Water Research 38 (2004) 41644172

Fouling mechanisms in the integrated system with softening

and ultraltration
Ji Hyang Kweon
a
, Desmond F. Lawler
b,
a
Department of Environmental Engineering, College of Engineering, Konkuk University, 1 Hwayang-dong, Gwangjin-gu,
Seoul 143-701, Korea
b
Department of Civil Engineering, Environmental and Water Resources Engineering, The University of Texas at Austin, ECJ 8.6, Austin,
Texas 78712,USA
Received 24 April 2003; received in revised form 23 June 2004; accepted 30 June 2004
Abstract
Softening is designed to remove hardness ions, but it can also remove NOM and particles, yielding the possibility to
use the process as a pretreatment for ultraltration. The objectives of this research were to understand the nature of the
fouling mechanisms for ultraltration when used for waters that either require softening or have been softened, and to
use that understanding to determine promising options for the use of softening as a pretreatment before ultraltration.
To understand fouling mechanisms in the integrated system with softening and ultraltration, three different levels of
softening performance in terms of removal of inorganics and organic matter were selected. Experiments were performed
with both natural waters and synthetic waters with similar (but separable) inorganic, organic, and particulate
characteristics. The synthetic waters were used to distinguish among inorganic fouling by precipitates, organic fouling,
particulate fouling, and combined fouling by particles and organic matter. The results showed that organic matter
played a major role in fouling, either by itself or by adsorption onto particles, and that softening pretreatment
effectively reduced the foulants prior to ultraltration.
r 2004 Elsevier Ltd. All rights reserved.
Keywords: Precipitative softening; Ultraltration; Pretreatment; Fouling mechanisms; Organic fouling
1. Introduction
Membrane processes are widely considered for
application in drinking water treatment, but fouling
often impedes their use. Pretreatment to remove foulants
(mainly particles and organic matter) before the
membrane process has become an important aspect of
membrane operations. Studies on pretreatment schemes
for membrane processes have included coagulation
followed by sedimentation (Wiesner et al., 1989;
Lahaussine-Turcaud et al., 1990; Carroll et al., 2000),
coagulation followed by dissolved air otation (Brag-
hetta et al., 1997), and activated carbon adsorption
(Adham et al., 1991; Jacangelo et al., 1995). Softening is
traditionally designed to remove hardness ions in hard
waters but it can also remove particles and organic
matter (Randtke et al., 1994). Therefore, the integrated
water treatment with softening and ultraltration is
proposed as a promising option for hard waters.
The objectives of this research were to understand the
nature of the fouling mechanisms for ultraltration
when used for waters that either require softening or
ARTICLE IN PRESS
www.elsevier.com/locate/watres
0043-1354/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2004.06.013

Corresponding author. Tel: +1-512-471-4595; Fax: +1-

512-471-5870.
E-mail address: dlawler@mail.utexas.edu (D.F. Lawler).
have been softened, and to use that understanding to
determine promising options for the use of softening as a
pretreatment before ultraltration. In this study, three
levels of softening were selected and four possible
fouling mechanisms after softening pretreatment were
considered. The four fouling mechanisms included
inorganic fouling by continued precipitative softening,
organic fouling, particulate fouling, and combined
fouling by particles and organic matter. Two natural
waters and several synthetic waters were used to
thoroughly investigate fouling in ultraltration.
2. Materials and methods
2.1. Source waters
Source waters included two natural waters and
synthetic waters that were designed to separate the
effects of specic constituents on membrane perfor-
mance. The natural waters were Lake Austin water
(Austin, TX) and Missouri River water (St. Louis, MO).
The turbidity values were substantially different
(1.8 NTU for Lake Austin water and 314 NTU for
Missouri River water) but most other water quality
parameters, such as TOC and Mg
2+
: Ca
2+
ratio, were
rather similar to each other (Kweon, 2002). Due to the
dramatic difference in particle concentrations, Missouri
River water was useful to investigate effects of particle
concentration on fouling in ultraltration. The synthetic
waters included one with inorganic constituents only,
waters with simple organic components, a water with
kaolin, and a water with all of these constituents. Since
Lake Austin water was the primary water source,
inorganic constituents of all synthetic waters matched
Lake Austin water. The synthetic organic waters were
produced with organic compounds at the same dissolved
organic carbon (DOC) concentration as Lake Austin.
To overcome the complexities of interpreting the
effects of natural organic matter (NOM) on membrane
fouling, simple organic components were considered as
surrogates for NOM. Previous research on NOM
components in membrane processes indicated that
polysaccharides formed an important part of the
macromolecules present in the fouling cake (Lahous-
sine-Turcaud et al., 1990; Mallevialle et al., 1989). Two
polysaccharides, dextran and alginic acid, were selected
as the surrogates, based on the criterion that the DOC
removal had to be similar for the NOM surrogates
and the Lake Austin water. The removal of dextran
at 4 mg/L C by softening at various lime doses was
almost identical to the NOM removal previously
reported for the Lake Austin water (Kweon and Lawler,
2002). The removal of alginic acid was greater than that
of NOM (and dextran) with increasing lime doses, but
nevertheless showed a trend of increasing removal with
increasing dose like NOM in Lake Austin water. Two
other simple organics, polygalacturonic acid and tannic
acid, were found to be unsuitable surrogates for NOM
for this research (Kweon, 2002).
2.2. Softening pretreatment
Softening as a pretreatment before ultraltration was
conducted in a batch mode; a large amount of water
(14 L) was softened in a 20 L cylindrical plastic reservoir
at the specied softening condition. Softening was
accomplished with 2 min of rapid mixing after lime
addition, 30 min of slow mixing, and up to several hours
of settling. Experiments at high lime doses included
soda ash (Na
2
CO
3
) addition one minute prior to the
lime. Since turbidity is usually in the range of 25 NTU
after sedimentation in real plants, the settling was
stopped when the supernatant was in this range before
supernatant was siphoned off. The supernatant was used
as the feed water for the subsequent ultraltration
operation.
2.3. Ultraltration
The ultraltration system was a bench-scale plate and
frame apparatus (Minitan-S, Millipore Company, Bed-
ford, MA) that was operated in crossow. The system
had 30 cm
2
of lter area and ports for the feed, ltrate,
and concentrate. A relatively tight (10 kDa) hydropho-
bic (polysulfone) membrane was purchased from Milli-
pore Company for this research.
A new membrane sheet was used in each experiment
after rigorous cleaning and stabilizing procedures
(Kweon, 2002). The clean water specic ux was
measured at a constant pressure (approximately
90 kPa) for 2 h. The clean water specic ux was
reported after the ux was corrected for temperature
(normalized to 20 1C) using the equation by Braghetta et
al. [1997]. Ultraltration was operated as a total recycle
system, i.e., water from both concentrate and ltrate
ports was returned to the feed reservoir. The water
volume in the feed reservoir (14 L) was signicantly
larger than that in the ultraltration system (approxi-
mately 30 mL), so that materials deposited on the
membrane would cause little change in water quality
of the feed reservoir.
The normalized specic ux (i.e., percent of the clean
water specic ux) was used to compare ux decline
behavior among the various experiments. Normalization
was also performed for operational periods; normalized
cumulative production (cm
3
/cm
2
) was used instead of
elapsed times (i.e., minutes or hours) or cumulative
volume produced. Besides accounting for the exposed
area of the membrane, this normalization accounts
appropriately for variations in pore area and transmem-
brane pressure (TMP) between experiments. The calcu-
ARTICLE IN PRESS
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4165
lations were performed as follows:
Normalized cumulative production to
time t
P
t
i
ot
Q
i
Dt
i
A
d
J
o;std
J
o
;
where Q
i
and Dt
i
are ow rates and time intervals
measured throughout the experiment, A is the surface
area of the membrane, J
o
is the measured clean water
ux for the membrane in use at the TMP of 90 kPa, and
J
o,std
is a theoretical calculated value of clean water ux
for a standard membrane (assuming 5% porosity and
Poiseuille ow in the pores). The assumption was that
the ratio of the clean water uxes of two membranes at a
standardized TMP equaled the ratio of their pore areas
(Kweon and Lawler, 2002).
2.4. Cleaning of fouled membrances
After each experiment, membranes were cleaned with
three cleaning methods, and the clean water ux was
measured after each cleaning to evaluate the extent of
three fouling mechanisms. These cleaning methods
included a one-hour surface wash with distilled/deio-
nized water for surface deposits, a two-hour caustic
wash with a high pH solution (0.5 N NaOH) for organic
fouling, and a one-hour acidic wash with a low pH
solution (0.1 N HNO
3
) for inorganic fouling. After each
cleaning, the specic ux was measured using distilled/
deionized water for 20 min.
Liquid sample analyses were performed according to
Standard Methods (APHA, AWWA and WEF, 1998)
for pH, turbidity, Ca
2+
, Mg
2+
, DOC, and UV
254
(for
natural waters). In addition, several solutions of dextran
were analyzed for molecular weight distribution using
size exclusion chromatography. The detailed methods
are described elsewhere (Kweon and Lawler, 2002).
3. Results and discussion
3.1. Selection of degree of softening
The degree to which the water can and should be
softened is different for waters from different sources.
Experiments at various lime doses were conducted with
the two natural waters to determine the extent of
softening. Three softening conditions were chosen
corresponding to different removals of hardness ions
(Ca
2+
and Mg
2+
) and organic matter: standard
softening (maximum Ca
2+
removal with no addition
of soda ash), enhanced softening (maximum lime
addition without massive precipitation of Mg(OH)
2
),
and Mg softening (lime dose to achieve approximately
95% Mg
2+
removal). Each condition represents a
different degree of pretreatment to ultraltration in
terms of organic matter and inorganics. The lime doses
for these three conditions were found for the Lake Austin
water to be 125 mg/L as CaO for standard softening,
170 mg/L for enhanced softening, and 230 mg/L for the
Mg softening conditions. Only two lime doses (90 and
165 mg/L as CaO) were selected for Missouri River
water, because the lower dose corresponded to both the
standard and the enhanced softening conditions. The
detailed water quality at these lime doses for the two
natural waters was reported elsewhere (Kweon, 2002).
3.2. Inorganic fouling by synthetic waters
Softening is commonly used to remove Ca
2+
and
Mg
2+
from drinking water sources. The hardness ions
remaining in softened water, if they continue to form
precipitates on the membrane surface or in the pores,
could cause fouling in ultraltration. The slow kinetics
of softening (Alexander and McClanahan, 1975) sug-
gests this possibility. Using a synthetic water to simulate
the inorganic constituents of Lake Austin, the effects of
the degree of softening and precipitation kinetics were
investigated with three lime doses and two shorter
occulation times (1.5 and 7.5 min) than the standard
condition (30 min). The results showed little effect on
water ux decline in ultraltration regardless of the
extent of softening and precipitation times. Scanning
electron microscopy images from the membrane at the
230 mg/L CaO dose and 7.5 min occulation time
revealed substantial deposits on the membrane surface,
but these deposits had virtually no effect on water ux
(Kweon, 2002).
3.3. Effects of extent of softening on organic fouling
The effects of softening on organic fouling were
investigated with both dextran and alginic acid. The
properties of the two chosen polysaccharides are
summarized in Table 1. The different nominal molecular
weights, shapes, and charges of these polysaccharides
might lead to different fouling characteristics in ultra-
ltration. The raw solutions were softened at the three
chosen levels of softening based on the softening
performance in the Lake Austin water. Soda ash was
added in some experiments at the high lime doses (170
and 230 mg/L CaO) to ensure that CaCO
3
precipitation
was not limited by the availability of CO
3
2
ions.
The water quality in the experiments with the dextran
and alginic acid is presented in Table 2. The water
quality achieved under each experimental condition was
different because of the different degrees of softening
(and resulting different degrees of removal of the
organic matter). Turbidity and inorganic constituents
such as pH, Ca
2+
, and Mg
2+
were reasonably
consistent with results from Lake Austin water. The
only exception to this consistency was that the calcium
ARTICLE IN PRESS
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4166
concentration in the experiment with raw alginic acid
was slightly higher than that in Lake Austin water.
The DOC removal of dextran by ultraltration was in
the range of 3366% of the feed water. The high
removals of alginic acid (more than 90%) and turbidity
(almost 100% removal) were expected because the
nominal molecular weight of the dextran (60 kDa) and
the alginic acid (1280 kDa) are relatively high com-
pared to the nominal pore size of the membrane
(10 kDa). Also, some Ca
2+
and Mg
2+
was removed
by ultraltration in the experiments with alginic acid,
even after softening, but such removal did not occur
with dextran. Apparently, Ca
2+
and Mg
2+
became
associated with alginic acid by complexation or pre-
cipitation, and these compounds were removed from the
solution by ultraltration.
The ux decline in the experiments of dextran or
alginic acid at different lime doses is shown in Fig. 1.
The results showed that simple organic components used
as NOM surrogates were detrimental foulants when raw
waters (without softening) were applied to the mem-
brane. Softening pretreatment signicantly reduced
organic fouling. For each water, increasing the extent
of softening continually reduced the fouling, but the
extent of the improvement of water ux varied depend-
ing on the feed water.
In the case of the dextran solution (Fig. 1a) softened
at 125 mg/L of lime, the ux leveled off at approximately
73% of the clean water specic ux at a normalized
cumulative production of approximately 60 cm
3
/cm
2
.
Without softening, the ux was 59% of the clean water
specic ux after the same production. As the degree of
softening was increased to the enhanced softening
condition (170 mg/L of lime), virtually no ux decline
was observed. Note that the DOC concentration in feed
water was 4.1 mg/L at the lime dose of 125 mg/L CaO
and 1.8 mg/L at the lime dose of 170 mg/L CaO. The
results imply that the portion of dextran solution that is
ARTICLE IN PRESS
Table 1
Characteristic properties of dextran and alginic acid
Compounds Nominal molecular
weights (kDa)
Charge Comments Ref.
Alginic acid (Aldrich, St.
Louis, MO)
1280 Negative,
COO

Polydisperse, semi-exible,
negatively charged
polyelectrolytes extended
random coils
Bufe et al. (1998)
Dextran (Fluka, St.
Louis, MO)
60 Neutral Neutral; dense coil
Table 2
Water quality of experiments with dextran and alginic acid at different degrees of softening
Organic
matter
Lime
(mg/L CaO)
Sample PH Turbidity
(NTU)
Ca
2+
(mg/L) Mg
2+
(mg/L) DOC
(mg/L)
Dextran 0 Feed 8.26 NA 59.2 16.9 5.4
Filtrate 8.25 45.0 15.6 2.7
125 Feed 10.56 2.03 12.8 13.6 4.1
Filtrate 10.52 0.014 14.0 13.6 2.7
170
a
Feed 10.81 4.20 11.1 2.2 1.8
Filtrate 10.95 0.013 7.8 1.9 0.7
230
a
Feed 11.37 3.4 4.4 0.6 1.0
Filtrate 11.42 0.011 4.4 0.5 0.3
Alginic acid 0 Feed 8.41 NA 70.4 16.7 4.7
Filtrate 8.48 66.1 16.9 0.3
125 Feed 10.54 3.78 19.3 14.1 3.8
Filtrate 10.53 0.020 15.6 13.7 D.L.
170
a
Feed 10.88 4.25 12.6 3.9 2.9
Filtrate 10.80 0.016 9.4 2.3 2.3
230
a
Feed 11.36 3.57 5.5 2.0 1.5
Filtrate 11.37 0.015 2.7 0.3 0.6
a
Soda ash (Na
2
CO
3
) was added at doses equivalent to 45 and 105 mg/L as CaO at the enhanced softening and Mg softening
condition, respectively.
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4167
primarily responsible for fouling in membranes was
preferentially removed by softening.
The ux improvement of alginic acid (Fig. 1b) was
similar to that of dextran. The only difference occurred
at the highest lime dose (230 mg/L CaO); the ux
continuously improved with increasing lime dose in the
experiments with alginic acid but it leveled off in the
experiments with dextran. The DOC in the experi-
ments with alginic acid constantly decreased with the
increasing extent of softening, so that the improved
ux appears directly related to the decreased DOC
concentration.
To test quantitatively whether the ux decline is
directly caused by the overall DOC, the results are re-
plotted in Fig. 2 (a) and (b) based on the cumulative
mass of DOC fed to the membrane (per unit area),
rather than the cumulative volume of water. This
measure was obtained by multiplying the cumulative
water production by the feed DOC concentration in
each case. These graphs are helpful in elucidating the
underlying causes of the DOC removal, as explained
subsequently. The result from the experiments with
dextran is almost the same as the previous graph
normalized with the cumulative water production. In
contrast, the graphs after softening alginic acid nearly
collapse into one line, but that line is quite different
from the no softening results. The difference between the
no softening and the softening results could arise from
four causespH differences leading to different beha-
vior of the organics interacting with the membrane
surface, differences in the Ca
2+
and Mg
2+
concentra-
tions, more particles being present in the softening cases
(from the precipitation of inorganic solids), or a
preferential (rst) removal in softening of that portion
of alginic acid that causes the most fouling (or a
combination of all of these).
The pH effects after softening were studied with the
Lake Austin water and reported elsewhere (Kweon,
2002; Kweon and Lawler, 2002). Adjusting the pH after
softening to pH 8.5 (the same as the raw water) had no
effect at all on the ux decline at the lower two lime
doses and caused a very slight worsening of the ux at
the highest lime dose. Hence, the pH does not appear to
be a signicant factor in explaining the results of Figs. 1
and 2. The Ca
2+
concentration is reduced by all three
softening conditions, and the Mg
2+
concentration is
reduced by the higher two lime doses; since the results
for these inorganic species were essentially the same in
the experiments for the two different organic materials,
they cannot explain the differences between them. On
the other hand, it is possible that the reduced inorganic
concentrations between the zero lime dose and the
ARTICLE IN PRESS
0
20
40
60
80
100
120
0 0.05 0.1 0.15 0.2 0.25 0.3
0 0.05 0.1 0.15 0.2 0.25 0.3
0 mg/L CaO
125 mg/L CaO
170 mg/L CaO
230 mg/L CaO
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
Normalized Cumulative DOC Mass (mg/cm
2
)
Normalized Cumulative DOC Mass (mg/cm
2
)
0
20
40
60
80
100
120
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
(a)
(b)
Fig. 2. Effects of softening at different lime doses: ux decline
normalized with cumulative DOC mass; (a) dextran and (b)
alginic acid.
0
20
40
60
80
100
120
0 mg/L CaO
125 mg/ L CaO
170 mg/ L CaO
230 mg/ L CaO
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
Normalized Cumulative Production (cm
3
/cm
2
)
0
20
40
60
80
100
120
100 80 60 40 20 0
100 80 60 40 20 0
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
Normalized Cumulative Production (cm
3
/cm
2
)
(a)
(b)
Fig. 1. Effects of softening at different lime doses: ux decline
normalized with cumulative water production; (a) dextran and
(b) alginic acid.
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4168
others adds to the improvement of the ux, perhaps by
the reduction in ionic strength and the consequent
effects on the double layer thickness near the membrane
surface. The particle concentrations (turbidity) were
nearly the same under the various softening conditions,
so turbidity also cannot explain the effects of softening
on the results. Hence, the ux decline results are
explained best by the idea that the portions of the
organics that are most responsible for ux decline are
removed rst in softening.
3.4. Effects of softening on molecular weight distribution
of dextran
Previous research indicated that organic matter with
high molecular weight was preferentially removed in
softening and coagulation (Liao and Randtke, 1986).
Therefore, the molecular weight distributions of the raw
and softened dextran solutions were analyzed with size
exclusion chromatography, as presented in Fig. 3. The
ordinate is a relative measure of the concentration
within each sample. Each sample originally had a
different peak height because the concentrations varied
after softening, and the differences were exaggerated
after concentration by rotary evaporation. To determine
whether softening selectively removed high molecular
weights of dextran, the maximum value of the ordinate
was tted to a common scale in this gure.
Fig. 3(a) shows the molecular weight distribution of
the raw dextran, and the modal value (55.5 kDa) is close
to the average MW reported by manufacturer (60 kDa).
As the extent of softening increased, the molecular
weight distribution clearly shifted to smaller values, as
shown in Fig. 3 (b)(d). The results clearly illustrate that
softening preferentially removes the higher molecular
weight fraction of dextran.
3.5. Efciency of cleaning methods on organic fouling
In addition to the improvement of hydraulic water
ux, another possible advantage of pretreatment for a
membrane process is to condition fouling cakes to make
them more easily recovered by cleaning procedures. The
results of the three cleaning methods for the membranes
fouled with dextran or alginic acid are shown in Fig. 4.
After each cleaning, the ux was measured at three times
(5, 10, and 20 min), and, at each time, three separate
measurements were performed. The means (bars) and
standard deviations (whiskers) of the nine measurements
are shown in the gure.
ARTICLE IN PRESS
0 mg/L CaO
MW = 55496
MW = 36915
125 mg/L CaO
170 mg/L CaO
MW = 29309
10
4
10
5
10
6
Molecular Weight (kDa)
230 mg/L CaO
MW = 10644
(a)
(b)
(c)
(d)
Fig. 3. Molecular weight distribution of dextran after softening
by size exclusion chromatography.
Fig. 4. Membrane cleaning: effects of softening at different
lime doses; (a) dextran, (b) alginic acid.
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4169
For the experiments with dextran, only the experi-
ments with lime doses at 125 and 170 mg/L were
evaluated along with the raw dextran, because the ux
decline at 230 mg/L of lime was almost the same as that
at 170 mg/L of lime. As shown in Fig. 4(a), surface wash
yielded little or no improvement in the ux, suggesting
either that fouling occurred inside the pores or the
fouling cake was attached chemically to the membrane
surface. After the caustic wash, 65% of the lost ux was
recovered for the raw dextran, while 46% of the lost ux
was recovered for the solution treated at 125 mg/L of
lime. The caustic wash, which removes organic matter in
the fouling layer, was quite effective in cleaning the
membranes fouled with the dextran solutions.
For the experiments with alginic acid (Fig. 4b),
surface wash was quite effective in recovering the lost
ux in all experiments; the best result was a recovery of
approximately 70% of the lost ux for the membrane
fouled with the softened water at 230 mg/L of lime. In
each case, the caustic wash and the subsequent acid
wash improved the recovery a few more percent.
Considering that the caustic wash is intended to remove
organic foulants, the effective cleaning by the acidic
wash was relatively signicant; this result is consistent
with the idea that inorganics complex with alginic acid
and therefore play a role in membrane fouling.
3.6. Particle fouling by synthetic inorganic waters with
clay
Particulate matter, which can be substantially re-
moved in the softening process, can be a major foulant
in membrane processes. Under any operational condi-
tion in ultraltration, a particular size range of particles
will have the minimum back transport velocity and is
therefore most susceptible to deposit on the membrane
surface. The effects of particles alone on membrane
fouling were investigated using two clay suspensions (at
approximately 50 and 500 NTU) in the synthetic
inorganic water with no organics present. Ultraltration
showed virtually no ux decline even when a highly
turbid water (500 NTU) was applied; only 5% of the
ux was lost after a cumulative production of 150 cm
3
/
cm
2
. Particulate fouling was inconsequential under the
operational conditions in this research.
3.7. Combined fouling by dextran with kaolin
Little research has investigated the combined effects
of particle and NOM fouling. NOM fouling is con-
sidered the most signicant foulant in membrane
processes, but particle fouling has been neglected based
on the results of fouling by particles only. As reported in
the previous section, the clay material exhibited little
ux decline even with the very high turbidity. However,
NOM is likely to be adsorbed on particles in natural
water. In the presence of NOM, the characteristics of
particles such as their stability (stickiness) and surface
charge will be affected; thus, the fouling by particles
with adsorbed NOM is expected to be different from
that with inorganic particles with no adsorbed NOM.
Therefore, a synthetic water with kaolin particles and
one organic component (dextran) was used to evaluate
effects of both organic matter and particles on
membrane fouling.
The experiments were done at lime doses of zero and
170 mg/L CaO, and the results for both experiments
with kaolin were indistinguishable from those shown in
Fig. 1a without kaolin. The similarity presumably stems
from the fact that the polysaccharide was not adsorbed
to any signicant extent on the clay (as indicated by
DOC measurements), so that the polysaccharide be-
haves the same with respect to the membrane with or
without clay material being present. Therefore, effects of
fouling by natural particles in association with natural
organic matter were investigated with experiments with
Missouri River water.
3.8. Combined fouling by a natural water
The combined fouling by natural particles was
examined with Missouri River water since it has
dramatic differences in particle concentrations from
untreated Lake Austin water. As shown in Fig. 5, the
much greater ux decline with Missouri River water
than Lake Austin water indicates the importance of
particle fouling. While the synthetic particles alone
showed a negligible effect, the natural particles showed a
dramatic effect on membrane fouling. The interpretation
implies that other organic components, which can be
adsorbed on the clay material, would act differently
from the polysaccharide.
In addition, effects of the extent of softening on
fouling by Missouri River water was investigated using
two lime doses (i.e., 90 and 165 mg/L CaO). Since the
Missouri River water had a high concentration of
ARTICLE IN PRESS
0
20
40
60
80
100
120
0 20 40 60 80 100 120
Lake Austin water
Missouri River water
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
Normalized Cumulative Production (cm
3
/cm
2
)
Fig. 5. Flux decline: two natural water sources.
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4170
particulate matter, lime softening alone showed ineffec-
tive removal of turbidity. To improve the turbidity
removal, iron at a dose of 3 mg/L Fe
3+
was introduced
to the Missouri River water during the rapid mixing
period.
The water quality for the experiments with the
Missouri River water is summarized in Table 3. The
raw water turbidity was in the range of 297405 NTU
and was signicantly reduced to 24 NTU by softening
with iron addition. As expected, the Ca
2+
and Mg
2+
removal also increased as lime doses were increased (and
soda ash was added at the highest dose). The DOC and
SUVA decreased as lime doses increased. The DOC
removal by ultraltration varied slightly between 13%
and 18%.
The results of ux decline from the experiments with
the raw and the softened Missouri River water are
shown in Fig. 6. Softening shows dramatic improvement
of the ux decline of the raw Missouri River water. At
the lime dose of 90 mg/L CaO, 75% of the clean water
specic ux was still maintained after 90 cm
3
/cm
2
of the
normalized cumulative production. The higher lime dose
(the Mg softening condition) yielded only a marginal
further improvement of the ux, despite the greater
precipitation of CaCO
3
and Mg(OH)
2
and the increased
DOC removal.
The substantial ux improvement between the raw
and softened waters was surprising, given the trivial
differences in DOC concentrations. Note that the DOC
concentration was 3.2 mg/L C in the raw water (total
organic carbon concentration was reported as 3.8 mg/L
C in Table 3) and the DOC concentration was 3.3 mg/L
C in the feed water for the experiments at the lime dose
of 90 mg/L CaO. The results suggest that the DOC that
is preferentially removed in softening has a greater role
in fouling of ultraltration than the amount of DOC
remaining after the softening process. In addition, the
removal of particles from the raw water by softening
might play a role in improving the ux. The particles in
the raw Missouri River might have crucial sizes to foul
the membrane, or could have properties that cause them
to accumulate and build a fouling layer with signicant
resistance. The adsorption of NOM on particle surfaces
in the natural water could cause the characteristics of the
particles to preferentially foul the membrane. Note that
this behavior in natural water is different than that
observed in the synthetic water; water with both kaolin
and dextran yielded the same ux decline as the water
with dextran only, apparently because kaolin showed
little adsorption capability of dextran. In addition, the
ux decline from the experiment with the raw Missouri
River water was also greater than in the experiment with
synthetic water made with dextran and kaolin. The
turbidity of synthetic water with dextran and kaolin was
actually higher (560 NTU) than the Missouri River
water (297 NTU), although both waters had similar
DOC.
The improved ux at the standard and enhanced
softening conditions also implies that softening changes
the particle and NOM properties. Particles are con-
verted to bigger ocs by coagulation, particle composi-
tion is changed from natural particles to the CaCO
3
and
Mg(OH)
2
, and NOM becomes associated with the fresh
precipitates.
ARTICLE IN PRESS
Table 3
Water quality of experiments with the Missouri River water
Lime Soda ash
(mg/L CaO)
Samples pH Turbidity
(NTU)
Ca
2+
(mg/L)
Mg
2+
(mg/L)
DOC
(mg/L)
SUVA
(L/m-mg)
0 0 Feed 8.31 297 53.3 16.7 3.2
#
3.01
ltrate 8.29 0.019 46.9 14.7 2.6 3.44
90* 0 Raw 8.05 325 46.8 14.2 3.7 2.47
feed 9.47 2.43 23.6 10.2 3.3 NA
ltrate 9.53 0.015 23.3 10.0 2.8 2.02
165* 75 Raw 7.95 405 60.8 17.4 3.1 3.61
feed 10.74 3.03 10.4 0.9 2.2 2.56
ltrate 10.78 0.026 7.8 0.7 1.8 2.07
*Addition of 3 mg/L Fe
3+
during softening,
#
TOC: 3.8 mg/L C.
0
20
40
60
80
100
120
0 20 40 60 80 100 120
0 mg/L CaO
90 mg/L CaO
165 mg/L CaO
%

o
f

C
l
e
a
n

W
a
t
e
r

S
p
e
c
i
f
i
c

F
l
u
x
Normalized Cumulative Production (cm
3
/cm
2
)
Fig. 6. Flux decline: effects of softening Missouri River water
at different lime doses.
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4171
4. Conclusions
Inorganic fouling by precipitates shows negligible
effect on membrane fouling.
Ultraltration is not susceptible to signicant parti-
culate fouling under the operational conditions in this
research. Ultraltration showed virtually no ux
decline when the synthetic inorganic water with only
a clay material (without NOM) was treated.
Organic matter is the most detrimental foulant of
ultralters. Softening pretreatment signicantly re-
duced organic fouling in ultraltration, but the extent
of improvement of water ux depends on the extent
of softening and the specic feed water.
The cause of ux improvement in the experiments
with the polysaccharide (dextran or alginic acid)
stemmed from the preferential removal of its high
molecular weight fraction in softening.
The combined fouling by synthetic particles (kaolin)
and dextran showed little difference from the foul-
ing with only the organic matter. However, the
natural particles showed a dramatic effect on
membrane fouling. The interpretation implies that
other organic components, which can be adsorbed
on the clay material, would act differently from the
polysaccharide.
Acknowledgements
The authors would like to acknowledge the American
Water Works Association Research Foundation (AW-
WARF) for nancial support of this work. The views
expressed herein are those of the authors alone and do
not necessarily reect the views of AWWARF. We also
acknowledge the cooperation and assistance of the City
of Austin Water and Wastewater Utility and the
Missouri American Water Company.
References
Adham, S.S., Snoeyink, V.L., Clark, M.M., Bersillon, J., 1991.
Predicting and verifying organics removal by PAC in an
ultraltration system. Am. Water Works Assoc. J. 83 (12),
8191.
APHA, AWWA, WEF (American Public Health Association,
American Water Works Association, and Water Environ-
ment Federation) 1998. Standard Methods for the Exam-
ination of Water and Wastewater, twentieth ed. APHA,
Washington, DC.
Alexander, H.J., McClanahan, M.A., 1975. Kinetics of calcium
carbonate precipitation in lime-soda ash softening. Am.
Water Works Assoc. J. 67 (11), 618621.
Bufe, J., Wilkinson, K.J., Stoll, S., Filella, M., Zhang, J., 1998.
A generalized description of aquatic colloidal interactions:
the three-colloidal component approach. Environ. Sci.
Technol. 32 (19), 28872899.
Braghetta, A., Jacangelo, J.G., Chellem, S., Hotaling, M.L.,
Utne, B.A., 1997. DAF pretreatment: its effects on MF
performance. Am. Water Works Assoc. J. 89 (10), 90101.
Carroll, T., King, S., Gray, S.R., Bolto, B.A., Booker, N.A.,
2000. The fouling of microltration membranes by NOM
after coagulation treatment. Water Res. 34 (11), 28612868.
Jacangelo, J.G., Laine, J., Cummings, E.W., Adham, S.S.,
1995. UF with pretreatment for removing DBP precursors.
Am. Water Works Assoc. J. 87 (3), 100112.
Kweon, J.H., 2002. Integrated water treatment softening
and ultraltration. Dissertation for the degree of Doctor
of Philosophy. Department of Civil Engineering, The
University of Texas at Austin, TX, USA.
Kweon, J.H., Lawler, D.F., 2002. Evaluating precipitative
softening pretreatment for ultraltration of a natural water.
Environ. Eng. Sci. 19 (6), 531544.
Lahoussine-Turcaud, V., Wiesner, M.R., Bottero, J., Mal-
levialle, J., 1990. Coagulation in pretreatment for ultral-
tration of a surface water. Am. Water Works Assoc. J. 82
(12), 7681.
Liao, M.Y., Randtke, S.J., 1986. Predicting the removal of
soluble organic contaminants by lime softening. Water Res.
20 (1), 2735.
Mallevialle, J., Anselme, C., Marsigny, O., 1989. Effects of
humic substances on membrane processes. Aquatic humic
substances. In: Proceedings of the 193rd Meeting of the
American Chemical Society, Denver, CO pp. 749767.
Randtke, S.J., Hoehn, R.C., Knocke, W.R., Dietrich, A.M.,
Long, B.W., Wang, N., 1994. A comprehensive assessment
of DBP precursor removal by enhanced coagulation/soft-
ening. In: Proceedings of the Annual AWWA Conference.
New York, NY, AWWA.
Wiesner, M.R., Clark, M.M., Mallevialle, J., 1989. Membrane
ltration of coagulated suspensions. J. Environ. Eng. 115
(1), 2040.
ARTICLE IN PRESS
J.H. Kweon, D.F. Lawler / Water Research 38 (2004) 41644172 4172