a
fixed calcium concentration. This indicated that the amount of gel formation was
limited by calcium concentration. When calcium concentration was increase to a
higher value of 10 mM, the turning point was not observed just because there was
sufficient calcium for reaction with alginate to form more gel. At the two turning
points, the weight ratios of alginate and calcium were about 8.3. Therefore, the
possible turning point for the case of calcium concentration of 10 mM was a high
alginate concentration of more than 3000 mg/L, which was beyond the tested range of
alginate concentration.
68
Polysaccharide fouling and chemical cleaning
8
Ca concentration:
0 mM
1 mM
2 mM
10 mM
(a)
7
Turbidity (NTU)
6
5
4
3
2
1
0
0
200
400
600
800
1000
800
1000
Sodium alginate (mg/L)
25
(b)
10 mM
Turbidity (NTU)
20
15
10
5
0
0
200
400
600
Sodium alginate (mg/L)
Figure 4.4 Turbidity change with gel formation by sodium alginate and calcium in
beaker tests. Figure 4.4b is the extended range of Y axis for 10 mM calcium. The
ovals in Figure 4.4a highlight the turning points.
69
Polysaccharide fouling and chemical cleaning
4.2
Chemical
cleaning
of
membranes
fouled
by
polysaccharide
4.2.1 Effects of calcium on membrane cleaning
Membrane cleaning using different cleaning agents was carried out right after the
fouling experiments. Figure 4.5 shows the flux restoration after the fouled membranes
was cleaned separately with DI water, EDTA, SDS and NaOH. When the membranes
were fouled with calcium-free feed water, flux restoration above 90% of initial flux
was achieved using the three cleaning agents and interestingly, even with DI water.
This suggests that the shear rate (45.7 s-1) was sufficient to overcome the bond
between the alginate and the membrane surface, allowing the removal of the foulant
layer on the membrane surface.
However, the same shear rate was not sufficient to physically remove the strongly
bonded alginate-calcium gel from the membrane surface when calcium was present in
the feed water. Flux could be only restored to 53% of its initial flux when DI water
was used to clean the membrane. SDS and NaOH performed slightly better by
achieving around 60% of the initial flux. EDTA, on the other hand, was able to almost
restore the flux to its initial level. The excellent performance of EDTA as a cleaning
agent for alginate removal is due to its strong metal chelating property that can
preferentially bond calcium in alginate-calcium gel through ligand-exchange reaction
to form soluble calcium-EDTA complexes (Ang et al., 2006). As a result, the crosslinked alginate-calcium structure was destroyed and the alginate chains were easily
detached and removed from the membrane surface. Here, the cleaning efficiency over
100% (that is, normalized flux is over 1) after EDTA cleaning was noted for the
70
Polysaccharide fouling and chemical cleaning
membranes . This is due to the fact that EDTA in contact with the membrane during
the cleaning process may make the membrane more hydrophilic and increase the
water permeability (Madaeni and Mansourpanah, 2004).
110
W ithout Ca
W ith Ca
Flux restoration, %
100
90
80
70
60
50
H 2O
EDTA 1mM
pH=4.5
SDS 1mM
pH=8
NaOH
pH=9
Cleaning solution
Figure 4.5 Comparison of the extent of flux restoration with and without calcium in
the feed water using different types of chemical cleaning agents. The fouling test
conditions: sodium alginate of 50 mg/L; calcium of 0 or 1.0 mM; ionic strength of 10
mM; pH unjusted (6.0±0.1); initial permeate flux of 1.38×10-5 m/s; crossflow velocity
of 0.0914 m/s. Cleaning conditions are described in Chapter 3.
Table 4.1 shows the atomic weight percentage of carbon, sulphur and calcium found
on the new, fouled and cleaned membrane surfaces obtained by SEM-EDX. It was
observed that the amount of calcium increased when the membrane was fouled, due to
the formation of a layer of alginate-calcium gel on the membrane surface. Membrane
cleaning using SDS and NaOH could not effectively remove the alginate gel, as the
carbon and calcium content in the cleaned membranes were similar to that of a fouled
71
Polysaccharide fouling and chemical cleaning
membrane. In the case of EDTA, calcium was not detected on the membrane surface
and the carbon content was similar to that of a new membrane after cleaning, i.e.,
EDTA was effective in removing the alginate gel layer from the membrane surface.
Table 4.1 Atomic weight percentage on the membrane surface by SEM-EDX.
Mean atomic weight percentage (%)
Membrane
a
b
C
S
Ca
New membrane
70.68
8.37
-a
Fouled membraneb
22.42
1.42
14.60
EDTA cleaning
66.80
7.60
-a
SDS cleaning
22.99
0.47
15.10
NaOH cleaning
22.39
0.54
15.75
not detected
Fouled membrane was obtained from the fouling run using 10
mM of calcium and 50 mg/L of sodium alginate.
4.2.2 Effects of pH and concentration of cleaning solution
The effects of pH and concentration of EDTA and SDS solutions on flux restoration
were investigated with fouling tests under identical operating conditions. Figures 4.6a
and 4.6b show the flux restoration at different pH and EDTA concentrations.
Membrane cleaning was observed to be more effective at pH above 4.5, where
deprotonation of COOH-groups was more extensive (Ang et al., 2006), and at higher
EDTA concentration, where more EDTA was available to form complexes with
calcium in the alginate gel (Ang et al., 2006). Based on the experimental conditions,
72
Polysaccharide fouling and chemical cleaning
including the sodium alginate concentration, 0.5 mM EDTA solution at pH 4.5 was
adequate for complete flux restoration.
Figs. 4.7a and 4.7b show the flux restoration under different pH and concentrations of
SDS solution. SDS was shown to perform better at lower pH than at higher pH. As an
anionic surfactant, the surface of the SDS micelle has a high density of negative
charges, which induces an electrical repulsive force to the negatively charged alginate
(Bu et al., 2005). At low pH the negative charge density could decrease, which in turn,
resulted in a larger adhesive force between these two compounds.
Flux restoration with SDS solution increased from 60% to almost 100% of its initial
value when SDS concentration was raised from 1 to 5 mM. The results were in
agreement with the findings of Lee et al. (2001), who reported poor flux restoration
with low concentrations of SDS on UF membrane fouled by organic matters in
surface water. They pointed out that 1 mM SDS was lower than critical micelle
concentration (CMC) of about 8.40 mM in DI water (Bu et al., 2006), which was
believed to be the most effective SDS concentration for organic fouling removal.
However, CMC is also dependent on the ionic strength of the solution (Thongngam et
al., 2005). Alginate dissolution might increase the ionic strength near the surface of
membrane and decrease the CMC. In this case, high removal efficiency could be
achieved at a SDS concentration less than 8.4 mM. Nevertheless, SDS cleaning would
consume more chemicals than EDTA cleaning to achieve the same cleaning
efficiency.
73
Polysaccharide fouling and chemical cleaning
110
(a)
Flux restoration, %
100
90
80
70
60
50
pH=3.7
pH=4.5
pH=7
pH=10
Cleaning solution pH
(b)
Flux restoration, %
100
90
80
70
60
50
0.1 mM
0.5 mM
1 mM
EDTA concentration
Figure 4.6 Flux restoration with EDTA cleaning at different (a) pH (the concentration
of EDTA solution was fixed at 1 mM); and (b) EDTA concentrations (the pH of
EDTA solution was fixed at 4.5). Fouling conditions: sodium alginate of 50 mg/L;
calcium of 1.0 mM; ionic strength of 10 mM; pH unjusted (6.0±0.1); initial permeate
flux 1.38×10-5 m/s; and crossflow velocity 0.0914 m/s. Cleaning conditions are
described in Chapter 3.
74
Polysaccharide fouling and chemical cleaning
100
(a)
Flux restoration, %
80
60
40
20
0
pH=4.5
pH=8
Cleaning solution pH
100
(b)
Flux restoration, %
90
80
70
60
50
1 mM
5 mM
SDS concentration
Figure 4.7 Flux restoration after membrane cleaning with SDS at different (a) pH (the
concentration of SDS solution was fixed at 1 mM); and (b) SDS concentrations (the
pH of SDS solution was fixed at 8). Fouling conditions: sodium alginate of 50 mg/L;
calcium of 1.0 mM; ionic strength of 10 mM; pH unjusted (6.0±0.1); initial permeate
flux of 1.38×10-5 m/s; crossflow velocity of 0.0914 m/s. Cleaning conditions are
described in Chapter 3.
75
Polysaccharide fouling and chemical cleaning
4.2.3 Impact of multiple cleaning cycles
The effectiveness of EDTA for membrane cleaning was investigated over a series of
fouling and cleaning cycles on the same piece of RO membrane in a single fouling
experiment consisting of three alternant fouling and cleaning. Permeate flux was
allowed to decline to half the initial value before the membrane was cleaned with
EDTA solution. Figure 4.8 shows the permeate fluxes over a period of 15 h, in which
three membrane cleanings were done to restore the permeate flux. The extent of flux
restoration for each cleaning is given in Figure 4.9.
1.0
Normalized flux, J/J0
0.8
0.6
0.4
st
1 cleaning
nd
2 cleaning
rd
3 cleaning
0.2
0.0
0
2
4
6
8
10
12
14
Time (h)
Figure 4.8 Normalized fluxes over three successive cycles of alginate fouling and
EDTA cleaning. Fouling conditions: alginate of 50 mg/L; calcium of 1.0 mM; ionic
strength of 10 mM; pH unjusted (6.0±0.1); initial permeate flux of 1.38×10-5 m/s; and
crossflow velocity of 0.0914 m/s.
76
Polysaccharide fouling and chemical cleaning
Flux restoration, %
100
90
80
70
60
50
1
2
3
Cleaning cycle
Figure 4.9 Comparison of flux restoration after each successive EDTA cleaning.
Cleaning conditions: EDTA concentration of 1 mM and pH 4.5. Other conditions are
described in Chapter 3.
It was found that although EDTA could restore the flux to over the initial value, the
extent of flux restoration decreased with the number of membrane cleaning.
Obviously, the amount of accumulated alginate on the membrane surface increased
with membrane cleaning cycle time. One possible explanation for such phenomenon
is that EDTA cleaning altered the membrane to be more hydrophilic to get a higher
cleaning efficiency on one side, and on the other side it strengthened the force
between the hydrophilic alginate and the more hydrophilic membrane. Since the
subsequent cleaning could not remove all the alginate which was more tightly retained
on the membrane surface, more alginate was accumulated and the resistance across
the membrane increased.
77
Polysaccharide fouling and chemical cleaning
4.3 Summary
Membrane fouling of polysaccharide (sodium alginate) in the presence of calcium
ions in feed water and membrane cleaning with different chemicals were investigated
in this study. Calcium is believed to provide the ‘crosslink” between alginate chains,
which resulted in the formation of high resistance alginate-calcium gel on the
membrane surface. In addition, alginate concentration was the dominant factor in the
initial fouling rate while calcium concentration determined the long term flux. Near
complete flux restoration could be easily achieved by flushing with deionized water
on membranes fouled with alginate in the absence of calcium. However, membranes
fouled with alginate-calcium gel could only restore the flux when cleaned with a
suitable chemical. EDTA was shown to be a better cleaning agent for organic fouling
in the presence of calcium because of its ability to form soluble complexes with
calcium, forming the alginate-calcium gel. Further decrease in flux due to the
interaction between polysaccharide and calcium highlighted the importance of
establishing and reducing the inorganic components, especially calcium ions, in the
feed water. Therefore, pre-treatment of feed water should not only target the foulants
directly, but also the ionic components that can interact with the foulants and magnify
the foulants’ fouling potential.
78
Protein fouling and chemical cleaning
Chapter 5
Protein Fouling and Chemical Cleaning
Literature review has shown that the structural stability of protein is determined by
covalent bonds and various non-covalent interactions (Compbell and Farrell, 2006). A
significant amount of studies on protein fouling in MF/UF confirmed the roles of
these non-covalent interactions in determining the protein structure in solution and the
fouling behavior. It was also noted that some conflicting results have been reported by
different researchers (Matthiasson, 1983; Mueller and Davis, 1996). Therefore, there
is a need to carry out a systematic study on protein fouling in RO process for effective
fouling control. This chapter presents the fouling behavior of protein using bovine
serum albumin (BSA) as a model protein in a small lab-scale crossflow RO filtration
setup. Some important chemical (pH, cation species and ionic strength) and physical
(temperature) parameters were investigated. The streaming potential of the membrane
surface in a solution with protein was measured to elucidate the fouling behavior. In
addition, the RO membrane fouled by BSA was cleaned by three types of cleaning
chemicals (EDTA, SDS and urea) to restore the permeate flux. Effects of pH,
concentration and cleaning time on the cleaning efficiency were investigated.
5.1 Protein fouling
5.1.1 Adsorption of BSA on membrane
The adsorption of BSA on membrane can be demonstrated by the change of the zeta
potential of the membrane surface. Figure 5.1 illustrates the surface zeta potential of a
79
Protein fouling and chemical cleaning
fresh RO membrane as a function of pH in the absence and presence of BSA (1 mM
or 10 mM) in the test solution. It was observed that the membrane surface zeta
potential generally decreased or became more negative when the pH was increased. In
the absence of BSA, the isoelectric point (IEP) of LFC-1 membrane was found to be
above 3.3, which was comparable to the value of 3.5 reported by Elimelech et al.
(1994). The change in the membrane surface charge was attributed to the protonation
(below IEP) and deprotonation (above IEP) of various functional groups, such as
carboxylic and amine groups, on the membrane surface. In the presence of 50 mg/L of
BSA, it was observed that the zeta potential profile shifted to the right of the graph,
i.e., the IEP of the membrane increased from 3.3 to about 4.5. The increased IEP was
similar to the IEP of BSA, which was about 4.74.9 (Nakamura and Matsumoto,
2006). This strongly suggested that BSA was adsorbed onto the membrane surface
and dominated the surface charge of the RO membrane.
10
NaCl 1mM
NaCl 10mM
NaCl 1mM & Protein
NaCl 10mM & Protein
Zeta potential (mV)
5
0
-5
-10
-15
-20
-25
2
3
4
5
6
7
8
pH
Figure 5.1 Zeta potential on RO membrane surface as a function of pH at different
ionic strength and in the presence/absence of BSA (50 mg/L BSA).
80
Protein fouling and chemical cleaning
The electrostatic interaction between the BSA and membrane surface at different pH
could be explained with the zeta potentials measured from both materials. At pH
between the IEP of BSA and RO membrane, electrostatic attraction occurred since the
protein and membrane surface have opposite charges in the aqueous solution. This
encouraged the adsorption of BSA onto the membrane. When the pH was above the
IEP of BSA, BSA and the membrane surface were negatively charged, and
electrostatic repulsion was the dominant interaction between the two materials. Hence,
any adsorption of BSA onto membrane surface in this pH range was attributed to
other interactions between BSA and membrane surface such as structural interaction
(Salgın et al., 2006). A previous study has reported that the apparent protein
adsorption at pH above the IEP of BSA could be attributed to the “soft” structure and
low internal stability of BSA (Bos et al., 1994). It has been demonstrated that BSA is
adsorbed on all surfaces regardless of electrostatic interactions, owing to a gain in
conformational entropy resulting from adsorption. Since the inevitable adsorption of
BSA on membrane surface, protein fouling in RO process would be dominated by the
interaction between BSA in the bulk solution and BSA adsorbed on membrane
surface, rather than direct interaction between BSA in the bulk solution and the fresh
membrane surface.
5.1.2 Effects of ionic strength
Figure 5.1 also shows the effect of ionic strength on zeta potential of RO membrane
surface in the presence of BSA at different pH. It was observed that ionic strength
could affect the zeta potential on membrane surface near to the extreme ends of the
tested pH range. At high pH, the zeta potential of the membrane surface was less
negative at 10 mM of NaCl as compared to 1 mM of NaCl. At low pH, the zeta
81
Protein fouling and chemical cleaning
potential of membrane surface at 10 mM NaCl was less positive than the surface zeta
potential at 1 mM of NaCl. However, there was no significant difference in the IEP of
the membrane surface at both ionic strengths. The smaller magnitude in zeta potential
at higher ionic strength could be attributed to the double layer compression or
shielding effect by an increase in counter-ions with an increase in ionic strength
(Jones and O’Melia, 2000).
The fouling behavior of BSA using RO membrane was determined with a crossflow
RO cell and shown in Figs. 5.2a and 5.2b. Both figures show that the highest rate of
flux decline occurred at pH 4.9, while permeate flux declined most slowly at pH 7.
These observations could be reasonably explained by the extent of electrostatic
repulsion between the BSA and the BSA-adsorbed RO membrane. The previous
section had shown that the measured IEP of the BSA-adsorbed RO membrane was
very much similar to the IEP of the BSA. Hence similar surface charge was expected
on both BSA and BSA-adsorbed RO membrane that would lead to electrostatic
repulsion. The extent of electrostatic repulsion was dependent on the magnitude of
surface charge, which in turn was a function of the difference between the pH and IEP.
Since zeta potential at pH 4.9 was very near the IEP, repulsion between BSA and RO
membrane was weakest and BSA could easily accumulate on the neutral membrane
surface under convectional flux. As pH 7 was more far away from the IEP of the
membrane (pH 4.6) compared to pH 3.9, higher surface charge and stronger
electrostatic repulsion between BSA and RO membrane at pH 7 led to much less
accumulation of BSA on the membrane surface. At pH 3.9, apart from the
electrostatic repulsion between BSA and BSA-adsorbed membrane surface, the
electrostatic attraction between BSA (positive charge) and the fresh membrane
82
Protein fouling and chemical cleaning
(negative charge) affected the initial flux decline rate at the initial fouling stage.
However, the electrostatic attraction was partially offset by the intermolecular
electrostatic repulsion of BSA and contributed to a slower fouling rate than at pH 4.9.
(a)
1.0
Normalized flux, J/J0
0.8
0.6
0.4
pH=3.9
pH=4.9
pH=7
0.2
0.0
0
5
10
15
20
25
Time (h)
1.0
(b)
Normalized flux, J/J0
0.8
0.6
0.4
pH=3.9
pH=4.9
pH=7
0.2
0.0
0
5
10
15
20
25
Time (h)
Figure 5.2 Normalized flux with time at different pH under different ionic strength of
(a) 1 mM and (b) 10 mM.
83
Protein fouling and chemical cleaning
When the ionic strength was increased from 1 to 10 mM, the rate of fouling was
observed to increase for feed water with pH 3.9 and 7. The shielding effect enhanced
by the increases in counter ions with ionic strength reduced the electrostatic repulsion
between the BSA-adsorbed membrane and BSA, or between BSA molecules, which
expedited the accumulation of BSA on the membrane surface. This effect even
changed the shape of the curves, especially for pH 3.9, from convex to concave. This
indicated the effects mostly occurred in the initial stage of fouling instead of the
subsequent accumulation of BSA after the membrane surface was fully occupied by
BSA in the late stage of the fouling. On the other hand, the fouling at pH 4.9 seemed
to be slightly alleviated with an increase in ionic strength. With the lack of active
electrostatic interactions at pH near IEP, this fouling behavior could be attributed to
the structural change in BSA molecule. At pH near the IEP of BSA, the functional
groups with opposite charges within the BSA molecule attracted one another to form
a compact molecular structure with zero net charge. The compact structure explained
the dense fouling layer often seen in cases with severe fouling (Jones and O’Melia,
2000). When the ionic strength was increased, the increased charge shielding reduced
the attractive force between functional groups, resulting in a slightly larger molecule
that formed a more porous fouling layer.
5.1.3 Effects of cations
The cation species present in the feed water could affect the zeta potential of the RO
membrane. Zeta potentials at different pH for monovalent (Na+) and divalent (Ca2+
and Mg2+) ions were measured under constant ionic strength in BSA solution (50
mg/L) and shown in Figure 5.3. Both Mg2+ and Ca2+ exerted similar effect on the zeta
potential over the range of pH tested. The IEP of the BSA-adsorbed RO membrane
84
Protein fouling and chemical cleaning
was observed to slightly shift to the right in the presence of divalent cations. When the
pH was around the higher end of the pH range tested, zeta potentials measured from
solutions containing divalent cations were found to be much less negative than the
one measured from sodium chloride solution. This could be attributed to the
preferential adsorption of divalent cations to the negatively charged membrane
(Lawrence et al., 2006). At the lower end of the pH range, zeta potentials of both
divalent cations were similar to that of sodium chloride. This is the result of low
adsorption of divalent ions as the membrane surface was positively charged.
10
+
Na 10mM
+
2+
Na 7mM & Ca 1mM
+
2+
Na 7mM & Mg 1mM
Zeta potential (mV)
5
0
-5
-10
-15
-20
3
4
5
6
7
8
9
pH
Figure 5.3 Zeta potential of RO membrane surface as a function of pH with different
cation species. Ionic strength was maintained at 10 mM for all cation species and
BSA concentration was maintained at 50 mg/L.
85
Protein fouling and chemical cleaning
The fouling tests using BSA solution with different cations were conducted at
different pH. The ionic strength was kept constant at 10 mM in all the feed waters.
The decline in the permeate flux with time for all feed waters is shown in Figure 5.4.
The most severe fouling for feed water containing only Na+, and Na+ and Ca2+ was
observed at pH 4.9. This was reasonable since the IEP of the RO membrane (about
4.8) was very similar in both types of feed water. The extent of fouling, however,
differed with the cations in the feed water. In the case of Ca2+ presence in the feed
water, permeate flux declined by 54 and 36% at pH 4.9 and 7, respectively, after
running for 25 h. Compared to 41% (pH 4.9) and 12% (pH 7) decline achieved with
only Na+ in the feed water, it was clear that fouling was more severe when Ca2+ was
present in the feed water. The more severe fouling occurred because Ca2+, unlike Na+,
could form ionic bridge between two adjacent carboxyl groups from different peptide
chains (Xiong, 1992), resulting in a denser fouling layer forming on the membrane
surface. In addition, the lower negative zeta potential due to preferential adsorption of
Ca2+ compared to Na+ that led to the reduction in electrostatic repulsion also
contributed to the higher fouling rate of Ca2+ at pH higher than the IEP. When fouling
tests were conducted at pH 3.9, the extent of fouling for both Ca2+ and Na+ was
similar. Results suggested that the effect of additional positive charge on the Ca2+ was
restricted in an environment where electrostatic repulsion was achieved with similar
positive zeta potential on both the membrane surface and BSA.
86
Protein fouling and chemical cleaning
(a)
1.0
Normalized flux, J/J0
0.8
0.6
0.4
pH=3.9
pH=4.9
pH=7
0.2
0.0
0
5
10
15
20
25
Time (h)
(b)
1.0
Normalized flux, J/J0
0.8
0.6
0.4
pH=3.9
pH=4.9
pH=7
0.2
0.0
0
5
10
15
20
25
Time (h)
Figure 5.4 Normalized flux decline with time at three different pH with different
cation species (a) 10 mM Na+; (b) 7 mM Na+ and 1 mM Ca2+; (c) 7 mM Na+ and 1
mM Mg2+ (to be continued).
87
Protein fouling and chemical cleaning
1.0
(c)
Normalized flux, J/J0
0.8
0.6
0.4
pH=3.9
pH=4.9
pH=7
0.2
0.0
0
5
10
15
20
25
Time (h)
Figure 5.4 (Cont.)
One interesting finding in this study is the fouling behavior of BSA in the presence of
Mg2+. The extent of fouling with Mg2+ was not as severe as the ones observed with
Ca2+, even though both were divalent ions. Li and Elimelech (2004), in their study on
NOM fouling using atomic force microscopy (AFM), had shown that the adhesion
force of the AFM probe on membrane in the presence of Mg2+ was much lower than
the force experienced with Ca2+. The slight interaction between the foulant (probe)
and membrane observed with Mg2+ demonstrated that Mg2+ functioned differently
from Ca2+ in the process of protein fouling. In another study, Lee et al. (2006) found a
thick layer of alginate gel layer on the membrane surface in the presence of Ca2+,
while no alginate layer was visible when test was done with Mg2+. Another interesting
finding was that in the presence of Mg2+, BSA fouling rates were similar for all the
pH tested, unlike the fouling behavior observed with Ca2+ and Na+. Hence, protein
88
Protein fouling and chemical cleaning
fouling did not seem to be affected by the change in pH in the presence of Mg2+. This
anomaly in the fouling profile in the presence of Mg2+ is probably attributed to other
interactions apart from the electrostatic interaction and binding with protein, most
likely salting-out effect. Based on the hydration theory, protein would precipitate out
of solution or exert self-aggregation due to the salting-out effect (Grover and Ryall,
2005). Mg2+ possesses higher charge density than Ca2+ and Na+ and shows greater
salting-out effect (Tran-Ha et al., 2005). It was hypothesized that the salting-out effect
resulted in faster decline of permeate flux at pH 7 (away from IEP) due to protein
precipitation out. However, at IEP or below IEP, the synergetic effects of electrostatic
repulsion, protein binding and salting-out led to complex the fouling profile due to
unknown interactions. To our best knowledge, few works on the effect of Mg2+ on
organic fouling was documented although Mg2+ was a typical cation in the secondary
effluent. Therefore further detailed investigation is needed to verify the above
hypothesis.
5.1.4 Effects of temperature
Zeta potential on membrane surface at different temperatures was measured at
different pH and the results are shown in Figure 5.5. It was observed that zeta
potential became less negative with increasing temperature at pH 4.9 and 7. Generally,
as the temperature increases, the structure of BSA molecule on membrane surface
undergoes conformational changes in the form of unfolding, which exposes the
charged functional groups to the aqueous solution (Kelly and Zedney, 1994; Meireles
et al., 1991; Simmons et al., 2007). The aqueous solution contains counter-ions that
can screen the charged functional groups, resulting in the decline of zeta potential at
pH range where zeta potential is predominantly negative. The decreasing zeta
89
Protein fouling and chemical cleaning
potential reduces the electrostatic repulsion between BSA molecules and between
BSA and membrane surface, which consequently accelerates the BSA aggregation
and accumulation on membrane surface. In addition, the unfolding of BSA molecules
at higher temperature exposes the hydrophobic residues and increases the
hydrophobic interactions between BSA molecules that lead to a higher rate of
aggregation (Veerman et al., 2003). The effect of increasing temperature on BSA
fouling is illustrated in Figure 5.6, which shows the fouling behavior of BSA at
different temperatures and pH. At any pH, normalized flux was observed to decline
much faster at higher temperature, indicating higher fouling rate at higher temperature.
8
Zeta potential (mV)
4
pH=3.9
pH=4.9
pH=7
0
-4
-8
-12
10
15
20
25
30
35
40
o
Temperature ( C)
Figure 5.5 Zeta potential on RO membrane surface as a function of temperature for
three different pH. BSA concentration was 50 mg/L and ionic strength of solution
was 10 mM (1 mM Ca2+ and 7 mM Na+).
90
Protein fouling and chemical cleaning
1.0
(a)
Normalized flux, J/J0
0.8
0.6
0.4
o
T=18 C
o
T=25 C
o
T=35 C
0.2
0.0
0
5
10
15
20
25
Time (h)
1.0
(b)
Normalized flux, J/J0
0.8
0.6
0.4
o
T=18 C
o
T=25 C
o
T=35 C
0.2
0.0
0
5
10
15
20
25
Time (h)
Figure 5.6 Normalized flux with time at different temperatures for different pH: (a)
pH=3.9; (b) pH=4.9; (c) pH=7. BSA concentration was 50 mg/L and ionic strength of
solution was 10 mM (1 mM Ca2+ and 7 mM Na+) (to be continued).
91
Protein fouling and chemical cleaning
1.0
(c)
Normalized flux, J/J0
0.8
0.6
0.4
o
T=18 C
o
T=25 C
o
T=35 C
0.2
0.0
0
5
10
15
20
25
Time (h)
Figure 5.6 (Cont.)
It was also revealed in the experiments that the change in temperature had a greater
effect on BSA fouling in feed water with higher pH. The reason for this behavior was
demonstrated in Figure 5.5 where zeta potential of membrane surface was observed to
become less negative with temperature at faster rate as pH was increased from 4.9 to 7.
No significant variation in zeta potential at pH 3.9 was observed over the temperature
range. The higher rate of decline in surface charge with temperature associated with
higher pH reduced the electrostatic repulsion between BSA molecules and between
BSA and BSA-adsorbed membrane surface, leading to a more significant increase in
the rate of fouling. The exact explanation to the combined effect of pH and
temperature on surface charge of BSA was unclear, although the ability of BSA
molecule to unfold at lower pH (Michnik et al., 2005; El Kadi et al., 2006) could
reasonably explain the different rate of surface charge decline with temperature at
92
Protein fouling and chemical cleaning
different pH. The more compact and folded nature of BSA molecule at higher pH
meant that more charged functional groups not in contact with the aqueous solution
would be exposed to charge screening when the BSA molecule began to unfold with
temperature increase, resulting in greater decrease in surface charge. On the other
hand, raising the temperature at lower pH might not significantly affect the surface
charge as the charged functional groups had already been exposed to the aqueous
solution with the pH decline.
5.2 Chemical cleaning of membranes fouled by protein
5.2.1 Effects of cleaning solution concentration
BSA-fouled membranes were cleaned using SDS, urea and EDTA at different
concentrations after the permeate flux declined to about 54% of the initial flux in the
fouling process of about 25 h. All the fouling runs were conducted at the identical
conditions (i.e., BSA concentration of 50 mg/L; ionic strength of 10 mM; calcium
concentration of 1.0 mM; pH of 4.9; and temperature of 25oC). These fouling
conditions were the scenario where most severe BSA fouling occurred in this study.
The cleaning efficiency, defined as the permeate flux of BSA-free solution after
cleaning divided by the initial flux before filtration, by different chemical agents (i.e.,
SDS, urea and EDTA) at different concentration are shown in Figure 5.7. In general,
all the cleaning efficiencies by the three different cleaning agents increased with
cleaning solution concentration even though three categories of cleaning agents had
their respective characteristics.
93
Protein fouling and chemical cleaning
100
Flux restoration (%)
90
80
70
SDS
urea
EDTA
60
50
0.1
1
10
Concentration (mM)
Figure 5.7 Variation of SDS, urea and EDTA cleaning efficiencies as a function of
cleaning solution concentration (Cleaning time of 60 min; operating pressure of 15 psi;
crossflow velocity of 0.29 m/s; and pH of 8.0 for SDS, 7.8 for urea and 5.2 for
EDTA).
EDTA cleaning achieved higher cleaning efficiencies at low concentrations (0.1–1
mM) compared to urea (10 mM) and SDS (5 mM). In addition, EDTA cleaning
efficiency depended slightly on its concentration in the tested range. In the previous
section it is observed that the fouling was enhanced in the presence of Ca2+ due to the
link between protein and Ca2+. Hence, EDTA is effective in breaking the fouling layer
through the preferential EDTA-calcium complexation.
SDS is a surface active agent and is expected to solubilize protein at a concentration
near its critical micelle concentration (CMC) of about 8.4 mM in DI water. The
results showed that there was an optimum concentration of SDS (5 mM) to achieve
94
Protein fouling and chemical cleaning
best cleaning efficiency. Apart from solubilizing protein and removing the protein
fouling layer, the high dosage of SDS would displaced the adsorbed protein layer,
thus possibly leading to the formation of a new fouling layer (Chilukuri et al., 2001).
The cleaning efficiency by urea highly depended on the concentration and the highest
flux restoration was obtained at a concentration of 10mM urea. Urea is not a common
cleaning agent for membrane fouling, but playes a particular role in solubilizing
insoluble or denatured proteins and eliminating the formation of albumin gels (Kelly
and Zydney, 1994). Therefore, urea could be a cheap alternative to clean the RO
membranes fouled by proteins.
5.2.2 Effects of cleaning solution pH
BSA-fouled membrane RO membranes were cleaned using SDS, urea and EDTA at
different pH when the flux decline to about of 54% of initial flux in the fouling
process of about 25 h. All the fouling runs were conducted at the same conditions (i.e.,
BSA concentration of 50 mg/L; ionic strength of 10 mM; and calcium concentration
of 1.0 mM; pH of 4.9; temperature of 25oC). The influence of solution pH on the
cleaning efficiencies of SDS, urea and EDTA is illustrated in Figure 5.8. It is shown
that cleaning efficiency increased noticeably with increasing pH from 4.5 to 10.6 for
SDS and urea, while there was almost no increase for EDTA. It has been reported that
the EDTA-calcium complex reaction is affected by the deprotonation degree of
carboxylic function groups of EDTA, which is related to pH (Ang et al., 2006). When
EDTA concentration was as low as 0.1 mM, EDTA was able to achieve a high
cleaning efficiency (above 95% of flux restoration) in a wide range of pH. The
anionic surfactant, SDS, is in its ionic form at the examined pH. Hence, pH has little
95
Protein fouling and chemical cleaning
effect on SDS itself. On the other hand, protein unfolds itself at high pH, exposing
more hydrophobic groups for binding interaction with the surfactant. Similarly, high
pH allows more opportunities for urea to bind with protein.
100
Flux restoration (%)
90
80
70
SDS 1mM
Urea 1mM
EDTA 0.1mM
60
50
4
5
6
7
8
9
10
11
pH
Figure 5.8 Variation of SDS, urea and EDTA cleaning efficiencies as a function of
cleaning solution pH (Cleaning conditions: cleaning time of 60 min; operating
pressure of 15 psi; crossflow velocity of 0.29 m/s; solution concentration of 1 mM for
both SDS and urea and 0.1 mM for EDTA).
5.2.3 Effects of cleaning time
BSA-fouled membranes were cleaned using SDS, and urea at different cleaning time
when the permeate flux dropped to about 54% of the initial flux in the fouling process
of about 25 h. All the fouling runs were also done at the same conditions (i.e., BSA
concentration of 50 mg/L; ionic strength of 10 mM; calcium of 1.0 mM; pH of 4.9;
and temperature of 25oC). The influence of cleaning time for SDS and urea was
96
Protein fouling and chemical cleaning
investigated and the results are showed in Figure 5.9. For SDS cleaning, an increase
in cleaning time from 15 to 60 min resulted in negligible difference in cleaning
efficiency when the SDS concentration was 1 mM. Interestingly, the cleaning time
became important when the SDS concentration increased to 5 mM. This influence of
cleaning time on cleaning efficiency subject to the chemical dosage is more obvious
for urea. Longer cleaning time resulted in much higher flux restoration (> 90%) when
the urea concentration increased from 1 to 10 mM.
100
Flux Restoration (%)
90
80
70
SDS 1mM
SDS 5mM
urea 1mM
urea 10mM
60
50
15
60
Time (min)
Figure 5.9 Variation of SDS and urea cleaning efficiencies as a function of cleaning
time (Cleaning conditions: operating pressure of 15 psi; crossflow velocity of 0.29
m/s; cleaning solution pH of 8.0 for SDS and 7.8 for urea).
97
Protein fouling and chemical cleaning
5.3 Summary
The effects of pH, cation species (Na+, Ca2+ and Mg2+), ionic strength and temperature
on protein fouling of RO membrane were investigated using biopolymer BSA in a
crossflow membrane cell. Zeta potentials of RO membrane surface in ionic solutions
at various pH revealed a shift in membrane surface IEP to a value similar to the IEP
of BSA when BSA was introduced into the solution. This observation clearly
demonstrated that BSA, with its “soft” structure and low internal stability, was readily
adsorbed onto the RO membrane surface regardless of the solution pH. Therefore,
instead of interactions between BSA and the fresh membrane, interactions between
BSA molecules and BSA-adsorbed membrane became the dominating mechanism for
fouling. Fouling experiments using the lab-scale crossflow membrane cell had shown
a higher rate of BSA fouling when the ionic strength was increased or calcium ion
was present in the ionic solution, due to increased charge-shielding effect that led to
weaker electrostatic repulsion between BSA molecules. In the presence of Mg2+,
fouling of RO membrane by BSA was not as severe as the ones observed with Ca2+ at
similar concentration. In addition, BSA fouling did not seem to be affected by the
change of pH in the presence of Mg2+. An important finding from this study was that
the most severe fouling occurred at pH near to the IEP of the BSA and this behavior
occurred for all the ionic strength and cation species tested.
This study also attempted to examine effects of temperature on protein fouling, which
was rarely an investigated parameter in RO process. Higher temperature was found to
increase the rate of BSA fouling at pH above the IEP of BSA. At pH below the IEP of
BSA, the fouling rates for different temperatures did not vary significantly. The
phenomenon was hypothetically attributed to the extent of unfolding of the compact
98
Protein fouling and chemical cleaning
BSA molecule at solution pH under different temperatures. Unfolding of the BSA
molecules exposed the charged functional groups to the aqueous solution and the
resulting charge shielding reduced the electrostatic repulsion between the BSA
molecules. The results pointed out the need to keep the operating temperature low
when the RO process was fouled largely under protein adsorption and accumulation.
The cleaning tests of three types of cleaning agents (i.e., SDS, urea, and EDTA) were
carried out to clean the membrane fouled by protein. EDTA was a very effective
cleaning agent through its preferential complexation with calcium. SDS and urea
performed cleaning through the solubilization of protein and the cleaning efficiencies
were affected by solution concentration and pH. At higher concentration, the cleaning
efficiency by urea could be substantially increased with a longer cleaning time.
99
Permeate behavior and concentration polarization in a long RO
membrane channel
Chapter 6
Permeate Behavior and Concentration
Polarization in a Long RO Membrane Channel
Behavior of a full-scale RO process has been described with basic transport theories
and models, which relate the permeate flux linearly with the driving pressure.
However, in some full-scale RO processes, it was observed that the average flux did
not increase linearly with driving force. Some studies attributed this phenomenon to
concentration polarization on the membrane surface (Turan, 2004), while some
studies proposed that a full-scale RO process was a heterogeneous system which
cannot be described by linear transport theories (Song et al., 2002). Both arguments
need to be supported by more evidences. Firstly, local permeate flux along the feed
channel should be determined to reflect the transport phenomenon along the
membrane channel. Secondly, the degree of concentration polarization should be
characterized along the membrane channel instead of measurement of a bulk permeate
such as the apparent flux decline. Finally, experimental approach should be
complemented to verify the theory deduced by numerical methodology.
In this chapter, the results of the experimental study of the behavior of permeate flux
in a long RO membrane channel are presented. A 1-m long channel RO membrane
cell was designed to study the variation of filtration performance in terms of local
permeate flux and salt rejection along the long channel. The local permeate flux
100
Permeate behavior and concentration polarization in a long RO
membrane channel
referred to the permeate water obtained in each 20-cm channel segment divided by the
respective filtration area. The fundamental membrane transfer principle and mass
balance were applied to each 20-cm channel segment to investigate how concentration
polarization grows along the channel. In addition, the relationship of the concentration
polarization and recovery are discussed. A good mastery of operation conditions and
membrane channel design would facilitate determining the optimal channel length to
control concentration polarization while maximizing recovery. Effects of spacer
including the additional perpendicular resistance and concentration polarization
reduction along the channel were also investigated and delineated.
6.1 Calculation of concentration polarization
Based on the measurable local permeate flux and salt concentration in the permeate of
a long RO membrane channel, the local concentration polarization modulus can be
calculated. The length and height of the membrane channel are indicated by L and H,
respectively. The feed flow (inlet crossflow velocity) and feed salt concentration are
denoted as u0 and cs0, respectively. Permeate flux and salt concentration averaged on
each 20-cm permeate segment along the channel are referred as the local variables
with the distance from the inlet.
According to the principle of membrane transport, permeate flux at any point along
the channel is linked to the net driving force by
v p (i ) 
p (i )   m (i )
Rm (i )
6.1
101
Permeate behavior and concentration polarization in a long RO
membrane channel
where νp(i) is the permeate flux experimentally measured (m/s), ∆πm(i) is the osmotic
pressure(Pa), Rm(i) is the hydrodynamic membrane resistance(Pa.s/m), and ∆p(i) is
the applied hydraulic pressure at permeation location i along the channel (Pa). In this
setup, the inside pressure in the feed side monitored by the 10 pressure transducers
along the channel indicated no significant transverse pressure variation along the
channel, hence ∆p(i) is constant (i.e., ∆p). Al-Bastaki and Addas (2000) also drew a
conclusion that ignoring pressure drops was not significant for spiral-wound
membrane after comparing the theoretical and experimental recovery. Besides, the
permeate side was open to the atmosphere and the pressure in the permeate side could
be assumed to be equal to one atm. Therefore, a constant applied pressure was
substituted for the hydraulic pressure over the whole channel. The net driving force
for water permeation is the difference between the applied hydraulic pressure and
osmotic pressure ∆πm(i), which is determined by (Song and Tay, 2006 )
 m (i )  c m (i )
6.2
where φ is the osmotic pressure coefficient (75.8 Pa.L/mg), and cm(i) is the salt
concentration at the membrane wall at location i(mg/L).
After rearranging Eq. 6.1 and 6.2, the net driving force, osmotic pressure and
membrane wall concentration can be determined since νp(i) can be measured
experimentally. Accordingly, concentration polarization modulus, CP(i) is also
determined by
102
Permeate behavior and concentration polarization in a long RO
membrane channel
CP (i ) 
c m (i )
cs 0
6.3
where cs0 is the salt concentration of the feed (mg/L).
By applying the mass conservation principle on water and salt, the cumulative
recovery and bulk concentration along the channel can be respectively expressed as
i
Rec (i ) 
v
n 1
p
(n)dx
u0 H
*100
6.4
and
i
c s (i ) 
c s 0 u 0 H   v p (n)dxc p (n)
n 1
i
u 0 H   v p (n)dx
6.5
n 1
where Rec(i) is the cumulative recovery, cs(i) and cp(i) are the salt concentration in the
bulk solution and permeate (mg/L), respectively, at permeate location i along the
channel, dx is the length of each permeate segment (m), which is equal to 0.2 m in this
study, and n is the segment number.
True rejection rtru(i) and observed rejection robs(i) at location i are defined respectively
as
103
Permeate behavior and concentration polarization in a long RO
membrane channel
rtru (i )  1 
c p (i )
c m (i )
6.6
and
robs (i )  1 
c p (i )
cs0
6.7
The concentration factor (CF) indicating the transverse concentration polarization
along the channel is determined by
CF (i ) 
c s (i )
cs0
6.8
6.2 Variation of permeate flux along the channel
Figure 6.1 plots the membrane resistance averaged over each 20-cm length in 1-m RO
membrane channel at the pressure from 50 to 350 psi.. It was found that membrane
resistance was not distributed randomly, but following a polynomial behavior with
higher resistance in two ends than in the middle, and higher resistance downstream
than upstream. The heterogeneity of membrane resistance can be ignored reasonably
for a lab-scale membrane cell with a small piece of membrane, but it is not the case
for a long and narrow piece of membrane. The mean resistance for the full membrane
104
Permeate behavior and concentration polarization in a long RO
membrane channel
used in this experiment was 9.17×1010 Pa.s/m with an error of 5.7%. The usage of
mean resistance is practical for calculating the global production in a full-scale RO
process (Song and Tay, 2006), but would yield potential errors caused by uneven
membrane resistance distribution when studying the local filtration performance along
the membrane channel. Therefore, for accuracy, the mean resistance was replaced by
the resistance polynomial distribution along the channel in this study for the
Resistance (Pa.s/m)
investigation of the local concentration polarization.
1.2x10
11
1.0x10
11
8.0x10
10
6.0x10
10
4.0x10
10
2.0x10
10
2
3
Y =9.64E10-5.53E10 X+7.92E10 X -1.78E10 X
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.1 Clean ESPA2 membrane hydrodynamic resistance averaged over a range
of pressure from 50-350 psi at different points along the channel.
105
Permeate behavior and concentration polarization in a long RO
membrane channel
The local permeate flux on each 20-cm permeate segment is plotted along the channel
as shown in Figure 6.2. The permeate flux in DI water filtration varied along the
channel in a polynomial manner due to the uneven distribution of membrane
resistance. The introduction of salt in the feed resulted in lower permeate fluxes due
to the osmotic pressure. A typical characteristic was that the dropped flux away from
corresponding DI water flux at the same applied pressure became larger from the inlet
to the outlet (for example, Flux dropped by 18.3% in the inlet and 36.4% in the outlet
for the case of 300 psi pressure and 0.091 m/s feed flow), which indicates that the salt
concentration build up in the boundary layer on the membrane surface was not even
along the channel.
Figure 6.2 also shows the different effects of pressure and feed flow on the permeate
flux along the membrane channel. At the upstream of the channel, the permeate flux
increased with applied pressure because a larger net driving force was supplied by
increasing the applied pressure (Figure 6.3a). Whereas at the end of channel, the
increased applied pressure from 140 to 300 psi cannot offer the same increment of the
net driving force for permeate (increase of 7.29×105 Pa) as it did in the inlet (increase
of 9.74×105 Pa), which indicates that the contribution of the applied pressure in
driving the permeate flux was fading along the channel due to increasing osmotic
pressure. On the other hand, as shown in Figure 6.2b the permeate flux was almost not
affected by the feed flow until the channel length was longer than 0.5 m, after which
the permeate flux decreased as feed flow decreased. This is because a lower feed flow
caused greater concentration buildup on the membrane surface at the downstream of
the channel, resulting in lesser effective driving force on the permeate accordingly.
106
Permeate behavior and concentration polarization in a long RO
membrane channel
-5
3.0x10
300 psi
260 psi
220 psi
(a)
-5
Permeat flux (m/s)
2.5x10
180 psi
140 psi
-5
2.0x10
-5
1.5x10
-5
1.0x10
-6
5.0x10
0.0
0.2
0.4
0.6
0.8
1.0
Permeate flux (m/s)
Channel length (m)
3.0x10
-5
2.5x10
-5
2.0x10
-5
1.5x10
-5
1.0x10
-5
5.0x10
-6
(b)
0.0
DI water
0.046 m/s
0.091 m/s
0.226 m/s
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.2 Permeate flux variation along the channel for DI water filtration (void
symbol) and salt solution filtration (feed concentration of 30 mM, solid symbol) at (a)
different pressure (feed flow of 0.091 m/s) and (b) different feed flow (applied
pressure of 300 psi).
107
Permeate behavior and concentration polarization in a long RO
membrane channel
2.0x10
6
1.6x10
6
1.2x10
6
8.0x10
5
5
5
4.0x10
5
300 psi
260 psi
220 psi
(a)
6
2.0x10
180 psi
140 psi
1.6x10
6
1.2x10
8.0x10
4.0x10
0.0
0.0
0.2
0.4
0.6
Osmotic pressure (Pa)
Net driving force (Pa)
6
0.0
1.0
0.8
(b)
2.0x10
6
1.6x10
6
1.2x10
6
2.0x10
6
1.6x10
6
1.2x10
6
5
8.0x10
5
8.0x10
4.0x10
5
4.0x10
5
0.0
0.0
0.046 m/s
0.091 m/s
0.226 m/s
0.2
0.4
0.6
0.8
Osmotic pressure (Pa)
Net driving force (Pa)
Channel length (m)
0.0
1.0
Channel length (m)
Figure 6.3 Net driving force on permeate (solid symbol) and trans-membrane osmotic
pressure (void symbol) variation along the channel for salt solution filtration (feed
concentration of 30 mM) at different (a) applied pressure (feed flow of 0.091 m/s) and
feed flow (applied pressure of 300 psi).
108
Permeate behavior and concentration polarization in a long RO
membrane channel
6.3 Variation of rejection along the channel
Membrane separation efficiency could be characterized by the membrane true
rejection and observed rejection. The true rejection depends on the feed solution,
membrane characteristics and operating pressure (Bouchard et al., 1994). Figure 6.4a
shows that high pressure resulted in higher true rejection due to the different driving
forces on water flux and solute flux, that is, pressure difference and concentration
difference, respectively. The feed flow had no effect on the true rejection (Figure
6.4b).
Figure 6.5 shows a general trend of the observed rejection decline along the channel
and the effect of operating conditions on this trend. The rejection slightly increased
with applied pressure upstream but decreased with pressure when the channel length
was longer than 0.8 m (Figure 6.5a). This could be attributed to the influence of the
uneven formation of concentration polarization along the channel on the different
driving forces for water flux and solute flux. In the upstream zone where
concentration polarization was weak, water flux increased with increasing pressure
while solute flux remained almost unchanged due to a relatively minor build-up of
boundary concentration; consequently, the rejection was increased. However, at the
downstream zone where concentration polarization had become severe, water flux
increase with increasing pressure was less than upstream while solute flux became
higher due to the higher salt concentration in the boundary layer near the membrane
surface. As a result, the rejection became lower when pressure increased for a long
membrane channel. Increasing feed flow could relieve the concentration polarization
downstream, leading to the increased rejection with feed flow as shown in Figure 6.5b.
109
Permeate behavior and concentration polarization in a long RO
membrane channel
1.00
(a)
True rejection
0.99
0.98
0.97
0.96
0.95
5
8.0x10
1.2x10
6
1.6x10
6
2.0x10
6
Pressure (Pa)
1.00
(b)
True rejection
0.99
0.98
0.97
0.96
0.95
5
10
15
20
25
Feed flow (m/s)
Figure 6.4 True rejection for salt solution filtration (NaCl of 30 mM) at (a) different
applied pressure (feed flow of 0.091 m/s) and (b) feed flow (applied pressure of 300
psi)
110
Permeate behavior and concentration polarization in a long RO
membrane channel
1.00
(a)
300
260
220
180
140
Observed rejection
0.99
0.98
psi
psi
psi
psi
psi
0.97
0.96
0.95
0.94
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
1.00
(b)
0.046 m/s
0.091 m/s
0.226 m/s
Observed rejection
0.99
0.98
0.97
0.96
0.95
0.94
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.5 Observed rejection variation along the channel for salt solution filtration
(feed concentration of 30 mM) at different (a) applied pressure (feed flow of 0.091
m/s) and (b) feed flow (applied pressure of 300 psi).
111
Permeate behavior and concentration polarization in a long RO
membrane channel
In comparison with the parametric study in modeling membrane separation conducted
by Bouchard et al. (1994), this experimental work shows more specific information of
membrane separation along the channel. In addition, our results suggest that particular
attention should be paid to both the channel length of membrane cell and operating
conditions in RO process to achieve a better salt rejection. If the overall rejection is
not satisfactory upon adjustment of operating condition, redesign of the channel
length (that is, to change the number of membrane modules in a pressure vessel)
would improve the separation performance.
6.4 CP and CF variation along the channel
Earlier results have shown the impact of concentration polarization on the local
permeate flux and rejection variation along the channel. More delineation on
concentration polarization along the channel will be given in this chapter.
Figure 6.6 shows the variation of CF (Concentration factor) and CP (Concentration
polarization) modulus along the channel. CF increased linearly for the full channel
while CP escalated exponentially. CF represented the increase in concentration of
bulk salt solution along the channel which also contributed to the enhanced osmotic
pressure; therefore it was convenient to predict permeate flux using CF or the salt
concentration in concentrate stream when designing and operating a process (Song
and Tay, 2006). Compared with CF, however, CP was far higher than CF in
magnitude which resulted in the sharp decrease of the permeate flux and rejection
downstream. Therefore, CP should be an indispensable element in design and
operation of RO process. A high CF is desirable while a low CP is preferred. In order
112
Permeate behavior and concentration polarization in a long RO
membrane channel
to achieve this goal, feed spacer or other turbulence promoter is applied in membrane
modules (Ahmad, 2005).
Figure 6.6 also demonstrates the CP development under different operating conditions.
CP at the inlet increased with pressure due to more salt accumulation on the boundary
layer at higher pressure. Additionally, the CP modulus incremental rate with pressure
was intensively enhanced along the channel as a result of continuous accumulation of
salt. For the different feed flow, the values of CP modulus at the inlet were very close
to each other as they were subjected to the same applied pressure, which imply that
the effect of the feed flow on the starting point was negligible. However, feed flow
exerted a significant role on the CP modulus incremental rate about half way along
the channel length with lower feed flow obviously associated with larger CP modulus
incremental rate for a long channel.
The concentration polarization profiles observed along the channel suggests that the
channel length should be shortened to minimize concentration polarization. However,
a short channel will reduce water recovery. Therefore, a better understanding of the
correlation between concentration polarization and recovery is needed to resolve this
contradicting effect and facilitate determination of an optimal channel length.
113
Permeate behavior and concentration polarization in a long RO
membrane channel
6
(a)
300 psi
260 psi
220 psi
180 psi
140 psi
5
CP, cm/cs0
4
5
4
3
3
2
2
1
0.0
0.2
0.4
0.6
0.8
1.0
CF, cs/cs0
6
1
Channel length (m)
7
(b)
6
5
CP , cm/c0
6
0.046 m/s
0.091 m/s
0.226 m/s
5
4
4
3
3
2
2
1
0.0
CF, cb/c0
7
1
0.2
0.4
0.6
0.8
1.0
0
Channel length (m)
Figure 6.6 CP (solid symbol) and CF (void symbol) variation along the channel for
salt solution filtration (feed concentration of 30 mM) at different (a) applied pressure
(feed flow of 0.091 m/s) and (b) feed flow (applied pressure of 300 psi).
114
Permeate behavior and concentration polarization in a long RO
membrane channel
6.5 Correlation between recovery and CP
Figures 6.7a and 6.8a show the cumulative recovery with channel length at different
applied pressure and feed flow. The recovery increased linearly with channel length
for the first half of the channel. Towards the second half of the channel, the recovery
changed only slightly because of the sharp drop of permeate flux (for example, from
1.87×10-5 m/s at the inlet to 1.38×10-5 m/s at the outlet for 300 psi pressure and 9.14
m/s feed flow as shown in Figure 6.2). In addition, cumulative recovery increased
when applied pressure was increased or feed flow was decreased.
The CP modulus presented in Figure 6.6 was re-plotted against the cumulative
recovery as shown in Figures 6.7b and 6.8b. At the initial stage of recovery
accumulation as shown in Figure 6.7b, CP value was larger at higher pressure. This is
because higher pressure resulted in considerable concentration polarization since
more salt was dragged onto the membrane surface by higher permeate flux in the 1-D
permeate direction. However, while the recovery increased further using an extended
filtration channel, CP incremental rate displayed a similar profile for all the tested
pressures. This could be attributed to the 2-D competitive interplay of permeate flux
and tangential crossflow instead of the effect of 1-D permeate flux only. When
applied pressure is low, permeate flux is low but crossflow is relatively high to
balance the flow according to mass balance. Higher crossflow will move more salt
downstream, leading to severe concentration polarization in the extended channel for
a required recovery. Therefore, for a certain recovery, the same amount of salt will be
retained by the membrane due to the same feed stream and thus the same
concentration polarization will be build up at all pressure. Compared to high pressure,
115
Permeate behavior and concentration polarization in a long RO
membrane channel
low pressure mitigated only the initial concentration polarization in a short feed
channel and subsequently led to more heterogeneous distribution of concentration
accumulation in an extended channel, but neither eliminating nor retarding
concentration polarization. Such a concentration accumulation behavior in long
channel was also reported in hollow fiber membrane module (Bessiere, 2008).
Figure 6.8b further highlighted the interaction between permeate flux and crossflow in
the concentration polarization formation at different feed flows. The initial values of
CP modulus for different feed flows were similar because the same operating pressure
was applied and the resulting similar permeate flux brought the same amount of salt
onto the membrane surface. For a higher recovery achieved in an extended channel,
the higher feed flow produced more water and more salt was retained on the
membrane surface. As a result, a faster CP incremental rate was observed with higher
crossflow. It is reasonable to conjecture that the feed flow determined the
concentration polarization incremental rate with cumulative recovery, that is, higher
feed flow resulted in a higher concentration polarization incremental rate for a
designed recovery in a long channel.
116
Permeate behavior and concentration polarization in a long RO
membrane channel
25
(a)
300
260
220
180
140
Cumulative recovery, %
20
15
psi
psi
psi
psi
psi
10
5
0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
6
(b)
5
CP, cm/cs0
4
3
300 psi
260 psi
220 psi
180 psi
140 psi
2
1
0
5
10
15
20
25
Cumulative recovery, %
Figure 6.7 (a) Cumulative recovery and (b) correlation between cumulative recovery
and CP modulus for salt solution filtration at different applied pressure (feed
concentration of 30 mM and feed flow of 0.091 m/s).
117
Permeate behavior and concentration polarization in a long RO
membrane channel
35
(a)
Cumulative recovery, %
30
0.046 m/s
0.091 m/s
0.226 m/s
25
20
15
10
5
0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
7
(b)
6
CP, cm/c0
5
4
3
0.046 m/s
0.091 m/s
0.226 m/s
2
1
0
5
10
15
20
25
30
35
Cumulative recovery, %
Figure 6.8 (a) Cumulative recovery and (b) correlation between cumulative recovery
and CP modulus for salt solution filtration at different feed flow (feed concentration
of 30 mM and applied pressure of 300 psi).
118
Permeate behavior and concentration polarization in a long RO
membrane channel
The effects of operating conditions on the correlation between recovery and CP
suggest that it is too general to say that low pressure and high cross flow reduce
concentration polarization, which may be practical in small RO cell but not always
suitable in a long membrane channel. Water recovery, an important design objective
of RO process, is determined by the filtration area (channel length) and operating
conditions. Hence, there are two means to achieve the recovery. One means is to
design a longer channel and adopt low pressure and high feed flow, which resulted in
less operating cost but at the expense of more capital investment on system
installment and more concentration polarization. The other way is to design a shorter
channel operated under high pressure and low feed flow, which consume more energy
but less concentration polarization is formed. Hence, apart from the cost on the
system installment and operation, the loss of membrane permeability caused by
concentration polarization should be taken into account when designing a RO system
to maximize the membrane permeability.
6.6 Permeate performance in a spacer-filled channel
Concentration polarization has been observed to grow exponentially along the 1-m
long membrane channel without feed spacer inserted. The effects of lateral mixing
were investigated using one rectangular spacer inserted into the long crossflow
channel from the inlet to the outlet. The spacer dimension was 1000×30×0.71 mm and
the pitch was 4 mm. The spacer filament was oriented at 45o to the direction of
crossflow. As the spacer height was less than the channel height, the spacer used were
simply inserted into the channel freely and not fixed to the membrane surface. It was
noted that the spacer did not cause significant difference in horizontal pressure drop
119
Permeate behavior and concentration polarization in a long RO
membrane channel
along the 1-m channel as monitored by pressure sensors compared to the spacer-free
channel.
Permeate fluxes obtained along the membrane when the RO membrane was used to
filter salt solutions of different concentrations (0, 10, 30, and 60 mM) are shown in
Figure 6.9. When no salt was present in the feed water, i.e., using DI water, the
presence of a spacer in the channel resulted in a lower permeate flux (reduction of 6%)
in the upstream, but no difference was observed in the downstream of the channel.
This indicates that the spacer adds additional vertical hydrodynamic resistance on the
permeate due to a reduction of effective membrane area. Because the spacer was not
fixed tightly on the membrane surface but floated between the membrane and the
impermeable wall, it is susceptible to the characteristic of the feed flow. In the inlet,
the spacer was forced and pressed against the membrane by the inlet steam; in the
outlet, the spacer was flushed away from the membrane by the outlet stream. As a
result, the spacer exerted different influence on the effective membrane area and
added more vertical resistance upstream in this study although such a phenomenon
can be avoided in a full-scale RO module. However, the effective membrane areas
reduction caused by spacer attaching to the membrane should not be neglected when
designing and optimizing a spacer.
When salt was introduced into the feed solution, concentration polarization developed
along the channel, which resulted in a decreasing local permeate flux distribution
along the channel as shown in Figure 6.9. The CP moduli were computed and the
results are shown in Figure 6.10. Inserting a spacer lessened intensively the formation
120
Permeate behavior and concentration polarization in a long RO
membrane channel
of concentration polarization due to its mixing effect, resulting in a thinner
concentration boundary layer and a lower CP modulus along the membrane surface.
Therefore, it is expected that the permeate flux with a spacer inserted would surpassed
one without a spacer inserted as concentration polarization was weaken significantly
(up to 50% reduction of CP) in the spacer-filled channel.
-5
2.4x10
-5
2.2x10
-5
Flux (m/s)
2.0x10
-5
1.8x10
-5
1.6x10
-5
1.4x10
-5
1.2x10
DI water
10 mM
30 mM
60 mM
DI water_spa
10 mM_spa
30 mM_spa
60 mM_spa
-5
1.0x10
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.9 Permeate flux variation along the channel with or without a spacer inserted
(applied pressure of 300 psi and feed flow of 0.091 m/s).
It is also noted that for the low feed salt concentration of 10 mM, inserting a spacer
decrease the permeate flux by 5% because the spacer contributed an extra resistance
in the upstream part of the channel (< 0.4 m in this study). On the contrary, at higher
salt concentration, 30-60 mM, the effect of a spacer on suppressing concentration
boundary layer was further enhanced. Accordingly, the benefit of concentration
121
Permeate behavior and concentration polarization in a long RO
membrane channel
polarization reduction was sufficient to offset deficiency of the additional resistance,
which resulted in an increase of permeate flux over the whole membrane channel with
a spacer (See Figure 6.9). Therefore, a spacer created a competition between an
additional vertical resistance and a reduction of concentration polarization, and the
overall net effect is positive especially when the salt concentration is high in the feed
water.
11
10
9
10 mM
30 mM
60 mM
10 mM_spa
30 mM_spa
60 mM_spa
8
CP, cm/cs0
7
6
5
4
3
2
1
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.10 CP modulus growth along the channel with or without a spacer inserted
(applied pressure of 300 psi and feed flow of 0.091 m/s).
As a result of concentration polarization reduction by the spacer along the channel,
the recovery and concentration factor were higher with a spacer than without a spacer
for feed concentrations greater than 10 mM (Figure 6.11). This effect was more
obvious when the channel was longer. Another performance indicator of RO process
122
Permeate behavior and concentration polarization in a long RO
membrane channel
was the observed salt rejection. Salt rejection was observed to decrease intensively
without a spacer due to salt accumulation along the channel. With a spacer, salt
rejections were increased intensively, particularly downstream of the channel,
ensuring an almost homogenous rejection over the whole channel (Figure 6.12).
1.35
10 mM
30 mM
60 mM
1.30
1.25
10 mM_spa
30 mM_spa
60 mM_spa
CF, cs/cs0
1.20
1.15
1.10
1.05
1.00
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 6.11 CF variation along the channel with or without a spacer inserted (applied
pressure of 300 psi and feed flow of 0.091 m/s).
123
Permeate behavior and concentration polarization in a long RO
membrane channel
0.995
0.990
Observed rejection
0.985
0.980
0.975
0.970
0.965
10 mM
30 mM
60 mM
10 mM_spa
30 mM_spa
60 mM_spa
0.2
0.4
0.960
0.0
0.6
0.8
1.0
Channel length (m)
Figure 6.12 Observed rejection variation along the channel with or without a spacer
inserted (applied pressure of 300 psi and feed flow of 0.091m/s).
6.6 Summary
This is the first experimental study to investigate the permeate characteristics in a
long RO membrane channel by designing and operating a 1-m long continuous
channel RO membrane cell with local permeate flux measurement when membrane
was used to filter salt solution. It was observed that both local permeate flux and
rejection decreased along the channel due to concentration polarization exponential
growth with the channel length.
The results also showed that effects of operating conditions on permeate
characteristics were complicated by the long channel. Higher applied pressure
124
Permeate behavior and concentration polarization in a long RO
membrane channel
increased the permeate flux, but the degree of increment of permeate flux decreased
downstream since more severe concentration polarization and lower resulting net
driving force occurred along the channel. Higher feed flow reduced concentration
polarization downstream but had almost no effect on permeate flux upstream. The
observed rejection was also affected by the coupling of applied pressure and feed flow
in a long channel filtration system. Higher applied pressure increased the rejection
upstream but was inversely affected downstream while higher feed flow increased the
rejection downstream and had no effect on the rejection upstream. These results imply
that special attention should be paid to the variation of the effects of operating
conditions along the membrane channel in order to achieve high production and good
separation.
The correlation between concentration polarization and the cumulative recovery under
various operating conditions was evaluated. Cumulative recovery increased with
channel length. For a designed recovery (> 10%), decreasing applied pressure did not
reduce concentration polarization but developed a heterogeneous development of
concentration polarization profile along the channel. In addition, feed flow played a
significant role on concentration polarization incremental profile. The higher the feed
flow, the more severe was the concentration polarization development with certain
recovery. These novel findings relating concentration polarization and cumulative
recovery could offer a better picture of concentration polarization in a long channel
RO filtration, which will be useful in designing a long life-span and high performance
RO process for practical industrial application by balancing the installment and
operating cost.
125
Permeate behavior and concentration polarization in a long RO
membrane channel
Insertion of spacers in the membrane channel could suppress the formation of severe
concentration polarization, thus promoting even distribution of local permeate and salt
rejection along the membrane channel. However, in the presence of a spacer, an
additional perpendicular resistance on the permeate flux would occur although the
difference of horizontal pressure drop was negligible.
126
Organic fouling development in a long RO membrane channel
Chapter 7
Organic Fouling Development in a Long RO
Membrane Channel
Prediction of organic fouling development along the RO membrane channel is
important to optimize RO process design and to adopt proper fouling control
measures. Most membrane fouling models developed for homogenous membrane
system assume uniform flow properties and fouling rate throughout the membrane
surface. Chapter 6 presented data that showed even in a 1-m membrane channel, the
flow and concentration polarization changed substantially along the channel, which
would affect fouling development. Hence, those models simplify the actual flow
characteristics for a full-scale RO process. This issue can potentially be resolved by
using two new models that were developed to describe the fouling characteristics of a
full-scale RO process (Chen et al., 2004; Hoek et al., 2008). These two models
worked impressively to predict the performance of a full-scale RO process, but some
assumption and claims made remained doubtful. These two models, based on simple
deposition of foulant on the membrane surface, did not incorporate a variety of
significant interactions involved in organic fouling. Therefore, more realistic
assumptions and definitions are needed in the modeling of the organic fouling
development in a long RO membrane channel.
127
Organic fouling development in a long RO membrane channel
This chapter presents research work on organic fouling development in 1-m long RO
membrane channel in two steps. The first step was to experimentally observe the
fouling behavior of alginate and BSA along the long RO membrane channel.
Important operating conditions including initial flux and crossflow velocity were
investigated and their effects on the organic fouling development along the channel
were discussed. All the fouling process was initiated after DI water compaction and
salt condition, i.e., the initial local permeate flow decreased and concentration
polarization of salt increased along the channel according to the study presented in
Chapter 6. The second step was to theoretically propose several key indexes to
improve previous modeling of organic fouling and to compare the experimental
results of alginate fouling with numerical results.
7.1 Organic fouling development along the channel
7.1.1 The permeate behavior in alginate and BSA fouling along the RO
membrane channel
Figure 7.1a shows the local permeate flux at five different permeate points along the
1-m RO membrane channel in alginate fouling process. The local permeate flux
defined as the flux averaged over each 20-cm filtration section. For each permeate
point, flux behavior was characterized by the same profile with one in the small RO
cell presented in Chapter 4 - rapid decline initially and gradually reached a (pseudo)state flux. Due to concentration polarization of salt becoming more severe along the
channel, the initial flux decreased along the channel as discussed in Chapter 6. Hence,
the flux was normalized to the corresponding initial flux in order to compare the
differences of fouling rate for each permeate point along the channel. Figure 7.1b
shows the final normalized flux was lower downstream than upstream, which
128
Organic fouling development in a long RO membrane channel
indicates that organic fouling developed faster downstream than upstream. This
observation of permeate flux becoming lower downstream was coincident with some
reports of more foulant deposition in the latter elements in a RO train (Schneider et al.,
2005).
Initial flux could be a possible factor affecting the organic fouling development along
the channel since the local initial flux at each permeate point was different along the
channel and was lower downstream, thus bringing less foulant to the membrane
surface and leading to less fouling towards the end of the channel. Hence, the
contribution of the initial flux to the fouling development conflicted with the
observation of alginate fouling development in this study. Therefore, another possible
factor for alginate fouling becoming more severe along the channel is the
concentration polarization of salt and organic foulant. On one hand, the increasing
concentration of organic foulant might directly contribute to faster flux reduction
along the channel but this effect was not significant because the organic concentration
was much less than salt concentration (Tang, 2007). On the other hand, elevated salt
concentration downstream led to the formation of a compact and thick fouling layer
on the membrane surface due to charge-screening and alginate-calcium interactions,
leaving less free alginate in the bulk stream and inducing more foulant deposition in
the downstream of the channel (Grant et al., 1973; Lee et al., 2006).
129
Organic fouling development in a long RO membrane channel
2.4x10
-5
Flux (m/s)
(a)
2.0x10
-5
1.6x10
-5
1.2x10
-5
8.0x10
-6
4.0x10
-6
0.2 m
0.4 m
0.6 m
0.8 m
1.0 m
0
5
10
15
20
25
Time (h)
1.0
(b)
Nornalized flux, J/J0
0.8
0.6
0.2m
0.4m
0.6m
0.4
0.8m
1.0m
0.2
0
5
10
15
20
25
Time (h)
Figure 7.1 Permeate flux (a) and normalized flux (b) evolution over 25 h of alginate
fouling at different locations along the channel (Fouling conditions: alginate of 50
mg/L; ionic strength of 10 mM; calcium of 1.0 mM; unjusted pH at 6.0±0.1;
temperature of 25oC; initial flux of 2.00×10-5 m/s; crossflow velocity of 0.09 m/s).
130
Organic fouling development in a long RO membrane channel
A previous study has shown that severe alginate fouling occured at high salt
concentration and in the presence of divalent cation (Lee et al., 2006). Our study
presented in Chapter 4 showed the importance of calcium concentration determining
the capacity to binding with alginate. When salt and calcium concentration increased
along the channel, the ratio of foulant (alginate) sticking onto the membrane over the
total foulant being convected towards the membrane surface increased along the
channel. This ratio is significant in investigating the organic fouling development
quantitatively.
BSA is another biopolymer and had different fouling behavior from alginate in the
small lab-scale RO membrane cell as presented in Chapter 5. Figure 7.2 shows the
local permeate flux profile at five different permeate points along the long RO
membrane channel in the BSA fouling process. Compared with alginate, BSA fouling
was not as severe as alginate since the total flux over the whole channel declined by
12% (Figure 7.2b) at most while the declined for alginate was 72% (Figure 7.1b) after
25-h fouling process. In addition, a more significant difference between BSA fouling
development and alginate fouling development along the long RO channel was
observed in terms of normalized flux as shown in Figure 7.2b. The flux declined
slightly faster upstream than downstream in BSA fouling development, in contrast to
alginate fouling development where the flux declining became faster towards the end
of the channel. However, BSA fouling development along the long channel is also in
agreement with the report of more protein deposition in the former elements in a RO
train (Schneider et al., 2005).
131
Flux (m/s)
Organic fouling development in a long RO membrane channel
2.2x10
-5
2.0x10
-5
1.8x10
-5
(a)
0.4m
0.6m
0.2m
0.8m
1.6x10
-5
1.0m
1.4x10
-5
0
5
10
15
20
25
Time (h)
1.0
(b)
1.0m
0.8m
0.6m
0.4m
0.2m
Normalized flux, J/J0
0.9
0.8
0.7
0.6
0
5
10
15
20
25
Time (h)
Figure 7.2 Permeate flux (a) and normalized flux (b) evolution over 25 h of protein
fouling at different locations along the channel (Fouling conditions: BSA of 50 mg/L;
ionic strength of 10 mM; calcium of 1.0 mM; unjusted pH at 6.0±0.1; temperature of
25oC; initial flux of 2.00×10-5 m/s; crossflow velocity of 0.09 m/s).
132
Organic fouling development in a long RO membrane channel
One possible explanation for this difference could be the relatively weaker
interactions between BSA and the membrane, which could be observed by comparing
the effect of salts on alginate and BSA fouling in Figure 4.1 and Figure 5.4. One
published paper also pointed out that polysaccharides adsorbed on the RO membrane
much more effectively than protein due to specific interactions between biopolymers
and salt (Herzberg et al., 2009). Hence, the increase of salt concentration downstream
would not substantially increase the amount of proteins adhering onto the membrane
surface. On the other hand, higher permeate flux upstream brought more protein
towards the membrane surface and create a more severe protein fouling layer. On
contrary to alginate fouling development, the initial flux therefore seemed to be
predominant in determining BSA fouling development along the RO membrane
channel. This suggests the influence of the flow and salt concentration variation along
the channel on the organic fouling are susceptible to the organic type and involved
interactions.
7.1.2 Effects of operating conditions on organic fouling development along the
channel
Operating conditions are of interest because it is easy to change operating conditions
to create different hydrodynamic conditions to study fouling at the macro-level and to
control fouling in practice. Hence, organic fouling tests were conducted to investigate
alginate and BSA fouling development along the RO membrane for different initial
fluxes (the total average flux over the whole channel) and crossflow velocities.
Effects of the initial flux on organic fouling have been shown by a number of fouling
tests in small lab-scale membrane cells (Hong and Elimelech, 1997; Lin et al., 2008).
133
Organic fouling development in a long RO membrane channel
This study investigated the effects of the initial flux (0.97×10-5, 1.32×10-5, and
2.00×10-5 m/s) on organic fouling development in the long channel RO membrane
cell. Figure 7.3 plots the fouling extent in terms of the normalized flux at each
permeate point after 25-h fouling process by alginate or BSA under different initial
flux. The normalized flux at a specific location was computed based on the ratio of
the flux at the end of 25-h fouling process and the respective initial permeate flux at
that permeate point.
The results show that higher initial flux resulted in severer fouling as identified by
previous studies. This might be mainly due to higher salt concentration under higher
initial flux (or higher applied pressure) as discussed in Chapter 6, which will
contribute to more alginate deposition. In addition, effect of the initial flux on the
development of organic fouling along the channel was affected by the long channel
effect. For alginate fouling, the decreased normalized flux almost remains the same at
low initial flux but had a difference of 4% at high initial flux after 25-h fouling
process in the first 0.2 m of the channel. Whereas for the latest 0.2 m of the channel,
the difference in the decreased normalized flux increased to 29% from 0 for low
initial flux but was only 8% for high initial flux. These results indicated the low initial
flux did not reduce global fouling development, but redistributed foulant deposition
on the membrane surface along the channel, either on the upstream or the downstream
of channel within the hydrodynamic retention time of foulant inside of the membrane
channel. This is also right for the effects of the initial flux on BSA fouling along the
channel. The difference in the decreased normalized flux along the channel was
bigger at the low initial flux than at high initial flux.
134
Organic fouling development in a long RO membrane channel
1.0
(a)
Normalized flux, J/J0
0.8
0.6
0.4
0.2
0.0
0.0
Initial flux:
-5
0.97x10 m/s
-5
1.32x10 m/s
-5
2.00x10 m/s
0.2
0.4
0.6
0.8
1.0
Channel length (m)
(b)
Normalized flux, J/J0
1.00
0.95
0.90
Initial flux:
-5
0.97x10 m/s
-5
1.32x10 m/s
-5
2.0x10 m/s
0.85
0.80
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 7.3 Effects of the initial flux on alginate (a) and BSA (b) fouling along the
channel (Fouling conditions: alginate or BSA of 50 mg/L; ionic strength of 10 mM;
calcium of 1.0 mM; unjusted pH at 6.0±0.1; temperature of 25oC; crossflow velocity
of 0.09 m/s).
135
Organic fouling development in a long RO membrane channel
The effect of crossflow velocity on the alginate and BSA fouling development along
the 1-m long RO membrane channel is shown in Figure 7.4, which plots the fouling
extent in terms of the normalized flux at each permeate point after 25-h organic
fouling process under different crossflow velocities (0.03, 0.09, 0.23, and 0.36 m/s for
alginate and 0.09 and 0.23 m/s for BSA). It was noted that the effect of crossflow
velocity on organic fouling was complicated by the channel effect. In the upstream of
the channel, lower crossflow velocity resulted in more severe fouling extent for both
alginate and BSA due to lower shear rate sweeping less foulant away from the
membrane surface, which is consistent with most of the other studies without
considering the membrane channel effect (Chilukuri, 2001; Hong and Elimelech,
1997). After the middle point of the channel, the fouling trend reversed, that is, higher
crossflow velocity contributed to more severe fouling. This is possibly due to the
higher crossflow velocity sweeping more foulant into the bulk solution in the
upstream of the membrane channel, which enabled the downstream permeate flux to
bring more foulant to the membrane surface.
Hence, the effect of crossflow velocity is to change the foulant distribution along the
membrane channel instead of mitigating the fouling for a certain membrane channel.
In the tested range of crossflow velocity in this study, higher crossflow velocity
moved and deposited more foulant towards the outlet of the channel. As a result,
higher crossflow velocity increased the heterogeneity of the development of alginate
fouling along the membrane channel but lowered the heterogeneity of the
development of BSA fouling. In addition, such an effect becomes more important
when membrane channel is longer.
136
Organic fouling development in a long RO membrane channel
1.0
(a)
Normalized flux, J/J0
0.8
0.6
0.4
Crossflow velocity:
0.03 m/s
0.09 m/s
0.23 m/s
0.36 m/s
0.2
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Normalized flux, J/J0
1.00
(b)
0.95
0.90
Crossflow velocity:
0.09 m/s
0.23 m/s
0.85
0.80
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 7.4 Effects of crossflow velocity on (a) alginate and (b) BSA fouling along the
channel (Fouling conditions: alginate or BSA of 50 mg/L; ionic strength of 10 mM;
calcium of 1.0 mM; unjusted pH at 6.0±0.1; temperature of 25oC; initial flux of
1.32×10-5 m/s).
137
Organic fouling development in a long RO membrane channel
7.1.3 Effects of feed spacer(s) on alginate fouling
The impact of spacer on the filtration of salt solution in the long RO membrane
channel has been discussed in Chapter 6, which shows that inserting a spacer
intensively suppressed the formation of concentration polarization, especially in the
rear part of the membrane channel where concentration polarization developed faster
in the absence of a spacer. The effect of inserting a spacer in the feed channel on
alginate fouling development was also investigated by inserting a similar piece of
spacer as the one used in the study in Chapter 6.
Figure 7.5 shows the results in terms of the normalized flux at the end of 25-h fouling
test in the long RO membrane channel in the presence of one or two spacers
compared with that without a spacer. As expected, the flux in the spacer-filled
channel was significantly higher by about 20% than those in the spacer-free channel
while flux was increased by about 30% in the two-spacers-filled channel after 25-h
alginate fouling. The beneficial effect of the spacer could be ascribed to the reduction
of salt and alginate accumulation on the membrane surface, which resulted in less
alginate depositing on the membrane surface. Two spacers promoted stronger
turbulence and further reduced salt and alginate concentration polarization more than
one spacer, resulting in much lesser alginate fouling.
However, Figure 7.5 does not reveal the difference in the extent of fouling reduction
along the channel although concentration polarization was reduced much more
downstream than upstream when a spacer was inserted in the membrane channel. This
disagreement might be caused by the adverse effect of the spacer. Inserting a spacer
will change local hydraulic conditions where the spacer filament contacts the
138
Organic fouling development in a long RO membrane channel
membrane surface, thus leading to different distribution of foulant deposition (Neal et
al., 2003). Hence, the optimization of spacer is necessary in order to minimize organic
fouling development in a heterogeneous RO membrane channel.
1.0
No spacer
One spacer
Two spacers
Normalized flux, J/J0
0.8
0.6
0.4
0.2
0.0
0.0
0.2
0.4
0.6
0.8
1.0
Channel length (m)
Figure 7.5 Effects of feed spacer on the alginate fouling development along the
spacer-free channel and channel with one or two spacers inserted (Fouling conditions:
alginate of 50 mg/L; ionic strength of 10 mM; calcium of 1.0 mM; unjusted pH at
6.0±0.1; temperature of 25oC; initial flux of 2.00×10-5 m/s; crossflow velocity of 0.09
m/s).
7.2 Key factors in organic fouling development in a long
membrane channel and numerical study
Our experimental study has shown that organic fouling does not developed evenly
along the RO membrane channel. The degree of organic foulant adhering on the
membrane surface is determined by interactions between foulant and between foulant
139
Organic fouling development in a long RO membrane channel
and membrane as shown in Chapter 4 & 5. The dynamic study of the long RO
membrane channel in Chapter 6 shows the flow and salt concentration variation along
the membrane. In addition, the flow and salt concentration variation along the channel
affected the amount of foulant deposited on the membrane surface along the channel
as shown in this chapter. Based on our experimental data on organic fouling
development along the RO membrane channel, some key factors are proposed to
model organic fouling in a 1-m RO membrane channel. An attempt was made to
quantitatively describe the fouling development in a long RO membrane channel.
7.2.1 Model development
7.2.1.1 Transport across membrane
Water flux through a membrane was a function of the pressure difference. Based on
resistance-in-series, the local permeate flux at location x and time t is given by
v p ( x, t ) 
p( x, t )   ( x, t )
R m ( x, t )
7.1
where vp(x, t), Δp(x, t), Δπ(x, t), Rm(x, t) are the permeate flux, transmembrane
pressure, osmotic pressure, and total membrane resistance at location x and time t. the
transmembrane pressure Δp(x, t) was assumed to be constant along the channel during
the whole fouling processes because the 10 pressure sensors did not detect significant
pressure variation. Hence, the applied pressure Δp was used to replace Δp(x, t).
Osmotic pressure Δπ(x, t) was assumed to be caused mainly by salt and not to be
affected by the organics in this study since its concentration is negligible compared to
the salt concentration in RO process. However, we recognize this may be a
140
Organic fouling development in a long RO membrane channel
simplification since the concentration of macromolecules may have appreciable
osmotic pressure in the same order of magnitude as the applied pressure generally
used in UF (Clifton et al., 1884; Sablani et al., 2001). Additionally, it was assumed
that the osmotic pressure was not affected by the formation of the organic fouling
layer (Lee et al., 2005). Hence, Δπ(x, t) is a function of the channel length Δπ(x). Δπ(x)
could be calculated from concentration polarization modulus based on mass transfer
coefficient estimation and Eq. 6.2 (Mulder, 1996; Murthy and Gupta, 1997), but, in
the light of non-proven application of mass transfer coefficient in long RO membrane
channel, an experimentally measured osmotic pressure in Chapter 6 was substituted to
calculate Δπ(x) in this study. Therefore, the osmotic pressure along the channel was
measured and fitted with the following equation
 ( x)  a x  bx  e
x
cc
7.2
where parameters ax, bx, and cx depend on the salt concentration, operating pressure
and feed flow velocity.
Rm(x, t) is the sum of clean membrane resistance and the fouling layer resistance.
When t is 0, Rm(x, 0) is the clean membrane resistance, Rm0, which is described as a
function of the channel length as discussed in Chapter 6. Hence, the initial membrane
resistance is obtained from a curve of membrane resistance with channel length
similar to Figure 6.1 and expressed in Eq. 7.3.
Rm ( x,0)  1.71  1010  x 3  1.06  1011  x 2  6.35  1010  x  9.36  1010
7.3
141
Organic fouling development in a long RO membrane channel
Therefore, the fouling layer resistance is given by
Rm ( x, t )  Rm ( x, t )  Rm ( x,0) 
p   ( x)
 R m ( x ,0 )
v p ( x, t )
7.4
The calculation of ΔRm(x, t) is delineated in the following section.
7.2.1.2 Calculation of fouling layer resistance
The increment in fouling layer resistance can be linearly related to the amount of
accumulated foulants, i.e.,
Rm  rs M
7.5
where M is the amount of foulants deposited on the membrane and rs is the specific
resistance of the fouling layer. The organic foulants in the bulk solution transferred to
the membrane was dominated by two types of forces (Elzo et al., 1996; Yamamura et
al., 2008). The attachment forces such as electrostatic repulsion, van der Waals
attraction, hydrogen bonding, and particles buoyangcy make the organic foulant
adhere on the membrane surface. The hydrodynamic forces counteract on the
attachment force and sweep the foulant away from the membrane or the fouling layer
surface. As a result, only a fraction of the organic foulant transferred to the membrane
finally deposit on the membrane surface and form the fouling layer, depending on the
chemical and physical conditions (Hong and Elimelech, 1997). Thus, a deposition
ratio, θ, representing the ratio of organic foulant deposited on the membrane to all
142
Organic fouling development in a long RO membrane channel
organic foulant being transferred to the membrane surface, is introduced to calculate
the deposited foulant amount. Hence, the increment in fouling layer resistance is
t
t
0
0
Rm  rsc f 0  v p dt  k m  v p dt
7.6
where rs is the specific resistance of fouling layer and assumed to be constant, cf0 is
the organic concentration in bulk solution, and km is a modified fouling potential in
place of the original fouling potential kf (Eq. 2.10) and experimentally determined
according to Eq. 2.11. Similarly, θ is calculated from organic concentration in the feed
water and the concentrate according to the mass balance using a simple RO filtration
in a small lab-scale crossflow setup where the channel hydrodynamic conditions are
homogeneous.
From Eq. 7.6, km is related to the organic concentration and organic foulant adhering
efficiency and θ depends on the physical and chemical conditions. In a long
membrane channel where the salt concentration, organic concentration and crossflow
velocity vary along the channel, the amount of organic foulant deposited on the
membrane surface is differentfrom the amount calculated with the feed solution.
Therefore, three indexes α, β, and λ are introduced to characterize the variation of the
organic concentration in the bulk solution, salt concentration at the membrane wall
and crossflow velocity along the channel
 ( x, t ) 
c f ( x, t )
c f (0,0)

c f ( x, t )
cf 0
7.7
143
Organic fouling development in a long RO membrane channel
 ( x, t ) 
 ( x, t ) 
c m ( x, t ) c m ( x, t )

c s (0,0)
cs0
7.8
u ( x, t ) u ( x , t )

u (0,0)
u0
7.9
where cs(x,t), cf(x,t), and u(x,t) are salt concentration in the bulk solution, organic
matter concentration in bulk solution, and the crossflow velocity at location x and
time t, respectively. α(x,t), β(x,t), and λ(x,t) are the ratio of organic concentration in
bulk solution, salt concentration at the membrane wall and crossflow velocity at
location x and time t over the feed concentration and feed flow velocity, respectively.
In this study, some assumptions were made: (i) the amount of organic foulant
deposited on the membrane surface is proportional to the organic concentration in the
bulk solution; and (ii) the deposition ratio is proportional to salt concentration and
inversely proportional to the crossflow velocity. Although the assumptions are
oversimplified, they incorporate the effects of attachment forces and hydrodynamic
forces on the organic foulant. Hence, the fouling layer resistance development in a
long membrane channel is expressed as follows
 a ( x, t )  b ( x , t ) t
Rm ( x, t ) 
k m  v p ( x, t )dt
 c ( x, t )
0
k
k
k
7.10
In Eq. 7.10, the value of coefficients ak, bk, and ck would be deduced from a series of
fouling tests to quantitatively represent the effects of organic concentration, salt
concentration and crossflow velocity. The quantitative study of these parameters on
organic fouling was not within the scope of this study and also limited in the
144
Organic fouling development in a long RO membrane channel
published literature by so far. So they were arbitrarily set in this study based on the
aforementioned assumption.
7.2.1.3 Mass balance on the flow and solute
The crossflow velocity and the salt and organic matter concentration in bulk solution
at any location and any time can be calculated by applying mass balance to the flow
and solute
x
u ( x, t ) 
Hu 0   v p ( x, t )dx
0
7.11
H
x
c s ( x, t ) 
Hu 0 c s 0  (1  rtru )  v p ( x, t )c m ( x, t )dx
0
Hu ( x, t )
x
c f ( x, t ) 
Hu 0 c f 0   v p ( x, t )
0
 ( x, t )  ( x, t )
k m dx
 ( x, t )
Hu ( x, t )
7.12
7.13
where H is the channel height and is a constant.
7.2.2 Comparison between experimental work and numerical simulation
A numerical simulation was carried out to investigate the organic fouling
development in the 1-m RO membrane channel. The parameter values used in the
simulations are listed in the Table 7.1. The modified fouling potential, km was
calculated from a simple RO filtration using a small lab-scale RO cell. Alternatively,
km could also be calculated from the flux behavior at the first permeate point in the 1m long RO membrane cell in this study. One example of km profile with time in a
145
Organic fouling development in a long RO membrane channel
fouling test is given in Figure 7.5. Contrary to the original definition in Eq. 2.10 (Tay
and Song, 2005; Tay, 2006), km was not a constant, but declined during the filtration
process like the permeate flux decline profile. This is because the constant fouling
potential kf was originated from the linear permeate flux decline in colloidal fouling
process and did not fit the organic fouling well. This difference implies the necessity
of fouling potential modification and addition of deposition ratio θ. This thesis will
not further characterize the fouling potential of feed solution to RO membrane
processes since this is not the objectives of this study. In this study, a varied km was
used to simulate the organic fouling in the long RO membrane channel. However,
further efforts are required to modify this index or develop a new general one which
can describe all the fouling types.
146
Organic fouling development in a long RO membrane channel
Table 7.1 RO parameter values for simulation.
RO membrane channel length, L
1
m
Channel height, H
0.8
mm
Applied pressure, Δp
300
psi
1.0(Calcium), 7.0(Sodium)
mM
50(alginate)
mg/L
Feed salt concentration, cs0
Feed organic concentration, cf0
0.09
Feed flow velocity, u0
Solute rejection, rtru
m/s
99.5%
Clean membrane resistance, Rm0(x)
Ref. Eq. 7.3
Operation time, T
25
Pa.s/m
h
Feed water fouling potential, km
1×1011
Pa.s/m2
5h
0.9×1011
Pa.s/m2
10h
0.7×1011
Pa.s/m2
15h
0.6×1011
Pa.s/m2
25h
0.5×1011
Pa.s/m2
0, 1, 3 h
Feed organic deposition ratio, θ
0.0526
Osmotic pressure fitting parameters
ax
148937
bx
16351
cx
0.34417
147
Organic fouling development in a long RO membrane channel
11
2.0x10
11
1.8x10
11
1.6x10
2
km (Pa.s/m )
11
1.4x10
11
1.2x10
11
1.0x10
10
8.0x10
10
6.0x10
10
4.0x10
0
5
10
15
20
25
Time (h)
Figure 7.6 An example of the modified fouling potential km evolution during alginate
fouling process.
The fouling experiments were conducted using 1 m-long channel RO membrane cell.
The feed solution consisted of sodium alginate of 50 mg/L with an ionic strength of
10 mM (7.0 mM of NaCl and 1.0 mM of calcium). The local permeate flux at five
permeate collect points along the channel was measured at constant pressure of 300
psi and crossflow velocity of 0.09 m/s. The experimental results were compared with
numerically simulated permeate flux as shown in Figure 7.7. It is found that the
simulated data predicted the experimental results well. This indicates the importance
of the channel effect including the flow and concentration variation along the channel
on the organic fouling development, which implies that the optimization of channel
length is a way to control organic fouling. It was also noted that the model results
were different from the experimental data for some locations of the channel, like the
later part of the channel for 1 hour. This might be caused by the arbitrary setting of
148
Organic fouling development in a long RO membrane channel
those three coefficients for the estimation of the effects of flow and concentration
along the channel. More efforts should be made on the quantitative study of these
coefficients in future work for accurate calculation of resistance of the fouling layer.
-5
2.4x10
0h
-5
2.0x10
-5
Flux (m/s)
1.6x10
-5
1.2x10
-6
8.0x10
-6
4.0x10
0.0
1h
3h
5h
10h
15h
25h
0h
3h
10 h
25 h
0.2
1h
5h
15 h
0.4
0.6
0.8
1.0
Channel length(m)
Figure 7.7 Numerically simulated permeate flux and experiential data of alginate
fouling in a long RO membrane channel.
7.3 Summary
The research work in this study experimentally investigated local permeate flux
decline in the fouling development of different organic matters – sodium alginate and
and BSA along a 1-m RO membrane channel for the first time. Local permeate flux
behavior shows organic fouling development was a heterogeneous process in a long
membrane channel. The fouling of RO membrane by alginate, which has a high
fouling potential, was faster and more severe downstream than upstream of the
149
Organic fouling development in a long RO membrane channel
channel although the downstream initial flux was low. This is attributed to the
concentration polarization formation and other hydrodynamic conditions in a long
membrane channel. Higher downstream salt concentration at the membrane surface
enhanced the interactions and caused a higher fraction of organic matters adhering
onto the membrane although the total organic matter transfer to the membrane surface
by lower permeate flux was low. In addition, high crossflow velocity due to rapid
decline in permeate flux in alginate fouling process swept more organic matter
towards downstream. On the contrary to alginate fouling, fouling of RO membrane by
BSA, which has a low fouling potential, was slightly faster upstream than downstream
because the high permeate flux brought more organic matter towards membrane
surface.
The prediction of membrane fouling in a full-length RO membrane channel is
indispensable for more effective fouling control in membrane design and operation.
Several key factors were proposed for incorporation in the permeate flux predictive
model to delineate the heterogeneous organic fouling development in a long RO
membrane channel. Together with the modified fouling potential of feed solution
obtained from a single RO lab-scale, numerical simulation data shows a good
prediction in organic fouling temporal and spatial development in the long membrane
channel with experimental results, which evidenced the effect of the long channel on
the organic fouling development. However, the fouling potential characterization of
organic solution still remains disputable and many efforts are required to address this
problem.
150
Conclusions and recommendations
Chapter 8
Conclusions and Recommendations
The research work reported in this thesis studied the fouling behavior of model
polysaccharide alginate and model protein BSA. This study also experimentally
investigated the membrane performance and organic fouling characteristics in a long
RO membrane channels. One main contribution of this research work is the
investigation of interactions involved in the organic fouling of RO membrane that will
enrich the knowledge to further characterize the feed fouling potential. The other
main contribution is the design and operation of a 1-m long RO membrane channel
with local permeate flux measurement to study fouling development. This is the first
report to experimentally show the exponential growth of concentration polarization
along the channel, which significantly influenced the membrane performance and
organic fouling development in a long membrane channel. Simulation work
demonstrates the benefit of incorporating concentration polarization and other
variables for accurate prediction of the organic fouling development along the RO
membrane channel. This chapter summarizes the main results and findings of the
research work and recommends possible future work that can extend this research
study.
151
Conclusions and recommendations
8.1 Conclusions
1. The formation of alginate-calcium gel in alginate fouling
Alginate fouling was a result of the formation of high resistance alginate-calcium gel
on the membrane surface where calcium is believed to provide the “crosslink”
between alginate chains. During alginate fouling process, alginate concentration was
the dominant factor in the initial fouling rate while calcium concentration determined
the long term flux. Near complete flux restoration could be easily achieved by
flushing with DI water on membranes fouled with alginate in the absence of calcium.
However, membranes fouled with alginate-calcium gel could only restore the flux
when cleaned with a suitable chemical. EDTA was shown to be a better cleaning
agent for organic fouling in the presence of calcium because of its ability to form
soluble complexes with calcium from the alginate-calcium gel. Therefore, the
interaction between polysaccharide and calcium highlighted the importance of
establishing and reducing the inorganic components, especially calcium ions, in the
feed water. Pre-treatment of feed water should not only target the foulants directly,
but also the ionic components that could interact with the foulants and magnify the
foulants’ fouling potential.
2. A variety of interactions involved in BSA fouling
Due to its “soft” structure and low internal stability, BSA was readily adsorbed onto
the RO membrane surface regardless of the solution pH. Therefore, instead of BSA–
membrane interaction, interactions between BSA molecules and the BSA adsorbed
membrane surface became the dominating mechanism for fouling. The rate of BSA
fouling was affected by various physical and chemical parameters. When the ionic
strength was increased or calcium ion was present in the ionic solution, BSA fouling
152
Conclusions and recommendations
of RO membrane was more severe due to increased charge-shielding effect that led to
a weaker electrostatic repulsion between BSA molecules. In the presence of Mg2+,
fouling of RO membrane by BSA was not as severe as that observed with Ca2+. An
important finding from the study of BSA fouling was that the most severe fouling
occurred at pH near to the IEP of BSA and this behavior occurred for all the ionic
strength and cation species tested. However, BSA fouling did not seem to be affected
by the change in pH in the presence of Mg2+, which requires further investigation to
explain it. Higher temperature was found to increase the rate of BSA fouling only at
pH above the IEP of BSA as a result of the effect of temperature on protein unfolding.
According to the interactions involved in the BSA fouling, SDS, urea, and EDTA can
effectively restore flux with cleaning efficiency of higher than 90%. EDTA had the
highest cleaning efficiency with a low dosage, mild pH, and shortest cleaning time.
3. Exponential growth of concentration polarization in a long RO membrane channel
A long RO membrane channel should be treated as a heterogeneous system because
the key hydrodynamic parameters varied substantially along the membrane channel
due to the concentration polarization exponential growth with the channel length.
Hence, RO membrane separation performance indicators, the permeate flux and
rejection was found to decrease exponentially from inlet to outlet along the channel
because the osmotic pressure increased and net driving force decreased dramatically
downstream. Effects of operating conditions on the membrane separation performance
were complicated by the long channel effect. High applied pressure increased the
global permeate flux, but compared with low pressure, it resulted in faster decrease of
local permeate flux along the channel at high pressure. High feed flow could weaken
concentration polarization which in turn increase the permeate flux. The apparent
153
Conclusions and recommendations
rejection was also affected by the coupling of applied pressure and feed flow in a long
channel filtration system. Higher applied pressure increased the rejection upstream
but inversely proportional downstream while higher feed flow increased the rejection
downstream. Concentration polarization profile was related to the recovery because a
high recovery could be achieved by increasing membrane area (i.e., channel length)
when designing RO process. Applied pressure affected the concentration polarization
formation for low recovery while feed flow played a significant role on concentration
polarization increment for high recovery. Hence, the design and operation of a long
channel RO system should take into account the channel length to choose favorable
range of pressure and feed flow to achieve both high production and good separation.
The introduction of a spacer into the membrane channel mitigated concentration
polarization formation whereby the increase of concentration polarization was
significantly reduced along the channel. However, the spacer also added nonnegligible vertical resistance which may counteract the effect of concentration
polarization mitigation by spacers, especially for low concentration salt solution at the
upstream end of the flow channel. This suggests that it is necessary to consider the
integral effect of the spacer for designing of an RO membrane module and overall
system.
4. Heterogeneous development of organic fouling in a long RO membrane channel
Two types of organic matters with different fouling potential, alginate and BSA,
displayed heterogeneous organic fouling development in the long channel membrane
cell. Globally, alginate fouling developed faster than BSA for the same operating
condition. Locally, the alginate fouling developed faster downstream than upstream as
154
Conclusions and recommendations
the local permeate flux declined faster downstream. This is primarily the result of the
exponential growth of concentration polarization, which significantly enhanced the
interactions between alginate and salt at the membrane wall. Also, high crossflow, as
a balance of fast decline of permeate flux, flushed more organic matters to
downstream and contributed to the faster alginate fouling development downstream.
Insertion of spacer in the long channel could mitigate algiante fouling development
due to concentration polarization reduction. On the contrary, BSA fouling developed
slightly faster upstream than downstream. It seemed that the high salt concentration
downstream did not improve BSA fouling due to the weak interaction between salt
and BSA. The main factor for BSA fouling development could be the higher initial
flux upstream which brought more BSA to the membrane surface. Therefore, the
prediction of organic fouling in a long RO membrane channel requires two group of
knowledge. One is the characterization of the feed solution fouling potential and the
other is the information of process parameters in the long membrane channel. Some
key fouling factors, modified fouling potential km and deposition ratio θ, were
proposed in the flux predictive model to address the first group of information. In
addition, three more parameters α, β, γ, were introduced to reflect the variation of
hydrodynamic condition. The simulation of alginate fouling development in the 1-m
long RO membrane channel predicted well the experimental results.
8.2 Recommendations
There are some interesting directions that can be followed from the research study
presented in this thesis. The following are recommendations for future work:
155
Conclusions and recommendations
1. Implementation of fouling studies using the secondary effluent
In this study, two model macromolecules polysaccharide (alginate) and protein (BSA)
were used to study the organic fouling of RO membrane which involved a variety of
interactions of organic fouling. However, in an actual RO process for water
reclamation, those interactions should behave differently because they are susceptible
to a more complex solution environment, a mixture of salts, colloids, bacteria, and
organic matters than a single organic matter solution. Organic fouling could be either
abated or accelerated depending on the binding effect of other components. Although
some studies have been conducted using a synthetic solution consisting more than one
model foulant, there are still a big gap from the actual fouling behavior. The first step
of the fouling study using the actual secondary effluents is to characterize the fouled
membrane and determine the foulants in the RO feed. The second step is to determine
each foulant’s individual contribution to the fouling potential of the feed and their
combined effects. The third step is to quantitatively link the foulant in the feed to the
fouling rate or extent in the RO process.
2. Improvement of fouling potential kf
Estimation of fouling strength of the RO feed is critical to predict the fouling in RO
process operation and design. In this study a fouling potential index kf was modified
and applied to estimate the organic solutions’ fouling strength which represented
collective effect of feed solution on fouling. However, Tay and Song (2005)
developed this index based on linear decline of flux for colloidal suspension filtration,
which might not be suitable for organic fouling. The flux decline in organic fouling
showed a transit from an initial fast decline to a pseudo-stable flux, which resulted in
the change of fouling potential kf with time. Therefore, the fouling potential kf has to
156
Conclusions and recommendations
be improved to characterize organic fouling apart from colloid fouling. If the fouling
potential characterization for all the fouling types can be expressed in a value or a
simple mathematical equation, the fouling in RO process can be estimated more
accurately. The costs and efforts invested to distinguish the different foulants in the
feed can be greatly saved.
3. Implementation of pilot-scale study
Much effort in this study is based on experimental test using lab-scale RO membrane
setup. Even the long channel membrane cell is only one meter in length and is much
shorter than the actual RO process, which typically consists of 6 modules in one
pressure vessel. Pilot scale studies have to be implemented to observe concentration
polarization growth and organic or other fouling development along the actual RO
membrane channel. Pilot scale studies can verify the fouling characteristics obtained
from this lab-scale study and the effectiveness of the organic fouling factors proposed
in the predictive model in this study. At present, the pilot study used a global average
flux to characterize the fouling and no additional samples except the final permeate
can be taken and measured. Therefore, a modified pilot setup is needed to measure
the permeate flux and concentration at the end of each RO membrane module.
157
References
References
Abu Tarboush, B.J., D. Rana, T. Matsuura, H.A. Arafat, and R.M. Narbaitz.
Preparation of thin-film-composite polyamide membranes for desalination using
novel hydrophilic surface modifying macromolecules. J. Membr. Sci., 325, 166-175,
2008.
Ahmad, A.L., K.K. Lau, and M.Z. Abu Bakar. Impact of different spacer filament
geometries on concentration polarization control in narrow membrane channel. J.
Membr. Sci., 262, 138-152, 2005.
Al-Bastaki, N.M., and Abbas, A. Predicting the performance of RO membranes.
Desalination, 132, 181-187, 2000.
Ang, W.S., S. Lee, and M. Elimelech. Chemical and physical aspects of cleaning of
organic-fouled reverse osmosis membranes. J. Membr. Sci., 272, 198-210, 2006.
Ang, W.S., and Elimelech, M. Protein (BSA) fouling of reverse osmosis membranes:
Implications for wastewater reclamation. J. Membr. Sci., 296, 83-92, 2007.
Aptel, P., and Buckley, C.A. Categories of membrane operations in Water Treatment
Membrane Process ed. By Mallevialle et al. New York: McGraw Hill. 1996.
Avlonitis, S., W.T. Hanbury, and M.B. Boudinar. Spiral wound modules performance
an analytical solution: part II. Desalination, 89, 227-246, 1993.
Barker, D.J., and Stuckey, D.C. A review of soluble microbial products (SMP) in
wastewater treatment systems. Wat. Res., 33, 3063-3082, 1999.
Bellona, C., J.E. Drewes, P. Xu, and G. Amy. Factors affecting the rejection of
organic solutes during NF/RO treatment — a literature review. Wat. Res., 38, 27952809, 2004.
Bessiere, Y., D.F. Fletcher, and P. Bacchin. Numerical simulation of colloid dead-end
filtration: Effect of membrane characteristics and operating conditions on matter
accumulation. J. Membr. Sci., 313, 52-59, 2008.
Bhattacharyya, D., S.L. Back, R.I. Kermode, and M.C. Roco. Prediction of
concentration polarization and flux behavior in reverse osmosis by numerical analysis.
J. Membr. Sci., 48, 231-262, 1990.
Bloomfield, V. The structure of bovine serum albumin at low pH. Biochemistry, 5,
684-689, 1966.
158
References
Bos, M.A., Z. Shervani, A.C.I. Anusiem, M. Giesbers, W. Norde, and J.M. Kleijn,
Influence of the electric potential of the interface on the adsorption of proteins.
Colloids Surf. B: Biointerf., 3, 91-100, 1994.
Bouchard, C.R., P.J. Carreau, T. Matsuura, and S. Sourirajan. Modeling of
ultrafiltration: predictions of concentration polarization effects. J. Membr. Sci., 97,
215-229, 1994.
Brauns, E., E.V. Hoof, B. Molenberghs, C. Dotremont, W. Doyen, and R. Leysen. A
new method of measuring and presenting the membrane fouling potential.
Desalination, 150, 31-43, 2002.
Brandt, D.C., G.F. Leitner, and W.E. Leitner. Reverse osmosis membranes state of the
art in Reverse osmosis membrane technology, water chemistry, and industrial
applications ed. by Z. Amjad. New York: Van Nostrand Reinhold. 1993.
Bruus, J.H., P.H. Nielsen, and K. Keiding. On the stability of activated sludge flocs
with implications to dewatering. Wat. Res., 26, 1597-1604, 1992.
Bu-Ali, Q., M. Al-Aseeri, and N. Al-Bastaki. An experimental study of performance
parameters and ion concentration along a reverse osmosis membrane. Chem. Eng.
Process., 46, 323-328, 2007.
Bu, H., A.-L. Kjøniksen, A. Elgsaeter, and B. Nyström. Interaction of unmodified and
hydrophobically modified alginate with sodium dodecyl sulfate in dilute aqueous
solution: Calorimetric, rheological, and turbidity studies. Colloid Surf. A:
Physicochem. Eng. Asp., 278, 166-174, 2006.
Bu, H., A.-L. Kjøniksen, K.D. Knudsen, and B. Nyström. Effects of surfactant and
temperature on rheological and structural properties of semidilute aqueous solutions
of unmodified and hydrophobically modified alginate. Langmuir, 21, 10923-10930,
2005.
Byrne, W. Reverse Osmosis- A practical guide for industrial users. Littleton: Tall
Oaks Publishing, Inc. 1995.
Campbell, M.J., R.P. Walter, R. McLoughlin, and C.J. Knowles. Effect of
temperature on protein conformation and activity during ultratfiltration. J. Membr.
Sci., 78, 3543, 1993.
Chen, K.L., L. Song, S.L. Ong, and W.J. Ng. The development of membrane fouling
in full-scale RO processes. J. Membr. Sci., 232, 63-74, 2004.
Chen, V., R. Chan, H. Li, and M.P. Buchnall. Spatial distribution of foulants on
membranes during ultrafiltration of protein mixtures and the influence of spacers in
the channel. J. Membr. Sci., 287, 79-87, 2007.
Chilukuri, V.V.S., A.D. Marshall, P.A. Munro, and H. Singh. Effect of sodium
dodecyl sulphate and cross-flow velocity on membrane fouling during cross-flow
microfiltration of lactoferrin solutions. Chem. Eng. Process., 40, 321-328, 2001.
159
References
Clifton, M.J., N. Abidine, P. Aptel, and V. Sanchez. Growth of the polarization layer
in ultrafiltration with hollow-fiber membranes. J. Membr. Sci., 21, 233-246, 1984.
Cohen, R.D., and Probstein, R.F. Colloidal fouling of reverse osmosis membranes. J.
Colloid Interface Sci., 114, 194-207, 1986.
Combe, C., E. Molis, P. Lucas, R. Riley, and M.M. Clark. The effect of CA
membrane properties on adsorptive fouling by humic acid. J. Membr. Sci., 154, 73-87,
1999.
Compbell, M.K., and Farrell, S.O. Biochemistry. Brooks/Cole, California, 2006.
Dacosta, A.R. Fluid flow and mass transfer in spacer-filled channels for
ultrafiltration. Ph.D. Thesis, School of Chemical Engineering and Industrial
Chemistry, University of New South Wales, 1993.
Davis, R.H. Modeling of fouling of crossflow microfiltration membranes. Sep. Purif.
Meth., 21, 75-126, 1992.
Davis, T.A., F. Llanes, B. Volesky, and A. Mucci. Metal selectivity of Sargassum spp.
and their alginates in relation to their α-L–guluronic acid content and conformation.
Environ. Sci. Technol., 37, 261-267, 2003.
de Pinho, M.N., V. Semião, and V. Geraldes. Integrated modeling of transport
processes in fluid/nanofiltration membrane systems. J. Membr. Sci., 206, 189-200,
2002.
Dignac, M.-F., V. Urbain, D. Rybacki, A. Bruchet, D. Snidaro, and P. Scribe.
Chemical description of extracellular polymers: implication on activated sludge floc
structure. Wat. Sci. Tech., 38, 45-53, 1998.
DuBois, M., K.A. Gilles, J.K. Hamilton, P.A. Rebers, and F. Smith. Colorimetric
method for determination of sugars and related substances. Anal. Chem., 28, 350-356,
1956.
El Kadi, N., N. Taulier, J.Y. Le Huérou, M. Gindre, W. Urbach, I. Nwigwe, P.C.
Kahn, and M. Waks. Unfolding and refolding of bovine serum albumin at acid pH:
ultrasound and structural studies. Biophys. J., 91, 3397-3404, 2006.
Elimelech, M., X. Zhu, A.E. Childress, and S. Hong. Role of membrane surface
morphology in colloidal fouling of cellulose acetate and composite aromatic
polyamide reverse osmosis membranes. J. Membr. Sci., 127, 101-109, 1997.
Elimelech, M., W.H. Chen, and J.J. Waypa. Measuring the zeta (electrokinetic)
potential of reverse osmosis membrane by a streaming potential analyzer.
Desalination, 95, 269286, 1994.
Elzo, D., P. Schmita, D. Houi, and S. Joscelyne. Measurement of particle/membrane
interactions by a hydrodynamics method. J. Membr. Sci., 109, 43-53, 1996.
160
References
Fairbrother, F., and Mastin, H. Studies in electro-endosmosis. J. Chem. Soc., 125,
2319-2330, 1924.
Fariñas, M., J.M. Granda, L. Gurtubai, and M.J. Villagra. Pilot experiences on the
recovery of polluted reverse osmosis membranes. Desalination, 66, 385-402, 1987.
Frank, B.P., and Belfort, G. Polysaccharides and sticky membrane surfaces: critical
ionic effects. J. Membr. Sci., 212, 205-212, 2003.
Fu, X., T. Maruyama, T. Sotani, and H. Matsuyama. Effect of surface morphology on
membrane fouling by humic acid with the use of cellulose acetate butyrate hollow
fiber membranes. J. Membr. Sci., 320, 483-491, 2008.
Geraldes, V., V. Semião, and M.N. de Pinho. Flow and mass transfer modeling of
nanofiltration. J. Membr. Sci., 191, 109-128, 2001.
Gerard, R., H. Hachisuka, and M. Hirose. New membrane developments expanding
the horizon for the application of reverse osmosis technology. Desalination, 119, 4755, 1998.
Goosen, M.F.A., S.S. Sablani, H. Ai-Hinai, S. Ai-Obeidani, R. Al-Belushi, and D.
Jackson. Fouling of reverse osmosis and ultrafiltration membranes: a critical review.
Sep. Sci. Technol., 39, 2261- 2297, 2004.
Grant, G.T., E.R. Morris, R.A. Rees, P.J.C. Smith, and D. Thom. Biological
interactions between polysaccharides and divalent cations: the egg-box model. FEBS
LETTERS, 32, 195-198, 1973.
Grover, P.K., and Ryall, R.L. Critical appraisal of salting-out and its implications for
chemical and biological sciences. Chem. Rev., 105, 1-10, 2005.
Gupta, S.K. Design and analysis of reverse osmosis system using three parameter
models for transport across the membrane. Desalination, 85, 283- 296, 1992.
Güell, C., and Davis, R.H. Membrane fouling during microfiltration of protein
mixtures. J. Membr. Sci., 119, 269284, 1996.
Haarhoff, J., and van der Merwe, B. Twenty-five years of wastewater reclamation in
Windhoek, Namibia. Wat. Sci. Tech., 33, 25-35, 1996.
Herzberg, M., and Elimelech, M. Biofouling of reverse osmosis membranes: role of
biofilm-enhanced osmotic pressure. J. Membr. Sci., 295, 11-20, 2007.
Herzberg, M., S. Kang, and M. Elimelech. Role of extracellular polymeric substances
(EPS) in biofouling of reverse osmosis membranes. Environ. Sci. Technol., 43, 43934398, 2009.
Hoek, E.M.V., and Elimelech, M. Cake-enhanced concentration polarization: a new
fouling mechanism for salt-rejecting membranes. Environ. Sci. Technol., 37, 55815588, 2003.
161
References
Hoek, E.M.V., J. Allred, T. Knoell, and B.-H. Jeong. Modeling the effects of fouling
on full-scale reverse osmosis processes. J. Membr. Sci., 314, 33-49, 2008.
Hong, S., and Elimelech, M. Chemical and physical aspects of natural organic matter
(NOM) fouling of nanofiltration membranes. J. Membr. Sci., 132, 159-181, 1997.
Hong, S., R.S. Faibish, and M. Elimelech. Kinetics of permeate flux decline in
crossflow membrane filtration of colloidal suspensions. J. Colloid Interface Sci., 196,
267-277, 1997.
Hu, J.Y., S.L. Ong, J.H. Shan, J.S. Kang, and W.J., Ng. Treatability of organic
fractions derived from secondary effluent by reverse osmosis membrane. Wat. Res.,
37, 4801-4809, 2003.
Imai, A., T. Fukushima, K. Matsushige, Y.-H. Kim, and K. Choi. Characterization of
dissolved organic matter in effluents from wastewater treatment plants. Wat. Res., 36,
859-870, 2002.
Jacangelo, J.G, J.-M. Laine, E.W. Cummins, and S.S. Adham. UF with pretreatment
for removing DSP precursors. J. AWWA, 3, 100-112, 1987.
Jaffrin, M.Y. Dynamic shear-enhanced membrane filtration: a review of rotating disks,
rotating membranes and vibrating systems. J. Membr. Sci., 324, 7-25, 2008.
Jarusutthirak, C., G. Amy, and J.-P. Croué. Fouling characteristics of wastewater
effluent organic matter (EfOM) isolates on NF and UF membranes. Desalination, 145,
247-255, 2002.
Jones, K.L., and O’Melia, C.R. Protein and humic acid adsorption onto hydrophilic
membrane surfaces: effects of pH and ionic strength. J. Membr. Sci., 165, 31-46, 2000.
Kelly, S.T., and Zydney, A.L. Effects of intermolecular thiol-disulfide interchange
reactions on BSA fouling during microfiltration. Biotechnol. Bioeng., 44, 972-982,
1994.
Kimura, K., Y. Hane, Y. Watanabe, G. Amy, and N. Ohkuma. Irreversible membrane
fouling during ultrsfiltrstion of surface water. Wat. Res., 38, 3431-3441, 2004.
Koutsou, C.P., S.G. Yiantsios, and A.J. Karabelas. A numerical and experimental
study of mass transfer in spacer-filled channels: Effects of spacer geometrical
characteristics and Schmidt number. J. Membr. Sci., 326, 234-251, 2009.
Kuzmenko, D., E. Arkhangelsky, S. Belfer, V. Freger, and V. Gitis. Chemical
cleaning of UF membranes fouled by BSA. Desalination, 179, 323-333, 2005.
Lapointe, J.-F., S.F. Gauthier, Y. Pouliot, and C. Bouchard. Fouling of a
nanofiltration membrane by β-lactoglobulin tryptic hydrolysate: impact on the
membrane sieving and electrostatic properties. J. Membr. Sci., 253, 89-102, 2005.
162
References
Lattner, D., H.-C. Flemming, and C. Mayer. 13C-NMR study of the interaction of
bacterial alginate with bivalent cations. Int. J. Biol. Macromol., 33, 81-88, 2003.
Lawrence, N.D., J.M. Perera, M. Iyer, M.W. Hickey, and G.W. Stevens. The use of
streaming potential measurements to study the fouling and cleaning of ultrafiltration
membranes. Sep. Purif. Technol., 48, 106-112, 2006.
Le Gouellec, Y.A., and Elimelech, M. Calcium sulfate (gypsum) scaling in
nanofiltration of agriculture drainage water. J. Membr. Sci., 205, 279-291, 2002.
Lee, H., G. Amy, J. Cho, Y. Yoon, S.-H. Moon, and I.S. Kim. Cleaning strategies for
flux recovery of an ultrafiltration membrane fouled by natural organic matter. Wat.
Res., 35, 3301-3308, 2001.
Lee, S., J. Cho, and M. Elimelech. Combined influence of natural organic matter
(NOM) and colloidal particles on nanofiltration membrane fouling. J. Membr. Sci.,
262, 27-41, 2005.
Lee, S., W.S. Ang, and M. Elimelech. Fouling of reverse osmosis membranes by
hydrophilic organic matter: implications for water reuse. Desalination, 187, 313-321,
2006.
Li, Q., and Elimelech, M. Organic fouling and chemical cleaning of nanofiltration
membrane: measurements and mechanisms. Environ. Sci. Technol., 38, 4683-4693,
2004.
Li, Q., Z. Xu, and I. Pinnau. Fouling of reverse osmosis membranes by biopolymers
in wastewater secondary effluent: Role of membrane surface properties and initial
permeate flux. J. Membr. Sci., 290, 173-181, 2007.
Li, X., J. Li, X. Fu, R. Wickramasinghe, and J. Chen. Chemical cleaning of PS
ultrafilters fouled by the fermentation broth of glutamic acid. Sep. Purif. Technol., 42,
181-187, 2005.
Lin, C.-F., S.-H. Liu, and O.J. Hao. Effect of functional groups of humic substances
on UF performance. Wat. Res., 35, 2395-2402, 2001.
Lin, S.-H., C.-L. Hung, and R.-S. Juang. Effect of operating parameters on the
separation of protein in aqueous solutions by dead-end ultrafiltration. Desalination,
234, 116-125, 2008.
Lowry, O.H., N.J. Rosebrough, A.L. Farr, and R.J. Randall. Protein measurement
with the folin phenol reagent. J. Biol. Chem., 193, 265-275, 1951.
Lyster, E., and Cohen, Y. Numerical study of concentration polarization in a
rectangular reverse osmosis membrane channel: Permeate flux variation and
hydrodynamics end effects. J. Membr. Sci., 303, 140-153, 2007.
163
References
Ma, H., H.E. Allen, and Y. Yin. Characterization of isolated fractions of dissolved
organic matter from natural waters and a wastewater effluent. Wat. Res., 35, 985-996,
2001.
Madaeni, S.S., and Mansourpanah, Y. Chemical cleaning of reverse osmosis
membranes fouled by whey. Desalination. 161, 13-24, 2004.
Madaeni, S.S., T. Mohamamdi, and M.K. Moghadam. Chemical cleaning of reverse
osmosis membrane. Desalination. 134, 77-82, 2001.
Maruyama, T., S. Katoh, M. Nakajima, H. Nabetani, T.P. Abbott, A. Shono and K.
Satoh. FT-IR analysis of BSA fouled on ultrafiltration and microfiltration membranes.
J. Membr. Sci., 192, 201-207, 2001.
Matthiasson, E. The role of macromolecular adsorption in fouling of ultrafiltration
membranes. J. Membr. Sci., 16, 23-36, 1983.
Meireles, M., P. Aimar, and V. Sanchez. Albumin denaturation during ultrafiltration:
Effects of operating conditions and consequences on membrane fouling. Biotechnol.
Bioeng., 38, 528534, 1991.
Metcalf & Eddy, Inc. Wastewater engineering treatment and reuse. 4th ed. McGrawHill, New York, 2004.
Michaels, A.S. New separation technique for the CPI. Chem. Eng. Prog., 64, 3143,1968.
Michnik, A., K. Michalik, and Z. Drzazga. Stability of bovine serum albumin at
different pH. J. Therm. Anal. Cal., 80, 399-406, 2005.
Mueller, J., and Davis, R.H. Protein fouling of surface-modified polymeric
microfiltration membranes. J. Membr. Sci., 116, 4760, 1996.
Mujeirigo, R. Achievements and challenges in the reuse of reclaimed water.
Presentations and articles Euro-CAE workshop “wastewater as resource”. Paris,
Institute de France, 7 July 2000.
Mulder, M. Basic Principles of Membrane technology. 2rd ed. London: Kluwer
Academic Publishers. 1996.
Muñoz-Aguado, M.J., D.E. Wiley, and A.G. Fane. Enzymatic and detergent cleaning
of a polysulfone ultrafiltration membrane fouled with BSA and whey. J. Membr. Sci.,
117, 175-187, 1996.
Murrer, J., and Latter, S. Reducing the costs of ultrapure water production. Alpheus
International Ltd. Available via www.Alpheus.co.uk, І-3. 2003.
Murthy, Z.V.P., and Gupta, S.K. Estimation of mass transfer coefficient using a
combined nonlinear membrane transport and film theory model. Desalination, 109,
39-49, 1997.
164
References
Nakamura, K., and Matsumoto, K. Properties of protein adsorption onto pore surface
during microfiltration: Effects of solution environment and membrane hydrophobicity.
J. Membr. Sci., 280, 363374, 2006.
Neal, P.R., H. Li, A.G. Fane, and D.E. Wiley. The effect of filament orientation on
critical flux and particle deposition in spacer-filled channels. J. Membr. Sci., 214,
165-178, 2003.
Nemeth, J.E. Innovative system designs to optimize performance of ultra-low
pressure reverse osmosis membranes. Desalination, 118, 63-71, 1998.
Ng, H.Y., T.W. Tan, and S.L. Ong. Membrane fouling of submerged membrane
bioreactors: impact of mean cell residence time and the contributing factors. Environ.
Sci. Technol., 40, 2706-2713, 2006.
Noguera, D.R., N. Araki, and B.E. Rittmann. Soluble microbial products (SMP) in
anaerobic chemostats. Biotechnol. Bioeng., 44, 1040-1047, 1994
Norberg, D., S. Hong, J. Taylor, and Y. Zhao. Surface characterization and
performation evaluation of commercial fouling resistance low-pressure RO
membranes. Desalination, 202, 45-52, 2007.
Núñez, C., R. León, J. Guzmán, G. Espín, and G. Soberón-Chávez. Role of
Azobobacter vinelandii mucA and mucC gene products in alginate production. J.
Bacteriol., 182, 6550-6556, 2000.
Nyström, M., K. Ruhomäki, and L. Kaipia. Humic acid as a fouling agent in filtration.
Desalination, 106, 79-87, 1996.
Ognier, S., C. Wisniewski, and A. Grasmick. Influence of macromolecule adsorption
during filtration of a membrane bioreactor mixed liquor suspension. J. Membr. Sci.,
209, 27-37, 2002.
Park, N., B. Kwon, I.S. Kim, and J. Cho. Biofouling potential of various NF
membranes with respect to bacteris and their soluble microbial products (SMP):
Characterizations, flux decline, and transport parameters. J. Membr. Sci., 258, 43-54,
2005.
Painter, H.A. Organic compound in solution in sewage effluents. Chem. & Ind.,
September, 818-822, 1973.
Public Utility Board. Annual Report, 2007/2008.
Rautenbach, R., T. Linn, and L. Eilers. Treatment of severely contaminated waste
water by a combination of RO, high–pressure RO and NF—potential and limits of the
process. J. Membr. Sci., 174, 231-241, 2000.
Rebhum, M., and Manka, J. Classification of organics in secondary effluents. Environ.
Sci. Techol., 5, 606-609, 1971.
165
References
Sablani, S.S., M.F.A. Goosen, R. Al-Relushi, and M. Wilf. Concentration polarization
in ultratfiltration and reverse osmosis: a critical review. Desalination, 141, 269-289,
2001.
Salgın, S., S. Takaç, and T.H. Özdamar. Adsorption of bovine serum albumin on
polyether sulfone ultrafiltration membranes: Determination of interfacial interaction
energy and effective diffusion coefficient. J. Membr. Sci., 278, 251-260, 2006.
Schiener, P., S. Nachaiyasit, and D.C. Stuckey. Production of soluble microbial
products (SMP) in an anaerobic baffled reactor: composition, biodegradability and the
effect of process parameters. Environ. Tech., 19, 391-400, 1998.
Schneider, R.P., L.M. Ferreira, P. Binder, and J.R. Ramos. Analysis of foulant layer
in all elements of an RO train. J. Membr. Sci., 261, 152-162, 2005.
Schneider, R.P., L.M. Ferreira, P. Binder, E.M. Bejarano, K.P. Góes, E. Slongo, C.R.
Machado, G.M.Z. Rosa. Dynamics of organic carbon and of bacterial populations in a
conventional pretreatment train of a reverse osmosis unit experiencing server
biofouling. J. Membr. Sci., 266, 18-29, 2005.
Schwinge, J., P.R. Neal, D.E. Wiley, D.F. Fletcher, and A.G. Fane. Spiral wound
modules and spacers review and analysis. J. Membr. Sci., 242, 129-153, 2004.
Seidel, A., and Elimelech, M. Coupling between chemical and physical interactions in
natural organic matter (NOM) fouling of nanofiltration membranes: implications for
fouling control. J. Membr. Sci., 203, 245-255, 2002.
Shahalam, A.M., A. Al-Harthy, and A. Al-Zawhry. Feed water pretreatment in RO
systems: unit processes in the Middle East. Desalination. 150, 235-245, 2002.
Simmons, M.J.H., P. Jayaraman, and P.J. Fryer. The effect of temperature and shear
rate upon the aggregation of whey protein and its implications for milk fouling. J.
Food Eng., 79, 517528, 2007.
Soltanieh, M., and Gill, W.N. Review of reverse osmosisi membranes and transport
models. Chem. Eng. Commun., 12, 279-363, 1981.
Song, L., J.Y. Hu, S.L. Ong, W.J. Ng, M. Elimelech, and M. Wilf. Performance
limitation of the full-scale reverse osmosis process. J. Membr. Sci., 214, 239-244,
2003.
Song, L., and Tay, K.G. Performance prediction of a long crossflow reverse osmosis
membrane channel. J. Membr. Sci., 281, 163-169, 2006.
Tang, C.Y., and Leckie, J.O. Membrane independent limiting flux for RO and NF
membranes fouled by humic acid. Environ. Sci. Technol., 41, 4767–4773, 2007.
Tang, C.Y., Y.-N. Kwon, and J.O. Leckie. The role of foulant-foulant electrostatic
interaction on limiting flux for RO and NF membranes during humic acid fouling166
References
Theoretical basis, experimental evidence, and AFM interaction force measurement. J.
Membr. Sci., 326, 526-532, 2009.
Tang, C.Y., Y.-N. Kwon, and J.O. Leckie. Fouling of reverse osmosis and
nanofiltrtion by humic acid — effects of solution composition and hydrodynamic
conditions. J. Membr. Sci., 290, 86-94, 2007.
Tasaka, K., T. katsura, H. Iwahori, and Y. Kamiyama. Analysis of RO elements
operated at more than 80 plants in Japan. Desalination. 96, 259-272, 1994.
Tay, K.G. Dynamics and characterization of membrane fouling in a long reverse
osmosis membrane channel. Ph.D. Thesis, Department of civil engineering, National
University of Singapore, 2006.
Tay, K.G., and Song, L. A more effective method for fouling characterization in a
full-scale reverse osmosis process. Desalination, 177, 95-107, 2005.
Thongngam, M., and McClements, D.J. Influence of pH, ionic strength, and
Temperature on self-association and Interactions of sodium dodecyl sulfate in the
absence and presence of chitosan. Langmuir, 21, 79-86, 2005.
Tran-Ha, M.H., V. Santos, and D.E. Wiley. The effect of multivalent cations on
membrane-protein interactions during cleaning with CTAB. J. Membr. Sci., 251, 179188, 2005.
Trägårdh, G. Membrane cleaning. Desalination, 71, 325-335, 1989.
Turan, M. Influence of filtration conditions on the performance of nanofiltration and
reverse osmosis membranes in dairy wastewater treatment. Desalination, 170, 83-90,
2004.
Vedavyasan, C.V. Combating water shortages with innovative uses of membranes.
Desalination, 132, 345347, 2000.
Veerman, C., L.M.C. Sagis, J. Heck, and E. van der Linden. Mesostructure of fibrillar
bovine serum albumin gels. Int. J. Biol. Macromol., 31, 139-146, 2003.
Velasco, C., M. Ouammou, J.I. Calvo, and A. Hernández. Protein fouling in
microfiltration: deposition mechanism as a function of pressure for different pH. J.
Colloid Interface Sci., 266, 148-152, 2003.
Vrijenhoek, E.M., S. Hong, and M. Elimelech. Influence of membrane surface
properties on initial rate of colloidal fouling of reverse osmosis and nanofiltration
membranes. J. Membr. Sci., 188, 115-128, 2001.
Vrouwenvelder, J.S., J.A.M. van Paassen, J.M.C. van Agtmaal, M.C.M. van
Loosdrecht, J.C. Kruithof. A critical flux to void biofouling of spiral wound
nanofiltration and reverse osmosis membranes: Fact or fiction? J. Membr. Sci., 326,
36-44, 2009.
167
References
Wang, B.-J., T.-C. Wei, and Z.-R. Yu. Effect of operating temperature on component
distribution of West Indian cherry juice in a microfiltration system. LWT-Food Sci.
Technol., 38, 683689, 2005.
Wehner, M. A research and demonstration project permit under the new California
groundwater recharge regulations. Desalination, 87, 37-52, 1992.
Wilbert, M.C., J. Pellegrino, and A. Zydney. Bench-scale testing of surfactantmodified reverse osmosis/nanofiltration membranes. Desalination, 115, 15-32, 1998.
Wilf, M. Design consequences of recent improvements in membrane performance.
Desalination, 113, 157-163, 1997.
Wilf, M., and Alt, S. Application of low fouling RO membrane elements for
reclamation of municipal wastewater. Desalination, 132, 1119, 2000.
Wilf, M., and Klinko, K. Effective new pretreatment for seawater reserve osmosis
systems. Desalination, 117, 323-331, 1998.
Xiong, Y.L. Influence of pH and ionic environment on thermal aggregation of whey
proteins. J. Agric. Food Chem., 40, 380-384, 1992.
Xue, J., X. Huang, and E.M.V. Hoek. Roles of specific ion interactions in seawater
RO membrane fouling by alginate acid. Environ. Sci. Technol., 43, 3580-3587, 2009.
Yamamura, H., K. Kimura, T. Okajima, H. Tokumoto, and Y. Watanabe. Affinity of
functional groups for membrane surfaces: Implications for physically irreversible
fouling. Environ. Sci. Technol., 42, 5310-5315, 2008.
Ye, Y., P.L. Clech, V. Chen, A.G. Fane, and B. Jefferson. Fouling mechanisms of
alginate solutions as model extracellular polymeric substances. Desalination, 175, 720, 2005.
Ye, Y., V. Chen, and A.G. Fane. Modeling long-term subcritical filtration of model
EPS solutions. Desalination, 191, 318-327, 2006.
Yiantsios, S.G., D. Sioutopoulos, and A.J. Karabelas. Colloidal fouling of RO
membranes: an overview of key issues and effects to develop improved prediction
techniques. Desalination, 183, 257-272, 2005.
Yoon, S-H., C-H. Lee, K-J. Kim, and A.G. Fane. Effect of calcium ion on the fouling
of nanofilter by humic acid in drinking water production. Wat. Res., 32, 2180-2186,
1998.
Zhao, Y., L. Song, and S.L. Ong. Fouling of RO membranes by effluent organic
matter (EfOM): Relating major components of EfOM to their characteristic fouling
behaviors. J. Membr. Sci., 349,75-82, 2010.
168
References
Zhou, W., L. Song, and K.G. Tay. A numerical study on concentration polarization
and system performance of spiral wound RO membrane modules. J. Membr. Sci., 271,
38-46, 2006.
Zhu, X., and Elimelech, M. Fouling of reverse osmosis membranes by aluminum
oxide colloids. J. Environ. Eng., 121, 884-892, 1995.
Zydney, A.L. Stagnant film model for concentration polarization in membrane system.
J. Membr. Sci., 130, 275-281, 1997.
169
Appendix
Appendix
A.
Relationship
between
NaCl
concentration
and
conductivity
Low range
25
Conductivity (s/cm)
20
15
10
5
0
0
2
4
6
8
10
NaCl conc. (mg/L)
170
Appendix
250
Middle range
Conductivity (s/cm)
200
150
100
50
0
0
20
40
60
80
100
800
1000
NaCl conc. (mg/L)
2000
High range
Conductivity (s/cm)
1600
1200
800
400
0
0
200
400
600
NaCl conc. (mg/L)
171
Appendix
10000
Superhigh range
Conductivity (s/cm)
8000
6000
4000
2000
1000
2000
3000
4000
5000
NaCl conc. (mg/L)
Range
R2
Conversion equation
COND(μs/cm)
Conc.(mg/L)
Conc. =0.4426×COND-0.4556
0.9999
<23.7
1~10
Conc.=0.4582×COND-0.8779
1
23.7~220
10~100
Conc.=0.5124×COND-20.218
0.9986
220~2010
100~1000
Conc.=0.5587×COND-146.53
0.9985
>2010
1000~5000
172
Download